mbox series

[v4,00/13] blk-mq: Reduce static requests memory footprint for shared sbitmap

Message ID 1632472110-244938-1-git-send-email-john.garry@huawei.com (mailing list archive)
Headers show
Series blk-mq: Reduce static requests memory footprint for shared sbitmap | expand

Message

John Garry Sept. 24, 2021, 8:28 a.m. UTC
Currently a full set of static requests are allocated per hw queue per
tagset when shared sbitmap is used.

However, only tagset->queue_depth number of requests may be active at
any given time. As such, only tagset->queue_depth number of static
requests are required.

The same goes for using an IO scheduler, which allocates a full set of
static requests per hw queue per request queue.

This series changes shared sbitmap support by using a shared tags per
tagset and request queue. Ming suggested something along those lines in
v1 review. But we'll keep name "shared sbitmap" name as it is familiar. In
using a shared tags, the static rqs also become shared, reducing the
number of sets of static rqs, reducing memory usage.

Patch "blk-mq: Use shared tags for shared sbitmap support" is a bit big,
and could potentially be broken down. But then maintaining ability to
bisect becomes harder and each sub-patch would get more convoluted.

For megaraid sas driver on my 128-CPU arm64 system with 1x SATA disk, we
save approx. 300MB(!) [370MB -> 60MB]

Baseline is v5.15-rc2

Changes since v3:
- Fix transient error handling issue in  05/13
- Add Hannes RB tags (thanks!)

Changes since v2:
- Make blk_mq_clear_rq_mapping() static again
- Various function renaming for conciseness and consistency
- Add/refactor alloc/free map and rqs function
- Drop the new blk_mq_ops init_request method in favour of passing an
  invalid HW queue index for shared sbitmap
- Add patch to not clear rq mapping for driver tags
- Remove blk_mq_init_bitmap_tags()
- Add some more RB tags (thanks!)

Changes since v1:
- Switch to use blk_mq_tags for shared sbitmap
- Add new blk_mq_ops init request callback
- Add some RB tags (thanks!)


John Garry (13):
  blk-mq: Change rqs check in blk_mq_free_rqs()
  block: Rename BLKDEV_MAX_RQ -> BLKDEV_DEFAULT_RQ
  blk-mq: Relocate shared sbitmap resize in blk_mq_update_nr_requests()
  blk-mq: Invert check in blk_mq_update_nr_requests()
  blk-mq-sched: Rename blk_mq_sched_alloc_{tags -> map_and_rqs}()
  blk-mq-sched: Rename blk_mq_sched_free_{requests -> rqs}()
  blk-mq: Pass driver tags to blk_mq_clear_rq_mapping()
  blk-mq: Don't clear driver tags own mapping
  blk-mq: Add blk_mq_tag_update_sched_shared_sbitmap()
  blk-mq: Add blk_mq_alloc_map_and_rqs()
  blk-mq: Refactor and rename blk_mq_free_map_and_{requests->rqs}()
  blk-mq: Use shared tags for shared sbitmap support
  blk-mq: Stop using pointers for blk_mq_tags bitmap tags

 block/bfq-iosched.c    |   4 +-
 block/blk-core.c       |   4 +-
 block/blk-mq-debugfs.c |   8 +-
 block/blk-mq-sched.c   | 116 +++++++++++------------
 block/blk-mq-sched.h   |   4 +-
 block/blk-mq-tag.c     | 125 +++++++++----------------
 block/blk-mq-tag.h     |  14 +--
 block/blk-mq.c         | 208 ++++++++++++++++++++++++-----------------
 block/blk-mq.h         |  18 ++--
 block/blk.h            |   2 +-
 block/kyber-iosched.c  |   4 +-
 block/mq-deadline.c    |   2 +-
 drivers/block/rbd.c    |   2 +-
 include/linux/blk-mq.h |  15 ++-
 include/linux/blkdev.h |   5 +-
 15 files changed, 255 insertions(+), 276 deletions(-)