> On 29 Aug 2018, at 19:39, Radim Krčmář <rkrcmar@xxxxxxxxxx> wrote: > > 2018-08-29 09:34-0700, Sean Christopherson: >> On Wed, Aug 29, 2018 at 06:18:20PM +0200, Radim Krčmář wrote: >>> 2018-08-29 18:43+0300, Liran Alon: >>>> Consider the case L1 had a pending event until it executed >>>> VMLAUNCH/VMRESUME which wasn't delivered because it was disallowed >>>> (e.g. interrupts disabled). When L1 executes VMLAUNCH/VMRESUME, >>>> L0 needs to evaluate if this pending event should cause an exit from >>>> L2 to L1 or delivered directly to L2 (In case L1 don't intercept >>>> EXTERNAL_INTERRUPT). >>>> >>>> Usually this would be handled by L0 requesting a window (e.g. IRQ >>>> window) by setting VMCS accordingly. However, this setting was done on >>>> VMCS01 and now VMCS02 is active instead. Thus, when L1 executes >>>> VMLAUNCH/VMRESUME we force L0 to perform pending event evaluation by >>>> requesting a KVM_REQ_EVENT. >>>> >>>> Note that above scenario exists when L1 KVM is about to enter L2 but >>>> requests an "immediate-exit". As in this case, L1 will >>>> disable-interrupts and then send a self-IPI before entering L2. >>> >>> Which makes it a big blunder, I'll add "Cc: stable@xxxxxxxxxxxxxxx". >> >> Please hold off on doing anything with this, I don't think this is the >> correct fix. I have a half-finished response to the preemption timer >> thread that prompted this patch, I'll get that sent ASAP. > > Sure, thanks for the heads-up. Sean, I think this is orthogonal to the “immediate-exit” mechanism implementation issue you suggest to replace with preemption-timer with interval of 0 instead of self-IPI. In my opinion, this patch handles a general issue of losing pending interrupt queued (And disallowed from being dispatched) in L1 before entering L2. This is not just related to immediate-exit mechanism. This is also true for example for a timer-interrupt that may be raised L1 during the timespan in which L1 disables interrupts until he VMRESUME into L2. I have actually written a small effective kvm-unit-test for this. It fails before this patch and passes after it. I will submit the unit-test and Cc you guys. -Liran
