Re: [PATCH] tests/acceptance: Add a 'virt_kvm' test using the GICv3

[Cleber Rosa <crosa@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 5126 bytes --]

On Tue, Apr 13, 2021 at 06:25:47PM +0200, Philippe Mathieu-Daudé wrote:
> Hi Alex,
> 
> On 4/12/21 7:55 PM, Philippe Mathieu-Daudé wrote:
> > On 4/6/21 7:12 PM, Alex Bennée wrote:
> >>
> >> Philippe Mathieu-Daudé <philmd@redhat.com> writes:
> >>
> >>> On 3/31/21 5:45 PM, Alex Bennée wrote:
> >>>>
> >>>> Philippe Mathieu-Daudé <philmd@redhat.com> writes:
> >>>>
> >>>>> The current 'virt_kvm' test is restricted to GICv2, but can also
> >>>>> work with a GICv3. Duplicate it but add a GICv3 test which can be
> >>>>> tested on some hardware.
> >>>>>
> >>>>> Noticed while running:
> >>>>>
> >>>>>  $ avocado --show=app run -t machine:virt tests/acceptance/
> >>>>>  ...
> >>>>>  (2/6) tests/acceptance/boot_linux.py:BootLinuxAarch64.test_virt_kvm: ERROR: Unexpected empty reply from server (1.82 s)
> >>>>>
> >>>>> The job.log content is:
> >>>>>
> >>>>>   L0351 DEBUG| Output: 'qemu-system-aarch64: host does not support in-kernel GICv2 emulation\n'
> >>>>>
> >>>>> With this patch:
> >>>>>
> >>>>>  $ avocado --show=app run -t device:gicv3 tests/acceptance/
> >>>>>  (1/1)
> >>>>>  tests/acceptance/boot_linux.py:BootLinuxAarch64.test_virt_kvm_gicv3:
> >>>>>  PASS (55.10 s)
> >>>>
> >>>> On the new aarch64 machine which is GICv3 I get the following:
> >>>>
> >>>>  (006/142) tests/acceptance/boot_linux.py:BootLinuxAarch64.test_virt_kvm_gicv2: ERROR: Unexpected empty reply from server (0.47 s)
> >>>>
> >>>> which it shouldn't have run. However:
> >>>>
> >>>>   ./tests/venv/bin/avocado --show=app run -t device:gic3 tests/acceptance/
> >>>>   Test Suite could not be create. No test references provided nor any other arguments resolved into tests
> >>>>
> >>>> Is this something that has regressed or am I doing it wrong?
> >>>
> >>> Typo in the tag: "device:gic3" -> "device:gicv3"
> >>
> >> Doh!
> >>
> >> But what about:
> >>
> >> /tests/venv/bin/avocado run tests/acceptance/boot_linux.py:BootLinuxAarch64.test_virt_kvm_gicv2
> >> JOB ID     : 396696d8f9d31d970878cb46025b2ced76f3623f
> >> JOB LOG    : /home/alex/avocado/job-results/job-2021-04-06T17.11-396696d/job.log
> >>  (1/1) tests/acceptance/boot_linux.py:BootLinuxAarch64.test_virt_kvm_gicv2: ERROR: Unexpected empty reply from server (0.65 s)
> >> RESULTS    : PASS 0 | ERROR 1 | FAIL 0 | SKIP 0 | WARN 0 | INTERRUPT 0 | CANCEL 0
> >> JOB TIME   : 0.96 s
> >>
> >> why doesn't that skip?
> > 
> > /home/phil/avocado/job-results/job-2021-04-12T17.51-efdca81/job.log
> > 2021-04-12 17:52:44,589 machine          L0389 DEBUG| Output:
> > "qemu-system-aarch64: Could not find ROM image
> > '/home/phil/qemu/build/host/pc-bios/edk2-aarch64-code.fd'\n"
> > 
> > Missing prerequisite:
> > 
> > $ ninja pc-bios/edk2-aarch64-code.fd
> > [1/1] Generating edk2-aarch64-code.fd with a custom command (wrapped by
> > meson to capture output)
> > 
> > Then we are good:
> > 
> > $ avocado --show=app,console run -t device:gicv3 tests/acceptance
> > JOB ID     : e84401e5cc3ae53a3094c79491e661385cc7b4a7
> > JOB LOG    :
> > /home/phil/avocado/job-results/job-2021-04-12T17.53-e84401e/job.log
> >  (1/1)
> > tests/acceptance/boot_linux.py:BootLinuxAarch64.test_virt_kvm_gicv3:
> > PASS (16.38 s)
> > RESULTS    : PASS 1 | ERROR 0 | FAIL 0 | SKIP 0 | WARN 0 | INTERRUPT 0 |
> > CANCEL 0
> > JOB TIME   : 16.70 s
> > 
> > Probably some missing dependency in Makefile/Meson?
> 
> Are you using multiple build directories?
> 
> I could reproduce doing:
> 
> $ mkdir A B
> $ cd A
> $ make check-qtest-aarch64
> $ avocado --show=app,console run -t device:gicv3 tests/acceptance
> $ cd ../B
> $ ninja qemu-system-aarch64
> $ avocado --show=app,console run -t device:gicv3 tests/acceptance
> 
> In A edk2-aarch64-code.fd has been expanded in A/pc-bios/,
> in B it isn't.
> 
> check-acceptance is a Makefile rule, not a ninja one...
> I suppose we need to convert it to ninja to be able to use the
> rest of the dependencies checks.
> 
> Cc'ing Paolo because I'm not sure what the best move and where
> to plug things.
> 
> 

While a proper build system conversion fix doesn't happen, what about:

diff --git a/tests/acceptance/boot_linux.py b/tests/acceptance/boot_linux.py
index 314370fd1f..5d42d98dc7 100644
--- a/tests/acceptance/boot_linux.py
+++ b/tests/acceptance/boot_linux.py
@@ -69,9 +69,10 @@ class BootLinuxAarch64(LinuxTest):
     chksum = '1e18d9c0cf734940c4b5d5ec592facaed2af0ad0329383d5639c997fdf16fe49'
 
     def add_common_args(self):
-        self.vm.add_args('-bios',
-                         os.path.join(BUILD_DIR, 'pc-bios',
-                                      'edk2-aarch64-code.fd'))
+        bios_path = os.path.join(BUILD_DIR, 'pc-bios', 'edk2-aarch64-code.fd')
+        if not os.path.exists(bios_path):
+            self.cancel('Missing bios file "%s"', bios_path)
+        self.vm.add_args('-bios', bios_path)
         self.vm.add_args('-device', 'virtio-rng-pci,rng=rng0')
         self.vm.add_args('-object', 'rng-random,id=rng0,filename=/dev/urandom')

I can send a patch if this looks reasonable.

- Cleber.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]