Merge branch 'for-next' of git://neil.brown.name/md

Pull md updates from NeilBrown.

* 'for-next' of git://neil.brown.name/md:
  DM RAID: Add support for MD RAID10
  md/RAID1: Add missing case for attempting to repair known bad blocks.
  md/raid5: For odirect-write performance, do not set STRIPE_PREREAD_ACTIVE.
  md/raid1: don't abort a resync on the first badblock.
  md: remove duplicated test on ->openers when calling do_md_stop()
  raid5: Add R5_ReadNoMerge flag which prevent bio from merging at block layer
  md/raid1: prevent merging too large request
  md/raid1: read balance chooses idlest disk for SSD
  md/raid1: make sequential read detection per disk based
  MD RAID10: Export md_raid10_congested
  MD: Move macros from raid1*.h to raid1*.c
  MD RAID1: rename mirror_info structure
  MD RAID10: rename mirror_info structure
  MD RAID10: Fix compiler warning.
  raid5: add a per-stripe lock
  raid5: remove unnecessary bitmap write optimization
  raid5: lockless access raid5 overrided bi_phys_segments
  raid5: reduce chance release_stripe() taking device_lock

9 files changed, 426 insertions, 219 deletions

diff --git a/Documentation/device-mapper/dm-raid.txt b/Documentation/device-mapper/dm-raid.txt
index 946c73342cd..1c184495716 100644
--- a/Documentation/device-mapper/dm-raid.txt
+++ b/Documentation/device-mapper/dm-raid.txt
@@ -27,6 +27,10 @@ The target is named "raid" and it accepts the following parameters:
 		- rotating parity N (right-to-left) with data restart
   raid6_nc	RAID6 N continue
 		- rotating parity N (right-to-left) with data continuation
+  raid10        Various RAID10 inspired algorithms chosen by additional params
+		- RAID10: Striped Mirrors (aka 'Striping on top of mirrors')
+		- RAID1E: Integrated Adjacent Stripe Mirroring
+		-  and other similar RAID10 variants
 
   Reference: Chapter 4 of
   http://www.snia.org/sites/default/files/SNIA_DDF_Technical_Position_v2.0.pdf
@@ -59,6 +63,28 @@ The target is named "raid" and it accepts the following parameters:
 		logical size of the array.  The bitmap records the device
 		synchronisation state for each region.
 
+        [raid10_copies   <# copies>]
+        [raid10_format   near]
+		These two options are used to alter the default layout of
+		a RAID10 configuration.  The number of copies is can be
+		specified, but the default is 2.  There are other variations
+		to how the copies are laid down - the default and only current
+		option is "near".  Near copies are what most people think of
+		with respect to mirroring.  If these options are left
+		unspecified, or 'raid10_copies 2' and/or 'raid10_format near'
+		are given, then the layouts for 2, 3 and 4 devices are:
+		2 drives         3 drives          4 drives
+		--------         ----------        --------------
+		A1  A1           A1  A1  A2        A1  A1  A2  A2
+		A2  A2           A2  A3  A3        A3  A3  A4  A4
+		A3  A3           A4  A4  A5        A5  A5  A6  A6
+		A4  A4           A5  A6  A6        A7  A7  A8  A8
+		..  ..           ..  ..  ..        ..  ..  ..  ..
+		The 2-device layout is equivalent 2-way RAID1.  The 4-device
+		layout is what a traditional RAID10 would look like.  The
+		3-device layout is what might be called a 'RAID1E - Integrated
+		Adjacent Stripe Mirroring'.
+
 <#raid_devs>: The number of devices composing the array.
 	Each device consists of two entries.  The first is the device
 	containing the metadata (if any); the second is the one containing the
diff --git a/drivers/md/dm-raid.c b/drivers/md/dm-raid.c
index f2f29c52654..982e3e390c4 100644
--- a/drivers/md/dm-raid.c
+++ b/drivers/md/dm-raid.c
@@ -11,6 +11,7 @@
 #include "md.h"
 #include "raid1.h"
 #include "raid5.h"
+#include "raid10.h"
 #include "bitmap.h"
 
 #include <linux/device-mapper.h>
@@ -52,7 +53,10 @@ struct raid_dev {
 #define DMPF_MAX_RECOVERY_RATE 0x20
 #define DMPF_MAX_WRITE_BEHIND  0x40
 #define DMPF_STRIPE_CACHE      0x80
-#define DMPF_REGION_SIZE       0X100
+#define DMPF_REGION_SIZE       0x100
+#define DMPF_RAID10_COPIES     0x200
+#define DMPF_RAID10_FORMAT     0x400
+
 struct raid_set {
 	struct dm_target *ti;
 
@@ -76,6 +80,7 @@ static struct raid_type {
 	const unsigned algorithm;	/* RAID algorithm. */
 } raid_types[] = {
 	{"raid1",    "RAID1 (mirroring)",               0, 2, 1, 0 /* NONE */},
+	{"raid10",   "RAID10 (striped mirrors)",        0, 2, 10, UINT_MAX /* Varies */},
 	{"raid4",    "RAID4 (dedicated parity disk)",	1, 2, 5, ALGORITHM_PARITY_0},
 	{"raid5_la", "RAID5 (left asymmetric)",		1, 2, 5, ALGORITHM_LEFT_ASYMMETRIC},
 	{"raid5_ra", "RAID5 (right asymmetric)",	1, 2, 5, ALGORITHM_RIGHT_ASYMMETRIC},
@@ -86,6 +91,17 @@ static struct raid_type {
 	{"raid6_nc", "RAID6 (N continue)",		2, 4, 6, ALGORITHM_ROTATING_N_CONTINUE}
 };
 
+static unsigned raid10_md_layout_to_copies(int layout)
+{
+	return layout & 0xFF;
+}
+
+static int raid10_format_to_md_layout(char *format, unsigned copies)
+{
+	/* 1 "far" copy, and 'copies' "near" copies */
+	return (1 << 8) | (copies & 0xFF);
+}
+
 static struct raid_type *get_raid_type(char *name)
 {
 	int i;
@@ -339,10 +355,16 @@ static int validate_region_size(struct raid_set *rs, unsigned long region_size)
  *    [max_write_behind <sectors>]	See '-write-behind=' (man mdadm)
  *    [stripe_cache <sectors>]		Stripe cache size for higher RAIDs
  *    [region_size <sectors>]           Defines granularity of bitmap
+ *
+ * RAID10-only options:
+ *    [raid10_copies <# copies>]        Number of copies.  (Default: 2)
+ *    [raid10_format <near>]            Layout algorithm.  (Default: near)
  */
 static int parse_raid_params(struct raid_set *rs, char **argv,
 			     unsigned num_raid_params)
 {
+	char *raid10_format = "near";
+	unsigned raid10_copies = 2;
 	unsigned i, rebuild_cnt = 0;
 	unsigned long value, region_size = 0;
 	sector_t sectors_per_dev = rs->ti->len;
@@ -416,11 +438,28 @@ static int parse_raid_params(struct raid_set *rs, char **argv,
 		}
 
 		key = argv[i++];
+
+		/* Parameters that take a string value are checked here. */
+		if (!strcasecmp(key, "raid10_format")) {
+			if (rs->raid_type->level != 10) {
+				rs->ti->error = "'raid10_format' is an invalid parameter for this RAID type";
+				return -EINVAL;
+			}
+			if (strcmp("near", argv[i])) {
+				rs->ti->error = "Invalid 'raid10_format' value given";
+				return -EINVAL;
+			}
+			raid10_format = argv[i];
+			rs->print_flags |= DMPF_RAID10_FORMAT;
+			continue;
+		}
+
 		if (strict_strtoul(argv[i], 10, &value) < 0) {
 			rs->ti->error = "Bad numerical argument given in raid params";
 			return -EINVAL;
 		}
 
+		/* Parameters that take a numeric value are checked here */
 		if (!strcasecmp(key, "rebuild")) {
 			rebuild_cnt++;
 
@@ -439,6 +478,7 @@ static int parse_raid_params(struct raid_set *rs, char **argv,
 					return -EINVAL;
 				}
 				break;
+			case 10:
 			default:
 				DMERR("The rebuild parameter is not supported for %s", rs->raid_type->name);
 				rs->ti->error = "Rebuild not supported for this RAID type";
@@ -495,7 +535,8 @@ static int parse_raid_params(struct raid_set *rs, char **argv,
 			 */
 			value /= 2;
 
-			if (rs->raid_type->level < 5) {
+			if ((rs->raid_type->level != 5) &&
+			    (rs->raid_type->level != 6)) {
 				rs->ti->error = "Inappropriate argument: stripe_cache";
 				return -EINVAL;
 			}
@@ -520,6 +561,14 @@ static int parse_raid_params(struct raid_set *rs, char **argv,
 		} else if (!strcasecmp(key, "region_size")) {
 			rs->print_flags |= DMPF_REGION_SIZE;
 			region_size = value;
+		} else if (!strcasecmp(key, "raid10_copies") &&
+			   (rs->raid_type->level == 10)) {
+			if ((value < 2) || (value > 0xFF)) {
+				rs->ti->error = "Bad value for 'raid10_copies'";
+				return -EINVAL;
+			}
+			rs->print_flags |= DMPF_RAID10_COPIES;
+			raid10_copies = value;
 		} else {
 			DMERR("Unable to parse RAID parameter: %s", key);
 			rs->ti->error = "Unable to parse RAID parameters";
@@ -538,8 +587,22 @@ static int parse_raid_params(struct raid_set *rs, char **argv,
 	if (dm_set_target_max_io_len(rs->ti, max_io_len))
 		return -EINVAL;
 
-	if ((rs->raid_type->level > 1) &&
-	    sector_div(sectors_per_dev, (rs->md.raid_disks - rs->raid_type->parity_devs))) {
+	if (rs->raid_type->level == 10) {
+		if (raid10_copies > rs->md.raid_disks) {
+			rs->ti->error = "Not enough devices to satisfy specification";
+			return -EINVAL;
+		}
+
+		/* (Len * #mirrors) / #devices */
+		sectors_per_dev = rs->ti->len * raid10_copies;
+		sector_div(sectors_per_dev, rs->md.raid_disks);
+
+		rs->md.layout = raid10_format_to_md_layout(raid10_format,
+							   raid10_copies);
+		rs->md.new_layout = rs->md.layout;
+	} else if ((rs->raid_type->level > 1) &&
+		   sector_div(sectors_per_dev,
+			      (rs->md.raid_disks - rs->raid_type->parity_devs))) {
 		rs->ti->error = "Target length not divisible by number of data devices";
 		return -EINVAL;
 	}
@@ -566,6 +629,9 @@ static int raid_is_congested(struct dm_target_callbacks *cb, int bits)
 	if (rs->raid_type->level == 1)
 		return md_raid1_congested(&rs->md, bits);
 
+	if (rs->raid_type->level == 10)
+		return md_raid10_congested(&rs->md, bits);
+
 	return md_raid5_congested(&rs->md, bits);
 }
 
@@ -884,6 +950,9 @@ static int analyse_superblocks(struct dm_target *ti, struct raid_set *rs)
 	case 6:
 		redundancy = rs->raid_type->parity_devs;
 		break;
+	case 10:
+		redundancy = raid10_md_layout_to_copies(mddev->layout) - 1;
+		break;
 	default:
 		ti->error = "Unknown RAID type";
 		return -EINVAL;
@@ -1049,12 +1118,19 @@ static int raid_ctr(struct dm_target *ti, unsigned argc, char **argv)
 		goto bad;
 	}
 
+	if (ti->len != rs->md.array_sectors) {
+		ti->error = "Array size does not match requested target length";
+		ret = -EINVAL;
+		goto size_mismatch;
+	}
 	rs->callbacks.congested_fn = raid_is_congested;
 	dm_table_add_target_callbacks(ti->table, &rs->callbacks);
 
 	mddev_suspend(&rs->md);
 	return 0;
 
+size_mismatch:
+	md_stop(&rs->md);
 bad:
 	context_free(rs);
 
@@ -1203,6 +1279,13 @@ static int raid_status(struct dm_target *ti, status_type_t type,
 			DMEMIT(" region_size %lu",
 			       rs->md.bitmap_info.chunksize >> 9);
 
+		if (rs->print_flags & DMPF_RAID10_COPIES)
+			DMEMIT(" raid10_copies %u",
+			       raid10_md_layout_to_copies(rs->md.layout));
+
+		if (rs->print_flags & DMPF_RAID10_FORMAT)
+			DMEMIT(" raid10_format near");
+
 		DMEMIT(" %d", rs->md.raid_disks);
 		for (i = 0; i < rs->md.raid_disks; i++) {
 			if (rs->dev[i].meta_dev)
@@ -1277,7 +1360,7 @@ static void raid_resume(struct dm_target *ti)
 
 static struct target_type raid_target = {
 	.name = "raid",
-	.version = {1, 2, 0},
+	.version = {1, 3, 0},
 	.module = THIS_MODULE,
 	.ctr = raid_ctr,
 	.dtr = raid_dtr,
@@ -1304,6 +1387,8 @@ module_init(dm_raid_init);
 module_exit(dm_raid_exit);
 
 MODULE_DESCRIPTION(DM_NAME " raid4/5/6 target");
+MODULE_ALIAS("dm-raid1");
+MODULE_ALIAS("dm-raid10");
 MODULE_ALIAS("dm-raid4");
 MODULE_ALIAS("dm-raid5");
 MODULE_ALIAS("dm-raid6");
diff --git a/drivers/md/md.c b/drivers/md/md.c
index d5ab4493c8b..f6c46109b07 100644
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -3942,17 +3942,13 @@ array_state_store(struct mddev *mddev, const char *buf, size_t len)
 		break;
 	case clear:
 		/* stopping an active array */
-		if (atomic_read(&mddev->openers) > 0)
-			return -EBUSY;
 		err = do_md_stop(mddev, 0, NULL);
 		break;
 	case inactive:
 		/* stopping an active array */
-		if (mddev->pers) {
-			if (atomic_read(&mddev->openers) > 0)
-				return -EBUSY;
+		if (mddev->pers)
 			err = do_md_stop(mddev, 2, NULL);
-		} else
+		else
 			err = 0; /* already inactive */
 		break;
 	case suspended:
diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c
index cacd008d686..197f62681db 100644
--- a/drivers/md/raid1.c
+++ b/drivers/md/raid1.c
@@ -46,6 +46,20 @@
  */
 #define	NR_RAID1_BIOS 256
 
+/* when we get a read error on a read-only array, we redirect to another
+ * device without failing the first device, or trying to over-write to
+ * correct the read error.  To keep track of bad blocks on a per-bio
+ * level, we store IO_BLOCKED in the appropriate 'bios' pointer
+ */
+#define IO_BLOCKED ((struct bio *)1)
+/* When we successfully write to a known bad-block, we need to remove the
+ * bad-block marking which must be done from process context.  So we record
+ * the success by setting devs[n].bio to IO_MADE_GOOD
+ */
+#define IO_MADE_GOOD ((struct bio *)2)
+
+#define BIO_SPECIAL(bio) ((unsigned long)bio <= 2)
+
 /* When there are this many requests queue to be written by
  * the raid1 thread, we become 'congested' to provide back-pressure
  * for writeback.
@@ -483,12 +497,14 @@ static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sect
 	const sector_t this_sector = r1_bio->sector;
 	int sectors;
 	int best_good_sectors;
-	int start_disk;
-	int best_disk;
-	int i;
+	int best_disk, best_dist_disk, best_pending_disk;
+	int has_nonrot_disk;
+	int disk;
 	sector_t best_dist;
+	unsigned int min_pending;
 	struct md_rdev *rdev;
 	int choose_first;
+	int choose_next_idle;
 
 	rcu_read_lock();
 	/*
@@ -499,26 +515,26 @@ static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sect
  retry:
 	sectors = r1_bio->sectors;
 	best_disk = -1;
+	best_dist_disk = -1;
 	best_dist = MaxSector;
+	best_pending_disk = -1;
+	min_pending = UINT_MAX;
 	best_good_sectors = 0;
+	has_nonrot_disk = 0;
+	choose_next_idle = 0;
 
 	if (conf->mddev->recovery_cp < MaxSector &&
-	    (this_sector + sectors >= conf->next_resync)) {
+	    (this_sector + sectors >= conf->next_resync))
 		choose_first = 1;
-		start_disk = 0;
-	} else {
+	else
 		choose_first = 0;
-		start_disk = conf->last_used;
-	}
 
