Re: intermittent qtest-aarch64/xlnx-canfd-test test failure

[Cédric Le Goater]

On 9/21/23 09:51, Michael Tokarev wrote:
> Hi!
>
> While doing stable-8.1.1 preparation, I've a CI failure of 
> ubuntu-20.04-s390x-all here:
>
> https://gitlab.com/qemu-project/qemu/-/jobs/5132720046
>
> ▶ ERROR:../tests/qtest/xlnx-canfd-test.c:265:match_rx_tx_data: \
>    assertion failed ((buf_rx[size] & DLC_FD_BIT_MASK) == (buf_tx[size] & 
> DLC_FD_BIT_MASK)): (2281701376 == 2147483648)
>
>   qemu:qtest+qtest-aarch64 / qtest-aarch64/xlnx-canfd-test  \
>             ERROR           0.77s   killed by signal 6 SIGABRT
>
> Re-run of this test went successfully.  However, the assert looks
> a bit scary, and the fact that it does not fail in a reliable way
> is a bit troubling too.  Is it a flaky test or there's something
> else in qemu with concurrent threads?
>
> I don't see this assert in previous test runs.  But ubuntu-20.04-s390x-all
> test fails quite often due to other reasons, so it isn't conclusive.
>
> Should something be done with this?
>
> The testing is done for commit f2fc49c302036315db6e8c9f74592decc3be0476,
> which is in staging-8.1 branch only temporarily.
>
> Thanks,
>
> /mjt

I had the same CI error in job 
https://gitlab.com/legoater/qemu/-/jobs/5416150239
 
▶ 170/258 ERROR:../tests/qtest/xlnx-canfd-test.c:265:match_rx_tx_data: assertion failed ((buf_rx[size] & DLC_FD_BIT_MASK) == (buf_tx[size] & DLC_FD_BIT_MASK)): (2281701376 == 2147483648) ERROR
170/258 qemu:qtest+qtest-aarch64 / qtest-aarch64/xlnx-canfd-test           
ERROR           0.49s   killed by signal 6 SIGABRT
> > > QTEST_QEMU_BINARY=./qemu-system-aarch64 
> > > PYTHON=/builds/legoater/qemu/build/pyvenv/bin/python3 MALLOC_PERTURB_=60 
> > > G_TEST_DBUS_DAEMON=/builds/legoater/qemu/tests/dbus-vmstate-daemon.sh 
> > > /builds/legoater/qemu/build/tests/qtest/xlnx-canfd-test --tap -k
――――――――――――――――――――――――――――――――――――― ✀  ―――――――――――――――――――――――――――――――――――――
stderr:
**
ERROR:../tests/qtest/xlnx-canfd-test.c:265:match_rx_tx_data: assertion failed 
((buf_rx[size] & DLC_FD_BIT_MASK) == (buf_tx[size] & DLC_FD_BIT_MASK)): 
(2281701376 == 2147483648)
(test program exited with status code -6)

I could reproduce on PPC64 and s390x systems.

C.