Enclosure LED fixes

- Pass $VDEV_ENC_SYSFS_PATH to 'zpool [iostat|status] -c' to include
  enclosure LED sysfs path.

- Set LEDs correctly after import.  This includes clearing any erroniously
  set LEDs prior to the import, and setting the LED for any UNAVAIL drives.

- Include symlink for vdev_attach-led.sh in Makefile.am.

- Print the VDEV path in all-syslog.sh, and fix it so the pool GUID actually
  prints.    

Reviewed-by: Don Brady <don.brady@intel.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Tony Hutter <hutter2@llnl.gov>
Closes #5716 
Closes #5751

cmd/zed/Makefile.am

@@ -67,7 +67,9 @@ dist_zedexec_SCRIPTS = \
 	zed.d/scrub_finish-notify.sh \
 	zed.d/statechange-led.sh \
 	zed.d/statechange-notify.sh \
-	zed.d/vdev_clear-led.sh
+	zed.d/vdev_clear-led.sh \
+	zed.d/vdev_attach-led.sh \
+	zed.d/pool_import-led.sh
 
 zedconfdefaults = \
 	all-syslog.sh \
@@ -76,7 +78,9 @@ zedconfdefaults = \
 	scrub_finish-notify.sh \
 	statechange-led.sh \
 	statechange-notify.sh \
-	vdev_clear-led.sh
+	vdev_clear-led.sh \
+	vdev_attach-led.sh \
+	pool_import-led.sh
 
 install-data-hook:
 	$(MKDIR_P) "$(DESTDIR)$(zedconfdir)"

cmd/zed/zed.d/all-syslog.sh

@@ -6,6 +6,7 @@
 . "${ZED_ZEDLET_DIR}/zed-functions.sh"
 
 zed_log_msg "eid=${ZEVENT_EID}" "class=${ZEVENT_SUBCLASS}" \
-    "${ZEVENT_POOL:+"pool=${ZEVENT_POOL}"}" \
+    "${ZEVENT_POOL_GUID:+"pool_guid=${ZEVENT_POOL_GUID}"}" \
+    "${ZEVENT_VDEV_PATH:+"vdev_path=${ZEVENT_VDEV_PATH}"}" \
     "${ZEVENT_VDEV_STATE_STR:+"vdev_state=${ZEVENT_VDEV_STATE_STR}"}"
 exit 0

cmd/zed/zed.d/pool_import-led.sh

@@ -0,0 +1 @@
+statechange-led.sh

cmd/zed/zed.d/statechange-led.sh

@@ -2,13 +2,19 @@
 #
 # Turn off/on the VDEV's enclosure fault LEDs when the pool's state changes.
 #
-# Turn LED on if the VDEV becomes faulted or degraded, and turn it back off
+# Turn the VDEV's fault LED on if it becomes FAULTED, DEGRADED or UNAVAIL.
-# when it's online again.  It will also turn on the LED (or keep it on) if
+# Turn the LED off when it's back ONLINE again.
-# the drive becomes unavailable, unless the drive was in was a previously
-# online state (online->unavail is a normal state transition during an
-# autoreplace).
 #
-# This script requires that your enclosure be supported by the
+# This script run in two basic modes:
+#
+# 1. If $ZEVENT_VDEV_ENC_SYSFS_PATH and $ZEVENT_VDEV_STATE_STR are set, then
+# only set the LED for that particular VDEV. This is the case for statechange
+# events and some vdev_* events.
+#
+# 2. If those vars are not set, then check the state of all VDEVs in the pool,i
+# and set the LEDs accordingly.  This is the case for pool_import events.
+#
+# Note that this script requires that your enclosure be supported by the
 # Linux SCSI enclosure services (ses) driver.  The script will do nothing
 # if you have no enclosure, or if your enclosure isn't supported.
 #
@@ -16,8 +22,8 @@
 #   0: enclosure led successfully set
 #   1: enclosure leds not not available
 #   2: enclosure leds administratively disabled
-#   3: ZED didn't pass enclosure sysfs path
+#   3: The led sysfs path passed from ZFS does not exist
-#   4: Enclosure sysfs path doesn't exist
+#   4: $ZPOOL not set
 
