Bug 4764
| Summary: | On Running LTP test suite testcase nptl comes out with failures | ||
|---|---|---|---|
| Product: | Other | Reporter: | Sharyathi (sharyathi) |
| Component: | Other | Assignee: | other_other |
| Status: | REJECTED INVALID | ||
| Severity: | normal | CC: | akpm, bnpoorni, bunk, sglass, sripathik |
| Priority: | P2 | ||
| Hardware: | i386 | ||
| OS: | Linux | ||
| Kernel Version: | 2.6.12-rc6-mm1 | Subsystem: | |
| Regression: | --- | Bisected commit-id: | |
I can reproduce this all the way back to 2.6.10. Are you sure LTP isn't busted? I lied. 2.6.10 is OK. Will keep searching. This has been reported even in SLES 9 SP2 RC3 (2.6.5-7-187) with the latest version of LTP test suite of month of June(ie ltp-full-20050608) A bug has been filed and can be seen bugzill #16713 bug This is a problem with the LTP testcase. The test sets an alarm for 300 seconds (MAXTIME) and ASSUMES that the test will finish before that. If test is still running after 300 seconds, it assumes that the test is hung, so kills the test with failure information. Looks like 300 seconds is too less on most systems. So the test will fail on any slow system! Ideally testcase should not make such an assumption about when the test should end. It should use some other way to detect hangs. But the simplest way to fix nptl01.c is to set MAXTIME to some obscenely large value. I would set it to 2 hours! As described in the bug log, this seems to be a LTP problem and not a kernel bug. |
Distribution: SLES 9 SP1 ----------------------------- Hardware Environment: 1 way, Pentium IV 2.8GHz, 2G RAM Network Interface(e1000) Disk I/O SCSI storage controller: Adaptec Ultra320 ----------------------------- Software Environment: Linux x206h 2.6.12-rc6-mm1-I #3 SMP Sat Jun 25 13:27:57 IST 2005 i686 i686 i386 GNU/Linux ----------------------------- Problem Description: While running the test suite, test case nptl01 fails. nptl01 is a testcase to test the bug in system call pthread_cond_timedwait() of the native posix library. The test was supposed to verify a condition where the sequence counters were getting updated without holding the internal condvar lock. A FAIL is indicated by test hangin and not completing execution. On running the test individually test hung for more than 300 secs. -----------------------------