Skip to content

blk-mq: improvement CPU hotplug

Ming Lei dm-devel linux-block 12 versions – In review

Get series with b4
b4 am [email protected]

Versions newest first · linked by author and title

  1. v11 Latest12 patches 18 repliesIn review
  2. v1011 patches 29 repliesSuperseded
  3. v911 patches 5 repliesSuperseded
Earlier versions · 9
VersionPostedPatchesRepliesStatus
v811 patches68 repliesSuperseded
v79 patches12 repliesSuperseded
v68 patches10 repliesSuperseded
v56 patches19 repliesSuperseded
v45 patches26 repliesSuperseded
v35 patches12 repliesSuperseded
v2 posting 2/25 patches10 repliesSuperseded
v2 posting 1/25 patches11 repliesSuperseded
v1 RFC7 patches8 repliesSuperseded

Changes in v11 from the cover letter

- drop new callback from blk_mq_all_tag_busy_iter, add new helper of blk_mq_all_tag_iter (5/12), as suggested by Bart
- fix request allocation hang in case of queue freeze(11/12), as reported by Bart

Patches in v11 No applies recorded yet

#Subject and filesSizeRepliesApplied
1/12block: clone nr_integrity_segments and write_hint in blk_rq_prep_cloneblock/blk-core.c+4 -00 replies
2/12block: add helper for copying requestblock/blk-core.c block/blk.h+20 -130 replies
3/12blk-mq: mark blk_mq_get_driver_tag as staticblock/blk-mq.c block/blk-mq.h+1 -20 replies
4/12blk-mq: assign rq->tag in blk_mq_get_driver_tagblock/blk-flush.c block/blk-mq.c block/blk-mq.h +1 more+51 -680 replies
5/12blk-mq: add blk_mq_all_tag_iterblock/blk-mq-tag.c block/blk-mq-tag.h+31 -41 reply
6/12blk-mq: prepare for draining IO when hctx's all CPUs are offlineblock/blk-mq-debugfs.c block/blk-mq.c drivers/block/loop.c +3 more+27 -23 replies
7/12blk-mq: stop to handle IO and drain IO before hctx becomes inactiveblock/blk-mq-debugfs.c block/blk-mq.c include/linux/blk-mq.h+110 -114 replies
8/12block: add blk_end_flush_machineryblock/blk-flush.c block/blk.h+120 -71 reply
9/12blk-mq: add blk_mq_hctx_handle_dead_cpu for handling cpu deadblock/blk-mq.c+17 -121 reply
10/12block: add request allocation flag of BLK_MQ_REQ_FORCEblock/blk-core.c include/linux/blk-mq.h+12 -01 reply
11/12blk-mq: re-submit IO in case that hctx is inactiveblock/blk-mq.c+111 -55 replies
12/12block: deactivate hctx when the hctx is actually inactiveblock/blk-mq.c+10 -200 replies

Reviews trailers per patch

9 of 12 patches have a Reviewed-by or Acked-by.

PatchHannes ReineckeChristoph HellwigMartin K. Petersen
1/12R-b R-b R-b
2/12R-b R-b R-b
3/12R-b R-b R-b
4/12R-b R-b
5/12R-b
6/12R-b
7/12
8/12R-b
9/12R-b R-b
10/12
11/12
12/12R-b R-b

Trailers given in replies are filled; those carried in the patches from earlier versions are outlined with dashes. Scroll the table sideways for all 3 reviewers.

Read the v11 thread

Cover letter [PATCH V11 00/12] blk-mq: improvement CPU hotplug

Hi,

Thomas mentioned:
"
That was the constraint of managed interrupts from the very beginning:

The driver/subsystem has to quiesce the interrupt line and the associated queue _before_ it gets shutdown in CPU unplug and not fiddle with it until it's restarted by the core when the CPU is plugged in again.
"

But no drivers or blk-mq do that before one hctx becomes inactive(all CPUs for one hctx are offline), and even it is worse, blk-mq stills tries to run hw queue after hctx is dead, see blk_mq_hctx_notify_dead().

