linux-kernel - Re: [PATCH v2] KVM: arm64: Constrain the host to the maximum shared SVE VL with pKVM

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] [day] [month] [year] [list]

Message-ID: <86ldzw7wfm.wl-maz@kernel.org>
Date: Thu, 12 Sep 2024 15:56:13 +0100
From: Marc Zyngier <maz@...nel.org>
To: Mark Brown <broonie@...nel.org>
Cc: Fuad Tabba <tabba@...gle.com>,
	Oliver Upton <oliver.upton@...ux.dev>,
	James Morse <james.morse@....com>,
	Suzuki K Poulose <suzuki.poulose@....com>,
	Catalin Marinas <catalin.marinas@....com>,
	Will Deacon <will@...nel.org>,
	Dave Martin <Dave.Martin@....com>,
	linux-arm-kernel@...ts.infradead.org,
	kvmarm@...ts.linux.dev,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2] KVM: arm64: Constrain the host to the maximum shared SVE VL with pKVM

On Thu, 12 Sep 2024 15:45:27 +0100,
Mark Brown <broonie@...nel.org> wrote:
> 
> [1  <text/plain; us-ascii (7bit)>]
> On Thu, Sep 12, 2024 at 03:21:43PM +0100, Fuad Tabba wrote:
> 
> > One part that you haven't changed is setting ZCR_EL2 during el2 setup:
> > arch/arm64/include/asm/el2_setup.h: .Linit_sve_ : lines 290/291
> 
> > I guess at that point it's not straightforward to figure sve_max_vl.
> > Is there a window after el2 setup where we might actually get the VL
> > implied by ZCR_ELx_LEN_MASK, or would it always get set to
> > sve_vq_from_vl(kvm_host_sve_max_vl) - 1 ?
> 
> Yeah, at that point we have no idea what any other cores might look like
> and there's just generally a bootstrapping issue - we need to see the
> actual maximum VLs for the PEs to work out what the maximum VL for the
> system is so we can tell KVM what to enforce.

That's fine. By the time KVM initialises, all the "early" CPUs will be
on, and pKVM actively prevents any late onlining of CPUs.

The only gotcha would be if "something" snapshots the "large" VL in
between two CPU boots, but we really don't expect that sort of things.

	M.

-- 
Without deviation from the norm, progress is not possible.