1MMAP(P) POSIX Programmer's Manual MMAP(P)
2
3
4
6 mmap - map pages of memory
7
9 #include <sys/mman.h>
10
11 void *mmap(void *addr, size_t len, int prot, int flags,
12 int fildes, off_t off);
13
14
16 The mmap() function shall establish a mapping between a process'
17 address space and a file, shared memory object, or typed memory
18 object. The format of the call is as follows:
19
20
21 pa=mmap(addr, len, prot, flags, fildes, off);
22
23 The mmap() function shall establish a mapping between the address space
24 of the process at an address pa for len bytes to the memory object rep‐
25 resented by the file descriptor fildes at offset off for len bytes.
26 The value of pa is an implementation-defined function of the parameter
27 addr and the values of flags, further described below. A successful
28 mmap() call shall return pa as its result. The address range starting
29 at pa and continuing for len bytes shall be legitimate for the possible
30 (not necessarily current) address space of the process. The range of
31 bytes starting at off and continuing for len bytes shall be legitimate
32 for the possible (not necessarily current) offsets in the file, shared
33 memory object, or typed memory object represented by fildes.
34
35 If fildes represents a typed memory object opened with either the
36 POSIX_TYPED_MEM_ALLOCATE flag or the POSIX_TYPED_MEM_ALLOCATE_CONTIG
37 flag, the memory object to be mapped shall be that portion of the typed
38 memory object allocated by the implementation as specified below. In
39 this case, if off is non-zero, the behavior of mmap() is undefined. If
40 fildes refers to a valid typed memory object that is not accessible
41 from the calling process, mmap() shall fail.
42
43 The mapping established by mmap() shall replace any previous mappings
44 for those whole pages containing any part of the address space of the
45 process starting at pa and continuing for len bytes.
46
47 If the size of the mapped file changes after the call to mmap() as a
48 result of some other operation on the mapped file, the effect of refer‐
49 ences to portions of the mapped region that correspond to added or
50 removed portions of the file is unspecified.
51
52 The mmap() function shall be supported for regular files, shared memory
53 objects, and typed memory objects. Support for any other type of
54 file is unspecified.
55
56 The parameter prot determines whether read, write, execute, or some
57 combination of accesses are permitted to the data being mapped. The
58 prot shall be either PROT_NONE or the bitwise-inclusive OR of one or
59 more of the other flags in the following table, defined in the
60 <sys/mman.h> header.
61
62 Symbolic Constant Description
63 PROT_READ Data can be read.
64 PROT_WRITE Data can be written.
65 PROT_EXEC Data can be executed.
66
67 PROT_NONE Data cannot be accessed.
68
69 If an implementation cannot support the combination of access types
70 specified by prot, the call to mmap() shall fail.
71
72 An implementation may permit accesses other than those specified by
73 prot; however, if the Memory Protection option is supported, the
74 implementation shall not permit a write to succeed where PROT_WRITE has
75 not been set or shall not permit any access where PROT_NONE alone has
76 been set. The implementation shall support at least the following val‐
77 ues of prot: PROT_NONE, PROT_READ, PROT_WRITE, and the bitwise-inclu‐
78 sive OR of PROT_READ and PROT_WRITE. If the Memory Protection option
79 is not supported, the result of any access that conflicts with the
80 specified protection is undefined. The file descriptor fildes shall
81 have been opened with read permission, regardless of the protection
82 options specified. If PROT_WRITE is specified, the application shall
83 ensure that it has opened the file descriptor fildes with write permis‐
84 sion unless MAP_PRIVATE is specified in the flags parameter as
85 described below.
86
87 The parameter flags provides other information about the handling of
88 the mapped data. The value of flags is the bitwise-inclusive OR of
89 these options, defined in <sys/mman.h>:
90
91 Symbolic Constant Description
92 MAP_SHARED Changes are shared.
93 MAP_PRIVATE Changes are private.
94 MAP_FIXED Interpret addr exactly.
95
96 Implementations that do not support the Memory Mapped Files option are
97 not required to support MAP_PRIVATE.
