[Intel-gfx] bisected: i915 modeset broken in ac9b8236551d1177fd07b56aef9b565d1864420d

[Ville Syrjälä]

On Mon, Dec 14, 2015 at 03:31:09PM +0200, Meelis Roos wrote:
> Between 4.4-rc3 and 4.4-rc4, i915 modesetting broke on my i5-2400 PC. 

That would seem to be SNB.

> Instead of seeing the new dense graphics mode, I see the last VGA text 
> lines and no X appears either.

That's a bit weird. SNB has no power power wells, so only runtime PM
could be a factor, but it should not kick in that fast during boot even
if you enable it before loading the driver since we set the delay to 10
seconds.

And in any case the commit you list shouldn't really change anything
for SNB. We used to grab a rpm reference for gmbus via
intel_aux_display_runtime_get() and now we get it via the GMBUS power
domain instead.

I would be more tempted to blame the live status check, except it was
already included in 4.4-rc1 I think. We didn't have a power well
reference for it until
commit 0f5a9be15797 ("drm/i915: take a power domain reference while checking the HDMI live status")
but that's already in v4.4-rc4.

In any case the live status check was already reported to cause failures
(at least on some SNB IIRC). There have been attempts to remedy it since:
http://lists.freedesktop.org/archives/intel-gfx/2015-December/082689.html
http://lists.freedesktop.org/archives/intel-gfx/2015-December/082932.html

So this bisect result is somewhat mysterious. A full dmesg with
drm.debug=0xe with and without the offending patch reverted would be
helpful. And might be best to attach those into a bug report
(https://bugs.freedesktop.org/ -> DRI -> DRM/Intel) so that we don't
lose track of them.

Oh, are we even talking about HDMI/DVI here, or something else?

> 
> I saw something similar on I865G but have not had time to check if it is 
> the same issue.
> 
> ac9b8236551d1177fd07b56aef9b565d1864420d is the first bad commit
> commit ac9b8236551d1177fd07b56aef9b565d1864420d
> Author: Ville Syrjälä <ville.syrjala at linux.intel.com>
> Date:   Fri Nov 27 18:55:26 2015 +0200
> 
>     drm/i915: Introduce a gmbus power domain
>     
>     Currently the gmbus code uses intel_aux_display_runtime_get/put in an
>     effort to make sure the hardware is powered up sufficiently for gmbus.
>     That function only takes the runtime PM reference which on VLV/CHV/BXT
>     is not enough. We need the disp2d/pipe-a well on VLV/CHV and power well
>     2 on BXT. So add a new power domnain for gmbus and kill off the now
>     unused intel_aux_display_runtime_get/put. And change
>     intel_hdmi_set_edid() to use the gmbus power domain too since that's all
>     we need there.
>     
>     Also toss in a BUILD_BUG_ON() to catch problems if we run out of
>     bits for power domains. We're already really close to the limit...
>     
>     [Patrik: Add gmbus string to debugfs output]
>     
>     Signed-off-by: Ville Syrjälä <ville.syrjala at linux.intel.com>
>     Reviewed-by: Patrik Jakobsson <patrik.jakobsson at linux.intel.com>
>     [Cherry-picked from drm-intel-next-queued f0ab43e6 (Imre)]
>     Signed-off-by: Imre Deak <imre.deak at intel.com>
>     Link: http://patchwork.freedesktop.org/patch/msgid/1448643329-18675-3-git-send-email-imre.deak@intel.com
>     Signed-off-by: Jani Nikula <jani.nikula at intel.com>
> 
> :040000 040000 39379146d7e6dda8a4d5f8781ee3d307cce8c47e f4f09fae0485ad6263d31d425296fa9cd7de343b M     drivers
> 
> 
> -- 
> Meelis Roos (mroos at linux.ee)

-- 
Ville Syrjälä
Intel OTC