1 files changed, 34 insertions, 13 deletions

diff --git a/drivers/md/dm-region-hash.c b/drivers/md/dm-region-hash.c
index 7b899be0b08..b929fd5f498 100644
--- a/drivers/md/dm-region-hash.c
+++ b/drivers/md/dm-region-hash.c
@@ -11,6 +11,7 @@
 #include <linux/ctype.h>
 #include <linux/init.h>
 #include <linux/module.h>
+#include <linux/slab.h>
 #include <linux/vmalloc.h>
 
 #include "dm.h"
@@ -79,6 +80,11 @@ struct dm_region_hash {
 	struct list_head recovered_regions;
 	struct list_head failed_recovered_regions;
 
+	/*
+	 * If there was a flush failure no regions can be marked clean.
+	 */
+	int flush_failure;
+
 	void *context;
 	sector_t target_begin;
 
@@ -120,7 +126,8 @@ EXPORT_SYMBOL_GPL(dm_rh_region_to_sector);
 
 region_t dm_rh_bio_to_region(struct dm_region_hash *rh, struct bio *bio)
 {
-	return dm_rh_sector_to_region(rh, bio->bi_sector - rh->target_begin);
+	return dm_rh_sector_to_region(rh, bio->bi_iter.bi_sector -
+				      rh->target_begin);
 }
 EXPORT_SYMBOL_GPL(dm_rh_bio_to_region);
 
@@ -211,6 +218,7 @@ struct dm_region_hash *dm_region_hash_create(
 	INIT_LIST_HEAD(&rh->quiesced_regions);
 	INIT_LIST_HEAD(&rh->recovered_regions);
 	INIT_LIST_HEAD(&rh->failed_recovered_regions);
+	rh->flush_failure = 0;
 
 	rh->region_pool = mempool_create_kmalloc_pool(MIN_REGIONS,
 						      sizeof(struct dm_region));
@@ -283,7 +291,7 @@ static struct dm_region *__rh_alloc(struct dm_region_hash *rh, region_t region)
 
 	nreg = mempool_alloc(rh->region_pool, GFP_ATOMIC);
 	if (unlikely(!nreg))
-		nreg = kmalloc(sizeof(*nreg), GFP_NOIO);
+		nreg = kmalloc(sizeof(*nreg), GFP_NOIO | __GFP_NOFAIL);
 
 	nreg->state = rh->log->type->in_sync(rh->log, region, 1) ?
 		      DM_RH_CLEAN : DM_RH_NOSYNC;
@@ -377,8 +385,6 @@ static void complete_resync_work(struct dm_region *reg, int success)
 /* dm_rh_mark_nosync
  * @ms
  * @bio
- * @done
- * @error
  *
  * The bio was written on some mirror(s) but failed on other mirror(s).
  * We can successfully endio the bio but should avoid the region being
@@ -386,8 +392,7 @@ static void complete_resync_work(struct dm_region *reg, int success)
  *
  * This function is _not_ safe in interrupt context!
  */
-void dm_rh_mark_nosync(struct dm_region_hash *rh,
-		       struct bio *bio, unsigned done, int error)
+void dm_rh_mark_nosync(struct dm_region_hash *rh, struct bio *bio)
 {
 	unsigned long flags;
 	struct dm_dirty_log *log = rh->log;
@@ -395,6 +400,14 @@ void dm_rh_mark_nosync(struct dm_region_hash *rh,
 	region_t region = dm_rh_bio_to_region(rh, bio);
 	int recovering = 0;
 
+	if (bio->bi_rw & REQ_FLUSH) {
+		rh->flush_failure = 1;
+		return;
+	}
+
+	if (bio->bi_rw & REQ_DISCARD)
+		return;
+
 	/* We must inform the log that the sync count has changed. */
 	log->type->set_region_sync(log, region, 0);
 
@@ -410,7 +423,7 @@ void dm_rh_mark_nosync(struct dm_region_hash *rh,
 	/*
 	 * Possible cases:
 	 *   1) DM_RH_DIRTY
-	 *   2) DM_RH_NOSYNC: was dirty, other preceeding writes failed
+	 *   2) DM_RH_NOSYNC: was dirty, other preceding writes failed
 	 *   3) DM_RH_RECOVERING: flushing pending writes
 	 * Either case, the region should have not been connected to list.
 	 */
@@ -419,7 +432,6 @@ void dm_rh_mark_nosync(struct dm_region_hash *rh,
 	BUG_ON(!list_empty(&reg->list));
 	spin_unlock_irqrestore(&rh->region_lock, flags);
 
-	bio_endio(bio, error);
 	if (recovering)
 		complete_resync_work(reg, 0);
 }
@@ -515,8 +527,11 @@ void dm_rh_inc_pending(struct dm_region_hash *rh, struct bio_list *bios)
 {
 	struct bio *bio;
 
-	for (bio = bios->head; bio; bio = bio->bi_next)
+	for (bio = bios->head; bio; bio = bio->bi_next) {
+		if (bio->bi_rw & (REQ_FLUSH | REQ_DISCARD))
+			continue;
 		rh_inc(rh, dm_rh_bio_to_region(rh, bio));
+	}
 }
 EXPORT_SYMBOL_GPL(dm_rh_inc_pending);
 
@@ -544,7 +559,14 @@ void dm_rh_dec(struct dm_region_hash *rh, region_t region)
 		 */
 
 		/* do nothing for DM_RH_NOSYNC */
-		if (reg->state == DM_RH_RECOVERING) {
+		if (unlikely(rh->flush_failure)) {
+			/*
+			 * If a write flush failed some time ago, we
+			 * don't know whether or not this write made it
+			 * to the disk, so we must resync the device.
+			 */
+			reg->state = DM_RH_NOSYNC;
+		} else if (reg->state == DM_RH_RECOVERING) {
 			list_add_tail(&reg->list, &rh->quiesced_regions);
 		} else if (reg->state == DM_RH_DIRTY) {
 			reg->state = DM_RH_CLEAN;
@@ -643,10 +665,9 @@ void dm_rh_recovery_end(struct dm_region *reg, int success)
 	spin_lock_irq(&rh->region_lock);
 	if (success)
 		list_add(&reg->list, &reg->rh->recovered_regions);
-	else {
-		reg->state = DM_RH_NOSYNC;
+	else
 		list_add(&reg->list, &reg->rh->failed_recovered_regions);
-	}
+
 	spin_unlock_irq(&rh->region_lock);
 
 	rh->wakeup_workers(rh->context);