]>
Commit | Line | Data |
---|---|---|
99a20616 UD |
1 | @node Memory, Character Handling, Error Reporting, Top |
2 | @chapter Virtual Memory Allocation And Paging | |
3 | @c %MENU% Allocating virtual memory and controlling paging | |
28f540f4 RM |
4 | @cindex memory allocation |
5 | @cindex storage allocation | |
6 | ||
99a20616 | 7 | This chapter describes how processes manage and use memory in a system |
1f77f049 | 8 | that uses @theglibc{}. |
99a20616 | 9 | |
1f77f049 | 10 | @Theglibc{} has several functions for dynamically allocating |
99a20616 UD |
11 | virtual memory in various ways. They vary in generality and in |
12 | efficiency. The library also provides functions for controlling paging | |
13 | and allocation of real memory. | |
14 | ||
15 | ||
16 | @menu | |
17 | * Memory Concepts:: An introduction to concepts and terminology. | |
18 | * Memory Allocation:: Allocating storage for your program data | |
99a20616 | 19 | * Resizing the Data Segment:: @code{brk}, @code{sbrk} |
4c23fed5 | 20 | * Locking Pages:: Preventing page faults |
99a20616 UD |
21 | @end menu |
22 | ||
23 | Memory mapped I/O is not discussed in this chapter. @xref{Memory-mapped I/O}. | |
24 | ||
25 | ||
26 | ||
27 | @node Memory Concepts | |
28 | @section Process Memory Concepts | |
29 | ||
30 | One of the most basic resources a process has available to it is memory. | |
31 | There are a lot of different ways systems organize memory, but in a | |
32 | typical one, each process has one linear virtual address space, with | |
33 | addresses running from zero to some huge maximum. It need not be | |
11bf311e | 34 | contiguous; i.e., not all of these addresses actually can be used to |
99a20616 UD |
35 | store data. |
36 | ||
37 | The virtual memory is divided into pages (4 kilobytes is typical). | |
38 | Backing each page of virtual memory is a page of real memory (called a | |
39 | @dfn{frame}) or some secondary storage, usually disk space. The disk | |
40 | space might be swap space or just some ordinary disk file. Actually, a | |
41 | page of all zeroes sometimes has nothing at all backing it -- there's | |
42 | just a flag saying it is all zeroes. | |
43 | @cindex page frame | |
44 | @cindex frame, real memory | |
45 | @cindex swap space | |
46 | @cindex page, virtual memory | |
47 | ||
48 | The same frame of real memory or backing store can back multiple virtual | |
49 | pages belonging to multiple processes. This is normally the case, for | |
1f77f049 | 50 | example, with virtual memory occupied by @glibcadj{} code. The same |
99a20616 UD |
51 | real memory frame containing the @code{printf} function backs a virtual |
52 | memory page in each of the existing processes that has a @code{printf} | |
53 | call in its program. | |
54 | ||
55 | In order for a program to access any part of a virtual page, the page | |
56 | must at that moment be backed by (``connected to'') a real frame. But | |
57 | because there is usually a lot more virtual memory than real memory, the | |
58 | pages must move back and forth between real memory and backing store | |
59 | regularly, coming into real memory when a process needs to access them | |
60 | and then retreating to backing store when not needed anymore. This | |
61 | movement is called @dfn{paging}. | |
62 | ||
63 | When a program attempts to access a page which is not at that moment | |
64 | backed by real memory, this is known as a @dfn{page fault}. When a page | |
65 | fault occurs, the kernel suspends the process, places the page into a | |
66 | real page frame (this is called ``paging in'' or ``faulting in''), then | |
67 | resumes the process so that from the process' point of view, the page | |
68 | was in real memory all along. In fact, to the process, all pages always | |
69 | seem to be in real memory. Except for one thing: the elapsed execution | |
70 | time of an instruction that would normally be a few nanoseconds is | |
71 | suddenly much, much, longer (because the kernel normally has to do I/O | |
72 | to complete the page-in). For programs sensitive to that, the functions | |
73 | described in @ref{Locking Pages} can control it. | |
74 | @cindex page fault | |
75 | @cindex paging | |
76 | ||
77 | Within each virtual address space, a process has to keep track of what | |
78 | is at which addresses, and that process is called memory allocation. | |
79 | Allocation usually brings to mind meting out scarce resources, but in | |
80 | the case of virtual memory, that's not a major goal, because there is | |
81 | generally much more of it than anyone needs. Memory allocation within a | |
68979757 | 82 | process is mainly just a matter of making sure that the same byte of |
99a20616 UD |
83 | memory isn't used to store two different things. |
84 | ||
85 | Processes allocate memory in two major ways: by exec and | |
86 | programmatically. Actually, forking is a third way, but it's not very | |
87 | interesting. @xref{Creating a Process}. | |
88 | ||
89 | Exec is the operation of creating a virtual address space for a process, | |
90 | loading its basic program into it, and executing the program. It is | |
91 | done by the ``exec'' family of functions (e.g. @code{execl}). The | |
92 | operation takes a program file (an executable), it allocates space to | |
93 | load all the data in the executable, loads it, and transfers control to | |
94 | it. That data is most notably the instructions of the program (the | |
95 | @dfn{text}), but also literals and constants in the program and even | |
96 | some variables: C variables with the static storage class (@pxref{Memory | |
97 | Allocation and C}). | |
98 | @cindex executable | |
99 | @cindex literals | |
100 | @cindex constants | |
101 | ||
102 | Once that program begins to execute, it uses programmatic allocation to | |
1f77f049 | 103 | gain additional memory. In a C program with @theglibc{}, there |
99a20616 UD |
104 | are two kinds of programmatic allocation: automatic and dynamic. |
105 | @xref{Memory Allocation and C}. | |
106 | ||
107 | Memory-mapped I/O is another form of dynamic virtual memory allocation. | |
108 | Mapping memory to a file means declaring that the contents of certain | |
109 | range of a process' addresses shall be identical to the contents of a | |
110 | specified regular file. The system makes the virtual memory initially | |
111 | contain the contents of the file, and if you modify the memory, the | |
112 | system writes the same modification to the file. Note that due to the | |
113 | magic of virtual memory and page faults, there is no reason for the | |
114 | system to do I/O to read the file, or allocate real memory for its | |
115 | contents, until the program accesses the virtual memory. | |
116 | @xref{Memory-mapped I/O}. | |
117 | @cindex memory mapped I/O | |
118 | @cindex memory mapped file | |
119 | @cindex files, accessing | |
120 | ||
121 | Just as it programmatically allocates memory, the program can | |
122 | programmatically deallocate (@dfn{free}) it. You can't free the memory | |
123 | that was allocated by exec. When the program exits or execs, you might | |
124 | say that all its memory gets freed, but since in both cases the address | |
125 | space ceases to exist, the point is really moot. @xref{Program | |
126 | Termination}. | |
127 | @cindex execing a program | |
128 | @cindex freeing memory | |
129 | @cindex exiting a program | |
130 | ||
131 | A process' virtual address space is divided into segments. A segment is | |
132 | a contiguous range of virtual addresses. Three important segments are: | |
28f540f4 | 133 | |
28f540f4 | 134 | @itemize @bullet |
28f540f4 | 135 | |
68979757 | 136 | @item |
99a20616 UD |
137 | |
138 | The @dfn{text segment} contains a program's instructions and literals and | |
139 | static constants. It is allocated by exec and stays the same size for | |
68979757 | 140 | the life of the virtual address space. |
28f540f4 RM |
141 | |
142 | @item | |
99a20616 UD |
143 | The @dfn{data segment} is working storage for the program. It can be |
144 | preallocated and preloaded by exec and the process can extend or shrink | |
145 | it by calling functions as described in @xref{Resizing the Data | |
146 | Segment}. Its lower end is fixed. | |
147 | ||
68979757 | 148 | @item |
99a20616 UD |
149 | The @dfn{stack segment} contains a program stack. It grows as the stack |
150 | grows, but doesn't shrink when the stack shrinks. | |
151 | ||
28f540f4 | 152 | @end itemize |
99a20616 UD |
153 | |
154 | ||
155 | ||
156 | @node Memory Allocation | |
68979757 | 157 | @section Allocating Storage For Program Data |
99a20616 UD |
158 | |
159 | This section covers how ordinary programs manage storage for their data, | |
160 | including the famous @code{malloc} function and some fancier facilities | |
1f77f049 | 161 | special @theglibc{} and GNU Compiler. |
28f540f4 RM |
162 | |
163 | @menu | |
99a20616 | 164 | * Memory Allocation and C:: How to get different kinds of allocation in C. |
28f540f4 RM |
165 | * Unconstrained Allocation:: The @code{malloc} facility allows fully general |
166 | dynamic allocation. | |
bd355af0 | 167 | * Allocation Debugging:: Finding memory leaks and not freed memory. |
28f540f4 RM |
168 | * Obstacks:: Obstacks are less general than malloc |
169 | but more efficient and convenient. | |
170 | * Variable Size Automatic:: Allocation of variable-sized blocks | |
171 | of automatic storage that are freed when the | |
172 | calling function returns. | |
28f540f4 RM |
173 | @end menu |
174 | ||
28f540f4 | 175 | |
99a20616 UD |
176 | @node Memory Allocation and C |
177 | @subsection Memory Allocation in C Programs | |
28f540f4 | 178 | |
99a20616 UD |
179 | The C language supports two kinds of memory allocation through the |
180 | variables in C programs: | |
28f540f4 RM |
181 | |
182 | @itemize @bullet | |
183 | @item | |
184 | @dfn{Static allocation} is what happens when you declare a static or | |
185 | global variable. Each static or global variable defines one block of | |
186 | space, of a fixed size. The space is allocated once, when your program | |
99a20616 UD |
187 | is started (part of the exec operation), and is never freed. |
188 | @cindex static memory allocation | |
189 | @cindex static storage class | |
28f540f4 RM |
190 | |
191 | @item | |
192 | @dfn{Automatic allocation} happens when you declare an automatic | |
193 | variable, such as a function argument or a local variable. The space | |
194 | for an automatic variable is allocated when the compound statement | |
195 | containing the declaration is entered, and is freed when that | |
196 | compound statement is exited. | |
99a20616 UD |
197 | @cindex automatic memory allocation |
198 | @cindex automatic storage class | |
28f540f4 | 199 | |
99a20616 | 200 | In GNU C, the size of the automatic storage can be an expression |
28f540f4 RM |
201 | that varies. In other C implementations, it must be a constant. |
202 | @end itemize | |
203 | ||
99a20616 | 204 | A third important kind of memory allocation, @dfn{dynamic allocation}, |
1f77f049 | 205 | is not supported by C variables but is available via @glibcadj{} |
99a20616 UD |
206 | functions. |
207 | @cindex dynamic memory allocation | |
208 | ||
209 | @subsubsection Dynamic Memory Allocation | |
210 | @cindex dynamic memory allocation | |
211 | ||
212 | @dfn{Dynamic memory allocation} is a technique in which programs | |
213 | determine as they are running where to store some information. You need | |
214 | dynamic allocation when the amount of memory you need, or how long you | |
215 | continue to need it, depends on factors that are not known before the | |
216 | program runs. | |
217 | ||
218 | For example, you may need a block to store a line read from an input | |
219 | file; since there is no limit to how long a line can be, you must | |
220 | allocate the memory dynamically and make it dynamically larger as you | |
221 | read more of the line. | |
222 | ||
223 | Or, you may need a block for each record or each definition in the input | |
224 | data; since you can't know in advance how many there will be, you must | |
225 | allocate a new block for each record or definition as you read it. | |
226 | ||
227 | When you use dynamic allocation, the allocation of a block of memory is | |
228 | an action that the program requests explicitly. You call a function or | |
229 | macro when you want to allocate space, and specify the size with an | |
230 | argument. If you want to free the space, you do so by calling another | |
231 | function or macro. You can do these things whenever you want, as often | |
232 | as you want. | |
233 | ||
28f540f4 RM |
234 | Dynamic allocation is not supported by C variables; there is no storage |
235 | class ``dynamic'', and there can never be a C variable whose value is | |
99a20616 | 236 | stored in dynamically allocated space. The only way to get dynamically |
1f77f049 JM |
237 | allocated memory is via a system call (which is generally via a @glibcadj{} |
238 | function call), and the only way to refer to dynamically | |
99a20616 UD |
239 | allocated space is through a pointer. Because it is less convenient, |
240 | and because the actual process of dynamic allocation requires more | |
241 | computation time, programmers generally use dynamic allocation only when | |
242 | neither static nor automatic allocation will serve. | |
28f540f4 RM |
243 | |
244 | For example, if you want to allocate dynamically some space to hold a | |
245 | @code{struct foobar}, you cannot declare a variable of type @code{struct | |
246 | foobar} whose contents are the dynamically allocated space. But you can | |
247 | declare a variable of pointer type @code{struct foobar *} and assign it the | |
248 | address of the space. Then you can use the operators @samp{*} and | |
249 | @samp{->} on this pointer variable to refer to the contents of the space: | |
250 | ||
251 | @smallexample | |
252 | @{ | |
253 | struct foobar *ptr | |
254 | = (struct foobar *) malloc (sizeof (struct foobar)); | |
255 | ptr->name = x; | |
256 | ptr->next = current_foobar; | |
257 | current_foobar = ptr; | |
258 | @} | |
259 | @end smallexample | |
260 | ||
261 | @node Unconstrained Allocation | |
99a20616 UD |
262 | @subsection Unconstrained Allocation |
263 | @cindex unconstrained memory allocation | |
28f540f4 RM |
264 | @cindex @code{malloc} function |
265 | @cindex heap, dynamic allocation from | |
266 | ||
267 | The most general dynamic allocation facility is @code{malloc}. It | |
268 | allows you to allocate blocks of memory of any size at any time, make | |
269 | them bigger or smaller at any time, and free the blocks individually at | |
270 | any time (or never). | |
271 | ||
272 | @menu | |
273 | * Basic Allocation:: Simple use of @code{malloc}. | |
274 | * Malloc Examples:: Examples of @code{malloc}. @code{xmalloc}. | |
275 | * Freeing after Malloc:: Use @code{free} to free a block you | |
276 | got with @code{malloc}. | |
277 | * Changing Block Size:: Use @code{realloc} to make a block | |
278 | bigger or smaller. | |
279 | * Allocating Cleared Space:: Use @code{calloc} to allocate a | |
280 | block and clear it. | |
281 | * Efficiency and Malloc:: Efficiency considerations in use of | |
282 | these functions. | |
68979757 | 283 | * Aligned Memory Blocks:: Allocating specially aligned memory. |
c131718c UD |
284 | * Malloc Tunable Parameters:: Use @code{mallopt} to adjust allocation |
285 | parameters. | |
28f540f4 RM |
286 | * Heap Consistency Checking:: Automatic checking for errors. |
287 | * Hooks for Malloc:: You can use these hooks for debugging | |
288 | programs that use @code{malloc}. | |
289 | * Statistics of Malloc:: Getting information about how much | |
290 | memory your program is using. | |
291 | * Summary of Malloc:: Summary of @code{malloc} and related functions. | |
292 | @end menu | |
293 | ||
294 | @node Basic Allocation | |
99a20616 | 295 | @subsubsection Basic Memory Allocation |
28f540f4 RM |
296 | @cindex allocation of memory with @code{malloc} |
297 | ||
298 | To allocate a block of memory, call @code{malloc}. The prototype for | |
299 | this function is in @file{stdlib.h}. | |
300 | @pindex stdlib.h | |
301 | ||
302 | @comment malloc.h stdlib.h | |
f65fd747 | 303 | @comment ISO |
28f540f4 RM |
304 | @deftypefun {void *} malloc (size_t @var{size}) |
305 | This function returns a pointer to a newly allocated block @var{size} | |
306 | bytes long, or a null pointer if the block could not be allocated. | |
307 | @end deftypefun | |
308 | ||
309 | The contents of the block are undefined; you must initialize it yourself | |
310 | (or use @code{calloc} instead; @pxref{Allocating Cleared Space}). | |
311 | Normally you would cast the value as a pointer to the kind of object | |
312 | that you want to store in the block. Here we show an example of doing | |
313 | so, and of initializing the space with zeros using the library function | |
314 | @code{memset} (@pxref{Copying and Concatenation}): | |
315 | ||
316 | @smallexample | |
317 | struct foo *ptr; | |
318 | @dots{} | |
319 | ptr = (struct foo *) malloc (sizeof (struct foo)); | |
320 | if (ptr == 0) abort (); | |
321 | memset (ptr, 0, sizeof (struct foo)); | |
322 | @end smallexample | |
323 | ||
324 | You can store the result of @code{malloc} into any pointer variable | |
f65fd747 | 325 | without a cast, because @w{ISO C} automatically converts the type |
28f540f4 RM |
326 | @code{void *} to another type of pointer when necessary. But the cast |
327 | is necessary in contexts other than assignment operators or if you might | |
328 | want your code to run in traditional C. | |
329 | ||
330 | Remember that when allocating space for a string, the argument to | |
331 | @code{malloc} must be one plus the length of the string. This is | |
332 | because a string is terminated with a null character that doesn't count | |
333 | in the ``length'' of the string but does need space. For example: | |
334 | ||
335 | @smallexample | |
336 | char *ptr; | |
337 | @dots{} | |
338 | ptr = (char *) malloc (length + 1); | |
339 | @end smallexample | |
340 | ||
341 | @noindent | |
342 | @xref{Representation of Strings}, for more information about this. | |
343 | ||
344 | @node Malloc Examples | |
99a20616 | 345 | @subsubsection Examples of @code{malloc} |
28f540f4 RM |
346 | |
347 | If no more space is available, @code{malloc} returns a null pointer. | |
348 | You should check the value of @emph{every} call to @code{malloc}. It is | |
349 | useful to write a subroutine that calls @code{malloc} and reports an | |
350 | error if the value is a null pointer, returning only if the value is | |
351 | nonzero. This function is conventionally called @code{xmalloc}. Here | |
352 | it is: | |
353 | ||
354 | @smallexample | |
355 | void * | |
356 | xmalloc (size_t size) | |
357 | @{ | |
358 | register void *value = malloc (size); | |
359 | if (value == 0) | |
360 | fatal ("virtual memory exhausted"); | |
361 | return value; | |
362 | @} | |
363 | @end smallexample | |
364 | ||
365 | Here is a real example of using @code{malloc} (by way of @code{xmalloc}). | |
366 | The function @code{savestring} will copy a sequence of characters into | |
367 | a newly allocated null-terminated string: | |
368 | ||
369 | @smallexample | |
370 | @group | |
371 | char * | |
372 | savestring (const char *ptr, size_t len) | |
373 | @{ | |
374 | register char *value = (char *) xmalloc (len + 1); | |
28f540f4 | 375 | value[len] = '\0'; |
390955cb | 376 | return (char *) memcpy (value, ptr, len); |
28f540f4 RM |
377 | @} |
378 | @end group | |
379 | @end smallexample | |
380 | ||
381 | The block that @code{malloc} gives you is guaranteed to be aligned so | |
a7a93d50 | 382 | that it can hold any type of data. On @gnusystems{}, the address is |
c131718c UD |
383 | always a multiple of eight on most systems, and a multiple of 16 on |
384 | 64-bit systems. Only rarely is any higher boundary (such as a page | |
68979757 UD |
385 | boundary) necessary; for those cases, use @code{memalign}, |
386 | @code{posix_memalign} or @code{valloc} (@pxref{Aligned Memory Blocks}). | |
28f540f4 RM |
387 | |
388 | Note that the memory located after the end of the block is likely to be | |
389 | in use for something else; perhaps a block already allocated by another | |
390 | call to @code{malloc}. If you attempt to treat the block as longer than | |
391 | you asked for it to be, you are liable to destroy the data that | |
392 | @code{malloc} uses to keep track of its blocks, or you may destroy the | |
393 | contents of another block. If you have already allocated a block and | |
394 | discover you want it to be bigger, use @code{realloc} (@pxref{Changing | |
395 | Block Size}). | |
396 | ||
397 | @node Freeing after Malloc | |
99a20616 | 398 | @subsubsection Freeing Memory Allocated with @code{malloc} |
28f540f4 RM |
399 | @cindex freeing memory allocated with @code{malloc} |
400 | @cindex heap, freeing memory from | |
401 | ||
402 | When you no longer need a block that you got with @code{malloc}, use the | |
403 | function @code{free} to make the block available to be allocated again. | |
404 | The prototype for this function is in @file{stdlib.h}. | |
405 | @pindex stdlib.h | |
406 | ||
407 | @comment malloc.h stdlib.h | |
f65fd747 | 408 | @comment ISO |
28f540f4 | 409 | @deftypefun void free (void *@var{ptr}) |
99a20616 | 410 | The @code{free} function deallocates the block of memory pointed at |
28f540f4 RM |
411 | by @var{ptr}. |
412 | @end deftypefun | |
413 | ||
414 | @comment stdlib.h | |
415 | @comment Sun | |
416 | @deftypefun void cfree (void *@var{ptr}) | |
417 | This function does the same thing as @code{free}. It's provided for | |
418 | backward compatibility with SunOS; you should use @code{free} instead. | |
419 | @end deftypefun | |
420 | ||
421 | Freeing a block alters the contents of the block. @strong{Do not expect to | |
422 | find any data (such as a pointer to the next block in a chain of blocks) in | |
423 | the block after freeing it.} Copy whatever you need out of the block before | |
424 | freeing it! Here is an example of the proper way to free all the blocks in | |
425 | a chain, and the strings that they point to: | |
426 | ||
427 | @smallexample | |
428 | struct chain | |
429 | @{ | |
430 | struct chain *next; | |
431 | char *name; | |
432 | @} | |
433 | ||
434 | void | |
435 | free_chain (struct chain *chain) | |
436 | @{ | |
437 | while (chain != 0) | |
438 | @{ | |
439 | struct chain *next = chain->next; | |
440 | free (chain->name); | |
441 | free (chain); | |
442 | chain = next; | |
443 | @} | |
444 | @} | |
445 | @end smallexample | |
446 | ||
447 | Occasionally, @code{free} can actually return memory to the operating | |
448 | system and make the process smaller. Usually, all it can do is allow a | |
449 | later call to @code{malloc} to reuse the space. In the meantime, the | |
450 | space remains in your program as part of a free-list used internally by | |
451 | @code{malloc}. | |
452 | ||
453 | There is no point in freeing blocks at the end of a program, because all | |
454 | of the program's space is given back to the system when the process | |
455 | terminates. | |
456 | ||
457 | @node Changing Block Size | |
99a20616 | 458 | @subsubsection Changing the Size of a Block |
28f540f4 RM |
459 | @cindex changing the size of a block (@code{malloc}) |
460 | ||
461 | Often you do not know for certain how big a block you will ultimately need | |
462 | at the time you must begin to use the block. For example, the block might | |
463 | be a buffer that you use to hold a line being read from a file; no matter | |
464 | how long you make the buffer initially, you may encounter a line that is | |
465 | longer. | |
466 | ||
467 | You can make the block longer by calling @code{realloc}. This function | |
468 | is declared in @file{stdlib.h}. | |
469 | @pindex stdlib.h | |
470 | ||
471 | @comment malloc.h stdlib.h | |
f65fd747 | 472 | @comment ISO |
28f540f4 RM |
473 | @deftypefun {void *} realloc (void *@var{ptr}, size_t @var{newsize}) |
474 | The @code{realloc} function changes the size of the block whose address is | |
475 | @var{ptr} to be @var{newsize}. | |
476 | ||
477 | Since the space after the end of the block may be in use, @code{realloc} | |
478 | may find it necessary to copy the block to a new address where more free | |
479 | space is available. The value of @code{realloc} is the new address of the | |
480 | block. If the block needs to be moved, @code{realloc} copies the old | |
481 | contents. | |
482 | ||
483 | If you pass a null pointer for @var{ptr}, @code{realloc} behaves just | |
484 | like @samp{malloc (@var{newsize})}. This can be convenient, but beware | |
f65fd747 | 485 | that older implementations (before @w{ISO C}) may not support this |
28f540f4 RM |
486 | behavior, and will probably crash when @code{realloc} is passed a null |
487 | pointer. | |
488 | @end deftypefun | |
489 | ||
490 | Like @code{malloc}, @code{realloc} may return a null pointer if no | |
491 | memory space is available to make the block bigger. When this happens, | |
492 | the original block is untouched; it has not been modified or relocated. | |
493 | ||
494 | In most cases it makes no difference what happens to the original block | |
495 | when @code{realloc} fails, because the application program cannot continue | |
496 | when it is out of memory, and the only thing to do is to give a fatal error | |
497 | message. Often it is convenient to write and use a subroutine, | |
498 | conventionally called @code{xrealloc}, that takes care of the error message | |
499 | as @code{xmalloc} does for @code{malloc}: | |
500 | ||
501 | @smallexample | |
502 | void * | |
503 | xrealloc (void *ptr, size_t size) | |
504 | @{ | |
505 | register void *value = realloc (ptr, size); | |
506 | if (value == 0) | |
507 | fatal ("Virtual memory exhausted"); | |
508 | return value; | |
509 | @} | |
510 | @end smallexample | |
511 | ||
512 | You can also use @code{realloc} to make a block smaller. The reason you | |
ed277b4e | 513 | would do this is to avoid tying up a lot of memory space when only a little |
c131718c UD |
514 | is needed. |
515 | @comment The following is no longer true with the new malloc. | |
516 | @comment But it seems wise to keep the warning for other implementations. | |
517 | In several allocation implementations, making a block smaller sometimes | |
518 | necessitates copying it, so it can fail if no other space is available. | |
28f540f4 RM |
519 | |
520 | If the new size you specify is the same as the old size, @code{realloc} | |
521 | is guaranteed to change nothing and return the same address that you gave. | |
522 | ||
523 | @node Allocating Cleared Space | |
99a20616 | 524 | @subsubsection Allocating Cleared Space |
28f540f4 RM |
525 | |
526 | The function @code{calloc} allocates memory and clears it to zero. It | |
527 | is declared in @file{stdlib.h}. | |
528 | @pindex stdlib.h | |
529 | ||
530 | @comment malloc.h stdlib.h | |
f65fd747 | 531 | @comment ISO |
28f540f4 RM |
532 | @deftypefun {void *} calloc (size_t @var{count}, size_t @var{eltsize}) |
533 | This function allocates a block long enough to contain a vector of | |
534 | @var{count} elements, each of size @var{eltsize}. Its contents are | |
535 | cleared to zero before @code{calloc} returns. | |
536 | @end deftypefun | |
537 | ||
538 | You could define @code{calloc} as follows: | |
539 | ||
540 | @smallexample | |
541 | void * | |
542 | calloc (size_t count, size_t eltsize) | |
543 | @{ | |
544 | size_t size = count * eltsize; | |
545 | void *value = malloc (size); | |
546 | if (value != 0) | |
547 | memset (value, 0, size); | |
548 | return value; | |
549 | @} | |
550 | @end smallexample | |
551 | ||
c131718c UD |
552 | But in general, it is not guaranteed that @code{calloc} calls |
553 | @code{malloc} internally. Therefore, if an application provides its own | |
554 | @code{malloc}/@code{realloc}/@code{free} outside the C library, it | |
555 | should always define @code{calloc}, too. | |
556 | ||
28f540f4 | 557 | @node Efficiency and Malloc |
99a20616 | 558 | @subsubsection Efficiency Considerations for @code{malloc} |
28f540f4 RM |
559 | @cindex efficiency and @code{malloc} |
560 | ||
99a20616 UD |
561 | |
562 | ||
563 | ||
c131718c UD |
564 | @ignore |
565 | ||
566 | @c No longer true, see below instead. | |
28f540f4 RM |
567 | To make the best use of @code{malloc}, it helps to know that the GNU |
568 | version of @code{malloc} always dispenses small amounts of memory in | |
569 | blocks whose sizes are powers of two. It keeps separate pools for each | |
570 | power of two. This holds for sizes up to a page size. Therefore, if | |
571 | you are free to choose the size of a small block in order to make | |
572 | @code{malloc} more efficient, make it a power of two. | |
573 | @c !!! xref getpagesize | |
574 | ||
575 | Once a page is split up for a particular block size, it can't be reused | |
576 | for another size unless all the blocks in it are freed. In many | |
577 | programs, this is unlikely to happen. Thus, you can sometimes make a | |
578 | program use memory more efficiently by using blocks of the same size for | |
579 | many different purposes. | |
580 | ||
581 | When you ask for memory blocks of a page or larger, @code{malloc} uses a | |
582 | different strategy; it rounds the size up to a multiple of a page, and | |
583 | it can coalesce and split blocks as needed. | |
584 | ||
585 | The reason for the two strategies is that it is important to allocate | |
586 | and free small blocks as fast as possible, but speed is less important | |
587 | for a large block since the program normally spends a fair amount of | |
588 | time using it. Also, large blocks are normally fewer in number. | |
589 | Therefore, for large blocks, it makes sense to use a method which takes | |
590 | more time to minimize the wasted space. | |
591 | ||
c131718c UD |
592 | @end ignore |
593 | ||
1f77f049 | 594 | As opposed to other versions, the @code{malloc} in @theglibc{} |
99a20616 UD |
595 | does not round up block sizes to powers of two, neither for large nor |
596 | for small sizes. Neighboring chunks can be coalesced on a @code{free} | |
597 | no matter what their size is. This makes the implementation suitable | |
598 | for all kinds of allocation patterns without generally incurring high | |
599 | memory waste through fragmentation. | |
c131718c UD |
600 | |
601 | Very large blocks (much larger than a page) are allocated with | |
602 | @code{mmap} (anonymous or via @code{/dev/zero}) by this implementation. | |
603 | This has the great advantage that these chunks are returned to the | |
604 | system immediately when they are freed. Therefore, it cannot happen | |
605 | that a large chunk becomes ``locked'' in between smaller ones and even | |
606 | after calling @code{free} wastes memory. The size threshold for | |
607 | @code{mmap} to be used can be adjusted with @code{mallopt}. The use of | |
608 | @code{mmap} can also be disabled completely. | |
609 | ||
28f540f4 | 610 | @node Aligned Memory Blocks |
99a20616 | 611 | @subsubsection Allocating Aligned Memory Blocks |
28f540f4 RM |
612 | |
613 | @cindex page boundary | |
614 | @cindex alignment (with @code{malloc}) | |
615 | @pindex stdlib.h | |
616 | The address of a block returned by @code{malloc} or @code{realloc} in | |
a7a93d50 | 617 | @gnusystems{} is always a multiple of eight (or sixteen on 64-bit |
c131718c | 618 | systems). If you need a block whose address is a multiple of a higher |
68979757 | 619 | power of two than that, use @code{memalign}, @code{posix_memalign}, or |
eab0f04c | 620 | @code{valloc}. @code{memalign} is declared in @file{malloc.h} and |
67197215 | 621 | @code{posix_memalign} is declared in @file{stdlib.h}. |
28f540f4 | 622 | |
1f77f049 | 623 | With @theglibc{}, you can use @code{free} to free the blocks that |
68979757 UD |
624 | @code{memalign}, @code{posix_memalign}, and @code{valloc} return. That |
625 | does not work in BSD, however---BSD does not provide any way to free | |
626 | such blocks. | |
28f540f4 | 627 | |
eab0f04c | 628 | @comment malloc.h |
28f540f4 | 629 | @comment BSD |
22a1292a | 630 | @deftypefun {void *} memalign (size_t @var{boundary}, size_t @var{size}) |
28f540f4 RM |
631 | The @code{memalign} function allocates a block of @var{size} bytes whose |
632 | address is a multiple of @var{boundary}. The @var{boundary} must be a | |
c131718c UD |
633 | power of two! The function @code{memalign} works by allocating a |
634 | somewhat larger block, and then returning an address within the block | |
635 | that is on the specified boundary. | |
28f540f4 RM |
636 | @end deftypefun |
637 | ||
68979757 UD |
638 | @comment stdlib.h |
639 | @comment POSIX | |
640 | @deftypefun int posix_memalign (void **@var{memptr}, size_t @var{alignment}, size_t @var{size}) | |
641 | The @code{posix_memalign} function is similar to the @code{memalign} | |
642 | function in that it returns a buffer of @var{size} bytes aligned to a | |
643 | multiple of @var{alignment}. But it adds one requirement to the | |
644 | parameter @var{alignment}: the value must be a power of two multiple of | |
645 | @code{sizeof (void *)}. | |
646 | ||
647 | If the function succeeds in allocation memory a pointer to the allocated | |
648 | memory is returned in @code{*@var{memptr}} and the return value is zero. | |
649 | Otherwise the function returns an error value indicating the problem. | |
650 | ||
651 | This function was introduced in POSIX 1003.1d. | |
652 | @end deftypefun | |
653 | ||
28f540f4 RM |
654 | @comment malloc.h stdlib.h |
655 | @comment BSD | |
656 | @deftypefun {void *} valloc (size_t @var{size}) | |
657 | Using @code{valloc} is like using @code{memalign} and passing the page size | |
658 | as the value of the second argument. It is implemented like this: | |
659 | ||
660 | @smallexample | |
661 | void * | |
662 | valloc (size_t size) | |
663 | @{ | |
22a1292a | 664 | return memalign (getpagesize (), size); |
28f540f4 RM |
665 | @} |
666 | @end smallexample | |
b642f101 UD |
667 | |
668 | @ref{Query Memory Parameters} for more information about the memory | |
669 | subsystem. | |
28f540f4 RM |
670 | @end deftypefun |
671 | ||
c131718c | 672 | @node Malloc Tunable Parameters |
99a20616 | 673 | @subsubsection Malloc Tunable Parameters |
c131718c UD |
674 | |
675 | You can adjust some parameters for dynamic memory allocation with the | |
676 | @code{mallopt} function. This function is the general SVID/XPG | |
677 | interface, defined in @file{malloc.h}. | |
678 | @pindex malloc.h | |
679 | ||
680 | @deftypefun int mallopt (int @var{param}, int @var{value}) | |
681 | When calling @code{mallopt}, the @var{param} argument specifies the | |
682 | parameter to be set, and @var{value} the new value to be set. Possible | |
683 | choices for @var{param}, as defined in @file{malloc.h}, are: | |
684 | ||
685 | @table @code | |
ec4ff04d CD |
686 | @comment TODO: @item M_ARENA_MAX |
687 | @comment - Document ARENA_MAX env var. | |
688 | @comment TODO: @item M_ARENA_TEST | |
689 | @comment - Document ARENA_TEST env var. | |
690 | @comment TODO: @item M_CHECK_ACTION | |
691 | @item M_MMAP_MAX | |
