On 3 January 2018 at 16:55, Greg KH gregkh@google.com wrote:
On Tue, Jan 02, 2018 at 09:50:02PM +0000, Linaro QA wrote:
Summary
git repo: https://android.googlesource.com/kernel/hikey-linaro git branch: android-hikey-linaro-4.4 git commit: 059105824d1a42b4a4e8879ce32ee678cc97c842 git describe: v4.4.108-568592-g059105824d1a Test details: https://qa-reports.linaro.org/lkft/android-hikey-linaro-4.4-aosp/build/v4.4....
Regressions (compared to build v4.4.107-568588-g587ae2c7e1ad)
hi6220-hikey - arm64: vts-test/arm64-v8a.VtsKernelLtp: * KernelLtpTest.syscalls.select04_64bit
* test src: not informedvts-test/armeabi-v7a.VtsKernelLtp: * KernelLtpTest.syscalls.poll02_32bit
LTP test case nanosleep01 and poll02 failed on OE userspace also.
* test src: not informedWhat does "not informed" mean?
That is default string from reports templates. Please ignore now for "AOSP builds".
And what broke here?
We have an open bug and the investigation is in progress. LKFT: 4.4, 4.9: Hikey: LTP: clock_nanosleep02 epoll_wait02 futex_wait05 nanosleep01 poll02 failed https://bugs.linaro.org/show_bug.cgi?id=3388#c21
Summary: The erroneous RCU_LOCKDEP_WARN() and the CONFIG_PROVE_RCU=y hided the fact that with the schedutil cpufreq governor, some of the timer related LTP fail on hikey620 with OPP_208MHz enabled. This issue can't be seen on juno r0.
Why this happens on hikey620 has still to be figured out ... Maybe the OPP_208Mhz delivers to little capacity for these testcases and the RCU_LOCKDEP_WARN() with CONFIG_PROVE_RCU=y in the wakeup path lets the system use this OPP_208Mhz less often.
- Naresh
thanks,
greg k-h _______________________________________________ Lts-dev mailing list Lts-dev@lists.linaro.org https://lists.linaro.org/mailman/listinfo/lts-dev