1 files changed, 1128 insertions, 153 deletions

diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
index 3f91c2e1dfe..987db37cb87 100644
--- a/drivers/md/raid10.c
+++ b/drivers/md/raid10.c
@@ -24,6 +24,7 @@
 #include <linux/module.h>
 #include <linux/seq_file.h>
 #include <linux/ratelimit.h>
+#include <linux/kthread.h>
 #include "md.h"
 #include "raid10.h"
 #include "raid0.h"
@@ -68,6 +69,11 @@ static int max_queued_requests = 1024;
 static void allow_barrier(struct r10conf *conf);
 static void lower_barrier(struct r10conf *conf);
 static int enough(struct r10conf *conf, 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);
+static void end_reshape_write(struct bio *bio, int error);
+static void end_reshape(struct r10conf *conf);
 
 static void * r10bio_pool_alloc(gfp_t gfp_flags, void *data)
 {
@@ -112,7 +118,8 @@ static void * r10buf_pool_alloc(gfp_t gfp_flags, void *data)
 	if (!r10_bio)
 		return NULL;
 
-	if (test_bit(MD_RECOVERY_SYNC, &conf->mddev->recovery))
+	if (test_bit(MD_RECOVERY_SYNC, &conf->mddev->recovery) ||
+	    test_bit(MD_RECOVERY_RESHAPE, &conf->mddev->recovery))
 		nalloc = conf->copies; /* resync */
 	else
 		nalloc = 2; /* recovery */
@@ -140,9 +147,10 @@ static void * r10buf_pool_alloc(gfp_t gfp_flags, void *data)
 		struct bio *rbio = r10_bio->devs[j].repl_bio;
 		bio = r10_bio->devs[j].bio;
 		for (i = 0; i < RESYNC_PAGES; i++) {
-			if (j == 1 && !test_bit(MD_RECOVERY_SYNC,
-						&conf->mddev->recovery)) {
-				/* we can share bv_page's during recovery */
+			if (j > 0 && !test_bit(MD_RECOVERY_SYNC,
+					       &conf->mddev->recovery)) {
+				/* we can share bv_page's during recovery
+				 * and reshape */
 				struct bio *rbio = r10_bio->devs[0].bio;
 				page = rbio->bi_io_vec[i].bv_page;
 				get_page(page);
@@ -165,10 +173,11 @@ out_free_pages:
 	while (j--)
 		for (i = 0; i < RESYNC_PAGES ; i++)
 			safe_put_page(r10_bio->devs[j].bio->bi_io_vec[i].bv_page);
-	j = -1;
+	j = 0;
 out_free_bio:
-	while (++j < nalloc) {
-		bio_put(r10_bio->devs[j].bio);
+	for ( ; j < nalloc; j++) {
+		if (r10_bio->devs[j].bio)
+			bio_put(r10_bio->devs[j].bio);
 		if (r10_bio->devs[j].repl_bio)
 			bio_put(r10_bio->devs[j].repl_bio);
 	}
@@ -504,79 +513,96 @@ static void raid10_end_write_request(struct bio *bio, int error)
  * sector offset to a virtual address
  */
 
-static void raid10_find_phys(struct r10conf *conf, struct r10bio *r10bio)
+static void __raid10_find_phys(struct geom *geo, struct r10bio *r10bio)
 {
 	int n,f;
 	sector_t sector;
 	sector_t chunk;
 	sector_t stripe;
 	int dev;
-
 	int slot = 0;
 
 	/* now calculate first sector/dev */
-	chunk = r10bio->sector >> conf->chunk_shift;
-	sector = r10bio->sector & conf->chunk_mask;
+	chunk = r10bio->sector >> geo->chunk_shift;
+	sector = r10bio->sector & geo->chunk_mask;
 
-	chunk *= conf->near_copies;
+	chunk *= geo->near_copies;
 	stripe = chunk;
-	dev = sector_div(stripe, conf->raid_disks);
-	if (conf->far_offset)
-		stripe *= conf->far_copies;
+	dev = sector_div(stripe, geo->raid_disks);
+	if (geo->far_offset)
+		stripe *= geo->far_copies;
 
-	sector += stripe << conf->chunk_shift;
+	sector += stripe << geo->chunk_shift;
 
 	/* and calculate all the others */
-	for (n=0; n < conf->near_copies; n++) {
+	for (n = 0; n < geo->near_copies; n++) {
 		int d = dev;
 		sector_t s = sector;
 		r10bio->devs[slot].addr = sector;
 		r10bio->devs[slot].devnum = d;
 		slot++;
 
-		for (f = 1; f < conf->far_copies; f++) {
-			d += conf->near_copies;
-			if (d >= conf->raid_disks)
-				d -= conf->raid_disks;
-			s += conf->stride;
+		for (f = 1; f < geo->far_copies; f++) {
+			d += geo->near_copies;
+			if (d >= geo->raid_disks)
+				d -= geo->raid_disks;
+			s += geo->stride;
 			r10bio->devs[slot].devnum = d;
 			r10bio->devs[slot].addr = s;
 			slot++;
 		}
 		dev++;
-		if (dev >= conf->raid_disks) {
+		if (dev >= geo->raid_disks) {
 			dev = 0;
-			sector += (conf->chunk_mask + 1);
+			sector += (geo->chunk_mask + 1);
 		}
 	}
-	BUG_ON(slot != conf->copies);
+}
+
+static void raid10_find_phys(struct r10conf *conf, struct r10bio *r10bio)
+{
+	struct geom *geo = &conf->geo;
+
+	if (conf->reshape_progress != MaxSector &&
+	    ((r10bio->sector >= conf->reshape_progress) !=
+	     conf->mddev->reshape_backwards)) {
+		set_bit(R10BIO_Previous, &r10bio->state);
+		geo = &conf->prev;
+	} else
+		clear_bit(R10BIO_Previous, &r10bio->state);
+
+	__raid10_find_phys(geo, r10bio);
 }
 
 static sector_t raid10_find_virt(struct r10conf *conf, sector_t sector, int dev)
 {
 	sector_t offset, chunk, vchunk;
+	/* Never use conf->prev as this is only called during resync
+	 * or recovery, so reshape isn't happening
+	 */
+	struct geom *geo = &conf->geo;
 
