NAS-132432 / 24.10.1 / Cherry-pick important zfs commits from upstream to stable/electriceel for 24.10.1 freeze

cmd/arc_summary

@@ -566,7 +566,7 @@ def section_arc(kstats_dict):
     l2_hdr_size = arc_stats['l2_hdr_size']
     abd_chunk_waste_size = arc_stats['abd_chunk_waste_size']
 
-    prt_1('ARC structal breakdown (current size):', f_bytes(arc_size))
+    prt_1('ARC structural breakdown (current size):', f_bytes(arc_size))
     prt_i2('Compressed size:',
            f_perc(compressed_size, arc_size), f_bytes(compressed_size))
     prt_i2('Overhead size:',

cmd/zed/agents/zfs_diagnosis.c

@@ -71,6 +71,7 @@ typedef struct zfs_case_data {
 	uint64_t	zc_ena;
 	uint64_t	zc_pool_guid;
 	uint64_t	zc_vdev_guid;
+	uint64_t	zc_parent_guid;
 	int		zc_pool_state;
 	char		zc_serd_checksum[MAX_SERDLEN];
 	char		zc_serd_io[MAX_SERDLEN];
@@ -181,10 +182,10 @@ zfs_case_unserialize(fmd_hdl_t *hdl, fmd_case_t *cp)
 }
 
 /*
- * count other unique slow-io cases in a pool
+ * Return count of other unique SERD cases under same vdev parent
  */
 static uint_t
-zfs_other_slow_cases(fmd_hdl_t *hdl, const zfs_case_data_t *zfs_case)
+zfs_other_serd_cases(fmd_hdl_t *hdl, const zfs_case_data_t *zfs_case)
 {
 	zfs_case_t *zcp;
 	uint_t cases = 0;
@@ -206,10 +207,32 @@ zfs_other_slow_cases(fmd_hdl_t *hdl, const zfs_case_data_t *zfs_case)
 
 	for (zcp = uu_list_first(zfs_cases); zcp != NULL;
 	    zcp = uu_list_next(zfs_cases, zcp)) {
-		if (zcp->zc_data.zc_pool_guid == zfs_case->zc_pool_guid &&
-		    zcp->zc_data.zc_vdev_guid != zfs_case->zc_vdev_guid &&
-		    zcp->zc_data.zc_serd_slow_io[0] != '\0' &&
-		    fmd_serd_active(hdl, zcp->zc_data.zc_serd_slow_io)) {
+		zfs_case_data_t *zcd = &zcp->zc_data;
+
+		/*
+		 * must be same pool and parent vdev but different leaf vdev
+		 */
+		if (zcd->zc_pool_guid != zfs_case->zc_pool_guid ||
+		    zcd->zc_parent_guid != zfs_case->zc_parent_guid ||
+		    zcd->zc_vdev_guid == zfs_case->zc_vdev_guid) {
+			continue;
+		}
+
+		/*
+		 * Check if there is another active serd case besides zfs_case
+		 *
+		 * Only one serd engine will be assigned to the case
+		 */
+		if (zcd->zc_serd_checksum[0] == zfs_case->zc_serd_checksum[0] &&
+		    fmd_serd_active(hdl, zcd->zc_serd_checksum)) {
+			cases++;
+		}
+		if (zcd->zc_serd_io[0] == zfs_case->zc_serd_io[0] &&
+		    fmd_serd_active(hdl, zcd->zc_serd_io)) {
+			cases++;
+		}
+		if (zcd->zc_serd_slow_io[0] == zfs_case->zc_serd_slow_io[0] &&
+		    fmd_serd_active(hdl, zcd->zc_serd_slow_io)) {
 			cases++;
 		}
 	}
@@ -502,6 +525,34 @@ zfs_ereport_when(fmd_hdl_t *hdl, nvlist_t *nvl, er_timeval_t *when)
 	}
 }
 
+/*
+ * Record the specified event in the SERD engine and return a
+ * boolean value indicating whether or not the engine fired as
+ * the result of inserting this event.
+ *
+ * When the pool has similar active cases on other vdevs, then
+ * the fired state is disregarded and the case is retired.
+ */
+static int
+zfs_fm_serd_record(fmd_hdl_t *hdl, const char *name, fmd_event_t *ep,
+    zfs_case_t *zcp, const char *err_type)
+{
+	int fired = fmd_serd_record(hdl, name, ep);
+	int peers = 0;
+
+	if (fired && (peers = zfs_other_serd_cases(hdl, &zcp->zc_data)) > 0) {
+		fmd_hdl_debug(hdl, "pool %llu is tracking %d other %s cases "
+		    "-- skip faulting the vdev %llu",
+		    (u_longlong_t)zcp->zc_data.zc_pool_guid,
+		    peers, err_type,
+		    (u_longlong_t)zcp->zc_data.zc_vdev_guid);
+		zfs_case_retire(hdl, zcp);
+		fired = 0;
+	}
+
+	return (fired);
+}
+
 /*
  * Main fmd entry point.
  */
@@ -510,7 +561,7 @@ zfs_fm_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl, const char *class)
 {
 	zfs_case_t *zcp, *dcp;
 	int32_t pool_state;
-	uint64_t ena, pool_guid, vdev_guid;
+	uint64_t ena, pool_guid, vdev_guid, parent_guid;
 	uint64_t checksum_n, checksum_t;
 	uint64_t io_n, io_t;
 	er_timeval_t pool_load;
@@ -600,6 +651,9 @@ zfs_fm_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl, const char *class)
 	if (nvlist_lookup_uint64(nvl,
 	    FM_EREPORT_PAYLOAD_ZFS_VDEV_GUID, &vdev_guid) != 0)
 		vdev_guid = 0;
+	if (nvlist_lookup_uint64(nvl,
+	    FM_EREPORT_PAYLOAD_ZFS_PARENT_GUID, &parent_guid) != 0)
+		parent_guid = 0;
 	if (nvlist_lookup_uint64(nvl, FM_EREPORT_ENA, &ena) != 0)
 		ena = 0;
 
@@ -710,6 +764,7 @@ zfs_fm_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl, const char *class)
 		data.zc_ena = ena;
 		data.zc_pool_guid = pool_guid;
 		data.zc_vdev_guid = vdev_guid;
+		data.zc_parent_guid = parent_guid;
 		data.zc_pool_state = (int)pool_state;
 
 		fmd_buf_write(hdl, cs, CASE_DATA, &data, sizeof (data));
@@ -872,8 +927,10 @@ zfs_fm_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl, const char *class)
 				    SEC2NSEC(io_t));
 				zfs_case_serialize(zcp);
 			}
-			if (fmd_serd_record(hdl, zcp->zc_data.zc_serd_io, ep))
+			if (zfs_fm_serd_record(hdl, zcp->zc_data.zc_serd_io,
+			    ep, zcp, "io error")) {
 				checkremove = B_TRUE;
+			}
 		} else if (fmd_nvl_class_match(hdl, nvl,
 		    ZFS_MAKE_EREPORT(FM_EREPORT_ZFS_DELAY))) {
 			uint64_t slow_io_n, slow_io_t;
@@ -899,25 +956,10 @@ zfs_fm_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl, const char *class)
 			}
 			/* Pass event to SERD engine and see if this triggers */
 			if (zcp->zc_data.zc_serd_slow_io[0] != '\0' &&
-			    fmd_serd_record(hdl, zcp->zc_data.zc_serd_slow_io,
-			    ep)) {
-				/*
-				 * Ignore a slow io diagnosis when other
-				 * VDEVs in the pool show signs of being slow.
-				 */
-				if (zfs_other_slow_cases(hdl, &zcp->zc_data)) {
-					zfs_case_retire(hdl, zcp);
-					fmd_hdl_debug(hdl, "pool %llu has "
-					    "multiple slow io cases -- skip "
-					    "degrading vdev %llu",
-					    (u_longlong_t)
-					    zcp->zc_data.zc_pool_guid,
-					    (u_longlong_t)
-					    zcp->zc_data.zc_vdev_guid);
-				} else {
-					zfs_case_solve(hdl, zcp,
-					    "fault.fs.zfs.vdev.slow_io");
-				}
+			    zfs_fm_serd_record(hdl,
+			    zcp->zc_data.zc_serd_slow_io, ep, zcp, "slow io")) {
+				zfs_case_solve(hdl, zcp,
+				    "fault.fs.zfs.vdev.slow_io");
 			}
 		} else if (fmd_nvl_class_match(hdl, nvl,
 		    ZFS_MAKE_EREPORT(FM_EREPORT_ZFS_CHECKSUM))) {
@@ -968,8 +1010,9 @@ zfs_fm_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl, const char *class)
 				    SEC2NSEC(checksum_t));
 				zfs_case_serialize(zcp);
 			}
-			if (fmd_serd_record(hdl,
-			    zcp->zc_data.zc_serd_checksum, ep)) {
+			if (zfs_fm_serd_record(hdl,
+			    zcp->zc_data.zc_serd_checksum, ep, zcp,
+			    "checksum")) {
 				zfs_case_solve(hdl, zcp,
 				    "fault.fs.zfs.vdev.checksum");
 			}