98
99 It is implementation-defined whether MAP_FIXED shall be supported.
100 MAP_FIXED shall be supported on XSI-conformant systems.
101
102 MAP_SHARED and MAP_PRIVATE describe the disposition of write references
103 to the memory object. If MAP_SHARED is specified, write references
104 shall change the underlying object. If MAP_PRIVATE is specified, modi‐
105 fications to the mapped data by the calling process shall be visible
106 only to the calling process and shall not change the underlying object.
107 It is unspecified whether modifications to the underlying object done
108 after the MAP_PRIVATE mapping is established are visible through the
109 MAP_PRIVATE mapping. Either MAP_SHARED or MAP_PRIVATE can be specified,
110 but not both. The mapping type is retained across fork().
111
112 When fildes represents a typed memory object opened with either the
113 POSIX_TYPED_MEM_ALLOCATE flag or the POSIX_TYPED_MEM_ALLOCATE_CONTIG
114 flag, mmap() shall, if there are enough resources available, map len
115 bytes allocated from the corresponding typed memory object which were
116 not previously allocated to any process in any processor that may
117 access that typed memory object. If there are not enough resources
118 available, the function shall fail. If fildes represents a typed memory
119 object opened with the POSIX_TYPED_MEM_ALLOCATE_CONTIG flag, these
120 allocated bytes shall be contiguous within the typed memory object. If
121 fildes represents a typed memory object opened with the
122 POSIX_TYPED_MEM_ALLOCATE flag, these allocated bytes may be composed of
123 non-contiguous fragments within the typed memory object. If fildes rep‐
124 resents a typed memory object opened with neither the
125 POSIX_TYPED_MEM_ALLOCATE_CONTIG flag nor the POSIX_TYPED_MEM_ALLOCATE
126 flag, len bytes starting at offset off within the typed memory object
127 are mapped, exactly as when mapping a file or shared memory object. In
128 this case, if two processes map an area of typed memory using the same
129 off and len values and using file descriptors that refer to the same
130 memory pool (either from the same port or from a different port), both
131 processes shall map the same region of storage.
132
133 When MAP_FIXED is set in the flags argument, the implementation is
134 informed that the value of pa shall be addr, exactly. If MAP_FIXED is
135 set, mmap() may return MAP_FAILED and set errno to [EINVAL]. If a
136 MAP_FIXED request is successful, the mapping established by mmap()
137 replaces any previous mappings for the process' pages in the range
138 [pa,pa+len).
139
140 When MAP_FIXED is not set, the implementation uses addr in an implemen‐
141 tation-defined manner to arrive at pa. The pa so chosen shall be an
142 area of the address space that the implementation deems suitable for a
143 mapping of len bytes to the file. All implementations interpret an addr
144 value of 0 as granting the implementation complete freedom in selecting
145 pa, subject to constraints described below. A non-zero value of addr is
146 taken to be a suggestion of a process address near which the mapping
147 should be placed. When the implementation selects a value for pa, it
148 never places a mapping at address 0, nor does it replace any extant
149 mapping.
150
151 The off argument is constrained to be aligned and sized according to
152 the value returned by sysconf() when passed _SC_PAGESIZE or
153 _SC_PAGE_SIZE. When MAP_FIXED is specified, the application shall
154 ensure that the argument addr also meets these constraints. The imple‐
155 mentation performs mapping operations over whole pages. Thus, while the
156 argument len need not meet a size or alignment constraint, the imple‐
157 mentation shall include, in any mapping operation, any partial page
158 specified by the range [pa,pa+len).
159
160 The system shall always zero-fill any partial page at the end of an
161 object. Further, the system shall never write out any modified portions
162 of the last page of an object which are beyond its end. References
163 within the address range starting at pa and continuing for len bytes to
164 whole pages following the end of an object shall result in delivery of
165 a SIGBUS signal.
