Merge branch 'md-next' into md-linus

drivers/md/bitmap.c

@@ -27,6 +27,7 @@
 #include <linux/mount.h>
 #include <linux/buffer_head.h>
 #include <linux/seq_file.h>
+#include <trace/events/block.h>
 #include "md.h"
 #include "bitmap.h"
 
@@ -208,11 +209,13 @@ static struct md_rdev *next_active_rdev(struct md_rdev *rdev, struct mddev *mdde
 
 static int write_sb_page(struct bitmap *bitmap, struct page *page, int wait)
 {
-	struct md_rdev *rdev = NULL;
+	struct md_rdev *rdev;
 	struct block_device *bdev;
 	struct mddev *mddev = bitmap->mddev;
 	struct bitmap_storage *store = &bitmap->storage;
 
+restart:
+	rdev = NULL;
 	while ((rdev = next_active_rdev(rdev, mddev)) != NULL) {
 		int size = PAGE_SIZE;
 		loff_t offset = mddev->bitmap_info.offset;
@@ -268,8 +271,8 @@ static int write_sb_page(struct bitmap *bitmap, struct page *page, int wait)
 			       page);
 	}
 
-	if (wait)
-		md_super_wait(mddev);
+	if (wait && md_super_wait(mddev) < 0)
+		goto restart;
 	return 0;
 
  bad_alignment:
@@ -405,10 +408,10 @@ static int read_page(struct file *file, unsigned long index,
 		ret = -EIO;
 out:
 	if (ret)
-		printk(KERN_ALERT "md: bitmap read error: (%dB @ %llu): %d\n",
-			(int)PAGE_SIZE,
-			(unsigned long long)index << PAGE_SHIFT,
-			ret);
+		pr_err("md: bitmap read error: (%dB @ %llu): %d\n",
+		       (int)PAGE_SIZE,
+		       (unsigned long long)index << PAGE_SHIFT,
+		       ret);
 	return ret;
 }
 
@@ -416,6 +419,28 @@ out:
  * bitmap file superblock operations
  */
 
+/*
+ * bitmap_wait_writes() should be called before writing any bitmap
+ * blocks, to ensure previous writes, particularly from
+ * bitmap_daemon_work(), have completed.
+ */
+static void bitmap_wait_writes(struct bitmap *bitmap)
+{
+	if (bitmap->storage.file)
+		wait_event(bitmap->write_wait,
+			   atomic_read(&bitmap->pending_writes)==0);
+	else
+		/* Note that we ignore the return value.  The writes
+		 * might have failed, but that would just mean that
+		 * some bits which should be cleared haven't been,
+		 * which is safe.  The relevant bitmap blocks will
+		 * probably get written again, but there is no great
+		 * loss if they aren't.
+		 */
+		md_super_wait(bitmap->mddev);
+}
+
+
 /* update the event counter and sync the superblock to disk */
 void bitmap_update_sb(struct bitmap *bitmap)
 {
@@ -455,24 +480,24 @@ void bitmap_print_sb(struct bitmap *bitmap)
 	if (!bitmap || !bitmap->storage.sb_page)
 		return;
 	sb = kmap_atomic(bitmap->storage.sb_page);
-	printk(KERN_DEBUG "%s: bitmap file superblock:\n", bmname(bitmap));
-	printk(KERN_DEBUG "         magic: %08x\n", le32_to_cpu(sb->magic));
-	printk(KERN_DEBUG "       version: %d\n", le32_to_cpu(sb->version));
-	printk(KERN_DEBUG "          uuid: %08x.%08x.%08x.%08x\n",
-					*(__u32 *)(sb->uuid+0),
-					*(__u32 *)(sb->uuid+4),
-					*(__u32 *)(sb->uuid+8),
-					*(__u32 *)(sb->uuid+12));
-	printk(KERN_DEBUG "        events: %llu\n",
-			(unsigned long long) le64_to_cpu(sb->events));
-	printk(KERN_DEBUG "events cleared: %llu\n",
-			(unsigned long long) le64_to_cpu(sb->events_cleared));
-	printk(KERN_DEBUG "         state: %08x\n", le32_to_cpu(sb->state));
-	printk(KERN_DEBUG "     chunksize: %d B\n", le32_to_cpu(sb->chunksize));
-	printk(KERN_DEBUG "  daemon sleep: %ds\n", le32_to_cpu(sb->daemon_sleep));
-	printk(KERN_DEBUG "     sync size: %llu KB\n",
-			(unsigned long long)le64_to_cpu(sb->sync_size)/2);
-	printk(KERN_DEBUG "max write behind: %d\n", le32_to_cpu(sb->write_behind));
+	pr_debug("%s: bitmap file superblock:\n", bmname(bitmap));
+	pr_debug("         magic: %08x\n", le32_to_cpu(sb->magic));
+	pr_debug("       version: %d\n", le32_to_cpu(sb->version));
+	pr_debug("          uuid: %08x.%08x.%08x.%08x\n",
+		 *(__u32 *)(sb->uuid+0),
+		 *(__u32 *)(sb->uuid+4),
+		 *(__u32 *)(sb->uuid+8),
+		 *(__u32 *)(sb->uuid+12));
+	pr_debug("        events: %llu\n",
+		 (unsigned long long) le64_to_cpu(sb->events));
+	pr_debug("events cleared: %llu\n",
+		 (unsigned long long) le64_to_cpu(sb->events_cleared));
+	pr_debug("         state: %08x\n", le32_to_cpu(sb->state));
+	pr_debug("     chunksize: %d B\n", le32_to_cpu(sb->chunksize));
+	pr_debug("  daemon sleep: %ds\n", le32_to_cpu(sb->daemon_sleep));
+	pr_debug("     sync size: %llu KB\n",
+		 (unsigned long long)le64_to_cpu(sb->sync_size)/2);
+	pr_debug("max write behind: %d\n", le32_to_cpu(sb->write_behind));
 	kunmap_atomic(sb);
 }
 
@@ -506,14 +531,14 @@ static int bitmap_new_disk_sb(struct bitmap *bitmap)
 	BUG_ON(!chunksize);
 	if (!is_power_of_2(chunksize)) {
 		kunmap_atomic(sb);
-		printk(KERN_ERR "bitmap chunksize not a power of 2\n");
+		pr_warn("bitmap chunksize not a power of 2\n");
 		return -EINVAL;
 	}
 	sb->chunksize = cpu_to_le32(chunksize);
 
 	daemon_sleep = bitmap->mddev->bitmap_info.daemon_sleep;
 	if (!daemon_sleep || (daemon_sleep > MAX_SCHEDULE_TIMEOUT)) {
-		printk(KERN_INFO "Choosing daemon_sleep default (5 sec)\n");
+		pr_debug("Choosing daemon_sleep default (5 sec)\n");
 		daemon_sleep = 5 * HZ;
 	}
 	sb->daemon_sleep = cpu_to_le32(daemon_sleep);
@@ -584,7 +609,7 @@ re_read:
 		/* to 4k blocks */
 		bm_blocks = DIV_ROUND_UP_SECTOR_T(bm_blocks, 4096);
 		offset = bitmap->mddev->bitmap_info.offset + (bitmap->cluster_slot * (bm_blocks << 3));
-		pr_info("%s:%d bm slot: %d offset: %llu\n", __func__, __LINE__,
+		pr_debug("%s:%d bm slot: %d offset: %llu\n", __func__, __LINE__,
 			bitmap->cluster_slot, offset);
 	}
 
@@ -634,7 +659,7 @@ re_read:
 	else if (write_behind > COUNTER_MAX)
 		reason = "write-behind limit out of range (0 - 16383)";
 	if (reason) {
-		printk(KERN_INFO "%s: invalid bitmap file superblock: %s\n",
+		pr_warn("%s: invalid bitmap file superblock: %s\n",
 			bmname(bitmap), reason);
 		goto out;
 	}
@@ -648,18 +673,15 @@ re_read:
 		 * bitmap's UUID and event counter to the mddev's
 		 */
 		if (memcmp(sb->uuid, bitmap->mddev->uuid, 16)) {
-			printk(KERN_INFO
-			       "%s: bitmap superblock UUID mismatch\n",
-			       bmname(bitmap));
+			pr_warn("%s: bitmap superblock UUID mismatch\n",
+				bmname(bitmap));
 			goto out;
 		}
 		events = le64_to_cpu(sb->events);
 		if (!nodes && (events < bitmap->mddev->events)) {
-			printk(KERN_INFO
-			       "%s: bitmap file is out of date (%llu < %llu) "
-			       "-- forcing full recovery\n",
-			       bmname(bitmap), events,
-			       (unsigned long long) bitmap->mddev->events);
+			pr_warn("%s: bitmap file is out of date (%llu < %llu) -- forcing full recovery\n",
+				bmname(bitmap), events,
+				(unsigned long long) bitmap->mddev->events);
 			set_bit(BITMAP_STALE, &bitmap->flags);
 		}
 	}
@@ -679,8 +701,8 @@ out:
 	if (err == 0 && nodes && (bitmap->cluster_slot < 0)) {
 		err = md_setup_cluster(bitmap->mddev, nodes);
 		if (err) {
-			pr_err("%s: Could not setup cluster service (%d)\n",
-					bmname(bitmap), err);
+			pr_warn("%s: Could not setup cluster service (%d)\n",
+				bmname(bitmap), err);
 			goto out_no_sb;
 		}
 		bitmap->cluster_slot = md_cluster_ops->slot_number(bitmap->mddev);
@@ -847,15 +869,13 @@ static void bitmap_file_kick(struct bitmap *bitmap)
 				ptr = file_path(bitmap->storage.file,
 					     path, PAGE_SIZE);
 
-			printk(KERN_ALERT
-			      "%s: kicking failed bitmap file %s from array!\n",
-			      bmname(bitmap), IS_ERR(ptr) ? "" : ptr);
+			pr_warn("%s: kicking failed bitmap file %s from array!\n",
+				bmname(bitmap), IS_ERR(ptr) ? "" : ptr);
 
 			kfree(path);
 		} else
-			printk(KERN_ALERT
-			       "%s: disabling internal bitmap due to errors\n",
-			       bmname(bitmap));
+			pr_warn("%s: disabling internal bitmap due to errors\n",
+				bmname(bitmap));
 	}
 }
 
@@ -983,6 +1003,7 @@ void bitmap_unplug(struct bitmap *bitmap)
 {
 	unsigned long i;
 	int dirty, need_write;
+	int writing = 0;
 
 	if (!bitmap || !bitmap->storage.filemap ||
 	    test_bit(BITMAP_STALE, &bitmap->flags))
@@ -997,15 +1018,19 @@ void bitmap_unplug(struct bitmap *bitmap)
 		need_write = test_and_clear_page_attr(bitmap, i,
 						      BITMAP_PAGE_NEEDWRITE);
 		if (dirty || need_write) {
+			if (!writing) {
+				bitmap_wait_writes(bitmap);
+				if (bitmap->mddev->queue)
+					blk_add_trace_msg(bitmap->mddev->queue,
+							  "md bitmap_unplug");
+			}
 			clear_page_attr(bitmap, i, BITMAP_PAGE_PENDING);
 			write_page(bitmap, bitmap->storage.filemap[i], 0);
+			writing = 1;
 		}
 	}
-	if (bitmap->storage.file)
-		wait_event(bitmap->write_wait,
-			   atomic_read(&bitmap->pending_writes)==0);
-	else
-		md_super_wait(bitmap->mddev);
+	if (writing)
+		bitmap_wait_writes(bitmap);
 
 	if (test_bit(BITMAP_WRITE_ERROR, &bitmap->flags))
 		bitmap_file_kick(bitmap);
@@ -1056,14 +1081,13 @@ static int bitmap_init_from_disk(struct bitmap *bitmap, sector_t start)
 
 	outofdate = test_bit(BITMAP_STALE, &bitmap->flags);
 	if (outofdate)
-		printk(KERN_INFO "%s: bitmap file is out of date, doing full "
-			"recovery\n", bmname(bitmap));
+		pr_warn("%s: bitmap file is out of date, doing full recovery\n", bmname(bitmap));
 
 	if (file && i_size_read(file->f_mapping->host) < store->bytes) {
-		printk(KERN_INFO "%s: bitmap file too short %lu < %lu\n",
-		       bmname(bitmap),
-		       (unsigned long) i_size_read(file->f_mapping->host),
-		       store->bytes);
+		pr_warn("%s: bitmap file too short %lu < %lu\n",
+			bmname(bitmap),
+			(unsigned long) i_size_read(file->f_mapping->host),
+			store->bytes);
 		goto err;
 	}
 
@@ -1137,16 +1161,15 @@ static int bitmap_init_from_disk(struct bitmap *bitmap, sector_t start)
 		offset = 0;
 	}
 
-	printk(KERN_INFO "%s: bitmap initialized from disk: "
-	       "read %lu pages, set %lu of %lu bits\n",
-	       bmname(bitmap), store->file_pages,
-	       bit_cnt, chunks);
+	pr_debug("%s: bitmap initialized from disk: read %lu pages, set %lu of %lu bits\n",
+		 bmname(bitmap), store->file_pages,
+		 bit_cnt, chunks);
 
 	return 0;
 
  err:
-	printk(KERN_INFO "%s: bitmap initialisation failed: %d\n",
-	       bmname(bitmap), ret);
+	pr_warn("%s: bitmap initialisation failed: %d\n",
+		bmname(bitmap), ret);
 	return ret;
 }
 
@@ -1225,6 +1248,10 @@ void bitmap_daemon_work(struct mddev *mddev)
 	}
 	bitmap->allclean = 1;
 
+	if (bitmap->mddev->queue)
+		blk_add_trace_msg(bitmap->mddev->queue,
+				  "md bitmap_daemon_work");
+
 	/* Any file-page which is PENDING now needs to be written.
 	 * So set NEEDWRITE now, then after we make any last-minute changes
 	 * we will write it.
@@ -1289,6 +1316,7 @@ void bitmap_daemon_work(struct mddev *mddev)
 	}
 	spin_unlock_irq(&counts->lock);
 
+	bitmap_wait_writes(bitmap);
 	/* Now start writeout on any page in NEEDWRITE that isn't DIRTY.
 	 * DIRTY pages need to be written by bitmap_unplug so it can wait
 	 * for them.
@@ -1595,7 +1623,7 @@ void bitmap_cond_end_sync(struct bitmap *bitmap, sector_t sector, bool force)
 		   atomic_read(&bitmap->mddev->recovery_active) == 0);
 
 	bitmap->mddev->curr_resync_completed = sector;
-	set_bit(MD_CHANGE_CLEAN, &bitmap->mddev->flags);
+	set_bit(MD_SB_CHANGE_CLEAN, &bitmap->mddev->sb_flags);
 	sector &= ~((1ULL << bitmap->counts.chunkshift) - 1);
 	s = 0;
 	while (s < sector && s < bitmap->mddev->resync_max_sectors) {
@@ -1825,8 +1853,8 @@ struct bitmap *bitmap_create(struct mddev *mddev, int slot)
 	if (err)
 		goto error;
 
-	printk(KERN_INFO "created bitmap (%lu pages) for device %s\n",
-	       bitmap->counts.pages, bmname(bitmap));
+	pr_debug("created bitmap (%lu pages) for device %s\n",
+		 bitmap->counts.pages, bmname(bitmap));
 
 	err = test_bit(BITMAP_WRITE_ERROR, &bitmap->flags) ? -EIO : 0;
 	if (err)
@@ -2029,8 +2057,10 @@ int bitmap_resize(struct bitmap *bitmap, sector_t blocks,
 					   !bitmap->mddev->bitmap_info.external,
 					   mddev_is_clustered(bitmap->mddev)
 					   ? bitmap->cluster_slot : 0);
-	if (ret)
+	if (ret) {
+		bitmap_file_unmap(&store);
 		goto err;
+	}
 
 	pages = DIV_ROUND_UP(chunks, PAGE_COUNTER_RATIO);
 
@@ -2089,7 +2119,7 @@ int bitmap_resize(struct bitmap *bitmap, sector_t blocks,
 				bitmap->mddev->bitmap_info.chunksize = 1 << (old_counts.chunkshift +
 									     BITMAP_BLOCK_SHIFT);
 				blocks = old_counts.chunks << old_counts.chunkshift;
-				pr_err("Could not pre-allocate in-memory bitmap for cluster raid\n");
+				pr_warn("Could not pre-allocate in-memory bitmap for cluster raid\n");
 				break;
 			} else
 				bitmap->counts.bp[page].count += 1;
@@ -2266,7 +2296,7 @@ location_store(struct mddev *mddev, const char *buf, size_t len)
 		/* Ensure new bitmap info is stored in
 		 * metadata promptly.
 		 */
-		set_bit(MD_CHANGE_DEVS, &mddev->flags);
+		set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
 		md_wakeup_thread(mddev->thread);
 	}
 	rv = 0;

drivers/md/dm-raid.c

@@ -2011,7 +2011,7 @@ static int super_load(struct md_rdev *rdev, struct md_rdev *refdev)
 		sb->compat_features = cpu_to_le32(FEATURE_FLAG_SUPPORTS_V190);
 
 		/* Force writing of superblocks to disk */
-		set_bit(MD_CHANGE_DEVS, &rdev->mddev->flags);
+		set_bit(MD_SB_CHANGE_DEVS, &rdev->mddev->sb_flags);
 
 		/* Any superblock is better than none, choose that if given */
 		return refdev ? 0 : 1;
@@ -3497,7 +3497,7 @@ static void rs_update_sbs(struct raid_set *rs)
 	struct mddev *mddev = &rs->md;
 	int ro = mddev->ro;
 
-	set_bit(MD_CHANGE_DEVS, &mddev->flags);
+	set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
 	mddev->ro = 0;
 	md_update_sb(mddev, 1);
 	mddev->ro = ro;

drivers/md/linear.c

@@ -21,6 +21,7 @@
 #include <linux/seq_file.h>
 #include <linux/module.h>
 #include <linux/slab.h>
+#include <trace/events/block.h>
 #include "md.h"
 #include "linear.h"
 
@@ -101,8 +102,8 @@ static struct linear_conf *linear_conf(struct mddev *mddev, int raid_disks)
 		sector_t sectors;
 
 		if (j < 0 || j >= raid_disks || disk->rdev) {
-			printk(KERN_ERR "md/linear:%s: disk numbering problem. Aborting!\n",
-			       mdname(mddev));
+			pr_warn("md/linear:%s: disk numbering problem. Aborting!\n",
+				mdname(mddev));
 			goto out;
 		}
 
@@ -123,8 +124,8 @@ static struct linear_conf *linear_conf(struct mddev *mddev, int raid_disks)
 			discard_supported = true;
 	}
 	if (cnt != raid_disks) {
-		printk(KERN_ERR "md/linear:%s: not enough drives present. Aborting!\n",
-		       mdname(mddev));
+		pr_warn("md/linear:%s: not enough drives present. Aborting!\n",
+			mdname(mddev));
 		goto out;
 	}
 
@@ -227,22 +228,22 @@ static void linear_make_request(struct mddev *mddev, struct bio *bio)
 	}
 
 	do {
-		tmp_dev = which_dev(mddev, bio->bi_iter.bi_sector);
+		sector_t bio_sector = bio->bi_iter.bi_sector;
+		tmp_dev = which_dev(mddev, bio_sector);
 		start_sector = tmp_dev->end_sector - tmp_dev->rdev->sectors;
 		end_sector = tmp_dev->end_sector;
 		data_offset = tmp_dev->rdev->data_offset;
 		bio->bi_bdev = tmp_dev->rdev->bdev;
 
-		if (unlikely(bio->bi_iter.bi_sector >= end_sector ||
-			     bio->bi_iter.bi_sector < start_sector))
+		if (unlikely(bio_sector >= end_sector ||
+			     bio_sector < start_sector))
 			goto out_of_bounds;
 
 		if (unlikely(bio_end_sector(bio) > end_sector)) {
 			/* This bio crosses a device boundary, so we have to
 			 * split it.
 			 */
-			split = bio_split(bio, end_sector -
-					  bio->bi_iter.bi_sector,
+			split = bio_split(bio, end_sector - bio_sector,
 					  GFP_NOIO, fs_bio_set);
 			bio_chain(split, bio);
 		} else {
@@ -256,15 +257,18 @@ static void linear_make_request(struct mddev *mddev, struct bio *bio)
 			 !blk_queue_discard(bdev_get_queue(split->bi_bdev)))) {
 			/* Just ignore it */
 			bio_endio(split);
-		} else
+		} else {
+			if (mddev->gendisk)
+				trace_block_bio_remap(bdev_get_queue(split->bi_bdev),
+						      split, disk_devt(mddev->gendisk),
+						      bio_sector);
 			generic_make_request(split);
+		}
 	} while (split != bio);
 	return;
 
 out_of_bounds:
-	printk(KERN_ERR
-	       "md/linear:%s: make_request: Sector %llu out of bounds on "
-	       "dev %s: %llu sectors, offset %llu\n",
+	pr_err("md/linear:%s: make_request: Sector %llu out of bounds on dev %s: %llu sectors, offset %llu\n",
 	       mdname(mddev),
 	       (unsigned long long)bio->bi_iter.bi_sector,
 	       bdevname(tmp_dev->rdev->bdev, b),
@@ -275,7 +279,6 @@ out_of_bounds:
 
 static void linear_status (struct seq_file *seq, struct mddev *mddev)
 {
-
 	seq_printf(seq, " %dk rounding", mddev->chunk_sectors / 2);
 }
 

drivers/md/md.c

@@ -30,6 +30,18 @@
    You should have received a copy of the GNU General Public License
    (for example /usr/src/linux/COPYING); if not, write to the Free
    Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+
+   Errors, Warnings, etc.
+   Please use:
+     pr_crit() for error conditions that risk data loss
+     pr_err() for error conditions that are unexpected, like an IO error
+         or internal inconsistency
+     pr_warn() for error conditions that could have been predicated, like
+         adding a device to an array when it has incompatible metadata
+     pr_info() for every interesting, very rare events, like an array starting
+         or stopping, or resync starting or stopping
+     pr_debug() for everything else.
+
 */
 
 #include <linux/kthread.h>
@@ -52,6 +64,7 @@
 #include <linux/raid/md_p.h>
 #include <linux/raid/md_u.h>
 #include <linux/slab.h>
+#include <trace/events/block.h>
 #include "md.h"
 #include "bitmap.h"
 #include "md-cluster.h"
@@ -684,11 +697,8 @@ static inline sector_t calc_dev_sboffset(struct md_rdev *rdev)
 static int alloc_disk_sb(struct md_rdev *rdev)
 {
 	rdev->sb_page = alloc_page(GFP_KERNEL);
-	if (!rdev->sb_page) {
-		printk(KERN_ALERT "md: out of memory.\n");
+	if (!rdev->sb_page)
 		return -ENOMEM;
-	}
-
 	return 0;
 }
 
@@ -715,9 +725,15 @@ static void super_written(struct bio *bio)
 	struct mddev *mddev = rdev->mddev;
 
 	if (bio->bi_error) {
-		printk("md: super_written gets error=%d\n", bio->bi_error);
+		pr_err("md: super_written gets error=%d\n", bio->bi_error);
 		md_error(mddev, rdev);
-	}
+		if (!test_bit(Faulty, &rdev->flags)
+		    && (bio->bi_opf & MD_FAILFAST)) {
+			set_bit(MD_SB_NEED_REWRITE, &mddev->sb_flags);
+			set_bit(LastDev, &rdev->flags);
+		}
+	} else
+		clear_bit(LastDev, &rdev->flags);
 
 	if (atomic_dec_and_test(&mddev->pending_writes))
 		wake_up(&mddev->sb_wait);
@@ -734,7 +750,13 @@ void md_super_write(struct mddev *mddev, struct md_rdev *rdev,
 	 * if zero is reached.
 	 * If an error occurred, call md_error
 	 */
-	struct bio *bio = bio_alloc_mddev(GFP_NOIO, 1, mddev);
+	struct bio *bio;
+	int ff = 0;
+
+	if (test_bit(Faulty, &rdev->flags))
+		return;
+
+	bio = bio_alloc_mddev(GFP_NOIO, 1, mddev);
 
 	atomic_inc(&rdev->nr_pending);
 
@@ -743,16 +765,24 @@ void md_super_write(struct mddev *mddev, struct md_rdev *rdev,
 	bio_add_page(bio, page, size, 0);
 	bio->bi_private = rdev;
 	bio->bi_end_io = super_written;
-	bio->bi_opf = REQ_OP_WRITE | REQ_PREFLUSH | REQ_FUA;
+
+	if (test_bit(MD_FAILFAST_SUPPORTED, &mddev->flags) &&
+	    test_bit(FailFast, &rdev->flags) &&
+	    !test_bit(LastDev, &rdev->flags))
+		ff = MD_FAILFAST;
+	bio->bi_opf = REQ_OP_WRITE | REQ_PREFLUSH | REQ_FUA | ff;
 
 	atomic_inc(&mddev->pending_writes);
 	submit_bio(bio);
 }
 
-void md_super_wait(struct mddev *mddev)
+int md_super_wait(struct mddev *mddev)
 {
 	/* wait for all superblock writes that were scheduled to complete */
 	wait_event(mddev->sb_wait, atomic_read(&mddev->pending_writes)==0);
+	if (test_and_clear_bit(MD_SB_NEED_REWRITE, &mddev->sb_flags))
+		return -EAGAIN;
+	return 0;
 }
 
 int sync_page_io(struct md_rdev *rdev, sector_t sector, int size,
@@ -795,8 +825,8 @@ static int read_disk_sb(struct md_rdev *rdev, int size)
 	return 0;
 
 fail:
-	printk(KERN_WARNING "md: disabled device %s, could not read superblock.\n",
-		bdevname(rdev->bdev,b));
+	pr_err("md: disabled device %s, could not read superblock.\n",
+	       bdevname(rdev->bdev,b));
 	return -EINVAL;
 }
 
@@ -818,7 +848,6 @@ static int sb_equal(mdp_super_t *sb1, mdp_super_t *sb2)
 
 	if (!tmp1 || !tmp2) {
 		ret = 0;
-		printk(KERN_INFO "md.c sb_equal(): failed to allocate memory!\n");
 		goto abort;
 	}
 
@@ -932,7 +961,7 @@ int md_check_no_bitmap(struct mddev *mddev)
 {
 	if (!mddev->bitmap_info.file && !mddev->bitmap_info.offset)
 		return 0;
-	printk(KERN_ERR "%s: bitmaps are not supported for %s\n",
+	pr_warn("%s: bitmaps are not supported for %s\n",
 		mdname(mddev), mddev->pers->name);
 	return 1;
 }
@@ -956,7 +985,8 @@ static int super_90_load(struct md_rdev *rdev, struct md_rdev *refdev, int minor
 	rdev->sb_start = calc_dev_sboffset(rdev);
 
 	ret = read_disk_sb(rdev, MD_SB_BYTES);
-	if (ret) return ret;
+	if (ret)
+		return ret;
 
 	ret = -EINVAL;
 
@@ -964,17 +994,15 @@ static int super_90_load(struct md_rdev *rdev, struct md_rdev *refdev, int minor
 	sb = page_address(rdev->sb_page);
 
 	if (sb->md_magic != MD_SB_MAGIC) {
-		printk(KERN_ERR "md: invalid raid superblock magic on %s\n",
-		       b);
+		pr_warn("md: invalid raid superblock magic on %s\n", b);
 		goto abort;
 	}
 
 	if (sb->major_version != 0 ||
 	    sb->minor_version < 90 ||
 	    sb->minor_version > 91) {
-		printk(KERN_WARNING "Bad version number %d.%d on %s\n",
-			sb->major_version, sb->minor_version,
-			b);
+		pr_warn("Bad version number %d.%d on %s\n",
+			sb->major_version, sb->minor_version, b);
 		goto abort;
 	}
 
@@ -982,8 +1010,7 @@ static int super_90_load(struct md_rdev *rdev, struct md_rdev *refdev, int minor
 		goto abort;
 
 	if (md_csum_fold(calc_sb_csum(sb)) != md_csum_fold(sb->sb_csum)) {
-		printk(KERN_WARNING "md: invalid superblock checksum on %s\n",
-			b);
+		pr_warn("md: invalid superblock checksum on %s\n", b);
 		goto abort;
 	}
 
@@ -1004,14 +1031,13 @@ static int super_90_load(struct md_rdev *rdev, struct md_rdev *refdev, int minor
 		__u64 ev1, ev2;
 		mdp_super_t *refsb = page_address(refdev->sb_page);
 		if (!uuid_equal(refsb, sb)) {
-			printk(KERN_WARNING "md: %s has different UUID to %s\n",
+			pr_warn("md: %s has different UUID to %s\n",
 				b, bdevname(refdev->bdev,b2));
 			goto abort;
 		}
 		if (!sb_equal(refsb, sb)) {
-			printk(KERN_WARNING "md: %s has same UUID"
-			       " but different superblock to %s\n",
-			       b, bdevname(refdev->bdev, b2));
+			pr_warn("md: %s has same UUID but different superblock to %s\n",
+				b, bdevname(refdev->bdev, b2));
 			goto abort;
 		}
 		ev1 = md_event(sb);
@@ -1158,6 +1184,8 @@ static int super_90_validate(struct mddev *mddev, struct md_rdev *rdev)
 		}
 		if (desc->state & (1<<MD_DISK_WRITEMOSTLY))
 			set_bit(WriteMostly, &rdev->flags);
+		if (desc->state & (1<<MD_DISK_FAILFAST))
+			set_bit(FailFast, &rdev->flags);
 	} else /* MULTIPATH are always insync */
 		set_bit(In_sync, &rdev->flags);
 	return 0;
@@ -1283,6 +1311,8 @@ static void super_90_sync(struct mddev *mddev, struct md_rdev *rdev)
 		}
 		if (test_bit(WriteMostly, &rdev2->flags))
 			d->state |= (1<<MD_DISK_WRITEMOSTLY);
+		if (test_bit(FailFast, &rdev2->flags))
+			d->state |= (1<<MD_DISK_FAILFAST);
 	}
 	/* now set the "removed" and "faulty" bits on any missing devices */
 	for (i=0 ; i < mddev->raid_disks ; i++) {
@@ -1324,9 +1354,10 @@ super_90_rdev_size_change(struct md_rdev *rdev, sector_t num_sectors)
 	if (IS_ENABLED(CONFIG_LBDAF) && (u64)num_sectors >= (2ULL << 32) &&
 	    rdev->mddev->level >= 1)
 		num_sectors = (sector_t)(2ULL << 32) - 2;
-	md_super_write(rdev->mddev, rdev, rdev->sb_start, rdev->sb_size,
+	do {
+		md_super_write(rdev->mddev, rdev, rdev->sb_start, rdev->sb_size,
 		       rdev->sb_page);
-	md_super_wait(rdev->mddev);
+	} while (md_super_wait(rdev->mddev) < 0);
 	return num_sectors;
 }
 
@@ -1413,13 +1444,13 @@ static int super_1_load(struct md_rdev *rdev, struct md_rdev *refdev, int minor_
 		return -EINVAL;
 
 	if (calc_sb_1_csum(sb) != sb->sb_csum) {
-		printk("md: invalid superblock checksum on %s\n",
+		pr_warn("md: invalid superblock checksum on %s\n",
 			bdevname(rdev->bdev,b));
 		return -EINVAL;
 	}
 	if (le64_to_cpu(sb->data_size) < 10) {
-		printk("md: data_size too small on %s\n",
-		       bdevname(rdev->bdev,b));
+		pr_warn("md: data_size too small on %s\n",
+			bdevname(rdev->bdev,b));
 		return -EINVAL;
 	}
 	if (sb->pad0 ||
@@ -1503,8 +1534,7 @@ static int super_1_load(struct md_rdev *rdev, struct md_rdev *refdev, int minor_
 		    sb->level != refsb->level ||
 		    sb->layout != refsb->layout ||
 		    sb->chunksize != refsb->chunksize) {
-			printk(KERN_WARNING "md: %s has strangely different"
-				" superblock to %s\n",
+			pr_warn("md: %s has strangely different superblock to %s\n",
 				bdevname(rdev->bdev,b),
 				bdevname(refdev->bdev,b2));
 			return -EINVAL;
@@ -1646,8 +1676,7 @@ static int super_1_validate(struct mddev *mddev, struct md_rdev *rdev)
 		case MD_DISK_ROLE_JOURNAL: /* journal device */
 			if (!(le32_to_cpu(sb->feature_map) & MD_FEATURE_JOURNAL)) {
 				/* journal device without journal feature */
-				printk(KERN_WARNING
-				  "md: journal device provided without journal feature, ignoring the device\n");
+				pr_warn("md: journal device provided without journal feature, ignoring the device\n");
 				return -EINVAL;
 			}
 			set_bit(Journal, &rdev->flags);
@@ -1669,6 +1698,8 @@ static int super_1_validate(struct mddev *mddev, struct md_rdev *rdev)
 		}
 		if (sb->devflags & WriteMostly1)
 			set_bit(WriteMostly, &rdev->flags);
+		if (sb->devflags & FailFast1)
+			set_bit(FailFast, &rdev->flags);
 		if (le32_to_cpu(sb->feature_map) & MD_FEATURE_REPLACEMENT)
 			set_bit(Replacement, &rdev->flags);
 	} else /* MULTIPATH are always insync */
@@ -1707,6 +1738,10 @@ static void super_1_sync(struct mddev *mddev, struct md_rdev *rdev)
 	sb->chunksize = cpu_to_le32(mddev->chunk_sectors);
 	sb->level = cpu_to_le32(mddev->level);
 	sb->layout = cpu_to_le32(mddev->layout);
+	if (test_bit(FailFast, &rdev->flags))
+		sb->devflags |= FailFast1;
+	else
+		sb->devflags &= ~FailFast1;
 
 	if (test_bit(WriteMostly, &rdev->flags))
 		sb->devflags |= WriteMostly1;
@@ -1863,9 +1898,10 @@ super_1_rdev_size_change(struct md_rdev *rdev, sector_t num_sectors)
 	sb->data_size = cpu_to_le64(num_sectors);
 	sb->super_offset = rdev->sb_start;
 	sb->sb_csum = calc_sb_1_csum(sb);
-	md_super_write(rdev->mddev, rdev, rdev->sb_start, rdev->sb_size,
-		       rdev->sb_page);
-	md_super_wait(rdev->mddev);
+	do {
+		md_super_write(rdev->mddev, rdev, rdev->sb_start, rdev->sb_size,
+			       rdev->sb_page);
+	} while (md_super_wait(rdev->mddev) < 0);
 	return num_sectors;
 
 }
@@ -2004,9 +2040,9 @@ int md_integrity_register(struct mddev *mddev)
 	blk_integrity_register(mddev->gendisk,
 			       bdev_get_integrity(reference->bdev));
 
-	printk(KERN_NOTICE "md: data integrity enabled on %s\n", mdname(mddev));
+	pr_debug("md: data integrity enabled on %s\n", mdname(mddev));
 	if (bioset_integrity_create(mddev->bio_set, BIO_POOL_SIZE)) {
-		printk(KERN_ERR "md: failed to create integrity pool for %s\n",
+		pr_err("md: failed to create integrity pool for %s\n",
 		       mdname(mddev));
 		return -EINVAL;
 	}
@@ -2034,8 +2070,8 @@ int md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev)
 		return 0;
 
 	if (blk_integrity_compare(mddev->gendisk, rdev->bdev->bd_disk) != 0) {
-		printk(KERN_NOTICE "%s: incompatible integrity profile for %s\n",
-				mdname(mddev), bdevname(rdev->bdev, name));
+		pr_err("%s: incompatible integrity profile for %s\n",
+		       mdname(mddev), bdevname(rdev->bdev, name));
 		return -ENXIO;
 	}
 
@@ -2089,15 +2125,15 @@ static int bind_rdev_to_array(struct md_rdev *rdev, struct mddev *mddev)
 	rcu_read_unlock();
 	if (!test_bit(Journal, &rdev->flags) &&
 	    mddev->max_disks && rdev->desc_nr >= mddev->max_disks) {
-		printk(KERN_WARNING "md: %s: array is limited to %d devices\n",
-		       mdname(mddev), mddev->max_disks);
+		pr_warn("md: %s: array is limited to %d devices\n",
+			mdname(mddev), mddev->max_disks);
 		return -EBUSY;
 	}
 	bdevname(rdev->bdev,b);
 	strreplace(b, '/', '!');
 
 	rdev->mddev = mddev;
-	printk(KERN_INFO "md: bind<%s>\n", b);
+	pr_debug("md: bind<%s>\n", b);
 
 	if ((err = kobject_add(&rdev->kobj, &mddev->kobj, "dev-%s", b)))
 		goto fail;
@@ -2116,8 +2152,8 @@ static int bind_rdev_to_array(struct md_rdev *rdev, struct mddev *mddev)
 	return 0;
 
  fail:
-	printk(KERN_WARNING "md: failed to register dev-%s for %s\n",
-	       b, mdname(mddev));
+	pr_warn("md: failed to register dev-%s for %s\n",
+		b, mdname(mddev));
 	return err;
 }
 
@@ -2134,7 +2170,7 @@ static void unbind_rdev_from_array(struct md_rdev *rdev)
 
 	bd_unlink_disk_holder(rdev->bdev, rdev->mddev->gendisk);
 	list_del_rcu(&rdev->same_set);
-	printk(KERN_INFO "md: unbind<%s>\n", bdevname(rdev->bdev,b));
+	pr_debug("md: unbind<%s>\n", bdevname(rdev->bdev,b));
 	rdev->mddev = NULL;
 	sysfs_remove_link(&rdev->kobj, "block");
 	sysfs_put(rdev->sysfs_state);
@@ -2164,8 +2200,7 @@ static int lock_rdev(struct md_rdev *rdev, dev_t dev, int shared)
 	bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL,
 				 shared ? (struct md_rdev *)lock_rdev : rdev);
 	if (IS_ERR(bdev)) {
-		printk(KERN_ERR "md: could not open %s.\n",
-			__bdevname(dev, b));
+		pr_warn("md: could not open %s.\n", __bdevname(dev, b));
 		return PTR_ERR(bdev);
 	}
 	rdev->bdev = bdev;
@@ -2185,8 +2220,7 @@ static void export_rdev(struct md_rdev *rdev)
 {
 	char b[BDEVNAME_SIZE];
 
-	printk(KERN_INFO "md: export_rdev(%s)\n",
-		bdevname(rdev->bdev,b));
+	pr_debug("md: export_rdev(%s)\n", bdevname(rdev->bdev,b));
 	md_rdev_clear(rdev);
 #ifndef MODULE
 	if (test_bit(AutoDetected, &rdev->flags))
@@ -2288,24 +2322,24 @@ void md_update_sb(struct mddev *mddev, int force_change)
 
 	if (mddev->ro) {
 		if (force_change)
-			set_bit(MD_CHANGE_DEVS, &mddev->flags);
+			set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
 		return;
 	}
 
 repeat:
 	if (mddev_is_clustered(mddev)) {
-		if (test_and_clear_bit(MD_CHANGE_DEVS, &mddev->flags))
+		if (test_and_clear_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags))
 			force_change = 1;
-		if (test_and_clear_bit(MD_CHANGE_CLEAN, &mddev->flags))
+		if (test_and_clear_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags))
 			nospares = 1;
 		ret = md_cluster_ops->metadata_update_start(mddev);
 		/* Has someone else has updated the sb */
 		if (!does_sb_need_changing(mddev)) {
 			if (ret == 0)
 				md_cluster_ops->metadata_update_cancel(mddev);
-			bit_clear_unless(&mddev->flags, BIT(MD_CHANGE_PENDING),
-							 BIT(MD_CHANGE_DEVS) |
-							 BIT(MD_CHANGE_CLEAN));
+			bit_clear_unless(&mddev->sb_flags, BIT(MD_SB_CHANGE_PENDING),
+							 BIT(MD_SB_CHANGE_DEVS) |
+							 BIT(MD_SB_CHANGE_CLEAN));
 			return;
 		}
 	}
@@ -2321,10 +2355,10 @@ repeat:
 
 	}
 	if (!mddev->persistent) {
-		clear_bit(MD_CHANGE_CLEAN, &mddev->flags);
-		clear_bit(MD_CHANGE_DEVS, &mddev->flags);
+		clear_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags);
+		clear_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
 		if (!mddev->external) {
-			clear_bit(MD_CHANGE_PENDING, &mddev->flags);
+			clear_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags);
 			rdev_for_each(rdev, mddev) {
 				if (rdev->badblocks.changed) {
 					rdev->badblocks.changed = 0;
@@ -2344,9 +2378,9 @@ repeat:
 
 	mddev->utime = ktime_get_real_seconds();
 
-	if (test_and_clear_bit(MD_CHANGE_DEVS, &mddev->flags))
+	if (test_and_clear_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags))
 		force_change = 1;
-	if (test_and_clear_bit(MD_CHANGE_CLEAN, &mddev->flags))
+	if (test_and_clear_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags))
 		/* just a clean<-> dirty transition, possibly leave spares alone,
 		 * though if events isn't the right even/odd, we will have to do
 		 * spares after all
@@ -2402,6 +2436,9 @@ repeat:
 	pr_debug("md: updating %s RAID superblock on device (in sync %d)\n",
 		 mdname(mddev), mddev->in_sync);
 
+	if (mddev->queue)
+		blk_add_trace_msg(mddev->queue, "md md_update_sb");
+rewrite:
 	bitmap_update_sb(mddev->bitmap);
 	rdev_for_each(rdev, mddev) {
 		char b[BDEVNAME_SIZE];
@@ -2433,15 +2470,16 @@ repeat:
 			/* only need to write one superblock... */
 			break;
 	}
-	md_super_wait(mddev);
-	/* if there was a failure, MD_CHANGE_DEVS was set, and we re-write super */
+	if (md_super_wait(mddev) < 0)
+		goto rewrite;
+	/* if there was a failure, MD_SB_CHANGE_DEVS was set, and we re-write super */
 
 	if (mddev_is_clustered(mddev) && ret == 0)
 		md_cluster_ops->metadata_update_finish(mddev);
 
 	if (mddev->in_sync != sync_req ||
-	    !bit_clear_unless(&mddev->flags, BIT(MD_CHANGE_PENDING),
-			       BIT(MD_CHANGE_DEVS) | BIT(MD_CHANGE_CLEAN)))
+	    !bit_clear_unless(&mddev->sb_flags, BIT(MD_SB_CHANGE_PENDING),
+			       BIT(MD_SB_CHANGE_DEVS) | BIT(MD_SB_CHANGE_CLEAN)))
 		/* have to write it out again */
 		goto repeat;
 	wake_up(&mddev->sb_wait);
@@ -2485,7 +2523,7 @@ static int add_bound_rdev(struct md_rdev *rdev)
 	}
 	sysfs_notify_dirent_safe(rdev->sysfs_state);
 
