Fix for Theater200 on Radeon x600 VIVO

[Alex Deucher]

On 9/7/07, Maciej Cencora <m.cencora at gmail.com> wrote:
> Hi all,
>
> If you are interested only in the solution, scroll to bottom.
>
> My name is Maciej Cencora. I've been using Linux since couple of years
> and since mainly I'm a programmer (mostly PHP/HTML, but also C/C++) I
> decided to start contributing a month ago. At that moment I knew nothing
> about how to write a kernel driver or how radeon or theater chip is build.
>
> I decided to choose one of my i-work-only-on-windows devices. First was
> bluetooth usb dongle, but after kernel upgrade it started working.
> Second - scanner in Lexmark x1195 - hmm, not very interesting. So I've
> chosen Rage Theater on my X600 Pro VIVO HDTV(rv380) card (in xorg.log it
> showed no response from all VIP buses).
>
> I began with reading "Linux Device Drivers" e-book, and studying xorg
> ati driver source code. I wrote small kernel module, that was supposed
> to check if theater chip exists (almost ripped off from theater_detect.c
> :) Unfortunatly it didn't worked.
>
> Using SoftIce I was investigating 2 regs (VIPH_REG_DATA, VIPH_REG_ADDR)
> in Windows, and with help of theater_detect source code I found out that
> theater chip is connected to second VIP bus (nr 1). Then my module was
> torturing that bus in Linux, but no success. I thought that there are
> some timing issues, that I won't be able to deal with without a
> documentation. I was very disappointed, and almost were about to abandon
> this short linux developer adventure.
> But after a week I decided to try one more time, with hope that it's not
> a timing issue. Logically thinking if I got no response from any bus,
> maybe those buses are disabled by default to save power. So there must
> be some register that controls those buses and surprisingly its called
> VIPH_CONTROL ;) In Linux that reg contains value 0x003F0004, and in
> Windows it is 0x003F000D. And that's it! Change two bits, and it works:)
> The rest was piece of cake: find where to program that reg and recompile
> radeon driver.
>
>
> SOLUTION:
> Additional case statement in RADEONVIP_reset function (radeon_vip.c)
> with OUTREG(VIPH_CONTROL, 0x003F000D) (remaining OUTREGS copied from
> default case).

send me a patch and I'll push it upstream.

Alex