project-base/server-skynet-source-3rd-jemalloc
2
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Move slabs out of chunks.

Browse Source
This commit is contained in:
Jason Evans 2016-05-29 18:34:50 -07:00
parent d28e5a6696
commit 498856f44a
21 changed files with 596 additions and 2332 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

98
doc/jemalloc.xml.in
View File

@ -509,26 +509,20 @@ for (i = 0; i < nbins; i++) {
<para>In addition to multiple arenas, unless <para>In addition to multiple arenas, unless
<option>--disable-tcache</option> is specified during configuration, this <option>--disable-tcache</option> is specified during configuration, this
allocator supports thread-specific caching for small and large objects, in allocator supports thread-specific caching, in order to make it possible to
order to make it possible to completely avoid synchronization for most completely avoid synchronization for most allocation requests. Such caching
allocation requests. Such caching allows very fast allocation in the allows very fast allocation in the common case, but it increases memory
common case, but it increases memory usage and fragmentation, since a usage and fragmentation, since a bounded number of objects can remain
bounded number of objects can remain allocated in each thread cache.</para> allocated in each thread cache.</para>
<para>Memory is conceptually broken into equal-sized chunks, where the chunk <para>Memory is conceptually broken into extents. Extents are always
size is a power of two that is greater than the page size. Chunks are aligned to multiples of the page size. This alignment makes it possible to
always aligned to multiples of the chunk size. This alignment makes it find metadata for user objects quickly. User objects are broken into two
possible to find metadata for user objects very quickly. User objects are categories according to size: small and large. Contiguous small objects
broken into three categories according to size: small, large, and huge. comprise a slab, which resides within a single extent, whereas large objects
Multiple small and large objects can reside within a single chunk, whereas each have their own extents backing them.</para>
huge objects each have one or more chunks backing them. Each chunk that
contains small and/or large objects tracks its contents as runs of
contiguous pages (unused, backing a set of small objects, or backing one
large object). The combination of chunk alignment and chunk page maps makes
it possible to determine all metadata regarding small and large allocations
in constant time.</para>
<para>Small objects are managed in groups by page runs. Each run maintains <para>Small objects are managed in groups by slabs. Each slab maintains
a bitmap to track which regions are in use. Allocation requests that are no a bitmap to track which regions are in use. Allocation requests that are no
more than half the quantum (8 or 16, depending on architecture) are rounded more than half the quantum (8 or 16, depending on architecture) are rounded
up to the nearest power of two that is at least <code up to the nearest power of two that is at least <code
@ -536,11 +530,9 @@ for (i = 0; i < nbins; i++) {
classes are multiples of the quantum, spaced such that there are four size classes are multiples of the quantum, spaced such that there are four size
classes for each doubling in size, which limits internal fragmentation to classes for each doubling in size, which limits internal fragmentation to
approximately 20% for all but the smallest size classes. Small size classes approximately 20% for all but the smallest size classes. Small size classes
are smaller than four times the page size, large size classes are smaller are smaller than four times the page size, and large size classes extend
than the chunk size (see the <link from four times the page size up to the largest size class that does not
linkend="opt.lg_chunk"><mallctl>opt.lg_chunk</mallctl></link> option), and exceed <constant>PTRDIFF_MAX</constant>.</para>
huge size classes extend from the chunk size up to the largest size class
that does not exceed <constant>PTRDIFF_MAX</constant>.</para>
<para>Allocations are packed tightly together, which can be an issue for <para>Allocations are packed tightly together, which can be an issue for
multi-threaded applications. If you need to assure that allocations do not multi-threaded applications. If you need to assure that allocations do not
@ -560,18 +552,16 @@ for (i = 0; i < nbins; i++) {
trivially succeeds in place as long as the pre-size and post-size both round trivially succeeds in place as long as the pre-size and post-size both round
up to the same size class. No other API guarantees are made regarding up to the same size class. No other API guarantees are made regarding
in-place resizing, but the current implementation also tries to resize large in-place resizing, but the current implementation also tries to resize large
and huge allocations in place, as long as the pre-size and post-size are allocations in place, as long as the pre-size and post-size are both large.
both large or both huge. In such cases shrinkage always succeeds for large For shrinkage to succeed, the extent allocator must support splitting (see
size classes, but for huge size classes the chunk allocator must support <link
splitting (see <link
linkend="arena.i.chunk_hooks"><mallctl>arena.&lt;i&gt;.chunk_hooks</mallctl></link>). linkend="arena.i.chunk_hooks"><mallctl>arena.&lt;i&gt;.chunk_hooks</mallctl></link>).
Growth only succeeds if the trailing memory is currently available, and Growth only succeeds if the trailing memory is currently available, and the
additionally for huge size classes the chunk allocator must support extent allocator supports merging.</para>
merging.</para>
<para>Assuming 2 MiB chunks, 4 KiB pages, and a 16-byte quantum on a <para>Assuming 4 KiB pages and a 16-byte quantum on a 64-bit system, the
64-bit system, the size classes in each category are as shown in <xref size classes in each category are as shown in <xref linkend="size_classes"
linkend="size_classes" xrefstyle="template:Table %n"/>.</para> xrefstyle="template:Table %n"/>.</para>
<table xml:id="size_classes" frame="all"> <table xml:id="size_classes" frame="all">
<title>Size classes</title> <title>Size classes</title>
@ -625,7 +615,7 @@ for (i = 0; i < nbins; i++) {
<entry>[10 KiB, 12 KiB, 14 KiB]</entry> <entry>[10 KiB, 12 KiB, 14 KiB]</entry>
</row> </row>
<row> <row>
<entry morerows="7">Large</entry> <entry morerows="15">Large</entry>
<entry>2 KiB</entry> <entry>2 KiB</entry>
<entry>[16 KiB]</entry> <entry>[16 KiB]</entry>
</row> </row>
@ -655,12 +645,7 @@ for (i = 0; i < nbins; i++) {
</row> </row>
<row> <row>
<entry>256 KiB</entry> <entry>256 KiB</entry>
<entry>[1280 KiB, 1536 KiB, 1792 KiB]</entry> <entry>[1280 KiB, 1536 KiB, 1792 KiB, 2 MiB]</entry>
</row>
<row>
<entry morerows="8">Huge</entry>
<entry>256 KiB</entry>
<entry>[2 MiB]</entry>
</row> </row>
<row> <row>
<entry>512 KiB</entry> <entry>512 KiB</entry>
@ -1875,16 +1860,16 @@ typedef struct {
(<type>uint32_t</type>) (<type>uint32_t</type>)
<literal>r-</literal> <literal>r-</literal>
</term> </term>
<listitem><para>Number of regions per page run.</para></listitem> <listitem><para>Number of regions per slab.</para></listitem>
</varlistentry> </varlistentry>
<varlistentry id="arenas.bin.i.run_size"> <varlistentry id="arenas.bin.i.slab_size">
<term> <term>
<mallctl>arenas.bin.&lt;i&gt;.run_size</mallctl> <mallctl>arenas.bin.&lt;i&gt;.slab_size</mallctl>
(<type>size_t</type>) (<type>size_t</type>)
<literal>r-</literal> <literal>r-</literal>
</term> </term>
<listitem><para>Number of bytes per page run.</para></listitem> <listitem><para>Number of bytes per slab.</para></listitem>
</varlistentry> </varlistentry>
<varlistentry id="arenas.nhchunks"> <varlistentry id="arenas.nhchunks">
@ -2185,7 +2170,7 @@ typedef struct {
(<type>size_t</type>) (<type>size_t</type>)
<literal>r-</literal> <literal>r-</literal>
</term> </term>
<listitem><para>Number of pages in active runs.</para></listitem> <listitem><para>Number of pages in active extents.</para></listitem>
</varlistentry> </varlistentry>
<varlistentry id="stats.arenas.i.pdirty"> <varlistentry id="stats.arenas.i.pdirty">
@ -2194,8 +2179,9 @@ typedef struct {
(<type>size_t</type>) (<type>size_t</type>)
<literal>r-</literal> <literal>r-</literal>
</term> </term>
<listitem><para>Number of pages within unused runs that are potentially <listitem><para>Number of pages within unused extents that are
dirty, and for which <function>madvise<parameter>...</parameter> potentially dirty, and for which
<function>madvise<parameter>...</parameter>
<parameter><constant>MADV_DONTNEED</constant></parameter></function> or <parameter><constant>MADV_DONTNEED</constant></parameter></function> or
similar has not been called.</para></listitem> similar has not been called.</para></listitem>
</varlistentry> </varlistentry>
@ -2483,35 +2469,35 @@ typedef struct {
<listitem><para>Cumulative number of tcache flushes.</para></listitem> <listitem><para>Cumulative number of tcache flushes.</para></listitem>
</varlistentry> </varlistentry>
<varlistentry id="stats.arenas.i.bins.j.nruns"> <varlistentry id="stats.arenas.i.bins.j.nslabs">
<term> <term>
<mallctl>stats.arenas.&lt;i&gt;.bins.&lt;j&gt;.nruns</mallctl> <mallctl>stats.arenas.&lt;i&gt;.bins.&lt;j&gt;.nslabs</mallctl>
(<type>uint64_t</type>) (<type>uint64_t</type>)
<literal>r-</literal> <literal>r-</literal>
[<option>--enable-stats</option>] [<option>--enable-stats</option>]
</term> </term>
<listitem><para>Cumulative number of runs created.</para></listitem> <listitem><para>Cumulative number of slabs created.</para></listitem>
</varlistentry> </varlistentry>
<varlistentry id="stats.arenas.i.bins.j.nreruns"> <varlistentry id="stats.arenas.i.bins.j.nreslabs">
<term> <term>
<mallctl>stats.arenas.&lt;i&gt;.bins.&lt;j&gt;.nreruns</mallctl> <mallctl>stats.arenas.&lt;i&gt;.bins.&lt;j&gt;.nreslabs</mallctl>
(<type>uint64_t</type>) (<type>uint64_t</type>)
<literal>r-</literal> <literal>r-</literal>
[<option>--enable-stats</option>] [<option>--enable-stats</option>]
</term> </term>
<listitem><para>Cumulative number of times the current run from which <listitem><para>Cumulative number of times the current slab from which
to allocate changed.</para></listitem> to allocate changed.</para></listitem>
</varlistentry> </varlistentry>
<varlistentry id="stats.arenas.i.bins.j.curruns"> <varlistentry id="stats.arenas.i.bins.j.curslabs">
<term> <term>
<mallctl>stats.arenas.&lt;i&gt;.bins.&lt;j&gt;.curruns</mallctl> <mallctl>stats.arenas.&lt;i&gt;.bins.&lt;j&gt;.curslabs</mallctl>
(<type>size_t</type>) (<type>size_t</type>)
<literal>r-</literal> <literal>r-</literal>
[<option>--enable-stats</option>] [<option>--enable-stats</option>]
</term> </term>
<listitem><para>Current number of runs.</para></listitem> <listitem><para>Current number of slabs.</para></listitem>
</varlistentry> </varlistentry>
<varlistentry id="stats.arenas.i.hchunks.j.nmalloc"> <varlistentry id="stats.arenas.i.hchunks.j.nmalloc">

