af-alg stressor from ubuntu_stress_smoke_test failed on B-hwe-5.4 ARM64

[Po-Hsu Lin]

Issue found on ARM64 node wright-kernel with B-hwe-5.4

The test will timeout and eventually leads to sut-test in the end of the test.

 Running '/home/ubuntu/autotest/client/tests/ubuntu_stress_smoke_test/ubuntu_stress_single_smoke_test.sh af-alg'
 [stdout] Free memory: 126834 MB
 [stdout] Memory used: 114150 MB
 [stderr] /home/ubuntu/autotest/client/tests/ubuntu_stress_smoke_test/ubuntu_stress_single_smoke_test.sh: line 32: [: too many arguments
 [stdout] Using cgroup version 1
 [stdout]
 [stdout] Machine Configuration
 [stdout] Physical Pages: 32918979
 [stdout] Pages available: 32469363
 [stdout] Page Size: 4096
 [stdout] Zswap enabled: Y
 [stdout]
 [stdout] Free memory:
 [stdout] total used free shared buff/cache available
 [stdout] Mem: 131675916 853736 129877200 1120 944980 129983100
 [stdout] Swap: 9437176 0 9437176
 [stdout]
 [stdout] Number of CPUs: 96
 [stdout] Number of CPUs Online: 96
 [stdout]
 [stdout] Maximum bogo ops: 3000
 [stdout]
 [stdout] af-alg STARTING
 Timer expired (2100 sec.), nuking pid 22602

[Po-Hsu Lin (cypressyew)]

With V0.17.06 this test will bail out with badret:

$ sudo /home/ubuntu/autotest/client/tests/ubuntu_stress_smoke_test/ubuntu_stress_single_smoke_test.sh af-alg
Free memory: 126868 MB
Memory used: 114181 MB
/home/ubuntu/autotest/client/tests/ubuntu_stress_smoke_test/ubuntu_stress_single_smoke_test.sh: line 32: [: too many arguments
Using cgroup version 1

Machine Configuration
Physical Pages: 32919011
Pages available: 32478365
Page Size: 4096
Zswap enabled: Y

Free memory:
              total used free shared buff/cache available
Mem: 131676044 844068 129913460 1120 918516 129986500
Swap: 8388604 0 8388604

Number of CPUs: 96
Number of CPUs Online: 96

Maximum bogo ops: 3000

/home/ubuntu/autotest/client/tests/ubuntu_stress_smoke_test/ubuntu_stress_single_smoke_test.sh: line 109: cgexec: command not found
WARNING: cgexec fails, is tmpfs
cgroup
cgroup
cgroup
cgroup
cgroup
cgroup
cgroup
cgroup
cgroup
cgroup
cgroup
cgroup
cgroup working correctly?
af-alg STARTING
af-alg RETURNED 127
af-alg BADRET (unknown return status 127)

Summary:
  Stressors run: 1
  Skipped: 0,
  Failed: 0,
  Oopsed: 0,
  Oomed: 0,
  Passed: 0,
  Badret: 1, af-alg

Tests took 0 seconds to run

V0.17.08 will hang.

Next is to install the cgroup-tools to see if it helps.

[Po-Hsu Lin (cypressyew)]

V0.17.06 can pass with cgroup-tools installed.

Bisecting V0.17.06 and V0.17.08

[Po-Hsu Lin (cypressyew)]

Bisect shows:

md32613839d3b69c0a1291353709f18bc6bb58a4e is the first bad commit
commit d32613839d3b69c0a1291353709f18bc6bb58a4e
Author: Colin Ian King <email address hidden>
Date: Fri May 3 10:54:32 2024 +0100

    core-lock: add pause/yield in spinlock for architectures other than x86

    Signed-off-by: Colin Ian King <email address hidden>

:100644 100644 8680fb4118fe556dec03db247fe1ac818f56f115 228c5b8b73789deece6da86c8c57b968d9ac39fa M core-lock.c

I can see this failure on ThunderX ARM64 boxes with Mantic 6.5.0-35-generic as well.

Issue reported upstream: https://github.com/ColinIanKing/stress-ng/issues/393

[Po-Hsu Lin (cypressyew)]

An identical report can be found here https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2067527