blob: 65b2f16e83babbe0a91b303a93baafc487aa202e [file] [log] [blame]
This is a version (aka dlmalloc) of malloc/free/realloc written by
Doug Lea and released to the public domain, as explained at Send questions,
comments, complaints, performance data, etc to
* Version 2.8.6 Wed Aug 29 06:57:58 2012 Doug Lea
Note: There may be an updated version of this malloc obtainable at
Check before installing!
* Quickstart
This library is all in one file to simplify the most common usage:
ftp it, compile it (-O3), and link it into another program. All of
the compile-time options default to reasonable values for use on
most platforms. You might later want to step through various
compile-time and dynamic tuning options.
For convenience, an include file for code using this malloc is at:
You don't really need this .h file unless you call functions not
defined in your system include files. The .h file contains only the
excerpts from this file needed for using this malloc on ANSI C/C++
systems, so long as you haven't changed compile-time options about
naming and tuning parameters. If you do, then you can create your
own malloc.h that does include all settings by cutting at the point
indicated below. Note that you may already by default be using a C
library containing a malloc that is based on some version of this
malloc (for example in linux). You might still want to use the one
in this file to customize settings or to avoid overheads associated
with library versions.
* Vital statistics:
Supported pointer/size_t representation: 4 or 8 bytes
size_t MUST be an unsigned type of the same width as
pointers. (If you are using an ancient system that declares
size_t as a signed type, or need it to be a different width
than pointers, you can use a previous release of this malloc
(e.g. 2.7.2) supporting these.)
Alignment: 8 bytes (minimum)
This suffices for nearly all current machines and C compilers.
However, you can define MALLOC_ALIGNMENT to be wider than this
if necessary (up to 128bytes), at the expense of using more space.
Minimum overhead per allocated chunk: 4 or 8 bytes (if 4byte sizes)
8 or 16 bytes (if 8byte sizes)
Each malloced chunk has a hidden word of overhead holding size
and status information, and additional cross-check word
if FOOTERS is defined.
Minimum allocated size: 4-byte ptrs: 16 bytes (including overhead)
8-byte ptrs: 32 bytes (including overhead)
Even a request for zero bytes (i.e., malloc(0)) returns a
pointer to something of the minimum allocatable size.
The maximum overhead wastage (i.e., number of extra bytes
allocated than were requested in malloc) is less than or equal
to the minimum size, except for requests >= mmap_threshold that
are serviced via mmap(), where the worst case wastage is about
32 bytes plus the remainder from a system page (the minimal
mmap unit); typically 4096 or 8192 bytes.
Security: static-safe; optionally more or less
The "security" of malloc refers to the ability of malicious
code to accentuate the effects of errors (for example, freeing
space that is not currently malloc'ed or overwriting past the
ends of chunks) in code that calls malloc. This malloc
guarantees not to modify any memory locations below the base of
heap, i.e., static variables, even in the presence of usage
errors. The routines additionally detect most improper frees
and reallocs. All this holds as long as the static bookkeeping
for malloc itself is not corrupted by some other means. This
is only one aspect of security -- these checks do not, and
cannot, detect all possible programming errors.
If FOOTERS is defined nonzero, then each allocated chunk
carries an additional check word to verify that it was malloced
from its space. These check words are the same within each
execution of a program using malloc, but differ across
executions, so externally crafted fake chunks cannot be
freed. This improves security by rejecting frees/reallocs that
could corrupt heap memory, in addition to the checks preventing
writes to statics that are always on. This may further improve
security at the expense of time and space overhead. (Note that
FOOTERS may also be worth using with MSPACES.)
By default detected errors cause the program to abort (calling
"abort()"). You can override this to instead proceed past
errors by defining PROCEED_ON_ERROR. In this case, a bad free
has no effect, and a malloc that encounters a bad address
caused by user overwrites will ignore the bad address by
dropping pointers and indices to all known memory. This may
be appropriate for programs that should continue if at all
possible in the face of programming errors, although they may
run out of memory because dropped memory is never reclaimed.
If you don't like either of these options, you can define
else. And if if you are sure that your program using malloc has
no errors or vulnerabilities, you can define INSECURE to 1,
which might (or might not) provide a small performance improvement.
It is also possible to limit the maximum total allocatable
space, using malloc_set_footprint_limit. This is not
designed as a security feature in itself (calls to set limits
are not screened or privileged), but may be useful as one
aspect of a secure implementation.
Thread-safety: NOT thread-safe unless USE_LOCKS defined non-zero
When USE_LOCKS is defined, each public call to malloc, free,
etc is surrounded with a lock. By default, this uses a plain
pthread mutex, win32 critical section, or a spin-lock if if
available for the platform and not disabled by setting
USE_SPIN_LOCKS=0. However, if USE_RECURSIVE_LOCKS is defined,
recursive versions are used instead (which are not required for
base functionality but may be needed in layered extensions).
Using a global lock is not especially fast, and can be a major
bottleneck. It is designed only to provide minimal protection
in concurrent environments, and to provide a basis for
extensions. If you are using malloc in a concurrent program,
consider instead using nedmalloc
( or
ptmalloc (See, which are derived from
versions of this malloc.
System requirements: Any combination of MORECORE and/or MMAP/MUNMAP
This malloc can use unix sbrk or any emulation (invoked using
the CALL_MORECORE macro) and/or mmap/munmap or any emulation
(invoked using CALL_MMAP/CALL_MUNMAP) to get and release system
memory. On most unix systems, it tends to work best if both
MORECORE and MMAP are enabled. On Win32, it uses emulations
based on VirtualAlloc. It also uses common C library functions
like memset.
Compliance: I believe it is compliant with the Single Unix Specification
(See Also SVID/XPG, ANSI C, and probably
others as well.
* Overview of algorithms
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 malloc-intensive programs.
In most ways, this malloc is a best-fit allocator. Generally, it
chooses the best-fitting existing chunk for a request, with ties
broken in approximately least-recently-used order. (This strategy
normally maintains low fragmentation.) However, for requests less
than 256bytes, it deviates from best-fit when there is not an
exactly fitting available chunk by preferring to use space adjacent
to that used for the previous small request, as well as by breaking
ties in approximately most-recently-used order. (These enhance
locality of series of small allocations.) And for very large requests
(>= 256Kb by default), it relies on system memory mapping
facilities, if supported. (This helps avoid carrying around and
possibly fragmenting memory used only for large chunks.)
All operations (except malloc_stats and mallinfo) have execution
times that are bounded by a constant factor of the number of bits in
a size_t, not counting any clearing in calloc or copying in realloc,
or actions surrounding MORECORE and MMAP that have times
proportional to the number of non-contiguous regions returned by
system allocation routines, which is often just 1. In real-time
applications, you can optionally suppress segment traversals using
NO_SEGMENT_TRAVERSAL, which assures bounded execution even when
system allocators return non-contiguous spaces, at the typical
expense of carrying around more memory and increased fragmentation.
The implementation is not very modular and seriously overuses
macros. Perhaps someday all C compilers will do as good a job
inlining modular code as can now be done by brute-force expansion,
but now, enough of them seem not to.
Some compilers issue a lot of warnings about code that is
dead/unreachable only on some platforms, and also about intentional
uses of negation on unsigned types. All known cases of each can be
For a longer but out of date high-level description, see
If MSPACES is defined, then in addition to malloc, free, etc.,
this file also defines mspace_malloc, mspace_free, etc. These
are versions of malloc routines that take an "mspace" argument
obtained using create_mspace, to control all internal bookkeeping.
If ONLY_MSPACES is defined, only these versions are compiled.
So if you would like to use this allocator for only some allocations,
and your system malloc for others, you can compile with
ONLY_MSPACES and then do something like...
static mspace mymspace = create_mspace(0,0); // for example
#define mymalloc(bytes) mspace_malloc(mymspace, bytes)
(Note: If you only need one instance of an mspace, you can instead
use "USE_DL_PREFIX" to relabel the global malloc.)
You can similarly create thread-local allocators by storing
mspaces as thread-locals. For example:
static __thread mspace tlms = 0;
void* tlmalloc(size_t bytes) {
if (tlms == 0) tlms = create_mspace(0, 0);
return mspace_malloc(tlms, bytes);
void tlfree(void* mem) { mspace_free(tlms, mem); }
Unless FOOTERS is defined, each mspace is completely independent.
You cannot allocate from one and free to another (although
conformance is only weakly checked, so usage errors are not always
caught). If FOOTERS is defined, then each chunk carries around a tag
indicating its originating mspace, and frees are directed to their
originating spaces. Normally, this requires use of locks.
------------------------- Compile-time options ---------------------------
Be careful in setting #define values for numerical constants of type
size_t. On some systems, literal values are not automatically extended
to size_t precision unless they are explicitly casted. You can also
use the symbolic values MAX_SIZE_T, SIZE_T_ONE, etc below.
WIN32 default: defined if _WIN32 defined
Defining WIN32 sets up defaults for MS environment and compilers.
Otherwise defaults are for unix. Beware that there seem to be some
cases where this malloc might not be a pure drop-in replacement for
Win32 malloc: Random-looking failures from Win32 GDI API's (eg;
SetDIBits()) may be due to bugs in some video driver implementations
when pixel buffers are malloc()ed, and the region spans more than
one VirtualAlloc()ed region. Because dlmalloc uses a small (64Kb)
default granularity, pixel buffers may straddle virtual allocation
regions more often than when using the Microsoft allocator. You can
avoid this by using VirtualAlloc() and VirtualFree() for all pixel
buffers rather than using malloc(). If this is not possible,
recompile this malloc with a larger DEFAULT_GRANULARITY. Note:
in cases where MSC and gcc (cygwin) are known to differ on WIN32,
conditions use _MSC_VER to distinguish them.
DLMALLOC_EXPORT default: extern
Defines how public APIs are declared. If you want to export via a
Windows DLL, you might define this as
#define DLMALLOC_EXPORT extern __declspec(dllexport)
If you want a POSIX ELF shared object, you might use
#define DLMALLOC_EXPORT extern __attribute__((visibility("default")))
MALLOC_ALIGNMENT default: (size_t)(2 * sizeof(void *))
Controls the minimum alignment for malloc'ed chunks. It must be a
power of two and at least 8, even on machines for which smaller
alignments would suffice. It may be defined as larger than this
though. Note however that code and data structures are optimized for
the case of 8-byte alignment.
MSPACES default: 0 (false)
If true, compile in support for independent allocation spaces.
This is only supported if HAVE_MMAP is true.
ONLY_MSPACES default: 0 (false)
If true, only compile in mspace versions, not regular versions.
USE_LOCKS default: 0 (false)
Causes each call to each public routine to be surrounded with
pthread or WIN32 mutex lock/unlock. (If set true, this can be
overridden on a per-mspace basis for mspace versions.) If set to a
non-zero value other than 1, locks are used, but their
implementation is left out, so lock functions must be supplied manually,
as described below.
USE_SPIN_LOCKS default: 1 iff USE_LOCKS and spin locks available
If true, uses custom spin locks for locking. This is currently
supported only gcc >= 4.1, older gccs on x86 platforms, and recent
MS compilers. Otherwise, posix locks or win32 critical sections are
USE_RECURSIVE_LOCKS default: not defined
If defined nonzero, uses recursive (aka reentrant) locks, otherwise
uses plain mutexes. This is not required for malloc proper, but may
be needed for layered allocators such as nedmalloc.
LOCK_AT_FORK default: not defined
If defined nonzero, performs pthread_atfork upon initialization
to initialize child lock while holding parent lock. The implementation
assumes that pthread locks (not custom locks) are being used. In other
cases, you may need to customize the implementation.
FOOTERS default: 0
If true, provide extra checking and dispatching by placing
information in the footers of allocated chunks. This adds
space and time overhead.
INSECURE default: 0
If true, omit checks for usage errors and heap space overwrites.
USE_DL_PREFIX default: NOT defined
Causes compiler to prefix all public routines with the string 'dl'.
This can be useful when you only want to use this malloc in one part
of a program, using your regular system malloc elsewhere.
MALLOC_INSPECT_ALL default: NOT defined
If defined, compiles malloc_inspect_all and mspace_inspect_all, that
perform traversal of all heap space. Unless access to these
functions is otherwise restricted, you probably do not want to
include them in secure implementations.
ABORT default: defined as abort()
Defines how to abort on failed checks. On most systems, a failed
check cannot die with an "assert" or even print an informative
message, because the underlying print routines in turn call malloc,
which will fail again. Generally, the best policy is to simply call
abort(). It's not very useful to do more than this because many
errors due to overwriting will show up as address faults (null, odd
addresses etc) rather than malloc-triggered checks, so will also
abort. Also, most compilers know that abort() does not return, so
can better optimize code conditionally calling it.
PROCEED_ON_ERROR default: defined as 0 (false)
Controls whether detected bad addresses cause them to bypassed
rather than aborting. If set, detected bad arguments to free and
realloc are ignored. And all bookkeeping information is zeroed out
upon a detected overwrite of freed heap space, thus losing the
ability to ever return it from malloc again, but enabling the
application to proceed. If PROCEED_ON_ERROR is defined, the
static variable malloc_corruption_error_count is compiled in
and can be examined to see if errors have occurred. This option
generates slower code than the default abort policy.
DEBUG default: NOT defined
The DEBUG setting is mainly intended for people trying to modify
this code or diagnose problems when porting to new platforms.
However, it may also be able to better isolate user errors than just
using runtime checks. The assertions in the check routines spell
out in more detail the assumptions and invariants underlying the
algorithms. The checking is fairly extensive, and will slow down
execution noticeably. Calling malloc_stats or mallinfo with DEBUG
set will attempt to check every non-mmapped allocated and free chunk
in the course of computing the summaries.
ABORT_ON_ASSERT_FAILURE default: defined as 1 (true)
Debugging assertion failures can be nearly impossible if your
version of the assert macro causes malloc to be called, which will
lead to a cascade of further failures, blowing the runtime stack.
