[thirdparty/man-pages.git] / man2 / mmap.2

.\" Copyright (C) 1996 Andries Brouwer <aeb@cwi.nl>
.\" and Copyright (C) 2006, 2007 Michael Kerrisk <mtk.manpages@gmail.com>
.\"
.\" SPDX-License-Identifier: Linux-man-pages-copyleft
.\"
.\" Modified 1997-01-31 by Eric S. Raymond <esr@thyrsus.com>
.\" Modified 2000-03-25 by Jim Van Zandt <jrv@vanzandt.mv.com>
.\" Modified 2001-10-04 by John Levon <moz@compsoc.man.ac.uk>
.\" Modified 2003-02-02 by Andi Kleen <ak@muc.de>
.\" Modified 2003-05-21 by Michael Kerrisk <mtk.manpages@gmail.com>
.\"	MAP_LOCKED works from 2.5.37
.\" Modified 2004-06-17 by Michael Kerrisk <mtk.manpages@gmail.com>
.\" Modified 2004-09-11 by aeb
.\" Modified 2004-12-08, from Eric Estievenart <eric.estievenart@free.fr>
.\" Modified 2004-12-08, mtk, formatting tidy-ups
.\" Modified 2006-12-04, mtk, various parts rewritten
.\" 2007-07-10, mtk, Added an example program.
.\" 2008-11-18, mtk, document MAP_STACK
.\"
.TH MMAP 2 2021-03-22 "Linux man-pages (unreleased)"
.SH NAME
mmap, munmap \- map or unmap files or devices into memory
.SH LIBRARY
Standard C library
.RI ( libc ", " \-lc )
.SH SYNOPSIS
.nf
.B #include <sys/mman.h>
.PP
.BI "void *mmap(void *" addr ", size_t " length \
", int " prot ", int " flags ,
.BI "           int " fd ", off_t " offset );
.BI "int munmap(void *" addr ", size_t " length );
.fi
.PP
See NOTES for information on feature test macro requirements.
.SH DESCRIPTION
.BR mmap ()
creates a new mapping in the virtual address space of
the calling process.
The starting address for the new mapping is specified in
.IR addr .
The
.I length
argument specifies the length of the mapping (which must be greater than 0).
.PP
If
.I addr
is NULL,
then the kernel chooses the (page-aligned) address
at which to create the mapping;
this is the most portable method of creating a new mapping.
If
.I addr
is not NULL,
then the kernel takes it as a hint about where to place the mapping;
on Linux, the kernel will pick a nearby page boundary (but always above
or equal to the value specified by
.IR /proc/sys/vm/mmap_min_addr )
and attempt to create the mapping there.
If another mapping already exists there, the kernel picks a new address that
may or may not depend on the hint.
.\" Before Linux 2.6.24, the address was rounded up to the next page
.\" boundary; since 2.6.24, it is rounded down!
The address of the new mapping is returned as the result of the call.
.PP
The contents of a file mapping (as opposed to an anonymous mapping; see
.B MAP_ANONYMOUS
below), are initialized using
.I length
bytes starting at offset
.I offset
in the file (or other object) referred to by the file descriptor
.IR fd .
.I offset
must be a multiple of the page size as returned by
.IR sysconf(_SC_PAGE_SIZE) .
.PP
After the
.BR mmap ()
call has returned, the file descriptor,
.IR fd ,
can be closed immediately without invalidating the mapping.
.PP
The
.I prot
argument describes the desired memory protection of the mapping
(and must not conflict with the open mode of the file).
It is either
.B PROT_NONE
or the bitwise OR of one or more of the following flags:
.TP 1.1i
.B PROT_EXEC
Pages may be executed.
.TP
.B PROT_READ
Pages may be read.
.TP
.B PROT_WRITE
Pages may be written.
.TP
.B PROT_NONE
Pages may not be accessed.
.\"
.SS The flags argument
The
.I flags
argument determines whether updates to the mapping
are visible to other processes mapping the same region,
and whether updates are carried through to the underlying file.
This behavior is determined by including exactly one
of the following values in
.IR flags :
.TP
.B MAP_SHARED
Share this mapping.
Updates to the mapping are visible to other processes mapping the same region,
and (in the case of file-backed mappings)
are carried through to the underlying file.
