Merge tag 'md-3.11' of git://neil.brown.name/md

Pull md updates from NeilBrown:
"Mostly fixes, with a few minor features (eg 'last_sync_action' sysfs
file)

A couple marked for -stable including one recent bug which causes a
RAID10 reshape to complete without moving any data :-(

A couple more bugfixes (at least) to come, but haven't confirmed the
right solution yet."

* tag 'md-3.11' of git://neil.brown.name/md:
md/raid10: fix bug which causes all RAID10 reshapes to move no data.
md/raid5: allow 5-device RAID6 to be reshaped to 4-device.
md/raid10: fix two bugs affecting RAID10 reshape.
md: remove doubled description for sync_max, merging it within sync_min/sync_max
MD: Remember the last sync operation that was performed
md: fix buglet in RAID5 -> RAID0 conversion.
md/raid10: check In_sync flag in 'enough()'.
md/raid10: locking changes for 'enough()'.
md: replace strict_strto*() with kstrto*()
md: Wait for md_check_recovery before attempting device removal.
dm-raid: silence compiler warning on rebuilds_per_group.
DM RAID: Fix raid_resume not reviving failed devices in all cases
DM RAID: Break-up untidy function
DM RAID: Add ability to restore transiently failed devices on resume

Documentation/device-mapper/dm-raid.txt

···
 1.4.2   Add RAID10 "far" and "offset" algorithm support.
 1.5.0   Add message interface to allow manipulation of the sync_action.
 	New status (STATUSTYPE_INFO) fields: sync_action and mismatch_cnt.
+1.5.1   Add ability to restore transiently failed devices on resume.
+1.5.2   'mismatch_cnt' is zero unless [last_]sync_action is "check".

Documentation/md.txt

···
      when it reaches the current sync_max (below) and possibly at
      other times.
 
-   sync_max
-     This is a number of sectors at which point a resync/recovery
-     process will pause.  When a resync is active, the value can
-     only ever be increased, never decreased.  The value of 'max'
-     effectively disables the limit.
-
-
    sync_speed
      This shows the current actual speed, in K/sec, of the current
      sync_action.  It is averaged over the last 30 seconds.
···
      You can use 'select' or 'poll' on "sync_completed" to wait for
      that number to reach sync_max.  Then you can either increase
      "sync_max", or can write 'idle' to "sync_action".
+
+     The value of 'max' for "sync_max" effectively disables the limit.
+     When a resync is active, the value can only ever be increased,
+     never decreased.
+     The value of '0' is the minimum for "sync_min".
+
 
 
 Each active md device may also have attributes specific to the

drivers/md/bitmap.c

···
 		} else {
 			int rv;
 			if (buf[0] == '+')
-				rv = strict_strtoll(buf+1, 10, &offset);
+				rv = kstrtoll(buf+1, 10, &offset);
 			else
-				rv = strict_strtoll(buf, 10, &offset);
+				rv = kstrtoll(buf, 10, &offset);
 			if (rv)
 				return rv;
 			if (offset == 0)
···
 backlog_store(struct mddev *mddev, const char *buf, size_t len)
 {
 	unsigned long backlog;
-	int rv = strict_strtoul(buf, 10, &backlog);
+	int rv = kstrtoul(buf, 10, &backlog);
 	if (rv)
 		return rv;
 	if (backlog > COUNTER_MAX)
···
 	unsigned long csize;
 	if (mddev->bitmap)
 		return -EBUSY;
-	rv = strict_strtoul(buf, 10, &csize);
+	rv = kstrtoul(buf, 10, &csize);
 	if (rv)
 		return rv;
 	if (csize < 512 ||

drivers/md/dm-raid.c

···
 static int validate_raid_redundancy(struct raid_set *rs)
 {
 	unsigned i, rebuild_cnt = 0;
-	unsigned rebuilds_per_group, copies, d;
+	unsigned rebuilds_per_group = 0, copies, d;
 	unsigned group_size, last_group_start;
 
 	for (i = 0; i < rs->md.raid_disks; i++)
···
 	 * First, parse the in-order required arguments
 	 * "chunk_size" is the only argument of this type.
 	 */
-	if ((strict_strtoul(argv[0], 10, &value) < 0)) {
+	if ((kstrtoul(argv[0], 10, &value) < 0)) {
 		rs->ti->error = "Bad chunk size";
 		return -EINVAL;
 	} else if (rs->raid_type->level == 1) {
···
 			continue;
 		}
 
