diff --git a/.gitignore b/.gitignore index 32b4c424..e6e8bb00 100644 --- a/.gitignore +++ b/.gitignore @@ -11,6 +11,7 @@ /lib/ /Makefile /include/jemalloc/internal/jemalloc_internal\.h +/include/jemalloc/internal/size_classes\.h /include/jemalloc/jemalloc\.h /include/jemalloc/jemalloc_defs\.h /test/jemalloc_test\.h @@ -21,3 +22,4 @@ !test/*.c !test/*.exp /VERSION +/bin/jemalloc.sh diff --git a/COPYING b/COPYING index 10ade120..e27fc4d6 100644 --- a/COPYING +++ b/COPYING @@ -1,9 +1,10 @@ Unless otherwise specified, files in the jemalloc source distribution are -subject to the following licenses: +subject to the following license: -------------------------------------------------------------------------------- -Copyright (C) 2002-2010 Jason Evans . +Copyright (C) 2002-2012 Jason Evans . All rights reserved. -Copyright (C) 2007-2010 Mozilla Foundation. All rights reserved. +Copyright (C) 2007-2012 Mozilla Foundation. All rights reserved. +Copyright (C) 2009-2012 Facebook, Inc. All rights reserved. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: @@ -24,28 +25,3 @@ LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. -------------------------------------------------------------------------------- -Copyright (C) 2009-2010 Facebook, Inc. -All rights reserved. - -Redistribution and use in source and binary forms, with or without modification, -are permitted provided that the following conditions are met: -* Redistributions of source code must retain the above copyright notice, this - list of conditions and the following disclaimer. -* Redistributions in binary form must reproduce the above copyright notice, this - list of conditions and the following disclaimer in the documentation and/or - other materials provided with the distribution. -* Neither the name of Facebook, Inc. nor the names of its contributors may be - used to endorse or promote products derived from this software without - specific prior written permission. - -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND -ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED -WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE -DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR -ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES -(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; -LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON -ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT -(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS -SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. --------------------------------------------------------------------------------- diff --git a/ChangeLog b/ChangeLog index 326ee7a9..231dd6da 100644 --- a/ChangeLog +++ b/ChangeLog @@ -6,6 +6,95 @@ found in the git revision history: http://www.canonware.com/cgi-bin/gitweb.cgi?p=jemalloc.git git://canonware.com/jemalloc.git +* 3.0.0 (May 11, 2012) + + Although this version adds some major new features, the primary focus is on + internal code cleanup that facilitates maintainability and portability, most + of which is not reflected in the ChangeLog. This is the first release to + incorporate substantial contributions from numerous other developers, and the + result is a more broadly useful allocator (see the git revision history for + contribution details). Note that the license has been unified, thanks to + Facebook granting a license under the same terms as the other copyright + holders (see COPYING). + + New features: + - Implement Valgrind support, redzones, and quarantine. + - Add support for additional platforms: + + FreeBSD + + Mac OS X Lion + + MinGW + + Windows (no support yet for replacing the system malloc) + - Add support for additional architectures: + + MIPS + + SH4 + + Tilera + - Add support for cross compiling. + - Add nallocm(), which rounds a request size up to the nearest size class + without actually allocating. + - Implement aligned_alloc() (blame C11). + - Add the "thread.tcache.enabled" mallctl. + - Add the "opt.prof_final" mallctl. + - Update pprof (from gperftools 2.0). + - Add the --with-mangling option. + - Add the --disable-experimental option. + - Add the --disable-munmap option, and make it the default on Linux. + - Add the --enable-mremap option, which disables use of mremap(2) by default. + + Incompatible changes: + - Enable stats by default. + - Enable fill by default. + - Disable lazy locking by default. + - Rename the "tcache.flush" mallctl to "thread.tcache.flush". + - Rename the "arenas.pagesize" mallctl to "arenas.page". + - Change the "opt.lg_prof_sample" default from 0 to 19 (1 B to 512 KiB). + - Change the "opt.prof_accum" default from true to false. + + Removed features: + - Remove the swap feature, including the "config.swap", "swap.avail", + "swap.prezeroed", "swap.nfds", and "swap.fds" mallctls. + - Remove highruns statistics, including the + "stats.arenas..bins..highruns" and + "stats.arenas..lruns..highruns" mallctls. + - As part of small size class refactoring, remove the "opt.lg_[qc]space_max", + "arenas.cacheline", "arenas.subpage", "arenas.[tqcs]space_{min,max}", and + "arenas.[tqcs]bins" mallctls. + - Remove the "arenas.chunksize" mallctl. + - Remove the "opt.lg_prof_tcmax" option. + - Remove the "opt.lg_prof_bt_max" option. + - Remove the "opt.lg_tcache_gc_sweep" option. + - Remove the --disable-tiny option, including the "config.tiny" mallctl. + - Remove the --enable-dynamic-page-shift configure option. + - Remove the --enable-sysv configure option. + + Bug fixes: + - Fix a statistics-related bug in the "thread.arena" mallctl that could cause + invalid statistics and crashes. + - Work around TLS deallocation via free() on Linux. This bug could cause + write-after-free memory corruption. + - Fix a potential deadlock that could occur during interval- and + growth-triggered heap profile dumps. + - Fix large calloc() zeroing bugs due to dropping chunk map unzeroed flags. + - Fix chunk_alloc_dss() to stop claiming memory is zeroed. This bug could + cause memory corruption and crashes with --enable-dss specified. + - Fix fork-related bugs that could cause deadlock in children between fork + and exec. + - Fix malloc_stats_print() to honor 'b' and 'l' in the opts parameter. + - Fix realloc(p, 0) to act like free(p). + - Do not enforce minimum alignment in memalign(). + - Check for NULL pointer in malloc_usable_size(). + - Fix an off-by-one heap profile statistics bug that could be observed in + interval- and growth-triggered heap profiles. + - Fix the "epoch" mallctl to update cached stats even if the passed in epoch + is 0. + - Fix bin->runcur management to fix a layout policy bug. This bug did not + affect correctness. + - Fix a bug in choose_arena_hard() that potentially caused more arenas to be + initialized than necessary. + - Add missing "opt.lg_tcache_max" mallctl implementation. + - Use glibc allocator hooks to make mixed allocator usage less likely. + - Fix build issues for --disable-tcache. + - Don't mangle pthread_create() when --with-private-namespace is specified. + * 2.2.5 (November 14, 2011) Bug fixes: diff --git a/INSTALL b/INSTALL index 2a1e469c..e40a7edd 100644 --- a/INSTALL +++ b/INSTALL @@ -26,6 +26,19 @@ any of the following arguments (not a definitive list) to 'configure': Embed one or more library paths, so that libjemalloc can find the libraries it is linked to. This works only on ELF-based systems. +--with-mangling= + Mangle public symbols specified in which is a comma-separated list of + name:mangled pairs. + + For example, to use ld's --wrap option as an alternative method for + overriding libc's malloc implementation, specify something like: + + --with-mangling=malloc:__wrap_malloc,free:__wrap_free[...] + + Note that mangling happens prior to application of the prefix specified by + --with-jemalloc-prefix, and mangled symbols are then ignored when applying + the prefix. + --with-jemalloc-prefix= Prefix all public APIs with . For example, if is "prefix_", API changes like the following occur: @@ -62,8 +75,8 @@ any of the following arguments (not a definitive list) to 'configure': Enable assertions and validation code. This incurs a substantial performance hit, but is very useful during application development. ---enable-stats - Enable statistics gathering functionality. See the "opt.stats_print" +--disable-stats + Disable statistics gathering functionality. See the "opt.stats_print" option documentation for usage details. --enable-prof @@ -90,51 +103,50 @@ any of the following arguments (not a definitive list) to 'configure': Statically link against the specified libunwind.a rather than dynamically linking with -lunwind. ---disable-tiny - Disable tiny (sub-quantum-sized) object support. Technically it is not - legal for a malloc implementation to allocate objects with less than - quantum alignment (8 or 16 bytes, depending on architecture), but in - practice it never causes any problems if, for example, 4-byte allocations - are 4-byte-aligned. - --disable-tcache Disable thread-specific caches for small objects. Objects are cached and released in bulk, thus reducing the total number of mutex operations. See the "opt.tcache" option for usage details. ---enable-swap - Enable mmap()ed swap file support. When this feature is built in, it is - possible to specify one or more files that act as backing store. This - effectively allows for per application swap files. +--enable-mremap + Enable huge realloc() via mremap(2). mremap() is disabled by default + because the flavor used is specific to Linux, which has a quirk in its + virtual memory allocation algorithm that causes semi-permanent VM map holes + under normal jemalloc operation. + +--disable-munmap + Disable virtual memory deallocation via munmap(2); instead keep track of + the virtual memory for later use. munmap() is disabled by default (i.e. + --disable-munmap is implied) on Linux, which has a quirk in its virtual + memory allocation algorithm that causes semi-permanent VM map holes under + normal jemalloc operation. --enable-dss Enable support for page allocation/deallocation via sbrk(2), in addition to mmap(2). ---enable-fill - Enable support for junk/zero filling of memory. See the "opt.junk"/ - "opt.zero" option documentation for usage details. +--disable-fill + Disable support for junk/zero filling of memory, quarantine, and redzones. + See the "opt.junk", "opt.zero", "opt.quarantine", and "opt.redzone" option + documentation for usage details. + +--disable-valgrind + Disable support for Valgrind. + +--disable-experimental + Disable support for the experimental API (*allocm()). + +--enable-utrace + Enable utrace(2)-based allocation tracing. This feature is not broadly + portable (FreeBSD has it, but Linux and OS X do not). --enable-xmalloc Enable support for optional immediate termination due to out-of-memory errors, as is commonly implemented by "xmalloc" wrapper function for malloc. See the "opt.xmalloc" option documentation for usage details. ---enable-sysv - Enable support for System V semantics, wherein malloc(0) returns NULL - rather than a minimal allocation. See the "opt.sysv" option documentation - for usage details. - ---enable-dynamic-page-shift - Under most conditions, the system page size never changes (usually 4KiB or - 8KiB, depending on architecture and configuration), and unless this option - is enabled, jemalloc assumes that page size can safely be determined during - configuration and hard-coded. Enabling dynamic page size determination has - a measurable impact on performance, since the compiler is forced to load - the page size from memory rather than embedding immediate values. - ---disable-lazy-lock - Disable code that wraps pthread_create() to detect when an application +--enable-lazy-lock + Enable code that wraps pthread_create() to detect when an application switches from single-threaded to multi-threaded mode, so that it can avoid mutex locking/unlocking operations while in single-threaded mode. In practice, this feature usually has little impact on performance unless @@ -181,11 +193,24 @@ PATH="?" === Advanced compilation ======================================================= +To build only parts of jemalloc, use the following targets: + + build_lib_shared + build_lib_static + build_lib + build_doc_html + build_doc_man + build_doc + To install only parts of jemalloc, use the following targets: install_bin install_include + install_lib_shared + install_lib_static install_lib + install_doc_html + install_doc_man install_doc To clean up build results to varying degrees, use the following make targets: @@ -248,10 +273,6 @@ directory, issue configuration and build commands: The manual page is generated in both html and roff formats. Any web browser can be used to view the html manual. The roff manual page can be formatted -prior to installation via any of the following commands: +prior to installation via the following command: nroff -man -t doc/jemalloc.3 - - groff -man -t -Tps doc/jemalloc.3 | ps2pdf - doc/jemalloc.3.pdf - - (cd doc; groff -man -man-ext -t -Thtml jemalloc.3 > jemalloc.3.html) diff --git a/Makefile.in b/Makefile.in index de7492f9..6675b596 100644 --- a/Makefile.in +++ b/Makefile.in @@ -17,130 +17,185 @@ INCLUDEDIR := $(DESTDIR)@INCLUDEDIR@ LIBDIR := $(DESTDIR)@LIBDIR@ DATADIR := $(DESTDIR)@DATADIR@ MANDIR := $(DESTDIR)@MANDIR@ +srcroot := @srcroot@ +objroot := @objroot@ +abs_srcroot := @abs_srcroot@ +abs_objroot := @abs_objroot@ # Build parameters. -CPPFLAGS := @CPPFLAGS@ -I@srcroot@include -I@objroot@include +CPPFLAGS := @CPPFLAGS@ -I$(srcroot)include -I$(objroot)include CFLAGS := @CFLAGS@ -ifeq (macho, @abi@) -CFLAGS += -dynamic -endif LDFLAGS := @LDFLAGS@ +EXTRA_LDFLAGS := @EXTRA_LDFLAGS@ LIBS := @LIBS@ RPATH_EXTRA := @RPATH_EXTRA@ -ifeq (macho, @abi@) -SO := dylib -WL_SONAME := dylib_install_name +SO := @so@ +IMPORTLIB := @importlib@ +O := @o@ +A := @a@ +EXE := @exe@ +LIBPREFIX := @libprefix@ +REV := @rev@ +install_suffix := @install_suffix@ +ABI := @abi@ +XSLTPROC := @XSLTPROC@ +AUTOCONF := @AUTOCONF@ +_RPATH = @RPATH@ +RPATH = $(if $(1),$(call _RPATH,$(1))) +cfghdrs_in := @cfghdrs_in@ +cfghdrs_out := @cfghdrs_out@ +cfgoutputs_in := @cfgoutputs_in@ +cfgoutputs_out := @cfgoutputs_out@ +enable_autogen := @enable_autogen@ +enable_experimental := @enable_experimental@ +DSO_LDFLAGS = @DSO_LDFLAGS@ +SOREV = @SOREV@ +PIC_CFLAGS = @PIC_CFLAGS@ +CTARGET = @CTARGET@ +LDTARGET = @LDTARGET@ +MKLIB = @MKLIB@ +CC_MM = @CC_MM@ + +ifeq (macho, $(ABI)) +TEST_LIBRARY_PATH := DYLD_FALLBACK_LIBRARY_PATH="$(objroot)lib" else -SO := so -WL_SONAME := soname -endif -REV := 1 -ifeq (macho, @abi@) -TEST_LIBRARY_PATH := DYLD_FALLBACK_LIBRARY_PATH=@objroot@lib +ifeq (pecoff, $(ABI)) +TEST_LIBRARY_PATH := PATH="$(PATH):$(objroot)lib" else TEST_LIBRARY_PATH := endif +endif + +LIBJEMALLOC := $(LIBPREFIX)jemalloc$(install_suffix) # Lists of files. -BINS := @srcroot@bin/pprof -CHDRS := @objroot@include/jemalloc/jemalloc@install_suffix@.h \ - @objroot@include/jemalloc/jemalloc_defs@install_suffix@.h -CSRCS := @srcroot@src/jemalloc.c @srcroot@src/arena.c @srcroot@src/atomic.c \ - @srcroot@src/base.c @srcroot@src/bitmap.c @srcroot@src/chunk.c \ - @srcroot@src/chunk_dss.c @srcroot@src/chunk_mmap.c \ - @srcroot@src/chunk_swap.c @srcroot@src/ckh.c @srcroot@src/ctl.c \ - @srcroot@src/extent.c @srcroot@src/hash.c @srcroot@src/huge.c \ - @srcroot@src/mb.c @srcroot@src/mutex.c @srcroot@src/prof.c \ - @srcroot@src/rtree.c @srcroot@src/stats.c @srcroot@src/tcache.c -ifeq (macho, @abi@) -CSRCS += @srcroot@src/zone.c +BINS := $(srcroot)bin/pprof $(objroot)bin/jemalloc.sh +CHDRS := $(objroot)include/jemalloc/jemalloc$(install_suffix).h \ + $(objroot)include/jemalloc/jemalloc_defs$(install_suffix).h +CSRCS := $(srcroot)src/jemalloc.c $(srcroot)src/arena.c $(srcroot)src/atomic.c \ + $(srcroot)src/base.c $(srcroot)src/bitmap.c $(srcroot)src/chunk.c \ + $(srcroot)src/chunk_dss.c $(srcroot)src/chunk_mmap.c \ + $(srcroot)src/ckh.c $(srcroot)src/ctl.c $(srcroot)src/extent.c \ + $(srcroot)src/hash.c $(srcroot)src/huge.c $(srcroot)src/mb.c \ + $(srcroot)src/mutex.c $(srcroot)src/prof.c $(srcroot)src/quarantine.c \ + $(srcroot)src/rtree.c $(srcroot)src/stats.c $(srcroot)src/tcache.c \ + $(srcroot)src/util.c $(srcroot)src/tsd.c +ifeq (macho, $(ABI)) +CSRCS += $(srcroot)src/zone.c endif -STATIC_LIBS := @objroot@lib/libjemalloc@install_suffix@.a -DSOS := @objroot@lib/libjemalloc@install_suffix@.$(SO).$(REV) \ - @objroot@lib/libjemalloc@install_suffix@.$(SO) \ - @objroot@lib/libjemalloc@install_suffix@_pic.a -MAN3 := @objroot@doc/jemalloc@install_suffix@.3 -DOCS_XML := @objroot@doc/jemalloc@install_suffix@.xml -DOCS_HTML := $(DOCS_XML:@objroot@%.xml=@srcroot@%.html) -DOCS_MAN3 := $(DOCS_XML:@objroot@%.xml=@srcroot@%.3) +ifeq ($(IMPORTLIB),$(SO)) +STATIC_LIBS := $(objroot)lib/$(LIBJEMALLOC).$(A) +endif +ifdef PIC_CFLAGS +STATIC_LIBS += $(objroot)lib/$(LIBJEMALLOC)_pic.$(A) +else +STATIC_LIBS += $(objroot)lib/$(LIBJEMALLOC)_s.$(A) +endif +DSOS := $(objroot)lib/$(LIBJEMALLOC).$(SOREV) +ifneq ($(SOREV),$(SO)) +DSOS += $(objroot)lib/$(LIBJEMALLOC).$(SO) +endif +MAN3 := $(objroot)doc/jemalloc$(install_suffix).3 +DOCS_XML := $(objroot)doc/jemalloc$(install_suffix).xml +DOCS_HTML := $(DOCS_XML:$(objroot)%.xml=$(srcroot)%.html) +DOCS_MAN3 := $(DOCS_XML:$(objroot)%.xml=$(srcroot)%.3) DOCS := $(DOCS_HTML) $(DOCS_MAN3) -CTESTS := @srcroot@test/allocated.c @srcroot@test/allocm.c \ - @srcroot@test/bitmap.c @srcroot@test/mremap.c \ - @srcroot@test/posix_memalign.c @srcroot@test/rallocm.c \ - @srcroot@test/thread_arena.c +CTESTS := $(srcroot)test/aligned_alloc.c $(srcroot)test/allocated.c \ + $(srcroot)test/bitmap.c $(srcroot)test/mremap.c \ + $(srcroot)test/posix_memalign.c $(srcroot)test/thread_arena.c \ + $(srcroot)test/thread_tcache_enabled.c +ifeq ($(enable_experimental), 1) +CTESTS += $(srcroot)test/allocm.c $(srcroot)test/rallocm.c +endif + +COBJS := $(CSRCS:$(srcroot)%.c=$(objroot)%.$(O)) +CPICOBJS := $(CSRCS:$(srcroot)%.c=$(objroot)%.pic.$(O)) +CTESTOBJS := $(CTESTS:$(srcroot)%.c=$(objroot)%.$(O)) .PHONY: all dist doc_html doc_man doc .PHONY: install_bin install_include install_lib .PHONY: install_html install_man install_doc install .PHONY: tests check clean distclean relclean -.SECONDARY : $(CTESTS:@srcroot@%.c=@objroot@%.o) +.SECONDARY : $(CTESTOBJS) # Default target. -all: $(DSOS) $(STATIC_LIBS) +all: build -dist: doc +dist: build_doc -@srcroot@doc/%.html : @objroot@doc/%.xml @srcroot@doc/stylesheet.xsl @objroot@doc/html.xsl - @XSLTPROC@ -o $@ @objroot@doc/html.xsl $< +$(srcroot)doc/%.html : $(objroot)doc/%.xml $(srcroot)doc/stylesheet.xsl $(objroot)doc/html.xsl + $(XSLTPROC) -o $@ $(objroot)doc/html.xsl $< -@srcroot@doc/%.3 : @objroot@doc/%.xml @srcroot@doc/stylesheet.xsl @objroot@doc/manpages.xsl - @XSLTPROC@ -o $@ @objroot@doc/manpages.xsl $< +$(srcroot)doc/%.3 : $(objroot)doc/%.xml $(srcroot)doc/stylesheet.xsl $(objroot)doc/manpages.xsl + $(XSLTPROC) -o $@ $(objroot)doc/manpages.xsl $< -doc_html: $(DOCS_HTML) -doc_man: $(DOCS_MAN3) -doc: $(DOCS) +build_doc_html: $(DOCS_HTML) +build_doc_man: $(DOCS_MAN3) +build_doc: $(DOCS) # # Include generated dependency files. # --include $(CSRCS:@srcroot@%.c=@objroot@%.d) --include $(CSRCS:@srcroot@%.c=@objroot@%.pic.d) --include $(CTESTS:@srcroot@%.c=@objroot@%.d) +ifdef CC_MM +-include $(COBJS:%.$(O)=%.d) +-include $(CPICOBJS:%.$(O)=%.d) +-include $(CTESTOBJS:%.$(O)=%.d) +endif -@objroot@src/%.o: @srcroot@src/%.c +$(COBJS): $(objroot)src/%.$(O): $(srcroot)src/%.c +$(CPICOBJS): $(objroot)src/%.pic.$(O): $(srcroot)src/%.c +$(CPICOBJS): CFLAGS += $(PIC_CFLAGS) +$(CTESTOBJS): $(objroot)test/%.$(O): $(srcroot)test/%.c +$(CTESTOBJS): CPPFLAGS += -I$(objroot)test +ifneq ($(IMPORTLIB),$(SO)) +$(COBJS): CPPFLAGS += -DDLLEXPORT +endif + +ifndef CC_MM +# Dependencies +HEADER_DIRS = $(srcroot)include/jemalloc/internal \ + $(objroot)include/jemalloc $(objroot)include/jemalloc/internal +HEADERS = $(wildcard $(foreach dir,$(HEADER_DIRS),$(dir)/*.h)) +$(COBJS) $(CPICOBJS) $(CTESTOBJS): $(HEADERS) +$(CTESTOBJS): $(objroot)test/jemalloc_test.h +endif + +$(COBJS) $(CPICOBJS) $(CTESTOBJS): %.$(O): @mkdir -p $(@D) - $(CC) $(CFLAGS) -c $(CPPFLAGS) -o $@ $< - @$(SHELL) -ec "$(CC) -MM $(CPPFLAGS) $< | sed \"s/\($(subst /,\/,$(notdir $(basename $@)))\)\.o\([ :]*\)/$(subst /,\/,$(strip $(dir $@)))\1.o \2/g\" > $(@:%.o=%.d)" + $(CC) $(CFLAGS) -c $(CPPFLAGS) $(CTARGET) $< +ifdef CC_MM + @$(CC) -MM $(CPPFLAGS) -MT $@ -o $(@:%.$(O)=%.d) $< +endif -@objroot@src/%.pic.o: @srcroot@src/%.c - @mkdir -p $(@D) - $(CC) $(CFLAGS) -fPIC -DPIC -c $(CPPFLAGS) -o $@ $< - @$(SHELL) -ec "$(CC) -MM $(CPPFLAGS) $< | sed \"s/\($(subst /,\/,$(notdir $(basename $(basename $@))))\)\.o\([ :]*\)/$(subst /,\/,$(strip $(dir $@)))\1.pic.o \2/g\" > $(@:%.o=%.d)" - -%.$(SO) : %.$(SO).$(REV) +ifneq ($(SOREV),$(SO)) +%.$(SO) : %.$(SOREV) @mkdir -p $(@D) ln -sf $( $(@:%.o=%.d)" - -# Automatic dependency generation misses #include "*.c". -@objroot@test/bitmap.o : @objroot@src/bitmap.o - -@objroot@test/%: @objroot@test/%.o \ - @objroot@lib/libjemalloc@install_suffix@.$(SO) - @mkdir -p $(@D) -ifneq (@RPATH@, ) - $(CC) -o $@ $< @RPATH@@objroot@lib -L@objroot@lib -ljemalloc@install_suffix@ -lpthread -else - $(CC) -o $@ $< -L@objroot@lib -ljemalloc@install_suffix@ -lpthread endif +$(objroot)lib/$(LIBJEMALLOC).$(SOREV) : $(if $(PIC_CFLAGS),$(CPICOBJS),$(COBJS)) + @mkdir -p $(@D) + $(CC) $(DSO_LDFLAGS) $(call RPATH,$(RPATH_EXTRA)) $(LDTARGET) $+ $(LDFLAGS) $(LIBS) $(EXTRA_LDFLAGS) + +$(objroot)lib/$(LIBJEMALLOC)_pic.$(A) : $(CPICOBJS) +$(objroot)lib/$(LIBJEMALLOC).$(A) : $(COBJS) +$(objroot)lib/$(LIBJEMALLOC)_s.$(A) : $(COBJS) + +$(STATIC_LIBS): + @mkdir -p $(@D) + $(MKLIB) $+ + +$(objroot)test/bitmap$(EXE): $(objroot)src/bitmap.$(O) + +$(objroot)test/%$(EXE): $(objroot)test/%.$(O) $(objroot)src/util.$(O) $(DSOS) + @mkdir -p $(@D) + $(CC) $(LDTARGET) $(filter %.$(O),$^) $(call RPATH,$(objroot)lib) $(objroot)lib/$(LIBJEMALLOC).$(IMPORTLIB) $(filter -lpthread,$(LIBS)) $(EXTRA_LDFLAGS) + +build_lib_shared: $(DSOS) +build_lib_static: $(STATIC_LIBS) +build: build_lib_shared build_lib_static + install_bin: install -d $(BINDIR) @for b in $(BINS); do \ @@ -155,46 +210,55 @@ install_include: install -m 644 $$h $(INCLUDEDIR)/jemalloc; \ done -install_lib: $(DSOS) $(STATIC_LIBS) +install_lib_shared: $(DSOS) install -d $(LIBDIR) - install -m 755 @objroot@lib/libjemalloc@install_suffix@.$(SO).$(REV) $(LIBDIR) - ln -sf libjemalloc@install_suffix@.$(SO).$(REV) $(LIBDIR)/libjemalloc@install_suffix@.$(SO) - install -m 755 @objroot@lib/libjemalloc@install_suffix@_pic.a $(LIBDIR) - install -m 755 @objroot@lib/libjemalloc@install_suffix@.a $(LIBDIR) + install -m 755 $(objroot)lib/$(LIBJEMALLOC).$(SOREV) $(LIBDIR) +ifneq ($(SOREV),$(SO)) + ln -sf $(LIBJEMALLOC).$(SOREV) $(LIBDIR)/$(LIBJEMALLOC).$(SO) +endif -install_html: - install -d $(DATADIR)/doc/jemalloc@install_suffix@ - @for d in $(DOCS_HTML); do \ - echo "install -m 644 $$d $(DATADIR)/doc/jemalloc@install_suffix@"; \ - install -m 644 $$d $(DATADIR)/doc/jemalloc@install_suffix@; \ +install_lib_static: $(STATIC_LIBS) + install -d $(LIBDIR) + @for l in $(STATIC_LIBS); do \ + echo "install -m 755 $$l $(LIBDIR)"; \ + install -m 755 $$l $(LIBDIR); \ done -install_man: +install_lib: install_lib_shared install_lib_static + +install_doc_html: + install -d $(DATADIR)/doc/jemalloc$(install_suffix) + @for d in $(DOCS_HTML); do \ + echo "install -m 644 $$d $(DATADIR)/doc/jemalloc$(install_suffix)"; \ + install -m 644 $$d $(DATADIR)/doc/jemalloc$(install_suffix); \ +done + +install_doc_man: install -d $(MANDIR)/man3 @for d in $(DOCS_MAN3); do \ echo "install -m 644 $$d $(MANDIR)/man3"; \ install -m 644 $$d $(MANDIR)/man3; \ done -install_doc: install_html install_man +install_doc: install_doc_html install_doc_man install: install_bin install_include install_lib install_doc -tests: $(CTESTS:@srcroot@%.c=@objroot@%) +tests: $(CTESTS:$(srcroot)%.c=$(objroot)%$(EXE)) check: tests - @mkdir -p @objroot@test + @mkdir -p $(objroot)test @$(SHELL) -c 'total=0; \ failures=0; \ echo "========================================="; \ - for t in $(CTESTS:@srcroot@%.c=@objroot@%); do \ + for t in $(CTESTS:$(srcroot)%.c=$(objroot)%); do \ total=`expr $$total + 1`; \ /bin/echo -n "$${t} ... "; \ - $(TEST_LIBRARY_PATH) $${t} @abs_srcroot@ @abs_objroot@ \ - > @objroot@$${t}.out 2>&1; \ - if test -e "@srcroot@$${t}.exp"; then \ - diff -u @srcroot@$${t}.exp \ - @objroot@$${t}.out >/dev/null 2>&1; \ + $(TEST_LIBRARY_PATH) $${t}$(EXE) $(abs_srcroot) \ + $(abs_objroot) > $(objroot)$${t}.out 2>&1; \ + if test -e "$(srcroot)$${t}.exp"; then \ + diff -w -u $(srcroot)$${t}.exp \ + $(objroot)$${t}.out >/dev/null 2>&1; \ fail=$$?; \ if test "$${fail}" -eq "1" ; then \ failures=`expr $${failures} + 1`; \ @@ -211,49 +275,49 @@ check: tests echo "Failures: $${failures}/$${total}"' clean: - rm -f $(CSRCS:@srcroot@%.c=@objroot@%.o) - rm -f $(CSRCS:@srcroot@%.c=@objroot@%.pic.o) - rm -f $(CSRCS:@srcroot@%.c=@objroot@%.d) - rm -f $(CSRCS:@srcroot@%.c=@objroot@%.pic.d) - rm -f $(CTESTS:@srcroot@%.c=@objroot@%) - rm -f $(CTESTS:@srcroot@%.c=@objroot@%.o) - rm -f $(CTESTS:@srcroot@%.c=@objroot@%.d) - rm -f $(CTESTS:@srcroot@%.c=@objroot@%.out) + rm -f $(COBJS) + rm -f $(CPICOBJS) + rm -f $(COBJS:%.$(O)=%.d) + rm -f $(CPICOBJS:%.$(O)=%.d) + rm -f $(CTESTOBJS:%.$(O)=%$(EXE)) + rm -f $(CTESTOBJS) + rm -f $(CTESTOBJS:%.$(O)=%.d) + rm -f $(CTESTOBJS:%.$(O)=%.out) rm -f $(DSOS) $(STATIC_LIBS) distclean: clean - rm -rf @objroot@autom4te.cache - rm -f @objroot@config.log - rm -f @objroot@config.status - rm -f @objroot@config.stamp - rm -f @cfghdrs_out@ - rm -f @cfgoutputs_out@ + rm -rf $(objroot)autom4te.cache + rm -f $(objroot)config.log + rm -f $(objroot)config.status + rm -f $(objroot)config.stamp + rm -f $(cfghdrs_out) + rm -f $(cfgoutputs_out) relclean: distclean - rm -f @objroot@configure - rm -f @srcroot@VERSION + rm -f $(objroot)configure + rm -f $(srcroot)VERSION rm -f $(DOCS_HTML) rm -f $(DOCS_MAN3) #=============================================================================== # Re-configuration rules. -ifeq (@enable_autogen@, 1) -@srcroot@configure : @srcroot@configure.ac - cd ./@srcroot@ && @AUTOCONF@ +ifeq ($(enable_autogen), 1) +$(srcroot)configure : $(srcroot)configure.ac + cd ./$(srcroot) && $(AUTOCONF) -@objroot@config.status : @srcroot@configure - ./@objroot@config.status --recheck +$(objroot)config.status : $(srcroot)configure + ./$(objroot)config.status --recheck -@srcroot@config.stamp.in : @srcroot@configure.ac - echo stamp > @srcroot@config.stamp.in +$(srcroot)config.stamp.in : $(srcroot)configure.ac + echo stamp > $(srcroot)config.stamp.in -@objroot@config.stamp : @cfgoutputs_in@ @cfghdrs_in@ @srcroot@configure - ./@objroot@config.status +$(objroot)config.stamp : $(cfgoutputs_in) $(cfghdrs_in) $(srcroot)configure + ./$(objroot)config.status @touch $@ # There must be some action in order for make to re-read Makefile when it is # out of date. -@cfgoutputs_out@ @cfghdrs_out@ : @objroot@config.stamp +$(cfgoutputs_out) $(cfghdrs_out) : $(objroot)config.stamp @true endif diff --git a/README b/README index 4d7b552b..7661683b 100644 --- a/README +++ b/README @@ -1,10 +1,10 @@ jemalloc is a general-purpose scalable concurrent malloc(3) implementation. -This distribution is a stand-alone "portable" implementation that currently -targets Linux and Apple OS X. jemalloc is included as the default allocator in -the FreeBSD and NetBSD operating systems, and it is used by the Mozilla Firefox -web browser on Microsoft Windows-related platforms. Depending on your needs, -one of the other divergent versions may suit your needs better than this -distribution. +This distribution is a "portable" implementation that currently targets +FreeBSD, Linux, Apple OS X, and MinGW. jemalloc is included as the default +allocator in the FreeBSD and NetBSD operating systems, and it is used by the +Mozilla Firefox web browser on Microsoft Windows-related platforms. Depending +on your needs, one of the other divergent versions may suit your needs better +than this distribution. The COPYING file contains copyright and licensing information. diff --git a/bin/jemalloc.sh.in b/bin/jemalloc.sh.in new file mode 100644 index 00000000..cdf36737 --- /dev/null +++ b/bin/jemalloc.sh.in @@ -0,0 +1,9 @@ +#!/bin/sh + +prefix=@prefix@ +exec_prefix=@exec_prefix@ +libdir=@libdir@ + +@LD_PRELOAD_VAR@=${libdir}/libjemalloc.@SOREV@ +export @LD_PRELOAD_VAR@ +exec "$@" diff --git a/bin/pprof b/bin/pprof index 280ddcc8..727eb437 100755 --- a/bin/pprof +++ b/bin/pprof @@ -72,7 +72,7 @@ use strict; use warnings; use Getopt::Long; -my $PPROF_VERSION = "1.7"; +my $PPROF_VERSION = "2.0"; # These are the object tools we use which can come from a # user-specified location using --tools, from the PPROF_TOOLS @@ -87,13 +87,14 @@ my %obj_tool_map = ( #"addr2line_pdb" => "addr2line-pdb", # ditto #"otool" => "otool", # equivalent of objdump on OS X ); -my $DOT = "dot"; # leave non-absolute, since it may be in /usr/local -my $GV = "gv"; -my $EVINCE = "evince"; # could also be xpdf or perhaps acroread -my $KCACHEGRIND = "kcachegrind"; -my $PS2PDF = "ps2pdf"; +# NOTE: these are lists, so you can put in commandline flags if you want. +my @DOT = ("dot"); # leave non-absolute, since it may be in /usr/local +my @GV = ("gv"); +my @EVINCE = ("evince"); # could also be xpdf or perhaps acroread +my @KCACHEGRIND = ("kcachegrind"); +my @PS2PDF = ("ps2pdf"); # These are used for dynamic profiles -my $URL_FETCHER = "curl -s"; +my @URL_FETCHER = ("curl", "-s"); # These are the web pages that servers need to support for dynamic profiles my $HEAP_PAGE = "/pprof/heap"; @@ -104,7 +105,10 @@ my $GROWTH_PAGE = "/pprof/growth"; my $CONTENTION_PAGE = "/pprof/contention"; my $WALL_PAGE = "/pprof/wall(?:\\?.*)?"; # accepts options like namefilter my $FILTEREDPROFILE_PAGE = "/pprof/filteredprofile(?:\\?.*)?"; -my $CENSUSPROFILE_PAGE = "/pprof/censusprofile"; # must support "?seconds=#" +my $CENSUSPROFILE_PAGE = "/pprof/censusprofile(?:\\?.*)?"; # must support cgi-param + # "?seconds=#", + # "?tags_regexp=#" and + # "?type=#". my $SYMBOL_PAGE = "/pprof/symbol"; # must support symbol lookup via POST my $PROGRAM_NAME_PAGE = "/pprof/cmdline"; @@ -122,6 +126,11 @@ my $UNKNOWN_BINARY = "(unknown)"; # 64-bit profiles. To err on the safe size, default to 64-bit here: my $address_length = 16; +my $dev_null = "/dev/null"; +if (! -e $dev_null && $^O =~ /MSWin/) { # $^O is the OS perl was built for + $dev_null = "nul"; +} + # A list of paths to search for shared object files my @prefix_list = (); @@ -151,7 +160,8 @@ pprof [options] The / can be $HEAP_PAGE, $PROFILE_PAGE, /pprof/pmuprofile, $GROWTH_PAGE, $CONTENTION_PAGE, /pprof/wall, $CENSUSPROFILE_PAGE, or /pprof/filteredprofile. - For instance: "pprof http://myserver.com:80$HEAP_PAGE". + For instance: + pprof http://myserver.com:80$HEAP_PAGE If / is omitted, the service defaults to $PROFILE_PAGE (cpu profiling). pprof --symbols Maps addresses to symbol names. In this mode, stdin should be a @@ -162,7 +172,7 @@ pprof --symbols For more help with querying remote servers, including how to add the necessary server-side support code, see this filename (or one like it): - /usr/doc/google-perftools-$PPROF_VERSION/pprof_remote_servers.html + /usr/doc/gperftools-$PPROF_VERSION/pprof_remote_servers.html Options: --cum Sort by cumulative data @@ -260,7 +270,7 @@ EOF sub version_string { return < 0) { + if (IsProfileURL($ARGV[0])) { + $main::use_symbol_page = 1; + } elsif (IsSymbolizedProfileFile($ARGV[0])) { + $main::use_symbolized_profile = 1; + $main::prog = $UNKNOWN_BINARY; # will be set later from the profile file + } } if ($main::use_symbol_page || $main::use_symbolized_profile) { @@ -540,7 +552,7 @@ sub Init() { ConfigureObjTools($main::prog) } - # Break the opt_list_prefix into the prefix_list array + # Break the opt_lib_prefix into the prefix_list array @prefix_list = split (',', $main::opt_lib_prefix); # Remove trailing / from the prefixes, in the list to prevent @@ -636,9 +648,9 @@ sub Main() { # Print if (!$main::opt_interactive) { if ($main::opt_disasm) { - PrintDisassembly($libs, $flat, $cumulative, $main::opt_disasm, $total); + PrintDisassembly($libs, $flat, $cumulative, $main::opt_disasm); } elsif ($main::opt_list) { - PrintListing($libs, $flat, $cumulative, $main::opt_list); + PrintListing($total, $libs, $flat, $cumulative, $main::opt_list, 0); } elsif ($main::opt_text) { # Make sure the output is empty when have nothing to report # (only matters when --heapcheck is given but we must be @@ -646,7 +658,7 @@ sub Main() { if ($total != 0) { printf("Total: %s %s\n", Unparse($total), Units()); } - PrintText($symbols, $flat, $cumulative, $total, -1); + PrintText($symbols, $flat, $cumulative, -1); } elsif ($main::opt_raw) { PrintSymbolizedProfile($symbols, $profile, $main::prog); } elsif ($main::opt_callgrind) { @@ -656,7 +668,7 @@ sub Main() { if ($main::opt_gv) { RunGV(TempName($main::next_tmpfile, "ps"), ""); } elsif ($main::opt_evince) { - RunEvince(TempName($main::next_tmpfile, "pdf"), ""); + RunEvince(TempName($main::next_tmpfile, "pdf"), ""); } elsif ($main::opt_web) { my $tmp = TempName($main::next_tmpfile, "svg"); RunWeb($tmp); @@ -705,24 +717,25 @@ sub ReadlineMightFail { sub RunGV { my $fname = shift; my $bg = shift; # "" or " &" if we should run in background - if (!system("$GV --version >/dev/null 2>&1")) { + if (!system(ShellEscape(@GV, "--version") . " >$dev_null 2>&1")) { # Options using double dash are supported by this gv version. # Also, turn on noantialias to better handle bug in gv for # postscript files with large dimensions. # TODO: Maybe we should not pass the --noantialias flag # if the gv version is known to work properly without the flag. - system("$GV --scale=$main::opt_scale --noantialias " . $fname . $bg); + system(ShellEscape(@GV, "--scale=$main::opt_scale", "--noantialias", $fname) + . $bg); } else { # Old gv version - only supports options that use single dash. - print STDERR "$GV -scale $main::opt_scale\n"; - system("$GV -scale $main::opt_scale " . $fname . $bg); + print STDERR ShellEscape(@GV, "-scale", $main::opt_scale) . "\n"; + system(ShellEscape(@GV, "-scale", "$main::opt_scale", $fname) . $bg); } } sub RunEvince { my $fname = shift; my $bg = shift; # "" or " &" if we should run in background - system("$EVINCE " . $fname . $bg); + system(ShellEscape(@EVINCE, $fname) . $bg); } sub RunWeb { @@ -756,8 +769,8 @@ sub RunWeb { sub RunKcachegrind { my $fname = shift; my $bg = shift; # "" or " &" if we should run in background - print STDERR "Starting '$KCACHEGRIND " . $fname . $bg . "'\n"; - system("$KCACHEGRIND " . $fname . $bg); + print STDERR "Starting '@KCACHEGRIND " . $fname . $bg . "'\n"; + system(ShellEscape(@KCACHEGRIND, $fname) . $bg); } @@ -834,14 +847,14 @@ sub InteractiveCommand { my $ignore; ($routine, $ignore) = ParseInteractiveArgs($3); - my $profile = ProcessProfile($orig_profile, $symbols, "", $ignore); + my $profile = ProcessProfile($total, $orig_profile, $symbols, "", $ignore); my $reduced = ReduceProfile($symbols, $profile); # Get derived profiles my $flat = FlatProfile($reduced); my $cumulative = CumulativeProfile($reduced); - PrintText($symbols, $flat, $cumulative, $total, $line_limit); + PrintText($symbols, $flat, $cumulative, $line_limit); return 1; } if (m/^\s*callgrind\s*([^ \n]*)/) { @@ -861,21 +874,22 @@ sub InteractiveCommand { return 1; } - if (m/^\s*list\s*(.+)/) { + if (m/^\s*(web)?list\s*(.+)/) { + my $html = (defined($1) && ($1 eq "web")); $main::opt_list = 1; my $routine; my $ignore; - ($routine, $ignore) = ParseInteractiveArgs($1); + ($routine, $ignore) = ParseInteractiveArgs($2); - my $profile = ProcessProfile($orig_profile, $symbols, "", $ignore); + my $profile = ProcessProfile($total, $orig_profile, $symbols, "", $ignore); my $reduced = ReduceProfile($symbols, $profile); # Get derived profiles my $flat = FlatProfile($reduced); my $cumulative = CumulativeProfile($reduced); - PrintListing($libs, $flat, $cumulative, $routine); + PrintListing($total, $libs, $flat, $cumulative, $routine, $html); return 1; } if (m/^\s*disasm\s*(.+)/) { @@ -886,14 +900,14 @@ sub InteractiveCommand { ($routine, $ignore) = ParseInteractiveArgs($1); # Process current profile to account for various settings - my $profile = ProcessProfile($orig_profile, $symbols, "", $ignore); + my $profile = ProcessProfile($total, $orig_profile, $symbols, "", $ignore); my $reduced = ReduceProfile($symbols, $profile); # Get derived profiles my $flat = FlatProfile($reduced); my $cumulative = CumulativeProfile($reduced); - PrintDisassembly($libs, $flat, $cumulative, $routine, $total); + PrintDisassembly($libs, $flat, $cumulative, $routine); return 1; } if (m/^\s*(gv|web|evince)\s*(.*)/) { @@ -913,7 +927,8 @@ sub InteractiveCommand { ($focus, $ignore) = ParseInteractiveArgs($2); # Process current profile to account for various settings - my $profile = ProcessProfile($orig_profile, $symbols, $focus, $ignore); + my $profile = ProcessProfile($total, $orig_profile, $symbols, + $focus, $ignore); my $reduced = ReduceProfile($symbols, $profile); # Get derived profiles @@ -941,6 +956,7 @@ sub InteractiveCommand { sub ProcessProfile { + my $total_count = shift; my $orig_profile = shift; my $symbols = shift; my $focus = shift; @@ -948,7 +964,6 @@ sub ProcessProfile { # Process current profile to account for various settings my $profile = $orig_profile; - my $total_count = TotalProfile($profile); printf("Total: %s %s\n", Unparse($total_count), Units()); if ($focus ne '') { $profile = FocusProfile($symbols, $profile, $focus); @@ -995,6 +1010,11 @@ Commands: list [routine_regexp] [-ignore1] [-ignore2] Show source listing of routines whose names match "routine_regexp" + weblist [routine_regexp] [-ignore1] [-ignore2] + Displays a source listing of routines whose names match "routine_regexp" + in a web browser. You can click on source lines to view the + corresponding disassembly. + top [--cum] [-ignore1] [-ignore2] top20 [--cum] [-ignore1] [-ignore2] top37 [--cum] [-ignore1] [-ignore2] @@ -1019,8 +1039,8 @@ parameters will be ignored. Further pprof details are available at this location (or one similar): - /usr/doc/google-perftools-$PPROF_VERSION/cpu_profiler.html - /usr/doc/google-perftools-$PPROF_VERSION/heap_profiler.html + /usr/doc/gperftools-$PPROF_VERSION/cpu_profiler.html + /usr/doc/gperftools-$PPROF_VERSION/heap_profiler.html ENDOFHELP } @@ -1137,9 +1157,10 @@ sub PrintText { my $symbols = shift; my $flat = shift; my $cumulative = shift; - my $total = shift; my $line_limit = shift; + my $total = TotalProfile($flat); + # Which profile to sort by? my $s = $main::opt_cum ? $cumulative : $flat; @@ -1169,7 +1190,29 @@ sub PrintText { $sym); } $lines++; - last if ($line_limit >= 0 && $lines > $line_limit); + last if ($line_limit >= 0 && $lines >= $line_limit); + } +} + +# Callgrind format has a compression for repeated function and file +# names. You show the name the first time, and just use its number +# subsequently. This can cut down the file to about a third or a +# quarter of its uncompressed size. $key and $val are the key/value +# pair that would normally be printed by callgrind; $map is a map from +# value to number. +sub CompressedCGName { + my($key, $val, $map) = @_; + my $idx = $map->{$val}; + # For very short keys, providing an index hurts rather than helps. + if (length($val) <= 3) { + return "$key=$val\n"; + } elsif (defined($idx)) { + return "$key=($idx)\n"; + } else { + # scalar(keys $map) gives the number of items in the map. + $idx = scalar(keys(%{$map})) + 1; + $map->{$val} = $idx; + return "$key=($idx) $val\n"; } } @@ -1177,13 +1220,16 @@ sub PrintText { sub PrintCallgrind { my $calls = shift; my $filename; + my %filename_to_index_map; + my %fnname_to_index_map; + if ($main::opt_interactive) { $filename = shift; print STDERR "Writing callgrind file to '$filename'.\n" } else { $filename = "&STDOUT"; } - open(CG, ">".$filename ); + open(CG, ">$filename"); printf CG ("events: Hits\n\n"); foreach my $call ( map { $_->[0] } sort { $a->[1] cmp $b ->[1] || @@ -1197,11 +1243,14 @@ sub PrintCallgrind { $callee_file, $callee_line, $callee_function ) = ( $1, $2, $3, $5, $6, $7 ); - - printf CG ("fl=$caller_file\nfn=$caller_function\n"); + # TODO(csilvers): for better compression, collect all the + # caller/callee_files and functions first, before printing + # anything, and only compress those referenced more than once. + printf CG CompressedCGName("fl", $caller_file, \%filename_to_index_map); + printf CG CompressedCGName("fn", $caller_function, \%fnname_to_index_map); if (defined $6) { - printf CG ("cfl=$callee_file\n"); - printf CG ("cfn=$callee_function\n"); + printf CG CompressedCGName("cfl", $callee_file, \%filename_to_index_map); + printf CG CompressedCGName("cfn", $callee_function, \%fnname_to_index_map); printf CG ("calls=$count $callee_line\n"); } printf CG ("$caller_line $count\n\n"); @@ -1214,7 +1263,8 @@ sub PrintDisassembly { my $flat = shift; my $cumulative = shift; my $disasm_opts = shift; - my $total = shift; + + my $total = TotalProfile($flat); foreach my $lib (@{$libs}) { my $symbol_table = GetProcedureBoundaries($lib->[0], $disasm_opts); @@ -1249,10 +1299,10 @@ sub Disassemble { my $end_addr = shift; my $objdump = $obj_tool_map{"objdump"}; - my $cmd = sprintf("$objdump -C -d -l --no-show-raw-insn " . - "--start-address=0x$start_addr " . - "--stop-address=0x$end_addr $prog"); - open(OBJDUMP, "$cmd |") || error("$objdump: $!\n"); + my $cmd = ShellEscape($objdump, "-C", "-d", "-l", "--no-show-raw-insn", + "--start-address=0x$start_addr", + "--stop-address=0x$end_addr", $prog); + open(OBJDUMP, "$cmd |") || error("$cmd: $!\n"); my @result = (); my $filename = ""; my $linenumber = -1; @@ -1315,13 +1365,33 @@ sub ByName { return ShortFunctionName($a) cmp ShortFunctionName($b); } -# Print source-listing for all all routines that match $main::opt_list +# Print source-listing for all all routines that match $list_opts sub PrintListing { + my $total = shift; my $libs = shift; my $flat = shift; my $cumulative = shift; my $list_opts = shift; + my $html = shift; + my $output = \*STDOUT; + my $fname = ""; + + if ($html) { + # Arrange to write the output to a temporary file + $fname = TempName($main::next_tmpfile, "html"); + $main::next_tmpfile++; + if (!open(TEMP, ">$fname")) { + print STDERR "$fname: $!\n"; + return; + } + $output = \*TEMP; + print $output HtmlListingHeader(); + printf $output ("
%s
Total: %s %s
\n", + $main::prog, Unparse($total), Units()); + } + + my $listed = 0; foreach my $lib (@{$libs}) { my $symbol_table = GetProcedureBoundaries($lib->[0], $list_opts); my $offset = AddressSub($lib->[1], $lib->[3]); @@ -1333,15 +1403,113 @@ sub PrintListing { my $addr = AddressAdd($start_addr, $offset); for (my $i = 0; $i < $length; $i++) { if (defined($cumulative->{$addr})) { - PrintSource($lib->[0], $offset, - $routine, $flat, $cumulative, - $start_addr, $end_addr); + $listed += PrintSource( + $lib->[0], $offset, + $routine, $flat, $cumulative, + $start_addr, $end_addr, + $html, + $output); last; } $addr = AddressInc($addr); } } } + + if ($html) { + if ($listed > 0) { + print $output HtmlListingFooter(); + close($output); + RunWeb($fname); + } else { + close($output); + unlink($fname); + } + } +} + +sub HtmlListingHeader { + return <<'EOF'; + + + +Pprof listing + + + + +EOF +} + +sub HtmlListingFooter { + return <<'EOF'; + + +EOF +} + +sub HtmlEscape { + my $text = shift; + $text =~ s/&/&/g; + $text =~ s//>/g; + return $text; } # Returns the indentation of the line, if it has any non-whitespace @@ -1355,6 +1523,45 @@ sub Indentation { } } +# If the symbol table contains inlining info, Disassemble() may tag an +# instruction with a location inside an inlined function. But for +# source listings, we prefer to use the location in the function we +# are listing. So use MapToSymbols() to fetch full location +# information for each instruction and then pick out the first +# location from a location list (location list contains callers before +# callees in case of inlining). +# +# After this routine has run, each entry in $instructions contains: +# [0] start address +# [1] filename for function we are listing +# [2] line number for function we are listing +# [3] disassembly +# [4] limit address +# [5] most specific filename (may be different from [1] due to inlining) +# [6] most specific line number (may be different from [2] due to inlining) +sub GetTopLevelLineNumbers { + my ($lib, $offset, $instructions) = @_; + my $pcs = []; + for (my $i = 0; $i <= $#{$instructions}; $i++) { + push(@{$pcs}, $instructions->[$i]->[0]); + } + my $symbols = {}; + MapToSymbols($lib, $offset, $pcs, $symbols); + for (my $i = 0; $i <= $#{$instructions}; $i++) { + my $e = $instructions->[$i]; + push(@{$e}, $e->[1]); + push(@{$e}, $e->[2]); + my $addr = $e->[0]; + my $sym = $symbols->{$addr}; + if (defined($sym)) { + if ($#{$sym} >= 2 && $sym->[1] =~ m/^(.*):(\d+)$/) { + $e->[1] = $1; # File name + $e->[2] = $2; # Line number + } + } + } +} + # Print source-listing for one routine sub PrintSource { my $prog = shift; @@ -1364,9 +1571,12 @@ sub PrintSource { my $cumulative = shift; my $start_addr = shift; my $end_addr = shift; + my $html = shift; + my $output = shift; # Disassemble all instructions (just to get line numbers) my @instructions = Disassemble($prog, $offset, $start_addr, $end_addr); + GetTopLevelLineNumbers($prog, $offset, \@instructions); # Hack 1: assume that the first source file encountered in the # disassembly contains the routine @@ -1379,7 +1589,7 @@ sub PrintSource { } if (!defined($filename)) { print STDERR "no filename found in $routine\n"; - return; + return 0; } # Hack 2: assume that the largest line number from $filename is the @@ -1412,7 +1622,7 @@ sub PrintSource { { if (!open(FILE, "<$filename")) { print STDERR "$filename: $!\n"; - return; + return 0; } my $l = 0; my $first_indentation = -1; @@ -1440,12 +1650,24 @@ sub PrintSource { # Assign all samples to the range $firstline,$lastline, # Hack 4: If an instruction does not occur in the range, its samples # are moved to the next instruction that occurs in the range. - my $samples1 = {}; - my $samples2 = {}; - my $running1 = 0; # Unassigned flat counts - my $running2 = 0; # Unassigned cumulative counts - my $total1 = 0; # Total flat counts - my $total2 = 0; # Total cumulative counts + my $samples1 = {}; # Map from line number to flat count + my $samples2 = {}; # Map from line number to cumulative count + my $running1 = 0; # Unassigned flat counts + my $running2 = 0; # Unassigned cumulative counts + my $total1 = 0; # Total flat counts + my $total2 = 0; # Total cumulative counts + my %disasm = (); # Map from line number to disassembly + my $running_disasm = ""; # Unassigned disassembly + my $skip_marker = "---\n"; + if ($html) { + $skip_marker = ""; + for (my $l = $firstline; $l <= $lastline; $l++) { + $disasm{$l} = ""; + } + } + my $last_dis_filename = ''; + my $last_dis_linenum = -1; + my $last_touched_line = -1; # To detect gaps in disassembly for a line foreach my $e (@instructions) { # Add up counts for all address that fall inside this instruction my $c1 = 0; @@ -1454,6 +1676,38 @@ sub PrintSource { $c1 += GetEntry($flat, $a); $c2 += GetEntry($cumulative, $a); } + + if ($html) { + my $dis = sprintf(" %6s %6s \t\t%8s: %s ", + HtmlPrintNumber($c1), + HtmlPrintNumber($c2), + UnparseAddress($offset, $e->[0]), + CleanDisassembly($e->[3])); + + # Append the most specific source line associated with this instruction + if (length($dis) < 80) { $dis .= (' ' x (80 - length($dis))) }; + $dis = HtmlEscape($dis); + my $f = $e->[5]; + my $l = $e->[6]; + if ($f ne $last_dis_filename) { + $dis .= sprintf("%s:%d", + HtmlEscape(CleanFileName($f)), $l); + } elsif ($l ne $last_dis_linenum) { + # De-emphasize the unchanged file name portion + $dis .= sprintf("%s" . + ":%d", + HtmlEscape(CleanFileName($f)), $l); + } else { + # De-emphasize the entire location + $dis .= sprintf("%s:%d", + HtmlEscape(CleanFileName($f)), $l); + } + $last_dis_filename = $f; + $last_dis_linenum = $l; + $running_disasm .= $dis; + $running_disasm .= "\n"; + } + $running1 += $c1; $running2 += $c2; $total1 += $c1; @@ -1468,23 +1722,49 @@ sub PrintSource { AddEntry($samples2, $line, $running2); $running1 = 0; $running2 = 0; + if ($html) { + if ($line != $last_touched_line && $disasm{$line} ne '') { + $disasm{$line} .= "\n"; + } + $disasm{$line} .= $running_disasm; + $running_disasm = ''; + $last_touched_line = $line; + } } } # Assign any leftover samples to $lastline AddEntry($samples1, $lastline, $running1); AddEntry($samples2, $lastline, $running2); + if ($html) { + if ($lastline != $last_touched_line && $disasm{$lastline} ne '') { + $disasm{$lastline} .= "\n"; + } + $disasm{$lastline} .= $running_disasm; + } - printf("ROUTINE ====================== %s in %s\n" . - "%6s %6s Total %s (flat / cumulative)\n", - ShortFunctionName($routine), - $filename, - Units(), - Unparse($total1), - Unparse($total2)); + if ($html) { + printf $output ( + "

%s

%s\n
\n" .
+      "Total:%6s %6s (flat / cumulative %s)\n",
+      HtmlEscape(ShortFunctionName($routine)),
+      HtmlEscape(CleanFileName($filename)),
+      Unparse($total1),
+      Unparse($total2),
+      Units());
+  } else {
+    printf $output (
+      "ROUTINE ====================== %s in %s\n" .
+      "%6s %6s Total %s (flat / cumulative)\n",
+      ShortFunctionName($routine),
+      CleanFileName($filename),
+      Unparse($total1),
+      Unparse($total2),
+      Units());
+  }
   if (!open(FILE, "<$filename")) {
     print STDERR "$filename: $!\n";
-    return;
+    return 0;
   }
   my $l = 0;
   while () {
@@ -1494,16 +1774,47 @@ sub PrintSource {
         (($l <= $oldlastline + 5) || ($l <= $lastline))) {
       chop;
       my $text = $_;
-      if ($l == $firstline) { printf("---\n"); }
-      printf("%6s %6s %4d: %s\n",
-             UnparseAlt(GetEntry($samples1, $l)),
-             UnparseAlt(GetEntry($samples2, $l)),
-             $l,
-             $text);
-      if ($l == $lastline)  { printf("---\n"); }
+      if ($l == $firstline) { print $output $skip_marker; }
+      my $n1 = GetEntry($samples1, $l);
+      my $n2 = GetEntry($samples2, $l);
+      if ($html) {
+        # Emit a span that has one of the following classes:
+        #    livesrc -- has samples
+        #    deadsrc -- has disassembly, but with no samples
+        #    nop     -- has no matching disasembly
+        # Also emit an optional span containing disassembly.
+        my $dis = $disasm{$l};
+        my $asm = "";
+        if (defined($dis) && $dis ne '') {
+          $asm = "" . $dis . "";
+        }
+        my $source_class = (($n1 + $n2 > 0) 
+                            ? "livesrc" 
+                            : (($asm ne "") ? "deadsrc" : "nop"));
+        printf $output (
+          "%5d " .
+          "%6s %6s %s%s\n",
+          $l, $source_class,
+          HtmlPrintNumber($n1),
+          HtmlPrintNumber($n2),
+          HtmlEscape($text),
+          $asm);
+      } else {
+        printf $output(
+          "%6s %6s %4d: %s\n",
+          UnparseAlt($n1),
+          UnparseAlt($n2),
+          $l,
+          $text);
+      }
+      if ($l == $lastline)  { print $output $skip_marker; }
     };
   }
   close(FILE);
+  if ($html) {
+    print $output "
\n"; + } + return 1; } # Return the source line for the specified file/linenumber. @@ -1646,21 +1957,11 @@ sub PrintDisassembledFunction { # Print disassembly for (my $x = $first_inst; $x <= $last_inst; $x++) { my $e = $instructions[$x]; - my $address = $e->[0]; - $address = AddressSub($address, $offset); # Make relative to section - $address =~ s/^0x//; - $address =~ s/^0*//; - - # Trim symbols - my $d = $e->[3]; - while ($d =~ s/\([^()%]*\)(\s*const)?//g) { } # Argument types, not (%rax) - while ($d =~ s/(\w+)<[^<>]*>/$1/g) { } # Remove template arguments - printf("%6s %6s %8s: %6s\n", UnparseAlt($flat_count[$x]), UnparseAlt($cum_count[$x]), - $address, - $d); + UnparseAddress($offset, $e->[0]), + CleanDisassembly($e->[3])); } } } @@ -1706,19 +2007,24 @@ sub PrintDot { # Open DOT output file my $output; + my $escaped_dot = ShellEscape(@DOT); + my $escaped_ps2pdf = ShellEscape(@PS2PDF); if ($main::opt_gv) { - $output = "| $DOT -Tps2 >" . TempName($main::next_tmpfile, "ps"); + my $escaped_outfile = ShellEscape(TempName($main::next_tmpfile, "ps")); + $output = "| $escaped_dot -Tps2 >$escaped_outfile"; } elsif ($main::opt_evince) { - $output = "| $DOT -Tps2 | $PS2PDF - " . TempName($main::next_tmpfile, "pdf"); + my $escaped_outfile = ShellEscape(TempName($main::next_tmpfile, "pdf")); + $output = "| $escaped_dot -Tps2 | $escaped_ps2pdf - $escaped_outfile"; } elsif ($main::opt_ps) { - $output = "| $DOT -Tps2"; + $output = "| $escaped_dot -Tps2"; } elsif ($main::opt_pdf) { - $output = "| $DOT -Tps2 | $PS2PDF - -"; + $output = "| $escaped_dot -Tps2 | $escaped_ps2pdf - -"; } elsif ($main::opt_web || $main::opt_svg) { # We need to post-process the SVG, so write to a temporary file always. - $output = "| $DOT -Tsvg >" . TempName($main::next_tmpfile, "svg"); + my $escaped_outfile = ShellEscape(TempName($main::next_tmpfile, "svg")); + $output = "| $escaped_dot -Tsvg >$escaped_outfile"; } elsif ($main::opt_gif) { - $output = "| $DOT -Tgif"; + $output = "| $escaped_dot -Tgif"; } else { $output = ">&STDOUT"; } @@ -1770,7 +2076,7 @@ sub PrintDot { if ($f != $c) { $extra = sprintf("\\rof %s (%s)", Unparse($c), - Percent($c, $overall_total)); + Percent($c, $local_total)); } my $style = ""; if ($main::opt_heapcheck) { @@ -1789,7 +2095,7 @@ sub PrintDot { $node{$a}, $sym, Unparse($f), - Percent($f, $overall_total), + Percent($f, $local_total), $extra, $fs, $style, @@ -1799,10 +2105,12 @@ sub PrintDot { # Get edges and counts per edge my %edge = (); my $n; + my $fullname_to_shortname_map = {}; + FillFullnameToShortnameMap($symbols, $fullname_to_shortname_map); foreach my $k (keys(%{$raw})) { # TODO: omit low %age edges $n = $raw->{$k}; - my @translated = TranslateStack($symbols, $k); + my @translated = TranslateStack($symbols, $fullname_to_shortname_map, $k); for (my $i = 1; $i <= $#translated; $i++) { my $src = $translated[$i]; my $dst = $translated[$i-1]; @@ -2186,6 +2494,50 @@ function handleMouseUp(evt) { EOF } +# Provides a map from fullname to shortname for cases where the +# shortname is ambiguous. The symlist has both the fullname and +# shortname for all symbols, which is usually fine, but sometimes -- +# such as overloaded functions -- two different fullnames can map to +# the same shortname. In that case, we use the address of the +# function to disambiguate the two. This function fills in a map that +# maps fullnames to modified shortnames in such cases. If a fullname +# is not present in the map, the 'normal' shortname provided by the +# symlist is the appropriate one to use. +sub FillFullnameToShortnameMap { + my $symbols = shift; + my $fullname_to_shortname_map = shift; + my $shortnames_seen_once = {}; + my $shortnames_seen_more_than_once = {}; + + foreach my $symlist (values(%{$symbols})) { + # TODO(csilvers): deal with inlined symbols too. + my $shortname = $symlist->[0]; + my $fullname = $symlist->[2]; + if ($fullname !~ /<[0-9a-fA-F]+>$/) { # fullname doesn't end in an address + next; # the only collisions we care about are when addresses differ + } + if (defined($shortnames_seen_once->{$shortname}) && + $shortnames_seen_once->{$shortname} ne $fullname) { + $shortnames_seen_more_than_once->{$shortname} = 1; + } else { + $shortnames_seen_once->{$shortname} = $fullname; + } + } + + foreach my $symlist (values(%{$symbols})) { + my $shortname = $symlist->[0]; + my $fullname = $symlist->[2]; + # TODO(csilvers): take in a list of addresses we care about, and only + # store in the map if $symlist->[1] is in that list. Saves space. + next if defined($fullname_to_shortname_map->{$fullname}); + if (defined($shortnames_seen_more_than_once->{$shortname})) { + if ($fullname =~ /<0*([^>]*)>$/) { # fullname has address at end of it + $fullname_to_shortname_map->{$fullname} = "$shortname\@$1"; + } + } + } +} + # Return a small number that identifies the argument. # Multiple calls with the same argument will return the same number. # Calls with different arguments will return different numbers. @@ -2202,6 +2554,7 @@ sub ShortIdFor { # Translate a stack of addresses into a stack of symbols sub TranslateStack { my $symbols = shift; + my $fullname_to_shortname_map = shift; my $k = shift; my @addrs = split(/\n/, $k); @@ -2233,6 +2586,9 @@ sub TranslateStack { my $func = $symlist->[$j-2]; my $fileline = $symlist->[$j-1]; my $fullfunc = $symlist->[$j]; + if (defined($fullname_to_shortname_map->{$fullfunc})) { + $func = $fullname_to_shortname_map->{$fullfunc}; + } if ($j > 2) { $func = "$func (inline)"; } @@ -2319,6 +2675,16 @@ sub UnparseAlt { } } +# Alternate pretty-printed form: 0 maps to "" +sub HtmlPrintNumber { + my $num = shift; + if ($num == 0) { + return ""; + } else { + return Unparse($num); + } +} + # Return output units sub Units { if ($main::profile_type eq 'heap' || $main::profile_type eq 'growth') { @@ -2475,6 +2841,13 @@ sub RemoveUninterestingFrames { '__builtin_vec_new', 'operator new', 'operator new[]', + # The entry to our memory-allocation routines on OS X + 'malloc_zone_malloc', + 'malloc_zone_calloc', + 'malloc_zone_valloc', + 'malloc_zone_realloc', + 'malloc_zone_memalign', + 'malloc_zone_free', # These mark the beginning/end of our custom sections '__start_google_malloc', '__stop_google_malloc', @@ -2566,9 +2939,11 @@ sub ReduceProfile { my $symbols = shift; my $profile = shift; my $result = {}; + my $fullname_to_shortname_map = {}; + FillFullnameToShortnameMap($symbols, $fullname_to_shortname_map); foreach my $k (keys(%{$profile})) { my $count = $profile->{$k}; - my @translated = TranslateStack($symbols, $k); + my @translated = TranslateStack($symbols, $fullname_to_shortname_map, $k); my @path = (); my %seen = (); $seen{''} = 1; # So that empty keys are skipped @@ -2775,7 +3150,8 @@ sub AddEntries { sub CheckSymbolPage { my $url = SymbolPageURL(); - open(SYMBOL, "$URL_FETCHER '$url' |"); + my $command = ShellEscape(@URL_FETCHER, $url); + open(SYMBOL, "$command |") or error($command); my $line = ; $line =~ s/\r//g; # turn windows-looking lines into unix-looking lines close(SYMBOL); @@ -2832,7 +3208,7 @@ sub SymbolPageURL { sub FetchProgramName() { my ($host, $baseURL, $path) = ParseProfileURL($main::pfile_args[0]); my $url = "$baseURL$PROGRAM_NAME_PAGE"; - my $command_line = "$URL_FETCHER '$url'"; + my $command_line = ShellEscape(@URL_FETCHER, $url); open(CMDLINE, "$command_line |") or error($command_line); my $cmdline = ; $cmdline =~ s/\r//g; # turn windows-looking lines into unix-looking lines @@ -2849,7 +3225,7 @@ sub FetchProgramName() { # curl. Redirection happens on borg hosts. sub ResolveRedirectionForCurl { my $url = shift; - my $command_line = "$URL_FETCHER --head '$url'"; + my $command_line = ShellEscape(@URL_FETCHER, "--head", $url); open(CMDLINE, "$command_line |") or error($command_line); while () { s/\r//g; # turn windows-looking lines into unix-looking lines @@ -2861,18 +3237,18 @@ sub ResolveRedirectionForCurl { return $url; } -# Add a timeout flat to URL_FETCHER +# Add a timeout flat to URL_FETCHER. Returns a new list. sub AddFetchTimeout { - my $fetcher = shift; my $timeout = shift; + my @fetcher = shift; if (defined($timeout)) { - if ($fetcher =~ m/\bcurl -s/) { - $fetcher .= sprintf(" --max-time %d", $timeout); - } elsif ($fetcher =~ m/\brpcget\b/) { - $fetcher .= sprintf(" --deadline=%d", $timeout); + if (join(" ", @fetcher) =~ m/\bcurl -s/) { + push(@fetcher, "--max-time", sprintf("%d", $timeout)); + } elsif (join(" ", @fetcher) =~ m/\brpcget\b/) { + push(@fetcher, sprintf("--deadline=%d", $timeout)); } } - return $fetcher; + return @fetcher; } # Reads a symbol map from the file handle name given as $1, returning @@ -2932,15 +3308,17 @@ sub FetchSymbols { my $url = SymbolPageURL(); my $command_line; - if ($URL_FETCHER =~ m/\bcurl -s/) { + if (join(" ", @URL_FETCHER) =~ m/\bcurl -s/) { $url = ResolveRedirectionForCurl($url); - $command_line = "$URL_FETCHER -d '\@$main::tmpfile_sym' '$url'"; + $command_line = ShellEscape(@URL_FETCHER, "-d", "\@$main::tmpfile_sym", + $url); } else { - $command_line = "$URL_FETCHER --post '$url' < '$main::tmpfile_sym'"; + $command_line = (ShellEscape(@URL_FETCHER, "--post", $url) + . " < " . ShellEscape($main::tmpfile_sym)); } # We use c++filt in case $SYMBOL_PAGE gives us mangled symbols. - my $cppfilt = $obj_tool_map{"c++filt"}; - open(SYMBOL, "$command_line | $cppfilt |") or error($command_line); + my $escaped_cppfilt = ShellEscape($obj_tool_map{"c++filt"}); + open(SYMBOL, "$command_line | $escaped_cppfilt |") or error($command_line); $symbol_map = ReadSymbols(*SYMBOL{IO}); close(SYMBOL); } @@ -2956,8 +3334,8 @@ sub FetchSymbols { my $shortpc = $pc; $shortpc =~ s/^0*//; # Each line may have a list of names, which includes the function - # and also other functions it has inlined. They are separated - # (in PrintSymbolizedFile), by --, which is illegal in function names. + # and also other functions it has inlined. They are separated (in + # PrintSymbolizedProfile), by --, which is illegal in function names. my $fullnames; if (defined($symbol_map->{$shortpc})) { $fullnames = $symbol_map->{$shortpc}; @@ -3035,8 +3413,8 @@ sub FetchDynamicProfile { return $real_profile; } - my $fetcher = AddFetchTimeout($URL_FETCHER, $fetch_timeout); - my $cmd = "$fetcher '$url' > '$tmp_profile'"; + my @fetcher = AddFetchTimeout($fetch_timeout, @URL_FETCHER); + my $cmd = ShellEscape(@fetcher, $url) . " > " . ShellEscape($tmp_profile); if ($path =~ m/$PROFILE_PAGE|$PMUPROFILE_PAGE|$CENSUSPROFILE_PAGE/){ print STDERR "Gathering CPU profile from $url for $main::opt_seconds seconds to\n ${real_profile}\n"; if ($encourage_patience) { @@ -3047,7 +3425,7 @@ sub FetchDynamicProfile { } (system($cmd) == 0) || error("Failed to get profile: $cmd: $!\n"); - (system("mv $tmp_profile $real_profile") == 0) || error("Unable to rename profile\n"); + (system("mv", $tmp_profile, $real_profile) == 0) || error("Unable to rename profile\n"); print STDERR "Wrote profile to $real_profile\n"; $main::collected_profile = $real_profile; return $main::collected_profile; @@ -3161,7 +3539,7 @@ BEGIN { my $has_q = 0; eval { $has_q = pack("Q", "1") ? 1 : 1; }; if (!$has_q) { - $self->{perl_is_64bit} = 0; + $self->{perl_is_64bit} = 0; } read($self->{file}, $str, 8); if (substr($str, 4, 4) eq chr(0)x4) { @@ -3197,17 +3575,17 @@ BEGIN { # TODO(csilvers): if this is a 32-bit perl, the math below # could end up in a too-large int, which perl will promote # to a double, losing necessary precision. Deal with that. - # Right now, we just die. - my ($lo, $hi) = ($b32_values[$i], $b32_values[$i+1]); + # Right now, we just die. + my ($lo, $hi) = ($b32_values[$i], $b32_values[$i+1]); if ($self->{unpack_code} eq 'N') { # big-endian - ($lo, $hi) = ($hi, $lo); - } - my $value = $lo + $hi * (2**32); - if (!$self->{perl_is_64bit} && # check value is exactly represented - (($value % (2**32)) != $lo || int($value / (2**32)) != $hi)) { - ::error("Need a 64-bit perl to process this 64-bit profile.\n"); - } - push(@b64_values, $value); + ($lo, $hi) = ($hi, $lo); + } + my $value = $lo + $hi * (2**32); + if (!$self->{perl_is_64bit} && # check value is exactly represented + (($value % (2**32)) != $lo || int($value / (2**32)) != $hi)) { + ::error("Need a 64-bit perl to process this 64-bit profile.\n"); + } + push(@b64_values, $value); } @$slots = @b64_values; } @@ -3335,7 +3713,7 @@ sub ReadProfile { if (!$main::use_symbolized_profile) { # we have both a binary and symbolized profiles, abort error("FATAL ERROR: Symbolized profile\n $fname\ncannot be used with " . - "a binary arg. Try again without passing\n $prog\n"); + "a binary arg. Try again without passing\n $prog\n"); } # Read the symbol section of the symbolized profile file. $symbols = ReadSymbols(*PROFILE{IO}); @@ -3636,18 +4014,18 @@ sub ReadHeapProfile { # The sampling frequency is the rate of a Poisson process. # This means that the probability of sampling an allocation of # size X with sampling rate Y is 1 - exp(-X/Y) - if ($n1 != 0) { - my $ratio = (($s1*1.0)/$n1)/($sample_adjustment); - my $scale_factor = 1/(1 - exp(-$ratio)); - $n1 *= $scale_factor; - $s1 *= $scale_factor; - } - if ($n2 != 0) { - my $ratio = (($s2*1.0)/$n2)/($sample_adjustment); - my $scale_factor = 1/(1 - exp(-$ratio)); - $n2 *= $scale_factor; - $s2 *= $scale_factor; - } + if ($n1 != 0) { + my $ratio = (($s1*1.0)/$n1)/($sample_adjustment); + my $scale_factor = 1/(1 - exp(-$ratio)); + $n1 *= $scale_factor; + $s1 *= $scale_factor; + } + if ($n2 != 0) { + my $ratio = (($s2*1.0)/$n2)/($sample_adjustment); + my $scale_factor = 1/(1 - exp(-$ratio)); + $n2 *= $scale_factor; + $s2 *= $scale_factor; + } } else { # Remote-heap version 1 my $ratio; @@ -3771,19 +4149,19 @@ sub ReadSynchProfile { return $r; } -# Given a hex value in the form "0x1abcd" return "0001abcd" or -# "000000000001abcd", depending on the current address length. -# There's probably a more idiomatic (or faster) way to do this... +# Given a hex value in the form "0x1abcd" or "1abcd", return either +# "0001abcd" or "000000000001abcd", depending on the current (global) +# address length. sub HexExtend { my $addr = shift; - $addr =~ s/^0x//; - - if (length $addr > $address_length) { - printf STDERR "Warning: address $addr is longer than address length $address_length\n"; + $addr =~ s/^(0x)?0*//; + my $zeros_needed = $address_length - length($addr); + if ($zeros_needed < 0) { + printf STDERR "Warning: address $addr is longer than address length $address_length\n"; + return $addr; } - - return substr("000000000000000".$addr, -$address_length); + return ("0" x $zeros_needed) . $addr; } ##### Symbol extraction ##### @@ -3834,9 +4212,8 @@ sub ParseTextSectionHeaderFromObjdump { my $file_offset; # Get objdump output from the library file to figure out how to # map between mapped addresses and addresses in the library. - my $objdump = $obj_tool_map{"objdump"}; - open(OBJDUMP, "$objdump -h $lib |") - || error("$objdump $lib: $!\n"); + my $cmd = ShellEscape($obj_tool_map{"objdump"}, "-h", $lib); + open(OBJDUMP, "$cmd |") || error("$cmd: $!\n"); while () { s/\r//g; # turn windows-looking lines into unix-looking lines # Idx Name Size VMA LMA File off Algn @@ -3874,9 +4251,8 @@ sub ParseTextSectionHeaderFromOtool { my $file_offset = undef; # Get otool output from the library file to figure out how to # map between mapped addresses and addresses in the library. - my $otool = $obj_tool_map{"otool"}; - open(OTOOL, "$otool -l $lib |") - || error("$otool $lib: $!\n"); + my $command = ShellEscape($obj_tool_map{"otool"}, "-l", $lib); + open(OTOOL, "$command |") || error("$command: $!\n"); my $cmd = ""; my $sectname = ""; my $segname = ""; @@ -4218,18 +4594,18 @@ sub ExtractSymbols { my ($start_pc_index, $finish_pc_index); # Find smallest finish_pc_index such that $finish < $pc[$finish_pc_index]. for ($finish_pc_index = $#pcs + 1; $finish_pc_index > 0; - $finish_pc_index--) { + $finish_pc_index--) { last if $pcs[$finish_pc_index - 1] le $finish; } # Find smallest start_pc_index such that $start <= $pc[$start_pc_index]. for ($start_pc_index = $finish_pc_index; $start_pc_index > 0; - $start_pc_index--) { + $start_pc_index--) { last if $pcs[$start_pc_index - 1] lt $start; } # This keeps PC values higher than $pc[$finish_pc_index] in @pcs, # in case there are overlaps in libraries and the main binary. @{$contained} = splice(@pcs, $start_pc_index, - $finish_pc_index - $start_pc_index); + $finish_pc_index - $start_pc_index); # Map to symbols MapToSymbols($libname, AddressSub($start, $offset), $contained, $symbols); } @@ -4251,15 +4627,15 @@ sub MapToSymbols { # Figure out the addr2line command to use my $addr2line = $obj_tool_map{"addr2line"}; - my $cmd = "$addr2line -f -C -e $image"; + my $cmd = ShellEscape($addr2line, "-f", "-C", "-e", $image); if (exists $obj_tool_map{"addr2line_pdb"}) { $addr2line = $obj_tool_map{"addr2line_pdb"}; - $cmd = "$addr2line --demangle -f -C -e $image"; + $cmd = ShellEscape($addr2line, "--demangle", "-f", "-C", "-e", $image); } # If "addr2line" isn't installed on the system at all, just use # nm to get what info we can (function names, but not line numbers). - if (system("$addr2line --help >/dev/null 2>&1") != 0) { + if (system(ShellEscape($addr2line, "--help") . " >$dev_null 2>&1") != 0) { MapSymbolsWithNM($image, $offset, $pclist, $symbols); return; } @@ -4273,11 +4649,10 @@ sub MapToSymbols { $sep_address = undef; # May be filled in by MapSymbolsWithNM() my $nm_symbols = {}; MapSymbolsWithNM($image, $offset, $pclist, $nm_symbols); - # TODO(csilvers): only add '-i' if addr2line supports it. if (defined($sep_address)) { # Only add " -i" to addr2line if the binary supports it. # addr2line --help returns 0, but not if it sees an unknown flag first. - if (system("$cmd -i --help >/dev/null 2>&1") == 0) { + if (system("$cmd -i --help >$dev_null 2>&1") == 0) { $cmd .= " -i"; } else { $sep_address = undef; # no need for sep_address if we don't support -i @@ -4299,13 +4674,14 @@ sub MapToSymbols { close(ADDRESSES); if ($debug) { print("----\n"); - system("cat $main::tmpfile_sym"); + system("cat", $main::tmpfile_sym); print("----\n"); - system("$cmd <$main::tmpfile_sym"); + system("$cmd < " . ShellEscape($main::tmpfile_sym)); print("----\n"); } - open(SYMBOLS, "$cmd <$main::tmpfile_sym |") || error("$cmd: $!\n"); + open(SYMBOLS, "$cmd <" . ShellEscape($main::tmpfile_sym) . " |") + || error("$cmd: $!\n"); my $count = 0; # Index in pclist while () { # Read fullfunction and filelineinfo from next pair of lines @@ -4325,15 +4701,29 @@ sub MapToSymbols { my $pcstr = $pclist->[$count]; my $function = ShortFunctionName($fullfunction); - if ($fullfunction eq '??') { - # See if nm found a symbol - my $nms = $nm_symbols->{$pcstr}; - if (defined($nms)) { + my $nms = $nm_symbols->{$pcstr}; + if (defined($nms)) { + if ($fullfunction eq '??') { + # nm found a symbol for us. $function = $nms->[0]; $fullfunction = $nms->[2]; + } else { + # MapSymbolsWithNM tags each routine with its starting address, + # useful in case the image has multiple occurrences of this + # routine. (It uses a syntax that resembles template paramters, + # that are automatically stripped out by ShortFunctionName().) + # addr2line does not provide the same information. So we check + # if nm disambiguated our symbol, and if so take the annotated + # (nm) version of the routine-name. TODO(csilvers): this won't + # catch overloaded, inlined symbols, which nm doesn't see. + # Better would be to do a check similar to nm's, in this fn. + if ($nms->[2] =~ m/^\Q$function\E/) { # sanity check it's the right fn + $function = $nms->[0]; + $fullfunction = $nms->[2]; + } } } - + # Prepend to accumulated symbols for pcstr # (so that caller comes before callee) my $sym = $symbols->{$pcstr}; @@ -4344,7 +4734,7 @@ sub MapToSymbols { unshift(@{$sym}, $function, $filelinenum, $fullfunction); if ($debug) { printf STDERR ("%s => [%s]\n", $pcstr, join(" ", @{$sym})); } if (!defined($sep_address)) { - # Inlining is off, se this entry ends immediately + # Inlining is off, so this entry ends immediately $count++; } } @@ -4407,6 +4797,31 @@ sub ShortFunctionName { return $function; } +# Trim overly long symbols found in disassembler output +sub CleanDisassembly { + my $d = shift; + while ($d =~ s/\([^()%]*\)(\s*const)?//g) { } # Argument types, not (%rax) + while ($d =~ s/(\w+)<[^<>]*>/$1/g) { } # Remove template arguments + return $d; +} + +# Clean file name for display +sub CleanFileName { + my ($f) = @_; + $f =~ s|^/proc/self/cwd/||; + $f =~ s|^\./||; + return $f; +} + +# Make address relative to section and clean up for display +sub UnparseAddress { + my ($offset, $address) = @_; + $address = AddressSub($address, $offset); + $address =~ s/^0x//; + $address =~ s/^0*//; + return $address; +} + ##### Miscellaneous ##### # Find the right versions of the above object tools to use. The @@ -4423,8 +4838,18 @@ sub ConfigureObjTools { # predictably return error status in prod. (-e $prog_file) || error("$prog_file does not exist.\n"); - # Follow symlinks (at least for systems where "file" supports that) - my $file_type = `/usr/bin/file -L $prog_file 2>/dev/null || /usr/bin/file $prog_file`; + my $file_type = undef; + if (-e "/usr/bin/file") { + # Follow symlinks (at least for systems where "file" supports that). + my $escaped_prog_file = ShellEscape($prog_file); + $file_type = `/usr/bin/file -L $escaped_prog_file 2>$dev_null || + /usr/bin/file $escaped_prog_file`; + } elsif ($^O == "MSWin32") { + $file_type = "MS Windows"; + } else { + print STDERR "WARNING: Can't determine the file type of $prog_file"; + } + if ($file_type =~ /64-bit/) { # Change $address_length to 16 if the program file is ELF 64-bit. # We can't detect this from many (most?) heap or lock contention @@ -4500,6 +4925,19 @@ sub ConfigureTool { return $path; } +sub ShellEscape { + my @escaped_words = (); + foreach my $word (@_) { + my $escaped_word = $word; + if ($word =~ m![^a-zA-Z0-9/.,_=-]!) { # check for anything not in whitelist + $escaped_word =~ s/'/'\\''/; + $escaped_word = "'$escaped_word'"; + } + push(@escaped_words, $escaped_word); + } + return join(" ", @escaped_words); +} + sub cleanup { unlink($main::tmpfile_sym); unlink(keys %main::tempnames); @@ -4537,11 +4975,11 @@ sub error { # names match "$regexp" and returns them in a hashtable mapping from # procedure name to a two-element vector of [start address, end address] sub GetProcedureBoundariesViaNm { - my $nm_command = shift; + my $escaped_nm_command = shift; # shell-escaped my $regexp = shift; my $symbol_table = {}; - open(NM, "$nm_command |") || error("$nm_command: $!\n"); + open(NM, "$escaped_nm_command |") || error("$escaped_nm_command: $!\n"); my $last_start = "0"; my $routine = ""; while () { @@ -4619,6 +5057,21 @@ sub GetProcedureBoundaries { my $image = shift; my $regexp = shift; + # If $image doesn't start with /, then put ./ in front of it. This works + # around an obnoxious bug in our probing of nm -f behavior. + # "nm -f $image" is supposed to fail on GNU nm, but if: + # + # a. $image starts with [BbSsPp] (for example, bin/foo/bar), AND + # b. you have a.out in your current directory (a not uncommon occurence) + # + # then "nm -f $image" succeeds because -f only looks at the first letter of + # the argument, which looks valid because it's [BbSsPp], and then since + # there's no image provided, it looks for a.out and finds it. + # + # This regex makes sure that $image starts with . or /, forcing the -f + # parsing to fail since . and / are not valid formats. + $image =~ s#^[^/]#./$&#; + # For libc libraries, the copy in /usr/lib/debug contains debugging symbols my $debugging = DebuggingLibrary($image); if ($debugging) { @@ -4636,28 +5089,29 @@ sub GetProcedureBoundaries { # --demangle and -f. my $demangle_flag = ""; my $cppfilt_flag = ""; - if (system("$nm --demangle $image >/dev/null 2>&1") == 0) { + my $to_devnull = ">$dev_null 2>&1"; + if (system(ShellEscape($nm, "--demangle", "image") . $to_devnull) == 0) { # In this mode, we do "nm --demangle " $demangle_flag = "--demangle"; $cppfilt_flag = ""; - } elsif (system("$cppfilt $image >/dev/null 2>&1") == 0) { + } elsif (system(ShellEscape($cppfilt, $image) . $to_devnull) == 0) { # In this mode, we do "nm | c++filt" - $cppfilt_flag = " | $cppfilt"; + $cppfilt_flag = " | " . ShellEscape($cppfilt); }; my $flatten_flag = ""; - if (system("$nm -f $image >/dev/null 2>&1") == 0) { + if (system(ShellEscape($nm, "-f", $image) . $to_devnull) == 0) { $flatten_flag = "-f"; } # Finally, in the case $imagie isn't a debug library, we try again with # -D to at least get *exported* symbols. If we can't use --demangle, # we use c++filt instead, if it exists on this system. - my @nm_commands = ("$nm -n $flatten_flag $demangle_flag" . - " $image 2>/dev/null $cppfilt_flag", - "$nm -D -n $flatten_flag $demangle_flag" . - " $image 2>/dev/null $cppfilt_flag", + my @nm_commands = (ShellEscape($nm, "-n", $flatten_flag, $demangle_flag, + $image) . " 2>$dev_null $cppfilt_flag", + ShellEscape($nm, "-D", "-n", $flatten_flag, $demangle_flag, + $image) . " 2>$dev_null $cppfilt_flag", # 6nm is for Go binaries - "6nm $image 2>/dev/null | sort", + ShellEscape("6nm", "$image") . " 2>$dev_null | sort", ); # If the executable is an MS Windows PDB-format executable, we'll @@ -4665,8 +5119,9 @@ sub GetProcedureBoundaries { # want to use both unix nm and windows-specific nm_pdb, since # PDB-format executables can apparently include dwarf .o files. if (exists $obj_tool_map{"nm_pdb"}) { - my $nm_pdb = $obj_tool_map{"nm_pdb"}; - push(@nm_commands, "$nm_pdb --demangle $image 2>/dev/null"); + push(@nm_commands, + ShellEscape($obj_tool_map{"nm_pdb"}, "--demangle", $image) + . " 2>$dev_null"); } foreach my $nm_command (@nm_commands) { diff --git a/config.guess b/config.guess index 0773d0f6..d622a44e 100755 --- a/config.guess +++ b/config.guess @@ -1,9 +1,10 @@ #! /bin/sh # Attempt to guess a canonical system name. # Copyright (C) 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, -# 2000, 2001, 2002, 2003 Free Software Foundation, Inc. +# 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, +# 2011, 2012 Free Software Foundation, Inc. -timestamp='2004-03-03' +timestamp='2012-02-10' # This file is free software; you can redistribute it and/or modify it # under the terms of the GNU General Public License as published by @@ -16,24 +17,24 @@ timestamp='2004-03-03' # General Public License for more details. # # You should have received a copy of the GNU General Public License -# along with this program; if not, write to the Free Software -# Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. +# along with this program; if not, see . # # As a special exception to the GNU General Public License, if you # distribute this file as part of a program that contains a # configuration script generated by Autoconf, you may include it under # the same distribution terms that you use for the rest of that program. -# Originally written by Per Bothner . -# Please send patches to . Submit a context -# diff and a properly formatted ChangeLog entry. + +# Originally written by Per Bothner. Please send patches (context +# diff format) to and include a ChangeLog +# entry. # # This script attempts to guess a canonical system name similar to # config.sub. If it succeeds, it prints the system name on stdout, and # exits with 0. Otherwise, it exits with 1. # -# The plan is that this can be called by configure scripts if you -# don't specify an explicit build system type. +# You can get the latest version of this script from: +# http://git.savannah.gnu.org/gitweb/?p=config.git;a=blob_plain;f=config.guess;hb=HEAD me=`echo "$0" | sed -e 's,.*/,,'` @@ -53,7 +54,8 @@ version="\ GNU config.guess ($timestamp) Originally written by Per Bothner. -Copyright (C) 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2001 +Copyright (C) 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000, +2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012 Free Software Foundation, Inc. This is free software; see the source for copying conditions. There is NO @@ -66,11 +68,11 @@ Try \`$me --help' for more information." while test $# -gt 0 ; do case $1 in --time-stamp | --time* | -t ) - echo "$timestamp" ; exit 0 ;; + echo "$timestamp" ; exit ;; --version | -v ) - echo "$version" ; exit 0 ;; + echo "$version" ; exit ;; --help | --h* | -h ) - echo "$usage"; exit 0 ;; + echo "$usage"; exit ;; -- ) # Stop option processing shift; break ;; - ) # Use stdin as input. @@ -104,7 +106,7 @@ set_cc_for_build=' trap "exitcode=\$?; (rm -f \$tmpfiles 2>/dev/null; rmdir \$tmp 2>/dev/null) && exit \$exitcode" 0 ; trap "rm -f \$tmpfiles 2>/dev/null; rmdir \$tmp 2>/dev/null; exit 1" 1 2 13 15 ; : ${TMPDIR=/tmp} ; - { tmp=`(umask 077 && mktemp -d -q "$TMPDIR/cgXXXXXX") 2>/dev/null` && test -n "$tmp" && test -d "$tmp" ; } || + { tmp=`(umask 077 && mktemp -d "$TMPDIR/cgXXXXXX") 2>/dev/null` && test -n "$tmp" && test -d "$tmp" ; } || { test -n "$RANDOM" && tmp=$TMPDIR/cg$$-$RANDOM && (umask 077 && mkdir $tmp) ; } || { tmp=$TMPDIR/cg-$$ && (umask 077 && mkdir $tmp) && echo "Warning: creating insecure temp directory" >&2 ; } || { echo "$me: cannot create a temporary directory in $TMPDIR" >&2 ; exit 1 ; } ; @@ -123,7 +125,7 @@ case $CC_FOR_BUILD,$HOST_CC,$CC in ;; ,,*) CC_FOR_BUILD=$CC ;; ,*,*) CC_FOR_BUILD=$HOST_CC ;; -esac ;' +esac ; set_cc_for_build= ;' # This is needed to find uname on a Pyramid OSx when run in the BSD universe. # (ghazi@noc.rutgers.edu 1994-08-24) @@ -141,7 +143,7 @@ UNAME_VERSION=`(uname -v) 2>/dev/null` || UNAME_VERSION=unknown case "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}" in *:NetBSD:*:*) # NetBSD (nbsd) targets should (where applicable) match one or - # more of the tupples: *-*-netbsdelf*, *-*-netbsdaout*, + # more of the tuples: *-*-netbsdelf*, *-*-netbsdaout*, # *-*-netbsdecoff* and *-*-netbsd*. For targets that recently # switched to ELF, *-*-netbsd* would select the old # object file format. This provides both forward @@ -158,6 +160,7 @@ case "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}" in arm*) machine=arm-unknown ;; sh3el) machine=shl-unknown ;; sh3eb) machine=sh-unknown ;; + sh5el) machine=sh5le-unknown ;; *) machine=${UNAME_MACHINE_ARCH}-unknown ;; esac # The Operating System including object format, if it has switched @@ -166,7 +169,7 @@ case "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}" in arm*|i386|m68k|ns32k|sh3*|sparc|vax) eval $set_cc_for_build if echo __ELF__ | $CC_FOR_BUILD -E - 2>/dev/null \ - | grep __ELF__ >/dev/null + | grep -q __ELF__ then # Once all utilities can be ECOFF (netbsdecoff) or a.out (netbsdaout). # Return netbsd for either. FIX? @@ -176,7 +179,7 @@ case "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}" in fi ;; *) - os=netbsd + os=netbsd ;; esac # The OS release @@ -196,71 +199,30 @@ case "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}" in # contains redundant information, the shorter form: # CPU_TYPE-MANUFACTURER-OPERATING_SYSTEM is used. echo "${machine}-${os}${release}" - exit 0 ;; - amd64:OpenBSD:*:*) - echo x86_64-unknown-openbsd${UNAME_RELEASE} - exit 0 ;; - amiga:OpenBSD:*:*) - echo m68k-unknown-openbsd${UNAME_RELEASE} - exit 0 ;; - arc:OpenBSD:*:*) - echo mipsel-unknown-openbsd${UNAME_RELEASE} - exit 0 ;; - cats:OpenBSD:*:*) - echo arm-unknown-openbsd${UNAME_RELEASE} - exit 0 ;; - hp300:OpenBSD:*:*) - echo m68k-unknown-openbsd${UNAME_RELEASE} - exit 0 ;; - mac68k:OpenBSD:*:*) - echo m68k-unknown-openbsd${UNAME_RELEASE} - exit 0 ;; - macppc:OpenBSD:*:*) - echo powerpc-unknown-openbsd${UNAME_RELEASE} - exit 0 ;; - mvme68k:OpenBSD:*:*) - echo m68k-unknown-openbsd${UNAME_RELEASE} - exit 0 ;; - mvme88k:OpenBSD:*:*) - echo m88k-unknown-openbsd${UNAME_RELEASE} - exit 0 ;; - mvmeppc:OpenBSD:*:*) - echo powerpc-unknown-openbsd${UNAME_RELEASE} - exit 0 ;; - pegasos:OpenBSD:*:*) - echo powerpc-unknown-openbsd${UNAME_RELEASE} - exit 0 ;; - pmax:OpenBSD:*:*) - echo mipsel-unknown-openbsd${UNAME_RELEASE} - exit 0 ;; - sgi:OpenBSD:*:*) - echo mipseb-unknown-openbsd${UNAME_RELEASE} - exit 0 ;; - sun3:OpenBSD:*:*) - echo m68k-unknown-openbsd${UNAME_RELEASE} - exit 0 ;; - wgrisc:OpenBSD:*:*) - echo mipsel-unknown-openbsd${UNAME_RELEASE} - exit 0 ;; + exit ;; *:OpenBSD:*:*) - echo ${UNAME_MACHINE}-unknown-openbsd${UNAME_RELEASE} - exit 0 ;; + UNAME_MACHINE_ARCH=`arch | sed 's/OpenBSD.//'` + echo ${UNAME_MACHINE_ARCH}-unknown-openbsd${UNAME_RELEASE} + exit ;; *:ekkoBSD:*:*) echo ${UNAME_MACHINE}-unknown-ekkobsd${UNAME_RELEASE} - exit 0 ;; + exit ;; + *:SolidBSD:*:*) + echo ${UNAME_MACHINE}-unknown-solidbsd${UNAME_RELEASE} + exit ;; macppc:MirBSD:*:*) - echo powerppc-unknown-mirbsd${UNAME_RELEASE} - exit 0 ;; + echo powerpc-unknown-mirbsd${UNAME_RELEASE} + exit ;; *:MirBSD:*:*) echo ${UNAME_MACHINE}-unknown-mirbsd${UNAME_RELEASE} - exit 0 ;; + exit ;; alpha:OSF1:*:*) case $UNAME_RELEASE in *4.0) UNAME_RELEASE=`/usr/sbin/sizer -v | awk '{print $3}'` ;; *5.*) - UNAME_RELEASE=`/usr/sbin/sizer -v | awk '{print $4}'` + UNAME_RELEASE=`/usr/sbin/sizer -v | awk '{print $4}'` ;; esac # According to Compaq, /usr/sbin/psrinfo has been available on @@ -306,40 +268,46 @@ case "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}" in # A Xn.n version is an unreleased experimental baselevel. # 1.2 uses "1.2" for uname -r. echo ${UNAME_MACHINE}-dec-osf`echo ${UNAME_RELEASE} | sed -e 's/^[PVTX]//' | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz'` - exit 0 ;; - Alpha*:OpenVMS:*:*) - echo alpha-hp-vms - exit 0 ;; + # Reset EXIT trap before exiting to avoid spurious non-zero exit code. + exitcode=$? + trap '' 0 + exit $exitcode ;; Alpha\ *:Windows_NT*:*) # How do we know it's Interix rather than the generic POSIX subsystem? # Should we change UNAME_MACHINE based on the output of uname instead # of the specific Alpha model? echo alpha-pc-interix - exit 0 ;; + exit ;; 21064:Windows_NT:50:3) echo alpha-dec-winnt3.5 - exit 0 ;; + exit ;; Amiga*:UNIX_System_V:4.0:*) echo m68k-unknown-sysv4 - exit 0;; + exit ;; *:[Aa]miga[Oo][Ss]:*:*) echo ${UNAME_MACHINE}-unknown-amigaos - exit 0 ;; + exit ;; *:[Mm]orph[Oo][Ss]:*:*) echo ${UNAME_MACHINE}-unknown-morphos - exit 0 ;; + exit ;; *:OS/390:*:*) echo i370-ibm-openedition - exit 0 ;; + exit ;; + *:z/VM:*:*) + echo s390-ibm-zvmoe + exit ;; *:OS400:*:*) - echo powerpc-ibm-os400 - exit 0 ;; + echo powerpc-ibm-os400 + exit ;; arm:RISC*:1.[012]*:*|arm:riscix:1.[012]*:*) echo arm-acorn-riscix${UNAME_RELEASE} - exit 0;; + exit ;; + arm:riscos:*:*|arm:RISCOS:*:*) + echo arm-unknown-riscos + exit ;; SR2?01:HI-UX/MPP:*:* | SR8000:HI-UX/MPP:*:*) echo hppa1.1-hitachi-hiuxmpp - exit 0;; + exit ;; Pyramid*:OSx*:*:* | MIS*:OSx*:*:* | MIS*:SMP_DC-OSx*:*:*) # akee@wpdis03.wpafb.af.mil (Earle F. Ake) contributed MIS and NILE. if test "`(/bin/universe) 2>/dev/null`" = att ; then @@ -347,32 +315,51 @@ case "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}" in else echo pyramid-pyramid-bsd fi - exit 0 ;; + exit ;; NILE*:*:*:dcosx) echo pyramid-pyramid-svr4 - exit 0 ;; + exit ;; DRS?6000:unix:4.0:6*) echo sparc-icl-nx6 - exit 0 ;; - DRS?6000:UNIX_SV:4.2*:7*) + exit ;; + DRS?6000:UNIX_SV:4.2*:7* | DRS?6000:isis:4.2*:7*) case `/usr/bin/uname -p` in - sparc) echo sparc-icl-nx7 && exit 0 ;; + sparc) echo sparc-icl-nx7; exit ;; esac ;; + s390x:SunOS:*:*) + echo ${UNAME_MACHINE}-ibm-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'` + exit ;; sun4H:SunOS:5.*:*) echo sparc-hal-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'` - exit 0 ;; + exit ;; sun4*:SunOS:5.*:* | tadpole*:SunOS:5.*:*) echo sparc-sun-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'` - exit 0 ;; - i86pc:SunOS:5.*:*) - echo i386-pc-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'` - exit 0 ;; + exit ;; + i86pc:AuroraUX:5.*:* | i86xen:AuroraUX:5.*:*) + echo i386-pc-auroraux${UNAME_RELEASE} + exit ;; + i86pc:SunOS:5.*:* | i86xen:SunOS:5.*:*) + eval $set_cc_for_build + SUN_ARCH="i386" + # If there is a compiler, see if it is configured for 64-bit objects. + # Note that the Sun cc does not turn __LP64__ into 1 like gcc does. + # This test works for both compilers. + if [ "$CC_FOR_BUILD" != 'no_compiler_found' ]; then + if (echo '#ifdef __amd64'; echo IS_64BIT_ARCH; echo '#endif') | \ + (CCOPTS= $CC_FOR_BUILD -E - 2>/dev/null) | \ + grep IS_64BIT_ARCH >/dev/null + then + SUN_ARCH="x86_64" + fi + fi + echo ${SUN_ARCH}-pc-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'` + exit ;; sun4*:SunOS:6*:*) # According to config.sub, this is the proper way to canonicalize # SunOS6. Hard to guess exactly what SunOS6 will be like, but # it's likely to be more like Solaris than SunOS4. echo sparc-sun-solaris3`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'` - exit 0 ;; + exit ;; sun4*:SunOS:*:*) case "`/usr/bin/arch -k`" in Series*|S4*) @@ -381,10 +368,10 @@ case "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}" in esac # Japanese Language versions have a version number like `4.1.3-JL'. echo sparc-sun-sunos`echo ${UNAME_RELEASE}|sed -e 's/-/_/'` - exit 0 ;; + exit ;; sun3*:SunOS:*:*) echo m68k-sun-sunos${UNAME_RELEASE} - exit 0 ;; + exit ;; sun*:*:4.2BSD:*) UNAME_RELEASE=`(sed 1q /etc/motd | awk '{print substr($5,1,3)}') 2>/dev/null` test "x${UNAME_RELEASE}" = "x" && UNAME_RELEASE=3 @@ -396,10 +383,10 @@ case "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}" in echo sparc-sun-sunos${UNAME_RELEASE} ;; esac - exit 0 ;; + exit ;; aushp:SunOS:*:*) echo sparc-auspex-sunos${UNAME_RELEASE} - exit 0 ;; + exit ;; # The situation for MiNT is a little confusing. The machine name # can be virtually everything (everything which is not # "atarist" or "atariste" at least should have a processor @@ -409,41 +396,41 @@ case "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}" in # MiNT. But MiNT is downward compatible to TOS, so this should # be no problem. atarist[e]:*MiNT:*:* | atarist[e]:*mint:*:* | atarist[e]:*TOS:*:*) - echo m68k-atari-mint${UNAME_RELEASE} - exit 0 ;; + echo m68k-atari-mint${UNAME_RELEASE} + exit ;; atari*:*MiNT:*:* | atari*:*mint:*:* | atarist[e]:*TOS:*:*) echo m68k-atari-mint${UNAME_RELEASE} - exit 0 ;; + exit ;; *falcon*:*MiNT:*:* | *falcon*:*mint:*:* | *falcon*:*TOS:*:*) - echo m68k-atari-mint${UNAME_RELEASE} - exit 0 ;; + echo m68k-atari-mint${UNAME_RELEASE} + exit ;; milan*:*MiNT:*:* | milan*:*mint:*:* | *milan*:*TOS:*:*) - echo m68k-milan-mint${UNAME_RELEASE} - exit 0 ;; + echo m68k-milan-mint${UNAME_RELEASE} + exit ;; hades*:*MiNT:*:* | hades*:*mint:*:* | *hades*:*TOS:*:*) - echo m68k-hades-mint${UNAME_RELEASE} - exit 0 ;; + echo m68k-hades-mint${UNAME_RELEASE} + exit ;; *:*MiNT:*:* | *:*mint:*:* | *:*TOS:*:*) - echo m68k-unknown-mint${UNAME_RELEASE} - exit 0 ;; + echo m68k-unknown-mint${UNAME_RELEASE} + exit ;; m68k:machten:*:*) echo m68k-apple-machten${UNAME_RELEASE} - exit 0 ;; + exit ;; powerpc:machten:*:*) echo powerpc-apple-machten${UNAME_RELEASE} - exit 0 ;; + exit ;; RISC*:Mach:*:*) echo mips-dec-mach_bsd4.3 - exit 0 ;; + exit ;; RISC*:ULTRIX:*:*) echo mips-dec-ultrix${UNAME_RELEASE} - exit 0 ;; + exit ;; VAX*:ULTRIX*:*:*) echo vax-dec-ultrix${UNAME_RELEASE} - exit 0 ;; + exit ;; 2020:CLIX:*:* | 2430:CLIX:*:*) echo clipper-intergraph-clix${UNAME_RELEASE} - exit 0 ;; + exit ;; mips:*:*:UMIPS | mips:*:*:RISCos) eval $set_cc_for_build sed 's/^ //' << EOF >$dummy.c @@ -467,35 +454,36 @@ case "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}" in exit (-1); } EOF - $CC_FOR_BUILD -o $dummy $dummy.c \ - && $dummy `echo "${UNAME_RELEASE}" | sed -n 's/\([0-9]*\).*/\1/p'` \ - && exit 0 + $CC_FOR_BUILD -o $dummy $dummy.c && + dummyarg=`echo "${UNAME_RELEASE}" | sed -n 's/\([0-9]*\).*/\1/p'` && + SYSTEM_NAME=`$dummy $dummyarg` && + { echo "$SYSTEM_NAME"; exit; } echo mips-mips-riscos${UNAME_RELEASE} - exit 0 ;; + exit ;; Motorola:PowerMAX_OS:*:*) echo powerpc-motorola-powermax - exit 0 ;; + exit ;; Motorola:*:4.3:PL8-*) echo powerpc-harris-powermax - exit 0 ;; + exit ;; Night_Hawk:*:*:PowerMAX_OS | Synergy:PowerMAX_OS:*:*) echo powerpc-harris-powermax - exit 0 ;; + exit ;; Night_Hawk:Power_UNIX:*:*) echo powerpc-harris-powerunix - exit 0 ;; + exit ;; m88k:CX/UX:7*:*) echo m88k-harris-cxux7 - exit 0 ;; + exit ;; m88k:*:4*:R4*) echo m88k-motorola-sysv4 - exit 0 ;; + exit ;; m88k:*:3*:R3*) echo m88k-motorola-sysv3 - exit 0 ;; + exit ;; AViiON:dgux:*:*) - # DG/UX returns AViiON for all architectures - UNAME_PROCESSOR=`/usr/bin/uname -p` + # DG/UX returns AViiON for all architectures + UNAME_PROCESSOR=`/usr/bin/uname -p` if [ $UNAME_PROCESSOR = mc88100 ] || [ $UNAME_PROCESSOR = mc88110 ] then if [ ${TARGET_BINARY_INTERFACE}x = m88kdguxelfx ] || \ @@ -508,29 +496,29 @@ EOF else echo i586-dg-dgux${UNAME_RELEASE} fi - exit 0 ;; + exit ;; M88*:DolphinOS:*:*) # DolphinOS (SVR3) echo m88k-dolphin-sysv3 - exit 0 ;; + exit ;; M88*:*:R3*:*) # Delta 88k system running SVR3 echo m88k-motorola-sysv3 - exit 0 ;; + exit ;; XD88*:*:*:*) # Tektronix XD88 system running UTekV (SVR3) echo m88k-tektronix-sysv3 - exit 0 ;; + exit ;; Tek43[0-9][0-9]:UTek:*:*) # Tektronix 4300 system running UTek (BSD) echo m68k-tektronix-bsd - exit 0 ;; + exit ;; *:IRIX*:*:*) echo mips-sgi-irix`echo ${UNAME_RELEASE}|sed -e 's/-/_/g'` - exit 0 ;; + exit ;; ????????:AIX?:[12].1:2) # AIX 2.2.1 or AIX 2.1.1 is RT/PC AIX. - echo romp-ibm-aix # uname -m gives an 8 hex-code CPU id - exit 0 ;; # Note that: echo "'`uname -s`'" gives 'AIX ' + echo romp-ibm-aix # uname -m gives an 8 hex-code CPU id + exit ;; # Note that: echo "'`uname -s`'" gives 'AIX ' i*86:AIX:*:*) echo i386-ibm-aix - exit 0 ;; + exit ;; ia64:AIX:*:*) if [ -x /usr/bin/oslevel ] ; then IBM_REV=`/usr/bin/oslevel` @@ -538,7 +526,7 @@ EOF IBM_REV=${UNAME_VERSION}.${UNAME_RELEASE} fi echo ${UNAME_MACHINE}-ibm-aix${IBM_REV} - exit 0 ;; + exit ;; *:AIX:2:3) if grep bos325 /usr/include/stdio.h >/dev/null 2>&1; then eval $set_cc_for_build @@ -553,15 +541,19 @@ EOF exit(0); } EOF - $CC_FOR_BUILD -o $dummy $dummy.c && $dummy && exit 0 - echo rs6000-ibm-aix3.2.5 + if $CC_FOR_BUILD -o $dummy $dummy.c && SYSTEM_NAME=`$dummy` + then + echo "$SYSTEM_NAME" + else + echo rs6000-ibm-aix3.2.5 + fi elif grep bos324 /usr/include/stdio.h >/dev/null 2>&1; then echo rs6000-ibm-aix3.2.4 else echo rs6000-ibm-aix3.2 fi - exit 0 ;; - *:AIX:*:[45]) + exit ;; + *:AIX:*:[4567]) IBM_CPU_ID=`/usr/sbin/lsdev -C -c processor -S available | sed 1q | awk '{ print $1 }'` if /usr/sbin/lsattr -El ${IBM_CPU_ID} | grep ' POWER' >/dev/null 2>&1; then IBM_ARCH=rs6000 @@ -574,28 +566,28 @@ EOF IBM_REV=${UNAME_VERSION}.${UNAME_RELEASE} fi echo ${IBM_ARCH}-ibm-aix${IBM_REV} - exit 0 ;; + exit ;; *:AIX:*:*) echo rs6000-ibm-aix - exit 0 ;; + exit ;; ibmrt:4.4BSD:*|romp-ibm:BSD:*) echo romp-ibm-bsd4.4 - exit 0 ;; + exit ;; ibmrt:*BSD:*|romp-ibm:BSD:*) # covers RT/PC BSD and echo romp-ibm-bsd${UNAME_RELEASE} # 4.3 with uname added to - exit 0 ;; # report: romp-ibm BSD 4.3 + exit ;; # report: romp-ibm BSD 4.3 *:BOSX:*:*) echo rs6000-bull-bosx - exit 0 ;; + exit ;; DPX/2?00:B.O.S.:*:*) echo m68k-bull-sysv3 - exit 0 ;; + exit ;; 9000/[34]??:4.3bsd:1.*:*) echo m68k-hp-bsd - exit 0 ;; + exit ;; hp300:4.4BSD:*:* | 9000/[34]??:4.3bsd:2.*:*) echo m68k-hp-bsd4.4 - exit 0 ;; + exit ;; 9000/[34678]??:HP-UX:*:*) HPUX_REV=`echo ${UNAME_RELEASE}|sed -e 's/[^.]*.[0B]*//'` case "${UNAME_MACHINE}" in @@ -604,52 +596,52 @@ EOF 9000/[678][0-9][0-9]) if [ -x /usr/bin/getconf ]; then sc_cpu_version=`/usr/bin/getconf SC_CPU_VERSION 2>/dev/null` - sc_kernel_bits=`/usr/bin/getconf SC_KERNEL_BITS 2>/dev/null` - case "${sc_cpu_version}" in - 523) HP_ARCH="hppa1.0" ;; # CPU_PA_RISC1_0 - 528) HP_ARCH="hppa1.1" ;; # CPU_PA_RISC1_1 - 532) # CPU_PA_RISC2_0 - case "${sc_kernel_bits}" in - 32) HP_ARCH="hppa2.0n" ;; - 64) HP_ARCH="hppa2.0w" ;; + sc_kernel_bits=`/usr/bin/getconf SC_KERNEL_BITS 2>/dev/null` + case "${sc_cpu_version}" in + 523) HP_ARCH="hppa1.0" ;; # CPU_PA_RISC1_0 + 528) HP_ARCH="hppa1.1" ;; # CPU_PA_RISC1_1 + 532) # CPU_PA_RISC2_0 + case "${sc_kernel_bits}" in + 32) HP_ARCH="hppa2.0n" ;; + 64) HP_ARCH="hppa2.0w" ;; '') HP_ARCH="hppa2.0" ;; # HP-UX 10.20 - esac ;; - esac + esac ;; + esac fi if [ "${HP_ARCH}" = "" ]; then eval $set_cc_for_build - sed 's/^ //' << EOF >$dummy.c + sed 's/^ //' << EOF >$dummy.c - #define _HPUX_SOURCE - #include - #include + #define _HPUX_SOURCE + #include + #include - int main () - { - #if defined(_SC_KERNEL_BITS) - long bits = sysconf(_SC_KERNEL_BITS); - #endif - long cpu = sysconf (_SC_CPU_VERSION); + int main () + { + #if defined(_SC_KERNEL_BITS) + long bits = sysconf(_SC_KERNEL_BITS); + #endif + long cpu = sysconf (_SC_CPU_VERSION); - switch (cpu) - { - case CPU_PA_RISC1_0: puts ("hppa1.0"); break; - case CPU_PA_RISC1_1: puts ("hppa1.1"); break; - case CPU_PA_RISC2_0: - #if defined(_SC_KERNEL_BITS) - switch (bits) - { - case 64: puts ("hppa2.0w"); break; - case 32: puts ("hppa2.0n"); break; - default: puts ("hppa2.0"); break; - } break; - #else /* !defined(_SC_KERNEL_BITS) */ - puts ("hppa2.0"); break; - #endif - default: puts ("hppa1.0"); break; - } - exit (0); - } + switch (cpu) + { + case CPU_PA_RISC1_0: puts ("hppa1.0"); break; + case CPU_PA_RISC1_1: puts ("hppa1.1"); break; + case CPU_PA_RISC2_0: + #if defined(_SC_KERNEL_BITS) + switch (bits) + { + case 64: puts ("hppa2.0w"); break; + case 32: puts ("hppa2.0n"); break; + default: puts ("hppa2.0"); break; + } break; + #else /* !defined(_SC_KERNEL_BITS) */ + puts ("hppa2.0"); break; + #endif + default: puts ("hppa1.0"); break; + } + exit (0); + } EOF (CCOPTS= $CC_FOR_BUILD -o $dummy $dummy.c 2>/dev/null) && HP_ARCH=`$dummy` test -z "$HP_ARCH" && HP_ARCH=hppa @@ -657,9 +649,19 @@ EOF esac if [ ${HP_ARCH} = "hppa2.0w" ] then - # avoid double evaluation of $set_cc_for_build - test -n "$CC_FOR_BUILD" || eval $set_cc_for_build - if echo __LP64__ | (CCOPTS= $CC_FOR_BUILD -E -) | grep __LP64__ >/dev/null + eval $set_cc_for_build + + # hppa2.0w-hp-hpux* has a 64-bit kernel and a compiler generating + # 32-bit code. hppa64-hp-hpux* has the same kernel and a compiler + # generating 64-bit code. GNU and HP use different nomenclature: + # + # $ CC_FOR_BUILD=cc ./config.guess + # => hppa2.0w-hp-hpux11.23 + # $ CC_FOR_BUILD="cc +DA2.0w" ./config.guess + # => hppa64-hp-hpux11.23 + + if echo __LP64__ | (CCOPTS= $CC_FOR_BUILD -E - 2>/dev/null) | + grep -q __LP64__ then HP_ARCH="hppa2.0w" else @@ -667,11 +669,11 @@ EOF fi fi echo ${HP_ARCH}-hp-hpux${HPUX_REV} - exit 0 ;; + exit ;; ia64:HP-UX:*:*) HPUX_REV=`echo ${UNAME_RELEASE}|sed -e 's/[^.]*.[0B]*//'` echo ia64-hp-hpux${HPUX_REV} - exit 0 ;; + exit ;; 3050*:HI-UX:*:*) eval $set_cc_for_build sed 's/^ //' << EOF >$dummy.c @@ -699,207 +701,173 @@ EOF exit (0); } EOF - $CC_FOR_BUILD -o $dummy $dummy.c && $dummy && exit 0 + $CC_FOR_BUILD -o $dummy $dummy.c && SYSTEM_NAME=`$dummy` && + { echo "$SYSTEM_NAME"; exit; } echo unknown-hitachi-hiuxwe2 - exit 0 ;; + exit ;; 9000/7??:4.3bsd:*:* | 9000/8?[79]:4.3bsd:*:* ) echo hppa1.1-hp-bsd - exit 0 ;; + exit ;; 9000/8??:4.3bsd:*:*) echo hppa1.0-hp-bsd - exit 0 ;; + exit ;; *9??*:MPE/iX:*:* | *3000*:MPE/iX:*:*) echo hppa1.0-hp-mpeix - exit 0 ;; + exit ;; hp7??:OSF1:*:* | hp8?[79]:OSF1:*:* ) echo hppa1.1-hp-osf - exit 0 ;; + exit ;; hp8??:OSF1:*:*) echo hppa1.0-hp-osf - exit 0 ;; + exit ;; i*86:OSF1:*:*) if [ -x /usr/sbin/sysversion ] ; then echo ${UNAME_MACHINE}-unknown-osf1mk else echo ${UNAME_MACHINE}-unknown-osf1 fi - exit 0 ;; + exit ;; parisc*:Lites*:*:*) echo hppa1.1-hp-lites - exit 0 ;; + exit ;; C1*:ConvexOS:*:* | convex:ConvexOS:C1*:*) echo c1-convex-bsd - exit 0 ;; + exit ;; C2*:ConvexOS:*:* | convex:ConvexOS:C2*:*) if getsysinfo -f scalar_acc then echo c32-convex-bsd else echo c2-convex-bsd fi - exit 0 ;; + exit ;; C34*:ConvexOS:*:* | convex:ConvexOS:C34*:*) echo c34-convex-bsd - exit 0 ;; + exit ;; C38*:ConvexOS:*:* | convex:ConvexOS:C38*:*) echo c38-convex-bsd - exit 0 ;; + exit ;; C4*:ConvexOS:*:* | convex:ConvexOS:C4*:*) echo c4-convex-bsd - exit 0 ;; + exit ;; CRAY*Y-MP:*:*:*) echo ymp-cray-unicos${UNAME_RELEASE} | sed -e 's/\.[^.]*$/.X/' - exit 0 ;; + exit ;; CRAY*[A-Z]90:*:*:*) echo ${UNAME_MACHINE}-cray-unicos${UNAME_RELEASE} \ | sed -e 's/CRAY.*\([A-Z]90\)/\1/' \ -e y/ABCDEFGHIJKLMNOPQRSTUVWXYZ/abcdefghijklmnopqrstuvwxyz/ \ -e 's/\.[^.]*$/.X/' - exit 0 ;; + exit ;; CRAY*TS:*:*:*) echo t90-cray-unicos${UNAME_RELEASE} | sed -e 's/\.[^.]*$/.X/' - exit 0 ;; + exit ;; CRAY*T3E:*:*:*) echo alphaev5-cray-unicosmk${UNAME_RELEASE} | sed -e 's/\.[^.]*$/.X/' - exit 0 ;; + exit ;; CRAY*SV1:*:*:*) echo sv1-cray-unicos${UNAME_RELEASE} | sed -e 's/\.[^.]*$/.X/' - exit 0 ;; + exit ;; *:UNICOS/mp:*:*) - echo nv1-cray-unicosmp${UNAME_RELEASE} | sed -e 's/\.[^.]*$/.X/' - exit 0 ;; + echo craynv-cray-unicosmp${UNAME_RELEASE} | sed -e 's/\.[^.]*$/.X/' + exit ;; F30[01]:UNIX_System_V:*:* | F700:UNIX_System_V:*:*) FUJITSU_PROC=`uname -m | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz'` - FUJITSU_SYS=`uname -p | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz' | sed -e 's/\///'` - FUJITSU_REL=`echo ${UNAME_RELEASE} | sed -e 's/ /_/'` - echo "${FUJITSU_PROC}-fujitsu-${FUJITSU_SYS}${FUJITSU_REL}" - exit 0 ;; + FUJITSU_SYS=`uname -p | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz' | sed -e 's/\///'` + FUJITSU_REL=`echo ${UNAME_RELEASE} | sed -e 's/ /_/'` + echo "${FUJITSU_PROC}-fujitsu-${FUJITSU_SYS}${FUJITSU_REL}" + exit ;; 5000:UNIX_System_V:4.*:*) - FUJITSU_SYS=`uname -p | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz' | sed -e 's/\///'` - FUJITSU_REL=`echo ${UNAME_RELEASE} | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz' | sed -e 's/ /_/'` - echo "sparc-fujitsu-${FUJITSU_SYS}${FUJITSU_REL}" - exit 0 ;; + FUJITSU_SYS=`uname -p | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz' | sed -e 's/\///'` + FUJITSU_REL=`echo ${UNAME_RELEASE} | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz' | sed -e 's/ /_/'` + echo "sparc-fujitsu-${FUJITSU_SYS}${FUJITSU_REL}" + exit ;; i*86:BSD/386:*:* | i*86:BSD/OS:*:* | *:Ascend\ Embedded/OS:*:*) echo ${UNAME_MACHINE}-pc-bsdi${UNAME_RELEASE} - exit 0 ;; + exit ;; sparc*:BSD/OS:*:*) echo sparc-unknown-bsdi${UNAME_RELEASE} - exit 0 ;; + exit ;; *:BSD/OS:*:*) echo ${UNAME_MACHINE}-unknown-bsdi${UNAME_RELEASE} - exit 0 ;; + exit ;; *:FreeBSD:*:*) - # Determine whether the default compiler uses glibc. - eval $set_cc_for_build - sed 's/^ //' << EOF >$dummy.c - #include - #if __GLIBC__ >= 2 - LIBC=gnu - #else - LIBC= - #endif -EOF - eval `$CC_FOR_BUILD -E $dummy.c 2>/dev/null | grep ^LIBC=` - # GNU/KFreeBSD systems have a "k" prefix to indicate we are using - # FreeBSD's kernel, but not the complete OS. - case ${LIBC} in gnu) kernel_only='k' ;; esac - echo ${UNAME_MACHINE}-unknown-${kernel_only}freebsd`echo ${UNAME_RELEASE}|sed -e 's/[-(].*//'`${LIBC:+-$LIBC} - exit 0 ;; + UNAME_PROCESSOR=`/usr/bin/uname -p` + case ${UNAME_PROCESSOR} in + amd64) + echo x86_64-unknown-freebsd`echo ${UNAME_RELEASE}|sed -e 's/[-(].*//'` ;; + *) + echo ${UNAME_PROCESSOR}-unknown-freebsd`echo ${UNAME_RELEASE}|sed -e 's/[-(].*//'` ;; + esac + exit ;; i*:CYGWIN*:*) echo ${UNAME_MACHINE}-pc-cygwin - exit 0 ;; - i*:MINGW*:*) + exit ;; + *:MINGW*:*) echo ${UNAME_MACHINE}-pc-mingw32 - exit 0 ;; + exit ;; + i*:MSYS*:*) + echo ${UNAME_MACHINE}-pc-msys + exit ;; + i*:windows32*:*) + # uname -m includes "-pc" on this system. + echo ${UNAME_MACHINE}-mingw32 + exit ;; i*:PW*:*) echo ${UNAME_MACHINE}-pc-pw32 - exit 0 ;; - x86:Interix*:[34]*) - echo i586-pc-interix${UNAME_RELEASE}|sed -e 's/\..*//' - exit 0 ;; + exit ;; + *:Interix*:*) + case ${UNAME_MACHINE} in + x86) + echo i586-pc-interix${UNAME_RELEASE} + exit ;; + authenticamd | genuineintel | EM64T) + echo x86_64-unknown-interix${UNAME_RELEASE} + exit ;; + IA64) + echo ia64-unknown-interix${UNAME_RELEASE} + exit ;; + esac ;; [345]86:Windows_95:* | [345]86:Windows_98:* | [345]86:Windows_NT:*) echo i${UNAME_MACHINE}-pc-mks - exit 0 ;; + exit ;; + 8664:Windows_NT:*) + echo x86_64-pc-mks + exit ;; i*:Windows_NT*:* | Pentium*:Windows_NT*:*) # How do we know it's Interix rather than the generic POSIX subsystem? # It also conflicts with pre-2.0 versions of AT&T UWIN. Should we # UNAME_MACHINE based on the output of uname instead of i386? echo i586-pc-interix - exit 0 ;; + exit ;; i*:UWIN*:*) echo ${UNAME_MACHINE}-pc-uwin - exit 0 ;; + exit ;; + amd64:CYGWIN*:*:* | x86_64:CYGWIN*:*:*) + echo x86_64-unknown-cygwin + exit ;; p*:CYGWIN*:*) echo powerpcle-unknown-cygwin - exit 0 ;; + exit ;; prep*:SunOS:5.*:*) echo powerpcle-unknown-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'` - exit 0 ;; + exit ;; *:GNU:*:*) # the GNU system echo `echo ${UNAME_MACHINE}|sed -e 's,[-/].*$,,'`-unknown-gnu`echo ${UNAME_RELEASE}|sed -e 's,/.*$,,'` - exit 0 ;; + exit ;; *:GNU/*:*:*) # other systems with GNU libc and userland echo ${UNAME_MACHINE}-unknown-`echo ${UNAME_SYSTEM} | sed 's,^[^/]*/,,' | tr '[A-Z]' '[a-z]'``echo ${UNAME_RELEASE}|sed -e 's/[-(].*//'`-gnu - exit 0 ;; + exit ;; i*86:Minix:*:*) echo ${UNAME_MACHINE}-pc-minix - exit 0 ;; - arm*:Linux:*:*) + exit ;; + aarch64:Linux:*:*) echo ${UNAME_MACHINE}-unknown-linux-gnu - exit 0 ;; - cris:Linux:*:*) - echo cris-axis-linux-gnu - exit 0 ;; - ia64:Linux:*:*) + exit ;; + aarch64_be:Linux:*:*) + UNAME_MACHINE=aarch64_be echo ${UNAME_MACHINE}-unknown-linux-gnu - exit 0 ;; - m68*:Linux:*:*) - echo ${UNAME_MACHINE}-unknown-linux-gnu - exit 0 ;; - mips:Linux:*:*) - eval $set_cc_for_build - sed 's/^ //' << EOF >$dummy.c - #undef CPU - #undef mips - #undef mipsel - #if defined(__MIPSEL__) || defined(__MIPSEL) || defined(_MIPSEL) || defined(MIPSEL) - CPU=mipsel - #else - #if defined(__MIPSEB__) || defined(__MIPSEB) || defined(_MIPSEB) || defined(MIPSEB) - CPU=mips - #else - CPU= - #endif - #endif -EOF - eval `$CC_FOR_BUILD -E $dummy.c 2>/dev/null | grep ^CPU=` - test x"${CPU}" != x && echo "${CPU}-unknown-linux-gnu" && exit 0 - ;; - mips64:Linux:*:*) - eval $set_cc_for_build - sed 's/^ //' << EOF >$dummy.c - #undef CPU - #undef mips64 - #undef mips64el - #if defined(__MIPSEL__) || defined(__MIPSEL) || defined(_MIPSEL) || defined(MIPSEL) - CPU=mips64el - #else - #if defined(__MIPSEB__) || defined(__MIPSEB) || defined(_MIPSEB) || defined(MIPSEB) - CPU=mips64 - #else - CPU= - #endif - #endif -EOF - eval `$CC_FOR_BUILD -E $dummy.c 2>/dev/null | grep ^CPU=` - test x"${CPU}" != x && echo "${CPU}-unknown-linux-gnu" && exit 0 - ;; - ppc:Linux:*:*) - echo powerpc-unknown-linux-gnu - exit 0 ;; - ppc64:Linux:*:*) - echo powerpc64-unknown-linux-gnu - exit 0 ;; + exit ;; alpha:Linux:*:*) case `sed -n '/^cpu model/s/^.*: \(.*\)/\1/p' < /proc/cpuinfo` in EV5) UNAME_MACHINE=alphaev5 ;; @@ -909,11 +877,90 @@ EOF EV6) UNAME_MACHINE=alphaev6 ;; EV67) UNAME_MACHINE=alphaev67 ;; EV68*) UNAME_MACHINE=alphaev68 ;; - esac - objdump --private-headers /bin/sh | grep ld.so.1 >/dev/null + esac + objdump --private-headers /bin/sh | grep -q ld.so.1 if test "$?" = 0 ; then LIBC="libc1" ; else LIBC="" ; fi echo ${UNAME_MACHINE}-unknown-linux-gnu${LIBC} - exit 0 ;; + exit ;; + arm*:Linux:*:*) + eval $set_cc_for_build + if echo __ARM_EABI__ | $CC_FOR_BUILD -E - 2>/dev/null \ + | grep -q __ARM_EABI__ + then + echo ${UNAME_MACHINE}-unknown-linux-gnu + else + if echo __ARM_PCS_VFP | $CC_FOR_BUILD -E - 2>/dev/null \ + | grep -q __ARM_PCS_VFP + then + echo ${UNAME_MACHINE}-unknown-linux-gnueabi + else + echo ${UNAME_MACHINE}-unknown-linux-gnueabihf + fi + fi + exit ;; + avr32*:Linux:*:*) + echo ${UNAME_MACHINE}-unknown-linux-gnu + exit ;; + cris:Linux:*:*) + echo ${UNAME_MACHINE}-axis-linux-gnu + exit ;; + crisv32:Linux:*:*) + echo ${UNAME_MACHINE}-axis-linux-gnu + exit ;; + frv:Linux:*:*) + echo ${UNAME_MACHINE}-unknown-linux-gnu + exit ;; + hexagon:Linux:*:*) + echo ${UNAME_MACHINE}-unknown-linux-gnu + exit ;; + i*86:Linux:*:*) + LIBC=gnu + eval $set_cc_for_build + sed 's/^ //' << EOF >$dummy.c + #ifdef __dietlibc__ + LIBC=dietlibc + #endif +EOF + eval `$CC_FOR_BUILD -E $dummy.c 2>/dev/null | grep '^LIBC'` + echo "${UNAME_MACHINE}-pc-linux-${LIBC}" + exit ;; + ia64:Linux:*:*) + echo ${UNAME_MACHINE}-unknown-linux-gnu + exit ;; + m32r*:Linux:*:*) + echo ${UNAME_MACHINE}-unknown-linux-gnu + exit ;; + m68*:Linux:*:*) + echo ${UNAME_MACHINE}-unknown-linux-gnu + exit ;; + mips:Linux:*:* | mips64:Linux:*:*) + eval $set_cc_for_build + sed 's/^ //' << EOF >$dummy.c + #undef CPU + #undef ${UNAME_MACHINE} + #undef ${UNAME_MACHINE}el + #if defined(__MIPSEL__) || defined(__MIPSEL) || defined(_MIPSEL) || defined(MIPSEL) + CPU=${UNAME_MACHINE}el + #else + #if defined(__MIPSEB__) || defined(__MIPSEB) || defined(_MIPSEB) || defined(MIPSEB) + CPU=${UNAME_MACHINE} + #else + CPU= + #endif + #endif +EOF + eval `$CC_FOR_BUILD -E $dummy.c 2>/dev/null | grep '^CPU'` + test x"${CPU}" != x && { echo "${CPU}-unknown-linux-gnu"; exit; } + ;; + or32:Linux:*:*) + echo ${UNAME_MACHINE}-unknown-linux-gnu + exit ;; + padre:Linux:*:*) + echo sparc-unknown-linux-gnu + exit ;; + parisc64:Linux:*:* | hppa64:Linux:*:*) + echo hppa64-unknown-linux-gnu + exit ;; parisc:Linux:*:* | hppa:Linux:*:*) # Look for CPU level case `grep '^cpu[^a-z]*:' /proc/cpuinfo 2>/dev/null | cut -d' ' -f2` in @@ -921,115 +968,71 @@ EOF PA8*) echo hppa2.0-unknown-linux-gnu ;; *) echo hppa-unknown-linux-gnu ;; esac - exit 0 ;; - parisc64:Linux:*:* | hppa64:Linux:*:*) - echo hppa64-unknown-linux-gnu - exit 0 ;; + exit ;; + ppc64:Linux:*:*) + echo powerpc64-unknown-linux-gnu + exit ;; + ppc:Linux:*:*) + echo powerpc-unknown-linux-gnu + exit ;; s390:Linux:*:* | s390x:Linux:*:*) echo ${UNAME_MACHINE}-ibm-linux - exit 0 ;; + exit ;; sh64*:Linux:*:*) - echo ${UNAME_MACHINE}-unknown-linux-gnu - exit 0 ;; + echo ${UNAME_MACHINE}-unknown-linux-gnu + exit ;; sh*:Linux:*:*) echo ${UNAME_MACHINE}-unknown-linux-gnu - exit 0 ;; + exit ;; sparc:Linux:*:* | sparc64:Linux:*:*) echo ${UNAME_MACHINE}-unknown-linux-gnu - exit 0 ;; + exit ;; + tile*:Linux:*:*) + echo ${UNAME_MACHINE}-unknown-linux-gnu + exit ;; + vax:Linux:*:*) + echo ${UNAME_MACHINE}-dec-linux-gnu + exit ;; x86_64:Linux:*:*) - echo x86_64-unknown-linux-gnu - exit 0 ;; - i*86:Linux:*:*) - # The BFD linker knows what the default object file format is, so - # first see if it will tell us. cd to the root directory to prevent - # problems with other programs or directories called `ld' in the path. - # Set LC_ALL=C to ensure ld outputs messages in English. - ld_supported_targets=`cd /; LC_ALL=C ld --help 2>&1 \ - | sed -ne '/supported targets:/!d - s/[ ][ ]*/ /g - s/.*supported targets: *// - s/ .*// - p'` - case "$ld_supported_targets" in - elf32-i386) - TENTATIVE="${UNAME_MACHINE}-pc-linux-gnu" - ;; - a.out-i386-linux) - echo "${UNAME_MACHINE}-pc-linux-gnuaout" - exit 0 ;; - coff-i386) - echo "${UNAME_MACHINE}-pc-linux-gnucoff" - exit 0 ;; - "") - # Either a pre-BFD a.out linker (linux-gnuoldld) or - # one that does not give us useful --help. - echo "${UNAME_MACHINE}-pc-linux-gnuoldld" - exit 0 ;; - esac - # Determine whether the default compiler is a.out or elf - eval $set_cc_for_build - sed 's/^ //' << EOF >$dummy.c - #include - #ifdef __ELF__ - # ifdef __GLIBC__ - # if __GLIBC__ >= 2 - LIBC=gnu - # else - LIBC=gnulibc1 - # endif - # else - LIBC=gnulibc1 - # endif - #else - #ifdef __INTEL_COMPILER - LIBC=gnu - #else - LIBC=gnuaout - #endif - #endif - #ifdef __dietlibc__ - LIBC=dietlibc - #endif -EOF - eval `$CC_FOR_BUILD -E $dummy.c 2>/dev/null | grep ^LIBC=` - test x"${LIBC}" != x && echo "${UNAME_MACHINE}-pc-linux-${LIBC}" && exit 0 - test x"${TENTATIVE}" != x && echo "${TENTATIVE}" && exit 0 - ;; + echo ${UNAME_MACHINE}-unknown-linux-gnu + exit ;; + xtensa*:Linux:*:*) + echo ${UNAME_MACHINE}-unknown-linux-gnu + exit ;; i*86:DYNIX/ptx:4*:*) # ptx 4.0 does uname -s correctly, with DYNIX/ptx in there. # earlier versions are messed up and put the nodename in both # sysname and nodename. echo i386-sequent-sysv4 - exit 0 ;; + exit ;; i*86:UNIX_SV:4.2MP:2.*) - # Unixware is an offshoot of SVR4, but it has its own version - # number series starting with 2... - # I am not positive that other SVR4 systems won't match this, + # Unixware is an offshoot of SVR4, but it has its own version + # number series starting with 2... + # I am not positive that other SVR4 systems won't match this, # I just have to hope. -- rms. - # Use sysv4.2uw... so that sysv4* matches it. + # Use sysv4.2uw... so that sysv4* matches it. echo ${UNAME_MACHINE}-pc-sysv4.2uw${UNAME_VERSION} - exit 0 ;; + exit ;; i*86:OS/2:*:*) # If we were able to find `uname', then EMX Unix compatibility # is probably installed. echo ${UNAME_MACHINE}-pc-os2-emx - exit 0 ;; + exit ;; i*86:XTS-300:*:STOP) echo ${UNAME_MACHINE}-unknown-stop - exit 0 ;; + exit ;; i*86:atheos:*:*) echo ${UNAME_MACHINE}-unknown-atheos - exit 0 ;; - i*86:syllable:*:*) + exit ;; + i*86:syllable:*:*) echo ${UNAME_MACHINE}-pc-syllable - exit 0 ;; - i*86:LynxOS:2.*:* | i*86:LynxOS:3.[01]*:* | i*86:LynxOS:4.0*:*) + exit ;; + i*86:LynxOS:2.*:* | i*86:LynxOS:3.[01]*:* | i*86:LynxOS:4.[02]*:*) echo i386-unknown-lynxos${UNAME_RELEASE} - exit 0 ;; + exit ;; i*86:*DOS:*:*) echo ${UNAME_MACHINE}-pc-msdosdjgpp - exit 0 ;; + exit ;; i*86:*:4.*:* | i*86:SYSTEM_V:4.*:*) UNAME_REL=`echo ${UNAME_RELEASE} | sed 's/\/MP$//'` if grep Novell /usr/include/link.h >/dev/null 2>/dev/null; then @@ -1037,15 +1040,16 @@ EOF else echo ${UNAME_MACHINE}-pc-sysv${UNAME_REL} fi - exit 0 ;; - i*86:*:5:[78]*) + exit ;; + i*86:*:5:[678]*) + # UnixWare 7.x, OpenUNIX and OpenServer 6. case `/bin/uname -X | grep "^Machine"` in *486*) UNAME_MACHINE=i486 ;; *Pentium) UNAME_MACHINE=i586 ;; *Pent*|*Celeron) UNAME_MACHINE=i686 ;; esac echo ${UNAME_MACHINE}-unknown-sysv${UNAME_RELEASE}${UNAME_SYSTEM}${UNAME_VERSION} - exit 0 ;; + exit ;; i*86:*:3.2:*) if test -f /usr/options/cb.name; then UNAME_REL=`sed -n 's/.*Version //p' /dev/null 2>&1 ; then echo i860-stardent-sysv${UNAME_RELEASE} # Stardent Vistra i860-SVR4 else # Add other i860-SVR4 vendors below as they are discovered. echo i860-unknown-sysv${UNAME_RELEASE} # Unknown i860-SVR4 fi - exit 0 ;; + exit ;; mini*:CTIX:SYS*5:*) # "miniframe" echo m68010-convergent-sysv - exit 0 ;; + exit ;; mc68k:UNIX:SYSTEM5:3.51m) echo m68k-convergent-sysv - exit 0 ;; + exit ;; M680?0:D-NIX:5.3:*) echo m68k-diab-dnix - exit 0 ;; - M68*:*:R3V[567]*:*) - test -r /sysV68 && echo 'm68k-motorola-sysv' && exit 0 ;; - 3[345]??:*:4.0:3.0 | 3[34]??A:*:4.0:3.0 | 3[34]??,*:*:4.0:3.0 | 3[34]??/*:*:4.0:3.0 | 4400:*:4.0:3.0 | 4850:*:4.0:3.0 | SKA40:*:4.0:3.0 | SDS2:*:4.0:3.0 | SHG2:*:4.0:3.0) + exit ;; + M68*:*:R3V[5678]*:*) + test -r /sysV68 && { echo 'm68k-motorola-sysv'; exit; } ;; + 3[345]??:*:4.0:3.0 | 3[34]??A:*:4.0:3.0 | 3[34]??,*:*:4.0:3.0 | 3[34]??/*:*:4.0:3.0 | 4400:*:4.0:3.0 | 4850:*:4.0:3.0 | SKA40:*:4.0:3.0 | SDS2:*:4.0:3.0 | SHG2:*:4.0:3.0 | S7501*:*:4.0:3.0) OS_REL='' test -r /etc/.relid \ && OS_REL=.`sed -n 's/[^ ]* [^ ]* \([0-9][0-9]\).*/\1/p' < /etc/.relid` /bin/uname -p 2>/dev/null | grep 86 >/dev/null \ - && echo i486-ncr-sysv4.3${OS_REL} && exit 0 + && { echo i486-ncr-sysv4.3${OS_REL}; exit; } /bin/uname -p 2>/dev/null | /bin/grep entium >/dev/null \ - && echo i586-ncr-sysv4.3${OS_REL} && exit 0 ;; + && { echo i586-ncr-sysv4.3${OS_REL}; exit; } ;; 3[34]??:*:4.0:* | 3[34]??,*:*:4.0:*) - /bin/uname -p 2>/dev/null | grep 86 >/dev/null \ - && echo i486-ncr-sysv4 && exit 0 ;; + /bin/uname -p 2>/dev/null | grep 86 >/dev/null \ + && { echo i486-ncr-sysv4; exit; } ;; + NCR*:*:4.2:* | MPRAS*:*:4.2:*) + OS_REL='.3' + test -r /etc/.relid \ + && OS_REL=.`sed -n 's/[^ ]* [^ ]* \([0-9][0-9]\).*/\1/p' < /etc/.relid` + /bin/uname -p 2>/dev/null | grep 86 >/dev/null \ + && { echo i486-ncr-sysv4.3${OS_REL}; exit; } + /bin/uname -p 2>/dev/null | /bin/grep entium >/dev/null \ + && { echo i586-ncr-sysv4.3${OS_REL}; exit; } + /bin/uname -p 2>/dev/null | /bin/grep pteron >/dev/null \ + && { echo i586-ncr-sysv4.3${OS_REL}; exit; } ;; m68*:LynxOS:2.*:* | m68*:LynxOS:3.0*:*) echo m68k-unknown-lynxos${UNAME_RELEASE} - exit 0 ;; + exit ;; mc68030:UNIX_System_V:4.*:*) echo m68k-atari-sysv4 - exit 0 ;; + exit ;; TSUNAMI:LynxOS:2.*:*) echo sparc-unknown-lynxos${UNAME_RELEASE} - exit 0 ;; + exit ;; rs6000:LynxOS:2.*:*) echo rs6000-unknown-lynxos${UNAME_RELEASE} - exit 0 ;; - PowerPC:LynxOS:2.*:* | PowerPC:LynxOS:3.[01]*:* | PowerPC:LynxOS:4.0*:*) + exit ;; + PowerPC:LynxOS:2.*:* | PowerPC:LynxOS:3.[01]*:* | PowerPC:LynxOS:4.[02]*:*) echo powerpc-unknown-lynxos${UNAME_RELEASE} - exit 0 ;; + exit ;; SM[BE]S:UNIX_SV:*:*) echo mips-dde-sysv${UNAME_RELEASE} - exit 0 ;; + exit ;; RM*:ReliantUNIX-*:*:*) echo mips-sni-sysv4 - exit 0 ;; + exit ;; RM*:SINIX-*:*:*) echo mips-sni-sysv4 - exit 0 ;; + exit ;; *:SINIX-*:*:*) if uname -p 2>/dev/null >/dev/null ; then UNAME_MACHINE=`(uname -p) 2>/dev/null` @@ -1137,68 +1154,94 @@ EOF else echo ns32k-sni-sysv fi - exit 0 ;; - PENTIUM:*:4.0*:*) # Unisys `ClearPath HMP IX 4000' SVR4/MP effort - # says - echo i586-unisys-sysv4 - exit 0 ;; + exit ;; + PENTIUM:*:4.0*:*) # Unisys `ClearPath HMP IX 4000' SVR4/MP effort + # says + echo i586-unisys-sysv4 + exit ;; *:UNIX_System_V:4*:FTX*) # From Gerald Hewes . # How about differentiating between stratus architectures? -djm echo hppa1.1-stratus-sysv4 - exit 0 ;; + exit ;; *:*:*:FTX*) # From seanf@swdc.stratus.com. echo i860-stratus-sysv4 - exit 0 ;; + exit ;; + i*86:VOS:*:*) + # From Paul.Green@stratus.com. + echo ${UNAME_MACHINE}-stratus-vos + exit ;; *:VOS:*:*) # From Paul.Green@stratus.com. echo hppa1.1-stratus-vos - exit 0 ;; + exit ;; mc68*:A/UX:*:*) echo m68k-apple-aux${UNAME_RELEASE} - exit 0 ;; + exit ;; news*:NEWS-OS:6*:*) echo mips-sony-newsos6 - exit 0 ;; + exit ;; R[34]000:*System_V*:*:* | R4000:UNIX_SYSV:*:* | R*000:UNIX_SV:*:*) if [ -d /usr/nec ]; then - echo mips-nec-sysv${UNAME_RELEASE} + echo mips-nec-sysv${UNAME_RELEASE} else - echo mips-unknown-sysv${UNAME_RELEASE} + echo mips-unknown-sysv${UNAME_RELEASE} fi - exit 0 ;; + exit ;; BeBox:BeOS:*:*) # BeOS running on hardware made by Be, PPC only. echo powerpc-be-beos - exit 0 ;; + exit ;; BeMac:BeOS:*:*) # BeOS running on Mac or Mac clone, PPC only. echo powerpc-apple-beos - exit 0 ;; + exit ;; BePC:BeOS:*:*) # BeOS running on Intel PC compatible. echo i586-pc-beos - exit 0 ;; + exit ;; + BePC:Haiku:*:*) # Haiku running on Intel PC compatible. + echo i586-pc-haiku + exit ;; SX-4:SUPER-UX:*:*) echo sx4-nec-superux${UNAME_RELEASE} - exit 0 ;; + exit ;; SX-5:SUPER-UX:*:*) echo sx5-nec-superux${UNAME_RELEASE} - exit 0 ;; + exit ;; SX-6:SUPER-UX:*:*) echo sx6-nec-superux${UNAME_RELEASE} - exit 0 ;; + exit ;; + SX-7:SUPER-UX:*:*) + echo sx7-nec-superux${UNAME_RELEASE} + exit ;; + SX-8:SUPER-UX:*:*) + echo sx8-nec-superux${UNAME_RELEASE} + exit ;; + SX-8R:SUPER-UX:*:*) + echo sx8r-nec-superux${UNAME_RELEASE} + exit ;; Power*:Rhapsody:*:*) echo powerpc-apple-rhapsody${UNAME_RELEASE} - exit 0 ;; + exit ;; *:Rhapsody:*:*) echo ${UNAME_MACHINE}-apple-rhapsody${UNAME_RELEASE} - exit 0 ;; + exit ;; *:Darwin:*:*) - case `uname -p` in - *86) UNAME_PROCESSOR=i686 ;; - powerpc) UNAME_PROCESSOR=powerpc ;; + UNAME_PROCESSOR=`uname -p` || UNAME_PROCESSOR=unknown + case $UNAME_PROCESSOR in + i386) + eval $set_cc_for_build + if [ "$CC_FOR_BUILD" != 'no_compiler_found' ]; then + if (echo '#ifdef __LP64__'; echo IS_64BIT_ARCH; echo '#endif') | \ + (CCOPTS= $CC_FOR_BUILD -E - 2>/dev/null) | \ + grep IS_64BIT_ARCH >/dev/null + then + UNAME_PROCESSOR="x86_64" + fi + fi ;; + unknown) UNAME_PROCESSOR=powerpc ;; esac echo ${UNAME_PROCESSOR}-apple-darwin${UNAME_RELEASE} - exit 0 ;; + exit ;; *:procnto*:*:* | *:QNX:[0123456789]*:*) UNAME_PROCESSOR=`uname -p` if test "$UNAME_PROCESSOR" = "x86"; then @@ -1206,22 +1249,28 @@ EOF UNAME_MACHINE=pc fi echo ${UNAME_PROCESSOR}-${UNAME_MACHINE}-nto-qnx${UNAME_RELEASE} - exit 0 ;; + exit ;; *:QNX:*:4*) echo i386-pc-qnx - exit 0 ;; + exit ;; + NEO-?:NONSTOP_KERNEL:*:*) + echo neo-tandem-nsk${UNAME_RELEASE} + exit ;; + NSE-?:NONSTOP_KERNEL:*:*) + echo nse-tandem-nsk${UNAME_RELEASE} + exit ;; NSR-?:NONSTOP_KERNEL:*:*) echo nsr-tandem-nsk${UNAME_RELEASE} - exit 0 ;; + exit ;; *:NonStop-UX:*:*) echo mips-compaq-nonstopux - exit 0 ;; + exit ;; BS2000:POSIX*:*:*) echo bs2000-siemens-sysv - exit 0 ;; + exit ;; DS/*:UNIX_System_V:*:*) echo ${UNAME_MACHINE}-${UNAME_SYSTEM}-${UNAME_RELEASE} - exit 0 ;; + exit ;; *:Plan9:*:*) # "uname -m" is not consistent, so use $cputype instead. 386 # is converted to i386 for consistency with other x86 @@ -1232,31 +1281,53 @@ EOF UNAME_MACHINE="$cputype" fi echo ${UNAME_MACHINE}-unknown-plan9 - exit 0 ;; + exit ;; *:TOPS-10:*:*) echo pdp10-unknown-tops10 - exit 0 ;; + exit ;; *:TENEX:*:*) echo pdp10-unknown-tenex - exit 0 ;; + exit ;; KS10:TOPS-20:*:* | KL10:TOPS-20:*:* | TYPE4:TOPS-20:*:*) echo pdp10-dec-tops20 - exit 0 ;; + exit ;; XKL-1:TOPS-20:*:* | TYPE5:TOPS-20:*:*) echo pdp10-xkl-tops20 - exit 0 ;; + exit ;; *:TOPS-20:*:*) echo pdp10-unknown-tops20 - exit 0 ;; + exit ;; *:ITS:*:*) echo pdp10-unknown-its - exit 0 ;; + exit ;; SEI:*:*:SEIUX) - echo mips-sei-seiux${UNAME_RELEASE} - exit 0 ;; + echo mips-sei-seiux${UNAME_RELEASE} + exit ;; *:DragonFly:*:*) echo ${UNAME_MACHINE}-unknown-dragonfly`echo ${UNAME_RELEASE}|sed -e 's/[-(].*//'` - exit 0 ;; + exit ;; + *:*VMS:*:*) + UNAME_MACHINE=`(uname -p) 2>/dev/null` + case "${UNAME_MACHINE}" in + A*) echo alpha-dec-vms ; exit ;; + I*) echo ia64-dec-vms ; exit ;; + V*) echo vax-dec-vms ; exit ;; + esac ;; + *:XENIX:*:SysV) + echo i386-pc-xenix + exit ;; + i*86:skyos:*:*) + echo ${UNAME_MACHINE}-pc-skyos`echo ${UNAME_RELEASE}` | sed -e 's/ .*$//' + exit ;; + i*86:rdos:*:*) + echo ${UNAME_MACHINE}-pc-rdos + exit ;; + i*86:AROS:*:*) + echo ${UNAME_MACHINE}-pc-aros + exit ;; + x86_64:VMkernel:*:*) + echo ${UNAME_MACHINE}-unknown-esx + exit ;; esac #echo '(No uname command or uname output not recognized.)' 1>&2 @@ -1279,16 +1350,16 @@ main () #include printf ("m68k-sony-newsos%s\n", #ifdef NEWSOS4 - "4" + "4" #else - "" + "" #endif - ); exit (0); + ); exit (0); #endif #endif #if defined (__arm) && defined (__acorn) && defined (__unix) - printf ("arm-acorn-riscix"); exit (0); + printf ("arm-acorn-riscix\n"); exit (0); #endif #if defined (hp300) && !defined (hpux) @@ -1377,11 +1448,12 @@ main () } EOF -$CC_FOR_BUILD -o $dummy $dummy.c 2>/dev/null && $dummy && exit 0 +$CC_FOR_BUILD -o $dummy $dummy.c 2>/dev/null && SYSTEM_NAME=`$dummy` && + { echo "$SYSTEM_NAME"; exit; } # Apollos put the system type in the environment. -test -d /usr/apollo && { echo ${ISP}-apollo-${SYSTYPE}; exit 0; } +test -d /usr/apollo && { echo ${ISP}-apollo-${SYSTYPE}; exit; } # Convex versions that predate uname can use getsysinfo(1) @@ -1390,22 +1462,22 @@ then case `getsysinfo -f cpu_type` in c1*) echo c1-convex-bsd - exit 0 ;; + exit ;; c2*) if getsysinfo -f scalar_acc then echo c32-convex-bsd else echo c2-convex-bsd fi - exit 0 ;; + exit ;; c34*) echo c34-convex-bsd - exit 0 ;; + exit ;; c38*) echo c38-convex-bsd - exit 0 ;; + exit ;; c4*) echo c4-convex-bsd - exit 0 ;; + exit ;; esac fi @@ -1416,7 +1488,9 @@ This script, last modified $timestamp, has failed to recognize the operating system you are using. It is advised that you download the most up to date version of the config scripts from - ftp://ftp.gnu.org/pub/gnu/config/ + http://git.savannah.gnu.org/gitweb/?p=config.git;a=blob_plain;f=config.guess;hb=HEAD +and + http://git.savannah.gnu.org/gitweb/?p=config.git;a=blob_plain;f=config.sub;hb=HEAD If the version you run ($0) is already up to date, please send the following data and any information you think might be diff --git a/config.sub b/config.sub index 264f820a..c894da45 100755 --- a/config.sub +++ b/config.sub @@ -1,9 +1,10 @@ #! /bin/sh # Configuration validation subroutine script. # Copyright (C) 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, -# 2000, 2001, 2002, 2003 Free Software Foundation, Inc. +# 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, +# 2011, 2012 Free Software Foundation, Inc. -timestamp='2004-02-23' +timestamp='2012-02-10' # This file is (in principle) common to ALL GNU software. # The presence of a machine in this file suggests that SOME GNU software @@ -20,23 +21,25 @@ timestamp='2004-02-23' # GNU General Public License for more details. # # You should have received a copy of the GNU General Public License -# along with this program; if not, write to the Free Software -# Foundation, Inc., 59 Temple Place - Suite 330, -# Boston, MA 02111-1307, USA. - +# along with this program; if not, see . +# # As a special exception to the GNU General Public License, if you # distribute this file as part of a program that contains a # configuration script generated by Autoconf, you may include it under # the same distribution terms that you use for the rest of that program. + # Please send patches to . Submit a context -# diff and a properly formatted ChangeLog entry. +# diff and a properly formatted GNU ChangeLog entry. # # Configuration subroutine to validate and canonicalize a configuration type. # Supply the specified configuration type as an argument. # If it is invalid, we print an error message on stderr and exit with code 1. # Otherwise, we print the canonical config type on stdout and succeed. +# You can get the latest version of this script from: +# http://git.savannah.gnu.org/gitweb/?p=config.git;a=blob_plain;f=config.sub;hb=HEAD + # This file is supposed to be the same for all GNU packages # and recognize all the CPU types, system types and aliases # that are meaningful with *any* GNU software. @@ -70,7 +73,8 @@ Report bugs and patches to ." version="\ GNU config.sub ($timestamp) -Copyright (C) 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2001 +Copyright (C) 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000, +2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012 Free Software Foundation, Inc. This is free software; see the source for copying conditions. There is NO @@ -83,11 +87,11 @@ Try \`$me --help' for more information." while test $# -gt 0 ; do case $1 in --time-stamp | --time* | -t ) - echo "$timestamp" ; exit 0 ;; + echo "$timestamp" ; exit ;; --version | -v ) - echo "$version" ; exit 0 ;; + echo "$version" ; exit ;; --help | --h* | -h ) - echo "$usage"; exit 0 ;; + echo "$usage"; exit ;; -- ) # Stop option processing shift; break ;; - ) # Use stdin as input. @@ -99,7 +103,7 @@ while test $# -gt 0 ; do *local*) # First pass through any local machine types. echo $1 - exit 0;; + exit ;; * ) break ;; @@ -118,11 +122,18 @@ esac # Here we must recognize all the valid KERNEL-OS combinations. maybe_os=`echo $1 | sed 's/^\(.*\)-\([^-]*-[^-]*\)$/\2/'` case $maybe_os in - nto-qnx* | linux-gnu* | linux-dietlibc | linux-uclibc* | uclinux-uclibc* | uclinux-gnu* | \ - kfreebsd*-gnu* | knetbsd*-gnu* | netbsd*-gnu* | storm-chaos* | os2-emx* | rtmk-nova*) + nto-qnx* | linux-gnu* | linux-android* | linux-dietlibc | linux-newlib* | \ + linux-uclibc* | uclinux-uclibc* | uclinux-gnu* | kfreebsd*-gnu* | \ + knetbsd*-gnu* | netbsd*-gnu* | \ + kopensolaris*-gnu* | \ + storm-chaos* | os2-emx* | rtmk-nova*) os=-$maybe_os basic_machine=`echo $1 | sed 's/^\(.*\)-\([^-]*-[^-]*\)$/\1/'` ;; + android-linux) + os=-linux-android + basic_machine=`echo $1 | sed 's/^\(.*\)-\([^-]*-[^-]*\)$/\1/'`-unknown + ;; *) basic_machine=`echo $1 | sed 's/-[^-]*$//'` if [ $basic_machine != $1 ] @@ -145,10 +156,13 @@ case $os in -convergent* | -ncr* | -news | -32* | -3600* | -3100* | -hitachi* |\ -c[123]* | -convex* | -sun | -crds | -omron* | -dg | -ultra | -tti* | \ -harris | -dolphin | -highlevel | -gould | -cbm | -ns | -masscomp | \ - -apple | -axis) + -apple | -axis | -knuth | -cray | -microblaze) os= basic_machine=$1 ;; + -bluegene*) + os=-cnk + ;; -sim | -cisco | -oki | -wec | -winbond) os= basic_machine=$1 @@ -163,13 +177,17 @@ case $os in os=-chorusos basic_machine=$1 ;; - -chorusrdb) - os=-chorusrdb + -chorusrdb) + os=-chorusrdb basic_machine=$1 - ;; + ;; -hiux*) os=-hiuxwe2 ;; + -sco6) + os=-sco5v6 + basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'` + ;; -sco5) os=-sco3.2v5 basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'` @@ -186,6 +204,10 @@ case $os in # Don't forget version if it is 3.2v4 or newer. basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'` ;; + -sco5v6*) + # Don't forget version if it is 3.2v4 or newer. + basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'` + ;; -sco*) os=-sco3.2v2 basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'` @@ -227,25 +249,36 @@ case $basic_machine in # Some are omitted here because they have special meanings below. 1750a | 580 \ | a29k \ + | aarch64 | aarch64_be \ | alpha | alphaev[4-8] | alphaev56 | alphaev6[78] | alphapca5[67] \ | alpha64 | alpha64ev[4-8] | alpha64ev56 | alpha64ev6[78] | alpha64pca5[67] \ | am33_2.0 \ - | arc | arm | arm[bl]e | arme[lb] | armv[2345] | armv[345][lb] | avr \ + | arc | arm | arm[bl]e | arme[lb] | armv[2345] | armv[345][lb] | avr | avr32 \ + | be32 | be64 \ + | bfin \ | c4x | clipper \ | d10v | d30v | dlx | dsp16xx \ - | fr30 | frv \ + | epiphany \ + | fido | fr30 | frv \ | h8300 | h8500 | hppa | hppa1.[01] | hppa2.0 | hppa2.0[nw] | hppa64 \ + | hexagon \ | i370 | i860 | i960 | ia64 \ | ip2k | iq2000 \ - | m32r | m68000 | m68k | m88k | mcore \ + | le32 | le64 \ + | lm32 \ + | m32c | m32r | m32rle | m68000 | m68k | m88k \ + | maxq | mb | microblaze | mcore | mep | metag \ | mips | mipsbe | mipseb | mipsel | mipsle \ | mips16 \ | mips64 | mips64el \ - | mips64vr | mips64vrel \ + | mips64octeon | mips64octeonel \ | mips64orion | mips64orionel \ + | mips64r5900 | mips64r5900el \ + | mips64vr | mips64vrel \ | mips64vr4100 | mips64vr4100el \ | mips64vr4300 | mips64vr4300el \ | mips64vr5000 | mips64vr5000el \ + | mips64vr5900 | mips64vr5900el \ | mipsisa32 | mipsisa32el \ | mipsisa32r2 | mipsisa32r2el \ | mipsisa64 | mipsisa64el \ @@ -254,30 +287,65 @@ case $basic_machine in | mipsisa64sr71k | mipsisa64sr71kel \ | mipstx39 | mipstx39el \ | mn10200 | mn10300 \ + | moxie \ + | mt \ | msp430 \ + | nds32 | nds32le | nds32be \ + | nios | nios2 \ | ns16k | ns32k \ - | openrisc | or32 \ + | open8 \ + | or32 \ | pdp10 | pdp11 | pj | pjl \ - | powerpc | powerpc64 | powerpc64le | powerpcle | ppcbe \ + | powerpc | powerpc64 | powerpc64le | powerpcle \ | pyramid \ - | sh | sh[1234] | sh[23]e | sh[34]eb | shbe | shle | sh[1234]le | sh3ele \ + | rl78 | rx \ + | score \ + | sh | sh[1234] | sh[24]a | sh[24]aeb | sh[23]e | sh[34]eb | sheb | shbe | shle | sh[1234]le | sh3ele \ | sh64 | sh64le \ - | sparc | sparc64 | sparc86x | sparclet | sparclite | sparcv9 | sparcv9b \ - | strongarm \ - | tahoe | thumb | tic4x | tic80 | tron \ - | v850 | v850e \ + | sparc | sparc64 | sparc64b | sparc64v | sparc86x | sparclet | sparclite \ + | sparcv8 | sparcv9 | sparcv9b | sparcv9v \ + | spu \ + | tahoe | tic4x | tic54x | tic55x | tic6x | tic80 | tron \ + | ubicom32 \ + | v850 | v850e | v850e1 | v850e2 | v850es | v850e2v3 \ | we32k \ - | x86 | xscale | xstormy16 | xtensa \ - | z8k) + | x86 | xc16x | xstormy16 | xtensa \ + | z8k | z80) basic_machine=$basic_machine-unknown ;; - m6811 | m68hc11 | m6812 | m68hc12) - # Motorola 68HC11/12. + c54x) + basic_machine=tic54x-unknown + ;; + c55x) + basic_machine=tic55x-unknown + ;; + c6x) + basic_machine=tic6x-unknown + ;; + m6811 | m68hc11 | m6812 | m68hc12 | m68hcs12x | picochip) basic_machine=$basic_machine-unknown os=-none ;; m88110 | m680[12346]0 | m683?2 | m68360 | m5200 | v70 | w65 | z8k) ;; + ms1) + basic_machine=mt-unknown + ;; + + strongarm | thumb | xscale) + basic_machine=arm-unknown + ;; + xgate) + basic_machine=$basic_machine-unknown + os=-none + ;; + xscaleeb) + basic_machine=armeb-unknown + ;; + + xscaleel) + basic_machine=armel-unknown + ;; # We use `pc' rather than `unknown' # because (1) that's what they normally are, and @@ -293,32 +361,40 @@ case $basic_machine in # Recognize the basic CPU types with company name. 580-* \ | a29k-* \ + | aarch64-* | aarch64_be-* \ | alpha-* | alphaev[4-8]-* | alphaev56-* | alphaev6[78]-* \ | alpha64-* | alpha64ev[4-8]-* | alpha64ev56-* | alpha64ev6[78]-* \ | alphapca5[67]-* | alpha64pca5[67]-* | arc-* \ | arm-* | armbe-* | armle-* | armeb-* | armv*-* \ - | avr-* \ - | bs2000-* \ - | c[123]* | c30-* | [cjt]90-* | c4x-* | c54x-* | c55x-* | c6x-* \ - | clipper-* | cydra-* \ + | avr-* | avr32-* \ + | be32-* | be64-* \ + | bfin-* | bs2000-* \ + | c[123]* | c30-* | [cjt]90-* | c4x-* \ + | clipper-* | craynv-* | cydra-* \ | d10v-* | d30v-* | dlx-* \ | elxsi-* \ - | f30[01]-* | f700-* | fr30-* | frv-* | fx80-* \ + | f30[01]-* | f700-* | fido-* | fr30-* | frv-* | fx80-* \ | h8300-* | h8500-* \ | hppa-* | hppa1.[01]-* | hppa2.0-* | hppa2.0[nw]-* | hppa64-* \ + | hexagon-* \ | i*86-* | i860-* | i960-* | ia64-* \ | ip2k-* | iq2000-* \ - | m32r-* \ + | le32-* | le64-* \ + | lm32-* \ + | m32c-* | m32r-* | m32rle-* \ | m68000-* | m680[012346]0-* | m68360-* | m683?2-* | m68k-* \ - | m88110-* | m88k-* | mcore-* \ + | m88110-* | m88k-* | maxq-* | mcore-* | metag-* | microblaze-* \ | mips-* | mipsbe-* | mipseb-* | mipsel-* | mipsle-* \ | mips16-* \ | mips64-* | mips64el-* \ - | mips64vr-* | mips64vrel-* \ + | mips64octeon-* | mips64octeonel-* \ | mips64orion-* | mips64orionel-* \ + | mips64r5900-* | mips64r5900el-* \ + | mips64vr-* | mips64vrel-* \ | mips64vr4100-* | mips64vr4100el-* \ | mips64vr4300-* | mips64vr4300el-* \ | mips64vr5000-* | mips64vr5000el-* \ + | mips64vr5900-* | mips64vr5900el-* \ | mipsisa32-* | mipsisa32el-* \ | mipsisa32r2-* | mipsisa32r2el-* \ | mipsisa64-* | mipsisa64el-* \ @@ -326,26 +402,39 @@ case $basic_machine in | mipsisa64sb1-* | mipsisa64sb1el-* \ | mipsisa64sr71k-* | mipsisa64sr71kel-* \ | mipstx39-* | mipstx39el-* \ + | mmix-* \ + | mt-* \ | msp430-* \ - | none-* | np1-* | nv1-* | ns16k-* | ns32k-* \ + | nds32-* | nds32le-* | nds32be-* \ + | nios-* | nios2-* \ + | none-* | np1-* | ns16k-* | ns32k-* \ + | open8-* \ | orion-* \ | pdp10-* | pdp11-* | pj-* | pjl-* | pn-* | power-* \ - | powerpc-* | powerpc64-* | powerpc64le-* | powerpcle-* | ppcbe-* \ + | powerpc-* | powerpc64-* | powerpc64le-* | powerpcle-* \ | pyramid-* \ - | romp-* | rs6000-* \ - | sh-* | sh[1234]-* | sh[23]e-* | sh[34]eb-* | shbe-* \ + | rl78-* | romp-* | rs6000-* | rx-* \ + | sh-* | sh[1234]-* | sh[24]a-* | sh[24]aeb-* | sh[23]e-* | sh[34]eb-* | sheb-* | shbe-* \ | shle-* | sh[1234]le-* | sh3ele-* | sh64-* | sh64le-* \ - | sparc-* | sparc64-* | sparc86x-* | sparclet-* | sparclite-* \ - | sparcv9-* | sparcv9b-* | strongarm-* | sv1-* | sx?-* \ - | tahoe-* | thumb-* \ + | sparc-* | sparc64-* | sparc64b-* | sparc64v-* | sparc86x-* | sparclet-* \ + | sparclite-* \ + | sparcv8-* | sparcv9-* | sparcv9b-* | sparcv9v-* | sv1-* | sx?-* \ + | tahoe-* \ | tic30-* | tic4x-* | tic54x-* | tic55x-* | tic6x-* | tic80-* \ + | tile*-* \ | tron-* \ - | v850-* | v850e-* | vax-* \ + | ubicom32-* \ + | v850-* | v850e-* | v850e1-* | v850es-* | v850e2-* | v850e2v3-* \ + | vax-* \ | we32k-* \ - | x86-* | x86_64-* | xps100-* | xscale-* | xstormy16-* \ - | xtensa-* \ + | x86-* | x86_64-* | xc16x-* | xps100-* \ + | xstormy16-* | xtensa*-* \ | ymp-* \ - | z8k-*) + | z8k-* | z80-*) + ;; + # Recognize the basic CPU types without company name, with glob match. + xtensa*) + basic_machine=$basic_machine-unknown ;; # Recognize the various machine names and aliases which stand # for a CPU type and a company and sometimes even an OS. @@ -363,7 +452,7 @@ case $basic_machine in basic_machine=a29k-amd os=-udi ;; - abacus) + abacus) basic_machine=abacus-unknown ;; adobe68k) @@ -409,6 +498,10 @@ case $basic_machine in basic_machine=m68k-apollo os=-bsd ;; + aros) + basic_machine=i386-pc + os=-aros + ;; aux) basic_machine=m68k-apple os=-aux @@ -417,10 +510,35 @@ case $basic_machine in basic_machine=ns32k-sequent os=-dynix ;; + blackfin) + basic_machine=bfin-unknown + os=-linux + ;; + blackfin-*) + basic_machine=bfin-`echo $basic_machine | sed 's/^[^-]*-//'` + os=-linux + ;; + bluegene*) + basic_machine=powerpc-ibm + os=-cnk + ;; + c54x-*) + basic_machine=tic54x-`echo $basic_machine | sed 's/^[^-]*-//'` + ;; + c55x-*) + basic_machine=tic55x-`echo $basic_machine | sed 's/^[^-]*-//'` + ;; + c6x-*) + basic_machine=tic6x-`echo $basic_machine | sed 's/^[^-]*-//'` + ;; c90) basic_machine=c90-cray os=-unicos ;; + cegcc) + basic_machine=arm-unknown + os=-cegcc + ;; convex-c1) basic_machine=c1-convex os=-bsd @@ -445,13 +563,20 @@ case $basic_machine in basic_machine=j90-cray os=-unicos ;; - cr16c) - basic_machine=cr16c-unknown + craynv) + basic_machine=craynv-cray + os=-unicosmp + ;; + cr16 | cr16-*) + basic_machine=cr16-unknown os=-elf ;; crds | unos) basic_machine=m68k-crds ;; + crisv32 | crisv32-* | etraxfs*) + basic_machine=crisv32-axis + ;; cris | cris-* | etrax*) basic_machine=cris-axis ;; @@ -481,6 +606,14 @@ case $basic_machine in basic_machine=m88k-motorola os=-sysv3 ;; + dicos) + basic_machine=i686-pc + os=-dicos + ;; + djgpp) + basic_machine=i586-pc + os=-msdosdjgpp + ;; dpx20 | dpx20-*) basic_machine=rs6000-bull os=-bosx @@ -592,7 +725,6 @@ case $basic_machine in i370-ibm* | ibm*) basic_machine=i370-ibm ;; -# I'm not sure what "Sysv32" means. Should this be sysv3.2? i*86v32) basic_machine=`echo $1 | sed -e 's/86.*/86-pc/'` os=-sysv32 @@ -631,6 +763,14 @@ case $basic_machine in basic_machine=m68k-isi os=-sysv ;; + m68knommu) + basic_machine=m68k-unknown + os=-linux + ;; + m68knommu-*) + basic_machine=m68k-`echo $basic_machine | sed 's/^[^-]*-//'` + os=-linux + ;; m88k-omron*) basic_machine=m88k-omron ;; @@ -642,10 +782,17 @@ case $basic_machine in basic_machine=ns32k-utek os=-sysv ;; + microblaze) + basic_machine=microblaze-xilinx + ;; mingw32) basic_machine=i386-pc os=-mingw32 ;; + mingw32ce) + basic_machine=arm-unknown + os=-mingw32ce + ;; miniframe) basic_machine=m68000-convergent ;; @@ -659,10 +806,6 @@ case $basic_machine in mips3*) basic_machine=`echo $basic_machine | sed -e 's/mips3/mips64/'`-unknown ;; - mmix*) - basic_machine=mmix-knuth - os=-mmixware - ;; monitor) basic_machine=m68k-rom68k os=-coff @@ -675,10 +818,21 @@ case $basic_machine in basic_machine=i386-pc os=-msdos ;; + ms1-*) + basic_machine=`echo $basic_machine | sed -e 's/ms1-/mt-/'` + ;; + msys) + basic_machine=i386-pc + os=-msys + ;; mvs) basic_machine=i370-ibm os=-mvs ;; + nacl) + basic_machine=le32-unknown + os=-nacl + ;; ncr3000) basic_machine=i486-ncr os=-sysv4 @@ -743,9 +897,11 @@ case $basic_machine in np1) basic_machine=np1-gould ;; - nv1) - basic_machine=nv1-cray - os=-unicosmp + neo-tandem) + basic_machine=neo-tandem + ;; + nse-tandem) + basic_machine=nse-tandem ;; nsr-tandem) basic_machine=nsr-tandem @@ -754,9 +910,8 @@ case $basic_machine in basic_machine=hppa1.1-oki os=-proelf ;; - or32 | or32-*) + openrisc | openrisc-*) basic_machine=or32-unknown - os=-coff ;; os400) basic_machine=powerpc-ibm @@ -778,6 +933,14 @@ case $basic_machine in basic_machine=i860-intel os=-osf ;; + parisc) + basic_machine=hppa-unknown + os=-linux + ;; + parisc-*) + basic_machine=hppa-`echo $basic_machine | sed 's/^[^-]*-//'` + os=-linux + ;; pbd) basic_machine=sparc-tti ;; @@ -787,6 +950,12 @@ case $basic_machine in pc532 | pc532-*) basic_machine=ns32k-pc532 ;; + pc98) + basic_machine=i386-pc + ;; + pc98-*) + basic_machine=i386-`echo $basic_machine | sed 's/^[^-]*-//'` + ;; pentium | p5 | k5 | k6 | nexgen | viac3) basic_machine=i586-pc ;; @@ -816,9 +985,10 @@ case $basic_machine in ;; power) basic_machine=power-ibm ;; - ppc) basic_machine=powerpc-unknown + ppc | ppcbe) basic_machine=powerpc-unknown ;; - ppc-*) basic_machine=powerpc-`echo $basic_machine | sed 's/^[^-]*-//'` + ppc-* | ppcbe-*) + basic_machine=powerpc-`echo $basic_machine | sed 's/^[^-]*-//'` ;; ppcle | powerpclittle | ppc-le | powerpc-little) basic_machine=powerpcle-unknown @@ -843,6 +1013,10 @@ case $basic_machine in basic_machine=i586-unknown os=-pw32 ;; + rdos) + basic_machine=i386-pc + os=-rdos + ;; rom68k) basic_machine=m68k-rom68k os=-coff @@ -869,6 +1043,10 @@ case $basic_machine in sb1el) basic_machine=mipsisa64sb1el-unknown ;; + sde) + basic_machine=mipsisa32-sde + os=-elf + ;; sei) basic_machine=mips-sei os=-seiux @@ -880,6 +1058,9 @@ case $basic_machine in basic_machine=sh-hitachi os=-hms ;; + sh5el) + basic_machine=sh5le-unknown + ;; sh64) basic_machine=sh64-unknown ;; @@ -901,6 +1082,9 @@ case $basic_machine in basic_machine=i860-stratus os=-sysv4 ;; + strongarm-* | thumb-*) + basic_machine=arm-`echo $basic_machine | sed 's/^[^-]*-//'` + ;; sun2) basic_machine=m68000-sun ;; @@ -957,17 +1141,9 @@ case $basic_machine in basic_machine=t90-cray os=-unicos ;; - tic54x | c54x*) - basic_machine=tic54x-unknown - os=-coff - ;; - tic55x | c55x*) - basic_machine=tic55x-unknown - os=-coff - ;; - tic6x | c6x*) - basic_machine=tic6x-unknown - os=-coff + tile*) + basic_machine=$basic_machine-unknown + os=-linux-gnu ;; tx39) basic_machine=mipstx39-unknown @@ -1029,9 +1205,16 @@ case $basic_machine in basic_machine=hppa1.1-winbond os=-proelf ;; + xbox) + basic_machine=i686-pc + os=-mingw32 + ;; xps | xps100) basic_machine=xps100-honeywell ;; + xscale-* | xscalee[bl]-*) + basic_machine=`echo $basic_machine | sed 's/^xscale/arm/'` + ;; ymp) basic_machine=ymp-cray os=-unicos @@ -1040,6 +1223,10 @@ case $basic_machine in basic_machine=z8k-unknown os=-sim ;; + z80-*-coff) + basic_machine=z80-unknown + os=-sim + ;; none) basic_machine=none-none os=-none @@ -1059,6 +1246,9 @@ case $basic_machine in romp) basic_machine=romp-ibm ;; + mmix) + basic_machine=mmix-knuth + ;; rs6000) basic_machine=rs6000-ibm ;; @@ -1075,13 +1265,10 @@ case $basic_machine in we32k) basic_machine=we32k-att ;; - sh3 | sh4 | sh[34]eb | sh[1234]le | sh[23]ele) + sh[1234] | sh[24]a | sh[24]aeb | sh[34]eb | sh[1234]le | sh[23]ele) basic_machine=sh-unknown ;; - sh64) - basic_machine=sh64-unknown - ;; - sparc | sparcv9 | sparcv9b) + sparc | sparcv8 | sparcv9 | sparcv9b | sparcv9v) basic_machine=sparc-sun ;; cydra) @@ -1125,9 +1312,12 @@ esac if [ x"$os" != x"" ] then case $os in - # First match some system type aliases - # that might get confused with valid system types. + # First match some system type aliases + # that might get confused with valid system types. # -solaris* is a basic system type, with this one exception. + -auroraux) + os=-auroraux + ;; -solaris1 | -solaris1.*) os=`echo $os | sed -e 's|solaris1|sunos4|'` ;; @@ -1148,26 +1338,31 @@ case $os in # Each alternative MUST END IN A *, to match a version number. # -sysv* is not here because it comes later, after sysvr4. -gnu* | -bsd* | -mach* | -minix* | -genix* | -ultrix* | -irix* \ - | -*vms* | -sco* | -esix* | -isc* | -aix* | -sunos | -sunos[34]*\ - | -hpux* | -unos* | -osf* | -luna* | -dgux* | -solaris* | -sym* \ + | -*vms* | -sco* | -esix* | -isc* | -aix* | -cnk* | -sunos | -sunos[34]*\ + | -hpux* | -unos* | -osf* | -luna* | -dgux* | -auroraux* | -solaris* \ + | -sym* | -kopensolaris* \ | -amigaos* | -amigados* | -msdos* | -newsos* | -unicos* | -aof* \ - | -aos* \ + | -aos* | -aros* \ | -nindy* | -vxsim* | -vxworks* | -ebmon* | -hms* | -mvs* \ | -clix* | -riscos* | -uniplus* | -iris* | -rtu* | -xenix* \ - | -hiux* | -386bsd* | -knetbsd* | -mirbsd* | -netbsd* | -openbsd* \ + | -hiux* | -386bsd* | -knetbsd* | -mirbsd* | -netbsd* \ + | -openbsd* | -solidbsd* \ | -ekkobsd* | -kfreebsd* | -freebsd* | -riscix* | -lynxos* \ | -bosx* | -nextstep* | -cxux* | -aout* | -elf* | -oabi* \ | -ptx* | -coff* | -ecoff* | -winnt* | -domain* | -vsta* \ | -udi* | -eabi* | -lites* | -ieee* | -go32* | -aux* \ - | -chorusos* | -chorusrdb* \ - | -cygwin* | -pe* | -psos* | -moss* | -proelf* | -rtems* \ - | -mingw32* | -linux-gnu* | -linux-uclibc* | -uxpv* | -beos* | -mpeix* | -udk* \ + | -chorusos* | -chorusrdb* | -cegcc* \ + | -cygwin* | -msys* | -pe* | -psos* | -moss* | -proelf* | -rtems* \ + | -mingw32* | -linux-gnu* | -linux-android* \ + | -linux-newlib* | -linux-uclibc* \ + | -uxpv* | -beos* | -mpeix* | -udk* \ | -interix* | -uwin* | -mks* | -rhapsody* | -darwin* | -opened* \ | -openstep* | -oskit* | -conix* | -pw32* | -nonstopux* \ | -storm-chaos* | -tops10* | -tenex* | -tops20* | -its* \ | -os2* | -vos* | -palmos* | -uclinux* | -nucleus* \ | -morphos* | -superux* | -rtmk* | -rtmk-nova* | -windiss* \ - | -powermax* | -dnix* | -nx6 | -nx7 | -sei* | -dragonfly*) + | -powermax* | -dnix* | -nx6 | -nx7 | -sei* | -dragonfly* \ + | -skyos* | -haiku* | -rdos* | -toppers* | -drops* | -es*) # Remember, each alternative MUST END IN *, to match a version number. ;; -qnx*) @@ -1185,7 +1380,7 @@ case $os in os=`echo $os | sed -e 's|nto|nto-qnx|'` ;; -sim | -es1800* | -hms* | -xray | -os68k* | -none* | -v88r* \ - | -windows* | -osx | -abug | -netware* | -os9* | -beos* \ + | -windows* | -osx | -abug | -netware* | -os9* | -beos* | -haiku* \ | -macos* | -mpw* | -magic* | -mmixware* | -mon960* | -lnews*) ;; -mac*) @@ -1206,7 +1401,7 @@ case $os in -opened*) os=-openedition ;; - -os400*) + -os400*) os=-os400 ;; -wince*) @@ -1255,7 +1450,7 @@ case $os in -sinix*) os=-sysv4 ;; - -tpf*) + -tpf*) os=-tpf ;; -triton*) @@ -1294,6 +1489,14 @@ case $os in -kaos*) os=-kaos ;; + -zvmoe) + os=-zvmoe + ;; + -dicos*) + os=-dicos + ;; + -nacl*) + ;; -none) ;; *) @@ -1316,6 +1519,12 @@ else # system, and we'll never get to this point. case $basic_machine in + score-*) + os=-elf + ;; + spu-*) + os=-elf + ;; *-acorn) os=-riscix1.2 ;; @@ -1325,9 +1534,18 @@ case $basic_machine in arm*-semi) os=-aout ;; - c4x-* | tic4x-*) - os=-coff - ;; + c4x-* | tic4x-*) + os=-coff + ;; + tic54x-*) + os=-coff + ;; + tic55x-*) + os=-coff + ;; + tic6x-*) + os=-coff + ;; # This must come before the *-dec entry. pdp10-*) os=-tops20 @@ -1346,13 +1564,13 @@ case $basic_machine in ;; m68000-sun) os=-sunos3 - # This also exists in the configure program, but was not the - # default. - # os=-sunos4 ;; m68*-cisco) os=-aout ;; + mep-*) + os=-elf + ;; mips*-cisco) os=-elf ;; @@ -1371,9 +1589,15 @@ case $basic_machine in *-be) os=-beos ;; + *-haiku) + os=-haiku + ;; *-ibm) os=-aix ;; + *-knuth) + os=-mmixware + ;; *-wec) os=-proelf ;; @@ -1476,7 +1700,7 @@ case $basic_machine in -sunos*) vendor=sun ;; - -aix*) + -cnk*|-aix*) vendor=ibm ;; -beos*) @@ -1539,7 +1763,7 @@ case $basic_machine in esac echo $basic_machine$os -exit 0 +exit # Local variables: # eval: (add-hook 'write-file-hooks 'time-stamp) diff --git a/configure.ac b/configure.ac index b58aa520..a72019e5 100644 --- a/configure.ac +++ b/configure.ac @@ -14,7 +14,7 @@ if test "x${CFLAGS}" = "x" ; then else CFLAGS="${CFLAGS} $1" fi -AC_RUN_IFELSE([AC_LANG_PROGRAM( +AC_COMPILE_IFELSE([AC_LANG_PROGRAM( [[ ]], [[ return 0; @@ -26,20 +26,25 @@ AC_RUN_IFELSE([AC_LANG_PROGRAM( ]) dnl JE_COMPILABLE(label, hcode, mcode, rvar) +dnl +dnl Use AC_LINK_IFELSE() rather than AC_COMPILE_IFELSE() so that linker errors +dnl cause failure. AC_DEFUN([JE_COMPILABLE], [ -AC_MSG_CHECKING([whether $1 is compilable]) -AC_RUN_IFELSE([AC_LANG_PROGRAM( -[$2], [$3])], - AC_MSG_RESULT([yes]) - [$4="yes"], - AC_MSG_RESULT([no]) - [$4="no"] -) +AC_CACHE_CHECK([whether $1 is compilable], + [$4], + [AC_LINK_IFELSE([AC_LANG_PROGRAM([$2], + [$3])], + [$4=yes], + [$4=no])]) ]) dnl ============================================================================ +dnl Library revision. +rev=1 +AC_SUBST([rev]) + srcroot=$srcdir if test "x${srcroot}" = "x." ; then srcroot="" @@ -82,14 +87,23 @@ AC_SUBST([MANDIR]) dnl Support for building documentation. AC_PATH_PROG([XSLTPROC], [xsltproc], , [$PATH]) +if test -d "/usr/share/xml/docbook/stylesheet/docbook-xsl" ; then + DEFAULT_XSLROOT="/usr/share/xml/docbook/stylesheet/docbook-xsl" +elif test -d "/usr/share/sgml/docbook/xsl-stylesheets" ; then + DEFAULT_XSLROOT="/usr/share/sgml/docbook/xsl-stylesheets" +else + dnl Documentation building will fail if this default gets used. + DEFAULT_XSLROOT="" +fi AC_ARG_WITH([xslroot], - [AS_HELP_STRING([--with-xslroot=], [XSL stylesheet root path])], + [AS_HELP_STRING([--with-xslroot=], [XSL stylesheet root path])], [ if test "x$with_xslroot" = "xno" ; then - XSLROOT="/usr/share/xml/docbook/stylesheet/docbook-xsl" + XSLROOT="${DEFAULT_XSLROOT}" else XSLROOT="${with_xslroot}" -fi, - XSLROOT="/usr/share/xml/docbook/stylesheet/docbook-xsl" +fi +], + XSLROOT="${DEFAULT_XSLROOT}" ) AC_SUBST([XSLROOT]) @@ -97,6 +111,19 @@ dnl If CFLAGS isn't defined, set CFLAGS to something reasonable. Otherwise, dnl just prevent autoconf from molesting CFLAGS. CFLAGS=$CFLAGS AC_PROG_CC +if test "x$GCC" != "xyes" ; then + AC_CACHE_CHECK([whether compiler is MSVC], + [je_cv_msvc], + [AC_COMPILE_IFELSE([AC_LANG_PROGRAM([], + [ +#ifndef _MSC_VER + int fail[-1]; +#endif +])], + [je_cv_msvc=yes], + [je_cv_msvc=no])]) +fi + if test "x$CFLAGS" = "x" ; then no_CFLAGS="yes" if test "x$GCC" = "xyes" ; then @@ -104,6 +131,12 @@ if test "x$CFLAGS" = "x" ; then JE_CFLAGS_APPEND([-Wall]) JE_CFLAGS_APPEND([-pipe]) JE_CFLAGS_APPEND([-g3]) + elif test "x$je_cv_msvc" = "xyes" ; then + CC="$CC -nologo" + JE_CFLAGS_APPEND([-Zi]) + JE_CFLAGS_APPEND([-MT]) + JE_CFLAGS_APPEND([-W3]) + CPPFLAGS="$CPPFLAGS -I${srcroot}/include/msvc_compat" fi fi dnl Append EXTRA_CFLAGS to CFLAGS, if defined. @@ -142,6 +175,18 @@ else fi AC_DEFINE_UNQUOTED([LG_SIZEOF_LONG], [$LG_SIZEOF_LONG]) +AC_CHECK_SIZEOF([intmax_t]) +if test "x${ac_cv_sizeof_intmax_t}" = "x16" ; then + LG_SIZEOF_INTMAX_T=4 +elif test "x${ac_cv_sizeof_intmax_t}" = "x8" ; then + LG_SIZEOF_INTMAX_T=3 +elif test "x${ac_cv_sizeof_intmax_t}" = "x4" ; then + LG_SIZEOF_INTMAX_T=2 +else + AC_MSG_ERROR([Unsupported intmax_t size: ${ac_cv_sizeof_intmax_t}]) +fi +AC_DEFINE_UNQUOTED([LG_SIZEOF_INTMAX_T], [$LG_SIZEOF_INTMAX_T]) + AC_CANONICAL_HOST dnl CPU-specific settings. CPU_SPINWAIT="" @@ -150,15 +195,15 @@ case "${host_cpu}" in ;; i686) JE_COMPILABLE([__asm__], [], [[__asm__ volatile("pause"); return 0;]], - [asm]) - if test "x${asm}" = "xyes" ; then + [je_cv_asm]) + if test "x${je_cv_asm}" = "xyes" ; then CPU_SPINWAIT='__asm__ volatile("pause")' fi ;; x86_64) JE_COMPILABLE([__asm__ syntax], [], - [[__asm__ volatile("pause"); return 0;]], [asm]) - if test "x${asm}" = "xyes" ; then + [[__asm__ volatile("pause"); return 0;]], [je_cv_asm]) + if test "x${je_cv_asm}" = "xyes" ; then CPU_SPINWAIT='__asm__ volatile("pause")' fi ;; @@ -167,6 +212,23 @@ case "${host_cpu}" in esac AC_DEFINE_UNQUOTED([CPU_SPINWAIT], [$CPU_SPINWAIT]) +LD_PRELOAD_VAR="LD_PRELOAD" +so="so" +importlib="${so}" +o="$ac_objext" +a="a" +exe="$ac_exeext" +libprefix="lib" +DSO_LDFLAGS='-shared -Wl,-soname,$(@F)' +RPATH='-Wl,-rpath,$(1)' +SOREV="${so}.${rev}" +PIC_CFLAGS='-fPIC -DPIC' +CTARGET='-o $@' +LDTARGET='-o $@' +EXTRA_LDFLAGS= +MKLIB='ar crus $@' +CC_MM=1 + dnl Platform-specific settings. abi and RPATH can probably be determined dnl programmatically, but doing so is error-prone, which makes it generally dnl not worth the trouble. @@ -174,25 +236,33 @@ dnl dnl Define cpp macros in CPPFLAGS, rather than doing AC_DEFINE(macro), since the dnl definitions need to be seen before any headers are included, which is a pain dnl to make happen otherwise. +default_munmap="1" case "${host}" in *-*-darwin*) - CFLAGS="$CFLAGS -fno-common -no-cpp-precomp" + CFLAGS="$CFLAGS" abi="macho" - AC_DEFINE([JEMALLOC_PURGE_MADVISE_FREE]) + AC_DEFINE([JEMALLOC_PURGE_MADVISE_FREE], [ ]) RPATH="" + LD_PRELOAD_VAR="DYLD_INSERT_LIBRARIES" + so="dylib" + importlib="${so}" + force_tls="0" + DSO_LDFLAGS='-shared -Wl,-dylib_install_name,$(@F)' + SOREV="${rev}.${so}" ;; *-*-freebsd*) CFLAGS="$CFLAGS" abi="elf" - AC_DEFINE([JEMALLOC_PURGE_MADVISE_FREE]) - RPATH="-Wl,-rpath," + AC_DEFINE([JEMALLOC_PURGE_MADVISE_FREE], [ ]) + force_lazy_lock="1" ;; *-*-linux*) CFLAGS="$CFLAGS" CPPFLAGS="$CPPFLAGS -D_GNU_SOURCE" abi="elf" - AC_DEFINE([JEMALLOC_PURGE_MADVISE_DONTNEED]) - RPATH="-Wl,-rpath," + AC_DEFINE([JEMALLOC_PURGE_MADVISE_DONTNEED], [ ]) + AC_DEFINE([JEMALLOC_THREADED_INIT], [ ]) + default_munmap="0" ;; *-*-netbsd*) AC_MSG_CHECKING([ABI]) @@ -206,45 +276,100 @@ case "${host}" in [CFLAGS="$CFLAGS"; abi="elf"], [abi="aout"]) AC_MSG_RESULT([$abi]) - AC_DEFINE([JEMALLOC_PURGE_MADVISE_FREE]) - RPATH="-Wl,-rpath," + AC_DEFINE([JEMALLOC_PURGE_MADVISE_FREE], [ ]) ;; *-*-solaris2*) CFLAGS="$CFLAGS" abi="elf" - RPATH="-Wl,-R," + RPATH='-Wl,-R,$(1)' dnl Solaris needs this for sigwait(). CPPFLAGS="$CPPFLAGS -D_POSIX_PTHREAD_SEMANTICS" LIBS="$LIBS -lposix4 -lsocket -lnsl" ;; + *-ibm-aix*) + if "$LG_SIZEOF_PTR" = "8"; then + dnl 64bit AIX + LD_PRELOAD_VAR="LDR_PRELOAD64" + else + dnl 32bit AIX + LD_PRELOAD_VAR="LDR_PRELOAD" + fi + abi="xcoff" + ;; + *-*-mingw*) + abi="pecoff" + force_tls="0" + RPATH="" + so="dll" + if test "x$je_cv_msvc" = "xyes" ; then + importlib="lib" + DSO_LDFLAGS="-LD" + EXTRA_LDFLAGS="-link -DEBUG" + CTARGET='-Fo$@' + LDTARGET='-Fe$@' + MKLIB='lib -nologo -out:$@' + CC_MM= + else + importlib="${so}" + DSO_LDFLAGS="-shared" + fi + a="lib" + libprefix="" + SOREV="${so}" + PIC_CFLAGS="" + ;; *) AC_MSG_RESULT([Unsupported operating system: ${host}]) abi="elf" - RPATH="-Wl,-rpath," ;; esac AC_SUBST([abi]) AC_SUBST([RPATH]) +AC_SUBST([LD_PRELOAD_VAR]) +AC_SUBST([so]) +AC_SUBST([importlib]) +AC_SUBST([o]) +AC_SUBST([a]) +AC_SUBST([exe]) +AC_SUBST([libprefix]) +AC_SUBST([DSO_LDFLAGS]) +AC_SUBST([EXTRA_LDFLAGS]) +AC_SUBST([SOREV]) +AC_SUBST([PIC_CFLAGS]) +AC_SUBST([CTARGET]) +AC_SUBST([LDTARGET]) +AC_SUBST([MKLIB]) +AC_SUBST([CC_MM]) + +if test "x$abi" != "xpecoff"; then + dnl Heap profiling uses the log(3) function. + LIBS="$LIBS -lm" +fi JE_COMPILABLE([__attribute__ syntax], [static __attribute__((unused)) void foo(void){}], [], - [attribute]) -if test "x${attribute}" = "xyes" ; then + [je_cv_attribute]) +if test "x${je_cv_attribute}" = "xyes" ; then AC_DEFINE([JEMALLOC_HAVE_ATTR], [ ]) if test "x${GCC}" = "xyes" -a "x${abi}" = "xelf"; then JE_CFLAGS_APPEND([-fvisibility=hidden]) fi fi - -JE_COMPILABLE([mremap(...MREMAP_FIXED...)], [ -#define _GNU_SOURCE -#include -], [ -void *p = mremap((void *)0, 0, 0, MREMAP_MAYMOVE|MREMAP_FIXED, (void *)0); -], [mremap_fixed]) -if test "x${mremap_fixed}" = "xyes" ; then - AC_DEFINE([JEMALLOC_MREMAP_FIXED]) +dnl Check for tls_model attribute support (clang 3.0 still lacks support). +SAVED_CFLAGS="${CFLAGS}" +JE_CFLAGS_APPEND([-Werror]) +JE_COMPILABLE([tls_model attribute], [], + [static __thread int + __attribute__((tls_model("initial-exec"))) foo; + foo = 0;], + [je_cv_tls_model]) +CFLAGS="${SAVED_CFLAGS}" +if test "x${je_cv_tls_model}" = "xyes" ; then + AC_DEFINE([JEMALLOC_TLS_MODEL], + [__attribute__((tls_model("initial-exec")))]) +else + AC_DEFINE([JEMALLOC_TLS_MODEL], [ ]) fi dnl Support optional additions to rpath. @@ -278,11 +403,52 @@ AC_PATH_PROG([AR], [ar], , [$PATH]) AC_PATH_PROG([LD], [ld], , [$PATH]) AC_PATH_PROG([AUTOCONF], [autoconf], , [$PATH]) +public_syms="malloc_conf malloc_message malloc calloc posix_memalign aligned_alloc realloc free malloc_usable_size malloc_stats_print mallctl mallctlnametomib mallctlbymib" + +dnl Check for allocator-related functions that should be wrapped. +AC_CHECK_FUNC([memalign], + [AC_DEFINE([JEMALLOC_OVERRIDE_MEMALIGN], [ ]) + public_syms="${public_syms} memalign"]) +AC_CHECK_FUNC([valloc], + [AC_DEFINE([JEMALLOC_OVERRIDE_VALLOC], [ ]) + public_syms="${public_syms} valloc"]) + +dnl Support the experimental API by default. +AC_ARG_ENABLE([experimental], + [AS_HELP_STRING([--disable-experimental], + [Disable support for the experimental API])], +[if test "x$enable_experimental" = "xno" ; then + enable_experimental="0" +else + enable_experimental="1" +fi +], +[enable_experimental="1"] +) +if test "x$enable_experimental" = "x1" ; then + AC_DEFINE([JEMALLOC_EXPERIMENTAL], [ ]) + public_syms="${public_syms} allocm dallocm nallocm rallocm sallocm" +fi +AC_SUBST([enable_experimental]) + +dnl Perform no name mangling by default. +AC_ARG_WITH([mangling], + [AS_HELP_STRING([--with-mangling=], [Mangle symbols in ])], + [mangling_map="$with_mangling"], [mangling_map=""]) +for nm in `echo ${mangling_map} |tr ',' ' '` ; do + k="`echo ${nm} |tr ':' ' ' |awk '{print $1}'`" + n="je_${k}" + m=`echo ${nm} |tr ':' ' ' |awk '{print $2}'` + AC_DEFINE_UNQUOTED([${n}], [${m}]) + dnl Remove key from public_syms so that it isn't redefined later. + public_syms=`for sym in ${public_syms}; do echo "${sym}"; done |grep -v "^${k}\$" |tr '\n' ' '` +done + dnl Do not prefix public APIs by default. AC_ARG_WITH([jemalloc_prefix], [AS_HELP_STRING([--with-jemalloc-prefix=], [Prefix to prepend to all public APIs])], [JEMALLOC_PREFIX="$with_jemalloc_prefix"], - [if test "x$abi" != "xmacho" ; then + [if test "x$abi" != "xmacho" -a "x$abi" != "xpecoff"; then JEMALLOC_PREFIX="" else JEMALLOC_PREFIX="je_" @@ -292,8 +458,15 @@ if test "x$JEMALLOC_PREFIX" != "x" ; then JEMALLOC_CPREFIX=`echo ${JEMALLOC_PREFIX} | tr "a-z" "A-Z"` AC_DEFINE_UNQUOTED([JEMALLOC_PREFIX], ["$JEMALLOC_PREFIX"]) AC_DEFINE_UNQUOTED([JEMALLOC_CPREFIX], ["$JEMALLOC_CPREFIX"]) - AC_DEFINE_UNQUOTED([JEMALLOC_P(string_that_no_one_should_want_to_use_as_a_jemalloc_API_prefix)], [${JEMALLOC_PREFIX}##string_that_no_one_should_want_to_use_as_a_jemalloc_API_prefix]) fi +dnl Generate macros to rename public symbols. All public symbols are prefixed +dnl with je_ in the source code, so these macro definitions are needed even if +dnl --with-jemalloc-prefix wasn't specified. +for stem in ${public_syms}; do + n="je_${stem}" + m="${JEMALLOC_PREFIX}${stem}" + AC_DEFINE_UNQUOTED([${n}], [${m}]) +done dnl Do not mangle library-private APIs by default. AC_ARG_WITH([private_namespace], @@ -342,8 +515,10 @@ cfgoutputs_tup="${cfgoutputs_tup} include/jemalloc/internal/jemalloc_internal.h" cfgoutputs_tup="${cfgoutputs_tup} test/jemalloc_test.h:test/jemalloc_test.h.in" cfghdrs_in="${srcroot}include/jemalloc/jemalloc_defs.h.in" +cfghdrs_in="${cfghdrs_in} ${srcroot}include/jemalloc/internal/size_classes.sh" cfghdrs_out="include/jemalloc/jemalloc_defs${install_suffix}.h" +cfghdrs_out="${cfghdrs_out} include/jemalloc/internal/size_classes.h" cfghdrs_tup="include/jemalloc/jemalloc_defs${install_suffix}.h:include/jemalloc/jemalloc_defs.h.in" @@ -361,7 +536,7 @@ fi [enable_cc_silence="0"] ) if test "x$enable_cc_silence" = "x1" ; then - AC_DEFINE([JEMALLOC_CC_SILENCE]) + AC_DEFINE([JEMALLOC_CC_SILENCE], [ ]) fi dnl Do not compile with debugging by default. @@ -390,22 +565,25 @@ if test "x$enable_debug" = "x0" -a "x$no_CFLAGS" = "xyes" ; then if test "x$GCC" = "xyes" ; then JE_CFLAGS_APPEND([-O3]) JE_CFLAGS_APPEND([-funroll-loops]) + elif test "x$je_cv_msvc" = "xyes" ; then + JE_CFLAGS_APPEND([-O2]) else JE_CFLAGS_APPEND([-O]) fi fi fi -dnl Do not enable statistics calculation by default. +dnl Enable statistics calculation by default. AC_ARG_ENABLE([stats], - [AS_HELP_STRING([--enable-stats], [Enable statistics calculation/reporting])], + [AS_HELP_STRING([--disable-stats], + [Disable statistics calculation/reporting])], [if test "x$enable_stats" = "xno" ; then enable_stats="0" else enable_stats="1" fi ], -[enable_stats="0"] +[enable_stats="1"] ) if test "x$enable_stats" = "x1" ; then AC_DEFINE([JEMALLOC_STATS], [ ]) @@ -531,27 +709,14 @@ fi AC_MSG_CHECKING([configured backtracing method]) AC_MSG_RESULT([$backtrace_method]) if test "x$enable_prof" = "x1" ; then - LIBS="$LIBS -lm" + if test "x${force_tls}" = "x0" ; then + AC_MSG_ERROR([Heap profiling requires TLS]); + fi + force_tls="1" AC_DEFINE([JEMALLOC_PROF], [ ]) fi AC_SUBST([enable_prof]) -dnl Enable tiny allocations by default. -AC_ARG_ENABLE([tiny], - [AS_HELP_STRING([--disable-tiny], [Disable tiny (sub-quantum) allocations])], -[if test "x$enable_tiny" = "xno" ; then - enable_tiny="0" -else - enable_tiny="1" -fi -], -[enable_tiny="1"] -) -if test "x$enable_tiny" = "x1" ; then - AC_DEFINE([JEMALLOC_TINY], [ ]) -fi -AC_SUBST([enable_tiny]) - dnl Enable thread-specific caching by default. AC_ARG_ENABLE([tcache], [AS_HELP_STRING([--disable-tcache], [Disable per thread caches])], @@ -568,21 +733,48 @@ if test "x$enable_tcache" = "x1" ; then fi AC_SUBST([enable_tcache]) -dnl Do not enable mmap()ped swap files by default. -AC_ARG_ENABLE([swap], - [AS_HELP_STRING([--enable-swap], [Enable mmap()ped swap files])], -[if test "x$enable_swap" = "xno" ; then - enable_swap="0" +dnl Disable mremap() for huge realloc() by default. +AC_ARG_ENABLE([mremap], + [AS_HELP_STRING([--enable-mremap], [Enable mremap(2) for huge realloc()])], +[if test "x$enable_mremap" = "xno" ; then + enable_mremap="0" else - enable_swap="1" + enable_mremap="1" fi ], -[enable_swap="0"] +[enable_mremap="0"] ) -if test "x$enable_swap" = "x1" ; then - AC_DEFINE([JEMALLOC_SWAP], [ ]) +if test "x$enable_mremap" = "x1" ; then + JE_COMPILABLE([mremap(...MREMAP_FIXED...)], [ +#define _GNU_SOURCE +#include +], [ +void *p = mremap((void *)0, 0, 0, MREMAP_MAYMOVE|MREMAP_FIXED, (void *)0); +], [je_cv_mremap_fixed]) + if test "x${je_cv_mremap_fixed}" = "xno" ; then + enable_mremap="0" + fi fi -AC_SUBST([enable_swap]) +if test "x$enable_mremap" = "x1" ; then + AC_DEFINE([JEMALLOC_MREMAP], [ ]) +fi +AC_SUBST([enable_mremap]) + +dnl Enable VM deallocation via munmap() by default. +AC_ARG_ENABLE([munmap], + [AS_HELP_STRING([--disable-munmap], [Disable VM deallocation via munmap(2)])], +[if test "x$enable_munmap" = "xno" ; then + enable_munmap="0" +else + enable_munmap="1" +fi +], +[enable_munmap="${default_munmap}"] +) +if test "x$enable_munmap" = "x1" ; then + AC_DEFINE([JEMALLOC_MUNMAP], [ ]) +fi +AC_SUBST([enable_munmap]) dnl Do not enable allocation from DSS by default. AC_ARG_ENABLE([dss], @@ -595,27 +787,93 @@ fi ], [enable_dss="0"] ) +dnl Check whether the BSD/SUSv1 sbrk() exists. If not, disable DSS support. +AC_CHECK_FUNC([sbrk], [have_sbrk="1"], [have_sbrk="0"]) +if test "x$have_sbrk" = "x1" ; then + AC_DEFINE([JEMALLOC_HAVE_SBRK], [ ]) +else + enable_dss="0" +fi + if test "x$enable_dss" = "x1" ; then AC_DEFINE([JEMALLOC_DSS], [ ]) fi AC_SUBST([enable_dss]) -dnl Do not support the junk/zero filling option by default. +dnl Support the junk/zero filling option by default. AC_ARG_ENABLE([fill], - [AS_HELP_STRING([--enable-fill], [Support junk/zero filling option])], + [AS_HELP_STRING([--disable-fill], + [Disable support for junk/zero filling, quarantine, and redzones])], [if test "x$enable_fill" = "xno" ; then enable_fill="0" else enable_fill="1" fi ], -[enable_fill="0"] +[enable_fill="1"] ) if test "x$enable_fill" = "x1" ; then AC_DEFINE([JEMALLOC_FILL], [ ]) fi AC_SUBST([enable_fill]) +dnl Disable utrace(2)-based tracing by default. +AC_ARG_ENABLE([utrace], + [AS_HELP_STRING([--enable-utrace], [Enable utrace(2)-based tracing])], +[if test "x$enable_utrace" = "xno" ; then + enable_utrace="0" +else + enable_utrace="1" +fi +], +[enable_utrace="0"] +) +JE_COMPILABLE([utrace(2)], [ +#include +#include +#include +#include +#include +], [ + utrace((void *)0, 0); +], [je_cv_utrace]) +if test "x${je_cv_utrace}" = "xno" ; then + enable_utrace="0" +fi +if test "x$enable_utrace" = "x1" ; then + AC_DEFINE([JEMALLOC_UTRACE], [ ]) +fi +AC_SUBST([enable_utrace]) + +dnl Support Valgrind by default. +AC_ARG_ENABLE([valgrind], + [AS_HELP_STRING([--disable-valgrind], [Disable support for Valgrind])], +[if test "x$enable_valgrind" = "xno" ; then + enable_valgrind="0" +else + enable_valgrind="1" +fi +], +[enable_valgrind="1"] +) +if test "x$enable_valgrind" = "x1" ; then + JE_COMPILABLE([valgrind], [ +#include +#include + +#if !defined(VALGRIND_RESIZEINPLACE_BLOCK) +# error "Incompatible Valgrind version" +#endif +], [], [je_cv_valgrind]) + if test "x${je_cv_valgrind}" = "xno" ; then + enable_valgrind="0" + fi + if test "x$enable_valgrind" = "x1" ; then + AC_DEFINE([JEMALLOC_VALGRIND], [ ]) + fi +fi +AC_SUBST([enable_valgrind]) + dnl Do not support the xmalloc option by default. AC_ARG_ENABLE([xmalloc], [AS_HELP_STRING([--enable-xmalloc], [Support xmalloc option])], @@ -632,65 +890,51 @@ if test "x$enable_xmalloc" = "x1" ; then fi AC_SUBST([enable_xmalloc]) -dnl Do not support the SYSV option by default. -AC_ARG_ENABLE([sysv], - [AS_HELP_STRING([--enable-sysv], [Support SYSV semantics option])], -[if test "x$enable_sysv" = "xno" ; then - enable_sysv="0" -else - enable_sysv="1" -fi -], -[enable_sysv="0"] -) -if test "x$enable_sysv" = "x1" ; then - AC_DEFINE([JEMALLOC_SYSV], [ ]) -fi -AC_SUBST([enable_sysv]) - -dnl Do not determine page shift at run time by default. -AC_ARG_ENABLE([dynamic_page_shift], - [AS_HELP_STRING([--enable-dynamic-page-shift], - [Determine page size at run time (don't trust configure result)])], -[if test "x$enable_dynamic_page_shift" = "xno" ; then - enable_dynamic_page_shift="0" -else - enable_dynamic_page_shift="1" -fi -], -[enable_dynamic_page_shift="0"] -) -if test "x$enable_dynamic_page_shift" = "x1" ; then - AC_DEFINE([DYNAMIC_PAGE_SHIFT], [ ]) -fi -AC_SUBST([enable_dynamic_page_shift]) - -AC_MSG_CHECKING([STATIC_PAGE_SHIFT]) -AC_RUN_IFELSE([AC_LANG_PROGRAM( -[[#include -#include +AC_CACHE_CHECK([STATIC_PAGE_SHIFT], + [je_cv_static_page_shift], + AC_RUN_IFELSE([AC_LANG_PROGRAM( +[[ #include -]], [[ +#ifdef _WIN32 +#include +#else +#include +#endif +#include +]], +[[ long result; FILE *f; +#ifdef _WIN32 + SYSTEM_INFO si; + GetSystemInfo(&si); + result = si.dwPageSize; +#else result = sysconf(_SC_PAGESIZE); +#endif if (result == -1) { return 1; } + result = ffsl(result) - 1; + f = fopen("conftest.out", "w"); if (f == NULL) { return 1; } - fprintf(f, "%u\n", ffs((int)result) - 1); - close(f); + fprintf(f, "%u\n", result); + fclose(f); return 0; ]])], - [STATIC_PAGE_SHIFT=`cat conftest.out`] - AC_MSG_RESULT([$STATIC_PAGE_SHIFT]) - AC_DEFINE_UNQUOTED([STATIC_PAGE_SHIFT], [$STATIC_PAGE_SHIFT]), - AC_MSG_RESULT([error])) + [je_cv_static_page_shift=`cat conftest.out`], + [je_cv_static_page_shift=undefined])) + +if test "x$je_cv_static_page_shift" != "xundefined"; then + AC_DEFINE_UNQUOTED([STATIC_PAGE_SHIFT], [$je_cv_static_page_shift]) +else + AC_MSG_ERROR([cannot determine value for STATIC_PAGE_SHIFT]) +fi dnl ============================================================================ dnl jemalloc configuration. @@ -716,28 +960,65 @@ AC_SUBST([jemalloc_version_gid]) dnl ============================================================================ dnl Configure pthreads. -AC_CHECK_HEADERS([pthread.h], , [AC_MSG_ERROR([pthread.h is missing])]) -AC_CHECK_LIB([pthread], [pthread_create], [LIBS="$LIBS -lpthread"], - [AC_MSG_ERROR([libpthread is missing])]) +if test "x$abi" != "xpecoff" ; then + AC_CHECK_HEADERS([pthread.h], , [AC_MSG_ERROR([pthread.h is missing])]) + dnl Some systems may embed pthreads functionality in libc; check for libpthread + dnl first, but try libc too before failing. + AC_CHECK_LIB([pthread], [pthread_create], [LIBS="$LIBS -lpthread"], + [AC_SEARCH_LIBS([pthread_create], , , + AC_MSG_ERROR([libpthread is missing]))]) +fi CPPFLAGS="$CPPFLAGS -D_REENTRANT" -dnl Enable lazy locking by default. +dnl Check whether the BSD-specific _malloc_thread_cleanup() exists. If so, use +dnl it rather than pthreads TSD cleanup functions to support cleanup during +dnl thread exit, in order to avoid pthreads library recursion during +dnl bootstrapping. +AC_CHECK_FUNC([_malloc_thread_cleanup], + [have__malloc_thread_cleanup="1"], + [have__malloc_thread_cleanup="0"] + ) +if test "x$have__malloc_thread_cleanup" = "x1" ; then + AC_DEFINE([JEMALLOC_MALLOC_THREAD_CLEANUP], [ ]) + force_tls="1" +fi + +dnl Check whether the BSD-specific _pthread_mutex_init_calloc_cb() exists. If +dnl so, mutex initialization causes allocation, and we need to implement this +dnl callback function in order to prevent recursive allocation. +AC_CHECK_FUNC([_pthread_mutex_init_calloc_cb], + [have__pthread_mutex_init_calloc_cb="1"], + [have__pthread_mutex_init_calloc_cb="0"] + ) +if test "x$have__pthread_mutex_init_calloc_cb" = "x1" ; then + AC_DEFINE([JEMALLOC_MUTEX_INIT_CB]) +fi + +dnl Disable lazy locking by default. AC_ARG_ENABLE([lazy_lock], - [AS_HELP_STRING([--disable-lazy-lock], - [Disable lazy locking (always lock, even when single-threaded)])], + [AS_HELP_STRING([--enable-lazy-lock], + [Enable lazy locking (only lock when multi-threaded)])], [if test "x$enable_lazy_lock" = "xno" ; then enable_lazy_lock="0" else enable_lazy_lock="1" fi ], -[enable_lazy_lock="1"] +[enable_lazy_lock="0"] ) +if test "x$enable_lazy_lock" = "x0" -a "x${force_lazy_lock}" = "x1" ; then + AC_MSG_RESULT([Forcing lazy-lock to avoid allocator/threading bootstrap issues]) + enable_lazy_lock="1" +fi if test "x$enable_lazy_lock" = "x1" ; then - AC_CHECK_HEADERS([dlfcn.h], , [AC_MSG_ERROR([dlfcn.h is missing])]) - AC_CHECK_LIB([dl], [dlopen], [LIBS="$LIBS -ldl"], - [AC_MSG_ERROR([libdl is missing])]) + if test "x$abi" != "xpecoff" ; then + AC_CHECK_HEADERS([dlfcn.h], , [AC_MSG_ERROR([dlfcn.h is missing])]) + AC_CHECK_FUNC([dlsym], [], + [AC_CHECK_LIB([dl], [dlsym], [LIBS="$LIBS -ldl"], + [AC_MSG_ERROR([libdl is missing])]) + ]) + fi AC_DEFINE([JEMALLOC_LAZY_LOCK], [ ]) fi AC_SUBST([enable_lazy_lock]) @@ -752,9 +1033,17 @@ fi , enable_tls="1" ) +if test "x${enable_tls}" = "x0" -a "x${force_tls}" = "x1" ; then + AC_MSG_RESULT([Forcing TLS to avoid allocator/threading bootstrap issues]) + enable_tls="1" +fi +if test "x${enable_tls}" = "x1" -a "x${force_tls}" = "x0" ; then + AC_MSG_RESULT([Forcing no TLS to avoid allocator/threading bootstrap issues]) + enable_tls="0" +fi if test "x${enable_tls}" = "x1" ; then AC_MSG_CHECKING([for TLS]) -AC_RUN_IFELSE([AC_LANG_PROGRAM( +AC_COMPILE_IFELSE([AC_LANG_PROGRAM( [[ __thread int x; ]], [[ @@ -767,17 +1056,50 @@ AC_RUN_IFELSE([AC_LANG_PROGRAM( enable_tls="0") fi AC_SUBST([enable_tls]) -if test "x${enable_tls}" = "x0" ; then - AC_DEFINE_UNQUOTED([NO_TLS], [ ]) +if test "x${enable_tls}" = "x1" ; then + AC_DEFINE_UNQUOTED([JEMALLOC_TLS], [ ]) +elif test "x${force_tls}" = "x1" ; then + AC_MSG_ERROR([Failed to configure TLS, which is mandatory for correct function]) fi dnl ============================================================================ dnl Check for ffsl(3), and fail if not found. This function exists on all dnl platforms that jemalloc currently has a chance of functioning on without dnl modification. +JE_COMPILABLE([a program using ffsl], [ +#include +#include +], [ + { + int rv = ffsl(0x08); + } +], [je_cv_function_ffsl]) +if test "x${je_cv_function_ffsl}" != "xyes" ; then + AC_MSG_ERROR([Cannot build without ffsl(3)]) +fi -AC_CHECK_FUNC([ffsl], [], - [AC_MSG_ERROR([Cannot build without ffsl(3)])]) +dnl ============================================================================ +dnl Check for atomic(9) operations as provided on FreeBSD. + +JE_COMPILABLE([atomic(9)], [ +#include +#include +#include +], [ + { + uint32_t x32 = 0; + volatile uint32_t *x32p = &x32; + atomic_fetchadd_32(x32p, 1); + } + { + unsigned long xlong = 0; + volatile unsigned long *xlongp = &xlong; + atomic_fetchadd_long(xlongp, 1); + } +], [je_cv_atomic9]) +if test "x${je_cv_atomic9}" = "xyes" ; then + AC_DEFINE([JEMALLOC_ATOMIC9]) +fi dnl ============================================================================ dnl Check for atomic(3) operations as provided on Darwin. @@ -796,9 +1118,43 @@ JE_COMPILABLE([Darwin OSAtomic*()], [ volatile int64_t *x64p = &x64; OSAtomicAdd64(1, x64p); } -], [osatomic]) -if test "x${osatomic}" = "xyes" ; then - AC_DEFINE([JEMALLOC_OSATOMIC]) +], [je_cv_osatomic]) +if test "x${je_cv_osatomic}" = "xyes" ; then + AC_DEFINE([JEMALLOC_OSATOMIC], [ ]) +fi + +dnl ============================================================================ +dnl Check whether __sync_{add,sub}_and_fetch() are available despite +dnl __GCC_HAVE_SYNC_COMPARE_AND_SWAP_n macros being undefined. + +AC_DEFUN([JE_SYNC_COMPARE_AND_SWAP_CHECK],[ + AC_CACHE_CHECK([whether to force $1-bit __sync_{add,sub}_and_fetch()], + [je_cv_sync_compare_and_swap_$2], + [AC_LINK_IFELSE([AC_LANG_PROGRAM([ + #include + ], + [ + #ifndef __GCC_HAVE_SYNC_COMPARE_AND_SWAP_$2 + { + uint$1_t x$1 = 0; + __sync_add_and_fetch(&x$1, 42); + __sync_sub_and_fetch(&x$1, 1); + } + #else + #error __GCC_HAVE_SYNC_COMPARE_AND_SWAP_$2 is defined, no need to force + #endif + ])], + [je_cv_sync_compare_and_swap_$2=yes], + [je_cv_sync_compare_and_swap_$2=no])]) + + if test "x${je_cv_sync_compare_and_swap_$2}" = "xyes" ; then + AC_DEFINE([JE_FORCE_SYNC_COMPARE_AND_SWAP_$2], [ ]) + fi +]) + +if test "x${je_cv_atomic9}" != "xyes" -a "x${je_cv_osatomic}" != "xyes" ; then + JE_SYNC_COMPARE_AND_SWAP_CHECK(32, 4) + JE_SYNC_COMPARE_AND_SWAP_CHECK(64, 8) fi dnl ============================================================================ @@ -811,69 +1167,59 @@ JE_COMPILABLE([Darwin OSSpin*()], [ OSSpinLock lock = 0; OSSpinLockLock(&lock); OSSpinLockUnlock(&lock); -], [osspin]) -if test "x${osspin}" = "xyes" ; then - AC_DEFINE([JEMALLOC_OSSPIN]) +], [je_cv_osspin]) +if test "x${je_cv_osspin}" = "xyes" ; then + AC_DEFINE([JEMALLOC_OSSPIN], [ ]) fi -dnl ============================================================================ -dnl Check for allocator-related functions that should be wrapped. - -AC_CHECK_FUNC([memalign], - [AC_DEFINE([JEMALLOC_OVERRIDE_MEMALIGN])]) -AC_CHECK_FUNC([valloc], - [AC_DEFINE([JEMALLOC_OVERRIDE_VALLOC])]) - dnl ============================================================================ dnl Darwin-related configuration. if test "x${abi}" = "xmacho" ; then - AC_DEFINE([JEMALLOC_IVSALLOC]) - AC_DEFINE([JEMALLOC_ZONE]) + AC_DEFINE([JEMALLOC_IVSALLOC], [ ]) + AC_DEFINE([JEMALLOC_ZONE], [ ]) dnl The szone version jumped from 3 to 6 between the OS X 10.5.x and 10.6 dnl releases. malloc_zone_t and malloc_introspection_t have new fields in dnl 10.6, which is the only source-level indication of the change. AC_MSG_CHECKING([malloc zone version]) - AC_TRY_COMPILE([#include -#include ], [ - static malloc_zone_t zone; - static struct malloc_introspection_t zone_introspect; + AC_DEFUN([JE_ZONE_PROGRAM], + [AC_LANG_PROGRAM( + [#include ], + [static foo[[sizeof($1) $2 sizeof(void *) * $3 ? 1 : -1]]] + )]) - zone.size = NULL; - zone.malloc = NULL; - zone.calloc = NULL; - zone.valloc = NULL; - zone.free = NULL; - zone.realloc = NULL; - zone.destroy = NULL; - zone.zone_name = "jemalloc_zone"; - zone.batch_malloc = NULL; - zone.batch_free = NULL; - zone.introspect = &zone_introspect; - zone.version = 6; - zone.memalign = NULL; - zone.free_definite_size = NULL; - - zone_introspect.enumerator = NULL; - zone_introspect.good_size = NULL; - zone_introspect.check = NULL; - zone_introspect.print = NULL; - zone_introspect.log = NULL; - zone_introspect.force_lock = NULL; - zone_introspect.force_unlock = NULL; - zone_introspect.statistics = NULL; - zone_introspect.zone_locked = NULL; -], [AC_DEFINE_UNQUOTED([JEMALLOC_ZONE_VERSION], [6]) - AC_MSG_RESULT([6])], - [AC_DEFINE_UNQUOTED([JEMALLOC_ZONE_VERSION], [3]) - AC_MSG_RESULT([3])]) + AC_COMPILE_IFELSE([JE_ZONE_PROGRAM(malloc_zone_t,==,14)],[JEMALLOC_ZONE_VERSION=3],[ + AC_COMPILE_IFELSE([JE_ZONE_PROGRAM(malloc_zone_t,==,15)],[JEMALLOC_ZONE_VERSION=5],[ + AC_COMPILE_IFELSE([JE_ZONE_PROGRAM(malloc_zone_t,==,16)],[ + AC_COMPILE_IFELSE([JE_ZONE_PROGRAM(malloc_introspection_t,==,9)],[JEMALLOC_ZONE_VERSION=6],[ + AC_COMPILE_IFELSE([JE_ZONE_PROGRAM(malloc_introspection_t,==,13)],[JEMALLOC_ZONE_VERSION=7],[JEMALLOC_ZONE_VERSION=] + )])],[ + AC_COMPILE_IFELSE([JE_ZONE_PROGRAM(malloc_zone_t,==,17)],[JEMALLOC_ZONE_VERSION=8],[ + AC_COMPILE_IFELSE([JE_ZONE_PROGRAM(malloc_zone_t,>,17)],[JEMALLOC_ZONE_VERSION=9],[JEMALLOC_ZONE_VERSION=] + )])])])]) + if test "x${JEMALLOC_ZONE_VERSION}" = "x"; then + AC_MSG_RESULT([unsupported]) + AC_MSG_ERROR([Unsupported malloc zone version]) + fi + if test "${JEMALLOC_ZONE_VERSION}" = 9; then + JEMALLOC_ZONE_VERSION=8 + AC_MSG_RESULT([> 8]) + else + AC_MSG_RESULT([$JEMALLOC_ZONE_VERSION]) + fi + AC_DEFINE_UNQUOTED(JEMALLOC_ZONE_VERSION, [$JEMALLOC_ZONE_VERSION]) fi dnl ============================================================================ dnl Check for typedefs, structures, and compiler characteristics. AC_HEADER_STDBOOL +AC_CONFIG_COMMANDS([include/jemalloc/internal/size_classes.h], [ + mkdir -p "include/jemalloc/internal" + "${srcdir}/include/jemalloc/internal/size_classes.sh" > "${objroot}include/jemalloc/internal/size_classes.h" +]) + dnl Process .in files. AC_SUBST([cfghdrs_in]) AC_SUBST([cfghdrs_out]) @@ -881,7 +1227,7 @@ AC_CONFIG_HEADERS([$cfghdrs_tup]) dnl ============================================================================ dnl Generate outputs. -AC_CONFIG_FILES([$cfgoutputs_tup config.stamp]) +AC_CONFIG_FILES([$cfgoutputs_tup config.stamp bin/jemalloc.sh]) AC_SUBST([cfgoutputs_in]) AC_SUBST([cfgoutputs_out]) AC_OUTPUT @@ -889,7 +1235,8 @@ AC_OUTPUT dnl ============================================================================ dnl Print out the results of configuration. AC_MSG_RESULT([===============================================================================]) -AC_MSG_RESULT([jemalloc version : $jemalloc_version]) +AC_MSG_RESULT([jemalloc version : ${jemalloc_version}]) +AC_MSG_RESULT([library revision : ${rev}]) AC_MSG_RESULT([]) AC_MSG_RESULT([CC : ${CC}]) AC_MSG_RESULT([CPPFLAGS : ${CPPFLAGS}]) @@ -918,6 +1265,7 @@ AC_MSG_RESULT([JEMALLOC_PRIVATE_NAMESPACE]) AC_MSG_RESULT([ : ${JEMALLOC_PRIVATE_NAMESPACE}]) AC_MSG_RESULT([install_suffix : ${install_suffix}]) AC_MSG_RESULT([autogen : ${enable_autogen}]) +AC_MSG_RESULT([experimental : ${enable_experimental}]) AC_MSG_RESULT([cc-silence : ${enable_cc_silence}]) AC_MSG_RESULT([debug : ${enable_debug}]) AC_MSG_RESULT([stats : ${enable_stats}]) @@ -925,14 +1273,14 @@ AC_MSG_RESULT([prof : ${enable_prof}]) AC_MSG_RESULT([prof-libunwind : ${enable_prof_libunwind}]) AC_MSG_RESULT([prof-libgcc : ${enable_prof_libgcc}]) AC_MSG_RESULT([prof-gcc : ${enable_prof_gcc}]) -AC_MSG_RESULT([tiny : ${enable_tiny}]) AC_MSG_RESULT([tcache : ${enable_tcache}]) AC_MSG_RESULT([fill : ${enable_fill}]) +AC_MSG_RESULT([utrace : ${enable_utrace}]) +AC_MSG_RESULT([valgrind : ${enable_valgrind}]) AC_MSG_RESULT([xmalloc : ${enable_xmalloc}]) -AC_MSG_RESULT([sysv : ${enable_sysv}]) -AC_MSG_RESULT([swap : ${enable_swap}]) +AC_MSG_RESULT([mremap : ${enable_mremap}]) +AC_MSG_RESULT([munmap : ${enable_munmap}]) AC_MSG_RESULT([dss : ${enable_dss}]) -AC_MSG_RESULT([dynamic_page_shift : ${enable_dynamic_page_shift}]) AC_MSG_RESULT([lazy_lock : ${enable_lazy_lock}]) AC_MSG_RESULT([tls : ${enable_tls}]) AC_MSG_RESULT([===============================================================================]) diff --git a/doc/jemalloc.xml.in b/doc/jemalloc.xml.in index 7a32879a..877c500f 100644 --- a/doc/jemalloc.xml.in +++ b/doc/jemalloc.xml.in @@ -30,6 +30,7 @@ malloc calloc posix_memalign + aligned_alloc realloc free malloc_usable_size @@ -41,6 +42,7 @@ rallocm sallocm dallocm + nallocm --> general purpose memory allocation functions @@ -72,6 +74,11 @@ size_t alignment size_t size + + void *aligned_alloc + size_t alignment + size_t size + void *realloc void *ptr @@ -154,6 +161,12 @@ void *ptr int flags + + int nallocm + size_t *rsize + size_t size + int flags + @@ -183,6 +196,14 @@ alignment must be a power of 2 at least as large as sizeof(void *). + The aligned_alloc function + allocates size bytes of memory such that the + allocation's base address is an even multiple of + alignment. The requested + alignment must be a power of 2. Behavior is + undefined if size is not an integral multiple of + alignment. + The realloc function changes the size of the previously allocated memory referenced by ptr to size bytes. The @@ -297,12 +318,15 @@ for (i = 0; i < nbins; i++) { Experimental API The experimental API is subject to change or removal without regard - for backward compatibility. + for backward compatibility. If + is specified during configuration, the experimental API is + omitted. The allocm, rallocm, - sallocm, and - dallocm functions all have a + sallocm, + dallocm, and + nallocm functions all have a flags argument that can be used to specify options. The functions only check the options that are contextually relevant. Use bitwise or (|) operations to @@ -351,7 +375,9 @@ for (i = 0; i < nbins; i++) { least size bytes of memory, sets *ptr to the base address of the allocation, and sets *rsize to the real size of the allocation if - rsize is not NULL. + rsize is not NULL. Behavior + is undefined if size is + 0. The rallocm function resizes the allocation at *ptr to be at least @@ -364,7 +390,8 @@ for (i = 0; i < nbins; i++) { language="C">size + extra) bytes, though inability to allocate the extra byte(s) will not by itself result in failure. Behavior is - undefined if (size + + undefined if size is 0, or if + (size + extra > SIZE_T_MAX). @@ -374,6 +401,15 @@ for (i = 0; i < nbins; i++) { The dallocm function causes the memory referenced by ptr to be made available for future allocations. + + The nallocm function allocates no + memory, but it performs the same size computation as the + allocm function, and if + rsize is not NULL it sets + *rsize to the real size of the allocation that + would result from the equivalent allocm + function call. Behavior is undefined if + size is 0. @@ -408,9 +444,9 @@ for (i = 0; i < nbins; i++) { suboptimal for several reasons, including race conditions, increased fragmentation, and artificial limitations on maximum usable memory. If is specified during configuration, this - allocator uses both sbrk + allocator uses both mmap 2 and - mmap + sbrk 2, in that order of preference; otherwise only mmap 2 is used. @@ -455,24 +491,14 @@ for (i = 0; i < nbins; i++) { allocations in constant time. Small objects are managed in groups by page runs. Each run maintains - a frontier and free list to track which regions are in use. Unless - is specified during configuration, - allocation requests that are no more than half the quantum (8 or 16, - depending on architecture) are rounded up to the nearest power of two that - is at least sizeof(void *). - Allocation requests that are more than half the quantum, but no more than - the minimum cacheline-multiple size class (see the opt.lg_qspace_max - option) are rounded up to the nearest multiple of the quantum. Allocation - requests that are more than the minimum cacheline-multiple size class, but - no more than the minimum subpage-multiple size class (see the opt.lg_cspace_max - option) are rounded up to the nearest multiple of the cacheline size (64). - Allocation requests that are more than the minimum subpage-multiple size - class, but no more than the maximum subpage-multiple size class are rounded - up to the nearest multiple of the subpage size (256). Allocation requests - that are more than the maximum subpage-multiple size class, but small - enough to fit in an arena-managed chunk (see the sizeof(double). All other small + object size classes are multiples of the quantum, spaced such that internal + fragmentation is limited to approximately 25% for all but the smallest size + classes. Allocation requests that are larger than the maximum small size + class, but small enough to fit in an arena-managed chunk (see the opt.lg_chunk option), are rounded up to the nearest run size. Allocation requests that are too large to fit in an arena-managed chunk are rounded up to the nearest multiple of @@ -490,41 +516,55 @@ for (i = 0; i < nbins; i++) { Size classes - - - - + + + + Category - Subcategory + Spacing Size - Small - Tiny + Small + lg [8] - Quantum-spaced + 16 [16, 32, 48, ..., 128] - Cacheline-spaced - [192, 256, 320, ..., 512] + 32 + [160, 192, 224, 256] - Subpage-spaced - [768, 1024, 1280, ..., 3840] + 64 + [320, 384, 448, 512] - Large + 128 + [640, 768, 896, 1024] + + + 256 + [1280, 1536, 1792, 2048] + + + 512 + [2560, 3072, 3584] + + + Large + 4 KiB [4 KiB, 8 KiB, 12 KiB, ..., 4072 KiB] - Huge + Huge + 4 MiB [4 MiB, 8 MiB, 12 MiB, ...] @@ -590,16 +630,6 @@ for (i = 0; i < nbins; i++) { build configuration. - - - config.dynamic_page_shift - (bool) - r- - - was - specified during build configuration. - - config.fill @@ -620,6 +650,26 @@ for (i = 0; i < nbins; i++) { during build configuration. + + + config.mremap + (bool) + r- + + was specified during + build configuration. + + + + + config.munmap + (bool) + r- + + was specified during + build configuration. + + config.prof @@ -660,26 +710,6 @@ for (i = 0; i < nbins; i++) { build configuration. - - - config.swap - (bool) - r- - - was specified during - build configuration. - - - - - config.sysv - (bool) - r- - - was specified during - build configuration. - - config.tcache @@ -690,16 +720,6 @@ for (i = 0; i < nbins; i++) { during build configuration. - - - config.tiny - (bool) - r- - - was not specified - during build configuration. - - config.tls @@ -710,6 +730,26 @@ for (i = 0; i < nbins; i++) { build configuration. + + + config.utrace + (bool) + r- + + was specified during + build configuration. + + + + + config.valgrind + (bool) + r- + + was specified during + build configuration. + + config.xmalloc @@ -735,30 +775,6 @@ for (i = 0; i < nbins; i++) { - - - opt.lg_qspace_max - (size_t) - r- - - Size (log base 2) of the maximum size class that is a - multiple of the quantum (8 or 16 bytes, depending on architecture). - Above this size, cacheline spacing is used for size classes. The - default value is 128 bytes (2^7). - - - - - opt.lg_cspace_max - (size_t) - r- - - Size (log base 2) of the maximum size class that is a - multiple of the cacheline size (64). Above this size, subpage spacing - (256 bytes) is used for size classes. The default value is 512 bytes - (2^9). - - opt.lg_chunk @@ -833,6 +849,45 @@ for (i = 0; i < nbins; i++) { configuration, in which case it is enabled by default. + + + opt.quarantine + (size_t) + r- + [] + + Per thread quarantine size in bytes. If non-zero, each + thread maintains a FIFO object quarantine that stores up to the + specified number of bytes of memory. The quarantined memory is not + freed until it is released from quarantine, though it is immediately + junk-filled if the opt.junk option is + enabled. This feature is of particular use in combination with Valgrind, which can detect attempts + to access quarantined objects. This is intended for debugging and will + impact performance negatively. The default quarantine size is + 0. + + + + + opt.redzone + (bool) + r- + [] + + Redzones enabled/disabled. If enabled, small + allocations have redzones before and after them. Furthermore, if the + opt.junk option is + enabled, the redzones are checked for corruption during deallocation. + However, the primary intended purpose of this feature is to be used in + combination with Valgrind, + which needs redzones in order to do effective buffer overflow/underflow + detection. This option is intended for debugging and will impact + performance negatively. This option is disabled by + default. + + opt.zero @@ -850,20 +905,36 @@ for (i = 0; i < nbins; i++) { - + - opt.sysv + opt.utrace (bool) r- - [] + [] - If enabled, attempting to allocate zero bytes will - return a NULL pointer instead of a valid pointer. - (The default behavior is to make a minimal allocation and return a - pointer to it.) This option is provided for System V compatibility. - This option is incompatible with the opt.xmalloc option. - This option is disabled by default. + Allocation tracing based on + utrace + 2 enabled/disabled. This option + is disabled by default. + + + + + opt.valgrind + (bool) + r- + [] + + Valgrind + support enabled/disabled. If enabled, several other options are + automatically modified during options processing to work well with + Valgrind: opt.junk + and opt.zero are set + to false, opt.quarantine is + set to 16 MiB, and opt.redzone is set to + true. This option is disabled by default. @@ -899,29 +970,11 @@ malloc_conf = "xmalloc:true";]]> allocations to be satisfied without performing any thread synchronization, at the cost of increased memory use. See the opt.lg_tcache_gc_sweep - and opt.lg_tcache_max - options for related tuning information. This option is enabled by + option for related tuning information. This option is enabled by default. - - - opt.lg_tcache_gc_sweep - (ssize_t) - r- - [] - - Approximate interval (log base 2) between full - thread-specific cache garbage collection sweeps, counted in terms of - thread-specific cache allocation/deallocation events. Garbage - collection is actually performed incrementally, one size class at a - time, in order to avoid large collection pauses. The default sweep - interval is 8192 (2^13); setting this option to -1 will disable garbage - collection. - - opt.lg_tcache_max @@ -943,31 +996,21 @@ malloc_conf = "xmalloc:true";]]> [] Memory profiling enabled/disabled. If enabled, profile - memory allocation activity, and use an - atexit - 3 function to dump final memory - usage to a file named according to the pattern - <prefix>.<pid>.<seq>.f.heap, - where <prefix> is controlled by the opt.prof_prefix - option. See the opt.lg_prof_bt_max - option for backtrace depth control. See the opt.prof_active option for on-the-fly activation/deactivation. See the opt.lg_prof_sample option for probabilistic sampling control. See the opt.prof_accum option for control of cumulative sample reporting. See the opt.lg_prof_tcmax - option for control of per thread backtrace caching. See the opt.lg_prof_interval - option for information on interval-triggered profile dumping, and the - opt.prof_gdump - option for information on high-water-triggered profile dumping. - Profile output is compatible with the included pprof - Perl script, which originates from the google-perftools + option for information on interval-triggered profile dumping, the opt.prof_gdump + option for information on high-water-triggered profile dumping, and the + opt.prof_final + option for final profile dumping. Profile output is compatible with + the included pprof Perl script, which originates + from the gperftools package. @@ -985,17 +1028,6 @@ malloc_conf = "xmalloc:true";]]> jeprof. - - - opt.lg_prof_bt_max - (size_t) - r- - [] - - Maximum backtrace depth (log base 2) when profiling - memory allocation activity. The default is 128 (2^7). - - opt.prof_active @@ -1023,8 +1055,8 @@ malloc_conf = "xmalloc:true";]]> Average interval (log base 2) between allocation samples, as measured in bytes of allocation activity. Increasing the sampling interval decreases profile fidelity, but also decreases the - computational overhead. The default sample interval is 1 (2^0) (i.e. - all allocations are sampled). + computational overhead. The default sample interval is 512 KiB (2^19 + B). @@ -1038,28 +1070,8 @@ malloc_conf = "xmalloc:true";]]> dumps enabled/disabled. If this option is enabled, every unique backtrace must be stored for the duration of execution. Depending on the application, this can impose a large memory overhead, and the - cumulative counts are not always of interest. See the - opt.lg_prof_tcmax - option for control of per thread backtrace caching, which has important - interactions. This option is enabled by default. - - - - - opt.lg_prof_tcmax - (ssize_t) - r- - [] - - Maximum per thread backtrace cache (log base 2) used - for heap profiling. A backtrace can only be discarded if the - opt.prof_accum - option is disabled, and no thread caches currently refer to the - backtrace. Therefore, a backtrace cache limit should be imposed if the - intention is to limit how much memory is used by backtraces. By - default, no limit is imposed (encoded as -1). - + cumulative counts are not always of interest. This option is disabled + by default. @@ -1099,6 +1111,23 @@ malloc_conf = "xmalloc:true";]]> option. This option is disabled by default. + + + opt.prof_final + (bool) + r- + [] + + Use an + atexit + 3 function to dump final memory + usage to a file named according to the pattern + <prefix>.<pid>.<seq>.f.heap, + where <prefix> is controlled by the opt.prof_prefix + option. This option is enabled by default. + + opt.prof_leak @@ -1110,51 +1139,11 @@ malloc_conf = "xmalloc:true";]]> atexit 3 function to report memory leaks detected by allocation sampling. See the - opt.lg_prof_bt_max - option for backtrace depth control. See the opt.prof option for information on analyzing heap profile output. This option is disabled by default. - - - opt.overcommit - (bool) - r- - [] - - Over-commit enabled/disabled. If enabled, over-commit - memory as a side effect of using anonymous - mmap - 2 or - sbrk - 2 for virtual memory allocation. - In order for overcommit to be disabled, the swap.fds mallctl must have - been successfully written to. This option is enabled by - default. - - - - - tcache.flush - (void) - -- - [] - - Flush calling thread's tcache. This interface releases - all cached objects and internal data structures associated with the - calling thread's thread-specific cache. Ordinarily, this interface - need not be called, since automatic periodic incremental garbage - collection occurs, and the thread cache is automatically discarded when - a thread exits. However, garbage collection is triggered by allocation - activity, so it is possible for a thread that stops - allocating/deallocating to retain its cache indefinitely, in which case - the developer may find manual flushing useful. - - thread.arena @@ -1226,6 +1215,38 @@ malloc_conf = "xmalloc:true";]]> mallctl* calls. + + + thread.tcache.enabled + (bool) + rw + [] + + Enable/disable calling thread's tcache. The tcache is + implicitly flushed as a side effect of becoming + disabled (see thread.tcache.flush). + + + + + + thread.tcache.flush + (void) + -- + [] + + Flush calling thread's tcache. This interface releases + all cached objects and internal data structures associated with the + calling thread's thread-specific cache. Ordinarily, this interface + need not be called, since automatic periodic incremental garbage + collection occurs, and the thread cache is automatically discarded when + a thread exits. However, garbage collection is triggered by allocation + activity, so it is possible for a thread that stops + allocating/deallocating to retain its cache indefinitely, in which case + the developer may find manual flushing useful. + + arenas.narenas @@ -1258,114 +1279,13 @@ malloc_conf = "xmalloc:true";]]> - arenas.cacheline - (size_t) - r- - - Assumed cacheline size. - - - - - arenas.subpage - (size_t) - r- - - Subpage size class interval. - - - - - arenas.pagesize + arenas.page (size_t) r- Page size. - - - arenas.chunksize - (size_t) - r- - - Chunk size. - - - - - arenas.tspace_min - (size_t) - r- - - Minimum tiny size class. Tiny size classes are powers - of two. - - - - - arenas.tspace_max - (size_t) - r- - - Maximum tiny size class. Tiny size classes are powers - of two. - - - - - arenas.qspace_min - (size_t) - r- - - Minimum quantum-spaced size class. - - - - - arenas.qspace_max - (size_t) - r- - - Maximum quantum-spaced size class. - - - - - arenas.cspace_min - (size_t) - r- - - Minimum cacheline-spaced size class. - - - - - arenas.cspace_max - (size_t) - r- - - Maximum cacheline-spaced size class. - - - - - arenas.sspace_min - (size_t) - r- - - Minimum subpage-spaced size class. - - - - - arenas.sspace_max - (size_t) - r- - - Maximum subpage-spaced size class. - - arenas.tcache_max @@ -1376,52 +1296,13 @@ malloc_conf = "xmalloc:true";]]> Maximum thread-cached size class. - - - arenas.ntbins - (unsigned) - r- - - Number of tiny bin size classes. - - - - - arenas.nqbins - (unsigned) - r- - - Number of quantum-spaced bin size - classes. - - - - - arenas.ncbins - (unsigned) - r- - - Number of cacheline-spaced bin size - classes. - - - - - arenas.nsbins - (unsigned) - r- - - Number of subpage-spaced bin size - classes. - - arenas.nbins (unsigned) r- - Total number of bin size classes. + Number of bin size classes. @@ -1590,8 +1471,7 @@ malloc_conf = "xmalloc:true";]]> application. This is a multiple of the chunk size, and is at least as large as stats.active. This - does not include inactive chunks backed by swap files. his does not - include inactive chunks embedded in the DSS. + does not include inactive chunks. @@ -1602,8 +1482,7 @@ malloc_conf = "xmalloc:true";]]> [] Total number of chunks actively mapped on behalf of the - application. This does not include inactive chunks backed by swap - files. This does not include inactive chunks embedded in the DSS. + application. This does not include inactive chunks. @@ -1908,17 +1787,6 @@ malloc_conf = "xmalloc:true";]]> to allocate changed. - - - stats.arenas.<i>.bins.<j>.highruns - (size_t) - r- - [] - - Maximum number of runs at any time thus far. - - - stats.arenas.<i>.bins.<j>.curruns @@ -1962,17 +1830,6 @@ malloc_conf = "xmalloc:true";]]> class. - - - stats.arenas.<i>.lruns.<j>.highruns - (size_t) - r- - [] - - Maximum number of runs at any time thus far for this - size class. - - stats.arenas.<i>.lruns.<j>.curruns @@ -1983,65 +1840,6 @@ malloc_conf = "xmalloc:true";]]> Current number of runs for this size class. - - - - swap.avail - (size_t) - r- - [] - - Number of swap file bytes that are currently not - associated with any chunk (i.e. mapped, but otherwise completely - unmanaged). - - - - - swap.prezeroed - (bool) - rw - [] - - If true, the allocator assumes that the swap file(s) - contain nothing but nil bytes. If this assumption is violated, - allocator behavior is undefined. This value becomes read-only after - swap.fds is - successfully written to. - - - - - swap.nfds - (size_t) - r- - [] - - Number of file descriptors in use for swap. - - - - - - swap.fds - (int *) - rw - [] - - When written to, the files associated with the - specified file descriptors are contiguously mapped via - mmap - 2. The resulting virtual memory - region is preferred over anonymous - mmap - 2 and - sbrk - 2 memory. Note that if a file's - size is not a multiple of the page size, it is automatically truncated - to the nearest page size multiple. See the - swap.prezeroed - mallctl for specifying that the files are pre-zeroed. - @@ -2065,10 +1863,11 @@ malloc_conf = "xmalloc:true";]]> This implementation does not provide much detail about the problems it detects, because the performance impact for storing such information - would be prohibitive. There are a number of allocator implementations - available on the Internet which focus on detecting and pinpointing problems - by trading performance for extra sanity checks and detailed - diagnostics. + would be prohibitive. However, jemalloc does integrate with the most + excellent Valgrind tool if the + configuration option is enabled and the + opt.valgrind option + is enabled. DIAGNOSTIC MESSAGES @@ -2124,6 +1923,27 @@ malloc_conf = "xmalloc:true";]]> + The aligned_alloc function returns + a pointer to the allocated memory if successful; otherwise a + NULL pointer is returned and + errno is set. The + aligned_alloc function will fail if: + + + EINVAL + + The alignment parameter is + not a power of 2. + + + + ENOMEM + + Memory allocation error. + + + + The realloc function returns a pointer, possibly identical to ptr, to the allocated memory if successful; otherwise a NULL @@ -2196,11 +2016,13 @@ malloc_conf = "xmalloc:true";]]> Experimental API The allocm, rallocm, - sallocm, and - dallocm functions return + sallocm, + dallocm, and + nallocm functions return ALLOCM_SUCCESS on success; otherwise they return an - error value. The allocm and - rallocm functions will fail if: + error value. The allocm, + rallocm, and + nallocm functions will fail if: ALLOCM_ERR_OOM @@ -2259,6 +2081,8 @@ malloc_conf = "lg_chunk:24";]]> 2, sbrk 2, + utrace + 2, alloca 3, atexit diff --git a/include/jemalloc/internal/arena.h b/include/jemalloc/internal/arena.h index b80c118d..0b0f640a 100644 --- a/include/jemalloc/internal/arena.h +++ b/include/jemalloc/internal/arena.h @@ -1,41 +1,6 @@ /******************************************************************************/ #ifdef JEMALLOC_H_TYPES -/* - * Subpages are an artificially designated partitioning of pages. Their only - * purpose is to support subpage-spaced size classes. - * - * There must be at least 4 subpages per page, due to the way size classes are - * handled. - */ -#define LG_SUBPAGE 8 -#define SUBPAGE ((size_t)(1U << LG_SUBPAGE)) -#define SUBPAGE_MASK (SUBPAGE - 1) - -/* Return the smallest subpage multiple that is >= s. */ -#define SUBPAGE_CEILING(s) \ - (((s) + SUBPAGE_MASK) & ~SUBPAGE_MASK) - -#ifdef JEMALLOC_TINY - /* Smallest size class to support. */ -# define LG_TINY_MIN LG_SIZEOF_PTR -# define TINY_MIN (1U << LG_TINY_MIN) -#endif - -/* - * Maximum size class that is a multiple of the quantum, but not (necessarily) - * a power of 2. Above this size, allocations are rounded up to the nearest - * power of 2. - */ -#define LG_QSPACE_MAX_DEFAULT 7 - -/* - * Maximum size class that is a multiple of the cacheline, but not (necessarily) - * a power of 2. Above this size, allocations are rounded up to the nearest - * power of 2. - */ -#define LG_CSPACE_MAX_DEFAULT 9 - /* * RUN_MAX_OVRHD indicates maximum desired run header overhead. Runs are sized * as small as possible such that this setting is still honored, without @@ -51,7 +16,7 @@ * constraint is relaxed (ignored) for runs that are so small that the * per-region overhead is greater than: * - * (RUN_MAX_OVRHD / (reg_size << (3+RUN_BFP)) + * (RUN_MAX_OVRHD / (reg_interval << (3+RUN_BFP)) */ #define RUN_BFP 12 /* \/ Implicit binary fixed point. */ @@ -62,6 +27,12 @@ #define LG_RUN_MAXREGS 11 #define RUN_MAXREGS (1U << LG_RUN_MAXREGS) +/* + * Minimum redzone size. Redzones may be larger than this if necessary to + * preserve region alignment. + */ +#define REDZONE_MINSIZE 16 + /* * The minimum ratio of active:dirty pages per arena is computed as: * @@ -85,6 +56,15 @@ typedef struct arena_s arena_t; /* Each element of the chunk map corresponds to one page within the chunk. */ struct arena_chunk_map_s { +#ifndef JEMALLOC_PROF + /* + * Overlay prof_ctx in order to allow it to be referenced by dead code. + * Such antics aren't warranted for per arena data structures, but + * chunk map overhead accounts for a percentage of memory, rather than + * being just a fixed cost. + */ + union { +#endif union { /* * Linkage for run trees. There are two disjoint uses: @@ -103,22 +83,23 @@ struct arena_chunk_map_s { ql_elm(arena_chunk_map_t) ql_link; } u; -#ifdef JEMALLOC_PROF /* Profile counters, used for large object runs. */ prof_ctx_t *prof_ctx; +#ifndef JEMALLOC_PROF + }; /* union { ... }; */ #endif /* * Run address (or size) and various flags are stored together. The bit * layout looks like (assuming 32-bit system): * - * ???????? ???????? ????