692 | The maximum number of chunks to allocate with @code{mmap}. Setting this | |
693 | to zero disables all use of @code{mmap}. | |
c131718c UD |
694 | @item M_MMAP_THRESHOLD |
695 | All chunks larger than this value are allocated outside the normal | |
696 | heap, using the @code{mmap} system call. This way it is guaranteed | |
697 | that the memory for these chunks can be returned to the system on | |
13c0f771 AJ |
698 | @code{free}. Note that requests smaller than this threshold might still |
699 | be allocated via @code{mmap}. | |
ec4ff04d | 700 | @comment TODO: @item M_MXFAST |
deb9cabb AS |
701 | @item M_PERTURB |
702 | If non-zero, memory blocks are filled with values depending on some | |
703 | low order bits of this parameter when they are allocated (except when | |
704 | allocated by @code{calloc}) and freed. This can be used to debug the | |
b741de23 SP |
705 | use of uninitialized or freed heap memory. Note that this option does not |
706 | guarantee that the freed block will have any specific values. It only | |
707 | guarantees that the content the block had before it was freed will be | |
708 | overwritten. | |
ec4ff04d CD |
709 | @item M_TOP_PAD |
710 | This parameter determines the amount of extra memory to obtain from the | |
711 | system when a call to @code{sbrk} is required. It also specifies the | |
712 | number of bytes to retain when shrinking the heap by calling @code{sbrk} | |
713 | with a negative argument. This provides the necessary hysteresis in | |
714 | heap size such that excessive amounts of system calls can be avoided. | |
715 | @item M_TRIM_THRESHOLD | |
716 | This is the minimum size (in bytes) of the top-most, releasable chunk | |
717 | that will cause @code{sbrk} to be called with a negative argument in | |
718 | order to return memory to the system. | |
c131718c UD |
719 | @end table |
720 | ||
721 | @end deftypefun | |
722 | ||
28f540f4 | 723 | @node Heap Consistency Checking |
99a20616 | 724 | @subsubsection Heap Consistency Checking |
28f540f4 RM |
725 | |
726 | @cindex heap consistency checking | |
727 | @cindex consistency checking, of heap | |
728 | ||
99a20616 | 729 | You can ask @code{malloc} to check the consistency of dynamic memory by |
28f540f4 | 730 | using the @code{mcheck} function. This function is a GNU extension, |
4775243a UD |
731 | declared in @file{mcheck.h}. |
732 | @pindex mcheck.h | |
28f540f4 | 733 | |
4775243a | 734 | @comment mcheck.h |
28f540f4 RM |
735 | @comment GNU |
736 | @deftypefun int mcheck (void (*@var{abortfn}) (enum mcheck_status @var{status})) | |
737 | Calling @code{mcheck} tells @code{malloc} to perform occasional | |
738 | consistency checks. These will catch things such as writing | |
739 | past the end of a block that was allocated with @code{malloc}. | |
740 | ||
741 | The @var{abortfn} argument is the function to call when an inconsistency | |
742 | is found. If you supply a null pointer, then @code{mcheck} uses a | |
743 | default function which prints a message and calls @code{abort} | |
744 | (@pxref{Aborting a Program}). The function you supply is called with | |
745 | one argument, which says what sort of inconsistency was detected; its | |
746 | type is described below. | |
747 | ||
748 | It is too late to begin allocation checking once you have allocated | |
749 | anything with @code{malloc}. So @code{mcheck} does nothing in that | |
750 | case. The function returns @code{-1} if you call it too late, and | |
751 | @code{0} otherwise (when it is successful). | |
752 | ||
753 | The easiest way to arrange to call @code{mcheck} early enough is to use | |
754 | the option @samp{-lmcheck} when you link your program; then you don't | |
bc938d3d | 755 | need to modify your program source at all. Alternatively you might use |
3cb07217 UD |
756 | a debugger to insert a call to @code{mcheck} whenever the program is |
757 | started, for example these gdb commands will automatically call @code{mcheck} | |
758 | whenever the program starts: | |
759 | ||
760 | @smallexample | |
761 | (gdb) break main | |
762 | Breakpoint 1, main (argc=2, argv=0xbffff964) at whatever.c:10 | |
763 | (gdb) command 1 | |
764 | Type commands for when breakpoint 1 is hit, one per line. | |
765 | End with a line saying just "end". | |
766 | >call mcheck(0) | |
767 | >continue | |
768 | >end | |
95fdc6a0 | 769 | (gdb) @dots{} |
3cb07217 UD |
770 | @end smallexample |
771 | ||
772 | This will however only work if no initialization function of any object | |
773 | involved calls any of the @code{malloc} functions since @code{mcheck} | |
774 | must be called before the first such function. | |
775 | ||
28f540f4 RM |
776 | @end deftypefun |
777 | ||
778 | @deftypefun {enum mcheck_status} mprobe (void *@var{pointer}) | |
779 | The @code{mprobe} function lets you explicitly check for inconsistencies | |
780 | in a particular allocated block. You must have already called | |
781 | @code{mcheck} at the beginning of the program, to do its occasional | |
782 | checks; calling @code{mprobe} requests an additional consistency check | |
783 | to be done at the time of the call. | |
784 | ||
785 | The argument @var{pointer} must be a pointer returned by @code{malloc} | |
786 | or @code{realloc}. @code{mprobe} returns a value that says what | |
787 | inconsistency, if any, was found. The values are described below. | |
788 | @end deftypefun | |
789 | ||
790 | @deftp {Data Type} {enum mcheck_status} | |
791 | This enumerated type describes what kind of inconsistency was detected | |
792 | in an allocated block, if any. Here are the possible values: | |
793 | ||
794 | @table @code | |
795 | @item MCHECK_DISABLED | |
796 | @code{mcheck} was not called before the first allocation. | |
797 | No consistency checking can be done. | |
798 | @item MCHECK_OK | |
799 | No inconsistency detected. | |
800 | @item MCHECK_HEAD | |
801 | The data immediately before the block was modified. | |
802 | This commonly happens when an array index or pointer | |
803 | is decremented too far. | |
804 | @item MCHECK_TAIL | |
805 | The data immediately after the block was modified. | |
806 | This commonly happens when an array index or pointer | |
807 | is incremented too far. | |
808 | @item MCHECK_FREE | |
809 | The block was already freed. | |
810 | @end table | |
811 | @end deftp | |
812 | ||
7551a1e5 UD |
813 | Another possibility to check for and guard against bugs in the use of |
814 | @code{malloc}, @code{realloc} and @code{free} is to set the environment | |
815 | variable @code{MALLOC_CHECK_}. When @code{MALLOC_CHECK_} is set, a | |
816 | special (less efficient) implementation is used which is designed to be | |
817 | tolerant against simple errors, such as double calls of @code{free} with | |
818 | the same argument, or overruns of a single byte (off-by-one bugs). Not | |
bc938d3d | 819 | all such errors can be protected against, however, and memory leaks can |
7551a1e5 UD |
820 | result. If @code{MALLOC_CHECK_} is set to @code{0}, any detected heap |
821 | corruption is silently ignored; if set to @code{1}, a diagnostic is | |
822 | printed on @code{stderr}; if set to @code{2}, @code{abort} is called | |
823 | immediately. This can be useful because otherwise a crash may happen | |
824 | much later, and the true cause for the problem is then very hard to | |
825 | track down. | |
826 | ||
68979757 UD |
827 | There is one problem with @code{MALLOC_CHECK_}: in SUID or SGID binaries |
828 | it could possibly be exploited since diverging from the normal programs | |
0bc93a2f | 829 | behavior it now writes something to the standard error descriptor. |
68979757 UD |
830 | Therefore the use of @code{MALLOC_CHECK_} is disabled by default for |
831 | SUID and SGID binaries. It can be enabled again by the system | |
832 | administrator by adding a file @file{/etc/suid-debug} (the content is | |
833 | not important it could be empty). | |
834 | ||
789b13c4 | 835 | So, what's the difference between using @code{MALLOC_CHECK_} and linking |
bc938d3d | 836 | with @samp{-lmcheck}? @code{MALLOC_CHECK_} is orthogonal with respect to |
789b13c4 UD |
837 | @samp{-lmcheck}. @samp{-lmcheck} has been added for backward |
838 | compatibility. Both @code{MALLOC_CHECK_} and @samp{-lmcheck} should | |
839 | uncover the same bugs - but using @code{MALLOC_CHECK_} you don't need to | |
840 | recompile your application. | |
841 | ||
28f540f4 | 842 | @node Hooks for Malloc |
99a20616 | 843 | @subsubsection Memory Allocation Hooks |
28f540f4 RM |
844 | @cindex allocation hooks, for @code{malloc} |
845 | ||
1f77f049 | 846 | @Theglibc{} lets you modify the behavior of @code{malloc}, |
28f540f4 RM |
847 | @code{realloc}, and @code{free} by specifying appropriate hook |
848 | functions. You can use these hooks to help you debug programs that use | |
99a20616 | 849 | dynamic memory allocation, for example. |
28f540f4 RM |
850 | |
851 | The hook variables are declared in @file{malloc.h}. | |
852 | @pindex malloc.h | |
853 | ||
854 | @comment malloc.h | |
855 | @comment GNU | |
856 | @defvar __malloc_hook | |
bc938d3d UD |
857 | The value of this variable is a pointer to the function that |
858 | @code{malloc} uses whenever it is called. You should define this | |
859 | function to look like @code{malloc}; that is, like: | |
28f540f4 RM |
860 | |
861 | @smallexample | |
18a3a9a3 | 862 | void *@var{function} (size_t @var{size}, const void *@var{caller}) |
28f540f4 | 863 | @end smallexample |
bd355af0 UD |
864 | |
865 | The value of @var{caller} is the return address found on the stack when | |
bc938d3d UD |
866 | the @code{malloc} function was called. This value allows you to trace |
867 | the memory consumption of the program. | |
28f540f4 RM |
868 | @end defvar |
869 | ||
870 | @comment malloc.h | |
871 | @comment GNU | |
872 | @defvar __realloc_hook | |
873 | The value of this variable is a pointer to function that @code{realloc} | |
874 | uses whenever it is called. You should define this function to look | |
875 | like @code{realloc}; that is, like: | |
876 | ||
877 | @smallexample | |
18a3a9a3 | 878 | void *@var{function} (void *@var{ptr}, size_t @var{size}, const void *@var{caller}) |
28f540f4 | 879 | @end smallexample |
bd355af0 UD |
880 | |
881 | The value of @var{caller} is the return address found on the stack when | |
e8b1163e | 882 | the @code{realloc} function was called. This value allows you to trace the |
bd355af0 | 883 | memory consumption of the program. |
28f540f4 RM |
884 | @end defvar |
885 | ||
886 | @comment malloc.h | |
887 | @comment GNU | |
888 | @defvar __free_hook | |
889 | The value of this variable is a pointer to function that @code{free} | |
890 | uses whenever it is called. You should define this function to look | |
891 | like @code{free}; that is, like: | |
892 | ||
893 | @smallexample | |
18a3a9a3 | 894 | void @var{function} (void *@var{ptr}, const void *@var{caller}) |
28f540f4 | 895 | @end smallexample |
bd355af0 UD |
896 | |
897 | The value of @var{caller} is the return address found on the stack when | |
e8b1163e | 898 | the @code{free} function was called. This value allows you to trace the |
bd355af0 | 899 | memory consumption of the program. |
28f540f4 RM |
900 | @end defvar |
901 | ||
3cb07217 UD |
902 | @comment malloc.h |
903 | @comment GNU | |
904 | @defvar __memalign_hook | |
905 | The value of this variable is a pointer to function that @code{memalign} | |
906 | uses whenever it is called. You should define this function to look | |
907 | like @code{memalign}; that is, like: | |
908 | ||
909 | @smallexample | |
46ca7a1c | 910 | void *@var{function} (size_t @var{alignment}, size_t @var{size}, const void *@var{caller}) |
3cb07217 | 911 | @end smallexample |
18a3a9a3 UD |
912 | |
913 | The value of @var{caller} is the return address found on the stack when | |
914 | the @code{memalign} function was called. This value allows you to trace the | |
915 | memory consumption of the program. | |
3cb07217 UD |
916 | @end defvar |
917 | ||
28f540f4 RM |
918 | You must make sure that the function you install as a hook for one of |
919 | these functions does not call that function recursively without restoring | |
920 | the old value of the hook first! Otherwise, your program will get stuck | |
3cb07217 UD |
921 | in an infinite recursion. Before calling the function recursively, one |
922 | should make sure to restore all the hooks to their previous value. When | |
923 | coming back from the recursive call, all the hooks should be resaved | |
924 | since a hook might modify itself. | |
28f540f4 | 925 | |
b2f46c3c UD |
926 | @comment malloc.h |
927 | @comment GNU | |
928 | @defvar __malloc_initialize_hook | |
929 | The value of this variable is a pointer to a function that is called | |
930 | once when the malloc implementation is initialized. This is a weak | |
931 | variable, so it can be overridden in the application with a definition | |
932 | like the following: | |
933 | ||
934 | @smallexample | |
935 | void (*@var{__malloc_initialize_hook}) (void) = my_init_hook; | |
936 | @end smallexample | |
937 | @end defvar | |
938 | ||
939 | An issue to look out for is the time at which the malloc hook functions | |
940 | can be safely installed. If the hook functions call the malloc-related | |
941 | functions recursively, it is necessary that malloc has already properly | |
942 | initialized itself at the time when @code{__malloc_hook} etc. is | |
943 | assigned to. On the other hand, if the hook functions provide a | |
944 | complete malloc implementation of their own, it is vital that the hooks | |
945 | are assigned to @emph{before} the very first @code{malloc} call has | |
946 | completed, because otherwise a chunk obtained from the ordinary, | |
947 | un-hooked malloc may later be handed to @code{__free_hook}, for example. | |
948 | ||
949 | In both cases, the problem can be solved by setting up the hooks from | |
950 | within a user-defined function pointed to by | |
951 | @code{__malloc_initialize_hook}---then the hooks will be set up safely | |
952 | at the right time. | |
953 | ||
3cb07217 UD |
954 | Here is an example showing how to use @code{__malloc_hook} and |
955 | @code{__free_hook} properly. It installs a function that prints out | |
956 | information every time @code{malloc} or @code{free} is called. We just | |
957 | assume here that @code{realloc} and @code{memalign} are not used in our | |
958 | program. | |
28f540f4 RM |
959 | |
960 | @smallexample | |
18a3a9a3 UD |
961 | /* Prototypes for __malloc_hook, __free_hook */ |
962 | #include <malloc.h> | |
3cb07217 UD |
963 | |
964 | /* Prototypes for our hooks. */ | |
2ac057a0 | 965 | static void my_init_hook (void); |
18a3a9a3 UD |
966 | static void *my_malloc_hook (size_t, const void *); |
967 | static void my_free_hook (void*, const void *); | |
b2f46c3c UD |
968 | |
969 | /* Override initializing hook from the C library. */ | |
970 | void (*__malloc_initialize_hook) (void) = my_init_hook; | |
971 | ||
972 | static void | |
973 | my_init_hook (void) | |
974 | @{ | |
975 | old_malloc_hook = __malloc_hook; | |
976 | old_free_hook = __free_hook; | |
977 | __malloc_hook = my_malloc_hook; | |
978 | __free_hook = my_free_hook; | |
979 | @} | |
3cb07217 | 980 | |
28f540f4 | 981 | static void * |
18a3a9a3 | 982 | my_malloc_hook (size_t size, const void *caller) |
28f540f4 RM |
983 | @{ |
984 | void *result; | |
3cb07217 | 985 | /* Restore all old hooks */ |
28f540f4 | 986 | __malloc_hook = old_malloc_hook; |
3cb07217 UD |
987 | __free_hook = old_free_hook; |
988 | /* Call recursively */ | |
28f540f4 | 989 | result = malloc (size); |
0bc93a2f | 990 | /* Save underlying hooks */ |
3cb07217 UD |
991 | old_malloc_hook = __malloc_hook; |
992 | old_free_hook = __free_hook; | |
28f540f4 RM |
993 | /* @r{@code{printf} might call @code{malloc}, so protect it too.} */ |
994 | printf ("malloc (%u) returns %p\n", (unsigned int) size, result); | |
3cb07217 | 995 | /* Restore our own hooks */ |
28f540f4 | 996 | __malloc_hook = my_malloc_hook; |
3cb07217 | 997 | __free_hook = my_free_hook; |
28f540f4 RM |
998 | return result; |
999 | @} | |
1000 | ||
2ac057a0 | 1001 | static void |
18a3a9a3 | 1002 | my_free_hook (void *ptr, const void *caller) |
3cb07217 UD |
1003 | @{ |
1004 | /* Restore all old hooks */ | |
1005 | __malloc_hook = old_malloc_hook; | |
1006 | __free_hook = old_free_hook; | |
1007 | /* Call recursively */ | |
1008 | free (ptr); | |
0bc93a2f | 1009 | /* Save underlying hooks */ |
3cb07217 UD |
1010 | old_malloc_hook = __malloc_hook; |
1011 | old_free_hook = __free_hook; | |
1012 | /* @r{@code{printf} might call @code{free}, so protect it too.} */ | |
1013 | printf ("freed pointer %p\n", ptr); | |
1014 | /* Restore our own hooks */ | |
1015 | __malloc_hook = my_malloc_hook; | |
1016 | __free_hook = my_free_hook; | |
1017 | @} | |
1018 | ||
28f540f4 RM |
1019 | main () |
1020 | @{ | |
95fdc6a0 | 1021 | @dots{} |
28f540f4 RM |
1022 | @} |
1023 | @end smallexample | |
1024 | ||
1025 | The @code{mcheck} function (@pxref{Heap Consistency Checking}) works by | |
1026 | installing such hooks. | |
1027 | ||
1028 | @c __morecore, __after_morecore_hook are undocumented | |
1029 | @c It's not clear whether to document them. | |
1030 | ||
1031 | @node Statistics of Malloc | |
99a20616 | 1032 | @subsubsection Statistics for Memory Allocation with @code{malloc} |
28f540f4 RM |
1033 | |
1034 | @cindex allocation statistics | |
99a20616 | 1035 | You can get information about dynamic memory allocation by calling the |
c131718c UD |
1036 | @code{mallinfo} function. This function and its associated data type |
