path: root/man/man2/mremap.2

.\" Copyright (c) 1996 Tom Bjorkholm <tomb@mydata.se>
.\"
.\" SPDX-License-Identifier: GPL-2.0-or-later
.\"
.\" 1996-04-11 Tom Bjorkholm <tomb@mydata.se>
.\"            First version written (1.3.86)
.\" 1996-04-12 Tom Bjorkholm <tomb@mydata.se>
.\"            Update for Linux 1.3.87 and later
.\" 2005-10-11 mtk: Added NOTES for MREMAP_FIXED; revised EINVAL text.
.\"
.TH mremap 2 (date) "Linux man-pages (unreleased)"
.SH NAME
mremap \- remap a virtual memory address
.SH LIBRARY
Standard C library
.RI ( libc ", " \-lc )
.SH SYNOPSIS
.nf
.BR "#define _GNU_SOURCE" "         /* See feature_test_macros(7) */"
.B #include <sys/mman.h>
.P
.BI "void *mremap(void " old_address [. old_size "], size_t " old_size ,
.BI "             size_t " new_size ", int " flags ", ... /* void *" new_address " */);"
.fi
.SH DESCRIPTION
.BR mremap ()
expands (or shrinks) an existing memory mapping, potentially
moving it at the same time (controlled by the \fIflags\fP argument and
the available virtual address space).
.P
\fIold_address\fP is the old address of the virtual memory block that you
want to expand (or shrink).
Note that \fIold_address\fP has to be page
aligned.
\fIold_size\fP is the old size of the
virtual memory block.
\fInew_size\fP is the requested size of the
virtual memory block after the resize.
An optional fifth argument,
.IR new_address ,
may be provided; see the description of
.B MREMAP_FIXED
below.
.P
If the value of \fIold_size\fP is zero, and \fIold_address\fP refers to
a shareable mapping
(see the description of
.B MAP_SHARED
in
.BR mmap (2)),
then
.BR mremap ()
will create a new mapping of the same pages.
\fInew_size\fP
will be the size of the new mapping and the location of the new mapping
may be specified with \fInew_address\fP; see the description of
.B MREMAP_FIXED
below.
If a new mapping is requested via this method, then the
.B MREMAP_MAYMOVE
flag must also be specified.
.P
The \fIflags\fP bit-mask argument may be 0, or include the following flags:
.TP
.B MREMAP_MAYMOVE
By default, if there is not sufficient space to expand a mapping
at its current location, then
.BR mremap ()
fails.
If this flag is specified, then the kernel is permitted to
relocate the mapping to a new virtual address, if necessary.
If the mapping is relocated,
then absolute pointers into the old mapping location
become invalid (offsets relative to the starting address of
the mapping should be employed).
.TP
.BR MREMAP_FIXED " (since Linux 2.3.31)"
This flag serves a similar purpose to the
.B MAP_FIXED
flag of
.BR mmap (2).
If this flag is specified, then
.BR mremap ()
accepts a fifth argument,
.IR "void\ *new_address" ,
which specifies a page-aligned address to which the mapping must
be moved.
Any previous mapping at the address range specified by
.I new_address
and
.I new_size
is unmapped.
.IP
If
.B MREMAP_FIXED
is specified, then
.B MREMAP_MAYMOVE
must also be specified.
.TP
.BR MREMAP_DONTUNMAP " (since Linux 5.7)"
.\" commit e346b3813067d4b17383f975f197a9aa28a3b077
This flag, which must be used in conjunction with
.BR MREMAP_MAYMOVE ,
remaps a mapping to a new address but does not unmap the mapping at
.IR old_address .
.IP
The
.B MREMAP_DONTUNMAP
flag can be used only with private anonymous mappings
(see the description of
.B MAP_PRIVATE
and
.B MAP_ANONYMOUS
in
.BR mmap (2)).
.IP
After completion,
any access to the range specified by
.I old_address
and
.I old_size
will result in a page fault.
The page fault will be handled by a
.BR userfaultfd (2)
handler
if the address is in a range previously registered with
.BR userfaultfd (2).
Otherwise, the kernel allocates a zero-filled page to handle the fault.
.IP
The
.B MREMAP_DONTUNMAP
flag may be used to atomically move a mapping while leaving the source
mapped.
See NOTES for some possible applications of
.BR MREMAP_DONTUNMAP .
.P
If the memory segment specified by
.I old_address
and
.I old_size
is locked (using
.BR mlock (2)
or similar), then this lock is maintained when the segment is
resized and/or relocated.
As a consequence, the amount of memory locked by the process may change.
.SH RETURN VALUE
On success
.BR mremap ()
returns a pointer to the new virtual memory area.
On error, the value
.B MAP_FAILED
(that is, \fI(void\ *)\ \-1\fP) is returned,
and \fIerrno\fP is set to indicate the error.
.SH ERRORS
.TP
.B EAGAIN
The caller tried to expand a memory segment that is locked,
but this was not possible without exceeding the
.B RLIMIT_MEMLOCK
resource limit.
.TP
.B EFAULT
Some address in the range
\fIold_address\fP to \fIold_address\fP+\fIold_size\fP is an invalid
virtual memory address for this process.
