xref: /freebsd-12.1/lib/libc/sys/mmap.2 (revision dd5ef26e)
1.\" Copyright (c) 1991, 1993
2.\"	The Regents of the University of California.  All rights reserved.
3.\"
4.\" Redistribution and use in source and binary forms, with or without
5.\" modification, are permitted provided that the following conditions
6.\" are met:
7.\" 1. Redistributions of source code must retain the above copyright
8.\"    notice, this list of conditions and the following disclaimer.
9.\" 2. Redistributions in binary form must reproduce the above copyright
10.\"    notice, this list of conditions and the following disclaimer in the
11.\"    documentation and/or other materials provided with the distribution.
12.\" 3. Neither the name of the University nor the names of its contributors
13.\"    may be used to endorse or promote products derived from this software
14.\"    without specific prior written permission.
15.\"
16.\" THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
17.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19.\" ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
20.\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26.\" SUCH DAMAGE.
27.\"
28.\"	@(#)mmap.2	8.4 (Berkeley) 5/11/95
29.\" $FreeBSD$
30.\"
31.Dd June 22, 2017
32.Dt MMAP 2
33.Os
34.Sh NAME
35.Nm mmap
36.Nd allocate memory, or map files or devices into memory
37.Sh LIBRARY
38.Lb libc
39.Sh SYNOPSIS
40.In sys/mman.h
41.Ft void *
42.Fn mmap "void *addr" "size_t len" "int prot" "int flags" "int fd" "off_t offset"
43.Sh DESCRIPTION
44The
45.Fn mmap
46system call causes the pages starting at
47.Fa addr
48and continuing for at most
49.Fa len
50bytes to be mapped from the object described by
51.Fa fd ,
52starting at byte offset
53.Fa offset .
54If
55.Fa len
56is not a multiple of the page size, the mapped region may extend past the
57specified range.
58Any such extension beyond the end of the mapped object will be zero-filled.
59.Pp
60If
61.Fa fd
62references a regular file or a shared memory object, the range of
63bytes starting at
64.Fa offset
65and continuing for
66.Fa len
67bytes must be legitimate for the possible (not necessarily
68current) offsets in the object.
69In particular, the
70.Fa offset
71value cannot be negative.
72If the object is truncated and the process later accesses a page that
73is wholly within the truncated region, the access is aborted and a
74.Dv SIGBUS
75signal is delivered to the process.
76.Pp
77If
78.Fa fd
79references a device file, the interpretation of the
80.Fa offset
81value is device specific and defined by the device driver.
82The virtual memory subsystem does not impose any restrictitions on the
83.Fa offset
84value in this case, passing it unchanged to the driver.
85.Pp
86If
87.Fa addr
88is non-zero, it is used as a hint to the system.
89(As a convenience to the system, the actual address of the region may differ
90from the address supplied.)
91If
92.Fa addr
93is zero, an address will be selected by the system.
94The actual starting address of the region is returned.
95A successful
96.Fa mmap
97deletes any previous mapping in the allocated address range.
98.Pp
99The protections (region accessibility) are specified in the
100.Fa prot
101argument by
102.Em or Ns 'ing
103the following values:
104.Pp
105.Bl -tag -width PROT_WRITE -compact
106.It Dv PROT_NONE
107Pages may not be accessed.
108.It Dv PROT_READ
109Pages may be read.
110.It Dv PROT_WRITE
111Pages may be written.
112.It Dv PROT_EXEC
113Pages may be executed.
114.El
115.Pp
116The
117.Fa flags
118argument specifies the type of the mapped object, mapping options and
119whether modifications made to the mapped copy of the page are private
120to the process or are to be shared with other references.
121Sharing, mapping type and options are specified in the
122.Fa flags
123argument by
124.Em or Ns 'ing
125the following values:
126.Bl -tag -width MAP_PREFAULT_READ
127.It Dv MAP_32BIT
128Request a region in the first 2GB of the current process's address space.
129If a suitable region cannot be found,
130.Fn mmap
131will fail.