-	for (i = 0 ; i < conf->raid_disks * 2 ; i++) {
+	for (disk = 0 ; disk < conf->raid_disks * 2 ; disk++) {
 		sector_t dist;
 		sector_t first_bad;
 		int bad_sectors;
-
-		int disk = start_disk + i;
-		if (disk >= conf->raid_disks * 2)
-			disk -= conf->raid_disks * 2;
+		unsigned int pending;
+		bool nonrot;
 
 		rdev = rcu_dereference(conf->mirrors[disk].rdev);
 		if (r1_bio->bios[disk] == IO_BLOCKED
@@ -577,22 +593,77 @@ static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sect
 		} else
 			best_good_sectors = sectors;
 
+		nonrot = blk_queue_nonrot(bdev_get_queue(rdev->bdev));
+		has_nonrot_disk |= nonrot;
+		pending = atomic_read(&rdev->nr_pending);
 		dist = abs(this_sector - conf->mirrors[disk].head_position);
-		if (choose_first
-		    /* Don't change to another disk for sequential reads */
-		    || conf->next_seq_sect == this_sector
-		    || dist == 0
-		    /* If device is idle, use it */
-		    || atomic_read(&rdev->nr_pending) == 0) {
+		if (choose_first) {
+			best_disk = disk;
+			break;
+		}
+		/* Don't change to another disk for sequential reads */
+		if (conf->mirrors[disk].next_seq_sect == this_sector
+		    || dist == 0) {
+			int opt_iosize = bdev_io_opt(rdev->bdev) >> 9;
+			struct raid1_info *mirror = &conf->mirrors[disk];
+
+			best_disk = disk;
+			/*
+			 * If buffered sequential IO size exceeds optimal
+			 * iosize, check if there is idle disk. If yes, choose
+			 * the idle disk. read_balance could already choose an
+			 * idle disk before noticing it's a sequential IO in
+			 * this disk. This doesn't matter because this disk
+			 * will idle, next time it will be utilized after the
+			 * first disk has IO size exceeds optimal iosize. In
+			 * this way, iosize of the first disk will be optimal
+			 * iosize at least. iosize of the second disk might be
+			 * small, but not a big deal since when the second disk
+			 * starts IO, the first disk is likely still busy.
+			 */
+			if (nonrot && opt_iosize > 0 &&
+			    mirror->seq_start != MaxSector &&
+			    mirror->next_seq_sect > opt_iosize &&
+			    mirror->next_seq_sect - opt_iosize >=
+			    mirror->seq_start) {
+				choose_next_idle = 1;
+				continue;
+			}
+			break;
+		}
+		/* If device is idle, use it */
+		if (pending == 0) {
 			best_disk = disk;
 			break;
 		}
+
+		if (choose_next_idle)
+			continue;
+
+		if (min_pending > pending) {
+			min_pending = pending;
+			best_pending_disk = disk;
+		}
+
 		if (dist < best_dist) {
 			best_dist = dist;
-			best_disk = disk;
+			best_dist_disk = disk;
 		}
 	}
 
+	/*
+	 * If all disks are rotational, choose the closest disk. If any disk is
+	 * non-rotational, choose the disk with less pending request even the
+	 * disk is rotational, which might/might not be optimal for raids with
+	 * mixed ratation/non-rotational disks depending on workload.
+	 */
+	if (best_disk == -1) {
+		if (has_nonrot_disk)
+			best_disk = best_pending_disk;
+		else
+			best_disk = best_dist_disk;
+	}
+
 	if (best_disk >= 0) {
 		rdev = rcu_dereference(conf->mirrors[best_disk].rdev);
 		if (!rdev)
@@ -606,8 +677,11 @@ static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sect
 			goto retry;
 		}
 		sectors = best_good_sectors;
-		conf->next_seq_sect = this_sector + sectors;
-		conf->last_used = best_disk;
+
+		if (conf->mirrors[best_disk].next_seq_sect != this_sector)
+			conf->mirrors[best_disk].seq_start = this_sector;
+
+		conf->mirrors[best_disk].next_seq_sect = this_sector + sectors;
 	}
 	rcu_read_unlock();
 	*max_sectors = sectors;
@@ -873,7 +947,7 @@ do_sync_io:
 static void make_request(struct mddev *mddev, struct bio * bio)
 {
 	struct r1conf *conf = mddev->private;
-	struct mirror_info *mirror;
+	struct raid1_info *mirror;
 	struct r1bio *r1_bio;
 	struct bio *read_bio;
 	int i, disks;
@@ -1364,7 +1438,7 @@ static int raid1_add_disk(struct mddev *mddev, struct md_rdev *rdev)
 	struct r1conf *conf = mddev->private;
 	int err = -EEXIST;
 	int mirror = 0;
-	struct mirror_info *p;
+	struct raid1_info *p;
 	int first = 0;
 	int last = conf->raid_disks - 1;
 	struct request_queue *q = bdev_get_queue(rdev->bdev);
@@ -1433,7 +1507,7 @@ static int raid1_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
 	struct r1conf *conf = mddev->private;
 	int err = 0;
 	int number = rdev->raid_disk;
-	struct mirror_info *p = conf->mirrors+ number;
+	struct raid1_info *p = conf->mirrors + number;
 
 	if (rdev != p->rdev)
 		p = conf->mirrors + conf->raid_disks + number;
@@ -2371,6 +2445,18 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int *skipp
 				bio->bi_rw = READ;
 				bio->bi_end_io = end_sync_read;
 				read_targets++;
+			} else if (!test_bit(WriteErrorSeen, &rdev->flags) &&
+				test_bit(MD_RECOVERY_SYNC, &mddev->recovery) &&
+				!test_bit(MD_RECOVERY_CHECK, &mddev->recovery)) {
+				/*
+				 * The device is suitable for reading (InSync),
+				 * but has bad block(s) here. Let's try to correct them,
+				 * if we are doing resync or repair. Otherwise, leave
+				 * this device alone for this sync request.
+				 */
+				bio->bi_rw = WRITE;
+				bio->bi_end_io = end_sync_write;
+				write_targets++;
 			}
 		}
 		if (bio->bi_end_io) {
@@ -2428,7 +2514,10 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int *skipp
 		/* There is nowhere to write, so all non-sync
 		 * drives must be failed - so we are finished
 		 */
-		sector_t rv = max_sector - sector_nr;
+		sector_t rv;
+		if (min_bad > 0)
+			max_sector = sector_nr + min_bad;
+		rv = max_sector - sector_nr;
 		*skipped = 1;
 		put_buf(r1_bio);
 		return rv;
@@ -2521,7 +2610,7 @@ static struct r1conf *setup_conf(struct mddev *mddev)
 {
 	struct r1conf *conf;
 	int i;
-	struct mirror_info *disk;
+	struct raid1_info *disk;
 	struct md_rdev *rdev;
 	int err = -ENOMEM;
 
@@ -2529,7 +2618,7 @@ static struct r1conf *setup_conf(struct mddev *mddev)
 	if (!conf)
 		goto abort;
 
