Re: [PATCH 3/7] x86/resctrl: Add a new node-scoped resource to rdt_resources_all[]

[Peter Newman]

Hi Tony,

On Fri, Jan 27, 2023 at 6:25 AM Yu, Fenghua <fenghua.yu@xxxxxxxxx> wrote:
>
> On 1/26/23 10:41, Tony Luck wrote:
> > diff --git a/arch/x86/kernel/cpu/resctrl/core.c b/arch/x86/kernel/cpu/resctrl/core.c
> > index 6914232acf84..19be6fe42ef3 100644
> > --- a/arch/x86/kernel/cpu/resctrl/core.c
> > +++ b/arch/x86/kernel/cpu/resctrl/core.c
> > @@ -100,6 +100,16 @@ struct rdt_hw_resource rdt_resources_all[] = {
> >                       .fflags                 = RFTYPE_RES_MB,
> >               },
> >       },
> > +     [RDT_RESOURCE_NODE] =
> > +     {
> > +             .r_resctrl = {
> > +                     .rid                    = RDT_RESOURCE_NODE,
> > +                     .name                   = "L3",
> "L3" was named as RDT_RESOURCE_L3 already. The duplicate name here may
> cause duplicate file names in info dir. Maybe rename it as "L3_NODE"?

I'm trying to get some feedback from our own users on whether changing
the directory names would bother them. At least from my own testing, I
did learn to appreciate the interface change a bit more: I needed an
SNC and non-SNC case to correctly predict which mon_data subdirectory
the data would appear in.

I was able to confirm that this change allows bandwidth to be counted
on RMID/CPU combos where it didn't work before on an SNC2
configuration.

If I'm understanding this correctly, it might be helpful to highlight
that the extra resource is needed to allow a different number of L3
domains in L3 monitoring vs allocation.

Thanks!
-Peter

Tested-By: Peter Newman <peternewman@xxxxxxxxxx>
Reviewed-By: Peter Newman <peternewman@xxxxxxxxxx>