1 .\" Copyright (C) 1996 Andries Brouwer <aeb@cwi.nl>
2 .\" and Copyright (C) 2006, 2007 Michael Kerrisk <mtk.manpages@gmail.com>
4 .\" %%%LICENSE_START(VERBATIM)
5 .\" Permission is granted to make and distribute verbatim copies of this
6 .\" manual provided the copyright notice and this permission notice are
7 .\" preserved on all copies.
9 .\" Permission is granted to copy and distribute modified versions of this
10 .\" manual under the conditions for verbatim copying, provided that the
11 .\" entire resulting derived work is distributed under the terms of a
12 .\" permission notice identical to this one.
14 .\" Since the Linux kernel and libraries are constantly changing, this
15 .\" manual page may be incorrect or out-of-date. The author(s) assume no
16 .\" responsibility for errors or omissions, or for damages resulting from
17 .\" the use of the information contained herein. The author(s) may not
18 .\" have taken the same level of care in the production of this manual,
19 .\" which is licensed free of charge, as they might when working
22 .\" Formatted or processed versions of this manual, if unaccompanied by
23 .\" the source, must acknowledge the copyright and authors of this work.
26 .\" Modified 1997-01-31 by Eric S. Raymond <esr@thyrsus.com>
27 .\" Modified 2000-03-25 by Jim Van Zandt <jrv@vanzandt.mv.com>
28 .\" Modified 2001-10-04 by John Levon <moz@compsoc.man.ac.uk>
29 .\" Modified 2003-02-02 by Andi Kleen <ak@muc.de>
30 .\" Modified 2003-05-21 by Michael Kerrisk <mtk.manpages@gmail.com>
31 .\" MAP_LOCKED works from 2.5.37
32 .\" Modified 2004-06-17 by Michael Kerrisk <mtk.manpages@gmail.com>
33 .\" Modified 2004-09-11 by aeb
34 .\" Modified 2004-12-08, from Eric Estievenart <eric.estievenart@free.fr>
35 .\" Modified 2004-12-08, mtk, formatting tidy-ups
36 .\" Modified 2006-12-04, mtk, various parts rewritten
37 .\" 2007-07-10, mtk, Added an example program.
38 .\" 2008-11-18, mtk, document MAP_STACK
40 .TH MMAP 2 2016-07-17 "Linux" "Linux Programmer's Manual"
42 mmap, munmap \- map or unmap files or devices into memory
45 .B #include <sys/mman.h>
47 .BI "void *mmap(void *" addr ", size_t " length \
48 ", int " prot ", int " flags ,
49 .BI " int " fd ", off_t " offset );
50 .BI "int munmap(void *" addr ", size_t " length );
53 See NOTES for information on feature test macro requirements.
56 creates a new mapping in the virtual address space of
58 The starting address for the new mapping is specified in
62 argument specifies the length of the mapping.
67 then the kernel chooses the address at which to create the mapping;
68 this is the most portable method of creating a new mapping.
72 then the kernel takes it as a hint about where to place the mapping;
73 on Linux, the mapping will be created at a nearby page boundary.
74 .\" Before Linux 2.6.24, the address was rounded up to the next page
75 .\" boundary; since 2.6.24, it is rounded down!
76 The address of the new mapping is returned as the result of the call.
78 The contents of a file mapping (as opposed to an anonymous mapping; see
80 below), are initialized using
82 bytes starting at offset
84 in the file (or other object) referred to by the file descriptor
87 must be a multiple of the page size as returned by
88 .IR sysconf(_SC_PAGE_SIZE) .
92 argument describes the desired memory protection of the mapping
93 (and must not conflict with the open mode of the file).
96 or the bitwise OR of one or more of the following flags:
99 Pages may be executed.
105 Pages may be written.
108 Pages may not be accessed.
112 argument determines whether updates to the mapping
113 are visible to other processes mapping the same region,
114 and whether updates are carried through to the underlying file.
115 This behavior is determined by including exactly one
116 of the following values in
121 Updates to the mapping are visible to other processes that map this file,
122 and are carried through to the underlying file.
123 (To precisely control when updates are carried through
124 to the underlying file requires the use of
128 Create a private copy-on-write mapping.
129 Updates to the mapping are not visible to other processes
130 mapping the same file, and are not carried through to
132 It is unspecified whether changes made to the file after the
134 call are visible in the mapped region.
