--- /dev/null
+/*\r
+ This is a version (aka dlmalloc) of malloc/free/realloc written by\r
+ Doug Lea and released to the public domain, as explained at\r
+ http://creativecommons.org/licenses/publicdomain. Send questions,\r
+ comments, complaints, performance data, etc to dl@cs.oswego.edu\r
+\r
+* Version 2.8.3 Thu Sep 22 11:16:15 2005 Doug Lea (dl at gee)\r
+\r
+ Note: There may be an updated version of this malloc obtainable at\r
+ ftp://gee.cs.oswego.edu/pub/misc/malloc.c\r
+ Check before installing!\r
+\r
+* Quickstart\r
+\r
+ This library is all in one file to simplify the most common usage:\r
+ ftp it, compile it (-O3), and link it into another program. All of\r
+ the compile-time options default to reasonable values for use on\r
+ most platforms. You might later want to step through various\r
+ compile-time and dynamic tuning options.\r
+\r
+ For convenience, an include file for code using this malloc is at:\r
+ ftp://gee.cs.oswego.edu/pub/misc/malloc-2.8.3.h\r
+ You don't really need this .h file unless you call functions not\r
+ defined in your system include files. The .h file contains only the\r
+ excerpts from this file needed for using this malloc on ANSI C/C++\r
+ systems, so long as you haven't changed compile-time options about\r
+ naming and tuning parameters. If you do, then you can create your\r
+ own malloc.h that does include all settings by cutting at the point\r
+ indicated below. Note that you may already by default be using a C\r
+ library containing a malloc that is based on some version of this\r
+ malloc (for example in linux). You might still want to use the one\r
+ in this file to customize settings or to avoid overheads associated\r
+ with library versions.\r
+\r
+* Vital statistics:\r
+\r
+ Supported pointer/size_t representation: 4 or 8 bytes\r
+ size_t MUST be an unsigned type of the same width as\r
+ pointers. (If you are using an ancient system that declares\r
+ size_t as a signed type, or need it to be a different width\r
+ than pointers, you can use a previous release of this malloc\r
+ (e.g. 2.7.2) supporting these.)\r
+\r
+ Alignment: 8 bytes (default)\r
+ This suffices for nearly all current machines and C compilers.\r
+ However, you can define MALLOC_ALIGNMENT to be wider than this\r
+ if necessary (up to 128bytes), at the expense of using more space.\r
+\r
+ Minimum overhead per allocated chunk: 4 or 8 bytes (if 4byte sizes)\r
+ 8 or 16 bytes (if 8byte sizes)\r
+ Each malloced chunk has a hidden word of overhead holding size\r
+ and status information, and additional cross-check word\r
+ if FOOTERS is defined.\r
+\r
+ Minimum allocated size: 4-byte ptrs: 16 bytes (including overhead)\r
+ 8-byte ptrs: 32 bytes (including overhead)\r
+\r
+ Even a request for zero bytes (i.e., malloc(0)) returns a\r
+ pointer to something of the minimum allocatable size.\r
+ The maximum overhead wastage (i.e., number of extra bytes\r
+ allocated than were requested in malloc) is less than or equal\r
+ to the minimum size, except for requests >= mmap_threshold that\r
+ are serviced via mmap(), where the worst case wastage is about\r
+ 32 bytes plus the remainder from a system page (the minimal\r
+ mmap unit); typically 4096 or 8192 bytes.\r
+\r
+ Security: static-safe; optionally more or less\r
+ The "security" of malloc refers to the ability of malicious\r
+ code to accentuate the effects of errors (for example, freeing\r
+ space that is not currently malloc'ed or overwriting past the\r
+ ends of chunks) in code that calls malloc. This malloc\r
+ guarantees not to modify any memory locations below the base of\r
+ heap, i.e., static variables, even in the presence of usage\r
+ errors. The routines additionally detect most improper frees\r
+ and reallocs. All this holds as long as the static bookkeeping\r
+ for malloc itself is not corrupted by some other means. This\r
+ is only one aspect of security -- these checks do not, and\r
+ cannot, detect all possible programming errors.\r
+\r
+ If FOOTERS is defined nonzero, then each allocated chunk\r
+ carries an additional check word to verify that it was malloced\r
+ from its space. These check words are the same within each\r
+ execution of a program using malloc, but differ across\r
+ executions, so externally crafted fake chunks cannot be\r
+ freed. This improves security by rejecting frees/reallocs that\r
+ could corrupt heap memory, in addition to the checks preventing\r
+ writes to statics that are always on. This may further improve\r
+ security at the expense of time and space overhead. (Note that\r
+ FOOTERS may also be worth using with MSPACES.)\r
+\r
+ By default detected errors cause the program to abort (calling\r
+ "abort()"). You can override this to instead proceed past\r
+ errors by defining PROCEED_ON_ERROR. In this case, a bad free\r
+ has no effect, and a malloc that encounters a bad address\r
+ caused by user overwrites will ignore the bad address by\r
+ dropping pointers and indices to all known memory. This may\r
+ be appropriate for programs that should continue if at all\r
+ possible in the face of programming errors, although they may\r
+ run out of memory because dropped memory is never reclaimed.\r
+\r
+ If you don't like either of these options, you can define\r
+ CORRUPTION_ERROR_ACTION and USAGE_ERROR_ACTION to do anything\r
+ else. And if if you are sure that your program using malloc has\r
+ no errors or vulnerabilities, you can define INSECURE to 1,\r
+ which might (or might not) provide a small performance improvement.\r
+\r
+ Thread-safety: NOT thread-safe unless USE_LOCKS defined\r
+ When USE_LOCKS is defined, each public call to malloc, free,\r
+ etc is surrounded with either a pthread mutex or a win32\r
+ spinlock (depending on WIN32). This is not especially fast, and\r
+ can be a major bottleneck. It is designed only to provide\r
+ minimal protection in concurrent environments, and to provide a\r
+ basis for extensions. If you are using malloc in a concurrent\r
+ program, consider instead using ptmalloc, which is derived from\r
+ a version of this malloc. (See http://www.malloc.de).\r
+\r
+ System requirements: Any combination of MORECORE and/or MMAP/MUNMAP\r
+ This malloc can use unix sbrk or any emulation (invoked using\r
+ the CALL_MORECORE macro) and/or mmap/munmap or any emulation\r
+ (invoked using CALL_MMAP/CALL_MUNMAP) to get and release system\r
+ memory. On most unix systems, it tends to work best if both\r
+ MORECORE and MMAP are enabled. On Win32, it uses emulations\r
+ based on VirtualAlloc. It also uses common C library functions\r
+ like memset.\r
+\r
+ Compliance: I believe it is compliant with the Single Unix Specification\r
+ (See http://www.unix.org). Also SVID/XPG, ANSI C, and probably\r
+ others as well.\r
+\r
+* Overview of algorithms\r
+\r
+ This is not the fastest, most space-conserving, most portable, or\r
+ most tunable malloc ever written. However it is among the fastest\r
+ while also being among the most space-conserving, portable and\r
+ tunable. Consistent balance across these factors results in a good\r
+ general-purpose allocator for malloc-intensive programs.\r
+\r
+ In most ways, this malloc is a best-fit allocator. Generally, it\r
+ chooses the best-fitting existing chunk for a request, with ties\r
+ broken in approximately least-recently-used order. (This strategy\r
+ normally maintains low fragmentation.) However, for requests less\r
+ than 256bytes, it deviates from best-fit when there is not an\r
+ exactly fitting available chunk by preferring to use space adjacent\r
+ to that used for the previous small request, as well as by breaking\r
+ ties in approximately most-recently-used order. (These enhance\r
+ locality of series of small allocations.) And for very large requests\r
+ (>= 256Kb by default), it relies on system memory mapping\r
+ facilities, if supported. (This helps avoid carrying around and\r
+ possibly fragmenting memory used only for large chunks.)\r
+\r
+ All operations (except malloc_stats and mallinfo) have execution\r
+ times that are bounded by a constant factor of the number of bits in\r
+ a size_t, not counting any clearing in calloc or copying in realloc,\r
+ or actions surrounding MORECORE and MMAP that have times\r
+ proportional to the number of non-contiguous regions returned by\r
+ system allocation routines, which is often just 1.\r
+\r
+ The implementation is not very modular and seriously overuses\r
+ macros. Perhaps someday all C compilers will do as good a job\r
+ inlining modular code as can now be done by brute-force expansion,\r
+ but now, enough of them seem not to.\r
+\r
+ Some compilers issue a lot of warnings about code that is\r
+ dead/unreachable only on some platforms, and also about intentional\r
+ uses of negation on unsigned types. All known cases of each can be\r
+ ignored.\r
+\r
+ For a longer but out of date high-level description, see\r
+ http://gee.cs.oswego.edu/dl/html/malloc.html\r
+\r
+* MSPACES\r
+ If MSPACES is defined, then in addition to malloc, free, etc.,\r
+ this file also defines mspace_malloc, mspace_free, etc. These\r
+ are versions of malloc routines that take an "mspace" argument\r
+ obtained using create_mspace, to control all internal bookkeeping.\r
+ If ONLY_MSPACES is defined, only these versions are compiled.\r
+ So if you would like to use this allocator for only some allocations,\r
+ and your system malloc for others, you can compile with\r
+ ONLY_MSPACES and then do something like...\r
+ static mspace mymspace = create_mspace(0,0); // for example\r
+ #define mymalloc(bytes) mspace_malloc(mymspace, bytes)\r
+\r
+ (Note: If you only need one instance of an mspace, you can instead\r
+ use "USE_DL_PREFIX" to relabel the global malloc.)\r
+\r
+ You can similarly create thread-local allocators by storing\r
+ mspaces as thread-locals. For example:\r
+ static __thread mspace tlms = 0;\r
+ void* tlmalloc(size_t bytes) {\r
+ if (tlms == 0) tlms = create_mspace(0, 0);\r
+ return mspace_malloc(tlms, bytes);\r
+ }\r
+ void tlfree(void* mem) { mspace_free(tlms, mem); }\r
+\r
+ Unless FOOTERS is defined, each mspace is completely independent.\r
+ You cannot allocate from one and free to another (although\r
+ conformance is only weakly checked, so usage errors are not always\r
+ caught). If FOOTERS is defined, then each chunk carries around a tag\r
+ indicating its originating mspace, and frees are directed to their\r
+ originating spaces.\r
+\r
+ ------------------------- Compile-time options ---------------------------\r
+\r
+Be careful in setting #define values for numerical constants of type\r
+size_t. On some systems, literal values are not automatically extended\r
+to size_t precision unless they are explicitly casted.\r
+\r
+WIN32 default: defined if _WIN32 defined\r
+ Defining WIN32 sets up defaults for MS environment and compilers.\r
+ Otherwise defaults are for unix.\r
+\r
+MALLOC_ALIGNMENT default: (size_t)8\r
+ Controls the minimum alignment for malloc'ed chunks. It must be a\r
+ power of two and at least 8, even on machines for which smaller\r
+ alignments would suffice. It may be defined as larger than this\r
+ though. Note however that code and data structures are optimized for\r
+ the case of 8-byte alignment.\r
+\r
+MSPACES default: 0 (false)\r
+ If true, compile in support for independent allocation spaces.\r
+ This is only supported if HAVE_MMAP is true.\r
+\r
+ONLY_MSPACES default: 0 (false)\r
+ If true, only compile in mspace versions, not regular versions.\r
+\r
+USE_LOCKS default: 0 (false)\r
+ Causes each call to each public routine to be surrounded with\r
+ pthread or WIN32 mutex lock/unlock. (If set true, this can be\r
+ overridden on a per-mspace basis for mspace versions.)\r
+\r
+FOOTERS default: 0\r
+ If true, provide extra checking and dispatching by placing\r
+ information in the footers of allocated chunks. This adds\r
+ space and time overhead.\r
+\r
+INSECURE default: 0\r
+ If true, omit checks for usage errors and heap space overwrites.\r
+\r
+USE_DL_PREFIX default: NOT defined\r
+ Causes compiler to prefix all public routines with the string 'dl'.\r
+ This can be useful when you only want to use this malloc in one part\r
+ of a program, using your regular system malloc elsewhere.\r
+\r
+ABORT default: defined as abort()\r
+ Defines how to abort on failed checks. On most systems, a failed\r
+ check cannot die with an "assert" or even print an informative\r
+ message, because the underlying print routines in turn call malloc,\r
+ which will fail again. Generally, the best policy is to simply call\r
+ abort(). It's not very useful to do more than this because many\r
+ errors due to overwriting will show up as address faults (null, odd\r
+ addresses etc) rather than malloc-triggered checks, so will also\r
+ abort. Also, most compilers know that abort() does not return, so\r
+ can better optimize code conditionally calling it.\r
+\r
+PROCEED_ON_ERROR default: defined as 0 (false)\r
+ Controls whether detected bad addresses cause them to bypassed\r
+ rather than aborting. If set, detected bad arguments to free and\r
+ realloc are ignored. And all bookkeeping information is zeroed out\r
+ upon a detected overwrite of freed heap space, thus losing the\r
+ ability to ever return it from malloc again, but enabling the\r
+ application to proceed. If PROCEED_ON_ERROR is defined, the\r
+ static variable malloc_corruption_error_count is compiled in\r
+ and can be examined to see if errors have occurred. This option\r
+ generates slower code than the default abort policy.\r
+\r
+DEBUG default: NOT defined\r
+ The DEBUG setting is mainly intended for people trying to modify\r
+ this code or diagnose problems when porting to new platforms.\r
+ However, it may also be able to better isolate user errors than just\r
+ using runtime checks. The assertions in the check routines spell\r
+ out in more detail the assumptions and invariants underlying the\r
+ algorithms. The checking is fairly extensive, and will slow down\r
+ execution noticeably. Calling malloc_stats or mallinfo with DEBUG\r
+ set will attempt to check every non-mmapped allocated and free chunk\r
+ in the course of computing the summaries.\r
+\r
+ABORT_ON_ASSERT_FAILURE default: defined as 1 (true)\r
+ Debugging assertion failures can be nearly impossible if your\r
+ version of the assert macro causes malloc to be called, which will\r
+ lead to a cascade of further failures, blowing the runtime stack.\r
+ ABORT_ON_ASSERT_FAILURE cause assertions failures to call abort(),\r
+ which will usually make debugging easier.\r
+\r
+MALLOC_FAILURE_ACTION default: sets errno to ENOMEM, or no-op on win32\r
+ The action to take before "return 0" when malloc fails to be able to\r
+ return memory because there is none available.\r
+\r
+HAVE_MORECORE default: 1 (true) unless win32 or ONLY_MSPACES\r
+ True if this system supports sbrk or an emulation of it.\r
+\r
+MORECORE default: sbrk\r
+ The name of the sbrk-style system routine to call to obtain more\r
+ memory. See below for guidance on writing custom MORECORE\r
+ functions. The type of the argument to sbrk/MORECORE varies across\r
+ systems. It cannot be size_t, because it supports negative\r
+ arguments, so it is normally the signed type of the same width as\r
+ size_t (sometimes declared as "intptr_t"). It doesn't much matter\r
+ though. Internally, we only call it with arguments less than half\r
+ the max value of a size_t, which should work across all reasonable\r
+ possibilities, although sometimes generating compiler warnings. See\r
+ near the end of this file for guidelines for creating a custom\r
+ version of MORECORE.\r
+\r
+MORECORE_CONTIGUOUS default: 1 (true)\r
+ If true, take advantage of fact that consecutive calls to MORECORE\r
+ with positive arguments always return contiguous increasing\r
+ addresses. This is true of unix sbrk. It does not hurt too much to\r
+ set it true anyway, since malloc copes with non-contiguities.\r
+ Setting it false when definitely non-contiguous saves time\r
+ and possibly wasted space it would take to discover this though.\r
+\r
+MORECORE_CANNOT_TRIM default: NOT defined\r
+ True if MORECORE cannot release space back to the system when given\r
+ negative arguments. This is generally necessary only if you are\r
+ using a hand-crafted MORECORE function that cannot handle negative\r
+ arguments.\r
+\r
+DONT_MERGE_SEGMENTS default: NOT defined\r
+ Defined if no attempt should be made to merge memory segments\r
+ returned by either MORECORE or CALL_MMAP. This is mostly an\r
+ optimization step to avoid having to walk through all the memory\r
+ segments when it is known in advance that a merge will not be possible.\r
+\r
+HAVE_MMAP default: 1 (true)\r
+ True if this system supports mmap or an emulation of it. If so, and\r
+ HAVE_MORECORE is not true, MMAP is used for all system\r
+ allocation. If set and HAVE_MORECORE is true as well, MMAP is\r
+ primarily used to directly allocate very large blocks. It is also\r
+ used as a backup strategy in cases where MORECORE fails to provide\r
+ space from system. Note: A single call to MUNMAP is assumed to be\r
+ able to unmap memory that may have be allocated using multiple calls\r
+ to MMAP, so long as they are adjacent.\r
+\r
+HAVE_MREMAP default: 1 on linux, else 0\r
+ If true realloc() uses mremap() to re-allocate large blocks and\r
+ extend or shrink allocation spaces.\r
+\r
+MMAP_CLEARS default: 1 on unix\r
+ True if mmap clears memory so calloc doesn't need to. This is true\r
+ for standard unix mmap using /dev/zero.\r
+\r
+USE_BUILTIN_FFS default: 0 (i.e., not used)\r
+ Causes malloc to use the builtin ffs() function to compute indices.\r
+ Some compilers may recognize and intrinsify ffs to be faster than the\r
+ supplied C version. Also, the case of x86 using gcc is special-cased\r
+ to an asm instruction, so is already as fast as it can be, and so\r
+ this setting has no effect. (On most x86s, the asm version is only\r
+ slightly faster than the C version.)\r
+\r
+malloc_getpagesize default: derive from system includes, or 4096.\r
+ The system page size. To the extent possible, this malloc manages\r
+ memory from the system in page-size units. This may be (and\r
+ usually is) a function rather than a constant. This is ignored\r
+ if WIN32, where page size is determined using getSystemInfo during\r
+ initialization.\r
+\r
+USE_DEV_RANDOM default: 0 (i.e., not used)\r
+ Causes malloc to use /dev/random to initialize secure magic seed for\r
+ stamping footers. Otherwise, the current time is used.\r
+\r
+NO_MALLINFO default: 0\r
+ If defined, don't compile "mallinfo". This can be a simple way\r
+ of dealing with mismatches between system declarations and\r
+ those in this file.\r
+\r
+MALLINFO_FIELD_TYPE default: size_t\r
+ The type of the fields in the mallinfo struct. This was originally\r
+ defined as "int" in SVID etc, but is more usefully defined as\r
+ size_t. The value is used only if HAVE_USR_INCLUDE_MALLOC_H is not set\r
+\r
+REALLOC_ZERO_BYTES_FREES default: not defined\r
+ This should be set if a call to realloc with zero bytes should \r
+ be the same as a call to free. Some people think it should. Otherwise, \r
+ since this malloc returns a unique pointer for malloc(0), so does \r
+ realloc(p, 0).\r
+\r
+LACKS_UNISTD_H, LACKS_FCNTL_H, LACKS_SYS_PARAM_H, LACKS_SYS_MMAN_H\r
+LACKS_STRINGS_H, LACKS_STRING_H, LACKS_SYS_TYPES_H, LACKS_ERRNO_H\r
+LACKS_STDLIB_H default: NOT defined unless on WIN32\r
+ Define these if your system does not have these header files.\r
+ You might need to manually insert some of the declarations they provide.\r
+\r
+DEFAULT_GRANULARITY default: page size if MORECORE_CONTIGUOUS,\r
+ system_info.dwAllocationGranularity in WIN32,\r
+ otherwise 64K.\r
+ Also settable using mallopt(M_GRANULARITY, x)\r
+ The unit for allocating and deallocating memory from the system. On\r
+ most systems with contiguous MORECORE, there is no reason to\r
+ make this more than a page. However, systems with MMAP tend to\r
+ either require or encourage larger granularities. You can increase\r
+ this value to prevent system allocation functions to be called so\r
+ often, especially if they are slow. The value must be at least one\r
+ page and must be a power of two. Setting to 0 causes initialization\r
+ to either page size or win32 region size. (Note: In previous\r
+ versions of malloc, the equivalent of this option was called\r
+ "TOP_PAD")\r
+\r
+DEFAULT_TRIM_THRESHOLD default: 2MB\r
+ Also settable using mallopt(M_TRIM_THRESHOLD, x)\r
+ The maximum amount of unused top-most memory to keep before\r
+ releasing via malloc_trim in free(). Automatic trimming is mainly\r
+ useful in long-lived programs using contiguous MORECORE. Because\r
+ trimming via sbrk can be slow on some systems, and can sometimes be\r
+ wasteful (in cases where programs immediately afterward allocate\r
+ more large chunks) the value should be high enough so that your\r
+ overall system performance would improve by releasing this much\r
+ memory. As a rough guide, you might set to a value close to the\r
+ average size of a process (program) running on your system.\r
+ Releasing this much memory would allow such a process to run in\r
+ memory. Generally, it is worth tuning trim thresholds when a\r
+ program undergoes phases where several large chunks are allocated\r
+ and released in ways that can reuse each other's storage, perhaps\r
+ mixed with phases where there are no such chunks at all. The trim\r
+ value must be greater than page size to have any useful effect. To\r
+ disable trimming completely, you can set to MAX_SIZE_T. Note that the trick\r
+ some people use of mallocing a huge space and then freeing it at\r
+ program startup, in an attempt to reserve system memory, doesn't\r
+ have the intended effect under automatic trimming, since that memory\r
+ will immediately be returned to the system.\r
+\r
+DEFAULT_MMAP_THRESHOLD default: 256K\r
+ Also settable using mallopt(M_MMAP_THRESHOLD, x)\r
+ The request size threshold for using MMAP to directly service a\r
+ request. Requests of at least this size that cannot be allocated\r
+ using already-existing space will be serviced via mmap. (If enough\r
+ normal freed space already exists it is used instead.) Using mmap\r
+ segregates relatively large chunks of memory so that they can be\r
+ individually obtained and released from the host system. A request\r
+ serviced through mmap is never reused by any other request (at least\r
+ not directly; the system may just so happen to remap successive\r
+ requests to the same locations). Segregating space in this way has\r
+ the benefits that: Mmapped space can always be individually released\r
+ back to the system, which helps keep the system level memory demands\r
+ of a long-lived program low. Also, mapped memory doesn't become\r
+ `locked' between other chunks, as can happen with normally allocated\r
+ chunks, which means that even trimming via malloc_trim would not\r
+ release them. However, it has the disadvantage that the space\r
