From owner-freebsd-multimedia  Thu May 15 13:25:56 1997
Return-Path: <owner-multimedia>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.5/8.8.5) id NAA12663
          for multimedia-outgoing; Thu, 15 May 1997 13:25:56 -0700 (PDT)
Received: from compaq.parker.boston.ma.us (bparker.ne.highway1.com [24.128.47.60])
          by hub.freebsd.org (8.8.5/8.8.5) with ESMTP id NAA12658
          for <multimedia@FreeBSD.ORG>; Thu, 15 May 1997 13:25:54 -0700 (PDT)
Received: from sparc (sparc [204.253.111.2]) by compaq.parker.boston.ma.us (8.7.6/8.7.3) with ESMTP id UAA02908; Thu, 15 May 1997 20:25:50 GMT
Message-Id: <199705152025.UAA02908@compaq.parker.boston.ma.us>
To: Amancio Hasty <hasty@rah.star-gate.com>
cc: multimedia@FreeBSD.ORG
Subject: Re: questions about current bt848 code 
In-reply-to: Your message of "Wed, 14 May 1997 20:52:31 PDT."
             <199705150352.UAA23946@rah.star-gate.com> 
Date: Thu, 15 May 1997 16:25:50 -0400
From: Brad Parker <brad@parker.boston.ma.us>
Sender: owner-multimedia@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk


You've been so kind listening to my questions, I hesistate to ask more
but (what the heck!)

- I have a "working" port of the freebsd snapshot bt848 driver for linux.
The signal and mmap interface work fine, as well as the read interface
if I comment out the code to wait for a frame (and assume that the capture
ioctl has been called, as the meteor examples/single.c program does).

- Naturally I have an S3 card.  I have XFree 3.2 also.  I can't get
"direct" mode to work right.

If I use "fxtv -disableDirectV" I get live video (since that uses the
mmap interface) which looks ok, but has a little "snow" all over the
screen at regular intervals...  (reminds me of when we used to write
into the frame buffer during on-screen time and the non-dual ported
ram would have to arbitrate between the bus and the video and the
video would loose).

If I use "fxtv" I get a blank screen - but when I hit the middle mouse
button in the title bar part of the screen fills in (dont' ask me why).
I did some experiments mapping the frame buffer into kernel space and
using memset, etc... I the address being given to the dma engine is
right... it looks like the dma is stopping at random places.  (I assume
the pci bus master operating in the "physical address space", i.e. no
mmu mapping is going on out there).

Any ideas?  I'm wondering if the X server has the addresses space cached and
I need to disable the caching...

-brad