Always wait for txg sync when umounting dataset

Currently, when unmounting a filesystem, ZFS will only wait for
a txg sync if the dataset is dirty and not readonly. However, this
can be problematic in cases where a dataset is remounted readonly
immediately before being unmounted, which often happens when the
system is being shut down. Since encrypted datasets require that
all I/O is completed before the dataset is disowned, this issue
causes problems when write I/Os leak into the txgs after the
dataset is disowned, which can happen when sync=disabled.

While looking into fixes for this issue, it was discovered that
dsl_dataset_is_dirty() does not return B_TRUE when the dataset has
been removed from the txg dirty datasets list, but has not actually
been processed yet. Furthermore, the implementation is comletely
different from dmu_objset_is_dirty(), adding to the confusion.
Rather than relying on this function, this patch forces the umount
code path (and the remount readonly code path) to always perform a
txg sync on read-write datasets and removes the function altogether.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Tom Caputi <tcaputi@datto.com>
Closes #7753
Closes #7795

include/sys/dsl_dataset.h

@@ -395,7 +395,6 @@ int dsl_dataset_space_written(dsl_dataset_t *oldsnap, dsl_dataset_t *new,
     uint64_t *usedp, uint64_t *compp, uint64_t *uncompp);
 int dsl_dataset_space_wouldfree(dsl_dataset_t *firstsnap, dsl_dataset_t *last,
     uint64_t *usedp, uint64_t *compp, uint64_t *uncompp);
-boolean_t dsl_dataset_is_dirty(dsl_dataset_t *ds);
 
 int dsl_dsobj_to_dsname(char *pname, uint64_t obj, char *buf);
 

module/zfs/dsl_dataset.c

@@ -794,15 +794,6 @@ dsl_dataset_rele_flags(dsl_dataset_t *ds, ds_hold_flags_t flags, void *tag)
 	    (flags & DS_HOLD_FLAG_DECRYPT)) {
 		(void) spa_keystore_remove_mapping(ds->ds_dir->dd_pool->dp_spa,
 		    ds->ds_object, ds);
-
-		/*
-		 * Encrypted datasets require that users only release their
-		 * decrypting reference after the dirty data has actually
-		 * been written out. This ensures that the mapping exists
-		 * when it is needed to write out dirty data.
-		 */
-		ASSERT(dmu_buf_user_refcount(ds->ds_dbuf) != 0 ||
-		    !dsl_dataset_is_dirty(ds));
 	}
 
 	dmu_buf_rele(ds->ds_dbuf, tag);
@@ -1168,17 +1159,6 @@ dsl_dataset_dirty(dsl_dataset_t *ds, dmu_tx_t *tx)
 	}
 }
 
-boolean_t
-dsl_dataset_is_dirty(dsl_dataset_t *ds)
-{
-	for (int t = 0; t < TXG_SIZE; t++) {
-		if (txg_list_member(&ds->ds_dir->dd_pool->dp_dirty_datasets,
-		    ds, t))
-			return (B_TRUE);
-	}
-	return (B_FALSE);
-}
-
 static int
 dsl_dataset_snapshot_reserve_space(dsl_dataset_t *ds, dmu_tx_t *tx)
 {

module/zfs/zfs_vfsops.c

@@ -1745,10 +1745,10 @@ zfsvfs_teardown(zfsvfs_t *zfsvfs, boolean_t unmounting)
 	zfs_unregister_callbacks(zfsvfs);
 
 	/*
-	 * Evict cached data
+	 * Evict cached data. We must write out any dirty data before
+	 * disowning the dataset.
 	 */
-	if (dsl_dataset_is_dirty(dmu_objset_ds(zfsvfs->z_os)) &&
+	if (!zfs_is_readonly(zfsvfs))
-	    !zfs_is_readonly(zfsvfs))
 		txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), 0);
 	dmu_objset_evict_dbufs(zfsvfs->z_os);
 
@@ -1970,6 +1970,9 @@ zfs_remount(struct super_block *sb, int *flags, zfs_mnt_t *zm)
 	if (error)
 		return (error);
 
+	if (!zfs_is_readonly(zfsvfs) && (*flags & MS_RDONLY))
+		txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), 0);
+
 	zfs_unregister_callbacks(zfsvfs);
 	zfsvfs_vfs_free(zfsvfs->z_vfs);
 

module/zfs/zvol.c

@@ -1193,10 +1193,10 @@ zvol_shutdown_zv(zvol_state_t *zv)
 	zv->zv_dn = NULL;
 
 	/*
-	 * Evict cached data
+	 * Evict cached data. We must write out any dirty data before
+	 * disowning the dataset.
 	 */
-	if (dsl_dataset_is_dirty(dmu_objset_ds(zv->zv_objset)) &&
+	if (!(zv->zv_flags & ZVOL_RDONLY))
-	    !(zv->zv_flags & ZVOL_RDONLY))
 		txg_wait_synced(dmu_objset_pool(zv->zv_objset), 0);
 	(void) dmu_objset_evict_dbufs(zv->zv_objset);
 }
@@ -1489,8 +1489,7 @@ zvol_ioctl(struct block_device *bdev, fmode_t mode,
 		invalidate_bdev(bdev);
 		rw_enter(&zv->zv_suspend_lock, RW_READER);
 
-		if (dsl_dataset_is_dirty(dmu_objset_ds(zv->zv_objset)) &&
+		if (!(zv->zv_flags & ZVOL_RDONLY))
-		    !(zv->zv_flags & ZVOL_RDONLY))
 			txg_wait_synced(dmu_objset_pool(zv->zv_objset), 0);
 
 		rw_exit(&zv->zv_suspend_lock);

tests/zfs-tests/tests/functional/cli_root/zfs_mount/zfs_mount_remount.ksh

@@ -36,8 +36,10 @@
 # 2. Verify we can (re)mount the dataset readonly/read-write
 # 3. Verify we can mount the snapshot and it's mounted readonly
 # 4. Verify we can't remount it read-write
-# 5. Re-import the pool readonly
+# 5. Verify we can remount a dataset readonly and unmount it with
-# 6. Verify we can't remount its filesystem read-write
+#    encryption=on and sync=disabled (issue #7753)
+# 6. Re-import the pool readonly
+# 7. Verify we can't remount its filesystem read-write
 #
 
 verify_runnable "both"
@@ -130,11 +132,21 @@ readonlyfs $MNTPSNAP
 checkmount $TESTSNAP 'ro'
 log_must umount $MNTPSNAP
 
-# 5. Re-import the pool readonly
+# 5. Verify we can remount a dataset readonly and unmount it with
+#    encryption=on and sync=disabled (issue #7753)
+log_must eval "echo 'password' | zfs create -o sync=disabled \
+    -o encryption=on -o keyformat=passphrase $TESTFS/crypt"
+CRYPT_MNTPFS="$(get_prop mountpoint $TESTFS/crypt)"
+log_must touch $CRYPT_MNTPFS/file.dat
+log_must mount -o remount,ro $TESTFS/crypt $CRYPT_MNTPFS
+log_must umount -f $CRYPT_MNTPFS
+zpool sync $TESTPOOL
+
+# 6. Re-import the pool readonly
 log_must zpool export $TESTPOOL
 log_must zpool import -o readonly=on $TESTPOOL
 
-# 6. Verify we can't remount its filesystem read-write
+# 7. Verify we can't remount its filesystem read-write
 readonlyfs $MNTPFS
 checkmount $TESTFS 'ro'
 log_mustnot mount -o remount,rw $MNTPFS