Re: [PATCH v3 1/2] KVM: powerpc: Use generic xfer to guest work function

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] Lockdep assertion failure in xfer_to_guest_mode_prepare() due to mismatched IRQ state on BookE architecture after emulated MTMSRWE.
--

--- Patch [1]: [PATCH v3 1/2] KVM: powerpc: Use generic xfer to guest work function ---
commit ad7243cc79a6e34d4784ae75e8ea35ce946a8343
Author: Vishal Chourasia <vishalc@linux.ibm.com>

KVM: powerpc: Use generic xfer to guest work function

This commit updates powerpc KVM run loops to use generic infrastructure for
checking and handling pending task_work before guest entry. It replaces
open-coded need_resched() checks and adds xfer_to_guest_mode_prepare() to
disable IRQs and handle TIF_NOTIFY_RESUME properly.

> diff --git a/arch/powerpc/kvm/powerpc.c b/arch/powerpc/kvm/powerpc.c
> index 00302399fc37b..be5e48ae0c6c2 100644
> --- a/arch/powerpc/kvm/powerpc.c
> +++ b/arch/powerpc/kvm/powerpc.c
> @@ -81,23 +81,41 @@ int kvmppc_prepare_to_enter(struct kvm_vcpu *vcpu)
>  	int r;
>  
>  	WARN_ON(irqs_disabled());
> +	/*
> +	 * local_irq_disable() first: on 32-bit, hard_irq_disable() alone is a
> +	 * raw MSR[EE] clear that bypasses the lockdep/irq-tracing state, and
> +	 * the xfer_to_guest_mode helpers assert IRQs are seen as disabled.
> +	 */
> +	local_irq_disable();
>  	hard_irq_disable();
>  
>  	while (true) {
> -		if (need_resched()) {
> +		xfer_to_guest_mode_prepare();

[Severity: High]
Can this trigger a lockdep assertion failure on BookE architecture?

When a BookE guest executes mtmsr to set the MSR_WE bit, it enters an idle
state and invokes kvmppc_core_prepare_to_enter(). This function enables
lockdep IRQs but disables hardware IRQs using a macro that bypasses lockdep 
on 32-bit:

arch/powerpc/kvm/booke.c:kvmppc_core_prepare_to_enter() {
    ...
    if (vcpu->arch.shared->msr & MSR_WE) {
        local_irq_enable();
        kvm_vcpu_halt(vcpu);
        hard_irq_disable();
        ...
        r = 1;
    }
}

If kvmppc_core_prepare_to_enter() returns 1, the while loop later in 
kvmppc_prepare_to_enter() continues:

arch/powerpc/kvm/powerpc.c:kvmppc_prepare_to_enter() {
    ...
    if (kvmppc_core_prepare_to_enter(vcpu)) {
        /* interrupts got enabled in between, so we
           are back at square 1 */
        continue;
    }
}

This continues back to the start of the loop and invokes the newly added 
xfer_to_guest_mode_prepare() check. Since hard_irq_disable() bypassed the 
lockdep state update, lockdep still believes IRQs are enabled. Will this 
cause xfer_to_guest_mode_prepare() to fail its lockdep_assert_irqs_disabled() 
check and trigger a splat?

[ ... ]

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260709062940.1679427-2-vishalc@linux.ibm.com?part=1