Menu
▾
▴
Re: [Audit-test-developer] current failures on z
|
From: Linda K. <lin...@hp...> - 2012-01-31 17:48:39
|
Linda Knippers wrote:
> Hi Debbie,
>
> Debora Velarde wrote:
>> Hi All,
>>
>> I am down to 5 tests failing on z in mls 64-bit mode. They are:
>>
>> 1. syscalls - clone
>> [38] clone perm=process_newns expres=fail user=test permtype=cap err=EPERM
>> tag=clone__cap_process_newns_fail_test
>> ERROR (2)
>> unexpected test result
>>
>> I think the syscall isn't failing as expected. Looking at the old IBM
>> tests, we only tested the success case and the following note:
>> * This test does not test the erroneous case, since the only two
>> * errors defined in the man page for clone() both involve
>> running
>> * out of system resources. Since running out of resources
>> would
>> * invalidate the test environment, we cannot produce the error
>> * condition.
>> Question: do we need to test the fail case for clone?
>
> Is the manpage for clone() different on zSeries? I see lots of failure
> cases documented in the manpage. The failure case we're testing is one
> where the NEWNS flag, which requires CAP_SYS_ADMIN, is attempted by
> a non-privileged user, so the call should fail with EPERM. That was
> probably new with RHEL5 so if your test cases are from an earlier
> evaluation, they might have overlooked some of the possible failure cases.
>
> Unless clone() or namespaces are different on zSeries, I'd expect the
> failure case to work. Stephan usually wants both a success and failure
> case.
>
> I wonder what you get if you just run 'do_clone newns' as a regular
> user outside of the test suite?
>> Here is part of the run.log file:
>> +++ /bin/su - testuser -c
>> '/usr/local/eal4_testing/audit-test/utils/bin/do_clone newns '
>> ++ [[ -z 0 ]]
>> ++ [[ -z 1646 ]]
>> ++ [[ -z 1626 ]]
>> ++ check_result fail 0 1646 EPERM
>> ++ declare suc=fail res=0 ext=1646 err=EPERM
>> ++ [[ -n EPERM ]]
>> +++ get_error_code EPERM
>> +++ gcc -E -dM /usr/include/asm-generic/errno-base.h
>> +++ grep EPERM
>> +++ awk '{print $3}'
>> ++ err=1
>> ++ case $suc in
>> ++ [[ 0 != 1 ]]
>> ++ exit_error 'unexpected test result'
>> ++ declare status
>> ++ [[ -n unexpected test result ]]
>> ++ [[ unexpected test result != *[!0-9]* ]]
>> ++ status=2
>> ++ [[ -n unexpected test result ]]
>> ++ printf '\nexit_error: %s\n' 'unexpected test result'
>>
>> exit_error: unexpected test result
>> ++ exit 2
>>
>>
>> 2. misc - securetty
>> ERROR - login attempt failed when it should have succeeded.
>> Looking at the test, it looks like it finds tty1 in /etc/securetty. It
>> doesn't look like /dev/tty1 exists.
>> I see this in the run.log:
>> + ./do_tty /dev/tty1 'exit\n'
>> open: No such file or directory
>> + sleep 2
>> + return 1
>> + '[' 1 -ne 0 ']'
>> + echo 'ERROR - login attempt failed when it should have succeeded.'
>> ERROR - login attempt failed when it should have succeeded.
>
> What do tty devices look like on your system, assuming you have tty
> devices? If you need a different device, we can probably change
> the securetty test to either take the tty device name as an argument
> and conditionalize the run.conf or just conditionalize the test case.
>
>> 3. trustedprograms - screen_locking[8]
>> [8] screen_locking
>> FAIL
>> screen clear before locking failed
>
> I don't know if this is related but the test case says something
> about 'kernel boot options contain "no-scroll" and "fbcon=scrollback:0'
>
> I saw something in Stephan's kickstart mail from this morning about
> having to make that work for zSeries. Do you have that boot option?
I'm pretty sure this failure is due to the missing boot option. With
Stephan's latest KS script, the option is missing for all platforms
and this test failed for me until I added the rebooted with the boot options.
-- ljk
>
>> I do still see /sbin/consoletype: Permission denied message. Looking at
>> the output in run.log, I think this could have something to do with why it
>> is failing:
>> + chown testuser:testuser /home/testuser/screen.exp
>> + /bin/su - -c 'expect /home/testuser/screen.exp' testuser
>> spawn screen^M
>> ^[[?1049h^[)0^[[4l^[[?1h^[=^[[0m^[(B^[[1;24r^[[H^[[J^[[H^
>> [[Jbash: /sbin/consoletype: Permission denied^M
>> [testuser/lspp_test_r/SystemLow@lnxccc1 ~]$ ^[[?1l^[>^[[24;1H^M
>> ^[[?1049l^[[H^[[J^[[23B^[[3mUtmp slot not found -> not removed^[[23m
>> ^H^H^M ^[[H^[[44C
>> ^M
>> Screen used by <testuser>.^M
>> Password:^G^M
>> ^[[?1049h^[)0^[[4l^[[?1h^[=^[[0m^[(B^[[1;24r^[[H^[[J^[[H^
>> [[Jbash: /sbin/consoletype: Permission denied^M
>> [testuser/lspp_test_r/SystemLow@lnxccc1 ~]$ xit^M
>> logout^M
>> ^[[20B^[[3mUtmp slot not found -> not removed^[[23m ^H^H^M
>> ^[[4;1H^[[?1l^[>^[[24;1H^M
>> ^[[?1049l[screen is terminating]^M
>> + RET=0
>> + rm -f /home/testuser/screen.exp
>> + return 0
>
> Yeah, we need to get to the bottom of that consoletype problem.
> Do you see any AVCs?
>
>> 4. audit-remote - remote-disk_full_action syslog immediate
>> I believe I once saw this one pass when run individually. But now I
>> am seeing most of these tests fail. I need to get the others working again
>> before I can look into this one more.
>
> We talked about this on irc but you definitely need the time to be
> synchronized between the two systems. There's also my outstanding
> question about the profile.bash contents.
>
>> 5. LTP - msgctl11
>> I haven't started looking into this one yet.
>
> We've seen that fail in the past. Look in our ltp Makefile
> and you'll see that we set sysctl kernel.msgmni. Maybe on
> your system we need something different?
>
> -- ljk
>
>>
>> Thanks for any help/ideas!
>> -debbie
>>
>>
>> ------------------------------------------------------------------------------
>> Keep Your Developer Skills Current with LearnDevNow!
>> The most comprehensive online learning library for Microsoft developers
>> is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
>> Metro Style Apps, more. Free future releases when you subscribe now!
>> http://p.sf.net/sfu/learndevnow-d2d
>> _______________________________________________
>> Audit-test-developer mailing list
>> Aud...@li...
>> https://lists.sourceforge.net/lists/listinfo/audit-test-developer
>
|