linux/drivers/md/dm-writecache.c, branch v5.14

dm writecache: make writeback pause configurable

2021-06-28T20:30:13Z

Commit 95b88f4d71cb953e02206be3c757083601391a0f ("dm writecache: pause writeback if cache full and origin being written directly") introduced a code that pauses cache flushing if we are issuing writes directly to the origin. Improve that initial commit by making the timeout code configurable (via the option "pause_writeback"). Also change the default from 1s to 3s because it performed better. Signed-off-by: Mikulas Patocka Signed-off-by: Mike Snitzer

dm writecache: pause writeback if cache full and origin being written directly

2021-06-25T20:04:01Z

Implementation reuses dm_io_tracker, that until now was only used by dm-cache, to track if any writes were issued directly to the origin (due to cache being full) within the last second. If so writeback is paused for a second. This change improves performance for when the cache is full and IO is issued directly to the origin device (rather than through the cache). Depends-on: d53f1fafec9d ("dm writecache: do direct write if the cache is full") Suggested-by: Joe Thornber Signed-off-by: Mikulas Patocka Signed-off-by: Mike Snitzer

dm writecache: add optional "metadata_only" parameter

2021-06-25T19:25:21Z

Add a "metadata_only" parameter that when present: only metadata is promoted to the cache. This option improves performance for heavier REQ_META workloads (e.g. device-mapper-test-suite's "git clone and checkout" benchmark improves from 341s to 312s). Signed-off-by: Mikulas Patocka Signed-off-by: Mike Snitzer

dm writecache: write at least 4k when committing

2021-06-21T20:15:10Z

SSDs perform badly with sub-4k writes (because they perfrorm read-modify-write internally), so make sure writecache writes at least 4k when committing. Fixes: 991bd8d7bc78 ("dm writecache: commit just one block, not a full page") Signed-off-by: Mikulas Patocka Signed-off-by: Mike Snitzer

dm writecache: flush origin device when writing and cache is full

2021-06-16T16:57:14Z

Commit d53f1fafec9d086f1c5166436abefdaef30e0363 ("dm writecache: do direct write if the cache is full") changed dm-writecache, so that it writes directly to the origin device if the cache is full. Unfortunately, it doesn't forward flush requests to the origin device, so that there is a bug where flushes are being ignored. Fix this by adding missing flush forwarding. For PMEM mode, we fix this bug by disabling direct writes to the origin device, because it performs better. Signed-off-by: Mikulas Patocka Fixes: d53f1fafec9d ("dm writecache: do direct write if the cache is full") Cc: stable@vger.kernel.org # v5.7+ Signed-off-by: Mike Snitzer

dm writecache: have ssd writeback wait if the kcopyd workqueue is busy

2021-06-15T19:42:03Z

Make dm-writecache wait if the kcopyd workqueue is busy (as will happen if waiting for page allocation or inside submit_bio). This change improves performance of "mkfs.ext2" by approximately 20% on one testbed. Signed-off-by: Mikulas Patocka Signed-off-by: Mike Snitzer

dm writecache: use list_move instead of list_del/list_add in writecache_writeback()

2021-06-14T15:49:04Z

Reported-by: Hulk Robot Signed-off-by: Baokun Li Signed-off-by: Mike Snitzer

dm writecache: commit just one block, not a full page

2021-06-14T15:45:30Z

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 Signed-off-by: Mike Snitzer

dm writecache: remove unused gfp_t argument from wc_add_block()

2021-06-14T15:44:06Z

Signed-off-by: Mikulas Patocka Signed-off-by: Mike Snitzer

dm writecache: interrupt writeback if suspended

2021-06-04T16:07:27Z

If the DM device is suspended, interrupt the writeback sequence so that there is no excessive suspend delay. Signed-off-by: Mikulas Patocka Signed-off-by: Mike Snitzer