-	offset = sector & conf->chunk_mask;
-	if (conf->far_offset) {
+	offset = sector & geo->chunk_mask;
+	if (geo->far_offset) {
 		int fc;
-		chunk = sector >> conf->chunk_shift;
-		fc = sector_div(chunk, conf->far_copies);
-		dev -= fc * conf->near_copies;
+		chunk = sector >> geo->chunk_shift;
+		fc = sector_div(chunk, geo->far_copies);
+		dev -= fc * geo->near_copies;
 		if (dev < 0)
-			dev += conf->raid_disks;
+			dev += geo->raid_disks;
 	} else {
-		while (sector >= conf->stride) {
-			sector -= conf->stride;
-			if (dev < conf->near_copies)
-				dev += conf->raid_disks - conf->near_copies;
+		while (sector >= geo->stride) {
+			sector -= geo->stride;
+			if (dev < geo->near_copies)
+				dev += geo->raid_disks - geo->near_copies;
 			else
-				dev -= conf->near_copies;
+				dev -= geo->near_copies;
 		}
-		chunk = sector >> conf->chunk_shift;
+		chunk = sector >> geo->chunk_shift;
 	}
-	vchunk = chunk * conf->raid_disks + dev;
-	sector_div(vchunk, conf->near_copies);
-	return (vchunk << conf->chunk_shift) + offset;
+	vchunk = chunk * geo->raid_disks + dev;
+	sector_div(vchunk, geo->near_copies);
+	return (vchunk << geo->chunk_shift) + offset;
 }
 
 /**
@@ -597,10 +623,17 @@ static int raid10_mergeable_bvec(struct request_queue *q,
 	struct r10conf *conf = mddev->private;
 	sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
 	int max;
-	unsigned int chunk_sectors = mddev->chunk_sectors;
+	unsigned int chunk_sectors;
 	unsigned int bio_sectors = bvm->bi_size >> 9;
+	struct geom *geo = &conf->geo;
+
+	chunk_sectors = (conf->geo.chunk_mask & conf->prev.chunk_mask) + 1;
+	if (conf->reshape_progress != MaxSector &&
+	    ((sector >= conf->reshape_progress) !=
+	     conf->mddev->reshape_backwards))
+		geo = &conf->prev;
 
-	if (conf->near_copies < conf->raid_disks) {
+	if (geo->near_copies < geo->raid_disks) {
 		max = (chunk_sectors - ((sector & (chunk_sectors - 1))
 					+ bio_sectors)) << 9;
 		if (max < 0)
@@ -614,6 +647,12 @@ static int raid10_mergeable_bvec(struct request_queue *q,
 	if (mddev->merge_check_needed) {
 		struct r10bio r10_bio;
 		int s;
+		if (conf->reshape_progress != MaxSector) {
+			/* Cannot give any guidance during reshape */
+			if (max <= biovec->bv_len && bio_sectors == 0)
+				return biovec->bv_len;
+			return 0;
+		}
 		r10_bio.sector = sector;
 		raid10_find_phys(conf, &r10_bio);
 		rcu_read_lock();
@@ -681,6 +720,7 @@ static struct md_rdev *read_balance(struct r10conf *conf,
 	struct md_rdev *rdev, *best_rdev;
 	int do_balance;
 	int best_slot;
+	struct geom *geo = &conf->geo;
 
 	raid10_find_phys(conf, r10_bio);
 	rcu_read_lock();
@@ -761,11 +801,11 @@ retry:
 		 * sequential read speed for 'far copies' arrays.  So only
 		 * keep it for 'near' arrays, and review those later.
 		 */
-		if (conf->near_copies > 1 && !atomic_read(&rdev->nr_pending))
+		if (geo->near_copies > 1 && !atomic_read(&rdev->nr_pending))
 			break;
 
 		/* for far > 1 always use the lowest address */
-		if (conf->far_copies > 1)
+		if (geo->far_copies > 1)
 			new_distance = r10_bio->devs[slot].addr;
 		else
 			new_distance = abs(r10_bio->devs[slot].addr -
@@ -812,7 +852,10 @@ static int raid10_congested(void *data, int bits)
 	if (mddev_congested(mddev, bits))
 		return 1;
 	rcu_read_lock();
-	for (i = 0; i < conf->raid_disks && ret == 0; i++) {
+	for (i = 0;
+	     (i < conf->geo.raid_disks || i < conf->prev.raid_disks)
+		     && ret == 0;
+	     i++) {
 		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
 		if (rdev && !test_bit(Faulty, &rdev->flags)) {
 			struct request_queue *q = bdev_get_queue(rdev->bdev);
@@ -973,13 +1016,24 @@ static void unfreeze_array(struct r10conf *conf)
 	spin_unlock_irq(&conf->resync_lock);
 }
 
