[LKP] 088fe47ce9: tests/openmp/openmp_sample: Segmentation fault

Eric W. Biederman ebiederm at xmission.com
Fri Oct 19 07:57:49 PDT 2018


Rong Chen <rong.a.chen at intel.com> writes:

> FYI, we noticed the following commit (built with gcc-7):
>
> commit: 088fe47ce952542389e604e83f533811750aaf7c ("signal: Add calculate_sigpending()")
> base:https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git  master
>
> in testcase: perf_event_tests
> with following parameters:
>
> 	paranoid: disallow_cpu_events
>
>
>
> on test machine: Intel(R) Xeon(R) CPU E5-2699 v3 @ 2.30GHz with 128G memory
>
> caused below changes (please refer to attached dmesg/kmsg for entire log/backtrace):
>
>
> * Checking OpenMP support
>   + tests/openmp/openmp_test
>     Testing OpenMP results...                                  PASSED
>   + tests/openmp/openmp_overflow
>       + tests/openmp/openmp_sample
>     * Checking proposed interface updates (not in any released kernel)
>
> To reproduce:
>
>         git clonehttps://github.com/intel/lkp-tests.git
>         cd lkp-tests
>         bin/lkp qemu -k <bzImage> job-script # job-script is attached
> in this email

I may dig into this more, but it looks like it will take a lot of
digging to make any sense of the test code, and I don't have the time
for that at the moment.

The patch you have pinpointed may change when a SIGSEGV is delivered but
it can not not cause a new SIGSEGV to be generated.

So I strongly suspect the test case made assumptions about the exact
time of signals that are no longer true, or that there is simply a race
in it's operation and the change caused that race to show up.

If the assumption shows up in production code rather than a test case I
would definitely consider even a race a regression that needs to be
fixed.  I don't know about this case.

Can you provide any help tracking down what openmp_overflow is doing
and why it is getting a SIGSEGV in a location it does not expect?

If this has uncovered a unreasonable race in the test the test needs to
get fixed.

Eric


More information about the LKP mailing list