Merge remote-tracking branch 'md/for-next'

8 files changed, 607 insertions, 67 deletions

diff --git a/drivers/md/dm-raid.c b/drivers/md/dm-raid.c
index a3e115b62186..9d6bf19ce6e8 100644
--- a/drivers/md/dm-raid.c
+++ b/drivers/md/dm-raid.c
@@ -91,15 +91,44 @@ static struct raid_type {
 	{"raid6_nc", "RAID6 (N continue)",		2, 4, 6, ALGORITHM_ROTATING_N_CONTINUE}
 };
 
+static char *raid10_md_layout_to_format(int layout)
+{
+	/*
+	 * Bit 16 and 17 stand for "offset" and "use_far_sets"
+	 * Refer to MD's raid10.c for details
+	 */
+	if ((layout & 0x10000) && (layout & 0x20000))
+		return "offset";
+
+	if ((layout & 0xFF) > 1)
+		return "near";
+
+	return "far";
+}
+
 static unsigned raid10_md_layout_to_copies(int layout)
 {
-	return layout & 0xFF;
+	if ((layout & 0xFF) > 1)
+		return layout & 0xFF;
+	return (layout >> 8) & 0xFF;
 }
 
 static int raid10_format_to_md_layout(char *format, unsigned copies)
 {
-	/* 1 "far" copy, and 'copies' "near" copies */
-	return (1 << 8) | (copies & 0xFF);
+	unsigned n = 1, f = 1;
+
+	if (!strcmp("near", format))
+		n = copies;
+	else
+		f = copies;
+
+	if (!strcmp("offset", format))
+		return 0x30000 | (f << 8) | n;
+
+	if (!strcmp("far", format))
+		return 0x20000 | (f << 8) | n;
+
+	return (f << 8) | n;
 }
 
 static struct raid_type *get_raid_type(char *name)
@@ -352,6 +381,7 @@ static int validate_raid_redundancy(struct raid_set *rs)
 {
 	unsigned i, rebuild_cnt = 0;
 	unsigned rebuilds_per_group, copies, d;
+	unsigned group_size, last_group_start;
 
 	for (i = 0; i < rs->md.raid_disks; i++)
 		if (!test_bit(In_sync, &rs->dev[i].rdev.flags) ||
@@ -379,9 +409,6 @@ static int validate_raid_redundancy(struct raid_set *rs)
 		 * as long as the failed devices occur in different mirror
 		 * groups (i.e. different stripes).
 		 *
-		 * Right now, we only allow for "near" copies.  When other
-		 * formats are added, we will have to check those too.
-		 *
 		 * When checking "near" format, make sure no adjacent devices
 		 * have failed beyond what can be handled.  In addition to the
 		 * simple case where the number of devices is a multiple of the
@@ -391,14 +418,41 @@ static int validate_raid_redundancy(struct raid_set *rs)
 		 *          A    A    B    B    C
 		 *          C    D    D    E    E
 		 */
-		for (i = 0; i < rs->md.raid_disks * copies; i++) {
-			if (!(i % copies))
+		if (!strcmp("near", raid10_md_layout_to_format(rs->md.layout))) {
+			for (i = 0; i < rs->md.raid_disks * copies; i++) {
+				if (!(i % copies))
+					rebuilds_per_group = 0;
+				d = i % rs->md.raid_disks;
+				if ((!rs->dev[d].rdev.sb_page ||
+				     !test_bit(In_sync, &rs->dev[d].rdev.flags)) &&
+				    (++rebuilds_per_group >= copies))
+					goto too_many;
+			}
+			break;
+		}
+
+		/*
+		 * When checking "far" and "offset" formats, we need to ensure
+		 * that the device that holds its copy is not also dead or
+		 * being rebuilt.  (Note that "far" and "offset" formats only
+		 * support two copies right now.  These formats also only ever
+		 * use the 'use_far_sets' variant.)
+		 *
+		 * This check is somewhat complicated by the need to account
+		 * for arrays that are not a multiple of (far) copies.  This
+		 * results in the need to treat the last (potentially larger)
+		 * set differently.
+		 */
+		group_size = (rs->md.raid_disks / copies);
+		last_group_start = (rs->md.raid_disks / group_size) - 1;
+		last_group_start *= group_size;
+		for (i = 0; i < rs->md.raid_disks; i++) {
+			if (!(i % copies) && !(i > last_group_start))
 				rebuilds_per_group = 0;
-			d = i % rs->md.raid_disks;
-			if ((!rs->dev[d].rdev.sb_page ||
-			     !test_bit(In_sync, &rs->dev[d].rdev.flags)) &&
+			if ((!rs->dev[i].rdev.sb_page ||
+			     !test_bit(In_sync, &rs->dev[i].rdev.flags)) &&
 			    (++rebuilds_per_group >= copies))
-				goto too_many;
+					goto too_many;
 		}
 		break;
 	default:
@@ -433,7 +487,7 @@ too_many:
  *
  * RAID10-only options:
  *    [raid10_copies <# copies>]        Number of copies.  (Default: 2)
- *    [raid10_format <near>]            Layout algorithm.  (Default: near)
+ *    [raid10_format <near|far|offset>] Layout algorithm.  (Default: near)
  */
 static int parse_raid_params(struct raid_set *rs, char **argv,
 			     unsigned num_raid_params)
@@ -520,7 +574,9 @@ static int parse_raid_params(struct raid_set *rs, char **argv,
 				rs->ti->error = "'raid10_format' is an invalid parameter for this RAID type";
 				return -EINVAL;
 			}
-			if (strcmp("near", argv[i])) {
+			if (strcmp("near", argv[i]) &&
+			    strcmp("far", argv[i]) &&
+			    strcmp("offset", argv[i])) {
 				rs->ti->error = "Invalid 'raid10_format' value given";
 				return -EINVAL;
 			}
@@ -644,6 +700,15 @@ static int parse_raid_params(struct raid_set *rs, char **argv,
 			return -EINVAL;
 		}
 
