Re: Repeated hard crashes of 2.2.* system related to tcp_keepalive

From: Kurt Garloff (garloff@suse.de)
Date: Sun Feb 27 2000 - 15:20:11 EST


On Thu, Feb 24, 2000 at 11:19:23PM -0800, Andre Hedrick wrote:
> > Hard crashes of system related to tcp_keepalive
> >
> > [2.] Full description of the problem/report:
> >
> > The following frozen screen occurring between once a week and once a
> > day (with only the most minor differences, has appeared running kernels
> > 1.2.13, 1.2.14 and now 1.2.15pre9, through replacement of the memory,
> > motherboard, CPU, brand of NICs, video card, and drive cables (what's
> > constant now is the Tekram DC-390F SCSI controller, the IBM HD, a
> > Western Digital HD used for /boot and swap, and the case and power supply):
>
> Kurt Garloff can comment on the "Tekram DC-390F SCSI controller" and
> remove that from the things in question.

OK. Tekram DC390F is a ncr53c875 based UW SCSI host adapter. Can be driven
by either ncr53c8xx or sym53c8xx (or even a solution of Tekram: tmscsiw, but
I don't recommend that). Those drivers are known to work quite well.

The only trouble reported so far seems to be that the driver in very special
circumstances does not report a unit attention for unknown reasons (and the
driver may be completely irresponsible for it), so that a disk change
(applies to removable disks, obviously) is not getting known to the kernel.

> > >>EIP: c0177d5c <tcp_keepalive+e0/180>
> > Trace: c01780e6 <tcp_sltimer_handler+32/70>
> > Trace: c01780b4 <tcp_sltimer_handler+0/70>
> > Trace: c0111271 <timer_bh+331/388>
> > Trace: c0117991 <do_bottom_half+45/64>
> > Trace: c0106000 <get_options+0/74>
> > Code: c0177d5c <tcp_keepalive+e0/180> 00000000 <_EIP>: <===
> > Code: c0177d5c <tcp_keepalive+e0/180> 0: 8b 40 50 mov 0x50(%eax),%eax <===

.. and this points to trouble in the network stack.

-- 
Kurt Garloff  <garloff@suse.de>                          Eindhoven, NL
GPG key: See mail header, key servers         Linux kernel development
SuSE GmbH, Nuernberg, FRG                               SCSI, Security


- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Tue Feb 29 2000 - 21:00:18 EST