@matheusbsbs / matheus-bagliano / commits / f0ed8226c9
Fork 0

Add custom memory allocator to MinGW and MacOS builds

The standard allocator on Windows is pretty bad prior to Windows Vista, and nedmalloc is better than the modified dlmalloc provided with newer versions of the MinGW libc. NedMalloc stats in Git ---------------------- All results are the best result out of 3 runs. The benchmarks have been done on different hardware, so the repack times are not comparable. These benchmarks are all based on 'git repack -adf' on the Linux kernel. XP ----------------------------------------------- MinGW Threads Total Time Speed ----------------------------------------------- 3.4.2 (1T) 00:12:28.422 3.4.2 + nedmalloc (1T) 00:07:25.437 1.68x 3.4.5 (1T) 00:12:20.718 3.4.5 + nedmalloc (1T) 00:07:24.809 1.67x 4.3.3-tdm (1T) 00:12:01.843 4.3.3-tdm + nedmalloc (1T) 00:07:16.468 1.65x 4.3.3-tdm (2T) 00:07:35.062 4.3.3-tdm + nedmalloc (2T) 00:04:57.874 1.54x Vista ----------------------------------------------- MinGW Threads Total Time Speed ----------------------------------------------- 4.3.3-tdm (1T) 00:07:40.844 4.3.3-tdm + nedmalloc (1T) 00:07:17.548 1.05x 4.3.3-tdm (2T) 00:05:33.746 4.3.3-tdm + nedmalloc (2T) 00:05:27.334 1.02x Mac Mini ----------------------------------------------- GCC Threads Total Time Speed ----------------------------------------------- i686-darwin9-4.0.1 (2T) 00:09:57.346 i686-darwin9-4.0.1+ned (2T) 00:08:51.072 1.12x Signed-off-by: Marius Storm-Olsen <marius@trolltech.com> Signed-off-by: Steffen Prohaska <prohaska@zib.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>

Marius Storm-Olsen committed May 31, 2009 at 18:15 UTC f0ed8226c9c65f2a324610258418258229e77fbe
6 files changed +7064
Makefile
+9
@@ -178,6 +178,9 @@ all::
178 #
179 # Define NO_CROSS_DIRECTORY_HARDLINKS if you plan to distribute the installed
180 # programs as a tar, where bin/ and libexec/ might be on different file systems.
181 +#
182 +# Define USE_NED_ALLOCATOR if you want to replace the platforms default
183 +# memory allocators with the nedmalloc allocator written by Niall Douglas.
184
185 GIT-VERSION-FILE: .FORCE-GIT-VERSION-FILE
186 @$(SHELL_PATH) ./GIT-VERSION-GEN
@@ -844,6 +847,7 @@ ifneq (,$(findstring MINGW,$(uname_S)))
847 NO_ST_BLOCKS_IN_STRUCT_STAT = YesPlease
848 NO_NSEC = YesPlease
849 USE_WIN32_MMAP = YesPlease
850 + USE_NED_ALLOCATOR = YesPlease
851 UNRELIABLE_FSTAT = UnfortunatelyYes
852 OBJECT_CREATION_USES_RENAMES = UnfortunatelyNeedsTo
853 COMPAT_CFLAGS += -D__USE_MINGW_ACCESS -DNOGDI -Icompat -Icompat/regex -Icompat/fnmatch
@@ -1130,6 +1134,11 @@ ifdef UNRELIABLE_FSTAT
1134 BASIC_CFLAGS += -DUNRELIABLE_FSTAT
1135 endif
1136
1137 +ifdef USE_NED_ALLOCATOR
1138 + COMPAT_CFLAGS += -DUSE_NED_ALLOCATOR -DOVERRIDE_STRDUP -DNDEBUG -DREPLACE_SYSTEM_ALLOCATOR -Icompat/nedmalloc
1139 + COMPAT_OBJS += compat/nedmalloc/nedmalloc.o
1140 +endif
1141 +
1142 ifeq ($(TCLTK_PATH),)
1143 NO_TCLTK=NoThanks
1144 endif
compat/nedmalloc/License.txt new
+23
@@ -0,0 +1,23 @@
1 +Boost Software License - Version 1.0 - August 17th, 2003
2 +
3 +Permission is hereby granted, free of charge, to any person or organization
4 +obtaining a copy of the software and accompanying documentation covered by
5 +this license (the "Software") to use, reproduce, display, distribute,
6 +execute, and transmit the Software, and to prepare derivative works of the
7 +Software, and to permit third-parties to whom the Software is furnished to
8 +do so, all subject to the following:
9 +
10 +The copyright notices in the Software and this entire statement, including
11 +the above license grant, this restriction and the following disclaimer,
12 +must be included in all copies of the Software, in whole or in part, and
13 +all derivative works of the Software, unless such copies or derivative
14 +works are solely in the form of machine-executable object code generated by
15 +a source language processor.
16 +
17 +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
18 +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19 +FITNESS FOR A PARTICULAR PURPOSE, TITLE AND NON-INFRINGEMENT. IN NO EVENT
20 +SHALL THE COPYRIGHT HOLDERS OR ANYONE DISTRIBUTING THE SOFTWARE BE LIABLE
21 +FOR ANY DAMAGES OR OTHER LIABILITY, WHETHER IN CONTRACT, TORT OR OTHERWISE,
22 +ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
23 +DEALINGS IN THE SOFTWARE.
compat/nedmalloc/Readme.txt new
+136
@@ -0,0 +1,136 @@
1 +nedalloc v1.05 15th June 2008:
2 +-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
3 +
4 +by Niall Douglas (http://www.nedprod.com/programs/portable/nedmalloc/)
5 +
6 +Enclosed is nedalloc, an alternative malloc implementation for multiple
7 +threads without lock contention based on dlmalloc v2.8.4. It is more
8 +or less a newer implementation of ptmalloc2, the standard allocator in
9 +Linux (which is based on dlmalloc v2.7.0) but also contains a per-thread
10 +cache for maximum CPU scalability.
11 +
12 +It is licensed under the Boost Software License which basically means
13 +you can do anything you like with it. This does not apply to the malloc.c.h
14 +file which remains copyright to others.
15 +
16 +It has been tested on win32 (x86), win64 (x64), Linux (x64), FreeBSD (x64)
17 +and Apple MacOS X (x86). It works very well on all of these and is very
18 +significantly faster than the system allocator on all of these platforms.
19 +
20 +By literally dropping in this allocator as a replacement for your system
21 +allocator, you can see real world improvements of up to three times in normal
22 +code!
23 +
24 +To use:
25 +-=-=-=-
26 +Drop in nedmalloc.h, nedmalloc.c and malloc.c.h into your project.
27 +Configure using the instructions in nedmalloc.h. Run and enjoy.
28 +
29 +To test, compile test.c. It will run a comparison between your system
30 +allocator and nedalloc and tell you how much faster nedalloc is. It also
31 +serves as an example of usage.
32 +
33 +Notes:
34 +-=-=-=
35 +If you want the very latest version of this allocator, get it from the
36 +TnFOX SVN repository at svn://svn.berlios.de/viewcvs/tnfox/trunk/src/nedmalloc
37 +
38 +Because of how nedalloc allocates an mspace per thread, it can cause
39 +severe bloating of memory usage under certain allocation patterns.
40 +You can substantially reduce this wastage by setting MAXTHREADSINPOOL
41 +or the threads parameter to nedcreatepool() to a fraction of the number of
42 +threads which would normally be in a pool at once. This will reduce
43 +bloating at the cost of an increase in lock contention. If allocated size
44 +is less than THREADCACHEMAX, locking is avoided 90-99% of the time and
45 +if most of your allocations are below this value, you can safely set
46 +MAXTHREADSINPOOL to one.
47 +
48 +You will suffer memory leakage unless you call neddisablethreadcache()
49 +per pool for every thread which exits. This is because nedalloc cannot
50 +portably know when a thread exits and thus when its thread cache can
51 +be returned for use by other code. Don't forget pool zero, the system pool.
52 +
53 +For C++ type allocation patterns (where the same sizes of memory are
54 +regularly allocated and deallocated as objects are created and destroyed),
55 +the threadcache always benefits performance. If however your allocation
56 +patterns are different, searching the threadcache may significantly slow
57 +down your code - as a rule of thumb, if cache utilisation is below 80%
58 +(see the source for neddisablethreadcache() for how to enable debug
59 +printing in release mode) then you should disable the thread cache for
60 +that thread. You can compile out the threadcache code by setting
61 +THREADCACHEMAX to zero.
62 +
63 +Speed comparisons:
64 +-=-=-=-=-=-=-=-=-=
65 +See Benchmarks.xls for details.
66 +
67 +The enclosed test.c can do two things: it can be a torture test or a speed
68 +test. The speed test is designed to be a representative synthetic
69 +memory allocator test. It works by randomly mixing allocations with frees
70 +with half of the allocation sizes being a two power multiple less than
71 +512 bytes (to mimic C++ stack instantiated objects) and the other half
72 +being a simple random value less than 16Kb.
73 +
74 +The real world code results are from Tn's TestIO benchmark. This is a
75 +heavily multithreaded and memory intensive benchmark with a lot of branching
76 +and other stuff modern processors don't like so much. As you'll note, the
77 +test doesn't show the benefits of the threadcache mostly due to the saturation
78 +of the memory bus being the limiting factor.
79 +
80 +ChangeLog:
81 +-=-=-=-=-=
82 +v1.05 15th June 2008:
83 + * { 1042 } Added error check for TLSSET() and TLSFREE() macros. Thanks to
84 +Markus Elfring for reporting this.
85 + * { 1043 } Fixed a segfault when freeing memory allocated using
86 +nedindependent_comalloc(). Thanks to Pavel Vozenilek for reporting this.
87 +
88 +v1.04 14th July 2007:
89 + * Fixed a bug with the new optimised implementation that failed to lock
90 +on a realloc under certain conditions.
91 + * Fixed lack of thread synchronisation in InitPool() causing pool corruption
92 + * Fixed a memory leak of thread cache contents on disabling. Thanks to Earl
93 +Chew for reporting this.
94 + * Added a sanity check for freed blocks being valid.
95 + * Reworked test.c into being a torture test.
96 + * Fixed GCC assembler optimisation misspecification
97 +
98 +v1.04alpha_svn915 7th October 2006:
99 + * Fixed failure to unlock thread cache list if allocating a new list failed.
100 +Thanks to Dmitry Chichkov for reporting this. Futher thanks to Aleksey Sanin.
101 + * Fixed realloc(0, <size>) segfaulting. Thanks to Dmitry Chichkov for
102 +reporting this.
103 + * Made config defines #ifndef so they can be overriden by the build system.
104 +Thanks to Aleksey Sanin for suggesting this.
105 + * Fixed deadlock in nedprealloc() due to unnecessary locking of preferred
106 +thread mspace when mspace_realloc() always uses the original block's mspace
107 +anyway. Thanks to Aleksey Sanin for reporting this.
108 + * Made some speed improvements by hacking mspace_malloc() to no longer lock
109 +its mspace, thus allowing the recursive mutex implementation to be removed
110 +with an associated speed increase. Thanks to Aleksey Sanin for suggesting this.
111 + * Fixed a bug where allocating mspaces overran its max limit. Thanks to
112 +Aleksey Sanin for reporting this.
113 +
114 +v1.03 10th July 2006:
115 + * Fixed memory corruption bug in threadcache code which only appeared with >4
116 +threads and in heavy use of the threadcache.
117 +
118 +v1.02 15th May 2006:
119 + * Integrated dlmalloc v2.8.4, fixing the win32 memory release problem and
120 +improving performance still further. Speed is now up to twice the speed of v1.01
121 +(average is 67% faster).
122 + * Fixed win32 critical section implementation. Thanks to Pavel Kuznetsov
123 +for reporting this.
124 + * Wasn't locking mspace if all mspaces were locked. Thanks to Pavel Kuznetsov
125 +for reporting this.
126 + * Added Apple Mac OS X support.
127 +
128 +v1.01 24th February 2006:
129 + * Fixed multiprocessor scaling problems by removing sources of cache sloshing
130 + * Earl Chew <earl_chew <at> agilent <dot> com> sent patches for the following:
131 + 1. size2binidx() wasn't working for default code path (non x86)
132 + 2. Fixed failure to release mspace lock under certain circumstances which
133 + caused a deadlock
134 +
135 +v1.00 1st January 2006:
136 + * First release
compat/nedmalloc/malloc.c.h new
+5750
@@ -0,0 +1,5750 @@
1 +/*
2 + This is a version (aka dlmalloc) of malloc/free/realloc written by
3 + Doug Lea and released to the public domain, as explained at
4 + http://creativecommons.org/licenses/publicdomain. Send questions,
5 + comments, complaints, performance data, etc to dl@cs.oswego.edu
6 +
7 +* Version pre-2.8.4 Mon Nov 27 11:22:37 2006 (dl at gee)
8 +
9 + Note: There may be an updated version of this malloc obtainable at
10 + ftp://gee.cs.oswego.edu/pub/misc/malloc.c
11 + Check before installing!
12 +
13 +* Quickstart
14 +
15 + This library is all in one file to simplify the most common usage:
16 + ftp it, compile it (-O3), and link it into another program. All of
17 + the compile-time options default to reasonable values for use on
18 + most platforms. You might later want to step through various
19 + compile-time and dynamic tuning options.
20 +
21 + For convenience, an include file for code using this malloc is at:
22 + ftp://gee.cs.oswego.edu/pub/misc/malloc-2.8.4.h
23 + You don't really need this .h file unless you call functions not
24 + defined in your system include files. The .h file contains only the
25 + excerpts from this file needed for using this malloc on ANSI C/C++
26 + systems, so long as you haven't changed compile-time options about
27 + naming and tuning parameters. If you do, then you can create your
28 + own malloc.h that does include all settings by cutting at the point
29 + indicated below. Note that you may already by default be using a C
30 + library containing a malloc that is based on some version of this
31 + malloc (for example in linux). You might still want to use the one
32 + in this file to customize settings or to avoid overheads associated
33 + with library versions.
34 +
35 +* Vital statistics:
36 +
37 + Supported pointer/size_t representation: 4 or 8 bytes
38 + size_t MUST be an unsigned type of the same width as
39 + pointers. (If you are using an ancient system that declares
40 + size_t as a signed type, or need it to be a different width
41 + than pointers, you can use a previous release of this malloc
42 + (e.g. 2.7.2) supporting these.)
43 +
44 + Alignment: 8 bytes (default)
45 + This suffices for nearly all current machines and C compilers.
46 + However, you can define MALLOC_ALIGNMENT to be wider than this
47 + if necessary (up to 128bytes), at the expense of using more space.
48 +
49 + Minimum overhead per allocated chunk: 4 or 8 bytes (if 4byte sizes)
50 + 8 or 16 bytes (if 8byte sizes)
51 + Each malloced chunk has a hidden word of overhead holding size
52 + and status information, and additional cross-check word
53 + if FOOTERS is defined.
54 +
55 + Minimum allocated size: 4-byte ptrs: 16 bytes (including overhead)
56 + 8-byte ptrs: 32 bytes (including overhead)
57 +
58 + Even a request for zero bytes (i.e., malloc(0)) returns a
59 + pointer to something of the minimum allocatable size.
60 + The maximum overhead wastage (i.e., number of extra bytes
61 + allocated than were requested in malloc) is less than or equal
62 + to the minimum size, except for requests >= mmap_threshold that
63 + are serviced via mmap(), where the worst case wastage is about
64 + 32 bytes plus the remainder from a system page (the minimal
65 + mmap unit); typically 4096 or 8192 bytes.
66 +
67 + Security: static-safe; optionally more or less
68 + The "security" of malloc refers to the ability of malicious
69 + code to accentuate the effects of errors (for example, freeing
70 + space that is not currently malloc'ed or overwriting past the
71 + ends of chunks) in code that calls malloc. This malloc
72 + guarantees not to modify any memory locations below the base of
73 + heap, i.e., static variables, even in the presence of usage
74 + errors. The routines additionally detect most improper frees
75 + and reallocs. All this holds as long as the static bookkeeping
76 + for malloc itself is not corrupted by some other means. This
77 + is only one aspect of security -- these checks do not, and
78 + cannot, detect all possible programming errors.
79 +
80 + If FOOTERS is defined nonzero, then each allocated chunk
81 + carries an additional check word to verify that it was malloced
82 + from its space. These check words are the same within each
83 + execution of a program using malloc, but differ across
84 + executions, so externally crafted fake chunks cannot be
85 + freed. This improves security by rejecting frees/reallocs that
86 + could corrupt heap memory, in addition to the checks preventing
87 + writes to statics that are always on. This may further improve
88 + security at the expense of time and space overhead. (Note that
89 + FOOTERS may also be worth using with MSPACES.)
90 +
91 + By default detected errors cause the program to abort (calling
92 + "abort()"). You can override this to instead proceed past
93 + errors by defining PROCEED_ON_ERROR. In this case, a bad free
94 + has no effect, and a malloc that encounters a bad address
95 + caused by user overwrites will ignore the bad address by
96 + dropping pointers and indices to all known memory. This may
97 + be appropriate for programs that should continue if at all
98 + possible in the face of programming errors, although they may
99 + run out of memory because dropped memory is never reclaimed.
100 +
101 + If you don't like either of these options, you can define
102 + CORRUPTION_ERROR_ACTION and USAGE_ERROR_ACTION to do anything
103 + else. And if if you are sure that your program using malloc has
104 + no errors or vulnerabilities, you can define INSECURE to 1,
105 + which might (or might not) provide a small performance improvement.
106 +
107 + Thread-safety: NOT thread-safe unless USE_LOCKS defined
108 + When USE_LOCKS is defined, each public call to malloc, free,
109 + etc is surrounded with either a pthread mutex or a win32
110 + spinlock (depending on WIN32). This is not especially fast, and
111 + can be a major bottleneck. It is designed only to provide
112 + minimal protection in concurrent environments, and to provide a
113 + basis for extensions. If you are using malloc in a concurrent
114 + program, consider instead using nedmalloc
115 + (http://www.nedprod.com/programs/portable/nedmalloc/) or
116 + ptmalloc (See http://www.malloc.de), which are derived
117 + from versions of this malloc.
118 +
119 + System requirements: Any combination of MORECORE and/or MMAP/MUNMAP
120 + This malloc can use unix sbrk or any emulation (invoked using
121 + the CALL_MORECORE macro) and/or mmap/munmap or any emulation
122 + (invoked using CALL_MMAP/CALL_MUNMAP) to get and release system
123 + memory. On most unix systems, it tends to work best if both
124 + MORECORE and MMAP are enabled. On Win32, it uses emulations
125 + based on VirtualAlloc. It also uses common C library functions
126 + like memset.
127 +
128 + Compliance: I believe it is compliant with the Single Unix Specification
129 + (See http://www.unix.org). Also SVID/XPG, ANSI C, and probably
130 + others as well.
131 +
132 +* Overview of algorithms
133 +
134 + This is not the fastest, most space-conserving, most portable, or
135 + most tunable malloc ever written. However it is among the fastest
136 + while also being among the most space-conserving, portable and
137 + tunable. Consistent balance across these factors results in a good
138 + general-purpose allocator for malloc-intensive programs.
139 +
140 + In most ways, this malloc is a best-fit allocator. Generally, it
141 + chooses the best-fitting existing chunk for a request, with ties
142 + broken in approximately least-recently-used order. (This strategy
143 + normally maintains low fragmentation.) However, for requests less
144 + than 256bytes, it deviates from best-fit when there is not an
145 + exactly fitting available chunk by preferring to use space adjacent
146 + to that used for the previous small request, as well as by breaking
147 + ties in approximately most-recently-used order. (These enhance
148 + locality of series of small allocations.) And for very large requests
149 + (>= 256Kb by default), it relies on system memory mapping
150 + facilities, if supported. (This helps avoid carrying around and
151 + possibly fragmenting memory used only for large chunks.)
152 +
153 + All operations (except malloc_stats and mallinfo) have execution
154 + times that are bounded by a constant factor of the number of bits in
155 + a size_t, not counting any clearing in calloc or copying in realloc,
156 + or actions surrounding MORECORE and MMAP that have times
157 + proportional to the number of non-contiguous regions returned by
158 + system allocation routines, which is often just 1. In real-time
159 + applications, you can optionally suppress segment traversals using
160 + NO_SEGMENT_TRAVERSAL, which assures bounded execution even when
161 + system allocators return non-contiguous spaces, at the typical
162 + expense of carrying around more memory and increased fragmentation.
163 +
164 + The implementation is not very modular and seriously overuses
165 + macros. Perhaps someday all C compilers will do as good a job
166 + inlining modular code as can now be done by brute-force expansion,
167 + but now, enough of them seem not to.
168 +
169 + Some compilers issue a lot of warnings about code that is
170 + dead/unreachable only on some platforms, and also about intentional
171 + uses of negation on unsigned types. All known cases of each can be
172 + ignored.
173 +
174 + For a longer but out of date high-level description, see
175 + http://gee.cs.oswego.edu/dl/html/malloc.html
176 +
177 +* MSPACES
178 + If MSPACES is defined, then in addition to malloc, free, etc.,
179 + this file also defines mspace_malloc, mspace_free, etc. These
180 + are versions of malloc routines that take an "mspace" argument
181 + obtained using create_mspace, to control all internal bookkeeping.
182 + If ONLY_MSPACES is defined, only these versions are compiled.
183 + So if you would like to use this allocator for only some allocations,
184 + and your system malloc for others, you can compile with
185 + ONLY_MSPACES and then do something like...
186 + static mspace mymspace = create_mspace(0,0); // for example
187 + #define mymalloc(bytes) mspace_malloc(mymspace, bytes)
188 +
189 + (Note: If you only need one instance of an mspace, you can instead
190 + use "USE_DL_PREFIX" to relabel the global malloc.)
191 +
192 + You can similarly create thread-local allocators by storing
193 + mspaces as thread-locals. For example:
194 + static __thread mspace tlms = 0;
195 + void* tlmalloc(size_t bytes) {
196 + if (tlms == 0) tlms = create_mspace(0, 0);
197 + return mspace_malloc(tlms, bytes);
198 + }
199 + void tlfree(void* mem) { mspace_free(tlms, mem); }
200 +
201 + Unless FOOTERS is defined, each mspace is completely independent.
202 + You cannot allocate from one and free to another (although
203 + conformance is only weakly checked, so usage errors are not always
204 + caught). If FOOTERS is defined, then each chunk carries around a tag
205 + indicating its originating mspace, and frees are directed to their
206 + originating spaces.
207 +
208 + ------------------------- Compile-time options ---------------------------
209 +
210 +Be careful in setting #define values for numerical constants of type
211 +size_t. On some systems, literal values are not automatically extended
212 +to size_t precision unless they are explicitly casted. You can also
213 +use the symbolic values MAX_SIZE_T, SIZE_T_ONE, etc below.
214 +
215 +WIN32 default: defined if _WIN32 defined
216 + Defining WIN32 sets up defaults for MS environment and compilers.
217 + Otherwise defaults are for unix. Beware that there seem to be some
218 + cases where this malloc might not be a pure drop-in replacement for
219 + Win32 malloc: Random-looking failures from Win32 GDI API's (eg;
220 + SetDIBits()) may be due to bugs in some video driver implementations
221 + when pixel buffers are malloc()ed, and the region spans more than
222 + one VirtualAlloc()ed region. Because dlmalloc uses a small (64Kb)
223 + default granularity, pixel buffers may straddle virtual allocation
224 + regions more often than when using the Microsoft allocator. You can
225 + avoid this by using VirtualAlloc() and VirtualFree() for all pixel
226 + buffers rather than using malloc(). If this is not possible,
227 + recompile this malloc with a larger DEFAULT_GRANULARITY.
228 +
229 +MALLOC_ALIGNMENT default: (size_t)8
230 + Controls the minimum alignment for malloc'ed chunks. It must be a
231 + power of two and at least 8, even on machines for which smaller
232 + alignments would suffice. It may be defined as larger than this
233 + though. Note however that code and data structures are optimized for
234 + the case of 8-byte alignment.
235 +
236 +MSPACES default: 0 (false)
237 + If true, compile in support for independent allocation spaces.
238 + This is only supported if HAVE_MMAP is true.
239 +
240 +ONLY_MSPACES default: 0 (false)
241 + If true, only compile in mspace versions, not regular versions.
242 +
243 +USE_LOCKS default: 0 (false)
244 + Causes each call to each public routine to be surrounded with
245 + pthread or WIN32 mutex lock/unlock. (If set true, this can be
246 + overridden on a per-mspace basis for mspace versions.) If set to a
247 + non-zero value other than 1, locks are used, but their
248 + implementation is left out, so lock functions must be supplied manually.
249 +
250 +USE_SPIN_LOCKS default: 1 iff USE_LOCKS and on x86 using gcc or MSC
251 + If true, uses custom spin locks for locking. This is currently
252 + supported only for x86 platforms using gcc or recent MS compilers.
253 + Otherwise, posix locks or win32 critical sections are used.
254 +
255 +FOOTERS default: 0
256 + If true, provide extra checking and dispatching by placing
257 + information in the footers of allocated chunks. This adds
258 + space and time overhead.
259 +
260 +INSECURE default: 0
261 + If true, omit checks for usage errors and heap space overwrites.
262 +
263 +USE_DL_PREFIX default: NOT defined
264 + Causes compiler to prefix all public routines with the string 'dl'.
265 + This can be useful when you only want to use this malloc in one part
266 + of a program, using your regular system malloc elsewhere.
267 +
268 +ABORT default: defined as abort()
269 + Defines how to abort on failed checks. On most systems, a failed
270 + check cannot die with an "assert" or even print an informative
271 + message, because the underlying print routines in turn call malloc,
272 + which will fail again. Generally, the best policy is to simply call
273 + abort(). It's not very useful to do more than this because many
274 + errors due to overwriting will show up as address faults (null, odd
275 + addresses etc) rather than malloc-triggered checks, so will also
276 + abort. Also, most compilers know that abort() does not return, so
277 + can better optimize code conditionally calling it.
278 +
279 +PROCEED_ON_ERROR default: defined as 0 (false)
280 + Controls whether detected bad addresses cause them to bypassed
281 + rather than aborting. If set, detected bad arguments to free and
282 + realloc are ignored. And all bookkeeping information is zeroed out
283 + upon a detected overwrite of freed heap space, thus losing the
284 + ability to ever return it from malloc again, but enabling the
285 + application to proceed. If PROCEED_ON_ERROR is defined, the
286 + static variable malloc_corruption_error_count is compiled in
287 + and can be examined to see if errors have occurred. This option
288 + generates slower code than the default abort policy.
289 +
290 +DEBUG default: NOT defined
291 + The DEBUG setting is mainly intended for people trying to modify
292 + this code or diagnose problems when porting to new platforms.
293 + However, it may also be able to better isolate user errors than just
294 + using runtime checks. The assertions in the check routines spell
295 + out in more detail the assumptions and invariants underlying the
296 + algorithms. The checking is fairly extensive, and will slow down
297 + execution noticeably. Calling malloc_stats or mallinfo with DEBUG
298 + set will attempt to check every non-mmapped allocated and free chunk
299 + in the course of computing the summaries.
300 +
301 +ABORT_ON_ASSERT_FAILURE default: defined as 1 (true)
302 + Debugging assertion failures can be nearly impossible if your
303 + version of the assert macro causes malloc to be called, which will
304 + lead to a cascade of further failures, blowing the runtime stack.
305 + ABORT_ON_ASSERT_FAILURE cause assertions failures to call abort(),
306 + which will usually make debugging easier.
307 +
308 +MALLOC_FAILURE_ACTION default: sets errno to ENOMEM, or no-op on win32
309 + The action to take before "return 0" when malloc fails to be able to
310 + return memory because there is none available.
311 +
312 +HAVE_MORECORE default: 1 (true) unless win32 or ONLY_MSPACES
313 + True if this system supports sbrk or an emulation of it.
314 +
315 +MORECORE default: sbrk
316 + The name of the sbrk-style system routine to call to obtain more
317 + memory. See below for guidance on writing custom MORECORE
318 + functions. The type of the argument to sbrk/MORECORE varies across
319 + systems. It cannot be size_t, because it supports negative
320 + arguments, so it is normally the signed type of the same width as
321 + size_t (sometimes declared as "intptr_t"). It doesn't much matter
322 + though. Internally, we only call it with arguments less than half
323 + the max value of a size_t, which should work across all reasonable
324 + possibilities, although sometimes generating compiler warnings.
325 +
326 +MORECORE_CONTIGUOUS default: 1 (true) if HAVE_MORECORE
327 + If true, take advantage of fact that consecutive calls to MORECORE
328 + with positive arguments always return contiguous increasing
329 + addresses. This is true of unix sbrk. It does not hurt too much to
330 + set it true anyway, since malloc copes with non-contiguities.
331 + Setting it false when definitely non-contiguous saves time
332 + and possibly wasted space it would take to discover this though.
333 +
334 +MORECORE_CANNOT_TRIM default: NOT defined
335 + True if MORECORE cannot release space back to the system when given
336 + negative arguments. This is generally necessary only if you are
337 + using a hand-crafted MORECORE function that cannot handle negative
338 + arguments.
339 +
340 +NO_SEGMENT_TRAVERSAL default: 0
341 + If non-zero, suppresses traversals of memory segments
342 + returned by either MORECORE or CALL_MMAP. This disables
343 + merging of segments that are contiguous, and selectively
344 + releasing them to the OS if unused, but bounds execution times.
345 +
346 +HAVE_MMAP default: 1 (true)
347 + True if this system supports mmap or an emulation of it. If so, and
348 + HAVE_MORECORE is not true, MMAP is used for all system
349 + allocation. If set and HAVE_MORECORE is true as well, MMAP is
350 + primarily used to directly allocate very large blocks. It is also
351 + used as a backup strategy in cases where MORECORE fails to provide
352 + space from system. Note: A single call to MUNMAP is assumed to be
353 + able to unmap memory that may have be allocated using multiple calls
354 + to MMAP, so long as they are adjacent.
355 +
356 +HAVE_MREMAP default: 1 on linux, else 0
357 + If true realloc() uses mremap() to re-allocate large blocks and
358 + extend or shrink allocation spaces.
359 +
360 +MMAP_CLEARS default: 1 except on WINCE.
361 + True if mmap clears memory so calloc doesn't need to. This is true
362 + for standard unix mmap using /dev/zero and on WIN32 except for WINCE.
363 +
364 +USE_BUILTIN_FFS default: 0 (i.e., not used)
365 + Causes malloc to use the builtin ffs() function to compute indices.
366 + Some compilers may recognize and intrinsify ffs to be faster than the
367 + supplied C version. Also, the case of x86 using gcc is special-cased
368 + to an asm instruction, so is already as fast as it can be, and so
369 + this setting has no effect. Similarly for Win32 under recent MS compilers.
370 + (On most x86s, the asm version is only slightly faster than the C version.)
371 +
372 +malloc_getpagesize default: derive from system includes, or 4096.
373 + The system page size. To the extent possible, this malloc manages
374 + memory from the system in page-size units. This may be (and
375 + usually is) a function rather than a constant. This is ignored
376 + if WIN32, where page size is determined using getSystemInfo during
377 + initialization.
378 +
379 +USE_DEV_RANDOM default: 0 (i.e., not used)
380 + Causes malloc to use /dev/random to initialize secure magic seed for
381 + stamping footers. Otherwise, the current time is used.
382 +
383 +NO_MALLINFO default: 0
384 + If defined, don't compile "mallinfo". This can be a simple way
385 + of dealing with mismatches between system declarations and
386 + those in this file.
387 +
388 +MALLINFO_FIELD_TYPE default: size_t
389 + The type of the fields in the mallinfo struct. This was originally
390 + defined as "int" in SVID etc, but is more usefully defined as
391 + size_t. The value is used only if HAVE_USR_INCLUDE_MALLOC_H is not set
392 +
393 +REALLOC_ZERO_BYTES_FREES default: not defined
394 + This should be set if a call to realloc with zero bytes should
395 + be the same as a call to free. Some people think it should. Otherwise,
396 + since this malloc returns a unique pointer for malloc(0), so does
397 + realloc(p, 0).
398 +
399 +LACKS_UNISTD_H, LACKS_FCNTL_H, LACKS_SYS_PARAM_H, LACKS_SYS_MMAN_H
400 +LACKS_STRINGS_H, LACKS_STRING_H, LACKS_SYS_TYPES_H, LACKS_ERRNO_H
401 +LACKS_STDLIB_H default: NOT defined unless on WIN32
402 + Define these if your system does not have these header files.
403 + You might need to manually insert some of the declarations they provide.
404 +
405 +DEFAULT_GRANULARITY default: page size if MORECORE_CONTIGUOUS,
406 + system_info.dwAllocationGranularity in WIN32,
407 + otherwise 64K.
408 + Also settable using mallopt(M_GRANULARITY, x)
409 + The unit for allocating and deallocating memory from the system. On
410 + most systems with contiguous MORECORE, there is no reason to
411 + make this more than a page. However, systems with MMAP tend to
412 + either require or encourage larger granularities. You can increase
413 + this value to prevent system allocation functions to be called so
414 + often, especially if they are slow. The value must be at least one
415 + page and must be a power of two. Setting to 0 causes initialization
416 + to either page size or win32 region size. (Note: In previous
417 + versions of malloc, the equivalent of this option was called
418 + "TOP_PAD")
419 +
420 +DEFAULT_TRIM_THRESHOLD default: 2MB
421 + Also settable using mallopt(M_TRIM_THRESHOLD, x)
422 + The maximum amount of unused top-most memory to keep before
423 + releasing via malloc_trim in free(). Automatic trimming is mainly
424 + useful in long-lived programs using contiguous MORECORE. Because
425 + trimming via sbrk can be slow on some systems, and can sometimes be
426 + wasteful (in cases where programs immediately afterward allocate
427 + more large chunks) the value should be high enough so that your
428 + overall system performance would improve by releasing this much
429 + memory. As a rough guide, you might set to a value close to the
430 + average size of a process (program) running on your system.
431 + Releasing this much memory would allow such a process to run in
432 + memory. Generally, it is worth tuning trim thresholds when a
433 + program undergoes phases where several large chunks are allocated
434 + and released in ways that can reuse each other's storage, perhaps
435 + mixed with phases where there are no such chunks at all. The trim
436 + value must be greater than page size to have any useful effect. To
437 + disable trimming completely, you can set to MAX_SIZE_T. Note that the trick
438 + some people use of mallocing a huge space and then freeing it at
439 + program startup, in an attempt to reserve system memory, doesn't
440 + have the intended effect under automatic trimming, since that memory
441 + will immediately be returned to the system.
442 +
443 +DEFAULT_MMAP_THRESHOLD default: 256K
444 + Also settable using mallopt(M_MMAP_THRESHOLD, x)
445 + The request size threshold for using MMAP to directly service a
446 + request. Requests of at least this size that cannot be allocated
447 + using already-existing space will be serviced via mmap. (If enough
448 + normal freed space already exists it is used instead.) Using mmap
449 + segregates relatively large chunks of memory so that they can be
450 + individually obtained and released from the host system. A request
451 + serviced through mmap is never reused by any other request (at least
452 + not directly; the system may just so happen to remap successive
453 + requests to the same locations). Segregating space in this way has
454 + the benefits that: Mmapped space can always be individually released
455 + back to the system, which helps keep the system level memory demands
456 + of a long-lived program low. Also, mapped memory doesn't become
457 + `locked' between other chunks, as can happen with normally allocated
458 + chunks, which means that even trimming via malloc_trim would not
459 + release them. However, it has the disadvantage that the space
460 + cannot be reclaimed, consolidated, and then used to service later
461 + requests, as happens with normal chunks. The advantages of mmap
462 + nearly always outweigh disadvantages for "large" chunks, but the
463 + value of "large" may vary across systems. The default is an
464 + empirically derived value that works well in most systems. You can
465 + disable mmap by setting to MAX_SIZE_T.