136 Both of these flags are described in POSIX.1-2001 and POSIX.1-2008.
138 In addition, zero or more of the following values can be ORed in
141 .BR MAP_32BIT " (since Linux 2.4.20, 2.6)"
142 Put the mapping into the first 2 Gigabytes of the process address space.
143 This flag is supported only on x86-64, for 64-bit programs.
144 It was added to allow thread stacks to be allocated somewhere
145 in the first 2GB of memory,
146 so as to improve context-switch performance on some early
148 .\" See http://lwn.net/Articles/294642 "Tangled up in threads", 19 Aug 08
149 Modern x86-64 processors no longer have this performance problem,
150 so use of this flag is not required on those systems.
163 The mapping is not backed by any file;
164 its contents are initialized to zero.
169 arguments are ignored;
170 however, some implementations require
177 and portable applications should ensure this.
182 is supported on Linux only since kernel 2.4.
185 This flag is ignored.
186 .\" Introduced in 1.1.36, removed in 1.3.24.
187 (Long ago, it signaled that attempts to write to the underlying file
190 But this was a source of denial-of-service attacks.)
193 This flag is ignored.
194 .\" Introduced in 1.1.38, removed in 1.3.24. Flag tested in proc_follow_link.
195 .\" (Long ago, it signaled that the underlying file is an executable.
196 .\" However, that information was not really used anywhere.)
197 .\" Linus talked about DOS related to MAP_EXECUTABLE, but he was thinking of
203 .\" On some systems, this was required as the opposite of
204 .\" MAP_ANONYMOUS -- mtk, 1 May 2007
209 as a hint: place the mapping at exactly that address.
211 must be a multiple of the page size.
212 If the memory region specified by
216 overlaps pages of any existing mapping(s), then the overlapped
217 part of the existing mapping(s) will be discarded.
218 If the specified address cannot be used,
221 Because requiring a fixed address for a mapping is less portable,
222 the use of this option is discouraged.
226 Indicates to the kernel virtual memory system that the mapping
227 should extend downward in memory.
229 .BR MAP_HUGETLB " (since Linux 2.6.32)"
230 Allocate the mapping using "huge pages."
231 See the Linux kernel source file
232 .I Documentation/vm/hugetlbpage.txt
233 for further information, as well as NOTES, below.
235 .BR MAP_HUGE_2MB ", " MAP_HUGE_1GB " (since Linux 3.8)"
236 .\" See https://lwn.net/Articles/533499/
237 Used in conjunction with
239 to select alternative hugetlb page sizes (respectively, 2 MB and 1 GB)
240 on systems that support multiple hugetlb page sizes.
242 More generally, the desired huge page size can be configured by encoding
243 the base-2 logarithm of the desired page size in the six bits at the offset
245 (A value of zero in this bit field provides the default huge page size;
246 the default huge page size can be discovered vie the
250 Thus, the above two constants are defined as:
254 #define MAP_HUGE_2MB (21 << MAP_HUGE_SHIFT)
255 #define MAP_HUGE_1GB (30 << MAP_HUGE_SHIFT)
259 The range of huge page sizes that are supported by the system
260 can be discovered by listing the subdirectories in
261 .IR /sys/kernel/mm/hugepages .
263 .BR MAP_LOCKED " (since Linux 2.5.37)"
264 Mark the mmaped region to be locked in the same way as
266 This implementation will try to populate (prefault) the whole range but
267 the mmap call doesn't fail with
270 Therefore major faults might happen later on.
271 So the semantic is not as strong as
277 when major faults are not acceptable after the initialization of the mapping.
280 flag is ignored in older kernels.
281 .\" If set, the mapped pages will not be swapped out.
283 .BR MAP_NONBLOCK " (since Linux 2.5.46)"
284 Only meaningful in conjunction with
286 Don't perform read-ahead:
287 create page tables entries only for pages
288 that are already present in RAM.
289 Since Linux 2.6.23, this flag causes
292 One day, the combination of
296 may be reimplemented.
299 Do not reserve swap space for this mapping.
300 When swap space is reserved, one has the guarantee
301 that it is possible to modify the mapping.
302 When swap space is not reserved one might get
305 if no physical memory is available.
306 See also the discussion of the file
307 .I /proc/sys/vm/overcommit_memory
310 In kernels before 2.6, this flag had effect only for
311 private writable mappings.
313 .BR MAP_POPULATE " (since Linux 2.5.46)"
314 Populate (prefault) page tables for a mapping.
315 For a file mapping, this causes read-ahead on the file.
316 This will help to reduce blocking on page faults later.
318 is supported for private mappings only since Linux 2.6.23.
320 .BR MAP_STACK " (since Linux 2.6.27)"
321 Allocate the mapping at an address suitable for a process
323 This flag is currently a no-op,
324 but is used in the glibc threading implementation so that
325 if some architectures require special treatment for stack allocations,
326 support can later be transparently implemented for glibc.
