I have found the "can't allocate llinfo" error to arise when the arp
code tried to create a routing table entry to record that an arp was
sent (and eventually to hold the arp reply when it arrives later), but
the call to add the entry fails.  The add can fail due to there being
another entry in the routing table for the same address.  If, for
example, you put in a host route for a.b.c.d, then allocating an arp
entry would fail.

I have seen non-normal masks due to byte-swapping errors in ppp (on
older FreeBSD).  If it happens again, you might scrutinize 'netstat
-nr' from this point of view, and perhaps check the netstat source to
make sure non-normal masks would be accurately displayed.

If the address were outside the netmask for the cloning route, I would
expect that the arp code wouldn't be entered at all.

This failure can happen if you are trying to run ospf p2mp on
wireless, for example; I have not seen this happen in a wired network.

        Greg Troxel <gdt@ir.bbn.com>