On 12/13/05, Benjamin LaHaise <[EMAIL PROTECTED]> wrote: > Hello Christoph, > > On Tue, Dec 13, 2005 at 07:29:36PM +0100, Christoph Hellwig wrote: > > Mid-term I plan to restructure the file_operations so that we don't need > > to have all these duplicate aio and vectored versions. This patch is > > a small step in that direction but also a worthwile cleanup on it's own: > > Good idea, but I think that we may want to do a bit more serious overhauling > of how socket read/write works. Right now the fast path case of networking > is pretty atrocious and winds its way through quite a number of layers > before any actual work is done. What I would like to do is something > along the following: have a protocol specified fast path (which can likely > be shared amongst many protocols) that provides a read() and write() > operation with the option for msghdr bits and flags. When it falls out of > the fast path, then we can incur the overhead of setting up the data > structures that would be shared with aio. I'm still hacking around on > af_unix right now, and will spend the time today to put together a bit of > a patch showing what I mean.
reducing the level of indirections is one of my long term goals with lots of the refactorings I did, idea is that eventually we can remove struct proto_ops and use struct proto (sk_prot) directly after more refactorings, I've did that in my private LLC tree and Aristeu Rozanski did some work on the x25 codebase, both dormant for now. I'll continue reducing the differences among DCCP and TCP and from this experience go on a storm on the legacy ones, SCTP being done after I'm confortable with the DCCP/TCP IPv4/IPv6 refactorings. - Arnaldo - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