-		if (strict_strtoul(argv[i], 10, &value) < 0) {
+		if (kstrtoul(argv[i], 10, &value) < 0) {
 			rs->ti->error = "Bad numerical argument given in raid params";
 			return -EINVAL;
 		}
···
 	argv++;
 
 	/* number of RAID parameters */
-	if (strict_strtoul(argv[0], 10, &num_raid_params) < 0) {
+	if (kstrtoul(argv[0], 10, &num_raid_params) < 0) {
 		ti->error = "Cannot understand number of RAID parameters";
 		return -EINVAL;
 	}
···
 		return -EINVAL;
 	}
 
-	if ((strict_strtoul(argv[num_raid_params], 10, &num_raid_devs) < 0) ||
+	if ((kstrtoul(argv[num_raid_params], 10, &num_raid_devs) < 0) ||
 	    (num_raid_devs >= INT_MAX)) {
 		ti->error = "Cannot understand number of raid devices";
 		return -EINVAL;
···
 		 *   performing a "check" of the array.
 		 */
 		DMEMIT(" %llu",
+		       (strcmp(rs->md.last_sync_action, "check")) ? 0 :
 		       (unsigned long long)
 		       atomic64_read(&rs->md.resync_mismatches));
 		break;
···
 	mddev_suspend(&rs->md);
 }
 
+static void attempt_restore_of_faulty_devices(struct raid_set *rs)
+{
+	int i;
+	uint64_t failed_devices, cleared_failed_devices = 0;
+	unsigned long flags;
+	struct dm_raid_superblock *sb;
+	struct md_rdev *r;
+
+	for (i = 0; i < rs->md.raid_disks; i++) {
+		r = &rs->dev[i].rdev;
+		if (test_bit(Faulty, &r->flags) && r->sb_page &&
+		    sync_page_io(r, 0, r->sb_size, r->sb_page, READ, 1)) {
+			DMINFO("Faulty %s device #%d has readable super block."
+			       "  Attempting to revive it.",
+			       rs->raid_type->name, i);
+
+			/*
+			 * Faulty bit may be set, but sometimes the array can
+			 * be suspended before the personalities can respond
+			 * by removing the device from the array (i.e. calling
+			 * 'hot_remove_disk').  If they haven't yet removed
+			 * the failed device, its 'raid_disk' number will be
+			 * '>= 0' - meaning we must call this function
+			 * ourselves.
+			 */
+			if ((r->raid_disk >= 0) &&
+			    (r->mddev->pers->hot_remove_disk(r->mddev, r) != 0))
+				/* Failed to revive this device, try next */
+				continue;
+
+			r->raid_disk = i;
+			r->saved_raid_disk = i;
+			flags = r->flags;
+			clear_bit(Faulty, &r->flags);
+			clear_bit(WriteErrorSeen, &r->flags);
+			clear_bit(In_sync, &r->flags);
+			if (r->mddev->pers->hot_add_disk(r->mddev, r)) {
+				r->raid_disk = -1;
+				r->saved_raid_disk = -1;
+				r->flags = flags;
+			} else {
+				r->recovery_offset = 0;
+				cleared_failed_devices |= 1 << i;
+			}
+		}
+	}
+	if (cleared_failed_devices) {
+		rdev_for_each(r, &rs->md) {
+			sb = page_address(r->sb_page);
+			failed_devices = le64_to_cpu(sb->failed_devices);
+			failed_devices &= ~cleared_failed_devices;
+			sb->failed_devices = cpu_to_le64(failed_devices);
+		}
+	}
+}
+
 static void raid_resume(struct dm_target *ti)
 {
 	struct raid_set *rs = ti->private;
···
 	if (!rs->bitmap_loaded) {
 		bitmap_load(&rs->md);
 		rs->bitmap_loaded = 1;
+	} else {
+		/*
+		 * A secondary resume while the device is active.
+		 * Take this opportunity to check whether any failed
+		 * devices are reachable again.
+		 */
+		attempt_restore_of_faulty_devices(rs);
 	}
 
