Allow stuck spa_namespace lock to be stolen

Signed-off-by: Don Brady <don.brady@klarasystems.com>
don-brady · Mar 6, 2024 · dfcfa28 · dfcfa28
1 parent 9e4aac1
commit dfcfa28
Show file tree

Hide file tree

Showing 17 changed files with 313 additions and 175 deletions.
diff --git a/cmd/zdb/zdb.c b/cmd/zdb/zdb.c
@@ -6912,11 +6912,11 @@ zdb_set_skip_mmp(char *target)
 	 * Disable the activity check to allow examination of
 	 * active pools.
 	 */
-	mutex_enter(&spa_namespace_lock);
+	spa_namespace_enter(FTAG);
 	if ((spa = spa_lookup(target)) != NULL) {
 		spa->spa_import_flags |= ZFS_IMPORT_SKIP_MMP;
 	}
-	mutex_exit(&spa_namespace_lock);
+	spa_namespace_exit(FTAG);
 }
 
 #define	BOGUS_SUFFIX "_CHECKPOINTED_UNIVERSE"
@@ -8710,13 +8710,13 @@ main(int argc, char **argv)
 				 * try opening the pool after clearing the
 				 * log state.
 				 */
-				mutex_enter(&spa_namespace_lock);
+				spa_namespace_enter(FTAG);
 				if ((spa = spa_lookup(target)) != NULL &&
 				    spa->spa_log_state == SPA_LOG_MISSING) {
 					spa->spa_log_state = SPA_LOG_CLEAR;
 					error = 0;
 				}
-				mutex_exit(&spa_namespace_lock);
+				spa_namespace_exit(FTAG);
 
 				if (!error) {
 					error = spa_open_rewind(target, &spa,

diff --git a/cmd/ztest/ztest.c b/cmd/ztest/ztest.c
@@ -1183,9 +1183,9 @@ ztest_kill(ztest_shared_t *zs)
 	 * Before we kill off ztest, make sure that the config is updated.
 	 * See comment above spa_write_cachefile().
 	 */
-	mutex_enter(&spa_namespace_lock);
+	spa_namespace_enter(FTAG);
 	spa_write_cachefile(ztest_spa, B_FALSE, B_FALSE, B_FALSE);
-	mutex_exit(&spa_namespace_lock);
+	spa_namespace_exit(FTAG);
 
 	(void) kill(getpid(), SIGKILL);
 }
@@ -3571,10 +3571,10 @@ ztest_split_pool(ztest_ds_t *zd, uint64_t id)
 
 	if (error == 0) {
 		(void) printf("successful split - results:\n");
-		mutex_enter(&spa_namespace_lock);
+		spa_namespace_enter(FTAG);
 		show_pool_stats(spa);
 		show_pool_stats(spa_lookup("splitp"));
-		mutex_exit(&spa_namespace_lock);
+		spa_namespace_exit(FTAG);
 		++zs->zs_splits;
 		--zs->zs_mirrors;
 	}
@@ -6858,11 +6858,11 @@ ztest_walk_pool_directory(char *header)
 	if (ztest_opts.zo_verbose >= 6)
 		(void) printf("%s\n", header);
 
-	mutex_enter(&spa_namespace_lock);
+	spa_namespace_enter(FTAG);
 	while ((spa = spa_next(spa)) != NULL)
 		if (ztest_opts.zo_verbose >= 6)
 			(void) printf("\t%s\n", spa_name(spa));
-	mutex_exit(&spa_namespace_lock);
+	spa_namespace_exit(FTAG);
 }
 
 static void
@@ -7604,11 +7604,11 @@ ztest_run(ztest_shared_t *zs)
 	/*
 	 * Verify that we can loop over all pools.
 	 */
-	mutex_enter(&spa_namespace_lock);
+	spa_namespace_enter(FTAG);
 	for (spa = spa_next(NULL); spa != NULL; spa = spa_next(spa))
 		if (ztest_opts.zo_verbose > 3)
 			(void) printf("spa_next: found %s\n", spa_name(spa));
-	mutex_exit(&spa_namespace_lock);
+	spa_namespace_exit(FTAG);
 
 	/*
 	 * Verify that we can export the pool and reimport it under a

diff --git a/include/sys/spa.h b/include/sys/spa.h
@@ -828,9 +828,6 @@ extern void spa_sync_allpools(void);
 
 extern int zfs_sync_pass_deferred_free;
 