ABORT_ON_ASSERT_FAILURE cause assertions failures to call abort(),
which will usually make debugging easier.
MALLOC_FAILURE_ACTION default: sets errno to ENOMEM, or no-op on win32
The action to take before "return 0" when malloc fails to be able to
return memory because there is none available.
HAVE_MORECORE default: 1 (true) unless win32 or ONLY_MSPACES
True if this system supports sbrk or an emulation of it.
MORECORE default: sbrk
The name of the sbrk-style system routine to call to obtain more
memory. See below for guidance on writing custom MORECORE
functions. The type of the argument to sbrk/MORECORE varies across
systems. It cannot be size_t, because it supports negative
arguments, so it is normally the signed type of the same width as
size_t (sometimes declared as "intptr_t"). It doesn't much matter
though. Internally, we only call it with arguments less than half
the max value of a size_t, which should work across all reasonable
possibilities, although sometimes generating compiler warnings.
If true, take advantage of fact that consecutive calls to MORECORE
with positive arguments always return contiguous increasing
addresses. This is true of unix sbrk. It does not hurt too much to
set it true anyway, since malloc copes with non-contiguities.
Setting it false when definitely non-contiguous saves time
and possibly wasted space it would take to discover this though.
True if MORECORE cannot release space back to the system when given
negative arguments. This is generally necessary only if you are
using a hand-crafted MORECORE function that cannot handle negative
If non-zero, suppresses traversals of memory segments
returned by either MORECORE or CALL_MMAP. This disables
merging of segments that are contiguous, and selectively
releasing them to the OS if unused, but bounds execution times.
HAVE_MMAP default: 1 (true)
True if this system supports mmap or an emulation of it. If so, and
HAVE_MORECORE is not true, MMAP is used for all system
allocation. If set and HAVE_MORECORE is true as well, MMAP is
primarily used to directly allocate very large blocks. It is also
used as a backup strategy in cases where MORECORE fails to provide
space from system. Note: A single call to MUNMAP is assumed to be
able to unmap memory that may have be allocated using multiple calls
to MMAP, so long as they are adjacent.
HAVE_MREMAP default: 1 on linux, else 0
If true realloc() uses mremap() to re-allocate large blocks and
extend or shrink allocation spaces.
MMAP_CLEARS default: 1 except on WINCE.
True if mmap clears memory so calloc doesn't need to. This is true
for standard unix mmap using /dev/zero and on WIN32 except for WINCE.
USE_BUILTIN_FFS default: 0 (i.e., not used)
Causes malloc to use the builtin ffs() function to compute indices.
Some compilers may recognize and intrinsify ffs to be faster than the
supplied C version. Also, the case of x86 using gcc is special-cased
to an asm instruction, so is already as fast as it can be, and so
this setting has no effect. Similarly for Win32 under recent MS compilers.
(On most x86s, the asm version is only slightly faster than the C version.)
malloc_getpagesize default: derive from system includes, or 4096.
The system page size. To the extent possible, this malloc manages
memory from the system in page-size units. This may be (and
usually is) a function rather than a constant. This is ignored
if WIN32, where page size is determined using getSystemInfo during
USE_DEV_RANDOM default: 0 (i.e., not used)
Causes malloc to use /dev/random to initialize secure magic seed for
stamping footers. Otherwise, the current time is used.
NO_MALLINFO default: 0
If defined, don't compile "mallinfo". This can be a simple way
of dealing with mismatches between system declarations and
those in this file.
MALLINFO_FIELD_TYPE default: size_t
The type of the fields in the mallinfo struct. This was originally
defined as "int" in SVID etc, but is more usefully defined as
size_t. The value is used only if HAVE_USR_INCLUDE_MALLOC_H is not set
NO_MALLOC_STATS default: 0
If defined, don't compile "malloc_stats". This avoids calls to
fprintf and bringing in stdio dependencies you might not want.
REALLOC_ZERO_BYTES_FREES default: not defined
This should be set if a call to realloc with zero bytes should
be the same as a call to free. Some people think it should. Otherwise,
since this malloc returns a unique pointer for malloc(0), so does
realloc(p, 0).
LACKS_STDLIB_H LACKS_SCHED_H LACKS_TIME_H default: NOT defined unless on WIN32
Define these if your system does not have these header files.
You might need to manually insert some of the declarations they provide.
system_info.dwAllocationGranularity in WIN32,
otherwise 64K.
Also settable using mallopt(M_GRANULARITY, x)
The unit for allocating and deallocating memory from the system. On
most systems with contiguous MORECORE, there is no reason to
make this more than a page. However, systems with MMAP tend to
either require or encourage larger granularities. You can increase
this value to prevent system allocation functions to be called so
often, especially if they are slow. The value must be at least one
page and must be a power of two. Setting to 0 causes initialization
to either page size or win32 region size. (Note: In previous
versions of malloc, the equivalent of this option was called
Also settable using mallopt(M_TRIM_THRESHOLD, x)
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 using contiguous MORECORE. 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 this much
memory. 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 is worth tuning trim thresholds 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. The trim
value must be greater than page size to have any useful effect. To
disable trimming completely, you can set to MAX_SIZE_T. Note that the trick
some people use of mallocing a huge space and then freeing it at
program startup, in an attempt to reserve system memory, doesn't
have the intended effect under automatic trimming, since that memory
will immediately be returned to the system.
Also settable using mallopt(M_MMAP_THRESHOLD, x)
The request size threshold for using MMAP to directly 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 benefits 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. Also, mapped memory doesn't become
`locked' between other chunks, as can happen with normally allocated
chunks, which means that even trimming via malloc_trim would not
release them. However, it has the disadvantage that the space
cannot be reclaimed, consolidated, and then used to service later
requests, as happens with normal chunks. The advantages of mmap
nearly always outweigh disadvantages for "large" chunks, but the
value of "large" may vary across systems. The default is an
empirically derived value that works well in most systems. You can
disable mmap by setting to MAX_SIZE_T.
MAX_RELEASE_CHECK_RATE default: 4095 unless not HAVE_MMAP
The number of consolidated frees between checks to release
unused segments when freeing. When using non-contiguous segments,
especially with multiple mspaces, checking only for topmost space
doesn't always suffice to trigger trimming. To compensate for this,
free() will, with a period of MAX_RELEASE_CHECK_RATE (or the
current number of segments, if greater) try to release unused
segments to the OS when freeing chunks that result in
consolidation. The best value for this parameter is a compromise
between slowing down frees with relatively costly checks that
rarely trigger versus holding on to unused memory. To effectively
disable, set to MAX_SIZE_T. This may lead to a very slight speed
improvement at the expense of carrying around more memory.
/* Version identifier to allow people to support multiple versions */
#define DLMALLOC_VERSION 20806
#define DLMALLOC_EXPORT extern
#if defined(STARBOARD)
// Just make sure we don't detect as these other platforms.
#elif defined(__LB_SHELL__)
#include <time.h>
#if defined(__LB_XB1__) || defined(__LB_XB360__)
// We don't want dlmalloc's WIN32, but
// we do need some Windows types.
#include <windows.h>
#undef WIN32
#endif // __LB_XB1__ || __LB_XB360__
#else // defined(STARBOARD)
#ifndef WIN32
#if defined(_WIN32) && !defined(__LB_XB1__) && !defined(__LB_XB360__)
#define WIN32 1
#endif /* _WIN32 */
#ifdef _WIN32_WCE
#define WIN32 1
#endif /* _WIN32_WCE */
#endif /* WIN32 */
#if defined(WIN32) && !defined(__LB_XB1__) && !defined(__LB_XB360__)
#include <windows.h>
#include <tchar.h>
#define HAVE_MMAP 1
#ifdef _WIN32_WCE /* WINCE reportedly does not clear */
#define MMAP_CLEARS 0
#define MMAP_CLEARS 1
#endif /* _WIN32_WCE */
#endif /*MMAP_CLEARS */
#endif /* WIN32 */
#if defined(DARWIN) || defined(_DARWIN)
/* Mac OSX docs advise not to use sbrk; it seems better to use mmap */
#define HAVE_MMAP 1
/* OSX allocators provide 16 byte alignment */
#define MALLOC_ALIGNMENT ((size_t)16U)
#endif /* HAVE_MORECORE */
#endif /* DARWIN */
#include <sys/types.h> /* For size_t */
#endif /* LACKS_SYS_TYPES_H */
#endif // defined(STARBOARD)
#if defined(STARBOARD)
#include "dlmalloc_config.h"
#include "starboard/log.h"
#include "starboard/memory.h"
#include "starboard/string.h"
#include "starboard/system.h"
#ifdef MAX
#undef MAX
#define MAX(a, b) (((a) > (b)) ? (a) : (b))
#define ABORT \
do { \
SbLogFormatF("%s:%s:%d - ABORT\n", __FUNCTION__, __FILE__, __LINE__); \
SbSystemBreakIntoDebugger(); \
} while(false)
#define CRASH() ABORT
#elif defined(__LB_SHELL__)
#include "dlmalloc_config.h"
#include "lb_memory_debug_platform.h"
#ifdef MAX
#undef MAX
#define MAX(a, b) (((a) > (b)) ? (a) : (b))
#define CRASH() ((void(*)())0)()
#endif // defined(STARBOARD)
/* The maximum possible size_t value has all bits set */
#define MAX_SIZE_T (~(size_t)0)
#ifndef USE_LOCKS /* ensure true if spin or recursive locks set */
#define USE_LOCKS ((defined(USE_SPIN_LOCKS) && USE_SPIN_LOCKS != 0) || \
#endif /* USE_LOCKS */
#if USE_LOCKS /* Spin locks for gcc >= 4.1, older gcc on x86, MSC >= 1310 */
#if ((defined(__GNUC__) && \
((__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 1)) || \
defined(__i386__) || defined(__x86_64__))) || \
(defined(_MSC_VER) && _MSC_VER>=1310))
#define USE_SPIN_LOCKS 1
#endif /* USE_SPIN_LOCKS */
#error "USE_SPIN_LOCKS defined without implementation"
#endif /* ... locks available... */
#elif !defined(USE_SPIN_LOCKS)
#define USE_SPIN_LOCKS 0
#endif /* USE_LOCKS */
#define ONLY_MSPACES 0
#endif /* ONLY_MSPACES */
#ifndef MSPACES
#define MSPACES 1
#else /* ONLY_MSPACES */
#define MSPACES 0
#endif /* ONLY_MSPACES */
#endif /* MSPACES */
#define MALLOC_ALIGNMENT ((size_t)(2 * sizeof(void *)))
#ifndef FOOTERS
#define FOOTERS 0
#endif /* FOOTERS */
#ifndef ABORT
#define ABORT abort()
#endif /* ABORT */
#endif /* PROCEED_ON_ERROR */
#ifndef INSECURE
#define INSECURE 0
#endif /* INSECURE */
#ifndef HAVE_MMAP
#define HAVE_MMAP 1
#endif /* HAVE_MMAP */
#define MMAP_CLEARS 1
#endif /* MMAP_CLEARS */
#ifdef linux
#define HAVE_MREMAP 1
#define _GNU_SOURCE /* Turns on mremap() definition */
#else /* linux */
#define HAVE_MREMAP 0
#endif /* linux */
#endif /* HAVE_MREMAP */
#else /* ONLY_MSPACES */
#endif /* ONLY_MSPACES */
#endif /* HAVE_MORECORE */
#else /* !HAVE_MORECORE */
#endif /* HAVE_MORECORE */
#if (MORECORE_CONTIGUOUS || defined(WIN32) || defined(STARBOARD))
#define DEFAULT_GRANULARITY (0) /* 0 means to compute in init_mparams */
#define DEFAULT_GRANULARITY ((size_t)64U * (size_t)1024U)
#define DEFAULT_TRIM_THRESHOLD ((size_t)2U * (size_t)1024U * (size_t)1024U)
#define DEFAULT_MMAP_THRESHOLD ((size_t)256U * (size_t)1024U)
#else /* HAVE_MMAP */
#endif /* HAVE_MMAP */
#endif /* HAVE_MMAP */
#endif /* USE_BUILTIN_FFS */
#define USE_DEV_RANDOM 0
#endif /* USE_DEV_RANDOM */
#define NO_MALLINFO 0
#endif /* NO_MALLINFO */
#define MALLINFO_FIELD_TYPE size_t
#endif /* NO_MALLOC_STATS */
mallopt tuning options. SVID/XPG defines four standard parameter
numbers for mallopt, normally defined in malloc.h. None of these
are used in this malloc, so setting them has no effect. But this
malloc does support the following options.
#define M_TRIM_THRESHOLD (-1)
#define M_GRANULARITY (-2)
#define M_MMAP_THRESHOLD (-3)
/* ------------------------ Mallinfo declarations ------------------------ */
This version of malloc supports the standard SVID/XPG mallinfo
routine that returns a struct containing usage properties and
statistics. It should work on any system that has a
/usr/include/malloc.h defining struct mallinfo. The main
declaration needed is the mallinfo struct that is returned (by-copy)
by mallinfo(). The malloinfo struct contains a bunch of fields that
are not even meaningful in this version of malloc. These fields are
are instead filled by mallinfo() with other numbers that might be of
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 a compliant version is
declared below. These must be precisely the same for mallinfo() to
work. The original SVID version of this struct, defined on most
systems with mallinfo, declares all fields as ints. But some others
define as unsigned long. If your system defines the fields using a
type of different width than listed here, you MUST #include your
system version and #define HAVE_USR_INCLUDE_MALLOC_H.
