Re: [RFC PATCH] arm64: unify WORKAROUND_SPECULATIVE_AT_{NVHE,VHE}

From: James Morse <james.morse@arm.com>
To: Andrew Scull <ascull@google.com>
Cc: qwandor@google.com, Marc Zyngier <maz@kernel.org>,
	Steven Price <steven.price@arm.com>,
	Will Deacon <will@kernel.org>,
	wedsonaf@google.com, linux-arm-kernel@lists.infradead.org,
	dbrazdil@google.com, kernel-team@android.com,
	kvmarm@lists.cs.columbia.edu, tabba@google.com
Subject: Re: [RFC PATCH] arm64: unify WORKAROUND_SPECULATIVE_AT_{NVHE,VHE}
Date: Fri, 27 Mar 2020 16:11:56 +0000	[thread overview]
Message-ID: <a8cc7a17-cb84-3e52-15f4-87b27a01876b@arm.com> (raw)
In-Reply-To: <20200327143941.195626-1-ascull@google.com>

Hi Andrew,

On 3/27/20 2:39 PM, Andrew Scull wrote:
> Errata 1165522, 1319367 and 1530923 each allow TLB entries to be
> allocated as a result of a speculative AT instruction. In order to
> avoid mandating VHE on certain affected CPUs, apply the workaround to
> both the nVHE and the VHE case for all affected CPUs.

You're booting a VHE capable system without VHE, and need KVM?
Do tell!

Would enabling the nVHE workaround on a VHE capable part solve your problem?
Merging the errata has some side effects...

> ---
> I'm not able to test the workarounds properly for the affected CPUs but
> have built and booted under qemu configs with and without VHE as well as
> the workaround being enabled and disabled.
> 
> As there exist work arounds for nVHE and VHE, it doesn't appear to be a
> technical limitation that meant VHE was being mandated. Please correct
> me if this understanding is inaccurate. Thanks!

The affected VHE parts came first. Then came those that didn't have VHE at all.

> diff --git a/arch/arm64/kvm/hyp/sysreg-sr.c b/arch/arm64/kvm/hyp/sysreg-sr.c
> index 7672a978926c..2c1436fc0830 100644
> --- a/arch/arm64/kvm/hyp/sysreg-sr.c
> +++ b/arch/arm64/kvm/hyp/sysreg-sr.c
> @@ -118,7 +118,7 @@ static void __hyp_text __sysreg_restore_el1_state(struct kvm_cpu_context *ctxt)
>  	write_sysreg(ctxt->sys_regs[MPIDR_EL1],		vmpidr_el2);
>  	write_sysreg(ctxt->sys_regs[CSSELR_EL1],	csselr_el1);
>  
> -	if (!cpus_have_const_cap(ARM64_WORKAROUND_SPECULATIVE_AT_NVHE)) {
> +	if (!cpus_have_const_cap(ARM64_WORKAROUND_SPECULATIVE_AT)) {
>  		write_sysreg_el1(ctxt->sys_regs[SCTLR_EL1],	SYS_SCTLR);
>  		write_sysreg_el1(ctxt->sys_regs[TCR_EL1],	SYS_TCR);
>  	} else	if (!ctxt->__hyp_running_vcpu) {

The comment just below here:
|		/*
|		 * Must only be done for guest registers, hence the context
|		 * test. We're coming from the host, so SCTLR.M is already
|		 * set. Pairs with __activate_traps_nvhe().
|		 */

The VHE parts aren't going to run __activate_traps_nvhe(), so you skip restoring
the guest's SCTLR_EL1 and TCR_EL1...

Thanks,

James
_______________________________________________
kvmarm mailing list
kvmarm@lists.cs.columbia.edu
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm