PSARC 2008/388 Short circuit for vdev probe failure
authoreschrock
Fri, 27 Jun 2008 17:05:29 -0700
changeset 6976 cae5f06df471
parent 6975 c511f317869e
child 6977 2827369c5456
PSARC 2008/388 Short circuit for vdev probe failure 6410382 ZFS rename should update ->v_path 6633592 vdev_root_open() faulted logic doesn't work 6674674 zfs needs to handle more than EIO 6705158 failed log device should not be treated like normal toplevel vdev 6705232 vdev_not_present erroneously cleared in vdev_validate() 6705649 libses/libscsi plugins should be included in ABI exception list 6707536 zfs_case_timeout and associated code is unused 6707637 vdev_propagate_state() shouldn't skip log devices 6708095 better FMA integration for vdev probe failure 6708192 typo in fix for 6646106 breaks pool I/O failure handling 6709112 zio_should_retry() should account for vdev_is_failing 6718542 zinject doesn't work for file I/O probe failure 6719105 ses2: incorrect parsing of page 15 (nickname)
usr/src/cmd/fm/modules/common/zfs-diagnosis/zfs_de.c
usr/src/lib/scsi/plugins/ses/ses2/common/ses2_enclosure.c
usr/src/tools/abi/etc/exceptions
usr/src/uts/common/fs/nfs/nfs3_srv.c
usr/src/uts/common/fs/nfs/nfs4_srv.c
usr/src/uts/common/fs/nfs/nfs_srv.c
usr/src/uts/common/fs/vnode.c
usr/src/uts/common/fs/zfs/vdev.c
usr/src/uts/common/fs/zfs/vdev_disk.c
usr/src/uts/common/fs/zfs/vdev_file.c
usr/src/uts/common/fs/zfs/vdev_label.c
usr/src/uts/common/fs/zfs/vdev_root.c
usr/src/uts/common/fs/zfs/zfs_fm.c
usr/src/uts/common/fs/zfs/zfs_vnops.c
usr/src/uts/common/fs/zfs/zio.c
usr/src/uts/common/sys/fm/fs/zfs.h
usr/src/uts/common/sys/vnode.h
--- a/usr/src/cmd/fm/modules/common/zfs-diagnosis/zfs_de.c	Fri Jun 27 14:18:08 2008 -0700
+++ b/usr/src/cmd/fm/modules/common/zfs-diagnosis/zfs_de.c	Fri Jun 27 17:05:29 2008 -0700
@@ -42,38 +42,49 @@
  */
 #define	MAX_SERDLEN	(16 * 2 + sizeof ("zfs___checksum"))
 
+/*
+ * On-disk case structure.  This must maintain backwards compatibility with
+ * previous versions of the DE.  By default, any members appended to the end
+ * will be filled with zeros if they don't exist in a previous version.
+ */
 typedef struct zfs_case_data {
 	uint64_t	zc_version;
 	uint64_t	zc_ena;
 	uint64_t	zc_pool_guid;
 	uint64_t	zc_vdev_guid;
-	int		zc_has_timer;
+	int		zc_has_timer;		/* defunct */
 	int		zc_pool_state;
 	char		zc_serd_checksum[MAX_SERDLEN];
 	char		zc_serd_io[MAX_SERDLEN];
-	int		zc_has_serd_timer;
+	int		zc_has_remove_timer;
 } zfs_case_data_t;
 
+/*
+ * In-core case structure.
+ */
 typedef struct zfs_case {
 	boolean_t	zc_present;
 	uint32_t	zc_version;
 	zfs_case_data_t	zc_data;
 	fmd_case_t	*zc_case;
 	uu_list_node_t	zc_node;
-	id_t		zc_timer;
-	id_t		zc_serd_timer;
+	id_t		zc_remove_timer;
 } zfs_case_t;
 
 #define	CASE_DATA			"data"
 #define	CASE_DATA_VERSION_INITIAL	1
 #define	CASE_DATA_VERSION_SERD		2
 
-static hrtime_t zfs_case_timeout;
-static hrtime_t zfs_serd_timeout;
+static hrtime_t zfs_remove_timeout;
 
 uu_list_pool_t *zfs_case_pool;
 uu_list_t *zfs_cases;
 
+#define	ZFS_MAKE_RSRC(type)	\
+    FM_RSRC_CLASS "." ZFS_ERROR_CLASS "." type
+#define	ZFS_MAKE_EREPORT(type)	\
+    FM_EREPORT_CLASS "." ZFS_ERROR_CLASS "." type
+
 /*
  * Write out the persistent representation of an active case.
  */
@@ -114,12 +125,9 @@
 	 * doesn't include the SERD engine name.
 	 */
 
-	if (zcp->zc_data.zc_has_timer)
-		zcp->zc_timer = fmd_timer_install(hdl, zcp,
-		    NULL, zfs_case_timeout);
-	if (zcp->zc_data.zc_has_serd_timer)
-		zcp->zc_serd_timer = fmd_timer_install(hdl, zcp,
-		    NULL, zfs_serd_timeout);
+	if (zcp->zc_data.zc_has_remove_timer)
+		zcp->zc_remove_timer = fmd_timer_install(hdl, zcp,
+		    NULL, zfs_remove_timeout);
 
 	(void) uu_list_insert_before(zfs_cases, NULL, zcp);
 
@@ -310,14 +318,9 @@
 	fmd_case_solve(hdl, zcp->zc_case);
 
 	serialize = B_FALSE;
-	if (zcp->zc_data.zc_has_timer) {
-		fmd_timer_remove(hdl, zcp->zc_timer);
-		zcp->zc_data.zc_has_timer = 0;
-		serialize = B_TRUE;
-	}
-	if (zcp->zc_data.zc_has_serd_timer) {
-		fmd_timer_remove(hdl, zcp->zc_serd_timer);
-		zcp->zc_data.zc_has_serd_timer = 0;
+	if (zcp->zc_data.zc_has_remove_timer) {
+		fmd_timer_remove(hdl, zcp->zc_remove_timer);
+		zcp->zc_data.zc_has_remove_timer = 0;
 		serialize = B_TRUE;
 	}
 	if (serialize)
@@ -338,7 +341,7 @@
 	uint64_t ena, pool_guid, vdev_guid;
 	nvlist_t *detector;
 	boolean_t isresource;
-	const char *serd;
+	boolean_t checkremove;
 
 	isresource = fmd_nvl_class_match(hdl, nvl, "resource.fs.zfs.*");
 
@@ -459,7 +462,7 @@
 
 	if (isresource) {
 		if (fmd_nvl_class_match(hdl, nvl,
-		    "resource.fs.zfs.autoreplace")) {
+		    ZFS_MAKE_RSRC(FM_RESOURCE_AUTOREPLACE))) {
 			/*
 			 * The 'resource.fs.zfs.autoreplace' event indicates
 			 * that the pool was loaded with the 'autoreplace'
@@ -469,7 +472,7 @@
 			 */
 			fmd_case_close(hdl, zcp->zc_case);
 		} else if (fmd_nvl_class_match(hdl, nvl,
-		    "resource.fs.zfs.removed")) {
+		    ZFS_MAKE_RSRC(FM_RESOURCE_REMOVED))) {
 			/*
 			 * The 'resource.fs.zfs.removed' event indicates that
 			 * device removal was detected, and the device was
@@ -479,9 +482,9 @@
 			 * We reset the SERD engine, and cancel any pending
 			 * timers.
 			 */
-			if (zcp->zc_data.zc_has_serd_timer) {
-				fmd_timer_remove(hdl, zcp->zc_serd_timer);
-				zcp->zc_data.zc_has_serd_timer = 0;
+			if (zcp->zc_data.zc_has_remove_timer) {
+				fmd_timer_remove(hdl, zcp->zc_remove_timer);
+				zcp->zc_data.zc_has_remove_timer = 0;
 				zfs_case_serialize(hdl, zcp);
 			}
 			if (zcp->zc_data.zc_serd_io[0] != '\0')
@@ -519,7 +522,8 @@
 	 * succeeded, we associate a timer with the event.  When it expires, we
 	 * close the case.
 	 */
-	if (fmd_nvl_class_match(hdl, nvl, "ereport.fs.zfs.zpool")) {
+	if (fmd_nvl_class_match(hdl, nvl,
+	    ZFS_MAKE_EREPORT(FM_EREPORT_ZFS_POOL))) {
 		/*
 		 * Pool level fault.  Before solving the case, go through and
 		 * close any open device cases that may be pending.
@@ -542,9 +546,14 @@
 		 * within the timeout period, then we'll solve the device case.
 		 */
 		zfs_case_solve(hdl, zcp, "fault.fs.zfs.device",  B_TRUE);
-	} else if (fmd_nvl_class_match(hdl, nvl, "ereport.fs.zfs.io") ||
-	    fmd_nvl_class_match(hdl, nvl, "ereport.fs.zfs.checksum") ||
-	    fmd_nvl_class_match(hdl, nvl, "ererpot.fs.zfs.io_failure")) {
+	} else if (fmd_nvl_class_match(hdl, nvl,
+	    ZFS_MAKE_EREPORT(FM_EREPORT_ZFS_IO)) ||
+	    fmd_nvl_class_match(hdl, nvl,
+	    ZFS_MAKE_EREPORT(FM_EREPORT_ZFS_CHECKSUM)) ||
+	    fmd_nvl_class_match(hdl, nvl,
+	    ZFS_MAKE_EREPORT(FM_EREPORT_ZFS_IO_FAILURE)) ||
+	    fmd_nvl_class_match(hdl, nvl,
+	    ZFS_MAKE_EREPORT(FM_EREPORT_ZFS_PROBE_FAILURE))) {
 		char *failmode = NULL;
 