+ cannot be reclaimed, consolidated, and then used to service later\r
+ requests, as happens with normal chunks. The advantages of mmap\r
+ nearly always outweigh disadvantages for "large" chunks, but the\r
+ value of "large" may vary across systems. The default is an\r
+ empirically derived value that works well in most systems. You can\r
+ disable mmap by setting to MAX_SIZE_T.\r
+\r
+*/\r
+\r
+#ifndef WIN32\r
+#ifdef _WIN32\r
+#define WIN32 1\r
+#endif /* _WIN32 */\r
+#endif /* WIN32 */\r
+#ifdef WIN32\r
+#define WIN32_LEAN_AND_MEAN\r
+#include <windows.h>\r
+#define HAVE_MMAP 1\r
+#define HAVE_MORECORE 0\r
+#define LACKS_UNISTD_H\r
+#define LACKS_SYS_PARAM_H\r
+#define LACKS_SYS_MMAN_H\r
+#define LACKS_STRING_H\r
+#define LACKS_STRINGS_H\r
+#define LACKS_SYS_TYPES_H\r
+#define LACKS_ERRNO_H\r
+#define MALLOC_FAILURE_ACTION\r
+#define MMAP_CLEARS 1 /* WINCE and some others apparently don't clear */\r
+#endif /* WIN32 */\r
+\r
+#if defined(DARWIN) || defined(_DARWIN)\r
+/* Mac OSX docs advise not to use sbrk; it seems better to use mmap */\r
+#ifndef HAVE_MORECORE\r
+#define HAVE_MORECORE 0\r
+#define HAVE_MMAP 1\r
+#endif /* HAVE_MORECORE */\r
+#endif /* DARWIN */\r
+\r
+#ifndef LACKS_SYS_TYPES_H\r
+#include <sys/types.h> /* For size_t */\r
+#endif /* LACKS_SYS_TYPES_H */\r
+\r
+/* The maximum possible size_t value has all bits set */\r
+#define MAX_SIZE_T (~(size_t)0)\r
+\r
+#ifndef ONLY_MSPACES\r
+#define ONLY_MSPACES 0\r
+#endif /* ONLY_MSPACES */\r
+#ifndef MSPACES\r
+#if ONLY_MSPACES\r
+#define MSPACES 1\r
+#else /* ONLY_MSPACES */\r
+#define MSPACES 0\r
+#endif /* ONLY_MSPACES */\r
+#endif /* MSPACES */\r
+#ifndef MALLOC_ALIGNMENT\r
+#define MALLOC_ALIGNMENT ((size_t)8U)\r
+#endif /* MALLOC_ALIGNMENT */\r
+#ifndef FOOTERS\r
+#define FOOTERS 0\r
+#endif /* FOOTERS */\r
+#ifndef ABORT\r
+#define ABORT abort()\r
+#endif /* ABORT */\r
+#ifndef ABORT_ON_ASSERT_FAILURE\r
+#define ABORT_ON_ASSERT_FAILURE 1\r
+#endif /* ABORT_ON_ASSERT_FAILURE */\r
+#ifndef PROCEED_ON_ERROR\r
+#define PROCEED_ON_ERROR 0\r
+#endif /* PROCEED_ON_ERROR */\r
+#ifndef USE_LOCKS\r
+#define USE_LOCKS 0\r
+#endif /* USE_LOCKS */\r
+#ifndef INSECURE\r
+#define INSECURE 0\r
+#endif /* INSECURE */\r
+#ifndef HAVE_MMAP\r
+#define HAVE_MMAP 1\r
+#endif /* HAVE_MMAP */\r
+#ifndef MMAP_CLEARS\r
+#define MMAP_CLEARS 1\r
+#endif /* MMAP_CLEARS */\r
+#ifndef HAVE_MREMAP\r
+#ifdef linux\r
+#define HAVE_MREMAP 1\r
+#else /* linux */\r
+#define HAVE_MREMAP 0\r
+#endif /* linux */\r
+#endif /* HAVE_MREMAP */\r
+#ifndef MALLOC_FAILURE_ACTION\r
+#define MALLOC_FAILURE_ACTION errno = ENOMEM;\r
+#endif /* MALLOC_FAILURE_ACTION */\r
+#ifndef HAVE_MORECORE\r
+#if ONLY_MSPACES\r
+#define HAVE_MORECORE 0\r
+#else /* ONLY_MSPACES */\r
+#define HAVE_MORECORE 1\r
+#endif /* ONLY_MSPACES */\r
+#endif /* HAVE_MORECORE */\r
+#if !HAVE_MORECORE\r
+#define MORECORE_CONTIGUOUS 0\r
+#else /* !HAVE_MORECORE */\r
+#ifndef MORECORE\r
+#define MORECORE sbrk\r
+#endif /* MORECORE */\r
+#ifndef MORECORE_CONTIGUOUS\r
+#define MORECORE_CONTIGUOUS 1\r
+#endif /* MORECORE_CONTIGUOUS */\r
+#endif /* HAVE_MORECORE */\r
+#ifndef DEFAULT_GRANULARITY\r
+#if MORECORE_CONTIGUOUS\r
+#define DEFAULT_GRANULARITY (0) /* 0 means to compute in init_mparams */\r
+#else /* MORECORE_CONTIGUOUS */\r
+#define DEFAULT_GRANULARITY ((size_t)64U * (size_t)1024U)\r
+#endif /* MORECORE_CONTIGUOUS */\r
+#endif /* DEFAULT_GRANULARITY */\r
+#ifndef DEFAULT_TRIM_THRESHOLD\r
+#ifndef MORECORE_CANNOT_TRIM\r
+#define DEFAULT_TRIM_THRESHOLD ((size_t)2U * (size_t)1024U * (size_t)1024U)\r
+#else /* MORECORE_CANNOT_TRIM */\r
+#define DEFAULT_TRIM_THRESHOLD MAX_SIZE_T\r
+#endif /* MORECORE_CANNOT_TRIM */\r
+#endif /* DEFAULT_TRIM_THRESHOLD */\r
+#ifndef DEFAULT_MMAP_THRESHOLD\r
+#if HAVE_MMAP\r
+#define DEFAULT_MMAP_THRESHOLD ((size_t)256U * (size_t)1024U)\r
+#else /* HAVE_MMAP */\r
+#define DEFAULT_MMAP_THRESHOLD MAX_SIZE_T\r
+#endif /* HAVE_MMAP */\r
+#endif /* DEFAULT_MMAP_THRESHOLD */\r
+#ifndef USE_BUILTIN_FFS\r
+#define USE_BUILTIN_FFS 0\r
+#endif /* USE_BUILTIN_FFS */\r
+#ifndef USE_DEV_RANDOM\r
+#define USE_DEV_RANDOM 0\r
+#endif /* USE_DEV_RANDOM */\r
+#ifndef NO_MALLINFO\r
+#define NO_MALLINFO 0\r
+#endif /* NO_MALLINFO */\r
+#ifndef MALLINFO_FIELD_TYPE\r
+#define MALLINFO_FIELD_TYPE size_t\r
+#endif /* MALLINFO_FIELD_TYPE */\r
+\r
+/*\r
+ mallopt tuning options. SVID/XPG defines four standard parameter\r
+ numbers for mallopt, normally defined in malloc.h. None of these\r
+ are used in this malloc, so setting them has no effect. But this\r
+ malloc does support the following options.\r
+*/\r
+\r
+#define M_TRIM_THRESHOLD (-1)\r
+#define M_GRANULARITY (-2)\r
+#define M_MMAP_THRESHOLD (-3)\r
+\r
+/* ------------------------ Mallinfo declarations ------------------------ */\r
+\r
+#if !NO_MALLINFO\r
+/*\r
+ This version of malloc supports the standard SVID/XPG mallinfo\r
+ routine that returns a struct containing usage properties and\r
+ statistics. It should work on any system that has a\r
+ /usr/include/malloc.h defining struct mallinfo. The main\r
+ declaration needed is the mallinfo struct that is returned (by-copy)\r
+ by mallinfo(). The malloinfo struct contains a bunch of fields that\r
+ are not even meaningful in this version of malloc. These fields are\r
+ are instead filled by mallinfo() with other numbers that might be of\r
+ interest.\r
+\r
+ HAVE_USR_INCLUDE_MALLOC_H should be set if you have a\r
+ /usr/include/malloc.h file that includes a declaration of struct\r
+ mallinfo. If so, it is included; else a compliant version is\r
+ declared below. These must be precisely the same for mallinfo() to\r
+ work. The original SVID version of this struct, defined on most\r
+ systems with mallinfo, declares all fields as ints. But some others\r
+ define as unsigned long. If your system defines the fields using a\r
+ type of different width than listed here, you MUST #include your\r
+ system version and #define HAVE_USR_INCLUDE_MALLOC_H.\r
+*/\r
+\r
+/* #define HAVE_USR_INCLUDE_MALLOC_H */\r
+\r
+#ifdef HAVE_USR_INCLUDE_MALLOC_H\r
+#include "/usr/include/malloc.h"\r
+#else /* HAVE_USR_INCLUDE_MALLOC_H */\r
+\r
+struct mallinfo {\r
+ MALLINFO_FIELD_TYPE arena; /* non-mmapped space allocated from system */\r
+ MALLINFO_FIELD_TYPE ordblks; /* number of free chunks */\r
+ MALLINFO_FIELD_TYPE smblks; /* always 0 */\r
+ MALLINFO_FIELD_TYPE hblks; /* always 0 */\r
+ MALLINFO_FIELD_TYPE hblkhd; /* space in mmapped regions */\r
+ MALLINFO_FIELD_TYPE usmblks; /* maximum total allocated space */\r
+ MALLINFO_FIELD_TYPE fsmblks; /* always 0 */\r
+ MALLINFO_FIELD_TYPE uordblks; /* total allocated space */\r
+ MALLINFO_FIELD_TYPE fordblks; /* total free space */\r
+ MALLINFO_FIELD_TYPE keepcost; /* releasable (via malloc_trim) space */\r
+};\r
+\r
+#endif /* HAVE_USR_INCLUDE_MALLOC_H */\r
+#endif /* NO_MALLINFO */\r
+\r
+#ifndef FORCEINLINE\r
+ #if defined(__GNUC__)\r
+#define FORCEINLINE __inline __attribute__ ((always_inline))\r
+ #elif defined(_MSC_VER)\r
+ #define FORCEINLINE __forceinline\r
+ #endif\r
+#endif\r
+#ifndef NOINLINE\r
+ #if defined(__GNUC__)\r
+ #define NOINLINE __attribute__ ((noinline))\r
+ #elif defined(_MSC_VER)\r
+ #define NOINLINE __declspec(noinline)\r
+ #else\r
+ #define NOINLINE\r
+ #endif\r
+#endif\r
+\r
+#ifdef __cplusplus\r
+extern "C" {\r
+#ifndef FORCEINLINE\r
+ #define FORCEINLINE inline\r
+#endif\r
+#endif /* __cplusplus */\r
+#ifndef FORCEINLINE\r
+ #define FORCEINLINE\r
+#endif\r
+\r
+#if !ONLY_MSPACES\r
+\r
+/* ------------------- Declarations of public routines ------------------- */\r
+\r
+#ifndef USE_DL_PREFIX\r
+#define dlcalloc calloc\r
+#define dlfree free\r
+#define dlmalloc malloc\r
+#define dlmemalign memalign\r
+#define dlrealloc realloc\r
+#define dlvalloc valloc\r
+#define dlpvalloc pvalloc\r
+#define dlmallinfo mallinfo\r
+#define dlmallopt mallopt\r
+#define dlmalloc_trim malloc_trim\r
+#define dlmalloc_stats malloc_stats\r
+#define dlmalloc_usable_size malloc_usable_size\r
+#define dlmalloc_footprint malloc_footprint\r
+#define dlmalloc_max_footprint malloc_max_footprint\r
+#define dlindependent_calloc independent_calloc\r
+#define dlindependent_comalloc independent_comalloc\r
+#endif /* USE_DL_PREFIX */\r
+\r
+\r
+/*\r
+ malloc(size_t n)\r
+ Returns a pointer to a newly allocated chunk of at least n bytes, or\r
+ null if no space is available, in which case errno is set to ENOMEM\r
+ on ANSI C systems.\r
+\r
+ If n is zero, malloc returns a minimum-sized chunk. (The minimum\r
+ size is 16 bytes on most 32bit systems, and 32 bytes on 64bit\r
+ systems.) Note that size_t is an unsigned type, so calls with\r
+ arguments that would be negative if signed are interpreted as\r
+ requests for huge amounts of space, which will often fail. The\r
+ maximum supported value of n differs across systems, but is in all\r
+ cases less than the maximum representable value of a size_t.\r
+*/\r
+void* dlmalloc(size_t);\r
+\r
+/*\r
+ free(void* p)\r
+ Releases the chunk of memory pointed to by p, that had been previously\r
+ allocated using malloc or a related routine such as realloc.\r
+ It has no effect if p is null. If p was not malloced or already\r
+ freed, free(p) will by default cause the current program to abort.\r
+*/\r
+void dlfree(void*);\r
+\r
+/*\r
+ calloc(size_t n_elements, size_t element_size);\r
+ Returns a pointer to n_elements * element_size bytes, with all locations\r
+ set to zero.\r
+*/\r
+void* dlcalloc(size_t, size_t);\r
+\r
+/*\r
+ realloc(void* p, size_t n)\r
+ Returns a pointer to a chunk of size n that contains the same data\r
+ as does chunk p up to the minimum of (n, p's size) bytes, or null\r
+ if no space is available.\r
+\r
+ The returned pointer may or may not be the same as p. The algorithm\r
+ prefers extending p in most cases when possible, otherwise it\r
+ employs the equivalent of a malloc-copy-free sequence.\r
+\r
+ If p is null, realloc is equivalent to malloc.\r
+\r
+ If space is not available, realloc returns null, errno is set (if on\r
+ ANSI) and p is NOT freed.\r
+\r
+ if n is for fewer bytes than already held by p, the newly unused\r
+ space is lopped off and freed if possible. realloc with a size\r
+ argument of zero (re)allocates a minimum-sized chunk.\r
+\r
+ The old unix realloc convention of allowing the last-free'd chunk\r
+ to be used as an argument to realloc is not supported.\r
+*/\r
+\r
+void* dlrealloc(void*, size_t);\r
+\r
+/*\r
+ memalign(size_t alignment, size_t n);\r
+ Returns a pointer to a newly allocated chunk of n bytes, aligned\r
+ in accord with the alignment argument.\r
+\r
+ The alignment argument should be a power of two. If the argument is\r
+ not a power of two, the nearest greater power is used.\r
+ 8-byte alignment is guaranteed by normal malloc calls, so don't\r
+ bother calling memalign with an argument of 8 or less.\r
+\r
+ Overreliance on memalign is a sure way to fragment space.\r
+*/\r
+void* dlmemalign(size_t, size_t);\r
+\r
+/*\r
+ valloc(size_t n);\r
+ Equivalent to memalign(pagesize, n), where pagesize is the page\r
+ size of the system. If the pagesize is unknown, 4096 is used.\r
+*/\r
+void* dlvalloc(size_t);\r
+\r
+/*\r
+ mallopt(int parameter_number, int parameter_value)\r
+ Sets tunable parameters The format is to provide a\r
+ (parameter-number, parameter-value) pair. mallopt then sets the\r
+ corresponding parameter to the argument value if it can (i.e., so\r
+ long as the value is meaningful), and returns 1 if successful else\r
+ 0. SVID/XPG/ANSI defines four standard param numbers for mallopt,\r
+ normally defined in malloc.h. None of these are use in this malloc,\r
+ so setting them has no effect. But this malloc also supports other\r
+ options in mallopt. See below for details. Briefly, supported\r
+ parameters are as follows (listed defaults are for "typical"\r
+ configurations).\r
+\r
+ Symbol param # default allowed param values\r
+ M_TRIM_THRESHOLD -1 2*1024*1024 any (MAX_SIZE_T disables)\r
+ M_GRANULARITY -2 page size any power of 2 >= page size\r
+ M_MMAP_THRESHOLD -3 256*1024 any (or 0 if no MMAP support)\r
+*/\r
+int dlmallopt(int, int);\r
+\r
+/*\r
+ malloc_footprint();\r
+ Returns the number of bytes obtained from the system. The total\r
+ number of bytes allocated by malloc, realloc etc., is less than this\r
+ value. Unlike mallinfo, this function returns only a precomputed\r
+ result, so can be called frequently to monitor memory consumption.\r
+ Even if locks are otherwise defined, this function does not use them,\r
+ so results might not be up to date.\r
+*/\r
+size_t dlmalloc_footprint(void);\r
+\r
+/*\r
+ malloc_max_footprint();\r
+ Returns the maximum number of bytes obtained from the system. This\r
+ value will be greater than current footprint if deallocated space\r
+ has been reclaimed by the system. The peak number of bytes allocated\r
+ by malloc, realloc etc., is less than this value. Unlike mallinfo,\r
+ this function returns only a precomputed result, so can be called\r
+ frequently to monitor memory consumption. Even if locks are\r
+ otherwise defined, this function does not use them, so results might\r
+ not be up to date.\r
+*/\r
+size_t dlmalloc_max_footprint(void);\r
+\r
+#if !NO_MALLINFO\r
+/*\r
+ mallinfo()\r
+ Returns (by copy) a struct containing various summary statistics:\r
+\r
+ arena: current total non-mmapped bytes allocated from system\r
+ ordblks: the number of free chunks\r
+ smblks: always zero.\r
+ hblks: current number of mmapped regions\r
+ hblkhd: total bytes held in mmapped regions\r
+ usmblks: the maximum total allocated space. This will be greater\r
+ than current total if trimming has occurred.\r
+ fsmblks: always zero\r
+ uordblks: current total allocated space (normal or mmapped)\r
+ fordblks: total free space\r
+ keepcost: the maximum number of bytes that could ideally be released\r
+ back to system via malloc_trim. ("ideally" means that\r
+ it ignores page restrictions etc.)\r
+\r
+ Because these fields are ints, but internal bookkeeping may\r
+ be kept as longs, the reported values may wrap around zero and\r
+ thus be inaccurate.\r
+*/\r
+struct mallinfo dlmallinfo(void);\r
+#endif /* NO_MALLINFO */\r
+\r
+/*\r
+ independent_calloc(size_t n_elements, size_t element_size, void* chunks[]);\r
+\r
+ independent_calloc is similar to calloc, but instead of returning a\r
+ single cleared space, it returns an array of pointers to n_elements\r
+ independent elements that can hold contents of size elem_size, each\r
+ of which starts out cleared, and can be independently freed,\r
+ realloc'ed etc. The elements are guaranteed to be adjacently\r
+ allocated (this is not guaranteed to occur with multiple callocs or\r
+ mallocs), which may also improve cache locality in some\r
+ applications.\r
+\r
+ The "chunks" argument is optional (i.e., may be null, which is\r
+ probably the most typical usage). If it is null, the returned array\r
+ is itself dynamically allocated and should also be freed when it is\r
+ no longer needed. Otherwise, the chunks array must be of at least\r
+ n_elements in length. It is filled in with the pointers to the\r
+ chunks.\r
+\r
+ In either case, independent_calloc returns this pointer array, or\r
+ null if the allocation failed. If n_elements is zero and "chunks"\r
+ is null, it returns a chunk representing an array with zero elements\r
+ (which should be freed if not wanted).\r
+\r
+ Each element must be individually freed when it is no longer\r
+ needed. If you'd like to instead be able to free all at once, you\r
+ should instead use regular calloc and assign pointers into this\r
+ space to represent elements. (In this case though, you cannot\r
+ independently free elements.)\r
+\r
+ independent_calloc simplifies and speeds up implementations of many\r
+ kinds of pools. It may also be useful when constructing large data\r
+ structures that initially have a fixed number of fixed-sized nodes,\r
+ but the number is not known at compile time, and some of the nodes\r
+ may later need to be freed. For example:\r
+\r
+ struct Node { int item; struct Node* next; };\r
+\r
+ struct Node* build_list() {\r
+ struct Node** pool;\r
+ int n = read_number_of_nodes_needed();\r
+ if (n <= 0) return 0;\r
+ pool = (struct Node**)(independent_calloc(n, sizeof(struct Node), 0);\r
+ if (pool == 0) die();\r
+ // organize into a linked list...\r
+ struct Node* first = pool[0];\r
+ for (i = 0; i < n-1; ++i)\r
+ pool[i]->next = pool[i+1];\r
+ free(pool); // Can now free the array (or not, if it is needed later)\r
+ return first;\r
+ }\r
+*/\r
+void** dlindependent_calloc(size_t, size_t, void**);\r
+\r
+/*\r
+ independent_comalloc(size_t n_elements, size_t sizes[], void* chunks[]);\r
+\r
+ independent_comalloc allocates, all at once, a set of n_elements\r
+ chunks with sizes indicated in the "sizes" array. It returns\r
+ an array of pointers to these elements, each of which can be\r
+ independently freed, realloc'ed etc. The elements are guaranteed to\r
+ be adjacently allocated (this is not guaranteed to occur with\r
+ multiple callocs or mallocs), which may also improve cache locality\r
+ in some applications.\r
+\r
+ The "chunks" argument is optional (i.e., may be null). If it is null\r
+ the returned array is itself dynamically allocated and should also\r
+ be freed when it is no longer needed. Otherwise, the chunks array\r
+ must be of at least n_elements in length. It is filled in with the\r
+ pointers to the chunks.\r
+\r
+ In either case, independent_comalloc returns this pointer array, or\r
+ null if the allocation failed. If n_elements is zero and chunks is\r
+ null, it returns a chunk representing an array with zero elements\r
+ (which should be freed if not wanted).\r
+\r
+ Each element must be individually freed when it is no longer\r
+ needed. If you'd like to instead be able to free all at once, you\r
+ should instead use a single regular malloc, and assign pointers at\r
+ particular offsets in the aggregate space. (In this case though, you\r
+ cannot independently free elements.)\r
+\r
+ independent_comallac differs from independent_calloc in that each\r
+ element may have a different size, and also that it does not\r
+ automatically clear elements.\r
+\r
+ independent_comalloc can be used to speed up allocation in cases\r
+ where several structs or objects must always be allocated at the\r
+ same time. For example:\r
+\r
+ struct Head { ... }\r
+ struct Foot { ... }\r
+\r
+ void send_message(char* msg) {\r
+ int msglen = strlen(msg);\r
+ size_t sizes[3] = { sizeof(struct Head), msglen, sizeof(struct Foot) };\r
+ void* chunks[3];\r
+ if (independent_comalloc(3, sizes, chunks) == 0)\r
+ die();\r
+ struct Head* head = (struct Head*)(chunks[0]);\r
+ char* body = (char*)(chunks[1]);\r
+ struct Foot* foot = (struct Foot*)(chunks[2]);\r
+ // ...\r
+ }\r
+\r
+ In general though, independent_comalloc is worth using only for\r
+ larger values of n_elements. For small values, you probably won't\r
+ detect enough difference from series of malloc calls to bother.\r
+\r
+ Overuse of independent_comalloc can increase overall memory usage,\r
+ since it cannot reuse existing noncontiguous small chunks that\r
+ might be available for some of the elements.\r
+*/\r
+void** dlindependent_comalloc(size_t, size_t*, void**);\r
+\r
+\r
+/*\r
+ pvalloc(size_t n);\r
+ Equivalent to valloc(minimum-page-that-holds(n)), that is,\r
+ round up n to nearest pagesize.\r
+ */\r
+void* dlpvalloc(size_t);\r
+\r
+/*\r
+ malloc_trim(size_t pad);\r
+\r
+ If possible, gives memory back to the system (via negative arguments\r
+ to sbrk) if there is unused memory at the `high' end of the malloc\r
+ pool or in unused MMAP segments. You can call this after freeing\r
+ large blocks of memory to potentially reduce the system-level memory\r
+ requirements of a program. However, it cannot guarantee to reduce\r
+ memory. Under some allocation patterns, some large free blocks of\r
+ memory will be locked between two used chunks, so they cannot be\r
+ given back to the system.\r
+\r
+ The `pad' argument to malloc_trim represents the amount of free\r
+ trailing space to leave untrimmed. If this argument is zero, only\r
+ the minimum amount of memory to maintain internal data structures\r
+ will be left. Non-zero arguments can be supplied to maintain enough\r
+ trailing space to service future expected allocations without having\r
+ to re-obtain memory from the system.\r
+\r
+ Malloc_trim returns 1 if it actually released any memory, else 0.\r
+*/\r
+int dlmalloc_trim(size_t);\r
+\r
+/*\r
+ malloc_usable_size(void* p);\r
+\r
+ Returns the number of bytes you can actually use in\r
+ an allocated chunk, which may be more than you requested (although\r
+ often not) due to alignment and minimum size constraints.\r
+ You can use this many bytes without worrying about\r
+ overwriting other allocated objects. This is not a particularly great\r
+ programming practice. malloc_usable_size can be more useful in\r
+ debugging and assertions, for example:\r
+\r
+ p = malloc(n);\r
+ assert(malloc_usable_size(p) >= 256);\r
+*/\r
+size_t dlmalloc_usable_size(void*);\r
+\r
+/*\r
+ malloc_stats();\r
+ Prints on stderr the amount of space obtained from the system (both\r
+ via sbrk and mmap), the maximum amount (which may be more than\r
+ current if malloc_trim and/or munmap got called), and the current\r
+ number of bytes allocated via malloc (or realloc, etc) but not yet\r
+ freed. Note that this is the number of bytes allocated, not the\r
+ number requested. It will be larger than the number requested\r
+ because of alignment and bookkeeping overhead. Because it includes\r
+ alignment wastage as being in use, this figure may be greater than\r
+ zero even when no user-level chunks are allocated.\r
+\r
+ The reported current and maximum system memory can be inaccurate if\r
+ a program makes other calls to system memory allocation functions\r
+ (normally sbrk) outside of malloc.\r
+\r
+ malloc_stats prints only the most commonly interesting statistics.\r
+ More information can be obtained by calling mallinfo.\r
+*/\r
+void dlmalloc_stats(void);\r
+\r
+#endif /* ONLY_MSPACES */\r
+\r
+#if MSPACES\r
+\r
+/*\r
+ mspace is an opaque type representing an independent\r
+ region of space that supports mspace_malloc, etc.\r
+*/\r
+typedef void* mspace;\r
+\r
+/*\r
+ create_mspace creates and returns a new independent space with the\r
+ given initial capacity, or, if 0, the default granularity size. It\r
+ returns null if there is no system memory available to create the\r
+ space. If argument locked is non-zero, the space uses a separate\r
+ lock to control access. The capacity of the space will grow\r
+ dynamically as needed to service mspace_malloc requests. You can\r
+ control the sizes of incremental increases of this space by\r
+ compiling with a different DEFAULT_GRANULARITY or dynamically\r
+ setting with mallopt(M_GRANULARITY, value).\r
+*/\r
+mspace create_mspace(size_t capacity, int locked);\r
+\r
+/*\r
+ destroy_mspace destroys the given space, and attempts to return all\r
