Skip to content

Instantly share code, notes, and snippets.

@davidxia

davidxia/malloc-2.6.4.c

Created Apr 20, 2020
Star
Embed
What would you like to do?
Learn more about clone URLs
Download ZIP
old malloc implementations
Raw
malloc-2.6.4.c
/* Malloc implementation for multiple threads without lock contention.
Copyright (C) 1996-2001, 2002 Free Software Foundation, Inc.
This file is part of the GNU C Library.
Contributed by Wolfram Gloger <wmglo@dent.med.uni-muenchen.de>
and Doug Lea <dl@cs.oswego.edu>, 1996.
The GNU C Library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
The GNU C Library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with the GNU C Library; if not, write to the Free
Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
02111-1307 USA. */
/* $Id$
This work is mainly derived from malloc-2.6.4 by Doug Lea
<dl@cs.oswego.edu>, which is available from:
ftp://g.oswego.edu/pub/misc/malloc.c
Most of the original comments are reproduced in the code below.
* Why use this malloc?
This is not the fastest, most space-conserving, most portable, or
most tunable malloc ever written. However it is among the fastest
while also being among the most space-conserving, portable and tunable.
Consistent balance across these factors results in a good general-purpose
allocator. For a high-level description, see
http://g.oswego.edu/dl/html/malloc.html
On many systems, the standard malloc implementation is by itself not
thread-safe, and therefore wrapped with a single global lock around
all malloc-related functions. In some applications, especially with
multiple available processors, this can lead to contention problems
and bad performance. This malloc version was designed with the goal
to avoid waiting for locks as much as possible. Statistics indicate
that this goal is achieved in many cases.
* Synopsis of public routines
(Much fuller descriptions are contained in the program documentation below.)
ptmalloc_init();
Initialize global configuration. When compiled for multiple threads,
this function must be called once before any other function in the
package. It is not required otherwise. It is called automatically
in the Linux/GNU C libray or when compiling with MALLOC_HOOKS.
malloc(size_t n);
Return a pointer to a newly allocated chunk of at least n bytes, or null
if no space is available.
free(Void_t* p);
Release the chunk of memory pointed to by p, or no effect if p is null.
realloc(Void_t* p, size_t n);
Return a pointer to a chunk of size n that contains the same data
as does chunk p up to the minimum of (n, p's size) bytes, or null
if no space is available. The returned pointer may or may not be
the same as p. If p is null, equivalent to malloc. Unless the
#define REALLOC_ZERO_BYTES_FREES below is set, realloc with a
size argument of zero (re)allocates a minimum-sized chunk.
memalign(size_t alignment, size_t n);
Return a pointer to a newly allocated chunk of n bytes, aligned
in accord with the alignment argument, which must be a power of
two.
valloc(size_t n);
Equivalent to memalign(pagesize, n), where pagesize is the page
size of the system (or as near to this as can be figured out from
all the includes/defines below.)
pvalloc(size_t n);
Equivalent to valloc(minimum-page-that-holds(n)), that is,
round up n to nearest pagesize.
calloc(size_t unit, size_t quantity);
Returns a pointer to quantity * unit bytes, with all locations
set to zero.
cfree(Void_t* p);
Equivalent to free(p).
malloc_trim(size_t pad);
Release all but pad bytes of freed top-most memory back
to the system. Return 1 if successful, else 0.
malloc_usable_size(Void_t* p);
Report the number usable allocated bytes associated with allocated
chunk p. This may or may not report more bytes than were requested,
due to alignment and minimum size constraints.
malloc_stats();
Prints brief summary statistics on stderr.
mallinfo()
Returns (by copy) a struct containing various summary statistics.
mallopt(int parameter_number, int parameter_value)
Changes one of the tunable parameters described below. Returns
1 if successful in changing the parameter, else 0.
* Vital statistics:
Alignment: 8-byte
8 byte alignment is currently hardwired into the design. This
seems to suffice for all current machines and C compilers.
Assumed pointer representation: 4 or 8 bytes
Code for 8-byte pointers is untested by me but has worked
reliably by Wolfram Gloger, who contributed most of the
changes supporting this.
Assumed size_t representation: 4 or 8 bytes
Note that size_t is allowed to be 4 bytes even if pointers are 8.
Minimum overhead per allocated chunk: 4 or 8 bytes
Each malloced chunk has a hidden overhead of 4 bytes holding size
and status information.
Minimum allocated size: 4-byte ptrs: 16 bytes (including 4 overhead)
8-byte ptrs: 24/32 bytes (including, 4/8 overhead)
When a chunk is freed, 12 (for 4byte ptrs) or 20 (for 8 byte
ptrs but 4 byte size) or 24 (for 8/8) additional bytes are
needed; 4 (8) for a trailing size field
and 8 (16) bytes for free list pointers. Thus, the minimum
allocatable size is 16/24/32 bytes.
Even a request for zero bytes (i.e., malloc(0)) returns a
pointer to something of the minimum allocatable size.
Maximum allocated size: 4-byte size_t: 2^31 - 8 bytes
8-byte size_t: 2^63 - 16 bytes
It is assumed that (possibly signed) size_t bit values suffice to
represent chunk sizes. `Possibly signed' is due to the fact
that `size_t' may be defined on a system as either a signed or
an unsigned type. To be conservative, values that would appear
as negative numbers are avoided.
Requests for sizes with a negative sign bit will return a
minimum-sized chunk.
Maximum overhead wastage per allocated chunk: normally 15 bytes
Alignment demands, plus the minimum allocatable size restriction
make the normal worst-case wastage 15 bytes (i.e., up to 15
more bytes will be allocated than were requested in malloc), with
two exceptions:
1. Because requests for zero bytes allocate non-zero space,
the worst case wastage for a request of zero bytes is 24 bytes.
2. For requests >= mmap_threshold that are serviced via
mmap(), the worst case wastage is 8 bytes plus the remainder
from a system page (the minimal mmap unit); typically 4096 bytes.
* Limitations
Here are some features that are NOT currently supported
* No automated mechanism for fully checking that all accesses
to malloced memory stay within their bounds.
* No support for compaction.
* Synopsis of compile-time options:
People have reported using previous versions of this malloc on all
versions of Unix, sometimes by tweaking some of the defines
below. It has been tested most extensively on Solaris and
Linux. People have also reported adapting this malloc for use in
stand-alone embedded systems.
The implementation is in straight, hand-tuned ANSI C. Among other
consequences, it uses a lot of macros. Because of this, to be at
all usable, this code should be compiled using an optimizing compiler
(for example gcc -O2) that can simplify expressions and control
paths.
__STD_C (default: derived from C compiler defines)
Nonzero if using ANSI-standard C compiler, a C++ compiler, or
a C compiler sufficiently close to ANSI to get away with it.
MALLOC_DEBUG (default: NOT defined)
Define to enable debugging. Adds fairly extensive assertion-based
checking to help track down memory errors, but noticeably slows down
execution.
MALLOC_HOOKS (default: NOT defined)
Define to enable support run-time replacement of the allocation
functions through user-defined `hooks'.
REALLOC_ZERO_BYTES_FREES (default: defined)
Define this if you think that realloc(p, 0) should be equivalent
to free(p). (The C standard requires this behaviour, therefore
it is the default.) Otherwise, since malloc returns a unique
pointer for malloc(0), so does realloc(p, 0).
HAVE_MEMCPY (default: defined)
Define if you are not otherwise using ANSI STD C, but still
have memcpy and memset in your C library and want to use them.
Otherwise, simple internal versions are supplied.
USE_MEMCPY (default: 1 if HAVE_MEMCPY is defined, 0 otherwise)
Define as 1 if you want the C library versions of memset and
memcpy called in realloc and calloc (otherwise macro versions are used).
At least on some platforms, the simple macro versions usually
outperform libc versions.
HAVE_MMAP (default: defined as 1)
Define to non-zero to optionally make malloc() use mmap() to
allocate very large blocks.
HAVE_MREMAP (default: defined as 0 unless Linux libc set)
Define to non-zero to optionally make realloc() use mremap() to
reallocate very large blocks.
USE_ARENAS (default: the same as HAVE_MMAP)
Enable support for multiple arenas, allocated using mmap().
malloc_getpagesize (default: derived from system #includes)
Either a constant or routine call returning the system page size.
HAVE_USR_INCLUDE_MALLOC_H (default: NOT defined)
Optionally define if you are on a system with a /usr/include/malloc.h
that declares struct mallinfo. It is not at all necessary to
define this even if you do, but will ensure consistency.
INTERNAL_SIZE_T (default: size_t)
Define to a 32-bit type (probably `unsigned int') if you are on a
64-bit machine, yet do not want or need to allow malloc requests of
greater than 2^31 to be handled. This saves space, especially for
very small chunks.
_LIBC (default: NOT defined)
Defined only when compiled as part of the Linux libc/glibc.
Also note that there is some odd internal name-mangling via defines
(for example, internally, `malloc' is named `mALLOc') needed
when compiling in this case. These look funny but don't otherwise
affect anything.
LACKS_UNISTD_H (default: undefined)
Define this if your system does not have a <unistd.h>.
MORECORE (default: sbrk)
The name of the routine to call to obtain more memory from the system.
MORECORE_FAILURE (default: -1)
The value returned upon failure of MORECORE.
MORECORE_CLEARS (default 1)
The degree to which the routine mapped to MORECORE zeroes out
memory: never (0), only for newly allocated space (1) or always
(2). The distinction between (1) and (2) is necessary because on
some systems, if the application first decrements and then
increments the break value, the contents of the reallocated space
are unspecified.
DEFAULT_TRIM_THRESHOLD
DEFAULT_TOP_PAD
DEFAULT_MMAP_THRESHOLD
DEFAULT_MMAP_MAX
Default values of tunable parameters (described in detail below)
controlling interaction with host system routines (sbrk, mmap, etc).
These values may also be changed dynamically via mallopt(). The
preset defaults are those that give best performance for typical
programs/systems.
DEFAULT_CHECK_ACTION
When the standard debugging hooks are in place, and a pointer is
detected as corrupt, do nothing (0), print an error message (1),
or call abort() (2).
*/
/*
* Compile-time options for multiple threads:
USE_PTHREADS, USE_THR, USE_SPROC
Define one of these as 1 to select the thread interface:
POSIX threads, Solaris threads or SGI sproc's, respectively.
If none of these is defined as non-zero, you get a `normal'
malloc implementation which is not thread-safe. Support for
multiple threads requires HAVE_MMAP=1. As an exception, when
compiling for GNU libc, i.e. when _LIBC is defined, then none of
the USE_... symbols have to be defined.
HEAP_MIN_SIZE
HEAP_MAX_SIZE
When thread support is enabled, additional `heap's are created
with mmap calls. These are limited in size; HEAP_MIN_SIZE should
be a multiple of the page size, while HEAP_MAX_SIZE must be a power
of two for alignment reasons. HEAP_MAX_SIZE should be at least
twice as large as the mmap threshold.
THREAD_STATS
When this is defined as non-zero, some statistics on mutex locking
are computed.