 		/*
@@ -554,8 +563,8 @@
 		 * (persistent errors for a single data block, etc).  For now,
 		 * a single SERD engine is sufficient.
 		 */
-		serd = NULL;
-		if (fmd_nvl_class_match(hdl, nvl, "ereport.fs.zfs.io")) {
+		if (fmd_nvl_class_match(hdl, nvl,
+		    ZFS_MAKE_EREPORT(FM_EREPORT_ZFS_IO))) {
 			if (zcp->zc_data.zc_serd_io[0] == '\0') {
 				zfs_serd_name(zcp->zc_data.zc_serd_io,
 				    pool_guid, vdev_guid, "io");
@@ -564,9 +573,10 @@
 				    fmd_prop_get_int64(hdl, "io_T"));
 				zfs_case_serialize(hdl, zcp);
 			}
-			serd = zcp->zc_data.zc_serd_io;
+			if (fmd_serd_record(hdl, zcp->zc_data.zc_serd_io, ep))
+				checkremove = B_TRUE;
 		} else if (fmd_nvl_class_match(hdl, nvl,
-		    "ereport.fs.zfs.checksum")) {
+		    ZFS_MAKE_EREPORT(FM_EREPORT_ZFS_CHECKSUM))) {
 			if (zcp->zc_data.zc_serd_checksum[0] == '\0') {
 				zfs_serd_name(zcp->zc_data.zc_serd_checksum,
 				    pool_guid, vdev_guid, "checksum");
@@ -576,9 +586,14 @@
 				    fmd_prop_get_int64(hdl, "checksum_T"));
 				zfs_case_serialize(hdl, zcp);
 			}
-			serd = zcp->zc_data.zc_serd_checksum;
+			if (fmd_serd_record(hdl,
+			    zcp->zc_data.zc_serd_checksum, ep)) {
+				zfs_case_solve(hdl, zcp,
+				    "fault.fs.zfs.vdev.checksum", B_FALSE);
+			}
 		} else if (fmd_nvl_class_match(hdl, nvl,
-		    "ereport.fs.zfs.io_failure") && (nvlist_lookup_string(nvl,
+		    ZFS_MAKE_EREPORT(FM_EREPORT_ZFS_IO_FAILURE)) &&
+		    (nvlist_lookup_string(nvl,
 		    FM_EREPORT_PAYLOAD_ZFS_POOL_FAILMODE, &failmode) == 0) &&
 		    failmode != NULL) {
 			if (strncmp(failmode, FM_EREPORT_FAILMODE_CONTINUE,
@@ -591,6 +606,9 @@
 				zfs_case_solve(hdl, zcp,
 				    "fault.fs.zfs.io_failure_wait", B_FALSE);
 			}
+		} else if (fmd_nvl_class_match(hdl, nvl,
+		    ZFS_MAKE_EREPORT(FM_EREPORT_ZFS_PROBE_FAILURE))) {
+			checkremove = B_TRUE;
 		}
 
 		/*
@@ -598,13 +616,13 @@
 		 * any diagnosis until we're sure that we aren't about to
 		 * receive a 'resource.fs.zfs.removed' event.
 		 */
-		if (serd && fmd_serd_record(hdl, serd, ep)) {
-			if (zcp->zc_data.zc_has_serd_timer)
-				fmd_timer_remove(hdl, zcp->zc_serd_timer);
-			zcp->zc_serd_timer = fmd_timer_install(hdl, zcp, NULL,
-			    zfs_serd_timeout);
-			if (!zcp->zc_data.zc_has_serd_timer) {
-				zcp->zc_data.zc_has_serd_timer = 1;
+		if (checkremove) {
+			if (zcp->zc_data.zc_has_remove_timer)
+				fmd_timer_remove(hdl, zcp->zc_remove_timer);
+			zcp->zc_remove_timer = fmd_timer_install(hdl, zcp, NULL,
+			    zfs_remove_timeout);
+			if (!zcp->zc_data.zc_has_remove_timer) {
+				zcp->zc_data.zc_has_remove_timer = 1;
 				zfs_case_serialize(hdl, zcp);
 			}
 		}
@@ -612,37 +630,17 @@
 }
 