#include "/usr/include/malloc.h"
/* HP-UX (and others?) redefines mallinfo unless _STRUCT_MALLINFO is defined */
struct mallinfo {
MALLINFO_FIELD_TYPE arena; /* non-mmapped space allocated from system */
MALLINFO_FIELD_TYPE ordblks; /* number of free chunks */
MALLINFO_FIELD_TYPE smblks; /* always 0 */
MALLINFO_FIELD_TYPE hblks; /* always 0 */
MALLINFO_FIELD_TYPE hblkhd; /* space in mmapped regions */
MALLINFO_FIELD_TYPE usmblks; /* maximum total allocated space */
MALLINFO_FIELD_TYPE fsmblks; /* always 0 */
MALLINFO_FIELD_TYPE uordblks; /* total allocated space */
MALLINFO_FIELD_TYPE fordblks; /* total free space */
MALLINFO_FIELD_TYPE keepcost; /* releasable (via malloc_trim) space */
#endif /* NO_MALLINFO */
Try to persuade compilers to inline. The most critical functions for
inlining are defined as macros, so these aren't used for them.
#if defined(__GNUC__)
#define FORCEINLINE __inline __attribute__ ((always_inline))
#elif defined(_MSC_VER)
#define FORCEINLINE __forceinline
#elif defined(__LB_XB360__)
#define FORCEINLINE __forceinline
#ifndef NOINLINE
#if defined(__GNUC__)
#define NOINLINE __attribute__ ((noinline))
#elif defined(_MSC_VER)
#define NOINLINE __declspec(noinline)
#define NOINLINE
#ifdef __cplusplus
extern "C" {
#define FORCEINLINE inline
#endif /* __cplusplus */
/* ------------------- Declarations of public routines ------------------- */
#define dlcalloc calloc
#define dlfree free
#define dlmalloc malloc
#define dlmemalign memalign
#define dlposix_memalign posix_memalign
#define dlrealloc realloc
#define dlrealloc_in_place realloc_in_place
#define dlvalloc valloc
#define dlpvalloc pvalloc
#define dlmallinfo mallinfo
#define dlmallopt mallopt
#define dlmalloc_trim malloc_trim
#define dlmalloc_stats malloc_stats
#define dlmalloc_usable_size malloc_usable_size
#define dlmalloc_footprint malloc_footprint
#define dlmalloc_max_footprint malloc_max_footprint
#define dlmalloc_footprint_limit malloc_footprint_limit
#define dlmalloc_set_footprint_limit malloc_set_footprint_limit
#define dlmalloc_inspect_all malloc_inspect_all
#define dlindependent_calloc independent_calloc
#define dlindependent_comalloc independent_comalloc
#define dlbulk_free bulk_free
#endif /* USE_DL_PREFIX */
malloc(size_t n)
Returns a pointer to a newly allocated chunk of at least n bytes, or
null if no space is available, in which case errno is set to ENOMEM
on ANSI C systems.
If n is zero, malloc returns a minimum-sized chunk. (The minimum
size is 16 bytes on most 32bit systems, and 32 bytes on 64bit
systems.) Note that size_t is an unsigned type, so calls with
arguments that would be negative if signed are interpreted as
requests for huge amounts of space, which will often fail. The
maximum supported value of n differs across systems, but is in all
cases less than the maximum representable value of a size_t.
DLMALLOC_EXPORT void* dlmalloc(size_t);
free(void* p)
Releases the chunk of memory pointed to by p, that had been previously
allocated using malloc or a related routine such as realloc.
It has no effect if p is null. If p was not malloced or already
freed, free(p) will by default cause the current program to abort.
DLMALLOC_EXPORT void dlfree(void*);
calloc(size_t n_elements, size_t element_size);
Returns a pointer to n_elements * element_size bytes, with all locations
set to zero.
DLMALLOC_EXPORT void* dlcalloc(size_t, size_t);
realloc(void* p, size_t n)
Returns 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. The algorithm
prefers extending p in most cases when possible, otherwise it
employs the equivalent of a malloc-copy-free sequence.
If p is null, realloc is equivalent to malloc.
If space is not available, realloc returns null, errno is set (if on
ANSI) and p is NOT freed.
if n is for fewer bytes than already held by p, the newly unused
space is lopped off and freed if possible. realloc with a size
argument of zero (re)allocates a minimum-sized chunk.
The old unix realloc convention of allowing the last-free'd chunk
to be used as an argument to realloc is not supported.
DLMALLOC_EXPORT void* dlrealloc(void*, size_t);
realloc_in_place(void* p, size_t n)
Resizes the space allocated for p to size n, only if this can be
done without moving p (i.e., only if there is adjacent space
available if n is greater than p's current allocated size, or n is
less than or equal to p's size). This may be used instead of plain
realloc if an alternative allocation strategy is needed upon failure
to expand space; for example, reallocation of a buffer that must be
memory-aligned or cleared. You can use realloc_in_place to trigger
these alternatives only when needed.
Returns p if successful; otherwise null.
DLMALLOC_EXPORT void* dlrealloc_in_place(void*, size_t);
memalign(size_t alignment, size_t n);
Returns a pointer to a newly allocated chunk of n bytes, aligned
in accord with the alignment argument.
The alignment argument should be a power of two. If the argument is
not a power of two, the nearest greater power is used.
8-byte alignment is guaranteed by normal malloc calls, so don't
bother calling memalign with an argument of 8 or less.
Overreliance on memalign is a sure way to fragment space.
DLMALLOC_EXPORT void* dlmemalign(size_t, size_t);
int posix_memalign(void** pp, size_t alignment, size_t n);
Allocates a chunk of n bytes, aligned in accord with the alignment
argument. Differs from memalign only in that it (1) assigns the
allocated memory to *pp rather than returning it, (2) fails and
returns EINVAL if the alignment is not a power of two (3) fails and
returns ENOMEM if memory cannot be allocated.
DLMALLOC_EXPORT int dlposix_memalign(void**, size_t, size_t);
valloc(size_t n);
Equivalent to memalign(pagesize, n), where pagesize is the page
size of the system. If the pagesize is unknown, 4096 is used.
DLMALLOC_EXPORT void* dlvalloc(size_t);
mallopt(int parameter_number, int parameter_value)
Sets tunable parameters The format is to provide a
(parameter-number, parameter-value) pair. mallopt then sets the
corresponding parameter to the argument value if it can (i.e., so
long as the value is meaningful), and returns 1 if successful else
0. To workaround the fact that mallopt is specified to use int,
not size_t parameters, the value -1 is specially treated as the
maximum unsigned size_t value.
SVID/XPG/ANSI defines four standard param numbers for mallopt,
normally defined in malloc.h. None of these are use in this malloc,
so setting them has no effect. But this malloc also supports other
options in mallopt. See below for details. Briefly, supported
parameters are as follows (listed defaults are for "typical"
Symbol param # default allowed param values
M_TRIM_THRESHOLD -1 2*1024*1024 any (-1 disables)
M_GRANULARITY -2 page size any power of 2 >= page size
M_MMAP_THRESHOLD -3 256*1024 any (or 0 if no MMAP support)
DLMALLOC_EXPORT int dlmallopt(int, int);
Returns the number of bytes obtained from the system. The total
number of bytes allocated by malloc, realloc etc., is less than this
value. Unlike mallinfo, this function returns only a precomputed
result, so can be called frequently to monitor memory consumption.
Even if locks are otherwise defined, this function does not use them,
so results might not be up to date.
DLMALLOC_EXPORT size_t dlmalloc_footprint(void);
Returns the maximum number of bytes obtained from the system. This
value will be greater than current footprint if deallocated space
has been reclaimed by the system. The peak number of bytes allocated
by malloc, realloc etc., is less than this value. Unlike mallinfo,
this function returns only a precomputed result, so can be called
frequently to monitor memory consumption. Even if locks are
otherwise defined, this function does not use them, so results might
not be up to date.
DLMALLOC_EXPORT size_t dlmalloc_max_footprint(void);
Returns the number of bytes that the heap is allowed to obtain from
the system, returning the last value returned by
malloc_set_footprint_limit, or the maximum size_t value if
never set. The returned value reflects a permission. There is no
guarantee that this number of bytes can actually be obtained from
the system.
DLMALLOC_EXPORT size_t dlmalloc_footprint_limit();
Sets the maximum number of bytes to obtain from the system, causing
failure returns from malloc and related functions upon attempts to
exceed this value. The argument value may be subject to page
rounding to an enforceable limit; this actual value is returned.
Using an argument of the maximum possible size_t effectively
disables checks. If the argument is less than or equal to the
current malloc_footprint, then all future allocations that require
additional system memory will fail. However, invocation cannot
retroactively deallocate existing used memory.
DLMALLOC_EXPORT size_t dlmalloc_set_footprint_limit(size_t bytes);
malloc_inspect_all(void(*handler)(void *start,
void *end,
size_t used_bytes,
void* callback_arg),
void* arg);
Traverses the heap and calls the given handler for each managed
region, skipping all bytes that are (or may be) used for bookkeeping
purposes. Traversal does not include include chunks that have been
directly memory mapped. Each reported region begins at the start
address, and continues up to but not including the end address. The
first used_bytes of the region contain allocated data. If
used_bytes is zero, the region is unallocated. The handler is
invoked with the given callback argument. If locks are defined, they
are held during the entire traversal. It is a bad idea to invoke
other malloc functions from within the handler.
For example, to count the number of in-use chunks with size greater
than 1000, you could write:
static int count = 0;
void count_chunks(void* start, void* end, size_t used, void* arg) {
if (used >= 1000) ++count;
malloc_inspect_all(count_chunks, NULL);
malloc_inspect_all is compiled only if MALLOC_INSPECT_ALL is defined.
DLMALLOC_EXPORT void dlmalloc_inspect_all(void(*handler)(void*, void *, size_t, void*),
void* arg);
Returns (by copy) a struct containing various summary statistics:
arena: current total non-mmapped bytes allocated from system
ordblks: the number of free chunks
smblks: always zero.
hblks: current number of mmapped regions
hblkhd: total bytes held in mmapped regions
usmblks: the maximum total allocated space. This will be greater
than current total if trimming has occurred.
fsmblks: always zero
uordblks: current total allocated space (normal or mmapped)
fordblks: total free space
keepcost: the maximum number of bytes that could ideally be released
back to system via malloc_trim. ("ideally" means that
it ignores page restrictions etc.)
Because these fields are ints, but internal bookkeeping may
be kept as longs, the reported values may wrap around zero and
thus be inaccurate.
DLMALLOC_EXPORT struct mallinfo dlmallinfo(void);
#endif /* NO_MALLINFO */
independent_calloc(size_t n_elements, size_t element_size, void* chunks[]);
independent_calloc is similar to calloc, but instead of returning a
single cleared space, it returns an array of pointers to n_elements
independent elements that can hold contents of size elem_size, each
of which starts out cleared, and can be independently freed,
realloc'ed etc. The elements are guaranteed to be adjacently
allocated (this is not guaranteed to occur with multiple callocs or
mallocs), which may also improve cache locality in some
The "chunks" argument is optional (i.e., may be null, which is
probably the most typical usage). If it is null, the returned array
is itself dynamically allocated and should also be freed when it is
no longer needed. Otherwise, the chunks array must be of at least
n_elements in length. It is filled in with the pointers to the
In either case, independent_calloc returns this pointer array, or
null if the allocation failed. If n_elements is zero and "chunks"
is null, it returns a chunk representing an array with zero elements
(which should be freed if not wanted).
Each element must be freed when it is no longer needed. This can be
done all at once using bulk_free.
independent_calloc simplifies and speeds up implementations of many
kinds of pools. It may also be useful when constructing large data
structures that initially have a fixed number of fixed-sized nodes,
but the number is not known at compile time, and some of the nodes
may later need to be freed. For example:
struct Node { int item; struct Node* next; };
struct Node* build_list() {
struct Node** pool;
int n = read_number_of_nodes_needed();
if (n <= 0) return 0;
pool = (struct Node**)(independent_calloc(n, sizeof(struct Node), 0);
if (pool == 0) die();
// organize into a linked list...
struct Node* first = pool[0];
for (i = 0; i < n-1; ++i)
pool[i]->next = pool[i+1];
free(pool); // Can now free the array (or not, if it is needed later)
return first;
DLMALLOC_EXPORT void** dlindependent_calloc(size_t, size_t, void**);
independent_comalloc(size_t n_elements, size_t sizes[], void* chunks[]);
independent_comalloc allocates, all at once, a set of n_elements
chunks with sizes indicated in the "sizes" array. It returns
an array of pointers to these elements, each of which can be
independently freed, realloc'ed etc. The elements are guaranteed to
be adjacently allocated (this is not guaranteed to occur with
multiple callocs or mallocs), which may also improve cache locality
in some applications.
The "chunks" argument is optional (i.e., may be null). If it is null
the returned array is itself dynamically allocated and should also
be freed when it is no longer needed. Otherwise, the chunks array
must be of at least n_elements in length. It is filled in with the
pointers to the chunks.
In either case, independent_comalloc returns this pointer array, or
null if the allocation failed. If n_elements is zero and chunks is
null, it returns a chunk representing an array with zero elements
(which should be freed if not wanted).
Each element must be freed when it is no longer needed. This can be
done all at once using bulk_free.
independent_comallac differs from independent_calloc in that each
element may have a different size, and also that it does not
automatically clear elements.
independent_comalloc can be used to speed up allocation in cases
where several structs or objects must always be allocated at the
same time. For example:
struct Head { ... }
struct Foot { ... }
void send_message(char* msg) {
int msglen = strlen(msg);
size_t sizes[3] = { sizeof(struct Head), msglen, sizeof(struct Foot) };
void* chunks[3];
if (independent_comalloc(3, sizes, chunks) == 0)
struct Head* head = (struct Head*)(chunks[0]);
char* body = (char*)(chunks[1]);
struct Foot* foot = (struct Foot*)(chunks[2]);
// ...
In general though, independent_comalloc is worth using only for
larger values of n_elements. For small values, you probably won't
detect enough difference from series of malloc calls to bother.
Overuse of independent_comalloc can increase overall memory usage,
since it cannot reuse existing noncontiguous small chunks that
might be available for some of the elements.