+		/*
+		 * If the format is not "near", we only support
+		 * two copies at the moment.
+		 */
+		if (strcmp("near", raid10_format) && (raid10_copies > 2)) {
+			rs->ti->error = "Too many copies for given RAID10 format.";
+			return -EINVAL;
+		}
+
 		/* (Len * #mirrors) / #devices */
 		sectors_per_dev = rs->ti->len * raid10_copies;
 		sector_div(sectors_per_dev, rs->md.raid_disks);
@@ -854,17 +919,30 @@ static int super_init_validation(struct mddev *mddev, struct md_rdev *rdev)
 	/*
 	 * Reshaping is not currently allowed
 	 */
-	if ((le32_to_cpu(sb->level) != mddev->level) ||
-	    (le32_to_cpu(sb->layout) != mddev->layout) ||
-	    (le32_to_cpu(sb->stripe_sectors) != mddev->chunk_sectors)) {
-		DMERR("Reshaping arrays not yet supported.");
+	if (le32_to_cpu(sb->level) != mddev->level) {
+		DMERR("Reshaping arrays not yet supported. (RAID level change)");
+		return -EINVAL;
+	}
+	if (le32_to_cpu(sb->layout) != mddev->layout) {
+		DMERR("Reshaping arrays not yet supported. (RAID layout change)");
+		DMERR("  0x%X vs 0x%X", le32_to_cpu(sb->layout), mddev->layout);
+		DMERR("  Old layout: %s w/ %d copies",
+		      raid10_md_layout_to_format(le32_to_cpu(sb->layout)),
+		      raid10_md_layout_to_copies(le32_to_cpu(sb->layout)));
+		DMERR("  New layout: %s w/ %d copies",
+		      raid10_md_layout_to_format(mddev->layout),
+		      raid10_md_layout_to_copies(mddev->layout));
+		return -EINVAL;
+	}
+	if (le32_to_cpu(sb->stripe_sectors) != mddev->chunk_sectors) {
+		DMERR("Reshaping arrays not yet supported. (stripe sectors change)");
 		return -EINVAL;
 	}
 
 	/* We can only change the number of devices in RAID1 right now */
 	if ((rs->raid_type->level != 1) &&
 	    (le32_to_cpu(sb->num_devices) != mddev->raid_disks)) {
-		DMERR("Reshaping arrays not yet supported.");
+		DMERR("Reshaping arrays not yet supported. (device count change)");
 		return -EINVAL;
 	}
 
@@ -1329,7 +1407,8 @@ static int raid_status(struct dm_target *ti, status_type_t type,
 			       raid10_md_layout_to_copies(rs->md.layout));
 
 		if (rs->print_flags & DMPF_RAID10_FORMAT)
-			DMEMIT(" raid10_format near");
+			DMEMIT(" raid10_format %s",
+			       raid10_md_layout_to_format(rs->md.layout));
 
 		DMEMIT(" %d", rs->md.raid_disks);
 		for (i = 0; i < rs->md.raid_disks; i++) {
@@ -1420,6 +1499,10 @@ static struct target_type raid_target = {
 
 static int __init dm_raid_init(void)
 {
+	DMINFO("Loading target version %u.%u.%u",
+	       raid_target.version[0],
+	       raid_target.version[1],
+	       raid_target.version[2]);
 	return dm_register_target(&raid_target);
 }
 
diff --git a/drivers/md/md.c b/drivers/md/md.c
index 1e634a68541e..8b557d2b6f99 100644
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -607,9 +607,9 @@ static struct mddev * mddev_find(dev_t unit)
 	goto retry;
 }
 
-static inline int mddev_lock(struct mddev * mddev)
+int md_queue_misc_work(struct work_struct *work)
 {
-	return mutex_lock_interruptible(&mddev->reconfig_mutex);
+	return queue_work(md_misc_wq, work);
 }
 
 static inline int mddev_is_locked(struct mddev *mddev)
@@ -624,7 +624,7 @@ static inline int mddev_trylock(struct mddev * mddev)
 
 static struct attribute_group md_redundancy_group;
 
-static void mddev_unlock(struct mddev * mddev)
+void mddev_unlock(struct mddev * mddev)
 {
 	if (mddev->to_remove) {
 		/* These cannot be removed under reconfig_mutex as
@@ -8631,6 +8631,8 @@ EXPORT_SYMBOL(md_register_thread);
 EXPORT_SYMBOL(md_unregister_thread);
 EXPORT_SYMBOL(md_wakeup_thread);
 EXPORT_SYMBOL(md_check_recovery);
+EXPORT_SYMBOL(mddev_unlock);
+EXPORT_SYMBOL(md_queue_misc_work);
 MODULE_LICENSE("GPL");
 MODULE_DESCRIPTION("MD RAID framework");
 MODULE_ALIAS("md");
diff --git a/drivers/md/md.h b/drivers/md/md.h
index eca59c3074ef..2afa1c34d35b 100644
--- a/drivers/md/md.h
+++ b/drivers/md/md.h
@@ -612,4 +612,12 @@ static inline int mddev_check_plugged(struct mddev *mddev)
 	return !!blk_check_plugged(md_unplug, mddev,
 				   sizeof(struct blk_plug_cb));
 }
+
+static inline int mddev_lock(struct mddev * mddev)
+{
+	return mutex_lock_interruptible(&mddev->reconfig_mutex);
+}
+extern void mddev_unlock(struct mddev *mddev);
+extern int md_queue_misc_work(struct work_struct *work);
+
 #endif /* _MD_MD_H */
diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c
index d5bddfc4010e..6e5d5a5f9cb4 100644
--- a/drivers/md/raid1.c
+++ b/drivers/md/raid1.c
@@ -1000,6 +1000,7 @@ static void make_request(struct mddev *mddev, struct bio * bio)
 	const unsigned long do_flush_fua = (bio->bi_rw & (REQ_FLUSH | REQ_FUA));
 	const unsigned long do_discard = (bio->bi_rw
 					  & (REQ_DISCARD | REQ_SECURE));
+	const unsigned long do_same = (bio->bi_rw & REQ_WRITE_SAME);
 	struct md_rdev *blocked_rdev;
 	struct blk_plug_cb *cb;
 	struct raid1_plug_cb *plug = NULL;
@@ -1301,7 +1302,8 @@ read_again:
 				   conf->mirrors[i].rdev->data_offset);
 		mbio->bi_bdev = conf->mirrors[i].rdev->bdev;
 		mbio->bi_end_io	= raid1_end_write_request;
-		mbio->bi_rw = WRITE | do_flush_fua | do_sync | do_discard;
+		mbio->bi_rw =
+			WRITE | do_flush_fua | do_sync | do_discard | do_same;
 		mbio->bi_private = r1_bio;
 
