Re: sched/fair: Hard lockup from idle_balance()/task_numa_migrate() race

2017-11-07 Thread Ed Swierk
On Mon, Oct 16, 2017 at 4:11 PM, Ed Swierk wrote: > To recap: a dual-socket Xeon (E5 v4) server system had been running a > bunch of KVM workloads just fine for over 6 weeks. Suddenly hard > lockups occurred on cpu 13 in task_numa_migrate(), and cpu 0 in > idle_balance(). That condition persisted

Re: sched/fair: Hard lockup from idle_balance()/task_numa_migrate() race

2017-11-06 Thread Peter Zijlstra
On Mon, Oct 16, 2017 at 04:11:39PM -0700, Ed Swierk wrote: > Ping for Peter, Ingo and other sched maintainers: > > I'd appreciate any feedback on this hard lockup issue, which occurred > on a system running kernel 4.4.52-grsec. So I'm currently somewhat tied up with things, and the other guy who

Re: sched/fair: Hard lockup from idle_balance()/task_numa_migrate() race

2017-11-03 Thread Ed Swierk
On Fri, Nov 3, 2017 at 8:03 AM, Linus Torvalds wrote: > I suspect that you'd get more responses if it was an upstream kernel > (and preferably newer), or even a distro one. > > If you run the grsec patches, you need to go to grsec to get support. > We don't even know what they are doing, since the

Re: sched/fair: Hard lockup from idle_balance()/task_numa_migrate() race

2017-11-03 Thread Linus Torvalds
I suspect that you'd get more responses if it was an upstream kernel (and preferably newer), or even a distro one. If you run the grsec patches, you need to go to grsec to get support. We don't even know what they are doing, since they're hiding their patches and not breaking them out.

Re: sched/fair: Hard lockup from idle_balance()/task_numa_migrate() race

2017-11-02 Thread Ed Swierk
Ping? On Wed, Oct 25, 2017 at 9:35 PM, Ed Swierk wrote: > > Ping? > > On Mon, Oct 16, 2017 at 4:11 PM, Ed Swierk wrote: > > > > Ping for Peter, Ingo and other sched maintainers: > > > > I'd appreciate any feedback on this hard lockup issue, which occurred > > on a system running kernel 4.4.5

Re: sched/fair: Hard lockup from idle_balance()/task_numa_migrate() race

2017-10-25 Thread Ed Swierk
Ping? On Mon, Oct 16, 2017 at 4:11 PM, Ed Swierk wrote: > > Ping for Peter, Ingo and other sched maintainers: > > I'd appreciate any feedback on this hard lockup issue, which occurred > on a system running kernel 4.4.52-grsec. > > To recap: a dual-socket Xeon (E5 v4) server system had been run

sched/fair: Hard lockup from idle_balance()/task_numa_migrate() race

2017-10-16 Thread Ed Swierk
Ping for Peter, Ingo and other sched maintainers: I'd appreciate any feedback on this hard lockup issue, which occurred on a system running kernel 4.4.52-grsec. To recap: a dual-socket Xeon (E5 v4) server system had been running a bunch of KVM workloads just fine for over 6 weeks. Suddenly hard l