freebsd-dev/lib/libc/sys/mmap.2
Mark Johnston 64910ddbff Launder VPO_NOSYNC pages upon vnode deactivation.
As of r234483, vnode deactivation causes non-VPO_NOSYNC pages to be
laundered. This behaviour has two problems:

1. Dirty VPO_NOSYNC pages must be laundered before the vnode can be
   reclaimed, and this work may be unfairly deferred to the vnlru process
   or an unrelated application when the system is under vnode pressure.
2. Deactivation of a vnode with dirty VPO_NOSYNC pages requires a scan of
   the corresponding VM object's memq for non-VPO_NOSYNC dirty pages; if
   the laundry thread needs to launder pages from an unreferenced such
   vnode, it will reactivate and deactivate the vnode with each laundering,
   potentially resulting in a large number of expensive scans.

Therefore, ensure that all dirty pages are laundered upon deactivation,
i.e., when all maps of the vnode are removed and all references are
released.

Reviewed by:	alc, kib
MFC after:	1 month
Differential Revision:	https://reviews.freebsd.org/D8641
2016-11-26 21:00:27 +00:00

453 lines
12 KiB
Groff

.\" Copyright (c) 1991, 1993
.\" The Regents of the University of California. All rights reserved.
.\"
.\" Redistribution and use in source and binary forms, with or without
.\" modification, are permitted provided that the following conditions
.\" are met:
.\" 1. Redistributions of source code must retain the above copyright
.\" notice, this list of conditions and the following disclaimer.
.\" 2. Redistributions in binary form must reproduce the above copyright
.\" notice, this list of conditions and the following disclaimer in the
.\" documentation and/or other materials provided with the distribution.
.\" 4. Neither the name of the University nor the names of its contributors
.\" may be used to endorse or promote products derived from this software
.\" without specific prior written permission.
.\"
.\" THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
.\" ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
.\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
.\" SUCH DAMAGE.
.\"
.\" @(#)mmap.2 8.4 (Berkeley) 5/11/95
.\" $FreeBSD$
.\"
.Dd November 25, 2016
.Dt MMAP 2
.Os
.Sh NAME
.Nm mmap
.Nd allocate memory, or map files or devices into memory
.Sh LIBRARY
.Lb libc
.Sh SYNOPSIS
.In sys/mman.h
.Ft void *
.Fn mmap "void *addr" "size_t len" "int prot" "int flags" "int fd" "off_t offset"
.Sh DESCRIPTION
The
.Fn mmap
system call causes the pages starting at
.Fa addr
and continuing for at most
.Fa len
bytes to be mapped from the object described by
.Fa fd ,
starting at byte offset
.Fa offset .
If
.Fa len
is not a multiple of the pagesize, the mapped region may extend past the
specified range.
Any such extension beyond the end of the mapped object will be zero-filled.
.Pp
If
.Fa addr
is non-zero, it is used as a hint to the system.
(As a convenience to the system, the actual address of the region may differ
from the address supplied.)
If
.Fa addr
is zero, an address will be selected by the system.
The actual starting address of the region is returned.
A successful
.Fa mmap
deletes any previous mapping in the allocated address range.
.Pp
The protections (region accessibility) are specified in the
.Fa prot
argument by
.Em or Ns 'ing
the following values:
.Pp
.Bl -tag -width PROT_WRITE -compact
.It Dv PROT_NONE
Pages may not be accessed.
.It Dv PROT_READ
Pages may be read.
.It Dv PROT_WRITE
Pages may be written.
.It Dv PROT_EXEC
Pages may be executed.
.El
.Pp
The
.Fa flags
argument specifies the type of the mapped object, mapping options and
whether modifications made to the mapped copy of the page are private
to the process or are to be shared with other references.
Sharing, mapping type and options are specified in the
.Fa flags
argument by
.Em or Ns 'ing
the following values:
.Bl -tag -width MAP_PREFAULT_READ
.It Dv MAP_32BIT
Request a region in the first 2GB of the current process's address space.
If a suitable region cannot be found,
.Fn mmap
will fail.
This flag is only available on 64-bit platforms.
.It Dv MAP_ALIGNED Ns Pq Fa n
Align the region on a requested boundary.
If a suitable region cannot be found,
.Fn mmap
will fail.
The
.Fa n
argument specifies the binary logarithm of the desired alignment.
.It Dv MAP_ALIGNED_SUPER
Align the region to maximize the potential use of large
.Pq Dq super
pages.
