> > There's no absolute guarantee of that, but that doesn't prevent making the 
> > common case optimal.
> 
> I'm not sure it would be more efficient. Even for small packets, you may have
> a mbuf chain (and this may be the common case, in fact).

If a small packet is in a long mbuf chain it could easily be a 'win' to copy
the fragments into a single buffer (of some form).  Indeed the 'gather'
DMA may not be also to copy with a large number of small fragments (due to
bus loading reading the descriptors) even if you have reserved enough
space for a descriptor per byte.

I have seen a sytem where an ethernet driver was given a 1500 byte message
that was in over 1400 fragments...
(and if it dropped it, the same packet would be presented again)

	David

-- 
David Laight: david@l8s.co.uk