Re: [PATCH v4 06/18] KVM: arm64: Skip fixed-feature state flush for protected vCPUs

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] kvm: arm64: host hardware breakpoints leak into protected guests
--

commit 678dfb9fd1576272693dba360820510bcb1ec938
Author: Fuad Tabba <fuad.tabba@linux.dev>

    KVM: arm64: Skip fixed-feature state flush for protected vCPUs

This commit moves the mdcr_el2 and TWI/TWE copies into the non-protected
branch. It returns early from flush_debug_state() and sync_debug_state() for
protected vCPUs, as their debug registers are hypervisor-owned, and
configures HPMN at vCPU load.

> diff --git a/arch/arm64/kvm/hyp/nvhe/hyp-main.c b/arch/arm64/kvm/hyp/nvhe/hyp-main.c
> index e3f4bbb623492..b7b9f5afb673b 100644
> --- a/arch/arm64/kvm/hyp/nvhe/hyp-main.c
> +++ b/arch/arm64/kvm/hyp/nvhe/hyp-main.c
> @@ -200,6 +200,9 @@ static void flush_debug_state(struct pkvm_hyp_vcpu *hyp_vcpu)
>  {
>  	struct kvm_vcpu *host_vcpu = hyp_vcpu->host_vcpu;
>  
> +	if (pkvm_hyp_vcpu_is_protected(hyp_vcpu))
> +		return;
> +

[Severity: High]
Could this early return cause a regression by allowing the host's hardware
breakpoints to leak into the protected guest?

By returning early here, the hyp vCPU's debug_owner remains VCPU_DEBUG_FREE (0)
for protected VMs.

During the world switch, this causes the hypervisor to skip context-switching
the hardware debug registers because kvm_debug_regs_in_use() evaluates to false:

arch/arm64/kvm/hyp/include/hyp/debug-sr.h:__debug_switch_to_guest_common() {
    ...
	if (!kvm_debug_regs_in_use(vcpu))
		return;
    ...
}

As a result, the host's physical breakpoints remain active in hardware.

If the guest implements DebugVer, MDCR_EL2.TDA is not set:

arch/arm64/kvm/hyp/nvhe/pkvm.c:pvm_init_traps_mdcr() {
    ...
	if (!kvm_has_feat(kvm, ID_AA64DFR0_EL1, DebugVer, IMP))
		val |= MDCR_EL2_TDRA | MDCR_EL2_TDA;
    ...
}

This gives the guest untrapped access to MDSCR_EL1. If the guest sets
MDSCR_EL1.MDE=1, the host's hardware breakpoints could fire inside the guest
at EL1, circumventing guest isolation.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261001135711.1640520-1-fuad.tabba@linux.dev?part=6