Add support for autoexpand property

While the autoexpand property may seem like a small feature it
depends on a significant amount of system infrastructure.  Enough
of that infrastructure is now in place with a few customizations
for Linux the autoexpand property for whole disk configurations
can be supported.

Autoexpand works as follows; when a block device is resized a
change event is generated by udev with the DISK_MEDIA_CHANGE key.
The ZED, which is monitoring udev events detects the event for
disks (but not partitions) and hands it off to zfs_deliver_dle().
The zfs_deliver_dle() function appends the exected whole disk
partition suffix, and if the partition can be matched against
a known pool vdev it re-opens it.

Re-opening the vdev with trigger a re-reading of the partition
table so the maximum possible expansion size can be reported.
Next if the property autoexpand is set to "on" a vdev expansion
will be attempted.  After performing some sanity checks on the
disk to verify it's safe to expand the ZFS partition (-part1) it
will be expanded an the partition table updated.  The partition
is then re-opened again to detect the updated size which allows
the new capacity to be used.

Added PHYS_PATH="/dev/zvol/dataset" to vdev configuration for
ZFS volumes.  This was required for the test cases which test
expansion by layering a new pool on top of ZFS volumes.

Enable the zpool_expand_001_pos and /zpool_expand_003_pos
test cases which excercise the autoexpand property.

Fixed zfs_zevent_wait() signal handling which could result
in the ZED spinning when a signal was not handled.

Removed vdev_disk_rrpart() functionality which can be abandoned
in favour of re-opening the device which trigger a re-read of
the partition table as long no other partitions are in use.
This will be true as long as we're working with hole disks.
As a bonus this allows us to remove to Linux kernel API checks.

Signed-off-by: Brian Behlendorf <[email protected]>
Issue openzfs#120
Issue openzfs#2437
Issue openzfs#5771
Issue openzfs#7582

cmd/zed/agents/zfs_mod.c

@@ -751,23 +751,30 @@ zfsdle_vdev_online(zpool_handle_t *zhp, void *data)
 }
 
 /*
- * This function handles the ESC_DEV_DLE event.
+ * This function handles the ESC_DEV_DLE (DISK_MEDIA_CHANGE) event which
+ * is only delivered for the disk itself, not for each partition.  Presume
+ * that a 'wholedisk' partition exists and append the expected partition
+ * suffix in order to attempt a match.
  */
 static int
 zfs_deliver_dle(nvlist_t *nvl)
 {
-	char *devname;
+	char *devname, pname[MAXPATHLEN];
 
 	if (nvlist_lookup_string(nvl, DEV_PHYS_PATH, &devname) != 0) {
 		zed_log_msg(LOG_INFO, "zfs_deliver_dle: no physpath");
 		return (-1);
 	}
 
-	if (zpool_iter(g_zfshdl, zfsdle_vdev_online, devname) != 1) {
+	strlcpy(pname, devname, MAXPATHLEN);
+	zfs_append_partition(pname, MAXPATHLEN);
+
+	if (zpool_iter(g_zfshdl, zfsdle_vdev_online, pname) != 1) {
 		zed_log_msg(LOG_INFO, "zfs_deliver_dle: device '%s' not "
-		    "found", devname);
+		    "found", pname);
 		return (1);
 	}
+
 	return (0);
 }
 

cmd/zed/zed_disk_event.c

@@ -165,11 +165,12 @@ zed_udev_monitor(void *arg)
 
 	while (1) {
 		struct udev_device *dev;
-		const char *action, *type, *part, *sectors;
+		const char *action, *type, *part, *sectors, *change;
 		const char *bus, *uuid;
 		const char *class, *subclass;
 		nvlist_t *nvl;
 		boolean_t is_zfs = B_FALSE;
+		boolean_t is_disk_media_change = B_FALSE;
 
 		/* allow a cancellation while blocked (recvmsg) */
 		pthread_setcancelstate(PTHREAD_CANCEL_ENABLE, NULL);
@@ -202,14 +203,26 @@ zed_udev_monitor(void *arg)
 		}
 
