1 @comment !!! describe mmap et al (here?)
4 @node Memory Allocation, Character Handling, Error Reporting, Top
5 @chapter Memory Allocation
6 @cindex memory allocation
7 @cindex storage allocation
9 The GNU system provides several methods for allocating memory space
10 under explicit program control. They vary in generality and in
16 The @code{malloc} facility allows fully general dynamic allocation.
17 @xref{Unconstrained Allocation}.
20 Obstacks are another facility, less general than @code{malloc} but more
21 efficient and convenient for stacklike allocation. @xref{Obstacks}.
24 The function @code{alloca} lets you allocate storage dynamically that
25 will be freed automatically. @xref{Variable Size Automatic}.
30 * Memory Concepts:: An introduction to concepts and terminology.
31 * Dynamic Allocation and C:: How to get different kinds of allocation in C.
32 * Unconstrained Allocation:: The @code{malloc} facility allows fully general
34 * Allocation Debugging:: Finding memory leaks and not freed memory.
35 * Obstacks:: Obstacks are less general than malloc
36 but more efficient and convenient.
37 * Variable Size Automatic:: Allocation of variable-sized blocks
38 of automatic storage that are freed when the
39 calling function returns.
40 * Relocating Allocator:: Waste less memory, if you can tolerate
41 automatic relocation of the blocks you get.
45 @section Dynamic Memory Allocation Concepts
46 @cindex dynamic allocation
47 @cindex static allocation
48 @cindex automatic allocation
50 @dfn{Dynamic memory allocation} is a technique in which programs
51 determine as they are running where to store some information. You need
52 dynamic allocation when the number of memory blocks you need, or how
53 long you continue to need them, depends on the data you are working on.
55 For example, you may need a block to store a line read from an input file;
56 since there is no limit to how long a line can be, you must allocate the
57 storage dynamically and make it dynamically larger as you read more of the
60 Or, you may need a block for each record or each definition in the input
61 data; since you can't know in advance how many there will be, you must
62 allocate a new block for each record or definition as you read it.
64 When you use dynamic allocation, the allocation of a block of memory is an
65 action that the program requests explicitly. You call a function or macro
66 when you want to allocate space, and specify the size with an argument. If
67 you want to free the space, you do so by calling another function or macro.
68 You can do these things whenever you want, as often as you want.
70 @node Dynamic Allocation and C
71 @section Dynamic Allocation and C
73 The C language supports two kinds of memory allocation through the variables
78 @dfn{Static allocation} is what happens when you declare a static or
79 global variable. Each static or global variable defines one block of
80 space, of a fixed size. The space is allocated once, when your program
81 is started, and is never freed.
84 @dfn{Automatic allocation} happens when you declare an automatic
85 variable, such as a function argument or a local variable. The space
86 for an automatic variable is allocated when the compound statement
87 containing the declaration is entered, and is freed when that
88 compound statement is exited.
90 In GNU C, the length of the automatic storage can be an expression
91 that varies. In other C implementations, it must be a constant.
94 Dynamic allocation is not supported by C variables; there is no storage
95 class ``dynamic'', and there can never be a C variable whose value is
96 stored in dynamically allocated space. The only way to refer to
97 dynamically allocated space is through a pointer. Because it is less
98 convenient, and because the actual process of dynamic allocation
99 requires more computation time, programmers generally use dynamic
100 allocation only when neither static nor automatic allocation will serve.
102 For example, if you want to allocate dynamically some space to hold a
103 @code{struct foobar}, you cannot declare a variable of type @code{struct
104 foobar} whose contents are the dynamically allocated space. But you can
105 declare a variable of pointer type @code{struct foobar *} and assign it the
106 address of the space. Then you can use the operators @samp{*} and
107 @samp{->} on this pointer variable to refer to the contents of the space:
112 = (struct foobar *) malloc (sizeof (struct foobar));
114 ptr->next = current_foobar;
115 current_foobar = ptr;
119 @node Unconstrained Allocation
120 @section Unconstrained Allocation
121 @cindex unconstrained storage allocation
122 @cindex @code{malloc} function
123 @cindex heap, dynamic allocation from
125 The most general dynamic allocation facility is @code{malloc}. It
126 allows you to allocate blocks of memory of any size at any time, make
127 them bigger or smaller at any time, and free the blocks individually at
131 * Basic Allocation:: Simple use of @code{malloc}.
132 * Malloc Examples:: Examples of @code{malloc}. @code{xmalloc}.
133 * Freeing after Malloc:: Use @code{free} to free a block you
134 got with @code{malloc}.
135 * Changing Block Size:: Use @code{realloc} to make a block
137 * Allocating Cleared Space:: Use @code{calloc} to allocate a
139 * Efficiency and Malloc:: Efficiency considerations in use of
141 * Aligned Memory Blocks:: Allocating specially aligned memory:
142 @code{memalign} and @code{valloc}.
143 * Malloc Tunable Parameters:: Use @code{mallopt} to adjust allocation
145 * Heap Consistency Checking:: Automatic checking for errors.
146 * Hooks for Malloc:: You can use these hooks for debugging
147 programs that use @code{malloc}.
148 * Statistics of Malloc:: Getting information about how much
149 memory your program is using.
150 * Summary of Malloc:: Summary of @code{malloc} and related functions.
153 @node Basic Allocation
154 @subsection Basic Storage Allocation
155 @cindex allocation of memory with @code{malloc}
157 To allocate a block of memory, call @code{malloc}. The prototype for
158 this function is in @file{stdlib.h}.
161 @comment malloc.h stdlib.h
163 @deftypefun {void *} malloc (size_t @var{size})
164 This function returns a pointer to a newly allocated block @var{size}
165 bytes long, or a null pointer if the block could not be allocated.
168 The contents of the block are undefined; you must initialize it yourself
169 (or use @code{calloc} instead; @pxref{Allocating Cleared Space}).
170 Normally you would cast the value as a pointer to the kind of object
171 that you want to store in the block. Here we show an example of doing
172 so, and of initializing the space with zeros using the library function
173 @code{memset} (@pxref{Copying and Concatenation}):
178 ptr = (struct foo *) malloc (sizeof (struct foo));
179 if (ptr == 0) abort ();
180 memset (ptr, 0, sizeof (struct foo));
183 You can store the result of @code{malloc} into any pointer variable
184 without a cast, because @w{ISO C} automatically converts the type
185 @code{void *} to another type of pointer when necessary. But the cast
186 is necessary in contexts other than assignment operators or if you might
187 want your code to run in traditional C.
189 Remember that when allocating space for a string, the argument to
190 @code{malloc} must be one plus the length of the string. This is
191 because a string is terminated with a null character that doesn't count
192 in the ``length'' of the string but does need space. For example:
197 ptr = (char *) malloc (length + 1);
201 @xref{Representation of Strings}, for more information about this.
203 @node Malloc Examples
204 @subsection Examples of @code{malloc}
206 If no more space is available, @code{malloc} returns a null pointer.
207 You should check the value of @emph{every} call to @code{malloc}. It is
208 useful to write a subroutine that calls @code{malloc} and reports an
209 error if the value is a null pointer, returning only if the value is
210 nonzero. This function is conventionally called @code{xmalloc}. Here
215 xmalloc (size_t size)
217 register void *value = malloc (size);
219 fatal ("virtual memory exhausted");
224 Here is a real example of using @code{malloc} (by way of @code{xmalloc}).
225 The function @code{savestring} will copy a sequence of characters into
226 a newly allocated null-terminated string:
231 savestring (const char *ptr, size_t len)
233 register char *value = (char *) xmalloc (len + 1);
234 memcpy (value, ptr, len);
241 The block that @code{malloc} gives you is guaranteed to be aligned so
242 that it can hold any type of data. In the GNU system, the address is
243 always a multiple of eight on most systems, and a multiple of 16 on
244 64-bit systems. Only rarely is any higher boundary (such as a page
245 boundary) necessary; for those cases, use @code{memalign} or
246 @code{valloc} (@pxref{Aligned Memory Blocks}).
248 Note that the memory located after the end of the block is likely to be
249 in use for something else; perhaps a block already allocated by another
250 call to @code{malloc}. If you attempt to treat the block as longer than
251 you asked for it to be, you are liable to destroy the data that
252 @code{malloc} uses to keep track of its blocks, or you may destroy the
253 contents of another block. If you have already allocated a block and
254 discover you want it to be bigger, use @code{realloc} (@pxref{Changing
257 @node Freeing after Malloc
258 @subsection Freeing Memory Allocated with @code{malloc}
259 @cindex freeing memory allocated with @code{malloc}
260 @cindex heap, freeing memory from
262 When you no longer need a block that you got with @code{malloc}, use the
263 function @code{free} to make the block available to be allocated again.
264 The prototype for this function is in @file{stdlib.h}.
267 @comment malloc.h stdlib.h
269 @deftypefun void free (void *@var{ptr})
270 The @code{free} function deallocates the block of storage pointed at
276 @deftypefun void cfree (void *@var{ptr})
277 This function does the same thing as @code{free}. It's provided for
278 backward compatibility with SunOS; you should use @code{free} instead.