(To precisely control when updates are carried through
to the underlying file requires the use of
.BR msync (2).)
.TP
.BR MAP_SHARED_VALIDATE " (since Linux 4.15)"
This flag provides the same behavior as
.B MAP_SHARED
except that
.B MAP_SHARED
mappings ignore unknown flags in
.IR flags .
By contrast, when creating a mapping using
.BR MAP_SHARED_VALIDATE ,
the kernel verifies all passed flags are known and fails the
mapping with the error
.B EOPNOTSUPP
for unknown flags.
This mapping type is also required to be able to use some mapping flags
(e.g.,
.BR MAP_SYNC ).
.TP
.B MAP_PRIVATE
Create a private copy-on-write mapping.
Updates to the mapping are not visible to other processes
mapping the same file, and are not carried through to
the underlying file.
It is unspecified whether changes made to the file after the
.BR mmap ()
call are visible in the mapped region.
.PP
Both
.B MAP_SHARED
and
.B MAP_PRIVATE
are described in POSIX.1-2001 and POSIX.1-2008.
.B MAP_SHARED_VALIDATE
is a Linux extension.
.PP
In addition, zero or more of the following values can be ORed in
.IR flags :
.TP
.BR MAP_32BIT " (since Linux 2.4.20, 2.6)"
Put the mapping into the first 2 Gigabytes of the process address space.
This flag is supported only on x86-64, for 64-bit programs.
It was added to allow thread stacks to be allocated somewhere
in the first 2\ GB of memory,
so as to improve context-switch performance on some early
64-bit processors.
.\" See http://lwn.net/Articles/294642 "Tangled up in threads", 19 Aug 08
Modern x86-64 processors no longer have this performance problem,
so use of this flag is not required on those systems.
The
.B MAP_32BIT
flag is ignored when
.B MAP_FIXED
is set.
.TP
.B MAP_ANON
Synonym for
.BR MAP_ANONYMOUS ;
provided for compatibility with other implementations.
.TP
.B MAP_ANONYMOUS
The mapping is not backed by any file;
its contents are initialized to zero.
The
.I fd
argument is ignored;
however, some implementations require
.I fd
to be \-1 if
.B MAP_ANONYMOUS
(or
.BR MAP_ANON )
is specified,
and portable applications should ensure this.
The
.I offset
argument should be zero.
.\" See the pgoff overflow check in do_mmap().
.\" See the offset check in sys_mmap in arch/x86/kernel/sys_x86_64.c.
Support for
.B MAP_ANONYMOUS
in conjunction with
.B MAP_SHARED
was added in Linux 2.4.
.TP
.B MAP_DENYWRITE
This flag is ignored.
.\" Introduced in 1.1.36, removed in 1.3.24.
(Long ago\(emLinux 2.0 and earlier\(emit signaled
that attempts to write to the underlying file should fail with
.BR ETXTBSY .
But this was a source of denial-of-service attacks.)
.TP
.B MAP_EXECUTABLE
This flag is ignored.
.\" Introduced in 1.1.38, removed in 1.3.24. Flag tested in proc_follow_link.
.\" (Long ago, it signaled that the underlying file is an executable.
.\" However, that information was not really used anywhere.)
.\" Linus talked about DOS related to MAP_EXECUTABLE, but he was thinking of
.\" MAP_DENYWRITE?
.TP
.B MAP_FILE
Compatibility flag.
Ignored.
.\" On some systems, this was required as the opposite of
.\" MAP_ANONYMOUS -- mtk, 1 May 2007
.TP
.B MAP_FIXED
Don't interpret
.I addr
as a hint: place the mapping at exactly that address.
.I addr
must be suitably aligned: for most architectures a multiple of the page
size is sufficient; however, some architectures may impose additional
restrictions.
If the memory region specified by
.I addr
and
.I length
overlaps pages of any existing mapping(s), then the overlapped
part of the existing mapping(s) will be discarded.
If the specified address cannot be used,
.BR mmap ()
will fail.
.IP
Software that aspires to be portable should use the
.B MAP_FIXED
flag with care,
keeping in mind that the exact layout of a process's memory mappings
