.\" Copyright 2003 Abhijit Menon-Sen <ams@wiw.org>
-.\" %%%LICENSE_START(verbatim)
+.\" and Copyright (C) 2010, 2015, 2017 Michael Kerrisk <mtk.manpages@gmail.com>
+.\"
+.\" %%%LICENSE_START(VERBATIM)
.\" Permission is granted to make and distribute verbatim copies of this
.\" manual provided the copyright notice and this permission notice are
.\" preserved on all copies.
.\" 2005-04-08 mtk, noted kernel version and added BUGS
.\" 2010-10-09, mtk, document arm_fadvise64_64()
.\"
-.TH POSIX_FADVISE 2 2010-10-09 "Linux" "Linux Programmer's Manual"
+.TH POSIX_FADVISE 2 2019-03-06 "Linux" "Linux Programmer's Manual"
.SH NAME
posix_fadvise \- predeclare an access pattern for file data
.SH SYNOPSIS
.nf
.B #include <fcntl.h>
-.sp
+.PP
.BI "int posix_fadvise(int " fd ", off_t " offset ", off_t " len \
", int " advice ");"
.fi
-.sp
+.PP
.ad l
.in -4n
Feature Test Macro Requirements for glibc (see
.BR feature_test_macros (7)):
.in
-.sp
+.PP
.BR posix_fadvise ():
.RS 4
-_XOPEN_SOURCE\ >=\ 600 || _POSIX_C_SOURCE\ >=\ 200112L
+_POSIX_C_SOURCE\ >=\ 200112L
.RE
.ad
.SH DESCRIPTION
to announce an intention to access
file data in a specific pattern in the future, thus allowing the kernel
to perform appropriate optimizations.
-
+.PP
The \fIadvice\fP applies to a (not necessarily existent) region starting
at \fIoffset\fP and extending for \fIlen\fP bytes (or until the end of
the file if \fIlen\fP is 0) within the file referred to by \fIfd\fP.
The \fIadvice\fP is not binding;
it merely constitutes an expectation on behalf of
the application.
-
+.PP
Permissible values for \fIadvice\fP include:
.TP
.B POSIX_FADV_NORMAL
.TP
.B POSIX_FADV_NOREUSE
The specified data will be accessed only once.
+.IP
+In kernels before 2.6.18, \fBPOSIX_FADV_NOREUSE\fP had the
+same semantics as \fBPOSIX_FADV_WILLNEED\fP.
+This was probably a bug; since kernel 2.6.18, this flag is a no-op.
.TP
.B POSIX_FADV_WILLNEED
The specified data will be accessed in the near future.
+.IP
+\fBPOSIX_FADV_WILLNEED\fP initiates a
+nonblocking read of the specified region into the page cache.
+The amount of data read may be decreased by the kernel depending
+on virtual memory load.
+(A few megabytes will usually be fully satisfied,
+and more is rarely useful.)
.TP
.B POSIX_FADV_DONTNEED
The specified data will not be accessed in the near future.
+.IP
+\fBPOSIX_FADV_DONTNEED\fP attempts to free cached pages associated with
+the specified region.
+This is useful, for example, while streaming large
+files.
+A program may periodically request the kernel to free cached data
+that has already been used, so that more useful cached pages are not
+discarded instead.
+.IP
+Requests to discard partial pages are ignored.
+It is preferable to preserve needed data than discard unneeded data.
+If the application requires that data be considered for discarding, then
+.I offset
+and
+.I len
+must be page-aligned.
+.IP
+The implementation
+.I may
+attempt to write back dirty pages in the specified region,
+but this is not guaranteed.
+Any unwritten dirty pages will not be freed.
+If the application wishes to ensure that dirty pages will be released,
+it should call
+.BR fsync (2)
+or
+.BR fdatasync (2)
+first.
.SH RETURN VALUE
On success, zero is returned.
On error, an error number is returned.
.TP
.B ESPIPE
The specified file descriptor refers to a pipe or FIFO.
-(Linux actually
-returns
+.RB ( ESPIPE
+is the error specified by POSIX,
+but before kernel version 2.6.16,
+.\" commit 87ba81dba431232548ce29d5d224115d0c2355ac
+Linux returned
.B EINVAL
in this case.)
.SH VERSIONS
Library support has been provided since glibc version 2.2,
via the wrapper function
.BR posix_fadvise ().
