On Thu, Sep 27, 2018 at 06:41:26AM -0300, Rafael David Tinoco wrote:
On 9/27/18 6:31 AM, Rafael David Tinoco wrote:
x15 - arm: ltp-quickhit-tests: * runltp_quickhit * select04
* test src: git://github.com/linux-test-project/ltp.git
Any hint as to what went wrong here?
I hope Naresh and the team can chime in with more information about that, but I just checked and the test did pass when ran again in the full test suite: https://lkft.validation.linaro.org/scheduler/job/439912#results_68482978
Just to be sure, I ran the sanity job again and all tests passed on X15: https://lkft.validation.linaro.org/scheduler/job/440164
select04 is a timing issue, already investigated: https://bugs.linaro.org/show_bug.cgi?id=3852 being worked in KV-114, together with other timing related issues.
Expectation is that timing issues will go away once we move from QEMU to KVM guests (for AARCH32 and AARCH64), having better context switching times (less VM_EXITS) and/or virtio drivers.
Forgot to say, for this particular issue, explanation on why this happens in x15 board is in here: https://bugs.linaro.org/show_bug.cgi?id=3852#c8 (few CPU cores can cause bad accuracy whenever there are just a few iterations for the deviation measurement).
This, sometimes, happen to timing tests (like RTC clock tests, also, failing with 64HZ frequency alarms because deviation is close to scheduler limit of 1ms per scheduled task).
That sounds like some messed up hardware :(
OBS: I usually do small comments, like this one, on all regressions coming from qa-reports@linaro.org. I haven't on this one since I saw this was yours only. Will do from now on, if it helps you.
It does, thanks.
greg k-h