Bug 60046 - [845G KMS] Blank screen using KMS
Summary: [845G KMS] Blank screen using KMS
Status: CLOSED INVALID
Alias: None
Product: DRI
Classification: Unclassified
Component: DRM/Intel (show other bugs)
Version: unspecified
Hardware: x86 (IA32) All
: medium normal
Assignee: Intel GFX Bugs mailing list
QA Contact: Intel GFX Bugs mailing list
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2013-01-30 00:28 UTC by Trey Ramsay
Modified: 2017-07-24 22:59 UTC (History)
1 user (show)

See Also:
i915 platform: I845G
i915 features: display/Other


Attachments
dmesg.log after X has come up using drm.debug=0x7 (333.21 KB, text/plain)
2013-01-30 00:28 UTC, Trey Ramsay
no flags Details
Xorg.0.log (31.94 KB, text/plain)
2013-01-30 00:29 UTC, Trey Ramsay
no flags Details
xrandr.verbose xrandr --verbose (5.33 KB, text/plain)
2013-01-30 00:30 UTC, Trey Ramsay
no flags Details
i915_error_state after GPU hang with SNA enabled (673.37 KB, text/plain)
2013-01-30 15:56 UTC, Trey Ramsay
no flags Details
dmesg-3.8-rc5 (49.48 KB, text/plain)
2013-01-31 15:22 UTC, Trey Ramsay
no flags Details
dmesg-3.8-rc5_before_X (85.65 KB, text/plain)
2013-01-31 17:50 UTC, Trey Ramsay
no flags Details
dmesg output from kernel v3.11 (90.37 KB, text/plain)
2013-09-24 21:24 UTC, Reza Arbab
no flags Details
output from intel_reg_dumper on kernel v3.11 (9.72 KB, text/plain)
2013-09-24 21:30 UTC, Reza Arbab
no flags Details
dmesg output from kernel e4c64bc (57.29 KB, text/plain)
2013-09-25 16:58 UTC, Reza Arbab
no flags Details
dmesg output from kernel 3.13 (59.30 KB, text/plain)
2014-01-20 17:09 UTC, Reza Arbab
no flags Details
dmesg output from kernel 0f7cc12 (58.68 KB, text/plain)
2014-09-25 16:25 UTC, Reza Arbab
no flags Details

Description Trey Ramsay 2013-01-30 00:28:32 UTC
Created attachment 73890 [details]
dmesg.log after X has come up using drm.debug=0x7

The screen is blank before and after X is started when using KMS.  The display comes up by disabling KMS with the "nomodeset" option. 

System environment:
Intel 845G
3.3.4-5.fc17.i686

I installed intel-gpu-tools-2.20.16-1.fc17.i686 but was unable to run intel_reg_dumper. 

# intel_reg_dumper
Gen2/3 Ranges are not supported. Please use unsafe access.Aborted (core dumped)
Comment 1 Trey Ramsay 2013-01-30 00:29:24 UTC
Created attachment 73891 [details]
Xorg.0.log
Comment 2 Trey Ramsay 2013-01-30 00:30:21 UTC
Created attachment 73892 [details]
xrandr.verbose  xrandr --verbose
Comment 3 Chris Wilson 2013-01-30 08:53:10 UTC
What's the actual display hardware you are using? It detects both an LVDS and VGA connection. As the 845g can only show on one pipe, it can only show on either LVDS or VGA (if we ignore the rather unhelpful and fragile clone mode). Currently it is using the LVDS, which I guess is a phantom output.

Perhaps xrandr --output LVDS1 --off --output VGA1 --preferred

To make that permanent you can add video=LVDS-1:d to you kernel parameters.
Comment 4 Chris Wilson 2013-01-30 08:54:28 UTC
Btw, neither of those versions of the kernel or ddx are stable on 845g. You need either a brand spanking new 3.8 or to enable SNA.
Comment 5 Daniel Vetter 2013-01-30 10:02:15 UTC
Btw, latst intel-gpu-tools git from http://cgit.freedesktop.org/xorg/app/intel-gpu-tools/ should work for your platform ...
Comment 6 Trey Ramsay 2013-01-30 14:25:48 UTC
There are two displays, one is a integrated display LVDS-1 and the other is external, VGA-1. Clone mode would be ideal if we can get that to work.

