Merge git://git.kernel.org/pub/scm/linux/kernel/git/agk/linux-2.6-dm

* git://git.kernel.org/pub/scm/linux/kernel/git/agk/linux-2.6-dm: (80 commits)
  dm snapshot: use merge origin if snapshot invalid
  dm snapshot: report merge failure in status
  dm snapshot: merge consecutive chunks together
  dm snapshot: trigger exceptions in remaining snapshots during merge
  dm snapshot: delay merging a chunk until writes to it complete
  dm snapshot: queue writes to chunks being merged
  dm snapshot: add merging
  dm snapshot: permit only one merge at once
  dm snapshot: support barriers in snapshot merge target
  dm snapshot: avoid allocating exceptions in merge
  dm snapshot: rework writing to origin
  dm snapshot: add merge target
  dm exception store: add merge specific methods
  dm snapshot: create function for chunk_is_tracked wait
  dm snapshot: make bio optional in __origin_write
  dm mpath: reject messages when device is suspended
  dm: export suspended state to targets
  dm: rename dm_suspended to dm_suspended_md
  dm: swap target postsuspend call and setting suspended flag
  dm crypt: add plain64 iv
  ...

1 files changed, 156 insertions, 63 deletions

diff --git a/drivers/md/dm-raid1.c b/drivers/md/dm-raid1.c
index cc9dc79..ad779bd 100644
--- a/drivers/md/dm-raid1.c
+++ b/drivers/md/dm-raid1.c
@@ -35,6 +35,7 @@ static DECLARE_WAIT_QUEUE_HEAD(_kmirrord_recovery_stopped);
  *---------------------------------------------------------------*/
 enum dm_raid1_error {
 	DM_RAID1_WRITE_ERROR,
+	DM_RAID1_FLUSH_ERROR,
 	DM_RAID1_SYNC_ERROR,
 	DM_RAID1_READ_ERROR
 };
@@ -57,6 +58,7 @@ struct mirror_set {
 	struct bio_list reads;
 	struct bio_list writes;
 	struct bio_list failures;
+	struct bio_list holds;	/* bios are waiting until suspend */
 
 	struct dm_region_hash *rh;
 	struct dm_kcopyd_client *kcopyd_client;
@@ -67,6 +69,7 @@ struct mirror_set {
 	region_t nr_regions;
 	int in_sync;
 	int log_failure;
+	int leg_failure;
 	atomic_t suspend;
 
 	atomic_t default_mirror;	/* Default mirror */
@@ -179,6 +182,17 @@ static void set_default_mirror(struct mirror *m)
 	atomic_set(&ms->default_mirror, m - m0);
 }
 
+static struct mirror *get_valid_mirror(struct mirror_set *ms)
+{
+	struct mirror *m;
+
+	for (m = ms->mirror; m < ms->mirror + ms->nr_mirrors; m++)
+		if (!atomic_read(&m->error_count))
+			return m;
+
+	return NULL;
+}
+
 /* fail_mirror
  * @m: mirror device to fail
  * @error_type: one of the enum's, DM_RAID1_*_ERROR
@@ -198,6 +212,8 @@ static void fail_mirror(struct mirror *m, enum dm_raid1_error error_type)
 	struct mirror_set *ms = m->ms;
 	struct mirror *new;
 
+	ms->leg_failure = 1;
+
 	/*
 	 * error_count is used for nothing more than a
 	 * simple way to tell if a device has encountered
@@ -224,19 +240,50 @@ static void fail_mirror(struct mirror *m, enum dm_raid1_error error_type)
 		goto out;
 	}
 
