Re: [PATCH 0/3] block: fix mismatch of figuring out physical segment number

[Ming Lei <tom.leiming@gmail.com>]

On Tue, Jul 31, 2018 at 6:49 PM, Ming Lei <ming.lei@redhat.com> wrote:
> Hi,
>
> This patchset fixes one issue related with physical segment computation,
> which is found by Mike. In case of dm-rq, the warning of 'blk_cloned_rq_check_limits:
> over max segments limit' can be triggered easily.
>
> Follows the cause:
>
> 1) in IO fast path(blk_queue_split()), we always figure out physical segment number
> no matter the flag of QUEUE_FLAG_NO_SG_MERGE is set or not.
>
> 2) only blk_recount_segments() and blk_recalc_rq_segments() uses the flag of
> QUEUE_FLAG_NO_SG_MERGE, but the two are only called in some unusual
> cases, such as request clone in dm-rq.
>
> 3) the above two computation don't match, and cause the warning of
> "blk_cloned_rq_check_limits: over max segments limit".
>
> This patchset fixes this issue by killing the queue flag since it is basically
> bypassed since v4.4, and no one complains that at all. Also multipage bvec will
> come soon, and it doesn't make sense to keep QUEUE_FLAG_NO_SG_MERGE any more.
>
>
> Ming Lei (3):
>   block: don't use bio->bi_vcnt to figure out segment number
>   block: kill QUEUE_FLAG_NO_SG_MERGE
>   block: kill BLK_MQ_F_SG_MERGE
>
>  block/blk-merge.c            | 39 +++++++--------------------------------
>  block/blk-mq-debugfs.c       |  2 --
>  block/blk-mq.c               |  3 ---
>  drivers/block/loop.c         |  2 +-
>  drivers/block/nbd.c          |  2 +-
>  drivers/block/rbd.c          |  2 +-
>  drivers/block/skd_main.c     |  1 -
>  drivers/block/xen-blkfront.c |  2 +-
>  drivers/md/dm-rq.c           |  2 +-
>  drivers/md/dm-table.c        | 13 -------------
>  drivers/mmc/core/queue.c     |  3 +--
>  drivers/scsi/scsi_lib.c      |  2 +-
>  include/linux/blk-mq.h       |  1 -
>  include/linux/blkdev.h       |  1 -
>  14 files changed, 14 insertions(+), 61 deletions(-)
>
> Cc: Christoph Hellwig <hch@lst.de>
> Cc: Mike Snitzer <snitzer@redhat.com>
> Cc: Kent Overstreet <kent.overstreet@gmail.com>

Hello Guys,

Ping...


Thanks,
Ming Lei