The ptests for mdadm are timing out rather than running: https://autobuilder.yocto.io/pub/non-release/20190529-5/testresults/testresult-report.txt I confirmed from the commandline that the tests do just hang, reason is unknown, its also not known when this started happening. I ruled out gcc8 verses 9 though.
The ptests for mdadm are timing out in 2.7.1 rc1 as well. https://autobuilder.yocto.io/pub/releases/yocto-2.7.1.rc1/testresults/testresult-report.txt
The below two cases make mdadm ptest hang: /usr/lib64/mdadm/ptest/tests/01r5integ /usr/lib64/mdadm/ptest/tests/01raid6integ Both cases have below logic to wait for the array to be fully optimal before proceeding. while ! (check state 'U*'); do check wait; sleep 0.2; done BTW, mdadm has lower pass rate in qemu env, it's better to run the mdadm test in the real board.
Fixed by https://git.openembedded.org/openembedded-core/commit/?id=b507793f4adde0c60588b06520d7d0b70acfd82a
QA team found mdadm timeout in qemuarm64 testresults below: https://autobuilder.yocto.io/pub/releases/yocto-2.8_M2.rc1/testresults/qemuarm64-ptest/testresults.json Log: PASS: /usr/lib/mdadm/ptest/tests/10ddf-sudden-degraded ERROR: Exit status is 9 DURATION: 736
seems the test 11spare-migration hangs on qemuarm64.
Grace, so are you going to debug the hang?
Any news Sandy?
I test on qemux86-64 and qemuarm64, mdadm timeouted with default cmd: ptest-runner mdadm. But it can completed if I set timeout longger: ptest-runner mdadm -t 6000 Below is result on qemuarm64: .... FAIL: /usr/lib/mdadm/ptest/tests/19repair-does-not-destroyFAILED - see //19repair-does-not-destroy.log and //fail19repair-does-not-destroy.log for details FAIL: /usr/lib/mdadm/ptest/tests/20raid5journalFAILED - see //20raid5journal.log and //fail20raid5journal.log for details FAIL: /usr/lib/mdadm/ptest/tests/21raid5cacheFAILED - see //21raid5cache.log and //fail21raid5cache.log for details DURATION: 1788 END: /usr/lib/mdadm/ptest 2019-08-30T06:19 STOP: ptest-runner @Randy @Richard As mingli also said before, "mdadm has lower pass rate in qemu env, it's better to run the mdadm test in the real board.", do you really need this test in PTESTS_SLOW? I am wondering if we only skip slow subcase or remove the whole mdadm test. if only skip subcase, I will skip 11spare-migration, I have test on qemuarm64, skip this one can make mdadm test not timeout
I don't object to skipping the slow test for 3.0 so you could send that for now. Best if you briefly explain what the 11spare-migration test is doing and report on how long it takes under qemuarm64. It would be nice to understand why it's so slow in case that reveals deeper problems. You could run the one test under [sl]trace -o /tmp/mdadm.trc -f -ttt ... to learn more today or under this defect during 3.1.
Please skip the test ASAP for M3 and we can downgrade the investigation to not block M3.
https://git.openembedded.org/openembedded-core/commit/?id=47c4ad9bd43d3b0539a19691bd4deb19ea63c3a7