DLMALLOC_EXPORT void** dlindependent_comalloc(size_t, size_t*, void**);
bulk_free(void* array[], size_t n_elements)
Frees and clears (sets to null) each non-null pointer in the given
array. This is likely to be faster than freeing them one-by-one.
If footers are used, pointers that have been allocated in different
mspaces are not freed or cleared, and the count of all such pointers
is returned. For large arrays of pointers with poor locality, it
may be worthwhile to sort this array before calling bulk_free.
DLMALLOC_EXPORT size_t dlbulk_free(void**, size_t n_elements);
pvalloc(size_t n);
Equivalent to valloc(minimum-page-that-holds(n)), that is,
round up n to nearest pagesize.
DLMALLOC_EXPORT void* dlpvalloc(size_t);
malloc_trim(size_t pad);
If possible, gives memory back to the system (via negative arguments
to sbrk) if there is unused memory at the `high' end of the malloc
pool or in unused MMAP segments. You can call this after freeing
large blocks of memory to potentially reduce the system-level memory
requirements of a program. However, it cannot guarantee to reduce
memory. Under some allocation patterns, some large free blocks of
memory will be locked between two used chunks, so they cannot be
given back to the system.
The `pad' argument to malloc_trim represents the amount of free
trailing space to leave untrimmed. If this argument is zero, only
the minimum amount of memory to maintain internal data structures
will be left. Non-zero arguments can be supplied to maintain enough
trailing space to service future expected allocations without having
to re-obtain memory from the system.
Malloc_trim returns 1 if it actually released any memory, else 0.
DLMALLOC_EXPORT int dlmalloc_trim(size_t);
Prints on stderr the amount of space obtained from the system (both
via sbrk and mmap), the maximum amount (which may be more than
current if malloc_trim and/or munmap got called), and the current
number of bytes allocated via malloc (or realloc, etc) but not yet
freed. Note that this is the number of bytes allocated, not the
number requested. It will be larger than the number requested
because of alignment and bookkeeping overhead. Because it includes
alignment wastage as being in use, this figure may be greater than
zero even when no user-level chunks are allocated.
The reported current and maximum system memory can be inaccurate if
a program makes other calls to system memory allocation functions
(normally sbrk) outside of malloc.
malloc_stats prints only the most commonly interesting statistics.
More information can be obtained by calling mallinfo.
DLMALLOC_EXPORT void dlmalloc_stats(void);
malloc_usable_size(void* p);
Returns the number of bytes you can actually use in
an allocated chunk, which may be more than you requested (although
often not) due to alignment and minimum size constraints.
You can use this many bytes without worrying about
overwriting other allocated objects. This is not a particularly great
programming practice. malloc_usable_size can be more useful in
debugging and assertions, for example:
p = malloc(n);
assert(malloc_usable_size(p) >= 256);
size_t dlmalloc_usable_size(void*);
#endif /* ONLY_MSPACES */
mspace is an opaque type representing an independent
region of space that supports mspace_malloc, etc.
typedef void* mspace;
create_mspace creates and returns a new independent space with the
given initial capacity, or, if 0, the default granularity size. It
returns null if there is no system memory available to create the
space. If argument locked is non-zero, the space uses a separate
lock to control access. The capacity of the space will grow
dynamically as needed to service mspace_malloc requests. You can
control the sizes of incremental increases of this space by
compiling with a different DEFAULT_GRANULARITY or dynamically
setting with mallopt(M_GRANULARITY, value).
DLMALLOC_EXPORT mspace create_mspace(size_t capacity, int locked);
destroy_mspace destroys the given space, and attempts to return all
of its memory back to the system, returning the total number of
bytes freed. After destruction, the results of access to all memory
used by the space become undefined.
DLMALLOC_EXPORT size_t destroy_mspace(mspace msp);
create_mspace_with_base uses the memory supplied as the initial base
of a new mspace. Part (less than 128*sizeof(size_t) bytes) of this
space is used for bookkeeping, so the capacity must be at least this
large. (Otherwise 0 is returned.) When this initial space is
exhausted, additional memory will be obtained from the system.
Destroying this space will deallocate all additionally allocated
space (if possible) but not the initial base.
DLMALLOC_EXPORT mspace create_mspace_with_base(void* base, size_t capacity, int locked);
mspace_track_large_chunks controls whether requests for large chunks
are allocated in their own untracked mmapped regions, separate from
others in this mspace. By default large chunks are not tracked,
which reduces fragmentation. However, such chunks are not
necessarily released to the system upon destroy_mspace. Enabling
tracking by setting to true may increase fragmentation, but avoids
leakage when relying on destroy_mspace to release all memory
allocated using this space. The function returns the previous
DLMALLOC_EXPORT int mspace_track_large_chunks(mspace msp, int enable);
mspace_malloc behaves as malloc, but operates within
the given space.
DLMALLOC_EXPORT void* mspace_malloc(mspace msp, size_t bytes);
mspace_free behaves as free, but operates within
the given space.
If compiled with FOOTERS==1, mspace_free is not actually needed.
free may be called instead of mspace_free because freed chunks from
any space are handled by their originating spaces.
DLMALLOC_EXPORT void mspace_free(mspace msp, void* mem);
mspace_realloc behaves as realloc, but operates within
the given space.
If compiled with FOOTERS==1, mspace_realloc is not actually
needed. realloc may be called instead of mspace_realloc because
realloced chunks from any space are handled by their originating
DLMALLOC_EXPORT void* mspace_realloc(mspace msp, void* mem, size_t newsize);
mspace_calloc behaves as calloc, but operates within
the given space.
DLMALLOC_EXPORT void* mspace_calloc(mspace msp, size_t n_elements, size_t elem_size);
mspace_memalign behaves as memalign, but operates within
the given space.
DLMALLOC_EXPORT void* mspace_memalign(mspace msp, size_t alignment, size_t bytes);
mspace_independent_calloc behaves as independent_calloc, but
operates within the given space.
DLMALLOC_EXPORT void** mspace_independent_calloc(mspace msp, size_t n_elements,
size_t elem_size, void* chunks[]);
mspace_independent_comalloc behaves as independent_comalloc, but
operates within the given space.
DLMALLOC_EXPORT void** mspace_independent_comalloc(mspace msp, size_t n_elements,
size_t sizes[], void* chunks[]);
mspace_footprint() returns the number of bytes obtained from the
system for this space.
DLMALLOC_EXPORT size_t mspace_footprint(mspace msp);
mspace_max_footprint() returns the peak number of bytes obtained from the
system for this space.
DLMALLOC_EXPORT size_t mspace_max_footprint(mspace msp);
mspace_mallinfo behaves as mallinfo, but reports properties of
the given space.
DLMALLOC_EXPORT struct mallinfo mspace_mallinfo(mspace msp);
#endif /* NO_MALLINFO */
malloc_usable_size(void* p) behaves the same as malloc_usable_size;
DLMALLOC_EXPORT size_t mspace_usable_size(const void* mem);
mspace_malloc_stats behaves as malloc_stats, but reports
properties of the given space.
DLMALLOC_EXPORT void mspace_malloc_stats(mspace msp);
mspace_trim behaves as malloc_trim, but
operates within the given space.
DLMALLOC_EXPORT int mspace_trim(mspace msp, size_t pad);
An alias for mallopt.
DLMALLOC_EXPORT int mspace_mallopt(int, int);
#endif /* MSPACES */
#ifdef __cplusplus
} /* end of extern "C" */
#endif /* __cplusplus */
To make a fully customizable malloc.h header file, cut everything
above this line, put into file malloc.h, edit to suit, and #include it
on the next line, as well as in programs that use this malloc.
/* #include "malloc.h" */
/*------------------------------ internal #includes ---------------------- */
#ifdef _MSC_VER
#pragma warning( disable : 4146 ) /* no "unsigned" warnings */
#endif /* _MSC_VER */
#include <stdio.h> /* for printing in malloc_stats */
#endif /* NO_MALLOC_STATS */
#include <errno.h> /* for MALLOC_FAILURE_ACTION */
#endif /* LACKS_ERRNO_H */
#ifdef DEBUG
#undef assert
#define assert(x) if(!(x)) ABORT
#include <assert.h>
#else /* DEBUG */
#ifndef assert
#define assert(x)
#define DEBUG 0
#endif /* DEBUG */
#if !defined(WIN32) && !defined(LACKS_TIME_H)
#include <time.h> /* for magic initialization */
#endif /* WIN32 */
#include <stdlib.h> /* for abort() */
#endif /* LACKS_STDLIB_H */
#include <string.h> /* for memset etc */
#endif /* LACKS_STRING_H */
#include <strings.h> /* for ffs */
#endif /* LACKS_STRINGS_H */
#endif /* USE_BUILTIN_FFS */
/* On some versions of linux, mremap decl in mman.h needs __USE_GNU set */
#if (defined(linux) && !defined(__USE_GNU))
#define __USE_GNU 1
#include <sys/mman.h> /* for mmap */
#undef __USE_GNU
#include <sys/mman.h> /* for mmap */
#endif /* linux */
#endif /* LACKS_SYS_MMAN_H */
#include <fcntl.h>
#endif /* LACKS_FCNTL_H */
#endif /* HAVE_MMAP */
#include <unistd.h> /* for sbrk, sysconf */
#else /* LACKS_UNISTD_H */
#if !defined(__FreeBSD__) && !defined(__OpenBSD__) && !defined(__NetBSD__)
extern void* sbrk(ptrdiff_t);
#endif /* FreeBSD etc */
#endif /* LACKS_UNISTD_H */
/* Declarations for locking */
#ifndef WIN32
#if defined (__SVR4) && defined (__sun) /* solaris */
#include <thread.h>
#elif !defined(LACKS_SCHED_H)
#include <sched.h>
#endif /* solaris or LACKS_SCHED_H */
#if !defined(STARBOARD)
#include <pthread.h>
#endif /* !STARBOARD */
#endif /* USE_RECURSIVE_LOCKS ... */
#elif defined(_MSC_VER)
#ifndef _M_AMD64
/* These are already defined on AMD64 builds */
#ifdef __cplusplus
extern "C" {
#endif /* __cplusplus */
LONG __cdecl _InterlockedCompareExchange(LONG volatile *Dest, LONG Exchange, LONG Comp);
LONG __cdecl _InterlockedExchange(LONG volatile *Target, LONG Value);
#ifdef __cplusplus
#endif /* __cplusplus */
#endif /* _M_AMD64 */
#pragma intrinsic (_InterlockedCompareExchange)
#pragma intrinsic (_InterlockedExchange)
#define interlockedcompareexchange _InterlockedCompareExchange
#define interlockedexchange _InterlockedExchange
#elif defined(WIN32) && defined(__GNUC__)
#define interlockedcompareexchange(a, b, c) __sync_val_compare_and_swap(a, c, b)
#define interlockedexchange __sync_lock_test_and_set
#endif /* Win32 */
#else /* USE_LOCKS */
#endif /* USE_LOCKS */
#ifndef LOCK_AT_FORK
#define LOCK_AT_FORK 0
/* Declarations for bit scanning on win32 */
#if defined(_MSC_VER) && _MSC_VER>=1300 && !defined(__LB_XB360__)
#ifndef BitScanForward /* Try to avoid pulling in WinNT.h */
#ifdef __cplusplus
extern "C" {
#endif /* __cplusplus */
unsigned char _BitScanForward(unsigned long *index, unsigned long mask);
unsigned char _BitScanReverse(unsigned long *index, unsigned long mask);
#ifdef __cplusplus
#endif /* __cplusplus */
#define BitScanForward _BitScanForward
#define BitScanReverse _BitScanReverse
#pragma intrinsic(_BitScanForward)
#pragma intrinsic(_BitScanReverse)
#endif /* BitScanForward */
#endif /* defined(_MSC_VER) && _MSC_VER>=1300 */
#ifndef WIN32
#ifndef malloc_getpagesize
# ifdef _SC_PAGESIZE /* some SVR4 systems omit an underscore */
# ifndef _SC_PAGE_SIZE
# 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
# ifdef WIN32 /* use supplied emulation of getpagesize */
# define malloc_getpagesize getpagesize()
# else
# include <sys/param.h>
# endif
# 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 /* just guess */
# define malloc_getpagesize ((size_t)4096U)
# endif
# endif
# endif
# endif
# endif
# endif
# endif
/* ------------------- size_t and alignment properties -------------------- */
/* The byte and bit size of a size_t */
#define SIZE_T_SIZE (sizeof(size_t))
#define SIZE_T_BITSIZE (sizeof(size_t) << 3)
/* Some constants coerced to size_t */
/* Annoying but necessary to avoid errors on some platforms */
#define SIZE_T_ZERO ((size_t)0)
#define SIZE_T_ONE ((size_t)1)
#define SIZE_T_TWO ((size_t)2)
#define SIZE_T_FOUR ((size_t)4)
/* The bit mask value corresponding to MALLOC_ALIGNMENT */
/* True if address a has acceptable alignment */
#define is_aligned(A) (((size_t)((A)) & (CHUNK_ALIGN_MASK)) == 0)
/* the number of bytes to offset an address to align it */
#define align_offset(A)\
((((size_t)(A) & CHUNK_ALIGN_MASK) == 0)? 0 :\
/* -------------------------- MMAP preliminaries ------------------------- */
If HAVE_MORECORE or HAVE_MMAP are false, we just define calls and
checks to fail so compiler optimizer can delete code rather than
using so many "#if"s.
