4.6 DRM/i915 update CFT (Sandy Bridge?)/IvyBridge/Haswell/Broadwell/SkyLake/KabyLake supported

[Matthew Macy]

 ---- On Mon, 23 May 2016 14:57:44 -0700 Juan Ramón Molina Menor <listjm at club.fr> wrote ---- 
 > Le 23/05/2016 20:46, Matthew Macy a écrit : 
 > > 
 > > 
 > > 
 > >   ---- On Mon, 23 May 2016 01:44:17 -0700 Juan Ramón Molina Menor <listjm at club.fr> wrote ---- 
 > >   > Hi! 
 > >   > 
 > >   > Your commits overnight have restored and even improved the performance 
 > >   > with my Haswell (HD 4400) GPU: Lower CPU usage and snappier resize / 
 > >   > full-screen changes. 
 > >   > 
 > >   > But on the other hand, loading the i915kms module through startx 
 > >   > (non-root) panics, I have to use kldload i915kms. 
 > >   > 
 > >   > Also, changing the power mode (plugging or unplugging the power cord) 
 > >   > leads to instant panic, while with the drm-next-3.9 branch, switching 
 > >   > AC/battery modifies backlight value up and down without issues. I have 
 > >   > replaced the screenshot: 
 > >   > 
 > >   > https://wiki.freebsd.org/Laptops/ThinkPad_S440?action=AttachFile&do=get&target=panic-DRM-46.jpg 
 > > 
 > > I'll need the backtrace from both panics. Neither of them occurs on my hardware. If you have a problem recovering a core dump try loading drm2 first and then setting dev.drm.skip_ddb=1. That will bypass much of the work that panic tries to do to quiesce the CPU before dumping core. 
 > > 
 > > -M 
 > > 
 >  
 > Thanks for your attention. 
 >  
 > Attached the screenshot resulting from the following: 
 >  
 > # kldload drm2 
 > # sysctl dev.drm.skip_ddb=1 
 > # kldload i915kms 
 > [plug / unplug the power cord  (the result is exactly the same)] 
 >  
 > panic: IMPLEMENT ME 
 > …then the backtrace: 
 >  
 > https://wiki.freebsd.org/Laptops/ThinkPad_S440?action=AttachFile&do=view&target=panic-DRM-46-AC-battery.jpg 
 >  
 > I have tried setting the same values for  
 > hw.acpi.video.lcd0.{economy,fullpower} but it still panics. 
 >  
 > Sorry for the ACPI noise in the image, that’s another story… 



Excellent. That tells me exactly what needs to be done. And the startx panic? If it's a race in I915 you'll still hit it doing the skip_ddb step first I'd think. I really should make that a tunable.

Thanks!

I probably won't have time to fix it today. Nonetheless, it should be fixed by the next update some time this week.

-M


 >  
 > Hope it helps, 
 > Juan 
 >