Subject: IP fragment circular lists.
To: None <tech-kern@NetBSD.ORG>
From: Darren Reed <darrenr@vitruvius.arbld.unimelb.edu.au>
List: tech-kern
Date: 05/14/1995 13:30:19
Upon looking at ip_reass(), there would seem to be some problems
with this routine.

My concern is that if fragments arrive out of order and/or overlap
it won't defragment them correctly.

The (what seems to me to be) a significant problem is that if the
first fragment queued isn't the first fragment and nothing can ever
be inserted in front of the first fragment received (without deque'ing
the queue and reque'ing it).

Can someone confirm that this is indeed actually true and what sort of
justification there is for this behaviour ?  This sort of thing could
impact on (for example) NFS performance significantly, I would imagine.

thanks,
Darren