WBT for SQ devices interface

WBT for SQ devices interface

[Jan Kara]

Hi,

when testing my fix for 0-day reports with writeback throttling I came
across somewhat unexpected behavior with user interface of writeback
throttling. So currently if CFQ is used as an IO scheduler, we disable
writeback throttling because they don't go well together. However when user
has CONFIG_BLK_WBT_SQ=y and switches IO scheduler to NOOP or DEADLINE
writeback throttling still stays disabled. This is somewhat unexpected
especially because the switching of IO scheduler from CFQ to something else
can have happened behind user's back by some udev rule or so. So basically
CONFIG_DEFAULT_CFQ=y and CONFIG_BLK_WBT_SQ=y don't make sense together.

So do people thing we should enable WBT if CONFIG_BLK_WBT_SQ=y and IO
scheduler is switched from CFQ to something else?

								Honza
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR

Re: WBT for SQ devices interface

[Holger Hoffstätte]

On Tue, 11 Apr 2017 11:38:27 +0200, Jan Kara wrote:

> when testing my fix for 0-day reports with writeback throttling I came
> across somewhat unexpected behavior with user interface of writeback
> throttling. So currently if CFQ is used as an IO scheduler, we disable
> writeback throttling because they don't go well together. However when user
> has CONFIG_BLK_WBT_SQ=y and switches IO scheduler to NOOP or DEADLINE
> writeback throttling still stays disabled. This is somewhat unexpected
> especially because the switching of IO scheduler from CFQ to something else
> can have happened behind user's back by some udev rule or so. So basically
> CONFIG_DEFAULT_CFQ=y and CONFIG_BLK_WBT_SQ=y don't make sense together.

:(

> So do people thing we should enable WBT if CONFIG_BLK_WBT_SQ=y and IO
> scheduler is switched from CFQ to something else?

Yes please. If a scheduler doesn't like WBT then that is its own problem.
Deadline with WBT has worked fine for me.

-h

Re: WBT for SQ devices interface

[Jens Axboe]

On Apr 11, 2017, at 3:38 AM, Jan Kara <jack@suse.cz> wrote:
>=20
> Hi,
>=20
> when testing my fix for 0-day reports with writeback throttling I came
> across somewhat unexpected behavior with user interface of writeback
> throttling. So currently if CFQ is used as an IO scheduler, we disable
> writeback throttling because they don't go well together. However when use=
r
> has CONFIG_BLK_WBT_SQ=3Dy and switches IO scheduler to NOOP or DEADLINE
> writeback throttling still stays disabled. This is somewhat unexpected
> especially because the switching of IO scheduler from CFQ to something els=
e
> can have happened behind user's back by some udev rule or so. So basically=

> CONFIG_DEFAULT_CFQ=3Dy and CONFIG_BLK_WBT_SQ=3Dy don't make sense together=
.
>=20
> So do people thing we should enable WBT if CONFIG_BLK_WBT_SQ=3Dy and IO
> scheduler is switched from CFQ to something else?

Yes, I think so. We should have consistent behavior in that area, regardless=
 of the starting point.

--=20
Jens Axboe