 	clear_bit(MD_RECOVERY_FROZEN, &rs->md.recovery);
···
 
 static struct target_type raid_target = {
 	.name = "raid",
-	.version = {1, 5, 0},
+	.version = {1, 5, 2},
 	.module = THIS_MODULE,
 	.ctr = raid_ctr,
 	.dtr = raid_dtr,

drivers/md/md.c

···
 	init_waitqueue_head(&mddev->recovery_wait);
 	mddev->reshape_position = MaxSector;
 	mddev->reshape_backwards = 0;
+	mddev->last_sync_action = "none";
 	mddev->resync_min = 0;
 	mddev->resync_max = MaxSector;
 	mddev->level = LEVEL_NONE;
···
 offset_store(struct md_rdev *rdev, const char *buf, size_t len)
 {
 	unsigned long long offset;
-	if (strict_strtoull(buf, 10, &offset) < 0)
+	if (kstrtoull(buf, 10, &offset) < 0)
 		return -EINVAL;
 	if (rdev->mddev->pers && rdev->raid_disk >= 0)
 		return -EBUSY;
···
 	unsigned long long new_offset;
 	struct mddev *mddev = rdev->mddev;
 
-	if (strict_strtoull(buf, 10, &new_offset) < 0)
+	if (kstrtoull(buf, 10, &new_offset) < 0)
 		return -EINVAL;
 
 	if (mddev->sync_thread)
···
 	unsigned long long blocks;
 	sector_t new;
 
-	if (strict_strtoull(buf, 10, &blocks) < 0)
+	if (kstrtoull(buf, 10, &blocks) < 0)
 		return -EINVAL;
 
 	if (blocks & 1ULL << (8 * sizeof(blocks) - 1))
···
 
 	if (cmd_match(buf, "none"))
 		recovery_start = MaxSector;
-	else if (strict_strtoull(buf, 10, &recovery_start))
+	else if (kstrtoull(buf, 10, &recovery_start))
 		return -EINVAL;
 
 	if (rdev->mddev->pers &&
···
 	if (clevel[len-1] == '\n')
 		len--;
 	clevel[len] = 0;
-	if (strict_strtol(clevel, 10, &level))
+	if (kstrtol(clevel, 10, &level))
 		level = LEVEL_NONE;
 
 	if (request_module("md-%s", clevel) != 0)
···
 	return len;
 }
 
+static struct md_sysfs_entry md_scan_mode =
+__ATTR(sync_action, S_IRUGO|S_IWUSR, action_show, action_store);
+
+static ssize_t
+last_sync_action_show(struct mddev *mddev, char *page)
+{
+	return sprintf(page, "%s\n", mddev->last_sync_action);
+}
+
+static struct md_sysfs_entry md_last_scan_mode = __ATTR_RO(last_sync_action);
+
 static ssize_t
 mismatch_cnt_show(struct mddev *mddev, char *page)
 {
···
 		       (unsigned long long)
 		       atomic64_read(&mddev->resync_mismatches));
 }
-
-static struct md_sysfs_entry md_scan_mode =
-__ATTR(sync_action, S_IRUGO|S_IWUSR, action_show, action_store);
-
 
 static struct md_sysfs_entry md_mismatches = __ATTR_RO(mismatch_cnt);
 
···
 {
 	long n;
 