is allowed to change significantly between kernel versions,
C library versions, and operating system releases.
.I Carefully read the discussion of this flag in NOTES!
.TP
.BR MAP_FIXED_NOREPLACE " (since Linux 4.17)"
.\" commit a4ff8e8620d3f4f50ac4b41e8067b7d395056843
This flag provides behavior that is similar to
.B MAP_FIXED
with respect to the
.I addr
enforcement, but differs in that
.B MAP_FIXED_NOREPLACE
never clobbers a preexisting mapped range.
If the requested range would collide with an existing mapping,
then this call fails with the error
.B EEXIST.
This flag can therefore be used as a way to atomically
(with respect to other threads) attempt to map an address range:
one thread will succeed; all others will report failure.
.IP
Note that older kernels which do not recognize the
.B MAP_FIXED_NOREPLACE
flag will typically (upon detecting a collision with a preexisting mapping)
fall back to a
.RB \(lqnon- MAP_FIXED \(rq
type of behavior:
they will return an address that is different from the requested address.
Therefore, backward-compatible software
should check the returned address against the requested address.
.TP
.B MAP_GROWSDOWN
This flag is used for stacks.
It indicates to the kernel virtual memory system that the mapping
should extend downward in memory.
The return address is one page lower than the memory area that is
actually created in the process's virtual address space.
Touching an address in the "guard" page below the mapping will cause
the mapping to grow by a page.
This growth can be repeated until the mapping grows to within a
page of the high end of the next lower mapping,
at which point touching the "guard" page will result in a
.B SIGSEGV
signal.
.TP
.BR MAP_HUGETLB " (since Linux 2.6.32)"
Allocate the mapping using "huge" pages.
See the Linux kernel source file
.I Documentation/admin\-guide/mm/hugetlbpage.rst
for further information, as well as NOTES, below.
.TP
.BR MAP_HUGE_2MB ", " MAP_HUGE_1GB " (since Linux 3.8)"
.\" See https://lwn.net/Articles/533499/
Used in conjunction with
.B MAP_HUGETLB
to select alternative hugetlb page sizes (respectively, 2\ MB and 1\ GB)
on systems that support multiple hugetlb page sizes.
.IP
More generally, the desired huge page size can be configured by encoding
the base-2 logarithm of the desired page size in the six bits at the offset
.BR MAP_HUGE_SHIFT .
(A value of zero in this bit field provides the default huge page size;
the default huge page size can be discovered via the
.I Hugepagesize
field exposed by
.IR /proc/meminfo .)
Thus, the above two constants are defined as:
.IP
.in +4n
.EX
#define MAP_HUGE_2MB    (21 << MAP_HUGE_SHIFT)
#define MAP_HUGE_1GB    (30 << MAP_HUGE_SHIFT)
.EE
.in
.IP
The range of huge page sizes that are supported by the system
can be discovered by listing the subdirectories in
.IR /sys/kernel/mm/hugepages .
.TP
.BR MAP_LOCKED " (since Linux 2.5.37)"
Mark the mapped region to be locked in the same way as
.BR mlock (2).
This implementation will try to populate (prefault) the whole range but the
.BR mmap ()
call doesn't fail with
.B ENOMEM
if this fails.
Therefore major faults might happen later on.
So the semantic is not as strong as
.BR mlock (2).
One should use
.BR mmap ()
plus
.BR mlock (2)
when major faults are not acceptable after the initialization of the mapping.
The
.B MAP_LOCKED
flag is ignored in older kernels.
.\" If set, the mapped pages will not be swapped out.
.TP
.BR MAP_NONBLOCK " (since Linux 2.5.46)"
This flag is meaningful only in conjunction with
.BR MAP_POPULATE .
Don't perform read-ahead:
create page tables entries only for pages
that are already present in RAM.
Since Linux 2.6.23,
.\" commit 54cb8821de07f2ffcd28c380ce9b93d5784b40d7
this flag causes
.B MAP_POPULATE
to do nothing.