-	set_bit(MD_CHANGE_DEVS, &mddev->flags);
+	set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
 	if (mddev->degraded)
 		set_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
 	set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
@@ -2523,51 +2561,41 @@ struct rdev_sysfs_entry {
 static ssize_t
 state_show(struct md_rdev *rdev, char *page)
 {
-	char *sep = "";
+	char *sep = ",";
 	size_t len = 0;
 	unsigned long flags = ACCESS_ONCE(rdev->flags);
 
 	if (test_bit(Faulty, &flags) ||
-	    rdev->badblocks.unacked_exist) {
-		len+= sprintf(page+len, "%sfaulty",sep);
-		sep = ",";
-	}
-	if (test_bit(In_sync, &flags)) {
-		len += sprintf(page+len, "%sin_sync",sep);
-		sep = ",";
-	}
-	if (test_bit(Journal, &flags)) {
-		len += sprintf(page+len, "%sjournal",sep);
-		sep = ",";
-	}
-	if (test_bit(WriteMostly, &flags)) {
-		len += sprintf(page+len, "%swrite_mostly",sep);
-		sep = ",";
-	}
+	    (!test_bit(ExternalBbl, &flags) &&
+	    rdev->badblocks.unacked_exist))
+		len += sprintf(page+len, "faulty%s", sep);
+	if (test_bit(In_sync, &flags))
+		len += sprintf(page+len, "in_sync%s", sep);
+	if (test_bit(Journal, &flags))
+		len += sprintf(page+len, "journal%s", sep);
+	if (test_bit(WriteMostly, &flags))
+		len += sprintf(page+len, "write_mostly%s", sep);
 	if (test_bit(Blocked, &flags) ||
 	    (rdev->badblocks.unacked_exist
-	     && !test_bit(Faulty, &flags))) {
-		len += sprintf(page+len, "%sblocked", sep);
-		sep = ",";
-	}
+	     && !test_bit(Faulty, &flags)))
+		len += sprintf(page+len, "blocked%s", sep);
 	if (!test_bit(Faulty, &flags) &&
 	    !test_bit(Journal, &flags) &&
-	    !test_bit(In_sync, &flags)) {
-		len += sprintf(page+len, "%sspare", sep);
-		sep = ",";
-	}
-	if (test_bit(WriteErrorSeen, &flags)) {
-		len += sprintf(page+len, "%swrite_error", sep);
-		sep = ",";
-	}
-	if (test_bit(WantReplacement, &flags)) {
-		len += sprintf(page+len, "%swant_replacement", sep);
-		sep = ",";
-	}
-	if (test_bit(Replacement, &flags)) {
-		len += sprintf(page+len, "%sreplacement", sep);
-		sep = ",";
-	}
+	    !test_bit(In_sync, &flags))
+		len += sprintf(page+len, "spare%s", sep);
+	if (test_bit(WriteErrorSeen, &flags))
+		len += sprintf(page+len, "write_error%s", sep);
+	if (test_bit(WantReplacement, &flags))
+		len += sprintf(page+len, "want_replacement%s", sep);
+	if (test_bit(Replacement, &flags))
+		len += sprintf(page+len, "replacement%s", sep);
+	if (test_bit(ExternalBbl, &flags))
+		len += sprintf(page+len, "external_bbl%s", sep);
+	if (test_bit(FailFast, &flags))
+		len += sprintf(page+len, "failfast%s", sep);
+
+	if (len)
+		len -= strlen(sep);
 
 	return len+sprintf(page+len, "\n");
 }
@@ -2587,6 +2615,7 @@ state_store(struct md_rdev *rdev, const char *buf, size_t len)
 	 *            so that it gets rebuilt based on bitmap
 	 *  write_error - sets WriteErrorSeen
 	 *  -write_error - clears WriteErrorSeen
+	 *  {,-}failfast - set/clear FailFast
 	 */
 	int err = -EINVAL;
 	if (cmd_match(buf, "faulty") && rdev->mddev->pers) {
@@ -2610,8 +2639,10 @@ state_store(struct md_rdev *rdev, const char *buf, size_t len)
 
 			if (err == 0) {
 				md_kick_rdev_from_array(rdev);
-				if (mddev->pers)
-					md_update_sb(mddev, 1);
+				if (mddev->pers) {
+					set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
+					md_wakeup_thread(mddev->thread);
+				}
 				md_new_event(mddev);
 			}
 		}
@@ -2626,6 +2657,7 @@ state_store(struct md_rdev *rdev, const char *buf, size_t len)
 		err = 0;
 	} else if (cmd_match(buf, "-blocked")) {
 		if (!test_bit(Faulty, &rdev->flags) &&
+		    !test_bit(ExternalBbl, &rdev->flags) &&
 		    rdev->badblocks.unacked_exist) {
 			/* metadata handler doesn't understand badblocks,
 			 * so we need to fail the device
@@ -2642,6 +2674,12 @@ state_store(struct md_rdev *rdev, const char *buf, size_t len)
 	} else if (cmd_match(buf, "insync") && rdev->raid_disk == -1) {
 		set_bit(In_sync, &rdev->flags);
 		err = 0;
+	} else if (cmd_match(buf, "failfast")) {
+		set_bit(FailFast, &rdev->flags);
+		err = 0;
+	} else if (cmd_match(buf, "-failfast")) {
+		clear_bit(FailFast, &rdev->flags);
+		err = 0;
 	} else if (cmd_match(buf, "-insync") && rdev->raid_disk >= 0 &&
 		   !test_bit(Journal, &rdev->flags)) {
 		if (rdev->mddev->pers == NULL) {
@@ -2708,6 +2746,13 @@ state_store(struct md_rdev *rdev, const char *buf, size_t len)
 			}
 		} else
 			err = -EBUSY;
+	} else if (cmd_match(buf, "external_bbl") && (rdev->mddev->external)) {
+		set_bit(ExternalBbl, &rdev->flags);
+		rdev->badblocks.shift = 0;
+		err = 0;
+	} else if (cmd_match(buf, "-external_bbl") && (rdev->mddev->external)) {
+		clear_bit(ExternalBbl, &rdev->flags);
+		err = 0;
 	}
 	if (!err)
 		sysfs_notify_dirent_safe(rdev->sysfs_state);
@@ -3211,10 +3256,8 @@ static struct md_rdev *md_import_device(dev_t newdev, int super_format, int supe
 	sector_t size;
 
 	rdev = kzalloc(sizeof(*rdev), GFP_KERNEL);
-	if (!rdev) {
-		printk(KERN_ERR "md: could not alloc mem for new device!\n");
+	if (!rdev)
 		return ERR_PTR(-ENOMEM);
-	}
 
 	err = md_rdev_init(rdev);
 	if (err)
@@ -3231,8 +3274,7 @@ static struct md_rdev *md_import_device(dev_t newdev, int super_format, int supe
 
 	size = i_size_read(rdev->bdev->bd_inode) >> BLOCK_SIZE_BITS;
 	if (!size) {
-		printk(KERN_WARNING
-			"md: %s has zero or unknown size, marking faulty!\n",
+		pr_warn("md: %s has zero or unknown size, marking faulty!\n",
 			bdevname(rdev->bdev,b));
 		err = -EINVAL;
 		goto abort_free;
@@ -3242,16 +3284,13 @@ static struct md_rdev *md_import_device(dev_t newdev, int super_format, int supe
 		err = super_types[super_format].
 			load_super(rdev, NULL, super_minor);
 		if (err == -EINVAL) {
-			printk(KERN_WARNING
-				"md: %s does not have a valid v%d.%d "
-			       "superblock, not importing!\n",
+			pr_warn("md: %s does not have a valid v%d.%d superblock, not importing!\n",
 				bdevname(rdev->bdev,b),
-			       super_format, super_minor);
+				super_format, super_minor);
 			goto abort_free;
 		}
 		if (err < 0) {
-			printk(KERN_WARNING
-				"md: could not read %s's sb, not importing!\n",
+			pr_warn("md: could not read %s's sb, not importing!\n",
 				bdevname(rdev->bdev,b));
 			goto abort_free;
 		}
@@ -3287,9 +3326,7 @@ static void analyze_sbs(struct mddev *mddev)
 		case 0:
 			break;
 		default:
-			printk( KERN_ERR \
-				"md: fatal superblock inconsistency in %s"
-				" -- removing from array\n",
+			pr_warn("md: fatal superblock inconsistency in %s -- removing from array\n",
 				bdevname(rdev->bdev,b));
 			md_kick_rdev_from_array(rdev);
 		}
@@ -3302,18 +3339,16 @@ static void analyze_sbs(struct mddev *mddev)
 		if (mddev->max_disks &&
 		    (rdev->desc_nr >= mddev->max_disks ||
 		     i > mddev->max_disks)) {
-			printk(KERN_WARNING
-			       "md: %s: %s: only %d devices permitted\n",
-			       mdname(mddev), bdevname(rdev->bdev, b),
-			       mddev->max_disks);
+			pr_warn("md: %s: %s: only %d devices permitted\n",
+				mdname(mddev), bdevname(rdev->bdev, b),
+				mddev->max_disks);
 			md_kick_rdev_from_array(rdev);
 			continue;
 		}
 		if (rdev != freshest) {
 			if (super_types[mddev->major_version].
 			    validate_super(mddev, rdev)) {
-				printk(KERN_WARNING "md: kicking non-fresh %s"
-					" from array!\n",
+				pr_warn("md: kicking non-fresh %s from array!\n",
 					bdevname(rdev->bdev,b));
 				md_kick_rdev_from_array(rdev);
 				continue;
@@ -3384,7 +3419,7 @@ safe_delay_store(struct mddev *mddev, const char *cbuf, size_t len)
 	unsigned long msec;
 
 	if (mddev_is_clustered(mddev)) {
-		pr_info("md: Safemode is disabled for clustered mode\n");
+		pr_warn("md: Safemode is disabled for clustered mode\n");
 		return -EINVAL;
 	}
 
@@ -3472,8 +3507,8 @@ level_store(struct mddev *mddev, const char *buf, size_t len)
 
 	rv = -EINVAL;
 	if (!mddev->pers->quiesce) {
-		printk(KERN_WARNING "md: %s: %s does not support online personality change\n",
-		       mdname(mddev), mddev->pers->name);
+		pr_warn("md: %s: %s does not support online personality change\n",
+			mdname(mddev), mddev->pers->name);
 		goto out_unlock;
 	}
 
@@ -3491,7 +3526,7 @@ level_store(struct mddev *mddev, const char *buf, size_t len)
 	pers = find_pers(level, clevel);
 	if (!pers || !try_module_get(pers->owner)) {
 		spin_unlock(&pers_lock);
-		printk(KERN_WARNING "md: personality %s not loaded\n", clevel);
+		pr_warn("md: personality %s not loaded\n", clevel);
 		rv = -EINVAL;
 		goto out_unlock;
 	}
@@ -3505,8 +3540,8 @@ level_store(struct mddev *mddev, const char *buf, size_t len)
 	}
 	if (!pers->takeover) {
 		module_put(pers->owner);
-		printk(KERN_WARNING "md: %s: %s does not support personality takeover\n",
-		       mdname(mddev), clevel);
+		pr_warn("md: %s: %s does not support personality takeover\n",
+			mdname(mddev), clevel);
 		rv = -EINVAL;
 		goto out_unlock;
 	}
@@ -3526,8 +3561,8 @@ level_store(struct mddev *mddev, const char *buf, size_t len)
 		mddev->delta_disks = 0;
 		mddev->reshape_backwards = 0;
 		module_put(pers->owner);
-		printk(KERN_WARNING "md: %s: %s would not accept array\n",
-		       mdname(mddev), clevel);
+		pr_warn("md: %s: %s would not accept array\n",
+			mdname(mddev), clevel);
 		rv = PTR_ERR(priv);
 		goto out_unlock;
 	}
@@ -3570,9 +3605,8 @@ level_store(struct mddev *mddev, const char *buf, size_t len)
 	    pers->sync_request != NULL) {
 		/* need to add the md_redundancy_group */
 		if (sysfs_create_group(&mddev->kobj, &md_redundancy_group))
-			printk(KERN_WARNING
-			       "md: cannot register extra attributes for %s\n",
-			       mdname(mddev));
+			pr_warn("md: cannot register extra attributes for %s\n",
+				mdname(mddev));
 		mddev->sysfs_action = sysfs_get_dirent(mddev->kobj.sd, "sync_action");
 	}
 	if (oldpers->sync_request != NULL &&
@@ -3603,9 +3637,8 @@ level_store(struct mddev *mddev, const char *buf, size_t len)
 			clear_bit(In_sync, &rdev->flags);
 		else {
 			if (sysfs_link_rdev(mddev, rdev))
-				printk(KERN_WARNING "md: cannot register rd%d"
-				       " for %s after level change\n",
-				       rdev->raid_disk, mdname(mddev));
+				pr_warn("md: cannot register rd%d for %s after level change\n",
+					rdev->raid_disk, mdname(mddev));
 		}
 	}
 
@@ -3618,7 +3651,7 @@ level_store(struct mddev *mddev, const char *buf, size_t len)
 	}
 	blk_set_stacking_limits(&mddev->queue->limits);
 	pers->run(mddev);
-	set_bit(MD_CHANGE_DEVS, &mddev->flags);
+	set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
 	mddev_resume(mddev);
 	if (!mddev->thread)
 		md_update_sb(mddev, 1);
@@ -3813,7 +3846,7 @@ resync_start_store(struct mddev *mddev, const char *buf, size_t len)
 	if (!err) {
 		mddev->recovery_cp = n;
 		if (mddev->pers)
-			set_bit(MD_CHANGE_CLEAN, &mddev->flags);
+			set_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags);
 	}
 	mddev_unlock(mddev);
 	return err ?: len;
@@ -3887,7 +3920,7 @@ array_state_show(struct mddev *mddev, char *page)
 			st = read_auto;
 			break;
 		case 0:
-			if (test_bit(MD_CHANGE_PENDING, &mddev->flags))
+			if (test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags))
 				st = write_pending;
 			else if (mddev->in_sync)
 				st = clean;
@@ -3925,7 +3958,8 @@ array_state_store(struct mddev *mddev, const char *buf, size_t len)
 		spin_lock(&mddev->lock);
 		if (st == active) {
 			restart_array(mddev);
-			clear_bit(MD_CHANGE_PENDING, &mddev->flags);
+			clear_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags);
+			md_wakeup_thread(mddev->thread);
 			wake_up(&mddev->sb_wait);
 			err = 0;
 		} else /* st == clean */ {
@@ -3935,7 +3969,7 @@ array_state_store(struct mddev *mddev, const char *buf, size_t len)
 					mddev->in_sync = 1;
 					if (mddev->safemode == 1)
 						mddev->safemode = 0;
-					set_bit(MD_CHANGE_CLEAN, &mddev->flags);
+					set_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags);
 				}
 				err = 0;
 			} else
@@ -4001,7 +4035,7 @@ array_state_store(struct mddev *mddev, const char *buf, size_t len)
 					mddev->in_sync = 1;
 					if (mddev->safemode == 1)
 						mddev->safemode = 0;
-					set_bit(MD_CHANGE_CLEAN, &mddev->flags);
+					set_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags);
 				}
 				err = 0;
 			} else
@@ -4015,7 +4049,7 @@ array_state_store(struct mddev *mddev, const char *buf, size_t len)
 			err = restart_array(mddev);
 			if (err)
 				break;
-			clear_bit(MD_CHANGE_PENDING, &mddev->flags);
+			clear_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags);
 			wake_up(&mddev->sb_wait);
 			err = 0;
 		} else {
@@ -5071,13 +5105,13 @@ static int md_alloc(dev_t dev, char *name)
 		/* This isn't possible, but as kobject_init_and_add is marked
 		 * __must_check, we must do something with the result
 		 */
-		printk(KERN_WARNING "md: cannot register %s/md - name in use\n",
-		       disk->disk_name);
+		pr_debug("md: cannot register %s/md - name in use\n",
+			 disk->disk_name);
 		error = 0;
 	}
 	if (mddev->kobj.sd &&
 	    sysfs_create_group(&mddev->kobj, &md_bitmap_group))
-		printk(KERN_DEBUG "pointless warning\n");
+		pr_debug("pointless warning\n");
 	mutex_unlock(&mddev->open_mutex);
  abort:
 	mutex_unlock(&disks_mutex);
@@ -5179,15 +5213,15 @@ int md_run(struct mddev *mddev)
 			if (mddev->dev_sectors &&
 			    rdev->data_offset + mddev->dev_sectors
 			    > rdev->sb_start) {
-				printk("md: %s: data overlaps metadata\n",
-				       mdname(mddev));
+				pr_warn("md: %s: data overlaps metadata\n",
+					mdname(mddev));
 				return -EINVAL;
 			}
 		} else {
 			if (rdev->sb_start + rdev->sb_size/512
 			    > rdev->data_offset) {
-				printk("md: %s: metadata overlaps data\n",
-				       mdname(mddev));
+				pr_warn("md: %s: metadata overlaps data\n",
+					mdname(mddev));
 				return -EINVAL;
 			}
 		}
@@ -5202,11 +5236,11 @@ int md_run(struct mddev *mddev)
 	if (!pers || !try_module_get(pers->owner)) {
 		spin_unlock(&pers_lock);
 		if (mddev->level != LEVEL_NONE)
-			printk(KERN_WARNING "md: personality for level %d is not loaded!\n",
-			       mddev->level);
+			pr_warn("md: personality for level %d is not loaded!\n",
+				mddev->level);
 		else
-			printk(KERN_WARNING "md: personality for level %s is not loaded!\n",
-			       mddev->clevel);
+			pr_warn("md: personality for level %s is not loaded!\n",
+				mddev->clevel);
 		return -EINVAL;
 	}
 	spin_unlock(&pers_lock);
@@ -5236,21 +5270,16 @@ int md_run(struct mddev *mddev)
 				if (rdev < rdev2 &&
 				    rdev->bdev->bd_contains ==
 				    rdev2->bdev->bd_contains) {
-					printk(KERN_WARNING
-					       "%s: WARNING: %s appears to be"
-					       " on the same physical disk as"
-					       " %s.\n",
-					       mdname(mddev),
-					       bdevname(rdev->bdev,b),
-					       bdevname(rdev2->bdev,b2));
+					pr_warn("%s: WARNING: %s appears to be on the same physical disk as %s.\n",
+						mdname(mddev),
+						bdevname(rdev->bdev,b),
+						bdevname(rdev2->bdev,b2));
 					warned = 1;
 				}
 			}
 
 		if (warned)
-			printk(KERN_WARNING
-			       "True protection against single-disk"
-			       " failure might be compromised.\n");
+			pr_warn("True protection against single-disk failure might be compromised.\n");
 	}
 
 	mddev->recovery = 0;
@@ -5264,14 +5293,14 @@ int md_run(struct mddev *mddev)
 
 	err = pers->run(mddev);
 	if (err)
-		printk(KERN_ERR "md: pers->run() failed ...\n");
+		pr_warn("md: pers->run() failed ...\n");
 	else if (pers->size(mddev, 0, 0) < mddev->array_sectors) {
-		WARN_ONCE(!mddev->external_size, "%s: default size too small,"
-			  " but 'external_size' not in effect?\n", __func__);
-		printk(KERN_ERR
-		       "md: invalid array_size %llu > default size %llu\n",
-		       (unsigned long long)mddev->array_sectors / 2,
-		       (unsigned long long)pers->size(mddev, 0, 0) / 2);
+		WARN_ONCE(!mddev->external_size,
+			  "%s: default size too small, but 'external_size' not in effect?\n",
+			  __func__);
+		pr_warn("md: invalid array_size %llu > default size %llu\n",
+			(unsigned long long)mddev->array_sectors / 2,
+			(unsigned long long)pers->size(mddev, 0, 0) / 2);
 		err = -EINVAL;
 	}
 	if (err == 0 && pers->sync_request &&
@@ -5281,8 +5310,8 @@ int md_run(struct mddev *mddev)
 		bitmap = bitmap_create(mddev, -1);
 		if (IS_ERR(bitmap)) {
 			err = PTR_ERR(bitmap);
-			printk(KERN_ERR "%s: failed to create bitmap (%d)\n",
-			       mdname(mddev), err);
+			pr_warn("%s: failed to create bitmap (%d)\n",
+				mdname(mddev), err);
 		} else
 			mddev->bitmap = bitmap;
 
@@ -5318,9 +5347,8 @@ int md_run(struct mddev *mddev)
 	if (pers->sync_request) {
 		if (mddev->kobj.sd &&
 		    sysfs_create_group(&mddev->kobj, &md_redundancy_group))
-			printk(KERN_WARNING
-			       "md: cannot register extra attributes for %s\n",
-			       mdname(mddev));
+			pr_warn("md: cannot register extra attributes for %s\n",
+				mdname(mddev));
 		mddev->sysfs_action = sysfs_get_dirent_safe(mddev->kobj.sd, "sync_action");
 	} else if (mddev->ro == 2) /* auto-readonly not meaningful */
 		mddev->ro = 0;
@@ -5350,7 +5378,7 @@ int md_run(struct mddev *mddev)
 		set_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
 	set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
 
-	if (mddev->flags & MD_UPDATE_SB_FLAGS)
+	if (mddev->sb_flags)
 		md_update_sb(mddev, 0);
 
 	md_new_event(mddev);
@@ -5421,8 +5449,7 @@ static int restart_array(struct mddev *mddev)
 	mddev->safemode = 0;
 	mddev->ro = 0;
 	set_disk_ro(disk, 0);
-	printk(KERN_INFO "md: %s switched to read-write mode.\n",
-		mdname(mddev));
+	pr_debug("md: %s switched to read-write mode.\n", mdname(mddev));
 	/* Kick recovery or resync if necessary */
 	set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
 	md_wakeup_thread(mddev->thread);
@@ -5446,6 +5473,7 @@ static void md_clean(struct mddev *mddev)
 	mddev->level = LEVEL_NONE;
 	mddev->clevel[0] = 0;
 	mddev->flags = 0;
+	mddev->sb_flags = 0;
 	mddev->ro = 0;
 	mddev->metadata_type[0] = 0;
 	mddev->chunk_sectors = 0;
@@ -5490,12 +5518,15 @@ static void __md_stop_writes(struct mddev *mddev)
 
 	del_timer_sync(&mddev->safemode_timer);
 
+	if (mddev->pers && mddev->pers->quiesce) {
+		mddev->pers->quiesce(mddev, 1);
+		mddev->pers->quiesce(mddev, 0);
+	}
 	bitmap_flush(mddev);
-	md_super_wait(mddev);
 
 	if (mddev->ro == 0 &&
 	    ((!mddev->in_sync && !mddev_is_clustered(mddev)) ||
-	     (mddev->flags & MD_UPDATE_SB_FLAGS))) {
+	     mddev->sb_flags)) {
 		/* mark array as shutdown cleanly */
 		if (!mddev_is_clustered(mddev))
 			mddev->in_sync = 1;
@@ -5516,8 +5547,8 @@ static void mddev_detach(struct mddev *mddev)
 	struct bitmap *bitmap = mddev->bitmap;
 	/* wait for behind writes to complete */
 	if (bitmap && atomic_read(&bitmap->behind_writes) > 0) {
-		printk(KERN_INFO "md:%s: behind writes in progress - waiting to stop.\n",
-		       mdname(mddev));
+		pr_debug("md:%s: behind writes in progress - waiting to stop.\n",
+			 mdname(mddev));
 		/* need to kick something here to make sure I/O goes? */
 		wait_event(bitmap->behind_wait,
 			   atomic_read(&bitmap->behind_writes) == 0);
@@ -5578,20 +5609,20 @@ static int md_set_readonly(struct mddev *mddev, struct block_device *bdev)
 		 * which will now never happen */
 		wake_up_process(mddev->sync_thread->tsk);
 
-	if (mddev->external && test_bit(MD_CHANGE_PENDING, &mddev->flags))
+	if (mddev->external && test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags))
 		return -EBUSY;
 	mddev_unlock(mddev);
 	wait_event(resync_wait, !test_bit(MD_RECOVERY_RUNNING,
 					  &mddev->recovery));
 	wait_event(mddev->sb_wait,
-		   !test_bit(MD_CHANGE_PENDING, &mddev->flags));
+		   !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags));
 	mddev_lock_nointr(mddev);
 
 	mutex_lock(&mddev->open_mutex);
 	if ((mddev->pers && atomic_read(&mddev->openers) > !!bdev) ||
 	    mddev->sync_thread ||
 	    test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) {
-		printk("md: %s still in use.\n",mdname(mddev));
+		pr_warn("md: %s still in use.\n",mdname(mddev));
 		if (did_freeze) {
 			clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
 			set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
@@ -5653,7 +5684,7 @@ static int do_md_stop(struct mddev *mddev, int mode,
 	    mddev->sysfs_active ||
 	    mddev->sync_thread ||
 	    test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) {
-		printk("md: %s still in use.\n",mdname(mddev));
+		pr_warn("md: %s still in use.\n",mdname(mddev));
 		mutex_unlock(&mddev->open_mutex);
 		if (did_freeze) {
 			clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
@@ -5690,7 +5721,7 @@ static int do_md_stop(struct mddev *mddev, int mode,
 	 * Free resources if final stop
 	 */
 	if (mode == 0) {
-		printk(KERN_INFO "md: %s stopped.\n", mdname(mddev));
+		pr_info("md: %s stopped.\n", mdname(mddev));
 
 		bitmap_destroy(mddev);
 		if (mddev->bitmap_info.file) {
@@ -5722,17 +5753,17 @@ static void autorun_array(struct mddev *mddev)
 	if (list_empty(&mddev->disks))
 		return;
 
-	printk(KERN_INFO "md: running: ");
+	pr_info("md: running: ");
 
 	rdev_for_each(rdev, mddev) {
 		char b[BDEVNAME_SIZE];
-		printk("<%s>", bdevname(rdev->bdev,b));
+		pr_cont("<%s>", bdevname(rdev->bdev,b));
 	}
-	printk("\n");
+	pr_cont("\n");
 
 	err = do_md_run(mddev);
 	if (err) {
-		printk(KERN_WARNING "md: do_md_run() returned %d\n", err);
+		pr_warn("md: do_md_run() returned %d\n", err);
 		do_md_stop(mddev, 0, NULL);
 	}
 }
@@ -5755,7 +5786,7 @@ static void autorun_devices(int part)
 	struct mddev *mddev;
 	char b[BDEVNAME_SIZE];
 
-	printk(KERN_INFO "md: autorun ...\n");
+	pr_info("md: autorun ...\n");
 	while (!list_empty(&pending_raid_disks)) {
 		int unit;
 		dev_t dev;
@@ -5763,13 +5794,12 @@ static void autorun_devices(int part)
 		rdev0 = list_entry(pending_raid_disks.next,
 					 struct md_rdev, same_set);
 
-		printk(KERN_INFO "md: considering %s ...\n",
-			bdevname(rdev0->bdev,b));
+		pr_debug("md: considering %s ...\n", bdevname(rdev0->bdev,b));
 		INIT_LIST_HEAD(&candidates);
 		rdev_for_each_list(rdev, tmp, &pending_raid_disks)
 			if (super_90_load(rdev, rdev0, 0) >= 0) {
-				printk(KERN_INFO "md:  adding %s ...\n",
-					bdevname(rdev->bdev,b));
+				pr_debug("md:  adding %s ...\n",
+					 bdevname(rdev->bdev,b));
 				list_move(&rdev->same_set, &candidates);
 			}
 		/*
@@ -5786,8 +5816,8 @@ static void autorun_devices(int part)
 			unit = MINOR(dev);
 		}
 		if (rdev0->preferred_minor != unit) {
-			printk(KERN_INFO "md: unit number in %s is bad: %d\n",
-			       bdevname(rdev0->bdev, b), rdev0->preferred_minor);
+			pr_warn("md: unit number in %s is bad: %d\n",
+				bdevname(rdev0->bdev, b), rdev0->preferred_minor);
 			break;
 		}
 
@@ -5796,21 +5826,17 @@ static void autorun_devices(int part)
 		if (!mddev || !mddev->gendisk) {
 			if (mddev)
 				mddev_put(mddev);
-			printk(KERN_ERR
-				"md: cannot allocate memory for md drive.\n");
 			break;
 		}
 		if (mddev_lock(mddev))
-			printk(KERN_WARNING "md: %s locked, cannot run\n",
-			       mdname(mddev));
+			pr_warn("md: %s locked, cannot run\n", mdname(mddev));
 		else if (mddev->raid_disks || mddev->major_version
 			 || !list_empty(&mddev->disks)) {
-			printk(KERN_WARNING
-				"md: %s already running, cannot run %s\n",
+			pr_warn("md: %s already running, cannot run %s\n",
 				mdname(mddev), bdevname(rdev0->bdev,b));
 			mddev_unlock(mddev);
 		} else {
-			printk(KERN_INFO "md: created %s\n", mdname(mddev));
+			pr_debug("md: created %s\n", mdname(mddev));
 			mddev->persistent = 1;
 			rdev_for_each_list(rdev, tmp, &candidates) {
 				list_del_init(&rdev->same_set);
@@ -5829,7 +5855,7 @@ static void autorun_devices(int part)
 		}
 		mddev_put(mddev);
 	}
-	printk(KERN_INFO "md: ... autorun DONE.\n");
+	pr_info("md: ... autorun DONE.\n");
 }
 #endif /* !MODULE */
 
@@ -5964,6 +5990,8 @@ static int get_disk_info(struct mddev *mddev, void __user * arg)
 			info.state |= (1<<MD_DISK_JOURNAL);
 		if (test_bit(WriteMostly, &rdev->flags))
 			info.state |= (1<<MD_DISK_WRITEMOSTLY);
+		if (test_bit(FailFast, &rdev->flags))
+			info.state |= (1<<MD_DISK_FAILFAST);
 	} else {
 		info.major = info.minor = 0;
 		info.raid_disk = -1;
@@ -5985,8 +6013,8 @@ static int add_new_disk(struct mddev *mddev, mdu_disk_info_t *info)
 
 	if (mddev_is_clustered(mddev) &&
 		!(info->state & ((1 << MD_DISK_CLUSTER_ADD) | (1 << MD_DISK_CANDIDATE)))) {
-		pr_err("%s: Cannot add to clustered mddev.\n",
-			       mdname(mddev));
+		pr_warn("%s: Cannot add to clustered mddev.\n",
+			mdname(mddev));
 		return -EINVAL;
 	}
 
@@ -5998,8 +6026,7 @@ static int add_new_disk(struct mddev *mddev, mdu_disk_info_t *info)
 		/* expecting a device which has a superblock */
 		rdev = md_import_device(dev, mddev->major_version, mddev->minor_version);
 		if (IS_ERR(rdev)) {
-			printk(KERN_WARNING
-				"md: md_import_device returned %ld\n",
+			pr_warn("md: md_import_device returned %ld\n",
 				PTR_ERR(rdev));
 			return PTR_ERR(rdev);
 		}
@@ -6010,8 +6037,7 @@ static int add_new_disk(struct mddev *mddev, mdu_disk_info_t *info)
 			err = super_types[mddev->major_version]
 				.load_super(rdev, rdev0, mddev->minor_version);
 			if (err < 0) {
-				printk(KERN_WARNING
-					"md: %s has different UUID to %s\n",
+				pr_warn("md: %s has different UUID to %s\n",
 					bdevname(rdev->bdev,b),
 					bdevname(rdev0->bdev,b2));
 				export_rdev(rdev);
@@ -6032,9 +6058,8 @@ static int add_new_disk(struct mddev *mddev, mdu_disk_info_t *info)
 	if (mddev->pers) {
 		int err;
 		if (!mddev->pers->hot_add_disk) {
-			printk(KERN_WARNING
-				"%s: personality does not support diskops!\n",
-			       mdname(mddev));
+			pr_warn("%s: personality does not support diskops!\n",
+				mdname(mddev));
 			return -EINVAL;
 		}
 		if (mddev->persistent)
@@ -6043,8 +6068,7 @@ static int add_new_disk(struct mddev *mddev, mdu_disk_info_t *info)
 		else
 			rdev = md_import_device(dev, -1, -1);
 		if (IS_ERR(rdev)) {
-			printk(KERN_WARNING
-				"md: md_import_device returned %ld\n",
+			pr_warn("md: md_import_device returned %ld\n",
 				PTR_ERR(rdev));
 			return PTR_ERR(rdev);
 		}
@@ -6075,6 +6099,10 @@ static int add_new_disk(struct mddev *mddev, mdu_disk_info_t *info)
 			set_bit(WriteMostly, &rdev->flags);
 		else
 			clear_bit(WriteMostly, &rdev->flags);
+		if (info->state & (1<<MD_DISK_FAILFAST))
+			set_bit(FailFast, &rdev->flags);
+		else
+			clear_bit(FailFast, &rdev->flags);
 
 		if (info->state & (1<<MD_DISK_JOURNAL)) {
 			struct md_rdev *rdev2;
@@ -6140,8 +6168,7 @@ static int add_new_disk(struct mddev *mddev, mdu_disk_info_t *info)
 	 * for major_version==0 superblocks
 	 */
 	if (mddev->major_version != 0) {
-		printk(KERN_WARNING "%s: ADD_NEW_DISK not supported\n",
-		       mdname(mddev));
+		pr_warn("%s: ADD_NEW_DISK not supported\n", mdname(mddev));
 		return -EINVAL;
 	}
 
@@ -6149,8 +6176,7 @@ static int add_new_disk(struct mddev *mddev, mdu_disk_info_t *info)
 		int err;
 		rdev = md_import_device(dev, -1, 0);
 		if (IS_ERR(rdev)) {
-			printk(KERN_WARNING
-				"md: error, md_import_device() returned %ld\n",
+			pr_warn("md: error, md_import_device() returned %ld\n",
 				PTR_ERR(rdev));
 			return PTR_ERR(rdev);
 		}
@@ -6166,9 +6192,11 @@ static int add_new_disk(struct mddev *mddev, mdu_disk_info_t *info)
 
 		if (info->state & (1<<MD_DISK_WRITEMOSTLY))
 			set_bit(WriteMostly, &rdev->flags);
+		if (info->state & (1<<MD_DISK_FAILFAST))
+			set_bit(FailFast, &rdev->flags);
 
 		if (!mddev->persistent) {
-			printk(KERN_INFO "md: nonpersistent superblock ...\n");
+			pr_debug("md: nonpersistent superblock ...\n");
 			rdev->sb_start = i_size_read(rdev->bdev->bd_inode) / 512;
 		} else
 			rdev->sb_start = calc_dev_sboffset(rdev);
@@ -6207,13 +6235,17 @@ kick_rdev:
 		md_cluster_ops->remove_disk(mddev, rdev);
 
 	md_kick_rdev_from_array(rdev);
-	md_update_sb(mddev, 1);
+	set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
+	if (mddev->thread)
+		md_wakeup_thread(mddev->thread);
+	else
+		md_update_sb(mddev, 1);
 	md_new_event(mddev);
 
 	return 0;
 busy:
-	printk(KERN_WARNING "md: cannot remove active disk %s from %s ...\n",
-		bdevname(rdev->bdev,b), mdname(mddev));
+	pr_debug("md: cannot remove active disk %s from %s ...\n",
+		 bdevname(rdev->bdev,b), mdname(mddev));
 	return -EBUSY;
 }
 
@@ -6227,22 +6259,19 @@ static int hot_add_disk(struct mddev *mddev, dev_t dev)
 		return -ENODEV;
 
 	if (mddev->major_version != 0) {
-		printk(KERN_WARNING "%s: HOT_ADD may only be used with"
-			" version-0 superblocks.\n",
+		pr_warn("%s: HOT_ADD may only be used with version-0 superblocks.\n",
 			mdname(mddev));
 		return -EINVAL;
 	}
 	if (!mddev->pers->hot_add_disk) {
-		printk(KERN_WARNING
-			"%s: personality does not support diskops!\n",
+		pr_warn("%s: personality does not support diskops!\n",
 			mdname(mddev));
 		return -EINVAL;
 	}
 
 	rdev = md_import_device(dev, -1, 0);
 	if (IS_ERR(rdev)) {
-		printk(KERN_WARNING
-			"md: error, md_import_device() returned %ld\n",
+		pr_warn("md: error, md_import_device() returned %ld\n",
 			PTR_ERR(rdev));
 		return -EINVAL;
 	}
@@ -6255,8 +6284,7 @@ static int hot_add_disk(struct mddev *mddev, dev_t dev)
 	rdev->sectors = rdev->sb_start;
 
 	if (test_bit(Faulty, &rdev->flags)) {
-		printk(KERN_WARNING
-			"md: can not hot-add faulty %s disk to %s!\n",
+		pr_warn("md: can not hot-add faulty %s disk to %s!\n",
 			bdevname(rdev->bdev,b), mdname(mddev));
 		err = -EINVAL;
 		goto abort_export;
@@ -6276,7 +6304,9 @@ static int hot_add_disk(struct mddev *mddev, dev_t dev)
 
 	rdev->raid_disk = -1;
 
-	md_update_sb(mddev, 1);
+	set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
+	if (!mddev->thread)
+		md_update_sb(mddev, 1);
 	/*
 	 * Kick recovery, maybe this spare has to be added to the
 	 * array immediately.
@@ -6312,23 +6342,23 @@ static int set_bitmap_file(struct mddev *mddev, int fd)
 		f = fget(fd);
 
 		if (f == NULL) {
-			printk(KERN_ERR "%s: error: failed to get bitmap file\n",
-			       mdname(mddev));
+			pr_warn("%s: error: failed to get bitmap file\n",
+				mdname(mddev));
 			return -EBADF;
 		}
 
 		inode = f->f_mapping->host;
 		if (!S_ISREG(inode->i_mode)) {
-			printk(KERN_ERR "%s: error: bitmap file must be a regular file\n",
-			       mdname(mddev));
+			pr_warn("%s: error: bitmap file must be a regular file\n",
+				mdname(mddev));
 			err = -EBADF;
 		} else if (!(f->f_mode & FMODE_WRITE)) {
-			printk(KERN_ERR "%s: error: bitmap file must open for write\n",
-			       mdname(mddev));
+			pr_warn("%s: error: bitmap file must open for write\n",
+				mdname(mddev));
 			err = -EBADF;
 		} else if (atomic_read(&inode->i_writecount) != 1) {
-			printk(KERN_ERR "%s: error: bitmap file is already in use\n",
-			       mdname(mddev));
+			pr_warn("%s: error: bitmap file is already in use\n",
+				mdname(mddev));
 			err = -EBUSY;
 		}
 		if (err) {
@@ -6393,8 +6423,7 @@ static int set_array_info(struct mddev *mddev, mdu_array_info_t *info)
 		    info->major_version >= ARRAY_SIZE(super_types) ||
 		    super_types[info->major_version].name == NULL) {
 			/* maybe try to auto-load a module? */
-			printk(KERN_INFO
-				"md: superblock version %d not known\n",
+			pr_warn("md: superblock version %d not known\n",
 				info->major_version);
 			return -EINVAL;
 		}
@@ -6432,9 +6461,11 @@ static int set_array_info(struct mddev *mddev, mdu_array_info_t *info)
 
 	mddev->max_disks     = MD_SB_DISKS;
 
-	if (mddev->persistent)
+	if (mddev->persistent) {
 		mddev->flags         = 0;
-	set_bit(MD_CHANGE_DEVS, &mddev->flags);
+		mddev->sb_flags         = 0;
+	}
+	set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
 
 	mddev->bitmap_info.default_offset = MD_SB_BYTES >> 9;
 	mddev->bitmap_info.default_space = 64*2 - (MD_SB_BYTES >> 9);
@@ -6660,8 +6691,7 @@ static int update_array_info(struct mddev *mddev, mdu_array_info_t *info)
 			if (mddev->bitmap_info.nodes) {
 				/* hold PW on all the bitmap lock */
 				if (md_cluster_ops->lock_all_bitmaps(mddev) <= 0) {
-					printk("md: can't change bitmap to none since the"
-					       " array is in use by more than one node\n");
+					pr_warn("md: can't change bitmap to none since the array is in use by more than one node\n");
 					rv = -EPERM;
 					md_cluster_ops->unlock_all_bitmaps(mddev);
 					goto err;
@@ -6829,7 +6859,7 @@ static int md_ioctl(struct block_device *bdev, fmode_t mode,
 		/* need to ensure recovery thread has run */
 		wait_event_interruptible_timeout(mddev->sb_wait,
 						 !test_bit(MD_RECOVERY_NEEDED,
-							   &mddev->flags),
+							   &mddev->recovery),
 						 msecs_to_jiffies(5000));
 	if (cmd == STOP_ARRAY || cmd == STOP_ARRAY_RO) {
 		/* Need to flush page cache, and ensure no-one else opens
@@ -6847,9 +6877,8 @@ static int md_ioctl(struct block_device *bdev, fmode_t mode,
 	}
 	err = mddev_lock(mddev);
 	if (err) {
-		printk(KERN_INFO
-			"md: ioctl lock interrupted, reason %d, cmd %d\n",
-			err, cmd);
+		pr_debug("md: ioctl lock interrupted, reason %d, cmd %d\n",
+			 err, cmd);
 		goto out;
 	}
 
@@ -6864,30 +6893,24 @@ static int md_ioctl(struct block_device *bdev, fmode_t mode,
 		if (mddev->pers) {
 			err = update_array_info(mddev, &info);
 			if (err) {
-				printk(KERN_WARNING "md: couldn't update"
-				       " array info. %d\n", err);
+				pr_warn("md: couldn't update array info. %d\n", err);
 				goto unlock;
 			}
 			goto unlock;
 		}
 		if (!list_empty(&mddev->disks)) {
-			printk(KERN_WARNING
-			       "md: array %s already has disks!\n",
-			       mdname(mddev));
+			pr_warn("md: array %s already has disks!\n", mdname(mddev));
 			err = -EBUSY;
 			goto unlock;
 		}
 		if (mddev->raid_disks) {
-			printk(KERN_WARNING
-			       "md: array %s already initialised!\n",
-			       mdname(mddev));
+			pr_warn("md: array %s already initialised!\n", mdname(mddev));
 			err = -EBUSY;
 			goto unlock;
 		}
 		err = set_array_info(mddev, &info);
 		if (err) {
-			printk(KERN_WARNING "md: couldn't set"
-			       " array info. %d\n", err);
+			pr_warn("md: couldn't set array info. %d\n", err);
 			goto unlock;
 		}
 		goto unlock;
@@ -6987,11 +7010,11 @@ static int md_ioctl(struct block_device *bdev, fmode_t mode,
 			/* If a device failed while we were read-only, we
 			 * need to make sure the metadata is updated now.
 			 */
-			if (test_bit(MD_CHANGE_DEVS, &mddev->flags)) {
+			if (test_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags)) {
 				mddev_unlock(mddev);
 				wait_event(mddev->sb_wait,
-					   !test_bit(MD_CHANGE_DEVS, &mddev->flags) &&
-					   !test_bit(MD_CHANGE_PENDING, &mddev->flags));
+					   !test_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags) &&
+					   !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags));
 				mddev_lock_nointr(mddev);
 			}
 		} else {
@@ -7092,7 +7115,8 @@ static int md_open(struct block_device *bdev, fmode_t mode)
 
 	if (test_bit(MD_CLOSING, &mddev->flags)) {
 		mutex_unlock(&mddev->open_mutex);
-		return -ENODEV;
+		err = -ENODEV;
+		goto out;
 	}
 
 	err = 0;
@@ -7101,6 +7125,8 @@ static int md_open(struct block_device *bdev, fmode_t mode)
 
 	check_disk_change(bdev);
  out:
+	if (err)
+		mddev_put(mddev);
 	return err;
 }
 
@@ -7171,10 +7197,12 @@ static int md_thread(void *arg)
 		wait_event_interruptible_timeout
 			(thread->wqueue,
 			 test_bit(THREAD_WAKEUP, &thread->flags)
-			 || kthread_should_stop(),
+			 || kthread_should_stop() || kthread_should_park(),
 			 thread->timeout);
 
 		clear_bit(THREAD_WAKEUP, &thread->flags);
+		if (kthread_should_park())
+			kthread_parkme();
 		if (!kthread_should_stop())
 			thread->run(thread);
 	}
@@ -7588,8 +7616,8 @@ static const struct file_operations md_seq_fops = {
 
 int register_md_personality(struct md_personality *p)
 {
-	printk(KERN_INFO "md: %s personality registered for level %d\n",
-						p->name, p->level);
+	pr_debug("md: %s personality registered for level %d\n",
+		 p->name, p->level);
 	spin_lock(&pers_lock);
 	list_add_tail(&p->list, &pers_list);
 	spin_unlock(&pers_lock);
@@ -7599,7 +7627,7 @@ EXPORT_SYMBOL(register_md_personality);
 
 int unregister_md_personality(struct md_personality *p)
 {
-	printk(KERN_INFO "md: %s personality unregistered\n", p->name);
+	pr_debug("md: %s personality unregistered\n", p->name);
 	spin_lock(&pers_lock);
 	list_del_init(&p->list);
 	spin_unlock(&pers_lock);
@@ -7639,7 +7667,7 @@ int md_setup_cluster(struct mddev *mddev, int nodes)
 	spin_lock(&pers_lock);
 	/* ensure module won't be unloaded */
 	if (!md_cluster_ops || !try_module_get(md_cluster_mod)) {
-		pr_err("can't find md-cluster module or get it's reference.\n");
+		pr_warn("can't find md-cluster module or get it's reference.\n");
 		spin_unlock(&pers_lock);
 		return -ENOENT;
 	}
@@ -7741,8 +7769,8 @@ void md_write_start(struct mddev *mddev, struct bio *bi)
 		spin_lock(&mddev->lock);
 		if (mddev->in_sync) {
 			mddev->in_sync = 0;
-			set_bit(MD_CHANGE_CLEAN, &mddev->flags);
-			set_bit(MD_CHANGE_PENDING, &mddev->flags);
+			set_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags);
+			set_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags);
 			md_wakeup_thread(mddev->thread);
 			did_change = 1;
 		}
@@ -7751,7 +7779,7 @@ void md_write_start(struct mddev *mddev, struct bio *bi)
 	if (did_change)
 		sysfs_notify_dirent_safe(mddev->sysfs_state);
 	wait_event(mddev->sb_wait,
-		   !test_bit(MD_CHANGE_PENDING, &mddev->flags));
+		   !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags));
 }
 EXPORT_SYMBOL(md_write_start);
 
@@ -7772,7 +7800,7 @@ EXPORT_SYMBOL(md_write_end);
  * attempting a GFP_KERNEL allocation while holding the mddev lock.
  * Must be called with mddev_lock held.
  *
- * In the ->external case MD_CHANGE_PENDING can not be cleared until mddev->lock
+ * In the ->external case MD_SB_CHANGE_PENDING can not be cleared until mddev->lock
  * is dropped, so return -EAGAIN after notifying userspace.
  */
 int md_allow_write(struct mddev *mddev)
@@ -7787,8 +7815,8 @@ int md_allow_write(struct mddev *mddev)
 	spin_lock(&mddev->lock);
 	if (mddev->in_sync) {
 		mddev->in_sync = 0;
-		set_bit(MD_CHANGE_CLEAN, &mddev->flags);
-		set_bit(MD_CHANGE_PENDING, &mddev->flags);
+		set_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags);
+		set_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags);
 		if (mddev->safemode_delay &&
 		    mddev->safemode == 0)
 			mddev->safemode = 1;
@@ -7798,7 +7826,7 @@ int md_allow_write(struct mddev *mddev)
 	} else
 		spin_unlock(&mddev->lock);
 
-	if (test_bit(MD_CHANGE_PENDING, &mddev->flags))
+	if (test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags))
 		return -EAGAIN;
 	else
 		return 0;
@@ -7914,11 +7942,9 @@ void md_do_sync(struct md_thread *thread)
 				    mddev2->curr_resync >= mddev->curr_resync) {
 					if (mddev2_minor != mddev2->md_minor) {
 						mddev2_minor = mddev2->md_minor;
-						printk(KERN_INFO "md: delaying %s of %s"
-						       " until %s has finished (they"
-						       " share one or more physical units)\n",
-						       desc, mdname(mddev),
-						       mdname(mddev2));
+						pr_info("md: delaying %s of %s until %s has finished (they share one or more physical units)\n",
+							desc, mdname(mddev),
+							mdname(mddev2));
 					}
 					mddev_put(mddev2);
 					if (signal_pending(current))
@@ -7975,12 +8001,10 @@ void md_do_sync(struct md_thread *thread)
 		}
 	}
 
-	printk(KERN_INFO "md: %s of RAID array %s\n", desc, mdname(mddev));
-	printk(KERN_INFO "md: minimum _guaranteed_  speed:"
-		" %d KB/sec/disk.\n", speed_min(mddev));
-	printk(KERN_INFO "md: using maximum available idle IO bandwidth "
-	       "(but not more than %d KB/sec) for %s.\n",
-	       speed_max(mddev), desc);
+	pr_info("md: %s of RAID array %s\n", desc, mdname(mddev));
+	pr_debug("md: minimum _guaranteed_  speed: %d KB/sec/disk.\n", speed_min(mddev));
+	pr_debug("md: using maximum available idle IO bandwidth (but not more than %d KB/sec) for %s.\n",
+		 speed_max(mddev), desc);
 
 	is_mddev_idle(mddev, 1); /* this initializes IO event counters */
 
@@ -7997,16 +8021,15 @@ void md_do_sync(struct md_thread *thread)
 	 * Tune reconstruction:
 	 */
 	window = 32*(PAGE_SIZE/512);
-	printk(KERN_INFO "md: using %dk window, over a total of %lluk.\n",
-		window/2, (unsigned long long)max_sectors/2);
+	pr_debug("md: using %dk window, over a total of %lluk.\n",
+		 window/2, (unsigned long long)max_sectors/2);
 
 	atomic_set(&mddev->recovery_active, 0);
 	last_check = 0;
 
 	if (j>2) {
-		printk(KERN_INFO
-		       "md: resuming %s of %s from checkpoint.\n",
-		       desc, mdname(mddev));
+		pr_debug("md: resuming %s of %s from checkpoint.\n",
+			 desc, mdname(mddev));
 		mddev->curr_resync = j;
 	} else
 		mddev->curr_resync = 3; /* no longer delayed */
@@ -8038,7 +8061,7 @@ void md_do_sync(struct md_thread *thread)
 			    j > mddev->recovery_cp)
 				mddev->recovery_cp = j;
 			update_time = jiffies;
-			set_bit(MD_CHANGE_CLEAN, &mddev->flags);
+			set_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags);
 			sysfs_notify(&mddev->kobj, NULL, "sync_completed");
 		}
 