If a suitable region cannot be found,
.Fn mmap
will fail.
The system will choose a suitable page size based on the size of
mapping.
The page size used as well as the alignment of the region may both be
affected by properties of the file being mapped.
In particular,
the physical address of existing pages of a file may require a specific
alignment.
The region is not guaranteed to be aligned on any specific boundary.
.It Dv MAP_ANON
Map anonymous memory not associated with any specific file.
The file descriptor used for creating
.Dv MAP_ANON
must be \-1.
The
.Fa offset
argument must be 0.
.\".It Dv MAP_FILE
.\"Mapped from a regular file or character-special device memory.
.It Dv MAP_ANONYMOUS
This flag is identical to
.Dv MAP_ANON
and is provided for compatibility.
.It Dv MAP_EXCL
This flag can only be used in combination with
.Dv MAP_FIXED .
Please see the definition of
.Dv MAP_FIXED
for the description of its effect.
.It Dv MAP_FIXED
Do not permit the system to select a different address than the one
specified.
If the specified address cannot be used,
.Fn mmap
will fail.
If
.Dv MAP_FIXED
is specified,
.Fa addr
must be a multiple of the pagesize.
If
.Dv MAP_EXCL
is not specified, a successful
.Dv MAP_FIXED
request replaces any previous mappings for the process'
pages in the range from
.Fa addr
to
.Fa addr
+
.Fa len .
In contrast, if
.Dv MAP_EXCL
is specified, the request will fail if a mapping
already exists within the range.
.It Dv MAP_HASSEMAPHORE
Notify the kernel that the region may contain semaphores and that special
handling may be necessary.
.It Dv MAP_NOCORE
Region is not included in a core file.
.It Dv MAP_NOSYNC
Causes data dirtied via this VM map to be flushed to physical media
only when necessary (usually by the pager) rather than gratuitously.
Typically this prevents the update daemons from flushing pages dirtied
through such maps and thus allows efficient sharing of memory across
unassociated processes using a file-backed shared memory map.
Without
this option any VM pages you dirty may be flushed to disk every so often
(every 30-60 seconds usually) which can create performance problems if you
do not need that to occur (such as when you are using shared file-backed
mmap regions for IPC purposes).
Dirty data will be flushed automatically when all mappings of an object are
removed and all descriptors referencing the object are closed.
Note that VM/file system coherency is
maintained whether you use
.Dv MAP_NOSYNC
or not.
This option is not portable
across
.Ux
platforms (yet), though some may implement the same behavior
by default.
.Pp
.Em WARNING !
Extending a file with
.Xr ftruncate 2 ,
thus creating a big hole, and then filling the hole by modifying a shared
.Fn mmap
can lead to severe file fragmentation.
In order to avoid such fragmentation you should always pre-allocate the
file's backing store by
.Fn write Ns ing
zero's into the newly extended area prior to modifying the area via your
.Fn mmap .
The fragmentation problem is especially sensitive to
.Dv MAP_NOSYNC
pages, because pages may be flushed to disk in a totally random order.
.Pp
The same applies when using
.Dv MAP_NOSYNC
to implement a file-based shared memory store.
It is recommended that you create the backing store by
.Fn write Ns ing
zero's to the backing file rather than
.Fn ftruncate Ns ing
it.
You can test file fragmentation by observing the KB/t (kilobytes per
transfer) results from an
.Dq Li iostat 1
while reading a large file sequentially, e.g.,\& using
.Dq Li dd if=filename of=/dev/null bs=32k .
.Pp
The
.Xr fsync 2
system call will flush all dirty data and metadata associated with a file,
including dirty NOSYNC VM data, to physical media.
The
.Xr sync 8
command and
.Xr sync 2
system call generally do not flush dirty NOSYNC VM data.
The
.Xr msync 2
system call is usually not needed since
.Bx
implements a coherent file system buffer cache.
However, it may be
used to associate dirty VM pages with file system buffers and thus cause
them to be flushed to physical media sooner rather than later.
.It Dv MAP_PREFAULT_READ
Immediately update the calling process's lowest-level virtual address
translation structures, such as its page table, so that every memory
resident page within the region is mapped for read access.
Ordinarily these structures are updated lazily.
The effect of this option is to eliminate any soft faults that would
