Re: [PATCH v2 02/20] KVM: x86/xen: Use read_trylock() for GPC locks in hardirq/atomic paths

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] Unbalanced interrupt disable causes permanent hard IRQ disable on PREEMPT_RT
--

commit 13e357726c8cbadaf777f9052a529db65b824338
Author: David Woodhouse <dwmw@amazon.co.uk>
KVM: x86/xen: Use read_trylock() for GPC locks in hardirq/atomic paths

This patch modifies the GPC lock acquisition paths to use read_trylock()
instead of read_lock_irqsave() in atomic contexts. It introduces an explicit
local_irq_save() to prevent issues on PREEMPT_RT kernels where rwlock_t
cannot be used with read_lock_irqsave() from hardirq context.

Reported-by: syzbot+208f7f3e5f59c11aeb90@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=208f7f3e5f59c11aeb90
Fixes: 14243b387137 ("KVM: x86/xen: Add KVM_IRQ_ROUTING_XEN_EVTCHN and event channel delivery")

> diff --git a/arch/x86/kvm/xen.c b/arch/x86/kvm/xen.c
> index 91fd3673c09a2..9bdb8e3cad58b 100644
> --- a/arch/x86/kvm/xen.c
> +++ b/arch/x86/kvm/xen.c
> @@ -713,7 +714,15 @@ int __kvm_xen_has_interrupt(struct kvm_vcpu *v)
>  	BUILD_BUG_ON(sizeof(rc) !=
>  		     sizeof_field(struct compat_vcpu_info, evtchn_upcall_pending));
>  
> -	read_lock_irqsave(&gpc->lock, flags);
> +	if (atomic) {
> +		local_irq_save(flags);
> +		if (!read_trylock(&gpc->lock)) {
> +			local_irq_restore(flags);
> +			return 1;
> +		}
> +	} else {
> +		read_lock_irqsave(&gpc->lock, flags);
> +	}
>  	while (!kvm_gpc_check(gpc, sizeof(struct vcpu_info))) {
>  		read_unlock_irqrestore(&gpc->lock, flags);

[Severity: High]
Does this result in unbalanced interrupt disables on PREEMPT_RT?

When atomic is true, local_irq_save() explicitly disables hardware interrupts.
On standard kernels, read_unlock_irqrestore() restores the interrupt state
correctly.

However, on PREEMPT_RT, rwlock_t is a sleepable lock and the implementation
of read_unlock_irqrestore() ignores the flags parameter, merely dropping the
lock without re-enabling interrupts. The interrupt state saved by
local_irq_save() is never restored, which would lead to hardware interrupts
being permanently disabled on the CPU.

I noticed this issue is resolved later in the series by commit 625a2fc07daa6
("KVM: x86/xen: Remove unnecessary irqsave from GPC lock usage in xen.c").

Should this fix be folded into this patch to ensure the kernel remains fully
functional and bisectable at this step?

[ ... ]

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260529165114.748639-1-seanjc@google.com?part=2