cmd/zed/zed_disk_event.c

@@ -139,7 +139,8 @@ dev_event_nvlist(struct udev_device *dev)
 		 * is /dev/sda.
 		 */
 		struct udev_device *parent_dev = udev_device_get_parent(dev);
-		if ((value = udev_device_get_sysattr_value(parent_dev, "size"))
+		if (parent_dev != NULL &&
+		    (value = udev_device_get_sysattr_value(parent_dev, "size"))
 		    != NULL) {
 			uint64_t numval = DEV_BSIZE;
 

cmd/zfs/zfs_main.c

@@ -2162,6 +2162,7 @@ zfs_do_get(int argc, char **argv)
 	cb.cb_type = ZFS_TYPE_DATASET;
 
 	struct option long_options[] = {
+		{"json", no_argument, NULL, 'j'},
 		{"json-int", no_argument, NULL, ZFS_OPTION_JSON_NUMS_AS_INT},
 		{0, 0, 0, 0}
 	};
@@ -3760,8 +3761,13 @@ collect_dataset(zfs_handle_t *zhp, list_cbdata_t *cb)
 		if (cb->cb_json) {
 			if (pl->pl_prop == ZFS_PROP_NAME)
 				continue;
+			const char *prop_name;
+			if (pl->pl_prop != ZPROP_USERPROP)
+				prop_name = zfs_prop_to_name(pl->pl_prop);
+			else
+				prop_name = pl->pl_user_prop;
 			if (zprop_nvlist_one_property(
-			    zfs_prop_to_name(pl->pl_prop), propstr,
+			    prop_name, propstr,
 			    sourcetype, source, NULL, props,
 			    cb->cb_json_as_int) != 0)
 				nomem();
@@ -3852,6 +3858,7 @@ zfs_do_list(int argc, char **argv)
 	nvlist_t *data = NULL;
 
 	struct option long_options[] = {
+		{"json", no_argument, NULL, 'j'},
 		{"json-int", no_argument, NULL, ZFS_OPTION_JSON_NUMS_AS_INT},
 		{0, 0, 0, 0}
 	};
@@ -7436,9 +7443,15 @@ share_mount(int op, int argc, char **argv)
 	uint_t nthr;
 	jsobj = data = item = NULL;
 
+	struct option long_options[] = {
+		{"json", no_argument, NULL, 'j'},
+		{0, 0, 0, 0}
+	};
+
 	/* check options */
-	while ((c = getopt(argc, argv, op == OP_MOUNT ? ":ajRlvo:Of" : "al"))
-	    != -1) {
+	while ((c = getopt_long(argc, argv,
+	    op == OP_MOUNT ? ":ajRlvo:Of" : "al",
+	    op == OP_MOUNT ? long_options : NULL, NULL)) != -1) {
 		switch (c) {
 		case 'a':
 			do_all = 1;
@@ -8374,8 +8387,14 @@ zfs_do_channel_program(int argc, char **argv)
 	boolean_t sync_flag = B_TRUE, json_output = B_FALSE;
 	zpool_handle_t *zhp;
 
+	struct option long_options[] = {
+		{"json", no_argument, NULL, 'j'},
+		{0, 0, 0, 0}
+	};
+
 	/* check options */
-	while ((c = getopt(argc, argv, "nt:m:j")) != -1) {
+	while ((c = getopt_long(argc, argv, "nt:m:j", long_options,
+	    NULL)) != -1) {
 		switch (c) {
 		case 't':
 		case 'm': {
@@ -9083,7 +9102,13 @@ zfs_do_version(int argc, char **argv)
 	int c;
 	nvlist_t *jsobj = NULL, *zfs_ver = NULL;
 	boolean_t json = B_FALSE;
-	while ((c = getopt(argc, argv, "j")) != -1) {
+
+	struct option long_options[] = {
+		{"json", no_argument, NULL, 'j'},
+		{0, 0, 0, 0}
+	};
+
+	while ((c = getopt_long(argc, argv, "j", long_options, NULL)) != -1) {
 		switch (c) {
 		case 'j':
 			json = B_TRUE;

cmd/zpool/zpool_main.c

@@ -6870,8 +6870,13 @@ collect_pool(zpool_handle_t *zhp, list_cbdata_t *cb)
 		if (cb->cb_json) {
 			if (pl->pl_prop == ZPOOL_PROP_NAME)
 				continue;
+			const char *prop_name;
+			if (pl->pl_prop != ZPROP_USERPROP)
+				prop_name = zpool_prop_to_name(pl->pl_prop);
+			else
+				prop_name = pl->pl_user_prop;
 			(void) zprop_nvlist_one_property(
-			    zpool_prop_to_name(pl->pl_prop), propstr,
+			    prop_name, propstr,
 			    sourcetype, NULL, NULL, props, cb->cb_json_as_int);
 		} else {
 			/*
@@ -7328,6 +7333,7 @@ zpool_do_list(int argc, char **argv)
 	current_prop_type = ZFS_TYPE_POOL;
 
 	struct option long_options[] = {
+		{"json", no_argument, NULL, 'j'},
 		{"json-int", no_argument, NULL, ZPOOL_OPTION_JSON_NUMS_AS_INT},
 		{"json-pool-key-guid", no_argument, NULL,
 		    ZPOOL_OPTION_POOL_KEY_GUID},
@@ -7953,8 +7959,11 @@ zpool_do_online(int argc, char **argv)
 
 	poolname = argv[0];
 