One day, the combination of
.B MAP_POPULATE
and
.B MAP_NONBLOCK
may be reimplemented.
.TP
.B MAP_NORESERVE
Do not reserve swap space for this mapping.
When swap space is reserved, one has the guarantee
that it is possible to modify the mapping.
When swap space is not reserved one might get
.B SIGSEGV
upon a write
if no physical memory is available.
See also the discussion of the file
.I /proc/sys/vm/overcommit_memory
in
.BR proc (5).
In kernels before 2.6, this flag had effect only for
private writable mappings.
.TP
.BR MAP_POPULATE " (since Linux 2.5.46)"
Populate (prefault) page tables for a mapping.
For a file mapping, this causes read-ahead on the file.
This will help to reduce blocking on page faults later.
The
.BR mmap ()
call doesn't fail if the mapping cannot be populated (for example, due
to limitations on the number of mapped huge pages when using
.BR MAP_HUGETLB ).
Support for
.B MAP_POPULATE
in conjunction with private mappings was added in Linux 2.6.23.
.TP
.BR MAP_STACK " (since Linux 2.6.27)"
Allocate the mapping at an address suitable for a process
or thread stack.
.IP
This flag is currently a no-op on Linux.
However, by employing this flag, applications can ensure that
they transparently obtain support if the flag
is implemented in the future.
Thus, it is used in the glibc threading implementation to allow for
the fact that some architectures may (later) require special treatment
for stack allocations.
.\" See http://lwn.net/Articles/294642 "Tangled up in threads", 19 Aug 08
.\" commit cd98a04a59e2f94fa64d5bf1e26498d27427d5e7
.\" http://thread.gmane.org/gmane.linux.kernel/720412
.\" "pthread_create() slow for many threads; also time to revisit 64b
.\"  context switch optimization?"
A further reason to employ this flag is portability:
.B MAP_STACK
exists (and has an effect) on some other systems (e.g., some of the BSDs).
.TP
.BR MAP_SYNC " (since Linux 4.15)"
This flag is available only with the
.B MAP_SHARED_VALIDATE
mapping type;
mappings of type
.B MAP_SHARED
will silently ignore this flag.
This flag is supported only for files supporting DAX
(direct mapping of persistent memory).
For other files, creating a mapping with this flag results in an
.B EOPNOTSUPP
error.
.IP
Shared file mappings with this flag provide the guarantee that while
some memory is mapped writable in the address space of the process,
it will be visible in the same file at the same offset even after
the system crashes or is rebooted.
In conjunction with the use of appropriate CPU instructions,
this provides users of such mappings with a more efficient way
of making data modifications persistent.
.TP
.BR MAP_UNINITIALIZED " (since Linux 2.6.33)"
Don't clear anonymous pages.
This flag is intended to improve performance on embedded devices.
This flag is honored only if the kernel was configured with the
.B CONFIG_MMAP_ALLOW_UNINITIALIZED
option.
Because of the security implications,
that option is normally enabled only on embedded devices
(i.e., devices where one has complete control of the contents of user memory).
.PP
Of the above flags, only
.B MAP_FIXED
is specified in POSIX.1-2001 and POSIX.1-2008.
However, most systems also support
.B MAP_ANONYMOUS
(or its synonym
.BR MAP_ANON ).
.\" FIXME . for later review when Issue 8 is one day released...
.\" POSIX may add MAP_ANON in the future
.\" http://austingroupbugs.net/tag_view_page.php?tag_id=8
.\" http://austingroupbugs.net/view.php?id=850
.SS munmap()
The
.BR munmap ()
system call deletes the mappings for the specified address range, and
causes further references to addresses within the range to generate
invalid memory references.
The region is also automatically unmapped
when the process is terminated.
On the other hand, closing the file
descriptor does not unmap the region.
