Fwd: [Bug 10799] sky2 general protection fault

It's a good cable and switch port - it was used on a different machine with no
issues. According to the switch it's negotiated to auto/100. The hardware is
also working OK with the sk98lin driver. There may also be some extra log
messages because of me trying to remotely recover the system by resetting the
switch port. 

What debugging is needed to help here? I need to keep this system up, and so
far sk98lin is working, but I could run sky2 again for a while and use another
nic to remotely reset the interface.


I've attached syslogs from the dates I was running sky2

Jan 16 08:41:22 kernel: [    0.000000] Linux version 2.6.28-4-server
(buildd@vernadsky) (gcc version 4.3.3 20090111 (prerelease) (Ubuntu
4.3.2-2ubuntu11) ) #10-Ubuntu SMP Mon Jan 12 20:35:38 UTC 2009 (Ubuntu
2.6.28-4.10-server)
Jan 16 08:41:34 kernel: [    6.398381] sky2 driver version 1.22
Jan 16 08:41:34 kernel: [    6.398407] sky2 0000:01:00.0: PCI INT A -> GSI 17
(level, low) -> IRQ 17
Jan 16 08:41:34 kernel: [    6.398443] sky2 0000:01:00.0: Yukon-2 EC Ultra chip
revision 3
Jan 16 08:41:34 kernel: [    6.430561] sky2 0000:01:00.0: Marvell Yukon 88E8056
Gigabit Ethernet Controller
Jan 16 08:41:34 kernel: [    6.431102] sky2 eth0: addr 00:1b:fc:a0:a6:cb
Jan 16 08:41:34 kernel: [   15.547168] sky2 eth0: enabling interface
Jan 16 08:41:34 kernel: [   15.550225] ADDRCONF(NETDEV_UP): eth0: link is not
ready
Jan 16 08:41:34 kernel: [   18.444818] sky2 eth0: Link is up at 100 Mbps, full
duplex, flow control rx
Jan 16 08:41:34 kernel: [   18.444952] ADDRCONF(NETDEV_CHANGE): eth0: link
becomes ready
Jan 16 08:41:34 kernel: [   18.510127] sky2 eth0: Link is down.
Jan 16 08:41:34 kernel: [   21.399482] sky2 eth0: Link is up at 100 Mbps, full
duplex, flow control rx
Jan 16 09:02:18 kernel: [    3.921621] sky2 0000:01:00.0: PCI INT A -> GSI 17
(level, low) -> IRQ 17
Jan 16 09:02:18 kernel: [    3.921657] sky2 0000:01:00.0: v1.22 addr 0xfeafc000
irq 17 Yukon-2 EC Ultra rev 3
Jan 16 09:02:18 kernel: [    3.922000] sky2 eth0: addr 00:1b:fc:a0:a6:cb
Jan 16 09:03:39 kernel: [    3.910116] sky2 0000:01:00.0: v1.13 addr 0xfeafc000
irq 16 Yukon-EC Ultra (0xb4) rev 3
Jan 16 09:03:39 kernel: [    3.910190] sky2 eth0: addr 00:1b:fc:a0:a6:cb
Jan 16 09:03:39 kernel: [   15.345272] sky2 eth0: enabling interface
Jan 16 09:03:39 kernel: [   15.394455] sky2 eth0: phy write timeout
Jan 16 09:03:39 kernel: [   15.395482] sky2 eth0: phy write timeout
Jan 16 09:03:39 kernel: [   15.396508] sky2 eth0: phy write timeout
Jan 16 09:03:39 kernel: [   15.397534] sky2 eth0: phy write timeout
Jan 16 09:03:39 kernel: [   15.398560] sky2 eth0: phy write timeout
Jan 16 09:03:39 kernel: [   15.399582] sky2 eth0: phy write timeout
Jan 16 09:03:39 kernel: [   15.400606] sky2 eth0: phy write timeout
Jan 16 09:03:39 kernel: [   15.401630] sky2 eth0: phy write timeout
Jan 16 09:03:39 kernel: [   15.402653] sky2 eth0: phy write timeout
Jan 16 09:03:39 kernel: [   15.403676] sky2 eth0: phy write timeout
Jan 16 09:03:39 kernel: [   15.403778] sky2 eth0: ram buffer 1020K
Jan 16 09:03:39 kernel: [   15.405023] ADDRCONF(NETDEV_UP): eth0: link is not
ready
Jan 16 09:32:37 kernel: [    3.060427] sky2 0000:01:00.0: v1.21 addr 0xfeafc000
irq 17 Yukon-EC Ultra (0xb4) rev 3

Taking this out of bugzilla and onto netdev where it belongs

Why is the link bouncing so much. Does your switch not support autoneg?

If you see phy write timeouts, there are lower level problems. Check that the
firmware is up to date. You need to get that from the motherboard mfg, there isn't
a generic version. The firmware does a bunch of register settings that are design
dependent, so I wouldn't just take it from any other motherboard.