+.PP
+Since Linux 3.18,
+.\" commit d3ac21cacc24790eb45d735769f35753f5b56ceb
+support for the underlying system call is optional,
+depending on the setting of the
+.B CONFIG_ADVISE_SYSCALLS
+configuration option.
.SH CONFORMING TO
-POSIX.1-2001.
+POSIX.1-2001, POSIX.1-2008.
Note that the type of the
.I len
argument was changed from
this size, and \fBPOSIX_FADV_RANDOM\fP disables file readahead entirely.
These changes affect the entire file, not just the specified region
(but other open file handles to the same file are unaffected).
-
-\fBPOSIX_FADV_WILLNEED\fP initiates a
-nonblocking read of the specified region into the page cache.
-The amount of data read may be decreased by the kernel depending
-on virtual memory load.
-(A few megabytes will usually be fully satisfied,
-and more is rarely useful.)
-
-In kernels before 2.6.18, \fBPOSIX_FADV_NOREUSE\fP had the
-same semantics as \fBPOSIX_FADV_WILLNEED\fP.
-This was probably a bug; since kernel 2.6.18, this flag is a no-op.
-
-\fBPOSIX_FADV_DONTNEED\fP attempts to free cached pages associated with
-the specified region.
-This is useful, for example, while streaming large
-files.
-A program may periodically request the kernel to free cached data
-that has already been used, so that more useful cached pages are not
-discarded instead.
-
-Pages that have not yet been written out will be unaffected, so if the
-application wishes to guarantee that pages will be released, it should
-call
-.BR fsync (2)
-or
-.BR fdatasync (2)
-first.
-.SS arm_fadvise()
-The ARM architecture
-needs 64-bit arguments to be aligned in a suitable pair of registers.
-On this architecture, the call signature of
+.PP
+The contents of the kernel buffer cache can be cleared via the
+.IR /proc/sys/vm/drop_caches
+interface described in
+.BR proc (5).
+.PP
+One can obtain a snapshot of which pages of a file are resident
+in the buffer cache by opening a file, mapping it with
+.BR mmap (2),
+and then applying
+.BR mincore (2)
+to the mapping.
+.SS C library/kernel differences
+The name of the wrapper function in the C library is
+.BR posix_fadvise ().
+The underlying system call is called
+.BR fadvise64 ()
+(or, on some architectures,
+.BR fadvise64_64 ());
+the difference between the two is that the former system call
+assumes that the type of the \fIlen\fP argument is \fIsize_t\fP,
+while the latter expects \fIloff_t\fP there.
+.SS Architecture-specific variants
+Some architectures require
+64-bit arguments to be aligned in a suitable pair of registers (see
+.BR syscall (2)
+for further detail).
+On such architectures, the call signature of
.BR posix_fadvise ()
-is flawed, since it forces a register to be wasted as padding between the
+shown in the SYNOPSIS would force
+a register to be wasted as padding between the
.I fd
and
-.I len
+.I offset
arguments.
-Therefore, since Linux 2.6.14, ARM defines a different
-system call that orders the arguments suitably:
+Therefore, these architectures define a version of the
+system call that orders the arguments suitably,
+but is otherwise exactly the same as
+.BR posix_fadvise ().
+.PP
+For example, since Linux 2.6.14, ARM has the following system call:
.PP
.in +4n
-.nf
+.EX
.BI "long arm_fadvise64_64(int " fd ", int " advice ,
.BI " loff_t " offset ", loff_t " len );
-.fi
+.EE
.in
.PP
-The behavior of this system call is otherwise exactly the same as
-.BR posix_fadvise ().
-No library support is provided for this system call in glibc.
-.\" No ARM support in glibc.
+These architecture-specific details are generally
+hidden from applications by the glibc
+.BR posix_fadvise ()
+wrapper function,
+which invokes the appropriate architecture-specific system call.
.SH BUGS
In kernels before 2.6.6, if
.I len
was specified as 0, then this was interpreted literally as "zero bytes",
rather than as meaning "all bytes through to the end of the file".
.SH SEE ALSO
+.BR fincore (1),
+.BR mincore (2),
.BR readahead (2),
.BR sync_file_range (2),
.BR posix_fallocate (3),
.BR posix_madvise (3)
-.\" FIXME . Write a posix_fadvise(3) page.