---- ----dula + * ???????? ???????? ????nnnn nnnndula * * ? : Unallocated: Run address for first/last pages, unset for internal * pages. * Small: Run page offset. * Large: Run size for first page, unset for trailing pages. - * - : Unused. + * n : binind for small size class, BININD_INVALID for large size class. * d : dirty? * u : unzeroed? * l : large? @@ -128,7 +109,8 @@ struct arena_chunk_map_s { * * p : run page offset * s : run size - * c : (binind+1) for size class (used only if prof_promote is true) + * n : binind for size class; large objects set these to BININD_INVALID + * except for promoted allocations (see prof_promote) * x : don't care * - : 0 * + : 1 @@ -136,37 +118,38 @@ struct arena_chunk_map_s { * [dula] : bit unset * * Unallocated (clean): - * ssssssss ssssssss ssss---- ----du-a - * xxxxxxxx xxxxxxxx xxxx---- -----Uxx - * ssssssss ssssssss ssss---- ----dU-a + * ssssssss ssssssss ssss++++ ++++du-a + * xxxxxxxx xxxxxxxx xxxxxxxx xxxx-Uxx + * ssssssss ssssssss ssss++++ ++++dU-a * * Unallocated (dirty): - * ssssssss ssssssss ssss---- ----D--a - * xxxxxxxx xxxxxxxx xxxx---- ----xxxx - * ssssssss ssssssss ssss---- ----D--a + * ssssssss ssssssss ssss++++ ++++D--a + * xxxxxxxx xxxxxxxx xxxxxxxx xxxxxxxx + * ssssssss ssssssss ssss++++ ++++D--a * * Small: - * pppppppp pppppppp pppp---- ----d--A - * pppppppp pppppppp pppp---- -------A - * pppppppp pppppppp pppp---- ----d--A + * pppppppp pppppppp ppppnnnn nnnnd--A + * pppppppp pppppppp ppppnnnn nnnn---A + * pppppppp pppppppp ppppnnnn nnnnd--A * * Large: - * ssssssss ssssssss ssss---- ----D-LA - * xxxxxxxx xxxxxxxx xxxx---- ----xxxx - * -------- -------- -------- ----D-LA + * ssssssss ssssssss ssss++++ ++++D-LA + * xxxxxxxx xxxxxxxx xxxxxxxx xxxxxxxx + * -------- -------- ----++++ ++++D-LA * - * Large (sampled, size <= PAGE_SIZE): - * ssssssss ssssssss sssscccc ccccD-LA + * Large (sampled, size <= PAGE): + * ssssssss ssssssss ssssnnnn nnnnD-LA * - * Large (not sampled, size == PAGE_SIZE): - * ssssssss ssssssss ssss---- ----D-LA + * Large (not sampled, size == PAGE): + * ssssssss ssssssss ssss++++ ++++D-LA */ size_t bits; -#ifdef JEMALLOC_PROF -#define CHUNK_MAP_CLASS_SHIFT 4 -#define CHUNK_MAP_CLASS_MASK ((size_t)0xff0U) -#endif -#define CHUNK_MAP_FLAGS_MASK ((size_t)0xfU) +#define CHUNK_MAP_BININD_SHIFT 4 +#define BININD_INVALID ((size_t)0xffU) +/* CHUNK_MAP_BININD_MASK == (BININD_INVALID << CHUNK_MAP_BININD_SHIFT) */ +#define CHUNK_MAP_BININD_MASK ((size_t)0xff0U) +#define CHUNK_MAP_BININD_INVALID CHUNK_MAP_BININD_MASK +#define CHUNK_MAP_FLAGS_MASK ((size_t)0xcU) #define CHUNK_MAP_DIRTY ((size_t)0x8U) #define CHUNK_MAP_UNZEROED ((size_t)0x4U) #define CHUNK_MAP_LARGE ((size_t)0x2U) @@ -205,11 +188,6 @@ struct arena_chunk_s { typedef rb_tree(arena_chunk_t) arena_chunk_tree_t; struct arena_run_s { -#ifdef JEMALLOC_DEBUG - uint32_t magic; -# define ARENA_RUN_MAGIC 0x384adf93 -#endif - /* Bin this run is associated with. */ arena_bin_t *bin; @@ -224,11 +202,50 @@ struct arena_run_s { * Read-only information associated with each element of arena_t's bins array * is stored separately, partly to reduce memory usage (only one copy, rather * than one per arena), but mainly to avoid false cacheline sharing. + * + * Each run has the following layout: + * + * /--------------------\ + * | arena_run_t header | + * | ... | + * bitmap_offset | bitmap | + * | ... | + * ctx0_offset | ctx map | + * | ... | + * |--------------------| + * | redzone | + * reg0_offset | region 0 | + * | redzone | + * |--------------------| \ + * | redzone | | + * | region 1 | > reg_interval + * | redzone | / + * |--------------------| + * | ... | + * | ... | + * | ... | + * |--------------------| + * | redzone | + * | region nregs-1 | + * | redzone | + * |--------------------| + * | alignment pad? | + * \--------------------/ + * + * reg_interval has at least the same minimum alignment as reg_size; this + * preserves the alignment constraint that sa2u() depends on. Alignment pad is + * either 0 or redzone_size; it is present only if needed to align reg0_offset. */ struct arena_bin_info_s { /* Size of regions in a run for this bin's size class. */ size_t reg_size; + /* Redzone size. */ + size_t redzone_size; + + /* Interval between regions (reg_size + (redzone_size << 1)). */ + size_t reg_interval; + /* Total size of a run for this bin's size class. */ size_t run_size; @@ -247,13 +264,11 @@ struct arena_bin_info_s { */ bitmap_info_t bitmap_info; -#ifdef JEMALLOC_PROF /* * Offset of first (prof_ctx_t *) in a run header for this bin's size - * class, or 0 if (opt_prof == false). + * class, or 0 if (config_prof == false || opt_prof == false). */ uint32_t ctx0_offset; -#endif /* Offset of first region in a run for this bin's size class. */ uint32_t reg0_offset; @@ -283,18 +298,11 @@ struct arena_bin_s { */ arena_run_tree_t runs; -#ifdef JEMALLOC_STATS /* Bin statistics. */ malloc_bin_stats_t stats; -#endif }; struct arena_s { -#ifdef JEMALLOC_DEBUG - uint32_t magic; -# define ARENA_MAGIC 0x947d3d24 -#endif - /* This arena's index within the arenas array. */ unsigned ind; @@ -314,20 +322,14 @@ struct arena_s { */ malloc_mutex_t lock; -#ifdef JEMALLOC_STATS arena_stats_t stats; -# ifdef JEMALLOC_TCACHE /* * List of tcaches for extant threads associated with this arena. * Stats from these are merged incrementally, and at exit. */ ql_head(tcache_t) tcache_ql; -# endif -#endif -#ifdef JEMALLOC_PROF uint64_t prof_accumbytes; -#endif /* List of dirty-page-containing chunks this arena manages. */ ql_head(arena_chunk_t) chunks_dirty; @@ -378,140 +380,334 @@ struct arena_s { arena_avail_tree_t runs_avail_clean; arena_avail_tree_t runs_avail_dirty; - /* - * bins is used to store trees of free regions of the following sizes, - * assuming a 64-bit system with 16-byte quantum, 4 KiB page size, and - * default MALLOC_CONF. - * - * bins[i] | size | - * --------+--------+ - * 0 | 8 | - * --------+--------+ - * 1 | 16 | - * 2 | 32 | - * 3 | 48 | - * : : - * 6 | 96 | - * 7 | 112 | - * 8 | 128 | - * --------+--------+ - * 9 | 192 | - * 10 | 256 | - * 11 | 320 | - * 12 | 384 | - * 13 | 448 | - * 14 | 512 | - * --------+--------+ - * 15 | 768 | - * 16 | 1024 | - * 17 | 1280 | - * : : - * 25 | 3328 | - * 26 | 3584 | - * 27 | 3840 | - * --------+--------+ - */ - arena_bin_t bins[1]; /* Dynamically sized. */ + /* bins is used to store trees of free regions. */ + arena_bin_t bins[NBINS]; }; #endif /* JEMALLOC_H_STRUCTS */ /******************************************************************************/ #ifdef JEMALLOC_H_EXTERNS -extern size_t opt_lg_qspace_max; -extern size_t opt_lg_cspace_max; extern ssize_t opt_lg_dirty_mult; /* * small_size2bin is a compact lookup table that rounds request sizes up to * size classes. In order to reduce cache footprint, the table is compressed, * and all accesses are via the SMALL_SIZE2BIN macro. */ -extern uint8_t const *small_size2bin; +extern uint8_t const small_size2bin[]; #define SMALL_SIZE2BIN(s) (small_size2bin[(s-1) >> LG_TINY_MIN]) -extern arena_bin_info_t *arena_bin_info; - -/* Various bin-related settings. */ -#ifdef JEMALLOC_TINY /* Number of (2^n)-spaced tiny bins. */ -# define ntbins ((unsigned)(LG_QUANTUM - LG_TINY_MIN)) -#else -# define ntbins 0 -#endif -extern unsigned nqbins; /* Number of quantum-spaced bins. */ -extern unsigned ncbins; /* Number of cacheline-spaced bins. */ -extern unsigned nsbins; /* Number of subpage-spaced bins. */ -extern unsigned nbins; -#ifdef JEMALLOC_TINY -# define tspace_max ((size_t)(QUANTUM >> 1)) -#endif -#define qspace_min QUANTUM -extern size_t qspace_max; -extern size_t cspace_min; -extern size_t cspace_max; -extern size_t sspace_min; -extern size_t sspace_max; -#define small_maxclass sspace_max +extern arena_bin_info_t arena_bin_info[NBINS]; +/* Number of large size classes. */ #define nlclasses (chunk_npages - map_bias) void arena_purge_all(arena_t *arena); -#ifdef JEMALLOC_PROF void arena_prof_accum(arena_t *arena, uint64_t accumbytes); -#endif -#ifdef JEMALLOC_TCACHE void arena_tcache_fill_small(arena_t *arena, tcache_bin_t *tbin, - size_t binind -# ifdef JEMALLOC_PROF - , uint64_t prof_accumbytes -# endif - ); -#endif + size_t binind, uint64_t prof_accumbytes); +void arena_alloc_junk_small(void *ptr, arena_bin_info_t *bin_info, + bool zero); +void arena_dalloc_junk_small(void *ptr, arena_bin_info_t *bin_info); void *arena_malloc_small(arena_t *arena, size_t size, bool zero); void *arena_malloc_large(arena_t *arena, size_t size, bool zero); -void *arena_malloc(size_t size, bool zero); -void *arena_palloc(arena_t *arena, size_t size, size_t alloc_size, - size_t alignment, bool zero); -size_t arena_salloc(const void *ptr); -#ifdef JEMALLOC_PROF +void *arena_palloc(arena_t *arena, size_t size, size_t alignment, bool zero); void arena_prof_promoted(const void *ptr, size_t size); -size_t arena_salloc_demote(const void *ptr); -#endif -void arena_dalloc_bin(arena_t *arena, arena_chunk_t *chunk, void *ptr, +void arena_dalloc_bin_locked(arena_t *arena, arena_chunk_t *chunk, void *ptr, arena_chunk_map_t *mapelm); +void arena_dalloc_bin(arena_t *arena, arena_chunk_t *chunk, void *ptr, + size_t pageind, arena_chunk_map_t *mapelm); +void arena_dalloc_small(arena_t *arena, arena_chunk_t *chunk, void *ptr, + size_t pageind); +void arena_dalloc_large_locked(arena_t *arena, arena_chunk_t *chunk, + void *ptr); void arena_dalloc_large(arena_t *arena, arena_chunk_t *chunk, void *ptr); -#ifdef JEMALLOC_STATS void arena_stats_merge(arena_t *arena, size_t *nactive, size_t *ndirty, arena_stats_t *astats, malloc_bin_stats_t *bstats, malloc_large_stats_t *lstats); -#endif void *arena_ralloc_no_move(void *ptr, size_t oldsize, size_t size, size_t extra, bool zero); void *arena_ralloc(void *ptr, size_t oldsize, size_t size, size_t extra, - size_t alignment, bool zero); + size_t alignment, bool zero, bool try_tcache); bool arena_new(arena_t *arena, unsigned ind); -bool arena_boot(void); +void arena_boot(void); +void arena_prefork(arena_t *arena); +void arena_postfork_parent(arena_t *arena); +void arena_postfork_child(arena_t *arena); #endif /* JEMALLOC_H_EXTERNS */ /******************************************************************************/ #ifdef JEMALLOC_H_INLINES #ifndef JEMALLOC_ENABLE_INLINE +arena_chunk_map_t *arena_mapp_get(arena_chunk_t *chunk, size_t pageind); +size_t *arena_mapbitsp_get(arena_chunk_t *chunk, size_t pageind); +size_t arena_mapbits_get(arena_chunk_t *chunk, size_t pageind); +size_t arena_mapbits_unallocated_size_get(arena_chunk_t *chunk, + size_t pageind); +size_t arena_mapbits_large_size_get(arena_chunk_t *chunk, size_t pageind); +size_t arena_mapbits_small_runind_get(arena_chunk_t *chunk, size_t pageind); +size_t arena_mapbits_binind_get(arena_chunk_t *chunk, size_t pageind); +size_t arena_mapbits_dirty_get(arena_chunk_t *chunk, size_t pageind); +size_t arena_mapbits_unzeroed_get(arena_chunk_t *chunk, size_t pageind); +size_t arena_mapbits_large_get(arena_chunk_t *chunk, size_t pageind); +size_t arena_mapbits_allocated_get(arena_chunk_t *chunk, size_t pageind); +void arena_mapbits_unallocated_set(arena_chunk_t *chunk, size_t pageind, + size_t size, size_t flags); +void arena_mapbits_unallocated_size_set(arena_chunk_t *chunk, size_t pageind, + size_t size); +void arena_mapbits_large_set(arena_chunk_t *chunk, size_t pageind, + size_t size, size_t flags); +void arena_mapbits_large_binind_set(arena_chunk_t *chunk, size_t pageind, + size_t binind); +void arena_mapbits_small_set(arena_chunk_t *chunk, size_t pageind, + size_t runind, size_t binind, size_t flags); +void arena_mapbits_unzeroed_set(arena_chunk_t *chunk, size_t pageind, + size_t unzeroed); +size_t arena_ptr_small_binind_get(const void *ptr, size_t mapbits); size_t arena_bin_index(arena_t *arena, arena_bin_t *bin); unsigned arena_run_regind(arena_run_t *run, arena_bin_info_t *bin_info, const void *ptr); -# ifdef JEMALLOC_PROF prof_ctx_t *arena_prof_ctx_get(const void *ptr); void arena_prof_ctx_set(const void *ptr, prof_ctx_t *ctx); -# endif -void arena_dalloc(arena_t *arena, arena_chunk_t *chunk, void *ptr); +void *arena_malloc(arena_t *arena, size_t size, bool zero, bool try_tcache); +size_t arena_salloc(const void *ptr, bool demote); +void arena_dalloc(arena_t *arena, arena_chunk_t *chunk, void *ptr, + bool try_tcache); #endif #if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_ARENA_C_)) +# ifdef JEMALLOC_ARENA_INLINE_A +JEMALLOC_INLINE arena_chunk_map_t * +arena_mapp_get(arena_chunk_t *chunk, size_t pageind) +{ + + assert(pageind >= map_bias); + assert(pageind < chunk_npages); + + return (&chunk->map[pageind-map_bias]); +} + +JEMALLOC_INLINE size_t * +arena_mapbitsp_get(arena_chunk_t *chunk, size_t pageind) +{ + + return (&arena_mapp_get(chunk, pageind)->bits); +} + +JEMALLOC_INLINE size_t +arena_mapbits_get(arena_chunk_t *chunk, size_t pageind) +{ + + return (*arena_mapbitsp_get(chunk, pageind)); +} + +JEMALLOC_INLINE size_t +arena_mapbits_unallocated_size_get(arena_chunk_t *chunk, size_t pageind) +{ + size_t mapbits; + + mapbits = arena_mapbits_get(chunk, pageind); + assert((mapbits & (CHUNK_MAP_LARGE|CHUNK_MAP_ALLOCATED)) == 0); + return (mapbits & ~PAGE_MASK); +} + +JEMALLOC_INLINE size_t +arena_mapbits_large_size_get(arena_chunk_t *chunk, size_t pageind) +{ + size_t mapbits; + + mapbits = arena_mapbits_get(chunk, pageind); + assert((mapbits & (CHUNK_MAP_LARGE|CHUNK_MAP_ALLOCATED)) == + (CHUNK_MAP_LARGE|CHUNK_MAP_ALLOCATED)); + return (mapbits & ~PAGE_MASK); +} + +JEMALLOC_INLINE size_t +arena_mapbits_small_runind_get(arena_chunk_t *chunk, size_t pageind) +{ + size_t mapbits; + + mapbits = arena_mapbits_get(chunk, pageind); + assert((mapbits & (CHUNK_MAP_LARGE|CHUNK_MAP_ALLOCATED)) == + CHUNK_MAP_ALLOCATED); + return (mapbits >> LG_PAGE); +} + +JEMALLOC_INLINE size_t +arena_mapbits_binind_get(arena_chunk_t *chunk, size_t pageind) +{ + size_t mapbits; + size_t binind; + + mapbits = arena_mapbits_get(chunk, pageind); + binind = (mapbits & CHUNK_MAP_BININD_MASK) >> CHUNK_MAP_BININD_SHIFT; + assert(binind < NBINS || binind == BININD_INVALID); + return (binind); +} + +JEMALLOC_INLINE size_t +arena_mapbits_dirty_get(arena_chunk_t *chunk, size_t pageind) +{ + size_t mapbits; + + mapbits = arena_mapbits_get(chunk, pageind); + return (mapbits & CHUNK_MAP_DIRTY); +} + +JEMALLOC_INLINE size_t +arena_mapbits_unzeroed_get(arena_chunk_t *chunk, size_t pageind) +{ + size_t mapbits; + + mapbits = arena_mapbits_get(chunk, pageind); + return (mapbits & CHUNK_MAP_UNZEROED); +} + +JEMALLOC_INLINE size_t +arena_mapbits_large_get(arena_chunk_t *chunk, size_t pageind) +{ + size_t mapbits; + + mapbits = arena_mapbits_get(chunk, pageind); + return (mapbits & CHUNK_MAP_LARGE); +} + +JEMALLOC_INLINE size_t +arena_mapbits_allocated_get(arena_chunk_t *chunk, size_t pageind) +{ + size_t mapbits; + + mapbits = arena_mapbits_get(chunk, pageind); + return (mapbits & CHUNK_MAP_ALLOCATED); +} + +JEMALLOC_INLINE void +arena_mapbits_unallocated_set(arena_chunk_t *chunk, size_t pageind, size_t size, + size_t flags) +{ + size_t *mapbitsp; + + mapbitsp = arena_mapbitsp_get(chunk, pageind); + assert((size & PAGE_MASK) == 0); + assert((flags & ~CHUNK_MAP_FLAGS_MASK) == 0); + assert((flags & (CHUNK_MAP_DIRTY|CHUNK_MAP_UNZEROED)) == flags); + *mapbitsp = size | CHUNK_MAP_BININD_INVALID | flags; +} + +JEMALLOC_INLINE void +arena_mapbits_unallocated_size_set(arena_chunk_t *chunk, size_t pageind, + size_t size) +{ + size_t *mapbitsp; + + mapbitsp = arena_mapbitsp_get(chunk, pageind); + assert((size & PAGE_MASK) == 0); + assert((*mapbitsp & (CHUNK_MAP_LARGE|CHUNK_MAP_ALLOCATED)) == 0); + *mapbitsp = size | (*mapbitsp & PAGE_MASK); +} + +JEMALLOC_INLINE void +arena_mapbits_large_set(arena_chunk_t *chunk, size_t pageind, size_t size, + size_t flags) +{ + size_t *mapbitsp; + size_t unzeroed; + + mapbitsp = arena_mapbitsp_get(chunk, pageind); + assert((size & PAGE_MASK) == 0); + assert((flags & CHUNK_MAP_DIRTY) == flags); + unzeroed = *mapbitsp & CHUNK_MAP_UNZEROED; /* Preserve unzeroed. */ + *mapbitsp = size | CHUNK_MAP_BININD_INVALID | flags | unzeroed | + CHUNK_MAP_LARGE | CHUNK_MAP_ALLOCATED; +} + +JEMALLOC_INLINE void +arena_mapbits_large_binind_set(arena_chunk_t *chunk, size_t pageind, + size_t binind) +{ + size_t *mapbitsp; + + assert(binind <= BININD_INVALID); + mapbitsp = arena_mapbitsp_get(chunk, pageind); + assert(arena_mapbits_large_size_get(chunk, pageind) == PAGE); + *mapbitsp = (*mapbitsp & ~CHUNK_MAP_BININD_MASK) | (binind << + CHUNK_MAP_BININD_SHIFT); +} + +JEMALLOC_INLINE void +arena_mapbits_small_set(arena_chunk_t *chunk, size_t pageind, size_t runind, + size_t binind, size_t flags) +{ + size_t *mapbitsp; + size_t unzeroed; + + assert(binind < BININD_INVALID); + mapbitsp = arena_mapbitsp_get(chunk, pageind); + assert(pageind - runind >= map_bias); + assert((flags & CHUNK_MAP_DIRTY) == flags); + unzeroed = *mapbitsp & CHUNK_MAP_UNZEROED; /* Preserve unzeroed. */ + *mapbitsp = (runind << LG_PAGE) | (binind << CHUNK_MAP_BININD_SHIFT) | + flags | unzeroed | CHUNK_MAP_ALLOCATED; +} + +JEMALLOC_INLINE void +arena_mapbits_unzeroed_set(arena_chunk_t *chunk, size_t pageind, + size_t unzeroed) +{ + size_t *mapbitsp; + + mapbitsp = arena_mapbitsp_get(chunk, pageind); + *mapbitsp = (*mapbitsp & ~CHUNK_MAP_UNZEROED) | unzeroed; +} + +JEMALLOC_INLINE size_t +arena_ptr_small_binind_get(const void *ptr, size_t mapbits) +{ + size_t binind; + + binind = (mapbits & CHUNK_MAP_BININD_MASK) >> CHUNK_MAP_BININD_SHIFT; + + if (config_debug) { + arena_chunk_t *chunk; + arena_t *arena; + size_t pageind; + size_t actual_mapbits; + arena_run_t *run; + arena_bin_t *bin; + size_t actual_binind; + arena_bin_info_t *bin_info; + + assert(binind != BININD_INVALID); + assert(binind < NBINS); + chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr); + arena = chunk->arena; + pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE; + actual_mapbits = arena_mapbits_get(chunk, pageind); + assert(mapbits == actual_mapbits); + assert(arena_mapbits_large_get(chunk, pageind) == 0); + assert(arena_mapbits_allocated_get(chunk, pageind) != 0); + run = (arena_run_t *)((uintptr_t)chunk + (uintptr_t)((pageind - + (actual_mapbits >> LG_PAGE)) << LG_PAGE)); + bin = run->bin; + actual_binind = bin - arena->bins; + assert(binind == actual_binind); + bin_info = &arena_bin_info[actual_binind]; + assert(((uintptr_t)ptr - ((uintptr_t)run + + (uintptr_t)bin_info->reg0_offset)) % bin_info->reg_interval + == 0); + } + + return (binind); +} +# endif /* JEMALLOC_ARENA_INLINE_A */ + +# ifdef JEMALLOC_ARENA_INLINE_B JEMALLOC_INLINE size_t arena_bin_index(arena_t *arena, arena_bin_t *bin) { size_t binind = bin - arena->bins; - assert(binind < nbins); + assert(binind < NBINS); return (binind); } @@ -519,9 +715,8 @@ JEMALLOC_INLINE unsigned arena_run_regind(arena_run_t *run, arena_bin_info_t *bin_info, const void *ptr) { unsigned shift, diff, regind; - size_t size; + size_t interval; - dassert(run->magic == ARENA_RUN_MAGIC); /* * Freeing a pointer lower than region zero can cause assertion * failure. @@ -537,12 +732,12 @@ arena_run_regind(arena_run_t *run, arena_bin_info_t *bin_info, const void *ptr) bin_info->reg0_offset); /* Rescale (factor powers of 2 out of the numerator and denominator). */ - size = bin_info->reg_size; - shift = ffs(size) - 1; + interval = bin_info->reg_interval; + shift = ffs(interval) - 1; diff >>= shift; - size >>= shift; + interval >>= shift; - if (size == 1) { + if (interval == 1) { /* The divisor was a power of 2. */ regind = diff; } else { @@ -554,7 +749,7 @@ arena_run_regind(arena_run_t *run, arena_bin_info_t *bin_info, const void *ptr) * * becomes * - * (X * size_invs[D - 3]) >> SIZE_INV_SHIFT + * (X * interval_invs[D - 3]) >> SIZE_INV_SHIFT * * We can omit the first three elements, because we never * divide by 0, and 1 and 2 are both powers of two, which are @@ -562,7 +757,7 @@ arena_run_regind(arena_run_t *run, arena_bin_info_t *bin_info, const void *ptr) */ #define SIZE_INV_SHIFT ((sizeof(unsigned) << 3) - LG_RUN_MAXREGS) #define SIZE_INV(s) (((1U << SIZE_INV_SHIFT) / (s)) + 1) - static const unsigned size_invs[] = { + static const unsigned interval_invs[] = { SIZE_INV(3), SIZE_INV(4), SIZE_INV(5), SIZE_INV(6), SIZE_INV(7), SIZE_INV(8), SIZE_INV(9), SIZE_INV(10), SIZE_INV(11), @@ -573,20 +768,21 @@ arena_run_regind(arena_run_t *run, arena_bin_info_t *bin_info, const void *ptr) SIZE_INV(28), SIZE_INV(29), SIZE_INV(30), SIZE_INV(31) }; - if (size <= ((sizeof(size_invs) / sizeof(unsigned)) + 2)) - regind = (diff * size_invs[size - 3]) >> SIZE_INV_SHIFT; - else - regind = diff / size; + if (interval <= ((sizeof(interval_invs) / sizeof(unsigned)) + + 2)) { + regind = (diff * interval_invs[interval - 3]) >> + SIZE_INV_SHIFT; + } else + regind = diff / interval; #undef SIZE_INV #undef SIZE_INV_SHIFT } - assert(diff == regind * size); + assert(diff == regind * interval); assert(regind < bin_info->nregs); return (regind); } -#ifdef JEMALLOC_PROF JEMALLOC_INLINE prof_ctx_t * arena_prof_ctx_get(const void *ptr) { @@ -594,32 +790,33 @@ arena_prof_ctx_get(const void *ptr) arena_chunk_t *chunk; size_t pageind, mapbits; + cassert(config_prof); assert(ptr != NULL); assert(CHUNK_ADDR2BASE(ptr) != ptr); chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr); - pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> PAGE_SHIFT; - mapbits = chunk->map[pageind-map_bias].bits; + pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE; + mapbits = arena_mapbits_get(chunk, pageind); assert((mapbits & CHUNK_MAP_ALLOCATED) != 0); if ((mapbits & CHUNK_MAP_LARGE) == 0) { if (prof_promote) ret = (prof_ctx_t *)(uintptr_t)1U; else { arena_run_t *run = (arena_run_t *)((uintptr_t)chunk + - (uintptr_t)((pageind - (mapbits >> PAGE_SHIFT)) << - PAGE_SHIFT)); - size_t binind = arena_bin_index(chunk->arena, run->bin); + (uintptr_t)((pageind - (mapbits >> LG_PAGE)) << + LG_PAGE)); + size_t binind = arena_ptr_small_binind_get(ptr, + mapbits); arena_bin_info_t *bin_info = &arena_bin_info[binind]; unsigned regind; - dassert(run->magic == ARENA_RUN_MAGIC); regind = arena_run_regind(run, bin_info, ptr); ret = *(prof_ctx_t **)((uintptr_t)run + bin_info->ctx0_offset + (regind * sizeof(prof_ctx_t *))); } } else - ret = chunk->map[pageind-map_bias].prof_ctx; + ret = arena_mapp_get(chunk, pageind)->prof_ctx; return (ret); } @@ -630,25 +827,24 @@ arena_prof_ctx_set(const void *ptr, prof_ctx_t *ctx) arena_chunk_t *chunk; size_t pageind, mapbits; + cassert(config_prof); assert(ptr != NULL); assert(CHUNK_ADDR2BASE(ptr) != ptr); chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr); - pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> PAGE_SHIFT; - mapbits = chunk->map[pageind-map_bias].bits; + pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE; + mapbits = arena_mapbits_get(chunk, pageind); assert((mapbits & CHUNK_MAP_ALLOCATED) != 0); if ((mapbits & CHUNK_MAP_LARGE) == 0) { if (prof_promote == false) { arena_run_t *run = (arena_run_t *)((uintptr_t)chunk + - (uintptr_t)((pageind - (mapbits >> PAGE_SHIFT)) << - PAGE_SHIFT)); - arena_bin_t *bin = run->bin; + (uintptr_t)((pageind - (mapbits >> LG_PAGE)) << + LG_PAGE)); size_t binind; arena_bin_info_t *bin_info; unsigned regind; - dassert(run->magic == ARENA_RUN_MAGIC); - binind = arena_bin_index(chunk->arena, bin); + binind = arena_ptr_small_binind_get(ptr, mapbits); bin_info = &arena_bin_info[binind]; regind = arena_run_regind(run, bin_info, ptr); @@ -657,86 +853,122 @@ arena_prof_ctx_set(const void *ptr, prof_ctx_t *ctx) } else assert((uintptr_t)ctx == (uintptr_t)1U); } else - chunk->map[pageind-map_bias].prof_ctx = ctx; + arena_mapp_get(chunk, pageind)->prof_ctx = ctx; +} + +JEMALLOC_INLINE void * +arena_malloc(arena_t *arena, size_t size, bool zero, bool try_tcache) +{ + tcache_t *tcache; + + assert(size != 0); + assert(size <= arena_maxclass); + + if (size <= SMALL_MAXCLASS) { + if (try_tcache && (tcache = tcache_get(true)) != NULL) + return (tcache_alloc_small(tcache, size, zero)); + else { + return (arena_malloc_small(choose_arena(arena), size, + zero)); + } + } else { + /* + * Initialize tcache after checking size in order to avoid + * infinite recursion during tcache initialization. + */ + if (try_tcache && size <= tcache_maxclass && (tcache = + tcache_get(true)) != NULL) + return (tcache_alloc_large(tcache, size, zero)); + else { + return (arena_malloc_large(choose_arena(arena), size, + zero)); + } + } +} + +/* Return the size of the allocation pointed to by ptr. */ +JEMALLOC_INLINE size_t +arena_salloc(const void *ptr, bool demote) +{ + size_t ret; + arena_chunk_t *chunk; + size_t pageind, binind; + + assert(ptr != NULL); + assert(CHUNK_ADDR2BASE(ptr) != ptr); + + chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr); + pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE; + assert(arena_mapbits_allocated_get(chunk, pageind) != 0); + binind = arena_mapbits_binind_get(chunk, pageind); + if (binind == BININD_INVALID || (config_prof && demote == false && + prof_promote && arena_mapbits_large_get(chunk, pageind) != 0)) { + /* + * Large allocation. In the common case (demote == true), and + * as this is an inline function, most callers will only end up + * looking at binind to determine that ptr is a small + * allocation. + */ + assert(((uintptr_t)ptr & PAGE_MASK) == 0); + ret = arena_mapbits_large_size_get(chunk, pageind); + assert(ret != 0); + assert(pageind + (ret>>LG_PAGE) <= chunk_npages); + assert(ret == PAGE || arena_mapbits_large_size_get(chunk, + pageind+(ret>>LG_PAGE)-1) == 0); + assert(binind == arena_mapbits_binind_get(chunk, + pageind+(ret>>LG_PAGE)-1)); + assert(arena_mapbits_dirty_get(chunk, pageind) == + arena_mapbits_dirty_get(chunk, pageind+(ret>>LG_PAGE)-1)); + } else { + /* + * Small allocation (possibly promoted to a large object due to + * prof_promote). + */ + assert(arena_mapbits_large_get(chunk, pageind) != 0 || + arena_ptr_small_binind_get(ptr, arena_mapbits_get(chunk, + pageind)) == binind); + ret = arena_bin_info[binind].reg_size; + } + + return (ret); } -#endif JEMALLOC_INLINE void -arena_dalloc(arena_t *arena, arena_chunk_t *chunk, void *ptr) +arena_dalloc(arena_t *arena, arena_chunk_t *chunk, void *ptr, bool try_tcache) { - size_t pageind; - arena_chunk_map_t *mapelm; + size_t pageind, mapbits; + tcache_t *tcache; assert(arena != NULL); - dassert(arena->magic == ARENA_MAGIC); assert(chunk->arena == arena); assert(ptr != NULL); assert(CHUNK_ADDR2BASE(ptr) != ptr); - pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> PAGE_SHIFT; - mapelm = &chunk->map[pageind-map_bias]; - assert((mapelm->bits & CHUNK_MAP_ALLOCATED) != 0); - if ((mapelm->bits & CHUNK_MAP_LARGE) == 0) { + pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE; + mapbits = arena_mapbits_get(chunk, pageind); + assert(arena_mapbits_allocated_get(chunk, pageind) != 0); + if ((mapbits & CHUNK_MAP_LARGE) == 0) { /* Small allocation. */ -#ifdef JEMALLOC_TCACHE - tcache_t *tcache; + if (try_tcache && (tcache = tcache_get(false)) != NULL) { + size_t binind; - if ((tcache = tcache_get()) != NULL) - tcache_dalloc_small(tcache, ptr); - else { -#endif - arena_run_t *run; - arena_bin_t *bin; - - run = (arena_run_t *)((uintptr_t)chunk + - (uintptr_t)((pageind - (mapelm->bits >> - PAGE_SHIFT)) << PAGE_SHIFT)); - dassert(run->magic == ARENA_RUN_MAGIC); - bin = run->bin; -#ifdef JEMALLOC_DEBUG - { - size_t binind = arena_bin_index(arena, bin); - arena_bin_info_t *bin_info = - &arena_bin_info[binind]; - assert(((uintptr_t)ptr - ((uintptr_t)run + - (uintptr_t)bin_info->reg0_offset)) % - bin_info->reg_size == 0); - } -#endif - malloc_mutex_lock(&bin->lock); - arena_dalloc_bin(arena, chunk, ptr, mapelm); - malloc_mutex_unlock(&bin->lock); -#ifdef JEMALLOC_TCACHE - } -#endif + binind = arena_ptr_small_binind_get(ptr, mapbits); + tcache_dalloc_small(tcache, ptr, binind); + } else + arena_dalloc_small(arena, chunk, ptr, pageind); } else { -#ifdef JEMALLOC_TCACHE - size_t size = mapelm->bits & ~PAGE_MASK; + size_t size = arena_mapbits_large_size_get(chunk, pageind); assert(((uintptr_t)ptr & PAGE_MASK) == 0); - if (size <= tcache_maxclass) { - tcache_t *tcache; - if ((tcache = tcache_get()) != NULL) - tcache_dalloc_large(tcache, ptr, size); - else { - malloc_mutex_lock(&arena->lock); - arena_dalloc_large(arena, chunk, ptr); - malloc_mutex_unlock(&arena->lock); - } - } else { - malloc_mutex_lock(&arena->lock); + if (try_tcache && size <= tcache_maxclass && (tcache = + tcache_get(false)) != NULL) { + tcache_dalloc_large(tcache, ptr, size); + } else arena_dalloc_large(arena, chunk, ptr); - malloc_mutex_unlock(&arena->lock); - } -#else - assert(((uintptr_t)ptr & PAGE_MASK) == 0); - malloc_mutex_lock(&arena->lock); - arena_dalloc_large(arena, chunk, ptr); - malloc_mutex_unlock(&arena->lock); -#endif } } +# endif /* JEMALLOC_ARENA_INLINE_B */ #endif #endif /* JEMALLOC_H_INLINES */ diff --git a/include/jemalloc/internal/atomic.h b/include/jemalloc/internal/atomic.h index 9a298623..11a7b47f 100644 --- a/include/jemalloc/internal/atomic.h +++ b/include/jemalloc/internal/atomic.h @@ -11,22 +11,8 @@ #define atomic_read_uint64(p) atomic_add_uint64(p, 0) #define atomic_read_uint32(p) atomic_add_uint32(p, 0) - -#if (LG_SIZEOF_PTR == 3) -# define atomic_read_z(p) \ - (size_t)atomic_add_uint64((uint64_t *)p, (uint64_t)0) -# define atomic_add_z(p, x) \ - (size_t)atomic_add_uint64((uint64_t *)p, (uint64_t)x) -# define atomic_sub_z(p, x) \ - (size_t)atomic_sub_uint64((uint64_t *)p, (uint64_t)x) -#elif (LG_SIZEOF_PTR == 2) -# define atomic_read_z(p) \ - (size_t)atomic_add_uint32((uint32_t *)p, (uint32_t)0) -# define atomic_add_z(p, x) \ - (size_t)atomic_add_uint32((uint32_t *)p, (uint32_t)x) -# define atomic_sub_z(p, x) \ - (size_t)atomic_sub_uint32((uint32_t *)p, (uint32_t)x) -#endif +#define atomic_read_z(p) atomic_add_z(p, 0) +#define atomic_read_u(p) atomic_add_u(p, 0) #endif /* JEMALLOC_H_EXTERNS */ /******************************************************************************/ @@ -37,12 +23,17 @@ uint64_t atomic_add_uint64(uint64_t *p, uint64_t x); uint64_t atomic_sub_uint64(uint64_t *p, uint64_t x); uint32_t atomic_add_uint32(uint32_t *p, uint32_t x); uint32_t atomic_sub_uint32(uint32_t *p, uint32_t x); +size_t atomic_add_z(size_t *p, size_t x); +size_t atomic_sub_z(size_t *p, size_t x); +unsigned atomic_add_u(unsigned *p, unsigned x); +unsigned atomic_sub_u(unsigned *p, unsigned x); #endif #if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_ATOMIC_C_)) /******************************************************************************/ /* 64-bit operations. */ -#ifdef __GCC_HAVE_SYNC_COMPARE_AND_SWAP_8 +#if (LG_SIZEOF_PTR == 3 || LG_SIZEOF_INT == 3) +# ifdef __GCC_HAVE_SYNC_COMPARE_AND_SWAP_8 JEMALLOC_INLINE uint64_t atomic_add_uint64(uint64_t *p, uint64_t x) { @@ -56,6 +47,20 @@ atomic_sub_uint64(uint64_t *p, uint64_t x) return (__sync_sub_and_fetch(p, x)); } +#elif (defined(_MSC_VER)) +JEMALLOC_INLINE uint64_t +atomic_add_uint64(uint64_t *p, uint64_t x) +{ + + return (InterlockedExchangeAdd64(p, x)); +} + +JEMALLOC_INLINE uint64_t +atomic_sub_uint64(uint64_t *p, uint64_t x) +{ + + return (InterlockedExchangeAdd64(p, -((int64_t)x))); +} #elif (defined(JEMALLOC_OSATOMIC)) JEMALLOC_INLINE uint64_t atomic_add_uint64(uint64_t *p, uint64_t x) @@ -70,7 +75,7 @@ atomic_sub_uint64(uint64_t *p, uint64_t x) return (OSAtomicAdd64(-((int64_t)x), (int64_t *)p)); } -#elif (defined(__amd64_) || defined(__x86_64__)) +# elif (defined(__amd64__) || defined(__x86_64__)) JEMALLOC_INLINE uint64_t atomic_add_uint64(uint64_t *p, uint64_t x) { @@ -97,8 +102,43 @@ atomic_sub_uint64(uint64_t *p, uint64_t x) return (x); } -#else -# if (LG_SIZEOF_PTR == 3) +# elif (defined(JEMALLOC_ATOMIC9)) +JEMALLOC_INLINE uint64_t +atomic_add_uint64(uint64_t *p, uint64_t x) +{ + + /* + * atomic_fetchadd_64() doesn't exist, but we only ever use this + * function on LP64 systems, so atomic_fetchadd_long() will do. + */ + assert(sizeof(uint64_t) == sizeof(unsigned long)); + + return (atomic_fetchadd_long(p, (unsigned long)x) + x); +} + +JEMALLOC_INLINE uint64_t +atomic_sub_uint64(uint64_t *p, uint64_t x) +{ + + assert(sizeof(uint64_t) == sizeof(unsigned long)); + + return (atomic_fetchadd_long(p, (unsigned long)(-(long)x)) - x); +} +# elif (defined(JE_FORCE_SYNC_COMPARE_AND_SWAP_8)) +JEMALLOC_INLINE uint64_t +atomic_add_uint64(uint64_t *p, uint64_t x) +{ + + return (__sync_add_and_fetch(p, x)); +} + +JEMALLOC_INLINE uint64_t +atomic_sub_uint64(uint64_t *p, uint64_t x) +{ + + return (__sync_sub_and_fetch(p, x)); +} +# else # error "Missing implementation for 64-bit atomic operations" # endif #endif @@ -119,6 +159,20 @@ atomic_sub_uint32(uint32_t *p, uint32_t x) return (__sync_sub_and_fetch(p, x)); } +#elif (defined(_MSC_VER)) +JEMALLOC_INLINE uint32_t +atomic_add_uint32(uint32_t *p, uint32_t x) +{ + + return (InterlockedExchangeAdd(p, x)); +} + +JEMALLOC_INLINE uint32_t +atomic_sub_uint32(uint32_t *p, uint32_t x) +{ + + return (InterlockedExchangeAdd(p, -((int32_t)x))); +} #elif (defined(JEMALLOC_OSATOMIC)) JEMALLOC_INLINE uint32_t atomic_add_uint32(uint32_t *p, uint32_t x) @@ -133,7 +187,7 @@ atomic_sub_uint32(uint32_t *p, uint32_t x) return (OSAtomicAdd32(-((int32_t)x), (int32_t *)p)); } -#elif (defined(__i386__) || defined(__amd64_) || defined(__x86_64__)) +#elif (defined(__i386__) || defined(__amd64__) || defined(__x86_64__)) JEMALLOC_INLINE uint32_t atomic_add_uint32(uint32_t *p, uint32_t x) { @@ -160,9 +214,90 @@ atomic_sub_uint32(uint32_t *p, uint32_t x) return (x); } +#elif (defined(JEMALLOC_ATOMIC9)) +JEMALLOC_INLINE uint32_t +atomic_add_uint32(uint32_t *p, uint32_t x) +{ + + return (atomic_fetchadd_32(p, x) + x); +} + +JEMALLOC_INLINE uint32_t +atomic_sub_uint32(uint32_t *p, uint32_t x) +{ + + return (atomic_fetchadd_32(p, (uint32_t)(-(int32_t)x)) - x); +} +#elif (defined(JE_FORCE_SYNC_COMPARE_AND_SWAP_4)) +JEMALLOC_INLINE uint32_t +atomic_add_uint32(uint32_t *p, uint32_t x) +{ + + return (__sync_add_and_fetch(p, x)); +} + +JEMALLOC_INLINE uint32_t +atomic_sub_uint32(uint32_t *p, uint32_t x) +{ + + return (__sync_sub_and_fetch(p, x)); +} #else # error "Missing implementation for 32-bit atomic operations" #endif + +/******************************************************************************/ +/* size_t operations. */ +JEMALLOC_INLINE size_t +atomic_add_z(size_t *p, size_t x) +{ + +#if (LG_SIZEOF_PTR == 3) + return ((size_t)atomic_add_uint64((uint64_t *)p, (uint64_t)x)); +#elif (LG_SIZEOF_PTR == 2) + return ((size_t)atomic_add_uint32((uint32_t *)p, (uint32_t)x)); +#endif +} + +JEMALLOC_INLINE size_t +atomic_sub_z(size_t *p, size_t x) +{ + +#if (LG_SIZEOF_PTR == 3) + return ((size_t)atomic_add_uint64((uint64_t *)p, + (uint64_t)-((int64_t)x))); +#elif (LG_SIZEOF_PTR == 2) + return ((size_t)atomic_add_uint32((uint32_t *)p, + (uint32_t)-((int32_t)x))); +#endif +} + +/******************************************************************************/ +/* unsigned operations. */ +JEMALLOC_INLINE unsigned +atomic_add_u(unsigned *p, unsigned x) +{ + +#if (LG_SIZEOF_INT == 3) + return ((unsigned)atomic_add_uint64((uint64_t *)p, (uint64_t)x)); +#elif (LG_SIZEOF_INT == 2) + return ((unsigned)atomic_add_uint32((uint32_t *)p, (uint32_t)x)); +#endif +} + +JEMALLOC_INLINE unsigned +atomic_sub_u(unsigned *p, unsigned x) +{ + +#if (LG_SIZEOF_INT == 3) + return ((unsigned)atomic_add_uint64((uint64_t *)p, + (uint64_t)-((int64_t)x))); +#elif (LG_SIZEOF_INT == 2) + return ((unsigned)atomic_add_uint32((uint32_t *)p, + (uint32_t)-((int32_t)x))); +#endif +} +/******************************************************************************/ #endif #endif /* JEMALLOC_H_INLINES */ diff --git a/include/jemalloc/internal/base.h b/include/jemalloc/internal/base.h index e353f309..9cf75ffb 100644 --- a/include/jemalloc/internal/base.h +++ b/include/jemalloc/internal/base.h @@ -9,12 +9,14 @@ /******************************************************************************/ #ifdef JEMALLOC_H_EXTERNS -extern malloc_mutex_t base_mtx; - void *base_alloc(size_t size); +void *base_calloc(size_t number, size_t size); extent_node_t *base_node_alloc(void); void base_node_dealloc(extent_node_t *node); bool base_boot(void); +void base_prefork(void); +void base_postfork_parent(void); +void base_postfork_child(void); #endif /* JEMALLOC_H_EXTERNS */ /******************************************************************************/ diff --git a/include/jemalloc/internal/chunk.h b/include/jemalloc/internal/chunk.h index 54b6a3ec..8fb1fe6d 100644 --- a/include/jemalloc/internal/chunk.h +++ b/include/jemalloc/internal/chunk.h @@ -28,20 +28,13 @@ #ifdef JEMALLOC_H_EXTERNS extern size_t opt_lg_chunk; -#ifdef JEMALLOC_SWAP -extern bool opt_overcommit; -#endif -#if (defined(JEMALLOC_STATS) || defined(JEMALLOC_PROF)) /* Protects stats_chunks; currently not used for any other purpose. */ extern malloc_mutex_t chunks_mtx; /* Chunk statistics. */ extern chunk_stats_t stats_chunks; -#endif -#ifdef JEMALLOC_IVSALLOC extern rtree_t *chunks_rtree; -#endif extern size_t chunksize; extern size_t chunksize_mask; /* (chunksize - 1). */ @@ -49,7 +42,7 @@ extern size_t chunk_npages; extern size_t map_bias; /* Number of arena chunk header pages. */ extern size_t arena_maxclass; /* Max size class for arenas. */ -void *chunk_alloc(size_t size, bool base, bool *zero); +void *chunk_alloc(size_t size, size_t alignment, bool base, bool *zero); void chunk_dealloc(void *chunk, size_t size, bool unmap); bool chunk_boot(void); @@ -60,6 +53,5 @@ bool chunk_boot(void); #endif /* JEMALLOC_H_INLINES */ /******************************************************************************/ -#include "jemalloc/internal/chunk_swap.h" #include "jemalloc/internal/chunk_dss.h" #include "jemalloc/internal/chunk_mmap.h" diff --git a/include/jemalloc/internal/chunk_dss.h b/include/jemalloc/internal/chunk_dss.h index 6f005222..6e2643b2 100644 --- a/include/jemalloc/internal/chunk_dss.h +++ b/include/jemalloc/internal/chunk_dss.h @@ -1,4 +1,3 @@ -#ifdef JEMALLOC_DSS /******************************************************************************/ #ifdef JEMALLOC_H_TYPES @@ -10,16 +9,12 @@ /******************************************************************************/ #ifdef JEMALLOC_H_EXTERNS -/* - * Protects sbrk() calls. This avoids malloc races among threads, though it - * does not protect against races with threads that call sbrk() directly. - */ -extern malloc_mutex_t dss_mtx; - -void *chunk_alloc_dss(size_t size, bool *zero); +void *chunk_alloc_dss(size_t size, size_t alignment, bool *zero); bool chunk_in_dss(void *chunk); -bool chunk_dealloc_dss(void *chunk, size_t size); bool chunk_dss_boot(void); +void chunk_dss_prefork(void); +void chunk_dss_postfork_parent(void); +void chunk_dss_postfork_child(void); #endif /* JEMALLOC_H_EXTERNS */ /******************************************************************************/ @@ -27,4 +22,3 @@ bool chunk_dss_boot(void); #endif /* JEMALLOC_H_INLINES */ /******************************************************************************/ -#endif /* JEMALLOC_DSS */ diff --git a/include/jemalloc/internal/chunk_mmap.h b/include/jemalloc/internal/chunk_mmap.h index 07b50a4d..b29f39e9 100644 --- a/include/jemalloc/internal/chunk_mmap.h +++ b/include/jemalloc/internal/chunk_mmap.h @@ -9,11 +9,10 @@ /******************************************************************************/ #ifdef JEMALLOC_H_EXTERNS -void *chunk_alloc_mmap(size_t size); -void *chunk_alloc_mmap_noreserve(size_t size); -void chunk_dealloc_mmap(void *chunk, size_t size); +void pages_purge(void *addr, size_t length); -bool chunk_mmap_boot(void); +void *chunk_alloc_mmap(size_t size, size_t alignment, bool *zero); +bool chunk_dealloc_mmap(void *chunk, size_t size); #endif /* JEMALLOC_H_EXTERNS */ /******************************************************************************/ diff --git a/include/jemalloc/internal/chunk_swap.h b/include/jemalloc/internal/chunk_swap.h deleted file mode 100644 index 9faa739f..00000000 --- a/include/jemalloc/internal/chunk_swap.h +++ /dev/null @@ -1,34 +0,0 @@ -#ifdef JEMALLOC_SWAP -/******************************************************************************/ -#ifdef JEMALLOC_H_TYPES - -#endif /* JEMALLOC_H_TYPES */ -/******************************************************************************/ -#ifdef JEMALLOC_H_STRUCTS - -#endif /* JEMALLOC_H_STRUCTS */ -/******************************************************************************/ -#ifdef JEMALLOC_H_EXTERNS - -extern malloc_mutex_t swap_mtx; -extern bool swap_enabled; -extern bool swap_prezeroed; -extern size_t swap_nfds; -extern int *swap_fds; -#ifdef JEMALLOC_STATS -extern size_t swap_avail; -#endif - -void *chunk_alloc_swap(size_t size, bool *zero); -bool chunk_in_swap(void *chunk); -bool chunk_dealloc_swap(void *chunk, size_t size); -bool chunk_swap_enable(const int *fds, unsigned nfds, bool prezeroed); -bool chunk_swap_boot(void); - -#endif /* JEMALLOC_H_EXTERNS */ -/******************************************************************************/ -#ifdef JEMALLOC_H_INLINES - -#endif /* JEMALLOC_H_INLINES */ -/******************************************************************************/ -#endif /* JEMALLOC_SWAP */ diff --git a/include/jemalloc/internal/ckh.h b/include/jemalloc/internal/ckh.h index 3e4ad4c8..05d1fc03 100644 --- a/include/jemalloc/internal/ckh.h +++ b/include/jemalloc/internal/ckh.h @@ -30,11 +30,6 @@ struct ckhc_s { }; struct ckh_s { -#ifdef JEMALLOC_DEBUG -#define CKH_MAGIC 0x3af2489d - uint32_t magic; -#endif - #ifdef CKH_COUNT /* Counters used to get an idea of performance. */ uint64_t ngrows; @@ -47,7 +42,7 @@ struct ckh_s { /* Used for pseudo-random number generation. */ #define CKH_A 1103515241 #define CKH_C 12347 - uint32_t prn_state; + uint32_t prng_state; /* Total number of items. */ size_t count; diff --git a/include/jemalloc/internal/ctl.h b/include/jemalloc/internal/ctl.h index f1f5eb70..adf3827f 100644 --- a/include/jemalloc/internal/ctl.h +++ b/include/jemalloc/internal/ctl.h @@ -2,6 +2,8 @@ #ifdef JEMALLOC_H_TYPES typedef struct ctl_node_s ctl_node_t; +typedef struct ctl_named_node_s ctl_named_node_t; +typedef struct ctl_indexed_node_s ctl_indexed_node_t; typedef struct ctl_arena_stats_s ctl_arena_stats_t; typedef struct ctl_stats_s ctl_stats_t; @@ -11,20 +13,21 @@ typedef struct ctl_stats_s ctl_stats_t; struct ctl_node_s { bool named; - union { - struct { - const char *name; - /* If (nchildren == 0), this is a terminal node. */ - unsigned nchildren; - const ctl_node_t *children; - } named; - struct { - const ctl_node_t *(*index)(const size_t *, size_t, - size_t); - } indexed; - } u; - int (*ctl)(const size_t *, size_t, void *, size_t *, void *, - size_t); +}; + +struct ctl_named_node_s { + struct ctl_node_s node; + const char *name; + /* If (nchildren == 0), this is a terminal node. */ + unsigned nchildren; + const ctl_node_t *children; + int (*ctl)(const size_t *, size_t, void *, size_t *, + void *, size_t); +}; + +struct ctl_indexed_node_s { + struct ctl_node_s node; + const ctl_named_node_t *(*index)(const size_t *, size_t, size_t); }; struct ctl_arena_stats_s { @@ -32,7 +35,6 @@ struct ctl_arena_stats_s { unsigned nthreads; size_t pactive; size_t pdirty; -#ifdef JEMALLOC_STATS arena_stats_t astats; /* Aggregate stats for small size classes, based on bin stats. */ @@ -41,13 +43,11 @@ struct ctl_arena_stats_s { uint64_t ndalloc_small; uint64_t nrequests_small; - malloc_bin_stats_t *bstats; /* nbins elements. */ + malloc_bin_stats_t bstats[NBINS]; malloc_large_stats_t *lstats; /* nlclasses elements. */ -#endif }; struct ctl_stats_s { -#ifdef JEMALLOC_STATS size_t allocated; size_t active; size_t mapped; @@ -61,11 +61,7 @@ struct ctl_stats_s { uint64_t nmalloc; /* huge_nmalloc */ uint64_t ndalloc; /* huge_ndalloc */ } huge; -#endif ctl_arena_stats_t *arenas; /* (narenas + 1) elements. */ -#ifdef JEMALLOC_SWAP - size_t swap_avail; -#endif }; #endif /* JEMALLOC_H_STRUCTS */ @@ -81,27 +77,25 @@ int ctl_bymib(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp, bool ctl_boot(void); #define xmallctl(name, oldp, oldlenp, newp, newlen) do { \ - if (JEMALLOC_P(mallctl)(name, oldp, oldlenp, newp, newlen) \ + if (je_mallctl(name, oldp, oldlenp, newp, newlen) \ != 0) { \ - malloc_write(": Failure in xmallctl(\""); \ - malloc_write(name); \ - malloc_write("\", ...)\n"); \ + malloc_printf( \ + ": Failure in xmallctl(\"%s\", ...)\n", \ + name); \ abort(); \ } \ } while (0) #define xmallctlnametomib(name, mibp, miblenp) do { \ - if (JEMALLOC_P(mallctlnametomib)(name, mibp, miblenp) != 0) { \ - malloc_write( \ - ": Failure in xmallctlnametomib(\""); \ - malloc_write(name); \ - malloc_write("\", ...)\n"); \ + if (je_mallctlnametomib(name, mibp, miblenp) != 0) { \ + malloc_printf(": Failure in " \ + "xmallctlnametomib(\"%s\", ...)\n", name); \ abort(); \ } \ } while (0) #define xmallctlbymib(mib, miblen, oldp, oldlenp, newp, newlen) do { \ - if (JEMALLOC_P(mallctlbymib)(mib, miblen, oldp, oldlenp, newp, \ + if (je_mallctlbymib(mib, miblen, oldp, oldlenp, newp, \ newlen) != 0) { \ malloc_write( \ ": Failure in xmallctlbymib()\n"); \ diff --git a/include/jemalloc/internal/extent.h b/include/jemalloc/internal/extent.h index 6fe9702b..36af8be8 100644 --- a/include/jemalloc/internal/extent.h +++ b/include/jemalloc/internal/extent.h @@ -9,18 +9,14 @@ typedef struct extent_node_s extent_node_t; /* Tree of extents. */ struct extent_node_s { -#if (defined(JEMALLOC_SWAP) || defined(JEMALLOC_DSS)) /* Linkage for the size/address-ordered tree. */ rb_node(extent_node_t) link_szad; -#endif /* Linkage for the address-ordered tree. */ rb_node(extent_node_t) link_ad; -#ifdef JEMALLOC_PROF /* Profile counters, used for huge objects. */ prof_ctx_t *prof_ctx; -#endif /* Pointer to the extent that this tree node is responsible for. */ void *addr; @@ -34,9 +30,7 @@ typedef rb_tree(extent_node_t) extent_tree_t; /******************************************************************************/ #ifdef JEMALLOC_H_EXTERNS -#if (defined(JEMALLOC_SWAP) || defined(JEMALLOC_DSS)) rb_proto(, extent_tree_szad_, extent_tree_t, extent_node_t) -#endif rb_proto(, extent_tree_ad_, extent_tree_t, extent_node_t) diff --git a/include/jemalloc/internal/hash.h b/include/jemalloc/internal/hash.h index 8a46ce30..2f501f5d 100644 --- a/include/jemalloc/internal/hash.h +++ b/include/jemalloc/internal/hash.h @@ -26,7 +26,7 @@ uint64_t hash(const void *key, size_t len, uint64_t seed); JEMALLOC_INLINE uint64_t hash(const void *key, size_t len, uint64_t seed) { - const uint64_t m = 0xc6a4a7935bd1e995LLU; + const uint64_t m = UINT64_C(0xc6a4a7935bd1e995); const int r = 47; uint64_t h = seed ^ (len * m); const uint64_t *data = (const uint64_t *)key; @@ -48,14 +48,14 @@ hash(const void *key, size_t len, uint64_t seed) data2 = (const unsigned char *)data; switch(len & 7) { - case 7: h ^= ((uint64_t)(data2[6])) << 48; - case 6: h ^= ((uint64_t)(data2[5])) << 40; - case 5: h ^= ((uint64_t)(data2[4])) << 32; - case 4: h ^= ((uint64_t)(data2[3])) << 24; - case 3: h ^= ((uint64_t)(data2[2])) << 16; - case 2: h ^= ((uint64_t)(data2[1])) << 8; - case 1: h ^= ((uint64_t)(data2[0])); - h *= m; + case 7: h ^= ((uint64_t)(data2[6])) << 48; + case 6: h ^= ((uint64_t)(data2[5])) << 40; + case 5: h ^= ((uint64_t)(data2[4])) << 32; + case 4: h ^= ((uint64_t)(data2[3])) << 24; + case 3: h ^= ((uint64_t)(data2[2])) << 16; + case 2: h ^= ((uint64_t)(data2[1])) << 8; + case 1: h ^= ((uint64_t)(data2[0])); + h *= m; } h ^= h >> r; diff --git a/include/jemalloc/internal/huge.h b/include/jemalloc/internal/huge.h index 66544cf8..e8513c93 100644 --- a/include/jemalloc/internal/huge.h +++ b/include/jemalloc/internal/huge.h @@ -9,12 +9,10 @@ /******************************************************************************/ #ifdef JEMALLOC_H_EXTERNS -#ifdef JEMALLOC_STATS /* Huge allocation statistics. */ extern uint64_t huge_nmalloc; extern uint64_t huge_ndalloc; extern size_t huge_allocated; -#endif /* Protects chunk-related data structures. */ extern malloc_mutex_t huge_mtx; @@ -27,11 +25,12 @@ void *huge_ralloc(void *ptr, size_t oldsize, size_t size, size_t extra, size_t alignment, bool zero); void huge_dalloc(void *ptr, bool unmap); size_t huge_salloc(const void *ptr); -#ifdef JEMALLOC_PROF prof_ctx_t *huge_prof_ctx_get(const void *ptr); void huge_prof_ctx_set(const void *ptr, prof_ctx_t *ctx); -#endif bool huge_boot(void); +void huge_prefork(void); +void huge_postfork_parent(void); +void huge_postfork_child(void); #endif /* JEMALLOC_H_EXTERNS */ /******************************************************************************/ diff --git a/include/jemalloc/internal/jemalloc_internal.h.in b/include/jemalloc/internal/jemalloc_internal.h.in index a44f0978..268cd146 100644 --- a/include/jemalloc/internal/jemalloc_internal.h.in +++ b/include/jemalloc/internal/jemalloc_internal.h.in @@ -1,17 +1,33 @@ -#include -#include -#include +#ifndef JEMALLOC_INTERNAL_H +#define JEMALLOC_INTERNAL_H +#include +#ifdef _WIN32 +# include +# define ENOENT ERROR_PATH_NOT_FOUND +# define EINVAL ERROR_BAD_ARGUMENTS +# define EAGAIN ERROR_OUTOFMEMORY +# define EPERM ERROR_WRITE_FAULT +# define EFAULT ERROR_INVALID_ADDRESS +# define ENOMEM ERROR_NOT_ENOUGH_MEMORY +# undef ERANGE +# define ERANGE ERROR_INVALID_DATA +#else +# include +# include +# include +# if !defined(SYS_write) && defined(__NR_write) +# define SYS_write __NR_write +# endif +# include +# include +# include +#endif #include -#include -#include -#include #include #ifndef SIZE_T_MAX # define SIZE_T_MAX SIZE_MAX #endif -#include -#include #include #include #include @@ -25,16 +41,156 @@ #include #include #include -#include +#ifdef _MSC_VER +# include +typedef intptr_t ssize_t; +# define PATH_MAX 1024 +# define STDERR_FILENO 2 +# define __func__ __FUNCTION__ +/* Disable warnings about deprecated system functions */ +# pragma warning(disable: 4996) +#else +# include +#endif #include -#include -#include -#define JEMALLOC_MANGLE +#define JEMALLOC_NO_DEMANGLE #include "../jemalloc@install_suffix@.h" +#ifdef JEMALLOC_UTRACE +#include +#endif + +#ifdef JEMALLOC_VALGRIND +#include +#include +#endif + #include "jemalloc/internal/private_namespace.h" +#ifdef JEMALLOC_CC_SILENCE +#define UNUSED JEMALLOC_ATTR(unused) +#else +#define UNUSED +#endif + +static const bool config_debug = +#ifdef JEMALLOC_DEBUG + true +#else + false +#endif + ; +static const bool config_dss = +#ifdef JEMALLOC_DSS + true +#else + false +#endif + ; +static const bool config_fill = +#ifdef JEMALLOC_FILL + true +#else + false +#endif + ; +static const bool config_lazy_lock = +#ifdef JEMALLOC_LAZY_LOCK + true +#else + false +#endif + ; +static const bool config_prof = +#ifdef JEMALLOC_PROF + true +#else + false +#endif + ; +static const bool config_prof_libgcc = +#ifdef JEMALLOC_PROF_LIBGCC + true +#else + false +#endif + ; +static const bool config_prof_libunwind = +#ifdef JEMALLOC_PROF_LIBUNWIND + true +#else + false +#endif + ; +static const bool config_mremap = +#ifdef JEMALLOC_MREMAP + true +#else + false +#endif + ; +static const bool config_munmap = +#ifdef JEMALLOC_MUNMAP + true +#else + false +#endif + ; +static const bool config_stats = +#ifdef JEMALLOC_STATS + true +#else + false +#endif + ; +static const bool config_tcache = +#ifdef JEMALLOC_TCACHE + true +#else + false +#endif + ; +static const bool config_tls = +#ifdef JEMALLOC_TLS + true +#else + false +#endif + ; +static const bool config_utrace = +#ifdef JEMALLOC_UTRACE + true +#else + false +#endif + ; +static const bool config_valgrind = +#ifdef JEMALLOC_VALGRIND + true +#else + false +#endif + ; +static const bool config_xmalloc = +#ifdef JEMALLOC_XMALLOC + true +#else + false +#endif + ; +static const bool config_ivsalloc = +#ifdef JEMALLOC_IVSALLOC + true +#else + false +#endif + ; + +#ifdef JEMALLOC_ATOMIC9 +#include +#endif + #if (defined(JEMALLOC_OSATOMIC) || defined(JEMALLOC_OSSPIN)) #include #endif @@ -46,48 +202,11 @@ #include #endif -#ifdef JEMALLOC_LAZY_LOCK -#include -#endif - #define RB_COMPACT #include "jemalloc/internal/rb.h" #include "jemalloc/internal/qr.h" #include "jemalloc/internal/ql.h" -extern void (*JEMALLOC_P(malloc_message))(void *wcbopaque, const char *s); - -/* - * Define a custom assert() in order to reduce the chances of deadlock during - * assertion failure. - */ -#ifndef assert -# ifdef JEMALLOC_DEBUG -# define assert(e) do { \ - if (!(e)) { \ - char line_buf[UMAX2S_BUFSIZE]; \ - malloc_write(": "); \ - malloc_write(__FILE__); \ - malloc_write(":"); \ - malloc_write(u2s(__LINE__, 10, line_buf)); \ - malloc_write(": Failed assertion: "); \ - malloc_write("\""); \ - malloc_write(#e); \ - malloc_write("\"\n"); \ - abort(); \ - } \ -} while (0) -# else -# define assert(e) -# endif -#endif - -#ifdef JEMALLOC_DEBUG -# define dassert(e) assert(e) -#else -# define dassert(e) -#endif - /* * jemalloc can conceptually be broken into components (arena, tcache, etc.), * but there are circular dependencies that cannot be broken without @@ -119,38 +238,56 @@ extern void (*JEMALLOC_P(malloc_message))(void *wcbopaque, const char *s); #else # define JEMALLOC_ENABLE_INLINE # define JEMALLOC_INLINE static inline +# ifdef _MSC_VER +# define inline _inline +# endif #endif -/* Size of stack-allocated buffer passed to buferror(). */ -#define BUFERROR_BUF 64 +/* Smallest size class to support. */ +#define LG_TINY_MIN 3 +#define TINY_MIN (1U << LG_TINY_MIN) -/* Minimum alignment of allocations is 2^LG_QUANTUM bytes. */ -#ifdef __i386__ -# define LG_QUANTUM 4 -#endif -#ifdef __ia64__ -# define LG_QUANTUM 4 -#endif -#ifdef __alpha__ -# define LG_QUANTUM 4 -#endif -#ifdef __sparc64__ -# define LG_QUANTUM 4 -#endif -#if (defined(__amd64__) || defined(__x86_64__)) -# define LG_QUANTUM 4 -#endif -#ifdef __arm__ -# define LG_QUANTUM 3 -#endif -#ifdef __mips__ -# define LG_QUANTUM 3 -#endif -#ifdef __powerpc__ -# define LG_QUANTUM 4 -#endif -#ifdef __s390x__ -# define LG_QUANTUM 4 +/* + * Minimum alignment of allocations is 2^LG_QUANTUM bytes (ignoring tiny size + * classes). + */ +#ifndef LG_QUANTUM +# if (defined(__i386__) || defined(_M_IX86)) +# define LG_QUANTUM 4 +# endif +# ifdef __ia64__ +# define LG_QUANTUM 4 +# endif +# ifdef __alpha__ +# define LG_QUANTUM 4 +# endif +# ifdef __sparc64__ +# define LG_QUANTUM 4 +# endif +# if (defined(__amd64__) || defined(__x86_64__) || defined(_M_X64)) +# define LG_QUANTUM 4 +# endif +# ifdef __arm__ +# define LG_QUANTUM 3 +# endif +# ifdef __mips__ +# define LG_QUANTUM 3 +# endif +# ifdef __powerpc__ +# define LG_QUANTUM 4 +# endif +# ifdef __s390x__ +# define LG_QUANTUM 4 +# endif +# ifdef __SH4__ +# define LG_QUANTUM 4 +# endif +# ifdef __tile__ +# define LG_QUANTUM 4 +# endif +# ifndef LG_QUANTUM +# error "No LG_QUANTUM definition for architecture; specify via CPPFLAGS" +# endif #endif #define QUANTUM ((size_t)(1U << LG_QUANTUM)) @@ -164,67 +301,149 @@ extern void (*JEMALLOC_P(malloc_message))(void *wcbopaque, const char *s); #define LONG_MASK (LONG - 1) /* Return the smallest long multiple that is >= a. */ -#define LONG_CEILING(a) \ +#define LONG_CEILING(a) \ (((a) + LONG_MASK) & ~LONG_MASK) #define SIZEOF_PTR (1U << LG_SIZEOF_PTR) #define PTR_MASK (SIZEOF_PTR - 1) /* Return the smallest (void *) multiple that is >= a. */ -#define PTR_CEILING(a) \ +#define PTR_CEILING(a) \ (((a) + PTR_MASK) & ~PTR_MASK) /* * Maximum size of L1 cache line. This is used to avoid cache line aliasing. * In addition, this controls the spacing of cacheline-spaced size classes. + * + * CACHELINE cannot be based on LG_CACHELINE because __declspec(align()) can + * only handle raw constants. */ #define LG_CACHELINE 6 -#define CACHELINE ((size_t)(1U << LG_CACHELINE)) +#define CACHELINE 64 #define CACHELINE_MASK (CACHELINE - 1) /* Return the smallest cacheline multiple that is >= s. */ #define CACHELINE_CEILING(s) \ (((s) + CACHELINE_MASK) & ~CACHELINE_MASK) -/* - * Page size. STATIC_PAGE_SHIFT is determined by the configure script. If - * DYNAMIC_PAGE_SHIFT is enabled, only use the STATIC_PAGE_* macros where - * compile-time values are required for the purposes of defining data - * structures. - */ -#define STATIC_PAGE_SIZE ((size_t)(1U << STATIC_PAGE_SHIFT)) -#define STATIC_PAGE_MASK ((size_t)(STATIC_PAGE_SIZE - 1)) - -#ifdef PAGE_SHIFT -# undef PAGE_SHIFT -#endif -#ifdef PAGE_SIZE -# undef PAGE_SIZE -#endif +/* Page size. STATIC_PAGE_SHIFT is determined by the configure script. */ #ifdef PAGE_MASK # undef PAGE_MASK #endif - -#ifdef DYNAMIC_PAGE_SHIFT -# define PAGE_SHIFT lg_pagesize -# define PAGE_SIZE pagesize -# define PAGE_MASK pagesize_mask -#else -# define PAGE_SHIFT STATIC_PAGE_SHIFT -# define PAGE_SIZE STATIC_PAGE_SIZE -# define PAGE_MASK STATIC_PAGE_MASK -#endif +#define LG_PAGE STATIC_PAGE_SHIFT +#define PAGE ((size_t)(1U << STATIC_PAGE_SHIFT)) +#define PAGE_MASK ((size_t)(PAGE - 1)) /* Return the smallest pagesize multiple that is >= s. */ #define PAGE_CEILING(s) \ (((s) + PAGE_MASK) & ~PAGE_MASK) +/* Return the nearest aligned address at or below a. */ +#define ALIGNMENT_ADDR2BASE(a, alignment) \ + ((void *)((uintptr_t)(a) & (-(alignment)))) + +/* Return the offset between a and the nearest aligned address at or below a. */ +#define ALIGNMENT_ADDR2OFFSET(a, alignment) \ + ((size_t)((uintptr_t)(a) & (alignment - 1))) + +/* Return the smallest alignment multiple that is >= s. */ +#define ALIGNMENT_CEILING(s, alignment) \ + (((s) + (alignment - 1)) & (-(alignment))) + +/* Declare a variable length array */ +#if __STDC_VERSION__ < 199901L +# ifdef _MSC_VER +# include +# define alloca _alloca +# else +# include +# endif +# define VARIABLE_ARRAY(type, name, count) \ + type *name = alloca(sizeof(type) * count) +#else +# define VARIABLE_ARRAY(type, name, count) type name[count] +#endif + +#ifdef JEMALLOC_VALGRIND +/* + * The JEMALLOC_VALGRIND_*() macros must be macros rather than functions + * so that when Valgrind reports errors, there are no extra stack frames + * in the backtraces. + * + * The size that is reported to valgrind must be consistent through a chain of + * malloc..realloc..realloc calls. Request size isn't recorded anywhere in + * jemalloc, so it is critical that all callers of these macros provide usize + * rather than request size. As a result, buffer overflow detection is + * technically weakened for the standard API, though it is generally accepted + * practice to consider any extra bytes reported by malloc_usable_size() as + * usable space. + */ +#define JEMALLOC_VALGRIND_MALLOC(cond, ptr, usize, zero) do { \ + if (config_valgrind && opt_valgrind && cond) \ + VALGRIND_MALLOCLIKE_BLOCK(ptr, usize, p2rz(ptr), zero); \ +} while (0) +#define JEMALLOC_VALGRIND_REALLOC(ptr, usize, old_ptr, old_usize, \ + old_rzsize, zero) do { \ + if (config_valgrind && opt_valgrind) { \ + size_t rzsize = p2rz(ptr); \ + \ + if (ptr == old_ptr) { \ + VALGRIND_RESIZEINPLACE_BLOCK(ptr, old_usize, \ + usize, rzsize); \ + if (zero && old_usize < usize) { \ + VALGRIND_MAKE_MEM_DEFINED( \ + (void *)((uintptr_t)ptr + \ + old_usize), usize - old_usize); \ + } \ + } else { \ + if (old_ptr != NULL) { \ + VALGRIND_FREELIKE_BLOCK(old_ptr, \ + old_rzsize); \ + } \ + if (ptr != NULL) { \ + size_t copy_size = (old_usize < usize) \ + ? old_usize : usize; \ + size_t tail_size = usize - copy_size; \ + VALGRIND_MALLOCLIKE_BLOCK(ptr, usize, \ + rzsize, false); \ + if (copy_size > 0) { \ + VALGRIND_MAKE_MEM_DEFINED(ptr, \ + copy_size); \ + } \ + if (zero && tail_size > 0) { \ + VALGRIND_MAKE_MEM_DEFINED( \ + (void *)((uintptr_t)ptr + \ + copy_size), tail_size); \ + } \ + } \ + } \ + } \ +} while (0) +#define JEMALLOC_VALGRIND_FREE(ptr, rzsize) do { \ + if (config_valgrind && opt_valgrind) \ + VALGRIND_FREELIKE_BLOCK(ptr, rzsize); \ +} while (0) +#else +#define VALGRIND_MALLOCLIKE_BLOCK(addr, sizeB, rzB, is_zeroed) +#define VALGRIND_RESIZEINPLACE_BLOCK(addr, oldSizeB, newSizeB, rzB) +#define VALGRIND_FREELIKE_BLOCK(addr, rzB) +#define VALGRIND_MAKE_MEM_UNDEFINED(_qzz_addr, _qzz_len) +#define VALGRIND_MAKE_MEM_DEFINED(_qzz_addr, _qzz_len) +#define JEMALLOC_VALGRIND_MALLOC(cond, ptr, usize, zero) +#define JEMALLOC_VALGRIND_REALLOC(ptr, usize, old_ptr, old_usize, \ + old_rzsize, zero) +#define JEMALLOC_VALGRIND_FREE(ptr, rzsize) +#endif + +#include "jemalloc/internal/util.h" #include "jemalloc/internal/atomic.h" -#include "jemalloc/internal/prn.h" +#include "jemalloc/internal/prng.h" #include "jemalloc/internal/ckh.h" +#include "jemalloc/internal/size_classes.h" #include "jemalloc/internal/stats.h" #include "jemalloc/internal/ctl.h" #include "jemalloc/internal/mutex.h" +#include "jemalloc/internal/tsd.h" #include "jemalloc/internal/mb.h" #include "jemalloc/internal/extent.h" #include "jemalloc/internal/arena.h" @@ -235,21 +454,22 @@ extern void (*JEMALLOC_P(malloc_message))(void *wcbopaque, const char *s); #include "jemalloc/internal/rtree.h" #include "jemalloc/internal/tcache.h" #include "jemalloc/internal/hash.h" -#ifdef JEMALLOC_ZONE -#include "jemalloc/internal/zone.h" -#endif +#include "jemalloc/internal/quarantine.h" #include "jemalloc/internal/prof.h" #undef JEMALLOC_H_TYPES /******************************************************************************/ #define JEMALLOC_H_STRUCTS +#include "jemalloc/internal/util.h" #include "jemalloc/internal/atomic.h" -#include "jemalloc/internal/prn.h" +#include "jemalloc/internal/prng.h" #include "jemalloc/internal/ckh.h" +#include "jemalloc/internal/size_classes.h" #include "jemalloc/internal/stats.h" #include "jemalloc/internal/ctl.h" #include "jemalloc/internal/mutex.h" +#include "jemalloc/internal/tsd.h" #include "jemalloc/internal/mb.h" #include "jemalloc/internal/bitmap.h" #include "jemalloc/internal/extent.h" @@ -260,66 +480,37 @@ extern void (*JEMALLOC_P(malloc_message))(void *wcbopaque, const char *s); #include "jemalloc/internal/rtree.h" #include "jemalloc/internal/tcache.h" #include "jemalloc/internal/hash.h" -#ifdef JEMALLOC_ZONE -#include "jemalloc/internal/zone.h" -#endif +#include "jemalloc/internal/quarantine.h" #include "jemalloc/internal/prof.h" -#ifdef JEMALLOC_STATS typedef struct { uint64_t allocated; uint64_t deallocated; } thread_allocated_t; -#endif +/* + * The JEMALLOC_CONCAT() wrapper is necessary to pass {0, 0} via a cpp macro + * argument. + */ +#define THREAD_ALLOCATED_INITIALIZER JEMALLOC_CONCAT({0, 0}) #undef JEMALLOC_H_STRUCTS /******************************************************************************/ #define JEMALLOC_H_EXTERNS extern bool opt_abort; -#ifdef JEMALLOC_FILL extern bool opt_junk; -#endif -#ifdef JEMALLOC_SYSV -extern bool opt_sysv; -#endif -#ifdef JEMALLOC_XMALLOC +extern size_t opt_quarantine; +extern bool opt_redzone; +extern bool opt_utrace; +extern bool opt_valgrind; extern bool opt_xmalloc; -#endif -#ifdef JEMALLOC_FILL extern bool opt_zero; -#endif extern size_t opt_narenas; -#ifdef DYNAMIC_PAGE_SHIFT -extern size_t pagesize; -extern size_t pagesize_mask; -extern size_t lg_pagesize; -#endif - /* Number of CPUs. */ extern unsigned ncpus; extern malloc_mutex_t arenas_lock; /* Protects arenas initialization. */ -extern pthread_key_t arenas_tsd; -#ifndef NO_TLS -/* - * Map of pthread_self() --> arenas[???], used for selecting an arena to use - * for allocations. - */ -extern __thread arena_t *arenas_tls JEMALLOC_ATTR(tls_model("initial-exec")); -# define ARENA_GET() arenas_tls -# define ARENA_SET(v) do { \ - arenas_tls = (v); \ - pthread_setspecific(arenas_tsd, (void *)(v)); \ -} while (0) -#else -# define ARENA_GET() ((arena_t *)pthread_getspecific(arenas_tsd)) -# define ARENA_SET(v) do { \ - pthread_setspecific(arenas_tsd, (void *)(v)); \ -} while (0) -#endif - /* * Arenas that are used to service external requests. Not all elements of the * arenas array are necessarily used; arenas are created lazily as needed. @@ -327,45 +518,22 @@ extern __thread arena_t *arenas_tls JEMALLOC_ATTR(tls_model("initial-exec")); extern arena_t **arenas; extern unsigned narenas; -#ifdef JEMALLOC_STATS -# ifndef NO_TLS -extern __thread thread_allocated_t thread_allocated_tls; -# define ALLOCATED_GET() (thread_allocated_tls.allocated) -# define ALLOCATEDP_GET() (&thread_allocated_tls.allocated) -# define DEALLOCATED_GET() (thread_allocated_tls.deallocated) -# define DEALLOCATEDP_GET() (&thread_allocated_tls.deallocated) -# define ALLOCATED_ADD(a, d) do { \ - thread_allocated_tls.allocated += a; \ - thread_allocated_tls.deallocated += d; \ -} while (0) -# else -extern pthread_key_t thread_allocated_tsd; -thread_allocated_t *thread_allocated_get_hard(void); - -# define ALLOCATED_GET() (thread_allocated_get()->allocated) -# define ALLOCATEDP_GET() (&thread_allocated_get()->allocated) -# define DEALLOCATED_GET() (thread_allocated_get()->deallocated) -# define DEALLOCATEDP_GET() (&thread_allocated_get()->deallocated) -# define ALLOCATED_ADD(a, d) do { \ - thread_allocated_t *thread_allocated = thread_allocated_get(); \ - thread_allocated->allocated += (a); \ - thread_allocated->deallocated += (d); \ -} while (0) -# endif -#endif - arena_t *arenas_extend(unsigned ind); +void arenas_cleanup(void *arg); arena_t *choose_arena_hard(void); -int buferror(int errnum, char *buf, size_t buflen); void jemalloc_prefork(void); -void jemalloc_postfork(void); +void jemalloc_postfork_parent(void); +void jemalloc_postfork_child(void); +#include "jemalloc/internal/util.h" #include "jemalloc/internal/atomic.h" -#include "jemalloc/internal/prn.h" +#include "jemalloc/internal/prng.h" #include "jemalloc/internal/ckh.h" +#include "jemalloc/internal/size_classes.h" #include "jemalloc/internal/stats.h" #include "jemalloc/internal/ctl.h" #include "jemalloc/internal/mutex.h" +#include "jemalloc/internal/tsd.h" #include "jemalloc/internal/mb.h" #include "jemalloc/internal/bitmap.h" #include "jemalloc/internal/extent.h" @@ -376,21 +544,22 @@ void jemalloc_postfork(void); #include "jemalloc/internal/rtree.h" #include "jemalloc/internal/tcache.h" #include "jemalloc/internal/hash.h" -#ifdef JEMALLOC_ZONE -#include "jemalloc/internal/zone.h" -#endif +#include "jemalloc/internal/quarantine.h" #include "jemalloc/internal/prof.h" #undef JEMALLOC_H_EXTERNS /******************************************************************************/ #define JEMALLOC_H_INLINES +#include "jemalloc/internal/util.h" #include "jemalloc/internal/atomic.h" -#include "jemalloc/internal/prn.h" +#include "jemalloc/internal/prng.h" #include "jemalloc/internal/ckh.h" +#include "jemalloc/internal/size_classes.h" #include "jemalloc/internal/stats.h" #include "jemalloc/internal/ctl.h" #include "jemalloc/internal/mutex.h" +#include "jemalloc/internal/tsd.h" #include "jemalloc/internal/mb.h" #include "jemalloc/internal/extent.h" #include "jemalloc/internal/base.h" @@ -398,34 +567,20 @@ void jemalloc_postfork(void); #include "jemalloc/internal/huge.h" #ifndef JEMALLOC_ENABLE_INLINE -size_t pow2_ceil(size_t x); +malloc_tsd_protos(JEMALLOC_ATTR(unused), arenas, arena_t *) + size_t s2u(size_t size); -size_t sa2u(size_t size, size_t alignment, size_t *run_size_p); -void malloc_write(const char *s); -arena_t *choose_arena(void); -# if (defined(JEMALLOC_STATS) && defined(NO_TLS)) -thread_allocated_t *thread_allocated_get(void); -# endif +size_t sa2u(size_t size, size_t alignment); +arena_t *choose_arena(arena_t *arena); #endif #if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_C_)) -/* Compute the smallest power of 2 that is >= x. */ -JEMALLOC_INLINE size_t -pow2_ceil(size_t x) -{ - - x--; - x |= x >> 1; - x |= x >> 2; - x |= x >> 4; - x |= x >> 8; - x |= x >> 16; -#if (LG_SIZEOF_PTR == 3) - x |= x >> 32; -#endif - x++; - return (x); -} +/* + * Map of pthread_self() --> arenas[???], used for selecting an arena to use + * for allocations. + */ +malloc_tsd_externs(arenas, arena_t *) +malloc_tsd_funcs(JEMALLOC_INLINE, arenas, arena_t *, NULL, arenas_cleanup) /* * Compute usable size that would result from allocating an object with the @@ -435,7 +590,7 @@ JEMALLOC_INLINE size_t s2u(size_t size) { - if (size <= small_maxclass) + if (size <= SMALL_MAXCLASS) return (arena_bin_info[SMALL_SIZE2BIN(size)].reg_size); if (size <= arena_maxclass) return (PAGE_CEILING(size)); @@ -447,10 +602,12 @@ s2u(size_t size) * specified size and alignment. */ JEMALLOC_INLINE size_t -sa2u(size_t size, size_t alignment, size_t *run_size_p) +sa2u(size_t size, size_t alignment) { size_t usize; + assert(alignment != 0 && ((alignment - 1) & alignment) == 0); + /* * Round size up to the nearest multiple of alignment. * @@ -464,12 +621,8 @@ sa2u(size_t size, size_t alignment, size_t *run_size_p) * 96 | 1100000 | 32 * 144 | 10100000 | 32 * 192 | 11000000 | 64 - * - * Depending on runtime settings, it is possible that arena_malloc() - * will further round up to a power of two, but that never causes - * correctness issues. */ - usize = (size + (alignment - 1)) & (-alignment); + usize = ALIGNMENT_CEILING(size, alignment); /* * (usize < size) protects against the combination of maximal * alignment and size greater than maximal alignment. @@ -479,8 +632,8 @@ sa2u(size_t size, size_t alignment, size_t *run_size_p) return (0); } - if (usize <= arena_maxclass && alignment <= PAGE_SIZE) { - if (usize <= small_maxclass) + if (usize <= arena_maxclass && alignment <= PAGE) { + if (usize <= SMALL_MAXCLASS) return (arena_bin_info[SMALL_SIZE2BIN(usize)].reg_size); return (PAGE_CEILING(usize)); } else { @@ -494,7 +647,7 @@ sa2u(size_t size, size_t alignment, size_t *run_size_p) usize = PAGE_CEILING(size); /* * (usize < size) protects against very large sizes within - * PAGE_SIZE of SIZE_T_MAX. + * PAGE of SIZE_T_MAX. * * (usize + alignment < usize) protects against the * combination of maximal alignment and usize large enough @@ -512,93 +665,63 @@ sa2u(size_t size, size_t alignment, size_t *run_size_p) /* * Calculate the size of the over-size run that arena_palloc() * would need to allocate in order to guarantee the alignment. + * If the run wouldn't fit within a chunk, round up to a huge + * allocation size. */ - if (usize >= alignment) - run_size = usize + alignment - PAGE_SIZE; - else { - /* - * It is possible that (alignment << 1) will cause - * overflow, but it doesn't matter because we also - * subtract PAGE_SIZE, which in the case of overflow - * leaves us with a very large run_size. That causes - * the first conditional below to fail, which means - * that the bogus run_size value never gets used for - * anything important. - */ - run_size = (alignment << 1) - PAGE_SIZE; - } - if (run_size_p != NULL) - *run_size_p = run_size; - + run_size = usize + alignment - PAGE; if (run_size <= arena_maxclass) return (PAGE_CEILING(usize)); return (CHUNK_CEILING(usize)); } } -/* - * Wrapper around malloc_message() that avoids the need for - * JEMALLOC_P(malloc_message)(...) throughout the code. - */ -JEMALLOC_INLINE void -malloc_write(const char *s) -{ - - JEMALLOC_P(malloc_message)(NULL, s); -} - -/* - * Choose an arena based on a per-thread value (fast-path code, calls slow-path - * code if necessary). - */ +/* Choose an arena based on a per-thread value. */ JEMALLOC_INLINE arena_t * -choose_arena(void) +choose_arena(arena_t *arena) { arena_t *ret; - ret = ARENA_GET(); - if (ret == NULL) { + if (arena != NULL) + return (arena); + + if ((ret = *arenas_tsd_get()) == NULL) { ret = choose_arena_hard(); assert(ret != NULL); } return (ret); } - -#if (defined(JEMALLOC_STATS) && defined(NO_TLS)) -JEMALLOC_INLINE thread_allocated_t * -thread_allocated_get(void) -{ - thread_allocated_t *thread_allocated = (thread_allocated_t *) - pthread_getspecific(thread_allocated_tsd); - - if (thread_allocated == NULL) - return (thread_allocated_get_hard()); - return (thread_allocated); -} -#endif #endif #include "jemalloc/internal/bitmap.h" #include "jemalloc/internal/rtree.h" -#include "jemalloc/internal/tcache.h" +/* + * Include arena.h twice in order to resolve circular dependencies with + * tcache.h. + */ +#define JEMALLOC_ARENA_INLINE_A #include "jemalloc/internal/arena.h" +#undef JEMALLOC_ARENA_INLINE_A +#include "jemalloc/internal/tcache.h" +#define JEMALLOC_ARENA_INLINE_B +#include "jemalloc/internal/arena.h" +#undef JEMALLOC_ARENA_INLINE_B #include "jemalloc/internal/hash.h" -#ifdef JEMALLOC_ZONE -#include "jemalloc/internal/zone.h" -#endif +#include "jemalloc/internal/quarantine.h" #ifndef JEMALLOC_ENABLE_INLINE void *imalloc(size_t size); void *icalloc(size_t size); void *ipalloc(size_t usize, size_t alignment, bool zero); -size_t isalloc(const void *ptr); -# ifdef JEMALLOC_IVSALLOC -size_t ivsalloc(const void *ptr); -# endif +size_t isalloc(const void *ptr, bool demote); +size_t ivsalloc(const void *ptr, bool demote); +size_t u2rz(size_t usize); +size_t p2rz(const void *ptr); void idalloc(void *ptr); +void iqalloc(void *ptr); void *iralloc(void *ptr, size_t size, size_t extra, size_t alignment, bool zero, bool no_move); +malloc_tsd_protos(JEMALLOC_ATTR(unused), thread_allocated, thread_allocated_t) #endif #if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_C_)) @@ -609,7 +732,7 @@ imalloc(size_t size) assert(size != 0); if (size <= arena_maxclass) - return (arena_malloc(size, false)); + return (arena_malloc(NULL, size, false, true)); else return (huge_malloc(size, false)); } @@ -619,7 +742,7 @@ icalloc(size_t size) { if (size <= arena_maxclass) - return (arena_malloc(size, true)); + return (arena_malloc(NULL, size, true, true)); else return (huge_malloc(size, true)); } @@ -630,75 +753,80 @@ ipalloc(size_t usize, size_t alignment, bool zero) void *ret; assert(usize != 0); - assert(usize == sa2u(usize, alignment, NULL)); + assert(usize == sa2u(usize, alignment)); - if (usize <= arena_maxclass && alignment <= PAGE_SIZE) - ret = arena_malloc(usize, zero); + if (usize <= arena_maxclass && alignment <= PAGE) + ret = arena_malloc(NULL, usize, zero, true); else { - size_t run_size -#ifdef JEMALLOC_CC_SILENCE - = 0 -#endif - ; - - /* - * Ideally we would only ever call sa2u() once per aligned - * allocation request, and the caller of this function has - * already done so once. However, it's rather burdensome to - * require every caller to pass in run_size, especially given - * that it's only relevant to large allocations. Therefore, - * just call it again here in order to get run_size. - */ - sa2u(usize, alignment, &run_size); - if (run_size <= arena_maxclass) { - ret = arena_palloc(choose_arena(), usize, run_size, - alignment, zero); + if (usize <= arena_maxclass) { + ret = arena_palloc(choose_arena(NULL), usize, alignment, + zero); } else if (alignment <= chunksize) ret = huge_malloc(usize, zero); else ret = huge_palloc(usize, alignment, zero); } - assert(((uintptr_t)ret & (alignment - 1)) == 0); + assert(ALIGNMENT_ADDR2BASE(ret, alignment) == ret); return (ret); } +/* + * Typical usage: + * void *ptr = [...] + * size_t sz = isalloc(ptr, config_prof); + */ JEMALLOC_INLINE size_t -isalloc(const void *ptr) +isalloc(const void *ptr, bool demote) { size_t ret; arena_chunk_t *chunk; assert(ptr != NULL); + /* Demotion only makes sense if config_prof is true. */ + assert(config_prof || demote == false); chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr); - if (chunk != ptr) { - /* Region. */ - dassert(chunk->arena->magic == ARENA_MAGIC); - -#ifdef JEMALLOC_PROF - ret = arena_salloc_demote(ptr); -#else - ret = arena_salloc(ptr); -#endif - } else + if (chunk != ptr) + ret = arena_salloc(ptr, demote); + else ret = huge_salloc(ptr); return (ret); } -#ifdef JEMALLOC_IVSALLOC JEMALLOC_INLINE size_t -ivsalloc(const void *ptr) +ivsalloc(const void *ptr, bool demote) { /* Return 0 if ptr is not within a chunk managed by jemalloc. */ if (rtree_get(chunks_rtree, (uintptr_t)CHUNK_ADDR2BASE(ptr)) == NULL) return (0); - return (isalloc(ptr)); + return (isalloc(ptr, demote)); +} + +JEMALLOC_INLINE size_t +u2rz(size_t usize) +{ + size_t ret; + + if (usize <= SMALL_MAXCLASS) { + size_t binind = SMALL_SIZE2BIN(usize); + ret = arena_bin_info[binind].redzone_size; + } else + ret = 0; + + return (ret); +} + +JEMALLOC_INLINE size_t +p2rz(const void *ptr) +{ + size_t usize = isalloc(ptr, false); + + return (u2rz(usize)); } -#endif JEMALLOC_INLINE void idalloc(void *ptr) @@ -709,11 +837,21 @@ idalloc(void *ptr) chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr); if (chunk != ptr) - arena_dalloc(chunk->arena, chunk, ptr); + arena_dalloc(chunk->arena, chunk, ptr, true); else huge_dalloc(ptr, true); } +JEMALLOC_INLINE void +iqalloc(void *ptr) +{ + + if (config_fill && opt_quarantine) + quarantine(ptr); + else + idalloc(ptr); +} + JEMALLOC_INLINE void * iralloc(void *ptr, size_t size, size_t extra, size_t alignment, bool zero, bool no_move) @@ -724,19 +862,19 @@ iralloc(void *ptr, size_t size, size_t extra, size_t alignment, bool zero, assert(ptr != NULL); assert(size != 0); - oldsize = isalloc(ptr); + oldsize = isalloc(ptr, config_prof); if (alignment != 0 && ((uintptr_t)ptr & ((uintptr_t)alignment-1)) != 0) { size_t usize, copysize; /* - * Existing object alignment is inadquate; allocate new space + * Existing object alignment is inadequate; allocate new space * and copy. */ if (no_move) return (NULL); - usize = sa2u(size + extra, alignment, NULL); + usize = sa2u(size + extra, alignment); if (usize == 0) return (NULL); ret = ipalloc(usize, alignment, zero); @@ -744,7 +882,7 @@ iralloc(void *ptr, size_t size, size_t extra, size_t alignment, bool zero, if (extra == 0) return (NULL); /* Try again, without extra this time. */ - usize = sa2u(size, alignment, NULL); + usize = sa2u(size, alignment); if (usize == 0) return (NULL); ret = ipalloc(usize, alignment, zero); @@ -758,7 +896,7 @@ iralloc(void *ptr, size_t size, size_t extra, size_t alignment, bool zero, */ copysize = (size < oldsize) ? size : oldsize; memcpy(ret, ptr, copysize); - idalloc(ptr); + iqalloc(ptr); return (ret); } @@ -773,16 +911,21 @@ iralloc(void *ptr, size_t size, size_t extra, size_t alignment, bool zero, } else { if (size + extra <= arena_maxclass) { return (arena_ralloc(ptr, oldsize, size, extra, - alignment, zero)); + alignment, zero, true)); } else { return (huge_ralloc(ptr, oldsize, size, extra, alignment, zero)); } } } + +malloc_tsd_externs(thread_allocated, thread_allocated_t) +malloc_tsd_funcs(JEMALLOC_INLINE, thread_allocated, thread_allocated_t, + THREAD_ALLOCATED_INITIALIZER, malloc_tsd_no_cleanup) #endif #include "jemalloc/internal/prof.h" #undef JEMALLOC_H_INLINES /******************************************************************************/ +#endif /* JEMALLOC_INTERNAL_H */ diff --git a/include/jemalloc/internal/mb.h b/include/jemalloc/internal/mb.h index dc9f2a54..3cfa7872 100644 --- a/include/jemalloc/internal/mb.h +++ b/include/jemalloc/internal/mb.h @@ -54,7 +54,7 @@ mb_write(void) ); #endif } -#elif (defined(__amd64_) || defined(__x86_64__)) +#elif (defined(__amd64__) || defined(__x86_64__)) JEMALLOC_INLINE void mb_write(void) { @@ -87,6 +87,13 @@ mb_write(void) : "memory" /* Clobbers. */ ); } +#elif defined(__tile__) +JEMALLOC_INLINE void +mb_write(void) +{ + + __sync_synchronize(); +} #else /* * This is much slower than a simple memory barrier, but the semantics of mutex diff --git a/include/jemalloc/internal/mutex.h b/include/jemalloc/internal/mutex.h index 62947ced..de44e143 100644 --- a/include/jemalloc/internal/mutex.h +++ b/include/jemalloc/internal/mutex.h @@ -1,22 +1,42 @@ /******************************************************************************/ #ifdef JEMALLOC_H_TYPES -#ifdef JEMALLOC_OSSPIN -typedef OSSpinLock malloc_mutex_t; -#else -typedef pthread_mutex_t malloc_mutex_t; -#endif +typedef struct malloc_mutex_s malloc_mutex_t; -#ifdef PTHREAD_ADAPTIVE_MUTEX_INITIALIZER_NP -# define MALLOC_MUTEX_INITIALIZER PTHREAD_ADAPTIVE_MUTEX_INITIALIZER_NP +#ifdef _WIN32 +# define MALLOC_MUTEX_INITIALIZER +#elif (defined(JEMALLOC_OSSPIN)) +# define MALLOC_MUTEX_INITIALIZER {0} +#elif (defined(JEMALLOC_MUTEX_INIT_CB)) +# define MALLOC_MUTEX_INITIALIZER {PTHREAD_MUTEX_INITIALIZER, NULL} #else -# define MALLOC_MUTEX_INITIALIZER PTHREAD_MUTEX_INITIALIZER +# if (defined(PTHREAD_MUTEX_ADAPTIVE_NP) && \ + defined(PTHREAD_ADAPTIVE_MUTEX_INITIALIZER_NP)) +# define MALLOC_MUTEX_TYPE PTHREAD_MUTEX_ADAPTIVE_NP +# define MALLOC_MUTEX_INITIALIZER {PTHREAD_ADAPTIVE_MUTEX_INITIALIZER_NP} +# else +# define MALLOC_MUTEX_TYPE PTHREAD_MUTEX_DEFAULT +# define MALLOC_MUTEX_INITIALIZER {PTHREAD_MUTEX_INITIALIZER} +# endif #endif #endif /* JEMALLOC_H_TYPES */ /******************************************************************************/ #ifdef JEMALLOC_H_STRUCTS +struct malloc_mutex_s { +#ifdef _WIN32 + CRITICAL_SECTION lock; +#elif (defined(JEMALLOC_OSSPIN)) + OSSpinLock lock; +#elif (defined(JEMALLOC_MUTEX_INIT_CB)) + pthread_mutex_t lock; + malloc_mutex_t *postponed_next; +#else + pthread_mutex_t lock; +#endif +}; + #endif /* JEMALLOC_H_STRUCTS */ /******************************************************************************/ #ifdef JEMALLOC_H_EXTERNS @@ -24,11 +44,15 @@ typedef pthread_mutex_t malloc_mutex_t; #ifdef JEMALLOC_LAZY_LOCK extern bool isthreaded; #else +# undef isthreaded /* Undo private_namespace.h definition. */ # define isthreaded true #endif bool malloc_mutex_init(malloc_mutex_t *mutex); -void malloc_mutex_destroy(malloc_mutex_t *mutex); +void malloc_mutex_prefork(malloc_mutex_t *mutex); +void malloc_mutex_postfork_parent(malloc_mutex_t *mutex); +void malloc_mutex_postfork_child(malloc_mutex_t *mutex); +bool mutex_boot(void); #endif /* JEMALLOC_H_EXTERNS */ /******************************************************************************/ @@ -36,7 +60,6 @@ void malloc_mutex_destroy(malloc_mutex_t *mutex); #ifndef JEMALLOC_ENABLE_INLINE void malloc_mutex_lock(malloc_mutex_t *mutex); -bool malloc_mutex_trylock(malloc_mutex_t *mutex); void malloc_mutex_unlock(malloc_mutex_t *mutex); #endif @@ -46,37 +69,27 @@ malloc_mutex_lock(malloc_mutex_t *mutex) { if (isthreaded) { -#ifdef JEMALLOC_OSSPIN - OSSpinLockLock(mutex); +#ifdef _WIN32 + EnterCriticalSection(&mutex->lock); +#elif (defined(JEMALLOC_OSSPIN)) + OSSpinLockLock(&mutex->lock); #else - pthread_mutex_lock(mutex); + pthread_mutex_lock(&mutex->lock); #endif } } -JEMALLOC_INLINE bool -malloc_mutex_trylock(malloc_mutex_t *mutex) -{ - - if (isthreaded) { -#ifdef JEMALLOC_OSSPIN - return (OSSpinLockTry(mutex) == false); -#else - return (pthread_mutex_trylock(mutex) != 0); -#endif - } else - return (false); -} - JEMALLOC_INLINE void malloc_mutex_unlock(malloc_mutex_t *mutex) { if (isthreaded) { -#ifdef JEMALLOC_OSSPIN - OSSpinLockUnlock(mutex); +#ifdef _WIN32 + LeaveCriticalSection(&mutex->lock); +#elif (defined(JEMALLOC_OSSPIN)) + OSSpinLockUnlock(&mutex->lock); #else - pthread_mutex_unlock(mutex); + pthread_mutex_unlock(&mutex->lock); #endif } } diff --git a/include/jemalloc/internal/private_namespace.h b/include/jemalloc/internal/private_namespace.h index d4f5f96d..b8166470 100644 --- a/include/jemalloc/internal/private_namespace.h +++ b/include/jemalloc/internal/private_namespace.h @@ -1,36 +1,84 @@ +#define a0calloc JEMALLOC_N(a0calloc) +#define a0free JEMALLOC_N(a0free) +#define a0malloc JEMALLOC_N(a0malloc) +#define arena_alloc_junk_small JEMALLOC_N(arena_alloc_junk_small) #define arena_bin_index JEMALLOC_N(arena_bin_index) +#define arena_bin_info JEMALLOC_N(arena_bin_info) #define arena_boot JEMALLOC_N(arena_boot) #define arena_dalloc JEMALLOC_N(arena_dalloc) #define arena_dalloc_bin JEMALLOC_N(arena_dalloc_bin) +#define arena_dalloc_bin_locked JEMALLOC_N(arena_dalloc_bin_locked) +#define arena_dalloc_junk_small JEMALLOC_N(arena_dalloc_junk_small) #define arena_dalloc_large JEMALLOC_N(arena_dalloc_large) +#define arena_dalloc_large_locked JEMALLOC_N(arena_dalloc_large_locked) +#define arena_dalloc_small JEMALLOC_N(arena_dalloc_small) #define arena_malloc JEMALLOC_N(arena_malloc) #define arena_malloc_large JEMALLOC_N(arena_malloc_large) #define arena_malloc_small JEMALLOC_N(arena_malloc_small) +#define arena_mapbits_allocated_get JEMALLOC_N(arena_mapbits_allocated_get) +#define arena_mapbits_binind_get JEMALLOC_N(arena_mapbits_binind_get) +#define arena_mapbits_dirty_get JEMALLOC_N(arena_mapbits_dirty_get) +#define arena_mapbits_get JEMALLOC_N(arena_mapbits_get) +#define arena_mapbits_large_binind_set JEMALLOC_N(arena_mapbits_large_binind_set) +#define arena_mapbits_large_get JEMALLOC_N(arena_mapbits_large_get) +#define arena_mapbits_large_set JEMALLOC_N(arena_mapbits_large_set) +#define arena_mapbits_large_size_get JEMALLOC_N(arena_mapbits_large_size_get) +#define arena_mapbits_small_runind_get JEMALLOC_N(arena_mapbits_small_runind_get) +#define arena_mapbits_small_set JEMALLOC_N(arena_mapbits_small_set) +#define arena_mapbits_unallocated_set JEMALLOC_N(arena_mapbits_unallocated_set) +#define arena_mapbits_unallocated_size_get JEMALLOC_N(arena_mapbits_unallocated_size_get) +#define arena_mapbits_unallocated_size_set JEMALLOC_N(arena_mapbits_unallocated_size_set) +#define arena_mapbits_unzeroed_get JEMALLOC_N(arena_mapbits_unzeroed_get) +#define arena_mapbits_unzeroed_set JEMALLOC_N(arena_mapbits_unzeroed_set) +#define arena_mapbitsp_get JEMALLOC_N(arena_mapbitsp_get) +#define arena_mapp_get JEMALLOC_N(arena_mapp_get) +#define arena_maxclass JEMALLOC_N(arena_maxclass) #define arena_new JEMALLOC_N(arena_new) #define arena_palloc JEMALLOC_N(arena_palloc) +#define arena_postfork_child JEMALLOC_N(arena_postfork_child) +#define arena_postfork_parent JEMALLOC_N(arena_postfork_parent) +#define arena_prefork JEMALLOC_N(arena_prefork) #define arena_prof_accum JEMALLOC_N(arena_prof_accum) #define arena_prof_ctx_get JEMALLOC_N(arena_prof_ctx_get) #define arena_prof_ctx_set JEMALLOC_N(arena_prof_ctx_set) #define arena_prof_promoted JEMALLOC_N(arena_prof_promoted) +#define arena_ptr_small_binind_get JEMALLOC_N(arena_ptr_small_binind_get) #define arena_purge_all JEMALLOC_N(arena_purge_all) #define arena_ralloc JEMALLOC_N(arena_ralloc) #define arena_ralloc_no_move JEMALLOC_N(arena_ralloc_no_move) #define arena_run_regind JEMALLOC_N(arena_run_regind) #define arena_salloc JEMALLOC_N(arena_salloc) -#define arena_salloc_demote JEMALLOC_N(arena_salloc_demote) #define arena_stats_merge JEMALLOC_N(arena_stats_merge) #define arena_tcache_fill_small JEMALLOC_N(arena_tcache_fill_small) +#define arenas JEMALLOC_N(arenas) #define arenas_bin_i_index JEMALLOC_N(arenas_bin_i_index) +#define arenas_booted JEMALLOC_N(arenas_booted) +#define arenas_cleanup JEMALLOC_N(arenas_cleanup) #define arenas_extend JEMALLOC_N(arenas_extend) +#define arenas_initialized JEMALLOC_N(arenas_initialized) +#define arenas_lock JEMALLOC_N(arenas_lock) #define arenas_lrun_i_index JEMALLOC_N(arenas_lrun_i_index) +#define arenas_tls JEMALLOC_N(arenas_tls) +#define arenas_tsd_boot JEMALLOC_N(arenas_tsd_boot) +#define arenas_tsd_cleanup_wrapper JEMALLOC_N(arenas_tsd_cleanup_wrapper) +#define arenas_tsd_get JEMALLOC_N(arenas_tsd_get) +#define arenas_tsd_set JEMALLOC_N(arenas_tsd_set) +#define atomic_add_u JEMALLOC_N(atomic_add_u) #define atomic_add_uint32 JEMALLOC_N(atomic_add_uint32) #define atomic_add_uint64 JEMALLOC_N(atomic_add_uint64) +#define atomic_add_z JEMALLOC_N(atomic_add_z) +#define atomic_sub_u JEMALLOC_N(atomic_sub_u) #define atomic_sub_uint32 JEMALLOC_N(atomic_sub_uint32) #define atomic_sub_uint64 JEMALLOC_N(atomic_sub_uint64) +#define atomic_sub_z JEMALLOC_N(atomic_sub_z) #define base_alloc JEMALLOC_N(base_alloc) #define base_boot JEMALLOC_N(base_boot) +#define base_calloc JEMALLOC_N(base_calloc) #define base_node_alloc JEMALLOC_N(base_node_alloc) #define base_node_dealloc JEMALLOC_N(base_node_dealloc) +#define base_postfork_child JEMALLOC_N(base_postfork_child) +#define base_postfork_parent JEMALLOC_N(base_postfork_parent) +#define base_prefork JEMALLOC_N(base_prefork) #define bitmap_full JEMALLOC_N(bitmap_full) #define bitmap_get JEMALLOC_N(bitmap_get) #define bitmap_info_init JEMALLOC_N(bitmap_info_init) @@ -47,19 +95,19 @@ #define chunk_alloc JEMALLOC_N(chunk_alloc) #define chunk_alloc_dss JEMALLOC_N(chunk_alloc_dss) #define chunk_alloc_mmap JEMALLOC_N(chunk_alloc_mmap) -#define chunk_alloc_mmap_noreserve JEMALLOC_N(chunk_alloc_mmap_noreserve) -#define chunk_alloc_swap JEMALLOC_N(chunk_alloc_swap) #define chunk_boot JEMALLOC_N(chunk_boot) #define chunk_dealloc JEMALLOC_N(chunk_dealloc) -#define chunk_dealloc_dss JEMALLOC_N(chunk_dealloc_dss) #define chunk_dealloc_mmap JEMALLOC_N(chunk_dealloc_mmap) -#define chunk_dealloc_swap JEMALLOC_N(chunk_dealloc_swap) #define chunk_dss_boot JEMALLOC_N(chunk_dss_boot) +#define chunk_dss_postfork_child JEMALLOC_N(chunk_dss_postfork_child) +#define chunk_dss_postfork_parent JEMALLOC_N(chunk_dss_postfork_parent) +#define chunk_dss_prefork JEMALLOC_N(chunk_dss_prefork) #define chunk_in_dss JEMALLOC_N(chunk_in_dss) -#define chunk_in_swap JEMALLOC_N(chunk_in_swap) -#define chunk_mmap_boot JEMALLOC_N(chunk_mmap_boot) -#define chunk_swap_boot JEMALLOC_N(chunk_swap_boot) -#define chunk_swap_enable JEMALLOC_N(chunk_swap_enable) +#define chunk_npages JEMALLOC_N(chunk_npages) +#define chunks_mtx JEMALLOC_N(chunks_mtx) +#define chunks_rtree JEMALLOC_N(chunks_rtree) +#define chunksize JEMALLOC_N(chunksize) +#define chunksize_mask JEMALLOC_N(chunksize_mask) #define ckh_bucket_search JEMALLOC_N(ckh_bucket_search) #define ckh_count JEMALLOC_N(ckh_count) #define ckh_delete JEMALLOC_N(ckh_delete) @@ -77,7 +125,6 @@ #define ckh_string_keycomp JEMALLOC_N(ckh_string_keycomp) #define ckh_try_bucket_insert JEMALLOC_N(ckh_try_bucket_insert) #define ckh_try_insert JEMALLOC_N(ckh_try_insert) -#define create_zone JEMALLOC_N(create_zone) #define ctl_boot JEMALLOC_N(ctl_boot) #define ctl_bymib JEMALLOC_N(ctl_bymib) #define ctl_byname JEMALLOC_N(ctl_byname) @@ -115,10 +162,17 @@ #define extent_tree_szad_reverse_iter_start JEMALLOC_N(extent_tree_szad_reverse_iter_start) #define extent_tree_szad_search JEMALLOC_N(extent_tree_szad_search) #define hash JEMALLOC_N(hash) +#define huge_allocated JEMALLOC_N(huge_allocated) #define huge_boot JEMALLOC_N(huge_boot) #define huge_dalloc JEMALLOC_N(huge_dalloc) #define huge_malloc JEMALLOC_N(huge_malloc) +#define huge_mtx JEMALLOC_N(huge_mtx) +#define huge_ndalloc JEMALLOC_N(huge_ndalloc) +#define huge_nmalloc JEMALLOC_N(huge_nmalloc) #define huge_palloc JEMALLOC_N(huge_palloc) +#define huge_postfork_child JEMALLOC_N(huge_postfork_child) +#define huge_postfork_parent JEMALLOC_N(huge_postfork_parent) +#define huge_prefork JEMALLOC_N(huge_prefork) #define huge_prof_ctx_get JEMALLOC_N(huge_prof_ctx_get) #define huge_prof_ctx_set JEMALLOC_N(huge_prof_ctx_set) #define huge_ralloc JEMALLOC_N(huge_ralloc) @@ -129,21 +183,63 @@ #define idalloc JEMALLOC_N(idalloc) #define imalloc JEMALLOC_N(imalloc) #define ipalloc JEMALLOC_N(ipalloc) +#define iqalloc JEMALLOC_N(iqalloc) #define iralloc JEMALLOC_N(iralloc) #define isalloc JEMALLOC_N(isalloc) +#define isthreaded JEMALLOC_N(isthreaded) #define ivsalloc JEMALLOC_N(ivsalloc) -#define jemalloc_darwin_init JEMALLOC_N(jemalloc_darwin_init) -#define jemalloc_postfork JEMALLOC_N(jemalloc_postfork) +#define jemalloc_postfork_child JEMALLOC_N(jemalloc_postfork_child) +#define jemalloc_postfork_parent JEMALLOC_N(jemalloc_postfork_parent) #define jemalloc_prefork JEMALLOC_N(jemalloc_prefork) #define malloc_cprintf JEMALLOC_N(malloc_cprintf) -#define malloc_mutex_destroy JEMALLOC_N(malloc_mutex_destroy) #define malloc_mutex_init JEMALLOC_N(malloc_mutex_init) #define malloc_mutex_lock JEMALLOC_N(malloc_mutex_lock) -#define malloc_mutex_trylock JEMALLOC_N(malloc_mutex_trylock) +#define malloc_mutex_postfork_child JEMALLOC_N(malloc_mutex_postfork_child) +#define malloc_mutex_postfork_parent JEMALLOC_N(malloc_mutex_postfork_parent) +#define malloc_mutex_prefork JEMALLOC_N(malloc_mutex_prefork) #define malloc_mutex_unlock JEMALLOC_N(malloc_mutex_unlock) #define malloc_printf JEMALLOC_N(malloc_printf) +#define malloc_snprintf JEMALLOC_N(malloc_snprintf) +#define malloc_strtoumax JEMALLOC_N(malloc_strtoumax) +#define malloc_tsd_boot JEMALLOC_N(malloc_tsd_boot) +#define malloc_tsd_cleanup_register JEMALLOC_N(malloc_tsd_cleanup_register) +#define malloc_tsd_dalloc JEMALLOC_N(malloc_tsd_dalloc) +#define malloc_tsd_malloc JEMALLOC_N(malloc_tsd_malloc) +#define malloc_tsd_no_cleanup JEMALLOC_N(malloc_tsd_no_cleanup) +#define malloc_vcprintf JEMALLOC_N(malloc_vcprintf) +#define malloc_vsnprintf JEMALLOC_N(malloc_vsnprintf) #define malloc_write JEMALLOC_N(malloc_write) +#define map_bias JEMALLOC_N(map_bias) #define mb_write JEMALLOC_N(mb_write) +#define mutex_boot JEMALLOC_N(mutex_boot) +#define narenas JEMALLOC_N(narenas) +#define ncpus JEMALLOC_N(ncpus) +#define nhbins JEMALLOC_N(nhbins) +#define opt_abort JEMALLOC_N(opt_abort) +#define opt_junk JEMALLOC_N(opt_junk) +#define opt_lg_chunk JEMALLOC_N(opt_lg_chunk) +#define opt_lg_dirty_mult JEMALLOC_N(opt_lg_dirty_mult) +#define opt_lg_prof_interval JEMALLOC_N(opt_lg_prof_interval) +#define opt_lg_prof_sample JEMALLOC_N(opt_lg_prof_sample) +#define opt_lg_tcache_max JEMALLOC_N(opt_lg_tcache_max) +#define opt_narenas JEMALLOC_N(opt_narenas) +#define opt_prof JEMALLOC_N(opt_prof) +#define opt_prof_accum JEMALLOC_N(opt_prof_accum) +#define opt_prof_active JEMALLOC_N(opt_prof_active) +#define opt_prof_final JEMALLOC_N(opt_prof_final) +#define opt_prof_gdump JEMALLOC_N(opt_prof_gdump) +#define opt_prof_leak JEMALLOC_N(opt_prof_leak) +#define opt_prof_prefix JEMALLOC_N(opt_prof_prefix) +#define opt_quarantine JEMALLOC_N(opt_quarantine) +#define opt_redzone JEMALLOC_N(opt_redzone) +#define opt_stats_print JEMALLOC_N(opt_stats_print) +#define opt_tcache JEMALLOC_N(opt_tcache) +#define opt_utrace JEMALLOC_N(opt_utrace) +#define opt_valgrind JEMALLOC_N(opt_valgrind) +#define opt_xmalloc JEMALLOC_N(opt_xmalloc) +#define opt_zero JEMALLOC_N(opt_zero) +#define p2rz JEMALLOC_N(p2rz) +#define pages_purge JEMALLOC_N(pages_purge) #define pow2_ceil JEMALLOC_N(pow2_ceil) #define prof_backtrace JEMALLOC_N(prof_backtrace) #define prof_boot0 JEMALLOC_N(prof_boot0) @@ -154,14 +250,31 @@ #define prof_free JEMALLOC_N(prof_free) #define prof_gdump JEMALLOC_N(prof_gdump) #define prof_idump JEMALLOC_N(prof_idump) +#define prof_interval JEMALLOC_N(prof_interval) #define prof_lookup JEMALLOC_N(prof_lookup) #define prof_malloc JEMALLOC_N(prof_malloc) #define prof_mdump JEMALLOC_N(prof_mdump) +#define prof_promote JEMALLOC_N(prof_promote) #define prof_realloc JEMALLOC_N(prof_realloc) #define prof_sample_accum_update JEMALLOC_N(prof_sample_accum_update) #define prof_sample_threshold_update JEMALLOC_N(prof_sample_threshold_update) +#define prof_tdata_booted JEMALLOC_N(prof_tdata_booted) +#define prof_tdata_cleanup JEMALLOC_N(prof_tdata_cleanup) +#define prof_tdata_get JEMALLOC_N(prof_tdata_get) #define prof_tdata_init JEMALLOC_N(prof_tdata_init) -#define pthread_create JEMALLOC_N(pthread_create) +#define prof_tdata_initialized JEMALLOC_N(prof_tdata_initialized) +#define prof_tdata_tls JEMALLOC_N(prof_tdata_tls) +#define prof_tdata_tsd_boot JEMALLOC_N(prof_tdata_tsd_boot) +#define prof_tdata_tsd_cleanup_wrapper JEMALLOC_N(prof_tdata_tsd_cleanup_wrapper) +#define prof_tdata_tsd_get JEMALLOC_N(prof_tdata_tsd_get) +#define prof_tdata_tsd_set JEMALLOC_N(prof_tdata_tsd_set) +#define quarantine JEMALLOC_N(quarantine) +#define quarantine_boot JEMALLOC_N(quarantine_boot) +#define quarantine_tsd_boot JEMALLOC_N(quarantine_tsd_boot) +#define quarantine_tsd_cleanup_wrapper JEMALLOC_N(quarantine_tsd_cleanup_wrapper) +#define quarantine_tsd_get JEMALLOC_N(quarantine_tsd_get) +#define quarantine_tsd_set JEMALLOC_N(quarantine_tsd_set) +#define register_zone JEMALLOC_N(register_zone) #define rtree_get JEMALLOC_N(rtree_get) #define rtree_get_locked JEMALLOC_N(rtree_get_locked) #define rtree_new JEMALLOC_N(rtree_new) @@ -171,25 +284,56 @@ #define stats_arenas_i_bins_j_index JEMALLOC_N(stats_arenas_i_bins_j_index) #define stats_arenas_i_index JEMALLOC_N(stats_arenas_i_index) #define stats_arenas_i_lruns_j_index JEMALLOC_N(stats_arenas_i_lruns_j_index) +#define stats_cactive JEMALLOC_N(stats_cactive) #define stats_cactive_add JEMALLOC_N(stats_cactive_add) #define stats_cactive_get JEMALLOC_N(stats_cactive_get) #define stats_cactive_sub JEMALLOC_N(stats_cactive_sub) +#define stats_chunks JEMALLOC_N(stats_chunks) #define stats_print JEMALLOC_N(stats_print) -#define szone2ozone JEMALLOC_N(szone2ozone) #define tcache_alloc_easy JEMALLOC_N(tcache_alloc_easy) #define tcache_alloc_large JEMALLOC_N(tcache_alloc_large) #define tcache_alloc_small JEMALLOC_N(tcache_alloc_small) #define tcache_alloc_small_hard JEMALLOC_N(tcache_alloc_small_hard) +#define tcache_arena_associate JEMALLOC_N(tcache_arena_associate) +#define tcache_arena_dissociate JEMALLOC_N(tcache_arena_dissociate) #define tcache_bin_flush_large JEMALLOC_N(tcache_bin_flush_large) #define tcache_bin_flush_small JEMALLOC_N(tcache_bin_flush_small) -#define tcache_boot JEMALLOC_N(tcache_boot) +#define tcache_bin_info JEMALLOC_N(tcache_bin_info) +#define