Merge FreeBSD's malloc(3) and posix_memalign(3) manpages to create jemalloc(3).

This commit is contained in:
Jason Evans 2009-06-22 17:44:39 -07:00
parent 804c9ec3af
commit 4450b830b6

589
jemalloc/doc/jemalloc.3 Normal file
View File

@ -0,0 +1,589 @@
.\" Copyright (c) 2006-2008 Jason Evans <jasone@canonware.com>.
.\" Copyright (c) 2009 Facebook, Inc. All rights reserved.
.\" All rights reserved.
.\" Copyright (c) 1980, 1991, 1993
.\" The Regents of the University of California. All rights reserved.
.\"
.\" This code is derived from software contributed to Berkeley by
.\" the American National Standards Committee X3, on Information
.\" Processing Systems.
.\"
.\" 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. Neither the name of the University 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 REGENTS 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 REGENTS 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.
.\"
.\" @(#)malloc.3 8.1 (Berkeley) 6/4/93
.\" $FreeBSD: head/lib/libc/stdlib/malloc.3 182225 2008-08-27 02:00:53Z jasone $
.\"
.Dd June 22, 2009
.Dt JEMALLOC 3
.Os
.Sh NAME
.Nm malloc , calloc , posix_memalign , realloc , free , malloc_usable_size
.Nd general purpose memory allocation functions
.Sh LIBRARY
.Lb libc
.Sh SYNOPSIS
.In stdlib.h
.Ft void *
.Fn malloc "size_t size"
.Ft void *
.Fn calloc "size_t number" "size_t size"
.Ft int
.Fn posix_memalign "void **ptr" "size_t alignment" "size_t size"
.Ft void *
.Fn realloc "void *ptr" "size_t size"
.Ft void
.Fn free "void *ptr"
.Ft const char *
.Va jemalloc_options ;
.Ft void
.Fo \*(lp*jemalloc_message\*(rp
.Fa "const char *p1" "const char *p2" "const char *p3" "const char *p4"
.Fc
.In malloc_np.h
.Ft size_t
.Fn malloc_usable_size "const void *ptr"
.Sh DESCRIPTION
The
.Fn malloc
function allocates
.Fa size
bytes of uninitialized memory.
The allocated space is suitably aligned (after possible pointer coercion)
for storage of any type of object.
.Pp
The
.Fn calloc
function allocates space for
.Fa number
objects,
each
.Fa size
bytes in length.
The result is identical to calling
.Fn malloc
with an argument of
.Dq "number * size" ,
with the exception that the allocated memory is explicitly initialized
to zero bytes.
.Pp
The
.Fn posix_memalign
function allocates
.Fa size
bytes of memory such that the allocation's base address is an even multiple of
.Fa alignment ,
and returns the allocation in the value pointed to by
.Fa ptr .
The requested
.Fa alignment
must be a power of 2 at least as large as
.Fn sizeof "void *" .
.Pp
The
.Fn realloc
function changes the size of the previously allocated memory referenced by
.Fa ptr
to
.Fa size
bytes.
The contents of the memory are unchanged up to the lesser of the new and
old sizes.
If the new size is larger,
the contents of the newly allocated portion of the memory are undefined.
Upon success, the memory referenced by
.Fa ptr
is freed and a pointer to the newly allocated memory is returned.
Note that
.Fn realloc
may move the memory allocation, resulting in a different return value than
.Fa ptr .
If
.Fa ptr
is
.Dv NULL ,
the
.Fn realloc
function behaves identically to
.Fn malloc
for the specified size.
.Pp
The
.Fn free
function causes the allocated memory referenced by
.Fa ptr
to be made available for future allocations.
If
.Fa ptr
is
.Dv NULL ,
no action occurs.
.Pp
The
.Fn malloc_usable_size
function returns the usable size of the allocation pointed to by
.Fa ptr .
The return value may be larger than the size that was requested during
allocation.
The
.Fn malloc_usable_size
function is not a mechanism for in-place
.Fn realloc ;
rather it is provided solely as a tool for introspection purposes.
