With a revert of b6b8a145 ('Rework interception of IRQs and NMIs'), the issue does not occur readily with the test case. I was able to run for 1+ hour. Generally I can reproduce within 15m.
With 9242b5b6 ('KVM: x86: Check for nested events if there is an injectable interrupt') applied, I can run for 1+ hour without issue.
Current 3.13.0 patchlevel is in between those two commits which allows for this bug to reproduce easily.
With a revert of b6b8a145 ('Rework interception of IRQs and NMIs'), the issue does not occur readily with the test case. I was able to run for 1+ hour. Generally I can reproduce within 15m.
With 9242b5b6 ('KVM: x86: Check for nested events if there is an injectable interrupt') applied, I can run for 1+ hour without issue.
Current 3.13.0 patchlevel is in between those two commits which allows for this bug to reproduce easily.