 /*
- * Timeout indicates one of two scenarios:
- *
- * 	- A device could not be opened while opening a pool, but the pool
- *	  itself was opened successfully.
- *
- * 	- We diagnosed an I/O error, and it was not due to device removal (which
- *	  would cause the timeout to be cancelled).
+ * The timeout is fired when we diagnosed an I/O error, and it was not due to
+ * device removal (which would cause the timeout to be cancelled).
  */
 /* ARGSUSED */
 static void
 zfs_fm_timeout(fmd_hdl_t *hdl, id_t id, void *data)
 {
 	zfs_case_t *zcp = data;
-	const char *faultname;
 
-	if (id == zcp->zc_timer) {
-		zcp->zc_data.zc_has_timer = 0;
-		zfs_case_solve(hdl, zcp, "fault.fs.zfs.device", B_TRUE);
-	}
-
-	if (id == zcp->zc_serd_timer) {
-		if (zcp->zc_data.zc_serd_io[0] != '\0' &&
-		    fmd_serd_fired(hdl, zcp->zc_data.zc_serd_io)) {
-			faultname = "fault.fs.zfs.vdev.io";
-		} else {
-			assert(fmd_serd_fired(hdl,
-			    zcp->zc_data.zc_serd_checksum));
-			faultname = "fault.fs.zfs.vdev.checksum";
-		}
-		zfs_case_solve(hdl, zcp, faultname, B_FALSE);
-	}
+	if (id == zcp->zc_remove_timer)
+		zfs_case_solve(hdl, zcp, "fault.fs.zfs.vdev.io", B_FALSE);
 }
 
 static void
@@ -654,10 +652,8 @@
 		fmd_serd_destroy(hdl, zcp->zc_data.zc_serd_checksum);
 	if (zcp->zc_data.zc_serd_io[0] != '\0')
 		fmd_serd_destroy(hdl, zcp->zc_data.zc_serd_io);
-	if (zcp->zc_data.zc_has_timer)
-		fmd_timer_remove(hdl, zcp->zc_timer);
-	if (zcp->zc_data.zc_has_serd_timer)
-		fmd_timer_remove(hdl, zcp->zc_serd_timer);
+	if (zcp->zc_data.zc_has_remove_timer)
+		fmd_timer_remove(hdl, zcp->zc_remove_timer);
 	uu_list_remove(zfs_cases, zcp);
 	fmd_hdl_free(hdl, zcp, sizeof (zfs_case_t));
 }
@@ -686,7 +682,7 @@
 	{ "checksum_T", FMD_TYPE_TIME, "10min" },
 	{ "io_N", FMD_TYPE_UINT32, "10" },
 	{ "io_T", FMD_TYPE_TIME, "10min" },
-	{ "serd_timeout", FMD_TYPE_TIME, "5sec" },
+	{ "remove_timeout", FMD_TYPE_TIME, "5sec" },
 	{ NULL, 0, NULL }
 };
 
@@ -738,8 +734,7 @@
 	 */
 	zfs_purge_cases(hdl);
 
-	zfs_case_timeout = fmd_prop_get_int64(hdl, "case_timeout");
-	zfs_serd_timeout = fmd_prop_get_int64(hdl, "serd_timeout");
+	zfs_remove_timeout = fmd_prop_get_int64(hdl, "remove_timeout");
 }
 
 void
--- a/usr/src/lib/scsi/plugins/ses/ses2/common/ses2_enclosure.c	Fri Jun 27 14:18:08 2008 -0700
+++ b/usr/src/lib/scsi/plugins/ses/ses2/common/ses2_enclosure.c	Fri Jun 27 17:05:29 2008 -0700
@@ -273,7 +273,7 @@
 		return (0);
 
 	for (dip = &spip->sspci_subnicks[0], i = 0;
-	    i < spip->sspci_n_subenclosures;
+	    i <= spip->sspci_n_subenclosures;
 	    i++, dip++) {
 		if (!SES_WITHIN_PAGE_STRUCT(dip, spip, len))
 			break;
--- a/usr/src/tools/abi/etc/exceptions	Fri Jun 27 14:18:08 2008 -0700
+++ b/usr/src/tools/abi/etc/exceptions	Fri Jun 27 17:05:29 2008 -0700
@@ -239,6 +239,7 @@
 4773624: RULE W1: usr/lib/locale/iso_8859_1/
 4773624: RULE W1: usr/lib/fm/topo/
 6233613: RULE W1: usr/platform/sun4v/lib/mdb/
+6705649: RULE W1: usr/lib/scsi/plugins/
 