-	conf->mirrors = kzalloc(sizeof(struct mirror_info)
+	conf->mirrors = kzalloc(sizeof(struct raid1_info)
 				* mddev->raid_disks * 2,
 				 GFP_KERNEL);
 	if (!conf->mirrors)
@@ -2572,6 +2661,7 @@ static struct r1conf *setup_conf(struct mddev *mddev)
 			mddev->merge_check_needed = 1;
 
 		disk->head_position = 0;
+		disk->seq_start = MaxSector;
 	}
 	conf->raid_disks = mddev->raid_disks;
 	conf->mddev = mddev;
@@ -2585,7 +2675,6 @@ static struct r1conf *setup_conf(struct mddev *mddev)
 	conf->recovery_disabled = mddev->recovery_disabled - 1;
 
 	err = -EIO;
-	conf->last_used = -1;
 	for (i = 0; i < conf->raid_disks * 2; i++) {
 
 		disk = conf->mirrors + i;
@@ -2611,19 +2700,9 @@ static struct r1conf *setup_conf(struct mddev *mddev)
 			if (disk->rdev &&
 			    (disk->rdev->saved_raid_disk < 0))
 				conf->fullsync = 1;
-		} else if (conf->last_used < 0)
-			/*
-			 * The first working device is used as a
-			 * starting point to read balancing.
-			 */
-			conf->last_used = i;
+		}
 	}
 
-	if (conf->last_used < 0) {
-		printk(KERN_ERR "md/raid1:%s: no operational mirrors\n",
-		       mdname(mddev));
-		goto abort;
-	}
 	err = -ENOMEM;
 	conf->thread = md_register_thread(raid1d, mddev, "raid1");
 	if (!conf->thread) {
@@ -2798,7 +2877,7 @@ static int raid1_reshape(struct mddev *mddev)
 	 */
 	mempool_t *newpool, *oldpool;
 	struct pool_info *newpoolinfo;
-	struct mirror_info *newmirrors;
+	struct raid1_info *newmirrors;
 	struct r1conf *conf = mddev->private;
 	int cnt, raid_disks;
 	unsigned long flags;
@@ -2841,7 +2920,7 @@ static int raid1_reshape(struct mddev *mddev)
 		kfree(newpoolinfo);
 		return -ENOMEM;
 	}
-	newmirrors = kzalloc(sizeof(struct mirror_info) * raid_disks * 2,
+	newmirrors = kzalloc(sizeof(struct raid1_info) * raid_disks * 2,
 			     GFP_KERNEL);
 	if (!newmirrors) {
 		kfree(newpoolinfo);
@@ -2880,7 +2959,6 @@ static int raid1_reshape(struct mddev *mddev)
 	conf->raid_disks = mddev->raid_disks = raid_disks;
 	mddev->delta_disks = 0;
 
-	conf->last_used = 0; /* just make sure it is in-range */
 	lower_barrier(conf);
 
 	set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
diff --git a/drivers/md/raid1.h b/drivers/md/raid1.h
index 80ded139314..0ff3715fb7e 100644
--- a/drivers/md/raid1.h
+++ b/drivers/md/raid1.h
@@ -1,9 +1,15 @@
 #ifndef _RAID1_H
 #define _RAID1_H
 
-struct mirror_info {
+struct raid1_info {
 	struct md_rdev	*rdev;
 	sector_t	head_position;
+
+	/* When choose the best device for a read (read_balance())
+	 * we try to keep sequential reads one the same device
+	 */
+	sector_t	next_seq_sect;
+	sector_t	seq_start;
 };
 
 /*
@@ -24,17 +30,11 @@ struct pool_info {
 
 struct r1conf {
 	struct mddev		*mddev;
-	struct mirror_info	*mirrors;	/* twice 'raid_disks' to
+	struct raid1_info	*mirrors;	/* twice 'raid_disks' to
 						 * allow for replacements.
 						 */
 	int			raid_disks;
 
-	/* When choose the best device for a read (read_balance())
-	 * we try to keep sequential reads one the same device
-	 * using 'last_used' and 'next_seq_sect'
-	 */
-	int			last_used;
-	sector_t		next_seq_sect;
 	/* During resync, read_balancing is only allowed on the part
 	 * of the array that has been resynced.  'next_resync' tells us
 	 * where that is.
@@ -135,20 +135,6 @@ struct r1bio {
 	/* DO NOT PUT ANY NEW FIELDS HERE - bios array is contiguously alloced*/
 };
 
-/* when we get a read error on a read-only array, we redirect to another
- * device without failing the first device, or trying to over-write to
- * correct the read error.  To keep track of bad blocks on a per-bio
- * level, we store IO_BLOCKED in the appropriate 'bios' pointer
- */
-#define IO_BLOCKED ((struct bio *)1)
-/* When we successfully write to a known bad-block, we need to remove the
- * bad-block marking which must be done from process context.  So we record
- * the success by setting bios[n] to IO_MADE_GOOD
- */
-#define IO_MADE_GOOD ((struct bio *)2)
-
-#define BIO_SPECIAL(bio) ((unsigned long)bio <= 2)
-
 /* bits for r1bio.state */
 #define	R1BIO_Uptodate	0
 #define	R1BIO_IsSync	1
diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
index 8da6282254c..e2549deab7c 100644
--- a/drivers/md/raid10.c
+++ b/drivers/md/raid10.c
@@ -60,7 +60,21 @@
  */
 #define	NR_RAID10_BIOS 256
 
-/* When there are this many requests queue to be written by
+/* when we get a read error on a read-only array, we redirect to another
+ * device without failing the first device, or trying to over-write to
+ * correct the read error.  To keep track of bad blocks on a per-bio
+ * level, we store IO_BLOCKED in the appropriate 'bios' pointer
+ */
+#define IO_BLOCKED ((struct bio *)1)
+/* When we successfully write to a known bad-block, we need to remove the
+ * bad-block marking which must be done from process context.  So we record
+ * the success by setting devs[n].bio to IO_MADE_GOOD
+ */
+#define IO_MADE_GOOD ((struct bio *)2)
+
+#define BIO_SPECIAL(bio) ((unsigned long)bio <= 2)
+
+/* When there are this many requests queued to be written by
  * the raid10 thread, we become 'congested' to provide back-pressure
  * for writeback.
  */
@@ -717,7 +731,7 @@ static struct md_rdev *read_balance(struct r10conf *conf,
 	int sectors = r10_bio->sectors;
 	int best_good_sectors;
 	sector_t new_distance, best_dist;
-	struct md_rdev *rdev, *best_rdev;
+	struct md_rdev *best_rdev, *rdev = NULL;
 	int do_balance;
 	int best_slot;
 	struct geom *geo = &conf->geo;
@@ -839,9 +853,8 @@ retry:
 	return rdev;
 }
 
-static int raid10_congested(void *data, int bits)
+int md_raid10_congested(struct mddev *mddev, int bits)
 {
-	struct mddev *mddev = data;
 	struct r10conf *conf = mddev->private;
 	int i, ret = 0;
 
@@ -849,8 +862,6 @@ static int raid10_congested(void *data, int bits)
 	    conf->pending_count >= max_queued_requests)
 		return 1;
 
-	if (mddev_congested(mddev, bits))
-		return 1;
 	rcu_read_lock();
 	for (i = 0;
 	     (i < conf->geo.raid_disks || i < conf->prev.raid_disks)
@@ -866,6 +877,15 @@ static int raid10_congested(void *data, int bits)
 	rcu_read_unlock();
 	return ret;
 }
+EXPORT_SYMBOL_GPL(md_raid10_congested);
+
+static int raid10_congested(void *data, int bits)
+{
+	struct mddev *mddev = data;
+
+	return mddev_congested(mddev, bits) ||
+		md_raid10_congested(mddev, bits);
+}
 
 static void flush_pending_writes(struct r10conf *conf)
 {
@@ -1546,7 +1566,7 @@ static void error(struct mddev *mddev, struct md_rdev *rdev)
 static void print_conf(struct r10conf *conf)
 {
 	int i;
-	struct mirror_info *tmp;
+	struct raid10_info *tmp;
 
 	printk(KERN_DEBUG "RAID10 conf printout:\n");
 	if (!conf) {
@@ -1580,7 +1600,7 @@ static int raid10_spare_active(struct mddev *mddev)
 {
 	int i;
 	struct r10conf *conf = mddev->private;
-	struct mirror_info *tmp;
+	struct raid10_info *tmp;
 	int count = 0;
 	unsigned long flags;
 
@@ -1655,7 +1675,7 @@ static int raid10_add_disk(struct mddev *mddev, struct md_rdev *rdev)
 	else
 		mirror = first;
 	for ( ; mirror <= last ; mirror++) {
-		struct mirror_info *p = &conf->mirrors[mirror];
+		struct raid10_info *p = &conf->mirrors[mirror];
 		if (p->recovery_disabled == mddev->recovery_disabled)
 			continue;
 		if (p->rdev) {
@@ -1709,7 +1729,7 @@ static int raid10_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
 	int err = 0;
 	int number = rdev->raid_disk;
 	struct md_rdev **rdevp;
-	struct mirror_info *p = conf->mirrors + number;
+	struct raid10_info *p = conf->mirrors + number;
 
 	print_conf(conf);
 	if (rdev == p->rdev)
@@ -2876,7 +2896,7 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr,
 			sector_t sect;
 			int must_sync;
 			int any_working;
-			struct mirror_info *mirror = &conf->mirrors[i];
+			struct raid10_info *mirror = &conf->mirrors[i];
 
 			if ((mirror->rdev == NULL ||
 			     test_bit(In_sync, &mirror->rdev->flags))
@@ -3388,7 +3408,7 @@ static struct r10conf *setup_conf(struct mddev *mddev)
 		goto out;
 
 	/* FIXME calc properly */
-	conf->mirrors = kzalloc(sizeof(struct mirror_info)*(mddev->raid_disks +
+	conf->mirrors = kzalloc(sizeof(struct raid10_info)*(mddev->raid_disks +
 							    max(0,mddev->delta_disks)),
 				GFP_KERNEL);
 	if (!conf->mirrors)
@@ -3452,7 +3472,7 @@ static int run(struct mddev *mddev)
 {
 	struct r10conf *conf;
 	int i, disk_idx, chunk_size;
-	struct mirror_info *disk;
+	struct raid10_info *disk;
 	struct md_rdev *rdev;
 	sector_t size;
 	sector_t min_offset_diff = 0;
@@ -3472,12 +3492,14 @@ static int run(struct mddev *mddev)
 	conf->thread = NULL;
 
 	chunk_size = mddev->chunk_sectors << 9;
-	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);
-	else
-		blk_queue_io_opt(mddev->queue, chunk_size *
-				 (conf->geo.raid_disks / conf->geo.near_copies));
+	if (mddev->queue) {
+		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);
+		else
+			blk_queue_io_opt(mddev->queue, chunk_size *
+					 (conf->geo.raid_disks / conf->geo.near_copies));
+	}
 
 	rdev_for_each(rdev, mddev) {
 		long long diff;
@@ -3511,8 +3533,9 @@ static int run(struct mddev *mddev)
 		if (first || diff < min_offset_diff)
 			min_offset_diff = diff;
 
-		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);
 
 		disk->head_position = 0;
 	}
@@ -3575,22 +3598,22 @@ static int run(struct mddev *mddev)
 	md_set_array_sectors(mddev, size);
 	mddev->resync_max_sectors = size;
 
-	mddev->queue->backing_dev_info.congested_fn = raid10_congested;
-	mddev->queue->backing_dev_info.congested_data = mddev;
-
-	/* Calculate max read-ahead size.
-	 * We need to readahead at least twice a whole stripe....
-	 * maybe...
-	 */
-	{
+	if (mddev->queue) {
 		int stripe = conf->geo.raid_disks *
 			((mddev->chunk_sectors << 9) / PAGE_SIZE);
+		mddev->queue->backing_dev_info.congested_fn = raid10_congested;
+		mddev->queue->backing_dev_info.congested_data = mddev;
+
+		/* Calculate max read-ahead size.
+		 * We need to readahead at least twice a whole stripe....
+		 * maybe...
+		 */
 		stripe /= conf->geo.near_copies;
 		if (mddev->queue->backing_dev_info.ra_pages < 2 * stripe)
 			mddev->queue->backing_dev_info.ra_pages = 2 * stripe;
+		blk_queue_merge_bvec(mddev->queue, raid10_mergeable_bvec);
 	}
 
-	blk_queue_merge_bvec(mddev->queue, raid10_mergeable_bvec);
 
 	if (md_integrity_register(mddev))
 		goto out_free_conf;
@@ -3641,7 +3664,10 @@ static int stop(struct mddev *mddev)
 	lower_barrier(conf);
 
 	md_unregister_thread(&mddev->thread);
-	blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
+	if (mddev->queue)
+		/* the unplug fn references 'conf'*/
+		blk_sync_queue(mddev->queue);
+
 	if (conf->r10bio_pool)
 		mempool_destroy(conf->r10bio_pool);
 	kfree(conf->mirrors);
@@ -3805,7 +3831,7 @@ static int raid10_check_reshape(struct mddev *mddev)
 	if (mddev->delta_disks > 0) {
 		/* allocate new 'mirrors' list */
 		conf->mirrors_new = kzalloc(
-			sizeof(struct mirror_info)
+			sizeof(struct raid10_info)
 			*(mddev->raid_disks +
 			  mddev->delta_disks),
 			GFP_KERNEL);
@@ -3930,7 +3956,7 @@ static int raid10_start_reshape(struct mddev *mddev)
 	spin_lock_irq(&conf->device_lock);
 	if (conf->mirrors_new) {
 		memcpy(conf->mirrors_new, conf->mirrors,
-		       sizeof(struct mirror_info)*conf->prev.raid_disks);
+		       sizeof(struct raid10_info)*conf->prev.raid_disks);
 		smp_mb();
 		kfree(conf->mirrors_old); /* FIXME and elsewhere */
 		conf->mirrors_old = conf->mirrors;
diff --git a/drivers/md/raid10.h b/drivers/md/raid10.h
index 135b1b0a155..007c2c68dd8 100644
--- a/drivers/md/raid10.h
+++ b/drivers/md/raid10.h
@@ -1,7 +1,7 @@
 #ifndef _RAID10_H
 #define _RAID10_H
 
-struct mirror_info {
+struct raid10_info {
 	struct md_rdev	*rdev, *replacement;
 	sector_t	head_position;
 	int		recovery_disabled;	/* matches
@@ -13,8 +13,8 @@ struct mirror_info {
 
 struct r10conf {
 	struct mddev		*mddev;
-	struct mirror_info	*mirrors;
-	struct mirror_info	*mirrors_new, *mirrors_old;
+	struct raid10_info	*mirrors;
+	struct raid10_info	*mirrors_new, *mirrors_old;
 	spinlock_t		device_lock;
 
 	/* geometry */
@@ -123,20 +123,6 @@ struct r10bio {
 	} devs[0];
 };
 
