I've tried your kernel over the night, but the problem of very bad network performance is still existing. Your patches have changed that the network is not totally dead after the watchdog trigger. After the boot the first ssh session was behaving completely normal, then i started 3 rsync jobs (rsync from /etc,/root and /boot to a rsyncd server). Normally they are completing in a few seconds or minutes, but now they are running hours (network throughput of one finished rsync: sent 10,640,117 bytes received 234,491 bytes 13,384.13 bytes/sec). Also the ssh session was beginnung to slow down. Again there are no network errors (netstat -i) reported, ethtool is showing a full duplex 1 GB link state. The rsync processes are not appearing in top. The CPU is totally unloaded (no iowait). Also after the rsync jobs were finished, the problem state is now still existing (after 10 hours uptime).