Symptoms were that we lost connectivity on the Gig-E port, things
were *really* slow accessing the 100bT port (different network), so
I grabbed the Serial console to try and troubleshoot. This very soon
stoped responding, so I had to power-cycle the box.
We'd been playing with the box all day testing a new configuration
(we are migrating a DB from an e4500 to this machine).
Here is the log file:
Nov 9 16:33:21 dbgsh01 kernel: Unable to handle kernel paging request at virtual address 8fab6104
Nov 9 16:33:21 dbgsh01 kernel: printing eip:
Nov 9 16:33:21 dbgsh01 kernel: c01087da
Nov 9 16:33:21 dbgsh01 kernel: *pde = 00000000
Nov 9 16:33:21 dbgsh01 kernel: Oops: 0002
Nov 9 16:33:21 dbgsh01 kernel: CPU: 1
Nov 9 16:33:21 dbgsh01 kernel: EIP: 0010:[do_IRQ+42/236] Not tainted
Nov 9 16:33:21 dbgsh01 kernel: EFLAGS: 00010093
Nov 9 16:33:21 dbgsh01 kernel: eax: cf792580 ebx: 000002a0 ecx: 0000011a edx: 0858ba10
Nov 9 16:33:21 dbgsh01 kernel: esi: c0303aa0 edi: 00000015 ebp: d9a63d1c esp: d9a63d08
Nov 9 16:33:21 dbgsh01 kernel: ds: 0018 es: 0018 ss: 0018
Nov 9 16:33:21 dbgsh01 kernel: Process xml_search_pull (pid: 2901, stackpage=d9a62000)
Nov 9 16:33:21 dbgsh01 kernel: Stack: 40156368 08597564 0858bff4 d7f2229d 081a28b0 bffff45c c010a8f4 40156368
Nov 9 16:33:21 dbgsh01 kernel: 0000011a 0858ba10 08597564 0858bff4 bffff45c 00000a4c c010002b 0000002b
Nov 9 16:33:21 dbgsh01 kernel: ffffff15 400b51d7 00000023 00010206 bffff454 0000002b ffffffff 0000000a
Nov 9 16:33:21 dbgsh01 kernel: Call Trace: [call_do_IRQ+5/13] [vsnprintf+991/1056] [reschedule_idle+533/540] [wake_up_parent+29/48] [do_notify_p arent+178/188]
Nov 9 16:33:21 dbgsh01 kernel: [__mmdrop+35/40] [do_exit+557/564] [sys_exit+14/16] [system_call+47/52]
Nov 9 16:33:21 dbgsh01 kernel:
Nov 9 16:33:21 dbgsh01 kernel: Code: ff 84 b8 30 3b 32 c0 f0 fe 8b 10 38 30 c0 0f 88 a3 60 14 00
Nov 9 16:35:58 dbgsh01 kernel: <5>nfs: server cart2-1 not responding, still trying
Nov 9 16:36:12 dbgsh01 kernel: nfs: server cart1-1 not responding, still trying
Hardware: VALinux 2200 dual CPUs, 2 gig ram.
Netgear GA620
IBM 9gig HD on /dev/sda (using Adaptec aic7896/97 Ultra2 SCSI adapter)
3ware 64xx card with 2 75 Gig IDE drives.
Kernel: 2.4.13-ac8. Using Acenic driver .83 for GA620. EEpro1.09 for
onboad 100bT.
I suspect that something caused the Acenic driver to, go bad.
Is this a case of "upgrade to the lastest version (IIRC .84) and
hope it doesn't happen again", or am I barking down the wrong
branch?
-- Share and Enjoy. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/