Merge tag 'md-3.3' of git://neil.brown.name/md

md update for 3.3

Big change is new hot-replacement.
A slot in an array can hold 2 devices - one that
wants-replacement and one that is the replacement.
Once the replacement is built - either from the
original or (in the case of errors) from elsewhere,
the wants-replacement device will be removed.

* tag 'md-3.3' of git://neil.brown.name/md: (36 commits)
  md/raid1: Mark device want_replacement when we see a write error.
  md/raid1: If there is a spare and a want_replacement device, start replacement.
  md/raid1: recognise replacements when assembling arrays.
  md/raid1: handle activation of replacement device when recovery completes.
  md/raid1: Allow a failed replacement device to be removed.
  md/raid1: Allocate spare to store replacement devices and their bios.
  md/raid1:  Replace use of mddev->raid_disks with conf->raid_disks.
  md/raid10: If there is a spare and a want_replacement device, start replacement.
  md/raid10: recognise replacements when assembling array.
  md/raid10: Allow replacement device to be replace old drive.
  md/raid10: handle recovery of replacement devices.
  md/raid10:  Handle replacement devices during resync.
  md/raid10: writes should get directed to replacement as well as original.
  md/raid10: allow removal of failed replacement devices.
  md/raid10: preferentially read from replacement device if possible.
  md/raid10:  change read_balance to return an rdev
  md/raid10: prepare data structures for handling replacement.
  md/raid5: Mark device want_replacement when we see a write error.
  md/raid5: If there is a spare and a want_replacement device, start replacement.
  md/raid5: recognise replacements when assembling array.
  ...

Documentation/md.txt

@@ -357,14 +357,14 @@ Each directory contains:
         written to, that device.
 
       state
-        A file recording the current state of the device in the array
+	A file recording the current state of the device in the array
 	which can be a comma separated list of
 	      faulty   - device has been kicked from active use due to
-                         a detected fault or it has unacknowledged bad
-                         blocks
+			 a detected fault, or it has unacknowledged bad
+			 blocks
 	      in_sync  - device is a fully in-sync member of the array
 	      writemostly - device will only be subject to read
-		         requests if there are no other options.
+			 requests if there are no other options.
 			 This applies only to raid1 arrays.
 	      blocked  - device has failed, and the failure hasn't been
 			 acknowledged yet by the metadata handler.
@@ -374,6 +374,13 @@ Each directory contains:
 			 This includes spares that are in the process
 			 of being recovered to
 	      write_error - device has ever seen a write error.
+	      want_replacement - device is (mostly) working but probably
+			 should be replaced, either due to errors or
+			 due to user request.
+	      replacement - device is a replacement for another active
+			 device with same raid_disk.
+
+
 	This list may grow in future.
 	This can be written to.
 	Writing "faulty"  simulates a failure on the device.
@@ -386,6 +393,13 @@ Each directory contains:
 	Writing "in_sync" sets the in_sync flag.
 	Writing "write_error" sets writeerrorseen flag.
 	Writing "-write_error" clears writeerrorseen flag.
+	Writing "want_replacement" is allowed at any time except to a
+		replacement device or a spare.  It sets the flag.
+	Writing "-want_replacement" is allowed at any time.  It clears
+		the flag.
+	Writing "replacement" or "-replacement" is only allowed before
+		starting the array.  It sets or clears the flag.
+
 
 	This file responds to select/poll. Any change to 'faulty'
 	or 'blocked' causes an event.

drivers/md/bitmap.c

@@ -1149,12 +1149,12 @@ void bitmap_daemon_work(struct mddev *mddev)
 		return;
 	}
 	if (time_before(jiffies, bitmap->daemon_lastrun
-			+ bitmap->mddev->bitmap_info.daemon_sleep))
+			+ mddev->bitmap_info.daemon_sleep))
 		goto done;
 
 	bitmap->daemon_lastrun = jiffies;
 	if (bitmap->allclean) {
-		bitmap->mddev->thread->timeout = MAX_SCHEDULE_TIMEOUT;
+		mddev->thread->timeout = MAX_SCHEDULE_TIMEOUT;
 		goto done;
 	}
 	bitmap->allclean = 1;
@@ -1206,7 +1206,7 @@ void bitmap_daemon_work(struct mddev *mddev)
 			 * sure that events_cleared is up-to-date.
 			 */
 			if (bitmap->need_sync &&
-			    bitmap->mddev->bitmap_info.external == 0) {
+			    mddev->bitmap_info.external == 0) {
 				bitmap_super_t *sb;
 				bitmap->need_sync = 0;
 				sb = kmap_atomic(bitmap->sb_page, KM_USER0);
@@ -1270,8 +1270,8 @@ void bitmap_daemon_work(struct mddev *mddev)
 
  done:
 	if (bitmap->allclean == 0)
-		bitmap->mddev->thread->timeout =
-			bitmap->mddev->bitmap_info.daemon_sleep;
+		mddev->thread->timeout =
+			mddev->bitmap_info.daemon_sleep;
 	mutex_unlock(&mddev->bitmap_info.mutex);
 }
 
@@ -1587,7 +1587,7 @@ static void bitmap_set_memory_bits(struct bitmap *bitmap, sector_t offset, int n
 	}
 	if (!*bmc) {
 		struct page *page;
-		*bmc = 1 | (needed ? NEEDED_MASK : 0);
+		*bmc = 2 | (needed ? NEEDED_MASK : 0);
 		bitmap_count_page(bitmap, offset, 1);
 		page = filemap_get_page(bitmap, offset >> CHUNK_BLOCK_SHIFT(bitmap));
 		set_page_attr(bitmap, page, BITMAP_PAGE_PENDING);

drivers/md/md.c

@@ -1713,6 +1713,8 @@ static int super_1_validate(struct mddev *mddev, struct md_rdev *rdev)
 		}
 		if (sb->devflags & WriteMostly1)
 			set_bit(WriteMostly, &rdev->flags);
+		if (le32_to_cpu(sb->feature_map) & MD_FEATURE_REPLACEMENT)
+			set_bit(Replacement, &rdev->flags);
 	} else /* MULTIPATH are always insync */
 		set_bit(In_sync, &rdev->flags);
 
@@ -1766,6 +1768,9 @@ static void super_1_sync(struct mddev *mddev, struct md_rdev *rdev)
 		sb->recovery_offset =
 			cpu_to_le64(rdev->recovery_offset);
 	}
+	if (test_bit(Replacement, &rdev->flags))
+		sb->feature_map |=
+			cpu_to_le32(MD_FEATURE_REPLACEMENT);
 
 	if (mddev->reshape_position != MaxSector) {
 		sb->feature_map |= cpu_to_le32(MD_FEATURE_RESHAPE_ACTIVE);
@@ -2559,6 +2564,15 @@ state_show(struct md_rdev *rdev, char *page)
 		len += sprintf(page+len, "%swrite_error", sep);
 		sep = ",";
 	}
+	if (test_bit(WantReplacement, &rdev->flags)) {
+		len += sprintf(page+len, "%swant_replacement", sep);
+		sep = ",";
+	}
+	if (test_bit(Replacement, &rdev->flags)) {
+		len += sprintf(page+len, "%sreplacement", sep);
+		sep = ",";
+	}
+
 	return len+sprintf(page+len, "\n");
 }
 
@@ -2627,6 +2641,42 @@ state_store(struct md_rdev *rdev, const char *buf, size_t len)
 	} else if (cmd_match(buf, "-write_error")) {
 		clear_bit(WriteErrorSeen, &rdev->flags);
 		err = 0;
+	} else if (cmd_match(buf, "want_replacement")) {
+		/* Any non-spare device that is not a replacement can
+		 * become want_replacement at any time, but we then need to
+		 * check if recovery is needed.
+		 */
+		if (rdev->raid_disk >= 0 &&
+		    !test_bit(Replacement, &rdev->flags))
+			set_bit(WantReplacement, &rdev->flags);
+		set_bit(MD_RECOVERY_NEEDED, &rdev->mddev->recovery);
+		md_wakeup_thread(rdev->mddev->thread);
+		err = 0;
+	} else if (cmd_match(buf, "-want_replacement")) {
+		/* Clearing 'want_replacement' is always allowed.
+		 * Once replacements starts it is too late though.
+		 */
+		err = 0;
+		clear_bit(WantReplacement, &rdev->flags);
+	} else if (cmd_match(buf, "replacement")) {
+		/* Can only set a device as a replacement when array has not
+		 * yet been started.  Once running, replacement is automatic
+		 * from spares, or by assigning 'slot'.
+		 */
+		if (rdev->mddev->pers)
+			err = -EBUSY;
+		else {
+			set_bit(Replacement, &rdev->flags);
+			err = 0;
+		}
+	} else if (cmd_match(buf, "-replacement")) {
+		/* Similarly, can only clear Replacement before start */
+		if (rdev->mddev->pers)
+			err = -EBUSY;
+		else {
+			clear_bit(Replacement, &rdev->flags);
+			err = 0;
+		}
 	}
 	if (!err)
 		sysfs_notify_dirent_safe(rdev->sysfs_state);
@@ -2688,7 +2738,7 @@ slot_store(struct md_rdev *rdev, const char *buf, size_t len)
 		if (rdev->mddev->pers->hot_remove_disk == NULL)
 			return -EINVAL;
 		err = rdev->mddev->pers->
-			hot_remove_disk(rdev->mddev, rdev->raid_disk);
+			hot_remove_disk(rdev->mddev, rdev);
 		if (err)
 			return err;
 		sysfs_unlink_rdev(rdev->mddev, rdev);
@@ -2696,7 +2746,6 @@ slot_store(struct md_rdev *rdev, const char *buf, size_t len)
 		set_bit(MD_RECOVERY_NEEDED, &rdev->mddev->recovery);
 		md_wakeup_thread(rdev->mddev->thread);
 	} else if (rdev->mddev->pers) {
-		struct md_rdev *rdev2;
 		/* Activating a spare .. or possibly reactivating
 		 * if we ever get bitmaps working here.
 		 */
@@ -2710,10 +2759,6 @@ slot_store(struct md_rdev *rdev, const char *buf, size_t len)
 		if (rdev->mddev->pers->hot_add_disk == NULL)
 			return -EINVAL;
 
-		list_for_each_entry(rdev2, &rdev->mddev->disks, same_set)
-			if (rdev2->raid_disk == slot)
-				return -EEXIST;
-
 		if (slot >= rdev->mddev->raid_disks &&
 		    slot >= rdev->mddev->raid_disks + rdev->mddev->delta_disks)
 			return -ENOSPC;
@@ -6053,8 +6098,15 @@ static int md_ioctl(struct block_device *bdev, fmode_t mode,
 	struct mddev *mddev = NULL;
 	int ro;
 
-	if (!capable(CAP_SYS_ADMIN))
-		return -EACCES;
+	switch (cmd) {
+	case RAID_VERSION:
+	case GET_ARRAY_INFO:
+	case GET_DISK_INFO:
+		break;
+	default:
+		if (!capable(CAP_SYS_ADMIN))
+			return -EACCES;
+	}
 
 	/*
 	 * Commands dealing with the RAID driver but not any
@@ -6714,8 +6766,11 @@ static int md_seq_show(struct seq_file *seq, void *v)
 			if (test_bit(Faulty, &rdev->flags)) {
 				seq_printf(seq, "(F)");
 				continue;
-			} else if (rdev->raid_disk < 0)
+			}
+			if (rdev->raid_disk < 0)
 				seq_printf(seq, "(S)"); /* spare */
+			if (test_bit(Replacement, &rdev->flags))
+				seq_printf(seq, "(R)");
 			sectors += rdev->sectors;
 		}
 
@@ -7337,29 +7392,27 @@ static int remove_and_add_spares(struct mddev *mddev)
 		     ! test_bit(In_sync, &rdev->flags)) &&
 		    atomic_read(&rdev->nr_pending)==0) {
 			if (mddev->pers->hot_remove_disk(
-				    mddev, rdev->raid_disk)==0) {
+				    mddev, rdev) == 0) {
 				sysfs_unlink_rdev(mddev, rdev);
 				rdev->raid_disk = -1;
 			}
 		}
 
-	if (mddev->degraded) {
-		list_for_each_entry(rdev, &mddev->disks, same_set) {
-			if (rdev->raid_disk >= 0 &&
-			    !test_bit(In_sync, &rdev->flags) &&
-			    !test_bit(Faulty, &rdev->flags))
+	list_for_each_entry(rdev, &mddev->disks, same_set) {
+		if (rdev->raid_disk >= 0 &&
+		    !test_bit(In_sync, &rdev->flags) &&
+		    !test_bit(Faulty, &rdev->flags))
+			spares++;
+		if (rdev->raid_disk < 0
+		    && !test_bit(Faulty, &rdev->flags)) {
+			rdev->recovery_offset = 0;
+			if (mddev->pers->
+			    hot_add_disk(mddev, rdev) == 0) {
+				if (sysfs_link_rdev(mddev, rdev))
+					/* failure here is OK */;
 				spares++;
-			if (rdev->raid_disk < 0
-			    && !test_bit(Faulty, &rdev->flags)) {
-				rdev->recovery_offset = 0;
-				if (mddev->pers->
-				    hot_add_disk(mddev, rdev) == 0) {
-					if (sysfs_link_rdev(mddev, rdev))
-						/* failure here is OK */;
-					spares++;
-					md_new_event(mddev);
-					set_bit(MD_CHANGE_DEVS, &mddev->flags);
-				}
+				md_new_event(mddev);
+				set_bit(MD_CHANGE_DEVS, &mddev->flags);
 			}
 		}
 	}
@@ -7474,7 +7527,7 @@ void md_check_recovery(struct mddev *mddev)
 				    test_bit(Faulty, &rdev->flags) &&
 				    atomic_read(&rdev->nr_pending)==0) {
 					if (mddev->pers->hot_remove_disk(
-						    mddev, rdev->raid_disk)==0) {
+						    mddev, rdev) == 0) {
 						sysfs_unlink_rdev(mddev, rdev);
 						rdev->raid_disk = -1;
 					}

drivers/md/md.h

@@ -72,34 +72,7 @@ struct md_rdev {
 	 * This reduces the burden of testing multiple flags in many cases
 	 */
 
-	unsigned long	flags;
-#define	Faulty		1		/* device is known to have a fault */
-#define	In_sync		2		/* device is in_sync with rest of array */
-#define	WriteMostly	4		/* Avoid reading if at all possible */
-#define	AutoDetected	7		/* added by auto-detect */
-#define Blocked		8		/* An error occurred but has not yet
-					 * been acknowledged by the metadata
-					 * handler, so don't allow writes
-					 * until it is cleared */
-#define WriteErrorSeen	9		/* A write error has been seen on this
-					 * device
-					 */
-#define FaultRecorded	10		/* Intermediate state for clearing
-					 * Blocked.  The Fault is/will-be
-					 * recorded in the metadata, but that
-					 * metadata hasn't been stored safely
-					 * on disk yet.
-					 */
-#define BlockedBadBlocks 11		/* A writer is blocked because they
-					 * found an unacknowledged bad-block.
-					 * This can safely be cleared at any
-					 * time, and the writer will re-check.
-					 * It may be set at any time, and at
-					 * worst the writer will timeout and
-					 * re-check.  So setting it as
-					 * accurately as possible is good, but
-					 * not absolutely critical.
-					 */
+	unsigned long	flags;	/* bit set of 'enum flag_bits' bits. */
 	wait_queue_head_t blocked_wait;
 
 	int desc_nr;			/* descriptor index in the superblock */
@@ -152,6 +125,44 @@ struct md_rdev {
 		sector_t size;		/* in sectors */
 	} badblocks;
 };
+enum flag_bits {
+	Faulty,			/* device is known to have a fault */
+	In_sync,		/* device is in_sync with rest of array */
+	WriteMostly,		/* Avoid reading if at all possible */
+	AutoDetected,		/* added by auto-detect */
+	Blocked,		/* An error occurred but has not yet
+				 * been acknowledged by the metadata
+				 * handler, so don't allow writes
+				 * until it is cleared */
+	WriteErrorSeen,		/* A write error has been seen on this
+				 * device
+				 */
+	FaultRecorded,		/* Intermediate state for clearing
+				 * Blocked.  The Fault is/will-be
+				 * recorded in the metadata, but that
+				 * metadata hasn't been stored safely
+				 * on disk yet.
+				 */
+	BlockedBadBlocks,	/* A writer is blocked because they
+				 * found an unacknowledged bad-block.
+				 * This can safely be cleared at any
+				 * time, and the writer will re-check.
+				 * It may be set at any time, and at
+				 * worst the writer will timeout and
+				 * re-check.  So setting it as
+				 * accurately as possible is good, but
+				 * not absolutely critical.
+				 */
+	WantReplacement,	/* This device is a candidate to be
+				 * hot-replaced, either because it has
+				 * reported some faults, or because
+				 * of explicit request.
+				 */
+	Replacement,		/* This device is a replacement for
+				 * a want_replacement device with same
+				 * raid_disk number.
+				 */
+};
 
