Permalink
Browse files

Clear PG_writeback after zil_commit() for sync I/O

When writing via ->writepage() the writeback bit was always cleared
as part of the txg commit callback.  However, when the I/O is also
being written synchronsously to the zil we can immediately clear this
bit.  There is no need to wait for the subsequent TXG sync since the
data is already safe on stable storage.

This has been observed to reduce the msync(2) delay from up to 5
seconds down 10s of miliseconds.  One workload which is expected
to benefit from this are the intermittent samba hands described
in issue #700.

Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #700
Closes #907
  • Loading branch information...
1 parent e6f2905 commit 2b2861362f7dd09cc3167df8fddb6e2cb475018a @behlendorf behlendorf committed Aug 29, 2012
Showing with 8 additions and 3 deletions.
  1. +8 −3 module/zfs/zfs_vnops.c
View
@@ -3797,6 +3797,7 @@ zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc)
uint64_t mtime[2], ctime[2];
sa_bulk_attr_t bulk[3];
int cnt = 0;
+ int sync;
ASSERT(PageLocked(pp));
@@ -3833,7 +3834,10 @@ zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc)
tx = dmu_tx_create(zsb->z_os);
- dmu_tx_callback_register(tx, zfs_putpage_commit_cb, pp);
+ sync = ((zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) ||
+ (wbc->sync_mode == WB_SYNC_ALL));
+ if (!sync)
+ dmu_tx_callback_register(tx, zfs_putpage_commit_cb, pp);
dmu_tx_hold_write(tx, zp->z_id, pgoff, pglen);
@@ -3862,9 +3866,10 @@ zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc)
dmu_tx_commit(tx);
ASSERT3S(err, ==, 0);
- if ((zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) ||
- (wbc->sync_mode == WB_SYNC_ALL))
+ if (sync) {
zil_commit(zsb->z_log, zp->z_id);
+ zfs_putpage_commit_cb(pp, err);
+ }
return (err);
}

0 comments on commit 2b28613

Please sign in to comment.