[dm-devel] [PATCH AUTOSEL 5.12 061/160] dm space maps: don't reset space map allocation cursor when committing

[dm-devel] [PATCH AUTOSEL 5.12 061/160] dm space maps: don't reset space map allocation cursor when committing

[Sasha Levin]

From: Joe Thornber <ejt@redhat.com>

[ Upstream commit 5faafc77f7de69147d1e818026b9a0cbf036a7b2 ]

Current commit code resets the place where the search for free blocks
will begin back to the start of the metadata device.  There are a couple
of repercussions to this:

- The first allocation after the commit is likely to take longer than
  normal as it searches for a free block in an area that is likely to
  have very few free blocks (if any).

- Any free blocks it finds will have been recently freed.  Reusing them
  means we have fewer old copies of the metadata to aid recovery from
  hardware error.

Fix these issues by leaving the cursor alone, only resetting when the
search hits the end of the metadata device.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/md/persistent-data/dm-space-map-disk.c     | 9 ++++++++-
 drivers/md/persistent-data/dm-space-map-metadata.c | 9 ++++++++-
 2 files changed, 16 insertions(+), 2 deletions(-)

diff --git a/drivers/md/persistent-data/dm-space-map-disk.c b/drivers/md/persistent-data/dm-space-map-disk.c
index bf4c5e2ccb6f..e0acae7a3815 100644
--- a/drivers/md/persistent-data/dm-space-map-disk.c
+++ b/drivers/md/persistent-data/dm-space-map-disk.c
@@ -171,6 +171,14 @@ static int sm_disk_new_block(struct dm_space_map *sm, dm_block_t *b)
 	 * Any block we allocate has to be free in both the old and current ll.
 	 */
 	r = sm_ll_find_common_free_block(&smd->old_ll, &smd->ll, smd->begin, smd->ll.nr_blocks, b);
+	if (r == -ENOSPC) {
+		/*
+		 * There's no free block between smd->begin and the end of the metadata device.
+		 * We search before smd->begin in case something has been freed.
+		 */
+		r = sm_ll_find_common_free_block(&smd->old_ll, &smd->ll, 0, smd->begin, b);
+	}
+
 	if (r)
 		return r;
 
@@ -199,7 +207,6 @@ static int sm_disk_commit(struct dm_space_map *sm)
 		return r;
 
 	memcpy(&smd->old_ll, &smd->ll, sizeof(smd->old_ll));
-	smd->begin = 0;
 	smd->nr_allocated_this_transaction = 0;
 
 	r = sm_disk_get_nr_free(sm, &nr_free);
diff --git a/drivers/md/persistent-data/dm-space-map-metadata.c b/drivers/md/persistent-data/dm-space-map-metadata.c
index 9e3c64ec2026..da439ac85796 100644
--- a/drivers/md/persistent-data/dm-space-map-metadata.c
+++ b/drivers/md/persistent-data/dm-space-map-metadata.c
@@ -452,6 +452,14 @@ static int sm_metadata_new_block_(struct dm_space_map *sm, dm_block_t *b)
 	 * Any block we allocate has to be free in both the old and current ll.
 	 */
 	r = sm_ll_find_common_free_block(&smm->old_ll, &smm->ll, smm->begin, smm->ll.nr_blocks, b);
+	if (r == -ENOSPC) {
+		/*
+		 * There's no free block between smm->begin and the end of the metadata device.
+		 * We search before smm->begin in case something has been freed.
+		 */
+		r = sm_ll_find_common_free_block(&smm->old_ll, &smm->ll, 0, smm->begin, b);
+	}
+
 	if (r)
 		return r;
 
@@ -503,7 +511,6 @@ static int sm_metadata_commit(struct dm_space_map *sm)
 		return r;
 
 	memcpy(&smm->old_ll, &smm->ll, sizeof(smm->old_ll));
-	smm->begin = 0;
 	smm->allocated_this_transaction = 0;
 
 	return 0;
-- 
2.30.2

--
dm-devel mailing list
dm-devel@redhat.com
https://listman.redhat.com/mailman/listinfo/dm-devel