.PP
The address
.I addr
must be a multiple of the page size (but
.I length
need not be).
All pages containing a part
of the indicated range are unmapped, and subsequent references
to these pages will generate
.BR SIGSEGV .
It is not an error if the
indicated range does not contain any mapped pages.
.SH RETURN VALUE
On success,
.BR mmap ()
returns a pointer to the mapped area.
On error, the value
.B MAP_FAILED
(that is,
.IR "(void\ *)\ \-1" )
is returned, and
.I errno
is set to indicate the error.
.PP
On success,
.BR munmap ()
returns 0.
On failure, it returns \-1, and
.I errno
is set to indicate the error (probably to
.BR EINVAL ).
.SH ERRORS
.TP
.B EACCES
A file descriptor refers to a non-regular file.
Or a file mapping was requested, but
.I fd
is not open for reading.
Or
.B MAP_SHARED
was requested and
.B PROT_WRITE
is set, but
.I fd
is not open in read/write
.RB ( O_RDWR )
mode.
Or
.B PROT_WRITE
is set, but the file is append-only.
.TP
.B EAGAIN
The file has been locked, or too much memory has been locked (see
.BR setrlimit (2)).
.TP
.B EBADF
.I fd
is not a valid file descriptor (and
.B MAP_ANONYMOUS
was not set).
.TP
.B EEXIST
.B MAP_FIXED_NOREPLACE
was specified in
.IR flags ,
and the range covered by
.I addr
and
.I length
clashes with an existing mapping.
.TP
.B EINVAL
We don't like
.IR addr ,
.IR length ,
or
.I offset
(e.g., they are too large, or not aligned on a page boundary).
.TP
.B EINVAL
(since Linux 2.6.12)
.I length
was 0.
.TP
.B EINVAL
.I flags
contained none of
.BR MAP_PRIVATE ,
.BR MAP_SHARED ,
or
.BR MAP_SHARED_VALIDATE .
.TP
.B ENFILE
.\" This is for shared anonymous segments
.\" [2.6.7] shmem_zero_setup()-->shmem_file_setup()-->get_empty_filp()
The system-wide limit on the total number of open files has been reached.
.\" .TP
.\" .B ENOEXEC
.\" A file could not be mapped for reading.
.TP
.B ENODEV
The underlying filesystem of the specified file does not support
memory mapping.
.TP
.B ENOMEM
No memory is available.
.TP
.B ENOMEM
The process's maximum number of mappings would have been exceeded.
This error can also occur for
.BR munmap (),
when unmapping a region in the middle of an existing mapping,
since this results in two smaller mappings on either side of
the region being unmapped.
.TP
.B ENOMEM
(since Linux 4.7)
The process's
.B RLIMIT_DATA
limit, described in
.BR getrlimit (2),
would have been exceeded.
.TP
.B ENOMEM
We don't like
.IR addr ,
because it exceeds the virtual address space of the CPU.
.TP
.B EOVERFLOW
On 32-bit architecture together with the large file extension
(i.e., using 64-bit
.IR off_t ):
the number of pages used for
.I length
plus number of pages used for
.I offset
would overflow
.I "unsigned long"
(32 bits).
.TP
.B EPERM
The
.I prot
argument asks for
.B PROT_EXEC
but the mapped area belongs to a file on a filesystem that
was mounted no-exec.
.\" (Since 2.4.25 / 2.6.0.)
.TP
.B EPERM
The operation was prevented by a file seal; see
.BR fcntl (2).
.TP
.B EPERM
The
.B MAP_HUGETLB
flag was specified, but the caller was not privileged (did not have the
.B CAP_IPC_LOCK
capability)
and is not a member of the
.I sysctl_hugetlb_shm_group
group; see the description of
.I /proc/sys/vm/sysctl_hugetlb_shm_group
in
.TP
.B ETXTBSY
.B MAP_DENYWRITE
was set but the object specified by
.I fd
is open for writing.
.PP
Use of a mapped region can result in these signals:
.TP
.B SIGSEGV
Attempted write into a region mapped as read-only.
.TP
.B SIGBUS
Attempted access to a page of the buffer that lies beyond the
end of the mapped file.
For an explanation of the treatment of the bytes in the page that
corresponds to the end of a mapped file that is not a multiple
of the page size, see NOTES.
.SH ATTRIBUTES
For an explanation of the terms used in this section, see