-/* spa namespace global mutex */
-extern kmutex_t spa_namespace_lock;
-
 /*
  * SPA configuration functions in spa_config.c
  */
@@ -984,6 +981,11 @@ extern void spa_config_enter(spa_t *spa, int locks, const void *tag, krw_t rw);
 extern void spa_config_exit(spa_t *spa, int locks, const void *tag);
 extern int spa_config_held(spa_t *spa, int locks, krw_t rw);
 
+extern void spa_namespace_enter(const char *tag);
+extern void spa_namespace_exit(const char *tag);
+extern boolean_t spa_namespace_held(const char *tag);
+extern boolean_t spa_namespace_tryenter(const char *tag);
+
 /* Pool vdev add/remove lock */
 extern uint64_t spa_vdev_enter(spa_t *spa);
 extern uint64_t spa_vdev_detach_enter(spa_t *spa, uint64_t guid);

diff --git a/module/os/linux/zfs/zvol_os.c b/module/os/linux/zfs/zvol_os.c
@@ -570,8 +570,8 @@ zvol_open(struct block_device *bdev, fmode_t flag)
 		 * the kernel so the only option is to return the error for
 		 * the caller to handle it.
 		 */
-		if (!mutex_owned(&spa_namespace_lock)) {
-			if (!mutex_tryenter(&spa_namespace_lock)) {
+		if (!spa_namespace_held(FTAG)) {
+			if (!spa_namespace_tryenter(FTAG)) {
 				mutex_exit(&zv->zv_state_lock);
 				rw_exit(&zv->zv_suspend_lock);
 
@@ -593,7 +593,7 @@ zvol_open(struct block_device *bdev, fmode_t flag)
 		error = -zvol_first_open(zv, !(flag & FMODE_WRITE));
 
 		if (drop_namespace)
-			mutex_exit(&spa_namespace_lock);
+			spa_namespace_exit(FTAG);
 	}
 
 	if (error == 0) {

diff --git a/module/zfs/arc.c b/module/zfs/arc.c
@@ -8533,7 +8533,7 @@ l2arc_dev_get_next(void)
 	 * of cache devices (l2arc_dev_mtx).  Once a device has been selected,
 	 * both locks will be dropped and a spa config lock held instead.
 	 */
-	mutex_enter(&spa_namespace_lock);
+	spa_namespace_enter(FTAG);
 	mutex_enter(&l2arc_dev_mtx);
 
 	/* if there are no vdevs, there is nothing to do */
@@ -8577,7 +8577,7 @@ l2arc_dev_get_next(void)
 	 */
 	if (next != NULL)
 		spa_config_enter(next->l2ad_spa, SCL_L2ARC, next, RW_READER);
-	mutex_exit(&spa_namespace_lock);
+	spa_namespace_exit(FTAG);
 
 	return (next);
 }
@@ -10188,7 +10188,7 @@ l2arc_stop(void)
 void
 l2arc_spa_rebuild_start(spa_t *spa)
 {
-	ASSERT(MUTEX_HELD(&spa_namespace_lock));
+	ASSERT(spa_namespace_held(FTAG));
 
 	/*
 	 * Locate the spa's l2arc devices and kick off rebuild threads.
@@ -10213,7 +10213,7 @@ l2arc_spa_rebuild_start(spa_t *spa)
 void
 l2arc_spa_rebuild_stop(spa_t *spa)
 {
-	ASSERT(MUTEX_HELD(&spa_namespace_lock));
+	ASSERT(spa_namespace_held(FTAG));
 
 	/*
 	 * Locate the spa's l2arc devices and kick off rebuild threads.

diff --git a/module/zfs/mmp.c b/module/zfs/mmp.c
@@ -723,12 +723,12 @@ mmp_signal_all_threads(void)
 {
 	spa_t *spa = NULL;
 
-	mutex_enter(&spa_namespace_lock);
+	spa_namespace_enter(FTAG);
 	while ((spa = spa_next(spa))) {
 		if (spa->spa_state == POOL_STATE_ACTIVE)
 			mmp_signal_thread(spa);
 	}
-	mutex_exit(&spa_namespace_lock);
+	spa_namespace_exit(FTAG);
 }
 
 /* BEGIN CSTYLED */