*/
/* Preliminaries */
#ifndef __STD_C
#if defined (__STDC__)
#define __STD_C 1
#else
#if __cplusplus
#define __STD_C 1
#else
#define __STD_C 0
#endif /*__cplusplus*/
#endif /*__STDC__*/
#endif /*__STD_C*/
#ifndef Void_t
#if __STD_C
#define Void_t void
#else
#define Void_t char
#endif
#endif /*Void_t*/
#define _GNU_SOURCE
#include <features.h>
#define _LIBC 1
#define NOT_IN_libc 1
#if __STD_C
# include <stddef.h> /* for size_t */
# if defined _LIBC || defined MALLOC_HOOKS
# include <stdlib.h> /* for getenv(), abort() */
# endif
#else
# include <sys/types.h>
# if defined _LIBC || defined MALLOC_HOOKS
extern char* getenv();
# endif
#endif
/* newlib modifications */
#include <libc-symbols.h>
#include <sys/types.h>
extern void __pthread_initialize (void) __attribute__((weak));
extern void *__mmap (void *__addr, size_t __len, int __prot,
int __flags, int __fd, off_t __offset);
extern int __munmap (void *__addr, size_t __len);
extern void *__mremap (void *__addr, size_t __old_len, size_t __new_len,
int __may_move);
extern int __getpagesize (void);
#define __libc_enable_secure 1
/* Macros for handling mutexes and thread-specific data. This is
included early, because some thread-related header files (such as
pthread.h) should be included before any others. */
#include <bits/libc-lock.h>
#include "thread-m.h"
void *(*__malloc_internal_tsd_get) (enum __libc_tsd_key_t) = NULL;
int (*__malloc_internal_tsd_set) (enum __libc_tsd_key_t,
__const void *) = NULL;
weak_alias(__malloc_internal_tsd_get, __libc_internal_tsd_get)
weak_alias(__malloc_internal_tsd_set, __libc_internal_tsd_set)
#ifdef __cplusplus
extern "C" {
#endif
#include <errno.h>
#include <stdio.h> /* needed for malloc_stats */
/*
Compile-time options
*/
/*
Debugging:
Because freed chunks may be overwritten with link fields, this
malloc will often die when freed memory is overwritten by user
programs. This can be very effective (albeit in an annoying way)
in helping track down dangling pointers.
If you compile with -DMALLOC_DEBUG, a number of assertion checks are
enabled that will catch more memory errors. You probably won't be
able to make much sense of the actual assertion errors, but they
should help you locate incorrectly overwritten memory. The
checking is fairly extensive, and will slow down execution
noticeably. Calling malloc_stats or mallinfo with MALLOC_DEBUG set will
attempt to check every non-mmapped allocated and free chunk in the
course of computing the summaries. (By nature, mmapped regions
cannot be checked very much automatically.)
Setting MALLOC_DEBUG may also be helpful if you are trying to modify
this code. The assertions in the check routines spell out in more
detail the assumptions and invariants underlying the algorithms.
*/
#if MALLOC_DEBUG
#include <assert.h>
#else
#define assert(x) ((void)0)
#endif
/*
INTERNAL_SIZE_T is the word-size used for internal bookkeeping
of chunk sizes. On a 64-bit machine, you can reduce malloc
overhead by defining INTERNAL_SIZE_T to be a 32 bit `unsigned int'
at the expense of not being able to handle requests greater than
2^31. This limitation is hardly ever a concern; you are encouraged
to set this. However, the default version is the same as size_t.
*/
#ifndef INTERNAL_SIZE_T
#define INTERNAL_SIZE_T size_t
#endif
/*
REALLOC_ZERO_BYTES_FREES should be set if a call to realloc with
zero bytes should be the same as a call to free. The C standard
requires this. Otherwise, since this malloc returns a unique pointer
for malloc(0), so does realloc(p, 0).
*/
#define REALLOC_ZERO_BYTES_FREES
/*
HAVE_MEMCPY should be defined if you are not otherwise using
ANSI STD C, but still have memcpy and memset in your C library
and want to use them in calloc and realloc. Otherwise simple
macro versions are defined here.
USE_MEMCPY should be defined as 1 if you actually want to
have memset and memcpy called. People report that the macro
versions are often enough faster than libc versions on many
systems that it is better to use them.
*/
#define HAVE_MEMCPY 1
#ifndef USE_MEMCPY
#ifdef HAVE_MEMCPY
#define USE_MEMCPY 1
#else
#define USE_MEMCPY 0
#endif
#endif
#if (__STD_C || defined(HAVE_MEMCPY))
#if __STD_C
void* memset(void*, int, size_t);
void* memcpy(void*, const void*, size_t);
void* memmove(void*, const void*, size_t);
#else
Void_t* memset();
Void_t* memcpy();
Void_t* memmove();
#endif
#endif
/* The following macros are only invoked with (2n+1)-multiples of
INTERNAL_SIZE_T units, with a positive integer n. This is exploited
for fast inline execution when n is small. If the regions to be
copied do overlap, the destination lies always _below_ the source. */
#if USE_MEMCPY
#define MALLOC_ZERO(charp, nbytes) \
do { \
INTERNAL_SIZE_T mzsz = (nbytes); \
if(mzsz <= 9*sizeof(mzsz)) { \
INTERNAL_SIZE_T* mz = (INTERNAL_SIZE_T*) (charp); \
if(mzsz >= 5*sizeof(mzsz)) { *mz++ = 0; \
*mz++ = 0; \
if(mzsz >= 7*sizeof(mzsz)) { *mz++ = 0; \
*mz++ = 0; \
if(mzsz >= 9*sizeof(mzsz)) { *mz++ = 0; \
*mz++ = 0; }}} \
*mz++ = 0; \
*mz++ = 0; \
*mz = 0; \
} else memset((charp), 0, mzsz); \
} while(0)
/* If the regions overlap, dest is always _below_ src. */
#define MALLOC_COPY(dest,src,nbytes,overlap) \
do { \
INTERNAL_SIZE_T mcsz = (nbytes); \
if(mcsz <= 9*sizeof(mcsz)) { \
INTERNAL_SIZE_T* mcsrc = (INTERNAL_SIZE_T*) (src); \
INTERNAL_SIZE_T* mcdst = (INTERNAL_SIZE_T*) (dest); \
if(mcsz >= 5*sizeof(mcsz)) { *mcdst++ = *mcsrc++; \
*mcdst++ = *mcsrc++; \
if(mcsz >= 7*sizeof(mcsz)) { *mcdst++ = *mcsrc++; \
*mcdst++ = *mcsrc++; \
if(mcsz >= 9*sizeof(mcsz)) { *mcdst++ = *mcsrc++; \
*mcdst++ = *mcsrc++; }}} \
*mcdst++ = *mcsrc++; \
*mcdst++ = *mcsrc++; \
*mcdst = *mcsrc ; \
} else if(overlap) \
memmove(dest, src, mcsz); \
else \
memcpy(dest, src, mcsz); \
} while(0)
#else /* !USE_MEMCPY */
/* Use Duff's device for good zeroing/copying performance. */
#define MALLOC_ZERO(charp, nbytes) \
do { \
INTERNAL_SIZE_T* mzp = (INTERNAL_SIZE_T*)(charp); \
long mctmp = (nbytes)/sizeof(INTERNAL_SIZE_T), mcn; \
if (mctmp < 8) mcn = 0; else { mcn = (mctmp-1)/8; mctmp %= 8; } \
switch (mctmp) { \
case 0: for(;;) { *mzp++ = 0; \
case 7: *mzp++ = 0; \
case 6: *mzp++ = 0; \
case 5: *mzp++ = 0; \
case 4: *mzp++ = 0; \
case 3: *mzp++ = 0; \
case 2: *mzp++ = 0; \
case 1: *mzp++ = 0; if(mcn <= 0) break; mcn--; } \
} \
} while(0)
/* If the regions overlap, dest is always _below_ src. */
#define MALLOC_COPY(dest,src,nbytes,overlap) \
do { \
INTERNAL_SIZE_T* mcsrc = (INTERNAL_SIZE_T*) src; \
INTERNAL_SIZE_T* mcdst = (INTERNAL_SIZE_T*) dest; \
long mctmp = (nbytes)/sizeof(INTERNAL_SIZE_T), mcn; \
if (mctmp < 8) mcn = 0; else { mcn = (mctmp-1)/8; mctmp %= 8; } \
switch (mctmp) { \
case 0: for(;;) { *mcdst++ = *mcsrc++; \
case 7: *mcdst++ = *mcsrc++; \
case 6: *mcdst++ = *mcsrc++; \
case 5: *mcdst++ = *mcsrc++; \
case 4: *mcdst++ = *mcsrc++; \
case 3: *mcdst++ = *mcsrc++; \
case 2: *mcdst++ = *mcsrc++; \
case 1: *mcdst++ = *mcsrc++; if(mcn <= 0) break; mcn--; } \
} \
} while(0)
#endif
#ifndef LACKS_UNISTD_H
# include <unistd.h>
#endif
/*
Define HAVE_MMAP to optionally make malloc() use mmap() to allocate
very large blocks. These will be returned to the operating system
immediately after a free(). HAVE_MMAP is also a prerequisite to
support multiple `arenas' (see USE_ARENAS below).
*/
#ifndef HAVE_MMAP
# ifdef _POSIX_MAPPED_FILES
# define HAVE_MMAP 1
# endif
#endif
/*
Define HAVE_MREMAP to make realloc() use mremap() to re-allocate
large blocks. This is currently only possible on Linux with
kernel versions newer than 1.3.77.
*/
#ifndef HAVE_MREMAP
#define HAVE_MREMAP defined(__linux__)
#endif
/* Define USE_ARENAS to enable support for multiple `arenas'. These
are allocated using mmap(), are necessary for threads and
occasionally useful to overcome address space limitations affecting
sbrk(). */
#ifndef USE_ARENAS
#define USE_ARENAS HAVE_MMAP
#endif
#if HAVE_MMAP
#include <unistd.h>
#include <fcntl.h>
#include <sys/mman.h>
#if !defined(MAP_ANONYMOUS) && defined(MAP_ANON)
#define MAP_ANONYMOUS MAP_ANON
#endif
#if !defined(MAP_FAILED)
#define MAP_FAILED ((char*)-1)
#endif
#ifndef MAP_NORESERVE
# ifdef MAP_AUTORESRV
# define MAP_NORESERVE MAP_AUTORESRV
# else
# define MAP_NORESERVE 0
# endif
#endif
#endif /* HAVE_MMAP */
/*
Access to system page size. To the extent possible, this malloc
manages memory from the system in page-size units.
The following mechanics for getpagesize were adapted from
bsd/gnu getpagesize.h
*/
#ifndef malloc_getpagesize
# ifdef _SC_PAGESIZE /* some SVR4 systems omit an underscore */
# ifndef _SC_PAGE_SIZE
# define _SC_PAGE_SIZE _SC_PAGESIZE
# endif
# endif
# ifdef _SC_PAGE_SIZE
# define malloc_getpagesize sysconf(_SC_PAGE_SIZE)
# else
# if defined(BSD) || defined(DGUX) || defined(HAVE_GETPAGESIZE)
extern size_t getpagesize();
# define malloc_getpagesize getpagesize()
# else
# include <sys/param.h>
# ifdef EXEC_PAGESIZE
# define malloc_getpagesize EXEC_PAGESIZE
# else
# ifdef NBPG
# ifndef CLSIZE
# define malloc_getpagesize NBPG
# else
# define malloc_getpagesize (NBPG * CLSIZE)
# endif
# else
# ifdef NBPC
# define malloc_getpagesize NBPC
# else
# ifdef PAGESIZE
# define malloc_getpagesize PAGESIZE
# else
# define malloc_getpagesize (4096) /* just guess */
# endif
# endif
# endif
# endif
# endif
# endif
#endif
/*
This version of malloc supports the standard SVID/XPG mallinfo
routine that returns a struct containing the same kind of
information you can get from malloc_stats. It should work on
any SVID/XPG compliant system that has a /usr/include/malloc.h
defining struct mallinfo. (If you'd like to install such a thing
yourself, cut out the preliminary declarations as described above
and below and save them in a malloc.h file. But there's no
compelling reason to bother to do this.)
The main declaration needed is the mallinfo struct that is returned
(by-copy) by mallinfo(). The SVID/XPG malloinfo struct contains a
bunch of fields, most of which are not even meaningful in this
version of malloc. Some of these fields are are instead filled by
mallinfo() with other numbers that might possibly be of interest.
HAVE_USR_INCLUDE_MALLOC_H should be set if you have a
/usr/include/malloc.h file that includes a declaration of struct
mallinfo. If so, it is included; else an SVID2/XPG2 compliant
version is declared below. These must be precisely the same for
mallinfo() to work.