 #define BB_LEN_MASK	(0x00000000000001FFULL)
 #define BB_OFFSET_MASK	(0x7FFFFFFFFFFFFE00ULL)
@@ -428,7 +439,7 @@ struct md_personality
 	 */
 	void (*error_handler)(struct mddev *mddev, struct md_rdev *rdev);
 	int (*hot_add_disk) (struct mddev *mddev, struct md_rdev *rdev);
-	int (*hot_remove_disk) (struct mddev *mddev, int number);
+	int (*hot_remove_disk) (struct mddev *mddev, struct md_rdev *rdev);
 	int (*spare_active) (struct mddev *mddev);
 	sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, int *skipped, int go_faster);
 	int (*resize) (struct mddev *mddev, sector_t sectors);
@@ -482,15 +493,20 @@ static inline char * mdname (struct mddev * mddev)
 static inline int sysfs_link_rdev(struct mddev *mddev, struct md_rdev *rdev)
 {
 	char nm[20];
-	sprintf(nm, "rd%d", rdev->raid_disk);
-	return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm);
+	if (!test_bit(Replacement, &rdev->flags)) {
+		sprintf(nm, "rd%d", rdev->raid_disk);
+		return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm);
+	} else
+		return 0;
 }
 
 static inline void sysfs_unlink_rdev(struct mddev *mddev, struct md_rdev *rdev)
 {
 	char nm[20];
-	sprintf(nm, "rd%d", rdev->raid_disk);
-	sysfs_remove_link(&mddev->kobj, nm);
+	if (!test_bit(Replacement, &rdev->flags)) {
+		sprintf(nm, "rd%d", rdev->raid_disk);
+		sysfs_remove_link(&mddev->kobj, nm);
+	}
 }
 
 /*

drivers/md/multipath.c

@@ -292,17 +292,16 @@ static int multipath_add_disk(struct mddev *mddev, struct md_rdev *rdev)
 	return err;
 }
 
-static int multipath_remove_disk(struct mddev *mddev, int number)
+static int multipath_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
 {
 	struct mpconf *conf = mddev->private;
 	int err = 0;
-	struct md_rdev *rdev;
+	int number = rdev->raid_disk;
 	struct multipath_info *p = conf->multipaths + number;
 
 	print_multipath_conf(conf);
 
-	rdev = p->rdev;
-	if (rdev) {
+	if (rdev == p->rdev) {
 		if (test_bit(In_sync, &rdev->flags) ||
 		    atomic_read(&rdev->nr_pending)) {
 			printk(KERN_ERR "hot-remove-disk, slot %d is identified"

drivers/md/raid1.c

@@ -135,7 +135,7 @@ static void * r1buf_pool_alloc(gfp_t gfp_flags, void *data)
 			put_page(r1_bio->bios[j]->bi_io_vec[i].bv_page);
 	j = -1;
 out_free_bio:
-	while ( ++j < pi->raid_disks )
+	while (++j < pi->raid_disks)
 		bio_put(r1_bio->bios[j]);
 	r1bio_pool_free(r1_bio, data);
 	return NULL;
@@ -164,7 +164,7 @@ static void put_all_bios(struct r1conf *conf, struct r1bio *r1_bio)
 {
 	int i;
 
-	for (i = 0; i < conf->raid_disks; i++) {
+	for (i = 0; i < conf->raid_disks * 2; i++) {
 		struct bio **bio = r1_bio->bios + i;
 		if (!BIO_SPECIAL(*bio))
 			bio_put(*bio);
@@ -185,7 +185,7 @@ static void put_buf(struct r1bio *r1_bio)
 	struct r1conf *conf = r1_bio->mddev->private;
 	int i;
 
-	for (i=0; i<conf->raid_disks; i++) {
+	for (i = 0; i < conf->raid_disks * 2; i++) {
 		struct bio *bio = r1_bio->bios[i];
 		if (bio->bi_end_io)
 			rdev_dec_pending(conf->mirrors[i].rdev, r1_bio->mddev);
@@ -277,13 +277,14 @@ static inline void update_head_pos(int disk, struct r1bio *r1_bio)
 static int find_bio_disk(struct r1bio *r1_bio, struct bio *bio)
 {
 	int mirror;
-	int raid_disks = r1_bio->mddev->raid_disks;
+	struct r1conf *conf = r1_bio->mddev->private;
+	int raid_disks = conf->raid_disks;
 
-	for (mirror = 0; mirror < raid_disks; mirror++)
+	for (mirror = 0; mirror < raid_disks * 2; mirror++)
 		if (r1_bio->bios[mirror] == bio)
 			break;
 
-	BUG_ON(mirror == raid_disks);
+	BUG_ON(mirror == raid_disks * 2);
 	update_head_pos(mirror, r1_bio);
 
 	return mirror;
@@ -390,6 +391,11 @@ static void raid1_end_write_request(struct bio *bio, int error)
 	if (!uptodate) {
 		set_bit(WriteErrorSeen,
 			&conf->mirrors[mirror].rdev->flags);
+		if (!test_and_set_bit(WantReplacement,
+				      &conf->mirrors[mirror].rdev->flags))
+			set_bit(MD_RECOVERY_NEEDED, &
+				conf->mddev->recovery);
+
 		set_bit(R1BIO_WriteError, &r1_bio->state);
 	} else {
 		/*
@@ -505,7 +511,7 @@ static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sect
 		start_disk = conf->last_used;
 	}
 
-	for (i = 0 ; i < conf->raid_disks ; i++) {
+	for (i = 0 ; i < conf->raid_disks * 2 ; i++) {
 		sector_t dist;
 		sector_t first_bad;
 		int bad_sectors;
@@ -609,7 +615,7 @@ int md_raid1_congested(struct mddev *mddev, int bits)
 		return 1;
 
 	rcu_read_lock();
-	for (i = 0; i < mddev->raid_disks; i++) {
+	for (i = 0; i < conf->raid_disks; 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);
@@ -974,7 +980,7 @@ static void make_request(struct mddev *mddev, struct bio * bio)
 	 */
 	plugged = mddev_check_plugged(mddev);
 
-	disks = conf->raid_disks;
+	disks = conf->raid_disks * 2;
  retry_write:
 	blocked_rdev = NULL;
 	rcu_read_lock();
@@ -988,7 +994,8 @@ static void make_request(struct mddev *mddev, struct bio * bio)
 		}
 		r1_bio->bios[i] = NULL;
 		if (!rdev || test_bit(Faulty, &rdev->flags)) {
-			set_bit(R1BIO_Degraded, &r1_bio->state);
+			if (i < conf->raid_disks)
+				set_bit(R1BIO_Degraded, &r1_bio->state);
 			continue;
 		}
 
@@ -1263,6 +1270,25 @@ static int raid1_spare_active(struct mddev *mddev)
 	 */
 	for (i = 0; i < conf->raid_disks; i++) {
 		struct md_rdev *rdev = conf->mirrors[i].rdev;
+		struct md_rdev *repl = conf->mirrors[conf->raid_disks + i].rdev;
+		if (repl
+		    && repl->recovery_offset == MaxSector
+		    && !test_bit(Faulty, &repl->flags)
+		    && !test_and_set_bit(In_sync, &repl->flags)) {
+			/* replacement has just become active */
+			if (!rdev ||
+			    !test_and_clear_bit(In_sync, &rdev->flags))
+				count++;
+			if (rdev) {
+				/* Replaced device not technically
+				 * faulty, but we need to be sure
+				 * it gets removed and never re-added
+				 */
+				set_bit(Faulty, &rdev->flags);
+				sysfs_notify_dirent_safe(
+					rdev->sysfs_state);
+			}
+		}
 		if (rdev
 		    && !test_bit(Faulty, &rdev->flags)
 		    && !test_and_set_bit(In_sync, &rdev->flags)) {
@@ -1286,7 +1312,7 @@ static int raid1_add_disk(struct mddev *mddev, struct md_rdev *rdev)
 	int mirror = 0;
 	struct mirror_info *p;
 	int first = 0;
-	int last = mddev->raid_disks - 1;
+	int last = conf->raid_disks - 1;
 
 	if (mddev->recovery_disabled == conf->recovery_disabled)
 		return -EBUSY;
@@ -1294,8 +1320,9 @@ static int raid1_add_disk(struct mddev *mddev, struct md_rdev *rdev)
 	if (rdev->raid_disk >= 0)
 		first = last = rdev->raid_disk;
 
-	for (mirror = first; mirror <= last; mirror++)
-		if ( !(p=conf->mirrors+mirror)->rdev) {
+	for (mirror = first; mirror <= last; mirror++) {
+		p = conf->mirrors+mirror;
+		if (!p->rdev) {
 
 			disk_stack_limits(mddev->gendisk, rdev->bdev,
 					  rdev->data_offset << 9);
@@ -1322,21 +1349,35 @@ static int raid1_add_disk(struct mddev *mddev, struct md_rdev *rdev)
 			rcu_assign_pointer(p->rdev, rdev);
 			break;
 		}
+		if (test_bit(WantReplacement, &p->rdev->flags) &&
+		    p[conf->raid_disks].rdev == NULL) {
+			/* Add this device as a replacement */
+			clear_bit(In_sync, &rdev->flags);
+			set_bit(Replacement, &rdev->flags);
+			rdev->raid_disk = mirror;
+			err = 0;
+			conf->fullsync = 1;
+			rcu_assign_pointer(p[conf->raid_disks].rdev, rdev);
+			break;
+		}
+	}
 	md_integrity_add_rdev(rdev, mddev);
 	print_conf(conf);
 	return err;
 }
 
-static int raid1_remove_disk(struct mddev *mddev, int number)
+static int raid1_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
 {
 	struct r1conf *conf = mddev->private;
 	int err = 0;
-	struct md_rdev *rdev;
+	int number = rdev->raid_disk;
 	struct mirror_info *p = conf->mirrors+ number;
 
+	if (rdev != p->rdev)
+		p = conf->mirrors + conf->raid_disks + number;
+
 	print_conf(conf);
-	rdev = p->rdev;
-	if (rdev) {
+	if (rdev == p->rdev) {
 		if (test_bit(In_sync, &rdev->flags) ||
 		    atomic_read(&rdev->nr_pending)) {
 			err = -EBUSY;
@@ -1358,7 +1399,21 @@ static int raid1_remove_disk(struct mddev *mddev, int number)
 			err = -EBUSY;
 			p->rdev = rdev;
 			goto abort;
-		}
+		} else if (conf->mirrors[conf->raid_disks + number].rdev) {
+			/* We just removed a device that is being replaced.
+			 * Move down the replacement.  We drain all IO before
+			 * doing this to avoid confusion.
+			 */
+			struct md_rdev *repl =
+				conf->mirrors[conf->raid_disks + number].rdev;
+			raise_barrier(conf);
+			clear_bit(Replacement, &repl->flags);
+			p->rdev = repl;
+			conf->mirrors[conf->raid_disks + number].rdev = NULL;
+			lower_barrier(conf);
+			clear_bit(WantReplacement, &rdev->flags);
+		} else
+			clear_bit(WantReplacement, &rdev->flags);
 		err = md_integrity_register(mddev);
 	}
 abort:
@@ -1411,6 +1466,10 @@ static void end_sync_write(struct bio *bio, int error)
 		} while (sectors_to_go > 0);
 		set_bit(WriteErrorSeen,
 			&conf->mirrors[mirror].rdev->flags);
+		if (!test_and_set_bit(WantReplacement,
+				      &conf->mirrors[mirror].rdev->flags))
+			set_bit(MD_RECOVERY_NEEDED, &
+				mddev->recovery);
 		set_bit(R1BIO_WriteError, &r1_bio->state);
 	} else if (is_badblock(conf->mirrors[mirror].rdev,
 			       r1_bio->sector,
@@ -1441,8 +1500,13 @@ static int r1_sync_page_io(struct md_rdev *rdev, sector_t sector,
 	if (sync_page_io(rdev, sector, sectors << 9, page, rw, false))
 		/* success */
 		return 1;
-	if (rw == WRITE)
+	if (rw == WRITE) {
 		set_bit(WriteErrorSeen, &rdev->flags);
+		if (!test_and_set_bit(WantReplacement,
+				      &rdev->flags))
+			set_bit(MD_RECOVERY_NEEDED, &
+				rdev->mddev->recovery);
+	}
 	/* need to record an error - either for the block or the device */
 	if (!rdev_set_badblocks(rdev, sector, sectors, 0))
 		md_error(rdev->mddev, rdev);
@@ -1493,7 +1557,7 @@ static int fix_sync_read_error(struct r1bio *r1_bio)
 				}
 			}
 			d++;
-			if (d == conf->raid_disks)
+			if (d == conf->raid_disks * 2)
 				d = 0;
 		} while (!success && d != r1_bio->read_disk);
 
@@ -1510,7 +1574,7 @@ static int fix_sync_read_error(struct r1bio *r1_bio)
 			       mdname(mddev),
 			       bdevname(bio->bi_bdev, b),
 			       (unsigned long long)r1_bio->sector);
-			for (d = 0; d < conf->raid_disks; d++) {
+			for (d = 0; d < conf->raid_disks * 2; d++) {
 				rdev = conf->mirrors[d].rdev;
 				if (!rdev || test_bit(Faulty, &rdev->flags))
 					continue;
@@ -1536,7 +1600,7 @@ static int fix_sync_read_error(struct r1bio *r1_bio)
 		/* write it back and re-read */
 		while (d != r1_bio->read_disk) {
 			if (d == 0)
-				d = conf->raid_disks;
+				d = conf->raid_disks * 2;
 			d--;
 			if (r1_bio->bios[d]->bi_end_io != end_sync_read)
 				continue;
@@ -1551,7 +1615,7 @@ static int fix_sync_read_error(struct r1bio *r1_bio)
 		d = start;
 		while (d != r1_bio->read_disk) {
 			if (d == 0)
-				d = conf->raid_disks;
+				d = conf->raid_disks * 2;
 			d--;
 			if (r1_bio->bios[d]->bi_end_io != end_sync_read)
 				continue;
@@ -1584,7 +1648,7 @@ static int process_checks(struct r1bio *r1_bio)
 	int primary;
 	int i;
 
-	for (primary = 0; primary < conf->raid_disks; primary++)
+	for (primary = 0; primary < conf->raid_disks * 2; primary++)
 		if (r1_bio->bios[primary]->bi_end_io == end_sync_read &&
 		    test_bit(BIO_UPTODATE, &r1_bio->bios[primary]->bi_flags)) {
 			r1_bio->bios[primary]->bi_end_io = NULL;
@@ -1592,7 +1656,7 @@ static int process_checks(struct r1bio *r1_bio)
 			break;
 		}
 	r1_bio->read_disk = primary;
-	for (i = 0; i < conf->raid_disks; i++) {
+	for (i = 0; i < conf->raid_disks * 2; i++) {
 		int j;
 		int vcnt = r1_bio->sectors >> (PAGE_SHIFT- 9);
 		struct bio *pbio = r1_bio->bios[primary];
@@ -1656,7 +1720,7 @@ static void sync_request_write(struct mddev *mddev, struct r1bio *r1_bio)
 {
 	struct r1conf *conf = mddev->private;
 	int i;
-	int disks = conf->raid_disks;
+	int disks = conf->raid_disks * 2;
 	struct bio *bio, *wbio;
 
 	bio = r1_bio->bios[r1_bio->read_disk];
@@ -1737,7 +1801,7 @@ static void fix_read_error(struct r1conf *conf, int read_disk,
 				success = 1;
 			else {
 				d++;
-				if (d == conf->raid_disks)
+				if (d == conf->raid_disks * 2)
 					d = 0;
 			}
 		} while (!success && d != read_disk);
@@ -1753,7 +1817,7 @@ static void fix_read_error(struct r1conf *conf, int read_disk,
 		start = d;
 		while (d != read_disk) {
 			if (d==0)
-				d = conf->raid_disks;
+				d = conf->raid_disks * 2;
 			d--;
 			rdev = conf->mirrors[d].rdev;
 			if (rdev &&
@@ -1765,7 +1829,7 @@ static void fix_read_error(struct r1conf *conf, int read_disk,
 		while (d != read_disk) {
 			char b[BDEVNAME_SIZE];
 			if (d==0)
-				d = conf->raid_disks;
+				d = conf->raid_disks * 2;
 			d--;
 			rdev = conf->mirrors[d].rdev;
 			if (rdev &&
@@ -1887,7 +1951,7 @@ static void handle_sync_write_finished(struct r1conf *conf, struct r1bio *r1_bio
 {
 	int m;
 	int s = r1_bio->sectors;
-	for (m = 0; m < conf->raid_disks ; m++) {
+	for (m = 0; m < conf->raid_disks * 2 ; m++) {
 		struct md_rdev *rdev = conf->mirrors[m].rdev;
 		struct bio *bio = r1_bio->bios[m];
 		if (bio->bi_end_io == NULL)
@@ -1909,7 +1973,7 @@ static void handle_sync_write_finished(struct r1conf *conf, struct r1bio *r1_bio
 static void handle_write_finished(struct r1conf *conf, struct r1bio *r1_bio)
 {
 	int m;
-	for (m = 0; m < conf->raid_disks ; m++)
+	for (m = 0; m < conf->raid_disks * 2 ; m++)
 		if (r1_bio->bios[m] == IO_MADE_GOOD) {
 			struct md_rdev *rdev = conf->mirrors[m].rdev;
 			rdev_clear_badblocks(rdev,
@@ -2184,7 +2248,7 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int *skipp
 	r1_bio->state = 0;
 	set_bit(R1BIO_IsSync, &r1_bio->state);
 