Any discrepancy between the requested allocation size and the size reported by
.Fn malloc_usable_size
should not be depended on, since such behavior is entirely
implementation-dependent.
.Sh TUNING
Once, when the first call is made to one of these memory allocation
routines, various flags will be set or reset, which affects the
workings of this allocator implementation.
.Pp
The
.Dq name
of the file referenced by the symbolic link named
.Pa /etc/jemalloc.conf ,
the value of the environment variable
.Ev JEMALLOC_OPTIONS ,
and the string pointed to by the global variable
.Va jemalloc_options
will be interpreted, in that order, from left to right as flags.
.Pp
Each flag is a single letter, optionally prefixed by a non-negative base 10
integer repetition count.
For example,
.Dq 3N
is equivalent to
.Dq NNN .
Some flags control parameter magnitudes, where uppercase increases the
magnitude, and lowercase decreases the magnitude.
Other flags control boolean parameters, where uppercase indicates that a
behavior is set, or on, and lowercase means that a behavior is not set, or off.
.Bl -tag -width indent
.It A
All warnings (except for the warning about unknown
flags being set) become fatal.
The process will call
.Xr abort 3
in these cases.
.It B
Double/halve the per-arena lock contention threshold at which a thread is
randomly re-assigned to an arena.
This dynamic load balancing tends to push threads away from highly contended
arenas, which avoids worst case contention scenarios in which threads
disproportionately utilize arenas.
However, due to the highly dynamic load that applications may place on the
allocator, it is impossible for the allocator to know in advance how sensitive
it should be to contention over arenas.
Therefore, some applications may benefit from increasing or decreasing this
threshold parameter.
This option is not available for some configurations (non-PIC).
.It C
Double/halve the size 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.
.It D
Use
.Xr sbrk 2
to acquire memory in the data storage segment (DSS).
This option is enabled by default.
See the
.Dq M
option for related information and interactions.
.It F
Double/halve the per-arena maximum number of dirty unused pages that are
allowed to accumulate before informing the kernel about at least half of those
pages via
.Xr madvise 2 .
This provides the kernel with sufficient information to recycle dirty pages if
physical memory becomes scarce and the pages remain unused.
The default is 512 pages per arena;
.Ev JEMALLOC_OPTIONS=10f
will prevent any dirty unused pages from accumulating.
.It G
When there are multiple threads, use thread-specific caching for objects that
are smaller than one page.
This option is enabled by default.
Thread-specific caching allows many allocations to be satisfied without
performing any thread synchronization, at the cost of increased memory use.
See the
.Dq R
option for related tuning information.
This option is not available for some configurations (non-PIC).
.It J
Each byte of new memory allocated by
.Fn malloc
or
.Fn realloc
will be initialized to 0xa5.
All memory returned by
.Fn free
or
.Fn realloc
will be initialized to 0x5a.
This is intended for debugging and will impact performance negatively.
.It K
Double/halve the virtual memory chunk size.
The default chunk size is 1 MB.
.It M
Use
.Xr mmap 2
to acquire anonymously mapped memory.
This option is enabled by default.
If both the
.Dq D
and
.Dq M
options are enabled, the allocator prefers the DSS over anonymous mappings,
but allocation only fails if memory cannot be acquired via either method.
If neither option is enabled, then the
.Dq M
option is implicitly enabled in order to assure that there is a method for
acquiring memory.
.It N
Double/halve the number of arenas.
The default number of arenas is two times the number of CPUs, or one if there
is a single CPU.
.It P
Various statistics are printed at program exit via an
.Xr atexit 3
function.
This has the potential to cause deadlock for a multi-threaded process that exits
while one or more threads are executing in the memory allocation functions.
Therefore, this option should only be used with care; it is primarily intended
as a performance tuning aid during application development.
.It Q
Double/halve the size 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.
.It R
Double/halve magazine size, which approximately doubles/halves the number of
rounds in each magazine.
Magazines are used by the thread-specific caching machinery to acquire and
release objects in bulk.
Increasing the magazine size decreases locking overhead, at the expense of
increased memory usage.
This option is not available for some configurations (non-PIC).
.It U
Generate
.Dq utrace
entries for
.Xr ktrace 1 ,
for all operations.
Consult the source for details on this option.
.It V
Attempting to allocate zero bytes will return a
.Dv 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
.Dq X
option.
.It X
Rather than return failure for any allocation function,
display a diagnostic message on
.Dv stderr
and cause the program to drop
core (using
.Xr abort 3 ) .
This option should be set at compile time by including the following in
the source code:
.Bd -literal -offset indent
jemalloc_options = "X";
.Ed
.It Z
Each byte of new memory allocated by
.Fn malloc
or
.Fn realloc
will be initialized to 0.
Note that this initialization only happens once for each byte, so
.Fn realloc
calls do not zero memory that was previously allocated.
This is intended for debugging and will impact performance negatively.
.El
.Pp
The
.Dq J
and
.Dq Z
options are intended for testing and debugging.
An application which changes its behavior when these options are used
is flawed.
.Sh IMPLEMENTATION NOTES
Traditionally, allocators have used
.Xr sbrk 2
to obtain memory, which is suboptimal for several reasons, including race
conditions, increased fragmentation, and artificial limitations on maximum
usable memory.
This allocator uses both
.Xr sbrk 2
and
.Xr mmap 2
by default, but it can be configured at run time to use only one or the other.
If resource limits are not a primary concern, the preferred configuration is
.Ev JEMALLOC_OPTIONS=dM
or
.Ev JEMALLOC_OPTIONS=DM .
When so configured, the
.Ar datasize
resource limit has little practical effect for typical applications; use
.Ev JEMALLOC_OPTIONS=Dm
if that is a concern.
Regardless of allocator configuration, the
.Ar vmemoryuse
resource limit can be used to bound the total virtual memory used by a
process, as described in
.Xr limits 1 .
.Pp
This allocator uses multiple arenas in order to reduce lock contention for
threaded programs on multi-processor systems.
This works well with regard to threading scalability, but incurs some costs.
There is a small fixed per-arena overhead, and additionally, arenas manage
memory completely independently of each other, which means a small fixed
increase in overall memory fragmentation.
These overheads are not generally an issue, given the number of arenas normally
used.
Note that using substantially more arenas than the default is not likely to
improve performance, mainly due to reduced cache performance.
However, it may make sense to reduce the number of arenas if an application
does not make much use of the allocation functions.
.Pp
In addition to multiple arenas, this allocator supports thread-specific
caching for small objects (smaller than one page), in order to make it
possible to completely avoid synchronization for most small allocation requests.
Such caching allows very fast allocation in the common case, but it increases
memory usage and fragmentation, since a bounded number of objects can remain
allocated in each thread cache.
.Pp
Memory is conceptually broken into equal-sized chunks, where the chunk size is
a power of two that is greater than the page size.
Chunks are always aligned to multiples of the chunk size.
This alignment makes it possible to find metadata for user objects very
quickly.
.Pp
User objects are broken into three categories according to size: small, large,
and huge.
Small objects are smaller than one page.
Large objects are smaller than the chunk size.
Huge objects are a multiple of the chunk size.
Small and large objects are managed by arenas; huge objects are managed
separately in a single data structure that is shared by all threads.
Huge objects are used by applications infrequently enough that this single
data structure is not a scalability issue.
.Pp
Each chunk that is managed by an arena 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.
.Pp
Small objects are managed in groups by page runs.
Each run maintains a bitmap that tracks which regions are in use.
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.
Allocation requests that are more than half the quantum, but no more than the
minimum cacheline-multiple size class (see the
.Dq Q
option) are rounded up to the nearest multiple of the quantum.
Allocation requests that are more than the minumum cacheline-multiple size
class, but no more than the minimum subpage-multiple size class (see the
.Dq C
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
are rounded up to the nearest multiple of the subpage size (256).
Allocation requests that are more than one page, but small enough to fit in
an arena-managed chunk (see the
.Dq K
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 the chunk size.
.Pp
Allocations are packed tightly together, which can be an issue for
multi-threaded applications.
If you need to assure that allocations do not suffer from cacheline sharing,
round your allocation requests up to the nearest multiple of the cacheline
size.