132This flag is only available on 64-bit platforms.
133.It Dv MAP_ALIGNED Ns Pq Fa n
134Align the region on a requested boundary.
135If a suitable region cannot be found,
136.Fn mmap
137will fail.
138The
139.Fa n
140argument specifies the binary logarithm of the desired alignment.
141.It Dv MAP_ALIGNED_SUPER
142Align the region to maximize the potential use of large
143.Pq Dq super
144pages.
145If a suitable region cannot be found,
146.Fn mmap
147will fail.
148The system will choose a suitable page size based on the size of
149mapping.
150The page size used as well as the alignment of the region may both be
151affected by properties of the file being mapped.
152In particular,
153the physical address of existing pages of a file may require a specific
154alignment.
155The region is not guaranteed to be aligned on any specific boundary.
156.It Dv MAP_ANON
157Map anonymous memory not associated with any specific file.
158The file descriptor used for creating
159.Dv MAP_ANON
160must be \-1.
161The
162.Fa offset
163argument must be 0.
164.\".It Dv MAP_FILE
165.\"Mapped from a regular file or character-special device memory.
166.It Dv MAP_ANONYMOUS
167This flag is identical to
168.Dv MAP_ANON
169and is provided for compatibility.
170.It Dv MAP_EXCL
171This flag can only be used in combination with
172.Dv MAP_FIXED .
173Please see the definition of
174.Dv MAP_FIXED
175for the description of its effect.
176.It Dv MAP_FIXED
177Do not permit the system to select a different address than the one
178specified.
179If the specified address cannot be used,
180.Fn mmap
181will fail.
182If
183.Dv MAP_FIXED
184is specified,
185.Fa addr
186must be a multiple of the page size.
187If
188.Dv MAP_EXCL
189is not specified, a successful
190.Dv MAP_FIXED
191request replaces any previous mappings for the process'
192pages in the range from
193.Fa addr
194to
195.Fa addr
196+
197.Fa len .
198In contrast, if
199.Dv MAP_EXCL
200is specified, the request will fail if a mapping
201already exists within the range.
202.It Dv MAP_GUARD
203Instead of a mapping, create a guard of the specified size.
204Guards allow a process to create reservations in its address space,
205which can later be replaced by actual mappings.
206.Pp
207.Fa mmap
208will not create mappings in the address range of a guard unless
209the request specifies
210.Dv MAP_FIXED .
211Guards can be destroyed with
212.Xr munmap 2 .
213Any memory access by a thread to the guarded range results
214in the delivery of a
215.Dv SIGSEGV
216signal to that thread.
217.It Dv MAP_NOCORE
218Region is not included in a core file.
219.It Dv MAP_NOSYNC
220Causes data dirtied via this VM map to be flushed to physical media
221only when necessary (usually by the pager) rather than gratuitously.
222Typically this prevents the update daemons from flushing pages dirtied
223through such maps and thus allows efficient sharing of memory across
224unassociated processes using a file-backed shared memory map.
225Without
226this option any VM pages you dirty may be flushed to disk every so often
227(every 30-60 seconds usually) which can create performance problems if you
228do not need that to occur (such as when you are using shared file-backed
229mmap regions for IPC purposes).
230Dirty data will be flushed automatically when all mappings of an object are
231removed and all descriptors referencing the object are closed.
232Note that VM/file system coherency is
233maintained whether you use
234.Dv MAP_NOSYNC
235or not.
236This option is not portable
237across
238.Ux
239platforms (yet), though some may implement the same behavior
240by default.
241.Pp
242.Em WARNING !
243Extending a file with
244.Xr ftruncate 2 ,
245thus creating a big hole, and then filling the hole by modifying a shared
246.Fn mmap
247can lead to severe file fragmentation.
248In order to avoid such fragmentation you should always pre-allocate the
249file's backing store by
250.Fn write Ns ing
251zero's into the newly extended area prior to modifying the area via your
252.Fn mmap .
