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

diff --git a/module/zfs/zfs_vnops.c b/module/zfs/zfs_vnops.c
index 2da5fec8672c454759bc05dd94497888f77a3be8..755d0575ab3ccaf8462898e834e71a3a206b4501 100644 (file)
--- a/module/zfs/zfs_vnops.c
+++ b/module/zfs/zfs_vnops.c
@@ -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);
 }