mbox series

[0/2] Minor fixes

Message ID 20221126025550.967914-1-damien.lemoal@opensource.wdc.com (mailing list archive)
Headers show
Series Minor fixes | expand

Message

Damien Le Moal Nov. 26, 2022, 2:55 a.m. UTC
Jens,

A couple of patches to:
(1) fix recent mq-deadline improvement. Feeel free to squash that one
    into commit 015d02f48537 ("block: mq-deadline: Do not break
    sequential write streams to zoned HDDs") in your for-6.2/block
    branch.
(2) Fix REQ_SYNC use for async direct IOs to block device files.

Damien Le Moal (2):
  block: mq-deadline: Rename deadline_is_seq_writes()
  block: fops: Do not set REQ_SYNC for async IOs

 block/fops.c        | 7 ++++++-
 block/mq-deadline.c | 4 ++--
 2 files changed, 8 insertions(+), 3 deletions(-)

Comments

Jens Axboe Nov. 28, 2022, 2:15 p.m. UTC | #1
On Sat, 26 Nov 2022 11:55:48 +0900, Damien Le Moal wrote:
> Jens,
> 
> A couple of patches to:
> (1) fix recent mq-deadline improvement. Feeel free to squash that one
>     into commit 015d02f48537 ("block: mq-deadline: Do not break
>     sequential write streams to zoned HDDs") in your for-6.2/block
>     branch.
> (2) Fix REQ_SYNC use for async direct IOs to block device files.
> 
> [...]

Applied, thanks!

[1/2] block: mq-deadline: Rename deadline_is_seq_writes()
      commit: 4b49197f9fbd2dd4e0452df6effaba665beecb45
[2/2] block: fops: Do not set REQ_SYNC for async IOs
      commit: 26a7bc153a19f3349fd8c2e262728b2eae6bfd6f

Best regards,
Christoph Hellwig Nov. 28, 2022, 3:29 p.m. UTC | #2
On Mon, Nov 28, 2022 at 07:15:47AM -0700, Jens Axboe wrote:
> [2/2] block: fops: Do not set REQ_SYNC for async IOs
>       commit: 26a7bc153a19f3349fd8c2e262728b2eae6bfd6f

Urgg.  I don't think this is right.  And even if it was, block device
I/O is not out of line with the other direct I/O implementations.
Jens Axboe Nov. 28, 2022, 3:36 p.m. UTC | #3
On 11/28/22 8:29 AM, Christoph Hellwig wrote:
> On Mon, Nov 28, 2022 at 07:15:47AM -0700, Jens Axboe wrote:
>> [2/2] block: fops: Do not set REQ_SYNC for async IOs
>>       commit: 26a7bc153a19f3349fd8c2e262728b2eae6bfd6f
> 
> Urgg.  I don't think this is right.  And even if it was, block device
> I/O is not out of line with the other direct I/O implementations.

I'll drop it for now.