281 Freeing a block alters the contents of the block. @strong{Do not expect to
282 find any data (such as a pointer to the next block in a chain of blocks) in
283 the block after freeing it.} Copy whatever you need out of the block before
284 freeing it! Here is an example of the proper way to free all the blocks in
285 a chain, and the strings that they point to:
295 free_chain (struct chain *chain)
299 struct chain *next = chain->next;
307 Occasionally, @code{free} can actually return memory to the operating
308 system and make the process smaller. Usually, all it can do is allow a
309 later call to @code{malloc} to reuse the space. In the meantime, the
310 space remains in your program as part of a free-list used internally by
313 There is no point in freeing blocks at the end of a program, because all
314 of the program's space is given back to the system when the process
317 @node Changing Block Size
318 @subsection Changing the Size of a Block
319 @cindex changing the size of a block (@code{malloc})
321 Often you do not know for certain how big a block you will ultimately need
322 at the time you must begin to use the block. For example, the block might
323 be a buffer that you use to hold a line being read from a file; no matter
324 how long you make the buffer initially, you may encounter a line that is
327 You can make the block longer by calling @code{realloc}. This function
328 is declared in @file{stdlib.h}.
331 @comment malloc.h stdlib.h
333 @deftypefun {void *} realloc (void *@var{ptr}, size_t @var{newsize})
334 The @code{realloc} function changes the size of the block whose address is
335 @var{ptr} to be @var{newsize}.
337 Since the space after the end of the block may be in use, @code{realloc}
338 may find it necessary to copy the block to a new address where more free
339 space is available. The value of @code{realloc} is the new address of the
340 block. If the block needs to be moved, @code{realloc} copies the old
343 If you pass a null pointer for @var{ptr}, @code{realloc} behaves just
344 like @samp{malloc (@var{newsize})}. This can be convenient, but beware
345 that older implementations (before @w{ISO C}) may not support this
346 behavior, and will probably crash when @code{realloc} is passed a null
350 Like @code{malloc}, @code{realloc} may return a null pointer if no
351 memory space is available to make the block bigger. When this happens,
352 the original block is untouched; it has not been modified or relocated.
354 In most cases it makes no difference what happens to the original block
355 when @code{realloc} fails, because the application program cannot continue
356 when it is out of memory, and the only thing to do is to give a fatal error
357 message. Often it is convenient to write and use a subroutine,
358 conventionally called @code{xrealloc}, that takes care of the error message
359 as @code{xmalloc} does for @code{malloc}:
363 xrealloc (void *ptr, size_t size)
365 register void *value = realloc (ptr, size);
367 fatal ("Virtual memory exhausted");
372 You can also use @code{realloc} to make a block smaller. The reason you
373 would do this is to avoid tying up a lot of memory space when only a little
375 @comment The following is no longer true with the new malloc.
376 @comment But it seems wise to keep the warning for other implementations.
377 In several allocation implementations, making a block smaller sometimes
378 necessitates copying it, so it can fail if no other space is available.
380 If the new size you specify is the same as the old size, @code{realloc}
381 is guaranteed to change nothing and return the same address that you gave.
383 @node Allocating Cleared Space
384 @subsection Allocating Cleared Space
386 The function @code{calloc} allocates memory and clears it to zero. It
387 is declared in @file{stdlib.h}.
390 @comment malloc.h stdlib.h
392 @deftypefun {void *} calloc (size_t @var{count}, size_t @var{eltsize})
393 This function allocates a block long enough to contain a vector of
394 @var{count} elements, each of size @var{eltsize}. Its contents are
395 cleared to zero before @code{calloc} returns.
398 You could define @code{calloc} as follows:
402 calloc (size_t count, size_t eltsize)
404 size_t size = count * eltsize;
405 void *value = malloc (size);
407 memset (value, 0, size);
412 But in general, it is not guaranteed that @code{calloc} calls
413 @code{malloc} internally. Therefore, if an application provides its own
414 @code{malloc}/@code{realloc}/@code{free} outside the C library, it
415 should always define @code{calloc}, too.
417 @node Efficiency and Malloc
418 @subsection Efficiency Considerations for @code{malloc}
419 @cindex efficiency and @code{malloc}
423 @c No longer true, see below instead.
424 To make the best use of @code{malloc}, it helps to know that the GNU
425 version of @code{malloc} always dispenses small amounts of memory in
426 blocks whose sizes are powers of two. It keeps separate pools for each
427 power of two. This holds for sizes up to a page size. Therefore, if
428 you are free to choose the size of a small block in order to make
429 @code{malloc} more efficient, make it a power of two.
430 @c !!! xref getpagesize
432 Once a page is split up for a particular block size, it can't be reused
433 for another size unless all the blocks in it are freed. In many
434 programs, this is unlikely to happen. Thus, you can sometimes make a
435 program use memory more efficiently by using blocks of the same size for
436 many different purposes.
438 When you ask for memory blocks of a page or larger, @code{malloc} uses a
439 different strategy; it rounds the size up to a multiple of a page, and
440 it can coalesce and split blocks as needed.
442 The reason for the two strategies is that it is important to allocate
443 and free small blocks as fast as possible, but speed is less important
444 for a large block since the program normally spends a fair amount of
445 time using it. Also, large blocks are normally fewer in number.
446 Therefore, for large blocks, it makes sense to use a method which takes
447 more time to minimize the wasted space.
451 As apposed to other versions, the @code{malloc} in GNU libc does not
452 round up block sizes to powers of two, neither for large nor for small
453 sizes. Neighboring chunks can be coalesced on a @code{free} no matter
454 what their size is. This makes the implementation suitable for all
455 kinds of allocation patterns without generally incurring high memory
456 waste through fragmentation.
458 Very large blocks (much larger than a page) are allocated with
459 @code{mmap} (anonymous or via @code{/dev/zero}) by this implementation.
460 This has the great advantage that these chunks are returned to the
461 system immediately when they are freed. Therefore, it cannot happen
462 that a large chunk becomes ``locked'' in between smaller ones and even
463 after calling @code{free} wastes memory. The size threshold for
464 @code{mmap} to be used can be adjusted with @code{mallopt}. The use of
465 @code{mmap} can also be disabled completely.
467 @node Aligned Memory Blocks
468 @subsection Allocating Aligned Memory Blocks
470 @cindex page boundary
471 @cindex alignment (with @code{malloc})
473 The address of a block returned by @code{malloc} or @code{realloc} in
474 the GNU system is always a multiple of eight (or sixteen on 64-bit
475 systems). If you need a block whose address is a multiple of a higher
476 power of two than that, use @code{memalign} or @code{valloc}. These
477 functions are declared in @file{stdlib.h}.
479 With the GNU library, you can use @code{free} to free the blocks that
480 @code{memalign} and @code{valloc} return. That does not work in BSD,
481 however---BSD does not provide any way to free such blocks.
483 @comment malloc.h stdlib.h
485 @deftypefun {void *} memalign (size_t @var{boundary}, size_t @var{size})
486 The @code{memalign} function allocates a block of @var{size} bytes whose
487 address is a multiple of @var{boundary}. The @var{boundary} must be a
488 power of two! The function @code{memalign} works by allocating a
489 somewhat larger block, and then returning an address within the block
490 that is on the specified boundary.
493 @comment malloc.h stdlib.h
495 @deftypefun {void *} valloc (size_t @var{size})
496 Using @code{valloc} is like using @code{memalign} and passing the page size
497 as the value of the second argument. It is implemented like this:
503 return memalign (getpagesize (), size);
506 @c !!! xref getpagesize
509 @node Malloc Tunable Parameters
510 @subsection Malloc Tunable Parameters
512 You can adjust some parameters for dynamic memory allocation with the
513 @code{mallopt} function. This function is the general SVID/XPG
514 interface, defined in @file{malloc.h}.
517 @deftypefun int mallopt (int @var{param}, int @var{value})
518 When calling @code{mallopt}, the @var{param} argument specifies the
519 parameter to be set, and @var{value} the new value to be set. Possible
520 choices for @var{param}, as defined in @file{malloc.h}, are:
523 @item M_TRIM_THRESHOLD
524 This is the minimum size (in bytes) of the top-most, releaseable chunk
525 that will cause @code{sbrk} to be called with a negative argument in
526 order to return memory to the system.
528 This parameter determines the amount of extra memory to obtain from the
529 system when a call to @code{sbrk} is required. It also specifies the
530 number of bytes to retain when shrinking the heap by calling @code{sbrk}
531 with a negative argument. This provides the necessary hysteresis in
532 heap size such that excessive amounts of system calls can be avoided.
533 @item M_MMAP_THRESHOLD
534 All chunks larger than this value are allocated outside the normal
535 heap, using the @code{mmap} system call. This way it is guaranteed
536 that the memory for these chunks can be returned to the system on
539 The maximum number of chunks to allocate with @code{mmap}. Setting this
540 to zero disables all use of @code{mmap}.
545 @node Heap Consistency Checking
546 @subsection Heap Consistency Checking
548 @cindex heap consistency checking
549 @cindex consistency checking, of heap
551 You can ask @code{malloc} to check the consistency of dynamic storage by
552 using the @code{mcheck} function. This function is a GNU extension,
553 declared in @file{mcheck.h}.