 [ -f "${ZED_ZEDLET_DIR}/zed.rc" ] && . "${ZED_ZEDLET_DIR}/zed.rc"
 . "${ZED_ZEDLET_DIR}/zed-functions.sh"
@@ -30,15 +36,54 @@ if [ "${ZED_USE_ENCLOSURE_LEDS}" != "1" ] ; then
 	exit 2
 fi
 
-[ -n "${ZEVENT_VDEV_ENC_SYSFS_PATH}" ] || exit 3
+zed_check_cmd "$ZPOOL" || exit 4
 
-[ -e "${ZEVENT_VDEV_ENC_SYSFS_PATH}/fault" ] || exit 4
+# Global used in set_led debug print
+vdev=""
 
-# Turn on/off enclosure LEDs
+# set_led (file)
-function led
+#
+# Write an enclosure sysfs file
+#
+# Arguments
+#   file: sysfs file to set (like /sys/class/enclosure/0:0:1:0/SLOT 10/fault)
+#   val: value to set it to
+#
+# Globals
+#   vdev: VDEV name for debug printing
+#
+# Return
+#   nothing
+#
+function set_led {
+	file="$1"
+	val="$2"
+
+	# Set the value twice.  I've seen enclosures that were
+	# flakey about setting it the first time.
+	echo "$val" > "$file"
+	echo "$val" > "$file"
+	zed_log_msg "vdev $vdev set '$file' LED to $val"
+}
+
+# check_and_set_led (file, val)
+#
+# Read an enclosure sysfs file, and write it if it's not already set to 'val'
+#
+# Arguments
+#   file: sysfs file to set (like /sys/class/enclosure/0:0:1:0/SLOT 10/fault)
+#   val: value to set it to
+#
+# Return
+#  0 on success, 3 on missing sysfs path
+#
+function check_and_set_led
 {
-	file="$1/fault"
+	file="$1"
-	val=$2
+	val="$2"
+	if [ ! -e "$ZEVENT_VDEV_ENC_SYSFS_PATH/fault" ] ; then
+		return 3
+	fi
 
 	# We want to check the current state first, since writing to the
 	# 'fault' entry always always causes a SES command, even if the
@@ -53,28 +98,88 @@ function led
 	fi
 
 	if [ "$current" != "$val" ] ; then
-		# Set the value twice.  I've seen enclosures that were
+		set_led "$file" "$val"
-		# flakey about setting it the first time.
-		echo "$val" > "$file"
-		echo "$val" > "$file"
 	fi
 }
 
-# Decide whether to turn on/off an LED based on the state
+function state_to_val {
-# Pass in path name and fault string ("ONLINE"/"FAULTED"/"DEGRADED"...etc)
+	state="$1"
+	if [ "$state" == "FAULTED" ] || [ "$state" == "DEGRADED" ] || \
+	   [ "$state" == "UNAVAIL" ] ; then
+		echo 1
+	elif [ "$state" == "ONLINE" ] ; then
+		echo 0
+	fi
+}
+
+# process_pool ([pool])
 #
-# We only turn on LEDs when a drive becomes FAULTED, DEGRADED, or UNAVAIL and
+# Iterate through a pool (or pools) and set the VDEV's enclosure slot LEDs to
-# only turn it on when it comes back ONLINE.  All other states are ignored, and
+# the VDEV's state.
-# keep the previous LED state.  
+#
-function process {
+# Arguments
-	path="$1"
+#   pool:	Optional pool name.  If not specified, iterate though all pools.
-	fault=$2
+#
-	if [ "$fault" == "FAULTED" ] || [ "$fault" == "DEGRADED" ] || \
+# Return
-	   [ "$fault" == "UNAVAIL" ] ; then
+#  0 on success, 3 on missing sysfs path
-		led "$path" 1
+#
-	elif [ "$fault" == "ONLINE" ] ; then
+function process_pool
-		led "$path" 0
+{
+	pool="$1"
+	rc=0
+
+	# Lookup all the current LED values and paths in parallel
+	cmd='echo led_token=$(cat "$VDEV_ENC_SYSFS_PATH/fault"),"$VDEV_ENC_SYSFS_PATH",'
+	out=$($ZPOOL status -vc "$cmd" $pool | grep 'led_token=')
+	while read -r vdev state read write chksum therest ; do
+
+		# Read out current LED value and path
+		tmp=$(echo "$therest" | sed 's/^.*led_token=//g')
+		IFS="," read -r current_val vdev_enc_sysfs_path <<< "$tmp"
+
+		if [ -z "$vdev_enc_sysfs_path" ] ; then
+			# Skip anything with no sysfs LED entries
+			continue
+		fi
+
+		# On some enclosures if you write 1 to fault, and read it back,
+		# it will return 2.  Treat all non-zero values as 1 for
+		# simplicity.
+		if [ "$current_val" != "0" ] ; then
+			current_val=1
+		fi
+
+		val=$(state_to_val "$state")
+
+		if [ "$current_val" == "$val" ] ; then
+			# LED is already set correctly
+			continue
+		fi
+
+		if [ ! -e "$vdev_enc_sysfs_path/fault" ] ; then
+			rc=1
+			zed_log_msg "vdev $vdev '$file/fault' doesn't exist"
+			continue;
+		fi
+
+		set_led "$vdev_enc_sysfs_path/fault" $val
+
+	done <<< "$out"
+	if [ "$rc" == "0" ] ; then
+		return 0
+	else
+		# We didn't see a sysfs entry that we wanted to set
+		return 3
 	fi
 }
 