253The fragmentation problem is especially sensitive to
254.Dv MAP_NOSYNC
255pages, because pages may be flushed to disk in a totally random order.
256.Pp
257The same applies when using
258.Dv MAP_NOSYNC
259to implement a file-based shared memory store.
260It is recommended that you create the backing store by
261.Fn write Ns ing
262zero's to the backing file rather than
263.Fn ftruncate Ns ing
264it.
265You can test file fragmentation by observing the KB/t (kilobytes per
266transfer) results from an
267.Dq Li iostat 1
268while reading a large file sequentially, e.g.,\& using
269.Dq Li dd if=filename of=/dev/null bs=32k .
270.Pp
271The
272.Xr fsync 2
273system call will flush all dirty data and metadata associated with a file,
274including dirty NOSYNC VM data, to physical media.
275The
276.Xr sync 8
277command and
278.Xr sync 2
279system call generally do not flush dirty NOSYNC VM data.
280The
281.Xr msync 2
282system call is usually not needed since
283.Bx
284implements a coherent file system buffer cache.
285However, it may be
286used to associate dirty VM pages with file system buffers and thus cause
287them to be flushed to physical media sooner rather than later.
288.It Dv MAP_PREFAULT_READ
289Immediately update the calling process's lowest-level virtual address
290translation structures, such as its page table, so that every memory
291resident page within the region is mapped for read access.
292Ordinarily these structures are updated lazily.
293The effect of this option is to eliminate any soft faults that would
294otherwise occur on the initial read accesses to the region.
295Although this option does not preclude
296.Fa prot
297from including
298.Dv PROT_WRITE ,
299it does not eliminate soft faults on the initial write accesses to the
300region.
301.It Dv MAP_PRIVATE
302Modifications are private.
303.It Dv MAP_SHARED
304Modifications are shared.
305.It Dv MAP_STACK
306.Dv MAP_STACK
307implies
308.Dv MAP_ANON ,
309and
310.Fa offset
311of 0.
312The
313.Fa fd
314argument
315must be -1 and
316.Fa prot
317must include at least
318.Dv PROT_READ
319and
320.Dv PROT_WRITE .
321.Pp
322This option creates
323a memory region that grows to at most
324.Fa len
325bytes in size, starting from the stack top and growing down.
326The
327stack top is the starting address returned by the call, plus
328.Fa len
329bytes.
330The bottom of the stack at maximum growth is the starting
331address returned by the call.
332.Pp
333Stacks created with
334.Dv MAP_STACK
335automatically grow.
336Guards prevent inadvertent use of the regions into which those
337stacks can grow without requiring mapping the whole stack in advance.
338.El
339.Pp
340The
341.Xr close 2
342system call does not unmap pages, see
343.Xr munmap 2
344for further information.
345.Sh NOTES
346Although this implementation does not impose any alignment restrictions on
347the
348.Fa offset
349argument, a portable program must only use page-aligned values.
350.Pp
351Large page mappings require that the pages backing an object be
352aligned in matching blocks in both the virtual address space and RAM.
353The system will automatically attempt to use large page mappings when
354mapping an object that is already backed by large pages in RAM by
355aligning the mapping request in the virtual address space to match the
356alignment of the large physical pages.
357The system may also use large page mappings when mapping portions of an
358object that are not yet backed by pages in RAM.
359The
360.Dv MAP_ALIGNED_SUPER
361flag is an optimization that will align the mapping request to the
362size of a large page similar to
363.Dv MAP_ALIGNED ,
364except that the system will override this alignment if an object already
365uses large pages so that the mapping will be consistent with the existing
366large pages.
367This flag is mostly useful for maximizing the use of large pages on the
368first mapping of objects that do not yet have pages present in RAM.
369.Sh RETURN VALUES
370Upon successful completion,
371.Fn mmap
372returns a pointer to the mapped region.
373Otherwise, a value of
374.Dv MAP_FAILED
375is returned and
376.Va errno
377is set to indicate the error.