otherwise occur on the initial read accesses to the region.
Although this option does not preclude
.Fa prot
from including
.Dv PROT_WRITE ,
it does not eliminate soft faults on the initial write accesses to the
region.
.It Dv MAP_PRIVATE
Modifications are private.
.It Dv MAP_SHARED
Modifications are shared.
.It Dv MAP_STACK
.Dv MAP_STACK
implies
.Dv MAP_ANON ,
and
.Fa offset
of 0.
The
.Fa fd
argument
must be -1 and
.Fa prot
must include at least
.Dv PROT_READ
and
.Dv PROT_WRITE .
This option creates
a memory region that grows to at most
.Fa len
bytes in size, starting from the stack top and growing down.
The
stack top is the starting address returned by the call, plus
.Fa len
bytes.
The bottom of the stack at maximum growth is the starting
address returned by the call.
.El
.Pp
The
.Xr close 2
system call does not unmap pages, see
.Xr munmap 2
for further information.
.Sh NOTES
Although this implementation does not impose any alignment restrictions on
the
.Fa offset
argument, a portable program must only use page-aligned values.
.Pp
Large page mappings require that the pages backing an object be
aligned in matching blocks in both the virtual address space and RAM.
The system will automatically attempt to use large page mappings when
mapping an object that is already backed by large pages in RAM by
aligning the mapping request in the virtual address space to match the
alignment of the large physical pages.
The system may also use large page mappings when mapping portions of an
object that are not yet backed by pages in RAM.
The
.Dv MAP_ALIGNED_SUPER
flag is an optimization that will align the mapping request to the
size of a large page similar to
.Dv MAP_ALIGNED ,
except that the system will override this alignment if an object already
uses large pages so that the mapping will be consistent with the existing
large pages.
This flag is mostly useful for maximizing the use of large pages on the
first mapping of objects that do not yet have pages present in RAM.
.Sh RETURN VALUES
Upon successful completion,
.Fn mmap
returns a pointer to the mapped region.
Otherwise, a value of
.Dv MAP_FAILED
is returned and
.Va errno
is set to indicate the error.
.Sh ERRORS
The
.Fn mmap
system call
will fail if:
.Bl -tag -width Er
.It Bq Er EACCES
The flag
.Dv PROT_READ
was specified as part of the
.Fa prot
argument and
.Fa fd
was not open for reading.
The flags
.Dv MAP_SHARED
and
.Dv PROT_WRITE
were specified as part of the
.Fa flags
and
.Fa prot
argument and
.Fa fd
was not open for writing.
.It Bq Er EBADF
The
.Fa fd
argument
is not a valid open file descriptor.
.It Bq Er EINVAL
An invalid value was passed in the
.Fa prot
argument.
.It Bq Er EINVAL
An undefined option was set in the
.Fa flags
argument.
.It Bq Er EINVAL
Both
.Dv MAP_PRIVATE
and
.Dv MAP_SHARED
were specified.
.It Bq Er EINVAL
None of
.Dv MAP_ANON ,
.Dv MAP_PRIVATE ,
.Dv MAP_SHARED ,
or
.Dv MAP_STACK
was specified.
At least one of these flags must be included.
.It Bq Er EINVAL
.Dv MAP_FIXED
was specified and the
.Fa addr
argument was not page aligned, or part of the desired address space
resides out of the valid address space for a user process.
.It Bq Er EINVAL
Both
.Dv MAP_FIXED
and
.Dv MAP_32BIT
were specified and part of the desired address space resides outside
of the first 2GB of user address space.
.It Bq Er EINVAL
The
.Fa len
argument
was equal to zero.
.It Bq Er EINVAL
.Dv MAP_ALIGNED
was specified and the desired alignment was either larger than the
virtual address size of the machine or smaller than a page.
.It Bq Er EINVAL
.Dv MAP_ANON
was specified and the
.Fa fd
argument was not -1.
.It Bq Er EINVAL
.Dv MAP_ANON
was specified and the
.Fa offset
argument was not 0.
.It Bq Er EINVAL
Both
.Dv MAP_FIXED
and
.Dv MAP_EXCL
were specified, but the requested region is already used by a mapping.
.It Bq Er EINVAL
.Dv MAP_EXCL
was specified, but
.Dv MAP_FIXED
was not.
.It Bq Er ENODEV
.Dv MAP_ANON
has not been specified and
.Fa fd
did not reference a regular or character special file.
.It Bq Er ENOMEM
.Dv MAP_FIXED
was specified and the
.Fa addr
argument was not available.
.Dv MAP_ANON
was specified and insufficient memory was available.
.El
.Sh SEE ALSO
.Xr madvise 2 ,
.Xr mincore 2 ,
.Xr minherit 2 ,
.Xr mlock 2 ,
.Xr mprotect 2 ,
.Xr msync 2 ,
.Xr munlock 2 ,
.Xr munmap 2 ,
.Xr getpagesize 3 ,
.Xr getpagesizes 3