-process "$ZEVENT_VDEV_ENC_SYSFS_PATH" "$ZEVENT_VDEV_STATE_STR"
+if [ ! -z "$ZEVENT_VDEV_ENC_SYSFS_PATH" ] && [ ! -z "$ZEVENT_VDEV_STATE_STR" ] ; then
+	# Got a statechange for an individual VDEV
+	val=$(state_to_val "$ZEVENT_VDEV_STATE_STR")
+
+	vdev="$(basename $ZEVENT_VDEV_PATH)"
+	check_and_set_led "$ZEVENT_VDEV_ENC_SYSFS_PATH/fault" "$val"
+else
+	# Process the entire pool
+	process_pool $(zed_guid_to_pool $ZEVENT_POOL_GUID)
+fi

cmd/zed/zed.d/zed-functions.sh

@@ -413,3 +413,25 @@ zed_rate_limit()
     zed_unlock "${lockfile}" "${lockfile_fd}"
     return "${rv}"
 }
+
+
+# zed_guid_to_pool (guid)
+#
+# Convert a pool GUID into its pool name (like "tank")
+# Arguments
+#   guid: pool GUID (decimal or hex)
+#
+# Return
+#   Pool name
+#
+function zed_guid_to_pool
+{
+	if [ -z "$1" ] ; then
+		return
+	fi
+
+	guid=$(printf "%llu" $1)
+	if [ ! -z "$guid" ] ; then
+		$ZPOOL get -H -ovalue,name guid | awk '$1=='$guid' {print $2}'
+	fi
+}

cmd/zpool/zpool_iter.c

@@ -332,9 +332,11 @@ vdev_run_cmd_thread(void *cb_cmd_data)
 	char cmd[_POSIX_ARG_MAX];
 
 	/* Set our VDEV_PATH and VDEV_UPATH env vars and run command */
-	if (snprintf(cmd, sizeof (cmd), "VDEV_PATH=%s && VDEV_UPATH=%s && %s",
+	if (snprintf(cmd, sizeof (cmd), "VDEV_PATH=%s && VDEV_UPATH=\"%s\" && "
-	    data->path, data->upath ? data->upath : "\"\"", data->cmd) >=
+	    "VDEV_ENC_SYSFS_PATH=\"%s\" && %s", data->path ? data->path : "",
-	    sizeof (cmd)) {
+	    data->upath ? data->upath : "",
+	    data->vdev_enc_sysfs_path ? data->vdev_enc_sysfs_path : "",
+	    data->cmd) >= sizeof (cmd)) {
 		/* Our string was truncated */
 		return;
 	}
@@ -364,11 +366,15 @@ for_each_vdev_run_cb(zpool_handle_t *zhp, nvlist_t *nv, void *cb_vcdl)
 	vdev_cmd_data_t *data;
 	char *path = NULL;
 	char *vname = NULL;
+	char *vdev_enc_sysfs_path = NULL;
 	int i, match = 0;
 
 	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) != 0)
 		return (1);
 