[dm-devel] [PATCH AUTOSEL 5.12 062/160] dm writecache: don't split bios when overwriting contiguous cache content

[Sasha Levin]

From: Mikulas Patocka <mpatocka@redhat.com>

[ Upstream commit ee50cc19d80e9b9a8283d1fb517a778faf2f6899 ]

If dm-writecache overwrites existing cached data, it splits the
incoming bio into many block-sized bios. The I/O scheduler does merge
these bios into one large request but this needless splitting and
merging causes performance degradation.

Fix this by avoiding bio splitting if the cache target area that is
being overwritten is contiguous.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/md/dm-writecache.c | 38 ++++++++++++++++++++++++++++++--------
 1 file changed, 30 insertions(+), 8 deletions(-)

diff --git a/drivers/md/dm-writecache.c b/drivers/md/dm-writecache.c
index 4f72b6f66c3a..7bb4d83e90cc 100644
--- a/drivers/md/dm-writecache.c
+++ b/drivers/md/dm-writecache.c
@@ -1360,14 +1360,18 @@ static int writecache_map(struct dm_target *ti, struct bio *bio)
 	} else {
 		do {
 			bool found_entry = false;
+			bool search_used = false;
 			if (writecache_has_error(wc))
 				goto unlock_error;
 			e = writecache_find_entry(wc, bio->bi_iter.bi_sector, 0);
 			if (e) {
-				if (!writecache_entry_is_committed(wc, e))
+				if (!writecache_entry_is_committed(wc, e)) {
+					search_used = true;
 					goto bio_copy;
+				}
 				if (!WC_MODE_PMEM(wc) && !e->write_in_progress) {
 					wc->overwrote_committed = true;
+					search_used = true;
 					goto bio_copy;
 				}
 				found_entry = true;
@@ -1404,13 +1408,31 @@ static int writecache_map(struct dm_target *ti, struct bio *bio)
 				sector_t current_cache_sec = start_cache_sec + (bio_size >> SECTOR_SHIFT);
 
 				while (bio_size < bio->bi_iter.bi_size) {
-					struct wc_entry *f = writecache_pop_from_freelist(wc, current_cache_sec);
-					if (!f)
-						break;
-					write_original_sector_seq_count(wc, f, bio->bi_iter.bi_sector +
-									(bio_size >> SECTOR_SHIFT), wc->seq_count);
-					writecache_insert_entry(wc, f);
-					wc->uncommitted_blocks++;
+					if (!search_used) {
+						struct wc_entry *f = writecache_pop_from_freelist(wc, current_cache_sec);
+						if (!f)
+							break;
+						write_original_sector_seq_count(wc, f, bio->bi_iter.bi_sector +
+										(bio_size >> SECTOR_SHIFT), wc->seq_count);
+						writecache_insert_entry(wc, f);
+						wc->uncommitted_blocks++;
+					} else {
+						struct wc_entry *f;
+						struct rb_node *next = rb_next(&e->rb_node);
+						if (!next)
+							break;
+						f = container_of(next, struct wc_entry, rb_node);
+						if (f != e + 1)
+							break;
+						if (read_original_sector(wc, f) !=
+						    read_original_sector(wc, e) + (wc->block_size >> SECTOR_SHIFT))
+							break;
+						if (unlikely(f->write_in_progress))
+							break;
+						if (writecache_entry_is_committed(wc, f))
+							wc->overwrote_committed = true;
+						e = f;
+					}
 					bio_size += wc->block_size;
 					current_cache_sec += wc->block_size >> SECTOR_SHIFT;
 				}
-- 
2.30.2

--
dm-devel mailing list
dm-devel@redhat.com
https://listman.redhat.com/mailman/listinfo/dm-devel

[dm-devel] [PATCH AUTOSEL 5.12 063/160] dm: Fix dm_accept_partial_bio() relative to zone management commands

[Sasha Levin]

From: Damien Le Moal <damien.lemoal@wdc.com>

[ Upstream commit 6842d264aa5205da338b6dcc6acfa2a6732558f1 ]

Fix dm_accept_partial_bio() to actually check that zone management
commands are not passed as explained in the function documentation
comment. Also, since a zone append operation cannot be split, add
REQ_OP_ZONE_APPEND as a forbidden command.

White lines are added around the group of BUG_ON() calls to make the
code more legible.

Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/md/dm.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/drivers/md/dm.c b/drivers/md/dm.c
index 3f3be9408afa..05747d2736a7 100644
--- a/drivers/md/dm.c
+++ b/drivers/md/dm.c
@@ -1239,8 +1239,8 @@ static int dm_dax_zero_page_range(struct dax_device *dax_dev, pgoff_t pgoff,
 
 /*
  * A target may call dm_accept_partial_bio only from the map routine.  It is
- * allowed for all bio types except REQ_PREFLUSH, REQ_OP_ZONE_RESET,
- * REQ_OP_ZONE_OPEN, REQ_OP_ZONE_CLOSE and REQ_OP_ZONE_FINISH.
+ * allowed for all bio types except REQ_PREFLUSH, REQ_OP_ZONE_* zone management
+ * operations and REQ_OP_ZONE_APPEND (zone append writes).
  *
  * dm_accept_partial_bio informs the dm that the target only wants to process
  * additional n_sectors sectors of the bio and the rest of the data should be
@@ -1270,9 +1270,13 @@ void dm_accept_partial_bio(struct bio *bio, unsigned n_sectors)
 {
 	struct dm_target_io *tio = container_of(bio, struct dm_target_io, clone);
 	unsigned bi_size = bio->bi_iter.bi_size >> SECTOR_SHIFT;
+
 	BUG_ON(bio->bi_opf & REQ_PREFLUSH);
+	BUG_ON(op_is_zone_mgmt(bio_op(bio)));
+	BUG_ON(bio_op(bio) == REQ_OP_ZONE_APPEND);
 	BUG_ON(bi_size > *tio->len_ptr);
 	BUG_ON(n_sectors > bi_size);
+
 	*tio->len_ptr -= bi_size - n_sectors;
 	bio->bi_iter.bi_size = n_sectors << SECTOR_SHIFT;
 }
-- 
2.30.2

--
dm-devel mailing list
dm-devel@redhat.com
https://listman.redhat.com/mailman/listinfo/dm-devel

[dm-devel] [PATCH AUTOSEL 5.12 091/160] dm writecache: commit just one block, not a full page

[Sasha Levin]

From: Mikulas Patocka <mpatocka@redhat.com>

[ Upstream commit 991bd8d7bc78966b4dc427b53a144f276bffcd52 ]

Some architectures have pages larger than 4k and committing a full
page causes needless overhead.

Fix this by writing a single block when committing the superblock.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/md/dm-writecache.c | 6 +-----
 1 file changed, 1 insertion(+), 5 deletions(-)

diff --git a/drivers/md/dm-writecache.c b/drivers/md/dm-writecache.c
index 7bb4d83e90cc..51b26db56ba9 100644
--- a/drivers/md/dm-writecache.c
+++ b/drivers/md/dm-writecache.c
@@ -532,11 +532,7 @@ static void ssd_commit_superblock(struct dm_writecache *wc)
 
 	region.bdev = wc->ssd_dev->bdev;
 	region.sector = 0;
-	region.count = PAGE_SIZE >> SECTOR_SHIFT;
-
-	if (unlikely(region.sector + region.count > wc->metadata_sectors))
-		region.count = wc->metadata_sectors - region.sector;
-
+	region.count = wc->block_size >> SECTOR_SHIFT;
 	region.sector += wc->start_sector;
 
 	req.bi_op = REQ_OP_WRITE;
-- 
2.30.2

--
dm-devel mailing list
dm-devel@redhat.com
https://listman.redhat.com/mailman/listinfo/dm-devel