Created attachment 96579 [details] dmesg Somewhere after 3.12, KMS has started to set garbage video mode with too large letters and vertical offset on Thinkpad X30 with Intel 830M graphics. 3.14-rc8 is still problematic. Bisecting is tricky - it seems the bug only happens with some kernels but not others from the buggy range. I did 2 bisections, one from 3.13 to 3.12 and one from first known-bad commits to 3.12 with lastest gcc 4.8 from Debian unstable. These lead me to differnet results that both reside in lustre commits, but lustre is not compiled for this laptop at all. Same kernel that exhibits the bug seems to exhibit it always. When X starts, it sets normal graphics mode and works with no problems. Attached are dmesg with drm.debug=0x0e, lspci -vvv, config and 2 bisection logs.
Created attachment 96580 [details] lspci
Created attachment 96581 [details] .config
Created attachment 96582 [details] bisect log 1
Created attachment 96583 [details] bisect log 2
Ville, you have a 830 right? I was going to check my 845g next week anyway, so I'll see if have something similar. Meelis, can you grab a photo of the distortion? It might help identify, if for example, it was an incorrect double-wide config.
Did you try restricting the bisect to drivers/gpu/drm/i915? Since it is a modesetting bug, it is likely to be there.
i.e. git bisect start -- drivers/gpu/drm/i915
Created attachment 96599 [details] Screenshot
Oops, forgot to attach the screenshot, now attached. Will try restricted bisect.
Bisect restricted to this path did not lead to an answer either: git bisect start '--' 'drivers//gpu/i915' # good: [5e01dc7b26d9f24f39abace5da98ccbd6a5ceb52] Linux 3.12 git bisect good 5e01dc7b26d9f24f39abace5da98ccbd6a5ceb52 # bad: [2c0514ee8a4a02650310eaffb378dd1fd08a7bb1] staging:lustre: Replace max() with max_t() git bisect bad 2c0514ee8a4a02650310eaffb378dd1fd08a7bb1 # good: [31d141e3a666269a3b6fcccddb0351caf7454240] Linux 3.12-rc6 git bisect good 31d141e3a666269a3b6fcccddb0351caf7454240 And this last git bisect good told me No testable commit found. Maybe you started with bad path parameters? So how else could I track this?
External merge commit? Odd. Lets do the basics and grab intel_reg_dumper before and after starting X.
Created attachment 96619 [details] Dump before X with distortion
Created attachment 96620 [details] dump after starting X and switching to VT1 (works)
Created attachment 96621 [details] Dump while running X (working)
Added 3 dumps - before X with distortion, after starting X and back to VT1 (no distortion) and while running X with no VT switch (working).
Initial fbcon is running at 8-bpp indexed pixel format, X switches to x888 xRGB format. This should be something to do with the pipeconfig reconstruction and inheritance from BIOS.
I observe exactly the same behavior with Slackware 14.1 (AFAIR kernel 3.10.7) on Fujitsu Lifebook S-6010. In my case setting the video mode triggers messages related to dpms, pins and connectors, whatever those are. Compiling kernel 3.15.1 and messing with configuration settings does not seem to change anything.
Is this still an issue with a more recent kernel, like 4.0?
Hi @Meelis Roos could you provide information to last comment
Just tested 4.2.0-rc6+git and the problem is gone - KMS works on this X30.
(In reply to Meelis Roos from comment #20) > Just tested 4.2.0-rc6+git and the problem is gone - KMS works on this > X30. Marking as fixed.
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.