+	nvlist_lookup_string(nv, ZPOOL_CONFIG_VDEV_ENC_SYSFS_PATH,
+	    &vdev_enc_sysfs_path);
+
 	/* Spares show more than once if they're in use, so skip if exists */
 	for (i = 0; i < vcdl->count; i++) {
 		if ((strcmp(vcdl->data[i].path, path) == 0) &&
@@ -406,6 +412,10 @@ for_each_vdev_run_cb(zpool_handle_t *zhp, nvlist_t *nv, void *cb_vcdl)
 	data->path = strdup(path);
 	data->upath = zfs_get_underlying_path(path);
 	data->cmd = vcdl->cmd;
+	if (vdev_enc_sysfs_path)
+		data->vdev_enc_sysfs_path = strdup(vdev_enc_sysfs_path);
+	else
+		data->vdev_enc_sysfs_path = NULL;
 
 	vcdl->count++;
 
@@ -500,6 +510,7 @@ free_vdev_cmd_data_list(vdev_cmd_data_list_t *vcdl)
 		free(vcdl->data[i].pool);
 		free(vcdl->data[i].upath);
 		free(vcdl->data[i].line);
+		free(vcdl->data[i].vdev_enc_sysfs_path);
 	}
 	free(vcdl->data);
 	free(vcdl);

cmd/zpool/zpool_util.h

@@ -80,6 +80,7 @@ typedef	struct vdev_cmd_data
 	char *upath;	/* vdev underlying path */
 	char *pool;	/* Pool name */
 	char *cmd;	/* backpointer to cmd */
+	char *vdev_enc_sysfs_path;	/* enclosure sysfs path (if any) */
 } vdev_cmd_data_t;
 
 typedef struct vdev_cmd_data_list

man/man8/zpool.8

@@ -1548,12 +1548,16 @@ base 1024.  To get the raw values, use the \fB-p\fR flag.
 Run a command on each vdev and include first line of output
 .sp
 The \fB-c\fR option allows you to run an arbitrary command on each vdev and
-display the first line of output in zpool iostat.  The environment vars
+display the first line of output in zpool iostat.  The following environment
-\fBVDEV_PATH\fR and \fBVDEV_UPATH\fR are set to the vdev's path and "underlying
+vars are set before running each command:
-path" before running the command.  For device mapper, multipath, or partitioned
+.sp
-vdevs, \fBVDEV_UPATH\fR is the actual underlying /dev/sd* disk.  This can be
+\fB$VDEV_PATH\fR: Full path to the vdev.
-useful if the command you're running requires a /dev/sd* device.  Commands run
+.LP
-in parallel for each vdev for performance.
+\fB$VDEV_UPATH\fR: "Underlying path" to the vdev.  For device mapper, multipath, or
+partitioned vdevs, \fBVDEV_UPATH\fR is the actual underlying /dev/sd* disk.
+This can be useful if the command you're running requires a /dev/sd* device.
+.LP
+\fB$VDEV_ENC_SYSFS_PATH\fR: The sysfs path to the vdev's enclosure LEDs (if any).
 .RE
 
 .sp
@@ -2116,12 +2120,16 @@ If a scrub or resilver is in progress, this command reports the percentage done
 Run a command on each vdev and include first line of output
 .sp
 The \fB-c\fR option allows you to run an arbitrary command on each vdev and
-display the first line of output in zpool status.  The environment vars
+display the first line of output in zpool iostat.  The following environment
-\fBVDEV_PATH\fR and \fBVDEV_UPATH\fR are set to the vdev's path and "underlying
+vars are set before running each command:
-path" before running the command.  For device mapper, multipath, or partitioned
+.sp
-vdevs, \fBVDEV_UPATH\fR is the actual underlying /dev/sd* disk.  This can be
+\fB$VDEV_PATH\fR: Full path to the vdev.
-useful if the command you're running requires a /dev/sd* device.  Commands run
+.LP
-in parallel for each vdev for performance.
+\fB$VDEV_UPATH\fR: "Underlying path" to the vdev.  For device mapper, multipath, or
+partitioned vdevs, \fBVDEV_UPATH\fR is the actual underlying /dev/sd* disk.
+This can be useful if the command you're running requires a /dev/sd* device.
+.LP
+\fB$VDEV_ENC_SYSFS_PATH\fR: The sysfs path to the vdev's enclosure LEDs (if any).
 .RE
 
 .sp