RFC: less special casing for flush requests

RFC: less special casing for flush requests

[Christoph Hellwig]

Hi all,

inspired by Barts quest for less reording during requeues, this series
aims to limit the scope of requeues.  This is mostly done by simply
sending down more than we currently do down the normal submission path
with the help of Bart's patch to allow requests that are in the flush
state machine to still be seen by the I/O schedulers.

Diffstat:
 block/blk-flush.c      |  115 +++++++++++++++++++++++++++----------------------
 block/blk-mq-debugfs.c |    1 
 block/blk-mq-sched.c   |   14 +++++
 block/blk-mq-sched.h   |    1 
 block/blk-mq.c         |   68 ++++++++++------------------
 block/blk-mq.h         |    5 --
 block/blk.h            |    2 
 include/linux/blk-mq.h |   31 +++++--------
 include/linux/blkdev.h |    1 
 9 files changed, 119 insertions(+), 119 deletions(-)

RFC: less special casing for flush requests

[Christoph Hellwig]

Hi all,

inspired by Barts quest for less reording during requeues, this series
aims to limit the scope of requeues.  This is mostly done by simply
sending down more than we currently do down the normal submission path
with the help of Bart's patch to allow requests that are in the flush
state machine to still be seen by the I/O schedulers.

Diffstat:
 block/blk-flush.c      |  115 +++++++++++++++++++++++++++----------------------
 block/blk-mq-debugfs.c |    1 
 block/blk-mq-sched.c   |   14 +++++
 block/blk-mq-sched.h   |    1 
 block/blk-mq.c         |   68 ++++++++++------------------
 block/blk-mq.h         |    5 --
 block/blk.h            |    2 
 include/linux/blk-mq.h |   31 +++++--------
 include/linux/blkdev.h |    1 
 9 files changed, 119 insertions(+), 119 deletions(-)

Re: RFC: less special casing for flush requests

[Bart Van Assche]

On 4/16/23 13:09, Christoph Hellwig wrote:
> inspired by Barts quest for less reording during requeues, this series
> aims to limit the scope of requeues.  This is mostly done by simply
> sending down more than we currently do down the normal submission path
> with the help of Bart's patch to allow requests that are in the flush
> state machine to still be seen by the I/O schedulers.

Hi Christoph,

Do you have the time to continue the work on this patch series? Please 
let me know in case you would prefer that I continue the work on this 
patch series. I just found and fixed a bug in my patch "block: Send FUA 
requests to the I/O scheduler" (not included in this series).

Thanks,

Bart.

Re: RFC: less special casing for flush requests

[Christoph Hellwig]

On Mon, May 15, 2023 at 05:08:27PM -0700, Bart Van Assche wrote:
> Do you have the time to continue the work on this patch series? Please let 
> me know in case you would prefer that I continue the work on this patch 
> series. I just found and fixed a bug in my patch "block: Send FUA requests 
> to the I/O scheduler" (not included in this series).

I did another pass last weekend and was planning to finish it off today.

Re: RFC: less special casing for flush requests

[Christoph Hellwig]

On Tue, May 16, 2023 at 06:03:51AM +0200, Christoph Hellwig wrote:
> On Mon, May 15, 2023 at 05:08:27PM -0700, Bart Van Assche wrote:
> > Do you have the time to continue the work on this patch series? Please let 
> > me know in case you would prefer that I continue the work on this patch 
> > series. I just found and fixed a bug in my patch "block: Send FUA requests 
> > to the I/O scheduler" (not included in this series).
> 
> I did another pass last weekend and was planning to finish it off today.

Turns out this all interacts with the scheduler tags vs passthrough
dicussion we had, but I now have a version of the series rebased on top
of that:

    http://git.infradead.org/users/hch/block.git/shortlog/refs/heads/blk-flush

I will send it out once we've settled the tags discussion.