[LTP] perf_event_open03.c:95: TFAIL: Likely kernel memory leak detected

[LTP] perf_event_open03.c:95: TFAIL: Likely kernel memory leak detected

[Li Wang]

Hi Martin, All,

Our automatic jobs keep throwing thus false positives on the daily test.
After checking perf_event_open03.c, it uses the diff of MemAvailable as the
final condition for memory leaks.

    perf_event_open03.c:95: TFAIL: Likely kernel memory leak detected

I think relying solely on the MemAvailable metric to detect a memory leak
can be imprecise, because available memory can be influenced by various
factors unrelated to the specific code being tested.

And "/sys/kernel/debug/kmemleak" maybe a good tool for diagnosing memory
leak,
but it is usually disabled on the stock kernel by default, so far I have no
better idea
how to improve that, any suggestions?


    diff = SAFE_READ_MEMINFO("MemAvailable:");

    /* leak about 100MB of RAM */
    for (i = 0; i < iterations; i++) {
        ioctl(fd, PERF_EVENT_IOC_SET_FILTER, "filter,0/0@abcd");
        check_progress(i);
    }

    diff -= SAFE_READ_MEMINFO("MemAvailable:");

    if (diff > 50 * 1024)
        tst_res(TFAIL, "Likely kernel memory leak detected");
    ...


-- 
Regards,
Li Wang

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] perf_event_open03.c:95: TFAIL: Likely kernel memory leak detected

[Cyril Hrubis]

Hi!
> Our automatic jobs keep throwing thus false positives on the daily test.
> After checking perf_event_open03.c, it uses the diff of MemAvailable as the
> final condition for memory leaks.
> 
>     perf_event_open03.c:95: TFAIL: Likely kernel memory leak detected
> 
> I think relying solely on the MemAvailable metric to detect a memory leak
> can be imprecise, because available memory can be influenced by various
> factors unrelated to the specific code being tested.
> 
> And "/sys/kernel/debug/kmemleak" maybe a good tool for diagnosing memory
> leak,
> but it is usually disabled on the stock kernel by default, so far I have no
> better idea
> how to improve that, any suggestions?
> 
> 
>     diff = SAFE_READ_MEMINFO("MemAvailable:");
> 
>     /* leak about 100MB of RAM */
>     for (i = 0; i < iterations; i++) {
>         ioctl(fd, PERF_EVENT_IOC_SET_FILTER, "filter,0/0@abcd");
>         check_progress(i);
>     }
> 
>     diff -= SAFE_READ_MEMINFO("MemAvailable:");
> 
>     if (diff > 50 * 1024)
>         tst_res(TFAIL, "Likely kernel memory leak detected");
>     ...

Hmm, maybe we can sample the MemAvailable a few times, e.g. every 20MB
leaked and fail the test if we found that most samples have increased
the value.

-- 
Cyril Hrubis
chrubis@suse.cz

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp