Re: [PATCH 1/5] arm64: cpufeature: Add scope for capability check

From: Will Deacon
Date: Fri Apr 15 2016 - 08:55:20 EST


On Thu, Apr 14, 2016 at 06:47:46PM +0100, Suzuki K Poulose wrote:
> On 14/04/16 18:38, Will Deacon wrote:
> >Hi Suzuki,
> >
> >On Wed, Apr 06, 2016 at 12:24:10PM +0100, Suzuki K Poulose wrote:
> >>Add scope parameter to the arm64_cpu_capabilities::matches(),
> >>so that this can be reused for checking the capability on a
> >>given CPU vs the system wide. By default, the system uses
> >>'system' wide values for setting the CPU_HWCAPs and ELF_HWCAPs.
>
> >> static bool __maybe_unused
> >>-is_affected_midr_range(const struct arm64_cpu_capabilities *entry)
> >>+is_affected_midr_range(const struct arm64_cpu_capabilities *entry, int __unused)
> >
> >Maybe it would be better to WARN if somebody passes SCOPE_SYSTEM, rather
> >than silently treat it as per-cpu?
>
> Should we worry about errata's which may not necessarily depend on per CPU or
> a local capability (GIC) ?

Why would they be calling is_affected_midr_range?

> If not, we could add a WARN after passing down LOCAL
> scope for errata.

But if we don't care about errata that aren't local, then why would we warn
on LOCAL?

> Right now we always do SCOPE_SYSTEM from update_cpu_capabilities(), even for
> cpu_errata table. There is no specific reason for that.

I'm totally confused. Can you define SCOPE_SYSTEM and SCOPE_LOCAL for me,
please?

Will