1037 | are declared in @file{malloc.h}; they are an extension of the standard | |
1038 | SVID/XPG version. | |
28f540f4 RM |
1039 | @pindex malloc.h |
1040 | ||
1041 | @comment malloc.h | |
1042 | @comment GNU | |
c131718c | 1043 | @deftp {Data Type} {struct mallinfo} |
28f540f4 | 1044 | This structure type is used to return information about the dynamic |
99a20616 | 1045 | memory allocator. It contains the following members: |
28f540f4 RM |
1046 | |
1047 | @table @code | |
c131718c UD |
1048 | @item int arena |
1049 | This is the total size of memory allocated with @code{sbrk} by | |
1050 | @code{malloc}, in bytes. | |
1051 | ||
1052 | @item int ordblks | |
99a20616 | 1053 | This is the number of chunks not in use. (The memory allocator |
c131718c UD |
1054 | internally gets chunks of memory from the operating system, and then |
1055 | carves them up to satisfy individual @code{malloc} requests; see | |
1056 | @ref{Efficiency and Malloc}.) | |
1057 | ||
1058 | @item int smblks | |
1059 | This field is unused. | |
1060 | ||
1061 | @item int hblks | |
1062 | This is the total number of chunks allocated with @code{mmap}. | |
1063 | ||
1064 | @item int hblkhd | |
1065 | This is the total size of memory allocated with @code{mmap}, in bytes. | |
1066 | ||
1067 | @item int usmblks | |
1068 | This field is unused. | |
28f540f4 | 1069 | |
c131718c UD |
1070 | @item int fsmblks |
1071 | This field is unused. | |
28f540f4 | 1072 | |
c131718c UD |
1073 | @item int uordblks |
1074 | This is the total size of memory occupied by chunks handed out by | |
1075 | @code{malloc}. | |
1076 | ||
1077 | @item int fordblks | |
1078 | This is the total size of memory occupied by free (not in use) chunks. | |
28f540f4 | 1079 | |
c131718c | 1080 | @item int keepcost |
e8b1163e | 1081 | This is the size of the top-most releasable chunk that normally |
11bf311e | 1082 | borders the end of the heap (i.e., the high end of the virtual address |
99a20616 | 1083 | space's data segment). |
28f540f4 | 1084 | |
28f540f4 RM |
1085 | @end table |
1086 | @end deftp | |
1087 | ||
1088 | @comment malloc.h | |
c131718c UD |
1089 | @comment SVID |
1090 | @deftypefun {struct mallinfo} mallinfo (void) | |
28f540f4 | 1091 | This function returns information about the current dynamic memory usage |
c131718c | 1092 | in a structure of type @code{struct mallinfo}. |
28f540f4 RM |
1093 | @end deftypefun |
1094 | ||
1095 | @node Summary of Malloc | |
99a20616 | 1096 | @subsubsection Summary of @code{malloc}-Related Functions |
28f540f4 RM |
1097 | |
1098 | Here is a summary of the functions that work with @code{malloc}: | |
1099 | ||
1100 | @table @code | |
1101 | @item void *malloc (size_t @var{size}) | |
1102 | Allocate a block of @var{size} bytes. @xref{Basic Allocation}. | |
1103 | ||
1104 | @item void free (void *@var{addr}) | |
1105 | Free a block previously allocated by @code{malloc}. @xref{Freeing after | |
1106 | Malloc}. | |
1107 | ||
1108 | @item void *realloc (void *@var{addr}, size_t @var{size}) | |
1109 | Make a block previously allocated by @code{malloc} larger or smaller, | |
1110 | possibly by copying it to a new location. @xref{Changing Block Size}. | |
1111 | ||
1112 | @item void *calloc (size_t @var{count}, size_t @var{eltsize}) | |
1113 | Allocate a block of @var{count} * @var{eltsize} bytes using | |
1114 | @code{malloc}, and set its contents to zero. @xref{Allocating Cleared | |
1115 | Space}. | |
1116 | ||
1117 | @item void *valloc (size_t @var{size}) | |
1118 | Allocate a block of @var{size} bytes, starting on a page boundary. | |
1119 | @xref{Aligned Memory Blocks}. | |
1120 | ||
1121 | @item void *memalign (size_t @var{size}, size_t @var{boundary}) | |
1122 | Allocate a block of @var{size} bytes, starting on an address that is a | |
1123 | multiple of @var{boundary}. @xref{Aligned Memory Blocks}. | |
1124 | ||
c131718c | 1125 | @item int mallopt (int @var{param}, int @var{value}) |
8b7fb588 | 1126 | Adjust a tunable parameter. @xref{Malloc Tunable Parameters}. |
c131718c | 1127 | |
28f540f4 RM |
1128 | @item int mcheck (void (*@var{abortfn}) (void)) |
1129 | Tell @code{malloc} to perform occasional consistency checks on | |
1130 | dynamically allocated memory, and to call @var{abortfn} when an | |
1131 | inconsistency is found. @xref{Heap Consistency Checking}. | |
1132 | ||
18a3a9a3 | 1133 | @item void *(*__malloc_hook) (size_t @var{size}, const void *@var{caller}) |
28f540f4 RM |
1134 | A pointer to a function that @code{malloc} uses whenever it is called. |
1135 | ||
18a3a9a3 | 1136 | @item void *(*__realloc_hook) (void *@var{ptr}, size_t @var{size}, const void *@var{caller}) |
28f540f4 RM |
1137 | A pointer to a function that @code{realloc} uses whenever it is called. |
1138 | ||
18a3a9a3 | 1139 | @item void (*__free_hook) (void *@var{ptr}, const void *@var{caller}) |
28f540f4 RM |
1140 | A pointer to a function that @code{free} uses whenever it is called. |
1141 | ||
18a3a9a3 | 1142 | @item void (*__memalign_hook) (size_t @var{size}, size_t @var{alignment}, const void *@var{caller}) |
3cb07217 UD |
1143 | A pointer to a function that @code{memalign} uses whenever it is called. |
1144 | ||
c131718c | 1145 | @item struct mallinfo mallinfo (void) |
28f540f4 RM |
1146 | Return information about the current dynamic memory usage. |
1147 | @xref{Statistics of Malloc}. | |
1148 | @end table | |
1149 | ||
bd355af0 | 1150 | @node Allocation Debugging |
99a20616 | 1151 | @subsection Allocation Debugging |
bd355af0 UD |
1152 | @cindex allocation debugging |
1153 | @cindex malloc debugger | |
1154 | ||
bc938d3d | 1155 | A complicated task when programming with languages which do not use |
bd355af0 UD |
1156 | garbage collected dynamic memory allocation is to find memory leaks. |
1157 | Long running programs must assure that dynamically allocated objects are | |
1158 | freed at the end of their lifetime. If this does not happen the system | |
1159 | runs out of memory, sooner or later. | |
1160 | ||
1f77f049 | 1161 | The @code{malloc} implementation in @theglibc{} provides some |
bc938d3d | 1162 | simple means to detect such leaks and obtain some information to find |
bd355af0 UD |
1163 | the location. To do this the application must be started in a special |
1164 | mode which is enabled by an environment variable. There are no speed | |
bc938d3d | 1165 | penalties for the program if the debugging mode is not enabled. |
bd355af0 UD |
1166 | |
1167 | @menu | |
1168 | * Tracing malloc:: How to install the tracing functionality. | |
1169 | * Using the Memory Debugger:: Example programs excerpts. | |
1170 | * Tips for the Memory Debugger:: Some more or less clever ideas. | |
1171 | * Interpreting the traces:: What do all these lines mean? | |
1172 | @end menu | |
1173 | ||
1174 | @node Tracing malloc | |
99a20616 | 1175 | @subsubsection How to install the tracing functionality |
bd355af0 UD |
1176 | |
1177 | @comment mcheck.h | |
1178 | @comment GNU | |
1179 | @deftypefun void mtrace (void) | |
1180 | When the @code{mtrace} function is called it looks for an environment | |
1181 | variable named @code{MALLOC_TRACE}. This variable is supposed to | |
1182 | contain a valid file name. The user must have write access. If the | |
1183 | file already exists it is truncated. If the environment variable is not | |
1184 | set or it does not name a valid file which can be opened for writing | |
0bc93a2f | 1185 | nothing is done. The behavior of @code{malloc} etc. is not changed. |
bc938d3d UD |
1186 | For obvious reasons this also happens if the application is installed |
1187 | with the SUID or SGID bit set. | |
bd355af0 | 1188 | |
e8b1163e | 1189 | If the named file is successfully opened, @code{mtrace} installs special |
bd355af0 | 1190 | handlers for the functions @code{malloc}, @code{realloc}, and |
e8b1163e | 1191 | @code{free} (@pxref{Hooks for Malloc}). From then on, all uses of these |
bd355af0 | 1192 | functions are traced and protocolled into the file. There is now of |
bc938d3d | 1193 | course a speed penalty for all calls to the traced functions so tracing |
e8b1163e | 1194 | should not be enabled during normal use. |
bd355af0 UD |
1195 | |
1196 | This function is a GNU extension and generally not available on other | |
1197 | systems. The prototype can be found in @file{mcheck.h}. | |
1198 | @end deftypefun | |
1199 | ||
1200 | @comment mcheck.h | |
1201 | @comment GNU | |
1202 | @deftypefun void muntrace (void) | |
1203 | The @code{muntrace} function can be called after @code{mtrace} was used | |
0bc93a2f | 1204 | to enable tracing the @code{malloc} calls. If no (successful) call of |
bd355af0 UD |
1205 | @code{mtrace} was made @code{muntrace} does nothing. |
1206 | ||
1207 | Otherwise it deinstalls the handlers for @code{malloc}, @code{realloc}, | |
1208 | and @code{free} and then closes the protocol file. No calls are | |
bc938d3d | 1209 | protocolled anymore and the program runs again at full speed. |
bd355af0 UD |
1210 | |
1211 | This function is a GNU extension and generally not available on other | |
1212 | systems. The prototype can be found in @file{mcheck.h}. | |
1213 | @end deftypefun | |
1214 | ||
1215 | @node Using the Memory Debugger | |
99a20616 | 1216 | @subsubsection Example program excerpts |
bd355af0 UD |
1217 | |
1218 | Even though the tracing functionality does not influence the runtime | |
0bc93a2f | 1219 | behavior of the program it is not a good idea to call @code{mtrace} in |
bc938d3d UD |
1220 | all programs. Just imagine that you debug a program using @code{mtrace} |
1221 | and all other programs used in the debugging session also trace their | |
1222 | @code{malloc} calls. The output file would be the same for all programs | |
1223 | and thus is unusable. Therefore one should call @code{mtrace} only if | |
1224 | compiled for debugging. A program could therefore start like this: | |
bd355af0 UD |
1225 | |
1226 | @example | |
1227 | #include <mcheck.h> | |
1228 | ||
1229 | int | |
1230 | main (int argc, char *argv[]) | |
1231 | @{ | |
1232 | #ifdef DEBUGGING | |
1233 | mtrace (); | |
1234 | #endif | |
1235 | @dots{} | |
1236 | @} | |
1237 | @end example | |
1238 | ||
1239 | This is all what is needed if you want to trace the calls during the | |
1240 | whole runtime of the program. Alternatively you can stop the tracing at | |
1241 | any time with a call to @code{muntrace}. It is even possible to restart | |
bc938d3d UD |
1242 | the tracing again with a new call to @code{mtrace}. But this can cause |
1243 | unreliable results since there may be calls of the functions which are | |
1244 | not called. Please note that not only the application uses the traced | |
1245 | functions, also libraries (including the C library itself) use these | |
1246 | functions. | |
bd355af0 UD |
1247 | |
1248 | This last point is also why it is no good idea to call @code{muntrace} | |
1249 | before the program terminated. The libraries are informed about the | |
1250 | termination of the program only after the program returns from | |
1251 | @code{main} or calls @code{exit} and so cannot free the memory they use | |
1252 | before this time. | |
1253 | ||
1254 | So the best thing one can do is to call @code{mtrace} as the very first | |
1255 | function in the program and never call @code{muntrace}. So the program | |
1256 | traces almost all uses of the @code{malloc} functions (except those | |
1257 | calls which are executed by constructors of the program or used | |
1258 | libraries). | |
1259 | ||
1260 | @node Tips for the Memory Debugger | |
99a20616 | 1261 | @subsubsection Some more or less clever ideas |
bd355af0 UD |
1262 | |
1263 | You know the situation. The program is prepared for debugging and in | |
1264 | all debugging sessions it runs well. But once it is started without | |
bc938d3d UD |
1265 | debugging the error shows up. A typical example is a memory leak that |
1266 | becomes visible only when we turn off the debugging. If you foresee | |
1267 | such situations you can still win. Simply use something equivalent to | |
1268 | the following little program: | |
bd355af0 UD |
1269 | |
1270 | @example | |
1271 | #include <mcheck.h> | |
1272 | #include <signal.h> | |
1273 | ||
1274 | static void | |
1275 | enable (int sig) | |
1276 | @{ | |
1277 | mtrace (); | |
1278 | signal (SIGUSR1, enable); | |
1279 | @} | |
1280 | ||
1281 | static void | |
1282 | disable (int sig) | |
1283 | @{ | |
1284 | muntrace (); | |
1285 | signal (SIGUSR2, disable); | |
1286 | @} | |
1287 | ||
1288 | int | |
1289 | main (int argc, char *argv[]) | |
1290 | @{ | |
1291 | @dots{} | |
1292 | ||
1293 | signal (SIGUSR1, enable); | |
1294 | signal (SIGUSR2, disable); | |
1295 | ||
1296 | @dots{} | |
1297 | @} | |
1298 | @end example | |
1299 | ||
9756dfe1 | 1300 | I.e., the user can start the memory debugger any time s/he wants if the |
bd355af0 UD |
1301 | program was started with @code{MALLOC_TRACE} set in the environment. |
1302 | The output will of course not show the allocations which happened before | |
1303 | the first signal but if there is a memory leak this will show up | |
1304 | nevertheless. | |
1305 | ||
1306 | @node Interpreting the traces | |
99a20616 | 1307 | @subsubsection Interpreting the traces |
bd355af0 UD |
1308 | |
1309 | If you take a look at the output it will look similar to this: | |
1310 | ||
1311 | @example | |
1312 | = Start | |
1313 | @ [0x8048209] - 0x8064cc8 | |
1314 | @ [0x8048209] - 0x8064ce0 | |
1315 | @ [0x8048209] - 0x8064cf8 | |
1316 | @ [0x80481eb] + 0x8064c48 0x14 | |
1317 | @ [0x80481eb] + 0x8064c60 0x14 | |
1318 | @ [0x80481eb] + 0x8064c78 0x14 | |
1319 | @ [0x80481eb] + 0x8064c90 0x14 | |
1320 | = End | |
1321 | @end example | |
1322 | ||
1323 | What this all means is not really important since the trace file is not | |
bc938d3d | 1324 | meant to be read by a human. Therefore no attention is given to |
1f77f049 JM |
1325 | readability. Instead there is a program which comes with @theglibc{} |
1326 | which interprets the traces and outputs a summary in an | |
bd355af0 UD |
1327 | user-friendly way. The program is called @code{mtrace} (it is in fact a |
1328 | Perl script) and it takes one or two arguments. In any case the name of | |
bc938d3d UD |
1329 | the file with the trace output must be specified. If an optional |
1330 | argument precedes the name of the trace file this must be the name of | |
1331 | the program which generated the trace. | |
bd355af0 UD |
1332 | |
1333 | @example | |
1334 | drepper$ mtrace tst-mtrace log | |
1335 | No memory leaks. | |
1336 | @end example | |
1337 | ||
1338 | In this case the program @code{tst-mtrace} was run and it produced a | |
1339 | trace file @file{log}. The message printed by @code{mtrace} shows there | |
1340 | are no problems with the code, all allocated memory was freed | |
1341 | afterwards. | |
1342 | ||
1343 | If we call @code{mtrace} on the example trace given above we would get a | |
1344 | different outout: | |
1345 | ||
1346 | @example | |
1347 | drepper$ mtrace errlog | |
1348 | - 0x08064cc8 Free 2 was never alloc'd 0x8048209 | |
1349 | - 0x08064ce0 Free 3 was never alloc'd 0x8048209 | |
1350 | - 0x08064cf8 Free 4 was never alloc'd 0x8048209 | |
1351 | ||
1352 | Memory not freed: | |
1353 | ----------------- | |
1354 | Address Size Caller | |
1355 | 0x08064c48 0x14 at 0x80481eb | |
1356 | 0x08064c60 0x14 at 0x80481eb | |
1357 | 0x08064c78 0x14 at 0x80481eb | |
1358 | 0x08064c90 0x14 at 0x80481eb | |
1359 | @end example | |
1360 | ||
1361 | We have called @code{mtrace} with only one argument and so the script | |
1362 | has no chance to find out what is meant with the addresses given in the | |
1363 | trace. We can do better: | |
1364 | ||
1365 | @example | |
bc938d3d UD |
1366 | drepper$ mtrace tst errlog |
1367 | - 0x08064cc8 Free 2 was never alloc'd /home/drepper/tst.c:39 | |
1368 | - 0x08064ce0 Free 3 was never alloc'd /home/drepper/tst.c:39 | |
1369 | - 0x08064cf8 Free 4 was never alloc'd /home/drepper/tst.c:39 | |
bd355af0 UD |
1370 | |
1371 | Memory not freed: | |
1372 | ----------------- | |
1373 | Address Size Caller | |
bc938d3d UD |
1374 | 0x08064c48 0x14 at /home/drepper/tst.c:33 |
1375 | 0x08064c60 0x14 at /home/drepper/tst.c:33 | |
1376 | 0x08064c78 0x14 at /home/drepper/tst.c:33 | |
1377 | 0x08064c90 0x14 at /home/drepper/tst.c:33 | |
bd355af0 UD |
1378 | @end example |
1379 | ||
1380 | Suddenly the output makes much more sense and the user can see | |
1381 | immediately where the function calls causing the trouble can be found. | |
1382 | ||
9756dfe1 UD |
1383 | Interpreting this output is not complicated. There are at most two |
1384 | different situations being detected. First, @code{free} was called for | |
1385 | pointers which were never returned by one of the allocation functions. | |
bc938d3d | 1386 | This is usually a very bad problem and what this looks like is shown in |
9756dfe1 UD |
1387 | the first three lines of the output. Situations like this are quite |
1388 | rare and if they appear they show up very drastically: the program | |
1389 | normally crashes. | |
1390 | ||
1391 | The other situation which is much harder to detect are memory leaks. As | |
1392 | you can see in the output the @code{mtrace} function collects all this | |
1393 | information and so can say that the program calls an allocation function | |
1394 | from line 33 in the source file @file{/home/drepper/tst-mtrace.c} four | |
1395 | times without freeing this memory before the program terminates. | |
bc938d3d | 1396 | Whether this is a real problem remains to be investigated. |
9756dfe1 | 1397 | |
28f540f4 | 1398 | @node Obstacks |
99a20616 | 1399 | @subsection Obstacks |
28f540f4 RM |
1400 | @cindex obstacks |
1401 | ||
1402 | An @dfn{obstack} is a pool of memory containing a stack of objects. You | |
1403 | can create any number of separate obstacks, and then allocate objects in | |
1404 | specified obstacks. Within each obstack, the last object allocated must | |
1405 | always be the first one freed, but distinct obstacks are independent of | |
1406 | each other. | |
1407 | ||
1408 | Aside from this one constraint of order of freeing, obstacks are totally | |
1409 | general: an obstack can contain any number of objects of any size. They | |
1410 | are implemented with macros, so allocation is usually very fast as long as | |
1411 | the objects are usually small. And the only space overhead per object is | |
1412 | the padding needed to start each object on a suitable boundary. | |
1413 | ||
1414 | @menu | |
1415 | * Creating Obstacks:: How to declare an obstack in your program. | |
1416 | * Preparing for Obstacks:: Preparations needed before you can | |
1417 | use obstacks. | |
1418 | * Allocation in an Obstack:: Allocating objects in an obstack. | |
1419 | * Freeing Obstack Objects:: Freeing objects in an obstack. | |
1420 | * Obstack Functions:: The obstack functions are both | |
1421 | functions and macros. | |
1422 | * Growing Objects:: Making an object bigger by stages. | |
1423 | * Extra Fast Growing:: Extra-high-efficiency (though more | |
1424 | complicated) growing objects. | |
1425 | * Status of an Obstack:: Inquiries about the status of an obstack. | |
1426 | * Obstacks Data Alignment:: Controlling alignment of objects in obstacks. | |
1427 | * Obstack Chunks:: How obstacks obtain and release chunks; | |
1428 | efficiency considerations. | |
a5113b14 | 1429 | * Summary of Obstacks:: |
28f540f4 RM |
1430 | @end menu |
1431 | ||
1432 | @node Creating Obstacks | |
99a20616 | 1433 | @subsubsection Creating Obstacks |
28f540f4 RM |
1434 | |
1435 | The utilities for manipulating obstacks are declared in the header | |
1436 | file @file{obstack.h}. | |
1437 | @pindex obstack.h | |
1438 | ||
1439 | @comment obstack.h | |
1440 | @comment GNU | |
1441 | @deftp {Data Type} {struct obstack} | |
1442 | An obstack is represented by a data structure of type @code{struct | |
1443 | obstack}. This structure has a small fixed size; it records the status | |
1444 | of the obstack and how to find the space in which objects are allocated. | |
1445 | It does not contain any of the objects themselves. You should not try | |
1446 | to access the contents of the structure directly; use only the functions | |
1447 | described in this chapter. | |
1448 | @end deftp | |
1449 | ||
1450 | You can declare variables of type @code{struct obstack} and use them as | |
1451 | obstacks, or you can allocate obstacks dynamically like any other kind | |
1452 | of object. Dynamic allocation of obstacks allows your program to have a | |
1453 | variable number of different stacks. (You can even allocate an | |
1454 | obstack structure in another obstack, but this is rarely useful.) | |
1455 | ||
1456 | All the functions that work with obstacks require you to specify which | |
1457 | obstack to use. You do this with a pointer of type @code{struct obstack | |
1458 | *}. In the following, we often say ``an obstack'' when strictly | |
1459 | speaking the object at hand is such a pointer. | |
1460 | ||
1461 | The objects in the obstack are packed into large blocks called | |
1462 | @dfn{chunks}. The @code{struct obstack} structure points to a chain of | |
1463 | the chunks currently in use. | |
1464 | ||
1465 | The obstack library obtains a new chunk whenever you allocate an object | |
1466 | that won't fit in the previous chunk. Since the obstack library manages | |
1467 | chunks automatically, you don't need to pay much attention to them, but | |
1468 | you do need to supply a function which the obstack library should use to | |
1469 | get a chunk. Usually you supply a function which uses @code{malloc} | |
1470 | directly or indirectly. You must also supply a function to free a chunk. | |
1471 | These matters are described in the following section. | |
1472 | ||
1473 | @node Preparing for Obstacks | |
99a20616 | 1474 | @subsubsection Preparing for Using Obstacks |
28f540f4 RM |
1475 | |
1476 | Each source file in which you plan to use the obstack functions | |
1477 | must include the header file @file{obstack.h}, like this: | |
1478 | ||
1479 | @smallexample | |
1480 | #include <obstack.h> | |
1481 | @end smallexample | |
1482 | ||
1483 | @findex obstack_chunk_alloc | |
1484 | @findex obstack_chunk_free | |
1485 | Also, if the source file uses the macro @code{obstack_init}, it must | |
1486 | declare or define two functions or macros that will be called by the | |
1487 | obstack library. One, @code{obstack_chunk_alloc}, is used to allocate | |
1488 | the chunks of memory into which objects are packed. The other, | |
1489 | @code{obstack_chunk_free}, is used to return chunks when the objects in | |
1490 | them are freed. These macros should appear before any use of obstacks | |
1491 | in the source file. | |
1492 | ||
1493 | Usually these are defined to use @code{malloc} via the intermediary | |
1494 | @code{xmalloc} (@pxref{Unconstrained Allocation}). This is done with | |
1495 | the following pair of macro definitions: | |
1496 | ||
1497 | @smallexample | |
1498 | #define obstack_chunk_alloc xmalloc | |
1499 | #define obstack_chunk_free free | |
1500 | @end smallexample | |
1501 | ||
1502 | @noindent | |
99a20616 | 1503 | Though the memory you get using obstacks really comes from @code{malloc}, |
28f540f4 RM |
1504 | using obstacks is faster because @code{malloc} is called less often, for |
1505 | larger blocks of memory. @xref{Obstack Chunks}, for full details. | |
1506 | ||
1507 | At run time, before the program can use a @code{struct obstack} object | |
1508 | as an obstack, it must initialize the obstack by calling | |
1509 | @code{obstack_init}. | |
1510 | ||
1511 | @comment obstack.h | |
1512 | @comment GNU | |
1513 | @deftypefun int obstack_init (struct obstack *@var{obstack-ptr}) | |
1514 | Initialize obstack @var{obstack-ptr} for allocation of objects. This | |
3cb07217 UD |
1515 | function calls the obstack's @code{obstack_chunk_alloc} function. If |
1516 | allocation of memory fails, the function pointed to by | |
1517 | @code{obstack_alloc_failed_handler} is called. The @code{obstack_init} | |
1518 | function always returns 1 (Compatibility notice: Former versions of | |
1519 | obstack returned 0 if allocation failed). | |
28f540f4 RM |
1520 | @end deftypefun |
1521 | ||
1522 | Here are two examples of how to allocate the space for an obstack and | |
1523 | initialize it. First, an obstack that is a static variable: | |
1524 | ||
1525 | @smallexample | |
1526 | static struct obstack myobstack; | |
1527 | @dots{} | |
1528 | obstack_init (&myobstack); | |
1529 | @end smallexample | |
1530 | ||
1531 | @noindent | |
1532 | Second, an obstack that is itself dynamically allocated: | |
1533 | ||
1534 | @smallexample | |
1535 | struct obstack *myobstack_ptr | |
1536 | = (struct obstack *) xmalloc (sizeof (struct obstack)); | |
1537 | ||
1538 | obstack_init (myobstack_ptr); | |
1539 | @end smallexample | |
1540 | ||
3cb07217 UD |
1541 | @comment obstack.h |
1542 | @comment GNU | |
1543 | @defvar obstack_alloc_failed_handler | |
1544 | The value of this variable is a pointer to a function that | |
1545 | @code{obstack} uses when @code{obstack_chunk_alloc} fails to allocate | |
1546 | memory. The default action is to print a message and abort. | |
1547 | You should supply a function that either calls @code{exit} | |
1548 | (@pxref{Program Termination}) or @code{longjmp} (@pxref{Non-Local | |
1549 | Exits}) and doesn't return. | |
1550 | ||
1551 | @smallexample | |
1552 | void my_obstack_alloc_failed (void) | |
1553 | @dots{} | |
1554 | obstack_alloc_failed_handler = &my_obstack_alloc_failed; | |
1555 | @end smallexample | |
1556 | ||
1557 | @end defvar | |
1558 | ||
28f540f4 | 1559 | @node Allocation in an Obstack |
99a20616 | 1560 | @subsubsection Allocation in an Obstack |
28f540f4 RM |
1561 | @cindex allocation (obstacks) |
1562 | ||
1563 | The most direct way to allocate an object in an obstack is with | |
1564 | @code{obstack_alloc}, which is invoked almost like @code{malloc}. | |
1565 | ||
1566 | @comment obstack.h | |
1567 | @comment GNU | |
1568 | @deftypefun {void *} obstack_alloc (struct obstack *@var{obstack-ptr}, int @var{size}) | |
1569 | This allocates an uninitialized block of @var{size} bytes in an obstack | |
1570 | and returns its address. Here @var{obstack-ptr} specifies which obstack | |
1571 | to allocate the block in; it is the address of the @code{struct obstack} | |
1572 | object which represents the obstack. Each obstack function or macro | |
1573 | requires you to specify an @var{obstack-ptr} as the first argument. | |
1574 | ||
1575 | This function calls the obstack's @code{obstack_chunk_alloc} function if | |
3cb07217 UD |
1576 | it needs to allocate a new chunk of memory; it calls |
1577 | @code{obstack_alloc_failed_handler} if allocation of memory by | |
1578 | @code{obstack_chunk_alloc} failed. | |
28f540f4 RM |
1579 | @end deftypefun |
1580 | ||
1581 | For example, here is a function that allocates a copy of a string @var{str} | |
1582 | in a specific obstack, which is in the variable @code{string_obstack}: | |
1583 | ||
1584 | @smallexample | |
1585 | struct obstack string_obstack; | |
1586 | ||
1587 | char * | |
1588 | copystring (char *string) | |
1589 | @{ | |
7cc27f44 UD |
1590 | size_t len = strlen (string) + 1; |
1591 | char *s = (char *) obstack_alloc (&string_obstack, len); | |
1592 | memcpy (s, string, len); | |
28f540f4 RM |
1593 | return s; |
1594 | @} | |
1595 | @end smallexample | |
1596 | ||
1597 | To allocate a block with specified contents, use the function | |
1598 | @code{obstack_copy}, declared like this: | |
1599 | ||
1600 | @comment obstack.h | |
1601 | @comment GNU | |
1602 | @deftypefun {void *} obstack_copy (struct obstack *@var{obstack-ptr}, void *@var{address}, int @var{size}) | |
1603 | This allocates a block and initializes it by copying @var{size} | |
3cb07217 UD |
1604 | bytes of data starting at @var{address}. It calls |
1605 | @code{obstack_alloc_failed_handler} if allocation of memory by | |
1606 | @code{obstack_chunk_alloc} failed. | |
28f540f4 RM |
1607 | @end deftypefun |
1608 | ||
1609 | @comment obstack.h | |
1610 | @comment GNU | |
1611 | @deftypefun {void *} obstack_copy0 (struct obstack *@var{obstack-ptr}, void *@var{address}, int @var{size}) | |
1612 | Like @code{obstack_copy}, but appends an extra byte containing a null | |
1613 | character. This extra byte is not counted in the argument @var{size}. | |
1614 | @end deftypefun | |
1615 | ||
1616 | The @code{obstack_copy0} function is convenient for copying a sequence | |
1617 | of characters into an obstack as a null-terminated string. Here is an | |
1618 | example of its use: | |
1619 | ||
1620 | @smallexample | |
1621 | char * | |
1622 | obstack_savestring (char *addr, int size) | |
1623 | @{ | |
1624 | return obstack_copy0 (&myobstack, addr, size); | |
1625 | @} | |
1626 | @end smallexample | |
1627 | ||
1628 | @noindent | |
1629 | Contrast this with the previous example of @code{savestring} using | |
1630 | @code{malloc} (@pxref{Basic Allocation}). | |
1631 | ||
1632 | @node Freeing Obstack Objects | |
99a20616 | 1633 | @subsubsection Freeing Objects in an Obstack |
28f540f4 RM |
1634 | @cindex freeing (obstacks) |
1635 | ||
1636 | To free an object allocated in an obstack, use the function | |
1637 | @code{obstack_free}. Since the obstack is a stack of objects, freeing | |
1638 | one object automatically frees all other objects allocated more recently | |
1639 | in the same obstack. | |
1640 | ||
1641 | @comment obstack.h | |
1642 | @comment GNU | |
1643 | @deftypefun void obstack_free (struct obstack *@var{obstack-ptr}, void *@var{object}) | |
1644 | If @var{object} is a null pointer, everything allocated in the obstack | |
1645 | is freed. Otherwise, @var{object} must be the address of an object | |
1646 | allocated in the obstack. Then @var{object} is freed, along with | |
1647 | everything allocated in @var{obstack} since @var{object}. | |
1648 | @end deftypefun | |
1649 | ||
1650 | Note that if @var{object} is a null pointer, the result is an | |
99a20616 | 1651 | uninitialized obstack. To free all memory in an obstack but leave it |
28f540f4 RM |
1652 | valid for further allocation, call @code{obstack_free} with the address |
1653 | of the first object allocated on the obstack: | |
1654 | ||
1655 | @smallexample | |
1656 | obstack_free (obstack_ptr, first_object_allocated_ptr); | |
1657 | @end smallexample | |
1658 | ||
1659 | Recall that the objects in an obstack are grouped into chunks. When all | |
1660 | the objects in a chunk become free, the obstack library automatically | |
1661 | frees the chunk (@pxref{Preparing for Obstacks}). Then other | |
1662 | obstacks, or non-obstack allocation, can reuse the space of the chunk. | |
1663 | ||
1664 | @node Obstack Functions | |
99a20616 | 1665 | @subsubsection Obstack Functions and Macros |
28f540f4 RM |
1666 | @cindex macros |
1667 | ||
1668 | The interfaces for using obstacks may be defined either as functions or | |
1669 | as macros, depending on the compiler. The obstack facility works with | |
f65fd747 | 1670 | all C compilers, including both @w{ISO C} and traditional C, but there are |
28f540f4 RM |
1671 | precautions you must take if you plan to use compilers other than GNU C. |
1672 | ||
f65fd747 | 1673 | If you are using an old-fashioned @w{non-ISO C} compiler, all the obstack |
28f540f4 RM |
1674 | ``functions'' are actually defined only as macros. You can call these |
1675 | macros like functions, but you cannot use them in any other way (for | |
1676 | example, you cannot take their address). | |
1677 | ||
1678 | Calling the macros requires a special precaution: namely, the first | |
1679 | operand (the obstack pointer) may not contain any side effects, because | |
1680 | it may be computed more than once. For example, if you write this: | |
1681 | ||
1682 | @smallexample | |
1683 | obstack_alloc (get_obstack (), 4); | |
1684 | @end smallexample | |
1685 | ||
1686 | @noindent | |
1687 | you will find that @code{get_obstack} may be called several times. | |
1688 | If you use @code{*obstack_list_ptr++} as the obstack pointer argument, | |
1689 | you will get very strange results since the incrementation may occur | |
1690 | several times. | |
1691 | ||
f65fd747 | 1692 | In @w{ISO C}, each function has both a macro definition and a function |
28f540f4 RM |
1693 | definition. The function definition is used if you take the address of the |
1694 | function without calling it. An ordinary call uses the macro definition by | |
1695 | default, but you can request the function definition instead by writing the | |
1696 | function name in parentheses, as shown here: | |
1697 | ||
1698 | @smallexample | |
1699 | char *x; | |
1700 | void *(*funcp) (); | |
1701 | /* @r{Use the macro}. */ | |
1702 | x = (char *) obstack_alloc (obptr, size); | |
1703 | /* @r{Call the function}. */ | |
1704 | x = (char *) (obstack_alloc) (obptr, size); | |
1705 | /* @r{Take the address of the function}. */ | |
1706 | funcp = obstack_alloc; | |
1707 | @end smallexample | |
1708 | ||
1709 | @noindent | |
f65fd747 | 1710 | This is the same situation that exists in @w{ISO C} for the standard library |
28f540f4 RM |
1711 | functions. @xref{Macro Definitions}. |
1712 | ||
1713 | @strong{Warning:} When you do use the macros, you must observe the | |
f65fd747 | 1714 | precaution of avoiding side effects in the first operand, even in @w{ISO C}. |
28f540f4 RM |
1715 | |
1716 | If you use the GNU C compiler, this precaution is not necessary, because | |
1717 | various language extensions in GNU C permit defining the macros so as to | |
1718 | compute each argument only once. | |
1719 | ||
1720 | @node Growing Objects | |
99a20616 | 1721 | @subsubsection Growing Objects |
28f540f4 RM |
1722 | @cindex growing objects (in obstacks) |
1723 | @cindex changing the size of a block (obstacks) | |
1724 | ||
99a20616 | 1725 | Because memory in obstack chunks is used sequentially, it is possible to |
28f540f4 RM |
1726 | build up an object step by step, adding one or more bytes at a time to the |
1727 | end of the object. With this technique, you do not need to know how much | |
1728 | data you will put in the object until you come to the end of it. We call | |
1729 | this the technique of @dfn{growing objects}. The special functions | |
1730 | for adding data to the growing object are described in this section. | |
1731 | ||
1732 | You don't need to do anything special when you start to grow an object. | |
1733 | Using one of the functions to add data to the object automatically | |
1734 | starts it. However, it is necessary to say explicitly when the object is | |
1735 | finished. This is done with the function @code{obstack_finish}. | |
1736 | ||
1737 | The actual address of the object thus built up is not known until the | |
1738 | object is finished. Until then, it always remains possible that you will | |
1739 | add so much data that the object must be copied into a new chunk. | |
1740 | ||
1741 | While the obstack is in use for a growing object, you cannot use it for | |
1742 | ordinary allocation of another object. If you try to do so, the space | |
1743 | already added to the growing object will become part of the other object. | |