 #############################################
 # WARNINGs exempted from RULE W2 (See RULES section of intf_check manpage) 
--- a/usr/src/uts/common/fs/nfs/nfs3_srv.c	Fri Jun 27 14:18:08 2008 -0700
+++ b/usr/src/uts/common/fs/nfs/nfs3_srv.c	Fri Jun 27 17:05:29 2008 -0700
@@ -2819,19 +2819,9 @@
 		}
 		nbl_end_crit(srcvp);
 	}
-	if (error == 0) {
-		char *tmp;
-
-		/* fix the path name for the renamed file */
-		mutex_enter(&srcvp->v_lock);
-		tmp = srcvp->v_path;
-		srcvp->v_path = NULL;
-		mutex_exit(&srcvp->v_lock);
-		vn_setpath(rootdir, tvp, srcvp, args->to.name,
+	if (error == 0)
+		vn_renamepath(tvp, srcvp, args->to.name,
 		    strlen(args->to.name));
-		if (tmp != NULL)
-			kmem_free(tmp, strlen(tmp) + 1);
-	}
 	VN_RELE(srcvp);
 	srcvp = NULL;
 
--- a/usr/src/uts/common/fs/nfs/nfs4_srv.c	Fri Jun 27 14:18:08 2008 -0700
+++ b/usr/src/uts/common/fs/nfs/nfs4_srv.c	Fri Jun 27 17:05:29 2008 -0700
@@ -4325,18 +4325,8 @@
 			VN_RELE(tvp);
 		}
 	}
-	if (error == 0) {
-		char *tmp;
-
-		/* fix the path name for the renamed file */
-		mutex_enter(&srcvp->v_lock);
-		tmp = srcvp->v_path;
-		srcvp->v_path = NULL;
-		mutex_exit(&srcvp->v_lock);
-		vn_setpath(rootdir, ndvp, srcvp, nnm, nlen - 1);
-		if (tmp != NULL)
-			kmem_free(tmp, strlen(tmp) + 1);
-	}
+	if (error == 0)
+		vn_renamepath(ndvp, srcvp, nnm, nlen - 1);
 
 	if (in_crit_src)
 		nbl_end_crit(srcvp);
--- a/usr/src/uts/common/fs/nfs/nfs_srv.c	Fri Jun 27 14:18:08 2008 -0700
+++ b/usr/src/uts/common/fs/nfs/nfs_srv.c	Fri Jun 27 17:05:29 2008 -0700
@@ -2236,19 +2236,9 @@
 	    tovp, args->rna_to.da_name, cr, NULL, 0);
 	TRACE_0(TR_FAC_NFS, TR_VOP_RENAME_END, "vop_rename_end:");
 
-	if (error == 0) {
-		char *tmp;
-
-		/* fix the path name for the renamed file */
-		mutex_enter(&srcvp->v_lock);
-		tmp = srcvp->v_path;
-		srcvp->v_path = NULL;
-		mutex_exit(&srcvp->v_lock);
-		vn_setpath(rootdir, tovp, srcvp, args->rna_to.da_name,
+	if (error == 0)
+		vn_renamepath(tovp, srcvp, args->rna_to.da_name,
 		    strlen(args->rna_to.da_name));
-		if (tmp != NULL)
-			kmem_free(tmp, strlen(tmp) + 1);
-	}
 
 	/*
 	 * Force modified data and metadata out to stable storage.
--- a/usr/src/uts/common/fs/vnode.c	Fri Jun 27 14:18:08 2008 -0700
+++ b/usr/src/uts/common/fs/vnode.c	Fri Jun 27 17:05:29 2008 -0700
@@ -2939,6 +2939,24 @@
 }
 
 /*
+ * Called from within filesystem's vop_rename() to handle renames once the
+ * target vnode is available.
+ */
+void
+vn_renamepath(vnode_t *dvp, vnode_t *vp, const char *nm, size_t len)
+{
+	char *tmp;
+
+	mutex_enter(&vp->v_lock);
+	tmp = vp->v_path;
+	vp->v_path = NULL;
+	mutex_exit(&vp->v_lock);
+	vn_setpath(rootdir, dvp, vp, nm, len);
+	if (tmp != NULL)
+		kmem_free(tmp, strlen(tmp) + 1);
+}
+
+/*
  * Similar to vn_setpath_str(), this function sets the path of the destination
  * vnode to the be the same as the source vnode.
  */
--- a/usr/src/uts/common/fs/zfs/vdev.c	Fri Jun 27 14:18:08 2008 -0700
+++ b/usr/src/uts/common/fs/zfs/vdev.c	Fri Jun 27 17:05:29 2008 -0700
@@ -1038,15 +1038,16 @@
 		if (spa->spa_load_state == SPA_LOAD_OPEN &&
 		    state != POOL_STATE_ACTIVE)
 			return (EBADF);
+
+		/*
+		 * If we were able to open and validate a vdev that was
+		 * previously marked permanently unavailable, clear that state
+		 * now.
+		 */
+		if (vd->vdev_not_present)
+			vd->vdev_not_present = 0;
 	}
 