793
include/jemalloc/internal/arena.h
View File

@ -3,9 +3,9 @@
#define LARGE_MINCLASS (ZU(1) << LG_LARGE_MINCLASS) #define LARGE_MINCLASS (ZU(1) << LG_LARGE_MINCLASS)
/* Maximum number of regions in one run. */ /* Maximum number of regions in one slab. */
#define LG_RUN_MAXREGS (LG_PAGE - LG_TINY_MIN) #define LG_SLAB_MAXREGS (LG_PAGE - LG_TINY_MIN)
#define RUN_MAXREGS (1U << LG_RUN_MAXREGS) #define SLAB_MAXREGS (1U << LG_SLAB_MAXREGS)
/* /*
* The minimum ratio of active:dirty pages per arena is computed as: * The minimum ratio of active:dirty pages per arena is computed as:
@ -29,12 +29,7 @@ typedef enum {
/* Number of event ticks between time checks. */ /* Number of event ticks between time checks. */
#define DECAY_NTICKS_PER_UPDATE 1000 #define DECAY_NTICKS_PER_UPDATE 1000
typedef struct arena_runs_dirty_link_s arena_runs_dirty_link_t; typedef struct arena_slab_data_s arena_slab_data_t;
typedef struct arena_avail_links_s arena_avail_links_t;
typedef struct arena_run_s arena_run_t;
typedef struct arena_chunk_map_bits_s arena_chunk_map_bits_t;
typedef struct arena_chunk_map_misc_s arena_chunk_map_misc_t;
typedef struct arena_chunk_s arena_chunk_t;
typedef struct arena_bin_info_s arena_bin_info_t; typedef struct arena_bin_info_s arena_bin_info_t;
typedef struct arena_bin_s arena_bin_t; typedef struct arena_bin_s arena_bin_t;
typedef struct arena_s arena_t; typedef struct arena_s arena_t;
@ -45,152 +40,25 @@ typedef struct arena_tdata_s arena_tdata_t;
#ifdef JEMALLOC_H_STRUCTS #ifdef JEMALLOC_H_STRUCTS
#ifdef JEMALLOC_ARENA_STRUCTS_A #ifdef JEMALLOC_ARENA_STRUCTS_A
struct arena_run_s { struct arena_slab_data_s {
/* Index of bin this run is associated with. */ /* Index of bin this slab is associated with. */
szind_t binind; szind_t binind;
/* Number of free regions in run. */ /* Number of free regions in slab. */
unsigned nfree; unsigned nfree;
/* Per region allocated/deallocated bitmap. */ /* Per region allocated/deallocated bitmap. */
bitmap_t bitmap[BITMAP_GROUPS_MAX]; bitmap_t bitmap[BITMAP_GROUPS_MAX];
}; };
/* Each element of the chunk map corresponds to one page within the chunk. */
struct arena_chunk_map_bits_s {
/*
* Run address (or size) and various flags are stored together. The bit
* layout looks like (assuming 32-bit system):
*
* ???????? ???????? ???nnnnn nnndumla
*
* ? : Unallocated: Run address for first/last pages, unset for internal
* pages.
* Small: Run page offset.
* Large: Run page count for first page, unset for trailing pages.
* n : binind for small size class, BININD_INVALID for large size class.
* d : dirty?
* u : unzeroed?
* m : decommitted?
* l : large?
* a : allocated?
*
* Following are example bit patterns for the three types of runs.
*
* p : run page offset
* s : run size
* n : binind for size class; large objects set these to BININD_INVALID
* x : don't care
* - : 0
* + : 1
* [DUMLA] : bit set
* [dumla] : bit unset
*
* Unallocated (clean):
* ssssssss ssssssss sss+++++ +++dum-a
* xxxxxxxx xxxxxxxx xxxxxxxx xxx-Uxxx
* ssssssss ssssssss sss+++++ +++dUm-a
*
* Unallocated (dirty):
* ssssssss ssssssss sss+++++ +++D-m-a
* xxxxxxxx xxxxxxxx xxxxxxxx xxxxxxxx
* ssssssss ssssssss sss+++++ +++D-m-a
*
* Small:
* pppppppp pppppppp pppnnnnn nnnd---A
* pppppppp pppppppp pppnnnnn nnn----A
* pppppppp pppppppp pppnnnnn nnnd---A
*
* Large:
* ssssssss ssssssss sss+++++ +++D--LA
* xxxxxxxx xxxxxxxx xxxxxxxx xxxxxxxx
* -------- -------- ---+++++ +++D--LA
*
* Large (sampled, size <= LARGE_MINCLASS):
* ssssssss ssssssss sssnnnnn nnnD--LA
* xxxxxxxx xxxxxxxx xxxxxxxx xxxxxxxx
* -------- -------- ---+++++ +++D--LA
*
* Large (not sampled, size == LARGE_MINCLASS):
* ssssssss ssssssss sss+++++ +++D--LA
* xxxxxxxx xxxxxxxx xxxxxxxx xxxxxxxx
* -------- -------- ---+++++ +++D--LA
*/
size_t bits;
#define CHUNK_MAP_ALLOCATED ((size_t)0x01U)
#define CHUNK_MAP_LARGE ((size_t)0x02U)
#define CHUNK_MAP_STATE_MASK ((size_t)0x3U)
#define CHUNK_MAP_DECOMMITTED ((size_t)0x04U)
#define CHUNK_MAP_UNZEROED ((size_t)0x08U)
#define CHUNK_MAP_DIRTY ((size_t)0x10U)
#define CHUNK_MAP_FLAGS_MASK ((size_t)0x1cU)
#define CHUNK_MAP_BININD_SHIFT 5
#define BININD_INVALID ((size_t)0xffU)
#define CHUNK_MAP_BININD_MASK (BININD_INVALID << CHUNK_MAP_BININD_SHIFT)
#define CHUNK_MAP_BININD_INVALID CHUNK_MAP_BININD_MASK
#define CHUNK_MAP_RUNIND_SHIFT (CHUNK_MAP_BININD_SHIFT + 8)
#define CHUNK_MAP_SIZE_SHIFT (CHUNK_MAP_RUNIND_SHIFT - LG_PAGE)
#define CHUNK_MAP_SIZE_MASK \
(~(CHUNK_MAP_BININD_MASK | CHUNK_MAP_FLAGS_MASK | CHUNK_MAP_STATE_MASK))
};
struct arena_runs_dirty_link_s {
qr(arena_runs_dirty_link_t) rd_link;
};
/*
* Each arena_chunk_map_misc_t corresponds to one page within the chunk, just
* like arena_chunk_map_bits_t. Two separate arrays are stored within each
* chunk header in order to improve cache locality.
*/
struct arena_chunk_map_misc_s {
/*
* Linkage for run heaps. There are two disjoint uses:
*
* 1) arena_t's runs_avail heaps.
* 2) arena_run_t conceptually uses this linkage for in-use non-full
* runs, rather than directly embedding linkage.
*/
phn(arena_chunk_map_misc_t) ph_link;
union {
/* Linkage for list of dirty runs. */
arena_runs_dirty_link_t rd;
/* Profile counters, used for large object runs. */
union {
void *prof_tctx_pun;
prof_tctx_t *prof_tctx;
};
/* Small region run metadata. */
arena_run_t run;
};
};
typedef ph(arena_chunk_map_misc_t) arena_run_heap_t;
#endif /* JEMALLOC_ARENA_STRUCTS_A */ #endif /* JEMALLOC_ARENA_STRUCTS_A */
#ifdef JEMALLOC_ARENA_STRUCTS_B #ifdef JEMALLOC_ARENA_STRUCTS_B
/* Arena chunk header. */
struct arena_chunk_s {
/*
* Map of pages within chunk that keeps track of free/large/small. The
* first map_bias entries are omitted, since the chunk header does not
* need to be tracked in the map. This omission saves a header page
* for common chunk sizes (e.g. 4 MiB).
*/
arena_chunk_map_bits_t map_bits[1]; /* Dynamically sized. */
};
/* /*
* Read-only information associated with each element of arena_t's bins array * 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 * is stored separately, partly to reduce memory usage (only one copy, rather
* than one per arena), but mainly to avoid false cacheline sharing. * than one per arena), but mainly to avoid false cacheline sharing.
* *
* Each run has the following layout: * Each slab has the following layout:
* *
* /--------------------\ * /--------------------\
* | region 0 | * | region 0 |
@ -205,45 +73,42 @@ struct arena_chunk_s {
* \--------------------/ * \--------------------/
*/ */
struct arena_bin_info_s { struct arena_bin_info_s {
/* Size of regions in a run for this bin's size class. */ /* Size of regions in a slab for this bin's size class. */
size_t reg_size; size_t reg_size;
/* Total size of a run for this bin's size class. */ /* Total size of a slab for this bin's size class. */
size_t run_size; size_t slab_size;
/* Total number of regions in a run for this bin's size class. */ /* Total number of regions in a slab for this bin's size class. */
uint32_t nregs; uint32_t nregs;
/* /*
* Metadata used to manipulate bitmaps for runs associated with this * Metadata used to manipulate bitmaps for slabs associated with this
* bin. * bin.
*/ */
bitmap_info_t bitmap_info; bitmap_info_t bitmap_info;
}; };
struct arena_bin_s { struct arena_bin_s {
/* /* All operations on arena_bin_t fields require lock ownership. */
* All operations on runcur, runs, and stats require that lock be
* locked. Run allocation/deallocation are protected by the arena lock,
* which may be acquired while holding one or more bin locks, but not
* vise versa.
*/
malloc_mutex_t lock; malloc_mutex_t lock;
/* /*
* Current run being used to service allocations of this bin's size * Current slab being used to service allocations of this bin's size
* class. * class. slabcur is independent of slabs_{nonfull,full}; whenever
* slabcur is reassigned, the previous slab must be deallocated or
* inserted into slabs_{nonfull,full}.
*/ */
arena_run_t *runcur; extent_t *slabcur;
/* /*
* Heap of non-full runs. This heap is used when looking for an * Heap of non-full slabs. This heap is used to assure that new
* existing run when runcur is no longer usable. We choose the * allocations come from the non-full slab that is lowest in memory.
* non-full run that is lowest in memory; this policy tends to keep
* objects packed well, and it can also help reduce the number of
* almost-empty chunks.
*/ */
arena_run_heap_t runs; extent_heap_t slabs_nonfull;
/* Ring sentinel used to track full slabs. */
extent_t slabs_full;
/* Bin statistics. */ /* Bin statistics. */
malloc_bin_stats_t stats; malloc_bin_stats_t stats;
@ -272,7 +137,7 @@ struct arena_s {
* perspective: * perspective:
* 1) Thread assignment (modifies nthreads) is synchronized via atomics. * 1) Thread assignment (modifies nthreads) is synchronized via atomics.
* 2) Bin-related operations are protected by bin locks. * 2) Bin-related operations are protected by bin locks.
* 3) Chunk- and run-related operations are protected by this mutex. * 3) Chunk-related operations are protected by this mutex.
*/ */
malloc_mutex_t lock; malloc_mutex_t lock;
@ -294,32 +159,17 @@ struct arena_s {
dss_prec_t dss_prec; dss_prec_t dss_prec;
/* Extant arena chunks. */
ql_head(extent_t) achunks;
/*
* In order to avoid rapid chunk allocation/deallocation when an arena
* oscillates right on the cusp of needing a new chunk, cache the most
* recently freed chunk. The spare is left in the arena's chunk trees
* until it is deleted.
*
* There is one spare chunk per arena, rather than one spare total, in
* order to avoid interactions between multiple threads that could make
* a single spare inadequate.
*/
extent_t *spare;
/* Minimum ratio (log base 2) of nactive:ndirty. */ /* Minimum ratio (log base 2) of nactive:ndirty. */
ssize_t lg_dirty_mult; ssize_t lg_dirty_mult;
/* True if a thread is currently executing arena_purge_to_limit(). */ /* True if a thread is currently executing arena_purge_to_limit(). */
bool purging; bool purging;
/* Number of pages in active runs and huge regions. */ /* Number of pages in active extents. */
size_t nactive; size_t nactive;
/* /*
* Current count of pages within unused runs that are potentially * Current count of pages within unused extents that are potentially
* dirty, and for which madvise(... MADV_DONTNEED) has not been called. * dirty, and for which madvise(... MADV_DONTNEED) has not been called.
* By tracking this, we can institute a limit on how much dirty unused * By tracking this, we can institute a limit on how much dirty unused
* memory is mapped for each arena. * memory is mapped for each arena.
@ -327,35 +177,10 @@ struct arena_s {
size_t ndirty; size_t ndirty;
/* /*
* Unused dirty memory this arena manages. Dirty memory is conceptually * Ring sentinel used to track unused dirty memory. Dirty memory is
* tracked as an arbitrarily interleaved LRU of dirty runs and cached * managed as an LRU of cached extents.
* chunks, but the list linkage is actually semi-duplicated in order to
* avoid extra arena_chunk_map_misc_t space overhead.
*
* LRU-----------------------------------------------------------MRU
*
* /-- arena ---\
* | |
* | |
* |------------| /-- chunk --\
* ...->|chunks_cache|<--------------------------->| /------\ |<--...
* |------------| | |extent| |
* | | | | | |
* | | /- run -\ /- run -\ | | | |
* | | | | | | | | | |
* | | | | | | | | | |
* |------------| |-------| |-------| | |------| |
* ...->|runs_dirty |<-->|rd |<-->|rd |<---->|rd |<----...
* |------------| |-------| |-------| | |------| |
* | | | | | | | | | |
* | | | | | | | \------/ |
* | | \-------/ \-------/ | |
* | | | |
* | | | |
* \------------/ \-----------/
*/ */
arena_runs_dirty_link_t runs_dirty; extent_t extents_dirty;
extent_t chunks_cache;
/* /*
* Approximate time in seconds from the creation of a set of unused * Approximate time in seconds from the creation of a set of unused
@ -424,16 +249,8 @@ struct arena_s {
/* User-configurable chunk hook functions. */ /* User-configurable chunk hook functions. */