466 +
467 +MAX_RELEASE_CHECK_RATE default: 4095 unless not HAVE_MMAP
468 + The number of consolidated frees between checks to release
469 + unused segments when freeing. When using non-contiguous segments,
470 + especially with multiple mspaces, checking only for topmost space
471 + doesn't always suffice to trigger trimming. To compensate for this,
472 + free() will, with a period of MAX_RELEASE_CHECK_RATE (or the
473 + current number of segments, if greater) try to release unused
474 + segments to the OS when freeing chunks that result in
475 + consolidation. The best value for this parameter is a compromise
476 + between slowing down frees with relatively costly checks that
477 + rarely trigger versus holding on to unused memory. To effectively
478 + disable, set to MAX_SIZE_T. This may lead to a very slight speed
479 + improvement at the expense of carrying around more memory.
480 +*/
481 +
482 +/* Version identifier to allow people to support multiple versions */
483 +#ifndef DLMALLOC_VERSION
484 +#define DLMALLOC_VERSION 20804
485 +#endif /* DLMALLOC_VERSION */
486 +
487 +#ifndef WIN32
488 +#ifdef _WIN32
489 +#define WIN32 1
490 +#endif /* _WIN32 */
491 +#ifdef _WIN32_WCE
492 +#define LACKS_FCNTL_H
493 +#define WIN32 1
494 +#endif /* _WIN32_WCE */
495 +#endif /* WIN32 */
496 +#ifdef WIN32
497 +#define WIN32_LEAN_AND_MEAN
498 +#define _WIN32_WINNT 0x403
499 +#include <windows.h>
500 +#define HAVE_MMAP 1
501 +#define HAVE_MORECORE 0
502 +#define LACKS_UNISTD_H
503 +#define LACKS_SYS_PARAM_H
504 +#define LACKS_SYS_MMAN_H
505 +#define LACKS_STRING_H
506 +#define LACKS_STRINGS_H
507 +#define LACKS_SYS_TYPES_H
508 +#define LACKS_ERRNO_H
509 +#ifndef MALLOC_FAILURE_ACTION
510 +#define MALLOC_FAILURE_ACTION
511 +#endif /* MALLOC_FAILURE_ACTION */
512 +#ifdef _WIN32_WCE /* WINCE reportedly does not clear */
513 +#define MMAP_CLEARS 0
514 +#else
515 +#define MMAP_CLEARS 1
516 +#endif /* _WIN32_WCE */
517 +#endif /* WIN32 */
518 +
519 +#if defined(DARWIN) || defined(_DARWIN)
520 +/* Mac OSX docs advise not to use sbrk; it seems better to use mmap */
521 +#ifndef HAVE_MORECORE
522 +#define HAVE_MORECORE 0
523 +#define HAVE_MMAP 1
524 +/* OSX allocators provide 16 byte alignment */
525 +#ifndef MALLOC_ALIGNMENT
526 +#define MALLOC_ALIGNMENT ((size_t)16U)
527 +#endif
528 +#endif /* HAVE_MORECORE */
529 +#endif /* DARWIN */
530 +
531 +#ifndef LACKS_SYS_TYPES_H
532 +#include <sys/types.h> /* For size_t */
533 +#endif /* LACKS_SYS_TYPES_H */
534 +
535 +/* The maximum possible size_t value has all bits set */
536 +#define MAX_SIZE_T (~(size_t)0)
537 +
538 +#ifndef ONLY_MSPACES
539 +#define ONLY_MSPACES 0 /* define to a value */
540 +#else
541 +#define ONLY_MSPACES 1
542 +#endif /* ONLY_MSPACES */
543 +#ifndef MSPACES
544 +#if ONLY_MSPACES
545 +#define MSPACES 1
546 +#else /* ONLY_MSPACES */
547 +#define MSPACES 0
548 +#endif /* ONLY_MSPACES */
549 +#endif /* MSPACES */
550 +#ifndef MALLOC_ALIGNMENT
551 +#define MALLOC_ALIGNMENT ((size_t)8U)
552 +#endif /* MALLOC_ALIGNMENT */
553 +#ifndef FOOTERS
554 +#define FOOTERS 0
555 +#endif /* FOOTERS */
556 +#ifndef ABORT
557 +#define ABORT abort()
558 +#endif /* ABORT */
559 +#ifndef ABORT_ON_ASSERT_FAILURE
560 +#define ABORT_ON_ASSERT_FAILURE 1
561 +#endif /* ABORT_ON_ASSERT_FAILURE */
562 +#ifndef PROCEED_ON_ERROR
563 +#define PROCEED_ON_ERROR 0
564 +#endif /* PROCEED_ON_ERROR */
565 +#ifndef USE_LOCKS
566 +#define USE_LOCKS 0
567 +#endif /* USE_LOCKS */
568 +#ifndef USE_SPIN_LOCKS
569 +#if USE_LOCKS && (defined(__GNUC__) && ((defined(__i386__) || defined(__x86_64__)))) || (defined(_MSC_VER) && _MSC_VER>=1310)
570 +#define USE_SPIN_LOCKS 1
571 +#else
572 +#define USE_SPIN_LOCKS 0
573 +#endif /* USE_LOCKS && ... */
574 +#endif /* USE_SPIN_LOCKS */
575 +#ifndef INSECURE
576 +#define INSECURE 0
577 +#endif /* INSECURE */
578 +#ifndef HAVE_MMAP
579 +#define HAVE_MMAP 1
580 +#endif /* HAVE_MMAP */
581 +#ifndef MMAP_CLEARS
582 +#define MMAP_CLEARS 1
583 +#endif /* MMAP_CLEARS */
584 +#ifndef HAVE_MREMAP
585 +#ifdef linux
586 +#define HAVE_MREMAP 1
587 +#else /* linux */
588 +#define HAVE_MREMAP 0
589 +#endif /* linux */
590 +#endif /* HAVE_MREMAP */
591 +#ifndef MALLOC_FAILURE_ACTION
592 +#define MALLOC_FAILURE_ACTION errno = ENOMEM;
593 +#endif /* MALLOC_FAILURE_ACTION */
594 +#ifndef HAVE_MORECORE
595 +#if ONLY_MSPACES
596 +#define HAVE_MORECORE 0
597 +#else /* ONLY_MSPACES */
598 +#define HAVE_MORECORE 1
599 +#endif /* ONLY_MSPACES */
600 +#endif /* HAVE_MORECORE */
601 +#if !HAVE_MORECORE
602 +#define MORECORE_CONTIGUOUS 0
603 +#else /* !HAVE_MORECORE */
604 +#define MORECORE_DEFAULT sbrk
605 +#ifndef MORECORE_CONTIGUOUS
606 +#define MORECORE_CONTIGUOUS 1
607 +#endif /* MORECORE_CONTIGUOUS */
608 +#endif /* HAVE_MORECORE */
609 +#ifndef DEFAULT_GRANULARITY
610 +#if (MORECORE_CONTIGUOUS || defined(WIN32))
611 +#define DEFAULT_GRANULARITY (0) /* 0 means to compute in init_mparams */
612 +#else /* MORECORE_CONTIGUOUS */
613 +#define DEFAULT_GRANULARITY ((size_t)64U * (size_t)1024U)
614 +#endif /* MORECORE_CONTIGUOUS */
615 +#endif /* DEFAULT_GRANULARITY */
616 +#ifndef DEFAULT_TRIM_THRESHOLD
617 +#ifndef MORECORE_CANNOT_TRIM
618 +#define DEFAULT_TRIM_THRESHOLD ((size_t)2U * (size_t)1024U * (size_t)1024U)
619 +#else /* MORECORE_CANNOT_TRIM */
620 +#define DEFAULT_TRIM_THRESHOLD MAX_SIZE_T
621 +#endif /* MORECORE_CANNOT_TRIM */
622 +#endif /* DEFAULT_TRIM_THRESHOLD */
623 +#ifndef DEFAULT_MMAP_THRESHOLD
624 +#if HAVE_MMAP
625 +#define DEFAULT_MMAP_THRESHOLD ((size_t)256U * (size_t)1024U)
626 +#else /* HAVE_MMAP */
627 +#define DEFAULT_MMAP_THRESHOLD MAX_SIZE_T
628 +#endif /* HAVE_MMAP */
629 +#endif /* DEFAULT_MMAP_THRESHOLD */
630 +#ifndef MAX_RELEASE_CHECK_RATE
631 +#if HAVE_MMAP
632 +#define MAX_RELEASE_CHECK_RATE 4095
633 +#else
634 +#define MAX_RELEASE_CHECK_RATE MAX_SIZE_T
635 +#endif /* HAVE_MMAP */
636 +#endif /* MAX_RELEASE_CHECK_RATE */
637 +#ifndef USE_BUILTIN_FFS
638 +#define USE_BUILTIN_FFS 0
639 +#endif /* USE_BUILTIN_FFS */
640 +#ifndef USE_DEV_RANDOM
641 +#define USE_DEV_RANDOM 0
642 +#endif /* USE_DEV_RANDOM */
643 +#ifndef NO_MALLINFO
644 +#define NO_MALLINFO 0
645 +#endif /* NO_MALLINFO */
646 +#ifndef MALLINFO_FIELD_TYPE
647 +#define MALLINFO_FIELD_TYPE size_t
648 +#endif /* MALLINFO_FIELD_TYPE */
649 +#ifndef NO_SEGMENT_TRAVERSAL
650 +#define NO_SEGMENT_TRAVERSAL 0
651 +#endif /* NO_SEGMENT_TRAVERSAL */
652 +
653 +/*
654 + mallopt tuning options. SVID/XPG defines four standard parameter
655 + numbers for mallopt, normally defined in malloc.h. None of these
656 + are used in this malloc, so setting them has no effect. But this
657 + malloc does support the following options.
658 +*/
659 +
660 +#define M_TRIM_THRESHOLD (-1)
661 +#define M_GRANULARITY (-2)
662 +#define M_MMAP_THRESHOLD (-3)
663 +
664 +/* ------------------------ Mallinfo declarations ------------------------ */
665 +
666 +#if !NO_MALLINFO
667 +/*
668 + This version of malloc supports the standard SVID/XPG mallinfo
669 + routine that returns a struct containing usage properties and
670 + statistics. It should work on any system that has a
671 + /usr/include/malloc.h defining struct mallinfo. The main
672 + declaration needed is the mallinfo struct that is returned (by-copy)
673 + by mallinfo(). The malloinfo struct contains a bunch of fields that
674 + are not even meaningful in this version of malloc. These fields are
675 + are instead filled by mallinfo() with other numbers that might be of
676 + interest.
677 +
678 + HAVE_USR_INCLUDE_MALLOC_H should be set if you have a
679 + /usr/include/malloc.h file that includes a declaration of struct
680 + mallinfo. If so, it is included; else a compliant version is
681 + declared below. These must be precisely the same for mallinfo() to
682 + work. The original SVID version of this struct, defined on most
683 + systems with mallinfo, declares all fields as ints. But some others
684 + define as unsigned long. If your system defines the fields using a
685 + type of different width than listed here, you MUST #include your
686 + system version and #define HAVE_USR_INCLUDE_MALLOC_H.
687 +*/
688 +
689 +/* #define HAVE_USR_INCLUDE_MALLOC_H */
690 +
691 +#ifdef HAVE_USR_INCLUDE_MALLOC_H
692 +#include "/usr/include/malloc.h"
693 +#else /* HAVE_USR_INCLUDE_MALLOC_H */
694 +#ifndef STRUCT_MALLINFO_DECLARED
695 +#define STRUCT_MALLINFO_DECLARED 1
696 +struct mallinfo {
697 + MALLINFO_FIELD_TYPE arena; /* non-mmapped space allocated from system */
698 + MALLINFO_FIELD_TYPE ordblks; /* number of free chunks */
699 + MALLINFO_FIELD_TYPE smblks; /* always 0 */
700 + MALLINFO_FIELD_TYPE hblks; /* always 0 */
701 + MALLINFO_FIELD_TYPE hblkhd; /* space in mmapped regions */
702 + MALLINFO_FIELD_TYPE usmblks; /* maximum total allocated space */
703 + MALLINFO_FIELD_TYPE fsmblks; /* always 0 */
704 + MALLINFO_FIELD_TYPE uordblks; /* total allocated space */
705 + MALLINFO_FIELD_TYPE fordblks; /* total free space */
706 + MALLINFO_FIELD_TYPE keepcost; /* releasable (via malloc_trim) space */
707 +};
708 +#endif /* STRUCT_MALLINFO_DECLARED */
709 +#endif /* HAVE_USR_INCLUDE_MALLOC_H */
710 +#endif /* NO_MALLINFO */
711 +
712 +/*
713 + Try to persuade compilers to inline. The most critical functions for
714 + inlining are defined as macros, so these aren't used for them.
715 +*/
716 +
717 +#ifndef FORCEINLINE
718 + #if defined(__GNUC__)
719 +#define FORCEINLINE __inline __attribute__ ((always_inline))
720 + #elif defined(_MSC_VER)
721 + #define FORCEINLINE __forceinline
722 + #endif
723 +#endif
724 +#ifndef NOINLINE
725 + #if defined(__GNUC__)
726 + #define NOINLINE __attribute__ ((noinline))
727 + #elif defined(_MSC_VER)
728 + #define NOINLINE __declspec(noinline)
729 + #else
730 + #define NOINLINE
731 + #endif
732 +#endif
733 +
734 +#ifdef __cplusplus
735 +extern "C" {
736 +#ifndef FORCEINLINE
737 + #define FORCEINLINE inline
738 +#endif
739 +#endif /* __cplusplus */
740 +#ifndef FORCEINLINE
741 + #define FORCEINLINE
742 +#endif
743 +
744 +#if !ONLY_MSPACES
745 +
746 +/* ------------------- Declarations of public routines ------------------- */
747 +
748 +#ifndef USE_DL_PREFIX
749 +#define dlcalloc calloc
750 +#define dlfree free
751 +#define dlmalloc malloc
752 +#define dlmemalign memalign
753 +#define dlrealloc realloc
754 +#define dlvalloc valloc
755 +#define dlpvalloc pvalloc
756 +#define dlmallinfo mallinfo
757 +#define dlmallopt mallopt
758 +#define dlmalloc_trim malloc_trim
759 +#define dlmalloc_stats malloc_stats
760 +#define dlmalloc_usable_size malloc_usable_size
761 +#define dlmalloc_footprint malloc_footprint
762 +#define dlmalloc_max_footprint malloc_max_footprint
763 +#define dlindependent_calloc independent_calloc
764 +#define dlindependent_comalloc independent_comalloc
765 +#endif /* USE_DL_PREFIX */
766 +
767 +
768 +/*
769 + malloc(size_t n)
770 + Returns a pointer to a newly allocated chunk of at least n bytes, or
771 + null if no space is available, in which case errno is set to ENOMEM
772 + on ANSI C systems.
773 +
774 + If n is zero, malloc returns a minimum-sized chunk. (The minimum
775 + size is 16 bytes on most 32bit systems, and 32 bytes on 64bit
776 + systems.) Note that size_t is an unsigned type, so calls with
777 + arguments that would be negative if signed are interpreted as
778 + requests for huge amounts of space, which will often fail. The
779 + maximum supported value of n differs across systems, but is in all
780 + cases less than the maximum representable value of a size_t.
781 +*/
782 +void* dlmalloc(size_t);
783 +
784 +/*
785 + free(void* p)
786 + Releases the chunk of memory pointed to by p, that had been previously
787 + allocated using malloc or a related routine such as realloc.
788 + It has no effect if p is null. If p was not malloced or already
789 + freed, free(p) will by default cause the current program to abort.
790 +*/
791 +void dlfree(void*);
792 +
793 +/*
794 + calloc(size_t n_elements, size_t element_size);
795 + Returns a pointer to n_elements * element_size bytes, with all locations
796 + set to zero.
797 +*/
798 +void* dlcalloc(size_t, size_t);
799 +
800 +/*
801 + realloc(void* p, size_t n)
802 + Returns a pointer to a chunk of size n that contains the same data
803 + as does chunk p up to the minimum of (n, p's size) bytes, or null
804 + if no space is available.
805 +
806 + The returned pointer may or may not be the same as p. The algorithm
807 + prefers extending p in most cases when possible, otherwise it
808 + employs the equivalent of a malloc-copy-free sequence.
809 +
810 + If p is null, realloc is equivalent to malloc.
811 +
812 + If space is not available, realloc returns null, errno is set (if on
813 + ANSI) and p is NOT freed.
814 +
815 + if n is for fewer bytes than already held by p, the newly unused
816 + space is lopped off and freed if possible. realloc with a size
817 + argument of zero (re)allocates a minimum-sized chunk.
818 +
819 + The old unix realloc convention of allowing the last-free'd chunk
820 + to be used as an argument to realloc is not supported.
821 +*/
822 +
823 +void* dlrealloc(void*, size_t);
824 +
825 +/*
826 + memalign(size_t alignment, size_t n);
827 + Returns a pointer to a newly allocated chunk of n bytes, aligned
828 + in accord with the alignment argument.
829 +
830 + The alignment argument should be a power of two. If the argument is
831 + not a power of two, the nearest greater power is used.
832 + 8-byte alignment is guaranteed by normal malloc calls, so don't
833 + bother calling memalign with an argument of 8 or less.
834 +
835 + Overreliance on memalign is a sure way to fragment space.
836 +*/
837 +void* dlmemalign(size_t, size_t);
838 +
839 +/*
840 + valloc(size_t n);
841 + Equivalent to memalign(pagesize, n), where pagesize is the page
842 + size of the system. If the pagesize is unknown, 4096 is used.
843 +*/
844 +void* dlvalloc(size_t);
845 +
846 +/*
847 + mallopt(int parameter_number, int parameter_value)
848 + Sets tunable parameters The format is to provide a
849 + (parameter-number, parameter-value) pair. mallopt then sets the
850 + corresponding parameter to the argument value if it can (i.e., so
851 + long as the value is meaningful), and returns 1 if successful else
852 + 0. To workaround the fact that mallopt is specified to use int,
853 + not size_t parameters, the value -1 is specially treated as the
854 + maximum unsigned size_t value.
855 +
856 + SVID/XPG/ANSI defines four standard param numbers for mallopt,
857 + normally defined in malloc.h. None of these are use in this malloc,
858 + so setting them has no effect. But this malloc also supports other
859 + options in mallopt. See below for details. Briefly, supported
860 + parameters are as follows (listed defaults are for "typical"
861 + configurations).
862 +
863 + Symbol param # default allowed param values
864 + M_TRIM_THRESHOLD -1 2*1024*1024 any (-1 disables)
865 + M_GRANULARITY -2 page size any power of 2 >= page size
866 + M_MMAP_THRESHOLD -3 256*1024 any (or 0 if no MMAP support)
867 +*/
868 +int dlmallopt(int, int);
869 +
870 +/*
871 + malloc_footprint();
872 + Returns the number of bytes obtained from the system. The total
873 + number of bytes allocated by malloc, realloc etc., is less than this
874 + value. Unlike mallinfo, this function returns only a precomputed
875 + result, so can be called frequently to monitor memory consumption.
876 + Even if locks are otherwise defined, this function does not use them,
877 + so results might not be up to date.
878 +*/
879 +size_t dlmalloc_footprint(void);
880 +
881 +/*
882 + malloc_max_footprint();
883 + Returns the maximum number of bytes obtained from the system. This
884 + value will be greater than current footprint if deallocated space
885 + has been reclaimed by the system. The peak number of bytes allocated
886 + by malloc, realloc etc., is less than this value. Unlike mallinfo,
887 + this function returns only a precomputed result, so can be called
888 + frequently to monitor memory consumption. Even if locks are
889 + otherwise defined, this function does not use them, so results might
890 + not be up to date.
891 +*/
892 +size_t dlmalloc_max_footprint(void);
893 +
894 +#if !NO_MALLINFO
895 +/*
896 + mallinfo()
897 + Returns (by copy) a struct containing various summary statistics:
898 +
899 + arena: current total non-mmapped bytes allocated from system
900 + ordblks: the number of free chunks
901 + smblks: always zero.
902 + hblks: current number of mmapped regions
903 + hblkhd: total bytes held in mmapped regions
904 + usmblks: the maximum total allocated space. This will be greater
905 + than current total if trimming has occurred.
906 + fsmblks: always zero
907 + uordblks: current total allocated space (normal or mmapped)
908 + fordblks: total free space
909 + keepcost: the maximum number of bytes that could ideally be released
910 + back to system via malloc_trim. ("ideally" means that
911 + it ignores page restrictions etc.)
912 +
913 + Because these fields are ints, but internal bookkeeping may
914 + be kept as longs, the reported values may wrap around zero and
915 + thus be inaccurate.
916 +*/
917 +struct mallinfo dlmallinfo(void);
918 +#endif /* NO_MALLINFO */
919 +
920 +/*
921 + independent_calloc(size_t n_elements, size_t element_size, void* chunks[]);
922 +
923 + independent_calloc is similar to calloc, but instead of returning a
924 + single cleared space, it returns an array of pointers to n_elements
925 + independent elements that can hold contents of size elem_size, each
926 + of which starts out cleared, and can be independently freed,
927 + realloc'ed etc. The elements are guaranteed to be adjacently
928 + allocated (this is not guaranteed to occur with multiple callocs or
929 + mallocs), which may also improve cache locality in some
930 + applications.
931 +
932 + The "chunks" argument is optional (i.e., may be null, which is
933 + probably the most typical usage). If it is null, the returned array
934 + is itself dynamically allocated and should also be freed when it is
935 + no longer needed. Otherwise, the chunks array must be of at least
936 + n_elements in length. It is filled in with the pointers to the
937 + chunks.
938 +
939 + In either case, independent_calloc returns this pointer array, or
940 + null if the allocation failed. If n_elements is zero and "chunks"
941 + is null, it returns a chunk representing an array with zero elements
942 + (which should be freed if not wanted).
943 +
944 + Each element must be individually freed when it is no longer
945 + needed. If you'd like to instead be able to free all at once, you
946 + should instead use regular calloc and assign pointers into this
947 + space to represent elements. (In this case though, you cannot
948 + independently free elements.)
949 +
950 + independent_calloc simplifies and speeds up implementations of many
951 + kinds of pools. It may also be useful when constructing large data
952 + structures that initially have a fixed number of fixed-sized nodes,
953 + but the number is not known at compile time, and some of the nodes
954 + may later need to be freed. For example:
955 +
956 + struct Node { int item; struct Node* next; };
957 +
958 + struct Node* build_list() {
959 + struct Node** pool;
960 + int n = read_number_of_nodes_needed();
961 + if (n <= 0) return 0;
962 + pool = (struct Node**)(independent_calloc(n, sizeof(struct Node), 0);
963 + if (pool == 0) die();
964 + // organize into a linked list...
965 + struct Node* first = pool[0];
966 + for (i = 0; i < n-1; ++i)
967 + pool[i]->next = pool[i+1];
968 + free(pool); // Can now free the array (or not, if it is needed later)
969 + return first;
970 + }
971 +*/
972 +void** dlindependent_calloc(size_t, size_t, void**);
973 +
974 +/*
975 + independent_comalloc(size_t n_elements, size_t sizes[], void* chunks[]);
976 +
977 + independent_comalloc allocates, all at once, a set of n_elements
978 + chunks with sizes indicated in the "sizes" array. It returns
979 + an array of pointers to these elements, each of which can be
980 + independently freed, realloc'ed etc. The elements are guaranteed to
981 + be adjacently allocated (this is not guaranteed to occur with
982 + multiple callocs or mallocs), which may also improve cache locality
983 + in some applications.
984 +
985 + The "chunks" argument is optional (i.e., may be null). If it is null
986 + the returned array is itself dynamically allocated and should also
987 + be freed when it is no longer needed. Otherwise, the chunks array
988 + must be of at least n_elements in length. It is filled in with the
989 + pointers to the chunks.
990 +
991 + In either case, independent_comalloc returns this pointer array, or
992 + null if the allocation failed. If n_elements is zero and chunks is
993 + null, it returns a chunk representing an array with zero elements
994 + (which should be freed if not wanted).
995 +
996 + Each element must be individually freed when it is no longer
997 + needed. If you'd like to instead be able to free all at once, you
998 + should instead use a single regular malloc, and assign pointers at
999 + particular offsets in the aggregate space. (In this case though, you
1000 + cannot independently free elements.)
1001 +
1002 + independent_comallac differs from independent_calloc in that each
1003 + element may have a different size, and also that it does not
1004 + automatically clear elements.
1005 +
1006 + independent_comalloc can be used to speed up allocation in cases
1007 + where several structs or objects must always be allocated at the
1008 + same time. For example:
1009 +
1010 + struct Head { ... }
1011 + struct Foot { ... }
1012 +
1013 + void send_message(char* msg) {
1014 + int msglen = strlen(msg);
1015 + size_t sizes[3] = { sizeof(struct Head), msglen, sizeof(struct Foot) };
1016 + void* chunks[3];
1017 + if (independent_comalloc(3, sizes, chunks) == 0)
1018 + die();
1019 + struct Head* head = (struct Head*)(chunks[0]);
1020 + char* body = (char*)(chunks[1]);
1021 + struct Foot* foot = (struct Foot*)(chunks[2]);
1022 + // ...
1023 + }
1024 +
1025 + In general though, independent_comalloc is worth using only for
1026 + larger values of n_elements. For small values, you probably won't
1027 + detect enough difference from series of malloc calls to bother.
1028 +
1029 + Overuse of independent_comalloc can increase overall memory usage,
1030 + since it cannot reuse existing noncontiguous small chunks that
1031 + might be available for some of the elements.
1032 +*/
1033 +void** dlindependent_comalloc(size_t, size_t*, void**);
1034 +
1035 +
1036 +/*
1037 + pvalloc(size_t n);
1038 + Equivalent to valloc(minimum-page-that-holds(n)), that is,
1039 + round up n to nearest pagesize.
1040 + */
1041 +void* dlpvalloc(size_t);
1042 +
1043 +/*
1044 + malloc_trim(size_t pad);
1045 +
1046 + If possible, gives memory back to the system (via negative arguments
1047 + to sbrk) if there is unused memory at the `high' end of the malloc
1048 + pool or in unused MMAP segments. You can call this after freeing
1049 + large blocks of memory to potentially reduce the system-level memory
1050 + requirements of a program. However, it cannot guarantee to reduce
1051 + memory. Under some allocation patterns, some large free blocks of
1052 + memory will be locked between two used chunks, so they cannot be
1053 + given back to the system.
1054 +
1055 + The `pad' argument to malloc_trim represents the amount of free
1056 + trailing space to leave untrimmed. If this argument is zero, only
1057 + the minimum amount of memory to maintain internal data structures
1058 + will be left. Non-zero arguments can be supplied to maintain enough
1059 + trailing space to service future expected allocations without having
1060 + to re-obtain memory from the system.
1061 +
1062 + Malloc_trim returns 1 if it actually released any memory, else 0.
1063 +*/
1064 +int dlmalloc_trim(size_t);
1065 +
1066 +/*
1067 + malloc_stats();
1068 + Prints on stderr the amount of space obtained from the system (both
1069 + via sbrk and mmap), the maximum amount (which may be more than
1070 + current if malloc_trim and/or munmap got called), and the current
1071 + number of bytes allocated via malloc (or realloc, etc) but not yet
1072 + freed. Note that this is the number of bytes allocated, not the
1073 + number requested. It will be larger than the number requested
1074 + because of alignment and bookkeeping overhead. Because it includes
1075 + alignment wastage as being in use, this figure may be greater than
1076 + zero even when no user-level chunks are allocated.
1077 +
1078 + The reported current and maximum system memory can be inaccurate if
1079 + a program makes other calls to system memory allocation functions
1080 + (normally sbrk) outside of malloc.
1081 +
1082 + malloc_stats prints only the most commonly interesting statistics.
1083 + More information can be obtained by calling mallinfo.
1084 +*/
1085 +void dlmalloc_stats(void);
1086 +
1087 +#endif /* ONLY_MSPACES */
1088 +
1089 +/*
1090 + malloc_usable_size(void* p);
1091 +
1092 + Returns the number of bytes you can actually use in
1093 + an allocated chunk, which may be more than you requested (although
1094 + often not) due to alignment and minimum size constraints.
1095 + You can use this many bytes without worrying about
1096 + overwriting other allocated objects. This is not a particularly great
1097 + programming practice. malloc_usable_size can be more useful in
1098 + debugging and assertions, for example:
1099 +
1100 + p = malloc(n);
1101 + assert(malloc_usable_size(p) >= 256);
1102 +*/
1103 +size_t dlmalloc_usable_size(void*);
1104 +
1105 +
1106 +#if MSPACES
1107 +
1108 +/*
1109 + mspace is an opaque type representing an independent
1110 + region of space that supports mspace_malloc, etc.
1111 +*/
1112 +typedef void* mspace;
1113 +
1114 +/*
1115 + create_mspace creates and returns a new independent space with the
1116 + given initial capacity, or, if 0, the default granularity size. It
1117 + returns null if there is no system memory available to create the
1118 + space. If argument locked is non-zero, the space uses a separate
1119 + lock to control access. The capacity of the space will grow
1120 + dynamically as needed to service mspace_malloc requests. You can
1121 + control the sizes of incremental increases of this space by
1122 + compiling with a different DEFAULT_GRANULARITY or dynamically
1123 + setting with mallopt(M_GRANULARITY, value).
1124 +*/
1125 +mspace create_mspace(size_t capacity, int locked);
1126 +
1127 +/*
1128 + destroy_mspace destroys the given space, and attempts to return all
1129 + of its memory back to the system, returning the total number of
1130 + bytes freed. After destruction, the results of access to all memory
1131 + used by the space become undefined.
1132 +*/
1133 +size_t destroy_mspace(mspace msp);
1134 +
1135 +/*
1136 + create_mspace_with_base uses the memory supplied as the initial base
1137 + of a new mspace. Part (less than 128*sizeof(size_t) bytes) of this
1138 + space is used for bookkeeping, so the capacity must be at least this
1139 + large. (Otherwise 0 is returned.) When this initial space is
1140 + exhausted, additional memory will be obtained from the system.
1141 + Destroying this space will deallocate all additionally allocated
1142 + space (if possible) but not the initial base.
1143 +*/
1144 +mspace create_mspace_with_base(void* base, size_t capacity, int locked);
1145 +
1146 +/*
1147 + mspace_mmap_large_chunks controls whether requests for large chunks
1148 + are allocated in their own mmapped regions, separate from others in
1149 + this mspace. By default this is enabled, which reduces
1150 + fragmentation. However, such chunks are not necessarily released to
1151 + the system upon destroy_mspace. Disabling by setting to false may
1152 + increase fragmentation, but avoids leakage when relying on
1153 + destroy_mspace to release all memory allocated using this space.
1154 +*/
1155 +int mspace_mmap_large_chunks(mspace msp, int enable);
1156 +
1157 +
1158 +/*
1159 + mspace_malloc behaves as malloc, but operates within
1160 + the given space.
1161 +*/
1162 +void* mspace_malloc(mspace msp, size_t bytes);
1163 +
1164 +/*
1165 + mspace_free behaves as free, but operates within
1166 + the given space.
1167 +
1168 + If compiled with FOOTERS==1, mspace_free is not actually needed.
1169 + free may be called instead of mspace_free because freed chunks from
1170 + any space are handled by their originating spaces.
1171 +*/
1172 +void mspace_free(mspace msp, void* mem);
1173 +
1174 +/*
1175 + mspace_realloc behaves as realloc, but operates within
1176 + the given space.
1177 +
1178 + If compiled with FOOTERS==1, mspace_realloc is not actually
1179 + needed. realloc may be called instead of mspace_realloc because
1180 + realloced chunks from any space are handled by their originating
1181 + spaces.
1182 +*/
1183 +void* mspace_realloc(mspace msp, void* mem, size_t newsize);
1184 +
1185 +/*
1186 + mspace_calloc behaves as calloc, but operates within
1187 + the given space.
1188 +*/
1189 +void* mspace_calloc(mspace msp, size_t n_elements, size_t elem_size);
1190 +
1191 +/*
1192 + mspace_memalign behaves as memalign, but operates within
1193 + the given space.
1194 +*/
1195 +void* mspace_memalign(mspace msp, size_t alignment, size_t bytes);
1196 +
1197 +/*
1198 + mspace_independent_calloc behaves as independent_calloc, but
1199 + operates within the given space.
1200 +*/
1201 +void** mspace_independent_calloc(mspace msp, size_t n_elements,
1202 + size_t elem_size, void* chunks[]);
1203 +
1204 +/*
1205 + mspace_independent_comalloc behaves as independent_comalloc, but
1206 + operates within the given space.
1207 +*/
1208 +void** mspace_independent_comalloc(mspace msp, size_t n_elements,
1209 + size_t sizes[], void* chunks[]);
1210 +
1211 +/*
1212 + mspace_footprint() returns the number of bytes obtained from the
1213 + system for this space.
1214 +*/
1215 +size_t mspace_footprint(mspace msp);
1216 +
1217 +/*
1218 + mspace_max_footprint() returns the peak number of bytes obtained from the
1219 + system for this space.
1220 +*/
1221 +size_t mspace_max_footprint(mspace msp);
1222 +
1223 +
1224 +#if !NO_MALLINFO
1225 +/*
1226 + mspace_mallinfo behaves as mallinfo, but reports properties of
1227 + the given space.
1228 +*/
1229 +struct mallinfo mspace_mallinfo(mspace msp);
1230 +#endif /* NO_MALLINFO */
1231 +
1232 +/*
1233 + malloc_usable_size(void* p) behaves the same as malloc_usable_size;
1234 +*/
1235 + size_t mspace_usable_size(void* mem);
1236 +
1237 +/*
1238 + mspace_malloc_stats behaves as malloc_stats, but reports
1239 + properties of the given space.
1240 +*/
1241 +void mspace_malloc_stats(mspace msp);
1242 +
1243 +/*
1244 + mspace_trim behaves as malloc_trim, but
1245 + operates within the given space.
1246 +*/
1247 +int mspace_trim(mspace msp, size_t pad);
1248 +
1249 +/*
1250 + An alias for mallopt.
1251 +*/
1252 +int mspace_mallopt(int, int);
1253 +
1254 +#endif /* MSPACES */
1255 +
1256 +#ifdef __cplusplus
1257 +}; /* end of extern "C" */
1258 +#endif /* __cplusplus */
1259 +
1260 +/*
1261 + ========================================================================
1262 + To make a fully customizable malloc.h header file, cut everything
1263 + above this line, put into file malloc.h, edit to suit, and #include it
1264 + on the next line, as well as in programs that use this malloc.