@@ -8133,9 +8156,9 @@ void md_do_sync(struct md_thread *thread)
 			}
 		}
 	}
-	printk(KERN_INFO "md: %s: %s %s.\n",mdname(mddev), desc,
-	       test_bit(MD_RECOVERY_INTR, &mddev->recovery)
-	       ? "interrupted" : "done");
+	pr_info("md: %s: %s %s.\n",mdname(mddev), desc,
+		test_bit(MD_RECOVERY_INTR, &mddev->recovery)
+		? "interrupted" : "done");
 	/*
 	 * this also signals 'finished resyncing' to md_stop
 	 */
@@ -8155,9 +8178,8 @@ void md_do_sync(struct md_thread *thread)
 		if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
 			if (test_bit(MD_RECOVERY_INTR, &mddev->recovery)) {
 				if (mddev->curr_resync >= mddev->recovery_cp) {
-					printk(KERN_INFO
-					       "md: checkpointing %s of %s.\n",
-					       desc, mdname(mddev));
+					pr_debug("md: checkpointing %s of %s.\n",
+						 desc, mdname(mddev));
 					if (test_bit(MD_RECOVERY_ERROR,
 						&mddev->recovery))
 						mddev->recovery_cp =
@@ -8187,8 +8209,8 @@ void md_do_sync(struct md_thread *thread)
 	/* set CHANGE_PENDING here since maybe another update is needed,
 	 * so other nodes are informed. It should be harmless for normal
 	 * raid */
-	set_mask_bits(&mddev->flags, 0,
-		      BIT(MD_CHANGE_PENDING) | BIT(MD_CHANGE_DEVS));
+	set_mask_bits(&mddev->sb_flags, 0,
+		      BIT(MD_SB_CHANGE_PENDING) | BIT(MD_SB_CHANGE_DEVS));
 
 	spin_lock(&mddev->lock);
 	if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery)) {
@@ -8288,12 +8310,12 @@ static int remove_and_add_spares(struct mddev *mddev,
 			if (!test_bit(Journal, &rdev->flags))
 				spares++;
 			md_new_event(mddev);
-			set_bit(MD_CHANGE_DEVS, &mddev->flags);
+			set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
 		}
 	}
 no_add:
 	if (removed)
-		set_bit(MD_CHANGE_DEVS, &mddev->flags);
+		set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
 	return spares;
 }
 
@@ -8305,8 +8327,8 @@ static void md_start_sync(struct work_struct *ws)
 						mddev,
 						"resync");
 	if (!mddev->sync_thread) {
-		printk(KERN_ERR "%s: could not start resync thread...\n",
-		       mdname(mddev));
+		pr_warn("%s: could not start resync thread...\n",
+			mdname(mddev));
 		/* leave the spares where they are, it shouldn't hurt */
 		clear_bit(MD_RECOVERY_SYNC, &mddev->recovery);
 		clear_bit(MD_RECOVERY_RESHAPE, &mddev->recovery);
@@ -8356,8 +8378,8 @@ void md_check_recovery(struct mddev *mddev)
 
 	if (signal_pending(current)) {
 		if (mddev->pers->sync_request && !mddev->external) {
-			printk(KERN_INFO "md: %s in immediate safe mode\n",
-			       mdname(mddev));
+			pr_debug("md: %s in immediate safe mode\n",
+				 mdname(mddev));
 			mddev->safemode = 2;
 		}
 		flush_signals(current);
@@ -8366,7 +8388,7 @@ void md_check_recovery(struct mddev *mddev)
 	if (mddev->ro && !test_bit(MD_RECOVERY_NEEDED, &mddev->recovery))
 		return;
 	if ( ! (
-		(mddev->flags & MD_UPDATE_SB_FLAGS & ~ (1<<MD_CHANGE_PENDING)) ||
+		(mddev->sb_flags & ~ (1<<MD_SB_CHANGE_PENDING)) ||
 		test_bit(MD_RECOVERY_NEEDED, &mddev->recovery) ||
 		test_bit(MD_RECOVERY_DONE, &mddev->recovery) ||
 		test_bit(MD_RELOAD_SB, &mddev->flags) ||
@@ -8404,7 +8426,7 @@ void md_check_recovery(struct mddev *mddev)
 			md_reap_sync_thread(mddev);
 			clear_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
 			clear_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
-			clear_bit(MD_CHANGE_PENDING, &mddev->flags);
+			clear_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags);
 			goto unlock;
 		}
 
@@ -8432,7 +8454,7 @@ void md_check_recovery(struct mddev *mddev)
 			    mddev->recovery_cp == MaxSector) {
 				mddev->in_sync = 1;
 				did_change = 1;
-				set_bit(MD_CHANGE_CLEAN, &mddev->flags);
+				set_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags);
 			}
 			if (mddev->safemode == 1)
 				mddev->safemode = 0;
@@ -8441,7 +8463,7 @@ void md_check_recovery(struct mddev *mddev)
 				sysfs_notify_dirent_safe(mddev->sysfs_state);
 		}
 
-		if (mddev->flags & MD_UPDATE_SB_FLAGS)
+		if (mddev->sb_flags)
 			md_update_sb(mddev, 0);
 
 		if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) &&
@@ -8537,7 +8559,7 @@ void md_reap_sync_thread(struct mddev *mddev)
 		if (mddev->pers->spare_active(mddev)) {
 			sysfs_notify(&mddev->kobj, NULL,
 				     "degraded");
-			set_bit(MD_CHANGE_DEVS, &mddev->flags);
+			set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
 		}
 	}
 	if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) &&
@@ -8552,7 +8574,7 @@ void md_reap_sync_thread(struct mddev *mddev)
 			rdev->saved_raid_disk = -1;
 
 	md_update_sb(mddev, 1);
-	/* MD_CHANGE_PENDING should be cleared by md_update_sb, so we can
+	/* MD_SB_CHANGE_PENDING should be cleared by md_update_sb, so we can
 	 * call resync_finish here if MD_CLUSTER_RESYNC_LOCKED is set by
 	 * clustered raid */
 	if (test_and_clear_bit(MD_CLUSTER_RESYNC_LOCKED, &mddev->flags))
@@ -8614,9 +8636,12 @@ int rdev_set_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
 	rv = badblocks_set(&rdev->badblocks, s, sectors, 0);
 	if (rv == 0) {
 		/* Make sure they get written out promptly */
+		if (test_bit(ExternalBbl, &rdev->flags))
+			sysfs_notify(&rdev->kobj, NULL,
+				     "unacknowledged_bad_blocks");
 		sysfs_notify_dirent_safe(rdev->sysfs_state);
-		set_mask_bits(&mddev->flags, 0,
-			      BIT(MD_CHANGE_CLEAN) | BIT(MD_CHANGE_PENDING));
+		set_mask_bits(&mddev->sb_flags, 0,
+			      BIT(MD_SB_CHANGE_CLEAN) | BIT(MD_SB_CHANGE_PENDING));
 		md_wakeup_thread(rdev->mddev->thread);
 		return 1;
 	} else
@@ -8627,12 +8652,15 @@ EXPORT_SYMBOL_GPL(rdev_set_badblocks);
 int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
 			 int is_new)
 {
+	int rv;
 	if (is_new)
 		s += rdev->new_data_offset;
 	else
 		s += rdev->data_offset;
-	return badblocks_clear(&rdev->badblocks,
-				  s, sectors);
+	rv = badblocks_clear(&rdev->badblocks, s, sectors);
+	if ((rv == 0) && test_bit(ExternalBbl, &rdev->flags))
+		sysfs_notify(&rdev->kobj, NULL, "bad_blocks");
+	return rv;
 }
 EXPORT_SYMBOL_GPL(rdev_clear_badblocks);
 
@@ -8749,7 +8777,7 @@ static void check_sb_changes(struct mddev *mddev, struct md_rdev *rdev)
 				rdev2->saved_raid_disk = role;
 				ret = remove_and_add_spares(mddev, rdev2);
 				pr_info("Activated spare: %s\n",
-						bdevname(rdev2->bdev,b));
+					bdevname(rdev2->bdev,b));
 				/* wakeup mddev->thread here, so array could
 				 * perform resync with the new activated disk */
 				set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
@@ -8785,15 +8813,18 @@ static int read_rdev(struct mddev *mddev, struct md_rdev *rdev)
 	 * variable in case we err in the future
 	 */
 	rdev->sb_page = NULL;
-	alloc_disk_sb(rdev);
-	ClearPageUptodate(rdev->sb_page);
-	rdev->sb_loaded = 0;
-	err = super_types[mddev->major_version].load_super(rdev, NULL, mddev->minor_version);
-
+	err = alloc_disk_sb(rdev);
+	if (err == 0) {
+		ClearPageUptodate(rdev->sb_page);
+		rdev->sb_loaded = 0;
+		err = super_types[mddev->major_version].
+			load_super(rdev, NULL, mddev->minor_version);
+	}
 	if (err < 0) {
 		pr_warn("%s: %d Could not reload rdev(%d) err: %d. Restoring old values\n",
 				__func__, __LINE__, rdev->desc_nr, err);
-		put_page(rdev->sb_page);
+		if (rdev->sb_page)
+			put_page(rdev->sb_page);
 		rdev->sb_page = swapout;
 		rdev->sb_loaded = 1;
 		return err;
@@ -8871,9 +8902,6 @@ void md_autodetect_dev(dev_t dev)
 		mutex_lock(&detected_devices_mutex);
 		list_add_tail(&node_detected_dev->list, &all_detected_devices);
 		mutex_unlock(&detected_devices_mutex);
-	} else {
-		printk(KERN_CRIT "md: md_autodetect_dev: kzalloc failed"
-			", skipping dev(%d,%d)\n", MAJOR(dev), MINOR(dev));
 	}
 }
 
@@ -8887,7 +8915,7 @@ static void autostart_arrays(int part)
 	i_scanned = 0;
 	i_passed = 0;
 
-	printk(KERN_INFO "md: Autodetecting RAID arrays.\n");
+	pr_info("md: Autodetecting RAID arrays.\n");
 
 	mutex_lock(&detected_devices_mutex);
 	while (!list_empty(&all_detected_devices) && i_scanned < INT_MAX) {
@@ -8912,8 +8940,7 @@ static void autostart_arrays(int part)
 	}
 	mutex_unlock(&detected_devices_mutex);
 
-	printk(KERN_INFO "md: Scanned %d and added %d devices.\n",
-						i_scanned, i_passed);
+	pr_debug("md: Scanned %d and added %d devices.\n", i_scanned, i_passed);
 
 	autorun_devices(part);
 }

drivers/md/md.h

@@ -29,6 +29,16 @@
 
 #define MaxSector (~(sector_t)0)
 
+/*
+ * These flags should really be called "NO_RETRY" rather than
+ * "FAILFAST" because they don't make any promise about time lapse,
+ * only about the number of retries, which will be zero.
+ * REQ_FAILFAST_DRIVER is not included because
+ * Commit: 4a27446f3e39 ("[SCSI] modify scsi to handle new fail fast flags.")
+ * seems to suggest that the errors it avoids retrying should usually
+ * be retried.
+ */
+#define	MD_FAILFAST	(REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT)
 /*
  * MD's 'extended' device
  */
@@ -168,6 +178,19 @@ enum flag_bits {
 				 * so it is safe to remove without
 				 * another synchronize_rcu() call.
 				 */
+	ExternalBbl,            /* External metadata provides bad
+				 * block management for a disk
+				 */
+	FailFast,		/* Minimal retries should be attempted on
+				 * this device, so use REQ_FAILFAST_DEV.
+				 * Also don't try to repair failed reads.
+				 * It is expects that no bad block log
+				 * is present.
+				 */
+	LastDev,		/* Seems to be the last working dev as
+				 * it didn't fail, so don't use FailFast
+				 * any more for metadata
+				 */
 };
 
 static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors,
@@ -189,6 +212,31 @@ extern int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
 				int is_new);
 struct md_cluster_info;
 
+enum mddev_flags {
+	MD_ARRAY_FIRST_USE,	/* First use of array, needs initialization */
+	MD_CLOSING,		/* If set, we are closing the array, do not open
+				 * it then */
+	MD_JOURNAL_CLEAN,	/* A raid with journal is already clean */
+	MD_HAS_JOURNAL,		/* The raid array has journal feature set */
+	MD_RELOAD_SB,		/* Reload the superblock because another node
+				 * updated it.
+				 */
+	MD_CLUSTER_RESYNC_LOCKED, /* cluster raid only, which means node
+				   * already took resync lock, need to
+				   * release the lock */
+	MD_FAILFAST_SUPPORTED,	/* Using MD_FAILFAST on metadata writes is
+				 * supported as calls to md_error() will
+				 * never cause the array to become failed.
+				 */
+};
+
+enum mddev_sb_flags {
+	MD_SB_CHANGE_DEVS,		/* Some device status has changed */
+	MD_SB_CHANGE_CLEAN,	/* transition to or from 'clean' */
+	MD_SB_CHANGE_PENDING,	/* switch from 'clean' to 'active' in progress */
+	MD_SB_NEED_REWRITE,	/* metadata write needs to be repeated */
+};
+
 struct mddev {
 	void				*private;
 	struct md_personality		*pers;
@@ -196,21 +244,7 @@ struct mddev {
 	int				md_minor;
 	struct list_head		disks;
 	unsigned long			flags;
-#define MD_CHANGE_DEVS	0	/* Some device status has changed */
-#define MD_CHANGE_CLEAN 1	/* transition to or from 'clean' */
-#define MD_CHANGE_PENDING 2	/* switch from 'clean' to 'active' in progress */
-#define MD_UPDATE_SB_FLAGS (1 | 2 | 4)	/* If these are set, md_update_sb needed */
-#define MD_ARRAY_FIRST_USE 3    /* First use of array, needs initialization */
-#define MD_CLOSING	4	/* If set, we are closing the array, do not open
-				 * it then */
-#define MD_JOURNAL_CLEAN 5	/* A raid with journal is already clean */
-#define MD_HAS_JOURNAL	6	/* The raid array has journal feature set */
-#define MD_RELOAD_SB	7	/* Reload the superblock because another node
-				 * updated it.
-				 */
-#define MD_CLUSTER_RESYNC_LOCKED 8 /* cluster raid only, which means node
-				    * already took resync lock, need to
-				    * release the lock */
+	unsigned long			sb_flags;
 
 	int				suspended;
 	atomic_t			active_io;
@@ -304,31 +338,6 @@ struct mddev {
 	int				parallel_resync;
 
 	int				ok_start_degraded;
-	/* recovery/resync flags
-	 * NEEDED:   we might need to start a resync/recover
-	 * RUNNING:  a thread is running, or about to be started
-	 * SYNC:     actually doing a resync, not a recovery
-	 * RECOVER:  doing recovery, or need to try it.
-	 * INTR:     resync needs to be aborted for some reason
-	 * DONE:     thread is done and is waiting to be reaped
-	 * REQUEST:  user-space has requested a sync (used with SYNC)
-	 * CHECK:    user-space request for check-only, no repair
-	 * RESHAPE:  A reshape is happening
-	 * ERROR:    sync-action interrupted because io-error
-	 *
-	 * If neither SYNC or RESHAPE are set, then it is a recovery.
-	 */
-#define	MD_RECOVERY_RUNNING	0
-#define	MD_RECOVERY_SYNC	1
-#define	MD_RECOVERY_RECOVER	2
-#define	MD_RECOVERY_INTR	3
-#define	MD_RECOVERY_DONE	4
-#define	MD_RECOVERY_NEEDED	5
-#define	MD_RECOVERY_REQUESTED	6
-#define	MD_RECOVERY_CHECK	7
-#define MD_RECOVERY_RESHAPE	8
-#define	MD_RECOVERY_FROZEN	9
-#define	MD_RECOVERY_ERROR	10
 
 	unsigned long			recovery;
 	/* If a RAID personality determines that recovery (of a particular
@@ -442,6 +451,23 @@ struct mddev {
 	unsigned int			good_device_nr;	/* good device num within cluster raid */
 };
 
+enum recovery_flags {
+	/*
+	 * If neither SYNC or RESHAPE are set, then it is a recovery.
+	 */
+	MD_RECOVERY_RUNNING,	/* a thread is running, or about to be started */
+	MD_RECOVERY_SYNC,	/* actually doing a resync, not a recovery */
+	MD_RECOVERY_RECOVER,	/* doing recovery, or need to try it. */
+	MD_RECOVERY_INTR,	/* resync needs to be aborted for some reason */
+	MD_RECOVERY_DONE,	/* thread is done and is waiting to be reaped */
+	MD_RECOVERY_NEEDED,	/* we might need to start a resync/recover */
+	MD_RECOVERY_REQUESTED,	/* user-space has requested a sync (used with SYNC) */
+	MD_RECOVERY_CHECK,	/* user-space request for check-only, no repair */
+	MD_RECOVERY_RESHAPE,	/* A reshape is happening */
+	MD_RECOVERY_FROZEN,	/* User request to abort, and not restart, any action */
+	MD_RECOVERY_ERROR,	/* sync-action interrupted because io-error */
+};
+
 static inline int __must_check mddev_lock(struct mddev *mddev)
 {
 	return mutex_lock_interruptible(&mddev->reconfig_mutex);
@@ -623,7 +649,7 @@ extern int mddev_congested(struct mddev *mddev, int bits);
 extern void md_flush_request(struct mddev *mddev, struct bio *bio);
 extern void md_super_write(struct mddev *mddev, struct md_rdev *rdev,
 			   sector_t sector, int size, struct page *page);
-extern void md_super_wait(struct mddev *mddev);
+extern int md_super_wait(struct mddev *mddev);
 extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size,
 			struct page *page, int op, int op_flags,
 			bool metadata_op);

drivers/md/multipath.c

@@ -52,7 +52,7 @@ static int multipath_map (struct mpconf *conf)
 	}
 	rcu_read_unlock();
 
-	printk(KERN_ERR "multipath_map(): no more operational IO paths?\n");
+	pr_crit_ratelimited("multipath_map(): no more operational IO paths?\n");
 	return (-1);
 }
 
@@ -97,9 +97,9 @@ static void multipath_end_request(struct bio *bio)
 		 */
 		char b[BDEVNAME_SIZE];
 		md_error (mp_bh->mddev, rdev);
-		printk(KERN_ERR "multipath: %s: rescheduling sector %llu\n",
-		       bdevname(rdev->bdev,b),
-		       (unsigned long long)bio->bi_iter.bi_sector);
+		pr_info("multipath: %s: rescheduling sector %llu\n",
+			bdevname(rdev->bdev,b),
+			(unsigned long long)bio->bi_iter.bi_sector);
 		multipath_reschedule_retry(mp_bh);
 	} else
 		multipath_end_bh_io(mp_bh, bio->bi_error);
@@ -194,8 +194,7 @@ static void multipath_error (struct mddev *mddev, struct md_rdev *rdev)
 		 * first check if this is a queued request for a device
 		 * which has just failed.
 		 */
-		printk(KERN_ALERT
-		       "multipath: only one IO path left and IO error.\n");
+		pr_warn("multipath: only one IO path left and IO error.\n");
 		/* leave it active... it's all we have */
 		return;
 	}
@@ -209,11 +208,9 @@ static void multipath_error (struct mddev *mddev, struct md_rdev *rdev)
 		spin_unlock_irqrestore(&conf->device_lock, flags);
 	}
 	set_bit(Faulty, &rdev->flags);
-	set_bit(MD_CHANGE_DEVS, &mddev->flags);
-	printk(KERN_ALERT "multipath: IO failure on %s,"
-	       " disabling IO path.\n"
-	       "multipath: Operation continuing"
-	       " on %d IO paths.\n",
+	set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
+	pr_err("multipath: IO failure on %s, disabling IO path.\n"
+	       "multipath: Operation continuing on %d IO paths.\n",
 	       bdevname(rdev->bdev, b),
 	       conf->raid_disks - mddev->degraded);
 }
@@ -223,21 +220,21 @@ static void print_multipath_conf (struct mpconf *conf)
 	int i;
 	struct multipath_info *tmp;
 
-	printk("MULTIPATH conf printout:\n");
+	pr_debug("MULTIPATH conf printout:\n");
 	if (!conf) {
-		printk("(conf==NULL)\n");
+		pr_debug("(conf==NULL)\n");
 		return;
 	}
-	printk(" --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
-			 conf->raid_disks);
+	pr_debug(" --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
+		 conf->raid_disks);
 
 	for (i = 0; i < conf->raid_disks; i++) {
 		char b[BDEVNAME_SIZE];
 		tmp = conf->multipaths + i;
 		if (tmp->rdev)
-			printk(" disk%d, o:%d, dev:%s\n",
-				i,!test_bit(Faulty, &tmp->rdev->flags),
-			       bdevname(tmp->rdev->bdev,b));
+			pr_debug(" disk%d, o:%d, dev:%s\n",
+				 i,!test_bit(Faulty, &tmp->rdev->flags),
+				 bdevname(tmp->rdev->bdev,b));
 	}
 }
 
@@ -292,8 +289,7 @@ static int multipath_remove_disk(struct mddev *mddev, struct md_rdev *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"
-			       " but is still operational!\n", number);
+			pr_warn("hot-remove-disk, slot %d is identified but is still operational!\n", number);
 			err = -EBUSY;
 			goto abort;
 		}
@@ -346,16 +342,14 @@ static void multipathd(struct md_thread *thread)
 		bio->bi_iter.bi_sector = mp_bh->master_bio->bi_iter.bi_sector;
 
 		if ((mp_bh->path = multipath_map (conf))<0) {
-			printk(KERN_ALERT "multipath: %s: unrecoverable IO read"
-				" error for block %llu\n",
-				bdevname(bio->bi_bdev,b),
-				(unsigned long long)bio->bi_iter.bi_sector);
+			pr_err("multipath: %s: unrecoverable IO read error for block %llu\n",
+			       bdevname(bio->bi_bdev,b),
+			       (unsigned long long)bio->bi_iter.bi_sector);
 			multipath_end_bh_io(mp_bh, -EIO);
 		} else {
-			printk(KERN_ERR "multipath: %s: redirecting sector %llu"
-				" to another IO path\n",
-				bdevname(bio->bi_bdev,b),
-				(unsigned long long)bio->bi_iter.bi_sector);
+			pr_err("multipath: %s: redirecting sector %llu to another IO path\n",
+			       bdevname(bio->bi_bdev,b),
+			       (unsigned long long)bio->bi_iter.bi_sector);
 			*bio = *(mp_bh->master_bio);
 			bio->bi_iter.bi_sector +=
 				conf->multipaths[mp_bh->path].rdev->data_offset;
@@ -389,8 +383,8 @@ static int multipath_run (struct mddev *mddev)
 		return -EINVAL;
 
 	if (mddev->level != LEVEL_MULTIPATH) {
-		printk("multipath: %s: raid level not set to multipath IO (%d)\n",
-		       mdname(mddev), mddev->level);
+		pr_warn("multipath: %s: raid level not set to multipath IO (%d)\n",
+			mdname(mddev), mddev->level);
 		goto out;
 	}
 	/*
@@ -401,21 +395,13 @@ static int multipath_run (struct mddev *mddev)
 
 	conf = kzalloc(sizeof(struct mpconf), GFP_KERNEL);
 	mddev->private = conf;
-	if (!conf) {
-		printk(KERN_ERR
-			"multipath: couldn't allocate memory for %s\n",
-			mdname(mddev));
+	if (!conf)
 		goto out;
-	}
 
 	conf->multipaths = kzalloc(sizeof(struct multipath_info)*mddev->raid_disks,
 				   GFP_KERNEL);
-	if (!conf->multipaths) {
-		printk(KERN_ERR
-			"multipath: couldn't allocate memory for %s\n",
-			mdname(mddev));
+	if (!conf->multipaths)
 		goto out_free_conf;
-	}
 
 	working_disks = 0;
 	rdev_for_each(rdev, mddev) {
@@ -439,7 +425,7 @@ static int multipath_run (struct mddev *mddev)
 	INIT_LIST_HEAD(&conf->retry_list);
 
 	if (!working_disks) {
-		printk(KERN_ERR "multipath: no operational IO paths for %s\n",
+		pr_warn("multipath: no operational IO paths for %s\n",
 			mdname(mddev));
 		goto out_free_conf;
 	}
@@ -447,27 +433,17 @@ static int multipath_run (struct mddev *mddev)
 
 	conf->pool = mempool_create_kmalloc_pool(NR_RESERVED_BUFS,
 						 sizeof(struct multipath_bh));
-	if (conf->pool == NULL) {
-		printk(KERN_ERR
-			"multipath: couldn't allocate memory for %s\n",
-			mdname(mddev));
+	if (conf->pool == NULL)
 		goto out_free_conf;
-	}
 
-	{
-		mddev->thread = md_register_thread(multipathd, mddev,
-						   "multipath");
-		if (!mddev->thread) {
-			printk(KERN_ERR "multipath: couldn't allocate thread"
-				" for %s\n", mdname(mddev));
-			goto out_free_conf;
-		}
-	}
+	mddev->thread = md_register_thread(multipathd, mddev,
+					   "multipath");
+	if (!mddev->thread)
+		goto out_free_conf;
 
-	printk(KERN_INFO
-		"multipath: array %s active with %d out of %d IO paths\n",
+	pr_info("multipath: array %s active with %d out of %d IO paths\n",
 		mdname(mddev), conf->raid_disks - mddev->degraded,
-	       mddev->raid_disks);
+		mddev->raid_disks);
 	/*
 	 * Ok, everything is just fine now
 	 */

drivers/md/raid0.c

@@ -21,6 +21,7 @@
 #include <linux/seq_file.h>
 #include <linux/module.h>
 #include <linux/slab.h>
+#include <trace/events/block.h>
 #include "md.h"
 #include "raid0.h"
 #include "raid5.h"
@@ -51,20 +52,21 @@ static void dump_zones(struct mddev *mddev)
 	char b[BDEVNAME_SIZE];
 	struct r0conf *conf = mddev->private;
 	int raid_disks = conf->strip_zone[0].nb_dev;
-	printk(KERN_INFO "md: RAID0 configuration for %s - %d zone%s\n",
-	       mdname(mddev),
-	       conf->nr_strip_zones, conf->nr_strip_zones==1?"":"s");
+	pr_debug("md: RAID0 configuration for %s - %d zone%s\n",
+		 mdname(mddev),
+		 conf->nr_strip_zones, conf->nr_strip_zones==1?"":"s");
 	for (j = 0; j < conf->nr_strip_zones; j++) {
-		printk(KERN_INFO "md: zone%d=[", j);
+		char line[200];
+		int len = 0;
+
 		for (k = 0; k < conf->strip_zone[j].nb_dev; k++)
-			printk(KERN_CONT "%s%s", k?"/":"",
-			bdevname(conf->devlist[j*raid_disks
-						+ k]->bdev, b));
-		printk(KERN_CONT "]\n");
+			len += snprintf(line+len, 200-len, "%s%s", k?"/":"",
+					bdevname(conf->devlist[j*raid_disks
+							       + k]->bdev, b));
+		pr_debug("md: zone%d=[%s]\n", j, line);
 
 		zone_size  = conf->strip_zone[j].zone_end - zone_start;
-		printk(KERN_INFO "      zone-offset=%10lluKB, "
-				"device-offset=%10lluKB, size=%10lluKB\n",
+		pr_debug("      zone-offset=%10lluKB, device-offset=%10lluKB, size=%10lluKB\n",
 			(unsigned long long)zone_start>>1,
 			(unsigned long long)conf->strip_zone[j].dev_start>>1,
 			(unsigned long long)zone_size>>1);
@@ -142,9 +144,9 @@ static int create_strip_zones(struct mddev *mddev, struct r0conf **private_conf)
 	 * chunk size is a multiple of that sector size
 	 */
 	if ((mddev->chunk_sectors << 9) % blksize) {
-		printk(KERN_ERR "md/raid0:%s: chunk_size of %d not multiple of block size %d\n",
-		       mdname(mddev),
-		       mddev->chunk_sectors << 9, blksize);
+		pr_warn("md/raid0:%s: chunk_size of %d not multiple of block size %d\n",
+			mdname(mddev),
+			mddev->chunk_sectors << 9, blksize);
 		err = -EINVAL;
 		goto abort;
 	}
@@ -186,19 +188,18 @@ static int create_strip_zones(struct mddev *mddev, struct r0conf **private_conf)
 		}
 
 		if (j < 0) {
-			printk(KERN_ERR
-			       "md/raid0:%s: remove inactive devices before converting to RAID0\n",
-			       mdname(mddev));
+			pr_warn("md/raid0:%s: remove inactive devices before converting to RAID0\n",
+				mdname(mddev));
 			goto abort;
 		}
 		if (j >= mddev->raid_disks) {
-			printk(KERN_ERR "md/raid0:%s: bad disk number %d - "
-			       "aborting!\n", mdname(mddev), j);
+			pr_warn("md/raid0:%s: bad disk number %d - aborting!\n",
+				mdname(mddev), j);
 			goto abort;
 		}
 		if (dev[j]) {
-			printk(KERN_ERR "md/raid0:%s: multiple devices for %d - "
-			       "aborting!\n", mdname(mddev), j);
+			pr_warn("md/raid0:%s: multiple devices for %d - aborting!\n",
+				mdname(mddev), j);
 			goto abort;
 		}
 		dev[j] = rdev1;
@@ -208,8 +209,8 @@ static int create_strip_zones(struct mddev *mddev, struct r0conf **private_conf)
 		cnt++;
 	}
 	if (cnt != mddev->raid_disks) {
-		printk(KERN_ERR "md/raid0:%s: too few disks (%d of %d) - "
-		       "aborting!\n", mdname(mddev), cnt, mddev->raid_disks);
+		pr_warn("md/raid0:%s: too few disks (%d of %d) - aborting!\n",
+			mdname(mddev), cnt, mddev->raid_disks);
 		goto abort;
 	}
 	zone->nb_dev = cnt;
@@ -357,8 +358,7 @@ static int raid0_run(struct mddev *mddev)
 	int ret;
 
 	if (mddev->chunk_sectors == 0) {
-		printk(KERN_ERR "md/raid0:%s: chunk size must be set.\n",
-		       mdname(mddev));
+		pr_warn("md/raid0:%s: chunk size must be set.\n", mdname(mddev));
 		return -EINVAL;
 	}
 	if (md_check_no_bitmap(mddev))
@@ -399,9 +399,9 @@ static int raid0_run(struct mddev *mddev)
 	/* calculate array device size */
 	md_set_array_sectors(mddev, raid0_size(mddev, 0, 0));
 
-	printk(KERN_INFO "md/raid0:%s: md_size is %llu sectors.\n",
-	       mdname(mddev),
-	       (unsigned long long)mddev->array_sectors);
+	pr_debug("md/raid0:%s: md_size is %llu sectors.\n",
+		 mdname(mddev),
+		 (unsigned long long)mddev->array_sectors);
 
 	if (mddev->queue) {
 		/* calculate the max read-ahead size.
@@ -464,7 +464,8 @@ static void raid0_make_request(struct mddev *mddev, struct bio *bio)
 	}
 
 	do {
-		sector_t sector = bio->bi_iter.bi_sector;
+		sector_t bio_sector = bio->bi_iter.bi_sector;
+		sector_t sector = bio_sector;
 		unsigned chunk_sects = mddev->chunk_sectors;
 
 		unsigned sectors = chunk_sects -
@@ -473,7 +474,7 @@ static void raid0_make_request(struct mddev *mddev, struct bio *bio)
 			 : sector_div(sector, chunk_sects));
 
 		/* Restore due to sector_div */
-		sector = bio->bi_iter.bi_sector;
+		sector = bio_sector;
 
 		if (sectors < bio_sectors(bio)) {
 			split = bio_split(bio, sectors, GFP_NOIO, fs_bio_set);
@@ -492,8 +493,13 @@ static void raid0_make_request(struct mddev *mddev, struct bio *bio)
 			 !blk_queue_discard(bdev_get_queue(split->bi_bdev)))) {
 			/* Just ignore it */
 			bio_endio(split);
-		} else
+		} else {
+			if (mddev->gendisk)
+				trace_block_bio_remap(bdev_get_queue(split->bi_bdev),
+						      split, disk_devt(mddev->gendisk),
+						      bio_sector);
 			generic_make_request(split);
+		}
 	} while (split != bio);
 }
 
@@ -509,17 +515,17 @@ static void *raid0_takeover_raid45(struct mddev *mddev)
 	struct r0conf *priv_conf;
 
 	if (mddev->degraded != 1) {
-		printk(KERN_ERR "md/raid0:%s: raid5 must be degraded! Degraded disks: %d\n",
-		       mdname(mddev),
-		       mddev->degraded);
+		pr_warn("md/raid0:%s: raid5 must be degraded! Degraded disks: %d\n",
+			mdname(mddev),
+			mddev->degraded);
 		return ERR_PTR(-EINVAL);
 	}
 
 	rdev_for_each(rdev, mddev) {
 		/* check slot number for a disk */
 		if (rdev->raid_disk == mddev->raid_disks-1) {
-			printk(KERN_ERR "md/raid0:%s: raid5 must have missing parity disk!\n",
-			       mdname(mddev));
+			pr_warn("md/raid0:%s: raid5 must have missing parity disk!\n",
+				mdname(mddev));
 			return ERR_PTR(-EINVAL);
 		}
 		rdev->sectors = mddev->dev_sectors;
@@ -533,8 +539,11 @@ static void *raid0_takeover_raid45(struct mddev *mddev)
 	mddev->delta_disks = -1;
 	/* make sure it will be not marked as dirty */
 	mddev->recovery_cp = MaxSector;
+	clear_bit(MD_HAS_JOURNAL, &mddev->flags);
+	clear_bit(MD_JOURNAL_CLEAN, &mddev->flags);
 
 	create_strip_zones(mddev, &priv_conf);
+
 	return priv_conf;
 }
 
@@ -549,19 +558,19 @@ static void *raid0_takeover_raid10(struct mddev *mddev)
 	 *  - all mirrors must be already degraded
 	 */
 	if (mddev->layout != ((1 << 8) + 2)) {
-		printk(KERN_ERR "md/raid0:%s:: Raid0 cannot takeover layout: 0x%x\n",
-		       mdname(mddev),
-		       mddev->layout);
+		pr_warn("md/raid0:%s:: Raid0 cannot takeover layout: 0x%x\n",
+			mdname(mddev),
+			mddev->layout);
 		return ERR_PTR(-EINVAL);
 	}
 	if (mddev->raid_disks & 1) {
-		printk(KERN_ERR "md/raid0:%s: Raid0 cannot takeover Raid10 with odd disk number.\n",
-		       mdname(mddev));
+		pr_warn("md/raid0:%s: Raid0 cannot takeover Raid10 with odd disk number.\n",
+			mdname(mddev));
 		return ERR_PTR(-EINVAL);
 	}
 	if (mddev->degraded != (mddev->raid_disks>>1)) {
-		printk(KERN_ERR "md/raid0:%s: All mirrors must be already degraded!\n",
-		       mdname(mddev));
+		pr_warn("md/raid0:%s: All mirrors must be already degraded!\n",
+			mdname(mddev));
 		return ERR_PTR(-EINVAL);
 	}
 
@@ -574,6 +583,7 @@ static void *raid0_takeover_raid10(struct mddev *mddev)
 	mddev->degraded = 0;
 	/* make sure it will be not marked as dirty */
 	mddev->recovery_cp = MaxSector;
+	clear_bit(MD_FAILFAST_SUPPORTED, &mddev->flags);
 
 	create_strip_zones(mddev, &priv_conf);
 	return priv_conf;
@@ -588,7 +598,7 @@ static void *raid0_takeover_raid1(struct mddev *mddev)
 	 *  - (N - 1) mirror drives must be already faulty
 	 */
 	if ((mddev->raid_disks - 1) != mddev->degraded) {
-		printk(KERN_ERR "md/raid0:%s: (N - 1) mirrors drives must be already faulty!\n",
+		pr_err("md/raid0:%s: (N - 1) mirrors drives must be already faulty!\n",
 		       mdname(mddev));
 		return ERR_PTR(-EINVAL);
 	}
@@ -616,6 +626,7 @@ static void *raid0_takeover_raid1(struct mddev *mddev)
 	mddev->raid_disks = 1;
 	/* make sure it will be not marked as dirty */
 	mddev->recovery_cp = MaxSector;
+	clear_bit(MD_FAILFAST_SUPPORTED, &mddev->flags);
 
 	create_strip_zones(mddev, &priv_conf);
 	return priv_conf;
@@ -631,8 +642,8 @@ static void *raid0_takeover(struct mddev *mddev)
 	 */
 
 	if (mddev->bitmap) {
-		printk(KERN_ERR "md/raid0: %s: cannot takeover array with bitmap\n",
-		       mdname(mddev));
+		pr_warn("md/raid0: %s: cannot takeover array with bitmap\n",
+			mdname(mddev));
 		return ERR_PTR(-EBUSY);
 	}
 	if (mddev->level == 4)
@@ -642,8 +653,8 @@ static void *raid0_takeover(struct mddev *mddev)
 		if (mddev->layout == ALGORITHM_PARITY_N)
 			return raid0_takeover_raid45(mddev);
 
-		printk(KERN_ERR "md/raid0:%s: Raid can only takeover Raid5 with layout: %d\n",
-		       mdname(mddev), ALGORITHM_PARITY_N);
+		pr_warn("md/raid0:%s: Raid can only takeover Raid5 with layout: %d\n",
+			mdname(mddev), ALGORITHM_PARITY_N);
 	}
 
 	if (mddev->level == 10)
@@ -652,7 +663,7 @@ static void *raid0_takeover(struct mddev *mddev)
 	if (mddev->level == 1)
 		return raid0_takeover_raid1(mddev);
 
-	printk(KERN_ERR "Takeover from raid%i to raid0 not supported\n",
+	pr_warn("Takeover from raid%i to raid0 not supported\n",
 		mddev->level);
 
 	return ERR_PTR(-EINVAL);

drivers/md/raid1.c

@@ -37,6 +37,7 @@
 #include <linux/module.h>
 #include <linux/seq_file.h>
 #include <linux/ratelimit.h>
+#include <trace/events/block.h>
 #include "md.h"
 #include "raid1.h"
 #include "bitmap.h"
@@ -70,6 +71,9 @@ static void allow_barrier(struct r1conf *conf, sector_t start_next_window,
 			  sector_t bi_sector);
 static void lower_barrier(struct r1conf *conf);
 
+#define raid1_log(md, fmt, args...)				\
+	do { if ((md)->queue) blk_add_trace_msg((md)->queue, "raid1 " fmt, ##args); } while (0)
+
 static void * r1bio_pool_alloc(gfp_t gfp_flags, void *data)
 {
 	struct pool_info *pi = data;
@@ -325,6 +329,11 @@ static void raid1_end_read_request(struct bio *bio)
 
 	if (uptodate)
 		set_bit(R1BIO_Uptodate, &r1_bio->state);
+	else if (test_bit(FailFast, &rdev->flags) &&
+		 test_bit(R1BIO_FailFast, &r1_bio->state))
+		/* This was a fail-fast read so we definitely
+		 * want to retry */
+		;
 	else {
 		/* If all other devices have failed, we want to return
 		 * the error upwards rather than fail the last device.
@@ -347,13 +356,10 @@ static void raid1_end_read_request(struct bio *bio)
 		 * oops, read error:
 		 */
 		char b[BDEVNAME_SIZE];
-		printk_ratelimited(
-			KERN_ERR "md/raid1:%s: %s: "
-			"rescheduling sector %llu\n",
-			mdname(conf->mddev),
-			bdevname(rdev->bdev,
-				 b),
-			(unsigned long long)r1_bio->sector);
+		pr_err_ratelimited("md/raid1:%s: %s: rescheduling sector %llu\n",
+				   mdname(conf->mddev),
+				   bdevname(rdev->bdev, b),
+				   (unsigned long long)r1_bio->sector);
 		set_bit(R1BIO_ReadError, &r1_bio->state);
 		reschedule_retry(r1_bio);
 		/* don't drop the reference on read_disk yet */
@@ -416,7 +422,24 @@ static void raid1_end_write_request(struct bio *bio)
 			set_bit(MD_RECOVERY_NEEDED, &
 				conf->mddev->recovery);
 
-		set_bit(R1BIO_WriteError, &r1_bio->state);
+		if (test_bit(FailFast, &rdev->flags) &&
+		    (bio->bi_opf & MD_FAILFAST) &&
+		    /* We never try FailFast to WriteMostly devices */
+		    !test_bit(WriteMostly, &rdev->flags)) {
+			md_error(r1_bio->mddev, rdev);
+			if (!test_bit(Faulty, &rdev->flags))
+				/* This is the only remaining device,
+				 * We need to retry the write without
+				 * FailFast
+				 */
+				set_bit(R1BIO_WriteError, &r1_bio->state);
+			else {
+				/* Finished with this branch */
+				r1_bio->bios[mirror] = NULL;
+				to_put = bio;
+			}
+		} else
+			set_bit(R1BIO_WriteError, &r1_bio->state);
 	} else {
 		/*
 		 * Set R1BIO_Uptodate in our master bio, so that we
@@ -534,6 +557,7 @@ static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sect
 	best_good_sectors = 0;
 	has_nonrot_disk = 0;
 	choose_next_idle = 0;
+	clear_bit(R1BIO_FailFast, &r1_bio->state);
 
 	if ((conf->mddev->recovery_cp < this_sector + sectors) ||
 	    (mddev_is_clustered(conf->mddev) &&
@@ -607,6 +631,10 @@ static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sect
 		} else
 			best_good_sectors = sectors;
 
+		if (best_disk >= 0)
+			/* At least two disks to choose from so failfast is OK */
+			set_bit(R1BIO_FailFast, &r1_bio->state);
+
 		nonrot = blk_queue_nonrot(bdev_get_queue(rdev->bdev));
 		has_nonrot_disk |= nonrot;
 		pending = atomic_read(&rdev->nr_pending);
@@ -645,11 +673,6 @@ static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sect
 			}
 			break;
 		}
-		/* If device is idle, use it */
-		if (pending == 0) {
-			best_disk = disk;
-			break;
-		}
 
 		if (choose_next_idle)
 			continue;
@@ -672,7 +695,7 @@ static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sect
 	 * mixed ratation/non-rotational disks depending on workload.
 	 */
 	if (best_disk == -1) {
-		if (has_nonrot_disk)
+		if (has_nonrot_disk || min_pending == 0)
 			best_disk = best_pending_disk;
 		else
 			best_disk = best_dist_disk;
@@ -745,9 +768,14 @@ static void flush_pending_writes(struct r1conf *conf)
 
 		while (bio) { /* submit pending writes */
 			struct bio *next = bio->bi_next;
+			struct md_rdev *rdev = (void*)bio->bi_bdev;
 			bio->bi_next = NULL;
-			if (unlikely((bio_op(bio) == REQ_OP_DISCARD) &&
-			    !blk_queue_discard(bdev_get_queue(bio->bi_bdev))))
+			bio->bi_bdev = rdev->bdev;
+			if (test_bit(Faulty, &rdev->flags)) {
+				bio->bi_error = -EIO;
+				bio_endio(bio);
+			} else if (unlikely((bio_op(bio) == REQ_OP_DISCARD) &&
+					    !blk_queue_discard(bdev_get_queue(bio->bi_bdev))))
 				/* Just ignore it */
 				bio_endio(bio);
 			else
@@ -832,7 +860,7 @@ static bool need_to_wait_for_sync(struct r1conf *conf, struct bio *bio)
 	else if (conf->barrier && bio_data_dir(bio) == WRITE) {
 		if ((conf->mddev->curr_resync_completed
 		     >= bio_end_sector(bio)) ||
-		    (conf->next_resync + NEXT_NORMALIO_DISTANCE
+		    (conf->start_next_window + NEXT_NORMALIO_DISTANCE
 		     <= bio->bi_iter.bi_sector))
 			wait = false;
 		else
@@ -858,6 +886,7 @@ static sector_t wait_barrier(struct r1conf *conf, struct bio *bio)
 		 * that queue to allow conf->start_next_window
 		 * to increase.
 		 */
+		raid1_log(conf->mddev, "wait barrier");
 		wait_event_lock_irq(conf->wait_barrier,
 				    !conf->array_frozen &&
 				    (!conf->barrier ||
@@ -937,6 +966,7 @@ static void freeze_array(struct r1conf *conf, int extra)
 	 */
 	spin_lock_irq(&conf->resync_lock);
 	conf->array_frozen = 1;
+	raid1_log(conf->mddev, "wait freeze");
 	wait_event_lock_irq_cmd(conf->wait_barrier,
 				conf->nr_pending == conf->nr_queued+extra,
 				conf->resync_lock,
@@ -1019,9 +1049,14 @@ static void raid1_unplug(struct blk_plug_cb *cb, bool from_schedule)
 
 	while (bio) { /* submit pending writes */
 		struct bio *next = bio->bi_next;
+		struct md_rdev *rdev = (void*)bio->bi_bdev;
 		bio->bi_next = NULL;
-		if (unlikely((bio_op(bio) == REQ_OP_DISCARD) &&
-		    !blk_queue_discard(bdev_get_queue(bio->bi_bdev))))
+		bio->bi_bdev = rdev->bdev;
+		if (test_bit(Faulty, &rdev->flags)) {
+			bio->bi_error = -EIO;
+			bio_endio(bio);
+		} else if (unlikely((bio_op(bio) == REQ_OP_DISCARD) &&
+				    !blk_queue_discard(bdev_get_queue(bio->bi_bdev))))
 			/* Just ignore it */
 			bio_endio(bio);
 		else
@@ -1136,6 +1171,7 @@ read_again:
 			 * take care not to over-take any writes
 			 * that are 'behind'
 			 */
+			raid1_log(mddev, "wait behind writes");
 			wait_event(bitmap->behind_wait,
 				   atomic_read(&bitmap->behind_writes) == 0);
 		}
@@ -1153,8 +1189,16 @@ read_again:
 		read_bio->bi_bdev = mirror->rdev->bdev;
 		read_bio->bi_end_io = raid1_end_read_request;
 		bio_set_op_attrs(read_bio, op, do_sync);
+		if (test_bit(FailFast, &mirror->rdev->flags) &&
+		    test_bit(R1BIO_FailFast, &r1_bio->state))
+			read_bio->bi_opf |= MD_FAILFAST;
 		read_bio->bi_private = r1_bio;
 
