Wailer <banshee@gabriella.abattoir.com> wrote:
[ Subject: weird looping inet bug ]

> ppp0 is v.34 -> hostA
> ppp1 is isdn -> hostB
> 
> default route -> hostB
> ip addresses are a subset from hostA so the loop for packets from 
> machines other than the gateway is:
> 
> 	  ---> ISDN -> hostB --> NET
> [machine]
> 	  <--- v.34 <- hostA <-- NET
> 
> unfortunately this sometimes leads to terrible loops where the
> ISDN TD light is solid on and the v.34 RD is solid on.  Mbuf use
> shoots up and everything locks down.  It seems to be ftp from a 
> mac on the local net that locks everything like this.  Ideas?
> Packets look like this:

For what it's worth here are my current troubles with a completely
rebuild userland and kernel (i386) from April 10:
ppp0 to my ISP. mail -v someone@somemachine.domain where
somemachine.domain is reached via the default route.
The ending dot is printed on the screen and then the SD (send data)
light on my V.34 modem stays ON for no apparent reason.
A tcpdump -i ppp0 shows nothing and killing the sendmail process
doesn't help either ..........

Twice I've seen a related message in /var/log/message:
Apr  9 19:35:24 sheep /netbsd: rtinit: wrong ifa (0xf8838680) was (0xf8823700)
Apr 11 19:41:24 sheep /netbsd: rtinit: wrong ifa (0xf881e680) was (0xf881e700)

I first thought that changing the routed_flags in /etc/rc.conf to "NO"
solved my problems because a mail -v to my brother went fine, but after
that a mail -v to a friend showed the same scenario again.

I've also seen the same `SD light ON' behaviour when searching with
hotbot or lycos: after the first bytes of a reply to my search the
SD light goes on for no reason at all, etc., etc., etc.

Still have a 1.2C kernel from the end of February lying around which
is why you see this message.

Onno van der Linden (onno@simplex.nl)