1265 + ========================================================================
1266 +*/
1267 +
1268 +/* #include "malloc.h" */
1269 +
1270 +/*------------------------------ internal #includes ---------------------- */
1271 +
1272 +#ifdef WIN32
1273 +#pragma warning( disable : 4146 ) /* no "unsigned" warnings */
1274 +#endif /* WIN32 */
1275 +
1276 +#include <stdio.h> /* for printing in malloc_stats */
1277 +
1278 +#ifndef LACKS_ERRNO_H
1279 +#include <errno.h> /* for MALLOC_FAILURE_ACTION */
1280 +#endif /* LACKS_ERRNO_H */
1281 +#if FOOTERS
1282 +#include <time.h> /* for magic initialization */
1283 +#endif /* FOOTERS */
1284 +#ifndef LACKS_STDLIB_H
1285 +#include <stdlib.h> /* for abort() */
1286 +#endif /* LACKS_STDLIB_H */
1287 +#ifdef DEBUG
1288 +#if ABORT_ON_ASSERT_FAILURE
1289 +#define assert(x) if(!(x)) ABORT
1290 +#else /* ABORT_ON_ASSERT_FAILURE */
1291 +#include <assert.h>
1292 +#endif /* ABORT_ON_ASSERT_FAILURE */
1293 +#else /* DEBUG */
1294 +#ifndef assert
1295 +#define assert(x)
1296 +#endif
1297 +#define DEBUG 0
1298 +#endif /* DEBUG */
1299 +#ifndef LACKS_STRING_H
1300 +#include <string.h> /* for memset etc */
1301 +#endif /* LACKS_STRING_H */
1302 +#if USE_BUILTIN_FFS
1303 +#ifndef LACKS_STRINGS_H
1304 +#include <strings.h> /* for ffs */
1305 +#endif /* LACKS_STRINGS_H */
1306 +#endif /* USE_BUILTIN_FFS */
1307 +#if HAVE_MMAP
1308 +#ifndef LACKS_SYS_MMAN_H
1309 +#include <sys/mman.h> /* for mmap */
1310 +#endif /* LACKS_SYS_MMAN_H */
1311 +#ifndef LACKS_FCNTL_H
1312 +#include <fcntl.h>
1313 +#endif /* LACKS_FCNTL_H */
1314 +#endif /* HAVE_MMAP */
1315 +#ifndef LACKS_UNISTD_H
1316 +#include <unistd.h> /* for sbrk, sysconf */
1317 +#else /* LACKS_UNISTD_H */
1318 +#if !defined(__FreeBSD__) && !defined(__OpenBSD__) && !defined(__NetBSD__)
1319 +extern void* sbrk(ptrdiff_t);
1320 +#endif /* FreeBSD etc */
1321 +#endif /* LACKS_UNISTD_H */
1322 +
1323 +/* Declarations for locking */
1324 +#if USE_LOCKS
1325 +#ifndef WIN32
1326 +#include <pthread.h>
1327 +#if defined (__SVR4) && defined (__sun) /* solaris */
1328 +#include <thread.h>
1329 +#endif /* solaris */
1330 +#else
1331 +#ifndef _M_AMD64
1332 +/* These are already defined on AMD64 builds */
1333 +#ifdef __cplusplus
1334 +extern "C" {
1335 +#endif /* __cplusplus */
1336 +#ifndef __MINGW32__
1337 +LONG __cdecl _InterlockedCompareExchange(LONG volatile *Dest, LONG Exchange, LONG Comp);
1338 +LONG __cdecl _InterlockedExchange(LONG volatile *Target, LONG Value);
1339 +#endif
1340 +#ifdef __cplusplus
1341 +}
1342 +#endif /* __cplusplus */
1343 +#endif /* _M_AMD64 */
1344 +#ifndef __MINGW32__
1345 +#pragma intrinsic (_InterlockedCompareExchange)
1346 +#pragma intrinsic (_InterlockedExchange)
1347 +#else
1348 + /* --[ start GCC compatibility ]----------------------------------------------
1349 + * Compatibility <intrin_x86.h> header for GCC -- GCC equivalents of intrinsic
1350 + * Microsoft Visual C++ functions. Originally developed for the ReactOS
1351 + * (<http://www.reactos.org/>) and TinyKrnl (<http://www.tinykrnl.org/>)
1352 + * projects.
1353 + *
1354 + * Copyright (c) 2006 KJK::Hyperion <hackbunny@reactos.com>
1355 + *
1356 + * Permission is hereby granted, free of charge, to any person obtaining a
1357 + * copy of this software and associated documentation files (the "Software"),
1358 + * to deal in the Software without restriction, including without limitation
1359 + * the rights to use, copy, modify, merge, publish, distribute, sublicense,
1360 + * and/or sell copies of the Software, and to permit persons to whom the
1361 + * Software is furnished to do so, subject to the following conditions:
1362 + *
1363 + * The above copyright notice and this permission notice shall be included in
1364 + * all copies or substantial portions of the Software.
1365 + *
1366 + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
1367 + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
1368 + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
1369 + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
1370 + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
1371 + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
1372 + * DEALINGS IN THE SOFTWARE.
1373 + */
1374 +
1375 + /*** Atomic operations ***/
1376 + #if (__GNUC__ * 10000 + __GNUC_MINOR__ * 100 + __GNUC_PATCHLEVEL__) > 40100
1377 + #define _ReadWriteBarrier() __sync_synchronize()
1378 + #else
1379 + static __inline__ __attribute__((always_inline)) long __sync_lock_test_and_set(volatile long * const Target, const long Value)
1380 + {
1381 + long res;
1382 + __asm__ __volatile__("xchg%z0 %2, %0" : "=g" (*(Target)), "=r" (res) : "1" (Value));
1383 + return res;
1384 + }
1385 + static void __inline__ __attribute__((always_inline)) _MemoryBarrier(void)
1386 + {
1387 + __asm__ __volatile__("" : : : "memory");
1388 + }
1389 + #define _ReadWriteBarrier() _MemoryBarrier()
1390 + #endif
1391 + /* BUGBUG: GCC only supports full barriers */
1392 + static __inline__ __attribute__((always_inline)) long _InterlockedExchange(volatile long * const Target, const long Value)
1393 + {
1394 + /* NOTE: __sync_lock_test_and_set would be an acquire barrier, so we force a full barrier */
1395 + _ReadWriteBarrier();
1396 + return __sync_lock_test_and_set(Target, Value);
1397 + }
1398 + /* --[ end GCC compatibility ]---------------------------------------------- */
1399 +#endif
1400 +#define interlockedcompareexchange _InterlockedCompareExchange
1401 +#define interlockedexchange _InterlockedExchange
1402 +#endif /* Win32 */
1403 +#endif /* USE_LOCKS */
1404 +
1405 +/* Declarations for bit scanning on win32 */
1406 +#if defined(_MSC_VER) && _MSC_VER>=1300
1407 +#ifndef BitScanForward /* Try to avoid pulling in WinNT.h */
1408 +#ifdef __cplusplus
1409 +extern "C" {
1410 +#endif /* __cplusplus */
1411 +unsigned char _BitScanForward(unsigned long *index, unsigned long mask);
1412 +unsigned char _BitScanReverse(unsigned long *index, unsigned long mask);
1413 +#ifdef __cplusplus
1414 +}
1415 +#endif /* __cplusplus */
1416 +
1417 +#define BitScanForward _BitScanForward
1418 +#define BitScanReverse _BitScanReverse
1419 +#pragma intrinsic(_BitScanForward)
1420 +#pragma intrinsic(_BitScanReverse)
1421 +#endif /* BitScanForward */
1422 +#endif /* defined(_MSC_VER) && _MSC_VER>=1300 */
1423 +
1424 +#ifndef WIN32
1425 +#ifndef malloc_getpagesize
1426 +# ifdef _SC_PAGESIZE /* some SVR4 systems omit an underscore */
1427 +# ifndef _SC_PAGE_SIZE
1428 +# define _SC_PAGE_SIZE _SC_PAGESIZE
1429 +# endif
1430 +# endif
1431 +# ifdef _SC_PAGE_SIZE
1432 +# define malloc_getpagesize sysconf(_SC_PAGE_SIZE)
1433 +# else
1434 +# if defined(BSD) || defined(DGUX) || defined(HAVE_GETPAGESIZE)
1435 + extern size_t getpagesize();
1436 +# define malloc_getpagesize getpagesize()
1437 +# else
1438 +# ifdef WIN32 /* use supplied emulation of getpagesize */
1439 +# define malloc_getpagesize getpagesize()
1440 +# else
1441 +# ifndef LACKS_SYS_PARAM_H
1442 +# include <sys/param.h>
1443 +# endif
1444 +# ifdef EXEC_PAGESIZE
1445 +# define malloc_getpagesize EXEC_PAGESIZE
1446 +# else
1447 +# ifdef NBPG
1448 +# ifndef CLSIZE
1449 +# define malloc_getpagesize NBPG
1450 +# else
1451 +# define malloc_getpagesize (NBPG * CLSIZE)
1452 +# endif
1453 +# else
1454 +# ifdef NBPC
1455 +# define malloc_getpagesize NBPC
1456 +# else
1457 +# ifdef PAGESIZE
1458 +# define malloc_getpagesize PAGESIZE
1459 +# else /* just guess */
1460 +# define malloc_getpagesize ((size_t)4096U)
1461 +# endif
1462 +# endif
1463 +# endif
1464 +# endif
1465 +# endif
1466 +# endif
1467 +# endif
1468 +#endif
1469 +#endif
1470 +
1471 +
1472 +
1473 +/* ------------------- size_t and alignment properties -------------------- */
1474 +
1475 +/* The byte and bit size of a size_t */
1476 +#define SIZE_T_SIZE (sizeof(size_t))
1477 +#define SIZE_T_BITSIZE (sizeof(size_t) << 3)
1478 +
1479 +/* Some constants coerced to size_t */
1480 +/* Annoying but necessary to avoid errors on some platforms */
1481 +#define SIZE_T_ZERO ((size_t)0)
1482 +#define SIZE_T_ONE ((size_t)1)
1483 +#define SIZE_T_TWO ((size_t)2)
1484 +#define SIZE_T_FOUR ((size_t)4)
1485 +#define TWO_SIZE_T_SIZES (SIZE_T_SIZE<<1)
1486 +#define FOUR_SIZE_T_SIZES (SIZE_T_SIZE<<2)
1487 +#define SIX_SIZE_T_SIZES (FOUR_SIZE_T_SIZES+TWO_SIZE_T_SIZES)
1488 +#define HALF_MAX_SIZE_T (MAX_SIZE_T / 2U)
1489 +
1490 +/* The bit mask value corresponding to MALLOC_ALIGNMENT */
1491 +#define CHUNK_ALIGN_MASK (MALLOC_ALIGNMENT - SIZE_T_ONE)
1492 +
1493 +/* True if address a has acceptable alignment */
1494 +#define is_aligned(A) (((size_t)((A)) & (CHUNK_ALIGN_MASK)) == 0)
1495 +
1496 +/* the number of bytes to offset an address to align it */
1497 +#define align_offset(A)\
1498 + ((((size_t)(A) & CHUNK_ALIGN_MASK) == 0)? 0 :\
1499 + ((MALLOC_ALIGNMENT - ((size_t)(A) & CHUNK_ALIGN_MASK)) & CHUNK_ALIGN_MASK))
1500 +
1501 +/* -------------------------- MMAP preliminaries ------------------------- */
1502 +
1503 +/*
1504 + If HAVE_MORECORE or HAVE_MMAP are false, we just define calls and
1505 + checks to fail so compiler optimizer can delete code rather than
1506 + using so many "#if"s.
1507 +*/
1508 +
1509 +
1510 +/* MORECORE and MMAP must return MFAIL on failure */
1511 +#define MFAIL ((void*)(MAX_SIZE_T))
1512 +#define CMFAIL ((char*)(MFAIL)) /* defined for convenience */
1513 +
1514 +#if HAVE_MMAP
1515 +
1516 +#ifndef WIN32
1517 +#define MUNMAP_DEFAULT(a, s) munmap((a), (s))
1518 +#define MMAP_PROT (PROT_READ|PROT_WRITE)
1519 +#if !defined(MAP_ANONYMOUS) && defined(MAP_ANON)
1520 +#define MAP_ANONYMOUS MAP_ANON
1521 +#endif /* MAP_ANON */
1522 +#ifdef MAP_ANONYMOUS
1523 +#define MMAP_FLAGS (MAP_PRIVATE|MAP_ANONYMOUS)
1524 +#define MMAP_DEFAULT(s) mmap(0, (s), MMAP_PROT, MMAP_FLAGS, -1, 0)
1525 +#else /* MAP_ANONYMOUS */
1526 +/*
1527 + Nearly all versions of mmap support MAP_ANONYMOUS, so the following
1528 + is unlikely to be needed, but is supplied just in case.
1529 +*/
1530 +#define MMAP_FLAGS (MAP_PRIVATE)
1531 +static int dev_zero_fd = -1; /* Cached file descriptor for /dev/zero. */
1532 +#define MMAP_DEFAULT(s) ((dev_zero_fd < 0) ? \
1533 + (dev_zero_fd = open("/dev/zero", O_RDWR), \
1534 + mmap(0, (s), MMAP_PROT, MMAP_FLAGS, dev_zero_fd, 0)) : \
1535 + mmap(0, (s), MMAP_PROT, MMAP_FLAGS, dev_zero_fd, 0))
1536 +#endif /* MAP_ANONYMOUS */
1537 +
1538 +#define DIRECT_MMAP_DEFAULT(s) MMAP_DEFAULT(s)
1539 +
1540 +#else /* WIN32 */
1541 +
1542 +/* Win32 MMAP via VirtualAlloc */
1543 +static FORCEINLINE void* win32mmap(size_t size) {
1544 + void* ptr = VirtualAlloc(0, size, MEM_RESERVE|MEM_COMMIT, PAGE_READWRITE);
1545 + return (ptr != 0)? ptr: MFAIL;
1546 +}
1547 +
1548 +/* For direct MMAP, use MEM_TOP_DOWN to minimize interference */
1549 +static FORCEINLINE void* win32direct_mmap(size_t size) {
1550 + void* ptr = VirtualAlloc(0, size, MEM_RESERVE|MEM_COMMIT|MEM_TOP_DOWN,
1551 + PAGE_READWRITE);
1552 + return (ptr != 0)? ptr: MFAIL;
1553 +}
1554 +
1555 +/* This function supports releasing coalesed segments */
1556 +static FORCEINLINE int win32munmap(void* ptr, size_t size) {
1557 + MEMORY_BASIC_INFORMATION minfo;
1558 + char* cptr = (char*)ptr;
1559 + while (size) {
1560 + if (VirtualQuery(cptr, &minfo, sizeof(minfo)) == 0)
1561 + return -1;
1562 + if (minfo.BaseAddress != cptr || minfo.AllocationBase != cptr ||
1563 + minfo.State != MEM_COMMIT || minfo.RegionSize > size)
1564 + return -1;
1565 + if (VirtualFree(cptr, 0, MEM_RELEASE) == 0)
1566 + return -1;
1567 + cptr += minfo.RegionSize;
1568 + size -= minfo.RegionSize;
1569 + }
1570 + return 0;
1571 +}
1572 +
1573 +#define MMAP_DEFAULT(s) win32mmap(s)
1574 +#define MUNMAP_DEFAULT(a, s) win32munmap((a), (s))
1575 +#define DIRECT_MMAP_DEFAULT(s) win32direct_mmap(s)
1576 +#endif /* WIN32 */
1577 +#endif /* HAVE_MMAP */
1578 +
1579 +#if HAVE_MREMAP
1580 +#ifndef WIN32
1581 +#define MREMAP_DEFAULT(addr, osz, nsz, mv) mremap((addr), (osz), (nsz), (mv))
1582 +#endif /* WIN32 */
1583 +#endif /* HAVE_MREMAP */
1584 +
1585 +
1586 +/**
1587 + * Define CALL_MORECORE
1588 + */
1589 +#if HAVE_MORECORE
1590 + #ifdef MORECORE
1591 + #define CALL_MORECORE(S) MORECORE(S)
1592 + #else /* MORECORE */
1593 + #define CALL_MORECORE(S) MORECORE_DEFAULT(S)
1594 + #endif /* MORECORE */
1595 +#else /* HAVE_MORECORE */
1596 + #define CALL_MORECORE(S) MFAIL
1597 +#endif /* HAVE_MORECORE */
1598 +
1599 +/**
1600 + * Define CALL_MMAP/CALL_MUNMAP/CALL_DIRECT_MMAP
1601 + */
1602 +#if HAVE_MMAP
1603 + #define IS_MMAPPED_BIT (SIZE_T_ONE)
1604 + #define USE_MMAP_BIT (SIZE_T_ONE)
1605 +
1606 + #ifdef MMAP
1607 + #define CALL_MMAP(s) MMAP(s)
1608 + #else /* MMAP */
1609 + #define CALL_MMAP(s) MMAP_DEFAULT(s)
1610 + #endif /* MMAP */
1611 + #ifdef MUNMAP
1612 + #define CALL_MUNMAP(a, s) MUNMAP((a), (s))
1613 + #else /* MUNMAP */
1614 + #define CALL_MUNMAP(a, s) MUNMAP_DEFAULT((a), (s))
1615 + #endif /* MUNMAP */
1616 + #ifdef DIRECT_MMAP
1617 + #define CALL_DIRECT_MMAP(s) DIRECT_MMAP(s)
1618 + #else /* DIRECT_MMAP */
1619 + #define CALL_DIRECT_MMAP(s) DIRECT_MMAP_DEFAULT(s)
1620 + #endif /* DIRECT_MMAP */
1621 +#else /* HAVE_MMAP */
1622 + #define IS_MMAPPED_BIT (SIZE_T_ZERO)
1623 + #define USE_MMAP_BIT (SIZE_T_ZERO)
1624 +
1625 + #define MMAP(s) MFAIL
1626 + #define MUNMAP(a, s) (-1)
1627 + #define DIRECT_MMAP(s) MFAIL
1628 + #define CALL_DIRECT_MMAP(s) DIRECT_MMAP(s)
1629 + #define CALL_MMAP(s) MMAP(s)
1630 + #define CALL_MUNMAP(a, s) MUNMAP((a), (s))
1631 +#endif /* HAVE_MMAP */
1632 +
1633 +/**
1634 + * Define CALL_MREMAP
1635 + */
1636 +#if HAVE_MMAP && HAVE_MREMAP
1637 + #ifdef MREMAP
1638 + #define CALL_MREMAP(addr, osz, nsz, mv) MREMAP((addr), (osz), (nsz), (mv))
1639 + #else /* MREMAP */
1640 + #define CALL_MREMAP(addr, osz, nsz, mv) MREMAP_DEFAULT((addr), (osz), (nsz), (mv))
1641 + #endif /* MREMAP */
1642 +#else /* HAVE_MMAP && HAVE_MREMAP */
1643 + #define CALL_MREMAP(addr, osz, nsz, mv) MFAIL
1644 +#endif /* HAVE_MMAP && HAVE_MREMAP */
1645 +
1646 +/* mstate bit set if continguous morecore disabled or failed */
1647 +#define USE_NONCONTIGUOUS_BIT (4U)
1648 +
1649 +/* segment bit set in create_mspace_with_base */
1650 +#define EXTERN_BIT (8U)
1651 +
1652 +
1653 +/* --------------------------- Lock preliminaries ------------------------ */
1654 +
1655 +/*
1656 + When locks are defined, there is one global lock, plus
1657 + one per-mspace lock.
1658 +
1659 + The global lock_ensures that mparams.magic and other unique
1660 + mparams values are initialized only once. It also protects
1661 + sequences of calls to MORECORE. In many cases sys_alloc requires
1662 + two calls, that should not be interleaved with calls by other
1663 + threads. This does not protect against direct calls to MORECORE
1664 + by other threads not using this lock, so there is still code to
1665 + cope the best we can on interference.
1666 +
1667 + Per-mspace locks surround calls to malloc, free, etc. To enable use
1668 + in layered extensions, per-mspace locks are reentrant.
1669 +
1670 + Because lock-protected regions generally have bounded times, it is
1671 + OK to use the supplied simple spinlocks in the custom versions for
1672 + x86.
1673 +
1674 + If USE_LOCKS is > 1, the definitions of lock routines here are
1675 + bypassed, in which case you will need to define at least
1676 + INITIAL_LOCK, ACQUIRE_LOCK, RELEASE_LOCK and possibly TRY_LOCK
1677 + (which is not used in this malloc, but commonly needed in
1678 + extensions.)
1679 +*/
1680 +
1681 +#if USE_LOCKS == 1
1682 +
1683 +#if USE_SPIN_LOCKS
1684 +#ifndef WIN32
1685 +
1686 +/* Custom pthread-style spin locks on x86 and x64 for gcc */
1687 +struct pthread_mlock_t {
1688 + volatile unsigned int l;
1689 + volatile unsigned int c;
1690 + volatile pthread_t threadid;
1691 +};
1692 +#define MLOCK_T struct pthread_mlock_t
1693 +#define CURRENT_THREAD pthread_self()
1694 +#define INITIAL_LOCK(sl) (memset(sl, 0, sizeof(MLOCK_T)), 0)
1695 +#define ACQUIRE_LOCK(sl) pthread_acquire_lock(sl)
1696 +#define RELEASE_LOCK(sl) pthread_release_lock(sl)
1697 +#define TRY_LOCK(sl) pthread_try_lock(sl)
1698 +#define SPINS_PER_YIELD 63
1699 +
1700 +static MLOCK_T malloc_global_mutex = { 0, 0, 0};
1701 +
1702 +static FORCEINLINE int pthread_acquire_lock (MLOCK_T *sl) {
1703 + int spins = 0;
1704 + volatile unsigned int* lp = &sl->l;
1705 + for (;;) {
1706 + if (*lp != 0) {
1707 + if (sl->threadid == CURRENT_THREAD) {
1708 + ++sl->c;
1709 + return 0;
1710 + }
1711 + }
1712 + else {
1713 + /* place args to cmpxchgl in locals to evade oddities in some gccs */
1714 + int cmp = 0;
1715 + int val = 1;
1716 + int ret;
1717 + __asm__ __volatile__ ("lock; cmpxchgl %1, %2"
1718 + : "=a" (ret)
1719 + : "r" (val), "m" (*(lp)), "0"(cmp)
1720 + : "memory", "cc");
1721 + if (!ret) {
1722 + assert(!sl->threadid);
1723 + sl->c = 1;
1724 + sl->threadid = CURRENT_THREAD;
1725 + return 0;
1726 + }
1727 + if ((++spins & SPINS_PER_YIELD) == 0) {
1728 +#if defined (__SVR4) && defined (__sun) /* solaris */
1729 + thr_yield();
1730 +#else
1731 +#if defined(__linux__) || defined(__FreeBSD__) || defined(__APPLE__)
1732 + sched_yield();
1733 +#else /* no-op yield on unknown systems */
1734 + ;
1735 +#endif /* __linux__ || __FreeBSD__ || __APPLE__ */
1736 +#endif /* solaris */
1737 + }
1738 + }
1739 + }
1740 +}
1741 +
1742 +static FORCEINLINE void pthread_release_lock (MLOCK_T *sl) {
1743 + assert(sl->l != 0);
1744 + assert(sl->threadid == CURRENT_THREAD);
1745 + if (--sl->c == 0) {
1746 + sl->threadid = 0;
1747 + volatile unsigned int* lp = &sl->l;
1748 + int prev = 0;
1749 + int ret;
1750 + __asm__ __volatile__ ("lock; xchgl %0, %1"
1751 + : "=r" (ret)
1752 + : "m" (*(lp)), "0"(prev)
1753 + : "memory");
1754 + }
1755 +}
1756 +
1757 +static FORCEINLINE int pthread_try_lock (MLOCK_T *sl) {
1758 + volatile unsigned int* lp = &sl->l;
1759 + if (*lp != 0) {
1760 + if (sl->threadid == CURRENT_THREAD) {
1761 + ++sl->c;
1762 + return 1;
1763 + }
1764 + }
1765 + else {
1766 + int cmp = 0;
1767 + int val = 1;
1768 + int ret;
1769 + __asm__ __volatile__ ("lock; cmpxchgl %1, %2"
1770 + : "=a" (ret)
1771 + : "r" (val), "m" (*(lp)), "0"(cmp)
1772 + : "memory", "cc");
1773 + if (!ret) {
1774 + assert(!sl->threadid);
1775 + sl->c = 1;
1776 + sl->threadid = CURRENT_THREAD;
1777 + return 1;
1778 + }
1779 + }
1780 + return 0;
1781 +}
1782 +
1783 +
1784 +#else /* WIN32 */
1785 +/* Custom win32-style spin locks on x86 and x64 for MSC */
1786 +struct win32_mlock_t
1787 +{
1788 + volatile long l;
1789 + volatile unsigned int c;
1790 + volatile long threadid;
1791 +};
1792 +
1793 +#define MLOCK_T struct win32_mlock_t
1794 +#define CURRENT_THREAD win32_getcurrentthreadid()
1795 +#define INITIAL_LOCK(sl) (memset(sl, 0, sizeof(MLOCK_T)), 0)
1796 +#define ACQUIRE_LOCK(sl) win32_acquire_lock(sl)
1797 +#define RELEASE_LOCK(sl) win32_release_lock(sl)
1798 +#define TRY_LOCK(sl) win32_try_lock(sl)
1799 +#define SPINS_PER_YIELD 63
1800 +
1801 +static MLOCK_T malloc_global_mutex = { 0, 0, 0};
1802 +
1803 +static FORCEINLINE long win32_getcurrentthreadid() {
1804 +#ifdef _MSC_VER
1805 +#if defined(_M_IX86)
1806 + long *threadstruct=(long *)__readfsdword(0x18);
1807 + long threadid=threadstruct[0x24/sizeof(long)];
1808 + return threadid;
1809 +#elif defined(_M_X64)
1810 + /* todo */
1811 + return GetCurrentThreadId();
1812 +#else
1813 + return GetCurrentThreadId();
1814 +#endif
1815 +#else
1816 + return GetCurrentThreadId();
1817 +#endif
1818 +}
1819 +
1820 +static FORCEINLINE int win32_acquire_lock (MLOCK_T *sl) {
1821 + int spins = 0;
1822 + for (;;) {
1823 + if (sl->l != 0) {
1824 + if (sl->threadid == CURRENT_THREAD) {
1825 + ++sl->c;
1826 + return 0;
1827 + }
1828 + }
1829 + else {
1830 + if (!interlockedexchange(&sl->l, 1)) {
1831 + assert(!sl->threadid);
1832 + sl->c=CURRENT_THREAD;
1833 + sl->threadid = CURRENT_THREAD;
1834 + sl->c = 1;
1835 + return 0;
1836 + }
1837 + }
1838 + if ((++spins & SPINS_PER_YIELD) == 0)
1839 + SleepEx(0, FALSE);
1840 + }
1841 +}
1842 +
1843 +static FORCEINLINE void win32_release_lock (MLOCK_T *sl) {
1844 + assert(sl->threadid == CURRENT_THREAD);
1845 + assert(sl->l != 0);
1846 + if (--sl->c == 0) {
1847 + sl->threadid = 0;
1848 + interlockedexchange (&sl->l, 0);
1849 + }
1850 +}
1851 +
1852 +static FORCEINLINE int win32_try_lock (MLOCK_T *sl) {
1853 + if(sl->l != 0) {
1854 + if (sl->threadid == CURRENT_THREAD) {
1855 + ++sl->c;
1856 + return 1;
1857 + }
1858 + }
1859 + else {
1860 + if (!interlockedexchange(&sl->l, 1)){
1861 + assert(!sl->threadid);
1862 + sl->threadid = CURRENT_THREAD;
1863 + sl->c = 1;
1864 + return 1;
1865 + }
1866 + }
1867 + return 0;
1868 +}
1869 +
1870 +#endif /* WIN32 */
1871 +#else /* USE_SPIN_LOCKS */
1872 +
1873 +#ifndef WIN32
1874 +/* pthreads-based locks */
1875 +
1876 +#define MLOCK_T pthread_mutex_t
1877 +#define CURRENT_THREAD pthread_self()
1878 +#define INITIAL_LOCK(sl) pthread_init_lock(sl)
1879 +#define ACQUIRE_LOCK(sl) pthread_mutex_lock(sl)
1880 +#define RELEASE_LOCK(sl) pthread_mutex_unlock(sl)
1881 +#define TRY_LOCK(sl) (!pthread_mutex_trylock(sl))
1882 +
1883 +static MLOCK_T malloc_global_mutex = PTHREAD_MUTEX_INITIALIZER;
1884 +
1885 +/* Cope with old-style linux recursive lock initialization by adding */
1886 +/* skipped internal declaration from pthread.h */
1887 +#ifdef linux
1888 +#ifndef PTHREAD_MUTEX_RECURSIVE
1889 +extern int pthread_mutexattr_setkind_np __P ((pthread_mutexattr_t *__attr,
1890 + int __kind));
1891 +#define PTHREAD_MUTEX_RECURSIVE PTHREAD_MUTEX_RECURSIVE_NP
1892 +#define pthread_mutexattr_settype(x,y) pthread_mutexattr_setkind_np(x,y)
1893 +#endif
1894 +#endif
1895 +
1896 +static int pthread_init_lock (MLOCK_T *sl) {
1897 + pthread_mutexattr_t attr;
1898 + if (pthread_mutexattr_init(&attr)) return 1;
1899 + if (pthread_mutexattr_settype(&attr, PTHREAD_MUTEX_RECURSIVE)) return 1;
1900 + if (pthread_mutex_init(sl, &attr)) return 1;
1901 + if (pthread_mutexattr_destroy(&attr)) return 1;
1902 + return 0;
1903 +}
1904 +
1905 +#else /* WIN32 */
1906 +/* Win32 critical sections */
1907 +#define MLOCK_T CRITICAL_SECTION
1908 +#define CURRENT_THREAD GetCurrentThreadId()
1909 +#define INITIAL_LOCK(s) (!InitializeCriticalSectionAndSpinCount((s), 0x80000000|4000))
1910 +#define ACQUIRE_LOCK(s) (EnterCriticalSection(s), 0)
1911 +#define RELEASE_LOCK(s) LeaveCriticalSection(s)
1912 +#define TRY_LOCK(s) TryEnterCriticalSection(s)
1913 +#define NEED_GLOBAL_LOCK_INIT
1914 +
1915 +static MLOCK_T malloc_global_mutex;
1916 +static volatile long malloc_global_mutex_status;
1917 +
1918 +/* Use spin loop to initialize global lock */
1919 +static void init_malloc_global_mutex() {
1920 + for (;;) {
1921 + long stat = malloc_global_mutex_status;
1922 + if (stat > 0)
1923 + return;
1924 + /* transition to < 0 while initializing, then to > 0) */
1925 + if (stat == 0 &&
1926 + interlockedcompareexchange(&malloc_global_mutex_status, -1, 0) == 0) {
1927 + InitializeCriticalSection(&malloc_global_mutex);
1928 + interlockedexchange(&malloc_global_mutex_status,1);
1929 + return;
1930 + }
1931 + SleepEx(0, FALSE);
1932 + }
1933 +}
1934 +
1935 +#endif /* WIN32 */
1936 +#endif /* USE_SPIN_LOCKS */
1937 +#endif /* USE_LOCKS == 1 */
1938 +
1939 +/* ----------------------- User-defined locks ------------------------ */
1940 +
1941 +#if USE_LOCKS > 1
1942 +/* Define your own lock implementation here */
1943 +/* #define INITIAL_LOCK(sl) ... */
1944 +/* #define ACQUIRE_LOCK(sl) ... */
1945 +/* #define RELEASE_LOCK(sl) ... */
1946 +/* #define TRY_LOCK(sl) ... */
1947 +/* static MLOCK_T malloc_global_mutex = ... */
1948 +#endif /* USE_LOCKS > 1 */
1949 +
1950 +/* ----------------------- Lock-based state ------------------------ */
1951 +
1952 +#if USE_LOCKS
1953 +#define USE_LOCK_BIT (2U)
1954 +#else /* USE_LOCKS */
1955 +#define USE_LOCK_BIT (0U)
1956 +#define INITIAL_LOCK(l)
1957 +#endif /* USE_LOCKS */
1958 +
1959 +#if USE_LOCKS
1960 +#define ACQUIRE_MALLOC_GLOBAL_LOCK() ACQUIRE_LOCK(&malloc_global_mutex);
1961 +#define RELEASE_MALLOC_GLOBAL_LOCK() RELEASE_LOCK(&malloc_global_mutex);
1962 +#else /* USE_LOCKS */
1963 +#define ACQUIRE_MALLOC_GLOBAL_LOCK()
1964 +#define RELEASE_MALLOC_GLOBAL_LOCK()
1965 +#endif /* USE_LOCKS */
1966 +
1967 +
1968 +/* ----------------------- Chunk representations ------------------------ */
1969 +
1970 +/*
1971 + (The following includes lightly edited explanations by Colin Plumb.)
1972 +
1973 + The malloc_chunk declaration below is misleading (but accurate and
1974 + necessary). It declares a "view" into memory allowing access to
1975 + necessary fields at known offsets from a given base.
1976 +
1977 + Chunks of memory are maintained using a `boundary tag' method as
1978 + originally described by Knuth. (See the paper by Paul Wilson
1979 + ftp://ftp.cs.utexas.edu/pub/garbage/allocsrv.ps for a survey of such
1980 + techniques.) Sizes of free chunks are stored both in the front of
1981 + each chunk and at the end. This makes consolidating fragmented
1982 + chunks into bigger chunks fast. The head fields also hold bits
1983 + representing whether chunks are free or in use.
1984 +
1985 + Here are some pictures to make it clearer. They are "exploded" to
1986 + show that the state of a chunk can be thought of as extending from
1987 + the high 31 bits of the head field of its header through the
1988 + prev_foot and PINUSE_BIT bit of the following chunk header.
1989 +
1990 + A chunk that's in use looks like:
1991 +
1992 + chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1993 + | Size of previous chunk (if P = 0) |
1994 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1995 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |P|
1996 + | Size of this chunk 1| +-+
1997 + mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1998 + | |
1999 + +- -+
2000 + | |
2001 + +- -+
2002 + | :
2003 + +- size - sizeof(size_t) available payload bytes -+
2004 + : |
2005 + chunk-> +- -+
2006 + | |
2007 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2008 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |1|
2009 + | Size of next chunk (may or may not be in use) | +-+
2010 + mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2011 +
2012 + And if it's free, it looks like this:
2013 +
2014 + chunk-> +- -+
2015 + | User payload (must be in use, or we would have merged!) |
2016 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2017 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |P|
2018 + | Size of this chunk 0| +-+
2019 + mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2020 + | Next pointer |
2021 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2022 + | Prev pointer |
2023 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2024 + | :
2025 + +- size - sizeof(struct chunk) unused bytes -+
2026 + : |
2027 + chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2028 + | Size of this chunk |
2029 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2030 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |0|
2031 + | Size of next chunk (must be in use, or we would have merged)| +-+
2032 + mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2033 + | :
2034 + +- User payload -+
2035 + : |
2036 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2037 + |0|
2038 + +-+
2039 + Note that since we always merge adjacent free chunks, the chunks
2040 + adjacent to a free chunk must be in use.
2041 +
2042 + Given a pointer to a chunk (which can be derived trivially from the
2043 + payload pointer) we can, in O(1) time, find out whether the adjacent
2044 + chunks are free, and if so, unlink them from the lists that they
2045 + are on and merge them with the current chunk.
2046 +
2047 + Chunks always begin on even word boundaries, so the mem portion
2048 + (which is returned to the user) is also on an even word boundary, and
2049 + thus at least double-word aligned.
2050 +
2051 + The P (PINUSE_BIT) bit, stored in the unused low-order bit of the
2052 + chunk size (which is always a multiple of two words), is an in-use
2053 + bit for the *previous* chunk. If that bit is *clear*, then the
2054 + word before the current chunk size contains the previous chunk
2055 + size, and can be used to find the front of the previous chunk.
2056 + The very first chunk allocated always has this bit set, preventing
2057 + access to non-existent (or non-owned) memory. If pinuse is set for
2058 + any given chunk, then you CANNOT determine the size of the
2059 + previous chunk, and might even get a memory addressing fault when
2060 + trying to do so.
2061 +
2062 + The C (CINUSE_BIT) bit, stored in the unused second-lowest bit of
2063 + the chunk size redundantly records whether the current chunk is
2064 + inuse. This redundancy enables usage checks within free and realloc,
2065 + and reduces indirection when freeing and consolidating chunks.
2066 +
2067 + Each freshly allocated chunk must have both cinuse and pinuse set.