166
167 An implementation may generate SIGBUS signals when a reference would
168 cause an error in the mapped object, such as out-of-space condition.
169
170 The mmap() function shall add an extra reference to the file associated
171 with the file descriptor fildes which is not removed by a subsequent
172 close() on that file descriptor. This reference shall be removed when
173 there are no more mappings to the file.
174
175 The st_atime field of the mapped file may be marked for update at any
176 time between the mmap() call and the corresponding munmap() call. The
177 initial read or write reference to a mapped region shall cause the
178 file's st_atime field to be marked for update if it has not already
179 been marked for update.
180
181 The st_ctime and st_mtime fields of a file that is mapped with
182 MAP_SHARED and PROT_WRITE shall be marked for update at some point in
183 the interval between a write reference to the mapped region and the
184 next call to msync() with MS_ASYNC or MS_SYNC for that portion of the
185 file by any process. If there is no such call and if the underlying
186 file is modified as a result of a write reference, then these fields
187 shall be marked for update at some time after the write reference.
188
189 There may be implementation-defined limits on the number of memory
190 regions that can be mapped (per process or per system).
191
192 If such a limit is imposed, whether the number of memory regions that
193 can be mapped by a process is decreased by the use of shmat() is imple‐
194 mentation-defined.
195
196 If mmap() fails for reasons other than [EBADF], [EINVAL], or [ENOTSUP],
197 some of the mappings in the address range starting at addr and continu‐
198 ing for len bytes may have been unmapped.
199
201 Upon successful completion, the mmap() function shall return the
202 address at which the mapping was placed ( pa); otherwise, it shall
203 return a value of MAP_FAILED and set errno to indicate the error. The
204 symbol MAP_FAILED is defined in the <sys/mman.h> header. No successful
205 return from mmap() shall return the value MAP_FAILED.
206
208 The mmap() function shall fail if:
209
210 EACCES The fildes argument is not open for read, regardless of the pro‐
211 tection specified, or fildes is not open for write and
212 PROT_WRITE was specified for a MAP_SHARED type mapping.
213
214 EAGAIN The mapping could not be locked in memory, if required by mlock‐
215 all(), due to a lack of resources.
216
217 EBADF The fildes argument is not a valid open file descriptor.
218
219 EINVAL The addr argument (if MAP_FIXED was specified) or off is not a
220 multiple of the page size as returned by sysconf(), or is con‐
221 sidered invalid by the implementation.
222
223 EINVAL The value of flags is invalid (neither MAP_PRIVATE nor
224 MAP_SHARED is set).
225
226 EMFILE The number of mapped regions would exceed an implementation-
227 defined limit (per process or per system).
228
229 ENODEV The fildes argument refers to a file whose type is not supported
230 by mmap().
231
232 ENOMEM MAP_FIXED was specified, and the range [addr,addr+len) exceeds
233 that allowed for the address space of a process; or, if
234 MAP_FIXED was not specified and there is insufficient room in
235 the address space to effect the mapping.
236
237 ENOMEM The mapping could not be locked in memory, if required by mlock‐
238 all(), because it would require more space than the system is
239 able to supply.
240
241 ENOMEM Not enough unallocated memory resources remain in the typed mem‐
242 ory object designated by fildes to allocate len bytes.
243
244 ENOTSUP
245 MAP_FIXED or MAP_PRIVATE was specified in the flags argument and
246 the implementation does not support this functionality.
247
248 The implementation does not support the combination of accesses
249 requested in the prot argument.
250
251 ENXIO Addresses in the range [off,off+len) are invalid for the object
252 specified by fildes.
253
254 ENXIO MAP_FIXED was specified in flags and the combination of addr,
255 len, and off is invalid for the object specified by fildes.
256
257 ENXIO The fildes argument refers to a typed memory object that is not
258 accessible from the calling process.
