diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
index 39b2058845f5..909450414c67 100644
--- a/drivers/md/raid10.c
+++ b/drivers/md/raid10.c
@@ -327,6 +327,16 @@ static void raid10_end_read_request(struct bio *bio, int error)
 	}
 }
 
+static void close_write(r10bio_t *r10_bio)
+{
+	/* clear the bitmap if all writes complete successfully */
+	bitmap_endwrite(r10_bio->mddev->bitmap, r10_bio->sector,
+			r10_bio->sectors,
+			!test_bit(R10BIO_Degraded, &r10_bio->state),
+			0);
+	md_write_end(r10_bio->mddev);
+}
+
 static void raid10_end_write_request(struct bio *bio, int error)
 {
 	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
@@ -342,9 +352,9 @@ static void raid10_end_write_request(struct bio *bio, int error)
 	 * this branch is our 'one mirror IO has finished' event handler:
 	 */
 	if (!uptodate) {
-		md_error(r10_bio->mddev, conf->mirrors[dev].rdev);
-		/* an I/O failed, we can't clear the bitmap */
-		set_bit(R10BIO_Degraded, &r10_bio->state);
+		set_bit(WriteErrorSeen,	&conf->mirrors[dev].rdev->flags);
+		set_bit(R10BIO_WriteError, &r10_bio->state);
+		dec_rdev = 0;
 	} else {
 		/*
 		 * Set R10BIO_Uptodate in our master bio, so that
@@ -378,16 +388,15 @@ static void raid10_end_write_request(struct bio *bio, int error)
 	 * already.
 	 */
 	if (atomic_dec_and_test(&r10_bio->remaining)) {
-		/* clear the bitmap if all writes complete successfully */
-		bitmap_endwrite(r10_bio->mddev->bitmap, r10_bio->sector,
-				r10_bio->sectors,
-				!test_bit(R10BIO_Degraded, &r10_bio->state),
-				0);
-		md_write_end(r10_bio->mddev);
-		if (test_bit(R10BIO_MadeGood, &r10_bio->state))
+		if (test_bit(R10BIO_WriteError, &r10_bio->state))
 			reschedule_retry(r10_bio);
-		else
-			raid_end_bio_io(r10_bio);
+		else {
+			close_write(r10_bio);
+			if (test_bit(R10BIO_MadeGood, &r10_bio->state))
+				reschedule_retry(r10_bio);
+			else
+				raid_end_bio_io(r10_bio);
+		}
 	}
 	if (dec_rdev)
 		rdev_dec_pending(conf->mirrors[dev].rdev, conf->mddev);
@@ -1839,6 +1848,82 @@ static void fix_read_error(conf_t *conf, mddev_t *mddev, r10bio_t *r10_bio)
 	}
 }
 