-	/*
-	 * If we were able to open and validate a vdev that was previously
-	 * marked permanently unavailable, clear that state now.
-	 */
-	if (vd->vdev_not_present)
-		vd->vdev_not_present = 0;
-
 	return (0);
 }
 
@@ -2102,12 +2103,21 @@
 	if (vd->vdev_children > 0) {
 		for (c = 0; c < vd->vdev_children; c++) {
 			child = vd->vdev_child[c];
-			if (vdev_is_dead(child) && !vdev_readable(child))
-				faulted++;
-			else if (child->vdev_stat.vs_aux == VDEV_AUX_IO_FAILURE)
-				faulted++;
-			else if (child->vdev_state <= VDEV_STATE_DEGRADED)
+
+			if ((vdev_is_dead(child) && !vdev_readable(child)) ||
+			    child->vdev_stat.vs_aux == VDEV_AUX_IO_FAILURE) {
+				/*
+				 * Root special: if there is a top-level log
+				 * device, treat the root vdev as if it were
+				 * degraded.
+				 */
+				if (child->vdev_islog && vd == rvd)
+					degraded++;
+				else
+					faulted++;
+			} else if (child->vdev_state <= VDEV_STATE_DEGRADED) {
 				degraded++;
+			}
 
 			if (child->vdev_stat.vs_aux == VDEV_AUX_CORRUPT_DATA)
 				corrupted++;
@@ -2127,7 +2137,7 @@
 			    VDEV_AUX_CORRUPT_DATA);
 	}
 
-	if (vd->vdev_parent && !vd->vdev_islog)
+	if (vd->vdev_parent)
 		vdev_propagate_state(vd->vdev_parent);
 }
 
--- a/usr/src/uts/common/fs/zfs/vdev_disk.c	Fri Jun 27 14:18:08 2008 -0700
+++ b/usr/src/uts/common/fs/zfs/vdev_disk.c	Fri Jun 27 17:05:29 2008 -0700
@@ -33,6 +33,7 @@
 #include <sys/fs/zfs.h>
 #include <sys/zio.h>
 #include <sys/sunldi.h>
+#include <sys/fm/fs/zfs.h>
 
 /*
  * Virtual device vector for disks.
@@ -405,7 +406,14 @@
 	vdev_disk_buf_t *vdb = (vdev_disk_buf_t *)bp;
 	zio_t *zio = vdb->vdb_io;
 
-	if ((zio->io_error = geterror(bp)) == 0 && bp->b_resid != 0)
+	/*
+	 * The rest of the zio stack only deals with EIO, ECKSUM, and ENXIO.
+	 * Rather than teach the rest of the stack about other error
+	 * possibilities (EFAULT, etc), we normalize the error value here.
+	 */
+	zio->io_error = (geterror(bp) != 0 ? EIO : 0);
+
+	if (zio->io_error == 0 && bp->b_resid != 0)
 		zio->io_error = EIO;
 
 	kmem_free(vdb, sizeof (vdev_disk_buf_t));
@@ -564,7 +572,11 @@
 			spa_async_request(zio->io_spa, SPA_ASYNC_REMOVE);
 		} else if (vdev_probe(vd) != 0) {
 			ASSERT(vd->vdev_ops->vdev_op_leaf);
-			vd->vdev_is_failing = B_TRUE;
+			if (!vd->vdev_is_failing) {
+				vd->vdev_is_failing = B_TRUE;
+				zfs_ereport_post(FM_EREPORT_ZFS_PROBE_FAILURE,
+				    vd->vdev_spa, vd, zio, 0, 0);
+			}
 		}
 	}
 
--- a/usr/src/uts/common/fs/zfs/vdev_file.c	Fri Jun 27 14:18:08 2008 -0700
+++ b/usr/src/uts/common/fs/zfs/vdev_file.c	Fri Jun 27 17:05:29 2008 -0700
@@ -31,6 +31,7 @@
 #include <sys/vdev_impl.h>
 #include <sys/zio.h>
 #include <sys/fs/zfs.h>
+#include <sys/fm/fs/zfs.h>
 
