Re: [RFC PATCH 1/2] arm64: kvm: Save/restore MTE registers

[Catalin Marinas]

On Wed, Jun 17, 2020 at 01:38:43PM +0100, Steven Price wrote:
> diff --git a/arch/arm64/kvm/hyp/sysreg-sr.c b/arch/arm64/kvm/hyp/sysreg-sr.c
> index 75b1925763f1..6ecee1528566 100644
> --- a/arch/arm64/kvm/hyp/sysreg-sr.c
> +++ b/arch/arm64/kvm/hyp/sysreg-sr.c
> @@ -26,6 +26,12 @@
>  static void __hyp_text __sysreg_save_common_state(struct kvm_cpu_context *ctxt)
>  {
>  	ctxt->sys_regs[MDSCR_EL1]	= read_sysreg(mdscr_el1);
> +	if (system_supports_mte()) {
> +		ctxt->sys_regs[RGSR_EL1] = read_sysreg_s(SYS_RGSR_EL1);
> +		ctxt->sys_regs[GCR_EL1] = read_sysreg_s(SYS_GCR_EL1);
> +		ctxt->sys_regs[TFSRE0_EL1] = read_sysreg_s(SYS_TFSRE0_EL1);
> +		ctxt->sys_regs[TFSR_EL1] = read_sysreg_s(SYS_TFSR_EL1);
> +	}

TFSR_EL1 is not a common register as we have the TFSR_EL2 as well. So
you'd have to access it as read_sysreg_el1(SYS_TFSR) so that, in the VHE
case, it generates TFSR_EL12, otherwise you just save the host register.

Also, since TFSR*_EL1 can be set asynchronously, I think we need to set
the SCTLR_EL2.ITFSB bit so that the register update is synchronised on
entry to EL2. With VHE we get this automatically as part of
SCTLR_EL1_SET but it turns out that we have another SCTLR_ELx_FLAGS
macro for the non-VHE case (why not calling this SCTLR_EL2_* I have no
idea).

>  	/*
>  	 * The host arm64 Linux uses sp_el0 to point to 'current' and it must
> @@ -99,6 +105,12 @@ NOKPROBE_SYMBOL(sysreg_save_guest_state_vhe);
>  static void __hyp_text __sysreg_restore_common_state(struct kvm_cpu_context *ctxt)
>  {
>  	write_sysreg(ctxt->sys_regs[MDSCR_EL1],	  mdscr_el1);
> +	if (system_supports_mte()) {
> +		write_sysreg_s(ctxt->sys_regs[RGSR_EL1], SYS_RGSR_EL1);
> +		write_sysreg_s(ctxt->sys_regs[GCR_EL1], SYS_GCR_EL1);
> +		write_sysreg_s(ctxt->sys_regs[TFSRE0_EL1], SYS_TFSRE0_EL1);
> +		write_sysreg_s(ctxt->sys_regs[TFSR_EL1], SYS_TFSR_EL1);
> +	}

Similarly here, you override the TFSR_EL2 with VHE enabled.

-- 
Catalin