Re: [PATCH v7 03/12] KVM: arm64: PMU: Clear PM{C,I}NTEN{SET,CLR} and PMOVS{SET,CLR} on vCPU reset

From: Oliver Upton <oliver.upton@linux.dev>
To: Raghavendra Rao Ananta <rananta@google.com>
Cc: Eric Auger <eauger@redhat.com>, Marc Zyngier <maz@kernel.org>,
	Alexandru Elisei <alexandru.elisei@arm.com>,
	James Morse <james.morse@arm.com>,
	Suzuki K Poulose <suzuki.poulose@arm.com>,
	Paolo Bonzini <pbonzini@redhat.com>,
	Zenghui Yu <yuzenghui@huawei.com>,
	Shaoqin Huang <shahuang@redhat.com>,
	Jing Zhang <jingzhangos@google.com>,
	Reiji Watanabe <reijiw@google.com>,
	Colton Lewis <coltonlewis@google.com>,
	linux-arm-kernel@lists.infradead.org, kvmarm@lists.linux.dev,
	linux-kernel@vger.kernel.org, kvm@vger.kernel.org
Subject: Re: [PATCH v7 03/12] KVM: arm64: PMU: Clear PM{C,I}NTEN{SET,CLR} and PMOVS{SET,CLR} on vCPU reset
Date: Thu, 19 Oct 2023 19:05:58 +0000	[thread overview]
Message-ID: <ZTF-FlDtvha-6Pw1@linux.dev> (raw)
In-Reply-To: <CAJHc60yQSzsuTJLcyzs5vffgRzR5i0vKQwLnhavAon6hoSkb+A@mail.gmail.com>

Hi Raghu,

Can you please make sure you include leading and trailing whitespace for
your inline replies? The message gets extremely dense and is difficult
to read.

Also -- delete any unrelated context from your replies. If there's a
localized conversation about a particular detail there's no reason to
keep the entire thread in the body.

On Thu, Oct 19, 2023 at 11:46:22AM -0700, Raghavendra Rao Ananta wrote:
> On Wed, Oct 18, 2023 at 2:16 PM Raghavendra Rao Ananta
> <rananta@google.com> wrote:
> > I had a brief discussion about this with Oliver, and it looks like we
> > might need a couple of additional changes for these register accesses:
> > - For the userspace accesses, we have to implement explicit get_user
> > and set_user callbacks that to filter out the unimplemented counters
> > using kvm_pmu_valid_counter_mask().
> Re-thinking the first case: Since these registers go through a reset
> (reset_pmu_reg()) during initialization, where the valid counter mask
> is applied, and since we are sanitizing the registers with the mask
> before running the guest (below case), will implementing the
> {get,set}_user() add any value, apart from just keeping userspace in
> sync with every update of PMCR.N?

KVM's sysreg emulation (as seen from userspace) fails to uphold the RES0
bits of these registers. That's a bug.

-- 
Thanks,
Oliver