chunk_hooks_t chunk_hooks; chunk_hooks_t chunk_hooks;
/* bins is used to store trees of free regions. */ /* bins is used to store heaps of free regions. */
arena_bin_t bins[NBINS]; arena_bin_t bins[NBINS];
/*
* Size-segregated address-ordered heaps of this arena's available runs,
* used for first-best-fit run allocation. Runs are quantized, i.e.
* they reside in the last heap which corresponds to a size class less
* than or equal to the run size.
*/
arena_run_heap_t runs_avail[NPSIZES];
}; };
/* Used in conjunction with tsd for fast arena-related context lookup. */ /* Used in conjunction with tsd for fast arena-related context lookup. */
@ -461,15 +278,6 @@ extern ssize_t opt_decay_time;
extern const arena_bin_info_t arena_bin_info[NBINS]; extern const arena_bin_info_t arena_bin_info[NBINS];
extern size_t map_bias; /* Number of arena chunk header pages. */
extern size_t map_misc_offset;
extern size_t arena_maxrun; /* Max run size for arenas. */
#ifdef JEMALLOC_JET
typedef size_t (run_quantize_t)(size_t);
extern run_quantize_t *run_quantize_floor;
extern run_quantize_t *run_quantize_ceil;
#endif
extent_t *arena_chunk_cache_alloc(tsdn_t *tsdn, arena_t *arena, extent_t *arena_chunk_cache_alloc(tsdn_t *tsdn, arena_t *arena,
chunk_hooks_t *chunk_hooks, void *new_addr, size_t size, size_t alignment, chunk_hooks_t *chunk_hooks, void *new_addr, size_t size, size_t alignment,
bool *zero); bool *zero);
@ -514,10 +322,9 @@ void arena_prof_promote(tsdn_t *tsdn, extent_t *extent, const void *ptr,
void arena_dalloc_promoted(tsdn_t *tsdn, extent_t *extent, void *ptr, void arena_dalloc_promoted(tsdn_t *tsdn, extent_t *extent, void *ptr,
tcache_t *tcache, bool slow_path); tcache_t *tcache, bool slow_path);
void arena_dalloc_bin_junked_locked(tsdn_t *tsdn, arena_t *arena, void arena_dalloc_bin_junked_locked(tsdn_t *tsdn, arena_t *arena,
arena_chunk_t *chunk, extent_t *extent, void *ptr, extent_t *extent, void *ptr);
arena_chunk_map_bits_t *bitselm); void arena_dalloc_small(tsdn_t *tsdn, arena_t *arena, extent_t *extent,
void arena_dalloc_small(tsdn_t *tsdn, arena_t *arena, arena_chunk_t *chunk, void *ptr);
extent_t *extent, void *ptr, size_t pageind);
bool arena_ralloc_no_move(tsdn_t *tsdn, extent_t *extent, void *ptr, bool arena_ralloc_no_move(tsdn_t *tsdn, extent_t *extent, void *ptr,
size_t oldsize, size_t size, size_t extra, bool zero); size_t oldsize, size_t size, size_t extra, bool zero);
void *arena_ralloc(tsdn_t *tsdn, arena_t *arena, extent_t *extent, void *ptr, void *arena_ralloc(tsdn_t *tsdn, arena_t *arena, extent_t *extent, void *ptr,
@ -552,70 +359,19 @@ void arena_postfork_child(tsdn_t *tsdn, arena_t *arena);
#ifdef JEMALLOC_H_INLINES #ifdef JEMALLOC_H_INLINES
#ifndef JEMALLOC_ENABLE_INLINE #ifndef JEMALLOC_ENABLE_INLINE
arena_chunk_map_bits_t *arena_bitselm_get_mutable(arena_chunk_t *chunk,
size_t pageind);
const arena_chunk_map_bits_t *arena_bitselm_get_const(
const arena_chunk_t *chunk, size_t pageind);
arena_chunk_map_misc_t *arena_miscelm_get_mutable(arena_chunk_t *chunk,
size_t pageind);
const arena_chunk_map_misc_t *arena_miscelm_get_const(
const arena_chunk_t *chunk, size_t pageind);
size_t arena_miscelm_to_pageind(const extent_t *extent,
const arena_chunk_map_misc_t *miscelm);
void *arena_miscelm_to_rpages(const extent_t *extent,
const arena_chunk_map_misc_t *miscelm);
arena_chunk_map_misc_t *arena_rd_to_miscelm(const extent_t *extent,
arena_runs_dirty_link_t *rd);
arena_chunk_map_misc_t *arena_run_to_miscelm(const extent_t *extent,
arena_run_t *run);
size_t *arena_mapbitsp_get_mutable(arena_chunk_t *chunk, size_t pageind);
const size_t *arena_mapbitsp_get_const(const arena_chunk_t *chunk,
size_t pageind);
size_t arena_mapbitsp_read(const size_t *mapbitsp);
size_t arena_mapbits_get(const arena_chunk_t *chunk, size_t pageind);
size_t arena_mapbits_size_decode(size_t mapbits);
size_t arena_mapbits_unallocated_size_get(const arena_chunk_t *chunk,
size_t pageind);
size_t arena_mapbits_large_size_get(const arena_chunk_t *chunk,
size_t pageind);
size_t arena_mapbits_small_runind_get(const arena_chunk_t *chunk,
size_t pageind);
szind_t arena_mapbits_binind_get(const arena_chunk_t *chunk, size_t pageind);
size_t arena_mapbits_dirty_get(const arena_chunk_t *chunk, size_t pageind);
size_t arena_mapbits_unzeroed_get(const arena_chunk_t *chunk, size_t pageind);
size_t arena_mapbits_decommitted_get(const arena_chunk_t *chunk,
size_t pageind);
size_t arena_mapbits_large_get(const arena_chunk_t *chunk, size_t pageind);
size_t arena_mapbits_allocated_get(const arena_chunk_t *chunk, size_t pageind);
void arena_mapbitsp_write(size_t *mapbitsp, size_t mapbits);
size_t arena_mapbits_size_encode(size_t size);
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_internal_set(arena_chunk_t *chunk, size_t pageind,
size_t flags);
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,
szind_t binind);
void arena_mapbits_small_set(arena_chunk_t *chunk, size_t pageind,
size_t runind, szind_t binind, size_t flags);
void arena_metadata_allocated_add(arena_t *arena, size_t size); void arena_metadata_allocated_add(arena_t *arena, size_t size);
void arena_metadata_allocated_sub(arena_t *arena, size_t size); void arena_metadata_allocated_sub(arena_t *arena, size_t size);
size_t arena_metadata_allocated_get(arena_t *arena); size_t arena_metadata_allocated_get(arena_t *arena);
bool arena_prof_accum_impl(arena_t *arena, uint64_t accumbytes); bool arena_prof_accum_impl(arena_t *arena, uint64_t accumbytes);
bool arena_prof_accum_locked(arena_t *arena, uint64_t accumbytes); bool arena_prof_accum_locked(arena_t *arena, uint64_t accumbytes);
bool arena_prof_accum(tsdn_t *tsdn, arena_t *arena, uint64_t accumbytes); bool arena_prof_accum(tsdn_t *tsdn, arena_t *arena, uint64_t accumbytes);
szind_t arena_ptr_small_binind_get(tsdn_t *tsdn, const void *ptr,
size_t mapbits);
szind_t arena_bin_index(arena_t *arena, arena_bin_t *bin); szind_t arena_bin_index(arena_t *arena, arena_bin_t *bin);
prof_tctx_t *arena_prof_tctx_get(tsdn_t *tsdn, const extent_t *extent, prof_tctx_t *arena_prof_tctx_get(tsdn_t *tsdn, const extent_t *extent,
const void *ptr); const void *ptr);
void arena_prof_tctx_set(tsdn_t *tsdn, extent_t *extent, const void *ptr, void arena_prof_tctx_set(tsdn_t *tsdn, extent_t *extent, const void *ptr,
size_t usize, prof_tctx_t *tctx); size_t usize, prof_tctx_t *tctx);
void arena_prof_tctx_reset(tsdn_t *tsdn, extent_t *extent, const void *ptr, void arena_prof_tctx_reset(tsdn_t *tsdn, extent_t *extent, const void *ptr,
size_t usize, const void *old_ptr, prof_tctx_t *old_tctx); prof_tctx_t *tctx);
void arena_decay_ticks(tsdn_t *tsdn, arena_t *arena, unsigned nticks); void arena_decay_ticks(tsdn_t *tsdn, arena_t *arena, unsigned nticks);
void arena_decay_tick(tsdn_t *tsdn, arena_t *arena); void arena_decay_tick(tsdn_t *tsdn, arena_t *arena);
void *arena_malloc(tsdn_t *tsdn, arena_t *arena, size_t size, szind_t ind, void *arena_malloc(tsdn_t *tsdn, arena_t *arena, size_t size, szind_t ind,
@ -630,330 +386,6 @@ void arena_sdalloc(tsdn_t *tsdn, extent_t *extent, void *ptr, size_t size,
#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_ARENA_C_)) #if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_ARENA_C_))
# ifdef JEMALLOC_ARENA_INLINE_A # ifdef JEMALLOC_ARENA_INLINE_A
JEMALLOC_ALWAYS_INLINE arena_chunk_map_bits_t *
arena_bitselm_get_mutable(arena_chunk_t *chunk, size_t pageind)
{
assert(pageind >= map_bias);
assert(pageind < chunk_npages);
return (&chunk->map_bits[pageind-map_bias]);
}
JEMALLOC_ALWAYS_INLINE const arena_chunk_map_bits_t *
arena_bitselm_get_const(const arena_chunk_t *chunk, size_t pageind)
{
return (arena_bitselm_get_mutable((arena_chunk_t *)chunk, pageind));
}
JEMALLOC_ALWAYS_INLINE arena_chunk_map_misc_t *
arena_miscelm_get_mutable(arena_chunk_t *chunk, size_t pageind)
{
assert(pageind >= map_bias);
assert(pageind < chunk_npages);
return ((arena_chunk_map_misc_t *)((uintptr_t)chunk +
(uintptr_t)map_misc_offset) + pageind-map_bias);
}
JEMALLOC_ALWAYS_INLINE const arena_chunk_map_misc_t *
arena_miscelm_get_const(const arena_chunk_t *chunk, size_t pageind)
{
return (arena_miscelm_get_mutable((arena_chunk_t *)chunk, pageind));
}
JEMALLOC_ALWAYS_INLINE size_t
arena_miscelm_to_pageind(const extent_t *extent,
const arena_chunk_map_misc_t *miscelm)
{
arena_chunk_t *chunk = (arena_chunk_t *)extent_base_get(extent);
size_t pageind = ((uintptr_t)miscelm - ((uintptr_t)chunk +
map_misc_offset)) / sizeof(arena_chunk_map_misc_t) + map_bias;
assert(pageind >= map_bias);
assert(pageind < chunk_npages);
return (pageind);
}
JEMALLOC_ALWAYS_INLINE void *
arena_miscelm_to_rpages(const extent_t *extent,
const arena_chunk_map_misc_t *miscelm)
{
arena_chunk_t *chunk = (arena_chunk_t *)extent_base_get(extent);
size_t pageind = arena_miscelm_to_pageind(extent, miscelm);
return ((void *)((uintptr_t)chunk + (pageind << LG_PAGE)));
}
JEMALLOC_ALWAYS_INLINE arena_chunk_map_misc_t *
arena_rd_to_miscelm(const extent_t *extent, arena_runs_dirty_link_t *rd)
{
arena_chunk_map_misc_t *miscelm = (arena_chunk_map_misc_t
*)((uintptr_t)rd - offsetof(arena_chunk_map_misc_t, rd));
assert(arena_miscelm_to_pageind(extent, miscelm) >= map_bias);
assert(arena_miscelm_to_pageind(extent, miscelm) < chunk_npages);
return (miscelm);
}
JEMALLOC_ALWAYS_INLINE arena_chunk_map_misc_t *
arena_run_to_miscelm(const extent_t *extent, arena_run_t *run)
{
arena_chunk_map_misc_t *miscelm = (arena_chunk_map_misc_t
*)((uintptr_t)run - offsetof(arena_chunk_map_misc_t, run));
assert(arena_miscelm_to_pageind(extent, miscelm) >= map_bias);
assert(arena_miscelm_to_pageind(extent, miscelm) < chunk_npages);
return (miscelm);
}
JEMALLOC_ALWAYS_INLINE size_t *
arena_mapbitsp_get_mutable(arena_chunk_t *chunk, size_t pageind)
{
return (&arena_bitselm_get_mutable(chunk, pageind)->bits);
}
JEMALLOC_ALWAYS_INLINE const size_t *
arena_mapbitsp_get_const(const arena_chunk_t *chunk, size_t pageind)
{
return (arena_mapbitsp_get_mutable((arena_chunk_t *)chunk, pageind));
}
JEMALLOC_ALWAYS_INLINE size_t
arena_mapbitsp_read(const size_t *mapbitsp)
{
return (*mapbitsp);
}
JEMALLOC_ALWAYS_INLINE size_t
arena_mapbits_get(const arena_chunk_t *chunk, size_t pageind)
{
return (arena_mapbitsp_read(arena_mapbitsp_get_const(chunk, pageind)));
}
JEMALLOC_ALWAYS_INLINE size_t
arena_mapbits_size_decode(size_t mapbits)
{
size_t size;
#if CHUNK_MAP_SIZE_SHIFT > 0
size = (mapbits & CHUNK_MAP_SIZE_MASK) >> CHUNK_MAP_SIZE_SHIFT;
#elif CHUNK_MAP_SIZE_SHIFT == 0
size = mapbits & CHUNK_MAP_SIZE_MASK;
#else
size = (mapbits & CHUNK_MAP_SIZE_MASK) << -CHUNK_MAP_SIZE_SHIFT;
#endif
return (size);
}
JEMALLOC_ALWAYS_INLINE size_t
arena_mapbits_unallocated_size_get(const 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 (arena_mapbits_size_decode(mapbits));
}
JEMALLOC_ALWAYS_INLINE size_t
arena_mapbits_large_size_get(const 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 (arena_mapbits_size_decode(mapbits));
}
JEMALLOC_ALWAYS_INLINE size_t
arena_mapbits_small_runind_get(const 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 >> CHUNK_MAP_RUNIND_SHIFT);
}
JEMALLOC_ALWAYS_INLINE szind_t
arena_mapbits_binind_get(const arena_chunk_t *chunk, size_t pageind)
{
size_t mapbits;
szind_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_ALWAYS_INLINE size_t
arena_mapbits_dirty_get(const arena_chunk_t *chunk, size_t pageind)
{
size_t mapbits;
mapbits = arena_mapbits_get(chunk, pageind);
assert((mapbits & CHUNK_MAP_DECOMMITTED) == 0 || (mapbits &
(CHUNK_MAP_DIRTY|CHUNK_MAP_UNZEROED)) == 0);
return (mapbits & CHUNK_MAP_DIRTY);
}
JEMALLOC_ALWAYS_INLINE size_t
arena_mapbits_unzeroed_get(const arena_chunk_t *chunk, size_t pageind)
{
size_t mapbits;
mapbits = arena_mapbits_get(chunk, pageind);
assert((mapbits & CHUNK_MAP_DECOMMITTED) == 0 || (mapbits &
(CHUNK_MAP_DIRTY|CHUNK_MAP_UNZEROED)) == 0);
return (mapbits & CHUNK_MAP_UNZEROED);
}
JEMALLOC_ALWAYS_INLINE size_t
arena_mapbits_decommitted_get(const arena_chunk_t *chunk, size_t pageind)