558 @deftypefun int mcheck (void (*@var{abortfn}) (enum mcheck_status @var{status}))
559 Calling @code{mcheck} tells @code{malloc} to perform occasional
560 consistency checks. These will catch things such as writing
561 past the end of a block that was allocated with @code{malloc}.
563 The @var{abortfn} argument is the function to call when an inconsistency
564 is found. If you supply a null pointer, then @code{mcheck} uses a
565 default function which prints a message and calls @code{abort}
566 (@pxref{Aborting a Program}). The function you supply is called with
567 one argument, which says what sort of inconsistency was detected; its
568 type is described below.
570 It is too late to begin allocation checking once you have allocated
571 anything with @code{malloc}. So @code{mcheck} does nothing in that
572 case. The function returns @code{-1} if you call it too late, and
573 @code{0} otherwise (when it is successful).
575 The easiest way to arrange to call @code{mcheck} early enough is to use
576 the option @samp{-lmcheck} when you link your program; then you don't
577 need to modify your program source at all.
580 @deftypefun {enum mcheck_status} mprobe (void *@var{pointer})
581 The @code{mprobe} function lets you explicitly check for inconsistencies
582 in a particular allocated block. You must have already called
583 @code{mcheck} at the beginning of the program, to do its occasional
584 checks; calling @code{mprobe} requests an additional consistency check
585 to be done at the time of the call.
587 The argument @var{pointer} must be a pointer returned by @code{malloc}
588 or @code{realloc}. @code{mprobe} returns a value that says what
589 inconsistency, if any, was found. The values are described below.
592 @deftp {Data Type} {enum mcheck_status}
593 This enumerated type describes what kind of inconsistency was detected
594 in an allocated block, if any. Here are the possible values:
597 @item MCHECK_DISABLED
598 @code{mcheck} was not called before the first allocation.
599 No consistency checking can be done.
601 No inconsistency detected.
603 The data immediately before the block was modified.
604 This commonly happens when an array index or pointer
605 is decremented too far.
607 The data immediately after the block was modified.
608 This commonly happens when an array index or pointer
609 is incremented too far.
611 The block was already freed.
615 Another possibility to check for and guard against bugs in the use of
616 @code{malloc}, @code{realloc} and @code{free} is to set the environment
617 variable @code{MALLOC_CHECK_}. When @code{MALLOC_CHECK_} is set, a
618 special (less efficient) implementation is used which is designed to be
619 tolerant against simple errors, such as double calls of @code{free} with
620 the same argument, or overruns of a single byte (off-by-one bugs). Not
621 all such errors can be proteced against, however, and memory leaks can
622 result. If @code{MALLOC_CHECK_} is set to @code{0}, any detected heap
623 corruption is silently ignored; if set to @code{1}, a diagnostic is
624 printed on @code{stderr}; if set to @code{2}, @code{abort} is called
625 immediately. This can be useful because otherwise a crash may happen
626 much later, and the true cause for the problem is then very hard to
629 So, what's the difference between using @code{MALLOC_CHECK_} and linking
630 with @samp{-lmcheck}? @code{MALLOC_CHECK_} is orthognal with respect to
631 @samp{-lmcheck}. @samp{-lmcheck} has been added for backward
632 compatibility. Both @code{MALLOC_CHECK_} and @samp{-lmcheck} should
633 uncover the same bugs - but using @code{MALLOC_CHECK_} you don't need to
634 recompile your application.
636 @node Hooks for Malloc
637 @subsection Storage Allocation Hooks
638 @cindex allocation hooks, for @code{malloc}
640 The GNU C library lets you modify the behavior of @code{malloc},
641 @code{realloc}, and @code{free} by specifying appropriate hook
642 functions. You can use these hooks to help you debug programs that use
643 dynamic storage allocation, for example.
645 The hook variables are declared in @file{malloc.h}.
650 @defvar __malloc_hook
651 The value of this variable is a pointer to function that @code{malloc}
652 uses whenever it is called. You should define this function to look
653 like @code{malloc}; that is, like:
656 void *@var{function} (size_t @var{size}, void *@var{caller})
659 The value of @var{caller} is the return address found on the stack when
660 the @code{malloc} function was called. This value allows to trace the
661 memory consumption of the program.
666 @defvar __realloc_hook
667 The value of this variable is a pointer to function that @code{realloc}
668 uses whenever it is called. You should define this function to look
669 like @code{realloc}; that is, like:
672 void *@var{function} (void *@var{ptr}, size_t @var{size}, void *@var{caller})
675 The value of @var{caller} is the return address found on the stack when
676 the @code{realloc} function was called. This value allows to trace the
677 memory consumption of the program.
683 The value of this variable is a pointer to function that @code{free}
684 uses whenever it is called. You should define this function to look
685 like @code{free}; that is, like:
688 void @var{function} (void *@var{ptr}, void *@var{caller})
691 The value of @var{caller} is the return address found on the stack when
692 the @code{free} function was called. This value allows to trace the
693 memory consumption of the program.
696 You must make sure that the function you install as a hook for one of
697 these functions does not call that function recursively without restoring
698 the old value of the hook first! Otherwise, your program will get stuck
699 in an infinite recursion.
701 Here is an example showing how to use @code{__malloc_hook} properly. It
702 installs a function that prints out information every time @code{malloc}
706 static void *(*old_malloc_hook) (size_t);
708 my_malloc_hook (size_t size)
711 __malloc_hook = old_malloc_hook;
712 result = malloc (size);
713 /* @r{@code{printf} might call @code{malloc}, so protect it too.} */
714 printf ("malloc (%u) returns %p\n", (unsigned int) size, result);
715 __malloc_hook = my_malloc_hook;
722 old_malloc_hook = __malloc_hook;
723 __malloc_hook = my_malloc_hook;
728 The @code{mcheck} function (@pxref{Heap Consistency Checking}) works by
729 installing such hooks.
731 @c __morecore, __after_morecore_hook are undocumented
732 @c It's not clear whether to document them.
734 @node Statistics of Malloc
735 @subsection Statistics for Storage Allocation with @code{malloc}
737 @cindex allocation statistics
738 You can get information about dynamic storage allocation by calling the
739 @code{mallinfo} function. This function and its associated data type
740 are declared in @file{malloc.h}; they are an extension of the standard
746 @deftp {Data Type} {struct mallinfo}
747 This structure type is used to return information about the dynamic
748 storage allocator. It contains the following members:
752 This is the total size of memory allocated with @code{sbrk} by
753 @code{malloc}, in bytes.
756 This is the number of chunks not in use. (The storage allocator
757 internally gets chunks of memory from the operating system, and then
758 carves them up to satisfy individual @code{malloc} requests; see
759 @ref{Efficiency and Malloc}.)
762 This field is unused.
765 This is the total number of chunks allocated with @code{mmap}.
768 This is the total size of memory allocated with @code{mmap}, in bytes.
771 This field is unused.
774 This field is unused.
777 This is the total size of memory occupied by chunks handed out by
781 This is the total size of memory occupied by free (not in use) chunks.
784 This is the size of the top-most, releaseable chunk that normally
785 borders the end of the heap (i.e. the ``brk'' of the process).
792 @deftypefun {struct mallinfo} mallinfo (void)
793 This function returns information about the current dynamic memory usage
794 in a structure of type @code{struct mallinfo}.
797 @node Summary of Malloc
798 @subsection Summary of @code{malloc}-Related Functions
800 Here is a summary of the functions that work with @code{malloc}:
803 @item void *malloc (size_t @var{size})
804 Allocate a block of @var{size} bytes. @xref{Basic Allocation}.
806 @item void free (void *@var{addr})
807 Free a block previously allocated by @code{malloc}. @xref{Freeing after
810 @item void *realloc (void *@var{addr}, size_t @var{size})
811 Make a block previously allocated by @code{malloc} larger or smaller,
812 possibly by copying it to a new location. @xref{Changing Block Size}.
814 @item void *calloc (size_t @var{count}, size_t @var{eltsize})
815 Allocate a block of @var{count} * @var{eltsize} bytes using
816 @code{malloc}, and set its contents to zero. @xref{Allocating Cleared
819 @item void *valloc (size_t @var{size})
820 Allocate a block of @var{size} bytes, starting on a page boundary.
821 @xref{Aligned Memory Blocks}.
823 @item void *memalign (size_t @var{size}, size_t @var{boundary})
824 Allocate a block of @var{size} bytes, starting on an address that is a
825 multiple of @var{boundary}. @xref{Aligned Memory Blocks}.
827 @item int mallopt (int @var{param}, int @var{value})
828 Adjust a tunable parameter. @xref{Malloc Tunable Parameters}
830 @item int mcheck (void (*@var{abortfn}) (void))
831 Tell @code{malloc} to perform occasional consistency checks on
832 dynamically allocated memory, and to call @var{abortfn} when an
833 inconsistency is found. @xref{Heap Consistency Checking}.
835 @item void *(*__malloc_hook) (size_t @var{size}, void *@var{caller})
836 A pointer to a function that @code{malloc} uses whenever it is called.
838 @item void *(*__realloc_hook) (void *@var{ptr}, size_t @var{size}, void *@var{caller})
839 A pointer to a function that @code{realloc} uses whenever it is called.
841 @item void (*__free_hook) (void *@var{ptr}, void *@var{caller})
842 A pointer to a function that @code{free} uses whenever it is called.
844 @item struct mallinfo mallinfo (void)
845 Return information about the current dynamic memory usage.
846 @xref{Statistics of Malloc}.
