fsync.2: Various improvements

- explain the situation with disk caches better
- remove the duplicate fdatasync() explanation in the NOTES
  section
- remove an incorrect note about fsync() generally requiring two
  writes
- remove an obsolete ext2 example note
- fsync() works on any file descriptor (doesn't need to be
  writable); correct the EBADF error code explanation

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

man2/fsync.2

@@ -63,12 +63,15 @@ transfers ("flushes") all modified in-core data of
 (i.e., modified buffer cache pages for) the
 file referred to by the file descriptor
 .I fd
-to the disk device (or other permanent storage device)
-where that file resides.
+to the disk device (or other permanent storage device) so that all
+changed information can be retrieved even after the system crashed or
+was rebooted.  This includes writing through or flushing a disk cache
+if present.
 The call blocks until the device reports that the transfer has completed.
 It also flushes metadata information associated with the file (see
 .BR stat (2)).
 
+
 Calling
 .BR fsync ()
 does not necessarily ensure
@@ -111,7 +114,7 @@ is set appropriately.
 .TP
 .B EBADF
 .I fd
-is not a valid file descriptor open for writing.
+is not a valid open file descriptor.
 .TP
 .B EIO
 An error occurred during synchronization.
@@ -135,49 +138,21 @@ to a value greater than 0.
 .\" -1: unavailable, 0: ask using sysconf().
 .\" glibc defines them to 1.
 .SH NOTES
-Applications that access databases or log files often write a tiny
-data fragment (e.g., one line in a log file) and then call
-.BR fsync ()
-immediately in order to ensure that the written data is physically
-stored on the harddisk.
-Unfortunately,
-.BR fsync ()
-will always initiate two write operations: one for the newly written
-data and another one in order to update the modification time stored
-in the inode.
-If the modification time is not a part of the transaction
-concept
-.BR fdatasync ()
-can be used to avoid unnecessary inode disk write operations.
-
-If the underlying hard disk has write caching enabled, then
-the data may not really be on permanent storage when
-.BR fsync ()
-/
-.BR fdatasync ()
-return.
-.\" See
-.\" .BR hdparm (8)
-.\" for how to disable that cache for IDE disks.
-.LP
-When an ext2 file system is mounted with the
-.I sync
-option, directory entries are also implicitly synced by
-.BR fsync ().
-.LP
-On kernels before 2.4,
-.BR fsync ()
-on big files can be inefficient.
-An alternative might be to use the
-.B O_SYNC
-flag to
-.BR open (2).
-
 In Linux 2.2 and earlier,
 .BR fdatasync ()
 is equivalent to
 .BR fsync (),
 and so has no performance advantage.
+
+The
+.BR fsync ()
+implementations in older kernels and lesser used filesystems
+does not know how to flush disk caches.  In these cases disk caches need to
+be disabled using
+.BR hdparm (8)
+or
+.BR sdparm (8)
+to guarantee safe operation.
 .SH "SEE ALSO"
 .BR bdflush (2),
 .BR open (2),