/* MORECORE and MMAP must return MFAIL on failure */
#define MFAIL ((void*)(MAX_SIZE_T))
#define CMFAIL ((char*)(MFAIL)) /* defined for convenience */
#if defined(STARBOARD)
#define MMAP_DEFAULT(s) SbPageMapUntracked(s, kSbMemoryMapProtectReadWrite, "dlmalloc_mmap")
#define MUNMAP_DEFAULT(a, s) (SbPageUnmapUntracked((a), (s)) ? 0 : 1)
#elif defined (__LB_SHELL__)
#define MMAP_DEFAULT(s) lb_mmap_untracked(s, "dlmalloc_mmap")
#define MUNMAP_DEFAULT(a, s) lb_munmap_untracked((a), (s))
#elif !defined(WIN32)
#define MUNMAP_DEFAULT(a, s) munmap((a), (s))
#if !defined(MAP_ANONYMOUS) && defined(MAP_ANON)
#endif /* MAP_ANON */
#define MMAP_DEFAULT(s) mmap(0, (s), MMAP_PROT, MMAP_FLAGS, -1, 0)
#else /* MAP_ANONYMOUS */
Nearly all versions of mmap support MAP_ANONYMOUS, so the following
is unlikely to be needed, but is supplied just in case.
static int dev_zero_fd = -1; /* Cached file descriptor for /dev/zero. */
#define MMAP_DEFAULT(s) ((dev_zero_fd < 0) ? \
(dev_zero_fd = open("/dev/zero", O_RDWR), \
mmap(0, (s), MMAP_PROT, MMAP_FLAGS, dev_zero_fd, 0)) : \
mmap(0, (s), MMAP_PROT, MMAP_FLAGS, dev_zero_fd, 0))
#endif /* MAP_ANONYMOUS */
#else /* WIN32 */
/* Win32 MMAP via VirtualAlloc */
static FORCEINLINE void* win32mmap(size_t size) {
void* ptr = VirtualAlloc(0, size, MEM_RESERVE|MEM_COMMIT, PAGE_READWRITE);
return (ptr != 0)? ptr: MFAIL;
/* For direct MMAP, use MEM_TOP_DOWN to minimize interference */
static FORCEINLINE void* win32direct_mmap(size_t size) {
void* ptr = VirtualAlloc(0, size, MEM_RESERVE|MEM_COMMIT|MEM_TOP_DOWN,
return (ptr != 0)? ptr: MFAIL;
/* This function supports releasing coalesed segments */
static FORCEINLINE int win32munmap(void* ptr, size_t size) {
char* cptr = (char*)ptr;
while (size) {
if (VirtualQuery(cptr, &minfo, sizeof(minfo)) == 0)
return -1;
if (minfo.BaseAddress != cptr || minfo.AllocationBase != cptr ||
minfo.State != MEM_COMMIT || minfo.RegionSize > size)
return -1;
if (VirtualFree(cptr, 0, MEM_RELEASE) == 0)
return -1;
cptr += minfo.RegionSize;
size -= minfo.RegionSize;
return 0;
#define MMAP_DEFAULT(s) win32mmap(s)
#define MUNMAP_DEFAULT(a, s) win32munmap((a), (s))
#define DIRECT_MMAP_DEFAULT(s) win32direct_mmap(s)
#endif /* WIN32 */
#endif /* HAVE_MMAP */
#ifndef WIN32
#define MREMAP_DEFAULT(addr, osz, nsz, mv) mremap((addr), (osz), (nsz), (mv))
#endif /* WIN32 */
#endif /* HAVE_MREMAP */
#else /* MORECORE */
#endif /* MORECORE */
#else /* HAVE_MORECORE */
#endif /* HAVE_MORECORE */
#ifdef MMAP
#define CALL_MMAP(s) MMAP(s)
#else /* MMAP */
#endif /* MMAP */
#ifdef MUNMAP
#define CALL_MUNMAP(a, s) MUNMAP((a), (s))
#else /* MUNMAP */
#define CALL_MUNMAP(a, s) MUNMAP_DEFAULT((a), (s))
#endif /* MUNMAP */
#else /* DIRECT_MMAP */
#endif /* DIRECT_MMAP */
#else /* HAVE_MMAP */
#define MMAP(s) MFAIL
#define MUNMAP(a, s) (-1)
#define CALL_MMAP(s) MMAP(s)
#define CALL_MUNMAP(a, s) MUNMAP((a), (s))
#endif /* HAVE_MMAP */
#ifdef MREMAP
#define CALL_MREMAP(addr, osz, nsz, mv) MREMAP((addr), (osz), (nsz), (mv))
#else /* MREMAP */
#define CALL_MREMAP(addr, osz, nsz, mv) MREMAP_DEFAULT((addr), (osz), (nsz), (mv))
#endif /* MREMAP */
#else /* HAVE_MMAP && HAVE_MREMAP */
#define CALL_MREMAP(addr, osz, nsz, mv) MFAIL
#endif /* HAVE_MMAP && HAVE_MREMAP */
/* mstate bit set if continguous morecore disabled or failed */
/* segment bit set in create_mspace_with_base */
#define EXTERN_BIT (8U)
/* --------------------------- Lock preliminaries ------------------------ */
When locks are defined, there is one global lock, plus
one per-mspace lock.
The global lock_ensures that mparams.magic and other unique
mparams values are initialized only once. It also protects
sequences of calls to MORECORE. In many cases sys_alloc requires
two calls, that should not be interleaved with calls by other
threads. This does not protect against direct calls to MORECORE
by other threads not using this lock, so there is still code to
cope the best we can on interference.
Per-mspace locks surround calls to malloc, free, etc.
By default, locks are simple non-reentrant mutexes.
Because lock-protected regions generally have bounded times, it is
OK to use the supplied simple spinlocks. Spinlocks are likely to
improve performance for lightly contended applications, but worsen
performance under heavy contention.
If USE_LOCKS is > 1, the definitions of lock routines here are
bypassed, in which case you will need to define the type MLOCK_T,
and TRY_LOCK. You must also declare a
static MLOCK_T malloc_global_mutex = { initialization values };.
#define USE_LOCK_BIT (0U)
#define INITIAL_LOCK(l) (0)
#define DESTROY_LOCK(l) (0)
#if USE_LOCKS > 1
/* ----------------------- User-defined locks ------------------------ */
/* Define your own lock implementation here */
/* #define INITIAL_LOCK(lk) ... */
/* #define DESTROY_LOCK(lk) ... */
/* #define ACQUIRE_LOCK(lk) ... */
/* #define RELEASE_LOCK(lk) ... */
/* #define TRY_LOCK(lk) ... */
/* static MLOCK_T malloc_global_mutex = ... */
// See dlmalloc_config.h for definitions of *_LOCK macros.
static MLOCK_T malloc_global_mutex;
static volatile int malloc_global_mutex_status;
static void init_malloc_global_mutex() {
if (malloc_global_mutex_status > 0) {
malloc_global_mutex_status = 1;
/* First, define CAS_LOCK and CLEAR_LOCK on ints */
/* Note CAS_LOCK defined to return 0 on success */
#if defined(__GNUC__)&& (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 1))
#define CAS_LOCK(sl) __sync_lock_test_and_set(sl, 1)
#define CLEAR_LOCK(sl) __sync_lock_release(sl)
#elif (defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__)))
/* Custom spin locks for older gcc on x86 */
static FORCEINLINE int x86_cas_lock(int *sl) {
int ret;
int val = 1;
int cmp = 0;
__asm__ __volatile__ ("lock; cmpxchgl %1, %2"
: "=a" (ret)
: "r" (val), "m" (*(sl)), "0"(cmp)
: "memory", "cc");
return ret;
static FORCEINLINE void x86_clear_lock(int* sl) {
assert(*sl != 0);
int prev = 0;
int ret;
__asm__ __volatile__ ("lock; xchgl %0, %1"
: "=r" (ret)
: "m" (*(sl)), "0"(prev)
: "memory");
#define CAS_LOCK(sl) x86_cas_lock(sl)
#define CLEAR_LOCK(sl) x86_clear_lock(sl)
#else /* Win32 MSC */
#define CAS_LOCK(sl) interlockedexchange(sl, (LONG)1)
#define CLEAR_LOCK(sl) interlockedexchange (sl, (LONG)0)
#endif /* ... gcc spins locks ... */
/* How to yield for a spin lock */
#define SPINS_PER_YIELD 63
#if defined(_MSC_VER)
#define SLEEP_EX_DURATION 50 /* delay for yield/sleep */
#elif defined (__SVR4) && defined (__sun) /* solaris */
#define SPIN_LOCK_YIELD thr_yield();
#elif !defined(LACKS_SCHED_H)
#define SPIN_LOCK_YIELD sched_yield();
#endif /* ... yield ... */
/* Plain spin locks use single word (embedded in malloc_states) */
static int spin_acquire_lock(int *sl) {
int spins = 0;
while (*(volatile int *)sl != 0 || CAS_LOCK(sl)) {
if ((++spins & SPINS_PER_YIELD) == 0) {
return 0;
#define MLOCK_T int
#define TRY_LOCK(sl) !CAS_LOCK(sl)
#define ACQUIRE_LOCK(sl) (CAS_LOCK(sl)? spin_acquire_lock(sl) : 0)
#define INITIAL_LOCK(sl) (*sl = 0)
#define DESTROY_LOCK(sl) (0)
static MLOCK_T malloc_global_mutex = 0;
/* types for lock owners */
#ifdef WIN32
#define CURRENT_THREAD GetCurrentThreadId()
#define EQ_OWNER(X,Y) ((X) == (Y))
Note: the following assume that pthread_t is a type that can be
initialized to (casted) zero. If this is not the case, you will need to
somehow redefine these or not use spin locks.
#define THREAD_ID_T pthread_t
#define CURRENT_THREAD pthread_self()
#define EQ_OWNER(X,Y) pthread_equal(X, Y)
struct malloc_recursive_lock {
int sl;
unsigned int c;
THREAD_ID_T threadid;
#define MLOCK_T struct malloc_recursive_lock
static MLOCK_T malloc_global_mutex = { 0, 0, (THREAD_ID_T)0};
static FORCEINLINE void recursive_release_lock(MLOCK_T *lk) {
assert(lk->sl != 0);
if (--lk->c == 0) {
static FORCEINLINE int recursive_acquire_lock(MLOCK_T *lk) {
int spins = 0;
for (;;) {
if (*((volatile int *)(&lk->sl)) == 0) {
if (!CAS_LOCK(&lk->sl)) {
lk->threadid = mythreadid;
lk->c = 1;
return 0;
else if (EQ_OWNER(lk->threadid, mythreadid)) {
return 0;
if ((++spins & SPINS_PER_YIELD) == 0) {
static FORCEINLINE int recursive_try_lock(MLOCK_T *lk) {
if (*((volatile int *)(&lk->sl)) == 0) {
if (!CAS_LOCK(&lk->sl)) {
lk->threadid = mythreadid;
lk->c = 1;
return 1;
else if (EQ_OWNER(lk->threadid, mythreadid)) {
return 1;
return 0;
#define RELEASE_LOCK(lk) recursive_release_lock(lk)
#define TRY_LOCK(lk) recursive_try_lock(lk)
#define ACQUIRE_LOCK(lk) recursive_acquire_lock(lk)
#define INITIAL_LOCK(lk) ((lk)->threadid = (THREAD_ID_T)0, (lk)->sl = 0, (lk)->c = 0)
#define DESTROY_LOCK(lk) (0)
#elif defined(WIN32) /* Win32 critical sections */
#define ACQUIRE_LOCK(lk) (EnterCriticalSection(lk), 0)
#define RELEASE_LOCK(lk) LeaveCriticalSection(lk)
#define TRY_LOCK(lk) TryEnterCriticalSection(lk)
#define INITIAL_LOCK(lk) (!InitializeCriticalSectionAndSpinCount((lk), 0x80000000|4000))
#define DESTROY_LOCK(lk) (DeleteCriticalSection(lk), 0)
static MLOCK_T malloc_global_mutex;
static volatile LONG malloc_global_mutex_status;
/* Use spin loop to initialize global lock */
static void init_malloc_global_mutex() {
for (;;) {
long stat = malloc_global_mutex_status;
if (stat > 0)
/* transition to < 0 while initializing, then to > 0) */
if (stat == 0 &&
interlockedcompareexchange(&malloc_global_mutex_status, (LONG)-1, (LONG)0) == 0) {
interlockedexchange(&malloc_global_mutex_status, (LONG)1);
SleepEx(0, FALSE);
#else /* pthreads-based locks */
#define MLOCK_T pthread_mutex_t
#define ACQUIRE_LOCK(lk) pthread_mutex_lock(lk)
#define RELEASE_LOCK(lk) pthread_mutex_unlock(lk)
#define TRY_LOCK(lk) (!pthread_mutex_trylock(lk))
#define INITIAL_LOCK(lk) pthread_init_lock(lk)
#define DESTROY_LOCK(lk) pthread_mutex_destroy(lk)
#if defined(USE_RECURSIVE_LOCKS) && USE_RECURSIVE_LOCKS != 0 && defined(linux) && !defined(PTHREAD_MUTEX_RECURSIVE)
/* Cope with old-style linux recursive lock initialization by adding */
/* skipped internal declaration from pthread.h */
extern int pthread_mutexattr_setkind_np __P ((pthread_mutexattr_t *__attr,
int __kind));
#define pthread_mutexattr_settype(x,y) pthread_mutexattr_setkind_np(x,y)
#endif /* USE_RECURSIVE_LOCKS ... */
static MLOCK_T malloc_global_mutex = PTHREAD_MUTEX_INITIALIZER;
static int pthread_init_lock (MLOCK_T *lk) {
pthread_mutexattr_t attr;
if (pthread_mutexattr_init(&attr)) return 1;
if (pthread_mutexattr_settype(&attr, PTHREAD_MUTEX_RECURSIVE)) return 1;
if (pthread_mutex_init(lk, &attr)) return 1;
if (pthread_mutexattr_destroy(&attr)) return 1;
return 0;
#endif /* ... lock types ... */
/* Common code for all lock types */
#define USE_LOCK_BIT (2U)
#define ACQUIRE_MALLOC_GLOBAL_LOCK() ACQUIRE_LOCK(&malloc_global_mutex);
#define RELEASE_MALLOC_GLOBAL_LOCK() RELEASE_LOCK(&malloc_global_mutex);
#endif /* USE_LOCKS */
/* ----------------------- Chunk representations ------------------------ */
(The following includes lightly edited explanations by Colin Plumb.)
The malloc_chunk declaration below is misleading (but accurate and
necessary). It declares a "view" into memory allowing access to
necessary fields at known offsets from a given base.
Chunks of memory are maintained using a `boundary tag' method as
originally described by Knuth. (See the paper by Paul Wilson 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 fast. The head fields also hold bits
representing whether chunks are free or in use.
Here are some pictures to make it clearer. They are "exploded" to
show that the state of a chunk can be thought of as extending from
the high 31 bits of the head field of its header through the
prev_foot and PINUSE_BIT bit of the following chunk header.
A chunk that's in use looks like:
chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Size of previous chunk (if P = 0) |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |P|
| Size of this chunk 1| +-+
mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
+- -+
| |
+- -+
| :
+- size - sizeof(size_t) available payload bytes -+
: |
chunk-> +- -+
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |1|
| Size of next chunk (may or may not be in use) | +-+
mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
And if it's free, it looks like this:
chunk-> +- -+
| User payload (must be in use, or we would have merged!) |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |P|
| Size of this chunk 0| +-+
mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Next pointer |
| Prev pointer |
| :
+- size - sizeof(struct chunk) unused bytes -+
: |
chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Size of this chunk |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |0|
| Size of next chunk (must be in use, or we would have merged)| +-+
mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| :
+- User payload -+
: |
Note that since we always merge adjacent free chunks, the chunks
adjacent to a free chunk must be in use.
Given a pointer to a chunk (which can be derived trivially from the
payload pointer) we can, in O(1) time, find out whether the adjacent
chunks are free, and if so, unlink them from the lists that they
are on and merge them with the current 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 at least double-word aligned.
The P (PINUSE_BIT) 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. If pinuse is set for
any given chunk, then you CANNOT determine the size of the
previous chunk, and might even get a memory addressing fault when
trying to do so.
The C (CINUSE_BIT) bit, stored in the unused second-lowest bit of
the chunk size redundantly records whether the current chunk is
inuse (unless the chunk is mmapped). This redundancy enables usage
checks within free and realloc, and reduces indirection when freeing
and consolidating chunks.
Each freshly allocated chunk must have both cinuse and pinuse set.
That is, each allocated chunk borders either a previously allocated
and still in-use chunk, or the base of its memory arena. This is
ensured by making all allocations from the `lowest' part of any
found chunk. Further, no free chunk physically borders another one,
so each free chunk is known to be preceded and followed by either
inuse chunks or the ends of memory.
Note that the `foot' of the current chunk is actually represented
as the prev_foot of the NEXT chunk. This makes it easier to
deal with alignments etc but can be very confusing when trying
to extend or adapt this code.
The exceptions to all this are
1. The special chunk `top' is the top-most available chunk (i.e.,
the one bordering the end of available memory). It is treated
specially. Top 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). In effect,
the top chunk is treated as larger (and thus less well
fitting) than any other available chunk. The top chunk
doesn't update its trailing size field since there is no next
contiguous chunk that would have to index off it. However,
space is still allocated for it (TOP_FOOT_SIZE) to enable
separation or merging when space is extended.
3. Chunks allocated via mmap, have both cinuse and pinuse bits
cleared in their head fields. Because they are allocated
one-by-one, each must carry its own prev_foot field, which is
also used to hold the offset this chunk has within its mmapped
region, which is needed to preserve alignment. Each mmapped
chunk is trailed by the first two fields of a fake next-chunk
for sake of usage checks.
struct malloc_chunk {
size_t prev_foot; /* Size of previous chunk (if free). */
size_t head; /* Size and inuse bits. */
struct malloc_chunk* fd; /* double links -- used only if free. */
struct malloc_chunk* bk;
typedef struct malloc_chunk mchunk;
typedef struct malloc_chunk* mchunkptr;
typedef struct malloc_chunk* sbinptr; /* The type of bins of chunks */
typedef unsigned int bindex_t; /* Described below */
typedef unsigned int binmap_t; /* Described below */
typedef unsigned int flag_t; /* The type of various bit flag sets */
/* ------------------- Chunks sizes and alignments ----------------------- */
#define MCHUNK_SIZE (sizeof(mchunk))
#else /* FOOTERS */
#endif /* FOOTERS */
/* MMapped chunks need a second word of overhead ... */
/* ... and additional padding for fake next-chunk at foot */
/* The smallest size we can malloc is an aligned minimal chunk */
/* conversion from malloc headers to user pointers, and back */
#define chunk2mem(p) ((void*)((char*)(p) + TWO_SIZE_T_SIZES))
#define mem2chunk(mem) ((mchunkptr)((char*)(mem) - TWO_SIZE_T_SIZES))
/* chunk associated with aligned address A */
#define align_as_chunk(A) (mchunkptr)((A) + align_offset(chunk2mem(A)))
/* Bounds on request (not chunk) sizes. */
#define MAX_REQUEST ((-MIN_CHUNK_SIZE) << 2)
/* pad request bytes into a usable size */
#define pad_request(req) \
/* pad request, checking for minimum (but not maximum) */
#define request2size(req) \
(((req) < MIN_REQUEST)? MIN_CHUNK_SIZE : pad_request(req))
/* ------------------ Operations on head and foot fields ----------------- */
The head field of a chunk is or'ed with PINUSE_BIT when previous
adjacent chunk in use, and or'ed with CINUSE_BIT if this chunk is in
use, unless mmapped, in which case both bits are cleared.
FLAG4_BIT is not used by this malloc, but might be useful in extensions.
/* Head value for fenceposts */
/* extraction of fields from head words */
#define cinuse(p) ((p)->head & CINUSE_BIT)
#define pinuse(p) ((p)->head & PINUSE_BIT)
#define flag4inuse(p) ((p)->head & FLAG4_BIT)
#define is_inuse(p) (((p)->head & INUSE_BITS) != PINUSE_BIT)
#define is_mmapped(p) (((p)->head & INUSE_BITS) == 0)
#define chunksize(p) ((p)->head & ~(FLAG_BITS))
#define clear_pinuse(p) ((p)->head &= ~PINUSE_BIT)
#define set_flag4(p) ((p)->head |= FLAG4_BIT)
#define clear_flag4(p) ((p)->head &= ~FLAG4_BIT)
/* Treat space at ptr +/- offset as a chunk */
#define chunk_plus_offset(p, s) ((mchunkptr)(((char*)(p)) + (s)))
#define chunk_minus_offset(p, s) ((mchunkptr)(((char*)(p)) - (s)))
/* Ptr to next or previous physical malloc_chunk. */
#define next_chunk(p) ((mchunkptr)( ((char*)(p)) + ((p)->head & ~FLAG_BITS)))
#define prev_chunk(p) ((mchunkptr)( ((char*)(p)) - ((p)->prev_foot) ))
/* extract next chunk's pinuse bit */
#define next_pinuse(p) ((next_chunk(p)->head) & PINUSE_BIT)
/* Get/set size at footer */
#define get_foot(p, s) (((mchunkptr)((char*)(p) + (s)))->prev_foot)
#define set_foot(p, s) (((mchunkptr)((char*)(p) + (s)))->prev_foot = (s))
/* Set size, pinuse bit, and foot */
#define set_size_and_pinuse_of_free_chunk(p, s)\
((p)->head = (s|PINUSE_BIT), set_foot(p, s))
/* Set size, pinuse bit, foot, and clear next pinuse */
#define set_free_with_pinuse(p, s, n)\
(clear_pinuse(n), set_size_and_pinuse_of_free_chunk(p, s))
/* Get the internal overhead associated with chunk p */
#define overhead_for(p)\
/* Return true if malloced space is not necessarily cleared */
#define calloc_must_clear(p) (!is_mmapped(p))
#else /* MMAP_CLEARS */
#define calloc_must_clear(p) (1)
#endif /* MMAP_CLEARS */
/* ---------------------- Overlaid data structures ----------------------- */
When chunks are not in use, they are treated as nodes of either
lists or trees.
"Small" 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 |
Larger chunks are kept in a form of bitwise digital trees (aka
tries) keyed on chunksizes. Because malloc_tree_chunks are only for
free chunks greater than 256 bytes, their size doesn't impose any
constraints on user chunk sizes. Each node looks like:
chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Size of previous chunk |
`head:' | Size of chunk, in bytes |P|
mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Forward pointer to next chunk of same size |
| Back pointer to previous chunk of same size |
| Pointer to left child (child[0]) |
| Pointer to right child (child[1]) |
| Pointer to parent |
| bin index of this chunk |
| Unused space .
. |
nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
`foot:' | Size of chunk, in bytes |
Each tree holding treenodes is a tree of unique chunk sizes. Chunks
of the same size are arranged in a circularly-linked list, with only
the oldest chunk (the next to be used, in our FIFO ordering)
actually in the tree. (Tree members are distinguished by a non-null
parent pointer.) If a chunk with the same size an an existing node
is inserted, it is linked off the existing node using pointers that
work in the same way as fd/bk pointers of small chunks.
Each tree contains a power of 2 sized range of chunk sizes (the
smallest is 0x100 <= x < 0x180), which is is divided in half at each
tree level, with the chunks in the smaller half of the range (0x100
<= x < 0x140 for the top nose) in the left subtree and the larger
half (0x140 <= x < 0x180) in the right subtree. This is, of course,
done by inspecting individual bits.
Using these rules, each node's left subtree contains all smaller
sizes than its right subtree. However, the node at the root of each
subtree has no particular ordering relationship to either. (The
dividing line between the subtree sizes is based on trie relation.)
If we remove the last chunk of a given size from the interior of the
tree, we need to replace it with a leaf node. The tree ordering
rules permit a node to be replaced by any leaf below it.
The smallest chunk in a tree (a common operation in a best-fit
allocator) can be found by walking a path to the leftmost leaf in
the tree. Unlike a usual binary tree, where we follow left child
pointers until we reach a null, here we follow the right child
pointer any time the left one is null, until we reach a leaf with
both child pointers null. The smallest chunk in the tree will be
somewhere along that path.
The worst case number of steps to add, find, or remove a node is
bounded by the number of bits differentiating chunks within
bins. Under current bin calculations, this ranges from 6 up to 21
(for 32 bit sizes) or up to 53 (for 64 bit sizes). The typical case
is of course much better.
struct malloc_tree_chunk {
/* The first four fields must be compatible with malloc_chunk */
size_t prev_foot;
size_t head;
struct malloc_tree_chunk* fd;
struct malloc_tree_chunk* bk;
struct malloc_tree_chunk* child[2];
struct malloc_tree_chunk* parent;
bindex_t index;
typedef struct malloc_tree_chunk tchunk;
typedef struct malloc_tree_chunk* tchunkptr;
typedef struct malloc_tree_chunk* tbinptr; /* The type of bins of trees */
/* A little helper macro for trees */
#define leftmost_child(t) ((t)->child[0] != 0? (t)->child[0] : (t)->child[1])
/* ----------------------------- Segments -------------------------------- */
Each malloc space may include non-contiguous segments, held in a
list headed by an embedded malloc_segment record representing the
top-most space. Segments also include flags holding properties of
the space. Large chunks that are directly allocated by mmap are not
included in this list. They are instead independently created and
destroyed without otherwise keeping track of them.
Segment management mainly comes into play for spaces allocated by
MMAP. Any call to MMAP might or might not return memory that is
adjacent to an existing segment. MORECORE normally contiguously
extends the current space, so this space is almost always adjacent,
which is simpler and faster to deal with. (This is why MORECORE is
used preferentially to MMAP when both are available -- see
sys_alloc.) When allocating using MMAP, we don't use any of the
hinting mechanisms (inconsistently) supported in various
implementations of unix mmap, or distinguish reserving from
committing memory. Instead, we just ask for space, and exploit
contiguity when we get it. It is probably possible to do
better than this on some systems, but no general scheme seems
to be significantly better.
Management entails a simpler variant of the consolidation scheme
used for chunks to reduce fragmentation -- new adjacent memory is
normally prepended or appended to an existing segment. However,
there are limitations compared to chunk consolidation that mostly
reflect the fact that segment processing is relatively infrequent
(occurring only when getting memory from system) and that we
don't expect to have huge numbers of segments:
* Segments are not indexed, so traversal requires linear scans. (It
would be possible to index these, but is not worth the extra
overhead and complexity for most programs on most platforms.)
* New segments are only appended to old ones when holding top-most
memory; if they cannot be prepended to others, they are held in
different segments.
Except for the top-most segment of an mstate, each segment record
is kept at the tail of its segment. Segments are added by pushing
segment records onto the list headed by &mstate.seg for the
containing mstate.
Segment flags control allocation/merge/deallocation policies:
* If EXTERN_BIT set, then we did not allocate this segment,
and so should not try to deallocate or merge with others.
(This currently holds only for the initial segment passed
into create_mspace_with_base.)
* If USE_MMAP_BIT set, the segment may be merged with
other surrounding mmapped segments and trimmed/de-allocated
using munmap.
* If neither bit is set, then the segment was obtained using
MORECORE so can be merged with surrounding MORECORE'd segments
and deallocated/trimmed using MORECORE with negative arguments.
struct malloc_segment {
char* base; /* base address */
size_t size; /* allocated size */
struct malloc_segment* next; /* ptr to next segment */
flag_t sflags; /* mmap and extern flag */
#define is_mmapped_segment(S) ((S)->sflags & USE_MMAP_BIT)
#define is_extern_segment(S) ((S)->sflags & EXTERN_BIT)
typedef struct malloc_segment msegment;
typedef struct malloc_segment* msegmentptr;
/* ---------------------------- malloc_state ----------------------------- */
A malloc_state holds all of the bookkeeping for a space.
The main fields are:
The topmost chunk of the currently active segment. Its size is
cached in topsize. The actual size of topmost space is
topsize+TOP_FOOT_SIZE, which includes space reserved for adding
fenceposts and segment records if necessary when getting more
space from the system. The size at which to autotrim top is
cached from mparams in trim_check, except that it is disabled if
an autotrim fails.
Designated victim (dv)
This is the preferred chunk for servicing small requests that
don't have exact fits. It is normally the chunk split off most
recently to service another small request. Its size is cached in
dvsize. The link fields of this chunk are not maintained since it
is not kept in a bin.
An array of bin headers for free chunks. These bins hold chunks
with sizes less than MIN_LARGE_SIZE bytes. Each bin contains
chunks of all the same size, spaced 8 bytes apart. To simplify
use in double-linked lists, each bin header acts as a malloc_chunk
pointing to the real first node, if it exists (else pointing to
itself). This avoids special-casing for headers. But to avoid
waste, we allocate only the fd/bk pointers of bins, and then use
repositioning tricks to treat these as the fields of a chunk.
Treebins are pointers to the roots of trees holding a range of
sizes. There are 2 equally spaced treebins for each power of two
from TREE_SHIFT to TREE_SHIFT+16. The last bin holds anything
Bin maps
There is one bit map for small bins ("smallmap") and one for
treebins ("treemap). Each bin sets its bit when non-empty, and
clears the bit when empty. Bit operations are then used to avoid
bin-by-bin searching -- nearly all "search" is done without ever
looking at bins that won't be selected. The bit maps
conservatively use 32 bits per map word, even if on 64bit system.
For a good description of some of the bit-based techniques used
here, see Henry S. Warren Jr's book "Hacker's Delight" (and
supplement at Many of these are
intended to reduce the branchiness of paths through malloc etc, as
well as to reduce the number of memory locations read or written.
A list of segments headed by an embedded malloc_segment record
representing the initial space.
Address check support
The least_addr field is the least address ever obtained from
MORECORE or MMAP. Attempted frees and reallocs of any address less
than this are trapped (unless INSECURE is defined).
Magic tag
A cross-check field that should always hold same value as mparams.magic.
Max allowed footprint
The maximum allowed bytes to allocate from system (zero means no limit)
Bits recording whether to use MMAP, locks, or contiguous MORECORE
Each space keeps track of current and maximum system memory
obtained via MORECORE or MMAP.
Trim support
Fields holding the amount of unused topmost memory that should trigger
trimming, and a counter to force periodic scanning to release unused
non-topmost segments.
If USE_LOCKS is defined, the "mutex" lock is acquired and released
around every public call using this mspace.
Extension support
A void* pointer and a size_t field that can be used to help implement
extensions to this malloc.
/* Bin types, widths and sizes */
#define NSMALLBINS (32U)
#define NTREEBINS (32U)
#define TREEBIN_SHIFT (8U)
struct malloc_state {
binmap_t smallmap;
binmap_t treemap;
size_t dvsize;
size_t topsize;
char* least_addr;
mchunkptr dv;
mchunkptr top;
size_t trim_check;
size_t release_checks;
size_t magic;
mchunkptr smallbins[(NSMALLBINS+1)*2];
tbinptr treebins[NTREEBINS];
size_t footprint;
size_t max_footprint;
size_t footprint_limit; /* zero means no limit */
flag_t mflags;
MLOCK_T mutex; /* locate lock among fields that rarely change */
#endif /* USE_LOCKS */
msegment seg;
void* extp; /* Unused but available for extensions */
size_t exts;
size_t inuse;
typedef struct malloc_state* mstate;
/* ------------- Global malloc_state and malloc_params ------------------- */
malloc_params holds global properties, including those that can be
dynamically set using mallopt. There is a single instance, mparams,
initialized in init_mparams. Note that the non-zeroness of "magic"
also serves as an initialization flag.
struct malloc_params {
size_t magic;
size_t page_size;
size_t granularity;
size_t mmap_threshold;
size_t trim_threshold;
flag_t default_mflags;
static struct malloc_params mparams;
/* Ensure mparams initialized */
#define ensure_initialization() (void)(mparams.magic != 0 || init_mparams())
/* The global malloc_state used for all non-"mspace" calls */
static struct malloc_state _gm_;
#define gm (&_gm_)
#define is_global(M) ((M) == &_gm_)
#endif /* !ONLY_MSPACES */
#define is_initialized(M) ((M)->top != 0)
/* -------------------------- system alloc setup ------------------------- */
/* Operations on mflags */
#define use_lock(M) ((M)->mflags & USE_LOCK_BIT)
#define enable_lock(M) ((M)->mflags |= USE_LOCK_BIT)
#define disable_lock(M) ((M)->mflags &= ~USE_LOCK_BIT)
#define disable_lock(M)
#define use_mmap(M) ((M)->mflags & USE_MMAP_BIT)
#define enable_mmap(M) ((M)->mflags |= USE_MMAP_BIT)
#define disable_mmap(M) ((M)->mflags &= ~USE_MMAP_BIT)
#define disable_mmap(M)
#define use_noncontiguous(M) ((M)->mflags & USE_NONCONTIGUOUS_BIT)
#define disable_contiguous(M) ((M)->mflags |= USE_NONCONTIGUOUS_BIT)
#define set_lock(M,L)\
((M)->mflags = (L)?\
((M)->mflags | USE_LOCK_BIT) :\
((M)->mflags & ~USE_LOCK_BIT))
/* page-align a size */
#define page_align(S)\
(((S) + (mparams.page_size - SIZE_T_ONE)) & ~(mparams.page_size - SIZE_T_ONE))
/* granularity-align a size */
#define granularity_align(S)\
(((S) + (mparams.granularity - SIZE_T_ONE))\
& ~(mparams.granularity - SIZE_T_ONE))
/* For mmap, use granularity alignment on windows, else page-align */
#ifdef WIN32
#define mmap_align(S) granularity_align(S)
#define mmap_align(S) page_align(S)
/* For sys_alloc, enough padding to ensure can malloc request on success */
#define is_page_aligned(S)\
(((size_t)(S) & (mparams.page_size - SIZE_T_ONE)) == 0)
#define is_granularity_aligned(S)\
(((size_t)(S) & (mparams.granularity - SIZE_T_ONE)) == 0)
/* True if segment S holds address A */
#define segment_holds(S, A)\
((char*)(A) >= S->base && (char*)(A) < S->base + S->size)
/* Return segment holding given address */
static msegmentptr segment_holding(mstate m, char* addr) {
msegmentptr sp = &m->seg;
for (;;) {
if (addr >= sp->base && addr < sp->base + sp->size)
return sp;
if ((sp = sp->next) == 0)
return 0;
/* Return true if segment contains a segment link */
static int has_segment_link(mstate m, msegmentptr ss) {
msegmentptr sp = &m->seg;
for (;;) {
if ((char*)sp >= ss->base && (char*)sp < ss->base + ss->size)
return 1;
if ((sp = sp->next) == 0)
return 0;
#define should_trim(M,s) ((s) > (M)->trim_check)
#define should_trim(M,s) (0)
TOP_FOOT_SIZE is padding at the end of a segment, including space
that may be needed to place segment records and fenceposts when new
noncontiguous segments are added.
#define TOP_FOOT_SIZE\
(align_offset(chunk2mem(0))+pad_request(sizeof(struct malloc_segment))+MIN_CHUNK_SIZE)
/* ------------------------------- Hooks -------------------------------- */
PREACTION should be defined to return 0 on success, and nonzero on
failure. If you are not using locking, you can redefine these to do
anything you like.
#define PREACTION(M) ((use_lock(M))? ACQUIRE_LOCK(&(M)->mutex) : 0)
#define POSTACTION(M) { if (use_lock(M)) RELEASE_LOCK(&(M)->mutex); }
#else /* USE_LOCKS */
#define PREACTION(M) (0)
#endif /* PREACTION */
#endif /* POSTACTION */
#endif /* USE_LOCKS */
CORRUPTION_ERROR_ACTION is triggered upon detected bad addresses.
USAGE_ERROR_ACTION is triggered on detected bad frees and
reallocs. The argument p is an address that might have triggered the
fault. It is ignored by the two predefined actions, but might be
useful in custom actions that try to help diagnose errors.
/* A count of the number of corruption errors causing resets */
int malloc_corruption_error_count;
/* default corruption action */
static void reset_on_error(mstate m);
#define CORRUPTION_ERROR_ACTION(m) reset_on_error(m)
#define USAGE_ERROR_ACTION(m, p)
#else /* PROCEED_ON_ERROR */
#endif /* PROCEED_ON_ERROR */
/* -------------------------- Debugging setup ---------------------------- */
#if ! DEBUG
#define check_free_chunk(M,P)
#define check_inuse_chunk(M,P)
#define check_malloced_chunk(M,P,N)
#define check_mmapped_chunk(M,P)
#define check_malloc_state(M)
#define check_top_chunk(M,P)
#else /* DEBUG */
#define check_free_chunk(M,P) do_check_free_chunk(M,P)
#define check_inuse_chunk(M,P) do_check_inuse_chunk(M,P)
#define check_top_chunk(M,P) do_check_top_chunk(M,P)
#define check_malloced_chunk(M,P,N) do_check_malloced_chunk(M,P,N)
#define check_mmapped_chunk(M,P) do_check_mmapped_chunk(M,P)
#define check_malloc_state(M) do_check_malloc_state(M)
static void do_check_any_chunk(mstate m, mchunkptr p);
static void do_check_top_chunk(mstate m, mchunkptr p);
static void do_check_mmapped_chunk(mstate m, mchunkptr p);
static void do_check_inuse_chunk(mstate m, mchunkptr p);
static void do_check_free_chunk(mstate m, mchunkptr p);
static void do_check_malloced_chunk(mstate m, void* mem, size_t s);
static void do_check_tree(mstate m, tchunkptr t);
static void do_check_treebin(mstate m, bindex_t i);
static void do_check_smallbin(mstate m, bindex_t i);
static void do_check_malloc_state(mstate m);
static int bin_find(mstate m, mchunkptr x);
static size_t traverse_and_check(mstate m);
#endif /* DEBUG */
/* ---------------------------- Indexing Bins ---------------------------- */
#define is_small(s) (((s) >> SMALLBIN_SHIFT) < NSMALLBINS)
#define small_index(s) (bindex_t)((s) >> SMALLBIN_SHIFT)
#define small_index2size(i) ((i) << SMALLBIN_SHIFT)
#define MIN_SMALL_INDEX (small_index(MIN_CHUNK_SIZE))
/* addressing by index. See above about smallbin repositioning */
#define smallbin_at(M, i) ((sbinptr)((char*)&((M)->smallbins[(i)<<1])))
#define treebin_at(M,i) (&((M)->treebins[i]))
/* assign tree index for size S to variable I. Use x86 asm if possible */
#if defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__))
#define compute_tree_index(S, I)\
unsigned int X = S >> TREEBIN_SHIFT;\
if (X == 0)\
I = 0;\
else if (X > 0xFFFF)\
else {\
unsigned int K = (unsigned) sizeof(X)*__CHAR_BIT__ - 1 - (unsigned) __builtin_clz(X); \
I = (bindex_t)((K << 1) + ((S >> (K + (TREEBIN_SHIFT-1)) & 1)));\
#elif defined (__INTEL_COMPILER)
#define compute_tree_index(S, I)\
size_t X = S >> TREEBIN_SHIFT;\
if (X == 0)\
I = 0;\
else if (X > 0xFFFF)\
else {\
unsigned int K = _bit_scan_reverse (X); \
I = (bindex_t)((K << 1) + ((S >> (K + (TREEBIN_SHIFT-1)) & 1)));\
#elif defined(_MSC_VER) && _MSC_VER>=1300 && !defined(__LB_XB360__)
#define compute_tree_index(S, I)\
size_t X = S >> TREEBIN_SHIFT;\
if (X == 0)\
I = 0;\
else if (X > 0xFFFF)\
else {\
unsigned int K;\
_BitScanReverse((DWORD *) &K, (DWORD) X);\
I = (bindex_t)((K << 1) + ((S >> (K + (TREEBIN_SHIFT-1)) & 1)));\
#else /* GNUC */
#define compute_tree_index(S, I)\
size_t X = S >> TREEBIN_SHIFT;\
if (X == 0)\
I = 0;\
else if (X > 0xFFFF)\
else {\
unsigned int Y = (unsigned int)X;\
unsigned int N = ((Y - 0x100) >> 16) & 8;\
unsigned int K = (((Y <<= N) - 0x1000) >> 16) & 4;\
N += K;\
N += K = (((Y <<= K) - 0x4000) >> 16) & 2;\
K = 14 - N + ((Y <<= K) >> 15);\
I = (K << 1) + ((S >> (K + (TREEBIN_SHIFT-1)) & 1));\
#endif /* GNUC */
/* Bit representing maximum resolved size in a treebin at i */
#define bit_for_tree_index(i) \
(i == NTREEBINS-1)? (SIZE_T_BITSIZE-1) : (((i) >> 1) + TREEBIN_SHIFT - 2)
/* Shift placing maximum resolved bit in a treebin at i as sign bit */
#define leftshift_for_tree_index(i) \
((i == NTREEBINS-1)? 0 : \
((SIZE_T_BITSIZE-SIZE_T_ONE) - (((i) >> 1) + TREEBIN_SHIFT - 2)))
/* The size of the smallest chunk held in bin with index i */
#define minsize_for_tree_index(i) \
((SIZE_T_ONE << (((i) >> 1) + TREEBIN_SHIFT)) | \
(((size_t)((i) & SIZE_T_ONE)) << (((i) >> 1) + TREEBIN_SHIFT - 1)))
/* ------------------------ Operations on bin maps ----------------------- */
/* bit corresponding to given index */
#define idx2bit(i) ((binmap_t)(1) << (i))
/* Mark/Clear bits with given index */
#define mark_smallmap(M,i) ((M)->smallmap |= idx2bit(i))
#define clear_smallmap(M,i) ((M)->smallmap &= ~idx2bit(i))
#define smallmap_is_marked(M,i) ((M)->smallmap & idx2bit(i))
#define mark_treemap(M,i) ((M)->treemap |= idx2bit(i))
#define clear_treemap(M,i) ((M)->treemap &= ~idx2bit(i))
#define treemap_is_marked(M,i) ((M)->treemap & idx2bit(i))
/* isolate the least set bit of a bitmap */
#define least_bit(x) ((x) & -(x))
/* mask with all bits to left of least bit of x on */
#define left_bits(x) ((x<<1) | -(x<<1))
/* mask with all bits to left of or equal to least bit of x on */
#define same_or_left_bits(x) ((x) | -(x))
/* index corresponding to given bit. Use x86 asm if possible */
#if defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__))
#define compute_bit2idx(X, I)\
unsigned int J;\
J = __builtin_ctz(X); \
I = (bindex_t)J;\
#elif defined (__INTEL_COMPILER)
#define compute_bit2idx(X, I)\
unsigned int J;\
J = _bit_scan_forward (X); \
I = (bindex_t)J;\
#elif defined(_MSC_VER) && _MSC_VER>=1300 && !defined(__LB_XB360__)
#define compute_bit2idx(X, I)\
unsigned int J;\
_BitScanForward((DWORD *) &J, X);\
I = (bindex_t)J;\
#define compute_bit2idx(X, I) I = ffs(X)-1
#define compute_bit2idx(X, I)\
unsigned int Y = X - 1;\
unsigned int K = Y >> (16-4) & 16;\
unsigned int N = K; Y >>= K;\
N += K = Y >> (8-3) & 8; Y >>= K;\
N += K = Y >> (4-2) & 4; Y >>= K;\
N += K = Y >> (2-1) & 2; Y >>= K;\
N += K = Y >> (1-0) & 1; Y >>= K;\
I = (bindex_t)(N + Y);\
#endif /* GNUC */
/* ----------------------- Runtime Check Support ------------------------- */
For security, the main invariant is that malloc/free/etc never
writes to a static address other than malloc_state, unless static
malloc_state itself has been corrupted, which cannot occur via
malloc (because of these checks). In essence this means that we
believe all pointers, sizes, maps etc held in malloc_state, but
check all of those linked or offsetted from other embedded data
structures. These checks are interspersed with main code in a way
that tends to minimize their run-time cost.
When FOOTERS is defined, in addition to range checking, we also
verify footer fields of inuse chunks, which can be used guarantee
that the mstate controlling malloc/free is intact. This is a
streamlined version of the approach described by William Robertson
et al in "Run-time Detection of Heap-based Overflows" LISA'03 The footer
of an inuse chunk holds the xor of its mstate and a random seed,
that is checked upon calls to free() and realloc(). This is
(probabalistically) unguessable from outside the program, but can be
computed by any code successfully malloc'ing any chunk, so does not
itself provide protection against code that has already broken
security through some other means. Unlike Robertson et al, we
always dynamically check addresses of all offset chunks (previous,
next, etc). This turns out to be cheaper than relying on hashes.
/* Check if address a is at least as high as any from MORECORE or MMAP */
#define ok_address(M, a) ((char*)(a) >= (M)->least_addr)
/* Check if address of next chunk n is higher than base chunk p */
#define ok_next(p, n) ((char*)(p) < (char*)(n))
/* Check if p has inuse status */
#define ok_inuse(p) is_inuse(p)
/* Check if p has its pinuse bit on */
#define ok_pinuse(p) pinuse(p)
#else /* !INSECURE */
#define ok_address(M, a) (1)
#define ok_next(b, n) (1)
#define ok_inuse(p) (1)
#define ok_pinuse(p) (1)
#endif /* !INSECURE */
/* Check if (alleged) mstate m has expected magic field */
#define ok_magic(M) ((M)->magic == mparams.magic)
#else /* (FOOTERS && !INSECURE) */
#define ok_magic(M) (1)
#endif /* (FOOTERS && !INSECURE) */
/* In gcc, use __builtin_expect to minimize impact of checks */
#if defined(__GNUC__) && __GNUC__ >= 3
#define RTCHECK(e) __builtin_expect(e, 1)
#else /* GNUC */
#define RTCHECK(e) (e)
#endif /* GNUC */
#else /* !INSECURE */
#define RTCHECK(e) (1)
#endif /* !INSECURE */
/* macros to set up inuse chunks with or without footers */
#define mark_inuse_foot(M,p,s)
/* Macros for setting head/foot of non-mmapped chunks */
/* Set cinuse bit and pinuse bit of next chunk */
#define set_inuse(M,p,s)\
((p)->head = (((p)->head & PINUSE_BIT)|s|CINUSE_BIT),\
((mchunkptr)(((char*)(p)) + (s)))->head |= PINUSE_BIT)
/* Set cinuse and pinuse of this chunk and pinuse of next chunk */
#define set_inuse_and_pinuse(M,p,s)\
((p)->head = (s|PINUSE_BIT|CINUSE_BIT),\
((mchunkptr)(((char*)(p)) + (s)))->head |= PINUSE_BIT)
/* Set size, cinuse and pinuse bit of this chunk */
#define set_size_and_pinuse_of_inuse_chunk(M, p, s)\
((p)->head = (s|PINUSE_BIT|CINUSE_BIT))
#else /* FOOTERS */
/* Set foot of inuse chunk to be xor of mstate and seed */
#define mark_inuse_foot(M,p,s)\
(((mchunkptr)((char*)(p) + (s)))->prev_foot = ((size_t)(M) ^ mparams.magic))
#define get_mstate_for(p)\
((mstate)(((mchunkptr)((char*)(p) +\
(chunksize(p))))->prev_foot ^ mparams.magic))
#define set_inuse(M,p,s)\
((p)->head = (((p)->head & PINUSE_BIT)|s|CINUSE_BIT),\
(((mchunkptr)(((char*)(p)) + (s)))->head |= PINUSE_BIT), \
#define set_inuse_and_pinuse(M,p,s)\
((p)->head = (s|PINUSE_BIT|CINUSE_BIT),\
(((mchunkptr)(((char*)(p)) + (s)))->head |= PINUSE_BIT),\
#define set_size_and_pinuse_of_inuse_chunk(M, p, s)\
((p)->head = (s|PINUSE_BIT|CINUSE_BIT),\
mark_inuse_foot(M, p, s))
#endif /* !FOOTERS */
/* ---------------------------- setting mparams -------------------------- */
static void pre_fork(void) { ACQUIRE_LOCK(&(gm)->mutex); }
static void post_fork_parent(void) { RELEASE_LOCK(&(gm)->mutex); }
static void post_fork_child(void) { INITIAL_LOCK(&(gm)->mutex); }
#endif /* LOCK_AT_FORK */
#if defined(__LB_SHELL__) || defined(STARBOARD)
static void dl_heap_init();
/* Initialize mparams */
static int init_mparams(void) {
if (malloc_global_mutex_status <= 0)
if (mparams.magic == 0) {
size_t magic;
size_t psize;
size_t gsize;
#ifndef WIN32
psize = malloc_getpagesize;
#else /* WIN32 */
SYSTEM_INFO system_info;
psize = system_info.dwPageSize;
gsize = ((DEFAULT_GRANULARITY != 0)?
DEFAULT_GRANULARITY : system_info.dwAllocationGranularity);
#endif /* WIN32 */
/* Sanity-check configuration:
size_t must be unsigned and as wide as pointer type.
ints must be at least 4 bytes.
alignment must be at least 8.
Alignment, min chunk size, and page size must all be powers of 2.
if ((sizeof(size_t) != sizeof(char*)) ||
(sizeof(int) < 4) ||
(MALLOC_ALIGNMENT < (size_t)8U) ||
((gsize & (gsize-SIZE_T_ONE)) != 0) ||
((psize & (psize-SIZE_T_ONE)) != 0))
mparams.granularity = gsize;
mparams.page_size = psize;
mparams.mmap_threshold = DEFAULT_MMAP_THRESHOLD;
mparams.trim_threshold = DEFAULT_TRIM_THRESHOLD;
mparams.default_mflags = USE_LOCK_BIT|USE_MMAP_BIT;
/* Set up lock for main malloc area */
gm->mflags = mparams.default_mflags;
pthread_atfork(&pre_fork, &post_fork_parent, &post_fork_child);
int fd;
unsigned char buf[sizeof(size_t)];
/* Try to use /dev/urandom, else fall back on using time */
if ((fd = open("/dev/urandom", O_RDONLY)) >= 0 &&
read(fd, buf, sizeof(buf)) == sizeof(buf)) {
magic = *((size_t *) buf);
#endif /* USE_DEV_RANDOM */
#ifdef WIN32
magic = (size_t)(GetTickCount() ^ (size_t)0x55555555U);
#elif defined(LACKS_TIME_H)
magic = (size_t)&magic ^ (size_t)0x55555555U;
magic = (size_t)(time(0) ^ (size_t)0x55555555U);
magic |= (size_t)8U; /* ensure nonzero */
magic &= ~(size_t)7U; /* improve chances of fault for bad values */
/* Until memory modes commonly available, use volatile-write */
(*(volatile size_t *)(&(mparams.magic))) = magic;
#if defined(__LB_SHELL__) || defined(STARBOARD)
// Call our own initialization functions.
return 1;
/* support for mallopt */
static int change_mparam(int param_number, int value) {
size_t val;
val = (value == -1)? MAX_SIZE_T : (size_t)value;
switch(param_number) {
mparams.trim_threshold = val;
return 1;
if (val >= mparams.page_size && ((val & (val-1)) == 0)) {
mparams.granularity = val;
return 1;
return 0;
mparams.mmap_threshold = val;
return 1;
return 0;
/* ------------------------- Debugging Support --------------------------- */
/* Check properties of any chunk, whether free, inuse, mmapped etc */
static void do_check_any_chunk(mstate m, mchunkptr p) {
assert((is_aligned(chunk2mem(p))) || (p->head == FENCEPOST_HEAD));
assert(ok_address(m, p));
/* Check properties of top chunk */
static void do_check_top_chunk(mstate m, mchunkptr p) {
msegmentptr sp = segment_holding(m, (char*)p);
size_t sz = p->head & ~INUSE_BITS; /* third-lowest bit can be set! */
assert(sp != 0);
assert((is_aligned(chunk2mem(p))) || (p->head == FENCEPOST_HEAD));
assert(ok_address(m, p));
assert(sz == m->topsize);
assert(sz > 0);
assert(sz == ((sp->base + sp->size) - (char*)p) - TOP_FOOT_SIZE);
assert(!pinuse(chunk_plus_offset(p, sz)));
/* Check properties of (inuse) mmapped chunks */
static void do_check_mmapped_chunk(mstate m, mchunkptr p) {
size_t sz = chunksize(p);
size_t len = (sz + (p->prev_foot) + MMAP_FOOT_PAD);
assert((is_aligned(chunk2mem(p))) || (p->head == FENCEPOST_HEAD));
assert(ok_address(m, p));
assert((len & (mparams.page_size-SIZE_T_ONE)) == 0);
assert(chunk_plus_offset(p, sz)->head == FENCEPOST_HEAD);
assert(chunk_plus_offset(p, sz+SIZE_T_SIZE)->head == 0);
/* Check properties of inuse chunks */
static void do_check_inuse_chunk(mstate m, mchunkptr p) {
do_check_any_chunk(m, p);
/* If not pinuse and not mmapped, previous chunk has OK offset */
assert(is_mmapped(p) || pinuse(p) || next_chunk(prev_chunk(p)) == p);
if (is_mmapped(p))
do_check_mmapped_chunk(m, p);
/* Check properties of free chunks */
static void do_check_free_chunk(mstate m, mchunkptr p) {
size_t sz = chunksize(p);
mchunkptr next = chunk_plus_offset(p, sz);
do_check_any_chunk(m, p);
assert (!is_mmapped(p));
if (p != m->dv && p != m->top) {
if (sz >= MIN_CHUNK_SIZE) {
assert((sz & CHUNK_ALIGN_MASK) == 0);
assert(next->prev_foot == sz);
assert (next == m->top || is_inuse(next));
assert(p->fd->bk == p);
assert(p->bk->fd == p);
else /* markers are always of size SIZE_T_SIZE */
assert(sz == SIZE_T_SIZE);
/* Check properties of malloced chunks at the point they are malloced */
static void do_check_malloced_chunk(mstate m, void* mem, size_t s) {
if (mem != 0) {
mchunkptr p = mem2chunk(mem);
size_t sz = p->head & ~INUSE_BITS;
do_check_inuse_chunk(m, p);
assert((sz & CHUNK_ALIGN_MASK) == 0);
assert(sz >= MIN_CHUNK_SIZE);
assert(sz >= s);
/* unless mmapped, size is less than MIN_CHUNK_SIZE more than request */
assert(is_mmapped(p) || sz < (s + MIN_CHUNK_SIZE));
/* Check a tree and its subtrees. */
static void do_check_tree(mstate m, tchunkptr t) {
tchunkptr head = 0;
tchunkptr u = t;
bindex_t tindex = t->index;
size_t tsize = chunksize(t);
bindex_t idx;
compute_tree_index(tsize, idx);
assert(tindex == idx);
assert(tsize >= MIN_LARGE_SIZE);
assert(tsize >= minsize_for_tree_index(idx));
assert((idx == NTREEBINS-1) || (tsize < minsize_for_tree_index((idx+1))));
do { /* traverse through chain of same-sized nodes */
do_check_any_chunk(m, ((mchunkptr)u));
assert(u->index == tindex);
assert(chunksize(u) == tsize);
assert(u->fd->bk == u);
assert(u->bk->fd == u);
if (u->parent == 0) {