bad-ioctl test in ubuntu_stress_smoke_tests stop responding on F-xilinx ZCU104

[Po-Hsu Lin]

The ubuntu_stress_smoke_tests was skipped before sru-20240108 on ZCU104 because it does not have the required amount of memory (2GB < 3GB)

After we have lower the test limit from 3GB to 1.5GB in mid January [1], it's now getting tested.

In sru-20240108 (5.4.0-1037-xilinx-zynqmp) the test stops at sigbus test.
In sru-20240205 (5.4.0-1039-xilinx-zynqmp) the test stops much earlier at bad-ioctl test.

Other xilinx devices are not affected as they have 4GB memory.

Test log:
21:31:17 INFO | GOOD ubuntu_stress_smoke_test.bad-altstack ubuntu_stress_smoke_test.bad-altstack timestamp=1709587877 localtime=Mar 04 21:31:17 completed successfully
21:31:17 INFO | END GOOD ubuntu_stress_smoke_test.bad-altstack ubuntu_stress_smoke_test.bad-altstack timestamp=1709587877 localtime=Mar 04 21:31:17
21:31:18 DEBUG| Persistent state client._record_indent now set to 1
21:31:18 DEBUG| Persistent state client.unexpected_reboot deleted
21:31:18 DEBUG| Test has timeout: 2100 sec.
21:31:19 INFO | START ubuntu_stress_smoke_test.bad-ioctl ubuntu_stress_smoke_test.bad-ioctl timestamp=1709587878 timeout=2100 localtime=Mar 04 21:31:18
21:31:19 DEBUG| Persistent state client._record_indent now set to 2
21:31:19 DEBUG| Persistent state client.unexpected_reboot now set to ('ubuntu_stress_smoke_test.bad-ioctl', 'ubuntu_stress_smoke_test.bad-ioctl')
21:31:19 DEBUG| Waiting for pid 18824 for 2100 seconds
21:31:21 WARNI| System python is too old, crash handling disabled
(system disconnects here)
rsync: connection unexpectedly closed (0 bytes received so far) [Receiver]
rsync error: unexplained error (code 255) at io.c(235) [Receiver=3.1.2]

This should not be considered as a regression, as we're not testing it in the old days.

I have bad-ioctl test started manually on ZCU104 and it has passed with 5.4.0-1039-xilinx-zynqmp. And another attempt to run the whole test suite can get the same interruption when it hits the bad-ioctl test. Unfortunately there is no useful information from demsg / syslog.

We will need to dig deeper to figure out what's going on here.

[1] https://git.launchpad.net/~canonical-kernel-team/+git/autotest-client-tests/commit/?id=a1dcc85311066ecda1cb5533b49b2edbaf6244e5

[Po-Hsu Lin (cypressyew)]

I have this tested with the updated stress-ng (V0.17.06) as well and this issue still exist.