1MMAP(2) Linux Programmer's Manual MMAP(2)
2
6 mmap, munmap - map or unmap files or devices into memory
7
9 #include <sys/mman.h>
10 void *mmap(void *start, size_t length, int prot, int flags,
12 int fd, off_t offset);
13 int munmap(void *start, size_t length);
15
17 mmap() creates a new mapping in the virtual address space of the call‐
18 ing process. The starting address for the new mapping is specified in
19 start. The length argument specifies the length of the mapping.
20 If start is NULL, then the kernel chooses the address at which to cre‐
22 ate the mapping; this is the most portable method of creating a new
23 mapping. If start is not NULL, then the kernel takes it as a hint
24 about where to place the mapping; on Linux, the mapping will be created
25 at the next higher page boundary. The address of the new mapping is
26 returned as the result of the call.
27 The contents of a file mapping (as opposed to an anonymous mapping; see
29 MAP_ANONYMOUS below), are initialised using length bytes starting at
30 offset offset in the file (or other object) referred to by the file
31 descriptor fd. offset must be a multiple of the page size as returned
32 by sysconf(_SC_PAGE_SIZE).
33 The prot argument describes the desired memory protection of the map‐
35 ping (and must not conflict with the open mode of the file). It is
36 either PROT_NONE or the bitwise OR of one or more of the following
37 flags:
38 PROT_EXEC Pages may be executed.
40 PROT_READ Pages may be read.
42 PROT_WRITE Pages may be written.
44 PROT_NONE Pages may not be accessed.
46 The flags argument determines whether updates to the mapping are visi‐
48 ble to other processes mapping the same region, and whether updates are
49 caried through to the underlying file. This behaviour is determined by
50 including exactly one of the following values in flags:
51 MAP_SHARED Share this mapping. Updates to the mapping are visible to
53 other processes that map this file, and are carried through
54 to the underlying file. The file may not actually be
55 updated until msync(2) or munmap(2) is called.
56 MAP_PRIVATE
58 Create a private copy-on-write mapping. Updates to the map‐
59 ping are not visible to other processes mapping the same
60 file, and are not carried through to the underlying file.
61 It is unspecified whether changes made to the file after the
62 mmap() call are visible in the mapped region.
63 Both of these flags are described in POSIX.1-2001.
65 In addition, zero or more of the following values can be ORed in flags:
67 MAP_FIXED
69 Don't interpret start as a hint: place the mapping at exactly
70 that address. start must be a multiple of the page size. If
71 the memory region specified by start and len overlaps pages of
72 any existing mapping(s), then the overlapped part of the exist‐
73 ing mapping(s) will be discarded. If the specified address can‐
74 not be used, mmap() will fail. Because requiring a fixed
75 address for a mapping is less portable, the use of this option
76 is discouraged.
77 MAP_DENYWRITE
79 This flag is ignored. (Long ago, it signalled that attempts to
80 write to the underlying file should fail with ETXTBUSY. But
81 this was a source of denial-of-service attacks.)
82 MAP_EXECUTABLE
84 This flag is ignored.
85 MAP_NORESERVE
87 Do not reserve swap space for this mapping. When swap space is
88 reserved, one has the guarantee that it is possible to modify
89 the mapping. When swap space is not reserved one might get
90 SIGSEGV upon a write if no physical memory is available. See
91 also the discussion of the file /proc/sys/vm/overcommit_memory
92 in proc(5). In kernels before 2.6, this flag only had effect
93 for private writable mappings.
94 MAP_LOCKED (since Linux 2.5.37)
96 Lock the pages of the mapped region into memory in the manner of
97 mlock(). This flag is ignored in older kernels.
98 MAP_GROWSDOWN
100 Used for stacks. Indicates to the kernel virtual memory system
101 that the mapping should extend downwards in memory.
102 MAP_ANON
104 Synonym for MAP_ANONYMOUS. Deprecated.
105 MAP_ANONYMOUS
107 The mapping is not backed by any file; its contents are ini‐
108 tialised to zero. The fd and offset arguments are ignored; how‐
109 ever, some implementations require fd to be -1 if MAP_ANONYMOUS
110 (or MAP_ANON) is specified, and portable applications should
111 ensure this. The use of MAP_ANONYMOUS in conjunction with
112 MAP_SHARED is only supported on Linux since kernel 2.4.
113 MAP_FILE
115 Compatibility flag. Ignored.
116 MAP_32BIT
118 Put the mapping into the first 2GB of the process address space.
119 Ignored when MAP_FIXED is set. This flag is currently only sup‐
120 ported on x86-64 for 64bit programs.
121 MAP_POPULATE (since Linux 2.5.46)
123 Populate (prefault) page tables for a file mapping, by perform‐
124 ing read-ahead on the file. Later accesses to the mapping will
125 not be blocked by page faults.
126 MAP_NONBLOCK (since Linux 2.5.46)
128 Only meaningful in conjunction with MAP_POPULATE. Don't perform
129 read-ahead: only create page tables entries for pages that are
130 already present in RAM.
