[Intel-gfx] 282dbf9b "drm/i915: Pass intel_plane and intel_crtc to plane hooks" breaks i945GM in 4.13 and above

[Ville Syrjälä]

On Fri, Jan 12, 2018 at 11:50:06AM +0200, Jani Nikula wrote:
> On Thu, 11 Jan 2018, Karl-Johan Karlsson <creideiki+intel-gfx at ferretporn.se> wrote:
> > Hello,
> >
> > I have an old Lenovo Thinkpad X60 laptop with a Core 2 Duo T7200 CPU and 
> > i945GM integrated graphics whose graphics broke when I tried to update the 
> > kernel to get the Meltdown fixes.
> >
> > GRUB2 shows a graphical boot menu, and Linux displays the Tux boot logos 
> > correctly at first. However, almost immediately after displaying the Tuxes, 
> > the actively scrolling area of the display shifts about two text lines 
> > downward, so that two decapitated penguin heads are left at the top of the 
> > screen and new text appears below the lower edge of the screen. When X starts, 
> > it too is shifted down the same amount, and only shows a black screen with a 
> > (working) mouse cursor. My display manager (SDDM) never shows up. Sometimes, 
> > while X is running, random pixels appear in the non-active area at the top of 
> > the screen containing the decapitated penguin heads.
> >
> > My hardware is:
> >
> > $ grep name /proc/cpuinfo
> > model name      : Intel(R) Core(TM)2 CPU         T7200  @ 2.00GHz
> > model name      : Intel(R) Core(TM)2 CPU         T7200  @ 2.00GHz
> >
> > # lspci | grep VGA
> > 00:02.0 VGA compatible controller: Intel Corporation Mobile 945GM/GMS, 
> > 943/940GML Express Integrated Graphics Controller (rev 03)
> >
> > # grep Chipset: /var/log/Xorg.0.log
> > [    33.085] (--) intel(0): Integrated Graphics Chipset: Intel(R) 945GM
> >
> > $ xrandr | grep primary
> > LVDS1 connected primary 1024x768+0+0 (normal left inverted right x axis y 
> > axis) 250mm x 180mm
> >
> > Kernel 4.12.14 does not have the bug. 4.13 and 4.14.11 have the bug in console 
> > and X mode. 4.15-rc7 has the bug in console mode, but crashes (completely 
> > black screen, Magic SysRq to reboot is the only thing that works) when 
> > starting X.
> >
> > Bisecting from 4.12 to 4.13 pointed to this as the first bad commit:
> >
> > commit 282dbf9b017bc6d5fdaeadf14e534c2fe22fee2d
> > Author: Ville Syrjälä <ville.syrjala at linux.intel.com>
> > Date:   Mon Mar 27 21:55:33 2017 +0300
> >
> >     drm/i915: Pass intel_plane and intel_crtc to plane hooks
> >
> > But I know nothing about the driver internals to explain why.
> 
> I'm afraid that bisect result doesn't seem plausible. Eyeballing through
> it, it's just straightforward mechanical non-functional changes.

The bisect reslut is actually correct. We have one or more bugs open
about this. https://bugs.freedesktop.org/show_bug.cgi?id=102929 at
least.

We need these commits to fix it:
51f5a0963984 ("drm/i915: Add .get_hw_state() method for planes")
b1e01595a66d ("drm/i915: Redo plane sanitation during readout")
5816d9cbc0a0 ("drm/i915: Fix deadlock in i830_disable_pipe()")

Now I recall that my plan was to get those into 4.15-rc but then
Christmas/etc. happened. Is there still a way to get those into
4.15 or has that ship sailed already?

-- 
Ville Syrjälä
Intel OTC