*/
/* #define HAVE_USR_INCLUDE_MALLOC_H */
#if HAVE_USR_INCLUDE_MALLOC_H
# include "/usr/include/malloc.h"
#else
# ifdef _LIBC
# include "malloc.h"
# else
# include "ptmalloc.h"
# endif
#endif
#include <bp-checks.h>
#ifndef DEFAULT_TRIM_THRESHOLD
#define DEFAULT_TRIM_THRESHOLD (128 * 1024)
#endif
/*
M_TRIM_THRESHOLD is the maximum amount of unused top-most memory
to keep before releasing via malloc_trim in free().
Automatic trimming is mainly useful in long-lived programs.
Because trimming via sbrk can be slow on some systems, and can
sometimes be wasteful (in cases where programs immediately
afterward allocate more large chunks) the value should be high
enough so that your overall system performance would improve by
releasing.
The trim threshold and the mmap control parameters (see below)
can be traded off with one another. Trimming and mmapping are
two different ways of releasing unused memory back to the
system. Between these two, it is often possible to keep
system-level demands of a long-lived program down to a bare
minimum. For example, in one test suite of sessions measuring
the XF86 X server on Linux, using a trim threshold of 128K and a
mmap threshold of 192K led to near-minimal long term resource
consumption.
If you are using this malloc in a long-lived program, it should
pay to experiment with these values. As a rough guide, you
might set to a value close to the average size of a process
(program) running on your system. Releasing this much memory
would allow such a process to run in memory. Generally, it's
worth it to tune for trimming rather than memory mapping when a
program undergoes phases where several large chunks are
allocated and released in ways that can reuse each other's
storage, perhaps mixed with phases where there are no such
chunks at all. And in well-behaved long-lived programs,
controlling release of large blocks via trimming versus mapping
is usually faster.
However, in most programs, these parameters serve mainly as
protection against the system-level effects of carrying around
massive amounts of unneeded memory. Since frequent calls to
sbrk, mmap, and munmap otherwise degrade performance, the default
parameters are set to relatively high values that serve only as
safeguards.
The default trim value is high enough to cause trimming only in
fairly extreme (by current memory consumption standards) cases.
It must be greater than page size to have any useful effect. To
disable trimming completely, you can set to (unsigned long)(-1);
*/
#ifndef DEFAULT_TOP_PAD
#define DEFAULT_TOP_PAD (0)
#endif
/*
M_TOP_PAD is the amount of extra `padding' space to allocate or
retain whenever sbrk is called. It is used in two ways internally:
* When sbrk is called to extend the top of the arena to satisfy
a new malloc request, this much padding is added to the sbrk
request.
* When malloc_trim is called automatically from free(),
it is used as the `pad' argument.
In both cases, the actual amount of padding is rounded
so that the end of the arena is always a system page boundary.
The main reason for using padding is to avoid calling sbrk so
often. Having even a small pad greatly reduces the likelihood
that nearly every malloc request during program start-up (or
after trimming) will invoke sbrk, which needlessly wastes
time.
Automatic rounding-up to page-size units is normally sufficient
to avoid measurable overhead, so the default is 0. However, in
systems where sbrk is relatively slow, it can pay to increase
this value, at the expense of carrying around more memory than
the program needs.
*/
#ifndef DEFAULT_MMAP_THRESHOLD
#define DEFAULT_MMAP_THRESHOLD (128 * 1024)
#endif
/*
M_MMAP_THRESHOLD is the request size threshold for using mmap()
to service a request. Requests of at least this size that cannot
be allocated using already-existing space will be serviced via mmap.
(If enough normal freed space already exists it is used instead.)
Using mmap segregates relatively large chunks of memory so that
they can be individually obtained and released from the host
system. A request serviced through mmap is never reused by any
other request (at least not directly; the system may just so
happen to remap successive requests to the same locations).
Segregating space in this way has the benefit that mmapped space
can ALWAYS be individually released back to the system, which
helps keep the system level memory demands of a long-lived
program low. Mapped memory can never become `locked' between
other chunks, as can happen with normally allocated chunks, which
menas that even trimming via malloc_trim would not release them.
However, it has the disadvantages that:
1. The space cannot be reclaimed, consolidated, and then
used to service later requests, as happens with normal chunks.
2. It can lead to more wastage because of mmap page alignment
requirements
3. It causes malloc performance to be more dependent on host
system memory management support routines which may vary in
implementation quality and may impose arbitrary
limitations. Generally, servicing a request via normal
malloc steps is faster than going through a system's mmap.
All together, these considerations should lead you to use mmap
only for relatively large requests.
*/
#ifndef DEFAULT_MMAP_MAX
#if HAVE_MMAP
#define DEFAULT_MMAP_MAX (1024)
#else
#define DEFAULT_MMAP_MAX (0)
#endif
#endif
/*
M_MMAP_MAX is the maximum number of requests to simultaneously
service using mmap. This parameter exists because:
1. Some systems have a limited number of internal tables for
use by mmap.
2. In most systems, overreliance on mmap can degrade overall
performance.
3. If a program allocates many large regions, it is probably
better off using normal sbrk-based allocation routines that
can reclaim and reallocate normal heap memory. Using a
small value allows transition into this mode after the
first few allocations.
Setting to 0 disables all use of mmap. If HAVE_MMAP is not set,
the default value is 0, and attempts to set it to non-zero values
in mallopt will fail.
*/
#ifndef DEFAULT_CHECK_ACTION
#define DEFAULT_CHECK_ACTION 1
#endif
/* What to do if the standard debugging hooks are in place and a
corrupt pointer is detected: do nothing (0), print an error message
(1), or call abort() (2). */
#define HEAP_MIN_SIZE (32*1024)
#define HEAP_MAX_SIZE (1024*1024) /* must be a power of two */
/* HEAP_MIN_SIZE and HEAP_MAX_SIZE limit the size of mmap()ed heaps
that are dynamically created for multi-threaded programs. The
maximum size must be a power of two, for fast determination of
which heap belongs to a chunk. It should be much larger than
the mmap threshold, so that requests with a size just below that
threshold can be fulfilled without creating too many heaps.
*/
#ifndef THREAD_STATS
#define THREAD_STATS 0
#endif
/* If THREAD_STATS is non-zero, some statistics on mutex locking are
computed. */
/* Macro to set errno. */
#ifndef __set_errno
# define __set_errno(val) errno = (val)
#endif
/* On some platforms we can compile internal, not exported functions better.
Let the environment provide a macro and define it to be empty if it
is not available. */
#ifndef internal_function
# define internal_function
#endif
/*
Special defines for the Linux/GNU C library.
*/
#ifdef _LIBC
#if __STD_C
Void_t * __default_morecore (ptrdiff_t);
Void_t *(*__morecore)(ptrdiff_t) = __default_morecore;
#else
Void_t * __default_morecore ();
Void_t *(*__morecore)() = __default_morecore;
#endif
#define MORECORE (*__morecore)
#define MORECORE_FAILURE 0
#ifndef MORECORE_CLEARS
#define MORECORE_CLEARS 1
#endif
static size_t __libc_pagesize;
#define access __access
#define mmap __mmap
#define munmap __munmap
#define mremap __mremap
#define mprotect __mprotect
#undef malloc_getpagesize
#define malloc_getpagesize __libc_pagesize
#else /* _LIBC */
#if __STD_C
extern Void_t* sbrk(ptrdiff_t);
#else
extern Void_t* sbrk();
#endif
#ifndef MORECORE
#define MORECORE sbrk
#endif
#ifndef MORECORE_FAILURE
#define MORECORE_FAILURE -1
#endif
#ifndef MORECORE_CLEARS
#define MORECORE_CLEARS 1
#endif
#endif /* _LIBC */
#ifdef _LIBC
#define cALLOc __libc_calloc
#define fREe __libc_free
#define mALLOc __libc_malloc
#define mEMALIGn __libc_memalign
#define rEALLOc __libc_realloc
#define vALLOc __libc_valloc
#define pvALLOc __libc_pvalloc
#define mALLINFo __libc_mallinfo
#define mALLOPt __libc_mallopt
#define mALLOC_STATs __malloc_stats
#define mALLOC_USABLE_SIZe __malloc_usable_size
#define mALLOC_TRIm __malloc_trim
#define mALLOC_GET_STATe __malloc_get_state
#define mALLOC_SET_STATe __malloc_set_state
#else
#define cALLOc calloc
#define fREe free
#define mALLOc malloc
#define mEMALIGn memalign
#define rEALLOc realloc
#define vALLOc valloc
#define pvALLOc pvalloc
#define mALLINFo mallinfo
#define mALLOPt mallopt
#define mALLOC_STATs malloc_stats
#define mALLOC_USABLE_SIZe malloc_usable_size
#define mALLOC_TRIm malloc_trim
#define mALLOC_GET_STATe malloc_get_state
#define mALLOC_SET_STATe malloc_set_state
#endif
/* Public routines */
#if __STD_C
#ifndef _LIBC
void ptmalloc_init(void);
#endif
Void_t* mALLOc(size_t);
void fREe(Void_t*);
Void_t* rEALLOc(Void_t*, size_t);
Void_t* mEMALIGn(size_t, size_t);
Void_t* vALLOc(size_t);
Void_t* pvALLOc(size_t);
Void_t* cALLOc(size_t, size_t);
void cfree(Void_t*);
int mALLOC_TRIm(size_t);
size_t mALLOC_USABLE_SIZe(Void_t*);
void mALLOC_STATs(void);
int mALLOPt(int, int);
struct mallinfo mALLINFo(void);
Void_t* mALLOC_GET_STATe(void);
int mALLOC_SET_STATe(Void_t*);
#else /* !__STD_C */
#ifndef _LIBC
void ptmalloc_init();
#endif
Void_t* mALLOc();
void fREe();
Void_t* rEALLOc();
Void_t* mEMALIGn();
Void_t* vALLOc();
Void_t* pvALLOc();
Void_t* cALLOc();
void cfree();
int mALLOC_TRIm();
size_t mALLOC_USABLE_SIZe();
void mALLOC_STATs();
int mALLOPt();
struct mallinfo mALLINFo();
Void_t* mALLOC_GET_STATe();
int mALLOC_SET_STATe();
#endif /* __STD_C */
#ifdef __cplusplus
} /* end of extern "C" */
#endif
#if !defined(NO_THREADS) && !HAVE_MMAP
"Can't have threads support without mmap"
#endif
#if USE_ARENAS && !HAVE_MMAP
"Can't have multiple arenas without mmap"
#endif
/*
Type declarations
*/
struct malloc_chunk
{
INTERNAL_SIZE_T prev_size; /* Size of previous chunk (if free). */
INTERNAL_SIZE_T size; /* Size in bytes, including overhead. */
struct malloc_chunk* fd; /* double links -- used only if free. */
struct malloc_chunk* bk;
};
typedef struct malloc_chunk* mchunkptr;
/*
malloc_chunk details:
(The following includes lightly edited explanations by Colin Plumb.)
Chunks of memory are maintained using a `boundary tag' method as
described in e.g., Knuth or Standish. (See the paper by Paul
Wilson ftp://ftp.cs.utexas.edu/pub/garbage/allocsrv.ps for a
survey of such techniques.) Sizes of free chunks are stored both
in the front of each chunk and at the end. This makes
consolidating fragmented chunks into bigger chunks very fast. The
size fields also hold bits representing whether chunks are free or
in use.
An allocated chunk looks like this:
chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Size of previous chunk, if allocated | |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Size of chunk, in bytes |P|
mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| User data starts here... .
. .
. (malloc_usable_space() bytes) .
. |
nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Size of chunk |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Where "chunk" is the front of the chunk for the purpose of most of
the malloc code, but "mem" is the pointer that is returned to the
user. "Nextchunk" is the beginning of the next contiguous chunk.
Chunks always begin on even word boundaries, so the mem portion
(which is returned to the user) is also on an even word boundary, and
thus double-word aligned.
