On Wed, 2007-07-25 at 09:45 -0700, Bart Smaalders wrote: > read side I/O buffer pre-posting
I like this. there's also promise for making it possible to constrain the amount of kernel memory tied up in buffers "in flight" to userspace without hurting performance -- if your tcp (or other) window *is* the amount of unused buffer space the application has posted, the kernel will have a place to put every incoming byte before you tell the peer it's okay to send it. _______________________________________________ networking-discuss mailing list networking-discuss@opensolaris.org
