Using sync_file_range means that neither any required metadata gets commited,
nor the disk cache gets flushed. Stop using it for the journal, and add
a comment on why a fsync_range system call would be helpful here.
Btw, why does the code use O_SYNC (and not even O_DSYNC!) if using direct
I/O, but fdatasync/fsync for buffered I/O? Avoiding cache flushes and
metadata updates for every writes is just as important for direct I/O
as it is for buffered I/O.
Signed-off-by: Christoph Hellwig <h...@lst.de>
Index: ceph/src/os/FileJournal.cc
===================================================================
--- ceph.orig/src/os/FileJournal.cc 2011-11-03 22:27:35.304690331 +0100
+++ ceph/src/os/FileJournal.cc 2011-11-03 22:40:02.244191569 +0100
@@ -842,23 +842,26 @@ void FileJournal::do_write(bufferlist& b
if (!directio) {
dout(20) << "do_write fsync" << dendl;
+
+ /*
+ * We'd really love to have a fsync_range or fdatasync_range and do a:
+ *
+ * if (split) {
+ * ::fsync_range(fd, header.max_size - split, split)l
+ * ::fsync_range(fd, get_top(), bl.length() - split);
+ * else
+ * ::fsync_range(fd, write_pos, bl.length())
+ *
+ * NetBSD and AIX apparently have it, and adding it to Linux wouldn't be
+ * too hard given all the underlying infrastructure already exist.
+ *
+ * NOTE: using sync_file_range here would not be safe as it does not
+ * flush disk caches or commits any sort of metadata.
+ */
#if defined(DARWIN) || defined(__FreeBSD__)
::fsync(fd);
#else
-# ifdef HAVE_SYNC_FILE_RANGE
- if (is_bdev) {
- if (split) {
- ::sync_file_range(fd, header.max_size - split, split,
SYNC_FILE_RANGE_WAIT_BEFORE|SYNC_FILE_RANGE_WRITE);
- ::sync_file_range(fd, get_top(), bl.length() - split,
SYNC_FILE_RANGE_WAIT_BEFORE|SYNC_FILE_RANGE_WRITE);
- ::sync_file_range(fd, header.max_size - split, split,
SYNC_FILE_RANGE_WAIT_AFTER);
- ::sync_file_range(fd, get_top(), bl.length() - split,
SYNC_FILE_RANGE_WAIT_AFTER);
- } else {
- ::sync_file_range(fd, write_pos, bl.length(),
-
SYNC_FILE_RANGE_WAIT_BEFORE|SYNC_FILE_RANGE_WRITE|SYNC_FILE_RANGE_WAIT_AFTER);
- }
- } else
-# endif
- ::fdatasync(fd);
+ ::fdatasync(fd);
#endif
}
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