-/* when we get a read error on a read-only array, we redirect to another
- * device without failing the first device, or trying to over-write to
- * correct the read error.  To keep track of bad blocks on a per-bio
- * level, we store IO_BLOCKED in the appropriate 'bios' pointer
- */
-#define IO_BLOCKED ((struct bio*)1)
-/* When we successfully write to a known bad-block, we need to remove the
- * bad-block marking which must be done from process context.  So we record
- * the success by setting devs[n].bio to IO_MADE_GOOD
- */
-#define IO_MADE_GOOD ((struct bio *)2)
-
-#define BIO_SPECIAL(bio) ((unsigned long)bio <= 2)
-
 /* bits for r10bio.state */
 enum r10bio_state {
 	R10BIO_Uptodate,
@@ -159,4 +145,7 @@ enum r10bio_state {
  */
 	R10BIO_Previous,
 };
+
+extern int md_raid10_congested(struct mddev *mddev, int bits);
+
 #endif
diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index 04348d76bb3..259f519814c 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -99,34 +99,40 @@ static inline struct bio *r5_next_bio(struct bio *bio, sector_t sector)
  * We maintain a biased count of active stripes in the bottom 16 bits of
  * bi_phys_segments, and a count of processed stripes in the upper 16 bits
  */
-static inline int raid5_bi_phys_segments(struct bio *bio)
+static inline int raid5_bi_processed_stripes(struct bio *bio)
 {
-	return bio->bi_phys_segments & 0xffff;
+	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
+	return (atomic_read(segments) >> 16) & 0xffff;
 }
 
-static inline int raid5_bi_hw_segments(struct bio *bio)
+static inline int raid5_dec_bi_active_stripes(struct bio *bio)
 {
-	return (bio->bi_phys_segments >> 16) & 0xffff;
+	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
+	return atomic_sub_return(1, segments) & 0xffff;
 }
 
-static inline int raid5_dec_bi_phys_segments(struct bio *bio)
+static inline void raid5_inc_bi_active_stripes(struct bio *bio)
 {
-	--bio->bi_phys_segments;
-	return raid5_bi_phys_segments(bio);
+	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
+	atomic_inc(segments);
 }
 
-static inline int raid5_dec_bi_hw_segments(struct bio *bio)
+static inline void raid5_set_bi_processed_stripes(struct bio *bio,
+	unsigned int cnt)
 {
-	unsigned short val = raid5_bi_hw_segments(bio);
+	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
+	int old, new;
 
-	--val;
-	bio->bi_phys_segments = (val << 16) | raid5_bi_phys_segments(bio);
-	return val;
+	do {
+		old = atomic_read(segments);
+		new = (old & 0xffff) | (cnt << 16);
+	} while (atomic_cmpxchg(segments, old, new) != old);
 }
 
-static inline void raid5_set_bi_hw_segments(struct bio *bio, unsigned int cnt)
+static inline void raid5_set_bi_stripes(struct bio *bio, unsigned int cnt)
 {
-	bio->bi_phys_segments = raid5_bi_phys_segments(bio) | (cnt << 16);
+	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
+	atomic_set(segments, cnt);
 }
 
 /* Find first data disk in a raid6 stripe */
@@ -190,49 +196,56 @@ static int stripe_operations_active(struct stripe_head *sh)
 	       test_bit(STRIPE_COMPUTE_RUN, &sh->state);
 }
 
-static void __release_stripe(struct r5conf *conf, struct stripe_head *sh)
+static void do_release_stripe(struct r5conf *conf, struct stripe_head *sh)
 {
-	if (atomic_dec_and_test(&sh->count)) {
-		BUG_ON(!list_empty(&sh->lru));
-		BUG_ON(atomic_read(&conf->active_stripes)==0);
-		if (test_bit(STRIPE_HANDLE, &sh->state)) {
-			if (test_bit(STRIPE_DELAYED, &sh->state) &&
-			    !test_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
-				list_add_tail(&sh->lru, &conf->delayed_list);
-			else if (test_bit(STRIPE_BIT_DELAY, &sh->state) &&
-				   sh->bm_seq - conf->seq_write > 0)
-				list_add_tail(&sh->lru, &conf->bitmap_list);
-			else {
-				clear_bit(STRIPE_DELAYED, &sh->state);
-				clear_bit(STRIPE_BIT_DELAY, &sh->state);
-				list_add_tail(&sh->lru, &conf->handle_list);
-			}
-			md_wakeup_thread(conf->mddev->thread);
-		} else {
-			BUG_ON(stripe_operations_active(sh));
-			if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
-				if (atomic_dec_return(&conf->preread_active_stripes)
-				    < IO_THRESHOLD)
-					md_wakeup_thread(conf->mddev->thread);
-			atomic_dec(&conf->active_stripes);
-			if (!test_bit(STRIPE_EXPANDING, &sh->state)) {
-				list_add_tail(&sh->lru, &conf->inactive_list);
-				wake_up(&conf->wait_for_stripe);
-				if (conf->retry_read_aligned)
-					md_wakeup_thread(conf->mddev->thread);
-			}
+	BUG_ON(!list_empty(&sh->lru));
+	BUG_ON(atomic_read(&conf->active_stripes)==0);
+	if (test_bit(STRIPE_HANDLE, &sh->state)) {
+		if (test_bit(STRIPE_DELAYED, &sh->state) &&
+		    !test_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
+			list_add_tail(&sh->lru, &conf->delayed_list);
+		else if (test_bit(STRIPE_BIT_DELAY, &sh->state) &&
+			   sh->bm_seq - conf->seq_write > 0)
+			list_add_tail(&sh->lru, &conf->bitmap_list);
+		else {
+			clear_bit(STRIPE_DELAYED, &sh->state);
+			clear_bit(STRIPE_BIT_DELAY, &sh->state);
+			list_add_tail(&sh->lru, &conf->handle_list);
+		}
+		md_wakeup_thread(conf->mddev->thread);
+	} else {
+		BUG_ON(stripe_operations_active(sh));
+		if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
+			if (atomic_dec_return(&conf->preread_active_stripes)
+			    < IO_THRESHOLD)
+				md_wakeup_thread(conf->mddev->thread);
+		atomic_dec(&conf->active_stripes);
+		if (!test_bit(STRIPE_EXPANDING, &sh->state)) {
+			list_add_tail(&sh->lru, &conf->inactive_list);
+			wake_up(&conf->wait_for_stripe);
+			if (conf->retry_read_aligned)
+				md_wakeup_thread(conf->mddev->thread);
 		}
 	}
 }
 
+static void __release_stripe(struct r5conf *conf, struct stripe_head *sh)
+{
+	if (atomic_dec_and_test(&sh->count))
+		do_release_stripe(conf, sh);
+}
+
 static void release_stripe(struct stripe_head *sh)
 {
 	struct r5conf *conf = sh->raid_conf;
 	unsigned long flags;
 
-	spin_lock_irqsave(&conf->device_lock, flags);
-	__release_stripe(conf, sh);
-	spin_unlock_irqrestore(&conf->device_lock, flags);
+	local_irq_save(flags);
+	if (atomic_dec_and_lock(&sh->count, &conf->device_lock)) {
+		do_release_stripe(conf, sh);
+		spin_unlock(&conf->device_lock);
+	}
+	local_irq_restore(flags);
 }
 
 static inline void remove_hash(struct stripe_head *sh)
@@ -640,6 +653,9 @@ static void ops_run_io(struct stripe_head *sh, struct stripe_head_state *s)
 			else
 				bi->bi_sector = (sh->sector
 						 + rdev->data_offset);
+			if (test_bit(R5_ReadNoMerge, &sh->dev[i].flags))
+				bi->bi_rw |= REQ_FLUSH;
+
 			bi->bi_flags = 1 << BIO_UPTODATE;
 			bi->bi_idx = 0;
 			bi->bi_io_vec[0].bv_len = STRIPE_SIZE;
@@ -749,14 +765,12 @@ static void ops_complete_biofill(void *stripe_head_ref)
 {
 	struct stripe_head *sh = stripe_head_ref;
 	struct bio *return_bi = NULL;
-	struct r5conf *conf = sh->raid_conf;
 	int i;
 
 	pr_debug("%s: stripe %llu\n", __func__,
 		(unsigned long long)sh->sector);
 
 	/* clear completed biofills */
-	spin_lock_irq(&conf->device_lock);
 	for (i = sh->disks; i--; ) {
 		struct r5dev *dev = &sh->dev[i];
 
