Re: [PATCH] KVM: VMX: Fall back to IRR scan when PIR is empty despite PID.ON being set

[Sean Christopherson <seanjc@google.com>]

On Wed, Apr 29, 2026, Chenyi Qiang wrote:
> On 4/28/2026 11:50 PM, Sean Christopherson wrote:
> > On Tue, Apr 28, 2026, Chenyi Qiang wrote:
> >> On 4/28/2026 3:45 PM, Paolo Bonzini wrote:
> >>> On 4/28/26 09:03, Chenyi Qiang wrote:
> >>>> The interrupt is not lost (it resides in the IRR from the first sync and
> >>>> is recovered on the next vcpu_enter_guest() iteration), but the incorrect
> >>>> max_irr causes a spurious WARNING and a wasted L2 VM-Enter/VM-Exit cycle.
> >>>>
> >>>> Fixes: b41f8638b9d3 ("KVM: VMX: Isolate pure loads from atomic XCHG when processing PIR")
> > 
> > Cc: stable@vger.kernel.org
> > 
> >>>> Reported-by: Farrah Chen <farrah.chen@intel.com>
> >>>> Assisted-by: GitHub Copilot:Claude Opus 4.6
> >>>> Signed-off-by: Chenyi Qiang <chenyi.qiang@intel.com>
> >>>>
> >>>> ---
> >>>> There is a WARNING call trace during a nested VM stress test.
> > 
> > Any chance the stress test is something that can be shared?  I've seen this WARN
> > 2-3 times over the last year, but it was never reproducible, and so intermittent
> > that I couldn't even correlate what I was doing at the time with the WARN.
> 
> It's not something special. The stress test is only to create and tear down
> 20 nested VM instance in a cycle.  It often takes more than 20 cycles to
> reproduce this issue.

Got it, thanks much!