849 @node Allocation Debugging
850 @section Allocation Debugging
851 @cindex allocation debugging
852 @cindex malloc debugger
854 An complicated task when programming with languages which do not use
855 garbage collected dynamic memory allocation is to find memory leaks.
856 Long running programs must assure that dynamically allocated objects are
857 freed at the end of their lifetime. If this does not happen the system
858 runs out of memory, sooner or later.
860 The @code{malloc} implementation in the GNU C library provides some
861 simple means to detect sich leaks and provide some information to find
862 the location. To do this the application must be started in a special
863 mode which is enabled by an environment variable. There are no speed
864 penalties if the program is compiled in preparation of the debugging if
865 the debug mode is not enabled.
868 * Tracing malloc:: How to install the tracing functionality.
869 * Using the Memory Debugger:: Example programs excerpts.
870 * Tips for the Memory Debugger:: Some more or less clever ideas.
871 * Interpreting the traces:: What do all these lines mean?
875 @subsection How to install the tracing functionality
879 @deftypefun void mtrace (void)
880 When the @code{mtrace} function is called it looks for an environment
881 variable named @code{MALLOC_TRACE}. This variable is supposed to
882 contain a valid file name. The user must have write access. If the
883 file already exists it is truncated. If the environment variable is not
884 set or it does not name a valid file which can be opened for writing
885 nothing is done. The behaviour of @code{malloc} etc. is not changed.
886 For obvious reasons this also happens if the application is install SUID
889 If the named file is successfully opened @code{mtrace} installs special
890 handlers for the functions @code{malloc}, @code{realloc}, and
891 @code{free} (@pxref{Hooks for Malloc}). From now on all uses of these
892 functions are traced and protocolled into the file. There is now of
893 course a speed penalty for all calls to the traced functions so that the
894 tracing should not be enabled during their normal use.
896 This function is a GNU extension and generally not available on other
897 systems. The prototype can be found in @file{mcheck.h}.
902 @deftypefun void muntrace (void)
903 The @code{muntrace} function can be called after @code{mtrace} was used
904 to enable tracing the @code{malloc} calls. If no (succesful) call of
905 @code{mtrace} was made @code{muntrace} does nothing.
907 Otherwise it deinstalls the handlers for @code{malloc}, @code{realloc},
908 and @code{free} and then closes the protocol file. No calls are
909 protocolled anymore and the programs runs again with the full speed.
911 This function is a GNU extension and generally not available on other
912 systems. The prototype can be found in @file{mcheck.h}.
915 @node Using the Memory Debugger
916 @subsection Example programs excerpts
918 Even though the tracing functionality does not influence the runtime
919 behaviour of the program it is no wise idea to call @code{mtrace} in all
920 programs. Just imagine you debug a program using @code{mtrace} and all
921 other programs used in the debug sessions also trace their @code{malloc}
922 calls. The output file would be the same for all programs and so is
923 unusable. Therefore one should call @code{mtrace} only if compiled for
924 debugging. A program could therefore start like this:
930 main (int argc, char *argv[])
939 This is all what is needed if you want to trace the calls during the
940 whole runtime of the program. Alternatively you can stop the tracing at
941 any time with a call to @code{muntrace}. It is even possible to restart
942 the tracing again with a new call to @code{mtrace}. But this can course
943 unreliable results since there are possibly calls of the functions which
944 are not called. Please note that not only the application uses the
945 traced functions, also libraries (including the C library itself) use
948 This last point is also why it is no good idea to call @code{muntrace}
949 before the program terminated. The libraries are informed about the
950 termination of the program only after the program returns from
951 @code{main} or calls @code{exit} and so cannot free the memory they use
954 So the best thing one can do is to call @code{mtrace} as the very first
955 function in the program and never call @code{muntrace}. So the program
956 traces almost all uses of the @code{malloc} functions (except those
957 calls which are executed by constructors of the program or used
960 @node Tips for the Memory Debugger
961 @subsection Some more or less clever ideas
963 You know the situation. The program is prepared for debugging and in
964 all debugging sessions it runs well. But once it is started without
965 debugging the error shows up. In our situation here: the memory leaks
966 becomes visible only when we just turned off the debugging. If you
967 foresee such situations you can still win. Simply use something
968 equivalent to the following little program:
978 signal (SIGUSR1, enable);
985 signal (SIGUSR2, disable);
989 main (int argc, char *argv[])
993 signal (SIGUSR1, enable);
994 signal (SIGUSR2, disable);
1000 I.e., the user can start the memory debugger any time s/he wants if the
1001 program was started with @code{MALLOC_TRACE} set in the environment.
1002 The output will of course not show the allocations which happened before
1003 the first signal but if there is a memory leak this will show up
1006 @node Interpreting the traces
1007 @subsection Interpreting the traces
1009 If you take a look at the output it will look similar to this:
1013 @ [0x8048209] - 0x8064cc8
1014 @ [0x8048209] - 0x8064ce0
1015 @ [0x8048209] - 0x8064cf8
1016 @ [0x80481eb] + 0x8064c48 0x14
1017 @ [0x80481eb] + 0x8064c60 0x14
1018 @ [0x80481eb] + 0x8064c78 0x14
1019 @ [0x80481eb] + 0x8064c90 0x14
1023 What this all means is not really important since the trace file is not
1024 meant to be read by a human. Therefore no attention is payed to good
1025 readability. Instead there is a program which comes with the GNU C
1026 library which interprets the traces and outputs a summary in on
1027 user-friendly way. The program is called @code{mtrace} (it is in fact a
1028 Perl script) and it takes one or two arguments. In any case the name of
1029 the file with the trace output must be specified. If an optional argument
1030 precedes the name of the trace file this must be the name of the program
1031 which generated the trace.
1034 drepper$ mtrace tst-mtrace log
1038 In this case the program @code{tst-mtrace} was run and it produced a
1039 trace file @file{log}. The message printed by @code{mtrace} shows there
1040 are no problems with the code, all allocated memory was freed
1043 If we call @code{mtrace} on the example trace given above we would get a
1047 drepper$ mtrace errlog
1048 - 0x08064cc8 Free 2 was never alloc'd 0x8048209
1049 - 0x08064ce0 Free 3 was never alloc'd 0x8048209
1050 - 0x08064cf8 Free 4 was never alloc'd 0x8048209
1055 0x08064c48 0x14 at 0x80481eb
1056 0x08064c60 0x14 at 0x80481eb
1057 0x08064c78 0x14 at 0x80481eb
1058 0x08064c90 0x14 at 0x80481eb
1061 We have called @code{mtrace} with only one argument and so the script
1062 has no chance to find out what is meant with the addresses given in the
1063 trace. We can do better:
1066 drepper$ mtrace tst-mtrace errlog
1067 - 0x08064cc8 Free 2 was never alloc'd /home/drepper/tst-mtrace.c:39
1068 - 0x08064ce0 Free 3 was never alloc'd /home/drepper/tst-mtrace.c:39
1069 - 0x08064cf8 Free 4 was never alloc'd /home/drepper/tst-mtrace.c:39
1074 0x08064c48 0x14 at /home/drepper/tst-mtrace.c:33
1075 0x08064c60 0x14 at /home/drepper/tst-mtrace.c:33
1076 0x08064c78 0x14 at /home/drepper/tst-mtrace.c:33
1077 0x08064c90 0x14 at /home/drepper/tst-mtrace.c:33
1080 Suddenly the output makes much more sense and the user can see
1081 immediately where the function calls causing the trouble can be found.
1083 Interpreting this output is not complicated. There are at most two
1084 different situations being detected. First, @code{free} was called for
1085 pointers which were never returned by one of the allocation functions.
1086 This is usually a very bad problem and how this looks like is shown in
1087 the first three lines of the output. Situations like this are quite
1088 rare and if they appear they show up very drastically: the program
1091 The other situation which is much harder to detect are memory leaks. As
1092 you can see in the output the @code{mtrace} function collects all this
1093 information and so can say that the program calls an allocation function
1094 from line 33 in the source file @file{/home/drepper/tst-mtrace.c} four
1095 times without freeing this memory before the program terminates.
1096 Whether this is a real problem keeps to be investigated.
1102 An @dfn{obstack} is a pool of memory containing a stack of objects. You
1103 can create any number of separate obstacks, and then allocate objects in
1104 specified obstacks. Within each obstack, the last object allocated must
1105 always be the first one freed, but distinct obstacks are independent of
1108 Aside from this one constraint of order of freeing, obstacks are totally
1109 general: an obstack can contain any number of objects of any size. They
1110 are implemented with macros, so allocation is usually very fast as long as
1111 the objects are usually small. And the only space overhead per object is
1112 the padding needed to start each object on a suitable boundary.
1115 * Creating Obstacks:: How to declare an obstack in your program.
1116 * Preparing for Obstacks:: Preparations needed before you can
1118 * Allocation in an Obstack:: Allocating objects in an obstack.
1119 * Freeing Obstack Objects:: Freeing objects in an obstack.
1120 * Obstack Functions:: The obstack functions are both
1121 functions and macros.
1122 * Growing Objects:: Making an object bigger by stages.
