On Fri, Aug 28, 1998 at 11:47:39AM -0700, Curt Sampson wrote: > On Fri, 28 Aug 1998, Ted Powell wrote: > [...] > > IP: always defragment > > CONFIG_IP_ALWAYS_DEFRAG > > No, this just determines whether fragments are reassembled in > unusual circumstances. My complaint is not about that, but about > the fact that when you do any reassembly (either for this option, > or because the fragments you received were destined to your machine) > you do a full (and unnecessary) copy of the data.
You've inspired me to go digging. The comment on the routine that does
the deed, ip_glue, is (for those who haven't already seen it):
/*
* Build a new IP datagram from all its fragments.
*
* FIXME: We copy here because we lack an effective way of handling lists
* of bits on input. Until the new skb data handling is in I'm not going
* to touch this with a bargepole.
*/