+static sector_t choose_data_offset(struct r10bio *r10_bio,
+				   struct md_rdev *rdev)
+{
+	if (!test_bit(MD_RECOVERY_RESHAPE, &rdev->mddev->recovery) ||
+	    test_bit(R10BIO_Previous, &r10_bio->state))
+		return rdev->data_offset;
+	else
+		return rdev->new_data_offset;
+}
+
 static void make_request(struct mddev *mddev, struct bio * bio)
 {
 	struct r10conf *conf = mddev->private;
 	struct r10bio *r10_bio;
 	struct bio *read_bio;
 	int i;
-	int chunk_sects = conf->chunk_mask + 1;
+	sector_t chunk_mask = (conf->geo.chunk_mask & conf->prev.chunk_mask);
+	int chunk_sects = chunk_mask + 1;
 	const int rw = bio_data_dir(bio);
 	const unsigned long do_sync = (bio->bi_rw & REQ_SYNC);
 	const unsigned long do_fua = (bio->bi_rw & REQ_FUA);
@@ -988,6 +1042,7 @@ static void make_request(struct mddev *mddev, struct bio * bio)
 	int plugged;
 	int sectors_handled;
 	int max_sectors;
+	int sectors;
 
 	if (unlikely(bio->bi_rw & REQ_FLUSH)) {
 		md_flush_request(mddev, bio);
@@ -997,9 +1052,10 @@ static void make_request(struct mddev *mddev, struct bio * bio)
 	/* If this request crosses a chunk boundary, we need to
 	 * split it.  This will only happen for 1 PAGE (or less) requests.
 	 */
-	if (unlikely( (bio->bi_sector & conf->chunk_mask) + (bio->bi_size >> 9)
-		      > chunk_sects &&
-		    conf->near_copies < conf->raid_disks)) {
+	if (unlikely((bio->bi_sector & chunk_mask) + (bio->bi_size >> 9)
+		     > chunk_sects
+		     && (conf->geo.near_copies < conf->geo.raid_disks
+			 || conf->prev.near_copies < conf->prev.raid_disks))) {
 		struct bio_pair *bp;
 		/* Sanity check -- queue functions should prevent this happening */
 		if (bio->bi_vcnt != 1 ||
@@ -1051,10 +1107,41 @@ static void make_request(struct mddev *mddev, struct bio * bio)
 	 */
 	wait_barrier(conf);
 
+	sectors = bio->bi_size >> 9;
+	while (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) &&
+	    bio->bi_sector < conf->reshape_progress &&
+	    bio->bi_sector + sectors > conf->reshape_progress) {
+		/* IO spans the reshape position.  Need to wait for
+		 * reshape to pass
+		 */
+		allow_barrier(conf);
+		wait_event(conf->wait_barrier,
+			   conf->reshape_progress <= bio->bi_sector ||
+			   conf->reshape_progress >= bio->bi_sector + sectors);
+		wait_barrier(conf);
+	}
+	if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) &&
+	    bio_data_dir(bio) == WRITE &&
+	    (mddev->reshape_backwards
+	     ? (bio->bi_sector < conf->reshape_safe &&
+		bio->bi_sector + sectors > conf->reshape_progress)
+	     : (bio->bi_sector + sectors > conf->reshape_safe &&
+		bio->bi_sector < conf->reshape_progress))) {
+		/* Need to update reshape_position in metadata */
+		mddev->reshape_position = conf->reshape_progress;
+		set_bit(MD_CHANGE_DEVS, &mddev->flags);
+		set_bit(MD_CHANGE_PENDING, &mddev->flags);
+		md_wakeup_thread(mddev->thread);
+		wait_event(mddev->sb_wait,
+			   !test_bit(MD_CHANGE_PENDING, &mddev->flags));
+
+		conf->reshape_safe = mddev->reshape_position;
+	}
+
 	r10_bio = mempool_alloc(conf->r10bio_pool, GFP_NOIO);
 
 	r10_bio->master_bio = bio;
-	r10_bio->sectors = bio->bi_size >> 9;
+	r10_bio->sectors = sectors;
 
 	r10_bio->mddev = mddev;
 	r10_bio->sector = bio->bi_sector;
@@ -1093,7 +1180,7 @@ read_again:
 		r10_bio->devs[slot].rdev = rdev;
 
 		read_bio->bi_sector = r10_bio->devs[slot].addr +
-			rdev->data_offset;
+			choose_data_offset(r10_bio, rdev);
 		read_bio->bi_bdev = rdev->bdev;
 		read_bio->bi_end_io = raid10_end_read_request;
 		read_bio->bi_rw = READ | do_sync;
@@ -1297,7 +1384,8 @@ retry_write:
 		r10_bio->devs[i].bio = mbio;
 
 		mbio->bi_sector	= (r10_bio->devs[i].addr+
-				   conf->mirrors[d].rdev->data_offset);
+				   choose_data_offset(r10_bio,
+						      conf->mirrors[d].rdev));
 		mbio->bi_bdev = conf->mirrors[d].rdev->bdev;
 		mbio->bi_end_io	= raid10_end_write_request;
 		mbio->bi_rw = WRITE | do_sync | do_fua;
@@ -1321,8 +1409,10 @@ retry_write:
 		 * so it cannot disappear, so the replacement cannot
 		 * become NULL here
 		 */
-		mbio->bi_sector	= (r10_bio->devs[i].addr+
-				   conf->mirrors[d].replacement->data_offset);
+		mbio->bi_sector	= (r10_bio->devs[i].addr +
+				   choose_data_offset(
+					   r10_bio,
+					   conf->mirrors[d].replacement));
 		mbio->bi_bdev = conf->mirrors[d].replacement->bdev;
 		mbio->bi_end_io	= raid10_end_write_request;
 		mbio->bi_rw = WRITE | do_sync | do_fua;
@@ -1368,19 +1458,19 @@ static void status(struct seq_file *seq, struct mddev *mddev)
 	struct r10conf *conf = mddev->private;
 	int i;
 
