Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1313233 > unrolled thread

Re: vmstat: make vmstat_updater deferrable again and shut down on idle

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  Re: vmstat: make vmstat_updater deferrable again and shut down on idle Shiraz Hashim <shiraz.linux.kernel@gmail.com> - 2016-01-20 15:00 +0100
    Re: vmstat: make vmstat_updater deferrable again and shut down on  idle Christoph Lameter <cl@linux.com> - 2016-01-20 16:20 +0100
      Re: vmstat: make vmstat_updater deferrable again and shut down on idle Shiraz Hashim <shiraz.linux.kernel@gmail.com> - 2016-01-21 07:30 +0100
        Re: vmstat: make vmstat_updater deferrable again and shut down on  idle Christoph Lameter <cl@linux.com> - 2016-01-21 16:50 +0100

#1313233 — Re: vmstat: make vmstat_updater deferrable again and shut down on idle

Hi Christoph,

On Fri, Dec 11, 2015 at 2:15 AM, Christoph Lameter <cl@linux.com> wrote:
> Currently the vmstat updater is not deferrable as a result of commit
> ba4877b9ca51f80b5d30f304a46762f0509e1635. This in turn can cause multiple
> interruptions of the applications because the vmstat updater may run at
> different times than tick processing. No good.
>
> Make vmstate_update deferrable again and provide a function that
> folds the differentials when the processor is going to idle mode thus
> addressing the issue of the above commit in a clean way.
>

The patch makes vmstat_shepherd deferable which if is quiesed
would not schedule vmstat update on other cpus. Wouldn't this
aggravate the problem of vmstat for rest cpus not gettng updated.

regards
Shiraz

[toc] | [next] | [standalone]

#1313289 — Re: vmstat: make vmstat_updater deferrable again and shut down on idle

On Wed, 20 Jan 2016, Shiraz Hashim wrote:

> The patch makes vmstat_shepherd deferable which if is quiesed
> would not schedule vmstat update on other cpus. Wouldn't this
> aggravate the problem of vmstat for rest cpus not gettng updated.

Its only "deferred" in order to make it at the next tick and not cause an
extra event. This means that vmstat will run periodically from tick
processing. It merely causes a synching so that we have one interruption
that does both.

On idle we fold counters immediately. So there is no loss of accuracy.

[toc] | [prev] | [next] | [standalone]

#1313941

On Wed, Jan 20, 2016 at 8:42 PM, Christoph Lameter <cl@linux.com> wrote:
> On Wed, 20 Jan 2016, Shiraz Hashim wrote:
>
>> The patch makes vmstat_shepherd deferable which if is quiesed
>> would not schedule vmstat update on other cpus. Wouldn't this
>> aggravate the problem of vmstat for rest cpus not gettng updated.
>
> Its only "deferred" in order to make it at the next tick and not cause an
> extra event. This means that vmstat will run periodically from tick
> processing. It merely causes a synching so that we have one interruption
> that does both.
>
> On idle we fold counters immediately. So there is no loss of accuracy.
>

vmstat is scheduled by shepherd or by itself (conditionally). In case shepherd
is deferred and vmstat doesn't schedule itself, then vmstat needs to wait
for shepherd to be up and then schedule it. This may end up in delayed status
update for all live cpus. Isn't it ?

-- 
regards
Shiraz Hashim

[toc] | [prev] | [next] | [standalone]

#1314288 — Re: vmstat: make vmstat_updater deferrable again and shut down on idle

On Thu, 21 Jan 2016, Shiraz Hashim wrote:

> > On idle we fold counters immediately. So there is no loss of accuracy.
>
> vmstat is scheduled by shepherd or by itself (conditionally). In case shepherd
> is deferred and vmstat doesn't schedule itself, then vmstat needs to wait
> for shepherd to be up and then schedule it. This may end up in delayed status
> update for all live cpus. Isn't it ?

The shepherd runs on a processor with an active tick and thus should do
its duty every 2 seconds as scheduled. Small milisecond range deferrals do
not matter much.

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web