327 .\" See http://lwn.net/Articles/294642 "Tangled up in threads", 19 Aug 08
328 .\" commit cd98a04a59e2f94fa64d5bf1e26498d27427d5e7
329 .\" http://thread.gmane.org/gmane.linux.kernel/720412
330 .\" "pthread_create() slow for many threads; also time to revisit 64b
331 .\" context switch optimization?"
333 .BR MAP_UNINITIALIZED " (since Linux 2.6.33)"
334 Don't clear anonymous pages.
335 This flag is intended to improve performance on embedded devices.
336 This flag is honored only if the kernel was configured with the
337 .B CONFIG_MMAP_ALLOW_UNINITIALIZED
339 Because of the security implications,
340 that option is normally enabled only on embedded devices
341 (i.e., devices where one has complete control of the contents of user memory).
343 Of the above flags, only
345 is specified in POSIX.1-2001 and POSIX.1-2008.
346 However, most systems also support
351 Some systems document the additional flags
362 with the same attributes.
364 A file is mapped in multiples of the page size.
365 For a file that is not
366 a multiple of the page size, the remaining memory is zeroed when mapped,
367 and writes to that region are not written out to the file.
369 changing the size of the underlying file of a mapping on the pages that
370 correspond to added or removed regions of the file is unspecified.
374 system call deletes the mappings for the specified address range, and
375 causes further references to addresses within the range to generate
376 invalid memory references.
377 The region is also automatically unmapped
378 when the process is terminated.
379 On the other hand, closing the file
380 descriptor does not unmap the region.
384 must be a multiple of the page size (but
387 All pages containing a part
388 of the indicated range are unmapped, and subsequent references
389 to these pages will generate
391 It is not an error if the
392 indicated range does not contain any mapped pages.
396 returns a pointer to the mapped area.
400 .IR "(void\ *)\ \-1" )
403 is set to indicate the cause of the error.
408 On failure, it returns \-1, and
410 is set to indicate the cause of the error (probably to
415 A file descriptor refers to a non-regular file.
416 Or a file mapping was requested, but
418 is not open for reading.
425 is not open in read/write
430 is set, but the file is append-only.
433 The file has been locked, or too much memory has been locked (see
438 is not a valid file descriptor (and
448 (e.g., they are too large, or not aligned on a page boundary).
461 or contained both of these values.
464 .\" This is for shared anonymous segments
465 .\" [2.6.7] shmem_zero_setup()-->shmem_file_setup()-->get_empty_filp()
466 The system-wide limit on the total number of open files has been reached.
469 .\" A file could not be mapped for reading.
472 The underlying filesystem of the specified file does not support
476 No memory is available.
479 The process's maximum number of mappings would have been exceeded.
480 This error can also occur for
482 when unmapping a region in the middle of an existing mapping,
483 since this results in two smaller mappings on either side of
484 the region being unmapped.
491 but the mapped area belongs to a file on a filesystem that
493 .\" (Since 2.4.25 / 2.6.0.)
496 The operation was prevented by a file seal; see
501 was set but the object specified by
506 On 32-bit architecture together with the large file extension
509 the number of pages used for
511 plus number of pages used for
517 Use of a mapped region can result in these signals:
520 Attempted write into a region mapped as read-only.
523 Attempted access to a portion of the buffer that does not correspond
524 to the file (for example, beyond the end of the file, including the
525 case where another process has truncated the file).
527 For an explanation of the terms used in this section, see
533 Interface Attribute Value
537 T} Thread safety MT-Safe
540 POSIX.1-2001, POSIX.1-2008, SVr4, 4.4BSD.
541 .\" SVr4 documents additional error codes ENXIO and ENODEV.
542 .\" SUSv2 documents additional error codes EMFILE and EOVERFLOW.
544 On POSIX systems on which
550 .B _POSIX_MAPPED_FILES
551 is defined in \fI<unistd.h>\fP to a value greater than 0.
554 .\" POSIX.1-2001: It shall be defined to -1 or 0 or 200112L.
555 .\" -1: unavailable, 0: ask using sysconf().
556 .\" glibc defines it to 1.
558 On some hardware architectures (e.g., i386),
562 It is architecture dependent whether
567 Portable programs should always set
569 if they intend to execute code in the new mapping.
571 The portable way to create a mapping is to specify
573 as 0 (NULL), and omit
577 In this case, the system chooses the address for the mapping;
578 the address is chosen so as not to conflict with any existing mapping,
582 flag is specified, and
584 is 0 (NULL), then the mapped address will be 0 (NULL).