tcache_boot0 JEMALLOC_N(tcache_boot0) +#define tcache_boot1 JEMALLOC_N(tcache_boot1) +#define tcache_booted JEMALLOC_N(tcache_booted) #define tcache_create JEMALLOC_N(tcache_create) #define tcache_dalloc_large JEMALLOC_N(tcache_dalloc_large) #define tcache_dalloc_small JEMALLOC_N(tcache_dalloc_small) #define tcache_destroy JEMALLOC_N(tcache_destroy) +#define tcache_enabled_booted JEMALLOC_N(tcache_enabled_booted) +#define tcache_enabled_get JEMALLOC_N(tcache_enabled_get) +#define tcache_enabled_initialized JEMALLOC_N(tcache_enabled_initialized) +#define tcache_enabled_set JEMALLOC_N(tcache_enabled_set) +#define tcache_enabled_tls JEMALLOC_N(tcache_enabled_tls) +#define tcache_enabled_tsd_boot JEMALLOC_N(tcache_enabled_tsd_boot) +#define tcache_enabled_tsd_cleanup_wrapper JEMALLOC_N(tcache_enabled_tsd_cleanup_wrapper) +#define tcache_enabled_tsd_get JEMALLOC_N(tcache_enabled_tsd_get) +#define tcache_enabled_tsd_set JEMALLOC_N(tcache_enabled_tsd_set) #define tcache_event JEMALLOC_N(tcache_event) +#define tcache_event_hard JEMALLOC_N(tcache_event_hard) +#define tcache_flush JEMALLOC_N(tcache_flush) #define tcache_get JEMALLOC_N(tcache_get) +#define tcache_initialized JEMALLOC_N(tcache_initialized) +#define tcache_maxclass JEMALLOC_N(tcache_maxclass) +#define tcache_salloc JEMALLOC_N(tcache_salloc) #define tcache_stats_merge JEMALLOC_N(tcache_stats_merge) -#define thread_allocated_get JEMALLOC_N(thread_allocated_get) -#define thread_allocated_get_hard JEMALLOC_N(thread_allocated_get_hard) -#define u2s JEMALLOC_N(u2s) +#define tcache_thread_cleanup JEMALLOC_N(tcache_thread_cleanup) +#define tcache_tls JEMALLOC_N(tcache_tls) +#define tcache_tsd_boot JEMALLOC_N(tcache_tsd_boot) +#define tcache_tsd_cleanup_wrapper JEMALLOC_N(tcache_tsd_cleanup_wrapper) +#define tcache_tsd_get JEMALLOC_N(tcache_tsd_get) +#define tcache_tsd_set JEMALLOC_N(tcache_tsd_set) +#define thread_allocated_booted JEMALLOC_N(thread_allocated_booted) +#define thread_allocated_initialized JEMALLOC_N(thread_allocated_initialized) +#define thread_allocated_tls JEMALLOC_N(thread_allocated_tls) +#define thread_allocated_tsd_boot JEMALLOC_N(thread_allocated_tsd_boot) +#define thread_allocated_tsd_cleanup_wrapper JEMALLOC_N(thread_allocated_tsd_cleanup_wrapper) +#define thread_allocated_tsd_get JEMALLOC_N(thread_allocated_tsd_get) +#define thread_allocated_tsd_set JEMALLOC_N(thread_allocated_tsd_set) +#define u2rz JEMALLOC_N(u2rz) diff --git a/include/jemalloc/internal/prn.h b/include/jemalloc/internal/prng.h similarity index 89% rename from include/jemalloc/internal/prn.h rename to include/jemalloc/internal/prng.h index 0709d708..83a5462b 100644 --- a/include/jemalloc/internal/prn.h +++ b/include/jemalloc/internal/prng.h @@ -4,7 +4,7 @@ /* * Simple linear congruential pseudo-random number generator: * - * prn(y) = (a*x + c) % m + * prng(y) = (a*x + c) % m * * where the following constants ensure maximal period: * @@ -25,7 +25,7 @@ * uint32_t state : Seed value. * const uint32_t a, c : See above discussion. */ -#define prn32(r, lg_range, state, a, c) do { \ +#define prng32(r, lg_range, state, a, c) do { \ assert(lg_range > 0); \ assert(lg_range <= 32); \ \ @@ -34,8 +34,8 @@ r >>= (32 - lg_range); \ } while (false) -/* Same as prn32(), but 64 bits of pseudo-randomness, using uint64_t. */ -#define prn64(r, lg_range, state, a, c) do { \ +/* Same as prng32(), but 64 bits of pseudo-randomness, using uint64_t. */ +#define prng64(r, lg_range, state, a, c) do { \ assert(lg_range > 0); \ assert(lg_range <= 64); \ \ diff --git a/include/jemalloc/internal/prof.h b/include/jemalloc/internal/prof.h index e9064ba6..c3e3f9e4 100644 --- a/include/jemalloc/internal/prof.h +++ b/include/jemalloc/internal/prof.h @@ -1,4 +1,3 @@ -#ifdef JEMALLOC_PROF /******************************************************************************/ #ifdef JEMALLOC_H_TYPES @@ -10,28 +9,41 @@ typedef struct prof_tdata_s prof_tdata_t; /* Option defaults. */ #define PROF_PREFIX_DEFAULT "jeprof" -#define LG_PROF_BT_MAX_DEFAULT 7 -#define LG_PROF_SAMPLE_DEFAULT 0 +#define LG_PROF_SAMPLE_DEFAULT 19 #define LG_PROF_INTERVAL_DEFAULT -1 -#define LG_PROF_TCMAX_DEFAULT -1 /* - * Hard limit on stack backtrace depth. Note that the version of - * prof_backtrace() that is based on __builtin_return_address() necessarily has - * a hard-coded number of backtrace frame handlers. + * Hard limit on stack backtrace depth. The version of prof_backtrace() that + * is based on __builtin_return_address() necessarily has a hard-coded number + * of backtrace frame handlers, and should be kept in sync with this setting. */ -#if (defined(JEMALLOC_PROF_LIBGCC) || defined(JEMALLOC_PROF_LIBUNWIND)) -# define LG_PROF_BT_MAX ((ZU(1) << (LG_SIZEOF_PTR+3)) - 1) -#else -# define LG_PROF_BT_MAX 7 /* >= LG_PROF_BT_MAX_DEFAULT */ -#endif -#define PROF_BT_MAX (1U << LG_PROF_BT_MAX) +#define PROF_BT_MAX 128 + +/* Maximum number of backtraces to store in each per thread LRU cache. */ +#define PROF_TCMAX 1024 /* Initial hash table size. */ -#define PROF_CKH_MINITEMS 64 +#define PROF_CKH_MINITEMS 64 /* Size of memory buffer to use when writing dump files. */ -#define PROF_DUMP_BUF_SIZE 65536 +#define PROF_DUMP_BUFSIZE 65536 + +/* Size of stack-allocated buffer used by prof_printf(). */ +#define PROF_PRINTF_BUFSIZE 128 + +/* + * Number of mutexes shared among all ctx's. No space is allocated for these + * unless profiling is enabled, so it's okay to over-provision. + */ +#define PROF_NCTX_LOCKS 1024 + +/* + * prof_tdata pointers close to NULL are used to encode state information that + * is used for cleaning up during thread shutdown. + */ +#define PROF_TDATA_STATE_REINCARNATED ((prof_tdata_t *)(uintptr_t)1) +#define PROF_TDATA_STATE_PURGATORY ((prof_tdata_t *)(uintptr_t)2) +#define PROF_TDATA_STATE_MAX PROF_TDATA_STATE_PURGATORY #endif /* JEMALLOC_H_TYPES */ /******************************************************************************/ @@ -109,8 +121,18 @@ struct prof_ctx_s { /* Associated backtrace. */ prof_bt_t *bt; - /* Protects cnt_merged and cnts_ql. */ - malloc_mutex_t lock; + /* Protects nlimbo, cnt_merged, and cnts_ql. */ + malloc_mutex_t *lock; + + /* + * Number of threads that currently cause this ctx to be in a state of + * limbo due to one of: + * - Initializing per thread counters associated with this ctx. + * - Preparing to destroy this ctx. + * nlimbo must be 1 (single destroyer) in order to safely destroy the + * ctx. + */ + unsigned nlimbo; /* Temporary storage for summation during dump. */ prof_cnt_t cnt_summed; @@ -145,9 +167,14 @@ struct prof_tdata_s { void **vec; /* Sampling state. */ - uint64_t prn_state; + uint64_t prng_state; uint64_t threshold; uint64_t accum; + + /* State used to avoid dumping while operating on prof internals. */ + bool enq; + bool enq_idump; + bool enq_gdump; }; #endif /* JEMALLOC_H_STRUCTS */ @@ -162,13 +189,12 @@ extern bool opt_prof; * to notice state changes. */ extern bool opt_prof_active; -extern size_t opt_lg_prof_bt_max; /* Maximum backtrace depth. */ extern size_t opt_lg_prof_sample; /* Mean bytes between samples. */ extern ssize_t opt_lg_prof_interval; /* lg(prof_interval). */ extern bool opt_prof_gdump; /* High-water memory dumping. */ +extern bool opt_prof_final; /* Final profile dumping. */ extern bool opt_prof_leak; /* Dump leak summary at exit. */ extern bool opt_prof_accum; /* Report cumulative bytes. */ -extern ssize_t opt_lg_prof_tcmax; /* lg(max per thread bactrace cache) */ extern char opt_prof_prefix[PATH_MAX + 1]; /* @@ -186,39 +212,14 @@ extern uint64_t prof_interval; */ extern bool prof_promote; -/* (1U << opt_lg_prof_bt_max). */ -extern unsigned prof_bt_max; - -/* Thread-specific backtrace cache, used to reduce bt2ctx contention. */ -#ifndef NO_TLS -extern __thread prof_tdata_t *prof_tdata_tls - JEMALLOC_ATTR(tls_model("initial-exec")); -# define PROF_TCACHE_GET() prof_tdata_tls -# define PROF_TCACHE_SET(v) do { \ - prof_tdata_tls = (v); \ - pthread_setspecific(prof_tdata_tsd, (void *)(v)); \ -} while (0) -#else -# define PROF_TCACHE_GET() \ - ((prof_tdata_t *)pthread_getspecific(prof_tdata_tsd)) -# define PROF_TCACHE_SET(v) do { \ - pthread_setspecific(prof_tdata_tsd, (void *)(v)); \ -} while (0) -#endif -/* - * Same contents as b2cnt_tls, but initialized such that the TSD destructor is - * called when a thread exits, so that prof_tdata_tls contents can be merged, - * unlinked, and deallocated. - */ -extern pthread_key_t prof_tdata_tsd; - void bt_init(prof_bt_t *bt, void **vec); -void prof_backtrace(prof_bt_t *bt, unsigned nignore, unsigned max); +void prof_backtrace(prof_bt_t *bt, unsigned nignore); prof_thr_cnt_t *prof_lookup(prof_bt_t *bt); void prof_idump(void); bool prof_mdump(const char *filename); void prof_gdump(void); prof_tdata_t *prof_tdata_init(void); +void prof_tdata_cleanup(void *arg); void prof_boot0(void); void prof_boot1(void); bool prof_boot2(void); @@ -233,13 +234,13 @@ bool prof_boot2(void); \ assert(size == s2u(size)); \ \ - prof_tdata = PROF_TCACHE_GET(); \ - if (prof_tdata == NULL) { \ - prof_tdata = prof_tdata_init(); \ - if (prof_tdata == NULL) { \ + prof_tdata = prof_tdata_get(); \ + if ((uintptr_t)prof_tdata <= (uintptr_t)PROF_TDATA_STATE_MAX) { \ + if (prof_tdata != NULL) \ + ret = (prof_thr_cnt_t *)(uintptr_t)1U; \ + else \ ret = NULL; \ - break; \ - } \ + break; \ } \ \ if (opt_prof_active == false) { \ @@ -249,13 +250,13 @@ bool prof_boot2(void); /* Don't bother with sampling logic, since sampling */\ /* interval is 1. */\ bt_init(&bt, prof_tdata->vec); \ - prof_backtrace(&bt, nignore, prof_bt_max); \ + prof_backtrace(&bt, nignore); \ ret = prof_lookup(&bt); \ } else { \ if (prof_tdata->threshold == 0) { \ /* Initialize. Seed the prng differently for */\ /* each thread. */\ - prof_tdata->prn_state = \ + prof_tdata->prng_state = \ (uint64_t)(uintptr_t)&size; \ prof_sample_threshold_update(prof_tdata); \ } \ @@ -272,7 +273,7 @@ bool prof_boot2(void); if (size >= prof_tdata->threshold - \ prof_tdata->accum) { \ bt_init(&bt, prof_tdata->vec); \ - prof_backtrace(&bt, nignore, prof_bt_max); \ + prof_backtrace(&bt, nignore); \ ret = prof_lookup(&bt); \ } else \ ret = (prof_thr_cnt_t *)(uintptr_t)1U; \ @@ -280,6 +281,9 @@ bool prof_boot2(void); } while (0) #ifndef JEMALLOC_ENABLE_INLINE +malloc_tsd_protos(JEMALLOC_ATTR(unused), prof_tdata, prof_tdata_t *) + +prof_tdata_t *prof_tdata_get(void); void prof_sample_threshold_update(prof_tdata_t *prof_tdata); prof_ctx_t *prof_ctx_get(const void *ptr); void prof_ctx_set(const void *ptr, prof_ctx_t *ctx); @@ -291,12 +295,35 @@ void prof_free(const void *ptr, size_t size); #endif #if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_PROF_C_)) +/* Thread-specific backtrace cache, used to reduce bt2ctx contention. */ +malloc_tsd_externs(prof_tdata, prof_tdata_t *) +malloc_tsd_funcs(JEMALLOC_INLINE, prof_tdata, prof_tdata_t *, NULL, + prof_tdata_cleanup) + +JEMALLOC_INLINE prof_tdata_t * +prof_tdata_get(void) +{ + prof_tdata_t *prof_tdata; + + cassert(config_prof); + + prof_tdata = *prof_tdata_tsd_get(); + if ((uintptr_t)prof_tdata <= (uintptr_t)PROF_TDATA_STATE_MAX) { + if (prof_tdata == NULL) + prof_tdata = prof_tdata_init(); + } + + return (prof_tdata); +} + JEMALLOC_INLINE void prof_sample_threshold_update(prof_tdata_t *prof_tdata) { uint64_t r; double u; + cassert(config_prof); + /* * Compute sample threshold as a geometrically distributed random * variable with mean (2^opt_lg_prof_sample). @@ -315,8 +342,8 @@ prof_sample_threshold_update(prof_tdata_t *prof_tdata) * pp 500 * (http://cg.scs.carleton.ca/~luc/rnbookindex.html) */ - prn64(r, 53, prof_tdata->prn_state, - (uint64_t)6364136223846793005LLU, (uint64_t)1442695040888963407LLU); + prng64(r, 53, prof_tdata->prng_state, + UINT64_C(6364136223846793005), UINT64_C(1442695040888963407)); u = (double)r * (1.0/9007199254740992.0L); prof_tdata->threshold = (uint64_t)(log(u) / log(1.0 - (1.0 / (double)((uint64_t)1U << opt_lg_prof_sample)))) @@ -329,13 +356,12 @@ prof_ctx_get(const void *ptr) prof_ctx_t *ret; arena_chunk_t *chunk; + cassert(config_prof); assert(ptr != NULL); chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr); if (chunk != ptr) { /* Region. */ - dassert(chunk->arena->magic == ARENA_MAGIC); - ret = arena_prof_ctx_get(ptr); } else ret = huge_prof_ctx_get(ptr); @@ -348,13 +374,12 @@ prof_ctx_set(const void *ptr, prof_ctx_t *ctx) { arena_chunk_t *chunk; + cassert(config_prof); assert(ptr != NULL); chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr); if (chunk != ptr) { /* Region. */ - dassert(chunk->arena->magic == ARENA_MAGIC); - arena_prof_ctx_set(ptr, ctx); } else huge_prof_ctx_set(ptr, ctx); @@ -365,11 +390,13 @@ prof_sample_accum_update(size_t size) { prof_tdata_t *prof_tdata; + cassert(config_prof); /* Sampling logic is unnecessary if the interval is 1. */ assert(opt_lg_prof_sample != 0); - prof_tdata = PROF_TCACHE_GET(); - assert(prof_tdata != NULL); + prof_tdata = *prof_tdata_tsd_get(); + if ((uintptr_t)prof_tdata <= (uintptr_t)PROF_TDATA_STATE_MAX) + return (true); /* Take care to avoid integer overflow. */ if (size >= prof_tdata->threshold - prof_tdata->accum) { @@ -391,8 +418,9 @@ JEMALLOC_INLINE void prof_malloc(const void *ptr, size_t size, prof_thr_cnt_t *cnt) { + cassert(config_prof); assert(ptr != NULL); - assert(size == isalloc(ptr)); + assert(size == isalloc(ptr, true)); if (opt_lg_prof_sample != 0) { if (prof_sample_accum_update(size)) { @@ -437,10 +465,11 @@ prof_realloc(const void *ptr, size_t size, prof_thr_cnt_t *cnt, { prof_thr_cnt_t *told_cnt; + cassert(config_prof); assert(ptr != NULL || (uintptr_t)cnt <= (uintptr_t)1U); if (ptr != NULL) { - assert(size == isalloc(ptr)); + assert(size == isalloc(ptr, true)); if (opt_lg_prof_sample != 0) { if (prof_sample_accum_update(size)) { /* @@ -463,10 +492,10 @@ prof_realloc(const void *ptr, size_t size, prof_thr_cnt_t *cnt, * It's too late to propagate OOM for this realloc(), * so operate directly on old_cnt->ctx->cnt_merged. */ - malloc_mutex_lock(&old_ctx->lock); + malloc_mutex_lock(old_ctx->lock); old_ctx->cnt_merged.curobjs--; old_ctx->cnt_merged.curbytes -= old_size; - malloc_mutex_unlock(&old_ctx->lock); + malloc_mutex_unlock(old_ctx->lock); told_cnt = (prof_thr_cnt_t *)(uintptr_t)1U; } } else @@ -510,9 +539,12 @@ prof_free(const void *ptr, size_t size) { prof_ctx_t *ctx = prof_ctx_get(ptr); + cassert(config_prof); + if ((uintptr_t)ctx > (uintptr_t)1) { - assert(size == isalloc(ptr)); - prof_thr_cnt_t *tcnt = prof_lookup(ctx->bt); + prof_thr_cnt_t *tcnt; + assert(size == isalloc(ptr, true)); + tcnt = prof_lookup(ctx->bt); if (tcnt != NULL) { tcnt->epoch++; @@ -533,10 +565,10 @@ prof_free(const void *ptr, size_t size) * OOM during free() cannot be propagated, so operate * directly on cnt->ctx->cnt_merged. */ - malloc_mutex_lock(&ctx->lock); + malloc_mutex_lock(ctx->lock); ctx->cnt_merged.curobjs--; ctx->cnt_merged.curbytes -= size; - malloc_mutex_unlock(&ctx->lock); + malloc_mutex_unlock(ctx->lock); } } } @@ -544,4 +576,3 @@ prof_free(const void *ptr, size_t size) #endif /* JEMALLOC_H_INLINES */ /******************************************************************************/ -#endif /* JEMALLOC_PROF */ diff --git a/include/jemalloc/internal/zone.h b/include/jemalloc/internal/quarantine.h similarity index 77% rename from include/jemalloc/internal/zone.h rename to include/jemalloc/internal/quarantine.h index 859b529d..38f3d696 100644 --- a/include/jemalloc/internal/zone.h +++ b/include/jemalloc/internal/quarantine.h @@ -1,9 +1,9 @@ -#ifndef JEMALLOC_ZONE -# error "This source file is for zones on Darwin (OS X)." -#endif /******************************************************************************/ #ifdef JEMALLOC_H_TYPES +/* Default per thread quarantine size if valgrind is enabled. */ +#define JEMALLOC_VALGRIND_QUARANTINE_DEFAULT (ZU(1) << 24) + #endif /* JEMALLOC_H_TYPES */ /******************************************************************************/ #ifdef JEMALLOC_H_STRUCTS @@ -12,8 +12,8 @@ /******************************************************************************/ #ifdef JEMALLOC_H_EXTERNS -malloc_zone_t *create_zone(void); -void szone2ozone(malloc_zone_t *zone); +void quarantine(void *ptr); +bool quarantine_boot(void); #endif /* JEMALLOC_H_EXTERNS */ /******************************************************************************/ @@ -21,3 +21,4 @@ void szone2ozone(malloc_zone_t *zone); #endif /* JEMALLOC_H_INLINES */ /******************************************************************************/ + diff --git a/include/jemalloc/internal/rb.h b/include/jemalloc/internal/rb.h index ee9b009d..7b675f09 100644 --- a/include/jemalloc/internal/rb.h +++ b/include/jemalloc/internal/rb.h @@ -223,88 +223,88 @@ a_prefix##reverse_iter(a_rbt_type *rbtree, a_type *start, \ * The following API is generated: * * static void - * ex_new(ex_t *extree); + * ex_new(ex_t *tree); * Description: Initialize a red-black tree structure. * Args: - * extree: Pointer to an uninitialized red-black tree object. + * tree: Pointer to an uninitialized red-black tree object. * * static ex_node_t * - * ex_first(ex_t *extree); + * ex_first(ex_t *tree); * static ex_node_t * - * ex_last(ex_t *extree); - * Description: Get the first/last node in extree. + * ex_last(ex_t *tree); + * Description: Get the first/last node in tree. * Args: - * extree: Pointer to an initialized red-black tree object. - * Ret: First/last node in extree, or NULL if extree is empty. + * tree: Pointer to an initialized red-black tree object. + * Ret: First/last node in tree, or NULL if tree is empty. * * static ex_node_t * - * ex_next(ex_t *extree, ex_node_t *node); + * ex_next(ex_t *tree, ex_node_t *node); * static ex_node_t * - * ex_prev(ex_t *extree, ex_node_t *node); + * ex_prev(ex_t *tree, ex_node_t *node); * Description: Get node's successor/predecessor. * Args: - * extree: Pointer to an initialized red-black tree object. - * node : A node in extree. - * Ret: node's successor/predecessor in extree, or NULL if node is + * tree: Pointer to an initialized red-black tree object. + * node: A node in tree. + * Ret: node's successor/predecessor in tree, or NULL if node is * last/first. * * static ex_node_t * - * ex_search(ex_t *extree, ex_node_t *key); + * ex_search(ex_t *tree, ex_node_t *key); * Description: Search for node that matches key. * Args: - * extree: Pointer to an initialized red-black tree object. - * key : Search key. - * Ret: Node in extree that matches key, or NULL if no match. + * tree: Pointer to an initialized red-black tree object. + * key : Search key. + * Ret: Node in tree that matches key, or NULL if no match. * * static ex_node_t * - * ex_nsearch(ex_t *extree, ex_node_t *key); + * ex_nsearch(ex_t *tree, ex_node_t *key); * static ex_node_t * - * ex_psearch(ex_t *extree, ex_node_t *key); + * ex_psearch(ex_t *tree, ex_node_t *key); * Description: Search for node that matches key. If no match is found, * return what would be key's successor/predecessor, were - * key in extree. + * key in tree. * Args: - * extree: Pointer to an initialized red-black tree object. - * key : Search key. - * Ret: Node in extree that matches key, or if no match, hypothetical - * node's successor/predecessor (NULL if no successor/predecessor). + * tree: Pointer to an initialized red-black tree object. + * key : Search key. + * Ret: Node in tree that matches key, or if no match, hypothetical node's + * successor/predecessor (NULL if no successor/predecessor). * * static void - * ex_insert(ex_t *extree, ex_node_t *node); - * Description: Insert node into extree. + * ex_insert(ex_t *tree, ex_node_t *node); + * Description: Insert node into tree. * Args: - * extree: Pointer to an initialized red-black tree object. - * node : Node to be inserted into extree. + * tree: Pointer to an initialized red-black tree object. + * node: Node to be inserted into tree. * * static void - * ex_remove(ex_t *extree, ex_node_t *node); - * Description: Remove node from extree. + * ex_remove(ex_t *tree, ex_node_t *node); + * Description: Remove node from tree. * Args: - * extree: Pointer to an initialized red-black tree object. - * node : Node in extree to be removed. + * tree: Pointer to an initialized red-black tree object. + * node: Node in tree to be removed. * * static ex_node_t * - * ex_iter(ex_t *extree, ex_node_t *start, ex_node_t *(*cb)(ex_t *, + * ex_iter(ex_t *tree, ex_node_t *start, ex_node_t *(*cb)(ex_t *, * ex_node_t *, void *), void *arg); * static ex_node_t * - * ex_reverse_iter(ex_t *extree, ex_node_t *start, ex_node *(*cb)(ex_t *, + * ex_reverse_iter(ex_t *tree, ex_node_t *start, ex_node *(*cb)(ex_t *, * ex_node_t *, void *), void *arg); - * Description: Iterate forward/backward over extree, starting at node. - * If extree is modified, iteration must be immediately + * Description: Iterate forward/backward over tree, starting at node. If + * tree is modified, iteration must be immediately * terminated by the callback function that causes the * modification. * Args: - * extree: Pointer to an initialized red-black tree object. - * start : Node at which to start iteration, or NULL to start at - * first/last node. - * cb : Callback function, which is called for each node during - * iteration. Under normal circumstances the callback function - * should return NULL, which causes iteration to continue. If a - * callback function returns non-NULL, iteration is immediately - * terminated and the non-NULL return value is returned by the - * iterator. This is useful for re-starting iteration after - * modifying extree. - * arg : Opaque pointer passed to cb(). + * tree : Pointer to an initialized red-black tree object. + * start: Node at which to start iteration, or NULL to start at + * first/last node. + * cb : Callback function, which is called for each node during + * iteration. Under normal circumstances the callback function + * should return NULL, which causes iteration to continue. If a + * callback function returns non-NULL, iteration is immediately + * terminated and the non-NULL return value is returned by the + * iterator. This is useful for re-starting iteration after + * modifying tree. + * arg : Opaque pointer passed to cb(). * Ret: NULL if iteration completed, or the non-NULL callback return value * that caused termination of the iteration. */ diff --git a/include/jemalloc/internal/size_classes.sh b/include/jemalloc/internal/size_classes.sh new file mode 100755 index 00000000..29c80c1f --- /dev/null +++ b/include/jemalloc/internal/size_classes.sh @@ -0,0 +1,122 @@ +#!/bin/sh + +# The following limits are chosen such that they cover all supported platforms. + +# Range of quanta. +lg_qmin=3 +lg_qmax=4 + +# The range of tiny size classes is [2^lg_tmin..2^(lg_q-1)]. +lg_tmin=3 + +# Range of page sizes. +lg_pmin=12 +lg_pmax=16 + +pow2() { + e=$1 + pow2_result=1 + while [ ${e} -gt 0 ] ; do + pow2_result=$((${pow2_result} + ${pow2_result})) + e=$((${e} - 1)) + done +} + +cat < 255) +# error "Too many small size classes" +#endif + +#endif /* JEMALLOC_H_TYPES */ +/******************************************************************************/ +#ifdef JEMALLOC_H_STRUCTS + + +#endif /* JEMALLOC_H_STRUCTS */ +/******************************************************************************/ +#ifdef JEMALLOC_H_EXTERNS + + +#endif /* JEMALLOC_H_EXTERNS */ +/******************************************************************************/ +#ifdef JEMALLOC_H_INLINES + + +#endif /* JEMALLOC_H_INLINES */ +/******************************************************************************/ +EOF diff --git a/include/jemalloc/internal/stats.h b/include/jemalloc/internal/stats.h index 2a9b31d9..27f68e36 100644 --- a/include/jemalloc/internal/stats.h +++ b/include/jemalloc/internal/stats.h @@ -1,25 +1,16 @@ /******************************************************************************/ #ifdef JEMALLOC_H_TYPES -#define UMAX2S_BUFSIZE 65 - -#ifdef JEMALLOC_STATS typedef struct tcache_bin_stats_s tcache_bin_stats_t; typedef struct malloc_bin_stats_s malloc_bin_stats_t; typedef struct malloc_large_stats_s malloc_large_stats_t; typedef struct arena_stats_s arena_stats_t; -#endif -#if (defined(JEMALLOC_STATS) || defined(JEMALLOC_PROF)) typedef struct chunk_stats_s chunk_stats_t; -#endif #endif /* JEMALLOC_H_TYPES */ /******************************************************************************/ #ifdef JEMALLOC_H_STRUCTS -#ifdef JEMALLOC_STATS - -#ifdef JEMALLOC_TCACHE struct tcache_bin_stats_s { /* * Number of allocation requests that corresponded to the size of this @@ -27,7 +18,6 @@ struct tcache_bin_stats_s { */ uint64_t nrequests; }; -#endif struct malloc_bin_stats_s { /* @@ -52,13 +42,11 @@ struct malloc_bin_stats_s { */ uint64_t nrequests; -#ifdef JEMALLOC_TCACHE /* Number of tcache fills from this bin. */ uint64_t nfills; /* Number of tcache flushes to this bin. */ uint64_t nflushes; -#endif /* Total number of runs created for this bin's size class. */ uint64_t nruns; @@ -69,9 +57,6 @@ struct malloc_bin_stats_s { */ uint64_t reruns; - /* High-water mark for this bin. */ - size_t highruns; - /* Current number of runs in this bin. */ size_t curruns; }; @@ -93,9 +78,6 @@ struct malloc_large_stats_s { */ uint64_t nrequests; - /* High-water mark for this size class. */ - size_t highruns; - /* Current number of runs of this size class. */ size_t curruns; }; @@ -127,14 +109,10 @@ struct arena_stats_s { */ malloc_large_stats_t *lstats; }; -#endif /* JEMALLOC_STATS */ -#if (defined(JEMALLOC_STATS) || defined(JEMALLOC_PROF)) struct chunk_stats_s { -# ifdef JEMALLOC_STATS /* Number of chunks that were allocated. */ uint64_t nchunks; -# endif /* High-water mark for number of chunks allocated. */ size_t highchunks; @@ -146,7 +124,6 @@ struct chunk_stats_s { */ size_t curchunks; }; -#endif /* JEMALLOC_STATS */ #endif /* JEMALLOC_H_STRUCTS */ /******************************************************************************/ @@ -154,24 +131,14 @@ struct chunk_stats_s { extern bool opt_stats_print; -#ifdef JEMALLOC_STATS extern size_t stats_cactive; -#endif -char *u2s(uint64_t x, unsigned base, char *s); -#ifdef JEMALLOC_STATS -void malloc_cprintf(void (*write)(void *, const char *), void *cbopaque, - const char *format, ...) JEMALLOC_ATTR(format(printf, 3, 4)); -void malloc_printf(const char *format, ...) - JEMALLOC_ATTR(format(printf, 1, 2)); -#endif void stats_print(void (*write)(void *, const char *), void *cbopaque, const char *opts); #endif /* JEMALLOC_H_EXTERNS */ /******************************************************************************/ #ifdef JEMALLOC_H_INLINES -#ifdef JEMALLOC_STATS #ifndef JEMALLOC_ENABLE_INLINE size_t stats_cactive_get(void); @@ -202,6 +169,5 @@ stats_cactive_sub(size_t size) } #endif -#endif /* JEMALLOC_STATS */ #endif /* JEMALLOC_H_INLINES */ /******************************************************************************/ diff --git a/include/jemalloc/internal/tcache.h b/include/jemalloc/internal/tcache.h index da3c68c5..38d735c8 100644 --- a/include/jemalloc/internal/tcache.h +++ b/include/jemalloc/internal/tcache.h @@ -1,4 +1,3 @@ -#ifdef JEMALLOC_TCACHE /******************************************************************************/ #ifdef JEMALLOC_H_TYPES @@ -6,6 +5,16 @@ typedef struct tcache_bin_info_s tcache_bin_info_t; typedef struct tcache_bin_s tcache_bin_t; typedef struct tcache_s tcache_t; +/* + * tcache pointers close to NULL are used to encode state information that is + * used for two purposes: preventing thread caching on a per thread basis and + * cleaning up during thread shutdown. + */ +#define TCACHE_STATE_DISABLED ((tcache_t *)(uintptr_t)1) +#define TCACHE_STATE_REINCARNATED ((tcache_t *)(uintptr_t)2) +#define TCACHE_STATE_PURGATORY ((tcache_t *)(uintptr_t)3) +#define TCACHE_STATE_MAX TCACHE_STATE_PURGATORY + /* * Absolute maximum number of cache slots for each small bin in the thread * cache. This is an additional constraint beyond that imposed as: twice the @@ -22,17 +31,26 @@ typedef struct tcache_s tcache_t; #define LG_TCACHE_MAXCLASS_DEFAULT 15 /* - * (1U << opt_lg_tcache_gc_sweep) is the approximate number of allocation - * events between full GC sweeps (-1: disabled). Integer rounding may cause - * the actual number to be slightly higher, since GC is performed - * incrementally. + * TCACHE_GC_SWEEP is the approximate number of allocation events between + * full GC sweeps. Integer rounding may cause the actual number to be + * slightly higher, since GC is performed incrementally. */ -#define LG_TCACHE_GC_SWEEP_DEFAULT 13 +#define TCACHE_GC_SWEEP 8192 + +/* Number of tcache allocation/deallocation events between incremental GCs. */ +#define TCACHE_GC_INCR \ + ((TCACHE_GC_SWEEP / NBINS) + ((TCACHE_GC_SWEEP / NBINS == 0) ? 0 : 1)) #endif /* JEMALLOC_H_TYPES */ /******************************************************************************/ #ifdef JEMALLOC_H_STRUCTS +typedef enum { + tcache_enabled_false = 0, /* Enable cast to/from bool. */ + tcache_enabled_true = 1, + tcache_enabled_default = 2 +} tcache_enabled_t; + /* * Read-only information associated with each element of tcache_t's tbins array * is stored separately, mainly to reduce memory usage. @@ -42,9 +60,7 @@ struct tcache_bin_info_s { }; struct tcache_bin_s { -# ifdef JEMALLOC_STATS tcache_bin_stats_t tstats; -# endif int low_water; /* Min # cached since last GC. */ unsigned lg_fill_div; /* Fill (ncached_max >> lg_fill_div). */ unsigned ncached; /* # of cached objects. */ @@ -52,12 +68,8 @@ struct tcache_bin_s { }; struct tcache_s { -# ifdef JEMALLOC_STATS ql_elm(tcache_t) link; /* Used for aggregating stats. */ -# endif -# ifdef JEMALLOC_PROF uint64_t prof_accumbytes;/* Cleared after arena_prof_accum() */ -# endif arena_t *arena; /* This thread's arena. */ unsigned ev_cnt; /* Event count since incremental GC. */ unsigned next_gc_bin; /* Next bin to GC. */ @@ -76,29 +88,11 @@ struct tcache_s { extern bool opt_tcache; extern ssize_t opt_lg_tcache_max; -extern ssize_t opt_lg_tcache_gc_sweep; extern tcache_bin_info_t *tcache_bin_info; -/* Map of thread-specific caches. */ -#ifndef NO_TLS -extern __thread tcache_t *tcache_tls - JEMALLOC_ATTR(tls_model("initial-exec")); -# define TCACHE_GET() tcache_tls -# define TCACHE_SET(v) do { \ - tcache_tls = (tcache_t *)(v); \ - pthread_setspecific(tcache_tsd, (void *)(v)); \ -} while (0) -#else -# define TCACHE_GET() ((tcache_t *)pthread_getspecific(tcache_tsd)) -# define TCACHE_SET(v) do { \ - pthread_setspecific(tcache_tsd, (void *)(v)); \ -} while (0) -#endif -extern pthread_key_t tcache_tsd; - /* - * Number of tcache bins. There are nbins small-object bins, plus 0 or more + * Number of tcache bins. There are NBINS small-object bins, plus 0 or more * large-object bins. */ extern size_t nhbins; @@ -106,68 +100,159 @@ extern size_t nhbins; /* Maximum cached size class. */ extern size_t tcache_maxclass; -/* Number of tcache allocation/deallocation events between incremental GCs. */ -extern unsigned tcache_gc_incr; - -void tcache_bin_flush_small(tcache_bin_t *tbin, size_t binind, unsigned rem -#if (defined(JEMALLOC_STATS) || defined(JEMALLOC_PROF)) - , tcache_t *tcache -#endif - ); -void tcache_bin_flush_large(tcache_bin_t *tbin, size_t binind, unsigned rem -#if (defined(JEMALLOC_STATS) || defined(JEMALLOC_PROF)) - , tcache_t *tcache -#endif - ); -tcache_t *tcache_create(arena_t *arena); +size_t tcache_salloc(const void *ptr); +void tcache_event_hard(tcache_t *tcache); void *tcache_alloc_small_hard(tcache_t *tcache, tcache_bin_t *tbin, size_t binind); +void tcache_bin_flush_small(tcache_bin_t *tbin, size_t binind, unsigned rem, + tcache_t *tcache); +void tcache_bin_flush_large(tcache_bin_t *tbin, size_t binind, unsigned rem, + tcache_t *tcache); +void tcache_arena_associate(tcache_t *tcache, arena_t *arena); +void tcache_arena_dissociate(tcache_t *tcache); +tcache_t *tcache_create(arena_t *arena); void tcache_destroy(tcache_t *tcache); -#ifdef JEMALLOC_STATS +void tcache_thread_cleanup(void *arg); void tcache_stats_merge(tcache_t *tcache, arena_t *arena); -#endif -bool tcache_boot(void); +bool tcache_boot0(void); +bool tcache_boot1(void); #endif /* JEMALLOC_H_EXTERNS */ /******************************************************************************/ #ifdef JEMALLOC_H_INLINES #ifndef JEMALLOC_ENABLE_INLINE +malloc_tsd_protos(JEMALLOC_ATTR(unused), tcache, tcache_t *) +malloc_tsd_protos(JEMALLOC_ATTR(unused), tcache_enabled, tcache_enabled_t) + void tcache_event(tcache_t *tcache); -tcache_t *tcache_get(void); +void tcache_flush(void); +bool tcache_enabled_get(void); +tcache_t *tcache_get(bool create); +void tcache_enabled_set(bool enabled); void *tcache_alloc_easy(tcache_bin_t *tbin); void *tcache_alloc_small(tcache_t *tcache, size_t size, bool zero); void *tcache_alloc_large(tcache_t *tcache, size_t size, bool zero); -void tcache_dalloc_small(tcache_t *tcache, void *ptr); +void tcache_dalloc_small(tcache_t *tcache, void *ptr, size_t binind); void tcache_dalloc_large(tcache_t *tcache, void *ptr, size_t size); #endif #if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_TCACHE_C_)) -JEMALLOC_INLINE tcache_t * -tcache_get(void) +/* Map of thread-specific caches. */ +malloc_tsd_externs(tcache, tcache_t *) +malloc_tsd_funcs(JEMALLOC_INLINE, tcache, tcache_t *, NULL, + tcache_thread_cleanup) +/* Per thread flag that allows thread caches to be disabled. */ +malloc_tsd_externs(tcache_enabled, tcache_enabled_t) +malloc_tsd_funcs(JEMALLOC_INLINE, tcache_enabled, tcache_enabled_t, + tcache_enabled_default, malloc_tsd_no_cleanup) + +JEMALLOC_INLINE void +tcache_flush(void) { tcache_t *tcache; - if ((isthreaded & opt_tcache) == false) + cassert(config_tcache); + + tcache = *tcache_tsd_get(); + if ((uintptr_t)tcache <= (uintptr_t)TCACHE_STATE_MAX) + return; + tcache_destroy(tcache); + tcache = NULL; + tcache_tsd_set(&tcache); +} + +JEMALLOC_INLINE bool +tcache_enabled_get(void) +{ + tcache_enabled_t tcache_enabled; + + cassert(config_tcache); + + tcache_enabled = *tcache_enabled_tsd_get(); + if (tcache_enabled == tcache_enabled_default) { + tcache_enabled = (tcache_enabled_t)opt_tcache; + tcache_enabled_tsd_set(&tcache_enabled); + } + + return ((bool)tcache_enabled); +} + +JEMALLOC_INLINE void +tcache_enabled_set(bool enabled) +{ + tcache_enabled_t tcache_enabled; + tcache_t *tcache; + + cassert(config_tcache); + + tcache_enabled = (tcache_enabled_t)enabled; + tcache_enabled_tsd_set(&tcache_enabled); + tcache = *tcache_tsd_get(); + if (enabled) { + if (tcache == TCACHE_STATE_DISABLED) { + tcache = NULL; + tcache_tsd_set(&tcache); + } + } else /* disabled */ { + if (tcache > TCACHE_STATE_MAX) { + tcache_destroy(tcache); + tcache = NULL; + } + if (tcache == NULL) { + tcache = TCACHE_STATE_DISABLED; + tcache_tsd_set(&tcache); + } + } +} + +JEMALLOC_INLINE tcache_t * +tcache_get(bool create) +{ + tcache_t *tcache; + + if (config_tcache == false) + return (NULL); + if (config_lazy_lock && isthreaded == false) return (NULL); - tcache = TCACHE_GET(); - if ((uintptr_t)tcache <= (uintptr_t)2) { + tcache = *tcache_tsd_get(); + if ((uintptr_t)tcache <= (uintptr_t)TCACHE_STATE_MAX) { + if (tcache == TCACHE_STATE_DISABLED) + return (NULL); if (tcache == NULL) { - tcache = tcache_create(choose_arena()); - if (tcache == NULL) - return (NULL); - } else { - if (tcache == (void *)(uintptr_t)1) { + if (create == false) { /* - * Make a note that an allocator function was - * called after the tcache_thread_cleanup() was - * called. + * Creating a tcache here would cause + * allocation as a side effect of free(). + * Ordinarily that would be okay since + * tcache_create() failure is a soft failure + * that doesn't propagate. However, if TLS + * data are freed via free() as in glibc, + * subtle corruption could result from setting + * a TLS variable after its backing memory is + * freed. */ - TCACHE_SET((uintptr_t)2); + return (NULL); } + if (tcache_enabled_get() == false) { + tcache_enabled_set(false); /* Memoize. */ + return (NULL); + } + return (tcache_create(choose_arena(NULL))); + } + if (tcache == TCACHE_STATE_PURGATORY) { + /* + * Make a note that an allocator function was called + * after tcache_thread_cleanup() was called. + */ + tcache = TCACHE_STATE_REINCARNATED; + tcache_tsd_set(&tcache); return (NULL); } + if (tcache == TCACHE_STATE_REINCARNATED) + return (NULL); + not_reached(); } return (tcache); @@ -177,60 +262,13 @@ JEMALLOC_INLINE void tcache_event(tcache_t *tcache) { - if (tcache_gc_incr == 0) + if (TCACHE_GC_INCR == 0) return; tcache->ev_cnt++; - assert(tcache->ev_cnt <= tcache_gc_incr); - if (tcache->ev_cnt == tcache_gc_incr) { - size_t binind = tcache->next_gc_bin; - tcache_bin_t *tbin = &tcache->tbins[binind]; - tcache_bin_info_t *tbin_info = &tcache_bin_info[binind]; - - if (tbin->low_water > 0) { - /* - * Flush (ceiling) 3/4 of the objects below the low - * water mark. - */ - if (binind < nbins) { - tcache_bin_flush_small(tbin, binind, - tbin->ncached - tbin->low_water + - (tbin->low_water >> 2) -#if (defined(JEMALLOC_STATS) || defined(JEMALLOC_PROF)) - , tcache -#endif - ); - } else { - tcache_bin_flush_large(tbin, binind, - tbin->ncached - tbin->low_water + - (tbin->low_water >> 2) -#if (defined(JEMALLOC_STATS) || defined(JEMALLOC_PROF)) - , tcache -#endif - ); - } - /* - * Reduce fill count by 2X. Limit lg_fill_div such that - * the fill count is always at least 1. - */ - if ((tbin_info->ncached_max >> (tbin->lg_fill_div+1)) - >= 1) - tbin->lg_fill_div++; - } else if (tbin->low_water < 0) { - /* - * Increase fill count by 2X. Make sure lg_fill_div - * stays greater than 0. - */ - if (tbin->lg_fill_div > 1) - tbin->lg_fill_div--; - } - tbin->low_water = tbin->ncached; - - tcache->next_gc_bin++; - if (tcache->next_gc_bin == nhbins) - tcache->next_gc_bin = 0; - tcache->ev_cnt = 0; - } + assert(tcache->ev_cnt <= TCACHE_GC_INCR); + if (tcache->ev_cnt == TCACHE_GC_INCR) + tcache_event_hard(tcache); } JEMALLOC_INLINE void * @@ -257,7 +295,7 @@ tcache_alloc_small(tcache_t *tcache, size_t size, bool zero) tcache_bin_t *tbin; binind = SMALL_SIZE2BIN(size); - assert(binind < nbins); + assert(binind < NBINS); tbin = &tcache->tbins[binind]; ret = tcache_alloc_easy(tbin); if (ret == NULL) { @@ -265,24 +303,29 @@ tcache_alloc_small(tcache_t *tcache, size_t size, bool zero) if (ret == NULL) return (NULL); } - assert(arena_salloc(ret) == arena_bin_info[binind].reg_size); + assert(tcache_salloc(ret) == arena_bin_info[binind].reg_size); if (zero == false) { -#ifdef JEMALLOC_FILL - if (opt_junk) - memset(ret, 0xa5, size); - else if (opt_zero) - memset(ret, 0, size); -#endif - } else + if (config_fill) { + if (opt_junk) { + arena_alloc_junk_small(ret, + &arena_bin_info[binind], false); + } else if (opt_zero) + memset(ret, 0, size); + } + } else { + if (config_fill && opt_junk) { + arena_alloc_junk_small(ret, &arena_bin_info[binind], + true); + } + VALGRIND_MAKE_MEM_UNDEFINED(ret, size); memset(ret, 0, size); + } -#ifdef JEMALLOC_STATS - tbin->tstats.nrequests++; -#endif -#ifdef JEMALLOC_PROF - tcache->prof_accumbytes += arena_bin_info[binind].reg_size; -#endif + if (config_stats) + tbin->tstats.nrequests++; + if (config_prof) + tcache->prof_accumbytes += arena_bin_info[binind].reg_size; tcache_event(tcache); return (ret); } @@ -296,7 +339,7 @@ tcache_alloc_large(tcache_t *tcache, size_t size, bool zero) size = PAGE_CEILING(size); assert(size <= tcache_maxclass); - binind = nbins + (size >> PAGE_SHIFT) - 1; + binind = NBINS + (size >> LG_PAGE) - 1; assert(binind < nhbins); tbin = &tcache->tbins[binind]; ret = tcache_alloc_easy(tbin); @@ -309,28 +352,30 @@ tcache_alloc_large(tcache_t *tcache, size_t size, bool zero) if (ret == NULL) return (NULL); } else { -#ifdef JEMALLOC_PROF - arena_chunk_t *chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ret); - size_t pageind = (((uintptr_t)ret - (uintptr_t)chunk) >> - PAGE_SHIFT); - chunk->map[pageind-map_bias].bits &= ~CHUNK_MAP_CLASS_MASK; -#endif + if (config_prof && prof_promote && size == PAGE) { + arena_chunk_t *chunk = + (arena_chunk_t *)CHUNK_ADDR2BASE(ret); + size_t pageind = (((uintptr_t)ret - (uintptr_t)chunk) >> + LG_PAGE); + arena_mapbits_large_binind_set(chunk, pageind, + BININD_INVALID); + } if (zero == false) { -#ifdef JEMALLOC_FILL - if (opt_junk) - memset(ret, 0xa5, size); - else if (opt_zero) - memset(ret, 0, size); -#endif - } else + if (config_fill) { + if (opt_junk) + memset(ret, 0xa5, size); + else if (opt_zero) + memset(ret, 0, size); + } + } else { + VALGRIND_MAKE_MEM_UNDEFINED(ret, size); memset(ret, 0, size); + } -#ifdef JEMALLOC_STATS - tbin->tstats.nrequests++; -#endif -#ifdef JEMALLOC_PROF - tcache->prof_accumbytes += size; -#endif + if (config_stats) + tbin->tstats.nrequests++; + if (config_prof) + tcache->prof_accumbytes += size; } tcache_event(tcache); @@ -338,45 +383,21 @@ tcache_alloc_large(tcache_t *tcache, size_t size, bool zero) } JEMALLOC_INLINE void -tcache_dalloc_small(tcache_t *tcache, void *ptr) +tcache_dalloc_small(tcache_t *tcache, void *ptr, size_t binind) { - arena_t *arena; - arena_chunk_t *chunk; - arena_run_t *run; - arena_bin_t *bin; tcache_bin_t *tbin; tcache_bin_info_t *tbin_info; - size_t pageind, binind; - arena_chunk_map_t *mapelm; - assert(arena_salloc(ptr) <= small_maxclass); + assert(tcache_salloc(ptr) <= SMALL_MAXCLASS); - chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr); - arena = chunk->arena; - pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> PAGE_SHIFT; - mapelm = &chunk->map[pageind-map_bias]; - run = (arena_run_t *)((uintptr_t)chunk + (uintptr_t)((pageind - - (mapelm->bits >> PAGE_SHIFT)) << PAGE_SHIFT)); - dassert(run->magic == ARENA_RUN_MAGIC); - bin = run->bin; - binind = ((uintptr_t)bin - (uintptr_t)&arena->bins) / - sizeof(arena_bin_t); - assert(binind < nbins); - -#ifdef JEMALLOC_FILL - if (opt_junk) - memset(ptr, 0x5a, arena_bin_info[binind].reg_size); -#endif + if (config_fill && opt_junk) + arena_dalloc_junk_small(ptr, &arena_bin_info[binind]); tbin = &tcache->tbins[binind]; tbin_info = &tcache_bin_info[binind]; if (tbin->ncached == tbin_info->ncached_max) { tcache_bin_flush_small(tbin, binind, (tbin_info->ncached_max >> - 1) -#if (defined(JEMALLOC_STATS) || defined(JEMALLOC_PROF)) - , tcache -#endif - ); + 1), tcache); } assert(tbin->ncached < tbin_info->ncached_max); tbin->avail[tbin->ncached] = ptr; @@ -388,35 +409,24 @@ tcache_dalloc_small(tcache_t *tcache, void *ptr) JEMALLOC_INLINE void tcache_dalloc_large(tcache_t *tcache, void *ptr, size_t size) { - arena_t *arena; - arena_chunk_t *chunk; - size_t pageind, binind; + size_t binind; tcache_bin_t *tbin; tcache_bin_info_t *tbin_info; assert((size & PAGE_MASK) == 0); - assert(arena_salloc(ptr) > small_maxclass); - assert(arena_salloc(ptr) <= tcache_maxclass); + assert(tcache_salloc(ptr) > SMALL_MAXCLASS); + assert(tcache_salloc(ptr) <= tcache_maxclass); - chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr); - arena = chunk->arena; - pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> PAGE_SHIFT; - binind = nbins + (size >> PAGE_SHIFT) - 1; + binind = NBINS + (size >> LG_PAGE) - 1; -#ifdef JEMALLOC_FILL - if (opt_junk) + if (config_fill && opt_junk) memset(ptr, 0x5a, size); -#endif tbin = &tcache->tbins[binind]; tbin_info = &tcache_bin_info[binind]; if (tbin->ncached == tbin_info->ncached_max) { tcache_bin_flush_large(tbin, binind, (tbin_info->ncached_max >> - 1) -#if (defined(JEMALLOC_STATS) || defined(JEMALLOC_PROF)) - , tcache -#endif - ); + 1), tcache); } assert(tbin->ncached < tbin_info->ncached_max); tbin->avail[tbin->ncached] = ptr; @@ -428,4 +438,3 @@ tcache_dalloc_large(tcache_t *tcache, void *ptr, size_t size) #endif /* JEMALLOC_H_INLINES */ /******************************************************************************/ -#endif /* JEMALLOC_TCACHE */ diff --git a/include/jemalloc/internal/tsd.h b/include/jemalloc/internal/tsd.h new file mode 100644 index 00000000..0037cf35 --- /dev/null +++ b/include/jemalloc/internal/tsd.h @@ -0,0 +1,397 @@ +/******************************************************************************/ +#ifdef JEMALLOC_H_TYPES + +/* Maximum number of malloc_tsd users with cleanup functions. */ +#define MALLOC_TSD_CLEANUPS_MAX 8 + +typedef bool (*malloc_tsd_cleanup_t)(void); + +/* + * TLS/TSD-agnostic macro-based implementation of thread-specific data. There + * are four macros that support (at least) three use cases: file-private, + * library-private, and library-private inlined. Following is an example + * library-private tsd variable: + * + * In example.h: + * typedef struct { + * int x; + * int y; + * } example_t; + * #define EX_INITIALIZER JEMALLOC_CONCAT({0, 0}) + * malloc_tsd_protos(, example, example_t *) + * malloc_tsd_externs(example, example_t *) + * In example.c: + * malloc_tsd_data(, example, example_t *, EX_INITIALIZER) + * malloc_tsd_funcs(, example, example_t *, EX_INITIALIZER, + * example_tsd_cleanup) + * + * The result is a set of generated functions, e.g.: + * + * bool example_tsd_boot(void) {...} + * example_t **example_tsd_get() {...} + * void example_tsd_set(example_t **val) {...} + * + * Note that all of the functions deal in terms of (a_type *) rather than + * (a_type) so that it is possible to support non-pointer types (unlike + * pthreads TSD). example_tsd_cleanup() is passed an (a_type *) pointer that is + * cast to (void *). This means that the cleanup function needs to cast *and* + * dereference the function argument, e.g.: + * + * void + * example_tsd_cleanup(void *arg) + * { + * example_t *example = *(example_t **)arg; + * + * [...] + * if ([want the cleanup function to be called again]) { + * example_tsd_set(&example); + * } + * } + * + * If example_tsd_set() is called within example_tsd_cleanup(), it will be + * called again. This is similar to how pthreads TSD destruction works, except + * that pthreads only calls the cleanup function again if the value was set to + * non-NULL. + */ + +/* malloc_tsd_protos(). */ +#define malloc_tsd_protos(a_attr, a_name, a_type) \ +a_attr bool \ +a_name##_tsd_boot(void); \ +a_attr a_type * \ +a_name##_tsd_get(void); \ +a_attr void \ +a_name##_tsd_set(a_type *val); + +/* malloc_tsd_externs(). */ +#ifdef JEMALLOC_MALLOC_THREAD_CLEANUP +#define malloc_tsd_externs(a_name, a_type) \ +extern __thread a_type a_name##_tls; \ +extern __thread bool a_name##_initialized; \ +extern bool a_name##_booted; +#elif (defined(JEMALLOC_TLS)) +#define malloc_tsd_externs(a_name, a_type) \ +extern __thread a_type a_name##_tls; \ +extern pthread_key_t a_name##_tsd; \ +extern bool a_name##_booted; +#elif (defined(_WIN32)) +#define malloc_tsd_externs(a_name, a_type) \ +extern DWORD a_name##_tsd; \ +extern bool a_name##_booted; +#else +#define malloc_tsd_externs(a_name, a_type) \ +extern pthread_key_t a_name##_tsd; \ +extern bool a_name##_booted; +#endif + +/* malloc_tsd_data(). */ +#ifdef JEMALLOC_MALLOC_THREAD_CLEANUP +#define malloc_tsd_data(a_attr, a_name, a_type, a_initializer) \ +a_attr __thread a_type JEMALLOC_TLS_MODEL \ + a_name##_tls = a_initializer; \ +a_attr __thread bool JEMALLOC_TLS_MODEL \ + a_name##_initialized = false; \ +a_attr bool a_name##_booted = false; +#elif (defined(JEMALLOC_TLS)) +#define malloc_tsd_data(a_attr, a_name, a_type, a_initializer) \ +a_attr __thread a_type JEMALLOC_TLS_MODEL \ + a_name##_tls = a_initializer; \ +a_attr pthread_key_t a_name##_tsd; \ +a_attr bool a_name##_booted = false; +#elif (defined(_WIN32)) +#define malloc_tsd_data(a_attr, a_name, a_type, a_initializer) \ +a_attr DWORD a_name##_tsd; \ +a_attr bool a_name##_booted = false; +#else +#define malloc_tsd_data(a_attr, a_name, a_type, a_initializer) \ +a_attr pthread_key_t a_name##_tsd; \ +a_attr bool a_name##_booted = false; +#endif + +/* malloc_tsd_funcs(). */ +#ifdef JEMALLOC_MALLOC_THREAD_CLEANUP +#define malloc_tsd_funcs(a_attr, a_name, a_type, a_initializer, \ + a_cleanup) \ +/* Initialization/cleanup. */ \ +a_attr bool \ +a_name##_tsd_cleanup_wrapper(void) \ +{ \ + \ + if (a_name##_initialized) { \ + a_name##_initialized = false; \ + a_cleanup(&a_name##_tls); \ + } \ + return (a_name##_initialized); \ +} \ +a_attr bool \ +a_name##_tsd_boot(void) \ +{ \ + \ + if (a_cleanup != malloc_tsd_no_cleanup) { \ + malloc_tsd_cleanup_register( \ + &a_name##_tsd_cleanup_wrapper); \ + } \ + a_name##_booted = true; \ + return (false); \ +} \ +/* Get/set. */ \ +a_attr a_type * \ +a_name##_tsd_get(void) \ +{ \ + \ + assert(a_name##_booted); \ + return (&a_name##_tls); \ +} \ +a_attr void \ +a_name##_tsd_set(a_type *val) \ +{ \ + \ + assert(a_name##_booted); \ + a_name##_tls = (*val); \ + if (a_cleanup != malloc_tsd_no_cleanup) \ + a_name##_initialized = true; \ +} +#elif (defined(JEMALLOC_TLS)) +#define malloc_tsd_funcs(a_attr, a_name, a_type, a_initializer, \ + a_cleanup) \ +/* Initialization/cleanup. */ \ +a_attr bool \ +a_name##_tsd_boot(void) \ +{ \ + \ + if (a_cleanup != malloc_tsd_no_cleanup) { \ + if (pthread_key_create(&a_name##_tsd, a_cleanup) != 0) \ + return (true); \ + } \ + a_name##_booted = true; \ + return (false); \ +} \ +/* Get/set. */ \ +a_attr a_type * \ +a_name##_tsd_get(void) \ +{ \ + \ + assert(a_name##_booted); \ + return (&a_name##_tls); \ +} \ +a_attr void \ +a_name##_tsd_set(a_type *val) \ +{ \ + \ + assert(a_name##_booted); \ + a_name##_tls = (*val); \ + if (a_cleanup != malloc_tsd_no_cleanup) { \ + if (pthread_setspecific(a_name##_tsd, \ + (void *)(&a_name##_tls))) { \ + malloc_write(": Error" \ + " setting TSD for "#a_name"\n"); \ + if (opt_abort) \ + abort(); \ + } \ + } \ +} +#elif (defined(_WIN32)) +#define malloc_tsd_funcs(a_attr, a_name, a_type, a_initializer, \ + a_cleanup) \ +/* Data structure. */ \ +typedef struct { \ + bool initialized; \ + a_type val; \ +} a_name##_tsd_wrapper_t; \ +/* Initialization/cleanup. */ \ +a_attr bool \ +a_name##_tsd_cleanup_wrapper(void) \ +{ \ + a_name##_tsd_wrapper_t *wrapper; \ + \ + wrapper = (a_name##_tsd_wrapper_t *) TlsGetValue(a_name##_tsd); \ + if (wrapper == NULL) \ + return (false); \ + if (a_cleanup != malloc_tsd_no_cleanup && \ + wrapper->initialized) { \ + a_type val = wrapper->val; \ + a_type tsd_static_data = a_initializer; \ + wrapper->initialized = false; \ + wrapper->val = tsd_static_data; \ + a_cleanup(&val); \ + if (wrapper->initialized) { \ + /* Trigger another cleanup round. */ \ + return (true); \ + } \ + } \ + malloc_tsd_dalloc(wrapper); \ + return (false); \ +} \ +a_attr bool \ +a_name##_tsd_boot(void) \ +{ \ + \ + a_name##_tsd = TlsAlloc(); \ + if (a_name##_tsd == TLS_OUT_OF_INDEXES) \ + return (true); \ + if (a_cleanup != malloc_tsd_no_cleanup) { \ + malloc_tsd_cleanup_register( \ + &a_name##_tsd_cleanup_wrapper); \ + } \ + a_name##_booted = true; \ + return (false); \ +} \ +/* Get/set. */ \ +a_attr a_name##_tsd_wrapper_t * \ +a_name##_tsd_get_wrapper(void) \ +{ \ + a_name##_tsd_wrapper_t *wrapper = (a_name##_tsd_wrapper_t *) \ + TlsGetValue(a_name##_tsd); \ + \ + if (wrapper == NULL) { \ + wrapper = (a_name##_tsd_wrapper_t *) \ + malloc_tsd_malloc(sizeof(a_name##_tsd_wrapper_t)); \ + if (wrapper == NULL) { \ + malloc_write(": Error allocating" \ + " TSD for "#a_name"\n"); \ + abort(); \ + } else { \ + static a_type tsd_static_data = a_initializer; \ + wrapper->initialized = false; \ + wrapper->val = tsd_static_data; \ + } \ + if (!TlsSetValue(a_name##_tsd, (void *)wrapper)) { \ + malloc_write(": Error setting" \ + " TSD for "#a_name"\n"); \ + abort(); \ + } \ + } \ + return (wrapper); \ +} \ +a_attr a_type * \ +a_name##_tsd_get(void) \ +{ \ + a_name##_tsd_wrapper_t *wrapper; \ + \ + assert(a_name##_booted); \ + wrapper = a_name##_tsd_get_wrapper(); \ + return (&wrapper->val); \ +} \ +a_attr void \ +a_name##_tsd_set(a_type *val) \ +{ \ + a_name##_tsd_wrapper_t *wrapper; \ + \ + assert(a_name##_booted); \ + wrapper = a_name##_tsd_get_wrapper(); \ + wrapper->val = *(val); \ + if (a_cleanup != malloc_tsd_no_cleanup) \ + wrapper->initialized = true; \ +} +#else +#define malloc_tsd_funcs(a_attr, a_name, a_type, a_initializer, \ + a_cleanup) \ +/* Data structure. */ \ +typedef struct { \ + bool initialized; \ + a_type val; \ +} a_name##_tsd_wrapper_t; \ +/* Initialization/cleanup. */ \ +a_attr void \ +a_name##_tsd_cleanup_wrapper(void *arg) \ +{ \ + a_name##_tsd_wrapper_t *wrapper = (a_name##_tsd_wrapper_t *)arg;\ + \ + if (a_cleanup != malloc_tsd_no_cleanup && \ + wrapper->initialized) { \ + wrapper->initialized = false; \ + a_cleanup(&wrapper->val); \ + if (wrapper->initialized) { \ + /* Trigger another cleanup round. */ \ + if (pthread_setspecific(a_name##_tsd, \ + (void *)wrapper)) { \ + malloc_write(": Error" \ + " setting TSD for "#a_name"\n"); \ + if (opt_abort) \ + abort(); \ + } \ + return; \ + } \ + } \ + malloc_tsd_dalloc(wrapper); \ +} \ +a_attr bool \ +a_name##_tsd_boot(void) \ +{ \ + \ + if (pthread_key_create(&a_name##_tsd, \ + a_name##_tsd_cleanup_wrapper) != 0) \ + return (true); \ + a_name##_booted = true; \ + return (false); \ +} \ +/* Get/set. */ \ +a_attr a_name##_tsd_wrapper_t * \ +a_name##_tsd_get_wrapper(void) \ +{ \ + a_name##_tsd_wrapper_t *wrapper = (a_name##_tsd_wrapper_t *) \ + pthread_getspecific(a_name##_tsd); \ + \ + if (wrapper == NULL) { \ + wrapper = (a_name##_tsd_wrapper_t *) \ + malloc_tsd_malloc(sizeof(a_name##_tsd_wrapper_t)); \ + if (wrapper == NULL) { \ + malloc_write(": Error allocating" \ + " TSD for "#a_name"\n"); \ + abort(); \ + } else { \ + static a_type tsd_static_data = a_initializer; \ + wrapper->initialized = false; \ + wrapper->val = tsd_static_data; \ + } \ + if (pthread_setspecific(a_name##_tsd, \ + (void *)wrapper)) { \ + malloc_write(": Error setting" \ + " TSD for "#a_name"\n"); \ + abort(); \ + } \ + } \ + return (wrapper); \ +} \ +a_attr a_type * \ +a_name##_tsd_get(void) \ +{ \ + a_name##_tsd_wrapper_t *wrapper; \ + \ + assert(a_name##_booted); \ + wrapper = a_name##_tsd_get_wrapper(); \ + return (&wrapper->val); \ +} \ +a_attr void \ +a_name##_tsd_set(a_type *val) \ +{ \ + a_name##_tsd_wrapper_t *wrapper; \ + \ + assert(a_name##_booted); \ + wrapper = a_name##_tsd_get_wrapper(); \ + wrapper->val = *(val); \ + if (a_cleanup != malloc_tsd_no_cleanup) \ + wrapper->initialized = true; \ +} +#endif + +#endif /* JEMALLOC_H_TYPES */ +/******************************************************************************/ +#ifdef JEMALLOC_H_STRUCTS + +#endif /* JEMALLOC_H_STRUCTS */ +/******************************************************************************/ +#ifdef JEMALLOC_H_EXTERNS + +void *malloc_tsd_malloc(size_t size); +void malloc_tsd_dalloc(void *wrapper); +void malloc_tsd_no_cleanup(void *); +void malloc_tsd_cleanup_register(bool (*f)(void)); +void malloc_tsd_boot(void); + +#endif /* JEMALLOC_H_EXTERNS */ +/******************************************************************************/ +#ifdef JEMALLOC_H_INLINES + +#endif /* JEMALLOC_H_INLINES */ +/******************************************************************************/ diff --git a/include/jemalloc/internal/util.h b/include/jemalloc/internal/util.h new file mode 100644 index 00000000..84796936 --- /dev/null +++ b/include/jemalloc/internal/util.h @@ -0,0 +1,160 @@ +/******************************************************************************/ +#ifdef JEMALLOC_H_TYPES + +/* Size of stack-allocated buffer passed to buferror(). */ +#define BUFERROR_BUF 64 + +/* + * Size of stack-allocated buffer used by malloc_{,v,vc}printf(). This must be + * large enough for all possible uses within jemalloc. + */ +#define MALLOC_PRINTF_BUFSIZE 4096 + +/* + * Wrap a cpp argument that contains commas such that it isn't broken up into + * multiple arguments. + */ +#define JEMALLOC_CONCAT(...) __VA_ARGS__ + +/* + * Silence compiler warnings due to uninitialized values. This is used + * wherever the compiler fails to recognize that the variable is never used + * uninitialized. + */ +#ifdef JEMALLOC_CC_SILENCE +# define JEMALLOC_CC_SILENCE_INIT(v) = v +#else +# define JEMALLOC_CC_SILENCE_INIT(v) +#endif + +/* + * Define a custom assert() in order to reduce the chances of deadlock during + * assertion failure. + */ +#ifndef assert +#define assert(e) do { \ + if (config_debug && !(e)) { \ + malloc_printf( \ + ": %s:%d: Failed assertion: \"%s\"\n", \ + __FILE__, __LINE__, #e); \ + abort(); \ + } \ +} while (0) +#endif + +/* Use to assert a particular configuration, e.g., cassert(config_debug). */ +#define cassert(c) do { \ + if ((c) == false) \ + assert(false); \ +} while (0) + +#ifndef not_reached +#define not_reached() do { \ + if (config_debug) { \ + malloc_printf( \ + ": %s:%d: Unreachable code reached\n", \ + __FILE__, __LINE__); \ + abort(); \ + } \ +} while (0) +#endif + +#ifndef not_implemented +#define not_implemented() do { \ + if (config_debug) { \ + malloc_printf(": %s:%d: Not implemented\n", \ + __FILE__, __LINE__); \ + abort(); \ + } \ +} while (0) +#endif + +#define assert_not_implemented(e) do { \ + if (config_debug && !(e)) \ + not_implemented(); \ +} while (0) + +#endif /* JEMALLOC_H_TYPES */ +/******************************************************************************/ +#ifdef JEMALLOC_H_STRUCTS + +#endif /* JEMALLOC_H_STRUCTS */ +/******************************************************************************/ +#ifdef JEMALLOC_H_EXTERNS + +int buferror(char *buf, size_t buflen); +uintmax_t malloc_strtoumax(const char *nptr, char **endptr, int base); +void malloc_write(const char *s); + +/* + * malloc_vsnprintf() supports a subset of snprintf(3) that avoids floating + * point math. + */ +int malloc_vsnprintf(char *str, size_t size, const char *format, + va_list ap); +int malloc_snprintf(char *str, size_t size, const char *format, ...) + JEMALLOC_ATTR(format(printf, 3, 4)); +void malloc_vcprintf(void (*write_cb)(void *, const char *), void *cbopaque, + const char *format, va_list ap); +void malloc_cprintf(void (*write)(void *, const char *), void *cbopaque, + const char *format, ...) JEMALLOC_ATTR(format(printf, 3, 4)); +void malloc_printf(const char *format, ...) + JEMALLOC_ATTR(format(printf, 1, 2)); + +#endif /* JEMALLOC_H_EXTERNS */ +/******************************************************************************/ +#ifdef JEMALLOC_H_INLINES + +#ifndef JEMALLOC_ENABLE_INLINE +size_t pow2_ceil(size_t x); +void malloc_write(const char *s); +void set_errno(int errnum); +int get_errno(void); +#endif + +#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_UTIL_C_)) +/* Compute the smallest power of 2 that is >= x. */ +JEMALLOC_INLINE size_t +pow2_ceil(size_t x) +{ + + x--; + x |= x >> 1; + x |= x >> 2; + x |= x >> 4; + x |= x >> 8; + x |= x >> 16; +#if (LG_SIZEOF_PTR == 3) + x |= x >> 32; +#endif + x++; + return (x); +} + +/* Sets error code */ +JEMALLOC_INLINE void +set_errno(int errnum) +{ + +#ifdef _WIN32 + SetLastError(errnum); +#else + errno = errnum; +#endif +} + +/* Get last error code */ +JEMALLOC_INLINE int +get_errno(void) +{ + +#ifdef _WIN32 + return (GetLastError()); +#else + return (errno); +#endif +} +#endif + +#endif /* JEMALLOC_H_INLINES */ +/******************************************************************************/ diff --git a/include/jemalloc/jemalloc.h.in b/include/jemalloc/jemalloc.h.in index 580a5ec5..ad069485 100644 --- a/include/jemalloc/jemalloc.h.in +++ b/include/jemalloc/jemalloc.h.in @@ -15,10 +15,8 @@ extern "C" { #define JEMALLOC_VERSION_GID "@jemalloc_version_gid@" #include "jemalloc_defs@install_suffix@.h" -#ifndef JEMALLOC_P -# define JEMALLOC_P(s) s -#endif +#ifdef JEMALLOC_EXPERIMENTAL #define ALLOCM_LG_ALIGN(la) (la) #if LG_SIZEOF_PTR == 2 #define ALLOCM_ALIGN(a) (ffs(a)-1) @@ -31,34 +29,124 @@ extern "C" { #define ALLOCM_SUCCESS 0 #define ALLOCM_ERR_OOM 1 #define ALLOCM_ERR_NOT_MOVED 2 +#endif -extern const char *JEMALLOC_P(malloc_conf); -extern void (*JEMALLOC_P(malloc_message))(void *, const char *); +/* + * The je_ prefix on the following public symbol declarations is an artifact of + * namespace management, and should be omitted in application code unless + * JEMALLOC_NO_DEMANGLE is defined (see below). + */ +extern JEMALLOC_EXPORT const char *je_malloc_conf; +extern JEMALLOC_EXPORT void (*je_malloc_message)(void *cbopaque, + const char *s); -void *JEMALLOC_P(malloc)(size_t size) JEMALLOC_ATTR(malloc); -void *JEMALLOC_P(calloc)(size_t num, size_t size) JEMALLOC_ATTR(malloc); -int JEMALLOC_P(posix_memalign)(void **memptr, size_t alignment, size_t size) - JEMALLOC_ATTR(nonnull(1)); -void *JEMALLOC_P(realloc)(void *ptr, size_t size); -void JEMALLOC_P(free)(void *ptr); +JEMALLOC_EXPORT void *je_malloc(size_t size) JEMALLOC_ATTR(malloc); +JEMALLOC_EXPORT void *je_calloc(size_t num, size_t size) + JEMALLOC_ATTR(malloc); +JEMALLOC_EXPORT int je_posix_memalign(void **memptr, size_t alignment, + size_t size) JEMALLOC_ATTR(nonnull(1)); +JEMALLOC_EXPORT void *je_aligned_alloc(size_t alignment, size_t size) + JEMALLOC_ATTR(malloc); +JEMALLOC_EXPORT void *je_realloc(void *ptr, size_t size); +JEMALLOC_EXPORT void je_free(void *ptr); -size_t JEMALLOC_P(malloc_usable_size)(const void *ptr); -void JEMALLOC_P(malloc_stats_print)(void (*write_cb)(void *, const char *), - void *cbopaque, const char *opts); -int JEMALLOC_P(mallctl)(const char *name, void *oldp, size_t *oldlenp, - void *newp, size_t newlen); -int JEMALLOC_P(mallctlnametomib)(const char *name, size_t *mibp, - size_t *miblenp); -int JEMALLOC_P(mallctlbymib)(const size_t *mib, size_t miblen, void *oldp, +#ifdef JEMALLOC_OVERRIDE_MEMALIGN +JEMALLOC_EXPORT void * je_memalign(size_t alignment, size_t size) + JEMALLOC_ATTR(malloc); +#endif + +#ifdef JEMALLOC_OVERRIDE_VALLOC +JEMALLOC_EXPORT void * je_valloc(size_t size) JEMALLOC_ATTR(malloc); +#endif + +JEMALLOC_EXPORT size_t je_malloc_usable_size(const void *ptr); +JEMALLOC_EXPORT void je_malloc_stats_print(void (*write_cb)(void *, + const char *), void *je_cbopaque, const char *opts); +JEMALLOC_EXPORT int je_mallctl(const char *name, void *oldp, size_t *oldlenp, void *newp, size_t newlen); +JEMALLOC_EXPORT int je_mallctlnametomib(const char *name, size_t *mibp, + size_t *miblenp); +JEMALLOC_EXPORT int je_mallctlbymib(const size_t *mib, size_t miblen, + void *oldp, size_t *oldlenp, void *newp, size_t newlen); -int JEMALLOC_P(allocm)(void **ptr, size_t *rsize, size_t size, int flags) - JEMALLOC_ATTR(nonnull(1)); -int JEMALLOC_P(rallocm)(void **ptr, size_t *rsize, size_t size, +#ifdef JEMALLOC_EXPERIMENTAL +JEMALLOC_EXPORT int je_allocm(void **ptr, size_t *rsize, size_t size, + int flags) JEMALLOC_ATTR(nonnull(1)); +JEMALLOC_EXPORT int je_rallocm(void **ptr, size_t *rsize, size_t size, size_t extra, int flags) JEMALLOC_ATTR(nonnull(1)); -int JEMALLOC_P(sallocm)(const void *ptr, size_t *rsize, int flags) +JEMALLOC_EXPORT int je_sallocm(const void *ptr, size_t *rsize, int flags) JEMALLOC_ATTR(nonnull(1)); -int JEMALLOC_P(dallocm)(void *ptr, int flags) JEMALLOC_ATTR(nonnull(1)); +JEMALLOC_EXPORT int je_dallocm(void *ptr, int flags) + JEMALLOC_ATTR(nonnull(1)); +JEMALLOC_EXPORT int je_nallocm(size_t *rsize, size_t size, int flags); +#endif + +/* + * By default application code must explicitly refer to mangled symbol names, + * so that it is possible to use jemalloc in conjunction with another allocator + * in the same application. Define JEMALLOC_MANGLE in order to cause automatic + * name mangling that matches the API prefixing that happened as a result of + * --with-mangling and/or --with-jemalloc-prefix configuration settings. + */ +#ifdef JEMALLOC_MANGLE +#ifndef JEMALLOC_NO_DEMANGLE +#define JEMALLOC_NO_DEMANGLE +#endif +#define malloc_conf je_malloc_conf +#define malloc_message je_malloc_message +#define malloc je_malloc +#define calloc je_calloc +#define posix_memalign je_posix_memalign +#define aligned_alloc je_aligned_alloc +#define realloc je_realloc +#define free je_free +#define malloc_usable_size je_malloc_usable_size +#define malloc_stats_print je_malloc_stats_print +#define mallctl je_mallctl +#define mallctlnametomib je_mallctlnametomib +#define mallctlbymib je_mallctlbymib +#define memalign je_memalign +#define valloc je_valloc +#ifdef JEMALLOC_EXPERIMENTAL +#define allocm je_allocm +#define rallocm je_rallocm +#define sallocm je_sallocm +#define dallocm je_dallocm +#define nallocm je_nallocm +#endif +#endif + +/* + * The je_* macros can be used as stable alternative names for the public + * jemalloc API if JEMALLOC_NO_DEMANGLE is defined. This is primarily meant + * for use in jemalloc itself, but it can be used by application code to + * provide isolation from the name mangling specified via --with-mangling + * and/or --with-jemalloc-prefix. + */ +#ifndef JEMALLOC_NO_DEMANGLE +#undef je_malloc_conf +#undef je_malloc_message +#undef je_malloc +#undef je_calloc +#undef je_posix_memalign +#undef je_aligned_alloc +#undef je_realloc +#undef je_free +#undef je_malloc_usable_size +#undef je_malloc_stats_print +#undef je_mallctl +#undef je_mallctlnametomib +#undef je_mallctlbymib +#undef je_memalign +#undef je_valloc +#ifdef JEMALLOC_EXPERIMENTAL +#undef je_allocm +#undef je_rallocm +#undef je_sallocm +#undef je_dallocm +#undef je_nallocm +#endif +#endif #ifdef __cplusplus }; diff --git a/include/jemalloc/jemalloc_defs.h.in b/include/jemalloc/jemalloc_defs.h.in index 9ac7e1c2..c469142a 100644 --- a/include/jemalloc/jemalloc_defs.h.in +++ b/include/jemalloc/jemalloc_defs.h.in @@ -1,22 +1,36 @@ -#ifndef JEMALLOC_DEFS_H_ -#define JEMALLOC_DEFS_H_ - /* - * If JEMALLOC_PREFIX is defined, it will cause all public APIs to be prefixed. - * This makes it possible, with some care, to use multiple allocators - * simultaneously. - * - * In many cases it is more convenient to manually prefix allocator function - * calls than to let macros do it automatically, particularly when using - * multiple allocators simultaneously. Define JEMALLOC_MANGLE before - * #include'ing jemalloc.h in order to cause name mangling that corresponds to - * the API prefixing. + * If JEMALLOC_PREFIX is defined via --with-jemalloc-prefix, it will cause all + * public APIs to be prefixed. This makes it possible, with some care, to use + * multiple allocators simultaneously. */ #undef JEMALLOC_PREFIX #undef JEMALLOC_CPREFIX -#if (defined(JEMALLOC_PREFIX) && defined(JEMALLOC_MANGLE)) -#undef JEMALLOC_P -#endif + +/* + * Name mangling for public symbols is controlled by --with-mangling and + * --with-jemalloc-prefix. With default settings the je_ prefix is stripped by + * these macro definitions. + */ +#undef je_malloc_conf +#undef je_malloc_message +#undef je_malloc +#undef je_calloc +#undef je_posix_memalign +#undef je_aligned_alloc +#undef je_realloc +#undef je_free +#undef je_malloc_usable_size +#undef je_malloc_stats_print +#undef je_mallctl +#undef je_mallctlnametomib +#undef je_mallctlbymib +#undef je_memalign +#undef je_valloc +#undef je_allocm +#undef je_rallocm +#undef je_sallocm +#undef je_dallocm +#undef je_nallocm /* * JEMALLOC_PRIVATE_NAMESPACE is used as a prefix for all library-private APIs. @@ -33,26 +47,92 @@ */ #undef CPU_SPINWAIT +/* Defined if the equivalent of FreeBSD's atomic(9) functions are available. */ +#undef JEMALLOC_ATOMIC9 + /* * Defined if OSAtomic*() functions are available, as provided by Darwin, and * documented in the atomic(3) manual page. */ #undef JEMALLOC_OSATOMIC +/* + * Defined if __sync_add_and_fetch(uint32_t *, uint32_t) and + * __sync_sub_and_fetch(uint32_t *, uint32_t) are available, despite + * __GCC_HAVE_SYNC_COMPARE_AND_SWAP_4 not being defined (which means the + * functions are defined in libgcc instead of being inlines) + */ +#undef JE_FORCE_SYNC_COMPARE_AND_SWAP_4 + +/* + * Defined if __sync_add_and_fetch(uint64_t *, uint64_t) and + * __sync_sub_and_fetch(uint64_t *, uint64_t) are available, despite + * __GCC_HAVE_SYNC_COMPARE_AND_SWAP_8 not being defined (which means the + * functions are defined in libgcc instead of being inlines) + */ +#undef JE_FORCE_SYNC_COMPARE_AND_SWAP_8 + /* * Defined if OSSpin*() functions are available, as provided by Darwin, and * documented in the spinlock(3) manual page. */ #undef JEMALLOC_OSSPIN +/* + * Defined if _malloc_thread_cleanup() exists. At least in the case of + * FreeBSD, pthread_key_create() allocates, which if used during malloc + * bootstrapping will cause recursion into the pthreads library. Therefore, if + * _malloc_thread_cleanup() exists, use it as the basis for thread cleanup in + * malloc_tsd. + */ +#undef JEMALLOC_MALLOC_THREAD_CLEANUP + +/* + * Defined if threaded initialization is known to be safe on this platform. + * Among other things, it must be possible to initialize a mutex without + * triggering allocation in order for threaded allocation to be safe. + */ +#undef JEMALLOC_THREADED_INIT + +/* + * Defined if the pthreads implementation defines + * _pthread_mutex_init_calloc_cb(), in which case the function is used in order + * to avoid recursive allocation during mutex initialization. + */ +#undef JEMALLOC_MUTEX_INIT_CB + /* Defined if __attribute__((...)) syntax is supported. */ #undef JEMALLOC_HAVE_ATTR #ifdef JEMALLOC_HAVE_ATTR # define JEMALLOC_ATTR(s) __attribute__((s)) +# define JEMALLOC_EXPORT JEMALLOC_ATTR(visibility("default")) +# define JEMALLOC_ALIGNED(s) JEMALLOC_ATTR(aligned(s)) +# define JEMALLOC_SECTION(s) JEMALLOC_ATTR(section(s)) +# define JEMALLOC_NOINLINE JEMALLOC_ATTR(noinline) +#elif _MSC_VER +# define JEMALLOC_ATTR(s) +# ifdef DLLEXPORT +# define JEMALLOC_EXPORT __declspec(dllexport) +# else +# define JEMALLOC_EXPORT __declspec(dllimport) +# endif +# define JEMALLOC_ALIGNED(s) __declspec(align(s)) +# define JEMALLOC_SECTION(s) __declspec(allocate(s)) +# define JEMALLOC_NOINLINE __declspec(noinline) #else # define JEMALLOC_ATTR(s) +# define JEMALLOC_EXPORT +# define JEMALLOC_ALIGNED(s) +# define JEMALLOC_SECTION(s) +# define JEMALLOC_NOINLINE #endif +/* Defined if sbrk() is supported. */ +#undef JEMALLOC_HAVE_SBRK + +/* Non-empty if the tls_model attribute is supported. */ +#undef JEMALLOC_TLS_MODEL + /* JEMALLOC_CC_SILENCE enables code that silences unuseful compiler warnings. */ #undef JEMALLOC_CC_SILENCE @@ -77,12 +157,6 @@ /* Use gcc intrinsics for profile backtracing if defined. */ #undef JEMALLOC_PROF_GCC -/* - * JEMALLOC_TINY enables support for tiny objects, which are smaller than one - * quantum. - */ -#undef JEMALLOC_TINY - /* * JEMALLOC_TCACHE enables a thread-specific caching layer for small objects. * This makes it possible to allocate/deallocate objects without any locking @@ -96,29 +170,43 @@ */ #undef JEMALLOC_DSS -/* JEMALLOC_SWAP enables mmap()ed swap file support. */ -#undef JEMALLOC_SWAP - -/* Support memory filling (junk/zero). */ +/* Support memory filling (junk/zero/quarantine/redzone). */ #undef JEMALLOC_FILL +/* Support the experimental API. */ +#undef JEMALLOC_EXPERIMENTAL + +/* Support utrace(2)-based tracing. */ +#undef JEMALLOC_UTRACE + +/* Support Valgrind. */ +#undef JEMALLOC_VALGRIND + /* Support optional abort() on OOM. */ #undef JEMALLOC_XMALLOC -/* Support SYSV semantics. */ -#undef JEMALLOC_SYSV - /* Support lazy locking (avoid locking unless a second thread is launched). */ #undef JEMALLOC_LAZY_LOCK -/* Determine page size at run time if defined. */ -#undef DYNAMIC_PAGE_SHIFT - /* One page is 2^STATIC_PAGE_SHIFT bytes. */ #undef STATIC_PAGE_SHIFT +/* + * If defined, use munmap() to unmap freed chunks, rather than storing them for + * later reuse. This is disabled by default on Linux because common sequences + * of mmap()/munmap() calls will cause virtual memory map holes. + */ +#undef JEMALLOC_MUNMAP + +/* + * If defined, use mremap(...MREMAP_FIXED...) for huge realloc(). This is + * disabled by default because it is Linux-specific and it will cause virtual + * memory map holes, much like munmap(2) does. + */ +#undef JEMALLOC_MREMAP + /* TLS is used to map arenas and magazine caches to threads. */ -#undef NO_TLS +#undef JEMALLOC_TLS /* * JEMALLOC_IVSALLOC enables ivsalloc(), which verifies that pointers reside @@ -139,9 +227,6 @@ #undef JEMALLOC_ZONE #undef JEMALLOC_ZONE_VERSION -/* If defined, use mremap(...MREMAP_FIXED...) for huge realloc(). */ -#undef JEMALLOC_MREMAP_FIXED - /* * Methods for purging unused pages differ between operating systems. * @@ -164,4 +249,5 @@ /* sizeof(long) == 2^LG_SIZEOF_LONG. */ #undef LG_SIZEOF_LONG -#endif /* JEMALLOC_DEFS_H_ */ +/* sizeof(intmax_t) == 2^LG_SIZEOF_INTMAX_T. */ +#undef LG_SIZEOF_INTMAX_T diff --git a/include/msvc_compat/inttypes.h b/include/msvc_compat/inttypes.h new file mode 100644 index 00000000..a4e6b75c --- /dev/null +++ b/include/msvc_compat/inttypes.h @@ -0,0 +1,313 @@ +// ISO C9x compliant inttypes.h for Microsoft Visual Studio +// Based on ISO/IEC 9899:TC2 Committee draft (May 6, 2005) WG14/N1124 +// +// Copyright (c) 2006 Alexander Chemeris +// +// Redistribution and use in source and binary forms, with or without +// modification, are permitted provided that the following conditions are met: +// +// 1. Redistributions of source code must retain the above copyright notice, +// this list of conditions and the following disclaimer. +// +// 2. Redistributions in binary form must reproduce the above copyright +// notice, this list of conditions and the following disclaimer in the +// documentation and/or other materials provided with the distribution. +// +// 3. The name of the author may be used to endorse or promote products +// derived from this software without specific prior written permission. +// +// THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED +// WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF +// MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO +// EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, +// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; +// OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, +// WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR +// OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF +// ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +// +/////////////////////////////////////////////////////////////////////////////// + +#ifndef _MSC_VER // [ +#error "Use this header only with Microsoft Visual C++ compilers!" +#endif // _MSC_VER ] + +#ifndef _MSC_INTTYPES_H_ // [ +#define _MSC_INTTYPES_H_ + +#if _MSC_VER > 1000 +#pragma once +#endif + +#include "stdint.h" + +// 7.8 Format conversion of integer types + +typedef struct { + intmax_t quot; + intmax_t rem; +} imaxdiv_t; + +// 7.8.1 Macros for format specifiers + +#if !defined(__cplusplus) || defined(__STDC_FORMAT_MACROS) // [ See footnote 185 at page 198 + +#ifdef _WIN64 +# define __PRI64_PREFIX "l" +# define __PRIPTR_PREFIX "l" +#else +# define __PRI64_PREFIX "ll" +# define __PRIPTR_PREFIX +#endif + +// The fprintf macros for signed integers are: +#define PRId8 "d" +#define PRIi8 "i" +#define PRIdLEAST8 "d" +#define PRIiLEAST8 "i" +#define PRIdFAST8 "d" +#define PRIiFAST8 "i" + +#define PRId16 "hd" +#define PRIi16 "hi" +#define PRIdLEAST16 "hd" +#define PRIiLEAST16 "hi" +#define PRIdFAST16 "hd" +#define PRIiFAST16 "hi" + +#define PRId32 "d" +#define PRIi32 "i" +#define PRIdLEAST32 "d" +#define PRIiLEAST32 "i" +#define PRIdFAST32 "d" +#define PRIiFAST32 "i" + +#define PRId64 __PRI64_PREFIX "d" +#define PRIi64 __PRI64_PREFIX "i" +#define PRIdLEAST64 __PRI64_PREFIX "d" +#define PRIiLEAST64 __PRI64_PREFIX "i" +#define PRIdFAST64 __PRI64_PREFIX "d" +#define PRIiFAST64 __PRI64_PREFIX "i" + +#define PRIdMAX __PRI64_PREFIX "d" +#define PRIiMAX __PRI64_PREFIX "i" + +#define PRIdPTR __PRIPTR_PREFIX "d" +#define PRIiPTR __PRIPTR_PREFIX "i" + +// The fprintf macros for unsigned integers are: +#define PRIo8 "o" +#define PRIu8 "u" +#define PRIx8 "x" +#define PRIX8 "X" +#define PRIoLEAST8 "o" +#define PRIuLEAST8 "u" +#define PRIxLEAST8 "x" +#define PRIXLEAST8 "X" +#define PRIoFAST8 "o" +#define PRIuFAST8 "u" +#define PRIxFAST8 "x" +#define PRIXFAST8 "X" + +#define PRIo16 "ho" +#define PRIu16 "hu" +#define PRIx16 "hx" +#define PRIX16 "hX" +#define PRIoLEAST16 "ho" +#define PRIuLEAST16 "hu" +#define PRIxLEAST16 "hx" +#define PRIXLEAST16 "hX" +#define PRIoFAST16 "ho" +#define PRIuFAST16 "hu" +#define PRIxFAST16 "hx" +#define PRIXFAST16 "hX" + +#define PRIo32 "o" +#define PRIu32 "u" +#define PRIx32 "x" +#define PRIX32 "X" +#define PRIoLEAST32 "o" +#define PRIuLEAST32 "u" +#define PRIxLEAST32 "x" +#define PRIXLEAST32 "X" +#define PRIoFAST32 "o" +#define PRIuFAST32 "u" +#define PRIxFAST32 "x" +#define PRIXFAST32 "X" + +#define PRIo64 __PRI64_PREFIX "o" +#define PRIu64 __PRI64_PREFIX "u" +#define PRIx64 __PRI64_PREFIX "x" +#define PRIX64 __PRI64_PREFIX "X" +#define PRIoLEAST64 __PRI64_PREFIX "o" +#define PRIuLEAST64 __PRI64_PREFIX "u" +#define PRIxLEAST64 __PRI64_PREFIX "x" +#define PRIXLEAST64 __PRI64_PREFIX "X" +#define PRIoFAST64 __PRI64_PREFIX "o" +#define PRIuFAST64 __PRI64_PREFIX "u" +#define PRIxFAST64 __PRI64_PREFIX "x" +#define PRIXFAST64 __PRI64_PREFIX "X" + +#define PRIoMAX __PRI64_PREFIX "o" +#define PRIuMAX __PRI64_PREFIX "u" +#define PRIxMAX __PRI64_PREFIX "x" +#define PRIXMAX __PRI64_PREFIX "X" + +#define PRIoPTR __PRIPTR_PREFIX "o" +#define PRIuPTR __PRIPTR_PREFIX "u" +#define PRIxPTR __PRIPTR_PREFIX "x" +#define PRIXPTR __PRIPTR_PREFIX "X" + +// The fscanf macros for signed integers are: +#define SCNd8 "d" +#define SCNi8 "i" +#define SCNdLEAST8 "d" +#define SCNiLEAST8 "i" +#define SCNdFAST8 "d" +#define SCNiFAST8 "i" + +#define SCNd16 "hd" +#define SCNi16 "hi" +#define SCNdLEAST16 "hd" +#define SCNiLEAST16 "hi" +#define SCNdFAST16 "hd" +#define SCNiFAST16 "hi" + +#define SCNd32 "ld" +#define SCNi32 "li" +#define SCNdLEAST32 "ld" +#define SCNiLEAST32 "li" +#define SCNdFAST32 "ld" +#define SCNiFAST32 "li" + +#define SCNd64 "I64d" +#define SCNi64 "I64i" +#define SCNdLEAST64 "I64d" +#define SCNiLEAST64 "I64i" +#define SCNdFAST64 "I64d" +#define SCNiFAST64 "I64i" + +#define SCNdMAX "I64d" +#define SCNiMAX "I64i" + +#ifdef _WIN64 // [ +# define SCNdPTR "I64d" +# define SCNiPTR "I64i" +#else // _WIN64 ][ +# define SCNdPTR "ld" +# define SCNiPTR "li" +#endif // _WIN64 ] + +// The fscanf macros for unsigned integers are: +#define SCNo8 "o" +#define SCNu8 "u" +#define SCNx8 "x" +#define SCNX8 "X" +#define SCNoLEAST8 "o" +#define SCNuLEAST8 "u" +#define SCNxLEAST8 "x" +#define SCNXLEAST8 "X" +#define SCNoFAST8 "o" +#define SCNuFAST8 "u" +#define SCNxFAST8 "x" +#define SCNXFAST8 "X" + +#define SCNo16 "ho" +#define SCNu16 "hu" +#define SCNx16 "hx" +#define SCNX16 "hX" +#define SCNoLEAST16 "ho" +#define SCNuLEAST16 "hu" +#define SCNxLEAST16 "hx" +#define SCNXLEAST16 "hX" +#define SCNoFAST16 "ho" +#define SCNuFAST16 "hu" +#define SCNxFAST16 "hx" +#define SCNXFAST16 "hX" + +#define SCNo32 "lo" +#define SCNu32 "lu" +#define SCNx32 "lx" +#define SCNX32 "lX" +#define SCNoLEAST32 "lo" +#define SCNuLEAST32 "lu" +#define SCNxLEAST32 "lx" +#define SCNXLEAST32 "lX" +#define SCNoFAST32 "lo" +#define SCNuFAST32 "lu" +#define SCNxFAST32 "lx" +#define SCNXFAST32 "lX" + +#define SCNo64 "I64o" +#define SCNu64 "I64u" +#define SCNx64 "I64x" +#define SCNX64 "I64X" +#define SCNoLEAST64 "I64o" +#define SCNuLEAST64 "I64u" +#define SCNxLEAST64 "I64x" +#define SCNXLEAST64 "I64X" +#define SCNoFAST64 "I64o" +#define SCNuFAST64 "I64u" +#define SCNxFAST64 "I64x" +#define SCNXFAST64 "I64X" + +#define SCNoMAX "I64o" +#define SCNuMAX "I64u" +#define SCNxMAX "I64x" +#define SCNXMAX "I64X" + +#ifdef _WIN64 // [ +# define SCNoPTR "I64o" +# define SCNuPTR "I64u" +# define SCNxPTR "I64x" +# define SCNXPTR "I64X" +#else // _WIN64 ][ +# define SCNoPTR "lo" +# define SCNuPTR "lu" +# define SCNxPTR "lx" +# define SCNXPTR "lX" +#endif // _WIN64 ] + +#endif // __STDC_FORMAT_MACROS ] + +// 7.8.2 Functions for greatest-width integer types + +// 7.8.2.1 The imaxabs function +#define imaxabs _abs64 + +// 7.8.2.2 The imaxdiv function + +// This is modified version of div() function from Microsoft's div.c found +// in %MSVC.NET%\crt\src\div.c +#ifdef STATIC_IMAXDIV // [ +static +#else // STATIC_IMAXDIV ][ +_inline +#endif // STATIC_IMAXDIV ] +imaxdiv_t __cdecl imaxdiv(intmax_t numer, intmax_t denom) +{ + imaxdiv_t result; + + result.quot = numer / denom; + result.rem = numer % denom; + + if (numer < 0 && result.rem > 0) { + // did division wrong; must fix up + ++result.quot; + result.rem -= denom; + } + + return result; +} + +// 7.8.2.3 The strtoimax and strtoumax functions +#define strtoimax _strtoi64 +#define strtoumax _strtoui64 + +// 7.8.2.4 The wcstoimax and wcstoumax functions +#define wcstoimax _wcstoi64 +#define wcstoumax _wcstoui64 + + +#endif // _MSC_INTTYPES_H_ ] diff --git a/include/msvc_compat/stdbool.h b/include/msvc_compat/stdbool.h new file mode 100644 index 00000000..da9ee8b8 --- /dev/null +++ b/include/msvc_compat/stdbool.h @@ -0,0 +1,16 @@ +#ifndef stdbool_h +#define stdbool_h + +#include + +/* MSVC doesn't define _Bool or bool in C, but does have BOOL */ +/* Note this doesn't pass autoconf's test because (bool) 0.5 != true */ +typedef BOOL _Bool; + +#define bool _Bool +#define true 1 +#define false 0 + +#define __bool_true_false_are_defined 1 + +#endif /* stdbool_h */ diff --git a/include/msvc_compat/stdint.h b/include/msvc_compat/stdint.h new file mode 100644 index 00000000..d02608a5 --- /dev/null +++ b/include/msvc_compat/stdint.h @@ -0,0 +1,247 @@ +// ISO C9x compliant stdint.h for Microsoft Visual Studio +// Based on ISO/IEC 9899:TC2 Committee draft (May 6, 2005) WG14/N1124 +// +// Copyright (c) 2006-2008 Alexander Chemeris +// +// Redistribution and use in source and binary forms, with or without +// modification, are permitted provided that the following conditions are met: +// +// 1. Redistributions of source code must retain the above copyright notice, +// this list of conditions and the following disclaimer. +// +// 2. Redistributions in binary form must reproduce the above copyright +// notice, this list of conditions and the following disclaimer in the +// documentation and/or other materials provided with the distribution. +// +// 3. The name of the author may be used to endorse or promote products +// derived from this software without specific prior written permission. +// +// THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED +// WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF +// MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO +// EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, +// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; +// OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, +// WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR +// OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF +// ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +// +/////////////////////////////////////////////////////////////////////////////// + +#ifndef _MSC_VER // [ +#error "Use this header only with Microsoft Visual C++ compilers!" +#endif // _MSC_VER ] + +#ifndef _MSC_STDINT_H_ // [ +#define _MSC_STDINT_H_ + +#if _MSC_VER > 1000 +#pragma once +#endif + +#include + +// For Visual Studio 6 in C++ mode and for many Visual Studio versions when +// compiling for ARM we should wrap include with 'extern "C++" {}' +// or compiler give many errors like this: +// error C2733: second C linkage of overloaded function 'wmemchr' not allowed +#ifdef __cplusplus +extern "C" { +#endif +# include +#ifdef __cplusplus +} +#endif + +// Define _W64 macros to mark types changing their size, like intptr_t. +#ifndef _W64 +# if !defined(__midl) && (defined(_X86_) || defined(_M_IX86)) && _MSC_VER >= 1300 +# define _W64 __w64 +# else +# define _W64 +# endif +#endif + + +// 7.18.1 Integer types + +// 7.18.1.1 Exact-width integer types + +// Visual Studio 6 and Embedded Visual C++ 4 doesn't +// realize that, e.g. char has the same size as __int8 +// so we give up on __intX for them. +#if (_MSC_VER < 1300) + typedef signed char int8_t; + typedef signed short int16_t; + typedef signed int int32_t; + typedef unsigned char uint8_t; + typedef unsigned short uint16_t; + typedef unsigned int uint32_t; +#else + typedef signed __int8 int8_t; + typedef signed __int16 int16_t; + typedef signed __int32 int32_t; + typedef unsigned __int8 uint8_t; + typedef unsigned __int16 uint16_t; + typedef unsigned __int32 uint32_t; +#endif +typedef signed __int64 int64_t; +typedef unsigned __int64 uint64_t; + + +// 7.18.1.2 Minimum-width integer types +typedef int8_t int_least8_t; +typedef int16_t int_least16_t; +typedef int32_t int_least32_t; +typedef int64_t int_least64_t; +typedef uint8_t uint_least8_t; +typedef uint16_t uint_least16_t; +typedef uint32_t uint_least32_t; +typedef uint64_t uint_least64_t; + +// 7.18.1.3 Fastest minimum-width integer types +typedef int8_t int_fast8_t; +typedef int16_t int_fast16_t; +typedef int32_t int_fast32_t; +typedef int64_t int_fast64_t; +typedef uint8_t uint_fast8_t; +typedef uint16_t uint_fast16_t; +typedef uint32_t uint_fast32_t; +typedef uint64_t uint_fast64_t; + +// 7.18.1.4 Integer types capable of holding object pointers +#ifdef _WIN64 // [ + typedef signed __int64 intptr_t; + typedef unsigned __int64 uintptr_t; +#else // _WIN64 ][ + typedef _W64 signed int intptr_t; + typedef _W64 unsigned int uintptr_t; +#endif // _WIN64 ] + +// 7.18.1.5 Greatest-width integer types +typedef int64_t intmax_t; +typedef uint64_t uintmax_t; + + +// 7.18.2 Limits of specified-width integer types + +#if !defined(__cplusplus) || defined(__STDC_LIMIT_MACROS) // [ See footnote 220 at page 257 and footnote 221 at page 259 + +// 7.18.2.1 Limits of exact-width integer types +#define INT8_MIN ((int8_t)_I8_MIN) +#define INT8_MAX _I8_MAX +#define INT16_MIN ((int16_t)_I16_MIN) +#define INT16_MAX _I16_MAX +#define INT32_MIN ((int32_t)_I32_MIN) +#define INT32_MAX _I32_MAX +#define INT64_MIN ((int64_t)_I64_MIN) +#define INT64_MAX _I64_MAX +#define UINT8_MAX _UI8_MAX +#define UINT16_MAX _UI16_MAX +#define UINT32_MAX _UI32_MAX +#define UINT64_MAX _UI64_MAX + +// 7.18.2.2 Limits of minimum-width integer types +#define INT_LEAST8_MIN INT8_MIN +#define INT_LEAST8_MAX INT8_MAX +#define INT_LEAST16_MIN INT16_MIN +#define INT_LEAST16_MAX INT16_MAX +#define INT_LEAST32_MIN INT32_MIN +#define INT_LEAST32_MAX INT32_MAX +#define INT_LEAST64_MIN INT64_MIN +#define INT_LEAST64_MAX INT64_MAX +#define UINT_LEAST8_MAX UINT8_MAX +#define UINT_LEAST16_MAX UINT16_MAX +#define UINT_LEAST32_MAX UINT32_MAX +#define UINT_LEAST64_MAX UINT64_MAX + +// 7.18.2.3 Limits of fastest minimum-width integer types +#define INT_FAST8_MIN INT8_MIN +#define INT_FAST8_MAX INT8_MAX +#define INT_FAST16_MIN INT16_MIN +#define INT_FAST16_MAX INT16_MAX +#define INT_FAST32_MIN INT32_MIN +#define INT_FAST32_MAX INT32_MAX +#define INT_FAST64_MIN INT64_MIN +#define INT_FAST64_MAX INT64_MAX +#define UINT_FAST8_MAX UINT8_MAX +#define UINT_FAST16_MAX UINT16_MAX +#define UINT_FAST32_MAX UINT32_MAX +#define UINT_FAST64_MAX UINT64_MAX + +// 7.18.2.4 Limits of integer types capable of holding object pointers +#ifdef _WIN64 // [ +# define INTPTR_MIN INT64_MIN +# define INTPTR_MAX INT64_MAX +# define UINTPTR_MAX UINT64_MAX +#else // _WIN64 ][ +# define INTPTR_MIN INT32_MIN +# define INTPTR_MAX INT32_MAX +# define UINTPTR_MAX UINT32_MAX +#endif // _WIN64 ] + +// 7.18.2.5 Limits of greatest-width integer types +#define INTMAX_MIN INT64_MIN +#define INTMAX_MAX INT64_MAX +#define UINTMAX_MAX UINT64_MAX + +// 7.18.3 Limits of other integer types + +#ifdef _WIN64 // [ +# define PTRDIFF_MIN _I64_MIN +# define PTRDIFF_MAX _I64_MAX +#else // _WIN64 ][ +# define PTRDIFF_MIN _I32_MIN +# define PTRDIFF_MAX _I32_MAX +#endif // _WIN64 ] + +#define SIG_ATOMIC_MIN INT_MIN +#define SIG_ATOMIC_MAX INT_MAX + +#ifndef SIZE_MAX // [ +# ifdef _WIN64 // [ +# define SIZE_MAX _UI64_MAX +# else // _WIN64 ][ +# define SIZE_MAX _UI32_MAX +# endif // _WIN64 ] +#endif // SIZE_MAX ] + +// WCHAR_MIN and WCHAR_MAX are also defined in +#ifndef WCHAR_MIN // [ +# define WCHAR_MIN 0 +#endif // WCHAR_MIN ] +#ifndef WCHAR_MAX // [ +# define WCHAR_MAX _UI16_MAX +#endif // WCHAR_MAX ] + +#define WINT_MIN 0 +#define WINT_MAX _UI16_MAX + +#endif // __STDC_LIMIT_MACROS ] + + +// 7.18.4 Limits of other integer types + +#if !defined(__cplusplus) || defined(__STDC_CONSTANT_MACROS) // [ See footnote 224 at page 260 + +// 7.18.4.1 Macros for minimum-width integer constants + +#define INT8_C(val) val##i8 +#define INT16_C(val) val##i16 +#define INT32_C(val) val##i32 +#define INT64_C(val) val##i64 + +#define UINT8_C(val) val##ui8 +#define UINT16_C(val) val##ui16 +#define UINT32_C(val) val##ui32 +#define UINT64_C(val) val##ui64 + +// 7.18.4.2 Macros for greatest-width integer constants +#define INTMAX_C INT64_C +#define UINTMAX_C UINT64_C + +#endif // __STDC_CONSTANT_MACROS ] + + +#endif // _MSC_STDINT_H_ ] diff --git a/include/msvc_compat/strings.h b/include/msvc_compat/strings.h new file mode 100644 index 00000000..c84975b6 --- /dev/null +++ b/include/msvc_compat/strings.h @@ -0,0 +1,23 @@ +#ifndef strings_h +#define strings_h + +/* MSVC doesn't define ffs/ffsl. This dummy strings.h header is provided + * for both */ +#include +#pragma intrinsic(_BitScanForward) +static __forceinline int ffsl(long x) +{ + unsigned long i; + + if (_BitScanForward(&i, x)) + return (i + 1); + return (0); +} + +static __forceinline int ffs(int x) +{ + + return (ffsl(x)); +} + +#endif diff --git a/src/arena.c b/src/arena.c index d166ca1e..2a6150f3 100644 --- a/src/arena.c +++ b/src/arena.c @@ -4,175 +4,60 @@ /******************************************************************************/ /* Data. */ -size_t opt_lg_qspace_max = LG_QSPACE_MAX_DEFAULT; -size_t opt_lg_cspace_max = LG_CSPACE_MAX_DEFAULT; ssize_t opt_lg_dirty_mult = LG_DIRTY_MULT_DEFAULT; -uint8_t const *small_size2bin; -arena_bin_info_t *arena_bin_info; +arena_bin_info_t arena_bin_info[NBINS]; -/* Various bin-related settings. */ -unsigned nqbins; -unsigned ncbins; -unsigned nsbins; -unsigned nbins; -size_t qspace_max; -size_t cspace_min; -size_t cspace_max; -size_t sspace_min; -size_t sspace_max; - -size_t lg_mspace; -size_t mspace_mask; - -/* - * const_small_size2bin is a static constant lookup table that in the common - * case can be used as-is for small_size2bin. - */ -#if (LG_TINY_MIN == 2) -#define S2B_4(i) i, -#define S2B_8(i) S2B_4(i) S2B_4(i) -#elif (LG_TINY_MIN == 3) +JEMALLOC_ALIGNED(CACHELINE) +const uint8_t small_size2bin[] = { #define S2B_8(i) i, -#else -# error "Unsupported LG_TINY_MIN" -#endif #define S2B_16(i) S2B_8(i) S2B_8(i) #define S2B_32(i) S2B_16(i) S2B_16(i) #define S2B_64(i) S2B_32(i) S2B_32(i) #define S2B_128(i) S2B_64(i) S2B_64(i) #define S2B_256(i) S2B_128(i) S2B_128(i) -/* - * The number of elements in const_small_size2bin is dependent on the - * definition for SUBPAGE. - */ -static JEMALLOC_ATTR(aligned(CACHELINE)) - const uint8_t const_small_size2bin[] = { -#if (LG_QUANTUM == 4) -/* 16-byte quantum **********************/ -# ifdef JEMALLOC_TINY -# if (LG_TINY_MIN == 2) - S2B_4(0) /* 4 */ - S2B_4(1) /* 8 */ - S2B_8(2) /* 16 */ -# define S2B_QMIN 2 -# elif (LG_TINY_MIN == 3) - S2B_8(0) /* 8 */ - S2B_8(1) /* 16 */ -# define S2B_QMIN 1 -# else -# error "Unsupported LG_TINY_MIN" -# endif -# else - S2B_16(0) /* 16 */ -# define S2B_QMIN 0 -# endif - S2B_16(S2B_QMIN + 1) /* 32 */ - S2B_16(S2B_QMIN + 2) /* 48 */ - S2B_16(S2B_QMIN + 3) /* 64 */ - S2B_16(S2B_QMIN + 4) /* 80 */ - S2B_16(S2B_QMIN + 5) /* 96 */ - S2B_16(S2B_QMIN + 6) /* 112 */ - S2B_16(S2B_QMIN + 7) /* 128 */ -# define S2B_CMIN (S2B_QMIN + 8) -#else -/* 8-byte quantum ***********************/ -# ifdef JEMALLOC_TINY -# if (LG_TINY_MIN == 2) - S2B_4(0) /* 4 */ - S2B_4(1) /* 8 */ -# define S2B_QMIN 1 -# else -# error "Unsupported LG_TINY_MIN" -# endif -# else - S2B_8(0) /* 8 */ -# define S2B_QMIN 0 -# endif - S2B_8(S2B_QMIN + 1) /* 16 */ - S2B_8(S2B_QMIN + 2) /* 24 */ - S2B_8(S2B_QMIN + 3) /* 32 */ - S2B_8(S2B_QMIN + 4) /* 40 */ - S2B_8(S2B_QMIN + 5) /* 48 */ - S2B_8(S2B_QMIN + 6) /* 56 */ - S2B_8(S2B_QMIN + 7) /* 64 */ - S2B_8(S2B_QMIN + 8) /* 72 */ - S2B_8(S2B_QMIN + 9) /* 80 */ - S2B_8(S2B_QMIN + 10) /* 88 */ - S2B_8(S2B_QMIN + 11) /* 96 */ - S2B_8(S2B_QMIN + 12) /* 104 */ - S2B_8(S2B_QMIN + 13) /* 112 */ - S2B_8(S2B_QMIN + 14) /* 120 */ - S2B_8(S2B_QMIN + 15) /* 128 */ -# define S2B_CMIN (S2B_QMIN + 16) -#endif -/****************************************/ - S2B_64(S2B_CMIN + 0) /* 192 */ - S2B_64(S2B_CMIN + 1) /* 256 */ - S2B_64(S2B_CMIN + 2) /* 320 */ - S2B_64(S2B_CMIN + 3) /* 384 */ - S2B_64(S2B_CMIN + 4) /* 448 */ - S2B_64(S2B_CMIN + 5) /* 512 */ -# define S2B_SMIN (S2B_CMIN + 6) - S2B_256(S2B_SMIN + 0) /* 768 */ - S2B_256(S2B_SMIN + 1) /* 1024 */ - S2B_256(S2B_SMIN + 2) /* 1280 */ - S2B_256(S2B_SMIN + 3) /* 1536 */ - S2B_256(S2B_SMIN + 4) /* 1792 */ - S2B_256(S2B_SMIN + 5) /* 2048 */ - S2B_256(S2B_SMIN + 6) /* 2304 */ - S2B_256(S2B_SMIN + 7) /* 2560 */ - S2B_256(S2B_SMIN + 8) /* 2816 */ - S2B_256(S2B_SMIN + 9) /* 3072 */ - S2B_256(S2B_SMIN + 10) /* 3328 */ - S2B_256(S2B_SMIN + 11) /* 3584 */ - S2B_256(S2B_SMIN + 12) /* 3840 */ -#if (STATIC_PAGE_SHIFT == 13) - S2B_256(S2B_SMIN + 13) /* 4096 */ - S2B_256(S2B_SMIN + 14) /* 4352 */ - S2B_256(S2B_SMIN + 15) /* 4608 */ - S2B_256(S2B_SMIN + 16) /* 4864 */ - S2B_256(S2B_SMIN + 17) /* 5120 */ - S2B_256(S2B_SMIN + 18) /* 5376 */ - S2B_256(S2B_SMIN + 19) /* 5632 */ - S2B_256(S2B_SMIN + 20) /* 5888 */ - S2B_256(S2B_SMIN + 21) /* 6144 */ - S2B_256(S2B_SMIN + 22) /* 6400 */ - S2B_256(S2B_SMIN + 23) /* 6656 */ - S2B_256(S2B_SMIN + 24) /* 6912 */ - S2B_256(S2B_SMIN + 25) /* 7168 */ - S2B_256(S2B_SMIN + 26) /* 7424 */ - S2B_256(S2B_SMIN + 27) /* 7680 */ - S2B_256(S2B_SMIN + 28) /* 7936 */ -#endif -}; -#undef S2B_1 -#undef S2B_2 -#undef S2B_4 +#define S2B_512(i) S2B_256(i) S2B_256(i) +#define S2B_1024(i) S2B_512(i) S2B_512(i) +#define S2B_2048(i) S2B_1024(i) S2B_1024(i) +#define S2B_4096(i) S2B_2048(i) S2B_2048(i) +#define S2B_8192(i) S2B_4096(i) S2B_4096(i) +#define SIZE_CLASS(bin, delta, size) \ + S2B_##delta(bin) + SIZE_CLASSES #undef S2B_8 #undef S2B_16 #undef S2B_32 #undef S2B_64 #undef S2B_128 #undef S2B_256 -#undef S2B_QMIN -#undef S2B_CMIN -#undef S2B_SMIN +#undef S2B_512 +#undef S2B_1024 +#undef S2B_2048 +#undef S2B_4096 +#undef S2B_8192 +#undef SIZE_CLASS +}; /******************************************************************************/ /* Function prototypes for non-inline static functions. */ static void arena_run_split(arena_t *arena, arena_run_t *run, size_t size, - bool large, bool zero); + bool large, size_t binind, bool zero); static arena_chunk_t *arena_chunk_alloc(arena_t *arena); static void arena_chunk_dealloc(arena_t *arena, arena_chunk_t *chunk); +static arena_run_t *arena_run_alloc_helper(arena_t *arena, size_t size, + bool large, size_t binind, bool zero); static arena_run_t *arena_run_alloc(arena_t *arena, size_t size, bool large, - bool zero); + size_t binind, bool zero); static void arena_purge(arena_t *arena, bool all); static void arena_run_dalloc(arena_t *arena, arena_run_t *run, bool dirty); static void arena_run_trim_head(arena_t *arena, arena_chunk_t *chunk, arena_run_t *run, size_t oldsize, size_t newsize); static void arena_run_trim_tail(arena_t *arena, arena_chunk_t *chunk, arena_run_t *run, size_t oldsize, size_t newsize, bool dirty); +static arena_run_t *arena_bin_runs_first(arena_bin_t *bin); +static void arena_bin_runs_insert(arena_bin_t *bin, arena_run_t *run); +static void arena_bin_runs_remove(arena_bin_t *bin, arena_run_t *run); +static arena_run_t *arena_bin_nonfull_run_tryget(arena_bin_t *bin); static arena_run_t *arena_bin_nonfull_run_get(arena_t *arena, arena_bin_t *bin); static void *arena_bin_malloc_hard(arena_t *arena, arena_bin_t *bin); static void arena_dissociate_bin_run(arena_chunk_t *chunk, arena_run_t *run, @@ -187,14 +72,9 @@ static bool arena_ralloc_large_grow(arena_t *arena, arena_chunk_t *chunk, void *ptr, size_t oldsize, size_t size, size_t extra, bool zero); static bool arena_ralloc_large(void *ptr, size_t oldsize, size_t size, size_t extra, bool zero); -static bool small_size2bin_init(void); -#ifdef JEMALLOC_DEBUG -static void small_size2bin_validate(void); -#endif -static bool small_size2bin_init_hard(void); static size_t bin_info_run_size_calc(arena_bin_info_t *bin_info, size_t min_run_size); -static bool bin_info_init(void); +static void bin_info_init(void); /******************************************************************************/ @@ -211,8 +91,8 @@ arena_run_comp(arena_chunk_map_t *a, arena_chunk_map_t *b) } /* Generate red-black tree functions. */ -rb_gen(static JEMALLOC_ATTR(unused), arena_run_tree_, arena_run_tree_t, - arena_chunk_map_t, u.rb_link, arena_run_comp) +rb_gen(static UNUSED, arena_run_tree_, arena_run_tree_t, arena_chunk_map_t, + u.rb_link, arena_run_comp) static inline int arena_avail_comp(arena_chunk_map_t *a, arena_chunk_map_t *b) @@ -246,8 +126,8 @@ arena_avail_comp(arena_chunk_map_t *a, arena_chunk_map_t *b) } /* Generate red-black tree functions. */ -rb_gen(static JEMALLOC_ATTR(unused), arena_avail_tree_, arena_avail_tree_t, - arena_chunk_map_t, u.rb_link, arena_avail_comp) +rb_gen(static UNUSED, arena_avail_tree_, arena_avail_tree_t, arena_chunk_map_t, + u.rb_link, arena_avail_comp) static inline void * arena_run_reg_alloc(arena_run_t *run, arena_bin_info_t *bin_info) @@ -257,13 +137,12 @@ arena_run_reg_alloc(arena_run_t *run, arena_bin_info_t *bin_info) bitmap_t *bitmap = (bitmap_t *)((uintptr_t)run + (uintptr_t)bin_info->bitmap_offset); - dassert(run->magic == ARENA_RUN_MAGIC); assert(run->nfree > 0); assert(bitmap_full(bitmap, &bin_info->bitmap_info) == false); regind = bitmap_sfu(bitmap, &bin_info->bitmap_info); ret = (void *)((uintptr_t)run + (uintptr_t)bin_info->reg0_offset + - (uintptr_t)(bin_info->reg_size * regind)); + (uintptr_t)(bin_info->reg_interval * regind)); run->nfree--; if (regind == run->nextind) run->nextind++; @@ -275,7 +154,9 @@ static inline void arena_run_reg_dalloc(arena_run_t *run, void *ptr) { arena_chunk_t *chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(run); - size_t binind = arena_bin_index(chunk->arena, run->bin); + size_t pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE; + size_t mapbits = arena_mapbits_get(chunk, pageind); + size_t binind = arena_ptr_small_binind_get(ptr, mapbits); arena_bin_info_t *bin_info = &arena_bin_info[binind]; unsigned regind = arena_run_regind(run, bin_info, ptr); bitmap_t *bitmap = (bitmap_t *)((uintptr_t)run + @@ -284,8 +165,8 @@ arena_run_reg_dalloc(arena_run_t *run, void *ptr) assert(run->nfree < bin_info->nregs); /* Freeing an interior pointer can cause assertion failure. */ assert(((uintptr_t)ptr - ((uintptr_t)run + - (uintptr_t)bin_info->reg0_offset)) % (uintptr_t)bin_info->reg_size - == 0); + (uintptr_t)bin_info->reg0_offset)) % + (uintptr_t)bin_info->reg_interval == 0); assert((uintptr_t)ptr >= (uintptr_t)run + (uintptr_t)bin_info->reg0_offset); /* Freeing an unallocated pointer can cause assertion failure. */ @@ -295,75 +176,76 @@ arena_run_reg_dalloc(arena_run_t *run, void *ptr) run->nfree++; } -#ifdef JEMALLOC_DEBUG static inline void arena_chunk_validate_zeroed(arena_chunk_t *chunk, size_t run_ind) { size_t i; - size_t *p = (size_t *)((uintptr_t)chunk + (run_ind << PAGE_SHIFT)); + UNUSED size_t *p = (size_t *)((uintptr_t)chunk + (run_ind << LG_PAGE)); - for (i = 0; i < PAGE_SIZE / sizeof(size_t); i++) + for (i = 0; i < PAGE / sizeof(size_t); i++) assert(p[i] == 0); } -#endif static void arena_run_split(arena_t *arena, arena_run_t *run, size_t size, bool large, - bool zero) + size_t binind, bool zero) { arena_chunk_t *chunk; - size_t old_ndirty, run_ind, total_pages, need_pages, rem_pages, i; + size_t run_ind, total_pages, need_pages, rem_pages, i; size_t flag_dirty; arena_avail_tree_t *runs_avail; -#ifdef JEMALLOC_STATS - size_t cactive_diff; -#endif + + assert((large && binind == BININD_INVALID) || (large == false && binind + != BININD_INVALID)); chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(run); - old_ndirty = chunk->ndirty; - run_ind = (unsigned)(((uintptr_t)run - (uintptr_t)chunk) - >> PAGE_SHIFT); - flag_dirty = chunk->map[run_ind-map_bias].bits & CHUNK_MAP_DIRTY; + run_ind = (unsigned)(((uintptr_t)run - (uintptr_t)chunk) >> LG_PAGE); + flag_dirty = arena_mapbits_dirty_get(chunk, run_ind); runs_avail = (flag_dirty != 0) ? &arena->runs_avail_dirty : &arena->runs_avail_clean; - total_pages = (chunk->map[run_ind-map_bias].bits & ~PAGE_MASK) >> - PAGE_SHIFT; - assert((chunk->map[run_ind+total_pages-1-map_bias].bits & - CHUNK_MAP_DIRTY) == flag_dirty); - need_pages = (size >> PAGE_SHIFT); + total_pages = arena_mapbits_unallocated_size_get(chunk, run_ind) >> + LG_PAGE; + assert(arena_mapbits_dirty_get(chunk, run_ind+total_pages-1) == + flag_dirty); + need_pages = (size >> LG_PAGE); assert(need_pages > 0); assert(need_pages <= total_pages); rem_pages = total_pages - need_pages; - arena_avail_tree_remove(runs_avail, &chunk->map[run_ind-map_bias]); -#ifdef JEMALLOC_STATS - /* Update stats_cactive if nactive is crossing a chunk multiple. */ - cactive_diff = CHUNK_CEILING((arena->nactive + need_pages) << - PAGE_SHIFT) - CHUNK_CEILING(arena->nactive << PAGE_SHIFT); - if (cactive_diff != 0) - stats_cactive_add(cactive_diff); -#endif + arena_avail_tree_remove(runs_avail, arena_mapp_get(chunk, run_ind)); + if (config_stats) { + /* + * Update stats_cactive if nactive is crossing a chunk + * multiple. + */ + size_t cactive_diff = CHUNK_CEILING((arena->nactive + + need_pages) << LG_PAGE) - CHUNK_CEILING(arena->nactive << + LG_PAGE); + if (cactive_diff != 0) + stats_cactive_add(cactive_diff); + } arena->nactive += need_pages; /* Keep track of trailing unused pages for later use. */ if (rem_pages > 0) { if (flag_dirty != 0) { - chunk->map[run_ind+need_pages-map_bias].bits = - (rem_pages << PAGE_SHIFT) | CHUNK_MAP_DIRTY; - chunk->map[run_ind+total_pages-1-map_bias].bits = - (rem_pages << PAGE_SHIFT) | CHUNK_MAP_DIRTY; + arena_mapbits_unallocated_set(chunk, run_ind+need_pages, + (rem_pages << LG_PAGE), CHUNK_MAP_DIRTY); + arena_mapbits_unallocated_set(chunk, + run_ind+total_pages-1, (rem_pages << LG_PAGE), + CHUNK_MAP_DIRTY); } else { - chunk->map[run_ind+need_pages-map_bias].bits = - (rem_pages << PAGE_SHIFT) | - (chunk->map[run_ind+need_pages-map_bias].bits & - CHUNK_MAP_UNZEROED); - chunk->map[run_ind+total_pages-1-map_bias].bits = - (rem_pages << PAGE_SHIFT) | - (chunk->map[run_ind+total_pages-1-map_bias].bits & - CHUNK_MAP_UNZEROED); + arena_mapbits_unallocated_set(chunk, run_ind+need_pages, + (rem_pages << LG_PAGE), + arena_mapbits_unzeroed_get(chunk, + run_ind+need_pages)); + arena_mapbits_unallocated_set(chunk, + run_ind+total_pages-1, (rem_pages << LG_PAGE), + arena_mapbits_unzeroed_get(chunk, + run_ind+total_pages-1)); } - arena_avail_tree_insert(runs_avail, - &chunk->map[run_ind+need_pages-map_bias]); + arena_avail_tree_insert(runs_avail, arena_mapp_get(chunk, + run_ind+need_pages)); } /* Update dirty page accounting. */ @@ -384,28 +266,34 @@ arena_run_split(arena_t *arena, arena_run_t *run, size_t size, bool large, * zeroed (i.e. never before touched). */ for (i = 0; i < need_pages; i++) { - if ((chunk->map[run_ind+i-map_bias].bits - & CHUNK_MAP_UNZEROED) != 0) { + if (arena_mapbits_unzeroed_get(chunk, + run_ind+i) != 0) { + VALGRIND_MAKE_MEM_UNDEFINED( + (void *)((uintptr_t) + chunk + ((run_ind+i) << + LG_PAGE)), PAGE); memset((void *)((uintptr_t) chunk + ((run_ind+i) << - PAGE_SHIFT)), 0, - PAGE_SIZE); - } -#ifdef JEMALLOC_DEBUG - else { + LG_PAGE)), 0, PAGE); + } else if (config_debug) { + VALGRIND_MAKE_MEM_DEFINED( + (void *)((uintptr_t) + chunk + ((run_ind+i) << + LG_PAGE)), PAGE); arena_chunk_validate_zeroed( chunk, run_ind+i); } -#endif } } else { /* * The run is dirty, so all pages must be * zeroed. */ + VALGRIND_MAKE_MEM_UNDEFINED((void + *)((uintptr_t)chunk + (run_ind << + LG_PAGE)), (need_pages << LG_PAGE)); memset((void *)((uintptr_t)chunk + (run_ind << - PAGE_SHIFT)), 0, (need_pages << - PAGE_SHIFT)); + LG_PAGE)), 0, (need_pages << LG_PAGE)); } } @@ -413,10 +301,9 @@ arena_run_split(arena_t *arena, arena_run_t *run, size_t size, bool large, * Set the last element first, in case the run only contains one * page (i.e. both statements set the same element). */ - chunk->map[run_ind+need_pages-1-map_bias].bits = - CHUNK_MAP_LARGE | CHUNK_MAP_ALLOCATED | flag_dirty; - chunk->map[run_ind-map_bias].bits = size | flag_dirty | - CHUNK_MAP_LARGE | CHUNK_MAP_ALLOCATED; + arena_mapbits_large_set(chunk, run_ind+need_pages-1, 0, + flag_dirty); + arena_mapbits_large_set(chunk, run_ind, size, flag_dirty); } else { assert(zero == false); /* @@ -424,43 +311,29 @@ arena_run_split(arena_t *arena, arena_run_t *run, size_t size, bool large, * small run, so that arena_dalloc_bin_run() has the ability to * conditionally trim clean pages. */ - chunk->map[run_ind-map_bias].bits = - (chunk->map[run_ind-map_bias].bits & CHUNK_MAP_UNZEROED) | - CHUNK_MAP_ALLOCATED | flag_dirty; -#ifdef JEMALLOC_DEBUG + arena_mapbits_small_set(chunk, run_ind, 0, binind, flag_dirty); /* * The first page will always be dirtied during small run * initialization, so a validation failure here would not * actually cause an observable failure. */ - if (flag_dirty == 0 && - (chunk->map[run_ind-map_bias].bits & CHUNK_MAP_UNZEROED) - == 0) + if (config_debug && flag_dirty == 0 && + arena_mapbits_unzeroed_get(chunk, run_ind) == 0) arena_chunk_validate_zeroed(chunk, run_ind); -#endif for (i = 1; i < need_pages - 1; i++) { - chunk->map[run_ind+i-map_bias].bits = (i << PAGE_SHIFT) - | (chunk->map[run_ind+i-map_bias].bits & - CHUNK_MAP_UNZEROED) | CHUNK_MAP_ALLOCATED; -#ifdef JEMALLOC_DEBUG - if (flag_dirty == 0 && - (chunk->map[run_ind+i-map_bias].bits & - CHUNK_MAP_UNZEROED) == 0) + arena_mapbits_small_set(chunk, run_ind+i, i, binind, 0); + if (config_debug && flag_dirty == 0 && + arena_mapbits_unzeroed_get(chunk, run_ind+i) == 0) arena_chunk_validate_zeroed(chunk, run_ind+i); -#endif } - chunk->map[run_ind+need_pages-1-map_bias].bits = ((need_pages - - 1) << PAGE_SHIFT) | - (chunk->map[run_ind+need_pages-1-map_bias].bits & - CHUNK_MAP_UNZEROED) | CHUNK_MAP_ALLOCATED | flag_dirty; -#ifdef JEMALLOC_DEBUG - if (flag_dirty == 0 && - (chunk->map[run_ind+need_pages-1-map_bias].bits & - CHUNK_MAP_UNZEROED) == 0) { + arena_mapbits_small_set(chunk, run_ind+need_pages-1, + need_pages-1, binind, flag_dirty); + if (config_debug && flag_dirty == 0 && + arena_mapbits_unzeroed_get(chunk, run_ind+need_pages-1) == + 0) { arena_chunk_validate_zeroed(chunk, run_ind+need_pages-1); } -#endif } } @@ -476,31 +349,35 @@ arena_chunk_alloc(arena_t *arena) chunk = arena->spare; arena->spare = NULL; + assert(arena_mapbits_allocated_get(chunk, map_bias) == 0); + assert(arena_mapbits_allocated_get(chunk, chunk_npages-1) == 0); + assert(arena_mapbits_unallocated_size_get(chunk, map_bias) == + arena_maxclass); + assert(arena_mapbits_unallocated_size_get(chunk, + chunk_npages-1) == arena_maxclass); + assert(arena_mapbits_dirty_get(chunk, map_bias) == + arena_mapbits_dirty_get(chunk, chunk_npages-1)); + /* Insert the run into the appropriate runs_avail_* tree. */ - if ((chunk->map[0].bits & CHUNK_MAP_DIRTY) == 0) + if (arena_mapbits_dirty_get(chunk, map_bias) == 0) runs_avail = &arena->runs_avail_clean; else runs_avail = &arena->runs_avail_dirty; - assert((chunk->map[0].bits & ~PAGE_MASK) == arena_maxclass); - assert((chunk->map[chunk_npages-1-map_bias].bits & ~PAGE_MASK) - == arena_maxclass); - assert((chunk->map[0].bits & CHUNK_MAP_DIRTY) == - (chunk->map[chunk_npages-1-map_bias].bits & - CHUNK_MAP_DIRTY)); - arena_avail_tree_insert(runs_avail, &chunk->map[0]); + arena_avail_tree_insert(runs_avail, arena_mapp_get(chunk, + map_bias)); } else { bool zero; size_t unzeroed; zero = false; malloc_mutex_unlock(&arena->lock); - chunk = (arena_chunk_t *)chunk_alloc(chunksize, false, &zero); + chunk = (arena_chunk_t *)chunk_alloc(chunksize, chunksize, + false, &zero); malloc_mutex_lock(&arena->lock); if (chunk == NULL) return (NULL); -#ifdef JEMALLOC_STATS - arena->stats.mapped += chunksize; -#endif + if (config_stats) + arena->stats.mapped += chunksize; chunk->arena = arena; ql_elm_new(chunk, link_dirty); @@ -518,27 +395,27 @@ arena_chunk_alloc(arena_t *arena) * chunk. */ unzeroed = zero ? 