2068 + That is, each allocated chunk borders either a previously allocated
2069 + and still in-use chunk, or the base of its memory arena. This is
2070 + ensured by making all allocations from the the `lowest' part of any
2071 + found chunk. Further, no free chunk physically borders another one,
2072 + so each free chunk is known to be preceded and followed by either
2073 + inuse chunks or the ends of memory.
2074 +
2075 + Note that the `foot' of the current chunk is actually represented
2076 + as the prev_foot of the NEXT chunk. This makes it easier to
2077 + deal with alignments etc but can be very confusing when trying
2078 + to extend or adapt this code.
2079 +
2080 + The exceptions to all this are
2081 +
2082 + 1. The special chunk `top' is the top-most available chunk (i.e.,
2083 + the one bordering the end of available memory). It is treated
2084 + specially. Top is never included in any bin, is used only if
2085 + no other chunk is available, and is released back to the
2086 + system if it is very large (see M_TRIM_THRESHOLD). In effect,
2087 + the top chunk is treated as larger (and thus less well
2088 + fitting) than any other available chunk. The top chunk
2089 + doesn't update its trailing size field since there is no next
2090 + contiguous chunk that would have to index off it. However,
2091 + space is still allocated for it (TOP_FOOT_SIZE) to enable
2092 + separation or merging when space is extended.
2093 +
2094 + 3. Chunks allocated via mmap, which have the lowest-order bit
2095 + (IS_MMAPPED_BIT) set in their prev_foot fields, and do not set
2096 + PINUSE_BIT in their head fields. Because they are allocated
2097 + one-by-one, each must carry its own prev_foot field, which is
2098 + also used to hold the offset this chunk has within its mmapped
2099 + region, which is needed to preserve alignment. Each mmapped
2100 + chunk is trailed by the first two fields of a fake next-chunk
2101 + for sake of usage checks.
2102 +
2103 +*/
2104 +
2105 +struct malloc_chunk {
2106 + size_t prev_foot; /* Size of previous chunk (if free). */
2107 + size_t head; /* Size and inuse bits. */
2108 + struct malloc_chunk* fd; /* double links -- used only if free. */
2109 + struct malloc_chunk* bk;
2110 +};
2111 +
2112 +typedef struct malloc_chunk mchunk;
2113 +typedef struct malloc_chunk* mchunkptr;
2114 +typedef struct malloc_chunk* sbinptr; /* The type of bins of chunks */
2115 +typedef unsigned int bindex_t; /* Described below */
2116 +typedef unsigned int binmap_t; /* Described below */
2117 +typedef unsigned int flag_t; /* The type of various bit flag sets */
2118 +
2119 +/* ------------------- Chunks sizes and alignments ----------------------- */
2120 +
2121 +#define MCHUNK_SIZE (sizeof(mchunk))
2122 +
2123 +#if FOOTERS
2124 +#define CHUNK_OVERHEAD (TWO_SIZE_T_SIZES)
2125 +#else /* FOOTERS */
2126 +#define CHUNK_OVERHEAD (SIZE_T_SIZE)
2127 +#endif /* FOOTERS */
2128 +
2129 +/* MMapped chunks need a second word of overhead ... */
2130 +#define MMAP_CHUNK_OVERHEAD (TWO_SIZE_T_SIZES)
2131 +/* ... and additional padding for fake next-chunk at foot */
2132 +#define MMAP_FOOT_PAD (FOUR_SIZE_T_SIZES)
2133 +
2134 +/* The smallest size we can malloc is an aligned minimal chunk */
2135 +#define MIN_CHUNK_SIZE\
2136 + ((MCHUNK_SIZE + CHUNK_ALIGN_MASK) & ~CHUNK_ALIGN_MASK)
2137 +
2138 +/* conversion from malloc headers to user pointers, and back */
2139 +#define chunk2mem(p) ((void*)((char*)(p) + TWO_SIZE_T_SIZES))
2140 +#define mem2chunk(mem) ((mchunkptr)((char*)(mem) - TWO_SIZE_T_SIZES))
2141 +/* chunk associated with aligned address A */
2142 +#define align_as_chunk(A) (mchunkptr)((A) + align_offset(chunk2mem(A)))
2143 +
2144 +/* Bounds on request (not chunk) sizes. */
2145 +#define MAX_REQUEST ((-MIN_CHUNK_SIZE) << 2)
2146 +#define MIN_REQUEST (MIN_CHUNK_SIZE - CHUNK_OVERHEAD - SIZE_T_ONE)
2147 +
2148 +/* pad request bytes into a usable size */
2149 +#define pad_request(req) \
2150 + (((req) + CHUNK_OVERHEAD + CHUNK_ALIGN_MASK) & ~CHUNK_ALIGN_MASK)
2151 +
2152 +/* pad request, checking for minimum (but not maximum) */
2153 +#define request2size(req) \
2154 + (((req) < MIN_REQUEST)? MIN_CHUNK_SIZE : pad_request(req))
2155 +
2156 +
2157 +/* ------------------ Operations on head and foot fields ----------------- */
2158 +
2159 +/*
2160 + The head field of a chunk is or'ed with PINUSE_BIT when previous
2161 + adjacent chunk in use, and or'ed with CINUSE_BIT if this chunk is in
2162 + use. If the chunk was obtained with mmap, the prev_foot field has
2163 + IS_MMAPPED_BIT set, otherwise holding the offset of the base of the
2164 + mmapped region to the base of the chunk.
2165 +
2166 + FLAG4_BIT is not used by this malloc, but might be useful in extensions.
2167 +*/
2168 +
2169 +#define PINUSE_BIT (SIZE_T_ONE)
2170 +#define CINUSE_BIT (SIZE_T_TWO)
2171 +#define FLAG4_BIT (SIZE_T_FOUR)
2172 +#define INUSE_BITS (PINUSE_BIT|CINUSE_BIT)
2173 +#define FLAG_BITS (PINUSE_BIT|CINUSE_BIT|FLAG4_BIT)
2174 +
2175 +/* Head value for fenceposts */
2176 +#define FENCEPOST_HEAD (INUSE_BITS|SIZE_T_SIZE)
2177 +
2178 +/* extraction of fields from head words */
2179 +#define cinuse(p) ((p)->head & CINUSE_BIT)
2180 +#define pinuse(p) ((p)->head & PINUSE_BIT)
2181 +#define chunksize(p) ((p)->head & ~(FLAG_BITS))
2182 +
2183 +#define clear_pinuse(p) ((p)->head &= ~PINUSE_BIT)
2184 +#define clear_cinuse(p) ((p)->head &= ~CINUSE_BIT)
2185 +
2186 +/* Treat space at ptr +/- offset as a chunk */
2187 +#define chunk_plus_offset(p, s) ((mchunkptr)(((char*)(p)) + (s)))
2188 +#define chunk_minus_offset(p, s) ((mchunkptr)(((char*)(p)) - (s)))
2189 +
2190 +/* Ptr to next or previous physical malloc_chunk. */
2191 +#define next_chunk(p) ((mchunkptr)( ((char*)(p)) + ((p)->head & ~FLAG_BITS)))
2192 +#define prev_chunk(p) ((mchunkptr)( ((char*)(p)) - ((p)->prev_foot) ))
2193 +
2194 +/* extract next chunk's pinuse bit */
2195 +#define next_pinuse(p) ((next_chunk(p)->head) & PINUSE_BIT)
2196 +
2197 +/* Get/set size at footer */
2198 +#define get_foot(p, s) (((mchunkptr)((char*)(p) + (s)))->prev_foot)
2199 +#define set_foot(p, s) (((mchunkptr)((char*)(p) + (s)))->prev_foot = (s))
2200 +
2201 +/* Set size, pinuse bit, and foot */
2202 +#define set_size_and_pinuse_of_free_chunk(p, s)\
2203 + ((p)->head = (s|PINUSE_BIT), set_foot(p, s))
2204 +
2205 +/* Set size, pinuse bit, foot, and clear next pinuse */
2206 +#define set_free_with_pinuse(p, s, n)\
2207 + (clear_pinuse(n), set_size_and_pinuse_of_free_chunk(p, s))
2208 +
2209 +#define is_mmapped(p)\
2210 + (!((p)->head & PINUSE_BIT) && ((p)->prev_foot & IS_MMAPPED_BIT))
2211 +
2212 +/* Get the internal overhead associated with chunk p */
2213 +#define overhead_for(p)\
2214 + (is_mmapped(p)? MMAP_CHUNK_OVERHEAD : CHUNK_OVERHEAD)
2215 +
2216 +/* Return true if malloced space is not necessarily cleared */
2217 +#if MMAP_CLEARS
2218 +#define calloc_must_clear(p) (!is_mmapped(p))
2219 +#else /* MMAP_CLEARS */
2220 +#define calloc_must_clear(p) (1)
2221 +#endif /* MMAP_CLEARS */
2222 +
2223 +/* ---------------------- Overlaid data structures ----------------------- */
2224 +
2225 +/*
2226 + When chunks are not in use, they are treated as nodes of either
2227 + lists or trees.
2228 +
2229 + "Small" chunks are stored in circular doubly-linked lists, and look
2230 + like this:
2231 +
2232 + chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2233 + | Size of previous chunk |
2234 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2235 + `head:' | Size of chunk, in bytes |P|
2236 + mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2237 + | Forward pointer to next chunk in list |
2238 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2239 + | Back pointer to previous chunk in list |
2240 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2241 + | Unused space (may be 0 bytes long) .
2242 + . .
2243 + . |
2244 +nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2245 + `foot:' | Size of chunk, in bytes |
2246 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2247 +
2248 + Larger chunks are kept in a form of bitwise digital trees (aka
2249 + tries) keyed on chunksizes. Because malloc_tree_chunks are only for
2250 + free chunks greater than 256 bytes, their size doesn't impose any
2251 + constraints on user chunk sizes. Each node looks like:
2252 +
2253 + chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2254 + | Size of previous chunk |
2255 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2256 + `head:' | Size of chunk, in bytes |P|
2257 + mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2258 + | Forward pointer to next chunk of same size |
2259 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2260 + | Back pointer to previous chunk of same size |
2261 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2262 + | Pointer to left child (child[0]) |
2263 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2264 + | Pointer to right child (child[1]) |
2265 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2266 + | Pointer to parent |
2267 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2268 + | bin index of this chunk |
2269 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2270 + | Unused space .
2271 + . |
2272 +nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2273 + `foot:' | Size of chunk, in bytes |
2274 + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2275 +
2276 + Each tree holding treenodes is a tree of unique chunk sizes. Chunks
2277 + of the same size are arranged in a circularly-linked list, with only
2278 + the oldest chunk (the next to be used, in our FIFO ordering)
2279 + actually in the tree. (Tree members are distinguished by a non-null
2280 + parent pointer.) If a chunk with the same size an an existing node
2281 + is inserted, it is linked off the existing node using pointers that
2282 + work in the same way as fd/bk pointers of small chunks.
2283 +
2284 + Each tree contains a power of 2 sized range of chunk sizes (the
2285 + smallest is 0x100 <= x < 0x180), which is is divided in half at each
2286 + tree level, with the chunks in the smaller half of the range (0x100
2287 + <= x < 0x140 for the top nose) in the left subtree and the larger
2288 + half (0x140 <= x < 0x180) in the right subtree. This is, of course,
2289 + done by inspecting individual bits.
2290 +
2291 + Using these rules, each node's left subtree contains all smaller
2292 + sizes than its right subtree. However, the node at the root of each
2293 + subtree has no particular ordering relationship to either. (The
2294 + dividing line between the subtree sizes is based on trie relation.)
2295 + If we remove the last chunk of a given size from the interior of the
2296 + tree, we need to replace it with a leaf node. The tree ordering
2297 + rules permit a node to be replaced by any leaf below it.
2298 +
2299 + The smallest chunk in a tree (a common operation in a best-fit
2300 + allocator) can be found by walking a path to the leftmost leaf in
2301 + the tree. Unlike a usual binary tree, where we follow left child
2302 + pointers until we reach a null, here we follow the right child
2303 + pointer any time the left one is null, until we reach a leaf with
2304 + both child pointers null. The smallest chunk in the tree will be
2305 + somewhere along that path.
2306 +
2307 + The worst case number of steps to add, find, or remove a node is
2308 + bounded by the number of bits differentiating chunks within
2309 + bins. Under current bin calculations, this ranges from 6 up to 21
2310 + (for 32 bit sizes) or up to 53 (for 64 bit sizes). The typical case
2311 + is of course much better.
2312 +*/
2313 +
2314 +struct malloc_tree_chunk {
2315 + /* The first four fields must be compatible with malloc_chunk */
2316 + size_t prev_foot;
2317 + size_t head;
2318 + struct malloc_tree_chunk* fd;
2319 + struct malloc_tree_chunk* bk;
2320 +
2321 + struct malloc_tree_chunk* child[2];
2322 + struct malloc_tree_chunk* parent;
2323 + bindex_t index;
2324 +};
2325 +
2326 +typedef struct malloc_tree_chunk tchunk;
2327 +typedef struct malloc_tree_chunk* tchunkptr;
2328 +typedef struct malloc_tree_chunk* tbinptr; /* The type of bins of trees */
2329 +
2330 +/* A little helper macro for trees */
2331 +#define leftmost_child(t) ((t)->child[0] != 0? (t)->child[0] : (t)->child[1])
2332 +
2333 +/* ----------------------------- Segments -------------------------------- */
2334 +
2335 +/*
2336 + Each malloc space may include non-contiguous segments, held in a
2337 + list headed by an embedded malloc_segment record representing the
2338 + top-most space. Segments also include flags holding properties of
2339 + the space. Large chunks that are directly allocated by mmap are not
2340 + included in this list. They are instead independently created and
2341 + destroyed without otherwise keeping track of them.
2342 +
2343 + Segment management mainly comes into play for spaces allocated by
2344 + MMAP. Any call to MMAP might or might not return memory that is
2345 + adjacent to an existing segment. MORECORE normally contiguously
2346 + extends the current space, so this space is almost always adjacent,
2347 + which is simpler and faster to deal with. (This is why MORECORE is
2348 + used preferentially to MMAP when both are available -- see
2349 + sys_alloc.) When allocating using MMAP, we don't use any of the
2350 + hinting mechanisms (inconsistently) supported in various
2351 + implementations of unix mmap, or distinguish reserving from
2352 + committing memory. Instead, we just ask for space, and exploit
2353 + contiguity when we get it. It is probably possible to do
2354 + better than this on some systems, but no general scheme seems
2355 + to be significantly better.
2356 +
2357 + Management entails a simpler variant of the consolidation scheme
2358 + used for chunks to reduce fragmentation -- new adjacent memory is
2359 + normally prepended or appended to an existing segment. However,
2360 + there are limitations compared to chunk consolidation that mostly
2361 + reflect the fact that segment processing is relatively infrequent
2362 + (occurring only when getting memory from system) and that we
2363 + don't expect to have huge numbers of segments:
2364 +
2365 + * Segments are not indexed, so traversal requires linear scans. (It
2366 + would be possible to index these, but is not worth the extra
2367 + overhead and complexity for most programs on most platforms.)
2368 + * New segments are only appended to old ones when holding top-most
2369 + memory; if they cannot be prepended to others, they are held in
2370 + different segments.
2371 +
2372 + Except for the top-most segment of an mstate, each segment record
2373 + is kept at the tail of its segment. Segments are added by pushing
2374 + segment records onto the list headed by &mstate.seg for the
2375 + containing mstate.
2376 +
2377 + Segment flags control allocation/merge/deallocation policies:
2378 + * If EXTERN_BIT set, then we did not allocate this segment,
2379 + and so should not try to deallocate or merge with others.
2380 + (This currently holds only for the initial segment passed
2381 + into create_mspace_with_base.)
2382 + * If IS_MMAPPED_BIT set, the segment may be merged with
2383 + other surrounding mmapped segments and trimmed/de-allocated
2384 + using munmap.
2385 + * If neither bit is set, then the segment was obtained using
2386 + MORECORE so can be merged with surrounding MORECORE'd segments
2387 + and deallocated/trimmed using MORECORE with negative arguments.
2388 +*/
2389 +
2390 +struct malloc_segment {
2391 + char* base; /* base address */
2392 + size_t size; /* allocated size */
2393 + struct malloc_segment* next; /* ptr to next segment */
2394 + flag_t sflags; /* mmap and extern flag */
2395 +};
2396 +
2397 +#define is_mmapped_segment(S) ((S)->sflags & IS_MMAPPED_BIT)
2398 +#define is_extern_segment(S) ((S)->sflags & EXTERN_BIT)
2399 +
2400 +typedef struct malloc_segment msegment;
2401 +typedef struct malloc_segment* msegmentptr;
2402 +
2403 +/* ---------------------------- malloc_state ----------------------------- */
2404 +
2405 +/*
2406 + A malloc_state holds all of the bookkeeping for a space.
2407 + The main fields are:
2408 +
2409 + Top
2410 + The topmost chunk of the currently active segment. Its size is
2411 + cached in topsize. The actual size of topmost space is
2412 + topsize+TOP_FOOT_SIZE, which includes space reserved for adding
2413 + fenceposts and segment records if necessary when getting more
2414 + space from the system. The size at which to autotrim top is
2415 + cached from mparams in trim_check, except that it is disabled if
2416 + an autotrim fails.
2417 +
2418 + Designated victim (dv)
2419 + This is the preferred chunk for servicing small requests that
2420 + don't have exact fits. It is normally the chunk split off most
2421 + recently to service another small request. Its size is cached in
2422 + dvsize. The link fields of this chunk are not maintained since it
2423 + is not kept in a bin.
2424 +
2425 + SmallBins
2426 + An array of bin headers for free chunks. These bins hold chunks
2427 + with sizes less than MIN_LARGE_SIZE bytes. Each bin contains
2428 + chunks of all the same size, spaced 8 bytes apart. To simplify
2429 + use in double-linked lists, each bin header acts as a malloc_chunk
2430 + pointing to the real first node, if it exists (else pointing to
2431 + itself). This avoids special-casing for headers. But to avoid
2432 + waste, we allocate only the fd/bk pointers of bins, and then use
2433 + repositioning tricks to treat these as the fields of a chunk.
2434 +
2435 + TreeBins
2436 + Treebins are pointers to the roots of trees holding a range of
2437 + sizes. There are 2 equally spaced treebins for each power of two
2438 + from TREE_SHIFT to TREE_SHIFT+16. The last bin holds anything
2439 + larger.
2440 +
2441 + Bin maps
2442 + There is one bit map for small bins ("smallmap") and one for
2443 + treebins ("treemap). Each bin sets its bit when non-empty, and
2444 + clears the bit when empty. Bit operations are then used to avoid
2445 + bin-by-bin searching -- nearly all "search" is done without ever
2446 + looking at bins that won't be selected. The bit maps
2447 + conservatively use 32 bits per map word, even if on 64bit system.
2448 + For a good description of some of the bit-based techniques used
2449 + here, see Henry S. Warren Jr's book "Hacker's Delight" (and
2450 + supplement at http://hackersdelight.org/). Many of these are
2451 + intended to reduce the branchiness of paths through malloc etc, as
2452 + well as to reduce the number of memory locations read or written.
2453 +
2454 + Segments
2455 + A list of segments headed by an embedded malloc_segment record
2456 + representing the initial space.
2457 +
2458 + Address check support
2459 + The least_addr field is the least address ever obtained from
2460 + MORECORE or MMAP. Attempted frees and reallocs of any address less
2461 + than this are trapped (unless INSECURE is defined).
2462 +
2463 + Magic tag
2464 + A cross-check field that should always hold same value as mparams.magic.
2465 +
2466 + Flags
2467 + Bits recording whether to use MMAP, locks, or contiguous MORECORE
2468 +
2469 + Statistics
2470 + Each space keeps track of current and maximum system memory
2471 + obtained via MORECORE or MMAP.
2472 +
2473 + Trim support
2474 + Fields holding the amount of unused topmost memory that should trigger
2475 + timming, and a counter to force periodic scanning to release unused
2476 + non-topmost segments.
2477 +
2478 + Locking
2479 + If USE_LOCKS is defined, the "mutex" lock is acquired and released
2480 + around every public call using this mspace.
2481 +
2482 + Extension support
2483 + A void* pointer and a size_t field that can be used to help implement
2484 + extensions to this malloc.
2485 +*/
2486 +
2487 +/* Bin types, widths and sizes */
2488 +#define NSMALLBINS (32U)
2489 +#define NTREEBINS (32U)
2490 +#define SMALLBIN_SHIFT (3U)
2491 +#define SMALLBIN_WIDTH (SIZE_T_ONE << SMALLBIN_SHIFT)
2492 +#define TREEBIN_SHIFT (8U)
2493 +#define MIN_LARGE_SIZE (SIZE_T_ONE << TREEBIN_SHIFT)
2494 +#define MAX_SMALL_SIZE (MIN_LARGE_SIZE - SIZE_T_ONE)
2495 +#define MAX_SMALL_REQUEST (MAX_SMALL_SIZE - CHUNK_ALIGN_MASK - CHUNK_OVERHEAD)
2496 +
2497 +struct malloc_state {
2498 + binmap_t smallmap;
2499 + binmap_t treemap;
2500 + size_t dvsize;
2501 + size_t topsize;
2502 + char* least_addr;
2503 + mchunkptr dv;
2504 + mchunkptr top;
2505 + size_t trim_check;
2506 + size_t release_checks;
2507 + size_t magic;
2508 + mchunkptr smallbins[(NSMALLBINS+1)*2];
2509 + tbinptr treebins[NTREEBINS];
2510 + size_t footprint;
2511 + size_t max_footprint;
2512 + flag_t mflags;
2513 +#if USE_LOCKS
2514 + MLOCK_T mutex; /* locate lock among fields that rarely change */
2515 +#endif /* USE_LOCKS */
2516 + msegment seg;
2517 + void* extp; /* Unused but available for extensions */
2518 + size_t exts;
2519 +};
2520 +
2521 +typedef struct malloc_state* mstate;
2522 +
2523 +/* ------------- Global malloc_state and malloc_params ------------------- */
2524 +
2525 +/*
2526 + malloc_params holds global properties, including those that can be
2527 + dynamically set using mallopt. There is a single instance, mparams,
2528 + initialized in init_mparams. Note that the non-zeroness of "magic"
2529 + also serves as an initialization flag.
2530 +*/
2531 +
2532 +struct malloc_params {
2533 + volatile size_t magic;
2534 + size_t page_size;
2535 + size_t granularity;
2536 + size_t mmap_threshold;
2537 + size_t trim_threshold;
2538 + flag_t default_mflags;
2539 +};
2540 +
2541 +static struct malloc_params mparams;
2542 +
2543 +/* Ensure mparams initialized */
2544 +#define ensure_initialization() (mparams.magic != 0 || init_mparams())
2545 +
2546 +#if !ONLY_MSPACES
2547 +
2548 +/* The global malloc_state used for all non-"mspace" calls */
2549 +static struct malloc_state _gm_;
2550 +#define gm (&_gm_)
2551 +#define is_global(M) ((M) == &_gm_)
2552 +
2553 +#endif /* !ONLY_MSPACES */
2554 +
2555 +#define is_initialized(M) ((M)->top != 0)
2556 +
2557 +/* -------------------------- system alloc setup ------------------------- */
2558 +
2559 +/* Operations on mflags */
2560 +
2561 +#define use_lock(M) ((M)->mflags & USE_LOCK_BIT)
2562 +#define enable_lock(M) ((M)->mflags |= USE_LOCK_BIT)
2563 +#define disable_lock(M) ((M)->mflags &= ~USE_LOCK_BIT)
2564 +
2565 +#define use_mmap(M) ((M)->mflags & USE_MMAP_BIT)
2566 +#define enable_mmap(M) ((M)->mflags |= USE_MMAP_BIT)
2567 +#define disable_mmap(M) ((M)->mflags &= ~USE_MMAP_BIT)
2568 +
2569 +#define use_noncontiguous(M) ((M)->mflags & USE_NONCONTIGUOUS_BIT)
2570 +#define disable_contiguous(M) ((M)->mflags |= USE_NONCONTIGUOUS_BIT)
2571 +
2572 +#define set_lock(M,L)\
2573 + ((M)->mflags = (L)?\
2574 + ((M)->mflags | USE_LOCK_BIT) :\
2575 + ((M)->mflags & ~USE_LOCK_BIT))
2576 +
2577 +/* page-align a size */
2578 +#define page_align(S)\
2579 + (((S) + (mparams.page_size - SIZE_T_ONE)) & ~(mparams.page_size - SIZE_T_ONE))
2580 +
2581 +/* granularity-align a size */
2582 +#define granularity_align(S)\
2583 + (((S) + (mparams.granularity - SIZE_T_ONE))\
2584 + & ~(mparams.granularity - SIZE_T_ONE))
2585 +
2586 +
2587 +/* For mmap, use granularity alignment on windows, else page-align */
2588 +#ifdef WIN32
2589 +#define mmap_align(S) granularity_align(S)
2590 +#else
2591 +#define mmap_align(S) page_align(S)
2592 +#endif
2593 +
2594 +/* For sys_alloc, enough padding to ensure can malloc request on success */
2595 +#define SYS_ALLOC_PADDING (TOP_FOOT_SIZE + MALLOC_ALIGNMENT)
2596 +
2597 +#define is_page_aligned(S)\
2598 + (((size_t)(S) & (mparams.page_size - SIZE_T_ONE)) == 0)
2599 +#define is_granularity_aligned(S)\
2600 + (((size_t)(S) & (mparams.granularity - SIZE_T_ONE)) == 0)
2601 +
2602 +/* True if segment S holds address A */
2603 +#define segment_holds(S, A)\
2604 + ((char*)(A) >= S->base && (char*)(A) < S->base + S->size)
2605 +
2606 +/* Return segment holding given address */
2607 +static msegmentptr segment_holding(mstate m, char* addr) {
2608 + msegmentptr sp = &m->seg;
2609 + for (;;) {
2610 + if (addr >= sp->base && addr < sp->base + sp->size)
2611 + return sp;
2612 + if ((sp = sp->next) == 0)
2613 + return 0;
2614 + }
2615 +}
2616 +
2617 +/* Return true if segment contains a segment link */
2618 +static int has_segment_link(mstate m, msegmentptr ss) {
2619 + msegmentptr sp = &m->seg;
2620 + for (;;) {
2621 + if ((char*)sp >= ss->base && (char*)sp < ss->base + ss->size)
2622 + return 1;
2623 + if ((sp = sp->next) == 0)
2624 + return 0;
2625 + }
2626 +}
2627 +
2628 +#ifndef MORECORE_CANNOT_TRIM
2629 +#define should_trim(M,s) ((s) > (M)->trim_check)
2630 +#else /* MORECORE_CANNOT_TRIM */
2631 +#define should_trim(M,s) (0)
2632 +#endif /* MORECORE_CANNOT_TRIM */
2633 +
2634 +/*
2635 + TOP_FOOT_SIZE is padding at the end of a segment, including space
2636 + that may be needed to place segment records and fenceposts when new
2637 + noncontiguous segments are added.
2638 +*/
2639 +#define TOP_FOOT_SIZE\
2640 + (align_offset(chunk2mem(0))+pad_request(sizeof(struct malloc_segment))+MIN_CHUNK_SIZE)
2641 +
2642 +
2643 +/* ------------------------------- Hooks -------------------------------- */
2644 +
2645 +/*
2646 + PREACTION should be defined to return 0 on success, and nonzero on
2647 + failure. If you are not using locking, you can redefine these to do
2648 + anything you like.
2649 +*/
2650 +
2651 +#if USE_LOCKS
2652 +
2653 +#define PREACTION(M) ((use_lock(M))? ACQUIRE_LOCK(&(M)->mutex) : 0)
2654 +#define POSTACTION(M) { if (use_lock(M)) RELEASE_LOCK(&(M)->mutex); }
2655 +#else /* USE_LOCKS */
2656 +
2657 +#ifndef PREACTION
2658 +#define PREACTION(M) (0)
2659 +#endif /* PREACTION */
2660 +
2661 +#ifndef POSTACTION
2662 +#define POSTACTION(M)
2663 +#endif /* POSTACTION */
2664 +
2665 +#endif /* USE_LOCKS */
2666 +
2667 +/*
2668 + CORRUPTION_ERROR_ACTION is triggered upon detected bad addresses.
2669 + USAGE_ERROR_ACTION is triggered on detected bad frees and
2670 + reallocs. The argument p is an address that might have triggered the
2671 + fault. It is ignored by the two predefined actions, but might be
2672 + useful in custom actions that try to help diagnose errors.
2673 +*/
2674 +
2675 +#if PROCEED_ON_ERROR
2676 +
2677 +/* A count of the number of corruption errors causing resets */
2678 +int malloc_corruption_error_count;
2679 +
2680 +/* default corruption action */
2681 +static void reset_on_error(mstate m);
2682 +
2683 +#define CORRUPTION_ERROR_ACTION(m) reset_on_error(m)
2684 +#define USAGE_ERROR_ACTION(m, p)
2685 +
2686 +#else /* PROCEED_ON_ERROR */
2687 +
2688 +#ifndef CORRUPTION_ERROR_ACTION
2689 +#define CORRUPTION_ERROR_ACTION(m) ABORT
2690 +#endif /* CORRUPTION_ERROR_ACTION */
2691 +
2692 +#ifndef USAGE_ERROR_ACTION
2693 +#define USAGE_ERROR_ACTION(m,p) ABORT
2694 +#endif /* USAGE_ERROR_ACTION */
2695 +
2696 +#endif /* PROCEED_ON_ERROR */
2697 +
2698 +/* -------------------------- Debugging setup ---------------------------- */
2699 +
2700 +#if ! DEBUG
2701 +
2702 +#define check_free_chunk(M,P)
2703 +#define check_inuse_chunk(M,P)
2704 +#define check_malloced_chunk(M,P,N)
2705 +#define check_mmapped_chunk(M,P)
2706 +#define check_malloc_state(M)
2707 +#define check_top_chunk(M,P)
2708 +
2709 +#else /* DEBUG */
2710 +#define check_free_chunk(M,P) do_check_free_chunk(M,P)
2711 +#define check_inuse_chunk(M,P) do_check_inuse_chunk(M,P)
2712 +#define check_top_chunk(M,P) do_check_top_chunk(M,P)
2713 +#define check_malloced_chunk(M,P,N) do_check_malloced_chunk(M,P,N)
2714 +#define check_mmapped_chunk(M,P) do_check_mmapped_chunk(M,P)
2715 +#define check_malloc_state(M) do_check_malloc_state(M)
2716 +
2717 +static void do_check_any_chunk(mstate m, mchunkptr p);
2718 +static void do_check_top_chunk(mstate m, mchunkptr p);
2719 +static void do_check_mmapped_chunk(mstate m, mchunkptr p);
2720 +static void do_check_inuse_chunk(mstate m, mchunkptr p);
2721 +static void do_check_free_chunk(mstate m, mchunkptr p);
2722 +static void do_check_malloced_chunk(mstate m, void* mem, size_t s);
2723 +static void do_check_tree(mstate m, tchunkptr t);
2724 +static void do_check_treebin(mstate m, bindex_t i);
2725 +static void do_check_smallbin(mstate m, bindex_t i);
2726 +static void do_check_malloc_state(mstate m);
2727 +static int bin_find(mstate m, mchunkptr x);
2728 +static size_t traverse_and_check(mstate m);
2729 +#endif /* DEBUG */
2730 +
2731 +/* ---------------------------- Indexing Bins ---------------------------- */
2732 +
2733 +#define is_small(s) (((s) >> SMALLBIN_SHIFT) < NSMALLBINS)
2734 +#define small_index(s) ((s) >> SMALLBIN_SHIFT)
2735 +#define small_index2size(i) ((i) << SMALLBIN_SHIFT)
2736 +#define MIN_SMALL_INDEX (small_index(MIN_CHUNK_SIZE))
2737 +
2738 +/* addressing by index. See above about smallbin repositioning */
2739 +#define smallbin_at(M, i) ((sbinptr)((char*)&((M)->smallbins[(i)<<1])))
2740 +#define treebin_at(M,i) (&((M)->treebins[i]))
2741 +
2742 +/* assign tree index for size S to variable I. Use x86 asm if possible */
2743 +#if defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__))
2744 +#define compute_tree_index(S, I)\
2745 +{\
2746 + unsigned int X = S >> TREEBIN_SHIFT;\
2747 + if (X == 0)\
2748 + I = 0;\
2749 + else if (X > 0xFFFF)\
2750 + I = NTREEBINS-1;\
2751 + else {\
2752 + unsigned int K;\
2753 + __asm__("bsrl\t%1, %0\n\t" : "=r" (K) : "rm" (X));\
2754 + I = (bindex_t)((K << 1) + ((S >> (K + (TREEBIN_SHIFT-1)) & 1)));\
2755 + }\
2756 +}
2757 +
2758 +#elif defined (__INTEL_COMPILER)
2759 +#define compute_tree_index(S, I)\
2760 +{\
2761 + size_t X = S >> TREEBIN_SHIFT;\
2762 + if (X == 0)\
2763 + I = 0;\
2764 + else if (X > 0xFFFF)\
2765 + I = NTREEBINS-1;\
2766 + else {\
2767 + unsigned int K = _bit_scan_reverse (X); \
2768 + I = (bindex_t)((K << 1) + ((S >> (K + (TREEBIN_SHIFT-1)) & 1)));\
2769 + }\
2770 +}
2771 +
2772 +#elif defined(_MSC_VER) && _MSC_VER>=1300
2773 +#define compute_tree_index(S, I)\
2774 +{\
2775 + size_t X = S >> TREEBIN_SHIFT;\
2776 + if (X == 0)\
2777 + I = 0;\
2778 + else if (X > 0xFFFF)\
2779 + I = NTREEBINS-1;\
2780 + else {\
2781 + unsigned int K;\
2782 + _BitScanReverse((DWORD *) &K, X);\
2783 + I = (bindex_t)((K << 1) + ((S >> (K + (TREEBIN_SHIFT-1)) & 1)));\
2784 + }\
2785 +}
2786 +
2787 +#else /* GNUC */
2788 +#define compute_tree_index(S, I)\
2789 +{\
2790 + size_t X = S >> TREEBIN_SHIFT;\
2791 + if (X == 0)\
2792 + I = 0;\
2793 + else if (X > 0xFFFF)\
2794 + I = NTREEBINS-1;\
2795 + else {\
2796 + unsigned int Y = (unsigned int)X;\
2797 + unsigned int N = ((Y - 0x100) >> 16) & 8;\
2798 + unsigned int K = (((Y <<= N) - 0x1000) >> 16) & 4;\
2799 + N += K;\
2800 + N += K = (((Y <<= K) - 0x4000) >> 16) & 2;\
2801 + K = 14 - N + ((Y <<= K) >> 15);\
2802 + I = (K << 1) + ((S >> (K + (TREEBIN_SHIFT-1)) & 1));\
2803 + }\
2804 +}
2805 +#endif /* GNUC */
2806 +
2807 +/* Bit representing maximum resolved size in a treebin at i */
2808 +#define bit_for_tree_index(i) \
2809 + (i == NTREEBINS-1)? (SIZE_T_BITSIZE-1) : (((i) >> 1) + TREEBIN_SHIFT - 2)
2810 +
2811 +/* Shift placing maximum resolved bit in a treebin at i as sign bit */
2812 +#define leftshift_for_tree_index(i) \
2813 + ((i == NTREEBINS-1)? 0 : \
2814 + ((SIZE_T_BITSIZE-SIZE_T_ONE) - (((i) >> 1) + TREEBIN_SHIFT - 2)))
2815 +
2816 +/* The size of the smallest chunk held in bin with index i */
2817 +#define minsize_for_tree_index(i) \
2818 + ((SIZE_T_ONE << (((i) >> 1) + TREEBIN_SHIFT)) | \
2819 + (((size_t)((i) & SIZE_T_ONE)) << (((i) >> 1) + TREEBIN_SHIFT - 1)))
2820 +
2821 +
2822 +/* ------------------------ Operations on bin maps ----------------------- */
2823 +
2824 +/* bit corresponding to given index */
2825 +#define idx2bit(i) ((binmap_t)(1) << (i))
2826 +
2827 +/* Mark/Clear bits with given index */
2828 +#define mark_smallmap(M,i) ((M)->smallmap |= idx2bit(i))
2829 +#define clear_smallmap(M,i) ((M)->smallmap &= ~idx2bit(i))
2830 +#define smallmap_is_marked(M,i) ((M)->smallmap & idx2bit(i))
2831 +
2832 +#define mark_treemap(M,i) ((M)->treemap |= idx2bit(i))
2833 +#define clear_treemap(M,i) ((M)->treemap &= ~idx2bit(i))
2834 +#define treemap_is_marked(M,i) ((M)->treemap & idx2bit(i))
2835 +
2836 +/* isolate the least set bit of a bitmap */
2837 +#define least_bit(x) ((x) & -(x))
2838 +
2839 +/* mask with all bits to left of least bit of x on */
2840 +#define left_bits(x) ((x<<1) | -(x<<1))
2841 +
2842 +/* mask with all bits to left of or equal to least bit of x on */
2843 +#define same_or_left_bits(x) ((x) | -(x))
2844 +
2845 +/* index corresponding to given bit. Use x86 asm if possible */
2846 +
2847 +#if defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__))
2848 +#define compute_bit2idx(X, I)\
2849 +{\
2850 + unsigned int J;\
2851 + __asm__("bsfl\t%1, %0\n\t" : "=r" (J) : "rm" (X));\
2852 + I = (bindex_t)J;\
2853 +}
2854 +
2855 +#elif defined (__INTEL_COMPILER)
2856 +#define compute_bit2idx(X, I)\
2857 +{\
2858 + unsigned int J;\
2859 + J = _bit_scan_forward (X); \
2860 + I = (bindex_t)J;\
2861 +}
2862 +
2863 +#elif defined(_MSC_VER) && _MSC_VER>=1300
2864 +#define compute_bit2idx(X, I)\
2865 +{\
2866 + unsigned int J;\
2867 + _BitScanForward((DWORD *) &J, X);\
2868 + I = (bindex_t)J;\
2869 +}
2870 +
2871 +#elif USE_BUILTIN_FFS
2872 +#define compute_bit2idx(X, I) I = ffs(X)-1
2873 +
2874 +#else
2875 +#define compute_bit2idx(X, I)\
2876 +{\
2877 + unsigned int Y = X - 1;\
2878 + unsigned int K = Y >> (16-4) & 16;\
2879 + unsigned int N = K; Y >>= K;\
2880 + N += K = Y >> (8-3) & 8; Y >>= K;\
2881 + N += K = Y >> (4-2) & 4; Y >>= K;\
2882 + N += K = Y >> (2-1) & 2; Y >>= K;\
2883 + N += K = Y >> (1-0) & 1; Y >>= K;\
2884 + I = (bindex_t)(N + Y);\
2885 +}
2886 +#endif /* GNUC */
2887 +
2888 +
2889 +/* ----------------------- Runtime Check Support ------------------------- */
2890 +
2891 +/*
2892 + For security, the main invariant is that malloc/free/etc never
2893 + writes to a static address other than malloc_state, unless static
2894 + malloc_state itself has been corrupted, which cannot occur via
2895 + malloc (because of these checks). In essence this means that we
2896 + believe all pointers, sizes, maps etc held in malloc_state, but
2897 + check all of those linked or offsetted from other embedded data
2898 + structures. These checks are interspersed with main code in a way
2899 + that tends to minimize their run-time cost.