.Sh DEBUGGING MALLOC PROBLEMS
The first thing to do is to set the
.Dq A
option.
This option forces a coredump (if possible) at the first sign of trouble,
rather than the normal policy of trying to continue if at all possible.
.Pp
It is probably also a good idea to recompile the program with suitable
options and symbols for debugger support.
.Pp
If the program starts to give unusual results, coredump or generally behave
differently without emitting any of the messages mentioned in the next
section, it is likely because it depends on the storage being filled with
zero bytes.
Try running it with the
.Dq Z
option set;
if that improves the situation, this diagnosis has been confirmed.
If the program still misbehaves,
the likely problem is accessing memory outside the allocated area.
.Pp
Alternatively, if the symptoms are not easy to reproduce, setting the
.Dq J
option may help provoke the problem.
.Pp
In truly difficult cases, the
.Dq U
option, if supported by the kernel, can provide a detailed trace of
all calls made to these functions.
.Pp
Unfortunately this implementation does not provide much detail about
the problems it detects; 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.
.Sh DIAGNOSTIC MESSAGES
If any of the memory allocation/deallocation functions detect an error or
warning condition, a message will be printed to file descriptor
.Dv STDERR_FILENO .
Errors will result in the process dumping core.
If the
.Dq A
option is set, all warnings are treated as errors.
.Pp
The
.Va _malloc_message
variable allows the programmer to override the function which emits
the text strings forming the errors and warnings if for some reason
the
.Dv stderr
file descriptor is not suitable for this.
Please note that doing anything which tries to allocate memory in
this function is likely to result in a crash or deadlock.
.Pp
All messages are prefixed by
.Dq Ao Ar progname Ac Ns Li : (malloc) .
.Sh RETURN VALUES
The
.Fn malloc
and
.Fn calloc
functions return a pointer to the allocated memory if successful; otherwise
a
.Dv NULL
pointer is returned and
.Va errno
is set to
.Er ENOMEM .
.Pp
The
.Fn posix_memalign
function returns the value 0 if successful; otherwise it returns an error value.
The
.Fn posix_memalign
function will fail if:
.Bl -tag -width Er
.It Bq Er EINVAL
The
.Fa alignment
parameter is not a power of 2 at least as large as
.Fn sizeof "void *" .
.It Bq Er ENOMEM
Memory allocation error.
.El
.Pp
The
.Fn realloc
function returns a pointer, possibly identical to
.Fa ptr ,
to the allocated memory
if successful; otherwise a
.Dv NULL
pointer is returned, and
.Va errno
is set to
.Er ENOMEM
if the error was the result of an allocation failure.
The
.Fn realloc
function always leaves the original buffer intact
when an error occurs.
.Pp
The
.Fn free
function returns no value.
.Pp
The
.Fn malloc_usable_size
function returns the usable size of the allocation pointed to by
.Fa ptr .
.Sh ENVIRONMENT
The following environment variables affect the execution of the allocation
functions:
.Bl -tag -width ".Ev JEMALLOC_OPTIONS"
.It Ev JEMALLOC_OPTIONS
If the environment variable
.Ev JEMALLOC_OPTIONS
is set, the characters it contains will be interpreted as flags to the
allocation functions.
.El
.Sh EXAMPLES
To dump core whenever a problem occurs:
.Pp
.Bd -literal -offset indent
ln -s 'A' /etc/jemalloc.conf
.Ed
.Pp
To specify in the source that a program does no return value checking
on calls to these functions:
.Bd -literal -offset indent
jemalloc_options = "X";
.Ed
.Sh SEE ALSO
.Xr limits 1 ,
.Xr madvise 2 ,
.Xr mmap 2 ,
.Xr sbrk 2 ,
.Xr alloca 3 ,
.Xr atexit 3 ,
.Xr getpagesize 3 ,
.Xr memory 3 ,
.Xr posix_memalign 3
.Sh STANDARDS
The
.Fn malloc ,
.Fn calloc ,
.Fn realloc
and
.Fn free
functions conform to
.St -isoC .
.Pp
The
.Fn posix_memalign
function conforms to
.St -p1003.1-2001 .