deadlock between spa_errlog_lock and dp_config_rwlock

There is a lock order inversion deadlock between `spa_errlog_lock` and
`dp_config_rwlock`:

A thread in `spa_delete_dataset_errlog()` is running from a sync task.
It is holding the `dp_config_rwlock` for writer (see
`dsl_sync_task_sync()`), and waiting for the `spa_errlog_lock`.

A thread in `dsl_pool_config_enter()` is holding the `spa_errlog_lock`
(see `spa_get_errlog_size()`) and waiting for the `dp_config_rwlock` (as
reader).

Note that this was introduced by #12812.

This commit address this by defining the lock ordering to be
dp_config_rwlock first, then spa_errlog_lock / spa_errlist_lock.
spa_get_errlog() and spa_get_errlog_size() can acquire the locks in this
order, and then process_error_block() and get_head_and_birth_txg() can
verify that the dp_config_rwlock is already held.

Additionally, a buffer overrun in `spa_get_errlog()` is corrected.  Many
code paths didn't check if `*count` got to zero, instead continuing to
overwrite past the beginning of the userspace buffer at `uaddr`.

Tested by having some errors in the pool (via `zinject -t data
/path/to/file`), one thread running `zpool iostat 0.001`, and another
thread runs `zfs destroy` (in a loop, although it hits the first time).
This reproduces the problem easily without the fix, and works with the
fix.

Closes #14239
Signed-off-by: Matthew Ahrens <mahrens@delphix.com>

Author: ahrens

cmd/ztest.c

@@ -6313,7 +6313,7 @@ ztest_scrub_impl(spa_t *spa)
 	while (dsl_scan_scrubbing(spa_get_dsl(spa)))
 		txg_wait_synced(spa_get_dsl(spa), 0);
 
-	if (spa_get_errlog_size(spa) > 0)
+	if (spa_approx_errlog_size(spa) > 0)
 		return (ECKSUM);
 
 	ztest_pool_scrubbed = B_TRUE;

include/sys/spa.h

@@ -1146,7 +1146,7 @@ extern nvlist_t *zfs_event_create(spa_t *spa, vdev_t *vd, const char *type,
 extern void zfs_post_remove(spa_t *spa, vdev_t *vd);
 extern void zfs_post_state_change(spa_t *spa, vdev_t *vd, uint64_t laststate);
 extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd);
-extern uint64_t spa_get_errlog_size(spa_t *spa);
+extern uint64_t spa_approx_errlog_size(spa_t *spa);
 extern int spa_get_errlog(spa_t *spa, void *uaddr, uint64_t *count);
 extern void spa_errlog_rotate(spa_t *spa);
 extern void spa_errlog_drain(spa_t *spa);

lib/libzfs/libzfs_pool.c

@@ -4133,33 +4133,28 @@ zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp)
 {
 	zfs_cmd_t zc = {"\0"};
 	libzfs_handle_t *hdl = zhp->zpool_hdl;
-	uint64_t count;
-	zbookmark_phys_t *zb = NULL;
-	int i;
+	zbookmark_phys_t *buf;
+	uint64_t buflen = 10000; /* approx. 1MB of RAM */
+
+	if (fnvlist_lookup_uint64(zhp->zpool_config,
+	    ZPOOL_CONFIG_ERRCOUNT) == 0)
+		return (0);
 
 	/*
-	 * Retrieve the raw error list from the kernel.  If the number of errors
-	 * has increased, allocate more space and continue until we get the
-	 * entire list.
+	 * Retrieve the raw error list from the kernel.  If it doesn't fit,
+	 * allocate a larger buffer and retry.
 	 */
-	count = fnvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT);
-	if (count == 0)
-		return (0);
-	zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl,
-	    count * sizeof (zbookmark_phys_t));
-	zc.zc_nvlist_dst_size = count;
 	(void) strcpy(zc.zc_name, zhp->zpool_name);
 	for (;;) {
+		buf = zfs_alloc(zhp->zpool_hdl,
+		    buflen * sizeof (zbookmark_phys_t));
+		zc.zc_nvlist_dst = (uintptr_t)buf;
+		zc.zc_nvlist_dst_size = buflen;
 		if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_ERROR_LOG,
 		    &zc) != 0) {
-			free((void *)(uintptr_t)zc.zc_nvlist_dst);
+			free(buf);
 			if (errno == ENOMEM) {
-				void *dst;
-
-				count = zc.zc_nvlist_dst_size;
-				dst = zfs_alloc(zhp->zpool_hdl, count *
-				    sizeof (zbookmark_phys_t));
-				zc.zc_nvlist_dst = (uintptr_t)dst;
+				buflen *= 2;
 			} else {
 				return (zpool_standard_error_fmt(hdl, errno,
 				    dgettext(TEXT_DOMAIN, "errors: List of "
@@ -4177,18 +4172,17 @@ zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp)
 	 * _not_ copied as part of the process.  So we point the start of our
 	 * array appropriate and decrement the total number of elements.
 	 */
-	zb = ((zbookmark_phys_t *)(uintptr_t)zc.zc_nvlist_dst) +
-	    zc.zc_nvlist_dst_size;
-	count -= zc.zc_nvlist_dst_size;
+	zbookmark_phys_t *zb = buf + zc.zc_nvlist_dst_size;
+	uint64_t zblen = buflen - zc.zc_nvlist_dst_size;
 