+		if (mddev->gendisk)
+			trace_block_bio_remap(bdev_get_queue(read_bio->bi_bdev),
+					      read_bio, disk_devt(mddev->gendisk),
+					      r1_bio->sector);
+
 		if (max_sectors < r1_bio->sectors) {
 			/* could not read all from this device, so we will
 			 * need another r1_bio.
@@ -1195,6 +1239,7 @@ read_again:
 	 */
 	if (conf->pending_count >= max_queued_requests) {
 		md_wakeup_thread(mddev->thread);
+		raid1_log(mddev, "wait queued");
 		wait_event(conf->wait_barrier,
 			   conf->pending_count < max_queued_requests);
 	}
@@ -1286,6 +1331,7 @@ read_again:
 				rdev_dec_pending(conf->mirrors[j].rdev, mddev);
 		r1_bio->state = 0;
 		allow_barrier(conf, start_next_window, bio->bi_iter.bi_sector);
+		raid1_log(mddev, "wait rdev %d blocked", blocked_rdev->raid_disk);
 		md_wait_for_blocked_rdev(blocked_rdev, mddev);
 		start_next_window = wait_barrier(conf, bio);
 		/*
@@ -1363,10 +1409,21 @@ read_again:
 		mbio->bi_bdev = conf->mirrors[i].rdev->bdev;
 		mbio->bi_end_io	= raid1_end_write_request;
 		bio_set_op_attrs(mbio, op, do_flush_fua | do_sync);
+		if (test_bit(FailFast, &conf->mirrors[i].rdev->flags) &&
+		    !test_bit(WriteMostly, &conf->mirrors[i].rdev->flags) &&
+		    conf->raid_disks - mddev->degraded > 1)
+			mbio->bi_opf |= MD_FAILFAST;
 		mbio->bi_private = r1_bio;
 
 		atomic_inc(&r1_bio->remaining);
 
+		if (mddev->gendisk)
+			trace_block_bio_remap(bdev_get_queue(mbio->bi_bdev),
+					      mbio, disk_devt(mddev->gendisk),
+					      r1_bio->sector);
+		/* flush_pending_writes() needs access to the rdev so...*/
+		mbio->bi_bdev = (void*)conf->mirrors[i].rdev;
+
 		cb = blk_check_plugged(raid1_unplug, mddev, sizeof(*plug));
 		if (cb)
 			plug = container_of(cb, struct raid1_plug_cb, cb);
@@ -1436,6 +1493,7 @@ static void raid1_error(struct mddev *mddev, struct md_rdev *rdev)
 	 * next level up know.
 	 * else mark the drive as failed
 	 */
+	spin_lock_irqsave(&conf->device_lock, flags);
 	if (test_bit(In_sync, &rdev->flags)
 	    && (conf->raid_disks - mddev->degraded) == 1) {
 		/*
@@ -1445,10 +1503,10 @@ static void raid1_error(struct mddev *mddev, struct md_rdev *rdev)
 		 * it is very likely to fail.
 		 */
 		conf->recovery_disabled = mddev->recovery_disabled;
+		spin_unlock_irqrestore(&conf->device_lock, flags);
 		return;
 	}
 	set_bit(Blocked, &rdev->flags);
-	spin_lock_irqsave(&conf->device_lock, flags);
 	if (test_and_clear_bit(In_sync, &rdev->flags)) {
 		mddev->degraded++;
 		set_bit(Faulty, &rdev->flags);
@@ -1459,36 +1517,35 @@ static void raid1_error(struct mddev *mddev, struct md_rdev *rdev)
 	 * if recovery is running, make sure it aborts.
 	 */
 	set_bit(MD_RECOVERY_INTR, &mddev->recovery);
-	set_mask_bits(&mddev->flags, 0,
-		      BIT(MD_CHANGE_DEVS) | BIT(MD_CHANGE_PENDING));
-	printk(KERN_ALERT
-	       "md/raid1:%s: Disk failure on %s, disabling device.\n"
-	       "md/raid1:%s: Operation continuing on %d devices.\n",
-	       mdname(mddev), bdevname(rdev->bdev, b),
-	       mdname(mddev), conf->raid_disks - mddev->degraded);
+	set_mask_bits(&mddev->sb_flags, 0,
+		      BIT(MD_SB_CHANGE_DEVS) | BIT(MD_SB_CHANGE_PENDING));
+	pr_crit("md/raid1:%s: Disk failure on %s, disabling device.\n"
+		"md/raid1:%s: Operation continuing on %d devices.\n",
+		mdname(mddev), bdevname(rdev->bdev, b),
+		mdname(mddev), conf->raid_disks - mddev->degraded);
 }
 
 static void print_conf(struct r1conf *conf)
 {
 	int i;
 
-	printk(KERN_DEBUG "RAID1 conf printout:\n");
+	pr_debug("RAID1 conf printout:\n");
 	if (!conf) {
-		printk(KERN_DEBUG "(!conf)\n");
+		pr_debug("(!conf)\n");
 		return;
 	}
-	printk(KERN_DEBUG " --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
-		conf->raid_disks);
+	pr_debug(" --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
+		 conf->raid_disks);
 
 	rcu_read_lock();
 	for (i = 0; i < conf->raid_disks; i++) {
 		char b[BDEVNAME_SIZE];
 		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
 		if (rdev)
-			printk(KERN_DEBUG " disk %d, wo:%d, o:%d, dev:%s\n",
-			       i, !test_bit(In_sync, &rdev->flags),
-			       !test_bit(Faulty, &rdev->flags),
-			       bdevname(rdev->bdev,b));
+			pr_debug(" disk %d, wo:%d, o:%d, dev:%s\n",
+				 i, !test_bit(In_sync, &rdev->flags),
+				 !test_bit(Faulty, &rdev->flags),
+				 bdevname(rdev->bdev,b));
 	}
 	rcu_read_unlock();
 }
@@ -1788,12 +1845,24 @@ static int fix_sync_read_error(struct r1bio *r1_bio)
 	sector_t sect = r1_bio->sector;
 	int sectors = r1_bio->sectors;
 	int idx = 0;
+	struct md_rdev *rdev;
+
+	rdev = conf->mirrors[r1_bio->read_disk].rdev;
+	if (test_bit(FailFast, &rdev->flags)) {
+		/* Don't try recovering from here - just fail it
+		 * ... unless it is the last working device of course */
+		md_error(mddev, rdev);
+		if (test_bit(Faulty, &rdev->flags))
+			/* Don't try to read from here, but make sure
+			 * put_buf does it's thing
+			 */
+			bio->bi_end_io = end_sync_write;
+	}
 
 	while(sectors) {
 		int s = sectors;
 		int d = r1_bio->read_disk;
 		int success = 0;
-		struct md_rdev *rdev;
 		int start;
 
 		if (s > (PAGE_SIZE>>9))
@@ -1825,11 +1894,10 @@ static int fix_sync_read_error(struct r1bio *r1_bio)
 			 * work just disable and interrupt the recovery.
 			 * Don't fail devices as that won't really help.
 			 */
-			printk(KERN_ALERT "md/raid1:%s: %s: unrecoverable I/O read error"
-			       " for block %llu\n",
-			       mdname(mddev),
-			       bdevname(bio->bi_bdev, b),
-			       (unsigned long long)r1_bio->sector);
+			pr_crit_ratelimited("md/raid1:%s: %s: unrecoverable I/O read error for block %llu\n",
+					    mdname(mddev),
+					    bdevname(bio->bi_bdev, b),
+					    (unsigned long long)r1_bio->sector);
 			for (d = 0; d < conf->raid_disks * 2; d++) {
 				rdev = conf->mirrors[d].rdev;
 				if (!rdev || test_bit(Faulty, &rdev->flags))
@@ -2013,6 +2081,9 @@ static void sync_request_write(struct mddev *mddev, struct r1bio *r1_bio)
 			continue;
 
 		bio_set_op_attrs(wbio, REQ_OP_WRITE, 0);
+		if (test_bit(FailFast, &conf->mirrors[i].rdev->flags))
+			wbio->bi_opf |= MD_FAILFAST;
+
 		wbio->bi_end_io = end_sync_write;
 		atomic_inc(&r1_bio->remaining);
 		md_sync_acct(conf->mirrors[i].rdev->bdev, bio_sectors(wbio));
@@ -2122,13 +2193,11 @@ static void fix_read_error(struct r1conf *conf, int read_disk,
 				if (r1_sync_page_io(rdev, sect, s,
 						    conf->tmppage, READ)) {
 					atomic_add(s, &rdev->corrected_errors);
-					printk(KERN_INFO
-					       "md/raid1:%s: read error corrected "
-					       "(%d sectors at %llu on %s)\n",
-					       mdname(mddev), s,
-					       (unsigned long long)(sect +
-								    rdev->data_offset),
-					       bdevname(rdev->bdev, b));
+					pr_info("md/raid1:%s: read error corrected (%d sectors at %llu on %s)\n",
+						mdname(mddev), s,
+						(unsigned long long)(sect +
+								     rdev->data_offset),
+						bdevname(rdev->bdev, b));
 				}
 				rdev_dec_pending(rdev, mddev);
 			} else
@@ -2287,6 +2356,8 @@ static void handle_read_error(struct r1conf *conf, struct r1bio *r1_bio)
 	struct bio *bio;
 	char b[BDEVNAME_SIZE];
 	struct md_rdev *rdev;
+	dev_t bio_dev;
+	sector_t bio_sector;
 
 	clear_bit(R1BIO_ReadError, &r1_bio->state);
 	/* we got a read error. Maybe the drive is bad.  Maybe just
@@ -2300,10 +2371,14 @@ static void handle_read_error(struct r1conf *conf, struct r1bio *r1_bio)
 
 	bio = r1_bio->bios[r1_bio->read_disk];
 	bdevname(bio->bi_bdev, b);
+	bio_dev = bio->bi_bdev->bd_dev;
+	bio_sector = conf->mirrors[r1_bio->read_disk].rdev->data_offset + r1_bio->sector;
 	bio_put(bio);
 	r1_bio->bios[r1_bio->read_disk] = NULL;
 
-	if (mddev->ro == 0) {
+	rdev = conf->mirrors[r1_bio->read_disk].rdev;
+	if (mddev->ro == 0
+	    && !test_bit(FailFast, &rdev->flags)) {
 		freeze_array(conf, 1);
 		fix_read_error(conf, r1_bio->read_disk,
 			       r1_bio->sector, r1_bio->sectors);
@@ -2312,14 +2387,13 @@ static void handle_read_error(struct r1conf *conf, struct r1bio *r1_bio)
 		r1_bio->bios[r1_bio->read_disk] = IO_BLOCKED;
 	}
 
-	rdev_dec_pending(conf->mirrors[r1_bio->read_disk].rdev, conf->mddev);
+	rdev_dec_pending(rdev, conf->mddev);
 
 read_more:
 	disk = read_balance(conf, r1_bio, &max_sectors);
 	if (disk == -1) {
-		printk(KERN_ALERT "md/raid1:%s: %s: unrecoverable I/O"
-		       " read error for block %llu\n",
-		       mdname(mddev), b, (unsigned long long)r1_bio->sector);
+		pr_crit_ratelimited("md/raid1:%s: %s: unrecoverable I/O read error for block %llu\n",
+				    mdname(mddev), b, (unsigned long long)r1_bio->sector);
 		raid_end_bio_io(r1_bio);
 	} else {
 		const unsigned long do_sync
@@ -2330,16 +2404,17 @@ read_more:
 			 max_sectors);
 		r1_bio->bios[r1_bio->read_disk] = bio;
 		rdev = conf->mirrors[disk].rdev;
-		printk_ratelimited(KERN_ERR
-				   "md/raid1:%s: redirecting sector %llu"
-				   " to other mirror: %s\n",
-				   mdname(mddev),
-				   (unsigned long long)r1_bio->sector,
-				   bdevname(rdev->bdev, b));
+		pr_info_ratelimited("md/raid1:%s: redirecting sector %llu to other mirror: %s\n",
+				    mdname(mddev),
+				    (unsigned long long)r1_bio->sector,
+				    bdevname(rdev->bdev, b));
 		bio->bi_iter.bi_sector = r1_bio->sector + rdev->data_offset;
 		bio->bi_bdev = rdev->bdev;
 		bio->bi_end_io = raid1_end_read_request;
 		bio_set_op_attrs(bio, REQ_OP_READ, do_sync);
+		if (test_bit(FailFast, &rdev->flags) &&
+		    test_bit(R1BIO_FailFast, &r1_bio->state))
+			bio->bi_opf |= MD_FAILFAST;
 		bio->bi_private = r1_bio;
 		if (max_sectors < r1_bio->sectors) {
 			/* Drat - have to split this up more */
@@ -2353,6 +2428,8 @@ read_more:
 			else
 				mbio->bi_phys_segments++;
 			spin_unlock_irq(&conf->device_lock);
+			trace_block_bio_remap(bdev_get_queue(bio->bi_bdev),
+					      bio, bio_dev, bio_sector);
 			generic_make_request(bio);
 			bio = NULL;
 
@@ -2367,8 +2444,11 @@ read_more:
 				sectors_handled;
 
 			goto read_more;
-		} else
+		} else {
+			trace_block_bio_remap(bdev_get_queue(bio->bi_bdev),
+					      bio, bio_dev, bio_sector);
 			generic_make_request(bio);
+		}
 	}
 }
 
@@ -2384,10 +2464,10 @@ static void raid1d(struct md_thread *thread)
 	md_check_recovery(mddev);
 
 	if (!list_empty_careful(&conf->bio_end_io_list) &&
-	    !test_bit(MD_CHANGE_PENDING, &mddev->flags)) {
+	    !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags)) {
 		LIST_HEAD(tmp);
 		spin_lock_irqsave(&conf->device_lock, flags);
-		if (!test_bit(MD_CHANGE_PENDING, &mddev->flags)) {
+		if (!test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags)) {
 			while (!list_empty(&conf->bio_end_io_list)) {
 				list_move(conf->bio_end_io_list.prev, &tmp);
 				conf->nr_queued--;
@@ -2441,7 +2521,7 @@ static void raid1d(struct md_thread *thread)
 			generic_make_request(r1_bio->bios[r1_bio->read_disk]);
 
 		cond_resched();
-		if (mddev->flags & ~(1<<MD_CHANGE_PENDING))
+		if (mddev->sb_flags & ~(1<<MD_SB_CHANGE_PENDING))
 			md_check_recovery(mddev);
 	}
 	blk_finish_plug(&plug);
@@ -2623,6 +2703,8 @@ static sector_t raid1_sync_request(struct mddev *mddev, sector_t sector_nr,
 			bio->bi_iter.bi_sector = sector_nr + rdev->data_offset;
 			bio->bi_bdev = rdev->bdev;
 			bio->bi_private = r1_bio;
+			if (test_bit(FailFast, &rdev->flags))
+				bio->bi_opf |= MD_FAILFAST;
 		}
 	}
 	rcu_read_unlock();
@@ -2642,7 +2724,7 @@ static sector_t raid1_sync_request(struct mddev *mddev, sector_t sector_nr,
 							min_bad, 0
 					) && ok;
 			}
-		set_bit(MD_CHANGE_DEVS, &mddev->flags);
+		set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
 		*skipped = 1;
 		put_buf(r1_bio);
 
@@ -2753,6 +2835,8 @@ static sector_t raid1_sync_request(struct mddev *mddev, sector_t sector_nr,
 			if (bio->bi_end_io == end_sync_read) {
 				read_targets--;
 				md_sync_acct(bio->bi_bdev, nr_sectors);
+				if (read_targets == 1)
+					bio->bi_opf &= ~MD_FAILFAST;
 				generic_make_request(bio);
 			}
 		}
@@ -2760,6 +2844,8 @@ static sector_t raid1_sync_request(struct mddev *mddev, sector_t sector_nr,
 		atomic_set(&r1_bio->remaining, 1);
 		bio = r1_bio->bios[r1_bio->read_disk];
 		md_sync_acct(bio->bi_bdev, nr_sectors);
+		if (read_targets == 1)
+			bio->bi_opf &= ~MD_FAILFAST;
 		generic_make_request(bio);
 
 	}
@@ -2875,12 +2961,8 @@ static struct r1conf *setup_conf(struct mddev *mddev)
 
 	err = -ENOMEM;
 	conf->thread = md_register_thread(raid1d, mddev, "raid1");
-	if (!conf->thread) {
-		printk(KERN_ERR
-		       "md/raid1:%s: couldn't allocate thread\n",
-		       mdname(mddev));
+	if (!conf->thread)
 		goto abort;
-	}
 
 	return conf;
 
@@ -2905,13 +2987,13 @@ static int raid1_run(struct mddev *mddev)
 	bool discard_supported = false;
 
 	if (mddev->level != 1) {
-		printk(KERN_ERR "md/raid1:%s: raid level not set to mirroring (%d)\n",
-		       mdname(mddev), mddev->level);
+		pr_warn("md/raid1:%s: raid level not set to mirroring (%d)\n",
+			mdname(mddev), mddev->level);
 		return -EIO;
 	}
 	if (mddev->reshape_position != MaxSector) {
-		printk(KERN_ERR "md/raid1:%s: reshape_position set but not supported\n",
-		       mdname(mddev));
+		pr_warn("md/raid1:%s: reshape_position set but not supported\n",
+			mdname(mddev));
 		return -EIO;
 	}
 	/*
@@ -2950,11 +3032,9 @@ static int raid1_run(struct mddev *mddev)
 		mddev->recovery_cp = MaxSector;
 
 	if (mddev->recovery_cp != MaxSector)
-		printk(KERN_NOTICE "md/raid1:%s: not clean"
-		       " -- starting background reconstruction\n",
-		       mdname(mddev));
-	printk(KERN_INFO
-		"md/raid1:%s: active with %d out of %d mirrors\n",
+		pr_info("md/raid1:%s: not clean -- starting background reconstruction\n",
+			mdname(mddev));
+	pr_info("md/raid1:%s: active with %d out of %d mirrors\n",
 		mdname(mddev), mddev->raid_disks - mddev->degraded,
 		mddev->raid_disks);
 
@@ -2964,6 +3044,7 @@ static int raid1_run(struct mddev *mddev)
 	mddev->thread = conf->thread;
 	conf->thread = NULL;
 	mddev->private = conf;
+	set_bit(MD_FAILFAST_SUPPORTED, &mddev->flags);
 
 	md_set_array_sectors(mddev, raid1_size(mddev, 0, 0));
 
@@ -3107,9 +3188,8 @@ static int raid1_reshape(struct mddev *mddev)
 			rdev->raid_disk = d2;
 			sysfs_unlink_rdev(mddev, rdev);
 			if (sysfs_link_rdev(mddev, rdev))
-				printk(KERN_WARNING
-				       "md/raid1:%s: cannot register rd%d\n",
-				       mdname(mddev), rdev->raid_disk);
+				pr_warn("md/raid1:%s: cannot register rd%d\n",
+					mdname(mddev), rdev->raid_disk);
 		}
 		if (rdev)
 			newmirrors[d2++].rdev = rdev;
@@ -3163,9 +3243,12 @@ static void *raid1_takeover(struct mddev *mddev)
 		mddev->new_layout = 0;
 		mddev->new_chunk_sectors = 0;
 		conf = setup_conf(mddev);
-		if (!IS_ERR(conf))
+		if (!IS_ERR(conf)) {
 			/* Array must appear to be quiesced */
 			conf->array_frozen = 1;
+			clear_bit(MD_HAS_JOURNAL, &mddev->flags);
+			clear_bit(MD_JOURNAL_CLEAN, &mddev->flags);
+		}
 		return conf;
 	}
 	return ERR_PTR(-EINVAL);

drivers/md/raid1.h

@@ -161,14 +161,15 @@ struct r1bio {
 };
 
 /* bits for r1bio.state */
-#define	R1BIO_Uptodate	0
-#define	R1BIO_IsSync	1
-#define	R1BIO_Degraded	2
-#define	R1BIO_BehindIO	3
+enum r1bio_state {
+	R1BIO_Uptodate,
+	R1BIO_IsSync,
+	R1BIO_Degraded,
+	R1BIO_BehindIO,
 /* Set ReadError on bios that experience a readerror so that
  * raid1d knows what to do with them.
  */
-#define R1BIO_ReadError 4
+	R1BIO_ReadError,
 /* For write-behind requests, we call bi_end_io when
  * the last non-write-behind device completes, providing
  * any write was successful.  Otherwise we call when
@@ -176,10 +177,12 @@ struct r1bio {
  * with failure when last write completes (and all failed).
  * Record that bi_end_io was called with this flag...
  */
-#define	R1BIO_Returned 6
+	R1BIO_Returned,
 /* If a write for this request means we can clear some
  * known-bad-block records, we set this flag
  */
-#define	R1BIO_MadeGood 7
-#define	R1BIO_WriteError 8
+	R1BIO_MadeGood,
+	R1BIO_WriteError,
+	R1BIO_FailFast,
+};
 #endif

drivers/md/raid10.c

@@ -25,6 +25,7 @@
 #include <linux/seq_file.h>
 #include <linux/ratelimit.h>
 #include <linux/kthread.h>
+#include <trace/events/block.h>
 #include "md.h"
 #include "raid10.h"
 #include "raid0.h"
@@ -99,12 +100,16 @@ static int max_queued_requests = 1024;
 static void allow_barrier(struct r10conf *conf);
 static void lower_barrier(struct r10conf *conf);
 static int _enough(struct r10conf *conf, int previous, int ignore);
+static int enough(struct r10conf *conf, int ignore);
 static sector_t reshape_request(struct mddev *mddev, sector_t sector_nr,
 				int *skipped);
 static void reshape_request_write(struct mddev *mddev, struct r10bio *r10_bio);
 static void end_reshape_write(struct bio *bio);
 static void end_reshape(struct r10conf *conf);
 
+#define raid10_log(md, fmt, args...)				\
+	do { if ((md)->queue) blk_add_trace_msg((md)->queue, "raid10 " fmt, ##args); } while (0)
+
 static void * r10bio_pool_alloc(gfp_t gfp_flags, void *data)
 {
 	struct r10conf *conf = data;
@@ -404,8 +409,7 @@ static void raid10_end_read_request(struct bio *bio)
 		 * oops, read error - keep the refcount on the rdev
 		 */
 		char b[BDEVNAME_SIZE];
-		printk_ratelimited(KERN_ERR
-				   "md/raid10:%s: %s: rescheduling sector %llu\n",
+		pr_err_ratelimited("md/raid10:%s: %s: rescheduling sector %llu\n",
 				   mdname(conf->mddev),
 				   bdevname(rdev->bdev, b),
 				   (unsigned long long)r10_bio->sector);
@@ -447,6 +451,7 @@ static void raid10_end_write_request(struct bio *bio)
 	struct r10conf *conf = r10_bio->mddev->private;
 	int slot, repl;
 	struct md_rdev *rdev = NULL;
+	struct bio *to_put = NULL;
 	bool discard_error;
 
 	discard_error = bio->bi_error && bio_op(bio) == REQ_OP_DISCARD;
@@ -474,8 +479,24 @@ static void raid10_end_write_request(struct bio *bio)
 			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;
+			if (test_bit(FailFast, &rdev->flags) &&
+			    (bio->bi_opf & MD_FAILFAST)) {
+				md_error(rdev->mddev, rdev);
+				if (!test_bit(Faulty, &rdev->flags))
+					/* This is the only remaining device,
+					 * We need to retry the write without
+					 * FailFast
+					 */
+					set_bit(R10BIO_WriteError, &r10_bio->state);
+				else {
+					r10_bio->devs[slot].bio = NULL;
+					to_put = bio;
+					dec_rdev = 1;
+				}
+			} else
+				set_bit(R10BIO_WriteError, &r10_bio->state);
 		}
 	} else {
 		/*
@@ -525,6 +546,8 @@ static void raid10_end_write_request(struct bio *bio)
 	one_write_done(r10_bio);
 	if (dec_rdev)
 		rdev_dec_pending(rdev, conf->mddev);
+	if (to_put)
+		bio_put(to_put);
 }
 
 /*
@@ -716,6 +739,7 @@ static struct md_rdev *read_balance(struct r10conf *conf,
 	best_dist = MaxSector;
 	best_good_sectors = 0;
 	do_balance = 1;
+	clear_bit(R10BIO_FailFast, &r10_bio->state);
 	/*
 	 * Check if we can balance. We can balance on the whole
 	 * device if no resync is going on (recovery is ok), or below
@@ -780,15 +804,18 @@ static struct md_rdev *read_balance(struct r10conf *conf,
 		if (!do_balance)
 			break;
 
+		if (best_slot >= 0)
+			/* At least 2 disks to choose from so failfast is OK */
+			set_bit(R10BIO_FailFast, &r10_bio->state);
 		/* This optimisation is debatable, and completely destroys
 		 * sequential read speed for 'far copies' arrays.  So only
 		 * keep it for 'near' arrays, and review those later.
 		 */
 		if (geo->near_copies > 1 && !atomic_read(&rdev->nr_pending))
-			break;
+			new_distance = 0;
 
 		/* for far > 1 always use the lowest address */
-		if (geo->far_copies > 1)
+		else if (geo->far_copies > 1)
 			new_distance = r10_bio->devs[slot].addr;
 		else
 			new_distance = abs(r10_bio->devs[slot].addr -
@@ -859,9 +886,14 @@ static void flush_pending_writes(struct r10conf *conf)
 
 		while (bio) { /* submit pending writes */
 			struct bio *next = bio->bi_next;
+			struct md_rdev *rdev = (void*)bio->bi_bdev;
 			bio->bi_next = NULL;
-			if (unlikely((bio_op(bio) ==  REQ_OP_DISCARD) &&
-			    !blk_queue_discard(bdev_get_queue(bio->bi_bdev))))
+			bio->bi_bdev = rdev->bdev;
+			if (test_bit(Faulty, &rdev->flags)) {
+				bio->bi_error = -EIO;
+				bio_endio(bio);
+			} else if (unlikely((bio_op(bio) ==  REQ_OP_DISCARD) &&
+					    !blk_queue_discard(bdev_get_queue(bio->bi_bdev))))
 				/* Just ignore it */
 				bio_endio(bio);
 			else
@@ -937,6 +969,7 @@ static void wait_barrier(struct r10conf *conf)
 		 * that queue to get the nr_pending
 		 * count down.
 		 */
+		raid10_log(conf->mddev, "wait barrier");
 		wait_event_lock_irq(conf->wait_barrier,
 				    !conf->barrier ||
 				    (atomic_read(&conf->nr_pending) &&
@@ -1037,9 +1070,14 @@ static void raid10_unplug(struct blk_plug_cb *cb, bool from_schedule)
 
 	while (bio) { /* submit pending writes */
 		struct bio *next = bio->bi_next;
+		struct md_rdev *rdev = (void*)bio->bi_bdev;
 		bio->bi_next = NULL;
-		if (unlikely((bio_op(bio) ==  REQ_OP_DISCARD) &&
-		    !blk_queue_discard(bdev_get_queue(bio->bi_bdev))))
+		bio->bi_bdev = rdev->bdev;
+		if (test_bit(Faulty, &rdev->flags)) {
+			bio->bi_error = -EIO;
+			bio_endio(bio);
+		} else if (unlikely((bio_op(bio) ==  REQ_OP_DISCARD) &&
+				    !blk_queue_discard(bdev_get_queue(bio->bi_bdev))))
 			/* Just ignore it */
 			bio_endio(bio);
 		else
@@ -1083,6 +1121,7 @@ static void __make_request(struct mddev *mddev, struct bio *bio)
 		/* IO spans the reshape position.  Need to wait for
 		 * reshape to pass
 		 */
+		raid10_log(conf->mddev, "wait reshape");
 		allow_barrier(conf);
 		wait_event(conf->wait_barrier,
 			   conf->reshape_progress <= bio->bi_iter.bi_sector ||
@@ -1099,11 +1138,12 @@ static void __make_request(struct mddev *mddev, struct bio *bio)
 		bio->bi_iter.bi_sector < conf->reshape_progress))) {
 		/* Need to update reshape_position in metadata */
 		mddev->reshape_position = conf->reshape_progress;
-		set_mask_bits(&mddev->flags, 0,
-			      BIT(MD_CHANGE_DEVS) | BIT(MD_CHANGE_PENDING));
+		set_mask_bits(&mddev->sb_flags, 0,
+			      BIT(MD_SB_CHANGE_DEVS) | BIT(MD_SB_CHANGE_PENDING));
 		md_wakeup_thread(mddev->thread);
+		raid10_log(conf->mddev, "wait reshape metadata");
 		wait_event(mddev->sb_wait,
-			   !test_bit(MD_CHANGE_PENDING, &mddev->flags));
+			   !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags));
 
 		conf->reshape_safe = mddev->reshape_position;
 	}
@@ -1154,8 +1194,15 @@ read_again:
 		read_bio->bi_bdev = rdev->bdev;
 		read_bio->bi_end_io = raid10_end_read_request;
 		bio_set_op_attrs(read_bio, op, do_sync);
+		if (test_bit(FailFast, &rdev->flags) &&
+		    test_bit(R10BIO_FailFast, &r10_bio->state))
+			read_bio->bi_opf |= MD_FAILFAST;
 		read_bio->bi_private = r10_bio;
 
+		if (mddev->gendisk)
+			trace_block_bio_remap(bdev_get_queue(read_bio->bi_bdev),
+					      read_bio, disk_devt(mddev->gendisk),
+					      r10_bio->sector);
 		if (max_sectors < r10_bio->sectors) {
 			/* Could not read all from this device, so we will
 			 * need another r10_bio.
@@ -1195,6 +1242,7 @@ read_again:
 	 */
 	if (conf->pending_count >= max_queued_requests) {
 		md_wakeup_thread(mddev->thread);
+		raid10_log(mddev, "wait queued");
 		wait_event(conf->wait_barrier,
 			   conf->pending_count < max_queued_requests);
 	}
@@ -1322,6 +1370,7 @@ retry_write:
 			}
 		}
 		allow_barrier(conf);
+		raid10_log(conf->mddev, "wait rdev %d blocked", blocked_rdev->raid_disk);
 		md_wait_for_blocked_rdev(blocked_rdev, mddev);
 		wait_barrier(conf);
 		goto retry_write;
@@ -1361,8 +1410,18 @@ retry_write:
 			mbio->bi_bdev = rdev->bdev;
 			mbio->bi_end_io	= raid10_end_write_request;
 			bio_set_op_attrs(mbio, op, do_sync | do_fua);
+			if (test_bit(FailFast, &conf->mirrors[d].rdev->flags) &&
+			    enough(conf, d))
+				mbio->bi_opf |= MD_FAILFAST;
 			mbio->bi_private = r10_bio;
 
+			if (conf->mddev->gendisk)
+				trace_block_bio_remap(bdev_get_queue(mbio->bi_bdev),
+						      mbio, disk_devt(conf->mddev->gendisk),
+						      r10_bio->sector);
+			/* flush_pending_writes() needs access to the rdev so...*/
+			mbio->bi_bdev = (void*)rdev;
+
 			atomic_inc(&r10_bio->remaining);
 
 			cb = blk_check_plugged(raid10_unplug, mddev,
@@ -1405,6 +1464,13 @@ retry_write:
 			bio_set_op_attrs(mbio, op, do_sync | do_fua);
 			mbio->bi_private = r10_bio;
 
+			if (conf->mddev->gendisk)
+				trace_block_bio_remap(bdev_get_queue(mbio->bi_bdev),
+						      mbio, disk_devt(conf->mddev->gendisk),
+						      r10_bio->sector);
+			/* flush_pending_writes() needs access to the rdev so...*/
+			mbio->bi_bdev = (void*)rdev;
+
 			atomic_inc(&r10_bio->remaining);
 			spin_lock_irqsave(&conf->device_lock, flags);
 			bio_list_add(&conf->pending_bio_list, mbio);
@@ -1586,14 +1652,13 @@ static void raid10_error(struct mddev *mddev, struct md_rdev *rdev)
 	set_bit(MD_RECOVERY_INTR, &mddev->recovery);
 	set_bit(Blocked, &rdev->flags);
 	set_bit(Faulty, &rdev->flags);
-	set_mask_bits(&mddev->flags, 0,
-		      BIT(MD_CHANGE_DEVS) | BIT(MD_CHANGE_PENDING));
+	set_mask_bits(&mddev->sb_flags, 0,
+		      BIT(MD_SB_CHANGE_DEVS) | BIT(MD_SB_CHANGE_PENDING));
 	spin_unlock_irqrestore(&conf->device_lock, flags);
-	printk(KERN_ALERT
-	       "md/raid10:%s: Disk failure on %s, disabling device.\n"
-	       "md/raid10:%s: Operation continuing on %d devices.\n",
-	       mdname(mddev), bdevname(rdev->bdev, b),
-	       mdname(mddev), conf->geo.raid_disks - mddev->degraded);
+	pr_crit("md/raid10:%s: Disk failure on %s, disabling device.\n"
+		"md/raid10:%s: Operation continuing on %d devices.\n",
+		mdname(mddev), bdevname(rdev->bdev, b),
+		mdname(mddev), conf->geo.raid_disks - mddev->degraded);
 }
 
 static void print_conf(struct r10conf *conf)
@@ -1601,13 +1666,13 @@ static void print_conf(struct r10conf *conf)
 	int i;
 	struct md_rdev *rdev;
 
-	printk(KERN_DEBUG "RAID10 conf printout:\n");
+	pr_debug("RAID10 conf printout:\n");
 	if (!conf) {
-		printk(KERN_DEBUG "(!conf)\n");
+		pr_debug("(!conf)\n");
 		return;
 	}
-	printk(KERN_DEBUG " --- wd:%d rd:%d\n", conf->geo.raid_disks - conf->mddev->degraded,
-		conf->geo.raid_disks);
+	pr_debug(" --- wd:%d rd:%d\n", conf->geo.raid_disks - conf->mddev->degraded,
+		 conf->geo.raid_disks);
 
 	/* This is only called with ->reconfix_mutex held, so
 	 * rcu protection of rdev is not needed */
@@ -1615,10 +1680,10 @@ static void print_conf(struct r10conf *conf)
 		char b[BDEVNAME_SIZE];
 		rdev = conf->mirrors[i].rdev;
 		if (rdev)
-			printk(KERN_DEBUG " disk %d, wo:%d, o:%d, dev:%s\n",
-				i, !test_bit(In_sync, &rdev->flags),
-			        !test_bit(Faulty, &rdev->flags),
-				bdevname(rdev->bdev,b));
+			pr_debug(" disk %d, wo:%d, o:%d, dev:%s\n",
+				 i, !test_bit(In_sync, &rdev->flags),
+				 !test_bit(Faulty, &rdev->flags),
+				 bdevname(rdev->bdev,b));
 	}
 }
 
@@ -1953,6 +2018,7 @@ static void sync_request_write(struct mddev *mddev, struct r10bio *r10_bio)
 	/* now find blocks with errors */
 	for (i=0 ; i < conf->copies ; i++) {
 		int  j, d;
+		struct md_rdev *rdev;
 
 		tbio = r10_bio->devs[i].bio;
 
@@ -1960,6 +2026,8 @@ static void sync_request_write(struct mddev *mddev, struct r10bio *r10_bio)
 			continue;
 		if (i == first)
 			continue;
+		d = r10_bio->devs[i].devnum;
+		rdev = conf->mirrors[d].rdev;
 		if (!r10_bio->devs[i].bio->bi_error) {
 			/* We know that the bi_io_vec layout is the same for
 			 * both 'first' and 'i', so we just compare them.
@@ -1982,6 +2050,10 @@ static void sync_request_write(struct mddev *mddev, struct r10bio *r10_bio)
 			if (test_bit(MD_RECOVERY_CHECK, &mddev->recovery))
 				/* Don't fix anything. */
 				continue;
+		} else if (test_bit(FailFast, &rdev->flags)) {
+			/* Just give up on this device */
+			md_error(rdev->mddev, rdev);
+			continue;
 		}
 		/* Ok, we need to write this bio, either to correct an
 		 * inconsistency or to correct an unreadable block.
@@ -1999,11 +2071,12 @@ static void sync_request_write(struct mddev *mddev, struct r10bio *r10_bio)
 
 		bio_copy_data(tbio, fbio);
 
-		d = r10_bio->devs[i].devnum;
 		atomic_inc(&conf->mirrors[d].rdev->nr_pending);
 		atomic_inc(&r10_bio->remaining);
 		md_sync_acct(conf->mirrors[d].rdev->bdev, bio_sectors(tbio));
 
+		if (test_bit(FailFast, &conf->mirrors[d].rdev->flags))
+			tbio->bi_opf |= MD_FAILFAST;
 		tbio->bi_iter.bi_sector += conf->mirrors[d].rdev->data_offset;
 		tbio->bi_bdev = conf->mirrors[d].rdev->bdev;
 		generic_make_request(tbio);
@@ -2109,10 +2182,8 @@ static void fix_recovery_read_error(struct r10bio *r10_bio)
 				ok = rdev_set_badblocks(rdev2, addr, s, 0);
 				if (!ok) {
 					/* just abort the recovery */
-					printk(KERN_NOTICE
-					       "md/raid10:%s: recovery aborted"
-					       " due to read error\n",
-					       mdname(mddev));
+					pr_notice("md/raid10:%s: recovery aborted due to read error\n",
+						  mdname(mddev));
 
 					conf->mirrors[dw].recovery_disabled
 						= mddev->recovery_disabled;
@@ -2259,14 +2330,11 @@ static void fix_read_error(struct r10conf *conf, struct mddev *mddev, struct r10
 		char b[BDEVNAME_SIZE];
 		bdevname(rdev->bdev, b);
 
-		printk(KERN_NOTICE
-		       "md/raid10:%s: %s: Raid device exceeded "
-		       "read_error threshold [cur %d:max %d]\n",
-		       mdname(mddev), b,
-		       atomic_read(&rdev->read_errors), max_read_errors);
-		printk(KERN_NOTICE
-		       "md/raid10:%s: %s: Failing raid device\n",
-		       mdname(mddev), b);
+		pr_notice("md/raid10:%s: %s: Raid device exceeded read_error threshold [cur %d:max %d]\n",
+			  mdname(mddev), b,
+			  atomic_read(&rdev->read_errors), max_read_errors);
+		pr_notice("md/raid10:%s: %s: Failing raid device\n",
+			  mdname(mddev), b);
 		md_error(mddev, rdev);
 		r10_bio->devs[r10_bio->read_slot].bio = IO_BLOCKED;
 		return;
@@ -2356,20 +2424,16 @@ static void fix_read_error(struct r10conf *conf, struct mddev *mddev, struct r10
 					     s, conf->tmppage, WRITE)
 			    == 0) {
 				/* Well, this device is dead */
-				printk(KERN_NOTICE
-				       "md/raid10:%s: read correction "
-				       "write failed"
-				       " (%d sectors at %llu on %s)\n",
-				       mdname(mddev), s,
-				       (unsigned long long)(
-					       sect +
-					       choose_data_offset(r10_bio,
-								  rdev)),
-				       bdevname(rdev->bdev, b));
-				printk(KERN_NOTICE "md/raid10:%s: %s: failing "
-				       "drive\n",
-				       mdname(mddev),
-				       bdevname(rdev->bdev, b));
+				pr_notice("md/raid10:%s: read correction write failed (%d sectors at %llu on %s)\n",
+					  mdname(mddev), s,
+					  (unsigned long long)(
+						  sect +
+						  choose_data_offset(r10_bio,
+								     rdev)),
+					  bdevname(rdev->bdev, b));
+				pr_notice("md/raid10:%s: %s: failing drive\n",
+					  mdname(mddev),
+					  bdevname(rdev->bdev, b));
 			}
 			rdev_dec_pending(rdev, mddev);
 			rcu_read_lock();
@@ -2397,24 +2461,18 @@ static void fix_read_error(struct r10conf *conf, struct mddev *mddev, struct r10
 						 READ)) {
 			case 0:
 				/* Well, this device is dead */
-				printk(KERN_NOTICE
-				       "md/raid10:%s: unable to read back "
-				       "corrected sectors"
-				       " (%d sectors at %llu on %s)\n",
+				pr_notice("md/raid10:%s: unable to read back corrected sectors (%d sectors at %llu on %s)\n",
 				       mdname(mddev), s,
 				       (unsigned long long)(
 					       sect +
 					       choose_data_offset(r10_bio, rdev)),
 				       bdevname(rdev->bdev, b));
-				printk(KERN_NOTICE "md/raid10:%s: %s: failing "
-				       "drive\n",
+				pr_notice("md/raid10:%s: %s: failing drive\n",
 				       mdname(mddev),
 				       bdevname(rdev->bdev, b));
 				break;
 			case 1:
-				printk(KERN_INFO
-				       "md/raid10:%s: read error corrected"
-				       " (%d sectors at %llu on %s)\n",
+				pr_info("md/raid10:%s: read error corrected (%d sectors at %llu on %s)\n",
 				       mdname(mddev), s,
 				       (unsigned long long)(
 					       sect +
@@ -2503,6 +2561,8 @@ static void handle_read_error(struct mddev *mddev, struct r10bio *r10_bio)
 	char b[BDEVNAME_SIZE];
 	unsigned long do_sync;
 	int max_sectors;
+	dev_t bio_dev;
+	sector_t bio_last_sector;
 
 	/* we got a read error. Maybe the drive is bad.  Maybe just
 	 * the block and we can fix it.
@@ -2514,38 +2574,38 @@ static void handle_read_error(struct mddev *mddev, struct r10bio *r10_bio)
 	 */
 	bio = r10_bio->devs[slot].bio;
 	bdevname(bio->bi_bdev, b);
+	bio_dev = bio->bi_bdev->bd_dev;
+	bio_last_sector = r10_bio->devs[slot].addr + rdev->data_offset + r10_bio->sectors;
 	bio_put(bio);
 	r10_bio->devs[slot].bio = NULL;
 
-	if (mddev->ro == 0) {
+	if (mddev->ro)
+		r10_bio->devs[slot].bio = IO_BLOCKED;
+	else if (!test_bit(FailFast, &rdev->flags)) {
 		freeze_array(conf, 1);
 		fix_read_error(conf, mddev, r10_bio);
 		unfreeze_array(conf);
 	} else
-		r10_bio->devs[slot].bio = IO_BLOCKED;
+		md_error(mddev, rdev);
 
 	rdev_dec_pending(rdev, mddev);
 
 read_more:
 	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,
-		       (unsigned long long)r10_bio->sector);
+		pr_crit_ratelimited("md/raid10:%s: %s: unrecoverable I/O read error for block %llu\n",
+				    mdname(mddev), b,
+				    (unsigned long long)r10_bio->sector);
 		raid_end_bio_io(r10_bio);
 		return;
 	}
 
 	do_sync = (r10_bio->master_bio->bi_opf & REQ_SYNC);
 	slot = r10_bio->read_slot;
-	printk_ratelimited(
-		KERN_ERR
-		"md/raid10:%s: %s: redirecting "
-		"sector %llu to another mirror\n",
-		mdname(mddev),
-		bdevname(rdev->bdev, b),
-		(unsigned long long)r10_bio->sector);
+	pr_err_ratelimited("md/raid10:%s: %s: redirecting sector %llu to another mirror\n",
+			   mdname(mddev),
+			   bdevname(rdev->bdev, b),
+			   (unsigned long long)r10_bio->sector);
 	bio = bio_clone_mddev(r10_bio->master_bio,
 			      GFP_NOIO, mddev);
 	bio_trim(bio, r10_bio->sector - bio->bi_iter.bi_sector, max_sectors);
@@ -2555,8 +2615,15 @@ read_more:
 		+ choose_data_offset(r10_bio, rdev);
 	bio->bi_bdev = rdev->bdev;
 	bio_set_op_attrs(bio, REQ_OP_READ, do_sync);
+	if (test_bit(FailFast, &rdev->flags) &&
+	    test_bit(R10BIO_FailFast, &r10_bio->state))
+		bio->bi_opf |= MD_FAILFAST;
 	bio->bi_private = r10_bio;
 	bio->bi_end_io = raid10_end_read_request;
+	trace_block_bio_remap(bdev_get_queue(bio->bi_bdev),
+			      bio, bio_dev,
+			      bio_last_sector - r10_bio->sectors);
+
 	if (max_sectors < r10_bio->sectors) {
 		/* Drat - have to split this up more */
 		struct bio *mbio = r10_bio->master_bio;
@@ -2694,10 +2761,10 @@ static void raid10d(struct md_thread *thread)
 	md_check_recovery(mddev);
 
 	if (!list_empty_careful(&conf->bio_end_io_list) &&
-	    !test_bit(MD_CHANGE_PENDING, &mddev->flags)) {
+	    !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags)) {
 		LIST_HEAD(tmp);
 		spin_lock_irqsave(&conf->device_lock, flags);
-		if (!test_bit(MD_CHANGE_PENDING, &mddev->flags)) {
+		if (!test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags)) {
 			while (!list_empty(&conf->bio_end_io_list)) {
 				list_move(conf->bio_end_io_list.prev, &tmp);
 				conf->nr_queued--;
@@ -2755,7 +2822,7 @@ static void raid10d(struct md_thread *thread)
 		}
 
 		cond_resched();
-		if (mddev->flags & ~(1<<MD_CHANGE_PENDING))
+		if (mddev->sb_flags & ~(1<<MD_SB_CHANGE_PENDING))
 			md_check_recovery(mddev);
 	}
 	blk_finish_plug(&plug);
@@ -3072,6 +3139,8 @@ static sector_t raid10_sync_request(struct mddev *mddev, sector_t sector_nr,
 				bio->bi_private = r10_bio;
 				bio->bi_end_io = end_sync_read;
 				bio_set_op_attrs(bio, REQ_OP_READ, 0);
+				if (test_bit(FailFast, &rdev->flags))
+					bio->bi_opf |= MD_FAILFAST;
 				from_addr = r10_bio->devs[j].addr;
 				bio->bi_iter.bi_sector = from_addr +
 					rdev->data_offset;
@@ -3160,8 +3229,7 @@ static sector_t raid10_sync_request(struct mddev *mddev, sector_t sector_nr,
 				if (!any_working)  {
 					if (!test_and_set_bit(MD_RECOVERY_INTR,
 							      &mddev->recovery))
-						printk(KERN_INFO "md/raid10:%s: insufficient "
-						       "working devices for recovery.\n",
+						pr_warn("md/raid10:%s: insufficient working devices for recovery.\n",
 						       mdname(mddev));
 					mirror->recovery_disabled
 						= mddev->recovery_disabled;
@@ -3178,6 +3246,23 @@ static sector_t raid10_sync_request(struct mddev *mddev, sector_t sector_nr,
 			rdev_dec_pending(mrdev, mddev);
 			if (mreplace)
 				rdev_dec_pending(mreplace, mddev);
+			if (r10_bio->devs[0].bio->bi_opf & MD_FAILFAST) {
+				/* Only want this if there is elsewhere to
+				 * read from. 'j' is currently the first
+				 * readable copy.
+				 */
+				int targets = 1;
+				for (; j < conf->copies; j++) {
+					int d = r10_bio->devs[j].devnum;
+					if (conf->mirrors[d].rdev &&
+					    test_bit(In_sync,
+						      &conf->mirrors[d].rdev->flags))
+						targets++;
+				}
+				if (targets == 1)
+					r10_bio->devs[0].bio->bi_opf
+						&= ~MD_FAILFAST;
+			}
 		}
 		if (biolist == NULL) {
 			while (r10_bio) {
@@ -3256,6 +3341,8 @@ static sector_t raid10_sync_request(struct mddev *mddev, sector_t sector_nr,
 			bio->bi_private = r10_bio;
 			bio->bi_end_io = end_sync_read;
 			bio_set_op_attrs(bio, REQ_OP_READ, 0);
+			if (test_bit(FailFast, &conf->mirrors[d].rdev->flags))
+				bio->bi_opf |= MD_FAILFAST;
 			bio->bi_iter.bi_sector = sector + rdev->data_offset;
 			bio->bi_bdev = rdev->bdev;
 			count++;
@@ -3279,6 +3366,8 @@ static sector_t raid10_sync_request(struct mddev *mddev, sector_t sector_nr,
 			bio->bi_private = r10_bio;
 			bio->bi_end_io = end_sync_write;
 			bio_set_op_attrs(bio, REQ_OP_WRITE, 0);
+			if (test_bit(FailFast, &conf->mirrors[d].rdev->flags))
+				bio->bi_opf |= MD_FAILFAST;
 			bio->bi_iter.bi_sector = sector + rdev->data_offset;
 			bio->bi_bdev = rdev->bdev;
 			count++;
@@ -3489,15 +3578,14 @@ static struct r10conf *setup_conf(struct mddev *mddev)
 	copies = setup_geo(&geo, mddev, geo_new);
 
 	if (copies == -2) {
-		printk(KERN_ERR "md/raid10:%s: chunk size must be "
-		       "at least PAGE_SIZE(%ld) and be a power of 2.\n",
-		       mdname(mddev), PAGE_SIZE);
+		pr_warn("md/raid10:%s: chunk size must be at least PAGE_SIZE(%ld) and be a power of 2.\n",
+			mdname(mddev), PAGE_SIZE);
 		goto out;
 	}
 
 	if (copies < 2 || copies > mddev->raid_disks) {
-		printk(KERN_ERR "md/raid10:%s: unsupported raid10 layout: 0x%8x\n",
-		       mdname(mddev), mddev->new_layout);
+		pr_warn("md/raid10:%s: unsupported raid10 layout: 0x%8x\n",
+			mdname(mddev), mddev->new_layout);
 		goto out;
 	}
 