+ of its memory back to the system, returning the total number of\r
+ bytes freed. After destruction, the results of access to all memory\r
+ used by the space become undefined.\r
+*/\r
+size_t destroy_mspace(mspace msp);\r
+\r
+/*\r
+ create_mspace_with_base uses the memory supplied as the initial base\r
+ of a new mspace. Part (less than 128*sizeof(size_t) bytes) of this\r
+ space is used for bookkeeping, so the capacity must be at least this\r
+ large. (Otherwise 0 is returned.) When this initial space is\r
+ exhausted, additional memory will be obtained from the system.\r
+ Destroying this space will deallocate all additionally allocated\r
+ space (if possible) but not the initial base.\r
+*/\r
+mspace create_mspace_with_base(void* base, size_t capacity, int locked);\r
+\r
+/*\r
+ mspace_malloc behaves as malloc, but operates within\r
+ the given space.\r
+*/\r
+void* mspace_malloc(mspace msp, size_t bytes);\r
+\r
+/*\r
+ mspace_free behaves as free, but operates within\r
+ the given space.\r
+\r
+ If compiled with FOOTERS==1, mspace_free is not actually needed.\r
+ free may be called instead of mspace_free because freed chunks from\r
+ any space are handled by their originating spaces.\r
+*/\r
+void mspace_free(mspace msp, void* mem);\r
+\r
+/*\r
+ mspace_realloc behaves as realloc, but operates within\r
+ the given space.\r
+\r
+ If compiled with FOOTERS==1, mspace_realloc is not actually\r
+ needed. realloc may be called instead of mspace_realloc because\r
+ realloced chunks from any space are handled by their originating\r
+ spaces.\r
+*/\r
+void* mspace_realloc(mspace msp, void* mem, size_t newsize);\r
+\r
+/*\r
+ mspace_calloc behaves as calloc, but operates within\r
+ the given space.\r
+*/\r
+void* mspace_calloc(mspace msp, size_t n_elements, size_t elem_size);\r
+\r
+/*\r
+ mspace_memalign behaves as memalign, but operates within\r
+ the given space.\r
+*/\r
+void* mspace_memalign(mspace msp, size_t alignment, size_t bytes);\r
+\r
+/*\r
+ mspace_independent_calloc behaves as independent_calloc, but\r
+ operates within the given space.\r
+*/\r
+void** mspace_independent_calloc(mspace msp, size_t n_elements,\r
+ size_t elem_size, void* chunks[]);\r
+\r
+/*\r
+ mspace_independent_comalloc behaves as independent_comalloc, but\r
+ operates within the given space.\r
+*/\r
+void** mspace_independent_comalloc(mspace msp, size_t n_elements,\r
+ size_t sizes[], void* chunks[]);\r
+\r
+/*\r
+ mspace_footprint() returns the number of bytes obtained from the\r
+ system for this space.\r
+*/\r
+size_t mspace_footprint(mspace msp);\r
+\r
+/*\r
+ mspace_max_footprint() returns the peak number of bytes obtained from the\r
+ system for this space.\r
+*/\r
+size_t mspace_max_footprint(mspace msp);\r
+\r
+\r
+#if !NO_MALLINFO\r
+/*\r
+ mspace_mallinfo behaves as mallinfo, but reports properties of\r
+ the given space.\r
+*/\r
+struct mallinfo mspace_mallinfo(mspace msp);\r
+#endif /* NO_MALLINFO */\r
+\r
+/*\r
+ mspace_malloc_stats behaves as malloc_stats, but reports\r
+ properties of the given space.\r
+*/\r
+void mspace_malloc_stats(mspace msp);\r
+\r
+/*\r
+ mspace_trim behaves as malloc_trim, but\r
+ operates within the given space.\r
+*/\r
+int mspace_trim(mspace msp, size_t pad);\r
+\r
+/*\r
+ An alias for mallopt.\r
+*/\r
+int mspace_mallopt(int, int);\r
+\r
+#endif /* MSPACES */\r
+\r
+#ifdef __cplusplus\r
+}; /* end of extern "C" */\r
+#endif /* __cplusplus */\r
+\r
+/*\r
+ ========================================================================\r
+ To make a fully customizable malloc.h header file, cut everything\r
+ above this line, put into file malloc.h, edit to suit, and #include it\r
+ on the next line, as well as in programs that use this malloc.\r
+ ========================================================================\r
+*/\r
+\r
+/* #include "malloc.h" */\r
+\r
+/*------------------------------ internal #includes ---------------------- */\r
+\r
+#ifdef WIN32\r
+#pragma warning( disable : 4146 ) /* no "unsigned" warnings */\r
+#endif /* WIN32 */\r
+\r
+#include <stdio.h> /* for printing in malloc_stats */\r
+\r
+#ifndef LACKS_ERRNO_H\r
+#include <errno.h> /* for MALLOC_FAILURE_ACTION */\r
+#endif /* LACKS_ERRNO_H */\r
+#if FOOTERS\r
+#include <time.h> /* for magic initialization */\r
+#endif /* FOOTERS */\r
+#ifndef LACKS_STDLIB_H\r
+#include <stdlib.h> /* for abort() */\r
+#endif /* LACKS_STDLIB_H */\r
+#ifdef DEBUG\r
+#if ABORT_ON_ASSERT_FAILURE\r
+#define assert(x) if(!(x)) ABORT\r
+#else /* ABORT_ON_ASSERT_FAILURE */\r
+#include <assert.h>\r
+#endif /* ABORT_ON_ASSERT_FAILURE */\r
+#else /* DEBUG */\r
+#define assert(x)\r
+#endif /* DEBUG */\r
+#ifndef LACKS_STRING_H\r
+#include <string.h> /* for memset etc */\r
+#endif /* LACKS_STRING_H */\r
+#if USE_BUILTIN_FFS\r
+#ifndef LACKS_STRINGS_H\r
+#include <strings.h> /* for ffs */\r
+#endif /* LACKS_STRINGS_H */\r
+#endif /* USE_BUILTIN_FFS */\r
+#if HAVE_MMAP\r
+#ifndef LACKS_SYS_MMAN_H\r
+#include <sys/mman.h> /* for mmap */\r
+#endif /* LACKS_SYS_MMAN_H */\r
+#ifndef LACKS_FCNTL_H\r
+#include <fcntl.h>\r
+#endif /* LACKS_FCNTL_H */\r
+#endif /* HAVE_MMAP */\r
+#if HAVE_MORECORE\r
+#ifndef LACKS_UNISTD_H\r
+#include <unistd.h> /* for sbrk */\r
+#else /* LACKS_UNISTD_H */\r
+#if !defined(__FreeBSD__) && !defined(__OpenBSD__) && !defined(__NetBSD__)\r
+extern void* sbrk(ptrdiff_t);\r
+#endif /* FreeBSD etc */\r
+#endif /* LACKS_UNISTD_H */\r
+#endif /* HAVE_MMAP */\r
+\r
+#ifndef WIN32\r
+#ifndef malloc_getpagesize\r
+# ifdef _SC_PAGESIZE /* some SVR4 systems omit an underscore */\r
+# ifndef _SC_PAGE_SIZE\r
+# define _SC_PAGE_SIZE _SC_PAGESIZE\r
+# endif\r
+# endif\r
+# ifdef _SC_PAGE_SIZE\r
+# define malloc_getpagesize sysconf(_SC_PAGE_SIZE)\r
+# else\r
+# if defined(BSD) || defined(DGUX) || defined(HAVE_GETPAGESIZE)\r
+ extern size_t getpagesize();\r
+# define malloc_getpagesize getpagesize()\r
+# else\r
+# ifdef WIN32 /* use supplied emulation of getpagesize */\r
+# define malloc_getpagesize getpagesize()\r
+# else\r
+# ifndef LACKS_SYS_PARAM_H\r
+# include <sys/param.h>\r
+# endif\r
+# ifdef EXEC_PAGESIZE\r
+# define malloc_getpagesize EXEC_PAGESIZE\r
+# else\r
+# ifdef NBPG\r
+# ifndef CLSIZE\r
+# define malloc_getpagesize NBPG\r
+# else\r
+# define malloc_getpagesize (NBPG * CLSIZE)\r
+# endif\r
+# else\r
+# ifdef NBPC\r
+# define malloc_getpagesize NBPC\r
+# else\r
+# ifdef PAGESIZE\r
+# define malloc_getpagesize PAGESIZE\r
+# else /* just guess */\r
+# define malloc_getpagesize ((size_t)4096U)\r
+# endif\r
+# endif\r
+# endif\r
+# endif\r
+# endif\r
+# endif\r
+# endif\r
+#endif\r
+#endif\r
+\r
+/* ------------------- size_t and alignment properties -------------------- */\r
+\r
+/* The byte and bit size of a size_t */\r
+#define SIZE_T_SIZE (sizeof(size_t))\r
+#define SIZE_T_BITSIZE (sizeof(size_t) << 3)\r
+\r
+/* Some constants coerced to size_t */\r
+/* Annoying but necessary to avoid errors on some plaftorms */\r
+#define SIZE_T_ZERO ((size_t)0)\r
+#define SIZE_T_ONE ((size_t)1)\r
+#define SIZE_T_TWO ((size_t)2)\r
+#define TWO_SIZE_T_SIZES (SIZE_T_SIZE<<1)\r
+#define FOUR_SIZE_T_SIZES (SIZE_T_SIZE<<2)\r
+#define SIX_SIZE_T_SIZES (FOUR_SIZE_T_SIZES+TWO_SIZE_T_SIZES)\r
+#define HALF_MAX_SIZE_T (MAX_SIZE_T / 2U)\r
+\r
+/* The bit mask value corresponding to MALLOC_ALIGNMENT */\r
+#define CHUNK_ALIGN_MASK (MALLOC_ALIGNMENT - SIZE_T_ONE)\r
+\r
+/* True if address a has acceptable alignment */\r
+#define is_aligned(A) (((size_t)((A)) & (CHUNK_ALIGN_MASK)) == 0)\r
+\r
+/* the number of bytes to offset an address to align it */\r
+#define align_offset(A)\\r
+ ((((size_t)(A) & CHUNK_ALIGN_MASK) == 0)? 0 :\\r
+ ((MALLOC_ALIGNMENT - ((size_t)(A) & CHUNK_ALIGN_MASK)) & CHUNK_ALIGN_MASK))\r
+\r
+/* -------------------------- MMAP preliminaries ------------------------- */\r
+\r
+/*\r
+ If HAVE_MORECORE or HAVE_MMAP are false, we just define calls and\r
+ checks to fail so compiler optimizer can delete code rather than\r
+ using so many "#if"s.\r
+*/\r
+\r
+\r
+/* MORECORE and MMAP must return MFAIL on failure */\r
+#define MFAIL ((void*)(MAX_SIZE_T))\r
+#define CMFAIL ((char*)(MFAIL)) /* defined for convenience */\r
+\r
+#if !HAVE_MMAP\r
+#define IS_MMAPPED_BIT (SIZE_T_ZERO)\r
+#define USE_MMAP_BIT (SIZE_T_ZERO)\r
+#define CALL_MMAP(s) MFAIL\r
+#define CALL_MUNMAP(a, s) (-1)\r
+#define DIRECT_MMAP(s) MFAIL\r
+\r
+#else /* HAVE_MMAP */\r
+#define IS_MMAPPED_BIT (SIZE_T_ONE)\r
+#define USE_MMAP_BIT (SIZE_T_ONE)\r
+\r
+#ifndef WIN32\r
+#define CALL_MUNMAP(a, s) munmap((a), (s))\r
+#define MMAP_PROT (PROT_READ|PROT_WRITE)\r
+#if !defined(MAP_ANONYMOUS) && defined(MAP_ANON)\r
+#define MAP_ANONYMOUS MAP_ANON\r
+#endif /* MAP_ANON */\r
+#ifdef MAP_ANONYMOUS\r
+#define MMAP_FLAGS (MAP_PRIVATE|MAP_ANONYMOUS)\r
+#define CALL_MMAP(s) mmap(0, (s), MMAP_PROT, MMAP_FLAGS, -1, 0)\r
+#else /* MAP_ANONYMOUS */\r
+/*\r
+ Nearly all versions of mmap support MAP_ANONYMOUS, so the following\r
+ is unlikely to be needed, but is supplied just in case.\r
+*/\r
+#define MMAP_FLAGS (MAP_PRIVATE)\r
+static int dev_zero_fd = -1; /* Cached file descriptor for /dev/zero. */\r
+#define CALL_MMAP(s) ((dev_zero_fd < 0) ? \\r
+ (dev_zero_fd = open("/dev/zero", O_RDWR), \\r
+ mmap(0, (s), MMAP_PROT, MMAP_FLAGS, dev_zero_fd, 0)) : \\r
+ mmap(0, (s), MMAP_PROT, MMAP_FLAGS, dev_zero_fd, 0))\r
+#endif /* MAP_ANONYMOUS */\r
+\r
+#define DIRECT_MMAP(s) CALL_MMAP(s)\r
+#else /* WIN32 */\r
+\r
+/* Win32 MMAP via VirtualAlloc */\r
+static FORCEINLINE void* win32mmap(size_t size) {\r
+ void* ptr = VirtualAlloc(0, size, MEM_RESERVE|MEM_COMMIT, PAGE_READWRITE);\r
+ return (ptr != 0)? ptr: MFAIL;\r
+}\r
+\r
+/* For direct MMAP, use MEM_TOP_DOWN to minimize interference */\r
+static FORCEINLINE void* win32direct_mmap(size_t size) {\r
+ void* ptr = VirtualAlloc(0, size, MEM_RESERVE|MEM_COMMIT|MEM_TOP_DOWN,\r
+ PAGE_READWRITE);\r
+ return (ptr != 0)? ptr: MFAIL;\r
+}\r
+\r
+/* This function supports releasing coalesed segments */\r
+static FORCEINLINE int win32munmap(void* ptr, size_t size) {\r
+ MEMORY_BASIC_INFORMATION minfo;\r
+ char* cptr = (char *) ptr;\r
+ while (size) {\r
+ if (VirtualQuery(cptr, &minfo, sizeof(minfo)) == 0)\r
+ return -1;\r
+ if (minfo.BaseAddress != cptr || minfo.AllocationBase != cptr ||\r
+ minfo.State != MEM_COMMIT || minfo.RegionSize > size)\r
+ return -1;\r
+ if (VirtualFree(cptr, 0, MEM_RELEASE) == 0)\r
+ return -1;\r
+ cptr += minfo.RegionSize;\r
+ size -= minfo.RegionSize;\r
+ }\r
+ return 0;\r
+}\r
+\r
+#define CALL_MMAP(s) win32mmap(s)\r
+#define CALL_MUNMAP(a, s) win32munmap((a), (s))\r
+#define DIRECT_MMAP(s) win32direct_mmap(s)\r
+#endif /* WIN32 */\r
+#endif /* HAVE_MMAP */\r
+\r
+#if HAVE_MMAP && HAVE_MREMAP\r
+#define CALL_MREMAP(addr, osz, nsz, mv) mremap((addr), (osz), (nsz), (mv))\r
+#else /* HAVE_MMAP && HAVE_MREMAP */\r
+#define CALL_MREMAP(addr, osz, nsz, mv) ((void)(addr),(void)(osz), \\r
+ (void)(nsz), (void)(mv),MFAIL)\r
+#endif /* HAVE_MMAP && HAVE_MREMAP */\r
+\r
+#if HAVE_MORECORE\r
+#define CALL_MORECORE(S) MORECORE(S)\r
+#else /* HAVE_MORECORE */\r
+#define CALL_MORECORE(S) MFAIL\r
+#endif /* HAVE_MORECORE */\r
+\r
+/* mstate bit set if continguous morecore disabled or failed */\r
+#define USE_NONCONTIGUOUS_BIT (4U)\r
+\r
+/* segment bit set in create_mspace_with_base */\r
+#define EXTERN_BIT (8U)\r
+\r
+\r
+/* --------------------------- Lock preliminaries ------------------------ */\r
+\r
+#if USE_LOCKS\r
+\r
+#if USE_LOCKS==1\r
+\r
+/*\r
+ When locks are defined, there are up to two global locks:\r
+\r
+ * If HAVE_MORECORE, morecore_mutex protects sequences of calls to\r
+ MORECORE. In many cases sys_alloc requires two calls, that should\r
+ not be interleaved with calls by other threads. This does not\r
+ protect against direct calls to MORECORE by other threads not\r
+ using this lock, so there is still code to cope the best we can on\r
+ interference.\r
+\r
+ * magic_init_mutex ensures that mparams.magic and other\r
+ unique mparams values are initialized only once.\r
+*/\r
+\r
+#ifndef WIN32\r
+/* By default use posix locks */\r
+#include <pthread.h>\r
+\r
+#if defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__))\r
+/* Go for assembler spin locks on x86 and x64 */\r
+struct pthread_mlock_t\r
+{\r
+ volatile pthread_t threadid;\r
+ volatile unsigned int c;\r
+ volatile unsigned int l;\r
+};\r
+#define MLOCK_T struct pthread_mlock_t\r
+#define CURRENT_THREAD pthread_self()\r
+static FORCEINLINE int pthread_acquire_lock (MLOCK_T *sl) {\r
+ if(CURRENT_THREAD==sl->threadid)\r
+ ++sl->c;\r
+ else for (;;) {\r
+ int ret;\r
+ __asm__ __volatile__ ("pause\n\tlock cmpxchgl %2,(%1)" : "=a" (ret) : "r" (&sl->l), "r" (1), "a" (0));\r
+ if(!ret) {\r
+ assert(!sl->threadid);\r
+ sl->threadid=CURRENT_THREAD;\r
+ sl->c=1;\r
+ break;\r
+ }\r
+ sched_yield();\r
+ }\r
+\r
+ return 0;\r
+}\r
+\r
+static FORCEINLINE void pthread_release_lock (MLOCK_T *sl) {\r
+ int ret;\r
+ assert(CURRENT_THREAD==sl->threadid);\r
+ if (!--sl->c) {\r
+ sl->threadid=0;\r
+ __asm__ __volatile__ ("xchgl %2,(%1)" : "=r" (ret) : "r" (&sl->l), "0" (0));\r
+ }\r
+}\r
+\r
+static FORCEINLINE int pthread_try_lock (MLOCK_T *sl) {\r
+ int ret;\r
+ __asm__ __volatile__ ("pause\n\tlock cmpxchgl %2,(%1)" : "=a" (ret) : "r" (&sl->l), "r" (1), "a" (0));\r
+ if(!ret){\r
+ assert(!sl->threadid);\r
+ sl->threadid=CURRENT_THREAD;\r
+ sl->c=1;\r
+ return 1;\r
+ }\r
+ return 0;\r
+}\r
+\r
+#define INITIAL_LOCK(sl) (memset((sl), 0, sizeof(MLOCK_T)), 0)\r
+#define ACQUIRE_LOCK(sl) pthread_acquire_lock(sl)\r
+#define RELEASE_LOCK(sl) pthread_release_lock(sl)\r
+#define TRY_LOCK(sl) pthread_try_lock(sl)\r
+#define IS_LOCKED(sl) ((sl)->l)\r
+\r
+#if HAVE_MORECORE\r
+static MLOCK_T morecore_mutex = {0, 0 };\r
+#endif /* HAVE_MORECORE */\r
+\r
+static MLOCK_T magic_init_mutex = {0, 0 };\r
+\r
+#else\r
+\r
+/* The POSIX threads implementation */\r
+struct pthread_mlock_t\r
+{\r
+ volatile unsigned int c;\r
+ pthread_mutex_t l;\r
+};\r
+#define MLOCK_T struct pthread_mlock_t\r
+#define CURRENT_THREAD pthread_self()\r
+static FORCEINLINE int pthread_acquire_lock (MLOCK_T *sl) {\r
+ if(!pthread_mutex_lock(&(sl)->l)){\r
+ sl->c++;\r
+ return 0;\r
+ }\r
+ return 1;\r
+}\r
+\r
+static FORCEINLINE void pthread_release_lock (MLOCK_T *sl) {\r
+ --sl->c;\r
+ pthread_mutex_unlock(&(sl)->l);\r
+}\r
+\r
+static FORCEINLINE int pthread_try_lock (MLOCK_T *sl) {\r
+ if(!pthread_mutex_trylock(&(sl)->l)){\r
+ sl->c++;\r
+ return 1;\r
+ }\r
+ return 0;\r
+}\r
+\r
+static FORCEINLINE int pthread_init_lock (MLOCK_T *sl) {\r
+ pthread_mutexattr_t attr;\r
+ sl->c=0;\r
+ if(pthread_mutexattr_init(&attr)) return 1;\r
+ if(pthread_mutexattr_settype(&attr, PTHREAD_MUTEX_RECURSIVE)) return 1;\r
+ if(pthread_mutex_init(&sl->l, &attr)) return 1;\r
+ pthread_mutexattr_destroy(&attr);\r
+ return 0;\r
+}\r
+\r
+static FORCEINLINE int pthread_islocked (MLOCK_T *sl) {\r
+#if 0 /* This will be faster on SMP systems which enforce cache consistency */\r
+ return sl->c!=0;\r
+#else\r
+ /* Doing this correctly portably means inefficient code :( */\r
+ if(!pthread_try_lock(sl)){\r
+ int ret=(sl->c!=0);\r
+ pthread_mutex_unlock(sl);\r
+ return ret;\r
+ }\r
+ return 0;\r
+#endif\r
+}\r
+\r
+\r
+#define INITIAL_LOCK(sl) pthread_init_lock(sl)\r
+#define ACQUIRE_LOCK(sl) pthread_acquire_lock(sl)\r
+#define RELEASE_LOCK(sl) pthread_release_lock(sl)\r
+#define TRY_LOCK(sl) pthread_try_lock(sl)\r
+#define IS_LOCKED(sl) pthread_islocked(sl)\r
+\r
+#if HAVE_MORECORE\r
+static MLOCK_T morecore_mutex = {0, PTHREAD_MUTEX_INITIALIZER };\r
+#endif /* HAVE_MORECORE */\r
+\r
+static MLOCK_T magic_init_mutex = {0, PTHREAD_MUTEX_INITIALIZER };\r
+#endif /* __GNUC__ && i386 */\r
+\r
+#else /* WIN32 */\r
+/*\r
+ Because lock-protected regions have bounded times, and there\r
+ are no recursive lock calls, we can use simple spinlocks.\r
+*/\r
+\r
+#if defined(_MSC_VER) && _MSC_VER>=1310\r
+#if 1\r
+ #ifndef _M_AMD64\r
+ // These are already defined on AMD64 builds\r
+ #ifdef __cplusplus\r
+extern "C" {\r
+ #endif\r
+ LONG __cdecl _InterlockedCompareExchange(LPLONG volatile Dest, LONG Exchange, LONG Comp);\r
+ LONG __cdecl _InterlockedExchange(LPLONG volatile Target, LONG Value);\r
+ #ifdef __cplusplus\r
+}\r
+ #endif\r
+ #endif\r
+ // MSVC7.1 Intrinsics are far faster than win32 functions\r
+ #pragma intrinsic (_InterlockedCompareExchange)\r
+ #pragma intrinsic (_InterlockedExchange)\r
+ #define interlockedcompareexchange _InterlockedCompareExchange\r
+ #define interlockedexchange _InterlockedExchange\r
+#else\r
+ #define interlockedcompareexchange InterlockedCompareExchange\r
+ #define interlockedexchange InterlockedExchange\r
+#endif\r
+\r
+struct win32_mlock_t\r
+{\r
+ volatile long threadid;\r
+ volatile unsigned int c;\r
+ long l;\r
+};\r
+#define MLOCK_T struct win32_mlock_t\r
+#define CURRENT_THREAD GetCurrentThreadId()\r
+static FORCEINLINE int win32_acquire_lock (MLOCK_T *sl) {\r
+ long mythreadid=CURRENT_THREAD;\r
+ if(mythreadid==sl->threadid)\r
+ ++sl->c;\r
+ else for (;;) {\r
+ if (!interlockedexchange(&sl->l, 1)) {\r
+ assert(!sl->threadid);\r
+ sl->threadid=mythreadid;\r
+ sl->c=1;\r
+ break;\r
+ }\r
+ /*YieldProcessor();*/\r
+ }\r
+\r
+ return 0;\r
+}\r
+\r
+static FORCEINLINE void win32_release_lock (MLOCK_T *sl) {\r
+ assert(CURRENT_THREAD==sl->threadid);\r
+ if (!--sl->c) {\r
+ sl->threadid=0;\r
+ interlockedexchange (&sl->l, 0);\r
+ }\r
+}\r
+\r
+static FORCEINLINE int win32_try_lock (MLOCK_T *sl) {\r
+ if (!interlockedexchange(&sl->l, 1)){\r
+ assert(!sl->threadid);\r
+ sl->threadid=CURRENT_THREAD;\r
+ sl->c=1;\r
+ return 1;\r
+ }\r
+ return 0;\r
+}\r
+\r
+\r
+#define INITIAL_LOCK(sl) (memset(sl, 0, sizeof(MLOCK_T)), 0)\r
+#define ACQUIRE_LOCK(sl) win32_acquire_lock(sl)\r
+#define RELEASE_LOCK(sl) win32_release_lock(sl)\r
+#define TRY_LOCK(sl) win32_try_lock(sl)\r
+#define IS_LOCKED(sl) ((sl)->l)\r
+#else\r
+/* Critical section implementation */\r
+#define MLOCK_T CRITICAL_SECTION\r
+#define CURRENT_THREAD GetCurrentThreadId()\r
+#define INITIAL_LOCK(s) (!InitializeCriticalSectionAndSpinCount((s), 4000)\r
+#define ACQUIRE_LOCK(s) ( (!((s))->DebugInfo ? INITIAL_LOCK((s)) : 0), !EnterCriticalSection((s)), 0)\r
+#define RELEASE_LOCK(s) ( LeaveCriticalSection((s)), 0 )\r
+#define TRY_LOCK(s) ( TryEnterCriticalSection((s)) )\r
+#define IS_LOCKED(s) ( (s)->LockCount >= 0 )\r
+#define NULL_LOCK_INITIALIZER\r
+\r
+#endif /* 1 */\r
+#if HAVE_MORECORE\r
+static MLOCK_T morecore_mutex;\r
+#endif /* HAVE_MORECORE */\r
+static MLOCK_T magic_init_mutex;\r
+#endif /* WIN32 */\r
+\r
+#else\r
+/* User defines their own locks */\r
+#if HAVE_MORECORE\r
+static MLOCK_T morecore_mutex NULL_LOCK_INITIALIZER;\r
+#endif /* HAVE_MORECORE */\r
+static MLOCK_T magic_init_mutex NULL_LOCK_INITIALIZER;\r
+\r
+#endif /* USE_LOCKS==1 */\r
+\r
+#define USE_LOCK_BIT (2U)\r
+#else /* USE_LOCKS */\r
+#define USE_LOCK_BIT (0U)\r
+#define INITIAL_LOCK(l)\r
+#endif /* USE_LOCKS */\r
+\r
+#if USE_LOCKS && HAVE_MORECORE\r
+#define ACQUIRE_MORECORE_LOCK() ACQUIRE_LOCK(&morecore_mutex);\r
+#define RELEASE_MORECORE_LOCK() RELEASE_LOCK(&morecore_mutex);\r
+#else /* USE_LOCKS && HAVE_MORECORE */\r
+#define ACQUIRE_MORECORE_LOCK()\r
+#define RELEASE_MORECORE_LOCK()\r
+#endif /* USE_LOCKS && HAVE_MORECORE */\r
+\r
+#if USE_LOCKS\r
+#define ACQUIRE_MAGIC_INIT_LOCK() ACQUIRE_LOCK(&magic_init_mutex);\r
+#define RELEASE_MAGIC_INIT_LOCK() RELEASE_LOCK(&magic_init_mutex);\r
+#else /* USE_LOCKS */\r
+#define ACQUIRE_MAGIC_INIT_LOCK()\r
+#define RELEASE_MAGIC_INIT_LOCK()\r
+#endif /* USE_LOCKS */\r
+\r
+\r
+/* ----------------------- Chunk representations ------------------------ */\r
+\r
+/*\r
+ (The following includes lightly edited explanations by Colin Plumb.)\r
+\r
+ The malloc_chunk declaration below is misleading (but accurate and\r
+ necessary). It declares a "view" into memory allowing access to\r
+ necessary fields at known offsets from a given base.\r
+\r
+ Chunks of memory are maintained using a `boundary tag' method as\r
+ originally described by Knuth. (See the paper by Paul Wilson\r
+ ftp://ftp.cs.utexas.edu/pub/garbage/allocsrv.ps for a survey of such\r
+ techniques.) Sizes of free chunks are stored both in the front of\r
+ each chunk and at the end. This makes consolidating fragmented\r
+ chunks into bigger chunks fast. The head fields also hold bits\r
+ representing whether chunks are free or in use.\r
+\r
+ Here are some pictures to make it clearer. They are "exploded" to\r
+ show that the state of a chunk can be thought of as extending from\r
+ the high 31 bits of the head field of its header through the\r
+ prev_foot and PINUSE_BIT bit of the following chunk header.\r
+\r
+ A chunk that's in use looks like:\r
+\r
+ chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ | Size of previous chunk (if P = 1) |\r
+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |P|\r
+ | Size of this chunk 1| +-+\r
+ mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ | |\r
+ +- -+\r
+ | |\r
+ +- -+\r
+ | :\r
+ +- size - sizeof(size_t) available payload bytes -+\r
+ : |\r
+ chunk-> +- -+\r
+ | |\r
+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |1|\r
+ | Size of next chunk (may or may not be in use) | +-+\r
+ mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+\r
+ And if it's free, it looks like this:\r
+\r
+ chunk-> +- -+\r
+ | User payload (must be in use, or we would have merged!) |\r
+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |P|\r
+ | Size of this chunk 0| +-+\r
+ mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ | Next pointer |\r
+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ | Prev pointer |\r
+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ | :\r
+ +- size - sizeof(struct chunk) unused bytes -+\r
+ : |\r
+ chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ | Size of this chunk |\r
+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |0|\r
+ | Size of next chunk (must be in use, or we would have merged)| +-+\r
+ mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ | :\r
+ +- User payload -+\r
+ : |\r
+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ |0|\r
+ +-+\r
+ Note that since we always merge adjacent free chunks, the chunks\r
+ adjacent to a free chunk must be in use.\r
+\r
+ Given a pointer to a chunk (which can be derived trivially from the\r
+ payload pointer) we can, in O(1) time, find out whether the adjacent\r
+ chunks are free, and if so, unlink them from the lists that they\r
+ are on and merge them with the current chunk.\r
+\r
+ Chunks always begin on even word boundaries, so the mem portion\r
+ (which is returned to the user) is also on an even word boundary, and\r
+ thus at least double-word aligned.\r
+\r
+ The P (PINUSE_BIT) bit, stored in the unused low-order bit of the\r
+ chunk size (which is always a multiple of two words), is an in-use\r
+ bit for the *previous* chunk. If that bit is *clear*, then the\r
+ word before the current chunk size contains the previous chunk\r
+ size, and can be used to find the front of the previous chunk.\r
+ The very first chunk allocated always has this bit set, preventing\r
