[PATCH] Radeon Xpress 200M freeze fix

[Alex Deucher]

On Tue, Dec 15, 2009 at 10:28 AM, Matthias Hopf <mhopf at suse.de> wrote:
> ATI Radeon XPRESS 200M freezes on VT switch (often even during initial
> Xserver start).  It turned out to be a regression, which was introduced
> with git commit 7d9f643ae3d07 for fixing xorg bug 21321.
>
> It seems that on this chip you cannot reliably disable crtcs. I tried to
> work around that with usleep()s, but I didn't succeed fully. Patch 2+3
> are different attempts, both fix the issue somewhat, but after 5-60 VT
> switches (depending on phase of the moon or whatever) the system freezes
> again. Note that the freeze happens not at register access time, but
> sometime later, sometimes a second later. Increasing the sleep time
> doesn't help much.
>
> The attached Patch 1 disables crtc switch-off for the affected chips -
> it seems on these chips there is not much potential for power saving
> anyway, at least my measurements (when they were successful with one of
> my other patches) were pretty inconclusive.
>
> I'd like to see Patch 1 applied. Patches 2 and 3 are just for reference.
> If there is a different potential fix, I can reproduce the issue here
> and would be able to test patches.

Thanks for looking at this Matthias.   Based on your findings, I think
the issue is that the crtc must be enabled when  the crtc timing is
programmed since you didn't touch the dpms code which also disables
the crtc when the head is dpms off.  Also, in the last hunk of the
patch do you mean to force crtc1 or crtc2 on?
You have
+	save->crtc_gen_cntl |= RADEON_CRTC2_EN;
presumably it should be:
+	save->crtc2_gen_cntl |= RADEON_CRTC2_EN;

Does this updated patch work?  If so, I'll apply it.

Alex
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-fix-200M-freezes-on-VT-switch-if-CRTC-is-disabled.patch
Type: text/x-diff
Size: 1511 bytes
Desc: not available
Url : http://lists.x.org/archives/xorg-driver-ati/attachments/20091215/25973f0b/attachment.patch