Hamachi on alpha bad packets

Pete Wyckoff pw@dancer.ca.sandia.gov
Thu Mar 18 22:55:46 1999

I'd appreciate a sanity check from other hamachi users.  On alpha,
either kernel 2.2.1 or 2.2.3, I get lots of Tcp InErrs when running
bandwidth trials (ttcp).   They don't happen over a different ethernet
interface in the machine.  Would you look at the InErrs setting by:

	% egrep '^Tcp' /proc/net/snmp
	Tcp: [...] InErrs OutRsts
	Tcp: 0 0 0 0 4 0 0 0 0 166421 84174 0 149 0

and see if that next-to-last number goes away from zero on your
machines as it does on mine?

I generate the badness by running ttcp as such:

	% nttcp2 -l 16384 -n 32768 -b 60000 -r

while the other side sends packets.  Only the receiver sees errors.
I'm using a Becker stock-0.07 hamachi driver, with no checksumming or
modified parameters, on an alpha miata 433.  The machines are directly
connected.  Some combinations of length/buffersize seem more error-
prone than others.  Either machine, acting as receiver, generates
errors.  Swapping the fibre didn't help.

Instrumenting the kernel shows the errors are mostly due to PAWS-based
discards when receiving a time stamp earlier than one which had
previously been sent out.  A couple of plain bad checksums happen every
so often too.  These errors don't keep the transfer from completing,
but probably slow it down.

		-- Pete
Pete Wyckoff          | wyckoff@ca.sandia.gov
Sandia National Labs  | 925 294 3503 (voice)
MS 9011, P.O. Box 969 | 925 294 1225 (fax)
Livermore, CA  94551  |
 | To unsubscribe, send mail to Majordomo@cesdis.gsfc.nasa.gov, and within the
 |  body of the mail, include only the text:
 |   unsubscribe this-list-name youraddress@wherever.org
 | You will be unsubscribed as speedily as possible.