Hi guys,
There is a problem (I believe the _last_ problem) with the DK8-HTX's IRQs on the 8131's B-bus. I have two 3ware 9550's. If they're both on the A-bus, things work fine and the relevant slice of /proc/interrupts looks like:
CPU0 CPU1 CPU2 CPU3 25: 41 0 0 0 IO-APIC-fasteoi 3w-9xxx 30: 17 0 0 0 IO-APIC-fasteoi 3w-9xxx
If I move one of them to the B-bus (there's only one physical slot on B), then this becomes:
26: 36650 0 0 0 IO-APIC-fasteoi 3w-9xxx 30: 17 0 0 0 IO-APIC-fasteoi 3w-9xxx
The 36650 number increases by about 20000 or so all at once every few minutes. The timer appears to still be ticking away correctly. Accesses to the RAID on the B-bus immediately (deterministically; not 'once every few minutes') lead to a nobody cared that takes the card offline:
irq 26: nobody cared (try booting with the "irqpoll" option)
Call Trace: <IRQ> [<ffffffff8024ffec>] __report_bad_irq+0x30/0x7d [<ffffffff8025021a>] note_interrupt+0x1e1/0x223 [<ffffffff80250a33>] handle_fasteoi_irq+0x9e/0xc5 [<ffffffff8020a69c>] call_softirq+0x1c/0x28 [<ffffffff8020bbaf>] do_IRQ+0x7b/0xc8 [<ffffffff802088b4>] default_idle+0x0/0x47 [<ffffffff80209a91>] ret_from_intr+0x0/0xa <EOI> [<ffffffff802088dd>] default_idle+0x29/0x47 [<ffffffff80208985>] cpu_idle+0x8a/0xad [<ffffffff8061a673>] start_kernel+0x202/0x207 [<ffffffff8061a15a>] _sinittext+0x15a/0x15e
handlers: [<ffffffff8800401e>] (twa_interrupt+0x0/0x5d8 [3w_9xxx]) Disabling IRQ #26
Interestingly enough, the behavior is the _same_ in the Iwill BIOS. I'm wondering if the LB routings are just a copy of broken Iwill routings? Can anybody think of anything to try other than concluding 'bad mobo/noisy wire'? (If a board swap is the next thing to try, that's possible but disruptive.) Bootlogs attached.
Thanks again,
-mcq
-
mcqmcqmcq＠fastmail.fm