Subject: bridging vs altq
To: None <>
From: Sean Doran <>
List: tech-net
Date: 04/28/2002 20:58:00
Hi -

  I was pleasantly surprised to find that the in-tree ALTQ
works in -current, and have experimentally deployed an
i386 peecee with a couple ex(4) in a bridge(4) with CBQ
turned on in the face of a large overload of traffic from
one side of the bridge to the other (and beyond).

  AFAICT, with options HZ=1000, CBQ works OK in terms of
rate-shaping, however I am having problems with apparent
data corruption.  The altq.conf is not particularly
complicated, and only red is in use.

  scp (protocol 1) can perform bulk transfers that md5(1)
compare OK, but scp (protocol 2) complains about corrupted
MACs on input.  This does not happen with ALTQ off and
bridging on.

  It's intermittent enough that instrumentation to tcpdump
both sides of the bridge is difficult, so I thought I'd
ask here if this is a known problem (i.e., "it's broken")
or if this is the first anyone has heard of it (i.e., "check everything"),
or something in between.