Re: [stable-4.14 00/23] block/scsi multiqueue performance enhancement and

Jens Axboe [off-list ref] =E4=BA=8E2018=E5=B9=B47=E6=9C=8823=E6=97=A5=E5=
=91=A8=E4=B8=80 =E4=B8=8B=E5=8D=885:31=E5=86=99=E9=81=93=EF=BC=9A

On 7/23/18 9:28 AM, Jack Wang wrote:

quoted

Jens Axboe [off-list ref] =E4=BA=8E2018=E5=B9=B47=E6=9C=8823=E6=97=

=A5=E5=91=A8=E4=B8=80 =E4=B8=8B=E5=8D=885:05=E5=86=99=E9=81=93=EF=BC=9A

quoted

quoted

On 7/23/18 9:00 AM, Jack Wang wrote:

quoted

Hi Greg,

Thanks for quick reply. Please see reply inline.



Greg KH [off-list ref] =E4=BA=8E2018=E5=B9=B47=E6=9C=88=

23=E6=97=A5=E5=91=A8=E4=B8=80 =E4=B8=8B=E5=8D=883:34=E5=86=99=E9=81=93=EF=
=BC=9A

quoted

quoted

quoted

quoted

On Mon, Jul 23, 2018 at 03:24:22PM +0200, Jack Wang wrote:

quoted

Hi Greg,

Please consider this patchset, which include block/scsi multiqueue =

performance

quoted

quoted

quoted

quoted

quoted

enhancement and bugfix.

What exactly is the performance enhancement?  How can you measure it=

?

quoted

quoted

quoted

quoted

How did you measure it?

I'm testing on SRP/IBNBD using fio, I've seen +10% with mix IO load,
and 50% improvement on small IO (bs=3D512.) with the patchset.

Big nak on backporting this huge series, it's a lot of core changes.
It's way beyond the scope of a stable fix backport.

--
Jens Axboe

OK, could you shed light on how could we fix the queue stall problem on=

 4.14?

quoted

My colleague Roman reported to upstream:
https://lkml.org/lkml/2017/10/18/263

It's still there on latest 4.14.

The proposed patch is a helluva lot simpler than doing a 23 patch selecti=

ve

backport.

--
Jens Axboe

Do you think it make sense to port Roman's patch to upstream 4.14?

So others would also be benifitial?
On the other side, stable tree only takes patches accepted into Linus tree.

Thanks,
Jack