-	for (new = ms->mirror; new < ms->mirror + ms->nr_mirrors; new++)
-		if (!atomic_read(&new->error_count)) {
-			set_default_mirror(new);
-			break;
-		}
-
-	if (unlikely(new == ms->mirror + ms->nr_mirrors))
+	new = get_valid_mirror(ms);
+	if (new)
+		set_default_mirror(new);
+	else
 		DMWARN("All sides of mirror have failed.");
 
 out:
 	schedule_work(&ms->trigger_event);
 }
 
+static int mirror_flush(struct dm_target *ti)
+{
+	struct mirror_set *ms = ti->private;
+	unsigned long error_bits;
+
+	unsigned int i;
+	struct dm_io_region io[ms->nr_mirrors];
+	struct mirror *m;
+	struct dm_io_request io_req = {
+		.bi_rw = WRITE_BARRIER,
+		.mem.type = DM_IO_KMEM,
+		.mem.ptr.bvec = NULL,
+		.client = ms->io_client,
+	};
+
+	for (i = 0, m = ms->mirror; i < ms->nr_mirrors; i++, m++) {
+		io[i].bdev = m->dev->bdev;
+		io[i].sector = 0;
+		io[i].count = 0;
+	}
+
+	error_bits = -1;
+	dm_io(&io_req, ms->nr_mirrors, io, &error_bits);
+	if (unlikely(error_bits != 0)) {
+		for (i = 0; i < ms->nr_mirrors; i++)
+			if (test_bit(i, &error_bits))
+				fail_mirror(ms->mirror + i,
+					    DM_RAID1_FLUSH_ERROR);
+		return -EIO;
+	}
+
+	return 0;
+}
+
 /*-----------------------------------------------------------------
  * Recovery.
  *
@@ -396,6 +443,8 @@ static int mirror_available(struct mirror_set *ms, struct bio *bio)
  */
 static sector_t map_sector(struct mirror *m, struct bio *bio)
 {
+	if (unlikely(!bio->bi_size))
+		return 0;
 	return m->offset + (bio->bi_sector - m->ms->ti->begin);
 }
 
@@ -413,6 +462,27 @@ static void map_region(struct dm_io_region *io, struct mirror *m,
 	io->count = bio->bi_size >> 9;
 }
 
+static void hold_bio(struct mirror_set *ms, struct bio *bio)
+{
+	/*
+	 * If device is suspended, complete the bio.
+	 */
+	if (atomic_read(&ms->suspend)) {
+		if (dm_noflush_suspending(ms->ti))
+			bio_endio(bio, DM_ENDIO_REQUEUE);
+		else
+			bio_endio(bio, -EIO);
+		return;
+	}
+
+	/*
+	 * Hold bio until the suspend is complete.
+	 */
+	spin_lock_irq(&ms->lock);
+	bio_list_add(&ms->holds, bio);
+	spin_unlock_irq(&ms->lock);
+}
+
 /*-----------------------------------------------------------------
  * Reads
  *---------------------------------------------------------------*/
@@ -511,7 +581,6 @@ static void write_callback(unsigned long error, void *context)
 	unsigned i, ret = 0;
 	struct bio *bio = (struct bio *) context;
 	struct mirror_set *ms;
-	int uptodate = 0;
 	int should_wake = 0;
 	unsigned long flags;
 
@@ -524,36 +593,27 @@ static void write_callback(unsigned long error, void *context)
 	 * This way we handle both writes to SYNC and NOSYNC
 	 * regions with the same code.
 	 */
-	if (likely(!error))
-		goto out;
+	if (likely(!error)) {
+		bio_endio(bio, ret);
+		return;
+	}
 
 	for (i = 0; i < ms->nr_mirrors; i++)
 		if (test_bit(i, &error))
 			fail_mirror(ms->mirror + i, DM_RAID1_WRITE_ERROR);
-		else
-			uptodate = 1;
 
-	if (unlikely(!uptodate)) {
-		DMERR("All replicated volumes dead, failing I/O");
-		/* None of the writes succeeded, fail the I/O. */
-		ret = -EIO;
-	} else if (errors_handled(ms)) {
-		/*
-		 * Need to raise event.  Since raising
-		 * events can block, we need to do it in
-		 * the main thread.
-		 */
-		spin_lock_irqsave(&ms->lock, flags);
-		if (!ms->failures.head)
-			should_wake = 1;
-		bio_list_add(&ms->failures, bio);
-		spin_unlock_irqrestore(&ms->lock, flags);
-		if (should_wake)
-			wakeup_mirrord(ms);
-		return;
-	}
-out:
-	bio_endio(bio, ret);
+	/*
+	 * Need to raise event.  Since raising
+	 * events can block, we need to do it in
+	 * the main thread.
+	 */
+	spin_lock_irqsave(&ms->lock, flags);
+	if (!ms->failures.head)
+		should_wake = 1;
+	bio_list_add(&ms->failures, bio);
+	spin_unlock_irqrestore(&ms->lock, flags);
+	if (should_wake)
+		wakeup_mirrord(ms);
 }
 