 		atomic_inc(&r1_bio->remaining);
@@ -2818,6 +2820,9 @@ static int run(struct mddev *mddev)
 	if (IS_ERR(conf))
 		return PTR_ERR(conf);
 
+	if (mddev->queue)
+		blk_queue_max_write_same_sectors(mddev->queue,
+						 mddev->chunk_sectors);
 	rdev_for_each(rdev, mddev) {
 		if (!mddev->gendisk)
 			continue;
diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
index 64d48249c03b..61ed150bd0cf 100644
--- a/drivers/md/raid10.c
+++ b/drivers/md/raid10.c
@@ -38,21 +38,36 @@
  *    near_copies (stored in low byte of layout)
  *    far_copies (stored in second byte of layout)
  *    far_offset (stored in bit 16 of layout )
+ *    use_far_sets (stored in bit 17 of layout )
  *
- * The data to be stored is divided into chunks using chunksize.
- * Each device is divided into far_copies sections.
- * In each section, chunks are laid out in a style similar to raid0, but
- * near_copies copies of each chunk is stored (each on a different drive).
- * The starting device for each section is offset near_copies from the starting
- * device of the previous section.
- * Thus they are (near_copies*far_copies) of each chunk, and each is on a different
- * drive.
- * near_copies and far_copies must be at least one, and their product is at most
- * raid_disks.
+ * The data to be stored is divided into chunks using chunksize.  Each device
+ * is divided into far_copies sections.   In each section, chunks are laid out
+ * in a style similar to raid0, but near_copies copies of each chunk is stored
+ * (each on a different drive).  The starting device for each section is offset
+ * near_copies from the starting device of the previous section.  Thus there
+ * are (near_copies * far_copies) of each chunk, and each is on a different
+ * drive.  near_copies and far_copies must be at least one, and their product
+ * is at most raid_disks.
  *
  * If far_offset is true, then the far_copies are handled a bit differently.
- * The copies are still in different stripes, but instead of be very far apart
- * on disk, there are adjacent stripes.
+ * The copies are still in different stripes, but instead of being very far
+ * apart on disk, there are adjacent stripes.
+ *
+ * The far and offset algorithms are handled slightly differently if
+ * 'use_far_sets' is true.  In this case, the array's devices are grouped into
+ * sets that are (near_copies * far_copies) in size.  The far copied stripes
+ * are still shifted by 'near_copies' devices, but this shifting stays confined
+ * to the set rather than the entire array.  This is done to improve the number
+ * of device combinations that can fail without causing the array to fail.
+ * Example 'far' algorithm w/o 'use_far_sets' (each letter represents a chunk
+ * on a device):
+ *    A B C D    A B C D E
+ *      ...         ...
+ *    D A B C    E A B C D
+ * Example 'far' algorithm w/ 'use_far_sets' enabled (sets illustrated w/ []'s):
+ *    [A B] [C D]    [A B] [C D E]
+ *    |...| |...|    |...| | ... |
+ *    [B A] [D C]    [B A] [E C D]
  */
 
 /*
@@ -535,6 +550,13 @@ static void __raid10_find_phys(struct geom *geo, struct r10bio *r10bio)
 	sector_t stripe;
 	int dev;
 	int slot = 0;
+	int last_far_set_start, last_far_set_size;
+
+	last_far_set_start = (geo->raid_disks / geo->far_set_size) - 1;
+	last_far_set_start *= geo->far_set_size;
+
+	last_far_set_size = geo->far_set_size;
+	last_far_set_size += (geo->raid_disks % geo->far_set_size);
 
 	/* now calculate first sector/dev */
 	chunk = r10bio->sector >> geo->chunk_shift;
@@ -551,15 +573,25 @@ static void __raid10_find_phys(struct geom *geo, struct r10bio *r10bio)
 	/* and calculate all the others */
 	for (n = 0; n < geo->near_copies; n++) {
 		int d = dev;
+		int set;
 		sector_t s = sector;
-		r10bio->devs[slot].addr = sector;
 		r10bio->devs[slot].devnum = d;
+		r10bio->devs[slot].addr = s;
 		slot++;
 