@@ -774,7 +788,7 @@ static void ops_complete_biofill(void *stripe_head_ref)
 			while (rbi && rbi->bi_sector <
 				dev->sector + STRIPE_SECTORS) {
 				rbi2 = r5_next_bio(rbi, dev->sector);
-				if (!raid5_dec_bi_phys_segments(rbi)) {
+				if (!raid5_dec_bi_active_stripes(rbi)) {
 					rbi->bi_next = return_bi;
 					return_bi = rbi;
 				}
@@ -782,7 +796,6 @@ static void ops_complete_biofill(void *stripe_head_ref)
 			}
 		}
 	}
-	spin_unlock_irq(&conf->device_lock);
 	clear_bit(STRIPE_BIOFILL_RUN, &sh->state);
 
 	return_io(return_bi);
@@ -794,7 +807,6 @@ static void ops_complete_biofill(void *stripe_head_ref)
 static void ops_run_biofill(struct stripe_head *sh)
 {
 	struct dma_async_tx_descriptor *tx = NULL;
-	struct r5conf *conf = sh->raid_conf;
 	struct async_submit_ctl submit;
 	int i;
 
@@ -805,10 +817,10 @@ static void ops_run_biofill(struct stripe_head *sh)
 		struct r5dev *dev = &sh->dev[i];
 		if (test_bit(R5_Wantfill, &dev->flags)) {
 			struct bio *rbi;
-			spin_lock_irq(&conf->device_lock);
+			spin_lock_irq(&sh->stripe_lock);
 			dev->read = rbi = dev->toread;
 			dev->toread = NULL;
-			spin_unlock_irq(&conf->device_lock);
+			spin_unlock_irq(&sh->stripe_lock);
 			while (rbi && rbi->bi_sector <
 				dev->sector + STRIPE_SECTORS) {
 				tx = async_copy_data(0, rbi, dev->page,
@@ -1144,12 +1156,12 @@ ops_run_biodrain(struct stripe_head *sh, struct dma_async_tx_descriptor *tx)
 		if (test_and_clear_bit(R5_Wantdrain, &dev->flags)) {
 			struct bio *wbi;
 
-			spin_lock_irq(&sh->raid_conf->device_lock);
+			spin_lock_irq(&sh->stripe_lock);
 			chosen = dev->towrite;
 			dev->towrite = NULL;
 			BUG_ON(dev->written);
 			wbi = dev->written = chosen;
-			spin_unlock_irq(&sh->raid_conf->device_lock);
+			spin_unlock_irq(&sh->stripe_lock);
 
 			while (wbi && wbi->bi_sector <
 				dev->sector + STRIPE_SECTORS) {
@@ -1454,6 +1466,8 @@ static int grow_one_stripe(struct r5conf *conf)
 	init_waitqueue_head(&sh->ops.wait_for_ops);
 	#endif
 
+	spin_lock_init(&sh->stripe_lock);
+
 	if (grow_buffers(sh)) {
 		shrink_buffers(sh);
 		kmem_cache_free(conf->slab_cache, sh);
@@ -1739,7 +1753,9 @@ static void raid5_end_read_request(struct bio * bi, int error)
 			atomic_add(STRIPE_SECTORS, &rdev->corrected_errors);
 			clear_bit(R5_ReadError, &sh->dev[i].flags);
 			clear_bit(R5_ReWrite, &sh->dev[i].flags);
-		}
+		} else if (test_bit(R5_ReadNoMerge, &sh->dev[i].flags))
+			clear_bit(R5_ReadNoMerge, &sh->dev[i].flags);
+
 		if (atomic_read(&rdev->read_errors))
 			atomic_set(&rdev->read_errors, 0);
 	} else {
@@ -1784,7 +1800,11 @@ static void raid5_end_read_request(struct bio * bi, int error)
 		else
 			retry = 1;
 		if (retry)
-			set_bit(R5_ReadError, &sh->dev[i].flags);
+			if (test_bit(R5_ReadNoMerge, &sh->dev[i].flags)) {
+				set_bit(R5_ReadError, &sh->dev[i].flags);
+				clear_bit(R5_ReadNoMerge, &sh->dev[i].flags);
+			} else
+				set_bit(R5_ReadNoMerge, &sh->dev[i].flags);
 		else {
 			clear_bit(R5_ReadError, &sh->dev[i].flags);
 			clear_bit(R5_ReWrite, &sh->dev[i].flags);
@@ -2340,11 +2360,18 @@ static int add_stripe_bio(struct stripe_head *sh, struct bio *bi, int dd_idx, in
 		(unsigned long long)bi->bi_sector,
 		(unsigned long long)sh->sector);
 
-
-	spin_lock_irq(&conf->device_lock);
+	/*
+	 * If several bio share a stripe. The bio bi_phys_segments acts as a
+	 * reference count to avoid race. The reference count should already be
+	 * increased before this function is called (for example, in
+	 * make_request()), so other bio sharing this stripe will not free the
+	 * stripe. If a stripe is owned by one stripe, the stripe lock will
+	 * protect it.
+	 */
+	spin_lock_irq(&sh->stripe_lock);
 	if (forwrite) {
 		bip = &sh->dev[dd_idx].towrite;
-		if (*bip == NULL && sh->dev[dd_idx].written == NULL)
+		if (*bip == NULL)
 			firstwrite = 1;
 	} else
 		bip = &sh->dev[dd_idx].toread;
@@ -2360,7 +2387,7 @@ static int add_stripe_bio(struct stripe_head *sh, struct bio *bi, int dd_idx, in
 	if (*bip)
 		bi->bi_next = *bip;
 	*bip = bi;
-	bi->bi_phys_segments++;
+	raid5_inc_bi_active_stripes(bi);
 
 	if (forwrite) {
 		/* check if page is covered */
@@ -2375,7 +2402,7 @@ static int add_stripe_bio(struct stripe_head *sh, struct bio *bi, int dd_idx, in
 		if (sector >= sh->dev[dd_idx].sector + STRIPE_SECTORS)
 			set_bit(R5_OVERWRITE, &sh->dev[dd_idx].flags);
 	}
-	spin_unlock_irq(&conf->device_lock);
+	spin_unlock_irq(&sh->stripe_lock);
 
 	pr_debug("added bi b#%llu to stripe s#%llu, disk %d.\n",
 		(unsigned long long)(*bip)->bi_sector,
@@ -2391,7 +2418,7 @@ static int add_stripe_bio(struct stripe_head *sh, struct bio *bi, int dd_idx, in
 
  overlap:
 	set_bit(R5_Overlap, &sh->dev[dd_idx].flags);
-	spin_unlock_irq(&conf->device_lock);
+	spin_unlock_irq(&sh->stripe_lock);
 	return 0;
 }
 