I didn't have any luck with disabling LVDS1 using xrandr
# xrandr --output LVDS1 --off --output VGA1 --preferred
xrandr: cannot find crtc for output VGA1

# xrandr --output LVDS1 --preferred --output VGA1 --off
Had no effect

I was able to get Fedora to come up on external display VGA-1 by disabling LVDS-1 in the kernel command line.  The integrated display acted strange though with pixels being illuminated to look like a white fog on the screen.
video=LVDS-1:d

Disabling VGA-1 in the kernel command line had no effect.
Comment 7 Trey Ramsay 2013-01-30 15:14:15 UTC
Enabling SNA in the xorg.conf file resulted in a GPU hang.  I'll compile and install the 3.8-rc5 kernel to see if that helps.
Comment 8 Chris Wilson 2013-01-30 15:18:40 UTC
Let's see the hang (i915_error_state), 2.20.16 is meant to be stable.
Comment 9 Trey Ramsay 2013-01-30 15:56:50 UTC
Created attachment 73937 [details]
i915_error_state after GPU hang with SNA enabled
Comment 10 Chris Wilson 2013-01-30 15:59:44 UTC
(In reply to comment #9)
> Created attachment 73937 [details]
> i915_error_state after GPU hang with SNA enabled

Ok, that's one of Daniel's KMS hangs, not due to SNA itself. :-p
Comment 11 Trey Ramsay 2013-01-31 03:18:58 UTC
Thanks for the help!  Looks like the problem is fixed in the v3.8-rc5!  I'll work on doing a bisect of the code to narrow down the patch that fixes this problem. :)
Comment 12 Trey Ramsay 2013-01-31 15:13:29 UTC
Looks like the reason it is working on 3.8-rc5 is because the drm indicated
that there was no driver support for vblank timestamp query.  So it doesn't appear to be using KMS in 3.8-rc5.

3.8-rc5
[    1.234348] [drm] Supports vblank timestamp caching Rev 1 (10.10.2010).
[    1.241052] [drm] No driver support for vblank timestamp query.
[    1.247000] [drm] Initialized i915 1.6.0 20080730 for 0000:00:02.0 on minor 0

3.3.4-5.fc17
[    3.301722] [drm] Supports vblank timestamp caching Rev 1 (10.10.2010).
[    3.308365] [drm] Driver supports precise vblank timestamp query.
[    3.354705] [drm:init_vbt_defaults], Set default to SSC at 66MHz
[    3.354752] [drm:parse_general_features], BDB_GENERAL_FEATURES int_tv_support
 1 int_crt_support 0 lvds_use_ssc 0 lvds_ssc_freq 48 display_clock_mode 0
[    3.354762] [drm:parse_general_definitions], crt_ddc_bus_pin: 2
[    3.354769] [drm:parse_sdvo_device_mapping], different child size is found. I
nvalid.
[    3.354774] [drm:parse_device_mapping], different child size is found. Invali
d.
[    3.354805] [drm:quirk_pipea_force], applying pipe a force quirk
[    3.354814] [drm:intel_modeset_init], 1 display pipe available.
[    3.354828] [drm:intel_modeset_init], plane 0 init failed: -19
[    3.354846] vgaarb: device changed decodes: PCI:0000:00:02.0,olddecodes=io+me
m,decodes=io+mem:owns=io+mem
[    3.364761] [drm:drm_sysfs_connector_add], adding "VGA-1" to sysfs
Comment 13 Daniel Vetter 2013-01-31 15:15:18 UTC
Precise vblank query support is optional, so kms seems to be still loading I think. If you're in doubt, please add the complete debug dmesg.
Comment 14 Trey Ramsay 2013-01-31 15:22:57 UTC
Created attachment 73998 [details]
dmesg-3.8-rc5
Comment 15 Daniel Vetter 2013-01-31 15:39:50 UTC
Your driver loads fine, though with kernel modesetting disabled. Do you have a lingering setting somewhere for that?
Comment 16 Trey Ramsay 2013-01-31 15:50:39 UTC
doh! Thanks!  
My .config was missing CONFIG_DRM_I915_KMS

CONFIG_DRM_I915=y
# CONFIG_DRM_I915_KMS is not set

I'll give 3.8-rc5 another try.
Comment 17 Trey Ramsay 2013-01-31 17:30:17 UTC
Please disregard comment 11 and comment 12.