{
size_t mapbits;
mapbits = arena_mapbits_get(chunk, pageind);
assert((mapbits & CHUNK_MAP_DECOMMITTED) == 0 || (mapbits &
(CHUNK_MAP_DIRTY|CHUNK_MAP_UNZEROED)) == 0);
return (mapbits & CHUNK_MAP_DECOMMITTED);
}
JEMALLOC_ALWAYS_INLINE size_t
arena_mapbits_large_get(const arena_chunk_t *chunk, size_t pageind)
{
size_t mapbits;
mapbits = arena_mapbits_get(chunk, pageind);
return (mapbits & CHUNK_MAP_LARGE);
}
JEMALLOC_ALWAYS_INLINE size_t
arena_mapbits_allocated_get(const arena_chunk_t *chunk, size_t pageind)
{
size_t mapbits;
mapbits = arena_mapbits_get(chunk, pageind);
return (mapbits & CHUNK_MAP_ALLOCATED);
}
JEMALLOC_ALWAYS_INLINE void
arena_mapbitsp_write(size_t *mapbitsp, size_t mapbits)
{
*mapbitsp = mapbits;
}
JEMALLOC_ALWAYS_INLINE size_t
arena_mapbits_size_encode(size_t size)
{
size_t mapbits;
#if CHUNK_MAP_SIZE_SHIFT > 0
mapbits = size << CHUNK_MAP_SIZE_SHIFT;
#elif CHUNK_MAP_SIZE_SHIFT == 0
mapbits = size;
#else
mapbits = size >> -CHUNK_MAP_SIZE_SHIFT;
#endif
assert((mapbits & ~CHUNK_MAP_SIZE_MASK) == 0);
return (mapbits);
}
JEMALLOC_ALWAYS_INLINE void
arena_mapbits_unallocated_set(arena_chunk_t *chunk, size_t pageind, size_t size,
size_t flags)
{
size_t *mapbitsp = arena_mapbitsp_get_mutable(chunk, pageind);
assert((size & PAGE_MASK) == 0);
assert((flags & CHUNK_MAP_FLAGS_MASK) == flags);
assert((flags & CHUNK_MAP_DECOMMITTED) == 0 || (flags &
(CHUNK_MAP_DIRTY|CHUNK_MAP_UNZEROED)) == 0);
arena_mapbitsp_write(mapbitsp, arena_mapbits_size_encode(size) |
CHUNK_MAP_BININD_INVALID | flags);
}
JEMALLOC_ALWAYS_INLINE void
arena_mapbits_unallocated_size_set(arena_chunk_t *chunk, size_t pageind,
size_t size)
{
size_t *mapbitsp = arena_mapbitsp_get_mutable(chunk, pageind);
size_t mapbits = arena_mapbitsp_read(mapbitsp);
assert((size & PAGE_MASK) == 0);
assert((mapbits & (CHUNK_MAP_LARGE|CHUNK_MAP_ALLOCATED)) == 0);
arena_mapbitsp_write(mapbitsp, arena_mapbits_size_encode(size) |
(mapbits & ~CHUNK_MAP_SIZE_MASK));
}
JEMALLOC_ALWAYS_INLINE void
arena_mapbits_internal_set(arena_chunk_t *chunk, size_t pageind, size_t flags)
{
size_t *mapbitsp = arena_mapbitsp_get_mutable(chunk, pageind);
assert((flags & CHUNK_MAP_UNZEROED) == flags);
arena_mapbitsp_write(mapbitsp, flags);
}
JEMALLOC_ALWAYS_INLINE void
arena_mapbits_large_set(arena_chunk_t *chunk, size_t pageind, size_t size,
size_t flags)
{
size_t *mapbitsp = arena_mapbitsp_get_mutable(chunk, pageind);
assert((size & PAGE_MASK) == 0);
assert((flags & CHUNK_MAP_FLAGS_MASK) == flags);
assert((flags & CHUNK_MAP_DECOMMITTED) == 0 || (flags &
(CHUNK_MAP_DIRTY|CHUNK_MAP_UNZEROED)) == 0);
arena_mapbitsp_write(mapbitsp, arena_mapbits_size_encode(size) |
CHUNK_MAP_BININD_INVALID | flags | CHUNK_MAP_LARGE |
CHUNK_MAP_ALLOCATED);
}
JEMALLOC_ALWAYS_INLINE void
arena_mapbits_large_binind_set(arena_chunk_t *chunk, size_t pageind,
szind_t binind)
{
size_t *mapbitsp = arena_mapbitsp_get_mutable(chunk, pageind);
size_t mapbits = arena_mapbitsp_read(mapbitsp);
assert(binind <= BININD_INVALID);
assert(arena_mapbits_large_size_get(chunk, pageind) == LARGE_MINCLASS +
large_pad);
arena_mapbitsp_write(mapbitsp, (mapbits & ~CHUNK_MAP_BININD_MASK) |
(binind << CHUNK_MAP_BININD_SHIFT));
}
JEMALLOC_ALWAYS_INLINE void
arena_mapbits_small_set(arena_chunk_t *chunk, size_t pageind, size_t runind,
szind_t binind, size_t flags)
{
size_t *mapbitsp = arena_mapbitsp_get_mutable(chunk, pageind);
assert(binind < BININD_INVALID);
assert(pageind - runind >= map_bias);
assert((flags & CHUNK_MAP_UNZEROED) == flags);
arena_mapbitsp_write(mapbitsp, (runind << CHUNK_MAP_RUNIND_SHIFT) |
(binind << CHUNK_MAP_BININD_SHIFT) | flags | CHUNK_MAP_ALLOCATED);
}
JEMALLOC_INLINE void JEMALLOC_INLINE void
arena_metadata_allocated_add(arena_t *arena, size_t size) arena_metadata_allocated_add(arena_t *arena, size_t size)
{ {
@ -1022,54 +454,6 @@ arena_prof_accum(tsdn_t *tsdn, arena_t *arena, uint64_t accumbytes)
# endif /* JEMALLOC_ARENA_INLINE_A */ # endif /* JEMALLOC_ARENA_INLINE_A */
# ifdef JEMALLOC_ARENA_INLINE_B # ifdef JEMALLOC_ARENA_INLINE_B
JEMALLOC_ALWAYS_INLINE szind_t
arena_ptr_small_binind_get(tsdn_t *tsdn, const void *ptr, size_t mapbits)
{
szind_t binind;
binind = (mapbits & CHUNK_MAP_BININD_MASK) >> CHUNK_MAP_BININD_SHIFT;
if (config_debug) {
const extent_t *extent;
arena_chunk_t *chunk;
arena_t *arena;
size_t pageind;
size_t actual_mapbits;
size_t rpages_ind;
const arena_run_t *run;
arena_bin_t *bin;
szind_t run_binind, actual_binind;
const arena_bin_info_t *bin_info;
const arena_chunk_map_misc_t *miscelm;
const void *rpages;
assert(binind != BININD_INVALID);
assert(binind < NBINS);
extent = iealloc(tsdn, ptr);
chunk = (arena_chunk_t *)extent_base_get(extent);
arena = extent_arena_get(extent);
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);
rpages_ind = pageind - arena_mapbits_small_runind_get(chunk,
pageind);
miscelm = arena_miscelm_get_const(chunk, rpages_ind);
run = &miscelm->run;
run_binind = run->binind;
bin = &arena->bins[run_binind];
actual_binind = (szind_t)(bin - arena->bins);
assert(run_binind == actual_binind);
bin_info = &arena_bin_info[actual_binind];
rpages = arena_miscelm_to_rpages(extent, miscelm);
assert(((uintptr_t)ptr - (uintptr_t)rpages) % bin_info->reg_size
== 0);
}
return (binind);
}
JEMALLOC_INLINE szind_t JEMALLOC_INLINE szind_t
arena_bin_index(arena_t *arena, arena_bin_t *bin) arena_bin_index(arena_t *arena, arena_bin_t *bin)
{ {
@ -1081,27 +465,13 @@ arena_bin_index(arena_t *arena, arena_bin_t *bin)
JEMALLOC_INLINE prof_tctx_t * JEMALLOC_INLINE prof_tctx_t *
arena_prof_tctx_get(tsdn_t *tsdn, const extent_t *extent, const void *ptr) arena_prof_tctx_get(tsdn_t *tsdn, const extent_t *extent, const void *ptr)
{ {
prof_tctx_t *ret;
cassert(config_prof); cassert(config_prof);
assert(ptr != NULL); assert(ptr != NULL);
if (likely(extent_slab_get(extent))) { if (unlikely(!extent_slab_get(extent)))
arena_chunk_t *chunk = (arena_chunk_t *)extent_base_get(extent); return (huge_prof_tctx_get(tsdn, extent));
size_t pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE; return ((prof_tctx_t *)(uintptr_t)1U);
size_t mapbits = arena_mapbits_get(chunk, pageind);
assert((mapbits & CHUNK_MAP_ALLOCATED) != 0);
if (likely((mapbits & CHUNK_MAP_LARGE) == 0))
ret = (prof_tctx_t *)(uintptr_t)1U;
else {
arena_chunk_map_misc_t *elm =
arena_miscelm_get_mutable(chunk, pageind);
ret = atomic_read_p(&elm->prof_tctx_pun);
}
} else
ret = huge_prof_tctx_get(tsdn, extent);
return (ret);
} }
JEMALLOC_INLINE void JEMALLOC_INLINE void
@ -1112,62 +482,21 @@ arena_prof_tctx_set(tsdn_t *tsdn, extent_t *extent, const void *ptr,
cassert(config_prof); cassert(config_prof);
assert(ptr != NULL); assert(ptr != NULL);
if (likely(extent_slab_get(extent))) { if (unlikely(!extent_slab_get(extent)))
arena_chunk_t *chunk = (arena_chunk_t *)extent_base_get(extent);
size_t pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE;
assert(arena_mapbits_allocated_get(chunk, pageind) != 0);
if (unlikely(usize > SMALL_MAXCLASS || (uintptr_t)tctx >
(uintptr_t)1U)) {
arena_chunk_map_misc_t *elm;
assert(arena_mapbits_large_get(chunk, pageind) != 0);
elm = arena_miscelm_get_mutable(chunk, pageind);
atomic_write_p(&elm->prof_tctx_pun, tctx);
} else {
/*
* tctx must always be initialized for large runs.
* Assert that the surrounding conditional logic is
* equivalent to checking whether ptr refers to a large
* run.
*/
assert(arena_mapbits_large_get(chunk, pageind) == 0);
}
} else
huge_prof_tctx_set(tsdn, extent, tctx); huge_prof_tctx_set(tsdn, extent, tctx);
} }
JEMALLOC_INLINE void JEMALLOC_INLINE void
arena_prof_tctx_reset(tsdn_t *tsdn, extent_t *extent, const void *ptr, arena_prof_tctx_reset(tsdn_t *tsdn, extent_t *extent, const void *ptr,
size_t usize, const void *old_ptr, prof_tctx_t *old_tctx) prof_tctx_t *tctx)
{ {
cassert(config_prof); cassert(config_prof);
assert(ptr != NULL); assert(ptr != NULL);
assert(!extent_slab_get(extent));
if (unlikely(usize > SMALL_MAXCLASS || (ptr == old_ptr &&
(uintptr_t)old_tctx > (uintptr_t)1U))) {
if (likely(extent_slab_get(extent))) {
arena_chunk_t *chunk =
(arena_chunk_t *)extent_base_get(extent);
size_t pageind;
arena_chunk_map_misc_t *elm;
pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >>
LG_PAGE;
assert(arena_mapbits_allocated_get(chunk, pageind) !=
0);
assert(arena_mapbits_large_get(chunk, pageind) != 0);
elm = arena_miscelm_get_mutable(chunk, pageind);
atomic_write_p(&elm->prof_tctx_pun,
(prof_tctx_t *)(uintptr_t)1U);
} else
huge_prof_tctx_reset(tsdn, extent); huge_prof_tctx_reset(tsdn, extent);
} }
}
JEMALLOC_ALWAYS_INLINE void JEMALLOC_ALWAYS_INLINE void
arena_decay_ticks(tsdn_t *tsdn, arena_t *arena, unsigned nticks) arena_decay_ticks(tsdn_t *tsdn, arena_t *arena, unsigned nticks)
@ -1231,20 +560,9 @@ arena_salloc(tsdn_t *tsdn, const extent_t *extent, const void *ptr)
assert(ptr != NULL); assert(ptr != NULL);
if (likely(extent_slab_get(extent))) { if (likely(extent_slab_get(extent)))
const arena_chunk_t *chunk = ret = index2size(extent_slab_data_get_const(extent)->binind);
(const arena_chunk_t *)extent_base_get(extent); else
size_t pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE;
szind_t binind;
assert(arena_mapbits_allocated_get(chunk, pageind) != 0);
binind = arena_mapbits_binind_get(chunk, pageind);
/* Small allocation. */
assert(arena_mapbits_large_get(chunk, pageind) != 0 ||
arena_ptr_small_binind_get(tsdn, ptr,
arena_mapbits_get(chunk, pageind)) == binind);
ret = index2size(binind);
} else
ret = huge_salloc(tsdn, extent); ret = huge_salloc(tsdn, extent);
return (ret); return (ret);
@ -1260,19 +578,13 @@ arena_dalloc(tsdn_t *tsdn, extent_t *extent, void *ptr, tcache_t *tcache,
if (likely(extent_slab_get(extent))) { if (likely(extent_slab_get(extent))) {
/* Small allocation. */ /* Small allocation. */
arena_chunk_t *chunk = (arena_chunk_t *)extent_base_get(extent);
size_t pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE;
size_t mapbits = arena_mapbits_get(chunk, pageind);
assert(arena_mapbits_allocated_get(chunk, pageind) != 0);
assert((mapbits & CHUNK_MAP_LARGE) == 0);
if (likely(tcache != NULL)) { if (likely(tcache != NULL)) {
szind_t binind = arena_ptr_small_binind_get(tsdn, ptr, szind_t binind = extent_slab_data_get(extent)->binind;
mapbits);
tcache_dalloc_small(tsdn_tsd(tsdn), tcache, ptr, binind, tcache_dalloc_small(tsdn_tsd(tsdn), tcache, ptr, binind,
slow_path); slow_path);
} else { } else {
arena_dalloc_small(tsdn, extent_arena_get(extent), arena_dalloc_small(tsdn, extent_arena_get(extent),
chunk, extent, ptr, pageind); extent, ptr);
} }
} else { } else {
size_t usize = extent_usize_get(extent); size_t usize = extent_usize_get(extent);
@ -1282,8 +594,8 @@ arena_dalloc(tsdn_t *tsdn, extent_t *extent, void *ptr, tcache_t *tcache,
arena_dalloc_promoted(tsdn, extent, ptr, arena_dalloc_promoted(tsdn, extent, ptr,
tcache, slow_path); tcache, slow_path);
} else { } else {
tcache_dalloc_huge(tsdn_tsd(tsdn), tcache, ptr, tcache_dalloc_huge(tsdn_tsd(tsdn), tcache,
usize, slow_path); ptr, usize, slow_path);
} }
} else } else
huge_dalloc(tsdn, extent); huge_dalloc(tsdn, extent);
@ -1302,15 +614,12 @@ arena_sdalloc(tsdn_t *tsdn, extent_t *extent, void *ptr, size_t size,
/* Small allocation. */ /* Small allocation. */
if (likely(tcache != NULL)) { if (likely(tcache != NULL)) {
szind_t binind = size2index(size); szind_t binind = size2index(size);
assert(binind == extent_slab_data_get(extent)->binind);
tcache_dalloc_small(tsdn_tsd(tsdn), tcache, ptr, binind, tcache_dalloc_small(tsdn_tsd(tsdn), tcache, ptr, binind,
slow_path); slow_path);
} else { } else {
arena_chunk_t *chunk =
(arena_chunk_t *)extent_base_get(extent);
size_t pageind = ((uintptr_t)ptr -
(uintptr_t)chunk) >> LG_PAGE;
arena_dalloc_small(tsdn, extent_arena_get(extent), arena_dalloc_small(tsdn, extent_arena_get(extent),
chunk, extent, ptr, pageind); extent, ptr);
} }
} else { } else {
if (likely(tcache != NULL) && size <= tcache_maxclass) { if (likely(tcache != NULL) && size <= tcache_maxclass) {