1123 * Extra Fast Growing:: Extra-high-efficiency (though more
1124 complicated) growing objects.
1125 * Status of an Obstack:: Inquiries about the status of an obstack.
1126 * Obstacks Data Alignment:: Controlling alignment of objects in obstacks.
1127 * Obstack Chunks:: How obstacks obtain and release chunks;
1128 efficiency considerations.
1129 * Summary of Obstacks::
1132 @node Creating Obstacks
1133 @subsection Creating Obstacks
1135 The utilities for manipulating obstacks are declared in the header
1136 file @file{obstack.h}.
1141 @deftp {Data Type} {struct obstack}
1142 An obstack is represented by a data structure of type @code{struct
1143 obstack}. This structure has a small fixed size; it records the status
1144 of the obstack and how to find the space in which objects are allocated.
1145 It does not contain any of the objects themselves. You should not try
1146 to access the contents of the structure directly; use only the functions
1147 described in this chapter.
1150 You can declare variables of type @code{struct obstack} and use them as
1151 obstacks, or you can allocate obstacks dynamically like any other kind
1152 of object. Dynamic allocation of obstacks allows your program to have a
1153 variable number of different stacks. (You can even allocate an
1154 obstack structure in another obstack, but this is rarely useful.)
1156 All the functions that work with obstacks require you to specify which
1157 obstack to use. You do this with a pointer of type @code{struct obstack
1158 *}. In the following, we often say ``an obstack'' when strictly
1159 speaking the object at hand is such a pointer.
1161 The objects in the obstack are packed into large blocks called
1162 @dfn{chunks}. The @code{struct obstack} structure points to a chain of
1163 the chunks currently in use.
1165 The obstack library obtains a new chunk whenever you allocate an object
1166 that won't fit in the previous chunk. Since the obstack library manages
1167 chunks automatically, you don't need to pay much attention to them, but
1168 you do need to supply a function which the obstack library should use to
1169 get a chunk. Usually you supply a function which uses @code{malloc}
1170 directly or indirectly. You must also supply a function to free a chunk.
1171 These matters are described in the following section.
1173 @node Preparing for Obstacks
1174 @subsection Preparing for Using Obstacks
1176 Each source file in which you plan to use the obstack functions
1177 must include the header file @file{obstack.h}, like this:
1180 #include <obstack.h>
1183 @findex obstack_chunk_alloc
1184 @findex obstack_chunk_free
1185 Also, if the source file uses the macro @code{obstack_init}, it must
1186 declare or define two functions or macros that will be called by the
1187 obstack library. One, @code{obstack_chunk_alloc}, is used to allocate
1188 the chunks of memory into which objects are packed. The other,
1189 @code{obstack_chunk_free}, is used to return chunks when the objects in
1190 them are freed. These macros should appear before any use of obstacks
1193 Usually these are defined to use @code{malloc} via the intermediary
1194 @code{xmalloc} (@pxref{Unconstrained Allocation}). This is done with
1195 the following pair of macro definitions:
1198 #define obstack_chunk_alloc xmalloc
1199 #define obstack_chunk_free free
1203 Though the storage you get using obstacks really comes from @code{malloc},
1204 using obstacks is faster because @code{malloc} is called less often, for
1205 larger blocks of memory. @xref{Obstack Chunks}, for full details.
1207 At run time, before the program can use a @code{struct obstack} object
1208 as an obstack, it must initialize the obstack by calling
1209 @code{obstack_init}.
1213 @deftypefun int obstack_init (struct obstack *@var{obstack-ptr})
1214 Initialize obstack @var{obstack-ptr} for allocation of objects. This
1215 function calls the obstack's @code{obstack_chunk_alloc} function. It
1216 returns 0 if @code{obstack_chunk_alloc} returns a null pointer, meaning
1217 that it is out of memory. Otherwise, it returns 1. If you supply an
1218 @code{obstack_chunk_alloc} function that calls @code{exit}
1219 (@pxref{Program Termination}) or @code{longjmp} (@pxref{Non-Local
1220 Exits}) when out of memory, you can safely ignore the value that
1221 @code{obstack_init} returns.
1224 Here are two examples of how to allocate the space for an obstack and
1225 initialize it. First, an obstack that is a static variable:
1228 static struct obstack myobstack;
1230 obstack_init (&myobstack);
1234 Second, an obstack that is itself dynamically allocated:
1237 struct obstack *myobstack_ptr
1238 = (struct obstack *) xmalloc (sizeof (struct obstack));
1240 obstack_init (myobstack_ptr);
1243 @node Allocation in an Obstack
1244 @subsection Allocation in an Obstack
1245 @cindex allocation (obstacks)
1247 The most direct way to allocate an object in an obstack is with
1248 @code{obstack_alloc}, which is invoked almost like @code{malloc}.
1252 @deftypefun {void *} obstack_alloc (struct obstack *@var{obstack-ptr}, int @var{size})
1253 This allocates an uninitialized block of @var{size} bytes in an obstack
1254 and returns its address. Here @var{obstack-ptr} specifies which obstack
1255 to allocate the block in; it is the address of the @code{struct obstack}
1256 object which represents the obstack. Each obstack function or macro
1257 requires you to specify an @var{obstack-ptr} as the first argument.
1259 This function calls the obstack's @code{obstack_chunk_alloc} function if
1260 it needs to allocate a new chunk of memory; it returns a null pointer if
1261 @code{obstack_chunk_alloc} returns one. In that case, it has not
1262 changed the amount of memory allocated in the obstack. If you supply an
1263 @code{obstack_chunk_alloc} function that calls @code{exit}
1264 (@pxref{Program Termination}) or @code{longjmp} (@pxref{Non-Local
1265 Exits}) when out of memory, then @code{obstack_alloc} will never return
1269 For example, here is a function that allocates a copy of a string @var{str}
1270 in a specific obstack, which is in the variable @code{string_obstack}:
1273 struct obstack string_obstack;
1276 copystring (char *string)
1278 size_t len = strlen (string) + 1;
1279 char *s = (char *) obstack_alloc (&string_obstack, len);
1280 memcpy (s, string, len);
1285 To allocate a block with specified contents, use the function
1286 @code{obstack_copy}, declared like this:
1290 @deftypefun {void *} obstack_copy (struct obstack *@var{obstack-ptr}, void *@var{address}, int @var{size})
1291 This allocates a block and initializes it by copying @var{size}
1292 bytes of data starting at @var{address}. It can return a null pointer
1293 under the same conditions as @code{obstack_alloc}.
1298 @deftypefun {void *} obstack_copy0 (struct obstack *@var{obstack-ptr}, void *@var{address}, int @var{size})
1299 Like @code{obstack_copy}, but appends an extra byte containing a null
1300 character. This extra byte is not counted in the argument @var{size}.
1303 The @code{obstack_copy0} function is convenient for copying a sequence
1304 of characters into an obstack as a null-terminated string. Here is an
1309 obstack_savestring (char *addr, int size)
1311 return obstack_copy0 (&myobstack, addr, size);
1316 Contrast this with the previous example of @code{savestring} using
1317 @code{malloc} (@pxref{Basic Allocation}).
1319 @node Freeing Obstack Objects
1320 @subsection Freeing Objects in an Obstack
1321 @cindex freeing (obstacks)
1323 To free an object allocated in an obstack, use the function
1324 @code{obstack_free}. Since the obstack is a stack of objects, freeing
1325 one object automatically frees all other objects allocated more recently
1326 in the same obstack.
1330 @deftypefun void obstack_free (struct obstack *@var{obstack-ptr}, void *@var{object})
1331 If @var{object} is a null pointer, everything allocated in the obstack
1332 is freed. Otherwise, @var{object} must be the address of an object
1333 allocated in the obstack. Then @var{object} is freed, along with
1334 everything allocated in @var{obstack} since @var{object}.
1337 Note that if @var{object} is a null pointer, the result is an
1338 uninitialized obstack. To free all storage in an obstack but leave it
1339 valid for further allocation, call @code{obstack_free} with the address
1340 of the first object allocated on the obstack:
1343 obstack_free (obstack_ptr, first_object_allocated_ptr);
1346 Recall that the objects in an obstack are grouped into chunks. When all
1347 the objects in a chunk become free, the obstack library automatically
1348 frees the chunk (@pxref{Preparing for Obstacks}). Then other
1349 obstacks, or non-obstack allocation, can reuse the space of the chunk.
1351 @node Obstack Functions
1352 @subsection Obstack Functions and Macros
1355 The interfaces for using obstacks may be defined either as functions or
1356 as macros, depending on the compiler. The obstack facility works with
1357 all C compilers, including both @w{ISO C} and traditional C, but there are
1358 precautions you must take if you plan to use compilers other than GNU C.
1360 If you are using an old-fashioned @w{non-ISO C} compiler, all the obstack
1361 ``functions'' are actually defined only as macros. You can call these
1362 macros like functions, but you cannot use them in any other way (for
1363 example, you cannot take their address).
1365 Calling the macros requires a special precaution: namely, the first
1366 operand (the obstack pointer) may not contain any side effects, because
1367 it may be computed more than once. For example, if you write this:
1370 obstack_alloc (get_obstack (), 4);
1374 you will find that @code{get_obstack} may be called several times.