You can also get
.B EFAULT
even if there exist mappings that cover the
whole address space requested, but those mappings are of different types.
.TP
.B EINVAL
An invalid argument was given.
Possible causes are:
.RS
.IP \[bu] 3
\fIold_address\fP was not
page aligned;
.IP \[bu]
a value other than
.B MREMAP_MAYMOVE
or
.B MREMAP_FIXED
or
.B MREMAP_DONTUNMAP
was specified in
.IR flags ;
.IP \[bu]
.I new_size
was zero;
.IP \[bu]
.I new_size
or
.I new_address
was invalid;
.IP \[bu]
the new address range specified by
.I new_address
and
.I new_size
overlapped the old address range specified by
.I old_address
and
.IR old_size ;
.IP \[bu]
.B MREMAP_FIXED
or
.B MREMAP_DONTUNMAP
was specified without also specifying
.BR MREMAP_MAYMOVE ;
.IP \[bu]
.B MREMAP_DONTUNMAP
was specified, but one or more pages in the range specified by
.I old_address
and
.I old_size
were not private anonymous;
.IP \[bu]
.B MREMAP_DONTUNMAP
was specified and
.I old_size
was not equal to
.IR new_size ;
.IP \[bu]
\fIold_size\fP was zero and \fIold_address\fP does not refer to a
shareable mapping (but see BUGS);
.IP \[bu]
\fIold_size\fP was zero and the
.B MREMAP_MAYMOVE
flag was not specified.
.RE
.TP
.B ENOMEM
Not enough memory was available to complete the operation.
Possible causes are:
.RS
.IP \[bu] 3
The memory area cannot be expanded at the current virtual address, and the
.B MREMAP_MAYMOVE
flag is not set in \fIflags\fP.
Or, there is not enough (virtual) memory available.
.IP \[bu]
.B MREMAP_DONTUNMAP
was used causing a new mapping to be created that would exceed the
(virtual) memory available.
Or, it would exceed the maximum number of allowed mappings.
.RE
.SH STANDARDS
Linux.
.SH HISTORY
.\" 4.2BSD had a (never actually implemented)
.\" .BR mremap (2)
.\" call with completely different semantics.
.\" .P
Prior to glibc 2.4, glibc did not expose the definition of
.BR MREMAP_FIXED ,
and the prototype for
.BR mremap ()
did not allow for the
.I new_address
argument.
.SH NOTES
.BR mremap ()
changes the
mapping between virtual addresses and memory pages.
This can be used to implement a very efficient
.BR realloc (3).
.P
In Linux, memory is divided into pages.
A process has (one or)
several linear virtual memory segments.
Each virtual memory segment has one
or more mappings to real memory pages (in the page table).
Each virtual memory segment has its own
protection (access rights), which may cause
a segmentation violation
.RB ( SIGSEGV )
if the memory is accessed incorrectly (e.g.,
writing to a read-only segment).
Accessing virtual memory outside of the
segments will also cause a segmentation violation.
.P
If
.BR mremap ()
is used to move or expand an area locked with
.BR mlock (2)
or equivalent, the
.BR mremap ()
call will make a best effort to populate the new area but will not fail
with
.B ENOMEM
if the area cannot be populated.
.\"
.SS MREMAP_DONTUNMAP use cases
Possible applications for
.B MREMAP_DONTUNMAP
include:
.IP \[bu] 3
Non-cooperative
.BR userfaultfd (2):
an application can yank out a virtual address range using
.B MREMAP_DONTUNMAP
and then employ a
.BR userfaultfd (2)
handler to handle the page faults that subsequently occur
as other threads in the process touch pages in the yanked range.
.IP \[bu]
Garbage collection:
.B MREMAP_DONTUNMAP
can be used in conjunction with
.BR userfaultfd (2)
to implement garbage collection algorithms (e.g., in a Java virtual machine).
Such an implementation can be cheaper (and simpler)
than conventional garbage collection techniques that involve
marking pages with protection
.B PROT_NONE
in conjunction with the use of a
.B SIGSEGV
handler to catch accesses to those pages.
.SH BUGS
Before Linux 4.14,
if
.I old_size
was zero and the mapping referred to by
.I old_address
was a private mapping
(see the description of
.B MAP_PRIVATE
in
.BR mmap (2)),
.BR mremap ()
created a new private mapping unrelated to the original mapping.
This behavior was unintended
and probably unexpected in user-space applications
(since the intention of
.BR mremap ()
is to create a new mapping based on the original mapping).
Since Linux 4.14,
.\" commit dba58d3b8c5045ad89c1c95d33d01451e3964db7
.BR mremap ()
fails with the error
.B EINVAL
in this scenario.
.SH SEE ALSO
.BR brk (2),
.BR getpagesize (2),
.BR getrlimit (2),
.BR mlock (2),
.BR mmap (2),
.BR sbrk (2),
.BR malloc (3),
.BR realloc (3)
.P
Your favorite text book on operating systems
for more information on paged memory
(e.g., \fIModern Operating Systems\fP by Andrew S.\& Tanenbaum,
\fIInside Linux\fP by Randolph Bentson,
\fIThe Design of the UNIX Operating System\fP by Maurice J.\& Bach)