-	if (conf->near_copies < conf->raid_disks)
+	if (conf->geo.near_copies < conf->geo.raid_disks)
 		seq_printf(seq, " %dK chunks", mddev->chunk_sectors / 2);
-	if (conf->near_copies > 1)
-		seq_printf(seq, " %d near-copies", conf->near_copies);
-	if (conf->far_copies > 1) {
-		if (conf->far_offset)
-			seq_printf(seq, " %d offset-copies", conf->far_copies);
+	if (conf->geo.near_copies > 1)
+		seq_printf(seq, " %d near-copies", conf->geo.near_copies);
+	if (conf->geo.far_copies > 1) {
+		if (conf->geo.far_offset)
+			seq_printf(seq, " %d offset-copies", conf->geo.far_copies);
 		else
-			seq_printf(seq, " %d far-copies", conf->far_copies);
+			seq_printf(seq, " %d far-copies", conf->geo.far_copies);
 	}
-	seq_printf(seq, " [%d/%d] [", conf->raid_disks,
-					conf->raid_disks - mddev->degraded);
-	for (i = 0; i < conf->raid_disks; i++)
+	seq_printf(seq, " [%d/%d] [", conf->geo.raid_disks,
+					conf->geo.raid_disks - mddev->degraded);
+	for (i = 0; i < conf->geo.raid_disks; i++)
 		seq_printf(seq, "%s",
 			      conf->mirrors[i].rdev &&
 			      test_bit(In_sync, &conf->mirrors[i].rdev->flags) ? "U" : "_");
@@ -1392,7 +1482,7 @@ static void status(struct seq_file *seq, struct mddev *mddev)
  * Don't consider the device numbered 'ignore'
  * as we might be about to remove it.
  */
-static int enough(struct r10conf *conf, int ignore)
+static int _enough(struct r10conf *conf, struct geom *geo, int ignore)
 {
 	int first = 0;
 
@@ -1403,7 +1493,7 @@ static int enough(struct r10conf *conf, int ignore)
 			if (conf->mirrors[first].rdev &&
 			    first != ignore)
 				cnt++;
-			first = (first+1) % conf->raid_disks;
+			first = (first+1) % geo->raid_disks;
 		}
 		if (cnt == 0)
 			return 0;
@@ -1411,6 +1501,12 @@ static int enough(struct r10conf *conf, int ignore)
 	return 1;
 }
 
+static int enough(struct r10conf *conf, int ignore)
+{
+	return _enough(conf, &conf->geo, ignore) &&
+		_enough(conf, &conf->prev, ignore);
+}
+
 static void error(struct mddev *mddev, struct md_rdev *rdev)
 {
 	char b[BDEVNAME_SIZE];
@@ -1445,7 +1541,7 @@ static void error(struct mddev *mddev, struct md_rdev *rdev)
 	       "md/raid10:%s: Disk failure on %s, disabling device.\n"
 	       "md/raid10:%s: Operation continuing on %d devices.\n",
 	       mdname(mddev), bdevname(rdev->bdev, b),
-	       mdname(mddev), conf->raid_disks - mddev->degraded);
+	       mdname(mddev), conf->geo.raid_disks - mddev->degraded);
 }
 
 static void print_conf(struct r10conf *conf)
@@ -1458,10 +1554,10 @@ static void print_conf(struct r10conf *conf)
 		printk(KERN_DEBUG "(!conf)\n");
 		return;
 	}
-	printk(KERN_DEBUG " --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
-		conf->raid_disks);
+	printk(KERN_DEBUG " --- wd:%d rd:%d\n", conf->geo.raid_disks - conf->mddev->degraded,
+		conf->geo.raid_disks);
 
-	for (i = 0; i < conf->raid_disks; i++) {
+	for (i = 0; i < conf->geo.raid_disks; i++) {
 		char b[BDEVNAME_SIZE];
 		tmp = conf->mirrors + i;
 		if (tmp->rdev)
@@ -1493,7 +1589,7 @@ static int raid10_spare_active(struct mddev *mddev)
 	 * Find all non-in_sync disks within the RAID10 configuration
 	 * and mark them in_sync
 	 */
-	for (i = 0; i < conf->raid_disks; i++) {
+	for (i = 0; i < conf->geo.raid_disks; i++) {
 		tmp = conf->mirrors + i;
 		if (tmp->replacement
 		    && tmp->replacement->recovery_offset == MaxSector
@@ -1535,7 +1631,7 @@ static int raid10_add_disk(struct mddev *mddev, struct md_rdev *rdev)
 	int err = -EEXIST;
 	int mirror;
 	int first = 0;
-	int last = conf->raid_disks - 1;
+	int last = conf->geo.raid_disks - 1;
 	struct request_queue *q = bdev_get_queue(rdev->bdev);
 
 	if (mddev->recovery_cp < MaxSector)
@@ -1543,7 +1639,7 @@ static int raid10_add_disk(struct mddev *mddev, struct md_rdev *rdev)
 		 * very different from resync
 		 */
 		return -EBUSY;
-	if (rdev->saved_raid_disk < 0 && !enough(conf, -1))
+	if (rdev->saved_raid_disk < 0 && !_enough(conf, &conf->prev, -1))
 		return -EINVAL;
 
 	if (rdev->raid_disk >= 0)
@@ -1635,6 +1731,7 @@ static int raid10_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
 	if (!test_bit(Faulty, &rdev->flags) &&
 	    mddev->recovery_disabled != p->recovery_disabled &&
 	    (!p->replacement || p->replacement == rdev) &&
+	    number < conf->geo.raid_disks &&
 	    enough(conf, -1)) {
 		err = -EBUSY;
 		goto abort;
@@ -1676,7 +1773,11 @@ static void end_sync_read(struct bio *bio, int error)
 	struct r10conf *conf = r10_bio->mddev->private;
 	int d;
 
