Red Hat Bugzilla – Bug 447993
/kernel/security/audit/audit-test: syscalls tests fail on ppc64
Last modified: 2008-07-25 15:10:40 EDT
Description of problem:
PPC64 systems are seen to fail in a consistent manner via rhts:
These show both syscalls and filter test failures...
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. Run the aforementioned rhts test on i386
This may be related to the bz 444000 which is against the filter subtest w/in
the same parent test (audit-test.) As this is a consistent failure and we have
no data to say it is a test or kernel failure, I am marking it as a blocker to
prompt futher investigation.
Hey Josh -- would you mind updating this bz with the info we talked about the
As noted in the bug form, this is ppc
**Out of Memory Error
I looked at the first log in comment#0. It seems that augrok didn't find what it
was looking for. The logs seem to have it, though. So, I wonder if this is a
race between doing ptrace and having results on disk? If it is, maybe it just
needs a sleep added.
Is the ppc machines slow to where they need more time? Are they shared by
virtualization to where the kernel is not getting enough run time?
------- Comment From email@example.com 2008-06-26 17:18 EDT-------
I don't have access to view the links. Is this one of the audit filter testcases
we used for EAL4 certification?
------- Comment From firstname.lastname@example.org 2008-06-26 23:57 EDT-------
(In reply to comment #3)
> Hi Steve,
> I don't have access to view the links. Is this one of the audit filter testcases
> we used for EAL4 certification?
Can we get the logs posted somewhere we can get at them?
------- Comment From email@example.com 2008-06-27 11:24 EDT-------
At this point in time, the most likely cause of the problem appears to be a
timing issue in the Redhat Test Suite itself. I'm working with RH to work out a
test case to positively identify the cause of the problem. If anyone has a
nagging curiosity and just really wants to tear into this,
I've posted several complete sets of logs from the test results here:
------- Comment From firstname.lastname@example.org 2008-07-17 11:53 EDT-------
We've been kicking around discussion on this between Steve Grubb, Jeff Burke,
Peter M. and myself on this topic, and have not been able to conclusively prove
it to be an RHTS problem (or, for that matter, that it is not).
Mike, at this point, I think the next step is to find someone that can spend
half a day or so doing further testing. If we can grab an identical system/RHEL
level and show that >1 of the failing tests work outside of RHTS, we'll have
pretty good evidence that it is indeed a RHTS problem.
------- Comment From email@example.com 2008-07-17 12:06 EDT-------
do we have the rhts test suite at IBM?
The test suite within RHTS that is failing can be found here:
------- Comment From firstname.lastname@example.org 2008-07-17 12:38 EDT-------
It looks like the testcases Steve pointed to were developed by HP, not the ones
we created and shared with RedHat.
True, but they had been running fine.
------- Comment From email@example.com 2008-07-17 12:44 EDT-------
BTW, I just mention that the testcase was created by HP because I had thought we
could try to reproduce the problem by running our filter testcases with our
normal setup. But they are not the same testcases.
------- Comment From firstname.lastname@example.org 2008-07-17 13:02 EDT-------
is the system to that I should test with a js20?
------- Comment From email@example.com 2008-07-17 14:06 EDT-------
Mike, I've walked back thru some of the RHTS logs, and it looks like the
problems have been witnessed on a variety of systems, including:
Two different P5's: models 911051A, 9117570, as well as on a JS20
This seems to be further evidence the problem is in RHTS; I don't see how such a
widely impacting problem could have gone unnoticed for so long.
I don't see this issue anymore. Running the latest kernel on RHEL4.7 distro I
see the syscalls tests are passing now.
This bug should be closed "CURRENTRELEASE"
NOTE: There is still failures with PPC and the audit-test. Specifically the
filter/process_attrs test. But that should be a entered in as a new BZ.
Closing per jburke comments above