Re: migration test fails for aarch64 target (was: Re: [PULL 08/30] tests: Add dirty page rate limit test)

[Peter Maydell]

On Mon, 1 Aug 2022 at 11:41, Thomas Huth <thuth@redhat.com> wrote:
>
> On 20/07/2022 13.19, Dr. David Alan Gilbert (git) wrote:
> > From: Hyman Huang(黄勇) <huangy81@chinatelecom.cn>
> >
> > Add dirty page rate limit test if kernel support dirty ring,
> >
> > The following qmp commands are covered by this test case:
> > "calc-dirty-rate", "query-dirty-rate", "set-vcpu-dirty-limit",
> > "cancel-vcpu-dirty-limit" and "query-vcpu-dirty-limit".

> This assert triggers on my x86 laptop when I run "make check" there. More
> precisely, it triggers when qemu-system-aarch64 is getting tested:
>
> $ QTEST_QEMU_BINARY=./qemu-system-aarch64 tests/qtest/migration-test
> /aarch64/migration/bad_dest: OK
> /aarch64/migration/fd_proto: OK
> /aarch64/migration/validate_uuid: OK
> /aarch64/migration/validate_uuid_error: OK
> /aarch64/migration/validate_uuid_src_not_set: OK
> /aarch64/migration/validate_uuid_dst_not_set: OK
> /aarch64/migration/auto_converge: OK
> /aarch64/migration/dirty_ring: OK
> /aarch64/migration/vcpu_dirty_limit: migration-test:
> ../../devel/qemu/tests/qtest/migration-test.c:2304: dirtylimit_start_vm:
> Assertion `(strcmp(arch, "x86_64") == 0)' failed.
> Aborted (core dumped)
>
> >   static bool kvm_dirty_ring_supported(void)
> >   {
> >   #if defined(__linux__) && defined(HOST_X86_64)
> > @@ -2204,6 +2458,8 @@ int main(int argc, char **argv)
> >       if (kvm_dirty_ring_supported()) {
> >           qtest_add_func("/migration/dirty_ring",
> >                          test_precopy_unix_dirty_ring);
> > +        qtest_add_func("/migration/vcpu_dirty_limit",
> > +                       test_vcpu_dirty_limit);
> >       }
>
> kvm_dirty_ring_supported() returns "true" if the KVM of the *host* is x86
> and has the dirty ring support, but it does *not* check the target
> architecture that is currently tested, so you also get here "true" for
> qemu-system-aarch64 being tested on a x86 host... Thus I guess
> kvm_dirty_ring_supported() needs also a check for the right target 
> architecture?

Presumably it should be checking qtest_has_accel("kvm") -- even
if you're on an x86 host and with an x86 guest, if you're using
TCG you're not going to get KVM dirty ring support...

thanks
-- PMM