On 2013-06-26 10:03, Kashyap Chamarthy wrote: >> Thanks for the note, it's very helpful! This test actually fails on >> older CPUs as well, and I can finally reproduce the issue that Jay also >> reported. I'm not able to "cure" it by going back to 3b656cf764^, > > Ok, you tried w/o this commit.. > ------------------------ > commit 3b656cf764cbc43d3efb9bf5f45c618d4cf0989f > Author: Jan Kiszka <jan.kiszka@xxxxxxxxxxx> > Date: Sun Apr 14 12:12:45 2013 +0200 > > KVM: nVMX: Fix injection of PENDING_INTERRUPT and NMI_WINDOW exits to L1 > > Check if the interrupt or NMI window exit is for L1 by testing if it has > the corresponding controls enabled. This is required when we allow > direct injection from L0 to L2 > ------------------------ I first tried by reverting to the commit before this one, just like Jay reported for https://bugzilla.kernel.org/show_bug.cgi?id=58941. But this just varied the error (kvm reports an internal error), didn't solve the issue. Now I simply reverted the commit on top of next, but without an effect. Looks like those problems are not directly related. Kashyap, you can do us a favor and try to find out if there was a commit in the recent history (roughly before I started to hack on nVMX this year) where this test cases succeeded. TIA, Jan
Attachment:
signature.asc
Description: OpenPGP digital signature