0 : CHUNK_MAP_UNZEROED; - chunk->map[0].bits = arena_maxclass | unzeroed; + arena_mapbits_unallocated_set(chunk, map_bias, arena_maxclass, + unzeroed); /* * There is no need to initialize the internal page map entries * unless the chunk is not zeroed. */ if (zero == false) { for (i = map_bias+1; i < chunk_npages-1; i++) - chunk->map[i-map_bias].bits = unzeroed; + arena_mapbits_unzeroed_set(chunk, i, unzeroed); + } else if (config_debug) { + for (i = map_bias+1; i < chunk_npages-1; i++) { + assert(arena_mapbits_unzeroed_get(chunk, i) == + unzeroed); + } } -#ifdef JEMALLOC_DEBUG - else { - for (i = map_bias+1; i < chunk_npages-1; i++) - assert(chunk->map[i-map_bias].bits == unzeroed); - } -#endif - chunk->map[chunk_npages-1-map_bias].bits = arena_maxclass | - unzeroed; + arena_mapbits_unallocated_set(chunk, chunk_npages-1, + arena_maxclass, unzeroed); /* Insert the run into the runs_avail_clean tree. */ arena_avail_tree_insert(&arena->runs_avail_clean, - &chunk->map[0]); + arena_mapp_get(chunk, map_bias)); } return (chunk); @@ -549,15 +426,24 @@ arena_chunk_dealloc(arena_t *arena, arena_chunk_t *chunk) { arena_avail_tree_t *runs_avail; + assert(arena_mapbits_allocated_get(chunk, map_bias) == 0); + assert(arena_mapbits_allocated_get(chunk, chunk_npages-1) == 0); + assert(arena_mapbits_unallocated_size_get(chunk, map_bias) == + arena_maxclass); + assert(arena_mapbits_unallocated_size_get(chunk, chunk_npages-1) == + arena_maxclass); + assert(arena_mapbits_dirty_get(chunk, map_bias) == + arena_mapbits_dirty_get(chunk, chunk_npages-1)); + /* * Remove run from the appropriate runs_avail_* tree, so that the arena * does not use it. */ - if ((chunk->map[0].bits & CHUNK_MAP_DIRTY) == 0) + if (arena_mapbits_dirty_get(chunk, map_bias) == 0) runs_avail = &arena->runs_avail_clean; else runs_avail = &arena->runs_avail_dirty; - arena_avail_tree_remove(runs_avail, &chunk->map[0]); + arena_avail_tree_remove(runs_avail, arena_mapp_get(chunk, map_bias)); if (arena->spare != NULL) { arena_chunk_t *spare = arena->spare; @@ -571,24 +457,19 @@ arena_chunk_dealloc(arena_t *arena, arena_chunk_t *chunk) malloc_mutex_unlock(&arena->lock); chunk_dealloc((void *)spare, chunksize, true); malloc_mutex_lock(&arena->lock); -#ifdef JEMALLOC_STATS - arena->stats.mapped -= chunksize; -#endif + if (config_stats) + arena->stats.mapped -= chunksize; } else arena->spare = chunk; } static arena_run_t * -arena_run_alloc(arena_t *arena, size_t size, bool large, bool zero) +arena_run_alloc_helper(arena_t *arena, size_t size, bool large, size_t binind, + bool zero) { - arena_chunk_t *chunk; arena_run_t *run; arena_chunk_map_t *mapelm, key; - assert(size <= arena_maxclass); - assert((size & PAGE_MASK) == 0); - - /* Search the arena's chunks for the lowest best fit. */ key.bits = size | CHUNK_MAP_KEY; mapelm = arena_avail_tree_nsearch(&arena->runs_avail_dirty, &key); if (mapelm != NULL) { @@ -598,8 +479,8 @@ arena_run_alloc(arena_t *arena, size_t size, bool large, bool zero) + map_bias; run = (arena_run_t *)((uintptr_t)run_chunk + (pageind << - PAGE_SHIFT)); - arena_run_split(arena, run, size, large, zero); + LG_PAGE)); + arena_run_split(arena, run, size, large, binind, zero); return (run); } mapelm = arena_avail_tree_nsearch(&arena->runs_avail_clean, &key); @@ -610,19 +491,38 @@ arena_run_alloc(arena_t *arena, size_t size, bool large, bool zero) + map_bias; run = (arena_run_t *)((uintptr_t)run_chunk + (pageind << - PAGE_SHIFT)); - arena_run_split(arena, run, size, large, zero); + LG_PAGE)); + arena_run_split(arena, run, size, large, binind, zero); return (run); } + return (NULL); +} + +static arena_run_t * +arena_run_alloc(arena_t *arena, size_t size, bool large, size_t binind, + bool zero) +{ + arena_chunk_t *chunk; + arena_run_t *run; + + assert(size <= arena_maxclass); + assert((size & PAGE_MASK) == 0); + assert((large && binind == BININD_INVALID) || (large == false && binind + != BININD_INVALID)); + + /* Search the arena's chunks for the lowest best fit. */ + run = arena_run_alloc_helper(arena, size, large, binind, zero); + if (run != NULL) + return (run); + /* * No usable runs. Create a new chunk from which to allocate the run. */ chunk = arena_chunk_alloc(arena); if (chunk != NULL) { - run = (arena_run_t *)((uintptr_t)chunk + (map_bias << - PAGE_SHIFT)); - arena_run_split(arena, run, size, large, zero); + run = (arena_run_t *)((uintptr_t)chunk + (map_bias << LG_PAGE)); + arena_run_split(arena, run, size, large, binind, zero); return (run); } @@ -631,32 +531,7 @@ arena_run_alloc(arena_t *arena, size_t size, bool large, bool zero) * sufficient memory available while this one dropped arena->lock in * arena_chunk_alloc(), so search one more time. */ - mapelm = arena_avail_tree_nsearch(&arena->runs_avail_dirty, &key); - if (mapelm != NULL) { - arena_chunk_t *run_chunk = CHUNK_ADDR2BASE(mapelm); - size_t pageind = (((uintptr_t)mapelm - - (uintptr_t)run_chunk->map) / sizeof(arena_chunk_map_t)) - + map_bias; - - run = (arena_run_t *)((uintptr_t)run_chunk + (pageind << - PAGE_SHIFT)); - arena_run_split(arena, run, size, large, zero); - return (run); - } - mapelm = arena_avail_tree_nsearch(&arena->runs_avail_clean, &key); - if (mapelm != NULL) { - arena_chunk_t *run_chunk = CHUNK_ADDR2BASE(mapelm); - size_t pageind = (((uintptr_t)mapelm - - (uintptr_t)run_chunk->map) / sizeof(arena_chunk_map_t)) - + map_bias; - - run = (arena_run_t *)((uintptr_t)run_chunk + (pageind << - PAGE_SHIFT)); - arena_run_split(arena, run, size, large, zero); - return (run); - } - - return (NULL); + return (arena_run_alloc_helper(arena, size, large, binind, zero)); } static inline void @@ -677,12 +552,8 @@ arena_chunk_purge(arena_t *arena, arena_chunk_t *chunk) ql_head(arena_chunk_map_t) mapelms; arena_chunk_map_t *mapelm; size_t pageind, flag_unzeroed; -#ifdef JEMALLOC_DEBUG size_t ndirty; -#endif -#ifdef JEMALLOC_STATS size_t nmadvise; -#endif ql_new(&mapelms); @@ -692,13 +563,11 @@ arena_chunk_purge(arena_t *arena, arena_chunk_t *chunk) * madvise(..., MADV_DONTNEED) results in zero-filled pages for anonymous * mappings, but not for file-backed mappings. */ -# ifdef JEMALLOC_SWAP - swap_enabled ? CHUNK_MAP_UNZEROED : -# endif - 0; + 0 #else - CHUNK_MAP_UNZEROED; + CHUNK_MAP_UNZEROED #endif + ; /* * If chunk is the spare, temporarily re-allocate it, 1) so that its @@ -716,56 +585,61 @@ arena_chunk_purge(arena_t *arena, arena_chunk_t *chunk) * run. */ if (chunk == arena->spare) { - assert((chunk->map[0].bits & CHUNK_MAP_DIRTY) != 0); + assert(arena_mapbits_dirty_get(chunk, map_bias) != 0); + assert(arena_mapbits_dirty_get(chunk, chunk_npages-1) != 0); + arena_chunk_alloc(arena); } /* Temporarily allocate all free dirty runs within chunk. */ for (pageind = map_bias; pageind < chunk_npages;) { - mapelm = &chunk->map[pageind-map_bias]; - if ((mapelm->bits & CHUNK_MAP_ALLOCATED) == 0) { + mapelm = arena_mapp_get(chunk, pageind); + if (arena_mapbits_allocated_get(chunk, pageind) == 0) { size_t npages; - npages = mapelm->bits >> PAGE_SHIFT; + npages = arena_mapbits_unallocated_size_get(chunk, + pageind) >> LG_PAGE; assert(pageind + npages <= chunk_npages); - if (mapelm->bits & CHUNK_MAP_DIRTY) { + assert(arena_mapbits_dirty_get(chunk, pageind) == + arena_mapbits_dirty_get(chunk, pageind+npages-1)); + if (arena_mapbits_dirty_get(chunk, pageind) != 0) { size_t i; -#ifdef JEMALLOC_STATS - size_t cactive_diff; -#endif arena_avail_tree_remove( &arena->runs_avail_dirty, mapelm); - mapelm->bits = (npages << PAGE_SHIFT) | - flag_unzeroed | CHUNK_MAP_LARGE | - CHUNK_MAP_ALLOCATED; + arena_mapbits_unzeroed_set(chunk, pageind, + flag_unzeroed); + arena_mapbits_large_set(chunk, pageind, + (npages << LG_PAGE), 0); /* * Update internal elements in the page map, so * that CHUNK_MAP_UNZEROED is properly set. */ for (i = 1; i < npages - 1; i++) { - chunk->map[pageind+i-map_bias].bits = - flag_unzeroed; + arena_mapbits_unzeroed_set(chunk, + pageind+i, flag_unzeroed); } if (npages > 1) { - chunk->map[ - pageind+npages-1-map_bias].bits = - flag_unzeroed | CHUNK_MAP_LARGE | - CHUNK_MAP_ALLOCATED; + arena_mapbits_unzeroed_set(chunk, + pageind+npages-1, flag_unzeroed); + arena_mapbits_large_set(chunk, + pageind+npages-1, 0, 0); } -#ifdef JEMALLOC_STATS - /* - * Update stats_cactive if nactive is crossing a - * chunk multiple. - */ - cactive_diff = CHUNK_CEILING((arena->nactive + - npages) << PAGE_SHIFT) - - CHUNK_CEILING(arena->nactive << PAGE_SHIFT); - if (cactive_diff != 0) - stats_cactive_add(cactive_diff); -#endif + if (config_stats) { + /* + * Update stats_cactive if nactive is + * crossing a chunk multiple. + */ + size_t cactive_diff = + CHUNK_CEILING((arena->nactive + + npages) << LG_PAGE) - + CHUNK_CEILING(arena->nactive << + LG_PAGE); + if (cactive_diff != 0) + stats_cactive_add(cactive_diff); + } arena->nactive += npages; /* Append to list for later processing. */ ql_elm_new(mapelm, u.ql_link); @@ -775,71 +649,57 @@ arena_chunk_purge(arena_t *arena, arena_chunk_t *chunk) pageind += npages; } else { /* Skip allocated run. */ - if (mapelm->bits & CHUNK_MAP_LARGE) - pageind += mapelm->bits >> PAGE_SHIFT; + if (arena_mapbits_large_get(chunk, pageind)) + pageind += arena_mapbits_large_size_get(chunk, + pageind) >> LG_PAGE; else { + size_t binind; + arena_bin_info_t *bin_info; arena_run_t *run = (arena_run_t *)((uintptr_t) - chunk + (uintptr_t)(pageind << PAGE_SHIFT)); + chunk + (uintptr_t)(pageind << LG_PAGE)); - assert((mapelm->bits >> PAGE_SHIFT) == 0); - dassert(run->magic == ARENA_RUN_MAGIC); - size_t binind = arena_bin_index(arena, - run->bin); - arena_bin_info_t *bin_info = - &arena_bin_info[binind]; - pageind += bin_info->run_size >> PAGE_SHIFT; + assert(arena_mapbits_small_runind_get(chunk, + pageind) == 0); + binind = arena_bin_index(arena, run->bin); + bin_info = &arena_bin_info[binind]; + pageind += bin_info->run_size >> LG_PAGE; } } } assert(pageind == chunk_npages); -#ifdef JEMALLOC_DEBUG - ndirty = chunk->ndirty; -#endif -#ifdef JEMALLOC_STATS - arena->stats.purged += chunk->ndirty; -#endif + if (config_debug) + ndirty = chunk->ndirty; + if (config_stats) + arena->stats.purged += chunk->ndirty; arena->ndirty -= chunk->ndirty; chunk->ndirty = 0; ql_remove(&arena->chunks_dirty, chunk, link_dirty); chunk->dirtied = false; malloc_mutex_unlock(&arena->lock); -#ifdef JEMALLOC_STATS - nmadvise = 0; -#endif + if (config_stats) + nmadvise = 0; ql_foreach(mapelm, &mapelms, u.ql_link) { size_t pageind = (((uintptr_t)mapelm - (uintptr_t)chunk->map) / sizeof(arena_chunk_map_t)) + map_bias; - size_t npages = mapelm->bits >> PAGE_SHIFT; + size_t npages = arena_mapbits_large_size_get(chunk, pageind) >> + LG_PAGE; assert(pageind + npages <= chunk_npages); -#ifdef JEMALLOC_DEBUG assert(ndirty >= npages); - ndirty -= npages; -#endif + if (config_debug) + ndirty -= npages; -#ifdef JEMALLOC_PURGE_MADVISE_DONTNEED - madvise((void *)((uintptr_t)chunk + (pageind << PAGE_SHIFT)), - (npages << PAGE_SHIFT), MADV_DONTNEED); -#elif defined(JEMALLOC_PURGE_MADVISE_FREE) - madvise((void *)((uintptr_t)chunk + (pageind << PAGE_SHIFT)), - (npages << PAGE_SHIFT), MADV_FREE); -#else -# error "No method defined for purging unused dirty pages." -#endif - -#ifdef JEMALLOC_STATS - nmadvise++; -#endif + pages_purge((void *)((uintptr_t)chunk + (pageind << LG_PAGE)), + (npages << LG_PAGE)); + if (config_stats) + nmadvise++; } -#ifdef JEMALLOC_DEBUG assert(ndirty == 0); -#endif malloc_mutex_lock(&arena->lock); -#ifdef JEMALLOC_STATS - arena->stats.nmadvise += nmadvise; -#endif + if (config_stats) + arena->stats.nmadvise += nmadvise; /* Deallocate runs. */ for (mapelm = ql_first(&mapelms); mapelm != NULL; @@ -847,7 +707,7 @@ arena_chunk_purge(arena_t *arena, arena_chunk_t *chunk) size_t pageind = (((uintptr_t)mapelm - (uintptr_t)chunk->map) / sizeof(arena_chunk_map_t)) + map_bias; arena_run_t *run = (arena_run_t *)((uintptr_t)chunk + - (uintptr_t)(pageind << PAGE_SHIFT)); + (uintptr_t)(pageind << LG_PAGE)); ql_remove(&mapelms, mapelm, u.ql_link); arena_run_dalloc(arena, run, false); @@ -859,23 +719,22 @@ arena_purge(arena_t *arena, bool all) { arena_chunk_t *chunk; size_t npurgatory; -#ifdef JEMALLOC_DEBUG - size_t ndirty = 0; + if (config_debug) { + size_t ndirty = 0; - ql_foreach(chunk, &arena->chunks_dirty, link_dirty) { - assert(chunk->dirtied); - ndirty += chunk->ndirty; + ql_foreach(chunk, &arena->chunks_dirty, link_dirty) { + assert(chunk->dirtied); + ndirty += chunk->ndirty; + } + assert(ndirty == arena->ndirty); } - assert(ndirty == arena->ndirty); -#endif assert(arena->ndirty > arena->npurgatory || all); assert(arena->ndirty - arena->npurgatory > chunk_npages || all); assert((arena->nactive >> opt_lg_dirty_mult) < (arena->ndirty - arena->npurgatory) || all); -#ifdef JEMALLOC_STATS - arena->stats.npurge++; -#endif + if (config_stats) + arena->stats.npurge++; /* * Compute the minimum number of pages that this thread should try to @@ -957,44 +816,41 @@ arena_run_dalloc(arena_t *arena, arena_run_t *run, bool dirty) arena_chunk_t *chunk; size_t size, run_ind, run_pages, flag_dirty; arena_avail_tree_t *runs_avail; -#ifdef JEMALLOC_STATS - size_t cactive_diff; -#endif chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(run); - run_ind = (size_t)(((uintptr_t)run - (uintptr_t)chunk) - >> PAGE_SHIFT); + run_ind = (size_t)(((uintptr_t)run - (uintptr_t)chunk) >> LG_PAGE); assert(run_ind >= map_bias); assert(run_ind < chunk_npages); - if ((chunk->map[run_ind-map_bias].bits & CHUNK_MAP_LARGE) != 0) { - size = chunk->map[run_ind-map_bias].bits & ~PAGE_MASK; - assert(size == PAGE_SIZE || - (chunk->map[run_ind+(size>>PAGE_SHIFT)-1-map_bias].bits & - ~PAGE_MASK) == 0); - assert((chunk->map[run_ind+(size>>PAGE_SHIFT)-1-map_bias].bits & - CHUNK_MAP_LARGE) != 0); - assert((chunk->map[run_ind+(size>>PAGE_SHIFT)-1-map_bias].bits & - CHUNK_MAP_ALLOCATED) != 0); + if (arena_mapbits_large_get(chunk, run_ind) != 0) { + size = arena_mapbits_large_size_get(chunk, run_ind); + assert(size == PAGE || + arena_mapbits_large_size_get(chunk, + run_ind+(size>>LG_PAGE)-1) == 0); } else { size_t binind = arena_bin_index(arena, run->bin); arena_bin_info_t *bin_info = &arena_bin_info[binind]; size = bin_info->run_size; } - run_pages = (size >> PAGE_SHIFT); -#ifdef JEMALLOC_STATS - /* Update stats_cactive if nactive is crossing a chunk multiple. */ - cactive_diff = CHUNK_CEILING(arena->nactive << PAGE_SHIFT) - - CHUNK_CEILING((arena->nactive - run_pages) << PAGE_SHIFT); - if (cactive_diff != 0) - stats_cactive_sub(cactive_diff); -#endif + run_pages = (size >> LG_PAGE); + if (config_stats) { + /* + * Update stats_cactive if nactive is crossing a chunk + * multiple. + */ + size_t cactive_diff = CHUNK_CEILING(arena->nactive << LG_PAGE) - + CHUNK_CEILING((arena->nactive - run_pages) << LG_PAGE); + if (cactive_diff != 0) + stats_cactive_sub(cactive_diff); + } arena->nactive -= run_pages; /* * The run is dirty if the caller claims to have dirtied it, as well as * if it was already dirty before being allocated. */ - if ((chunk->map[run_ind-map_bias].bits & CHUNK_MAP_DIRTY) != 0) + assert(arena_mapbits_dirty_get(chunk, run_ind) == + arena_mapbits_dirty_get(chunk, run_ind+run_pages-1)); + if (arena_mapbits_dirty_get(chunk, run_ind) != 0) dirty = true; flag_dirty = dirty ? CHUNK_MAP_DIRTY : 0; runs_avail = dirty ? &arena->runs_avail_dirty : @@ -1002,59 +858,53 @@ arena_run_dalloc(arena_t *arena, arena_run_t *run, bool dirty) /* Mark pages as unallocated in the chunk map. */ if (dirty) { - chunk->map[run_ind-map_bias].bits = size | CHUNK_MAP_DIRTY; - chunk->map[run_ind+run_pages-1-map_bias].bits = size | - CHUNK_MAP_DIRTY; + arena_mapbits_unallocated_set(chunk, run_ind, size, + CHUNK_MAP_DIRTY); + arena_mapbits_unallocated_set(chunk, run_ind+run_pages-1, size, + CHUNK_MAP_DIRTY); chunk->ndirty += run_pages; arena->ndirty += run_pages; } else { - chunk->map[run_ind-map_bias].bits = size | - (chunk->map[run_ind-map_bias].bits & CHUNK_MAP_UNZEROED); - chunk->map[run_ind+run_pages-1-map_bias].bits = size | - (chunk->map[run_ind+run_pages-1-map_bias].bits & - CHUNK_MAP_UNZEROED); + arena_mapbits_unallocated_set(chunk, run_ind, size, + arena_mapbits_unzeroed_get(chunk, run_ind)); + arena_mapbits_unallocated_set(chunk, run_ind+run_pages-1, size, + arena_mapbits_unzeroed_get(chunk, run_ind+run_pages-1)); } /* Try to coalesce forward. */ if (run_ind + run_pages < chunk_npages && - (chunk->map[run_ind+run_pages-map_bias].bits & CHUNK_MAP_ALLOCATED) - == 0 && (chunk->map[run_ind+run_pages-map_bias].bits & - CHUNK_MAP_DIRTY) == flag_dirty) { - size_t nrun_size = chunk->map[run_ind+run_pages-map_bias].bits & - ~PAGE_MASK; - size_t nrun_pages = nrun_size >> PAGE_SHIFT; + arena_mapbits_allocated_get(chunk, run_ind+run_pages) == 0 && + arena_mapbits_dirty_get(chunk, run_ind+run_pages) == flag_dirty) { + size_t nrun_size = arena_mapbits_unallocated_size_get(chunk, + run_ind+run_pages); + size_t nrun_pages = nrun_size >> LG_PAGE; /* * Remove successor from runs_avail; the coalesced run is * inserted later. */ - assert((chunk->map[run_ind+run_pages+nrun_pages-1-map_bias].bits - & ~PAGE_MASK) == nrun_size); - assert((chunk->map[run_ind+run_pages+nrun_pages-1-map_bias].bits - & CHUNK_MAP_ALLOCATED) == 0); - assert((chunk->map[run_ind+run_pages+nrun_pages-1-map_bias].bits - & CHUNK_MAP_DIRTY) == flag_dirty); + assert(arena_mapbits_unallocated_size_get(chunk, + run_ind+run_pages+nrun_pages-1) == nrun_size); + assert(arena_mapbits_dirty_get(chunk, + run_ind+run_pages+nrun_pages-1) == flag_dirty); arena_avail_tree_remove(runs_avail, - &chunk->map[run_ind+run_pages-map_bias]); + arena_mapp_get(chunk, run_ind+run_pages)); size += nrun_size; run_pages += nrun_pages; - chunk->map[run_ind-map_bias].bits = size | - (chunk->map[run_ind-map_bias].bits & CHUNK_MAP_FLAGS_MASK); - chunk->map[run_ind+run_pages-1-map_bias].bits = size | - (chunk->map[run_ind+run_pages-1-map_bias].bits & - CHUNK_MAP_FLAGS_MASK); + arena_mapbits_unallocated_size_set(chunk, run_ind, size); + arena_mapbits_unallocated_size_set(chunk, run_ind+run_pages-1, + size); } /* Try to coalesce backward. */ - if (run_ind > map_bias && (chunk->map[run_ind-1-map_bias].bits & - CHUNK_MAP_ALLOCATED) == 0 && (chunk->map[run_ind-1-map_bias].bits & - CHUNK_MAP_DIRTY) == flag_dirty) { - size_t prun_size = chunk->map[run_ind-1-map_bias].bits & - ~PAGE_MASK; - size_t prun_pages = prun_size >> PAGE_SHIFT; + if (run_ind > map_bias && arena_mapbits_allocated_get(chunk, run_ind-1) + == 0 && arena_mapbits_dirty_get(chunk, run_ind-1) == flag_dirty) { + size_t prun_size = arena_mapbits_unallocated_size_get(chunk, + run_ind-1); + size_t prun_pages = prun_size >> LG_PAGE; run_ind -= prun_pages; @@ -1062,31 +912,26 @@ arena_run_dalloc(arena_t *arena, arena_run_t *run, bool dirty) * Remove predecessor from runs_avail; the coalesced run is * inserted later. */ - assert((chunk->map[run_ind-map_bias].bits & ~PAGE_MASK) - == prun_size); - assert((chunk->map[run_ind-map_bias].bits & CHUNK_MAP_ALLOCATED) - == 0); - assert((chunk->map[run_ind-map_bias].bits & CHUNK_MAP_DIRTY) - == flag_dirty); - arena_avail_tree_remove(runs_avail, - &chunk->map[run_ind-map_bias]); + assert(arena_mapbits_unallocated_size_get(chunk, run_ind) == + prun_size); + assert(arena_mapbits_dirty_get(chunk, run_ind) == flag_dirty); + arena_avail_tree_remove(runs_avail, arena_mapp_get(chunk, + run_ind)); size += prun_size; run_pages += prun_pages; - chunk->map[run_ind-map_bias].bits = size | - (chunk->map[run_ind-map_bias].bits & CHUNK_MAP_FLAGS_MASK); - chunk->map[run_ind+run_pages-1-map_bias].bits = size | - (chunk->map[run_ind+run_pages-1-map_bias].bits & - CHUNK_MAP_FLAGS_MASK); + arena_mapbits_unallocated_size_set(chunk, run_ind, size); + arena_mapbits_unallocated_size_set(chunk, run_ind+run_pages-1, + size); } /* Insert into runs_avail, now that coalescing is complete. */ - assert((chunk->map[run_ind-map_bias].bits & ~PAGE_MASK) == - (chunk->map[run_ind+run_pages-1-map_bias].bits & ~PAGE_MASK)); - assert((chunk->map[run_ind-map_bias].bits & CHUNK_MAP_DIRTY) == - (chunk->map[run_ind+run_pages-1-map_bias].bits & CHUNK_MAP_DIRTY)); - arena_avail_tree_insert(runs_avail, &chunk->map[run_ind-map_bias]); + assert(arena_mapbits_unallocated_size_get(chunk, run_ind) == + arena_mapbits_unallocated_size_get(chunk, run_ind+run_pages-1)); + assert(arena_mapbits_dirty_get(chunk, run_ind) == + arena_mapbits_dirty_get(chunk, run_ind+run_pages-1)); + arena_avail_tree_insert(runs_avail, arena_mapp_get(chunk, run_ind)); if (dirty) { /* @@ -1100,14 +945,12 @@ arena_run_dalloc(arena_t *arena, arena_run_t *run, bool dirty) } } - /* - * Deallocate chunk if it is now completely unused. The bit - * manipulation checks whether the first run is unallocated and extends - * to the end of the chunk. - */ - if ((chunk->map[0].bits & (~PAGE_MASK | CHUNK_MAP_ALLOCATED)) == - arena_maxclass) + /* Deallocate chunk if it is now completely unused. */ + if (size == arena_maxclass) { + assert(run_ind == map_bias); + assert(run_pages == (arena_maxclass >> LG_PAGE)); arena_chunk_dealloc(arena, chunk); + } /* * It is okay to do dirty page processing here even if the chunk was @@ -1124,9 +967,9 @@ static void arena_run_trim_head(arena_t *arena, arena_chunk_t *chunk, arena_run_t *run, size_t oldsize, size_t newsize) { - size_t pageind = ((uintptr_t)run - (uintptr_t)chunk) >> PAGE_SHIFT; - size_t head_npages = (oldsize - newsize) >> PAGE_SHIFT; - size_t flag_dirty = chunk->map[pageind-map_bias].bits & CHUNK_MAP_DIRTY; + size_t pageind = ((uintptr_t)run - (uintptr_t)chunk) >> LG_PAGE; + size_t head_npages = (oldsize - newsize) >> LG_PAGE; + size_t flag_dirty = arena_mapbits_dirty_get(chunk, pageind); assert(oldsize > newsize); @@ -1135,31 +978,19 @@ arena_run_trim_head(arena_t *arena, arena_chunk_t *chunk, arena_run_t *run, * leading run as separately allocated. Set the last element of each * run first, in case of single-page runs. */ - assert((chunk->map[pageind-map_bias].bits & CHUNK_MAP_LARGE) != 0); - assert((chunk->map[pageind-map_bias].bits & CHUNK_MAP_ALLOCATED) != 0); - chunk->map[pageind+head_npages-1-map_bias].bits = flag_dirty | - (chunk->map[pageind+head_npages-1-map_bias].bits & - CHUNK_MAP_UNZEROED) | CHUNK_MAP_LARGE | CHUNK_MAP_ALLOCATED; - chunk->map[pageind-map_bias].bits = (oldsize - newsize) - | flag_dirty | (chunk->map[pageind-map_bias].bits & - CHUNK_MAP_UNZEROED) | CHUNK_MAP_LARGE | CHUNK_MAP_ALLOCATED; + assert(arena_mapbits_large_size_get(chunk, pageind) == oldsize); + arena_mapbits_large_set(chunk, pageind+head_npages-1, 0, flag_dirty); + arena_mapbits_large_set(chunk, pageind, oldsize-newsize, flag_dirty); -#ifdef JEMALLOC_DEBUG - { - size_t tail_npages = newsize >> PAGE_SHIFT; - assert((chunk->map[pageind+head_npages+tail_npages-1-map_bias] - .bits & ~PAGE_MASK) == 0); - assert((chunk->map[pageind+head_npages+tail_npages-1-map_bias] - .bits & CHUNK_MAP_DIRTY) == flag_dirty); - assert((chunk->map[pageind+head_npages+tail_npages-1-map_bias] - .bits & CHUNK_MAP_LARGE) != 0); - assert((chunk->map[pageind+head_npages+tail_npages-1-map_bias] - .bits & CHUNK_MAP_ALLOCATED) != 0); + if (config_debug) { + UNUSED size_t tail_npages = newsize >> LG_PAGE; + assert(arena_mapbits_large_size_get(chunk, + pageind+head_npages+tail_npages-1) == 0); + assert(arena_mapbits_dirty_get(chunk, + pageind+head_npages+tail_npages-1) == flag_dirty); } -#endif - chunk->map[pageind+head_npages-map_bias].bits = newsize | flag_dirty | - (chunk->map[pageind+head_npages-map_bias].bits & - CHUNK_MAP_FLAGS_MASK) | CHUNK_MAP_LARGE | CHUNK_MAP_ALLOCATED; + arena_mapbits_large_set(chunk, pageind+head_npages, newsize, + flag_dirty); arena_run_dalloc(arena, run, false); } @@ -1168,11 +999,9 @@ static void arena_run_trim_tail(arena_t *arena, arena_chunk_t *chunk, arena_run_t *run, size_t oldsize, size_t newsize, bool dirty) { - size_t pageind = ((uintptr_t)run - (uintptr_t)chunk) >> PAGE_SHIFT; - size_t head_npages = newsize >> PAGE_SHIFT; - size_t tail_npages = (oldsize - newsize) >> PAGE_SHIFT; - size_t flag_dirty = chunk->map[pageind-map_bias].bits & - CHUNK_MAP_DIRTY; + size_t pageind = ((uintptr_t)run - (uintptr_t)chunk) >> LG_PAGE; + size_t head_npages = newsize >> LG_PAGE; + size_t flag_dirty = arena_mapbits_dirty_get(chunk, pageind); assert(oldsize > newsize); @@ -1181,61 +1010,92 @@ arena_run_trim_tail(arena_t *arena, arena_chunk_t *chunk, arena_run_t *run, * trailing run as separately allocated. Set the last element of each * run first, in case of single-page runs. */ - assert((chunk->map[pageind-map_bias].bits & CHUNK_MAP_LARGE) != 0); - assert((chunk->map[pageind-map_bias].bits & CHUNK_MAP_ALLOCATED) != 0); - chunk->map[pageind+head_npages-1-map_bias].bits = flag_dirty | - (chunk->map[pageind+head_npages-1-map_bias].bits & - CHUNK_MAP_UNZEROED) | CHUNK_MAP_LARGE | CHUNK_MAP_ALLOCATED; - chunk->map[pageind-map_bias].bits = newsize | flag_dirty | - (chunk->map[pageind-map_bias].bits & CHUNK_MAP_UNZEROED) | - CHUNK_MAP_LARGE | CHUNK_MAP_ALLOCATED; + assert(arena_mapbits_large_size_get(chunk, pageind) == oldsize); + arena_mapbits_large_set(chunk, pageind+head_npages-1, 0, flag_dirty); + arena_mapbits_large_set(chunk, pageind, newsize, flag_dirty); - assert((chunk->map[pageind+head_npages+tail_npages-1-map_bias].bits & - ~PAGE_MASK) == 0); - assert((chunk->map[pageind+head_npages+tail_npages-1-map_bias].bits & - CHUNK_MAP_LARGE) != 0); - assert((chunk->map[pageind+head_npages+tail_npages-1-map_bias].bits & - CHUNK_MAP_ALLOCATED) != 0); - chunk->map[pageind+head_npages+tail_npages-1-map_bias].bits = - flag_dirty | - (chunk->map[pageind+head_npages+tail_npages-1-map_bias].bits & - CHUNK_MAP_UNZEROED) | CHUNK_MAP_LARGE | CHUNK_MAP_ALLOCATED; - chunk->map[pageind+head_npages-map_bias].bits = (oldsize - newsize) | - flag_dirty | (chunk->map[pageind+head_npages-map_bias].bits & - CHUNK_MAP_UNZEROED) | CHUNK_MAP_LARGE | CHUNK_MAP_ALLOCATED; + if (config_debug) { + UNUSED size_t tail_npages = (oldsize - newsize) >> LG_PAGE; + assert(arena_mapbits_large_size_get(chunk, + pageind+head_npages+tail_npages-1) == 0); + assert(arena_mapbits_dirty_get(chunk, + pageind+head_npages+tail_npages-1) == flag_dirty); + } + arena_mapbits_large_set(chunk, pageind+head_npages, oldsize-newsize, + flag_dirty); arena_run_dalloc(arena, (arena_run_t *)((uintptr_t)run + newsize), dirty); } static arena_run_t * -arena_bin_nonfull_run_get(arena_t *arena, arena_bin_t *bin) +arena_bin_runs_first(arena_bin_t *bin) { - arena_chunk_map_t *mapelm; - arena_run_t *run; - size_t binind; - arena_bin_info_t *bin_info; - - /* Look for a usable run. */ - mapelm = arena_run_tree_first(&bin->runs); + arena_chunk_map_t *mapelm = arena_run_tree_first(&bin->runs); if (mapelm != NULL) { arena_chunk_t *chunk; size_t pageind; - - /* run is guaranteed to have available space. */ - arena_run_tree_remove(&bin->runs, mapelm); + arena_run_t *run; chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(mapelm); pageind = ((((uintptr_t)mapelm - (uintptr_t)chunk->map) / sizeof(arena_chunk_map_t))) + map_bias; run = (arena_run_t *)((uintptr_t)chunk + (uintptr_t)((pageind - - (mapelm->bits >> PAGE_SHIFT)) - << PAGE_SHIFT)); -#ifdef JEMALLOC_STATS - bin->stats.reruns++; -#endif + arena_mapbits_small_runind_get(chunk, pageind)) << + LG_PAGE)); return (run); } + + return (NULL); +} + +static void +arena_bin_runs_insert(arena_bin_t *bin, arena_run_t *run) +{ + arena_chunk_t *chunk = CHUNK_ADDR2BASE(run); + size_t pageind = ((uintptr_t)run - (uintptr_t)chunk) >> LG_PAGE; + arena_chunk_map_t *mapelm = arena_mapp_get(chunk, pageind); + + assert(arena_run_tree_search(&bin->runs, mapelm) == NULL); + + arena_run_tree_insert(&bin->runs, mapelm); +} + +static void +arena_bin_runs_remove(arena_bin_t *bin, arena_run_t *run) +{ + arena_chunk_t *chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(run); + size_t pageind = ((uintptr_t)run - (uintptr_t)chunk) >> LG_PAGE; + arena_chunk_map_t *mapelm = arena_mapp_get(chunk, pageind); + + assert(arena_run_tree_search(&bin->runs, mapelm) != NULL); + + arena_run_tree_remove(&bin->runs, mapelm); +} + +static arena_run_t * +arena_bin_nonfull_run_tryget(arena_bin_t *bin) +{ + arena_run_t *run = arena_bin_runs_first(bin); + if (run != NULL) { + arena_bin_runs_remove(bin, run); + if (config_stats) + bin->stats.reruns++; + } + return (run); +} + +static arena_run_t * +arena_bin_nonfull_run_get(arena_t *arena, arena_bin_t *bin) +{ + arena_run_t *run; + size_t binind; + arena_bin_info_t *bin_info; + + /* Look for a usable run. */ + run = arena_bin_nonfull_run_tryget(bin); + if (run != NULL) + return (run); /* No existing runs have any space available. */ binind = arena_bin_index(arena, bin); @@ -1245,30 +1105,27 @@ arena_bin_nonfull_run_get(arena_t *arena, arena_bin_t *bin) malloc_mutex_unlock(&bin->lock); /******************************/ malloc_mutex_lock(&arena->lock); - run = arena_run_alloc(arena, bin_info->run_size, false, false); + run = arena_run_alloc(arena, bin_info->run_size, false, binind, false); if (run != NULL) { bitmap_t *bitmap = (bitmap_t *)((uintptr_t)run + (uintptr_t)bin_info->bitmap_offset); /* Initialize run internals. */ + VALGRIND_MAKE_MEM_UNDEFINED(run, bin_info->reg0_offset - + bin_info->redzone_size); run->bin = bin; run->nextind = 0; run->nfree = bin_info->nregs; bitmap_init(bitmap, &bin_info->bitmap_info); -#ifdef JEMALLOC_DEBUG - run->magic = ARENA_RUN_MAGIC; -#endif } malloc_mutex_unlock(&arena->lock); /********************************/ malloc_mutex_lock(&bin->lock); if (run != NULL) { -#ifdef JEMALLOC_STATS - bin->stats.nruns++; - bin->stats.curruns++; - if (bin->stats.curruns > bin->stats.highruns) - bin->stats.highruns = bin->stats.curruns; -#endif + if (config_stats) { + bin->stats.nruns++; + bin->stats.curruns++; + } return (run); } @@ -1277,25 +1134,9 @@ arena_bin_nonfull_run_get(arena_t *arena, arena_bin_t *bin) * sufficient memory available while this one dropped bin->lock above, * so search one more time. */ - mapelm = arena_run_tree_first(&bin->runs); - if (mapelm != NULL) { - arena_chunk_t *chunk; - size_t pageind; - - /* run is guaranteed to have available space. */ - arena_run_tree_remove(&bin->runs, mapelm); - - chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(mapelm); - pageind = ((((uintptr_t)mapelm - (uintptr_t)chunk->map) / - sizeof(arena_chunk_map_t))) + map_bias; - run = (arena_run_t *)((uintptr_t)chunk + (uintptr_t)((pageind - - (mapelm->bits >> PAGE_SHIFT)) - << PAGE_SHIFT)); -#ifdef JEMALLOC_STATS - bin->stats.reruns++; -#endif + run = arena_bin_nonfull_run_tryget(bin); + if (run != NULL) return (run); - } return (NULL); } @@ -1318,7 +1159,6 @@ arena_bin_malloc_hard(arena_t *arena, arena_bin_t *bin) * Another thread updated runcur while this one ran without the * bin lock in arena_bin_nonfull_run_get(). */ - dassert(bin->runcur->magic == ARENA_RUN_MAGIC); assert(bin->runcur->nfree > 0); ret = arena_run_reg_alloc(bin->runcur, bin_info); if (run != NULL) { @@ -1346,18 +1186,18 @@ arena_bin_malloc_hard(arena_t *arena, arena_bin_t *bin) bin->runcur = run; - dassert(bin->runcur->magic == ARENA_RUN_MAGIC); assert(bin->runcur->nfree > 0); return (arena_run_reg_alloc(bin->runcur, bin_info)); } -#ifdef JEMALLOC_PROF void arena_prof_accum(arena_t *arena, uint64_t accumbytes) { - if (prof_interval != 0) { + cassert(config_prof); + + if (config_prof && prof_interval != 0) { arena->prof_accumbytes += accumbytes; if (arena->prof_accumbytes >= prof_interval) { prof_idump(); @@ -1365,15 +1205,10 @@ arena_prof_accum(arena_t *arena, uint64_t accumbytes) } } } -#endif -#ifdef JEMALLOC_TCACHE void -arena_tcache_fill_small(arena_t *arena, tcache_bin_t *tbin, size_t binind -# ifdef JEMALLOC_PROF - , uint64_t prof_accumbytes -# endif - ) +arena_tcache_fill_small(arena_t *arena, tcache_bin_t *tbin, size_t binind, + uint64_t prof_accumbytes) { unsigned i, nfill; arena_bin_t *bin; @@ -1382,11 +1217,11 @@ arena_tcache_fill_small(arena_t *arena, tcache_bin_t *tbin, size_t binind assert(tbin->ncached == 0); -#ifdef JEMALLOC_PROF - malloc_mutex_lock(&arena->lock); - arena_prof_accum(arena, prof_accumbytes); - malloc_mutex_unlock(&arena->lock); -#endif + if (config_prof) { + malloc_mutex_lock(&arena->lock); + arena_prof_accum(arena, prof_accumbytes); + malloc_mutex_unlock(&arena->lock); + } bin = &arena->bins[binind]; malloc_mutex_lock(&bin->lock); for (i = 0, nfill = (tcache_bin_info[binind].ncached_max >> @@ -1397,20 +1232,72 @@ arena_tcache_fill_small(arena_t *arena, tcache_bin_t *tbin, size_t binind ptr = arena_bin_malloc_hard(arena, bin); if (ptr == NULL) break; + if (config_fill && opt_junk) { + arena_alloc_junk_small(ptr, &arena_bin_info[binind], + true); + } /* Insert such that low regions get used first. */ tbin->avail[nfill - 1 - i] = ptr; } -#ifdef JEMALLOC_STATS - bin->stats.allocated += i * arena_bin_info[binind].reg_size; - bin->stats.nmalloc += i; - bin->stats.nrequests += tbin->tstats.nrequests; - bin->stats.nfills++; - tbin->tstats.nrequests = 0; -#endif + if (config_stats) { + bin->stats.allocated += i * arena_bin_info[binind].reg_size; + bin->stats.nmalloc += i; + bin->stats.nrequests += tbin->tstats.nrequests; + bin->stats.nfills++; + tbin->tstats.nrequests = 0; + } malloc_mutex_unlock(&bin->lock); tbin->ncached = i; } -#endif + +void +arena_alloc_junk_small(void *ptr, arena_bin_info_t *bin_info, bool zero) +{ + + if (zero) { + size_t redzone_size = bin_info->redzone_size; + memset((void *)((uintptr_t)ptr - redzone_size), 0xa5, + redzone_size); + memset((void *)((uintptr_t)ptr + bin_info->reg_size), 0xa5, + redzone_size); + } else { + memset((void *)((uintptr_t)ptr - bin_info->redzone_size), 0xa5, + bin_info->reg_interval); + } +} + +void +arena_dalloc_junk_small(void *ptr, arena_bin_info_t *bin_info) +{ + size_t size = bin_info->reg_size; + size_t redzone_size = bin_info->redzone_size; + size_t i; + bool error = false; + + for (i = 1; i <= redzone_size; i++) { + unsigned byte; + if ((byte = *(uint8_t *)((uintptr_t)ptr - i)) != 0xa5) { + error = true; + malloc_printf(": Corrupt redzone " + "%zu byte%s before %p (size %zu), byte=%#x\n", i, + (i == 1) ? "" : "s", ptr, size, byte); + } + } + for (i = 0; i < redzone_size; i++) { + unsigned byte; + if ((byte = *(uint8_t *)((uintptr_t)ptr + size + i)) != 0xa5) { + error = true; + malloc_printf(": Corrupt redzone " + "%zu byte%s after end of %p (size %zu), byte=%#x\n", + i, (i == 1) ? "" : "s", ptr, size, byte); + } + } + if (opt_abort && error) + abort(); + + memset((void *)((uintptr_t)ptr - redzone_size), 0x5a, + bin_info->reg_interval); +} void * arena_malloc_small(arena_t *arena, size_t size, bool zero) @@ -1421,7 +1308,7 @@ arena_malloc_small(arena_t *arena, size_t size, bool zero) size_t binind; binind = SMALL_SIZE2BIN(size); - assert(binind < nbins); + assert(binind < NBINS); bin = &arena->bins[binind]; size = arena_bin_info[binind].reg_size; @@ -1436,29 +1323,34 @@ arena_malloc_small(arena_t *arena, size_t size, bool zero) return (NULL); } -#ifdef JEMALLOC_STATS - bin->stats.allocated += size; - bin->stats.nmalloc++; - bin->stats.nrequests++; -#endif + if (config_stats) { + bin->stats.allocated += size; + bin->stats.nmalloc++; + bin->stats.nrequests++; + } malloc_mutex_unlock(&bin->lock); -#ifdef JEMALLOC_PROF - if (isthreaded == false) { + if (config_prof && isthreaded == false) { malloc_mutex_lock(&arena->lock); arena_prof_accum(arena, size); malloc_mutex_unlock(&arena->lock); } -#endif if (zero == false) { -#ifdef JEMALLOC_FILL - if (opt_junk) - memset(ret, 0xa5, size); - else if (opt_zero) - memset(ret, 0, size); -#endif - } else + if (config_fill) { + if (opt_junk) { + arena_alloc_junk_small(ret, + &arena_bin_info[binind], false); + } else if (opt_zero) + memset(ret, 0, size); + } + } else { + if (config_fill && opt_junk) { + arena_alloc_junk_small(ret, &arena_bin_info[binind], + true); + } + VALGRIND_MAKE_MEM_UNDEFINED(ret, size); memset(ret, 0, size); + } return (ret); } @@ -1471,244 +1363,113 @@ arena_malloc_large(arena_t *arena, size_t size, bool zero) /* Large allocation. */ size = PAGE_CEILING(size); malloc_mutex_lock(&arena->lock); - ret = (void *)arena_run_alloc(arena, size, true, zero); + ret = (void *)arena_run_alloc(arena, size, true, BININD_INVALID, zero); if (ret == NULL) { malloc_mutex_unlock(&arena->lock); return (NULL); } -#ifdef JEMALLOC_STATS - arena->stats.nmalloc_large++; - arena->stats.nrequests_large++; - arena->stats.allocated_large += size; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].nmalloc++; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].nrequests++; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].curruns++; - if (arena->stats.lstats[(size >> PAGE_SHIFT) - 1].curruns > - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].highruns) { - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].highruns = - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].curruns; + if (config_stats) { + arena->stats.nmalloc_large++; + arena->stats.nrequests_large++; + arena->stats.allocated_large += size; + arena->stats.lstats[(size >> LG_PAGE) - 1].nmalloc++; + arena->stats.lstats[(size >> LG_PAGE) - 1].nrequests++; + arena->stats.lstats[(size >> LG_PAGE) - 1].curruns++; } -#endif -#ifdef JEMALLOC_PROF - arena_prof_accum(arena, size); -#endif + if (config_prof) + arena_prof_accum(arena, size); malloc_mutex_unlock(&arena->lock); if (zero == false) { -#ifdef JEMALLOC_FILL - if (opt_junk) - memset(ret, 0xa5, size); - else if (opt_zero) - memset(ret, 0, size); -#endif + if (config_fill) { + if (opt_junk) + memset(ret, 0xa5, size); + else if (opt_zero) + memset(ret, 0, size); + } } return (ret); } -void * -arena_malloc(size_t size, bool zero) -{ - - assert(size != 0); - assert(QUANTUM_CEILING(size) <= arena_maxclass); - - if (size <= small_maxclass) { -#ifdef JEMALLOC_TCACHE - tcache_t *tcache; - - if ((tcache = tcache_get()) != NULL) - return (tcache_alloc_small(tcache, size, zero)); - else - -#endif - return (arena_malloc_small(choose_arena(), size, zero)); - } else { -#ifdef JEMALLOC_TCACHE - if (size <= tcache_maxclass) { - tcache_t *tcache; - - if ((tcache = tcache_get()) != NULL) - return (tcache_alloc_large(tcache, size, zero)); - else { - return (arena_malloc_large(choose_arena(), - size, zero)); - } - } else -#endif - return (arena_malloc_large(choose_arena(), size, zero)); - } -} - /* Only handles large allocations that require more than page alignment. */ void * -arena_palloc(arena_t *arena, size_t size, size_t alloc_size, size_t alignment, - bool zero) +arena_palloc(arena_t *arena, size_t size, size_t alignment, bool zero) { void *ret; - size_t offset; + size_t alloc_size, leadsize, trailsize; + arena_run_t *run; arena_chunk_t *chunk; assert((size & PAGE_MASK) == 0); alignment = PAGE_CEILING(alignment); + alloc_size = size + alignment - PAGE; malloc_mutex_lock(&arena->lock); - ret = (void *)arena_run_alloc(arena, alloc_size, true, zero); - if (ret == NULL) { + run = arena_run_alloc(arena, alloc_size, true, BININD_INVALID, zero); + if (run == NULL) { malloc_mutex_unlock(&arena->lock); return (NULL); } + chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(run); - chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ret); - - offset = (uintptr_t)ret & (alignment - 1); - assert((offset & PAGE_MASK) == 0); - assert(offset < alloc_size); - if (offset == 0) - arena_run_trim_tail(arena, chunk, ret, alloc_size, size, false); - else { - size_t leadsize, trailsize; - - leadsize = alignment - offset; - if (leadsize > 0) { - arena_run_trim_head(arena, chunk, ret, alloc_size, - alloc_size - leadsize); - ret = (void *)((uintptr_t)ret + leadsize); - } - - trailsize = alloc_size - leadsize - size; - if (trailsize != 0) { - /* Trim trailing space. */ - assert(trailsize < alloc_size); - arena_run_trim_tail(arena, chunk, ret, size + trailsize, - size, false); - } + leadsize = ALIGNMENT_CEILING((uintptr_t)run, alignment) - + (uintptr_t)run; + assert(alloc_size >= leadsize + size); + trailsize = alloc_size - leadsize - size; + ret = (void *)((uintptr_t)run + leadsize); + if (leadsize != 0) { + arena_run_trim_head(arena, chunk, run, alloc_size, alloc_size - + leadsize); + } + if (trailsize != 0) { + arena_run_trim_tail(arena, chunk, ret, size + trailsize, size, + false); } -#ifdef JEMALLOC_STATS - arena->stats.nmalloc_large++; - arena->stats.nrequests_large++; - arena->stats.allocated_large += size; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].nmalloc++; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].nrequests++; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].curruns++; - if (arena->stats.lstats[(size >> PAGE_SHIFT) - 1].curruns > - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].highruns) { - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].highruns = - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].curruns; + if (config_stats) { + arena->stats.nmalloc_large++; + arena->stats.nrequests_large++; + arena->stats.allocated_large += size; + arena->stats.lstats[(size >> LG_PAGE) - 1].nmalloc++; + arena->stats.lstats[(size >> LG_PAGE) - 1].nrequests++; + arena->stats.lstats[(size >> LG_PAGE) - 1].curruns++; } -#endif malloc_mutex_unlock(&arena->lock); -#ifdef JEMALLOC_FILL - if (zero == false) { + if (config_fill && zero == false) { if (opt_junk) memset(ret, 0xa5, size); else if (opt_zero) memset(ret, 0, size); } -#endif return (ret); } -/* Return the size of the allocation pointed to by ptr. */ -size_t -arena_salloc(const void *ptr) -{ - size_t ret; - arena_chunk_t *chunk; - size_t pageind, mapbits; - - assert(ptr != NULL); - assert(CHUNK_ADDR2BASE(ptr) != ptr); - - chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr); - pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> PAGE_SHIFT; - mapbits = chunk->map[pageind-map_bias].bits; - assert((mapbits & CHUNK_MAP_ALLOCATED) != 0); - if ((mapbits & CHUNK_MAP_LARGE) == 0) { - arena_run_t *run = (arena_run_t *)((uintptr_t)chunk + - (uintptr_t)((pageind - (mapbits >> PAGE_SHIFT)) << - PAGE_SHIFT)); - dassert(run->magic == ARENA_RUN_MAGIC); - size_t binind = arena_bin_index(chunk->arena, run->bin); - arena_bin_info_t *bin_info = &arena_bin_info[binind]; - assert(((uintptr_t)ptr - ((uintptr_t)run + - (uintptr_t)bin_info->reg0_offset)) % bin_info->reg_size == - 0); - ret = bin_info->reg_size; - } else { - assert(((uintptr_t)ptr & PAGE_MASK) == 0); - ret = mapbits & ~PAGE_MASK; - assert(ret != 0); - } - - return (ret); -} - -#ifdef JEMALLOC_PROF void arena_prof_promoted(const void *ptr, size_t size) { arena_chunk_t *chunk; size_t pageind, binind; + cassert(config_prof); assert(ptr != NULL); assert(CHUNK_ADDR2BASE(ptr) != ptr); - assert(isalloc(ptr) == PAGE_SIZE); - assert(size <= small_maxclass); + assert(isalloc(ptr, false) == PAGE); + assert(isalloc(ptr, true) == PAGE); + assert(size <= SMALL_MAXCLASS); chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr); - pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> PAGE_SHIFT; + pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE; binind = SMALL_SIZE2BIN(size); - assert(binind < nbins); - chunk->map[pageind-map_bias].bits = (chunk->map[pageind-map_bias].bits & - ~CHUNK_MAP_CLASS_MASK) | ((binind+1) << CHUNK_MAP_CLASS_SHIFT); + assert(binind < NBINS); + arena_mapbits_large_binind_set(chunk, pageind, binind); + + assert(isalloc(ptr, false) == PAGE); + assert(isalloc(ptr, true) == size); } -size_t -arena_salloc_demote(const void *ptr) -{ - size_t ret; - arena_chunk_t *chunk; - size_t pageind, mapbits; - - assert(ptr != NULL); - assert(CHUNK_ADDR2BASE(ptr) != ptr); - - chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr); - pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> PAGE_SHIFT; - mapbits = chunk->map[pageind-map_bias].bits; - assert((mapbits & CHUNK_MAP_ALLOCATED) != 0); - if ((mapbits & CHUNK_MAP_LARGE) == 0) { - arena_run_t *run = (arena_run_t *)((uintptr_t)chunk + - (uintptr_t)((pageind - (mapbits >> PAGE_SHIFT)) << - PAGE_SHIFT)); - dassert(run->magic == ARENA_RUN_MAGIC); - size_t binind = arena_bin_index(chunk->arena, run->bin); - arena_bin_info_t *bin_info = &arena_bin_info[binind]; - assert(((uintptr_t)ptr - ((uintptr_t)run + - (uintptr_t)bin_info->reg0_offset)) % bin_info->reg_size == - 0); - ret = bin_info->reg_size; - } else { - assert(((uintptr_t)ptr & PAGE_MASK) == 0); - ret = mapbits & ~PAGE_MASK; - if (prof_promote && ret == PAGE_SIZE && (mapbits & - CHUNK_MAP_CLASS_MASK) != 0) { - size_t binind = ((mapbits & CHUNK_MAP_CLASS_MASK) >> - CHUNK_MAP_CLASS_SHIFT) - 1; - assert(binind < nbins); - ret = arena_bin_info[binind].reg_size; - } - assert(ret != 0); - } - - return (ret); -} -#endif - static void arena_dissociate_bin_run(arena_chunk_t *chunk, arena_run_t *run, arena_bin_t *bin) @@ -1722,16 +1483,12 @@ arena_dissociate_bin_run(arena_chunk_t *chunk, arena_run_t *run, arena_bin_info_t *bin_info = &arena_bin_info[binind]; if (bin_info->nregs != 1) { - size_t run_pageind = (((uintptr_t)run - - (uintptr_t)chunk)) >> PAGE_SHIFT; - arena_chunk_map_t *run_mapelm = - &chunk->map[run_pageind-map_bias]; /* * This block's conditional is necessary because if the * run only contains one region, then it never gets * inserted into the non-full runs tree. */ - arena_run_tree_remove(&bin->runs, run_mapelm); + arena_bin_runs_remove(bin, run); } } } @@ -1745,19 +1502,21 @@ arena_dalloc_bin_run(arena_t *arena, arena_chunk_t *chunk, arena_run_t *run, size_t npages, run_ind, past; assert(run != bin->runcur); - assert(arena_run_tree_search(&bin->runs, &chunk->map[ - (((uintptr_t)run-(uintptr_t)chunk)>>PAGE_SHIFT)-map_bias]) == NULL); + assert(arena_run_tree_search(&bin->runs, + arena_mapp_get(chunk, ((uintptr_t)run-(uintptr_t)chunk)>>LG_PAGE)) + == NULL); binind = arena_bin_index(chunk->arena, run->bin); bin_info = &arena_bin_info[binind]; malloc_mutex_unlock(&bin->lock); /******************************/ - npages = bin_info->run_size >> PAGE_SHIFT; - run_ind = (size_t)(((uintptr_t)run - (uintptr_t)chunk) >> PAGE_SHIFT); + npages = bin_info->run_size >> LG_PAGE; + run_ind = (size_t)(((uintptr_t)run - (uintptr_t)chunk) >> LG_PAGE); past = (size_t)(PAGE_CEILING((uintptr_t)run + (uintptr_t)bin_info->reg0_offset + (uintptr_t)(run->nextind * - bin_info->reg_size) - (uintptr_t)chunk) >> PAGE_SHIFT); + bin_info->reg_interval - bin_info->redzone_size) - + (uintptr_t)chunk) >> LG_PAGE); malloc_mutex_lock(&arena->lock); /* @@ -1765,32 +1524,24 @@ arena_dalloc_bin_run(arena_t *arena, arena_chunk_t *chunk, arena_run_t *run, * trim the clean pages before deallocating the dirty portion of the * run. */ - if ((chunk->map[run_ind-map_bias].bits & CHUNK_MAP_DIRTY) == 0 && past - - run_ind < npages) { - /* - * Trim clean pages. Convert to large run beforehand. Set the - * last map element first, in case this is a one-page run. - */ - chunk->map[run_ind+npages-1-map_bias].bits = CHUNK_MAP_LARGE | - (chunk->map[run_ind+npages-1-map_bias].bits & - CHUNK_MAP_FLAGS_MASK); - chunk->map[run_ind-map_bias].bits = bin_info->run_size | - CHUNK_MAP_LARGE | (chunk->map[run_ind-map_bias].bits & - CHUNK_MAP_FLAGS_MASK); - arena_run_trim_tail(arena, chunk, run, (npages << PAGE_SHIFT), - ((past - run_ind) << PAGE_SHIFT), false); + assert(arena_mapbits_dirty_get(chunk, run_ind) == + arena_mapbits_dirty_get(chunk, run_ind+npages-1)); + if (arena_mapbits_dirty_get(chunk, run_ind) == 0 && past - run_ind < + npages) { + /* Trim clean pages. Convert to large run beforehand. */ + assert(npages > 0); + arena_mapbits_large_set(chunk, run_ind, bin_info->run_size, 0); + arena_mapbits_large_set(chunk, run_ind+npages-1, 0, 0); + arena_run_trim_tail(arena, chunk, run, (npages << LG_PAGE), + ((past - run_ind) << LG_PAGE), false); /* npages = past - run_ind; */ } -#ifdef JEMALLOC_DEBUG - run->magic = 0; -#endif arena_run_dalloc(arena, run, true); malloc_mutex_unlock(&arena->lock); /****************************/ malloc_mutex_lock(&bin->lock); -#ifdef JEMALLOC_STATS - bin->stats.curruns--; -#endif + if (config_stats) + bin->stats.curruns--; } static void @@ -1799,62 +1550,42 @@ arena_bin_lower_run(arena_t *arena, arena_chunk_t *chunk, arena_run_t *run, { /* - * Make sure that bin->runcur always refers to the lowest non-full run, - * if one exists. + * Make sure that if bin->runcur is non-NULL, it refers to the lowest + * non-full run. It is okay to NULL runcur out rather than proactively + * keeping it pointing at the lowest non-full run. */ - if (bin->runcur == NULL) - bin->runcur = run; - else if ((uintptr_t)run < (uintptr_t)bin->runcur) { + if ((uintptr_t)run < (uintptr_t)bin->runcur) { /* Switch runcur. */ - if (bin->runcur->nfree > 0) { - arena_chunk_t *runcur_chunk = - CHUNK_ADDR2BASE(bin->runcur); - size_t runcur_pageind = (((uintptr_t)bin->runcur - - (uintptr_t)runcur_chunk)) >> PAGE_SHIFT; - arena_chunk_map_t *runcur_mapelm = - &runcur_chunk->map[runcur_pageind-map_bias]; - - /* Insert runcur. */ - arena_run_tree_insert(&bin->runs, runcur_mapelm); - } + if (bin->runcur->nfree > 0) + arena_bin_runs_insert(bin, bin->runcur); bin->runcur = run; - } else { - size_t run_pageind = (((uintptr_t)run - - (uintptr_t)chunk)) >> PAGE_SHIFT; - arena_chunk_map_t *run_mapelm = - &chunk->map[run_pageind-map_bias]; - - assert(arena_run_tree_search(&bin->runs, run_mapelm) == NULL); - arena_run_tree_insert(&bin->runs, run_mapelm); - } + if (config_stats) + bin->stats.reruns++; + } else + arena_bin_runs_insert(bin, run); } void -arena_dalloc_bin(arena_t *arena, arena_chunk_t *chunk, void *ptr, +arena_dalloc_bin_locked(arena_t *arena, arena_chunk_t *chunk, void *ptr, arena_chunk_map_t *mapelm) { size_t pageind; arena_run_t *run; arena_bin_t *bin; -#if (defined(JEMALLOC_FILL) || defined(JEMALLOC_STATS)) - size_t size; -#endif + arena_bin_info_t *bin_info; + size_t size, binind; - pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> PAGE_SHIFT; + pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE; run = (arena_run_t *)((uintptr_t)chunk + (uintptr_t)((pageind - - (mapelm->bits >> PAGE_SHIFT)) << PAGE_SHIFT)); - dassert(run->magic == ARENA_RUN_MAGIC); + arena_mapbits_small_runind_get(chunk, pageind)) << LG_PAGE)); bin = run->bin; - size_t binind = arena_bin_index(arena, bin); - arena_bin_info_t *bin_info = &arena_bin_info[binind]; -#if (defined(JEMALLOC_FILL) || defined(JEMALLOC_STATS)) - size = bin_info->reg_size; -#endif + binind = arena_ptr_small_binind_get(ptr, mapelm->bits); + bin_info = &arena_bin_info[binind]; + if (config_fill || config_stats) + size = bin_info->reg_size; -#ifdef JEMALLOC_FILL - if (opt_junk) - memset(ptr, 0x5a, size); -#endif + if (config_fill && opt_junk) + arena_dalloc_junk_small(ptr, bin_info); arena_run_reg_dalloc(run, ptr); if (run->nfree == bin_info->nregs) { @@ -1863,13 +1594,41 @@ arena_dalloc_bin(arena_t *arena, arena_chunk_t *chunk, void *ptr, } else if (run->nfree == 1 && run != bin->runcur) arena_bin_lower_run(arena, chunk, run, bin); -#ifdef JEMALLOC_STATS - bin->stats.allocated -= size; - bin->stats.ndalloc++; -#endif + if (config_stats) { + bin->stats.allocated -= size; + bin->stats.ndalloc++; + } } -#ifdef JEMALLOC_STATS +void +arena_dalloc_bin(arena_t *arena, arena_chunk_t *chunk, void *ptr, + size_t pageind, arena_chunk_map_t *mapelm) +{ + arena_run_t *run; + arena_bin_t *bin; + + run = (arena_run_t *)((uintptr_t)chunk + (uintptr_t)((pageind - + arena_mapbits_small_runind_get(chunk, pageind)) << LG_PAGE)); + bin = run->bin; + malloc_mutex_lock(&bin->lock); + arena_dalloc_bin_locked(arena, chunk, ptr, mapelm); + malloc_mutex_unlock(&bin->lock); +} + +void +arena_dalloc_small(arena_t *arena, arena_chunk_t *chunk, void *ptr, + size_t pageind) +{ + arena_chunk_map_t *mapelm; + + if (config_debug) { + /* arena_ptr_small_binind_get() does extra sanity checking. */ + assert(arena_ptr_small_binind_get(ptr, arena_mapbits_get(chunk, + pageind)) != BININD_INVALID); + } + mapelm = arena_mapp_get(chunk, pageind); + arena_dalloc_bin(arena, chunk, ptr, pageind, mapelm); +} void arena_stats_merge(arena_t *arena, size_t *nactive, size_t *ndirty, arena_stats_t *astats, malloc_bin_stats_t *bstats, @@ -1894,12 +1653,11 @@ arena_stats_merge(arena_t *arena, size_t *nactive, size_t *ndirty, lstats[i].nmalloc += arena->stats.lstats[i].nmalloc; lstats[i].ndalloc += arena->stats.lstats[i].ndalloc; lstats[i].nrequests += arena->stats.lstats[i].nrequests; - lstats[i].highruns += arena->stats.lstats[i].highruns; lstats[i].curruns += arena->stats.lstats[i].curruns; } malloc_mutex_unlock(&arena->lock); - for (i = 0; i < nbins; i++) { + for (i = 0; i < NBINS; i++) { arena_bin_t *bin = &arena->bins[i]; malloc_mutex_lock(&bin->lock); @@ -1907,51 +1665,45 @@ arena_stats_merge(arena_t *arena, size_t *nactive, size_t *ndirty, bstats[i].nmalloc += bin->stats.nmalloc; bstats[i].ndalloc += bin->stats.ndalloc; bstats[i].nrequests += bin->stats.nrequests; -#ifdef JEMALLOC_TCACHE - bstats[i].nfills += bin->stats.nfills; - bstats[i].nflushes += bin->stats.nflushes; -#endif + if (config_tcache) { + bstats[i].nfills += bin->stats.nfills; + bstats[i].nflushes += bin->stats.nflushes; + } bstats[i].nruns += bin->stats.nruns; bstats[i].reruns += bin->stats.reruns; - bstats[i].highruns += bin->stats.highruns; bstats[i].curruns += bin->stats.curruns; malloc_mutex_unlock(&bin->lock); } } -#endif + +void +arena_dalloc_large_locked(arena_t *arena, arena_chunk_t *chunk, void *ptr) +{ + + if (config_fill || config_stats) { + size_t pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE; + size_t size = arena_mapbits_large_size_get(chunk, pageind); + + if (config_fill && config_stats && opt_junk) + memset(ptr, 0x5a, size); + if (config_stats) { + arena->stats.ndalloc_large++; + arena->stats.allocated_large -= size; + arena->stats.lstats[(size >> LG_PAGE) - 1].ndalloc++; + arena->stats.lstats[(size >> LG_PAGE) - 1].curruns--; + } + } + + arena_run_dalloc(arena, (arena_run_t *)ptr, true); +} void arena_dalloc_large(arena_t *arena, arena_chunk_t *chunk, void *ptr) { - /* Large allocation. */ -#ifdef JEMALLOC_FILL -# ifndef JEMALLOC_STATS - if (opt_junk) -# endif -#endif - { -#if (defined(JEMALLOC_FILL) || defined(JEMALLOC_STATS)) - size_t pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> - PAGE_SHIFT; - size_t size = chunk->map[pageind-map_bias].bits & ~PAGE_MASK; -#endif - -#ifdef JEMALLOC_FILL -# ifdef JEMALLOC_STATS - if (opt_junk) -# endif - memset(ptr, 0x5a, size); -#endif -#ifdef JEMALLOC_STATS - arena->stats.ndalloc_large++; - arena->stats.allocated_large -= size; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].ndalloc++; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].curruns--; -#endif - } - - arena_run_dalloc(arena, (arena_run_t *)ptr, true); + malloc_mutex_lock(&arena->lock); + arena_dalloc_large_locked(arena, chunk, ptr); + malloc_mutex_unlock(&arena->lock); } static void @@ -1968,24 +1720,19 @@ arena_ralloc_large_shrink(arena_t *arena, arena_chunk_t *chunk, void *ptr, malloc_mutex_lock(&arena->lock); arena_run_trim_tail(arena, chunk, (arena_run_t *)ptr, oldsize, size, true); -#ifdef JEMALLOC_STATS - arena->stats.ndalloc_large++; - arena->stats.allocated_large -= oldsize; - arena->stats.lstats[(oldsize >> PAGE_SHIFT) - 1].ndalloc++; - arena->stats.lstats[(oldsize >> PAGE_SHIFT) - 1].curruns--; + if (config_stats) { + arena->stats.ndalloc_large++; + arena->stats.allocated_large -= oldsize; + arena->stats.lstats[(oldsize >> LG_PAGE) - 1].ndalloc++; + arena->stats.lstats[(oldsize >> LG_PAGE) - 1].curruns--; - arena->stats.nmalloc_large++; - arena->stats.nrequests_large++; - arena->stats.allocated_large += size; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].nmalloc++; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].nrequests++; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].curruns++; - if (arena->stats.lstats[(size >> PAGE_SHIFT) - 1].curruns > - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].highruns) { - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].highruns = - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].curruns; + arena->stats.nmalloc_large++; + arena->stats.nrequests_large++; + arena->stats.allocated_large += size; + arena->stats.lstats[(size >> LG_PAGE) - 1].nmalloc++; + arena->stats.lstats[(size >> LG_PAGE) - 1].nrequests++; + arena->stats.lstats[(size >> LG_PAGE) - 1].curruns++; } -#endif malloc_mutex_unlock(&arena->lock); } @@ -1993,20 +1740,19 @@ static bool arena_ralloc_large_grow(arena_t *arena, arena_chunk_t *chunk, void *ptr, size_t oldsize, size_t size, size_t extra, bool zero) { - size_t pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> PAGE_SHIFT; - size_t npages = oldsize >> PAGE_SHIFT; + size_t pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE; + size_t npages = oldsize >> LG_PAGE; size_t followsize; - assert(oldsize == (chunk->map[pageind-map_bias].bits & ~PAGE_MASK)); + assert(oldsize == arena_mapbits_large_size_get(chunk, pageind)); /* Try to extend the run. */ assert(size + extra > oldsize); malloc_mutex_lock(&arena->lock); if (pageind + npages < chunk_npages && - (chunk->map[pageind+npages-map_bias].bits - & CHUNK_MAP_ALLOCATED) == 0 && (followsize = - chunk->map[pageind+npages-map_bias].bits & ~PAGE_MASK) >= size - - oldsize) { + arena_mapbits_allocated_get(chunk, pageind+npages) == 0 && + (followsize = arena_mapbits_unallocated_size_get(chunk, + pageind+npages)) >= size - oldsize) { /* * The next run is available and sufficiently large. Split the * following run, then merge the first part with the existing @@ -2016,10 +1762,11 @@ arena_ralloc_large_grow(arena_t *arena, arena_chunk_t *chunk, void *ptr, size_t splitsize = (oldsize + followsize <= size + extra) ? followsize : size + extra - oldsize; arena_run_split(arena, (arena_run_t *)((uintptr_t)chunk + - ((pageind+npages) << PAGE_SHIFT)), splitsize, true, zero); + ((pageind+npages) << LG_PAGE)), splitsize, true, + BININD_INVALID, zero); size = oldsize + splitsize; - npages = size >> PAGE_SHIFT; + npages = size >> LG_PAGE; /* * Mark the extended run as dirty if either portion of the run @@ -2029,34 +1776,24 @@ arena_ralloc_large_grow(arena_t *arena, arena_chunk_t *chunk, void *ptr, * arena_run_dalloc() with the dirty argument set to false * (which is when dirty flag consistency would really matter). */ - flag_dirty = (chunk->map[pageind-map_bias].bits & - CHUNK_MAP_DIRTY) | - (chunk->map[pageind+npages-1-map_bias].bits & - CHUNK_MAP_DIRTY); - chunk->map[pageind-map_bias].bits = size | flag_dirty - | CHUNK_MAP_LARGE | CHUNK_MAP_ALLOCATED; - chunk->map[pageind+npages-1-map_bias].bits = flag_dirty | - CHUNK_MAP_LARGE | CHUNK_MAP_ALLOCATED; + flag_dirty = arena_mapbits_dirty_get(chunk, pageind) | + arena_mapbits_dirty_get(chunk, pageind+npages-1); + arena_mapbits_large_set(chunk, pageind, size, flag_dirty); + arena_mapbits_large_set(chunk, pageind+npages-1, 0, flag_dirty); -#ifdef JEMALLOC_STATS - arena->stats.ndalloc_large++; - arena->stats.allocated_large -= oldsize; - arena->stats.lstats[(oldsize >> PAGE_SHIFT) - 1].ndalloc++; - arena->stats.lstats[(oldsize >> PAGE_SHIFT) - 1].curruns--; + if (config_stats) { + arena->stats.ndalloc_large++; + arena->stats.allocated_large -= oldsize; + arena->stats.lstats[(oldsize >> LG_PAGE) - 1].ndalloc++; + arena->stats.lstats[(oldsize >> LG_PAGE) - 1].curruns--; - arena->stats.nmalloc_large++; - arena->stats.nrequests_large++; - arena->stats.allocated_large += size; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].nmalloc++; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].nrequests++; - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].curruns++; - if (arena->stats.lstats[(size >> PAGE_SHIFT) - 1].curruns > - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].highruns) { - arena->stats.lstats[(size >> PAGE_SHIFT) - 1].highruns = - arena->stats.lstats[(size >> PAGE_SHIFT) - - 1].curruns; + arena->stats.nmalloc_large++; + arena->stats.nrequests_large++; + arena->stats.allocated_large += size; + arena->stats.lstats[(size >> LG_PAGE) - 1].nmalloc++; + arena->stats.lstats[(size >> LG_PAGE) - 1].nrequests++; + arena->stats.lstats[(size >> LG_PAGE) - 1].curruns++; } -#endif malloc_mutex_unlock(&arena->lock); return (false); } @@ -2078,12 +1815,10 @@ arena_ralloc_large(void *ptr, size_t oldsize, size_t size, size_t extra, psize = PAGE_CEILING(size + extra); if (psize == oldsize) { /* Same size class. */ -#ifdef JEMALLOC_FILL - if (opt_junk && size < oldsize) { + if (config_fill && opt_junk && size < oldsize) { memset((void *)((uintptr_t)ptr + size), 0x5a, oldsize - size); } -#endif return (false); } else { arena_chunk_t *chunk; @@ -2091,16 +1826,13 @@ arena_ralloc_large(void *ptr, size_t oldsize, size_t size, size_t extra, chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr); arena = chunk->arena; - dassert(arena->magic == ARENA_MAGIC); if (psize < oldsize) { -#ifdef JEMALLOC_FILL /* Fill before shrinking in order avoid a race. */ - if (opt_junk) { + if (config_fill && opt_junk) { memset((void *)((uintptr_t)ptr + size), 0x5a, oldsize - size); } -#endif arena_ralloc_large_shrink(arena, chunk, ptr, oldsize, psize); return (false); @@ -2108,12 +1840,11 @@ arena_ralloc_large(void *ptr, size_t oldsize, size_t size, size_t extra, bool ret = arena_ralloc_large_grow(arena, chunk, ptr, oldsize, PAGE_CEILING(size), psize - PAGE_CEILING(size), zero); -#ifdef JEMALLOC_FILL - if (ret == false && zero == false && opt_zero) { + if (config_fill && ret == false && zero == false && + opt_zero) { memset((void *)((uintptr_t)ptr + oldsize), 0, size - oldsize); } -#endif return (ret); } } @@ -2128,24 +1859,22 @@ arena_ralloc_no_move(void *ptr, size_t oldsize, size_t size, size_t extra, * Avoid moving the allocation if the size class can be left the same. */ if (oldsize <= arena_maxclass) { - if (oldsize <= small_maxclass) { + if (oldsize <= SMALL_MAXCLASS) { assert(arena_bin_info[SMALL_SIZE2BIN(oldsize)].reg_size == oldsize); - if ((size + extra <= small_maxclass && + if ((size + extra <= SMALL_MAXCLASS && SMALL_SIZE2BIN(size + extra) == SMALL_SIZE2BIN(oldsize)) || (size <= oldsize && size + extra >= oldsize)) { -#ifdef JEMALLOC_FILL - if (opt_junk && size < oldsize) { + if (config_fill && opt_junk && size < oldsize) { memset((void *)((uintptr_t)ptr + size), 0x5a, oldsize - size); } -#endif return (ptr); } } else { assert(size <= arena_maxclass); - if (size + extra > small_maxclass) { + if (size + extra > SMALL_MAXCLASS) { if (arena_ralloc_large(ptr, oldsize, size, extra, zero) == false) return (ptr); @@ -2159,7 +1888,7 @@ arena_ralloc_no_move(void *ptr, size_t oldsize, size_t size, size_t extra, void * arena_ralloc(void *ptr, size_t oldsize, size_t size, size_t extra, - size_t alignment, bool zero) + size_t alignment, bool zero, bool try_tcache) { void *ret; size_t copysize; @@ -2175,24 +1904,24 @@ arena_ralloc(void *ptr, size_t oldsize, size_t size, size_t extra, * copying. */ if (alignment != 0) { - size_t usize = sa2u(size + extra, alignment, NULL); + size_t usize = sa2u(size + extra, alignment); if (usize == 0) return (NULL); ret = ipalloc(usize, alignment, zero); } else - ret = arena_malloc(size + extra, zero); + ret = arena_malloc(NULL, size + extra, zero, try_tcache); if (ret == NULL) { if (extra == 0) return (NULL); /* Try again, this time without extra. */ if (alignment != 0) { - size_t usize = sa2u(size, alignment, NULL); + size_t usize = sa2u(size, alignment); if (usize == 0) return (NULL); ret = ipalloc(usize, alignment, zero); } else - ret = arena_malloc(size, zero); + ret = arena_malloc(NULL, size, zero, try_tcache); if (ret == NULL) return (NULL); @@ -2205,8 +1934,9 @@ arena_ralloc(void *ptr, size_t oldsize, size_t size, size_t extra, * expectation that the extra bytes will be reliably preserved. */ copysize = (size < oldsize) ? size : oldsize; + VALGRIND_MAKE_MEM_UNDEFINED(ret, copysize); memcpy(ret, ptr, copysize); - idalloc(ptr); + iqalloc(ptr); return (ret); } @@ -2222,22 +1952,21 @@ arena_new(arena_t *arena, unsigned ind) if (malloc_mutex_init(&arena->lock)) return (true); -#ifdef JEMALLOC_STATS - memset(&arena->stats, 0, sizeof(arena_stats_t)); - arena->stats.lstats = (malloc_large_stats_t *)base_alloc(nlclasses * - sizeof(malloc_large_stats_t)); - if (arena->stats.lstats == NULL) - return (true); - memset(arena->stats.lstats, 0, nlclasses * - sizeof(malloc_large_stats_t)); -# ifdef JEMALLOC_TCACHE - ql_new(&arena->tcache_ql); -# endif -#endif + if (config_stats) { + memset(&arena->stats, 0, sizeof(arena_stats_t)); + arena->stats.lstats = + (malloc_large_stats_t *)base_alloc(nlclasses * + sizeof(malloc_large_stats_t)); + if (arena->stats.lstats == NULL) + return (true); + memset(arena->stats.lstats, 0, nlclasses * + sizeof(malloc_large_stats_t)); + if (config_tcache) + ql_new(&arena->tcache_ql); + } -#ifdef JEMALLOC_PROF - arena->prof_accumbytes = 0; -#endif + if (config_prof) + arena->prof_accumbytes = 0; /* Initialize chunks. */ ql_new(&arena->chunks_dirty); @@ -2251,185 +1980,19 @@ arena_new(arena_t *arena, unsigned ind) arena_avail_tree_new(&arena->runs_avail_dirty); /* Initialize bins. */ - i = 0; -#ifdef JEMALLOC_TINY - /* (2^n)-spaced tiny bins. */ - for (; i < ntbins; i++) { + for (i = 0; i < NBINS; i++) { bin = &arena->bins[i]; if (malloc_mutex_init(&bin->lock)) return (true); bin->runcur = NULL; arena_run_tree_new(&bin->runs); -#ifdef JEMALLOC_STATS - memset(&bin->stats, 0, sizeof(malloc_bin_stats_t)); -#endif + if (config_stats) + memset(&bin->stats, 0, sizeof(malloc_bin_stats_t)); } -#endif - - /* Quantum-spaced bins. */ - for (; i < ntbins + nqbins; i++) { - bin = &arena->bins[i]; - if (malloc_mutex_init(&bin->lock)) - return (true); - bin->runcur = NULL; - arena_run_tree_new(&bin->runs); -#ifdef JEMALLOC_STATS - memset(&bin->stats, 0, sizeof(malloc_bin_stats_t)); -#endif - } - - /* Cacheline-spaced bins. */ - for (; i < ntbins + nqbins + ncbins; i++) { - bin = &arena->bins[i]; - if (malloc_mutex_init(&bin->lock)) - return (true); - bin->runcur = NULL; - arena_run_tree_new(&bin->runs); -#ifdef JEMALLOC_STATS - memset(&bin->stats, 0, sizeof(malloc_bin_stats_t)); -#endif - } - - /* Subpage-spaced bins. */ - for (; i < nbins; i++) { - bin = &arena->bins[i]; - if (malloc_mutex_init(&bin->lock)) - return (true); - bin->runcur = NULL; - arena_run_tree_new(&bin->runs); -#ifdef JEMALLOC_STATS - memset(&bin->stats, 0, sizeof(malloc_bin_stats_t)); -#endif - } - -#ifdef JEMALLOC_DEBUG - arena->magic = ARENA_MAGIC; -#endif return (false); } -#ifdef JEMALLOC_DEBUG -static void -small_size2bin_validate(void) -{ - size_t i, size, binind; - - i = 1; -# ifdef JEMALLOC_TINY - /* Tiny. */ - for (; i < (1U << LG_TINY_MIN); i++) { - size = pow2_ceil(1U << LG_TINY_MIN); - binind = ffs((int)(size >> (LG_TINY_MIN + 1))); - assert(SMALL_SIZE2BIN(i) == binind); - } - for (; i < qspace_min; i++) { - size = pow2_ceil(i); - binind = ffs((int)(size >> (LG_TINY_MIN + 1))); - assert(SMALL_SIZE2BIN(i) == binind); - } -# endif - /* Quantum-spaced. */ - for (; i <= qspace_max; i++) { - size = QUANTUM_CEILING(i); - binind = ntbins + (size >> LG_QUANTUM) - 1; - assert(SMALL_SIZE2BIN(i) == binind); - } - /* Cacheline-spaced. */ - for (; i <= cspace_max; i++) { - size = CACHELINE_CEILING(i); - binind = ntbins + nqbins + ((size - cspace_min) >> - LG_CACHELINE); - assert(SMALL_SIZE2BIN(i) == binind); - } - /* Sub-page. */ - for (; i <= sspace_max; i++) { - size = SUBPAGE_CEILING(i); - binind = ntbins + nqbins + ncbins + ((size - sspace_min) - >> LG_SUBPAGE); - assert(SMALL_SIZE2BIN(i) == binind); - } -} -#endif - -static bool -small_size2bin_init(void) -{ - - if (opt_lg_qspace_max != LG_QSPACE_MAX_DEFAULT - || opt_lg_cspace_max != LG_CSPACE_MAX_DEFAULT - || (sizeof(const_small_size2bin) != ((small_maxclass-1) >> - LG_TINY_MIN) + 1)) - return (small_size2bin_init_hard()); - - small_size2bin = const_small_size2bin; -#ifdef JEMALLOC_DEBUG - small_size2bin_validate(); -#endif - return (false); -} - -static bool -small_size2bin_init_hard(void) -{ - size_t i, size, binind; - uint8_t *custom_small_size2bin; -#define CUSTOM_SMALL_SIZE2BIN(s) \ - custom_small_size2bin[(s-1) >> LG_TINY_MIN] - - assert(opt_lg_qspace_max != LG_QSPACE_MAX_DEFAULT - || opt_lg_cspace_max != LG_CSPACE_MAX_DEFAULT - || (sizeof(const_small_size2bin) != ((small_maxclass-1) >> - LG_TINY_MIN) + 1)); - - custom_small_size2bin = (uint8_t *) - base_alloc(small_maxclass >> LG_TINY_MIN); - if (custom_small_size2bin == NULL) - return (true); - - i = 1; -#ifdef JEMALLOC_TINY - /* Tiny. */ - for (; i < (1U << LG_TINY_MIN); i += TINY_MIN) { - size = pow2_ceil(1U << LG_TINY_MIN); - binind = ffs((int)(size >> (LG_TINY_MIN + 1))); - CUSTOM_SMALL_SIZE2BIN(i) = binind; - } - for (; i < qspace_min; i += TINY_MIN) { - size = pow2_ceil(i); - binind = ffs((int)(size >> (LG_TINY_MIN + 1))); - CUSTOM_SMALL_SIZE2BIN(i) = binind; - } -#endif - /* Quantum-spaced. */ - for (; i <= qspace_max; i += TINY_MIN) { - size = QUANTUM_CEILING(i); - binind = ntbins + (size >> LG_QUANTUM) - 1; - CUSTOM_SMALL_SIZE2BIN(i) = binind; - } - /* Cacheline-spaced. */ - for (; i <= cspace_max; i += TINY_MIN) { - size = CACHELINE_CEILING(i); - binind = ntbins + nqbins + ((size - cspace_min) >> - LG_CACHELINE); - CUSTOM_SMALL_SIZE2BIN(i) = binind; - } - /* Sub-page. */ - for (; i <= sspace_max; i += TINY_MIN) { - size = SUBPAGE_CEILING(i); - binind = ntbins + nqbins + ncbins + ((size - sspace_min) >> - LG_SUBPAGE); - CUSTOM_SMALL_SIZE2BIN(i) = binind; - } - - small_size2bin = custom_small_size2bin; -#ifdef JEMALLOC_DEBUG - small_size2bin_validate(); -#endif - return (false); -#undef CUSTOM_SMALL_SIZE2BIN -} - /* * Calculate bin_info->run_size such that it meets the following constraints: * @@ -2444,18 +2007,40 @@ small_size2bin_init_hard(void) static size_t bin_info_run_size_calc(arena_bin_info_t *bin_info, size_t min_run_size) { + size_t pad_size; size_t try_run_size, good_run_size; uint32_t try_nregs, good_nregs; uint32_t try_hdr_size, good_hdr_size; uint32_t try_bitmap_offset, good_bitmap_offset; -#ifdef JEMALLOC_PROF uint32_t try_ctx0_offset, good_ctx0_offset; -#endif - uint32_t try_reg0_offset, good_reg0_offset; + uint32_t try_redzone0_offset, good_redzone0_offset; - assert(min_run_size >= PAGE_SIZE); + assert(min_run_size >= PAGE); assert(min_run_size <= arena_maxclass); + /* + * Determine redzone size based on minimum alignment and minimum + * redzone size. Add padding to the end of the run if it is needed to + * align the regions. The padding allows each redzone to be half the + * minimum alignment; without the padding, each redzone would have to + * be twice as large in order to maintain alignment. + */ + if (config_fill && opt_redzone) { + size_t align_min = ZU(1) << (ffs(bin_info->reg_size) - 1); + if (align_min <= REDZONE_MINSIZE) { + bin_info->redzone_size = REDZONE_MINSIZE; + pad_size = 0; + } else { + bin_info->redzone_size = align_min >> 1; + pad_size = bin_info->redzone_size; + } + } else { + bin_info->redzone_size = 0; + pad_size = 0; + } + bin_info->reg_interval = bin_info->reg_size + + (bin_info->redzone_size << 1); + /* * Calculate known-valid settings before entering the run_size * expansion loop, so that the first part of the loop always copies @@ -2467,7 +2052,8 @@ bin_info_run_size_calc(arena_bin_info_t *bin_info, size_t min_run_size) * header's mask length and the number of regions. */ try_run_size = min_run_size; - try_nregs = ((try_run_size - sizeof(arena_run_t)) / bin_info->reg_size) + try_nregs = ((try_run_size - sizeof(arena_run_t)) / + bin_info->reg_interval) + 1; /* Counter-act try_nregs-- in loop. */ if (try_nregs > RUN_MAXREGS) { try_nregs = RUN_MAXREGS @@ -2481,8 +2067,7 @@ bin_info_run_size_calc(arena_bin_info_t *bin_info, size_t min_run_size) try_bitmap_offset = try_hdr_size; /* Add space for bitmap. */ try_hdr_size += bitmap_size(try_nregs); -#ifdef JEMALLOC_PROF - if (opt_prof && prof_promote == false) { + if (config_prof && opt_prof && prof_promote == false) { /* Pad to a quantum boundary. */ try_hdr_size = QUANTUM_CEILING(try_hdr_size); try_ctx0_offset = try_hdr_size; @@ -2490,10 +2075,9 @@ bin_info_run_size_calc(arena_bin_info_t *bin_info, size_t min_run_size) try_hdr_size += try_nregs * sizeof(prof_ctx_t *); } else try_ctx0_offset = 0; -#endif - try_reg0_offset = try_run_size - (try_nregs * - bin_info->reg_size); - } while (try_hdr_size > try_reg0_offset); + try_redzone0_offset = try_run_size - (try_nregs * + bin_info->reg_interval) - pad_size; + } while (try_hdr_size > try_redzone0_offset); /* run_size expansion loop. */ do { @@ -2504,15 +2088,13 @@ bin_info_run_size_calc(arena_bin_info_t *bin_info, size_t min_run_size) good_nregs = try_nregs; good_hdr_size = try_hdr_size; good_bitmap_offset = try_bitmap_offset; -#ifdef JEMALLOC_PROF good_ctx0_offset = try_ctx0_offset; -#endif - good_reg0_offset = try_reg0_offset; + good_redzone0_offset = try_redzone0_offset; /* Try more aggressive settings. */ - try_run_size += PAGE_SIZE; - try_nregs = ((try_run_size - sizeof(arena_run_t)) / - bin_info->reg_size) + try_run_size += PAGE; + try_nregs = ((try_run_size - sizeof(arena_run_t) - pad_size) / + bin_info->reg_interval) + 1; /* Counter-act try_nregs-- in loop. */ if (try_nregs > RUN_MAXREGS) { try_nregs = RUN_MAXREGS @@ -2526,8 +2108,7 @@ bin_info_run_size_calc(arena_bin_info_t *bin_info, size_t min_run_size) try_bitmap_offset = try_hdr_size; /* Add space for bitmap. */ try_hdr_size += bitmap_size(try_nregs); -#ifdef JEMALLOC_PROF - if (opt_prof && prof_promote == false) { + if (config_prof && opt_prof && prof_promote == false) { /* Pad to a quantum boundary. */ try_hdr_size = QUANTUM_CEILING(try_hdr_size); try_ctx0_offset = try_hdr_size; @@ -2537,140 +2118,52 @@ bin_info_run_size_calc(arena_bin_info_t *bin_info, size_t min_run_size) try_hdr_size += try_nregs * sizeof(prof_ctx_t *); } -#endif - try_reg0_offset = try_run_size - (try_nregs * - bin_info->reg_size); - } while (try_hdr_size > try_reg0_offset); + try_redzone0_offset = try_run_size - (try_nregs * + bin_info->reg_interval) - pad_size; + } while (try_hdr_size > try_redzone0_offset); } while (try_run_size <= arena_maxclass && try_run_size <= arena_maxclass - && RUN_MAX_OVRHD * (bin_info->reg_size << 3) > RUN_MAX_OVRHD_RELAX - && (try_reg0_offset << RUN_BFP) > RUN_MAX_OVRHD * try_run_size + && RUN_MAX_OVRHD * (bin_info->reg_interval << 3) > + RUN_MAX_OVRHD_RELAX + && (try_redzone0_offset << RUN_BFP) > RUN_MAX_OVRHD * try_run_size && try_nregs < RUN_MAXREGS); - assert(good_hdr_size <= good_reg0_offset); + assert(good_hdr_size <= good_redzone0_offset); /* Copy final settings. */ bin_info->run_size = good_run_size; bin_info->nregs = good_nregs; bin_info->bitmap_offset = good_bitmap_offset; -#ifdef JEMALLOC_PROF bin_info->ctx0_offset = good_ctx0_offset; -#endif - bin_info->reg0_offset = good_reg0_offset; + bin_info->reg0_offset = good_redzone0_offset + bin_info->redzone_size; + + assert(bin_info->reg0_offset - bin_info->redzone_size + (bin_info->nregs + * bin_info->reg_interval) + pad_size == bin_info->run_size); return (good_run_size); } -static bool +static void bin_info_init(void) { arena_bin_info_t *bin_info; - unsigned i; - size_t prev_run_size; + size_t prev_run_size = PAGE; - arena_bin_info = base_alloc(sizeof(arena_bin_info_t) * nbins); - if (arena_bin_info == NULL) - return (true); - - prev_run_size = PAGE_SIZE; - i = 0; -#ifdef JEMALLOC_TINY - /* (2^n)-spaced tiny bins. */ - for (; i < ntbins; i++) { - bin_info = &arena_bin_info[i]; - bin_info->reg_size = (1U << (LG_TINY_MIN + i)); - prev_run_size = bin_info_run_size_calc(bin_info, prev_run_size); - bitmap_info_init(&bin_info->bitmap_info, bin_info->nregs); - } -#endif - - /* Quantum-spaced bins. */ - for (; i < ntbins + nqbins; i++) { - bin_info = &arena_bin_info[i]; - bin_info->reg_size = (i - ntbins + 1) << LG_QUANTUM; - prev_run_size = bin_info_run_size_calc(bin_info, prev_run_size); - bitmap_info_init(&bin_info->bitmap_info, bin_info->nregs); - } - - /* Cacheline-spaced bins. */ - for (; i < ntbins + nqbins + ncbins; i++) { - bin_info = &arena_bin_info[i]; - bin_info->reg_size = cspace_min + ((i - (ntbins + nqbins)) << - LG_CACHELINE); - prev_run_size = bin_info_run_size_calc(bin_info, prev_run_size); - bitmap_info_init(&bin_info->bitmap_info, bin_info->nregs); - } - - /* Subpage-spaced bins. */ - for (; i < nbins; i++) { - bin_info = &arena_bin_info[i]; - bin_info->reg_size = sspace_min + ((i - (ntbins + nqbins + - ncbins)) << LG_SUBPAGE); - prev_run_size = bin_info_run_size_calc(bin_info, prev_run_size); - bitmap_info_init(&bin_info->bitmap_info, bin_info->nregs); - } - - return (false); +#define SIZE_CLASS(bin, delta, size) \ + bin_info = &arena_bin_info[bin]; \ + bin_info->reg_size = size; \ + prev_run_size = bin_info_run_size_calc(bin_info, prev_run_size);\ + bitmap_info_init(&bin_info->bitmap_info, bin_info->nregs); + SIZE_CLASSES +#undef SIZE_CLASS } -bool +void arena_boot(void) { size_t header_size; unsigned i; - /* Set variables according to the value of opt_lg_[qc]space_max. */ - qspace_max = (1U << opt_lg_qspace_max); - cspace_min = CACHELINE_CEILING(qspace_max); - if (cspace_min == qspace_max) - cspace_min += CACHELINE; - cspace_max = (1U << opt_lg_cspace_max); - sspace_min = SUBPAGE_CEILING(cspace_max); - if (sspace_min == cspace_max) - sspace_min += SUBPAGE; - assert(sspace_min < PAGE_SIZE); - sspace_max = PAGE_SIZE - SUBPAGE; - -#ifdef JEMALLOC_TINY - assert(LG_QUANTUM >= LG_TINY_MIN); -#endif - assert(ntbins <= LG_QUANTUM); - nqbins = qspace_max >> LG_QUANTUM; - ncbins = ((cspace_max - cspace_min) >> LG_CACHELINE) + 1; - nsbins = ((sspace_max - sspace_min) >> LG_SUBPAGE) + 1; - nbins = ntbins + nqbins + ncbins + nsbins; - - /* - * The small_size2bin lookup table uses uint8_t to encode each bin - * index, so we cannot support more than 256 small size classes. This - * limit is difficult to exceed (not even possible with 16B quantum and - * 4KiB pages), and such configurations are impractical, but - * nonetheless we need to protect against this case in order to avoid - * undefined behavior. - * - * Further constrain nbins to 255 if prof_promote is true, since all - * small size classes, plus a "not small" size class must be stored in - * 8 bits of arena_chunk_map_t's bits field. - */ -#ifdef JEMALLOC_PROF - if (opt_prof && prof_promote) { - if (nbins > 255) { - char line_buf[UMAX2S_BUFSIZE]; - malloc_write(": Too many small size classes ("); - malloc_write(u2s(nbins, 10, line_buf)); - malloc_write(" > max 255)\n"); - abort(); - } - } else -#endif - if (nbins > 256) { - char line_buf[UMAX2S_BUFSIZE]; - malloc_write(": Too many small size classes ("); - malloc_write(u2s(nbins, 10, line_buf)); - malloc_write(" > max 256)\n"); - abort(); - } - /* * Compute the header size such that it is large enough to contain the * page map. The page map is biased to omit entries for the header @@ -2685,20 +2178,44 @@ arena_boot(void) */ map_bias = 0; for (i = 0; i < 3; i++) { - header_size = offsetof(arena_chunk_t, map) - + (sizeof(arena_chunk_map_t) * (chunk_npages-map_bias)); - map_bias = (header_size >> PAGE_SHIFT) + ((header_size & - PAGE_MASK) != 0); + header_size = offsetof(arena_chunk_t, map) + + (sizeof(arena_chunk_map_t) * (chunk_npages-map_bias)); + map_bias = (header_size >> LG_PAGE) + ((header_size & PAGE_MASK) + != 0); } assert(map_bias > 0); - arena_maxclass = chunksize - (map_bias << PAGE_SHIFT); + arena_maxclass = chunksize - (map_bias << LG_PAGE); - if (small_size2bin_init()) - return (true); - - if (bin_info_init()) - return (true); - - return (false); + bin_info_init(); +} + +void +arena_prefork(arena_t *arena) +{ + unsigned i; + + malloc_mutex_prefork(&arena->lock); + for (i = 0; i < NBINS; i++) + malloc_mutex_prefork(&arena->bins[i].lock); +} + +void +arena_postfork_parent(arena_t *arena) +{ + unsigned i; + + for (i = 0; i < NBINS; i++) + malloc_mutex_postfork_parent(&arena->bins[i].lock); + malloc_mutex_postfork_parent(&arena->lock); +} + +void +arena_postfork_child(arena_t *arena) +{ + unsigned i; + + for (i = 0; i < NBINS; i++) + malloc_mutex_postfork_child(&arena->bins[i].lock); + malloc_mutex_postfork_child(&arena->lock); } diff --git a/src/base.c b/src/base.c index cc85e849..bafaa743 100644 --- a/src/base.c +++ b/src/base.c @@ -4,7 +4,7 @@ /******************************************************************************/ /* Data. */ -malloc_mutex_t base_mtx; +static malloc_mutex_t base_mtx; /* * Current pages that are being used for internal memory allocations. These @@ -32,7 +32,7 @@ base_pages_alloc(size_t minsize) assert(minsize != 0); csize = CHUNK_CEILING(minsize); zero = false; - base_pages = chunk_alloc(csize, true, &zero); + base_pages = chunk_alloc(csize, chunksize, true, &zero); if (base_pages == NULL) return (true); base_next_addr = base_pages; @@ -66,6 +66,17 @@ base_alloc(size_t size) return (ret); } +void * +base_calloc(size_t number, size_t size) +{ + void *ret = base_alloc(number * size); + + if (ret != NULL) + memset(ret, 0, number * size); + + return (ret); +} + extent_node_t * base_node_alloc(void) { @@ -104,3 +115,24 @@ base_boot(void) return (false); } + +void +base_prefork(void) +{ + + malloc_mutex_prefork(&base_mtx); +} + +void +base_postfork_parent(void) +{ + + malloc_mutex_postfork_parent(&base_mtx); +} + +void +base_postfork_child(void) +{ + + malloc_mutex_postfork_child(&base_mtx); +} diff --git a/src/chunk.c b/src/chunk.c index d190c6f4..6bc24544 100644 --- a/src/chunk.c +++ b/src/chunk.c @@ -5,18 +5,20 @@ /* Data. */ size_t opt_lg_chunk = LG_CHUNK_DEFAULT; -#ifdef JEMALLOC_SWAP -bool opt_overcommit = true; -#endif -#if (defined(JEMALLOC_STATS) || defined(JEMALLOC_PROF)) malloc_mutex_t chunks_mtx; chunk_stats_t stats_chunks; -#endif -#ifdef JEMALLOC_IVSALLOC +/* + * Trees of chunks that were previously allocated (trees differ only in node + * ordering). These are used when allocating chunks, in an attempt to re-use + * address space. Depending on function, different tree orderings are needed, + * which is why there are two trees with the same contents. + */ +static extent_tree_t chunks_szad; +static extent_tree_t chunks_ad; + rtree_t *chunks_rtree; -#endif /* Various chunk-related settings. */ size_t chunksize; @@ -26,6 +28,98 @@ size_t map_bias; size_t arena_maxclass; /* Max size class for arenas. */ /******************************************************************************/ +/* Function prototypes for non-inline static functions. */ + +static void *chunk_recycle(size_t size, size_t alignment, bool base, + bool *zero); +static void chunk_record(void *chunk, size_t size); + +/******************************************************************************/ + +static void * +chunk_recycle(size_t size, size_t alignment, bool base, bool *zero) +{ + void *ret; + extent_node_t *node; + extent_node_t key; + size_t alloc_size, leadsize, trailsize; + + if (base) { + /* + * This function may need to call base_node_{,de}alloc(), but + * the current chunk allocation request is on behalf of the + * base allocator. Avoid deadlock (and if that weren't an + * issue, potential for infinite recursion) by returning NULL. + */ + return (NULL); + } + + alloc_size = size + alignment - chunksize; + /* Beware size_t wrap-around. */ + if (alloc_size < size) + return (NULL); + key.addr = NULL; + key.size = alloc_size; + malloc_mutex_lock(&chunks_mtx); + node = extent_tree_szad_nsearch(&chunks_szad, &key); + if (node == NULL) { + malloc_mutex_unlock(&chunks_mtx); + return (NULL); + } + leadsize = ALIGNMENT_CEILING((uintptr_t)node->addr, alignment) - + (uintptr_t)node->addr; + assert(node->size >= leadsize + size); + trailsize = node->size - leadsize - size; + ret = (void *)((uintptr_t)node->addr + leadsize); + /* Remove node from the tree. */ + extent_tree_szad_remove(&chunks_szad, node); + extent_tree_ad_remove(&chunks_ad, node); + if (leadsize != 0) { + /* Insert the leading space as a smaller chunk. */ + node->size = leadsize; + extent_tree_szad_insert(&chunks_szad, node); + extent_tree_ad_insert(&chunks_ad, node); + node = NULL; + } + if (trailsize != 0) { + /* Insert the trailing space as a smaller chunk. */ + if (node == NULL) { + /* + * An additional node is required, but + * base_node_alloc() can cause a new base chunk to be + * allocated. Drop chunks_mtx in order to avoid + * deadlock, and if node allocation fails, deallocate + * the result before returning an error. + */ + malloc_mutex_unlock(&chunks_mtx); + node = base_node_alloc(); + if (node == NULL) { + chunk_dealloc(ret, size, true); + return (NULL); + } + malloc_mutex_lock(&chunks_mtx); + } + node->addr = (void *)((uintptr_t)(ret) + size); + node->size = trailsize; + extent_tree_szad_insert(&chunks_szad, node); + extent_tree_ad_insert(&chunks_ad, node); + node = NULL; + } + malloc_mutex_unlock(&chunks_mtx); + + if (node != NULL) + base_node_dealloc(node); +#ifdef JEMALLOC_PURGE_MADVISE_DONTNEED + /* Pages are zeroed as a side effect of pages_purge(). */ + *zero = true; +#else + if (*zero) { + VALGRIND_MAKE_MEM_UNDEFINED(ret, size); + memset(ret, 0, size); + } +#endif + return (ret); +} /* * If the caller specifies (*zero == false), it is still possible to receive @@ -34,79 +128,138 @@ size_t arena_maxclass; /* Max size class for arenas. */ * advantage of them if they are returned. */ void * -chunk_alloc(size_t size, bool base, bool *zero) +chunk_alloc(size_t size, size_t alignment, bool base, bool *zero) { void *ret; assert(size != 0); assert((size & chunksize_mask) == 0); + assert(alignment != 0); + assert((alignment & chunksize_mask) == 0); -#ifdef JEMALLOC_SWAP - if (swap_enabled) { - ret = chunk_alloc_swap(size, zero); - if (ret != NULL) - goto RETURN; - } + ret = chunk_recycle(size, alignment, base, zero); + if (ret != NULL) + goto label_return; - if (swap_enabled == false || opt_overcommit) { -#endif -#ifdef JEMALLOC_DSS - ret = chunk_alloc_dss(size, zero); + ret = chunk_alloc_mmap(size, alignment, zero); + if (ret != NULL) + goto label_return; + + if (config_dss) { + ret = chunk_alloc_dss(size, alignment, zero); if (ret != NULL) - goto RETURN; -#endif - ret = chunk_alloc_mmap(size); - if (ret != NULL) { - *zero = true; - goto RETURN; - } -#ifdef JEMALLOC_SWAP + goto label_return; } -#endif /* All strategies for allocation failed. */ ret = NULL; -RETURN: -#ifdef JEMALLOC_IVSALLOC - if (base == false && ret != NULL) { +label_return: + if (config_ivsalloc && base == false && ret != NULL) { if (rtree_set(chunks_rtree, (uintptr_t)ret, ret)) { chunk_dealloc(ret, size, true); return (NULL); } } -#endif -#if (defined(JEMALLOC_STATS) || defined(JEMALLOC_PROF)) - if (ret != NULL) { -# ifdef JEMALLOC_PROF + if ((config_stats || config_prof) && ret != NULL) { bool gdump; -# endif malloc_mutex_lock(&chunks_mtx); -# ifdef JEMALLOC_STATS - stats_chunks.nchunks += (size / chunksize); -# endif + if (config_stats) + stats_chunks.nchunks += (size / chunksize); stats_chunks.curchunks += (size / chunksize); if (stats_chunks.curchunks > stats_chunks.highchunks) { stats_chunks.highchunks = stats_chunks.curchunks; -# ifdef JEMALLOC_PROF - gdump = true; -# endif - } -# ifdef JEMALLOC_PROF - else + if (config_prof) + gdump = true; + } else if (config_prof) gdump = false; -# endif malloc_mutex_unlock(&chunks_mtx); -# ifdef JEMALLOC_PROF - if (opt_prof && opt_prof_gdump && gdump) + if (config_prof && opt_prof && opt_prof_gdump && gdump) prof_gdump(); -# endif } -#endif + if (config_debug && *zero && ret != NULL) { + size_t i; + size_t *p = (size_t *)(uintptr_t)ret; + VALGRIND_MAKE_MEM_DEFINED(ret, size); + for (i = 0; i < size / sizeof(size_t); i++) + assert(p[i] == 0); + } assert(CHUNK_ADDR2BASE(ret) == ret); return (ret); } +static void +chunk_record(void *chunk, size_t size) +{ + extent_node_t *xnode, *node, *prev, key; + + pages_purge(chunk, size); + + /* + * Allocate a node before acquiring chunks_mtx even though it might not + * be needed, because base_node_alloc() may cause a new base chunk to + * be allocated, which could cause deadlock if chunks_mtx were already + * held. + */ + xnode = base_node_alloc(); + + malloc_mutex_lock(&chunks_mtx); + key.addr = (void *)((uintptr_t)chunk + size); + node = extent_tree_ad_nsearch(&chunks_ad, &key); + /* Try to coalesce forward. */ + if (node != NULL && node->addr == key.addr) { + /* + * Coalesce chunk with the following address range. This does + * not change the position within chunks_ad, so only + * remove/insert from/into chunks_szad. + */ + extent_tree_szad_remove(&chunks_szad, node); + node->addr = chunk; + node->size += size; + extent_tree_szad_insert(&chunks_szad, node); + if (xnode != NULL) + base_node_dealloc(xnode); + } else { + /* Coalescing forward failed, so insert a new node. */ + if (xnode == NULL) { + /* + * base_node_alloc() failed, which is an exceedingly + * unlikely failure. Leak chunk; its pages have + * already been purged, so this is only a virtual + * memory leak. + */ + malloc_mutex_unlock(&chunks_mtx); + return; + } + node = xnode; + node->addr = chunk; + node->size = size; + extent_tree_ad_insert(&chunks_ad, node); + extent_tree_szad_insert(&chunks_szad, node); + } + + /* Try to coalesce backward. */ + prev = extent_tree_ad_prev(&chunks_ad, node); + if (prev != NULL && (void *)((uintptr_t)prev->addr + prev->size) == + chunk) { + /* + * Coalesce chunk with the previous address range. This does + * not change the position within chunks_ad, so only + * remove/insert node from/into chunks_szad. + */ + extent_tree_szad_remove(&chunks_szad, prev); + extent_tree_ad_remove(&chunks_ad, prev); + + extent_tree_szad_remove(&chunks_szad, node); + node->addr = prev->addr; + node->size += prev->size; + extent_tree_szad_insert(&chunks_szad, node); + + base_node_dealloc(prev); + } + malloc_mutex_unlock(&chunks_mtx); +} + void chunk_dealloc(void *chunk, size_t size, bool unmap) { @@ -116,25 +269,18 @@ chunk_dealloc(void *chunk, size_t size, bool unmap) assert(size != 0); assert((size & chunksize_mask) == 0); -#ifdef JEMALLOC_IVSALLOC - rtree_set(chunks_rtree, (uintptr_t)chunk, NULL); -#endif -#if (defined(JEMALLOC_STATS) || defined(JEMALLOC_PROF)) - malloc_mutex_lock(&chunks_mtx); - stats_chunks.curchunks -= (size / chunksize); - malloc_mutex_unlock(&chunks_mtx); -#endif + if (config_ivsalloc) + rtree_set(chunks_rtree, (uintptr_t)chunk, NULL); + if (config_stats || config_prof) { + malloc_mutex_lock(&chunks_mtx); + stats_chunks.curchunks -= (size / chunksize); + malloc_mutex_unlock(&chunks_mtx); + } if (unmap) { -#ifdef JEMALLOC_SWAP - if (swap_enabled && chunk_dealloc_swap(chunk, size) == false) - return; -#endif -#ifdef JEMALLOC_DSS - if (chunk_dealloc_dss(chunk, size) == false) - return; -#endif - chunk_dealloc_mmap(chunk, size); + if ((config_dss && chunk_in_dss(chunk)) || + chunk_dealloc_mmap(chunk, size)) + chunk_record(chunk, size); } } @@ -144,30 +290,25 @@ chunk_boot(void) /* Set variables according to the value of opt_lg_chunk. */ chunksize = (ZU(1) << opt_lg_chunk); - assert(chunksize >= PAGE_SIZE); + assert(chunksize >= PAGE); chunksize_mask = chunksize - 1; - chunk_npages = (chunksize >> PAGE_SHIFT); + chunk_npages = (chunksize >> LG_PAGE); -#if (defined(JEMALLOC_STATS) || defined(JEMALLOC_PROF)) - if (malloc_mutex_init(&chunks_mtx)) + if (config_stats || config_prof) { + if (malloc_mutex_init(&chunks_mtx)) + return (true); + memset(&stats_chunks, 0, sizeof(chunk_stats_t)); + } + if (config_dss && chunk_dss_boot()) return (true); - memset(&stats_chunks, 0, sizeof(chunk_stats_t)); -#endif -#ifdef JEMALLOC_SWAP - if (chunk_swap_boot()) - return (true); -#endif - if (chunk_mmap_boot()) - return (true); -#ifdef JEMALLOC_DSS - if (chunk_dss_boot()) - return (true); -#endif -#ifdef JEMALLOC_IVSALLOC - chunks_rtree = rtree_new((ZU(1) << (LG_SIZEOF_PTR+3)) - opt_lg_chunk); - if (chunks_rtree == NULL) - return (true); -#endif + extent_tree_szad_new(&chunks_szad); + extent_tree_ad_new(&chunks_ad); + if (config_ivsalloc) { + chunks_rtree = rtree_new((ZU(1) << (LG_SIZEOF_PTR+3)) - + opt_lg_chunk); + if (chunks_rtree == NULL) + return (true); + } return (false); } diff --git a/src/chunk_dss.c b/src/chunk_dss.c index 5c0e290e..2d68e480 100644 --- a/src/chunk_dss.c +++ b/src/chunk_dss.c @@ -1,82 +1,42 @@ #define JEMALLOC_CHUNK_DSS_C_ #include "jemalloc/internal/jemalloc_internal.h" -#ifdef JEMALLOC_DSS /******************************************************************************/ /* Data. */ -malloc_mutex_t dss_mtx; +/* + * Protects sbrk() calls. This avoids malloc races among threads, though it + * does not protect against races with threads that call sbrk() directly. + */ +static malloc_mutex_t dss_mtx; /* Base address of the DSS. */ -static void *dss_base; +static void *dss_base; /* Current end of the DSS, or ((void *)-1) if the DSS is exhausted. */ -static void *dss_prev; +static void *dss_prev; /* Current upper limit on DSS addresses. */ -static void *dss_max; - -/* - * Trees of chunks that were previously allocated (trees differ only in node - * ordering). These are used when allocating chunks, in an attempt to re-use - * address space. Depending on function, different tree orderings are needed, - * which is why there are two trees with the same contents. - */ -static extent_tree_t dss_chunks_szad; -static extent_tree_t dss_chunks_ad; - -/******************************************************************************/ -/* Function prototypes for non-inline static functions. */ - -static void *chunk_recycle_dss(size_t size, bool *zero); -static extent_node_t *chunk_dealloc_dss_record(void *chunk, size_t size); +static void *dss_max; /******************************************************************************/ +#ifndef JEMALLOC_HAVE_SBRK static void * -chunk_recycle_dss(size_t size, bool *zero) +sbrk(intptr_t increment) { - extent_node_t *node, key; - key.addr = NULL; - key.size = size; - malloc_mutex_lock(&dss_mtx); - node = extent_tree_szad_nsearch(&dss_chunks_szad, &key); - if (node != NULL) { - void *ret = node->addr; - - /* Remove node from the tree. */ - extent_tree_szad_remove(&dss_chunks_szad, node); - if (node->size == size) { - extent_tree_ad_remove(&dss_chunks_ad, node); - base_node_dealloc(node); - } else { - /* - * Insert the remainder of node's address range as a - * smaller chunk. Its position within dss_chunks_ad - * does not change. - */ - assert(node->size > size); - node->addr = (void *)((uintptr_t)node->addr + size); - node->size -= size; - extent_tree_szad_insert(&dss_chunks_szad, node); - } - malloc_mutex_unlock(&dss_mtx); - - if (*zero) - memset(ret, 0, size); - return (ret); - } - malloc_mutex_unlock(&dss_mtx); + not_implemented(); return (NULL); } +#endif void * -chunk_alloc_dss(size_t size, bool *zero) +chunk_alloc_dss(size_t size, size_t alignment, bool *zero) { void *ret; - ret = chunk_recycle_dss(size, zero); - if (ret != NULL) - return (ret); + cassert(config_dss); + assert(size > 0 && (size & chunksize_mask) == 0); + assert(alignment > 0 && (alignment & chunksize_mask) == 0); /* * sbrk() uses a signed increment argument, so take care not to @@ -87,6 +47,8 @@ chunk_alloc_dss(size_t size, bool *zero) malloc_mutex_lock(&dss_mtx); if (dss_prev != (void *)-1) { + size_t gap_size, cpad_size; + void *cpad, *dss_next; intptr_t incr; /* @@ -97,26 +59,40 @@ chunk_alloc_dss(size_t size, bool *zero) do { /* Get the current end of the DSS. */ dss_max = sbrk(0); - /* * Calculate how much padding is necessary to * chunk-align the end of the DSS. */ - incr = (intptr_t)size - - (intptr_t)CHUNK_ADDR2OFFSET(dss_max); - if (incr == (intptr_t)size) - ret = dss_max; - else { - ret = (void *)((intptr_t)dss_max + incr); - incr += size; + gap_size = (chunksize - CHUNK_ADDR2OFFSET(dss_max)) & + chunksize_mask; + /* + * Compute how much chunk-aligned pad space (if any) is + * necessary to satisfy alignment. This space can be + * recycled for later use. + */ + cpad = (void *)((uintptr_t)dss_max + gap_size); + ret = (void *)ALIGNMENT_CEILING((uintptr_t)dss_max, + alignment); + cpad_size = (uintptr_t)ret - (uintptr_t)cpad; + dss_next = (void *)((uintptr_t)ret + size); + if ((uintptr_t)ret < (uintptr_t)dss_max || + (uintptr_t)dss_next < (uintptr_t)dss_max) { + /* Wrap-around. */ + malloc_mutex_unlock(&dss_mtx); + return (NULL); } - + incr = gap_size + cpad_size + size; dss_prev = sbrk(incr); if (dss_prev == dss_max) { /* Success. */ - dss_max = (void *)((intptr_t)dss_prev + incr); + dss_max = dss_next; malloc_mutex_unlock(&dss_mtx); - *zero = true; + if (cpad_size != 0) + chunk_dealloc(cpad, cpad_size, true); + if (*zero) { + VALGRIND_MAKE_MEM_UNDEFINED(ret, size); + memset(ret, 0, size); + } return (ret); } } while (dss_prev != (void *)-1); @@ -126,84 +102,13 @@ chunk_alloc_dss(size_t size, bool *zero) return (NULL); } -static extent_node_t * -chunk_dealloc_dss_record(void *chunk, size_t size) -{ - extent_node_t *xnode, *node, *prev, key; - - xnode = NULL; - while (true) { - key.addr = (void *)((uintptr_t)chunk + size); - node = extent_tree_ad_nsearch(&dss_chunks_ad, &key); - /* Try to coalesce forward. */ - if (node != NULL && node->addr == key.addr) { - /* - * Coalesce chunk with the following address range. - * This does not change the position within - * dss_chunks_ad, so only remove/insert from/into - * dss_chunks_szad. - */ - extent_tree_szad_remove(&dss_chunks_szad, node); - node->addr = chunk; - node->size += size; - extent_tree_szad_insert(&dss_chunks_szad, node); - break; - } else if (xnode == NULL) { - /* - * It is possible that base_node_alloc() will cause a - * new base chunk to be allocated, so take care not to - * deadlock on dss_mtx, and recover if another thread - * deallocates an adjacent chunk while this one is busy - * allocating xnode. - */ - malloc_mutex_unlock(&dss_mtx); - xnode = base_node_alloc(); - malloc_mutex_lock(&dss_mtx); - if (xnode == NULL) - return (NULL); - } else { - /* Coalescing forward failed, so insert a new node. */ - node = xnode; - xnode = NULL; - node->addr = chunk; - node->size = size; - extent_tree_ad_insert(&dss_chunks_ad, node); - extent_tree_szad_insert(&dss_chunks_szad, node); - break; - } - } - /* Discard xnode if it ended up unused do to a race. */ - if (xnode != NULL) - base_node_dealloc(xnode); - - /* Try to coalesce backward. */ - prev = extent_tree_ad_prev(&dss_chunks_ad, node); - if (prev != NULL && (void *)((uintptr_t)prev->addr + prev->size) == - chunk) { - /* - * Coalesce chunk with the previous address range. This does - * not change the position within dss_chunks_ad, so only - * remove/insert node from/into dss_chunks_szad. - */ - extent_tree_szad_remove(&dss_chunks_szad, prev); - extent_tree_ad_remove(&dss_chunks_ad, prev); - - extent_tree_szad_remove(&dss_chunks_szad, node); - node->addr = prev->addr; - node->size += prev->size; - extent_tree_szad_insert(&dss_chunks_szad, node); - - base_node_dealloc(prev); - } - - return (node); -} - bool chunk_in_dss(void *chunk) { bool ret; + cassert(config_dss); + malloc_mutex_lock(&dss_mtx); if ((uintptr_t)chunk >= (uintptr_t)dss_base && (uintptr_t)chunk < (uintptr_t)dss_max) @@ -215,70 +120,43 @@ chunk_in_dss(void *chunk) return (ret); } -bool -chunk_dealloc_dss(void *chunk, size_t size) -{ - bool ret; - - malloc_mutex_lock(&dss_mtx); - if ((uintptr_t)chunk >= (uintptr_t)dss_base - && (uintptr_t)chunk < (uintptr_t)dss_max) { - extent_node_t *node; - - /* Try to coalesce with other unused chunks. */ - node = chunk_dealloc_dss_record(chunk, size); - if (node != NULL) { - chunk = node->addr; - size = node->size; - } - - /* Get the current end of the DSS. */ - dss_max = sbrk(0); - - /* - * Try to shrink the DSS if this chunk is at the end of the - * DSS. The sbrk() call here is subject to a race condition - * with threads that use brk(2) or sbrk(2) directly, but the - * alternative would be to leak memory for the sake of poorly - * designed multi-threaded programs. - */ - if ((void *)((uintptr_t)chunk + size) == dss_max - && (dss_prev = sbrk(-(intptr_t)size)) == dss_max) { - /* Success. */ - dss_max = (void *)((intptr_t)dss_prev - (intptr_t)size); - - if (node != NULL) { - extent_tree_szad_remove(&dss_chunks_szad, node); - extent_tree_ad_remove(&dss_chunks_ad, node); - base_node_dealloc(node); - } - } else - madvise(chunk, size, MADV_DONTNEED); - - ret = false; - goto RETURN; - } - - ret = true; -RETURN: - malloc_mutex_unlock(&dss_mtx); - return (ret); -} - bool chunk_dss_boot(void) { + cassert(config_dss); + if (malloc_mutex_init(&dss_mtx)) return (true); dss_base = sbrk(0); dss_prev = dss_base; dss_max = dss_base; - extent_tree_szad_new(&dss_chunks_szad); - extent_tree_ad_new(&dss_chunks_ad); return (false); } +void +chunk_dss_prefork(void) +{ + + if (config_dss) + malloc_mutex_prefork(&dss_mtx); +} + +void +chunk_dss_postfork_parent(void) +{ + + if (config_dss) + malloc_mutex_postfork_parent(&dss_mtx); +} + +void +chunk_dss_postfork_child(void) +{ + + if (config_dss) + malloc_mutex_postfork_child(&dss_mtx); +} + /******************************************************************************/ -#endif /* JEMALLOC_DSS */ diff --git a/src/chunk_mmap.c b/src/chunk_mmap.c index 164e86e7..c8da6556 100644 --- a/src/chunk_mmap.c +++ b/src/chunk_mmap.c @@ -1,54 +1,37 @@ #define JEMALLOC_CHUNK_MMAP_C_ #include "jemalloc/internal/jemalloc_internal.h" -/******************************************************************************/ -/* Data. */ - -/* - * Used by chunk_alloc_mmap() to decide whether to attempt the fast path and - * potentially avoid some system calls. - */ -#ifndef NO_TLS -static __thread bool mmap_unaligned_tls - JEMALLOC_ATTR(tls_model("initial-exec")); -#define MMAP_UNALIGNED_GET() mmap_unaligned_tls -#define MMAP_UNALIGNED_SET(v) do { \ - mmap_unaligned_tls = (v); \ -} while (0) -#else -static pthread_key_t mmap_unaligned_tsd; -#define MMAP_UNALIGNED_GET() ((bool)pthread_getspecific(mmap_unaligned_tsd)) -#define MMAP_UNALIGNED_SET(v) do { \ - pthread_setspecific(mmap_unaligned_tsd, (void *)(v)); \ -} while (0) -#endif - /******************************************************************************/ /* Function prototypes for non-inline static functions. */ -static void *pages_map(void *addr, size_t size, bool noreserve); +static void *pages_map(void *addr, size_t size); static void pages_unmap(void *addr, size_t size); -static void *chunk_alloc_mmap_slow(size_t size, bool unaligned, - bool noreserve); -static void *chunk_alloc_mmap_internal(size_t size, bool noreserve); +static void *chunk_alloc_mmap_slow(size_t size, size_t alignment, + bool *zero); /******************************************************************************/ static void * -pages_map(void *addr, size_t size, bool noreserve) +pages_map(void *addr, size_t size) { void *ret; + assert(size != 0); + +#ifdef _WIN32 + /* + * If VirtualAlloc can't allocate at the given address when one is + * given, it fails and returns NULL. + */ + ret = VirtualAlloc(addr, size, MEM_COMMIT | MEM_RESERVE, + PAGE_READWRITE); +#else /* * We don't use MAP_FIXED here, because it can cause the *replacement* * of existing mappings, and we only want to create new mappings. */ - int flags = MAP_PRIVATE | MAP_ANON; -#ifdef MAP_NORESERVE - if (noreserve) - flags |= MAP_NORESERVE; -#endif - ret = mmap(addr, size, PROT_READ | PROT_WRITE, flags, -1, 0); + ret = mmap(addr, size, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANON, + -1, 0); assert(ret != NULL); if (ret == MAP_FAILED) @@ -60,16 +43,15 @@ pages_map(void *addr, size_t size, bool noreserve) if (munmap(ret, size) == -1) { char buf[BUFERROR_BUF]; - buferror(errno, buf, sizeof(buf)); - malloc_write(": Error in munmap(): "); - malloc_write(buf); - malloc_write("\n"); + buferror(buf, sizeof(buf)); + malloc_printf(": Error in munmap(): "); - malloc_write(buf); - malloc_write("\n"); + buferror(buf, sizeof(buf)); + malloc_printf(": Error in " +#ifdef _WIN32 + "VirtualFree" +#else + "munmap" +#endif + "(): %s\n", buf); if (opt_abort) abort(); } } static void * -chunk_alloc_mmap_slow(size_t size, bool unaligned, bool noreserve) +pages_trim(void *addr, size_t alloc_size, size_t leadsize, size_t size) { - void *ret; - size_t offset; + void *ret = (void *)((uintptr_t)addr + leadsize); - /* Beware size_t wrap-around. */ - if (size + chunksize <= size) + assert(alloc_size >= leadsize + size); +#ifdef _WIN32 + { + void *new_addr; + + pages_unmap(addr, alloc_size); + new_addr = pages_map(ret, size); + if (new_addr == ret) + return (ret); + if (new_addr) + pages_unmap(new_addr, size); return (NULL); - - ret = pages_map(NULL, size + chunksize, noreserve); - if (ret == NULL) - return (NULL); - - /* Clean up unneeded leading/trailing space. */ - offset = CHUNK_ADDR2OFFSET(ret); - if (offset != 0) { - /* Note that mmap() returned an unaligned mapping. */ - unaligned = true; - - /* Leading space. */ - pages_unmap(ret, chunksize - offset); - - ret = (void *)((uintptr_t)ret + - (chunksize - offset)); - - /* Trailing space. */ - pages_unmap((void *)((uintptr_t)ret + size), - offset); - } else { - /* Trailing space only. */ - pages_unmap((void *)((uintptr_t)ret + size), - chunksize); } +#else + { + size_t trailsize = alloc_size - leadsize - size; - /* - * If mmap() returned an aligned mapping, reset mmap_unaligned so that - * the next chunk_alloc_mmap() execution tries the fast allocation - * method. - */ - if (unaligned == false) - MMAP_UNALIGNED_SET(false); + if (leadsize != 0) + pages_unmap(addr, leadsize); + if (trailsize != 0) + pages_unmap((void *)((uintptr_t)ret + size), trailsize); + return (ret); + } +#endif +} - return (ret); +void +pages_purge(void *addr, size_t length) +{ + +#ifdef _WIN32 + VirtualAlloc(addr, length, MEM_RESET, PAGE_READWRITE); +#else +# ifdef JEMALLOC_PURGE_MADVISE_DONTNEED +# define JEMALLOC_MADV_PURGE MADV_DONTNEED +# elif defined(JEMALLOC_PURGE_MADVISE_FREE) +# define JEMALLOC_MADV_PURGE MADV_FREE +# else +# error "No method defined for purging unused dirty pages." +# endif + madvise(addr, length, JEMALLOC_MADV_PURGE); +#endif } static void * -chunk_alloc_mmap_internal(size_t size, bool noreserve) +chunk_alloc_mmap_slow(size_t size, size_t alignment, bool *zero) +{ + void *ret, *pages; + size_t alloc_size, leadsize; + + alloc_size = size + alignment - PAGE; + /* Beware size_t wrap-around. */ + if (alloc_size < size) + return (NULL); + do { + pages = pages_map(NULL, alloc_size); + if (pages == NULL) + return (NULL); + leadsize = ALIGNMENT_CEILING((uintptr_t)pages, alignment) - + (uintptr_t)pages; + ret = pages_trim(pages, alloc_size, leadsize, size); + } while (ret == NULL); + + assert(ret != NULL); + *zero = true; + return (ret); +} + +void * +chunk_alloc_mmap(size_t size, size_t alignment, bool *zero) { void *ret; + size_t offset; /* * Ideally, there would be a way to specify alignment to mmap() (like * NetBSD has), but in the absence of such a feature, we have to work * hard to efficiently create aligned mappings. The reliable, but * slow method is to create a mapping that is over-sized, then trim the - * excess. However, that always results in at least one call to + * excess. However, that always results in one or two calls to * pages_unmap(). * - * A more optimistic approach is to try mapping precisely the right - * amount, then try to append another mapping if alignment is off. In - * practice, this works out well as long as the application is not - * interleaving mappings via direct mmap() calls. If we do run into a - * situation where there is an interleaved mapping and we are unable to - * extend an unaligned mapping, our best option is to switch to the - * slow method until mmap() returns another aligned mapping. This will - * tend to leave a gap in the memory map that is too small to cause - * later problems for the optimistic method. - * - * Another possible confounding factor is address space layout - * randomization (ASLR), which causes mmap(2) to disregard the - * requested address. mmap_unaligned tracks whether the previous - * chunk_alloc_mmap() execution received any unaligned or relocated - * mappings, and if so, the current execution will immediately fall - * back to the slow method. However, we keep track of whether the fast - * method would have succeeded, and if so, we make a note to try the - * fast method next time. + * Optimistically try mapping precisely the right amount before falling + * back to the slow method, with the expectation that the optimistic + * approach works most of the time. */ - if (MMAP_UNALIGNED_GET() == false) { - size_t offset; + assert(alignment != 0); + assert((alignment & chunksize_mask) == 0); - ret = pages_map(NULL, size, noreserve); - if (ret == NULL) - return (NULL); - - offset = CHUNK_ADDR2OFFSET(ret); - if (offset != 0) { - MMAP_UNALIGNED_SET(true); - /* Try to extend chunk boundary. */ - if (pages_map((void *)((uintptr_t)ret + size), - chunksize - offset, noreserve) == NULL) { - /* - * Extension failed. Clean up, then revert to - * the reliable-but-expensive method. - */ - pages_unmap(ret, size); - ret = chunk_alloc_mmap_slow(size, true, - noreserve); - } else { - /* Clean up unneeded leading space. */ - pages_unmap(ret, chunksize - offset); - ret = (void *)((uintptr_t)ret + (chunksize - - offset)); - } - } - } else - ret = chunk_alloc_mmap_slow(size, false, noreserve); + ret = pages_map(NULL, size); + if (ret == NULL) + return (NULL); + offset = ALIGNMENT_ADDR2OFFSET(ret, alignment); + if (offset != 0) { + pages_unmap(ret, size); + return (chunk_alloc_mmap_slow(size, alignment, zero)); + } + assert(ret != NULL); + *zero = true; return (ret); } -void * -chunk_alloc_mmap(size_t size) -{ - - return (chunk_alloc_mmap_internal(size, false)); -} - -void * -chunk_alloc_mmap_noreserve(size_t size) -{ - - return (chunk_alloc_mmap_internal(size, true)); -} - -void +bool chunk_dealloc_mmap(void *chunk, size_t size) { - pages_unmap(chunk, size); -} - -bool -chunk_mmap_boot(void) -{ - -#ifdef NO_TLS - if (pthread_key_create(&mmap_unaligned_tsd, NULL) != 0) { - malloc_write(": Error in pthread_key_create()\n"); - return (true); - } -#endif - - return (false); + if (config_munmap) + pages_unmap(chunk, size); + + return (config_munmap == false); } diff --git a/src/chunk_swap.c b/src/chunk_swap.c deleted file mode 100644 index cb25ae0d..00000000 --- a/src/chunk_swap.c +++ /dev/null @@ -1,402 +0,0 @@ -#define JEMALLOC_CHUNK_SWAP_C_ -#include "jemalloc/internal/jemalloc_internal.h" -#ifdef JEMALLOC_SWAP -/******************************************************************************/ -/* Data. */ - -malloc_mutex_t swap_mtx; -bool swap_enabled; -bool swap_prezeroed; -size_t swap_nfds; -int *swap_fds; -#ifdef JEMALLOC_STATS -size_t swap_avail; -#endif - -/* Base address of the mmap()ed file(s). */ -static void *swap_base; -/* Current end of the space in use (<= swap_max). */ -static void *swap_end; -/* Absolute upper limit on file-backed addresses. */ -static void *swap_max; - -/* - * Trees of chunks that were previously allocated (trees differ only in node - * ordering). These are used when allocating chunks, in an attempt to re-use - * address space. Depending on function, different tree orderings are needed, - * which is why there are two trees with the same contents. - */ -static extent_tree_t swap_chunks_szad; -static extent_tree_t swap_chunks_ad; - -/******************************************************************************/ -/* Function prototypes for non-inline static functions. */ - -static void *chunk_recycle_swap(size_t size, bool *zero); -static extent_node_t *chunk_dealloc_swap_record(void *chunk, size_t size); - -/******************************************************************************/ - -static void * -chunk_recycle_swap(size_t size, bool *zero) -{ - extent_node_t *node, key; - - key.addr = NULL; - key.size = size; - malloc_mutex_lock(&swap_mtx); - node = extent_tree_szad_nsearch(&swap_chunks_szad, &key); - if (node != NULL) { - void *ret = node->addr; - - /* Remove node from the tree. */ - extent_tree_szad_remove(&swap_chunks_szad, node); - if (node->size == size) { - extent_tree_ad_remove(&swap_chunks_ad, node); - base_node_dealloc(node); - } else { - /* - * Insert the remainder of node's address range as a - * smaller chunk. Its position within swap_chunks_ad - * does not change. - */ - assert(node->size > size); - node->addr = (void *)((uintptr_t)node->addr + size); - node->size -= size; - extent_tree_szad_insert(&swap_chunks_szad, node); - } -#ifdef JEMALLOC_STATS - swap_avail -= size; -#endif - malloc_mutex_unlock(&swap_mtx); - - if (*zero) - memset(ret, 0, size); - return (ret); - } - malloc_mutex_unlock(&swap_mtx); - - return (NULL); -} - -void * -chunk_alloc_swap(size_t size, bool *zero) -{ - void *ret; - - assert(swap_enabled); - - ret = chunk_recycle_swap(size, zero); - if (ret != NULL) - return (ret); - - malloc_mutex_lock(&swap_mtx); - if ((uintptr_t)swap_end + size <= (uintptr_t)swap_max) { - ret = swap_end; - swap_end = (void *)((uintptr_t)swap_end + size); -#ifdef JEMALLOC_STATS - swap_avail -= size; -#endif - malloc_mutex_unlock(&swap_mtx); - - if (swap_prezeroed) - *zero = true; - else if (*zero) - memset(ret, 0, size); - } else { - malloc_mutex_unlock(&swap_mtx); - return (NULL); - } - - return (ret); -} - -static extent_node_t * -chunk_dealloc_swap_record(void *chunk, size_t size) -{ - extent_node_t *xnode, *node, *prev, key; - - xnode = NULL; - while (true) { - key.addr = (void *)((uintptr_t)chunk + size); - node = extent_tree_ad_nsearch(&swap_chunks_ad, &key); - /* Try to coalesce forward. */ - if (node != NULL && node->addr == key.addr) { - /* - * Coalesce chunk with the following address range. - * This does not change the position within - * swap_chunks_ad, so only remove/insert from/into - * swap_chunks_szad. - */ - extent_tree_szad_remove(&swap_chunks_szad, node); - node->addr = chunk; - node->size += size; - extent_tree_szad_insert(&swap_chunks_szad, node); - break; - } else if (xnode == NULL) { - /* - * It is possible that base_node_alloc() will cause a - * new base chunk to be allocated, so take care not to - * deadlock on swap_mtx, and recover if another thread - * deallocates an adjacent chunk while this one is busy - * allocating xnode. - */ - malloc_mutex_unlock(&swap_mtx); - xnode = base_node_alloc(); - malloc_mutex_lock(&swap_mtx); - if (xnode == NULL) - return (NULL); - } else { - /* Coalescing forward failed, so insert a new node. */ - node = xnode; - xnode = NULL; - node->addr = chunk; - node->size = size; - extent_tree_ad_insert(&swap_chunks_ad, node); - extent_tree_szad_insert(&swap_chunks_szad, node); - break; - } - } - /* Discard xnode if it ended up unused do to a race. */ - if (xnode != NULL) - base_node_dealloc(xnode); - - /* Try to coalesce backward. */ - prev = extent_tree_ad_prev(&swap_chunks_ad, node); - if (prev != NULL && (void *)((uintptr_t)prev->addr + prev->size) == - chunk) { - /* - * Coalesce chunk with the previous address range. This does - * not change the position within swap_chunks_ad, so only - * remove/insert node from/into swap_chunks_szad. - */ - extent_tree_szad_remove(&swap_chunks_szad, prev); - extent_tree_ad_remove(&swap_chunks_ad, prev); - - extent_tree_szad_remove(&swap_chunks_szad, node); - node->addr = prev->addr; - node->size += prev->size; - extent_tree_szad_insert(&swap_chunks_szad, node); - - base_node_dealloc(prev); - } - - return (node); -} - -bool -chunk_in_swap(void *chunk) -{ - bool ret; - - assert(swap_enabled); - - malloc_mutex_lock(&swap_mtx); - if ((uintptr_t)chunk >= (uintptr_t)swap_base - && (uintptr_t)chunk < (uintptr_t)swap_max) - ret = true; - else - ret = false; - malloc_mutex_unlock(&swap_mtx); - - return (ret); -} - -bool -chunk_dealloc_swap(void *chunk, size_t size) -{ - bool ret; - - assert(swap_enabled); - - malloc_mutex_lock(&swap_mtx); - if ((uintptr_t)chunk >= (uintptr_t)swap_base - && (uintptr_t)chunk < (uintptr_t)swap_max) { - extent_node_t *node; - - /* Try to coalesce with other unused chunks. */ - node = chunk_dealloc_swap_record(chunk, size); - if (node != NULL) { - chunk = node->addr; - size = node->size; - } - - /* - * Try to shrink the in-use memory if this chunk is at the end - * of the in-use memory. - */ - if ((void *)((uintptr_t)chunk + size) == swap_end) { - swap_end = (void *)((uintptr_t)swap_end - size); - - if (node != NULL) { - extent_tree_szad_remove(&swap_chunks_szad, - node); - extent_tree_ad_remove(&swap_chunks_ad, node); - base_node_dealloc(node); - } - } else - madvise(chunk, size, MADV_DONTNEED); - -#ifdef JEMALLOC_STATS - swap_avail += size; -#endif - ret = false; - goto RETURN; - } - - ret = true; -RETURN: - malloc_mutex_unlock(&swap_mtx); - return (ret); -} - -bool -chunk_swap_enable(const int *fds, unsigned nfds, bool prezeroed) -{ - bool ret; - unsigned i; - off_t off; - void *vaddr; - size_t cumsize, voff; - size_t sizes[nfds]; - - malloc_mutex_lock(&swap_mtx); - - /* Get file sizes. */ - for (i = 0, cumsize = 0; i < nfds; i++) { - off = lseek(fds[i], 0, SEEK_END); - if (off == ((off_t)-1)) { - ret = true; - goto RETURN; - } - if (PAGE_CEILING(off) != off) { - /* Truncate to a multiple of the page size. */ - off &= ~PAGE_MASK; - if (ftruncate(fds[i], off) != 0) { - ret = true; - goto RETURN; - } - } - sizes[i] = off; - if (cumsize + off < cumsize) { - /* - * Cumulative file size is greater than the total - * address space. Bail out while it's still obvious - * what the problem is. - */ - ret = true; - goto RETURN; - } - cumsize += off; - } - - /* Round down to a multiple of the chunk size. */ - cumsize &= ~chunksize_mask; - if (cumsize == 0) { - ret = true; - goto RETURN; - } - - /* - * Allocate a chunk-aligned region of anonymous memory, which will - * be the final location for the memory-mapped files. - */ - vaddr = chunk_alloc_mmap_noreserve(cumsize); - if (vaddr == NULL) { - ret = true; - goto RETURN; - } - - /* Overlay the files onto the anonymous mapping. */ - for (i = 0, voff = 0; i < nfds; i++) { - void *addr = mmap((void *)((uintptr_t)vaddr + voff), sizes[i], - PROT_READ | PROT_WRITE, MAP_SHARED | MAP_FIXED, fds[i], 0); - if (addr == MAP_FAILED) { - char buf[BUFERROR_BUF]; - - - buferror(errno, buf, sizeof(buf)); - malloc_write( - ": Error in mmap(..., MAP_FIXED, ...): "); - malloc_write(buf); - malloc_write("\n"); - if (opt_abort) - abort(); - if (munmap(vaddr, voff) == -1) { - buferror(errno, buf, sizeof(buf)); - malloc_write(": Error in munmap(): "); - malloc_write(buf); - malloc_write("\n"); - } - ret = true; - goto RETURN; - } - assert(addr == (void *)((uintptr_t)vaddr + voff)); - - /* - * Tell the kernel that the mapping will be accessed randomly, - * and that it should not gratuitously sync pages to the - * filesystem. - */ -#ifdef MADV_RANDOM - madvise(addr, sizes[i], MADV_RANDOM); -#endif -#ifdef MADV_NOSYNC - madvise(addr, sizes[i], MADV_NOSYNC); -#endif - - voff += sizes[i]; - } - - swap_prezeroed = prezeroed; - swap_base = vaddr; - swap_end = swap_base; - swap_max = (void *)((uintptr_t)vaddr + cumsize); - - /* Copy the fds array for mallctl purposes. */ - swap_fds = (int *)base_alloc(nfds * sizeof(int)); - if (swap_fds == NULL) { - ret = true; - goto RETURN; - } - memcpy(swap_fds, fds, nfds * sizeof(int)); - swap_nfds = nfds; - -#ifdef JEMALLOC_STATS - swap_avail = cumsize; -#endif - - swap_enabled = true; - - ret = false; -RETURN: - malloc_mutex_unlock(&swap_mtx); - return (ret); -} - -bool -chunk_swap_boot(void) -{ - - if (malloc_mutex_init(&swap_mtx)) - return (true); - - swap_enabled = false; - swap_prezeroed = false; /* swap.* mallctl's depend on this. */ - swap_nfds = 0; - swap_fds = NULL; -#ifdef JEMALLOC_STATS - swap_avail = 0; -#endif - swap_base = NULL; - swap_end = NULL; - swap_max = NULL; - - extent_tree_szad_new(&swap_chunks_szad); - extent_tree_ad_new(&swap_chunks_ad); - - return (false); -} - -/******************************************************************************/ -#endif /* JEMALLOC_SWAP */ diff --git a/src/ckh.c b/src/ckh.c index 43fcc252..742a950b 100644 --- a/src/ckh.c +++ b/src/ckh.c @@ -73,7 +73,6 @@ ckh_isearch(ckh_t *ckh, const void *key) size_t hash1, hash2, bucket, cell; assert(ckh != NULL); - dassert(ckh->magic == CKH_MAGIC); ckh->hash(key, ckh->lg_curbuckets, &hash1, &hash2); @@ -100,7 +99,7 @@ ckh_try_bucket_insert(ckh_t *ckh, size_t bucket, const void *key, * Cycle through the cells in the bucket, starting at a random position. * The randomness avoids worst-case search overhead as buckets fill up. */ - prn32(offset, LG_CKH_BUCKET_CELLS, ckh->prn_state, CKH_A, CKH_C); + prng32(offset, LG_CKH_BUCKET_CELLS, ckh->prng_state, CKH_A, CKH_C); for (i = 0; i < (ZU(1) << LG_CKH_BUCKET_CELLS); i++) { cell = &ckh->tab[(bucket << LG_CKH_BUCKET_CELLS) + ((i + offset) & ((ZU(1) << LG_CKH_BUCKET_CELLS) - 1))]; @@ -142,7 +141,7 @@ ckh_evict_reloc_insert(ckh_t *ckh, size_t argbucket, void const **argkey, * were an item for which both hashes indicated the same * bucket. */ - prn32(i, LG_CKH_BUCKET_CELLS, ckh->prn_state, CKH_A, CKH_C); + prng32(i, LG_CKH_BUCKET_CELLS, ckh->prng_state, CKH_A, CKH_C); cell = &ckh->tab[(bucket << LG_CKH_BUCKET_CELLS) + i]; assert(cell->key != NULL); @@ -265,15 +264,15 @@ ckh_grow(ckh_t *ckh) size_t usize; lg_curcells++; - usize = sa2u(sizeof(ckhc_t) << lg_curcells, CACHELINE, NULL); + usize = sa2u(sizeof(ckhc_t) << lg_curcells, CACHELINE); if (usize == 0) { ret = true; - goto RETURN; + goto label_return; } tab = (ckhc_t *)ipalloc(usize, CACHELINE, true); if (tab == NULL) { ret = true; - goto RETURN; + goto label_return; } /* Swap in new table. */ ttab = ckh->tab; @@ -293,7 +292,7 @@ ckh_grow(ckh_t *ckh) } ret = false; -RETURN: +label_return: return (ret); } @@ -310,7 +309,7 @@ ckh_shrink(ckh_t *ckh) */ lg_prevbuckets = ckh->lg_curbuckets; lg_curcells = ckh->lg_curbuckets + LG_CKH_BUCKET_CELLS - 1; - usize = sa2u(sizeof(ckhc_t) << lg_curcells, CACHELINE, NULL); + usize = sa2u(sizeof(ckhc_t) << lg_curcells, CACHELINE); if (usize == 0) return; tab = (ckhc_t *)ipalloc(usize, CACHELINE, true); @@ -362,7 +361,7 @@ ckh_new(ckh_t *ckh, size_t minitems, ckh_hash_t *hash, ckh_keycomp_t *keycomp) ckh->ninserts = 0; ckh->nrelocs = 0; #endif - ckh->prn_state = 42; /* Value doesn't really matter. */ + ckh->prng_state = 42; /* Value doesn't really matter. */ ckh->count = 0; /* @@ -383,23 +382,19 @@ ckh_new(ckh_t *ckh, size_t minitems, ckh_hash_t *hash, ckh_keycomp_t *keycomp) ckh->hash = hash; ckh->keycomp = keycomp; - usize = sa2u(sizeof(ckhc_t) << lg_mincells, CACHELINE, NULL); + usize = sa2u(sizeof(ckhc_t) << lg_mincells, CACHELINE); if (usize == 0) { ret = true; - goto RETURN; + goto label_return; } ckh->tab = (ckhc_t *)ipalloc(usize, CACHELINE, true); if (ckh->tab == NULL) { ret = true; - goto RETURN; + goto label_return; } -#ifdef JEMALLOC_DEBUG - ckh->magic = CKH_MAGIC; -#endif - ret = false; -RETURN: +label_return: return (ret); } @@ -408,7 +403,6 @@ ckh_delete(ckh_t *ckh) { assert(ckh != NULL); - dassert(ckh->magic == CKH_MAGIC); #ifdef CKH_VERBOSE malloc_printf( @@ -433,7 +427,6 @@ ckh_count(ckh_t *ckh) { assert(ckh != NULL); - dassert(ckh->magic == CKH_MAGIC); return (ckh->count); } @@ -464,7 +457,6 @@ ckh_insert(ckh_t *ckh, const void *key, const void *data) bool ret; assert(ckh != NULL); - dassert(ckh->magic == CKH_MAGIC); assert(ckh_search(ckh, key, NULL, NULL)); #ifdef CKH_COUNT @@ -474,12 +466,12 @@ ckh_insert(ckh_t *ckh, const void *key, const void *data) while (ckh_try_insert(ckh, &key, &data)) { if (ckh_grow(ckh)) { ret = true; - goto RETURN; + goto label_return; } } ret = false; -RETURN: +label_return: return (ret); } @@ -489,7 +481,6 @@ ckh_remove(ckh_t *ckh, const void *searchkey, void **key, void **data) size_t cell; assert(ckh != NULL); - dassert(ckh->magic == CKH_MAGIC); cell = ckh_isearch(ckh, searchkey); if (cell != SIZE_T_MAX) { @@ -521,7 +512,6 @@ ckh_search(ckh_t *ckh, const void *searchkey, void **key, void **data) size_t cell; assert(ckh != NULL); - dassert(ckh->magic == CKH_MAGIC); cell = ckh_isearch(ckh, searchkey); if (cell != SIZE_T_MAX) { @@ -545,7 +535,7 @@ ckh_string_hash(const void *key, unsigned minbits, size_t *hash1, size_t *hash2) assert(hash1 != NULL); assert(hash2 != NULL); - h = hash(key, strlen((const char *)key), 0x94122f335b332aeaLLU); + h = hash(key, strlen((const char *)key), UINT64_C(0x94122f335b332aea)); if (minbits <= 32) { /* * Avoid doing multiple hashes, since a single hash provides @@ -556,7 +546,7 @@ ckh_string_hash(const void *key, unsigned minbits, size_t *hash1, size_t *hash2) } else { ret1 = h; ret2 = hash(key, strlen((const char *)key), - 0x8432a476666bbc13LLU); + UINT64_C(0x8432a476666bbc13)); } *hash1 = ret1; @@ -593,7 +583,7 @@ ckh_pointer_hash(const void *key, unsigned minbits, size_t *hash1, u.i = 0; #endif u.v = key; - h = hash(&u.i, sizeof(u.i), 0xd983396e68886082LLU); + h = hash(&u.i, sizeof(u.i), UINT64_C(0xd983396e68886082)); if (minbits <= 32) { /* * Avoid doing multiple hashes, since a single hash provides @@ -604,7 +594,7 @@ ckh_pointer_hash(const void *key, unsigned minbits, size_t *hash1, } else { assert(SIZEOF_PTR == 8); ret1 = h; - ret2 = hash(&u.i, sizeof(u.i), 0x5e2be9aff8709a5dLLU); + ret2 = hash(&u.i, sizeof(u.i), UINT64_C(0x5e2be9aff8709a5d)); } *hash1 = ret1; diff --git a/src/ctl.c b/src/ctl.c index e5336d36..55e76677 100644 --- a/src/ctl.c +++ b/src/ctl.c @@ -8,14 +8,38 @@ * ctl_mtx protects the following: * - ctl_stats.* * - opt_prof_active - * - swap_enabled - * - swap_prezeroed */ static malloc_mutex_t ctl_mtx; static bool ctl_initialized; static uint64_t ctl_epoch; static ctl_stats_t ctl_stats; +/******************************************************************************/ +/* Helpers for named and indexed nodes. */ + +static inline const ctl_named_node_t * +ctl_named_node(const ctl_node_t *node) +{ + + return ((node->named) ? (const ctl_named_node_t *)node : NULL); +} + +static inline const ctl_named_node_t * +ctl_named_children(const ctl_named_node_t *node, int index) +{ + const ctl_named_node_t *children = ctl_named_node(node->children); + + return (children ? &children[index] : NULL); +} + +static inline const ctl_indexed_node_t * +ctl_indexed_node(const ctl_node_t *node) +{ + + return ((node->named == false) ? (const ctl_indexed_node_t *)node : + NULL); +} + /******************************************************************************/ /* Function prototypes for non-inline static functions. */ @@ -24,19 +48,15 @@ static int n##_ctl(const size_t *mib, size_t miblen, void *oldp, \ size_t *oldlenp, void *newp, size_t newlen); #define INDEX_PROTO(n) \ -const ctl_node_t *n##_index(const size_t *mib, size_t miblen, \ +const ctl_named_node_t *n##_index(const size_t *mib, size_t miblen, \ size_t i); -#ifdef JEMALLOC_STATS static bool ctl_arena_init(ctl_arena_stats_t *astats); -#endif static void ctl_arena_clear(ctl_arena_stats_t *astats); -#ifdef JEMALLOC_STATS static void ctl_arena_stats_amerge(ctl_arena_stats_t *cstats, arena_t *arena); static void ctl_arena_stats_smerge(ctl_arena_stats_t *sstats, ctl_arena_stats_t *astats); -#endif static void ctl_arena_refresh(arena_t *arena, unsigned i); static void ctl_refresh(void); static bool ctl_init(void); @@ -45,67 +65,51 @@ static int ctl_lookup(const char *name, ctl_node_t const **nodesp, CTL_PROTO(version) CTL_PROTO(epoch) -#ifdef JEMALLOC_TCACHE -CTL_PROTO(tcache_flush) -#endif +CTL_PROTO(thread_tcache_enabled) +CTL_PROTO(thread_tcache_flush) CTL_PROTO(thread_arena) -#ifdef JEMALLOC_STATS CTL_PROTO(thread_allocated) CTL_PROTO(thread_allocatedp) CTL_PROTO(thread_deallocated) CTL_PROTO(thread_deallocatedp) -#endif CTL_PROTO(config_debug) CTL_PROTO(config_dss) -CTL_PROTO(config_dynamic_page_shift) CTL_PROTO(config_fill) CTL_PROTO(config_lazy_lock) +CTL_PROTO(config_mremap) +CTL_PROTO(config_munmap) CTL_PROTO(config_prof) CTL_PROTO(config_prof_libgcc) CTL_PROTO(config_prof_libunwind) CTL_PROTO(config_stats) -CTL_PROTO(config_swap) -CTL_PROTO(config_sysv) CTL_PROTO(config_tcache) -CTL_PROTO(config_tiny) CTL_PROTO(config_tls) +CTL_PROTO(config_utrace) +CTL_PROTO(config_valgrind) CTL_PROTO(config_xmalloc) CTL_PROTO(opt_abort) -CTL_PROTO(opt_lg_qspace_max) -CTL_PROTO(opt_lg_cspace_max) CTL_PROTO(opt_lg_chunk) CTL_PROTO(opt_narenas) CTL_PROTO(opt_lg_dirty_mult) CTL_PROTO(opt_stats_print) -#ifdef JEMALLOC_FILL CTL_PROTO(opt_junk) CTL_PROTO(opt_zero) -#endif -#ifdef JEMALLOC_SYSV -CTL_PROTO(opt_sysv) -#endif -#ifdef JEMALLOC_XMALLOC +CTL_PROTO(opt_quarantine) +CTL_PROTO(opt_redzone) +CTL_PROTO(opt_utrace) +CTL_PROTO(opt_valgrind) CTL_PROTO(opt_xmalloc) -#endif -#ifdef JEMALLOC_TCACHE CTL_PROTO(opt_tcache) -CTL_PROTO(opt_lg_tcache_gc_sweep) -#endif -#ifdef JEMALLOC_PROF +CTL_PROTO(opt_lg_tcache_max) CTL_PROTO(opt_prof) CTL_PROTO(opt_prof_prefix) CTL_PROTO(opt_prof_active) -CTL_PROTO(opt_lg_prof_bt_max) CTL_PROTO(opt_lg_prof_sample) CTL_PROTO(opt_lg_prof_interval) CTL_PROTO(opt_prof_gdump) +CTL_PROTO(opt_prof_final) CTL_PROTO(opt_prof_leak) CTL_PROTO(opt_prof_accum) -CTL_PROTO(opt_lg_prof_tcmax) -#endif -#ifdef JEMALLOC_SWAP -CTL_PROTO(opt_overcommit) -#endif CTL_PROTO(arenas_bin_i_size) CTL_PROTO(arenas_bin_i_nregs) CTL_PROTO(arenas_bin_i_run_size) @@ -115,39 +119,15 @@ INDEX_PROTO(arenas_lrun_i) CTL_PROTO(arenas_narenas) CTL_PROTO(arenas_initialized) CTL_PROTO(arenas_quantum) -CTL_PROTO(arenas_cacheline) -CTL_PROTO(arenas_subpage) -CTL_PROTO(arenas_pagesize) -CTL_PROTO(arenas_chunksize) -#ifdef JEMALLOC_TINY -CTL_PROTO(arenas_tspace_min) -CTL_PROTO(arenas_tspace_max) -#endif -CTL_PROTO(arenas_qspace_min) -CTL_PROTO(arenas_qspace_max) -CTL_PROTO(arenas_cspace_min) -CTL_PROTO(arenas_cspace_max) -CTL_PROTO(arenas_sspace_min) -CTL_PROTO(arenas_sspace_max) -#ifdef JEMALLOC_TCACHE +CTL_PROTO(arenas_page) CTL_PROTO(arenas_tcache_max) -#endif -CTL_PROTO(arenas_ntbins) -CTL_PROTO(arenas_nqbins) -CTL_PROTO(arenas_ncbins) -CTL_PROTO(arenas_nsbins) CTL_PROTO(arenas_nbins) -#ifdef JEMALLOC_TCACHE CTL_PROTO(arenas_nhbins) -#endif CTL_PROTO(arenas_nlruns) CTL_PROTO(arenas_purge) -#ifdef JEMALLOC_PROF CTL_PROTO(prof_active) CTL_PROTO(prof_dump) CTL_PROTO(prof_interval) -#endif -#ifdef JEMALLOC_STATS CTL_PROTO(stats_chunks_current) CTL_PROTO(stats_chunks_total) CTL_PROTO(stats_chunks_high) @@ -166,46 +146,29 @@ CTL_PROTO(stats_arenas_i_bins_j_allocated) CTL_PROTO(stats_arenas_i_bins_j_nmalloc) CTL_PROTO(stats_arenas_i_bins_j_ndalloc) CTL_PROTO(stats_arenas_i_bins_j_nrequests) -#ifdef JEMALLOC_TCACHE CTL_PROTO(stats_arenas_i_bins_j_nfills) CTL_PROTO(stats_arenas_i_bins_j_nflushes) -#endif CTL_PROTO(stats_arenas_i_bins_j_nruns) CTL_PROTO(stats_arenas_i_bins_j_nreruns) -CTL_PROTO(stats_arenas_i_bins_j_highruns) CTL_PROTO(stats_arenas_i_bins_j_curruns) INDEX_PROTO(stats_arenas_i_bins_j) CTL_PROTO(stats_arenas_i_lruns_j_nmalloc) CTL_PROTO(stats_arenas_i_lruns_j_ndalloc) CTL_PROTO(stats_arenas_i_lruns_j_nrequests) -CTL_PROTO(stats_arenas_i_lruns_j_highruns) CTL_PROTO(stats_arenas_i_lruns_j_curruns) INDEX_PROTO(stats_arenas_i_lruns_j) -#endif CTL_PROTO(stats_arenas_i_nthreads) CTL_PROTO(stats_arenas_i_pactive) CTL_PROTO(stats_arenas_i_pdirty) -#ifdef JEMALLOC_STATS CTL_PROTO(stats_arenas_i_mapped) CTL_PROTO(stats_arenas_i_npurge) CTL_PROTO(stats_arenas_i_nmadvise) CTL_PROTO(stats_arenas_i_purged) -#endif INDEX_PROTO(stats_arenas_i) -#ifdef JEMALLOC_STATS CTL_PROTO(stats_cactive) CTL_PROTO(stats_allocated) CTL_PROTO(stats_active) CTL_PROTO(stats_mapped) -#endif -#ifdef JEMALLOC_SWAP -# ifdef JEMALLOC_STATS -CTL_PROTO(swap_avail) -# endif -CTL_PROTO(swap_prezeroed) -CTL_PROTO(swap_nfds) -CTL_PROTO(swap_fds) -#endif /******************************************************************************/ /* mallctl tree. */ @@ -213,296 +176,223 @@ CTL_PROTO(swap_fds) /* Maximum tree depth. */ #define CTL_MAX_DEPTH 6 -#define NAME(n) true, {.named = {n -#define CHILD(c) sizeof(c##_node) / sizeof(ctl_node_t), c##_node}}, NULL -#define CTL(c) 0, NULL}}, c##_ctl +#define NAME(n) {true}, n +#define CHILD(t, c) \ + sizeof(c##_node) / sizeof(ctl_##t##_node_t), \ + (ctl_node_t *)c##_node, \ + NULL +#define CTL(c) 0, NULL, c##_ctl /* * Only handles internal indexed nodes, since there are currently no external * ones. */ -#define INDEX(i) false, {.indexed = {i##_index}}, NULL +#define INDEX(i) {false}, i##_index -#ifdef JEMALLOC_TCACHE -static const ctl_node_t tcache_node[] = { - {NAME("flush"), CTL(tcache_flush)} +static const ctl_named_node_t tcache_node[] = { + {NAME("enabled"), CTL(thread_tcache_enabled)}, + {NAME("flush"), CTL(thread_tcache_flush)} }; -#endif -static const ctl_node_t thread_node[] = { - {NAME("arena"), CTL(thread_arena)} -#ifdef JEMALLOC_STATS - , +static const ctl_named_node_t thread_node[] = { + {NAME("arena"), CTL(thread_arena)}, {NAME("allocated"), CTL(thread_allocated)}, {NAME("allocatedp"), CTL(thread_allocatedp)}, {NAME("deallocated"), CTL(thread_deallocated)}, - {NAME("deallocatedp"), CTL(thread_deallocatedp)} -#endif + {NAME("deallocatedp"), CTL(thread_deallocatedp)}, + {NAME("tcache"), CHILD(named, tcache)} }; -static const ctl_node_t config_node[] = { +static const ctl_named_node_t config_node[] = { {NAME("debug"), CTL(config_debug)}, {NAME("dss"), CTL(config_dss)}, - {NAME("dynamic_page_shift"), CTL(config_dynamic_page_shift)}, {NAME("fill"), CTL(config_fill)}, {NAME("lazy_lock"), CTL(config_lazy_lock)}, + {NAME("mremap"), CTL(config_mremap)}, + {NAME("munmap"), CTL(config_munmap)}, {NAME("prof"), CTL(config_prof)}, {NAME("prof_libgcc"), CTL(config_prof_libgcc)}, {NAME("prof_libunwind"), CTL(config_prof_libunwind)}, {NAME("stats"), CTL(config_stats)}, - {NAME("swap"), CTL(config_swap)}, - {NAME("sysv"), CTL(config_sysv)}, {NAME("tcache"), CTL(config_tcache)}, - {NAME("tiny"), CTL(config_tiny)}, {NAME("tls"), CTL(config_tls)}, + {NAME("utrace"), CTL(config_utrace)}, + {NAME("valgrind"), CTL(config_valgrind)}, {NAME("xmalloc"), CTL(config_xmalloc)} }; -static const ctl_node_t opt_node[] = { +static const ctl_named_node_t opt_node[] = { {NAME("abort"), CTL(opt_abort)}, - {NAME("lg_qspace_max"), CTL(opt_lg_qspace_max)}, - {NAME("lg_cspace_max"), CTL(opt_lg_cspace_max)}, {NAME("lg_chunk"), CTL(opt_lg_chunk)}, {NAME("narenas"), CTL(opt_narenas)}, {NAME("lg_dirty_mult"), CTL(opt_lg_dirty_mult)}, - {NAME("stats_print"), CTL(opt_stats_print)} -#ifdef JEMALLOC_FILL - , + {NAME("stats_print"), CTL(opt_stats_print)}, {NAME("junk"), CTL(opt_junk)}, - {NAME("zero"), CTL(opt_zero)} -#endif -#ifdef JEMALLOC_SYSV - , - {NAME("sysv"), CTL(opt_sysv)} -#endif -#ifdef JEMALLOC_XMALLOC - , - {NAME("xmalloc"), CTL(opt_xmalloc)} -#endif -#ifdef JEMALLOC_TCACHE - , + {NAME("zero"), CTL(opt_zero)}, + {NAME("quarantine"), CTL(opt_quarantine)}, + {NAME("redzone"), CTL(opt_redzone)}, + {NAME("utrace"), CTL(opt_utrace)}, + {NAME("valgrind"), CTL(opt_valgrind)}, + {NAME("xmalloc"), CTL(opt_xmalloc)}, {NAME("tcache"), CTL(opt_tcache)}, - {NAME("lg_tcache_gc_sweep"), CTL(opt_lg_tcache_gc_sweep)} -#endif -#ifdef JEMALLOC_PROF - , + {NAME("lg_tcache_max"), CTL(opt_lg_tcache_max)}, {NAME("prof"), CTL(opt_prof)}, {NAME("prof_prefix"), CTL(opt_prof_prefix)}, {NAME("prof_active"), CTL(opt_prof_active)}, - {NAME("lg_prof_bt_max"), CTL(opt_lg_prof_bt_max)}, {NAME("lg_prof_sample"), CTL(opt_lg_prof_sample)}, {NAME("lg_prof_interval"), CTL(opt_lg_prof_interval)}, {NAME("prof_gdump"), CTL(opt_prof_gdump)}, + {NAME("prof_final"), CTL(opt_prof_final)}, {NAME("prof_leak"), CTL(opt_prof_leak)}, - {NAME("prof_accum"), CTL(opt_prof_accum)}, - {NAME("lg_prof_tcmax"), CTL(opt_lg_prof_tcmax)} -#endif -#ifdef JEMALLOC_SWAP - , - {NAME("overcommit"), CTL(opt_overcommit)} -#endif + {NAME("prof_accum"), CTL(opt_prof_accum)} }; -static const ctl_node_t arenas_bin_i_node[] = { +static const ctl_named_node_t arenas_bin_i_node[] = { {NAME("size"), CTL(arenas_bin_i_size)}, {NAME("nregs"), CTL(arenas_bin_i_nregs)}, {NAME("run_size"), CTL(arenas_bin_i_run_size)} }; -static const ctl_node_t super_arenas_bin_i_node[] = { - {NAME(""), CHILD(arenas_bin_i)} +static const ctl_named_node_t super_arenas_bin_i_node[] = { + {NAME(""), CHILD(named, arenas_bin_i)} }; -static const ctl_node_t arenas_bin_node[] = { +static const ctl_indexed_node_t arenas_bin_node[] = { {INDEX(arenas_bin_i)} }; -static const ctl_node_t arenas_lrun_i_node[] = { +static const ctl_named_node_t arenas_lrun_i_node[] = { {NAME("size"), CTL(arenas_lrun_i_size)} }; -static const ctl_node_t super_arenas_lrun_i_node[] = { - {NAME(""), CHILD(arenas_lrun_i)} +static const ctl_named_node_t super_arenas_lrun_i_node[] = { + {NAME(""), CHILD(named, arenas_lrun_i)} }; -static const ctl_node_t arenas_lrun_node[] = { +static const ctl_indexed_node_t arenas_lrun_node[] = { {INDEX(arenas_lrun_i)} }; -static const ctl_node_t arenas_node[] = { +static const ctl_named_node_t arenas_node[] = { {NAME("narenas"), CTL(arenas_narenas)}, {NAME("initialized"), CTL(arenas_initialized)}, {NAME("quantum"), CTL(arenas_quantum)}, - {NAME("cacheline"), CTL(arenas_cacheline)}, - {NAME("subpage"), CTL(arenas_subpage)}, - {NAME("pagesize"), CTL(arenas_pagesize)}, - {NAME("chunksize"), CTL(arenas_chunksize)}, -#ifdef JEMALLOC_TINY - {NAME("tspace_min"), CTL(arenas_tspace_min)}, - {NAME("tspace_max"), CTL(arenas_tspace_max)}, -#endif - {NAME("qspace_min"), CTL(arenas_qspace_min)}, - {NAME("qspace_max"), CTL(arenas_qspace_max)}, - {NAME("cspace_min"), CTL(arenas_cspace_min)}, - {NAME("cspace_max"), CTL(arenas_cspace_max)}, - {NAME("sspace_min"), CTL(arenas_sspace_min)}, - {NAME("sspace_max"), CTL(arenas_sspace_max)}, -#ifdef JEMALLOC_TCACHE + {NAME("page"), CTL(arenas_page)}, {NAME("tcache_max"), CTL(arenas_tcache_max)}, -#endif - {NAME("ntbins"), CTL(arenas_ntbins)}, - {NAME("nqbins"), CTL(arenas_nqbins)}, - {NAME("ncbins"), CTL(arenas_ncbins)}, - {NAME("nsbins"), CTL(arenas_nsbins)}, {NAME("nbins"), CTL(arenas_nbins)}, -#ifdef JEMALLOC_TCACHE {NAME("nhbins"), CTL(arenas_nhbins)}, -#endif - {NAME("bin"), CHILD(arenas_bin)}, + {NAME("bin"), CHILD(indexed, arenas_bin)}, {NAME("nlruns"), CTL(arenas_nlruns)}, - {NAME("lrun"), CHILD(arenas_lrun)}, + {NAME("lrun"), CHILD(indexed, arenas_lrun)}, {NAME("purge"), CTL(arenas_purge)} }; -#ifdef JEMALLOC_PROF -static const ctl_node_t prof_node[] = { +static const ctl_named_node_t prof_node[] = { {NAME("active"), CTL(prof_active)}, {NAME("dump"), CTL(prof_dump)}, {NAME("interval"), CTL(prof_interval)} }; -#endif -#ifdef JEMALLOC_STATS -static const ctl_node_t stats_chunks_node[] = { +static const ctl_named_node_t stats_chunks_node[] = { {NAME("current"), CTL(stats_chunks_current)}, {NAME("total"), CTL(stats_chunks_total)}, {NAME("high"), CTL(stats_chunks_high)} }; -static const ctl_node_t stats_huge_node[] = { +static const ctl_named_node_t stats_huge_node[] = { {NAME("allocated"), CTL(stats_huge_allocated)}, {NAME("nmalloc"), CTL(stats_huge_nmalloc)}, {NAME("ndalloc"), CTL(stats_huge_ndalloc)} }; -static const ctl_node_t stats_arenas_i_small_node[] = { +static const ctl_named_node_t stats_arenas_i_small_node[] = { {NAME("allocated"), CTL(stats_arenas_i_small_allocated)}, {NAME("nmalloc"), CTL(stats_arenas_i_small_nmalloc)}, {NAME("ndalloc"), CTL(stats_arenas_i_small_ndalloc)}, {NAME("nrequests"), CTL(stats_arenas_i_small_nrequests)} }; -static const ctl_node_t stats_arenas_i_large_node[] = { +static const ctl_named_node_t stats_arenas_i_large_node[] = { {NAME("allocated"), CTL(stats_arenas_i_large_allocated)}, {NAME("nmalloc"), CTL(stats_arenas_i_large_nmalloc)}, {NAME("ndalloc"), CTL(stats_arenas_i_large_ndalloc)}, {NAME("nrequests"), CTL(stats_arenas_i_large_nrequests)} }; -static const ctl_node_t stats_arenas_i_bins_j_node[] = { +static const ctl_named_node_t stats_arenas_i_bins_j_node[] = { {NAME("allocated"), CTL(stats_arenas_i_bins_j_allocated)}, {NAME("nmalloc"), CTL(stats_arenas_i_bins_j_nmalloc)}, {NAME("ndalloc"), CTL(stats_arenas_i_bins_j_ndalloc)}, {NAME("nrequests"), CTL(stats_arenas_i_bins_j_nrequests)}, -#ifdef JEMALLOC_TCACHE {NAME("nfills"), CTL(stats_arenas_i_bins_j_nfills)}, {NAME("nflushes"), CTL(stats_arenas_i_bins_j_nflushes)}, -#endif {NAME("nruns"), CTL(stats_arenas_i_bins_j_nruns)}, {NAME("nreruns"), CTL(stats_arenas_i_bins_j_nreruns)}, - {NAME("highruns"), CTL(stats_arenas_i_bins_j_highruns)}, {NAME("curruns"), CTL(stats_arenas_i_bins_j_curruns)} }; -static const ctl_node_t super_stats_arenas_i_bins_j_node[] = { - {NAME(""), CHILD(stats_arenas_i_bins_j)} +static const ctl_named_node_t super_stats_arenas_i_bins_j_node[] = { + {NAME(""), CHILD(named, stats_arenas_i_bins_j)} }; -static const ctl_node_t stats_arenas_i_bins_node[] = { +static const ctl_indexed_node_t stats_arenas_i_bins_node[] = { {INDEX(stats_arenas_i_bins_j)} }; -static const ctl_node_t stats_arenas_i_lruns_j_node[] = { +static const ctl_named_node_t stats_arenas_i_lruns_j_node[] = { {NAME("nmalloc"), CTL(stats_arenas_i_lruns_j_nmalloc)}, {NAME("ndalloc"), CTL(stats_arenas_i_lruns_j_ndalloc)}, {NAME("nrequests"), CTL(stats_arenas_i_lruns_j_nrequests)}, - {NAME("highruns"), CTL(stats_arenas_i_lruns_j_highruns)}, {NAME("curruns"), CTL(stats_arenas_i_lruns_j_curruns)} }; -static const ctl_node_t super_stats_arenas_i_lruns_j_node[] = { - {NAME(""), CHILD(stats_arenas_i_lruns_j)} +static const ctl_named_node_t super_stats_arenas_i_lruns_j_node[] = { + {NAME(""), CHILD(named, stats_arenas_i_lruns_j)} }; -static const ctl_node_t stats_arenas_i_lruns_node[] = { +static const ctl_indexed_node_t stats_arenas_i_lruns_node[] = { {INDEX(stats_arenas_i_lruns_j)} }; -#endif -static const ctl_node_t stats_arenas_i_node[] = { +static const ctl_named_node_t stats_arenas_i_node[] = { {NAME("nthreads"), CTL(stats_arenas_i_nthreads)}, {NAME("pactive"), CTL(stats_arenas_i_pactive)}, - {NAME("pdirty"), CTL(stats_arenas_i_pdirty)} -#ifdef JEMALLOC_STATS - , + {NAME("pdirty"), CTL(stats_arenas_i_pdirty)}, {NAME("mapped"), CTL(stats_arenas_i_mapped)}, {NAME("npurge"), CTL(stats_arenas_i_npurge)}, {NAME("nmadvise"), CTL(stats_arenas_i_nmadvise)}, {NAME("purged"), CTL(stats_arenas_i_purged)}, - {NAME("small"), CHILD(stats_arenas_i_small)}, - {NAME("large"), CHILD(stats_arenas_i_large)}, - {NAME("bins"), CHILD(stats_arenas_i_bins)}, - {NAME("lruns"), CHILD(stats_arenas_i_lruns)} -#endif + {NAME("small"), CHILD(named, stats_arenas_i_small)}, + {NAME("large"), CHILD(named, stats_arenas_i_large)}, + {NAME("bins"), CHILD(indexed, stats_arenas_i_bins)}, + {NAME("lruns"), CHILD(indexed, stats_arenas_i_lruns)} }; -static const ctl_node_t super_stats_arenas_i_node[] = { - {NAME(""), CHILD(stats_arenas_i)} +static const ctl_named_node_t super_stats_arenas_i_node[] = { + {NAME(""), CHILD(named, stats_arenas_i)} }; -static const ctl_node_t stats_arenas_node[] = { +static const ctl_indexed_node_t stats_arenas_node[] = { {INDEX(stats_arenas_i)} }; -static const ctl_node_t stats_node[] = { -#ifdef JEMALLOC_STATS +static const ctl_named_node_t stats_node[] = { {NAME("cactive"), CTL(stats_cactive)}, {NAME("allocated"), CTL(stats_allocated)}, {NAME("active"), CTL(stats_active)}, {NAME("mapped"), CTL(stats_mapped)}, - {NAME("chunks"), CHILD(stats_chunks)}, - {NAME("huge"), CHILD(stats_huge)}, -#endif - {NAME("arenas"), CHILD(stats_arenas)} + {NAME("chunks"), CHILD(named, stats_chunks)}, + {NAME("huge"), CHILD(named, stats_huge)}, + {NAME("arenas"), CHILD(indexed, stats_arenas)} }; -#ifdef JEMALLOC_SWAP -static const ctl_node_t swap_node[] = { -# ifdef JEMALLOC_STATS - {NAME("avail"), CTL(swap_avail)}, -# endif - {NAME("prezeroed"), CTL(swap_prezeroed)}, - {NAME("nfds"), CTL(swap_nfds)}, - {NAME("fds"), CTL(swap_fds)} -}; -#endif - -static const ctl_node_t root_node[] = { +static const ctl_named_node_t root_node[] = { {NAME("version"), CTL(version)}, {NAME("epoch"), CTL(epoch)}, -#ifdef JEMALLOC_TCACHE - {NAME("tcache"), CHILD(tcache)}, -#endif - {NAME("thread"), CHILD(thread)}, - {NAME("config"), CHILD(config)}, - {NAME("opt"), CHILD(opt)}, - {NAME("arenas"), CHILD(arenas)}, -#ifdef JEMALLOC_PROF - {NAME("prof"), CHILD(prof)}, -#endif - {NAME("stats"), CHILD(stats)} -#ifdef JEMALLOC_SWAP - , - {NAME("swap"), CHILD(swap)} -#endif + {NAME("thread"), CHILD(named, thread)}, + {NAME("config"), CHILD(named, config)}, + {NAME("opt"), CHILD(named, opt)}, + {NAME("arenas"), CHILD(named, arenas)}, + {NAME("prof"), CHILD(named, prof)}, + {NAME("stats"), CHILD(named, stats)} }; -static const ctl_node_t super_root_node[] = { - {NAME(""), CHILD(root)} +static const ctl_named_node_t super_root_node[] = { + {NAME(""), CHILD(named, root)} }; #undef NAME @@ -512,17 +402,10 @@ static const ctl_node_t super_root_node[] = { /******************************************************************************/ -#ifdef JEMALLOC_STATS static bool ctl_arena_init(ctl_arena_stats_t *astats) { - if (astats->bstats == NULL) { - astats->bstats = (malloc_bin_stats_t *)base_alloc(nbins * - sizeof(malloc_bin_stats_t)); - if (astats->bstats == NULL) - return (true); - } if (astats->lstats == NULL) { astats->lstats = (malloc_large_stats_t *)base_alloc(nlclasses * sizeof(malloc_large_stats_t)); @@ -532,7 +415,6 @@ ctl_arena_init(ctl_arena_stats_t *astats) return (false); } -#endif static void ctl_arena_clear(ctl_arena_stats_t *astats) @@ -540,18 +422,18 @@ ctl_arena_clear(ctl_arena_stats_t *astats) astats->pactive = 0; astats->pdirty = 0; -#ifdef JEMALLOC_STATS - memset(&astats->astats, 0, sizeof(arena_stats_t)); - astats->allocated_small = 0; - astats->nmalloc_small = 0; - astats->ndalloc_small = 0; - astats->nrequests_small = 0; - memset(astats->bstats, 0, nbins * sizeof(malloc_bin_stats_t)); - memset(astats->lstats, 0, nlclasses * sizeof(malloc_large_stats_t)); -#endif + if (config_stats) { + memset(&astats->astats, 0, sizeof(arena_stats_t)); + astats->allocated_small = 0; + astats->nmalloc_small = 0; + astats->ndalloc_small = 0; + astats->nrequests_small = 0; + memset(astats->bstats, 0, NBINS * sizeof(malloc_bin_stats_t)); + memset(astats->lstats, 0, nlclasses * + sizeof(malloc_large_stats_t)); + } } -#ifdef JEMALLOC_STATS static void ctl_arena_stats_amerge(ctl_arena_stats_t *cstats, arena_t *arena) { @@ -560,7 +442,7 @@ ctl_arena_stats_amerge(ctl_arena_stats_t *cstats, arena_t *arena) arena_stats_merge(arena, &cstats->pactive, &cstats->pdirty, &cstats->astats, cstats->bstats, cstats->lstats); - for (i = 0; i < nbins; i++) { + for (i = 0; i < NBINS; i++) { cstats->allocated_small += cstats->bstats[i].allocated; cstats->nmalloc_small += cstats->bstats[i].nmalloc; cstats->ndalloc_small += cstats->bstats[i].ndalloc; @@ -595,26 +477,24 @@ ctl_arena_stats_smerge(ctl_arena_stats_t *sstats, ctl_arena_stats_t *astats) sstats->lstats[i].nmalloc += astats->lstats[i].nmalloc; sstats->lstats[i].ndalloc += astats->lstats[i].ndalloc; sstats->lstats[i].nrequests += astats->lstats[i].nrequests; - sstats->lstats[i].highruns += astats->lstats[i].highruns; sstats->lstats[i].curruns += astats->lstats[i].curruns; } - for (i = 0; i < nbins; i++) { + for (i = 0; i < NBINS; i++) { sstats->bstats[i].allocated += astats->bstats[i].allocated; sstats->bstats[i].nmalloc += astats->bstats[i].nmalloc; sstats->bstats[i].ndalloc += astats->bstats[i].ndalloc; sstats->bstats[i].nrequests += astats->bstats[i].nrequests; -#ifdef JEMALLOC_TCACHE - sstats->bstats[i].nfills += astats->bstats[i].nfills; - sstats->bstats[i].nflushes += astats->bstats[i].nflushes; -#endif + if (config_tcache) { + sstats->bstats[i].nfills += astats->bstats[i].nfills; + sstats->bstats[i].nflushes += + astats->bstats[i].nflushes; + } sstats->bstats[i].nruns += astats->bstats[i].nruns; sstats->bstats[i].reruns += astats->bstats[i].reruns; - sstats->bstats[i].highruns += astats->bstats[i].highruns; sstats->bstats[i].curruns += astats->bstats[i].curruns; } } -#endif static void ctl_arena_refresh(arena_t *arena, unsigned i) @@ -625,38 +505,38 @@ ctl_arena_refresh(arena_t *arena, unsigned i) ctl_arena_clear(astats); sstats->nthreads += astats->nthreads; -#ifdef JEMALLOC_STATS - ctl_arena_stats_amerge(astats, arena); - /* Merge into sum stats as well. */ - ctl_arena_stats_smerge(sstats, astats); -#else - astats->pactive += arena->nactive; - astats->pdirty += arena->ndirty; - /* Merge into sum stats as well. */ - sstats->pactive += arena->nactive; - sstats->pdirty += arena->ndirty; -#endif + if (config_stats) { + ctl_arena_stats_amerge(astats, arena); + /* Merge into sum stats as well. */ + ctl_arena_stats_smerge(sstats, astats); + } else { + astats->pactive += arena->nactive; + astats->pdirty += arena->ndirty; + /* Merge into sum stats as well. */ + sstats->pactive += arena->nactive; + sstats->pdirty += arena->ndirty; + } } static void ctl_refresh(void) { unsigned i; - arena_t *tarenas[narenas]; + VARIABLE_ARRAY(arena_t *, tarenas, narenas); -#ifdef JEMALLOC_STATS - malloc_mutex_lock(&chunks_mtx); - ctl_stats.chunks.current = stats_chunks.curchunks; - ctl_stats.chunks.total = stats_chunks.nchunks; - ctl_stats.chunks.high = stats_chunks.highchunks; - malloc_mutex_unlock(&chunks_mtx); + if (config_stats) { + malloc_mutex_lock(&chunks_mtx); + ctl_stats.chunks.current = stats_chunks.curchunks; + ctl_stats.chunks.total = stats_chunks.nchunks; + ctl_stats.chunks.high = stats_chunks.highchunks; + malloc_mutex_unlock(&chunks_mtx); - malloc_mutex_lock(&huge_mtx); - ctl_stats.huge.allocated = huge_allocated; - ctl_stats.huge.nmalloc = huge_nmalloc; - ctl_stats.huge.ndalloc = huge_ndalloc; - malloc_mutex_unlock(&huge_mtx); -#endif + malloc_mutex_lock(&huge_mtx); + ctl_stats.huge.allocated = huge_allocated; + ctl_stats.huge.nmalloc = huge_nmalloc; + ctl_stats.huge.ndalloc = huge_ndalloc; + malloc_mutex_unlock(&huge_mtx); + } /* * Clear sum stats, since they will be merged into by @@ -682,20 +562,14 @@ ctl_refresh(void) ctl_arena_refresh(tarenas[i], i); } -#ifdef JEMALLOC_STATS - ctl_stats.allocated = ctl_stats.arenas[narenas].allocated_small - + ctl_stats.arenas[narenas].astats.allocated_large - + ctl_stats.huge.allocated; - ctl_stats.active = (ctl_stats.arenas[narenas].pactive << PAGE_SHIFT) - + ctl_stats.huge.allocated; - ctl_stats.mapped = (ctl_stats.chunks.current << opt_lg_chunk); - -# ifdef JEMALLOC_SWAP - malloc_mutex_lock(&swap_mtx); - ctl_stats.swap_avail = swap_avail; - malloc_mutex_unlock(&swap_mtx); -# endif -#endif + if (config_stats) { + ctl_stats.allocated = ctl_stats.arenas[narenas].allocated_small + + ctl_stats.arenas[narenas].astats.allocated_large + + ctl_stats.huge.allocated; + ctl_stats.active = (ctl_stats.arenas[narenas].pactive << + LG_PAGE) + ctl_stats.huge.allocated; + ctl_stats.mapped = (ctl_stats.chunks.current << opt_lg_chunk); + } ctl_epoch++; } @@ -707,10 +581,6 @@ ctl_init(void) malloc_mutex_lock(&ctl_mtx); if (ctl_initialized == false) { -#ifdef JEMALLOC_STATS - unsigned i; -#endif - /* * Allocate space for one extra arena stats element, which * contains summed stats across all arenas. @@ -719,7 +589,7 @@ ctl_init(void) (narenas + 1) * sizeof(ctl_arena_stats_t)); if (ctl_stats.arenas == NULL) { ret = true; - goto RETURN; + goto label_return; } memset(ctl_stats.arenas, 0, (narenas + 1) * sizeof(ctl_arena_stats_t)); @@ -729,14 +599,15 @@ ctl_init(void) * ever get used. Lazy initialization would allow errors to * cause inconsistent state to be viewable by the application. */ -#ifdef JEMALLOC_STATS - for (i = 0; i <= narenas; i++) { - if (ctl_arena_init(&ctl_stats.arenas[i])) { - ret = true; - goto RETURN; + if (config_stats) { + unsigned i; + for (i = 0; i <= narenas; i++) { + if (ctl_arena_init(&ctl_stats.arenas[i])) { + ret = true; + goto label_return; + } } } -#endif ctl_stats.arenas[narenas].initialized = true; ctl_epoch = 0; @@ -745,7 +616,7 @@ ctl_init(void) } ret = false; -RETURN: +label_return: malloc_mutex_unlock(&ctl_mtx); return (ret); } @@ -757,7 +628,7 @@ ctl_lookup(const char *name, ctl_node_t const **nodesp, size_t *mibp, int ret; const char *elm, *tdot, *dot; size_t elen, i, j; - const ctl_node_t *node; + const ctl_named_node_t *node; elm = name; /* Equivalent to strchrnul(). */ @@ -765,54 +636,53 @@ ctl_lookup(const char *name, ctl_node_t const **nodesp, size_t *mibp, elen = (size_t)((uintptr_t)dot - (uintptr_t)elm); if (elen == 0) { ret = ENOENT; - goto RETURN; + goto label_return; } node = super_root_node; for (i = 0; i < *depthp; i++) { - assert(node->named); - assert(node->u.named.nchildren > 0); - if (node->u.named.children[0].named) { - const ctl_node_t *pnode = node; + assert(node); + assert(node->nchildren > 0); + if (ctl_named_node(node->children) != NULL) { + const ctl_named_node_t *pnode = node; /* Children are named. */ - for (j = 0; j < node->u.named.nchildren; j++) { - const ctl_node_t *child = - &node->u.named.children[j]; - if (strlen(child->u.named.name) == elen - && strncmp(elm, child->u.named.name, - elen) == 0) { + for (j = 0; j < node->nchildren; j++) { + const ctl_named_node_t *child = + ctl_named_children(node, j); + if (strlen(child->name) == elen && + strncmp(elm, child->name, elen) == 0) { node = child; if (nodesp != NULL) - nodesp[i] = node; + nodesp[i] = + (const ctl_node_t *)node; mibp[i] = j; break; } } if (node == pnode) { ret = ENOENT; - goto RETURN; + goto label_return; } } else { - unsigned long index; - const ctl_node_t *inode; + uintmax_t index; + const ctl_indexed_node_t *inode; /* Children are indexed. */ - index = strtoul(elm, NULL, 10); - if (index == ULONG_MAX) { + index = malloc_strtoumax(elm, NULL, 10); + if (index == UINTMAX_MAX || index > SIZE_T_MAX) { ret = ENOENT; - goto RETURN; + goto label_return; } - inode = &node->u.named.children[0]; - node = inode->u.indexed.index(mibp, *depthp, - index); + inode = ctl_indexed_node(node->children); + node = inode->index(mibp, *depthp, (size_t)index); if (node == NULL) { ret = ENOENT; - goto RETURN; + goto label_return; } if (nodesp != NULL) - nodesp[i] = node; + nodesp[i] = (const ctl_node_t *)node; mibp[i] = (size_t)index; } @@ -824,7 +694,7 @@ ctl_lookup(const char *name, ctl_node_t const **nodesp, size_t *mibp, * in this path through the tree. */ ret = ENOENT; - goto RETURN; + goto label_return; } /* Complete lookup successful. */ *depthp = i + 1; @@ -835,7 +705,7 @@ ctl_lookup(const char *name, ctl_node_t const **nodesp, size_t *mibp, if (*dot == '\0') { /* No more elements. */ ret = ENOENT; - goto RETURN; + goto label_return; } elm = &dot[1]; dot = ((tdot = strchr(elm, '.')) != NULL) ? tdot : @@ -844,7 +714,7 @@ ctl_lookup(const char *name, ctl_node_t const **nodesp, size_t *mibp, } ret = 0; -RETURN: +label_return: return (ret); } @@ -856,25 +726,27 @@ ctl_byname(const char *name, void *oldp, size_t *oldlenp, void *newp, size_t depth; ctl_node_t const *nodes[CTL_MAX_DEPTH]; size_t mib[CTL_MAX_DEPTH]; + const ctl_named_node_t *node; if (ctl_initialized == false && ctl_init()) { ret = EAGAIN; - goto RETURN; + goto label_return; } depth = CTL_MAX_DEPTH; ret = ctl_lookup(name, nodes, mib, &depth); if (ret != 0) - goto RETURN; + goto label_return; - if (nodes[depth-1]->ctl == NULL) { + node = ctl_named_node(nodes[depth-1]); + if (node != NULL && node->ctl) + ret = node->ctl(mib, depth, oldp, oldlenp, newp, newlen); + else { /* The name refers to a partial path through the ctl tree. */ ret = ENOENT; - goto RETURN; } - ret = nodes[depth-1]->ctl(mib, depth, oldp, oldlenp, newp, newlen); -RETURN: +label_return: return(ret); } @@ -885,11 +757,11 @@ ctl_nametomib(const char *name, size_t *mibp, size_t *miblenp) if (ctl_initialized == false && ctl_init()) { ret = EAGAIN; - goto RETURN; + goto label_return; } ret = ctl_lookup(name, NULL, mibp, miblenp); -RETURN: +label_return: return(ret); } @@ -898,46 +770,48 @@ ctl_bymib(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp, void *newp, size_t newlen) { int ret; - const ctl_node_t *node; + const ctl_named_node_t *node; size_t i; if (ctl_initialized == false && ctl_init()) { ret = EAGAIN; - goto RETURN; + goto label_return; } /* Iterate down the tree. */ node = super_root_node; for (i = 0; i < miblen; i++) { - if (node->u.named.children[0].named) { + assert(node); + assert(node->nchildren > 0); + if (ctl_named_node(node->children) != NULL) { /* Children are named. */ - if (node->u.named.nchildren <= mib[i]) { + if (node->nchildren <= mib[i]) { ret = ENOENT; - goto RETURN; + goto label_return; } - node = &node->u.named.children[mib[i]]; + node = ctl_named_children(node, mib[i]); } else { - const ctl_node_t *inode; + const ctl_indexed_node_t *inode; /* Indexed element. */ - inode = &node->u.named.children[0]; - node = inode->u.indexed.index(mib, miblen, mib[i]); + inode = ctl_indexed_node(node->children); + node = inode->index(mib, miblen, mib[i]); if (node == NULL) { ret = ENOENT; - goto RETURN; + goto label_return; } } } /* Call the ctl function. */ - if (node->ctl == NULL) { + if (node && node->ctl) + ret = node->ctl(mib, miblen, oldp, oldlenp, newp, newlen); + else { /* Partial MIB. */ ret = ENOENT; - goto RETURN; } - ret = node->ctl(mib, miblen, oldp, oldlenp, newp, newlen); -RETURN: +label_return: return(ret); } @@ -959,22 +833,17 @@ ctl_boot(void) #define READONLY() do { \ if (newp != NULL || newlen != 0) { \ ret = EPERM; \ - goto RETURN; \ + goto label_return; \ } \ } while (0) #define WRITEONLY() do { \ if (oldp != NULL || oldlenp != NULL) { \ ret = EPERM; \ - goto RETURN; \ + goto label_return; \ } \ } while (0) -#define VOID() do { \ - READONLY(); \ - WRITEONLY(); \ -} while (0) - #define READ(v, t) do { \ if (oldp != NULL && oldlenp != NULL) { \ if (*oldlenp != sizeof(t)) { \ @@ -982,7 +851,7 @@ ctl_boot(void) ? sizeof(t) : *oldlenp; \ memcpy(oldp, (void *)&v, copylen); \ ret = EINVAL; \ - goto RETURN; \ + goto label_return; \ } else \ *(t *)oldp = v; \ } \ @@ -992,12 +861,60 @@ ctl_boot(void) if (newp != NULL) { \ if (newlen != sizeof(t)) { \ ret = EINVAL; \ - goto RETURN; \ + goto label_return; \ } \ v = *(t *)newp; \ } \ } while (0) +/* + * There's a lot of code duplication in the following macros due to limitations + * in how nested cpp macros are expanded. + */ +#define CTL_RO_CLGEN(c, l, n, v, t) \ +static int \ +n##_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp, \ + void *newp, size_t newlen) \ +{ \ + int ret; \ + t oldval; \ + \ + if ((c) == false) \ + return (ENOENT); \ + if (l) \ + malloc_mutex_lock(&ctl_mtx); \ + READONLY(); \ + oldval = v; \ + READ(oldval, t); \ + \ + ret = 0; \ +label_return: \ + if (l) \ + malloc_mutex_unlock(&ctl_mtx); \ + return (ret); \ +} + +#define CTL_RO_CGEN(c, n, v, t) \ +static int \ +n##_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp, \ + void *newp, size_t newlen) \ +{ \ + int ret; \ + t oldval; \ + \ + if ((c) == false) \ + return (ENOENT); \ + malloc_mutex_lock(&ctl_mtx); \ + READONLY(); \ + oldval = v; \ + READ(oldval, t); \ + \ + ret = 0; \ +label_return: \ + malloc_mutex_unlock(&ctl_mtx); \ + return (ret); \ +} + #define CTL_RO_GEN(n, v, t) \ static int \ n##_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp, \ @@ -1012,7 +929,7 @@ n##_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp, \ READ(oldval, t); \ \ ret = 0; \ -RETURN: \ +label_return: \ malloc_mutex_unlock(&ctl_mtx); \ return (ret); \ } @@ -1021,7 +938,26 @@ RETURN: \ * ctl_mtx is not acquired, under the assumption that no pertinent data will * mutate during the call. */ -#define CTL_RO_NL_GEN(n, v, t) \ +#define CTL_RO_NL_CGEN(c, n, v, t) \ +static int \ +n##_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp, \ + void *newp, size_t newlen) \ +{ \ + int ret; \ + t oldval; \ + \ + if ((c) == false) \ + return (ENOENT); \ + READONLY(); \ + oldval = v; \ + READ(oldval, t); \ + \ + ret = 0; \ +label_return: \ + return (ret); \ +} + +#define CTL_RO_NL_GEN(n, v, t) \ static int \ n##_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp, \ void *newp, size_t newlen) \ @@ -1034,11 +970,11 @@ n##_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp, \ READ(oldval, t); \ \ ret = 0; \ -RETURN: \ +label_return: \ return (ret); \ } -#define CTL_RO_TRUE_GEN(n) \ +#define CTL_RO_BOOL_CONFIG_GEN(n) \ static int \ n##_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp, \ void *newp, size_t newlen) \ @@ -1047,28 +983,11 @@ n##_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp, \ bool oldval; \ \ READONLY(); \ - oldval = true; \ + oldval = n; \ READ(oldval, bool); \ \ ret = 0; \ -RETURN: \ - return (ret); \ -} - -#define CTL_RO_FALSE_GEN(n) \ -static int \ -n##_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp, \ - void *newp, size_t newlen) \ -{ \ - int ret; \ - bool oldval; \ - \ - READONLY(); \ - oldval = false; \ - READ(oldval, bool); \ - \ - ret = 0; \ -RETURN: \ +label_return: \ return (ret); \ } @@ -1082,41 +1001,60 @@ epoch_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp, uint64_t newval; malloc_mutex_lock(&ctl_mtx); - newval = 0; WRITE(newval, uint64_t); - if (newval != 0) + if (newp != NULL) ctl_refresh(); READ(ctl_epoch, uint64_t); ret = 0; -RETURN: +label_return: malloc_mutex_unlock(&ctl_mtx); return (ret); } -#ifdef JEMALLOC_TCACHE static int -tcache_flush_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp, - void *newp, size_t newlen) +thread_tcache_enabled_ctl(const size_t *mib, size_t miblen, void *oldp, + size_t *oldlenp, void *newp, size_t newlen) { int ret; - tcache_t *tcache; + bool oldval; - VOID(); + if (config_tcache == false) + return (ENOENT); - tcache = TCACHE_GET(); - if (tcache == NULL) { - ret = 0; - goto RETURN; + oldval = tcache_enabled_get(); + if (newp != NULL) { + if (newlen != sizeof(bool)) { + ret = EINVAL; + goto label_return; + } + tcache_enabled_set(*(bool *)newp); } - tcache_destroy(tcache); - TCACHE_SET(NULL); + READ(oldval, bool); +label_return: ret = 0; -RETURN: return (ret); } -#endif + +static int +thread_tcache_flush_ctl(const size_t *mib, size_t miblen, void *oldp, + size_t *oldlenp, void *newp, size_t newlen) +{ + int ret; + + if (config_tcache == false) + return (ENOENT); + + READONLY(); + WRITEONLY(); + + tcache_flush(); + + ret = 0; +label_return: + return (ret); +} static int thread_arena_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp, @@ -1125,7 +1063,7 @@ thread_arena_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp, int ret; unsigned newind, oldind; - newind = oldind = choose_arena()->ind; + newind = oldind = choose_arena(NULL)->ind; WRITE(newind, unsigned); READ(oldind, unsigned); if (newind != oldind) { @@ -1134,191 +1072,108 @@ thread_arena_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp, if (newind >= narenas) { /* New arena index is out of range. */ ret = EFAULT; - goto RETURN; + goto label_return; } /* Initialize arena if necessary. */ malloc_mutex_lock(&arenas_lock); - if ((arena = arenas[newind]) == NULL) - arena = arenas_extend(newind); + if ((arena = arenas[newind]) == NULL && (arena = + arenas_extend(newind)) == NULL) { + malloc_mutex_unlock(&arenas_lock); + ret = EAGAIN; + goto label_return; + } + assert(arena == arenas[newind]); arenas[oldind]->nthreads--; arenas[newind]->nthreads++; malloc_mutex_unlock(&arenas_lock); - if (arena == NULL) { - ret = EAGAIN; - goto RETURN; - } /* Set new arena association. */ - ARENA_SET(arena); -#ifdef JEMALLOC_TCACHE - { - tcache_t *tcache = TCACHE_GET(); - if (tcache != NULL) - tcache->arena = arena; + if (config_tcache) { + tcache_t *tcache; + if ((uintptr_t)(tcache = *tcache_tsd_get()) > + (uintptr_t)TCACHE_STATE_MAX) { + tcache_arena_dissociate(tcache); + tcache_arena_associate(tcache, arena); + } } -#endif + arenas_tsd_set(&arena); } ret = 0; -RETURN: +label_return: return (ret); } -#ifdef JEMALLOC_STATS -CTL_RO_NL_GEN(thread_allocated, ALLOCATED_GET(), uint64_t); -CTL_RO_NL_GEN(thread_allocatedp, ALLOCATEDP_GET(), uint64_t *); -CTL_RO_NL_GEN(thread_deallocated, DEALLOCATED_GET(), uint64_t); -CTL_RO_NL_GEN(thread_deallocatedp, DEALLOCATEDP_GET(), uint64_t *); -#endif +CTL_RO_NL_CGEN(config_stats, thread_allocated, + thread_allocated_tsd_get()->allocated, uint64_t) +CTL_RO_NL_CGEN(config_stats, thread_allocatedp, + &thread_allocated_tsd_get()->allocated, uint64_t *) +CTL_RO_NL_CGEN(config_stats, thread_deallocated, + thread_allocated_tsd_get()->deallocated, uint64_t) +CTL_RO_NL_CGEN(config_stats, thread_deallocatedp, + &thread_allocated_tsd_get()->deallocated, uint64_t *) /******************************************************************************/ -#ifdef JEMALLOC_DEBUG -CTL_RO_TRUE_GEN(config_debug) -#else -CTL_RO_FALSE_GEN(config_debug) -#endif - -#ifdef JEMALLOC_DSS -CTL_RO_TRUE_GEN(config_dss) -#else -CTL_RO_FALSE_GEN(config_dss) -#endif - -#ifdef JEMALLOC_DYNAMIC_PAGE_SHIFT -CTL_RO_TRUE_GEN(config_dynamic_page_shift) -#else -CTL_RO_FALSE_GEN(config_dynamic_page_shift) -#endif - -#ifdef JEMALLOC_FILL -CTL_RO_TRUE_GEN(config_fill) -#else -CTL_RO_FALSE_GEN(config_fill) -#endif - -#ifdef JEMALLOC_LAZY_LOCK -CTL_RO_TRUE_GEN(config_lazy_lock) -#else -CTL_RO_FALSE_GEN(config_lazy_lock) -#endif - -#ifdef JEMALLOC_PROF -CTL_RO_TRUE_GEN(config_prof) -#else -CTL_RO_FALSE_GEN(config_prof) -#endif - -#ifdef JEMALLOC_PROF_LIBGCC -CTL_RO_TRUE_GEN(config_prof_libgcc) -#else -CTL_RO_FALSE_GEN(config_prof_libgcc) -#endif - -#ifdef JEMALLOC_PROF_LIBUNWIND -CTL_RO_TRUE_GEN(config_prof_libunwind) -#else -CTL_RO_FALSE_GEN(config_prof_libunwind) -#endif - -#ifdef JEMALLOC_STATS -CTL_RO_TRUE_GEN(config_stats) -#else -CTL_RO_FALSE_GEN(config_stats) -#endif - -#ifdef JEMALLOC_SWAP -CTL_RO_TRUE_GEN(config_swap) -#else -CTL_RO_FALSE_GEN(config_swap) -#endif - -#ifdef JEMALLOC_SYSV -CTL_RO_TRUE_GEN(config_sysv) -#else -CTL_RO_FALSE_GEN(config_sysv) -#endif - -#ifdef JEMALLOC_TCACHE -CTL_RO_TRUE_GEN(config_tcache) -#else -CTL_RO_FALSE_GEN(config_tcache) -#endif - -#ifdef JEMALLOC_TINY -CTL_RO_TRUE_GEN(config_tiny) -#else -CTL_RO_FALSE_GEN(config_tiny) -#endif - -#ifdef JEMALLOC_TLS -CTL_RO_TRUE_GEN(config_tls) -#else -CTL_RO_FALSE_GEN(config_tls) -#endif - -#ifdef JEMALLOC_XMALLOC -CTL_RO_TRUE_GEN(config_xmalloc) -#else -CTL_RO_FALSE_GEN(config_xmalloc) -#endif +CTL_RO_BOOL_CONFIG_GEN(config_debug) +CTL_RO_BOOL_CONFIG_GEN(config_dss) +CTL_RO_BOOL_CONFIG_GEN(config_fill) +CTL_RO_BOOL_CONFIG_GEN(config_lazy_lock) +CTL_RO_BOOL_CONFIG_GEN(config_mremap) +CTL_RO_BOOL_CONFIG_GEN(config_munmap) +CTL_RO_BOOL_CONFIG_GEN(config_prof) +CTL_RO_BOOL_CONFIG_GEN(config_prof_libgcc) +CTL_RO_BOOL_CONFIG_GEN(config_prof_libunwind) +CTL_RO_BOOL_CONFIG_GEN(config_stats) +CTL_RO_BOOL_CONFIG_GEN(config_tcache) +CTL_RO_BOOL_CONFIG_GEN(config_tls) +CTL_RO_BOOL_CONFIG_GEN(config_utrace) +CTL_RO_BOOL_CONFIG_GEN(config_valgrind) +CTL_RO_BOOL_CONFIG_GEN(config_xmalloc) /******************************************************************************/ CTL_RO_NL_GEN(opt_abort, opt_abort, bool) -CTL_RO_NL_GEN(opt_lg_qspace_max, opt_lg_qspace_max, size_t) -CTL_RO_NL_GEN(opt_lg_cspace_max, opt_lg_cspace_max, size_t) CTL_RO_NL_GEN(opt_lg_chunk, opt_lg_chunk, size_t) CTL_RO_NL_GEN(opt_narenas, opt_narenas, size_t) CTL_RO_NL_GEN(opt_lg_dirty_mult, opt_lg_dirty_mult, ssize_t) CTL_RO_NL_GEN(opt_stats_print, opt_stats_print, bool) -#ifdef JEMALLOC_FILL -CTL_RO_NL_GEN(opt_junk, opt_junk, bool) -CTL_RO_NL_GEN(opt_zero, opt_zero, bool) -#endif -#ifdef JEMALLOC_SYSV -CTL_RO_NL_GEN(opt_sysv, opt_sysv, bool) -#endif -#ifdef JEMALLOC_XMALLOC -CTL_RO_NL_GEN(opt_xmalloc, opt_xmalloc, bool) -#endif -#ifdef JEMALLOC_TCACHE -CTL_RO_NL_GEN(opt_tcache, opt_tcache, bool) -CTL_RO_NL_GEN(opt_lg_tcache_gc_sweep, opt_lg_tcache_gc_sweep, ssize_t) -#endif -#ifdef JEMALLOC_PROF -CTL_RO_NL_GEN(opt_prof, opt_prof, bool) -CTL_RO_NL_GEN(opt_prof_prefix, opt_prof_prefix, const char *) -CTL_RO_GEN(opt_prof_active, opt_prof_active, bool) /* Mutable. */ -CTL_RO_NL_GEN(opt_lg_prof_bt_max, opt_lg_prof_bt_max, size_t) -CTL_RO_NL_GEN(opt_lg_prof_sample, opt_lg_prof_sample, size_t) -CTL_RO_NL_GEN(opt_lg_prof_interval, opt_lg_prof_interval, ssize_t) -CTL_RO_NL_GEN(opt_prof_gdump, opt_prof_gdump, bool) -CTL_RO_NL_GEN(opt_prof_leak, opt_prof_leak, bool) -CTL_RO_NL_GEN(opt_prof_accum, opt_prof_accum, bool) -CTL_RO_NL_GEN(opt_lg_prof_tcmax, opt_lg_prof_tcmax, ssize_t) -#endif -#ifdef JEMALLOC_SWAP -CTL_RO_NL_GEN(opt_overcommit, opt_overcommit, bool) -#endif +CTL_RO_NL_CGEN(config_fill, opt_junk, opt_junk, bool) +CTL_RO_NL_CGEN(config_fill, opt_zero, opt_zero, bool) +CTL_RO_NL_CGEN(config_fill, opt_quarantine, opt_quarantine, size_t) +CTL_RO_NL_CGEN(config_fill, opt_redzone, opt_redzone, bool) +CTL_RO_NL_CGEN(config_utrace, opt_utrace, opt_utrace, bool) +CTL_RO_NL_CGEN(config_valgrind, opt_valgrind, opt_valgrind, bool) +CTL_RO_NL_CGEN(config_xmalloc, opt_xmalloc, opt_xmalloc, bool) +CTL_RO_NL_CGEN(config_tcache, opt_tcache, opt_tcache, bool) +CTL_RO_NL_CGEN(config_tcache, opt_lg_tcache_max, opt_lg_tcache_max, ssize_t) +CTL_RO_NL_CGEN(config_prof, opt_prof, opt_prof, bool) +CTL_RO_NL_CGEN(config_prof, opt_prof_prefix, opt_prof_prefix, const char *) +CTL_RO_CGEN(config_prof, opt_prof_active, opt_prof_active, bool) /* Mutable. */ +CTL_RO_NL_CGEN(config_prof, opt_lg_prof_sample, opt_lg_prof_sample, size_t) +CTL_RO_NL_CGEN(config_prof, opt_lg_prof_interval, opt_lg_prof_interval, ssize_t) +CTL_RO_NL_CGEN(config_prof, opt_prof_gdump, opt_prof_gdump, bool) +CTL_RO_NL_CGEN(config_prof, opt_prof_final, opt_prof_final, bool) +CTL_RO_NL_CGEN(config_prof, opt_prof_leak, opt_prof_leak, bool) +CTL_RO_NL_CGEN(config_prof, opt_prof_accum, opt_prof_accum, bool) /******************************************************************************/ CTL_RO_NL_GEN(arenas_bin_i_size, arena_bin_info[mib[2]].reg_size, size_t) CTL_RO_NL_GEN(arenas_bin_i_nregs, arena_bin_info[mib[2]].nregs, uint32_t) CTL_RO_NL_GEN(arenas_bin_i_run_size, arena_bin_info[mib[2]].run_size, size_t) -const ctl_node_t * +const ctl_named_node_t * arenas_bin_i_index(const size_t *mib, size_t miblen, size_t i) { - if (i > nbins) + if (i > NBINS) return (NULL); return (super_arenas_bin_i_node); } -CTL_RO_NL_GEN(arenas_lrun_i_size, ((mib[2]+1) << PAGE_SHIFT), size_t) -const ctl_node_t * +CTL_RO_NL_GEN(arenas_lrun_i_size, ((mib[2]+1) << LG_PAGE), size_t) +const ctl_named_node_t * arenas_lrun_i_index(const size_t *mib, size_t miblen, size_t i) { @@ -1350,37 +1205,16 @@ arenas_initialized_ctl(const size_t *mib, size_t miblen, void *oldp, for (i = 0; i < nread; i++) ((bool *)oldp)[i] = ctl_stats.arenas[i].initialized; -RETURN: +label_return: malloc_mutex_unlock(&ctl_mtx); return (ret); } CTL_RO_NL_GEN(arenas_quantum, QUANTUM, size_t) -CTL_RO_NL_GEN(arenas_cacheline, CACHELINE, size_t) -CTL_RO_NL_GEN(arenas_subpage, SUBPAGE, size_t) -CTL_RO_NL_GEN(arenas_pagesize, PAGE_SIZE, size_t) -CTL_RO_NL_GEN(arenas_chunksize, chunksize, size_t) -#ifdef JEMALLOC_TINY -CTL_RO_NL_GEN(arenas_tspace_min, (1U << LG_TINY_MIN), size_t) -CTL_RO_NL_GEN(arenas_tspace_max, (qspace_min >> 1), size_t) -#endif -CTL_RO_NL_GEN(arenas_qspace_min, qspace_min, size_t) -CTL_RO_NL_GEN(arenas_qspace_max, qspace_max, size_t) -CTL_RO_NL_GEN(arenas_cspace_min, cspace_min, size_t) -CTL_RO_NL_GEN(arenas_cspace_max, cspace_max, size_t) -CTL_RO_NL_GEN(arenas_sspace_min, sspace_min, size_t) -CTL_RO_NL_GEN(arenas_sspace_max, sspace_max, size_t) -#ifdef JEMALLOC_TCACHE -CTL_RO_NL_GEN(arenas_tcache_max, tcache_maxclass, size_t) -#endif -CTL_RO_NL_GEN(arenas_ntbins, ntbins, unsigned) -CTL_RO_NL_GEN(arenas_nqbins, nqbins, unsigned) -CTL_RO_NL_GEN(arenas_ncbins, ncbins, unsigned) -CTL_RO_NL_GEN(arenas_nsbins, nsbins, unsigned) -CTL_RO_NL_GEN(arenas_nbins, nbins, unsigned) -#ifdef JEMALLOC_TCACHE -CTL_RO_NL_GEN(arenas_nhbins, nhbins, unsigned) -#endif +CTL_RO_NL_GEN(arenas_page, PAGE, size_t) +CTL_RO_NL_CGEN(config_tcache, arenas_tcache_max, tcache_maxclass, size_t) +CTL_RO_NL_GEN(arenas_nbins, NBINS, unsigned) +CTL_RO_NL_CGEN(config_tcache, arenas_nhbins, nhbins, unsigned) CTL_RO_NL_GEN(arenas_nlruns, nlclasses, size_t) static int @@ -1395,9 +1229,9 @@ arenas_purge_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp, WRITE(arena, unsigned); if (newp != NULL && arena >= narenas) { ret = EFAULT; - goto RETURN; + goto label_return; } else { - arena_t *tarenas[narenas]; + VARIABLE_ARRAY(arena_t *, tarenas, narenas); malloc_mutex_lock(&arenas_lock); memcpy(tarenas, arenas, sizeof(arena_t *) * narenas); @@ -1417,13 +1251,12 @@ arenas_purge_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp, } ret = 0; -RETURN: +label_return: return (ret); } /******************************************************************************/ -#ifdef JEMALLOC_PROF static int prof_active_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp, void *newp, size_t newlen) @@ -1431,6 +1264,9 @@ prof_active_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp, int ret; bool oldval; + if (config_prof == false) + return (ENOENT); + malloc_mutex_lock(&ctl_mtx); /* Protect opt_prof_active. */ oldval = opt_prof_active; if (newp != NULL) { @@ -1445,7 +1281,7 @@ prof_active_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp, READ(oldval, bool); ret = 0; -RETURN: +label_return: malloc_mutex_unlock(&ctl_mtx); return (ret); } @@ -1457,92 +1293,88 @@ prof_dump_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp, int ret; const char *filename = NULL; + if (config_prof == false) + return (ENOENT); + WRITEONLY(); WRITE(filename, const char *); if (prof_mdump(filename)) { ret = EFAULT; - goto RETURN; + goto label_return; } ret = 0; -RETURN: +label_return: return (ret); } -CTL_RO_NL_GEN(prof_interval, prof_interval, uint64_t) -#endif +CTL_RO_NL_CGEN(config_prof, prof_interval, prof_interval, uint64_t) /******************************************************************************/ -#ifdef JEMALLOC_STATS -CTL_RO_GEN(stats_chunks_current, ctl_stats.chunks.current, size_t) -CTL_RO_GEN(stats_chunks_total, ctl_stats.chunks.total, uint64_t) -CTL_RO_GEN(stats_chunks_high, ctl_stats.chunks.high, size_t) -CTL_RO_GEN(stats_huge_allocated, huge_allocated, size_t) -CTL_RO_GEN(stats_huge_nmalloc, huge_nmalloc, uint64_t) -CTL_RO_GEN(stats_huge_ndalloc, huge_ndalloc, uint64_t) -CTL_RO_GEN(stats_arenas_i_small_allocated, +CTL_RO_CGEN(config_stats, stats_chunks_current, ctl_stats.chunks.current, + size_t) +CTL_RO_CGEN(config_stats, stats_chunks_total, ctl_stats.chunks.total, uint64_t) +CTL_RO_CGEN(config_stats, stats_chunks_high, ctl_stats.chunks.high, size_t) +CTL_RO_CGEN(config_stats, stats_huge_allocated, huge_allocated, size_t) +CTL_RO_CGEN(config_stats, stats_huge_nmalloc, huge_nmalloc, uint64_t) +CTL_RO_CGEN(config_stats, stats_huge_ndalloc, huge_ndalloc, uint64_t) +CTL_RO_CGEN(config_stats, stats_arenas_i_small_allocated, ctl_stats.arenas[mib[2]].allocated_small, size_t) -CTL_RO_GEN(stats_arenas_i_small_nmalloc, +CTL_RO_CGEN(config_stats, stats_arenas_i_small_nmalloc, ctl_stats.arenas[mib[2]].nmalloc_small, uint64_t) -CTL_RO_GEN(stats_arenas_i_small_ndalloc, +CTL_RO_CGEN(config_stats, stats_arenas_i_small_ndalloc, ctl_stats.arenas[mib[2]].ndalloc_small, uint64_t) -CTL_RO_GEN(stats_arenas_i_small_nrequests, +CTL_RO_CGEN(config_stats, stats_arenas_i_small_nrequests, ctl_stats.arenas[mib[2]].nrequests_small, uint64_t) -CTL_RO_GEN(stats_arenas_i_large_allocated, +CTL_RO_CGEN(config_stats, stats_arenas_i_large_allocated, ctl_stats.arenas[mib[2]].astats.allocated_large, size_t) -CTL_RO_GEN(stats_arenas_i_large_nmalloc, +CTL_RO_CGEN(config_stats, stats_arenas_i_large_nmalloc, ctl_stats.arenas[mib[2]].astats.nmalloc_large, uint64_t) -CTL_RO_GEN(stats_arenas_i_large_ndalloc, +CTL_RO_CGEN(config_stats, stats_arenas_i_large_ndalloc, ctl_stats.arenas[mib[2]].astats.ndalloc_large, uint64_t) -CTL_RO_GEN(stats_arenas_i_large_nrequests, +CTL_RO_CGEN(config_stats, stats_arenas_i_large_nrequests, ctl_stats.arenas[mib[2]].astats.nrequests_large, uint64_t) -CTL_RO_GEN(stats_arenas_i_bins_j_allocated, +CTL_RO_CGEN(config_stats, stats_arenas_i_bins_j_allocated, ctl_stats.arenas[mib[2]].bstats[mib[4]].allocated, size_t) -CTL_RO_GEN(stats_arenas_i_bins_j_nmalloc, +CTL_RO_CGEN(config_stats, stats_arenas_i_bins_j_nmalloc, ctl_stats.arenas[mib[2]].bstats[mib[4]].nmalloc, uint64_t) -CTL_RO_GEN(stats_arenas_i_bins_j_ndalloc, +CTL_RO_CGEN(config_stats, stats_arenas_i_bins_j_ndalloc, ctl_stats.arenas[mib[2]].bstats[mib[4]].ndalloc, uint64_t) -CTL_RO_GEN(stats_arenas_i_bins_j_nrequests, +CTL_RO_CGEN(config_stats, stats_arenas_i_bins_j_nrequests, ctl_stats.arenas[mib[2]].bstats[mib[4]].nrequests, uint64_t) -#ifdef JEMALLOC_TCACHE -CTL_RO_GEN(stats_arenas_i_bins_j_nfills, +CTL_RO_CGEN(config_stats && config_tcache, stats_arenas_i_bins_j_nfills, ctl_stats.arenas[mib[2]].bstats[mib[4]].nfills, uint64_t) -CTL_RO_GEN(stats_arenas_i_bins_j_nflushes, +CTL_RO_CGEN(config_stats && config_tcache, stats_arenas_i_bins_j_nflushes, ctl_stats.arenas[mib[2]].bstats[mib[4]].nflushes, uint64_t) -#endif -CTL_RO_GEN(stats_arenas_i_bins_j_nruns, +CTL_RO_CGEN(config_stats, stats_arenas_i_bins_j_nruns, ctl_stats.arenas[mib[2]].bstats[mib[4]].nruns, uint64_t) -CTL_RO_GEN(stats_arenas_i_bins_j_nreruns, +CTL_RO_CGEN(config_stats, stats_arenas_i_bins_j_nreruns, ctl_stats.arenas[mib[2]].bstats[mib[4]].reruns, uint64_t) -CTL_RO_GEN(stats_arenas_i_bins_j_highruns, - ctl_stats.arenas[mib[2]].bstats[mib[4]].highruns, size_t) -CTL_RO_GEN(stats_arenas_i_bins_j_curruns, +CTL_RO_CGEN(config_stats, stats_arenas_i_bins_j_curruns, ctl_stats.arenas[mib[2]].bstats[mib[4]].curruns, size_t) -const ctl_node_t * +const ctl_named_node_t * stats_arenas_i_bins_j_index(const size_t *mib, size_t miblen, size_t j) { - if (j > nbins) + if (j > NBINS) return (NULL); return (super_stats_arenas_i_bins_j_node); } -CTL_RO_GEN(stats_arenas_i_lruns_j_nmalloc, +CTL_RO_CGEN(config_stats, stats_arenas_i_lruns_j_nmalloc, ctl_stats.arenas[mib[2]].lstats[mib[4]].nmalloc, uint64_t) -CTL_RO_GEN(stats_arenas_i_lruns_j_ndalloc, +CTL_RO_CGEN(config_stats, stats_arenas_i_lruns_j_ndalloc, ctl_stats.arenas[mib[2]].lstats[mib[4]].ndalloc, uint64_t) -CTL_RO_GEN(stats_arenas_i_lruns_j_nrequests, +CTL_RO_CGEN(config_stats, stats_arenas_i_lruns_j_nrequests, ctl_stats.arenas[mib[2]].lstats[mib[4]].nrequests, uint64_t) -CTL_RO_GEN(stats_arenas_i_lruns_j_curruns, +CTL_RO_CGEN(config_stats, stats_arenas_i_lruns_j_curruns, ctl_stats.arenas[mib[2]].lstats[mib[4]].curruns, size_t) -CTL_RO_GEN(stats_arenas_i_lruns_j_highruns, - ctl_stats.arenas[mib[2]].lstats[mib[4]].highruns, size_t) -const ctl_node_t * +const ctl_named_node_t * stats_arenas_i_lruns_j_index(const size_t *mib, size_t miblen, size_t j) { @@ -1551,120 +1383,36 @@ stats_arenas_i_lruns_j_index(const size_t *mib, size_t miblen, size_t j) return (super_stats_arenas_i_lruns_j_node); } -#endif CTL_RO_GEN(stats_arenas_i_nthreads, ctl_stats.arenas[mib[2]].nthreads, unsigned) CTL_RO_GEN(stats_arenas_i_pactive, ctl_stats.arenas[mib[2]].pactive, size_t) CTL_RO_GEN(stats_arenas_i_pdirty, ctl_stats.arenas[mib[2]].pdirty, size_t) -#ifdef JEMALLOC_STATS -CTL_RO_GEN(stats_arenas_i_mapped, ctl_stats.arenas[mib[2]].astats.mapped, - size_t) -CTL_RO_GEN(stats_arenas_i_npurge, ctl_stats.arenas[mib[2]].astats.npurge, - uint64_t) -CTL_RO_GEN(stats_arenas_i_nmadvise, ctl_stats.arenas[mib[2]].astats.nmadvise, - uint64_t) -CTL_RO_GEN(stats_arenas_i_purged, ctl_stats.arenas[mib[2]].astats.purged, - uint64_t) -#endif +CTL_RO_CGEN(config_stats, stats_arenas_i_mapped, + ctl_stats.arenas[mib[2]].astats.mapped, size_t) +CTL_RO_CGEN(config_stats, stats_arenas_i_npurge, + ctl_stats.arenas[mib[2]].astats.npurge, uint64_t) +CTL_RO_CGEN(config_stats, stats_arenas_i_nmadvise, + ctl_stats.arenas[mib[2]].astats.nmadvise, uint64_t) +CTL_RO_CGEN(config_stats, stats_arenas_i_purged, + ctl_stats.arenas[mib[2]].astats.purged, uint64_t) -const ctl_node_t * +const ctl_named_node_t * stats_arenas_i_index(const size_t *mib, size_t miblen, size_t i) { - const ctl_node_t * ret; + const ctl_named_node_t * ret; malloc_mutex_lock(&ctl_mtx); if (ctl_stats.arenas[i].initialized == false) { ret = NULL; - goto RETURN; + goto label_return; } ret = super_stats_arenas_i_node; -RETURN: +label_return: malloc_mutex_unlock(&ctl_mtx); return (ret); } -#ifdef JEMALLOC_STATS -CTL_RO_GEN(stats_cactive, &stats_cactive, size_t *) -CTL_RO_GEN(stats_allocated, ctl_stats.allocated, size_t) -CTL_RO_GEN(stats_active, ctl_stats.active, size_t) -CTL_RO_GEN(stats_mapped, ctl_stats.mapped, size_t) -#endif - -/******************************************************************************/ - -#ifdef JEMALLOC_SWAP -# ifdef JEMALLOC_STATS -CTL_RO_GEN(swap_avail, ctl_stats.swap_avail, size_t) -# endif - -static int -swap_prezeroed_ctl(const size_t *mib, size_t miblen, void *oldp, - size_t *oldlenp, void *newp, size_t newlen) -{ - int ret; - - malloc_mutex_lock(&ctl_mtx); - if (swap_enabled) { - READONLY(); - } else { - /* - * swap_prezeroed isn't actually used by the swap code until it - * is set during a successful chunk_swap_enabled() call. We - * use it here to store the value that we'll pass to - * chunk_swap_enable() in a swap.fds mallctl(). This is not - * very clean, but the obvious alternatives are even worse. - */ - WRITE(swap_prezeroed, bool); - } - - READ(swap_prezeroed, bool); - - ret = 0; -RETURN: - malloc_mutex_unlock(&ctl_mtx); - return (ret); -} - -CTL_RO_GEN(swap_nfds, swap_nfds, size_t) - -static int -swap_fds_ctl(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp, - void *newp, size_t newlen) -{ - int ret; - - malloc_mutex_lock(&ctl_mtx); - if (swap_enabled) { - READONLY(); - } else if (newp != NULL) { - size_t nfds = newlen / sizeof(int); - - { - int fds[nfds]; - - memcpy(fds, newp, nfds * sizeof(int)); - if (chunk_swap_enable(fds, nfds, swap_prezeroed)) { - ret = EFAULT; - goto RETURN; - } - } - } - - if (oldp != NULL && oldlenp != NULL) { - if (*oldlenp != swap_nfds * sizeof(int)) { - size_t copylen = (swap_nfds * sizeof(int) <= *oldlenp) - ? swap_nfds * sizeof(int) : *oldlenp; - - memcpy(oldp, swap_fds, copylen); - ret = EINVAL; - goto RETURN; - } else - memcpy(oldp, swap_fds, *oldlenp); - } - - ret = 0; -RETURN: - malloc_mutex_unlock(&ctl_mtx); - return (ret); -} -#endif +CTL_RO_CGEN(config_stats, stats_cactive, &stats_cactive, size_t *) +CTL_RO_CGEN(config_stats, stats_allocated, ctl_stats.allocated, size_t) +CTL_RO_CGEN(config_stats, stats_active, ctl_stats.active, size_t) +CTL_RO_CGEN(config_stats, stats_mapped, ctl_stats.mapped, size_t) diff --git a/src/extent.c b/src/extent.c index 3c04d3aa..8c09b486 100644 --- a/src/extent.c +++ b/src/extent.c @@ -3,7 +3,6 @@ /******************************************************************************/ -#if (defined(JEMALLOC_SWAP) || defined(JEMALLOC_DSS)) static inline int extent_szad_comp(extent_node_t *a, extent_node_t *b) { @@ -25,7 +24,6 @@ extent_szad_comp(extent_node_t *a, extent_node_t *b) /* Generate red-black tree functions. */ rb_gen(, extent_tree_szad_, extent_tree_t, extent_node_t, link_szad, extent_szad_comp) -#endif static inline int extent_ad_comp(extent_node_t *a, extent_node_t *b) diff --git a/src/huge.c b/src/huge.c index a4f9b054..8a4ec942 100644 --- a/src/huge.c +++ b/src/huge.c @@ -4,11 +4,9 @@ /******************************************************************************/ /* Data. */ -#ifdef JEMALLOC_STATS uint64_t huge_nmalloc; uint64_t huge_ndalloc; size_t huge_allocated; -#endif malloc_mutex_t huge_mtx; @@ -19,10 +17,18 @@ static extent_tree_t huge; void * huge_malloc(size_t size, bool zero) +{ + + return (huge_palloc(size, chunksize, zero)); +} + +void * +huge_palloc(size_t size, size_t alignment, bool zero) { void *ret; size_t csize; extent_node_t *node; + bool is_zeroed; /* Allocate one or more contiguous chunks for this request. */ @@ -37,7 +43,12 @@ huge_malloc(size_t size, bool zero) if (node == NULL) return (NULL); - ret = chunk_alloc(csize, false, &zero); + /* + * Copy zero into is_zeroed and pass the copy to chunk_alloc(), so that + * it is possible to make correct junk/zero fill decisions below. + */ + is_zeroed = zero; + ret = chunk_alloc(csize, alignment, false, &is_zeroed); if (ret == NULL) { base_node_dealloc(node); return (NULL); @@ -49,106 +60,19 @@ huge_malloc(size_t size, bool zero) malloc_mutex_lock(&huge_mtx); extent_tree_ad_insert(&huge, node); -#ifdef JEMALLOC_STATS - stats_cactive_add(csize); - huge_nmalloc++; - huge_allocated += csize; -#endif + if (config_stats) { + stats_cactive_add(csize); + huge_nmalloc++; + huge_allocated += csize; + } malloc_mutex_unlock(&huge_mtx); -#ifdef JEMALLOC_FILL - if (zero == false) { + if (config_fill && zero == false) { if (opt_junk) memset(ret, 0xa5, csize); - else if (opt_zero) + else if (opt_zero && is_zeroed == false) memset(ret, 0, csize); } -#endif - - return (ret); -} - -/* Only handles large allocations that require more than chunk alignment. */ -void * -huge_palloc(size_t size, size_t alignment, bool zero) -{ - void *ret; - size_t alloc_size, chunk_size, offset; - extent_node_t *node; - - /* - * This allocation requires alignment that is even larger than chunk - * alignment. This means that huge_malloc() isn't good enough. - * - * Allocate almost twice as many chunks as are demanded by the size or - * alignment, in order to assure the alignment can be achieved, then - * unmap leading and trailing chunks. - */ - assert(alignment > chunksize); - - chunk_size = CHUNK_CEILING(size); - - if (size >= alignment) - alloc_size = chunk_size + alignment - chunksize; - else - alloc_size = (alignment << 1) - chunksize; - - /* Allocate an extent node with which to track the chunk. */ - node = base_node_alloc(); - if (node == NULL) - return (NULL); - - ret = chunk_alloc(alloc_size, false, &zero); - if (ret == NULL) { - base_node_dealloc(node); - return (NULL); - } - - offset = (uintptr_t)ret & (alignment - 1); - assert((offset & chunksize_mask) == 0); - assert(offset < alloc_size); - if (offset == 0) { - /* Trim trailing space. */ - chunk_dealloc((void *)((uintptr_t)ret + chunk_size), alloc_size - - chunk_size, true); - } else { - size_t trailsize; - - /* Trim leading space. */ - chunk_dealloc(ret, alignment - offset, true); - - ret = (void *)((uintptr_t)ret + (alignment - offset)); - - trailsize = alloc_size - (alignment - offset) - chunk_size; - if (trailsize != 0) { - /* Trim trailing space. */ - assert(trailsize < alloc_size); - chunk_dealloc((void *)((uintptr_t)ret + chunk_size), - trailsize, true); - } - } - - /* Insert node into huge. */ - node->addr = ret; - node->size = chunk_size; - - malloc_mutex_lock(&huge_mtx); - extent_tree_ad_insert(&huge, node); -#ifdef JEMALLOC_STATS - stats_cactive_add(chunk_size); - huge_nmalloc++; - huge_allocated += chunk_size; -#endif - malloc_mutex_unlock(&huge_mtx); - -#ifdef JEMALLOC_FILL - if (zero == false) { - if (opt_junk) - memset(ret, 0xa5, chunk_size); - else if (opt_zero) - memset(ret, 0, chunk_size); - } -#endif return (ret); } @@ -164,12 +88,10 @@ huge_ralloc_no_move(void *ptr, size_t oldsize, size_t size, size_t extra) && CHUNK_CEILING(oldsize) >= CHUNK_CEILING(size) && CHUNK_CEILING(oldsize) <= CHUNK_CEILING(size+extra)) { assert(CHUNK_CEILING(oldsize) == oldsize); -#ifdef JEMALLOC_FILL - if (opt_junk && size < oldsize) { + if (config_fill && opt_junk && size < oldsize) { memset((void *)((uintptr_t)ptr + size), 0x5a, oldsize - size); } -#endif return (ptr); } @@ -218,20 +140,13 @@ huge_ralloc(void *ptr, size_t oldsize, size_t size, size_t extra, */ copysize = (size < oldsize) ? size : oldsize; +#ifdef JEMALLOC_MREMAP /* * Use mremap(2) if this is a huge-->huge reallocation, and neither the - * source nor the destination are in swap or dss. + * source nor the destination are in dss. */ -#ifdef JEMALLOC_MREMAP_FIXED - if (oldsize >= chunksize -# ifdef JEMALLOC_SWAP - && (swap_enabled == false || (chunk_in_swap(ptr) == false && - chunk_in_swap(ret) == false)) -# endif -# ifdef JEMALLOC_DSS - && chunk_in_dss(ptr) == false && chunk_in_dss(ret) == false -# endif - ) { + if (oldsize >= chunksize && (config_dss == false || (chunk_in_dss(ptr) + == false && chunk_in_dss(ret) == false))) { size_t newsize = huge_salloc(ret); /* @@ -253,10 +168,9 @@ huge_ralloc(void *ptr, size_t oldsize, size_t size, size_t extra, */ char buf[BUFERROR_BUF]; - buferror(errno, buf, sizeof(buf)); - malloc_write(": Error in mremap(): "); - malloc_write(buf); - malloc_write("\n"); + buferror(buf, sizeof(buf)); + malloc_printf(": Error in mremap(): %s\n", + buf); if (opt_abort) abort(); memcpy(ret, ptr, copysize); @@ -266,7 +180,7 @@ huge_ralloc(void *ptr, size_t oldsize, size_t size, size_t extra, #endif { memcpy(ret, ptr, copysize); - idalloc(ptr); + iqalloc(ptr); } return (ret); } @@ -285,23 +199,16 @@ huge_dalloc(void *ptr, bool unmap) assert(node->addr == ptr); extent_tree_ad_remove(&huge, node); -#ifdef JEMALLOC_STATS - stats_cactive_sub(node->size); - huge_ndalloc++; - huge_allocated -= node->size; -#endif + if (config_stats) { + stats_cactive_sub(node->size); + huge_ndalloc++; + huge_allocated -= node->size; + } malloc_mutex_unlock(&huge_mtx); - if (unmap) { - /* Unmap chunk. */ -#ifdef JEMALLOC_FILL -#if (defined(JEMALLOC_SWAP) || defined(JEMALLOC_DSS)) - if (opt_junk) - memset(node->addr, 0x5a, node->size); -#endif -#endif - } + if (unmap && config_fill && config_dss && opt_junk) + memset(node->addr, 0x5a, node->size); chunk_dealloc(node->addr, node->size, unmap); @@ -328,7 +235,6 @@ huge_salloc(const void *ptr) return (ret); } -#ifdef JEMALLOC_PROF prof_ctx_t * huge_prof_ctx_get(const void *ptr) { @@ -365,7 +271,6 @@ huge_prof_ctx_set(const void *ptr, prof_ctx_t *ctx) malloc_mutex_unlock(&huge_mtx); } -#endif bool huge_boot(void) @@ -376,11 +281,32 @@ huge_boot(void) return (true); extent_tree_ad_new(&huge); -#ifdef JEMALLOC_STATS - huge_nmalloc = 0; - huge_ndalloc = 0; - huge_allocated = 0; -#endif + if (config_stats) { + huge_nmalloc = 0; + huge_ndalloc = 0; + huge_allocated = 0; + } return (false); } + +void +huge_prefork(void) +{ + + malloc_mutex_prefork(&huge_mtx); +} + +void +huge_postfork_parent(void) +{ + + malloc_mutex_postfork_parent(&huge_mtx); +} + +void +huge_postfork_child(void) +{ + + malloc_mutex_postfork_child(&huge_mtx); +} diff --git a/src/jemalloc.c b/src/jemalloc.c index a161c2e2..bc54cd7c 100644 --- a/src/jemalloc.c +++ b/src/jemalloc.c @@ -4,111 +4,108 @@ /******************************************************************************/ /* Data. */ -malloc_mutex_t arenas_lock; -arena_t **arenas; -unsigned narenas; - -pthread_key_t arenas_tsd; -#ifndef NO_TLS -__thread arena_t *arenas_tls JEMALLOC_ATTR(tls_model("initial-exec")); -#endif - -#ifdef JEMALLOC_STATS -# ifndef NO_TLS -__thread thread_allocated_t thread_allocated_tls; -# else -pthread_key_t thread_allocated_tsd; -# endif -#endif - -/* Set to true once the allocator has been initialized. */ -static bool malloc_initialized = false; - -/* Used to let the initializing thread recursively allocate. */ -static pthread_t malloc_initializer = (unsigned long)0; - -/* Used to avoid initialization races. */ -static malloc_mutex_t init_lock = -#ifdef JEMALLOC_OSSPIN - 0 -#else - MALLOC_MUTEX_INITIALIZER -#endif - ; - -#ifdef DYNAMIC_PAGE_SHIFT -size_t pagesize; -size_t pagesize_mask; -size_t lg_pagesize; -#endif - -unsigned ncpus; +malloc_tsd_data(, arenas, arena_t *, NULL) +malloc_tsd_data(, thread_allocated, thread_allocated_t, + THREAD_ALLOCATED_INITIALIZER) /* Runtime configuration options. */ -const char *JEMALLOC_P(malloc_conf) JEMALLOC_ATTR(visibility("default")); +const char *je_malloc_conf; #ifdef JEMALLOC_DEBUG bool opt_abort = true; # ifdef JEMALLOC_FILL bool opt_junk = true; +# else +bool opt_junk = false; # endif #else bool opt_abort = false; -# ifdef JEMALLOC_FILL bool opt_junk = false; -# endif #endif -#ifdef JEMALLOC_SYSV -bool opt_sysv = false; -#endif -#ifdef JEMALLOC_XMALLOC +size_t opt_quarantine = ZU(0); +bool opt_redzone = false; +bool opt_utrace = false; +bool opt_valgrind = false; bool opt_xmalloc = false; -#endif -#ifdef JEMALLOC_FILL bool opt_zero = false; -#endif size_t opt_narenas = 0; +unsigned ncpus; + +malloc_mutex_t arenas_lock; +arena_t **arenas; +unsigned narenas; + +/* Set to true once the allocator has been initialized. */ +static bool malloc_initialized = false; + +#ifdef JEMALLOC_THREADED_INIT +/* Used to let the initializing thread recursively allocate. */ +# define NO_INITIALIZER ((unsigned long)0) +# define INITIALIZER pthread_self() +# define IS_INITIALIZER (malloc_initializer == pthread_self()) +static pthread_t malloc_initializer = NO_INITIALIZER; +#else +# define NO_INITIALIZER false +# define INITIALIZER true +# define IS_INITIALIZER malloc_initializer +static bool malloc_initializer = NO_INITIALIZER; +#endif + +/* Used to avoid initialization races. */ +#ifdef _WIN32 +static malloc_mutex_t init_lock; + +JEMALLOC_ATTR(constructor) +static void WINAPI +_init_init_lock(void) +{ + + malloc_mutex_init(&init_lock); +} + +#ifdef _MSC_VER +# pragma section(".CRT$XCU", read) +JEMALLOC_SECTION(".CRT$XCU") JEMALLOC_ATTR(used) +static const void (WINAPI *init_init_lock)(void) = _init_init_lock; +#endif + +#else +static malloc_mutex_t init_lock = MALLOC_MUTEX_INITIALIZER; +#endif + +typedef struct { + void *p; /* Input pointer (as in realloc(p, s)). */ + size_t s; /* Request size. */ + void *r; /* Result pointer. */ +} malloc_utrace_t; + +#ifdef JEMALLOC_UTRACE +# define UTRACE(a, b, c) do { \ + if (opt_utrace) { \ + malloc_utrace_t ut; \ + ut.p = (a); \ + ut.s = (b); \ + ut.r = (c); \ + utrace(&ut, sizeof(ut)); \ + } \ +} while (0) +#else +# define UTRACE(a, b, c) +#endif + /******************************************************************************/ /* Function prototypes for non-inline static functions. */ -static void wrtmessage(void *cbopaque, const char *s); static void stats_print_atexit(void); static unsigned malloc_ncpus(void); -static void arenas_cleanup(void *arg); -#if (defined(JEMALLOC_STATS) && defined(NO_TLS)) -static void thread_allocated_cleanup(void *arg); -#endif static bool malloc_conf_next(char const **opts_p, char const **k_p, size_t *klen_p, char const **v_p, size_t *vlen_p); static void malloc_conf_error(const char *msg, const char *k, size_t klen, const char *v, size_t vlen); static void malloc_conf_init(void); static bool malloc_init_hard(void); -static int imemalign(void **memptr, size_t alignment, size_t size); - -/******************************************************************************/ -/* malloc_message() setup. */ - -#ifdef JEMALLOC_HAVE_ATTR -JEMALLOC_ATTR(visibility("hidden")) -#else -static -#endif -void -wrtmessage(void *cbopaque, const char *s) -{ -#ifdef JEMALLOC_CC_SILENCE - int result = -#endif - write(STDERR_FILENO, s, strlen(s)); -#ifdef JEMALLOC_CC_SILENCE - if (result < 0) - result = errno; -#endif -} - -void (*JEMALLOC_P(malloc_message))(void *, const char *s) - JEMALLOC_ATTR(visibility("default")) = wrtmessage; +static int imemalign(void **memptr, size_t alignment, size_t size, + size_t min_alignment); /******************************************************************************/ /* @@ -121,9 +118,7 @@ arenas_extend(unsigned ind) { arena_t *ret; - /* Allocate enough space for trailing bins. */ - ret = (arena_t *)base_alloc(offsetof(arena_t, bins) - + (sizeof(arena_bin_t) * nbins)); + ret = (arena_t *)base_alloc(sizeof(arena_t)); if (ret != NULL && arena_new(ret, ind) == false) { arenas[ind] = ret; return (ret); @@ -143,10 +138,7 @@ arenas_extend(unsigned ind) return (arenas[0]); } -/* - * Choose an arena based on a per-thread value (slow-path code only, called - * only by choose_arena()). - */ +/* Slow path, called only by choose_arena(). */ arena_t * choose_arena_hard(void) { @@ -182,7 +174,7 @@ choose_arena_hard(void) } } - if (arenas[choose] == 0 || first_null == narenas) { + if (arenas[choose]->nthreads == 0 || first_null == narenas) { /* * Use an unloaded arena, or the least loaded arena if * all arenas are already initialized. @@ -201,86 +193,47 @@ choose_arena_hard(void) malloc_mutex_unlock(&arenas_lock); } - ARENA_SET(ret); + arenas_tsd_set(&ret); return (ret); } -/* - * glibc provides a non-standard strerror_r() when _GNU_SOURCE is defined, so - * provide a wrapper. - */ -int -buferror(int errnum, char *buf, size_t buflen) -{ -#ifdef _GNU_SOURCE - char *b = strerror_r(errno, buf, buflen); - if (b != buf) { - strncpy(buf, b, buflen); - buf[buflen-1] = '\0'; - } - return (0); -#else - return (strerror_r(errno, buf, buflen)); -#endif -} - static void stats_print_atexit(void) { -#if (defined(JEMALLOC_TCACHE) && defined(JEMALLOC_STATS)) - unsigned i; + if (config_tcache && config_stats) { + unsigned i; - /* - * Merge stats from extant threads. This is racy, since individual - * threads do not lock when recording tcache stats events. As a - * consequence, the final stats may be slightly out of date by the time - * they are reported, if other threads continue to allocate. - */ - for (i = 0; i < narenas; i++) { - arena_t *arena = arenas[i]; - if (arena != NULL) { - tcache_t *tcache; + /* + * Merge stats from extant threads. This is racy, since + * individual threads do not lock when recording tcache stats + * events. As a consequence, the final stats may be slightly + * out of date by the time they are reported, if other threads + * continue to allocate. + */ + for (i = 0; i < narenas; i++) { + arena_t *arena = arenas[i]; + if (arena != NULL) { + tcache_t *tcache; - /* - * tcache_stats_merge() locks bins, so if any code is - * introduced that acquires both arena and bin locks in - * the opposite order, deadlocks may result. - */ - malloc_mutex_lock(&arena->lock); - ql_foreach(tcache, &arena->tcache_ql, link) { - tcache_stats_merge(tcache, arena); + /* + * tcache_stats_merge() locks bins, so if any + * code is introduced that acquires both arena + * and bin locks in the opposite order, + * deadlocks may result. + */ + malloc_mutex_lock(&arena->lock); + ql_foreach(tcache, &arena->tcache_ql, link) { + tcache_stats_merge(tcache, arena); + } + malloc_mutex_unlock(&arena->lock); } - malloc_mutex_unlock(&arena->lock); } } -#endif - JEMALLOC_P(malloc_stats_print)(NULL, NULL, NULL); + je_malloc_stats_print(NULL, NULL, NULL); } -#if (defined(JEMALLOC_STATS) && defined(NO_TLS)) -thread_allocated_t * -thread_allocated_get_hard(void) -{ - thread_allocated_t *thread_allocated = (thread_allocated_t *) - imalloc(sizeof(thread_allocated_t)); - if (thread_allocated == NULL) { - static thread_allocated_t static_thread_allocated = {0, 0}; - malloc_write(": Error allocating TSD;" - " mallctl(\"thread.