Free chunks are stored in circular doubly-linked lists, and look like this:
chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Size of previous chunk |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
`head:' | Size of chunk, in bytes |P|
mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Forward pointer to next chunk in list |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Back pointer to previous chunk in list |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Unused space (may be 0 bytes long) .
. .
. |
nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
`foot:' | Size of chunk, in bytes |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
The P (PREV_INUSE) bit, stored in the unused low-order bit of the
chunk size (which is always a multiple of two words), is an in-use
bit for the *previous* chunk. If that bit is *clear*, then the
word before the current chunk size contains the previous chunk
size, and can be used to find the front of the previous chunk.
(The very first chunk allocated always has this bit set,
preventing access to non-existent (or non-owned) memory.)
Note that the `foot' of the current chunk is actually represented
as the prev_size of the NEXT chunk. (This makes it easier to
deal with alignments etc).
The two exceptions to all this are
1. The special chunk `top', which doesn't bother using the
trailing size field since there is no
next contiguous chunk that would have to index off it. (After
initialization, `top' is forced to always exist. If it would
become less than MINSIZE bytes long, it is replenished via
malloc_extend_top.)
2. Chunks allocated via mmap, which have the second-lowest-order
bit (IS_MMAPPED) set in their size fields. Because they are
never merged or traversed from any other chunk, they have no
foot size or inuse information.
Available chunks are kept in any of several places (all declared below):
* `av': An array of chunks serving as bin headers for consolidated
chunks. Each bin is doubly linked. The bins are approximately
proportionally (log) spaced. There are a lot of these bins
(128). This may look excessive, but works very well in
practice. All procedures maintain the invariant that no
consolidated chunk physically borders another one. Chunks in
bins are kept in size order, with ties going to the
approximately least recently used chunk.
The chunks in each bin are maintained in decreasing sorted order by
size. This is irrelevant for the small bins, which all contain
the same-sized chunks, but facilitates best-fit allocation for
larger chunks. (These lists are just sequential. Keeping them in
order almost never requires enough traversal to warrant using
fancier ordered data structures.) Chunks of the same size are
linked with the most recently freed at the front, and allocations
are taken from the back. This results in LRU or FIFO allocation
order, which tends to give each chunk an equal opportunity to be
consolidated with adjacent freed chunks, resulting in larger free
chunks and less fragmentation.
* `top': The top-most available chunk (i.e., the one bordering the
end of available memory) is treated specially. It is never
included in any bin, is used only if no other chunk is
available, and is released back to the system if it is very
large (see M_TRIM_THRESHOLD).
* `last_remainder': A bin holding only the remainder of the
most recently split (non-top) chunk. This bin is checked
before other non-fitting chunks, so as to provide better
locality for runs of sequentially allocated chunks.
* Implicitly, through the host system's memory mapping tables.
If supported, requests greater than a threshold are usually
serviced via calls to mmap, and then later released via munmap.
*/
/*
Bins
The bins are an array of pairs of pointers serving as the
heads of (initially empty) doubly-linked lists of chunks, laid out
in a way so that each pair can be treated as if it were in a
malloc_chunk. (This way, the fd/bk offsets for linking bin heads
and chunks are the same).
Bins for sizes < 512 bytes contain chunks of all the same size, spaced
8 bytes apart. Larger bins are approximately logarithmically
spaced. (See the table below.)
Bin layout:
64 bins of size 8
32 bins of size 64
16 bins of size 512
8 bins of size 4096
4 bins of size 32768
2 bins of size 262144
1 bin of size what's left
There is actually a little bit of slop in the numbers in bin_index
for the sake of speed. This makes no difference elsewhere.
The special chunks `top' and `last_remainder' get their own bins,
(this is implemented via yet more trickery with the av array),
although `top' is never properly linked to its bin since it is
always handled specially.
*/
#define NAV 128 /* number of bins */
typedef struct malloc_chunk* mbinptr;
/* An arena is a configuration of malloc_chunks together with an array
of bins. With multiple threads, it must be locked via a mutex
before changing its data structures. One or more `heaps' are
associated with each arena, except for the main_arena, which is
associated only with the `main heap', i.e. the conventional free
store obtained with calls to MORECORE() (usually sbrk). The `av'
array is never mentioned directly in the code, but instead used via
bin access macros. */
typedef struct _arena {
mbinptr av[2*NAV + 2];
struct _arena *next;
size_t size;
#if THREAD_STATS
long stat_lock_direct, stat_lock_loop, stat_lock_wait;
#endif
mutex_t mutex;
} arena;
/* A heap is a single contiguous memory region holding (coalesceable)
malloc_chunks. It is allocated with mmap() and always starts at an
address aligned to HEAP_MAX_SIZE. Not used unless compiling with
USE_ARENAS. */
typedef struct _heap_info {
arena *ar_ptr; /* Arena for this heap. */
struct _heap_info *prev; /* Previous heap. */
size_t size; /* Current size in bytes. */
size_t pad; /* Make sure the following data is properly aligned. */
} heap_info;
/*
Static functions (forward declarations)
*/
#if __STD_C
static void chunk_free(arena *ar_ptr, mchunkptr p) internal_function;
static mchunkptr chunk_alloc(arena *ar_ptr, INTERNAL_SIZE_T size)
internal_function;
static mchunkptr chunk_realloc(arena *ar_ptr, mchunkptr oldp,
INTERNAL_SIZE_T oldsize, INTERNAL_SIZE_T nb)
internal_function;
static mchunkptr chunk_align(arena *ar_ptr, INTERNAL_SIZE_T nb,
size_t alignment) internal_function;
static int main_trim(size_t pad) internal_function;
#if USE_ARENAS
static int heap_trim(heap_info *heap, size_t pad) internal_function;
#endif
#if defined _LIBC || defined MALLOC_HOOKS
static Void_t* malloc_check(size_t sz, const Void_t *caller);
static void free_check(Void_t* mem, const Void_t *caller);
static Void_t* realloc_check(Void_t* oldmem, size_t bytes,
const Void_t *caller);
static Void_t* memalign_check(size_t alignment, size_t bytes,
const Void_t *caller);
#ifndef NO_THREADS
static Void_t* malloc_starter(size_t sz, const Void_t *caller);
static void free_starter(Void_t* mem, const Void_t *caller);
static Void_t* malloc_atfork(size_t sz, const Void_t *caller);
static void free_atfork(Void_t* mem, const Void_t *caller);
#endif
#endif
#else
static void chunk_free();
static mchunkptr chunk_alloc();
static mchunkptr chunk_realloc();
static mchunkptr chunk_align();
static int main_trim();
#if USE_ARENAS
static int heap_trim();
#endif
#if defined _LIBC || defined MALLOC_HOOKS
static Void_t* malloc_check();
static void free_check();
static Void_t* realloc_check();
static Void_t* memalign_check();
#ifndef NO_THREADS
static Void_t* malloc_starter();
static void free_starter();
static Void_t* malloc_atfork();
static void free_atfork();
#endif
#endif
#endif
/* sizes, alignments */
#define SIZE_SZ (sizeof(INTERNAL_SIZE_T))
/* Allow the default to be overwritten on the compiler command line. */
#ifndef MALLOC_ALIGNMENT
# define MALLOC_ALIGNMENT (SIZE_SZ + SIZE_SZ)
#endif
#define MALLOC_ALIGN_MASK (MALLOC_ALIGNMENT - 1)
#define MINSIZE (sizeof(struct malloc_chunk))
/* conversion from malloc headers to user pointers, and back */
#define chunk2mem(p) ((Void_t*)((char*)(p) + 2*SIZE_SZ))
#define mem2chunk(mem) chunk_at_offset((mem), -2*SIZE_SZ)
/* pad request bytes into a usable size, return non-zero on overflow */
#define request2size(req, nb) \
((nb = (req) + (SIZE_SZ + MALLOC_ALIGN_MASK)),\
((long)nb <= 0 || nb < (INTERNAL_SIZE_T) (req) \
? (__set_errno (ENOMEM), 1) \
: ((nb < (MINSIZE + MALLOC_ALIGN_MASK) \
? (nb = MINSIZE) : (nb &= ~MALLOC_ALIGN_MASK)), 0)))
/* Check if m has acceptable alignment */
#define aligned_OK(m) (((unsigned long)((m)) & (MALLOC_ALIGN_MASK)) == 0)
/*
Physical chunk operations
*/
/* size field is or'ed with PREV_INUSE when previous adjacent chunk in use */
#define PREV_INUSE 0x1UL
/* size field is or'ed with IS_MMAPPED if the chunk was obtained with mmap() */
#define IS_MMAPPED 0x2UL
/* Bits to mask off when extracting size */
#define SIZE_BITS (PREV_INUSE|IS_MMAPPED)
/* Ptr to next physical malloc_chunk. */
#define next_chunk(p) chunk_at_offset((p), (p)->size & ~PREV_INUSE)
/* Ptr to previous physical malloc_chunk */
#define prev_chunk(p) chunk_at_offset((p), -(p)->prev_size)
/* Treat space at ptr + offset as a chunk */
#define chunk_at_offset(p, s) BOUNDED_1((mchunkptr)(((char*)(p)) + (s)))
/*
Dealing with use bits
*/
/* extract p's inuse bit */
#define inuse(p) (next_chunk(p)->size & PREV_INUSE)
/* extract inuse bit of previous chunk */
#define prev_inuse(p) ((p)->size & PREV_INUSE)
/* check for mmap()'ed chunk */
#define chunk_is_mmapped(p) ((p)->size & IS_MMAPPED)
/* set/clear chunk as in use without otherwise disturbing */
#define set_inuse(p) (next_chunk(p)->size |= PREV_INUSE)
#define clear_inuse(p) (next_chunk(p)->size &= ~PREV_INUSE)
/* check/set/clear inuse bits in known places */
#define inuse_bit_at_offset(p, s) \
(chunk_at_offset((p), (s))->size & PREV_INUSE)
#define set_inuse_bit_at_offset(p, s) \
(chunk_at_offset((p), (s))->size |= PREV_INUSE)
#define clear_inuse_bit_at_offset(p, s) \
(chunk_at_offset((p), (s))->size &= ~(PREV_INUSE))
/*
Dealing with size fields
*/
/* Get size, ignoring use bits */
#define chunksize(p) ((p)->size & ~(SIZE_BITS))
/* Set size at head, without disturbing its use bit */
#define set_head_size(p, s) ((p)->size = (((p)->size & PREV_INUSE) | (s)))
/* Set size/use ignoring previous bits in header */
#define set_head(p, s) ((p)->size = (s))
/* Set size at footer (only when chunk is not in use) */
#define set_foot(p, s) (chunk_at_offset(p, s)->prev_size = (s))
/* access macros */
#define bin_at(a, i) BOUNDED_1(_bin_at(a, i))
#define _bin_at(a, i) ((mbinptr)((char*)&(((a)->av)[2*(i)+2]) - 2*SIZE_SZ))
#define init_bin(a, i) ((a)->av[2*(i)+2] = (a)->av[2*(i)+3] = bin_at((a), (i)))
#define next_bin(b) ((mbinptr)((char*)(b) + 2 * sizeof(((arena*)0)->av[0])))
#define prev_bin(b) ((mbinptr)((char*)(b) - 2 * sizeof(((arena*)0)->av[0])))
/*
The first 2 bins are never indexed. The corresponding av cells are instead
used for bookkeeping. This is not to save space, but to simplify
indexing, maintain locality, and avoid some initialization tests.
*/
#define binblocks(a) (bin_at(a,0)->size)/* bitvector of nonempty blocks */
#define top(a) (bin_at(a,0)->fd) /* The topmost chunk */
#define last_remainder(a) (bin_at(a,1)) /* remainder from last split */
/*
Because top initially points to its own bin with initial
zero size, thus forcing extension on the first malloc request,
we avoid having any special code in malloc to check whether
it even exists yet. But we still need to in malloc_extend_top.
