On Mon, Sep 18, 2017 at 06:00:28PM +0100, Milosz Wasilewski wrote:
On 18 September 2017 at 17:25, Greg KH gregkh@google.com wrote:
On Mon, Sep 18, 2017 at 11:47:16AM +0000, Linaro QA wrote:
Summary
git repo: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable-rc.git git branch: linux-4.9.y git commit: 8d96ea41a3ee57a7a145054e57fe0fb1a5d19861 git describe: v4.9.50-79-g8d96ea41a3ee Test details: https://qa-reports.linaro.org/lkft/linux-stable-rc-4.9-oe/build/v4.9.50-79-g...
No regressions (compared to build v4.9.50)
No regressions is good, but why do we have so many actual failures overall? Seeing that we are failing 15 kselftests and 12 ltp syscall
The LTP tests fail mostly because of NFS. This only applies to x86 where we mount rootfs this way. These tests pass on HiKey
flock01 flock04 linkat01 open12 openat02 renameat201 renameat202 utime01 utime02 utime03
Why would NFS matter for these syscalls? Why are you using NFS at all? :)
Two more failed because there is not enough space available: sendfile09 sendfile09_64
How can you find that in the logs? I dug in there, but nothing seemed to be marked with the red "error" color, and then I think I found the error with a blue box of text surrounded by red, but then digging into that didn't result in anything that I could see.
Any hints on how to read the logs?
In these results I can see flock actually passing and ioctl03 failing. Naresh, could you take a look? I didn't see the bug for that in our bugzilla.
I did see the ioctl03 "failure" in the log (after digging as I said above), but that was pure luck, and I couldn't see anything other than "failure". Are the test logs stored anywhere?
I'm not 100% sure about kselftest but these failures are most likely due to incompatibility between test version (coming from 4.13) and kernel version. Not all tests can fail gracefully when the tested feature is not available.
Ok, but finding out which ones are failing would be good, to verify this.
tests would be good to know, and something that I want to get fixed for 4.9.y as I'm guessing this is not the case for 4.14-rc1?
For LTP the situation is very similar in terms of results. The only difference I can see is ioctl03 test (which Naresh will check shortly). With kselftest the results also look pretty consistent. There are a few more passes on 4.14-rc1 but I assume they're there because the tested features exist.
Getting LTP running would be great, and then adding in the "real" LTP would also be good, as I have caused some recent regressions that our internal testing caught weeks later. We are also starting to see people run VTS and report other real issues that LTP does catch...
thanks,
greg k-h