378.Sh ERRORS
379The
380.Fn mmap
381system call
382will fail if:
383.Bl -tag -width Er
384.It Bq Er EACCES
385The flag
386.Dv PROT_READ
387was specified as part of the
388.Fa prot
389argument and
390.Fa fd
391was not open for reading.
392The flags
393.Dv MAP_SHARED
394and
395.Dv PROT_WRITE
396were specified as part of the
397.Fa flags
398and
399.Fa prot
400argument and
401.Fa fd
402was not open for writing.
403.It Bq Er EBADF
404The
405.Fa fd
406argument
407is not a valid open file descriptor.
408.It Bq Er EINVAL
409An invalid (negative) value was passed in the
410.Fa offset
411argument, when
412.Fa fd
413referenced a regular file or shared memory.
414.It Bq Er EINVAL
415An invalid value was passed in the
416.Fa prot
417argument.
418.It Bq Er EINVAL
419An undefined option was set in the
420.Fa flags
421argument.
422.It Bq Er EINVAL
423Both
424.Dv MAP_PRIVATE
425and
426.Dv MAP_SHARED
427were specified.
428.It Bq Er EINVAL
429None of
430.Dv MAP_ANON ,
431.Dv MAP_GUARD ,
432.Dv MAP_PRIVATE ,
433.Dv MAP_SHARED ,
434or
435.Dv MAP_STACK
436was specified.
437At least one of these flags must be included.
438.It Bq Er EINVAL
439.Dv MAP_FIXED
440was specified and the
441.Fa addr
442argument was not page aligned, or part of the desired address space
443resides out of the valid address space for a user process.
444.It Bq Er EINVAL
445Both
446.Dv MAP_FIXED
447and
448.Dv MAP_32BIT
449were specified and part of the desired address space resides outside
450of the first 2GB of user address space.
451.It Bq Er EINVAL
452The
453.Fa len
454argument
455was equal to zero.
456.It Bq Er EINVAL
457.Dv MAP_ALIGNED
458was specified and the desired alignment was either larger than the
459virtual address size of the machine or smaller than a page.
460.It Bq Er EINVAL
461.Dv MAP_ANON
462was specified and the
463.Fa fd
464argument was not -1.
465.It Bq Er EINVAL
466.Dv MAP_ANON
467was specified and the
468.Fa offset
469argument was not 0.
470.It Bq Er EINVAL
471Both
472.Dv MAP_FIXED
473and
474.Dv MAP_EXCL
475were specified, but the requested region is already used by a mapping.
476.It Bq Er EINVAL
477.Dv MAP_EXCL
478was specified, but
479.Dv MAP_FIXED
480was not.
481.It Bq Er EINVAL
482.Dv MAP_GUARD
483was specified, but the
484.Fa offset
485argument was not zero, the
486.Fa fd
487argument was not -1, or the
488.Fa prot
489argument was not
490.Dv PROT_NONE .
491.It Bq Er EINVAL
492.Dv MAP_GUARD
493was specified together with one of the flags
494.Dv MAP_ANON ,
495.Dv MAP_PREFAULT ,
496.Dv MAP_PREFAULT_READ ,
497.Dv MAP_PRIVATE ,
498.Dv MAP_SHARED ,
499.Dv MAP_STACK .
500.It Bq Er ENODEV
501.Dv MAP_ANON
502has not been specified and
503.Fa fd
504did not reference a regular or character special file.
505.It Bq Er ENOMEM
506.Dv MAP_FIXED
507was specified and the
508.Fa addr
509argument was not available.
510.Dv MAP_ANON
511was specified and insufficient memory was available.
512.El
513.Sh SEE ALSO
514.Xr madvise 2 ,
515.Xr mincore 2 ,
516.Xr minherit 2 ,
517.Xr mlock 2 ,
518.Xr mprotect 2 ,
519.Xr msync 2 ,
520.Xr munlock 2 ,
521.Xr munmap 2 ,
522.Xr getpagesize 3 ,
523.Xr getpagesizes 3
524