On 9/11/2010 8:15 PM, Ben Hutchings wrote:
Please try this while running kernel version 2.6.32:
1. Enable additional error logging by running: ethtool -s eth0 msglvl 0x3f
2. Enable autoneg again
3. Disconnect and reconnect the cable
4. Check for any new error messages in the kernel log
Also send the output of each of these commands when autoneg is enabled
and the cable connected, under 2.6.30 and under 2.6.32:
1. ethtool eth0
2. ethtool -d eth0
3. mii-tool eth0
Sorry this took me so long, I didn't even see it until today. I was
thinking about the bug and wondering why I hadn't heard anything. I
checked the website, then searched my email, and there it was. Although
the system has been stable since I implemented the 100/Full workaround,
I do miss gigabit speed between local machines. I am now running the
following kernel:
linux-image-2.6.32-bpo.5-amd64 2.6.32-26~bpo50+1
When I enabled autoneg, I moved the cable to a switchport that's still
doing auto. The error logging was not very interesting, but hopefully
it'll be helpful. It repeats the same message every 10 seconds:
Nov 20 09:53:27 frodo kernel: [4443395.744137] eth0: PHY reset until link up
Nov 20 09:53:37 frodo kernel: [4443405.744132] eth0: PHY reset until link up
Nov 20 09:53:47 frodo kernel: [4443415.744132] eth0: PHY reset until link up
Nov 20 09:53:57 frodo kernel: [4443425.744132] eth0: PHY reset until link up
Nov 20 09:54:07 frodo kernel: [4443435.744133] eth0: PHY reset until link up
Nov 20 09:54:17 frodo kernel: [4443445.744805] eth0: PHY reset until link up
Nov 20 09:54:27 frodo kernel: [4443455.744805] eth0: PHY reset until link up
Nov 20 09:54:37 frodo kernel: [4443465.744805] eth0: PHY reset until link up
Output of the other commands you wanted, which also don't look very
helpful to me:
ethtool eth0:
Settings for eth0:
Supported ports: [ TP MII ]
Supported link modes: 10baseT/Half 10baseT/Full
100baseT/Half 100baseT/Full
1000baseT/Half 1000baseT/Full
Supports auto-negotiation: Yes
Advertised link modes: 10baseT/Half 10baseT/Full
100baseT/Half 100baseT/Full
1000baseT/Half 1000baseT/Full
Advertised auto-negotiation: Yes
Speed: 10Mb/s
Duplex: Half
Port: MII
PHYAD: 0
Transceiver: internal
Auto-negotiation: on
Supports Wake-on: pumbg
Wake-on: d
Current message level: 0x0000003f (63)
Link detected: no
ethtool -d eth0:
Unknown RealTek chip (mask: 0x28000000)
mii-tool eth0:
eth0: no link
After disabling autoneg, hard-setting 100/Full again, and moving it back
to the hard-set switchport, the system was having trouble. It was
similar to what happens when one side is hard-set and the other is auto,
except the packet loss was much worse. I could not see anything in
either the switch log or dmesg, but the switch statistics were
incrementing CRC errors. Ultimately a reboot fixed it.
Are you going to want me to install an run a special kernel version?
--
To UNSUBSCRIBE, email to debian-kernel-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/4ce80340.4010...@elyograg.org
