[PATCH 2/4] md: add flush_idx support for stacked devices

[PATCH 2/4] md: add flush_idx support for stacked devices

[Dmitry Monakhov]

For single device mapping it is safe to introduce get_flush_idx conception.
This cover most common case linear mapping for a single dev.

CC: Alasdair Kergon <agk@redhat.com>
CC: Mike Snitzer <snitzer@redhat.com>
CC: dm-devel@redhat.com
Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
---
 drivers/md/dm.c |   31 +++++++++++++++++++++++++++++++
 1 files changed, 31 insertions(+), 0 deletions(-)

diff --git a/drivers/md/dm.c b/drivers/md/dm.c
index 32b958d..042d172 100644
--- a/drivers/md/dm.c
+++ b/drivers/md/dm.c
@@ -1843,6 +1843,36 @@ static int dm_any_congested(void *congested_data, int bdi_bits)
 	return r;
 }
 
+static int dm_iter_get_flush_idx(struct dm_target *ti,
+				struct dm_dev *dev, sector_t start,
+				sector_t len, void *data)
+{
+	struct block_device *bdev = dev->bdev;
+	struct request_queue *q = bdev_get_queue(bdev);
+
+	return blk_get_flush_idx(q,  *((bool *)data));
+}
+
+static unsigned dm_get_flush_idx(struct request_queue *q, bool queued)
+{
+	struct dm_target *ti;
+	struct mapped_device *md = q->queuedata;
+	struct dm_table *map = dm_get_live_table_fast(md);
+	unsigned fid = 0;
+
+	if (unlikely(!map || dm_table_get_num_targets(map) != 1))
+		goto out;
+
+	ti = dm_table_get_target(map, 0);
+	if (unlikely(!ti || !ti->type->iterate_devices))
+		goto out;
+
+	fid = ti->type->iterate_devices(ti, dm_iter_get_flush_idx, &queued);
+out:
+	dm_put_live_table_fast(md);
+	return fid;
+}
+
 /*-----------------------------------------------------------------
  * An IDR is used to keep track of allocated minor numbers.
  *---------------------------------------------------------------*/
@@ -1962,6 +1992,7 @@ static struct mapped_device *alloc_dev(int minor)
 		goto bad_queue;
 
 	dm_init_md_queue(md);
+	blk_queue_get_flush_idx(md->queue, dm_get_flush_idx);
 
 	md->disk = alloc_disk(1);
 	if (!md->disk)
-- 
1.7.1

Re: [PATCH 2/4] md: add flush_idx support for stacked devices

[Mike Snitzer]

On Tue, Oct 14 2014 at  6:03am -0400,
Dmitry Monakhov <dmonakhov@openvz.org> wrote:

> For single device mapping it is safe to introduce get_flush_idx conception.
> This cover most common case linear mapping for a single dev.

You really should email the entire patchset rather than forcing the need
to hunt for the remainder of the patches that actually provide context
for this change.