2900 +
2901 + When FOOTERS is defined, in addition to range checking, we also
2902 + verify footer fields of inuse chunks, which can be used guarantee
2903 + that the mstate controlling malloc/free is intact. This is a
2904 + streamlined version of the approach described by William Robertson
2905 + et al in "Run-time Detection of Heap-based Overflows" LISA'03
2906 + http://www.usenix.org/events/lisa03/tech/robertson.html The footer
2907 + of an inuse chunk holds the xor of its mstate and a random seed,
2908 + that is checked upon calls to free() and realloc(). This is
2909 + (probablistically) unguessable from outside the program, but can be
2910 + computed by any code successfully malloc'ing any chunk, so does not
2911 + itself provide protection against code that has already broken
2912 + security through some other means. Unlike Robertson et al, we
2913 + always dynamically check addresses of all offset chunks (previous,
2914 + next, etc). This turns out to be cheaper than relying on hashes.
2915 +*/
2916 +
2917 +#if !INSECURE
2918 +/* Check if address a is at least as high as any from MORECORE or MMAP */
2919 +#define ok_address(M, a) ((char*)(a) >= (M)->least_addr)
2920 +/* Check if address of next chunk n is higher than base chunk p */
2921 +#define ok_next(p, n) ((char*)(p) < (char*)(n))
2922 +/* Check if p has its cinuse bit on */
2923 +#define ok_cinuse(p) cinuse(p)
2924 +/* Check if p has its pinuse bit on */
2925 +#define ok_pinuse(p) pinuse(p)
2926 +
2927 +#else /* !INSECURE */
2928 +#define ok_address(M, a) (1)
2929 +#define ok_next(b, n) (1)
2930 +#define ok_cinuse(p) (1)
2931 +#define ok_pinuse(p) (1)
2932 +#endif /* !INSECURE */
2933 +
2934 +#if (FOOTERS && !INSECURE)
2935 +/* Check if (alleged) mstate m has expected magic field */
2936 +#define ok_magic(M) ((M)->magic == mparams.magic)
2937 +#else /* (FOOTERS && !INSECURE) */
2938 +#define ok_magic(M) (1)
2939 +#endif /* (FOOTERS && !INSECURE) */
2940 +
2941 +
2942 +/* In gcc, use __builtin_expect to minimize impact of checks */
2943 +#if !INSECURE
2944 +#if defined(__GNUC__) && __GNUC__ >= 3
2945 +#define RTCHECK(e) __builtin_expect(e, 1)
2946 +#else /* GNUC */
2947 +#define RTCHECK(e) (e)
2948 +#endif /* GNUC */
2949 +#else /* !INSECURE */
2950 +#define RTCHECK(e) (1)
2951 +#endif /* !INSECURE */
2952 +
2953 +/* macros to set up inuse chunks with or without footers */
2954 +
2955 +#if !FOOTERS
2956 +
2957 +#define mark_inuse_foot(M,p,s)
2958 +
2959 +/* Set cinuse bit and pinuse bit of next chunk */
2960 +#define set_inuse(M,p,s)\
2961 + ((p)->head = (((p)->head & PINUSE_BIT)|s|CINUSE_BIT),\
2962 + ((mchunkptr)(((char*)(p)) + (s)))->head |= PINUSE_BIT)
2963 +
2964 +/* Set cinuse and pinuse of this chunk and pinuse of next chunk */
2965 +#define set_inuse_and_pinuse(M,p,s)\
2966 + ((p)->head = (s|PINUSE_BIT|CINUSE_BIT),\
2967 + ((mchunkptr)(((char*)(p)) + (s)))->head |= PINUSE_BIT)
2968 +
2969 +/* Set size, cinuse and pinuse bit of this chunk */
2970 +#define set_size_and_pinuse_of_inuse_chunk(M, p, s)\
2971 + ((p)->head = (s|PINUSE_BIT|CINUSE_BIT))
2972 +
2973 +#else /* FOOTERS */
2974 +
2975 +/* Set foot of inuse chunk to be xor of mstate and seed */
2976 +#define mark_inuse_foot(M,p,s)\
2977 + (((mchunkptr)((char*)(p) + (s)))->prev_foot = ((size_t)(M) ^ mparams.magic))
2978 +
2979 +#define get_mstate_for(p)\
2980 + ((mstate)(((mchunkptr)((char*)(p) +\
2981 + (chunksize(p))))->prev_foot ^ mparams.magic))
2982 +
2983 +#define set_inuse(M,p,s)\
2984 + ((p)->head = (((p)->head & PINUSE_BIT)|s|CINUSE_BIT),\
2985 + (((mchunkptr)(((char*)(p)) + (s)))->head |= PINUSE_BIT), \
2986 + mark_inuse_foot(M,p,s))
2987 +
2988 +#define set_inuse_and_pinuse(M,p,s)\
2989 + ((p)->head = (s|PINUSE_BIT|CINUSE_BIT),\
2990 + (((mchunkptr)(((char*)(p)) + (s)))->head |= PINUSE_BIT),\
2991 + mark_inuse_foot(M,p,s))
2992 +
2993 +#define set_size_and_pinuse_of_inuse_chunk(M, p, s)\
2994 + ((p)->head = (s|PINUSE_BIT|CINUSE_BIT),\
2995 + mark_inuse_foot(M, p, s))
2996 +
2997 +#endif /* !FOOTERS */
2998 +
2999 +/* ---------------------------- setting mparams -------------------------- */
3000 +
3001 +/* Initialize mparams */
3002 +static int init_mparams(void) {
3003 +#ifdef NEED_GLOBAL_LOCK_INIT
3004 + if (malloc_global_mutex_status <= 0)
3005 + init_malloc_global_mutex();
3006 +#endif
3007 +
3008 + ACQUIRE_MALLOC_GLOBAL_LOCK();
3009 + if (mparams.magic == 0) {
3010 + size_t magic;
3011 + size_t psize;
3012 + size_t gsize;
3013 +
3014 +#ifndef WIN32
3015 + psize = malloc_getpagesize;
3016 + gsize = ((DEFAULT_GRANULARITY != 0)? DEFAULT_GRANULARITY : psize);
3017 +#else /* WIN32 */
3018 + {
3019 + SYSTEM_INFO system_info;
3020 + GetSystemInfo(&system_info);
3021 + psize = system_info.dwPageSize;
3022 + gsize = ((DEFAULT_GRANULARITY != 0)?
3023 + DEFAULT_GRANULARITY : system_info.dwAllocationGranularity);
3024 + }
3025 +#endif /* WIN32 */
3026 +
3027 + /* Sanity-check configuration:
3028 + size_t must be unsigned and as wide as pointer type.
3029 + ints must be at least 4 bytes.
3030 + alignment must be at least 8.
3031 + Alignment, min chunk size, and page size must all be powers of 2.
3032 + */
3033 + if ((sizeof(size_t) != sizeof(char*)) ||
3034 + (MAX_SIZE_T < MIN_CHUNK_SIZE) ||
3035 + (sizeof(int) < 4) ||
3036 + (MALLOC_ALIGNMENT < (size_t)8U) ||
3037 + ((MALLOC_ALIGNMENT & (MALLOC_ALIGNMENT-SIZE_T_ONE)) != 0) ||
3038 + ((MCHUNK_SIZE & (MCHUNK_SIZE-SIZE_T_ONE)) != 0) ||
3039 + ((gsize & (gsize-SIZE_T_ONE)) != 0) ||
3040 + ((psize & (psize-SIZE_T_ONE)) != 0))
3041 + ABORT;
3042 +
3043 + mparams.granularity = gsize;
3044 + mparams.page_size = psize;
3045 + mparams.mmap_threshold = DEFAULT_MMAP_THRESHOLD;
3046 + mparams.trim_threshold = DEFAULT_TRIM_THRESHOLD;
3047 +#if MORECORE_CONTIGUOUS
3048 + mparams.default_mflags = USE_LOCK_BIT|USE_MMAP_BIT;
3049 +#else /* MORECORE_CONTIGUOUS */
3050 + mparams.default_mflags = USE_LOCK_BIT|USE_MMAP_BIT|USE_NONCONTIGUOUS_BIT;
3051 +#endif /* MORECORE_CONTIGUOUS */
3052 +
3053 +#if !ONLY_MSPACES
3054 + /* Set up lock for main malloc area */
3055 + gm->mflags = mparams.default_mflags;
3056 + INITIAL_LOCK(&gm->mutex);
3057 +#endif
3058 +
3059 +#if (FOOTERS && !INSECURE)
3060 + {
3061 +#if USE_DEV_RANDOM
3062 + int fd;
3063 + unsigned char buf[sizeof(size_t)];
3064 + /* Try to use /dev/urandom, else fall back on using time */
3065 + if ((fd = open("/dev/urandom", O_RDONLY)) >= 0 &&
3066 + read(fd, buf, sizeof(buf)) == sizeof(buf)) {
3067 + magic = *((size_t *) buf);
3068 + close(fd);
3069 + }
3070 + else
3071 +#endif /* USE_DEV_RANDOM */
3072 +#ifdef WIN32
3073 + magic = (size_t)(GetTickCount() ^ (size_t)0x55555555U);
3074 +#else
3075 + magic = (size_t)(time(0) ^ (size_t)0x55555555U);
3076 +#endif
3077 + magic |= (size_t)8U; /* ensure nonzero */
3078 + magic &= ~(size_t)7U; /* improve chances of fault for bad values */
3079 + }
3080 +#else /* (FOOTERS && !INSECURE) */
3081 + magic = (size_t)0x58585858U;
3082 +#endif /* (FOOTERS && !INSECURE) */
3083 +
3084 + mparams.magic = magic;
3085 + }
3086 +
3087 + RELEASE_MALLOC_GLOBAL_LOCK();
3088 + return 1;
3089 +}
3090 +
3091 +/* support for mallopt */
3092 +static int change_mparam(int param_number, int value) {
3093 + size_t val = (value == -1)? MAX_SIZE_T : (size_t)value;
3094 + ensure_initialization();
3095 + switch(param_number) {
3096 + case M_TRIM_THRESHOLD:
3097 + mparams.trim_threshold = val;
3098 + return 1;
3099 + case M_GRANULARITY:
3100 + if (val >= mparams.page_size && ((val & (val-1)) == 0)) {
3101 + mparams.granularity = val;
3102 + return 1;
3103 + }
3104 + else
3105 + return 0;
3106 + case M_MMAP_THRESHOLD:
3107 + mparams.mmap_threshold = val;
3108 + return 1;
3109 + default:
3110 + return 0;
3111 + }
3112 +}
3113 +
3114 +#if DEBUG
3115 +/* ------------------------- Debugging Support --------------------------- */
3116 +
3117 +/* Check properties of any chunk, whether free, inuse, mmapped etc */
3118 +static void do_check_any_chunk(mstate m, mchunkptr p) {
3119 + assert((is_aligned(chunk2mem(p))) || (p->head == FENCEPOST_HEAD));
3120 + assert(ok_address(m, p));
3121 +}
3122 +
3123 +/* Check properties of top chunk */
3124 +static void do_check_top_chunk(mstate m, mchunkptr p) {
3125 + msegmentptr sp = segment_holding(m, (char*)p);
3126 + size_t sz = p->head & ~INUSE_BITS; /* third-lowest bit can be set! */
3127 + assert(sp != 0);
3128 + assert((is_aligned(chunk2mem(p))) || (p->head == FENCEPOST_HEAD));
3129 + assert(ok_address(m, p));
3130 + assert(sz == m->topsize);
3131 + assert(sz > 0);
3132 + assert(sz == ((sp->base + sp->size) - (char*)p) - TOP_FOOT_SIZE);
3133 + assert(pinuse(p));
3134 + assert(!pinuse(chunk_plus_offset(p, sz)));
3135 +}
3136 +
3137 +/* Check properties of (inuse) mmapped chunks */
3138 +static void do_check_mmapped_chunk(mstate m, mchunkptr p) {
3139 + size_t sz = chunksize(p);
3140 + size_t len = (sz + (p->prev_foot & ~IS_MMAPPED_BIT) + MMAP_FOOT_PAD);
3141 + assert(is_mmapped(p));
3142 + assert(use_mmap(m));
3143 + assert((is_aligned(chunk2mem(p))) || (p->head == FENCEPOST_HEAD));
3144 + assert(ok_address(m, p));
3145 + assert(!is_small(sz));
3146 + assert((len & (mparams.page_size-SIZE_T_ONE)) == 0);
3147 + assert(chunk_plus_offset(p, sz)->head == FENCEPOST_HEAD);
3148 + assert(chunk_plus_offset(p, sz+SIZE_T_SIZE)->head == 0);
3149 +}
3150 +
3151 +/* Check properties of inuse chunks */
3152 +static void do_check_inuse_chunk(mstate m, mchunkptr p) {
3153 + do_check_any_chunk(m, p);
3154 + assert(cinuse(p));
3155 + assert(next_pinuse(p));
3156 + /* If not pinuse and not mmapped, previous chunk has OK offset */
3157 + assert(is_mmapped(p) || pinuse(p) || next_chunk(prev_chunk(p)) == p);
3158 + if (is_mmapped(p))
3159 + do_check_mmapped_chunk(m, p);
3160 +}
3161 +
3162 +/* Check properties of free chunks */
3163 +static void do_check_free_chunk(mstate m, mchunkptr p) {
3164 + size_t sz = chunksize(p);
3165 + mchunkptr next = chunk_plus_offset(p, sz);
3166 + do_check_any_chunk(m, p);
3167 + assert(!cinuse(p));
3168 + assert(!next_pinuse(p));
3169 + assert (!is_mmapped(p));
3170 + if (p != m->dv && p != m->top) {
3171 + if (sz >= MIN_CHUNK_SIZE) {
3172 + assert((sz & CHUNK_ALIGN_MASK) == 0);
3173 + assert(is_aligned(chunk2mem(p)));
3174 + assert(next->prev_foot == sz);
3175 + assert(pinuse(p));
3176 + assert (next == m->top || cinuse(next));
3177 + assert(p->fd->bk == p);
3178 + assert(p->bk->fd == p);
3179 + }
3180 + else /* markers are always of size SIZE_T_SIZE */
3181 + assert(sz == SIZE_T_SIZE);
3182 + }
3183 +}
3184 +
3185 +/* Check properties of malloced chunks at the point they are malloced */
3186 +static void do_check_malloced_chunk(mstate m, void* mem, size_t s) {
3187 + if (mem != 0) {
3188 + mchunkptr p = mem2chunk(mem);
3189 + size_t sz = p->head & ~(PINUSE_BIT|CINUSE_BIT);
3190 + do_check_inuse_chunk(m, p);
3191 + assert((sz & CHUNK_ALIGN_MASK) == 0);
3192 + assert(sz >= MIN_CHUNK_SIZE);
3193 + assert(sz >= s);
3194 + /* unless mmapped, size is less than MIN_CHUNK_SIZE more than request */
3195 + assert(is_mmapped(p) || sz < (s + MIN_CHUNK_SIZE));
3196 + }
3197 +}
3198 +
3199 +/* Check a tree and its subtrees. */
3200 +static void do_check_tree(mstate m, tchunkptr t) {
3201 + tchunkptr head = 0;
3202 + tchunkptr u = t;
3203 + bindex_t tindex = t->index;
3204 + size_t tsize = chunksize(t);
3205 + bindex_t idx;
3206 + compute_tree_index(tsize, idx);
3207 + assert(tindex == idx);
3208 + assert(tsize >= MIN_LARGE_SIZE);
3209 + assert(tsize >= minsize_for_tree_index(idx));
3210 + assert((idx == NTREEBINS-1) || (tsize < minsize_for_tree_index((idx+1))));
3211 +
3212 + do { /* traverse through chain of same-sized nodes */
3213 + do_check_any_chunk(m, ((mchunkptr)u));
3214 + assert(u->index == tindex);
3215 + assert(chunksize(u) == tsize);
3216 + assert(!cinuse(u));
3217 + assert(!next_pinuse(u));
3218 + assert(u->fd->bk == u);
3219 + assert(u->bk->fd == u);
3220 + if (u->parent == 0) {
3221 + assert(u->child[0] == 0);
3222 + assert(u->child[1] == 0);
3223 + }
3224 + else {
3225 + assert(head == 0); /* only one node on chain has parent */
3226 + head = u;
3227 + assert(u->parent != u);
3228 + assert (u->parent->child[0] == u ||
3229 + u->parent->child[1] == u ||
3230 + *((tbinptr*)(u->parent)) == u);
3231 + if (u->child[0] != 0) {
3232 + assert(u->child[0]->parent == u);
3233 + assert(u->child[0] != u);
3234 + do_check_tree(m, u->child[0]);
3235 + }
3236 + if (u->child[1] != 0) {
3237 + assert(u->child[1]->parent == u);
3238 + assert(u->child[1] != u);
3239 + do_check_tree(m, u->child[1]);
3240 + }
3241 + if (u->child[0] != 0 && u->child[1] != 0) {
3242 + assert(chunksize(u->child[0]) < chunksize(u->child[1]));
3243 + }
3244 + }
3245 + u = u->fd;
3246 + } while (u != t);
3247 + assert(head != 0);
3248 +}
3249 +
3250 +/* Check all the chunks in a treebin. */
3251 +static void do_check_treebin(mstate m, bindex_t i) {
3252 + tbinptr* tb = treebin_at(m, i);
3253 + tchunkptr t = *tb;
3254 + int empty = (m->treemap & (1U << i)) == 0;
3255 + if (t == 0)
3256 + assert(empty);
3257 + if (!empty)
3258 + do_check_tree(m, t);
3259 +}
3260 +
3261 +/* Check all the chunks in a smallbin. */
3262 +static void do_check_smallbin(mstate m, bindex_t i) {
3263 + sbinptr b = smallbin_at(m, i);
3264 + mchunkptr p = b->bk;
3265 + unsigned int empty = (m->smallmap & (1U << i)) == 0;
3266 + if (p == b)
3267 + assert(empty);
3268 + if (!empty) {
3269 + for (; p != b; p = p->bk) {
3270 + size_t size = chunksize(p);
3271 + mchunkptr q;
3272 + /* each chunk claims to be free */
3273 + do_check_free_chunk(m, p);
3274 + /* chunk belongs in bin */
3275 + assert(small_index(size) == i);
3276 + assert(p->bk == b || chunksize(p->bk) == chunksize(p));
3277 + /* chunk is followed by an inuse chunk */
3278 + q = next_chunk(p);
3279 + if (q->head != FENCEPOST_HEAD)
3280 + do_check_inuse_chunk(m, q);
3281 + }
3282 + }
3283 +}
3284 +
3285 +/* Find x in a bin. Used in other check functions. */
3286 +static int bin_find(mstate m, mchunkptr x) {
3287 + size_t size = chunksize(x);
3288 + if (is_small(size)) {
3289 + bindex_t sidx = small_index(size);
3290 + sbinptr b = smallbin_at(m, sidx);
3291 + if (smallmap_is_marked(m, sidx)) {
3292 + mchunkptr p = b;
3293 + do {
3294 + if (p == x)
3295 + return 1;
3296 + } while ((p = p->fd) != b);
3297 + }
3298 + }
3299 + else {
3300 + bindex_t tidx;
3301 + compute_tree_index(size, tidx);
3302 + if (treemap_is_marked(m, tidx)) {
3303 + tchunkptr t = *treebin_at(m, tidx);
3304 + size_t sizebits = size << leftshift_for_tree_index(tidx);
3305 + while (t != 0 && chunksize(t) != size) {
3306 + t = t->child[(sizebits >> (SIZE_T_BITSIZE-SIZE_T_ONE)) & 1];
3307 + sizebits <<= 1;
3308 + }
3309 + if (t != 0) {
3310 + tchunkptr u = t;
3311 + do {
3312 + if (u == (tchunkptr)x)
3313 + return 1;
3314 + } while ((u = u->fd) != t);
3315 + }
3316 + }
3317 + }
3318 + return 0;
3319 +}
3320 +
3321 +/* Traverse each chunk and check it; return total */
3322 +static size_t traverse_and_check(mstate m) {
3323 + size_t sum = 0;
3324 + if (is_initialized(m)) {
3325 + msegmentptr s = &m->seg;
3326 + sum += m->topsize + TOP_FOOT_SIZE;
3327 + while (s != 0) {
3328 + mchunkptr q = align_as_chunk(s->base);
3329 + mchunkptr lastq = 0;
3330 + assert(pinuse(q));
3331 + while (segment_holds(s, q) &&
3332 + q != m->top && q->head != FENCEPOST_HEAD) {
3333 + sum += chunksize(q);
3334 + if (cinuse(q)) {
3335 + assert(!bin_find(m, q));
3336 + do_check_inuse_chunk(m, q);
3337 + }
3338 + else {
3339 + assert(q == m->dv || bin_find(m, q));
3340 + assert(lastq == 0 || cinuse(lastq)); /* Not 2 consecutive free */
3341 + do_check_free_chunk(m, q);
3342 + }
3343 + lastq = q;
3344 + q = next_chunk(q);
3345 + }
3346 + s = s->next;
3347 + }
3348 + }
3349 + return sum;
3350 +}
3351 +
3352 +/* Check all properties of malloc_state. */
3353 +static void do_check_malloc_state(mstate m) {
3354 + bindex_t i;
3355 + size_t total;
3356 + /* check bins */
3357 + for (i = 0; i < NSMALLBINS; ++i)
3358 + do_check_smallbin(m, i);
3359 + for (i = 0; i < NTREEBINS; ++i)
3360 + do_check_treebin(m, i);
3361 +
3362 + if (m->dvsize != 0) { /* check dv chunk */
3363 + do_check_any_chunk(m, m->dv);
3364 + assert(m->dvsize == chunksize(m->dv));
3365 + assert(m->dvsize >= MIN_CHUNK_SIZE);
3366 + assert(bin_find(m, m->dv) == 0);
3367 + }
3368 +
3369 + if (m->top != 0) { /* check top chunk */
3370 + do_check_top_chunk(m, m->top);
3371 + /*assert(m->topsize == chunksize(m->top)); redundant */
3372 + assert(m->topsize > 0);
3373 + assert(bin_find(m, m->top) == 0);
3374 + }
3375 +
3376 + total = traverse_and_check(m);
3377 + assert(total <= m->footprint);
3378 + assert(m->footprint <= m->max_footprint);
3379 +}
3380 +#endif /* DEBUG */
3381 +
3382 +/* ----------------------------- statistics ------------------------------ */
3383 +
3384 +#if !NO_MALLINFO
3385 +static struct mallinfo internal_mallinfo(mstate m) {
3386 + struct mallinfo nm = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 };
3387 + ensure_initialization();
3388 + if (!PREACTION(m)) {
3389 + check_malloc_state(m);
3390 + if (is_initialized(m)) {
3391 + size_t nfree = SIZE_T_ONE; /* top always free */
3392 + size_t mfree = m->topsize + TOP_FOOT_SIZE;
3393 + size_t sum = mfree;
3394 + msegmentptr s = &m->seg;
3395 + while (s != 0) {
3396 + mchunkptr q = align_as_chunk(s->base);
3397 + while (segment_holds(s, q) &&
3398 + q != m->top && q->head != FENCEPOST_HEAD) {
3399 + size_t sz = chunksize(q);
3400 + sum += sz;
3401 + if (!cinuse(q)) {
3402 + mfree += sz;
3403 + ++nfree;
3404 + }
3405 + q = next_chunk(q);
3406 + }
3407 + s = s->next;
3408 + }
3409 +
3410 + nm.arena = sum;
3411 + nm.ordblks = nfree;
3412 + nm.hblkhd = m->footprint - sum;
3413 + nm.usmblks = m->max_footprint;
3414 + nm.uordblks = m->footprint - mfree;
3415 + nm.fordblks = mfree;
3416 + nm.keepcost = m->topsize;
3417 + }
3418 +
3419 + POSTACTION(m);
3420 + }
3421 + return nm;
3422 +}
3423 +#endif /* !NO_MALLINFO */
3424 +
3425 +static void internal_malloc_stats(mstate m) {
3426 + ensure_initialization();
3427 + if (!PREACTION(m)) {
3428 + size_t maxfp = 0;
3429 + size_t fp = 0;
3430 + size_t used = 0;
3431 + check_malloc_state(m);
3432 + if (is_initialized(m)) {
3433 + msegmentptr s = &m->seg;
3434 + maxfp = m->max_footprint;
3435 + fp = m->footprint;
3436 + used = fp - (m->topsize + TOP_FOOT_SIZE);
3437 +
3438 + while (s != 0) {
3439 + mchunkptr q = align_as_chunk(s->base);
3440 + while (segment_holds(s, q) &&
3441 + q != m->top && q->head != FENCEPOST_HEAD) {
3442 + if (!cinuse(q))
3443 + used -= chunksize(q);
3444 + q = next_chunk(q);
3445 + }
3446 + s = s->next;
3447 + }
3448 + }
3449 +
3450 + fprintf(stderr, "max system bytes = %10lu\n", (unsigned long)(maxfp));
3451 + fprintf(stderr, "system bytes = %10lu\n", (unsigned long)(fp));
3452 + fprintf(stderr, "in use bytes = %10lu\n", (unsigned long)(used));
3453 +
3454 + POSTACTION(m);
3455 + }
3456 +}
3457 +
3458 +/* ----------------------- Operations on smallbins ----------------------- */
3459 +
3460 +/*
3461 + Various forms of linking and unlinking are defined as macros. Even
3462 + the ones for trees, which are very long but have very short typical
3463 + paths. This is ugly but reduces reliance on inlining support of
3464 + compilers.
3465 +*/
3466 +
3467 +/* Link a free chunk into a smallbin */
3468 +#define insert_small_chunk(M, P, S) {\
3469 + bindex_t I = small_index(S);\
3470 + mchunkptr B = smallbin_at(M, I);\
3471 + mchunkptr F = B;\
3472 + assert(S >= MIN_CHUNK_SIZE);\
3473 + if (!smallmap_is_marked(M, I))\
3474 + mark_smallmap(M, I);\
3475 + else if (RTCHECK(ok_address(M, B->fd)))\
3476 + F = B->fd;\
3477 + else {\
3478 + CORRUPTION_ERROR_ACTION(M);\
3479 + }\
3480 + B->fd = P;\
3481 + F->bk = P;\
3482 + P->fd = F;\
3483 + P->bk = B;\
3484 +}
3485 +
3486 +/* Unlink a chunk from a smallbin */
3487 +#define unlink_small_chunk(M, P, S) {\
3488 + mchunkptr F = P->fd;\
3489 + mchunkptr B = P->bk;\
3490 + bindex_t I = small_index(S);\
3491 + assert(P != B);\
3492 + assert(P != F);\
3493 + assert(chunksize(P) == small_index2size(I));\
3494 + if (F == B)\
3495 + clear_smallmap(M, I);\
3496 + else if (RTCHECK((F == smallbin_at(M,I) || ok_address(M, F)) &&\
3497 + (B == smallbin_at(M,I) || ok_address(M, B)))) {\
3498 + F->bk = B;\
3499 + B->fd = F;\
3500 + }\
3501 + else {\
3502 + CORRUPTION_ERROR_ACTION(M);\
3503 + }\
3504 +}
3505 +
3506 +/* Unlink the first chunk from a smallbin */
3507 +#define unlink_first_small_chunk(M, B, P, I) {\
3508 + mchunkptr F = P->fd;\
3509 + assert(P != B);\
3510 + assert(P != F);\
3511 + assert(chunksize(P) == small_index2size(I));\
3512 + if (B == F)\
3513 + clear_smallmap(M, I);\
3514 + else if (RTCHECK(ok_address(M, F))) {\
3515 + B->fd = F;\
3516 + F->bk = B;\
3517 + }\
3518 + else {\
3519 + CORRUPTION_ERROR_ACTION(M);\
3520 + }\
3521 +}
3522 +
3523 +
3524 +
3525 +/* Replace dv node, binning the old one */
3526 +/* Used only when dvsize known to be small */
3527 +#define replace_dv(M, P, S) {\
3528 + size_t DVS = M->dvsize;\
3529 + if (DVS != 0) {\
3530 + mchunkptr DV = M->dv;\
3531 + assert(is_small(DVS));\
3532 + insert_small_chunk(M, DV, DVS);\
3533 + }\
3534 + M->dvsize = S;\
3535 + M->dv = P;\
3536 +}
3537 +
3538 +/* ------------------------- Operations on trees ------------------------- */
3539 +
3540 +/* Insert chunk into tree */
3541 +#define insert_large_chunk(M, X, S) {\
3542 + tbinptr* H;\
3543 + bindex_t I;\
3544 + compute_tree_index(S, I);\
3545 + H = treebin_at(M, I);\
3546 + X->index = I;\
3547 + X->child[0] = X->child[1] = 0;\
3548 + if (!treemap_is_marked(M, I)) {\
3549 + mark_treemap(M, I);\
3550 + *H = X;\
3551 + X->parent = (tchunkptr)H;\
3552 + X->fd = X->bk = X;\
3553 + }\
3554 + else {\
3555 + tchunkptr T = *H;\
3556 + size_t K = S << leftshift_for_tree_index(I);\
3557 + for (;;) {\
3558 + if (chunksize(T) != S) {\
3559 + tchunkptr* C = &(T->child[(K >> (SIZE_T_BITSIZE-SIZE_T_ONE)) & 1]);\
3560 + K <<= 1;\
3561 + if (*C != 0)\
3562 + T = *C;\
3563 + else if (RTCHECK(ok_address(M, C))) {\
3564 + *C = X;\
3565 + X->parent = T;\
3566 + X->fd = X->bk = X;\
3567 + break;\
3568 + }\
3569 + else {\
3570 + CORRUPTION_ERROR_ACTION(M);\
3571 + break;\
3572 + }\
3573 + }\
3574 + else {\
3575 + tchunkptr F = T->fd;\
3576 + if (RTCHECK(ok_address(M, T) && ok_address(M, F))) {\
3577 + T->fd = F->bk = X;\
3578 + X->fd = F;\
3579 + X->bk = T;\
3580 + X->parent = 0;\
3581 + break;\
3582 + }\
3583 + else {\
3584 + CORRUPTION_ERROR_ACTION(M);\
3585 + break;\
3586 + }\
3587 + }\
3588 + }\
3589 + }\
3590 +}
3591 +
3592 +/*
3593 + Unlink steps:
3594 +
3595 + 1. If x is a chained node, unlink it from its same-sized fd/bk links
3596 + and choose its bk node as its replacement.
3597 + 2. If x was the last node of its size, but not a leaf node, it must
3598 + be replaced with a leaf node (not merely one with an open left or
3599 + right), to make sure that lefts and rights of descendents
3600 + correspond properly to bit masks. We use the rightmost descendent
3601 + of x. We could use any other leaf, but this is easy to locate and
3602 + tends to counteract removal of leftmosts elsewhere, and so keeps
3603 + paths shorter than minimally guaranteed. This doesn't loop much
3604 + because on average a node in a tree is near the bottom.
3605 + 3. If x is the base of a chain (i.e., has parent links) relink
3606 + x's parent and children to x's replacement (or null if none).
