Created attachment 67357 [details]
dmesg

First of all sorry for reporting this once latest drm-intel-next + xf86-video-intel works nifty. However this bug is critical for HP and all help are welcome.
The machine here has a hybrid GPU. When changing to intel one using i915 and fglrx X module and resuming it, it get frozen.
Even on pm_test=device it freezes.
I couldn't get any useful info neither with netconsole.

At this machine when using our tree I noticed that the bisecting point that fixes the s3-resume point is the 8d715f0024f64ad1b1be85d8c081cf577944c847 commit related to forcewake. However on 3.0.36-0.7.9.4471.1.PTF-default   from SuSE I'm using here Keith's patch is already applied although Ben's atomic patch is not there yet.

I also noticed that 
When using fglrx + i915 I just got this message when entering S3

[  103.127085] [drm:i915_driver_open],
[  103.217307] [drm:intel_crtc_cursor_set], [  103.217325] [drm:intel_crtc_cursor_set], cursor off

While using our driver I just can see one like that [drm:intel_crtc_cursor_set], cursor off

Using a debug=0xf there is no much difference except that using our code I see bunch of gem_close:
[18407.836342] [drm:drm_ioctl], pid=2827, cmd=0x40086409, nr=0x09, dev 0xe200, auth=1

And no one when using fglrx.
Kernel: 3.0.36-0.7.9.4471.1.PTF-default
X: xorg-x11-7.4-9.47.1
x11-video-fglrxG02-HP-8.982-1.1
xf86-video-intel: 2.17.0