131 Of the above flags, only MAP_FIXED is specified in POSIX.1-2001. How‐
133 ever, most systems also support MAP_ANONYMOUS (or its synonym
134 MAP_ANON).
135 Some systems document the additional flags MAP_AUTOGROW, MAP_AUTORESRV,
137 MAP_COPY, and MAP_LOCAL.
138 Memory mapped by mmap() is preserved across fork(2), with the same
140 attributes.
141 A file is mapped in multiples of the page size. For a file that is not
143 a multiple of the page size, the remaining memory is zeroed when
144 mapped, and writes to that region are not written out to the file. The
145 effect of changing the size of the underlying file of a mapping on the
146 pages that correspond to added or removed regions of the file is
147 unspecified.
148 The munmap() system call deletes the mappings for the specified address
150 range, and causes further references to addresses within the range to
151 generate invalid memory references. The region is also automatically
152 unmapped when the process is terminated. On the other hand, closing
153 the file descriptor does not unmap the region.
154 If MAP_FIXED is specified, start must be a multiple of the page size.
156 In all other cases start address is rounded up to the next page size
157 boundary. All pages containing a part of the indicated range are
158 unmapped, and subsequent references to these pages will generate
159 SIGSEGV. It is not an error if the indicated range does not contain any
160 mapped pages.
161 For file-backed mappings, the st_atime field for the mapped file may be
163 updated at any time between the mmap() and the corresponding unmapping;
164 the first reference to a mapped page will update the field if it has
165 not been already.
166 The st_ctime and st_mtime field for a file mapped with PROT_WRITE and
168 MAP_SHARED will be updated after a write to the mapped region, and
169 before a subsequent msync() with the MS_SYNC or MS_ASYNC flag, if one
170 occurs.
171
173 On success, mmap() returns a pointer to the mapped area. On error, the
174 value MAP_FAILED (that is, (void *) -1) is returned, and errno is set
175 appropriately. On success, munmap() returns 0, on failure -1, and
176 errno is set (probably to EINVAL).
177
179 It is architecture dependent whether PROT_READ implies PROT_EXEC or
180 not. Portable programs should always set PROT_EXEC if they intend to
181 execute code in the new mapping.
182
184 EACCES A file descriptor refers to a non-regular file. Or MAP_PRIVATE
185 was requested, but fd is not open for reading. Or MAP_SHARED
186 was requested and PROT_WRITE is set, but fd is not open in
187 read/write (O_RDWR) mode. Or PROT_WRITE is set, but the file is
188 append-only.
189 EAGAIN The file has been locked, or too much memory has been locked
191 (see setrlimit(2)).
192 EBADF fd is not a valid file descriptor (and MAP_ANONYMOUS was not
194 set).
195 EINVAL We don't like start, length, or offset (e.g., they are too
197 large, or not aligned on a page boundary).
198 EINVAL (since Linux 2.6.12), length was 0.
200 EINVAL flags contained neither MAP_PRIVATE or MAP_SHARED, or contained
202 both of these values.
203 ENFILE The system limit on the total number of open files has been
205 reached.
206 ENODEV The underlying filesystem of the specified file does not support
208 memory mapping.
209 ENOMEM No memory is available, or the process's maximum number of map‐
211 pings would have been exceeded.
212 EPERM The prot argument asks for PROT_EXEC but the mapped area belongs
214 to a file on a filesystem that was mounted no-exec.
215 ETXTBSY
217 MAP_DENYWRITE was set but the object specified by fd is open for
218 writing.
219 Use of a mapped region can result in these signals:
221 SIGSEGV
223 Attempted write into a region mapped as read-only.
224 SIGBUS Attempted access to a portion of the buffer that does not corre‐
226 spond to the file (for example, beyond the end of the file,
227 including the case where another process has truncated the
228 file).
229
231 On POSIX systems on which mmap(), msync() and munmap() are available,
232 _POSIX_MAPPED_FILES is defined in <unistd.h> to a value greater than 0.
233 (See also sysconf(3).)
234
236 SVr4, 4.4BSD, POSIX.1-2001.
237
239 On Linux there are no guarantees like those suggested above under
240 MAP_NORESERVE. By default, any process can be killed at any moment
241 when the system runs out of memory.
242 In kernels before 2.6.7, the MAP_POPULATE flag only has effect if prot
244 is specified as PROT_NONE.
245 SUSv3 specifies that mmap() should fail if length is 0. However, in
247 kernels before 2.6.12, mmap() succeeded in this case: no mapping was
248 created and the call returned start. Since kernel 2.6.12, mmap() fails
249 with the error EINVAL for this case.
250
252 getpagesize(2), mincore(2), mlock(2), mmap2(2), mremap(2), msync(2),
253 remap_file_pages(2), setrlimit(2), shm_open(3)
254 B.O. Gallmeister, POSIX.4, O'Reilly, pp. 128-129 and 389-391.
255Linux 2.6.19 2006-12-04 MMAP(2)