 		/*
-		 * if this is a disk and it is partitioned, then the
+		 * Disk media change events are allowed for auto-expand.
+		 * Whether the device contains a zfs_member is determined
+		 * at the time of the attempted expansion.
+		 */
+		change = udev_device_get_property_value(dev,
+		    "DISK_MEDIA_CHANGE");
+		if (change != NULL && change[0] == '1')
+			is_disk_media_change = B_TRUE;
+
+		/*
+		 * If this is a disk and it is partitioned, then the
 		 * zfs label will reside in a DEVTYPE=partition and
-		 * we can skip passing this event
+		 * we can skip passing this event.  Unless it's a disk
+		 * media changes event which is expected for auto-expand.
 		 */
 		type = udev_device_get_property_value(dev, "DEVTYPE");
 		part = udev_device_get_property_value(dev,
 		    "ID_PART_TABLE_TYPE");
-		if (type != NULL && type[0] != '\0' &&
+		if (!is_disk_media_change &&
+		    type != NULL && type[0] != '\0' &&
 		    strcmp(type, "disk") == 0 &&
 		    part != NULL && part[0] != '\0') {
 			/* skip and wait for partition event */
@@ -231,14 +244,15 @@ zed_udev_monitor(void *arg)
 		}
 
 		/*
-		 * If the blkid probe didn't find ZFS, then a persistent
-		 * device id string is required in the message schema
-		 * for matching with vdevs. Preflight here for expected
-		 * udev information.
+		 * If the blkid probe didn't find ZFS and this is not a
+		 * disk media change event.  Then a persistent device id
+		 * string is required in the message schema for matching
+		 * with vdevs. Preflight here for expected udev information.
 		 */
 		bus = udev_device_get_property_value(dev, "ID_BUS");
 		uuid = udev_device_get_property_value(dev, "DM_UUID");
-		if (!is_zfs && (bus == NULL && uuid == NULL)) {
+		if (!is_zfs && !is_disk_media_change &&
+		    bus == NULL && uuid == NULL) {
 			zed_log_msg(LOG_INFO, "zed_udev_monitor: %s no devid "
 			    "source", udev_device_get_devnode(dev));
 			udev_device_unref(dev);

config/kernel.m4

@@ -44,7 +44,6 @@ AC_DEFUN([ZFS_AC_CONFIG_KERNEL], [
 	ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS_CHECK_EVENTS
 	ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS_RELEASE_VOID
 	ZFS_AC_KERNEL_TYPE_FMODE_T
-	ZFS_AC_KERNEL_3ARG_BLKDEV_GET
 	ZFS_AC_KERNEL_BLKDEV_GET_BY_PATH
 	ZFS_AC_KERNEL_OPEN_BDEV_EXCLUSIVE
 	ZFS_AC_KERNEL_LOOKUP_BDEV
@@ -73,7 +72,6 @@ AC_DEFUN([ZFS_AC_CONFIG_KERNEL], [
 	ZFS_AC_KERNEL_BLK_QUEUE_HAVE_BLK_PLUG
 	ZFS_AC_KERNEL_GET_DISK_AND_MODULE
 	ZFS_AC_KERNEL_GET_DISK_RO
-	ZFS_AC_KERNEL_GET_GENDISK
 	ZFS_AC_KERNEL_HAVE_BIO_SET_OP_ATTRS
 	ZFS_AC_KERNEL_GENERIC_READLINK_GLOBAL
 	ZFS_AC_KERNEL_DISCARD_GRANULARITY

lib/libzfs/libzfs_import.c

@@ -145,6 +145,21 @@ zfs_device_get_devid(struct udev_device *dev, char *bufptr, size_t buflen)
 			return (0);
 		}
 
+		/*
+		 * For volumes use the persistent /dev/zvol/dataset identifier
+		 */
+		entry = udev_device_get_devlinks_list_entry(dev);
+		while (entry != NULL) {
+			const char *name;
+
+			name = udev_list_entry_get_name(entry);
+			if (strncmp(name, ZVOL_ROOT, strlen(ZVOL_ROOT)) == 0) {
+				(void) strlcpy(bufptr, name, buflen);
+				return (0);
+			}
+			entry = udev_list_entry_get_next(entry);
+		}
+
 		/*
 		 * NVME 'by-id' symlinks are similar to bus case
 		 */
@@ -187,26 +202,44 @@ int
 zfs_device_get_physical(struct udev_device *dev, char *bufptr, size_t buflen)
 {
 	const char *physpath = NULL;
+	struct udev_list_entry *entry;
 
 	/*
-	 * Normal disks use ID_PATH for their physical path.  Device mapper
-	 * devices are virtual and don't have a physical path.  For them we
-	 * use ID_VDEV instead, which is setup via the /etc/vdev_id.conf file.
-	 * ID_VDEV provides a persistent path to a virtual device.  If you
-	 * don't have vdev_id.conf setup, you cannot use multipath autoreplace.
+	 * Normal disks use ID_PATH for their physical path.
 	 */
-	if (!((physpath = udev_device_get_property_value(dev, "ID_PATH")) &&
-	    physpath[0])) {
-		if (!((physpath =
-		    udev_device_get_property_value(dev, "ID_VDEV")) &&
-		    physpath[0])) {
-			return (ENODATA);
-		}
+	physpath = udev_device_get_property_value(dev, "ID_PATH");
+	if (physpath != NULL && strlen(physpath) > 0) {
+		(void) strlcpy(bufptr, physpath, buflen);
+		return (0);
+	}
+
+	/*
+	 * Device mapper devices are virtual and don't have a physical
+	 * path. For them we use ID_VDEV instead, which is setup via the
+	 * /etc/vdev_id.conf file.  ID_VDEV provides a persistent path
+	 * to a virtual device.  If you don't have vdev_id.conf setup,
+	 * you cannot use multipath autoreplace with device mapper.
+	 */
+	physpath = udev_device_get_property_value(dev, "ID_VDEV");
+	if (physpath != NULL && strlen(physpath) > 0) {
+		(void) strlcpy(bufptr, physpath, buflen);
+		return (0);
 	}
 
-	(void) strlcpy(bufptr, physpath, buflen);
+	/*
+	 * For volumes use the persistent /dev/zvol/dataset identifier
+	 */
+	entry = udev_device_get_devlinks_list_entry(dev);
+	while (entry != NULL) {
+		physpath = udev_list_entry_get_name(entry);
+		if (strncmp(physpath, ZVOL_ROOT, strlen(ZVOL_ROOT)) == 0) {
+			(void) strlcpy(bufptr, physpath, buflen);
+			return (0);
+		}
+		entry = udev_list_entry_get_next(entry);
+	}
 
-	return (0);
+	return (ENODATA);
 }
 
 boolean_t

module/zfs/fm.c

@@ -671,19 +671,31 @@ zfs_zevent_wait(zfs_zevent_t *ze)
 	int error = 0;
 
 	mutex_enter(&zevent_lock);
+	zevent_waiters++;
 
-	if (zevent_flags & ZEVENT_SHUTDOWN) {
-		error = ESHUTDOWN;
-		goto out;
-	}
+	while (error == 0) {
+		if (zevent_flags & ZEVENT_SHUTDOWN) {
+			error = SET_ERROR(ESHUTDOWN);
+			break;
+		}
 
-	zevent_waiters++;
-	cv_wait_sig(&zevent_cv, &zevent_lock);
-	if (issig(JUSTLOOKING))
-		error = EINTR;
+		error = cv_timedwait_sig(&zevent_cv, &zevent_lock,
+		    ddi_get_lbolt() + hz);
+		if (signal_pending(current) || fatal_signal_pending(current)) {
+			error = SET_ERROR(EINTR);
+			break;
+		} else {
+			if (error == -1) {
+				error = 0;
+				continue;
+			} else {
+				error = 0;
+				break;
+			}
+		}
+	}
 
 	zevent_waiters--;
-out:
 	mutex_exit(&zevent_lock);
 
 	return (error);

module/zfs/vdev.c

@@ -3097,7 +3097,8 @@ vdev_online(spa_t *spa, uint64_t guid, uint64_t flags, vdev_state_t *newstate)
 	/* XXX - L2ARC 1.0 does not support expansion */
 	if (!vd->vdev_aux) {
 		for (pvd = vd; pvd != rvd; pvd = pvd->vdev_parent)
-			pvd->vdev_expanding = !!(flags & ZFS_ONLINE_EXPAND);
+			pvd->vdev_expanding = !!((flags & ZFS_ONLINE_EXPAND) ||
+			    spa->spa_autoexpand);
 	}
 
 	vdev_reopen(tvd);