.BR attributes (7).
.ad l
.nh
.TS
allbox;
lbx lb lb
l l l.
Interface	Attribute	Value
T{
.BR mmap (),
.BR munmap ()
T}	Thread safety	MT-Safe
.TE
.hy
.ad
.sp 1
.SH STANDARDS
POSIX.1-2001, POSIX.1-2008, SVr4, 4.4BSD.
.\" SVr4 documents additional error codes ENXIO and ENODEV.
.\" SUSv2 documents additional error codes EMFILE and EOVERFLOW.
.PP
On POSIX systems on which
.BR mmap (),
.BR msync (2),
and
.BR munmap ()
are available,
.B _POSIX_MAPPED_FILES
is defined in \fI<unistd.h>\fP to a value greater than 0.
(See also
.BR sysconf (3).)
.\" POSIX.1-2001: It shall be defined to -1 or 0 or 200112L.
.\" -1: unavailable, 0: ask using sysconf().
.\" glibc defines it to 1.
.SH NOTES
Memory mapped by
.BR mmap ()
is preserved across
.BR fork (2),
with the same attributes.
.PP
A file is mapped in multiples of the page size.
For a file that is not
a multiple of the page size,
the remaining bytes in the partial page at the end of the mapping
are zeroed when mapped,
and modifications to that region are not written out to the file.
The effect of
changing the size of the underlying file of a mapping on the pages that
correspond to added or removed regions of the file is unspecified.
.PP
On some hardware architectures (e.g., i386),
.B PROT_WRITE
implies
.BR PROT_READ .
It is architecture dependent whether
.B PROT_READ
implies
.B PROT_EXEC
or not.
Portable programs should always set
.B PROT_EXEC
if they intend to execute code in the new mapping.
.PP
The portable way to create a mapping is to specify
.I addr
as 0 (NULL), and omit
.B MAP_FIXED
from
.IR flags .
In this case, the system chooses the address for the mapping;
the address is chosen so as not to conflict with any existing mapping,
and will not be 0.
If the
.B MAP_FIXED
flag is specified, and
.I addr
is 0 (NULL), then the mapped address will be 0 (NULL).
.PP
Certain
.I flags
constants are defined only if suitable feature test macros are defined
(possibly by default):
.B _DEFAULT_SOURCE
with glibc 2.19 or later;
or
.B _BSD_SOURCE
or
.B _SVID_SOURCE
in glibc 2.19 and earlier.
(Employing
.B _GNU_SOURCE
also suffices,
and requiring that macro specifically would have been more logical,
since these flags are all Linux-specific.)
The relevant flags are:
.BR MAP_32BIT ,
.B MAP_ANONYMOUS
(and the synonym
.BR MAP_ANON ),
.BR MAP_DENYWRITE ,
.BR MAP_EXECUTABLE ,
.BR MAP_FILE ,
.BR MAP_GROWSDOWN ,
.BR MAP_HUGETLB ,
.BR MAP_LOCKED ,
.BR MAP_NONBLOCK ,
.BR MAP_NORESERVE ,
.BR MAP_POPULATE ,
and
.BR MAP_STACK .
.PP
An application can determine which pages of a mapping are
currently resident in the buffer/page cache using
.BR mincore (2).
.\"
.SS Using MAP_FIXED safely
The only safe use for
.B MAP_FIXED
is where the address range specified by
.I addr
and
.I length
was previously reserved using another mapping;
otherwise, the use of
.B MAP_FIXED
is hazardous because it forcibly removes preexisting mappings,
making it easy for a multithreaded process to corrupt its own address space.
.PP
For example, suppose that thread A looks through
.I /proc/<pid>/maps
in order to locate an unused address range that it can map using
.BR MAP_FIXED ,
while thread B simultaneously acquires part or all of that same
address range.
When thread A subsequently employs
.BR mmap(MAP_FIXED) ,
it will effectively clobber the mapping that thread B created.
In this scenario,
thread B need not create a mapping directly; simply making a library call
that, internally, uses
.BR dlopen (3)
to load some other shared library, will suffice.
The
.BR dlopen (3)
call will map the library into the process's address space.
Furthermore, almost any library call may be implemented in a way that
adds memory mappings to the address space, either with this technique,
or by simply allocating memory.
Examples include
.BR brk (2),
.BR malloc (3),
.BR pthread_create (3),
and the PAM libraries
.UR http://www.linux-pam.org
.UE .
.PP
Since Linux 4.17, a multithreaded program can use the
.B MAP_FIXED_NOREPLACE
flag to avoid the hazard described above
when attempting to create a mapping at a fixed address
that has not been reserved by a preexisting mapping.
.\"
.SS Timestamps changes for file-backed mappings
For file-backed mappings, the
.I st_atime
field for the mapped file may be updated at any time between the
.BR mmap ()
and the corresponding unmapping; the first reference to a mapped
page will update the field if it has not been already.
.PP
The
.I st_ctime
and
.I st_mtime
field for a file mapped with
.B PROT_WRITE
and
.B MAP_SHARED
will be updated after
a write to the mapped region, and before a subsequent
.BR msync (2)
with the
.B MS_SYNC
or
.B MS_ASYNC
flag, if one occurs.
.\"
.SS Huge page (Huge TLB) mappings
For mappings that employ huge pages, the requirements for the arguments of
.BR mmap ()
and
.BR munmap ()
differ somewhat from the requirements for mappings
that use the native system page size.
.PP
For
.BR mmap (),
.I offset
must be a multiple of the underlying huge page size.
The system automatically aligns
.I length
to be a multiple of the underlying huge page size.
.PP
For
.BR munmap (),
.IR addr ,
and
.I length
must both be a multiple of the underlying huge page size.
.\"
.SS C library/kernel differences
This page describes the interface provided by the glibc
.BR mmap ()
wrapper function.
Originally, this function invoked a system call of the same name.
Since kernel 2.4, that system call has been superseded by
.BR mmap2 (2),
and nowadays
.\" Since around glibc 2.1/2.2, depending on the platform.
the glibc
.BR mmap ()
wrapper function invokes
.BR mmap2 (2)
with a suitably adjusted value for
.IR offset .
.SH BUGS
On Linux, there are no guarantees like those suggested above under
.BR MAP_NORESERVE .
By default, any process can be killed
at any moment when the system runs out of memory.
.PP
In kernels before 2.6.7, the
.B MAP_POPULATE
flag has effect only if
.I prot
is specified as
.BR PROT_NONE .
.PP
SUSv3 specifies that
.BR mmap ()
should fail if
.I length
is 0.
However, in kernels before 2.6.12,
.BR mmap ()
succeeded in this case: no mapping was created and the call returned
.IR addr .
Since kernel 2.6.12,
.BR mmap ()
fails with the error
.B EINVAL
for this case.
.PP
POSIX specifies that the system shall always
zero fill any partial page at the end
of the object and that system will never write any modification of the
object beyond its end.
On Linux, when you write data to such partial page after the end
of the object, the data stays in the page cache even after the file
is closed and unmapped
and even though the data is never written to the file itself,
subsequent mappings may see the modified content.
In some cases, this could be fixed by calling
.BR msync (2)
before the unmap takes place;
however, this doesn't work on
.BR tmpfs (5)
(for example, when using the POSIX shared memory interface documented in
.BR shm_overview (7)).
.SH EXAMPLES
.\" FIXME . Add an example here that uses an anonymous shared region for
.\" IPC between parent and child.
The following program prints part of the file specified in
its first command-line argument to standard output.
The range of bytes to be printed is specified via offset and length
values in the second and third command-line arguments.
The program creates a memory mapping of the required
pages of the file and then uses
.BR write (2)
to output the desired bytes.
.SS Program source
.\" SRC BEGIN (mmap.c)
.EX
#include <fcntl.h>
#include <stdio.h>
#include <stdlib.h>
#include <sys/mman.h>
#include <sys/stat.h>
#include <unistd.h>

