mremap01 is TBROK on ppc64le about ltp HOT 5 CLOSED

SIGBUS usually means that we attempted to access memory outside the mapped file but I don't see anything wrong with the test here. It opens a file lseek()s and write()s it in order to resize it, then mmap()s it, and then mremap()s it so that the mapping is twice it's size and writes to the memory. And then it fills the memory that corresponds to the mapped file which is likely the place where we get the SIGBUS.

The only thing that caught my eye is that the memory is mapped PROT_WRITE only, but we do not attempt to read from the mapping at all.

Is that problem 100% reproducible? Can you get a strace of the failure?

from ltp.

What kernel are you using? What filesystem is on TMPDIR?

from ltp.

Log: runltp.log.txt
Kernel is 4.4 (from SLES12SP2 RC2), heavily patched. There's ext4 on the loopback connected to TMPDIR.

from ltp.

@mnowaksuse any chance to get the strace from the failing testcase? Otherwise I doubt that we can figure out what is wrong there.

from ltp.

Haven't seen it fail for quite some time.

from ltp.