This patchset tries to address the issue by two stages:

1) add one new cpuhp state of CPUHP_AP_BLK_MQ_ONLINE

- mark the hctx as internal stopped, and drain all in-flight requests if the hctx is going to be dead.

2) re-submit IO in the state of CPUHP_BLK_MQ_DEAD after the hctx becomes dead

- steal bios from the request, and resubmit them via generic_make_request(), then these IO will be mapped to other live hctx for dispatch

Thanks John Garry for running lots of tests on arm64 with this patchset and co-working on investigating all kinds of issues.

Thanks Christoph's review on V7 & V8.

Please consider it for v5.8.

https://github.com/ming1/linux/commits/v5.7-rc-blk-mq-improve-cpu-hotplug

V11:
- drop new callback from blk_mq_all_tag_busy_iter, add new helper of blk_mq_all_tag_iter (5/12), as suggested by Bart
- fix request allocation hang in case of queue freeze(11/12), as reported by Bart

V10:
- fix double bio complete in request resubmission(10/11)
- add tested-by tag

V9:
- add Reviewed-by tag
- document more on memory barrier usage between getting driver tag and handling cpu offline(7/11)
- small code cleanup as suggested by Chritoph(7/11)
- rebase against for-5.8/block(1/11, 2/11)
V8:
- add patches to share code with blk_rq_prep_clone
- code re-organization as suggested by Christoph, most of them are in 04/11, 10/11
- add reviewed-by tag

V7:
- fix updating .nr_active in get_driver_tag
- add hctx->cpumask check in cpuhp handler
- only drain requests which tag is >= 0
- pass more aggressive cpuhotplug&io test

V6:
- simplify getting driver tag, so that we can drain in-flight requests correctly without using synchronize_rcu()
- handle re-submission of flush & passthrough request correctly

V5:
- rename BLK_MQ_S_INTERNAL_STOPPED as BLK_MQ_S_INACTIVE
- re-factor code for re-submit requests in cpu dead hotplug handler
- address requeue corner case

V4:
- resubmit IOs in dispatch list in case that this hctx is dead

V3:
- re-organize patch 2 & 3 a bit for addressing Hannes's comment
- fix patch 4 for avoiding potential deadlock, as found by Hannes

V2:
- patch4 & patch 5 in V1 have been merged to block tree, so remove them
- address comments from John Garry and Minwoo

*** BLURB HERE ***

Ming Lei (12):
block: clone nr_integrity_segments and write_hint in blk_rq_prep_clone
block: add helper for copying request
blk-mq: mark blk_mq_get_driver_tag as static
blk-mq: assign rq->tag in blk_mq_get_driver_tag
blk-mq: add blk_mq_all_tag_iter
blk-mq: prepare for draining IO when hctx's all CPUs are offline
blk-mq: stop to handle IO and drain IO before hctx becomes inactive
block: add blk_end_flush_machinery
blk-mq: add blk_mq_hctx_handle_dead_cpu for handling cpu dead
block: add request allocation flag of BLK_MQ_REQ_FORCE
blk-mq: re-submit IO in case that hctx is inactive
block: deactivate hctx when the hctx is actually inactive

 block/blk-core.c           |  32 +++-
 block/blk-flush.c          | 141 ++++++++++++---
 block/blk-mq-debugfs.c     |   2 +
 block/blk-mq-tag.c         |  33 +++-
 block/blk-mq-tag.h         |   2 +
 block/blk-mq.c             | 356 +++++++++++++++++++++++++++++--------
 block/blk-mq.h             |  22 ++-
 block/blk.h                |  11 +-
 drivers/block/loop.c       |   2 +-
 drivers/md/dm-rq.c         |   2 +-
 include/linux/blk-mq.h     |  14 ++
 include/linux/cpuhotplug.h |   1 +
 12 files changed, 494 insertions(+), 124 deletions(-)
Signature · 1 line
2.25.2