1744 | ||
1745 | @comment obstack.h | |
1746 | @comment GNU | |
1747 | @deftypefun void obstack_blank (struct obstack *@var{obstack-ptr}, int @var{size}) | |
1748 | The most basic function for adding to a growing object is | |
1749 | @code{obstack_blank}, which adds space without initializing it. | |
1750 | @end deftypefun | |
1751 | ||
1752 | @comment obstack.h | |
1753 | @comment GNU | |
1754 | @deftypefun void obstack_grow (struct obstack *@var{obstack-ptr}, void *@var{data}, int @var{size}) | |
1755 | To add a block of initialized space, use @code{obstack_grow}, which is | |
1756 | the growing-object analogue of @code{obstack_copy}. It adds @var{size} | |
1757 | bytes of data to the growing object, copying the contents from | |
1758 | @var{data}. | |
1759 | @end deftypefun | |
1760 | ||
1761 | @comment obstack.h | |
1762 | @comment GNU | |
1763 | @deftypefun void obstack_grow0 (struct obstack *@var{obstack-ptr}, void *@var{data}, int @var{size}) | |
1764 | This is the growing-object analogue of @code{obstack_copy0}. It adds | |
1765 | @var{size} bytes copied from @var{data}, followed by an additional null | |
1766 | character. | |
1767 | @end deftypefun | |
1768 | ||
1769 | @comment obstack.h | |
1770 | @comment GNU | |
1771 | @deftypefun void obstack_1grow (struct obstack *@var{obstack-ptr}, char @var{c}) | |
1772 | To add one character at a time, use the function @code{obstack_1grow}. | |
1773 | It adds a single byte containing @var{c} to the growing object. | |
1774 | @end deftypefun | |
1775 | ||
2c6fe0bd UD |
1776 | @comment obstack.h |
1777 | @comment GNU | |
1778 | @deftypefun void obstack_ptr_grow (struct obstack *@var{obstack-ptr}, void *@var{data}) | |
1779 | Adding the value of a pointer one can use the function | |
1780 | @code{obstack_ptr_grow}. It adds @code{sizeof (void *)} bytes | |
1781 | containing the value of @var{data}. | |
1782 | @end deftypefun | |
1783 | ||
1784 | @comment obstack.h | |
1785 | @comment GNU | |
1786 | @deftypefun void obstack_int_grow (struct obstack *@var{obstack-ptr}, int @var{data}) | |
1787 | A single value of type @code{int} can be added by using the | |
1788 | @code{obstack_int_grow} function. It adds @code{sizeof (int)} bytes to | |
1789 | the growing object and initializes them with the value of @var{data}. | |
1790 | @end deftypefun | |
1791 | ||
28f540f4 RM |
1792 | @comment obstack.h |
1793 | @comment GNU | |
1794 | @deftypefun {void *} obstack_finish (struct obstack *@var{obstack-ptr}) | |
1795 | When you are finished growing the object, use the function | |
1796 | @code{obstack_finish} to close it off and return its final address. | |
1797 | ||
1798 | Once you have finished the object, the obstack is available for ordinary | |
1799 | allocation or for growing another object. | |
1800 | ||
1801 | This function can return a null pointer under the same conditions as | |
1802 | @code{obstack_alloc} (@pxref{Allocation in an Obstack}). | |
1803 | @end deftypefun | |
1804 | ||
1805 | When you build an object by growing it, you will probably need to know | |
1806 | afterward how long it became. You need not keep track of this as you grow | |
1807 | the object, because you can find out the length from the obstack just | |
1808 | before finishing the object with the function @code{obstack_object_size}, | |
1809 | declared as follows: | |
1810 | ||
1811 | @comment obstack.h | |
1812 | @comment GNU | |
1813 | @deftypefun int obstack_object_size (struct obstack *@var{obstack-ptr}) | |
1814 | This function returns the current size of the growing object, in bytes. | |
1815 | Remember to call this function @emph{before} finishing the object. | |
1816 | After it is finished, @code{obstack_object_size} will return zero. | |
1817 | @end deftypefun | |
1818 | ||
1819 | If you have started growing an object and wish to cancel it, you should | |
1820 | finish it and then free it, like this: | |
1821 | ||
1822 | @smallexample | |
1823 | obstack_free (obstack_ptr, obstack_finish (obstack_ptr)); | |
1824 | @end smallexample | |
1825 | ||
1826 | @noindent | |
1827 | This has no effect if no object was growing. | |
1828 | ||
1829 | @cindex shrinking objects | |
1830 | You can use @code{obstack_blank} with a negative size argument to make | |
1831 | the current object smaller. Just don't try to shrink it beyond zero | |
1832 | length---there's no telling what will happen if you do that. | |
1833 | ||
1834 | @node Extra Fast Growing | |
99a20616 | 1835 | @subsubsection Extra Fast Growing Objects |
28f540f4 RM |
1836 | @cindex efficiency and obstacks |
1837 | ||
1838 | The usual functions for growing objects incur overhead for checking | |
1839 | whether there is room for the new growth in the current chunk. If you | |
1840 | are frequently constructing objects in small steps of growth, this | |
1841 | overhead can be significant. | |
1842 | ||
1843 | You can reduce the overhead by using special ``fast growth'' | |
1844 | functions that grow the object without checking. In order to have a | |
1845 | robust program, you must do the checking yourself. If you do this checking | |
1846 | in the simplest way each time you are about to add data to the object, you | |
1847 | have not saved anything, because that is what the ordinary growth | |
1848 | functions do. But if you can arrange to check less often, or check | |
1849 | more efficiently, then you make the program faster. | |
1850 | ||
1851 | The function @code{obstack_room} returns the amount of room available | |
1852 | in the current chunk. It is declared as follows: | |
1853 | ||
1854 | @comment obstack.h | |
1855 | @comment GNU | |
1856 | @deftypefun int obstack_room (struct obstack *@var{obstack-ptr}) | |
1857 | This returns the number of bytes that can be added safely to the current | |
1858 | growing object (or to an object about to be started) in obstack | |
1859 | @var{obstack} using the fast growth functions. | |
1860 | @end deftypefun | |
1861 | ||
1862 | While you know there is room, you can use these fast growth functions | |
1863 | for adding data to a growing object: | |
1864 | ||
1865 | @comment obstack.h | |
1866 | @comment GNU | |
1867 | @deftypefun void obstack_1grow_fast (struct obstack *@var{obstack-ptr}, char @var{c}) | |
1868 | The function @code{obstack_1grow_fast} adds one byte containing the | |
1869 | character @var{c} to the growing object in obstack @var{obstack-ptr}. | |
1870 | @end deftypefun | |
1871 | ||
2c6fe0bd UD |
1872 | @comment obstack.h |
1873 | @comment GNU | |
1874 | @deftypefun void obstack_ptr_grow_fast (struct obstack *@var{obstack-ptr}, void *@var{data}) | |
1875 | The function @code{obstack_ptr_grow_fast} adds @code{sizeof (void *)} | |
1876 | bytes containing the value of @var{data} to the growing object in | |
1877 | obstack @var{obstack-ptr}. | |
1878 | @end deftypefun | |
1879 | ||
1880 | @comment obstack.h | |
1881 | @comment GNU | |
1882 | @deftypefun void obstack_int_grow_fast (struct obstack *@var{obstack-ptr}, int @var{data}) | |
1883 | The function @code{obstack_int_grow_fast} adds @code{sizeof (int)} bytes | |
1884 | containing the value of @var{data} to the growing object in obstack | |
1885 | @var{obstack-ptr}. | |
1886 | @end deftypefun | |
1887 | ||
28f540f4 RM |
1888 | @comment obstack.h |
1889 | @comment GNU | |
1890 | @deftypefun void obstack_blank_fast (struct obstack *@var{obstack-ptr}, int @var{size}) | |
1891 | The function @code{obstack_blank_fast} adds @var{size} bytes to the | |
1892 | growing object in obstack @var{obstack-ptr} without initializing them. | |
1893 | @end deftypefun | |
1894 | ||
1895 | When you check for space using @code{obstack_room} and there is not | |
1896 | enough room for what you want to add, the fast growth functions | |
1897 | are not safe. In this case, simply use the corresponding ordinary | |
1898 | growth function instead. Very soon this will copy the object to a | |
a5113b14 | 1899 | new chunk; then there will be lots of room available again. |
28f540f4 RM |
1900 | |
1901 | So, each time you use an ordinary growth function, check afterward for | |
1902 | sufficient space using @code{obstack_room}. Once the object is copied | |
1903 | to a new chunk, there will be plenty of space again, so the program will | |
1904 | start using the fast growth functions again. | |
1905 | ||
1906 | Here is an example: | |
1907 | ||
1908 | @smallexample | |
1909 | @group | |
1910 | void | |
1911 | add_string (struct obstack *obstack, const char *ptr, int len) | |
1912 | @{ | |
1913 | while (len > 0) | |
1914 | @{ | |
1915 | int room = obstack_room (obstack); | |
1916 | if (room == 0) | |
1917 | @{ | |
1918 | /* @r{Not enough room. Add one character slowly,} | |
1919 | @r{which may copy to a new chunk and make room.} */ | |
1920 | obstack_1grow (obstack, *ptr++); | |
1921 | len--; | |
1922 | @} | |
a5113b14 | 1923 | else |
28f540f4 RM |
1924 | @{ |
1925 | if (room > len) | |
1926 | room = len; | |
1927 | /* @r{Add fast as much as we have room for.} */ | |
1928 | len -= room; | |
1929 | while (room-- > 0) | |
1930 | obstack_1grow_fast (obstack, *ptr++); | |
1931 | @} | |
1932 | @} | |
1933 | @} | |
1934 | @end group | |
1935 | @end smallexample | |
1936 | ||
1937 | @node Status of an Obstack | |
99a20616 | 1938 | @subsubsection Status of an Obstack |
28f540f4 RM |
1939 | @cindex obstack status |
1940 | @cindex status of obstack | |
1941 | ||
1942 | Here are functions that provide information on the current status of | |
1943 | allocation in an obstack. You can use them to learn about an object while | |
1944 | still growing it. | |
1945 | ||
1946 | @comment obstack.h | |
1947 | @comment GNU | |
1948 | @deftypefun {void *} obstack_base (struct obstack *@var{obstack-ptr}) | |
1949 | This function returns the tentative address of the beginning of the | |
1950 | currently growing object in @var{obstack-ptr}. If you finish the object | |
1951 | immediately, it will have that address. If you make it larger first, it | |
1952 | may outgrow the current chunk---then its address will change! | |
1953 | ||
1954 | If no object is growing, this value says where the next object you | |
1955 | allocate will start (once again assuming it fits in the current | |
1956 | chunk). | |
1957 | @end deftypefun | |
1958 | ||
1959 | @comment obstack.h | |
1960 | @comment GNU | |
1961 | @deftypefun {void *} obstack_next_free (struct obstack *@var{obstack-ptr}) | |
1962 | This function returns the address of the first free byte in the current | |
1963 | chunk of obstack @var{obstack-ptr}. This is the end of the currently | |
1964 | growing object. If no object is growing, @code{obstack_next_free} | |
1965 | returns the same value as @code{obstack_base}. | |
1966 | @end deftypefun | |
1967 | ||
1968 | @comment obstack.h | |
1969 | @comment GNU | |
1970 | @deftypefun int obstack_object_size (struct obstack *@var{obstack-ptr}) | |
1971 | This function returns the size in bytes of the currently growing object. | |
1972 | This is equivalent to | |
1973 | ||
1974 | @smallexample | |
1975 | obstack_next_free (@var{obstack-ptr}) - obstack_base (@var{obstack-ptr}) | |
1976 | @end smallexample | |
1977 | @end deftypefun | |
1978 | ||
1979 | @node Obstacks Data Alignment | |
99a20616 | 1980 | @subsubsection Alignment of Data in Obstacks |
28f540f4 RM |
1981 | @cindex alignment (in obstacks) |
1982 | ||
1983 | Each obstack has an @dfn{alignment boundary}; each object allocated in | |
1984 | the obstack automatically starts on an address that is a multiple of the | |
11883883 RM |
1985 | specified boundary. By default, this boundary is aligned so that |
1986 | the object can hold any type of data. | |
28f540f4 RM |
1987 | |
1988 | To access an obstack's alignment boundary, use the macro | |
1989 | @code{obstack_alignment_mask}, whose function prototype looks like | |
1990 | this: | |
1991 | ||
1992 | @comment obstack.h | |
1993 | @comment GNU | |
1994 | @deftypefn Macro int obstack_alignment_mask (struct obstack *@var{obstack-ptr}) | |
1995 | The value is a bit mask; a bit that is 1 indicates that the corresponding | |
1996 | bit in the address of an object should be 0. The mask value should be one | |
1997 | less than a power of 2; the effect is that all object addresses are | |
11883883 RM |
1998 | multiples of that power of 2. The default value of the mask is a value |
1999 | that allows aligned objects to hold any type of data: for example, if | |
2000 | its value is 3, any type of data can be stored at locations whose | |
28f540f4 RM |
2001 | addresses are multiples of 4. A mask value of 0 means an object can start |
2002 | on any multiple of 1 (that is, no alignment is required). | |
2003 | ||
2004 | The expansion of the macro @code{obstack_alignment_mask} is an lvalue, | |
2005 | so you can alter the mask by assignment. For example, this statement: | |
2006 | ||
2007 | @smallexample | |
2008 | obstack_alignment_mask (obstack_ptr) = 0; | |
2009 | @end smallexample | |
2010 | ||
2011 | @noindent | |
2012 | has the effect of turning off alignment processing in the specified obstack. | |
2013 | @end deftypefn | |
2014 | ||
2015 | Note that a change in alignment mask does not take effect until | |
2016 | @emph{after} the next time an object is allocated or finished in the | |
2017 | obstack. If you are not growing an object, you can make the new | |
2018 | alignment mask take effect immediately by calling @code{obstack_finish}. | |
2019 | This will finish a zero-length object and then do proper alignment for | |
2020 | the next object. | |
2021 | ||
2022 | @node Obstack Chunks | |
99a20616 | 2023 | @subsubsection Obstack Chunks |
28f540f4 RM |
2024 | @cindex efficiency of chunks |
2025 | @cindex chunks | |
2026 | ||
2027 | Obstacks work by allocating space for themselves in large chunks, and | |
2028 | then parceling out space in the chunks to satisfy your requests. Chunks | |
2029 | are normally 4096 bytes long unless you specify a different chunk size. | |
2030 | The chunk size includes 8 bytes of overhead that are not actually used | |
2031 | for storing objects. Regardless of the specified size, longer chunks | |
2032 | will be allocated when necessary for long objects. | |
2033 | ||
2034 | The obstack library allocates chunks by calling the function | |
2035 | @code{obstack_chunk_alloc}, which you must define. When a chunk is no | |
2036 | longer needed because you have freed all the objects in it, the obstack | |
2037 | library frees the chunk by calling @code{obstack_chunk_free}, which you | |
2038 | must also define. | |
2039 | ||
2040 | These two must be defined (as macros) or declared (as functions) in each | |
2041 | source file that uses @code{obstack_init} (@pxref{Creating Obstacks}). | |
2042 | Most often they are defined as macros like this: | |
2043 | ||
2044 | @smallexample | |
bd355af0 | 2045 | #define obstack_chunk_alloc malloc |
28f540f4 RM |
2046 | #define obstack_chunk_free free |
2047 | @end smallexample | |
2048 | ||
2049 | Note that these are simple macros (no arguments). Macro definitions with | |
2050 | arguments will not work! It is necessary that @code{obstack_chunk_alloc} | |
2051 | or @code{obstack_chunk_free}, alone, expand into a function name if it is | |
2052 | not itself a function name. | |
2053 | ||
2054 | If you allocate chunks with @code{malloc}, the chunk size should be a | |
2055 | power of 2. The default chunk size, 4096, was chosen because it is long | |
2056 | enough to satisfy many typical requests on the obstack yet short enough | |
2057 | not to waste too much memory in the portion of the last chunk not yet used. | |
2058 | ||
2059 | @comment obstack.h | |
2060 | @comment GNU | |
2061 | @deftypefn Macro int obstack_chunk_size (struct obstack *@var{obstack-ptr}) | |
2062 | This returns the chunk size of the given obstack. | |
2063 | @end deftypefn | |
2064 | ||
2065 | Since this macro expands to an lvalue, you can specify a new chunk size by | |
2066 | assigning it a new value. Doing so does not affect the chunks already | |
2067 | allocated, but will change the size of chunks allocated for that particular | |
2068 | obstack in the future. It is unlikely to be useful to make the chunk size | |
2069 | smaller, but making it larger might improve efficiency if you are | |
2070 | allocating many objects whose size is comparable to the chunk size. Here | |
2071 | is how to do so cleanly: | |
2072 | ||
2073 | @smallexample | |
2074 | if (obstack_chunk_size (obstack_ptr) < @var{new-chunk-size}) | |
2075 | obstack_chunk_size (obstack_ptr) = @var{new-chunk-size}; | |
2076 | @end smallexample | |
2077 | ||
2078 | @node Summary of Obstacks | |
99a20616 | 2079 | @subsubsection Summary of Obstack Functions |
28f540f4 RM |
2080 | |
2081 | Here is a summary of all the functions associated with obstacks. Each | |
2082 | takes the address of an obstack (@code{struct obstack *}) as its first | |
2083 | argument. | |
2084 | ||
2085 | @table @code | |
2086 | @item void obstack_init (struct obstack *@var{obstack-ptr}) | |
2087 | Initialize use of an obstack. @xref{Creating Obstacks}. | |
2088 | ||
2089 | @item void *obstack_alloc (struct obstack *@var{obstack-ptr}, int @var{size}) | |
2090 | Allocate an object of @var{size} uninitialized bytes. | |
2091 | @xref{Allocation in an Obstack}. | |
2092 | ||
2093 | @item void *obstack_copy (struct obstack *@var{obstack-ptr}, void *@var{address}, int @var{size}) | |
2094 | Allocate an object of @var{size} bytes, with contents copied from | |
2095 | @var{address}. @xref{Allocation in an Obstack}. | |
2096 | ||
2097 | @item void *obstack_copy0 (struct obstack *@var{obstack-ptr}, void *@var{address}, int @var{size}) | |