2
include/jemalloc/internal/bitmap.h
View File

@ -2,7 +2,7 @@
#ifdef JEMALLOC_H_TYPES #ifdef JEMALLOC_H_TYPES
/* Maximum bitmap bit count is 2^LG_BITMAP_MAXBITS. */ /* Maximum bitmap bit count is 2^LG_BITMAP_MAXBITS. */
#define LG_BITMAP_MAXBITS LG_RUN_MAXREGS #define LG_BITMAP_MAXBITS LG_SLAB_MAXREGS
#define BITMAP_MAXBITS (ZU(1) << LG_BITMAP_MAXBITS) #define BITMAP_MAXBITS (ZU(1) << LG_BITMAP_MAXBITS)
typedef struct bitmap_level_s bitmap_level_t; typedef struct bitmap_level_s bitmap_level_t;

80
include/jemalloc/internal/extent.h
View File

@ -27,9 +27,6 @@ struct extent_s {
/* True if extent is active (in use). */ /* True if extent is active (in use). */
bool e_active; bool e_active;
/* True if extent is dirty (touched). */
bool e_dirty;
/* /*
* The zeroed flag is used by chunk recycling code to track whether * The zeroed flag is used by chunk recycling code to track whether
* memory is zero-filled. * memory is zero-filled.
@ -50,21 +47,27 @@ struct extent_s {
*/ */
bool e_slab; bool e_slab;
union {
/* Small region slab metadata. */
arena_slab_data_t e_slab_data;
/* Profile counters, used for huge objects. */ /* Profile counters, used for huge objects. */
union { union {
void *e_prof_tctx_pun; void *e_prof_tctx_pun;
prof_tctx_t *e_prof_tctx; prof_tctx_t *e_prof_tctx;
}; };
};
/* Linkage for arena's runs_dirty and chunks_cache rings. */ /*
arena_runs_dirty_link_t rd; * Linkage for arena's extents_dirty and arena_bin_t's slabs_full rings.
qr(extent_t) cc_link; */
qr(extent_t) qr_link;
union { union {
/* Linkage for per size class address-ordered heaps. */ /* Linkage for per size class address-ordered heaps. */
phn(extent_t) ph_link; phn(extent_t) ph_link;
/* Linkage for arena's achunks, huge, and node_cache lists. */ /* Linkage for arena's huge and extent_cache lists. */
ql_elm(extent_t) ql_link; ql_elm(extent_t) ql_link;
}; };
}; };
@ -102,11 +105,12 @@ void *extent_before_get(const extent_t *extent);
void *extent_last_get(const extent_t *extent); void *extent_last_get(const extent_t *extent);
void *extent_past_get(const extent_t *extent); void *extent_past_get(const extent_t *extent);
bool extent_active_get(const extent_t *extent); bool extent_active_get(const extent_t *extent);
bool extent_dirty_get(const extent_t *extent);
bool extent_retained_get(const extent_t *extent); bool extent_retained_get(const extent_t *extent);
bool extent_zeroed_get(const extent_t *extent); bool extent_zeroed_get(const extent_t *extent);
bool extent_committed_get(const extent_t *extent); bool extent_committed_get(const extent_t *extent);
bool extent_slab_get(const extent_t *extent); bool extent_slab_get(const extent_t *extent);
arena_slab_data_t *extent_slab_data_get(extent_t *extent);
const arena_slab_data_t *extent_slab_data_get_const(const extent_t *extent);
prof_tctx_t *extent_prof_tctx_get(const extent_t *extent); prof_tctx_t *extent_prof_tctx_get(const extent_t *extent);
void extent_arena_set(extent_t *extent, arena_t *arena); void extent_arena_set(extent_t *extent, arena_t *arena);
void extent_addr_set(extent_t *extent, void *addr); void extent_addr_set(extent_t *extent, void *addr);
@ -114,17 +118,15 @@ void extent_addr_randomize(tsdn_t *tsdn, extent_t *extent, size_t alignment);
void extent_size_set(extent_t *extent, size_t size); void extent_size_set(extent_t *extent, size_t size);
void extent_usize_set(extent_t *extent, size_t usize); void extent_usize_set(extent_t *extent, size_t usize);
void extent_active_set(extent_t *extent, bool active); void extent_active_set(extent_t *extent, bool active);
void extent_dirty_set(extent_t *extent, bool dirty);
void extent_zeroed_set(extent_t *extent, bool zeroed); void extent_zeroed_set(extent_t *extent, bool zeroed);
void extent_committed_set(extent_t *extent, bool committed); void extent_committed_set(extent_t *extent, bool committed);
void extent_slab_set(extent_t *extent, bool slab); void extent_slab_set(extent_t *extent, bool slab);
void extent_prof_tctx_set(extent_t *extent, prof_tctx_t *tctx); void extent_prof_tctx_set(extent_t *extent, prof_tctx_t *tctx);
void extent_init(extent_t *extent, arena_t *arena, void *addr, void extent_init(extent_t *extent, arena_t *arena, void *addr,
size_t size, size_t usize, bool active, bool dirty, bool zeroed, size_t size, size_t usize, bool active, bool zeroed, bool committed,
bool committed, bool slab); bool slab);
void extent_dirty_insert(extent_t *extent, void extent_ring_insert(extent_t *sentinel, extent_t *extent);
arena_runs_dirty_link_t *runs_dirty, extent_t *chunks_dirty); void extent_ring_remove(extent_t *extent);
void extent_dirty_remove(extent_t *extent);
#endif #endif
#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_EXTENT_C_)) #if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_EXTENT_C_))
@ -197,18 +199,11 @@ extent_active_get(const extent_t *extent)
return (extent->e_active); return (extent->e_active);
} }
JEMALLOC_INLINE bool
extent_dirty_get(const extent_t *extent)
{
return (extent->e_dirty);
}
JEMALLOC_INLINE bool JEMALLOC_INLINE bool
extent_retained_get(const extent_t *extent) extent_retained_get(const extent_t *extent)
{ {
return (qr_next(&extent->rd, rd_link) == &extent->rd); return (qr_next(extent, qr_link) == extent);
} }
JEMALLOC_INLINE bool JEMALLOC_INLINE bool
@ -232,6 +227,22 @@ extent_slab_get(const extent_t *extent)
return (extent->e_slab); return (extent->e_slab);
} }
JEMALLOC_INLINE arena_slab_data_t *
extent_slab_data_get(extent_t *extent)
{
assert(extent->e_slab);
return (&extent->e_slab_data);
}
JEMALLOC_INLINE const arena_slab_data_t *
extent_slab_data_get_const(const extent_t *extent)
{
assert(extent->e_slab);
return (&extent->e_slab_data);
}
JEMALLOC_INLINE prof_tctx_t * JEMALLOC_INLINE prof_tctx_t *
extent_prof_tctx_get(const extent_t *extent) extent_prof_tctx_get(const extent_t *extent)
{ {
@ -296,13 +307,6 @@ extent_active_set(extent_t *extent, bool active)
extent->e_active = active; extent->e_active = active;
} }
JEMALLOC_INLINE void
extent_dirty_set(extent_t *extent, bool dirty)
{
extent->e_dirty = dirty;
}
JEMALLOC_INLINE void JEMALLOC_INLINE void
extent_zeroed_set(extent_t *extent, bool zeroed) extent_zeroed_set(extent_t *extent, bool zeroed)
{ {
@ -333,8 +337,7 @@ extent_prof_tctx_set(extent_t *extent, prof_tctx_t *tctx)
JEMALLOC_INLINE void JEMALLOC_INLINE void
extent_init(extent_t *extent, arena_t *arena, void *addr, size_t size, extent_init(extent_t *extent, arena_t *arena, void *addr, size_t size,
size_t usize, bool active, bool dirty, bool zeroed, bool committed, size_t usize, bool active, bool zeroed, bool committed, bool slab)
bool slab)
{ {
assert(addr == PAGE_ADDR2BASE(addr) || !slab); assert(addr == PAGE_ADDR2BASE(addr) || !slab);
@ -344,31 +347,26 @@ extent_init(extent_t *extent, arena_t *arena, void *addr, size_t size,
extent_size_set(extent, size); extent_size_set(extent, size);
extent_usize_set(extent, usize); extent_usize_set(extent, usize);
extent_active_set(extent, active); extent_active_set(extent, active);
extent_dirty_set(extent, dirty);
extent_zeroed_set(extent, zeroed); extent_zeroed_set(extent, zeroed);
extent_committed_set(extent, committed); extent_committed_set(extent, committed);
extent_slab_set(extent, slab); extent_slab_set(extent, slab);
if (config_prof) if (config_prof)
extent_prof_tctx_set(extent, NULL); extent_prof_tctx_set(extent, NULL);
qr_new(&extent->rd, rd_link); qr_new(extent, qr_link);
qr_new(extent, cc_link);
} }
JEMALLOC_INLINE void JEMALLOC_INLINE void
extent_dirty_insert(extent_t *extent, extent_ring_insert(extent_t *sentinel, extent_t *extent)
arena_runs_dirty_link_t *runs_dirty, extent_t *chunks_dirty)
{ {
qr_meld(runs_dirty, &extent->rd, rd_link); qr_meld(sentinel, extent, qr_link);
qr_meld(chunks_dirty, extent, cc_link);
} }
JEMALLOC_INLINE void JEMALLOC_INLINE void
extent_dirty_remove(extent_t *extent) extent_ring_remove(extent_t *extent)
{ {
qr_remove(&extent->rd, rd_link); qr_remove(extent, qr_link);
qr_remove(extent, cc_link);
} }
#endif #endif

45
include/jemalloc/internal/private_symbols.txt
View File

@ -5,8 +5,6 @@ arena_alloc_junk_small
arena_basic_stats_merge arena_basic_stats_merge
arena_bin_index arena_bin_index
arena_bin_info arena_bin_info
arena_bitselm_get_const
arena_bitselm_get_mutable
arena_boot arena_boot
arena_choose arena_choose
arena_choose_hard arena_choose_hard
@ -43,38 +41,11 @@ arena_lg_dirty_mult_get
arena_lg_dirty_mult_set arena_lg_dirty_mult_set
arena_malloc arena_malloc
arena_malloc_hard arena_malloc_hard
arena_mapbits_allocated_get
arena_mapbits_binind_get
arena_mapbits_decommitted_get
arena_mapbits_dirty_get
arena_mapbits_get
arena_mapbits_internal_set
arena_mapbits_large_binind_set
arena_mapbits_large_get
arena_mapbits_large_set
arena_mapbits_large_size_get
arena_mapbits_size_decode
arena_mapbits_size_encode
arena_mapbits_small_runind_get
arena_mapbits_small_set
arena_mapbits_unallocated_set
arena_mapbits_unallocated_size_get
arena_mapbits_unallocated_size_set
arena_mapbits_unzeroed_get
arena_mapbitsp_get_const
arena_mapbitsp_get_mutable
arena_mapbitsp_read
arena_mapbitsp_write
arena_maxrun
arena_maybe_purge arena_maybe_purge
arena_metadata_allocated_add arena_metadata_allocated_add
arena_metadata_allocated_get arena_metadata_allocated_get
arena_metadata_allocated_sub arena_metadata_allocated_sub
arena_migrate arena_migrate
arena_miscelm_get_const
arena_miscelm_get_mutable
arena_miscelm_to_pageind
arena_miscelm_to_rpages
arena_new arena_new
arena_nthreads_dec arena_nthreads_dec
arena_nthreads_get arena_nthreads_get
@ -93,14 +64,11 @@ arena_prof_promote
arena_prof_tctx_get arena_prof_tctx_get
arena_prof_tctx_reset arena_prof_tctx_reset
arena_prof_tctx_set arena_prof_tctx_set
arena_ptr_small_binind_get
arena_purge arena_purge
arena_ralloc arena_ralloc
arena_ralloc_junk_large arena_ralloc_junk_large
arena_ralloc_no_move arena_ralloc_no_move
arena_rd_to_miscelm
arena_reset arena_reset
arena_run_to_miscelm
arena_salloc arena_salloc
arena_sdalloc arena_sdalloc
arena_stats_merge arena_stats_merge
@ -213,22 +181,23 @@ extent_before_get
extent_committed_get extent_committed_get
extent_committed_set extent_committed_set
extent_dalloc extent_dalloc
extent_dirty_get
extent_dirty_insert
extent_dirty_remove
extent_dirty_set
extent_init extent_init
extent_last_get extent_last_get
extent_past_get extent_past_get
extent_prof_tctx_get extent_prof_tctx_get
extent_prof_tctx_set extent_prof_tctx_set
extent_retained_get extent_retained_get
extent_ring_insert
extent_ring_remove
extent_size_get extent_size_get
extent_size_set extent_size_set
extent_size_quantize_ceil extent_size_quantize_ceil
extent_size_quantize_floor extent_size_quantize_floor
extent_slab_data_get
extent_slab_data_get_const
extent_slab_get extent_slab_get
extent_slab_set extent_slab_set
extent_slab_data_get
extent_usize_get extent_usize_get
extent_zeroed_get extent_zeroed_get
extent_zeroed_set extent_zeroed_set
@ -309,8 +278,6 @@ malloc_tsd_no_cleanup
malloc_vcprintf malloc_vcprintf
malloc_vsnprintf malloc_vsnprintf
malloc_write malloc_write
map_bias
map_misc_offset
mb_write mb_write
narenas_auto narenas_auto
narenas_tdata_cleanup narenas_tdata_cleanup
@ -451,8 +418,6 @@ rtree_subtree_read
rtree_subtree_read_hard rtree_subtree_read_hard
rtree_subtree_tryread rtree_subtree_tryread
rtree_write rtree_write
run_quantize_ceil
run_quantize_floor
s2u s2u
s2u_compute s2u_compute
s2u_lookup s2u_lookup