-	for (i=0; i < conf->raid_disks; i++) {
+	for (i = 0; i < conf->raid_disks * 2; i++) {
 		struct md_rdev *rdev;
 		bio = r1_bio->bios[i];
 
@@ -2203,7 +2267,8 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int *skipp
 		rdev = rcu_dereference(conf->mirrors[i].rdev);
 		if (rdev == NULL ||
 		    test_bit(Faulty, &rdev->flags)) {
-			still_degraded = 1;
+			if (i < conf->raid_disks)
+				still_degraded = 1;
 		} else if (!test_bit(In_sync, &rdev->flags)) {
 			bio->bi_rw = WRITE;
 			bio->bi_end_io = end_sync_write;
@@ -2254,7 +2319,7 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int *skipp
 		 * need to mark them bad on all write targets
 		 */
 		int ok = 1;
-		for (i = 0 ; i < conf->raid_disks ; i++)
+		for (i = 0 ; i < conf->raid_disks * 2 ; i++)
 			if (r1_bio->bios[i]->bi_end_io == end_sync_write) {
 				struct md_rdev *rdev =
 					rcu_dereference(conf->mirrors[i].rdev);
@@ -2323,7 +2388,7 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int *skipp
 				len = sync_blocks<<9;
 		}
 
-		for (i=0 ; i < conf->raid_disks; i++) {
+		for (i = 0 ; i < conf->raid_disks * 2; i++) {
 			bio = r1_bio->bios[i];
 			if (bio->bi_end_io) {
 				page = bio->bi_io_vec[bio->bi_vcnt].bv_page;
@@ -2356,7 +2421,7 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int *skipp
 	 */
 	if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
 		atomic_set(&r1_bio->remaining, read_targets);
-		for (i=0; i<conf->raid_disks; i++) {
+		for (i = 0; i < conf->raid_disks * 2; i++) {
 			bio = r1_bio->bios[i];
 			if (bio->bi_end_io == end_sync_read) {
 				md_sync_acct(bio->bi_bdev, nr_sectors);
@@ -2393,7 +2458,8 @@ static struct r1conf *setup_conf(struct mddev *mddev)
 	if (!conf)
 		goto abort;
 
-	conf->mirrors = kzalloc(sizeof(struct mirror_info)*mddev->raid_disks,
+	conf->mirrors = kzalloc(sizeof(struct mirror_info)
+				* mddev->raid_disks * 2,
 				 GFP_KERNEL);
 	if (!conf->mirrors)
 		goto abort;
@@ -2405,7 +2471,7 @@ static struct r1conf *setup_conf(struct mddev *mddev)
 	conf->poolinfo = kzalloc(sizeof(*conf->poolinfo), GFP_KERNEL);
 	if (!conf->poolinfo)
 		goto abort;
-	conf->poolinfo->raid_disks = mddev->raid_disks;
+	conf->poolinfo->raid_disks = mddev->raid_disks * 2;
 	conf->r1bio_pool = mempool_create(NR_RAID1_BIOS, r1bio_pool_alloc,
 					  r1bio_pool_free,
 					  conf->poolinfo);
@@ -2414,14 +2480,20 @@ static struct r1conf *setup_conf(struct mddev *mddev)
 
 	conf->poolinfo->mddev = mddev;
 
+	err = -EINVAL;
 	spin_lock_init(&conf->device_lock);
 	list_for_each_entry(rdev, &mddev->disks, same_set) {
 		int disk_idx = rdev->raid_disk;
 		if (disk_idx >= mddev->raid_disks
 		    || disk_idx < 0)
 			continue;
-		disk = conf->mirrors + disk_idx;
+		if (test_bit(Replacement, &rdev->flags))
+			disk = conf->mirrors + conf->raid_disks + disk_idx;
+		else
+			disk = conf->mirrors + disk_idx;
 
+		if (disk->rdev)
+			goto abort;
 		disk->rdev = rdev;
 
 		disk->head_position = 0;
@@ -2437,11 +2509,27 @@ static struct r1conf *setup_conf(struct mddev *mddev)
 	conf->pending_count = 0;
 	conf->recovery_disabled = mddev->recovery_disabled - 1;
 
+	err = -EIO;
 	conf->last_used = -1;
-	for (i = 0; i < conf->raid_disks; i++) {
+	for (i = 0; i < conf->raid_disks * 2; i++) {
 
 		disk = conf->mirrors + i;
 
+		if (i < conf->raid_disks &&
+		    disk[conf->raid_disks].rdev) {
+			/* This slot has a replacement. */
+			if (!disk->rdev) {
+				/* No original, just make the replacement
+				 * a recovering spare
+				 */
+				disk->rdev =
+					disk[conf->raid_disks].rdev;
+				disk[conf->raid_disks].rdev = NULL;
+			} else if (!test_bit(In_sync, &disk->rdev->flags))
+				/* Original is not in_sync - bad */
+				goto abort;
+		}
+
 		if (!disk->rdev ||
 		    !test_bit(In_sync, &disk->rdev->flags)) {
 			disk->head_position = 0;
@@ -2455,7 +2543,6 @@ static struct r1conf *setup_conf(struct mddev *mddev)
 			conf->last_used = i;
 	}
 
-	err = -EIO;
 	if (conf->last_used < 0) {
 		printk(KERN_ERR "md/raid1:%s: no operational mirrors\n",
 		       mdname(mddev));
@@ -2665,7 +2752,7 @@ static int raid1_reshape(struct mddev *mddev)
 	if (!newpoolinfo)
 		return -ENOMEM;
 	newpoolinfo->mddev = mddev;
-	newpoolinfo->raid_disks = raid_disks;
+	newpoolinfo->raid_disks = raid_disks * 2;
 
 	newpool = mempool_create(NR_RAID1_BIOS, r1bio_pool_alloc,
 				 r1bio_pool_free, newpoolinfo);
@@ -2673,7 +2760,8 @@ static int raid1_reshape(struct mddev *mddev)
 		kfree(newpoolinfo);
 		return -ENOMEM;
 	}
-	newmirrors = kzalloc(sizeof(struct mirror_info) * raid_disks, GFP_KERNEL);
+	newmirrors = kzalloc(sizeof(struct mirror_info) * raid_disks * 2,
+			     GFP_KERNEL);
 	if (!newmirrors) {
 		kfree(newpoolinfo);
 		mempool_destroy(newpool);

drivers/md/raid1.h

@@ -12,6 +12,9 @@ struct mirror_info {
  * pool was allocated for, so they know how much to allocate and free.
  * mddev->raid_disks cannot be used, as it can change while a pool is active
  * These two datums are stored in a kmalloced struct.
+ * The 'raid_disks' here is twice the raid_disks in r1conf.
+ * This allows space for each 'real' device can have a replacement in the
+ * second half of the array.
  */
 
 struct pool_info {
@@ -21,7 +24,9 @@ struct pool_info {
 
 struct r1conf {
 	struct mddev		*mddev;
-	struct mirror_info		*mirrors;
+	struct mirror_info	*mirrors;	/* twice 'raid_disks' to
+						 * allow for replacements.
+						 */
 	int			raid_disks;
 
 	/* When choose the best device for a read (read_balance())

drivers/md/raid10.c

@@ -73,7 +73,8 @@ static void * r10bio_pool_alloc(gfp_t gfp_flags, void *data)
 	struct r10conf *conf = data;
 	int size = offsetof(struct r10bio, devs[conf->copies]);
 
-	/* allocate a r10bio with room for raid_disks entries in the bios array */
+	/* allocate a r10bio with room for raid_disks entries in the
+	 * bios array */
 	return kzalloc(size, gfp_flags);
 }
 
@@ -123,12 +124,19 @@ static void * r10buf_pool_alloc(gfp_t gfp_flags, void *data)
 		if (!bio)
 			goto out_free_bio;
 		r10_bio->devs[j].bio = bio;
+		if (!conf->have_replacement)
+			continue;
+		bio = bio_kmalloc(gfp_flags, RESYNC_PAGES);
+		if (!bio)
+			goto out_free_bio;
+		r10_bio->devs[j].repl_bio = bio;
 	}
 	/*
 	 * Allocate RESYNC_PAGES data pages and attach them
 	 * where needed.
 	 */
 	for (j = 0 ; j < nalloc; j++) {
+		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,
@@ -143,6 +151,8 @@ static void * r10buf_pool_alloc(gfp_t gfp_flags, void *data)
 				goto out_free_pages;
 
 			bio->bi_io_vec[i].bv_page = page;
+			if (rbio)
+				rbio->bi_io_vec[i].bv_page = page;
 		}
 	}
 
@@ -156,8 +166,11 @@ static void * r10buf_pool_alloc(gfp_t gfp_flags, void *data)
 			safe_put_page(r10_bio->devs[j].bio->bi_io_vec[i].bv_page);
 	j = -1;
 out_free_bio:
-	while ( ++j < nalloc )
+	while (++j < nalloc) {
 		bio_put(r10_bio->devs[j].bio);
+		if (r10_bio->devs[j].repl_bio)
+			bio_put(r10_bio->devs[j].repl_bio);
+	}
 	r10bio_pool_free(r10_bio, conf);
 	return NULL;
 }
@@ -178,6 +191,9 @@ static void r10buf_pool_free(void *__r10_bio, void *data)
 			}
 			bio_put(bio);
 		}
+		bio = r10bio->devs[j].repl_bio;
+		if (bio)
+			bio_put(bio);
 	}
 	r10bio_pool_free(r10bio, conf);
 }
@@ -191,6 +207,10 @@ static void put_all_bios(struct r10conf *conf, struct r10bio *r10_bio)
 		if (!BIO_SPECIAL(*bio))
 			bio_put(*bio);
 		*bio = NULL;
+		bio = &r10_bio->devs[i].repl_bio;
+		if (r10_bio->read_slot < 0 && !BIO_SPECIAL(*bio))
+			bio_put(*bio);
+		*bio = NULL;
 	}
 }
 
@@ -275,19 +295,27 @@ static inline void update_head_pos(int slot, struct r10bio *r10_bio)
  * Find the disk number which triggered given bio
  */
 static int find_bio_disk(struct r10conf *conf, struct r10bio *r10_bio,
-			 struct bio *bio, int *slotp)
+			 struct bio *bio, int *slotp, int *replp)
 {
 	int slot;
+	int repl = 0;
 
-	for (slot = 0; slot < conf->copies; slot++)
+	for (slot = 0; slot < conf->copies; slot++) {
 		if (r10_bio->devs[slot].bio == bio)
 			break;
+		if (r10_bio->devs[slot].repl_bio == bio) {
+			repl = 1;
+			break;
+		}
+	}
 
 	BUG_ON(slot == conf->copies);
 	update_head_pos(slot, r10_bio);
 
 	if (slotp)
 		*slotp = slot;
+	if (replp)
+		*replp = repl;
 	return r10_bio->devs[slot].devnum;
 }
 
@@ -296,11 +324,13 @@ static void raid10_end_read_request(struct bio *bio, int error)
 	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
 	struct r10bio *r10_bio = bio->bi_private;
 	int slot, dev;
+	struct md_rdev *rdev;
 	struct r10conf *conf = r10_bio->mddev->private;
 
 
 	slot = r10_bio->read_slot;
 	dev = r10_bio->devs[slot].devnum;
+	rdev = r10_bio->devs[slot].rdev;
 	/*
 	 * this branch is our 'one mirror IO has finished' event handler:
 	 */
@@ -318,7 +348,7 @@ static void raid10_end_read_request(struct bio *bio, int error)
 		 */
 		set_bit(R10BIO_Uptodate, &r10_bio->state);
 		raid_end_bio_io(r10_bio);
-		rdev_dec_pending(conf->mirrors[dev].rdev, conf->mddev);
+		rdev_dec_pending(rdev, conf->mddev);
 	} else {
 		/*
 		 * oops, read error - keep the refcount on the rdev
@@ -327,7 +357,7 @@ static void raid10_end_read_request(struct bio *bio, int error)
 		printk_ratelimited(KERN_ERR
 				   "md/raid10:%s: %s: rescheduling sector %llu\n",
 				   mdname(conf->mddev),
-				   bdevname(conf->mirrors[dev].rdev->bdev, b),
+				   bdevname(rdev->bdev, b),
 				   (unsigned long long)r10_bio->sector);
 		set_bit(R10BIO_ReadError, &r10_bio->state);
 		reschedule_retry(r10_bio);
@@ -366,17 +396,35 @@ static void raid10_end_write_request(struct bio *bio, int error)
 	int dev;
 	int dec_rdev = 1;
 	struct r10conf *conf = r10_bio->mddev->private;
-	int slot;
+	int slot, repl;
+	struct md_rdev *rdev = NULL;
 
-	dev = find_bio_disk(conf, r10_bio, bio, &slot);
+	dev = find_bio_disk(conf, r10_bio, bio, &slot, &repl);
 
+	if (repl)
+		rdev = conf->mirrors[dev].replacement;
+	if (!rdev) {
+		smp_rmb();
+		repl = 0;
+		rdev = conf->mirrors[dev].rdev;
+	}
 	/*
 	 * this branch is our 'one mirror IO has finished' event handler:
 	 */
 	if (!uptodate) {
-		set_bit(WriteErrorSeen,	&conf->mirrors[dev].rdev->flags);
-		set_bit(R10BIO_WriteError, &r10_bio->state);
-		dec_rdev = 0;
+		if (repl)
+			/* Never record new bad blocks to replacement,
+			 * just fail it.
+			 */
+			md_error(rdev->mddev, rdev);
+		else {
+			set_bit(WriteErrorSeen,	&rdev->flags);
+			if (!test_and_set_bit(WantReplacement, &rdev->flags))
+				set_bit(MD_RECOVERY_NEEDED,
+					&rdev->mddev->recovery);
+			set_bit(R10BIO_WriteError, &r10_bio->state);
+			dec_rdev = 0;
+		}
 	} else {
 		/*
 		 * Set R10BIO_Uptodate in our master bio, so that
@@ -393,12 +441,15 @@ static void raid10_end_write_request(struct bio *bio, int error)
 		set_bit(R10BIO_Uptodate, &r10_bio->state);
 
 		/* Maybe we can clear some bad blocks. */
-		if (is_badblock(conf->mirrors[dev].rdev,
+		if (is_badblock(rdev,
 				r10_bio->devs[slot].addr,
 				r10_bio->sectors,
 				&first_bad, &bad_sectors)) {
 			bio_put(bio);
-			r10_bio->devs[slot].bio = IO_MADE_GOOD;
+			if (repl)
+				r10_bio->devs[slot].repl_bio = IO_MADE_GOOD;
+			else
+				r10_bio->devs[slot].bio = IO_MADE_GOOD;
 			dec_rdev = 0;
 			set_bit(R10BIO_MadeGood, &r10_bio->state);
 		}
@@ -414,7 +465,6 @@ static void raid10_end_write_request(struct bio *bio, int error)
 		rdev_dec_pending(conf->mirrors[dev].rdev, conf->mddev);
 }
 
-
 /*
  * RAID10 layout manager
  * As well as the chunksize and raid_disks count, there are two
@@ -562,14 +612,16 @@ static int raid10_mergeable_bvec(struct request_queue *q,
  * FIXME: possibly should rethink readbalancing and do it differently
  * depending on near_copies / far_copies geometry.
  */
-static int read_balance(struct r10conf *conf, struct r10bio *r10_bio, int *max_sectors)
+static struct md_rdev *read_balance(struct r10conf *conf,
+				    struct r10bio *r10_bio,
+				    int *max_sectors)
 {
 	const sector_t this_sector = r10_bio->sector;
 	int disk, slot;
 	int sectors = r10_bio->sectors;
 	int best_good_sectors;
 	sector_t new_distance, best_dist;
-	struct md_rdev *rdev;
+	struct md_rdev *rdev, *best_rdev;
 	int do_balance;
 	int best_slot;
 