259
260 EOVERFLOW
261 The file is a regular file and the value of off plus len exceeds
262 the offset maximum established in the open file description
263 associated with fildes.
264
265
266 The following sections are informative.
267
269 None.
270
272 Use of mmap() may reduce the amount of memory available to other memory
273 allocation functions.
274
275 Use of MAP_FIXED may result in unspecified behavior in further use of
276 malloc() and shmat(). The use of MAP_FIXED is discouraged, as it may
277 prevent an implementation from making the most effective use of
278 resources.
279
280 The application must ensure correct synchronization when using mmap()
281 in conjunction with any other file access method, such as read() and
282 write(), standard input/output, and shmat().
283
284 The mmap() function allows access to resources via address space manip‐
285 ulations, instead of read()/ write(). Once a file is mapped, all a
286 process has to do to access it is use the data at the address to which
287 the file was mapped. So, using pseudo-code to illustrate the way in
288 which an existing program might be changed to use mmap(), the follow‐
289 ing:
290
291
292 fildes = open(...)
293 lseek(fildes, some_offset)
294 read(fildes, buf, len)
295 /* Use data in buf. */
296
297 becomes:
298
299
300 fildes = open(...)
301 address = mmap(0, len, PROT_READ, MAP_PRIVATE, fildes, some_offset)
302 /* Use data at address. */
303
305 After considering several other alternatives, it was decided to adopt
306 the mmap() definition found in SVR4 for mapping memory objects into
307 process address spaces. The SVR4 definition is minimal, in that it
308 describes only what has been built, and what appears to be necessary
309 for a general and portable mapping facility.
310
311 Note that while mmap() was first designed for mapping files, it is
312 actually a general-purpose mapping facility. It can be used to map any
313 appropriate object, such as memory, files, devices, and so on, into the
314 address space of a process.
315
316 When a mapping is established, it is possible that the implementation
317 may need to map more than is requested into the address space of the
318 process because of hardware requirements. An application, however, can‐
319 not count on this behavior. Implementations that do not use a paged
320 architecture may simply allocate a common memory region and return the
321 address of it; such implementations probably do not allocate any more
322 than is necessary. References past the end of the requested area are
323 unspecified.
324
325 If an application requests a mapping that would overlay existing map‐
326 pings in the process, it might be desirable that an implementation
327 detect this and inform the application. However, the default, portable
328 (not MAP_FIXED) operation does not overlay existing mappings. On the
329 other hand, if the program specifies a fixed address mapping (which
330 requires some implementation knowledge to determine a suitable address,
331 if the function is supported at all), then the program is presumed to
332 be successfully managing its own address space and should be trusted
333 when it asks to map over existing data structures. Furthermore, it is
334 also desirable to make as few system calls as possible, and it might be
335 considered onerous to require an munmap() before an mmap() to the same
336 address range. This volume of IEEE Std 1003.1-2001 specifies that the
337 new mappings replace any existing mappings, following existing practice
338 in this regard.
339
340 It is not expected, when the Memory Protection option is supported,
341 that all hardware implementations are able to support all combinations
342 of permissions at all addresses. When this option is supported, imple‐
343 mentations are required to disallow write access to mappings without
344 write permission and to disallow access to mappings without any access
345 permission. Other than these restrictions, implementations may allow
346 access types other than those requested by the application. For exam‐
347 ple, if the application requests only PROT_WRITE, the implementation
348 may also allow read access. A call to mmap() fails if the implementa‐
349 tion cannot support allowing all the access requested by the applica‐
350 tion. For example, some implementations cannot support a request for
351 both write access and execute access simultaneously. All implementa‐
352 tions supporting the Memory Protection option must support requests for
353 no access, read access, write access, and both read and write access.
354 Strictly conforming code must only rely on the required checks. These
355 restrictions allow for portability across a wide range of hardware.