 		for (f = 1; f < geo->far_copies; f++) {
+			set = d / geo->far_set_size;
 			d += geo->near_copies;
-			if (d >= geo->raid_disks)
-				d -= geo->raid_disks;
+
+			if ((geo->raid_disks % geo->far_set_size) &&
+			    (d > last_far_set_start)) {
+				d -= last_far_set_start;
+				d %= last_far_set_size;
+				d += last_far_set_start;
+			} else {
+				d %= geo->far_set_size;
+				d += geo->far_set_size * set;
+			}
 			s += geo->stride;
 			r10bio->devs[slot].devnum = d;
 			r10bio->devs[slot].addr = s;
@@ -595,6 +627,20 @@ static sector_t raid10_find_virt(struct r10conf *conf, sector_t sector, int dev)
 	 * or recovery, so reshape isn't happening
 	 */
 	struct geom *geo = &conf->geo;
+	int far_set_start = (dev / geo->far_set_size) * geo->far_set_size;
+	int far_set_size = geo->far_set_size;
+	int last_far_set_start;
+
+	if (geo->raid_disks % geo->far_set_size) {
+		last_far_set_start = (geo->raid_disks / geo->far_set_size) - 1;
+		last_far_set_start *= geo->far_set_size;
+
+		if (dev >= last_far_set_start) {
+			far_set_size = geo->far_set_size;
+			far_set_size += (geo->raid_disks % geo->far_set_size);
+			far_set_start = last_far_set_start;
+		}
+	}
 
 	offset = sector & geo->chunk_mask;
 	if (geo->far_offset) {
@@ -602,13 +648,13 @@ static sector_t raid10_find_virt(struct r10conf *conf, sector_t sector, int dev)
 		chunk = sector >> geo->chunk_shift;
 		fc = sector_div(chunk, geo->far_copies);
 		dev -= fc * geo->near_copies;
-		if (dev < 0)
-			dev += geo->raid_disks;
+		if (dev < far_set_start)
+			dev += far_set_size;
 	} else {
 		while (sector >= geo->stride) {
 			sector -= geo->stride;
-			if (dev < geo->near_copies)
-				dev += geo->raid_disks - geo->near_copies;
+			if (dev < (geo->near_copies + far_set_start))
+				dev += far_set_size - geo->near_copies;
 			else
 				dev -= geo->near_copies;
 		}
@@ -1105,6 +1151,7 @@ static void make_request(struct mddev *mddev, struct bio * bio)
 	const unsigned long do_fua = (bio->bi_rw & REQ_FUA);
 	const unsigned long do_discard = (bio->bi_rw
 					  & (REQ_DISCARD | REQ_SECURE));
+	const unsigned long do_same = (bio->bi_rw & REQ_WRITE_SAME);
 	unsigned long flags;
 	struct md_rdev *blocked_rdev;
 	struct blk_plug_cb *cb;
@@ -1460,7 +1507,8 @@ retry_write:
 							      rdev));
 			mbio->bi_bdev = rdev->bdev;
 			mbio->bi_end_io	= raid10_end_write_request;
-			mbio->bi_rw = WRITE | do_sync | do_fua | do_discard;
+			mbio->bi_rw =
+				WRITE | do_sync | do_fua | do_discard | do_same;
 			mbio->bi_private = r10_bio;
 
 			atomic_inc(&r10_bio->remaining);
@@ -1502,7 +1550,8 @@ retry_write:
 						   r10_bio, rdev));
 			mbio->bi_bdev = rdev->bdev;
 			mbio->bi_end_io	= raid10_end_write_request;
-			mbio->bi_rw = WRITE | do_sync | do_fua | do_discard;
+			mbio->bi_rw =
+				WRITE | do_sync | do_fua | do_discard | do_same;
 			mbio->bi_private = r10_bio;
 
 			atomic_inc(&r10_bio->remaining);
@@ -3436,7 +3485,7 @@ static int setup_geo(struct geom *geo, struct mddev *mddev, enum geo_type new)
 		disks = mddev->raid_disks + mddev->delta_disks;
 		break;
 	}
-	if (layout >> 17)
+	if (layout >> 18)
 		return -1;
 	if (chunk < (PAGE_SIZE >> 9) ||
 	    !is_power_of_2(chunk))
@@ -3448,6 +3497,7 @@ static int setup_geo(struct geom *geo, struct mddev *mddev, enum geo_type new)
 	geo->near_copies = nc;
 	geo->far_copies = fc;
 	geo->far_offset = fo;
+	geo->far_set_size = (layout & (1<<17)) ? disks / fc : disks;
 	geo->chunk_mask = chunk - 1;
 	geo->chunk_shift = ffz(~chunk);
 	return nc*fc;
@@ -3569,6 +3619,8 @@ static int run(struct mddev *mddev)
 	if (mddev->queue) {
 		blk_queue_max_discard_sectors(mddev->queue,
 					      mddev->chunk_sectors);
+		blk_queue_max_write_same_sectors(mddev->queue,
+						 mddev->chunk_sectors);
 		blk_queue_io_min(mddev->queue, chunk_size);
 		if (conf->geo.raid_disks % conf->geo.near_copies)
 			blk_queue_io_opt(mddev->queue, chunk_size * conf->geo.raid_disks);
diff --git a/drivers/md/raid10.h b/drivers/md/raid10.h
index 1054cf602345..157d69e83ff4 100644
--- a/drivers/md/raid10.h
+++ b/drivers/md/raid10.h
@@ -33,6 +33,11 @@ struct r10conf {
 					       * far_offset, in which case it is
 					       * 1 stripe.
 					       */
+		int             far_set_size; /* The number of devices in a set,
+					       * where a 'set' are devices that
+					       * contain far/offset copies of
+					       * each other.
+					       */
 		int		chunk_shift; /* shift from chunks to sectors */
 		sector_t	chunk_mask;
 	} prev, geo;
diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index 9ab506df42da..77c4a585fb82 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -198,6 +198,21 @@ static int stripe_operations_active(struct stripe_head *sh)
 	       test_bit(STRIPE_COMPUTE_RUN, &sh->state);
 }
 