@@ -578,6 +630,7 @@ static int read_balance(struct r10conf *conf, struct r10bio *r10_bio, int *max_s
 retry:
 	sectors = r10_bio->sectors;
 	best_slot = -1;
+	best_rdev = NULL;
 	best_dist = MaxSector;
 	best_good_sectors = 0;
 	do_balance = 1;
@@ -599,10 +652,16 @@ static int read_balance(struct r10conf *conf, struct r10bio *r10_bio, int *max_s
 		if (r10_bio->devs[slot].bio == IO_BLOCKED)
 			continue;
 		disk = r10_bio->devs[slot].devnum;
-		rdev = rcu_dereference(conf->mirrors[disk].rdev);
+		rdev = rcu_dereference(conf->mirrors[disk].replacement);
+		if (rdev == NULL || test_bit(Faulty, &rdev->flags) ||
+		    r10_bio->devs[slot].addr + sectors > rdev->recovery_offset)
+			rdev = rcu_dereference(conf->mirrors[disk].rdev);
 		if (rdev == NULL)
 			continue;
-		if (!test_bit(In_sync, &rdev->flags))
+		if (test_bit(Faulty, &rdev->flags))
+			continue;
+		if (!test_bit(In_sync, &rdev->flags) &&
+		    r10_bio->devs[slot].addr + sectors > rdev->recovery_offset)
 			continue;
 
 		dev_sector = r10_bio->devs[slot].addr;
@@ -627,6 +686,7 @@ static int read_balance(struct r10conf *conf, struct r10bio *r10_bio, int *max_s
 				if (good_sectors > best_good_sectors) {
 					best_good_sectors = good_sectors;
 					best_slot = slot;
+					best_rdev = rdev;
 				}
 				if (!do_balance)
 					/* Must read from here */
@@ -655,16 +715,15 @@ static int read_balance(struct r10conf *conf, struct r10bio *r10_bio, int *max_s
 		if (new_distance < best_dist) {
 			best_dist = new_distance;
 			best_slot = slot;
+			best_rdev = rdev;
 		}
 	}
-	if (slot == conf->copies)
+	if (slot >= conf->copies) {
 		slot = best_slot;
+		rdev = best_rdev;
+	}
 
 	if (slot >= 0) {
-		disk = r10_bio->devs[slot].devnum;
-		rdev = rcu_dereference(conf->mirrors[disk].rdev);
-		if (!rdev)
-			goto retry;
 		atomic_inc(&rdev->nr_pending);
 		if (test_bit(Faulty, &rdev->flags)) {
 			/* Cannot risk returning a device that failed
@@ -675,11 +734,11 @@ static int read_balance(struct r10conf *conf, struct r10bio *r10_bio, int *max_s
 		}
 		r10_bio->read_slot = slot;
 	} else
-		disk = -1;
+		rdev = NULL;
 	rcu_read_unlock();
 	*max_sectors = best_good_sectors;
 
-	return disk;
+	return rdev;
 }
 
 static int raid10_congested(void *data, int bits)
@@ -846,7 +905,6 @@ static void unfreeze_array(struct r10conf *conf)
 static void make_request(struct mddev *mddev, struct bio * bio)
 {
 	struct r10conf *conf = mddev->private;
-	struct mirror_info *mirror;
 	struct r10bio *r10_bio;
 	struct bio *read_bio;
 	int i;
@@ -945,27 +1003,27 @@ static void make_request(struct mddev *mddev, struct bio * bio)
 		/*
 		 * read balancing logic:
 		 */
-		int disk;
+		struct md_rdev *rdev;
 		int slot;
 
 read_again:
-		disk = read_balance(conf, r10_bio, &max_sectors);
-		slot = r10_bio->read_slot;
-		if (disk < 0) {
+		rdev = read_balance(conf, r10_bio, &max_sectors);
+		if (!rdev) {
 			raid_end_bio_io(r10_bio);
 			return;
 		}
-		mirror = conf->mirrors + disk;
+		slot = r10_bio->read_slot;
 
 		read_bio = bio_clone_mddev(bio, GFP_NOIO, mddev);
 		md_trim_bio(read_bio, r10_bio->sector - bio->bi_sector,
 			    max_sectors);
 
 		r10_bio->devs[slot].bio = read_bio;
+		r10_bio->devs[slot].rdev = rdev;
 
 		read_bio->bi_sector = r10_bio->devs[slot].addr +
-			mirror->rdev->data_offset;
-		read_bio->bi_bdev = mirror->rdev->bdev;
+			rdev->data_offset;
+		read_bio->bi_bdev = rdev->bdev;
 		read_bio->bi_end_io = raid10_end_read_request;
 		read_bio->bi_rw = READ | do_sync;
 		read_bio->bi_private = r10_bio;
@@ -1025,6 +1083,7 @@ static void make_request(struct mddev *mddev, struct bio * bio)
 	 */
 	plugged = mddev_check_plugged(mddev);
 
+	r10_bio->read_slot = -1; /* make sure repl_bio gets freed */
 	raid10_find_phys(conf, r10_bio);
 retry_write:
 	blocked_rdev = NULL;
@@ -1034,12 +1093,25 @@ static void make_request(struct mddev *mddev, struct bio * bio)
 	for (i = 0;  i < conf->copies; i++) {
 		int d = r10_bio->devs[i].devnum;
 		struct md_rdev *rdev = rcu_dereference(conf->mirrors[d].rdev);
+		struct md_rdev *rrdev = rcu_dereference(
+			conf->mirrors[d].replacement);
+		if (rdev == rrdev)
+			rrdev = NULL;
 		if (rdev && unlikely(test_bit(Blocked, &rdev->flags))) {
 			atomic_inc(&rdev->nr_pending);
 			blocked_rdev = rdev;
 			break;
 		}
+		if (rrdev && unlikely(test_bit(Blocked, &rrdev->flags))) {
+			atomic_inc(&rrdev->nr_pending);
+			blocked_rdev = rrdev;
+			break;
+		}
+		if (rrdev && test_bit(Faulty, &rrdev->flags))
+			rrdev = NULL;
+
 		r10_bio->devs[i].bio = NULL;
+		r10_bio->devs[i].repl_bio = NULL;
 		if (!rdev || test_bit(Faulty, &rdev->flags)) {
 			set_bit(R10BIO_Degraded, &r10_bio->state);
 			continue;
@@ -1088,6 +1160,10 @@ static void make_request(struct mddev *mddev, struct bio * bio)
 		}
 		r10_bio->devs[i].bio = bio;
 		atomic_inc(&rdev->nr_pending);
+		if (rrdev) {
+			r10_bio->devs[i].repl_bio = bio;
+			atomic_inc(&rrdev->nr_pending);
+		}
 	}
 	rcu_read_unlock();
 
@@ -1096,11 +1172,23 @@ static void make_request(struct mddev *mddev, struct bio * bio)
 		int j;
 		int d;
 
-		for (j = 0; j < i; j++)
+		for (j = 0; j < i; j++) {
 			if (r10_bio->devs[j].bio) {
 				d = r10_bio->devs[j].devnum;
 				rdev_dec_pending(conf->mirrors[d].rdev, mddev);
 			}
+			if (r10_bio->devs[j].repl_bio) {
+				struct md_rdev *rdev;
+				d = r10_bio->devs[j].devnum;
+				rdev = conf->mirrors[d].replacement;
+				if (!rdev) {
+					/* Race with remove_disk */
+					smp_mb();
+					rdev = conf->mirrors[d].rdev;
+				}
+				rdev_dec_pending(rdev, mddev);
+			}
+		}
 		allow_barrier(conf);
 		md_wait_for_blocked_rdev(blocked_rdev, mddev);
 		wait_barrier(conf);
@@ -1147,6 +1235,31 @@ static void make_request(struct mddev *mddev, struct bio * bio)
 		bio_list_add(&conf->pending_bio_list, mbio);
 		conf->pending_count++;
 		spin_unlock_irqrestore(&conf->device_lock, flags);
+
+		if (!r10_bio->devs[i].repl_bio)
+			continue;
+
+		mbio = bio_clone_mddev(bio, GFP_NOIO, mddev);
+		md_trim_bio(mbio, r10_bio->sector - bio->bi_sector,
+			    max_sectors);
+		r10_bio->devs[i].repl_bio = mbio;
+
+		/* We are actively writing to the original device
+		 * 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_bdev = conf->mirrors[d].replacement->bdev;
+		mbio->bi_end_io	= raid10_end_write_request;
+		mbio->bi_rw = WRITE | do_sync | do_fua;
+		mbio->bi_private = r10_bio;
+
+		atomic_inc(&r10_bio->remaining);
+		spin_lock_irqsave(&conf->device_lock, flags);
+		bio_list_add(&conf->pending_bio_list, mbio);
+		conf->pending_count++;
+		spin_unlock_irqrestore(&conf->device_lock, flags);
 	}
 
 	/* Don't remove the bias on 'remaining' (one_write_done) until
@@ -1309,9 +1422,27 @@ static int raid10_spare_active(struct mddev *mddev)
 	 */
 	for (i = 0; i < conf->raid_disks; i++) {
 		tmp = conf->mirrors + i;
-		if (tmp->rdev
-		    && !test_bit(Faulty, &tmp->rdev->flags)
-		    && !test_and_set_bit(In_sync, &tmp->rdev->flags)) {
+		if (tmp->replacement
+		    && tmp->replacement->recovery_offset == MaxSector
+		    && !test_bit(Faulty, &tmp->replacement->flags)
+		    && !test_and_set_bit(In_sync, &tmp->replacement->flags)) {
+			/* Replacement has just become active */
+			if (!tmp->rdev
+			    || !test_and_clear_bit(In_sync, &tmp->rdev->flags))
+				count++;
+			if (tmp->rdev) {
+				/* Replaced device not technically faulty,
+				 * but we need to be sure it gets removed
+				 * and never re-added.
+				 */
+				set_bit(Faulty, &tmp->rdev->flags);
+				sysfs_notify_dirent_safe(
+					tmp->rdev->sysfs_state);
+			}
+			sysfs_notify_dirent_safe(tmp->replacement->sysfs_state);
+		} else if (tmp->rdev
+			   && !test_bit(Faulty, &tmp->rdev->flags)
+			   && !test_and_set_bit(In_sync, &tmp->rdev->flags)) {
 			count++;
 			sysfs_notify_dirent(tmp->rdev->sysfs_state);
 		}
@@ -1353,8 +1484,25 @@ static int raid10_add_disk(struct mddev *mddev, struct md_rdev *rdev)
 		struct mirror_info *p = &conf->mirrors[mirror];
 		if (p->recovery_disabled == mddev->recovery_disabled)
 			continue;
-		if (p->rdev)
-			continue;
+		if (p->rdev) {
+			if (!test_bit(WantReplacement, &p->rdev->flags) ||
+			    p->replacement != NULL)
+				continue;
+			clear_bit(In_sync, &rdev->flags);
+			set_bit(Replacement, &rdev->flags);
+			rdev->raid_disk = mirror;
+			err = 0;
+			disk_stack_limits(mddev->gendisk, rdev->bdev,
+					  rdev->data_offset << 9);
+			if (rdev->bdev->bd_disk->queue->merge_bvec_fn) {
+				blk_queue_max_segments(mddev->queue, 1);
+				blk_queue_segment_boundary(mddev->queue,
+							   PAGE_CACHE_SIZE - 1);
+			}
+			conf->fullsync = 1;
+			rcu_assign_pointer(p->replacement, rdev);
+			break;
+		}
 
 		disk_stack_limits(mddev->gendisk, rdev->bdev,
 				  rdev->data_offset << 9);
@@ -1385,40 +1533,61 @@ static int raid10_add_disk(struct mddev *mddev, struct md_rdev *rdev)
 	return err;
 }
 
-static int raid10_remove_disk(struct mddev *mddev, int number)
+static int raid10_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
 {
 	struct r10conf *conf = mddev->private;
 	int err = 0;
-	struct md_rdev *rdev;
-	struct mirror_info *p = conf->mirrors+ number;
+	int number = rdev->raid_disk;
+	struct md_rdev **rdevp;
+	struct mirror_info *p = conf->mirrors + number;
 
 	print_conf(conf);
-	rdev = p->rdev;
-	if (rdev) {
-		if (test_bit(In_sync, &rdev->flags) ||
-		    atomic_read(&rdev->nr_pending)) {
-			err = -EBUSY;
-			goto abort;
-		}
-		/* Only remove faulty devices in recovery
-		 * is not possible.
-		 */
-		if (!test_bit(Faulty, &rdev->flags) &&
-		    mddev->recovery_disabled != p->recovery_disabled &&
-		    enough(conf, -1)) {
-			err = -EBUSY;
-			goto abort;
-		}
-		p->rdev = NULL;
-		synchronize_rcu();
-		if (atomic_read(&rdev->nr_pending)) {
-			/* lost the race, try later */
-			err = -EBUSY;
-			p->rdev = rdev;
-			goto abort;
-		}
-		err = md_integrity_register(mddev);
+	if (rdev == p->rdev)
+		rdevp = &p->rdev;
+	else if (rdev == p->replacement)
+		rdevp = &p->replacement;
+	else
+		return 0;
+
+	if (test_bit(In_sync, &rdev->flags) ||
+	    atomic_read(&rdev->nr_pending)) {
+		err = -EBUSY;
+		goto abort;
 	}
+	/* Only remove faulty devices if recovery
+	 * is not possible.
+	 */
+	if (!test_bit(Faulty, &rdev->flags) &&
+	    mddev->recovery_disabled != p->recovery_disabled &&
+	    (!p->replacement || p->replacement == rdev) &&
+	    enough(conf, -1)) {
+		err = -EBUSY;
+		goto abort;
+	}
+	*rdevp = NULL;
+	synchronize_rcu();
+	if (atomic_read(&rdev->nr_pending)) {
+		/* lost the race, try later */
+		err = -EBUSY;
+		*rdevp = rdev;
+		goto abort;
+	} else if (p->replacement) {
+		/* We must have just cleared 'rdev' */
+		p->rdev = p->replacement;
+		clear_bit(Replacement, &p->replacement->flags);
+		smp_mb(); /* Make sure other CPUs may see both as identical
+			   * but will never see neither -- if they are careful.
+			   */
+		p->replacement = NULL;
+		clear_bit(WantReplacement, &rdev->flags);
+	} else
+		/* We might have just remove the Replacement as faulty
+		 * Clear the flag just in case
+		 */
+		clear_bit(WantReplacement, &rdev->flags);
+
+	err = md_integrity_register(mddev);
+
 abort:
 
 	print_conf(conf);
@@ -1432,7 +1601,7 @@ 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);
+	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);
@@ -1493,19 +1662,34 @@ static void end_sync_write(struct bio *bio, int error)
 	sector_t first_bad;
 	int bad_sectors;
 	int slot;
-
-	d = find_bio_disk(conf, r10_bio, bio, &slot);
+	int repl;
+	struct md_rdev *rdev = NULL;
+
+	d = find_bio_disk(conf, r10_bio, bio, &slot, &repl);
+	if (repl)
+		rdev = conf->mirrors[d].replacement;
+	if (!rdev) {
+		smp_mb();
+		rdev = conf->mirrors[d].rdev;
+	}
 
 	if (!uptodate) {
-		set_bit(WriteErrorSeen, &conf->mirrors[d].rdev->flags);
-		set_bit(R10BIO_WriteError, &r10_bio->state);
-	} else if (is_badblock(conf->mirrors[d].rdev,
+		if (repl)
+			md_error(mddev, rdev);
+		else {
+			set_bit(WriteErrorSeen, &rdev->flags);
+			if (!test_and_set_bit(WantReplacement, &rdev->flags))
+				set_bit(MD_RECOVERY_NEEDED,
+					&rdev->mddev->recovery);
+			set_bit(R10BIO_WriteError, &r10_bio->state);
+		}
+	} else if (is_badblock(rdev,
 			     r10_bio->devs[slot].addr,
 			     r10_bio->sectors,
 			     &first_bad, &bad_sectors))
 		set_bit(R10BIO_MadeGood, &r10_bio->state);
 
-	rdev_dec_pending(conf->mirrors[d].rdev, mddev);
+	rdev_dec_pending(rdev, mddev);
 
 	end_sync_request(r10_bio);
 }
@@ -1609,6 +1793,29 @@ static void sync_request_write(struct mddev *mddev, struct r10bio *r10_bio)
 		generic_make_request(tbio);
 	}
 
+	/* Now write out to any replacement devices
+	 * that are active
+	 */
+	for (i = 0; i < conf->copies; i++) {
+		int j, d;
+		int vcnt = r10_bio->sectors >> (PAGE_SHIFT-9);
+
+		tbio = r10_bio->devs[i].repl_bio;
+		if (!tbio || !tbio->bi_end_io)
+			continue;
+		if (r10_bio->devs[i].bio->bi_end_io != end_sync_write
+		    && r10_bio->devs[i].bio != fbio)
+			for (j = 0; j < vcnt; j++)
+				memcpy(page_address(tbio->bi_io_vec[j].bv_page),
+				       page_address(fbio->bi_io_vec[j].bv_page),
+				       PAGE_SIZE);
+		d = r10_bio->devs[i].devnum;
+		atomic_inc(&r10_bio->remaining);
+		md_sync_acct(conf->mirrors[d].replacement->bdev,
+			     tbio->bi_size >> 9);
+		generic_make_request(tbio);
+	}
+
 done:
 	if (atomic_dec_and_test(&r10_bio->remaining)) {
 		md_done_sync(mddev, r10_bio->sectors, 1);
@@ -1668,8 +1875,13 @@ static void fix_recovery_read_error(struct r10bio *r10_bio)
 					  s << 9,
 					  bio->bi_io_vec[idx].bv_page,
 					  WRITE, false);
-			if (!ok)
+			if (!ok) {
 				set_bit(WriteErrorSeen, &rdev->flags);
+				if (!test_and_set_bit(WantReplacement,
+						      &rdev->flags))
+					set_bit(MD_RECOVERY_NEEDED,
+						&rdev->mddev->recovery);
+			}
 		}
 		if (!ok) {
 			/* We don't worry if we cannot set a bad block -
@@ -1709,7 +1921,7 @@ static void recovery_request_write(struct mddev *mddev, struct r10bio *r10_bio)
 {
 	struct r10conf *conf = mddev->private;
 	int d;
-	struct bio *wbio;
+	struct bio *wbio, *wbio2;
 
 	if (!test_bit(R10BIO_Uptodate, &r10_bio->state)) {
 		fix_recovery_read_error(r10_bio);
@@ -1721,12 +1933,20 @@ static void recovery_request_write(struct mddev *mddev, struct r10bio *r10_bio)
 	 * share the pages with the first bio
 	 * and submit the write request
 	 */
-	wbio = r10_bio->devs[1].bio;
 	d = r10_bio->devs[1].devnum;
-
-	atomic_inc(&conf->mirrors[d].rdev->nr_pending);
-	md_sync_acct(conf->mirrors[d].rdev->bdev, wbio->bi_size >> 9);
-	generic_make_request(wbio);
+	wbio = r10_bio->devs[1].bio;
+	wbio2 = r10_bio->devs[1].repl_bio;
+	if (wbio->bi_end_io) {
+		atomic_inc(&conf->mirrors[d].rdev->nr_pending);
+		md_sync_acct(conf->mirrors[d].rdev->bdev, wbio->bi_size >> 9);
+		generic_make_request(wbio);
+	}
+	if (wbio2 && wbio2->bi_end_io) {
+		atomic_inc(&conf->mirrors[d].replacement->nr_pending);
+		md_sync_acct(conf->mirrors[d].replacement->bdev,
+			     wbio2->bi_size >> 9);
+		generic_make_request(wbio2);
+	}
 }
 
