My motherboard has a built-in ethernet port that is handled by the sk driver. I'm running FreeBSD 6.1-RC on amd64. When the network card is under high load, it often kills any current connections and leaves the message "sk0: watchdog timeout" in dmesg. I've seen previous posts on this issue, but the problem doesn't seem to be resolved. This issue makes these network cards unusable in production environments. Relevent dmesg output: skc0: <Marvell Gigabit Ethernet> port 0xa000-0xa0ff mem 0xfbb00000-0xfbb03fff ir q 17 at device 10.0 on pci0 skc0: Marvell Yukon Lite Gigabit Ethernet rev. (0x9) sk0: <Marvell Semiconductor, Inc. Yukon> on skc0 Any help would be appreciated, Miles
On Fri, May 05, 2006 at 01:03:55PM -0400, Miles Lubin wrote..> My motherboard has a built-in ethernet port that is handled by the sk > driver. I'm running FreeBSD 6.1-RC on amd64. When the network card is > under high load, it often kills any current connections and leaves the > message "sk0: watchdog timeout" in dmesg. I've seen previous posts on this > issue, but the problem doesn't seem to be resolved. This issue makes these > network cards unusable in production environments. > > Relevent dmesg output: > skc0: <Marvell Gigabit Ethernet> port 0xa000-0xa0ff mem > 0xfbb00000-0xfbb03fff ir > q 17 at device 10.0 on pci0 > skc0: Marvell Yukon Lite Gigabit Ethernet rev. (0x9) > sk0: <Marvell Semiconductor, Inc. Yukon> on skc0I had a similar event last night, on a P4 on an Asus P4P800. The current driver is much less prone to this lockup problem than it used to be. In my case it does not have to be in a high-load situation, it appears to happen rather randomly (and *very* infrequently) FreeBSD freebie.xs4all.nl 6.1-PRERELEASE FreeBSD 6.1-PRERELEASE #2: Thu May 4 22:37:22 CEST 2006 skc0: <3Com 3C940 Gigabit Ethernet> port 0xe800-0xe8ff mem 0xf7ffc000-0xf7ffffff irq 22 at device 5.0 on pci2 skc0: 3Com Gigabit LOM (3C940) rev. (0x1) sk0: <Marvell Semiconductor, Inc. Yukon> on skc0 sk0: Ethernet address: 00:0c:6e:4f:77:0c miibus0: <MII bus> on sk0 e1000phy0: <Marvell 88E1000 Gigabit PHY> on miibus0 e1000phy0: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, 1000baseTX-FDX, auto For the time being I stuck a em(4) in this machine, it is my primary server and I need to have it running for my work as re-builder for the Alpha platform :) Wilko -- Wilko Bulte wilko@FreeBSD.org
On Sat, May 06, 2006 at 12:10:46AM -0400, Miles Lubin wrote:> I know this is rather late in the release process, but given that this > issue will affect many people, I think it should be considered fixing > this driver for 6.1.The release is imminent. The last Ts are being crossed and Is dotted. mcl
On Sat, May 06, 2006 at 12:10:46AM -0400, Miles Lubin wrote: > On Fri, 5 May 2006, Bjoern A. Zeeb wrote: > > >On Fri, 5 May 2006, Wilko Bulte wrote: > > > >>I had a similar event last night, on a P4 on an Asus P4P800. The current > >>driver is much less prone to this lockup problem than it used to be. In > >>my > >>case it does not have to be in a high-load situation, it appears to happen > >>rather randomly (and *very* infrequently) > >> > >>FreeBSD freebie.xs4all.nl 6.1-PRERELEASE FreeBSD 6.1-PRERELEASE #2: Thu > >>May > >>4 22:37:22 CEST 2006 > > > >The updated driver has been in HEAD for some days and the problem > >should be fixed there thanks to Pyun. See last commits to src/sys/dev/sk/* > > > >-- > >Bjoern A. Zeeb bzeeb at Zabbadoz dot NeT > > > > > > > Is is possible and/or a good idea to just take the driver in head and > compile it with 6.1-RC? > If you have issues related with sk(4), you can take the driver in HEAD and compile, it should work on 6.1R. You may need to edit Makefile in order to reflect new location of sk(/sys/dev/sk) driver. > I know this is rather late in the release process, but given that this > issue will affect many people, I think it should be considered fixixing > this driver for 6.1. > > > Miles -- Regards, Pyun YongHyeon