*/
#define initial_top(a) ((mchunkptr)bin_at(a, 0))
/* field-extraction macros */
#define first(b) ((b)->fd)
#define last(b) ((b)->bk)
/*
Indexing into bins
*/
#define bin_index(sz) \
(((((unsigned long)(sz)) >> 9) == 0) ? (((unsigned long)(sz)) >> 3):\
((((unsigned long)(sz)) >> 9) <= 4) ? 56 + (((unsigned long)(sz)) >> 6):\
((((unsigned long)(sz)) >> 9) <= 20) ? 91 + (((unsigned long)(sz)) >> 9):\
((((unsigned long)(sz)) >> 9) <= 84) ? 110 + (((unsigned long)(sz)) >> 12):\
((((unsigned long)(sz)) >> 9) <= 340) ? 119 + (((unsigned long)(sz)) >> 15):\
((((unsigned long)(sz)) >> 9) <= 1364) ? 124 + (((unsigned long)(sz)) >> 18):\
126)
/*
bins for chunks < 512 are all spaced 8 bytes apart, and hold
identically sized chunks. This is exploited in malloc.
*/
#define MAX_SMALLBIN 63
#define MAX_SMALLBIN_SIZE 512
#define SMALLBIN_WIDTH 8
#define smallbin_index(sz) (((unsigned long)(sz)) >> 3)
/*
Requests are `small' if both the corresponding and the next bin are small
*/
#define is_small_request(nb) ((nb) < MAX_SMALLBIN_SIZE - SMALLBIN_WIDTH)
/*
To help compensate for the large number of bins, a one-level index
structure is used for bin-by-bin searching. `binblocks' is a
one-word bitvector recording whether groups of BINBLOCKWIDTH bins
have any (possibly) non-empty bins, so they can be skipped over
all at once during during traversals. The bits are NOT always
cleared as soon as all bins in a block are empty, but instead only
when all are noticed to be empty during traversal in malloc.
*/
#define BINBLOCKWIDTH 4 /* bins per block */
/* bin<->block macros */
#define idx2binblock(ix) ((unsigned)1 << ((ix) / BINBLOCKWIDTH))
#define mark_binblock(a, ii) (binblocks(a) |= idx2binblock(ii))
#define clear_binblock(a, ii) (binblocks(a) &= ~(idx2binblock(ii)))
/* Static bookkeeping data */
/* Helper macro to initialize bins */
#define IAV(i) _bin_at(&main_arena, i), _bin_at(&main_arena, i)
static arena main_arena = {
{
0, 0,
IAV(0), IAV(1), IAV(2), IAV(3), IAV(4), IAV(5), IAV(6), IAV(7),
IAV(8), IAV(9), IAV(10), IAV(11), IAV(12), IAV(13), IAV(14), IAV(15),
IAV(16), IAV(17), IAV(18), IAV(19), IAV(20), IAV(21), IAV(22), IAV(23),
IAV(24), IAV(25), IAV(26), IAV(27), IAV(28), IAV(29), IAV(30), IAV(31),
IAV(32), IAV(33), IAV(34), IAV(35), IAV(36), IAV(37), IAV(38), IAV(39),
IAV(40), IAV(41), IAV(42), IAV(43), IAV(44), IAV(45), IAV(46), IAV(47),
IAV(48), IAV(49), IAV(50), IAV(51), IAV(52), IAV(53), IAV(54), IAV(55),
IAV(56), IAV(57), IAV(58), IAV(59), IAV(60), IAV(61), IAV(62), IAV(63),
IAV(64), IAV(65), IAV(66), IAV(67), IAV(68), IAV(69), IAV(70), IAV(71),
IAV(72), IAV(73), IAV(74), IAV(75), IAV(76), IAV(77), IAV(78), IAV(79),
IAV(80), IAV(81), IAV(82), IAV(83), IAV(84), IAV(85), IAV(86), IAV(87),
IAV(88), IAV(89), IAV(90), IAV(91), IAV(92), IAV(93), IAV(94), IAV(95),
IAV(96), IAV(97), IAV(98), IAV(99), IAV(100), IAV(101), IAV(102), IAV(103),
IAV(104), IAV(105), IAV(106), IAV(107), IAV(108), IAV(109), IAV(110), IAV(111),
IAV(112), IAV(113), IAV(114), IAV(115), IAV(116), IAV(117), IAV(118), IAV(119),
IAV(120), IAV(121), IAV(122), IAV(123), IAV(124), IAV(125), IAV(126), IAV(127)
},
&main_arena, /* next */
0, /* size */
#if THREAD_STATS
0, 0, 0, /* stat_lock_direct, stat_lock_loop, stat_lock_wait */
#endif
MUTEX_INITIALIZER /* mutex */
};
#undef IAV
/* Thread specific data */
static tsd_key_t arena_key;
static mutex_t list_lock = MUTEX_INITIALIZER;
#if THREAD_STATS
static int stat_n_heaps;
#define THREAD_STAT(x) x
#else
#define THREAD_STAT(x) do ; while(0)
#endif
/* variables holding tunable values */
static unsigned long trim_threshold = DEFAULT_TRIM_THRESHOLD;
static unsigned long top_pad = DEFAULT_TOP_PAD;
static unsigned int n_mmaps_max = DEFAULT_MMAP_MAX;
static unsigned long mmap_threshold = DEFAULT_MMAP_THRESHOLD;
static int check_action = DEFAULT_CHECK_ACTION;
/* The first value returned from sbrk */
static char* sbrk_base = (char*)(-1);
/* The maximum memory obtained from system via sbrk */
static unsigned long max_sbrked_mem;
/* The maximum via either sbrk or mmap (too difficult to track with threads) */
#ifdef NO_THREADS
static unsigned long max_total_mem;
#endif
/* The total memory obtained from system via sbrk */
#define sbrked_mem (main_arena.size)
/* Tracking mmaps */
static unsigned int n_mmaps;
static unsigned int max_n_mmaps;
static unsigned long mmapped_mem;
static unsigned long max_mmapped_mem;
/* Mapped memory in non-main arenas (reliable only for NO_THREADS). */
static unsigned long arena_mem;
#ifndef _LIBC
#define weak_variable
#else
/* In GNU libc we want the hook variables to be weak definitions to
avoid a problem with Emacs. */
#define weak_variable weak_function
#endif
/* Already initialized? */
int __malloc_initialized = -1;
#ifndef NO_THREADS
/* Magic value for the thread-specific arena pointer when
malloc_atfork() is in use. */
#define ATFORK_ARENA_PTR ((Void_t*)-1)
/* The following two functions are registered via thread_atfork() to
make sure that the mutexes remain in a consistent state in the
fork()ed version of a thread. Also adapt the malloc and free hooks
temporarily, because the `atfork' handler mechanism may use
malloc/free internally (e.g. in LinuxThreads). */
#if defined _LIBC || defined MALLOC_HOOKS
static __malloc_ptr_t (*save_malloc_hook) __MALLOC_P ((size_t __size,
const __malloc_ptr_t));
static void (*save_free_hook) __MALLOC_P ((__malloc_ptr_t __ptr,
const __malloc_ptr_t));
static Void_t* save_arena;
#endif
static void
ptmalloc_lock_all __MALLOC_P((void))
{
arena *ar_ptr;
(void)mutex_lock(&list_lock);
for(ar_ptr = &main_arena;;) {
(void)mutex_lock(&ar_ptr->mutex);
ar_ptr = ar_ptr->next;
if(ar_ptr == &main_arena) break;
}
#if defined _LIBC || defined MALLOC_HOOKS
save_malloc_hook = __malloc_hook;
save_free_hook = __free_hook;
__malloc_hook = malloc_atfork;
__free_hook = free_atfork;
/* Only the current thread may perform malloc/free calls now. */
tsd_getspecific(arena_key, save_arena);
tsd_setspecific(arena_key, ATFORK_ARENA_PTR);
#endif
}
static void
ptmalloc_unlock_all __MALLOC_P((void))
{
arena *ar_ptr;
#if defined _LIBC || defined MALLOC_HOOKS
tsd_setspecific(arena_key, save_arena);
__malloc_hook = save_malloc_hook;
__free_hook = save_free_hook;
#endif
for(ar_ptr = &main_arena;;) {
(void)mutex_unlock(&ar_ptr->mutex);
ar_ptr = ar_ptr->next;
if(ar_ptr == &main_arena) break;
}
(void)mutex_unlock(&list_lock);
}
static void
ptmalloc_init_all __MALLOC_P((void))
{
arena *ar_ptr;
#if defined _LIBC || defined MALLOC_HOOKS
tsd_setspecific(arena_key, save_arena);
__malloc_hook = save_malloc_hook;
__free_hook = save_free_hook;
#endif
for(ar_ptr = &main_arena;;) {
(void)mutex_init(&ar_ptr->mutex);
ar_ptr = ar_ptr->next;
if(ar_ptr == &main_arena) break;
}
(void)mutex_init(&list_lock);
}
#endif /* !defined NO_THREADS */
/* Initialization routine. */
#if defined(_LIBC)
#if 0
static void ptmalloc_init __MALLOC_P ((void)) __attribute__ ((constructor));
#endif
#ifdef _LIBC
#include <string.h>
extern char **environ;
static char *
internal_function
next_env_entry (char ***position)
{
char **current = *position;
char *result = NULL;
while (*current != NULL)
{
if (__builtin_expect ((*current)[0] == 'M', 0)
&& (*current)[1] == 'A'
&& (*current)[2] == 'L'
&& (*current)[3] == 'L'
&& (*current)[4] == 'O'
&& (*current)[5] == 'C'
&& (*current)[6] == '_')
{
result = &(*current)[7];
/* Save current position for next visit. */
*position = ++current;
break;
}
++current;
}
return result;
}
#endif
static void
ptmalloc_init __MALLOC_P((void))
#else
void
ptmalloc_init __MALLOC_P((void))
#endif
{
#if defined _LIBC || defined MALLOC_HOOKS
# if __STD_C
const char* s;
# else
char* s;
# endif
#endif
int secure;
if(__malloc_initialized >= 0) return;
__malloc_initialized = 0;
#ifdef _LIBC
__libc_pagesize = __getpagesize();
#endif
#ifndef NO_THREADS
#if defined _LIBC || defined MALLOC_HOOKS
/* With some threads implementations, creating thread-specific data
or initializing a mutex may call malloc() itself. Provide a
simple starter version (realloc() won't work). */
save_malloc_hook = __malloc_hook;
save_free_hook = __free_hook;
__malloc_hook = malloc_starter;
__free_hook = free_starter;
#endif
#ifdef _LIBC
/* Initialize the pthreads interface. */
if (__pthread_initialize != NULL)
__pthread_initialize();
#endif
#endif /* !defined NO_THREADS */
mutex_init(&main_arena.mutex);
mutex_init(&list_lock);
tsd_key_create(&arena_key, NULL);
tsd_setspecific(arena_key, (Void_t *)&main_arena);
thread_atfork(ptmalloc_lock_all, ptmalloc_unlock_all, ptmalloc_init_all);
#if defined _LIBC || defined MALLOC_HOOKS
#ifndef NO_THREADS
__malloc_hook = save_malloc_hook;
__free_hook = save_free_hook;
#endif
secure = __libc_enable_secure;
#ifdef _LIBC
s = NULL;
if (environ != NULL)
{
char **runp = environ;
char *envline;
while (__builtin_expect ((envline = next_env_entry (&runp)) != NULL, 0))
{
size_t len = strcspn (envline, "=");
if (envline[len] != '=')
/* This is a "MALLOC_" variable at the end of the string
without a '=' character. Ignore it since otherwise we
will access invalid memory below. */
continue;
switch (len)
{
case 6:
if (memcmp (envline, "CHECK_", 6) == 0)
s = &envline[7];
break;
case 8:
if (! secure && memcmp (envline, "TOP_PAD_", 8) == 0)
mALLOPt(M_TOP_PAD, atoi(&envline[9]));
break;
case 9:
if (! secure && memcmp (envline, "MMAP_MAX_", 9) == 0)
mALLOPt(M_MMAP_MAX, atoi(&envline[10]));
break;
case 15:
if (! secure)
{
if (memcmp (envline, "TRIM_THRESHOLD_", 15) == 0)
mALLOPt(M_TRIM_THRESHOLD, atoi(&envline[16]));
else if (memcmp (envline, "MMAP_THRESHOLD_", 15) == 0)
mALLOPt(M_MMAP_THRESHOLD, atoi(&envline[16]));
}
break;
default:
break;
}
}
}
#else
if (! secure)
{
if((s = getenv("MALLOC_TRIM_THRESHOLD_")))
mALLOPt(M_TRIM_THRESHOLD, atoi(s));
if((s = getenv("MALLOC_TOP_PAD_")))
mALLOPt(M_TOP_PAD, atoi(s));
if((s = getenv("MALLOC_MMAP_THRESHOLD_")))
mALLOPt(M_MMAP_THRESHOLD, atoi(s));
if((s = getenv("MALLOC_MMAP_MAX_")))
mALLOPt(M_MMAP_MAX, atoi(s));
}
s = getenv("MALLOC_CHECK_");
#endif
if(s) {
if(s[0]) mALLOPt(M_CHECK_ACTION, (int)(s[0] - '0'));
__malloc_check_init();
}
if(__malloc_initialize_hook != NULL)
(*__malloc_initialize_hook)();
#endif
__malloc_initialized = 1;
}
/* There are platforms (e.g. Hurd) with a link-time hook mechanism. */
#ifdef thread_atfork_static
thread_atfork_static(ptmalloc_lock_all, ptmalloc_unlock_all, \
ptmalloc_init_all)
#endif
#if defined _LIBC || defined MALLOC_HOOKS
/* Hooks for debugging versions. The initial hooks just call the
initialization routine, then do the normal work. */
static Void_t*
#if __STD_C
malloc_hook_ini(size_t sz, const __malloc_ptr_t caller)
#else
malloc_hook_ini(sz, caller)
size_t sz; const __malloc_ptr_t caller;
#endif
{
__malloc_hook = NULL;
ptmalloc_init();
return mALLOc(sz);
}
static Void_t*
#if __STD_C
realloc_hook_ini(Void_t* ptr, size_t sz, const __malloc_ptr_t caller)
#else
realloc_hook_ini(ptr, sz, caller)
Void_t* ptr; size_t sz; const __malloc_ptr_t caller;
#endif
{
__malloc_hook = NULL;
__realloc_hook = NULL;
ptmalloc_init();
return rEALLOc(ptr, sz);
}
static Void_t*
#if __STD_C
memalign_hook_ini(size_t alignment, size_t sz, const __malloc_ptr_t caller)
#else
memalign_hook_ini(alignment, sz, caller)
size_t alignment; size_t sz; const __malloc_ptr_t caller;
#endif
{
__memalign_hook = NULL;
ptmalloc_init();
return mEMALIGn(alignment, sz);
}
void weak_variable (*__malloc_initialize_hook) __MALLOC_P ((void)) = NULL;
void weak_variable (*__free_hook) __MALLOC_P ((__malloc_ptr_t __ptr,
const __malloc_ptr_t)) = NULL;
__malloc_ptr_t weak_variable (*__malloc_hook)
__MALLOC_P ((size_t __size, const __malloc_ptr_t)) = malloc_hook_ini;
__malloc_ptr_t weak_variable (*__realloc_hook)
__MALLOC_P ((__malloc_ptr_t __ptr, size_t __size, const __malloc_ptr_t))
= realloc_hook_ini;
__malloc_ptr_t weak_variable (*__memalign_hook)
__MALLOC_P ((size_t __alignment, size_t __size, const __malloc_ptr_t))
= memalign_hook_ini;
void weak_variable (*__after_morecore_hook) __MALLOC_P ((void)) = NULL;
/* Whether we are using malloc checking. */
static int using_malloc_checking;
/* A flag that is set by malloc_set_state, to signal that malloc checking
must not be enabled on the request from the user (via the MALLOC_CHECK_
environment variable). It is reset by __malloc_check_init to tell
malloc_set_state that the user has requested malloc checking.
The purpose of this flag is to make sure that malloc checking is not
enabled when the heap to be restored was constructed without malloc
checking, and thus does not contain the required magic bytes.
Otherwise the heap would be corrupted by calls to free and realloc. If
it turns out that the heap was created with malloc checking and the
user has requested it malloc_set_state just calls __malloc_check_init
again to enable it. On the other hand, reusing such a heap without
further malloc checking is safe. */
static int disallow_malloc_check;
/* Activate a standard set of debugging hooks. */
void
__malloc_check_init()
{
if (disallow_malloc_check) {
disallow_malloc_check = 0;
return;
}
using_malloc_checking = 1;
__malloc_hook = malloc_check;
__free_hook = free_check;
__realloc_hook = realloc_check;
__memalign_hook = memalign_check;
if(check_action & 1)
fprintf(stderr, "malloc: using debugging hooks\n");
}
#endif
/* Routines dealing with mmap(). */
#if HAVE_MMAP
#ifndef MAP_ANONYMOUS
static int dev_zero_fd = -1; /* Cached file descriptor for /dev/zero. */
#define MMAP(addr, size, prot, flags) ((dev_zero_fd < 0) ? \
(dev_zero_fd = open("/dev/zero", O_RDWR), \
mmap((addr), (size), (prot), (flags), dev_zero_fd, 0)) : \
mmap((addr), (size), (prot), (flags), dev_zero_fd, 0))
#else
#define MMAP(addr, size, prot, flags) \
(mmap((addr), (size), (prot), (flags)|MAP_ANONYMOUS, -1, 0))
#endif
#if defined __GNUC__ && __GNUC__ >= 2
/* This function is only called from one place, inline it. */
__inline__
#endif
static mchunkptr
internal_function
#if __STD_C
mmap_chunk(size_t size)
#else
mmap_chunk(size) size_t size;
#endif
{
size_t page_mask = malloc_getpagesize - 1;
mchunkptr p;
/* For mmapped chunks, the overhead is one SIZE_SZ unit larger, because
* there is no following chunk whose prev_size field could be used.
*/
size = (size + SIZE_SZ + page_mask) & ~page_mask;
p = (mchunkptr)MMAP(0, size, PROT_READ|PROT_WRITE, MAP_PRIVATE);
if(p == (mchunkptr) MAP_FAILED) return 0;
n_mmaps++;
if (n_mmaps > max_n_mmaps) max_n_mmaps = n_mmaps;
/* We demand that eight bytes into a page must be 8-byte aligned. */
assert(aligned_OK(chunk2mem(p)));
/* The offset to the start of the mmapped region is stored
* in the prev_size field of the chunk; normally it is zero,
* but that can be changed in memalign().
*/
p->prev_size = 0;
set_head(p, size|IS_MMAPPED);
mmapped_mem += size;
if ((unsigned long)mmapped_mem > (unsigned long)max_mmapped_mem)
max_mmapped_mem = mmapped_mem;
#ifdef NO_THREADS
if ((unsigned long)(mmapped_mem + arena_mem + sbrked_mem) > max_total_mem)
max_total_mem = mmapped_mem + arena_mem + sbrked_mem;
#endif
return p;
}
static void
internal_function
#if __STD_C
munmap_chunk(mchunkptr p)
#else
munmap_chunk(p) mchunkptr p;
#endif
{
INTERNAL_SIZE_T size = chunksize(p);
int ret;
assert (chunk_is_mmapped(p));
assert(! ((char*)p >= sbrk_base && (char*)p < sbrk_base + sbrked_mem));
assert((n_mmaps > 0));
assert(((p->prev_size + size) & (malloc_getpagesize-1)) == 0);
n_mmaps--;
mmapped_mem -= (size + p->prev_size);
ret = munmap((char *)p - p->prev_size, size + p->prev_size);
/* munmap returns non-zero on failure */
assert(ret == 0);
}
#if HAVE_MREMAP
static mchunkptr
internal_function
#if __STD_C
mremap_chunk(mchunkptr p, size_t new_size)
#else
mremap_chunk(p, new_size) mchunkptr p; size_t new_size;
#endif
{
size_t page_mask = malloc_getpagesize - 1;
INTERNAL_SIZE_T offset = p->prev_size;
INTERNAL_SIZE_T size = chunksize(p);
char *cp;
assert (chunk_is_mmapped(p));
assert(! ((char*)p >= sbrk_base && (char*)p < sbrk_base + sbrked_mem));
assert((n_mmaps > 0));
assert(((size + offset) & (malloc_getpagesize-1)) == 0);
/* Note the extra SIZE_SZ overhead as in mmap_chunk(). */
new_size = (new_size + offset + SIZE_SZ + page_mask) & ~page_mask;
cp = (char *)mremap((char *)p - offset, size + offset, new_size,
MREMAP_MAYMOVE);
if (cp == MAP_FAILED) return 0;
p = (mchunkptr)(cp + offset);
assert(aligned_OK(chunk2mem(p)));
assert((p->prev_size == offset));
set_head(p, (new_size - offset)|IS_MMAPPED);
mmapped_mem -= size + offset;
mmapped_mem += new_size;
if ((unsigned long)mmapped_mem > (unsigned long)max_mmapped_mem)
max_mmapped_mem = mmapped_mem;
#ifdef NO_THREADS
if ((unsigned long)(mmapped_mem + arena_mem + sbrked_mem) > max_total_mem)
max_total_mem = mmapped_mem + arena_mem + sbrked_mem;
#endif
return p;
}
#endif /* HAVE_MREMAP */
#endif /* HAVE_MMAP */
/* Managing heaps and arenas (for concurrent threads) */
#if USE_ARENAS
/* Create a new heap. size is automatically rounded up to a multiple
of the page size. */
static heap_info *
internal_function
#if __STD_C
new_heap(size_t size)
#else
new_heap(size) size_t size;
#endif
{
size_t page_mask = malloc_getpagesize - 1;
char *p1, *p2;
unsigned long ul;
heap_info *h;
if(size+top_pad < HEAP_MIN_SIZE)
size = HEAP_MIN_SIZE;
else if(size+top_pad <= HEAP_MAX_SIZE)
size += top_pad;
else if(size > HEAP_MAX_SIZE)
return 0;
else
size = HEAP_MAX_SIZE;
size = (size + page_mask) & ~page_mask;
/* A memory region aligned to a multiple of HEAP_MAX_SIZE is needed.
No swap space needs to be reserved for the following large
mapping (on Linux, this is the case for all non-writable mappings
anyway). */
p1 = (char *)MMAP(0, HEAP_MAX_SIZE<<1, PROT_NONE, MAP_PRIVATE|MAP_NORESERVE);
if(p1 != MAP_FAILED) {
p2 = (char *)(((unsigned long)p1 + (HEAP_MAX_SIZE-1)) & ~(HEAP_MAX_SIZE-1));
ul = p2 - p1;
if (ul)
munmap(p1, ul);
munmap(p2 + HEAP_MAX_SIZE, HEAP_MAX_SIZE - ul);
} else {
/* Try to take the chance that an allocation of only HEAP_MAX_SIZE
is already aligned. */
p2 = (char *)MMAP(0, HEAP_MAX_SIZE, PROT_NONE, MAP_PRIVATE|MAP_NORESERVE);
if(p2 == MAP_FAILED)
return 0;
if((unsigned long)p2 & (HEAP_MAX_SIZE-1)) {
munmap(p2, HEAP_MAX_SIZE);
return 0;
}
}
if(MMAP(p2, size, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED)
== (char *) MAP_FAILED) {
munmap(p2, HEAP_MAX_SIZE);
return 0;
}
h = (heap_info *)p2;
h->size = size;
THREAD_STAT(stat_n_heaps++);
return h;
}
/* Grow or shrink a heap. size is automatically rounded up to a
multiple of the page size if it is positive. */
static int
#if __STD_C
grow_heap(heap_info *h, long diff)
#else
grow_heap(h, diff) heap_info *h; long diff;
#endif
{
size_t page_mask = malloc_getpagesize - 1;
long new_size;
if(diff >= 0) {
diff = (diff + page_mask) & ~page_mask;
new_size = (long)h->size + diff;
if(new_size > HEAP_MAX_SIZE)
return -1;
if(MMAP((char *)h + h->size, diff, PROT_READ|PROT_WRITE,
MAP_PRIVATE|MAP_FIXED) == (char *) MAP_FAILED)
return -2;
} else {
new_size = (long)h->size + diff;
if(new_size < (long)sizeof(*h))
return -1;
/* Try to re-map the extra heap space freshly to save memory, and
make it inaccessible. */
if((char *)MMAP((char *)h + new_size, -diff, PROT_NONE,
MAP_PRIVATE|MAP_FIXED) == (char *) MAP_FAILED)
return -2;
}
h->size = new_size;
return 0;
}
/* Delete a heap. */
#define delete_heap(heap) munmap((char*)(heap), HEAP_MAX_SIZE)
/* arena_get() acquires an arena and locks the corresponding mutex.