 
@@ -1779,8 +1999,12 @@ static int r10_sync_page_io(struct md_rdev *rdev, sector_t sector,
 	if (sync_page_io(rdev, sector, sectors << 9, page, rw, false))
 		/* success */
 		return 1;
-	if (rw == WRITE)
+	if (rw == WRITE) {
 		set_bit(WriteErrorSeen, &rdev->flags);
+		if (!test_and_set_bit(WantReplacement, &rdev->flags))
+			set_bit(MD_RECOVERY_NEEDED,
+				&rdev->mddev->recovery);
+	}
 	/* need to record an error - either for the block or the device */
 	if (!rdev_set_badblocks(rdev, sector, sectors, 0))
 		md_error(rdev->mddev, rdev);
@@ -2060,10 +2284,9 @@ static int narrow_write_error(struct r10bio *r10_bio, int i)
 static void handle_read_error(struct mddev *mddev, struct r10bio *r10_bio)
 {
 	int slot = r10_bio->read_slot;
-	int mirror = r10_bio->devs[slot].devnum;
 	struct bio *bio;
 	struct r10conf *conf = mddev->private;
-	struct md_rdev *rdev;
+	struct md_rdev *rdev = r10_bio->devs[slot].rdev;
 	char b[BDEVNAME_SIZE];
 	unsigned long do_sync;
 	int max_sectors;
@@ -2081,15 +2304,15 @@ static void handle_read_error(struct mddev *mddev, struct r10bio *r10_bio)
 		fix_read_error(conf, mddev, r10_bio);
 		unfreeze_array(conf);
 	}
-	rdev_dec_pending(conf->mirrors[mirror].rdev, mddev);
+	rdev_dec_pending(rdev, mddev);
 
 	bio = r10_bio->devs[slot].bio;
 	bdevname(bio->bi_bdev, b);
 	r10_bio->devs[slot].bio =
 		mddev->ro ? IO_BLOCKED : NULL;
 read_more:
-	mirror = read_balance(conf, r10_bio, &max_sectors);
-	if (mirror == -1) {
+	rdev = read_balance(conf, r10_bio, &max_sectors);
+	if (rdev == NULL) {
 		printk(KERN_ALERT "md/raid10:%s: %s: unrecoverable I/O"
 		       " read error for block %llu\n",
 		       mdname(mddev), b,
@@ -2103,7 +2326,6 @@ static void handle_read_error(struct mddev *mddev, struct r10bio *r10_bio)
 	if (bio)
 		bio_put(bio);
 	slot = r10_bio->read_slot;
-	rdev = conf->mirrors[mirror].rdev;
 	printk_ratelimited(
 		KERN_ERR
 		"md/raid10:%s: %s: redirecting"
@@ -2117,6 +2339,7 @@ static void handle_read_error(struct mddev *mddev, struct r10bio *r10_bio)
 		    r10_bio->sector - bio->bi_sector,
 		    max_sectors);
 	r10_bio->devs[slot].bio = bio;
+	r10_bio->devs[slot].rdev = rdev;
 	bio->bi_sector = r10_bio->devs[slot].addr
 		+ rdev->data_offset;
 	bio->bi_bdev = rdev->bdev;
@@ -2187,6 +2410,22 @@ static void handle_write_completed(struct r10conf *conf, struct r10bio *r10_bio)
 					    r10_bio->sectors, 0))
 					md_error(conf->mddev, rdev);
 			}
+			rdev = conf->mirrors[dev].replacement;
+			if (r10_bio->devs[m].repl_bio == NULL)
+				continue;
+			if (test_bit(BIO_UPTODATE,
+				     &r10_bio->devs[m].repl_bio->bi_flags)) {
+				rdev_clear_badblocks(
+					rdev,
+					r10_bio->devs[m].addr,
+					r10_bio->sectors);
+			} else {
+				if (!rdev_set_badblocks(
+					    rdev,
+					    r10_bio->devs[m].addr,
+					    r10_bio->sectors, 0))
+					md_error(conf->mddev, rdev);
+			}
 		}
 		put_buf(r10_bio);
 	} else {
@@ -2209,6 +2448,15 @@ static void handle_write_completed(struct r10conf *conf, struct r10bio *r10_bio)
 				}
 				rdev_dec_pending(rdev, conf->mddev);
 			}
+			bio = r10_bio->devs[m].repl_bio;
+			rdev = conf->mirrors[dev].replacement;
+			if (rdev && bio == IO_MADE_GOOD) {
+				rdev_clear_badblocks(
+					rdev,
+					r10_bio->devs[m].addr,
+					r10_bio->sectors);
+				rdev_dec_pending(rdev, conf->mddev);
+			}
 		}
 		if (test_bit(R10BIO_WriteError,
 			     &r10_bio->state))
@@ -2272,9 +2520,14 @@ static void raid10d(struct mddev *mddev)
 static int init_resync(struct r10conf *conf)
 {
 	int buffs;
+	int i;
 
 	buffs = RESYNC_WINDOW / RESYNC_BLOCK_SIZE;
 	BUG_ON(conf->r10buf_pool);
+	conf->have_replacement = 0;
+	for (i = 0; i < conf->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);
 	if (!conf->r10buf_pool)
 		return -ENOMEM;
@@ -2355,9 +2608,22 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr,
 				bitmap_end_sync(mddev->bitmap, sect,
 						&sync_blocks, 1);
 			}
-		} else /* completed sync */
+		} else {
+			/* completed sync */
+			if ((!mddev->bitmap || conf->fullsync)
+			    && conf->have_replacement
+			    && test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
+				/* Completed a full sync so the replacements
+				 * are now fully recovered.
+				 */
+				for (i = 0; i < conf->raid_disks; i++)
+					if (conf->mirrors[i].replacement)
+						conf->mirrors[i].replacement
+							->recovery_offset
+							= MaxSector;
+			}
 			conf->fullsync = 0;
-
+		}
 		bitmap_close_sync(mddev->bitmap);
 		close_sync(conf);
 		*skipped = 1;
@@ -2414,23 +2680,30 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr,
 			sector_t sect;
 			int must_sync;
 			int any_working;
-
-			if (conf->mirrors[i].rdev == NULL ||
-			    test_bit(In_sync, &conf->mirrors[i].rdev->flags)) 
+			struct mirror_info *mirror = &conf->mirrors[i];
+
+			if ((mirror->rdev == NULL ||
+			     test_bit(In_sync, &mirror->rdev->flags))
+			    &&
+			    (mirror->replacement == NULL ||
+			     test_bit(Faulty,
+				      &mirror->replacement->flags)))
 				continue;
 
 			still_degraded = 0;
 			/* want to reconstruct this device */
 			rb2 = r10_bio;
 			sect = raid10_find_virt(conf, sector_nr, i);
-			/* Unless we are doing a full sync, we only need
-			 * to recover the block if it is set in the bitmap
+			/* Unless we are doing a full sync, or a replacement
+			 * we only need to recover the block if it is set in
+			 * the bitmap
 			 */
 			must_sync = bitmap_start_sync(mddev->bitmap, sect,
 						      &sync_blocks, 1);
 			if (sync_blocks < max_sync)
 				max_sync = sync_blocks;
 			if (!must_sync &&
+			    mirror->replacement == NULL &&
 			    !conf->fullsync) {
 				/* yep, skip the sync_blocks here, but don't assume
 				 * that there will never be anything to do here
@@ -2500,33 +2773,60 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr,
 				bio->bi_end_io = end_sync_read;
 				bio->bi_rw = READ;
 				from_addr = r10_bio->devs[j].addr;
-				bio->bi_sector = from_addr +
-					conf->mirrors[d].rdev->data_offset;
-				bio->bi_bdev = conf->mirrors[d].rdev->bdev;
-				atomic_inc(&conf->mirrors[d].rdev->nr_pending);
-				atomic_inc(&r10_bio->remaining);
-				/* and we write to 'i' */
+				bio->bi_sector = from_addr + rdev->data_offset;
+				bio->bi_bdev = rdev->bdev;
+				atomic_inc(&rdev->nr_pending);
+				/* and we write to 'i' (if not in_sync) */
 
 				for (k=0; k<conf->copies; k++)
 					if (r10_bio->devs[k].devnum == i)
 						break;
 				BUG_ON(k == conf->copies);
-				bio = r10_bio->devs[1].bio;
-				bio->bi_next = biolist;
-				biolist = bio;
-				bio->bi_private = r10_bio;
-				bio->bi_end_io = end_sync_write;
-				bio->bi_rw = WRITE;
 				to_addr = r10_bio->devs[k].addr;
-				bio->bi_sector = to_addr +
-					conf->mirrors[i].rdev->data_offset;
-				bio->bi_bdev = conf->mirrors[i].rdev->bdev;
-
 				r10_bio->devs[0].devnum = d;
 				r10_bio->devs[0].addr = from_addr;
 				r10_bio->devs[1].devnum = i;
 				r10_bio->devs[1].addr = to_addr;
 
+				rdev = mirror->rdev;
+				if (!test_bit(In_sync, &rdev->flags)) {
+					bio = r10_bio->devs[1].bio;
+					bio->bi_next = biolist;
+					biolist = bio;
+					bio->bi_private = r10_bio;
+					bio->bi_end_io = end_sync_write;
+					bio->bi_rw = WRITE;
+					bio->bi_sector = to_addr
+						+ rdev->data_offset;
+					bio->bi_bdev = rdev->bdev;
+					atomic_inc(&r10_bio->remaining);
+				} else
+					r10_bio->devs[1].bio->bi_end_io = NULL;
+
+				/* and maybe write to replacement */
+				bio = r10_bio->devs[1].repl_bio;
+				if (bio)
+					bio->bi_end_io = NULL;
+				rdev = mirror->replacement;
+				/* Note: if rdev != NULL, then bio
+				 * cannot be NULL as r10buf_pool_alloc will
+				 * have allocated it.
+				 * So the second test here is pointless.
+				 * But it keeps semantic-checkers happy, and
+				 * this comment keeps human reviewers
+				 * happy.
+				 */
+				if (rdev == NULL || bio == NULL ||
+				    test_bit(Faulty, &rdev->flags))
+					break;
+				bio->bi_next = biolist;
+				biolist = bio;
+				bio->bi_private = r10_bio;
+				bio->bi_end_io = end_sync_write;
+				bio->bi_rw = WRITE;
+				bio->bi_sector = to_addr + rdev->data_offset;
+				bio->bi_bdev = rdev->bdev;
+				atomic_inc(&r10_bio->remaining);
 				break;
 			}
 			if (j == conf->copies) {
@@ -2544,8 +2844,16 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr,
 					for (k = 0; k < conf->copies; k++)
 						if (r10_bio->devs[k].devnum == i)
 							break;
-					if (!rdev_set_badblocks(
-						    conf->mirrors[i].rdev,
+					if (!test_bit(In_sync,
+						      &mirror->rdev->flags)
+					    && !rdev_set_badblocks(
+						    mirror->rdev,
+						    r10_bio->devs[k].addr,
+						    max_sync, 0))
+						any_working = 0;
+					if (mirror->replacement &&
+					    !rdev_set_badblocks(
+						    mirror->replacement,
 						    r10_bio->devs[k].addr,
 						    max_sync, 0))
 						any_working = 0;
@@ -2556,7 +2864,7 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr,
 						printk(KERN_INFO "md/raid10:%s: insufficient "
 						       "working devices for recovery.\n",
 						       mdname(mddev));
-					conf->mirrors[i].recovery_disabled
+					mirror->recovery_disabled
 						= mddev->recovery_disabled;
 				}
 				break;
@@ -2605,6 +2913,9 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr,
 			sector_t first_bad, sector;
 			int bad_sectors;
 
+			if (r10_bio->devs[i].repl_bio)
+				r10_bio->devs[i].repl_bio->bi_end_io = NULL;
+
 			bio = r10_bio->devs[i].bio;
 			bio->bi_end_io = NULL;
 			clear_bit(BIO_UPTODATE, &bio->bi_flags);
@@ -2635,6 +2946,27 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr,
 				conf->mirrors[d].rdev->data_offset;
 			bio->bi_bdev = conf->mirrors[d].rdev->bdev;
 			count++;
+
+			if (conf->mirrors[d].replacement == NULL ||
+			    test_bit(Faulty,
+				     &conf->mirrors[d].replacement->flags))
+				continue;
+
+			/* Need to set up for writing to the replacement */
+			bio = r10_bio->devs[i].repl_bio;
+			clear_bit(BIO_UPTODATE, &bio->bi_flags);
+
+			sector = r10_bio->devs[i].addr;
+			atomic_inc(&conf->mirrors[d].rdev->nr_pending);
+			bio->bi_next = biolist;
+			biolist = bio;
+			bio->bi_private = r10_bio;
+			bio->bi_end_io = end_sync_write;
+			bio->bi_rw = WRITE;
+			bio->bi_sector = sector +
+				conf->mirrors[d].replacement->data_offset;
+			bio->bi_bdev = conf->mirrors[d].replacement->bdev;
+			count++;
 		}
 
 		if (count < 2) {
@@ -2643,6 +2975,11 @@ static sector_t sync_request(struct mddev *mddev, sector_t sector_nr,
 				if (r10_bio->devs[i].bio->bi_end_io)
 					rdev_dec_pending(conf->mirrors[d].rdev,
 							 mddev);
+				if (r10_bio->devs[i].repl_bio &&
+				    r10_bio->devs[i].repl_bio->bi_end_io)
+					rdev_dec_pending(
+						conf->mirrors[d].replacement,
+						mddev);
 			}
 			put_buf(r10_bio);
 			biolist = NULL;
@@ -2896,6 +3233,16 @@ static int run(struct mddev *mddev)
 			continue;
 		disk = conf->mirrors + disk_idx;
 
+		if (test_bit(Replacement, &rdev->flags)) {
+			if (disk->replacement)
+				goto out_free_conf;
+			disk->replacement = rdev;
+		} else {
+			if (disk->rdev)
+				goto out_free_conf;
+			disk->rdev = rdev;
+		}
+
 		disk->rdev = rdev;
 		disk_stack_limits(mddev->gendisk, rdev->bdev,
 				  rdev->data_offset << 9);
@@ -2923,6 +3270,13 @@ static int run(struct mddev *mddev)
 
 		disk = conf->mirrors + i;
 
+		if (!disk->rdev && disk->replacement) {
+			/* The replacement is all we have - use it */
+			disk->rdev = disk->replacement;
+			disk->replacement = NULL;
+			clear_bit(Replacement, &disk->rdev->flags);
+		}
+
 		if (!disk->rdev ||
 		    !test_bit(In_sync, &disk->rdev->flags)) {
 			disk->head_position = 0;

drivers/md/raid10.h

@@ -2,7 +2,7 @@
 #define _RAID10_H
 
 struct mirror_info {
-	struct md_rdev	*rdev;
+	struct md_rdev	*rdev, *replacement;
 	sector_t	head_position;
 	int		recovery_disabled;	/* matches
 						 * mddev->recovery_disabled
@@ -18,12 +18,13 @@ struct r10conf {
 	spinlock_t		device_lock;
 
 	/* geometry */
-	int			near_copies;  /* number of copies laid out raid0 style */
+	int			near_copies;  /* number of copies laid out
+					       * raid0 style */
 	int 			far_copies;   /* number of copies laid out
 					       * at large strides across drives
 					       */
-	int			far_offset;   /* far_copies are offset by 1 stripe
-					       * instead of many
+	int			far_offset;   /* far_copies are offset by 1
+					       * stripe instead of many
 					       */
 	int			copies;	      /* near_copies * far_copies.
 					       * must be <= raid_disks
@@ -34,10 +35,11 @@ struct r10conf {
 					       * 1 stripe.
 					       */
 
