linux-block.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Weiping Zhang <zhangweiping@didiglobal.com>
To: <axboe@kernel.dk>, <tom.leiming@gmail.com>, <bvanassche@acm.org>
Cc: <linux-block@vger.kernel.org>
Subject: [PATCH v4 0/6] Fix potential kernel panic when increase hardware queue
Date: Tue, 28 Apr 2020 21:25:19 +0800	[thread overview]
Message-ID: <cover.1588078594.git.zhangweiping@didiglobal.com> (raw)

Hi,

This series do some improvement base on V2, and also fix two memleaks.
And V2 import a regression for blktest/block/test/029, this test case
will increase and decrease hardware queue count quickly.


Memleak 1:

__blk_mq_alloc_rq_maps
	__blk_mq_alloc_rq_map

if fail
	blk_mq_free_rq_map

Actually, __blk_mq_alloc_rq_map alloc both map and request, here
also need free request.

Memleak 2:
When driver decrease hardware queue, set->nr_hw_queues will be changed
firstly in blk_mq_realloc_tag_set_tags or __blk_mq_update_nr_hw_queues,
then blk_mq_realloc_hw_ctxs and blk_mq_map_swqueue, even
blk_mq_free_tag_set have no chance to free these hardware queue resource,
because they iterate hardware queue by
for (i = 0; i < set->nr_hw_queues; i++).

Patch1: fix Memleak 1.
Patch2: fix prev_nr_hw_queues issue, need be saved before change.
Patch3: refactor __blk_mq_alloc_rq_maps and fix Memleak 2.
Patch4: fix potential kernel panic when increase hardware queue.
Patch5~6: rename two function, because these two function alloc both
map and request, and keep in pair with blk_mq_free_map_and_request(s).

Changes since V3:
* record patchset, fix issue fistly then rename.
* rename function to blk_mq_alloc_map_and_request

Changes since V2:
 * rename some functions name and fix memleak when free map and requests
 * Not free new allocated map and request, they will be relased when tagset gone

Changes since V1:
 * Add fix for potential kernel panic when increase hardware queue

Weiping Zhang (6):
  block: free both rq_map and request
  block: save previous hardware queue count before udpate
  block: refactor __blk_mq_alloc_rq_maps
  block: alloc map and request for new hardware queue
  block: rename __blk_mq_alloc_rq_map
  block: rename blk_mq_alloc_rq_maps

 block/blk-mq.c         | 47 +++++++++++++++++++++++++++++-------------
 include/linux/blk-mq.h |  1 +
 2 files changed, 34 insertions(+), 14 deletions(-)

-- 
2.18.1


WARNING: multiple messages have this Message-ID (diff)
From: Weiping Zhang <zhangweiping@didiglobal.com>
To: <axboe@kernel.dk>, <tom.leiming@gmail.com>, <bvanassche@acm.org>
Cc: <linux-block@vger.kernel.org>
Subject: [RESEND PATCH v4 0/6] Fix potential kernel panic when increase hardware queue
Date: Tue, 28 Apr 2020 21:29:19 +0800	[thread overview]
Message-ID: <cover.1588078594.git.zhangweiping@didiglobal.com> (raw)
Message-ID: <20200428132919.9bHRgAY3SfYS2tSRPt1En2xSg9DiFTW2zmfXG-Ot570@z> (raw)

Hi,

This series do some improvement base on V2, and also fix two memleaks.
And V2 import a regression for blktest/block/test/029, this test case
will increase and decrease hardware queue count quickly.


Memleak 1:

__blk_mq_alloc_rq_maps
	__blk_mq_alloc_rq_map

if fail
	blk_mq_free_rq_map

Actually, __blk_mq_alloc_rq_map alloc both map and request, here
also need free request.

Memleak 2:
When driver decrease hardware queue, set->nr_hw_queues will be changed
firstly in blk_mq_realloc_tag_set_tags or __blk_mq_update_nr_hw_queues,
then blk_mq_realloc_hw_ctxs and blk_mq_map_swqueue, even
blk_mq_free_tag_set have no chance to free these hardware queue resource,
because they iterate hardware queue by
for (i = 0; i < set->nr_hw_queues; i++).

Patch1: fix Memleak 1.
Patch2: fix prev_nr_hw_queues issue, need be saved before change.
Patch3: refactor __blk_mq_alloc_rq_maps and fix Memleak 2.
Patch4: fix potential kernel panic when increase hardware queue.
Patch5~6: rename two function, because these two function alloc both
map and request, and keep in pair with blk_mq_free_map_and_request(s).

Changes since V3:
* record patchset, fix issue fistly then rename.
* rename function to blk_mq_alloc_map_and_request

Changes since V2:
 * rename some functions name and fix memleak when free map and requests
 * Not free new allocated map and request, they will be relased when tagset gone

Changes since V1:
 * Add fix for potential kernel panic when increase hardware queue

Weiping Zhang (6):
  block: free both rq_map and request
  block: save previous hardware queue count before udpate
  block: refactor __blk_mq_alloc_rq_maps
  block: alloc map and request for new hardware queue
  block: rename __blk_mq_alloc_rq_map
  block: rename blk_mq_alloc_rq_maps

 block/blk-mq.c         | 47 +++++++++++++++++++++++++++++-------------
 include/linux/blk-mq.h |  1 +
 2 files changed, 34 insertions(+), 14 deletions(-)

-- 
2.18.1


             reply	other threads:[~2020-04-28 13:25 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-04-28 13:25 Weiping Zhang [this message]
2020-04-28 13:29 ` [RESEND PATCH v4 0/6] Fix potential kernel panic when increase hardware queue Weiping Zhang

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=cover.1588078594.git.zhangweiping@didiglobal.com \
    --to=zhangweiping@didiglobal.com \
    --cc=axboe@kernel.dk \
    --cc=bvanassche@acm.org \
    --cc=linux-block@vger.kernel.org \
    --cc=tom.leiming@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).