Ben Greear wrote:
I finally had some time to debug this one a little more
thoroughly. On two different machines (Shuttle SB61G1) I
get the same results, so I do not believe it is bad hardware...
The bug is as follows:
I have 1 4-port tulip NIC in the machine. If I generate traffic
between two interfaces, it runs fine. But, if I start running traffic
on all 4 interfaces, the keyboard quits taking input, and ethernet
traffic stops on at least a few of the interfaces. I can still ssh
into the machine (via the rtl8139 interface), so at least one of the
processors (I'm using SMP on an P4 HT processor) is working. I also
enabled NMI and that does not trigger.
This was my bug. I was holding a lock that was required for receiving a packet
while calling the hard_start_xmit method. When an IRQ happened while I was in
the hard_start_xmit method, the IRQ could not grab the lock, and just sat there
spinning...
Thanks,
Ben
--
Ben Greear <greearb@xxxxxxxxxxxxxxx>
Candela Technologies Inc http://www.candelatech.com
|