>  I'm managing my machine over hme0.  I create the bridge, put hme0 in it, and
> bring it up (without ipf) and things are fine.  As soon as I put hme1 (which
> is ifconfig down, no IP address, no link) into the bridge my connections to
> the host hang.  This is NetBSD-current/sparc64 on an Ultra10.

 I put GATEWAY in my kernel.  With the bridge up I can kill the management
connections by toggling net.inet.ip.forwarding and net.inet.ip.checkinterface:

 Works:
  checkinterface = 0, forwarding = 0
  checkinterface = 0, forwarding = 1
  checkinterface = 1, forwarding = 1
 Fails:
  checkinterface = 1, forwarding = 0

>  My -current/i386 box works as expected....

 All four variations work fine on my /i386 laptop (with ex interfaces).  Except
when the laptop pagefaults in ex_intr, which it's done twice now.

> Could you check the interface flags that your hme* interface have in the
> broken state? The hme chips switches it's simplex behaviour depending on link
> type, and the bridge code has severe problems with !simplex interfaces.

 For my most recent test:
	UP,BROADCAST,NOTRAILERS,RUNNING,PROMISC,ALLMULTI,SIMPLEX,MULTICAST