system lockup with RTL8139
Martin Ehmke
ehmke@technologist.com
Mon Sep 20 09:22:42 1999
On Fri, Sep 17, 1999 at 01:59:44PM -0400, Donald Becker wrote:
> On Fri, 17 Sep 1999, Ron Luse wrote:
>
> > I am running kernel 2.2.10, now with the 1.08f rtl8139 driver mentioned by
> > Donald in his reply to the above quoted message.
> ...
> > When the interface fails, the statistics from ifconfig reveal that there
> > was an RX over-run. I have loaded the driver with various levels of
> > debug, sometimes resulting in the system being MORE resistant to interface
> > hanging (particularly at debug > 4).
>
> More messages might be changing the timing, especially by causing more
> activity that allows the card to have more PCI bus bandwidth.
>
> > Some failures also give me the
> > error "Oversized Ethernet frame, status XXXXXXXX with varying values
> > of XXXXXXXX. In perusing the code, near the "Oversized" printf I see
> > the comment:
> > /* A.C.: The chip hangs here. */
> > Which seems to be exactly what's happening.
>
> The Realtek people say the chip doesn't hang with oversized frames.
> I haven't been able to reproduce the problem. I've been running
> continuous tests for two days on both SMP and uniprocessors systems.
>
> Perhaps under heavy PCI bus load something is happening that corrupts the Rx
> packet list. The status we are seeing (what are the exact numbers?) is
> actually packet data, or something of the sort.
>
> > Is this a known problem with the chip? Is there a work around possible?
> > Does anyone want more specifics of this failure? Can anyone offer any
> > tips on what I might try to fix this ? I have device driver experience,
> > but would appreciate any tips on where to start.
>
> I'll change the RxTooLong check so that it resets the Rx hardware...
> ftp://cesdis.gsfc.nasa.gov/pub/linux/drivers/kern-2.3/rtl8139.c
>
Hi,
I have tested the new version, compiling it into the kernel, but it seems
that it doesn't help anything to solve the network hangup problem.
But someone I know head the same problem and he solved it swichting the
"CPU to PCI write buf" (I dont know the exact name off the option) option
in the BIOS off and it worked fine. But in the bios off the machine where
the problem occures there isn't such a option. I think this information
could be usefull.
By the way, since I updated the rtl8139 driver included in the
2.2.9 kernel with your newest version of your driver, the
"oversized ethernet frame"-problem doesn't occure so often I think, but it
happens.
Thanks
Martin
| To unsubscribe, send mail to Majordomo@cesdis.gsfc.nasa.gov, and within the
| body of the mail, include only the text:
| unsubscribe this-list-name youraddress@wherever.org
| You will be unsubscribed as speedily as possible.