 /*
  * Virtual device vector for files.
@@ -144,9 +145,14 @@
 
 	error = vn_rdwr(rw, vf->vf_vnode, data, size, offset, UIO_SYSSPACE,
 	    0, RLIM64_INFINITY, kcred, &resid);
+
 	if (error || resid != 0)
 		return (EIO);
-	return (0);
+
+	if (zio_injection_enabled)
+		error = zio_handle_device_injection(vd, EIO);
+
+	return (error);
 }
 
 /*
@@ -189,10 +195,10 @@
 		nvd = kmem_zalloc(sizeof (vdev_t), KM_SLEEP);
 		if (vd->vdev_path)
 			nvd->vdev_path = spa_strdup(vd->vdev_path);
+		nvd->vdev_guid = vd->vdev_guid;
 		retries++;
 
-		error = vdev_file_open_common(nvd);
-		if (error)
+		if (vdev_file_open_common(nvd) != 0)
 			break;
 	}
 
@@ -295,8 +301,13 @@
 	 * If an error has been encountered then attempt to probe the device
 	 * to determine if it's still accessible.
 	 */
-	if (zio->io_error == EIO && vdev_probe(vd) != 0)
-		vd->vdev_is_failing = B_TRUE;
+	if (zio->io_error == EIO && vdev_probe(vd) != 0) {
+		if (!vd->vdev_is_failing) {
+			vd->vdev_is_failing = B_TRUE;
+			zfs_ereport_post(FM_EREPORT_ZFS_PROBE_FAILURE,
+			    vd->vdev_spa, vd, zio, 0, 0);
+		}
+	}
 
 	vdev_queue_io_done(zio);
 
--- a/usr/src/uts/common/fs/zfs/vdev_label.c	Fri Jun 27 14:18:08 2008 -0700
+++ b/usr/src/uts/common/fs/zfs/vdev_label.c	Fri Jun 27 17:05:29 2008 -0700
@@ -913,6 +913,15 @@
 }
 
 /*
+ * We ignore errors for log devices, simply free the private data.
+ */
+static void
+vdev_label_sync_log_done(zio_t *zio)
+{
+	kmem_free(zio->io_private, sizeof (uint64_t));
+}
+
+/*
  * Write all even or odd labels to all leaves of the specified vdev.
  */
 static void
@@ -984,7 +993,8 @@
 	for (vd = list_head(dl); vd != NULL; vd = list_next(dl, vd)) {
 		uint64_t *good_writes = kmem_zalloc(sizeof (uint64_t),
 		    KM_SLEEP);
-		zio_t *vio = zio_null(nio, spa, vdev_label_sync_top_done,
+		zio_t *vio = zio_null(nio, spa, vd->vdev_islog ?
+		    vdev_label_sync_log_done : vdev_label_sync_top_done,
 		    good_writes, flags);
 		vdev_label_sync(vio, vd, l, txg);
 		zio_nowait(vio);
--- a/usr/src/uts/common/fs/zfs/vdev_root.c	Fri Jun 27 14:18:08 2008 -0700
+++ b/usr/src/uts/common/fs/zfs/vdev_root.c	Fri Jun 27 17:05:29 2008 -0700
@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -48,7 +48,7 @@
 too_many_errors(vdev_t *vd, int numerrors)
 {
 	ASSERT3U(numerrors, <=, vd->vdev_children);
-	return (numerrors == vd->vdev_children);
+	return (numerrors > 0);
 }
 
 static int
@@ -67,22 +67,17 @@
 		vdev_t *cvd = vd->vdev_child[c];
 		int error;
 
-		if ((error = vdev_open(cvd)) != 0) {
+		if ((error = vdev_open(cvd)) != 0 &&
+		    !cvd->vdev_islog) {
 			lasterror = error;
 			numerrors++;
 			continue;
 		}
 	}
 
-	if (numerrors > 0) {
-		if (!too_many_errors(vd, numerrors)) {
-			/* XXX - should not be explicitly setting this state */
-			vdev_set_state(vd, B_FALSE, VDEV_STATE_FAULTED,
-			    VDEV_AUX_NO_REPLICAS);
-		} else {
-			vd->vdev_stat.vs_aux = VDEV_AUX_NO_REPLICAS;
-			return (lasterror);
-		}
+	if (too_many_errors(vd, numerrors)) {
+		vd->vdev_stat.vs_aux = VDEV_AUX_NO_REPLICAS;
+		return (lasterror);
 	}
 
