kern/49592: EDID doesn't working in Xorg on HD2500 in 7.0_BETA and -current (amd64)

[aniou%smutek.pl@localhost]

>Number:         49592
>Category:       kern
>Synopsis:       EDID doesn't working in Xorg on HD2500 in 7.0_BETA and -current (amd64)
>Confidential:   no
>Severity:       critical
>Priority:       medium
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Tue Jan 20 21:15:00 +0000 2015
>Originator:     Piotr Meyer
>Release:        7.0_BETA
>Organization:
>Environment:
>Description:
amd64, G2030 (Ivy Bridge) CPU with HD2500 graphics (integrated)

Regression: EDID reads had stopped after import DRM from linux kernel 3.15. I think it is a kernel regression because same userland can read EDID data with older kernel (see below).

Now, starting X session ends in blank screen. Creating mode manually (via xrandr) gave me working system - but nowadays users are used to automatic mode configuration.

Unfortunately, DRM in debug mode create more logs than buffer size and I lost some data.

kernel log before 'startx' command: http://smutek.pl/netbsd/drmkms-regression-201501/dmesg.txt
kernel log after 'startx' command: http://smutek.pl/netbsd/drmkms-regression-201501/dmesg-after-x.txt
Xorg log: http://smutek.pl/netbsd/drmkms-regression-201501/Xorg.0.log

for comparision similar logs from older kernel:

http://smutek.pl/netbsd/drmkms-regression-201501/last-working-edid-messages.txt
http://smutek.pl/netbsd/drmkms-regression-201501/last-working-edid-Xorg.log.txt

I tried all three ports (DVI, HDMI and D-SUB) without success.

Point in time, when new code was commited: https://github.com/jsonn/src/commit/740a2c2859

>How-To-Repeat:

>Fix:
A workaround (mode definition should be taken from current system):

$ cat .xinitrc
xrandr --newmode 1920x1080 148.50  1920 2008 2052 2200  1080 1084 1089 1125 +hsync +vsync
xrandr --addmode HDMI1 1920x1080
xrandr --output HDMI1 --mode 1920x1080