3607 +*/
3608 +
3609 +#define unlink_large_chunk(M, X) {\
3610 + tchunkptr XP = X->parent;\
3611 + tchunkptr R;\
3612 + if (X->bk != X) {\
3613 + tchunkptr F = X->fd;\
3614 + R = X->bk;\
3615 + if (RTCHECK(ok_address(M, F))) {\
3616 + F->bk = R;\
3617 + R->fd = F;\
3618 + }\
3619 + else {\
3620 + CORRUPTION_ERROR_ACTION(M);\
3621 + }\
3622 + }\
3623 + else {\
3624 + tchunkptr* RP;\
3625 + if (((R = *(RP = &(X->child[1]))) != 0) ||\
3626 + ((R = *(RP = &(X->child[0]))) != 0)) {\
3627 + tchunkptr* CP;\
3628 + while ((*(CP = &(R->child[1])) != 0) ||\
3629 + (*(CP = &(R->child[0])) != 0)) {\
3630 + R = *(RP = CP);\
3631 + }\
3632 + if (RTCHECK(ok_address(M, RP)))\
3633 + *RP = 0;\
3634 + else {\
3635 + CORRUPTION_ERROR_ACTION(M);\
3636 + }\
3637 + }\
3638 + }\
3639 + if (XP != 0) {\
3640 + tbinptr* H = treebin_at(M, X->index);\
3641 + if (X == *H) {\
3642 + if ((*H = R) == 0) \
3643 + clear_treemap(M, X->index);\
3644 + }\
3645 + else if (RTCHECK(ok_address(M, XP))) {\
3646 + if (XP->child[0] == X) \
3647 + XP->child[0] = R;\
3648 + else \
3649 + XP->child[1] = R;\
3650 + }\
3651 + else\
3652 + CORRUPTION_ERROR_ACTION(M);\
3653 + if (R != 0) {\
3654 + if (RTCHECK(ok_address(M, R))) {\
3655 + tchunkptr C0, C1;\
3656 + R->parent = XP;\
3657 + if ((C0 = X->child[0]) != 0) {\
3658 + if (RTCHECK(ok_address(M, C0))) {\
3659 + R->child[0] = C0;\
3660 + C0->parent = R;\
3661 + }\
3662 + else\
3663 + CORRUPTION_ERROR_ACTION(M);\
3664 + }\
3665 + if ((C1 = X->child[1]) != 0) {\
3666 + if (RTCHECK(ok_address(M, C1))) {\
3667 + R->child[1] = C1;\
3668 + C1->parent = R;\
3669 + }\
3670 + else\
3671 + CORRUPTION_ERROR_ACTION(M);\
3672 + }\
3673 + }\
3674 + else\
3675 + CORRUPTION_ERROR_ACTION(M);\
3676 + }\
3677 + }\
3678 +}
3679 +
3680 +/* Relays to large vs small bin operations */
3681 +
3682 +#define insert_chunk(M, P, S)\
3683 + if (is_small(S)) insert_small_chunk(M, P, S)\
3684 + else { tchunkptr TP = (tchunkptr)(P); insert_large_chunk(M, TP, S); }
3685 +
3686 +#define unlink_chunk(M, P, S)\
3687 + if (is_small(S)) unlink_small_chunk(M, P, S)\
3688 + else { tchunkptr TP = (tchunkptr)(P); unlink_large_chunk(M, TP); }
3689 +
3690 +
3691 +/* Relays to internal calls to malloc/free from realloc, memalign etc */
3692 +
3693 +#if ONLY_MSPACES
3694 +#define internal_malloc(m, b) mspace_malloc(m, b)
3695 +#define internal_free(m, mem) mspace_free(m,mem);
3696 +#else /* ONLY_MSPACES */
3697 +#if MSPACES
3698 +#define internal_malloc(m, b)\
3699 + (m == gm)? dlmalloc(b) : mspace_malloc(m, b)
3700 +#define internal_free(m, mem)\
3701 + if (m == gm) dlfree(mem); else mspace_free(m,mem);
3702 +#else /* MSPACES */
3703 +#define internal_malloc(m, b) dlmalloc(b)
3704 +#define internal_free(m, mem) dlfree(mem)
3705 +#endif /* MSPACES */
3706 +#endif /* ONLY_MSPACES */
3707 +
3708 +/* ----------------------- Direct-mmapping chunks ----------------------- */
3709 +
3710 +/*
3711 + Directly mmapped chunks are set up with an offset to the start of
3712 + the mmapped region stored in the prev_foot field of the chunk. This
3713 + allows reconstruction of the required argument to MUNMAP when freed,
3714 + and also allows adjustment of the returned chunk to meet alignment
3715 + requirements (especially in memalign). There is also enough space
3716 + allocated to hold a fake next chunk of size SIZE_T_SIZE to maintain
3717 + the PINUSE bit so frees can be checked.
3718 +*/
3719 +
3720 +/* Malloc using mmap */
3721 +static void* mmap_alloc(mstate m, size_t nb) {
3722 + size_t mmsize = mmap_align(nb + SIX_SIZE_T_SIZES + CHUNK_ALIGN_MASK);
3723 + if (mmsize > nb) { /* Check for wrap around 0 */
3724 + char* mm = (char*)(CALL_DIRECT_MMAP(mmsize));
3725 + if (mm != CMFAIL) {
3726 + size_t offset = align_offset(chunk2mem(mm));
3727 + size_t psize = mmsize - offset - MMAP_FOOT_PAD;
3728 + mchunkptr p = (mchunkptr)(mm + offset);
3729 + p->prev_foot = offset | IS_MMAPPED_BIT;
3730 + (p)->head = (psize|CINUSE_BIT);
3731 + mark_inuse_foot(m, p, psize);
3732 + chunk_plus_offset(p, psize)->head = FENCEPOST_HEAD;
3733 + chunk_plus_offset(p, psize+SIZE_T_SIZE)->head = 0;
3734 +
3735 + if (mm < m->least_addr)
3736 + m->least_addr = mm;
3737 + if ((m->footprint += mmsize) > m->max_footprint)
3738 + m->max_footprint = m->footprint;
3739 + assert(is_aligned(chunk2mem(p)));
3740 + check_mmapped_chunk(m, p);
3741 + return chunk2mem(p);
3742 + }
3743 + }
3744 + return 0;
3745 +}
3746 +
3747 +/* Realloc using mmap */
3748 +static mchunkptr mmap_resize(mstate m, mchunkptr oldp, size_t nb) {
3749 + size_t oldsize = chunksize(oldp);
3750 + if (is_small(nb)) /* Can't shrink mmap regions below small size */
3751 + return 0;
3752 + /* Keep old chunk if big enough but not too big */
3753 + if (oldsize >= nb + SIZE_T_SIZE &&
3754 + (oldsize - nb) <= (mparams.granularity << 1))
3755 + return oldp;
3756 + else {
3757 + size_t offset = oldp->prev_foot & ~IS_MMAPPED_BIT;
3758 + size_t oldmmsize = oldsize + offset + MMAP_FOOT_PAD;
3759 + size_t newmmsize = mmap_align(nb + SIX_SIZE_T_SIZES + CHUNK_ALIGN_MASK);
3760 + char* cp = (char*)CALL_MREMAP((char*)oldp - offset,
3761 + oldmmsize, newmmsize, 1);
3762 + if (cp != CMFAIL) {
3763 + mchunkptr newp = (mchunkptr)(cp + offset);
3764 + size_t psize = newmmsize - offset - MMAP_FOOT_PAD;
3765 + newp->head = (psize|CINUSE_BIT);
3766 + mark_inuse_foot(m, newp, psize);
3767 + chunk_plus_offset(newp, psize)->head = FENCEPOST_HEAD;
3768 + chunk_plus_offset(newp, psize+SIZE_T_SIZE)->head = 0;
3769 +
3770 + if (cp < m->least_addr)
3771 + m->least_addr = cp;
3772 + if ((m->footprint += newmmsize - oldmmsize) > m->max_footprint)
3773 + m->max_footprint = m->footprint;
3774 + check_mmapped_chunk(m, newp);
3775 + return newp;
3776 + }
3777 + }
3778 + return 0;
3779 +}
3780 +
3781 +/* -------------------------- mspace management -------------------------- */
3782 +
3783 +/* Initialize top chunk and its size */
3784 +static void init_top(mstate m, mchunkptr p, size_t psize) {
3785 + /* Ensure alignment */
3786 + size_t offset = align_offset(chunk2mem(p));
3787 + p = (mchunkptr)((char*)p + offset);
3788 + psize -= offset;
3789 +
3790 + m->top = p;
3791 + m->topsize = psize;
3792 + p->head = psize | PINUSE_BIT;
3793 + /* set size of fake trailing chunk holding overhead space only once */
3794 + chunk_plus_offset(p, psize)->head = TOP_FOOT_SIZE;
3795 + m->trim_check = mparams.trim_threshold; /* reset on each update */
3796 +}
3797 +
3798 +/* Initialize bins for a new mstate that is otherwise zeroed out */
3799 +static void init_bins(mstate m) {
3800 + /* Establish circular links for smallbins */
3801 + bindex_t i;
3802 + for (i = 0; i < NSMALLBINS; ++i) {
3803 + sbinptr bin = smallbin_at(m,i);
3804 + bin->fd = bin->bk = bin;
3805 + }
3806 +}
3807 +
3808 +#if PROCEED_ON_ERROR
3809 +
3810 +/* default corruption action */
3811 +static void reset_on_error(mstate m) {
3812 + int i;
3813 + ++malloc_corruption_error_count;
3814 + /* Reinitialize fields to forget about all memory */
3815 + m->smallbins = m->treebins = 0;
3816 + m->dvsize = m->topsize = 0;
3817 + m->seg.base = 0;
3818 + m->seg.size = 0;
3819 + m->seg.next = 0;
3820 + m->top = m->dv = 0;
3821 + for (i = 0; i < NTREEBINS; ++i)
3822 + *treebin_at(m, i) = 0;
3823 + init_bins(m);
3824 +}
3825 +#endif /* PROCEED_ON_ERROR */
3826 +
3827 +/* Allocate chunk and prepend remainder with chunk in successor base. */
3828 +static void* prepend_alloc(mstate m, char* newbase, char* oldbase,
3829 + size_t nb) {
3830 + mchunkptr p = align_as_chunk(newbase);
3831 + mchunkptr oldfirst = align_as_chunk(oldbase);
3832 + size_t psize = (char*)oldfirst - (char*)p;
3833 + mchunkptr q = chunk_plus_offset(p, nb);
3834 + size_t qsize = psize - nb;
3835 + set_size_and_pinuse_of_inuse_chunk(m, p, nb);
3836 +
3837 + assert((char*)oldfirst > (char*)q);
3838 + assert(pinuse(oldfirst));
3839 + assert(qsize >= MIN_CHUNK_SIZE);
3840 +
3841 + /* consolidate remainder with first chunk of old base */
3842 + if (oldfirst == m->top) {
3843 + size_t tsize = m->topsize += qsize;
3844 + m->top = q;
3845 + q->head = tsize | PINUSE_BIT;
3846 + check_top_chunk(m, q);
3847 + }
3848 + else if (oldfirst == m->dv) {
3849 + size_t dsize = m->dvsize += qsize;
3850 + m->dv = q;
3851 + set_size_and_pinuse_of_free_chunk(q, dsize);
3852 + }
3853 + else {
3854 + if (!cinuse(oldfirst)) {
3855 + size_t nsize = chunksize(oldfirst);
3856 + unlink_chunk(m, oldfirst, nsize);
3857 + oldfirst = chunk_plus_offset(oldfirst, nsize);
3858 + qsize += nsize;
3859 + }
3860 + set_free_with_pinuse(q, qsize, oldfirst);
3861 + insert_chunk(m, q, qsize);
3862 + check_free_chunk(m, q);
3863 + }
3864 +
3865 + check_malloced_chunk(m, chunk2mem(p), nb);
3866 + return chunk2mem(p);
3867 +}
3868 +
3869 +/* Add a segment to hold a new noncontiguous region */
3870 +static void add_segment(mstate m, char* tbase, size_t tsize, flag_t mmapped) {
3871 + /* Determine locations and sizes of segment, fenceposts, old top */
3872 + char* old_top = (char*)m->top;
3873 + msegmentptr oldsp = segment_holding(m, old_top);
3874 + char* old_end = oldsp->base + oldsp->size;
3875 + size_t ssize = pad_request(sizeof(struct malloc_segment));
3876 + char* rawsp = old_end - (ssize + FOUR_SIZE_T_SIZES + CHUNK_ALIGN_MASK);
3877 + size_t offset = align_offset(chunk2mem(rawsp));
3878 + char* asp = rawsp + offset;
3879 + char* csp = (asp < (old_top + MIN_CHUNK_SIZE))? old_top : asp;
3880 + mchunkptr sp = (mchunkptr)csp;
3881 + msegmentptr ss = (msegmentptr)(chunk2mem(sp));
3882 + mchunkptr tnext = chunk_plus_offset(sp, ssize);
3883 + mchunkptr p = tnext;
3884 + int nfences = 0;
3885 +
3886 + /* reset top to new space */
3887 + init_top(m, (mchunkptr)tbase, tsize - TOP_FOOT_SIZE);
3888 +
3889 + /* Set up segment record */
3890 + assert(is_aligned(ss));
3891 + set_size_and_pinuse_of_inuse_chunk(m, sp, ssize);
3892 + *ss = m->seg; /* Push current record */
3893 + m->seg.base = tbase;
3894 + m->seg.size = tsize;
3895 + m->seg.sflags = mmapped;
3896 + m->seg.next = ss;
3897 +
3898 + /* Insert trailing fenceposts */
3899 + for (;;) {
3900 + mchunkptr nextp = chunk_plus_offset(p, SIZE_T_SIZE);
3901 + p->head = FENCEPOST_HEAD;
3902 + ++nfences;
3903 + if ((char*)(&(nextp->head)) < old_end)
3904 + p = nextp;
3905 + else
3906 + break;
3907 + }
3908 + assert(nfences >= 2);
3909 +
3910 + /* Insert the rest of old top into a bin as an ordinary free chunk */
3911 + if (csp != old_top) {
3912 + mchunkptr q = (mchunkptr)old_top;
3913 + size_t psize = csp - old_top;
3914 + mchunkptr tn = chunk_plus_offset(q, psize);
3915 + set_free_with_pinuse(q, psize, tn);
3916 + insert_chunk(m, q, psize);
3917 + }
3918 +
3919 + check_top_chunk(m, m->top);
3920 +}
3921 +
3922 +/* -------------------------- System allocation -------------------------- */
3923 +
3924 +/* Get memory from system using MORECORE or MMAP */
3925 +static void* sys_alloc(mstate m, size_t nb) {
3926 + char* tbase = CMFAIL;
3927 + size_t tsize = 0;
3928 + flag_t mmap_flag = 0;
3929 +
3930 + ensure_initialization();
3931 +
3932 + /* Directly map large chunks */
3933 + if (use_mmap(m) && nb >= mparams.mmap_threshold) {
3934 + void* mem = mmap_alloc(m, nb);
3935 + if (mem != 0)
3936 + return mem;
3937 + }
3938 +
3939 + /*
3940 + Try getting memory in any of three ways (in most-preferred to
3941 + least-preferred order):
3942 + 1. A call to MORECORE that can normally contiguously extend memory.
3943 + (disabled if not MORECORE_CONTIGUOUS or not HAVE_MORECORE or
3944 + or main space is mmapped or a previous contiguous call failed)
3945 + 2. A call to MMAP new space (disabled if not HAVE_MMAP).
3946 + Note that under the default settings, if MORECORE is unable to
3947 + fulfill a request, and HAVE_MMAP is true, then mmap is
3948 + used as a noncontiguous system allocator. This is a useful backup
3949 + strategy for systems with holes in address spaces -- in this case
3950 + sbrk cannot contiguously expand the heap, but mmap may be able to
3951 + find space.
3952 + 3. A call to MORECORE that cannot usually contiguously extend memory.
3953 + (disabled if not HAVE_MORECORE)
3954 +
3955 + In all cases, we need to request enough bytes from system to ensure
3956 + we can malloc nb bytes upon success, so pad with enough space for
3957 + top_foot, plus alignment-pad to make sure we don't lose bytes if
3958 + not on boundary, and round this up to a granularity unit.
3959 + */
3960 +
3961 + if (MORECORE_CONTIGUOUS && !use_noncontiguous(m)) {
3962 + char* br = CMFAIL;
3963 + msegmentptr ss = (m->top == 0)? 0 : segment_holding(m, (char*)m->top);
3964 + size_t asize = 0;
3965 + ACQUIRE_MALLOC_GLOBAL_LOCK();
3966 +
3967 + if (ss == 0) { /* First time through or recovery */
3968 + char* base = (char*)CALL_MORECORE(0);
3969 + if (base != CMFAIL) {
3970 + asize = granularity_align(nb + SYS_ALLOC_PADDING);
3971 + /* Adjust to end on a page boundary */
3972 + if (!is_page_aligned(base))
3973 + asize += (page_align((size_t)base) - (size_t)base);
3974 + /* Can't call MORECORE if size is negative when treated as signed */
3975 + if (asize < HALF_MAX_SIZE_T &&
3976 + (br = (char*)(CALL_MORECORE(asize))) == base) {
3977 + tbase = base;
3978 + tsize = asize;
3979 + }
3980 + }
3981 + }
3982 + else {
3983 + /* Subtract out existing available top space from MORECORE request. */
3984 + asize = granularity_align(nb - m->topsize + SYS_ALLOC_PADDING);
3985 + /* Use mem here only if it did continuously extend old space */
3986 + if (asize < HALF_MAX_SIZE_T &&
3987 + (br = (char*)(CALL_MORECORE(asize))) == ss->base+ss->size) {
3988 + tbase = br;
3989 + tsize = asize;
3990 + }
3991 + }
3992 +
3993 + if (tbase == CMFAIL) { /* Cope with partial failure */
3994 + if (br != CMFAIL) { /* Try to use/extend the space we did get */
3995 + if (asize < HALF_MAX_SIZE_T &&
3996 + asize < nb + SYS_ALLOC_PADDING) {
3997 + size_t esize = granularity_align(nb + SYS_ALLOC_PADDING - asize);
3998 + if (esize < HALF_MAX_SIZE_T) {
3999 + char* end = (char*)CALL_MORECORE(esize);
4000 + if (end != CMFAIL)
4001 + asize += esize;
4002 + else { /* Can't use; try to release */
4003 + (void) CALL_MORECORE(-asize);
4004 + br = CMFAIL;
4005 + }
4006 + }
4007 + }
4008 + }
4009 + if (br != CMFAIL) { /* Use the space we did get */
4010 + tbase = br;
4011 + tsize = asize;
4012 + }
4013 + else
4014 + disable_contiguous(m); /* Don't try contiguous path in the future */
4015 + }
4016 +
4017 + RELEASE_MALLOC_GLOBAL_LOCK();
4018 + }
4019 +
4020 + if (HAVE_MMAP && tbase == CMFAIL) { /* Try MMAP */
4021 + size_t rsize = granularity_align(nb + SYS_ALLOC_PADDING);
4022 + if (rsize > nb) { /* Fail if wraps around zero */
4023 + char* mp = (char*)(CALL_MMAP(rsize));
4024 + if (mp != CMFAIL) {
4025 + tbase = mp;
4026 + tsize = rsize;
4027 + mmap_flag = IS_MMAPPED_BIT;
4028 + }
4029 + }
4030 + }
4031 +
4032 + if (HAVE_MORECORE && tbase == CMFAIL) { /* Try noncontiguous MORECORE */
4033 + size_t asize = granularity_align(nb + SYS_ALLOC_PADDING);
4034 + if (asize < HALF_MAX_SIZE_T) {
4035 + char* br = CMFAIL;
4036 + char* end = CMFAIL;
4037 + ACQUIRE_MALLOC_GLOBAL_LOCK();
4038 + br = (char*)(CALL_MORECORE(asize));
4039 + end = (char*)(CALL_MORECORE(0));
4040 + RELEASE_MALLOC_GLOBAL_LOCK();
4041 + if (br != CMFAIL && end != CMFAIL && br < end) {
4042 + size_t ssize = end - br;
4043 + if (ssize > nb + TOP_FOOT_SIZE) {
4044 + tbase = br;
4045 + tsize = ssize;
4046 + }
4047 + }
4048 + }
4049 + }
4050 +
4051 + if (tbase != CMFAIL) {
4052 +
4053 + if ((m->footprint += tsize) > m->max_footprint)
4054 + m->max_footprint = m->footprint;
4055 +
4056 + if (!is_initialized(m)) { /* first-time initialization */
4057 + m->seg.base = m->least_addr = tbase;
4058 + m->seg.size = tsize;
4059 + m->seg.sflags = mmap_flag;
4060 + m->magic = mparams.magic;
4061 + m->release_checks = MAX_RELEASE_CHECK_RATE;
4062 + init_bins(m);
4063 +#if !ONLY_MSPACES
4064 + if (is_global(m))
4065 + init_top(m, (mchunkptr)tbase, tsize - TOP_FOOT_SIZE);
4066 + else
4067 +#endif
4068 + {
4069 + /* Offset top by embedded malloc_state */
4070 + mchunkptr mn = next_chunk(mem2chunk(m));
4071 + init_top(m, mn, (size_t)((tbase + tsize) - (char*)mn) -TOP_FOOT_SIZE);
4072 + }
4073 + }
4074 +
4075 + else {
4076 + /* Try to merge with an existing segment */
4077 + msegmentptr sp = &m->seg;
4078 + /* Only consider most recent segment if traversal suppressed */
4079 + while (sp != 0 && tbase != sp->base + sp->size)
4080 + sp = (NO_SEGMENT_TRAVERSAL) ? 0 : sp->next;
4081 + if (sp != 0 &&
4082 + !is_extern_segment(sp) &&
4083 + (sp->sflags & IS_MMAPPED_BIT) == mmap_flag &&
4084 + segment_holds(sp, m->top)) { /* append */
4085 + sp->size += tsize;
4086 + init_top(m, m->top, m->topsize + tsize);
4087 + }
4088 + else {
4089 + if (tbase < m->least_addr)
4090 + m->least_addr = tbase;
4091 + sp = &m->seg;
4092 + while (sp != 0 && sp->base != tbase + tsize)
4093 + sp = (NO_SEGMENT_TRAVERSAL) ? 0 : sp->next;
4094 + if (sp != 0 &&
4095 + !is_extern_segment(sp) &&
4096 + (sp->sflags & IS_MMAPPED_BIT) == mmap_flag) {
4097 + char* oldbase = sp->base;
4098 + sp->base = tbase;
4099 + sp->size += tsize;
4100 + return prepend_alloc(m, tbase, oldbase, nb);
4101 + }
4102 + else
4103 + add_segment(m, tbase, tsize, mmap_flag);
4104 + }
4105 + }
4106 +
4107 + if (nb < m->topsize) { /* Allocate from new or extended top space */
4108 + size_t rsize = m->topsize -= nb;
4109 + mchunkptr p = m->top;
4110 + mchunkptr r = m->top = chunk_plus_offset(p, nb);
4111 + r->head = rsize | PINUSE_BIT;
4112 + set_size_and_pinuse_of_inuse_chunk(m, p, nb);
4113 + check_top_chunk(m, m->top);
4114 + check_malloced_chunk(m, chunk2mem(p), nb);
4115 + return chunk2mem(p);
4116 + }
4117 + }
4118 +
4119 + MALLOC_FAILURE_ACTION;
4120 + return 0;
4121 +}
4122 +
4123 +/* ----------------------- system deallocation -------------------------- */
4124 +
4125 +/* Unmap and unlink any mmapped segments that don't contain used chunks */
4126 +static size_t release_unused_segments(mstate m) {
4127 + size_t released = 0;
4128 + int nsegs = 0;
4129 + msegmentptr pred = &m->seg;
4130 + msegmentptr sp = pred->next;
4131 + while (sp != 0) {
4132 + char* base = sp->base;
4133 + size_t size = sp->size;
4134 + msegmentptr next = sp->next;
4135 + ++nsegs;
4136 + if (is_mmapped_segment(sp) && !is_extern_segment(sp)) {
4137 + mchunkptr p = align_as_chunk(base);
4138 + size_t psize = chunksize(p);
4139 + /* Can unmap if first chunk holds entire segment and not pinned */
4140 + if (!cinuse(p) && (char*)p + psize >= base + size - TOP_FOOT_SIZE) {
4141 + tchunkptr tp = (tchunkptr)p;
4142 + assert(segment_holds(sp, (char*)sp));
4143 + if (p == m->dv) {
4144 + m->dv = 0;
4145 + m->dvsize = 0;
4146 + }
4147 + else {
4148 + unlink_large_chunk(m, tp);
4149 + }
4150 + if (CALL_MUNMAP(base, size) == 0) {
4151 + released += size;
4152 + m->footprint -= size;
4153 + /* unlink obsoleted record */
4154 + sp = pred;
4155 + sp->next = next;
4156 + }
4157 + else { /* back out if cannot unmap */
4158 + insert_large_chunk(m, tp, psize);
4159 + }
4160 + }
4161 + }
4162 + if (NO_SEGMENT_TRAVERSAL) /* scan only first segment */
4163 + break;
4164 + pred = sp;
4165 + sp = next;
4166 + }
4167 + /* Reset check counter */
4168 + m->release_checks = ((nsegs > MAX_RELEASE_CHECK_RATE)?
4169 + nsegs : MAX_RELEASE_CHECK_RATE);
4170 + return released;
4171 +}
4172 +
4173 +static int sys_trim(mstate m, size_t pad) {
4174 + size_t released = 0;
4175 + ensure_initialization();
4176 + if (pad < MAX_REQUEST && is_initialized(m)) {
4177 + pad += TOP_FOOT_SIZE; /* ensure enough room for segment overhead */
4178 +
4179 + if (m->topsize > pad) {
4180 + /* Shrink top space in granularity-size units, keeping at least one */
4181 + size_t unit = mparams.granularity;
4182 + size_t extra = ((m->topsize - pad + (unit - SIZE_T_ONE)) / unit -
4183 + SIZE_T_ONE) * unit;
4184 + msegmentptr sp = segment_holding(m, (char*)m->top);
4185 +
4186 + if (!is_extern_segment(sp)) {
4187 + if (is_mmapped_segment(sp)) {
4188 + if (HAVE_MMAP &&
4189 + sp->size >= extra &&
4190 + !has_segment_link(m, sp)) { /* can't shrink if pinned */
4191 + size_t newsize = sp->size - extra;
4192 + /* Prefer mremap, fall back to munmap */
4193 + if ((CALL_MREMAP(sp->base, sp->size, newsize, 0) != MFAIL) ||
4194 + (CALL_MUNMAP(sp->base + newsize, extra) == 0)) {
4195 + released = extra;
4196 + }
4197 + }
4198 + }
4199 + else if (HAVE_MORECORE) {
4200 + if (extra >= HALF_MAX_SIZE_T) /* Avoid wrapping negative */
4201 + extra = (HALF_MAX_SIZE_T) + SIZE_T_ONE - unit;
4202 + ACQUIRE_MALLOC_GLOBAL_LOCK();
4203 + {
4204 + /* Make sure end of memory is where we last set it. */
4205 + char* old_br = (char*)(CALL_MORECORE(0));
4206 + if (old_br == sp->base + sp->size) {
4207 + char* rel_br = (char*)(CALL_MORECORE(-extra));
4208 + char* new_br = (char*)(CALL_MORECORE(0));
4209 + if (rel_br != CMFAIL && new_br < old_br)
4210 + released = old_br - new_br;
4211 + }
4212 + }
4213 + RELEASE_MALLOC_GLOBAL_LOCK();
4214 + }
4215 + }
4216 +
4217 + if (released != 0) {
4218 + sp->size -= released;
4219 + m->footprint -= released;
4220 + init_top(m, m->top, m->topsize - released);
4221 + check_top_chunk(m, m->top);
4222 + }
4223 + }
4224 +
4225 + /* Unmap any unused mmapped segments */
4226 + if (HAVE_MMAP)
4227 + released += release_unused_segments(m);
4228 +
4229 + /* On failure, disable autotrim to avoid repeated failed future calls */
4230 + if (released == 0 && m->topsize > m->trim_check)
4231 + m->trim_check = MAX_SIZE_T;
4232 + }
4233 +
4234 + return (released != 0)? 1 : 0;
4235 +}
4236 +
4237 +
4238 +/* ---------------------------- malloc support --------------------------- */
4239 +
4240 +/* allocate a large request from the best fitting chunk in a treebin */
4241 +static void* tmalloc_large(mstate m, size_t nb) {
4242 + tchunkptr v = 0;
4243 + size_t rsize = -nb; /* Unsigned negation */
4244 + tchunkptr t;
4245 + bindex_t idx;
4246 + compute_tree_index(nb, idx);
4247 + if ((t = *treebin_at(m, idx)) != 0) {
4248 + /* Traverse tree for this bin looking for node with size == nb */
4249 + size_t sizebits = nb << leftshift_for_tree_index(idx);
4250 + tchunkptr rst = 0; /* The deepest untaken right subtree */
4251 + for (;;) {
4252 + tchunkptr rt;
4253 + size_t trem = chunksize(t) - nb;
4254 + if (trem < rsize) {
4255 + v = t;
4256 + if ((rsize = trem) == 0)
4257 + break;
4258 + }
4259 + rt = t->child[1];
4260 + t = t->child[(sizebits >> (SIZE_T_BITSIZE-SIZE_T_ONE)) & 1];
4261 + if (rt != 0 && rt != t)
4262 + rst = rt;
4263 + if (t == 0) {
4264 + t = rst; /* set t to least subtree holding sizes > nb */
4265 + break;
4266 + }
4267 + sizebits <<= 1;
4268 + }
4269 + }
4270 + if (t == 0 && v == 0) { /* set t to root of next non-empty treebin */
4271 + binmap_t leftbits = left_bits(idx2bit(idx)) & m->treemap;
4272 + if (leftbits != 0) {
4273 + bindex_t i;
4274 + binmap_t leastbit = least_bit(leftbits);
4275 + compute_bit2idx(leastbit, i);
4276 + t = *treebin_at(m, i);
4277 + }
4278 + }
4279 +
4280 + while (t != 0) { /* find smallest of tree or subtree */
4281 + size_t trem = chunksize(t) - nb;
4282 + if (trem < rsize) {
4283 + rsize = trem;
4284 + v = t;
4285 + }
4286 + t = leftmost_child(t);
4287 + }
4288 +
4289 + /* If dv is a better fit, return 0 so malloc will use it */
4290 + if (v != 0 && rsize < (size_t)(m->dvsize - nb)) {
4291 + if (RTCHECK(ok_address(m, v))) { /* split */
4292 + mchunkptr r = chunk_plus_offset(v, nb);
4293 + assert(chunksize(v) == rsize + nb);
4294 + if (RTCHECK(ok_next(v, r))) {
4295 + unlink_large_chunk(m, v);
4296 + if (rsize < MIN_CHUNK_SIZE)
4297 + set_inuse_and_pinuse(m, v, (rsize + nb));
4298 + else {
4299 + set_size_and_pinuse_of_inuse_chunk(m, v, nb);
4300 + set_size_and_pinuse_of_free_chunk(r, rsize);
4301 + insert_chunk(m, r, rsize);
4302 + }
4303 + return chunk2mem(v);
4304 + }
4305 + }
4306 + CORRUPTION_ERROR_ACTION(m);
4307 + }
4308 + return 0;
4309 +}
4310 +
4311 +/* allocate a small request from the best fitting chunk in a treebin */
4312 +static void* tmalloc_small(mstate m, size_t nb) {
4313 + tchunkptr t, v;
4314 + size_t rsize;
4315 + bindex_t i;
4316 + binmap_t leastbit = least_bit(m->treemap);
4317 + compute_bit2idx(leastbit, i);
4318 + v = t = *treebin_at(m, i);
4319 + rsize = chunksize(t) - nb;
4320 +
4321 + while ((t = leftmost_child(t)) != 0) {
4322 + size_t trem = chunksize(t) - nb;
4323 + if (trem < rsize) {
4324 + rsize = trem;
4325 + v = t;
4326 + }
4327 + }
4328 +
4329 + if (RTCHECK(ok_address(m, v))) {
4330 + mchunkptr r = chunk_plus_offset(v, nb);
4331 + assert(chunksize(v) == rsize + nb);
4332 + if (RTCHECK(ok_next(v, r))) {
4333 + unlink_large_chunk(m, v);
4334 + if (rsize < MIN_CHUNK_SIZE)
4335 + set_inuse_and_pinuse(m, v, (rsize + nb));
4336 + else {
4337 + set_size_and_pinuse_of_inuse_chunk(m, v, nb);
4338 + set_size_and_pinuse_of_free_chunk(r, rsize);
4339 + replace_dv(m, r, rsize);
4340 + }
4341 + return chunk2mem(v);
4342 + }
4343 + }
4344 +
4345 + CORRUPTION_ERROR_ACTION(m);
4346 + return 0;
4347 +}
4348 +
4349 +/* --------------------------- realloc support --------------------------- */
4350 +
4351 +static void* internal_realloc(mstate m, void* oldmem, size_t bytes) {
4352 + if (bytes >= MAX_REQUEST) {
4353 + MALLOC_FAILURE_ACTION;
4354 + return 0;
4355 + }
4356 + if (!PREACTION(m)) {
4357 + mchunkptr oldp = mem2chunk(oldmem);
4358 + size_t oldsize = chunksize(oldp);
4359 + mchunkptr next = chunk_plus_offset(oldp, oldsize);
4360 + mchunkptr newp = 0;
4361 + void* extra = 0;
4362 +
4363 + /* Try to either shrink or extend into top. Else malloc-copy-free */
4364 +
4365 + if (RTCHECK(ok_address(m, oldp) && ok_cinuse(oldp) &&
4366 + ok_next(oldp, next) && ok_pinuse(next))) {
4367 + size_t nb = request2size(bytes);
4368 + if (is_mmapped(oldp))
4369 + newp = mmap_resize(m, oldp, nb);
4370 + else if (oldsize >= nb) { /* already big enough */
4371 + size_t rsize = oldsize - nb;
4372 + newp = oldp;
4373 + if (rsize >= MIN_CHUNK_SIZE) {
4374 + mchunkptr remainder = chunk_plus_offset(newp, nb);
4375 + set_inuse(m, newp, nb);
4376 + set_inuse(m, remainder, rsize);
4377 + extra = chunk2mem(remainder);
4378 + }
4379 + }
4380 + else if (next == m->top && oldsize + m->topsize > nb) {
4381 + /* Expand into top */
4382 + size_t newsize = oldsize + m->topsize;
4383 + size_t newtopsize = newsize - nb;
4384 + mchunkptr newtop = chunk_plus_offset(oldp, nb);
4385 + set_inuse(m, oldp, nb);
4386 + newtop->head = newtopsize |PINUSE_BIT;
4387 + m->top = newtop;
4388 + m->topsize = newtopsize;
4389 + newp = oldp;
4390 + }
4391 + }
4392 + else {
4393 + USAGE_ERROR_ACTION(m, oldmem);
4394 + POSTACTION(m);
4395 + return 0;
4396 + }
4397 +
4398 + POSTACTION(m);
4399 +
4400 + if (newp != 0) {
4401 + if (extra != 0) {
4402 + internal_free(m, extra);
4403 + }
4404 + check_inuse_chunk(m, newp);
4405 + return chunk2mem(newp);
4406 + }
4407 + else {
4408 + void* newmem = internal_malloc(m, bytes);
4409 + if (newmem != 0) {
4410 + size_t oc = oldsize - overhead_for(oldp);
4411 + memcpy(newmem, oldmem, (oc < bytes)? oc : bytes);
4412 + internal_free(m, oldmem);
4413 + }
4414 + return newmem;
4415 + }
4416 + }
4417 + return 0;
4418 +}
4419 +
4420 +/* --------------------------- memalign support -------------------------- */
4421 +
4422 +static void* internal_memalign(mstate m, size_t alignment, size_t bytes) {
4423 + if (alignment <= MALLOC_ALIGNMENT) /* Can just use malloc */
4424 + return internal_malloc(m, bytes);
4425 + if (alignment < MIN_CHUNK_SIZE) /* must be at least a minimum chunk size */
4426 + alignment = MIN_CHUNK_SIZE;
4427 + if ((alignment & (alignment-SIZE_T_ONE)) != 0) {/* Ensure a power of 2 */
4428 + size_t a = MALLOC_ALIGNMENT << 1;
4429 + while (a < alignment) a <<= 1;
4430 + alignment = a;
4431 + }
4432 +
4433 + if (bytes >= MAX_REQUEST - alignment) {
4434 + if (m != 0) { /* Test isn't needed but avoids compiler warning */
4435 + MALLOC_FAILURE_ACTION;
4436 + }
4437 + }
4438 + else {
4439 + size_t nb = request2size(bytes);
4440 + size_t req = nb + alignment + MIN_CHUNK_SIZE - CHUNK_OVERHEAD;
4441 + char* mem = (char*)internal_malloc(m, req);
4442 + if (mem != 0) {
4443 + void* leader = 0;
4444 + void* trailer = 0;
4445 + mchunkptr p = mem2chunk(mem);
4446 +
4447 + if (PREACTION(m)) return 0;
4448 + if ((((size_t)(mem)) % alignment) != 0) { /* misaligned */
4449 + /*
4450 + Find an aligned spot inside chunk. Since we need to give
4451 + back leading space in a chunk of at least MIN_CHUNK_SIZE, if
4452 + the first calculation places us at a spot with less than
4453 + MIN_CHUNK_SIZE leader, we can move to the next aligned spot.
