I upgraded my kernel from 2.6.34-r12 to 2.6.36-r5, and now I see quite a few messages in /var/log/messages like this: Jan 3 12:47:23 scurvy kernel: [ 5071.845516] asix 1-2:1.0: eth1: asix_rx_fixup() Bad RX Length 1434 eth1 is a USB NIC (Trendnet TU2-ET100) that is supported by the CONFIG_USB_NET_AX8817X driver. When I boot into 2.6.36, I will see this error in /var/log/messages repeatedly, and ifconfig shows errors on the interface: scurvy ~ # ifconfig eth1 eth1 Link encap:Ethernet HWaddr XX:XX:XX:XX:XX:XX inet addr:XX.XX.XX.XX Bcast:255.255.255.255 Mask:255.255.248.0 UP BROADCAST RUNNING MULTICAST MTU:576 Metric:1 RX packets:92385 errors:418 dropped:0 overruns:0 frame:1 TX packets:25827 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:1000 RX bytes:14596881 (13.9 MiB) TX bytes:3488884 (3.3 MiB) This causes a lot of dropped packets, most noticeably preventing me from being able to VPN to my work. I found a discussion about a patch to my driver, and I wonder if it is what is causing this issue: http://help.lockergnome.com/linux/PATCH-usbnet-rx_process-ignore-packets--ftopict525698.html Reproducible: Always Steps to Reproduce: 1. Boot into 2.6.36-r5. 2. Attempt to establish network connections over the NIC. In particular, my VPN seems to cause this problem. I chose severity as critical, because it's a kernel driver bug.
its the same in 2.6.37. seems it's nothing to do with gentoo kernel, it seems to be upstream bug (same in vanilla). i see it for my DUB-E100 dlink usb2.0 100Mb/s card (ASIX AX88772). this works for me although I have no idea if its correct for this type of NIC: - default MTU for the card seems to be 576, same here in .37 and same in .34 where there where no errors etc. - ethernet standard is 1500, so it's how I set my card: ifconfig ethX mtu 1500 (or you can put it in your /etc/conf.d/net - no errors under ifconfig, seems to work for some time - i will update later if sth breaks again. if somebody would have anything to add here/correct then it would be nice. also i guess it would be good to open bug for kernel devs...
oh, well - it was annoying enough, so there you go: https://bugzilla.kernel.org/show_bug.cgi?id=30952 it contains a link to a proper workaround as well.