-	sector_t		dev_sectors;  /* temp copy of mddev->dev_sectors */
+	sector_t		dev_sectors;  /* temp copy of
+					       * mddev->dev_sectors */
 
-	int chunk_shift; /* shift from chunks to sectors */
-	sector_t chunk_mask;
+	int			chunk_shift; /* shift from chunks to sectors */
+	sector_t		chunk_mask;
 
 	struct list_head	retry_list;
 	/* queue pending writes and submit them on unplug */
@@ -45,20 +47,22 @@ struct r10conf {
 	int			pending_count;
 
 	spinlock_t		resync_lock;
-	int nr_pending;
-	int nr_waiting;
-	int nr_queued;
-	int barrier;
+	int			nr_pending;
+	int			nr_waiting;
+	int			nr_queued;
+	int			barrier;
 	sector_t		next_resync;
 	int			fullsync;  /* set to 1 if a full sync is needed,
 					    * (fresh device added).
 					    * Cleared when a sync completes.
 					    */
-
+	int			have_replacement; /* There is at least one
+						   * replacement device.
+						   */
 	wait_queue_head_t	wait_barrier;
 
-	mempool_t *r10bio_pool;
-	mempool_t *r10buf_pool;
+	mempool_t		*r10bio_pool;
+	mempool_t		*r10buf_pool;
 	struct page		*tmppage;
 
 	/* When taking over an array from a different personality, we store
@@ -98,11 +102,18 @@ struct r10bio {
 	 * When resyncing we also use one for each copy.
 	 * When reconstructing, we use 2 bios, one for read, one for write.
 	 * We choose the number when they are allocated.
+	 * We sometimes need an extra bio to write to the replacement.
 	 */
 	struct {
-		struct bio		*bio;
-		sector_t addr;
-		int devnum;
+		struct bio	*bio;
+		union {
+			struct bio	*repl_bio; /* used for resync and
+						    * writes */
+			struct md_rdev	*rdev;	   /* used for reads
+						    * (read_slot >= 0) */
+		};
+		sector_t	addr;
+		int		devnum;
 	} devs[0];
 };
 
@@ -121,17 +132,19 @@ struct r10bio {
 #define BIO_SPECIAL(bio) ((unsigned long)bio <= 2)
 
 /* bits for r10bio.state */
-#define	R10BIO_Uptodate	0
-#define	R10BIO_IsSync	1
-#define	R10BIO_IsRecover 2
-#define	R10BIO_Degraded 3
+enum r10bio_state {
+	R10BIO_Uptodate,
+	R10BIO_IsSync,
+	R10BIO_IsRecover,
+	R10BIO_Degraded,
 /* Set ReadError on bios that experience a read error
  * so that raid10d knows what to do with them.
  */
-#define	R10BIO_ReadError 4
+	R10BIO_ReadError,
 /* If a write for this request means we can clear some
  * known-bad-block records, we set this flag.
  */
-#define	R10BIO_MadeGood 5
-#define	R10BIO_WriteError 6
+	R10BIO_MadeGood,
+	R10BIO_WriteError,
+};
 #endif

drivers/md/raid5.c

@@ -370,12 +370,10 @@ static struct stripe_head *__find_stripe(struct r5conf *conf, sector_t sector,
  * of the two sections, and some non-in_sync devices may
  * be insync in the section most affected by failed devices.
  */
-static int has_failed(struct r5conf *conf)
+static int calc_degraded(struct r5conf *conf)
 {
-	int degraded;
+	int degraded, degraded2;
 	int i;
-	if (conf->mddev->reshape_position == MaxSector)
-		return conf->mddev->degraded > conf->max_degraded;
 
 	rcu_read_lock();
 	degraded = 0;
@@ -399,14 +397,14 @@ static int has_failed(struct r5conf *conf)
 				degraded++;
 	}
 	rcu_read_unlock();
-	if (degraded > conf->max_degraded)
-		return 1;
+	if (conf->raid_disks == conf->previous_raid_disks)
+		return degraded;
 	rcu_read_lock();
-	degraded = 0;
+	degraded2 = 0;
 	for (i = 0; i < conf->raid_disks; i++) {
 		struct md_rdev *rdev = rcu_dereference(conf->disks[i].rdev);
 		if (!rdev || test_bit(Faulty, &rdev->flags))
-			degraded++;
+			degraded2++;
 		else if (test_bit(In_sync, &rdev->flags))
 			;
 		else
@@ -416,9 +414,22 @@ static int has_failed(struct r5conf *conf)
 			 * almost certainly hasn't.
 			 */
 			if (conf->raid_disks <= conf->previous_raid_disks)
-				degraded++;
+				degraded2++;
 	}
 	rcu_read_unlock();
+	if (degraded2 > degraded)
+		return degraded2;
+	return degraded;
+}
+
+static int has_failed(struct r5conf *conf)
+{
+	int degraded;
+
+	if (conf->mddev->reshape_position == MaxSector)
+		return conf->mddev->degraded > conf->max_degraded;
+
+	degraded = calc_degraded(conf);
 	if (degraded > conf->max_degraded)
 		return 1;
 	return 0;
@@ -492,8 +503,9 @@ static void ops_run_io(struct stripe_head *sh, struct stripe_head_state *s)
 
 	for (i = disks; i--; ) {
 		int rw;
-		struct bio *bi;
-		struct md_rdev *rdev;
+		int replace_only = 0;
+		struct bio *bi, *rbi;
+		struct md_rdev *rdev, *rrdev = NULL;
 		if (test_and_clear_bit(R5_Wantwrite, &sh->dev[i].flags)) {
 			if (test_and_clear_bit(R5_WantFUA, &sh->dev[i].flags))
 				rw = WRITE_FUA;
@@ -501,27 +513,57 @@ static void ops_run_io(struct stripe_head *sh, struct stripe_head_state *s)
 				rw = WRITE;
 		} else if (test_and_clear_bit(R5_Wantread, &sh->dev[i].flags))
 			rw = READ;
-		else
+		else if (test_and_clear_bit(R5_WantReplace,
+					    &sh->dev[i].flags)) {
+			rw = WRITE;
+			replace_only = 1;
+		} else
 			continue;
 
 		bi = &sh->dev[i].req;
+		rbi = &sh->dev[i].rreq; /* For writing to replacement */
 
 		bi->bi_rw = rw;
-		if (rw & WRITE)
+		rbi->bi_rw = rw;
+		if (rw & WRITE) {
 			bi->bi_end_io = raid5_end_write_request;
-		else
+			rbi->bi_end_io = raid5_end_write_request;
+		} else
 			bi->bi_end_io = raid5_end_read_request;
 
 		rcu_read_lock();
+		rrdev = rcu_dereference(conf->disks[i].replacement);
+		smp_mb(); /* Ensure that if rrdev is NULL, rdev won't be */
 		rdev = rcu_dereference(conf->disks[i].rdev);
+		if (!rdev) {
+			rdev = rrdev;
+			rrdev = NULL;
+		}
+		if (rw & WRITE) {
+			if (replace_only)
+				rdev = NULL;
+			if (rdev == rrdev)
+				/* We raced and saw duplicates */
+				rrdev = NULL;
+		} else {
+			if (test_bit(R5_ReadRepl, &sh->dev[i].flags) && rrdev)
+				rdev = rrdev;
+			rrdev = NULL;
+		}
+
 		if (rdev && test_bit(Faulty, &rdev->flags))
 			rdev = NULL;
 		if (rdev)
 			atomic_inc(&rdev->nr_pending);
+		if (rrdev && test_bit(Faulty, &rrdev->flags))
+			rrdev = NULL;
+		if (rrdev)
+			atomic_inc(&rrdev->nr_pending);
 		rcu_read_unlock();
 
 		/* We have already checked bad blocks for reads.  Now
-		 * need to check for writes.
+		 * need to check for writes.  We never accept write errors
+		 * on the replacement, so we don't to check rrdev.
 		 */
 		while ((rw & WRITE) && rdev &&
 		       test_bit(WriteErrorSeen, &rdev->flags)) {
@@ -551,7 +593,8 @@ static void ops_run_io(struct stripe_head *sh, struct stripe_head_state *s)
 		}
 
 		if (rdev) {
-			if (s->syncing || s->expanding || s->expanded)
+			if (s->syncing || s->expanding || s->expanded
+			    || s->replacing)
 				md_sync_acct(rdev->bdev, STRIPE_SECTORS);
 
 			set_bit(STRIPE_IO_STARTED, &sh->state);
@@ -563,16 +606,38 @@ static void ops_run_io(struct stripe_head *sh, struct stripe_head_state *s)
 			atomic_inc(&sh->count);
 			bi->bi_sector = sh->sector + rdev->data_offset;
 			bi->bi_flags = 1 << BIO_UPTODATE;
-			bi->bi_vcnt = 1;
-			bi->bi_max_vecs = 1;
 			bi->bi_idx = 0;
-			bi->bi_io_vec = &sh->dev[i].vec;
 			bi->bi_io_vec[0].bv_len = STRIPE_SIZE;
 			bi->bi_io_vec[0].bv_offset = 0;
 			bi->bi_size = STRIPE_SIZE;
 			bi->bi_next = NULL;
+			if (rrdev)
+				set_bit(R5_DOUBLE_LOCKED, &sh->dev[i].flags);
 			generic_make_request(bi);
-		} else {
+		}
+		if (rrdev) {
+			if (s->syncing || s->expanding || s->expanded
+			    || s->replacing)
+				md_sync_acct(rrdev->bdev, STRIPE_SECTORS);
+
+			set_bit(STRIPE_IO_STARTED, &sh->state);
+
+			rbi->bi_bdev = rrdev->bdev;
+			pr_debug("%s: for %llu schedule op %ld on "
+				 "replacement disc %d\n",
+				__func__, (unsigned long long)sh->sector,
+				rbi->bi_rw, i);
+			atomic_inc(&sh->count);
+			rbi->bi_sector = sh->sector + rrdev->data_offset;
+			rbi->bi_flags = 1 << BIO_UPTODATE;
+			rbi->bi_idx = 0;
+			rbi->bi_io_vec[0].bv_len = STRIPE_SIZE;
+			rbi->bi_io_vec[0].bv_offset = 0;
+			rbi->bi_size = STRIPE_SIZE;
+			rbi->bi_next = NULL;
+			generic_make_request(rbi);
+		}
+		if (!rdev && !rrdev) {
 			if (rw & WRITE)
 				set_bit(STRIPE_DEGRADED, &sh->state);
 			pr_debug("skip op %ld on disc %d for sector %llu\n",
@@ -1583,7 +1648,7 @@ static void raid5_end_read_request(struct bio * bi, int error)
 	int disks = sh->disks, i;
 	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
 	char b[BDEVNAME_SIZE];
-	struct md_rdev *rdev;
+	struct md_rdev *rdev = NULL;
 
 
 	for (i=0 ; i<disks; i++)
@@ -1597,11 +1662,23 @@ static void raid5_end_read_request(struct bio * bi, int error)
 		BUG();
 		return;
 	}
+	if (test_bit(R5_ReadRepl, &sh->dev[i].flags))
+		/* If replacement finished while this request was outstanding,
+		 * 'replacement' might be NULL already.
+		 * In that case it moved down to 'rdev'.
+		 * rdev is not removed until all requests are finished.
+		 */
+		rdev = conf->disks[i].replacement;
+	if (!rdev)
+		rdev = conf->disks[i].rdev;
 
 	if (uptodate) {
 		set_bit(R5_UPTODATE, &sh->dev[i].flags);
 		if (test_bit(R5_ReadError, &sh->dev[i].flags)) {
-			rdev = conf->disks[i].rdev;
+			/* Note that this cannot happen on a
+			 * replacement device.  We just fail those on
+			 * any error
+			 */
 			printk_ratelimited(
 				KERN_INFO
 				"md/raid:%s: read error corrected"
@@ -1614,16 +1691,24 @@ static void raid5_end_read_request(struct bio * bi, int error)
 			clear_bit(R5_ReadError, &sh->dev[i].flags);
 			clear_bit(R5_ReWrite, &sh->dev[i].flags);
 		}
-		if (atomic_read(&conf->disks[i].rdev->read_errors))
-			atomic_set(&conf->disks[i].rdev->read_errors, 0);
+		if (atomic_read(&rdev->read_errors))
+			atomic_set(&rdev->read_errors, 0);
 	} else {
-		const char *bdn = bdevname(conf->disks[i].rdev->bdev, b);
+		const char *bdn = bdevname(rdev->bdev, b);
 		int retry = 0;
-		rdev = conf->disks[i].rdev;
 
 		clear_bit(R5_UPTODATE, &sh->dev[i].flags);
 		atomic_inc(&rdev->read_errors);
-		if (conf->mddev->degraded >= conf->max_degraded)
+		if (test_bit(R5_ReadRepl, &sh->dev[i].flags))
+			printk_ratelimited(
+				KERN_WARNING
+				"md/raid:%s: read error on replacement device "
+				"(sector %llu on %s).\n",
+				mdname(conf->mddev),
+				(unsigned long long)(sh->sector
+						     + rdev->data_offset),
+				bdn);
+		else if (conf->mddev->degraded >= conf->max_degraded)
 			printk_ratelimited(
 				KERN_WARNING
 				"md/raid:%s: read error not correctable "
@@ -1657,7 +1742,7 @@ static void raid5_end_read_request(struct bio * bi, int error)
 			md_error(conf->mddev, rdev);
 		}
 	}
-	rdev_dec_pending(conf->disks[i].rdev, conf->mddev);
+	rdev_dec_pending(rdev, conf->mddev);
 	clear_bit(R5_LOCKED, &sh->dev[i].flags);
 	set_bit(STRIPE_HANDLE, &sh->state);
 	release_stripe(sh);
@@ -1668,14 +1753,30 @@ static void raid5_end_write_request(struct bio *bi, int error)
 	struct stripe_head *sh = bi->bi_private;
 	struct r5conf *conf = sh->raid_conf;
 	int disks = sh->disks, i;
+	struct md_rdev *uninitialized_var(rdev);
 	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
 	sector_t first_bad;
 	int bad_sectors;
+	int replacement = 0;
 
-	for (i=0 ; i<disks; i++)
-		if (bi == &sh->dev[i].req)
+	for (i = 0 ; i < disks; i++) {
+		if (bi == &sh->dev[i].req) {
+			rdev = conf->disks[i].rdev;
 			break;
-
+		}
+		if (bi == &sh->dev[i].rreq) {
+			rdev = conf->disks[i].replacement;
+			if (rdev)
+				replacement = 1;
+			else
+				/* rdev was removed and 'replacement'
+				 * replaced it.  rdev is not removed
+				 * until all requests are finished.
+				 */
+				rdev = conf->disks[i].rdev;
+			break;
+		}
+	}
 	pr_debug("end_write_request %llu/%d, count %d, uptodate: %d.\n",
 		(unsigned long long)sh->sector, i, atomic_read(&sh->count),
 		uptodate);
@@ -1684,21 +1785,33 @@ static void raid5_end_write_request(struct bio *bi, int error)
 		return;
 	}
 
-	if (!uptodate) {
-		set_bit(WriteErrorSeen, &conf->disks[i].rdev->flags);
-		set_bit(R5_WriteError, &sh->dev[i].flags);
-	} else if (is_badblock(conf->disks[i].rdev, sh->sector, STRIPE_SECTORS,
-			       &first_bad, &bad_sectors))
-		set_bit(R5_MadeGood, &sh->dev[i].flags);
+	if (replacement) {
+		if (!uptodate)
+			md_error(conf->mddev, rdev);
+		else if (is_badblock(rdev, sh->sector,
+				     STRIPE_SECTORS,
+				     &first_bad, &bad_sectors))
+			set_bit(R5_MadeGoodRepl, &sh->dev[i].flags);
+	} else {
+		if (!uptodate) {
+			set_bit(WriteErrorSeen, &rdev->flags);
+			set_bit(R5_WriteError, &sh->dev[i].flags);
+			if (!test_and_set_bit(WantReplacement, &rdev->flags))
+				set_bit(MD_RECOVERY_NEEDED,
+					&rdev->mddev->recovery);
+		} else if (is_badblock(rdev, sh->sector,
+				       STRIPE_SECTORS,
+				       &first_bad, &bad_sectors))
+			set_bit(R5_MadeGood, &sh->dev[i].flags);
+	}
+	rdev_dec_pending(rdev, conf->mddev);
 
-	rdev_dec_pending(conf->disks[i].rdev, conf->mddev);
-	
-	clear_bit(R5_LOCKED, &sh->dev[i].flags);
+	if (!test_and_clear_bit(R5_DOUBLE_LOCKED, &sh->dev[i].flags))
+		clear_bit(R5_LOCKED, &sh->dev[i].flags);
 	set_bit(STRIPE_HANDLE, &sh->state);
 	release_stripe(sh);
 }
 
