RE: Tux and TCP

"Collins, Tom" <[email protected]>
Newsgroups gmane.network.tux
Message-ID <[email protected]>
Hello...

I have been looking at the source for TUX and its dependencies
on the TCP stack.  I am trying to remap/extract the TCP 
dependencies of TUX so that it is socket-based only, with tcp-specific
function calls/data references mapped to our own network api.

In an earlier reply to a question I had, Mr. Molnar stated that TUX
was not really dependent on the TCP stack, that any socket-based 
protocol should be O.K., and cleaning up TUX to get rid of TCP-only
dependencies should not affect performance in any way.

My question is in regards to the differences between tcp_send_page()
and send_sync_buf().  I know that the call to tcp_sendpage() is 
predicated on tux_zerocopy_header being true.  I have forced TUX to 
use the send_sync_buf call, and everything works fine.  Does the 
tcp_sendpage() afford zero-copy functionality, whereas send_sync_buf
does not?  Perhaps there is something in the tcp_sendpage that would
be worth our time and effort to implement in our own networking layer,
if there were substantail benefits to doing so.

So, my question is what is the difference in TUX behavior if I use the
send_sync_buf() call as opposed to the tcp_sendpage() call.  Is this
performance related (judging from an earlier email response -- excerpted
as the 2nd paragraph above, I think not)?  Is it a matter of
convenience? 

Thank you for your time

Tom



-------------------------------
This e-mail is for the sole use of the intended recipient and may
contain confidential and privileged information. Any unauthorized
review, use or disclosure is strictly prohibited. If you are not the 
intended recipient of this e-mail, please notify sender and delete 
all copies immediately.
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.