-	d = find_bio_disk(conf, r10_bio, bio, NULL, NULL);
+	if (bio == r10_bio->master_bio) {
+		/* this is a reshape read */
+		d = r10_bio->read_slot; /* really the read dev */
+	} else
+		d = find_bio_disk(conf, r10_bio, bio, NULL, NULL);
 
 	if (test_bit(BIO_UPTODATE, &bio->bi_flags))
 		set_bit(R10BIO_Uptodate, &r10_bio->state);
@@ -2218,7 +2319,9 @@ static void fix_read_error(struct r10conf *conf, struct mddev *mddev, struct r10
 				       " (%d sectors at %llu on %s)\n",
 				       mdname(mddev), s,
 				       (unsigned long long)(
-					       sect + rdev->data_offset),
+					       sect +
+					       choose_data_offset(r10_bio,
+								  rdev)),
 				       bdevname(rdev->bdev, b));
 				printk(KERN_NOTICE "md/raid10:%s: %s: failing "
 				       "drive\n",
@@ -2256,7 +2359,8 @@ static void fix_read_error(struct r10conf *conf, struct mddev *mddev, struct r10
 				       " (%d sectors at %llu on %s)\n",
 				       mdname(mddev), s,
 				       (unsigned long long)(
-					       sect + rdev->data_offset),
+					       sect +
+					       choose_data_offset(r10_bio, rdev)),
 				       bdevname(rdev->bdev, b));
 				printk(KERN_NOTICE "md/raid10:%s: %s: failing "
 				       "drive\n",
@@ -2269,7 +2373,8 @@ static void fix_read_error(struct r10conf *conf, struct mddev *mddev, struct r10
 				       " (%d sectors at %llu on %s)\n",
 				       mdname(mddev), s,
 				       (unsigned long long)(
-					       sect + rdev->data_offset),
+					       sect +
+					       choose_data_offset(r10_bio, rdev)),
 				       bdevname(rdev->bdev, b));
 				atomic_add(s, &rdev->corrected_errors);
 			}
@@ -2343,7 +2448,7 @@ static int narrow_write_error(struct r10bio *r10_bio, int i)
 		wbio = bio_clone_mddev(bio, GFP_NOIO, mddev);
 		md_trim_bio(wbio, sector - bio->bi_sector, sectors);
 		wbio->bi_sector = (r10_bio->devs[i].addr+
-				   rdev->data_offset+
+				   choose_data_offset(r10_bio, rdev) +
 				   (sector - r10_bio->sector));
 		wbio->bi_bdev = rdev->bdev;
 		if (submit_bio_wait(WRITE, wbio) == 0)
@@ -2420,7 +2525,7 @@ read_more:
 	r10_bio->devs[slot].bio = bio;
 	r10_bio->devs[slot].rdev = rdev;
 	bio->bi_sector = r10_bio->devs[slot].addr
-		+ rdev->data_offset;
+		+ choose_data_offset(r10_bio, rdev);
 	bio->bi_bdev = rdev->bdev;
 	bio->bi_rw = READ | do_sync;
 	bio->bi_private = r10_bio;
@@ -2480,7 +2585,7 @@ static void handle_write_completed(struct r10conf *conf, struct r10bio *r10_bio)
 				rdev_clear_badblocks(
 					rdev,
 					r10_bio->devs[m].addr,
-					r10_bio->sectors);
+					r10_bio->sectors, 0);
 			} else {
 				if (!rdev_set_badblocks(
 					    rdev,
@@ -2496,7 +2601,7 @@ static void handle_write_completed(struct r10conf *conf, struct r10bio *r10_bio)
 				rdev_clear_badblocks(
 					rdev,
 					r10_bio->devs[m].addr,
-					r10_bio->sectors);
+					r10_bio->sectors, 0);
 			} else {
 				if (!rdev_set_badblocks(
 					    rdev,
@@ -2515,7 +2620,7 @@ static void handle_write_completed(struct r10conf *conf, struct r10bio *r10_bio)
 				rdev_clear_badblocks(
 					rdev,
 					r10_bio->devs[m].addr,
-					r10_bio->sectors);
+					r10_bio->sectors, 0);
 				rdev_dec_pending(rdev, conf->mddev);
 			} else if (bio != NULL &&
 				   !test_bit(BIO_UPTODATE, &bio->bi_flags)) {
@@ -2532,7 +2637,7 @@ static void handle_write_completed(struct r10conf *conf, struct r10bio *r10_bio)
 				rdev_clear_badblocks(
 					rdev,
 					r10_bio->devs[m].addr,
-					r10_bio->sectors);
+					r10_bio->sectors, 0);
 				rdev_dec_pending(rdev, conf->mddev);
 			}
 		}
@@ -2573,6 +2678,8 @@ static void raid10d(struct mddev *mddev)
 		if (test_bit(R10BIO_MadeGood, &r10_bio->state) ||
 		    test_bit(R10BIO_WriteError, &r10_bio->state))
 			handle_write_completed(conf, r10_bio);
+		else if (test_bit(R10BIO_IsReshape, &r10_bio->state))
+			reshape_request_write(mddev, r10_bio);
 		else if (test_bit(R10BIO_IsSync, &r10_bio->state))
 			sync_request_write(mddev, r10_bio);
 		else if (test_bit(R10BIO_IsRecover, &r10_bio->state))
@@ -2603,7 +2710,7 @@ static int init_resync(struct r10conf *conf)
 	buffs = RESYNC_WINDOW / RESYNC_BLOCK_SIZE;
 	BUG_ON(conf->r10buf_pool);
 	conf->have_replacement = 0;
-	for (i = 0; i < conf->raid_disks; i++)
+	for (i = 0; i < conf->geo.raid_disks; i++)
 		if (conf->mirrors[i].replacement)
 			conf->have_replacement = 1;
 	conf->r10buf_pool = mempool_create(buffs, r10buf_pool_alloc, r10buf_pool_free, conf);
@@ -2657,6 +2764,7 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr,
 	sector_t sync_blocks;
 	sector_t sectors_skipped = 0;
 	int chunks_skipped = 0;
+	sector_t chunk_mask = conf->geo.chunk_mask;
 