2098 | Allocate an object of @var{size}+1 bytes, with @var{size} of them copied | |
2099 | from @var{address}, followed by a null character at the end. | |
2100 | @xref{Allocation in an Obstack}. | |
2101 | ||
2102 | @item void obstack_free (struct obstack *@var{obstack-ptr}, void *@var{object}) | |
2103 | Free @var{object} (and everything allocated in the specified obstack | |
2104 | more recently than @var{object}). @xref{Freeing Obstack Objects}. | |
2105 | ||
2106 | @item void obstack_blank (struct obstack *@var{obstack-ptr}, int @var{size}) | |
2107 | Add @var{size} uninitialized bytes to a growing object. | |
2108 | @xref{Growing Objects}. | |
2109 | ||
2110 | @item void obstack_grow (struct obstack *@var{obstack-ptr}, void *@var{address}, int @var{size}) | |
2111 | Add @var{size} bytes, copied from @var{address}, to a growing object. | |
2112 | @xref{Growing Objects}. | |
2113 | ||
2114 | @item void obstack_grow0 (struct obstack *@var{obstack-ptr}, void *@var{address}, int @var{size}) | |
2115 | Add @var{size} bytes, copied from @var{address}, to a growing object, | |
2116 | and then add another byte containing a null character. @xref{Growing | |
2117 | Objects}. | |
2118 | ||
2119 | @item void obstack_1grow (struct obstack *@var{obstack-ptr}, char @var{data-char}) | |
2120 | Add one byte containing @var{data-char} to a growing object. | |
2121 | @xref{Growing Objects}. | |
2122 | ||
2123 | @item void *obstack_finish (struct obstack *@var{obstack-ptr}) | |
2124 | Finalize the object that is growing and return its permanent address. | |
2125 | @xref{Growing Objects}. | |
2126 | ||
2127 | @item int obstack_object_size (struct obstack *@var{obstack-ptr}) | |
2128 | Get the current size of the currently growing object. @xref{Growing | |
2129 | Objects}. | |
2130 | ||
2131 | @item void obstack_blank_fast (struct obstack *@var{obstack-ptr}, int @var{size}) | |
2132 | Add @var{size} uninitialized bytes to a growing object without checking | |
2133 | that there is enough room. @xref{Extra Fast Growing}. | |
2134 | ||
2135 | @item void obstack_1grow_fast (struct obstack *@var{obstack-ptr}, char @var{data-char}) | |
2136 | Add one byte containing @var{data-char} to a growing object without | |
2137 | checking that there is enough room. @xref{Extra Fast Growing}. | |
2138 | ||
2139 | @item int obstack_room (struct obstack *@var{obstack-ptr}) | |
2140 | Get the amount of room now available for growing the current object. | |
2141 | @xref{Extra Fast Growing}. | |
2142 | ||
2143 | @item int obstack_alignment_mask (struct obstack *@var{obstack-ptr}) | |
2144 | The mask used for aligning the beginning of an object. This is an | |
2145 | lvalue. @xref{Obstacks Data Alignment}. | |
2146 | ||
2147 | @item int obstack_chunk_size (struct obstack *@var{obstack-ptr}) | |
2148 | The size for allocating chunks. This is an lvalue. @xref{Obstack Chunks}. | |
2149 | ||
2150 | @item void *obstack_base (struct obstack *@var{obstack-ptr}) | |
2151 | Tentative starting address of the currently growing object. | |
2152 | @xref{Status of an Obstack}. | |
2153 | ||
2154 | @item void *obstack_next_free (struct obstack *@var{obstack-ptr}) | |
2155 | Address just after the end of the currently growing object. | |
2156 | @xref{Status of an Obstack}. | |
2157 | @end table | |
2158 | ||
2159 | @node Variable Size Automatic | |
99a20616 | 2160 | @subsection Automatic Storage with Variable Size |
28f540f4 RM |
2161 | @cindex automatic freeing |
2162 | @cindex @code{alloca} function | |
2163 | @cindex automatic storage with variable size | |
2164 | ||
2165 | The function @code{alloca} supports a kind of half-dynamic allocation in | |
2166 | which blocks are allocated dynamically but freed automatically. | |
2167 | ||
2168 | Allocating a block with @code{alloca} is an explicit action; you can | |
2169 | allocate as many blocks as you wish, and compute the size at run time. But | |
2170 | all the blocks are freed when you exit the function that @code{alloca} was | |
2171 | called from, just as if they were automatic variables declared in that | |
2172 | function. There is no way to free the space explicitly. | |
2173 | ||
2174 | The prototype for @code{alloca} is in @file{stdlib.h}. This function is | |
2175 | a BSD extension. | |
2176 | @pindex stdlib.h | |
2177 | ||
2178 | @comment stdlib.h | |
2179 | @comment GNU, BSD | |
cc6e48bc | 2180 | @deftypefun {void *} alloca (size_t @var{size}) |
28f540f4 | 2181 | The return value of @code{alloca} is the address of a block of @var{size} |
99a20616 | 2182 | bytes of memory, allocated in the stack frame of the calling function. |
28f540f4 RM |
2183 | @end deftypefun |
2184 | ||
2185 | Do not use @code{alloca} inside the arguments of a function call---you | |
2186 | will get unpredictable results, because the stack space for the | |
2187 | @code{alloca} would appear on the stack in the middle of the space for | |
2188 | the function arguments. An example of what to avoid is @code{foo (x, | |
2189 | alloca (4), y)}. | |
2190 | @c This might get fixed in future versions of GCC, but that won't make | |
2191 | @c it safe with compilers generally. | |
2192 | ||
2193 | @menu | |
2194 | * Alloca Example:: Example of using @code{alloca}. | |
2195 | * Advantages of Alloca:: Reasons to use @code{alloca}. | |
2196 | * Disadvantages of Alloca:: Reasons to avoid @code{alloca}. | |
2197 | * GNU C Variable-Size Arrays:: Only in GNU C, here is an alternative | |
2198 | method of allocating dynamically and | |
2199 | freeing automatically. | |
2200 | @end menu | |
2201 | ||
2202 | @node Alloca Example | |
99a20616 | 2203 | @subsubsection @code{alloca} Example |
28f540f4 | 2204 | |
bc938d3d UD |
2205 | As an example of the use of @code{alloca}, here is a function that opens |
2206 | a file name made from concatenating two argument strings, and returns a | |
2207 | file descriptor or minus one signifying failure: | |
28f540f4 RM |
2208 | |
2209 | @smallexample | |
2210 | int | |
2211 | open2 (char *str1, char *str2, int flags, int mode) | |
2212 | @{ | |
2213 | char *name = (char *) alloca (strlen (str1) + strlen (str2) + 1); | |
a5113b14 | 2214 | stpcpy (stpcpy (name, str1), str2); |
28f540f4 RM |
2215 | return open (name, flags, mode); |
2216 | @} | |
2217 | @end smallexample | |
2218 | ||
2219 | @noindent | |
2220 | Here is how you would get the same results with @code{malloc} and | |
2221 | @code{free}: | |
2222 | ||
2223 | @smallexample | |
2224 | int | |
2225 | open2 (char *str1, char *str2, int flags, int mode) | |
2226 | @{ | |
2227 | char *name = (char *) malloc (strlen (str1) + strlen (str2) + 1); | |
2228 | int desc; | |
2229 | if (name == 0) | |
2230 | fatal ("virtual memory exceeded"); | |
a5113b14 | 2231 | stpcpy (stpcpy (name, str1), str2); |
28f540f4 RM |
2232 | desc = open (name, flags, mode); |
2233 | free (name); | |
2234 | return desc; | |
2235 | @} | |
2236 | @end smallexample | |
2237 | ||
2238 | As you can see, it is simpler with @code{alloca}. But @code{alloca} has | |
2239 | other, more important advantages, and some disadvantages. | |
2240 | ||
2241 | @node Advantages of Alloca | |
99a20616 | 2242 | @subsubsection Advantages of @code{alloca} |
28f540f4 RM |
2243 | |
2244 | Here are the reasons why @code{alloca} may be preferable to @code{malloc}: | |
2245 | ||
2246 | @itemize @bullet | |
2247 | @item | |
2248 | Using @code{alloca} wastes very little space and is very fast. (It is | |
2249 | open-coded by the GNU C compiler.) | |
2250 | ||
2251 | @item | |
2252 | Since @code{alloca} does not have separate pools for different sizes of | |
2253 | block, space used for any size block can be reused for any other size. | |
99a20616 | 2254 | @code{alloca} does not cause memory fragmentation. |
28f540f4 RM |
2255 | |
2256 | @item | |
2257 | @cindex longjmp | |
2258 | Nonlocal exits done with @code{longjmp} (@pxref{Non-Local Exits}) | |
2259 | automatically free the space allocated with @code{alloca} when they exit | |
2260 | through the function that called @code{alloca}. This is the most | |
2261 | important reason to use @code{alloca}. | |
2262 | ||
2263 | To illustrate this, suppose you have a function | |
2264 | @code{open_or_report_error} which returns a descriptor, like | |
2265 | @code{open}, if it succeeds, but does not return to its caller if it | |
2266 | fails. If the file cannot be opened, it prints an error message and | |
2267 | jumps out to the command level of your program using @code{longjmp}. | |
2268 | Let's change @code{open2} (@pxref{Alloca Example}) to use this | |
2269 | subroutine:@refill | |
2270 | ||
2271 | @smallexample | |
2272 | int | |
2273 | open2 (char *str1, char *str2, int flags, int mode) | |
2274 | @{ | |
2275 | char *name = (char *) alloca (strlen (str1) + strlen (str2) + 1); | |
a5113b14 | 2276 | stpcpy (stpcpy (name, str1), str2); |
28f540f4 RM |
2277 | return open_or_report_error (name, flags, mode); |
2278 | @} | |
2279 | @end smallexample | |
2280 | ||
2281 | @noindent | |
99a20616 | 2282 | Because of the way @code{alloca} works, the memory it allocates is |
28f540f4 RM |
2283 | freed even when an error occurs, with no special effort required. |
2284 | ||
2285 | By contrast, the previous definition of @code{open2} (which uses | |
99a20616 | 2286 | @code{malloc} and @code{free}) would develop a memory leak if it were |
28f540f4 RM |
2287 | changed in this way. Even if you are willing to make more changes to |
2288 | fix it, there is no easy way to do so. | |
2289 | @end itemize | |
2290 | ||
2291 | @node Disadvantages of Alloca | |
99a20616 | 2292 | @subsubsection Disadvantages of @code{alloca} |
28f540f4 RM |
2293 | |
2294 | @cindex @code{alloca} disadvantages | |
2295 | @cindex disadvantages of @code{alloca} | |
2296 | These are the disadvantages of @code{alloca} in comparison with | |
2297 | @code{malloc}: | |
2298 | ||
2299 | @itemize @bullet | |
2300 | @item | |
99a20616 | 2301 | If you try to allocate more memory than the machine can provide, you |
28f540f4 RM |
2302 | don't get a clean error message. Instead you get a fatal signal like |
2303 | the one you would get from an infinite recursion; probably a | |
2304 | segmentation violation (@pxref{Program Error Signals}). | |
2305 | ||
2306 | @item | |
a7a93d50 | 2307 | Some @nongnusystems{} fail to support @code{alloca}, so it is less |
28f540f4 RM |
2308 | portable. However, a slower emulation of @code{alloca} written in C |
2309 | is available for use on systems with this deficiency. | |
2310 | @end itemize | |
2311 | ||
2312 | @node GNU C Variable-Size Arrays | |
99a20616 | 2313 | @subsubsection GNU C Variable-Size Arrays |
28f540f4 RM |
2314 | @cindex variable-sized arrays |
2315 | ||
2316 | In GNU C, you can replace most uses of @code{alloca} with an array of | |
2317 | variable size. Here is how @code{open2} would look then: | |
2318 | ||
2319 | @smallexample | |
2320 | int open2 (char *str1, char *str2, int flags, int mode) | |
2321 | @{ | |
2322 | char name[strlen (str1) + strlen (str2) + 1]; | |
a5113b14 | 2323 | stpcpy (stpcpy (name, str1), str2); |
28f540f4 RM |
2324 | return open (name, flags, mode); |
2325 | @} | |
2326 | @end smallexample | |
2327 | ||
2328 | But @code{alloca} is not always equivalent to a variable-sized array, for | |
2329 | several reasons: | |
2330 | ||
2331 | @itemize @bullet | |
2332 | @item | |
2333 | A variable size array's space is freed at the end of the scope of the | |
2334 | name of the array. The space allocated with @code{alloca} | |
2335 | remains until the end of the function. | |
2336 | ||
2337 | @item | |
2338 | It is possible to use @code{alloca} within a loop, allocating an | |
2339 | additional block on each iteration. This is impossible with | |
2340 | variable-sized arrays. | |
2341 | @end itemize | |
2342 | ||
48b22986 | 2343 | @strong{NB:} If you mix use of @code{alloca} and variable-sized arrays |
28f540f4 RM |
2344 | within one function, exiting a scope in which a variable-sized array was |
2345 | declared frees all blocks allocated with @code{alloca} during the | |
2346 | execution of that scope. | |
2347 | ||
99a20616 UD |
2348 | |
2349 | @node Resizing the Data Segment | |
2350 | @section Resizing the Data Segment | |
2351 | ||
2352 | The symbols in this section are declared in @file{unistd.h}. | |
2353 | ||
2354 | You will not normally use the functions in this section, because the | |
2355 | functions described in @ref{Memory Allocation} are easier to use. Those | |
1f77f049 | 2356 | are interfaces to a @glibcadj{} memory allocator that uses the |
99a20616 UD |
2357 | functions below itself. The functions below are simple interfaces to |
2358 | system calls. | |
2359 | ||
2360 | @comment unistd.h | |
2361 | @comment BSD | |
2362 | @deftypefun int brk (void *@var{addr}) | |
2363 | ||
2364 | @code{brk} sets the high end of the calling process' data segment to | |
2365 | @var{addr}. | |
2366 | ||
2367 | The address of the end of a segment is defined to be the address of the | |
2368 | last byte in the segment plus 1. | |
2369 | ||
2370 | The function has no effect if @var{addr} is lower than the low end of | |
2371 | the data segment. (This is considered success, by the way). | |
2372 | ||
2373 | The function fails if it would cause the data segment to overlap another | |
68979757 | 2374 | segment or exceed the process' data storage limit (@pxref{Limits on |
99a20616 UD |
2375 | Resources}). |
2376 | ||
2377 | The function is named for a common historical case where data storage | |
2378 | and the stack are in the same segment. Data storage allocation grows | |
2379 | upward from the bottom of the segment while the stack grows downward | |
2380 | toward it from the top of the segment and the curtain between them is | |
2381 | called the @dfn{break}. | |
2382 | ||
2383 | The return value is zero on success. On failure, the return value is | |
68979757 | 2384 | @code{-1} and @code{errno} is set accordingly. The following @code{errno} |
99a20616 UD |
2385 | values are specific to this function: |
2386 | ||
2387 | @table @code | |
2388 | @item ENOMEM | |
2389 | The request would cause the data segment to overlap another segment or | |
2390 | exceed the process' data storage limit. | |
2391 | @end table | |
2392 | ||
2393 | @c The Brk system call in Linux (as opposed to the GNU C Library function) | |
2394 | @c is considerably different. It always returns the new end of the data | |
2395 | @c segment, whether it succeeds or fails. The GNU C library Brk determines | |
bbf70ae9 | 2396 | @c it's a failure if and only if the system call returns an address less |
99a20616 UD |
2397 | @c than the address requested. |
2398 | ||
2399 | @end deftypefun | |
2400 | ||
2401 | ||
2402 | @comment unistd.h | |
2403 | @comment BSD | |
d6868416 | 2404 | @deftypefun void *sbrk (ptrdiff_t @var{delta}) |
99a20616 UD |
2405 | This function is the same as @code{brk} except that you specify the new |
2406 | end of the data segment as an offset @var{delta} from the current end | |
2407 | and on success the return value is the address of the resulting end of | |
2408 | the data segment instead of zero. | |
2409 | ||
2410 | This means you can use @samp{sbrk(0)} to find out what the current end | |
2411 | of the data segment is. | |
2412 | ||
2413 | @end deftypefun | |
2414 | ||
2415 | ||
2416 | ||
2417 | @node Locking Pages | |
2418 | @section Locking Pages | |
2419 | @cindex locking pages | |
2420 | @cindex memory lock | |
2421 | @cindex paging | |
2422 | ||
2423 | You can tell the system to associate a particular virtual memory page | |
11bf311e | 2424 | with a real page frame and keep it that way --- i.e., cause the page to |
99a20616 UD |
2425 | be paged in if it isn't already and mark it so it will never be paged |
2426 | out and consequently will never cause a page fault. This is called | |
2427 | @dfn{locking} a page. | |
2428 | ||
2429 | The functions in this chapter lock and unlock the calling process' | |
2430 | pages. | |
2431 | ||
2432 | @menu | |
2433 | * Why Lock Pages:: Reasons to read this section. | |
2434 | * Locked Memory Details:: Everything you need to know locked | |
2435 | memory | |
2436 | * Page Lock Functions:: Here's how to do it. | |
2437 | @end menu | |
2438 | ||
2439 | @node Why Lock Pages | |
2440 | @subsection Why Lock Pages | |
2441 | ||
2442 | Because page faults cause paged out pages to be paged in transparently, | |
68979757 | 2443 | a process rarely needs to be concerned about locking pages. However, |
99a20616 UD |
2444 | there are two reasons people sometimes are: |
2445 | ||
2446 | @itemize @bullet | |
2447 | ||
2448 | @item | |
2449 | Speed. A page fault is transparent only insofar as the process is not | |
2450 | sensitive to how long it takes to do a simple memory access. Time-critical | |
2451 | processes, especially realtime processes, may not be able to wait or | |
2452 | may not be able to tolerate variance in execution speed. | |
2453 | @cindex realtime processing | |
2454 | @cindex speed of execution | |
2455 | ||
2456 | A process that needs to lock pages for this reason probably also needs | |
2457 | priority among other processes for use of the CPU. @xref{Priority}. | |
2458 | ||
2459 | In some cases, the programmer knows better than the system's demand | |
2460 | paging allocator which pages should remain in real memory to optimize | |
2461 | system performance. In this case, locking pages can help. | |
2462 | ||
2463 | @item | |
2464 | Privacy. If you keep secrets in virtual memory and that virtual memory | |
2465 | gets paged out, that increases the chance that the secrets will get out. | |