-	if (strict_strtol(buf, 10, &n))
+	if (kstrtol(buf, 10, &n))
 		return -EINVAL;
 
 	if (n != 0 && n != 1)
···
 min_sync_store(struct mddev *mddev, const char *buf, size_t len)
 {
 	unsigned long long min;
-	if (strict_strtoull(buf, 10, &min))
+	if (kstrtoull(buf, 10, &min))
 		return -EINVAL;
 	if (min > mddev->resync_max)
 		return -EINVAL;
···
 		mddev->resync_max = MaxSector;
 	else {
 		unsigned long long max;
-		if (strict_strtoull(buf, 10, &max))
+		if (kstrtoull(buf, 10, &max))
 			return -EINVAL;
 		if (max < mddev->resync_min)
 			return -EINVAL;
···
 
 static struct attribute *md_redundancy_attrs[] = {
 	&md_scan_mode.attr,
+	&md_last_scan_mode.attr,
 	&md_mismatches.attr,
 	&md_sync_min.attr,
 	&md_sync_max.attr,
···
 		/* need to ensure md_delayed_delete() has completed */
 		flush_workqueue(md_misc_wq);
 
+	if (cmd == HOT_REMOVE_DISK)
+		/* need to ensure recovery thread has run */
+		wait_event_interruptible_timeout(mddev->sb_wait,
+						 !test_bit(MD_RECOVERY_NEEDED,
+							   &mddev->flags),
+						 msecs_to_jiffies(5000));
 	err = mddev_lock(mddev);
 	if (err) {
 		printk(KERN_INFO 
···
 	sector_t last_check;
 	int skipped = 0;
 	struct md_rdev *rdev;
-	char *desc;
+	char *desc, *action = NULL;
 	struct blk_plug plug;
 
 	/* just incase thread restarts... */
···
 		return;
 
 	if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
-		if (test_bit(MD_RECOVERY_CHECK, &mddev->recovery))
+		if (test_bit(MD_RECOVERY_CHECK, &mddev->recovery)) {
 			desc = "data-check";
-		else if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
+			action = "check";
+		} else if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
 			desc = "requested-resync";
-		else
+			action = "repair";
+		} else
 			desc = "resync";
 	} else if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
 		desc = "reshape";
 	else
 		desc = "recovery";
+
+	mddev->last_sync_action = action ?: desc;
 
 	/* we overload curr_resync somewhat here.
 	 * 0 == not engaged in resync at all
···
 			md_new_event(mddev);
 		}
 	unlock:
+		wake_up(&mddev->sb_wait);
+
 		if (!mddev->sync_thread) {
 			clear_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
 			if (test_and_clear_bit(MD_RECOVERY_RECOVER,

drivers/md/md.h

···
 
 	struct md_thread		*thread;	/* management thread */
 	struct md_thread		*sync_thread;	/* doing resync or reconstruct */
+
+	/* 'last_sync_action' is initialized to "none".  It is set when a
+	 * sync operation (i.e "data-check", "requested-resync", "resync",
+	 * "recovery", or "reshape") is started.  It holds this value even
+	 * when the sync thread is "frozen" (interrupted) or "idle" (stopped
+	 * or finished).  It is overwritten when a new sync operation is begun.
+	 */
+	char				*last_sync_action;
 	sector_t			curr_resync;	/* last block scheduled */
 	/* As resync requests can complete out of order, we cannot easily track
 	 * how much resync has been completed.  So we occasionally pause until

drivers/md/raid0.c

···
 			       mdname(mddev));
 			return ERR_PTR(-EINVAL);
 		}
+		rdev->sectors = mddev->dev_sectors;
 	}
 
 	/* Set new parameters */

drivers/md/raid1.c

···
 		p = conf->mirrors+mirror;
 		if (!p->rdev) {
 
-			disk_stack_limits(mddev->gendisk, rdev->bdev,
-					  rdev->data_offset << 9);
+			if (mddev->gendisk)
+				disk_stack_limits(mddev->gendisk, rdev->bdev,
+						  rdev->data_offset << 9);
 
 			p->head_position = 0;
 			rdev->raid_disk = mirror;
···
 		clear_bit(Unmerged, &rdev->flags);
 	}
 	md_integrity_add_rdev(rdev, mddev);
-	if (blk_queue_discard(bdev_get_queue(rdev->bdev)))
+	if (mddev->queue && blk_queue_discard(bdev_get_queue(rdev->bdev)))
 		queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, mddev->queue);
 	print_conf(conf);
 	return err;