44
include/jemalloc/internal/prof.h
View File

@ -335,8 +335,8 @@ prof_tctx_t *prof_tctx_get(tsdn_t *tsdn, const extent_t *extent,
void prof_tctx_set(tsdn_t *tsdn, extent_t *extent, const void *ptr, void prof_tctx_set(tsdn_t *tsdn, extent_t *extent, const void *ptr,
size_t usize, prof_tctx_t *tctx); size_t usize, prof_tctx_t *tctx);
void prof_tctx_reset(tsdn_t *tsdn, extent_t *extent, const void *ptr, void prof_tctx_reset(tsdn_t *tsdn, extent_t *extent, const void *ptr,
size_t usize, const void *old_ptr, prof_tctx_t *tctx); prof_tctx_t *tctx);
bool prof_sample_accum_update(tsd_t *tsd, size_t usize, bool commit, bool prof_sample_accum_update(tsd_t *tsd, size_t usize, bool update,
prof_tdata_t **tdata_out); prof_tdata_t **tdata_out);
prof_tctx_t *prof_alloc_prep(tsd_t *tsd, size_t usize, bool prof_active, prof_tctx_t *prof_alloc_prep(tsd_t *tsd, size_t usize, bool prof_active,
bool update); bool update);
@ -344,7 +344,8 @@ void prof_malloc(tsdn_t *tsdn, extent_t *extent, const void *ptr,
size_t usize, prof_tctx_t *tctx); size_t usize, prof_tctx_t *tctx);
void prof_realloc(tsd_t *tsd, extent_t *extent, const void *ptr, void prof_realloc(tsd_t *tsd, extent_t *extent, const void *ptr,
size_t usize, prof_tctx_t *tctx, bool prof_active, bool updated, size_t usize, prof_tctx_t *tctx, bool prof_active, bool updated,
const void *old_ptr, size_t old_usize, prof_tctx_t *old_tctx); extent_t *old_extent, const void *old_ptr, size_t old_usize,
prof_tctx_t *old_tctx);
void prof_free(tsd_t *tsd, const extent_t *extent, const void *ptr, void prof_free(tsd_t *tsd, const extent_t *extent, const void *ptr,
size_t usize); size_t usize);
#endif #endif
@ -421,14 +422,14 @@ prof_tctx_set(tsdn_t *tsdn, extent_t *extent, const void *ptr, size_t usize,
} }
JEMALLOC_ALWAYS_INLINE void JEMALLOC_ALWAYS_INLINE void
prof_tctx_reset(tsdn_t *tsdn, extent_t *extent, const void *ptr, size_t usize, prof_tctx_reset(tsdn_t *tsdn, extent_t *extent, const void *ptr,
const void *old_ptr, prof_tctx_t *old_tctx) prof_tctx_t *tctx)
{ {
cassert(config_prof); cassert(config_prof);
assert(ptr != NULL); assert(ptr != NULL);
arena_prof_tctx_reset(tsdn, extent, ptr, usize, old_ptr, old_tctx); arena_prof_tctx_reset(tsdn, extent, ptr, tctx);
} }
JEMALLOC_ALWAYS_INLINE bool JEMALLOC_ALWAYS_INLINE bool
@ -501,10 +502,10 @@ prof_malloc(tsdn_t *tsdn, extent_t *extent, const void *ptr, size_t usize,
JEMALLOC_ALWAYS_INLINE void JEMALLOC_ALWAYS_INLINE void
prof_realloc(tsd_t *tsd, extent_t *extent, const void *ptr, size_t usize, prof_realloc(tsd_t *tsd, extent_t *extent, const void *ptr, size_t usize,
prof_tctx_t *tctx, bool prof_active, bool updated, const void *old_ptr, prof_tctx_t *tctx, bool prof_active, bool updated, extent_t *old_extent,
size_t old_usize, prof_tctx_t *old_tctx) const void *old_ptr, size_t old_usize, prof_tctx_t *old_tctx)
{ {
bool sampled, old_sampled; bool sampled, old_sampled, moved;
cassert(config_prof); cassert(config_prof);
assert(ptr != NULL || (uintptr_t)tctx <= (uintptr_t)1U); assert(ptr != NULL || (uintptr_t)tctx <= (uintptr_t)1U);
@ -523,19 +524,30 @@ prof_realloc(tsd_t *tsd, extent_t *extent, const void *ptr, size_t usize,
} }
} }
/*
* The following code must differentiate among eight possible cases,
* based on three boolean conditions.
*/
sampled = ((uintptr_t)tctx > (uintptr_t)1U); sampled = ((uintptr_t)tctx > (uintptr_t)1U);
old_sampled = ((uintptr_t)old_tctx > (uintptr_t)1U); old_sampled = ((uintptr_t)old_tctx > (uintptr_t)1U);
moved = (ptr != old_ptr);
/*
* The following block must only execute if this is a non-moving
* reallocation, because for moving reallocation the old allocation will
* be deallocated via a separate call.
*/
if (unlikely(old_sampled) && !moved)
prof_free_sampled_object(tsd, old_usize, old_tctx);
if (unlikely(sampled)) { if (unlikely(sampled)) {
prof_malloc_sample_object(tsd_tsdn(tsd), extent, ptr, usize, prof_malloc_sample_object(tsd_tsdn(tsd), extent, ptr, usize,
tctx); tctx);
} else { } else if (moved) {
prof_tctx_reset(tsd_tsdn(tsd), extent, ptr, usize, old_ptr, prof_tctx_set(tsd_tsdn(tsd), extent, ptr, usize,
old_tctx); (prof_tctx_t *)(uintptr_t)1U);
} } else if (unlikely(old_sampled))
prof_tctx_reset(tsd_tsdn(tsd), extent, ptr, tctx);
if (unlikely(old_sampled))
prof_free_sampled_object(tsd, old_usize, old_tctx);
} }
JEMALLOC_ALWAYS_INLINE void JEMALLOC_ALWAYS_INLINE void

22
include/jemalloc/internal/size_classes.sh
View File

@ -50,7 +50,7 @@ reg_size_compute() {
reg_size=$((${grp} + ${delta}*${ndelta})) reg_size=$((${grp} + ${delta}*${ndelta}))
} }
run_size() { slab_size() {
lg_p=$1 lg_p=$1
lg_grp=$2 lg_grp=$2
lg_delta=$3 lg_delta=$3
@ -59,22 +59,22 @@ run_size() {
pow2 ${lg_p}; p=${pow2_result} pow2 ${lg_p}; p=${pow2_result}
reg_size_compute ${lg_grp} ${lg_delta} ${ndelta} reg_size_compute ${lg_grp} ${lg_delta} ${ndelta}
# Compute smallest run size that is an integer multiple of reg_size. # Compute smallest slab size that is an integer multiple of reg_size.
try_run_size=${p} try_slab_size=${p}
try_nregs=$((${try_run_size} / ${reg_size})) try_nregs=$((${try_slab_size} / ${reg_size}))
perfect=0 perfect=0
while [ ${perfect} -eq 0 ] ; do while [ ${perfect} -eq 0 ] ; do
perfect_run_size=${try_run_size} perfect_slab_size=${try_slab_size}
perfect_nregs=${try_nregs} perfect_nregs=${try_nregs}
try_run_size=$((${try_run_size} + ${p})) try_slab_size=$((${try_slab_size} + ${p}))
try_nregs=$((${try_run_size} / ${reg_size})) try_nregs=$((${try_slab_size} / ${reg_size}))
if [ ${perfect_run_size} -eq $((${perfect_nregs} * ${reg_size})) ] ; then if [ ${perfect_slab_size} -eq $((${perfect_nregs} * ${reg_size})) ] ; then
perfect=1 perfect=1
fi fi
done done
run_size_pgs=$((${perfect_run_size} / ${p})) slab_size_pgs=$((${perfect_slab_size} / ${p}))
} }
size_class() { size_class() {
@ -117,7 +117,7 @@ size_class() {
if [ ${lg_size} -lt $((${lg_p} + ${lg_g})) ] ; then if [ ${lg_size} -lt $((${lg_p} + ${lg_g})) ] ; then
bin="yes" bin="yes"
run_size ${lg_p} ${lg_grp} ${lg_delta} ${ndelta}; pgs=${run_size_pgs} slab_size ${lg_p} ${lg_grp} ${lg_delta} ${ndelta}; pgs=${slab_size_pgs}
else else
bin="no" bin="no"
pgs=0 pgs=0
@ -278,7 +278,7 @@ cat <<EOF
* ndelta: Delta multiplier. size == 1<<lg_grp + ndelta<<lg_delta * ndelta: Delta multiplier. size == 1<<lg_grp + ndelta<<lg_delta
* psz: 'yes' if a multiple of the page size, 'no' otherwise. * psz: 'yes' if a multiple of the page size, 'no' otherwise.
* bin: 'yes' if a small bin size class, 'no' otherwise. * bin: 'yes' if a small bin size class, 'no' otherwise.
* pgs: Run page count if a small bin size class, 0 otherwise. * pgs: Slab page count if a small bin size class, 0 otherwise.
* lg_delta_lookup: Same as lg_delta if a lookup table size class, 'no' * lg_delta_lookup: Same as lg_delta if a lookup table size class, 'no'
* otherwise. * otherwise.
* NTBINS: Number of tiny bins. * NTBINS: Number of tiny bins.

14
include/jemalloc/internal/stats.h
View File

@ -48,17 +48,17 @@ struct malloc_bin_stats_s {
/* Number of tcache flushes to this bin. */ /* Number of tcache flushes to this bin. */
uint64_t nflushes; uint64_t nflushes;
/* Total number of runs created for this bin's size class. */ /* Total number of slabs created for this bin's size class. */
uint64_t nruns; uint64_t nslabs;
/* /*
* Total number of runs reused by extracting them from the runs tree for * Total number of slabs reused by extracting them from the slabs heap
* this bin's size class. * for this bin's size class.
*/ */
uint64_t reruns; uint64_t reslabs;
/* Current number of runs in this bin. */ /* Current number of slabs in this bin. */
size_t curruns; size_t curslabs;
}; };
struct malloc_huge_stats_s { struct malloc_huge_stats_s {

2
include/jemalloc/internal/tcache.h
View File

@ -24,7 +24,7 @@ typedef struct tcaches_s tcaches_t;
/* /*
* Absolute maximum number of cache slots for each small bin in the thread * 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 * cache. This is an additional constraint beyond that imposed as: twice the
* number of regions per run for this size class. * number of regions per slab for this size class.
* *
* This constant must be an even number. * This constant must be an even number.
*/ */

1615
src/arena.c
View File

File diff suppressed because it is too large Load Diff

3
src/base.c
View File

@ -74,8 +74,7 @@ base_chunk_alloc(tsdn_t *tsdn, size_t minsize)
base_resident += PAGE_CEILING(nsize); base_resident += PAGE_CEILING(nsize);
} }
} }
extent_init(extent, NULL, addr, csize, 0, true, false, true, true, extent_init(extent, NULL, addr, csize, 0, true, true, true, false);
false);
return (extent); return (extent);
} }

11
src/chunk.c
View File

@ -558,8 +558,7 @@ chunk_alloc_wrapper_hard(tsdn_t *tsdn, arena_t *arena,
extent_dalloc(tsdn, arena, extent); extent_dalloc(tsdn, arena, extent);
return (NULL); return (NULL);
} }
extent_init(extent, arena, addr, size, usize, true, false, zero, commit, extent_init(extent, arena, addr, size, usize, true, zero, commit, slab);
slab);
if (pad != 0) if (pad != 0)
extent_addr_randomize(tsdn, extent, alignment); extent_addr_randomize(tsdn, extent, alignment);
if (chunk_register(tsdn, extent)) { if (chunk_register(tsdn, extent)) {
@ -828,8 +827,8 @@ chunk_split_wrapper(tsdn_t *tsdn, arena_t *arena, chunk_hooks_t *chunk_hooks,
extent_init(&lead, arena, extent_addr_get(extent), size_a, extent_init(&lead, arena, extent_addr_get(extent), size_a,
usize_a, extent_active_get(extent), usize_a, extent_active_get(extent),
extent_dirty_get(extent), extent_zeroed_get(extent), extent_zeroed_get(extent), extent_committed_get(extent),
extent_committed_get(extent), extent_slab_get(extent)); extent_slab_get(extent));
if (extent_rtree_acquire(tsdn, &lead, false, true, &lead_elm_a, if (extent_rtree_acquire(tsdn, &lead, false, true, &lead_elm_a,
&lead_elm_b)) &lead_elm_b))
@ -838,8 +837,8 @@ chunk_split_wrapper(tsdn_t *tsdn, arena_t *arena, chunk_hooks_t *chunk_hooks,
extent_init(trail, arena, (void *)((uintptr_t)extent_base_get(extent) + extent_init(trail, arena, (void *)((uintptr_t)extent_base_get(extent) +
size_a), size_b, usize_b, extent_active_get(extent), size_a), size_b, usize_b, extent_active_get(extent),
extent_dirty_get(extent), extent_zeroed_get(extent), extent_zeroed_get(extent), extent_committed_get(extent),
extent_committed_get(extent), extent_slab_get(extent)); extent_slab_get(extent));
if (extent_rtree_acquire(tsdn, trail, false, true, &trail_elm_a, if (extent_rtree_acquire(tsdn, trail, false, true, &trail_elm_a,
&trail_elm_b)) &trail_elm_b))
goto label_error_c; goto label_error_c;

