An update on the current status of OpenBSD Symmetric Multiprocessor support was recently posted to the OpenBSD smp mailing list. At this time, it still looks to be quite a ways off in the future, with currently only i386-specific code that will detect and spinup a 2'nd processor, but not yet actually use it. Due to personal reasons, the effort has been mostly on hold for the past year, but a new synchronization effort between the SMP branch and the main kernel tree is tentatively aimed for next month.

Also addressed was the frequent question, "Why not just port SMP support from FreeBSD?" Niklas Hallqvist, the primay OpenBSD SMP developer, explained that there were too many differences between the two operating systems for this to be a possibility, though there was plenty of excellent code there to learn from. Instead, the project is pulling from NetBSD's SMP code [story], though this still is a major effort, tracking down each area that is different between the two kernels. Niklas explains that the primary issue is, "SMP support radically changes the system and is not a localized subsystem." He also discusses areas where the project can use coding help.

From: Niklas Hallqvist
To: smp
Subject: Status
Date: Fri, 21 Mar 2003 14:37:24 +0100

I have refrained from posting here because strictly, I have had
nothing to say.  Now as the occasional periodic flurry of messages
have surfaced once more, I feel I ought to say something.

Being the primary coder of the MP support of OpenBSD so far I can tell
you that the state is basically:

* All the working code there is, is in the SMP branch.  It is only
  i386-specific parts, no machine-independent parts have been worked
  upon.  The code manages to detect and spinup a 2nd processor, but
  not let it do any useful work.

* The SMP branch have not been synchronized with the trunk for over a
  year.  There are always troubles doing this synchronization due to
  CVS bugs which makes it a little tedious and boring.  A new sync is
  planned to occur in April, but do not count on it.

* We do have some non-working not committed code to deal with TLB
  shootdown.  This is as far as I can recall the final prerequisite to
  have the 2nd processor to do anything useful.  After that it is
  possible to do crypto off-loading or start on the MI parts of SMP.

* We don't need testers at this point.  When and if, we will ask for
  them.

* Anyone willing to help codewise, there is one thing noone currently has
  responsibility for.  Look for bug-fixes/changes in NetBSD's
  i386-specific MP-parts since the birth of the SMP branch of OpenBSD.
  Our work is based on theirs, and if they have identified problems,
  we likely have them and need fixes similar to theirs.  This is
  tedious and boring, sure, but if you want to help, this is
  valuable work.

* We are, after a year of no time at all, a few people in the team
  that see a few openings in our agenda to be able to at least start
  moving again.  This is not to say something will be visible for
  the foreseeable future.  But it means that you may be stepped on
  your toes if you try to do something major on your own.  If you have
  that kind of action in your plans, write down your design thoughts
  and send them to me, and I can stop you from reinventing the wheel
  in time.

I probably won't react on messages coming here in a while.  I still
have my hands full with real life, and cannot deal well with OpenBSD
issues at the moment.

Please remember that what Miod and Toby has tried to say, we set our
own priorities.  That what open source is all about, the freedom of
programmers to work on what they want, and share the results.  Open
source was never about anything else.  It's not about free software,
it's about software freedom, and that means requirements must not be
externally imposed, but come from ourselves.  Please help make
developers feel development is fun.  Disputes does not help that goal.

Niklas



From: Niklas Hallqvist
Subject: Re: Status 
Date: Fri, 21 Mar 2003 16:04:01 +0100

I got this question private:

> Call me uneducated,  but why can't for SMP support from Free be cloned?

In order to save on my time, I'll reply here:

Because SMP support radically changes the system and is not a
localized subsystem.  SMP support from NetBSD is possible to use,
because we are more similar to them, but finding all the thousands of
changes spread out all over the kernel source tree is tedious at best,
and you still need to think things through and understand them,
because details are different in the two systems.  SMP from FreeBSD is
not an easy fit into our kernel.  However, they have quite some good
code to be inspired by.

Part of the thing about us developers not posting here is that we
always get lots of followups to deal with, and we can't, sad to say.
Please respect our wish to not want to answer questions about things
like this.  Of course these question have a right to be posed, but not
here.  These questions are better done in misc@, I think. 
I do not at all want to be elitistic or condoning, but
believe me, it stresses me, just like Toby explained, when I almost
cannot manage life without OpenBSD.  I am deeply disturbed that I have
made next to no contributions at all the last year due to real-life
issues.  When I finally get to see the light in the tunnel, I want to
get down to coding business, not do meta-things like talk on
mailinglists, that just takes away coding time for me.

This mailing list was meant as a low volume list, for announcements or
tech. discussions, not a FAQ forum, and believe me, when we have
something to say, we'll say it.

Niklas

• Online Archive Of Above Thread