Re: [PATCH] x86/resctrl: avoid compiler optimization in __resctrl_sched_in

[Stephane Eranian]

On Tue, Mar 7, 2023 at 1:06 PM Linus Torvalds
<torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:
>
> On Tue, Mar 7, 2023 at 12:54 PM Linus Torvalds
> <torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:
> >
> > I think the problem is that the <asm/resctrl.h> code is disgusting and
> > horrible in multiple ways:
> >
> >  (a) it shouldn't define and declare a static function in a header file
> >
> >  (b) the resctrl_sched_in() inline function is misdesigned to begin with
>
> Ok, so here's a *ttoally* untested and mindless patch to maybe fix
> what I dislike about that resctl code.
>
> Does it fix the code generation issue? I have no idea. But this is
> what I would suggest is the right answer, without actually knowing the
> code any better, and just going on a mindless rampage.
>
> It seems to compile for me, fwiw.
>

On Tue, Mar 7, 2023 at 3:01 PM Nick Desaulniers <ndesaulniers@xxxxxxxxxx> wrote:
>
> On Tue, Mar 7, 2023 at 2:56 PM Linus Torvalds
> <torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:
> >
> > On Tue, Mar 7, 2023 at 2:03 PM Nick Desaulniers <ndesaulniers@xxxxxxxxxx> wrote:
> > >
> > > Sounds like Stephane is going to re-run the internal tests he used to
> > > discover the issue with your diff applied, if you don't mind holding
> > > out for another Tested-by tag. EOM
> >
> > Ack. I am in no hurry.
> >
> > In fact, I'd prefer to just get the patch sent back to me with a
> > commit message too, if somebody has the energy. I don't need the
> > credit for a trivial thing like that.
>
> Sure, then maybe Stephane you can supply a v2 with updated commit message and a
>
> Suggested-by: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
>
I verified Linus' patch on my test case on AMD Zen3 and it works as
 expected, i.e., the limit is enforced. I had tried a similar approach myself
as well and it worked.

I think passing the task pointer is the proper approach because we are
in a sched_in routine
and I would expect the task scheduled in to be passed as argument
instead of having the function
retrieve it from the current pointer.

Thanks.

Tested-by: Stephane Eranian <eranian@xxxxxxxxxx>

>              Linus