NO_HZ_FULL not yet working on 3.14 kernel

Bug #1321948 reported by Gary S. Robertson
10
This bug affects 2 people
Affects Status Importance Assigned to Milestone
linaro-networking
Fix Released
Medium
viresh kumar

Bug Description

The 3.14 kernel still needs to be patched with the latest NO_HZ_FULL and CPU isolation patches.

Revision history for this message
Mike Holmes (mike-holmes) wrote :

In migrating to the 3.14 kernel the hotpulg patches developed on 3.10 have been dropped and we will pull in Vireshes changes that he is pushing upstream.

Changed in linaro-networking:
status: New → Confirmed
importance: Undecided → Medium
assignee: nobody → viresh kumar (viresh.kumar)
Revision history for this message
Mike Holmes (mike-holmes) wrote :

Viresh, when do you think you may have some patches for 3.14 ?

Revision history for this message
Santosh Shukla (santosh-shukla) wrote :

Gary,

Few question,

> The 3.14 kernel still needs to be patched with the latest NO_HZ_FULL and CPU isolation patches.

I assume, NO_HZ_FULL patches are in 3.14 then what other patch we're expecting here, Are we maintaining viresh's proposed under review in lng kernel?

I ran some test on x86 with little modification on viresh script for x86, running 3.15 kernel, could see secondary cpu 1 to 15 isolated but not fully, it ticks every 1 sec, are you expecting this behavior or complete isolation?

Revision history for this message
Mike Holmes (mike-holmes) wrote : Re: [Bug 1321948] Re: NO_HZ_FULL not yet working on 3.14 kernel

On https://validation.linaro.org/dashboard/image-charts/LNG-NO_HZ the X86 (
3.10.37-rt38) has been seeing complete isolation of 5000 seconds.
So maybe the test is broken in CI or there is some new difference, CI
should move to 3.14 today.

On 30 May 2014 02:35, Santosh Shukla <email address hidden> wrote:

> Gary,
>
> Few question,
>
> > The 3.14 kernel still needs to be patched with the latest NO_HZ_FULL
> and CPU isolation patches.
>
> I assume, NO_HZ_FULL patches are in 3.14 then what other patch we're
> expecting here, Are we maintaining viresh's proposed under review in lng
> kernel?
>
> I ran some test on x86 with little modification on viresh script for
> x86, running 3.15 kernel, could see secondary cpu 1 to 15 isolated but
> not fully, it ticks every 1 sec, are you expecting this behavior or
> complete isolation?
>
> --
> You received this bug notification because you are a member of Linaro
> Networking Group, which is subscribed to linaro-networking.
> Matching subscriptions: LNG all, all issues
> https://bugs.launchpad.net/bugs/1321948
>
> Title:
> NO_HZ_FULL not yet working on 3.14 kernel
>
> Status in Linaro networking Group:
> Confirmed
>
> Bug description:
> The 3.14 kernel still needs to be patched with the latest NO_HZ_FULL
> and CPU isolation patches.
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/linaro-networking/+bug/1321948/+subscriptions
>

--
*Mike Holmes*
Linaro Technical Manager / Lead
LNG - ODP

Revision history for this message
Gary S. Robertson (gary-robertson) wrote :

NO_HZ patches have not yet been integrated into 3.14... expect changes
there shortly. This time we decided to upgrade to the newer kernel version
even with some feature patches still missing - since we missed 3.12 because
we were still integrating all the feature patches when 3.14 RT patches
became available. So a number of features which are/were working in 3.10
are still being integrated into 3.14.

On Fri, May 30, 2014 at 5:17 AM, Mike Holmes <email address hidden> wrote:

> On https://validation.linaro.org/dashboard/image-charts/LNG-NO_HZ the
> X86 (
> 3.10.37-rt38) has been seeing complete isolation of 5000 seconds.
> So maybe the test is broken in CI or there is some new difference, CI
> should move to 3.14 today.
>
>
> On 30 May 2014 02:35, Santosh Shukla <email address hidden> wrote:
>
> > Gary,
> >
> > Few question,
> >
> > > The 3.14 kernel still needs to be patched with the latest NO_HZ_FULL
> > and CPU isolation patches.
> >
> > I assume, NO_HZ_FULL patches are in 3.14 then what other patch we're
> > expecting here, Are we maintaining viresh's proposed under review in lng
> > kernel?
> >
> > I ran some test on x86 with little modification on viresh script for
> > x86, running 3.15 kernel, could see secondary cpu 1 to 15 isolated but
> > not fully, it ticks every 1 sec, are you expecting this behavior or
> > complete isolation?
> >
> > --
> > You received this bug notification because you are a member of Linaro
> > Networking Group, which is subscribed to linaro-networking.
> > Matching subscriptions: LNG all, all issues
> > https://bugs.launchpad.net/bugs/1321948
> >
> > Title:
> > NO_HZ_FULL not yet working on 3.14 kernel
> >
> > Status in Linaro networking Group:
> > Confirmed
> >
> > Bug description:
> > The 3.14 kernel still needs to be patched with the latest NO_HZ_FULL
> > and CPU isolation patches.
> >
> > To manage notifications about this bug go to:
> > https://bugs.launchpad.net/linaro-networking/+bug/1321948/+subscriptions
> >
>
>
> --
> *Mike Holmes*
> Linaro Technical Manager / Lead
> LNG - ODP
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> Matching subscriptions: LNG all
> https://bugs.launchpad.net/bugs/1321948
>
> Title:
> NO_HZ_FULL not yet working on 3.14 kernel
>
> Status in Linaro networking Group:
> Confirmed
>
> Bug description:
> The 3.14 kernel still needs to be patched with the latest NO_HZ_FULL
> and CPU isolation patches.
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/linaro-networking/+bug/1321948/+subscriptions
>

Revision history for this message
viresh kumar (viresh.kumar) wrote :

#2: I am back today from my holidays and have lots of stuff to finish for the upstream stuff.. Give me sometime to test this again and see how to get things fixed for LNG kernel..

#3: Santosh, yes they were looking to pick my patches under review. I just need to pass only the relevant patches. You are getting 1 second tick because you might not have Kevin's patch that I told you about, which makes it possible to make it tickless.

#4: don't think CI is broken :)

Revision history for this message
Mike Holmes (mike-holmes) wrote :

Viresh has posted a patch set, Anders applied them, it remains to fix merge conflicts with RT given the current LNG git structure.

Revision history for this message
Mike Holmes (mike-holmes) wrote :

waiting on confirmation from CI before closing.

Revision history for this message
Gary S. Robertson (gary-robertson) wrote :

The latest CI jobs show NO_HZ_FULL working on 3.14 in LE and BE for the 'non-RT' branches.
See this LAVA test run log: https://validation.linaro.org/scheduler/job/136861/log_file#L_23_118
NO_HZ_FULL still does not work in conjunction with PREEMPT_RT.

Changed in linaro-networking:
status: Confirmed → Fix Released
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.