@@ -3557,9 +3645,6 @@ static struct r10conf *setup_conf(struct mddev *mddev)
 	return conf;
 
  out:
-	if (err == -ENOMEM)
-		printk(KERN_ERR "md/raid10:%s: couldn't allocate memory.\n",
-		       mdname(mddev));
 	if (conf) {
 		mempool_destroy(conf->r10bio_pool);
 		kfree(conf->mirrors);
@@ -3656,7 +3741,7 @@ static int raid10_run(struct mddev *mddev)
 	}
 	/* need to check that every block has at least one working mirror */
 	if (!enough(conf, -1)) {
-		printk(KERN_ERR "md/raid10:%s: not enough operational mirrors.\n",
+		pr_err("md/raid10:%s: not enough operational mirrors.\n",
 		       mdname(mddev));
 		goto out_free_conf;
 	}
@@ -3698,11 +3783,9 @@ static int raid10_run(struct mddev *mddev)
 	}
 
 	if (mddev->recovery_cp != MaxSector)
-		printk(KERN_NOTICE "md/raid10:%s: not clean"
-		       " -- starting background reconstruction\n",
-		       mdname(mddev));
-	printk(KERN_INFO
-		"md/raid10:%s: active with %d out of %d devices\n",
+		pr_notice("md/raid10:%s: not clean -- starting background reconstruction\n",
+			  mdname(mddev));
+	pr_info("md/raid10:%s: active with %d out of %d devices\n",
 		mdname(mddev), conf->geo.raid_disks - mddev->degraded,
 		conf->geo.raid_disks);
 	/*
@@ -3712,6 +3795,7 @@ static int raid10_run(struct mddev *mddev)
 	size = raid10_size(mddev, 0, 0);
 	md_set_array_sectors(mddev, size);
 	mddev->resync_max_sectors = size;
+	set_bit(MD_FAILFAST_SUPPORTED, &mddev->flags);
 
 	if (mddev->queue) {
 		int stripe = conf->geo.raid_disks *
@@ -3739,7 +3823,7 @@ static int raid10_run(struct mddev *mddev)
 
 		if (max(before_length, after_length) > min_offset_diff) {
 			/* This cannot work */
-			printk("md/raid10: offset difference not enough to continue reshape\n");
+			pr_warn("md/raid10: offset difference not enough to continue reshape\n");
 			goto out_free_conf;
 		}
 		conf->offset_diff = min_offset_diff;
@@ -3846,8 +3930,8 @@ static void *raid10_takeover_raid0(struct mddev *mddev, sector_t size, int devs)
 	struct r10conf *conf;
 
 	if (mddev->degraded > 0) {
-		printk(KERN_ERR "md/raid10:%s: Error: degraded raid0!\n",
-		       mdname(mddev));
+		pr_warn("md/raid10:%s: Error: degraded raid0!\n",
+			mdname(mddev));
 		return ERR_PTR(-EINVAL);
 	}
 	sector_div(size, devs);
@@ -3887,9 +3971,8 @@ static void *raid10_takeover(struct mddev *mddev)
 		/* for raid0 takeover only one zone is supported */
 		raid0_conf = mddev->private;
 		if (raid0_conf->nr_strip_zones > 1) {
-			printk(KERN_ERR "md/raid10:%s: cannot takeover raid 0"
-			       " with more than one zone.\n",
-			       mdname(mddev));
+			pr_warn("md/raid10:%s: cannot takeover raid 0 with more than one zone.\n",
+				mdname(mddev));
 			return ERR_PTR(-EINVAL);
 		}
 		return raid10_takeover_raid0(mddev,
@@ -4078,8 +4161,8 @@ static int raid10_start_reshape(struct mddev *mddev)
 		sector_t size = raid10_size(mddev, 0, 0);
 		if (size < mddev->array_sectors) {
 			spin_unlock_irq(&conf->device_lock);
-			printk(KERN_ERR "md/raid10:%s: array size must be reduce before number of disks\n",
-			       mdname(mddev));
+			pr_warn("md/raid10:%s: array size must be reduce before number of disks\n",
+				mdname(mddev));
 			return -EINVAL;
 		}
 		mddev->resync_max_sectors = size;
@@ -4126,7 +4209,7 @@ static int raid10_start_reshape(struct mddev *mddev)
 	spin_unlock_irq(&conf->device_lock);
 	mddev->raid_disks = conf->geo.raid_disks;
 	mddev->reshape_position = conf->reshape_progress;
-	set_bit(MD_CHANGE_DEVS, &mddev->flags);
+	set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
 
 	clear_bit(MD_RECOVERY_SYNC, &mddev->recovery);
 	clear_bit(MD_RECOVERY_CHECK, &mddev->recovery);
@@ -4321,9 +4404,9 @@ static sector_t reshape_request(struct mddev *mddev, sector_t sector_nr,
 		else
 			mddev->curr_resync_completed = conf->reshape_progress;
 		conf->reshape_checkpoint = jiffies;
-		set_bit(MD_CHANGE_DEVS, &mddev->flags);
+		set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
 		md_wakeup_thread(mddev->thread);
-		wait_event(mddev->sb_wait, mddev->flags == 0 ||
+		wait_event(mddev->sb_wait, mddev->sb_flags == 0 ||
 			   test_bit(MD_RECOVERY_INTR, &mddev->recovery));
 		if (test_bit(MD_RECOVERY_INTR, &mddev->recovery)) {
 			allow_barrier(conf);

drivers/md/raid10.h

@@ -156,5 +156,7 @@ enum r10bio_state {
  * flag is set
  */
 	R10BIO_Previous,
+/* failfast devices did receive failfast requests. */
+	R10BIO_FailFast,
 };
 #endif

drivers/md/raid5-cache.c

@@ -1,5 +1,6 @@
 /*
  * Copyright (C) 2015 Shaohua Li <shli@fb.com>
+ * Copyright (C) 2016 Song Liu <songliubraving@fb.com>
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms and conditions of the GNU General Public License,
@@ -18,8 +19,10 @@
 #include <linux/raid/md_p.h>
 #include <linux/crc32c.h>
 #include <linux/random.h>
+#include <linux/kthread.h>
 #include "md.h"
 #include "raid5.h"
+#include "bitmap.h"
 
 /*
  * metadata/data stored in disk with 4k size unit (a block) regardless
@@ -28,18 +31,70 @@
 #define BLOCK_SECTORS (8)
 
 /*
- * reclaim runs every 1/4 disk size or 10G reclaimable space. This can prevent
- * recovery scans a very long log
+ * log->max_free_space is min(1/4 disk size, 10G reclaimable space).
+ *
+ * In write through mode, the reclaim runs every log->max_free_space.
+ * This can prevent the recovery scans for too long
  */
 #define RECLAIM_MAX_FREE_SPACE (10 * 1024 * 1024 * 2) /* sector */
 #define RECLAIM_MAX_FREE_SPACE_SHIFT (2)
 
+/* wake up reclaim thread periodically */
+#define R5C_RECLAIM_WAKEUP_INTERVAL (30 * HZ)
+/* start flush with these full stripes */
+#define R5C_FULL_STRIPE_FLUSH_BATCH 256
+/* reclaim stripes in groups */
+#define R5C_RECLAIM_STRIPE_GROUP (NR_STRIPE_HASH_LOCKS * 2)
+
 /*
  * We only need 2 bios per I/O unit to make progress, but ensure we
  * have a few more available to not get too tight.
  */
 #define R5L_POOL_SIZE	4
 
+/*
+ * r5c journal modes of the array: write-back or write-through.
+ * write-through mode has identical behavior as existing log only
+ * implementation.
+ */
+enum r5c_journal_mode {
+	R5C_JOURNAL_MODE_WRITE_THROUGH = 0,
+	R5C_JOURNAL_MODE_WRITE_BACK = 1,
+};
+
+static char *r5c_journal_mode_str[] = {"write-through",
+				       "write-back"};
+/*
+ * raid5 cache state machine
+ *
+ * With the RAID cache, each stripe works in two phases:
+ *	- caching phase
+ *	- writing-out phase
+ *
+ * These two phases are controlled by bit STRIPE_R5C_CACHING:
+ *   if STRIPE_R5C_CACHING == 0, the stripe is in writing-out phase
+ *   if STRIPE_R5C_CACHING == 1, the stripe is in caching phase
+ *
+ * When there is no journal, or the journal is in write-through mode,
+ * the stripe is always in writing-out phase.
+ *
+ * For write-back journal, the stripe is sent to caching phase on write
+ * (r5c_try_caching_write). r5c_make_stripe_write_out() kicks off
+ * the write-out phase by clearing STRIPE_R5C_CACHING.
+ *
+ * Stripes in caching phase do not write the raid disks. Instead, all
+ * writes are committed from the log device. Therefore, a stripe in
+ * caching phase handles writes as:
+ *	- write to log device
+ *	- return IO
+ *
+ * Stripes in writing-out phase handle writes as:
+ *	- calculate parity
+ *	- write pending data and parity to journal
+ *	- write data and parity to raid disks
+ *	- return IO for pending writes
+ */
+
 struct r5l_log {
 	struct md_rdev *rdev;
 
@@ -58,7 +113,6 @@ struct r5l_log {
 	u64 seq;			/* log head sequence */
 
 	sector_t next_checkpoint;
-	u64 next_cp_seq;
 
 	struct mutex io_mutex;
 	struct r5l_io_unit *current_io;	/* current io_unit accepting new data */
@@ -96,6 +150,18 @@ struct r5l_log {
 	spinlock_t no_space_stripes_lock;
 
 	bool need_cache_flush;
+
+	/* for r5c_cache */
+	enum r5c_journal_mode r5c_journal_mode;
+
+	/* all stripes in r5cache, in the order of seq at sh->log_start */
+	struct list_head stripe_in_journal_list;
+
+	spinlock_t stripe_in_journal_lock;
+	atomic_t stripe_in_journal_count;
+
+	/* to submit async io_units, to fulfill ordering of flush */
+	struct work_struct deferred_io_work;
 };
 
 /*
@@ -122,6 +188,18 @@ struct r5l_io_unit {
 
 	int state;
 	bool need_split_bio;
+	struct bio *split_bio;
+
+	unsigned int has_flush:1;      /* include flush request */
+	unsigned int has_fua:1;        /* include fua request */
+	unsigned int has_null_flush:1; /* include empty flush request */
+	/*
+	 * io isn't sent yet, flush/fua request can only be submitted till it's
+	 * the first IO in running_ios list
+	 */
+	unsigned int io_deferred:1;
+
+	struct bio_list flush_barriers;   /* size == 0 flush bios */
 };
 
 /* r5l_io_unit state */
@@ -133,6 +211,12 @@ enum r5l_io_unit_state {
 	IO_UNIT_STRIPE_END = 3,	/* stripes data finished writing to raid */
 };
 
+bool r5c_is_writeback(struct r5l_log *log)
+{
+	return (log != NULL &&
+		log->r5c_journal_mode == R5C_JOURNAL_MODE_WRITE_BACK);
+}
+
 static sector_t r5l_ring_add(struct r5l_log *log, sector_t start, sector_t inc)
 {
 	start += inc;
@@ -168,12 +252,235 @@ static void __r5l_set_io_unit_state(struct r5l_io_unit *io,
 	io->state = state;
 }
 
+static void
+r5c_return_dev_pending_writes(struct r5conf *conf, struct r5dev *dev,
+			      struct bio_list *return_bi)
+{
+	struct bio *wbi, *wbi2;
+
+	wbi = dev->written;
+	dev->written = NULL;
+	while (wbi && wbi->bi_iter.bi_sector <
+	       dev->sector + STRIPE_SECTORS) {
+		wbi2 = r5_next_bio(wbi, dev->sector);
+		if (!raid5_dec_bi_active_stripes(wbi)) {
+			md_write_end(conf->mddev);
+			bio_list_add(return_bi, wbi);
+		}
+		wbi = wbi2;
+	}
+}
+
+void r5c_handle_cached_data_endio(struct r5conf *conf,
+	  struct stripe_head *sh, int disks, struct bio_list *return_bi)
+{
+	int i;
+
+	for (i = sh->disks; i--; ) {
+		if (sh->dev[i].written) {
+			set_bit(R5_UPTODATE, &sh->dev[i].flags);
+			r5c_return_dev_pending_writes(conf, &sh->dev[i],
+						      return_bi);
+			bitmap_endwrite(conf->mddev->bitmap, sh->sector,
+					STRIPE_SECTORS,
+					!test_bit(STRIPE_DEGRADED, &sh->state),
+					0);
+		}
+	}
+}
+
+/* Check whether we should flush some stripes to free up stripe cache */
+void r5c_check_stripe_cache_usage(struct r5conf *conf)
+{
+	int total_cached;
+
+	if (!r5c_is_writeback(conf->log))
+		return;
+
+	total_cached = atomic_read(&conf->r5c_cached_partial_stripes) +
+		atomic_read(&conf->r5c_cached_full_stripes);
+
+	/*
+	 * The following condition is true for either of the following:
+	 *   - stripe cache pressure high:
+	 *          total_cached > 3/4 min_nr_stripes ||
+	 *          empty_inactive_list_nr > 0
+	 *   - stripe cache pressure moderate:
+	 *          total_cached > 1/2 min_nr_stripes
+	 */
+	if (total_cached > conf->min_nr_stripes * 1 / 2 ||
+	    atomic_read(&conf->empty_inactive_list_nr) > 0)
+		r5l_wake_reclaim(conf->log, 0);
+}
+
+/*
+ * flush cache when there are R5C_FULL_STRIPE_FLUSH_BATCH or more full
+ * stripes in the cache
+ */
+void r5c_check_cached_full_stripe(struct r5conf *conf)
+{
+	if (!r5c_is_writeback(conf->log))
+		return;
+
+	/*
+	 * wake up reclaim for R5C_FULL_STRIPE_FLUSH_BATCH cached stripes
+	 * or a full stripe (chunk size / 4k stripes).
+	 */
+	if (atomic_read(&conf->r5c_cached_full_stripes) >=
+	    min(R5C_FULL_STRIPE_FLUSH_BATCH,
+		conf->chunk_sectors >> STRIPE_SHIFT))
+		r5l_wake_reclaim(conf->log, 0);
+}
+
+/*
+ * Total log space (in sectors) needed to flush all data in cache
+ *
+ * Currently, writing-out phase automatically includes all pending writes
+ * to the same sector. So the reclaim of each stripe takes up to
+ * (conf->raid_disks + 1) pages of log space.
+ *
+ * To totally avoid deadlock due to log space, the code reserves
+ * (conf->raid_disks + 1) pages for each stripe in cache, which is not
+ * necessary in most cases.
+ *
+ * To improve this, we will need writing-out phase to be able to NOT include
+ * pending writes, which will reduce the requirement to
+ * (conf->max_degraded + 1) pages per stripe in cache.
+ */
+static sector_t r5c_log_required_to_flush_cache(struct r5conf *conf)
+{
+	struct r5l_log *log = conf->log;
+
+	if (!r5c_is_writeback(log))
+		return 0;
+
+	return BLOCK_SECTORS * (conf->raid_disks + 1) *
+		atomic_read(&log->stripe_in_journal_count);
+}
+
+/*
+ * evaluate log space usage and update R5C_LOG_TIGHT and R5C_LOG_CRITICAL
+ *
+ * R5C_LOG_TIGHT is set when free space on the log device is less than 3x of
+ * reclaim_required_space. R5C_LOG_CRITICAL is set when free space on the log
+ * device is less than 2x of reclaim_required_space.
+ */
+static inline void r5c_update_log_state(struct r5l_log *log)
+{
+	struct r5conf *conf = log->rdev->mddev->private;
+	sector_t free_space;
+	sector_t reclaim_space;
+	bool wake_reclaim = false;
+
+	if (!r5c_is_writeback(log))
+		return;
+
+	free_space = r5l_ring_distance(log, log->log_start,
+				       log->last_checkpoint);
+	reclaim_space = r5c_log_required_to_flush_cache(conf);
+	if (free_space < 2 * reclaim_space)
+		set_bit(R5C_LOG_CRITICAL, &conf->cache_state);
+	else {
+		if (test_bit(R5C_LOG_CRITICAL, &conf->cache_state))
+			wake_reclaim = true;
+		clear_bit(R5C_LOG_CRITICAL, &conf->cache_state);
+	}
+	if (free_space < 3 * reclaim_space)
+		set_bit(R5C_LOG_TIGHT, &conf->cache_state);
+	else
+		clear_bit(R5C_LOG_TIGHT, &conf->cache_state);
+
+	if (wake_reclaim)
+		r5l_wake_reclaim(log, 0);
+}
+
+/*
+ * Put the stripe into writing-out phase by clearing STRIPE_R5C_CACHING.
+ * This function should only be called in write-back mode.
+ */
+void r5c_make_stripe_write_out(struct stripe_head *sh)
+{
+	struct r5conf *conf = sh->raid_conf;
+	struct r5l_log *log = conf->log;
+
+	BUG_ON(!r5c_is_writeback(log));
+
+	WARN_ON(!test_bit(STRIPE_R5C_CACHING, &sh->state));
+	clear_bit(STRIPE_R5C_CACHING, &sh->state);
+
+	if (!test_and_set_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
+		atomic_inc(&conf->preread_active_stripes);
+
+	if (test_and_clear_bit(STRIPE_R5C_PARTIAL_STRIPE, &sh->state)) {
+		BUG_ON(atomic_read(&conf->r5c_cached_partial_stripes) == 0);
+		atomic_dec(&conf->r5c_cached_partial_stripes);
+	}
+
+	if (test_and_clear_bit(STRIPE_R5C_FULL_STRIPE, &sh->state)) {
+		BUG_ON(atomic_read(&conf->r5c_cached_full_stripes) == 0);
+		atomic_dec(&conf->r5c_cached_full_stripes);
+	}
+}
+
+static void r5c_handle_data_cached(struct stripe_head *sh)
+{
+	int i;
+
+	for (i = sh->disks; i--; )
+		if (test_and_clear_bit(R5_Wantwrite, &sh->dev[i].flags)) {
+			set_bit(R5_InJournal, &sh->dev[i].flags);
+			clear_bit(R5_LOCKED, &sh->dev[i].flags);
+		}
+	clear_bit(STRIPE_LOG_TRAPPED, &sh->state);
+}
+
+/*
+ * this journal write must contain full parity,
+ * it may also contain some data pages
+ */
+static void r5c_handle_parity_cached(struct stripe_head *sh)
+{
+	int i;
+
+	for (i = sh->disks; i--; )
+		if (test_bit(R5_InJournal, &sh->dev[i].flags))
+			set_bit(R5_Wantwrite, &sh->dev[i].flags);
+}
+
+/*
+ * Setting proper flags after writing (or flushing) data and/or parity to the
+ * log device. This is called from r5l_log_endio() or r5l_log_flush_endio().
+ */
+static void r5c_finish_cache_stripe(struct stripe_head *sh)
+{
+	struct r5l_log *log = sh->raid_conf->log;
+
+	if (log->r5c_journal_mode == R5C_JOURNAL_MODE_WRITE_THROUGH) {
+		BUG_ON(test_bit(STRIPE_R5C_CACHING, &sh->state));
+		/*
+		 * Set R5_InJournal for parity dev[pd_idx]. This means
+		 * all data AND parity in the journal. For RAID 6, it is
+		 * NOT necessary to set the flag for dev[qd_idx], as the
+		 * two parities are written out together.
+		 */
+		set_bit(R5_InJournal, &sh->dev[sh->pd_idx].flags);
+	} else if (test_bit(STRIPE_R5C_CACHING, &sh->state)) {
+		r5c_handle_data_cached(sh);
+	} else {
+		r5c_handle_parity_cached(sh);
+		set_bit(R5_InJournal, &sh->dev[sh->pd_idx].flags);
+	}
+}
+
 static void r5l_io_run_stripes(struct r5l_io_unit *io)
 {
 	struct stripe_head *sh, *next;
 
 	list_for_each_entry_safe(sh, next, &io->stripe_list, log_list) {
 		list_del_init(&sh->log_list);
+
+		r5c_finish_cache_stripe(sh);
+
 		set_bit(STRIPE_HANDLE, &sh->state);
 		raid5_release_stripe(sh);
 	}
@@ -209,9 +516,11 @@ static void r5l_move_to_end_ios(struct r5l_log *log)
 	}
 }
 
+static void __r5l_stripe_write_finished(struct r5l_io_unit *io);
 static void r5l_log_endio(struct bio *bio)
 {
 	struct r5l_io_unit *io = bio->bi_private;
+	struct r5l_io_unit *io_deferred;
 	struct r5l_log *log = io->log;
 	unsigned long flags;
 
@@ -227,18 +536,89 @@ static void r5l_log_endio(struct bio *bio)
 		r5l_move_to_end_ios(log);
 	else
 		r5l_log_run_stripes(log);
+	if (!list_empty(&log->running_ios)) {
+		/*
+		 * FLUSH/FUA io_unit is deferred because of ordering, now we
+		 * can dispatch it
+		 */
+		io_deferred = list_first_entry(&log->running_ios,
+					       struct r5l_io_unit, log_sibling);
+		if (io_deferred->io_deferred)
+			schedule_work(&log->deferred_io_work);
+	}
+
 	spin_unlock_irqrestore(&log->io_list_lock, flags);
 
 	if (log->need_cache_flush)
 		md_wakeup_thread(log->rdev->mddev->thread);
+
+	if (io->has_null_flush) {
+		struct bio *bi;
+
+		WARN_ON(bio_list_empty(&io->flush_barriers));
+		while ((bi = bio_list_pop(&io->flush_barriers)) != NULL) {
+			bio_endio(bi);
+			atomic_dec(&io->pending_stripe);
+		}
+		if (atomic_read(&io->pending_stripe) == 0)
+			__r5l_stripe_write_finished(io);
+	}
+}
+
+static void r5l_do_submit_io(struct r5l_log *log, struct r5l_io_unit *io)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&log->io_list_lock, flags);
+	__r5l_set_io_unit_state(io, IO_UNIT_IO_START);
+	spin_unlock_irqrestore(&log->io_list_lock, flags);
+
+	if (io->has_flush)
+		io->current_bio->bi_opf |= REQ_PREFLUSH;
+	if (io->has_fua)
+		io->current_bio->bi_opf |= REQ_FUA;
+	submit_bio(io->current_bio);
+
+	if (!io->split_bio)
+		return;
+
+	if (io->has_flush)
+		io->split_bio->bi_opf |= REQ_PREFLUSH;
+	if (io->has_fua)
+		io->split_bio->bi_opf |= REQ_FUA;
+	submit_bio(io->split_bio);
+}
+
+/* deferred io_unit will be dispatched here */
+static void r5l_submit_io_async(struct work_struct *work)
+{
+	struct r5l_log *log = container_of(work, struct r5l_log,
+					   deferred_io_work);
+	struct r5l_io_unit *io = NULL;
+	unsigned long flags;
+
+	spin_lock_irqsave(&log->io_list_lock, flags);
+	if (!list_empty(&log->running_ios)) {
+		io = list_first_entry(&log->running_ios, struct r5l_io_unit,
+				      log_sibling);
+		if (!io->io_deferred)
+			io = NULL;
+		else
+			io->io_deferred = 0;
+	}
+	spin_unlock_irqrestore(&log->io_list_lock, flags);
+	if (io)
+		r5l_do_submit_io(log, io);
 }
 
 static void r5l_submit_current_io(struct r5l_log *log)
 {
 	struct r5l_io_unit *io = log->current_io;
+	struct bio *bio;
 	struct r5l_meta_block *block;
 	unsigned long flags;
 	u32 crc;
+	bool do_submit = true;
 
 	if (!io)
 		return;
@@ -247,13 +627,20 @@ static void r5l_submit_current_io(struct r5l_log *log)
 	block->meta_size = cpu_to_le32(io->meta_offset);
 	crc = crc32c_le(log->uuid_checksum, block, PAGE_SIZE);
 	block->checksum = cpu_to_le32(crc);
+	bio = io->current_bio;
 
 	log->current_io = NULL;
 	spin_lock_irqsave(&log->io_list_lock, flags);
-	__r5l_set_io_unit_state(io, IO_UNIT_IO_START);
+	if (io->has_flush || io->has_fua) {
+		if (io != list_first_entry(&log->running_ios,
+					   struct r5l_io_unit, log_sibling)) {
+			io->io_deferred = 1;
+			do_submit = false;
+		}
+	}
 	spin_unlock_irqrestore(&log->io_list_lock, flags);
-
-	submit_bio(io->current_bio);
+	if (do_submit)
+		r5l_do_submit_io(log, io);
 }
 
 static struct bio *r5l_bio_alloc(struct r5l_log *log)
@@ -271,6 +658,7 @@ static void r5_reserve_log_entry(struct r5l_log *log, struct r5l_io_unit *io)
 {
 	log->log_start = r5l_ring_add(log, log->log_start, BLOCK_SECTORS);
 
+	r5c_update_log_state(log);
 	/*
 	 * If we filled up the log device start from the beginning again,
 	 * which will require a new bio.
@@ -297,6 +685,7 @@ static struct r5l_io_unit *r5l_new_meta(struct r5l_log *log)
 	io->log = log;
 	INIT_LIST_HEAD(&io->log_sibling);
 	INIT_LIST_HEAD(&io->stripe_list);
+	bio_list_init(&io->flush_barriers);
 	io->state = IO_UNIT_RUNNING;
 
 	io->meta_page = mempool_alloc(log->meta_pool, GFP_NOIO);
@@ -367,12 +756,11 @@ static void r5l_append_payload_page(struct r5l_log *log, struct page *page)
 	struct r5l_io_unit *io = log->current_io;
 
 	if (io->need_split_bio) {
-		struct bio *prev = io->current_bio;
-
+		BUG_ON(io->split_bio);
+		io->split_bio = io->current_bio;
 		io->current_bio = r5l_bio_alloc(log);
-		bio_chain(io->current_bio, prev);
-
-		submit_bio(prev);
+		bio_chain(io->current_bio, io->split_bio);
+		io->need_split_bio = false;
 	}
 
 	if (!bio_add_page(io->current_bio, page, PAGE_SIZE, 0))
@@ -401,50 +789,85 @@ static int r5l_log_stripe(struct r5l_log *log, struct stripe_head *sh,
 
 	io = log->current_io;
 
+	if (test_and_clear_bit(STRIPE_R5C_PREFLUSH, &sh->state))
+		io->has_flush = 1;
+
 	for (i = 0; i < sh->disks; i++) {
-		if (!test_bit(R5_Wantwrite, &sh->dev[i].flags))
+		if (!test_bit(R5_Wantwrite, &sh->dev[i].flags) ||
+		    test_bit(R5_InJournal, &sh->dev[i].flags))
 			continue;
 		if (i == sh->pd_idx || i == sh->qd_idx)
 			continue;
+		if (test_bit(R5_WantFUA, &sh->dev[i].flags) &&
+		    log->r5c_journal_mode == R5C_JOURNAL_MODE_WRITE_BACK) {
+			io->has_fua = 1;
+			/*
+			 * we need to flush journal to make sure recovery can
+			 * reach the data with fua flag
+			 */
+			io->has_flush = 1;
+		}
 		r5l_append_payload_meta(log, R5LOG_PAYLOAD_DATA,
 					raid5_compute_blocknr(sh, i, 0),
 					sh->dev[i].log_checksum, 0, false);
 		r5l_append_payload_page(log, sh->dev[i].page);
 	}
 
-	if (sh->qd_idx >= 0) {
+	if (parity_pages == 2) {
 		r5l_append_payload_meta(log, R5LOG_PAYLOAD_PARITY,
 					sh->sector, sh->dev[sh->pd_idx].log_checksum,
 					sh->dev[sh->qd_idx].log_checksum, true);
 		r5l_append_payload_page(log, sh->dev[sh->pd_idx].page);
 		r5l_append_payload_page(log, sh->dev[sh->qd_idx].page);
-	} else {
+	} else if (parity_pages == 1) {
 		r5l_append_payload_meta(log, R5LOG_PAYLOAD_PARITY,
 					sh->sector, sh->dev[sh->pd_idx].log_checksum,
 					0, false);
 		r5l_append_payload_page(log, sh->dev[sh->pd_idx].page);
-	}
+	} else  /* Just writing data, not parity, in caching phase */
+		BUG_ON(parity_pages != 0);
 
 	list_add_tail(&sh->log_list, &io->stripe_list);
 	atomic_inc(&io->pending_stripe);
 	sh->log_io = io;
 
+	if (log->r5c_journal_mode == R5C_JOURNAL_MODE_WRITE_THROUGH)
+		return 0;
+
+	if (sh->log_start == MaxSector) {
+		BUG_ON(!list_empty(&sh->r5c));
+		sh->log_start = io->log_start;
+		spin_lock_irq(&log->stripe_in_journal_lock);
+		list_add_tail(&sh->r5c,
+			      &log->stripe_in_journal_list);
+		spin_unlock_irq(&log->stripe_in_journal_lock);
+		atomic_inc(&log->stripe_in_journal_count);
+	}
 	return 0;
 }
 
-static void r5l_wake_reclaim(struct r5l_log *log, sector_t space);
+/* add stripe to no_space_stripes, and then wake up reclaim */
+static inline void r5l_add_no_space_stripe(struct r5l_log *log,
+					   struct stripe_head *sh)
+{
+	spin_lock(&log->no_space_stripes_lock);
+	list_add_tail(&sh->log_list, &log->no_space_stripes);
+	spin_unlock(&log->no_space_stripes_lock);
+}
+
 /*
  * running in raid5d, where reclaim could wait for raid5d too (when it flushes
  * data from log to raid disks), so we shouldn't wait for reclaim here
  */
 int r5l_write_stripe(struct r5l_log *log, struct stripe_head *sh)
 {
+	struct r5conf *conf = sh->raid_conf;
 	int write_disks = 0;
 	int data_pages, parity_pages;
-	int meta_size;
 	int reserve;
 	int i;
 	int ret = 0;
+	bool wake_reclaim = false;
 
 	if (!log)
 		return -EAGAIN;
@@ -456,11 +879,15 @@ int r5l_write_stripe(struct r5l_log *log, struct stripe_head *sh)
 		return -EAGAIN;
 	}
 
+	WARN_ON(test_bit(STRIPE_R5C_CACHING, &sh->state));
+
 	for (i = 0; i < sh->disks; i++) {
 		void *addr;
 
-		if (!test_bit(R5_Wantwrite, &sh->dev[i].flags))
+		if (!test_bit(R5_Wantwrite, &sh->dev[i].flags) ||
+		    test_bit(R5_InJournal, &sh->dev[i].flags))
 			continue;
+
 		write_disks++;
 		/* checksum is already calculated in last run */
 		if (test_bit(STRIPE_LOG_TRAPPED, &sh->state))
@@ -473,15 +900,6 @@ int r5l_write_stripe(struct r5l_log *log, struct stripe_head *sh)
 	parity_pages = 1 + !!(sh->qd_idx >= 0);
 	data_pages = write_disks - parity_pages;
 
-	meta_size =
-		((sizeof(struct r5l_payload_data_parity) + sizeof(__le32))
-		 * data_pages) +
-		sizeof(struct r5l_payload_data_parity) +
-		sizeof(__le32) * parity_pages;
-	/* Doesn't work with very big raid array */
-	if (meta_size + sizeof(struct r5l_meta_block) > PAGE_SIZE)
-		return -EINVAL;
-
 	set_bit(STRIPE_LOG_TRAPPED, &sh->state);
 	/*
 	 * The stripe must enter state machine again to finish the write, so
@@ -493,22 +911,49 @@ int r5l_write_stripe(struct r5l_log *log, struct stripe_head *sh)
 	mutex_lock(&log->io_mutex);
 	/* meta + data */
 	reserve = (1 + write_disks) << (PAGE_SHIFT - 9);
-	if (!r5l_has_free_space(log, reserve)) {
-		spin_lock(&log->no_space_stripes_lock);
-		list_add_tail(&sh->log_list, &log->no_space_stripes);
-		spin_unlock(&log->no_space_stripes_lock);
 
-		r5l_wake_reclaim(log, reserve);
-	} else {
-		ret = r5l_log_stripe(log, sh, data_pages, parity_pages);
-		if (ret) {
-			spin_lock_irq(&log->io_list_lock);
-			list_add_tail(&sh->log_list, &log->no_mem_stripes);
-			spin_unlock_irq(&log->io_list_lock);
+	if (log->r5c_journal_mode == R5C_JOURNAL_MODE_WRITE_THROUGH) {
+		if (!r5l_has_free_space(log, reserve)) {
+			r5l_add_no_space_stripe(log, sh);
+			wake_reclaim = true;
+		} else {
+			ret = r5l_log_stripe(log, sh, data_pages, parity_pages);
+			if (ret) {
+				spin_lock_irq(&log->io_list_lock);
+				list_add_tail(&sh->log_list,
+					      &log->no_mem_stripes);
+				spin_unlock_irq(&log->io_list_lock);
+			}
+		}
+	} else {  /* R5C_JOURNAL_MODE_WRITE_BACK */
+		/*
+		 * log space critical, do not process stripes that are
+		 * not in cache yet (sh->log_start == MaxSector).
+		 */
+		if (test_bit(R5C_LOG_CRITICAL, &conf->cache_state) &&
+		    sh->log_start == MaxSector) {
+			r5l_add_no_space_stripe(log, sh);
+			wake_reclaim = true;
+			reserve = 0;
+		} else if (!r5l_has_free_space(log, reserve)) {
+			if (sh->log_start == log->last_checkpoint)
+				BUG();
+			else
+				r5l_add_no_space_stripe(log, sh);
+		} else {
+			ret = r5l_log_stripe(log, sh, data_pages, parity_pages);
+			if (ret) {
+				spin_lock_irq(&log->io_list_lock);
+				list_add_tail(&sh->log_list,
+					      &log->no_mem_stripes);
+				spin_unlock_irq(&log->io_list_lock);
+			}
 		}
 	}
 
 	mutex_unlock(&log->io_mutex);
+	if (wake_reclaim)
+		r5l_wake_reclaim(log, reserve);
 	return 0;
 }
 
@@ -525,17 +970,34 @@ int r5l_handle_flush_request(struct r5l_log *log, struct bio *bio)
 {
 	if (!log)
 		return -ENODEV;
-	/*
-	 * we flush log disk cache first, then write stripe data to raid disks.
-	 * So if bio is finished, the log disk cache is flushed already. The
-	 * recovery guarantees we can recovery the bio from log disk, so we
-	 * don't need to flush again
-	 */
-	if (bio->bi_iter.bi_size == 0) {
-		bio_endio(bio);
-		return 0;
+
+	if (log->r5c_journal_mode == R5C_JOURNAL_MODE_WRITE_THROUGH) {
+		/*
+		 * in write through (journal only)
+		 * we flush log disk cache first, then write stripe data to
+		 * raid disks. So if bio is finished, the log disk cache is
+		 * flushed already. The recovery guarantees we can recovery
+		 * the bio from log disk, so we don't need to flush again
+		 */
+		if (bio->bi_iter.bi_size == 0) {
+			bio_endio(bio);
+			return 0;
+		}
+		bio->bi_opf &= ~REQ_PREFLUSH;
+	} else {
+		/* write back (with cache) */
+		if (bio->bi_iter.bi_size == 0) {
+			mutex_lock(&log->io_mutex);
+			r5l_get_meta(log, 0);
+			bio_list_add(&log->current_io->flush_barriers, bio);
+			log->current_io->has_flush = 1;
+			log->current_io->has_null_flush = 1;
+			atomic_inc(&log->current_io->pending_stripe);
+			r5l_submit_current_io(log);
+			mutex_unlock(&log->io_mutex);
+			return 0;
+		}
 	}
-	bio->bi_opf &= ~REQ_PREFLUSH;
 	return -EAGAIN;
 }
 
@@ -555,10 +1017,40 @@ static void r5l_run_no_space_stripes(struct r5l_log *log)
 	spin_unlock(&log->no_space_stripes_lock);
 }
 
+/*
+ * calculate new last_checkpoint
+ * for write through mode, returns log->next_checkpoint
+ * for write back, returns log_start of first sh in stripe_in_journal_list
+ */
+static sector_t r5c_calculate_new_cp(struct r5conf *conf)
+{
+	struct stripe_head *sh;
+	struct r5l_log *log = conf->log;
+	sector_t new_cp;
+	unsigned long flags;
+
+	if (log->r5c_journal_mode == R5C_JOURNAL_MODE_WRITE_THROUGH)
+		return log->next_checkpoint;
+
+	spin_lock_irqsave(&log->stripe_in_journal_lock, flags);
+	if (list_empty(&conf->log->stripe_in_journal_list)) {
+		/* all stripes flushed */
+		spin_unlock_irqrestore(&log->stripe_in_journal_lock, flags);
+		return log->next_checkpoint;
+	}
+	sh = list_first_entry(&conf->log->stripe_in_journal_list,
+			      struct stripe_head, r5c);
+	new_cp = sh->log_start;
+	spin_unlock_irqrestore(&log->stripe_in_journal_lock, flags);
+	return new_cp;
+}
+
 static sector_t r5l_reclaimable_space(struct r5l_log *log)
 {
+	struct r5conf *conf = log->rdev->mddev->private;
+
 	return r5l_ring_distance(log, log->last_checkpoint,
-				 log->next_checkpoint);
+				 r5c_calculate_new_cp(conf));
 }
 
 static void r5l_run_no_mem_stripe(struct r5l_log *log)
@@ -589,7 +1081,6 @@ static bool r5l_complete_finished_ios(struct r5l_log *log)
 			break;
 
 		log->next_checkpoint = io->log_start;
-		log->next_cp_seq = io->seq;
 
 		list_del(&io->log_sibling);
 		mempool_free(io, log->io_pool);
@@ -604,6 +1095,7 @@ static bool r5l_complete_finished_ios(struct r5l_log *log)
 static void __r5l_stripe_write_finished(struct r5l_io_unit *io)
 {
 	struct r5l_log *log = io->log;
+	struct r5conf *conf = log->rdev->mddev->private;
 	unsigned long flags;
 
 	spin_lock_irqsave(&log->io_list_lock, flags);
@@ -614,7 +1106,8 @@ static void __r5l_stripe_write_finished(struct r5l_io_unit *io)
 		return;
 	}
 
-	if (r5l_reclaimable_space(log) > log->max_free_space)
+	if (r5l_reclaimable_space(log) > log->max_free_space ||
+	    test_bit(R5C_LOG_TIGHT, &conf->cache_state))
 		r5l_wake_reclaim(log, 0);
 
 	spin_unlock_irqrestore(&log->io_list_lock, flags);
@@ -713,8 +1206,8 @@ static void r5l_write_super_and_discard_space(struct r5l_log *log,
 	 * there is a deadlock. We workaround this issue with a trylock.
 	 * FIXME: we could miss discard if we can't take reconfig mutex
 	 */
-	set_mask_bits(&mddev->flags, 0,
-		BIT(MD_CHANGE_DEVS) | BIT(MD_CHANGE_PENDING));
+	set_mask_bits(&mddev->sb_flags, 0,
+		BIT(MD_SB_CHANGE_DEVS) | BIT(MD_SB_CHANGE_PENDING));
 	if (!mddev_trylock(mddev))
 		return;
 	md_update_sb(mddev, 1);
@@ -735,45 +1228,178 @@ static void r5l_write_super_and_discard_space(struct r5l_log *log,
 	}
 }
 
-
-static void r5l_do_reclaim(struct r5l_log *log)
+/*
+ * r5c_flush_stripe moves stripe from cached list to handle_list. When called,
+ * the stripe must be on r5c_cached_full_stripes or r5c_cached_partial_stripes.
+ *
+ * must hold conf->device_lock
+ */
+static void r5c_flush_stripe(struct r5conf *conf, struct stripe_head *sh)
 {
-	sector_t reclaim_target = xchg(&log->reclaim_target, 0);
-	sector_t reclaimable;
-	sector_t next_checkpoint;
-	u64 next_cp_seq;
+	BUG_ON(list_empty(&sh->lru));
+	BUG_ON(!test_bit(STRIPE_R5C_CACHING, &sh->state));
+	BUG_ON(test_bit(STRIPE_HANDLE, &sh->state));
 
-	spin_lock_irq(&log->io_list_lock);
 	/*
-	 * move proper io_unit to reclaim list. We should not change the order.
-	 * reclaimable/unreclaimable io_unit can be mixed in the list, we
-	 * shouldn't reuse space of an unreclaimable io_unit
+	 * The stripe is not ON_RELEASE_LIST, so it is safe to call
+	 * raid5_release_stripe() while holding conf->device_lock
 	 */
-	while (1) {
-		reclaimable = r5l_reclaimable_space(log);
-		if (reclaimable >= reclaim_target ||
-		    (list_empty(&log->running_ios) &&
-		     list_empty(&log->io_end_ios) &&
-		     list_empty(&log->flushing_ios) &&
-		     list_empty(&log->finished_ios)))
-			break;
+	BUG_ON(test_bit(STRIPE_ON_RELEASE_LIST, &sh->state));
+	assert_spin_locked(&conf->device_lock);
 
-		md_wakeup_thread(log->rdev->mddev->thread);
-		wait_event_lock_irq(log->iounit_wait,
-				    r5l_reclaimable_space(log) > reclaimable,
-				    log->io_list_lock);
-	}
+	list_del_init(&sh->lru);
+	atomic_inc(&sh->count);
 
-	next_checkpoint = log->next_checkpoint;
-	next_cp_seq = log->next_cp_seq;
-	spin_unlock_irq(&log->io_list_lock);
+	set_bit(STRIPE_HANDLE, &sh->state);
+	atomic_inc(&conf->active_stripes);
+	r5c_make_stripe_write_out(sh);
 
-	BUG_ON(reclaimable < 0);
-	if (reclaimable == 0)
+	raid5_release_stripe(sh);
+}
+
+/*
+ * if num == 0, flush all full stripes
+ * if num > 0, flush all full stripes. If less than num full stripes are
+ *             flushed, flush some partial stripes until totally num stripes are
+ *             flushed or there is no more cached stripes.
+ */
+void r5c_flush_cache(struct r5conf *conf, int num)
+{
+	int count;
+	struct stripe_head *sh, *next;
+
+	assert_spin_locked(&conf->device_lock);
+	if (!conf->log)
 		return;
 