@@ -2441,10 +2468,11 @@ handle_failed_stripe(struct r5conf *conf, struct stripe_head *sh,
 				rdev_dec_pending(rdev, conf->mddev);
 			}
 		}
-		spin_lock_irq(&conf->device_lock);
+		spin_lock_irq(&sh->stripe_lock);
 		/* fail all writes first */
 		bi = sh->dev[i].towrite;
 		sh->dev[i].towrite = NULL;
+		spin_unlock_irq(&sh->stripe_lock);
 		if (bi) {
 			s->to_write--;
 			bitmap_end = 1;
@@ -2457,13 +2485,17 @@ handle_failed_stripe(struct r5conf *conf, struct stripe_head *sh,
 			sh->dev[i].sector + STRIPE_SECTORS) {
 			struct bio *nextbi = r5_next_bio(bi, sh->dev[i].sector);
 			clear_bit(BIO_UPTODATE, &bi->bi_flags);
-			if (!raid5_dec_bi_phys_segments(bi)) {
+			if (!raid5_dec_bi_active_stripes(bi)) {
 				md_write_end(conf->mddev);
 				bi->bi_next = *return_bi;
 				*return_bi = bi;
 			}
 			bi = nextbi;
 		}
+		if (bitmap_end)
+			bitmap_endwrite(conf->mddev->bitmap, sh->sector,
+				STRIPE_SECTORS, 0, 0);
+		bitmap_end = 0;
 		/* and fail all 'written' */
 		bi = sh->dev[i].written;
 		sh->dev[i].written = NULL;
@@ -2472,7 +2504,7 @@ handle_failed_stripe(struct r5conf *conf, struct stripe_head *sh,
 		       sh->dev[i].sector + STRIPE_SECTORS) {
 			struct bio *bi2 = r5_next_bio(bi, sh->dev[i].sector);
 			clear_bit(BIO_UPTODATE, &bi->bi_flags);
-			if (!raid5_dec_bi_phys_segments(bi)) {
+			if (!raid5_dec_bi_active_stripes(bi)) {
 				md_write_end(conf->mddev);
 				bi->bi_next = *return_bi;
 				*return_bi = bi;
@@ -2496,14 +2528,13 @@ handle_failed_stripe(struct r5conf *conf, struct stripe_head *sh,
 				struct bio *nextbi =
 					r5_next_bio(bi, sh->dev[i].sector);
 				clear_bit(BIO_UPTODATE, &bi->bi_flags);
-				if (!raid5_dec_bi_phys_segments(bi)) {
+				if (!raid5_dec_bi_active_stripes(bi)) {
 					bi->bi_next = *return_bi;
 					*return_bi = bi;
 				}
 				bi = nextbi;
 			}
 		}
-		spin_unlock_irq(&conf->device_lock);
 		if (bitmap_end)
 			bitmap_endwrite(conf->mddev->bitmap, sh->sector,
 					STRIPE_SECTORS, 0, 0);
@@ -2707,30 +2738,23 @@ static void handle_stripe_clean_event(struct r5conf *conf,
 				test_bit(R5_UPTODATE, &dev->flags)) {
 				/* We can return any write requests */
 				struct bio *wbi, *wbi2;
-				int bitmap_end = 0;
 				pr_debug("Return write for disc %d\n", i);
-				spin_lock_irq(&conf->device_lock);
 				wbi = dev->written;
 				dev->written = NULL;
 				while (wbi && wbi->bi_sector <
 					dev->sector + STRIPE_SECTORS) {
 					wbi2 = r5_next_bio(wbi, dev->sector);
-					if (!raid5_dec_bi_phys_segments(wbi)) {
+					if (!raid5_dec_bi_active_stripes(wbi)) {
 						md_write_end(conf->mddev);
 						wbi->bi_next = *return_bi;
 						*return_bi = wbi;
 					}
 					wbi = wbi2;
 				}
-				if (dev->towrite == NULL)
-					bitmap_end = 1;
-				spin_unlock_irq(&conf->device_lock);
-				if (bitmap_end)
-					bitmap_endwrite(conf->mddev->bitmap,
-							sh->sector,
-							STRIPE_SECTORS,
+				bitmap_endwrite(conf->mddev->bitmap, sh->sector,
+						STRIPE_SECTORS,
 					 !test_bit(STRIPE_DEGRADED, &sh->state),
-							0);
+						0);
 			}
 		}
 
@@ -3182,7 +3206,6 @@ static void analyse_stripe(struct stripe_head *sh, struct stripe_head_state *s)
 
 	/* Now to look around and see what can be done */
 	rcu_read_lock();
-	spin_lock_irq(&conf->device_lock);
 	for (i=disks; i--; ) {
 		struct md_rdev *rdev;
 		sector_t first_bad;
@@ -3328,7 +3351,6 @@ static void analyse_stripe(struct stripe_head *sh, struct stripe_head_state *s)
 				do_recovery = 1;
 		}
 	}
-	spin_unlock_irq(&conf->device_lock);
 	if (test_bit(STRIPE_SYNCING, &sh->state)) {
 		/* If there is a failed device being replaced,
 		 *     we must be recovering.
@@ -3791,7 +3813,7 @@ static struct bio *remove_bio_from_retry(struct r5conf *conf)
 		 * this sets the active strip count to 1 and the processed
 		 * strip count to zero (upper 8 bits)
 		 */
-		bi->bi_phys_segments = 1; /* biased count of active stripes */
+		raid5_set_bi_stripes(bi, 1); /* biased count of active stripes */
 	}
 
 	return bi;
@@ -4113,7 +4135,7 @@ static void make_request(struct mddev *mddev, struct bio * bi)
 			finish_wait(&conf->wait_for_overlap, &w);
 			set_bit(STRIPE_HANDLE, &sh->state);
 			clear_bit(STRIPE_DELAYED, &sh->state);
-			if ((bi->bi_rw & REQ_SYNC) &&
+			if ((bi->bi_rw & REQ_NOIDLE) &&
 			    !test_and_set_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
 				atomic_inc(&conf->preread_active_stripes);
 			mddev_check_plugged(mddev);
@@ -4126,9 +4148,7 @@ static void make_request(struct mddev *mddev, struct bio * bi)
 		}
 	}
 
-	spin_lock_irq(&conf->device_lock);
-	remaining = raid5_dec_bi_phys_segments(bi);
-	spin_unlock_irq(&conf->device_lock);
+	remaining = raid5_dec_bi_active_stripes(bi);
 	if (remaining == 0) {
 
 		if ( rw == WRITE )
@@ -4484,7 +4504,7 @@ static int  retry_aligned_read(struct r5conf *conf, struct bio *raid_bio)
 		     sector += STRIPE_SECTORS,
 		     scnt++) {
 
-		if (scnt < raid5_bi_hw_segments(raid_bio))
+		if (scnt < raid5_bi_processed_stripes(raid_bio))
 			/* already done this stripe */
 			continue;
 
@@ -4492,25 +4512,24 @@ static int  retry_aligned_read(struct r5conf *conf, struct bio *raid_bio)
 
 		if (!sh) {
 			/* failed to get a stripe - must wait */
-			raid5_set_bi_hw_segments(raid_bio, scnt);
+			raid5_set_bi_processed_stripes(raid_bio, scnt);
 			conf->retry_read_aligned = raid_bio;
 			return handled;
 		}
 
 		if (!add_stripe_bio(sh, raid_bio, dd_idx, 0)) {
 			release_stripe(sh);
-			raid5_set_bi_hw_segments(raid_bio, scnt);
+			raid5_set_bi_processed_stripes(raid_bio, scnt);
 			conf->retry_read_aligned = raid_bio;
 			return handled;
 		}
 
+		set_bit(R5_ReadNoMerge, &sh->dev[dd_idx].flags);
 		handle_stripe(sh);
 		release_stripe(sh);
 		handled++;
 	}
-	spin_lock_irq(&conf->device_lock);
-	remaining = raid5_dec_bi_phys_segments(raid_bio);
-	spin_unlock_irq(&conf->device_lock);
+	remaining = raid5_dec_bi_active_stripes(raid_bio);
 	if (remaining == 0)
 		bio_endio(raid_bio, 0);
 	if (atomic_dec_and_test(&conf->active_aligned_reads))
diff --git a/drivers/md/raid5.h b/drivers/md/raid5.h
index 2164021f3b5..61dbb615c30 100644
--- a/drivers/md/raid5.h
+++ b/drivers/md/raid5.h
@@ -210,6 +210,7 @@ struct stripe_head {
 	int			disks;		/* disks in stripe */
 	enum check_states	check_state;
 	enum reconstruct_states reconstruct_state;
+	spinlock_t		stripe_lock;
 	/**
 	 * struct stripe_operations
 	 * @target - STRIPE_OP_COMPUTE_BLK target
@@ -273,6 +274,7 @@ enum r5dev_flags {
 	R5_Wantwrite,
 	R5_Overlap,	/* There is a pending overlapping request
 			 * on this block */
+	R5_ReadNoMerge, /* prevent bio from merging in block-layer */
 	R5_ReadError,	/* seen a read error here recently */
 	R5_ReWrite,	/* have tried to over-write the readerror */