procfs test in ubuntu_stress_smoke_test hang with F-gcp-5.4 on instance n2d-standard-2

[Po-Hsu Lin]

Issue found with Focal GCP 5.4.0-1093.102 this cycle (2022.10.10), on instance n2d-standard-2 only (we have another n2d instance n2d-standard-64 but it's working fine with this test on this kernel)

Reproducible rate: 6/6

stress-ng test suite HEAD SHA1: 346518c

If you just run the procfs test directly, it will pass within seconds:
$ time sudo ./stress-ng -v -t 5 --procfs 4 --procfs-ops 3000 --ignite-cpu --syslog --verbose --verify --oomable
stress-ng: debug: [1569] invoked with './stress-ng -v -t 5 --procfs 4 --procfs-ops 3000 --ignite-cpu --syslog --verbose --verify --oomable' by user 0 'root'
stress-ng: debug: [1569] stress-ng 0.14.06 g17935d0d1cf5
stress-ng: debug: [1569] system: Linux f-lgcp-gcp-5-4-n2dstd2-u-stress-smk-test 5.4.0-1093-gcp #102-Ubuntu SMP Mon Oct 24 11:36:45 UTC 2022 x86_64
stress-ng: debug: [1569] RAM total: 7.8G, RAM free: 7.2G, swap free: 0.0
stress-ng: debug: [1569] temporary file path: '.', filesystem type: ext2
stress-ng: debug: [1569] 2 processors online, 2 processors configured
stress-ng: info: [1569] setting to a 5 second run per stressor
stress-ng: info: [1569] dispatching hogs: 4 procfs
stress-ng: debug: [1569] cache allocate: shared cache buffer size: 16384K
stress-ng: debug: [1569] starting stressors
stress-ng: debug: [1570] procfs: started [1570] (instance 0)
stress-ng: debug: [1569] 4 stressors started
stress-ng: debug: [1571] procfs: started [1571] (instance 1)
stress-ng: debug: [1572] procfs: started [1572] (instance 2)
stress-ng: debug: [1573] procfs: started [1573] (instance 3)
stress-ng: debug: [1570] procfs: exited [1570] (instance 0)
stress-ng: debug: [1569] process [1570] terminated
stress-ng: debug: [1571] procfs: exited [1571] (instance 1)
stress-ng: debug: [1569] process [1571] terminated
stress-ng: debug: [1572] procfs: exited [1572] (instance 2)
stress-ng: debug: [1569] process [1572] terminated
stress-ng: debug: [1573] procfs: exited [1573] (instance 3)
stress-ng: debug: [1569] process [1573] terminated
stress-ng: debug: [1569] metrics-check: all stressor metrics validated and sane
stress-ng: info: [1569] successful run completed in 5.02s

real 0m5.315s
user 0m1.554s
sys 0m8.381s

But if you run the whole test suite, this procfs test will never finish. I guess this might be caused by some other test before the procfs test.

Steps:
  git clone --depth=1 https://git.launchpad.net/~canonical-kernel-team/+git/autotest-client-tests
  git clone --depth=1 https://git.launchpad.net/~canonical-kernel-team/+git/autotest
  rm -fr autotest/client/tests
  ln -sf ~/autotest-client-tests autotest/client/tests
  sudo apt-get install git python-minimal -y || sudo apt install python2 -y
  AUTOTEST_PATH=/home/ubuntu/autotest sudo -E autotest/client/autotest-local --verbose autotest/client/tests/ubuntu_stress_smoke_test/control

Look back into test history, I can see that:
  * cycle 2022.09.19 F-gcp:
    - stress-ng test suite HEAD SHA1: 346518c
    - 5.4.0-1090.98: all green
    - 5.4.0-1090.98+fips1: hang with procfs on n2d-standard-2, passed on other instances
  * cycle 2022.08.29 F-gco
    - stress-ng test suite HEAD SHA1: 346518c
    - 5.4.0-1089.97: hang with procfs on n2d-standard-2, n1-standard-8, t2d-standard-4, passed on other instances
    - 5.4.0-1089.97+fips1: all green
  * cycle 2022.08.08 F-gcp:
    - stress-ng test suite HEAD SHA1: 346518c
    - 5.4.0-1087.95: all green
    - 5.4.0-1087.95+fips1: all green
  * cycle 2022.07.11 F-gcp:
    - stress-ng test suite HEAD SHA1: 5239ae6
    - 5.4.0-1085.93: all green
    - 5.4.0-1085.93+fips1: all green