 	if (!conf->r10buf_pool)
 		if (init_resync(conf))
@@ -2664,7 +2772,8 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr,
 
  skipped:
 	max_sector = mddev->dev_sectors;
-	if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery))
+	if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery) ||
+	    test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
 		max_sector = mddev->resync_max_sectors;
 	if (sector_nr >= max_sector) {
 		/* If we aborted, we need to abort the
@@ -2676,11 +2785,16 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr,
 		 * we need to convert that to several
 		 * virtual addresses.
 		 */
+		if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)) {
+			end_reshape(conf);
+			return 0;
+		}
+
 		if (mddev->curr_resync < max_sector) { /* aborted */
 			if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery))
 				bitmap_end_sync(mddev->bitmap, mddev->curr_resync,
 						&sync_blocks, 1);
-			else for (i=0; i<conf->raid_disks; i++) {
+			else for (i = 0; i < conf->geo.raid_disks; i++) {
 				sector_t sect =
 					raid10_find_virt(conf, mddev->curr_resync, i);
 				bitmap_end_sync(mddev->bitmap, sect,
@@ -2694,7 +2808,7 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr,
 				/* Completed a full sync so the replacements
 				 * are now fully recovered.
 				 */
-				for (i = 0; i < conf->raid_disks; i++)
+				for (i = 0; i < conf->geo.raid_disks; i++)
 					if (conf->mirrors[i].replacement)
 						conf->mirrors[i].replacement
 							->recovery_offset
@@ -2707,7 +2821,11 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr,
 		*skipped = 1;
 		return sectors_skipped;
 	}
-	if (chunks_skipped >= conf->raid_disks) {
+
+	if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
+		return reshape_request(mddev, sector_nr, skipped);
+
+	if (chunks_skipped >= conf->geo.raid_disks) {
 		/* if there has been nothing to do on any drive,
 		 * then there is nothing to do at all..
 		 */
@@ -2721,9 +2839,9 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr,
 	/* make sure whole request will fit in a chunk - if chunks
 	 * are meaningful
 	 */
-	if (conf->near_copies < conf->raid_disks &&
-	    max_sector > (sector_nr | conf->chunk_mask))
-		max_sector = (sector_nr | conf->chunk_mask) + 1;
+	if (conf->geo.near_copies < conf->geo.raid_disks &&
+	    max_sector > (sector_nr | chunk_mask))
+		max_sector = (sector_nr | chunk_mask) + 1;
 	/*
 	 * If there is non-resync activity waiting for us then
 	 * put in a delay to throttle resync.
@@ -2752,7 +2870,7 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr,
 		int j;
 		r10_bio = NULL;
 
-		for (i=0 ; i<conf->raid_disks; i++) {
+		for (i = 0 ; i < conf->geo.raid_disks; i++) {
 			int still_degraded;
 			struct r10bio *rb2;
 			sector_t sect;
@@ -2806,7 +2924,7 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr,
 			/* Need to check if the array will still be
 			 * degraded
 			 */
-			for (j=0; j<conf->raid_disks; j++)
+			for (j = 0; j < conf->geo.raid_disks; j++)
 				if (conf->mirrors[j].rdev == NULL ||
 				    test_bit(Faulty, &conf->mirrors[j].rdev->flags)) {
 					still_degraded = 1;
@@ -2984,9 +3102,9 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr,
 		r10_bio->sector = sector_nr;
 		set_bit(R10BIO_IsSync, &r10_bio->state);
 		raid10_find_phys(conf, r10_bio);
-		r10_bio->sectors = (sector_nr | conf->chunk_mask) - sector_nr +1;
+		r10_bio->sectors = (sector_nr | chunk_mask) - sector_nr + 1;
 
-		for (i=0; i<conf->copies; i++) {
+		for (i = 0; i < conf->copies; i++) {
 			int d = r10_bio->devs[i].devnum;
 			sector_t first_bad, sector;
 			int bad_sectors;
@@ -3152,16 +3270,17 @@ raid10_size(struct mddev *mddev, sector_t sectors, int raid_disks)
 	struct r10conf *conf = mddev->private;
 
 	if (!raid_disks)
-		raid_disks = conf->raid_disks;
+		raid_disks = min(conf->geo.raid_disks,
+				 conf->prev.raid_disks);
 	if (!sectors)
 		sectors = conf->dev_sectors;
 
-	size = sectors >> conf->chunk_shift;
-	sector_div(size, conf->far_copies);
+	size = sectors >> conf->geo.chunk_shift;
+	sector_div(size, conf->geo.far_copies);
 	size = size * raid_disks;
-	sector_div(size, conf->near_copies);
+	sector_div(size, conf->geo.near_copies);
 
-	return size << conf->chunk_shift;
+	return size << conf->geo.chunk_shift;
 }
 
 static void calc_sectors(struct r10conf *conf, sector_t size)
@@ -3171,10 +3290,10 @@ static void calc_sectors(struct r10conf *conf, sector_t size)
 	 * conf->stride
 	 */
 
