[openbeosnetteam] Re: FreeBSD *netperf* stack port summer project anyone?

  • From: Nathan Whitehorn <nathanw@xxxxxxxxxxxx>
  • To: openbeosnetteam@xxxxxxxxxxxxx
  • Date: Fri, 15 Jul 2005 17:32:04 -0400

FreeBSD uses all-userland DHCP and just switched from ISC DHCPv3 to OpenBSD's fork of ISC DHCP. So, yes, they have one, but then again, ISC works on pretty much everything.

On Jul 15, 2005, at 5:16 PM, David Enderson wrote:


I think that sounds like a really good idea.

How does this affect the work I'm doing on a DHCP client?  Does their
stack that you are proposing porting have one?

Let me be clear: I'm happy to continue working on the DHCP client I'm
working on, and I am just as happy to use theirs if they have one and it
is better. I just want what is best for Haiku.

Just let me know what you want me to do.  I don't think I'll have time
to help with this port.

--David Enderson

-----Original Message-----
From: openbeosnetteam-bounce@xxxxxxxxxxxxx
[mailto:openbeosnetteam-bounce@xxxxxxxxxxxxx] On Behalf Of
Philippe Houdoin
Sent: Friday, July 01, 2005 11:21
To: openbeosnetteam@xxxxxxxxxxxxx
Subject: [openbeosnetteam] FreeBSD *netperf* stack port
summer project anyone?

Hi guys,

Indeed, it's really quiet.

At least, Nathan Whitehorn and I had some talks on #haiku IRC
channel recently about (again) porting their net stack code,
but the FreeBSD 5.x one because now it's multithreaded code.
The idea main objective is quickly resolve our main issues
with the current stack code:
- stability (proven one minus our own, smaller, code)
- maintenance (FreeBSD guys do it for us)
- features progress

You'll find more info about FreeBSD network stack "netperf"
project here: http://www.freebsd.org/projects/netperf

Robert Watson from the netperf project have some slides
presenting the project. It worth reading them:
http://www.watson.org/~robert/freebsd/netperf/20050513-bsdcan- netperf/

As I just start to look into their source code, it's too
early for me to comment in details the "what, who and how"
about such project. Anyway, I think it's my duty to throw
some plan at least:

a) Quick port. If it's not done after, lets say, 2
(man-)months of effort, we should drop the idea. That's why
we talked Nathan and I about a summer project: let's try to
do this during the summer time ;-)

b) Braindead port as possible: no redesign, no
modularisation, nothing except the mandatory stuff to make
run their code under BeOS and/or Haiku.

That mean we'll drop the protocols & interfaces as modules
design. Instead, let's put all into a single kernel module.
"fat" core module design, if you

/dev/net/stack <-> add-ons/kernel/network/freebsd_stack <->
/dev/net/* interfaces devices.

We must reuse some code from our current
network/interface/ethernet module to adapt it to bridge
FreeBSD interface API and BeOS/Haiku /dev/net/* network devices.

c) Limited userland stuffs impacted.

Let's focus on the stack internals first, right.

So, what do you think?
Want to join!?


- Philippe Houdoin

Other related posts: