We've had a spate of these lately - about 3 or 4 boxes. I have a feeling we just got a bad run of motherboards, but thought I'd try to get some feedback.
These are servers running with the Tyan Thunder K8SD (dual core AMD's, S2882 model). They have three onboard NIC's, two Broadcom gig NIC's BCM5704C's and an Intel 82551QM which is a 10/100 board. We use the Broadcom NIC's in this case with the tg3 driver. The kernel panics happen somewhat at random, though most frequently seen when scp'ing a large number of files (after a couple gigs they'll crash pretty reliably). Last night we had a box go down under very little load.
I couldn't copy and paste from the console or scroll up - the one thing I got that seems relevant is the following:
RIP <ffffffffa0092774> {:tg3:tg3_poll+177{ RSP <ffffffff8045a878>
The tg3 bit is why I suspect something with the network ports. Has anyone run into this? What's the best way to track this down - is there a setting where I can save a core file, or do I just need to hook up a terminal and leave it on 24-7? Thanks for any help.