#define handle_error(msg) \e
    do { perror(msg); exit(EXIT_FAILURE); } while (0)

int
main(int argc, char *argv[])
{
    char *addr;
    int fd;
    struct stat sb;
    off_t offset, pa_offset;
    size_t length;
    ssize_t s;

    if (argc < 3 || argc > 4) {
        fprintf(stderr, "%s file offset [length]\en", argv[0]);
        exit(EXIT_FAILURE);
    }

    fd = open(argv[1], O_RDONLY);
    if (fd == \-1)
        handle_error("open");

    if (fstat(fd, &sb) == \-1)           /* To obtain file size */
        handle_error("fstat");

    offset = atoi(argv[2]);
    pa_offset = offset & \(ti(sysconf(_SC_PAGE_SIZE) \- 1);
        /* offset for mmap() must be page aligned */

    if (offset >= sb.st_size) {
        fprintf(stderr, "offset is past end of file\en");
        exit(EXIT_FAILURE);
    }

    if (argc == 4) {
        length = atoi(argv[3]);
        if (offset + length > sb.st_size)
            length = sb.st_size \- offset;
                /* Can\(aqt display bytes past end of file */

    } else {    /* No length arg ==> display to end of file */
        length = sb.st_size \- offset;
    }

    addr = mmap(NULL, length + offset \- pa_offset, PROT_READ,
                MAP_PRIVATE, fd, pa_offset);
    if (addr == MAP_FAILED)
        handle_error("mmap");

    s = write(STDOUT_FILENO, addr + offset \- pa_offset, length);
    if (s != length) {
        if (s == \-1)
            handle_error("write");

        fprintf(stderr, "partial write");
        exit(EXIT_FAILURE);
    }

    munmap(addr, length + offset \- pa_offset);
    close(fd);

    exit(EXIT_SUCCESS);
}
.EE
.\" SRC END
.SH SEE ALSO
.BR ftruncate (2),
.BR getpagesize (2),
.BR memfd_create (2),
.BR mincore (2),
.BR mlock (2),
.BR mmap2 (2),
.BR mprotect (2),
.BR mremap (2),
.BR msync (2),
.BR remap_file_pages (2),
.BR setrlimit (2),
.BR shmat (2),
.BR userfaultfd (2),
.BR shm_open (3),
.BR shm_overview (7)
.PP
The descriptions of the following files in
.BR proc (5):
.IR /proc/[pid]/maps ,
.IR /proc/[pid]/map_files ,
and
.IR /proc/[pid]/smaps .
.PP
B.O. Gallmeister, POSIX.4, O'Reilly, pp. 128\(en129 and 389\(en391.
.\"
.\" Repeat after me: private read-only mappings are 100% equivalent to
.\" shared read-only mappings. No ifs, buts, or maybes. -- Linus