First, try the one last locked successfully by this thread. (This
is the common case and handled with a macro for speed.) Then, loop
once over the circularly linked list of arenas. If no arena is
readily available, create a new one. In this latter case, `size'
is just a hint as to how much memory will be required immediately
in the new arena. */
#define arena_get(ptr, size) do { \
Void_t *vptr = NULL; \
ptr = (arena *)tsd_getspecific(arena_key, vptr); \
if(ptr && !mutex_trylock(&ptr->mutex)) { \
THREAD_STAT(++(ptr->stat_lock_direct)); \
} else \
ptr = arena_get2(ptr, (size)); \
} while(0)
static arena *
internal_function
#if __STD_C
arena_get2(arena *a_tsd, size_t size)
#else
arena_get2(a_tsd, size) arena *a_tsd; size_t size;
#endif
{
arena *a;
heap_info *h;
char *ptr;
int i;
unsigned long misalign;
if(!a_tsd)
a = a_tsd = &main_arena;
else {
a = a_tsd->next;
if(!a) {
/* This can only happen while initializing the new arena. */
(void)mutex_lock(&main_arena.mutex);
THREAD_STAT(++(main_arena.stat_lock_wait));
return &main_arena;
}
}
/* Check the global, circularly linked list for available arenas. */
repeat:
do {
if(!mutex_trylock(&a->mutex)) {
THREAD_STAT(++(a->stat_lock_loop));
tsd_setspecific(arena_key, (Void_t *)a);
return a;
}
a = a->next;
} while(a != a_tsd);
/* If not even the list_lock can be obtained, try again. This can
happen during `atfork', or for example on systems where thread
creation makes it temporarily impossible to obtain _any_
locks. */
if(mutex_trylock(&list_lock)) {
a = a_tsd;
goto repeat;
}
(void)mutex_unlock(&list_lock);
/* Nothing immediately available, so generate a new arena. */
h = new_heap(size + (sizeof(*h) + sizeof(*a) + MALLOC_ALIGNMENT));
if(!h) {
/* Maybe size is too large to fit in a single heap. So, just try
to create a minimally-sized arena and let chunk_alloc() attempt
to deal with the large request via mmap_chunk(). */
h = new_heap(sizeof(*h) + sizeof(*a) + MALLOC_ALIGNMENT);
if(!h)
return 0;
}
a = h->ar_ptr = (arena *)(h+1);
for(i=0; i<NAV; i++)
init_bin(a, i);
a->next = NULL;
a->size = h->size;
arena_mem += h->size;
#ifdef NO_THREADS
if((unsigned long)(mmapped_mem + arena_mem + sbrked_mem) > max_total_mem)
max_total_mem = mmapped_mem + arena_mem + sbrked_mem;
#endif
tsd_setspecific(arena_key, (Void_t *)a);
mutex_init(&a->mutex);
i = mutex_lock(&a->mutex); /* remember result */
/* Set up the top chunk, with proper alignment. */
ptr = (char *)(a + 1);
misalign = (unsigned long)chunk2mem(ptr) & MALLOC_ALIGN_MASK;
if (misalign > 0)
ptr += MALLOC_ALIGNMENT - misalign;
top(a) = (mchunkptr)ptr;
set_head(top(a), (((char*)h + h->size) - ptr) | PREV_INUSE);
/* Add the new arena to the list. */
(void)mutex_lock(&list_lock);
a->next = main_arena.next;
main_arena.next = a;
(void)mutex_unlock(&list_lock);
if(i) /* locking failed; keep arena for further attempts later */
return 0;
THREAD_STAT(++(a->stat_lock_loop));
return a;
}
/* find the heap and corresponding arena for a given ptr */
#define heap_for_ptr(ptr) \
((heap_info *)((unsigned long)(ptr) & ~(HEAP_MAX_SIZE-1)))
#define arena_for_ptr(ptr) \
(((mchunkptr)(ptr) < top(&main_arena) && (char *)(ptr) >= sbrk_base) ? \
&main_arena : heap_for_ptr(ptr)->ar_ptr)
#else /* !USE_ARENAS */
/* There is only one arena, main_arena. */
#define arena_get(ptr, sz) (ptr = &main_arena)
#define arena_for_ptr(ptr) (&main_arena)
#endif /* USE_ARENAS */
/*
Debugging support
*/
#if MALLOC_DEBUG
/*
These routines make a number of assertions about the states
of data structures that should be true at all times. If any
are not true, it's very likely that a user program has somehow
trashed memory. (It's also possible that there is a coding error
in malloc. In which case, please report it!)
*/
#if __STD_C
static void do_check_chunk(arena *ar_ptr, mchunkptr p)
#else
static void do_check_chunk(ar_ptr, p) arena *ar_ptr; mchunkptr p;
#endif
{
INTERNAL_SIZE_T sz = p->size & ~PREV_INUSE;
/* No checkable chunk is mmapped */
assert(!chunk_is_mmapped(p));
#if USE_ARENAS
if(ar_ptr != &main_arena) {
heap_info *heap = heap_for_ptr(p);
assert(heap->ar_ptr == ar_ptr);
if(p != top(ar_ptr))
assert((char *)p + sz <= (char *)heap + heap->size);
else
assert((char *)p + sz == (char *)heap + heap->size);
return;
}
#endif
/* Check for legal address ... */
assert((char*)p >= sbrk_base);
if (p != top(ar_ptr))
assert((char*)p + sz <= (char*)top(ar_ptr));
else
assert((char*)p + sz <= sbrk_base + sbrked_mem);
}
#if __STD_C
static void do_check_free_chunk(arena *ar_ptr, mchunkptr p)
#else
static void do_check_free_chunk(ar_ptr, p) arena *ar_ptr; mchunkptr p;
#endif
{
INTERNAL_SIZE_T sz = p->size & ~PREV_INUSE;
mchunkptr next = chunk_at_offset(p, sz);
do_check_chunk(ar_ptr, p);
/* Check whether it claims to be free ... */
assert(!inuse(p));
/* Must have OK size and fields */
assert((long)sz >= (long)MINSIZE);
assert((sz & MALLOC_ALIGN_MASK) == 0);
assert(aligned_OK(chunk2mem(p)));
/* ... matching footer field */
assert(next->prev_size == sz);
/* ... and is fully consolidated */
assert(prev_inuse(p));
assert (next == top(ar_ptr) || inuse(next));
/* ... and has minimally sane links */
assert(p->fd->bk == p);
assert(p->bk->fd == p);
}
#if __STD_C
static void do_check_inuse_chunk(arena *ar_ptr, mchunkptr p)
#else
static void do_check_inuse_chunk(ar_ptr, p) arena *ar_ptr; mchunkptr p;
#endif
{
mchunkptr next = next_chunk(p);
do_check_chunk(ar_ptr, p);
/* Check whether it claims to be in use ... */
assert(inuse(p));
/* ... whether its size is OK (it might be a fencepost) ... */
assert(chunksize(p) >= MINSIZE || next->size == (0|PREV_INUSE));
/* ... and is surrounded by OK chunks.
Since more things can be checked with free chunks than inuse ones,
if an inuse chunk borders them and debug is on, it's worth doing them.
*/
if (!prev_inuse(p))
{
mchunkptr prv = prev_chunk(p);
assert(next_chunk(prv) == p);
do_check_free_chunk(ar_ptr, prv);
}
if (next == top(ar_ptr))
{
assert(prev_inuse(next));
assert(chunksize(next) >= MINSIZE);
}
else if (!inuse(next))
do_check_free_chunk(ar_ptr, next);
}
#if __STD_C
static void do_check_malloced_chunk(arena *ar_ptr,
mchunkptr p, INTERNAL_SIZE_T s)
#else
static void do_check_malloced_chunk(ar_ptr, p, s)
arena *ar_ptr; mchunkptr p; INTERNAL_SIZE_T s;
#endif
{
INTERNAL_SIZE_T sz = p->size & ~PREV_INUSE;
long room = sz - s;
do_check_inuse_chunk(ar_ptr, p);
/* Legal size ... */
assert((long)sz >= (long)MINSIZE);
assert((sz & MALLOC_ALIGN_MASK) == 0);
assert(room >= 0);
assert(room < (long)MINSIZE);
/* ... and alignment */
assert(aligned_OK(chunk2mem(p)));
/* ... and was allocated at front of an available chunk */
assert(prev_inuse(p));
}
#define check_free_chunk(A,P) do_check_free_chunk(A,P)
#define check_inuse_chunk(A,P) do_check_inuse_chunk(A,P)
#define check_chunk(A,P) do_check_chunk(A,P)
#define check_malloced_chunk(A,P,N) do_check_malloced_chunk(A,P,N)
#else
#define check_free_chunk(A,P)
#define check_inuse_chunk(A,P)
#define check_chunk(A,P)
#define check_malloced_chunk(A,P,N)
#endif
/*
Macro-based internal utilities
*/
/*
Linking chunks in bin lists.
Call these only with variables, not arbitrary expressions, as arguments.
*/
/*
Place chunk p of size s in its bin, in size order,
putting it ahead of others of same size.
*/
#define frontlink(A, P, S, IDX, BK, FD) \
{ \
if (S < MAX_SMALLBIN_SIZE) \
{ \
IDX = smallbin_index(S); \
mark_binblock(A, IDX); \
BK = bin_at(A, IDX); \
FD = BK->fd; \
P->bk = BK; \
P->fd = FD; \
FD->bk = BK->fd = P; \
} \
else \
{ \
IDX = bin_index(S); \
BK = bin_at(A, IDX); \
FD = BK->fd; \
if (FD == BK) mark_binblock(A, IDX); \
else \
{ \
while (FD != BK && S < chunksize(FD)) FD = FD->fd; \
BK = FD->bk; \
} \
P->bk = BK; \
P->fd = FD; \
FD->bk = BK->fd = P; \
} \
}
/* take a chunk off a list */
#define unlink(P, BK, FD) \
{ \
BK = P->bk; \
FD = P->fd; \
FD->bk = BK; \
BK->fd = FD; \
} \
/* Place p as the last remainder */
#define link_last_remainder(A, P) \
{ \
last_remainder(A)->fd = last_remainder(A)->bk = P; \
P->fd = P->bk = last_remainder(A); \
}
/* Clear the last_remainder bin */
#define clear_last_remainder(A) \
(last_remainder(A)->fd = last_remainder(A)->bk = last_remainder(A))
/*
Extend the top-most chunk by obtaining memory from system.
Main interface to sbrk (but see also malloc_trim).