4454 + We've allocated enough total room so that this is always
4455 + possible.
4456 + */
4457 + char* br = (char*)mem2chunk((size_t)(((size_t)(mem +
4458 + alignment -
4459 + SIZE_T_ONE)) &
4460 + -alignment));
4461 + char* pos = ((size_t)(br - (char*)(p)) >= MIN_CHUNK_SIZE)?
4462 + br : br+alignment;
4463 + mchunkptr newp = (mchunkptr)pos;
4464 + size_t leadsize = pos - (char*)(p);
4465 + size_t newsize = chunksize(p) - leadsize;
4466 +
4467 + if (is_mmapped(p)) { /* For mmapped chunks, just adjust offset */
4468 + newp->prev_foot = p->prev_foot + leadsize;
4469 + newp->head = (newsize|CINUSE_BIT);
4470 + }
4471 + else { /* Otherwise, give back leader, use the rest */
4472 + set_inuse(m, newp, newsize);
4473 + set_inuse(m, p, leadsize);
4474 + leader = chunk2mem(p);
4475 + }
4476 + p = newp;
4477 + }
4478 +
4479 + /* Give back spare room at the end */
4480 + if (!is_mmapped(p)) {
4481 + size_t size = chunksize(p);
4482 + if (size > nb + MIN_CHUNK_SIZE) {
4483 + size_t remainder_size = size - nb;
4484 + mchunkptr remainder = chunk_plus_offset(p, nb);
4485 + set_inuse(m, p, nb);
4486 + set_inuse(m, remainder, remainder_size);
4487 + trailer = chunk2mem(remainder);
4488 + }
4489 + }
4490 +
4491 + assert (chunksize(p) >= nb);
4492 + assert((((size_t)(chunk2mem(p))) % alignment) == 0);
4493 + check_inuse_chunk(m, p);
4494 + POSTACTION(m);
4495 + if (leader != 0) {
4496 + internal_free(m, leader);
4497 + }
4498 + if (trailer != 0) {
4499 + internal_free(m, trailer);
4500 + }
4501 + return chunk2mem(p);
4502 + }
4503 + }
4504 + return 0;
4505 +}
4506 +
4507 +/* ------------------------ comalloc/coalloc support --------------------- */
4508 +
4509 +static void** ialloc(mstate m,
4510 + size_t n_elements,
4511 + size_t* sizes,
4512 + int opts,
4513 + void* chunks[]) {
4514 + /*
4515 + This provides common support for independent_X routines, handling
4516 + all of the combinations that can result.
4517 +
4518 + The opts arg has:
4519 + bit 0 set if all elements are same size (using sizes[0])
4520 + bit 1 set if elements should be zeroed
4521 + */
4522 +
4523 + size_t element_size; /* chunksize of each element, if all same */
4524 + size_t contents_size; /* total size of elements */
4525 + size_t array_size; /* request size of pointer array */
4526 + void* mem; /* malloced aggregate space */
4527 + mchunkptr p; /* corresponding chunk */
4528 + size_t remainder_size; /* remaining bytes while splitting */
4529 + void** marray; /* either "chunks" or malloced ptr array */
4530 + mchunkptr array_chunk; /* chunk for malloced ptr array */
4531 + flag_t was_enabled; /* to disable mmap */
4532 + size_t size;
4533 + size_t i;
4534 +
4535 + ensure_initialization();
4536 + /* compute array length, if needed */
4537 + if (chunks != 0) {
4538 + if (n_elements == 0)
4539 + return chunks; /* nothing to do */
4540 + marray = chunks;
4541 + array_size = 0;
4542 + }
4543 + else {
4544 + /* if empty req, must still return chunk representing empty array */
4545 + if (n_elements == 0)
4546 + return (void**)internal_malloc(m, 0);
4547 + marray = 0;
4548 + array_size = request2size(n_elements * (sizeof(void*)));
4549 + }
4550 +
4551 + /* compute total element size */
4552 + if (opts & 0x1) { /* all-same-size */
4553 + element_size = request2size(*sizes);
4554 + contents_size = n_elements * element_size;
4555 + }
4556 + else { /* add up all the sizes */
4557 + element_size = 0;
4558 + contents_size = 0;
4559 + for (i = 0; i != n_elements; ++i)
4560 + contents_size += request2size(sizes[i]);
4561 + }
4562 +
4563 + size = contents_size + array_size;
4564 +
4565 + /*
4566 + Allocate the aggregate chunk. First disable direct-mmapping so
4567 + malloc won't use it, since we would not be able to later
4568 + free/realloc space internal to a segregated mmap region.
4569 + */
4570 + was_enabled = use_mmap(m);
4571 + disable_mmap(m);
4572 + mem = internal_malloc(m, size - CHUNK_OVERHEAD);
4573 + if (was_enabled)
4574 + enable_mmap(m);
4575 + if (mem == 0)
4576 + return 0;
4577 +
4578 + if (PREACTION(m)) return 0;
4579 + p = mem2chunk(mem);
4580 + remainder_size = chunksize(p);
4581 +
4582 + assert(!is_mmapped(p));
4583 +
4584 + if (opts & 0x2) { /* optionally clear the elements */
4585 + memset((size_t*)mem, 0, remainder_size - SIZE_T_SIZE - array_size);
4586 + }
4587 +
4588 + /* If not provided, allocate the pointer array as final part of chunk */
4589 + if (marray == 0) {
4590 + size_t array_chunk_size;
4591 + array_chunk = chunk_plus_offset(p, contents_size);
4592 + array_chunk_size = remainder_size - contents_size;
4593 + marray = (void**) (chunk2mem(array_chunk));
4594 + set_size_and_pinuse_of_inuse_chunk(m, array_chunk, array_chunk_size);
4595 + remainder_size = contents_size;
4596 + }
4597 +
4598 + /* split out elements */
4599 + for (i = 0; ; ++i) {
4600 + marray[i] = chunk2mem(p);
4601 + if (i != n_elements-1) {
4602 + if (element_size != 0)
4603 + size = element_size;
4604 + else
4605 + size = request2size(sizes[i]);
4606 + remainder_size -= size;
4607 + set_size_and_pinuse_of_inuse_chunk(m, p, size);
4608 + p = chunk_plus_offset(p, size);
4609 + }
4610 + else { /* the final element absorbs any overallocation slop */
4611 + set_size_and_pinuse_of_inuse_chunk(m, p, remainder_size);
4612 + break;
4613 + }
4614 + }
4615 +
4616 +#if DEBUG
4617 + if (marray != chunks) {
4618 + /* final element must have exactly exhausted chunk */
4619 + if (element_size != 0) {
4620 + assert(remainder_size == element_size);
4621 + }
4622 + else {
4623 + assert(remainder_size == request2size(sizes[i]));
4624 + }
4625 + check_inuse_chunk(m, mem2chunk(marray));
4626 + }
4627 + for (i = 0; i != n_elements; ++i)
4628 + check_inuse_chunk(m, mem2chunk(marray[i]));
4629 +
4630 +#endif /* DEBUG */
4631 +
4632 + POSTACTION(m);
4633 + return marray;
4634 +}
4635 +
4636 +
4637 +/* -------------------------- public routines ---------------------------- */
4638 +
4639 +#if !ONLY_MSPACES
4640 +
4641 +void* dlmalloc(size_t bytes) {
4642 + /*
4643 + Basic algorithm:
4644 + If a small request (< 256 bytes minus per-chunk overhead):
4645 + 1. If one exists, use a remainderless chunk in associated smallbin.
4646 + (Remainderless means that there are too few excess bytes to
4647 + represent as a chunk.)
4648 + 2. If it is big enough, use the dv chunk, which is normally the
4649 + chunk adjacent to the one used for the most recent small request.
4650 + 3. If one exists, split the smallest available chunk in a bin,
4651 + saving remainder in dv.
4652 + 4. If it is big enough, use the top chunk.
4653 + 5. If available, get memory from system and use it
4654 + Otherwise, for a large request:
4655 + 1. Find the smallest available binned chunk that fits, and use it
4656 + if it is better fitting than dv chunk, splitting if necessary.
4657 + 2. If better fitting than any binned chunk, use the dv chunk.
4658 + 3. If it is big enough, use the top chunk.
4659 + 4. If request size >= mmap threshold, try to directly mmap this chunk.
4660 + 5. If available, get memory from system and use it
4661 +
4662 + The ugly goto's here ensure that postaction occurs along all paths.
4663 + */
4664 +
4665 +#if USE_LOCKS
4666 + ensure_initialization(); /* initialize in sys_alloc if not using locks */
4667 +#endif
4668 +
4669 + if (!PREACTION(gm)) {
4670 + void* mem;
4671 + size_t nb;
4672 + if (bytes <= MAX_SMALL_REQUEST) {
4673 + bindex_t idx;
4674 + binmap_t smallbits;
4675 + nb = (bytes < MIN_REQUEST)? MIN_CHUNK_SIZE : pad_request(bytes);
4676 + idx = small_index(nb);
4677 + smallbits = gm->smallmap >> idx;
4678 +
4679 + if ((smallbits & 0x3U) != 0) { /* Remainderless fit to a smallbin. */
4680 + mchunkptr b, p;
4681 + idx += ~smallbits & 1; /* Uses next bin if idx empty */
4682 + b = smallbin_at(gm, idx);
4683 + p = b->fd;
4684 + assert(chunksize(p) == small_index2size(idx));
4685 + unlink_first_small_chunk(gm, b, p, idx);
4686 + set_inuse_and_pinuse(gm, p, small_index2size(idx));
4687 + mem = chunk2mem(p);
4688 + check_malloced_chunk(gm, mem, nb);
4689 + goto postaction;
4690 + }
4691 +
4692 + else if (nb > gm->dvsize) {
4693 + if (smallbits != 0) { /* Use chunk in next nonempty smallbin */
4694 + mchunkptr b, p, r;
4695 + size_t rsize;
4696 + bindex_t i;
4697 + binmap_t leftbits = (smallbits << idx) & left_bits(idx2bit(idx));
4698 + binmap_t leastbit = least_bit(leftbits);
4699 + compute_bit2idx(leastbit, i);
4700 + b = smallbin_at(gm, i);
4701 + p = b->fd;
4702 + assert(chunksize(p) == small_index2size(i));
4703 + unlink_first_small_chunk(gm, b, p, i);
4704 + rsize = small_index2size(i) - nb;
4705 + /* Fit here cannot be remainderless if 4byte sizes */
4706 + if (SIZE_T_SIZE != 4 && rsize < MIN_CHUNK_SIZE)
4707 + set_inuse_and_pinuse(gm, p, small_index2size(i));
4708 + else {
4709 + set_size_and_pinuse_of_inuse_chunk(gm, p, nb);
4710 + r = chunk_plus_offset(p, nb);
4711 + set_size_and_pinuse_of_free_chunk(r, rsize);
4712 + replace_dv(gm, r, rsize);
4713 + }
4714 + mem = chunk2mem(p);
4715 + check_malloced_chunk(gm, mem, nb);
4716 + goto postaction;
4717 + }
4718 +
4719 + else if (gm->treemap != 0 && (mem = tmalloc_small(gm, nb)) != 0) {
4720 + check_malloced_chunk(gm, mem, nb);
4721 + goto postaction;
4722 + }
4723 + }
4724 + }
4725 + else if (bytes >= MAX_REQUEST)
4726 + nb = MAX_SIZE_T; /* Too big to allocate. Force failure (in sys alloc) */
4727 + else {
4728 + nb = pad_request(bytes);
4729 + if (gm->treemap != 0 && (mem = tmalloc_large(gm, nb)) != 0) {
4730 + check_malloced_chunk(gm, mem, nb);
4731 + goto postaction;
4732 + }
4733 + }
4734 +
4735 + if (nb <= gm->dvsize) {
4736 + size_t rsize = gm->dvsize - nb;
4737 + mchunkptr p = gm->dv;
4738 + if (rsize >= MIN_CHUNK_SIZE) { /* split dv */
4739 + mchunkptr r = gm->dv = chunk_plus_offset(p, nb);
4740 + gm->dvsize = rsize;
4741 + set_size_and_pinuse_of_free_chunk(r, rsize);
4742 + set_size_and_pinuse_of_inuse_chunk(gm, p, nb);
4743 + }
4744 + else { /* exhaust dv */
4745 + size_t dvs = gm->dvsize;
4746 + gm->dvsize = 0;
4747 + gm->dv = 0;
4748 + set_inuse_and_pinuse(gm, p, dvs);
4749 + }
4750 + mem = chunk2mem(p);
4751 + check_malloced_chunk(gm, mem, nb);
4752 + goto postaction;
4753 + }
4754 +
4755 + else if (nb < gm->topsize) { /* Split top */
4756 + size_t rsize = gm->topsize -= nb;
4757 + mchunkptr p = gm->top;
4758 + mchunkptr r = gm->top = chunk_plus_offset(p, nb);
4759 + r->head = rsize | PINUSE_BIT;
4760 + set_size_and_pinuse_of_inuse_chunk(gm, p, nb);
4761 + mem = chunk2mem(p);
4762 + check_top_chunk(gm, gm->top);
4763 + check_malloced_chunk(gm, mem, nb);
4764 + goto postaction;
4765 + }
4766 +
4767 + mem = sys_alloc(gm, nb);
4768 +
4769 + postaction:
4770 + POSTACTION(gm);
4771 + return mem;
4772 + }
4773 +
4774 + return 0;
4775 +}
4776 +
4777 +void dlfree(void* mem) {
4778 + /*
4779 + Consolidate freed chunks with preceeding or succeeding bordering
4780 + free chunks, if they exist, and then place in a bin. Intermixed
4781 + with special cases for top, dv, mmapped chunks, and usage errors.
4782 + */
4783 +
4784 + if (mem != 0) {
4785 + mchunkptr p = mem2chunk(mem);
4786 +#if FOOTERS
4787 + mstate fm = get_mstate_for(p);
4788 + if (!ok_magic(fm)) {
4789 + USAGE_ERROR_ACTION(fm, p);
4790 + return;
4791 + }
4792 +#else /* FOOTERS */
4793 +#define fm gm
4794 +#endif /* FOOTERS */
4795 + if (!PREACTION(fm)) {
4796 + check_inuse_chunk(fm, p);
4797 + if (RTCHECK(ok_address(fm, p) && ok_cinuse(p))) {
4798 + size_t psize = chunksize(p);
4799 + mchunkptr next = chunk_plus_offset(p, psize);
4800 + if (!pinuse(p)) {
4801 + size_t prevsize = p->prev_foot;
4802 + if ((prevsize & IS_MMAPPED_BIT) != 0) {
4803 + prevsize &= ~IS_MMAPPED_BIT;
4804 + psize += prevsize + MMAP_FOOT_PAD;
4805 + if (CALL_MUNMAP((char*)p - prevsize, psize) == 0)
4806 + fm->footprint -= psize;
4807 + goto postaction;
4808 + }
4809 + else {
4810 + mchunkptr prev = chunk_minus_offset(p, prevsize);
4811 + psize += prevsize;
4812 + p = prev;
4813 + if (RTCHECK(ok_address(fm, prev))) { /* consolidate backward */
4814 + if (p != fm->dv) {
4815 + unlink_chunk(fm, p, prevsize);
4816 + }
4817 + else if ((next->head & INUSE_BITS) == INUSE_BITS) {
4818 + fm->dvsize = psize;
4819 + set_free_with_pinuse(p, psize, next);
4820 + goto postaction;
4821 + }
4822 + }
4823 + else
4824 + goto erroraction;
4825 + }
4826 + }
4827 +
4828 + if (RTCHECK(ok_next(p, next) && ok_pinuse(next))) {
4829 + if (!cinuse(next)) { /* consolidate forward */
4830 + if (next == fm->top) {
4831 + size_t tsize = fm->topsize += psize;
4832 + fm->top = p;
4833 + p->head = tsize | PINUSE_BIT;
4834 + if (p == fm->dv) {
4835 + fm->dv = 0;
4836 + fm->dvsize = 0;
4837 + }
4838 + if (should_trim(fm, tsize))
4839 + sys_trim(fm, 0);
4840 + goto postaction;
4841 + }
4842 + else if (next == fm->dv) {
4843 + size_t dsize = fm->dvsize += psize;
4844 + fm->dv = p;
4845 + set_size_and_pinuse_of_free_chunk(p, dsize);
4846 + goto postaction;
4847 + }
4848 + else {
4849 + size_t nsize = chunksize(next);
4850 + psize += nsize;
4851 + unlink_chunk(fm, next, nsize);
4852 + set_size_and_pinuse_of_free_chunk(p, psize);
4853 + if (p == fm->dv) {
4854 + fm->dvsize = psize;
4855 + goto postaction;
4856 + }
4857 + }
4858 + }
4859 + else
4860 + set_free_with_pinuse(p, psize, next);
4861 +
4862 + if (is_small(psize)) {
4863 + insert_small_chunk(fm, p, psize);
4864 + check_free_chunk(fm, p);
4865 + }
4866 + else {
4867 + tchunkptr tp = (tchunkptr)p;
4868 + insert_large_chunk(fm, tp, psize);
4869 + check_free_chunk(fm, p);
4870 + if (--fm->release_checks == 0)
4871 + release_unused_segments(fm);
4872 + }
4873 + goto postaction;
4874 + }
4875 + }
4876 + erroraction:
4877 + USAGE_ERROR_ACTION(fm, p);
4878 + postaction:
4879 + POSTACTION(fm);
4880 + }
4881 + }
4882 +#if !FOOTERS
4883 +#undef fm
4884 +#endif /* FOOTERS */
4885 +}
4886 +
4887 +void* dlcalloc(size_t n_elements, size_t elem_size) {
4888 + void* mem;
4889 + size_t req = 0;
4890 + if (n_elements != 0) {
4891 + req = n_elements * elem_size;
4892 + if (((n_elements | elem_size) & ~(size_t)0xffff) &&
4893 + (req / n_elements != elem_size))
4894 + req = MAX_SIZE_T; /* force downstream failure on overflow */
4895 + }
4896 + mem = dlmalloc(req);
4897 + if (mem != 0 && calloc_must_clear(mem2chunk(mem)))
4898 + memset(mem, 0, req);
4899 + return mem;
4900 +}
4901 +
4902 +void* dlrealloc(void* oldmem, size_t bytes) {
4903 + if (oldmem == 0)
4904 + return dlmalloc(bytes);
4905 +#ifdef REALLOC_ZERO_BYTES_FREES
4906 + if (bytes == 0) {
4907 + dlfree(oldmem);
4908 + return 0;
4909 + }
4910 +#endif /* REALLOC_ZERO_BYTES_FREES */
4911 + else {
4912 +#if ! FOOTERS
4913 + mstate m = gm;
4914 +#else /* FOOTERS */
4915 + mstate m = get_mstate_for(mem2chunk(oldmem));
4916 + if (!ok_magic(m)) {
4917 + USAGE_ERROR_ACTION(m, oldmem);
4918 + return 0;
4919 + }
4920 +#endif /* FOOTERS */
4921 + return internal_realloc(m, oldmem, bytes);
4922 + }
4923 +}
4924 +
4925 +void* dlmemalign(size_t alignment, size_t bytes) {
4926 + return internal_memalign(gm, alignment, bytes);
4927 +}
4928 +
4929 +void** dlindependent_calloc(size_t n_elements, size_t elem_size,
4930 + void* chunks[]) {
4931 + size_t sz = elem_size; /* serves as 1-element array */
4932 + return ialloc(gm, n_elements, &sz, 3, chunks);
4933 +}
4934 +
4935 +void** dlindependent_comalloc(size_t n_elements, size_t sizes[],
4936 + void* chunks[]) {
4937 + return ialloc(gm, n_elements, sizes, 0, chunks);
4938 +}
4939 +
4940 +void* dlvalloc(size_t bytes) {
4941 + size_t pagesz;
4942 + ensure_initialization();
4943 + pagesz = mparams.page_size;
4944 + return dlmemalign(pagesz, bytes);
4945 +}
4946 +
4947 +void* dlpvalloc(size_t bytes) {
4948 + size_t pagesz;
4949 + ensure_initialization();
4950 + pagesz = mparams.page_size;
4951 + return dlmemalign(pagesz, (bytes + pagesz - SIZE_T_ONE) & ~(pagesz - SIZE_T_ONE));
4952 +}
4953 +
4954 +int dlmalloc_trim(size_t pad) {
4955 + ensure_initialization();
4956 + int result = 0;
4957 + if (!PREACTION(gm)) {
4958 + result = sys_trim(gm, pad);
4959 + POSTACTION(gm);
4960 + }
4961 + return result;
4962 +}
4963 +
4964 +size_t dlmalloc_footprint(void) {
4965 + return gm->footprint;
4966 +}
4967 +
4968 +size_t dlmalloc_max_footprint(void) {
4969 + return gm->max_footprint;
4970 +}
4971 +
4972 +#if !NO_MALLINFO
4973 +struct mallinfo dlmallinfo(void) {
4974 + return internal_mallinfo(gm);
4975 +}
4976 +#endif /* NO_MALLINFO */
4977 +
4978 +void dlmalloc_stats() {
4979 + internal_malloc_stats(gm);
4980 +}
4981 +
4982 +int dlmallopt(int param_number, int value) {
4983 + return change_mparam(param_number, value);
4984 +}
4985 +
4986 +#endif /* !ONLY_MSPACES */
4987 +
4988 +size_t dlmalloc_usable_size(void* mem) {
4989 + if (mem != 0) {
4990 + mchunkptr p = mem2chunk(mem);
4991 + if (cinuse(p))
4992 + return chunksize(p) - overhead_for(p);
4993 + }
4994 + return 0;
4995 +}
4996 +
4997 +/* ----------------------------- user mspaces ---------------------------- */
4998 +
4999 +#if MSPACES

This file is too large to show in full.

compat/nedmalloc/nedmalloc.c new
+966
@@ -0,0 +1,966 @@
1 +/* Alternative malloc implementation for multiple threads without
2 +lock contention based on dlmalloc. (C) 2005-2006 Niall Douglas
3 +
4 +Boost Software License - Version 1.0 - August 17th, 2003
5 +
6 +Permission is hereby granted, free of charge, to any person or organization
7 +obtaining a copy of the software and accompanying documentation covered by
8 +this license (the "Software") to use, reproduce, display, distribute,
9 +execute, and transmit the Software, and to prepare derivative works of the
10 +Software, and to permit third-parties to whom the Software is furnished to
11 +do so, all subject to the following:
12 +
13 +The copyright notices in the Software and this entire statement, including
14 +the above license grant, this restriction and the following disclaimer,
15 +must be included in all copies of the Software, in whole or in part, and
16 +all derivative works of the Software, unless such copies or derivative
17 +works are solely in the form of machine-executable object code generated by
18 +a source language processor.
19 +
20 +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
21 +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
22 +FITNESS FOR A PARTICULAR PURPOSE, TITLE AND NON-INFRINGEMENT. IN NO EVENT
23 +SHALL THE COPYRIGHT HOLDERS OR ANYONE DISTRIBUTING THE SOFTWARE BE LIABLE
24 +FOR ANY DAMAGES OR OTHER LIABILITY, WHETHER IN CONTRACT, TORT OR OTHERWISE,
25 +ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
26 +DEALINGS IN THE SOFTWARE.
27 +*/
28 +
29 +#ifdef _MSC_VER
30 +/* Enable full aliasing on MSVC */
31 +/*#pragma optimize("a", on)*/
32 +#endif
33 +
34 +/*#define FULLSANITYCHECKS*/
35 +
36 +#include "nedmalloc.h"
37 +#if defined(WIN32) && !defined(__MINGW32__)
38 + #include <malloc.h>
39 +#endif
40 +#define MSPACES 1
41 +#define ONLY_MSPACES 1
42 +#ifndef USE_LOCKS
43 + #define USE_LOCKS 1
44 +#endif
45 +#define FOOTERS 1 /* Need to enable footers so frees lock the right mspace */
46 +#undef DEBUG /* dlmalloc wants DEBUG either 0 or 1 */
47 +#ifdef _DEBUG
48 + #define DEBUG 1
49 +#else
50 + #define DEBUG 0
51 +#endif
52 +#ifdef NDEBUG /* Disable assert checking on release builds */
53 + #undef DEBUG
54 +#endif
55 +/* The default of 64Kb means we spend too much time kernel-side */
56 +#ifndef DEFAULT_GRANULARITY
57 +#define DEFAULT_GRANULARITY (1*1024*1024)
58 +#endif
59 +/*#define USE_SPIN_LOCKS 0*/
60 +
61 +
62 +/*#define FORCEINLINE*/
63 +#include "malloc.c.h"
64 +#ifdef NDEBUG /* Disable assert checking on release builds */
65 + #undef DEBUG
66 +#endif
67 +
68 +/* The maximum concurrent threads in a pool possible */
69 +#ifndef MAXTHREADSINPOOL
70 +#define MAXTHREADSINPOOL 16
71 +#endif
72 +/* The maximum number of threadcaches which can be allocated */
73 +#ifndef THREADCACHEMAXCACHES
74 +#define THREADCACHEMAXCACHES 256
75 +#endif
76 +/* The maximum size to be allocated from the thread cache */
77 +#ifndef THREADCACHEMAX
78 +#define THREADCACHEMAX 8192
79 +#endif
80 +#if 0
81 +/* The number of cache entries for finer grained bins. This is (topbitpos(THREADCACHEMAX)-4)*2 */
82 +#define THREADCACHEMAXBINS ((13-4)*2)
83 +#else
84 +/* The number of cache entries. This is (topbitpos(THREADCACHEMAX)-4) */
85 +#define THREADCACHEMAXBINS (13-4)
86 +#endif
87 +/* Point at which the free space in a thread cache is garbage collected */
88 +#ifndef THREADCACHEMAXFREESPACE
89 +#define THREADCACHEMAXFREESPACE (512*1024)
90 +#endif
91 +
92 +
93 +#ifdef WIN32
94 + #define TLSVAR DWORD
95 + #define TLSALLOC(k) (*(k)=TlsAlloc(), TLS_OUT_OF_INDEXES==*(k))
96 + #define TLSFREE(k) (!TlsFree(k))
97 + #define TLSGET(k) TlsGetValue(k)
98 + #define TLSSET(k, a) (!TlsSetValue(k, a))
99 + #ifdef DEBUG
100 +static LPVOID ChkedTlsGetValue(DWORD idx)
101 +{
102 + LPVOID ret=TlsGetValue(idx);
103 + assert(S_OK==GetLastError());
104 + return ret;
105 +}
106 + #undef TLSGET
107 + #define TLSGET(k) ChkedTlsGetValue(k)
108 + #endif
109 +#else
110 + #define TLSVAR pthread_key_t
111 + #define TLSALLOC(k) pthread_key_create(k, 0)
112 + #define TLSFREE(k) pthread_key_delete(k)
113 + #define TLSGET(k) pthread_getspecific(k)
114 + #define TLSSET(k, a) pthread_setspecific(k, a)
115 +#endif
116 +
117 +#if 0
118 +/* Only enable if testing with valgrind. Causes misoperation */
119 +#define mspace_malloc(p, s) malloc(s)
120 +#define mspace_realloc(p, m, s) realloc(m, s)
121 +#define mspace_calloc(p, n, s) calloc(n, s)
122 +#define mspace_free(p, m) free(m)
123 +#endif
124 +
125 +
126 +#if defined(__cplusplus)
127 +#if !defined(NO_NED_NAMESPACE)
128 +namespace nedalloc {
129 +#else
130 +extern "C" {
131 +#endif
132 +#endif
133 +
134 +size_t nedblksize(void *mem) THROWSPEC
135 +{
136 +#if 0
137 + /* Only enable if testing with valgrind. Causes misoperation */
138 + return THREADCACHEMAX;
139 +#else
140 + if(mem)
141 + {
142 + mchunkptr p=mem2chunk(mem);
143 + assert(cinuse(p)); /* If this fails, someone tried to free a block twice */
144 + if(cinuse(p))
145 + return chunksize(p)-overhead_for(p);
146 + }
147 + return 0;
148 +#endif
149 +}
150 +
151 +void nedsetvalue(void *v) THROWSPEC { nedpsetvalue(0, v); }
152 +void * nedmalloc(size_t size) THROWSPEC { return nedpmalloc(0, size); }
153 +void * nedcalloc(size_t no, size_t size) THROWSPEC { return nedpcalloc(0, no, size); }
154 +void * nedrealloc(void *mem, size_t size) THROWSPEC { return nedprealloc(0, mem, size); }
155 +void nedfree(void *mem) THROWSPEC { nedpfree(0, mem); }
156 +void * nedmemalign(size_t alignment, size_t bytes) THROWSPEC { return nedpmemalign(0, alignment, bytes); }
157 +#if !NO_MALLINFO
158 +struct mallinfo nedmallinfo(void) THROWSPEC { return nedpmallinfo(0); }
159 +#endif
160 +int nedmallopt(int parno, int value) THROWSPEC { return nedpmallopt(0, parno, value); }
161 +int nedmalloc_trim(size_t pad) THROWSPEC { return nedpmalloc_trim(0, pad); }
162 +void nedmalloc_stats() THROWSPEC { nedpmalloc_stats(0); }
163 +size_t nedmalloc_footprint() THROWSPEC { return nedpmalloc_footprint(0); }
164 +void **nedindependent_calloc(size_t elemsno, size_t elemsize, void **chunks) THROWSPEC { return nedpindependent_calloc(0, elemsno, elemsize, chunks); }
165 +void **nedindependent_comalloc(size_t elems, size_t *sizes, void **chunks) THROWSPEC { return nedpindependent_comalloc(0, elems, sizes, chunks); }
166 +
167 +struct threadcacheblk_t;
168 +typedef struct threadcacheblk_t threadcacheblk;
169 +struct threadcacheblk_t
170 +{ /* Keep less than 16 bytes on 32 bit systems and 32 bytes on 64 bit systems */
171 +#ifdef FULLSANITYCHECKS
172 + unsigned int magic;
173 +#endif
174 + unsigned int lastUsed, size;
175 + threadcacheblk *next, *prev;
176 +};
177 +typedef struct threadcache_t
178 +{
179 +#ifdef FULLSANITYCHECKS
180 + unsigned int magic1;
181 +#endif
182 + int mymspace; /* Last mspace entry this thread used */
183 + long threadid;
184 + unsigned int mallocs, frees, successes;
185 + size_t freeInCache; /* How much free space is stored in this cache */
186 + threadcacheblk *bins[(THREADCACHEMAXBINS+1)*2];
187 +#ifdef FULLSANITYCHECKS
188 + unsigned int magic2;
189 +#endif
190 +} threadcache;
191 +struct nedpool_t
192 +{
193 + MLOCK_T mutex;
194 + void *uservalue;
195 + int threads; /* Max entries in m to use */
196 + threadcache *caches[THREADCACHEMAXCACHES];
197 + TLSVAR mycache; /* Thread cache for this thread. 0 for unset, negative for use mspace-1 directly, otherwise is cache-1 */
198 + mstate m[MAXTHREADSINPOOL+1]; /* mspace entries for this pool */
199 +};
200 +static nedpool syspool;
201 +
202 +static FORCEINLINE unsigned int size2binidx(size_t _size) THROWSPEC
203 +{ /* 8=1000 16=10000 20=10100 24=11000 32=100000 48=110000 4096=1000000000000 */
204 + unsigned int topbit, size=(unsigned int)(_size>>4);
205 + /* 16=1 20=1 24=1 32=10 48=11 64=100 96=110 128=1000 4096=100000000 */
206 +
207 +#if defined(__GNUC__)
208 + topbit = sizeof(size)*__CHAR_BIT__ - 1 - __builtin_clz(size);
209 +#elif defined(_MSC_VER) && _MSC_VER>=1300
210 + {
211 + unsigned long bsrTopBit;
212 +
213 + _BitScanReverse(&bsrTopBit, size);
214 +
215 + topbit = bsrTopBit;
216 + }
217 +#else
218 +#if 0
219 + union {
220 + unsigned asInt[2];
221 + double asDouble;
222 + };
223 + int n;
224 +
225 + asDouble = (double)size + 0.5;
226 + topbit = (asInt[!FOX_BIGENDIAN] >> 20) - 1023;
227 +#else
228 + {
229 + unsigned int x=size;
230 + x = x | (x >> 1);
231 + x = x | (x >> 2);
232 + x = x | (x >> 4);
233 + x = x | (x >> 8);
234 + x = x | (x >>16);
235 + x = ~x;
236 + x = x - ((x >> 1) & 0x55555555);
237 + x = (x & 0x33333333) + ((x >> 2) & 0x33333333);
238 + x = (x + (x >> 4)) & 0x0F0F0F0F;
239 + x = x + (x << 8);
240 + x = x + (x << 16);
241 + topbit=31 - (x >> 24);
242 + }
243 +#endif
244 +#endif
245 + return topbit;
246 +}
247 +
248 +
249 +#ifdef FULLSANITYCHECKS
250 +static void tcsanitycheck(threadcacheblk **ptr) THROWSPEC
251 +{
252 + assert((ptr[0] && ptr[1]) || (!ptr[0] && !ptr[1]));
253 + if(ptr[0] && ptr[1])
254 + {
255 + assert(nedblksize(ptr[0])>=sizeof(threadcacheblk));
256 + assert(nedblksize(ptr[1])>=sizeof(threadcacheblk));
257 + assert(*(unsigned int *) "NEDN"==ptr[0]->magic);
258 + assert(*(unsigned int *) "NEDN"==ptr[1]->magic);
259 + assert(!ptr[0]->prev);
260 + assert(!ptr[1]->next);
261 + if(ptr[0]==ptr[1])
262 + {
263 + assert(!ptr[0]->next);
264 + assert(!ptr[1]->prev);
265 + }
266 + }
267 +}
268 +static void tcfullsanitycheck(threadcache *tc) THROWSPEC
269 +{
270 + threadcacheblk **tcbptr=tc->bins;
271 + int n;
272 + for(n=0; n<=THREADCACHEMAXBINS; n++, tcbptr+=2)
273 + {
274 + threadcacheblk *b, *ob=0;
275 + tcsanitycheck(tcbptr);
276 + for(b=tcbptr[0]; b; ob=b, b=b->next)
277 + {
278 + assert(*(unsigned int *) "NEDN"==b->magic);
279 + assert(!ob || ob->next==b);
280 + assert(!ob || b->prev==ob);
281 + }
282 + }
283 +}
284 +#endif
285 +
286 +static NOINLINE void RemoveCacheEntries(nedpool *p, threadcache *tc, unsigned int age) THROWSPEC
287 +{
288 +#ifdef FULLSANITYCHECKS
