Re: [PATCH v3 36/38] fs/resctrl: Add boiler plate for external resctrl code

[Reinette Chatre]

Hi James,

On 7/4/24 9:40 AM, James Morse wrote:
> Hi Reinette,
>
> On 28/06/2024 17:54, Reinette Chatre wrote:
> > On 6/14/24 8:00 AM, James Morse wrote:
> > > Add Makefile and Kconfig for fs/resctrl. Add ARCH_HAS_CPU_RESCTRL
> > > for the common parts of the resctrl interface and make X86_CPU_RESCTRL
> > > depend on this.
> > >
> > > Adding an include of asm/resctrl.h to linux/resctrl.h allows some
> > > of the files to switch over to using this header instead.
>
>
> > > diff --git a/fs/resctrl/Kconfig b/fs/resctrl/Kconfig
> > > new file mode 100644
> > > index 000000000000..a5fbda54d32f
> > > --- /dev/null
> > > +++ b/fs/resctrl/Kconfig
> > > @@ -0,0 +1,36 @@
> > > +config RESCTRL_FS
> > > +    bool "CPU Resource Control Filesystem (resctrl)"
> > > +    depends on ARCH_HAS_CPU_RESCTRL
> > > +    select KERNFS
> > > +    select PROC_CPU_RESCTRL if PROC_FS
> > > +    help
> > > +      Some architectures provide hardware facilities to group tasks and
> > > +      monitor and control their usage of memory system resources such as
> > > +      caches and memory bandwidth.  Examples of such facilities include
> > > +      Intel's Resource Director Technology (Intel(R) RDT) and AMD's
> > > +      Platform Quality of Service (AMD QoS).
> > > +
> > > +      If your system has the necessary support and you want to be able to
> > > +      assign tasks to groups and manipulate the associated resource
> > > +      monitors and controls from userspace, say Y here to get a mountable
> > > +      'resctrl' filesystem that lets you do just that.
> > > +
> > > +      If nothing mounts or prods the 'resctrl' filesystem, resource
> > > +      controls and monitors are left in a quiescent, permissive state.
> > > +
> > > +      If unsure, it is safe to say N.
> > > +
> >
> > Will user ever get opportunity to say "Y" or "N"?
> > It looks to me that
> > RESCTRL_FS will be "forced" on user as it is selected by the arch specific
> > config X86_CPU_RESCTRL and be invisble otherwise because of the dependency
> > on ARCH_HAS_CPU_RESCTRL.
>
> I did it like this so that this change is invisible for x86 config files on the principle
> of 'least noise'. Users can't enable RDT but disable resctrl today.
> It isn't actually possible to enable RDT and disable resctrl until after the code has been
> split from the architecture code.
>
> I have ended up supporting this for MPAM - you can enable the architecture's MPAM code and
> the driver, but not resctrl. This will eventually be for in-kernel users of resources that
> resctrl doesn't understand.
>
>
> > The text about when to select "Y" or "N" thus does
> > not look practical to me and it may be helpful to instead provide
> > information about when it is selected? I do not know the customs for this
> > text and if it is intended to document any future usages also.
>
> I think Dave wrote this text because its traditional for Kconfig options to say this.
>
> Describing when it is selected gets messy as this varies by architecture, and Kconfig can
> already tell you this:
> | Selected by [y]:
> │   - X86_CPU_RESCTRL [=y] && X86 [=y] && (CPU_SUP_INTEL [=y] || CPU_SUP_AMD [=y]) &&
> MISC_FILESYSTEMS [=y]

Right.

> I don't think it makes sense for resctrl to be enabled/disabled independently on x86.

I was not asking for resctrl to be enabled/disabled independently on x86. I commented on this
patch that adds text to guide user for options that the user is never able to select.

> If you want to support this, we need a few more IS_ENABLED() checks and stubs to make it

I did not intend to suggest this at all.

> build. The only reason I can see to do it is to ensure the architecture code is self
> contained.
>
> I'll reword this as "On architectures where this can be disabled independently, it is safe
> to say N".

ok

Reinette