cfq-iosched: a regression

cfq-iosched: a regression

[Shaohua Li]

We have a kernel build regression since 3.1-rc1, which is about 10%
regression. The kernel source is in an ext3 filesystem.
Alex Shi bisect it to commit:
commit a07405b7802691d29ab3b23bdc76ee6d006aad0b
Author: Justin TerAvest <teravest@google.com>
Date:   Sun Jul 10 22:09:19 2011 +0200

    cfq: Remove special treatment for metadata rqs.

Apparently this is caused by lack metadata preemption, where ext3/ext4
do use READ_META. I didn't see a way to fix the issue, so suggest
reverting the patch.

Reported-by: Alex Shi<alex.shi@intel.com>
Reported-by: Shaohua Li<shaohua.li@intel.com>

Re: cfq-iosched: a regression

[Justin TerAvest]

Hi Shaohua,

What was the workload you were running to find this regression?

On Mon, Aug 15, 2011 at 7:47 PM, Shaohua Li <shaohua.li@intel.com> wrote:
> We have a kernel build regression since 3.1-rc1, which is about 10%
> regression. The kernel source is in an ext3 filesystem.
> Alex Shi bisect it to commit:
> commit a07405b7802691d29ab3b23bdc76ee6d006aad0b
> Author: Justin TerAvest <teravest@google.com>
> Date:   Sun Jul 10 22:09:19 2011 +0200
>
>    cfq: Remove special treatment for metadata rqs.
>
> Apparently this is caused by lack metadata preemption, where ext3/ext4
> do use READ_META. I didn't see a way to fix the issue, so suggest
> reverting the patch.
>
> Reported-by: Alex Shi<alex.shi@intel.com>
> Reported-by: Shaohua Li<shaohua.li@intel.com>
>
>

Re: cfq-iosched: a regression

[Shaohua Li]

On Fri, 2011-08-19 at 04:22 +0800, Justin TerAvest wrote:
> Hi Shaohua,
> 
> What was the workload you were running to find this regression?
kernel build as I declared below.

Thanks,
Shaohua

> On Mon, Aug 15, 2011 at 7:47 PM, Shaohua Li <shaohua.li@intel.com> wrote:
> > We have a kernel build regression since 3.1-rc1, which is about 10%
> > regression. The kernel source is in an ext3 filesystem.
> > Alex Shi bisect it to commit:
> > commit a07405b7802691d29ab3b23bdc76ee6d006aad0b
> > Author: Justin TerAvest <teravest@google.com>
> > Date:   Sun Jul 10 22:09:19 2011 +0200
> >
> >    cfq: Remove special treatment for metadata rqs.
> >
> > Apparently this is caused by lack metadata preemption, where ext3/ext4
> > do use READ_META. I didn't see a way to fix the issue, so suggest
> > reverting the patch.
> >
> > Reported-by: Alex Shi<alex.shi@intel.com>
> > Reported-by: Shaohua Li<shaohua.li@intel.com>
> >
> >

Re: cfq-iosched: a regression

[Jens Axboe]

On 2011-08-16 04:47, Shaohua Li wrote:
> We have a kernel build regression since 3.1-rc1, which is about 10%
> regression. The kernel source is in an ext3 filesystem.
> Alex Shi bisect it to commit:
> commit a07405b7802691d29ab3b23bdc76ee6d006aad0b
> Author: Justin TerAvest <teravest@google.com>
> Date:   Sun Jul 10 22:09:19 2011 +0200
> 
>     cfq: Remove special treatment for metadata rqs.
> 
> Apparently this is caused by lack metadata preemption, where ext3/ext4
> do use READ_META. I didn't see a way to fix the issue, so suggest
> reverting the patch.
> 
> Reported-by: Alex Shi<alex.shi@intel.com>
> Reported-by: Shaohua Li<shaohua.li@intel.com>

I had a gut feeling on that meta data, it's hard to know exactly where
it helps at this point. I'll revert the commit and then we can diagnose
this on the side.

-- 
Jens Axboe