On 4/14/20 12:11 PM, Vitaly Kuznetsov wrote: > Nitesh Narayan Lal <nitesh@xxxxxxxxxx> writes: > >> On 4/14/20 11:30 AM, Vitaly Kuznetsov wrote: >>> Nitesh Narayan Lal <nitesh@xxxxxxxxxx> writes: >>> >>>> On 4/14/20 10:01 AM, Vitaly Kuznetsov wrote: >>>>> Also, this patch could've been split. >>>> I can divide it 2 parts: >>>> 1. support for logical destination mode. >>>> 2. support for physical destination mode. I can also fix the above issue in >>>> this patch itself. >>>> Does that make sense? >>> Too late, it's already commited :-) I just meant to say that >>> e.g. spinlock part could've been split into its own patch, unittests.cfg >>> - another one,... >> Ah, I see. I will be more careful. >> For now, I will just move the physical destination mode test back under >> the check. Will that be acceptable as a standalone patch? > This is already in Paolo's patch: > https://lore.kernel.org/kvm/87zhbexh3v.fsf@xxxxxxxxxxxxxxxxxxxx/T/#m9791cd50a9d82fabdaddcb9259d14df3b89ed250 Yeap, this patch is already in the tree. I was referring to a new patch in which I would move the physical destination mode test under 'if (cpu_count() > 1)', that should fixed the ioapic-split hang. Sorry about the confusion. > >> In between I have a question is it normal for test_ioapic_self_reconfigure() >> to fail when executed with irqchip split? >> If so do we expect that it will leave the VM in some sort of dirty state >> that causes the following test to fail? > Not sure I got your question but IMO when someone does > ./run_tests.sh > all tests are supposed to pass -- unless there is a bug in KVM (e.g. the > person is running an old kernel). In case we're seeing failures (or, > even worse, hangs) with the latest upstream kernel -- something is > broken, either KVM or kvm-unit-tests. Yeap, I am not sure what is the exact issue that causes test_ioapic_self_reconfigure() to fail when used with kernel_irqchip=split. I can again take a look to find out the root cause later. -- Nitesh
