Subject: README: New bus dma framework committed to NetBSD-current
To: None <current-users@NetBSD.ORG>
From: Jason Thorpe <>
List: current-users
Date: 06/06/1997 17:38:11
[ Sorry about the cross-post, but I want to make sure the word gets out! ]

Hi folks,

You may have noticed a flurry of activity lately, and especially today,
on the thorpej-bus-dma branch of the source tree.  Well, today that
code was pulled down into the public NetBSD-current sources.

This has been _several_ months of very hard work... I have been working
on the API spec since last year, and actively coding on the project
since mid-Februrary.  There were many issues to consider in the design
and implementation of the subsystem, and I must say, I'm very happy with
the way it turned out.  I am not aware of any other free OS that has done
anything like this before, and very few, if any, commercial OSes have done
it to this scale, if at all.

Anyhow, here is what this means to you:

----- i386 users -----

There is now COMPLETE support for ISA bounce buffers on Adaptec,
BusLogic, and UltraStor SCSI controllers!  If you have one of these
controllers, you will now be able to install > 16M of RAM in your
system!  Yes, about time, for sure.  But, this is the "proper way"
to do it that has been talked about on the mailing lists all this time.

----- alpha users -----

There is now support for SGMAP-mapped DMA!  This means a couple of

	(a) DMA (mostly) works on the ISA bus.  This has been tested
	    with an Adaptec 1542 in an Alpha AXPpci33 (thanks, Curt!)

	(b) This paves the way for using > 1G of RAM on the AlphaServer
	    8200/8400.  (This is of interest to all you people running
	    NetBSD on your TurboLasers out there!  :-)

Now, for the caveat... While ISA DMA is basically working on the Alpha,
there is a slight problem I'm still trying to find... After some time,
an unexpected machine check occurs.  I'm working with Curt right now
to find the problem, and will address it ASAP.  However, no "known working"
drivers on the Alpha are affected by this change.  After discussing it
with Chris Demetriou (the Alpha portmaster), it was decided that pulling
the code down now would be OK, and allows i386 users to take advantage
of everthing that it gains over there.

I will make another announcement to port-alpha when the bugs are
ironed out.


I would like to take this opportunity to thank the people who helped
test and debug the code:

	Juergen Hannken-Illjes
	Manuel Bouyer
	Bill Sommerfeld
	Michael VanLoon
	John F. Woods
	Tracy J. Di Marco White
	Brian Hechinger
	Scott Reynolds
	Dave Burgess
	Matthias Drochner
	Curt Sampson
	Enami Tsugutomo

Without their help, this task would have been MUCH harder.  Because of
these folks, this code is fairly well-tested.  They risked file-system
life-and-limb, and deserve a round of applause.  Thanks, folks!

Now, even though this has been well-tested, that doesn't mean that
there aren't bugs.  With that in mind, if you encounter any problems,
a bug report using send-pr(1).  I will address it ASAP.  I simply
cannot emphasize enough how important lots of details are.  Please
include as many as you can.

Ok, enough of the long-windedness... It's Friday evening, and I'm going
to go treat myself to a trip down to the pub... merging 105 files is a
bit of work!  :-)


Jason R. Thorpe                             
NASA Ames Research Center                               Home: 408.866.1912
NAS: M/S 258-6                                          Work: 415.604.0935
Moffett Field, CA 94035                                Pager: 415.428.6939