+static void bi_complete(struct bio *bio, int error)
+{
+	complete((struct completion *)bio->bi_private);
+}
+
+static int submit_bio_wait(int rw, struct bio *bio)
+{
+	struct completion event;
+	rw |= REQ_SYNC;
+
+	init_completion(&event);
+	bio->bi_private = &event;
+	bio->bi_end_io = bi_complete;
+	submit_bio(rw, bio);
+	wait_for_completion(&event);
+
+	return test_bit(BIO_UPTODATE, &bio->bi_flags);
+}
+
+static int narrow_write_error(r10bio_t *r10_bio, int i)
+{
+	struct bio *bio = r10_bio->master_bio;
+	mddev_t *mddev = r10_bio->mddev;
+	conf_t *conf = mddev->private;
+	mdk_rdev_t *rdev = conf->mirrors[r10_bio->devs[i].devnum].rdev;
+	/* bio has the data to be written to slot 'i' where
+	 * we just recently had a write error.
+	 * We repeatedly clone the bio and trim down to one block,
+	 * then try the write.  Where the write fails we record
+	 * a bad block.
+	 * It is conceivable that the bio doesn't exactly align with
+	 * blocks.  We must handle this.
+	 *
+	 * We currently own a reference to the rdev.
+	 */
+
+	int block_sectors;
+	sector_t sector;
+	int sectors;
+	int sect_to_write = r10_bio->sectors;
+	int ok = 1;
+
+	if (rdev->badblocks.shift < 0)
+		return 0;
+
+	block_sectors = 1 << rdev->badblocks.shift;
+	sector = r10_bio->sector;
+	sectors = ((r10_bio->sector + block_sectors)
+		   & ~(sector_t)(block_sectors - 1))
+		- sector;
+
+	while (sect_to_write) {
+		struct bio *wbio;
+		if (sectors > sect_to_write)
+			sectors = sect_to_write;
+		/* Write at 'sector' for 'sectors' */
+		wbio = bio_clone_mddev(bio, GFP_NOIO, mddev);
+		md_trim_bio(wbio, sector - bio->bi_sector, sectors);
+		wbio->bi_sector = (r10_bio->devs[i].addr+
+				   rdev->data_offset+
+				   (sector - r10_bio->sector));
+		wbio->bi_bdev = rdev->bdev;
+		if (submit_bio_wait(WRITE, wbio) == 0)
+			/* Failure! */
+			ok = rdev_set_badblocks(rdev, sector,
+						sectors, 0)
+				&& ok;
+
+		bio_put(wbio);
+		sect_to_write -= sectors;
+		sector += sectors;
+		sectors = block_sectors;
+	}
+	return ok;
+}
+
 static void handle_read_error(mddev_t *mddev, r10bio_t *r10_bio)
 {
 	int slot = r10_bio->read_slot;
@@ -1962,16 +2047,29 @@ static void handle_write_completed(conf_t *conf, r10bio_t *r10_bio)
 			}
 		put_buf(r10_bio);
 	} else {
-		for (m = 0; m < conf->copies; m++)
-			if (r10_bio->devs[m].bio == IO_MADE_GOOD) {
-				int dev = r10_bio->devs[m].devnum;
-				rdev = conf->mirrors[dev].rdev;
+		for (m = 0; m < conf->copies; m++) {
+			int dev = r10_bio->devs[m].devnum;
+			struct bio *bio = r10_bio->devs[m].bio;
+			rdev = conf->mirrors[dev].rdev;
+			if (bio == IO_MADE_GOOD) {
 				rdev_clear_badblocks(
 					rdev,
 					r10_bio->devs[m].addr,
 					r10_bio->sectors);
 				rdev_dec_pending(rdev, conf->mddev);
+			} else if (bio != NULL &&
+				   !test_bit(BIO_UPTODATE, &bio->bi_flags)) {
+				if (!narrow_write_error(r10_bio, m)) {
+					md_error(conf->mddev, rdev);
+					set_bit(R10BIO_Degraded,
+						&r10_bio->state);
+				}
+				rdev_dec_pending(rdev, conf->mddev);
 			}
+		}
+		if (test_bit(R10BIO_WriteError,
+			     &r10_bio->state))
+			close_write(r10_bio);
 		raid_end_bio_io(r10_bio);
 	}
 }
@@ -2003,7 +2101,8 @@ static void raid10d(mddev_t *mddev)
 
 		mddev = r10_bio->mddev;
 		conf = mddev->private;
-		if (test_bit(R10BIO_MadeGood, &r10_bio->state))
+		if (test_bit(R10BIO_MadeGood, &r10_bio->state) ||
+		    test_bit(R10BIO_WriteError, &r10_bio->state))
 			handle_write_completed(conf, r10_bio);
 		else if (test_bit(R10BIO_IsSync, &r10_bio->state))
 			sync_request_write(mddev, r10_bio);
diff --git a/drivers/md/raid10.h b/drivers/md/raid10.h
index d8b7f9af92d5..79cb52a0d4a2 100644
--- a/drivers/md/raid10.h
+++ b/drivers/md/raid10.h
@@ -139,4 +139,5 @@ struct r10bio_s {
  * known-bad-block records, we set this flag.
  */
 #define	R10BIO_MadeGood 5
+#define	R10BIO_WriteError 6
 #endif