port async unlinked drain from illumos-nexenta

This patch is an async implementation of the existing sync
zfs_unlinked_drain() function. This function is called at mount time and
is responsible for freeing znodes that we didn't get to freeing before.
We don't have to hold mounting of the dataset until the unlinked list is
fully drained as is done now. Since we can process the unlinked set
asynchronously this results in a better user experience when mounting a
dataset with entries in the unlinked set.

Reviewed by: Jorgen Lundman <[email protected]>
Reviewed by: Tom Caputi <[email protected]>
Reviewed by: Brian Behlendorf <[email protected]>
Reviewed-by: Matt Ahrens <[email protected]>
Reviewed by: Paul Dagnelie <[email protected]>
Signed-off-by: Alek Pinchuk <[email protected]>
Closes #8142 

1 files changed, 31 insertions, 5 deletions

diff --git a/module/zfs/zfs_vfsops.c b/module/zfs/zfs_vfsops.c
index 766cbab74..cdc1bc707 100644
--- a/module/zfs/zfs_vfsops.c
+++ b/module/zfs/zfs_vfsops.c
@@ -1178,6 +1178,10 @@ zfsvfs_create_impl(zfsvfs_t **zfvp, zfsvfs_t *zfsvfs, objset_t *os)
 		return (error);
 	}
 
+	zfsvfs->z_drain_task = TASKQID_INVALID;
+	zfsvfs->z_draining = B_FALSE;
+	zfsvfs->z_drain_cancel = B_TRUE;
+
 	*zfvp = zfsvfs;
 	return (0);
 }
@@ -1200,14 +1204,27 @@ zfsvfs_setup(zfsvfs_t *zfsvfs, boolean_t mounting)
 	 * operations out since we closed the ZIL.
 	 */
 	if (mounting) {
+		ASSERT3P(zfsvfs->z_kstat.dk_kstats, ==, NULL);
+		dataset_kstats_create(&zfsvfs->z_kstat, zfsvfs->z_os);
+
 		/*
 		 * During replay we remove the read only flag to
 		 * allow replays to succeed.
 		 */
-		if (readonly != 0)
+		if (readonly != 0) {
 			readonly_changed_cb(zfsvfs, B_FALSE);
-		else
+		} else {
+			zap_stats_t zs;
+			if (zap_get_stats(zfsvfs->z_os, zfsvfs->z_unlinkedobj,
+			    &zs) == 0) {
+				dataset_kstats_update_nunlinks_kstat(
+				    &zfsvfs->z_kstat, zs.zs_num_entries);
+			}
+			dprintf_ds(zfsvfs->z_os->os_dsl_dataset,
+			    "num_entries in unlinked set: %llu",
+			    zs.zs_num_entries);
 			zfs_unlinked_drain(zfsvfs);
+		}
 
 		/*
 		 * Parse and replay the intent log.
@@ -1250,9 +1267,6 @@ zfsvfs_setup(zfsvfs_t *zfsvfs, boolean_t mounting)
 		/* restore readonly bit */
 		if (readonly != 0)
 			readonly_changed_cb(zfsvfs, B_TRUE);
-
-		ASSERT3P(zfsvfs->z_kstat.dk_kstats, ==, NULL);
-		dataset_kstats_create(&zfsvfs->z_kstat, zfsvfs->z_os);
 	}
 
 	/*
@@ -1633,6 +1647,8 @@ zfsvfs_teardown(zfsvfs_t *zfsvfs, boolean_t unmounting)
 {
 	znode_t	*zp;
 
+	zfs_unlinked_drain_stop_wait(zfsvfs);
+
 	/*
 	 * If someone has not already unmounted this file system,
 	 * drain the iput_taskq to ensure all active references to the
@@ -1884,6 +1900,7 @@ zfs_preumount(struct super_block *sb)
 
 	/* zfsvfs is NULL when zfs_domount fails during mount */
 	if (zfsvfs) {
+		zfs_unlinked_drain_stop_wait(zfsvfs);
 		zfsctl_destroy(sb->s_fs_info);
 		/*
 		 * Wait for iput_async before entering evict_inodes in
@@ -2159,6 +2176,15 @@ zfs_resume_fs(zfsvfs_t *zfsvfs, dsl_dataset_t *ds)
 	}
 	mutex_exit(&zfsvfs->z_znodes_lock);
 
+	if (!zfs_is_readonly(zfsvfs) && !zfsvfs->z_unmounted) {
+		/*
+		 * zfs_suspend_fs() could have interrupted freeing
+		 * of dnodes. We need to restart this freeing so
+		 * that we don't "leak" the space.
+		 */
+		zfs_unlinked_drain(zfsvfs);
+	}
+
 bail:
 	/* release the VFS ops */
 	rw_exit(&zfsvfs->z_teardown_inactive_lock);