With KMS enabled on 3.8-rc5 in the kernel config, I got similar results to comment 6 when LVDS-1 was disabled on the kernel command line. It came up on external display VGA-1 and the integrated display acted strange with pixels being illuminated with a white fog coming and going on the screen.
Comment 18 Trey Ramsay 2013-01-31 17:50:22 UTC
Created attachment 74008 [details]
dmesg-3.8-rc5_before_X

This is the dmesg output after the system is up and running and before X is started.
Comment 19 Daniel Vetter 2013-03-20 11:57:28 UTC
Ok, it sounds like we have a bit a sync issue on your LVDS output still. Can you please test http://cgit.freedesktop.org/~danvet/drm/commit/?h=pll-limits-mess&id=c2965c7fae71ac7a06dc6fe1b4c2e9b93dc28ed0
Comment 20 Trey Ramsay 2013-04-04 19:20:37 UTC
I don't have the failing system anymore but I'll see if I can have someone test the change.  Thanks!
Comment 21 Trey Ramsay 2013-05-08 20:50:51 UTC
The change in Comment 19 did not fix the problem.
Comment 22 Chris Wilson 2013-07-07 12:55:05 UTC
Sounds like we are not disabling the LVDS and its backlight entirely when switching the pipe over to VGA. Can you please grab intel_reg_dumper with LVDS on and LVDS off (but with the spooky glow effect).
Comment 23 Trey Ramsay 2013-07-11 17:32:59 UTC
Sorry Chris,
I don't have access to the system anymore.  I'll see if someone can get the information you need.

Thanks,
Trey
Comment 24 Reza Arbab 2013-09-24 21:24:35 UTC
Created attachment 86486 [details]
dmesg output from kernel v3.11

Sorry for the delay, I'll be picking up from here.

I've recreated with drm.debug=14 on kernel 3.11.  On boot, the screen alternates between a few minutes of blackness and an animated plasma/lava lamp effect.

Something new in the dmesg output that may be helpful:

[    1.737127] WARNING: CPU: 0 PID: 66 at drivers/gpu/drm/i915/intel_display.c:1074 assert_pipe+0x82/0xe0 [i915]()
[    1.737129] pipe A assertion failure (expected on, current off)

Please let me know what else I could do to help.
Comment 25 Reza Arbab 2013-09-24 21:30:12 UTC
Created attachment 86487 [details]
output from intel_reg_dumper on kernel v3.11

I ran it while the screen was doing the spooky glow/fade to black dance.
Comment 26 Daniel Vetter 2013-09-25 10:33:33 UTC
We've just fixed a bug in the drm-intel-fixes branch which applies for this situation. Can you please retest with that? git tree is at

http://cgit.freedesktop.org/~danvet/drm-intel/
Comment 27 Reza Arbab 2013-09-25 16:58:45 UTC
Created attachment 86578 [details]
dmesg output from kernel e4c64bc

No problem.  I recreated using the drm-intel-fixes branch.  git says I'm at v3.12-rc2-4-ge4c64bc.

The assertion failure is gone, but the bug behavior is still there.  The screen goes to black and periodically does the spooky glow/plasma/lava lamp thing.
Comment 28 Reza Arbab 2014-01-20 17:09:11 UTC
Created attachment 92468 [details]
dmesg output from kernel 3.13

Still having this problem.  

Can't tell from the dmesg exactly where things go bad.  Is there anything else I can try or do to help?
Comment 29 Ben Widawsky 2014-01-29 16:49:50 UTC
Eeek 845.

Daniel got his info. Reassigning to him.
Comment 30 Rodrigo Vivi 2014-09-24 20:29:39 UTC
Is this still happening on latest drm-intel-nightly?
Comment 31 Reza Arbab 2014-09-25 16:25:14 UTC
Created attachment 106862 [details]
dmesg output from kernel 0f7cc12

Yes, it does. I just verified using drm-intel-next-2014-09-19-379-g0f7cc12. Same behavior.
Comment 32 Elio 2015-08-12 18:02:23 UTC
Trying to contact Devs in order to see if there is any solution or if we need to close it since it was a long time without answer.
Comment 33 Jani Nikula 2016-04-21 12:43:49 UTC
(In reply to Elio from comment #32)
> Trying to contact Devs in order to see if there is any solution or if we
> need to close it since it was a long time without answer.

Timeout, closing. Please reopen if the problem persists with latest kernels.


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.