[TAG] ATI/OpenGL/swsusp/etc.

[Jimmy O'Regan]

Ben Okopnik wrote:

>For the past day or so, I've been experimenting with the latest
>incarnation of the ATI server that comes with Xfree86 - hurrah, finally!
>- I've been stuck with the proprietary "fglrx" server until now. One of
>the wonderful things this gives me is the ability to suspend the laptop,
>something that was not possible with "fglrx" and was my biggest
>complaint about this thing until now.
>
>The problem I'm running into is this: I can have GLX, or I can have the
>suspend feature. Not both. Suspending with GLX running works, actually;
>you can even see it progress through "waking up" - until it comes time
>to do that final "chvt", at which point the display partially or
>completely turns into a multicolored hash and freezes (nope, can't even
>SSH in - although the mouse will occasionally remain alive.)
>  
>
>For the moment, I've got two XF86Config files, one with 
>
>Load "dri"
>
>enabled, and one hashed out. I've also defined a 'startg' alias with an
>obvious function attached (for games.) My question is, does anyone know
>of a mechanism to load/unload X modules on the fly? This would be the
>*perfect* solution until this is fixed in the code. They're called
>"dynamically loadable modules", and this gives me hope.
>
>  
>
Have you looked here: http://cpbotha.net/dri_resume.html ?
"*NOTE3:* It seems that the swsusp code doesn't correctly restore all 
MTRRs: in your suspend/resume script, do something similar to what I do 
at the end of my script 
<http://cpbotha.net/clevo5600/scripts/suspend.txt>. Replace the 
addresses and sizes with what you see when you do a "cat /proc/mtrr" on 
your system. This is not crucial and you should only do it if you see a 
2D slow down during full screen repainting (at virtual desktop switches 
for instance). It also seems (from the changelog) that swsusp v17 is 
restoring MTRRs now. I haven't been able to test this yet."