-	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
+	if ((zhp = zpool_open(g_zfs, poolname)) == NULL) {
+		(void) fprintf(stderr, gettext("failed to open pool "
+		    "\"%s\""), poolname);
 		return (1);
+	}
 
 	for (i = 1; i < argc; i++) {
 		vdev_state_t oldstate;
@@ -7975,12 +7984,15 @@ zpool_do_online(int argc, char **argv)
 		    &l2cache, NULL);
 		if (tgt == NULL) {
 			ret = 1;
+			(void) fprintf(stderr, gettext("couldn't find device "
+			"\"%s\" in pool \"%s\"\n"), argv[i], poolname);
 			continue;
 		}
 		uint_t vsc;
 		oldstate = ((vdev_stat_t *)fnvlist_lookup_uint64_array(tgt,
 		    ZPOOL_CONFIG_VDEV_STATS, &vsc))->vs_state;
-		if (zpool_vdev_online(zhp, argv[i], flags, &newstate) == 0) {
+		if ((rc = zpool_vdev_online(zhp, argv[i], flags,
+		    &newstate)) == 0) {
 			if (newstate != VDEV_STATE_HEALTHY) {
 				(void) printf(gettext("warning: device '%s' "
 				    "onlined, but remains in faulted state\n"),
@@ -8006,6 +8018,9 @@ zpool_do_online(int argc, char **argv)
 				}
 			}
 		} else {
+			(void) fprintf(stderr, gettext("Failed to online "
+			    "\"%s\" in pool \"%s\": %d\n"),
+			    argv[i], poolname, rc);
 			ret = 1;
 		}
 	}
@@ -8090,8 +8105,11 @@ zpool_do_offline(int argc, char **argv)
 
 	poolname = argv[0];
 
-	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
+	if ((zhp = zpool_open(g_zfs, poolname)) == NULL) {
+		(void) fprintf(stderr, gettext("failed to open pool "
+		    "\"%s\""), poolname);
 		return (1);
+	}
 
 	for (i = 1; i < argc; i++) {
 		uint64_t guid = zpool_vdev_path_to_guid(zhp, argv[i]);
@@ -10958,6 +10976,7 @@ zpool_do_status(int argc, char **argv)
 
 	struct option long_options[] = {
 		{"power", no_argument, NULL, ZPOOL_OPTION_POWER},
+		{"json", no_argument, NULL, 'j'},
 		{"json-int", no_argument, NULL, ZPOOL_OPTION_JSON_NUMS_AS_INT},
 		{"json-flat-vdevs", no_argument, NULL,
 		    ZPOOL_OPTION_JSON_FLAT_VDEVS},
@@ -12563,6 +12582,7 @@ zpool_do_get(int argc, char **argv)
 	current_prop_type = cb.cb_type;
 
 	struct option long_options[] = {
+		{"json", no_argument, NULL, 'j'},
 		{"json-int", no_argument, NULL, ZPOOL_OPTION_JSON_NUMS_AS_INT},
 		{"json-pool-key-guid", no_argument, NULL,
 		    ZPOOL_OPTION_POOL_KEY_GUID},
@@ -13477,7 +13497,12 @@ zpool_do_version(int argc, char **argv)
 	int c;
 	nvlist_t *jsobj = NULL, *zfs_ver = NULL;
 	boolean_t json = B_FALSE;
-	while ((c = getopt(argc, argv, "j")) != -1) {
+
+	struct option long_options[] = {
+		{"json", no_argument, NULL, 'j'},
+	};
+
+	while ((c = getopt_long(argc, argv, "j", long_options, NULL)) != -1) {
 		switch (c) {
 		case 'j':
 			json = B_TRUE;

cmd/zstream/zstream_recompress.c

@@ -288,7 +288,8 @@ zstream_do_recompress(int argc, char *argv[])
 				abd_t *pabd =
 				    abd_get_from_buf_struct(&abd, buf, bufsz);
 				size_t csize = zio_compress_data(ctype, &dabd,
-				    &pabd, drrw->drr_logical_size, level);
+				    &pabd, drrw->drr_logical_size,
+				    drrw->drr_logical_size, level);
 				size_t rounded =
 				    P2ROUNDUP(csize, SPA_MINBLOCKSIZE);
 				if (rounded >= drrw->drr_logical_size) {