2
src/chunk_dss.c
View File

@ -121,7 +121,7 @@ chunk_alloc_dss(tsdn_t *tsdn, arena_t *arena, void *new_addr, size_t size,
pad_size = (uintptr_t)ret - (uintptr_t)pad_addr; pad_size = (uintptr_t)ret - (uintptr_t)pad_addr;
if (pad_size != 0) { if (pad_size != 0) {
extent_init(pad, arena, pad_addr, pad_size, extent_init(pad, arena, pad_addr, pad_size,
pad_size, false, true, false, true, false); pad_size, false, false, true, false);
} }
dss_next = (void *)((uintptr_t)ret + size); dss_next = (void *)((uintptr_t)ret + size);
if ((uintptr_t)ret < (uintptr_t)dss_max || if ((uintptr_t)ret < (uintptr_t)dss_max ||

37
src/ctl.c
View File

@ -124,7 +124,7 @@ CTL_PROTO(arena_i_chunk_hooks)
INDEX_PROTO(arena_i) INDEX_PROTO(arena_i)
CTL_PROTO(arenas_bin_i_size) CTL_PROTO(arenas_bin_i_size)
CTL_PROTO(arenas_bin_i_nregs) CTL_PROTO(arenas_bin_i_nregs)
CTL_PROTO(arenas_bin_i_run_size) CTL_PROTO(arenas_bin_i_slab_size)
INDEX_PROTO(arenas_bin_i) INDEX_PROTO(arenas_bin_i)
CTL_PROTO(arenas_hchunk_i_size) CTL_PROTO(arenas_hchunk_i_size)
INDEX_PROTO(arenas_hchunk_i) INDEX_PROTO(arenas_hchunk_i)
@ -160,9 +160,9 @@ CTL_PROTO(stats_arenas_i_bins_j_nrequests)
CTL_PROTO(stats_arenas_i_bins_j_curregs) CTL_PROTO(stats_arenas_i_bins_j_curregs)
CTL_PROTO(stats_arenas_i_bins_j_nfills) CTL_PROTO(stats_arenas_i_bins_j_nfills)
CTL_PROTO(stats_arenas_i_bins_j_nflushes) CTL_PROTO(stats_arenas_i_bins_j_nflushes)
CTL_PROTO(stats_arenas_i_bins_j_nruns) CTL_PROTO(stats_arenas_i_bins_j_nslabs)
CTL_PROTO(stats_arenas_i_bins_j_nreruns) CTL_PROTO(stats_arenas_i_bins_j_nreslabs)
CTL_PROTO(stats_arenas_i_bins_j_curruns) CTL_PROTO(stats_arenas_i_bins_j_curslabs)
INDEX_PROTO(stats_arenas_i_bins_j) INDEX_PROTO(stats_arenas_i_bins_j)
CTL_PROTO(stats_arenas_i_hchunks_j_nmalloc) CTL_PROTO(stats_arenas_i_hchunks_j_nmalloc)
CTL_PROTO(stats_arenas_i_hchunks_j_ndalloc) CTL_PROTO(stats_arenas_i_hchunks_j_ndalloc)
@ -300,7 +300,7 @@ static const ctl_indexed_node_t arena_node[] = {
static const ctl_named_node_t arenas_bin_i_node[] = { static const ctl_named_node_t arenas_bin_i_node[] = {
{NAME("size"), CTL(arenas_bin_i_size)}, {NAME("size"), CTL(arenas_bin_i_size)},
{NAME("nregs"), CTL(arenas_bin_i_nregs)}, {NAME("nregs"), CTL(arenas_bin_i_nregs)},
{NAME("run_size"), CTL(arenas_bin_i_run_size)} {NAME("slab_size"), CTL(arenas_bin_i_slab_size)}
}; };
static const ctl_named_node_t super_arenas_bin_i_node[] = { static const ctl_named_node_t super_arenas_bin_i_node[] = {
{NAME(""), CHILD(named, arenas_bin_i)} {NAME(""), CHILD(named, arenas_bin_i)}
@ -373,9 +373,9 @@ static const ctl_named_node_t stats_arenas_i_bins_j_node[] = {
{NAME("curregs"), CTL(stats_arenas_i_bins_j_curregs)}, {NAME("curregs"), CTL(stats_arenas_i_bins_j_curregs)},
{NAME("nfills"), CTL(stats_arenas_i_bins_j_nfills)}, {NAME("nfills"), CTL(stats_arenas_i_bins_j_nfills)},
{NAME("nflushes"), CTL(stats_arenas_i_bins_j_nflushes)}, {NAME("nflushes"), CTL(stats_arenas_i_bins_j_nflushes)},
{NAME("nruns"), CTL(stats_arenas_i_bins_j_nruns)}, {NAME("nslabs"), CTL(stats_arenas_i_bins_j_nslabs)},
{NAME("nreruns"), CTL(stats_arenas_i_bins_j_nreruns)}, {NAME("nreslabs"), CTL(stats_arenas_i_bins_j_nreslabs)},
{NAME("curruns"), CTL(stats_arenas_i_bins_j_curruns)} {NAME("curslabs"), CTL(stats_arenas_i_bins_j_curslabs)}
}; };
static const ctl_named_node_t super_stats_arenas_i_bins_j_node[] = { static const ctl_named_node_t super_stats_arenas_i_bins_j_node[] = {
{NAME(""), CHILD(named, stats_arenas_i_bins_j)} {NAME(""), CHILD(named, stats_arenas_i_bins_j)}
@ -549,9 +549,10 @@ ctl_arena_stats_smerge(ctl_arena_stats_t *sstats, ctl_arena_stats_t *astats)
sstats->bstats[i].nflushes += sstats->bstats[i].nflushes +=
astats->bstats[i].nflushes; astats->bstats[i].nflushes;
} }
sstats->bstats[i].nruns += astats->bstats[i].nruns; sstats->bstats[i].nslabs += astats->bstats[i].nslabs;
sstats->bstats[i].reruns += astats->bstats[i].reruns; sstats->bstats[i].reslabs += astats->bstats[i].reslabs;
sstats->bstats[i].curruns += astats->bstats[i].curruns; sstats->bstats[i].curslabs +=
astats->bstats[i].curslabs;
} }
for (i = 0; i < NSIZES - NBINS; i++) { for (i = 0; i < NSIZES - NBINS; i++) {
@ -1801,7 +1802,7 @@ CTL_RO_NL_GEN(arenas_nbins, NBINS, unsigned)
CTL_RO_NL_CGEN(config_tcache, arenas_nhbins, nhbins, unsigned) CTL_RO_NL_CGEN(config_tcache, arenas_nhbins, nhbins, unsigned)
CTL_RO_NL_GEN(arenas_bin_i_size, arena_bin_info[mib[2]].reg_size, size_t) 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_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) CTL_RO_NL_GEN(arenas_bin_i_slab_size, arena_bin_info[mib[2]].slab_size, size_t)
static const ctl_named_node_t * static const ctl_named_node_t *
arenas_bin_i_index(tsdn_t *tsdn, const size_t *mib, size_t miblen, size_t i) arenas_bin_i_index(tsdn_t *tsdn, const size_t *mib, size_t miblen, size_t i)
{ {
@ -2032,12 +2033,12 @@ 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_stats.arenas[mib[2]].bstats[mib[4]].nfills, uint64_t)
CTL_RO_CGEN(config_stats && config_tcache, 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) ctl_stats.arenas[mib[2]].bstats[mib[4]].nflushes, uint64_t)
CTL_RO_CGEN(config_stats, stats_arenas_i_bins_j_nruns, CTL_RO_CGEN(config_stats, stats_arenas_i_bins_j_nslabs,
ctl_stats.arenas[mib[2]].bstats[mib[4]].nruns, uint64_t) ctl_stats.arenas[mib[2]].bstats[mib[4]].nslabs, uint64_t)
CTL_RO_CGEN(config_stats, stats_arenas_i_bins_j_nreruns, CTL_RO_CGEN(config_stats, stats_arenas_i_bins_j_nreslabs,
ctl_stats.arenas[mib[2]].bstats[mib[4]].reruns, uint64_t) ctl_stats.arenas[mib[2]].bstats[mib[4]].reslabs, uint64_t)
CTL_RO_CGEN(config_stats, stats_arenas_i_bins_j_curruns, CTL_RO_CGEN(config_stats, stats_arenas_i_bins_j_curslabs,
ctl_stats.arenas[mib[2]].bstats[mib[4]].curruns, size_t) ctl_stats.arenas[mib[2]].bstats[mib[4]].curslabs, size_t)
static const ctl_named_node_t * static const ctl_named_node_t *
stats_arenas_i_bins_j_index(tsdn_t *tsdn, const size_t *mib, size_t miblen, stats_arenas_i_bins_j_index(tsdn_t *tsdn, const size_t *mib, size_t miblen,

4
src/huge.c
View File

@ -153,8 +153,8 @@ huge_ralloc_no_move_expand(tsdn_t *tsdn, extent_t *extent, size_t usize,
* Zero the trailing bytes of the original allocation's * Zero the trailing bytes of the original allocation's
* last page, since they are in an indeterminate state. * last page, since they are in an indeterminate state.
* There will always be trailing bytes, because ptr's * There will always be trailing bytes, because ptr's
* offset from the beginning of the run is a multiple of * offset from the beginning of the extent is a multiple
* CACHELINE in [0 .. PAGE). * of CACHELINE in [0 .. PAGE).
*/ */
void *zbase = (void *) void *zbase = (void *)
((uintptr_t)extent_addr_get(extent) + oldusize); ((uintptr_t)extent_addr_get(extent) + oldusize);

46
src/jemalloc.c
View File

@ -1707,28 +1707,30 @@ irealloc_prof_sample(tsd_t *tsd, extent_t *extent, void *old_ptr,
} }
JEMALLOC_ALWAYS_INLINE_C void * JEMALLOC_ALWAYS_INLINE_C void *
irealloc_prof(tsd_t *tsd, extent_t *extent, void *old_ptr, size_t old_usize, irealloc_prof(tsd_t *tsd, extent_t *old_extent, void *old_ptr, size_t old_usize,
size_t usize) size_t usize)
{ {
void *p; void *p;
extent_t *e; extent_t *extent;
bool prof_active; bool prof_active;
prof_tctx_t *old_tctx, *tctx; prof_tctx_t *old_tctx, *tctx;
prof_active = prof_active_get_unlocked(); prof_active = prof_active_get_unlocked();
old_tctx = prof_tctx_get(tsd_tsdn(tsd), extent, old_ptr); old_tctx = prof_tctx_get(tsd_tsdn(tsd), old_extent, old_ptr);
tctx = prof_alloc_prep(tsd, usize, prof_active, true); tctx = prof_alloc_prep(tsd, usize, prof_active, true);
if (unlikely((uintptr_t)tctx != (uintptr_t)1U)) { if (unlikely((uintptr_t)tctx != (uintptr_t)1U)) {
p = irealloc_prof_sample(tsd, extent, old_ptr, old_usize, usize, p = irealloc_prof_sample(tsd, old_extent, old_ptr, old_usize,
tctx); usize, tctx);
} else } else {
p = iralloc(tsd, extent, old_ptr, old_usize, usize, 0, false); p = iralloc(tsd, old_extent, old_ptr, old_usize, usize, 0,
false);
}
if (unlikely(p == NULL)) { if (unlikely(p == NULL)) {
prof_alloc_rollback(tsd, tctx, true); prof_alloc_rollback(tsd, tctx, true);
return (NULL); return (NULL);
} }
e = (p == old_ptr) ? extent : iealloc(tsd_tsdn(tsd), p); extent = (p == old_ptr) ? old_extent : iealloc(tsd_tsdn(tsd), p);
prof_realloc(tsd, e, p, usize, tctx, prof_active, true, prof_realloc(tsd, extent, p, usize, tctx, prof_active, true, old_extent,
old_ptr, old_usize, old_tctx); old_ptr, old_usize, old_tctx);
return (p); return (p);
@ -2146,24 +2148,24 @@ irallocx_prof_sample(tsdn_t *tsdn, extent_t *extent, void *old_ptr,
} }
JEMALLOC_ALWAYS_INLINE_C void * JEMALLOC_ALWAYS_INLINE_C void *
irallocx_prof(tsd_t *tsd, extent_t *extent, void *old_ptr, size_t old_usize, irallocx_prof(tsd_t *tsd, extent_t *old_extent, void *old_ptr, size_t old_usize,
size_t size, size_t alignment, size_t *usize, bool zero, tcache_t *tcache, size_t size, size_t alignment, size_t *usize, bool zero, tcache_t *tcache,
arena_t *arena) arena_t *arena)
{ {
void *p; void *p;
extent_t *e; extent_t *extent;
bool prof_active; bool prof_active;
prof_tctx_t *old_tctx, *tctx; prof_tctx_t *old_tctx, *tctx;
prof_active = prof_active_get_unlocked(); prof_active = prof_active_get_unlocked();
old_tctx = prof_tctx_get(tsd_tsdn(tsd), extent, old_ptr); old_tctx = prof_tctx_get(tsd_tsdn(tsd), old_extent, old_ptr);
tctx = prof_alloc_prep(tsd, *usize, prof_active, true); tctx = prof_alloc_prep(tsd, *usize, prof_active, true);
if (unlikely((uintptr_t)tctx != (uintptr_t)1U)) { if (unlikely((uintptr_t)tctx != (uintptr_t)1U)) {
p = irallocx_prof_sample(tsd_tsdn(tsd), extent, old_ptr, p = irallocx_prof_sample(tsd_tsdn(tsd), old_extent, old_ptr,
old_usize, *usize, alignment, zero, tcache, arena, tctx); old_usize, *usize, alignment, zero, tcache, arena, tctx);
} else { } else {
p = iralloct(tsd_tsdn(tsd), extent, old_ptr, old_usize, size, p = iralloct(tsd_tsdn(tsd), old_extent, old_ptr, old_usize,
alignment, zero, tcache, arena); size, alignment, zero, tcache, arena);
} }
if (unlikely(p == NULL)) { if (unlikely(p == NULL)) {
prof_alloc_rollback(tsd, tctx, true); prof_alloc_rollback(tsd, tctx, true);
@ -2179,12 +2181,12 @@ irallocx_prof(tsd_t *tsd, extent_t *extent, void *old_ptr, size_t old_usize,
* be the same as the current usize because of in-place large * be the same as the current usize because of in-place large
* reallocation. Therefore, query the actual value of usize. * reallocation. Therefore, query the actual value of usize.
*/ */
e = extent; extent = old_extent;
*usize = isalloc(tsd_tsdn(tsd), e, p); *usize = isalloc(tsd_tsdn(tsd), extent, p);
} else } else
e = iealloc(tsd_tsdn(tsd), p); extent = iealloc(tsd_tsdn(tsd), p);
prof_realloc(tsd, e, p, *usize, tctx, prof_active, true, old_ptr, prof_realloc(tsd, extent, p, *usize, tctx, prof_active, true,
old_usize, old_tctx); old_extent, old_ptr, old_usize, old_tctx);
return (p); return (p);
} }
@ -2338,8 +2340,8 @@ ixallocx_prof(tsd_t *tsd, extent_t *extent, void *ptr, size_t old_usize,
prof_alloc_rollback(tsd, tctx, false); prof_alloc_rollback(tsd, tctx, false);
return (usize); return (usize);
} }
prof_realloc(tsd, extent, ptr, usize, tctx, prof_active, false, ptr, prof_realloc(tsd, extent, ptr, usize, tctx, prof_active, false, extent,
old_usize, old_tctx); ptr, old_usize, old_tctx);
return (usize); return (usize);
} }

