Re: [PATCH v2 07/22] arm64: Keep track of CPU feature registers

From: Catalin Marinas <catalin.marinas@arm.com>
To: Mark Rutland <mark.rutland@arm.com>
Cc: "Suzuki K. Poulose" <Suzuki.Poulose@arm.com>,
	Vladimir.Murzin@arm.com, steve.capper@linaro.org,
	ard.biesheuvel@linaro.org, marc.zyngier@arm.com,
	will.deacon@arm.com, linux-kernel@vger.kernel.org,
	edward.nevill@linaro.org, aph@redhat.com, james.morse@arm.com,
	andre.przywara@arm.com, dave.martin@arm.com,
	linux-arm-kernel@lists.infradead.org
Subject: Re: [PATCH v2 07/22] arm64: Keep track of CPU feature registers
Date: Tue, 13 Oct 2015 10:40:26 +0100	[thread overview]
Message-ID: <20151013094025.GA18105@localhost> (raw)
In-Reply-To: <20151012172104.GC3659@leverpostej>

On Mon, Oct 12, 2015 at 06:21:04PM +0100, Mark Rutland wrote:
> On Mon, Oct 12, 2015 at 06:01:25PM +0100, Suzuki K. Poulose wrote:
> > On 08/10/15 16:03, Catalin Marinas wrote:
> > >On Thu, Oct 08, 2015 at 10:55:11AM +0100, Suzuki K. Poulose wrote:
> > 
> > ...
> > >
> > >So we have three types of fields in these registers:
> > >
> > >a) features defined but not something we care about in Linux
> > >b) reserved fields
> > >c) features important to Linux
> > >
> > >I guess for (a), Linux may not even care if they don't match (though we
> > >need to be careful which fields we ignore). As for (b), even if they
> > >differ, since we don't know the meaning at this point, I think we should
> > >just ignore them. If, for example, they add a feature that Linux doesn't
> > >care about, they practically fall under the (a) category.
> > >
> > >Regarding exposing reserved CPUID fields to user, I assume we would
> > >always return 0.
> > 
> > Mark,
> > 
> > Do you have any comments on this ? The list I have here is what you came
> > up with in SANITY checks.
> 
> My feeling was that we should play it safe with fields which are
> currently reserved (warning if they differ for now).
> 
> If they turn out to be irrelevant, it's simple to backport a patch to
> ignore them, whereas if they matter we get instant visibility, which is
> the entire point of the sanity checks.
> 
> So I think we should warn if reserved fields differ. I'd rather have a
> few spurious warnings until kernels get updated than miss an issue that
> could have been dealt with and avoided.

The warnings are indeed harmless. I think the main danger is when a
field goes negative which means an existing feature without CPUID field
allocated is removed (though I don't expect this for ARMv8). But you are
right, let the warnings in for now, let's re-assess when/if a difference
happens.

-- 
Catalin