Date: Mon, 29 Oct 2001 12:40:27 +0000 From: Mike Bristow <michaelb@thus.net> To: FreeBSD-gnats-submit@freebsd.org Subject: kern/31586: netgraph bridges cause connectivity probs from bridge Message-ID: <E15yBiN-0000cK-00@singsing.eng.demon.net>
next in thread | raw e-mail | index | archive | help
>Number: 31586
>Category: kern
>Synopsis: netgraph bridges cause connectivity probs from bridge
>Confidential: no
>Severity: serious
>Priority: low
>Responsible: freebsd-bugs
>State: open
>Quarter:
>Keywords:
>Date-Required:
>Class: sw-bug
>Submitter-Id: current-users
>Arrival-Date: Mon Oct 29 04:50:03 PST 2001
>Closed-Date:
>Last-Modified:
>Originator: Mike Bristow
>Release: FreeBSD 4.4-STABLE-20011027T210032 i386
>Organization:
THUS PLC
>Environment:
System: FreeBSD singsing 4.4-STABLE-20011027T210032 FreeBSD 4.4-STABLE-20011027T210032 #0: Sat Oct 27 22:33:54 BST 2001 michaelb@singsing:/usr/obj/usr/src/sys/SINGSING-ENG i386
>Description:
Using a suitibly modified copy of
/usr/share/examples/netgraph/ether.bridge, I brought up a bridge
between de0, de1, de2, and xl0 (with xl0 being the "local interface".
The topology is :
[ singsing ] <-x de0 -> [ win2k box]
<-x de1 -> [ solaris box ]
<-x de2 -> [ utterly borked freebsd-current box ]
<-- xl0 -> [ summit 48 + rest of network, default route etc ]
where <-x interface -> indicates a crossover cat5 cable, and <--
interface -> indicates a normal bit of cat5
[ this is all done because of a lack of wall ports in the office ]
Machines attached to the de0, de1 or de2 have connectivity problems
from singsing, but they are reachable from other machines on the network.
A "snoop -o snoop.out host singsing and host solaris_box" run on
the solaris box while simultanously running "telnet singsing imap"
in another windown suggest that the TCP checksum is wrong (acording to
ethereal). I haven't "snooped" icmp echo requests/responses in
either direction yet; nor did I run tcpdump on "singsing".
Traffic out of the xl0 interface appears to be unaffected.
http://www.urgle.com/~mike/netgraph/ has my modified ether.bridge;
the snoop output as seen by the solaris box; and dmesg output from "singsing".
This problem has appeared between 4.4-STABLE-20011001T111411 &
4.4-STABLE-20011027T210032
(where the timestamp is the time "cvs update" was run against a local
CVS mirror that cvsups hourly)
If additional snoops/tcpdumps on assorted hosts would be useful,
or if anything else would help track this down, let me know.
>How-To-Repeat:
Set up a bridge using /usr/share/examples/netgraph/ether.bridge
and a LOCAL_IFACE; try and connect from the bridge to machines that
are attached to BRIDGE_IFACES but not LOCAL_IFACE
>Fix:
An evil workaround is:
for i in machines attached to de{0,1,2} ; do
route add $i $default_route
done
but this is obviously less than ideal.
>Release-Note:
>Audit-Trail:
>Unformatted:
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-bugs" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?E15yBiN-0000cK-00>