proc.5: Document /proc/[pid]/io file

Attempt to document fields in the /proc/[pid]/io file, based on
the Documentation/filesystems/proc.txt. The text will probably
need some grammar corrections.

Signed-off-by: Peter Schiffer <pschiffe@redhat.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

man5/proc.5

@@ -350,9 +350,96 @@ file access mode and file status flags (see
 .BR open (2)).
 
 The files in this directory are readable only by the owner of the process.
-.\" FIXME document /proc/[pid]/io
-.\" .TP
-.\" .IR /proc/[pid]/io " (since kernel 2.6.20)"
+.TP
+.IR /proc/[pid]/io " (since kernel 2.6.20)"
+This file contains IO statistics for each running process, for example:
+.in +4n
+.nf
+
+.RB "#" " cat /proc/3828/io"
+rchar: 323934931
+wchar: 323929600
+syscr: 632687
+syscw: 632675
+read_bytes: 0
+write_bytes: 323932160
+cancelled_write_bytes: 0
+.fi
+.in
+
+The fields are as follows:
+.RS
+.IP * 2
+.IR rchar :
+chars read
+
+The number of bytes which this task has caused to be read from storage.
+This is simply the sum of bytes which this process passed to
+.BR read (2)
+and
+.BR pread (2).
+It includes things like tty IO and it is unaffected by whether or not actual
+physical disk IO was required (the read might have been satisfied from
+pagecache).
+.IP *
+.IR wchar :
+chars written
+
+The number of bytes which this task has caused, or shall cause to be written
+to disk.
+Similar caveats apply here as with
+.IR rchar .
+.IP *
+.IR syscr :
+read syscalls
+
+Attempt to count the number of read I/O operations, i.e. syscalls like
+.BR read (2)
+and
+.BR pread (2).
+.IP *
+.IR syscw :
+write syscalls
+
+Attempt to count the number of write I/O operations, i.e. syscalls like
+.BR write (2)
+and
+.BR pwrite (2).
+.IP *
+.IR read_bytes :
+bytes read
+
+Attempt to count the number of bytes which this process really did cause to
+be fetched from the storage layer.
+This is accurate for block-backed filesystems.
+.IP *
+.IR write_bytes :
+bytes written
+
+Attempt to count the number of bytes which this process caused to be sent to
+the storage layer.
+.IP *
+.IR cancelled_write_bytes :
+
+The big inaccuracy here is truncate.
+If a process writes 1MB to a file and then deletes the file,
+it will in fact perform no writeout.
+But it will have been accounted as having caused 1MB of write.
+In other words: this field represents number of bytes which this process
+caused to not happen, by truncating pagecache.
+A task can cause "negative" IO too.
+If this task truncates some dirty pagecache,
+some IO which another task has been accounted for
+(in its write_bytes) will not be happening.
+.IP
+.IR Note :
+
+At its current implementation state, this is a bit racy on 32-bit systems:
+if process A reads process B's
+.I /proc/[pid]/io
+while process B is updating one of those 64-bit counters,
+process A could see an intermediate result.
+.RE
 .TP
 .IR /proc/[pid]/limits " (since kernel 2.6.24)"
 This file displays the soft limit, hard limit, and units of measurement