ARM perf events spin locks

ARM perf events spin locks

[Jamie Iles]

Hi Will,

I think we should convert the spinlocks in the ARM perf events code to raw 
spinlocks for realtime. Should we wait for your split set to get merged first 
before doing this?

Jamie

ARM perf events spin locks

[Will Deacon]

> Hi Will,

Hello,

> I think we should convert the spinlocks in the ARM perf events code to raw
> spinlocks for realtime. Should we wait for your split set to get merged first
> before doing this?

Since this is a logically separate change, I think we're better off waiting
until the split stuff has been merged. As for the raw spinlocks, by realtime
do you mean PREEMPT_RT? Also, do we actually *need* raw spinlocks in the perf
code?

Will

ARM perf events spin locks

[Jamie Iles]

On Tue, Nov 30, 2010 at 10:41:04AM -0000, Will Deacon wrote:
> > Hi Will,
> 
> Hello,
> 
> > I think we should convert the spinlocks in the ARM perf events code to raw
> > spinlocks for realtime. Should we wait for your split set to get merged first
> > before doing this?
> 
> Since this is a logically separate change, I think we're better off waiting
> until the split stuff has been merged. As for the raw spinlocks, by realtime
> do you mean PREEMPT_RT? Also, do we actually *need* raw spinlocks in the perf
> code?
Yes, I meant PREEMPT_RT. It won't stop working without raw spinlocks but I'm 
not convinced that we couldn't lose too much accuracy with normal spinlocks. I 
am however willing to be convinced otherwise!

Jamie

ARM perf events spin locks

[Will Deacon]

Jamie,

> On Tue, Nov 30, 2010 at 10:41:04AM -0000, Will Deacon wrote:
> > > Hi Will,
> >
> > Hello,
> >
> > > I think we should convert the spinlocks in the ARM perf events code to raw
> > > spinlocks for realtime. Should we wait for your split set to get merged first
> > > before doing this?
> >
> > Since this is a logically separate change, I think we're better off waiting
> > until the split stuff has been merged. As for the raw spinlocks, by realtime
> > do you mean PREEMPT_RT? Also, do we actually *need* raw spinlocks in the perf
> > code?
> Yes, I meant PREEMPT_RT. It won't stop working without raw spinlocks but I'm
> not convinced that we couldn't lose too much accuracy with normal spinlocks. I
> am however willing to be convinced otherwise!

Well struct perf_event_ctx has a lock field which is of type raw_spinlock_t.
I *think* this is always held by the core perf code before calling the backend,
however IRQs may still be enabled so we probably do need to change our pmu_lock.

Is that a sane analysis?

Will

ARM perf events spin locks

[Jamie Iles]

On Tue, Nov 30, 2010 at 01:49:32PM -0000, Will Deacon wrote:
> Jamie,
> 
> > On Tue, Nov 30, 2010 at 10:41:04AM -0000, Will Deacon wrote:
> > > > Hi Will,
> > >
> > > Hello,
> > >
> > > > I think we should convert the spinlocks in the ARM perf events code to raw
> > > > spinlocks for realtime. Should we wait for your split set to get merged first
> > > > before doing this?
> > >
> > > Since this is a logically separate change, I think we're better off waiting
> > > until the split stuff has been merged. As for the raw spinlocks, by realtime
> > > do you mean PREEMPT_RT? Also, do we actually *need* raw spinlocks in the perf
> > > code?
> > Yes, I meant PREEMPT_RT. It won't stop working without raw spinlocks but I'm
> > not convinced that we couldn't lose too much accuracy with normal spinlocks. I
> > am however willing to be convinced otherwise!
> 
> Well struct perf_event_ctx has a lock field which is of type raw_spinlock_t.
> I *think* this is always held by the core perf code before calling the backend,
> however IRQs may still be enabled so we probably do need to change our pmu_lock.
> 
> Is that a sane analysis?
That sounds perfectly reasonable to me! We don't hold these locks for any 
potentially long periods of time so it's probably better to be safe! Do you 
want me to make this change after perf-split is merged or is that something 
you're happy to do?

Jamie

ARM perf events spin locks

[Will Deacon]

> > Well struct perf_event_ctx has a lock field which is of type raw_spinlock_t.
> > I *think* this is always held by the core perf code before calling the backend,
> > however IRQs may still be enabled so we probably do need to change our pmu_lock.
> >
> > Is that a sane analysis?
> That sounds perfectly reasonable to me! We don't hold these locks for any
> potentially long periods of time so it's probably better to be safe! Do you
> want me to make this change after perf-split is merged or is that something
> you're happy to do?

I don't mind doing it as a separate patch after Russell has pulled my
perf-split branch. I don't think it should form part of the pull though
because it's unrelated.

Will

ARM perf events spin locks

[Russell King - ARM Linux]

On Tue, Nov 30, 2010 at 01:59:10PM -0000, Will Deacon wrote:
> > > Well struct perf_event_ctx has a lock field which is of type raw_spinlock_t.
> > > I *think* this is always held by the core perf code before calling the backend,
> > > however IRQs may still be enabled so we probably do need to change our pmu_lock.
> > >
> > > Is that a sane analysis?
> > That sounds perfectly reasonable to me! We don't hold these locks for any
> > potentially long periods of time so it's probably better to be safe! Do you
> > want me to make this change after perf-split is merged or is that something
> > you're happy to do?
> 
> I don't mind doing it as a separate patch after Russell has pulled my
> perf-split branch.

You mean this?

commit f1690d17d2c7afa2a2079e3c91eb2bca8c1c5ecd
Merge: 612275a 43eab87
Author: Russell King <rmk+kernel@arm.linux.org.uk>
Date:   Fri Nov 26 10:26:43 2010 +0000

    Merge branch 'perf-split' of git://linux-arm.org/linux-2.6-wd into devel-stable

Sorry, I didn't push the tree out for a week or so because I was busy with
other things (like the pgtable stuff.)

ARM perf events spin locks

[Will Deacon]

> > I don't mind doing it as a separate patch after Russell has pulled my
> > perf-split branch.
> 
> You mean this?
> 
> commit f1690d17d2c7afa2a2079e3c91eb2bca8c1c5ecd
> Merge: 612275a 43eab87
> Author: Russell King <rmk+kernel@arm.linux.org.uk>
> Date:   Fri Nov 26 10:26:43 2010 +0000
> 
>     Merge branch 'perf-split' of git://linux-arm.org/linux-2.6-wd into devel-stable

Aha, brilliant - thanks.

> Sorry, I didn't push the tree out for a week or so because I was busy with
> other things (like the pgtable stuff.)

There's no rest for the weary! I'll submit the patch to fix perf for sparse
to the patch system (I didn't realise you'd pulled my changes on Friday) and
tackle this spinlock thing at the same time.

Cheers,

Will