1 files changed, 27 insertions, 10 deletions

diff --git a/drivers/md/dm-region-hash.c b/drivers/md/dm-region-hash.c
index 36dbe29f2fd6..bd5c58b28868 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 barrier failure no regions can be marked clean.
+         */
+        int barrier_failure;
+
         void *context;
         sector_t target_begin;
 
@@ -211,6 +217,7 @@ 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->barrier_failure = 0;
 
         rh->region_pool = mempool_create_kmalloc_pool(MIN_REGIONS,
                                                       sizeof(struct dm_region));
@@ -377,8 +384,6 @@ 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 +391,7 @@ 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 +399,11 @@ 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_empty_barrier(bio)) {
+                rh->barrier_failure = 1;
+                return;
+        }
+
         /* We must inform the log that the sync count has changed. */
         log->type->set_region_sync(log, region, 0);
 
@@ -419,7 +428,6 @@ 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 +523,11 @@ 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_empty_barrier(bio))
+                        continue;
                 rh_inc(rh, dm_rh_bio_to_region(rh, bio));
+        }
 }
 EXPORT_SYMBOL_GPL(dm_rh_inc_pending);
 
@@ -544,7 +555,14 @@ 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->barrier_failure)) {
+                        /*
+                         * If a write barrier 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 +661,9 @@ 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);