Ati FireMV 2400 PCI card

[Alex Deucher]

On 3/14/07, Erwin Rol <mailinglists at erwinrol.com> wrote:
> On Sat, 2007-03-10 at 01:39 +0100, Erwin Rol wrote:
>
> > Now i made two xorg.conf files, one xorg.conf.0 with the device entry
> > for head 1&2 (the first M9) and one xorg.conf.1 with the device entry
> > for head 3&4 (the second M9).
> >
> > I wanted to try if they work separately and to my surprise head 3&4
> > worked when run on their own. So the card/chip was initialized
> > correctly. After that i tried;
> >
> > Xorg :0 vt07 -config xorg.conf.0 &
> > Xorg :1 vt07 -config xorg.conf.1 &
> >
> > And to my even bigger surprise that worked two. I got a all 4 screens
> > working. The mouse of course is in kind of clone mode, but i can start
> > applications and direct them to :0.0 or :1.0 and they shop up on the
> > correct monitors.
>
> It seems that what i said there was partly incorrect, I can start two X
> servers but the order has to be first 1 then 0. If i do it the other way
> around the start of the second X server will hang my machine. Putting
> the two configs in one file will hang the machine no matter what order
> the chips are initialized. So i first need to start, and wait until the
> screen is there, the server that uses the  Option "BiosLocation"
> "primary:0xc000" line, and than the one that doesn't. If i do it like
> that i get two working X server with 4 monitors in total with DRI
> working.
>
> The question is why does it hang when the order is not exactly like
> that ? Could it be that the Option "BiosLocation"
> "primary:0xc000" causes the full card to be reset, and so this has the
> be done first?  Any ideas on how to debug this?

Since the card has 2 chips on one card, perhaps they share the same
bios and that causes some funkiness if the ordering is not correct.

Alex