drivers/md/raid10.c

···
 
 static void allow_barrier(struct r10conf *conf);
 static void lower_barrier(struct r10conf *conf);
-static int enough(struct r10conf *conf, int ignore);
+static int _enough(struct r10conf *conf, int previous, int ignore);
 static sector_t reshape_request(struct mddev *mddev, sector_t sector_nr,
 				int *skipped);
 static void reshape_request_write(struct mddev *mddev, struct r10bio *r10_bio);
···
 		 * than fail the last device.  Here we redefine
 		 * "uptodate" to mean "Don't want to retry"
 		 */
-		unsigned long flags;
-		spin_lock_irqsave(&conf->device_lock, flags);
-		if (!enough(conf, rdev->raid_disk))
+		if (!_enough(conf, test_bit(R10BIO_Previous, &r10_bio->state),
+			     rdev->raid_disk))
 			uptodate = 1;
-		spin_unlock_irqrestore(&conf->device_lock, flags);
 	}
 	if (uptodate) {
 		raid_end_bio_io(r10_bio);
···
  * Don't consider the device numbered 'ignore'
  * as we might be about to remove it.
  */
-static int _enough(struct r10conf *conf, struct geom *geo, int ignore)
+static int _enough(struct r10conf *conf, int previous, int ignore)
 {
 	int first = 0;
+	int has_enough = 0;
+	int disks, ncopies;
+	if (previous) {
+		disks = conf->prev.raid_disks;
+		ncopies = conf->prev.near_copies;
+	} else {
+		disks = conf->geo.raid_disks;
+		ncopies = conf->geo.near_copies;
+	}
 
+	rcu_read_lock();
 	do {
 		int n = conf->copies;
 		int cnt = 0;
 		int this = first;
 		while (n--) {
-			if (conf->mirrors[this].rdev &&
-			    this != ignore)
+			struct md_rdev *rdev;
+			if (this != ignore &&
+			    (rdev = rcu_dereference(conf->mirrors[this].rdev)) &&
+			    test_bit(In_sync, &rdev->flags))
 				cnt++;
-			this = (this+1) % geo->raid_disks;
+			this = (this+1) % disks;
 		}
 		if (cnt == 0)
-			return 0;
-		first = (first + geo->near_copies) % geo->raid_disks;
+			goto out;
+		first = (first + ncopies) % disks;
 	} while (first != 0);
-	return 1;
+	has_enough = 1;
+out:
+	rcu_read_unlock();
+	return has_enough;
 }
 
 static int enough(struct r10conf *conf, int ignore)
 {
-	return _enough(conf, &conf->geo, ignore) &&
-		_enough(conf, &conf->prev, ignore);
+	/* when calling 'enough', both 'prev' and 'geo' must
+	 * be stable.
+	 * This is ensured if ->reconfig_mutex or ->device_lock
+	 * is held.
+	 */
+	return _enough(conf, 0, ignore) &&
+		_enough(conf, 1, ignore);
 }
 
 static void error(struct mddev *mddev, struct md_rdev *rdev)
 {
 	char b[BDEVNAME_SIZE];
 	struct r10conf *conf = mddev->private;
+	unsigned long flags;
 
