joerg@britannica.bec.de --> tech-net (2006-02-24 15:09:35 +0100):
> On Fri, Feb 24, 2006 at 03:03:02PM +0100, Jukka Salmi wrote:
> > I don't know whether wol(1) depends on that order or not, but I can
> > still reproduce the problem: running tcpdump while running wol on the
> > host I restarted today shows the magic packet (UDP to broadcast address,
> > port 40k); on the other host tcpdump doesn't show it. And I bet it
> > would be shown if I restarted the system...
> 
> Does it have an option to explicitly specify the source address?
> That might help.

It does help. At least I think I know now what's going on...

I can't specify the source address with wol(1), but I can specify the
target address. If I let wol(1) explicitly send the packet to the
correct broadcast address, everything works fine again.

By default, wol(1) sends the packet to 255.255.255.255 - what causes
trouble in my case because the interface's first address has a /32
netmask. Thus the packet was sent the local host... After restart,
when the first address had its correct /28 netmask again, this worked
again of course.

So, how do I correctly configure alias IP addresses? According to the
FreeBSD [1]handbook only the first address should have a real netmask,
and all aliases should be /32. But this seems to not work correctly
if the order of the addresses might change suddenly...


Regards, Jukka

[1] http://www.freebsd.org/doc/en_US.ISO8859-1/books/handbook/configtuning-virtual-hosts.html

-- 
bashian roulette:
$ ((RANDOM%6)) || rm -rf ~