1375 If you use @code{*obstack_list_ptr++} as the obstack pointer argument,
1376 you will get very strange results since the incrementation may occur
1379 In @w{ISO C}, each function has both a macro definition and a function
1380 definition. The function definition is used if you take the address of the
1381 function without calling it. An ordinary call uses the macro definition by
1382 default, but you can request the function definition instead by writing the
1383 function name in parentheses, as shown here:
1388 /* @r{Use the macro}. */
1389 x = (char *) obstack_alloc (obptr, size);
1390 /* @r{Call the function}. */
1391 x = (char *) (obstack_alloc) (obptr, size);
1392 /* @r{Take the address of the function}. */
1393 funcp = obstack_alloc;
1397 This is the same situation that exists in @w{ISO C} for the standard library
1398 functions. @xref{Macro Definitions}.
1400 @strong{Warning:} When you do use the macros, you must observe the
1401 precaution of avoiding side effects in the first operand, even in @w{ISO C}.
1403 If you use the GNU C compiler, this precaution is not necessary, because
1404 various language extensions in GNU C permit defining the macros so as to
1405 compute each argument only once.
1407 @node Growing Objects
1408 @subsection Growing Objects
1409 @cindex growing objects (in obstacks)
1410 @cindex changing the size of a block (obstacks)
1412 Because storage in obstack chunks is used sequentially, it is possible to
1413 build up an object step by step, adding one or more bytes at a time to the
1414 end of the object. With this technique, you do not need to know how much
1415 data you will put in the object until you come to the end of it. We call
1416 this the technique of @dfn{growing objects}. The special functions
1417 for adding data to the growing object are described in this section.
1419 You don't need to do anything special when you start to grow an object.
1420 Using one of the functions to add data to the object automatically
1421 starts it. However, it is necessary to say explicitly when the object is
1422 finished. This is done with the function @code{obstack_finish}.
1424 The actual address of the object thus built up is not known until the
1425 object is finished. Until then, it always remains possible that you will
1426 add so much data that the object must be copied into a new chunk.
1428 While the obstack is in use for a growing object, you cannot use it for
1429 ordinary allocation of another object. If you try to do so, the space
1430 already added to the growing object will become part of the other object.
1434 @deftypefun void obstack_blank (struct obstack *@var{obstack-ptr}, int @var{size})
1435 The most basic function for adding to a growing object is
1436 @code{obstack_blank}, which adds space without initializing it.
1441 @deftypefun void obstack_grow (struct obstack *@var{obstack-ptr}, void *@var{data}, int @var{size})
1442 To add a block of initialized space, use @code{obstack_grow}, which is
1443 the growing-object analogue of @code{obstack_copy}. It adds @var{size}
1444 bytes of data to the growing object, copying the contents from
1450 @deftypefun void obstack_grow0 (struct obstack *@var{obstack-ptr}, void *@var{data}, int @var{size})
1451 This is the growing-object analogue of @code{obstack_copy0}. It adds
1452 @var{size} bytes copied from @var{data}, followed by an additional null
1458 @deftypefun void obstack_1grow (struct obstack *@var{obstack-ptr}, char @var{c})
1459 To add one character at a time, use the function @code{obstack_1grow}.
1460 It adds a single byte containing @var{c} to the growing object.
1465 @deftypefun void obstack_ptr_grow (struct obstack *@var{obstack-ptr}, void *@var{data})
1466 Adding the value of a pointer one can use the function
1467 @code{obstack_ptr_grow}. It adds @code{sizeof (void *)} bytes
1468 containing the value of @var{data}.
1473 @deftypefun void obstack_int_grow (struct obstack *@var{obstack-ptr}, int @var{data})
1474 A single value of type @code{int} can be added by using the
1475 @code{obstack_int_grow} function. It adds @code{sizeof (int)} bytes to
1476 the growing object and initializes them with the value of @var{data}.
1481 @deftypefun {void *} obstack_finish (struct obstack *@var{obstack-ptr})
1482 When you are finished growing the object, use the function
1483 @code{obstack_finish} to close it off and return its final address.
1485 Once you have finished the object, the obstack is available for ordinary
1486 allocation or for growing another object.
1488 This function can return a null pointer under the same conditions as
1489 @code{obstack_alloc} (@pxref{Allocation in an Obstack}).
1492 When you build an object by growing it, you will probably need to know
1493 afterward how long it became. You need not keep track of this as you grow
1494 the object, because you can find out the length from the obstack just
1495 before finishing the object with the function @code{obstack_object_size},
1496 declared as follows:
1500 @deftypefun int obstack_object_size (struct obstack *@var{obstack-ptr})
1501 This function returns the current size of the growing object, in bytes.
1502 Remember to call this function @emph{before} finishing the object.
1503 After it is finished, @code{obstack_object_size} will return zero.
1506 If you have started growing an object and wish to cancel it, you should
1507 finish it and then free it, like this:
1510 obstack_free (obstack_ptr, obstack_finish (obstack_ptr));
1514 This has no effect if no object was growing.
1516 @cindex shrinking objects
1517 You can use @code{obstack_blank} with a negative size argument to make
1518 the current object smaller. Just don't try to shrink it beyond zero
1519 length---there's no telling what will happen if you do that.
1521 @node Extra Fast Growing
1522 @subsection Extra Fast Growing Objects
1523 @cindex efficiency and obstacks
1525 The usual functions for growing objects incur overhead for checking
1526 whether there is room for the new growth in the current chunk. If you
1527 are frequently constructing objects in small steps of growth, this
1528 overhead can be significant.
1530 You can reduce the overhead by using special ``fast growth''
1531 functions that grow the object without checking. In order to have a
1532 robust program, you must do the checking yourself. If you do this checking
1533 in the simplest way each time you are about to add data to the object, you
1534 have not saved anything, because that is what the ordinary growth
1535 functions do. But if you can arrange to check less often, or check
1536 more efficiently, then you make the program faster.
1538 The function @code{obstack_room} returns the amount of room available
1539 in the current chunk. It is declared as follows:
1543 @deftypefun int obstack_room (struct obstack *@var{obstack-ptr})
1544 This returns the number of bytes that can be added safely to the current
1545 growing object (or to an object about to be started) in obstack
1546 @var{obstack} using the fast growth functions.
1549 While you know there is room, you can use these fast growth functions
1550 for adding data to a growing object:
1554 @deftypefun void obstack_1grow_fast (struct obstack *@var{obstack-ptr}, char @var{c})
1555 The function @code{obstack_1grow_fast} adds one byte containing the
1556 character @var{c} to the growing object in obstack @var{obstack-ptr}.
1561 @deftypefun void obstack_ptr_grow_fast (struct obstack *@var{obstack-ptr}, void *@var{data})
1562 The function @code{obstack_ptr_grow_fast} adds @code{sizeof (void *)}
1563 bytes containing the value of @var{data} to the growing object in
1564 obstack @var{obstack-ptr}.
1569 @deftypefun void obstack_int_grow_fast (struct obstack *@var{obstack-ptr}, int @var{data})
1570 The function @code{obstack_int_grow_fast} adds @code{sizeof (int)} bytes
1571 containing the value of @var{data} to the growing object in obstack
1577 @deftypefun void obstack_blank_fast (struct obstack *@var{obstack-ptr}, int @var{size})
1578 The function @code{obstack_blank_fast} adds @var{size} bytes to the
1579 growing object in obstack @var{obstack-ptr} without initializing them.
1582 When you check for space using @code{obstack_room} and there is not
1583 enough room for what you want to add, the fast growth functions
1584 are not safe. In this case, simply use the corresponding ordinary
1585 growth function instead. Very soon this will copy the object to a
1586 new chunk; then there will be lots of room available again.
1588 So, each time you use an ordinary growth function, check afterward for
1589 sufficient space using @code{obstack_room}. Once the object is copied
1590 to a new chunk, there will be plenty of space again, so the program will
1591 start using the fast growth functions again.
1598 add_string (struct obstack *obstack, const char *ptr, int len)
1602 int room = obstack_room (obstack);
1605 /* @r{Not enough room. Add one character slowly,}
1606 @r{which may copy to a new chunk and make room.} */
1607 obstack_1grow (obstack, *ptr++);
1614 /* @r{Add fast as much as we have room for.} */
1617 obstack_1grow_fast (obstack, *ptr++);
1624 @node Status of an Obstack
1625 @subsection Status of an Obstack
1626 @cindex obstack status
1627 @cindex status of obstack
1629 Here are functions that provide information on the current status of
1630 allocation in an obstack. You can use them to learn about an object while
1635 @deftypefun {void *} obstack_base (struct obstack *@var{obstack-ptr})
1636 This function returns the tentative address of the beginning of the
1637 currently growing object in @var{obstack-ptr}. If you finish the object
1638 immediately, it will have that address. If you make it larger first, it
1639 may outgrow the current chunk---then its address will change!
1641 If no object is growing, this value says where the next object you
1642 allocate will start (once again assuming it fits in the current
1648 @deftypefun {void *} obstack_next_free (struct obstack *@var{obstack-ptr})
1649 This function returns the address of the first free byte in the current
1650 chunk of obstack @var{obstack-ptr}. This is the end of the currently
1651 growing object. If no object is growing, @code{obstack_next_free}
1652 returns the same value as @code{obstack_base}.
