arm64: unhandled level 3 permission fault on ltp mmap16 test

[Catalin Marinas]

On Thu, May 19, 2016 at 10:14:17AM +0100, Suzuki K. Poulose wrote:
> On 19/05/16 06:34, fixed-term.Oleksij.Rempel wrote:
> >if i run latest ltp tests on any of latest kernel version i get
> >fallowing warning:
> >
> >root at rcar1:/opt/ltp# dmesg
> >root at rcar1:/opt/ltp# ./testcases/bin/mmap16
> >mmap16      0  TINFO  :  Found free device '/dev/loop0'
> >mmap16      0  TINFO  :  Formatting /dev/loop0 with ext4 opts='-b 1024'
> >extra opts='10240'
> >mke2fs 1.43-WIP (15-Mar-2016)
> >[  889.736679] EXT4-fs (loop0): mounted filesystem with ordered data
> >mode. Opts: nodelalloc
> >[  889.843995] mmap16[2164]: unhandled level 3 permission fault (7) at
> >0x7f8397d400, esr 0x9200004f
> 
> IIRC, this is expected. The test performs the access on purpose. It spawns
> a child which does this access and expects SIGBUS to be delivered to it. The
> parent makes sure that the SIGBUS is delivered to by wait()-ing for it.
> 
> See :
>                 /*
>                  * If child process was killed by SIGBUS, bug is not reproduced.
>                  */
>                 if (!WIFSIGNALED(status) || WTERMSIG(status) != SIGBUS) {
>                         tst_brkm(TBROK | TERRNO, cleanup,
>                                  "child process terminate unexpectedly");
>                 }
> 
> >Can some one please help me to understand this? If i see it correctly,
> >some thing wrong happened by level 3 address translation. But i have no
> >idea how to solve it (if it should be solved)
> 
> On arm64 we print unhandled faults to give enough information to debug the
> problem, which is why you see those messages in dmesg.
> 
> So, you can ignore those messages as long as the faults are generated by
> mmap16 and thetest reports success.

You can disable it with:

sysctl debug.exception-trace=0

(or via the sysctl files like /etc/sysctl.d/99-sysctl.conf)

-- 
Catalin