.\" Modified Sat Jul 12 20:45:39 1997 by Michael Haardt
.\" <michael@cantor.informatik.rwth-aachen.de>
.\"
-.TH READ 2 2013-02-12 "Linux" "Linux Programmer's Manual"
+.TH READ 2 2018-02-02 "Linux" "Linux Programmer's Manual"
.SH NAME
read \- read from a file descriptor
.SH SYNOPSIS
.nf
.B #include <unistd.h>
-.sp
+.PP
.BI "ssize_t read(int " fd ", void *" buf ", size_t " count );
.fi
.SH DESCRIPTION
.I fd
into the buffer starting at
.IR buf .
-
+.PP
On files that support seeking,
-the read operation commences at the current file offset,
+the read operation commences at the file offset,
and the file offset is incremented by the number of bytes read.
-If the current file offset is at or past the end of file,
+If the file offset is at or past the end of file,
no bytes are read, and
.BR read ()
returns zero.
-
+.PP
If
.I count
is zero,
with a
.I count
of 0 returns zero and has no other effects.
-
-If
+.PP
+According to POSIX.1, if
.I count
is greater than
.BR SSIZE_MAX ,
-the result is unspecified.
+the result is implementation-defined;
+see NOTES for the upper limit on Linux.
.SH RETURN VALUE
On success, the number of bytes read is returned (zero indicates end of
file), and the file position is advanced by this number.
because we are reading from a pipe, or from a terminal), or because
.BR read ()
was interrupted by a signal.
+See also NOTES.
+.PP
On error, \-1 is returned, and
.I errno
is set appropriately.
-In this case it is left unspecified whether
+In this case, it is left unspecified whether
the file position (if any) changes.
.SH ERRORS
.TP
refers to a file other than a socket and has been marked nonblocking
.RB ( O_NONBLOCK ),
and the read would block.
+See
+.BR open (2)
+for further details on the
+.BR O_NONBLOCK
+flag.
.TP
.BR EAGAIN " or " EWOULDBLOCK
.\" Actually EAGAIN on Linux
.IR buf ,
the value specified in
.IR count ,
-or the current file offset is not suitably aligned.
+or the file offset is not suitably aligned.
.TP
.B EINVAL
.I fd
is orphaned.
It may also occur when there is a low-level I/O error
while reading from a disk or tape.
+A further possible cause of
+.B EIO
+on networked filesystems is when an advisory lock had been taken
+out on the file descriptor and this lock has been lost.
+See the
+.I "Lost locks"
+section of
+.BR fcntl (2)
+for further details.
.TP
.B EISDIR
.I fd
.PP
Other errors may occur, depending on the object connected to
.IR fd .
-POSIX allows a
-.BR read ()
-that is interrupted after reading some data
-to return \-1 (with
-.I errno
-set to
-.BR EINTR )
-or to return the number of bytes already read.
.SH CONFORMING TO
SVr4, 4.3BSD, POSIX.1-2001.
.SH NOTES
-On NFS file systems, reading small amounts of data will update the
+The types
+.I size_t
+and
+.I ssize_t
+are, respectively,
+unsigned and signed integer data types specified by POSIX.1.
+.PP
+On Linux,
+.BR read ()
+(and similar system calls) will transfer at most
+0x7ffff000 (2,147,479,552) bytes,
+returning the number of bytes actually transferred.
+.\" commit e28cc71572da38a5a12c1cfe4d7032017adccf69
+(This is true on both 32-bit and 64-bit systems.)
+.PP
+On NFS filesystems, reading small amounts of data will update the
timestamp only the first time, subsequent calls may not do so.
This is caused
by client side attribute caching, because most if not all NFS clients
-leave st_atime (last file access time)
-updates to the server and client side reads satisfied from the
-client's cache will not cause st_atime updates on the server as there are no
-server side reads.
-UNIX semantics can be obtained by disabling client
-side attribute caching, but in most situations this will substantially
+leave
+.I st_atime
+(last file access time)
+updates to the server, and client side reads satisfied from the
+client's cache will not cause
+.I st_atime
+updates on the server as there are no
+server-side reads.
+UNIX semantics can be obtained by disabling client-side attribute caching,
+but in most situations this will substantially
increase server load and decrease performance.
+.SH BUGS
+According to POSIX.1-2008/SUSv4 Section XSI 2.9.7
+("Thread Interactions with Regular File Operations"):
+.PP
+.RS 4
+All of the following functions shall be atomic with respect to
+each other in the effects specified in POSIX.1-2008 when they
+operate on regular files or symbolic links: ...
+.RE
+.PP
+Among the APIs subsequently listed are
+.BR read ()
+and
+.BR readv (2).
+And among the effects that should be atomic across threads (and processes)
+are updates of the file offset.
+However, on Linux before version 3.14,
+this was not the case: if two processes that share
+an open file description (see
+.BR open (2))
+perform a
+.BR read ()
+(or
+.BR readv (2))
+at the same time, then the I/O operations were not atomic
+with respect updating the file offset,
+with the result that the reads in the two processes
+might (incorrectly) overlap in the blocks of data that they obtained.
+This problem was fixed in Linux 3.14.
+.\" http://thread.gmane.org/gmane.linux.kernel/1649458
+.\" From: Michael Kerrisk (man-pages <mtk.manpages <at> gmail.com>
+.\" Subject: Update of file offset on write() etc. is non-atomic with I/O
+.\" Date: 2014-02-17 15:41:37 GMT
+.\" Newsgroups: gmane.linux.kernel, gmane.linux.file-systems
+.\" commit 9c225f2655e36a470c4f58dbbc99244c5fc7f2d4
+.\" Author: Linus Torvalds <torvalds@linux-foundation.org>
+.\" Date: Mon Mar 3 09:36:58 2014 -0800
+.\"
+.\" vfs: atomic f_pos accesses as per POSIX
.SH SEE ALSO
.BR close (2),
.BR fcntl (2),