Re: [LTP] [PATCH v3] mm/ksm05: new testcase

[Cyril H. <ch...@su...>]

Hi!
> >Commited, however it seems like half of the ksm tests are missing some
> >checks. On my workstation (which is obviously not a numa system) half of
> >them prints that numa system is required while the other half crashes
> >and segfaults.
> >
> Hmmm, ksm0[1-5] work well on my workstation (Gentoo + 3.0.0 / RHEL6);
> ksm01/05: normal ksm
> ksm02: ksm + NUMA
> ksm03: ksm + memcg
> ksm04: ksm + memcg + NUMA
> 
> Thus, non-NUMA system showing "numa system is required" for ksm02
> and ksm04 is expected.
> 
> What's the exact crash/segfaults info and which they occur in which tests?

ksm01

TBROK : unexpected signal 1 recieved (pid = 27438)
...

ksm03

TBROK : mount /dev/cgroup: errno=ENOENT
 and much more ENOENTS

ksm05

TBROK : open: errno=ENOENT(2): No such file or directory
TBROK : open: errno=ENOENT(2): No such file or directory
TBROK : open: errno=ENOENT(2): No such file or directory
TBROK : open: errno=ENOENT(2): No such file or directory
TBROK : open: errno=ENOENT(2): No such file or directory
TBROK : open: errno=ENOENT(2): No such file or directory
TBROK : open: errno=ENOENT(2): No such file or directory
...
Segfault

 - The infinite loop is actually caused by passing the
   cleanup function to tst_brkm() that is called from cleanup.


I suppose that at least ksm03 and ksm05 should return TCONF instead of
TBROK.

-- 
Cyril Hrubis
ch...@su...