-
 static sector_t compute_blocknr(struct stripe_head *sh, int i, int previous);
 	
 static void raid5_build_block(struct stripe_head *sh, int i, int previous)
@@ -1709,12 +1822,15 @@ static void raid5_build_block(struct stripe_head *sh, int i, int previous)
 	dev->req.bi_io_vec = &dev->vec;
 	dev->req.bi_vcnt++;
 	dev->req.bi_max_vecs++;
+	dev->req.bi_private = sh;
 	dev->vec.bv_page = dev->page;
-	dev->vec.bv_len = STRIPE_SIZE;
-	dev->vec.bv_offset = 0;
 
-	dev->req.bi_sector = sh->sector;
-	dev->req.bi_private = sh;
+	bio_init(&dev->rreq);
+	dev->rreq.bi_io_vec = &dev->rvec;
+	dev->rreq.bi_vcnt++;
+	dev->rreq.bi_max_vecs++;
+	dev->rreq.bi_private = sh;
+	dev->rvec.bv_page = dev->page;
 
 	dev->flags = 0;
 	dev->sector = compute_blocknr(sh, i, previous);
@@ -1724,18 +1840,15 @@ static void error(struct mddev *mddev, struct md_rdev *rdev)
 {
 	char b[BDEVNAME_SIZE];
 	struct r5conf *conf = mddev->private;
+	unsigned long flags;
 	pr_debug("raid456: error called\n");
 
-	if (test_and_clear_bit(In_sync, &rdev->flags)) {
-		unsigned long flags;
-		spin_lock_irqsave(&conf->device_lock, flags);
-		mddev->degraded++;
-		spin_unlock_irqrestore(&conf->device_lock, flags);
-		/*
-		 * if recovery was running, make sure it aborts.
-		 */
-		set_bit(MD_RECOVERY_INTR, &mddev->recovery);
-	}
+	spin_lock_irqsave(&conf->device_lock, flags);
+	clear_bit(In_sync, &rdev->flags);
+	mddev->degraded = calc_degraded(conf);
+	spin_unlock_irqrestore(&conf->device_lock, flags);
+	set_bit(MD_RECOVERY_INTR, &mddev->recovery);
+
 	set_bit(Blocked, &rdev->flags);
 	set_bit(Faulty, &rdev->flags);
 	set_bit(MD_CHANGE_DEVS, &mddev->flags);
@@ -2362,8 +2475,9 @@ handle_failed_sync(struct r5conf *conf, struct stripe_head *sh,
 	md_done_sync(conf->mddev, STRIPE_SECTORS, 0);
 	clear_bit(STRIPE_SYNCING, &sh->state);
 	s->syncing = 0;
+	s->replacing = 0;
 	/* There is nothing more to do for sync/check/repair.
-	 * For recover we need to record a bad block on all
+	 * For recover/replace we need to record a bad block on all
 	 * non-sync devices, or abort the recovery
 	 */
 	if (!test_bit(MD_RECOVERY_RECOVER, &conf->mddev->recovery))
@@ -2373,12 +2487,18 @@ handle_failed_sync(struct r5conf *conf, struct stripe_head *sh,
 	 */
 	for (i = 0; i < conf->raid_disks; i++) {
 		struct md_rdev *rdev = conf->disks[i].rdev;
-		if (!rdev
-		    || test_bit(Faulty, &rdev->flags)
-		    || test_bit(In_sync, &rdev->flags))
-			continue;
-		if (!rdev_set_badblocks(rdev, sh->sector,
-					STRIPE_SECTORS, 0))
+		if (rdev
+		    && !test_bit(Faulty, &rdev->flags)
+		    && !test_bit(In_sync, &rdev->flags)
+		    && !rdev_set_badblocks(rdev, sh->sector,
+					   STRIPE_SECTORS, 0))
+			abort = 1;
+		rdev = conf->disks[i].replacement;
+		if (rdev
+		    && !test_bit(Faulty, &rdev->flags)
+		    && !test_bit(In_sync, &rdev->flags)
+		    && !rdev_set_badblocks(rdev, sh->sector,
+					   STRIPE_SECTORS, 0))
 			abort = 1;
 	}
 	if (abort) {
@@ -2387,6 +2507,22 @@ handle_failed_sync(struct r5conf *conf, struct stripe_head *sh,
 	}
 }
 
+static int want_replace(struct stripe_head *sh, int disk_idx)
+{
+	struct md_rdev *rdev;
+	int rv = 0;
+	/* Doing recovery so rcu locking not required */
+	rdev = sh->raid_conf->disks[disk_idx].replacement;
+	if (rdev
+	    && !test_bit(Faulty, &rdev->flags)
+	    && !test_bit(In_sync, &rdev->flags)
+	    && (rdev->recovery_offset <= sh->sector
+		|| rdev->mddev->recovery_cp <= sh->sector))
+		rv = 1;
+
+	return rv;
+}
+
 /* fetch_block - checks the given member device to see if its data needs
  * to be read or computed to satisfy a request.
  *
@@ -2406,6 +2542,7 @@ static int fetch_block(struct stripe_head *sh, struct stripe_head_state *s,
 	    (dev->toread ||
 	     (dev->towrite && !test_bit(R5_OVERWRITE, &dev->flags)) ||
 	     s->syncing || s->expanding ||
+	     (s->replacing && want_replace(sh, disk_idx)) ||
 	     (s->failed >= 1 && fdev[0]->toread) ||
 	     (s->failed >= 2 && fdev[1]->toread) ||
 	     (sh->raid_conf->level <= 5 && s->failed && fdev[0]->towrite &&
@@ -2959,22 +3096,18 @@ static void handle_stripe_expansion(struct r5conf *conf, struct stripe_head *sh)
 	}
 }
 
-
 /*
  * handle_stripe - do things to a stripe.
  *
- * We lock the stripe and then examine the state of various bits
- * to see what needs to be done.
+ * We lock the stripe by setting STRIPE_ACTIVE and then examine the
+ * state of various bits to see what needs to be done.
  * Possible results:
- *    return some read request which now have data
- *    return some write requests which are safely on disc
+ *    return some read requests which now have data
+ *    return some write requests which are safely on storage
  *    schedule a read on some buffers
  *    schedule a write of some buffers
  *    return confirmation of parity correctness
  *
- * buffers are taken off read_list or write_list, and bh_cache buffers
- * get BH_Lock set before the stripe lock is released.
- *
  */
 
 static void analyse_stripe(struct stripe_head *sh, struct stripe_head_state *s)
@@ -2983,10 +3116,10 @@ static void analyse_stripe(struct stripe_head *sh, struct stripe_head_state *s)
 	int disks = sh->disks;
 	struct r5dev *dev;
 	int i;
+	int do_recovery = 0;
 
 	memset(s, 0, sizeof(*s));
 
-	s->syncing = test_bit(STRIPE_SYNCING, &sh->state);
 	s->expanding = test_bit(STRIPE_EXPAND_SOURCE, &sh->state);
 	s->expanded = test_bit(STRIPE_EXPAND_READY, &sh->state);
 	s->failed_num[0] = -1;
@@ -3004,7 +3137,8 @@ static void analyse_stripe(struct stripe_head *sh, struct stripe_head_state *s)
 		dev = &sh->dev[i];
 
 		pr_debug("check %d: state 0x%lx read %p write %p written %p\n",
-			i, dev->flags, dev->toread, dev->towrite, dev->written);
+			 i, dev->flags,
+			 dev->toread, dev->towrite, dev->written);
 		/* maybe we can reply to a read
 		 *
 		 * new wantfill requests are only permitted while
@@ -3035,7 +3169,21 @@ static void analyse_stripe(struct stripe_head *sh, struct stripe_head_state *s)
 		}
 		if (dev->written)
 			s->written++;
-		rdev = rcu_dereference(conf->disks[i].rdev);
+		/* Prefer to use the replacement for reads, but only
+		 * if it is recovered enough and has no bad blocks.
+		 */
+		rdev = rcu_dereference(conf->disks[i].replacement);
+		if (rdev && !test_bit(Faulty, &rdev->flags) &&
+		    rdev->recovery_offset >= sh->sector + STRIPE_SECTORS &&
+		    !is_badblock(rdev, sh->sector, STRIPE_SECTORS,
+				 &first_bad, &bad_sectors))
+			set_bit(R5_ReadRepl, &dev->flags);
+		else {
+			if (rdev)
+				set_bit(R5_NeedReplace, &dev->flags);
+			rdev = rcu_dereference(conf->disks[i].rdev);
+			clear_bit(R5_ReadRepl, &dev->flags);
+		}
 		if (rdev && test_bit(Faulty, &rdev->flags))
 			rdev = NULL;
 		if (rdev) {
@@ -3077,20 +3225,38 @@ static void analyse_stripe(struct stripe_head *sh, struct stripe_head_state *s)
 			set_bit(R5_Insync, &dev->flags);
 
 		if (rdev && test_bit(R5_WriteError, &dev->flags)) {
-			clear_bit(R5_Insync, &dev->flags);
-			if (!test_bit(Faulty, &rdev->flags)) {
+			/* This flag does not apply to '.replacement'
+			 * only to .rdev, so make sure to check that*/
+			struct md_rdev *rdev2 = rcu_dereference(
+				conf->disks[i].rdev);
+			if (rdev2 == rdev)
+				clear_bit(R5_Insync, &dev->flags);
+			if (rdev2 && !test_bit(Faulty, &rdev2->flags)) {
 				s->handle_bad_blocks = 1;
-				atomic_inc(&rdev->nr_pending);
+				atomic_inc(&rdev2->nr_pending);
 			} else
 				clear_bit(R5_WriteError, &dev->flags);
 		}
 		if (rdev && test_bit(R5_MadeGood, &dev->flags)) {
-			if (!test_bit(Faulty, &rdev->flags)) {
+			/* This flag does not apply to '.replacement'
+			 * only to .rdev, so make sure to check that*/
+			struct md_rdev *rdev2 = rcu_dereference(
+				conf->disks[i].rdev);
+			if (rdev2 && !test_bit(Faulty, &rdev2->flags)) {
 				s->handle_bad_blocks = 1;
-				atomic_inc(&rdev->nr_pending);
+				atomic_inc(&rdev2->nr_pending);
 			} else
 				clear_bit(R5_MadeGood, &dev->flags);
 		}
+		if (test_bit(R5_MadeGoodRepl, &dev->flags)) {
+			struct md_rdev *rdev2 = rcu_dereference(
+				conf->disks[i].replacement);
+			if (rdev2 && !test_bit(Faulty, &rdev2->flags)) {
+				s->handle_bad_blocks = 1;
+				atomic_inc(&rdev2->nr_pending);
+			} else
+				clear_bit(R5_MadeGoodRepl, &dev->flags);
+		}
 		if (!test_bit(R5_Insync, &dev->flags)) {
 			/* The ReadError flag will just be confusing now */
 			clear_bit(R5_ReadError, &dev->flags);
@@ -3102,9 +3268,25 @@ static void analyse_stripe(struct stripe_head *sh, struct stripe_head_state *s)
 			if (s->failed < 2)
 				s->failed_num[s->failed] = i;
 			s->failed++;
+			if (rdev && !test_bit(Faulty, &rdev->flags))
+				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.
+		 * else if we are after recovery_cp, we must be syncing
+		 * else we can only be replacing
+		 * sync and recovery both need to read all devices, and so
+		 * use the same flag.
+		 */
+		if (do_recovery ||
+		    sh->sector >= conf->mddev->recovery_cp)
+			s->syncing = 1;
+		else
+			s->replacing = 1;
+	}
 	rcu_read_unlock();
 }
 
@@ -3146,7 +3328,7 @@ static void handle_stripe(struct stripe_head *sh)
 
 	if (unlikely(s.blocked_rdev)) {
 		if (s.syncing || s.expanding || s.expanded ||
-		    s.to_write || s.written) {
+		    s.replacing || s.to_write || s.written) {
 			set_bit(STRIPE_HANDLE, &sh->state);
 			goto finish;
 		}
@@ -3172,7 +3354,7 @@ static void handle_stripe(struct stripe_head *sh)
 		sh->reconstruct_state = 0;
 		if (s.to_read+s.to_write+s.written)
 			handle_failed_stripe(conf, sh, &s, disks, &s.return_bi);
-		if (s.syncing)
+		if (s.syncing + s.replacing)
 			handle_failed_sync(conf, sh, &s);
 	}
 
@@ -3203,7 +3385,9 @@ static void handle_stripe(struct stripe_head *sh)
 	 */
 	if (s.to_read || s.non_overwrite
 	    || (conf->level == 6 && s.to_write && s.failed)
-	    || (s.syncing && (s.uptodate + s.compute < disks)) || s.expanding)
+	    || (s.syncing && (s.uptodate + s.compute < disks))
+	    || s.replacing
+	    || s.expanding)
 		handle_stripe_fill(sh, &s, disks);
 
 	/* Now we check to see if any write operations have recently
@@ -3265,7 +3449,20 @@ static void handle_stripe(struct stripe_head *sh)
 			handle_parity_checks5(conf, sh, &s, disks);
 	}
 
-	if (s.syncing && s.locked == 0 && test_bit(STRIPE_INSYNC, &sh->state)) {
+	if (s.replacing && s.locked == 0
+	    && !test_bit(STRIPE_INSYNC, &sh->state)) {
+		/* Write out to replacement devices where possible */
+		for (i = 0; i < conf->raid_disks; i++)
+			if (test_bit(R5_UPTODATE, &sh->dev[i].flags) &&
+			    test_bit(R5_NeedReplace, &sh->dev[i].flags)) {
+				set_bit(R5_WantReplace, &sh->dev[i].flags);
+				set_bit(R5_LOCKED, &sh->dev[i].flags);
+				s.locked++;
+			}
+		set_bit(STRIPE_INSYNC, &sh->state);
+	}
+	if ((s.syncing || s.replacing) && s.locked == 0 &&
+	    test_bit(STRIPE_INSYNC, &sh->state)) {
 		md_done_sync(conf->mddev, STRIPE_SECTORS, 1);
 		clear_bit(STRIPE_SYNCING, &sh->state);
 	}
@@ -3363,6 +3560,15 @@ static void handle_stripe(struct stripe_head *sh)
 						     STRIPE_SECTORS);
 				rdev_dec_pending(rdev, conf->mddev);
 			}
+			if (test_and_clear_bit(R5_MadeGoodRepl, &dev->flags)) {
+				rdev = conf->disks[i].replacement;
+				if (!rdev)
+					/* rdev have been moved down */
+					rdev = conf->disks[i].rdev;
+				rdev_clear_badblocks(rdev, sh->sector,
+						     STRIPE_SECTORS);
+				rdev_dec_pending(rdev, conf->mddev);
+			}
 		}
 
 	if (s.ops_request)
@@ -3586,6 +3792,7 @@ static int chunk_aligned_read(struct mddev *mddev, struct bio * raid_bio)
 	int dd_idx;
 	struct bio* align_bi;
 	struct md_rdev *rdev;
+	sector_t end_sector;
 
 	if (!in_chunk_boundary(mddev, raid_bio)) {
 		pr_debug("chunk_aligned_read : non aligned\n");
@@ -3610,9 +3817,19 @@ static int chunk_aligned_read(struct mddev *mddev, struct bio * raid_bio)
 						    0,
 						    &dd_idx, NULL);
 
+	end_sector = align_bi->bi_sector + (align_bi->bi_size >> 9);
 	rcu_read_lock();
-	rdev = rcu_dereference(conf->disks[dd_idx].rdev);
-	if (rdev && test_bit(In_sync, &rdev->flags)) {
+	rdev = rcu_dereference(conf->disks[dd_idx].replacement);
+	if (!rdev || test_bit(Faulty, &rdev->flags) ||
+	    rdev->recovery_offset < end_sector) {
+		rdev = rcu_dereference(conf->disks[dd_idx].rdev);
+		if (rdev &&
+		    (test_bit(Faulty, &rdev->flags) ||
+		    !(test_bit(In_sync, &rdev->flags) ||
+		      rdev->recovery_offset >= end_sector)))
+			rdev = NULL;
+	}
+	if (rdev) {
 		sector_t first_bad;
 		int bad_sectors;
 
@@ -4137,7 +4354,6 @@ static inline sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int
 		return sync_blocks * STRIPE_SECTORS; /* keep things rounded to whole stripes */
 	}
 
-
 	bitmap_cond_end_sync(mddev->bitmap, sector_nr);
 
 	sh = get_active_stripe(conf, sector_nr, 0, 1, 0);
@@ -4208,7 +4424,6 @@ static int  retry_aligned_read(struct r5conf *conf, struct bio *raid_bio)
 			return handled;
 		}
 
-		set_bit(R5_ReadError, &sh->dev[dd_idx].flags);
 		if (!add_stripe_bio(sh, raid_bio, dd_idx, 0)) {
 			release_stripe(sh);
 			raid5_set_bi_hw_segments(raid_bio, scnt);
@@ -4635,7 +4850,15 @@ static struct r5conf *setup_conf(struct mddev *mddev)
 			continue;
 		disk = conf->disks + raid_disk;
 