+ access to non-existent (or non-owned) memory. If pinuse is set for\r
+ any given chunk, then you CANNOT determine the size of the\r
+ previous chunk, and might even get a memory addressing fault when\r
+ trying to do so.\r
+\r
+ The C (CINUSE_BIT) bit, stored in the unused second-lowest bit of\r
+ the chunk size redundantly records whether the current chunk is\r
+ inuse. This redundancy enables usage checks within free and realloc,\r
+ and reduces indirection when freeing and consolidating chunks.\r
+\r
+ Each freshly allocated chunk must have both cinuse and pinuse set.\r
+ That is, each allocated chunk borders either a previously allocated\r
+ and still in-use chunk, or the base of its memory arena. This is\r
+ ensured by making all allocations from the the `lowest' part of any\r
+ found chunk. Further, no free chunk physically borders another one,\r
+ so each free chunk is known to be preceded and followed by either\r
+ inuse chunks or the ends of memory.\r
+\r
+ Note that the `foot' of the current chunk is actually represented\r
+ as the prev_foot of the NEXT chunk. This makes it easier to\r
+ deal with alignments etc but can be very confusing when trying\r
+ to extend or adapt this code.\r
+\r
+ The exceptions to all this are\r
+\r
+ 1. The special chunk `top' is the top-most available chunk (i.e.,\r
+ the one bordering the end of available memory). It is treated\r
+ specially. Top is never included in any bin, is used only if\r
+ no other chunk is available, and is released back to the\r
+ system if it is very large (see M_TRIM_THRESHOLD). In effect,\r
+ the top chunk is treated as larger (and thus less well\r
+ fitting) than any other available chunk. The top chunk\r
+ doesn't update its trailing size field since there is no next\r
+ contiguous chunk that would have to index off it. However,\r
+ space is still allocated for it (TOP_FOOT_SIZE) to enable\r
+ separation or merging when space is extended.\r
+\r
+ 3. Chunks allocated via mmap, which have the lowest-order bit\r
+ (IS_MMAPPED_BIT) set in their prev_foot fields, and do not set\r
+ PINUSE_BIT in their head fields. Because they are allocated\r
+ one-by-one, each must carry its own prev_foot field, which is\r
+ also used to hold the offset this chunk has within its mmapped\r
+ region, which is needed to preserve alignment. Each mmapped\r
+ chunk is trailed by the first two fields of a fake next-chunk\r
+ for sake of usage checks.\r
+\r
+*/\r
+\r
+struct malloc_chunk {\r
+ size_t prev_foot; /* Size of previous chunk (if free). */\r
+ size_t head; /* Size and inuse bits. */\r
+ struct malloc_chunk* fd; /* double links -- used only if free. */\r
+ struct malloc_chunk* bk;\r
+};\r
+\r
+typedef struct malloc_chunk mchunk;\r
+typedef struct malloc_chunk* mchunkptr;\r
+typedef struct malloc_chunk* sbinptr; /* The type of bins of chunks */\r
+typedef unsigned int bindex_t; /* Described below */\r
+typedef unsigned int binmap_t; /* Described below */\r
+typedef unsigned int flag_t; /* The type of various bit flag sets */\r
+\r
+/* ------------------- Chunks sizes and alignments ----------------------- */\r
+\r
+#define MCHUNK_SIZE (sizeof(mchunk))\r
+\r
+#if FOOTERS\r
+#define CHUNK_OVERHEAD (TWO_SIZE_T_SIZES)\r
+#else /* FOOTERS */\r
+#define CHUNK_OVERHEAD (SIZE_T_SIZE)\r
+#endif /* FOOTERS */\r
+\r
+/* MMapped chunks need a second word of overhead ... */\r
+#define MMAP_CHUNK_OVERHEAD (TWO_SIZE_T_SIZES)\r
+/* ... and additional padding for fake next-chunk at foot */\r
+#define MMAP_FOOT_PAD (FOUR_SIZE_T_SIZES)\r
+\r
+/* The smallest size we can malloc is an aligned minimal chunk */\r
+#define MIN_CHUNK_SIZE\\r
+ ((MCHUNK_SIZE + CHUNK_ALIGN_MASK) & ~CHUNK_ALIGN_MASK)\r
+\r
+/* conversion from malloc headers to user pointers, and back */\r
+#define chunk2mem(p) ((void*)((char*)(p) + TWO_SIZE_T_SIZES))\r
+#define mem2chunk(mem) ((mchunkptr)((char*)(mem) - TWO_SIZE_T_SIZES))\r
+/* chunk associated with aligned address A */\r
+#define align_as_chunk(A) (mchunkptr)((A) + align_offset(chunk2mem(A)))\r
+\r
+/* Bounds on request (not chunk) sizes. */\r
+#define MAX_REQUEST ((-MIN_CHUNK_SIZE) << 2)\r
+#define MIN_REQUEST (MIN_CHUNK_SIZE - CHUNK_OVERHEAD - SIZE_T_ONE)\r
+\r
+/* pad request bytes into a usable size */\r
+#define pad_request(req) \\r
+ (((req) + CHUNK_OVERHEAD + CHUNK_ALIGN_MASK) & ~CHUNK_ALIGN_MASK)\r
+\r
+/* pad request, checking for minimum (but not maximum) */\r
+#define request2size(req) \\r
+ (((req) < MIN_REQUEST)? MIN_CHUNK_SIZE : pad_request(req))\r
+\r
+\r
+/* ------------------ Operations on head and foot fields ----------------- */\r
+\r
+/*\r
+ The head field of a chunk is or'ed with PINUSE_BIT when previous\r
+ adjacent chunk in use, and or'ed with CINUSE_BIT if this chunk is in\r
+ use. If the chunk was obtained with mmap, the prev_foot field has\r
+ IS_MMAPPED_BIT set, otherwise holding the offset of the base of the\r
+ mmapped region to the base of the chunk.\r
+*/\r
+\r
+#define PINUSE_BIT (SIZE_T_ONE)\r
+#define CINUSE_BIT (SIZE_T_TWO)\r
+#define INUSE_BITS (PINUSE_BIT|CINUSE_BIT)\r
+\r
+/* Head value for fenceposts */\r
+#define FENCEPOST_HEAD (INUSE_BITS|SIZE_T_SIZE)\r
+\r
+/* extraction of fields from head words */\r
+#define cinuse(p) ((p)->head & CINUSE_BIT)\r
+#define pinuse(p) ((p)->head & PINUSE_BIT)\r
+#define chunksize(p) ((p)->head & ~(INUSE_BITS))\r
+\r
+#define clear_pinuse(p) ((p)->head &= ~PINUSE_BIT)\r
+#define clear_cinuse(p) ((p)->head &= ~CINUSE_BIT)\r
+\r
+/* Treat space at ptr +/- offset as a chunk */\r
+#define chunk_plus_offset(p, s) ((mchunkptr)(((char*)(p)) + (s)))\r
+#define chunk_minus_offset(p, s) ((mchunkptr)(((char*)(p)) - (s)))\r
+\r
+/* Ptr to next or previous physical malloc_chunk. */\r
+#define next_chunk(p) ((mchunkptr)( ((char*)(p)) + ((p)->head & ~INUSE_BITS)))\r
+#define prev_chunk(p) ((mchunkptr)( ((char*)(p)) - ((p)->prev_foot) ))\r
+\r
+/* extract next chunk's pinuse bit */\r
+#define next_pinuse(p) ((next_chunk(p)->head) & PINUSE_BIT)\r
+\r
+/* Get/set size at footer */\r
+#define get_foot(p, s) (((mchunkptr)((char*)(p) + (s)))->prev_foot)\r
+#define set_foot(p, s) (((mchunkptr)((char*)(p) + (s)))->prev_foot = (s))\r
+\r
+/* Set size, pinuse bit, and foot */\r
+#define set_size_and_pinuse_of_free_chunk(p, s)\\r
+ ((p)->head = (s|PINUSE_BIT), set_foot(p, s))\r
+\r
+/* Set size, pinuse bit, foot, and clear next pinuse */\r
+#define set_free_with_pinuse(p, s, n)\\r
+ (clear_pinuse(n), set_size_and_pinuse_of_free_chunk(p, s))\r
+\r
+#define is_mmapped(p)\\r
+ (!((p)->head & PINUSE_BIT) && ((p)->prev_foot & IS_MMAPPED_BIT))\r
+\r
+/* Get the internal overhead associated with chunk p */\r
+#define overhead_for(p)\\r
+ (is_mmapped(p)? MMAP_CHUNK_OVERHEAD : CHUNK_OVERHEAD)\r
+\r
+/* Return true if malloced space is not necessarily cleared */\r
+#if MMAP_CLEARS\r
+#define calloc_must_clear(p) (!is_mmapped(p))\r
+#else /* MMAP_CLEARS */\r
+#define calloc_must_clear(p) (1)\r
+#endif /* MMAP_CLEARS */\r
+\r
+/* ---------------------- Overlaid data structures ----------------------- */\r
+\r
+/*\r
+ When chunks are not in use, they are treated as nodes of either\r
+ lists or trees.\r
+\r
+ "Small" chunks are stored in circular doubly-linked lists, and look\r
+ like this:\r
+\r
+ chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ | Size of previous chunk |\r
+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ `head:' | Size of chunk, in bytes |P|\r
+ mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ | Forward pointer to next chunk in list |\r
+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ | Back pointer to previous chunk in list |\r
+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ | Unused space (may be 0 bytes long) .\r
+ . .\r
+ . |\r
+nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ `foot:' | Size of chunk, in bytes |\r
+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+\r
+ Larger chunks are kept in a form of bitwise digital trees (aka\r
+ tries) keyed on chunksizes. Because malloc_tree_chunks are only for\r
+ free chunks greater than 256 bytes, their size doesn't impose any\r
+ constraints on user chunk sizes. Each node looks like:\r
+\r
+ chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ | Size of previous chunk |\r
+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ `head:' | Size of chunk, in bytes |P|\r
+ mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ | Forward pointer to next chunk of same size |\r
+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ | Back pointer to previous chunk of same size |\r
+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ | Pointer to left child (child[0]) |\r
+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ | Pointer to right child (child[1]) |\r
+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ | Pointer to parent |\r
+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ | bin index of this chunk |\r
+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ | Unused space .\r
+ . |\r
+nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+ `foot:' | Size of chunk, in bytes |\r
+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+\r
+\r
+ Each tree holding treenodes is a tree of unique chunk sizes. Chunks\r
+ of the same size are arranged in a circularly-linked list, with only\r
+ the oldest chunk (the next to be used, in our FIFO ordering)\r
+ actually in the tree. (Tree members are distinguished by a non-null\r
+ parent pointer.) If a chunk with the same size an an existing node\r
+ is inserted, it is linked off the existing node using pointers that\r
+ work in the same way as fd/bk pointers of small chunks.\r
+\r
+ Each tree contains a power of 2 sized range of chunk sizes (the\r
+ smallest is 0x100 <= x < 0x180), which is is divided in half at each\r
+ tree level, with the chunks in the smaller half of the range (0x100\r
+ <= x < 0x140 for the top nose) in the left subtree and the larger\r
+ half (0x140 <= x < 0x180) in the right subtree. This is, of course,\r
+ done by inspecting individual bits.\r
+\r
+ Using these rules, each node's left subtree contains all smaller\r
+ sizes than its right subtree. However, the node at the root of each\r
+ subtree has no particular ordering relationship to either. (The\r
+ dividing line between the subtree sizes is based on trie relation.)\r
+ If we remove the last chunk of a given size from the interior of the\r
+ tree, we need to replace it with a leaf node. The tree ordering\r
+ rules permit a node to be replaced by any leaf below it.\r
+\r
+ The smallest chunk in a tree (a common operation in a best-fit\r
+ allocator) can be found by walking a path to the leftmost leaf in\r
+ the tree. Unlike a usual binary tree, where we follow left child\r
+ pointers until we reach a null, here we follow the right child\r
+ pointer any time the left one is null, until we reach a leaf with\r
+ both child pointers null. The smallest chunk in the tree will be\r
+ somewhere along that path.\r
+\r
+ The worst case number of steps to add, find, or remove a node is\r
+ bounded by the number of bits differentiating chunks within\r
+ bins. Under current bin calculations, this ranges from 6 up to 21\r
+ (for 32 bit sizes) or up to 53 (for 64 bit sizes). The typical case\r
+ is of course much better.\r
+*/\r
+\r
+struct malloc_tree_chunk {\r
+ /* The first four fields must be compatible with malloc_chunk */\r
+ size_t prev_foot;\r
+ size_t head;\r
+ struct malloc_tree_chunk* fd;\r
+ struct malloc_tree_chunk* bk;\r
+\r
+ struct malloc_tree_chunk* child[2];\r
+ struct malloc_tree_chunk* parent;\r
+ bindex_t index;\r
+};\r
+\r
+typedef struct malloc_tree_chunk tchunk;\r
+typedef struct malloc_tree_chunk* tchunkptr;\r
+typedef struct malloc_tree_chunk* tbinptr; /* The type of bins of trees */\r
+\r
+/* A little helper macro for trees */\r
+#define leftmost_child(t) ((t)->child[0] != 0? (t)->child[0] : (t)->child[1])\r
+\r
+/* ----------------------------- Segments -------------------------------- */\r
+\r
+/*\r
+ Each malloc space may include non-contiguous segments, held in a\r
+ list headed by an embedded malloc_segment record representing the\r
+ top-most space. Segments also include flags holding properties of\r
+ the space. Large chunks that are directly allocated by mmap are not\r
+ included in this list. They are instead independently created and\r
+ destroyed without otherwise keeping track of them.\r
+\r
+ Segment management mainly comes into play for spaces allocated by\r
+ MMAP. Any call to MMAP might or might not return memory that is\r
+ adjacent to an existing segment. MORECORE normally contiguously\r
+ extends the current space, so this space is almost always adjacent,\r
+ which is simpler and faster to deal with. (This is why MORECORE is\r
+ used preferentially to MMAP when both are available -- see\r
+ sys_alloc.) When allocating using MMAP, we don't use any of the\r
+ hinting mechanisms (inconsistently) supported in various\r
+ implementations of unix mmap, or distinguish reserving from\r
+ committing memory. Instead, we just ask for space, and exploit\r
+ contiguity when we get it. It is probably possible to do\r
+ better than this on some systems, but no general scheme seems\r
+ to be significantly better.\r
+\r
+ Management entails a simpler variant of the consolidation scheme\r
+ used for chunks to reduce fragmentation -- new adjacent memory is\r
+ normally prepended or appended to an existing segment. However,\r
+ there are limitations compared to chunk consolidation that mostly\r
+ reflect the fact that segment processing is relatively infrequent\r
+ (occurring only when getting memory from system) and that we\r
+ don't expect to have huge numbers of segments:\r
+\r
+ * Segments are not indexed, so traversal requires linear scans. (It\r
+ would be possible to index these, but is not worth the extra\r
+ overhead and complexity for most programs on most platforms.)\r
+ * New segments are only appended to old ones when holding top-most\r
+ memory; if they cannot be prepended to others, they are held in\r
+ different segments.\r
+\r
+ Except for the top-most segment of an mstate, each segment record\r
+ is kept at the tail of its segment. Segments are added by pushing\r
+ segment records onto the list headed by &mstate.seg for the\r
+ containing mstate.\r
+\r
+ Segment flags control allocation/merge/deallocation policies:\r
+ * If EXTERN_BIT set, then we did not allocate this segment,\r
+ and so should not try to deallocate or merge with others.\r
+ (This currently holds only for the initial segment passed\r
+ into create_mspace_with_base.)\r
+ * If IS_MMAPPED_BIT set, the segment may be merged with\r
+ other surrounding mmapped segments and trimmed/de-allocated\r
+ using munmap.\r
+ * If neither bit is set, then the segment was obtained using\r
+ MORECORE so can be merged with surrounding MORECORE'd segments\r
+ and deallocated/trimmed using MORECORE with negative arguments.\r
+*/\r
+\r
+struct malloc_segment {\r
+ char* base; /* base address */\r
+ size_t size; /* allocated size */\r
+ struct malloc_segment* next; /* ptr to next segment */\r
+ flag_t sflags; /* mmap and extern flag */\r
+};\r
+\r
+#define is_mmapped_segment(S) ((S)->sflags & IS_MMAPPED_BIT)\r
+#define is_extern_segment(S) ((S)->sflags & EXTERN_BIT)\r
+\r
+typedef struct malloc_segment msegment;\r
+typedef struct malloc_segment* msegmentptr;\r
+\r
+/* ---------------------------- malloc_state ----------------------------- */\r
+\r
+/*\r
+ A malloc_state holds all of the bookkeeping for a space.\r
+ The main fields are:\r
+\r
+ Top\r
+ The topmost chunk of the currently active segment. Its size is\r
+ cached in topsize. The actual size of topmost space is\r
+ topsize+TOP_FOOT_SIZE, which includes space reserved for adding\r
+ fenceposts and segment records if necessary when getting more\r
+ space from the system. The size at which to autotrim top is\r
+ cached from mparams in trim_check, except that it is disabled if\r
+ an autotrim fails.\r
+\r
+ Designated victim (dv)\r
+ This is the preferred chunk for servicing small requests that\r
+ don't have exact fits. It is normally the chunk split off most\r
+ recently to service another small request. Its size is cached in\r
+ dvsize. The link fields of this chunk are not maintained since it\r
+ is not kept in a bin.\r
+\r
+ SmallBins\r
+ An array of bin headers for free chunks. These bins hold chunks\r
+ with sizes less than MIN_LARGE_SIZE bytes. Each bin contains\r
+ chunks of all the same size, spaced 8 bytes apart. To simplify\r
+ use in double-linked lists, each bin header acts as a malloc_chunk\r
+ pointing to the real first node, if it exists (else pointing to\r
+ itself). This avoids special-casing for headers. But to avoid\r
+ waste, we allocate only the fd/bk pointers of bins, and then use\r
+ repositioning tricks to treat these as the fields of a chunk.\r
+\r
+ TreeBins\r
+ Treebins are pointers to the roots of trees holding a range of\r
+ sizes. There are 2 equally spaced treebins for each power of two\r
+ from TREE_SHIFT to TREE_SHIFT+16. The last bin holds anything\r
+ larger.\r
+\r
+ Bin maps\r
+ There is one bit map for small bins ("smallmap") and one for\r
+ treebins ("treemap). Each bin sets its bit when non-empty, and\r
+ clears the bit when empty. Bit operations are then used to avoid\r
+ bin-by-bin searching -- nearly all "search" is done without ever\r
+ looking at bins that won't be selected. The bit maps\r
+ conservatively use 32 bits per map word, even if on 64bit system.\r
+ For a good description of some of the bit-based techniques used\r
+ here, see Henry S. Warren Jr's book "Hacker's Delight" (and\r
+ supplement at http://hackersdelight.org/). Many of these are\r
+ intended to reduce the branchiness of paths through malloc etc, as\r
+ well as to reduce the number of memory locations read or written.\r
+\r
+ Segments\r
+ A list of segments headed by an embedded malloc_segment record\r
+ representing the initial space.\r
+\r
+ Address check support\r
+ The least_addr field is the least address ever obtained from\r
+ MORECORE or MMAP. Attempted frees and reallocs of any address less\r
+ than this are trapped (unless INSECURE is defined).\r
+\r
+ Magic tag\r
+ A cross-check field that should always hold same value as mparams.magic.\r
+\r
+ Flags\r
+ Bits recording whether to use MMAP, locks, or contiguous MORECORE\r
+\r
+ Statistics\r
+ Each space keeps track of current and maximum system memory\r
+ obtained via MORECORE or MMAP.\r
+\r
+ Locking\r
+ If USE_LOCKS is defined, the "mutex" lock is acquired and released\r
+ around every public call using this mspace.\r
+*/\r
+\r
+/* Bin types, widths and sizes */\r
+#define NSMALLBINS (32U)\r
+#define NTREEBINS (32U)\r
+#define SMALLBIN_SHIFT (3U)\r
+#define SMALLBIN_WIDTH (SIZE_T_ONE << SMALLBIN_SHIFT)\r
+#define TREEBIN_SHIFT (8U)\r
+#define MIN_LARGE_SIZE (SIZE_T_ONE << TREEBIN_SHIFT)\r
+#define MAX_SMALL_SIZE (MIN_LARGE_SIZE - SIZE_T_ONE)\r
+#define MAX_SMALL_REQUEST (MAX_SMALL_SIZE - CHUNK_ALIGN_MASK - CHUNK_OVERHEAD)\r
+\r
+struct malloc_state {\r
+ binmap_t smallmap;\r
+ binmap_t treemap;\r
+ size_t dvsize;\r
+ size_t topsize;\r
+ char* least_addr;\r
+ mchunkptr dv;\r
+ mchunkptr top;\r
+ size_t trim_check;\r
+ size_t magic;\r
+ mchunkptr smallbins[(NSMALLBINS+1)*2];\r
+ tbinptr treebins[NTREEBINS];\r
+ size_t footprint;\r
+ size_t max_footprint;\r
+ flag_t mflags;\r
+ char cachesync[128];\r
+#if USE_LOCKS\r
+ MLOCK_T mutex; /* locate lock among fields that rarely change */\r
+#endif /* USE_LOCKS */\r
+ msegment seg;\r
+ void* nedpool; /* Points back to nedpool owning this mstate */\r
+};\r
+\r
+typedef struct malloc_state* mstate;\r
+\r
+/* ------------- Global malloc_state and malloc_params ------------------- */\r
+\r
+/*\r
+ malloc_params holds global properties, including those that can be\r
+ dynamically set using mallopt. There is a single instance, mparams,\r
+ initialized in init_mparams.\r
+*/\r
+\r
+struct malloc_params {\r
+ size_t magic;\r
+ size_t page_size;\r
+ size_t granularity;\r
+ size_t mmap_threshold;\r
+ size_t trim_threshold;\r
+ flag_t default_mflags;\r
+};\r
+\r
+static struct malloc_params mparams;\r
+\r
+/* The global malloc_state used for all non-"mspace" calls */\r
+static struct malloc_state _gm_;\r
+#define gm (&_gm_)\r
+#define is_global(M) ((M) == &_gm_)\r
+#define is_initialized(M) ((M)->top != 0)\r
+\r
+/* -------------------------- system alloc setup ------------------------- */\r
+\r
+/* Operations on mflags */\r
+\r
+#define use_lock(M) ((M)->mflags & USE_LOCK_BIT)\r
+#define enable_lock(M) ((M)->mflags |= USE_LOCK_BIT)\r
+#define disable_lock(M) ((M)->mflags &= ~USE_LOCK_BIT)\r
+\r
+#define use_mmap(M) ((M)->mflags & USE_MMAP_BIT)\r
+#define enable_mmap(M) ((M)->mflags |= USE_MMAP_BIT)\r
+#define disable_mmap(M) ((M)->mflags &= ~USE_MMAP_BIT)\r
+\r
+#define use_noncontiguous(M) ((M)->mflags & USE_NONCONTIGUOUS_BIT)\r
+#define disable_contiguous(M) ((M)->mflags |= USE_NONCONTIGUOUS_BIT)\r
+\r
+#define set_lock(M,L)\\r
+ ((M)->mflags = (L)?\\r
+ ((M)->mflags | USE_LOCK_BIT) :\\r
+ ((M)->mflags & ~USE_LOCK_BIT))\r
+\r
+/* page-align a size */\r
+#define page_align(S)\\r
+ (((S) + (mparams.page_size)) & ~(mparams.page_size - SIZE_T_ONE))\r
+\r
+/* granularity-align a size */\r
+#define granularity_align(S)\\r
+ (((S) + (mparams.granularity)) & ~(mparams.granularity - SIZE_T_ONE))\r
+\r
+#define is_page_aligned(S)\\r
+ (((size_t)(S) & (mparams.page_size - SIZE_T_ONE)) == 0)\r
+#define is_granularity_aligned(S)\\r
+ (((size_t)(S) & (mparams.granularity - SIZE_T_ONE)) == 0)\r
+\r
+/* True if segment S holds address A */\r
+#define segment_holds(S, A)\\r
+ ((char*)(A) >= S->base && (char*)(A) < S->base + S->size)\r
+\r
+/* Return segment holding given address */\r
+static msegmentptr segment_holding(mstate m, char* addr) {\r
+ msegmentptr sp = &m->seg;\r
+ for (;;) {\r
+ if (addr >= sp->base && addr < sp->base + sp->size)\r
+ return sp;\r
+ if ((sp = sp->next) == 0)\r
+ return 0;\r
+ }\r
+}\r
+\r
+/* Return true if segment contains a segment link */\r
+static int has_segment_link(mstate m, msegmentptr ss) {\r
+ msegmentptr sp = &m->seg;\r
+ for (;;) {\r
+ if ((char*)sp >= ss->base && (char*)sp < ss->base + ss->size)\r
+ return 1;\r
+ if ((sp = sp->next) == 0)\r
+ return 0;\r
+ }\r
+}\r
+\r
+#ifndef MORECORE_CANNOT_TRIM\r
+#define should_trim(M,s) ((s) > (M)->trim_check)\r
+#else /* MORECORE_CANNOT_TRIM */\r
+#define should_trim(M,s) (0)\r
+#endif /* MORECORE_CANNOT_TRIM */\r
+\r
+/*\r
+ TOP_FOOT_SIZE is padding at the end of a segment, including space\r
+ that may be needed to place segment records and fenceposts when new\r
+ noncontiguous segments are added.\r
+*/\r
+#define TOP_FOOT_SIZE\\r
+ (align_offset(chunk2mem(0))+pad_request(sizeof(struct malloc_segment))+MIN_CHUNK_SIZE)\r
+\r
+\r
+/* ------------------------------- Hooks -------------------------------- */\r
+\r
+/*\r
+ PREACTION should be defined to return 0 on success, and nonzero on\r
+ failure. If you are not using locking, you can redefine these to do\r
+ anything you like.\r
+*/\r
+\r
+#if USE_LOCKS\r
+\r
+/* Ensure locks are initialized */\r
+#define GLOBALLY_INITIALIZE() (mparams.page_size == 0 && init_mparams())\r
+\r
+#define PREACTION(M) ((GLOBALLY_INITIALIZE() || use_lock(M))? ACQUIRE_LOCK(&(M)->mutex) : 0)\r
+#define POSTACTION(M) { if (use_lock(M)) RELEASE_LOCK(&(M)->mutex); }\r
+#else /* USE_LOCKS */\r
+\r
+#ifndef PREACTION\r
+#define PREACTION(M) (0)\r
+#endif /* PREACTION */\r
+\r
+#ifndef POSTACTION\r
+#define POSTACTION(M)\r
+#endif /* POSTACTION */\r
+\r
+#endif /* USE_LOCKS */\r
+\r
+/*\r
+ CORRUPTION_ERROR_ACTION is triggered upon detected bad addresses.\r
+ USAGE_ERROR_ACTION is triggered on detected bad frees and\r
+ reallocs. The argument p is an address that might have triggered the\r
+ fault. It is ignored by the two predefined actions, but might be\r
+ useful in custom actions that try to help diagnose errors.\r
+*/\r
+\r
+#if PROCEED_ON_ERROR\r
+\r
+/* A count of the number of corruption errors causing resets */\r
+int malloc_corruption_error_count;\r
+\r
+/* default corruption action */\r
+static void reset_on_error(mstate m);\r
+\r
+#define CORRUPTION_ERROR_ACTION(m) reset_on_error(m)\r
+#define USAGE_ERROR_ACTION(m, p)\r
+\r
+#else /* PROCEED_ON_ERROR */\r
+\r
+#ifndef CORRUPTION_ERROR_ACTION\r
+#define CORRUPTION_ERROR_ACTION(m) ABORT\r
+#endif /* CORRUPTION_ERROR_ACTION */\r
+\r
+#ifndef USAGE_ERROR_ACTION\r
+#define USAGE_ERROR_ACTION(m,p) ABORT\r
+#endif /* USAGE_ERROR_ACTION */\r
+\r
+#endif /* PROCEED_ON_ERROR */\r
+\r
+/* -------------------------- Debugging setup ---------------------------- */\r
+\r
+#if ! DEBUG\r
+\r
+#define check_free_chunk(M,P)\r
+#define check_inuse_chunk(M,P)\r
+#define check_malloced_chunk(M,P,N)\r
+#define check_mmapped_chunk(M,P)\r
+#define check_malloc_state(M)\r
+#define check_top_chunk(M,P)\r
+\r
+#else /* DEBUG */\r
+#define check_free_chunk(M,P) do_check_free_chunk(M,P)\r
+#define check_inuse_chunk(M,P) do_check_inuse_chunk(M,P)\r
+#define check_top_chunk(M,P) do_check_top_chunk(M,P)\r
+#define check_malloced_chunk(M,P,N) do_check_malloced_chunk(M,P,N)\r
+#define check_mmapped_chunk(M,P) do_check_mmapped_chunk(M,P)\r
+#define check_malloc_state(M) do_check_malloc_state(M)\r
+\r
+static void do_check_any_chunk(mstate m, mchunkptr p);\r
+static void do_check_top_chunk(mstate m, mchunkptr p);\r
+static void do_check_mmapped_chunk(mstate m, mchunkptr p);\r
+static void do_check_inuse_chunk(mstate m, mchunkptr p);\r
+static void do_check_free_chunk(mstate m, mchunkptr p);\r
+static void do_check_malloced_chunk(mstate m, void* mem, size_t s);\r
+static void do_check_tree(mstate m, tchunkptr t);\r
+static void do_check_treebin(mstate m, bindex_t i);\r
+static void do_check_smallbin(mstate m, bindex_t i);\r
+static void do_check_malloc_state(mstate m);\r
+static int bin_find(mstate m, mchunkptr x);\r
+static size_t traverse_and_check(mstate m);\r
+#endif /* DEBUG */\r
+\r
+/* ---------------------------- Indexing Bins ---------------------------- */\r
+\r
+#define is_small(s) (((s) >> SMALLBIN_SHIFT) < NSMALLBINS)\r
+#define small_index(s) ((s) >> SMALLBIN_SHIFT)\r
+#define small_index2size(i) ((i) << SMALLBIN_SHIFT)\r
+#define MIN_SMALL_INDEX (small_index(MIN_CHUNK_SIZE))\r
+\r
+/* addressing by index. See above about smallbin repositioning */\r
+#define smallbin_at(M, i) ((sbinptr)((char*)&((M)->smallbins[(i)<<1])))\r
+#define treebin_at(M,i) (&((M)->treebins[i]))\r
+\r
+/* assign tree index for size S to variable I */\r
+#if defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__))\r
+#define compute_tree_index(S, I)\\r
+{\\r
+ unsigned int X = S >> TREEBIN_SHIFT;\\r
+ if (X == 0)\\r
+ I = 0;\\r
+ else if (X > 0xFFFF)\\r
+ I = NTREEBINS-1;\\r
+ else {\\r
+ unsigned int K;\\r
+ __asm__("bsrl\t%1, %0\n\t" : "=r" (K) : "g" (X));\\r
+ I = (bindex_t)((K << 1) + ((S >> (K + (TREEBIN_SHIFT-1)) & 1)));\\r
+ }\\r
+}\r
+#elif defined(_MSC_VER) && _MSC_VER>=1300\r
+ #ifndef BitScanForward /* Try to avoid pulling in WinNT.h */\r
+ #ifdef __cplusplus\r
+extern "C" {\r
+ #endif\r
+unsigned char _BitScanForward(unsigned long *index, unsigned long mask);\r
+unsigned char _BitScanReverse(unsigned long *index, unsigned long mask);\r
+ #ifdef __cplusplus\r
+}\r
+ #endif\r
+ #define BitScanForward _BitScanForward\r
+ #define BitScanReverse _BitScanReverse\r
+ #pragma intrinsic(_BitScanForward)\r
+ #pragma intrinsic(_BitScanReverse)\r
+ #endif\r
+#define compute_tree_index(S, I)\\r
+{\\r
+ size_t X = S >> TREEBIN_SHIFT;\\r
+ if (X == 0)\\r
+ I = 0;\\r
+ else if (X > 0xFFFF)\\r
+ I = NTREEBINS-1;\\r
+ else {\\r
+ unsigned int K;\\r
+ _BitScanReverse((DWORD *) &K, X);\\r
+ I = (bindex_t)((K << 1) + ((S >> (K + (TREEBIN_SHIFT-1)) & 1)));\\r
+ }\\r
+}\r
+#else /* GNUC */\r
+#define compute_tree_index(S, I)\\r
+{\\r
+ size_t X = S >> TREEBIN_SHIFT;\\r
+ if (X == 0)\\r
+ I = 0;\\r
+ else if (X > 0xFFFF)\\r
+ I = NTREEBINS-1;\\r
+ else {\\r
+ unsigned int Y = (unsigned int)X;\\r
+ unsigned int N = ((Y - 0x100) >> 16) & 8;\\r
+ unsigned int K = (((Y <<= N) - 0x1000) >> 16) & 4;\\r
+ N += K;\\r
+ N += K = (((Y <<= K) - 0x4000) >> 16) & 2;\\r
+ K = 14 - N + ((Y <<= K) >> 15);\\r
+ I = (K << 1) + ((S >> (K + (TREEBIN_SHIFT-1)) & 1));\\r
+ }\\r
+}\r
+#endif /* GNUC */\r
+\r
+/* Bit representing maximum resolved size in a treebin at i */\r
+#define bit_for_tree_index(i) \\r
+ (i == NTREEBINS-1)? (SIZE_T_BITSIZE-1) : (((i) >> 1) + TREEBIN_SHIFT - 2)\r
+\r
+/* Shift placing maximum resolved bit in a treebin at i as sign bit */\r
+#define leftshift_for_tree_index(i) \\r
+ ((i == NTREEBINS-1)? 0 : \\r
+ ((SIZE_T_BITSIZE-SIZE_T_ONE) - (((i) >> 1) + TREEBIN_SHIFT - 2)))\r
+\r
+/* The size of the smallest chunk held in bin with index i */\r
+#define minsize_for_tree_index(i) \\r
+ ((SIZE_T_ONE << (((i) >> 1) + TREEBIN_SHIFT)) | \\r
+ (((size_t)((i) & SIZE_T_ONE)) << (((i) >> 1) + TREEBIN_SHIFT - 1)))\r
+\r
+\r
+/* ------------------------ Operations on bin maps ----------------------- */\r
+\r
+/* bit corresponding to given index */\r
+#define idx2bit(i) ((binmap_t)(1) << (i))\r
+\r
+/* Mark/Clear bits with given index */\r
+#define mark_smallmap(M,i) ((M)->smallmap |= idx2bit(i))\r
+#define clear_smallmap(M,i) ((M)->smallmap &= ~idx2bit(i))\r
+#define smallmap_is_marked(M,i) ((M)->smallmap & idx2bit(i))\r
+\r
+#define mark_treemap(M,i) ((M)->treemap |= idx2bit(i))\r
+#define clear_treemap(M,i) ((M)->treemap &= ~idx2bit(i))\r
+#define treemap_is_marked(M,i) ((M)->treemap & idx2bit(i))\r
+\r
+/* index corresponding to given bit */\r
+\r
+#if defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__))\r
+#define compute_bit2idx(X, I)\\r
+{\\r
+ unsigned int J;\\r
+ __asm__("bsfl\t%1, %0\n\t" : "=r" (J) : "g" (X));\\r
+ I = (bindex_t)J;\\r
+}\r
+#elif defined(_MSC_VER) && _MSC_VER>=1300\r
+#define compute_bit2idx(X, I)\\r
+{\\r
+ unsigned int J;\\r
+ _BitScanForward((DWORD *) &J, X);\\r
+ I = (bindex_t)J;\\r
+}\r
+\r
+#else /* GNUC */\r
+#if USE_BUILTIN_FFS\r
+#define compute_bit2idx(X, I) I = ffs(X)-1\r
+\r
+#else /* USE_BUILTIN_FFS */\r
+#define compute_bit2idx(X, I)\\r
+{\\r
+ unsigned int Y = X - 1;\\r
+ unsigned int K = Y >> (16-4) & 16;\\r
+ unsigned int N = K; Y >>= K;\\r
+ N += K = Y >> (8-3) & 8; Y >>= K;\\r
+ N += K = Y >> (4-2) & 4; Y >>= K;\\r
+ N += K = Y >> (2-1) & 2; Y >>= K;\\r
+ N += K = Y >> (1-0) & 1; Y >>= K;\\r
+ I = (bindex_t)(N + Y);\\r
+}\r
+#endif /* USE_BUILTIN_FFS */\r
+#endif /* GNUC */\r
+\r
+/* isolate the least set bit of a bitmap */\r
+#define least_bit(x) ((x) & -(x))\r
+\r
+/* mask with all bits to left of least bit of x on */\r
+#define left_bits(x) ((x<<1) | -(x<<1))\r
+\r
+/* mask with all bits to left of or equal to least bit of x on */\r
+#define same_or_left_bits(x) ((x) | -(x))\r
+\r
+\r
+/* ----------------------- Runtime Check Support ------------------------- */\r
+\r
+/*\r
+ For security, the main invariant is that malloc/free/etc never\r
+ writes to a static address other than malloc_state, unless static\r
+ malloc_state itself has been corrupted, which cannot occur via\r
+ malloc (because of these checks). In essence this means that we\r
+ believe all pointers, sizes, maps etc held in malloc_state, but\r
+ check all of those linked or offsetted from other embedded data\r
+ structures. These checks are interspersed with main code in a way\r
+ that tends to minimize their run-time cost.\r
+\r
+ When FOOTERS is defined, in addition to range checking, we also\r
+ verify footer fields of inuse chunks, which can be used guarantee\r
+ that the mstate controlling malloc/free is intact. This is a\r
+ streamlined version of the approach described by William Robertson\r
+ et al in "Run-time Detection of Heap-based Overflows" LISA'03\r
+ http://www.usenix.org/events/lisa03/tech/robertson.html The footer\r
+ of an inuse chunk holds the xor of its mstate and a random seed,\r
+ that is checked upon calls to free() and realloc(). This is\r
+ (probablistically) unguessable from outside the program, but can be\r
+ computed by any code successfully malloc'ing any chunk, so does not\r
+ itself provide protection against code that has already broken\r
+ security through some other means. Unlike Robertson et al, we\r
+ always dynamically check addresses of all offset chunks (previous,\r
+ next, etc). This turns out to be cheaper than relying on hashes.\r
+*/\r
+\r
+#if !INSECURE\r
+/* Check if address a is at least as high as any from MORECORE or MMAP */\r
+#define ok_address(M, a) ((char*)(a) >= (M)->least_addr)\r
+/* Check if address of next chunk n is higher than base chunk p */\r
+#define ok_next(p, n) ((char*)(p) < (char*)(n))\r
+/* Check if p has its cinuse bit on */\r
+#define ok_cinuse(p) cinuse(p)\r
+/* Check if p has its pinuse bit on */\r
+#define ok_pinuse(p) pinuse(p)\r
+\r
+#else /* !INSECURE */\r
+#define ok_address(M, a) (1)\r
+#define ok_next(b, n) (1)\r
+#define ok_cinuse(p) (1)\r
+#define ok_pinuse(p) (1)\r
+#endif /* !INSECURE */\r
+\r
+#if (FOOTERS && !INSECURE)\r
+/* Check if (alleged) mstate m has expected magic field */\r
+#define ok_magic(M) ((M)->magic == mparams.magic)\r
+#else /* (FOOTERS && !INSECURE) */\r
+#define ok_magic(M) (1)\r
+#endif /* (FOOTERS && !INSECURE) */\r
+\r
+\r
+/* In gcc, use __builtin_expect to minimize impact of checks */\r
+#if !INSECURE\r
+#if defined(__GNUC__) && __GNUC__ >= 3\r
+#define RTCHECK(e) __builtin_expect(e, 1)\r
+#else /* GNUC */\r
+#define RTCHECK(e) (e)\r
+#endif /* GNUC */\r
+#else /* !INSECURE */\r
+#define RTCHECK(e) (1)\r
+#endif /* !INSECURE */\r
+\r
+/* macros to set up inuse chunks with or without footers */\r
+\r
+#if !FOOTERS\r
+\r
+#define mark_inuse_foot(M,p,s)\r
+\r
+/* Set cinuse bit and pinuse bit of next chunk */\r
+#define set_inuse(M,p,s)\\r
+ ((p)->head = (((p)->head & PINUSE_BIT)|s|CINUSE_BIT),\\r
+ ((mchunkptr)(((char*)(p)) + (s)))->head |= PINUSE_BIT)\r
+\r
+/* Set cinuse and pinuse of this chunk and pinuse of next chunk */\r
+#define set_inuse_and_pinuse(M,p,s)\\r
+ ((p)->head = (s|PINUSE_BIT|CINUSE_BIT),\\r
+ ((mchunkptr)(((char*)(p)) + (s)))->head |= PINUSE_BIT)\r
+\r
+/* Set size, cinuse and pinuse bit of this chunk */\r
+#define set_size_and_pinuse_of_inuse_chunk(M, p, s)\\r
+ ((p)->head = (s|PINUSE_BIT|CINUSE_BIT))\r
+\r
+#else /* FOOTERS */\r
+\r
+/* Set foot of inuse chunk to be xor of mstate and seed */\r
+#define mark_inuse_foot(M,p,s)\\r
+ (((mchunkptr)((char*)(p) + (s)))->prev_foot = ((size_t)(M) ^ mparams.magic))\r
+\r
+#define get_mstate_for(p)\\r
+ ((mstate)(((mchunkptr)((char*)(p) +\\r
+ (chunksize(p))))->prev_foot ^ mparams.magic))\r
+\r
+#define set_inuse(M,p,s)\\r
+ ((p)->head = (((p)->head & PINUSE_BIT)|s|CINUSE_BIT),\\r
+ (((mchunkptr)(((char*)(p)) + (s)))->head |= PINUSE_BIT), \\r
+ mark_inuse_foot(M,p,s))\r
+\r
+#define set_inuse_and_pinuse(M,p,s)\\r
+ ((p)->head = (s|PINUSE_BIT|CINUSE_BIT),\\r
+ (((mchunkptr)(((char*)(p)) + (s)))->head |= PINUSE_BIT),\\r
+ mark_inuse_foot(M,p,s))\r
+\r
+#define set_size_and_pinuse_of_inuse_chunk(M, p, s)\\r
+ ((p)->head = (s|PINUSE_BIT|CINUSE_BIT),\\r
+ mark_inuse_foot(M, p, s))\r
+\r
+#endif /* !FOOTERS */\r
+\r
+/* ---------------------------- setting mparams -------------------------- */\r
+\r
+/* Initialize mparams */\r
+static int init_mparams(void) {\r
+ if (mparams.page_size == 0) {\r
+ size_t s;\r
+\r
+ mparams.mmap_threshold = DEFAULT_MMAP_THRESHOLD;\r
+ mparams.trim_threshold = DEFAULT_TRIM_THRESHOLD;\r
+#if MORECORE_CONTIGUOUS\r
+ mparams.default_mflags = USE_LOCK_BIT|USE_MMAP_BIT;\r
+#else /* MORECORE_CONTIGUOUS */\r
+ mparams.default_mflags = USE_LOCK_BIT|USE_MMAP_BIT|USE_NONCONTIGUOUS_BIT;\r
+#endif /* MORECORE_CONTIGUOUS */\r
+\r
+#if (FOOTERS && !INSECURE)\r
+ {\r
+#if USE_DEV_RANDOM\r
+ int fd;\r
+ unsigned char buf[sizeof(size_t)];\r
+ /* Try to use /dev/urandom, else fall back on using time */\r
+ if ((fd = open("/dev/urandom", O_RDONLY)) >= 0 &&\r
+ read(fd, buf, sizeof(buf)) == sizeof(buf)) {\r
+ s = *((size_t *) buf);\r
+ close(fd);\r
+ }\r
+ else\r
+#endif /* USE_DEV_RANDOM */\r
+ s = (size_t)(time(0) ^ (size_t)0x55555555U);\r
+\r
+ s |= (size_t)8U; /* ensure nonzero */\r
+ s &= ~(size_t)7U; /* improve chances of fault for bad values */\r
+\r
+ }\r
+#else /* (FOOTERS && !INSECURE) */\r
+ s = (size_t)0x58585858U;\r
+#endif /* (FOOTERS && !INSECURE) */\r
+ ACQUIRE_MAGIC_INIT_LOCK();\r
+ if (mparams.magic == 0) {\r
+ mparams.magic = s;\r
+ /* Set up lock for main malloc area */\r
+ INITIAL_LOCK(&gm->mutex);\r
+ gm->mflags = mparams.default_mflags;\r
+ }\r
+ RELEASE_MAGIC_INIT_LOCK();\r
+\r
+#ifndef WIN32\r
+ mparams.page_size = malloc_getpagesize;\r
+ mparams.granularity = ((DEFAULT_GRANULARITY != 0)?\r
+ DEFAULT_GRANULARITY : mparams.page_size);\r
+#else /* WIN32 */\r
+ {\r
+ SYSTEM_INFO system_info;\r
+ GetSystemInfo(&system_info);\r
+ mparams.page_size = system_info.dwPageSize;\r
+ mparams.granularity = system_info.dwAllocationGranularity;\r
+ }\r
+#endif /* WIN32 */\r
+\r
+ /* Sanity-check configuration:\r
+ size_t must be unsigned and as wide as pointer type.\r
+ ints must be at least 4 bytes.\r
+ alignment must be at least 8.\r
+ Alignment, min chunk size, and page size must all be powers of 2.\r
+ */\r
+ if ((sizeof(size_t) != sizeof(char*)) ||\r
+ (MAX_SIZE_T < MIN_CHUNK_SIZE) ||\r
+ (sizeof(int) < 4) ||\r
+ (MALLOC_ALIGNMENT < (size_t)8U) ||\r
+ ((MALLOC_ALIGNMENT & (MALLOC_ALIGNMENT-SIZE_T_ONE)) != 0) ||\r
+ ((MCHUNK_SIZE & (MCHUNK_SIZE-SIZE_T_ONE)) != 0) ||\r
+ ((mparams.granularity & (mparams.granularity-SIZE_T_ONE)) != 0) ||\r
+ ((mparams.page_size & (mparams.page_size-SIZE_T_ONE)) != 0))\r
+ ABORT;\r
+ }\r
+ return 0;\r
+}\r
+\r
+/* support for mallopt */\r
+static int change_mparam(int param_number, int value) {\r
+ size_t val = (size_t)value;\r
+ init_mparams();\r
+ switch(param_number) {\r
+ case M_TRIM_THRESHOLD:\r
+ mparams.trim_threshold = val;\r
+ return 1;\r
+ case M_GRANULARITY:\r
+ if (val >= mparams.page_size && ((val & (val-1)) == 0)) {\r
+ mparams.granularity = val;\r
+ return 1;\r
+ }\r
+ else\r
+ return 0;\r
+ case M_MMAP_THRESHOLD:\r
+ mparams.mmap_threshold = val;\r
+ return 1;\r
+ default:\r
+ return 0;\r
+ }\r
+}\r
+\r
+#if DEBUG\r
+/* ------------------------- Debugging Support --------------------------- */\r
+\r
+/* Check properties of any chunk, whether free, inuse, mmapped etc */\r
+static void do_check_any_chunk(mstate m, mchunkptr p) {\r
+ assert((is_aligned(chunk2mem(p))) || (p->head == FENCEPOST_HEAD));\r
+ assert(ok_address(m, p));\r
+}\r
+\r
+/* Check properties of top chunk */\r
+static void do_check_top_chunk(mstate m, mchunkptr p) {\r
+ msegmentptr sp = segment_holding(m, (char*)p);\r
+ size_t sz = chunksize(p);\r
+ assert(sp != 0);\r
+ assert((is_aligned(chunk2mem(p))) || (p->head == FENCEPOST_HEAD));\r
+ assert(ok_address(m, p));\r
+ assert(sz == m->topsize);\r
+ assert(sz > 0);\r
+ assert(sz == ((sp->base + sp->size) - (char*)p) - TOP_FOOT_SIZE);\r
+ assert(pinuse(p));\r
+ assert(!next_pinuse(p));\r
+}\r
+\r
+/* Check properties of (inuse) mmapped chunks */\r
+static void do_check_mmapped_chunk(mstate m, mchunkptr p) {\r
+ size_t sz = chunksize(p);\r
+ size_t len = (sz + (p->prev_foot & ~IS_MMAPPED_BIT) + MMAP_FOOT_PAD);\r
+ assert(is_mmapped(p));\r
+ assert(use_mmap(m));\r
+ assert((is_aligned(chunk2mem(p))) || (p->head == FENCEPOST_HEAD));\r
+ assert(ok_address(m, p));\r
+ assert(!is_small(sz));\r
+ assert((len & (mparams.page_size-SIZE_T_ONE)) == 0);\r
+ assert(chunk_plus_offset(p, sz)->head == FENCEPOST_HEAD);\r
+ assert(chunk_plus_offset(p, sz+SIZE_T_SIZE)->head == 0);\r
+}\r
+\r
+/* Check properties of inuse chunks */\r
+static void do_check_inuse_chunk(mstate m, mchunkptr p) {\r
+ do_check_any_chunk(m, p);\r
+ assert(cinuse(p));\r
+ assert(next_pinuse(p));\r
+ /* If not pinuse and not mmapped, previous chunk has OK offset */\r
+ assert(is_mmapped(p) || pinuse(p) || next_chunk(prev_chunk(p)) == p);\r
+ if (is_mmapped(p))\r
+ do_check_mmapped_chunk(m, p);\r
+}\r
+\r
+/* Check properties of free chunks */\r
+static void do_check_free_chunk(mstate m, mchunkptr p) {\r
+ size_t sz = p->head & ~(PINUSE_BIT|CINUSE_BIT);\r
+ mchunkptr next = chunk_plus_offset(p, sz);\r
+ do_check_any_chunk(m, p);\r
+ assert(!cinuse(p));\r
+ assert(!next_pinuse(p));\r
+ assert (!is_mmapped(p));\r
+ if (p != m->dv && p != m->top) {\r
+ if (sz >= MIN_CHUNK_SIZE) {\r
+ assert((sz & CHUNK_ALIGN_MASK) == 0);\r
+ assert(is_aligned(chunk2mem(p)));\r
+ assert(next->prev_foot == sz);\r
+ assert(pinuse(p));\r
+ assert (next == m->top || cinuse(next));\r
+ assert(p->fd->bk == p);\r
+ assert(p->bk->fd == p);\r
+ }\r
+ else /* markers are always of size SIZE_T_SIZE */\r
+ assert(sz == SIZE_T_SIZE);\r
+ }\r
+}\r
+\r
+/* Check properties of malloced chunks at the point they are malloced */\r
+static void do_check_malloced_chunk(mstate m, void* mem, size_t s) {\r
+ if (mem != 0) {\r
+ mchunkptr p = mem2chunk(mem);\r
+ size_t sz = p->head & ~(PINUSE_BIT|CINUSE_BIT);\r
+ do_check_inuse_chunk(m, p);\r
+ assert((sz & CHUNK_ALIGN_MASK) == 0);\r
+ assert(sz >= MIN_CHUNK_SIZE);\r
+ assert(sz >= s);\r
+ /* unless mmapped, size is less than MIN_CHUNK_SIZE more than request */\r
+ assert(is_mmapped(p) || sz < (s + MIN_CHUNK_SIZE));\r
+ }\r
+}\r
+\r
+/* Check a tree and its subtrees. */\r
+static void do_check_tree(mstate m, tchunkptr t) {\r
+ tchunkptr head = 0;\r
+ tchunkptr u = t;\r
+ bindex_t tindex = t->index;\r
+ size_t tsize = chunksize(t);\r
+ bindex_t idx;\r
+ compute_tree_index(tsize, idx);\r
+ assert(tindex == idx);\r
+ assert(tsize >= MIN_LARGE_SIZE);\r
+ assert(tsize >= minsize_for_tree_index(idx));\r
+ assert((idx == NTREEBINS-1) || (tsize < minsize_for_tree_index((idx+1))));\r
+\r
+ do { /* traverse through chain of same-sized nodes */\r
+ do_check_any_chunk(m, ((mchunkptr)u));\r
+ assert(u->index == tindex);\r
+ assert(chunksize(u) == tsize);\r
+ assert(!cinuse(u));\r
+ assert(!next_pinuse(u));\r
+ assert(u->fd->bk == u);\r
+ assert(u->bk->fd == u);\r
+ if (u->parent == 0) {\r
+ assert(u->child[0] == 0);\r
+ assert(u->child[1] == 0);\r
+ }\r
+ else {\r
+ assert(head == 0); /* only one node on chain has parent */\r
+ head = u;\r
+ assert(u->parent != u);\r
+ assert (u->parent->child[0] == u ||\r
+ u->parent->child[1] == u ||\r
+ *((tbinptr*)(u->parent)) == u);\r
+ if (u->child[0] != 0) {\r
+ assert(u->child[0]->parent == u);\r
+ assert(u->child[0] != u);\r
+ do_check_tree(m, u->child[0]);\r
+ }\r
+ if (u->child[1] != 0) {\r
+ assert(u->child[1]->parent == u);\r
+ assert(u->child[1] != u);\r
+ do_check_tree(m, u->child[1]);\r
+ }\r
+ if (u->child[0] != 0 && u->child[1] != 0) {\r
+ assert(chunksize(u->child[0]) < chunksize(u->child[1]));\r
+ }\r
+ }\r
+ u = u->fd;\r
+ } while (u != t);\r
+ assert(head != 0);\r
+}\r
+\r
+/* Check all the chunks in a treebin. */\r
+static void do_check_treebin(mstate m, bindex_t i) {\r
+ tbinptr* tb = treebin_at(m, i);\r
+ tchunkptr t = *tb;\r
+ int empty = (m->treemap & (1U << i)) == 0;\r
+ if (t == 0)\r
+ assert(empty);\r
+ if (!empty)\r
+ do_check_tree(m, t);\r
+}\r
+\r
+/* Check all the chunks in a smallbin. */\r
+static void do_check_smallbin(mstate m, bindex_t i) {\r
+ sbinptr b = smallbin_at(m, i);\r
+ mchunkptr p = b->bk;\r
+ unsigned int empty = (m->smallmap & (1U << i)) == 0;\r
+ if (p == b)\r
+ assert(empty);\r
+ if (!empty) {\r
+ for (; p != b; p = p->bk) {\r
+ size_t size = chunksize(p);\r
+ mchunkptr q;\r
+ /* each chunk claims to be free */\r
+ do_check_free_chunk(m, p);\r
+ /* chunk belongs in bin */\r
+ assert(small_index(size) == i);\r
+ assert(p->bk == b || chunksize(p->bk) == chunksize(p));\r
+ /* chunk is followed by an inuse chunk */\r
+ q = next_chunk(p);\r
+ if (q->head != FENCEPOST_HEAD)\r
+ do_check_inuse_chunk(m, q);\r
+ }\r
+ }\r
+}\r
+\r
+/* Find x in a bin. Used in other check functions. */\r
+static int bin_find(mstate m, mchunkptr x) {\r
+ size_t size = chunksize(x);\r
+ if (is_small(size)) {\r
+ bindex_t sidx = small_index(size);\r
+ sbinptr b = smallbin_at(m, sidx);\r
+ if (smallmap_is_marked(m, sidx)) {\r
+ mchunkptr p = b;\r
+ do {\r
+ if (p == x)\r
+ return 1;\r
+ } while ((p = p->fd) != b);\r
+ }\r
+ }\r
+ else {\r
+ bindex_t tidx;\r
+ compute_tree_index(size, tidx);\r
+ if (treemap_is_marked(m, tidx)) {\r
+ tchunkptr t = *treebin_at(m, tidx);\r
+ size_t sizebits = size << leftshift_for_tree_index(tidx);\r
+ while (t != 0 && chunksize(t) != size) {\r
+ t = t->child[(sizebits >> (SIZE_T_BITSIZE-SIZE_T_ONE)) & 1];\r
+ sizebits <<= 1;\r
+ }\r
+ if (t != 0) {\r
+ tchunkptr u = t;\r
+ do {\r
+ if (u == (tchunkptr)x)\r
+ return 1;\r
+ } while ((u = u->fd) != t);\r
+ }\r
+ }\r
+ }\r
+ return 0;\r
+}\r
+\r
+/* Traverse each chunk and check it; return total */\r
+static size_t traverse_and_check(mstate m) {\r
+ size_t sum = 0;\r
+ if (is_initialized(m)) {\r
+ msegmentptr s = &m->seg;\r
+ sum += m->topsize + TOP_FOOT_SIZE;\r
+ while (s != 0) {\r
+ mchunkptr q = align_as_chunk(s->base);\r
+ mchunkptr lastq = 0;\r
+ assert(pinuse(q));\r
+ while (segment_holds(s, q) &&\r
+ q != m->top && q->head != FENCEPOST_HEAD) {\r
+ sum += chunksize(q);\r
+ if (cinuse(q)) {\r
+ assert(!bin_find(m, q));\r
+ do_check_inuse_chunk(m, q);\r
+ }\r
+ else {\r
+ assert(q == m->dv || bin_find(m, q));\r
+ assert(lastq == 0 || cinuse(lastq)); /* Not 2 consecutive free */\r
+ do_check_free_chunk(m, q);\r
+ }\r
+ lastq = q;\r
+ q = next_chunk(q);\r
+ }\r
+ s = s->next;\r
+ }\r
+ }\r
+ return sum;\r
+}\r
+\r
+/* Check all properties of malloc_state. */\r
+static void do_check_malloc_state(mstate m) {\r
+ bindex_t i;\r
+ size_t total;\r
+ /* check bins */\r
+ for (i = 0; i < NSMALLBINS; ++i)\r
+ do_check_smallbin(m, i);\r
+ for (i = 0; i < NTREEBINS; ++i)\r
+ do_check_treebin(m, i);\r
+\r
+ if (m->dvsize != 0) { /* check dv chunk */\r
+ do_check_any_chunk(m, m->dv);\r
+ assert(m->dvsize == chunksize(m->dv));\r
+ assert(m->dvsize >= MIN_CHUNK_SIZE);\r
+ assert(bin_find(m, m->dv) == 0);\r
+ }\r
+\r
+ if (m->top != 0) { /* check top chunk */\r
+ do_check_top_chunk(m, m->top);\r
+ assert(m->topsize == chunksize(m->top));\r
+ assert(m->topsize > 0);\r
+ assert(bin_find(m, m->top) == 0);\r
+ }\r
+\r
+ total = traverse_and_check(m);\r
+ assert(total <= m->footprint);\r
+ assert(m->footprint <= m->max_footprint);\r
+}\r
+#endif /* DEBUG */\r
+\r
+/* ----------------------------- statistics ------------------------------ */\r
+\r
+#if !NO_MALLINFO\r
+static struct mallinfo internal_mallinfo(mstate m) {\r
+ struct mallinfo nm = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 };\r
+ if (!PREACTION(m)) {\r
+ check_malloc_state(m);\r
+ if (is_initialized(m)) {\r
+ size_t nfree = SIZE_T_ONE; /* top always free */\r
+ size_t mfree = m->topsize + TOP_FOOT_SIZE;\r
+ size_t sum = mfree;\r
+ msegmentptr s = &m->seg;\r
+ while (s != 0) {\r
+ mchunkptr q = align_as_chunk(s->base);\r
+ while (segment_holds(s, q) &&\r
+ q != m->top && q->head != FENCEPOST_HEAD) {\r
+ size_t sz = chunksize(q);\r
+ sum += sz;\r
+ if (!cinuse(q)) {\r
+ mfree += sz;\r
+ ++nfree;\r
+ }\r
+ q = next_chunk(q);\r
+ }\r
+ s = s->next;\r
+ }\r
+\r
+ nm.arena = sum;\r
+ nm.ordblks = nfree;\r
+ nm.hblkhd = m->footprint - sum;\r
+ nm.usmblks = m->max_footprint;\r
+ nm.uordblks = m->footprint - mfree;\r
+ nm.fordblks = mfree;\r
+ nm.keepcost = m->topsize;\r
+ }\r
+\r
+ POSTACTION(m);\r
+ }\r
+ return nm;\r
+}\r
+#endif /* !NO_MALLINFO */\r
+\r
+static void internal_malloc_stats(mstate m) {\r
+ if (!PREACTION(m)) {\r
+ size_t maxfp = 0;\r
+ size_t fp = 0;\r
+ size_t used = 0;\r
+ check_malloc_state(m);\r
+ if (is_initialized(m)) {\r
+ msegmentptr s = &m->seg;\r
+ maxfp = m->max_footprint;\r
+ fp = m->footprint;\r
+ used = fp - (m->topsize + TOP_FOOT_SIZE);\r
+\r
+ while (s != 0) {\r
+ mchunkptr q = align_as_chunk(s->base);\r
+ while (segment_holds(s, q) &&\r
+ q != m->top && q->head != FENCEPOST_HEAD) {\r
+ if (!cinuse(q))\r
+ used -= chunksize(q);\r
+ q = next_chunk(q);\r
+ }\r
+ s = s->next;\r
+ }\r
+ }\r
+\r
+ fprintf(stderr, "max system bytes = %10lu\n", (unsigned long)(maxfp));\r
+ fprintf(stderr, "system bytes = %10lu\n", (unsigned long)(fp));\r
+ fprintf(stderr, "in use bytes = %10lu\n", (unsigned long)(used));\r
+\r
+ POSTACTION(m);\r
+ }\r
+}\r
+\r
+/* ----------------------- Operations on smallbins ----------------------- */\r
+\r
+/*\r
+ Various forms of linking and unlinking are defined as macros. Even\r
+ the ones for trees, which are very long but have very short typical\r
+ paths. This is ugly but reduces reliance on inlining support of\r
+ compilers.\r
+*/\r
+\r
+/* Link a free chunk into a smallbin */\r
+#define insert_small_chunk(M, P, S) {\\r
+ bindex_t I = small_index(S);\\r
+ mchunkptr B = smallbin_at(M, I);\\r
+ mchunkptr F = B;\\r
+ assert(S >= MIN_CHUNK_SIZE);\\r
+ if (!smallmap_is_marked(M, I))\\r
+ mark_smallmap(M, I);\\r
+ else if (RTCHECK(ok_address(M, B->fd)))\\r
+ F = B->fd;\\r
+ else {\\r
+ CORRUPTION_ERROR_ACTION(M);\\r
+ }\\r
+ B->fd = P;\\r
+ F->bk = P;\\r
+ P->fd = F;\\r
+ P->bk = B;\\r
+}\r
+\r
+/* Unlink a chunk from a smallbin */\r
+#define unlink_small_chunk(M, P, S) {\\r
+ mchunkptr F = P->fd;\\r
+ mchunkptr B = P->bk;\\r
+ bindex_t I = small_index(S);\\r
+ assert(P != B);\\r
+ assert(P != F);\\r
+ assert(chunksize(P) == small_index2size(I));\\r
+ if (F == B)\\r
+ clear_smallmap(M, I);\\r
+ else if (RTCHECK((F == smallbin_at(M,I) || ok_address(M, F)) &&\\r
+ (B == smallbin_at(M,I) || ok_address(M, B)))) {\\r
+ F->bk = B;\\r
+ B->fd = F;\\r
+ }\\r
+ else {\\r
+ CORRUPTION_ERROR_ACTION(M);\\r
+ }\\r
+}\r
+\r
+/* Unlink the first chunk from a smallbin */\r
+#define unlink_first_small_chunk(M, B, P, I) {\\r
+ mchunkptr F = P->fd;\\r
+ assert(P != B);\\r
+ assert(P != F);\\r
+ assert(chunksize(P) == small_index2size(I));\\r
+ if (B == F)\\r
+ clear_smallmap(M, I);\\r
+ else if (RTCHECK(ok_address(M, F))) {\\r
+ B->fd = F;\\r
+ F->bk = B;\\r
+ }\\r
+ else {\\r
+ CORRUPTION_ERROR_ACTION(M);\\r
+ }\\r
+}\r
+\r
+/* Replace dv node, binning the old one */\r
+/* Used only when dvsize known to be small */\r
+#define replace_dv(M, P, S) {\\r
+ size_t DVS = M->dvsize;\\r
+ if (DVS != 0) {\\r
+ mchunkptr DV = M->dv;\\r
+ assert(is_small(DVS));\\r
+ insert_small_chunk(M, DV, DVS);\\r
+ }\\r
+ M->dvsize = S;\\r
+ M->dv = P;\\r
+}\r
+\r
+/* ------------------------- Operations on trees ------------------------- */\r
+\r
+/* Insert chunk into tree */\r
+#define insert_large_chunk(M, X, S) {\\r
+ tbinptr* H;\\r
+ bindex_t I;\\r
+ compute_tree_index(S, I);\\r
+ H = treebin_at(M, I);\\r
+ X->index = I;\\r
+ X->child[0] = X->child[1] = 0;\\r
+ if (!treemap_is_marked(M, I)) {\\r
+ mark_treemap(M, I);\\r
+ *H = X;\\r
+ X->parent = (tchunkptr)H;\\r
+ X->fd = X->bk = X;\\r
+ }\\r
+ else {\\r
+ tchunkptr T = *H;\\r
+ size_t K = S << leftshift_for_tree_index(I);\\r
+ for (;;) {\\r
+ if (chunksize(T) != S) {\\r
+ tchunkptr* C = &(T->child[(K >> (SIZE_T_BITSIZE-SIZE_T_ONE)) & 1]);\\r
+ K <<= 1;\\r
+ if (*C != 0)\\r
+ T = *C;\\r
+ else if (RTCHECK(ok_address(M, C))) {\\r
+ *C = X;\\r
+ X->parent = T;\\r
+ X->fd = X->bk = X;\\r
+ break;\\r
+ }\\r
+ else {\\r
+ CORRUPTION_ERROR_ACTION(M);\\r
+ break;\\r
+ }\\r
+ }\\r
+ else {\\r
+ tchunkptr F = T->fd;\\r
+ if (RTCHECK(ok_address(M, T) && ok_address(M, F))) {\\r
+ T->fd = F->bk = X;\\r
+ X->fd = F;\\r
+ X->bk = T;\\r
+ X->parent = 0;\\r
+ break;\\r
+ }\\r
+ else {\\r
+ CORRUPTION_ERROR_ACTION(M);\\r
+ break;\\r
+ }\\r
+ }\\r
+ }\\r
+ }\\r
+}\r
+\r
+/*\r
+ Unlink steps:\r
+\r
+ 1. If x is a chained node, unlink it from its same-sized fd/bk links\r
+ and choose its bk node as its replacement.\r
+ 2. If x was the last node of its size, but not a leaf node, it must\r
+ be replaced with a leaf node (not merely one with an open left or\r
+ right), to make sure that lefts and rights of descendents\r
+ correspond properly to bit masks. We use the rightmost descendent\r
+ of x. We could use any other leaf, but this is easy to locate and\r
+ tends to counteract removal of leftmosts elsewhere, and so keeps\r
+ paths shorter than minimally guaranteed. This doesn't loop much\r
+ because on average a node in a tree is near the bottom.\r
+ 3. If x is the base of a chain (i.e., has parent links) relink\r
+ x's parent and children to x's replacement (or null if none).\r
+*/\r
+\r
+#define unlink_large_chunk(M, X) {\\r
+ tchunkptr XP = X->parent;\\r
+ tchunkptr R;\\r
+ if (X->bk != X) {\\r
+ tchunkptr F = X->fd;\\r
+ R = X->bk;\\r
+ if (RTCHECK(ok_address(M, F))) {\\r
+ F->bk = R;\\r
+ R->fd = F;\\r
+ }\\r
+ else {\\r
+ CORRUPTION_ERROR_ACTION(M);\\r
+ }\\r
+ }\\r
+ else {\\r
+ tchunkptr* RP;\\r
+ if (((R = *(RP = &(X->child[1]))) != 0) ||\\r
+ ((R = *(RP = &(X->child[0]))) != 0)) {\\r
+ tchunkptr* CP;\\r
+ while ((*(CP = &(R->child[1])) != 0) ||\\r
+ (*(CP = &(R->child[0])) != 0)) {\\r
+ R = *(RP = CP);\\r
+ }\\r
+ if (RTCHECK(ok_address(M, RP)))\\r
+ *RP = 0;\\r
+ else {\\r
+ CORRUPTION_ERROR_ACTION(M);\\r
+ }\\r
+ }\\r
+ }\\r
+ if (XP != 0) {\\r
+ tbinptr* H = treebin_at(M, X->index);\\r
+ if (X == *H) {\\r
+ if ((*H = R) == 0) \\r
+ clear_treemap(M, X->index);\\r
+ }\\r
+ else if (RTCHECK(ok_address(M, XP))) {\\r
+ if (XP->child[0] == X) \\r
+ XP->child[0] = R;\\r
+ else \\r
+ XP->child[1] = R;\\r
+ }\\r
+ else\\r
+ CORRUPTION_ERROR_ACTION(M);\\r
+ if (R != 0) {\\r
+ if (RTCHECK(ok_address(M, R))) {\\r
+ tchunkptr C0, C1;\\r
+ R->parent = XP;\\r
+ if ((C0 = X->child[0]) != 0) {\\r
+ if (RTCHECK(ok_address(M, C0))) {\\r
+ R->child[0] = C0;\\r
+ C0->parent = R;\\r
+ }\\r
+ else\\r
+ CORRUPTION_ERROR_ACTION(M);\\r
+ }\\r
+ if ((C1 = X->child[1]) != 0) {\\r
+ if (RTCHECK(ok_address(M, C1))) {\\r
+ R->child[1] = C1;\\r
+ C1->parent = R;\\r
+ }\\r
+ else\\r
+ CORRUPTION_ERROR_ACTION(M);\\r
+ }\\r
+ }\\r
+ else\\r
+ CORRUPTION_ERROR_ACTION(M);\\r
+ }\\r
+ }\\r
+}\r
+\r
+/* Relays to large vs small bin operations */\r
+\r
+#define insert_chunk(M, P, S)\\r
+ if (is_small(S)) insert_small_chunk(M, P, S)\\r
+ else { tchunkptr TP = (tchunkptr)(P); insert_large_chunk(M, TP, S); }\r
+\r
+#define unlink_chunk(M, P, S)\\r
+ if (is_small(S)) unlink_small_chunk(M, P, S)\\r
+ else { tchunkptr TP = (tchunkptr)(P); unlink_large_chunk(M, TP); }\r
+\r
+\r
+/* Relays to internal calls to malloc/free from realloc, memalign etc */\r
+\r
+#if ONLY_MSPACES\r
+#define internal_malloc(m, b) mspace_malloc(m, b)\r
+#define internal_free(m, mem) mspace_free(m,mem);\r
+#else /* ONLY_MSPACES */\r
+#if MSPACES\r
+#define internal_malloc(m, b)\\r
+ (m == gm)? dlmalloc(b) : mspace_malloc(m, b)\r
+#define internal_free(m, mem)\\r
+ if (m == gm) dlfree(mem); else mspace_free(m,mem);\r
+#else /* MSPACES */\r
+#define internal_malloc(m, b) dlmalloc(b)\r
+#define internal_free(m, mem) dlfree(mem)\r
+#endif /* MSPACES */\r
+#endif /* ONLY_MSPACES */\r
+\r
+/* ----------------------- Direct-mmapping chunks ----------------------- */\r
+\r
+/*\r
+ Directly mmapped chunks are set up with an offset to the start of\r
+ the mmapped region stored in the prev_foot field of the chunk. This\r
+ allows reconstruction of the required argument to MUNMAP when freed,\r
+ and also allows adjustment of the returned chunk to meet alignment\r
+ requirements (especially in memalign). There is also enough space\r
+ allocated to hold a fake next chunk of size SIZE_T_SIZE to maintain\r
+ the PINUSE bit so frees can be checked.\r
+*/\r
+\r
+/* Malloc using mmap */\r
+static void* mmap_alloc(mstate m, size_t nb) {\r
+ size_t mmsize = granularity_align(nb + SIX_SIZE_T_SIZES + CHUNK_ALIGN_MASK);\r
+ if (mmsize > nb) { /* Check for wrap around 0 */\r
+ char* mm = (char*)(DIRECT_MMAP(mmsize));\r
+ if (mm != CMFAIL) {\r
+ size_t offset = align_offset(chunk2mem(mm));\r
+ size_t psize = mmsize - offset - MMAP_FOOT_PAD;\r
+ mchunkptr p = (mchunkptr)(mm + offset);\r
+ p->prev_foot = offset | IS_MMAPPED_BIT;\r
+ (p)->head = (psize|CINUSE_BIT);\r
+ mark_inuse_foot(m, p, psize);\r
+ chunk_plus_offset(p, psize)->head = FENCEPOST_HEAD;\r
+ chunk_plus_offset(p, psize+SIZE_T_SIZE)->head = 0;\r
+\r
+ if (mm < m->least_addr)\r
+ m->least_addr = mm;\r
+ if ((m->footprint += mmsize) > m->max_footprint)\r
+ m->max_footprint = m->footprint;\r
+ assert(is_aligned(chunk2mem(p)));\r
+ check_mmapped_chunk(m, p);\r
+ return chunk2mem(p);\r
+ }\r
+ }\r
+ return 0;\r
+}\r
+\r
+/* Realloc using mmap */\r
+static mchunkptr mmap_resize(mstate m, mchunkptr oldp, size_t nb) {\r
+ size_t oldsize = chunksize(oldp);\r
+ if (is_small(nb)) /* Can't shrink mmap regions below small size */\r
+ return 0;\r
+ /* Keep old chunk if big enough but not too big */\r
+ if (oldsize >= nb + SIZE_T_SIZE &&\r
+ (oldsize - nb) <= (mparams.granularity << 1))\r
+ return oldp;\r
+ else {\r
+ size_t offset = oldp->prev_foot & ~IS_MMAPPED_BIT;\r
+ size_t oldmmsize = oldsize + offset + MMAP_FOOT_PAD;\r
+ size_t newmmsize = granularity_align(nb + SIX_SIZE_T_SIZES +\r
+ CHUNK_ALIGN_MASK);\r
+ char* cp = (char*)CALL_MREMAP((char*)oldp - offset,\r
+ oldmmsize, newmmsize, 1);\r
+ if (cp != CMFAIL) {\r
+ mchunkptr newp = (mchunkptr)(cp + offset);\r
+ size_t psize = newmmsize - offset - MMAP_FOOT_PAD;\r
+ newp->head = (psize|CINUSE_BIT);\r
+ mark_inuse_foot(m, newp, psize);\r
+ chunk_plus_offset(newp, psize)->head = FENCEPOST_HEAD;\r
+ chunk_plus_offset(newp, psize+SIZE_T_SIZE)->head = 0;\r
+\r
+ if (cp < m->least_addr)\r
+ m->least_addr = cp;\r
+ if ((m->footprint += newmmsize - oldmmsize) > m->max_footprint)\r
+ m->max_footprint = m->footprint;\r
+ check_mmapped_chunk(m, newp);\r
+ return newp;\r
+ }\r
+ }\r
+ return 0;\r
+}\r
+\r
+/* -------------------------- mspace management -------------------------- */\r
+\r
+/* Initialize top chunk and its size */\r
+static void init_top(mstate m, mchunkptr p, size_t psize) {\r
+ /* Ensure alignment */\r
+ size_t offset = align_offset(chunk2mem(p));\r
+ p = (mchunkptr)((char*)p + offset);\r
+ psize -= offset;\r
+\r
+ m->top = p;\r
+ m->topsize = psize;\r
+ p->head = psize | PINUSE_BIT;\r
+ /* set size of fake trailing chunk holding overhead space only once */\r
+ chunk_plus_offset(p, psize)->head = TOP_FOOT_SIZE;\r
+ m->trim_check = mparams.trim_threshold; /* reset on each update */\r
+}\r
+\r
+/* Initialize bins for a new mstate that is otherwise zeroed out */\r
+static void init_bins(mstate m) {\r
+ /* Establish circular links for smallbins */\r
+ bindex_t i;\r
+ for (i = 0; i < NSMALLBINS; ++i) {\r
+ sbinptr bin = smallbin_at(m,i);\r
+ bin->fd = bin->bk = bin;\r
+ }\r
+}\r
+\r
+#if PROCEED_ON_ERROR\r
+\r
+/* default corruption action */\r
+static void reset_on_error(mstate m) {\r
+ int i;\r
+ ++malloc_corruption_error_count;\r
+ /* Reinitialize fields to forget about all memory */\r
+ m->smallbins = m->treebins = 0;\r
+ m->dvsize = m->topsize = 0;\r
+ m->seg.base = 0;\r
+ m->seg.size = 0;\r
+ m->seg.next = 0;\r
+ m->top = m->dv = 0;\r
+ for (i = 0; i < NTREEBINS; ++i)\r
+ *treebin_at(m, i) = 0;\r
+ init_bins(m);\r
+}\r
+#endif /* PROCEED_ON_ERROR */\r
+\r
+/* Allocate chunk and prepend remainder with chunk in successor base. */\r
+static void* prepend_alloc(mstate m, char* newbase, char* oldbase,\r
+ size_t nb) {\r
+ mchunkptr p = align_as_chunk(newbase);\r
+ mchunkptr oldfirst = align_as_chunk(oldbase);\r
+ size_t psize = (char*)oldfirst - (char*)p;\r
+ mchunkptr q = chunk_plus_offset(p, nb);\r
+ size_t qsize = psize - nb;\r
+ set_size_and_pinuse_of_inuse_chunk(m, p, nb);\r
+\r
+ assert((char*)oldfirst > (char*)q);\r
+ assert(pinuse(oldfirst));\r
+ assert(qsize >= MIN_CHUNK_SIZE);\r
+\r
+ /* consolidate remainder with first chunk of old base */\r
+ if (oldfirst == m->top) {\r
+ size_t tsize = m->topsize += qsize;\r
+ m->top = q;\r
+ q->head = tsize | PINUSE_BIT;\r
+ check_top_chunk(m, q);\r
+ }\r
+ else if (oldfirst == m->dv) {\r
+ size_t dsize = m->dvsize += qsize;\r
+ m->dv = q;\r
+ set_size_and_pinuse_of_free_chunk(q, dsize);\r
+ }\r
+ else {\r
+ if (!cinuse(oldfirst)) {\r
+ size_t nsize = chunksize(oldfirst);\r
+ unlink_chunk(m, oldfirst, nsize);\r
+ oldfirst = chunk_plus_offset(oldfirst, nsize);\r
+ qsize += nsize;\r
+ }\r
+ set_free_with_pinuse(q, qsize, oldfirst);\r
+ insert_chunk(m, q, qsize);\r
+ check_free_chunk(m, q);\r
+ }\r
+\r
+ check_malloced_chunk(m, chunk2mem(p), nb);\r
+ return chunk2mem(p);\r
+}\r
+\r
+\r
+/* Add a segment to hold a new noncontiguous region */\r
+static void add_segment(mstate m, char* tbase, size_t tsize, flag_t mmapped) {\r
+ /* Determine locations and sizes of segment, fenceposts, old top */\r
+ char* old_top = (char*)m->top;\r
+ msegmentptr oldsp = segment_holding(m, old_top);\r
+ char* old_end = oldsp->base + oldsp->size;\r
+ size_t ssize = pad_request(sizeof(struct malloc_segment));\r
+ char* rawsp = old_end - (ssize + FOUR_SIZE_T_SIZES + CHUNK_ALIGN_MASK);\r
+ size_t offset = align_offset(chunk2mem(rawsp));\r
+ char* asp = rawsp + offset;\r
+ char* csp = (asp < (old_top + MIN_CHUNK_SIZE))? old_top : asp;\r
+ mchunkptr sp = (mchunkptr)csp;\r
+ msegmentptr ss = (msegmentptr)(chunk2mem(sp));\r
+ mchunkptr tnext = chunk_plus_offset(sp, ssize);\r
+ mchunkptr p = tnext;\r
+ int nfences = 0;\r
+\r
+ /* reset top to new space */\r
+ init_top(m, (mchunkptr)tbase, tsize - TOP_FOOT_SIZE);\r
+\r
+ /* Set up segment record */\r
+ assert(is_aligned(ss));\r
+ set_size_and_pinuse_of_inuse_chunk(m, sp, ssize);\r
+ *ss = m->seg; /* Push current record */\r
+ m->seg.base = tbase;\r
+ m->seg.size = tsize;\r
+ m->seg.sflags = mmapped;\r
+ m->seg.next = ss;\r
+\r
+ /* Insert trailing fenceposts */\r
+ for (;;) {\r
+ mchunkptr nextp = chunk_plus_offset(p, SIZE_T_SIZE);\r
+ p->head = FENCEPOST_HEAD;\r
+ ++nfences;\r
+ if ((char*)(&(nextp->head)) < old_end)\r
+ p = nextp;\r
+ else\r
+ break;\r
+ }\r
+ assert(nfences >= 2);\r
+\r
+ /* Insert the rest of old top into a bin as an ordinary free chunk */\r
+ if (csp != old_top) {\r
+ mchunkptr q = (mchunkptr)old_top;\r
+ size_t psize = csp - old_top;\r
+ mchunkptr tn = chunk_plus_offset(q, psize);\r
+ set_free_with_pinuse(q, psize, tn);\r
+ insert_chunk(m, q, psize);\r
+ }\r
+\r
+ check_top_chunk(m, m->top);\r
+}\r
+\r
+/* -------------------------- System allocation -------------------------- */\r
+\r
+/* Get memory from system using MORECORE or MMAP */\r
+static void* sys_alloc(mstate m, size_t nb) {\r
+ char* tbase = CMFAIL;\r
+ size_t tsize = 0;\r
+ flag_t mmap_flag = 0;\r
+\r
+ init_mparams();\r
+\r
+ /* Directly map large chunks */\r
+ if (use_mmap(m) && nb >= mparams.mmap_threshold) {\r
+ void* mem = mmap_alloc(m, nb);\r
+ if (mem != 0)\r
+ return mem;\r
+ }\r
+\r
+ /*\r
+ Try getting memory in any of three ways (in most-preferred to\r
+ least-preferred order):\r
+ 1. A call to MORECORE that can normally contiguously extend memory.\r
+ (disabled if not MORECORE_CONTIGUOUS or not HAVE_MORECORE or\r
+ or main space is mmapped or a previous contiguous call failed)\r
+ 2. A call to MMAP new space (disabled if not HAVE_MMAP).\r
+ Note that under the default settings, if MORECORE is unable to\r
+ fulfill a request, and HAVE_MMAP is true, then mmap is\r
+ used as a noncontiguous system allocator. This is a useful backup\r
+ strategy for systems with holes in address spaces -- in this case\r
+ sbrk cannot contiguously expand the heap, but mmap may be able to\r
+ find space.\r
+ 3. A call to MORECORE that cannot usually contiguously extend memory.\r
+ (disabled if not HAVE_MORECORE)\r
+ */\r
+\r
+ if (MORECORE_CONTIGUOUS && !use_noncontiguous(m)) {\r
+ char* br = CMFAIL;\r
+ msegmentptr ss = (m->top == 0)? 0 : segment_holding(m, (char*)m->top);\r
+ size_t asize = 0;\r
+ ACQUIRE_MORECORE_LOCK();\r
+\r
+ if (ss == 0) { /* First time through or recovery */\r
+ char* base = (char*)CALL_MORECORE(0);\r
+ if (base != CMFAIL) {\r
+ asize = granularity_align(nb + TOP_FOOT_SIZE + SIZE_T_ONE);\r
+ /* Adjust to end on a page boundary */\r
+ if (!is_page_aligned(base))\r
+ asize += (page_align((size_t)base) - (size_t)base);\r
+ /* Can't call MORECORE if size is negative when treated as signed */\r
+ if (asize < HALF_MAX_SIZE_T &&\r
+ (br = (char*)(CALL_MORECORE(asize))) == base) {\r
+ tbase = base;\r
+ tsize = asize;\r
+ }\r
+ }\r
+ }\r
+ else {\r
+ /* Subtract out existing available top space from MORECORE request. */\r
+ asize = granularity_align(nb - m->topsize + TOP_FOOT_SIZE + SIZE_T_ONE);\r
+ /* Use mem here only if it did continuously extend old space */\r
+ if (asize < HALF_MAX_SIZE_T &&\r
+ (br = (char*)(CALL_MORECORE(asize))) == ss->base+ss->size) {\r
+ tbase = br;\r
+ tsize = asize;\r
+ }\r
+ }\r
+\r
+ if (tbase == CMFAIL) { /* Cope with partial failure */\r
+ if (br != CMFAIL) { /* Try to use/extend the space we did get */\r
+ if (asize < HALF_MAX_SIZE_T &&\r
+ asize < nb + TOP_FOOT_SIZE + SIZE_T_ONE) {\r
+ size_t esize = granularity_align(nb + TOP_FOOT_SIZE + SIZE_T_ONE - asize);\r
+ if (esize < HALF_MAX_SIZE_T) {\r
+ char* end = (char*)CALL_MORECORE(esize);\r
+ if (end != CMFAIL)\r
+ asize += esize;\r
+ else { /* Can't use; try to release */\r
+ (void) CALL_MORECORE(-asize);\r
+ br = CMFAIL;\r
+ }\r
+ }\r
+ }\r
+ }\r
+ if (br != CMFAIL) { /* Use the space we did get */\r
+ tbase = br;\r
+ tsize = asize;\r
+ }\r
+ else\r
+ disable_contiguous(m); /* Don't try contiguous path in the future */\r
+ }\r
+\r
+ RELEASE_MORECORE_LOCK();\r
+ }\r
+\r
+ if (HAVE_MMAP && tbase == CMFAIL) { /* Try MMAP */\r
+ size_t req = nb + TOP_FOOT_SIZE + SIZE_T_ONE;\r
+ size_t rsize = granularity_align(req);\r
+ if (rsize > nb) { /* Fail if wraps around zero */\r
+ char* mp = (char*)(CALL_MMAP(rsize));\r
+ if (mp != CMFAIL) {\r
+ tbase = mp;\r
+ tsize = rsize;\r
+ mmap_flag = IS_MMAPPED_BIT;\r
+ }\r
+ }\r
+ }\r
+\r
+ if (HAVE_MORECORE && tbase == CMFAIL) { /* Try noncontiguous MORECORE */\r
+ size_t asize = granularity_align(nb + TOP_FOOT_SIZE + SIZE_T_ONE);\r
+ if (asize < HALF_MAX_SIZE_T) {\r
+ char* br = CMFAIL;\r
+ char* end = CMFAIL;\r
+ ACQUIRE_MORECORE_LOCK();\r
+ br = (char*)(CALL_MORECORE(asize));\r
+ end = (char*)(CALL_MORECORE(0));\r
+ RELEASE_MORECORE_LOCK();\r
+ if (br != CMFAIL && end != CMFAIL && br < end) {\r
+ size_t ssize = end - br;\r
+ if (ssize > nb + TOP_FOOT_SIZE) {\r
+ tbase = br;\r
+ tsize = ssize;\r
+ }\r
+ }\r
+ }\r
+ }\r
+\r
+ if (tbase != CMFAIL) {\r
+\r
+ if ((m->footprint += tsize) > m->max_footprint)\r
+ m->max_footprint = m->footprint;\r
+\r
+ if (!is_initialized(m)) { /* first-time initialization */\r
+ m->seg.base = m->least_addr = tbase;\r
+ m->seg.size = tsize;\r
+ m->seg.sflags = mmap_flag;\r
+ m->magic = mparams.magic;\r
+ init_bins(m);\r
+ if (is_global(m)) \r
+ init_top(m, (mchunkptr)tbase, tsize - TOP_FOOT_SIZE);\r
+ else {\r
+ /* Offset top by embedded malloc_state */\r
+ mchunkptr mn = next_chunk(mem2chunk(m));\r
+ init_top(m, mn, (size_t)((tbase + tsize) - (char*)mn) -TOP_FOOT_SIZE);\r
+ }\r
+ }\r
+\r
+ else {\r
+ /* Try to merge with an existing segment */\r
+#ifdef DONT_MERGE_SEGMENTS\r
+ (void) prepend_alloc;\r
+ add_segment(m, tbase, tsize, mmap_flag);\r
+#else\r
+ msegmentptr sp = &m->seg;\r
+ while (sp != 0 && tbase != sp->base + sp->size)\r
+ sp = sp->next;\r
+ if (sp != 0 &&\r
+ !is_extern_segment(sp) &&\r
+ (sp->sflags & IS_MMAPPED_BIT) == mmap_flag &&\r
+ segment_holds(sp, m->top)) { /* append */\r
+ sp->size += tsize;\r
+ init_top(m, m->top, m->topsize + tsize);\r
+ }\r
+ else {\r
+ if (tbase < m->least_addr)\r
+ m->least_addr = tbase;\r
+ sp = &m->seg;\r
+ while (sp != 0 && sp->base != tbase + tsize)\r
+ sp = sp->next;\r
+ if (sp != 0 &&\r
+ !is_extern_segment(sp) &&\r
+ (sp->sflags & IS_MMAPPED_BIT) == mmap_flag) {\r
+ char* oldbase = sp->base;\r
+ sp->base = tbase;\r
+ sp->size += tsize;\r
+ return prepend_alloc(m, tbase, oldbase, nb);\r
+ }\r
+ else\r
+ add_segment(m, tbase, tsize, mmap_flag);\r
+ }\r
+#endif /* DONT_MERGE_SEGMENTS */\r
+ }\r
+\r
+ if (nb < m->topsize) { /* Allocate from new or extended top space */\r
+ size_t rsize = m->topsize -= nb;\r
+ mchunkptr p = m->top;\r
+ mchunkptr r = m->top = chunk_plus_offset(p, nb);\r
+ r->head = rsize | PINUSE_BIT;\r
+ set_size_and_pinuse_of_inuse_chunk(m, p, nb);\r
+ check_top_chunk(m, m->top);\r
+ check_malloced_chunk(m, chunk2mem(p), nb);\r
+ return chunk2mem(p);\r
+ }\r
+ }\r
+\r
+ MALLOC_FAILURE_ACTION;\r
+ return 0;\r
+}\r
+\r
+/* ----------------------- system deallocation -------------------------- */\r
+\r
+/* Unmap and unlink any mmapped segments that don't contain used chunks */\r
+static size_t release_unused_segments(mstate m) {\r
+ size_t released = 0;\r
+ msegmentptr pred = &m->seg;\r
+ msegmentptr sp = pred->next;\r
+ while (sp != 0) {\r
+ char* base = sp->base;\r
+ size_t size = sp->size;\r
+ msegmentptr next = sp->next;\r
+ if (is_mmapped_segment(sp) && !is_extern_segment(sp)) {\r
+ mchunkptr p = align_as_chunk(base);\r
+ size_t psize = chunksize(p);\r
+ /* Can unmap if first chunk holds entire segment and not pinned */\r
+ if (!cinuse(p) && (char*)p + psize >= base + size - TOP_FOOT_SIZE) {\r
+ tchunkptr tp = (tchunkptr)p;\r
+ assert(segment_holds(sp, (char*)sp));\r
+ if (p == m->dv) {\r
+ m->dv = 0;\r
+ m->dvsize = 0;\r
+ }\r
+ else {\r
+ unlink_large_chunk(m, tp);\r
+ }\r
+ if (CALL_MUNMAP(base, size) == 0) {\r
+ released += size;\r
+ m->footprint -= size;\r
+ /* unlink obsoleted record */\r
+ sp = pred;\r
+ sp->next = next;\r
+ }\r
+ else { /* back out if cannot unmap */\r
+ insert_large_chunk(m, tp, psize);\r
+ }\r
+ }\r
+ }\r
+ pred = sp;\r
+ sp = next;\r
+ }\r
+ return released;\r
+}\r
+\r
+static int sys_trim(mstate m, size_t pad) {\r
+ size_t released = 0;\r
+ if (pad < MAX_REQUEST && is_initialized(m)) {\r
+ pad += TOP_FOOT_SIZE; /* ensure enough room for segment overhead */\r
+\r
+ if (m->topsize > pad) {\r
+ /* Shrink top space in granularity-size units, keeping at least one */\r
+ size_t unit = mparams.granularity;\r
+ size_t extra = ((m->topsize - pad + (unit - SIZE_T_ONE)) / unit -\r
+ SIZE_T_ONE) * unit;\r
+ msegmentptr sp = segment_holding(m, (char*)m->top);\r
+\r
+ if (!is_extern_segment(sp)) {\r
+ if (is_mmapped_segment(sp)) {\r
+ if (HAVE_MMAP &&\r
+ sp->size >= extra &&\r
+ !has_segment_link(m, sp)) { /* can't shrink if pinned */\r
+ size_t newsize = sp->size - extra;\r
+ /* Prefer mremap, fall back to munmap */\r
+ if ((CALL_MREMAP(sp->base, sp->size, newsize, 0) != MFAIL) ||\r
+ (CALL_MUNMAP(sp->base + newsize, extra) == 0)) {\r
+ released = extra;\r
+ }\r
+ }\r
+ }\r
+ else if (HAVE_MORECORE) {\r
+ if (extra >= HALF_MAX_SIZE_T) /* Avoid wrapping negative */\r
+ extra = (HALF_MAX_SIZE_T) + SIZE_T_ONE - unit;\r
+ ACQUIRE_MORECORE_LOCK();\r
+ {\r
+ /* Make sure end of memory is where we last set it. */\r
+ char* old_br = (char*)(CALL_MORECORE(0));\r
+ if (old_br == sp->base + sp->size) {\r
+ char* rel_br = (char*)(CALL_MORECORE(-extra));\r
+ char* new_br = (char*)(CALL_MORECORE(0));\r
+ if (rel_br != CMFAIL && new_br < old_br)\r
+ released = old_br - new_br;\r
+ }\r
+ }\r
+ RELEASE_MORECORE_LOCK();\r
+ }\r
+ }\r
+\r
+ if (released != 0) {\r
+ sp->size -= released;\r
+ m->footprint -= released;\r
+ init_top(m, m->top, m->topsize - released);\r
+ check_top_chunk(m, m->top);\r
+ }\r
+ }\r
+\r
+ /* Unmap any unused mmapped segments */\r
+ if (HAVE_MMAP) \r
+ released += release_unused_segments(m);\r
+\r
+ /* On failure, disable autotrim to avoid repeated failed future calls */\r
+ if (released == 0)\r
+ m->trim_check = MAX_SIZE_T;\r
+ }\r
+\r
+ return (released != 0)? 1 : 0;\r
+}\r
+\r
+/* ---------------------------- malloc support --------------------------- */\r
+\r
+/* allocate a large request from the best fitting chunk in a treebin */\r
+static void* tmalloc_large(mstate m, size_t nb) {\r
+ tchunkptr v = 0;\r
+ size_t rsize = -nb; /* Unsigned negation */\r
+ tchunkptr t;\r
+ bindex_t idx;\r
+ compute_tree_index(nb, idx);\r
+\r
+ if ((t = *treebin_at(m, idx)) != 0) {\r
+ /* Traverse tree for this bin looking for node with size == nb */\r
+ size_t sizebits = nb << leftshift_for_tree_index(idx);\r
+ tchunkptr rst = 0; /* The deepest untaken right subtree */\r
+ for (;;) {\r
+ tchunkptr rt;\r
+ size_t trem = chunksize(t) - nb;\r
+ if (trem < rsize) {\r
+ v = t;\r
+ if ((rsize = trem) == 0)\r
+ break;\r
+ }\r
+ rt = t->child[1];\r
+ t = t->child[(sizebits >> (SIZE_T_BITSIZE-SIZE_T_ONE)) & 1];\r
+ if (rt != 0 && rt != t)\r
+ rst = rt;\r
+ if (t == 0) {\r
+ t = rst; /* set t to least subtree holding sizes > nb */\r
+ break;\r
+ }\r
+ sizebits <<= 1;\r
+ }\r
+ }\r
+\r
+ if (t == 0 && v == 0) { /* set t to root of next non-empty treebin */\r
+ binmap_t leftbits = left_bits(idx2bit(idx)) & m->treemap;\r
+ if (leftbits != 0) {\r
+ bindex_t i;\r
+ binmap_t leastbit = least_bit(leftbits);\r
+ compute_bit2idx(leastbit, i);\r
+ t = *treebin_at(m, i);\r
+ }\r
+ }\r
+\r
+ while (t != 0) { /* find smallest of tree or subtree */\r
+ size_t trem = chunksize(t) - nb;\r
+ if (trem < rsize) {\r
+ rsize = trem;\r
+ v = t;\r
+ }\r
+ t = leftmost_child(t);\r
+ }\r
+\r
+ /* If dv is a better fit, return 0 so malloc will use it */\r
+ if (v != 0 && rsize < (size_t)(m->dvsize - nb)) {\r
+ if (RTCHECK(ok_address(m, v))) { /* split */\r
+ mchunkptr r = chunk_plus_offset(v, nb);\r
+ assert(chunksize(v) == rsize + nb);\r
+ if (RTCHECK(ok_next(v, r))) {\r
+ unlink_large_chunk(m, v);\r
+ if (rsize < MIN_CHUNK_SIZE)\r
+ set_inuse_and_pinuse(m, v, (rsize + nb));\r
+ else {\r
+ set_size_and_pinuse_of_inuse_chunk(m, v, nb);\r
+ set_size_and_pinuse_of_free_chunk(r, rsize);\r
+ insert_chunk(m, r, rsize);\r
+ }\r
+ return chunk2mem(v);\r
+ }\r
+ }\r
+ CORRUPTION_ERROR_ACTION(m);\r
+ }\r
+ return 0;\r
+}\r
+\r
+/* allocate a small request from the best fitting chunk in a treebin */\r
+static void* tmalloc_small(mstate m, size_t nb) {\r
+ tchunkptr t, v;\r
+ size_t rsize;\r
+ bindex_t i;\r
+ binmap_t leastbit = least_bit(m->treemap);\r
+ compute_bit2idx(leastbit, i);\r
+\r
+ v = t = *treebin_at(m, i);\r
+ rsize = chunksize(t) - nb;\r
+\r
+ while ((t = leftmost_child(t)) != 0) {\r
+ size_t trem = chunksize(t) - nb;\r
+ if (trem < rsize) {\r
+ rsize = trem;\r
+ v = t;\r
+ }\r
+ }\r
+\r
+ if (RTCHECK(ok_address(m, v))) {\r
+ mchunkptr r = chunk_plus_offset(v, nb);\r
+ assert(chunksize(v) == rsize + nb);\r
+ if (RTCHECK(ok_next(v, r))) {\r
+ unlink_large_chunk(m, v);\r
+ if (rsize < MIN_CHUNK_SIZE)\r
+ set_inuse_and_pinuse(m, v, (rsize + nb));\r
+ else {\r
+ set_size_and_pinuse_of_inuse_chunk(m, v, nb);\r
+ set_size_and_pinuse_of_free_chunk(r, rsize);\r
+ replace_dv(m, r, rsize);\r
+ }\r
+ return chunk2mem(v);\r
+ }\r
+ }\r
+\r
+ CORRUPTION_ERROR_ACTION(m);\r
+ return 0;\r
+}\r
+\r
+/* --------------------------- realloc support --------------------------- */\r
+\r
+static void* internal_realloc(mstate m, void* oldmem, size_t bytes) {\r
+ if (bytes >= MAX_REQUEST) {\r
+ MALLOC_FAILURE_ACTION;\r
+ return 0;\r
+ }\r
+ if (!PREACTION(m)) {\r
+ mchunkptr oldp = mem2chunk(oldmem);\r
+ size_t oldsize = chunksize(oldp);\r
+ mchunkptr next = chunk_plus_offset(oldp, oldsize);\r
+ mchunkptr newp = 0;\r
+ void* extra = 0;\r
+\r
+ /* Try to either shrink or extend into top. Else malloc-copy-free */\r
+\r
+ if (RTCHECK(ok_address(m, oldp) && ok_cinuse(oldp) &&\r
+ ok_next(oldp, next) && ok_pinuse(next))) {\r
+ size_t nb = request2size(bytes);\r
+ if (is_mmapped(oldp))\r
+ newp = mmap_resize(m, oldp, nb);\r
+ else if (oldsize >= nb) { /* already big enough */\r
+ size_t rsize = oldsize - nb;\r
+ newp = oldp;\r
+ if (rsize >= MIN_CHUNK_SIZE) {\r
+ mchunkptr remainder = chunk_plus_offset(newp, nb);\r
+ set_inuse(m, newp, nb);\r
+ set_inuse(m, remainder, rsize);\r
+ extra = chunk2mem(remainder);\r
+ }\r
+ }\r
+ else if (next == m->top && oldsize + m->topsize > nb) {\r
+ /* Expand into top */\r
+ size_t newsize = oldsize + m->topsize;\r
+ size_t newtopsize = newsize - nb;\r
+ mchunkptr newtop = chunk_plus_offset(oldp, nb);\r
+ set_inuse(m, oldp, nb);\r
+ newtop->head = newtopsize |PINUSE_BIT;\r
+ m->top = newtop;\r
+ m->topsize = newtopsize;\r
+ newp = oldp;\r
+ }\r
+ }\r
+ else {\r
+ USAGE_ERROR_ACTION(m, oldmem);\r
+ POSTACTION(m);\r
+ return 0;\r
+ }\r
+\r
+ POSTACTION(m);\r
+\r
+ if (newp != 0) {\r
+ if (extra != 0) {\r
+ internal_free(m, extra);\r
+ }\r
+ check_inuse_chunk(m, newp);\r
+ return chunk2mem(newp);\r
+ }\r
+ else {\r
+ void* newmem = internal_malloc(m, bytes);\r
+ if (newmem != 0) {\r
+ size_t oc = oldsize - overhead_for(oldp);\r
+ memcpy(newmem, oldmem, (oc < bytes)? oc : bytes);\r
+ internal_free(m, oldmem);\r
+ }\r
+ return newmem;\r
+ }\r
+ }\r
+ return 0;\r
+}\r
+\r
+/* --------------------------- memalign support -------------------------- */\r
+\r
+static void* internal_memalign(mstate m, size_t alignment, size_t bytes) {\r
+ if (alignment <= MALLOC_ALIGNMENT) /* Can just use malloc */\r
+ return internal_malloc(m, bytes);\r
+ if (alignment < MIN_CHUNK_SIZE) /* must be at least a minimum chunk size */\r
+ alignment = MIN_CHUNK_SIZE;\r
+ if ((alignment & (alignment-SIZE_T_ONE)) != 0) {/* Ensure a power of 2 */\r
+ size_t a = MALLOC_ALIGNMENT << 1;\r
+ while (a < alignment) a <<= 1;\r
+ alignment = a;\r
+ }\r
+ \r
+ if (bytes >= MAX_REQUEST - alignment) {\r
+ if (m != 0) { /* Test isn't needed but avoids compiler warning */\r
+ MALLOC_FAILURE_ACTION;\r
+ }\r
+ }\r
+ else {\r
+ size_t nb = request2size(bytes);\r
+ size_t req = nb + alignment + MIN_CHUNK_SIZE - CHUNK_OVERHEAD;\r
+ char* mem = (char*)internal_malloc(m, req);\r
+ if (mem != 0) {\r
+ void* leader = 0;\r
+ void* trailer = 0;\r
+ mchunkptr p = mem2chunk(mem);\r
+\r
+ if (PREACTION(m)) return 0;\r
+ if ((((size_t)(mem)) % alignment) != 0) { /* misaligned */\r
+ /*\r
+ Find an aligned spot inside chunk. Since we need to give\r
+ back leading space in a chunk of at least MIN_CHUNK_SIZE, if\r
+ the first calculation places us at a spot with less than\r
+ MIN_CHUNK_SIZE leader, we can move to the next aligned spot.\r
+ We've allocated enough total room so that this is always\r
+ possible.\r
+ */\r
+ char* br = (char*)mem2chunk((size_t)(((size_t)(mem +\r
+ alignment -\r
+ SIZE_T_ONE)) &\r
+ -alignment));\r
+ char* pos = ((size_t)(br - (char*)(p)) >= MIN_CHUNK_SIZE)?\r
+ br : br+alignment;\r
+ mchunkptr newp = (mchunkptr)pos;\r
+ size_t leadsize = pos - (char*)(p);\r
+ size_t newsize = chunksize(p) - leadsize;\r
+\r
+ if (is_mmapped(p)) { /* For mmapped chunks, just adjust offset */\r
+ newp->prev_foot = p->prev_foot + leadsize;\r
+ newp->head = (newsize|CINUSE_BIT);\r
+ }\r
+ else { /* Otherwise, give back leader, use the rest */\r
+ set_inuse(m, newp, newsize);\r
+ set_inuse(m, p, leadsize);\r
+ leader = chunk2mem(p);\r
+ }\r
+ p = newp;\r
+ }\r
+\r
+ /* Give back spare room at the end */\r
+ if (!is_mmapped(p)) {\r
+ size_t size = chunksize(p);\r
+ if (size > nb + MIN_CHUNK_SIZE) {\r
+ size_t remainder_size = size - nb;\r
+ mchunkptr remainder = chunk_plus_offset(p, nb);\r
+ set_inuse(m, p, nb);\r
+ set_inuse(m, remainder, remainder_size);\r
+ trailer = chunk2mem(remainder);\r
+ }\r
+ }\r
+\r
+ assert (chunksize(p) >= nb);\r
+ assert((((size_t)(chunk2mem(p))) % alignment) == 0);\r
+ check_inuse_chunk(m, p);\r
+ POSTACTION(m);\r
+ if (leader != 0) {\r
+ internal_free(m, leader);\r
+ }\r
+ if (trailer != 0) {\r
+ internal_free(m, trailer);\r
+ }\r
+ return chunk2mem(p);\r
+ }\r
+ }\r
+ return 0;\r
+}\r
+\r
+/* ------------------------ comalloc/coalloc support --------------------- */\r
+\r
+static void** ialloc(mstate m,\r
+ size_t n_elements,\r
+ size_t* sizes,\r
+ int opts,\r
+ void* chunks[]) {\r
+ /*\r
+ This provides common support for independent_X routines, handling\r
+ all of the combinations that can result.\r
+\r
+ The opts arg has:\r
+ bit 0 set if all elements are same size (using sizes[0])\r
+ bit 1 set if elements should be zeroed\r
+ */\r
+\r
+ size_t element_size; /* chunksize of each element, if all same */\r
+ size_t contents_size; /* total size of elements */\r
+ size_t array_size; /* request size of pointer array */\r
+ void* mem; /* malloced aggregate space */\r
+ mchunkptr p; /* corresponding chunk */\r
+ size_t remainder_size; /* remaining bytes while splitting */\r
+ void** marray; /* either "chunks" or malloced ptr array */\r
+ mchunkptr array_chunk; /* chunk for malloced ptr array */\r
+ flag_t was_enabled; /* to disable mmap */\r
+ size_t size;\r
+ size_t i;\r
+\r
+ /* compute array length, if needed */\r
+ if (chunks != 0) {\r
+ if (n_elements == 0)\r
+ return chunks; /* nothing to do */\r
+ marray = chunks;\r
+ array_size = 0;\r
+ }\r
+ else {\r
+ /* if empty req, must still return chunk representing empty array */\r
+ if (n_elements == 0)\r
+ return (void**)internal_malloc(m, 0);\r
+ marray = 0;\r
+ array_size = request2size(n_elements * (sizeof(void*)));\r
+ }\r
+\r
+ /* compute total element size */\r
+ if (opts & 0x1) { /* all-same-size */\r
+ element_size = request2size(*sizes);\r
+ contents_size = n_elements * element_size;\r
+ }\r
+ else { /* add up all the sizes */\r
+ element_size = 0;\r
+ contents_size = 0;\r
+ for (i = 0; i != n_elements; ++i)\r
+ contents_size += request2size(sizes[i]);\r
+ }\r
+\r
+ size = contents_size + array_size;\r
+\r
+ /*\r
+ Allocate the aggregate chunk. First disable direct-mmapping so\r
+ malloc won't use it, since we would not be able to later\r
+ free/realloc space internal to a segregated mmap region.\r
+ */\r
+ was_enabled = use_mmap(m);\r
+ disable_mmap(m);\r
+ mem = internal_malloc(m, size - CHUNK_OVERHEAD);\r
+ if (was_enabled)\r
+ enable_mmap(m);\r
+ if (mem == 0)\r
+ return 0;\r
+\r
+ if (PREACTION(m)) return 0;\r
+ p = mem2chunk(mem);\r
+ remainder_size = chunksize(p);\r
+\r
+ assert(!is_mmapped(p));\r
+\r
+ if (opts & 0x2) { /* optionally clear the elements */\r
+ memset((size_t*)mem, 0, remainder_size - SIZE_T_SIZE - array_size);\r
+ }\r
+\r
+ /* If not provided, allocate the pointer array as final part of chunk */\r
+ if (marray == 0) {\r
+ size_t array_chunk_size;\r
+ array_chunk = chunk_plus_offset(p, contents_size);\r
+ array_chunk_size = remainder_size - contents_size;\r
+ marray = (void**) (chunk2mem(array_chunk));\r
+ set_size_and_pinuse_of_inuse_chunk(m, array_chunk, array_chunk_size);\r
+ remainder_size = contents_size;\r
+ }\r
+\r
+ /* split out elements */\r
+ for (i = 0; ; ++i) {\r
+ marray[i] = chunk2mem(p);\r
+ if (i != n_elements-1) {\r
+ if (element_size != 0)\r
+ size = element_size;\r
+ else\r
+ size = request2size(sizes[i]);\r
+ remainder_size -= size;\r
+ set_size_and_pinuse_of_inuse_chunk(m, p, size);\r
+ p = chunk_plus_offset(p, size);\r
+ }\r
+ else { /* the final element absorbs any overallocation slop */\r
+ set_size_and_pinuse_of_inuse_chunk(m, p, remainder_size);\r
+ break;\r
+ }\r
+ }\r
+\r
+#if DEBUG\r
+ if (marray != chunks) {\r
+ /* final element must have exactly exhausted chunk */\r
+ if (element_size != 0) {\r
+ assert(remainder_size == element_size);\r
+ }\r
+ else {\r
+ assert(remainder_size == request2size(sizes[i]));\r
+ }\r
+ check_inuse_chunk(m, mem2chunk(marray));\r
+ }\r
+ for (i = 0; i != n_elements; ++i)\r
+ check_inuse_chunk(m, mem2chunk(marray[i]));\r
+\r
+#endif /* DEBUG */\r
+\r
+ POSTACTION(m);\r
+ return marray;\r
+}\r
+\r
+\r
+/* -------------------------- public routines ---------------------------- */\r
+\r
+#if !ONLY_MSPACES\r
+\r
+void* dlmalloc(size_t bytes) {\r
+ /*\r
+ Basic algorithm:\r
+ If a small request (< 256 bytes minus per-chunk overhead):\r
+ 1. If one exists, use a remainderless chunk in associated smallbin.\r
+ (Remainderless means that there are too few excess bytes to\r
+ represent as a chunk.)\r
+ 2. If it is big enough, use the dv chunk, which is normally the\r
+ chunk adjacent to the one used for the most recent small request.\r
+ 3. If one exists, split the smallest available chunk in a bin,\r
+ saving remainder in dv.\r
+ 4. If it is big enough, use the top chunk.\r
+ 5. If available, get memory from system and use it\r
+ Otherwise, for a large request:\r
+ 1. Find the smallest available binned chunk that fits, and use it\r
+ if it is better fitting than dv chunk, splitting if necessary.\r
+ 2. If better fitting than any binned chunk, use the dv chunk.\r
+ 3. If it is big enough, use the top chunk.\r
+ 4. If request size >= mmap threshold, try to directly mmap this chunk.\r
+ 5. If available, get memory from system and use it\r
+\r
+ The ugly goto's here ensure that postaction occurs along all paths.\r
+ */\r
+\r
+ if (!PREACTION(gm)) {\r
+ void* mem;\r
+ size_t nb;\r
+ if (bytes <= MAX_SMALL_REQUEST) {\r
+ bindex_t idx;\r
+ binmap_t smallbits;\r
+ nb = (bytes < MIN_REQUEST)? MIN_CHUNK_SIZE : pad_request(bytes);\r
+ idx = small_index(nb);\r
+ smallbits = gm->smallmap >> idx;\r
+\r
+ if ((smallbits & 0x3U) != 0) { /* Remainderless fit to a smallbin. */\r
+ mchunkptr b, p;\r
+ idx += ~smallbits & 1; /* Uses next bin if idx empty */\r
+ b = smallbin_at(gm, idx);\r
+ p = b->fd;\r
+ assert(chunksize(p) == small_index2size(idx));\r
+ unlink_first_small_chunk(gm, b, p, idx);\r
+ set_inuse_and_pinuse(gm, p, small_index2size(idx));\r
+ mem = chunk2mem(p);\r
+ check_malloced_chunk(gm, mem, nb);\r
+ goto postaction;\r
+ }\r
+\r
+ else if (nb > gm->dvsize) {\r
+ if (smallbits != 0) { /* Use chunk in next nonempty smallbin */\r
+ mchunkptr b, p, r;\r
+ size_t rsize;\r
+ bindex_t i;\r
+ binmap_t leftbits = (smallbits << idx) & left_bits(idx2bit(idx));\r
+ binmap_t leastbit = least_bit(leftbits);\r
+ compute_bit2idx(leastbit, i);\r
+ b = smallbin_at(gm, i);\r
+ p = b->fd;\r
+ assert(chunksize(p) == small_index2size(i));\r
+ unlink_first_small_chunk(gm, b, p, i);\r
+ rsize = small_index2size(i) - nb;\r
+ /* Fit here cannot be remainderless if 4byte sizes */\r
+ if (SIZE_T_SIZE != 4 && rsize < MIN_CHUNK_SIZE)\r
+ set_inuse_and_pinuse(gm, p, small_index2size(i));\r
+ else {\r
+ set_size_and_pinuse_of_inuse_chunk(gm, p, nb);\r
+ r = chunk_plus_offset(p, nb);\r
+ set_size_and_pinuse_of_free_chunk(r, rsize);\r
+ replace_dv(gm, r, rsize);\r
+ }\r
+ mem = chunk2mem(p);\r
+ check_malloced_chunk(gm, mem, nb);\r
+ goto postaction;\r
+ }\r
+\r
+ else if (gm->treemap != 0 && (mem = tmalloc_small(gm, nb)) != 0) {\r
+ check_malloced_chunk(gm, mem, nb);\r
+ goto postaction;\r
+ }\r
+ }\r
+ }\r
+ else if (bytes >= MAX_REQUEST)\r
+ nb = MAX_SIZE_T; /* Too big to allocate. Force failure (in sys alloc) */\r
+ else {\r
+ nb = pad_request(bytes);\r
+ if (gm->treemap != 0 && (mem = tmalloc_large(gm, nb)) != 0) {\r
+ check_malloced_chunk(gm, mem, nb);\r
+ goto postaction;\r
+ }\r
+ }\r
+\r
+ if (nb <= gm->dvsize) {\r
+ size_t rsize = gm->dvsize - nb;\r
+ mchunkptr p = gm->dv;\r
+ if (rsize >= MIN_CHUNK_SIZE) { /* split dv */\r
+ mchunkptr r = gm->dv = chunk_plus_offset(p, nb);\r
+ gm->dvsize = rsize;\r
+ set_size_and_pinuse_of_free_chunk(r, rsize);\r
+ set_size_and_pinuse_of_inuse_chunk(gm, p, nb);\r
+ }\r
+ else { /* exhaust dv */\r
+ size_t dvs = gm->dvsize;\r
+ gm->dvsize = 0;\r
+ gm->dv = 0;\r
+ set_inuse_and_pinuse(gm, p, dvs);\r
+ }\r
+ mem = chunk2mem(p);\r
+ check_malloced_chunk(gm, mem, nb);\r
+ goto postaction;\r
+ }\r
+\r
+ else if (nb < gm->topsize) { /* Split top */\r
+ size_t rsize = gm->topsize -= nb;\r
+ mchunkptr p = gm->top;\r
+ mchunkptr r = gm->top = chunk_plus_offset(p, nb);\r
+ r->head = rsize | PINUSE_BIT;\r
+ set_size_and_pinuse_of_inuse_chunk(gm, p, nb);\r
+ mem = chunk2mem(p);\r
+ check_top_chunk(gm, gm->top);\r
+ check_malloced_chunk(gm, mem, nb);\r
+ goto postaction;\r
+ }\r
+\r
+ mem = sys_alloc(gm, nb);\r
+\r
+ postaction:\r
+ POSTACTION(gm);\r
+ return mem;\r
+ }\r
+\r
+ return 0;\r
+}\r
+\r
+void dlfree(void* mem) {\r
+ /*\r
+ Consolidate freed chunks with preceeding or succeeding bordering\r
+ free chunks, if they exist, and then place in a bin. Intermixed\r
+ with special cases for top, dv, mmapped chunks, and usage errors.\r
+ */\r
+\r
+ if (mem != 0) {\r
+ mchunkptr p = mem2chunk(mem);\r
+#if FOOTERS\r
+ mstate fm = get_mstate_for(p);\r
+ if (!ok_magic(fm)) {\r
+ USAGE_ERROR_ACTION(fm, p);\r
+ return;\r
+ }\r
+#else /* FOOTERS */\r
+#define fm gm\r
+#endif /* FOOTERS */\r
+ if (!PREACTION(fm)) {\r
+ check_inuse_chunk(fm, p);\r
+ if (RTCHECK(ok_address(fm, p) && ok_cinuse(p))) {\r
+ size_t psize = chunksize(p);\r
+ mchunkptr next = chunk_plus_offset(p, psize);\r
+ if (!pinuse(p)) {\r
+ size_t prevsize = p->prev_foot;\r
+ if ((prevsize & IS_MMAPPED_BIT) != 0) {\r
+ prevsize &= ~IS_MMAPPED_BIT;\r
+ psize += prevsize + MMAP_FOOT_PAD;\r
+ if (CALL_MUNMAP((char*)p - prevsize, psize) == 0)\r
+ fm->footprint -= psize;\r
+ goto postaction;\r
+ }\r
+ else {\r
+ mchunkptr prev = chunk_minus_offset(p, prevsize);\r
+ psize += prevsize;\r
+ p = prev;\r
+ if (RTCHECK(ok_address(fm, prev))) { /* consolidate backward */\r
+ if (p != fm->dv) {\r
+ unlink_chunk(fm, p, prevsize);\r
+ }\r
+ else if ((next->head & INUSE_BITS) == INUSE_BITS) {\r
+ fm->dvsize = psize;\r
+ set_free_with_pinuse(p, psize, next);\r
+ goto postaction;\r
+ }\r
+ }\r
+ else\r
+ goto erroraction;\r
+ }\r
+ }\r
+\r
+ if (RTCHECK(ok_next(p, next) && ok_pinuse(next))) {\r
+ if (!cinuse(next)) { /* consolidate forward */\r
+ if (next == fm->top) {\r
+ size_t tsize = fm->topsize += psize;\r
+ fm->top = p;\r
+ p->head = tsize | PINUSE_BIT;\r
+ if (p == fm->dv) {\r
+ fm->dv = 0;\r
+ fm->dvsize = 0;\r
+ }\r
+ if (should_trim(fm, tsize))\r
+ sys_trim(fm, 0);\r
+ goto postaction;\r
+ }\r
+ else if (next == fm->dv) {\r
+ size_t dsize = fm->dvsize += psize;\r
+ fm->dv = p;\r
+ set_size_and_pinuse_of_free_chunk(p, dsize);\r
+ goto postaction;\r
+ }\r
+ else {\r
+ size_t nsize = chunksize(next);\r
+ psize += nsize;\r
+ unlink_chunk(fm, next, nsize);\r
+ set_size_and_pinuse_of_free_chunk(p, psize);\r
+ if (p == fm->dv) {\r
+ fm->dvsize = psize;\r
+ goto postaction;\r
+ }\r
+ }\r
+ }\r
+ else\r
+ set_free_with_pinuse(p, psize, next);\r
+ insert_chunk(fm, p, psize);\r
+ check_free_chunk(fm, p);\r
+ goto postaction;\r
+ }\r
+ }\r
+ erroraction:\r
+ USAGE_ERROR_ACTION(fm, p);\r
+ postaction:\r
+ POSTACTION(fm);\r
+ }\r
+ }\r
+#if !FOOTERS\r
+#undef fm\r
+#endif /* FOOTERS */\r
+}\r
+\r
+void* dlcalloc(size_t n_elements, size_t elem_size) {\r
+ void* mem;\r
+ size_t req = 0;\r
+ if (n_elements != 0) {\r
+ req = n_elements * elem_size;\r
+ if (((n_elements | elem_size) & ~(size_t)0xffff) &&\r
+ (req / n_elements != elem_size))\r
+ req = MAX_SIZE_T; /* force downstream failure on overflow */\r
+ }\r
+ mem = dlmalloc(req);\r
+ if (mem != 0 && calloc_must_clear(mem2chunk(mem)))\r
+ memset(mem, 0, req);\r
+ return mem;\r
+}\r
+\r
+void* dlrealloc(void* oldmem, size_t bytes) {\r
+ if (oldmem == 0)\r
+ return dlmalloc(bytes);\r
+#ifdef REALLOC_ZERO_BYTES_FREES\r
+ if (bytes == 0) {\r
+ dlfree(oldmem);\r
+ return 0;\r
+ }\r
+#endif /* REALLOC_ZERO_BYTES_FREES */\r
+ else {\r
+#if ! FOOTERS\r
+ mstate m = gm;\r
+#else /* FOOTERS */\r
+ mstate m = get_mstate_for(mem2chunk(oldmem));\r
+ if (!ok_magic(m)) {\r
+ USAGE_ERROR_ACTION(m, oldmem);\r
+ return 0;\r
+ }\r
+#endif /* FOOTERS */\r
+ return internal_realloc(m, oldmem, bytes);\r
+ }\r
+}\r
+\r
+void* dlmemalign(size_t alignment, size_t bytes) {\r
+ return internal_memalign(gm, alignment, bytes);\r
+}\r
+\r
+void** dlindependent_calloc(size_t n_elements, size_t elem_size,\r
+ void* chunks[]) {\r
+ size_t sz = elem_size; /* serves as 1-element array */\r
+ return ialloc(gm, n_elements, &sz, 3, chunks);\r
+}\r
+\r
+void** dlindependent_comalloc(size_t n_elements, size_t sizes[],\r
+ void* chunks[]) {\r
+ return ialloc(gm, n_elements, sizes, 0, chunks);\r
+}\r
+\r
+void* dlvalloc(size_t bytes) {\r
+ size_t pagesz;\r
+ init_mparams();\r
+ pagesz = mparams.page_size;\r
+ return dlmemalign(pagesz, bytes);\r
+}\r
+\r
+void* dlpvalloc(size_t bytes) {\r
+ size_t pagesz;\r
+ init_mparams();\r
+ pagesz = mparams.page_size;\r
+ return dlmemalign(pagesz, (bytes + pagesz - SIZE_T_ONE) & ~(pagesz - SIZE_T_ONE));\r
+}\r
+\r
+int dlmalloc_trim(size_t pad) {\r
+ int result = 0;\r
+ if (!PREACTION(gm)) {\r
+ result = sys_trim(gm, pad);\r
+ POSTACTION(gm);\r
+ }\r
+ return result;\r
+}\r
+\r
+size_t dlmalloc_footprint(void) {\r
+ return gm->footprint;\r
+}\r
+\r
+size_t dlmalloc_max_footprint(void) {\r
+ return gm->max_footprint;\r
+}\r
+\r
+#if !NO_MALLINFO\r
+struct mallinfo dlmallinfo(void) {\r
+ return internal_mallinfo(gm);\r
+}\r
+#endif /* NO_MALLINFO */\r
+\r
+void dlmalloc_stats() {\r
+ internal_malloc_stats(gm);\r
+}\r
+\r
+size_t dlmalloc_usable_size(void* mem) {\r
+ if (mem != 0) {\r
+ mchunkptr p = mem2chunk(mem);\r
+ if (cinuse(p))\r
+ return chunksize(p) - overhead_for(p);\r
+ }\r
+ return 0;\r
+}\r
+\r
+int dlmallopt(int param_number, int value) {\r
+ return change_mparam(param_number, value);\r
+}\r
+\r
+#endif /* !ONLY_MSPACES */\r
+\r
+/* ----------------------------- user mspaces ---------------------------- */\r
+\r
+#if MSPACES\r
+\r
+static mstate init_user_mstate(char* tbase, size_t tsize) {\r
+ size_t msize = pad_request(sizeof(struct malloc_state));\r
+ mchunkptr mn;\r
+ mchunkptr msp = align_as_chunk(tbase);\r
+ mstate m = (mstate)(chunk2mem(msp));\r
+ memset(m, 0, msize);\r
+ INITIAL_LOCK(&m->mutex);\r
+ msp->head = (msize|PINUSE_BIT|CINUSE_BIT);\r
+ m->seg.base = m->least_addr = tbase;\r
+ m->seg.size = m->footprint = m->max_footprint = tsize;\r
+ m->magic = mparams.magic;\r
+ m->mflags = mparams.default_mflags;\r
+ m->nedpool = 0;\r
+ disable_contiguous(m);\r
+ init_bins(m);\r
+ mn = next_chunk(mem2chunk(m));\r
+ init_top(m, mn, (size_t)((tbase + tsize) - (char*)mn) - TOP_FOOT_SIZE);\r
+ check_top_chunk(m, m->top);\r
+ return m;\r
+}\r
+\r
+mspace create_mspace(size_t capacity, int locked) {\r
+ mstate m = 0;\r
+ size_t msize = pad_request(sizeof(struct malloc_state));\r
+ init_mparams(); /* Ensure pagesize etc initialized */\r
+\r
+ if (capacity < (size_t) -(msize + TOP_FOOT_SIZE + mparams.page_size)) {\r
+ size_t rs = ((capacity == 0)? mparams.granularity :\r
+ (capacity + TOP_FOOT_SIZE + msize));\r
+ size_t tsize = granularity_align(rs);\r
+ char* tbase = (char*)(CALL_MMAP(tsize));\r
+ if (tbase != CMFAIL) {\r
+ m = init_user_mstate(tbase, tsize);\r
+ m->seg.sflags = IS_MMAPPED_BIT;\r
+ set_lock(m, locked);\r
+ }\r
+ }\r
+ return (mspace)m;\r
+}\r
+\r
+mspace create_mspace_with_base(void* base, size_t capacity, int locked) {\r
+ mstate m = 0;\r
+ size_t msize = pad_request(sizeof(struct malloc_state));\r
+ init_mparams(); /* Ensure pagesize etc initialized */\r
+\r
+ if (capacity > msize + TOP_FOOT_SIZE &&\r
+ capacity < (size_t) -(msize + TOP_FOOT_SIZE + mparams.page_size)) {\r
+ m = init_user_mstate((char*)base, capacity);\r
+ m->seg.sflags = EXTERN_BIT;\r
+ set_lock(m, locked);\r
+ }\r
+ return (mspace)m;\r
+}\r
+\r
+size_t destroy_mspace(mspace msp) {\r
+ size_t freed = 0;\r
+ mstate ms = (mstate)msp;\r
+ if (ok_magic(ms)) {\r
+ msegmentptr sp = &ms->seg;\r
+ while (sp != 0) {\r
+ char* base = sp->base;\r
+ size_t size = sp->size;\r
+ flag_t flag = sp->sflags;\r
+ sp = sp->next;\r
+ if ((flag & IS_MMAPPED_BIT) && !(flag & EXTERN_BIT) &&\r
+ CALL_MUNMAP(base, size) == 0)\r
+ freed += size;\r
+ }\r
+ }\r
+ else {\r
+ USAGE_ERROR_ACTION(ms,ms);\r
+ }\r
+ return freed;\r
+}\r
+\r
+/*\r
+ mspace versions of routines are near-clones of the global\r
+ versions. This is not so nice but better than the alternatives.\r
+*/\r
+\r
+\r
+void* mspace_malloc(mspace msp, size_t bytes) {\r
+ mstate ms = (mstate)msp;\r
+ if (!ok_magic(ms)) {\r
+ USAGE_ERROR_ACTION(ms,ms);\r
+ return 0;\r
+ }\r
+ if (!PREACTION(ms)) {\r
+ void* mem;\r
+ size_t nb;\r
+ if (bytes <= MAX_SMALL_REQUEST) {\r
+ bindex_t idx;\r
+ binmap_t smallbits;\r
+ nb = (bytes < MIN_REQUEST)? MIN_CHUNK_SIZE : pad_request(bytes);\r
+ idx = small_index(nb);\r
+ smallbits = ms->smallmap >> idx;\r
+\r
+ if ((smallbits & 0x3U) != 0) { /* Remainderless fit to a smallbin. */\r
+ mchunkptr b, p;\r
+ idx += ~smallbits & 1; /* Uses next bin if idx empty */\r
+ b = smallbin_at(ms, idx);\r
+ p = b->fd;\r
+ assert(chunksize(p) == small_index2size(idx));\r
+ unlink_first_small_chunk(ms, b, p, idx);\r
+ set_inuse_and_pinuse(ms, p, small_index2size(idx));\r
+ mem = chunk2mem(p);\r
+ check_malloced_chunk(ms, mem, nb);\r
+ goto postaction;\r
+ }\r
+\r
+ else if (nb > ms->dvsize) {\r
+ if (smallbits != 0) { /* Use chunk in next nonempty smallbin */\r
+ mchunkptr b, p, r;\r
+ size_t rsize;\r
+ bindex_t i;\r
+ binmap_t leftbits = (smallbits << idx) & left_bits(idx2bit(idx));\r
+ binmap_t leastbit = least_bit(leftbits);\r
+ compute_bit2idx(leastbit, i);\r
+ b = smallbin_at(ms, i);\r
+ p = b->fd;\r
+ assert(chunksize(p) == small_index2size(i));\r
+ unlink_first_small_chunk(ms, b, p, i);\r
+ rsize = small_index2size(i) - nb;\r
+ /* Fit here cannot be remainderless if 4byte sizes */\r
+ if (SIZE_T_SIZE != 4 && rsize < MIN_CHUNK_SIZE)\r
+ set_inuse_and_pinuse(ms, p, small_index2size(i));\r
+ else {\r
+ set_size_and_pinuse_of_inuse_chunk(ms, p, nb);\r
+ r = chunk_plus_offset(p, nb);\r
+ set_size_and_pinuse_of_free_chunk(r, rsize);\r
+ replace_dv(ms, r, rsize);\r
+ }\r
+ mem = chunk2mem(p);\r
+ check_malloced_chunk(ms, mem, nb);\r
+ goto postaction;\r
+ }\r
+\r
+ else if (ms->treemap != 0 && (mem = tmalloc_small(ms, nb)) != 0) {\r
+ check_malloced_chunk(ms, mem, nb);\r
+ goto postaction;\r
+ }\r
+ }\r
+ }\r
+ else if (bytes >= MAX_REQUEST)\r
+ nb = MAX_SIZE_T; /* Too big to allocate. Force failure (in sys alloc) */\r
+ else {\r
+ nb = pad_request(bytes);\r
+ if (ms->treemap != 0 && (mem = tmalloc_large(ms, nb)) != 0) {\r
+ check_malloced_chunk(ms, mem, nb);\r
+ goto postaction;\r
+ }\r
+ }\r
+\r
+ if (nb <= ms->dvsize) {\r
+ size_t rsize = ms->dvsize - nb;\r
+ mchunkptr p = ms->dv;\r
+ if (rsize >= MIN_CHUNK_SIZE) { /* split dv */\r
+ mchunkptr r = ms->dv = chunk_plus_offset(p, nb);\r
+ ms->dvsize = rsize;\r
+ set_size_and_pinuse_of_free_chunk(r, rsize);\r
+ set_size_and_pinuse_of_inuse_chunk(ms, p, nb);\r
+ }\r
+ else { /* exhaust dv */\r
+ size_t dvs = ms->dvsize;\r
+ ms->dvsize = 0;\r
+ ms->dv = 0;\r
+ set_inuse_and_pinuse(ms, p, dvs);\r
+ }\r
+ mem = chunk2mem(p);\r
+ check_malloced_chunk(ms, mem, nb);\r
+ goto postaction;\r
+ }\r
+\r
+ else if (nb < ms->topsize) { /* Split top */\r
+ size_t rsize = ms->topsize -= nb;\r
+ mchunkptr p = ms->top;\r
+ mchunkptr r = ms->top = chunk_plus_offset(p, nb);\r
+ r->head = rsize | PINUSE_BIT;\r
+ set_size_and_pinuse_of_inuse_chunk(ms, p, nb);\r
+ mem = chunk2mem(p);\r
+ check_top_chunk(ms, ms->top);\r
+ check_malloced_chunk(ms, mem, nb);\r
+ goto postaction;\r
+ }\r
+\r
+ mem = sys_alloc(ms, nb);\r
+\r
+ postaction:\r
+ POSTACTION(ms);\r
+ return mem;\r
+ }\r
+\r
+ return 0;\r
+}\r
+\r
+void mspace_free(mspace msp, void* mem) {\r
+ if (mem != 0) {\r
+ mchunkptr p = mem2chunk(mem);\r
+#if FOOTERS\r
+ mstate fm = get_mstate_for(p);\r
+#else /* FOOTERS */\r
+ mstate fm = (mstate)msp;\r
+#endif /* FOOTERS */\r
+ if (!ok_magic(fm)) {\r
+ USAGE_ERROR_ACTION(fm, p);\r
+ return;\r
+ }\r
+ if (!PREACTION(fm)) {\r
+ check_inuse_chunk(fm, p);\r
+ if (RTCHECK(ok_address(fm, p) && ok_cinuse(p))) {\r
+ size_t psize = chunksize(p);\r
+ mchunkptr next = chunk_plus_offset(p, psize);\r
+ if (!pinuse(p)) {\r
+ size_t prevsize = p->prev_foot;\r
+ if ((prevsize & IS_MMAPPED_BIT) != 0) {\r
+ prevsize &= ~IS_MMAPPED_BIT;\r
+ psize += prevsize + MMAP_FOOT_PAD;\r
+ if (CALL_MUNMAP((char*)p - prevsize, psize) == 0)\r
+ fm->footprint -= psize;\r
+ goto postaction;\r
+ }\r
+ else {\r
+ mchunkptr prev = chunk_minus_offset(p, prevsize);\r
+ psize += prevsize;\r
+ p = prev;\r
+ if (RTCHECK(ok_address(fm, prev))) { /* consolidate backward */\r
+ if (p != fm->dv) {\r
+ unlink_chunk(fm, p, prevsize);\r
+ }\r
+ else if ((next->head & INUSE_BITS) == INUSE_BITS) {\r
+ fm->dvsize = psize;\r
+ set_free_with_pinuse(p, psize, next);\r
+ goto postaction;\r
+ }\r
+ }\r
+ else\r
+ goto erroraction;\r
+ }\r
+ }\r
+\r
+ if (RTCHECK(ok_next(p, next) && ok_pinuse(next))) {\r
+ if (!cinuse(next)) { /* consolidate forward */\r
+ if (next == fm->top) {\r
+ size_t tsize = fm->topsize += psize;\r
+ fm->top = p;\r
+ p->head = tsize | PINUSE_BIT;\r
+ if (p == fm->dv) {\r
+ fm->dv = 0;\r
+ fm->dvsize = 0;\r
+ }\r
+ if (should_trim(fm, tsize))\r
+ sys_trim(fm, 0);\r
+ goto postaction;\r
+ }\r
+ else if (next == fm->dv) {\r
+ size_t dsize = fm->dvsize += psize;\r
+ fm->dv = p;\r
+ set_size_and_pinuse_of_free_chunk(p, dsize);\r
+ goto postaction;\r
+ }\r
+ else {\r
+ size_t nsize = chunksize(next);\r
+ psize += nsize;\r
+ unlink_chunk(fm, next, nsize);\r
+ set_size_and_pinuse_of_free_chunk(p, psize);\r
+ if (p == fm->dv) {\r
+ fm->dvsize = psize;\r
+ goto postaction;\r
+ }\r
+ }\r
+ }\r
+ else\r
+ set_free_with_pinuse(p, psize, next);\r
+ insert_chunk(fm, p, psize);\r
+ check_free_chunk(fm, p);\r
+ goto postaction;\r
+ }\r
+ }\r
+ erroraction:\r
+ USAGE_ERROR_ACTION(fm, p);\r
+ postaction:\r
+ POSTACTION(fm);\r
+ }\r
+ }\r
+}\r
+\r
+void* mspace_calloc(mspace msp, size_t n_elements, size_t elem_size) {\r
+ void* mem;\r
+ size_t req = 0;\r
+ mstate ms = (mstate)msp;\r
+ if (!ok_magic(ms)) {\r
+ USAGE_ERROR_ACTION(ms,ms);\r
+ return 0;\r
+ }\r
+ if (n_elements != 0) {\r
+ req = n_elements * elem_size;\r
+ if (((n_elements | elem_size) & ~(size_t)0xffff) &&\r
+ (req / n_elements != elem_size))\r
+ req = MAX_SIZE_T; /* force downstream failure on overflow */\r
+ }\r
+ mem = internal_malloc(ms, req);\r
+ if (mem != 0 && calloc_must_clear(mem2chunk(mem)))\r
+ memset(mem, 0, req);\r
+ return mem;\r
+}\r
+\r
+void* mspace_realloc(mspace msp, void* oldmem, size_t bytes) {\r
+ if (oldmem == 0)\r
+ return mspace_malloc(msp, bytes);\r
+#ifdef REALLOC_ZERO_BYTES_FREES\r
+ if (bytes == 0) {\r
+ mspace_free(msp, oldmem);\r
+ return 0;\r
+ }\r
+#endif /* REALLOC_ZERO_BYTES_FREES */\r
+ else {\r
+#if FOOTERS\r
+ mchunkptr p = mem2chunk(oldmem);\r
+ mstate ms = get_mstate_for(p);\r
+#else /* FOOTERS */\r
+ mstate ms = (mstate)msp;\r
+#endif /* FOOTERS */\r
+ if (!ok_magic(ms)) {\r
+ USAGE_ERROR_ACTION(ms,ms);\r
+ return 0;\r
+ }\r
+ return internal_realloc(ms, oldmem, bytes);\r
+ }\r
+}\r
+\r
+void* mspace_memalign(mspace msp, size_t alignment, size_t bytes) {\r
+ mstate ms = (mstate)msp;\r
+ if (!ok_magic(ms)) {\r
+ USAGE_ERROR_ACTION(ms,ms);\r
+ return 0;\r
+ }\r
+ return internal_memalign(ms, alignment, bytes);\r
+}\r
+\r
+void** mspace_independent_calloc(mspace msp, size_t n_elements,\r
+ size_t elem_size, void* chunks[]) {\r
+ size_t sz = elem_size; /* serves as 1-element array */\r
+ mstate ms = (mstate)msp;\r
+ if (!ok_magic(ms)) {\r
+ USAGE_ERROR_ACTION(ms,ms);\r
+ return 0;\r
+ }\r
+ return ialloc(ms, n_elements, &sz, 3, chunks);\r
+}\r
+\r
+void** mspace_independent_comalloc(mspace msp, size_t n_elements,\r
+ size_t sizes[], void* chunks[]) {\r
+ mstate ms = (mstate)msp;\r
+ if (!ok_magic(ms)) {\r
+ USAGE_ERROR_ACTION(ms,ms);\r
+ return 0;\r
+ }\r
+ return ialloc(ms, n_elements, sizes, 0, chunks);\r
+}\r
+\r
+int mspace_trim(mspace msp, size_t pad) {\r
+ int result = 0;\r
+ mstate ms = (mstate)msp;\r
+ if (ok_magic(ms)) {\r
+ if (!PREACTION(ms)) {\r
+ result = sys_trim(ms, pad);\r
+ POSTACTION(ms);\r
+ }\r
+ }\r
+ else {\r
+ USAGE_ERROR_ACTION(ms,ms);\r
+ }\r
+ return result;\r
+}\r
+\r
+void mspace_malloc_stats(mspace msp) {\r
+ mstate ms = (mstate)msp;\r
+ if (ok_magic(ms)) {\r
+ internal_malloc_stats(ms);\r
+ }\r
+ else {\r
+ USAGE_ERROR_ACTION(ms,ms);\r
+ }\r
+}\r
+\r
+size_t mspace_footprint(mspace msp) {\r
+ size_t result = 0;\r
+ mstate ms = (mstate)msp;\r
+ if (ok_magic(ms)) {\r
+ result = ms->footprint;\r
+ }\r
+ USAGE_ERROR_ACTION(ms,ms);\r
+ return result;\r
+}\r
+\r
+\r
+size_t mspace_max_footprint(mspace msp) {\r
+ size_t result = 0;\r
+ mstate ms = (mstate)msp;\r
+ if (ok_magic(ms)) {\r
+ result = ms->max_footprint;\r
+ }\r
+ USAGE_ERROR_ACTION(ms,ms);\r
+ return result;\r
+}\r
+\r
+\r
+#if !NO_MALLINFO\r
+struct mallinfo mspace_mallinfo(mspace msp) {\r
+ mstate ms = (mstate)msp;\r
+ if (!ok_magic(ms)) {\r
+ USAGE_ERROR_ACTION(ms,ms);\r
+ }\r
+ return internal_mallinfo(ms);\r
+}\r
+#endif /* NO_MALLINFO */\r
+\r
+int mspace_mallopt(int param_number, int value) {\r
+ return change_mparam(param_number, value);\r
+}\r
+\r
+#endif /* MSPACES */\r
+\r
+/* -------------------- Alternative MORECORE functions ------------------- */\r
+\r
+/*\r
+ Guidelines for creating a custom version of MORECORE:\r
+\r
+ * For best performance, MORECORE should allocate in multiples of pagesize.\r
+ * MORECORE may allocate more memory than requested. (Or even less,\r
+ but this will usually result in a malloc failure.)\r
+ * MORECORE must not allocate memory when given argument zero, but\r
+ instead return one past the end address of memory from previous\r
+ nonzero call.\r
+ * For best performance, consecutive calls to MORECORE with positive\r
+ arguments should return increasing addresses, indicating that\r
+ space has been contiguously extended.\r
+ * Even though consecutive calls to MORECORE need not return contiguous\r
+ addresses, it must be OK for malloc'ed chunks to span multiple\r
+ regions in those cases where they do happen to be contiguous.\r
+ * MORECORE need not handle negative arguments -- it may instead\r
+ just return MFAIL when given negative arguments.\r
+ Negative arguments are always multiples of pagesize. MORECORE\r
+ must not misinterpret negative args as large positive unsigned\r
+ args. You can suppress all such calls from even occurring by defining\r
+ MORECORE_CANNOT_TRIM,\r
+\r
+ As an example alternative MORECORE, here is a custom allocator\r
+ kindly contributed for pre-OSX macOS. It uses virtually but not\r
+ necessarily physically contiguous non-paged memory (locked in,\r
+ present and won't get swapped out). You can use it by uncommenting\r
+ this section, adding some #includes, and setting up the appropriate\r
+ defines above:\r
+\r
+ #define MORECORE osMoreCore\r
+\r
+ There is also a shutdown routine that should somehow be called for\r
+ cleanup upon program exit.\r
+\r
+ #define MAX_POOL_ENTRIES 100\r
+ #define MINIMUM_MORECORE_SIZE (64 * 1024U)\r
+ static int next_os_pool;\r
+ void *our_os_pools[MAX_POOL_ENTRIES];\r
+\r
+ void *osMoreCore(int size)\r
+ {\r
+ void *ptr = 0;\r
+ static void *sbrk_top = 0;\r
+\r
+ if (size > 0)\r
+ {\r
+ if (size < MINIMUM_MORECORE_SIZE)\r
+ size = MINIMUM_MORECORE_SIZE;\r
+ if (CurrentExecutionLevel() == kTaskLevel)\r
+ ptr = PoolAllocateResident(size + RM_PAGE_SIZE, 0);\r
+ if (ptr == 0)\r
+ {\r
+ return (void *) MFAIL;\r
+ }\r
+ // save ptrs so they can be freed during cleanup\r
+ our_os_pools[next_os_pool] = ptr;\r
+ next_os_pool++;\r
+ ptr = (void *) ((((size_t) ptr) + RM_PAGE_MASK) & ~RM_PAGE_MASK);\r
+ sbrk_top = (char *) ptr + size;\r
+ return ptr;\r
+ }\r
+ else if (size < 0)\r
+ {\r
+ // we don't currently support shrink behavior\r
+ return (void *) MFAIL;\r
+ }\r
+ else\r
+ {\r
+ return sbrk_top;\r
+ }\r
+ }\r
+\r
+ // cleanup any allocated memory pools\r
+ // called as last thing before shutting down driver\r
+\r
+ void osCleanupMem(void)\r
+ {\r
+ void **ptr;\r
+\r
+ for (ptr = our_os_pools; ptr < &our_os_pools[MAX_POOL_ENTRIES]; ptr++)\r
+ if (*ptr)\r
+ {\r
+ PoolDeallocate(*ptr);\r
+ *ptr = 0;\r
+ }\r
+ }\r
+\r
+*/\r
+\r
+\r
+/* -----------------------------------------------------------------------\r
+History:\r
+ V2.8.3 Thu Sep 22 11:16:32 2005 Doug Lea (dl at gee)\r
+ * Add max_footprint functions\r
+ * Ensure all appropriate literals are size_t\r
+ * Fix conditional compilation problem for some #define settings\r
+ * Avoid concatenating segments with the one provided\r
+ in create_mspace_with_base\r
+ * Rename some variables to avoid compiler shadowing warnings\r
+ * Use explicit lock initialization.\r
+ * Better handling of sbrk interference.\r
+ * Simplify and fix segment insertion, trimming and mspace_destroy\r
+ * Reinstate REALLOC_ZERO_BYTES_FREES option from 2.7.x\r
+ * Thanks especially to Dennis Flanagan for help on these.\r
+\r
+ V2.8.2 Sun Jun 12 16:01:10 2005 Doug Lea (dl at gee)\r
+ * Fix memalign brace error.\r
+\r
+ V2.8.1 Wed Jun 8 16:11:46 2005 Doug Lea (dl at gee)\r
+ * Fix improper #endif nesting in C++\r
+ * Add explicit casts needed for C++\r
+\r
+ V2.8.0 Mon May 30 14:09:02 2005 Doug Lea (dl at gee)\r
+ * Use trees for large bins\r
+ * Support mspaces\r
+ * Use segments to unify sbrk-based and mmap-based system allocation,\r
+ removing need for emulation on most platforms without sbrk.\r
+ * Default safety checks\r
+ * Optional footer checks. Thanks to William Robertson for the idea.\r
+ * Internal code refactoring\r
+ * Incorporate suggestions and platform-specific changes.\r
+ Thanks to Dennis Flanagan, Colin Plumb, Niall Douglas,\r
+ Aaron Bachmann, Emery Berger, and others.\r
+ * Speed up non-fastbin processing enough to remove fastbins.\r
+ * Remove useless cfree() to avoid conflicts with other apps.\r
+ * Remove internal memcpy, memset. Compilers handle builtins better.\r
+ * Remove some options that no one ever used and rename others.\r
+\r
+ V2.7.2 Sat Aug 17 09:07:30 2002 Doug Lea (dl at gee)\r
+ * Fix malloc_state bitmap array misdeclaration\r
+\r
+ V2.7.1 Thu Jul 25 10:58:03 2002 Doug Lea (dl at gee)\r
+ * Allow tuning of FIRST_SORTED_BIN_SIZE\r
+ * Use PTR_UINT as type for all ptr->int casts. Thanks to John Belmonte.\r
+ * Better detection and support for non-contiguousness of MORECORE.\r
+ Thanks to Andreas Mueller, Conal Walsh, and Wolfram Gloger\r
+ * Bypass most of malloc if no frees. Thanks To Emery Berger.\r
+ * Fix freeing of old top non-contiguous chunk im sysmalloc.\r
+ * Raised default trim and map thresholds to 256K.\r
+ * Fix mmap-related #defines. Thanks to Lubos Lunak.\r
+ * Fix copy macros; added LACKS_FCNTL_H. Thanks to Neal Walfield.\r
+ * Branch-free bin calculation\r
+ * Default trim and mmap thresholds now 256K.\r
+\r
+ V2.7.0 Sun Mar 11 14:14:06 2001 Doug Lea (dl at gee)\r
+ * Introduce independent_comalloc and independent_calloc.\r
+ Thanks to Michael Pachos for motivation and help.\r
+ * Make optional .h file available\r
+ * Allow > 2GB requests on 32bit systems.\r
+ * new WIN32 sbrk, mmap, munmap, lock code from <Walter@GeNeSys-e.de>.\r
+ Thanks also to Andreas Mueller <a.mueller at paradatec.de>,\r
+ and Anonymous.\r
+ * Allow override of MALLOC_ALIGNMENT (Thanks to Ruud Waij for\r
+ helping test this.)\r
+ * memalign: check alignment arg\r
+ * realloc: don't try to shift chunks backwards, since this\r
+ leads to more fragmentation in some programs and doesn't\r
+ seem to help in any others.\r
+ * Collect all cases in malloc requiring system memory into sysmalloc\r
+ * Use mmap as backup to sbrk\r
+ * Place all internal state in malloc_state\r
+ * Introduce fastbins (although similar to 2.5.1)\r
+ * Many minor tunings and cosmetic improvements\r
+ * Introduce USE_PUBLIC_MALLOC_WRAPPERS, USE_MALLOC_LOCK\r
+ * Introduce MALLOC_FAILURE_ACTION, MORECORE_CONTIGUOUS\r
+ Thanks to Tony E. Bennett <tbennett@nvidia.com> and others.\r
+ * Include errno.h to support default failure action.\r
+\r
+ V2.6.6 Sun Dec 5 07:42:19 1999 Doug Lea (dl at gee)\r
+ * return null for negative arguments\r
+ * Added Several WIN32 cleanups from Martin C. Fong <mcfong at yahoo.com>\r
+ * Add 'LACKS_SYS_PARAM_H' for those systems without 'sys/param.h'\r
+ (e.g. WIN32 platforms)\r
+ * Cleanup header file inclusion for WIN32 platforms\r
+ * Cleanup code to avoid Microsoft Visual C++ compiler complaints\r
+ * Add 'USE_DL_PREFIX' to quickly allow co-existence with existing\r
+ memory allocation routines\r
+ * Set 'malloc_getpagesize' for WIN32 platforms (needs more work)\r
+ * Use 'assert' rather than 'ASSERT' in WIN32 code to conform to\r
+ usage of 'assert' in non-WIN32 code\r
+ * Improve WIN32 'sbrk()' emulation's 'findRegion()' routine to\r
+ avoid infinite loop\r
+ * Always call 'fREe()' rather than 'free()'\r
+\r
+ V2.6.5 Wed Jun 17 15:57:31 1998 Doug Lea (dl at gee)\r
+ * Fixed ordering problem with boundary-stamping\r
+\r
+ V2.6.3 Sun May 19 08:17:58 1996 Doug Lea (dl at gee)\r
+ * Added pvalloc, as recommended by H.J. Liu\r
+ * Added 64bit pointer support mainly from Wolfram Gloger\r
+ * Added anonymously donated WIN32 sbrk emulation\r
+ * Malloc, calloc, getpagesize: add optimizations from Raymond Nijssen\r
+ * malloc_extend_top: fix mask error that caused wastage after\r
+ foreign sbrks\r
+ * Add linux mremap support code from HJ Liu\r
+\r
+ V2.6.2 Tue Dec 5 06:52:55 1995 Doug Lea (dl at gee)\r
+ * Integrated most documentation with the code.\r
+ * Add support for mmap, with help from\r
+ Wolfram Gloger (Gloger@lrz.uni-muenchen.de).\r
+ * Use last_remainder in more cases.\r
+ * Pack bins using idea from colin@nyx10.cs.du.edu\r
+ * Use ordered bins instead of best-fit threshhold\r
+ * Eliminate block-local decls to simplify tracing and debugging.\r
+ * Support another case of realloc via move into top\r
+ * Fix error occuring when initial sbrk_base not word-aligned.\r
+ * Rely on page size for units instead of SBRK_UNIT to\r
+ avoid surprises about sbrk alignment conventions.\r
+ * Add mallinfo, mallopt. Thanks to Raymond Nijssen\r
+ (raymond@es.ele.tue.nl) for the suggestion.\r
+ * Add `pad' argument to malloc_trim and top_pad mallopt parameter.\r
+ * More precautions for cases where other routines call sbrk,\r
+ courtesy of Wolfram Gloger (Gloger@lrz.uni-muenchen.de).\r
+ * Added macros etc., allowing use in linux libc from\r
+ H.J. Lu (hjl@gnu.ai.mit.edu)\r
+ * Inverted this history list\r
+\r
+ V2.6.1 Sat Dec 2 14:10:57 1995 Doug Lea (dl at gee)\r
+ * Re-tuned and fixed to behave more nicely with V2.6.0 changes.\r
+ * Removed all preallocation code since under current scheme\r
+ the work required to undo bad preallocations exceeds\r
+ the work saved in good cases for most test programs.\r
+ * No longer use return list or unconsolidated bins since\r
+ no scheme using them consistently outperforms those that don't\r
+ given above changes.\r
+ * Use best fit for very large chunks to prevent some worst-cases.\r
+ * Added some support for debugging\r
+\r
+ V2.6.0 Sat Nov 4 07:05:23 1995 Doug Lea (dl at gee)\r
+ * Removed footers when chunks are in use. Thanks to\r
+ Paul Wilson (wilson@cs.texas.edu) for the suggestion.\r
+\r
+ V2.5.4 Wed Nov 1 07:54:51 1995 Doug Lea (dl at gee)\r
+ * Added malloc_trim, with help from Wolfram Gloger\r
+ (wmglo@Dent.MED.Uni-Muenchen.DE).\r
+\r
+ V2.5.3 Tue Apr 26 10:16:01 1994 Doug Lea (dl at g)\r
+\r
+ V2.5.2 Tue Apr 5 16:20:40 1994 Doug Lea (dl at g)\r
+ * realloc: try to expand in both directions\r
+ * malloc: swap order of clean-bin strategy;\r
+ * realloc: only conditionally expand backwards\r
+ * Try not to scavenge used bins\r
+ * Use bin counts as a guide to preallocation\r
+ * Occasionally bin return list chunks in first scan\r
+ * Add a few optimizations from colin@nyx10.cs.du.edu\r
+\r
+ V2.5.1 Sat Aug 14 15:40:43 1993 Doug Lea (dl at g)\r
+ * faster bin computation & slightly different binning\r
+ * merged all consolidations to one part of malloc proper\r
+ (eliminating old malloc_find_space & malloc_clean_bin)\r
+ * Scan 2 returns chunks (not just 1)\r
+ * Propagate failure in realloc if malloc returns 0\r
+ * Add stuff to allow compilation on non-ANSI compilers\r
+ from kpv@research.att.com\r
+\r
+ V2.5 Sat Aug 7 07:41:59 1993 Doug Lea (dl at g.oswego.edu)\r
+ * removed potential for odd address access in prev_chunk\r
+ * removed dependency on getpagesize.h\r
+ * misc cosmetics and a bit more internal documentation\r
+ * anticosmetics: mangled names in macros to evade debugger strangeness\r
+ * tested on sparc, hp-700, dec-mips, rs6000\r
+ with gcc & native cc (hp, dec only) allowing\r
+ Detlefs & Zorn comparison study (in SIGPLAN Notices.)\r
+\r
+ Trial version Fri Aug 28 13:14:29 1992 Doug Lea (dl at g.oswego.edu)\r
+ * Based loosely on libg++-1.2X malloc. (It retains some of the overall\r
+ structure of old version, but most details differ.)\r
+ \r
+*/\r
projects / pdns / commitdiff