OpenBSD, SMP and LWKT

Matthew Dillon dillon at apollo.backplane.com
Thu Apr 1 09:08:10 PST 2004


:As you might know, I've been meaning to get OpenBSD into the SMP realm, and to have better
:SMP support. I've had my eye on your little project for a while, and I must say that your LWKT
:ideas, and XIO seem very interesting. Henceforth, I will try to put your LWKT and XIO technology
:into OpenBSD, because it seems the right thing to do.
:
:So, if any of you would like to get funded for getting this into OpenBSD, just drop me a line, 
:and I'll see what I can arrange. I hope this collaboration will be fruitful.
:
:Greetings,
:Theo.
:-- 

    Hello Theo!  Hey, that's great.  I think the LWKT technology and approach
    is the way to go (at least compared to FreeBSD-5's model).  It's a much
    saner programming model to be sure.

    On XIO.. it's based on Alan Cox's target-side sf_buf work in FreeBSD-5
    and obviously very new in terms of how I intend to use it for other
    parts of the system.  I have had great success adapting it to our CAPS
    (ipc) code and the PIPE code (mainly Alan's work in regards to PIPE, but
    the XIO encapsulation of it was very easy to adapt and will be comitted
    today).  I think it's a winner, but the proof isn't going to be in the
    pudding until I am able to convert the buffer cache over to using it. 
    The buffer cache is the ten thousand pound gorilla, so to speak.

    There are still some issues with XIO, in particular how to handle 
    writable pages and copy-on-write issues, but it seems well on its way.

					-Matt
					Matthew Dillon 
					<dillon at xxxxxxxxxxxxx>





More information about the Kernel mailing list