356
357 The MAP_FIXED address treatment is likely to fail for non-page-aligned
358 values and for certain architecture-dependent address ranges. Conform‐
359 ing implementations cannot count on being able to choose address values
360 for MAP_FIXED without utilizing non-portable, implementation-defined
361 knowledge. Nonetheless, MAP_FIXED is provided as a standard interface
362 conforming to existing practice for utilizing such knowledge when it is
363 available.
364
365 Similarly, in order to allow implementations that do not support vir‐
366 tual addresses, support for directly specifying any mapping addresses
367 via MAP_FIXED is not required and thus a conforming application may not
368 count on it.
369
370 The MAP_PRIVATE function can be implemented efficiently when memory
371 protection hardware is available. When such hardware is not available,
372 implementations can implement such "mappings" by simply making a real
373 copy of the relevant data into process private memory, though this
374 tends to behave similarly to read().
375
376 The function has been defined to allow for many different models of
377 using shared memory. However, all uses are not equally portable across
378 all machine architectures. In particular, the mmap() function allows
379 the system as well as the application to specify the address at which
380 to map a specific region of a memory object. The most portable way to
381 use the function is always to let the system choose the address, speci‐
382 fying NULL as the value for the argument addr and not to specify
383 MAP_FIXED.
384
385 If it is intended that a particular region of a memory object be mapped
386 at the same address in a group of processes (on machines where this is
387 even possible), then MAP_FIXED can be used to pass in the desired map‐
388 ping address. The system can still be used to choose the desired
389 address if the first such mapping is made without specifying MAP_FIXED,
390 and then the resulting mapping address can be passed to subsequent pro‐
391 cesses for them to pass in via MAP_FIXED. The availability of a spe‐
392 cific address range cannot be guaranteed, in general.
393
394 The mmap() function can be used to map a region of memory that is
395 larger than the current size of the object. Memory access within the
396 mapping but beyond the current end of the underlying objects may result
397 in SIGBUS signals being sent to the process. The reason for this is
398 that the size of the object can be manipulated by other processes and
399 can change at any moment. The implementation should tell the applica‐
400 tion that a memory reference is outside the object where this can be
401 detected; otherwise, written data may be lost and read data may not
402 reflect actual data in the object.
403
404 Note that references beyond the end of the object do not extend the
405 object as the new end cannot be determined precisely by most virtual
406 memory hardware. Instead, the size can be directly manipulated by
407 ftruncate().
408
409 Process memory locking does apply to shared memory regions, and the
410 MEMLOCK_FUTURE argument to mlockall() can be relied upon to cause new
411 shared memory regions to be automatically locked.
412
413 Existing implementations of mmap() return the value -1 when unsuccess‐
414 ful. Since the casting of this value to type void * cannot be guaran‐
415 teed by the ISO C standard to be distinct from a successful value, this
416 volume of IEEE Std 1003.1-2001 defines the symbol MAP_FAILED, which a
417 conforming implementation does not return as the result of a successful
418 call.
419
421 None.
422
424 exec() , fcntl() , fork() , lockf() , msync() , munmap() , mprotect() ,
425 posix_typed_mem_open() , shmat() , sysconf() , the Base Definitions
426 volume of IEEE Std 1003.1-2001, <sys/mman.h>
427
429 Portions of this text are reprinted and reproduced in electronic form
430 from IEEE Std 1003.1, 2003 Edition, Standard for Information Technology
431 -- Portable Operating System Interface (POSIX), The Open Group Base
432 Specifications Issue 6, Copyright (C) 2001-2003 by the Institute of
433 Electrical and Electronics Engineers, Inc and The Open Group. In the
434 event of any discrepancy between this version and the original IEEE and
435 The Open Group Standard, the original IEEE and The Open Group Standard
436 is the referee document. The original Standard can be obtained online
437 at http://www.opengroup.org/unix/online.html .
438
439
440
441IEEE/The Open Group 2003 MMAP(P)