-	/*
-	 * write_super will flush cache of each raid disk. We must write super
+	count = 0;
+	list_for_each_entry_safe(sh, next, &conf->r5c_full_stripe_list, lru) {
+		r5c_flush_stripe(conf, sh);
+		count++;
+	}
+
+	if (count >= num)
+		return;
+	list_for_each_entry_safe(sh, next,
+				 &conf->r5c_partial_stripe_list, lru) {
+		r5c_flush_stripe(conf, sh);
+		if (++count >= num)
+			break;
+	}
+}
+
+static void r5c_do_reclaim(struct r5conf *conf)
+{
+	struct r5l_log *log = conf->log;
+	struct stripe_head *sh;
+	int count = 0;
+	unsigned long flags;
+	int total_cached;
+	int stripes_to_flush;
+
+	if (!r5c_is_writeback(log))
+		return;
+
+	total_cached = atomic_read(&conf->r5c_cached_partial_stripes) +
+		atomic_read(&conf->r5c_cached_full_stripes);
+
+	if (total_cached > conf->min_nr_stripes * 3 / 4 ||
+	    atomic_read(&conf->empty_inactive_list_nr) > 0)
+		/*
+		 * if stripe cache pressure high, flush all full stripes and
+		 * some partial stripes
+		 */
+		stripes_to_flush = R5C_RECLAIM_STRIPE_GROUP;
+	else if (total_cached > conf->min_nr_stripes * 1 / 2 ||
+		 atomic_read(&conf->r5c_cached_full_stripes) >
+		 R5C_FULL_STRIPE_FLUSH_BATCH)
+		/*
+		 * if stripe cache pressure moderate, or if there is many full
+		 * stripes,flush all full stripes
+		 */
+		stripes_to_flush = 0;
+	else
+		/* no need to flush */
+		stripes_to_flush = -1;
+
+	if (stripes_to_flush >= 0) {
+		spin_lock_irqsave(&conf->device_lock, flags);
+		r5c_flush_cache(conf, stripes_to_flush);
+		spin_unlock_irqrestore(&conf->device_lock, flags);
+	}
+
+	/* if log space is tight, flush stripes on stripe_in_journal_list */
+	if (test_bit(R5C_LOG_TIGHT, &conf->cache_state)) {
+		spin_lock_irqsave(&log->stripe_in_journal_lock, flags);
+		spin_lock(&conf->device_lock);
+		list_for_each_entry(sh, &log->stripe_in_journal_list, r5c) {
+			/*
+			 * stripes on stripe_in_journal_list could be in any
+			 * state of the stripe_cache state machine. In this
+			 * case, we only want to flush stripe on
+			 * r5c_cached_full/partial_stripes. The following
+			 * condition makes sure the stripe is on one of the
+			 * two lists.
+			 */
+			if (!list_empty(&sh->lru) &&
+			    !test_bit(STRIPE_HANDLE, &sh->state) &&
+			    atomic_read(&sh->count) == 0) {
+				r5c_flush_stripe(conf, sh);
+			}
+			if (count++ >= R5C_RECLAIM_STRIPE_GROUP)
+				break;
+		}
+		spin_unlock(&conf->device_lock);
+		spin_unlock_irqrestore(&log->stripe_in_journal_lock, flags);
+	}
+
+	if (!test_bit(R5C_LOG_CRITICAL, &conf->cache_state))
+		r5l_run_no_space_stripes(log);
+
+	md_wakeup_thread(conf->mddev->thread);
+}
+
+static void r5l_do_reclaim(struct r5l_log *log)
+{
+	struct r5conf *conf = log->rdev->mddev->private;
+	sector_t reclaim_target = xchg(&log->reclaim_target, 0);
+	sector_t reclaimable;
+	sector_t next_checkpoint;
+	bool write_super;
+
+	spin_lock_irq(&log->io_list_lock);
+	write_super = r5l_reclaimable_space(log) > log->max_free_space ||
+		reclaim_target != 0 || !list_empty(&log->no_space_stripes);
+	/*
+	 * move proper io_unit to reclaim list. We should not change the order.
+	 * reclaimable/unreclaimable io_unit can be mixed in the list, we
+	 * shouldn't reuse space of an unreclaimable io_unit
+	 */
+	while (1) {
+		reclaimable = r5l_reclaimable_space(log);
+		if (reclaimable >= reclaim_target ||
+		    (list_empty(&log->running_ios) &&
+		     list_empty(&log->io_end_ios) &&
+		     list_empty(&log->flushing_ios) &&
+		     list_empty(&log->finished_ios)))
+			break;
+
+		md_wakeup_thread(log->rdev->mddev->thread);
+		wait_event_lock_irq(log->iounit_wait,
+				    r5l_reclaimable_space(log) > reclaimable,
+				    log->io_list_lock);
+	}
+
+	next_checkpoint = r5c_calculate_new_cp(conf);
+	spin_unlock_irq(&log->io_list_lock);
+
+	BUG_ON(reclaimable < 0);
+
+	if (reclaimable == 0 || !write_super)
+		return;
+
+	/*
+	 * write_super will flush cache of each raid disk. We must write super
 	 * here, because the log area might be reused soon and we don't want to
 	 * confuse recovery
 	 */
@@ -781,7 +1407,7 @@ static void r5l_do_reclaim(struct r5l_log *log)
 
 	mutex_lock(&log->io_mutex);
 	log->last_checkpoint = next_checkpoint;
-	log->last_cp_seq = next_cp_seq;
+	r5c_update_log_state(log);
 	mutex_unlock(&log->io_mutex);
 
 	r5l_run_no_space_stripes(log);
@@ -795,14 +1421,17 @@ static void r5l_reclaim_thread(struct md_thread *thread)
 
 	if (!log)
 		return;
+	r5c_do_reclaim(conf);
 	r5l_do_reclaim(log);
 }
 
-static void r5l_wake_reclaim(struct r5l_log *log, sector_t space)
+void r5l_wake_reclaim(struct r5l_log *log, sector_t space)
 {
 	unsigned long target;
 	unsigned long new = (unsigned long)space; /* overflow in theory */
 
+	if (!log)
+		return;
 	do {
 		target = log->reclaim_target;
 		if (new < target)
@@ -816,22 +1445,14 @@ void r5l_quiesce(struct r5l_log *log, int state)
 	struct mddev *mddev;
 	if (!log || state == 2)
 		return;
-	if (state == 0) {
-		/*
-		 * This is a special case for hotadd. In suspend, the array has
-		 * no journal. In resume, journal is initialized as well as the
-		 * reclaim thread.
-		 */
-		if (log->reclaim_thread)
-			return;
-		log->reclaim_thread = md_register_thread(r5l_reclaim_thread,
-					log->rdev->mddev, "reclaim");
-	} else if (state == 1) {
+	if (state == 0)
+		kthread_unpark(log->reclaim_thread->tsk);
+	else if (state == 1) {
 		/* make sure r5l_write_super_and_discard_space exits */
 		mddev = log->rdev->mddev;
 		wake_up(&mddev->sb_wait);
-		r5l_wake_reclaim(log, -1L);
-		md_unregister_thread(&log->reclaim_thread);
+		kthread_park(log->reclaim_thread->tsk);
+		r5l_wake_reclaim(log, MaxSector);
 		r5l_do_reclaim(log);
 	}
 }
@@ -857,10 +1478,13 @@ struct r5l_recovery_ctx {
 	sector_t meta_total_blocks;	/* total size of current meta and data */
 	sector_t pos;			/* recovery position */
 	u64 seq;			/* recovery position seq */
+	int data_parity_stripes;	/* number of data_parity stripes */
+	int data_only_stripes;		/* number of data_only stripes */
+	struct list_head cached_list;
 };
 
-static int r5l_read_meta_block(struct r5l_log *log,
-			       struct r5l_recovery_ctx *ctx)
+static int r5l_recovery_read_meta_block(struct r5l_log *log,
+					struct r5l_recovery_ctx *ctx)
 {
 	struct page *page = ctx->meta_page;
 	struct r5l_meta_block *mb;
@@ -892,170 +1516,618 @@ static int r5l_read_meta_block(struct r5l_log *log,
 	return 0;
 }
 
-static int r5l_recovery_flush_one_stripe(struct r5l_log *log,
-					 struct r5l_recovery_ctx *ctx,
-					 sector_t stripe_sect,
-					 int *offset, sector_t *log_offset)
+static void
+r5l_recovery_create_empty_meta_block(struct r5l_log *log,
+				     struct page *page,
+				     sector_t pos, u64 seq)
 {
-	struct r5conf *conf = log->rdev->mddev->private;
-	struct stripe_head *sh;
-	struct r5l_payload_data_parity *payload;
-	int disk_index;
+	struct r5l_meta_block *mb;
 
-	sh = raid5_get_active_stripe(conf, stripe_sect, 0, 0, 0);
-	while (1) {
-		payload = page_address(ctx->meta_page) + *offset;
+	mb = page_address(page);
+	clear_page(mb);
+	mb->magic = cpu_to_le32(R5LOG_MAGIC);
+	mb->version = R5LOG_VERSION;
+	mb->meta_size = cpu_to_le32(sizeof(struct r5l_meta_block));
+	mb->seq = cpu_to_le64(seq);
+	mb->position = cpu_to_le64(pos);
+}
 
-		if (le16_to_cpu(payload->header.type) == R5LOG_PAYLOAD_DATA) {
-			raid5_compute_sector(conf,
-					     le64_to_cpu(payload->location), 0,
-					     &disk_index, sh);
+static int r5l_log_write_empty_meta_block(struct r5l_log *log, sector_t pos,
+					  u64 seq)
+{
+	struct page *page;
+	struct r5l_meta_block *mb;
 
-			sync_page_io(log->rdev, *log_offset, PAGE_SIZE,
-				     sh->dev[disk_index].page, REQ_OP_READ, 0,
-				     false);
-			sh->dev[disk_index].log_checksum =
-				le32_to_cpu(payload->checksum[0]);
-			set_bit(R5_Wantwrite, &sh->dev[disk_index].flags);
-			ctx->meta_total_blocks += BLOCK_SECTORS;
-		} else {
-			disk_index = sh->pd_idx;
-			sync_page_io(log->rdev, *log_offset, PAGE_SIZE,
-				     sh->dev[disk_index].page, REQ_OP_READ, 0,
-				     false);
-			sh->dev[disk_index].log_checksum =
-				le32_to_cpu(payload->checksum[0]);
-			set_bit(R5_Wantwrite, &sh->dev[disk_index].flags);
-
-			if (sh->qd_idx >= 0) {
-				disk_index = sh->qd_idx;
-				sync_page_io(log->rdev,
-					     r5l_ring_add(log, *log_offset, BLOCK_SECTORS),
-					     PAGE_SIZE, sh->dev[disk_index].page,
-					     REQ_OP_READ, 0, false);
-				sh->dev[disk_index].log_checksum =
-					le32_to_cpu(payload->checksum[1]);
-				set_bit(R5_Wantwrite,
-					&sh->dev[disk_index].flags);
-			}
-			ctx->meta_total_blocks += BLOCK_SECTORS * conf->max_degraded;
-		}
+	page = alloc_page(GFP_KERNEL);
+	if (!page)
+		return -ENOMEM;
+	r5l_recovery_create_empty_meta_block(log, page, pos, seq);
+	mb = page_address(page);
+	mb->checksum = cpu_to_le32(crc32c_le(log->uuid_checksum,
+					     mb, PAGE_SIZE));
+	if (!sync_page_io(log->rdev, pos, PAGE_SIZE, page, REQ_OP_WRITE,
+			  REQ_FUA, false)) {
+		__free_page(page);
+		return -EIO;
+	}
+	__free_page(page);
+	return 0;
+}
 
-		*log_offset = r5l_ring_add(log, *log_offset,
-					   le32_to_cpu(payload->size));
-		*offset += sizeof(struct r5l_payload_data_parity) +
-			sizeof(__le32) *
-			(le32_to_cpu(payload->size) >> (PAGE_SHIFT - 9));
-		if (le16_to_cpu(payload->header.type) == R5LOG_PAYLOAD_PARITY)
-			break;
+/*
+ * r5l_recovery_load_data and r5l_recovery_load_parity uses flag R5_Wantwrite
+ * to mark valid (potentially not flushed) data in the journal.
+ *
+ * We already verified checksum in r5l_recovery_verify_data_checksum_for_mb,
+ * so there should not be any mismatch here.
+ */
+static void r5l_recovery_load_data(struct r5l_log *log,
+				   struct stripe_head *sh,
+				   struct r5l_recovery_ctx *ctx,
+				   struct r5l_payload_data_parity *payload,
+				   sector_t log_offset)
+{
+	struct mddev *mddev = log->rdev->mddev;
+	struct r5conf *conf = mddev->private;
+	int dd_idx;
+
+	raid5_compute_sector(conf,
+			     le64_to_cpu(payload->location), 0,
+			     &dd_idx, sh);
+	sync_page_io(log->rdev, log_offset, PAGE_SIZE,
+		     sh->dev[dd_idx].page, REQ_OP_READ, 0, false);
+	sh->dev[dd_idx].log_checksum =
+		le32_to_cpu(payload->checksum[0]);
+	ctx->meta_total_blocks += BLOCK_SECTORS;
+
+	set_bit(R5_Wantwrite, &sh->dev[dd_idx].flags);
+	set_bit(STRIPE_R5C_CACHING, &sh->state);
+}
+
+static void r5l_recovery_load_parity(struct r5l_log *log,
+				     struct stripe_head *sh,
+				     struct r5l_recovery_ctx *ctx,
+				     struct r5l_payload_data_parity *payload,
+				     sector_t log_offset)
+{
+	struct mddev *mddev = log->rdev->mddev;
+	struct r5conf *conf = mddev->private;
+
+	ctx->meta_total_blocks += BLOCK_SECTORS * conf->max_degraded;
+	sync_page_io(log->rdev, log_offset, PAGE_SIZE,
+		     sh->dev[sh->pd_idx].page, REQ_OP_READ, 0, false);
+	sh->dev[sh->pd_idx].log_checksum =
+		le32_to_cpu(payload->checksum[0]);
+	set_bit(R5_Wantwrite, &sh->dev[sh->pd_idx].flags);
+
+	if (sh->qd_idx >= 0) {
+		sync_page_io(log->rdev,
+			     r5l_ring_add(log, log_offset, BLOCK_SECTORS),
+			     PAGE_SIZE, sh->dev[sh->qd_idx].page,
+			     REQ_OP_READ, 0, false);
+		sh->dev[sh->qd_idx].log_checksum =
+			le32_to_cpu(payload->checksum[1]);
+		set_bit(R5_Wantwrite, &sh->dev[sh->qd_idx].flags);
 	}
+	clear_bit(STRIPE_R5C_CACHING, &sh->state);
+}
 
-	for (disk_index = 0; disk_index < sh->disks; disk_index++) {
-		void *addr;
-		u32 checksum;
+static void r5l_recovery_reset_stripe(struct stripe_head *sh)
+{
+	int i;
+
+	sh->state = 0;
+	sh->log_start = MaxSector;
+	for (i = sh->disks; i--; )
+		sh->dev[i].flags = 0;
+}
+
+static void
+r5l_recovery_replay_one_stripe(struct r5conf *conf,
+			       struct stripe_head *sh,
+			       struct r5l_recovery_ctx *ctx)
+{
+	struct md_rdev *rdev, *rrdev;
+	int disk_index;
+	int data_count = 0;
 
+	for (disk_index = 0; disk_index < sh->disks; disk_index++) {
 		if (!test_bit(R5_Wantwrite, &sh->dev[disk_index].flags))
 			continue;
-		addr = kmap_atomic(sh->dev[disk_index].page);
-		checksum = crc32c_le(log->uuid_checksum, addr, PAGE_SIZE);
-		kunmap_atomic(addr);
-		if (checksum != sh->dev[disk_index].log_checksum)
-			goto error;
+		if (disk_index == sh->qd_idx || disk_index == sh->pd_idx)
+			continue;
+		data_count++;
 	}
 
-	for (disk_index = 0; disk_index < sh->disks; disk_index++) {
-		struct md_rdev *rdev, *rrdev;
+	/*
+	 * stripes that only have parity must have been flushed
+	 * before the crash that we are now recovering from, so
+	 * there is nothing more to recovery.
+	 */
+	if (data_count == 0)
+		goto out;
 
-		if (!test_and_clear_bit(R5_Wantwrite,
-					&sh->dev[disk_index].flags))
+	for (disk_index = 0; disk_index < sh->disks; disk_index++) {
+		if (!test_bit(R5_Wantwrite, &sh->dev[disk_index].flags))
 			continue;
 
 		/* in case device is broken */
+		rcu_read_lock();
 		rdev = rcu_dereference(conf->disks[disk_index].rdev);
-		if (rdev)
-			sync_page_io(rdev, stripe_sect, PAGE_SIZE,
+		if (rdev) {
+			atomic_inc(&rdev->nr_pending);
+			rcu_read_unlock();
+			sync_page_io(rdev, sh->sector, PAGE_SIZE,
 				     sh->dev[disk_index].page, REQ_OP_WRITE, 0,
 				     false);
+			rdev_dec_pending(rdev, rdev->mddev);
+			rcu_read_lock();
+		}
 		rrdev = rcu_dereference(conf->disks[disk_index].replacement);
-		if (rrdev)
-			sync_page_io(rrdev, stripe_sect, PAGE_SIZE,
+		if (rrdev) {
+			atomic_inc(&rrdev->nr_pending);
+			rcu_read_unlock();
+			sync_page_io(rrdev, sh->sector, PAGE_SIZE,
 				     sh->dev[disk_index].page, REQ_OP_WRITE, 0,
 				     false);
+			rdev_dec_pending(rrdev, rrdev->mddev);
+			rcu_read_lock();
+		}
+		rcu_read_unlock();
 	}
-	raid5_release_stripe(sh);
+	ctx->data_parity_stripes++;
+out:
+	r5l_recovery_reset_stripe(sh);
+}
+
+static struct stripe_head *
+r5c_recovery_alloc_stripe(struct r5conf *conf,
+			  sector_t stripe_sect,
+			  sector_t log_start)
+{
+	struct stripe_head *sh;
+
+	sh = raid5_get_active_stripe(conf, stripe_sect, 0, 1, 0);
+	if (!sh)
+		return NULL;  /* no more stripe available */
+
+	r5l_recovery_reset_stripe(sh);
+	sh->log_start = log_start;
+
+	return sh;
+}
+
+static struct stripe_head *
+r5c_recovery_lookup_stripe(struct list_head *list, sector_t sect)
+{
+	struct stripe_head *sh;
+
+	list_for_each_entry(sh, list, lru)
+		if (sh->sector == sect)
+			return sh;
+	return NULL;
+}
+
+static void
+r5c_recovery_drop_stripes(struct list_head *cached_stripe_list,
+			  struct r5l_recovery_ctx *ctx)
+{
+	struct stripe_head *sh, *next;
+
+	list_for_each_entry_safe(sh, next, cached_stripe_list, lru) {
+		r5l_recovery_reset_stripe(sh);
+		list_del_init(&sh->lru);
+		raid5_release_stripe(sh);
+	}
+}
+
+static void
+r5c_recovery_replay_stripes(struct list_head *cached_stripe_list,
+			    struct r5l_recovery_ctx *ctx)
+{
+	struct stripe_head *sh, *next;
+
+	list_for_each_entry_safe(sh, next, cached_stripe_list, lru)
+		if (!test_bit(STRIPE_R5C_CACHING, &sh->state)) {
+			r5l_recovery_replay_one_stripe(sh->raid_conf, sh, ctx);
+			list_del_init(&sh->lru);
+			raid5_release_stripe(sh);
+		}
+}
+
+/* if matches return 0; otherwise return -EINVAL */
+static int
+r5l_recovery_verify_data_checksum(struct r5l_log *log, struct page *page,
+				  sector_t log_offset, __le32 log_checksum)
+{
+	void *addr;
+	u32 checksum;
+
+	sync_page_io(log->rdev, log_offset, PAGE_SIZE,
+		     page, REQ_OP_READ, 0, false);
+	addr = kmap_atomic(page);
+	checksum = crc32c_le(log->uuid_checksum, addr, PAGE_SIZE);
+	kunmap_atomic(addr);
+	return (le32_to_cpu(log_checksum) == checksum) ? 0 : -EINVAL;
+}
+
+/*
+ * before loading data to stripe cache, we need verify checksum for all data,
+ * if there is mismatch for any data page, we drop all data in the mata block
+ */
+static int
+r5l_recovery_verify_data_checksum_for_mb(struct r5l_log *log,
+					 struct r5l_recovery_ctx *ctx)
+{
+	struct mddev *mddev = log->rdev->mddev;
+	struct r5conf *conf = mddev->private;
+	struct r5l_meta_block *mb = page_address(ctx->meta_page);
+	sector_t mb_offset = sizeof(struct r5l_meta_block);
+	sector_t log_offset = r5l_ring_add(log, ctx->pos, BLOCK_SECTORS);
+	struct page *page;
+	struct r5l_payload_data_parity *payload;
+
+	page = alloc_page(GFP_KERNEL);
+	if (!page)
+		return -ENOMEM;
+
+	while (mb_offset < le32_to_cpu(mb->meta_size)) {
+		payload = (void *)mb + mb_offset;
+
+		if (payload->header.type == R5LOG_PAYLOAD_DATA) {
+			if (r5l_recovery_verify_data_checksum(
+				    log, page, log_offset,
+				    payload->checksum[0]) < 0)
+				goto mismatch;
+		} else if (payload->header.type == R5LOG_PAYLOAD_PARITY) {
+			if (r5l_recovery_verify_data_checksum(
+				    log, page, log_offset,
+				    payload->checksum[0]) < 0)
+				goto mismatch;
+			if (conf->max_degraded == 2 && /* q for RAID 6 */
+			    r5l_recovery_verify_data_checksum(
+				    log, page,
+				    r5l_ring_add(log, log_offset,
+						 BLOCK_SECTORS),
+				    payload->checksum[1]) < 0)
+				goto mismatch;
+		} else /* not R5LOG_PAYLOAD_DATA or R5LOG_PAYLOAD_PARITY */
+			goto mismatch;
+
+		log_offset = r5l_ring_add(log, log_offset,
+					  le32_to_cpu(payload->size));
+
+		mb_offset += sizeof(struct r5l_payload_data_parity) +
+			sizeof(__le32) *
+			(le32_to_cpu(payload->size) >> (PAGE_SHIFT - 9));
+	}
+
+	put_page(page);
 	return 0;
 
-error:
-	for (disk_index = 0; disk_index < sh->disks; disk_index++)
-		sh->dev[disk_index].flags = 0;
-	raid5_release_stripe(sh);
+mismatch:
+	put_page(page);
 	return -EINVAL;
 }
 
-static int r5l_recovery_flush_one_meta(struct r5l_log *log,
-				       struct r5l_recovery_ctx *ctx)
+/*
+ * Analyze all data/parity pages in one meta block
+ * Returns:
+ * 0 for success
+ * -EINVAL for unknown playload type
+ * -EAGAIN for checksum mismatch of data page
+ * -ENOMEM for run out of memory (alloc_page failed or run out of stripes)
+ */
+static int
+r5c_recovery_analyze_meta_block(struct r5l_log *log,
+				struct r5l_recovery_ctx *ctx,
+				struct list_head *cached_stripe_list)
 {
-	struct r5conf *conf = log->rdev->mddev->private;
-	struct r5l_payload_data_parity *payload;
+	struct mddev *mddev = log->rdev->mddev;
+	struct r5conf *conf = mddev->private;
 	struct r5l_meta_block *mb;
-	int offset;
+	struct r5l_payload_data_parity *payload;
+	int mb_offset;
 	sector_t log_offset;
-	sector_t stripe_sector;
+	sector_t stripe_sect;
+	struct stripe_head *sh;
+	int ret;
+
+	/*
+	 * for mismatch in data blocks, we will drop all data in this mb, but
+	 * we will still read next mb for other data with FLUSH flag, as
+	 * io_unit could finish out of order.
+	 */
+	ret = r5l_recovery_verify_data_checksum_for_mb(log, ctx);
+	if (ret == -EINVAL)
+		return -EAGAIN;
+	else if (ret)
+		return ret;   /* -ENOMEM duo to alloc_page() failed */
 
 	mb = page_address(ctx->meta_page);
-	offset = sizeof(struct r5l_meta_block);
+	mb_offset = sizeof(struct r5l_meta_block);
 	log_offset = r5l_ring_add(log, ctx->pos, BLOCK_SECTORS);
 
-	while (offset < le32_to_cpu(mb->meta_size)) {
+	while (mb_offset < le32_to_cpu(mb->meta_size)) {
 		int dd;
 
-		payload = (void *)mb + offset;
-		stripe_sector = raid5_compute_sector(conf,
-						     le64_to_cpu(payload->location), 0, &dd, NULL);
-		if (r5l_recovery_flush_one_stripe(log, ctx, stripe_sector,
-						  &offset, &log_offset))
+		payload = (void *)mb + mb_offset;
+		stripe_sect = (payload->header.type == R5LOG_PAYLOAD_DATA) ?
+			raid5_compute_sector(
+				conf, le64_to_cpu(payload->location), 0, &dd,
+				NULL)
+			: le64_to_cpu(payload->location);
+
+		sh = r5c_recovery_lookup_stripe(cached_stripe_list,
+						stripe_sect);
+
+		if (!sh) {
+			sh = r5c_recovery_alloc_stripe(conf, stripe_sect, ctx->pos);
+			/*
+			 * cannot get stripe from raid5_get_active_stripe
+			 * try replay some stripes
+			 */
+			if (!sh) {
+				r5c_recovery_replay_stripes(
+					cached_stripe_list, ctx);
+				sh = r5c_recovery_alloc_stripe(
+					conf, stripe_sect, ctx->pos);
+			}
+			if (!sh) {
+				pr_debug("md/raid:%s: Increasing stripe cache size to %d to recovery data on journal.\n",
+					mdname(mddev),
+					conf->min_nr_stripes * 2);
+				raid5_set_cache_size(mddev,
+						     conf->min_nr_stripes * 2);
+				sh = r5c_recovery_alloc_stripe(
+					conf, stripe_sect, ctx->pos);
+			}
+			if (!sh) {
+				pr_err("md/raid:%s: Cannot get enough stripes due to memory pressure. Recovery failed.\n",
+				       mdname(mddev));
+				return -ENOMEM;
+			}
+			list_add_tail(&sh->lru, cached_stripe_list);
+		}
+
+		if (payload->header.type == R5LOG_PAYLOAD_DATA) {
+			if (!test_bit(STRIPE_R5C_CACHING, &sh->state) &&
+			    test_bit(R5_Wantwrite, &sh->dev[sh->pd_idx].flags)) {
+				r5l_recovery_replay_one_stripe(conf, sh, ctx);
+				sh->log_start = ctx->pos;
+				list_move_tail(&sh->lru, cached_stripe_list);
+			}
+			r5l_recovery_load_data(log, sh, ctx, payload,
+					       log_offset);
+		} else if (payload->header.type == R5LOG_PAYLOAD_PARITY)
+			r5l_recovery_load_parity(log, sh, ctx, payload,
+						 log_offset);
+		else
 			return -EINVAL;
+
+		log_offset = r5l_ring_add(log, log_offset,
+					  le32_to_cpu(payload->size));
+
+		mb_offset += sizeof(struct r5l_payload_data_parity) +
+			sizeof(__le32) *
+			(le32_to_cpu(payload->size) >> (PAGE_SHIFT - 9));
 	}
+
 	return 0;
 }
 
-/* copy data/parity from log to raid disks */
-static void r5l_recovery_flush_log(struct r5l_log *log,
-				   struct r5l_recovery_ctx *ctx)
+/*
+ * Load the stripe into cache. The stripe will be written out later by
+ * the stripe cache state machine.
+ */
+static void r5c_recovery_load_one_stripe(struct r5l_log *log,
+					 struct stripe_head *sh)
 {
+	struct r5dev *dev;
+	int i;
+
+	for (i = sh->disks; i--; ) {
+		dev = sh->dev + i;
+		if (test_and_clear_bit(R5_Wantwrite, &dev->flags)) {
+			set_bit(R5_InJournal, &dev->flags);
+			set_bit(R5_UPTODATE, &dev->flags);
+		}
+	}
+	list_add_tail(&sh->r5c, &log->stripe_in_journal_list);
+	atomic_inc(&log->stripe_in_journal_count);
+}
+
+/*
+ * Scan through the log for all to-be-flushed data
+ *
+ * For stripes with data and parity, namely Data-Parity stripe
+ * (STRIPE_R5C_CACHING == 0), we simply replay all the writes.
+ *
+ * For stripes with only data, namely Data-Only stripe
+ * (STRIPE_R5C_CACHING == 1), we load them to stripe cache state machine.
+ *
+ * For a stripe, if we see data after parity, we should discard all previous
+ * data and parity for this stripe, as these data are already flushed to
+ * the array.
+ *
+ * At the end of the scan, we return the new journal_tail, which points to
+ * first data-only stripe on the journal device, or next invalid meta block.
+ */
+static int r5c_recovery_flush_log(struct r5l_log *log,
+				  struct r5l_recovery_ctx *ctx)
+{
+	struct stripe_head *sh;
+	int ret = 0;
+
+	/* scan through the log */
 	while (1) {
-		if (r5l_read_meta_block(log, ctx))
-			return;
-		if (r5l_recovery_flush_one_meta(log, ctx))
-			return;
+		if (r5l_recovery_read_meta_block(log, ctx))
+			break;
+
+		ret = r5c_recovery_analyze_meta_block(log, ctx,
+						      &ctx->cached_list);
+		/*
+		 * -EAGAIN means mismatch in data block, in this case, we still
+		 * try scan the next metablock
+		 */
+		if (ret && ret != -EAGAIN)
+			break;   /* ret == -EINVAL or -ENOMEM */
 		ctx->seq++;
 		ctx->pos = r5l_ring_add(log, ctx->pos, ctx->meta_total_blocks);
 	}
+
+	if (ret == -ENOMEM) {
+		r5c_recovery_drop_stripes(&ctx->cached_list, ctx);
+		return ret;
+	}
+
+	/* replay data-parity stripes */
+	r5c_recovery_replay_stripes(&ctx->cached_list, ctx);
+
+	/* load data-only stripes to stripe cache */
+	list_for_each_entry(sh, &ctx->cached_list, lru) {
+		WARN_ON(!test_bit(STRIPE_R5C_CACHING, &sh->state));
+		r5c_recovery_load_one_stripe(log, sh);
+		ctx->data_only_stripes++;
+	}
+
+	return 0;
 }
 
-static int r5l_log_write_empty_meta_block(struct r5l_log *log, sector_t pos,
-					  u64 seq)
+/*
+ * we did a recovery. Now ctx.pos points to an invalid meta block. New
+ * log will start here. but we can't let superblock point to last valid
+ * meta block. The log might looks like:
+ * | meta 1| meta 2| meta 3|
+ * meta 1 is valid, meta 2 is invalid. meta 3 could be valid. If
+ * superblock points to meta 1, we write a new valid meta 2n.  if crash
+ * happens again, new recovery will start from meta 1. Since meta 2n is
+ * valid now, recovery will think meta 3 is valid, which is wrong.
+ * The solution is we create a new meta in meta2 with its seq == meta
+ * 1's seq + 10000 and let superblock points to meta2. The same recovery
+ * will not think meta 3 is a valid meta, because its seq doesn't match
+ */
+
+/*
+ * Before recovery, the log looks like the following
+ *
+ *   ---------------------------------------------
+ *   |           valid log        | invalid log  |
+ *   ---------------------------------------------
+ *   ^
+ *   |- log->last_checkpoint
+ *   |- log->last_cp_seq
+ *
+ * Now we scan through the log until we see invalid entry
+ *
+ *   ---------------------------------------------
+ *   |           valid log        | invalid log  |
+ *   ---------------------------------------------
+ *   ^                            ^
+ *   |- log->last_checkpoint      |- ctx->pos
+ *   |- log->last_cp_seq          |- ctx->seq
+ *
+ * From this point, we need to increase seq number by 10 to avoid
+ * confusing next recovery.
+ *
+ *   ---------------------------------------------
+ *   |           valid log        | invalid log  |
+ *   ---------------------------------------------
+ *   ^                              ^
+ *   |- log->last_checkpoint        |- ctx->pos+1
+ *   |- log->last_cp_seq            |- ctx->seq+10001
+ *
+ * However, it is not safe to start the state machine yet, because data only
+ * parities are not yet secured in RAID. To save these data only parities, we
+ * rewrite them from seq+11.
+ *
+ *   -----------------------------------------------------------------
+ *   |           valid log        | data only stripes | invalid log  |
+ *   -----------------------------------------------------------------
+ *   ^                                                ^
+ *   |- log->last_checkpoint                          |- ctx->pos+n
+ *   |- log->last_cp_seq                              |- ctx->seq+10000+n
+ *
+ * If failure happens again during this process, the recovery can safe start
+ * again from log->last_checkpoint.
+ *
+ * Once data only stripes are rewritten to journal, we move log_tail
+ *
+ *   -----------------------------------------------------------------
+ *   |     old log        |    data only stripes    | invalid log  |
+ *   -----------------------------------------------------------------
+ *                        ^                         ^
+ *                        |- log->last_checkpoint   |- ctx->pos+n
+ *                        |- log->last_cp_seq       |- ctx->seq+10000+n
+ *
+ * Then we can safely start the state machine. If failure happens from this
+ * point on, the recovery will start from new log->last_checkpoint.
+ */
+static int
+r5c_recovery_rewrite_data_only_stripes(struct r5l_log *log,
+				       struct r5l_recovery_ctx *ctx)
 {
+	struct stripe_head *sh, *next;
+	struct mddev *mddev = log->rdev->mddev;
 	struct page *page;
-	struct r5l_meta_block *mb;
-	u32 crc;
 
-	page = alloc_page(GFP_KERNEL | __GFP_ZERO);
-	if (!page)
+	page = alloc_page(GFP_KERNEL);
+	if (!page) {
+		pr_err("md/raid:%s: cannot allocate memory to rewrite data only stripes\n",
+		       mdname(mddev));
 		return -ENOMEM;
-	mb = page_address(page);
-	mb->magic = cpu_to_le32(R5LOG_MAGIC);
-	mb->version = R5LOG_VERSION;
-	mb->meta_size = cpu_to_le32(sizeof(struct r5l_meta_block));
-	mb->seq = cpu_to_le64(seq);
-	mb->position = cpu_to_le64(pos);
-	crc = crc32c_le(log->uuid_checksum, mb, PAGE_SIZE);
-	mb->checksum = cpu_to_le32(crc);
+	}
 
-	if (!sync_page_io(log->rdev, pos, PAGE_SIZE, page, REQ_OP_WRITE,
-			  REQ_FUA, false)) {
-		__free_page(page);
-		return -EIO;
+	list_for_each_entry_safe(sh, next, &ctx->cached_list, lru) {
+		struct r5l_meta_block *mb;
+		int i;
+		int offset;
+		sector_t write_pos;
+
+		WARN_ON(!test_bit(STRIPE_R5C_CACHING, &sh->state));
+		r5l_recovery_create_empty_meta_block(log, page,
+						     ctx->pos, ctx->seq);
+		mb = page_address(page);
+		offset = le32_to_cpu(mb->meta_size);
+		write_pos = r5l_ring_add(log, ctx->pos, BLOCK_SECTORS);
+
+		for (i = sh->disks; i--; ) {
+			struct r5dev *dev = &sh->dev[i];
+			struct r5l_payload_data_parity *payload;
+			void *addr;
+
+			if (test_bit(R5_InJournal, &dev->flags)) {
+				payload = (void *)mb + offset;
+				payload->header.type = cpu_to_le16(
+					R5LOG_PAYLOAD_DATA);
+				payload->size = BLOCK_SECTORS;
+				payload->location = cpu_to_le64(
+					raid5_compute_blocknr(sh, i, 0));
+				addr = kmap_atomic(dev->page);
+				payload->checksum[0] = cpu_to_le32(
+					crc32c_le(log->uuid_checksum, addr,
+						  PAGE_SIZE));
+				kunmap_atomic(addr);
+				sync_page_io(log->rdev, write_pos, PAGE_SIZE,
+					     dev->page, REQ_OP_WRITE, 0, false);
+				write_pos = r5l_ring_add(log, write_pos,
+							 BLOCK_SECTORS);
+				offset += sizeof(__le32) +
+					sizeof(struct r5l_payload_data_parity);
+
+			}
+		}
+		mb->meta_size = cpu_to_le32(offset);
+		mb->checksum = cpu_to_le32(crc32c_le(log->uuid_checksum,
+						     mb, PAGE_SIZE));
+		sync_page_io(log->rdev, ctx->pos, PAGE_SIZE, page,
+			     REQ_OP_WRITE, REQ_FUA, false);
+		sh->log_start = ctx->pos;
+		ctx->pos = write_pos;
+		ctx->seq += 1;
+
+		list_del_init(&sh->lru);
+		raid5_release_stripe(sh);
 	}
 	__free_page(page);
 	return 0;
@@ -1063,45 +2135,60 @@ static int r5l_log_write_empty_meta_block(struct r5l_log *log, sector_t pos,
 
 static int r5l_recovery_log(struct r5l_log *log)
 {
+	struct mddev *mddev = log->rdev->mddev;
 	struct r5l_recovery_ctx ctx;
+	int ret;
+	sector_t pos;
+	struct stripe_head *sh;
 
 	ctx.pos = log->last_checkpoint;
 	ctx.seq = log->last_cp_seq;
 	ctx.meta_page = alloc_page(GFP_KERNEL);
+	ctx.data_only_stripes = 0;
+	ctx.data_parity_stripes = 0;
+	INIT_LIST_HEAD(&ctx.cached_list);
+
 	if (!ctx.meta_page)
 		return -ENOMEM;
 
-	r5l_recovery_flush_log(log, &ctx);
+	ret = r5c_recovery_flush_log(log, &ctx);
 	__free_page(ctx.meta_page);
 
-	/*
-	 * we did a recovery. Now ctx.pos points to an invalid meta block. New
-	 * log will start here. but we can't let superblock point to last valid
-	 * meta block. The log might looks like:
-	 * | meta 1| meta 2| meta 3|
-	 * meta 1 is valid, meta 2 is invalid. meta 3 could be valid. If
-	 * superblock points to meta 1, we write a new valid meta 2n.  if crash
-	 * happens again, new recovery will start from meta 1. Since meta 2n is
-	 * valid now, recovery will think meta 3 is valid, which is wrong.
-	 * The solution is we create a new meta in meta2 with its seq == meta
-	 * 1's seq + 10 and let superblock points to meta2. The same recovery will
-	 * not think meta 3 is a valid meta, because its seq doesn't match
-	 */
-	if (ctx.seq > log->last_cp_seq) {
-		int ret;
-
-		ret = r5l_log_write_empty_meta_block(log, ctx.pos, ctx.seq + 10);
-		if (ret)
-			return ret;
-		log->seq = ctx.seq + 11;
-		log->log_start = r5l_ring_add(log, ctx.pos, BLOCK_SECTORS);
-		r5l_write_super(log, ctx.pos);
-		log->last_checkpoint = ctx.pos;
+	if (ret)
+		return ret;
+
+	pos = ctx.pos;
+	ctx.seq += 10000;
+
+	if (ctx.data_only_stripes == 0) {
 		log->next_checkpoint = ctx.pos;
+		r5l_log_write_empty_meta_block(log, ctx.pos, ctx.seq++);
+		ctx.pos = r5l_ring_add(log, ctx.pos, BLOCK_SECTORS);
 	} else {
-		log->log_start = ctx.pos;
-		log->seq = ctx.seq;
+		sh = list_last_entry(&ctx.cached_list, struct stripe_head, lru);
+		log->next_checkpoint = sh->log_start;
 	}
+
+	if ((ctx.data_only_stripes == 0) && (ctx.data_parity_stripes == 0))
+		pr_debug("md/raid:%s: starting from clean shutdown\n",
+			 mdname(mddev));
+	else {
+		pr_debug("md/raid:%s: recoverying %d data-only stripes and %d data-parity stripes\n",
+			 mdname(mddev), ctx.data_only_stripes,
+			 ctx.data_parity_stripes);
+
+		if (ctx.data_only_stripes > 0)
+			if (r5c_recovery_rewrite_data_only_stripes(log, &ctx)) {
+				pr_err("md/raid:%s: failed to rewrite stripes to journal\n",
+				       mdname(mddev));
+				return -EIO;
+			}
+	}
+
+	log->log_start = ctx.pos;
+	log->seq = ctx.seq;
+	log->last_checkpoint = pos;
+	r5l_write_super(log, pos);
 	return 0;
 }
 
@@ -1110,7 +2197,293 @@ static void r5l_write_super(struct r5l_log *log, sector_t cp)
 	struct mddev *mddev = log->rdev->mddev;
 
 	log->rdev->journal_tail = cp;
-	set_bit(MD_CHANGE_DEVS, &mddev->flags);
+	set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
+}
+
+static ssize_t r5c_journal_mode_show(struct mddev *mddev, char *page)
+{
+	struct r5conf *conf = mddev->private;
+	int ret;
+
+	if (!conf->log)
+		return 0;
+
+	switch (conf->log->r5c_journal_mode) {
+	case R5C_JOURNAL_MODE_WRITE_THROUGH:
+		ret = snprintf(
+			page, PAGE_SIZE, "[%s] %s\n",
+			r5c_journal_mode_str[R5C_JOURNAL_MODE_WRITE_THROUGH],
+			r5c_journal_mode_str[R5C_JOURNAL_MODE_WRITE_BACK]);
+		break;
+	case R5C_JOURNAL_MODE_WRITE_BACK:
+		ret = snprintf(
+			page, PAGE_SIZE, "%s [%s]\n",
+			r5c_journal_mode_str[R5C_JOURNAL_MODE_WRITE_THROUGH],
+			r5c_journal_mode_str[R5C_JOURNAL_MODE_WRITE_BACK]);
+		break;
+	default:
+		ret = 0;
+	}
+	return ret;
+}
+
+static ssize_t r5c_journal_mode_store(struct mddev *mddev,
+				      const char *page, size_t length)
+{
+	struct r5conf *conf = mddev->private;
+	struct r5l_log *log = conf->log;
+	int val = -1, i;
+	int len = length;
+
+	if (!log)
+		return -ENODEV;
+
+	if (len && page[len - 1] == '\n')
+		len -= 1;
+	for (i = 0; i < ARRAY_SIZE(r5c_journal_mode_str); i++)
+		if (strlen(r5c_journal_mode_str[i]) == len &&
+		    strncmp(page, r5c_journal_mode_str[i], len) == 0) {
+			val = i;
+			break;
+		}
+	if (val < R5C_JOURNAL_MODE_WRITE_THROUGH ||
+	    val > R5C_JOURNAL_MODE_WRITE_BACK)
+		return -EINVAL;
+
+	mddev_suspend(mddev);
+	conf->log->r5c_journal_mode = val;
+	mddev_resume(mddev);
+
+	pr_debug("md/raid:%s: setting r5c cache mode to %d: %s\n",
+		 mdname(mddev), val, r5c_journal_mode_str[val]);
+	return length;
+}
+
+struct md_sysfs_entry
+r5c_journal_mode = __ATTR(journal_mode, 0644,
+			  r5c_journal_mode_show, r5c_journal_mode_store);
+
+/*
+ * Try handle write operation in caching phase. This function should only
+ * be called in write-back mode.
+ *
+ * If all outstanding writes can be handled in caching phase, returns 0
+ * If writes requires write-out phase, call r5c_make_stripe_write_out()
+ * and returns -EAGAIN
+ */
+int r5c_try_caching_write(struct r5conf *conf,
+			  struct stripe_head *sh,
+			  struct stripe_head_state *s,
+			  int disks)
+{
+	struct r5l_log *log = conf->log;
+	int i;
+	struct r5dev *dev;
+	int to_cache = 0;
+
+	BUG_ON(!r5c_is_writeback(log));
+
+	if (!test_bit(STRIPE_R5C_CACHING, &sh->state)) {
+		/*
+		 * There are two different scenarios here:
+		 *  1. The stripe has some data cached, and it is sent to
+		 *     write-out phase for reclaim
+		 *  2. The stripe is clean, and this is the first write
+		 *
+		 * For 1, return -EAGAIN, so we continue with
+		 * handle_stripe_dirtying().
+		 *
+		 * For 2, set STRIPE_R5C_CACHING and continue with caching
+		 * write.
+		 */
+
+		/* case 1: anything injournal or anything in written */
+		if (s->injournal > 0 || s->written > 0)
+			return -EAGAIN;
+		/* case 2 */
+		set_bit(STRIPE_R5C_CACHING, &sh->state);
+	}
+
+	for (i = disks; i--; ) {
+		dev = &sh->dev[i];
+		/* if non-overwrite, use writing-out phase */
+		if (dev->towrite && !test_bit(R5_OVERWRITE, &dev->flags) &&
+		    !test_bit(R5_InJournal, &dev->flags)) {
+			r5c_make_stripe_write_out(sh);
+			return -EAGAIN;
+		}
+	}
+
+	for (i = disks; i--; ) {
+		dev = &sh->dev[i];
+		if (dev->towrite) {
+			set_bit(R5_Wantwrite, &dev->flags);
+			set_bit(R5_Wantdrain, &dev->flags);
+			set_bit(R5_LOCKED, &dev->flags);
+			to_cache++;
+		}
+	}
+
+	if (to_cache) {
+		set_bit(STRIPE_OP_BIODRAIN, &s->ops_request);
+		/*
+		 * set STRIPE_LOG_TRAPPED, which triggers r5c_cache_data()
+		 * in ops_run_io(). STRIPE_LOG_TRAPPED will be cleared in
+		 * r5c_handle_data_cached()
+		 */
+		set_bit(STRIPE_LOG_TRAPPED, &sh->state);
+	}
+
+	return 0;
+}
+
+/*
+ * free extra pages (orig_page) we allocated for prexor
+ */
+void r5c_release_extra_page(struct stripe_head *sh)
+{
+	struct r5conf *conf = sh->raid_conf;
+	int i;
+	bool using_disk_info_extra_page;
+
+	using_disk_info_extra_page =
+		sh->dev[0].orig_page == conf->disks[0].extra_page;
+
+	for (i = sh->disks; i--; )
+		if (sh->dev[i].page != sh->dev[i].orig_page) {
+			struct page *p = sh->dev[i].orig_page;
+
+			sh->dev[i].orig_page = sh->dev[i].page;
+			if (!using_disk_info_extra_page)
+				put_page(p);
+		}
+
+	if (using_disk_info_extra_page) {
+		clear_bit(R5C_EXTRA_PAGE_IN_USE, &conf->cache_state);
+		md_wakeup_thread(conf->mddev->thread);
+	}
+}
+
+void r5c_use_extra_page(struct stripe_head *sh)
+{
+	struct r5conf *conf = sh->raid_conf;
+	int i;
+	struct r5dev *dev;
+
+	for (i = sh->disks; i--; ) {
+		dev = &sh->dev[i];
+		if (dev->orig_page != dev->page)
+			put_page(dev->orig_page);
+		dev->orig_page = conf->disks[i].extra_page;
+	}
+}
+
+/*
+ * clean up the stripe (clear R5_InJournal for dev[pd_idx] etc.) after the
+ * stripe is committed to RAID disks.
+ */
+void r5c_finish_stripe_write_out(struct r5conf *conf,
+				 struct stripe_head *sh,
+				 struct stripe_head_state *s)
+{
+	int i;
+	int do_wakeup = 0;
+
+	if (!conf->log ||
+	    !test_bit(R5_InJournal, &sh->dev[sh->pd_idx].flags))
+		return;
+
+	WARN_ON(test_bit(STRIPE_R5C_CACHING, &sh->state));
+	clear_bit(R5_InJournal, &sh->dev[sh->pd_idx].flags);
+
+	if (conf->log->r5c_journal_mode == R5C_JOURNAL_MODE_WRITE_THROUGH)
+		return;
+
+	for (i = sh->disks; i--; ) {
+		clear_bit(R5_InJournal, &sh->dev[i].flags);
+		if (test_and_clear_bit(R5_Overlap, &sh->dev[i].flags))
+			do_wakeup = 1;
+	}
+
+	/*
+	 * analyse_stripe() runs before r5c_finish_stripe_write_out(),
+	 * We updated R5_InJournal, so we also update s->injournal.
+	 */
+	s->injournal = 0;
+
+	if (test_and_clear_bit(STRIPE_FULL_WRITE, &sh->state))
+		if (atomic_dec_and_test(&conf->pending_full_writes))
+			md_wakeup_thread(conf->mddev->thread);
+
+	if (do_wakeup)
+		wake_up(&conf->wait_for_overlap);
+
+	if (conf->log->r5c_journal_mode == R5C_JOURNAL_MODE_WRITE_THROUGH)
+		return;
+
+	spin_lock_irq(&conf->log->stripe_in_journal_lock);
+	list_del_init(&sh->r5c);
+	spin_unlock_irq(&conf->log->stripe_in_journal_lock);
+	sh->log_start = MaxSector;
+	atomic_dec(&conf->log->stripe_in_journal_count);
+	r5c_update_log_state(conf->log);
+}
+
+int
+r5c_cache_data(struct r5l_log *log, struct stripe_head *sh,
+	       struct stripe_head_state *s)
+{
+	struct r5conf *conf = sh->raid_conf;
+	int pages = 0;
+	int reserve;
+	int i;
+	int ret = 0;
+
+	BUG_ON(!log);
+
+	for (i = 0; i < sh->disks; i++) {
+		void *addr;
+
+		if (!test_bit(R5_Wantwrite, &sh->dev[i].flags))
+			continue;
+		addr = kmap_atomic(sh->dev[i].page);
+		sh->dev[i].log_checksum = crc32c_le(log->uuid_checksum,
+						    addr, PAGE_SIZE);
+		kunmap_atomic(addr);
+		pages++;
+	}
+	WARN_ON(pages == 0);
+
+	/*
+	 * The stripe must enter state machine again to call endio, so
+	 * don't delay.
+	 */
+	clear_bit(STRIPE_DELAYED, &sh->state);
+	atomic_inc(&sh->count);
+
+	mutex_lock(&log->io_mutex);
+	/* meta + data */
+	reserve = (1 + pages) << (PAGE_SHIFT - 9);
+
+	if (test_bit(R5C_LOG_CRITICAL, &conf->cache_state) &&
+	    sh->log_start == MaxSector)
+		r5l_add_no_space_stripe(log, sh);
+	else if (!r5l_has_free_space(log, reserve)) {
+		if (sh->log_start == log->last_checkpoint)
+			BUG();
+		else
+			r5l_add_no_space_stripe(log, sh);
+	} else {
+		ret = r5l_log_stripe(log, sh, pages, 0);
+		if (ret) {
+			spin_lock_irq(&log->io_list_lock);
+			list_add_tail(&sh->log_list, &log->no_mem_stripes);
+			spin_unlock_irq(&log->io_list_lock);
+		}
+	}
+
+	mutex_unlock(&log->io_mutex);
+	return 0;
 }
 
 static int r5l_load_log(struct r5l_log *log)
@@ -1121,7 +2494,7 @@ static int r5l_load_log(struct r5l_log *log)
 	sector_t cp = log->rdev->journal_tail;
 	u32 stored_crc, expected_crc;
 	bool create_super = false;
-	int ret;
+	int ret = 0;
 
 	/* Make sure it's valid */
 	if (cp >= rdev->sectors || round_down(cp, BLOCK_SECTORS) != cp)
@@ -1171,11 +2544,18 @@ create:
 	if (log->max_free_space > RECLAIM_MAX_FREE_SPACE)
 		log->max_free_space = RECLAIM_MAX_FREE_SPACE;
 	log->last_checkpoint = cp;
-	log->next_checkpoint = cp;
 
 	__free_page(page);
 
-	return r5l_recovery_log(log);
+	if (create_super) {
+		log->log_start = r5l_ring_add(log, cp, BLOCK_SECTORS);
+		log->seq = log->last_cp_seq + 1;
+		log->next_checkpoint = cp;
+	} else
+		ret = r5l_recovery_log(log);
+
+	r5c_update_log_state(log);
+	return ret;
 ioerr:
 	__free_page(page);
 	return ret;
@@ -1188,6 +2568,22 @@ int r5l_init_log(struct r5conf *conf, struct md_rdev *rdev)
 
 	if (PAGE_SIZE != 4096)
 		return -EINVAL;
+
+	/*
+	 * The PAGE_SIZE must be big enough to hold 1 r5l_meta_block and
+	 * raid_disks r5l_payload_data_parity.
+	 *
+	 * Write journal and cache does not work for very big array
+	 * (raid_disks > 203)
+	 */
+	if (sizeof(struct r5l_meta_block) +
+	    ((sizeof(struct r5l_payload_data_parity) + sizeof(__le32)) *
+	     conf->raid_disks) > PAGE_SIZE) {
+		pr_err("md/raid:%s: write journal/cache doesn't work for array with %d disks\n",
+		       mdname(conf->mddev), conf->raid_disks);
+		return -EINVAL;
+	}
+
 	log = kzalloc(sizeof(*log), GFP_KERNEL);
 	if (!log)
 		return -ENOMEM;
@@ -1227,6 +2623,8 @@ int r5l_init_log(struct r5conf *conf, struct md_rdev *rdev)
 						 log->rdev->mddev, "reclaim");
 	if (!log->reclaim_thread)
 		goto reclaim_thread;
+	log->reclaim_thread->timeout = R5C_RECLAIM_WAKEUP_INTERVAL;
+
 	init_waitqueue_head(&log->iounit_wait);
 
 	INIT_LIST_HEAD(&log->no_mem_stripes);
@@ -1234,6 +2632,13 @@ int r5l_init_log(struct r5conf *conf, struct md_rdev *rdev)
 	INIT_LIST_HEAD(&log->no_space_stripes);
 	spin_lock_init(&log->no_space_stripes_lock);
 
+	INIT_WORK(&log->deferred_io_work, r5l_submit_io_async);
+
+	log->r5c_journal_mode = R5C_JOURNAL_MODE_WRITE_THROUGH;
+	INIT_LIST_HEAD(&log->stripe_in_journal_list);
+	spin_lock_init(&log->stripe_in_journal_lock);
+	atomic_set(&log->stripe_in_journal_count, 0);
+
 	if (r5l_load_log(log))
 		goto error;
 

drivers/md/raid5.c

@@ -70,19 +70,6 @@ module_param(devices_handle_discard_safely, bool, 0644);
 MODULE_PARM_DESC(devices_handle_discard_safely,
 		 "Set to Y if all devices in each array reliably return zeroes on reads from discarded regions");
 static struct workqueue_struct *raid5_wq;
-/*
- * Stripe cache
- */
-
-#define NR_STRIPES		256
-#define STRIPE_SIZE		PAGE_SIZE
-#define STRIPE_SHIFT		(PAGE_SHIFT - 9)
-#define STRIPE_SECTORS		(STRIPE_SIZE>>9)
-#define	IO_THRESHOLD		1
-#define BYPASS_THRESHOLD	1
-#define NR_HASH			(PAGE_SIZE / sizeof(struct hlist_head))
-#define HASH_MASK		(NR_HASH - 1)
-#define MAX_STRIPE_BATCH	8
 
 static inline struct hlist_head *stripe_hash(struct r5conf *conf, sector_t sect)
 {
@@ -126,64 +113,6 @@ static inline void unlock_all_device_hash_locks_irq(struct r5conf *conf)
 	local_irq_enable();
 }
 
-/* bio's attached to a stripe+device for I/O are linked together in bi_sector
- * order without overlap.  There may be several bio's per stripe+device, and
- * a bio could span several devices.
- * When walking this list for a particular stripe+device, we must never proceed
- * beyond a bio that extends past this device, as the next bio might no longer
- * be valid.
- * This function is used to determine the 'next' bio in the list, given the sector
- * of the current stripe+device
- */
-static inline struct bio *r5_next_bio(struct bio *bio, sector_t sector)
-{
-	int sectors = bio_sectors(bio);
-	if (bio->bi_iter.bi_sector + sectors < sector + STRIPE_SECTORS)
-		return bio->bi_next;
-	else
-		return NULL;
-}
-
-/*
- * We maintain a biased count of active stripes in the bottom 16 bits of
- * bi_phys_segments, and a count of processed stripes in the upper 16 bits
- */
-static inline int raid5_bi_processed_stripes(struct bio *bio)
-{
-	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
-	return (atomic_read(segments) >> 16) & 0xffff;
-}
-
-static inline int raid5_dec_bi_active_stripes(struct bio *bio)
-{
-	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
-	return atomic_sub_return(1, segments) & 0xffff;
-}
-
-static inline void raid5_inc_bi_active_stripes(struct bio *bio)
-{
-	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
-	atomic_inc(segments);
-}
-
-static inline void raid5_set_bi_processed_stripes(struct bio *bio,
-	unsigned int cnt)
-{
-	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
-	int old, new;
-
-	do {
-		old = atomic_read(segments);
-		new = (old & 0xffff) | (cnt << 16);
-	} while (atomic_cmpxchg(segments, old, new) != old);
-}
-
-static inline void raid5_set_bi_stripes(struct bio *bio, unsigned int cnt)
-{
-	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
-	atomic_set(segments, cnt);
-}
-
 /* Find first data disk in a raid6 stripe */
 static inline int raid6_d0(struct stripe_head *sh)
 {
@@ -289,8 +218,27 @@ static void raid5_wakeup_stripe_thread(struct stripe_head *sh)
 static void do_release_stripe(struct r5conf *conf, struct stripe_head *sh,
 			      struct list_head *temp_inactive_list)
 {
+	int i;
+	int injournal = 0;	/* number of date pages with R5_InJournal */
+
 	BUG_ON(!list_empty(&sh->lru));
 	BUG_ON(atomic_read(&conf->active_stripes)==0);
+
+	if (r5c_is_writeback(conf->log))
+		for (i = sh->disks; i--; )
+			if (test_bit(R5_InJournal, &sh->dev[i].flags))
+				injournal++;
+	/*
+	 * When quiesce in r5c write back, set STRIPE_HANDLE for stripes with
+	 * data in journal, so they are not released to cached lists
+	 */
+	if (conf->quiesce && r5c_is_writeback(conf->log) &&
+	    !test_bit(STRIPE_HANDLE, &sh->state) && injournal != 0) {
+		if (test_bit(STRIPE_R5C_CACHING, &sh->state))
+			r5c_make_stripe_write_out(sh);
+		set_bit(STRIPE_HANDLE, &sh->state);
+	}
+
 	if (test_bit(STRIPE_HANDLE, &sh->state)) {
 		if (test_bit(STRIPE_DELAYED, &sh->state) &&
 		    !test_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
@@ -316,8 +264,30 @@ static void do_release_stripe(struct r5conf *conf, struct stripe_head *sh,
 			    < IO_THRESHOLD)
 				md_wakeup_thread(conf->mddev->thread);
 		atomic_dec(&conf->active_stripes);
-		if (!test_bit(STRIPE_EXPANDING, &sh->state))
-			list_add_tail(&sh->lru, temp_inactive_list);
+		if (!test_bit(STRIPE_EXPANDING, &sh->state)) {
+			if (!r5c_is_writeback(conf->log))
+				list_add_tail(&sh->lru, temp_inactive_list);
+			else {
+				WARN_ON(test_bit(R5_InJournal, &sh->dev[sh->pd_idx].flags));
+				if (injournal == 0)
+					list_add_tail(&sh->lru, temp_inactive_list);
+				else if (injournal == conf->raid_disks - conf->max_degraded) {
+					/* full stripe */
+					if (!test_and_set_bit(STRIPE_R5C_FULL_STRIPE, &sh->state))
+						atomic_inc(&conf->r5c_cached_full_stripes);
+					if (test_and_clear_bit(STRIPE_R5C_PARTIAL_STRIPE, &sh->state))
+						atomic_dec(&conf->r5c_cached_partial_stripes);
+					list_add_tail(&sh->lru, &conf->r5c_full_stripe_list);
+					r5c_check_cached_full_stripe(conf);
+				} else {
+					/* partial stripe */
+					if (!test_and_set_bit(STRIPE_R5C_PARTIAL_STRIPE,
+							      &sh->state))
+						atomic_inc(&conf->r5c_cached_partial_stripes);
+					list_add_tail(&sh->lru, &conf->r5c_partial_stripe_list);
+				}
+			}
+		}
 	}
 }
 
@@ -541,7 +511,7 @@ retry:
 
 		if (dev->toread || dev->read || dev->towrite || dev->written ||
 		    test_bit(R5_LOCKED, &dev->flags)) {
-			printk(KERN_ERR "sector=%llx i=%d %p %p %p %p %d\n",
+			pr_err("sector=%llx i=%d %p %p %p %p %d\n",
 			       (unsigned long long)sh->sector, i, dev->toread,
 			       dev->read, dev->towrite, dev->written,
 			       test_bit(R5_LOCKED, &dev->flags));
@@ -680,9 +650,12 @@ raid5_get_active_stripe(struct r5conf *conf, sector_t sector,
 			}
 			if (noblock && sh == NULL)
 				break;
+
+			r5c_check_stripe_cache_usage(conf);
 			if (!sh) {
 				set_bit(R5_INACTIVE_BLOCKED,
 					&conf->cache_state);
+				r5l_wake_reclaim(conf->log, 0);
 				wait_event_lock_irq(
 					conf->wait_for_stripe,
 					!list_empty(conf->inactive_list + hash) &&
@@ -901,8 +874,19 @@ static void ops_run_io(struct stripe_head *sh, struct stripe_head_state *s)
 
 	might_sleep();
 
-	if (r5l_write_stripe(conf->log, sh) == 0)
-		return;
+	if (!test_bit(STRIPE_R5C_CACHING, &sh->state)) {
+		/* writing out phase */
+		if (s->waiting_extra_page)
+			return;
+		if (r5l_write_stripe(conf->log, sh) == 0)
+			return;
+	} else {  /* caching phase */
+		if (test_bit(STRIPE_LOG_TRAPPED, &sh->state)) {
+			r5c_cache_data(conf->log, sh, s);
+			return;
+		}
+	}
+
 	for (i = disks; i--; ) {
 		int op, op_flags = 0;
 		int replace_only = 0;
@@ -977,7 +961,7 @@ again:
 			if (bad < 0) {
 				set_bit(BlockedBadBlocks, &rdev->flags);
 				if (!conf->mddev->external &&
-				    conf->mddev->flags) {
+				    conf->mddev->sb_flags) {
 					/* It is very unlikely, but we might
 					 * still need to write out the
 					 * bad block log - better give it
@@ -1115,7 +1099,7 @@ again:
 static struct dma_async_tx_descriptor *
 async_copy_data(int frombio, struct bio *bio, struct page **page,
 	sector_t sector, struct dma_async_tx_descriptor *tx,
-	struct stripe_head *sh)
+	struct stripe_head *sh, int no_skipcopy)
 {
 	struct bio_vec bvl;
 	struct bvec_iter iter;
@@ -1155,7 +1139,8 @@ async_copy_data(int frombio, struct bio *bio, struct page **page,
 			if (frombio) {
 				if (sh->raid_conf->skip_copy &&
 				    b_offset == 0 && page_offset == 0 &&
-				    clen == STRIPE_SIZE)
+				    clen == STRIPE_SIZE &&
+				    !no_skipcopy)
 					*page = bio_page;
 				else
 					tx = async_memcpy(*page, bio_page, page_offset,
@@ -1237,7 +1222,7 @@ static void ops_run_biofill(struct stripe_head *sh)
 			while (rbi && rbi->bi_iter.bi_sector <
 				dev->sector + STRIPE_SECTORS) {
 				tx = async_copy_data(0, rbi, &dev->page,
-					dev->sector, tx, sh);
+						     dev->sector, tx, sh, 0);
 				rbi = r5_next_bio(rbi, dev->sector);
 			}
 		}
@@ -1364,10 +1349,15 @@ static int set_syndrome_sources(struct page **srcs,
 		if (i == sh->qd_idx || i == sh->pd_idx ||
 		    (srctype == SYNDROME_SRC_ALL) ||
 		    (srctype == SYNDROME_SRC_WANT_DRAIN &&
-		     test_bit(R5_Wantdrain, &dev->flags)) ||
+		     (test_bit(R5_Wantdrain, &dev->flags) ||
+		      test_bit(R5_InJournal, &dev->flags))) ||
 		    (srctype == SYNDROME_SRC_WRITTEN &&
-		     dev->written))
-			srcs[slot] = sh->dev[i].page;
+		     dev->written)) {
+			if (test_bit(R5_InJournal, &dev->flags))
+				srcs[slot] = sh->dev[i].orig_page;
+			else
+				srcs[slot] = sh->dev[i].page;
+		}
 		i = raid6_next_disk(i, disks);
 	} while (i != d0_idx);
 
@@ -1546,6 +1536,13 @@ static void ops_complete_prexor(void *stripe_head_ref)
 
 	pr_debug("%s: stripe %llu\n", __func__,
 		(unsigned long long)sh->sector);
+
+	if (r5c_is_writeback(sh->raid_conf->log))
+		/*
+		 * raid5-cache write back uses orig_page during prexor.
+		 * After prexor, it is time to free orig_page
+		 */
+		r5c_release_extra_page(sh);
 }
 
 static struct dma_async_tx_descriptor *
@@ -1567,7 +1564,9 @@ ops_run_prexor5(struct stripe_head *sh, struct raid5_percpu *percpu,
 	for (i = disks; i--; ) {
 		struct r5dev *dev = &sh->dev[i];
 		/* Only process blocks that are known to be uptodate */
-		if (test_bit(R5_Wantdrain, &dev->flags))
+		if (test_bit(R5_InJournal, &dev->flags))
+			xor_srcs[count++] = dev->orig_page;
+		else if (test_bit(R5_Wantdrain, &dev->flags))
 			xor_srcs[count++] = dev->page;
 	}
 
@@ -1601,6 +1600,7 @@ ops_run_prexor6(struct stripe_head *sh, struct raid5_percpu *percpu,
 static struct dma_async_tx_descriptor *
 ops_run_biodrain(struct stripe_head *sh, struct dma_async_tx_descriptor *tx)
 {
+	struct r5conf *conf = sh->raid_conf;
 	int disks = sh->disks;
 	int i;
 	struct stripe_head *head_sh = sh;
@@ -1618,6 +1618,11 @@ ops_run_biodrain(struct stripe_head *sh, struct dma_async_tx_descriptor *tx)
 
 again:
 			dev = &sh->dev[i];
+			/*
+			 * clear R5_InJournal, so when rewriting a page in
+			 * journal, it is not skipped by r5l_log_stripe()
+			 */
+			clear_bit(R5_InJournal, &dev->flags);
 			spin_lock_irq(&sh->stripe_lock);
 			chosen = dev->towrite;
 			dev->towrite = NULL;
@@ -1637,8 +1642,10 @@ again:
 					set_bit(R5_Discard, &dev->flags);
 				else {
 					tx = async_copy_data(1, wbi, &dev->page,
-						dev->sector, tx, sh);
-					if (dev->page != dev->orig_page) {
+							     dev->sector, tx, sh,
+							     r5c_is_writeback(conf->log));
+					if (dev->page != dev->orig_page &&
+					    !r5c_is_writeback(conf->log)) {
 						set_bit(R5_SkipCopy, &dev->flags);
 						clear_bit(R5_UPTODATE, &dev->flags);
 						clear_bit(R5_OVERWRITE, &dev->flags);
@@ -1746,7 +1753,8 @@ again:
 		xor_dest = xor_srcs[count++] = sh->dev[pd_idx].page;
 		for (i = disks; i--; ) {
 			struct r5dev *dev = &sh->dev[i];
-			if (head_sh->dev[i].written)
+			if (head_sh->dev[i].written ||
+			    test_bit(R5_InJournal, &head_sh->dev[i].flags))
 				xor_srcs[count++] = dev->page;
 		}
 	} else {
@@ -2000,7 +2008,10 @@ static struct stripe_head *alloc_stripe(struct kmem_cache *sc, gfp_t gfp,
 		spin_lock_init(&sh->batch_lock);
 		INIT_LIST_HEAD(&sh->batch_list);
 		INIT_LIST_HEAD(&sh->lru);
+		INIT_LIST_HEAD(&sh->r5c);
+		INIT_LIST_HEAD(&sh->log_list);
 		atomic_set(&sh->count, 1);
+		sh->log_start = MaxSector;
 		for (i = 0; i < disks; i++) {
 			struct r5dev *dev = &sh->dev[i];
 
@@ -2240,10 +2251,24 @@ static int resize_stripes(struct r5conf *conf, int newsize)
 	 */
 	ndisks = kzalloc(newsize * sizeof(struct disk_info), GFP_NOIO);
 	if (ndisks) {
-		for (i=0; i<conf->raid_disks; i++)
+		for (i = 0; i < conf->pool_size; i++)
 			ndisks[i] = conf->disks[i];
-		kfree(conf->disks);
-		conf->disks = ndisks;
+
+		for (i = conf->pool_size; i < newsize; i++) {
+			ndisks[i].extra_page = alloc_page(GFP_NOIO);
+			if (!ndisks[i].extra_page)
+				err = -ENOMEM;
+		}
+
+		if (err) {
+			for (i = conf->pool_size; i < newsize; i++)
+				if (ndisks[i].extra_page)
+					put_page(ndisks[i].extra_page);
+			kfree(ndisks);
+		} else {
+			kfree(conf->disks);
+			conf->disks = ndisks;
+		}
 	} else
 		err = -ENOMEM;
 
@@ -2342,10 +2367,8 @@ static void raid5_end_read_request(struct bio * bi)
 			 * replacement device.  We just fail those on
 			 * any error
 			 */
-			printk_ratelimited(
-				KERN_INFO
-				"md/raid:%s: read error corrected"
-				" (%lu sectors at %llu on %s)\n",
+			pr_info_ratelimited(
+				"md/raid:%s: read error corrected (%lu sectors at %llu on %s)\n",
 				mdname(conf->mddev), STRIPE_SECTORS,
 				(unsigned long long)s,
 				bdevname(rdev->bdev, b));
@@ -2365,36 +2388,29 @@ static void raid5_end_read_request(struct bio * bi)
 		clear_bit(R5_UPTODATE, &sh->dev[i].flags);
 		atomic_inc(&rdev->read_errors);
 		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",
+			pr_warn_ratelimited(
+				"md/raid:%s: read error on replacement device (sector %llu on %s).\n",
 				mdname(conf->mddev),
 				(unsigned long long)s,
 				bdn);
 		else if (conf->mddev->degraded >= conf->max_degraded) {
 			set_bad = 1;
-			printk_ratelimited(
-				KERN_WARNING
-				"md/raid:%s: read error not correctable "
-				"(sector %llu on %s).\n",
+			pr_warn_ratelimited(
+				"md/raid:%s: read error not correctable (sector %llu on %s).\n",
 				mdname(conf->mddev),
 				(unsigned long long)s,
 				bdn);
 		} else if (test_bit(R5_ReWrite, &sh->dev[i].flags)) {
 			/* Oh, no!!! */
 			set_bad = 1;
-			printk_ratelimited(
-				KERN_WARNING
-				"md/raid:%s: read error NOT corrected!! "
-				"(sector %llu on %s).\n",
+			pr_warn_ratelimited(
+				"md/raid:%s: read error NOT corrected!! (sector %llu on %s).\n",
 				mdname(conf->mddev),
 				(unsigned long long)s,
 				bdn);
 		} else if (atomic_read(&rdev->read_errors)
 			 > conf->max_nr_stripes)
-			printk(KERN_WARNING
-			       "md/raid:%s: Too many read errors, failing device %s.\n",
+			pr_warn("md/raid:%s: Too many read errors, failing device %s.\n",
 			       mdname(conf->mddev), bdn);
 		else
 			retry = 1;
@@ -2526,15 +2542,14 @@ static void raid5_error(struct mddev *mddev, struct md_rdev *rdev)
 
 	set_bit(Blocked, &rdev->flags);
 	set_bit(Faulty, &rdev->flags);
-	set_mask_bits(&mddev->flags, 0,
-		      BIT(MD_CHANGE_DEVS) | BIT(MD_CHANGE_PENDING));
-	printk(KERN_ALERT
-	       "md/raid:%s: Disk failure on %s, disabling device.\n"
-	       "md/raid:%s: Operation continuing on %d devices.\n",
-	       mdname(mddev),
-	       bdevname(rdev->bdev, b),
-	       mdname(mddev),
-	       conf->raid_disks - mddev->degraded);
+	set_mask_bits(&mddev->sb_flags, 0,
+		      BIT(MD_SB_CHANGE_DEVS) | BIT(MD_SB_CHANGE_PENDING));
+	pr_crit("md/raid:%s: Disk failure on %s, disabling device.\n"
+		"md/raid:%s: Operation continuing on %d devices.\n",
+		mdname(mddev),
+		bdevname(rdev->bdev, b),
+		mdname(mddev),
+		conf->raid_disks - mddev->degraded);
 }
 
 /*
@@ -2856,8 +2871,8 @@ sector_t raid5_compute_blocknr(struct stripe_head *sh, int i, int previous)
 				     previous, &dummy1, &sh2);
 	if (check != sh->sector || dummy1 != dd_idx || sh2.pd_idx != sh->pd_idx
 		|| sh2.qd_idx != sh->qd_idx) {
-		printk(KERN_ERR "md/raid:%s: compute_blocknr: map not correct\n",
-		       mdname(conf->mddev));
+		pr_warn("md/raid:%s: compute_blocknr: map not correct\n",
+			mdname(conf->mddev));
 		return 0;
 	}
 	return r_sector;
@@ -2872,6 +2887,13 @@ schedule_reconstruction(struct stripe_head *sh, struct stripe_head_state *s,
 	int level = conf->level;
 
 	if (rcw) {
+		/*
+		 * In some cases, handle_stripe_dirtying initially decided to
+		 * run rmw and allocates extra page for prexor. However, rcw is
+		 * cheaper later on. We need to free the extra page now,
+		 * because we won't be able to do that in ops_complete_prexor().
+		 */
+		r5c_release_extra_page(sh);
 
 		for (i = disks; i--; ) {
 			struct r5dev *dev = &sh->dev[i];
@@ -2882,6 +2904,9 @@ schedule_reconstruction(struct stripe_head *sh, struct stripe_head_state *s,
 				if (!expand)
 					clear_bit(R5_UPTODATE, &dev->flags);
 				s->locked++;
+			} else if (test_bit(R5_InJournal, &dev->flags)) {
+				set_bit(R5_LOCKED, &dev->flags);
+				s->locked++;
 			}
 		}
 		/* if we are not expanding this is a proper write request, and
@@ -2921,6 +2946,9 @@ schedule_reconstruction(struct stripe_head *sh, struct stripe_head_state *s,
 				set_bit(R5_LOCKED, &dev->flags);
 				clear_bit(R5_UPTODATE, &dev->flags);
 				s->locked++;
+			} else if (test_bit(R5_InJournal, &dev->flags)) {
+				set_bit(R5_LOCKED, &dev->flags);
+				s->locked++;
 			}
 		}
 		if (!s->locked)
@@ -3564,10 +3592,10 @@ unhash:
 		break_stripe_batch_list(head_sh, STRIPE_EXPAND_SYNC_FLAGS);
 }
 
-static void handle_stripe_dirtying(struct r5conf *conf,
-				   struct stripe_head *sh,
-				   struct stripe_head_state *s,
-				   int disks)
+static int handle_stripe_dirtying(struct r5conf *conf,
+				  struct stripe_head *sh,
+				  struct stripe_head_state *s,
+				  int disks)
 {
 	int rmw = 0, rcw = 0, i;
 	sector_t recovery_cp = conf->mddev->recovery_cp;
@@ -3592,9 +3620,12 @@ static void handle_stripe_dirtying(struct r5conf *conf,
 	} else for (i = disks; i--; ) {
 		/* would I have to read this buffer for read_modify_write */
 		struct r5dev *dev = &sh->dev[i];
-		if ((dev->towrite || i == sh->pd_idx || i == sh->qd_idx) &&
+		if ((dev->towrite || i == sh->pd_idx || i == sh->qd_idx ||
+		     test_bit(R5_InJournal, &dev->flags)) &&
 		    !test_bit(R5_LOCKED, &dev->flags) &&
-		    !(test_bit(R5_UPTODATE, &dev->flags) ||
+		    !((test_bit(R5_UPTODATE, &dev->flags) &&
+		       (!test_bit(R5_InJournal, &dev->flags) ||
+			dev->page != dev->orig_page)) ||
 		      test_bit(R5_Wantcompute, &dev->flags))) {
 			if (test_bit(R5_Insync, &dev->flags))
 				rmw++;
@@ -3606,13 +3637,15 @@ static void handle_stripe_dirtying(struct r5conf *conf,
 		    i != sh->pd_idx && i != sh->qd_idx &&
 		    !test_bit(R5_LOCKED, &dev->flags) &&
 		    !(test_bit(R5_UPTODATE, &dev->flags) ||
-		    test_bit(R5_Wantcompute, &dev->flags))) {
+		      test_bit(R5_InJournal, &dev->flags) ||
+		      test_bit(R5_Wantcompute, &dev->flags))) {
 			if (test_bit(R5_Insync, &dev->flags))
 				rcw++;
 			else
 				rcw += 2*disks;
 		}
 	}
+
 	pr_debug("for sector %llu, rmw=%d rcw=%d\n",
 		(unsigned long long)sh->sector, rmw, rcw);
 	set_bit(STRIPE_HANDLE, &sh->state);
@@ -3624,10 +3657,44 @@ static void handle_stripe_dirtying(struct r5conf *conf,
 					  (unsigned long long)sh->sector, rmw);
 		for (i = disks; i--; ) {
 			struct r5dev *dev = &sh->dev[i];
-			if ((dev->towrite || i == sh->pd_idx || i == sh->qd_idx) &&
+			if (test_bit(R5_InJournal, &dev->flags) &&
+			    dev->page == dev->orig_page &&
+			    !test_bit(R5_LOCKED, &sh->dev[sh->pd_idx].flags)) {
+				/* alloc page for prexor */
+				struct page *p = alloc_page(GFP_NOIO);
+
+				if (p) {
+					dev->orig_page = p;
+					continue;
+				}
+
+				/*
+				 * alloc_page() failed, try use
+				 * disk_info->extra_page
+				 */
+				if (!test_and_set_bit(R5C_EXTRA_PAGE_IN_USE,
+						      &conf->cache_state)) {
+					r5c_use_extra_page(sh);
+					break;
+				}
+
+				/* extra_page in use, add to delayed_list */
+				set_bit(STRIPE_DELAYED, &sh->state);
+				s->waiting_extra_page = 1;
+				return -EAGAIN;
+			}
+		}
+
+		for (i = disks; i--; ) {
+			struct r5dev *dev = &sh->dev[i];
+			if ((dev->towrite ||
+			     i == sh->pd_idx || i == sh->qd_idx ||
+			     test_bit(R5_InJournal, &dev->flags)) &&
 			    !test_bit(R5_LOCKED, &dev->flags) &&
-			    !(test_bit(R5_UPTODATE, &dev->flags) ||
-			    test_bit(R5_Wantcompute, &dev->flags)) &&
+			    !((test_bit(R5_UPTODATE, &dev->flags) &&
+			       (!test_bit(R5_InJournal, &dev->flags) ||
+				dev->page != dev->orig_page)) ||
+			      test_bit(R5_Wantcompute, &dev->flags)) &&
 			    test_bit(R5_Insync, &dev->flags)) {
 				if (test_bit(STRIPE_PREREAD_ACTIVE,
 					     &sh->state)) {
@@ -3653,6 +3720,7 @@ static void handle_stripe_dirtying(struct r5conf *conf,
 			    i != sh->pd_idx && i != sh->qd_idx &&
 			    !test_bit(R5_LOCKED, &dev->flags) &&
 			    !(test_bit(R5_UPTODATE, &dev->flags) ||
+			      test_bit(R5_InJournal, &dev->flags) ||
 			      test_bit(R5_Wantcompute, &dev->flags))) {
 				rcw++;
 				if (test_bit(R5_Insync, &dev->flags) &&
@@ -3692,8 +3760,9 @@ static void handle_stripe_dirtying(struct r5conf *conf,
 	 */
 	if ((s->req_compute || !test_bit(STRIPE_COMPUTE_RUN, &sh->state)) &&
 	    (s->locked == 0 && (rcw == 0 || rmw == 0) &&
-	    !test_bit(STRIPE_BIT_DELAY, &sh->state)))
+	     !test_bit(STRIPE_BIT_DELAY, &sh->state)))
 		schedule_reconstruction(sh, s, rcw == 0, 0);
+	return 0;
 }
 
 static void handle_parity_checks5(struct r5conf *conf, struct stripe_head *sh,
@@ -3777,7 +3846,7 @@ static void handle_parity_checks5(struct r5conf *conf, struct stripe_head *sh,
 	case check_state_compute_run:
 		break;
 	default:
-		printk(KERN_ERR "%s: unknown check_state: %d sector: %llu\n",
+		pr_err("%s: unknown check_state: %d sector: %llu\n",
 		       __func__, sh->check_state,
 		       (unsigned long long) sh->sector);
 		BUG();
@@ -3941,9 +4010,9 @@ static void handle_parity_checks6(struct r5conf *conf, struct stripe_head *sh,
 	case check_state_compute_run:
 		break;
 	default:
-		printk(KERN_ERR "%s: unknown check_state: %d sector: %llu\n",
-		       __func__, sh->check_state,
-		       (unsigned long long) sh->sector);
+		pr_warn("%s: unknown check_state: %d sector: %llu\n",
+			__func__, sh->check_state,
+			(unsigned long long) sh->sector);
 		BUG();
 	}
 }
@@ -4183,6 +4252,11 @@ static void analyse_stripe(struct stripe_head *sh, struct stripe_head_state *s)
 			if (rdev && !test_bit(Faulty, &rdev->flags))
 				do_recovery = 1;
 		}
+
+		if (test_bit(R5_InJournal, &dev->flags))
+			s->injournal++;
+		if (test_bit(R5_InJournal, &dev->flags) && dev->written)
+			s->just_cached++;
 	}
 	if (test_bit(STRIPE_SYNCING, &sh->state)) {
 		/* If there is a failed device being replaced,
@@ -4411,7 +4485,8 @@ static void handle_stripe(struct stripe_head *sh)
 			struct r5dev *dev = &sh->dev[i];
 			if (test_bit(R5_LOCKED, &dev->flags) &&
 				(i == sh->pd_idx || i == sh->qd_idx ||
-				 dev->written)) {
+				 dev->written || test_bit(R5_InJournal,
+							  &dev->flags))) {
 				pr_debug("Writing block %d\n", i);
 				set_bit(R5_Wantwrite, &dev->flags);
 				if (prexor)
@@ -4451,6 +4526,10 @@ static void handle_stripe(struct stripe_head *sh)
 				 test_bit(R5_Discard, &qdev->flags))))))
 		handle_stripe_clean_event(conf, sh, disks, &s.return_bi);
 
+	if (s.just_cached)
+		r5c_handle_cached_data_endio(conf, sh, disks, &s.return_bi);
+	r5l_stripe_write_finished(sh);
+
 	/* Now we might consider reading some blocks, either to check/generate
 	 * parity, or to satisfy requests
 	 * or to load a block that is being partially written.
@@ -4462,14 +4541,51 @@ static void handle_stripe(struct stripe_head *sh)
 	    || s.expanding)
 		handle_stripe_fill(sh, &s, disks);
 
-	/* Now to consider new write requests and what else, if anything
-	 * should be read.  We do not handle new writes when:
+	/*
+	 * When the stripe finishes full journal write cycle (write to journal
+	 * and raid disk), this is the clean up procedure so it is ready for
+	 * next operation.
+	 */
+	r5c_finish_stripe_write_out(conf, sh, &s);
+
+	/*
+	 * Now to consider new write requests, cache write back and what else,
+	 * if anything should be read.  We do not handle new writes when:
 	 * 1/ A 'write' operation (copy+xor) is already in flight.
 	 * 2/ A 'check' operation is in flight, as it may clobber the parity
 	 *    block.
+	 * 3/ A r5c cache log write is in flight.
 	 */
-	if (s.to_write && !sh->reconstruct_state && !sh->check_state)
-		handle_stripe_dirtying(conf, sh, &s, disks);
+
+	if (!sh->reconstruct_state && !sh->check_state && !sh->log_io) {
+		if (!r5c_is_writeback(conf->log)) {
+			if (s.to_write)
+				handle_stripe_dirtying(conf, sh, &s, disks);
+		} else { /* write back cache */
+			int ret = 0;
+
+			/* First, try handle writes in caching phase */
+			if (s.to_write)
+				ret = r5c_try_caching_write(conf, sh, &s,
+							    disks);
+			/*
+			 * If caching phase failed: ret == -EAGAIN
+			 *    OR
+			 * stripe under reclaim: !caching && injournal
+			 *
+			 * fall back to handle_stripe_dirtying()
+			 */
+			if (ret == -EAGAIN ||
+			    /* stripe under reclaim: !caching && injournal */
+			    (!test_bit(STRIPE_R5C_CACHING, &sh->state) &&
+			     s.injournal > 0)) {
+				ret = handle_stripe_dirtying(conf, sh, &s,
+							     disks);
+				if (ret == -EAGAIN)
+					goto finish;
+			}
+		}
+	}
 
 	/* maybe we need to check and possibly fix the parity for this stripe
 	 * Any reads will already have been scheduled, so we just see if enough
@@ -4640,9 +4756,7 @@ finish:
 	}
 
 	if (!bio_list_empty(&s.return_bi)) {
-		if (test_bit(MD_CHANGE_PENDING, &conf->mddev->flags) &&
-				(s.failed <= conf->max_degraded ||
-					conf->mddev->external == 0)) {
+		if (test_bit(MD_SB_CHANGE_PENDING, &conf->mddev->sb_flags)) {
 			spin_lock_irq(&conf->device_lock);
 			bio_list_merge(&conf->return_bi, &s.return_bi);
 			spin_unlock_irq(&conf->device_lock);
@@ -4698,6 +4812,10 @@ static int raid5_congested(struct mddev *mddev, int bits)
 
 	if (test_bit(R5_INACTIVE_BLOCKED, &conf->cache_state))
 		return 1;
+
+	/* Also checks whether there is pressure on r5cache log space */
+	if (test_bit(R5C_LOG_TIGHT, &conf->cache_state))
+		return 1;
 	if (conf->quiesce)
 		return 1;
 	if (atomic_read(&conf->empty_inactive_list_nr))
@@ -5167,6 +5285,7 @@ static void raid5_make_request(struct mddev *mddev, struct bio * bi)
 	int remaining;
 	DEFINE_WAIT(w);
 	bool do_prepare;
+	bool do_flush = false;
 
 	if (unlikely(bi->bi_opf & REQ_PREFLUSH)) {
 		int ret = r5l_handle_flush_request(conf->log, bi);
@@ -5178,6 +5297,11 @@ static void raid5_make_request(struct mddev *mddev, struct bio * bi)
 			return;
 		}
 		/* ret == -EAGAIN, fallback */
+		/*
+		 * if r5l_handle_flush_request() didn't clear REQ_PREFLUSH,
+		 * we need to flush journal device
+		 */
+		do_flush = bi->bi_opf & REQ_PREFLUSH;
 	}
 
 	md_write_start(mddev, bi);
@@ -5188,6 +5312,7 @@ static void raid5_make_request(struct mddev *mddev, struct bio * bi)
 	 * data on failed drives.
 	 */
 	if (rw == READ && mddev->degraded == 0 &&
+	    !r5c_is_writeback(conf->log) &&
 	    mddev->reshape_position == MaxSector) {
 		bi = chunk_aligned_read(mddev, bi);
 		if (!bi)
@@ -5316,6 +5441,12 @@ static void raid5_make_request(struct mddev *mddev, struct bio * bi)
 				do_prepare = true;
 				goto retry;
 			}
+			if (do_flush) {
+				set_bit(STRIPE_R5C_PREFLUSH, &sh->state);
+				/* we only need flush for one stripe */
+				do_flush = false;
+			}
+
 			set_bit(STRIPE_HANDLE, &sh->state);
 			clear_bit(STRIPE_DELAYED, &sh->state);
 			if ((!sh->batch_head || sh == sh->batch_head) &&
@@ -5481,9 +5612,9 @@ static sector_t reshape_request(struct mddev *mddev, sector_t sector_nr, int *sk
 		mddev->reshape_position = conf->reshape_progress;
 		mddev->curr_resync_completed = sector_nr;
 		conf->reshape_checkpoint = jiffies;
-		set_bit(MD_CHANGE_DEVS, &mddev->flags);
+		set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
 		md_wakeup_thread(mddev->thread);
-		wait_event(mddev->sb_wait, mddev->flags == 0 ||
+		wait_event(mddev->sb_wait, mddev->sb_flags == 0 ||
 			   test_bit(MD_RECOVERY_INTR, &mddev->recovery));
 		if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
 			return 0;
@@ -5579,10 +5710,10 @@ finish:
 		mddev->reshape_position = conf->reshape_progress;
 		mddev->curr_resync_completed = sector_nr;
 		conf->reshape_checkpoint = jiffies;
-		set_bit(MD_CHANGE_DEVS, &mddev->flags);
+		set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
 		md_wakeup_thread(mddev->thread);
 		wait_event(mddev->sb_wait,
-			   !test_bit(MD_CHANGE_DEVS, &mddev->flags)
+			   !test_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags)
 			   || test_bit(MD_RECOVERY_INTR, &mddev->recovery));
 		if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
 			goto ret;
@@ -5857,10 +5988,10 @@ static void raid5d(struct md_thread *thread)
 	md_check_recovery(mddev);
 
 	if (!bio_list_empty(&conf->return_bi) &&
-	    !test_bit(MD_CHANGE_PENDING, &mddev->flags)) {
+	    !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags)) {
 		struct bio_list tmp = BIO_EMPTY_LIST;
 		spin_lock_irq(&conf->device_lock);
-		if (!test_bit(MD_CHANGE_PENDING, &mddev->flags)) {
+		if (!test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags)) {
 			bio_list_merge(&tmp, &conf->return_bi);
 			bio_list_init(&conf->return_bi);
 		}
@@ -5907,7 +6038,7 @@ static void raid5d(struct md_thread *thread)
 			break;
 		handled += batch_size;
 
-		if (mddev->flags & ~(1<<MD_CHANGE_PENDING)) {
+		if (mddev->sb_flags & ~(1 << MD_SB_CHANGE_PENDING)) {
 			spin_unlock_irq(&conf->device_lock);
 			md_check_recovery(mddev);
 			spin_lock_irq(&conf->device_lock);
@@ -6237,6 +6368,7 @@ static struct attribute *raid5_attrs[] =  {
 	&raid5_group_thread_cnt.attr,
 	&raid5_skip_copy.attr,
 	&raid5_rmw_level.attr,
+	&r5c_journal_mode.attr,
 	NULL,
 };
 static struct attribute_group raid5_attrs_group = {
@@ -6363,6 +6495,8 @@ static void raid5_free_percpu(struct r5conf *conf)
 
 static void free_conf(struct r5conf *conf)
 {
+	int i;
+
 	if (conf->log)
 		r5l_exit_log(conf->log);
 	if (conf->shrinker.nr_deferred)
@@ -6371,6 +6505,9 @@ static void free_conf(struct r5conf *conf)
 	free_thread_groups(conf);
 	shrink_stripes(conf);
 	raid5_free_percpu(conf);
+	for (i = 0; i < conf->pool_size; i++)
+		if (conf->disks[i].extra_page)
+			put_page(conf->disks[i].extra_page);
 	kfree(conf->disks);
 	kfree(conf->stripe_hashtbl);
 	kfree(conf);
@@ -6382,8 +6519,8 @@ static int raid456_cpu_up_prepare(unsigned int cpu, struct hlist_node *node)
 	struct raid5_percpu *percpu = per_cpu_ptr(conf->percpu, cpu);
 
 	if (alloc_scratch_buffer(conf, percpu)) {
-		pr_err("%s: failed memory allocation for cpu%u\n",
-		       __func__, cpu);
+		pr_warn("%s: failed memory allocation for cpu%u\n",
+			__func__, cpu);
 		return -ENOMEM;
 	}
 	return 0;
@@ -6453,29 +6590,29 @@ static struct r5conf *setup_conf(struct mddev *mddev)
 	if (mddev->new_level != 5
 	    && mddev->new_level != 4
 	    && mddev->new_level != 6) {
-		printk(KERN_ERR "md/raid:%s: raid level not set to 4/5/6 (%d)\n",
-		       mdname(mddev), mddev->new_level);
+		pr_warn("md/raid:%s: raid level not set to 4/5/6 (%d)\n",
+			mdname(mddev), mddev->new_level);
 		return ERR_PTR(-EIO);
 	}
 	if ((mddev->new_level == 5
 	     && !algorithm_valid_raid5(mddev->new_layout)) ||
 	    (mddev->new_level == 6
 	     && !algorithm_valid_raid6(mddev->new_layout))) {
-		printk(KERN_ERR "md/raid:%s: layout %d not supported\n",
-		       mdname(mddev), mddev->new_layout);
+		pr_warn("md/raid:%s: layout %d not supported\n",
+			mdname(mddev), mddev->new_layout);
 		return ERR_PTR(-EIO);
 	}
 	if (mddev->new_level == 6 && mddev->raid_disks < 4) {
-		printk(KERN_ERR "md/raid:%s: not enough configured devices (%d, minimum 4)\n",
-		       mdname(mddev), mddev->raid_disks);
+		pr_warn("md/raid:%s: not enough configured devices (%d, minimum 4)\n",
+			mdname(mddev), mddev->raid_disks);
 		return ERR_PTR(-EINVAL);
 	}
 
 	if (!mddev->new_chunk_sectors ||
 	    (mddev->new_chunk_sectors << 9) % PAGE_SIZE ||
 	    !is_power_of_2(mddev->new_chunk_sectors)) {
-		printk(KERN_ERR "md/raid:%s: invalid chunk size %d\n",
-		       mdname(mddev), mddev->new_chunk_sectors << 9);
+		pr_warn("md/raid:%s: invalid chunk size %d\n",
+			mdname(mddev), mddev->new_chunk_sectors << 9);
 		return ERR_PTR(-EINVAL);
 	}
 
