TCP checksum errors occur naturally under VJ compression when an
earlier frame is dropped because it fails the PPP CRC check. The most
likely cause is dropping of received characters because of missed
interrupts. Disabling VJ compression will make the symptom disappear
but will not fix the underlying problem.
In particular, somewhere in the 2.1.92 to 2.1.106 transition, most
SCSI drivers were modified to be more SMP-friendly. The result was
that some drivers, in particular "fdomain.c", now disable interrupts
for extended periods of time. It's possible that "fdomain" has been
fixed---I submitted a patch to Linus about this---but I haven't
checked. However, there are probably other SCSI drivers with the same
problem.
Can I ask, are you using SCSI disks and, if so, does the number of
checksum errors noticeably increase when the disks are active?
Kevin <buhr@stat.wisc.edu>
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html