this issue is blocking us from shipping. bumping the importance to high+critical . please let me know if this is not appropriate.

thanks
-YK

Did you use the correct disable RC6 parameter in kernel: i915.enable_rc6=0?

yes..

The most likely cause is that the BIOS, at least partially, setup rc6, and we don't actually try to disable rc6 if passed enable_rc6=0 (just don't try to enable it). For testing purposes like this, it would be better if shutdown down rc6 completely.

Perhaps something like

diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
index 0a4bc51cebdc..7489da3fb9fa 100644
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -13470,6 +13470,7 @@ void intel_modeset_gem_init(struct drm_device *dev)
        intel_init_gt_powersave(dev);
        mutex_unlock(&dev->struct_mutex);
 
+       intel_disable_gt_powersave(dev);
        intel_modeset_init_hw(dev);
 
        intel_setup_overlay(dev);

added that but still get the gpu hang..

Really strange. Looks like BIOS issue.

But what is the output of /sys/kernel/debug/dri/0/i915_dp_mst_info
when it is enabled on BIOS and disabled on kernel with i915.enable_rc6=0?

Or better what is the output of intel_reg_dumper (of intel-gpu-tools)?

Created attachment 106124 [details]
register dump

/sys/kernel/debug/dri/0/i915_dp_mst_info is empty, register dump attached

the bios on my wilson beach is:

BDW-E2R1.86C.0072.R04.1405280639 05/28/2014

I verified it on Ubuntu 14.10 beta-1 (3.16.0-10-generic #15-Ubuntu) for Intel CRB with BDW-Y/U CPUs) by "i915_bdw.enable_rc6=0", and cat the value in /sys/module/i915/parameter/enable_rc6 is 0. if i915_bdw.enable_rc6=1, its value is also 1.

It seems to work well in drm/i915 driver layer.

I correct the information for #10 (with "i915.enable_rc6"), sorry!

I verified it on Ubuntu 14.10 beta-1 (3.16.0-10-generic #15-Ubuntu) for Intel CRB with BDW-Y/U CPUs) by "i915.enable_rc6=0", and cat the value in /sys/module/i915/parameter/enable_rc6 is 0. if i915.enable_rc6=1, its value is also 1.

It seems to work well in drm/i915 driver layer.

Right, I think this is indeed working properly with at least 3.16 and up. I can't reproduce #83207 myself anymore which was why this bug was opened in the first place, when rc6=0 didn't seem to make a difference but at least power consumption is up by ~2W (which I didn't test at the time) so it must work..