linux-kernel - Re: [PATCH v7] KVM: arm64: Fix confusion in documentation for pKVM SME assert

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <Z62NxRzbOyt-nBmK@linux.dev>
Date: Wed, 12 Feb 2025 22:14:29 -0800
From: Oliver Upton <oliver.upton@...ux.dev>
To: Mark Rutland <mark.rutland@....com>
Cc: Mark Brown <broonie@...nel.org>, Marc Zyngier <maz@...nel.org>,
	linux-arm-kernel@...ts.infradead.org, kvmarm@...ts.linux.dev,
	linux-kernel@...r.kernel.org, Fuad Tabba <tabba@...gle.com>,
	James Morse <james.morse@....com>,
	Suzuki K Poulose <suzuki.poulose@....com>,
	Catalin Marinas <catalin.marinas@....com>,
	Will Deacon <will@...nel.org>
Subject: Re: [PATCH v7] KVM: arm64: Fix confusion in documentation for pKVM
 SME assert

On Wed, Feb 12, 2025 at 11:11:04AM +0000, Mark Rutland wrote:
> On Wed, Feb 12, 2025 at 12:44:57AM +0000, Mark Brown wrote:
> > As raised in the review comments for the original patch the assert and
> > comment added in afb91f5f8ad7 ("KVM: arm64: Ensure that SME controls are
> > disabled in protected mode") are bogus. The comments says that we check
> > that we do not have SME enabled for a pKVM guest but the assert actually
> > checks to see if the host has anything set in SVCR which is unrelated to
> > the guest features or state, regardless of if those guests are protected
> > or not. This check is also made in the hypervisor, it will refuse to run
> > a guest if the check fails, so it appears that the assert here is
> > intended to improve diagnostics.
> > 
> > Update the comment to reflect the check in the code, and to clarify that
> > we do actually enforce this in the hypervisor. While we're here also
> > update to use a WARN_ON_ONCE() to avoid log spam if this triggers.
> > 
> > Fixes: afb91f5f8ad7 ("KVM: arm64: Ensure that SME controls are disabled in protected mode")
> > Reviewed-by: Fuad Tabba <tabba@...gle.com>
> > Signed-off-by: Mark Brown <broonie@...nel.org>

I don't think a Fixes tag is warranted here, this doesn't fix any
functional issue.

> > diff --git a/arch/arm64/kvm/fpsimd.c b/arch/arm64/kvm/fpsimd.c
> > index 4d3d1a2eb157047b4b2488e9c4ffaabc6f5a0818..e37e53883c357093ff4455f5afdaec90e662d744 100644
> > --- a/arch/arm64/kvm/fpsimd.c
> > +++ b/arch/arm64/kvm/fpsimd.c
> > @@ -93,11 +93,14 @@ void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu)
> >  	}
> >  
> >  	/*
> > -	 * If normal guests gain SME support, maintain this behavior for pKVM
> > -	 * guests, which don't support SME.
> > +	 * Protected and non-protected KVM modes require that
> > +	 * SVCR.{SM,ZA} == {0,0} when entering a guest so that no
> > +	 * host/guest SME state needs to be saved/restored by hyp code.
> > +	 *
> > +	 * In protected mode, hyp code will verify this later.
> >  	 */
> > -	WARN_ON(is_protected_kvm_enabled() && system_supports_sme() &&
> > -		read_sysreg_s(SYS_SVCR));
> > +	WARN_ON_ONCE(is_protected_kvm_enabled() && system_supports_sme() &&
> > +		     read_sysreg_s(SYS_SVCR));
> 
> As I mentioned on the last round, we can drop the is_protected_kvm_enabled()
> check, i.e. have:
> 
> 	/*
> 	 * Protected and non-protected KVM modes require that
> 	 * SVCR.{SM,ZA} == {0,0} when entering a guest so that no
> 	 * host/guest SME state needs to be saved/restored by hyp code.
> 	 *
> 	 * In protected mode, hyp code will verify this later.
> 	 */
> 	WARN_ON_ONCE(system_supports_sme() && read_sysreg_s(SYS_SVCR));
> 
> Either way:
> 
> Acked-by: Mark Rutland <mark.rutland@....com>
> 
> Marc, are you happy to queue this atop the recent fixes from me? Those
> try to ensure SVCR.{SM,ZA} == {0,0} regardless of whether KVM is in
> protected mode.

I'll pick it up for 6.15 if Marc doesn't grab it as a fix.

-- 
Thanks,
Oliver