*/
#if defined __GNUC__ && __GNUC__ >= 2
/* This function is called only from one place, inline it. */
__inline__
#endif
static void
internal_function
#if __STD_C
malloc_extend_top(arena *ar_ptr, INTERNAL_SIZE_T nb)
#else
malloc_extend_top(ar_ptr, nb) arena *ar_ptr; INTERNAL_SIZE_T nb;
#endif
{
unsigned long pagesz = malloc_getpagesize;
mchunkptr old_top = top(ar_ptr); /* Record state of old top */
INTERNAL_SIZE_T old_top_size = chunksize(old_top);
INTERNAL_SIZE_T top_size; /* new size of top chunk */
#if USE_ARENAS
if(ar_ptr == &main_arena) {
#endif
char* brk; /* return value from sbrk */
INTERNAL_SIZE_T front_misalign; /* unusable bytes at front of sbrked space */
INTERNAL_SIZE_T correction; /* bytes for 2nd sbrk call */
char* new_brk; /* return of 2nd sbrk call */
char* old_end = (char*)(chunk_at_offset(old_top, old_top_size));
/* Pad request with top_pad plus minimal overhead */
INTERNAL_SIZE_T sbrk_size = nb + top_pad + MINSIZE;
/* If not the first time through, round to preserve page boundary */
/* Otherwise, we need to correct to a page size below anyway. */
/* (We also correct below if an intervening foreign sbrk call.) */
if (sbrk_base != (char*)(-1))
sbrk_size = (sbrk_size + (pagesz - 1)) & ~(pagesz - 1);
brk = (char*)(MORECORE (sbrk_size));
/* Fail if sbrk failed or if a foreign sbrk call killed our space */
if (brk == (char*)(MORECORE_FAILURE) ||
(brk < old_end && old_top != initial_top(&main_arena)))
return;
#if defined _LIBC || defined MALLOC_HOOKS
/* Call the `morecore' hook if necessary. */
if (__after_morecore_hook)
(*__after_morecore_hook) ();
#endif
sbrked_mem += sbrk_size;
if (brk == old_end) { /* can just add bytes to current top */
top_size = sbrk_size + old_top_size;
set_head(old_top, top_size | PREV_INUSE);
old_top = 0; /* don't free below */
} else {
if (sbrk_base == (char*)(-1)) /* First time through. Record base */
sbrk_base = brk;
else
/* Someone else called sbrk(). Count those bytes as sbrked_mem. */
sbrked_mem += brk - (char*)old_end;
/* Guarantee alignment of first new chunk made from this space */
front_misalign = (unsigned long)chunk2mem(brk) & MALLOC_ALIGN_MASK;
if (front_misalign > 0) {
correction = (MALLOC_ALIGNMENT) - front_misalign;
brk += correction;
} else
correction = 0;
/* Guarantee the next brk will be at a page boundary */
correction += pagesz - ((unsigned long)(brk + sbrk_size) & (pagesz - 1));
/* Allocate correction */
new_brk = (char*)(MORECORE (correction));
if (new_brk == (char*)(MORECORE_FAILURE)) return;
#if defined _LIBC || defined MALLOC_HOOKS
/* Call the `morecore' hook if necessary. */
if (__after_morecore_hook)
(*__after_morecore_hook) ();
#endif
sbrked_mem += correction;
top(&main_arena) = chunk_at_offset(brk, 0);
top_size = new_brk - brk + correction;
set_head(top(&main_arena), top_size | PREV_INUSE);
if (old_top == initial_top(&main_arena))
old_top = 0; /* don't free below */
}
if ((unsigned long)sbrked_mem > (unsigned long)max_sbrked_mem)
max_sbrked_mem = sbrked_mem;
#ifdef NO_THREADS
if ((unsigned long)(mmapped_mem + arena_mem + sbrked_mem) > max_total_mem)
max_total_mem = mmapped_mem + arena_mem + sbrked_mem;
#endif
#if USE_ARENAS
} else { /* ar_ptr != &main_arena */
heap_info *old_heap, *heap;
size_t old_heap_size;
if(old_top_size < MINSIZE) /* this should never happen */
return;
/* First try to extend the current heap. */
if(MINSIZE + nb <= old_top_size)
return;
old_heap = heap_for_ptr(old_top);
old_heap_size = old_heap->size;
if(grow_heap(old_heap, MINSIZE + nb - old_top_size) == 0) {
ar_ptr->size += old_heap->size - old_heap_size;
arena_mem += old_heap->size - old_heap_size;
#ifdef NO_THREADS
if(mmapped_mem + arena_mem + sbrked_mem > max_total_mem)
max_total_mem = mmapped_mem + arena_mem + sbrked_mem;
#endif
top_size = ((char *)old_heap + old_heap->size) - (char *)old_top;
set_head(old_top, top_size | PREV_INUSE);
return;
}
/* A new heap must be created. */
heap = new_heap(nb + (MINSIZE + sizeof(*heap)));
if(!heap)
return;
heap->ar_ptr = ar_ptr;
heap->prev = old_heap;
ar_ptr->size += heap->size;
arena_mem += heap->size;
#ifdef NO_THREADS
if((unsigned long)(mmapped_mem + arena_mem + sbrked_mem) > max_total_mem)
max_total_mem = mmapped_mem + arena_mem + sbrked_mem;
#endif
/* Set up the new top, so we can safely use chunk_free() below. */
top(ar_ptr) = chunk_at_offset(heap, sizeof(*heap));
top_size = heap->size - sizeof(*heap);
set_head(top(ar_ptr), top_size | PREV_INUSE);
}
#endif /* USE_ARENAS */
/* We always land on a page boundary */
assert(((unsigned long)((char*)top(ar_ptr) + top_size) & (pagesz-1)) == 0);
/* Setup fencepost and free the old top chunk. */
if(old_top) {
/* The fencepost takes at least MINSIZE bytes, because it might
become the top chunk again later. Note that a footer is set
up, too, although the chunk is marked in use. */
old_top_size -= MINSIZE;
set_head(chunk_at_offset(old_top, old_top_size + 2*SIZE_SZ), 0|PREV_INUSE);
if(old_top_size >= MINSIZE) {
set_head(chunk_at_offset(old_top, old_top_size), (2*SIZE_SZ)|PREV_INUSE);
set_foot(chunk_at_offset(old_top, old_top_size), (2*SIZE_SZ));
set_head_size(old_top, old_top_size);
chunk_free(ar_ptr, old_top);
} else {
set_head(old_top, (old_top_size + 2*SIZE_SZ)|PREV_INUSE);
set_foot(old_top, (old_top_size + 2*SIZE_SZ));
}
}
}
/* Main public routines */
/*
Malloc Algorithm:
The requested size is first converted into a usable form, `nb'.
This currently means to add 4 bytes overhead plus possibly more to
obtain 8-byte alignment and/or to obtain a size of at least
MINSIZE (currently 16, 24, or 32 bytes), the smallest allocatable
size. (All fits are considered `exact' if they are within MINSIZE
bytes.)
From there, the first successful of the following steps is taken:
1. The bin corresponding to the request size is scanned, and if
a chunk of exactly the right size is found, it is taken.
2. The most recently remaindered chunk is used if it is big
enough. This is a form of (roving) first fit, used only in
the absence of exact fits. Runs of consecutive requests use
the remainder of the chunk used for the previous such request
whenever possible. This limited use of a first-fit style
allocation strategy tends to give contiguous chunks
coextensive lifetimes, which improves locality and can reduce
fragmentation in the long run.
3. Other bins are scanned in increasing size order, using a
chunk big enough to fulfill the request, and splitting off
any remainder. This search is strictly by best-fit; i.e.,
the smallest (with ties going to approximately the least
recently used) chunk that fits is selected.
4. If large enough, the chunk bordering the end of memory
(`top') is split off. (This use of `top' is in accord with
the best-fit search rule. In effect, `top' is treated as
larger (and thus less well fitting) than any other available
chunk since it can be extended to be as large as necessary
(up to system limitations).
5. If the request size meets the mmap threshold and the
system supports mmap, and there are few enough currently
allocated mmapped regions, and a call to mmap succeeds,
the request is allocated via direct memory mapping.
6. Otherwise, the top of memory is extended by
obtaining more space from the system (normally using sbrk,
but definable to anything else via the MORECORE macro).
Memory is gathered from the system (in system page-sized
units) in a way that allows chunks obtained across different
sbrk calls to be consolidated, but does not require
contiguous memory. Thus, it should be safe to intersperse
mallocs with other sbrk calls.
All allocations are made from the `lowest' part of any found
chunk. (The implementation invariant is that prev_inuse is
always true of any allocated chunk; i.e., that each allocated
chunk borders either a previously allocated and still in-use chunk,
or the base of its memory arena.)
*/
#if __STD_C
Void_t* mALLOc(size_t bytes)
#else
Void_t* mALLOc(bytes) size_t bytes;
#endif
{
arena *ar_ptr;
INTERNAL_SIZE_T nb; /* padded request size */
mchunkptr victim;
#if defined _LIBC || defined MALLOC_HOOKS
__malloc_ptr_t (*hook) __MALLOC_PMT ((size_t, __const __malloc_ptr_t)) =
__malloc_hook;
if (hook != NULL) {
Void_t* result;
#if defined __GNUC__ && __GNUC__ >= 2
result = (*hook)(bytes, RETURN_ADDRESS (0));
#else
result = (*hook)(bytes, NULL);
#endif
return result;
}
#endif
if(request2size(bytes, nb))
return 0;
arena_get(ar_ptr, nb);
if(!ar_ptr)
return 0;
victim = chunk_alloc(ar_ptr, nb);
if(!victim) {
/* Maybe the failure is due to running out of mmapped areas. */
if(ar_ptr != &main_arena) {
(void)mutex_unlock(&ar_ptr->mutex);
(void)mutex_lock(&main_arena.mutex);
victim = chunk_alloc(&main_arena, nb);
(void)mutex_unlock(&main_arena.mutex);
} else {
#if USE_ARENAS
/* ... or sbrk() has failed and there is still a chance to mmap() */
ar_ptr = arena_get2(ar_ptr->next ? ar_ptr : 0, nb);
(void)mutex_unlock(&main_arena.mutex);
if(ar_ptr) {
victim = chunk_alloc(ar_ptr, nb);
(void)mutex_unlock(&ar_ptr->mutex);
}
#endif
}
if(!victim) return 0;
} else
(void)mutex_unlock(&ar_ptr->mutex);
return BOUNDED_N(chunk2mem(victim), bytes);
}
static mchunkptr
internal_function
#if __STD_C
chunk_alloc(arena *ar_ptr, INTERNAL_SIZE_T nb)
#else
chunk_alloc(ar_ptr, nb) arena *ar_ptr; INTERNAL_SIZE_T nb;
#endif
{
mchunkptr victim; /* inspected/selected chunk */
INTERNAL_SIZE_T victim_size; /* its size */
int idx; /* index for bin traversal */
mbinptr bin; /* associated bin */
mchunkptr remainder; /* remainder from a split */
long remainder_size; /* its size */
int remainder_index; /* its bin index */
unsigned long block; /* block traverser bit */
int startidx; /* first bin of a traversed block */
mchunkptr fwd; /* misc temp for linking */
mchunkptr bck; /* misc temp for linking */
mbinptr q; /* misc temp */
/* Check for exact match in a bin */
if (is_small_request(nb)) /* Faster version for small requests */
{
idx = smallbin_index(nb);
/* No traversal or size check necessary for small bins. */
q = _bin_at(ar_ptr, idx);
victim = last(q);
/* Also scan the next one, since it would have a remainder < MINSIZE */
if (victim == q)
{
q = next_bin(q);
victim = last(q);
}
if (victim != q)
{
victim_size = chunksize(victim);
unlink(victim, bck, fwd);
set_inuse_bit_at_offset(victim, victim_size);
check_malloced_chunk(ar_ptr, victim, nb);
return victim;
}
idx += 2; /* Set for bin scan below. We've already scanned 2 bins. */
}
else
{
idx = bin_index(nb);
bin = bin_at(ar_ptr, idx);
for (victim = last(bin); victim != bin; victim = victim->bk)
{
victim_size = chunksize(victim);
remainder_size = victim_size - nb;