45
src/stats.c
View File

@ -58,29 +58,29 @@ stats_arena_bins_print(void (*write_cb)(void *, const char *), void *cbopaque,
if (config_tcache) { if (config_tcache) {
malloc_cprintf(write_cb, cbopaque, malloc_cprintf(write_cb, cbopaque,
"bins: size ind allocated nmalloc" "bins: size ind allocated nmalloc"
" ndalloc nrequests curregs curruns regs" " ndalloc nrequests curregs curslabs regs"
" pgs util nfills nflushes newruns" " pgs util nfills nflushes newslabs"
" reruns\n"); " reslabs\n");
} else { } else {
malloc_cprintf(write_cb, cbopaque, malloc_cprintf(write_cb, cbopaque,
"bins: size ind allocated nmalloc" "bins: size ind allocated nmalloc"
" ndalloc nrequests curregs curruns regs" " ndalloc nrequests curregs curslabs regs"
" pgs util newruns reruns\n"); " pgs util newslabs reslabs\n");
} }
CTL_GET("arenas.nbins", &nbins, unsigned); CTL_GET("arenas.nbins", &nbins, unsigned);
for (j = 0, in_gap = false; j < nbins; j++) { for (j = 0, in_gap = false; j < nbins; j++) {
uint64_t nruns; uint64_t nslabs;
CTL_M2_M4_GET("stats.arenas.0.bins.0.nruns", i, j, &nruns, CTL_M2_M4_GET("stats.arenas.0.bins.0.nslabs", i, j, &nslabs,
uint64_t); uint64_t);
if (nruns == 0) if (nslabs == 0)
in_gap = true; in_gap = true;
else { else {
size_t reg_size, run_size, curregs, availregs, milli; size_t reg_size, slab_size, curregs, availregs, milli;
size_t curruns; size_t curslabs;
uint32_t nregs; uint32_t nregs;
uint64_t nmalloc, ndalloc, nrequests, nfills, nflushes; uint64_t nmalloc, ndalloc, nrequests, nfills, nflushes;
uint64_t reruns; uint64_t reslabs;
char util[6]; /* "x.yyy". */ char util[6]; /* "x.yyy". */
if (in_gap) { if (in_gap) {
@ -90,7 +90,7 @@ stats_arena_bins_print(void (*write_cb)(void *, const char *), void *cbopaque,
} }
CTL_M2_GET("arenas.bin.0.size", j, &reg_size, size_t); CTL_M2_GET("arenas.bin.0.size", j, &reg_size, size_t);
CTL_M2_GET("arenas.bin.0.nregs", j, &nregs, uint32_t); CTL_M2_GET("arenas.bin.0.nregs", j, &nregs, uint32_t);
CTL_M2_GET("arenas.bin.0.run_size", j, &run_size, CTL_M2_GET("arenas.bin.0.slab_size", j, &slab_size,
size_t); size_t);
CTL_M2_M4_GET("stats.arenas.0.bins.0.nmalloc", i, j, CTL_M2_M4_GET("stats.arenas.0.bins.0.nmalloc", i, j,
&nmalloc, uint64_t); &nmalloc, uint64_t);
@ -106,12 +106,12 @@ stats_arena_bins_print(void (*write_cb)(void *, const char *), void *cbopaque,
CTL_M2_M4_GET("stats.arenas.0.bins.0.nflushes", CTL_M2_M4_GET("stats.arenas.0.bins.0.nflushes",
i, j, &nflushes, uint64_t); i, j, &nflushes, uint64_t);
} }
CTL_M2_M4_GET("stats.arenas.0.bins.0.nreruns", i, j, CTL_M2_M4_GET("stats.arenas.0.bins.0.nreslabs", i, j,
&reruns, uint64_t); &reslabs, uint64_t);
CTL_M2_M4_GET("stats.arenas.0.bins.0.curruns", i, j, CTL_M2_M4_GET("stats.arenas.0.bins.0.curslabs", i, j,
&curruns, size_t); &curslabs, size_t);
availregs = nregs * curruns; availregs = nregs * curslabs;
milli = (availregs != 0) ? (1000 * curregs) / availregs milli = (availregs != 0) ? (1000 * curregs) / availregs
: 1000; : 1000;
assert(milli <= 1000); assert(milli <= 1000);
@ -134,9 +134,9 @@ stats_arena_bins_print(void (*write_cb)(void *, const char *), void *cbopaque,
" %12zu %4u %3zu %-5s %12"FMTu64 " %12zu %4u %3zu %-5s %12"FMTu64
" %12"FMTu64" %12"FMTu64" %12"FMTu64"\n", " %12"FMTu64" %12"FMTu64" %12"FMTu64"\n",
reg_size, j, curregs * reg_size, nmalloc, reg_size, j, curregs * reg_size, nmalloc,
ndalloc, nrequests, curregs, curruns, nregs, ndalloc, nrequests, curregs, curslabs,
run_size / page, util, nfills, nflushes, nregs, slab_size / page, util, nfills,
nruns, reruns); nflushes, nslabs, reslabs);
} else { } else {
malloc_cprintf(write_cb, cbopaque, malloc_cprintf(write_cb, cbopaque,
"%20zu %3u %12zu %12"FMTu64 "%20zu %3u %12zu %12"FMTu64
@ -144,8 +144,9 @@ stats_arena_bins_print(void (*write_cb)(void *, const char *), void *cbopaque,
" %12zu %4u %3zu %-5s %12"FMTu64 " %12zu %4u %3zu %-5s %12"FMTu64
" %12"FMTu64"\n", " %12"FMTu64"\n",
reg_size, j, curregs * reg_size, nmalloc, reg_size, j, curregs * reg_size, nmalloc,
ndalloc, nrequests, curregs, curruns, nregs, ndalloc, nrequests, curregs, curslabs,
run_size / page, util, nruns, reruns); nregs, slab_size / page, util, nslabs,
reslabs);
} }
} }
} }

8
src/tcache.c
View File

@ -127,14 +127,8 @@ tcache_bin_flush_small(tsd_t *tsd, tcache_t *tcache, tcache_bin_t *tbin,
extent = iealloc(tsd_tsdn(tsd), ptr); extent = iealloc(tsd_tsdn(tsd), ptr);
if (extent_arena_get(extent) == bin_arena) { if (extent_arena_get(extent) == bin_arena) {
arena_chunk_t *chunk =
(arena_chunk_t *)extent_base_get(extent);
size_t pageind = ((uintptr_t)ptr -
(uintptr_t)chunk) >> LG_PAGE;
arena_chunk_map_bits_t *bitselm =
arena_bitselm_get_mutable(chunk, pageind);
arena_dalloc_bin_junked_locked(tsd_tsdn(tsd), arena_dalloc_bin_junked_locked(tsd_tsdn(tsd),
bin_arena, chunk, extent, ptr, bitselm); bin_arena, extent, ptr);
} else { } else {
/* /*
* This object was allocated via a different * This object was allocated via a different

10
test/unit/extent_quantize.c
View File

@ -16,7 +16,7 @@ TEST_BEGIN(test_small_extent_size)
assert_d_eq(mallctl("arenas.nbins", &nbins, &sz, NULL, 0), 0, assert_d_eq(mallctl("arenas.nbins", &nbins, &sz, NULL, 0), 0,
"Unexpected mallctl failure"); "Unexpected mallctl failure");
assert_d_eq(mallctlnametomib("arenas.bin.0.run_size", mib, &miblen), 0, assert_d_eq(mallctlnametomib("arenas.bin.0.slab_size", mib, &miblen), 0,
"Unexpected mallctlnametomib failure"); "Unexpected mallctlnametomib failure");
for (i = 0; i < nbins; i++) { for (i = 0; i < nbins; i++) {
mib[2] = i; mib[2] = i;
@ -71,12 +71,12 @@ TEST_BEGIN(test_huge_extent_size)
ceil = extent_size_quantize_ceil(extent_size); ceil = extent_size_quantize_ceil(extent_size);
assert_zu_eq(extent_size, floor, assert_zu_eq(extent_size, floor,
"Large run quantization should be a no-op for precise " "Extent quantization should be a no-op for precise size "
"size (lextent_size=%zu, extent_size=%zu)", lextent_size, "(lextent_size=%zu, extent_size=%zu)", lextent_size,
extent_size); extent_size);
assert_zu_eq(extent_size, ceil, assert_zu_eq(extent_size, ceil,
"Large run quantization should be a no-op for precise " "Extent quantization should be a no-op for precise size "
"size (lextent_size=%zu, extent_size=%zu)", lextent_size, "(lextent_size=%zu, extent_size=%zu)", lextent_size,
extent_size); extent_size);
if (i > 0) { if (i > 0) {

3
test/unit/mallctl.c
View File

@ -615,7 +615,8 @@ TEST_BEGIN(test_arenas_bin_constants)
TEST_ARENAS_BIN_CONSTANT(size_t, size, arena_bin_info[0].reg_size); TEST_ARENAS_BIN_CONSTANT(size_t, size, arena_bin_info[0].reg_size);
TEST_ARENAS_BIN_CONSTANT(uint32_t, nregs, arena_bin_info[0].nregs); TEST_ARENAS_BIN_CONSTANT(uint32_t, nregs, arena_bin_info[0].nregs);
TEST_ARENAS_BIN_CONSTANT(size_t, run_size, arena_bin_info[0].run_size); TEST_ARENAS_BIN_CONSTANT(size_t, slab_size,
arena_bin_info[0].slab_size);
#undef TEST_ARENAS_BIN_CONSTANT #undef TEST_ARENAS_BIN_CONSTANT
} }

18
test/unit/stats.c
View File

@ -229,9 +229,9 @@ TEST_BEGIN(test_stats_arenas_bins)
{ {
unsigned arena; unsigned arena;
void *p; void *p;
size_t sz, curruns, curregs; size_t sz, curslabs, curregs;
uint64_t epoch, nmalloc, ndalloc, nrequests, nfills, nflushes; uint64_t epoch, nmalloc, ndalloc, nrequests, nfills, nflushes;
uint64_t nruns, nreruns; uint64_t nslabs, nreslabs;
int expected = config_stats ? 0 : ENOENT; int expected = config_stats ? 0 : ENOENT;
arena = 0; arena = 0;
@ -266,12 +266,12 @@ TEST_BEGIN(test_stats_arenas_bins)
NULL, 0), config_tcache ? expected : ENOENT, NULL, 0), config_tcache ? expected : ENOENT,
"Unexpected mallctl() result"); "Unexpected mallctl() result");
assert_d_eq(mallctl("stats.arenas.0.bins.0.nruns", &nruns, &sz, assert_d_eq(mallctl("stats.arenas.0.bins.0.nslabs", &nslabs, &sz,
NULL, 0), expected, "Unexpected mallctl() result"); NULL, 0), expected, "Unexpected mallctl() result");
assert_d_eq(mallctl("stats.arenas.0.bins.0.nreruns", &nreruns, &sz, assert_d_eq(mallctl("stats.arenas.0.bins.0.nreslabs", &nreslabs, &sz,
NULL, 0), expected, "Unexpected mallctl() result"); NULL, 0), expected, "Unexpected mallctl() result");
sz = sizeof(size_t); sz = sizeof(size_t);
assert_d_eq(mallctl("stats.arenas.0.bins.0.curruns", &curruns, &sz, assert_d_eq(mallctl("stats.arenas.0.bins.0.curslabs", &curslabs, &sz,
NULL, 0), expected, "Unexpected mallctl() result"); NULL, 0), expected, "Unexpected mallctl() result");
if (config_stats) { if (config_stats) {
@ -289,10 +289,10 @@ TEST_BEGIN(test_stats_arenas_bins)
assert_u64_gt(nflushes, 0, assert_u64_gt(nflushes, 0,
"At least one flush should have occurred"); "At least one flush should have occurred");
} }
assert_u64_gt(nruns, 0, assert_u64_gt(nslabs, 0,
"At least one run should have been allocated"); "At least one slab should have been allocated");
assert_zu_gt(curruns, 0, assert_zu_gt(curslabs, 0,
"At least one run should be currently allocated"); "At least one slab should be currently allocated");
} }
dallocx(p, 0); dallocx(p, 0);