{de,}allocated[p]\", ...)" - " will be inaccurate\n"); - if (opt_abort) - abort(); - return (&static_thread_allocated); - } - pthread_setspecific(thread_allocated_tsd, thread_allocated); - thread_allocated->allocated = 0; - thread_allocated->deallocated = 0; - return (thread_allocated); -} -#endif - /* * End miscellaneous support functions. */ @@ -295,42 +248,32 @@ malloc_ncpus(void) unsigned ret; long result; +#ifdef _WIN32 + SYSTEM_INFO si; + GetSystemInfo(&si); + result = si.dwNumberOfProcessors; +#else result = sysconf(_SC_NPROCESSORS_ONLN); if (result == -1) { /* Error. */ ret = 1; } +#endif ret = (unsigned)result; return (ret); } -static void +void arenas_cleanup(void *arg) { - arena_t *arena = (arena_t *)arg; + arena_t *arena = *(arena_t **)arg; malloc_mutex_lock(&arenas_lock); arena->nthreads--; malloc_mutex_unlock(&arenas_lock); } -#if (defined(JEMALLOC_STATS) && defined(NO_TLS)) -static void -thread_allocated_cleanup(void *arg) -{ - uint64_t *allocated = (uint64_t *)arg; - - if (allocated != NULL) - idalloc(allocated); -} -#endif - -/* - * FreeBSD's pthreads implementation calls malloc(3), so the malloc - * implementation has to take pains to avoid infinite recursion during - * initialization. - */ static inline bool malloc_init(void) { @@ -352,68 +295,64 @@ malloc_conf_next(char const **opts_p, char const **k_p, size_t *klen_p, for (accept = false; accept == false;) { switch (*opts) { - case 'A': case 'B': case 'C': case 'D': case 'E': - case 'F': case 'G': case 'H': case 'I': case 'J': - case 'K': case 'L': case 'M': case 'N': case 'O': - case 'P': case 'Q': case 'R': case 'S': case 'T': - case 'U': case 'V': case 'W': case 'X': case 'Y': - case 'Z': - case 'a': case 'b': case 'c': case 'd': case 'e': - case 'f': case 'g': case 'h': case 'i': case 'j': - case 'k': case 'l': case 'm': case 'n': case 'o': - case 'p': case 'q': case 'r': case 's': case 't': - case 'u': case 'v': case 'w': case 'x': case 'y': - case 'z': - case '0': case '1': case '2': case '3': case '4': - case '5': case '6': case '7': case '8': case '9': - case '_': - opts++; - break; - case ':': - opts++; - *klen_p = (uintptr_t)opts - 1 - (uintptr_t)*k_p; - *v_p = opts; - accept = true; - break; - case '\0': - if (opts != *opts_p) { - malloc_write(": Conf string " - "ends with key\n"); - } - return (true); - default: - malloc_write(": Malformed conf " - "string\n"); - return (true); + case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': + case 'G': case 'H': case 'I': case 'J': case 'K': case 'L': + case 'M': case 'N': case 'O': case 'P': case 'Q': case 'R': + case 'S': case 'T': case 'U': case 'V': case 'W': case 'X': + case 'Y': case 'Z': + case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': + case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': + case 'm': case 'n': case 'o': case 'p': case 'q': case 'r': + case 's': case 't': case 'u': case 'v': case 'w': case 'x': + case 'y': case 'z': + case '0': case '1': case '2': case '3': case '4': case '5': + case '6': case '7': case '8': case '9': + case '_': + opts++; + break; + case ':': + opts++; + *klen_p = (uintptr_t)opts - 1 - (uintptr_t)*k_p; + *v_p = opts; + accept = true; + break; + case '\0': + if (opts != *opts_p) { + malloc_write(": Conf string ends " + "with key\n"); + } + return (true); + default: + malloc_write(": Malformed conf string\n"); + return (true); } } for (accept = false; accept == false;) { switch (*opts) { - case ',': - opts++; - /* - * Look ahead one character here, because the - * next time this function is called, it will - * assume that end of input has been cleanly - * reached if no input remains, but we have - * optimistically already consumed the comma if - * one exists. - */ - if (*opts == '\0') { - malloc_write(": Conf string " - "ends with comma\n"); - } - *vlen_p = (uintptr_t)opts - 1 - (uintptr_t)*v_p; - accept = true; - break; - case '\0': - *vlen_p = (uintptr_t)opts - (uintptr_t)*v_p; - accept = true; - break; - default: - opts++; - break; + case ',': + opts++; + /* + * Look ahead one character here, because the next time + * this function is called, it will assume that end of + * input has been cleanly reached if no input remains, + * but we have optimistically already consumed the + * comma if one exists. + */ + if (*opts == '\0') { + malloc_write(": Conf string ends " + "with comma\n"); + } + *vlen_p = (uintptr_t)opts - 1 - (uintptr_t)*v_p; + accept = true; + break; + case '\0': + *vlen_p = (uintptr_t)opts - (uintptr_t)*v_p; + accept = true; + break; + default: + opts++; + break; } } @@ -425,17 +364,9 @@ static void malloc_conf_error(const char *msg, const char *k, size_t klen, const char *v, size_t vlen) { - char buf[PATH_MAX + 1]; - malloc_write(": "); - malloc_write(msg); - malloc_write(": "); - memcpy(buf, k, klen); - memcpy(&buf[klen], ":", 1); - memcpy(&buf[klen+1], v, vlen); - buf[klen+1+vlen] = '\0'; - malloc_write(buf); - malloc_write("\n"); + malloc_printf(": %s: %.*s:%.*s\n", msg, (int)klen, k, + (int)vlen, v); } static void @@ -450,12 +381,12 @@ malloc_conf_init(void) /* Get runtime configuration. */ switch (i) { case 0: - if (JEMALLOC_P(malloc_conf) != NULL) { + if (je_malloc_conf != NULL) { /* * Use options that were compiled into the * program. */ - opts = JEMALLOC_P(malloc_conf); + opts = je_malloc_conf; } else { /* No configuration specified. */ buf[0] = '\0'; @@ -463,13 +394,14 @@ malloc_conf_init(void) } break; case 1: { +#ifndef _WIN32 int linklen; const char *linkname = -#ifdef JEMALLOC_PREFIX +# ifdef JEMALLOC_PREFIX "/etc/"JEMALLOC_PREFIX"malloc.conf" -#else +# else "/etc/malloc.conf" -#endif +# endif ; if ((linklen = readlink(linkname, buf, @@ -480,14 +412,15 @@ malloc_conf_init(void) */ buf[linklen] = '\0'; opts = buf; - } else { + } else +#endif + { /* No configuration specified. */ buf[0] = '\0'; opts = buf; } break; - } - case 2: { + } case 2: { const char *envname = #ifdef JEMALLOC_PREFIX JEMALLOC_CPREFIX"MALLOC_CONF" @@ -508,8 +441,7 @@ malloc_conf_init(void) opts = buf; } break; - } - default: + } default: /* NOTREACHED */ assert(false); buf[0] = '\0'; @@ -518,52 +450,59 @@ malloc_conf_init(void) while (*opts != '\0' && malloc_conf_next(&opts, &k, &klen, &v, &vlen) == false) { -#define CONF_HANDLE_BOOL(n) \ - if (sizeof(#n)-1 == klen && strncmp(#n, k, \ +#define CONF_HANDLE_BOOL_HIT(o, n, hit) \ + if (sizeof(n)-1 == klen && strncmp(n, k, \ klen) == 0) { \ if (strncmp("true", v, vlen) == 0 && \ vlen == sizeof("true")-1) \ - opt_##n = true; \ + o = true; \ else if (strncmp("false", v, vlen) == \ 0 && vlen == sizeof("false")-1) \ - opt_##n = false; \ + o = false; \ else { \ malloc_conf_error( \ "Invalid conf value", \ k, klen, v, vlen); \ } \ + hit = true; \ + } else \ + hit = false; +#define CONF_HANDLE_BOOL(o, n) { \ + bool hit; \ + CONF_HANDLE_BOOL_HIT(o, n, hit); \ + if (hit) \ continue; \ - } -#define CONF_HANDLE_SIZE_T(n, min, max) \ - if (sizeof(#n)-1 == klen && strncmp(#n, k, \ +} +#define CONF_HANDLE_SIZE_T(o, n, min, max) \ + if (sizeof(n)-1 == klen && strncmp(n, k, \ klen) == 0) { \ - unsigned long ul; \ + uintmax_t um; \ char *end; \ \ - errno = 0; \ - ul = strtoul(v, &end, 0); \ - if (errno != 0 || (uintptr_t)end - \ + set_errno(0); \ + um = malloc_strtoumax(v, &end, 0); \ + if (get_errno() != 0 || (uintptr_t)end -\ (uintptr_t)v != vlen) { \ malloc_conf_error( \ "Invalid conf value", \ k, klen, v, vlen); \ - } else if (ul < min || ul > max) { \ + } else if (um < min || um > max) { \ malloc_conf_error( \ "Out-of-range conf value", \ k, klen, v, vlen); \ } else \ - opt_##n = ul; \ + o = um; \ continue; \ } -#define CONF_HANDLE_SSIZE_T(n, min, max) \ - if (sizeof(#n)-1 == klen && strncmp(#n, k, \ +#define CONF_HANDLE_SSIZE_T(o, n, min, max) \ + if (sizeof(n)-1 == klen && strncmp(n, k, \ klen) == 0) { \ long l; \ char *end; \ \ - errno = 0; \ + set_errno(0); \ l = strtol(v, &end, 0); \ - if (errno != 0 || (uintptr_t)end - \ + if (get_errno() != 0 || (uintptr_t)end -\ (uintptr_t)v != vlen) { \ malloc_conf_error( \ "Invalid conf value", \ @@ -574,70 +513,86 @@ malloc_conf_init(void) "Out-of-range conf value", \ k, klen, v, vlen); \ } else \ - opt_##n = l; \ + o = l; \ continue; \ } -#define CONF_HANDLE_CHAR_P(n, d) \ - if (sizeof(#n)-1 == klen && strncmp(#n, k, \ +#define CONF_HANDLE_CHAR_P(o, n, d) \ + if (sizeof(n)-1 == klen && strncmp(n, k, \ klen) == 0) { \ size_t cpylen = (vlen <= \ - sizeof(opt_##n)-1) ? vlen : \ - sizeof(opt_##n)-1; \ - strncpy(opt_##n, v, cpylen); \ - opt_##n[cpylen] = '\0'; \ + sizeof(o)-1) ? vlen : \ + sizeof(o)-1; \ + strncpy(o, v, cpylen); \ + o[cpylen] = '\0'; \ continue; \ } - CONF_HANDLE_BOOL(abort) - CONF_HANDLE_SIZE_T(lg_qspace_max, LG_QUANTUM, - PAGE_SHIFT-1) - CONF_HANDLE_SIZE_T(lg_cspace_max, LG_QUANTUM, - PAGE_SHIFT-1) + CONF_HANDLE_BOOL(opt_abort, "abort") /* - * Chunks always require at least one * header page, - * plus one data page. + * Chunks always require at least one header page, plus + * one data page in the absence of redzones, or three + * pages in the presence of redzones. In order to + * simplify options processing, fix the limit based on + * config_fill. */ - CONF_HANDLE_SIZE_T(lg_chunk, PAGE_SHIFT+1, - (sizeof(size_t) << 3) - 1) - CONF_HANDLE_SIZE_T(narenas, 1, SIZE_T_MAX) - CONF_HANDLE_SSIZE_T(lg_dirty_mult, -1, - (sizeof(size_t) << 3) - 1) - CONF_HANDLE_BOOL(stats_print) -#ifdef JEMALLOC_FILL - CONF_HANDLE_BOOL(junk) - CONF_HANDLE_BOOL(zero) -#endif -#ifdef JEMALLOC_SYSV - CONF_HANDLE_BOOL(sysv) -#endif -#ifdef JEMALLOC_XMALLOC - CONF_HANDLE_BOOL(xmalloc) -#endif -#ifdef JEMALLOC_TCACHE - CONF_HANDLE_BOOL(tcache) - CONF_HANDLE_SSIZE_T(lg_tcache_gc_sweep, -1, - (sizeof(size_t) << 3) - 1) - CONF_HANDLE_SSIZE_T(lg_tcache_max, -1, - (sizeof(size_t) << 3) - 1) -#endif -#ifdef JEMALLOC_PROF - CONF_HANDLE_BOOL(prof) - CONF_HANDLE_CHAR_P(prof_prefix, "jeprof") - CONF_HANDLE_SIZE_T(lg_prof_bt_max, 0, LG_PROF_BT_MAX) - CONF_HANDLE_BOOL(prof_active) - CONF_HANDLE_SSIZE_T(lg_prof_sample, 0, - (sizeof(uint64_t) << 3) - 1) - CONF_HANDLE_BOOL(prof_accum) - CONF_HANDLE_SSIZE_T(lg_prof_tcmax, -1, - (sizeof(size_t) << 3) - 1) - CONF_HANDLE_SSIZE_T(lg_prof_interval, -1, - (sizeof(uint64_t) << 3) - 1) - CONF_HANDLE_BOOL(prof_gdump) - CONF_HANDLE_BOOL(prof_leak) -#endif -#ifdef JEMALLOC_SWAP - CONF_HANDLE_BOOL(overcommit) -#endif + CONF_HANDLE_SIZE_T(opt_lg_chunk, "lg_chunk", LG_PAGE + + (config_fill ? 2 : 1), (sizeof(size_t) << 3) - 1) + CONF_HANDLE_SIZE_T(opt_narenas, "narenas", 1, + SIZE_T_MAX) + CONF_HANDLE_SSIZE_T(opt_lg_dirty_mult, "lg_dirty_mult", + -1, (sizeof(size_t) << 3) - 1) + CONF_HANDLE_BOOL(opt_stats_print, "stats_print") + if (config_fill) { + CONF_HANDLE_BOOL(opt_junk, "junk") + CONF_HANDLE_SIZE_T(opt_quarantine, "quarantine", + 0, SIZE_T_MAX) + CONF_HANDLE_BOOL(opt_redzone, "redzone") + CONF_HANDLE_BOOL(opt_zero, "zero") + } + if (config_utrace) { + CONF_HANDLE_BOOL(opt_utrace, "utrace") + } + if (config_valgrind) { + bool hit; + CONF_HANDLE_BOOL_HIT(opt_valgrind, + "valgrind", hit) + if (config_fill && opt_valgrind && hit) { + opt_junk = false; + opt_zero = false; + if (opt_quarantine == 0) { + opt_quarantine = + JEMALLOC_VALGRIND_QUARANTINE_DEFAULT; + } + opt_redzone = true; + } + if (hit) + continue; + } + if (config_xmalloc) { + CONF_HANDLE_BOOL(opt_xmalloc, "xmalloc") + } + if (config_tcache) { + CONF_HANDLE_BOOL(opt_tcache, "tcache") + CONF_HANDLE_SSIZE_T(opt_lg_tcache_max, + "lg_tcache_max", -1, + (sizeof(size_t) << 3) - 1) + } + if (config_prof) { + CONF_HANDLE_BOOL(opt_prof, "prof") + CONF_HANDLE_CHAR_P(opt_prof_prefix, + "prof_prefix", "jeprof") + CONF_HANDLE_BOOL(opt_prof_active, "prof_active") + CONF_HANDLE_SSIZE_T(opt_lg_prof_sample, + "lg_prof_sample", 0, + (sizeof(uint64_t) << 3) - 1) + CONF_HANDLE_BOOL(opt_prof_accum, "prof_accum") + CONF_HANDLE_SSIZE_T(opt_lg_prof_interval, + "lg_prof_interval", -1, + (sizeof(uint64_t) << 3) - 1) + CONF_HANDLE_BOOL(opt_prof_gdump, "prof_gdump") + CONF_HANDLE_BOOL(opt_prof_final, "prof_final") + CONF_HANDLE_BOOL(opt_prof_leak, "prof_leak") + } malloc_conf_error("Invalid conf pair", k, klen, v, vlen); #undef CONF_HANDLE_BOOL @@ -645,14 +600,6 @@ malloc_conf_init(void) #undef CONF_HANDLE_SSIZE_T #undef CONF_HANDLE_CHAR_P } - - /* Validate configuration of options that are inter-related. */ - if (opt_lg_qspace_max+1 >= opt_lg_cspace_max) { - malloc_write(": Invalid lg_[qc]space_max " - "relationship; restoring defaults\n"); - opt_lg_qspace_max = LG_QSPACE_MAX_DEFAULT; - opt_lg_cspace_max = LG_CSPACE_MAX_DEFAULT; - } } } @@ -662,7 +609,7 @@ malloc_init_hard(void) arena_t *init_arenas[1]; malloc_mutex_lock(&init_lock); - if (malloc_initialized || malloc_initializer == pthread_self()) { + if (malloc_initialized || IS_INITIALIZER) { /* * Another thread initialized the allocator before this one * acquired init_lock, or this thread is the initializing @@ -671,7 +618,8 @@ malloc_init_hard(void) malloc_mutex_unlock(&init_lock); return (false); } - if (malloc_initializer != (unsigned long)0) { +#ifdef JEMALLOC_THREADED_INIT + if (malloc_initializer != NO_INITIALIZER && IS_INITIALIZER == false) { /* Busy-wait until the initializing thread completes. */ do { malloc_mutex_unlock(&init_lock); @@ -681,44 +629,25 @@ malloc_init_hard(void) malloc_mutex_unlock(&init_lock); return (false); } - -#ifdef DYNAMIC_PAGE_SHIFT - /* Get page size. */ - { - long result; - - result = sysconf(_SC_PAGESIZE); - assert(result != -1); - pagesize = (size_t)result; - - /* - * We assume that pagesize is a power of 2 when calculating - * pagesize_mask and lg_pagesize. - */ - assert(((result - 1) & result) == 0); - pagesize_mask = result - 1; - lg_pagesize = ffs((int)result) - 1; - } #endif + malloc_initializer = INITIALIZER; -#ifdef JEMALLOC_PROF - prof_boot0(); -#endif + malloc_tsd_boot(); + if (config_prof) + prof_boot0(); malloc_conf_init(); +#if (!defined(JEMALLOC_MUTEX_INIT_CB) && !defined(JEMALLOC_ZONE) \ + && !defined(_WIN32)) /* Register fork handlers. */ - if (pthread_atfork(jemalloc_prefork, jemalloc_postfork, - jemalloc_postfork) != 0) { + if (pthread_atfork(jemalloc_prefork, jemalloc_postfork_parent, + jemalloc_postfork_child) != 0) { malloc_write(": Error in pthread_atfork()\n"); if (opt_abort) abort(); } - - if (ctl_boot()) { - malloc_mutex_unlock(&init_lock); - return (true); - } +#endif if (opt_stats_print) { /* Print statistics at exit. */ @@ -729,54 +658,39 @@ malloc_init_hard(void) } } - if (chunk_boot()) { - malloc_mutex_unlock(&init_lock); - return (true); - } - if (base_boot()) { malloc_mutex_unlock(&init_lock); return (true); } -#ifdef JEMALLOC_PROF - prof_boot1(); -#endif - - if (arena_boot()) { + if (chunk_boot()) { malloc_mutex_unlock(&init_lock); return (true); } -#ifdef JEMALLOC_TCACHE - if (tcache_boot()) { + if (ctl_boot()) { + malloc_mutex_unlock(&init_lock); + return (true); + } + + if (config_prof) + prof_boot1(); + + arena_boot(); + + if (config_tcache && tcache_boot0()) { malloc_mutex_unlock(&init_lock); return (true); } -#endif if (huge_boot()) { malloc_mutex_unlock(&init_lock); return (true); } -#if (defined(JEMALLOC_STATS) && defined(NO_TLS)) - /* Initialize allocation counters before any allocations can occur. */ - if (pthread_key_create(&thread_allocated_tsd, thread_allocated_cleanup) - != 0) { - malloc_mutex_unlock(&init_lock); - return (true); - } -#endif - if (malloc_mutex_init(&arenas_lock)) return (true); - if (pthread_key_create(&arenas_tsd, arenas_cleanup) != 0) { - malloc_mutex_unlock(&init_lock); - return (true); - } - /* * Create enough scaffolding to allow recursive allocation in * malloc_ncpus(). @@ -795,27 +709,42 @@ malloc_init_hard(void) return (true); } - /* - * Assign the initial arena to the initial thread, in order to avoid - * spurious creation of an extra arena if the application switches to - * threaded mode. - */ - ARENA_SET(arenas[0]); - arenas[0]->nthreads++; - -#ifdef JEMALLOC_PROF - if (prof_boot2()) { + /* Initialize allocation counters before any allocations can occur. */ + if (config_stats && thread_allocated_tsd_boot()) { + malloc_mutex_unlock(&init_lock); + return (true); + } + + if (arenas_tsd_boot()) { + malloc_mutex_unlock(&init_lock); + return (true); + } + + if (config_tcache && tcache_boot1()) { + malloc_mutex_unlock(&init_lock); + return (true); + } + + if (config_fill && quarantine_boot()) { + malloc_mutex_unlock(&init_lock); + return (true); + } + + if (config_prof && prof_boot2()) { malloc_mutex_unlock(&init_lock); return (true); } -#endif /* Get number of CPUs. */ - malloc_initializer = pthread_self(); malloc_mutex_unlock(&init_lock); ncpus = malloc_ncpus(); malloc_mutex_lock(&init_lock); + if (mutex_boot()) { + malloc_mutex_unlock(&init_lock); + return (true); + } + if (opt_narenas == 0) { /* * For SMP systems, create more than one arena per CPU by @@ -833,12 +762,9 @@ malloc_init_hard(void) * machinery will fail to allocate memory at far lower limits. */ if (narenas > chunksize / sizeof(arena_t *)) { - char buf[UMAX2S_BUFSIZE]; - narenas = chunksize / sizeof(arena_t *); - malloc_write(": Reducing narenas to limit ("); - malloc_write(u2s(narenas, 10, buf)); - malloc_write(")\n"); + malloc_printf(": Reducing narenas to limit (%d)\n", + narenas); } /* Allocate and initialize arenas. */ @@ -855,34 +781,11 @@ malloc_init_hard(void) /* Copy the pointer to the one arena that was already initialized. */ arenas[0] = init_arenas[0]; -#ifdef JEMALLOC_ZONE - /* Register the custom zone. */ - malloc_zone_register(create_zone()); - - /* - * Convert the default szone to an "overlay zone" that is capable of - * deallocating szone-allocated objects, but allocating new objects - * from jemalloc. - */ - szone2ozone(malloc_default_zone()); -#endif - malloc_initialized = true; malloc_mutex_unlock(&init_lock); return (false); } -#ifdef JEMALLOC_ZONE -JEMALLOC_ATTR(constructor) -void -jemalloc_darwin_init(void) -{ - - if (malloc_init_hard()) - abort(); -} -#endif - /* * End initialization functions. */ @@ -891,191 +794,118 @@ jemalloc_darwin_init(void) * Begin malloc(3)-compatible functions. */ -JEMALLOC_ATTR(malloc) -JEMALLOC_ATTR(visibility("default")) void * -JEMALLOC_P(malloc)(size_t size) +je_malloc(size_t size) { void *ret; -#if (defined(JEMALLOC_PROF) || defined(JEMALLOC_STATS)) - size_t usize -# ifdef JEMALLOC_CC_SILENCE - = 0 -# endif - ; -#endif -#ifdef JEMALLOC_PROF - prof_thr_cnt_t *cnt -# ifdef JEMALLOC_CC_SILENCE - = NULL -# endif - ; -#endif + size_t usize JEMALLOC_CC_SILENCE_INIT(0); + prof_thr_cnt_t *cnt JEMALLOC_CC_SILENCE_INIT(NULL); if (malloc_init()) { ret = NULL; - goto OOM; + goto label_oom; } - if (size == 0) { -#ifdef JEMALLOC_SYSV - if (opt_sysv == false) -#endif - size = 1; -#ifdef JEMALLOC_SYSV - else { -# ifdef JEMALLOC_XMALLOC - if (opt_xmalloc) { - malloc_write(": Error in malloc(): " - "invalid size 0\n"); - abort(); - } -# endif - ret = NULL; - goto RETURN; - } -#endif - } + if (size == 0) + size = 1; -#ifdef JEMALLOC_PROF - if (opt_prof) { + if (config_prof && opt_prof) { usize = s2u(size); PROF_ALLOC_PREP(1, usize, cnt); if (cnt == NULL) { ret = NULL; - goto OOM; + goto label_oom; } if (prof_promote && (uintptr_t)cnt != (uintptr_t)1U && usize <= - small_maxclass) { - ret = imalloc(small_maxclass+1); + SMALL_MAXCLASS) { + ret = imalloc(SMALL_MAXCLASS+1); if (ret != NULL) arena_prof_promoted(ret, usize); } else ret = imalloc(size); - } else -#endif - { -#ifdef JEMALLOC_STATS - usize = s2u(size); -#endif + } else { + if (config_stats || (config_valgrind && opt_valgrind)) + usize = s2u(size); ret = imalloc(size); } -OOM: +label_oom: if (ret == NULL) { -#ifdef JEMALLOC_XMALLOC - if (opt_xmalloc) { + if (config_xmalloc && opt_xmalloc) { malloc_write(": Error in malloc(): " "out of memory\n"); abort(); } -#endif - errno = ENOMEM; + set_errno(ENOMEM); } - -#ifdef JEMALLOC_SYSV -RETURN: -#endif -#ifdef JEMALLOC_PROF - if (opt_prof && ret != NULL) + if (config_prof && opt_prof && ret != NULL) prof_malloc(ret, usize, cnt); -#endif -#ifdef JEMALLOC_STATS - if (ret != NULL) { - assert(usize == isalloc(ret)); - ALLOCATED_ADD(usize, 0); + if (config_stats && ret != NULL) { + assert(usize == isalloc(ret, config_prof)); + thread_allocated_tsd_get()->allocated += usize; } -#endif + UTRACE(0, size, ret); + JEMALLOC_VALGRIND_MALLOC(ret != NULL, ret, usize, false); return (ret); } JEMALLOC_ATTR(nonnull(1)) #ifdef JEMALLOC_PROF /* - * Avoid any uncertainty as to how many backtrace frames to ignore in + * Avoid any uncertainty as to how many backtrace frames to ignore in * PROF_ALLOC_PREP(). */ JEMALLOC_ATTR(noinline) #endif static int -imemalign(void **memptr, size_t alignment, size_t size) +imemalign(void **memptr, size_t alignment, size_t size, + size_t min_alignment) { int ret; - size_t usize -#ifdef JEMALLOC_CC_SILENCE - = 0 -#endif - ; + size_t usize; void *result; -#ifdef JEMALLOC_PROF - prof_thr_cnt_t *cnt -# ifdef JEMALLOC_CC_SILENCE - = NULL -# endif - ; -#endif + prof_thr_cnt_t *cnt JEMALLOC_CC_SILENCE_INIT(NULL); + + assert(min_alignment != 0); if (malloc_init()) result = NULL; else { - if (size == 0) { -#ifdef JEMALLOC_SYSV - if (opt_sysv == false) -#endif - size = 1; -#ifdef JEMALLOC_SYSV - else { -# ifdef JEMALLOC_XMALLOC - if (opt_xmalloc) { - malloc_write(": Error in " - "posix_memalign(): invalid size " - "0\n"); - abort(); - } -# endif - result = NULL; - *memptr = NULL; - ret = 0; - goto RETURN; - } -#endif - } + if (size == 0) + size = 1; /* Make sure that alignment is a large enough power of 2. */ if (((alignment - 1) & alignment) != 0 - || alignment < sizeof(void *)) { -#ifdef JEMALLOC_XMALLOC - if (opt_xmalloc) { - malloc_write(": Error in " - "posix_memalign(): invalid alignment\n"); + || (alignment < min_alignment)) { + if (config_xmalloc && opt_xmalloc) { + malloc_write(": Error allocating " + "aligned memory: invalid alignment\n"); abort(); } -#endif result = NULL; ret = EINVAL; - goto RETURN; + goto label_return; } - usize = sa2u(size, alignment, NULL); + usize = sa2u(size, alignment); if (usize == 0) { result = NULL; ret = ENOMEM; - goto RETURN; + goto label_return; } -#ifdef JEMALLOC_PROF - if (opt_prof) { + if (config_prof && opt_prof) { PROF_ALLOC_PREP(2, usize, cnt); if (cnt == NULL) { result = NULL; ret = EINVAL; } else { if (prof_promote && (uintptr_t)cnt != - (uintptr_t)1U && usize <= small_maxclass) { - assert(sa2u(small_maxclass+1, - alignment, NULL) != 0); - result = ipalloc(sa2u(small_maxclass+1, - alignment, NULL), alignment, false); + (uintptr_t)1U && usize <= SMALL_MAXCLASS) { + assert(sa2u(SMALL_MAXCLASS+1, + alignment) != 0); + result = ipalloc(sa2u(SMALL_MAXCLASS+1, + alignment), alignment, false); if (result != NULL) { arena_prof_promoted(result, usize); @@ -1086,88 +916,79 @@ imemalign(void **memptr, size_t alignment, size_t size) } } } else -#endif result = ipalloc(usize, alignment, false); } if (result == NULL) { -#ifdef JEMALLOC_XMALLOC - if (opt_xmalloc) { - malloc_write(": Error in posix_memalign(): " - "out of memory\n"); + if (config_xmalloc && opt_xmalloc) { + malloc_write(": Error allocating aligned " + "memory: out of memory\n"); abort(); } -#endif ret = ENOMEM; - goto RETURN; + goto label_return; } *memptr = result; ret = 0; -RETURN: -#ifdef JEMALLOC_STATS - if (result != NULL) { - assert(usize == isalloc(result)); - ALLOCATED_ADD(usize, 0); +label_return: + if (config_stats && result != NULL) { + assert(usize == isalloc(result, config_prof)); + thread_allocated_tsd_get()->allocated += usize; } -#endif -#ifdef JEMALLOC_PROF - if (opt_prof && result != NULL) + if (config_prof && opt_prof && result != NULL) prof_malloc(result, usize, cnt); -#endif + UTRACE(0, size, result); return (ret); } -JEMALLOC_ATTR(nonnull(1)) -JEMALLOC_ATTR(visibility("default")) int -JEMALLOC_P(posix_memalign)(void **memptr, size_t alignment, size_t size) +je_posix_memalign(void **memptr, size_t alignment, size_t size) { - - return imemalign(memptr, alignment, size); + int ret = imemalign(memptr, alignment, size, sizeof(void *)); + JEMALLOC_VALGRIND_MALLOC(ret == 0, *memptr, isalloc(*memptr, + config_prof), false); + return (ret); } -JEMALLOC_ATTR(malloc) -JEMALLOC_ATTR(visibility("default")) void * -JEMALLOC_P(calloc)(size_t num, size_t size) +je_aligned_alloc(size_t alignment, size_t size) +{ + void *ret; + int err; + + if ((err = imemalign(&ret, alignment, size, 1)) != 0) { + ret = NULL; + set_errno(err); + } + JEMALLOC_VALGRIND_MALLOC(err == 0, ret, isalloc(ret, config_prof), + false); + return (ret); +} + +void * +je_calloc(size_t num, size_t size) { void *ret; size_t num_size; -#if (defined(JEMALLOC_PROF) || defined(JEMALLOC_STATS)) - size_t usize -# ifdef JEMALLOC_CC_SILENCE - = 0 -# endif - ; -#endif -#ifdef JEMALLOC_PROF - prof_thr_cnt_t *cnt -# ifdef JEMALLOC_CC_SILENCE - = NULL -# endif - ; -#endif + size_t usize JEMALLOC_CC_SILENCE_INIT(0); + prof_thr_cnt_t *cnt JEMALLOC_CC_SILENCE_INIT(NULL); if (malloc_init()) { num_size = 0; ret = NULL; - goto RETURN; + goto label_return; } num_size = num * size; if (num_size == 0) { -#ifdef JEMALLOC_SYSV - if ((opt_sysv == false) && ((num == 0) || (size == 0))) -#endif + if (num == 0 || size == 0) num_size = 1; -#ifdef JEMALLOC_SYSV else { ret = NULL; - goto RETURN; + goto label_return; } -#endif /* * Try to avoid division here. We know that it isn't possible to * overflow during multiplication if neither operand uses any of the @@ -1177,135 +998,113 @@ JEMALLOC_P(calloc)(size_t num, size_t size) && (num_size / size != num)) { /* size_t overflow. */ ret = NULL; - goto RETURN; + goto label_return; } -#ifdef JEMALLOC_PROF - if (opt_prof) { + if (config_prof && opt_prof) { usize = s2u(num_size); PROF_ALLOC_PREP(1, usize, cnt); if (cnt == NULL) { ret = NULL; - goto RETURN; + goto label_return; } if (prof_promote && (uintptr_t)cnt != (uintptr_t)1U && usize - <= small_maxclass) { - ret = icalloc(small_maxclass+1); + <= SMALL_MAXCLASS) { + ret = icalloc(SMALL_MAXCLASS+1); if (ret != NULL) arena_prof_promoted(ret, usize); } else ret = icalloc(num_size); - } else -#endif - { -#ifdef JEMALLOC_STATS - usize = s2u(num_size); -#endif + } else { + if (config_stats || (config_valgrind && opt_valgrind)) + usize = s2u(num_size); ret = icalloc(num_size); } -RETURN: +label_return: if (ret == NULL) { -#ifdef JEMALLOC_XMALLOC - if (opt_xmalloc) { + if (config_xmalloc && opt_xmalloc) { malloc_write(": Error in calloc(): out of " "memory\n"); abort(); } -#endif - errno = ENOMEM; + set_errno(ENOMEM); } -#ifdef JEMALLOC_PROF - if (opt_prof && ret != NULL) + if (config_prof && opt_prof && ret != NULL) prof_malloc(ret, usize, cnt); -#endif -#ifdef JEMALLOC_STATS - if (ret != NULL) { - assert(usize == isalloc(ret)); - ALLOCATED_ADD(usize, 0); + if (config_stats && ret != NULL) { + assert(usize == isalloc(ret, config_prof)); + thread_allocated_tsd_get()->allocated += usize; } -#endif + UTRACE(0, num_size, ret); + JEMALLOC_VALGRIND_MALLOC(ret != NULL, ret, usize, true); return (ret); } -JEMALLOC_ATTR(visibility("default")) void * -JEMALLOC_P(realloc)(void *ptr, size_t size) +je_realloc(void *ptr, size_t size) { void *ret; -#if (defined(JEMALLOC_PROF) || defined(JEMALLOC_STATS)) - size_t usize -# ifdef JEMALLOC_CC_SILENCE - = 0 -# endif - ; + size_t usize JEMALLOC_CC_SILENCE_INIT(0); size_t old_size = 0; -#endif -#ifdef JEMALLOC_PROF - prof_thr_cnt_t *cnt -# ifdef JEMALLOC_CC_SILENCE - = NULL -# endif - ; - prof_ctx_t *old_ctx -# ifdef JEMALLOC_CC_SILENCE - = NULL -# endif - ; -#endif + size_t old_rzsize JEMALLOC_CC_SILENCE_INIT(0); + prof_thr_cnt_t *cnt JEMALLOC_CC_SILENCE_INIT(NULL); + prof_ctx_t *old_ctx JEMALLOC_CC_SILENCE_INIT(NULL); if (size == 0) { -#ifdef JEMALLOC_SYSV - if (opt_sysv == false) -#endif - size = 1; -#ifdef JEMALLOC_SYSV - else { - if (ptr != NULL) { -#if (defined(JEMALLOC_PROF) || defined(JEMALLOC_STATS)) - old_size = isalloc(ptr); -#endif -#ifdef JEMALLOC_PROF - if (opt_prof) { - old_ctx = prof_ctx_get(ptr); - cnt = NULL; - } -#endif - idalloc(ptr); + if (ptr != NULL) { + /* realloc(ptr, 0) is equivalent to free(p). */ + if (config_prof) { + old_size = isalloc(ptr, true); + if (config_valgrind && opt_valgrind) + old_rzsize = p2rz(ptr); + } else if (config_stats) { + old_size = isalloc(ptr, false); + if (config_valgrind && opt_valgrind) + old_rzsize = u2rz(old_size); + } else if (config_valgrind && opt_valgrind) { + old_size = isalloc(ptr, false); + old_rzsize = u2rz(old_size); } -#ifdef JEMALLOC_PROF - else if (opt_prof) { - old_ctx = NULL; + if (config_prof && opt_prof) { + old_ctx = prof_ctx_get(ptr); cnt = NULL; } -#endif + iqalloc(ptr); ret = NULL; - goto RETURN; - } -#endif + goto label_return; + } else + size = 1; } if (ptr != NULL) { - assert(malloc_initialized || malloc_initializer == - pthread_self()); + assert(malloc_initialized || IS_INITIALIZER); -#if (defined(JEMALLOC_PROF) || defined(JEMALLOC_STATS)) - old_size = isalloc(ptr); -#endif -#ifdef JEMALLOC_PROF - if (opt_prof) { + if (config_prof) { + old_size = isalloc(ptr, true); + if (config_valgrind && opt_valgrind) + old_rzsize = p2rz(ptr); + } else if (config_stats) { + old_size = isalloc(ptr, false); + if (config_valgrind && opt_valgrind) + old_rzsize = u2rz(old_size); + } else if (config_valgrind && opt_valgrind) { + old_size = isalloc(ptr, false); + old_rzsize = u2rz(old_size); + } + if (config_prof && opt_prof) { usize = s2u(size); old_ctx = prof_ctx_get(ptr); PROF_ALLOC_PREP(1, usize, cnt); if (cnt == NULL) { old_ctx = NULL; ret = NULL; - goto OOM; + goto label_oom; } if (prof_promote && (uintptr_t)cnt != (uintptr_t)1U && - usize <= small_maxclass) { - ret = iralloc(ptr, small_maxclass+1, 0, 0, + usize <= SMALL_MAXCLASS) { + ret = iralloc(ptr, SMALL_MAXCLASS+1, 0, 0, false, false); if (ret != NULL) arena_prof_promoted(ret, usize); @@ -1316,42 +1115,31 @@ JEMALLOC_P(realloc)(void *ptr, size_t size) if (ret == NULL) old_ctx = NULL; } - } else -#endif - { -#ifdef JEMALLOC_STATS - usize = s2u(size); -#endif + } else { + if (config_stats || (config_valgrind && opt_valgrind)) + usize = s2u(size); ret = iralloc(ptr, size, 0, 0, false, false); } -#ifdef JEMALLOC_PROF -OOM: -#endif +label_oom: if (ret == NULL) { -#ifdef JEMALLOC_XMALLOC - if (opt_xmalloc) { + if (config_xmalloc && opt_xmalloc) { malloc_write(": Error in realloc(): " "out of memory\n"); abort(); } -#endif - errno = ENOMEM; + set_errno(ENOMEM); } } else { -#ifdef JEMALLOC_PROF - if (opt_prof) + /* realloc(NULL, size) is equivalent to malloc(size). */ + if (config_prof && opt_prof) old_ctx = NULL; -#endif if (malloc_init()) { -#ifdef JEMALLOC_PROF - if (opt_prof) + if (config_prof && opt_prof) cnt = NULL; -#endif ret = NULL; } else { -#ifdef JEMALLOC_PROF - if (opt_prof) { + if (config_prof && opt_prof) { usize = s2u(size); PROF_ALLOC_PREP(1, usize, cnt); if (cnt == NULL) @@ -1359,8 +1147,8 @@ OOM: else { if (prof_promote && (uintptr_t)cnt != (uintptr_t)1U && usize <= - small_maxclass) { - ret = imalloc(small_maxclass+1); + SMALL_MAXCLASS) { + ret = imalloc(SMALL_MAXCLASS+1); if (ret != NULL) { arena_prof_promoted(ret, usize); @@ -1368,72 +1156,61 @@ OOM: } else ret = imalloc(size); } - } else -#endif - { -#ifdef JEMALLOC_STATS - usize = s2u(size); -#endif + } else { + if (config_stats || (config_valgrind && + opt_valgrind)) + usize = s2u(size); ret = imalloc(size); } } if (ret == NULL) { -#ifdef JEMALLOC_XMALLOC - if (opt_xmalloc) { + if (config_xmalloc && opt_xmalloc) { malloc_write(": Error in realloc(): " "out of memory\n"); abort(); } -#endif - errno = ENOMEM; + set_errno(ENOMEM); } } -#ifdef JEMALLOC_SYSV -RETURN: -#endif -#ifdef JEMALLOC_PROF - if (opt_prof) +label_return: + if (config_prof && opt_prof) prof_realloc(ret, usize, cnt, old_size, old_ctx); -#endif -#ifdef JEMALLOC_STATS - if (ret != NULL) { - assert(usize == isalloc(ret)); - ALLOCATED_ADD(usize, old_size); + if (config_stats && ret != NULL) { + thread_allocated_t *ta; + assert(usize == isalloc(ret, config_prof)); + ta = thread_allocated_tsd_get(); + ta->allocated += usize; + ta->deallocated += old_size; } -#endif + UTRACE(ptr, size, ret); + JEMALLOC_VALGRIND_REALLOC(ret, usize, ptr, old_size, old_rzsize, false); return (ret); } -JEMALLOC_ATTR(visibility("default")) void -JEMALLOC_P(free)(void *ptr) +je_free(void *ptr) { + UTRACE(ptr, 0, 0); if (ptr != NULL) { -#if (defined(JEMALLOC_PROF) || defined(JEMALLOC_STATS)) size_t usize; -#endif + size_t rzsize JEMALLOC_CC_SILENCE_INIT(0); - assert(malloc_initialized || malloc_initializer == - pthread_self()); + assert(malloc_initialized || IS_INITIALIZER); -#ifdef JEMALLOC_STATS - usize = isalloc(ptr); -#endif -#ifdef JEMALLOC_PROF - if (opt_prof) { -# ifndef JEMALLOC_STATS - usize = isalloc(ptr); -# endif + if (config_prof && opt_prof) { + usize = isalloc(ptr, config_prof); prof_free(ptr, usize); - } -#endif -#ifdef JEMALLOC_STATS - ALLOCATED_ADD(0, usize); -#endif - idalloc(ptr); + } else if (config_stats || config_valgrind) + usize = isalloc(ptr, config_prof); + if (config_stats) + thread_allocated_tsd_get()->deallocated += usize; + if (config_valgrind && opt_valgrind) + rzsize = p2rz(ptr); + iqalloc(ptr); + JEMALLOC_VALGRIND_FREE(ptr, rzsize); } } @@ -1443,51 +1220,56 @@ JEMALLOC_P(free)(void *ptr) /******************************************************************************/ /* * Begin non-standard override functions. - * - * These overrides are omitted if the JEMALLOC_PREFIX is defined, since the - * entire point is to avoid accidental mixed allocator usage. */ -#ifndef JEMALLOC_PREFIX #ifdef JEMALLOC_OVERRIDE_MEMALIGN -JEMALLOC_ATTR(malloc) -JEMALLOC_ATTR(visibility("default")) void * -JEMALLOC_P(memalign)(size_t alignment, size_t size) +je_memalign(size_t alignment, size_t size) { - void *ret; -#ifdef JEMALLOC_CC_SILENCE - int result = -#endif - imemalign(&ret, alignment, size); -#ifdef JEMALLOC_CC_SILENCE - if (result != 0) - return (NULL); -#endif + void *ret JEMALLOC_CC_SILENCE_INIT(NULL); + imemalign(&ret, alignment, size, 1); + JEMALLOC_VALGRIND_MALLOC(ret != NULL, ret, size, false); return (ret); } #endif #ifdef JEMALLOC_OVERRIDE_VALLOC -JEMALLOC_ATTR(malloc) -JEMALLOC_ATTR(visibility("default")) void * -JEMALLOC_P(valloc)(size_t size) +je_valloc(size_t size) { - void *ret; -#ifdef JEMALLOC_CC_SILENCE - int result = -#endif - imemalign(&ret, PAGE_SIZE, size); -#ifdef JEMALLOC_CC_SILENCE - if (result != 0) - return (NULL); -#endif + void *ret JEMALLOC_CC_SILENCE_INIT(NULL); + imemalign(&ret, PAGE, size, 1); + JEMALLOC_VALGRIND_MALLOC(ret != NULL, ret, size, false); return (ret); } #endif -#endif /* JEMALLOC_PREFIX */ +/* + * is_malloc(je_malloc) is some macro magic to detect if jemalloc_defs.h has + * #define je_malloc malloc + */ +#define malloc_is_malloc 1 +#define is_malloc_(a) malloc_is_ ## a +#define is_malloc(a) is_malloc_(a) + +#if ((is_malloc(je_malloc) == 1) && defined(__GLIBC__) && !defined(__UCLIBC__)) +/* + * glibc provides the RTLD_DEEPBIND flag for dlopen which can make it possible + * to inconsistently reference libc's malloc(3)-compatible functions + * (https://bugzilla.mozilla.org/show_bug.cgi?id=493541). + * + * These definitions interpose hooks in glibc. The functions are actually + * passed an extra argument for the caller return address, which will be + * ignored. + */ +JEMALLOC_EXPORT void (* const __free_hook)(void *ptr) = je_free; +JEMALLOC_EXPORT void *(* const __malloc_hook)(size_t size) = je_malloc; +JEMALLOC_EXPORT void *(* const __realloc_hook)(void *ptr, size_t size) = + je_realloc; +JEMALLOC_EXPORT void *(* const __memalign_hook)(size_t alignment, size_t size) = + je_memalign; +#endif + /* * End non-standard override functions. */ @@ -1496,36 +1278,31 @@ JEMALLOC_P(valloc)(size_t size) * Begin non-standard functions. */ -JEMALLOC_ATTR(visibility("default")) size_t -JEMALLOC_P(malloc_usable_size)(const void *ptr) +je_malloc_usable_size(const void *ptr) { size_t ret; - assert(malloc_initialized || malloc_initializer == pthread_self()); + assert(malloc_initialized || IS_INITIALIZER); -#ifdef JEMALLOC_IVSALLOC - ret = ivsalloc(ptr); -#else - assert(ptr != NULL); - ret = isalloc(ptr); -#endif + if (config_ivsalloc) + ret = ivsalloc(ptr, config_prof); + else + ret = (ptr != NULL) ? isalloc(ptr, config_prof) : 0; return (ret); } -JEMALLOC_ATTR(visibility("default")) void -JEMALLOC_P(malloc_stats_print)(void (*write_cb)(void *, const char *), - void *cbopaque, const char *opts) +je_malloc_stats_print(void (*write_cb)(void *, const char *), void *cbopaque, + const char *opts) { stats_print(write_cb, cbopaque, opts); } -JEMALLOC_ATTR(visibility("default")) int -JEMALLOC_P(mallctl)(const char *name, void *oldp, size_t *oldlenp, void *newp, +je_mallctl(const char *name, void *oldp, size_t *oldlenp, void *newp, size_t newlen) { @@ -1535,9 +1312,8 @@ JEMALLOC_P(mallctl)(const char *name, void *oldp, size_t *oldlenp, void *newp, return (ctl_byname(name, oldp, oldlenp, newp, newlen)); } -JEMALLOC_ATTR(visibility("default")) int -JEMALLOC_P(mallctlnametomib)(const char *name, size_t *mibp, size_t *miblenp) +je_mallctlnametomib(const char *name, size_t *mibp, size_t *miblenp) { if (malloc_init()) @@ -1546,10 +1322,9 @@ JEMALLOC_P(mallctlnametomib)(const char *name, size_t *mibp, size_t *miblenp) return (ctl_nametomib(name, mibp, miblenp)); } -JEMALLOC_ATTR(visibility("default")) int -JEMALLOC_P(mallctlbymib)(const size_t *mib, size_t miblen, void *oldp, - size_t *oldlenp, void *newp, size_t newlen) +je_mallctlbymib(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp, + void *newp, size_t newlen) { if (malloc_init()) @@ -1558,12 +1333,21 @@ JEMALLOC_P(mallctlbymib)(const size_t *mib, size_t miblen, void *oldp, return (ctl_bymib(mib, miblen, oldp, oldlenp, newp, newlen)); } +/* + * End non-standard functions. + */ +/******************************************************************************/ +/* + * Begin experimental functions. + */ +#ifdef JEMALLOC_EXPERIMENTAL + JEMALLOC_INLINE void * iallocm(size_t usize, size_t alignment, bool zero) { - assert(usize == ((alignment == 0) ? s2u(usize) : sa2u(usize, alignment, - NULL))); + assert(usize == ((alignment == 0) ? s2u(usize) : sa2u(usize, + alignment))); if (alignment != 0) return (ipalloc(usize, alignment, zero)); @@ -1573,116 +1357,96 @@ iallocm(size_t usize, size_t alignment, bool zero) return (imalloc(usize)); } -JEMALLOC_ATTR(nonnull(1)) -JEMALLOC_ATTR(visibility("default")) int -JEMALLOC_P(allocm)(void **ptr, size_t *rsize, size_t size, int flags) +je_allocm(void **ptr, size_t *rsize, size_t size, int flags) { void *p; size_t usize; size_t alignment = (ZU(1) << (flags & ALLOCM_LG_ALIGN_MASK) & (SIZE_T_MAX-1)); bool zero = flags & ALLOCM_ZERO; -#ifdef JEMALLOC_PROF - prof_thr_cnt_t *cnt; -#endif assert(ptr != NULL); assert(size != 0); if (malloc_init()) - goto OOM; + goto label_oom; - usize = (alignment == 0) ? s2u(size) : sa2u(size, alignment, NULL); + usize = (alignment == 0) ? s2u(size) : sa2u(size, alignment); if (usize == 0) - goto OOM; + goto label_oom; + + if (config_prof && opt_prof) { + prof_thr_cnt_t *cnt; -#ifdef JEMALLOC_PROF - if (opt_prof) { PROF_ALLOC_PREP(1, usize, cnt); if (cnt == NULL) - goto OOM; + goto label_oom; if (prof_promote && (uintptr_t)cnt != (uintptr_t)1U && usize <= - small_maxclass) { + SMALL_MAXCLASS) { size_t usize_promoted = (alignment == 0) ? - s2u(small_maxclass+1) : sa2u(small_maxclass+1, - alignment, NULL); + s2u(SMALL_MAXCLASS+1) : sa2u(SMALL_MAXCLASS+1, + alignment); assert(usize_promoted != 0); p = iallocm(usize_promoted, alignment, zero); if (p == NULL) - goto OOM; + goto label_oom; arena_prof_promoted(p, usize); } else { p = iallocm(usize, alignment, zero); if (p == NULL) - goto OOM; + goto label_oom; } prof_malloc(p, usize, cnt); - if (rsize != NULL) - *rsize = usize; - } else -#endif - { + } else { p = iallocm(usize, alignment, zero); if (p == NULL) - goto OOM; -#ifndef JEMALLOC_STATS - if (rsize != NULL) -#endif - { -#ifdef JEMALLOC_STATS - if (rsize != NULL) -#endif - *rsize = usize; - } + goto label_oom; } + if (rsize != NULL) + *rsize = usize; *ptr = p; -#ifdef JEMALLOC_STATS - assert(usize == isalloc(p)); - ALLOCATED_ADD(usize, 0); -#endif + if (config_stats) { + assert(usize == isalloc(p, config_prof)); + thread_allocated_tsd_get()->allocated += usize; + } + UTRACE(0, size, p); + JEMALLOC_VALGRIND_MALLOC(true, p, usize, zero); return (ALLOCM_SUCCESS); -OOM: -#ifdef JEMALLOC_XMALLOC - if (opt_xmalloc) { +label_oom: + if (config_xmalloc && opt_xmalloc) { malloc_write(": Error in allocm(): " "out of memory\n"); abort(); } -#endif *ptr = NULL; + UTRACE(0, size, 0); return (ALLOCM_ERR_OOM); } -JEMALLOC_ATTR(nonnull(1)) -JEMALLOC_ATTR(visibility("default")) int -JEMALLOC_P(rallocm)(void **ptr, size_t *rsize, size_t size, size_t extra, - int flags) +je_rallocm(void **ptr, size_t *rsize, size_t size, size_t extra, int flags) { void *p, *q; size_t usize; -#if (defined(JEMALLOC_PROF) || defined(JEMALLOC_STATS)) size_t old_size; -#endif + size_t old_rzsize JEMALLOC_CC_SILENCE_INIT(0); size_t alignment = (ZU(1) << (flags & ALLOCM_LG_ALIGN_MASK) & (SIZE_T_MAX-1)); bool zero = flags & ALLOCM_ZERO; bool no_move = flags & ALLOCM_NO_MOVE; -#ifdef JEMALLOC_PROF - prof_thr_cnt_t *cnt; -#endif assert(ptr != NULL); assert(*ptr != NULL); assert(size != 0); assert(SIZE_T_MAX - size >= extra); - assert(malloc_initialized || malloc_initializer == pthread_self()); + assert(malloc_initialized || IS_INITIALIZER); p = *ptr; -#ifdef JEMALLOC_PROF - if (opt_prof) { + if (config_prof && opt_prof) { + prof_thr_cnt_t *cnt; + /* * usize isn't knowable before iralloc() returns when extra is * non-zero. Therefore, compute its maximum possible value and @@ -1691,191 +1455,284 @@ JEMALLOC_P(rallocm)(void **ptr, size_t *rsize, size_t size, size_t extra, * decide whether to sample. */ size_t max_usize = (alignment == 0) ? s2u(size+extra) : - sa2u(size+extra, alignment, NULL); + sa2u(size+extra, alignment); prof_ctx_t *old_ctx = prof_ctx_get(p); - old_size = isalloc(p); + old_size = isalloc(p, true); + if (config_valgrind && opt_valgrind) + old_rzsize = p2rz(p); PROF_ALLOC_PREP(1, max_usize, cnt); if (cnt == NULL) - goto OOM; + goto label_oom; /* * Use minimum usize to determine whether promotion may happen. */ if (prof_promote && (uintptr_t)cnt != (uintptr_t)1U - && ((alignment == 0) ? s2u(size) : sa2u(size, - alignment, NULL)) <= small_maxclass) { - q = iralloc(p, small_maxclass+1, (small_maxclass+1 >= - size+extra) ? 0 : size+extra - (small_maxclass+1), + && ((alignment == 0) ? s2u(size) : sa2u(size, alignment)) + <= SMALL_MAXCLASS) { + q = iralloc(p, SMALL_MAXCLASS+1, (SMALL_MAXCLASS+1 >= + size+extra) ? 0 : size+extra - (SMALL_MAXCLASS+1), alignment, zero, no_move); if (q == NULL) - goto ERR; - if (max_usize < PAGE_SIZE) { + goto label_err; + if (max_usize < PAGE) { usize = max_usize; arena_prof_promoted(q, usize); } else - usize = isalloc(q); + usize = isalloc(q, config_prof); } else { q = iralloc(p, size, extra, alignment, zero, no_move); if (q == NULL) - goto ERR; - usize = isalloc(q); + goto label_err; + usize = isalloc(q, config_prof); } prof_realloc(q, usize, cnt, old_size, old_ctx); if (rsize != NULL) *rsize = usize; - } else -#endif - { -#ifdef JEMALLOC_STATS - old_size = isalloc(p); -#endif + } else { + if (config_stats) { + old_size = isalloc(p, false); + if (config_valgrind && opt_valgrind) + old_rzsize = u2rz(old_size); + } else if (config_valgrind && opt_valgrind) { + old_size = isalloc(p, false); + old_rzsize = u2rz(old_size); + } q = iralloc(p, size, extra, alignment, zero, no_move); if (q == NULL) - goto ERR; -#ifndef JEMALLOC_STATS - if (rsize != NULL) -#endif - { - usize = isalloc(q); -#ifdef JEMALLOC_STATS - if (rsize != NULL) -#endif - *rsize = usize; + goto label_err; + if (config_stats) + usize = isalloc(q, config_prof); + if (rsize != NULL) { + if (config_stats == false) + usize = isalloc(q, config_prof); + *rsize = usize; } } *ptr = q; -#ifdef JEMALLOC_STATS - ALLOCATED_ADD(usize, old_size); -#endif + if (config_stats) { + thread_allocated_t *ta; + ta = thread_allocated_tsd_get(); + ta->allocated += usize; + ta->deallocated += old_size; + } + UTRACE(p, size, q); + JEMALLOC_VALGRIND_REALLOC(q, usize, p, old_size, old_rzsize, zero); return (ALLOCM_SUCCESS); -ERR: - if (no_move) +label_err: + if (no_move) { + UTRACE(p, size, q); return (ALLOCM_ERR_NOT_MOVED); -#ifdef JEMALLOC_PROF -OOM: -#endif -#ifdef JEMALLOC_XMALLOC - if (opt_xmalloc) { + } +label_oom: + if (config_xmalloc && opt_xmalloc) { malloc_write(": Error in rallocm(): " "out of memory\n"); abort(); } -#endif + UTRACE(p, size, 0); return (ALLOCM_ERR_OOM); } -JEMALLOC_ATTR(nonnull(1)) -JEMALLOC_ATTR(visibility("default")) int -JEMALLOC_P(sallocm)(const void *ptr, size_t *rsize, int flags) +je_sallocm(const void *ptr, size_t *rsize, int flags) { size_t sz; - assert(malloc_initialized || malloc_initializer == pthread_self()); + assert(malloc_initialized || IS_INITIALIZER); -#ifdef JEMALLOC_IVSALLOC - sz = ivsalloc(ptr); -#else - assert(ptr != NULL); - sz = isalloc(ptr); -#endif + if (config_ivsalloc) + sz = ivsalloc(ptr, config_prof); + else { + assert(ptr != NULL); + sz = isalloc(ptr, config_prof); + } assert(rsize != NULL); *rsize = sz; return (ALLOCM_SUCCESS); } -JEMALLOC_ATTR(nonnull(1)) -JEMALLOC_ATTR(visibility("default")) int -JEMALLOC_P(dallocm)(void *ptr, int flags) +je_dallocm(void *ptr, int flags) { -#if (defined(JEMALLOC_PROF) || defined(JEMALLOC_STATS)) size_t usize; -#endif + size_t rzsize JEMALLOC_CC_SILENCE_INIT(0); assert(ptr != NULL); - assert(malloc_initialized || malloc_initializer == pthread_self()); + assert(malloc_initialized || IS_INITIALIZER); -#ifdef JEMALLOC_STATS - usize = isalloc(ptr); -#endif -#ifdef JEMALLOC_PROF - if (opt_prof) { -# ifndef JEMALLOC_STATS - usize = isalloc(ptr); -# endif + UTRACE(ptr, 0, 0); + if (config_stats || config_valgrind) + usize = isalloc(ptr, config_prof); + if (config_prof && opt_prof) { + if (config_stats == false && config_valgrind == false) + usize = isalloc(ptr, config_prof); prof_free(ptr, usize); } -#endif -#ifdef JEMALLOC_STATS - ALLOCATED_ADD(0, usize); -#endif - idalloc(ptr); + if (config_stats) + thread_allocated_tsd_get()->deallocated += usize; + if (config_valgrind && opt_valgrind) + rzsize = p2rz(ptr); + iqalloc(ptr); + JEMALLOC_VALGRIND_FREE(ptr, rzsize); return (ALLOCM_SUCCESS); } +int +je_nallocm(size_t *rsize, size_t size, int flags) +{ + size_t usize; + size_t alignment = (ZU(1) << (flags & ALLOCM_LG_ALIGN_MASK) + & (SIZE_T_MAX-1)); + + assert(size != 0); + + if (malloc_init()) + return (ALLOCM_ERR_OOM); + + usize = (alignment == 0) ? s2u(size) : sa2u(size, alignment); + if (usize == 0) + return (ALLOCM_ERR_OOM); + + if (rsize != NULL) + *rsize = usize; + return (ALLOCM_SUCCESS); +} + +#endif /* - * End non-standard functions. + * End experimental functions. */ /******************************************************************************/ - /* * The following functions are used by threading libraries for protection of * malloc during fork(). */ +#ifndef JEMALLOC_MUTEX_INIT_CB void jemalloc_prefork(void) +#else +JEMALLOC_EXPORT void +_malloc_prefork(void) +#endif { unsigned i; - /* Acquire all mutexes in a safe order. */ +#ifdef JEMALLOC_MUTEX_INIT_CB + if (malloc_initialized == false) + return; +#endif + assert(malloc_initialized); - malloc_mutex_lock(&arenas_lock); + /* Acquire all mutexes in a safe order. */ + malloc_mutex_prefork(&arenas_lock); for (i = 0; i < narenas; i++) { if (arenas[i] != NULL) - malloc_mutex_lock(&arenas[i]->lock); + arena_prefork(arenas[i]); } + base_prefork(); + huge_prefork(); + chunk_dss_prefork(); +} - malloc_mutex_lock(&base_mtx); - - malloc_mutex_lock(&huge_mtx); - -#ifdef JEMALLOC_DSS - malloc_mutex_lock(&dss_mtx); +#ifndef JEMALLOC_MUTEX_INIT_CB +void +jemalloc_postfork_parent(void) +#else +JEMALLOC_EXPORT void +_malloc_postfork(void) #endif +{ + unsigned i; -#ifdef JEMALLOC_SWAP - malloc_mutex_lock(&swap_mtx); +#ifdef JEMALLOC_MUTEX_INIT_CB + if (malloc_initialized == false) + return; #endif + assert(malloc_initialized); + + /* Release all mutexes, now that fork() has completed. */ + chunk_dss_postfork_parent(); + huge_postfork_parent(); + base_postfork_parent(); + for (i = 0; i < narenas; i++) { + if (arenas[i] != NULL) + arena_postfork_parent(arenas[i]); + } + malloc_mutex_postfork_parent(&arenas_lock); } void -jemalloc_postfork(void) +jemalloc_postfork_child(void) { unsigned i; + assert(malloc_initialized); + /* Release all mutexes, now that fork() has completed. */ - -#ifdef JEMALLOC_SWAP - malloc_mutex_unlock(&swap_mtx); -#endif - -#ifdef JEMALLOC_DSS - malloc_mutex_unlock(&dss_mtx); -#endif - - malloc_mutex_unlock(&huge_mtx); - - malloc_mutex_unlock(&base_mtx); - + chunk_dss_postfork_child(); + huge_postfork_child(); + base_postfork_child(); for (i = 0; i < narenas; i++) { if (arenas[i] != NULL) - malloc_mutex_unlock(&arenas[i]->lock); + arena_postfork_child(arenas[i]); } - malloc_mutex_unlock(&arenas_lock); + malloc_mutex_postfork_child(&arenas_lock); +} + +/******************************************************************************/ +/* + * The following functions are used for TLS allocation/deallocation in static + * binaries on FreeBSD. The primary difference between these and i[mcd]alloc() + * is that these avoid accessing TLS variables. + */ + +static void * +a0alloc(size_t size, bool zero) +{ + + if (malloc_init()) + return (NULL); + + if (size == 0) + size = 1; + + if (size <= arena_maxclass) + return (arena_malloc(arenas[0], size, zero, false)); + else + return (huge_malloc(size, zero)); +} + +void * +a0malloc(size_t size) +{ + + return (a0alloc(size, false)); +} + +void * +a0calloc(size_t num, size_t size) +{ + + return (a0alloc(num * size, true)); +} + +void +a0free(void *ptr) +{ + arena_chunk_t *chunk; + + if (ptr == NULL) + return; + + chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr); + if (chunk != ptr) + arena_dalloc(chunk->arena, chunk, ptr, false); + else + huge_dalloc(ptr, true); } /******************************************************************************/ diff --git a/src/mutex.c b/src/mutex.c index ca89ef1c..37a843e6 100644 --- a/src/mutex.c +++ b/src/mutex.c @@ -1,14 +1,26 @@ #define JEMALLOC_MUTEX_C_ #include "jemalloc/internal/jemalloc_internal.h" +#if defined(JEMALLOC_LAZY_LOCK) && !defined(_WIN32) +#include +#endif + +#ifndef _CRT_SPINCOUNT +#define _CRT_SPINCOUNT 4000 +#endif + /******************************************************************************/ /* Data. */ #ifdef JEMALLOC_LAZY_LOCK bool isthreaded = false; #endif +#ifdef JEMALLOC_MUTEX_INIT_CB +static bool postpone_init = true; +static malloc_mutex_t *postponed_mutexes = NULL; +#endif -#ifdef JEMALLOC_LAZY_LOCK +#if defined(JEMALLOC_LAZY_LOCK) && !defined(_WIN32) static void pthread_create_once(void); #endif @@ -18,7 +30,7 @@ static void pthread_create_once(void); * process goes multi-threaded. */ -#ifdef JEMALLOC_LAZY_LOCK +#if defined(JEMALLOC_LAZY_LOCK) && !defined(_WIN32) static int (*pthread_create_fptr)(pthread_t *__restrict, const pthread_attr_t *, void *(*)(void *), void *__restrict); @@ -36,8 +48,7 @@ pthread_create_once(void) isthreaded = true; } -JEMALLOC_ATTR(visibility("default")) -int +JEMALLOC_EXPORT int pthread_create(pthread_t *__restrict thread, const pthread_attr_t *__restrict attr, void *(*start_routine)(void *), void *__restrict arg) @@ -52,39 +63,87 @@ pthread_create(pthread_t *__restrict thread, /******************************************************************************/ +#ifdef JEMALLOC_MUTEX_INIT_CB +int _pthread_mutex_init_calloc_cb(pthread_mutex_t *mutex, + void *(calloc_cb)(size_t, size_t)); +#endif + bool malloc_mutex_init(malloc_mutex_t *mutex) { -#ifdef JEMALLOC_OSSPIN - *mutex = 0; + +#ifdef _WIN32 + if (!InitializeCriticalSectionAndSpinCount(&mutex->lock, + _CRT_SPINCOUNT)) + return (true); +#elif (defined(JEMALLOC_OSSPIN)) + mutex->lock = 0; +#elif (defined(JEMALLOC_MUTEX_INIT_CB)) + if (postpone_init) { + mutex->postponed_next = postponed_mutexes; + postponed_mutexes = mutex; + } else { + if (_pthread_mutex_init_calloc_cb(&mutex->lock, base_calloc) != + 0) + return (true); + } #else pthread_mutexattr_t attr; if (pthread_mutexattr_init(&attr) != 0) return (true); -#ifdef PTHREAD_MUTEX_ADAPTIVE_NP - pthread_mutexattr_settype(&attr, PTHREAD_MUTEX_ADAPTIVE_NP); -#else - pthread_mutexattr_settype(&attr, PTHREAD_MUTEX_DEFAULT); -#endif - if (pthread_mutex_init(mutex, &attr) != 0) { + pthread_mutexattr_settype(&attr, MALLOC_MUTEX_TYPE); + if (pthread_mutex_init(&mutex->lock, &attr) != 0) { pthread_mutexattr_destroy(&attr); return (true); } pthread_mutexattr_destroy(&attr); - #endif return (false); } void -malloc_mutex_destroy(malloc_mutex_t *mutex) +malloc_mutex_prefork(malloc_mutex_t *mutex) { -#ifndef JEMALLOC_OSSPIN - if (pthread_mutex_destroy(mutex) != 0) { - malloc_write(": Error in pthread_mutex_destroy()\n"); - abort(); + malloc_mutex_lock(mutex); +} + +void +malloc_mutex_postfork_parent(malloc_mutex_t *mutex) +{ + + malloc_mutex_unlock(mutex); +} + +void +malloc_mutex_postfork_child(malloc_mutex_t *mutex) +{ + +#ifdef JEMALLOC_MUTEX_INIT_CB + malloc_mutex_unlock(mutex); +#else + if (malloc_mutex_init(mutex)) { + malloc_printf(": Error re-initializing mutex in " + "child\n"); + if (opt_abort) + abort(); } #endif } + +bool +mutex_boot(void) +{ + +#ifdef JEMALLOC_MUTEX_INIT_CB + postpone_init = false; + while (postponed_mutexes != NULL) { + if (_pthread_mutex_init_calloc_cb(&postponed_mutexes->lock, + base_calloc) != 0) + return (true); + postponed_mutexes = postponed_mutexes->postponed_next; + } +#endif + return (false); +} diff --git a/src/prof.c b/src/prof.c index 8a144b4e..de1d3929 100644 --- a/src/prof.c +++ b/src/prof.c @@ -1,6 +1,5 @@ #define JEMALLOC_PROF_C_ #include "jemalloc/internal/jemalloc_internal.h" -#ifdef JEMALLOC_PROF /******************************************************************************/ #ifdef JEMALLOC_PROF_LIBUNWIND @@ -15,27 +14,30 @@ /******************************************************************************/ /* Data. */ +malloc_tsd_data(, prof_tdata, prof_tdata_t *, NULL) + bool opt_prof = false; bool opt_prof_active = true; -size_t opt_lg_prof_bt_max = LG_PROF_BT_MAX_DEFAULT; size_t opt_lg_prof_sample = LG_PROF_SAMPLE_DEFAULT; ssize_t opt_lg_prof_interval = LG_PROF_INTERVAL_DEFAULT; bool opt_prof_gdump = false; +bool opt_prof_final = true; bool opt_prof_leak = false; -bool opt_prof_accum = true; -ssize_t opt_lg_prof_tcmax = LG_PROF_TCMAX_DEFAULT; +bool opt_prof_accum = false; char opt_prof_prefix[PATH_MAX + 1]; uint64_t prof_interval; bool prof_promote; -unsigned prof_bt_max; - -#ifndef NO_TLS -__thread prof_tdata_t *prof_tdata_tls - JEMALLOC_ATTR(tls_model("initial-exec")); -#endif -pthread_key_t prof_tdata_tsd; +/* + * Table of mutexes that are shared among ctx's. These are leaf locks, so + * there is no problem with using them for more than one ctx at the same time. + * The primary motivation for this sharing though is that ctx's are ephemeral, + * and destroying mutexes causes complications for systems that allocate when + * creating/destroying mutexes. + */ +static malloc_mutex_t *ctx_locks; +static unsigned cum_ctxs; /* Atomic counter. */ /* * Global hash of (prof_bt_t *)-->(prof_ctx_t *). This is the master data @@ -55,18 +57,13 @@ static uint64_t prof_dump_useq; * all profile dumps. The buffer is implicitly protected by bt2ctx_mtx, since * it must be locked anyway during dumping. */ -static char prof_dump_buf[PROF_DUMP_BUF_SIZE]; +static char prof_dump_buf[PROF_DUMP_BUFSIZE]; static unsigned prof_dump_buf_end; static int prof_dump_fd; /* Do not dump any profiles until bootstrapping is complete. */ static bool prof_booted = false; -static malloc_mutex_t enq_mtx; -static bool enq; -static bool enq_idump; -static bool enq_gdump; - /******************************************************************************/ /* Function prototypes for non-inline static functions. */ @@ -79,22 +76,24 @@ static _Unwind_Reason_Code prof_unwind_callback( struct _Unwind_Context *context, void *arg); #endif static bool prof_flush(bool propagate_err); -static bool prof_write(const char *s, bool propagate_err); +static bool prof_write(bool propagate_err, const char *s); +static bool prof_printf(bool propagate_err, const char *format, ...) + JEMALLOC_ATTR(format(printf, 2, 3)); static void prof_ctx_sum(prof_ctx_t *ctx, prof_cnt_t *cnt_all, size_t *leak_nctx); static void prof_ctx_destroy(prof_ctx_t *ctx); static void prof_ctx_merge(prof_ctx_t *ctx, prof_thr_cnt_t *cnt); -static bool prof_dump_ctx(prof_ctx_t *ctx, prof_bt_t *bt, - bool propagate_err); +static bool prof_dump_ctx(bool propagate_err, prof_ctx_t *ctx, + prof_bt_t *bt); static bool prof_dump_maps(bool propagate_err); -static bool prof_dump(const char *filename, bool leakcheck, - bool propagate_err); +static bool prof_dump(bool propagate_err, const char *filename, + bool leakcheck); static void prof_dump_filename(char *filename, char v, int64_t vseq); static void prof_fdump(void); static void prof_bt_hash(const void *key, unsigned minbits, size_t *hash1, size_t *hash2); static bool prof_bt_keycomp(const void *k1, const void *k2); -static void prof_tdata_cleanup(void *arg); +static malloc_mutex_t *prof_ctx_mutex_choose(void); /******************************************************************************/ @@ -102,6 +101,8 @@ void bt_init(prof_bt_t *bt, void **vec) { + cassert(config_prof); + bt->vec = vec; bt->len = 0; } @@ -110,6 +111,8 @@ static void bt_destroy(prof_bt_t *bt) { + cassert(config_prof); + idalloc(bt); } @@ -118,6 +121,8 @@ bt_dup(prof_bt_t *bt) { prof_bt_t *ret; + cassert(config_prof); + /* * Create a single allocation that has space for vec immediately * following the prof_bt_t structure. The backtraces that get @@ -138,30 +143,32 @@ bt_dup(prof_bt_t *bt) } static inline void -prof_enter(void) +prof_enter(prof_tdata_t *prof_tdata) { - malloc_mutex_lock(&enq_mtx); - enq = true; - malloc_mutex_unlock(&enq_mtx); + cassert(config_prof); + + assert(prof_tdata->enq == false); + prof_tdata->enq = true; malloc_mutex_lock(&bt2ctx_mtx); } static inline void -prof_leave(void) +prof_leave(prof_tdata_t *prof_tdata) { bool idump, gdump; + cassert(config_prof); + malloc_mutex_unlock(&bt2ctx_mtx); - malloc_mutex_lock(&enq_mtx); - enq = false; - idump = enq_idump; - enq_idump = false; - gdump = enq_gdump; - enq_gdump = false; - malloc_mutex_unlock(&enq_mtx); + assert(prof_tdata->enq); + prof_tdata->enq = false; + idump = prof_tdata->enq_idump; + prof_tdata->enq_idump = false; + gdump = prof_tdata->enq_gdump; + prof_tdata->enq_gdump = false; if (idump) prof_idump(); @@ -171,16 +178,16 @@ prof_leave(void) #ifdef JEMALLOC_PROF_LIBUNWIND void -prof_backtrace(prof_bt_t *bt, unsigned nignore, unsigned max) +prof_backtrace(prof_bt_t *bt, unsigned nignore) { unw_context_t uc; unw_cursor_t cursor; unsigned i; int err; + cassert(config_prof); assert(bt->len == 0); assert(bt->vec != NULL); - assert(max <= (1U << opt_lg_prof_bt_max)); unw_getcontext(&uc); unw_init_local(&cursor, &uc); @@ -196,7 +203,7 @@ prof_backtrace(prof_bt_t *bt, unsigned nignore, unsigned max) * Iterate over stack frames until there are no more, or until no space * remains in bt. */ - for (i = 0; i < max; i++) { + for (i = 0; i < PROF_BT_MAX; i++) { unw_get_reg(&cursor, UNW_REG_IP, (unw_word_t *)&bt->vec[i]); bt->len++; err = unw_step(&cursor); @@ -204,12 +211,13 @@ prof_backtrace(prof_bt_t *bt, unsigned nignore, unsigned max) break; } } -#endif -#ifdef JEMALLOC_PROF_LIBGCC +#elif (defined(JEMALLOC_PROF_LIBGCC)) static _Unwind_Reason_Code prof_unwind_init_callback(struct _Unwind_Context *context, void *arg) { + cassert(config_prof); + return (_URC_NO_REASON); } @@ -218,6 +226,8 @@ prof_unwind_callback(struct _Unwind_Context *context, void *arg) { prof_unwind_data_t *data = (prof_unwind_data_t *)arg; + cassert(config_prof); + if (data->nignore > 0) data->nignore--; else { @@ -231,19 +241,20 @@ prof_unwind_callback(struct _Unwind_Context *context, void *arg) } void -prof_backtrace(prof_bt_t *bt, unsigned nignore, unsigned max) +prof_backtrace(prof_bt_t *bt, unsigned nignore) { - prof_unwind_data_t data = {bt, nignore, max}; + prof_unwind_data_t data = {bt, nignore, PROF_BT_MAX}; + + cassert(config_prof); _Unwind_Backtrace(prof_unwind_callback, &data); } -#endif -#ifdef JEMALLOC_PROF_GCC +#elif (defined(JEMALLOC_PROF_GCC)) void -prof_backtrace(prof_bt_t *bt, unsigned nignore, unsigned max) +prof_backtrace(prof_bt_t *bt, unsigned nignore) { #define BT_FRAME(i) \ - if ((i) < nignore + max) { \ + if ((i) < nignore + PROF_BT_MAX) { \ void *p; \ if (__builtin_frame_address(i) == 0) \ return; \ @@ -257,8 +268,8 @@ prof_backtrace(prof_bt_t *bt, unsigned nignore, unsigned max) } else \ return; + cassert(config_prof); assert(nignore <= 3); - assert(max <= (1U << opt_lg_prof_bt_max)); BT_FRAME(0) BT_FRAME(1) @@ -407,6 +418,14 @@ prof_backtrace(prof_bt_t *bt, unsigned nignore, unsigned max) BT_FRAME(130) #undef BT_FRAME } +#else +void +prof_backtrace(prof_bt_t *bt, unsigned nignore) +{ + + cassert(config_prof); + assert(false); +} #endif prof_thr_cnt_t * @@ -418,12 +437,11 @@ prof_lookup(prof_bt_t *bt) } ret; prof_tdata_t *prof_tdata; - prof_tdata = PROF_TCACHE_GET(); - if (prof_tdata == NULL) { - prof_tdata = prof_tdata_init(); - if (prof_tdata == NULL) - return (NULL); - } + cassert(config_prof); + + prof_tdata = prof_tdata_get(); + if ((uintptr_t)prof_tdata <= (uintptr_t)PROF_TDATA_STATE_MAX) + return (NULL); if (ckh_search(&prof_tdata->bt2cnt, bt, NULL, &ret.v)) { union { @@ -440,62 +458,51 @@ prof_lookup(prof_bt_t *bt) * This thread's cache lacks bt. Look for it in the global * cache. */ - prof_enter(); + prof_enter(prof_tdata); if (ckh_search(&bt2ctx, bt, &btkey.v, &ctx.v)) { /* bt has never been seen before. Insert it. */ ctx.v = imalloc(sizeof(prof_ctx_t)); if (ctx.v == NULL) { - prof_leave(); + prof_leave(prof_tdata); return (NULL); } btkey.p = bt_dup(bt); if (btkey.v == NULL) { - prof_leave(); + prof_leave(prof_tdata); idalloc(ctx.v); return (NULL); } ctx.p->bt = btkey.p; - if (malloc_mutex_init(&ctx.p->lock)) { - prof_leave(); - idalloc(btkey.v); - idalloc(ctx.v); - return (NULL); - } + ctx.p->lock = prof_ctx_mutex_choose(); + /* + * Set nlimbo to 1, in order to avoid a race condition + * with prof_ctx_merge()/prof_ctx_destroy(). + */ + ctx.p->nlimbo = 1; memset(&ctx.p->cnt_merged, 0, sizeof(prof_cnt_t)); ql_new(&ctx.p->cnts_ql); if (ckh_insert(&bt2ctx, btkey.v, ctx.v)) { /* OOM. */ - prof_leave(); - malloc_mutex_destroy(&ctx.p->lock); + prof_leave(prof_tdata); idalloc(btkey.v); idalloc(ctx.v); return (NULL); } - /* - * Artificially raise curobjs, in order to avoid a race - * condition with prof_ctx_merge()/prof_ctx_destroy(). - * - * No locking is necessary for ctx here because no other - * threads have had the opportunity to fetch it from - * bt2ctx yet. - */ - ctx.p->cnt_merged.curobjs++; new_ctx = true; } else { /* - * Artificially raise curobjs, in order to avoid a race - * condition with prof_ctx_merge()/prof_ctx_destroy(). + * Increment nlimbo, in order to avoid a race condition + * with prof_ctx_merge()/prof_ctx_destroy(). */ - malloc_mutex_lock(&ctx.p->lock); - ctx.p->cnt_merged.curobjs++; - malloc_mutex_unlock(&ctx.p->lock); + malloc_mutex_lock(ctx.p->lock); + ctx.p->nlimbo++; + malloc_mutex_unlock(ctx.p->lock); new_ctx = false; } - prof_leave(); + prof_leave(prof_tdata); /* Link a prof_thd_cnt_t into ctx for this thread. */ - if (opt_lg_prof_tcmax >= 0 && ckh_count(&prof_tdata->bt2cnt) - == (ZU(1) << opt_lg_prof_tcmax)) { + if (ckh_count(&prof_tdata->bt2cnt) == PROF_TCMAX) { assert(ckh_count(&prof_tdata->bt2cnt) > 0); /* * Flush the least recently used cnt in order to keep @@ -510,9 +517,7 @@ prof_lookup(prof_bt_t *bt) prof_ctx_merge(ret.p->ctx, ret.p); /* ret can now be re-used. */ } else { - assert(opt_lg_prof_tcmax < 0 || - ckh_count(&prof_tdata->bt2cnt) < (ZU(1) << - opt_lg_prof_tcmax)); + assert(ckh_count(&prof_tdata->bt2cnt) < PROF_TCMAX); /* Allocate and partially initialize a new cnt. */ ret.v = imalloc(sizeof(prof_thr_cnt_t)); if (ret.p == NULL) { @@ -534,10 +539,10 @@ prof_lookup(prof_bt_t *bt) return (NULL); } ql_head_insert(&prof_tdata->lru_ql, ret.p, lru_link); - malloc_mutex_lock(&ctx.p->lock); + malloc_mutex_lock(ctx.p->lock); ql_tail_insert(&ctx.p->cnts_ql, ret.p, cnts_link); - ctx.p->cnt_merged.curobjs--; - malloc_mutex_unlock(&ctx.p->lock); + ctx.p->nlimbo--; + malloc_mutex_unlock(ctx.p->lock); } else { /* Move ret to the front of the LRU. */ ql_remove(&prof_tdata->lru_ql, ret.p, lru_link); @@ -553,6 +558,8 @@ prof_flush(bool propagate_err) bool ret = false; ssize_t err; + cassert(config_prof); + err = write(prof_dump_fd, prof_dump_buf, prof_dump_buf_end); if (err == -1) { if (propagate_err == false) { @@ -569,24 +576,26 @@ prof_flush(bool propagate_err) } static bool -prof_write(const char *s, bool propagate_err) +prof_write(bool propagate_err, const char *s) { unsigned i, slen, n; + cassert(config_prof); + i = 0; slen = strlen(s); while (i < slen) { /* Flush the buffer if it is full. */ - if (prof_dump_buf_end == PROF_DUMP_BUF_SIZE) + if (prof_dump_buf_end == PROF_DUMP_BUFSIZE) if (prof_flush(propagate_err) && propagate_err) return (true); - if (prof_dump_buf_end + slen <= PROF_DUMP_BUF_SIZE) { + if (prof_dump_buf_end + slen <= PROF_DUMP_BUFSIZE) { /* Finish writing. */ n = slen - i; } else { /* Write as much of s as will fit. */ - n = PROF_DUMP_BUF_SIZE - prof_dump_buf_end; + n = PROF_DUMP_BUFSIZE - prof_dump_buf_end; } memcpy(&prof_dump_buf[prof_dump_buf_end], &s[i], n); prof_dump_buf_end += n; @@ -596,13 +605,31 @@ prof_write(const char *s, bool propagate_err) return (false); } +JEMALLOC_ATTR(format(printf, 2, 3)) +static bool +prof_printf(bool propagate_err, const char *format, ...) +{ + bool ret; + va_list ap; + char buf[PROF_PRINTF_BUFSIZE]; + + va_start(ap, format); + malloc_vsnprintf(buf, sizeof(buf), format, ap); + va_end(ap); + ret = prof_write(propagate_err, buf); + + return (ret); +} + static void prof_ctx_sum(prof_ctx_t *ctx, prof_cnt_t *cnt_all, size_t *leak_nctx) { prof_thr_cnt_t *thr_cnt; prof_cnt_t tcnt; - malloc_mutex_lock(&ctx->lock); + cassert(config_prof); + + malloc_mutex_lock(ctx->lock); memcpy(&ctx->cnt_summed, &ctx->cnt_merged, sizeof(prof_cnt_t)); ql_foreach(thr_cnt, &ctx->cnts_ql, cnts_link) { @@ -641,43 +668,48 @@ prof_ctx_sum(prof_ctx_t *ctx, prof_cnt_t *cnt_all, size_t *leak_nctx) cnt_all->accumbytes += ctx->cnt_summed.accumbytes; } - malloc_mutex_unlock(&ctx->lock); + malloc_mutex_unlock(ctx->lock); } static void prof_ctx_destroy(prof_ctx_t *ctx) { + prof_tdata_t *prof_tdata; + + cassert(config_prof); /* * Check that ctx is still unused by any thread cache before destroying - * it. prof_lookup() artificially raises ctx->cnt_merge.curobjs in - * order to avoid a race condition with this function, as does - * prof_ctx_merge() in order to avoid a race between the main body of - * prof_ctx_merge() and entry into this function. + * it. prof_lookup() increments ctx->nlimbo in order to avoid a race + * condition with this function, as does prof_ctx_merge() in order to + * avoid a race between the main body of prof_ctx_merge() and entry + * into this function. */ - prof_enter(); - malloc_mutex_lock(&ctx->lock); - if (ql_first(&ctx->cnts_ql) == NULL && ctx->cnt_merged.curobjs == 1) { + prof_tdata = *prof_tdata_tsd_get(); + assert((uintptr_t)prof_tdata > (uintptr_t)PROF_TDATA_STATE_MAX); + prof_enter(prof_tdata); + malloc_mutex_lock(ctx->lock); + if (ql_first(&ctx->cnts_ql) == NULL && ctx->cnt_merged.curobjs == 0 && + ctx->nlimbo == 1) { assert(ctx->cnt_merged.curbytes == 0); assert(ctx->cnt_merged.accumobjs == 0); assert(ctx->cnt_merged.accumbytes == 0); /* Remove ctx from bt2ctx. */ if (ckh_remove(&bt2ctx, ctx->bt, NULL, NULL)) assert(false); - prof_leave(); + prof_leave(prof_tdata); /* Destroy ctx. */ - malloc_mutex_unlock(&ctx->lock); + malloc_mutex_unlock(ctx->lock); bt_destroy(ctx->bt); - malloc_mutex_destroy(&ctx->lock); idalloc(ctx); } else { /* * Compensate for increment in prof_ctx_merge() or * prof_lookup(). */ - ctx->cnt_merged.curobjs--; - malloc_mutex_unlock(&ctx->lock); - prof_leave(); + ctx->nlimbo--; + malloc_mutex_unlock(ctx->lock); + prof_leave(prof_tdata); } } @@ -686,20 +718,22 @@ prof_ctx_merge(prof_ctx_t *ctx, prof_thr_cnt_t *cnt) { bool destroy; + cassert(config_prof); + /* Merge cnt stats and detach from ctx. */ - malloc_mutex_lock(&ctx->lock); + malloc_mutex_lock(ctx->lock); ctx->cnt_merged.curobjs += cnt->cnts.curobjs; ctx->cnt_merged.curbytes += cnt->cnts.curbytes; ctx->cnt_merged.accumobjs += cnt->cnts.accumobjs; ctx->cnt_merged.accumbytes += cnt->cnts.accumbytes; ql_remove(&ctx->cnts_ql, cnt, cnts_link); if (opt_prof_accum == false && ql_first(&ctx->cnts_ql) == NULL && - ctx->cnt_merged.curobjs == 0) { + ctx->cnt_merged.curobjs == 0 && ctx->nlimbo == 0) { /* - * Artificially raise ctx->cnt_merged.curobjs in order to keep - * another thread from winning the race to destroy ctx while - * this one has ctx->lock dropped. Without this, it would be - * possible for another thread to: + * Increment ctx->nlimbo in order to keep another thread from + * winning the race to destroy ctx while this one has ctx->lock + * dropped. Without this, it would be possible for another + * thread to: * * 1) Sample an allocation associated with ctx. * 2) Deallocate the sampled object. @@ -708,49 +742,51 @@ prof_ctx_merge(prof_ctx_t *ctx, prof_thr_cnt_t *cnt) * The result would be that ctx no longer exists by the time * this thread accesses it in prof_ctx_destroy(). */ - ctx->cnt_merged.curobjs++; + ctx->nlimbo++; destroy = true; } else destroy = false; - malloc_mutex_unlock(&ctx->lock); + malloc_mutex_unlock(ctx->lock); if (destroy) prof_ctx_destroy(ctx); } static bool -prof_dump_ctx(prof_ctx_t *ctx, prof_bt_t *bt, bool propagate_err) +prof_dump_ctx(bool propagate_err, prof_ctx_t *ctx, prof_bt_t *bt) { - char buf[UMAX2S_BUFSIZE]; unsigned i; - if (opt_prof_accum == false && ctx->cnt_summed.curobjs == 0) { + cassert(config_prof); + + /* + * Current statistics can sum to 0 as a result of unmerged per thread + * statistics. Additionally, interval- and growth-triggered dumps can + * occur between the time a ctx is created and when its statistics are + * filled in. Avoid dumping any ctx that is an artifact of either + * implementation detail. + */ + if ((opt_prof_accum == false && ctx->cnt_summed.curobjs == 0) || + (opt_prof_accum && ctx->cnt_summed.accumobjs == 0)) { + assert(ctx->cnt_summed.curobjs == 0); assert(ctx->cnt_summed.curbytes == 0); assert(ctx->cnt_summed.accumobjs == 0); assert(ctx->cnt_summed.accumbytes == 0); return (false); } - if (prof_write(u2s(ctx->cnt_summed.curobjs, 10, buf), propagate_err) - || prof_write(": ", propagate_err) - || prof_write(u2s(ctx->cnt_summed.curbytes, 10, buf), - propagate_err) - || prof_write(" [", propagate_err) - || prof_write(u2s(ctx->cnt_summed.accumobjs, 10, buf), - propagate_err) - || prof_write(": ", propagate_err) - || prof_write(u2s(ctx->cnt_summed.accumbytes, 10, buf), - propagate_err) - || prof_write("] @", propagate_err)) + if (prof_printf(propagate_err, "%"PRId64": %"PRId64 + " [%"PRIu64": %"PRIu64"] @", + ctx->cnt_summed.curobjs, ctx->cnt_summed.curbytes, + ctx->cnt_summed.accumobjs, ctx->cnt_summed.accumbytes)) return (true); for (i = 0; i < bt->len; i++) { - if (prof_write(" 0x", propagate_err) - || prof_write(u2s((uintptr_t)bt->vec[i], 16, buf), - propagate_err)) + if (prof_printf(propagate_err, " %#"PRIxPTR, + (uintptr_t)bt->vec[i])) return (true); } - if (prof_write("\n", propagate_err)) + if (prof_write(propagate_err, "\n")) return (true); return (false); @@ -760,49 +796,29 @@ static bool prof_dump_maps(bool propagate_err) { int mfd; - char buf[UMAX2S_BUFSIZE]; - char *s; - unsigned i, slen; - /* /proc//maps\0 */ - char mpath[6 + UMAX2S_BUFSIZE - + 5 + 1]; + char filename[PATH_MAX + 1]; - i = 0; + cassert(config_prof); - s = "/proc/"; - slen = strlen(s); - memcpy(&mpath[i], s, slen); - i += slen; - - s = u2s(getpid(), 10, buf); - slen = strlen(s); - memcpy(&mpath[i], s, slen); - i += slen; - - s = "/maps"; - slen = strlen(s); - memcpy(&mpath[i], s, slen); - i += slen; - - mpath[i] = '\0'; - - mfd = open(mpath, O_RDONLY); + malloc_snprintf(filename, sizeof(filename), "/proc/%d/maps", + (int)getpid()); + mfd = open(filename, O_RDONLY); if (mfd != -1) { ssize_t nread; - if (prof_write("\nMAPPED_LIBRARIES:\n", propagate_err) && + if (prof_write(propagate_err, "\nMAPPED_LIBRARIES:\n") && propagate_err) return (true); nread = 0; do { prof_dump_buf_end += nread; - if (prof_dump_buf_end == PROF_DUMP_BUF_SIZE) { + if (prof_dump_buf_end == PROF_DUMP_BUFSIZE) { /* Make space in prof_dump_buf before read(). */ if (prof_flush(propagate_err) && propagate_err) return (true); } nread = read(mfd, &prof_dump_buf[prof_dump_buf_end], - PROF_DUMP_BUF_SIZE - prof_dump_buf_end); + PROF_DUMP_BUFSIZE - prof_dump_buf_end); } while (nread > 0); close(mfd); } else @@ -812,8 +828,9 @@ prof_dump_maps(bool propagate_err) } static bool -prof_dump(const char *filename, bool leakcheck, bool propagate_err) +prof_dump(bool propagate_err, const char *filename, bool leakcheck) { + prof_tdata_t *prof_tdata; prof_cnt_t cnt_all; size_t tabind; union { @@ -824,20 +841,24 @@ prof_dump(const char *filename, bool leakcheck, bool propagate_err) prof_ctx_t *p; void *v; } ctx; - char buf[UMAX2S_BUFSIZE]; size_t leak_nctx; - prof_enter(); + cassert(config_prof); + + prof_tdata = prof_tdata_get(); + if ((uintptr_t)prof_tdata <= (uintptr_t)PROF_TDATA_STATE_MAX) + return (true); + prof_enter(prof_tdata); prof_dump_fd = creat(filename, 0644); if (prof_dump_fd == -1) { if (propagate_err == false) { - malloc_write(": creat(\""); - malloc_write(filename); - malloc_write("\", 0644) failed\n"); + malloc_printf( + ": creat(\"%s\"), 0644) failed\n", + filename); if (opt_abort) abort(); } - goto ERROR; + goto label_error; } /* Merge per thread profile stats, and sum them in cnt_all. */ @@ -847,131 +868,75 @@ prof_dump(const char *filename, bool leakcheck, bool propagate_err) prof_ctx_sum(ctx.p, &cnt_all, &leak_nctx); /* Dump profile header. */ - if (prof_write("heap profile: ", propagate_err) - || prof_write(u2s(cnt_all.curobjs, 10, buf), propagate_err) - || prof_write(": ", propagate_err) - || prof_write(u2s(cnt_all.curbytes, 10, buf), propagate_err) - || prof_write(" [", propagate_err) - || prof_write(u2s(cnt_all.accumobjs, 10, buf), propagate_err) - || prof_write(": ", propagate_err) - || prof_write(u2s(cnt_all.accumbytes, 10, buf), propagate_err)) - goto ERROR; - if (opt_lg_prof_sample == 0) { - if (prof_write("] @ heapprofile\n", propagate_err)) - goto ERROR; + if (prof_printf(propagate_err, + "heap profile: %"PRId64": %"PRId64 + " [%"PRIu64": %"PRIu64"] @ heapprofile\n", + cnt_all.curobjs, cnt_all.curbytes, + cnt_all.accumobjs, cnt_all.accumbytes)) + goto label_error; } else { - if (prof_write("] @ heap_v2/", propagate_err) - || prof_write(u2s((uint64_t)1U << opt_lg_prof_sample, 10, - buf), propagate_err) - || prof_write("\n", propagate_err)) - goto ERROR; + if (prof_printf(propagate_err, + "heap profile: %"PRId64": %"PRId64 + " [%"PRIu64": %"PRIu64"] @ heap_v2/%"PRIu64"\n", + cnt_all.curobjs, cnt_all.curbytes, + cnt_all.accumobjs, cnt_all.accumbytes, + ((uint64_t)1U << opt_lg_prof_sample))) + goto label_error; } /* Dump per ctx profile stats. */ for (tabind = 0; ckh_iter(&bt2ctx, &tabind, &bt.v, &ctx.v) == false;) { - if (prof_dump_ctx(ctx.p, bt.p, propagate_err)) - goto ERROR; + if (prof_dump_ctx(propagate_err, ctx.p, bt.p)) + goto label_error; } /* Dump /proc//maps if possible. */ if (prof_dump_maps(propagate_err)) - goto ERROR; + goto label_error; if (prof_flush(propagate_err)) - goto ERROR; + goto label_error; close(prof_dump_fd); - prof_leave(); + prof_leave(prof_tdata); if (leakcheck && cnt_all.curbytes != 0) { - malloc_write(": Leak summary: "); - malloc_write(u2s(cnt_all.curbytes, 10, buf)); - malloc_write((cnt_all.curbytes != 1) ? " bytes, " : " byte, "); - malloc_write(u2s(cnt_all.curobjs, 10, buf)); - malloc_write((cnt_all.curobjs != 1) ? " objects, " : - " object, "); - malloc_write(u2s(leak_nctx, 10, buf)); - malloc_write((leak_nctx != 1) ? " contexts\n" : " context\n"); - malloc_write(": Run pprof on \""); - malloc_write(filename); - malloc_write("\" for leak detail\n"); + malloc_printf(": Leak summary: %"PRId64" byte%s, %" + PRId64" object%s, %zu context%s\n", + cnt_all.curbytes, (cnt_all.curbytes != 1) ? "s" : "", + cnt_all.curobjs, (cnt_all.curobjs != 1) ? "s" : "", + leak_nctx, (leak_nctx != 1) ? "s" : ""); + malloc_printf( + ": Run pprof on \"%s\" for leak detail\n", + filename); } return (false); -ERROR: - prof_leave(); +label_error: + prof_leave(prof_tdata); return (true); } -#define DUMP_FILENAME_BUFSIZE (PATH_MAX+ UMAX2S_BUFSIZE \ - + 1 \ - + UMAX2S_BUFSIZE \ - + 2 \ - + UMAX2S_BUFSIZE \ - + 5 + 1) +#define DUMP_FILENAME_BUFSIZE (PATH_MAX + 1) static void prof_dump_filename(char *filename, char v, int64_t vseq) { - char buf[UMAX2S_BUFSIZE]; - char *s; - unsigned i, slen; - /* - * Construct a filename of the form: - * - * ...v.heap\0 - */ + cassert(config_prof); - i = 0; - - s = opt_prof_prefix; - slen = strlen(s); - memcpy(&filename[i], s, slen); - i += slen; - - s = "."; - slen = strlen(s); - memcpy(&filename[i], s, slen); - i += slen; - - s = u2s(getpid(), 10, buf); - slen = strlen(s); - memcpy(&filename[i], s, slen); - i += slen; - - s = "."; - slen = strlen(s); - memcpy(&filename[i], s, slen); - i += slen; - - s = u2s(prof_dump_seq, 10, buf); - prof_dump_seq++; - slen = strlen(s); - memcpy(&filename[i], s, slen); - i += slen; - - s = "."; - slen = strlen(s); - memcpy(&filename[i], s, slen); - i += slen; - - filename[i] = v; - i++; - - if (vseq != 0xffffffffffffffffLLU) { - s = u2s(vseq, 10, buf); - slen = strlen(s); - memcpy(&filename[i], s, slen); - i += slen; + if (vseq != UINT64_C(0xffffffffffffffff)) { + /* "...v.heap" */ + malloc_snprintf(filename, DUMP_FILENAME_BUFSIZE, + "%s.%d.%"PRIu64".%c%"PRId64".heap", + opt_prof_prefix, (int)getpid(), prof_dump_seq, v, vseq); + } else { + /* "....heap" */ + malloc_snprintf(filename, DUMP_FILENAME_BUFSIZE, + "%s.%d.%"PRIu64".%c.heap", + opt_prof_prefix, (int)getpid(), prof_dump_seq, v); } - - s = ".heap"; - slen = strlen(s); - memcpy(&filename[i], s, slen); - i += slen; - - filename[i] = '\0'; + prof_dump_seq++; } static void @@ -979,38 +944,47 @@ prof_fdump(void) { char filename[DUMP_FILENAME_BUFSIZE]; + cassert(config_prof); + if (prof_booted == false) return; - if (opt_prof_prefix[0] != '\0') { + if (opt_prof_final && opt_prof_prefix[0] != '\0') { malloc_mutex_lock(&prof_dump_seq_mtx); - prof_dump_filename(filename, 'f', 0xffffffffffffffffLLU); + prof_dump_filename(filename, 'f', UINT64_C(0xffffffffffffffff)); malloc_mutex_unlock(&prof_dump_seq_mtx); - prof_dump(filename, opt_prof_leak, false); + prof_dump(false, filename, opt_prof_leak); } } void prof_idump(void) { - char filename[DUMP_FILENAME_BUFSIZE]; + prof_tdata_t *prof_tdata; + char filename[PATH_MAX + 1]; + + cassert(config_prof); if (prof_booted == false) return; - malloc_mutex_lock(&enq_mtx); - if (enq) { - enq_idump = true; - malloc_mutex_unlock(&enq_mtx); + /* + * Don't call prof_tdata_get() here, because it could cause recursive + * allocation. + */ + prof_tdata = *prof_tdata_tsd_get(); + if ((uintptr_t)prof_tdata <= (uintptr_t)PROF_TDATA_STATE_MAX) + return; + if (prof_tdata->enq) { + prof_tdata->enq_idump = true; return; } - malloc_mutex_unlock(&enq_mtx); if (opt_prof_prefix[0] != '\0') { malloc_mutex_lock(&prof_dump_seq_mtx); prof_dump_filename(filename, 'i', prof_dump_iseq); prof_dump_iseq++; malloc_mutex_unlock(&prof_dump_seq_mtx); - prof_dump(filename, false, false); + prof_dump(false, filename, false); } } @@ -1019,6 +993,8 @@ prof_mdump(const char *filename) { char filename_buf[DUMP_FILENAME_BUFSIZE]; + cassert(config_prof); + if (opt_prof == false || prof_booted == false) return (true); @@ -1032,30 +1008,37 @@ prof_mdump(const char *filename) malloc_mutex_unlock(&prof_dump_seq_mtx); filename = filename_buf; } - return (prof_dump(filename, false, true)); + return (prof_dump(true, filename, false)); } void prof_gdump(void) { + prof_tdata_t *prof_tdata; char filename[DUMP_FILENAME_BUFSIZE]; + cassert(config_prof); + if (prof_booted == false) return; - malloc_mutex_lock(&enq_mtx); - if (enq) { - enq_gdump = true; - malloc_mutex_unlock(&enq_mtx); + /* + * Don't call prof_tdata_get() here, because it could cause recursive + * allocation. + */ + prof_tdata = *prof_tdata_tsd_get(); + if ((uintptr_t)prof_tdata <= (uintptr_t)PROF_TDATA_STATE_MAX) + return; + if (prof_tdata->enq) { + prof_tdata->enq_gdump = true; return; } - malloc_mutex_unlock(&enq_mtx); if (opt_prof_prefix[0] != '\0') { malloc_mutex_lock(&prof_dump_seq_mtx); prof_dump_filename(filename, 'u', prof_dump_useq); prof_dump_useq++; malloc_mutex_unlock(&prof_dump_seq_mtx); - prof_dump(filename, false, false); + prof_dump(false, filename, false); } } @@ -1066,11 +1049,13 @@ prof_bt_hash(const void *key, unsigned minbits, size_t *hash1, size_t *hash2) uint64_t h; prof_bt_t *bt = (prof_bt_t *)key; + cassert(config_prof); assert(minbits <= 32 || (SIZEOF_PTR == 8 && minbits <= 64)); assert(hash1 != NULL); assert(hash2 != NULL); - h = hash(bt->vec, bt->len * sizeof(void *), 0x94122f335b332aeaLLU); + h = hash(bt->vec, bt->len * sizeof(void *), + UINT64_C(0x94122f335b332aea)); if (minbits <= 32) { /* * Avoid doing multiple hashes, since a single hash provides @@ -1081,7 +1066,7 @@ prof_bt_hash(const void *key, unsigned minbits, size_t *hash1, size_t *hash2) } else { ret1 = h; ret2 = hash(bt->vec, bt->len * sizeof(void *), - 0x8432a476666bbc13LLU); + UINT64_C(0x8432a476666bbc13)); } *hash1 = ret1; @@ -1094,16 +1079,28 @@ prof_bt_keycomp(const void *k1, const void *k2) const prof_bt_t *bt1 = (prof_bt_t *)k1; const prof_bt_t *bt2 = (prof_bt_t *)k2; + cassert(config_prof); + if (bt1->len != bt2->len) return (false); return (memcmp(bt1->vec, bt2->vec, bt1->len * sizeof(void *)) == 0); } +static malloc_mutex_t * +prof_ctx_mutex_choose(void) +{ + unsigned nctxs = atomic_add_u(&cum_ctxs, 1); + + return (&ctx_locks[(nctxs - 1) % PROF_NCTX_LOCKS]); +} + prof_tdata_t * prof_tdata_init(void) { prof_tdata_t *prof_tdata; + cassert(config_prof); + /* Initialize an empty cache for this thread. */ prof_tdata = (prof_tdata_t *)imalloc(sizeof(prof_tdata_t)); if (prof_tdata == NULL) @@ -1116,51 +1113,77 @@ prof_tdata_init(void) } ql_new(&prof_tdata->lru_ql); - prof_tdata->vec = imalloc(sizeof(void *) * prof_bt_max); + prof_tdata->vec = imalloc(sizeof(void *) * PROF_BT_MAX); if (prof_tdata->vec == NULL) { ckh_delete(&prof_tdata->bt2cnt); idalloc(prof_tdata); return (NULL); } - prof_tdata->prn_state = 0; + prof_tdata->prng_state = 0; prof_tdata->threshold = 0; prof_tdata->accum = 0; - PROF_TCACHE_SET(prof_tdata); + prof_tdata->enq = false; + prof_tdata->enq_idump = false; + prof_tdata->enq_gdump = false; + + prof_tdata_tsd_set(&prof_tdata); return (prof_tdata); } -static void +void prof_tdata_cleanup(void *arg) { prof_thr_cnt_t *cnt; - prof_tdata_t *prof_tdata = (prof_tdata_t *)arg; + prof_tdata_t *prof_tdata = *(prof_tdata_t **)arg; - /* - * Delete the hash table. All of its contents can still be iterated - * over via the LRU. - */ - ckh_delete(&prof_tdata->bt2cnt); + cassert(config_prof); - /* Iteratively merge cnt's into the global stats and delete them. */ - while ((cnt = ql_last(&prof_tdata->lru_ql, lru_link)) != NULL) { - ql_remove(&prof_tdata->lru_ql, cnt, lru_link); - prof_ctx_merge(cnt->ctx, cnt); - idalloc(cnt); + if (prof_tdata == PROF_TDATA_STATE_REINCARNATED) { + /* + * Another destructor deallocated memory after this destructor + * was called. Reset prof_tdata to PROF_TDATA_STATE_PURGATORY + * in order to receive another callback. + */ + prof_tdata = PROF_TDATA_STATE_PURGATORY; + prof_tdata_tsd_set(&prof_tdata); + } else if (prof_tdata == PROF_TDATA_STATE_PURGATORY) { + /* + * The previous time this destructor was called, we set the key + * to PROF_TDATA_STATE_PURGATORY so that other destructors + * wouldn't cause re-creation of the prof_tdata. This time, do + * nothing, so that the destructor will not be called again. + */ + } else if (prof_tdata != NULL) { + /* + * Delete the hash table. All of its contents can still be + * iterated over via the LRU. + */ + ckh_delete(&prof_tdata->bt2cnt); + /* + * Iteratively merge cnt's into the global stats and delete + * them. + */ + while ((cnt = ql_last(&prof_tdata->lru_ql, lru_link)) != NULL) { + ql_remove(&prof_tdata->lru_ql, cnt, lru_link); + prof_ctx_merge(cnt->ctx, cnt); + idalloc(cnt); + } + idalloc(prof_tdata->vec); + idalloc(prof_tdata); + prof_tdata = PROF_TDATA_STATE_PURGATORY; + prof_tdata_tsd_set(&prof_tdata); } - - idalloc(prof_tdata->vec); - - idalloc(prof_tdata); - PROF_TCACHE_SET(NULL); } void prof_boot0(void) { + cassert(config_prof); + memcpy(opt_prof_prefix, PROF_PREFIX_DEFAULT, sizeof(PROF_PREFIX_DEFAULT)); } @@ -1169,6 +1192,8 @@ void prof_boot1(void) { + cassert(config_prof); + /* * opt_prof and prof_promote must be in their final state before any * arenas are initialized, so this function must be executed early. @@ -1190,41 +1215,46 @@ prof_boot1(void) prof_interval = 0; } - prof_promote = (opt_prof && opt_lg_prof_sample > PAGE_SHIFT); + prof_promote = (opt_prof && opt_lg_prof_sample > LG_PAGE); } bool prof_boot2(void) { + cassert(config_prof); + if (opt_prof) { + unsigned i; + if (ckh_new(&bt2ctx, PROF_CKH_MINITEMS, prof_bt_hash, prof_bt_keycomp)) return (true); if (malloc_mutex_init(&bt2ctx_mtx)) return (true); - if (pthread_key_create(&prof_tdata_tsd, prof_tdata_cleanup) - != 0) { + if (prof_tdata_tsd_boot()) { malloc_write( ": Error in pthread_key_create()\n"); abort(); } - prof_bt_max = (1U << opt_lg_prof_bt_max); if (malloc_mutex_init(&prof_dump_seq_mtx)) return (true); - if (malloc_mutex_init(&enq_mtx)) - return (true); - enq = false; - enq_idump = false; - enq_gdump = false; - if (atexit(prof_fdump) != 0) { malloc_write(": Error in atexit()\n"); if (opt_abort) abort(); } + + ctx_locks = (malloc_mutex_t *)base_alloc(PROF_NCTX_LOCKS * + sizeof(malloc_mutex_t)); + if (ctx_locks == NULL) + return (true); + for (i = 0; i < PROF_NCTX_LOCKS; i++) { + if (malloc_mutex_init(&ctx_locks[i])) + return (true); + } } #ifdef JEMALLOC_PROF_LIBGCC @@ -1241,4 +1271,3 @@ prof_boot2(void) } /******************************************************************************/ -#endif /* JEMALLOC_PROF */ diff --git a/src/quarantine.c b/src/quarantine.c new file mode 100644 index 00000000..9005ab3b --- /dev/null +++ b/src/quarantine.c @@ -0,0 +1,210 @@ +#include "jemalloc/internal/jemalloc_internal.h" + +/* + * quarantine pointers close to NULL are used to encode state information that + * is used for cleaning up during thread shutdown. + */ +#define QUARANTINE_STATE_REINCARNATED ((quarantine_t *)(uintptr_t)1) +#define QUARANTINE_STATE_PURGATORY ((quarantine_t *)(uintptr_t)2) +#define QUARANTINE_STATE_MAX QUARANTINE_STATE_PURGATORY + +/******************************************************************************/ +/* Data. */ + +typedef struct quarantine_obj_s quarantine_obj_t; +typedef struct quarantine_s quarantine_t; + +struct quarantine_obj_s { + void *ptr; + size_t usize; +}; + +struct quarantine_s { + size_t curbytes; + size_t curobjs; + size_t first; +#define LG_MAXOBJS_INIT 10 + size_t lg_maxobjs; + quarantine_obj_t objs[1]; /* Dynamically sized ring buffer. */ +}; + +static void quarantine_cleanup(void *arg); + +malloc_tsd_data(static, quarantine, quarantine_t *, NULL) +malloc_tsd_funcs(JEMALLOC_INLINE, quarantine, quarantine_t *, NULL, + quarantine_cleanup) + +/******************************************************************************/ +/* Function prototypes for non-inline static functions. */ + +static quarantine_t *quarantine_init(size_t lg_maxobjs); +static quarantine_t *quarantine_grow(quarantine_t *quarantine); +static void quarantine_drain(quarantine_t *quarantine, size_t upper_bound); + +/******************************************************************************/ + +static quarantine_t * +quarantine_init(size_t lg_maxobjs) +{ + quarantine_t *quarantine; + + quarantine = (quarantine_t *)imalloc(offsetof(quarantine_t, objs) + + ((ZU(1) << lg_maxobjs) * sizeof(quarantine_obj_t))); + if (quarantine == NULL) + return (NULL); + quarantine->curbytes = 0; + quarantine->curobjs = 0; + quarantine->first = 0; + quarantine->lg_maxobjs = lg_maxobjs; + + quarantine_tsd_set(&quarantine); + + return (quarantine); +} + +static quarantine_t * +quarantine_grow(quarantine_t *quarantine) +{ + quarantine_t *ret; + + ret = quarantine_init(quarantine->lg_maxobjs + 1); + if (ret == NULL) + return (quarantine); + + ret->curbytes = quarantine->curbytes; + ret->curobjs = quarantine->curobjs; + if (quarantine->first + quarantine->curobjs <= (ZU(1) << + quarantine->lg_maxobjs)) { + /* objs ring buffer data are contiguous. */ + memcpy(ret->objs, &quarantine->objs[quarantine->first], + quarantine->curobjs * sizeof(quarantine_obj_t)); + } else { + /* objs ring buffer data wrap around. */ + size_t ncopy_a = (ZU(1) << quarantine->lg_maxobjs) - + quarantine->first; + size_t ncopy_b = quarantine->curobjs - ncopy_a; + + memcpy(ret->objs, &quarantine->objs[quarantine->first], ncopy_a + * sizeof(quarantine_obj_t)); + memcpy(&ret->objs[ncopy_a], quarantine->objs, ncopy_b * + sizeof(quarantine_obj_t)); + } + + return (ret); +} + +static void +quarantine_drain(quarantine_t *quarantine, size_t upper_bound) +{ + + while (quarantine->curbytes > upper_bound && quarantine->curobjs > 0) { + quarantine_obj_t *obj = &quarantine->objs[quarantine->first]; + assert(obj->usize == isalloc(obj->ptr, config_prof)); + idalloc(obj->ptr); + quarantine->curbytes -= obj->usize; + quarantine->curobjs--; + quarantine->first = (quarantine->first + 1) & ((ZU(1) << + quarantine->lg_maxobjs) - 1); + } +} + +void +quarantine(void *ptr) +{ + quarantine_t *quarantine; + size_t usize = isalloc(ptr, config_prof); + + cassert(config_fill); + assert(opt_quarantine); + + quarantine = *quarantine_tsd_get(); + if ((uintptr_t)quarantine <= (uintptr_t)QUARANTINE_STATE_MAX) { + if (quarantine == NULL) { + if ((quarantine = quarantine_init(LG_MAXOBJS_INIT)) == + NULL) { + idalloc(ptr); + return; + } + } else { + if (quarantine == QUARANTINE_STATE_PURGATORY) { + /* + * Make a note that quarantine() was called + * after quarantine_cleanup() was called. + */ + quarantine = QUARANTINE_STATE_REINCARNATED; + quarantine_tsd_set(&quarantine); + } + idalloc(ptr); + return; + } + } + /* + * Drain one or more objects if the quarantine size limit would be + * exceeded by appending ptr. + */ + if (quarantine->curbytes + usize > opt_quarantine) { + size_t upper_bound = (opt_quarantine >= usize) ? opt_quarantine + - usize : 0; + quarantine_drain(quarantine, upper_bound); + } + /* Grow the quarantine ring buffer if it's full. */ + if (quarantine->curobjs == (ZU(1) << quarantine->lg_maxobjs)) + quarantine = quarantine_grow(quarantine); + /* quarantine_grow() must free a slot if it fails to grow. */ + assert(quarantine->curobjs < (ZU(1) << quarantine->lg_maxobjs)); + /* Append ptr if its size doesn't exceed the quarantine size. */ + if (quarantine->curbytes + usize <= opt_quarantine) { + size_t offset = (quarantine->first + quarantine->curobjs) & + ((ZU(1) << quarantine->lg_maxobjs) - 1); + quarantine_obj_t *obj = &quarantine->objs[offset]; + obj->ptr = ptr; + obj->usize = usize; + quarantine->curbytes += usize; + quarantine->curobjs++; + if (opt_junk) + memset(ptr, 0x5a, usize); + } else { + assert(quarantine->curbytes == 0); + idalloc(ptr); + } +} + +static void +quarantine_cleanup(void *arg) +{ + quarantine_t *quarantine = *(quarantine_t **)arg; + + if (quarantine == QUARANTINE_STATE_REINCARNATED) { + /* + * Another destructor deallocated memory after this destructor + * was called. Reset quarantine to QUARANTINE_STATE_PURGATORY + * in order to receive another callback. + */ + quarantine = QUARANTINE_STATE_PURGATORY; + quarantine_tsd_set(&quarantine); + } else if (quarantine == QUARANTINE_STATE_PURGATORY) { + /* + * The previous time this destructor was called, we set the key + * to QUARANTINE_STATE_PURGATORY so that other destructors + * wouldn't cause re-creation of the quarantine. This time, do + * nothing, so that the destructor will not be called again. + */ + } else if (quarantine != NULL) { + quarantine_drain(quarantine, 0); + idalloc(quarantine); + quarantine = QUARANTINE_STATE_PURGATORY; + quarantine_tsd_set(&quarantine); + } +} + +bool +quarantine_boot(void) +{ + + cassert(config_fill); + + if (quarantine_tsd_boot()) + return (true); + + return (false); +} diff --git a/src/stats.c b/src/stats.c index dc172e42..433b80d1 100644 --- a/src/stats.c +++ b/src/stats.c @@ -39,140 +39,40 @@ bool opt_stats_print = false; -#ifdef JEMALLOC_STATS size_t stats_cactive = 0; -#endif /******************************************************************************/ /* Function prototypes for non-inline static functions. */ -#ifdef JEMALLOC_STATS -static void malloc_vcprintf(void (*write_cb)(void *, const char *), - void *cbopaque, const char *format, va_list ap); static void stats_arena_bins_print(void (*write_cb)(void *, const char *), void *cbopaque, unsigned i); static void stats_arena_lruns_print(void (*write_cb)(void *, const char *), void *cbopaque, unsigned i); static void stats_arena_print(void (*write_cb)(void *, const char *), - void *cbopaque, unsigned i); -#endif + void *cbopaque, unsigned i, bool bins, bool large); /******************************************************************************/ -/* - * We don't want to depend on vsnprintf() for production builds, since that can - * cause unnecessary bloat for static binaries. u2s() provides minimal integer - * printing functionality, so that malloc_printf() use can be limited to - * JEMALLOC_STATS code. - */ -char * -u2s(uint64_t x, unsigned base, char *s) -{ - unsigned i; - - i = UMAX2S_BUFSIZE - 1; - s[i] = '\0'; - switch (base) { - case 10: - do { - i--; - s[i] = "0123456789"[x % (uint64_t)10]; - x /= (uint64_t)10; - } while (x > 0); - break; - case 16: - do { - i--; - s[i] = "0123456789abcdef"[x & 0xf]; - x >>= 4; - } while (x > 0); - break; - default: - do { - i--; - s[i] = "0123456789abcdefghijklmnopqrstuvwxyz"[x % - (uint64_t)base]; - x /= (uint64_t)base; - } while (x > 0); - } - - return (&s[i]); -} - -#ifdef JEMALLOC_STATS -static void -malloc_vcprintf(void (*write_cb)(void *, const char *), void *cbopaque, - const char *format, va_list ap) -{ - char buf[4096]; - - if (write_cb == NULL) { - /* - * The caller did not provide an alternate write_cb callback - * function, so use the default one. malloc_write() is an - * inline function, so use malloc_message() directly here. - */ - write_cb = JEMALLOC_P(malloc_message); - cbopaque = NULL; - } - - vsnprintf(buf, sizeof(buf), format, ap); - write_cb(cbopaque, buf); -} - -/* - * Print to a callback function in such a way as to (hopefully) avoid memory - * allocation. - */ -JEMALLOC_ATTR(format(printf, 3, 4)) -void -malloc_cprintf(void (*write_cb)(void *, const char *), void *cbopaque, - const char *format, ...) -{ - va_list ap; - - va_start(ap, format); - malloc_vcprintf(write_cb, cbopaque, format, ap); - va_end(ap); -} - -/* - * Print to stderr in such a way as to (hopefully) avoid memory allocation. - */ -JEMALLOC_ATTR(format(printf, 1, 2)) -void -malloc_printf(const char *format, ...) -{ - va_list ap; - - va_start(ap, format); - malloc_vcprintf(NULL, NULL, format, ap); - va_end(ap); -} -#endif - -#ifdef JEMALLOC_STATS static void stats_arena_bins_print(void (*write_cb)(void *, const char *), void *cbopaque, unsigned i) { - size_t pagesize; + size_t page; bool config_tcache; unsigned nbins, j, gap_start; - CTL_GET("arenas.pagesize", &pagesize, size_t); + CTL_GET("arenas.page", &page, size_t); CTL_GET("config.tcache", &config_tcache, bool); if (config_tcache) { malloc_cprintf(write_cb, cbopaque, - "bins: bin size regs pgs allocated nmalloc" + "bins: bin size regs pgs allocated nmalloc" " ndalloc nrequests nfills nflushes" - " newruns reruns maxruns curruns\n"); + " newruns reruns curruns\n"); } else { malloc_cprintf(write_cb, cbopaque, - "bins: bin size regs pgs allocated nmalloc" - " ndalloc newruns reruns maxruns" - " curruns\n"); + "bins: bin size regs pgs allocated nmalloc" + " ndalloc newruns reruns curruns\n"); } CTL_GET("arenas.nbins", &nbins, unsigned); for (j = 0, gap_start = UINT_MAX; j < nbins; j++) { @@ -183,12 +83,11 @@ stats_arena_bins_print(void (*write_cb)(void *, const char *), void *cbopaque, if (gap_start == UINT_MAX) gap_start = j; } else { - unsigned ntbins_, nqbins, ncbins, nsbins; size_t reg_size, run_size, allocated; uint32_t nregs; uint64_t nmalloc, ndalloc, nrequests, nfills, nflushes; uint64_t reruns; - size_t highruns, curruns; + size_t curruns; if (gap_start != UINT_MAX) { if (j > gap_start + 1) { @@ -203,10 +102,6 @@ stats_arena_bins_print(void (*write_cb)(void *, const char *), void *cbopaque, } gap_start = UINT_MAX; } - CTL_GET("arenas.ntbins", &ntbins_, unsigned); - CTL_GET("arenas.nqbins", &nqbins, unsigned); - CTL_GET("arenas.ncbins", &ncbins, unsigned); - CTL_GET("arenas.nsbins", &nsbins, unsigned); CTL_J_GET("arenas.bin.0.size", ®_size, size_t); CTL_J_GET("arenas.bin.0.nregs", &nregs, uint32_t); CTL_J_GET("arenas.bin.0.run_size", &run_size, size_t); @@ -226,36 +121,25 @@ stats_arena_bins_print(void (*write_cb)(void *, const char *), void *cbopaque, } CTL_IJ_GET("stats.arenas.0.bins.0.nreruns", &reruns, uint64_t); - CTL_IJ_GET("stats.arenas.0.bins.0.highruns", &highruns, - size_t); CTL_IJ_GET("stats.arenas.0.bins.0.curruns", &curruns, size_t); if (config_tcache) { malloc_cprintf(write_cb, cbopaque, - "%13u %1s %5zu %4u %3zu %12zu %12"PRIu64 + "%13u %5zu %4u %3zu %12zu %12"PRIu64 " %12"PRIu64" %12"PRIu64" %12"PRIu64 " %12"PRIu64" %12"PRIu64" %12"PRIu64 - " %12zu %12zu\n", - j, - j < ntbins_ ? "T" : j < ntbins_ + nqbins ? - "Q" : j < ntbins_ + nqbins + ncbins ? "C" : - "S", - reg_size, nregs, run_size / pagesize, + " %12zu\n", + j, reg_size, nregs, run_size / page, allocated, nmalloc, ndalloc, nrequests, - nfills, nflushes, nruns, reruns, highruns, - curruns); + nfills, nflushes, nruns, reruns, curruns); } else { malloc_cprintf(write_cb, cbopaque, - "%13u %1s %5zu %4u %3zu %12zu %12"PRIu64 + "%13u %5zu %4u %3zu %12zu %12"PRIu64 " %12"PRIu64" %12"PRIu64" %12"PRIu64 - " %12zu %12zu\n", - j, - j < ntbins_ ? "T" : j < ntbins_ + nqbins ? - "Q" : j < ntbins_ + nqbins + ncbins ? "C" : - "S", - reg_size, nregs, run_size / pagesize, + " %12zu\n", + j, reg_size, nregs, run_size / page, allocated, nmalloc, ndalloc, nruns, reruns, - highruns, curruns); + curruns); } } } @@ -275,18 +159,18 @@ static void stats_arena_lruns_print(void (*write_cb)(void *, const char *), void *cbopaque, unsigned i) { - size_t pagesize, nlruns, j; + size_t page, nlruns, j; ssize_t gap_start; - CTL_GET("arenas.pagesize", &pagesize, size_t); + CTL_GET("arenas.page", &page, size_t); malloc_cprintf(write_cb, cbopaque, "large: size pages nmalloc ndalloc nrequests" - " maxruns curruns\n"); + " curruns\n"); CTL_GET("arenas.nlruns", &nlruns, size_t); for (j = 0, gap_start = -1; j < nlruns; j++) { uint64_t nmalloc, ndalloc, nrequests; - size_t run_size, highruns, curruns; + size_t run_size, curruns; CTL_IJ_GET("stats.arenas.0.lruns.0.nmalloc", &nmalloc, uint64_t); @@ -299,8 +183,6 @@ stats_arena_lruns_print(void (*write_cb)(void *, const char *), void *cbopaque, gap_start = j; } else { CTL_J_GET("arenas.lrun.0.size", &run_size, size_t); - CTL_IJ_GET("stats.arenas.0.lruns.0.highruns", &highruns, - size_t); CTL_IJ_GET("stats.arenas.0.lruns.0.curruns", &curruns, size_t); if (gap_start != -1) { @@ -310,9 +192,9 @@ stats_arena_lruns_print(void (*write_cb)(void *, const char *), void *cbopaque, } malloc_cprintf(write_cb, cbopaque, "%13zu %5zu %12"PRIu64" %12"PRIu64" %12"PRIu64 - " %12zu %12zu\n", - run_size, run_size / pagesize, nmalloc, ndalloc, - nrequests, highruns, curruns); + " %12zu\n", + run_size, run_size / page, nmalloc, ndalloc, + nrequests, curruns); } } if (gap_start != -1) @@ -321,17 +203,17 @@ stats_arena_lruns_print(void (*write_cb)(void *, const char *), void *cbopaque, static void stats_arena_print(void (*write_cb)(void *, const char *), void *cbopaque, - unsigned i) + unsigned i, bool bins, bool large) { unsigned nthreads; - size_t pagesize, pactive, pdirty, mapped; + size_t page, pactive, pdirty, mapped; uint64_t npurge, nmadvise, purged; size_t small_allocated; uint64_t small_nmalloc, small_ndalloc, small_nrequests; size_t large_allocated; uint64_t large_nmalloc, large_ndalloc, large_nrequests; - CTL_GET("arenas.pagesize", &pagesize, size_t); + CTL_GET("arenas.page", &page, size_t); CTL_I_GET("stats.arenas.0.nthreads", &nthreads, unsigned); malloc_cprintf(write_cb, cbopaque, @@ -369,15 +251,15 @@ stats_arena_print(void (*write_cb)(void *, const char *), void *cbopaque, small_nmalloc + large_nmalloc, small_ndalloc + large_ndalloc, small_nrequests + large_nrequests); - malloc_cprintf(write_cb, cbopaque, "active: %12zu\n", - pactive * pagesize ); + malloc_cprintf(write_cb, cbopaque, "active: %12zu\n", pactive * page); CTL_I_GET("stats.arenas.0.mapped", &mapped, size_t); malloc_cprintf(write_cb, cbopaque, "mapped: %12zu\n", mapped); - stats_arena_bins_print(write_cb, cbopaque, i); - stats_arena_lruns_print(write_cb, cbopaque, i); + if (bins) + stats_arena_bins_print(write_cb, cbopaque, i); + if (large) + stats_arena_lruns_print(write_cb, cbopaque, i); } -#endif void stats_print(void (*write_cb)(void *, const char *), void *cbopaque, @@ -386,7 +268,6 @@ stats_print(void (*write_cb)(void *, const char *), void *cbopaque, int err; uint64_t epoch; size_t u64sz; - char s[UMAX2S_BUFSIZE]; bool general = true; bool merged = true; bool unmerged = true; @@ -402,8 +283,7 @@ stats_print(void (*write_cb)(void *, const char *), void *cbopaque, * */ epoch = 1; u64sz = sizeof(uint64_t); - err = JEMALLOC_P(mallctl)("epoch", &epoch, &u64sz, &epoch, - sizeof(uint64_t)); + err = je_mallctl("epoch", &epoch, &u64sz, &epoch, sizeof(uint64_t)); if (err != 0) { if (err == EAGAIN) { malloc_write(": Memory allocation failure in " @@ -415,42 +295,33 @@ stats_print(void (*write_cb)(void *, const char *), void *cbopaque, abort(); } - if (write_cb == NULL) { - /* - * The caller did not provide an alternate write_cb callback - * function, so use the default one. malloc_write() is an - * inline function, so use malloc_message() directly here. - */ - write_cb = JEMALLOC_P(malloc_message); - cbopaque = NULL; - } - if (opts != NULL) { unsigned i; for (i = 0; opts[i] != '\0'; i++) { switch (opts[i]) { - case 'g': - general = false; - break; - case 'm': - merged = false; - break; - case 'a': - unmerged = false; - break; - case 'b': - bins = false; - break; - case 'l': - large = false; - break; - default:; + case 'g': + general = false; + break; + case 'm': + merged = false; + break; + case 'a': + unmerged = false; + break; + case 'b': + bins = false; + break; + case 'l': + large = false; + break; + default:; } } } - write_cb(cbopaque, "___ Begin jemalloc statistics ___\n"); + malloc_cprintf(write_cb, cbopaque, + "___ Begin jemalloc statistics ___\n"); if (general) { int err; const char *cpv; @@ -465,229 +336,126 @@ stats_print(void (*write_cb)(void *, const char *), void *cbopaque, cpsz = sizeof(const char *); CTL_GET("version", &cpv, const char *); - write_cb(cbopaque, "Version: "); - write_cb(cbopaque, cpv); - write_cb(cbopaque, "\n"); + malloc_cprintf(write_cb, cbopaque, "Version: %s\n", cpv); CTL_GET("config.debug", &bv, bool); - write_cb(cbopaque, "Assertions "); - write_cb(cbopaque, bv ? "enabled" : "disabled"); - write_cb(cbopaque, "\n"); + malloc_cprintf(write_cb, cbopaque, "Assertions %s\n", + bv ? "enabled" : "disabled"); #define OPT_WRITE_BOOL(n) \ - if ((err = JEMALLOC_P(mallctl)("opt."