On Mon, Jul 30, 2018 at 10:41:45AM +0200, Greg Kurz wrote: > On Mon, 30 Jul 2018 15:57:15 +1000 > David Gibson wrote: > > > On Fri, Jul 27, 2018 at 09:54:52AM +0200, Greg Kurz wrote: > > > On Fri, 27 Jul 2018 15:27:24 +1000 > > > David Gibson wrote: > > > > > > > On Wed, Jul 25, 2018 at 04:45:26PM +0200, Greg Kurz wrote: > > > > > Commit b585395b655 fixed a regression introduced by some recent changes > > > > > in the XICS code, that was causing QEMU to crash instantly during CPU > > > > > hotplug with KVM. This is typically the kind of bug we'd like our > > > > > test suite to detect before it gets merged. Unfortunately, the current > > > > > tests run with '-machine accel=qtest' and don't exercise KVM specific > > > > > paths in QEMU. > > > > > > > > > > This patch hence changes add_pseries_test_case() to launch QEMU with > > > > > '-machine accel=kvm' if KVM is available. > > > > > > > > > > A notable consequence is that the guest will execute SLOF, but for some > > > > > reasons SLOF sometimes hits a program exception. This causes the guest > > > > > to loop forever and the test to be stuck. Since we don't need the guest > > > > > to be truely running, let's pass -S to QEMU to avoid that. > > > > > > > > > > Also disable machine capabilities that could be unavailable in KVM, eg, > > > > > when using PR KVM. > > > > > > > > > > Signed-off-by: Greg Kurz > > > > > > > > I'm pretty sure trying to change the accelerator on a qtest test just > > > > doesn't make sense. We'd need a different approach for testing cpu > > > > hotplug against kvm & tcg backends. > > > > > > > > > > The test starts QEMU, triggers the CPU hotplug code with a QMP command > > > and checks the command didn't fail (or QEMU didn't crash, as it would > > > have before commit b585395b655a)... I really don't understand what > > > is wrong with that... Please elaborate. > > > > Well, ok, let me turn that around. A test that doesn't rely on > > controlling the guest side behaviour at all probably shouldn't be a > > qtest based test, since that's what qtest is all about. > > > > The CPU hotplug test doesn't seem to do anything on the guest side: it > just checks that 'device_add' returns a response that isn't an > error. Right, which makes this ill suited to being a qtest test. The whole point of qtest is making it easier to test qemu peripherals without having to have specific test code within the guest, by essentially replacing the guest's cpu with a stub controlled by the test harness. That's what the qtest accel is all about. I think it's confusing to have a test which tries things with both qtest and kvm accelerators. Looking like a qtest test, people might reasonably think they can extend/refine the test to check behaviour when the guest does respond to the hotplug events. But such an extension won't work with the kvm accel, because the qtest code used to simulate that guest response won't have any effect with kvm. > I'm not aware that the guest is expected to have a specific behavior > during 'device_add', apart from not crashing or hanging. That was the > initial idea behind passing '-S' to ensure the guest doesn't run. Note that with qtest (or -S) we don't even test that minimal condition. We only test that *qemu* doesn't crash - it could fatally compromise the guest and the test would never know. > Your remark seems to be more general though... are you meaning that > doing something like qtest_start("-machine accel=kvm:tcg") is just > wrong ? Pretty much, yes. A non-qtest test which does that is reasonable, but not a qtest test. -- David Gibson | I'll have my music baroque, and my code david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_ | _way_ _around_! http://www.ozlabs.org/~dgibson