[Date Prev][Date Next] [Thread Prev][Thread Next]
[Thread Index]
[Date Index]
[Author Index]
RE: UP2000 crashes -- update
- From: "Greg Lindahl" <glindahl hpti com>
- To: <konrad pelimbert tssc univie ac at>, "RedHat AXP List" <axp-list redhat com>
- Subject: RE: UP2000 crashes -- update
- Date: Fri, 24 Dec 1999 23:43:00 -0700
> I found out that the guys using running that big number
> crunching program on the UP2000, which regularly caused the
> machine to commit suicide with "Too many files open", always set
> the stack size to unlimited. Now I tried setting the limit to
> 16M et voilá: the sucker's running stable for days now...
When it crashes, do you get these exceptions on the console and in
/var/log/messages:
RUC10_10.c0006.: Exception at [<fffffc0000317068>] (fffffc000031706c)
RUC10_10.c0006.: Exception at [<fffffc0000317068>] (fffffc000031706c)
RUC10_10.c0006.: Exception at [<fffffc0000317088>] (fffffc000031708c)
If you look in /var/log/messages it gives a better clue:
Dec 24 21:56:08 c21 kernel: RUC10_10.c0006.: Exception at
[setup_sigcontext+40/736] \
(fffffc0000316e2c)
Dec 24 21:56:08 c21 kernel: RUC10_10.c0006.: Exception at
[setup_sigcontext+48/736] \
(fffffc0000316e34)
Dec 24 21:56:08 c21 kernel: RUC10_10.c0006.: Exception at
[setup_sigcontext+64/736] \
(fffffc0000316e44)
Basically, setup_sigcontext is getting a pointer with illegal alignment.
I would guess that this frame sits on the stack?
Apparently the resulting memory clobber results in trashing the # of open
files. The files aren't actually opened; you can cd /proc ; ls */fds | wc -l
and see that they don't realy exist...
-- greg
[Date Prev][Date Next] [Thread Prev][Thread Next]
[Thread Index]
[Date Index]
[Author Index]
[]