On Fri, 2026-09-11 at 14:36 -0700, Sohil Mehta wrote: > Opens > ----- > INT1 event type: David Woodhouse shared a selftest which shows [2] KVM > re-injects an intercepted ICEBP (INT1) #DB as a hardware exception, so a > FRED guest reads event type 3 instead of 5. IIUC, the issue is not FRED > specific but it becomes visible due to FRED. Is that understanding > correct? Also, I am unsure of where the fix should reside or if a > generic fix is already in the works. Any insights here would be helpful.
That's my understanding too, yes. It doesn't *matter* what type it's re-injected as, until the guest has FRED support and can actually see the difference. So while the issue is pre-existing, it's only actually a *bug* in KVM once your series lands. So I think you need to plumb the event type through the KVM exception queue as an explicit field.
smime.p7s
Description: S/MIME cryptographic signature
