On Fri, Sep 05, 2014 at 10:31:49AM -0700, Linus Torvalds wrote:
> So quite frankly, the whole perf_pmu_migrate_context() thing looks
> completely and fundamentally broken.
>
> Your patch doesn't really solve anything in general, and would need to
> be extended to do that get_online_cpus() around e
On Fri, Sep 05, 2014 at 10:31:49AM -0700, Linus Torvalds wrote:
> So quite frankly, the whole perf_pmu_migrate_context() thing looks
> completely and fundamentally broken.
Yes, agreed. We can go play very nasty games, but fundamentally agreed.
> Or even just say: "if somebody takes down a CPU wit
On Fri, Sep 5, 2014 at 9:59 AM, Mark Rutland wrote:
>
> As you point out below, the race on event->ctx is the fundamental issue. That
> is what results in decrementing the refcount twice (once on a stale event->ctx
> pointer).
So quite frankly, the whole perf_pmu_migrate_context() thing looks
com
On Fri, Sep 05, 2014 at 04:41:43PM +0100, Linus Torvalds wrote:
> On Fri, Sep 5, 2014 at 8:16 AM, Peter Zijlstra wrote:
> >
> > How horrible is the below patch (performance wise). It does pretty much
> > the same thing except that percpu_rw_semaphore is a lot saner, its
> > read side performance s
On Fri, 5 Sep 2014, Linus Torvalds wrote:
> However, the more fundamental question is "what protects accesses to
> 'events->ctx'". Why is "put_event()" so special that *it* gets locking
> for the reading of "event->ctx", but none of the other cases of
> reading the ctx pointer gets it or needs it
On Fri, Sep 5, 2014 at 8:16 AM, Peter Zijlstra wrote:
>
> How horrible is the below patch (performance wise). It does pretty much
> the same thing except that percpu_rw_semaphore is a lot saner, its
> read side performance should be minimal in the absence of writes.
Ugh. Why do any locking at all
On Thu, Sep 04, 2014 at 12:07:40PM +0100, Mark Rutland wrote:
> Thanks for taking a look. If you have any ideas I'm happy to try another
> approach.
How horrible is the below patch (performance wise). It does pretty much
the same thing except that percpu_rw_semaphore is a lot saner, its
read side
On Thu, Sep 04, 2014 at 11:44:02AM +0100, Peter Zijlstra wrote:
> On Wed, Sep 03, 2014 at 12:50:14PM +0100, Mark Rutland wrote:
> > From 6465beace3ad9b12039127468f4596b8e87a53e8 Mon Sep 17 00:00:00 2001
> > From: Mark Rutland
> > Date: Wed, 3 Sep 2014 11:06:22 +0100
> > Subject: [PATCH] perf: prev
On Wed, Sep 03, 2014 at 12:50:14PM +0100, Mark Rutland wrote:
> From 6465beace3ad9b12039127468f4596b8e87a53e8 Mon Sep 17 00:00:00 2001
> From: Mark Rutland
> Date: Wed, 3 Sep 2014 11:06:22 +0100
> Subject: [PATCH] perf: prevent hotplug race on event->ctx
>
> The perf_pmu_migrate_context code intr
Hi all,
Further to my earlier reply I've come up with a potential fix below,
which has survived my stress test for both my WIP driver and the intel
uncore imc driver.
As it's impossible to synchronize with the event->ctx I'd hoped it would
be possible to synchronize with a field on the event itse
On Mon, Sep 01, 2014 at 08:05:34PM +0100, Peter Zijlstra wrote:
> On Mon, Sep 01, 2014 at 07:18:08PM +0100, Mark Rutland wrote:
> > Hi all,
>
> > [ 66.780759] [] rcu_process_callbacks+0x1e3/0x540
>
> > Has anything seen anything like this before? Is this a known issue?
>
> I've not seen it re
On Mon, Sep 01, 2014 at 07:18:08PM +0100, Mark Rutland wrote:
> Hi all,
> [ 66.780759] [] rcu_process_callbacks+0x1e3/0x540
> Has anything seen anything like this before? Is this a known issue?
I've not seen it reported.. sounds like 'fun' though.
--
To unsubscribe from this list: send the li
Hi all,
While trying some rework of the ARM CCI PMU driver on v3.17-rc2, I
encountered what seems to be a race between CPU hotplug and perf event
context migration, which results in a BUG in mm/slub.c.
It looks like this is a generic issue as I'm able to cause the same
splat with the uncore_imc d
13 matches
Mail list logo