Comment 50 for bug 76489

It looks likely this is related to bug #326891 "2.6.27.11 kernel breaks r8169 support for rtl8102e". The fix for that issue was to introduce five patches that correct the handling of the various realtek devices. Some devices are recognised by PCI ID and some by their MAC, with work-arounds for the "unknown MAC" issue (reported in the dmesg log).

Apparently this caused code meant for the 8169 to be executed for other devices which shouldn't have been used I suspect maybe there is a transmit-side issue remaining to be patched.

Looking at all the upstream and other-distribution bug reports it seems to be a widespread issue.