On 12/12/20 00:41, Oliver Upton wrote:
On Fri, Dec 11, 2020 at 5:20 PM Paolo Bonzini <pbonzini@xxxxxxxxxx> wrote:
On 06/10/20 23:25, Oliver Upton wrote:
If a guest blocks interrupts for the entirety of running in root mode
(RFLAGS.IF=0), a pending interrupt corresponding to the posted-interrupt
vector set in the VMCS should result in an interrupt posting to the vIRR
at VM-entry. However, on KVM this is not the case. The pending interrupt
is not recognized as the posted-interrupt vector and instead results in
an external interrupt VM-exit.
Add a regression test to exercise this issue.
Signed-off-by: Oliver Upton <oupton@xxxxxxxxxx>
I am a bit confused. Is this testing the KVM or the bare metal
behavior? Or was this fixed in KVM already?
This is a directed test case for
25bb2cf97139 ("KVM: nVMX: Morph notification vector IRQ on nested
VM-Enter to pending PI")
Ok, thanks. However, the patch currently fails like this:
Test suite: vmx_posted_interrupt_test
VM-Fail on vmlaunch: error number is 7. See Intel 30.4.
I haven't debugged it, so for now I suggest that you just move it to a
separate suite.
Paolo