 	*asize = 0;
@@ -103,13 +98,9 @@
 static void
 vdev_root_state_change(vdev_t *vd, int faulted, int degraded)
 {
-	if (faulted) {
-		if (too_many_errors(vd, faulted))
-			vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
-			    VDEV_AUX_NO_REPLICAS);
-		else
-			vdev_set_state(vd, B_FALSE, VDEV_STATE_FAULTED,
-			    VDEV_AUX_NO_REPLICAS);
+	if (too_many_errors(vd, faulted)) {
+		vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
+		    VDEV_AUX_NO_REPLICAS);
 	} else if (degraded) {
 		vdev_set_state(vd, B_FALSE, VDEV_STATE_DEGRADED, VDEV_AUX_NONE);
 	} else {
--- a/usr/src/uts/common/fs/zfs/zfs_fm.c	Fri Jun 27 14:18:08 2008 -0700
+++ b/usr/src/uts/common/fs/zfs/zfs_fm.c	Fri Jun 27 17:05:29 2008 -0700
@@ -141,6 +141,16 @@
 		 */
 		if (zio->io_flags & ZIO_FLAG_SPECULATIVE)
 			return;
+
+		/*
+		 * If the vdev has already been marked as failing due to a
+		 * failed probe, then ignore any subsequent I/O errors, as the
+		 * DE will automatically fault the vdev on the first such
+		 * failure.
+		 */
+		if (vd != NULL && vd->vdev_is_failing &&
+		    strcmp(subclass, FM_EREPORT_ZFS_PROBE_FAILURE) != 0)
+			return;
 	}
 
 	if ((ereport = fm_nvlist_create(NULL)) == NULL)
--- a/usr/src/uts/common/fs/zfs/zfs_vnops.c	Fri Jun 27 14:18:08 2008 -0700
+++ b/usr/src/uts/common/fs/zfs/zfs_vnops.c	Fri Jun 27 17:05:29 2008 -0700
@@ -3140,6 +3140,9 @@
 			zfs_log_rename(zilog, tx,
 			    TX_RENAME | (flags & FIGNORECASE ? TX_CI : 0),
 			    sdzp, sdl->dl_name, tdzp, tdl->dl_name, szp);
+
+			/* Update path information for the target vnode */
+			vn_renamepath(tdvp, ZTOV(szp), tnm, strlen(tnm));
 		}
 	}
 
--- a/usr/src/uts/common/fs/zfs/zio.c	Fri Jun 27 14:18:08 2008 -0700
+++ b/usr/src/uts/common/fs/zfs/zio.c	Fri Jun 27 17:05:29 2008 -0700
@@ -1834,8 +1834,12 @@
 		return (B_FALSE);
 	if (zio->io_delegate_list != NULL)
 		return (B_FALSE);
-	if (vd && vd != vd->vdev_top)
-		return (B_FALSE);
+	if (vd != NULL) {
+		if (vd != vd->vdev_top)
+			return (B_FALSE);
+		if (vd->vdev_is_failing)
+			return (B_FALSE);
+	}
 	if (zio->io_flags & ZIO_FLAG_DONT_RETRY)
 		return (B_FALSE);
 	if (zio->io_retries > 0)
--- a/usr/src/uts/common/sys/fm/fs/zfs.h	Fri Jun 27 14:18:08 2008 -0700
+++ b/usr/src/uts/common/sys/fm/fs/zfs.h	Fri Jun 27 17:05:29 2008 -0700
@@ -46,6 +46,7 @@
 #define	FM_EREPORT_ZFS_DEVICE_TOO_SMALL		"vdev.too_small"
 #define	FM_EREPORT_ZFS_DEVICE_BAD_LABEL		"vdev.bad_label"
 #define	FM_EREPORT_ZFS_IO_FAILURE		"io_failure"
+#define	FM_EREPORT_ZFS_PROBE_FAILURE		"probe_failure"
 
 #define	FM_EREPORT_PAYLOAD_ZFS_POOL		"pool"
 #define	FM_EREPORT_PAYLOAD_ZFS_POOL_FAILMODE	"pool_failmode"
--- a/usr/src/uts/common/sys/vnode.h	Fri Jun 27 14:18:08 2008 -0700
+++ b/usr/src/uts/common/sys/vnode.h	Fri Jun 27 17:05:29 2008 -0700
@@ -1225,6 +1225,7 @@
 void vn_setpath_str(struct vnode *vp, const char *str, size_t len);
 void vn_setpath(vnode_t *rootvp, struct vnode *startvp, struct vnode *vp,
     const char *path, size_t plen);
+void vn_renamepath(vnode_t *dvp, vnode_t *vp, const char *nm, size_t len);
 
 /* Vnode event notification */
 void	vnevent_rename_src(vnode_t *, vnode_t *, char *, caller_context_t *);