linux-kernel - Re: [PATCH v3 0/4] KVM: arm64: writable MIDR/REVIDR

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 for Android: free password hash cracker in your pocket

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

Message-ID: <Z7zxSdW5b6rqe2bt@linux.dev>
Date: Mon, 24 Feb 2025 14:23:05 -0800
From: Oliver Upton <oliver.upton@...ux.dev>
To: Sebastian Ott <sebott@...hat.com>
Cc: Marc Zyngier <maz@...nel.org>, Joey Gouly <joey.gouly@....com>,
	Suzuki K Poulose <suzuki.poulose@....com>,
	Zenghui Yu <yuzenghui@...wei.com>,
	Catalin Marinas <catalin.marinas@....com>,
	Will Deacon <will@...nel.org>,
	Shameer Kolothum <shameerali.kolothum.thodi@...wei.com>,
	Cornelia Huck <cohuck@...hat.com>,
	Eric Auger <eric.auger@...hat.com>,
	linux-arm-kernel@...ts.infradead.org, kvmarm@...ts.linux.dev,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH v3 0/4] KVM: arm64: writable MIDR/REVIDR

Hi Sebastian,

On Tue, Feb 18, 2025 at 05:34:39PM +0100, Sebastian Ott wrote:
> Based on prior discussion [1] this series allows VMMs to change
> MIDR/REVIDR to enable migration between machines that differ in
> these registers. Since these are used for errata handling the
> errata management series [2] is a prerequisite for this one.
> 
> changes for V3:
> * handle VPIDR_EL2 as part of vcpu ctxt - thanks Oliver!

Thanks for respinning. While your changes are looking good, as I got
ready to apply this series I wound up peeling the onion a bit further
and have a few more concerns:

 - Current KVM allows guests to read SMIDR_EL1 despite the fact that we
   do not support SME (this is part of TID1 traps)

 - The "invariant" values that KVM presents to userspace originate from
   the boot CPU, not the CPU that resets the ID registers for a VM

 - A VMM that wants to present big-little can do so on current KVM by
   affining vCPUs, but cannot with this series

All of this is to say, I think your series is going to collide with
the pre-existing pile of crap we have. I'm going to pick up these
changes and rework them so we can send a fix for #1 to stable trees and
(hopefully) avoid breaking the old "invariant" behavior.

I'll post what I have as soon as I test it, hopefully we can get this
shaped up for 6.15.

Thanks,
Oliver