Hi,
On Fri 27-01-12 16:15:48, Jeff Moyer wrote:
If a file is opened with O_SYNC|O_DIRECT, the drive cache does not get
flushed after the write completion. Instead, it's flushed *before* the
I/O is sent to the disk (in __generic_file_aio_write). This patch
attempts to fix that problem by marking an I/O as requiring a cache
flush in endio processing. I'll send a follow-on patch to the
generic write code to get rid of the bogus generic_write_sync call
when EIOCBQUEUED is returned.
Thanks for the patch!
quoted hunk ↗ jump to hunk
Signed-off-by: Jeff Moyer <redacted>
---
fs/ext4/ext4.h | 4 ++++
fs/ext4/inode.c | 11 +++++++++--
fs/ext4/page-io.c | 39 ++++++++++++++++++++++++++++++++-------
fs/ext4/super.c | 11 +++++++++++
4 files changed, 56 insertions(+), 9 deletions(-)
diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index 2d55d7c..4377ed3 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -185,6 +185,7 @@ struct mpage_da_data {
#define EXT4_IO_END_ERROR 0x0002
#define EXT4_IO_END_QUEUED 0x0004
#define EXT4_IO_END_DIRECT 0x0008
+#define EXT4_IO_END_NEEDS_SYNC 0x0010
struct ext4_io_page {
struct page *p_page;@@ -1247,6 +1248,9 @@ struct ext4_sb_info {
/* workqueue for dio unwritten */
struct workqueue_struct *dio_unwritten_wq;
+ /* workqueue for aio+dio+o_sync disk cache flushing */
+ struct workqueue_struct *aio_dio_flush_wq;
+
Hmm, looking at the patch I'm wondering why did you introduce the new
workqueue? It seems dio_unwritten_wq would be enough? You just need to
rename it to something more appropriate ;)
quoted hunk ↗ jump to hunk
diff --git a/fs/ext4/page-io.c b/fs/ext4/page-io.c
index 9e1b8eb..d07cd40 100644
--- a/fs/ext4/page-io.c
+++ b/fs/ext4/page-io.c
@@ -98,15 +98,40 @@ int ext4_end_io_nolock(ext4_io_end_t *io)
"list->prev 0x%p\n",
io, inode->i_ino, io->list.next, io->list.prev);
- ret = ext4_convert_unwritten_extents(inode, offset, size);
- if (ret < 0) {
- ext4_msg(inode->i_sb, KERN_EMERG,
- "failed to convert unwritten extents to written "
- "extents -- potential data loss! "
- "(inode %lu, offset %llu, size %zd, error %d)",
- inode->i_ino, offset, size, ret);
+ if (io->flag & EXT4_IO_END_UNWRITTEN) {
+
+ ret = ext4_convert_unwritten_extents(inode, offset, size);
+ if (ret < 0) {
+ ext4_msg(inode->i_sb, KERN_EMERG,
+ "failed to convert unwritten extents to "
+ "written extents -- potential data loss! "
+ "(inode %lu, offset %llu, size %zd, error %d)",
+ inode->i_ino, offset, size, ret);
+ goto endio;
+ }
+ }
+
+ /*
+ * This function has two callers. The first is the end_io_work
+ * routine just below. This is an asynchronous completion context.
+ * The second is in the fsync path. For the latter path, we can't
+ * return from here until the job is done. Hence, we issue a
+ * blocking blkdev_issue_flush call.
+ */
+ if (io->flag & EXT4_IO_END_NEEDS_SYNC) {
+ /*
+ * Ideally, we'd like to know if the force_commit routine
+ * actually did send something to disk. If it didn't,
+ * then we need to issue the cache flush by hand. For now,
+ * play it safe and do both.
+ */
+ ret = ext4_force_commit(inode->i_sb);
+ if (ret)
+ goto endio;
+ ret = blkdev_issue_flush(inode->i_sb->s_bdev, GFP_NOIO, NULL);
Look at what ext4_sync_file() does. It's more efficient than this.
You need something like:
commit_tid = file->f_flags & __O_SYNC ? EXT4_I(inode)->i_sync_tid :
EXT4_I(inode)->i_datasync_tid;
if (journal->j_flags & JBD2_BARRIER &&
!jbd2_trans_will_send_data_barrier(journal, commit_tid))
needs_barrier = true;
jbd2_log_start_commit(journal, commit_tid);
jbd2_log_wait_commit(journal, commit_tid);
if (needs_barrier)
blkdev_issue_flush(inode->i_sb->s_bdev, GFP_NOIO, NULL);
Honza
--
Jan Kara [off-list ref]
SUSE Labs, CR