+static void raid5_wakeup_stripe_thread(struct stripe_head *sh)
+{
+	struct r5conf *conf = sh->raid_conf;
+	struct raid5_percpu *percpu;
+	int i, orphaned = 1;
+
+	percpu = per_cpu_ptr(conf->percpu, sh->cpu);
+	for_each_cpu(i, &percpu->handle_threads) {
+		md_wakeup_thread(conf->aux_threads[i]->thread);
+		orphaned = 0;
+	}
+	if (orphaned)
+		md_wakeup_thread(conf->mddev->thread);
+}
+
 static void do_release_stripe(struct r5conf *conf, struct stripe_head *sh)
 {
 	BUG_ON(!list_empty(&sh->lru));
@@ -210,9 +225,19 @@ static void do_release_stripe(struct r5conf *conf, struct stripe_head *sh)
 			   sh->bm_seq - conf->seq_write > 0)
 			list_add_tail(&sh->lru, &conf->bitmap_list);
 		else {
+			int cpu = sh->cpu;
+			struct raid5_percpu *percpu;
+			if (!cpu_online(cpu)) {
+				cpu = cpumask_any(cpu_online_mask);
+				sh->cpu = cpu;
+			}
+			percpu = per_cpu_ptr(conf->percpu, cpu);
+
 			clear_bit(STRIPE_DELAYED, &sh->state);
 			clear_bit(STRIPE_BIT_DELAY, &sh->state);
-			list_add_tail(&sh->lru, &conf->handle_list);
+			list_add_tail(&sh->lru, &percpu->handle_list);
+			raid5_wakeup_stripe_thread(sh);
+			return;
 		}
 		md_wakeup_thread(conf->mddev->thread);
 	} else {
@@ -357,6 +382,7 @@ static void init_stripe(struct stripe_head *sh, sector_t sector, int previous)
 		raid5_build_block(sh, i, previous);
 	}
 	insert_hash(conf, sh);
+	sh->cpu = smp_processor_id();
 }
 
 static struct stripe_head *__find_stripe(struct r5conf *conf, sector_t sector,
@@ -3759,12 +3785,19 @@ static void raid5_activate_delayed(struct r5conf *conf)
 		while (!list_empty(&conf->delayed_list)) {
 			struct list_head *l = conf->delayed_list.next;
 			struct stripe_head *sh;
+			int cpu;
 			sh = list_entry(l, struct stripe_head, lru);
 			list_del_init(l);
 			clear_bit(STRIPE_DELAYED, &sh->state);
 			if (!test_and_set_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
 				atomic_inc(&conf->preread_active_stripes);
 			list_add_tail(&sh->lru, &conf->hold_list);
+			cpu = sh->cpu;
+			if (!cpu_online(cpu)) {
+				cpu = cpumask_any(cpu_online_mask);
+				sh->cpu = cpu;
+			}
+			raid5_wakeup_stripe_thread(sh);
 		}
 	}
 }
@@ -4041,18 +4074,29 @@ static int chunk_aligned_read(struct mddev *mddev, struct bio * raid_bio)
  * head of the hold_list has changed, i.e. the head was promoted to the
  * handle_list.
  */
-static struct stripe_head *__get_priority_stripe(struct r5conf *conf)
-{
-	struct stripe_head *sh;
-
+static struct stripe_head *__get_priority_stripe(struct r5conf *conf,
+	cpumask_t *mask)
+{
+	struct stripe_head *sh = NULL, *tmp;
+	struct list_head *handle_list = NULL;
+	int cpu;
+
+	/* Should we take action to avoid starvation of latter CPUs ? */
+	for_each_cpu(cpu, mask) {
+		struct raid5_percpu *percpu = per_cpu_ptr(conf->percpu, cpu);
+		if (!list_empty(&percpu->handle_list)) {
+			handle_list = &percpu->handle_list;
+			break;
+		}
+	}
 	pr_debug("%s: handle: %s hold: %s full_writes: %d bypass_count: %d\n",
 		  __func__,
-		  list_empty(&conf->handle_list) ? "empty" : "busy",
+		  !handle_list ? "empty" : "busy",
 		  list_empty(&conf->hold_list) ? "empty" : "busy",
 		  atomic_read(&conf->pending_full_writes), conf->bypass_count);
 
-	if (!list_empty(&conf->handle_list)) {
-		sh = list_entry(conf->handle_list.next, typeof(*sh), lru);
+	if (handle_list) {
+		sh = list_entry(handle_list->next, typeof(*sh), lru);
 
 		if (list_empty(&conf->hold_list))
 			conf->bypass_count = 0;
@@ -4070,12 +4114,23 @@ static struct stripe_head *__get_priority_stripe(struct r5conf *conf)
 		   ((conf->bypass_threshold &&
 		     conf->bypass_count > conf->bypass_threshold) ||
 		    atomic_read(&conf->pending_full_writes) == 0)) {
-		sh = list_entry(conf->hold_list.next,
-				typeof(*sh), lru);
-		conf->bypass_count -= conf->bypass_threshold;
-		if (conf->bypass_count < 0)
-			conf->bypass_count = 0;
-	} else
+
+		list_for_each_entry(tmp, &conf->hold_list,  lru) {
+			if (cpumask_test_cpu(tmp->cpu, mask) ||
+			    !cpu_online(tmp->cpu)) {
+				sh = tmp;
+				break;
+			}
+		}
+
+		if (sh) {
+			conf->bypass_count -= conf->bypass_threshold;
+			if (conf->bypass_count < 0)
+				conf->bypass_count = 0;
+		}
+	}
+
+	if (!sh)
 		return NULL;
 
 	list_del_init(&sh->lru);
@@ -4757,13 +4812,13 @@ static int  retry_aligned_read(struct r5conf *conf, struct bio *raid_bio)
 }
 
 #define MAX_STRIPE_BATCH 8
-static int handle_active_stripes(struct r5conf *conf)
+static int handle_active_stripes(struct r5conf *conf, cpumask_t *mask)
 {
 	struct stripe_head *batch[MAX_STRIPE_BATCH], *sh;
 	int i, batch_size = 0;
 
 	while (batch_size < MAX_STRIPE_BATCH &&
-			(sh = __get_priority_stripe(conf)) != NULL)
+			(sh = __get_priority_stripe(conf, mask)) != NULL)
 		batch[batch_size++] = sh;
 
 	if (batch_size == 0)
@@ -4781,6 +4836,35 @@ static int handle_active_stripes(struct r5conf *conf)
 	return batch_size;
 }
 
