On Thu, Mar 22, 2007 at 01:01:23PM -0800, Bill Stouder-Studenmund wrote:
> Yes. This suggestion doesn't eliminate the data copy (DMA) from card to 
> mbuf, it eliminates a possible datacopy from mbuf to kernel socket buffer. 
> It also releases the mbuf sooner (if we are in the regime where we have 
> the data hang around in mbufs) because the to-userspace copy happens 
> sooner.

Now that I read sbcompress(), I see where copies may happen.  It sure
would be neat to eliminate those copies for readv(2); before & after
benchmarks may show that it is even useful. :-)

Dave

-- 
David Young             OJC Technologies
dyoung@ojctech.com      Urbana, IL * (217) 278-3933