2019-03-14 13:13:38

by Peter Zijlstra

[permalink] [raw]
Subject: [RFC][PATCH 0/8] perf/x86: event scheduling cleanups

These here patches are the result of chasing a bug in the TSX Force Abort
patches.

The first patch is simple and fixes the issue and should be marked /urgent and
backported.

The rest is the result of having a hard look at how all that event scheduling
stuff worked while trying to figure out wth was going wrong.

Stephane, since you were involved with writing much of that code, can you
please have a good look? I _think_ I got it, but... :-)



2019-03-15 07:16:05

by Stephane Eranian

[permalink] [raw]
Subject: Re: [RFC][PATCH 0/8] perf/x86: event scheduling cleanups

On Thu, Mar 14, 2019 at 6:11 AM Peter Zijlstra <[email protected]> wrote:
>
> These here patches are the result of chasing a bug in the TSX Force Abort
> patches.
>
> The first patch is simple and fixes the issue and should be marked /urgent and
> backported.
>
> The rest is the result of having a hard look at how all that event scheduling
> stuff worked while trying to figure out wth was going wrong.
>
> Stephane, since you were involved with writing much of that code, can you
> please have a good look? I _think_ I got it, but... :-)
>
Will do today.
But you did not send me PATCH 1/8 apparently.

2019-03-15 07:17:20

by Stephane Eranian

[permalink] [raw]
Subject: Re: [RFC][PATCH 0/8] perf/x86: event scheduling cleanups

On Fri, Mar 15, 2019 at 12:15 AM Stephane Eranian <[email protected]> wrote:
>
> On Thu, Mar 14, 2019 at 6:11 AM Peter Zijlstra <[email protected]> wrote:
> >
> > These here patches are the result of chasing a bug in the TSX Force Abort
> > patches.
> >
> > The first patch is simple and fixes the issue and should be marked /urgent and
> > backported.
> >
> > The rest is the result of having a hard look at how all that event scheduling
> > stuff worked while trying to figure out wth was going wrong.
> >
> > Stephane, since you were involved with writing much of that code, can you
> > please have a good look? I _think_ I got it, but... :-)
> >
> Will do today.
> But you did not send me PATCH 1/8 apparently.

Correction : I got it. Mail client had just moved it somewhere else!

2019-03-15 08:02:16

by Peter Zijlstra

[permalink] [raw]
Subject: Re: [RFC][PATCH 0/8] perf/x86: event scheduling cleanups

On Fri, Mar 15, 2019 at 12:15:55AM -0700, Stephane Eranian wrote:
> On Fri, Mar 15, 2019 at 12:15 AM Stephane Eranian <[email protected]> wrote:
> >
> > On Thu, Mar 14, 2019 at 6:11 AM Peter Zijlstra <[email protected]> wrote:
> > >
> > > These here patches are the result of chasing a bug in the TSX Force Abort
> > > patches.
> > >
> > > The first patch is simple and fixes the issue and should be marked /urgent and
> > > backported.
> > >
> > > The rest is the result of having a hard look at how all that event scheduling
> > > stuff worked while trying to figure out wth was going wrong.
> > >
> > > Stephane, since you were involved with writing much of that code, can you
> > > please have a good look? I _think_ I got it, but... :-)
> > >
> > Will do today.
> > But you did not send me PATCH 1/8 apparently.
>
> Correction : I got it. Mail client had just moved it somewhere else!

Also please look at all 8, I'm reasonably sure 1/8 is correct, but it is
rather more ugly than I'd wish. So the remaining 7 clean it all up (and
effectively revert 1 again) but are much more invasive.