+static void raid5auxd(struct md_thread *thread)
+{
+	struct mddev *mddev = thread->mddev;
+	struct r5conf *conf = mddev->private;
+	struct blk_plug plug;
+	int handled;
+	struct raid5_auxth *auxth = thread->private;
+
+	pr_debug("+++ raid5auxd active\n");
+
+	blk_start_plug(&plug);
+	handled = 0;
+	spin_lock_irq(&conf->device_lock);
+	while (1) {
+		int batch_size;
+
+		batch_size = handle_active_stripes(conf, &auxth->work_mask);
+		if (!batch_size)
+			break;
+		handled += batch_size;
+	}
+	pr_debug("%d stripes handled\n", handled);
+
+	spin_unlock_irq(&conf->device_lock);
+	blk_finish_plug(&plug);
+
+	pr_debug("--- raid5auxd inactive\n");
+}
+
 /*
  * This is our raid5 kernel thread.
  *
@@ -4828,7 +4912,7 @@ static void raid5d(struct md_thread *thread)
 			handled++;
 		}
 
-		batch_size = handle_active_stripes(conf);
+		batch_size = handle_active_stripes(conf, &conf->work_mask);
 		if (!batch_size)
 			break;
 		handled += batch_size;
@@ -4957,10 +5041,270 @@ stripe_cache_active_show(struct mddev *mddev, char *page)
 static struct md_sysfs_entry
 raid5_stripecache_active = __ATTR_RO(stripe_cache_active);
 
+static void raid5_update_threads_handle_mask(struct mddev *mddev)
+{
+	int cpu, i;
+	struct raid5_percpu *percpu;
+	struct r5conf *conf = mddev->private;
+
+	for_each_online_cpu(cpu) {
+		percpu = per_cpu_ptr(conf->percpu, cpu);
+		cpumask_clear(&percpu->handle_threads);
+	}
+	cpumask_copy(&conf->work_mask, cpu_online_mask);
+
+	for (i = 0; i < conf->aux_thread_num; i++) {
+		cpumask_t *work_mask = &conf->aux_threads[i]->work_mask;
+		for_each_cpu(cpu, work_mask) {
+			percpu = per_cpu_ptr(conf->percpu, cpu);
+			cpumask_set_cpu(i, &percpu->handle_threads);
+		}
+		cpumask_andnot(&conf->work_mask, &conf->work_mask,
+				work_mask);
+	}
+}
+
+struct raid5_auxth_sysfs {
+	struct attribute attr;
+	ssize_t (*show)(struct mddev *, struct raid5_auxth *, char *);
+	ssize_t (*store)(struct mddev *, struct raid5_auxth *,
+		const char *, size_t);
+};
+
+static ssize_t raid5_show_thread_cpulist(struct mddev *mddev,
+	struct raid5_auxth *thread, char *page)
+{
+	if (!mddev->private)
+		return 0;
+	return cpulist_scnprintf(page, PAGE_SIZE, &thread->work_mask);
+}
+
+static ssize_t
+raid5_store_thread_cpulist(struct mddev *mddev, struct raid5_auxth *thread,
+	const char *page, size_t len)
+{
+	struct r5conf *conf = mddev->private;
+	cpumask_var_t mask;
+
+	if (!conf)
+		return -ENODEV;
+	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
+		return -ENOMEM;
+
+	if (cpulist_parse(page, mask)) {
+		free_cpumask_var(mask);
+		return -EINVAL;
+	}
+
+	get_online_cpus();
+	spin_lock_irq(&conf->device_lock);
+	cpumask_copy(&thread->work_mask, mask);
+	raid5_update_threads_handle_mask(mddev);
+	spin_unlock_irq(&conf->device_lock);
+	put_online_cpus();
+	set_cpus_allowed_ptr(thread->thread->tsk, mask);
+
+	free_cpumask_var(mask);
+	return len;
+}
+
+static struct raid5_auxth_sysfs thread_cpulist =
+__ATTR(cpulist, S_IRUGO|S_IWUSR, raid5_show_thread_cpulist,
+	raid5_store_thread_cpulist);
+
+static struct attribute *auxth_attrs[] = {
+	&thread_cpulist.attr,
+	NULL,
+};
+
+static ssize_t
+raid5_auxth_attr_show(struct kobject *kobj, struct attribute *attr, char *page)
+{
+	struct raid5_auxth_sysfs *entry = container_of(attr,
+		struct raid5_auxth_sysfs, attr);
+	struct raid5_auxth *thread = container_of(kobj,
+		struct raid5_auxth, kobj);
+	struct mddev *mddev = thread->thread->mddev;
+	ssize_t ret;
+
+	if (!entry->show)
+		return -EIO;
+	mddev_lock(mddev);
+	ret = entry->show(mddev, thread, page);
+	mddev_unlock(mddev);
+	return ret;
+}
+
+static ssize_t
+raid5_auxth_attr_store(struct kobject *kobj, struct attribute *attr,
+	      const char *page, size_t length)
+{
+	struct raid5_auxth_sysfs *entry = container_of(attr,
+		struct raid5_auxth_sysfs, attr);
+	struct raid5_auxth *thread = container_of(kobj,
+		struct raid5_auxth, kobj);
+	struct mddev *mddev = thread->thread->mddev;
+	ssize_t ret;
+
+	if (!entry->store)
+		return -EIO;
+	if (!capable(CAP_SYS_ADMIN))
+		return -EACCES;
+	mddev_lock(mddev);
+	ret = entry->store(mddev, thread, page, length);
+	mddev_unlock(mddev);
+	return ret;
+}
+
+static void raid5_auxth_release(struct kobject *kobj)
+{
+	struct raid5_auxth *thread = container_of(kobj,
+		struct raid5_auxth, kobj);
+	kfree(thread);
+}
+
+static const struct sysfs_ops raid5_auxth_sysfsops = {
+	.show = raid5_auxth_attr_show,
+	.store = raid5_auxth_attr_store,
+};
+static struct kobj_type raid5_auxth_ktype = {
+	.release = raid5_auxth_release,
+	.sysfs_ops = &raid5_auxth_sysfsops,
+	.default_attrs = auxth_attrs,
+};
+
+static ssize_t
+raid5_show_auxthread_number(struct mddev *mddev, char *page)
+{
+	struct r5conf *conf = mddev->private;
+	if (conf)
+		return sprintf(page, "%d\n", conf->aux_thread_num);
+	else
+		return 0;
+}
+
+static void raid5_auxth_delete(struct work_struct *ws)
+{
+	struct raid5_auxth *thread = container_of(ws, struct raid5_auxth,
+		del_work);
+
+	kobject_del(&thread->kobj);
+	kobject_put(&thread->kobj);
+}
+
+static void __free_aux_thread(struct mddev *mddev, struct raid5_auxth *thread)
+{
+	md_unregister_thread(&thread->thread);
+	INIT_WORK(&thread->del_work, raid5_auxth_delete);
+	kobject_get(&thread->kobj);
+	md_queue_misc_work(&thread->del_work);
+}
+
+static struct raid5_auxth *__create_aux_thread(struct mddev *mddev, int i)
+{
+	struct raid5_auxth *thread;
+	char name[10];
+
+	thread = kzalloc(sizeof(*thread), GFP_KERNEL);
+	if (!thread)
+		return NULL;
+	snprintf(name, 10, "aux%d", i);
+	thread->thread = md_register_thread(raid5auxd, mddev, name);
+	if (!thread->thread) {
+		kfree(thread);
+		return NULL;
+	}
+	thread->thread->private = thread;
+
+	cpumask_copy(&thread->work_mask, cpu_online_mask);
+
+	if (kobject_init_and_add(&thread->kobj, &raid5_auxth_ktype,
+	    &mddev->kobj, "auxth%d", i)) {
+		md_unregister_thread(&thread->thread);
+		kfree(thread);
+		return NULL;
+	}
+	return thread;
+}
+
+static ssize_t
+raid5_store_auxthread_number(struct mddev *mddev, const char *page, size_t len)
+{
+	struct r5conf *conf = mddev->private;
+	unsigned long new;
+	int i;
+	struct raid5_auxth **threads;
+
+	if (len >= PAGE_SIZE)
+		return -EINVAL;
+	if (!conf)
+		return -ENODEV;
+
+	if (kstrtoul(page, 10, &new))
+		return -EINVAL;
+
+	if (new == conf->aux_thread_num)
+		return len;
+
+	/* There is no point creating more threads than cpu number */
+	if (new > num_online_cpus())
+		return -EINVAL;
+
+	if (new > conf->aux_thread_num) {
+		threads = kzalloc(sizeof(struct raid5_auxth *) * new,
+				GFP_KERNEL);
+		if (!threads)
+			return -ENOMEM;
+
+		i = conf->aux_thread_num;
+		while (i < new) {
+			threads[i] = __create_aux_thread(mddev, i);
+			if (!threads[i])
+				goto error;
+
+			i++;
+		}
+		memcpy(threads, conf->aux_threads,
+			sizeof(struct raid5_auxth *) * conf->aux_thread_num);
+		get_online_cpus();
+		spin_lock_irq(&conf->device_lock);
+		kfree(conf->aux_threads);
+		conf->aux_threads = threads;
+		conf->aux_thread_num = new;
+		raid5_update_threads_handle_mask(mddev);
+		spin_unlock_irq(&conf->device_lock);
+		put_online_cpus();
+	} else {
+		int old = conf->aux_thread_num;
+
+		get_online_cpus();
+		spin_lock_irq(&conf->device_lock);
+		conf->aux_thread_num = new;
+		raid5_update_threads_handle_mask(mddev);
+		spin_unlock_irq(&conf->device_lock);
+		put_online_cpus();
+		for (i = new; i < old; i++)
+			__free_aux_thread(mddev, conf->aux_threads[i]);
+	}
+
+	return len;
+error:
+	while (--i >= conf->aux_thread_num)
+		__free_aux_thread(mddev, threads[i]);
+	kfree(threads);
+	return -ENOMEM;
+}
+
+static struct md_sysfs_entry
+raid5_auxthread_number = __ATTR(auxthread_number, S_IRUGO|S_IWUSR,
+				raid5_show_auxthread_number,
+				raid5_store_auxthread_number);
+
 static struct attribute *raid5_attrs[] =  {
 	&raid5_stripecache_size.attr,
 	&raid5_stripecache_active.attr,
 	&raid5_preread_bypass_threshold.attr,
+	&raid5_auxthread_number.attr,
 	NULL,
 };
 static struct attribute_group raid5_attrs_group = {
@@ -5008,6 +5352,7 @@ static void raid5_free_percpu(struct r5conf *conf)
 
 static void free_conf(struct r5conf *conf)
 {
+	kfree(conf->aux_threads);
 	shrink_stripes(conf);
 	raid5_free_percpu(conf);
 	kfree(conf->disks);
@@ -5020,7 +5365,7 @@ static int raid456_cpu_notify(struct notifier_block *nfb, unsigned long action,
 			      void *hcpu)
 {
 	struct r5conf *conf = container_of(nfb, struct r5conf, cpu_notify);
-	long cpu = (long)hcpu;
+	long cpu = (long)hcpu, anycpu;
 	struct raid5_percpu *percpu = per_cpu_ptr(conf->percpu, cpu);
 