1657 @deftypefun int obstack_object_size (struct obstack *@var{obstack-ptr})
1658 This function returns the size in bytes of the currently growing object.
1659 This is equivalent to
1662 obstack_next_free (@var{obstack-ptr}) - obstack_base (@var{obstack-ptr})
1666 @node Obstacks Data Alignment
1667 @subsection Alignment of Data in Obstacks
1668 @cindex alignment (in obstacks)
1670 Each obstack has an @dfn{alignment boundary}; each object allocated in
1671 the obstack automatically starts on an address that is a multiple of the
1672 specified boundary. By default, this boundary is 4 bytes.
1674 To access an obstack's alignment boundary, use the macro
1675 @code{obstack_alignment_mask}, whose function prototype looks like
1680 @deftypefn Macro int obstack_alignment_mask (struct obstack *@var{obstack-ptr})
1681 The value is a bit mask; a bit that is 1 indicates that the corresponding
1682 bit in the address of an object should be 0. The mask value should be one
1683 less than a power of 2; the effect is that all object addresses are
1684 multiples of that power of 2. The default value of the mask is 3, so that
1685 addresses are multiples of 4. A mask value of 0 means an object can start
1686 on any multiple of 1 (that is, no alignment is required).
1688 The expansion of the macro @code{obstack_alignment_mask} is an lvalue,
1689 so you can alter the mask by assignment. For example, this statement:
1692 obstack_alignment_mask (obstack_ptr) = 0;
1696 has the effect of turning off alignment processing in the specified obstack.
1699 Note that a change in alignment mask does not take effect until
1700 @emph{after} the next time an object is allocated or finished in the
1701 obstack. If you are not growing an object, you can make the new
1702 alignment mask take effect immediately by calling @code{obstack_finish}.
1703 This will finish a zero-length object and then do proper alignment for
1706 @node Obstack Chunks
1707 @subsection Obstack Chunks
1708 @cindex efficiency of chunks
1711 Obstacks work by allocating space for themselves in large chunks, and
1712 then parceling out space in the chunks to satisfy your requests. Chunks
1713 are normally 4096 bytes long unless you specify a different chunk size.
1714 The chunk size includes 8 bytes of overhead that are not actually used
1715 for storing objects. Regardless of the specified size, longer chunks
1716 will be allocated when necessary for long objects.
1718 The obstack library allocates chunks by calling the function
1719 @code{obstack_chunk_alloc}, which you must define. When a chunk is no
1720 longer needed because you have freed all the objects in it, the obstack
1721 library frees the chunk by calling @code{obstack_chunk_free}, which you
1724 These two must be defined (as macros) or declared (as functions) in each
1725 source file that uses @code{obstack_init} (@pxref{Creating Obstacks}).
1726 Most often they are defined as macros like this:
1729 #define obstack_chunk_alloc malloc
1730 #define obstack_chunk_free free
1733 Note that these are simple macros (no arguments). Macro definitions with
1734 arguments will not work! It is necessary that @code{obstack_chunk_alloc}
1735 or @code{obstack_chunk_free}, alone, expand into a function name if it is
1736 not itself a function name.
1738 If you allocate chunks with @code{malloc}, the chunk size should be a
1739 power of 2. The default chunk size, 4096, was chosen because it is long
1740 enough to satisfy many typical requests on the obstack yet short enough
1741 not to waste too much memory in the portion of the last chunk not yet used.
1745 @deftypefn Macro int obstack_chunk_size (struct obstack *@var{obstack-ptr})
1746 This returns the chunk size of the given obstack.
1749 Since this macro expands to an lvalue, you can specify a new chunk size by
1750 assigning it a new value. Doing so does not affect the chunks already
1751 allocated, but will change the size of chunks allocated for that particular
1752 obstack in the future. It is unlikely to be useful to make the chunk size
1753 smaller, but making it larger might improve efficiency if you are
1754 allocating many objects whose size is comparable to the chunk size. Here
1755 is how to do so cleanly:
1758 if (obstack_chunk_size (obstack_ptr) < @var{new-chunk-size})
1759 obstack_chunk_size (obstack_ptr) = @var{new-chunk-size};
1762 @node Summary of Obstacks
1763 @subsection Summary of Obstack Functions
1765 Here is a summary of all the functions associated with obstacks. Each
1766 takes the address of an obstack (@code{struct obstack *}) as its first
1770 @item void obstack_init (struct obstack *@var{obstack-ptr})
1771 Initialize use of an obstack. @xref{Creating Obstacks}.
1773 @item void *obstack_alloc (struct obstack *@var{obstack-ptr}, int @var{size})
1774 Allocate an object of @var{size} uninitialized bytes.
1775 @xref{Allocation in an Obstack}.
1777 @item void *obstack_copy (struct obstack *@var{obstack-ptr}, void *@var{address}, int @var{size})
1778 Allocate an object of @var{size} bytes, with contents copied from
1779 @var{address}. @xref{Allocation in an Obstack}.
1781 @item void *obstack_copy0 (struct obstack *@var{obstack-ptr}, void *@var{address}, int @var{size})
1782 Allocate an object of @var{size}+1 bytes, with @var{size} of them copied
1783 from @var{address}, followed by a null character at the end.
1784 @xref{Allocation in an Obstack}.
1786 @item void obstack_free (struct obstack *@var{obstack-ptr}, void *@var{object})
1787 Free @var{object} (and everything allocated in the specified obstack
1788 more recently than @var{object}). @xref{Freeing Obstack Objects}.
1790 @item void obstack_blank (struct obstack *@var{obstack-ptr}, int @var{size})
1791 Add @var{size} uninitialized bytes to a growing object.
1792 @xref{Growing Objects}.
1794 @item void obstack_grow (struct obstack *@var{obstack-ptr}, void *@var{address}, int @var{size})
1795 Add @var{size} bytes, copied from @var{address}, to a growing object.
1796 @xref{Growing Objects}.
1798 @item void obstack_grow0 (struct obstack *@var{obstack-ptr}, void *@var{address}, int @var{size})
1799 Add @var{size} bytes, copied from @var{address}, to a growing object,
1800 and then add another byte containing a null character. @xref{Growing
1803 @item void obstack_1grow (struct obstack *@var{obstack-ptr}, char @var{data-char})
1804 Add one byte containing @var{data-char} to a growing object.
1805 @xref{Growing Objects}.
1807 @item void *obstack_finish (struct obstack *@var{obstack-ptr})
1808 Finalize the object that is growing and return its permanent address.
1809 @xref{Growing Objects}.
1811 @item int obstack_object_size (struct obstack *@var{obstack-ptr})
1812 Get the current size of the currently growing object. @xref{Growing
1815 @item void obstack_blank_fast (struct obstack *@var{obstack-ptr}, int @var{size})
1816 Add @var{size} uninitialized bytes to a growing object without checking
1817 that there is enough room. @xref{Extra Fast Growing}.
1819 @item void obstack_1grow_fast (struct obstack *@var{obstack-ptr}, char @var{data-char})
1820 Add one byte containing @var{data-char} to a growing object without
1821 checking that there is enough room. @xref{Extra Fast Growing}.
1823 @item int obstack_room (struct obstack *@var{obstack-ptr})
1824 Get the amount of room now available for growing the current object.
1825 @xref{Extra Fast Growing}.
1827 @item int obstack_alignment_mask (struct obstack *@var{obstack-ptr})
1828 The mask used for aligning the beginning of an object. This is an
1829 lvalue. @xref{Obstacks Data Alignment}.
1831 @item int obstack_chunk_size (struct obstack *@var{obstack-ptr})
1832 The size for allocating chunks. This is an lvalue. @xref{Obstack Chunks}.
1834 @item void *obstack_base (struct obstack *@var{obstack-ptr})
1835 Tentative starting address of the currently growing object.
1836 @xref{Status of an Obstack}.
1838 @item void *obstack_next_free (struct obstack *@var{obstack-ptr})
1839 Address just after the end of the currently growing object.
1840 @xref{Status of an Obstack}.
1843 @node Variable Size Automatic
1844 @section Automatic Storage with Variable Size
1845 @cindex automatic freeing
1846 @cindex @code{alloca} function
1847 @cindex automatic storage with variable size
1849 The function @code{alloca} supports a kind of half-dynamic allocation in
1850 which blocks are allocated dynamically but freed automatically.
1852 Allocating a block with @code{alloca} is an explicit action; you can
1853 allocate as many blocks as you wish, and compute the size at run time. But
1854 all the blocks are freed when you exit the function that @code{alloca} was
1855 called from, just as if they were automatic variables declared in that
1856 function. There is no way to free the space explicitly.
1858 The prototype for @code{alloca} is in @file{stdlib.h}. This function is
1864 @deftypefun {void *} alloca (size_t @var{size});
1865 The return value of @code{alloca} is the address of a block of @var{size}
1866 bytes of storage, allocated in the stack frame of the calling function.
1869 Do not use @code{alloca} inside the arguments of a function call---you
1870 will get unpredictable results, because the stack space for the
1871 @code{alloca} would appear on the stack in the middle of the space for
1872 the function arguments. An example of what to avoid is @code{foo (x,
1874 @c This might get fixed in future versions of GCC, but that won't make
1875 @c it safe with compilers generally.