-	size = size >> conf->chunk_shift;
-	sector_div(size, conf->far_copies);
-	size = size * conf->raid_disks;
-	sector_div(size, conf->near_copies);
+	size = size >> conf->geo.chunk_shift;
+	sector_div(size, conf->geo.far_copies);
+	size = size * conf->geo.raid_disks;
+	sector_div(size, conf->geo.near_copies);
 	/* 'size' is now the number of chunks in the array */
 	/* calculate "used chunks per device" */
 	size = size * conf->copies;
@@ -3182,38 +3301,76 @@ static void calc_sectors(struct r10conf *conf, sector_t size)
 	/* We need to round up when dividing by raid_disks to
 	 * get the stride size.
 	 */
-	size = DIV_ROUND_UP_SECTOR_T(size, conf->raid_disks);
+	size = DIV_ROUND_UP_SECTOR_T(size, conf->geo.raid_disks);
 
-	conf->dev_sectors = size << conf->chunk_shift;
+	conf->dev_sectors = size << conf->geo.chunk_shift;
 
-	if (conf->far_offset)
-		conf->stride = 1 << conf->chunk_shift;
+	if (conf->geo.far_offset)
+		conf->geo.stride = 1 << conf->geo.chunk_shift;
 	else {
-		sector_div(size, conf->far_copies);
-		conf->stride = size << conf->chunk_shift;
+		sector_div(size, conf->geo.far_copies);
+		conf->geo.stride = size << conf->geo.chunk_shift;
 	}
 }
 
+enum geo_type {geo_new, geo_old, geo_start};
+static int setup_geo(struct geom *geo, struct mddev *mddev, enum geo_type new)
+{
+	int nc, fc, fo;
+	int layout, chunk, disks;
+	switch (new) {
+	case geo_old:
+		layout = mddev->layout;
+		chunk = mddev->chunk_sectors;
+		disks = mddev->raid_disks - mddev->delta_disks;
+		break;
+	case geo_new:
+		layout = mddev->new_layout;
+		chunk = mddev->new_chunk_sectors;
+		disks = mddev->raid_disks;
+		break;
+	default: /* avoid 'may be unused' warnings */
+	case geo_start: /* new when starting reshape - raid_disks not
+			 * updated yet. */
+		layout = mddev->new_layout;
+		chunk = mddev->new_chunk_sectors;
+		disks = mddev->raid_disks + mddev->delta_disks;
+		break;
+	}
+	if (layout >> 17)
+		return -1;
+	if (chunk < (PAGE_SIZE >> 9) ||
+	    !is_power_of_2(chunk))
+		return -2;
+	nc = layout & 255;
+	fc = (layout >> 8) & 255;
+	fo = layout & (1<<16);
+	geo->raid_disks = disks;
+	geo->near_copies = nc;
+	geo->far_copies = fc;
+	geo->far_offset = fo;
+	geo->chunk_mask = chunk - 1;
+	geo->chunk_shift = ffz(~chunk);
+	return nc*fc;
+}
+
 static struct r10conf *setup_conf(struct mddev *mddev)
 {
 	struct r10conf *conf = NULL;
-	int nc, fc, fo;
 	int err = -EINVAL;
+	struct geom geo;
+	int copies;
+
+	copies = setup_geo(&geo, mddev, geo_new);
 
-	if (mddev->new_chunk_sectors < (PAGE_SIZE >> 9) ||
-	    !is_power_of_2(mddev->new_chunk_sectors)) {
+	if (copies == -2) {
 		printk(KERN_ERR "md/raid10:%s: chunk size must be "
 		       "at least PAGE_SIZE(%ld) and be a power of 2.\n",
 		       mdname(mddev), PAGE_SIZE);
 		goto out;
 	}
 
-	nc = mddev->new_layout & 255;
-	fc = (mddev->new_layout >> 8) & 255;
-	fo = mddev->new_layout & (1<<16);
-
-	if ((nc*fc) <2 || (nc*fc) > mddev->raid_disks ||
-	    (mddev->new_layout >> 17)) {
+	if (copies < 2 || copies > mddev->raid_disks) {
 		printk(KERN_ERR "md/raid10:%s: unsupported raid10 layout: 0x%8x\n",
 		       mdname(mddev), mddev->new_layout);
 		goto out;
@@ -3224,7 +3381,9 @@ static struct r10conf *setup_conf(struct mddev *mddev)
 	if (!conf)
 		goto out;
 
-	conf->mirrors = kzalloc(sizeof(struct mirror_info)*mddev->raid_disks,
+	/* FIXME calc properly */
+	conf->mirrors = kzalloc(sizeof(struct mirror_info)*(mddev->raid_disks +
+							    max(0,mddev->delta_disks)),
 				GFP_KERNEL);
 	if (!conf->mirrors)
 		goto out;
@@ -3233,22 +3392,29 @@ static struct r10conf *setup_conf(struct mddev *mddev)
 	if (!conf->tmppage)
 		goto out;
 
