In February 2006 in comp.os.linux.misc I wrote:
> I just got a whole bunch of these messages:
>
> eth0: Out-of-sync dirty pointer, number1 vs. number2.
>
> Where number1 and number2 don't match. They're each about 25000,
> and they mismatch by about 15. Too much of the time I also get a
> kernel oops right after this. [At that point the system locks up
> hard, requiring a power-off reset.]
>
> This happens on a 2.6.7 kernel equipped Celeron machine using a
> RealTeK RTL8139 equipped network card.
A number of people in USENET suspected that this was hardware,
and I never did manage to find any hardware troubles. I've since
discovered that this problem ONLY occurs if I happen to have the
Cisco IPSec VPN client running at the time. I'm running client
version 4.8.00.0490.
I've since tried the Cisco client on a dual-core Athlon machine
running a 10/100/Gigabit Ethernet VIA VT6122 on-board ethernet
adapter in an ABIT AV8 3rd-Eye motherboard. This system is
running Mandriva 2006.0, Christmas Club edition, with a security
upgraded kernel to version 2.6.12-27mdk. While this system never
locks up hard, the network service will die, and traffic to/from
everywhere is disabled. Restarting the "network" service brings
this system back to usable. There are no suspicious messages in
/var/log/syslog nor /var/log/messages.
In both cases the badness occurs in the event of heavy network
traffic, e.g. copying a 50MB file to or from a machine.
Has anyone run into something like this? How about how to fix
it or work around it?
Thanks....
--
PLEASE post a SUMMARY of the answer(s) to your question(s)!
Show Windows & Gates to the exit door.
Unless otherwise noted, the statements herein reflect my personal
opinions and not those of any organization with which I may be affiliated.
|
Similar Threads
Linear Mode