 	/*
 	 * If it is not operational, then we have already marked it as dead
···
 	 * next level up know.
 	 * else mark the drive as failed
 	 */
+	spin_lock_irqsave(&conf->device_lock, flags);
 	if (test_bit(In_sync, &rdev->flags)
-	    && !enough(conf, rdev->raid_disk))
+	    && !enough(conf, rdev->raid_disk)) {
 		/*
 		 * Don't fail the drive, just return an IO error.
 		 */
-		return;
-	if (test_and_clear_bit(In_sync, &rdev->flags)) {
-		unsigned long flags;
-		spin_lock_irqsave(&conf->device_lock, flags);
-		mddev->degraded++;
 		spin_unlock_irqrestore(&conf->device_lock, flags);
-		/*
+		return;
+	}
+	if (test_and_clear_bit(In_sync, &rdev->flags)) {
+		mddev->degraded++;
+			/*
 		 * if recovery is running, make sure it aborts.
 		 */
 		set_bit(MD_RECOVERY_INTR, &mddev->recovery);
···
 	set_bit(Blocked, &rdev->flags);
 	set_bit(Faulty, &rdev->flags);
 	set_bit(MD_CHANGE_DEVS, &mddev->flags);
+	spin_unlock_irqrestore(&conf->device_lock, flags);
 	printk(KERN_ALERT
 	       "md/raid10:%s: Disk failure on %s, disabling device.\n"
 	       "md/raid10:%s: Operation continuing on %d devices.\n",
···
 		 * very different from resync
 		 */
 		return -EBUSY;
-	if (rdev->saved_raid_disk < 0 && !_enough(conf, &conf->prev, -1))
+	if (rdev->saved_raid_disk < 0 && !_enough(conf, 1, -1))
 		return -EINVAL;
 
 	if (rdev->raid_disk >= 0)
···
 			set_bit(Replacement, &rdev->flags);
 			rdev->raid_disk = mirror;
 			err = 0;
-			disk_stack_limits(mddev->gendisk, rdev->bdev,
-					  rdev->data_offset << 9);
+			if (mddev->gendisk)
+				disk_stack_limits(mddev->gendisk, rdev->bdev,
+						  rdev->data_offset << 9);
 			conf->fullsync = 1;
 			rcu_assign_pointer(p->replacement, rdev);
 			break;
 		}
 
-		disk_stack_limits(mddev->gendisk, rdev->bdev,
-				  rdev->data_offset << 9);
+		if (mddev->gendisk)
+			disk_stack_limits(mddev->gendisk, rdev->bdev,
+					  rdev->data_offset << 9);
 
 		p->head_position = 0;
 		p->recovery_disabled = mddev->recovery_disabled - 1;
···
 	 */
 	if (mddev->bitmap == NULL &&
 	    mddev->recovery_cp == MaxSector &&
+	    mddev->reshape_position == MaxSector &&
+	    !test_bit(MD_RECOVERY_SYNC, &mddev->recovery) &&
 	    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
+	    !test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) &&
 	    conf->fullsync == 0) {
 		*skipped = 1;
-		max_sector = mddev->dev_sectors;
-		if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery) ||
-		    test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
-			max_sector = mddev->resync_max_sectors;
-		return max_sector - sector_nr;
+		return mddev->dev_sectors - sector_nr;
 	}
 
  skipped:
···
 
 	/* FIXME calc properly */
 	conf->mirrors = kzalloc(sizeof(struct raid10_info)*(mddev->raid_disks +
-							    max(0,mddev->delta_disks)),
+							    max(0,-mddev->delta_disks)),
 				GFP_KERNEL);
 	if (!conf->mirrors)
 		goto out;
···
 		    conf->geo.far_offset == 0)
 			goto out_free_conf;
 		if (conf->prev.far_copies != 1 &&
-		    conf->geo.far_offset == 0)
+		    conf->prev.far_offset == 0)
 			goto out_free_conf;
 	}
 

drivers/md/raid5.c

···
 	if (!conf)
 		return -ENODEV;
 
-	if (strict_strtoul(page, 10, &new))
+	if (kstrtoul(page, 10, &new))
 		return -EINVAL;
 	err = raid5_set_cache_size(mddev, new);
 	if (err)
···
 	if (!conf)
 		return -ENODEV;
 
-	if (strict_strtoul(page, 10, &new))
+	if (kstrtoul(page, 10, &new))
 		return -EINVAL;
 	if (new > conf->max_nr_stripes)
 		return -EINVAL;
···
 		return 0; /* nothing to do */
 	if (has_failed(conf))
 		return -EINVAL;
-	if (mddev->delta_disks < 0) {
+	if (mddev->delta_disks < 0 && mddev->reshape_position == MaxSector) {
 		/* We might be able to shrink, but the devices must
 		 * be made bigger first.
 		 * For raid6, 4 is the minimum size.