Long hold the dataset during upgrade

If the receive or rollback is performed while filesystem is upgrading
the objset may be evicted in `dsl_dataset_clone_swap_sync_impl`. This
will lead to NULL pointer dereference when upgrade tries to access
evicted objset.

This commit adds long hold of dataset during whole upgrade process.
The receive and rollback will return an EBUSY error until the
upgrade is not finished.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Arkadiusz Bubała <arkadiusz.bubala@open-e.com>
Closes #5295
Closes #6837

diff --git a/module/zfs/dmu_objset.c b/module/zfs/dmu_objset.c
index 9a7a6968d6319163b39ab2e5ea2be6f023ae731c..1706937aded919b5e6e7c260f16b7d481811ba1c 100644 (file)
--- a/module/zfs/dmu_objset.c
+++ b/module/zfs/dmu_objset.c
@@ -28,6 +28,7 @@
  * Copyright (c) 2015, STRATO AG, Inc. All rights reserved.
  * Copyright (c) 2016 Actifio, Inc. All rights reserved.
  * Copyright 2017 Nexenta Systems, Inc.
+ * Copyright (c) 2017 Open-E, Inc. All Rights Reserved.
  */
 
 /* Portions Copyright 2010 Robert Milkowski */
@@ -78,6 +79,8 @@ int dmu_find_threads = 0;
  */
 int dmu_rescan_dnode_threshold = 1 << DN_MAX_INDBLKSHIFT;
 
+static char *upgrade_tag = "upgrade_tag";
+
 static void dmu_objset_find_dp_cb(void *arg);
 
 static void dmu_objset_upgrade(objset_t *os, dmu_objset_upgrade_cb_t cb);
@@ -1157,6 +1160,7 @@ dmu_objset_upgrade_task_cb(void *data)
        os->os_upgrade_exit = B_TRUE;
        os->os_upgrade_id = 0;
        mutex_exit(&os->os_upgrade_lock);
+       dsl_dataset_long_rele(dmu_objset_ds(os), upgrade_tag);
 }
 
 static void
@@ -1165,6 +1169,9 @@ dmu_objset_upgrade(objset_t *os, dmu_objset_upgrade_cb_t cb)
        if (os->os_upgrade_id != 0)
                return;
 
+       ASSERT(dsl_pool_config_held(dmu_objset_pool(os)));
+       dsl_dataset_long_hold(dmu_objset_ds(os), upgrade_tag);
+
        mutex_enter(&os->os_upgrade_lock);
        if (os->os_upgrade_id == 0 && os->os_upgrade_status == 0) {
                os->os_upgrade_exit = B_FALSE;
@@ -1172,8 +1179,10 @@ dmu_objset_upgrade(objset_t *os, dmu_objset_upgrade_cb_t cb)
                os->os_upgrade_id = taskq_dispatch(
                    os->os_spa->spa_upgrade_taskq,
                    dmu_objset_upgrade_task_cb, os, TQ_SLEEP);
-               if (os->os_upgrade_id == TASKQID_INVALID)
+               if (os->os_upgrade_id == TASKQID_INVALID) {
+                       dsl_dataset_long_rele(dmu_objset_ds(os), upgrade_tag);
                        os->os_upgrade_status = ENOMEM;
+               }
        }
        mutex_exit(&os->os_upgrade_lock);
 }
@@ -1189,7 +1198,10 @@ dmu_objset_upgrade_stop(objset_t *os)
                os->os_upgrade_id = 0;
                mutex_exit(&os->os_upgrade_lock);
 
-               taskq_cancel_id(os->os_spa->spa_upgrade_taskq, id);
+               if ((taskq_cancel_id(os->os_spa->spa_upgrade_taskq, id)) == 0) {
+                       dsl_dataset_long_rele(dmu_objset_ds(os), upgrade_tag);
+               }
+               txg_wait_synced(os->os_spa->spa_dsl_pool, 0);
        } else {
                mutex_exit(&os->os_upgrade_lock);
        }

diff --git a/module/zfs/zfs_ioctl.c b/module/zfs/zfs_ioctl.c
index d195eded76dc818f2f5da68677ce53915449ba41..7106576c49e3755aff471cf38b470669f95ebdee 100644 (file)
--- a/module/zfs/zfs_ioctl.c
+++ b/module/zfs/zfs_ioctl.c
@@ -5181,9 +5181,6 @@ zfs_ioc_userobjspace_upgrade(zfs_cmd_t *zc)
        if (error != 0)
                return (error);
 
-       dsl_dataset_long_hold(dmu_objset_ds(os), FTAG);
-       dsl_pool_rele(dmu_objset_pool(os), FTAG);
-
        if (dmu_objset_userobjspace_upgradable(os)) {
                mutex_enter(&os->os_upgrade_lock);
                if (os->os_upgrade_id == 0) {
@@ -5196,8 +5193,12 @@ zfs_ioc_userobjspace_upgrade(zfs_cmd_t *zc)
                        mutex_exit(&os->os_upgrade_lock);
                }
 
+               dsl_pool_rele(dmu_objset_pool(os), FTAG);
+
                taskq_wait_id(os->os_spa->spa_upgrade_taskq, os->os_upgrade_id);
                error = os->os_upgrade_status;
+       } else {
+               dsl_pool_rele(dmu_objset_pool(os), FTAG);
        }
 
        dsl_dataset_long_rele(dmu_objset_ds(os), FTAG);

diff --git a/module/zfs/zfs_vfsops.c b/module/zfs/zfs_vfsops.c
index f97660f37a69466f70369a40bda3d60cd5980fa5..ea7d4ee59c1f1467154b3155e52bf6f34d29860b 100644 (file)
--- a/module/zfs/zfs_vfsops.c
+++ b/module/zfs/zfs_vfsops.c
@@ -834,8 +834,13 @@ zfs_fuid_overobjquota(zfsvfs_t *zfsvfs, boolean_t isgroup, uint64_t fuid)
        int err;
 
        if (!dmu_objset_userobjspace_present(zfsvfs->z_os)) {
-               if (dmu_objset_userobjspace_upgradable(zfsvfs->z_os))
+               if (dmu_objset_userobjspace_upgradable(zfsvfs->z_os)) {
+                       dsl_pool_config_enter(
+                           dmu_objset_pool(zfsvfs->z_os), FTAG);
                        dmu_objset_userobjspace_upgrade(zfsvfs->z_os);
+                       dsl_pool_config_exit(
+                           dmu_objset_pool(zfsvfs->z_os), FTAG);
+               }
                return (B_FALSE);
        }