588 constants are defined only if suitable feature test macros are defined
589 (possibly by default):
591 with glibc 2.19 or later;
596 in glibc 2.19 and earlier.
600 and requiring that macro specifically would have been more logical,
601 since these flags are all Linux-specific.)
602 The relevant flags are:
619 .SS Timestamps changes for file-backed mappings
620 For file-backed mappings, the
622 field for the mapped file may be updated at any time between the
624 and the corresponding unmapping; the first reference to a mapped
625 page will update the field if it has not been already.
631 field for a file mapped with
635 will be updated after
636 a write to the mapped region, and before a subsequent
644 .SS Huge page (Huge TLB) mappings
645 For mappings that employ huge pages, the requirements for the arguments of
649 differ somewhat from the requirements for mappings
650 that use the native system page size.
655 must be a multiple of the underlying huge page size.
656 The system automatically aligns
658 to be a multiple of the underlying huge page size.
665 must both be a multiple of the underlying huge page size.
667 .SS C library/kernel differences
668 This page describes the interface provided by the glibc
671 Originally, this function invoked a system call of the same name.
672 Since kernel 2.4, that system call has been superseded by
675 .\" Since around glibc 2.1/2.2, depending on the platform.
678 wrapper function invokes
680 with a suitably adjusted value for
683 On Linux, there are no guarantees like those suggested above under
685 By default, any process can be killed
686 at any moment when the system runs out of memory.
688 In kernels before 2.6.7, the
690 flag has effect only if
700 However, in kernels before 2.6.12,
702 succeeded in this case: no mapping was created and the call returned
710 POSIX specifies that the system shall always
711 zero fill any partial page at the end
712 of the object and that system will never write any modification of the
713 object beyond its end.
714 On Linux, when you write data to such partial page after the end
715 of the object, the data stays in the page cache even after the file
716 is closed and unmapped
717 and even though the data is never written to the file itself,
718 subsequent mappings may see the modified content.
719 In some cases, this could be fixed by calling
721 before the unmap takes place;
722 however, this doesn't work on tmpfs
723 (for example, when using POSIX shared memory interface documented in
724 .BR shm_overview (7)).
726 .\" FIXME . Add an example here that uses an anonymous shared region for
727 .\" IPC between parent and child.
729 The following program prints part of the file specified in
730 its first command-line argument to standard output.
731 The range of bytes to be printed is specified via offset and length
732 values in the second and third command-line arguments.
733 The program creates a memory mapping of the required
734 pages of the file and then uses
736 to output the desired bytes.
739 #include <sys/mman.h>
740 #include <sys/stat.h>
746 #define handle_error(msg) \\
747 do { perror(msg); exit(EXIT_FAILURE); } while (0)
750 main(int argc, char *argv[])
755 off_t offset, pa_offset;
759 if (argc < 3 || argc > 4) {
760 fprintf(stderr, "%s file offset [length]\\n", argv[0]);
764 fd = open(argv[1], O_RDONLY);
766 handle_error("open");
768 if (fstat(fd, &sb) == \-1) /* To obtain file size */
769 handle_error("fstat");
771 offset = atoi(argv[2]);
772 pa_offset = offset & ~(sysconf(_SC_PAGE_SIZE) \- 1);
773 /* offset for mmap() must be page aligned */
775 if (offset >= sb.st_size) {
776 fprintf(stderr, "offset is past end of file\\n");
781 length = atoi(argv[3]);
782 if (offset + length > sb.st_size)
783 length = sb.st_size \- offset;
784 /* Can\(aqt display bytes past end of file */
786 } else { /* No length arg ==> display to end of file */
787 length = sb.st_size \- offset;
790 addr = mmap(NULL, length + offset \- pa_offset, PROT_READ,
791 MAP_PRIVATE, fd, pa_offset);
792 if (addr == MAP_FAILED)
793 handle_error("mmap");
795 s = write(STDOUT_FILENO, addr + offset \- pa_offset, length);
798 handle_error("write");
800 fprintf(stderr, "partial write");
804 munmap(addr, length + offset \- pa_offset);
812 .BR memfd_create (2),
819 .BR remap_file_pages (2),
825 The descriptions of the following files in
827 .IR /proc/[pid]/maps ,
828 .IR /proc/[pid]/map_files ,
830 .IR /proc/[pid]/smaps .
832 B.O. Gallmeister, POSIX.4, O'Reilly, pp. 128-129 and 389-391.
834 .\" Repeat after me: private read-only mappings are 100% equivalent to
835 .\" shared read-only mappings. No ifs, buts, or maybes. -- Linus