-
-	conf->raid_disks = mddev->raid_disks;
-	conf->near_copies = nc;
-	conf->far_copies = fc;
-	conf->copies = nc*fc;
-	conf->far_offset = fo;
-	conf->chunk_mask = mddev->new_chunk_sectors - 1;
-	conf->chunk_shift = ffz(~mddev->new_chunk_sectors);
-
+	conf->geo = geo;
+	conf->copies = copies;
 	conf->r10bio_pool = mempool_create(NR_RAID10_BIOS, r10bio_pool_alloc,
 					   r10bio_pool_free, conf);
 	if (!conf->r10bio_pool)
 		goto out;
 
 	calc_sectors(conf, mddev->dev_sectors);
-
+	if (mddev->reshape_position == MaxSector) {
+		conf->prev = conf->geo;
+		conf->reshape_progress = MaxSector;
+	} else {
+		if (setup_geo(&conf->prev, mddev, geo_old) != conf->copies) {
+			err = -EINVAL;
+			goto out;
+		}
+		conf->reshape_progress = mddev->reshape_position;
+		if (conf->prev.far_offset)
+			conf->prev.stride = 1 << conf->prev.chunk_shift;
+		else
+			/* far_copies must be 1 */
+			conf->prev.stride = conf->dev_sectors;
+	}
 	spin_lock_init(&conf->device_lock);
 	INIT_LIST_HEAD(&conf->retry_list);
 
@@ -3263,8 +3429,9 @@ static struct r10conf *setup_conf(struct mddev *mddev)
 	return conf;
 
  out:
-	printk(KERN_ERR "md/raid10:%s: couldn't allocate memory.\n",
-	       mdname(mddev));
+	if (err == -ENOMEM)
+		printk(KERN_ERR "md/raid10:%s: couldn't allocate memory.\n",
+		       mdname(mddev));
 	if (conf) {
 		if (conf->r10bio_pool)
 			mempool_destroy(conf->r10bio_pool);
@@ -3282,12 +3449,8 @@ static int run(struct mddev *mddev)
 	struct mirror_info *disk;
 	struct md_rdev *rdev;
 	sector_t size;
-
-	/*
-	 * copy the already verified devices into our private RAID10
-	 * bookkeeping area. [whatever we allocate in run(),
-	 * should be freed in stop()]
-	 */
+	sector_t min_offset_diff = 0;
+	int first = 1;
 
 	if (mddev->private == NULL) {
 		conf = setup_conf(mddev);
@@ -3304,17 +3467,20 @@ static int run(struct mddev *mddev)
 
 	chunk_size = mddev->chunk_sectors << 9;
 	blk_queue_io_min(mddev->queue, chunk_size);
-	if (conf->raid_disks % conf->near_copies)
-		blk_queue_io_opt(mddev->queue, chunk_size * conf->raid_disks);
+	if (conf->geo.raid_disks % conf->geo.near_copies)
+		blk_queue_io_opt(mddev->queue, chunk_size * conf->geo.raid_disks);
 	else
 		blk_queue_io_opt(mddev->queue, chunk_size *
-				 (conf->raid_disks / conf->near_copies));
+				 (conf->geo.raid_disks / conf->geo.near_copies));
 
 	rdev_for_each(rdev, mddev) {
+		long long diff;
 
 		disk_idx = rdev->raid_disk;
-		if (disk_idx >= conf->raid_disks
-		    || disk_idx < 0)
+		if (disk_idx < 0)
+			continue;
+		if (disk_idx >= conf->geo.raid_disks &&
+		    disk_idx >= conf->prev.raid_disks)
 			continue;
 		disk = conf->mirrors + disk_idx;
 
@@ -3327,12 +3493,20 @@ static int run(struct mddev *mddev)
 				goto out_free_conf;
 			disk->rdev = rdev;
 		}
+		diff = (rdev->new_data_offset - rdev->data_offset);
+		if (!mddev->reshape_backwards)
+			diff = -diff;
+		if (diff < 0)
+			diff = 0;
+		if (first || diff < min_offset_diff)
+			min_offset_diff = diff;
 
 		disk_stack_limits(mddev->gendisk, rdev->bdev,
 				  rdev->data_offset << 9);
 
 		disk->head_position = 0;
 	}
+
 	/* need to check that every block has at least one working mirror */
 	if (!enough(conf, -1)) {
 		printk(KERN_ERR "md/raid10:%s: not enough operational mirrors.\n",
@@ -3340,8 +3514,21 @@ static int run(struct mddev *mddev)
 		goto out_free_conf;
 	}
 
+	if (conf->reshape_progress != MaxSector) {
+		/* must ensure that shape change is supported */
+		if (conf->geo.far_copies != 1 &&
+		    conf->geo.far_offset == 0)
+			goto out_free_conf;
+		if (conf->prev.far_copies != 1 &&
+		    conf->geo.far_offset == 0)
+			goto out_free_conf;
+	}
+
 	mddev->degraded = 0;
-	for (i = 0; i < conf->raid_disks; i++) {
+	for (i = 0;
+	     i < conf->geo.raid_disks
+		     || i < conf->prev.raid_disks;
+	     i++) {
 
 		disk = conf->mirrors + i;
 
@@ -3368,8 +3555,8 @@ static int run(struct mddev *mddev)
 		       mdname(mddev));
 	printk(KERN_INFO
 		"md/raid10:%s: active with %d out of %d devices\n",
-		mdname(mddev), conf->raid_disks - mddev->degraded,
-		conf->raid_disks);
+		mdname(mddev), conf->geo.raid_disks - mddev->degraded,
+		conf->geo.raid_disks);
 	/*
 	 * Ok, everything is just fine now
 	 */
@@ -3386,11 +3573,11 @@ static int run(struct mddev *mddev)
 	 * maybe...
 	 */
 	{
-		int stripe = conf->raid_disks *
+		int stripe = conf->geo.raid_disks *
 			((mddev->chunk_sectors << 9) / PAGE_SIZE);
-		stripe /= conf->near_copies;
-		if (mddev->queue->backing_dev_info.ra_pages < 2* stripe)
-			mddev->queue->backing_dev_info.ra_pages = 2* stripe;
+		stripe /= conf->geo.near_copies;
+		if (mddev-