Bug 60573

Summary: Outputs still alive and scanning out as their backing memory is overwritten during KMS takeover
Product: Drivers Reporter: KernelBug (77777)
Component: Video(DRI - Intel)Assignee: Jesse Barnes (jbarnes)
Status: CLOSED CODE_FIX    
Severity: high CC: alan, chris, daniel, intel-gfx-bugs
Priority: P1    
Hardware: x86-64   
OS: Linux   
Kernel Version: 3.9 Subsystem:
Regression: No Bisected commit-id:
Attachments: config-3.9.9
config-3.10.1

Description KernelBug 2013-07-18 01:01:11 UTC
Created attachment 106915 [details]
config-3.9.9

Hi,

I'm running Slackware 14.0 and I noticed under both x86 & x64, that on around 8 different laptops from various makers, right when I was looking for a new laptop to buy and trying them out, every one of them using an Intel HD 4000 chip had this problem with the 3.9x kernels, and I"ve also noticed with my latest laptop the same problem under 3.10x

With 3.2x and 3.4x kernels I never saw this problem and I don't recall seeing it under 3.8x, but the problem might of begun there, not sure...

When the computers first boot the console for one second is 640x480 size, then it goes blank for a brief second, then the KMS kicks in. Right at the point it kicks in and starts to go into KMS, just for a brief second across about 1/4 of the screen at the top going all the way from left to right there's a flicker of the screen, typically in a hue of colors, it doesn't look good and it makes me feel like this could possibly cause screen damage over long periods of time, not sure...

Again, I do know for a fact this never happend in the 3.2x & 3.4x kernels, and it might of started at 3.8x, but it's certainly in all of the 3.9x and 3.10x kernels...

I truly hope something can be done to fix this...

I'm attaching the config files from the latest 3.9.9 and 3.10.1

THANKS
Comment 1 KernelBug 2013-07-18 01:01:39 UTC
Created attachment 106916 [details]
config-3.10.1
Comment 2 Daniel Vetter 2013-11-13 22:19:31 UTC
We occasionally scan out a bit of dirt when taking over the console, especially when coming from VGA mode. Should be harmless besides the eye-strain it causes ;-)

Jesse is working on patches to more smoothly take over the screen state.

If you can meanwhile please test latest kernels to make sure the bug is still there?
Comment 3 Jani Nikula 2013-12-12 18:40:17 UTC
(In reply to Daniel Vetter from comment #2)
> Jesse is working on patches to more smoothly take over the screen state.
> 
> If you can meanwhile please test latest kernels to make sure the bug is
> still there?

Reporter, please try the latest kernels.
Comment 4 Chris Wilson 2014-02-10 13:22:39 UTC
No need. Daniel's regression is still present.
Comment 5 Daniel Vetter 2014-02-11 10:27:13 UTC
Jesse has the patches for this.
Comment 6 Chris Wilson 2014-03-31 07:52:40 UTC
And they have been applied.