Hi, our QA has just encountered that testcases/kernel/mem/thp/thp04.c test case is failing in their testing environment. That made me look into what the test is doing actually and I have to say that I am quite confused about what is the objective here.
I think the test is simply misleading. You cannot expect 100% success rate for the THP pages. System memory might be fragmented and THP allocation might fail even when there is sufficient memory available. So from my perspective the test is simply bogus because its success rate highly depends on the current state of the system and a so the failure doesn't really mean a bug/misbehavior of the kernel. But before I suggest dropping this test I would have some questions. What has motivated the test case originally? What a failure would tell the tester? Under which conditions is the test supposed to be run? Fresh boot, idle syste, system under a light load? It is also not clear whether the testcase is supposed to test page fault THP path or khugepaded. I think it would make much more sense to test THP success rate with a different variety of workloads to see how well is the mm code able to cope with different levels of fragmentation and collect that data over time to potentially notice regressions. This however doesn't fit well inside LTP project which is stateless AFAIU (but my understanding is very limited). Thanks! -- Michal Hocko SUSE Labs ------------------------------------------------------------------------------ _______________________________________________ Ltp-list mailing list Ltp-list@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/ltp-list