289 + tcfullsanitycheck(tc);
290 +#endif
291 + if(tc->freeInCache)
292 + {
293 + threadcacheblk **tcbptr=tc->bins;
294 + int n;
295 + for(n=0; n<=THREADCACHEMAXBINS; n++, tcbptr+=2)
296 + {
297 + threadcacheblk **tcb=tcbptr+1; /* come from oldest end of list */
298 + /*tcsanitycheck(tcbptr);*/
299 + for(; *tcb && tc->frees-(*tcb)->lastUsed>=age; )
300 + {
301 + threadcacheblk *f=*tcb;
302 + size_t blksize=f->size; /*nedblksize(f);*/
303 + assert(blksize<=nedblksize(f));
304 + assert(blksize);
305 +#ifdef FULLSANITYCHECKS
306 + assert(*(unsigned int *) "NEDN"==(*tcb)->magic);
307 +#endif
308 + *tcb=(*tcb)->prev;
309 + if(*tcb)
310 + (*tcb)->next=0;
311 + else
312 + *tcbptr=0;
313 + tc->freeInCache-=blksize;
314 + assert((long) tc->freeInCache>=0);
315 + mspace_free(0, f);
316 + /*tcsanitycheck(tcbptr);*/
317 + }
318 + }
319 + }
320 +#ifdef FULLSANITYCHECKS
321 + tcfullsanitycheck(tc);
322 +#endif
323 +}
324 +static void DestroyCaches(nedpool *p) THROWSPEC
325 +{
326 + if(p->caches)
327 + {
328 + threadcache *tc;
329 + int n;
330 + for(n=0; n<THREADCACHEMAXCACHES; n++)
331 + {
332 + if((tc=p->caches[n]))
333 + {
334 + tc->frees++;
335 + RemoveCacheEntries(p, tc, 0);
336 + assert(!tc->freeInCache);
337 + tc->mymspace=-1;
338 + tc->threadid=0;
339 + mspace_free(0, tc);
340 + p->caches[n]=0;
341 + }
342 + }
343 + }
344 +}
345 +
346 +static NOINLINE threadcache *AllocCache(nedpool *p) THROWSPEC
347 +{
348 + threadcache *tc=0;
349 + int n, end;
350 + ACQUIRE_LOCK(&p->mutex);
351 + for(n=0; n<THREADCACHEMAXCACHES && p->caches[n]; n++);
352 + if(THREADCACHEMAXCACHES==n)
353 + { /* List exhausted, so disable for this thread */
354 + RELEASE_LOCK(&p->mutex);
355 + return 0;
356 + }
357 + tc=p->caches[n]=(threadcache *) mspace_calloc(p->m[0], 1, sizeof(threadcache));
358 + if(!tc)
359 + {
360 + RELEASE_LOCK(&p->mutex);
361 + return 0;
362 + }
363 +#ifdef FULLSANITYCHECKS
364 + tc->magic1=*(unsigned int *)"NEDMALC1";
365 + tc->magic2=*(unsigned int *)"NEDMALC2";
366 +#endif
367 + tc->threadid=(long)(size_t)CURRENT_THREAD;
368 + for(end=0; p->m[end]; end++);
369 + tc->mymspace=tc->threadid % end;
370 + RELEASE_LOCK(&p->mutex);
371 + if(TLSSET(p->mycache, (void *)(size_t)(n+1))) abort();
372 + return tc;
373 +}
374 +
375 +static void *threadcache_malloc(nedpool *p, threadcache *tc, size_t *size) THROWSPEC
376 +{
377 + void *ret=0;
378 + unsigned int bestsize;
379 + unsigned int idx=size2binidx(*size);
380 + size_t blksize=0;
381 + threadcacheblk *blk, **binsptr;
382 +#ifdef FULLSANITYCHECKS
383 + tcfullsanitycheck(tc);
384 +#endif
385 + /* Calculate best fit bin size */
386 + bestsize=1<<(idx+4);
387 +#if 0
388 + /* Finer grained bin fit */
389 + idx<<=1;
390 + if(*size>bestsize)
391 + {
392 + idx++;
393 + bestsize+=bestsize>>1;
394 + }
395 + if(*size>bestsize)
396 + {
397 + idx++;
398 + bestsize=1<<(4+(idx>>1));
399 + }
400 +#else
401 + if(*size>bestsize)
402 + {
403 + idx++;
404 + bestsize<<=1;
405 + }
406 +#endif
407 + assert(bestsize>=*size);
408 + if(*size<bestsize) *size=bestsize;
409 + assert(*size<=THREADCACHEMAX);
410 + assert(idx<=THREADCACHEMAXBINS);
411 + binsptr=&tc->bins[idx*2];
412 + /* Try to match close, but move up a bin if necessary */
413 + blk=*binsptr;
414 + if(!blk || blk->size<*size)
415 + { /* Bump it up a bin */
416 + if(idx<THREADCACHEMAXBINS)
417 + {
418 + idx++;
419 + binsptr+=2;
420 + blk=*binsptr;
421 + }
422 + }
423 + if(blk)
424 + {
425 + blksize=blk->size; /*nedblksize(blk);*/
426 + assert(nedblksize(blk)>=blksize);
427 + assert(blksize>=*size);
428 + if(blk->next)
429 + blk->next->prev=0;
430 + *binsptr=blk->next;
431 + if(!*binsptr)
432 + binsptr[1]=0;
433 +#ifdef FULLSANITYCHECKS
434 + blk->magic=0;
435 +#endif
436 + assert(binsptr[0]!=blk && binsptr[1]!=blk);
437 + assert(nedblksize(blk)>=sizeof(threadcacheblk) && nedblksize(blk)<=THREADCACHEMAX+CHUNK_OVERHEAD);
438 + /*printf("malloc: %p, %p, %p, %lu\n", p, tc, blk, (long) size);*/
439 + ret=(void *) blk;
440 + }
441 + ++tc->mallocs;
442 + if(ret)
443 + {
444 + assert(blksize>=*size);
445 + ++tc->successes;
446 + tc->freeInCache-=blksize;
447 + assert((long) tc->freeInCache>=0);
448 + }
449 +#if defined(DEBUG) && 0
450 + if(!(tc->mallocs & 0xfff))
451 + {
452 + printf("*** threadcache=%u, mallocs=%u (%f), free=%u (%f), freeInCache=%u\n", (unsigned int) tc->threadid, tc->mallocs,
453 + (float) tc->successes/tc->mallocs, tc->frees, (float) tc->successes/tc->frees, (unsigned int) tc->freeInCache);
454 + }
455 +#endif
456 +#ifdef FULLSANITYCHECKS
457 + tcfullsanitycheck(tc);
458 +#endif
459 + return ret;
460 +}
461 +static NOINLINE void ReleaseFreeInCache(nedpool *p, threadcache *tc, int mymspace) THROWSPEC
462 +{
463 + unsigned int age=THREADCACHEMAXFREESPACE/8192;
464 + /*ACQUIRE_LOCK(&p->m[mymspace]->mutex);*/
465 + while(age && tc->freeInCache>=THREADCACHEMAXFREESPACE)
466 + {
467 + RemoveCacheEntries(p, tc, age);
468 + /*printf("*** Removing cache entries older than %u (%u)\n", age, (unsigned int) tc->freeInCache);*/
469 + age>>=1;
470 + }
471 + /*RELEASE_LOCK(&p->m[mymspace]->mutex);*/
472 +}
473 +static void threadcache_free(nedpool *p, threadcache *tc, int mymspace, void *mem, size_t size) THROWSPEC
474 +{
475 + unsigned int bestsize;
476 + unsigned int idx=size2binidx(size);
477 + threadcacheblk **binsptr, *tck=(threadcacheblk *) mem;
478 + assert(size>=sizeof(threadcacheblk) && size<=THREADCACHEMAX+CHUNK_OVERHEAD);
479 +#ifdef DEBUG
480 + { /* Make sure this is a valid memory block */
481 + mchunkptr p = mem2chunk(mem);
482 + mstate fm = get_mstate_for(p);
483 + if (!ok_magic(fm)) {
484 + USAGE_ERROR_ACTION(fm, p);
485 + return;
486 + }
487 + }
488 +#endif
489 +#ifdef FULLSANITYCHECKS
490 + tcfullsanitycheck(tc);
491 +#endif
492 + /* Calculate best fit bin size */
493 + bestsize=1<<(idx+4);
494 +#if 0
495 + /* Finer grained bin fit */
496 + idx<<=1;
497 + if(size>bestsize)
498 + {
499 + unsigned int biggerbestsize=bestsize+bestsize<<1;
500 + if(size>=biggerbestsize)
501 + {
502 + idx++;
503 + bestsize=biggerbestsize;
504 + }
505 + }
506 +#endif
507 + if(bestsize!=size) /* dlmalloc can round up, so we round down to preserve indexing */
508 + size=bestsize;
509 + binsptr=&tc->bins[idx*2];
510 + assert(idx<=THREADCACHEMAXBINS);
511 + if(tck==*binsptr)
512 + {
513 + fprintf(stderr, "Attempt to free already freed memory block %p - aborting!\n", tck);
514 + abort();
515 + }
516 +#ifdef FULLSANITYCHECKS
517 + tck->magic=*(unsigned int *) "NEDN";
518 +#endif
519 + tck->lastUsed=++tc->frees;
520 + tck->size=(unsigned int) size;
521 + tck->next=*binsptr;
522 + tck->prev=0;
523 + if(tck->next)
524 + tck->next->prev=tck;
525 + else
526 + binsptr[1]=tck;
527 + assert(!*binsptr || (*binsptr)->size==tck->size);
528 + *binsptr=tck;
529 + assert(tck==tc->bins[idx*2]);
530 + assert(tc->bins[idx*2+1]==tck || binsptr[0]->next->prev==tck);
531 + /*printf("free: %p, %p, %p, %lu\n", p, tc, mem, (long) size);*/
532 + tc->freeInCache+=size;
533 +#ifdef FULLSANITYCHECKS
534 + tcfullsanitycheck(tc);
535 +#endif
536 +#if 1
537 + if(tc->freeInCache>=THREADCACHEMAXFREESPACE)
538 + ReleaseFreeInCache(p, tc, mymspace);
539 +#endif
540 +}
541 +
542 +
543 +
544 +
545 +static NOINLINE int InitPool(nedpool *p, size_t capacity, int threads) THROWSPEC
546 +{ /* threads is -1 for system pool */
547 + ensure_initialization();
548 + ACQUIRE_MALLOC_GLOBAL_LOCK();
549 + if(p->threads) goto done;
550 + if(INITIAL_LOCK(&p->mutex)) goto err;
551 + if(TLSALLOC(&p->mycache)) goto err;
552 + if(!(p->m[0]=(mstate) create_mspace(capacity, 1))) goto err;
553 + p->m[0]->extp=p;
554 + p->threads=(threads<1 || threads>MAXTHREADSINPOOL) ? MAXTHREADSINPOOL : threads;
555 +done:
556 + RELEASE_MALLOC_GLOBAL_LOCK();
557 + return 1;
558 +err:
559 + if(threads<0)
560 + abort(); /* If you can't allocate for system pool, we're screwed */
561 + DestroyCaches(p);
562 + if(p->m[0])
563 + {
564 + destroy_mspace(p->m[0]);
565 + p->m[0]=0;
566 + }
567 + if(p->mycache)
568 + {
569 + if(TLSFREE(p->mycache)) abort();
570 + p->mycache=0;
571 + }
572 + RELEASE_MALLOC_GLOBAL_LOCK();
573 + return 0;
574 +}
575 +static NOINLINE mstate FindMSpace(nedpool *p, threadcache *tc, int *lastUsed, size_t size) THROWSPEC
576 +{ /* Gets called when thread's last used mspace is in use. The strategy
577 + is to run through the list of all available mspaces looking for an
578 + unlocked one and if we fail, we create a new one so long as we don't
579 + exceed p->threads */
580 + int n, end;
581 + for(n=end=*lastUsed+1; p->m[n]; end=++n)
582 + {
583 + if(TRY_LOCK(&p->m[n]->mutex)) goto found;
584 + }
585 + for(n=0; n<*lastUsed && p->m[n]; n++)
586 + {
587 + if(TRY_LOCK(&p->m[n]->mutex)) goto found;
588 + }
589 + if(end<p->threads)
590 + {
591 + mstate temp;
592 + if(!(temp=(mstate) create_mspace(size, 1)))
593 + goto badexit;
594 + /* Now we're ready to modify the lists, we lock */
595 + ACQUIRE_LOCK(&p->mutex);
596 + while(p->m[end] && end<p->threads)
597 + end++;
598 + if(end>=p->threads)
599 + { /* Drat, must destroy it now */
600 + RELEASE_LOCK(&p->mutex);
601 + destroy_mspace((mspace) temp);
602 + goto badexit;
603 + }
604 + /* We really want to make sure this goes into memory now but we
605 + have to be careful of breaking aliasing rules, so write it twice */
606 + *((volatile struct malloc_state **) &p->m[end])=p->m[end]=temp;
607 + ACQUIRE_LOCK(&p->m[end]->mutex);
608 + /*printf("Created mspace idx %d\n", end);*/
609 + RELEASE_LOCK(&p->mutex);
610 + n=end;
611 + goto found;
612 + }
613 + /* Let it lock on the last one it used */
614 +badexit:
615 + ACQUIRE_LOCK(&p->m[*lastUsed]->mutex);
616 + return p->m[*lastUsed];
617 +found:
618 + *lastUsed=n;
619 + if(tc)
620 + tc->mymspace=n;
621 + else
622 + {
623 + if(TLSSET(p->mycache, (void *)(size_t)(-(n+1)))) abort();
624 + }
625 + return p->m[n];
626 +}
627 +
628 +nedpool *nedcreatepool(size_t capacity, int threads) THROWSPEC
629 +{
630 + nedpool *ret;
631 + if(!(ret=(nedpool *) nedpcalloc(0, 1, sizeof(nedpool)))) return 0;
632 + if(!InitPool(ret, capacity, threads))
633 + {
634 + nedpfree(0, ret);
635 + return 0;
636 + }
637 + return ret;
638 +}
639 +void neddestroypool(nedpool *p) THROWSPEC
640 +{
641 + int n;
642 + ACQUIRE_LOCK(&p->mutex);
643 + DestroyCaches(p);
644 + for(n=0; p->m[n]; n++)
645 + {
646 + destroy_mspace(p->m[n]);
647 + p->m[n]=0;
648 + }
649 + RELEASE_LOCK(&p->mutex);
650 + if(TLSFREE(p->mycache)) abort();
651 + nedpfree(0, p);
652 +}
653 +
654 +void nedpsetvalue(nedpool *p, void *v) THROWSPEC
655 +{
656 + if(!p) { p=&syspool; if(!syspool.threads) InitPool(&syspool, 0, -1); }
657 + p->uservalue=v;
658 +}
659 +void *nedgetvalue(nedpool **p, void *mem) THROWSPEC
660 +{
661 + nedpool *np=0;
662 + mchunkptr mcp=mem2chunk(mem);
663 + mstate fm;
664 + if(!(is_aligned(chunk2mem(mcp))) && mcp->head != FENCEPOST_HEAD) return 0;
665 + if(!cinuse(mcp)) return 0;
666 + if(!next_pinuse(mcp)) return 0;
667 + if(!is_mmapped(mcp) && !pinuse(mcp))
668 + {
669 + if(next_chunk(prev_chunk(mcp))!=mcp) return 0;
670 + }
671 + fm=get_mstate_for(mcp);
672 + if(!ok_magic(fm)) return 0;
673 + if(!ok_address(fm, mcp)) return 0;
674 + if(!fm->extp) return 0;
675 + np=(nedpool *) fm->extp;
676 + if(p) *p=np;
677 + return np->uservalue;
678 +}
679 +
680 +void neddisablethreadcache(nedpool *p) THROWSPEC
681 +{
682 + int mycache;
683 + if(!p)
684 + {
685 + p=&syspool;
686 + if(!syspool.threads) InitPool(&syspool, 0, -1);
687 + }
688 + mycache=(int)(size_t) TLSGET(p->mycache);
689 + if(!mycache)
690 + { /* Set to mspace 0 */
691 + if(TLSSET(p->mycache, (void *)-1)) abort();
692 + }
693 + else if(mycache>0)
694 + { /* Set to last used mspace */
695 + threadcache *tc=p->caches[mycache-1];
696 +#if defined(DEBUG)
697 + printf("Threadcache utilisation: %lf%% in cache with %lf%% lost to other threads\n",
698 + 100.0*tc->successes/tc->mallocs, 100.0*((double) tc->mallocs-tc->frees)/tc->mallocs);
699 +#endif
700 + if(TLSSET(p->mycache, (void *)(size_t)(-tc->mymspace))) abort();
701 + tc->frees++;
702 + RemoveCacheEntries(p, tc, 0);
703 + assert(!tc->freeInCache);
704 + tc->mymspace=-1;
705 + tc->threadid=0;
706 + mspace_free(0, p->caches[mycache-1]);
707 + p->caches[mycache-1]=0;
708 + }
709 +}
710 +
711 +#define GETMSPACE(m,p,tc,ms,s,action) \
712 + do \
713 + { \
714 + mstate m = GetMSpace((p),(tc),(ms),(s)); \
715 + action; \
716 + RELEASE_LOCK(&m->mutex); \
717 + } while (0)
718 +
719 +static FORCEINLINE mstate GetMSpace(nedpool *p, threadcache *tc, int mymspace, size_t size) THROWSPEC
720 +{ /* Returns a locked and ready for use mspace */
721 + mstate m=p->m[mymspace];
722 + assert(m);
723 + if(!TRY_LOCK(&p->m[mymspace]->mutex)) m=FindMSpace(p, tc, &mymspace, size);\
724 + /*assert(IS_LOCKED(&p->m[mymspace]->mutex));*/
725 + return m;
726 +}
727 +static FORCEINLINE void GetThreadCache(nedpool **p, threadcache **tc, int *mymspace, size_t *size) THROWSPEC
728 +{
729 + int mycache;
730 + if(size && *size<sizeof(threadcacheblk)) *size=sizeof(threadcacheblk);
731 + if(!*p)
732 + {
733 + *p=&syspool;
734 + if(!syspool.threads) InitPool(&syspool, 0, -1);
735 + }
736 + mycache=(int)(size_t) TLSGET((*p)->mycache);
737 + if(mycache>0)
738 + {
739 + *tc=(*p)->caches[mycache-1];
740 + *mymspace=(*tc)->mymspace;
741 + }
742 + else if(!mycache)
743 + {
744 + *tc=AllocCache(*p);
745 + if(!*tc)
746 + { /* Disable */
747 + if(TLSSET((*p)->mycache, (void *)-1)) abort();
748 + *mymspace=0;
749 + }
750 + else
751 + *mymspace=(*tc)->mymspace;
752 + }
753 + else
754 + {
755 + *tc=0;
756 + *mymspace=-mycache-1;
757 + }
758 + assert(*mymspace>=0);
759 + assert((long)(size_t)CURRENT_THREAD==(*tc)->threadid);
760 +#ifdef FULLSANITYCHECKS
761 + if(*tc)
762 + {
763 + if(*(unsigned int *)"NEDMALC1"!=(*tc)->magic1 || *(unsigned int *)"NEDMALC2"!=(*tc)->magic2)
764 + {
765 + abort();
766 + }
767 + }
768 +#endif
769 +}
770 +
771 +void * nedpmalloc(nedpool *p, size_t size) THROWSPEC
772 +{
773 + void *ret=0;
774 + threadcache *tc;
775 + int mymspace;
776 + GetThreadCache(&p, &tc, &mymspace, &size);
777 +#if THREADCACHEMAX
778 + if(tc && size<=THREADCACHEMAX)
779 + { /* Use the thread cache */
780 + ret=threadcache_malloc(p, tc, &size);
781 + }
782 +#endif
783 + if(!ret)
784 + { /* Use this thread's mspace */
785 + GETMSPACE(m, p, tc, mymspace, size,
786 + ret=mspace_malloc(m, size));
787 + }
788 + return ret;
789 +}
790 +void * nedpcalloc(nedpool *p, size_t no, size_t size) THROWSPEC
791 +{
792 + size_t rsize=size*no;
793 + void *ret=0;
794 + threadcache *tc;
795 + int mymspace;
796 + GetThreadCache(&p, &tc, &mymspace, &rsize);
797 +#if THREADCACHEMAX
798 + if(tc && rsize<=THREADCACHEMAX)
799 + { /* Use the thread cache */
800 + if((ret=threadcache_malloc(p, tc, &rsize)))
801 + memset(ret, 0, rsize);
802 + }
803 +#endif
804 + if(!ret)
805 + { /* Use this thread's mspace */
806 + GETMSPACE(m, p, tc, mymspace, rsize,
807 + ret=mspace_calloc(m, 1, rsize));
808 + }
809 + return ret;
810 +}
811 +void * nedprealloc(nedpool *p, void *mem, size_t size) THROWSPEC
812 +{
813 + void *ret=0;
814 + threadcache *tc;
815 + int mymspace;
816 + if(!mem) return nedpmalloc(p, size);
817 + GetThreadCache(&p, &tc, &mymspace, &size);
818 +#if THREADCACHEMAX
819 + if(tc && size && size<=THREADCACHEMAX)
820 + { /* Use the thread cache */
821 + size_t memsize=nedblksize(mem);
822 + assert(memsize);
823 + if((ret=threadcache_malloc(p, tc, &size)))
824 + {
825 + memcpy(ret, mem, memsize<size ? memsize : size);
826 + if(memsize<=THREADCACHEMAX)
827 + threadcache_free(p, tc, mymspace, mem, memsize);
828 + else
829 + mspace_free(0, mem);
830 + }
831 + }
832 +#endif
833 + if(!ret)
834 + { /* Reallocs always happen in the mspace they happened in, so skip
835 + locking the preferred mspace for this thread */
836 + ret=mspace_realloc(0, mem, size);
837 + }
838 + return ret;
839 +}
840 +void nedpfree(nedpool *p, void *mem) THROWSPEC
841 +{ /* Frees always happen in the mspace they happened in, so skip
842 + locking the preferred mspace for this thread */
843 + threadcache *tc;
844 + int mymspace;
845 + size_t memsize;
846 + assert(mem);
847 + GetThreadCache(&p, &tc, &mymspace, 0);
848 +#if THREADCACHEMAX
849 + memsize=nedblksize(mem);
850 + assert(memsize);
851 + if(mem && tc && memsize<=(THREADCACHEMAX+CHUNK_OVERHEAD))
852 + threadcache_free(p, tc, mymspace, mem, memsize);
853 + else
854 +#endif
855 + mspace_free(0, mem);
856 +}
857 +void * nedpmemalign(nedpool *p, size_t alignment, size_t bytes) THROWSPEC
858 +{
859 + void *ret;
860 + threadcache *tc;
861 + int mymspace;
862 + GetThreadCache(&p, &tc, &mymspace, &bytes);
863 + { /* Use this thread's mspace */
864 + GETMSPACE(m, p, tc, mymspace, bytes,
865 + ret=mspace_memalign(m, alignment, bytes));
866 + }
867 + return ret;
868 +}
869 +#if !NO_MALLINFO
870 +struct mallinfo nedpmallinfo(nedpool *p) THROWSPEC
871 +{
872 + int n;
873 + struct mallinfo ret={0};
874 + if(!p) { p=&syspool; if(!syspool.threads) InitPool(&syspool, 0, -1); }
875 + for(n=0; p->m[n]; n++)
876 + {
877 + struct mallinfo t=mspace_mallinfo(p->m[n]);
878 + ret.arena+=t.arena;
879 + ret.ordblks+=t.ordblks;
880 + ret.hblkhd+=t.hblkhd;
881 + ret.usmblks+=t.usmblks;
882 + ret.uordblks+=t.uordblks;
883 + ret.fordblks+=t.fordblks;
884 + ret.keepcost+=t.keepcost;
885 + }
886 + return ret;
887 +}
888 +#endif
889 +int nedpmallopt(nedpool *p, int parno, int value) THROWSPEC
890 +{
891 + return mspace_mallopt(parno, value);
892 +}
893 +int nedpmalloc_trim(nedpool *p, size_t pad) THROWSPEC
894 +{
895 + int n, ret=0;
896 + if(!p) { p=&syspool; if(!syspool.threads) InitPool(&syspool, 0, -1); }
897 + for(n=0; p->m[n]; n++)
898 + {
899 + ret+=mspace_trim(p->m[n], pad);
900 + }
901 + return ret;
902 +}
903 +void nedpmalloc_stats(nedpool *p) THROWSPEC
904 +{
905 + int n;
906 + if(!p) { p=&syspool; if(!syspool.threads) InitPool(&syspool, 0, -1); }
907 + for(n=0; p->m[n]; n++)
908 + {
909 + mspace_malloc_stats(p->m[n]);
910 + }
911 +}
912 +size_t nedpmalloc_footprint(nedpool *p) THROWSPEC
913 +{
914 + size_t ret=0;
915 + int n;
916 + if(!p) { p=&syspool; if(!syspool.threads) InitPool(&syspool, 0, -1); }
917 + for(n=0; p->m[n]; n++)
918 + {
919 + ret+=mspace_footprint(p->m[n]);
920 + }
921 + return ret;
922 +}
923 +void **nedpindependent_calloc(nedpool *p, size_t elemsno, size_t elemsize, void **chunks) THROWSPEC
924 +{
925 + void **ret;
926 + threadcache *tc;
927 + int mymspace;
928 + GetThreadCache(&p, &tc, &mymspace, &elemsize);
929 + GETMSPACE(m, p, tc, mymspace, elemsno*elemsize,
930 + ret=mspace_independent_calloc(m, elemsno, elemsize, chunks));
931 + return ret;
932 +}
933 +void **nedpindependent_comalloc(nedpool *p, size_t elems, size_t *sizes, void **chunks) THROWSPEC
934 +{
935 + void **ret;
936 + threadcache *tc;
937 + int mymspace;
938 + size_t i, *adjustedsizes=(size_t *) alloca(elems*sizeof(size_t));
939 + if(!adjustedsizes) return 0;
940 + for(i=0; i<elems; i++)
941 + adjustedsizes[i]=sizes[i]<sizeof(threadcacheblk) ? sizeof(threadcacheblk) : sizes[i];
942 + GetThreadCache(&p, &tc, &mymspace, 0);
943 + GETMSPACE(m, p, tc, mymspace, 0,
944 + ret=mspace_independent_comalloc(m, elems, adjustedsizes, chunks));
945 + return ret;
946 +}
947 +
948 +#ifdef OVERRIDE_STRDUP
949 +/*
950 + * This implementation is purely there to override the libc version, to
951 + * avoid a crash due to allocation and free on different 'heaps'.
952 + */
953 +char *strdup(const char *s1)
954 +{
955 + char *s2 = 0;
956 + if (s1) {
957 + s2 = malloc(strlen(s1) + 1);
958 + strcpy(s2, s1);
959 + }
960 + return s2;
961 +}
962 +#endif
963 +
964 +#if defined(__cplusplus)
965 +}
966 +#endif
compat/nedmalloc/nedmalloc.h new
+180
@@ -0,0 +1,180 @@
1 +/* nedalloc, an alternative malloc implementation for multiple threads without
2 +lock contention based on dlmalloc v2.8.3. (C) 2005 Niall Douglas
3 +
4 +Boost Software License - Version 1.0 - August 17th, 2003
5 +
6 +Permission is hereby granted, free of charge, to any person or organization
7 +obtaining a copy of the software and accompanying documentation covered by
8 +this license (the "Software") to use, reproduce, display, distribute,
9 +execute, and transmit the Software, and to prepare derivative works of the
10 +Software, and to permit third-parties to whom the Software is furnished to
11 +do so, all subject to the following:
12 +
13 +The copyright notices in the Software and this entire statement, including
14 +the above license grant, this restriction and the following disclaimer,
15 +must be included in all copies of the Software, in whole or in part, and
16 +all derivative works of the Software, unless such copies or derivative
17 +works are solely in the form of machine-executable object code generated by
18 +a source language processor.
19 +
20 +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
21 +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
22 +FITNESS FOR A PARTICULAR PURPOSE, TITLE AND NON-INFRINGEMENT. IN NO EVENT
23 +SHALL THE COPYRIGHT HOLDERS OR ANYONE DISTRIBUTING THE SOFTWARE BE LIABLE
24 +FOR ANY DAMAGES OR OTHER LIABILITY, WHETHER IN CONTRACT, TORT OR OTHERWISE,
25 +ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
26 +DEALINGS IN THE SOFTWARE.
27 +*/
28 +
29 +#ifndef NEDMALLOC_H
30 +#define NEDMALLOC_H
31 +
32 +
33 +/* See malloc.c.h for what each function does.
34 +
35 +REPLACE_SYSTEM_ALLOCATOR causes nedalloc's functions to be called malloc,
36 +free etc. instead of nedmalloc, nedfree etc. You may or may not want this.
37 +
38 +NO_NED_NAMESPACE prevents the functions from being defined in the nedalloc
39 +namespace when in C++ (uses the global namespace instead).
40 +
41 +EXTSPEC can be defined to be __declspec(dllexport) or
42 +__attribute__ ((visibility("default"))) or whatever you like. It defaults
43 +to extern.
44 +
45 +USE_LOCKS can be 2 if you want to define your own MLOCK_T, INITIAL_LOCK,
46 +ACQUIRE_LOCK, RELEASE_LOCK, TRY_LOCK, IS_LOCKED and NULL_LOCK_INITIALIZER.
47 +
48 +*/
49 +
50 +#include <stddef.h> /* for size_t */
51 +
52 +#ifndef EXTSPEC
53 + #define EXTSPEC extern
54 +#endif
55 +
56 +#if defined(_MSC_VER) && _MSC_VER>=1400
57 + #define MALLOCATTR __declspec(restrict)
58 +#endif
59 +#ifdef __GNUC__
60 + #define MALLOCATTR __attribute__ ((malloc))
61 +#endif
62 +#ifndef MALLOCATTR
63 + #define MALLOCATTR
64 +#endif
65 +
66 +#ifdef REPLACE_SYSTEM_ALLOCATOR
67 + #define nedmalloc malloc
68 + #define nedcalloc calloc
69 + #define nedrealloc realloc
70 + #define nedfree free
71 + #define nedmemalign memalign
72 + #define nedmallinfo mallinfo
73 + #define nedmallopt mallopt
74 + #define nedmalloc_trim malloc_trim
75 + #define nedmalloc_stats malloc_stats
76 + #define nedmalloc_footprint malloc_footprint
77 + #define nedindependent_calloc independent_calloc
78 + #define nedindependent_comalloc independent_comalloc
79 + #ifdef _MSC_VER
80 + #define nedblksize _msize
81 + #endif
82 +#endif
83 +
84 +#ifndef NO_MALLINFO
85 +#define NO_MALLINFO 0
86 +#endif
87 +
88 +#if !NO_MALLINFO
89 +struct mallinfo;
90 +#endif
91 +
92 +#if defined(__cplusplus)
93 + #if !defined(NO_NED_NAMESPACE)
94 +namespace nedalloc {
95 + #else
96 +extern "C" {
97 + #endif
98 + #define THROWSPEC throw()
99 +#else
100 + #define THROWSPEC
101 +#endif
102 +
103 +/* These are the global functions */
104 +
105 +/* Gets the usable size of an allocated block. Note this will always be bigger than what was
106 +asked for due to rounding etc.
107 +*/
108 +EXTSPEC size_t nedblksize(void *mem) THROWSPEC;
109 +
110 +EXTSPEC void nedsetvalue(void *v) THROWSPEC;
111 +
112 +EXTSPEC MALLOCATTR void * nedmalloc(size_t size) THROWSPEC;
113 +EXTSPEC MALLOCATTR void * nedcalloc(size_t no, size_t size) THROWSPEC;
114 +EXTSPEC MALLOCATTR void * nedrealloc(void *mem, size_t size) THROWSPEC;
115 +EXTSPEC void nedfree(void *mem) THROWSPEC;
116 +EXTSPEC MALLOCATTR void * nedmemalign(size_t alignment, size_t bytes) THROWSPEC;
117 +#if !NO_MALLINFO
118 +EXTSPEC struct mallinfo nedmallinfo(void) THROWSPEC;
119 +#endif
120 +EXTSPEC int nedmallopt(int parno, int value) THROWSPEC;
121 +EXTSPEC int nedmalloc_trim(size_t pad) THROWSPEC;
122 +EXTSPEC void nedmalloc_stats(void) THROWSPEC;
123 +EXTSPEC size_t nedmalloc_footprint(void) THROWSPEC;
124 +EXTSPEC MALLOCATTR void **nedindependent_calloc(size_t elemsno, size_t elemsize, void **chunks) THROWSPEC;
125 +EXTSPEC MALLOCATTR void **nedindependent_comalloc(size_t elems, size_t *sizes, void **chunks) THROWSPEC;
126 +
127 +/* These are the pool functions */
128 +struct nedpool_t;
129 +typedef struct nedpool_t nedpool;
130 +
131 +/* Creates a memory pool for use with the nedp* functions below.
132 +Capacity is how much to allocate immediately (if you know you'll be allocating a lot
133 +of memory very soon) which you can leave at zero. Threads specifies how many threads
134 +will *normally* be accessing the pool concurrently. Setting this to zero means it
135 +extends on demand, but be careful of this as it can rapidly consume system resources
136 +where bursts of concurrent threads use a pool at once.
137 +*/
138 +EXTSPEC MALLOCATTR nedpool *nedcreatepool(size_t capacity, int threads) THROWSPEC;
139 +
140 +/* Destroys a memory pool previously created by nedcreatepool().
141 +*/
142 +EXTSPEC void neddestroypool(nedpool *p) THROWSPEC;
143 +
144 +/* Sets a value to be associated with a pool. You can retrieve this value by passing
145 +any memory block allocated from that pool.
146 +*/
147 +EXTSPEC void nedpsetvalue(nedpool *p, void *v) THROWSPEC;
148 +/* Gets a previously set value using nedpsetvalue() or zero if memory is unknown.
149 +Optionally can also retrieve pool.
150 +*/
151 +EXTSPEC void *nedgetvalue(nedpool **p, void *mem) THROWSPEC;
152 +
153 +/* Disables the thread cache for the calling thread, returning any existing cache
154 +data to the central pool.
155 +*/
156 +EXTSPEC void neddisablethreadcache(nedpool *p) THROWSPEC;
157 +
158 +EXTSPEC MALLOCATTR void * nedpmalloc(nedpool *p, size_t size) THROWSPEC;
159 +EXTSPEC MALLOCATTR void * nedpcalloc(nedpool *p, size_t no, size_t size) THROWSPEC;
160 +EXTSPEC MALLOCATTR void * nedprealloc(nedpool *p, void *mem, size_t size) THROWSPEC;
161 +EXTSPEC void nedpfree(nedpool *p, void *mem) THROWSPEC;
162 +EXTSPEC MALLOCATTR void * nedpmemalign(nedpool *p, size_t alignment, size_t bytes) THROWSPEC;
163 +#if !NO_MALLINFO
164 +EXTSPEC struct mallinfo nedpmallinfo(nedpool *p) THROWSPEC;
165 +#endif
166 +EXTSPEC int nedpmallopt(nedpool *p, int parno, int value) THROWSPEC;
167 +EXTSPEC int nedpmalloc_trim(nedpool *p, size_t pad) THROWSPEC;
168 +EXTSPEC void nedpmalloc_stats(nedpool *p) THROWSPEC;
169 +EXTSPEC size_t nedpmalloc_footprint(nedpool *p) THROWSPEC;
170 +EXTSPEC MALLOCATTR void **nedpindependent_calloc(nedpool *p, size_t elemsno, size_t elemsize, void **chunks) THROWSPEC;
171 +EXTSPEC MALLOCATTR void **nedpindependent_comalloc(nedpool *p, size_t elems, size_t *sizes, void **chunks) THROWSPEC;
172 +
173 +#if defined(__cplusplus)
174 +}
175 +#endif
176 +
177 +#undef MALLOCATTR
178 +#undef EXTSPEC
179 +
180 +#endif