Re: trinity cores in watchdog while syslog'ing

Dave Jones <[email protected]> Wed, 6 Aug 2014 14:28:04 -0400
Newsgroups org.kernel.vger.trinity
Message-ID <[email protected]>
On Wed, Aug 06, 2014 at 08:16:27PM +0200, Toralf Förster wrote:
 > Just FWIW :
 > 
 > Fuzzying a 32 bit Gentoo Linux let the trinity process in side the UML guest sometimes core dump with this back trace:
 > 
 > Core was generated by `trinity -C 4 -N 100000 -x mremap -q'.
 > Program terminated with signal 11, Segmentation fault.
 > #0  __GI__IO_fflush (fp=0x8154200) at iofflush.c:40
 > 40      iofflush.c: No such file or directory.
 > (gdb) bt
 > #0  __GI__IO_fflush (fp=0x8154200) at iofflush.c:40
 > #1  0x080582c3 in synclogs () at log.c:163
 > #2  0x0805e93d in watchdog () at watchdog.c:363
 > #3  init_watchdog () at watchdog.c:430
 > #4  0x080538d9 in main (argc=8, argv=0xbfc8fd94) at trinity.c:132
 > (gdb) quit

ugh, when I switched the log file creation to happen from within
the child, I forgot to take care of this function.

I'm surprised it's taken this long for anyone to notice, as it can't
possibly work.

I think I'm just going to tear that whole thing out, as syncing
from watchdog context just isn't going to be possible.

	Dave