2466 | If a password gets written out to disk swap space, for example, it might | |
2467 | still be there long after virtual and real memory have been wiped clean. | |
2468 | ||
2469 | @end itemize | |
2470 | ||
2471 | Be aware that when you lock a page, that's one fewer page frame that can | |
2472 | be used to back other virtual memory (by the same or other processes), | |
2473 | which can mean more page faults, which means the system runs more | |
2474 | slowly. In fact, if you lock enough memory, some programs may not be | |
2475 | able to run at all for lack of real memory. | |
2476 | ||
2477 | @node Locked Memory Details | |
2478 | @subsection Locked Memory Details | |
2479 | ||
2480 | A memory lock is associated with a virtual page, not a real frame. The | |
2481 | paging rule is: If a frame backs at least one locked page, don't page it | |
2482 | out. | |
2483 | ||
11bf311e | 2484 | Memory locks do not stack. I.e., you can't lock a particular page twice |
99a20616 UD |
2485 | so that it has to be unlocked twice before it is truly unlocked. It is |
2486 | either locked or it isn't. | |
2487 | ||
2488 | A memory lock persists until the process that owns the memory explicitly | |
2489 | unlocks it. (But process termination and exec cause the virtual memory | |
2490 | to cease to exist, which you might say means it isn't locked any more). | |
2491 | ||
2492 | Memory locks are not inherited by child processes. (But note that on a | |
2493 | modern Unix system, immediately after a fork, the parent's and the | |
2494 | child's virtual address space are backed by the same real page frames, | |
2495 | so the child enjoys the parent's locks). @xref{Creating a Process}. | |
2496 | ||
2497 | Because of its ability to impact other processes, only the superuser can | |
2498 | lock a page. Any process can unlock its own page. | |
2499 | ||
2500 | The system sets limits on the amount of memory a process can have locked | |
2501 | and the amount of real memory it can have dedicated to it. @xref{Limits | |
2502 | on Resources}. | |
2503 | ||
2504 | In Linux, locked pages aren't as locked as you might think. | |
2505 | Two virtual pages that are not shared memory can nonetheless be backed | |
2506 | by the same real frame. The kernel does this in the name of efficiency | |
2507 | when it knows both virtual pages contain identical data, and does it | |
68979757 | 2508 | even if one or both of the virtual pages are locked. |
99a20616 UD |
2509 | |
2510 | But when a process modifies one of those pages, the kernel must get it a | |
2511 | separate frame and fill it with the page's data. This is known as a | |
2512 | @dfn{copy-on-write page fault}. It takes a small amount of time and in | |
2513 | a pathological case, getting that frame may require I/O. | |
2514 | @cindex copy-on-write page fault | |
2515 | @cindex page fault, copy-on-write | |
2516 | ||
2517 | To make sure this doesn't happen to your program, don't just lock the | |
2518 | pages. Write to them as well, unless you know you won't write to them | |
2519 | ever. And to make sure you have pre-allocated frames for your stack, | |
2520 | enter a scope that declares a C automatic variable larger than the | |
2521 | maximum stack size you will need, set it to something, then return from | |
2522 | its scope. | |
2523 | ||
2524 | @node Page Lock Functions | |
2525 | @subsection Functions To Lock And Unlock Pages | |
2526 | ||
2527 | The symbols in this section are declared in @file{sys/mman.h}. These | |
2528 | functions are defined by POSIX.1b, but their availability depends on | |
2529 | your kernel. If your kernel doesn't allow these functions, they exist | |
2530 | but always fail. They @emph{are} available with a Linux kernel. | |
2531 | ||
2532 | @strong{Portability Note:} POSIX.1b requires that when the @code{mlock} | |
2533 | and @code{munlock} functions are available, the file @file{unistd.h} | |
2534 | define the macro @code{_POSIX_MEMLOCK_RANGE} and the file | |
2535 | @code{limits.h} define the macro @code{PAGESIZE} to be the size of a | |
2536 | memory page in bytes. It requires that when the @code{mlockall} and | |
2537 | @code{munlockall} functions are available, the @file{unistd.h} file | |
1f77f049 | 2538 | define the macro @code{_POSIX_MEMLOCK}. @Theglibc{} conforms to |
99a20616 UD |
2539 | this requirement. |
2540 | ||
2541 | @comment sys/mman.h | |
2542 | @comment POSIX.1b | |
2543 | @deftypefun int mlock (const void *@var{addr}, size_t @var{len}) | |
2544 | ||
2545 | @code{mlock} locks a range of the calling process' virtual pages. | |
2546 | ||
2547 | The range of memory starts at address @var{addr} and is @var{len} bytes | |
2548 | long. Actually, since you must lock whole pages, it is the range of | |
2549 | pages that include any part of the specified range. | |
2550 | ||
2551 | When the function returns successfully, each of those pages is backed by | |
2552 | (connected to) a real frame (is resident) and is marked to stay that | |
2553 | way. This means the function may cause page-ins and have to wait for | |
2554 | them. | |
2555 | ||
2556 | When the function fails, it does not affect the lock status of any | |
2557 | pages. | |
2558 | ||
2559 | The return value is zero if the function succeeds. Otherwise, it is | |
2560 | @code{-1} and @code{errno} is set accordingly. @code{errno} values | |
2561 | specific to this function are: | |
2562 | ||
2563 | @table @code | |
2564 | @item ENOMEM | |
2565 | @itemize @bullet | |
2566 | @item | |
2567 | At least some of the specified address range does not exist in the | |
2568 | calling process' virtual address space. | |
2569 | @item | |
2570 | The locking would cause the process to exceed its locked page limit. | |
2571 | @end itemize | |
2572 | ||
2573 | @item EPERM | |
2574 | The calling process is not superuser. | |
2575 | ||
2576 | @item EINVAL | |
2577 | @var{len} is not positive. | |
2578 | ||
2579 | @item ENOSYS | |
2580 | The kernel does not provide @code{mlock} capability. | |
2581 | ||
2582 | @end table | |
2583 | ||
2584 | You can lock @emph{all} a process' memory with @code{mlockall}. You | |
2585 | unlock memory with @code{munlock} or @code{munlockall}. | |
2586 | ||
2587 | To avoid all page faults in a C program, you have to use | |
2588 | @code{mlockall}, because some of the memory a program uses is hidden | |
2589 | from the C code, e.g. the stack and automatic variables, and you | |
2590 | wouldn't know what address to tell @code{mlock}. | |
2591 | ||
2592 | @end deftypefun | |
2593 | ||
2594 | @comment sys/mman.h | |
2595 | @comment POSIX.1b | |
2596 | @deftypefun int munlock (const void *@var{addr}, size_t @var{len}) | |
2597 | ||
10e0498e | 2598 | @code{munlock} unlocks a range of the calling process' virtual pages. |
99a20616 UD |
2599 | |
2600 | @code{munlock} is the inverse of @code{mlock} and functions completely | |
2601 | analogously to @code{mlock}, except that there is no @code{EPERM} | |
2602 | failure. | |
2603 | ||
2604 | @end deftypefun | |
2605 | ||
2606 | @comment sys/mman.h | |
2607 | @comment POSIX.1b | |
2608 | @deftypefun int mlockall (int @var{flags}) | |
2609 | ||
2610 | @code{mlockall} locks all the pages in a process' virtual memory address | |
2611 | space, and/or any that are added to it in the future. This includes the | |
2612 | pages of the code, data and stack segment, as well as shared libraries, | |
2613 | user space kernel data, shared memory, and memory mapped files. | |
2614 | ||
2615 | @var{flags} is a string of single bit flags represented by the following | |
2616 | macros. They tell @code{mlockall} which of its functions you want. All | |
2617 | other bits must be zero. | |
2618 | ||
2619 | @table @code | |
2620 | ||
2621 | @item MCL_CURRENT | |
2622 | Lock all pages which currently exist in the calling process' virtual | |
2623 | address space. | |
2624 | ||
2625 | @item MCL_FUTURE | |
2626 | Set a mode such that any pages added to the process' virtual address | |
2627 | space in the future will be locked from birth. This mode does not | |
2628 | affect future address spaces owned by the same process so exec, which | |
2629 | replaces a process' address space, wipes out @code{MCL_FUTURE}. | |
2630 | @xref{Executing a File}. | |
2631 | ||
2632 | @end table | |
2633 | ||
2634 | When the function returns successfully, and you specified | |
2635 | @code{MCL_CURRENT}, all of the process' pages are backed by (connected | |
2636 | to) real frames (they are resident) and are marked to stay that way. | |
2637 | This means the function may cause page-ins and have to wait for them. | |
2638 | ||
2639 | When the process is in @code{MCL_FUTURE} mode because it successfully | |
2640 | executed this function and specified @code{MCL_CURRENT}, any system call | |
2641 | by the process that requires space be added to its virtual address space | |
2642 | fails with @code{errno} = @code{ENOMEM} if locking the additional space | |
2643 | would cause the process to exceed its locked page limit. In the case | |
0bc93a2f | 2644 | that the address space addition that can't be accommodated is stack |
99a20616 UD |
2645 | expansion, the stack expansion fails and the kernel sends a |
2646 | @code{SIGSEGV} signal to the process. | |
2647 | ||
2648 | When the function fails, it does not affect the lock status of any pages | |
2649 | or the future locking mode. | |
2650 | ||
2651 | The return value is zero if the function succeeds. Otherwise, it is | |
2652 | @code{-1} and @code{errno} is set accordingly. @code{errno} values | |
2653 | specific to this function are: | |
2654 | ||
2655 | @table @code | |
2656 | @item ENOMEM | |
2657 | @itemize @bullet | |
2658 | @item | |
2659 | At least some of the specified address range does not exist in the | |
2660 | calling process' virtual address space. | |
2661 | @item | |
2662 | The locking would cause the process to exceed its locked page limit. | |
2663 | @end itemize | |
2664 | ||
2665 | @item EPERM | |
2666 | The calling process is not superuser. | |
2667 | ||
2668 | @item EINVAL | |
2669 | Undefined bits in @var{flags} are not zero. | |
2670 | ||
2671 | @item ENOSYS | |
2672 | The kernel does not provide @code{mlockall} capability. | |
2673 | ||
2674 | @end table | |
2675 | ||
2676 | You can lock just specific pages with @code{mlock}. You unlock pages | |
2677 | with @code{munlockall} and @code{munlock}. | |
2678 | ||
2679 | @end deftypefun | |
2680 | ||
2681 | ||
2682 | @comment sys/mman.h | |
2683 | @comment POSIX.1b | |
2684 | @deftypefun int munlockall (void) | |
2685 | ||
2686 | @code{munlockall} unlocks every page in the calling process' virtual | |
2687 | address space and turn off @code{MCL_FUTURE} future locking mode. | |
2688 | ||
2689 | The return value is zero if the function succeeds. Otherwise, it is | |
68979757 | 2690 | @code{-1} and @code{errno} is set accordingly. The only way this |
99a20616 UD |
2691 | function can fail is for generic reasons that all functions and system |
2692 | calls can fail, so there are no specific @code{errno} values. | |
2693 | ||
2694 | @end deftypefun | |
2695 | ||
2696 | ||
2697 | ||
2698 | ||
a9ddb793 UD |
2699 | @ignore |
2700 | @c This was never actually implemented. -zw | |
28f540f4 RM |
2701 | @node Relocating Allocator |
2702 | @section Relocating Allocator | |
2703 | ||
2704 | @cindex relocating memory allocator | |
2705 | Any system of dynamic memory allocation has overhead: the amount of | |
2706 | space it uses is more than the amount the program asks for. The | |
2707 | @dfn{relocating memory allocator} achieves very low overhead by moving | |
2708 | blocks in memory as necessary, on its own initiative. | |
2709 | ||
a9ddb793 UD |
2710 | @c @menu |
2711 | @c * Relocator Concepts:: How to understand relocating allocation. | |
2712 | @c * Using Relocator:: Functions for relocating allocation. | |
2713 | @c @end menu | |
28f540f4 RM |
2714 | |
2715 | @node Relocator Concepts | |
2716 | @subsection Concepts of Relocating Allocation | |
2717 | ||
2718 | @ifinfo | |
2719 | The @dfn{relocating memory allocator} achieves very low overhead by | |
2720 | moving blocks in memory as necessary, on its own initiative. | |
2721 | @end ifinfo | |
2722 | ||
2723 | When you allocate a block with @code{malloc}, the address of the block | |
2724 | never changes unless you use @code{realloc} to change its size. Thus, | |
2725 | you can safely store the address in various places, temporarily or | |
2726 | permanently, as you like. This is not safe when you use the relocating | |
2727 | memory allocator, because any and all relocatable blocks can move | |
2728 | whenever you allocate memory in any fashion. Even calling @code{malloc} | |
2729 | or @code{realloc} can move the relocatable blocks. | |
2730 | ||
2731 | @cindex handle | |
2732 | For each relocatable block, you must make a @dfn{handle}---a pointer | |
2733 | object in memory, designated to store the address of that block. The | |
2734 | relocating allocator knows where each block's handle is, and updates the | |
2735 | address stored there whenever it moves the block, so that the handle | |
2736 | always points to the block. Each time you access the contents of the | |
2737 | block, you should fetch its address anew from the handle. | |
2738 | ||
2739 | To call any of the relocating allocator functions from a signal handler | |
2740 | is almost certainly incorrect, because the signal could happen at any | |
2741 | time and relocate all the blocks. The only way to make this safe is to | |
2742 | block the signal around any access to the contents of any relocatable | |
2743 | block---not a convenient mode of operation. @xref{Nonreentrancy}. | |
2744 | ||
2745 | @node Using Relocator | |
2746 | @subsection Allocating and Freeing Relocatable Blocks | |
2747 | ||
2748 | @pindex malloc.h | |
2749 | In the descriptions below, @var{handleptr} designates the address of the | |
2750 | handle. All the functions are declared in @file{malloc.h}; all are GNU | |
2751 | extensions. | |
2752 | ||
2753 | @comment malloc.h | |
2754 | @comment GNU | |
a9ddb793 | 2755 | @c @deftypefun {void *} r_alloc (void **@var{handleptr}, size_t @var{size}) |
28f540f4 RM |
2756 | This function allocates a relocatable block of size @var{size}. It |
2757 | stores the block's address in @code{*@var{handleptr}} and returns | |
2758 | a non-null pointer to indicate success. | |
2759 | ||
2760 | If @code{r_alloc} can't get the space needed, it stores a null pointer | |
2761 | in @code{*@var{handleptr}}, and returns a null pointer. | |
2762 | @end deftypefun | |
2763 | ||
2764 | @comment malloc.h | |
2765 | @comment GNU | |
a9ddb793 | 2766 | @c @deftypefun void r_alloc_free (void **@var{handleptr}) |
28f540f4 RM |
2767 | This function is the way to free a relocatable block. It frees the |
2768 | block that @code{*@var{handleptr}} points to, and stores a null pointer | |
2769 | in @code{*@var{handleptr}} to show it doesn't point to an allocated | |
2770 | block any more. | |
2771 | @end deftypefun | |
2772 | ||
2773 | @comment malloc.h | |
2774 | @comment GNU | |
a9ddb793 | 2775 | @c @deftypefun {void *} r_re_alloc (void **@var{handleptr}, size_t @var{size}) |
28f540f4 RM |
2776 | The function @code{r_re_alloc} adjusts the size of the block that |
2777 | @code{*@var{handleptr}} points to, making it @var{size} bytes long. It | |
2778 | stores the address of the resized block in @code{*@var{handleptr}} and | |
2779 | returns a non-null pointer to indicate success. | |
2780 | ||
2781 | If enough memory is not available, this function returns a null pointer | |
2782 | and does not modify @code{*@var{handleptr}}. | |
2783 | @end deftypefun | |
a9ddb793 | 2784 | @end ignore |
28f540f4 | 2785 | |
99a20616 UD |
2786 | |
2787 | ||
2788 | ||
c131718c UD |
2789 | @ignore |
2790 | @comment No longer available... | |
2791 | ||
2792 | @comment @node Memory Warnings | |
2793 | @comment @section Memory Usage Warnings | |
2794 | @comment @cindex memory usage warnings | |
2795 | @comment @cindex warnings of memory almost full | |
28f540f4 RM |
2796 | |
2797 | @pindex malloc.c | |
2798 | You can ask for warnings as the program approaches running out of memory | |
2799 | space, by calling @code{memory_warnings}. This tells @code{malloc} to | |
2800 | check memory usage every time it asks for more memory from the operating | |
2801 | system. This is a GNU extension declared in @file{malloc.h}. | |
2802 | ||
2803 | @comment malloc.h | |
2804 | @comment GNU | |
c131718c | 2805 | @comment @deftypefun void memory_warnings (void *@var{start}, void (*@var{warn-func}) (const char *)) |
28f540f4 RM |
2806 | Call this function to request warnings for nearing exhaustion of virtual |
2807 | memory. | |
2808 | ||
2809 | The argument @var{start} says where data space begins, in memory. The | |
2810 | allocator compares this against the last address used and against the | |
2811 | limit of data space, to determine the fraction of available memory in | |
2812 | use. If you supply zero for @var{start}, then a default value is used | |
2813 | which is right in most circumstances. | |
2814 | ||
2815 | For @var{warn-func}, supply a function that @code{malloc} can call to | |
2816 | warn you. It is called with a string (a warning message) as argument. | |
2817 | Normally it ought to display the string for the user to read. | |
2818 | @end deftypefun | |
2819 | ||
2820 | The warnings come when memory becomes 75% full, when it becomes 85% | |
2821 | full, and when it becomes 95% full. Above 95% you get another warning | |
2822 | each time memory usage increases. | |
c131718c UD |
2823 | |
2824 | @end ignore |