Subject: Re: FreeBSD Bus DMA (was Re: AdvanSys board support)
To: Justin T. Gibbs <>
From: Jonathan Stone <jonathan@DSG.Stanford.EDU>
List: tech-kern
Date: 06/11/1998 18:16:02

Please see the message I sent before this one.  
I think it outlines the technical areas of disagreement fairly
accurately.  If you find the "yes or no" format abrasive, my apologies.

>I think you need to become less confused first. 8-)

Not at all.  I think you need to look at non-x86 implementations of
NetBSD's bus_dma, and at non-SCSI uses.  Again, please see the messege
I sent just before this one.

>As "dm_segs" is provided to me as an argument to my callback, the 
>requirement that this information be stored in the dma map object
>is removed.

No, it isn't.  Without the clarification that "on a machine where bus
addresses are identical to sytem memory addresses", your claim is
easily shown to be false.  Several such machines exist, and the NetBSD
bus_dma API was designed to handle them.

Justin, I am reluctantly beginning to think that we haven't really
been communicating *at all*.  Unfortunately, I genuinely don't think
we are likely to, unless you let go of this notion that DMA the
segment list is an "S/G list".  That's only _part_ of what it can do,
and the semantics associated with it _ARE_ different from a SCSI
device's "S/G list".

This is not a putdown, Justin, they really _are_ different.

And the notion that I' not familiar with any NetBSD bus.h is an
insulting ad-hominem which was completely undeserved. I was quite
involved in the design discussions; Id have to check the logs, but I
think `heavily involved' is a fair description.

I would hope John Dyson has the integrity to condemn this piece of
arrogance, but somehow, I doubt it.