 static void do_write(struct mirror_set *ms, struct bio *bio)
@@ -562,7 +622,7 @@ static void do_write(struct mirror_set *ms, struct bio *bio)
 	struct dm_io_region io[ms->nr_mirrors], *dest = io;
 	struct mirror *m;
 	struct dm_io_request io_req = {
-		.bi_rw = WRITE,
+		.bi_rw = WRITE | (bio->bi_rw & WRITE_BARRIER),
 		.mem.type = DM_IO_BVEC,
 		.mem.ptr.bvec = bio->bi_io_vec + bio->bi_idx,
 		.notify.fn = write_callback,
@@ -603,6 +663,11 @@ static void do_writes(struct mirror_set *ms, struct bio_list *writes)
 	bio_list_init(&requeue);
 
 	while ((bio = bio_list_pop(writes))) {
+		if (unlikely(bio_empty_barrier(bio))) {
+			bio_list_add(&sync, bio);
+			continue;
+		}
+
 		region = dm_rh_bio_to_region(ms->rh, bio);
 
 		if (log->type->is_remote_recovering &&
@@ -672,8 +737,12 @@ static void do_writes(struct mirror_set *ms, struct bio_list *writes)
 		dm_rh_delay(ms->rh, bio);
 
 	while ((bio = bio_list_pop(&nosync))) {
-		map_bio(get_default_mirror(ms), bio);
-		generic_make_request(bio);
+		if (unlikely(ms->leg_failure) && errors_handled(ms))
+			hold_bio(ms, bio);
+		else {
+			map_bio(get_default_mirror(ms), bio);
+			generic_make_request(bio);
+		}
 	}
 }
 
@@ -681,20 +750,12 @@ static void do_failures(struct mirror_set *ms, struct bio_list *failures)
 {
 	struct bio *bio;
 
-	if (!failures->head)
-		return;
-
-	if (!ms->log_failure) {
-		while ((bio = bio_list_pop(failures))) {
-			ms->in_sync = 0;
-			dm_rh_mark_nosync(ms->rh, bio, bio->bi_size, 0);
-		}
+	if (likely(!failures->head))
 		return;
-	}
 
 	/*
 	 * If the log has failed, unattempted writes are being
-	 * put on the failures list.  We can't issue those writes
+	 * put on the holds list.  We can't issue those writes
 	 * until a log has been marked, so we must store them.
 	 *
 	 * If a 'noflush' suspend is in progress, we can requeue
@@ -709,23 +770,27 @@ static void do_failures(struct mirror_set *ms, struct bio_list *failures)
 	 * for us to treat them the same and requeue them
 	 * as well.
 	 */
-	if (dm_noflush_suspending(ms->ti)) {
-		while ((bio = bio_list_pop(failures)))
-			bio_endio(bio, DM_ENDIO_REQUEUE);
-		return;
-	}
+	while ((bio = bio_list_pop(failures))) {
+		if (!ms->log_failure) {
+			ms->in_sync = 0;
+			dm_rh_mark_nosync(ms->rh, bio);
+		}
 
-	if (atomic_read(&ms->suspend)) {
-		while ((bio = bio_list_pop(failures)))
+		/*
+		 * If all the legs are dead, fail the I/O.
+		 * If we have been told to handle errors, hold the bio
+		 * and wait for userspace to deal with the problem.
+		 * Otherwise pretend that the I/O succeeded. (This would
+		 * be wrong if the failed leg returned after reboot and
+		 * got replicated back to the good legs.)
+		 */
+		if (!get_valid_mirror(ms))
 			bio_endio(bio, -EIO);
-		return;
+		else if (errors_handled(ms))
+			hold_bio(ms, bio);
+		else
+			bio_endio(bio, 0);
 	}
-
-	spin_lock_irq(&ms->lock);
-	bio_list_merge(&ms->failures, failures);
-	spin_unlock_irq(&ms->lock);
-
-	delayed_wake(ms);
 }
 
 static void trigger_event(struct work_struct *work)
@@ -784,12 +849,17 @@ static struct mirror_set *alloc_context(unsigned int nr_mirrors,
 	}
 
 	spin_lock_init(&ms->lock);
+	bio_list_init(&ms->reads);
+	bio_list_init(&ms->writes);
+	bio_list_init(&ms->failures);
+	bio_list_init(&ms->holds);
 
 	ms->ti = ti;
 	ms->nr_mirrors = nr_mirrors;
 	ms->nr_regions = dm_sector_div_up(ti->len, region_size);
 	ms->in_sync = 0;
 	ms->log_failure = 0;
+	ms->leg_failure = 0;
 	atomic_set(&ms->suspend, 0);
 	atomic_set(&ms->default_mirror, DEFAULT_MIRROR);
 
@@ -889,7 +959,8 @@ static struct dm_dirty_log *create_dirty_log(struct dm_target *ti,
 		return NULL;
 	}
 
-	dl = dm_dirty_log_create(argv[0], ti, param_count, argv + 2);
+	dl = dm_dirty_log_create(argv[0], ti, mirror_flush, param_count,
+				 argv + 2);
 	if (!dl) {
 		ti->error = "Error creating mirror dirty log";
 		return NULL;
@@ -995,6 +1066,7 @@ static int mirror_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 
 	ti->private = ms;
 	ti->split_io = dm_rh_get_region_size(ms->rh);
+	ti->num_flush_requests = 1;
 
 	ms->kmirrord_wq = create_singlethread_workqueue("kmirrord");
 	if (!ms->kmirrord_wq) {
@@ -1122,7 +1194,8 @@ static int mirror_end_io(struct dm_target *ti, struct bio *bio,
 	 * We need to dec pending if this was a write.
 	 */
 	if (rw == WRITE) {
-		dm_rh_dec(ms->rh, map_context->ll);
+		if (likely(!bio_empty_barrier(bio)))
+			dm_rh_dec(ms->rh, map_context->ll);
 		return error;
 	}
 
@@ -1180,6 +1253,9 @@ static void mirror_presuspend(struct dm_target *ti)
 	struct mirror_set *ms = (struct mirror_set *) ti->private;
 	struct dm_dirty_log *log = dm_rh_dirty_log(ms->rh);
 
+	struct bio_list holds;
+	struct bio *bio;
+
 	atomic_set(&ms->suspend, 1);
 
 	/*
@@ -1202,6 +1278,22 @@ static void mirror_presuspend(struct dm_target *ti)
 	 * we know that all of our I/O has been pushed.
 	 */
 	flush_workqueue(ms->kmirrord_wq);
+
+	/*
+	 * Now set ms->suspend is set and the workqueue flushed, no more
+	 * entries can be added to ms->hold list, so process it.
+	 *
+	 * Bios can still arrive concurrently with or after this
+	 * presuspend function, but they cannot join the hold list
+	 * because ms->suspend is set.
+	 */
+	spin_lock_irq(&ms->lock);
+	holds = ms->holds;
+	bio_list_init(&ms->holds);
+	spin_unlock_irq(&ms->lock);
+
+	while ((bio = bio_list_pop(&holds)))
+		hold_bio(ms, bio);
 }
 
 static void mirror_postsuspend(struct dm_target *ti)
@@ -1244,7 +1336,8 @@ static char device_status_char(struct mirror *m)
 	if (!atomic_read(&(m->error_count)))
 		return 'A';
 
-	return (test_bit(DM_RAID1_WRITE_ERROR, &(m->error_type))) ? 'D' :
+	return (test_bit(DM_RAID1_FLUSH_ERROR, &(m->error_type))) ? 'F' :
+		(test_bit(DM_RAID1_WRITE_ERROR, &(m->error_type))) ? 'D' :
 		(test_bit(DM_RAID1_SYNC_ERROR, &(m->error_type))) ? 'S' :
 		(test_bit(DM_RAID1_READ_ERROR, &(m->error_type))) ? 'R' : 'U';
 }