1878 * Alloca Example:: Example of using @code{alloca}.
1879 * Advantages of Alloca:: Reasons to use @code{alloca}.
1880 * Disadvantages of Alloca:: Reasons to avoid @code{alloca}.
1881 * GNU C Variable-Size Arrays:: Only in GNU C, here is an alternative
1882 method of allocating dynamically and
1883 freeing automatically.
1886 @node Alloca Example
1887 @subsection @code{alloca} Example
1889 As an example of use of @code{alloca}, here is a function that opens a file
1890 name made from concatenating two argument strings, and returns a file
1891 descriptor or minus one signifying failure:
1895 open2 (char *str1, char *str2, int flags, int mode)
1897 char *name = (char *) alloca (strlen (str1) + strlen (str2) + 1);
1898 stpcpy (stpcpy (name, str1), str2);
1899 return open (name, flags, mode);
1904 Here is how you would get the same results with @code{malloc} and
1909 open2 (char *str1, char *str2, int flags, int mode)
1911 char *name = (char *) malloc (strlen (str1) + strlen (str2) + 1);
1914 fatal ("virtual memory exceeded");
1915 stpcpy (stpcpy (name, str1), str2);
1916 desc = open (name, flags, mode);
1922 As you can see, it is simpler with @code{alloca}. But @code{alloca} has
1923 other, more important advantages, and some disadvantages.
1925 @node Advantages of Alloca
1926 @subsection Advantages of @code{alloca}
1928 Here are the reasons why @code{alloca} may be preferable to @code{malloc}:
1932 Using @code{alloca} wastes very little space and is very fast. (It is
1933 open-coded by the GNU C compiler.)
1936 Since @code{alloca} does not have separate pools for different sizes of
1937 block, space used for any size block can be reused for any other size.
1938 @code{alloca} does not cause storage fragmentation.
1942 Nonlocal exits done with @code{longjmp} (@pxref{Non-Local Exits})
1943 automatically free the space allocated with @code{alloca} when they exit
1944 through the function that called @code{alloca}. This is the most
1945 important reason to use @code{alloca}.
1947 To illustrate this, suppose you have a function
1948 @code{open_or_report_error} which returns a descriptor, like
1949 @code{open}, if it succeeds, but does not return to its caller if it
1950 fails. If the file cannot be opened, it prints an error message and
1951 jumps out to the command level of your program using @code{longjmp}.
1952 Let's change @code{open2} (@pxref{Alloca Example}) to use this
1957 open2 (char *str1, char *str2, int flags, int mode)
1959 char *name = (char *) alloca (strlen (str1) + strlen (str2) + 1);
1960 stpcpy (stpcpy (name, str1), str2);
1961 return open_or_report_error (name, flags, mode);
1966 Because of the way @code{alloca} works, the storage it allocates is
1967 freed even when an error occurs, with no special effort required.
1969 By contrast, the previous definition of @code{open2} (which uses
1970 @code{malloc} and @code{free}) would develop a storage leak if it were
1971 changed in this way. Even if you are willing to make more changes to
1972 fix it, there is no easy way to do so.
1975 @node Disadvantages of Alloca
1976 @subsection Disadvantages of @code{alloca}
1978 @cindex @code{alloca} disadvantages
1979 @cindex disadvantages of @code{alloca}
1980 These are the disadvantages of @code{alloca} in comparison with
1985 If you try to allocate more storage than the machine can provide, you
1986 don't get a clean error message. Instead you get a fatal signal like
1987 the one you would get from an infinite recursion; probably a
1988 segmentation violation (@pxref{Program Error Signals}).
1991 Some non-GNU systems fail to support @code{alloca}, so it is less
1992 portable. However, a slower emulation of @code{alloca} written in C
1993 is available for use on systems with this deficiency.
1996 @node GNU C Variable-Size Arrays
1997 @subsection GNU C Variable-Size Arrays
1998 @cindex variable-sized arrays
2000 In GNU C, you can replace most uses of @code{alloca} with an array of
2001 variable size. Here is how @code{open2} would look then:
2004 int open2 (char *str1, char *str2, int flags, int mode)
2006 char name[strlen (str1) + strlen (str2) + 1];
2007 stpcpy (stpcpy (name, str1), str2);
2008 return open (name, flags, mode);
2012 But @code{alloca} is not always equivalent to a variable-sized array, for
2017 A variable size array's space is freed at the end of the scope of the
2018 name of the array. The space allocated with @code{alloca}
2019 remains until the end of the function.
2022 It is possible to use @code{alloca} within a loop, allocating an
2023 additional block on each iteration. This is impossible with
2024 variable-sized arrays.
2027 @strong{Note:} If you mix use of @code{alloca} and variable-sized arrays
2028 within one function, exiting a scope in which a variable-sized array was
2029 declared frees all blocks allocated with @code{alloca} during the
2030 execution of that scope.
2033 @node Relocating Allocator
2034 @section Relocating Allocator
2036 @cindex relocating memory allocator
2037 Any system of dynamic memory allocation has overhead: the amount of
2038 space it uses is more than the amount the program asks for. The
2039 @dfn{relocating memory allocator} achieves very low overhead by moving
2040 blocks in memory as necessary, on its own initiative.
2043 * Relocator Concepts:: How to understand relocating allocation.
2044 * Using Relocator:: Functions for relocating allocation.
2047 @node Relocator Concepts
2048 @subsection Concepts of Relocating Allocation
2051 The @dfn{relocating memory allocator} achieves very low overhead by
2052 moving blocks in memory as necessary, on its own initiative.
2055 When you allocate a block with @code{malloc}, the address of the block
2056 never changes unless you use @code{realloc} to change its size. Thus,
2057 you can safely store the address in various places, temporarily or
2058 permanently, as you like. This is not safe when you use the relocating
2059 memory allocator, because any and all relocatable blocks can move
2060 whenever you allocate memory in any fashion. Even calling @code{malloc}
2061 or @code{realloc} can move the relocatable blocks.
2064 For each relocatable block, you must make a @dfn{handle}---a pointer
2065 object in memory, designated to store the address of that block. The
2066 relocating allocator knows where each block's handle is, and updates the
2067 address stored there whenever it moves the block, so that the handle
2068 always points to the block. Each time you access the contents of the
2069 block, you should fetch its address anew from the handle.
2071 To call any of the relocating allocator functions from a signal handler
2072 is almost certainly incorrect, because the signal could happen at any
2073 time and relocate all the blocks. The only way to make this safe is to
2074 block the signal around any access to the contents of any relocatable
2075 block---not a convenient mode of operation. @xref{Nonreentrancy}.
2077 @node Using Relocator
2078 @subsection Allocating and Freeing Relocatable Blocks
2081 In the descriptions below, @var{handleptr} designates the address of the
2082 handle. All the functions are declared in @file{malloc.h}; all are GNU
2087 @deftypefun {void *} r_alloc (void **@var{handleptr}, size_t @var{size})
2088 This function allocates a relocatable block of size @var{size}. It
2089 stores the block's address in @code{*@var{handleptr}} and returns
2090 a non-null pointer to indicate success.
2092 If @code{r_alloc} can't get the space needed, it stores a null pointer
2093 in @code{*@var{handleptr}}, and returns a null pointer.
2098 @deftypefun void r_alloc_free (void **@var{handleptr})
2099 This function is the way to free a relocatable block. It frees the
2100 block that @code{*@var{handleptr}} points to, and stores a null pointer
2101 in @code{*@var{handleptr}} to show it doesn't point to an allocated
2107 @deftypefun {void *} r_re_alloc (void **@var{handleptr}, size_t @var{size})
2108 The function @code{r_re_alloc} adjusts the size of the block that
2109 @code{*@var{handleptr}} points to, making it @var{size} bytes long. It
2110 stores the address of the resized block in @code{*@var{handleptr}} and
2111 returns a non-null pointer to indicate success.
2113 If enough memory is not available, this function returns a null pointer
2114 and does not modify @code{*@var{handleptr}}.
2118 @comment No longer available...
2120 @comment @node Memory Warnings
2121 @comment @section Memory Usage Warnings
2122 @comment @cindex memory usage warnings
2123 @comment @cindex warnings of memory almost full
2126 You can ask for warnings as the program approaches running out of memory
2127 space, by calling @code{memory_warnings}. This tells @code{malloc} to
2128 check memory usage every time it asks for more memory from the operating
2129 system. This is a GNU extension declared in @file{malloc.h}.
2133 @comment @deftypefun void memory_warnings (void *@var{start}, void (*@var{warn-func}) (const char *))
2134 Call this function to request warnings for nearing exhaustion of virtual
2137 The argument @var{start} says where data space begins, in memory. The
2138 allocator compares this against the last address used and against the
2139 limit of data space, to determine the fraction of available memory in
2140 use. If you supply zero for @var{start}, then a default value is used
2141 which is right in most circumstances.
2143 For @var{warn-func}, supply a function that @code{malloc} can call to
2144 warn you. It is called with a string (a warning message) as argument.
2145 Normally it ought to display the string for the user to read.
2148 The warnings come when memory becomes 75% full, when it becomes 85%
2149 full, and when it becomes 95% full. Above 95% you get another warning
2150 each time memory usage increases.