Sean Christopherson <sean.j.christopherson@xxxxxxxxx> writes: >> >> The crash from dmesg is: >> kernel: Virtual processor ID = 0x0001 >> kernel: PLE Gap=00000080 Window=00001000 >> kernel: EPT pointer = 0x000000015defb05e >> kernel: TPR Threshold = 0x00 >> kernel: TSC Offset = 0xffffff7fb064e3f6 >> kernel: IDTVectoring: info=00000000 errcode=00000000 >> kernel: reason=80000021 qualification=0000000000000000 >> kernel: VMExit: intr_info=80000306 errcode=00000000 ilen=00000002 > > Likely emulation related. It's a failed VM-Entry due to invalid guest > state, and there's a #UD logged in VM=Exit interrupt info (which doesn't > get cleared in this case). > > On a related topic, dumping the bytes at GUEST_RIP would be helpful > for things like this. > We had an off-list discussion with Jon and I was able to reproduce the issue even with the latest kvm/queue. The problem appears on 'rsm' emulation (rsm_load_state_64() fails) and I'm going to take a deeper look. Initially, I thought that it may again be related to MMU split which happened in 4.19 but the issue reproduces even when it is disabled. In other words: work in progress, stay tuned. -- Vitaly
