From mboxrd@z Thu Jan  1 00:00:00 1970
From: Christian Smith <csmith@micromuse.com>
To: 9fans@cse.psu.edu
Subject: Re: [9fans] Plan9 in VMware?
In-Reply-To: <20010213195620.0362919A05@mail.cse.psu.edu>
Message-ID: <Pine.LNX.4.21.0102132118140.9468-100000@erol.dev.net>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Date: Tue, 13 Feb 2001 21:28:03 +0000
Topicbox-Message-UUID: 621f8842-eac9-11e9-9e20-41e7f4b1d025

Is acceleration really that important to get bootstrapped? At least when
in a VESA mode, you can then develop any acceleration on top of it. Linux
has no problems running at useable speeds in VESA modes.

As far as your other reservations, VESA defines a protected mode interface
which requires only (from memory):
1. Permissions to the ports used to program the controller.
2. Valid ds, ss and cs segments.

Once the mode is set up, acceleration details can be creamed off from
XFree86 sources.

So long as entry to the interface is serialised by the plan9 VESA driver,
then SMP should not be a problem.

Christian

On Tue, 13 Feb 2001, Russ Cox wrote:

>	Grr, all plan9 needs is a VESA 2.0 video driver. I know someone has talked
>	about doing this (can't remember the reference,) and I intend to have a go
>	myself once I've got plan9 up and running. Then we can dump all this
>	reliance on bios signatures etc.
>
>	I have to say, the way plan9 handles this stuff is surprisingly archaic!
>
>I believe Roger Peppe did this for Inferno.
>I explored doing it for Plan 9 a month ago
>and got frustrated: by the x86 arcana needed
>to pull it off, by the prospect of needing to
>set up the graphics card in the boot loader
>before doing the 32-bit mode switch, by the
>blind faith you need to trust the video calls to
>come back to you, by the fact that the VESA
>interface doesn't handle any acceleration so
>you'd still have to obtain chipset manuals to do that,
>by the fact that nowhere does the VESA spec
>guarantee that it will work on multiprocessors.
>
>I agree that card identification using BIOS strings
>should go, in favor of something like PCI ids,
>but it looks like we'll be stuck writing card-specific
>drivers for a long while still.  Everyone else does.
>
>The hard part is not writing the drivers.  The hard
>part (as discussed before) is getting the manuals
>from the manufacturers.  Since acceleration requires
>this anyway, the VESA stuff just doesn't seem worth it.
>
>Please, prove me wrong.  I'd love to see it.
>
>Russ
>

--
    /"\
    \ /    ASCII RIBBON CAMPAIGN - AGAINST HTML MAIL
     X                           - AGAINST MS ATTACHMENTS
    / \