@@ -6517,9 +6654,16 @@ static struct r5conf *setup_conf(struct mddev *mddev)
 
 	conf->disks = kzalloc(max_disks * sizeof(struct disk_info),
 			      GFP_KERNEL);
+
 	if (!conf->disks)
 		goto abort;
 
+	for (i = 0; i < max_disks; i++) {
+		conf->disks[i].extra_page = alloc_page(GFP_KERNEL);
+		if (!conf->disks[i].extra_page)
+			goto abort;
+	}
+
 	conf->mddev = mddev;
 
 	if ((conf->stripe_hashtbl = kzalloc(PAGE_SIZE, GFP_KERNEL)) == NULL)
@@ -6540,6 +6684,11 @@ static struct r5conf *setup_conf(struct mddev *mddev)
 	for (i = 0; i < NR_STRIPE_HASH_LOCKS; i++)
 		INIT_LIST_HEAD(conf->temp_inactive_list + i);
 
+	atomic_set(&conf->r5c_cached_full_stripes, 0);
+	INIT_LIST_HEAD(&conf->r5c_full_stripe_list);
+	atomic_set(&conf->r5c_cached_partial_stripes, 0);
+	INIT_LIST_HEAD(&conf->r5c_partial_stripe_list);
+
 	conf->level = mddev->new_level;
 	conf->chunk_sectors = mddev->new_chunk_sectors;
 	if (raid5_alloc_percpu(conf) != 0)
@@ -6566,9 +6715,8 @@ static struct r5conf *setup_conf(struct mddev *mddev)
 
 		if (test_bit(In_sync, &rdev->flags)) {
 			char b[BDEVNAME_SIZE];
-			printk(KERN_INFO "md/raid:%s: device %s operational as raid"
-			       " disk %d\n",
-			       mdname(mddev), bdevname(rdev->bdev, b), raid_disk);
+			pr_info("md/raid:%s: device %s operational as raid disk %d\n",
+				mdname(mddev), bdevname(rdev->bdev, b), raid_disk);
 		} else if (rdev->saved_raid_disk != raid_disk)
 			/* Cannot rely on bitmap to complete recovery */
 			conf->fullsync = 1;
@@ -6602,21 +6750,18 @@ static struct r5conf *setup_conf(struct mddev *mddev)
 			((mddev->new_chunk_sectors << 9) / STRIPE_SIZE) * 4);
 		conf->min_nr_stripes = max(NR_STRIPES, stripes);
 		if (conf->min_nr_stripes != NR_STRIPES)
-			printk(KERN_INFO
-				"md/raid:%s: force stripe size %d for reshape\n",
+			pr_info("md/raid:%s: force stripe size %d for reshape\n",
 				mdname(mddev), conf->min_nr_stripes);
 	}
 	memory = conf->min_nr_stripes * (sizeof(struct stripe_head) +
 		 max_disks * ((sizeof(struct bio) + PAGE_SIZE))) / 1024;
 	atomic_set(&conf->empty_inactive_list_nr, NR_STRIPE_HASH_LOCKS);
 	if (grow_stripes(conf, conf->min_nr_stripes)) {
-		printk(KERN_ERR
-		       "md/raid:%s: couldn't allocate %dkB for buffers\n",
-		       mdname(mddev), memory);
+		pr_warn("md/raid:%s: couldn't allocate %dkB for buffers\n",
+			mdname(mddev), memory);
 		goto abort;
 	} else
-		printk(KERN_INFO "md/raid:%s: allocated %dkB\n",
-		       mdname(mddev), memory);
+		pr_debug("md/raid:%s: allocated %dkB\n", mdname(mddev), memory);
 	/*
 	 * Losing a stripe head costs more than the time to refill it,
 	 * it reduces the queue depth and so can hurt throughput.
@@ -6628,18 +6773,16 @@ static struct r5conf *setup_conf(struct mddev *mddev)
 	conf->shrinker.batch = 128;
 	conf->shrinker.flags = 0;
 	if (register_shrinker(&conf->shrinker)) {
-		printk(KERN_ERR
-		       "md/raid:%s: couldn't register shrinker.\n",
-		       mdname(mddev));
+		pr_warn("md/raid:%s: couldn't register shrinker.\n",
+			mdname(mddev));
 		goto abort;
 	}
 
 	sprintf(pers_name, "raid%d", mddev->new_level);
 	conf->thread = md_register_thread(raid5d, mddev, pers_name);
 	if (!conf->thread) {
-		printk(KERN_ERR
-		       "md/raid:%s: couldn't allocate thread.\n",
-		       mdname(mddev));
+		pr_warn("md/raid:%s: couldn't allocate thread.\n",
+			mdname(mddev));
 		goto abort;
 	}
 
@@ -6692,9 +6835,8 @@ static int raid5_run(struct mddev *mddev)
 	int first = 1;
 
 	if (mddev->recovery_cp != MaxSector)
-		printk(KERN_NOTICE "md/raid:%s: not clean"
-		       " -- starting background reconstruction\n",
-		       mdname(mddev));
+		pr_notice("md/raid:%s: not clean -- starting background reconstruction\n",
+			  mdname(mddev));
 
 	rdev_for_each(rdev, mddev) {
 		long long diff;
@@ -6737,15 +6879,14 @@ static int raid5_run(struct mddev *mddev)
 		int new_data_disks;
 
 		if (journal_dev) {
-			printk(KERN_ERR "md/raid:%s: don't support reshape with journal - aborting.\n",
-			       mdname(mddev));
+			pr_warn("md/raid:%s: don't support reshape with journal - aborting.\n",
+				mdname(mddev));
 			return -EINVAL;
 		}
 
 		if (mddev->new_level != mddev->level) {
-			printk(KERN_ERR "md/raid:%s: unsupported reshape "
-			       "required - aborting.\n",
-			       mdname(mddev));
+			pr_warn("md/raid:%s: unsupported reshape required - aborting.\n",
+				mdname(mddev));
 			return -EINVAL;
 		}
 		old_disks = mddev->raid_disks - mddev->delta_disks;
@@ -6760,8 +6901,8 @@ static int raid5_run(struct mddev *mddev)
 		chunk_sectors = max(mddev->chunk_sectors, mddev->new_chunk_sectors);
 		new_data_disks = mddev->raid_disks - max_degraded;
 		if (sector_div(here_new, chunk_sectors * new_data_disks)) {
-			printk(KERN_ERR "md/raid:%s: reshape_position not "
-			       "on a stripe boundary\n", mdname(mddev));
+			pr_warn("md/raid:%s: reshape_position not on a stripe boundary\n",
+				mdname(mddev));
 			return -EINVAL;
 		}
 		reshape_offset = here_new * chunk_sectors;
@@ -6782,10 +6923,8 @@ static int raid5_run(struct mddev *mddev)
 			    abs(min_offset_diff) >= mddev->new_chunk_sectors)
 				/* not really in-place - so OK */;
 			else if (mddev->ro == 0) {
-				printk(KERN_ERR "md/raid:%s: in-place reshape "
-				       "must be started in read-only mode "
-				       "- aborting\n",
-				       mdname(mddev));
+				pr_warn("md/raid:%s: in-place reshape must be started in read-only mode - aborting\n",
+					mdname(mddev));
 				return -EINVAL;
 			}
 		} else if (mddev->reshape_backwards
@@ -6794,13 +6933,11 @@ static int raid5_run(struct mddev *mddev)
 		    : (here_new * chunk_sectors >=
 		       here_old * chunk_sectors + (-min_offset_diff))) {
 			/* Reading from the same stripe as writing to - bad */
-			printk(KERN_ERR "md/raid:%s: reshape_position too early for "
-			       "auto-recovery - aborting.\n",
-			       mdname(mddev));
+			pr_warn("md/raid:%s: reshape_position too early for auto-recovery - aborting.\n",
+				mdname(mddev));
 			return -EINVAL;
 		}
-		printk(KERN_INFO "md/raid:%s: reshape will continue\n",
-		       mdname(mddev));
+		pr_debug("md/raid:%s: reshape will continue\n", mdname(mddev));
 		/* OK, we should be able to continue; */
 	} else {
 		BUG_ON(mddev->level != mddev->new_level);
@@ -6819,8 +6956,8 @@ static int raid5_run(struct mddev *mddev)
 
 	if (test_bit(MD_HAS_JOURNAL, &mddev->flags)) {
 		if (!journal_dev) {
-			pr_err("md/raid:%s: journal disk is missing, force array readonly\n",
-			       mdname(mddev));
+			pr_warn("md/raid:%s: journal disk is missing, force array readonly\n",
+				mdname(mddev));
 			mddev->ro = 1;
 			set_disk_ro(mddev->gendisk, 1);
 		} else if (mddev->recovery_cp == MaxSector)
@@ -6847,8 +6984,7 @@ static int raid5_run(struct mddev *mddev)
 		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");
+			pr_warn("md: cannot handle concurrent replacement and reshape.\n");
 			goto abort;
 		}
 		if (test_bit(In_sync, &rdev->flags)) {
@@ -6890,8 +7026,7 @@ static int raid5_run(struct mddev *mddev)
 	mddev->degraded = calc_degraded(conf);
 
 	if (has_failed(conf)) {
-		printk(KERN_ERR "md/raid:%s: not enough operational devices"
-			" (%d/%d failed)\n",
+		pr_crit("md/raid:%s: not enough operational devices (%d/%d failed)\n",
 			mdname(mddev), mddev->degraded, conf->raid_disks);
 		goto abort;
 	}
@@ -6903,29 +7038,19 @@ static int raid5_run(struct mddev *mddev)
 	if (mddev->degraded > dirty_parity_disks &&
 	    mddev->recovery_cp != MaxSector) {
 		if (mddev->ok_start_degraded)
-			printk(KERN_WARNING
-			       "md/raid:%s: starting dirty degraded array"
-			       " - data corruption possible.\n",
-			       mdname(mddev));
+			pr_crit("md/raid:%s: starting dirty degraded array - data corruption possible.\n",
+				mdname(mddev));
 		else {
-			printk(KERN_ERR
-			       "md/raid:%s: cannot start dirty degraded array.\n",
-			       mdname(mddev));
+			pr_crit("md/raid:%s: cannot start dirty degraded array.\n",
+				mdname(mddev));
 			goto abort;
 		}
 	}
 
-	if (mddev->degraded == 0)
-		printk(KERN_INFO "md/raid:%s: raid level %d active with %d out of %d"
-		       " devices, algorithm %d\n", mdname(mddev), conf->level,
-		       mddev->raid_disks-mddev->degraded, mddev->raid_disks,
-		       mddev->new_layout);
-	else
-		printk(KERN_ALERT "md/raid:%s: raid level %d active with %d"
-		       " out of %d devices, algorithm %d\n",
-		       mdname(mddev), conf->level,
-		       mddev->raid_disks - mddev->degraded,
-		       mddev->raid_disks, mddev->new_layout);
+	pr_info("md/raid:%s: raid level %d active with %d out of %d devices, algorithm %d\n",
+		mdname(mddev), conf->level,
+		mddev->raid_disks-mddev->degraded, mddev->raid_disks,
+		mddev->new_layout);
 
 	print_raid5_conf(conf);
 
@@ -6945,9 +7070,8 @@ static int raid5_run(struct mddev *mddev)
 		mddev->to_remove = NULL;
 	else if (mddev->kobj.sd &&
 	    sysfs_create_group(&mddev->kobj, &raid5_attrs_group))
-		printk(KERN_WARNING
-		       "raid5: failed to create sysfs attributes for %s\n",
-		       mdname(mddev));
+		pr_warn("raid5: failed to create sysfs attributes for %s\n",
+			mdname(mddev));
 	md_set_array_sectors(mddev, raid5_size(mddev, 0, 0));
 
 	if (mddev->queue) {
@@ -6979,6 +7103,15 @@ static int raid5_run(struct mddev *mddev)
 			stripe = (stripe | (stripe-1)) + 1;
 		mddev->queue->limits.discard_alignment = stripe;
 		mddev->queue->limits.discard_granularity = stripe;
+
+		/*
+		 * We use 16-bit counter of active stripes in bi_phys_segments
+		 * (minus one for over-loaded initialization)
+		 */
+		blk_queue_max_hw_sectors(mddev->queue, 0xfffe * STRIPE_SECTORS);
+		blk_queue_max_discard_sectors(mddev->queue,
+					      0xfffe * STRIPE_SECTORS);
+
 		/*
 		 * unaligned part of discard request will be ignored, so can't
 		 * guarantee discard_zeroes_data
@@ -7035,9 +7168,10 @@ static int raid5_run(struct mddev *mddev)
 	if (journal_dev) {
 		char b[BDEVNAME_SIZE];
 
-		printk(KERN_INFO"md/raid:%s: using device %s as journal\n",
-		       mdname(mddev), bdevname(journal_dev->bdev, b));
-		r5l_init_log(conf, journal_dev);
+		pr_debug("md/raid:%s: using device %s as journal\n",
+			 mdname(mddev), bdevname(journal_dev->bdev, b));
+		if (r5l_init_log(conf, journal_dev))
+			goto abort;
 	}
 
 	return 0;
@@ -7046,7 +7180,7 @@ abort:
 	print_raid5_conf(conf);
 	free_conf(conf);
 	mddev->private = NULL;
-	printk(KERN_ALERT "md/raid:%s: failed to run raid set.\n", mdname(mddev));
+	pr_warn("md/raid:%s: failed to run raid set.\n", mdname(mddev));
 	return -EIO;
 }
 
@@ -7080,12 +7214,12 @@ static void print_raid5_conf (struct r5conf *conf)
 	int i;
 	struct disk_info *tmp;
 
-	printk(KERN_DEBUG "RAID conf printout:\n");
+	pr_debug("RAID conf printout:\n");
 	if (!conf) {
-		printk("(conf==NULL)\n");
+		pr_debug("(conf==NULL)\n");
 		return;
 	}
-	printk(KERN_DEBUG " --- level:%d rd:%d wd:%d\n", conf->level,
+	pr_debug(" --- level:%d rd:%d wd:%d\n", conf->level,
 	       conf->raid_disks,
 	       conf->raid_disks - conf->mddev->degraded);
 
@@ -7093,7 +7227,7 @@ static void print_raid5_conf (struct r5conf *conf)
 		char b[BDEVNAME_SIZE];
 		tmp = conf->disks + i;
 		if (tmp->rdev)
-			printk(KERN_DEBUG " disk %d, o:%d, dev:%s\n",
+			pr_debug(" disk %d, o:%d, dev:%s\n",
 			       i, !test_bit(Faulty, &tmp->rdev->flags),
 			       bdevname(tmp->rdev->bdev, b));
 	}
@@ -7241,8 +7375,8 @@ static int raid5_add_disk(struct mddev *mddev, struct md_rdev *rdev)
 		 * write requests running. We should be safe
 		 */
 		r5l_init_log(conf, rdev);
-		printk(KERN_INFO"md/raid:%s: using device %s as journal\n",
-		       mdname(mddev), bdevname(rdev->bdev, b));
+		pr_debug("md/raid:%s: using device %s as journal\n",
+			 mdname(mddev), bdevname(rdev->bdev, b));
 		return 0;
 	}
 	if (mddev->recovery_disabled == conf->recovery_disabled)
@@ -7346,10 +7480,10 @@ static int check_stripe_cache(struct mddev *mddev)
 	    > conf->min_nr_stripes ||
 	    ((mddev->new_chunk_sectors << 9) / STRIPE_SIZE) * 4
 	    > conf->min_nr_stripes) {
-		printk(KERN_WARNING "md/raid:%s: reshape: not enough stripes.  Needed %lu\n",
-		       mdname(mddev),
-		       ((max(mddev->chunk_sectors, mddev->new_chunk_sectors) << 9)
-			/ STRIPE_SIZE)*4);
+		pr_warn("md/raid:%s: reshape: not enough stripes.  Needed %lu\n",
+			mdname(mddev),
+			((max(mddev->chunk_sectors, mddev->new_chunk_sectors) << 9)
+			 / STRIPE_SIZE)*4);
 		return 0;
 	}
 	return 1;
@@ -7430,8 +7564,8 @@ static int raid5_start_reshape(struct mddev *mddev)
 	 */
 	if (raid5_size(mddev, 0, conf->raid_disks + mddev->delta_disks)
 	    < mddev->array_sectors) {
-		printk(KERN_ERR "md/raid:%s: array size must be reduced "
-		       "before number of disks\n", mdname(mddev));
+		pr_warn("md/raid:%s: array size must be reduced before number of disks\n",
+			mdname(mddev));
 		return -EINVAL;
 	}
 
@@ -7501,7 +7635,7 @@ static int raid5_start_reshape(struct mddev *mddev)
 	}
 	mddev->raid_disks = conf->raid_disks;
 	mddev->reshape_position = conf->reshape_progress;
-	set_bit(MD_CHANGE_DEVS, &mddev->flags);
+	set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
 
 	clear_bit(MD_RECOVERY_SYNC, &mddev->recovery);
 	clear_bit(MD_RECOVERY_CHECK, &mddev->recovery);
@@ -7619,6 +7753,7 @@ static void raid5_quiesce(struct mddev *mddev, int state)
 		/* '2' tells resync/reshape to pause so that all
 		 * active stripes can drain
 		 */
+		r5c_flush_cache(conf, INT_MAX);
 		conf->quiesce = 2;
 		wait_event_cmd(conf->wait_for_quiescent,
 				    atomic_read(&conf->active_stripes) == 0 &&
@@ -7649,8 +7784,8 @@ static void *raid45_takeover_raid0(struct mddev *mddev, int level)
 
 	/* for raid0 takeover only one zone is supported */
 	if (raid0_conf->nr_strip_zones > 1) {
-		printk(KERN_ERR "md/raid:%s: cannot takeover raid0 with more than one zone.\n",
-		       mdname(mddev));
+		pr_warn("md/raid:%s: cannot takeover raid0 with more than one zone.\n",
+			mdname(mddev));
 		return ERR_PTR(-EINVAL);
 	}
 
@@ -7671,6 +7806,7 @@ static void *raid45_takeover_raid0(struct mddev *mddev, int level)
 static void *raid5_takeover_raid1(struct mddev *mddev)
 {
 	int chunksect;
+	void *ret;
 
 	if (mddev->raid_disks != 2 ||
 	    mddev->degraded > 1)
@@ -7692,7 +7828,10 @@ static void *raid5_takeover_raid1(struct mddev *mddev)
 	mddev->new_layout = ALGORITHM_LEFT_SYMMETRIC;
 	mddev->new_chunk_sectors = chunksect;
 
-	return setup_conf(mddev);
+	ret = setup_conf(mddev);
+	if (!IS_ERR_VALUE(ret))
+		clear_bit(MD_FAILFAST_SUPPORTED, &mddev->flags);
+	return ret;
 }
 
 static void *raid5_takeover_raid6(struct mddev *mddev)
@@ -7762,7 +7901,7 @@ static int raid5_check_reshape(struct mddev *mddev)
 			conf->chunk_sectors = new_chunk ;
 			mddev->chunk_sectors = new_chunk;
 		}
-		set_bit(MD_CHANGE_DEVS, &mddev->flags);
+		set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags);
 		md_wakeup_thread(mddev->thread);
 	}
 	return check_reshape(mddev);

drivers/md/raid5.h

@@ -226,6 +226,8 @@ struct stripe_head {
 
 	struct r5l_io_unit	*log_io;
 	struct list_head	log_list;
+	sector_t		log_start; /* first meta block on the journal */
+	struct list_head	r5c; /* for r5c_cache->stripe_in_journal */
 	/**
 	 * struct stripe_operations
 	 * @target - STRIPE_OP_COMPUTE_BLK target
@@ -264,6 +266,7 @@ struct stripe_head_state {
 	int syncing, expanding, expanded, replacing;
 	int locked, uptodate, to_read, to_write, failed, written;
 	int to_fill, compute, req_compute, non_overwrite;
+	int injournal, just_cached;
 	int failed_num[2];
 	int p_failed, q_failed;
 	int dec_preread_active;
@@ -273,6 +276,7 @@ struct stripe_head_state {
 	struct md_rdev *blocked_rdev;
 	int handle_bad_blocks;
 	int log_failed;
+	int waiting_extra_page;
 };
 
 /* Flags for struct r5dev.flags */
@@ -313,6 +317,11 @@ enum r5dev_flags {
 			 */
 	R5_Discard,	/* Discard the stripe */
 	R5_SkipCopy,	/* Don't copy data from bio to stripe cache */
+	R5_InJournal,	/* data being written is in the journal device.
+			 * if R5_InJournal is set for parity pd_idx, all the
+			 * data and parity being written are in the journal
+			 * device
+			 */
 };
 
 /*
@@ -345,7 +354,30 @@ enum {
 	STRIPE_BITMAP_PENDING,	/* Being added to bitmap, don't add
 				 * to batch yet.
 				 */
-	STRIPE_LOG_TRAPPED, /* trapped into log */
+	STRIPE_LOG_TRAPPED,	/* trapped into log (see raid5-cache.c)
+				 * this bit is used in two scenarios:
+				 *
+				 * 1. write-out phase
+				 *  set in first entry of r5l_write_stripe
+				 *  clear in second entry of r5l_write_stripe
+				 *  used to bypass logic in handle_stripe
+				 *
+				 * 2. caching phase
+				 *  set in r5c_try_caching_write()
+				 *  clear when journal write is done
+				 *  used to initiate r5c_cache_data()
+				 *  also used to bypass logic in handle_stripe
+				 */
+	STRIPE_R5C_CACHING,	/* the stripe is in caching phase
+				 * see more detail in the raid5-cache.c
+				 */
+	STRIPE_R5C_PARTIAL_STRIPE,	/* in r5c cache (to-be/being handled or
+					 * in conf->r5c_partial_stripe_list)
+					 */
+	STRIPE_R5C_FULL_STRIPE,	/* in r5c cache (to-be/being handled or
+				 * in conf->r5c_full_stripe_list)
+				 */
+	STRIPE_R5C_PREFLUSH,	/* need to flush journal device */
 };
 
 #define STRIPE_EXPAND_SYNC_FLAGS \
@@ -408,8 +440,86 @@ enum {
 
 struct disk_info {
 	struct md_rdev	*rdev, *replacement;
+	struct page	*extra_page; /* extra page to use in prexor */
 };
 
+/*
+ * Stripe cache
+ */
+
+#define NR_STRIPES		256
+#define STRIPE_SIZE		PAGE_SIZE
+#define STRIPE_SHIFT		(PAGE_SHIFT - 9)
+#define STRIPE_SECTORS		(STRIPE_SIZE>>9)
+#define	IO_THRESHOLD		1
+#define BYPASS_THRESHOLD	1
+#define NR_HASH			(PAGE_SIZE / sizeof(struct hlist_head))
+#define HASH_MASK		(NR_HASH - 1)
+#define MAX_STRIPE_BATCH	8
+
+/* bio's attached to a stripe+device for I/O are linked together in bi_sector
+ * order without overlap.  There may be several bio's per stripe+device, and
+ * a bio could span several devices.
+ * When walking this list for a particular stripe+device, we must never proceed
+ * beyond a bio that extends past this device, as the next bio might no longer
+ * be valid.
+ * This function is used to determine the 'next' bio in the list, given the
+ * sector of the current stripe+device
+ */
+static inline struct bio *r5_next_bio(struct bio *bio, sector_t sector)
+{
+	int sectors = bio_sectors(bio);
+
+	if (bio->bi_iter.bi_sector + sectors < sector + STRIPE_SECTORS)
+		return bio->bi_next;
+	else
+		return NULL;
+}
+
+/*
+ * We maintain a biased count of active stripes in the bottom 16 bits of
+ * bi_phys_segments, and a count of processed stripes in the upper 16 bits
+ */
+static inline int raid5_bi_processed_stripes(struct bio *bio)
+{
+	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
+
+	return (atomic_read(segments) >> 16) & 0xffff;
+}
+
+static inline int raid5_dec_bi_active_stripes(struct bio *bio)
+{
+	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
+
+	return atomic_sub_return(1, segments) & 0xffff;
+}
+
+static inline void raid5_inc_bi_active_stripes(struct bio *bio)
+{
+	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
+
+	atomic_inc(segments);
+}
+
+static inline void raid5_set_bi_processed_stripes(struct bio *bio,
+	unsigned int cnt)
+{
+	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
+	int old, new;
+
+	do {
+		old = atomic_read(segments);
+		new = (old & 0xffff) | (cnt << 16);
+	} while (atomic_cmpxchg(segments, old, new) != old);
+}
+
+static inline void raid5_set_bi_stripes(struct bio *bio, unsigned int cnt)
+{
+	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
+
+	atomic_set(segments, cnt);
+}
+
 /* NOTE NR_STRIPE_HASH_LOCKS must remain below 64.
  * This is because we sometimes take all the spinlocks
  * and creating that much locking depth can cause
@@ -432,6 +542,30 @@ struct r5worker_group {
 	int stripes_cnt;
 };
 
+enum r5_cache_state {
+	R5_INACTIVE_BLOCKED,	/* release of inactive stripes blocked,
+				 * waiting for 25% to be free
+				 */
+	R5_ALLOC_MORE,		/* It might help to allocate another
+				 * stripe.
+				 */
+	R5_DID_ALLOC,		/* A stripe was allocated, don't allocate
+				 * more until at least one has been
+				 * released.  This avoids flooding
+				 * the cache.
+				 */
+	R5C_LOG_TIGHT,		/* log device space tight, need to
+				 * prioritize stripes at last_checkpoint
+				 */
+	R5C_LOG_CRITICAL,	/* log device is running out of space,
+				 * only process stripes that are already
+				 * occupying the log
+				 */
+	R5C_EXTRA_PAGE_IN_USE,	/* a stripe is using disk_info.extra_page
+				 * for prexor
+				 */
+};
+
 struct r5conf {
 	struct hlist_head	*stripe_hashtbl;
 	/* only protect corresponding hash list and inactive_list */
@@ -519,23 +653,18 @@ struct r5conf {
 	 */
 	atomic_t		active_stripes;
 	struct list_head	inactive_list[NR_STRIPE_HASH_LOCKS];
+
+	atomic_t		r5c_cached_full_stripes;
+	struct list_head	r5c_full_stripe_list;
+	atomic_t		r5c_cached_partial_stripes;
+	struct list_head	r5c_partial_stripe_list;
+
 	atomic_t		empty_inactive_list_nr;
 	struct llist_head	released_stripes;
 	wait_queue_head_t	wait_for_quiescent;
 	wait_queue_head_t	wait_for_stripe;
 	wait_queue_head_t	wait_for_overlap;
 	unsigned long		cache_state;
-#define R5_INACTIVE_BLOCKED	1	/* release of inactive stripes blocked,
-					 * waiting for 25% to be free
-					 */
-#define R5_ALLOC_MORE		2	/* It might help to allocate another
-					 * stripe.
-					 */
-#define R5_DID_ALLOC		4	/* A stripe was allocated, don't allocate
-					 * more until at least one has been
-					 * released.  This avoids flooding
-					 * the cache.
-					 */
 	struct shrinker		shrinker;
 	int			pool_size; /* number of disks in stripeheads in pool */
 	spinlock_t		device_lock;
@@ -633,4 +762,23 @@ extern void r5l_stripe_write_finished(struct stripe_head *sh);
 extern int r5l_handle_flush_request(struct r5l_log *log, struct bio *bio);
 extern void r5l_quiesce(struct r5l_log *log, int state);
 extern bool r5l_log_disk_error(struct r5conf *conf);
+extern bool r5c_is_writeback(struct r5l_log *log);
+extern int
+r5c_try_caching_write(struct r5conf *conf, struct stripe_head *sh,
+		      struct stripe_head_state *s, int disks);
+extern void
+r5c_finish_stripe_write_out(struct r5conf *conf, struct stripe_head *sh,
+			    struct stripe_head_state *s);
+extern void r5c_release_extra_page(struct stripe_head *sh);
+extern void r5c_use_extra_page(struct stripe_head *sh);
+extern void r5l_wake_reclaim(struct r5l_log *log, sector_t space);
+extern void r5c_handle_cached_data_endio(struct r5conf *conf,
+	struct stripe_head *sh, int disks, struct bio_list *return_bi);
+extern int r5c_cache_data(struct r5l_log *log, struct stripe_head *sh,
+			  struct stripe_head_state *s);
+extern void r5c_make_stripe_write_out(struct stripe_head *sh);
+extern void r5c_flush_cache(struct r5conf *conf, int num);
+extern void r5c_check_stripe_cache_usage(struct r5conf *conf);
+extern void r5c_check_cached_full_stripe(struct r5conf *conf);
+extern struct md_sysfs_entry r5c_journal_mode;
 #endif

include/uapi/linux/raid/md_p.h

@@ -84,6 +84,10 @@
 #define MD_DISK_CANDIDATE	5 /* disk is added as spare (local) until confirmed
 				   * For clustered enviroments only.
 				   */
+#define MD_DISK_FAILFAST	10 /* Send REQ_FAILFAST if there are multiple
+				    * devices available - and don't try to
+				    * correct read errors.
+				    */
 
 #define	MD_DISK_WRITEMOSTLY	9 /* disk is "write-mostly" is RAID1 config.
 				   * read requests will only be sent here in
@@ -265,8 +269,9 @@ struct mdp_superblock_1 {
 	__le32	dev_number;	/* permanent identifier of this  device - not role in raid */
 	__le32	cnt_corrected_read; /* number of read errors that were corrected by re-writing */
 	__u8	device_uuid[16]; /* user-space setable, ignored by kernel */
-	__u8	devflags;	/* per-device flags.  Only one defined...*/
+	__u8	devflags;	/* per-device flags.  Only two defined...*/
 #define	WriteMostly1	1	/* mask for writemostly flag in above */
+#define	FailFast1	2	/* Should avoid retries and fixups and just fail */
 	/* Bad block log.  If there are any bad blocks the feature flag is set.
 	 * If offset and size are non-zero, that space is reserved and available
 	 */

lib/raid6/avx2.c

@@ -87,9 +87,57 @@ static void raid6_avx21_gen_syndrome(int disks, size_t bytes, void **ptrs)
 	kernel_fpu_end();
 }
 
+static void raid6_avx21_xor_syndrome(int disks, int start, int stop,
+				     size_t bytes, void **ptrs)
+{
+	u8 **dptr = (u8 **)ptrs;
+	u8 *p, *q;
+	int d, z, z0;
+
+	z0 = stop;		/* P/Q right side optimization */
+	p = dptr[disks-2];	/* XOR parity */
+	q = dptr[disks-1];	/* RS syndrome */
+
+	kernel_fpu_begin();
+
+	asm volatile("vmovdqa %0,%%ymm0" : : "m" (raid6_avx2_constants.x1d[0]));
+
+	for (d = 0 ; d < bytes ; d += 32) {
+		asm volatile("vmovdqa %0,%%ymm4" :: "m" (dptr[z0][d]));
+		asm volatile("vmovdqa %0,%%ymm2" : : "m" (p[d]));
+		asm volatile("vpxor %ymm4,%ymm2,%ymm2");
+		/* P/Q data pages */
+		for (z = z0-1 ; z >= start ; z--) {
+			asm volatile("vpxor %ymm5,%ymm5,%ymm5");
+			asm volatile("vpcmpgtb %ymm4,%ymm5,%ymm5");
+			asm volatile("vpaddb %ymm4,%ymm4,%ymm4");
+			asm volatile("vpand %ymm0,%ymm5,%ymm5");
+			asm volatile("vpxor %ymm5,%ymm4,%ymm4");
+			asm volatile("vmovdqa %0,%%ymm5" :: "m" (dptr[z][d]));
+			asm volatile("vpxor %ymm5,%ymm2,%ymm2");
+			asm volatile("vpxor %ymm5,%ymm4,%ymm4");
+		}
+		/* P/Q left side optimization */
+		for (z = start-1 ; z >= 0 ; z--) {
+			asm volatile("vpxor %ymm5,%ymm5,%ymm5");
+			asm volatile("vpcmpgtb %ymm4,%ymm5,%ymm5");
+			asm volatile("vpaddb %ymm4,%ymm4,%ymm4");
+			asm volatile("vpand %ymm0,%ymm5,%ymm5");
+			asm volatile("vpxor %ymm5,%ymm4,%ymm4");
+		}
+		asm volatile("vpxor %0,%%ymm4,%%ymm4" : : "m" (q[d]));
+		/* Don't use movntdq for r/w memory area < cache line */
+		asm volatile("vmovdqa %%ymm4,%0" : "=m" (q[d]));
+		asm volatile("vmovdqa %%ymm2,%0" : "=m" (p[d]));
+	}
+
+	asm volatile("sfence" : : : "memory");
+	kernel_fpu_end();
+}
+
 const struct raid6_calls raid6_avx2x1 = {
 	raid6_avx21_gen_syndrome,
-	NULL,			/* XOR not yet implemented */
+	raid6_avx21_xor_syndrome,
 	raid6_have_avx2,
 	"avx2x1",
 	1			/* Has cache hints */
@@ -149,9 +197,77 @@ static void raid6_avx22_gen_syndrome(int disks, size_t bytes, void **ptrs)
 	kernel_fpu_end();
 }
 
+static void raid6_avx22_xor_syndrome(int disks, int start, int stop,
+				     size_t bytes, void **ptrs)
+{
+	u8 **dptr = (u8 **)ptrs;
+	u8 *p, *q;
+	int d, z, z0;
+
+	z0 = stop;		/* P/Q right side optimization */
+	p = dptr[disks-2];	/* XOR parity */
+	q = dptr[disks-1];	/* RS syndrome */
+
+	kernel_fpu_begin();
+
+	asm volatile("vmovdqa %0,%%ymm0" : : "m" (raid6_avx2_constants.x1d[0]));
+
+	for (d = 0 ; d < bytes ; d += 64) {
+		asm volatile("vmovdqa %0,%%ymm4" :: "m" (dptr[z0][d]));
+		asm volatile("vmovdqa %0,%%ymm6" :: "m" (dptr[z0][d+32]));
+		asm volatile("vmovdqa %0,%%ymm2" : : "m" (p[d]));
+		asm volatile("vmovdqa %0,%%ymm3" : : "m" (p[d+32]));
+		asm volatile("vpxor %ymm4,%ymm2,%ymm2");
+		asm volatile("vpxor %ymm6,%ymm3,%ymm3");
+		/* P/Q data pages */
+		for (z = z0-1 ; z >= start ; z--) {
+			asm volatile("vpxor %ymm5,%ymm5,%ymm5");
+			asm volatile("vpxor %ymm7,%ymm7,%ymm7");
+			asm volatile("vpcmpgtb %ymm4,%ymm5,%ymm5");
+			asm volatile("vpcmpgtb %ymm6,%ymm7,%ymm7");
+			asm volatile("vpaddb %ymm4,%ymm4,%ymm4");
+			asm volatile("vpaddb %ymm6,%ymm6,%ymm6");
+			asm volatile("vpand %ymm0,%ymm5,%ymm5");
+			asm volatile("vpand %ymm0,%ymm7,%ymm7");
+			asm volatile("vpxor %ymm5,%ymm4,%ymm4");
+			asm volatile("vpxor %ymm7,%ymm6,%ymm6");
+			asm volatile("vmovdqa %0,%%ymm5" :: "m" (dptr[z][d]));
+			asm volatile("vmovdqa %0,%%ymm7"
+				     :: "m" (dptr[z][d+32]));
+			asm volatile("vpxor %ymm5,%ymm2,%ymm2");
+			asm volatile("vpxor %ymm7,%ymm3,%ymm3");
+			asm volatile("vpxor %ymm5,%ymm4,%ymm4");
+			asm volatile("vpxor %ymm7,%ymm6,%ymm6");
+		}
+		/* P/Q left side optimization */
+		for (z = start-1 ; z >= 0 ; z--) {
+			asm volatile("vpxor %ymm5,%ymm5,%ymm5");
+			asm volatile("vpxor %ymm7,%ymm7,%ymm7");
+			asm volatile("vpcmpgtb %ymm4,%ymm5,%ymm5");
+			asm volatile("vpcmpgtb %ymm6,%ymm7,%ymm7");
+			asm volatile("vpaddb %ymm4,%ymm4,%ymm4");
+			asm volatile("vpaddb %ymm6,%ymm6,%ymm6");
+			asm volatile("vpand %ymm0,%ymm5,%ymm5");
+			asm volatile("vpand %ymm0,%ymm7,%ymm7");
+			asm volatile("vpxor %ymm5,%ymm4,%ymm4");
+			asm volatile("vpxor %ymm7,%ymm6,%ymm6");
+		}
+		asm volatile("vpxor %0,%%ymm4,%%ymm4" : : "m" (q[d]));
+		asm volatile("vpxor %0,%%ymm6,%%ymm6" : : "m" (q[d+32]));
+		/* Don't use movntdq for r/w memory area < cache line */
+		asm volatile("vmovdqa %%ymm4,%0" : "=m" (q[d]));
+		asm volatile("vmovdqa %%ymm6,%0" : "=m" (q[d+32]));
+		asm volatile("vmovdqa %%ymm2,%0" : "=m" (p[d]));
+		asm volatile("vmovdqa %%ymm3,%0" : "=m" (p[d+32]));
+	}
+
+	asm volatile("sfence" : : : "memory");
+	kernel_fpu_end();
+}
+
 const struct raid6_calls raid6_avx2x2 = {
 	raid6_avx22_gen_syndrome,
-	NULL,			/* XOR not yet implemented */
+	raid6_avx22_xor_syndrome,
 	raid6_have_avx2,
 	"avx2x2",
 	1			/* Has cache hints */
@@ -242,9 +358,119 @@ static void raid6_avx24_gen_syndrome(int disks, size_t bytes, void **ptrs)
 	kernel_fpu_end();
 }
 
+static void raid6_avx24_xor_syndrome(int disks, int start, int stop,
+				     size_t bytes, void **ptrs)
+{
+	u8 **dptr = (u8 **)ptrs;
+	u8 *p, *q;
+	int d, z, z0;
+
+	z0 = stop;		/* P/Q right side optimization */
+	p = dptr[disks-2];	/* XOR parity */
+	q = dptr[disks-1];	/* RS syndrome */
+
+	kernel_fpu_begin();
+
+	asm volatile("vmovdqa %0,%%ymm0" :: "m" (raid6_avx2_constants.x1d[0]));
+
+	for (d = 0 ; d < bytes ; d += 128) {
+		asm volatile("vmovdqa %0,%%ymm4" :: "m" (dptr[z0][d]));
+		asm volatile("vmovdqa %0,%%ymm6" :: "m" (dptr[z0][d+32]));
+		asm volatile("vmovdqa %0,%%ymm12" :: "m" (dptr[z0][d+64]));
+		asm volatile("vmovdqa %0,%%ymm14" :: "m" (dptr[z0][d+96]));
+		asm volatile("vmovdqa %0,%%ymm2" : : "m" (p[d]));
+		asm volatile("vmovdqa %0,%%ymm3" : : "m" (p[d+32]));
+		asm volatile("vmovdqa %0,%%ymm10" : : "m" (p[d+64]));
+		asm volatile("vmovdqa %0,%%ymm11" : : "m" (p[d+96]));
+		asm volatile("vpxor %ymm4,%ymm2,%ymm2");
+		asm volatile("vpxor %ymm6,%ymm3,%ymm3");
+		asm volatile("vpxor %ymm12,%ymm10,%ymm10");
+		asm volatile("vpxor %ymm14,%ymm11,%ymm11");
+		/* P/Q data pages */
+		for (z = z0-1 ; z >= start ; z--) {
+			asm volatile("prefetchnta %0" :: "m" (dptr[z][d]));
+			asm volatile("prefetchnta %0" :: "m" (dptr[z][d+64]));
+			asm volatile("vpxor %ymm5,%ymm5,%ymm5");
+			asm volatile("vpxor %ymm7,%ymm7,%ymm7");
+			asm volatile("vpxor %ymm13,%ymm13,%ymm13");
+			asm volatile("vpxor %ymm15,%ymm15,%ymm15");
+			asm volatile("vpcmpgtb %ymm4,%ymm5,%ymm5");
+			asm volatile("vpcmpgtb %ymm6,%ymm7,%ymm7");
+			asm volatile("vpcmpgtb %ymm12,%ymm13,%ymm13");
+			asm volatile("vpcmpgtb %ymm14,%ymm15,%ymm15");
+			asm volatile("vpaddb %ymm4,%ymm4,%ymm4");
+			asm volatile("vpaddb %ymm6,%ymm6,%ymm6");
+			asm volatile("vpaddb %ymm12,%ymm12,%ymm12");
+			asm volatile("vpaddb %ymm14,%ymm14,%ymm14");
+			asm volatile("vpand %ymm0,%ymm5,%ymm5");
+			asm volatile("vpand %ymm0,%ymm7,%ymm7");
+			asm volatile("vpand %ymm0,%ymm13,%ymm13");
+			asm volatile("vpand %ymm0,%ymm15,%ymm15");
+			asm volatile("vpxor %ymm5,%ymm4,%ymm4");
+			asm volatile("vpxor %ymm7,%ymm6,%ymm6");
+			asm volatile("vpxor %ymm13,%ymm12,%ymm12");
+			asm volatile("vpxor %ymm15,%ymm14,%ymm14");
+			asm volatile("vmovdqa %0,%%ymm5" :: "m" (dptr[z][d]));
+			asm volatile("vmovdqa %0,%%ymm7"
+				     :: "m" (dptr[z][d+32]));
+			asm volatile("vmovdqa %0,%%ymm13"
+				     :: "m" (dptr[z][d+64]));
+			asm volatile("vmovdqa %0,%%ymm15"
+				     :: "m" (dptr[z][d+96]));
+			asm volatile("vpxor %ymm5,%ymm2,%ymm2");
+			asm volatile("vpxor %ymm7,%ymm3,%ymm3");
+			asm volatile("vpxor %ymm13,%ymm10,%ymm10");
+			asm volatile("vpxor %ymm15,%ymm11,%ymm11");
+			asm volatile("vpxor %ymm5,%ymm4,%ymm4");
+			asm volatile("vpxor %ymm7,%ymm6,%ymm6");
+			asm volatile("vpxor %ymm13,%ymm12,%ymm12");
+			asm volatile("vpxor %ymm15,%ymm14,%ymm14");
+		}
+		asm volatile("prefetchnta %0" :: "m" (q[d]));
+		asm volatile("prefetchnta %0" :: "m" (q[d+64]));
+		/* P/Q left side optimization */
+		for (z = start-1 ; z >= 0 ; z--) {
+			asm volatile("vpxor %ymm5,%ymm5,%ymm5");
+			asm volatile("vpxor %ymm7,%ymm7,%ymm7");
+			asm volatile("vpxor %ymm13,%ymm13,%ymm13");
+			asm volatile("vpxor %ymm15,%ymm15,%ymm15");
+			asm volatile("vpcmpgtb %ymm4,%ymm5,%ymm5");
+			asm volatile("vpcmpgtb %ymm6,%ymm7,%ymm7");
+			asm volatile("vpcmpgtb %ymm12,%ymm13,%ymm13");
+			asm volatile("vpcmpgtb %ymm14,%ymm15,%ymm15");
+			asm volatile("vpaddb %ymm4,%ymm4,%ymm4");
+			asm volatile("vpaddb %ymm6,%ymm6,%ymm6");
+			asm volatile("vpaddb %ymm12,%ymm12,%ymm12");
+			asm volatile("vpaddb %ymm14,%ymm14,%ymm14");
+			asm volatile("vpand %ymm0,%ymm5,%ymm5");
+			asm volatile("vpand %ymm0,%ymm7,%ymm7");
+			asm volatile("vpand %ymm0,%ymm13,%ymm13");
+			asm volatile("vpand %ymm0,%ymm15,%ymm15");
+			asm volatile("vpxor %ymm5,%ymm4,%ymm4");
+			asm volatile("vpxor %ymm7,%ymm6,%ymm6");
+			asm volatile("vpxor %ymm13,%ymm12,%ymm12");
+			asm volatile("vpxor %ymm15,%ymm14,%ymm14");
+		}
+		asm volatile("vmovntdq %%ymm2,%0" : "=m" (p[d]));
+		asm volatile("vmovntdq %%ymm3,%0" : "=m" (p[d+32]));
+		asm volatile("vmovntdq %%ymm10,%0" : "=m" (p[d+64]));
+		asm volatile("vmovntdq %%ymm11,%0" : "=m" (p[d+96]));
+		asm volatile("vpxor %0,%%ymm4,%%ymm4" : : "m" (q[d]));
+		asm volatile("vpxor %0,%%ymm6,%%ymm6" : : "m" (q[d+32]));
+		asm volatile("vpxor %0,%%ymm12,%%ymm12" : : "m" (q[d+64]));
+		asm volatile("vpxor %0,%%ymm14,%%ymm14" : : "m" (q[d+96]));
+		asm volatile("vmovntdq %%ymm4,%0" : "=m" (q[d]));
+		asm volatile("vmovntdq %%ymm6,%0" : "=m" (q[d+32]));
+		asm volatile("vmovntdq %%ymm12,%0" : "=m" (q[d+64]));
+		asm volatile("vmovntdq %%ymm14,%0" : "=m" (q[d+96]));
+	}
+	asm volatile("sfence" : : : "memory");
+	kernel_fpu_end();
+}
+
 const struct raid6_calls raid6_avx2x4 = {
 	raid6_avx24_gen_syndrome,
-	NULL,			/* XOR not yet implemented */
+	raid6_avx24_xor_syndrome,
 	raid6_have_avx2,
 	"avx2x4",
 	1			/* Has cache hints */