-	qsort(zb, count, sizeof (zbookmark_phys_t), zbookmark_mem_compare);
+	qsort(zb, zblen, sizeof (zbookmark_phys_t), zbookmark_mem_compare);
 
 	verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0);
 
 	/*
 	 * Fill in the nverrlistp with nvlist's of dataset and object numbers.
 	 */
-	for (i = 0; i < count; i++) {
+	for (uint64_t i = 0; i < zblen; i++) {
 		nvlist_t *nv;
 
 		/* ignoring zb_blkid and zb_level for now */
@@ -4215,11 +4209,11 @@ zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp)
 		nvlist_free(nv);
 	}
 
-	free((void *)(uintptr_t)zc.zc_nvlist_dst);
+	free(buf);
 	return (0);
 
 nomem:
-	free((void *)(uintptr_t)zc.zc_nvlist_dst);
+	free(buf);
 	return (no_memory(zhp->zpool_hdl));
 }
 

lib/libzfs/libzfs_status.c

@@ -222,7 +222,6 @@ check_status(nvlist_t *config, boolean_t isimport,
 {
 	pool_scan_stat_t *ps = NULL;
 	uint_t vsc, psc;
-	uint64_t nerr;
 	uint64_t suspended;
 	uint64_t hostid = 0;
 	uint64_t errata = 0;
@@ -392,6 +391,7 @@ check_status(nvlist_t *config, boolean_t isimport,
 	 * Persistent data errors.
 	 */
 	if (!isimport) {
+		uint64_t nerr;
 		if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_ERRCOUNT,
 		    &nerr) == 0 && nerr != 0)
 			return (ZPOOL_STATUS_CORRUPT_DATA);

module/zfs/dsl_scan.c

@@ -944,13 +944,13 @@ dsl_scan_done(dsl_scan_t *scn, boolean_t complete, dmu_tx_t *tx)
 
 	if (dsl_scan_restarting(scn, tx))
 		spa_history_log_internal(spa, "scan aborted, restarting", tx,
-		    "errors=%llu", (u_longlong_t)spa_get_errlog_size(spa));
+		    "errors=%llu", (u_longlong_t)spa_approx_errlog_size(spa));
 	else if (!complete)
 		spa_history_log_internal(spa, "scan cancelled", tx,
-		    "errors=%llu", (u_longlong_t)spa_get_errlog_size(spa));
+		    "errors=%llu", (u_longlong_t)spa_approx_errlog_size(spa));
 	else
 		spa_history_log_internal(spa, "scan done", tx,
-		    "errors=%llu", (u_longlong_t)spa_get_errlog_size(spa));
+		    "errors=%llu", (u_longlong_t)spa_approx_errlog_size(spa));
 
 	if (DSL_SCAN_IS_SCRUB_RESILVER(scn)) {
 		spa->spa_scrub_active = B_FALSE;
@@ -1013,7 +1013,7 @@ dsl_scan_done(dsl_scan_t *scn, boolean_t complete, dmu_tx_t *tx)
 		    vdev_clear_resilver_deferred(spa->spa_root_vdev, tx)) {
 			spa_history_log_internal(spa,
 			    "starting deferred resilver", tx, "errors=%llu",
-			    (u_longlong_t)spa_get_errlog_size(spa));
+			    (u_longlong_t)spa_approx_errlog_size(spa));
 			spa_async_request(spa, SPA_ASYNC_RESILVER);
 		}
 

module/zfs/spa.c

@@ -5543,7 +5543,7 @@ spa_get_stats(const char *name, nvlist_t **config,
 
 			fnvlist_add_uint64(*config,
 			    ZPOOL_CONFIG_ERRCOUNT,
-			    spa_get_errlog_size(spa));
+			    spa_approx_errlog_size(spa));
 
 			if (spa_suspended(spa)) {
 				fnvlist_add_uint64(*config,