dm thin: fix bug where bio that overwrites thin block ignores FUA

When provisioning a new data block for a virtual block, either because
the block was previously unallocated or because we are breaking sharing,
if the whole block of data is being overwritten the bio that triggered
the provisioning is issued immediately, skipping copying or zeroing of
the data block.

When this bio completes the new mapping is inserted in to the pool's
metadata by process_prepared_mapping(), where the bio completion is
signaled to the upper layers.

This completion is signaled without first committing the metadata.  If
the bio in question has the REQ_FUA flag set and the system crashes
right after its completion and before the next metadata commit, then the
write is lost despite the REQ_FUA flag requiring that I/O completion for
this request must only be signaled after the data has been committed to
non-volatile storage.

Fix this by deferring the completion of overwrite bios, with the REQ_FUA
flag set, until after the metadata has been committed.

Cc: stable@vger.kernel.org
Signed-off-by: Nikos Tsironis <ntsironis@arrikto.com>
Acked-by: Joe Thornber <ejt@redhat.com>
Acked-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>

drivers/md/dm-thin.c

@@ -257,6 +257,7 @@ struct pool {
 
 	spinlock_t lock;
 	struct bio_list deferred_flush_bios;
+	struct bio_list deferred_flush_completions;
 	struct list_head prepared_mappings;
 	struct list_head prepared_discards;
 	struct list_head prepared_discards_pt2;
@@ -956,6 +957,39 @@ static void process_prepared_mapping_fail(struct dm_thin_new_mapping *m)
 	mempool_free(m, &m->tc->pool->mapping_pool);
 }
 
+static void complete_overwrite_bio(struct thin_c *tc, struct bio *bio)
+{
+	struct pool *pool = tc->pool;
+	unsigned long flags;
+
+	/*
+	 * If the bio has the REQ_FUA flag set we must commit the metadata
+	 * before signaling its completion.
+	 */
+	if (!bio_triggers_commit(tc, bio)) {
+		bio_endio(bio);
+		return;
+	}
+
+	/*
+	 * Complete bio with an error if earlier I/O caused changes to the
+	 * metadata that can't be committed, e.g, due to I/O errors on the
+	 * metadata device.
+	 */
+	if (dm_thin_aborted_changes(tc->td)) {
+		bio_io_error(bio);
+		return;
+	}
+
+	/*
+	 * Batch together any bios that trigger commits and then issue a
+	 * single commit for them in process_deferred_bios().
+	 */
+	spin_lock_irqsave(&pool->lock, flags);
+	bio_list_add(&pool->deferred_flush_completions, bio);
+	spin_unlock_irqrestore(&pool->lock, flags);
+}
+
 static void process_prepared_mapping(struct dm_thin_new_mapping *m)
 {
 	struct thin_c *tc = m->tc;
@@ -988,7 +1022,7 @@ static void process_prepared_mapping(struct dm_thin_new_mapping *m)
 	 */
 	if (bio) {
 		inc_remap_and_issue_cell(tc, m->cell, m->data_block);
-		bio_endio(bio);
+		complete_overwrite_bio(tc, bio);
 	} else {
 		inc_all_io_entry(tc->pool, m->cell->holder);
 		remap_and_issue(tc, m->cell->holder, m->data_block);
@@ -2317,7 +2351,7 @@ static void process_deferred_bios(struct pool *pool)
 {
 	unsigned long flags;
 	struct bio *bio;
-	struct bio_list bios;
+	struct bio_list bios, bio_completions;
 	struct thin_c *tc;
 
 	tc = get_first_thin(pool);
@@ -2328,26 +2362,36 @@ static void process_deferred_bios(struct pool *pool)
 	}
 
 	/*
-	 * If there are any deferred flush bios, we must commit
-	 * the metadata before issuing them.
+	 * If there are any deferred flush bios, we must commit the metadata
+	 * before issuing them or signaling their completion.
 	 */
 	bio_list_init(&bios);
+	bio_list_init(&bio_completions);
+
 	spin_lock_irqsave(&pool->lock, flags);
 	bio_list_merge(&bios, &pool->deferred_flush_bios);
 	bio_list_init(&pool->deferred_flush_bios);
+
+	bio_list_merge(&bio_completions, &pool->deferred_flush_completions);
+	bio_list_init(&pool->deferred_flush_completions);
 	spin_unlock_irqrestore(&pool->lock, flags);
 
-	if (bio_list_empty(&bios) &&
+	if (bio_list_empty(&bios) && bio_list_empty(&bio_completions) &&
 	    !(dm_pool_changed_this_transaction(pool->pmd) && need_commit_due_to_time(pool)))
 		return;
 
 	if (commit(pool)) {
+		bio_list_merge(&bios, &bio_completions);
+
 		while ((bio = bio_list_pop(&bios)))
 			bio_io_error(bio);
 		return;
 	}
 	pool->last_commit_jiffies = jiffies;
 
+	while ((bio = bio_list_pop(&bio_completions)))
+		bio_endio(bio);
+
 	while ((bio = bio_list_pop(&bios)))
 		generic_make_request(bio);
 }
@@ -2954,6 +2998,7 @@ static struct pool *pool_create(struct mapped_device *pool_md,
 	INIT_DELAYED_WORK(&pool->no_space_timeout, do_no_space_timeout);
 	spin_lock_init(&pool->lock);
 	bio_list_init(&pool->deferred_flush_bios);
+	bio_list_init(&pool->deferred_flush_completions);
 	INIT_LIST_HEAD(&pool->prepared_mappings);
 	INIT_LIST_HEAD(&pool->prepared_discards);
 	INIT_LIST_HEAD(&pool->prepared_discards_pt2);