#n, &bv, &bsz, \ - NULL, 0)) == 0) { \ - write_cb(cbopaque, " opt."#n": "); \ - write_cb(cbopaque, bv ? "true" : "false"); \ - write_cb(cbopaque, "\n"); \ + if ((err = je_mallctl("opt."#n, &bv, &bsz, NULL, 0)) \ + == 0) { \ + malloc_cprintf(write_cb, cbopaque, \ + " opt."#n": %s\n", bv ? "true" : "false"); \ } #define OPT_WRITE_SIZE_T(n) \ - if ((err = JEMALLOC_P(mallctl)("opt."#n, &sv, &ssz, \ - NULL, 0)) == 0) { \ - write_cb(cbopaque, " opt."#n": "); \ - write_cb(cbopaque, u2s(sv, 10, s)); \ - write_cb(cbopaque, "\n"); \ + if ((err = je_mallctl("opt."#n, &sv, &ssz, NULL, 0)) \ + == 0) { \ + malloc_cprintf(write_cb, cbopaque, \ + " opt."#n": %zu\n", sv); \ } #define OPT_WRITE_SSIZE_T(n) \ - if ((err = JEMALLOC_P(mallctl)("opt."#n, &ssv, &sssz, \ - NULL, 0)) == 0) { \ - if (ssv >= 0) { \ - write_cb(cbopaque, " opt."#n": "); \ - write_cb(cbopaque, u2s(ssv, 10, s)); \ - } else { \ - write_cb(cbopaque, " opt."#n": -"); \ - write_cb(cbopaque, u2s(-ssv, 10, s)); \ - } \ - write_cb(cbopaque, "\n"); \ + if ((err = je_mallctl("opt."#n, &ssv, &sssz, NULL, 0)) \ + == 0) { \ + malloc_cprintf(write_cb, cbopaque, \ + " opt."#n": %zd\n", ssv); \ } #define OPT_WRITE_CHAR_P(n) \ - if ((err = JEMALLOC_P(mallctl)("opt."#n, &cpv, &cpsz, \ - NULL, 0)) == 0) { \ - write_cb(cbopaque, " opt."#n": \""); \ - write_cb(cbopaque, cpv); \ - write_cb(cbopaque, "\"\n"); \ + if ((err = je_mallctl("opt."#n, &cpv, &cpsz, NULL, 0)) \ + == 0) { \ + malloc_cprintf(write_cb, cbopaque, \ + " opt."#n": \"%s\"\n", cpv); \ } - write_cb(cbopaque, "Run-time option settings:\n"); + malloc_cprintf(write_cb, cbopaque, + "Run-time option settings:\n"); OPT_WRITE_BOOL(abort) - OPT_WRITE_SIZE_T(lg_qspace_max) - OPT_WRITE_SIZE_T(lg_cspace_max) OPT_WRITE_SIZE_T(lg_chunk) OPT_WRITE_SIZE_T(narenas) OPT_WRITE_SSIZE_T(lg_dirty_mult) OPT_WRITE_BOOL(stats_print) OPT_WRITE_BOOL(junk) + OPT_WRITE_SIZE_T(quarantine) + OPT_WRITE_BOOL(redzone) OPT_WRITE_BOOL(zero) - OPT_WRITE_BOOL(sysv) + OPT_WRITE_BOOL(utrace) + OPT_WRITE_BOOL(valgrind) OPT_WRITE_BOOL(xmalloc) OPT_WRITE_BOOL(tcache) - OPT_WRITE_SSIZE_T(lg_tcache_gc_sweep) OPT_WRITE_SSIZE_T(lg_tcache_max) OPT_WRITE_BOOL(prof) OPT_WRITE_CHAR_P(prof_prefix) - OPT_WRITE_SIZE_T(lg_prof_bt_max) OPT_WRITE_BOOL(prof_active) OPT_WRITE_SSIZE_T(lg_prof_sample) OPT_WRITE_BOOL(prof_accum) - OPT_WRITE_SSIZE_T(lg_prof_tcmax) OPT_WRITE_SSIZE_T(lg_prof_interval) OPT_WRITE_BOOL(prof_gdump) + OPT_WRITE_BOOL(prof_final) OPT_WRITE_BOOL(prof_leak) - OPT_WRITE_BOOL(overcommit) #undef OPT_WRITE_BOOL #undef OPT_WRITE_SIZE_T #undef OPT_WRITE_SSIZE_T #undef OPT_WRITE_CHAR_P - write_cb(cbopaque, "CPUs: "); - write_cb(cbopaque, u2s(ncpus, 10, s)); - write_cb(cbopaque, "\n"); + malloc_cprintf(write_cb, cbopaque, "CPUs: %u\n", ncpus); CTL_GET("arenas.narenas", &uv, unsigned); - write_cb(cbopaque, "Max arenas: "); - write_cb(cbopaque, u2s(uv, 10, s)); - write_cb(cbopaque, "\n"); + malloc_cprintf(write_cb, cbopaque, "Max arenas: %u\n", uv); - write_cb(cbopaque, "Pointer size: "); - write_cb(cbopaque, u2s(sizeof(void *), 10, s)); - write_cb(cbopaque, "\n"); + malloc_cprintf(write_cb, cbopaque, "Pointer size: %zu\n", + sizeof(void *)); CTL_GET("arenas.quantum", &sv, size_t); - write_cb(cbopaque, "Quantum size: "); - write_cb(cbopaque, u2s(sv, 10, s)); - write_cb(cbopaque, "\n"); + malloc_cprintf(write_cb, cbopaque, "Quantum size: %zu\n", sv); - CTL_GET("arenas.cacheline", &sv, size_t); - write_cb(cbopaque, "Cacheline size (assumed): "); - write_cb(cbopaque, u2s(sv, 10, s)); - write_cb(cbopaque, "\n"); - - CTL_GET("arenas.subpage", &sv, size_t); - write_cb(cbopaque, "Subpage spacing: "); - write_cb(cbopaque, u2s(sv, 10, s)); - write_cb(cbopaque, "\n"); - - if ((err = JEMALLOC_P(mallctl)("arenas.tspace_min", &sv, &ssz, - NULL, 0)) == 0) { - write_cb(cbopaque, "Tiny 2^n-spaced sizes: ["); - write_cb(cbopaque, u2s(sv, 10, s)); - write_cb(cbopaque, ".."); - - CTL_GET("arenas.tspace_max", &sv, size_t); - write_cb(cbopaque, u2s(sv, 10, s)); - write_cb(cbopaque, "]\n"); - } - - CTL_GET("arenas.qspace_min", &sv, size_t); - write_cb(cbopaque, "Quantum-spaced sizes: ["); - write_cb(cbopaque, u2s(sv, 10, s)); - write_cb(cbopaque, ".."); - CTL_GET("arenas.qspace_max", &sv, size_t); - write_cb(cbopaque, u2s(sv, 10, s)); - write_cb(cbopaque, "]\n"); - - CTL_GET("arenas.cspace_min", &sv, size_t); - write_cb(cbopaque, "Cacheline-spaced sizes: ["); - write_cb(cbopaque, u2s(sv, 10, s)); - write_cb(cbopaque, ".."); - CTL_GET("arenas.cspace_max", &sv, size_t); - write_cb(cbopaque, u2s(sv, 10, s)); - write_cb(cbopaque, "]\n"); - - CTL_GET("arenas.sspace_min", &sv, size_t); - write_cb(cbopaque, "Subpage-spaced sizes: ["); - write_cb(cbopaque, u2s(sv, 10, s)); - write_cb(cbopaque, ".."); - CTL_GET("arenas.sspace_max", &sv, size_t); - write_cb(cbopaque, u2s(sv, 10, s)); - write_cb(cbopaque, "]\n"); + CTL_GET("arenas.page", &sv, size_t); + malloc_cprintf(write_cb, cbopaque, "Page size: %zu\n", sv); CTL_GET("opt.lg_dirty_mult", &ssv, ssize_t); if (ssv >= 0) { - write_cb(cbopaque, - "Min active:dirty page ratio per arena: "); - write_cb(cbopaque, u2s((1U << ssv), 10, s)); - write_cb(cbopaque, ":1\n"); + malloc_cprintf(write_cb, cbopaque, + "Min active:dirty page ratio per arena: %u:1\n", + (1U << ssv)); } else { - write_cb(cbopaque, + malloc_cprintf(write_cb, cbopaque, "Min active:dirty page ratio per arena: N/A\n"); } - if ((err = JEMALLOC_P(mallctl)("arenas.tcache_max", &sv, - &ssz, NULL, 0)) == 0) { - write_cb(cbopaque, - "Maximum thread-cached size class: "); - write_cb(cbopaque, u2s(sv, 10, s)); - write_cb(cbopaque, "\n"); + if ((err = je_mallctl("arenas.tcache_max", &sv, &ssz, NULL, 0)) + == 0) { + malloc_cprintf(write_cb, cbopaque, + "Maximum thread-cached size class: %zu\n", sv); } - if ((err = JEMALLOC_P(mallctl)("opt.lg_tcache_gc_sweep", &ssv, - &ssz, NULL, 0)) == 0) { - size_t tcache_gc_sweep = (1U << ssv); - bool tcache_enabled; - CTL_GET("opt.tcache", &tcache_enabled, bool); - write_cb(cbopaque, "Thread cache GC sweep interval: "); - write_cb(cbopaque, tcache_enabled && ssv >= 0 ? - u2s(tcache_gc_sweep, 10, s) : "N/A"); - write_cb(cbopaque, "\n"); - } - if ((err = JEMALLOC_P(mallctl)("opt.prof", &bv, &bsz, NULL, 0)) - == 0 && bv) { - CTL_GET("opt.lg_prof_bt_max", &sv, size_t); - write_cb(cbopaque, "Maximum profile backtrace depth: "); - write_cb(cbopaque, u2s((1U << sv), 10, s)); - write_cb(cbopaque, "\n"); - - CTL_GET("opt.lg_prof_tcmax", &ssv, ssize_t); - write_cb(cbopaque, - "Maximum per thread backtrace cache: "); - if (ssv >= 0) { - write_cb(cbopaque, u2s((1U << ssv), 10, s)); - write_cb(cbopaque, " (2^"); - write_cb(cbopaque, u2s(ssv, 10, s)); - write_cb(cbopaque, ")\n"); - } else - write_cb(cbopaque, "N/A\n"); - + if ((err = je_mallctl("opt.prof", &bv, &bsz, NULL, 0)) == 0 && + bv) { CTL_GET("opt.lg_prof_sample", &sv, size_t); - write_cb(cbopaque, "Average profile sample interval: "); - write_cb(cbopaque, u2s((((uint64_t)1U) << sv), 10, s)); - write_cb(cbopaque, " (2^"); - write_cb(cbopaque, u2s(sv, 10, s)); - write_cb(cbopaque, ")\n"); + malloc_cprintf(write_cb, cbopaque, + "Average profile sample interval: %"PRIu64 + " (2^%zu)\n", (((uint64_t)1U) << sv), sv); CTL_GET("opt.lg_prof_interval", &ssv, ssize_t); - write_cb(cbopaque, "Average profile dump interval: "); if (ssv >= 0) { - write_cb(cbopaque, u2s((((uint64_t)1U) << ssv), - 10, s)); - write_cb(cbopaque, " (2^"); - write_cb(cbopaque, u2s(ssv, 10, s)); - write_cb(cbopaque, ")\n"); - } else - write_cb(cbopaque, "N/A\n"); + malloc_cprintf(write_cb, cbopaque, + "Average profile dump interval: %"PRIu64 + " (2^%zd)\n", + (((uint64_t)1U) << ssv), ssv); + } else { + malloc_cprintf(write_cb, cbopaque, + "Average profile dump interval: N/A\n"); + } } - CTL_GET("arenas.chunksize", &sv, size_t); - write_cb(cbopaque, "Chunk size: "); - write_cb(cbopaque, u2s(sv, 10, s)); CTL_GET("opt.lg_chunk", &sv, size_t); - write_cb(cbopaque, " (2^"); - write_cb(cbopaque, u2s(sv, 10, s)); - write_cb(cbopaque, ")\n"); + malloc_cprintf(write_cb, cbopaque, "Chunk size: %zu (2^%zu)\n", + (ZU(1) << sv), sv); } -#ifdef JEMALLOC_STATS - { - int err; - size_t sszp, ssz; + if (config_stats) { size_t *cactive; size_t allocated, active, mapped; - size_t chunks_current, chunks_high, swap_avail; + size_t chunks_current, chunks_high; uint64_t chunks_total; size_t huge_allocated; uint64_t huge_nmalloc, huge_ndalloc; - sszp = sizeof(size_t *); - ssz = sizeof(size_t); - CTL_GET("stats.cactive", &cactive, size_t *); CTL_GET("stats.allocated", &allocated, size_t); CTL_GET("stats.active", &active, size_t); @@ -702,24 +470,10 @@ stats_print(void (*write_cb)(void *, const char *), void *cbopaque, CTL_GET("stats.chunks.total", &chunks_total, uint64_t); CTL_GET("stats.chunks.high", &chunks_high, size_t); CTL_GET("stats.chunks.current", &chunks_current, size_t); - if ((err = JEMALLOC_P(mallctl)("swap.avail", &swap_avail, &ssz, - NULL, 0)) == 0) { - size_t lg_chunk; - - malloc_cprintf(write_cb, cbopaque, "chunks: nchunks " - "highchunks curchunks swap_avail\n"); - CTL_GET("opt.lg_chunk", &lg_chunk, size_t); - malloc_cprintf(write_cb, cbopaque, - " %13"PRIu64"%13zu%13zu%13zu\n", - chunks_total, chunks_high, chunks_current, - swap_avail << lg_chunk); - } else { - malloc_cprintf(write_cb, cbopaque, "chunks: nchunks " - "highchunks curchunks\n"); - malloc_cprintf(write_cb, cbopaque, - " %13"PRIu64"%13zu%13zu\n", - chunks_total, chunks_high, chunks_current); - } + malloc_cprintf(write_cb, cbopaque, "chunks: nchunks " + "highchunks curchunks\n"); + malloc_cprintf(write_cb, cbopaque, " %13"PRIu64"%13zu%13zu\n", + chunks_total, chunks_high, chunks_current); /* Print huge stats. */ CTL_GET("stats.huge.nmalloc", &huge_nmalloc, uint64_t); @@ -736,11 +490,11 @@ stats_print(void (*write_cb)(void *, const char *), void *cbopaque, CTL_GET("arenas.narenas", &narenas, unsigned); { - bool initialized[narenas]; + VARIABLE_ARRAY(bool, initialized, narenas); size_t isz; unsigned i, ninitialized; - isz = sizeof(initialized); + isz = sizeof(bool) * narenas; xmallctl("arenas.initialized", initialized, &isz, NULL, 0); for (i = ninitialized = 0; i < narenas; i++) { @@ -753,7 +507,7 @@ stats_print(void (*write_cb)(void *, const char *), void *cbopaque, malloc_cprintf(write_cb, cbopaque, "\nMerged arenas stats:\n"); stats_arena_print(write_cb, cbopaque, - narenas); + narenas, bins, large); } } } @@ -765,11 +519,11 @@ stats_print(void (*write_cb)(void *, const char *), void *cbopaque, CTL_GET("arenas.narenas", &narenas, unsigned); { - bool initialized[narenas]; + VARIABLE_ARRAY(bool, initialized, narenas); size_t isz; unsigned i; - isz = sizeof(initialized); + isz = sizeof(bool) * narenas; xmallctl("arenas.initialized", initialized, &isz, NULL, 0); @@ -779,12 +533,11 @@ stats_print(void (*write_cb)(void *, const char *), void *cbopaque, cbopaque, "\narenas[%u]:\n", i); stats_arena_print(write_cb, - cbopaque, i); + cbopaque, i, bins, large); } } } } } -#endif /* #ifdef JEMALLOC_STATS */ - write_cb(cbopaque, "--- End jemalloc statistics ---\n"); + malloc_cprintf(write_cb, cbopaque, "--- End jemalloc statistics ---\n"); } diff --git a/src/tcache.c b/src/tcache.c index 31c329e1..60244c45 100644 --- a/src/tcache.c +++ b/src/tcache.c @@ -1,70 +1,92 @@ #define JEMALLOC_TCACHE_C_ #include "jemalloc/internal/jemalloc_internal.h" -#ifdef JEMALLOC_TCACHE + /******************************************************************************/ /* Data. */ +malloc_tsd_data(, tcache, tcache_t *, NULL) +malloc_tsd_data(, tcache_enabled, tcache_enabled_t, tcache_enabled_default) + bool opt_tcache = true; ssize_t opt_lg_tcache_max = LG_TCACHE_MAXCLASS_DEFAULT; -ssize_t opt_lg_tcache_gc_sweep = LG_TCACHE_GC_SWEEP_DEFAULT; tcache_bin_info_t *tcache_bin_info; static unsigned stack_nelms; /* Total stack elms per tcache. */ -/* Map of thread-specific caches. */ -#ifndef NO_TLS -__thread tcache_t *tcache_tls JEMALLOC_ATTR(tls_model("initial-exec")); -#endif - -/* - * Same contents as tcache, but initialized such that the TSD destructor is - * called when a thread exits, so that the cache can be cleaned up. - */ -pthread_key_t tcache_tsd; - -size_t nhbins; -size_t tcache_maxclass; -unsigned tcache_gc_incr; +size_t nhbins; +size_t tcache_maxclass; /******************************************************************************/ -/* Function prototypes for non-inline static functions. */ -static void tcache_thread_cleanup(void *arg); +size_t tcache_salloc(const void *ptr) +{ -/******************************************************************************/ + return (arena_salloc(ptr, false)); +} + +void +tcache_event_hard(tcache_t *tcache) +{ + size_t binind = tcache->next_gc_bin; + tcache_bin_t *tbin = &tcache->tbins[binind]; + tcache_bin_info_t *tbin_info = &tcache_bin_info[binind]; + + if (tbin->low_water > 0) { + /* + * Flush (ceiling) 3/4 of the objects below the low water mark. + */ + if (binind < NBINS) { + tcache_bin_flush_small(tbin, binind, tbin->ncached - + tbin->low_water + (tbin->low_water >> 2), tcache); + } else { + tcache_bin_flush_large(tbin, binind, tbin->ncached - + tbin->low_water + (tbin->low_water >> 2), tcache); + } + /* + * Reduce fill count by 2X. Limit lg_fill_div such that the + * fill count is always at least 1. + */ + if ((tbin_info->ncached_max >> (tbin->lg_fill_div+1)) >= 1) + tbin->lg_fill_div++; + } else if (tbin->low_water < 0) { + /* + * Increase fill count by 2X. Make sure lg_fill_div stays + * greater than 0. + */ + if (tbin->lg_fill_div > 1) + tbin->lg_fill_div--; + } + tbin->low_water = tbin->ncached; + + tcache->next_gc_bin++; + if (tcache->next_gc_bin == nhbins) + tcache->next_gc_bin = 0; + tcache->ev_cnt = 0; +} void * tcache_alloc_small_hard(tcache_t *tcache, tcache_bin_t *tbin, size_t binind) { void *ret; - arena_tcache_fill_small(tcache->arena, tbin, binind -#ifdef JEMALLOC_PROF - , tcache->prof_accumbytes -#endif - ); -#ifdef JEMALLOC_PROF - tcache->prof_accumbytes = 0; -#endif + arena_tcache_fill_small(tcache->arena, tbin, binind, + config_prof ? tcache->prof_accumbytes : 0); + if (config_prof) + tcache->prof_accumbytes = 0; ret = tcache_alloc_easy(tbin); return (ret); } void -tcache_bin_flush_small(tcache_bin_t *tbin, size_t binind, unsigned rem -#if (defined(JEMALLOC_STATS) || defined(JEMALLOC_PROF)) - , tcache_t *tcache -#endif - ) +tcache_bin_flush_small(tcache_bin_t *tbin, size_t binind, unsigned rem, + tcache_t *tcache) { void *ptr; unsigned i, nflush, ndeferred; -#ifdef JEMALLOC_STATS bool merged_stats = false; -#endif - assert(binind < nbins); + assert(binind < NBINS); assert(rem <= tbin->ncached); for (nflush = tbin->ncached - rem; nflush > 0; nflush = ndeferred) { @@ -74,25 +96,21 @@ tcache_bin_flush_small(tcache_bin_t *tbin, size_t binind, unsigned rem arena_t *arena = chunk->arena; arena_bin_t *bin = &arena->bins[binind]; -#ifdef JEMALLOC_PROF - if (arena == tcache->arena) { + if (config_prof && arena == tcache->arena) { malloc_mutex_lock(&arena->lock); arena_prof_accum(arena, tcache->prof_accumbytes); malloc_mutex_unlock(&arena->lock); tcache->prof_accumbytes = 0; } -#endif malloc_mutex_lock(&bin->lock); -#ifdef JEMALLOC_STATS - if (arena == tcache->arena) { + if (config_stats && arena == tcache->arena) { assert(merged_stats == false); merged_stats = true; bin->stats.nflushes++; bin->stats.nrequests += tbin->tstats.nrequests; tbin->tstats.nrequests = 0; } -#endif ndeferred = 0; for (i = 0; i < nflush; i++) { ptr = tbin->avail[i]; @@ -100,10 +118,15 @@ tcache_bin_flush_small(tcache_bin_t *tbin, size_t binind, unsigned rem chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr); if (chunk->arena == arena) { size_t pageind = ((uintptr_t)ptr - - (uintptr_t)chunk) >> PAGE_SHIFT; + (uintptr_t)chunk) >> LG_PAGE; arena_chunk_map_t *mapelm = - &chunk->map[pageind-map_bias]; - arena_dalloc_bin(arena, chunk, ptr, mapelm); + arena_mapp_get(chunk, pageind); + if (config_fill && opt_junk) { + arena_alloc_junk_small(ptr, + &arena_bin_info[binind], true); + } + arena_dalloc_bin_locked(arena, chunk, ptr, + mapelm); } else { /* * This object was allocated via a different @@ -117,8 +140,7 @@ tcache_bin_flush_small(tcache_bin_t *tbin, size_t binind, unsigned rem } malloc_mutex_unlock(&bin->lock); } -#ifdef JEMALLOC_STATS - if (merged_stats == false) { + if (config_stats && merged_stats == false) { /* * The flush loop didn't happen to flush to this thread's * arena, so the stats didn't get merged. Manually do so now. @@ -130,7 +152,6 @@ tcache_bin_flush_small(tcache_bin_t *tbin, size_t binind, unsigned rem tbin->tstats.nrequests = 0; malloc_mutex_unlock(&bin->lock); } -#endif memmove(tbin->avail, &tbin->avail[tbin->ncached - rem], rem * sizeof(void *)); @@ -140,17 +161,12 @@ tcache_bin_flush_small(tcache_bin_t *tbin, size_t binind, unsigned rem } void -tcache_bin_flush_large(tcache_bin_t *tbin, size_t binind, unsigned rem -#if (defined(JEMALLOC_STATS) || defined(JEMALLOC_PROF)) - , tcache_t *tcache -#endif - ) +tcache_bin_flush_large(tcache_bin_t *tbin, size_t binind, unsigned rem, + tcache_t *tcache) { void *ptr; unsigned i, nflush, ndeferred; -#ifdef JEMALLOC_STATS bool merged_stats = false; -#endif assert(binind < nhbins); assert(rem <= tbin->ncached); @@ -162,30 +178,28 @@ tcache_bin_flush_large(tcache_bin_t *tbin, size_t binind, unsigned rem arena_t *arena = chunk->arena; malloc_mutex_lock(&arena->lock); -#if (defined(JEMALLOC_PROF) || defined(JEMALLOC_STATS)) - if (arena == tcache->arena) { -#endif -#ifdef JEMALLOC_PROF - arena_prof_accum(arena, tcache->prof_accumbytes); - tcache->prof_accumbytes = 0; -#endif -#ifdef JEMALLOC_STATS - merged_stats = true; - arena->stats.nrequests_large += tbin->tstats.nrequests; - arena->stats.lstats[binind - nbins].nrequests += - tbin->tstats.nrequests; - tbin->tstats.nrequests = 0; -#endif -#if (defined(JEMALLOC_PROF) || defined(JEMALLOC_STATS)) + if ((config_prof || config_stats) && arena == tcache->arena) { + if (config_prof) { + arena_prof_accum(arena, + tcache->prof_accumbytes); + tcache->prof_accumbytes = 0; + } + if (config_stats) { + merged_stats = true; + arena->stats.nrequests_large += + tbin->tstats.nrequests; + arena->stats.lstats[binind - NBINS].nrequests += + tbin->tstats.nrequests; + tbin->tstats.nrequests = 0; + } } -#endif ndeferred = 0; for (i = 0; i < nflush; i++) { ptr = tbin->avail[i]; assert(ptr != NULL); chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr); if (chunk->arena == arena) - arena_dalloc_large(arena, chunk, ptr); + arena_dalloc_large_locked(arena, chunk, ptr); else { /* * This object was allocated via a different @@ -199,8 +213,7 @@ tcache_bin_flush_large(tcache_bin_t *tbin, size_t binind, unsigned rem } malloc_mutex_unlock(&arena->lock); } -#ifdef JEMALLOC_STATS - if (merged_stats == false) { + if (config_stats && merged_stats == false) { /* * The flush loop didn't happen to flush to this thread's * arena, so the stats didn't get merged. Manually do so now. @@ -208,12 +221,11 @@ tcache_bin_flush_large(tcache_bin_t *tbin, size_t binind, unsigned rem arena_t *arena = tcache->arena; malloc_mutex_lock(&arena->lock); arena->stats.nrequests_large += tbin->tstats.nrequests; - arena->stats.lstats[binind - nbins].nrequests += + arena->stats.lstats[binind - NBINS].nrequests += tbin->tstats.nrequests; tbin->tstats.nrequests = 0; malloc_mutex_unlock(&arena->lock); } -#endif memmove(tbin->avail, &tbin->avail[tbin->ncached - rem], rem * sizeof(void *)); @@ -222,6 +234,33 @@ tcache_bin_flush_large(tcache_bin_t *tbin, size_t binind, unsigned rem tbin->low_water = tbin->ncached; } +void +tcache_arena_associate(tcache_t *tcache, arena_t *arena) +{ + + if (config_stats) { + /* Link into list of extant tcaches. */ + malloc_mutex_lock(&arena->lock); + ql_elm_new(tcache, link); + ql_tail_insert(&arena->tcache_ql, tcache, link); + malloc_mutex_unlock(&arena->lock); + } + tcache->arena = arena; +} + +void +tcache_arena_dissociate(tcache_t *tcache) +{ + + if (config_stats) { + /* Unlink from list of extant tcaches. */ + malloc_mutex_lock(&tcache->arena->lock); + ql_remove(&tcache->arena->tcache_ql, tcache, link); + malloc_mutex_unlock(&tcache->arena->lock); + tcache_stats_merge(tcache, tcache->arena); + } +} + tcache_t * tcache_create(arena_t *arena) { @@ -244,7 +283,7 @@ tcache_create(arena_t *arena) */ size = (size + CACHELINE_MASK) & (-CACHELINE); - if (size <= small_maxclass) + if (size <= SMALL_MAXCLASS) tcache = (tcache_t *)arena_malloc_small(arena, size, true); else if (size <= tcache_maxclass) tcache = (tcache_t *)arena_malloc_large(arena, size, true); @@ -254,15 +293,8 @@ tcache_create(arena_t *arena) if (tcache == NULL) return (NULL); -#ifdef JEMALLOC_STATS - /* Link into list of extant tcaches. */ - malloc_mutex_lock(&arena->lock); - ql_elm_new(tcache, link); - ql_tail_insert(&arena->tcache_ql, tcache, link); - malloc_mutex_unlock(&arena->lock); -#endif + tcache_arena_associate(tcache, arena); - tcache->arena = arena; assert((TCACHE_NSLOTS_SMALL_MAX & 1U) == 0); for (i = 0; i < nhbins; i++) { tcache->tbins[i].lg_fill_div = 1; @@ -271,7 +303,7 @@ tcache_create(arena_t *arena) stack_offset += tcache_bin_info[i].ncached_max * sizeof(void *); } - TCACHE_SET(tcache); + tcache_tsd_set(&tcache); return (tcache); } @@ -282,121 +314,96 @@ tcache_destroy(tcache_t *tcache) unsigned i; size_t tcache_size; -#ifdef JEMALLOC_STATS - /* Unlink from list of extant tcaches. */ - malloc_mutex_lock(&tcache->arena->lock); - ql_remove(&tcache->arena->tcache_ql, tcache, link); - malloc_mutex_unlock(&tcache->arena->lock); - tcache_stats_merge(tcache, tcache->arena); -#endif + tcache_arena_dissociate(tcache); - for (i = 0; i < nbins; i++) { + for (i = 0; i < NBINS; i++) { tcache_bin_t *tbin = &tcache->tbins[i]; - tcache_bin_flush_small(tbin, i, 0 -#if (defined(JEMALLOC_STATS) || defined(JEMALLOC_PROF)) - , tcache -#endif - ); + tcache_bin_flush_small(tbin, i, 0, tcache); -#ifdef JEMALLOC_STATS - if (tbin->tstats.nrequests != 0) { + if (config_stats && tbin->tstats.nrequests != 0) { arena_t *arena = tcache->arena; arena_bin_t *bin = &arena->bins[i]; malloc_mutex_lock(&bin->lock); bin->stats.nrequests += tbin->tstats.nrequests; malloc_mutex_unlock(&bin->lock); } -#endif } for (; i < nhbins; i++) { tcache_bin_t *tbin = &tcache->tbins[i]; - tcache_bin_flush_large(tbin, i, 0 -#if (defined(JEMALLOC_STATS) || defined(JEMALLOC_PROF)) - , tcache -#endif - ); + tcache_bin_flush_large(tbin, i, 0, tcache); -#ifdef JEMALLOC_STATS - if (tbin->tstats.nrequests != 0) { + if (config_stats && tbin->tstats.nrequests != 0) { arena_t *arena = tcache->arena; malloc_mutex_lock(&arena->lock); arena->stats.nrequests_large += tbin->tstats.nrequests; - arena->stats.lstats[i - nbins].nrequests += + arena->stats.lstats[i - NBINS].nrequests += tbin->tstats.nrequests; malloc_mutex_unlock(&arena->lock); } -#endif } -#ifdef JEMALLOC_PROF - if (tcache->prof_accumbytes > 0) { + if (config_prof && tcache->prof_accumbytes > 0) { malloc_mutex_lock(&tcache->arena->lock); arena_prof_accum(tcache->arena, tcache->prof_accumbytes); malloc_mutex_unlock(&tcache->arena->lock); } -#endif - tcache_size = arena_salloc(tcache); - if (tcache_size <= small_maxclass) { + tcache_size = arena_salloc(tcache, false); + if (tcache_size <= SMALL_MAXCLASS) { arena_chunk_t *chunk = CHUNK_ADDR2BASE(tcache); arena_t *arena = chunk->arena; size_t pageind = ((uintptr_t)tcache - (uintptr_t)chunk) >> - PAGE_SHIFT; - arena_chunk_map_t *mapelm = &chunk->map[pageind-map_bias]; - arena_run_t *run = (arena_run_t *)((uintptr_t)chunk + - (uintptr_t)((pageind - (mapelm->bits >> PAGE_SHIFT)) << - PAGE_SHIFT)); - arena_bin_t *bin = run->bin; + LG_PAGE; + arena_chunk_map_t *mapelm = arena_mapp_get(chunk, pageind); - malloc_mutex_lock(&bin->lock); - arena_dalloc_bin(arena, chunk, tcache, mapelm); - malloc_mutex_unlock(&bin->lock); + arena_dalloc_bin(arena, chunk, tcache, pageind, mapelm); } else if (tcache_size <= tcache_maxclass) { arena_chunk_t *chunk = CHUNK_ADDR2BASE(tcache); arena_t *arena = chunk->arena; - malloc_mutex_lock(&arena->lock); arena_dalloc_large(arena, chunk, tcache); - malloc_mutex_unlock(&arena->lock); } else idalloc(tcache); } -static void +void tcache_thread_cleanup(void *arg) { - tcache_t *tcache = (tcache_t *)arg; + tcache_t *tcache = *(tcache_t **)arg; - if (tcache == (void *)(uintptr_t)1) { - /* - * The previous time this destructor was called, we set the key - * to 1 so that other destructors wouldn't cause re-creation of - * the tcache. This time, do nothing, so that the destructor - * will not be called again. - */ - } else if (tcache == (void *)(uintptr_t)2) { + if (tcache == TCACHE_STATE_DISABLED) { + /* Do nothing. */ + } else if (tcache == TCACHE_STATE_REINCARNATED) { /* * Another destructor called an allocator function after this - * destructor was called. Reset tcache to 1 in order to - * receive another callback. + * destructor was called. Reset tcache to + * TCACHE_STATE_PURGATORY in order to receive another callback. + */ + tcache = TCACHE_STATE_PURGATORY; + tcache_tsd_set(&tcache); + } else if (tcache == TCACHE_STATE_PURGATORY) { + /* + * The previous time this destructor was called, we set the key + * to TCACHE_STATE_PURGATORY so that other destructors wouldn't + * cause re-creation of the tcache. This time, do nothing, so + * that the destructor will not be called again. */ - TCACHE_SET((uintptr_t)1); } else if (tcache != NULL) { - assert(tcache != (void *)(uintptr_t)1); + assert(tcache != TCACHE_STATE_PURGATORY); tcache_destroy(tcache); - TCACHE_SET((uintptr_t)1); + tcache = TCACHE_STATE_PURGATORY; + tcache_tsd_set(&tcache); } } -#ifdef JEMALLOC_STATS void tcache_stats_merge(tcache_t *tcache, arena_t *arena) { unsigned i; /* Merge and reset tcache stats. */ - for (i = 0; i < nbins; i++) { + for (i = 0; i < NBINS; i++) { arena_bin_t *bin = &arena->bins[i]; tcache_bin_t *tbin = &tcache->tbins[i]; malloc_mutex_lock(&bin->lock); @@ -406,75 +413,62 @@ tcache_stats_merge(tcache_t *tcache, arena_t *arena) } for (; i < nhbins; i++) { - malloc_large_stats_t *lstats = &arena->stats.lstats[i - nbins]; + malloc_large_stats_t *lstats = &arena->stats.lstats[i - NBINS]; tcache_bin_t *tbin = &tcache->tbins[i]; arena->stats.nrequests_large += tbin->tstats.nrequests; lstats->nrequests += tbin->tstats.nrequests; tbin->tstats.nrequests = 0; } } -#endif bool -tcache_boot(void) +tcache_boot0(void) { + unsigned i; - if (opt_tcache) { - unsigned i; + /* + * If necessary, clamp opt_lg_tcache_max, now that arena_maxclass is + * known. + */ + if (opt_lg_tcache_max < 0 || (1U << opt_lg_tcache_max) < SMALL_MAXCLASS) + tcache_maxclass = SMALL_MAXCLASS; + else if ((1U << opt_lg_tcache_max) > arena_maxclass) + tcache_maxclass = arena_maxclass; + else + tcache_maxclass = (1U << opt_lg_tcache_max); - /* - * If necessary, clamp opt_lg_tcache_max, now that - * small_maxclass and arena_maxclass are known. - */ - if (opt_lg_tcache_max < 0 || (1U << - opt_lg_tcache_max) < small_maxclass) - tcache_maxclass = small_maxclass; - else if ((1U << opt_lg_tcache_max) > arena_maxclass) - tcache_maxclass = arena_maxclass; - else - tcache_maxclass = (1U << opt_lg_tcache_max); + nhbins = NBINS + (tcache_maxclass >> LG_PAGE); - nhbins = nbins + (tcache_maxclass >> PAGE_SHIFT); - - /* Initialize tcache_bin_info. */ - tcache_bin_info = (tcache_bin_info_t *)base_alloc(nhbins * - sizeof(tcache_bin_info_t)); - if (tcache_bin_info == NULL) - return (true); - stack_nelms = 0; - for (i = 0; i < nbins; i++) { - if ((arena_bin_info[i].nregs << 1) <= - TCACHE_NSLOTS_SMALL_MAX) { - tcache_bin_info[i].ncached_max = - (arena_bin_info[i].nregs << 1); - } else { - tcache_bin_info[i].ncached_max = - TCACHE_NSLOTS_SMALL_MAX; - } - stack_nelms += tcache_bin_info[i].ncached_max; - } - for (; i < nhbins; i++) { - tcache_bin_info[i].ncached_max = TCACHE_NSLOTS_LARGE; - stack_nelms += tcache_bin_info[i].ncached_max; - } - - /* Compute incremental GC event threshold. */ - if (opt_lg_tcache_gc_sweep >= 0) { - tcache_gc_incr = ((1U << opt_lg_tcache_gc_sweep) / - nbins) + (((1U << opt_lg_tcache_gc_sweep) % nbins == - 0) ? 0 : 1); - } else - tcache_gc_incr = 0; - - if (pthread_key_create(&tcache_tsd, tcache_thread_cleanup) != - 0) { - malloc_write( - ": Error in pthread_key_create()\n"); - abort(); + /* Initialize tcache_bin_info. */ + tcache_bin_info = (tcache_bin_info_t *)base_alloc(nhbins * + sizeof(tcache_bin_info_t)); + if (tcache_bin_info == NULL) + return (true); + stack_nelms = 0; + for (i = 0; i < NBINS; i++) { + if ((arena_bin_info[i].nregs << 1) <= TCACHE_NSLOTS_SMALL_MAX) { + tcache_bin_info[i].ncached_max = + (arena_bin_info[i].nregs << 1); + } else { + tcache_bin_info[i].ncached_max = + TCACHE_NSLOTS_SMALL_MAX; } + stack_nelms += tcache_bin_info[i].ncached_max; + } + for (; i < nhbins; i++) { + tcache_bin_info[i].ncached_max = TCACHE_NSLOTS_LARGE; + stack_nelms += tcache_bin_info[i].ncached_max; } return (false); } -/******************************************************************************/ -#endif /* JEMALLOC_TCACHE */ + +bool +tcache_boot1(void) +{ + + if (tcache_tsd_boot() || tcache_enabled_tsd_boot()) + return (true); + + return (false); +} diff --git a/src/tsd.c b/src/tsd.c new file mode 100644 index 00000000..961a5463 --- /dev/null +++ b/src/tsd.c @@ -0,0 +1,107 @@ +#define JEMALLOC_TSD_C_ +#include "jemalloc/internal/jemalloc_internal.h" + +/******************************************************************************/ +/* Data. */ + +static unsigned ncleanups; +static malloc_tsd_cleanup_t cleanups[MALLOC_TSD_CLEANUPS_MAX]; + +/******************************************************************************/ + +void * +malloc_tsd_malloc(size_t size) +{ + + /* Avoid choose_arena() in order to dodge bootstrapping issues. */ + return (arena_malloc(arenas[0], size, false, false)); +} + +void +malloc_tsd_dalloc(void *wrapper) +{ + + idalloc(wrapper); +} + +void +malloc_tsd_no_cleanup(void *arg) +{ + + not_reached(); +} + +#if defined(JEMALLOC_MALLOC_THREAD_CLEANUP) || defined(_WIN32) +#ifndef _WIN32 +JEMALLOC_EXPORT +#endif +void +_malloc_thread_cleanup(void) +{ + bool pending[MALLOC_TSD_CLEANUPS_MAX], again; + unsigned i; + + for (i = 0; i < ncleanups; i++) + pending[i] = true; + + do { + again = false; + for (i = 0; i < ncleanups; i++) { + if (pending[i]) { + pending[i] = cleanups[i](); + if (pending[i]) + again = true; + } + } + } while (again); +} +#endif + +void +malloc_tsd_cleanup_register(bool (*f)(void)) +{ + + assert(ncleanups < MALLOC_TSD_CLEANUPS_MAX); + cleanups[ncleanups] = f; + ncleanups++; +} + +void +malloc_tsd_boot(void) +{ + + ncleanups = 0; +} + +#ifdef _WIN32 +static BOOL WINAPI +_tls_callback(HINSTANCE hinstDLL, DWORD fdwReason, LPVOID lpvReserved) +{ + + switch (fdwReason) { +#ifdef JEMALLOC_LAZY_LOCK + case DLL_THREAD_ATTACH: + isthreaded = true; + break; +#endif + case DLL_THREAD_DETACH: + _malloc_thread_cleanup(); + break; + default: + break; + } + return (true); +} + +#ifdef _MSC_VER +# ifdef _M_IX86 +# pragma comment(linker, "/INCLUDE:__tls_used") +# else +# pragma comment(linker, "/INCLUDE:_tls_used") +# endif +# pragma section(".CRT$XLY",long,read) +#endif +JEMALLOC_SECTION(".CRT$XLY") JEMALLOC_ATTR(used) +static const BOOL (WINAPI *tls_callback)(HINSTANCE hinstDLL, + DWORD fdwReason, LPVOID lpvReserved) = _tls_callback; +#endif diff --git a/src/util.c b/src/util.c new file mode 100644 index 00000000..9b73c3ec --- /dev/null +++ b/src/util.c @@ -0,0 +1,646 @@ +#define assert(e) do { \ + if (config_debug && !(e)) { \ + malloc_write(": Failed assertion\n"); \ + abort(); \ + } \ +} while (0) + +#define not_reached() do { \ + if (config_debug) { \ + malloc_write(": Unreachable code reached\n"); \ + abort(); \ + } \ +} while (0) + +#define not_implemented() do { \ + if (config_debug) { \ + malloc_write(": Not implemented\n"); \ + abort(); \ + } \ +} while (0) + +#define JEMALLOC_UTIL_C_ +#include "jemalloc/internal/jemalloc_internal.h" + +/******************************************************************************/ +/* Function prototypes for non-inline static functions. */ + +static void wrtmessage(void *cbopaque, const char *s); +#define U2S_BUFSIZE ((1U << (LG_SIZEOF_INTMAX_T + 3)) + 1) +static char *u2s(uintmax_t x, unsigned base, bool uppercase, char *s, + size_t *slen_p); +#define D2S_BUFSIZE (1 + U2S_BUFSIZE) +static char *d2s(intmax_t x, char sign, char *s, size_t *slen_p); +#define O2S_BUFSIZE (1 + U2S_BUFSIZE) +static char *o2s(uintmax_t x, bool alt_form, char *s, size_t *slen_p); +#define X2S_BUFSIZE (2 + U2S_BUFSIZE) +static char *x2s(uintmax_t x, bool alt_form, bool uppercase, char *s, + size_t *slen_p); + +/******************************************************************************/ + +/* malloc_message() setup. */ +static void +wrtmessage(void *cbopaque, const char *s) +{ + +#ifdef SYS_write + /* + * Use syscall(2) rather than write(2) when possible in order to avoid + * the possibility of memory allocation within libc. This is necessary + * on FreeBSD; most operating systems do not have this problem though. + */ + UNUSED int result = syscall(SYS_write, STDERR_FILENO, s, strlen(s)); +#else + UNUSED int result = write(STDERR_FILENO, s, strlen(s)); +#endif +} + +JEMALLOC_EXPORT void (*je_malloc_message)(void *, const char *s); + +/* + * Wrapper around malloc_message() that avoids the need for + * je_malloc_message(...) throughout the code. + */ +void +malloc_write(const char *s) +{ + + if (je_malloc_message != NULL) + je_malloc_message(NULL, s); + else + wrtmessage(NULL, s); +} + +/* + * glibc provides a non-standard strerror_r() when _GNU_SOURCE is defined, so + * provide a wrapper. + */ +int +buferror(char *buf, size_t buflen) +{ + +#ifdef _WIN32 + FormatMessageA(FORMAT_MESSAGE_FROM_SYSTEM, NULL, GetLastError(), 0, + (LPSTR)buf, buflen, NULL); + return (0); +#elif defined(_GNU_SOURCE) + char *b = strerror_r(errno, buf, buflen); + if (b != buf) { + strncpy(buf, b, buflen); + buf[buflen-1] = '\0'; + } + return (0); +#else + return (strerror_r(errno, buf, buflen)); +#endif +} + +uintmax_t +malloc_strtoumax(const char *nptr, char **endptr, int base) +{ + uintmax_t ret, digit; + int b; + bool neg; + const char *p, *ns; + + if (base < 0 || base == 1 || base > 36) { + set_errno(EINVAL); + return (UINTMAX_MAX); + } + b = base; + + /* Swallow leading whitespace and get sign, if any. */ + neg = false; + p = nptr; + while (true) { + switch (*p) { + case '\t': case '\n': case '\v': case '\f': case '\r': case ' ': + p++; + break; + case '-': + neg = true; + /* Fall through. */ + case '+': + p++; + /* Fall through. */ + default: + goto label_prefix; + } + } + + /* Get prefix, if any. */ + label_prefix: + /* + * Note where the first non-whitespace/sign character is so that it is + * possible to tell whether any digits are consumed (e.g., " 0" vs. + * " -x"). + */ + ns = p; + if (*p == '0') { + switch (p[1]) { + case '0': case '1': case '2': case '3': case '4': case '5': + case '6': case '7': + if (b == 0) + b = 8; + if (b == 8) + p++; + break; + case 'x': + switch (p[2]) { + case '0': case '1': case '2': case '3': case '4': + case '5': case '6': case '7': case '8': case '9': + case 'A': case 'B': case 'C': case 'D': case 'E': + case 'F': + case 'a': case 'b': case 'c': case 'd': case 'e': + case 'f': + if (b == 0) + b = 16; + if (b == 16) + p += 2; + break; + default: + break; + } + break; + default: + break; + } + } + if (b == 0) + b = 10; + + /* Convert. */ + ret = 0; + while ((*p >= '0' && *p <= '9' && (digit = *p - '0') < b) + || (*p >= 'A' && *p <= 'Z' && (digit = 10 + *p - 'A') < b) + || (*p >= 'a' && *p <= 'z' && (digit = 10 + *p - 'a') < b)) { + uintmax_t pret = ret; + ret *= b; + ret += digit; + if (ret < pret) { + /* Overflow. */ + set_errno(ERANGE); + return (UINTMAX_MAX); + } + p++; + } + if (neg) + ret = -ret; + + if (endptr != NULL) { + if (p == ns) { + /* No characters were converted. */ + *endptr = (char *)nptr; + } else + *endptr = (char *)p; + } + + return (ret); +} + +static char * +u2s(uintmax_t x, unsigned base, bool uppercase, char *s, size_t *slen_p) +{ + unsigned i; + + i = U2S_BUFSIZE - 1; + s[i] = '\0'; + switch (base) { + case 10: + do { + i--; + s[i] = "0123456789"[x % (uint64_t)10]; + x /= (uint64_t)10; + } while (x > 0); + break; + case 16: { + const char *digits = (uppercase) + ? "0123456789ABCDEF" + : "0123456789abcdef"; + + do { + i--; + s[i] = digits[x & 0xf]; + x >>= 4; + } while (x > 0); + break; + } default: { + const char *digits = (uppercase) + ? "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ" + : "0123456789abcdefghijklmnopqrstuvwxyz"; + + assert(base >= 2 && base <= 36); + do { + i--; + s[i] = digits[x % (uint64_t)base]; + x /= (uint64_t)base; + } while (x > 0); + }} + + *slen_p = U2S_BUFSIZE - 1 - i; + return (&s[i]); +} + +static char * +d2s(intmax_t x, char sign, char *s, size_t *slen_p) +{ + bool neg; + + if ((neg = (x < 0))) + x = -x; + s = u2s(x, 10, false, s, slen_p); + if (neg) + sign = '-'; + switch (sign) { + case '-': + if (neg == false) + break; + /* Fall through. */ + case ' ': + case '+': + s--; + (*slen_p)++; + *s = sign; + break; + default: not_reached(); + } + return (s); +} + +static char * +o2s(uintmax_t x, bool alt_form, char *s, size_t *slen_p) +{ + + s = u2s(x, 8, false, s, slen_p); + if (alt_form && *s != '0') { + s--; + (*slen_p)++; + *s = '0'; + } + return (s); +} + +static char * +x2s(uintmax_t x, bool alt_form, bool uppercase, char *s, size_t *slen_p) +{ + + s = u2s(x, 16, uppercase, s, slen_p); + if (alt_form) { + s -= 2; + (*slen_p) += 2; + memcpy(s, uppercase ? "0X" : "0x", 2); + } + return (s); +} + +int +malloc_vsnprintf(char *str, size_t size, const char *format, va_list ap) +{ + int ret; + size_t i; + const char *f; + +#define APPEND_C(c) do { \ + if (i < size) \ + str[i] = (c); \ + i++; \ +} while (0) +#define APPEND_S(s, slen) do { \ + if (i < size) { \ + size_t cpylen = (slen <= size - i) ? slen : size - i; \ + memcpy(&str[i], s, cpylen); \ + } \ + i += slen; \ +} while (0) +#define APPEND_PADDED_S(s, slen, width, left_justify) do { \ + /* Left padding. */ \ + size_t pad_len = (width == -1) ? 0 : ((slen < (size_t)width) ? \ + (size_t)width - slen : 0); \ + if (left_justify == false && pad_len != 0) { \ + size_t j; \ + for (j = 0; j < pad_len; j++) \ + APPEND_C(' '); \ + } \ + /* Value. */ \ + APPEND_S(s, slen); \ + /* Right padding. */ \ + if (left_justify && pad_len != 0) { \ + size_t j; \ + for (j = 0; j < pad_len; j++) \ + APPEND_C(' '); \ + } \ +} while (0) +#define GET_ARG_NUMERIC(val, len) do { \ + switch (len) { \ + case '?': \ + val = va_arg(ap, int); \ + break; \ + case '?' | 0x80: \ + val = va_arg(ap, unsigned int); \ + break; \ + case 'l': \ + val = va_arg(ap, long); \ + break; \ + case 'l' | 0x80: \ + val = va_arg(ap, unsigned long); \ + break; \ + case 'q': \ + val = va_arg(ap, long long); \ + break; \ + case 'q' | 0x80: \ + val = va_arg(ap, unsigned long long); \ + break; \ + case 'j': \ + val = va_arg(ap, intmax_t); \ + break; \ + case 't': \ + val = va_arg(ap, ptrdiff_t); \ + break; \ + case 'z': \ + val = va_arg(ap, ssize_t); \ + break; \ + case 'z' | 0x80: \ + val = va_arg(ap, size_t); \ + break; \ + case 'p': /* Synthetic; used for %p. */ \ + val = va_arg(ap, uintptr_t); \ + break; \ + default: not_reached(); \ + } \ +} while (0) + + i = 0; + f = format; + while (true) { + switch (*f) { + case '\0': goto label_out; + case '%': { + bool alt_form = false; + bool zero_pad = false; + bool left_justify = false; + bool plus_space = false; + bool plus_plus = false; + int prec = -1; + int width = -1; + unsigned char len = '?'; + + f++; + if (*f == '%') { + /* %% */ + APPEND_C(*f); + break; + } + /* Flags. */ + while (true) { + switch (*f) { + case '#': + assert(alt_form == false); + alt_form = true; + break; + case '0': + assert(zero_pad == false); + zero_pad = true; + break; + case '-': + assert(left_justify == false); + left_justify = true; + break; + case ' ': + assert(plus_space == false); + plus_space = true; + break; + case '+': + assert(plus_plus == false); + plus_plus = true; + break; + default: goto label_width; + } + f++; + } + /* Width. */ + label_width: + switch (*f) { + case '*': + width = va_arg(ap, int); + f++; + break; + case '0': case '1': case '2': case '3': case '4': + case '5': case '6': case '7': case '8': case '9': { + uintmax_t uwidth; + set_errno(0); + uwidth = malloc_strtoumax(f, (char **)&f, 10); + assert(uwidth != UINTMAX_MAX || get_errno() != + ERANGE); + width = (int)uwidth; + if (*f == '.') { + f++; + goto label_precision; + } else + goto label_length; + break; + } case '.': + f++; + goto label_precision; + default: goto label_length; + } + /* Precision. */ + label_precision: + switch (*f) { + case '*': + prec = va_arg(ap, int); + f++; + break; + case '0': case '1': case '2': case '3': case '4': + case '5': case '6': case '7': case '8': case '9': { + uintmax_t uprec; + set_errno(0); + uprec = malloc_strtoumax(f, (char **)&f, 10); + assert(uprec != UINTMAX_MAX || get_errno() != + ERANGE); + prec = (int)uprec; + break; + } + default: break; + } + /* Length. */ + label_length: + switch (*f) { + case 'l': + f++; + if (*f == 'l') { + len = 'q'; + f++; + } else + len = 'l'; + break; + case 'j': + len = 'j'; + f++; + break; + case 't': + len = 't'; + f++; + break; + case 'z': + len = 'z'; + f++; + break; + default: break; + } + /* Conversion specifier. */ + switch (*f) { + char *s; + size_t slen; + case 'd': case 'i': { + intmax_t val JEMALLOC_CC_SILENCE_INIT(0); + char buf[D2S_BUFSIZE]; + + GET_ARG_NUMERIC(val, len); + s = d2s(val, (plus_plus ? '+' : (plus_space ? + ' ' : '-')), buf, &slen); + APPEND_PADDED_S(s, slen, width, left_justify); + f++; + break; + } case 'o': { + uintmax_t val JEMALLOC_CC_SILENCE_INIT(0); + char buf[O2S_BUFSIZE]; + + GET_ARG_NUMERIC(val, len | 0x80); + s = o2s(val, alt_form, buf, &slen); + APPEND_PADDED_S(s, slen, width, left_justify); + f++; + break; + } case 'u': { + uintmax_t val JEMALLOC_CC_SILENCE_INIT(0); + char buf[U2S_BUFSIZE]; + + GET_ARG_NUMERIC(val, len | 0x80); + s = u2s(val, 10, false, buf, &slen); + APPEND_PADDED_S(s, slen, width, left_justify); + f++; + break; + } case 'x': case 'X': { + uintmax_t val JEMALLOC_CC_SILENCE_INIT(0); + char buf[X2S_BUFSIZE]; + + GET_ARG_NUMERIC(val, len | 0x80); + s = x2s(val, alt_form, *f == 'X', buf, &slen); + APPEND_PADDED_S(s, slen, width, left_justify); + f++; + break; + } case 'c': { + unsigned char val; + char buf[2]; + + assert(len == '?' || len == 'l'); + assert_not_implemented(len != 'l'); + val = va_arg(ap, int); + buf[0] = val; + buf[1] = '\0'; + APPEND_PADDED_S(buf, 1, width, left_justify); + f++; + break; + } case 's': + assert(len == '?' || len == 'l'); + assert_not_implemented(len != 'l'); + s = va_arg(ap, char *); + slen = (prec == -1) ? strlen(s) : prec; + APPEND_PADDED_S(s, slen, width, left_justify); + f++; + break; + case 'p': { + uintmax_t val; + char buf[X2S_BUFSIZE]; + + GET_ARG_NUMERIC(val, 'p'); + s = x2s(val, true, false, buf, &slen); + APPEND_PADDED_S(s, slen, width, left_justify); + f++; + break; + } + default: not_implemented(); + } + break; + } default: { + APPEND_C(*f); + f++; + break; + }} + } + label_out: + if (i < size) + str[i] = '\0'; + else + str[size - 1] = '\0'; + ret = i; + +#undef APPEND_C +#undef APPEND_S +#undef APPEND_PADDED_S +#undef GET_ARG_NUMERIC + return (ret); +} + +JEMALLOC_ATTR(format(printf, 3, 4)) +int +malloc_snprintf(char *str, size_t size, const char *format, ...) +{ + int ret; + va_list ap; + + va_start(ap, format); + ret = malloc_vsnprintf(str, size, format, ap); + va_end(ap); + + return (ret); +} + +void +malloc_vcprintf(void (*write_cb)(void *, const char *), void *cbopaque, + const char *format, va_list ap) +{ + char buf[MALLOC_PRINTF_BUFSIZE]; + + if (write_cb == NULL) { + /* + * The caller did not provide an alternate write_cb callback + * function, so use the default one. malloc_write() is an + * inline function, so use malloc_message() directly here. + */ + write_cb = (je_malloc_message != NULL) ? je_malloc_message : + wrtmessage; + cbopaque = NULL; + } + + malloc_vsnprintf(buf, sizeof(buf), format, ap); + write_cb(cbopaque, buf); +} + +/* + * Print to a callback function in such a way as to (hopefully) avoid memory + * allocation. + */ +JEMALLOC_ATTR(format(printf, 3, 4)) +void +malloc_cprintf(void (*write_cb)(void *, const char *), void *cbopaque, + const char *format, ...) +{ + va_list ap; + + va_start(ap, format); + malloc_vcprintf(write_cb, cbopaque, format, ap); + va_end(ap); +} + +/* Print to stderr in such a way as to avoid memory allocation. */ +JEMALLOC_ATTR(format(printf, 1, 2)) +void +malloc_printf(const char *format, ...) +{ + va_list ap; + + va_start(ap, format); + malloc_vcprintf(NULL, NULL, format, ap); + va_end(ap); +} diff --git a/src/zone.c b/src/zone.c index 2c1b2318..cde5d49a 100644 --- a/src/zone.c +++ b/src/zone.c @@ -3,11 +3,18 @@ # error "This source file is for zones on Darwin (OS X)." #endif +/* + * The malloc_default_purgeable_zone function is only available on >= 10.6. + * We need to check whether it is present at runtime, thus the weak_import. + */ +extern malloc_zone_t *malloc_default_purgeable_zone(void) +JEMALLOC_ATTR(weak_import); + /******************************************************************************/ /* Data. */ -static malloc_zone_t zone, szone; -static struct malloc_introspection_t zone_introspect, ozone_introspect; +static malloc_zone_t zone; +static struct malloc_introspection_t zone_introspect; /******************************************************************************/ /* Function prototypes for non-inline static functions. */ @@ -18,8 +25,10 @@ static void *zone_calloc(malloc_zone_t *zone, size_t num, size_t size); static void *zone_valloc(malloc_zone_t *zone, size_t size); static void zone_free(malloc_zone_t *zone, void *ptr); static void *zone_realloc(malloc_zone_t *zone, void *ptr, size_t size); -#if (JEMALLOC_ZONE_VERSION >= 6) +#if (JEMALLOC_ZONE_VERSION >= 5) static void *zone_memalign(malloc_zone_t *zone, size_t alignment, +#endif +#if (JEMALLOC_ZONE_VERSION >= 6) size_t size); static void zone_free_definite_size(malloc_zone_t *zone, void *ptr, size_t size); @@ -28,19 +37,6 @@ static void *zone_destroy(malloc_zone_t *zone); static size_t zone_good_size(malloc_zone_t *zone, size_t size); static void zone_force_lock(malloc_zone_t *zone); static void zone_force_unlock(malloc_zone_t *zone); -static size_t ozone_size(malloc_zone_t *zone, void *ptr); -static void ozone_free(malloc_zone_t *zone, void *ptr); -static void *ozone_realloc(malloc_zone_t *zone, void *ptr, size_t size); -static unsigned ozone_batch_malloc(malloc_zone_t *zone, size_t size, - void **results, unsigned num_requested); -static void ozone_batch_free(malloc_zone_t *zone, void **to_be_freed, - unsigned num); -#if (JEMALLOC_ZONE_VERSION >= 6) -static void ozone_free_definite_size(malloc_zone_t *zone, void *ptr, - size_t size); -#endif -static void ozone_force_lock(malloc_zone_t *zone); -static void ozone_force_unlock(malloc_zone_t *zone); /******************************************************************************/ /* @@ -60,21 +56,21 @@ zone_size(malloc_zone_t *zone, void *ptr) * not work in practice, we must check all pointers to assure that they * reside within a mapped chunk before determining size. */ - return (ivsalloc(ptr)); + return (ivsalloc(ptr, config_prof)); } static void * zone_malloc(malloc_zone_t *zone, size_t size) { - return (JEMALLOC_P(malloc)(size)); + return (je_malloc(size)); } static void * zone_calloc(malloc_zone_t *zone, size_t num, size_t size) { - return (JEMALLOC_P(calloc)(num, size)); + return (je_calloc(num, size)); } static void * @@ -82,7 +78,7 @@ zone_valloc(malloc_zone_t *zone, size_t size) { void *ret = NULL; /* Assignment avoids useless compiler warning. */ - JEMALLOC_P(posix_memalign)(&ret, PAGE_SIZE, size); + je_posix_memalign(&ret, PAGE, size); return (ret); } @@ -91,33 +87,48 @@ static void zone_free(malloc_zone_t *zone, void *ptr) { - JEMALLOC_P(free)(ptr); + if (ivsalloc(ptr, config_prof) != 0) { + je_free(ptr); + return; + } + + free(ptr); } static void * zone_realloc(malloc_zone_t *zone, void *ptr, size_t size) { - return (JEMALLOC_P(realloc)(ptr, size)); + if (ivsalloc(ptr, config_prof) != 0) + return (je_realloc(ptr, size)); + + return (realloc(ptr, size)); } -#if (JEMALLOC_ZONE_VERSION >= 6) +#if (JEMALLOC_ZONE_VERSION >= 5) static void * zone_memalign(malloc_zone_t *zone, size_t alignment, size_t size) { void *ret = NULL; /* Assignment avoids useless compiler warning. */ - JEMALLOC_P(posix_memalign)(&ret, alignment, size); + je_posix_memalign(&ret, alignment, size); return (ret); } +#endif +#if (JEMALLOC_ZONE_VERSION >= 6) static void zone_free_definite_size(malloc_zone_t *zone, void *ptr, size_t size) { - assert(ivsalloc(ptr) == size); - JEMALLOC_P(free)(ptr); + if (ivsalloc(ptr, config_prof) != 0) { + assert(ivsalloc(ptr, config_prof) == size); + je_free(ptr); + return; + } + + free(ptr); } #endif @@ -133,22 +144,10 @@ zone_destroy(malloc_zone_t *zone) static size_t zone_good_size(malloc_zone_t *zone, size_t size) { - size_t ret; - void *p; - /* - * Actually create an object of the appropriate size, then find out - * how large it could have been without moving up to the next size - * class. - */ - p = JEMALLOC_P(malloc)(size); - if (p != NULL) { - ret = isalloc(p); - JEMALLOC_P(free)(p); - } else - ret = size; - - return (ret); + if (size == 0) + size = 1; + return (s2u(size)); } static void @@ -164,11 +163,12 @@ zone_force_unlock(malloc_zone_t *zone) { if (isthreaded) - jemalloc_postfork(); + jemalloc_postfork_parent(); } -malloc_zone_t * -create_zone(void) +JEMALLOC_ATTR(constructor) +void +register_zone(void) { zone.size = (void *)zone_size; @@ -183,10 +183,15 @@ create_zone(void) zone.batch_free = NULL; zone.introspect = &zone_introspect; zone.version = JEMALLOC_ZONE_VERSION; -#if (JEMALLOC_ZONE_VERSION >= 6) +#if (JEMALLOC_ZONE_VERSION >= 5) zone.memalign = zone_memalign; +#endif +#if (JEMALLOC_ZONE_VERSION >= 6) zone.free_definite_size = zone_free_definite_size; #endif +#if (JEMALLOC_ZONE_VERSION >= 8) + zone.pressure_relief = NULL; +#endif zone_introspect.enumerator = NULL; zone_introspect.good_size = (void *)zone_good_size; @@ -199,156 +204,45 @@ create_zone(void) #if (JEMALLOC_ZONE_VERSION >= 6) zone_introspect.zone_locked = NULL; #endif - - return (&zone); -} - -static size_t -ozone_size(malloc_zone_t *zone, void *ptr) -{ - size_t ret; - - ret = ivsalloc(ptr); - if (ret == 0) - ret = szone.size(zone, ptr); - - return (ret); -} - -static void -ozone_free(malloc_zone_t *zone, void *ptr) -{ - - if (ivsalloc(ptr) != 0) - JEMALLOC_P(free)(ptr); - else { - size_t size = szone.size(zone, ptr); - if (size != 0) - (szone.free)(zone, ptr); - } -} - -static void * -ozone_realloc(malloc_zone_t *zone, void *ptr, size_t size) -{ - size_t oldsize; - - if (ptr == NULL) - return (JEMALLOC_P(malloc)(size)); - - oldsize = ivsalloc(ptr); - if (oldsize != 0) - return (JEMALLOC_P(realloc)(ptr, size)); - else { - oldsize = szone.size(zone, ptr); - if (oldsize == 0) - return (JEMALLOC_P(malloc)(size)); - else { - void *ret = JEMALLOC_P(malloc)(size); - if (ret != NULL) { - memcpy(ret, ptr, (oldsize < size) ? oldsize : - size); - (szone.free)(zone, ptr); - } - return (ret); - } - } -} - -static unsigned -ozone_batch_malloc(malloc_zone_t *zone, size_t size, void **results, - unsigned num_requested) -{ - - /* Don't bother implementing this interface, since it isn't required. */ - return (0); -} - -static void -ozone_batch_free(malloc_zone_t *zone, void **to_be_freed, unsigned num) -{ - unsigned i; - - for (i = 0; i < num; i++) - ozone_free(zone, to_be_freed[i]); -} - -#if (JEMALLOC_ZONE_VERSION >= 6) -static void -ozone_free_definite_size(malloc_zone_t *zone, void *ptr, size_t size) -{ - - if (ivsalloc(ptr) != 0) { - assert(ivsalloc(ptr) == size); - JEMALLOC_P(free)(ptr); - } else { - assert(size == szone.size(zone, ptr)); - szone.free_definite_size(zone, ptr, size); - } -} +#if (JEMALLOC_ZONE_VERSION >= 7) + zone_introspect.enable_discharge_checking = NULL; + zone_introspect.disable_discharge_checking = NULL; + zone_introspect.discharge = NULL; +#ifdef __BLOCKS__ + zone_introspect.enumerate_discharged_pointers = NULL; +#else + zone_introspect.enumerate_unavailable_without_blocks = NULL; +#endif #endif - -static void -ozone_force_lock(malloc_zone_t *zone) -{ - - /* jemalloc locking is taken care of by the normal jemalloc zone. */ - szone.introspect->force_lock(zone); -} - -static void -ozone_force_unlock(malloc_zone_t *zone) -{ - - /* jemalloc locking is taken care of by the normal jemalloc zone. */ - szone.introspect->force_unlock(zone); -} - -/* - * Overlay the default scalable zone (szone) such that existing allocations are - * drained, and further allocations come from jemalloc. This is necessary - * because Core Foundation directly accesses and uses the szone before the - * jemalloc library is even loaded. - */ -void -szone2ozone(malloc_zone_t *zone) -{ /* - * Stash a copy of the original szone so that we can call its - * functions as needed. Note that the internally, the szone stores its - * bookkeeping data structures immediately following the malloc_zone_t - * header, so when calling szone functions, we need to pass a pointer - * to the original zone structure. + * The default purgeable zone is created lazily by OSX's libc. It uses + * the default zone when it is created for "small" allocations + * (< 15 KiB), but assumes the default zone is a scalable_zone. This + * obviously fails when the default zone is the jemalloc zone, so + * malloc_default_purgeable_zone is called beforehand so that the + * default purgeable zone is created when the default zone is still + * a scalable_zone. As purgeable zones only exist on >= 10.6, we need + * to check for the existence of malloc_default_purgeable_zone() at + * run time. */ - memcpy(&szone, zone, sizeof(malloc_zone_t)); + if (malloc_default_purgeable_zone != NULL) + malloc_default_purgeable_zone(); - zone->size = (void *)ozone_size; - zone->malloc = (void *)zone_malloc; - zone->calloc = (void *)zone_calloc; - zone->valloc = (void *)zone_valloc; - zone->free = (void *)ozone_free; - zone->realloc = (void *)ozone_realloc; - zone->destroy = (void *)zone_destroy; - zone->zone_name = "jemalloc_ozone"; - zone->batch_malloc = ozone_batch_malloc; - zone->batch_free = ozone_batch_free; - zone->introspect = &ozone_introspect; - zone->version = JEMALLOC_ZONE_VERSION; -#if (JEMALLOC_ZONE_VERSION >= 6) - zone->memalign = zone_memalign; - zone->free_definite_size = ozone_free_definite_size; -#endif + /* Register the custom zone. At this point it won't be the default. */ + malloc_zone_register(&zone); - ozone_introspect.enumerator = NULL; - ozone_introspect.good_size = (void *)zone_good_size; - ozone_introspect.check = NULL; - ozone_introspect.print = NULL; - ozone_introspect.log = NULL; - ozone_introspect.force_lock = (void *)ozone_force_lock; - ozone_introspect.force_unlock = (void *)ozone_force_unlock; - ozone_introspect.statistics = NULL; -#if (JEMALLOC_ZONE_VERSION >= 6) - ozone_introspect.zone_locked = NULL; -#endif + /* + * Unregister and reregister the default zone. On OSX >= 10.6, + * unregistering takes the last registered zone and places it at the + * location of the specified zone. Unregistering the default zone thus + * makes the last registered one the default. On OSX < 10.6, + * unregistering shifts all registered zones. The first registered zone + * then becomes the default. + */ + do { + malloc_zone_t *default_zone = malloc_default_zone(); + malloc_zone_unregister(default_zone); + malloc_zone_register(default_zone); + } while (malloc_default_zone() != &zone); } diff --git a/test/aligned_alloc.c b/test/aligned_alloc.c new file mode 100644 index 00000000..5a9b0cae --- /dev/null +++ b/test/aligned_alloc.c @@ -0,0 +1,119 @@ +#define JEMALLOC_MANGLE +#include "jemalloc_test.h" + +#define CHUNK 0x400000 +/* #define MAXALIGN ((size_t)UINT64_C(0x80000000000)) */ +#define MAXALIGN ((size_t)0x2000000LU) +#define NITER 4 + +int +main(void) +{ + size_t alignment, size, total; + unsigned i; + void *p, *ps[NITER]; + + malloc_printf("Test begin\n"); + + /* Test error conditions. */ + alignment = 0; + set_errno(0); + p = aligned_alloc(alignment, 1); + if (p != NULL || get_errno() != EINVAL) { + malloc_printf( + "Expected error for invalid alignment %zu\n", alignment); + } + + for (alignment = sizeof(size_t); alignment < MAXALIGN; + alignment <<= 1) { + set_errno(0); + p = aligned_alloc(alignment + 1, 1); + if (p != NULL || get_errno() != EINVAL) { + malloc_printf( + "Expected error for invalid alignment %zu\n", + alignment + 1); + } + } + +#if LG_SIZEOF_PTR == 3 + alignment = UINT64_C(0x8000000000000000); + size = UINT64_C(0x8000000000000000); +#else + alignment = 0x80000000LU; + size = 0x80000000LU; +#endif + set_errno(0); + p = aligned_alloc(alignment, size); + if (p != NULL || get_errno() != ENOMEM) { + malloc_printf( + "Expected error for aligned_alloc(%zu, %zu)\n", + alignment, size); + } + +#if LG_SIZEOF_PTR == 3 + alignment = UINT64_C(0x4000000000000000); + size = UINT64_C(0x8400000000000001); +#else + alignment = 0x40000000LU; + size = 0x84000001LU; +#endif + set_errno(0); + p = aligned_alloc(alignment, size); + if (p != NULL || get_errno() != ENOMEM) { + malloc_printf( + "Expected error for aligned_alloc(%zu, %zu)\n", + alignment, size); + } + + alignment = 0x10LU; +#if LG_SIZEOF_PTR == 3 + size = UINT64_C(0xfffffffffffffff0); +#else + size = 0xfffffff0LU; +#endif + set_errno(0); + p = aligned_alloc(alignment, size); + if (p != NULL || get_errno() != ENOMEM) { + malloc_printf( + "Expected error for aligned_alloc(&p, %zu, %zu)\n", + alignment, size); + } + + for (i = 0; i < NITER; i++) + ps[i] = NULL; + + for (alignment = 8; + alignment <= MAXALIGN; + alignment <<= 1) { + total = 0; + malloc_printf("Alignment: %zu\n", alignment); + for (size = 1; + size < 3 * alignment && size < (1U << 31); + size += (alignment >> (LG_SIZEOF_PTR-1)) - 1) { + for (i = 0; i < NITER; i++) { + ps[i] = aligned_alloc(alignment, size); + if (ps[i] == NULL) { + char buf[BUFERROR_BUF]; + + buferror(buf, sizeof(buf)); + malloc_printf( + "Error for size %zu (%#zx): %s\n", + size, size, buf); + exit(1); + } + total += malloc_usable_size(ps[i]); + if (total >= (MAXALIGN << 1)) + break; + } + for (i = 0; i < NITER; i++) { + if (ps[i] != NULL) { + free(ps[i]); + ps[i] = NULL; + } + } + } + } + + malloc_printf("Test end\n"); + return (0); +} diff --git a/test/aligned_alloc.exp b/test/aligned_alloc.exp new file mode 100644 index 00000000..b5061c72 --- /dev/null +++ b/test/aligned_alloc.exp @@ -0,0 +1,25 @@ +Test begin +Alignment: 8 +Alignment: 16 +Alignment: 32 +Alignment: 64 +Alignment: 128 +Alignment: 256 +Alignment: 512 +Alignment: 1024 +Alignment: 2048 +Alignment: 4096 +Alignment: 8192 +Alignment: 16384 +Alignment: 32768 +Alignment: 65536 +Alignment: 131072 +Alignment: 262144 +Alignment: 524288 +Alignment: 1048576 +Alignment: 2097152 +Alignment: 4194304 +Alignment: 8388608 +Alignment: 16777216 +Alignment: 33554432 +Test end diff --git a/test/allocated.c b/test/allocated.c index b1e40e47..9884905d 100644 --- a/test/allocated.c +++ b/test/allocated.c @@ -1,17 +1,8 @@ -#include -#include -#include -#include -#include -#include -#include -#include - #define JEMALLOC_MANGLE #include "jemalloc_test.h" void * -thread_start(void *arg) +je_thread_start(void *arg) { int err; void *p; @@ -20,89 +11,85 @@ thread_start(void *arg) size_t sz, usize; sz = sizeof(a0); - if ((err = JEMALLOC_P(mallctl)("thread.allocated", &a0, &sz, NULL, - 0))) { + if ((err = mallctl("thread.allocated", &a0, &sz, NULL, 0))) { if (err == ENOENT) { #ifdef JEMALLOC_STATS assert(false); #endif - goto RETURN; + goto label_return; } - fprintf(stderr, "%s(): Error in mallctl(): %s\n", __func__, + malloc_printf("%s(): Error in mallctl(): %s\n", __func__, strerror(err)); exit(1); } sz = sizeof(ap0); - if ((err = JEMALLOC_P(mallctl)("thread.allocatedp", &ap0, &sz, NULL, - 0))) { + if ((err = mallctl("thread.allocatedp", &ap0, &sz, NULL, 0))) { if (err == ENOENT) { #ifdef JEMALLOC_STATS assert(false); #endif - goto RETURN; + goto label_return; } - fprintf(stderr, "%s(): Error in mallctl(): %s\n", __func__, + malloc_printf("%s(): Error in mallctl(): %s\n", __func__, strerror(err)); exit(1); } assert(*ap0 == a0); sz = sizeof(d0); - if ((err = JEMALLOC_P(mallctl)("thread.deallocated", &d0, &sz, NULL, - 0))) { + if ((err = mallctl("thread.deallocated", &d0, &sz, NULL, 0))) { if (err == ENOENT) { #ifdef JEMALLOC_STATS assert(false); #endif - goto RETURN; + goto label_return; } - fprintf(stderr, "%s(): Error in mallctl(): %s\n", __func__, + malloc_printf("%s(): Error in mallctl(): %s\n", __func__, strerror(err)); exit(1); } sz = sizeof(dp0); - if ((err = JEMALLOC_P(mallctl)("thread.deallocatedp", &dp0, &sz, NULL, - 0))) { + if ((err = mallctl("thread.deallocatedp", &dp0, &sz, NULL, 0))) { if (err == ENOENT) { #ifdef JEMALLOC_STATS assert(false); #endif - goto RETURN; + goto label_return; } - fprintf(stderr, "%s(): Error in mallctl(): %s\n", __func__, + malloc_printf("%s(): Error in mallctl(): %s\n", __func__, strerror(err)); exit(1); } assert(*dp0 == d0); - p = JEMALLOC_P(malloc)(1); + p = malloc(1); if (p == NULL) { - fprintf(stderr, "%s(): Error in malloc()\n", __func__); + malloc_printf("%s(): Error in malloc()\n", __func__); exit(1); } sz = sizeof(a1); - JEMALLOC_P(mallctl)("thread.allocated", &a1, &sz, NULL, 0); + mallctl("thread.allocated", &a1, &sz, NULL, 0); sz = sizeof(ap1); - JEMALLOC_P(mallctl)("thread.allocatedp", &ap1, &sz, NULL, 0); + mallctl("thread.allocatedp", &ap1, &sz, NULL, 0); assert(*ap1 == a1); assert(ap0 == ap1); - usize = JEMALLOC_P(malloc_usable_size)(p); + usize = malloc_usable_size(p); assert(a0 + usize <= a1); - JEMALLOC_P(free)(p); + free(p); sz = sizeof(d1); - JEMALLOC_P(mallctl)("thread.deallocated", &d1, &sz, NULL, 0); + mallctl("thread.deallocated", &d1, &sz, NULL, 0); sz = sizeof(dp1); - JEMALLOC_P(mallctl)("thread.deallocatedp", &dp1, &sz, NULL, 0); + mallctl("thread.deallocatedp", &dp1, &sz, NULL, 0); assert(*dp1 == d1); assert(dp0 == dp1); assert(d0 + usize <= d1); -RETURN: +label_return: return (NULL); } @@ -110,33 +97,22 @@ int main(void) { int ret = 0; - pthread_t thread; + je_thread_t thread; - fprintf(stderr, "Test begin\n"); + malloc_printf("Test begin\n"); - thread_start(NULL); + je_thread_start(NULL); - if (pthread_create(&thread, NULL, thread_start, NULL) - != 0) { - fprintf(stderr, "%s(): Error in pthread_create()\n", __func__); - ret = 1; - goto RETURN; - } - pthread_join(thread, (void *)&ret); + je_thread_create(&thread, je_thread_start, NULL); + je_thread_join(thread, (void *)&ret); - thread_start(NULL); + je_thread_start(NULL); - if (pthread_create(&thread, NULL, thread_start, NULL) - != 0) { - fprintf(stderr, "%s(): Error in pthread_create()\n", __func__); - ret = 1; - goto RETURN; - } - pthread_join(thread, (void *)&ret); + je_thread_create(&thread, je_thread_start, NULL); + je_thread_join(thread, (void *)&ret); - thread_start(NULL); + je_thread_start(NULL); -RETURN: - fprintf(stderr, "Test end\n"); + malloc_printf("Test end\n"); return (ret); } diff --git a/test/allocm.c b/test/allocm.c index 59d0002e..80be673b 100644 --- a/test/allocm.c +++ b/test/allocm.c @@ -1,13 +1,9 @@ -#include -#include -#include - #define JEMALLOC_MANGLE #include "jemalloc_test.h" #define CHUNK 0x400000 -/* #define MAXALIGN ((size_t)0x80000000000LLU) */ -#define MAXALIGN ((size_t)0x2000000LLU) +/* #define MAXALIGN ((size_t)UINT64_C(0x80000000000)) */ +#define MAXALIGN ((size_t)0x2000000LU) #define NITER 4 int @@ -15,79 +11,122 @@ main(void) { int r; void *p; - size_t sz, alignment, total, tsz; + size_t nsz, rsz, sz, alignment, total; unsigned i; void *ps[NITER]; - fprintf(stderr, "Test begin\n"); + malloc_printf("Test begin\n"); - sz = 0; - r = JEMALLOC_P(allocm)(&p, &sz, 42, 0); + sz = 42; + nsz = 0; + r = nallocm(&nsz, sz, 0); if (r != ALLOCM_SUCCESS) { - fprintf(stderr, "Unexpected allocm() error\n"); + malloc_printf("Unexpected nallocm() error\n"); abort(); } - if (sz < 42) - fprintf(stderr, "Real size smaller than expected\n"); - if (JEMALLOC_P(dallocm)(p, 0) != ALLOCM_SUCCESS) - fprintf(stderr, "Unexpected dallocm() error\n"); - - r = JEMALLOC_P(allocm)(&p, NULL, 42, 0); + rsz = 0; + r = allocm(&p, &rsz, sz, 0); if (r != ALLOCM_SUCCESS) { - fprintf(stderr, "Unexpected allocm() error\n"); + malloc_printf("Unexpected allocm() error\n"); abort(); } - if (JEMALLOC_P(dallocm)(p, 0) != ALLOCM_SUCCESS) - fprintf(stderr, "Unexpected dallocm() error\n"); + if (rsz < sz) + malloc_printf("Real size smaller than expected\n"); + if (nsz != rsz) + malloc_printf("nallocm()/allocm() rsize mismatch\n"); + if (dallocm(p, 0) != ALLOCM_SUCCESS) + malloc_printf("Unexpected dallocm() error\n"); - r = JEMALLOC_P(allocm)(&p, NULL, 42, ALLOCM_ZERO); + r = allocm(&p, NULL, sz, 0); if (r != ALLOCM_SUCCESS) { - fprintf(stderr, "Unexpected allocm() error\n"); + malloc_printf("Unexpected allocm() error\n"); abort(); } - if (JEMALLOC_P(dallocm)(p, 0) != ALLOCM_SUCCESS) - fprintf(stderr, "Unexpected dallocm() error\n"); + if (dallocm(p, 0) != ALLOCM_SUCCESS) + malloc_printf("Unexpected dallocm() error\n"); + + nsz = 0; + r = nallocm(&nsz, sz, ALLOCM_ZERO); + if (r != ALLOCM_SUCCESS) { + malloc_printf("Unexpected nallocm() error\n"); + abort(); + } + rsz = 0; + r = allocm(&p, &rsz, sz, ALLOCM_ZERO); + if (r != ALLOCM_SUCCESS) { + malloc_printf("Unexpected allocm() error\n"); + abort(); + } + if (nsz != rsz) + malloc_printf("nallocm()/allocm() rsize mismatch\n"); + if (dallocm(p, 0) != ALLOCM_SUCCESS) + malloc_printf("Unexpected dallocm() error\n"); #if LG_SIZEOF_PTR == 3 - alignment = 0x8000000000000000LLU; - sz = 0x8000000000000000LLU; + alignment = UINT64_C(0x8000000000000000); + sz = UINT64_C(0x8000000000000000); #else alignment = 0x80000000LU; sz = 0x80000000LU; #endif - r = JEMALLOC_P(allocm)(&p, NULL, sz, ALLOCM_ALIGN(alignment)); + nsz = 0; + r = nallocm(&nsz, sz, ALLOCM_ALIGN(alignment)); if (r == ALLOCM_SUCCESS) { - fprintf(stderr, - "Expected error for allocm(&p, %zu, 0x%x)\n", + malloc_printf( + "Expected error for nallocm(&nsz, %zu, %#x)\n", sz, ALLOCM_ALIGN(alignment)); } + rsz = 0; + r = allocm(&p, &rsz, sz, ALLOCM_ALIGN(alignment)); + if (r == ALLOCM_SUCCESS) { + malloc_printf( + "Expected error for allocm(&p, %zu, %#x)\n", + sz, ALLOCM_ALIGN(alignment)); + } + if (nsz != rsz) + malloc_printf("nallocm()/allocm() rsize mismatch\n"); #if LG_SIZEOF_PTR == 3 - alignment = 0x4000000000000000LLU; - sz = 0x8400000000000001LLU; + alignment = UINT64_C(0x4000000000000000); + sz = UINT64_C(0x8400000000000001); #else alignment = 0x40000000LU; sz = 0x84000001LU; #endif - r = JEMALLOC_P(allocm)(&p, NULL, sz, ALLOCM_ALIGN(alignment)); + nsz = 0; + r = nallocm(&nsz, sz, ALLOCM_ALIGN(alignment)); + if (r != ALLOCM_SUCCESS) + malloc_printf("Unexpected nallocm() error\n"); + rsz = 0; + r = allocm(&p, &rsz, sz, ALLOCM_ALIGN(alignment)); if (r == ALLOCM_SUCCESS) { - fprintf(stderr, - "Expected error for allocm(&p, %zu, 0x%x)\n", + malloc_printf( + "Expected error for allocm(&p, %zu, %#x)\n", sz, ALLOCM_ALIGN(alignment)); } - alignment = 0x10LLU; + alignment = 0x10LU; #if LG_SIZEOF_PTR == 3 - sz = 0xfffffffffffffff0LLU; + sz = UINT64_C(0xfffffffffffffff0); #else - sz = 0xfffffff0LU; + sz = 0xfffffff0LU; #endif - r = JEMALLOC_P(allocm)(&p, NULL, sz, ALLOCM_ALIGN(alignment)); + nsz = 0; + r = nallocm(&nsz, sz, ALLOCM_ALIGN(alignment)); if (r == ALLOCM_SUCCESS) { - fprintf(stderr, - "Expected error for allocm(&p, %zu, 0x%x)\n", + malloc_printf( + "Expected error for nallocm(&nsz, %zu, %#x)\n", sz, ALLOCM_ALIGN(alignment)); } + rsz = 0; + r = allocm(&p, &rsz, sz, ALLOCM_ALIGN(alignment)); + if (r == ALLOCM_SUCCESS) { + malloc_printf( + "Expected error for allocm(&p, %zu, %#x)\n", + sz, ALLOCM_ALIGN(alignment)); + } + if (nsz != rsz) + malloc_printf("nallocm()/allocm() rsize mismatch\n"); for (i = 0; i < NITER; i++) ps[i] = NULL; @@ -96,38 +135,60 @@ main(void) alignment <= MAXALIGN; alignment <<= 1) { total = 0; - fprintf(stderr, "Alignment: %zu\n", alignment); + malloc_printf("Alignment: %zu\n", alignment); for (sz = 1; sz < 3 * alignment && sz < (1U << 31); sz += (alignment >> (LG_SIZEOF_PTR-1)) - 1) { for (i = 0; i < NITER; i++) { - r = JEMALLOC_P(allocm)(&ps[i], NULL, sz, + nsz = 0; + r = nallocm(&nsz, sz, ALLOCM_ALIGN(alignment) | ALLOCM_ZERO); if (r != ALLOCM_SUCCESS) { - fprintf(stderr, - "Error for size %zu (0x%zx): %d\n", + malloc_printf( + "nallocm() error for size %zu" + " (%#zx): %d\n", sz, sz, r); exit(1); } + rsz = 0; + r = allocm(&ps[i], &rsz, sz, + ALLOCM_ALIGN(alignment) | ALLOCM_ZERO); + if (r != ALLOCM_SUCCESS) { + malloc_printf( + "allocm() error for size %zu" + " (%#zx): %d\n", + sz, sz, r); + exit(1); + } + if (rsz < sz) { + malloc_printf( + "Real size smaller than" + " expected\n"); + } + if (nsz != rsz) { + malloc_printf( + "nallocm()/allocm() rsize" + " mismatch\n"); + } if ((uintptr_t)p & (alignment-1)) { - fprintf(stderr, + malloc_printf( "%p inadequately aligned for" " alignment: %zu\n", p, alignment); } - JEMALLOC_P(sallocm)(ps[i], &tsz, 0); - total += tsz; + sallocm(ps[i], &rsz, 0); + total += rsz; if (total >= (MAXALIGN << 1)) break; } for (i = 0; i < NITER; i++) { if (ps[i] != NULL) { - JEMALLOC_P(dallocm)(ps[i], 0); + dallocm(ps[i], 0); ps[i] = NULL; } } } } - fprintf(stderr, "Test end\n"); + malloc_printf("Test end\n"); return (0); } diff --git a/test/bitmap.c b/test/bitmap.c index adfaacfe..b2cb6300 100644 --- a/test/bitmap.c +++ b/test/bitmap.c @@ -1,18 +1,6 @@ #define JEMALLOC_MANGLE #include "jemalloc_test.h" -/* - * Avoid using the assert() from jemalloc_internal.h, since it requires - * internal libjemalloc functionality. - * */ -#include - -/* - * Directly include the bitmap code, since it isn't exposed outside - * libjemalloc. - */ -#include "../src/bitmap.c" - #if (LG_BITMAP_MAXBITS > 12) # define MAXBITS 4500 #else @@ -42,11 +30,13 @@ test_bitmap_init(void) bitmap_info_init(&binfo, i); { size_t j; - bitmap_t bitmap[bitmap_info_ngroups(&binfo)]; + bitmap_t *bitmap = malloc(sizeof(bitmap_t) * + bitmap_info_ngroups(&binfo)); bitmap_init(bitmap, &binfo); for (j = 0; j < i; j++) assert(bitmap_get(bitmap, &binfo, j) == false); + free(bitmap); } } @@ -62,12 +52,14 @@ test_bitmap_set(void) bitmap_info_init(&binfo, i); { size_t j; - bitmap_t bitmap[bitmap_info_ngroups(&binfo)]; + bitmap_t *bitmap = malloc(sizeof(bitmap_t) * + bitmap_info_ngroups(&binfo)); bitmap_init(bitmap, &binfo); for (j = 0; j < i; j++) bitmap_set(bitmap, &binfo, j); assert(bitmap_full(bitmap, &binfo)); + free(bitmap); } } } @@ -82,7 +74,8 @@ test_bitmap_unset(void) bitmap_info_init(&binfo, i); { size_t j; - bitmap_t bitmap[bitmap_info_ngroups(&binfo)]; + bitmap_t *bitmap = malloc(sizeof(bitmap_t) * + bitmap_info_ngroups(&binfo)); bitmap_init(bitmap, &binfo); for (j = 0; j < i; j++) @@ -93,6 +86,7 @@ test_bitmap_unset(void) for (j = 0; j < i; j++) bitmap_set(bitmap, &binfo, j); assert(bitmap_full(bitmap, &binfo)); + free(bitmap); } } } @@ -107,7 +101,8 @@ test_bitmap_sfu(void) bitmap_info_init(&binfo, i); { ssize_t j; - bitmap_t bitmap[bitmap_info_ngroups(&binfo)]; + bitmap_t *bitmap = malloc(sizeof(bitmap_t) * + bitmap_info_ngroups(&binfo)); bitmap_init(bitmap, &binfo); /* Iteratively set bits starting at the beginning. */ @@ -137,6 +132,7 @@ test_bitmap_sfu(void) } assert(bitmap_sfu(bitmap, &binfo) == i - 1); assert(bitmap_full(bitmap, &binfo)); + free(bitmap); } } } @@ -144,7 +140,7 @@ test_bitmap_sfu(void) int main(void) { - fprintf(stderr, "Test begin\n"); + malloc_printf("Test begin\n"); test_bitmap_size(); test_bitmap_init(); @@ -152,6 +148,6 @@ main(void) test_bitmap_unset(); test_bitmap_sfu(); - fprintf(stderr, "Test end\n"); + malloc_printf("Test end\n"); return (0); } diff --git a/test/jemalloc_test.h.in b/test/jemalloc_test.h.in index 0c48895e..e38b48ef 100644 --- a/test/jemalloc_test.h.in +++ b/test/jemalloc_test.h.in @@ -4,3 +4,50 @@ * have a different name. */ #include "jemalloc/jemalloc@install_suffix@.h" +#include "jemalloc/internal/jemalloc_internal.h" + +/* Abstraction layer for threading in tests */ +#ifdef _WIN32 +#include + +typedef HANDLE je_thread_t; + +void +je_thread_create(je_thread_t *thread, void *(*proc)(void *), void *arg) +{ + LPTHREAD_START_ROUTINE routine = (LPTHREAD_START_ROUTINE)proc; + *thread = CreateThread(NULL, 0, routine, arg, 0, NULL); + if (*thread == NULL) { + malloc_printf("Error in CreateThread()\n"); + exit(1); + } +} + +void +je_thread_join(je_thread_t thread, void **ret) +{ + WaitForSingleObject(thread, INFINITE); +} + +#else +#include + +typedef pthread_t je_thread_t; + +void +je_thread_create(je_thread_t *thread, void *(*proc)(void *), void *arg) +{ + + if (pthread_create(thread, NULL, proc, arg) != 0) { + malloc_printf("Error in pthread_create()\n"); + exit(1); + } +} + +void +je_thread_join(je_thread_t thread, void **ret) +{ + + pthread_join(thread, ret); +} +#endif diff --git a/test/mremap.c b/test/mremap.c index 146c66f4..47efa7c4 100644 --- a/test/mremap.c +++ b/test/mremap.c @@ -1,9 +1,3 @@ -#include -#include -#include -#include -#include - #define JEMALLOC_MANGLE #include "jemalloc_test.h" @@ -14,33 +8,32 @@ main(void) size_t sz, lg_chunk, chunksize, i; char *p, *q; - fprintf(stderr, "Test begin\n"); + malloc_printf("Test begin\n"); sz = sizeof(lg_chunk); - if ((err = JEMALLOC_P(mallctl)("opt.lg_chunk", &lg_chunk, &sz, NULL, - 0))) { + if ((err = mallctl("opt.lg_chunk", &lg_chunk, &sz, NULL, 0))) { assert(err != ENOENT); - fprintf(stderr, "%s(): Error in mallctl(): %s\n", __func__, + malloc_printf("%s(): Error in mallctl(): %s\n", __func__, strerror(err)); ret = 1; - goto RETURN; + goto label_return; } chunksize = ((size_t)1U) << lg_chunk; p = (char *)malloc(chunksize); if (p == NULL) { - fprintf(stderr, "malloc(%zu) --> %p\n", chunksize, p); + malloc_printf("malloc(%zu) --> %p\n", chunksize, p); ret = 1; - goto RETURN; + goto label_return; } memset(p, 'a', chunksize); q = (char *)realloc(p, chunksize * 2); if (q == NULL) { - fprintf(stderr, "realloc(%p, %zu) --> %p\n", p, chunksize * 2, + malloc_printf("realloc(%p, %zu) --> %p\n", p, chunksize * 2, q); ret = 1; - goto RETURN; + goto label_return; } for (i = 0; i < chunksize; i++) { assert(q[i] == 'a'); @@ -50,9 +43,9 @@ main(void) q = (char *)realloc(p, chunksize); if (q == NULL) { - fprintf(stderr, "realloc(%p, %zu) --> %p\n", p, chunksize, q); + malloc_printf("realloc(%p, %zu) --> %p\n", p, chunksize, q); ret = 1; - goto RETURN; + goto label_return; } for (i = 0; i < chunksize; i++) { assert(q[i] == 'a'); @@ -61,7 +54,7 @@ main(void) free(q); ret = 0; -RETURN: - fprintf(stderr, "Test end\n"); +label_return: + malloc_printf("Test end\n"); return (ret); } diff --git a/test/posix_memalign.c b/test/posix_memalign.c index 3e306c01..2185bcf7 100644 --- a/test/posix_memalign.c +++ b/test/posix_memalign.c @@ -1,15 +1,9 @@ -#include -#include -#include -#include -#include - #define JEMALLOC_MANGLE #include "jemalloc_test.h" #define CHUNK 0x400000 -/* #define MAXALIGN ((size_t)0x80000000000LLU) */ -#define MAXALIGN ((size_t)0x2000000LLU) +/* #define MAXALIGN ((size_t)UINT64_C(0x80000000000)) */ +#define MAXALIGN ((size_t)0x2000000LU) #define NITER 4 int @@ -20,13 +14,13 @@ main(void) int err; void *p, *ps[NITER]; - fprintf(stderr, "Test begin\n"); + malloc_printf("Test begin\n"); /* Test error conditions. */ for (alignment = 0; alignment < sizeof(void *); alignment++) { - err = JEMALLOC_P(posix_memalign)(&p, alignment, 1); + err = posix_memalign(&p, alignment, 1); if (err != EINVAL) { - fprintf(stderr, + malloc_printf( "Expected error for invalid alignment %zu\n", alignment); } @@ -34,51 +28,51 @@ main(void) for (alignment = sizeof(size_t); alignment < MAXALIGN; alignment <<= 1) { - err = JEMALLOC_P(posix_memalign)(&p, alignment + 1, 1); + err = posix_memalign(&p, alignment + 1, 1); if (err == 0) { - fprintf(stderr, + malloc_printf( "Expected error for invalid alignment %zu\n", alignment + 1); } } #if LG_SIZEOF_PTR == 3 - alignment = 0x8000000000000000LLU; - size = 0x8000000000000000LLU; + alignment = UINT64_C(0x8000000000000000); + size = UINT64_C(0x8000000000000000); #else alignment = 0x80000000LU; size = 0x80000000LU; #endif - err = JEMALLOC_P(posix_memalign)(&p, alignment, size); + err = posix_memalign(&p, alignment, size); if (err == 0) { - fprintf(stderr, + malloc_printf( "Expected error for posix_memalign(&p, %zu, %zu)\n", alignment, size); } #if LG_SIZEOF_PTR == 3 - alignment = 0x4000000000000000LLU; - size = 0x8400000000000001LLU; + alignment = UINT64_C(0x4000000000000000); + size = UINT64_C(0x8400000000000001); #else alignment = 0x40000000LU; size = 0x84000001LU; #endif - err = JEMALLOC_P(posix_memalign)(&p, alignment, size); + err = posix_memalign(&p, alignment, size); if (err == 0) { - fprintf(stderr, + malloc_printf( "Expected error for posix_memalign(&p, %zu, %zu)\n", alignment, size); } - alignment = 0x10LLU; + alignment = 0x10LU; #if LG_SIZEOF_PTR == 3 - size = 0xfffffffffffffff0LLU; + size = UINT64_C(0xfffffffffffffff0); #else size = 0xfffffff0LU; #endif - err = JEMALLOC_P(posix_memalign)(&p, alignment, size); + err = posix_memalign(&p, alignment, size); if (err == 0) { - fprintf(stderr, + malloc_printf( "Expected error for posix_memalign(&p, %zu, %zu)\n", alignment, size); } @@ -90,32 +84,32 @@ main(void) alignment <= MAXALIGN; alignment <<= 1) { total = 0; - fprintf(stderr, "Alignment: %zu\n", alignment); + malloc_printf("Alignment: %zu\n", alignment); for (size = 1; size < 3 * alignment && size < (1U << 31); size += (alignment >> (LG_SIZEOF_PTR-1)) - 1) { for (i = 0; i < NITER; i++) { - err = JEMALLOC_P(posix_memalign)(&ps[i], + err = posix_memalign(&ps[i], alignment, size); if (err) { - fprintf(stderr, - "Error for size %zu (0x%zx): %s\n", + malloc_printf( + "Error for size %zu (%#zx): %s\n", size, size, strerror(err)); exit(1); } - total += JEMALLOC_P(malloc_usable_size)(ps[i]); + total += malloc_usable_size(ps[i]); if (total >= (MAXALIGN << 1)) break; } for (i = 0; i < NITER; i++) { if (ps[i] != NULL) { - JEMALLOC_P(free)(ps[i]); + free(ps[i]); ps[i] = NULL; } } } } - fprintf(stderr, "Test end\n"); + malloc_printf("Test end\n"); return (0); } diff --git a/test/rallocm.c b/test/rallocm.c index ccf326bb..c5dedf48 100644 --- a/test/rallocm.c +++ b/test/rallocm.c @@ -1,9 +1,3 @@ -#include -#include -#include -#include -#include - #define JEMALLOC_MANGLE #include "jemalloc_test.h" @@ -15,113 +9,119 @@ main(void) size_t sz, tsz; int r; - fprintf(stderr, "Test begin\n"); + malloc_printf("Test begin\n"); /* Get page size. */ { +#ifdef _WIN32 + SYSTEM_INFO si; + GetSystemInfo(&si); + pagesize = (size_t)si.dwPageSize; +#else long result = sysconf(_SC_PAGESIZE); assert(result != -1); pagesize = (size_t)result; +#endif } - r = JEMALLOC_P(allocm)(&p, &sz, 42, 0); + r = allocm(&p, &sz, 42, 0); if (r != ALLOCM_SUCCESS) { - fprintf(stderr, "Unexpected allocm() error\n"); + malloc_printf("Unexpected allocm() error\n"); abort(); } q = p; - r = JEMALLOC_P(rallocm)(&q, &tsz, sz, 0, ALLOCM_NO_MOVE); + r = rallocm(&q, &tsz, sz, 0, ALLOCM_NO_MOVE); if (r != ALLOCM_SUCCESS) - fprintf(stderr, "Unexpected rallocm() error\n"); + malloc_printf("Unexpected rallocm() error\n"); if (q != p) - fprintf(stderr, "Unexpected object move\n"); + malloc_printf("Unexpected object move\n"); if (tsz != sz) { - fprintf(stderr, "Unexpected size change: %zu --> %zu\n", + malloc_printf("Unexpected size change: %zu --> %zu\n", sz, tsz); } q = p; - r = JEMALLOC_P(rallocm)(&q, &tsz, sz, 5, ALLOCM_NO_MOVE); + r = rallocm(&q, &tsz, sz, 5, ALLOCM_NO_MOVE); if (r != ALLOCM_SUCCESS) - fprintf(stderr, "Unexpected rallocm() error\n"); + malloc_printf("Unexpected rallocm() error\n"); if (q != p) - fprintf(stderr, "Unexpected object move\n"); + malloc_printf("Unexpected object move\n"); if (tsz != sz) { - fprintf(stderr, "Unexpected size change: %zu --> %zu\n", + malloc_printf("Unexpected size change: %zu --> %zu\n", sz, tsz); } q = p; - r = JEMALLOC_P(rallocm)(&q, &tsz, sz + 5, 0, ALLOCM_NO_MOVE); + r = rallocm(&q, &tsz, sz + 5, 0, ALLOCM_NO_MOVE); if (r != ALLOCM_ERR_NOT_MOVED) - fprintf(stderr, "Unexpected rallocm() result\n"); + malloc_printf("Unexpected rallocm() result\n"); if (q != p) - fprintf(stderr, "Unexpected object move\n"); + malloc_printf("Unexpected object move\n"); if (tsz != sz) { - fprintf(stderr, "Unexpected size change: %zu --> %zu\n", + malloc_printf("Unexpected size change: %zu --> %zu\n", sz, tsz); } q = p; - r = JEMALLOC_P(rallocm)(&q, &tsz, sz + 5, 0, 0); + r = rallocm(&q, &tsz, sz + 5, 0, 0); if (r != ALLOCM_SUCCESS) - fprintf(stderr, "Unexpected rallocm() error\n"); + malloc_printf("Unexpected rallocm() error\n"); if (q == p) - fprintf(stderr, "Expected object move\n"); + malloc_printf("Expected object move\n"); if (tsz == sz) { - fprintf(stderr, "Expected size change: %zu --> %zu\n", + malloc_printf("Expected size change: %zu --> %zu\n", sz, tsz); } p = q; sz = tsz; - r = JEMALLOC_P(rallocm)(&q, &tsz, pagesize*2, 0, 0); + r = rallocm(&q, &tsz, pagesize*2, 0, 0); if (r != ALLOCM_SUCCESS) - fprintf(stderr, "Unexpected rallocm() error\n"); + malloc_printf("Unexpected rallocm() error\n"); if (q == p) - fprintf(stderr, "Expected object move\n"); + malloc_printf("Expected object move\n"); if (tsz == sz) { - fprintf(stderr, "Expected size change: %zu --> %zu\n", + malloc_printf("Expected size change: %zu --> %zu\n", sz, tsz); } p = q; sz = tsz; - r = JEMALLOC_P(rallocm)(&q, &tsz, pagesize*4, 0, 0); + r = rallocm(&q, &tsz, pagesize*4, 0, 0); if (r != ALLOCM_SUCCESS) - fprintf(stderr, "Unexpected rallocm() error\n"); + malloc_printf("Unexpected rallocm() error\n"); if (tsz == sz) { - fprintf(stderr, "Expected size change: %zu --> %zu\n", + malloc_printf("Expected size change: %zu --> %zu\n", sz, tsz); } p = q; sz = tsz; - r = JEMALLOC_P(rallocm)(&q, &tsz, pagesize*2, 0, ALLOCM_NO_MOVE); + r = rallocm(&q, &tsz, pagesize*2, 0, ALLOCM_NO_MOVE); if (r != ALLOCM_SUCCESS) - fprintf(stderr, "Unexpected rallocm() error\n"); + malloc_printf("Unexpected rallocm() error\n"); if (q != p) - fprintf(stderr, "Unexpected object move\n"); + malloc_printf("Unexpected object move\n"); if (tsz == sz) { - fprintf(stderr, "Expected size change: %zu --> %zu\n", + malloc_printf("Expected size change: %zu --> %zu\n", sz, tsz); } sz = tsz; - r = JEMALLOC_P(rallocm)(&q, &tsz, pagesize*4, 0, ALLOCM_NO_MOVE); + r = rallocm(&q, &tsz, pagesize*4, 0, ALLOCM_NO_MOVE); if (r != ALLOCM_SUCCESS) - fprintf(stderr, "Unexpected rallocm() error\n"); + malloc_printf("Unexpected rallocm() error\n"); if (q != p) - fprintf(stderr, "Unexpected object move\n"); + malloc_printf("Unexpected object move\n"); if (tsz == sz) { - fprintf(stderr, "Expected size change: %zu --> %zu\n", + malloc_printf("Expected size change: %zu --> %zu\n", sz, tsz); } sz = tsz; - JEMALLOC_P(dallocm)(p, 0); + dallocm(p, 0); - fprintf(stderr, "Test end\n"); + malloc_printf("Test end\n"); return (0); } diff --git a/test/thread_arena.c b/test/thread_arena.c index ef8d6817..2020d994 100644 --- a/test/thread_arena.c +++ b/test/thread_arena.c @@ -1,16 +1,10 @@ -#include -#include -#include -#include -#include - #define JEMALLOC_MANGLE #include "jemalloc_test.h" #define NTHREADS 10 void * -thread_start(void *arg) +je_thread_start(void *arg) { unsigned main_arena_ind = *(unsigned *)arg; void *p; @@ -18,24 +12,24 @@ thread_start(void *arg) size_t size; int err; - p = JEMALLOC_P(malloc)(1); + p = malloc(1); if (p == NULL) { - fprintf(stderr, "%s(): Error in malloc()\n", __func__); + malloc_printf("%s(): Error in malloc()\n", __func__); return (void *)1; } size = sizeof(arena_ind); - if ((err = JEMALLOC_P(mallctl)("thread.arena", &arena_ind, &size, - &main_arena_ind, sizeof(main_arena_ind)))) { - fprintf(stderr, "%s(): Error in mallctl(): %s\n", __func__, + if ((err = mallctl("thread.arena", &arena_ind, &size, &main_arena_ind, + sizeof(main_arena_ind)))) { + malloc_printf("%s(): Error in mallctl(): %s\n", __func__, strerror(err)); return (void *)1; } size = sizeof(arena_ind); - if ((err = JEMALLOC_P(mallctl)("thread.arena", &arena_ind, &size, NULL, + if ((err = mallctl("thread.arena", &arena_ind, &size, NULL, 0))) { - fprintf(stderr, "%s(): Error in mallctl(): %s\n", __func__, + malloc_printf("%s(): Error in mallctl(): %s\n", __func__, strerror(err)); return (void *)1; } @@ -52,41 +46,33 @@ main(void) unsigned arena_ind; size_t size; int err; - pthread_t threads[NTHREADS]; + je_thread_t threads[NTHREADS]; unsigned i; - fprintf(stderr, "Test begin\n"); + malloc_printf("Test begin\n"); - p = JEMALLOC_P(malloc)(1); + p = malloc(1); if (p == NULL) { - fprintf(stderr, "%s(): Error in malloc()\n", __func__); + malloc_printf("%s(): Error in malloc()\n", __func__); ret = 1; - goto RETURN; + goto label_return; } size = sizeof(arena_ind); - if ((err = JEMALLOC_P(mallctl)("thread.arena", &arena_ind, &size, NULL, - 0))) { - fprintf(stderr, "%s(): Error in mallctl(): %s\n", __func__, + if ((err = mallctl("thread.arena", &arena_ind, &size, NULL, 0))) { + malloc_printf("%s(): Error in mallctl(): %s\n", __func__, strerror(err)); ret = 1; - goto RETURN; - } - - for (i = 0; i < NTHREADS; i++) { - if (pthread_create(&threads[i], NULL, thread_start, - (void *)&arena_ind) != 0) { - fprintf(stderr, "%s(): Error in pthread_create()\n", - __func__); - ret = 1; - goto RETURN; - } + goto label_return; } for (i = 0; i < NTHREADS; i++) - pthread_join(threads[i], (void *)&ret); + je_thread_create(&threads[i], je_thread_start, (void *)&arena_ind); -RETURN: - fprintf(stderr, "Test end\n"); + for (i = 0; i < NTHREADS; i++) + je_thread_join(threads[i], (void *)&ret); + +label_return: + malloc_printf("Test end\n"); return (ret); } diff --git a/test/thread_tcache_enabled.c b/test/thread_tcache_enabled.c new file mode 100644 index 00000000..2061b7bb --- /dev/null +++ b/test/thread_tcache_enabled.c @@ -0,0 +1,91 @@ +#define JEMALLOC_MANGLE +#include "jemalloc_test.h" + +void * +je_thread_start(void *arg) +{ + int err; + size_t sz; + bool e0, e1; + + sz = sizeof(bool); + if ((err = mallctl("thread.tcache.enabled", &e0, &sz, NULL, 0))) { + if (err == ENOENT) { +#ifdef JEMALLOC_TCACHE + assert(false); +#endif + } + goto label_return; + } + + if (e0) { + e1 = false; + assert(mallctl("thread.tcache.enabled", &e0, &sz, &e1, sz) + == 0); + assert(e0); + } + + e1 = true; + assert(mallctl("thread.tcache.enabled", &e0, &sz, &e1, sz) == 0); + assert(e0 == false); + + e1 = true; + assert(mallctl("thread.tcache.enabled", &e0, &sz, &e1, sz) == 0); + assert(e0); + + e1 = false; + assert(mallctl("thread.tcache.enabled", &e0, &sz, &e1, sz) == 0); + assert(e0); + + e1 = false; + assert(mallctl("thread.tcache.enabled", &e0, &sz, &e1, sz) == 0); + assert(e0 == false); + + free(malloc(1)); + e1 = true; + assert(mallctl("thread.tcache.enabled", &e0, &sz, &e1, sz) == 0); + assert(e0 == false); + + free(malloc(1)); + e1 = true; + assert(mallctl("thread.tcache.enabled", &e0, &sz, &e1, sz) == 0); + assert(e0); + + free(malloc(1)); + e1 = false; + assert(mallctl("thread.tcache.enabled", &e0, &sz, &e1, sz) == 0); + assert(e0); + + free(malloc(1)); + e1 = false; + assert(mallctl("thread.tcache.enabled", &e0, &sz, &e1, sz) == 0); + assert(e0 == false); + + free(malloc(1)); +label_return: + return (NULL); +} + +int +main(void) +{ + int ret = 0; + je_thread_t thread; + + malloc_printf("Test begin\n"); + + je_thread_start(NULL); + + je_thread_create(&thread, je_thread_start, NULL); + je_thread_join(thread, (void *)&ret); + + je_thread_start(NULL); + + je_thread_create(&thread, je_thread_start, NULL); + je_thread_join(thread, (void *)&ret); + + je_thread_start(NULL); + + malloc_printf("Test end\n"); + return (ret); +} diff --git a/test/thread_tcache_enabled.exp b/test/thread_tcache_enabled.exp new file mode 100644 index 00000000..369a88dd --- /dev/null +++ b/test/thread_tcache_enabled.exp @@ -0,0 +1,2 @@ +Test begin +Test end