-		disk->rdev = rdev;
+		if (test_bit(Replacement, &rdev->flags)) {
+			if (disk->replacement)
+				goto abort;
+			disk->replacement = rdev;
+		} else {
+			if (disk->rdev)
+				goto abort;
+			disk->rdev = rdev;
+		}
 
 		if (test_bit(In_sync, &rdev->flags)) {
 			char b[BDEVNAME_SIZE];
@@ -4724,6 +4947,7 @@ static int run(struct mddev *mddev)
 	int dirty_parity_disks = 0;
 	struct md_rdev *rdev;
 	sector_t reshape_offset = 0;
+	int i;
 
 	if (mddev->recovery_cp != MaxSector)
 		printk(KERN_NOTICE "md/raid:%s: not clean"
@@ -4813,12 +5037,25 @@ static int run(struct mddev *mddev)
 	conf->thread = NULL;
 	mddev->private = conf;
 
-	/*
-	 * 0 for a fully functional array, 1 or 2 for a degraded array.
-	 */
-	list_for_each_entry(rdev, &mddev->disks, same_set) {
-		if (rdev->raid_disk < 0)
+	for (i = 0; i < conf->raid_disks && conf->previous_raid_disks;
+	     i++) {
+		rdev = conf->disks[i].rdev;
+		if (!rdev && conf->disks[i].replacement) {
+			/* The replacement is all we have yet */
+			rdev = conf->disks[i].replacement;
+			conf->disks[i].replacement = NULL;
+			clear_bit(Replacement, &rdev->flags);
+			conf->disks[i].rdev = rdev;
+		}
+		if (!rdev)
 			continue;
+		if (conf->disks[i].replacement &&
+		    conf->reshape_progress != MaxSector) {
+			/* replacements and reshape simply do not mix. */
+			printk(KERN_ERR "md: cannot handle concurrent "
+			       "replacement and reshape.\n");
+			goto abort;
+		}
 		if (test_bit(In_sync, &rdev->flags)) {
 			working_disks++;
 			continue;
@@ -4852,8 +5089,10 @@ static int run(struct mddev *mddev)
 		dirty_parity_disks++;
 	}
 
-	mddev->degraded = (max(conf->raid_disks, conf->previous_raid_disks)
-			   - working_disks);
+	/*
+	 * 0 for a fully functional array, 1 or 2 for a degraded array.
+	 */
+	mddev->degraded = calc_degraded(conf);
 
 	if (has_failed(conf)) {
 		printk(KERN_ERR "md/raid:%s: not enough operational devices"
@@ -5016,7 +5255,25 @@ static int raid5_spare_active(struct mddev *mddev)
 
 	for (i = 0; i < conf->raid_disks; i++) {
 		tmp = conf->disks + i;
-		if (tmp->rdev
+		if (tmp->replacement
+		    && tmp->replacement->recovery_offset == MaxSector
+		    && !test_bit(Faulty, &tmp->replacement->flags)
+		    && !test_and_set_bit(In_sync, &tmp->replacement->flags)) {
+			/* Replacement has just become active. */
+			if (!tmp->rdev
+			    || !test_and_clear_bit(In_sync, &tmp->rdev->flags))
+				count++;
+			if (tmp->rdev) {
+				/* Replaced device not technically faulty,
+				 * but we need to be sure it gets removed
+				 * and never re-added.
+				 */
+				set_bit(Faulty, &tmp->rdev->flags);
+				sysfs_notify_dirent_safe(
+					tmp->rdev->sysfs_state);
+			}
+			sysfs_notify_dirent_safe(tmp->replacement->sysfs_state);
+		} else if (tmp->rdev
 		    && tmp->rdev->recovery_offset == MaxSector
 		    && !test_bit(Faulty, &tmp->rdev->flags)
 		    && !test_and_set_bit(In_sync, &tmp->rdev->flags)) {
@@ -5025,49 +5282,68 @@ static int raid5_spare_active(struct mddev *mddev)
 		}
 	}
 	spin_lock_irqsave(&conf->device_lock, flags);
-	mddev->degraded -= count;
+	mddev->degraded = calc_degraded(conf);
 	spin_unlock_irqrestore(&conf->device_lock, flags);
 	print_raid5_conf(conf);
 	return count;
 }
 
-static int raid5_remove_disk(struct mddev *mddev, int number)
+static int raid5_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
 {
 	struct r5conf *conf = mddev->private;
 	int err = 0;
-	struct md_rdev *rdev;
+	int number = rdev->raid_disk;
+	struct md_rdev **rdevp;
 	struct disk_info *p = conf->disks + number;
 
 	print_raid5_conf(conf);
-	rdev = p->rdev;
-	if (rdev) {
-		if (number >= conf->raid_disks &&
-		    conf->reshape_progress == MaxSector)
-			clear_bit(In_sync, &rdev->flags);
+	if (rdev == p->rdev)
+		rdevp = &p->rdev;
+	else if (rdev == p->replacement)
+		rdevp = &p->replacement;
+	else
+		return 0;
 
-		if (test_bit(In_sync, &rdev->flags) ||
-		    atomic_read(&rdev->nr_pending)) {
-			err = -EBUSY;
-			goto abort;
-		}
-		/* Only remove non-faulty devices if recovery
-		 * isn't possible.
-		 */
-		if (!test_bit(Faulty, &rdev->flags) &&
-		    mddev->recovery_disabled != conf->recovery_disabled &&
-		    !has_failed(conf) &&
-		    number < conf->raid_disks) {
-			err = -EBUSY;
-			goto abort;
-		}
-		p->rdev = NULL;
-		synchronize_rcu();
-		if (atomic_read(&rdev->nr_pending)) {
-			/* lost the race, try later */
-			err = -EBUSY;
-			p->rdev = rdev;
-		}
+	if (number >= conf->raid_disks &&
+	    conf->reshape_progress == MaxSector)
+		clear_bit(In_sync, &rdev->flags);
+
+	if (test_bit(In_sync, &rdev->flags) ||
+	    atomic_read(&rdev->nr_pending)) {
+		err = -EBUSY;
+		goto abort;
 	}
+	/* Only remove non-faulty devices if recovery
+	 * isn't possible.
+	 */
+	if (!test_bit(Faulty, &rdev->flags) &&
+	    mddev->recovery_disabled != conf->recovery_disabled &&
+	    !has_failed(conf) &&
+	    (!p->replacement || p->replacement == rdev) &&
+	    number < conf->raid_disks) {
+		err = -EBUSY;
+		goto abort;
+	}
+	*rdevp = NULL;
+	synchronize_rcu();
+	if (atomic_read(&rdev->nr_pending)) {
+		/* lost the race, try later */
+		err = -EBUSY;
+		*rdevp = rdev;
+	} else if (p->replacement) {
+		/* We must have just cleared 'rdev' */
+		p->rdev = p->replacement;
+		clear_bit(Replacement, &p->replacement->flags);
+		smp_mb(); /* Make sure other CPUs may see both as identical
+			   * but will never see neither - if they are careful
+			   */
+		p->replacement = NULL;
+		clear_bit(WantReplacement, &rdev->flags);
+	} else
+		/* We might have just removed the Replacement as faulty-
+		 * clear the bit just in case
+		 */
+		clear_bit(WantReplacement, &rdev->flags);
 abort:
 
 	print_raid5_conf(conf);
@@ -5103,8 +5379,9 @@ static int raid5_add_disk(struct mddev *mddev, struct md_rdev *rdev)
 		disk = rdev->saved_raid_disk;
 	else
 		disk = first;
-	for ( ; disk <= last ; disk++)
-		if ((p=conf->disks + disk)->rdev == NULL) {
+	for ( ; disk <= last ; disk++) {
+		p = conf->disks + disk;
+		if (p->rdev == NULL) {
 			clear_bit(In_sync, &rdev->flags);
 			rdev->raid_disk = disk;
 			err = 0;
@@ -5113,6 +5390,17 @@ static int raid5_add_disk(struct mddev *mddev, struct md_rdev *rdev)
 			rcu_assign_pointer(p->rdev, rdev);
 			break;
 		}
+		if (test_bit(WantReplacement, &p->rdev->flags) &&
+		    p->replacement == NULL) {
+			clear_bit(In_sync, &rdev->flags);
+			set_bit(Replacement, &rdev->flags);
+			rdev->raid_disk = disk;
+			err = 0;
+			conf->fullsync = 1;
+			rcu_assign_pointer(p->replacement, rdev);
+			break;
+		}
+	}
 	print_raid5_conf(conf);
 	return err;
 }
@@ -5286,8 +5574,7 @@ static int raid5_start_reshape(struct mddev *mddev)
 		 * pre and post number of devices.
 		 */
 		spin_lock_irqsave(&conf->device_lock, flags);
-		mddev->degraded += (conf->raid_disks - conf->previous_raid_disks)
-			- added_devices;
+		mddev->degraded = calc_degraded(conf);
 		spin_unlock_irqrestore(&conf->device_lock, flags);
 	}
 	mddev->raid_disks = conf->raid_disks;
@@ -5356,17 +5643,15 @@ static void raid5_finish_reshape(struct mddev *mddev)
 			revalidate_disk(mddev->gendisk);
 		} else {
 			int d;
-			mddev->degraded = conf->raid_disks;
-			for (d = 0; d < conf->raid_disks ; d++)
-				if (conf->disks[d].rdev &&
-				    test_bit(In_sync,
-					     &conf->disks[d].rdev->flags))
-					mddev->degraded--;
+			spin_lock_irq(&conf->device_lock);
+			mddev->degraded = calc_degraded(conf);
+			spin_unlock_irq(&conf->device_lock);
 			for (d = conf->raid_disks ;
 			     d < conf->raid_disks - mddev->delta_disks;
 			     d++) {
 				struct md_rdev *rdev = conf->disks[d].rdev;
-				if (rdev && raid5_remove_disk(mddev, d) == 0) {
+				if (rdev &&
+				    raid5_remove_disk(mddev, rdev) == 0) {
 					sysfs_unlink_rdev(mddev, rdev);
 					rdev->raid_disk = -1;
 				}

drivers/md/raid5.h

@@ -27,7 +27,7 @@
  * The possible state transitions are:
  *
  *  Empty -> Want   - on read or write to get old data for  parity calc
- *  Empty -> Dirty  - on compute_parity to satisfy write/sync request.(RECONSTRUCT_WRITE)
+ *  Empty -> Dirty  - on compute_parity to satisfy write/sync request.
  *  Empty -> Clean  - on compute_block when computing a block for failed drive
  *  Want  -> Empty  - on failed read
  *  Want  -> Clean  - on successful completion of read request
@@ -226,8 +226,11 @@ struct stripe_head {
 		#endif
 	} ops;
 	struct r5dev {
-		struct bio	req;
-		struct bio_vec	vec;
+		/* rreq and rvec are used for the replacement device when
+		 * writing data to both devices.
+		 */
+		struct bio	req, rreq;
+		struct bio_vec	vec, rvec;
 		struct page	*page;
 		struct bio	*toread, *read, *towrite, *written;
 		sector_t	sector;			/* sector of this page */
@@ -239,7 +242,13 @@ struct stripe_head {
  *     for handle_stripe.
  */
 struct stripe_head_state {
-	int syncing, expanding, expanded;
+	/* 'syncing' means that we need to read all devices, either
+	 * to check/correct parity, or to reconstruct a missing device.
+	 * 'replacing' means we are replacing one or more drives and
+	 * the source is valid at this point so we don't need to
+	 * read all devices, just the replacement targets.
+	 */
+	int syncing, expanding, expanded, replacing;
 	int locked, uptodate, to_read, to_write, failed, written;
 	int to_fill, compute, req_compute, non_overwrite;
 	int failed_num[2];
@@ -252,38 +261,41 @@ struct stripe_head_state {
 	int handle_bad_blocks;
 };
 
-/* Flags */
-#define	R5_UPTODATE	0	/* page contains current data */
-#define	R5_LOCKED	1	/* IO has been submitted on "req" */
-#define	R5_OVERWRITE	2	/* towrite covers whole page */
+/* Flags for struct r5dev.flags */
+enum r5dev_flags {
+	R5_UPTODATE,	/* page contains current data */
+	R5_LOCKED,	/* IO has been submitted on "req" */
+	R5_DOUBLE_LOCKED,/* Cannot clear R5_LOCKED until 2 writes complete */
+	R5_OVERWRITE,	/* towrite covers whole page */
 /* and some that are internal to handle_stripe */
-#define	R5_Insync	3	/* rdev && rdev->in_sync at start */
-#define	R5_Wantread	4	/* want to schedule a read */
-#define	R5_Wantwrite	5
-#define	R5_Overlap	7	/* There is a pending overlapping request on this block */
-#define	R5_ReadError	8	/* seen a read error here recently */
-#define	R5_ReWrite	9	/* have tried to over-write the readerror */
+	R5_Insync,	/* rdev && rdev->in_sync at start */
+	R5_Wantread,	/* want to schedule a read */
+	R5_Wantwrite,
+	R5_Overlap,	/* There is a pending overlapping request
+			 * on this block */
+	R5_ReadError,	/* seen a read error here recently */
+	R5_ReWrite,	/* have tried to over-write the readerror */
 
-#define	R5_Expanded	10	/* This block now has post-expand data */
-#define	R5_Wantcompute	11	/* compute_block in progress treat as
-				 * uptodate
-				 */
-#define	R5_Wantfill	12	/* dev->toread contains a bio that needs
-				 * filling
-				 */
-#define	R5_Wantdrain	13	/* dev->towrite needs to be drained */
-#define	R5_WantFUA	14	/* Write should be FUA */
-#define	R5_WriteError	15	/* got a write error - need to record it */
-#define	R5_MadeGood	16	/* A bad block has been fixed by writing to it*/
-/*
- * Write method
- */
-#define RECONSTRUCT_WRITE	1
-#define READ_MODIFY_WRITE	2
-/* not a write method, but a compute_parity mode */
-#define	CHECK_PARITY		3
-/* Additional compute_parity mode -- updates the parity w/o LOCKING */
-#define UPDATE_PARITY		4
+	R5_Expanded,	/* This block now has post-expand data */
+	R5_Wantcompute,	/* compute_block in progress treat as
+			 * uptodate
+			 */
+	R5_Wantfill,	/* dev->toread contains a bio that needs
+			 * filling
+			 */
+	R5_Wantdrain,	/* dev->towrite needs to be drained */
+	R5_WantFUA,	/* Write should be FUA */
+	R5_WriteError,	/* got a write error - need to record it */
+	R5_MadeGood,	/* A bad block has been fixed by writing to it */
+	R5_ReadRepl,	/* Will/did read from replacement rather than orig */
+	R5_MadeGoodRepl,/* A bad block on the replacement device has been
+			 * fixed by writing to it */
+	R5_NeedReplace,	/* This device has a replacement which is not
+			 * up-to-date at this stripe. */
+	R5_WantReplace, /* We need to update the replacement, we have read
+			 * data in, and now is a good time to write it out.
+			 */
+};
 
 /*
  * Stripe state
@@ -311,13 +323,14 @@ enum {
 /*
  * Operation request flags
  */
-#define STRIPE_OP_BIOFILL	0
-#define STRIPE_OP_COMPUTE_BLK	1
-#define STRIPE_OP_PREXOR	2
-#define STRIPE_OP_BIODRAIN	3
-#define STRIPE_OP_RECONSTRUCT	4
-#define STRIPE_OP_CHECK	5
-
+enum {
+	STRIPE_OP_BIOFILL,
+	STRIPE_OP_COMPUTE_BLK,
+	STRIPE_OP_PREXOR,
+	STRIPE_OP_BIODRAIN,
+	STRIPE_OP_RECONSTRUCT,
+	STRIPE_OP_CHECK,
+};
 /*
  * Plugging:
  *
@@ -344,13 +357,12 @@ enum {
 
 
 struct disk_info {
-	struct md_rdev	*rdev;
+	struct md_rdev	*rdev, *replacement;
 };
 
 struct r5conf {
 	struct hlist_head	*stripe_hashtbl;
 	struct mddev		*mddev;
-	struct disk_info	*spare;
 	int			chunk_sectors;
 	int			level, algorithm;
 	int			max_degraded;

include/linux/raid/md_p.h

@@ -277,7 +277,10 @@ struct mdp_superblock_1 {
 					   */
 #define	MD_FEATURE_RESHAPE_ACTIVE	4
 #define	MD_FEATURE_BAD_BLOCKS		8 /* badblock list is not empty */
-
-#define	MD_FEATURE_ALL			(1|2|4|8)
+#define	MD_FEATURE_REPLACEMENT		16 /* This device is replacing an
+					    * active device with same 'role'.
+					    * 'recovery_offset' is also set.
+					    */
+#define	MD_FEATURE_ALL			(1|2|4|8|16)
 
 #endif 

include/linux/raid/pq.h

@@ -132,7 +132,7 @@ void raid6_dual_recov(int disks, size_t bytes, int faila, int failb,
 						     PROT_READ|PROT_WRITE,   \
 						     MAP_PRIVATE|MAP_ANONYMOUS,\
 						     0, 0))
-# define free_pages(x, y)	munmap((void *)(x), (y)*PAGE_SIZE)
+# define free_pages(x, y)	munmap((void *)(x), PAGE_SIZE << (y))
 
 static inline void cpu_relax(void)
 {