This box, becomes unresponsive to the network every few days, and when i go back there and plug in a monitor/keyboard it wont reboot, i have to unplug it. This box runs squid, djbdns (cache), and samba, and it's basically a nas/cache for my lan. Anyway. Basically, eth0 Broadcom BCM5721 via tg3 on irq 16, takes a crap and locks the box up. And i have no idea what to do to fix it. When this happens is frequent, but somewhat random, and is perhaps related to network load, for instance after 500-600 gb have been put through it, and it also seems to corrupt data, it may not be the nic, it could just be the fact that cache's are not flushed when i do a hard reboot. Reproducible: Always Steps to Reproduce: 1.Use the nic 2. 3. Actual Results: After a few days, a week, or even a few hours, the box takes a crap after then nic dies. Expected Results: To not crash! emerge --info , lspci -vv, dmesg (containing the nic cracking up)and /proc/interrupts will be attached.
Created attachment 125521 [details] emerge --info
Created attachment 125523 [details] cat /proc/interrupts
Created attachment 125524 [details] lspci -vv
Created attachment 125526 [details] dmesg containing the ugly stuff
Can you test with gentoo-sources-2.6.22-r1 please? If the problem persists, please add "pci=routeirq" to your kernel command line (ie. in your grub.conf).
This appears to not be a problem in linux-2.6.19-gentoo-r5 (gentoo-sources) I will give 2.6.22-r1 a try in day or so, i need to do further testing to make sure it's stable, and that it's not (in some amazing way) a hardware problem. The hardware in particular is a hp dgl 145 g2, i have several of these, but this is the only one i have that's not 2 way/4 way etc. (has 2 physical sockets and supports dual core opties)
I was wrong, problem still persists, on all tested kernels. Going to attempt to get hp's firmware updating rpm to work in gentoo, and attempt to flash the nics.