 	switch (action) {
@@ -5039,9 +5384,17 @@ static int raid456_cpu_notify(struct notifier_block *nfb, unsigned long action,
 			       __func__, cpu);
 			return notifier_from_errno(-ENOMEM);
 		}
+		INIT_LIST_HEAD(&(percpu->handle_list));
+		cpumask_clear(&percpu->handle_threads);
 		break;
 	case CPU_DEAD:
 	case CPU_DEAD_FROZEN:
+		spin_lock_irq(&conf->device_lock);
+		anycpu = cpumask_any(cpu_online_mask);
+		list_splice_tail_init(&percpu->handle_list,
+			&per_cpu_ptr(conf->percpu, anycpu)->handle_list);
+		spin_unlock_irq(&conf->device_lock);
+
 		safe_put_page(percpu->spare_page);
 		kfree(percpu->scribble);
 		percpu->spare_page = NULL;
@@ -5050,6 +5403,10 @@ static int raid456_cpu_notify(struct notifier_block *nfb, unsigned long action,
 	default:
 		break;
 	}
+
+	spin_lock_irq(&conf->device_lock);
+	raid5_update_threads_handle_mask(conf->mddev);
+	spin_unlock_irq(&conf->device_lock);
 	return NOTIFY_OK;
 }
 #endif
@@ -5070,20 +5427,24 @@ static int raid5_alloc_percpu(struct r5conf *conf)
 	get_online_cpus();
 	err = 0;
 	for_each_present_cpu(cpu) {
+		struct raid5_percpu *percpu = per_cpu_ptr(conf->percpu, cpu);
+
 		if (conf->level == 6) {
 			spare_page = alloc_page(GFP_KERNEL);
 			if (!spare_page) {
 				err = -ENOMEM;
 				break;
 			}
-			per_cpu_ptr(conf->percpu, cpu)->spare_page = spare_page;
+			percpu->spare_page = spare_page;
 		}
 		scribble = kmalloc(conf->scribble_len, GFP_KERNEL);
 		if (!scribble) {
 			err = -ENOMEM;
 			break;
 		}
-		per_cpu_ptr(conf->percpu, cpu)->scribble = scribble;
+		percpu->scribble = scribble;
+		INIT_LIST_HEAD(&percpu->handle_list);
+		cpumask_clear(&percpu->handle_threads);
 	}
 #ifdef CONFIG_HOTPLUG_CPU
 	conf->cpu_notify.notifier_call = raid456_cpu_notify;
@@ -5139,7 +5500,6 @@ static struct r5conf *setup_conf(struct mddev *mddev)
 	spin_lock_init(&conf->device_lock);
 	init_waitqueue_head(&conf->wait_for_stripe);
 	init_waitqueue_head(&conf->wait_for_overlap);
-	INIT_LIST_HEAD(&conf->handle_list);
 	INIT_LIST_HEAD(&conf->hold_list);
 	INIT_LIST_HEAD(&conf->delayed_list);
 	INIT_LIST_HEAD(&conf->bitmap_list);
@@ -5150,6 +5510,8 @@ static struct r5conf *setup_conf(struct mddev *mddev)
 	conf->bypass_threshold = BYPASS_THRESHOLD;
 	conf->recovery_disabled = mddev->recovery_disabled - 1;
 
+	cpumask_copy(&conf->work_mask, cpu_online_mask);
+
 	conf->raid_disks = mddev->raid_disks;
 	if (mddev->reshape_position == MaxSector)
 		conf->previous_raid_disks = mddev->raid_disks;
@@ -5606,6 +5968,10 @@ abort:
 static int stop(struct mddev *mddev)
 {
 	struct r5conf *conf = mddev->private;
+	int i;
+
+	for (i = 0; i < conf->aux_thread_num; i++)
+		__free_aux_thread(mddev, conf->aux_threads[i]);
 
 	md_unregister_thread(&mddev->thread);
 	if (mddev->queue)
diff --git a/drivers/md/raid5.h b/drivers/md/raid5.h
index 18b2c4a8a1fd..2afd8358556b 100644
--- a/drivers/md/raid5.h
+++ b/drivers/md/raid5.h
@@ -211,6 +211,7 @@ struct stripe_head {
 	enum check_states	check_state;
 	enum reconstruct_states reconstruct_state;
 	spinlock_t		stripe_lock;
+	int			cpu;
 	/**
 	 * struct stripe_operations
 	 * @target - STRIPE_OP_COMPUTE_BLK target
@@ -365,6 +366,14 @@ struct disk_info {
 	struct md_rdev	*rdev, *replacement;
 };
 
+struct raid5_auxth {
+	struct md_thread	*thread;
+	/* which CPUs should the auxiliary thread handle stripes from */
+	cpumask_t		work_mask;
+	struct kobject		kobj;
+	struct work_struct	del_work;
+};
+
 struct r5conf {
 	struct hlist_head	*stripe_hashtbl;
 	struct mddev		*mddev;
@@ -433,6 +442,12 @@ struct r5conf {
 					      * lists and performing address
 					      * conversions
 					      */
+		struct list_head handle_list;
+		cpumask_t	handle_threads; /* Which threads can the CPU's
+						 * stripes be handled. It really
+						 * is a bitmap to aux_threads[],
+						 * but has max bits NR_CPUS
+						 */
 	} __percpu *percpu;
 	size_t			scribble_len; /* size of scribble region must be
 					       * associated with conf to handle
@@ -460,6 +475,10 @@ struct r5conf {
 	 * the new thread here until we fully activate the array.
 	 */
 	struct md_thread	*thread;
+	int			aux_thread_num;
+	struct raid5_auxth	**aux_threads;
+	/* which CPUs should raid5d thread handle stripes from */
+	cpumask_t		work_mask;
 };
 
 /*