Bug 81951 - [gma500] "No Signal" to monitor on boot
Summary: [gma500] "No Signal" to monitor on boot
Status: NEW
Alias: None
Product: Drivers
Classification: Unclassified
Component: Video(DRI - Intel) (show other bugs)
Hardware: IA-32 Linux
: P3 normal
Assignee: Patrik Jakobsson
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2014-08-08 18:51 UTC by TJ
Modified: 2016-03-23 18:29 UTC (History)
2 users (show)

See Also:
Kernel Version: 3.13.0
Subsystem:
Regression: No
Bisected commit-id:


Attachments
DSA-3020 dmesg with ""video=DVI-D-1:1920x1080D" (60.02 KB, text/x-log)
2014-08-08 18:51 UTC, TJ
Details

Description TJ 2014-08-08 18:51:16 UTC
Created attachment 145901 [details]
DSA-3020 dmesg with ""video=DVI-D-1:1920x1080D"

I've just been assisting a user with an Advantech DSA-3020 board [1][user manual] which
has the Poulsbro chip-set and uses the gma500_gfx driver on a small embedded server currently running Ubuntu 14.04.1 LTS.

The device has 2 outputs: LVDS-1 and DVI-D-1. Nothing is connected to LVDS-1. A 1920x1080 monitor is connected to DVI-D-1.

During POST and bootloader the display is on DVI-D-1. As soon as gma500_gfx initialises the DVI-D-1 display reports "No Signal". Nothing we've done in diagnosing this has helped.

One workaround is to blacklist gma500_gfx and allow the vesafb driver to handle the GPU. The other is to totally disable ACPI in BIOS. If ACPI is enabled in BIOS but disabled in the OS ("acpi=off") the symptoms still occur. Disabling ACPI is not acceptable since other issues arise.

We found:

/sys/class/drm/card0-DVI-D-1/enabled=disabled
/sys/class/drm/card0-DVI-D-1/status=connected
/sys/class/drm/card0-LVDS-1/enabled=disabled
/sys/class/drm/card0-LVDS-1/status=connected

We used "video=DVI-D-1:1920x1080D video=LVDS-1:1920x1080e" (both independently and together) but there was no improvement despite:

[drm] forcing DVI-D-1 connector ON

/sys/class/drm/card0-DVI-D-1/dpms=On
/sys/class/drm/card0-DVI-D-1/edid=<... Samsung EDID block ...>
/sys/class/drm/card0-DVI-D-1/enabled=enabled
/sys/class/drm/card0-DVI-D-1/modes=1920x1080
1920x1080
1920x1080
1920x1080
1920x1080
1920x1080
1920x1080
1920x1080
1680x1050
1600x900
1280x1024
1280x1024
1440x900
1366x768
1280x800
1152x864
1280x720
1280x720
1280x720
1024x768
1024x768
1024x768
832x624
800x600
800x600
800x600
720x576
720x480
720x480
640x480
640x480
640x480
640x480
640x480
720x400
/sys/class/drm/card0-DVI-D-1/status=connected
/sys/class/drm/card0-DVI-D-1/uevent=
/sys/class/drm/card0-LVDS-1/dpms=On
/sys/class/drm/card0-LVDS-1/edid=
/sys/class/drm/card0-LVDS-1/enabled=enabled
/sys/class/drm/card0-LVDS-1/modes=800x600
/sys/class/drm/card0-LVDS-1/status=connected
/sys/class/drm/card0-LVDS-1/uevent=

We also tried "vbetool dpms off".


[1] http://support.elmark.com.pl/advantech/pdf/DSA-3020man.pdf
Comment 1 Alan 2014-08-21 18:29:45 UTC
Unfortunately for a board and BIOS ACPI setting specific bug like that you'd pretty much need someone in front of the board to debug it. Poulsbo uses an Imagination graphics core so there is no documentation and the driver is built out of bits that were required to be released under the GPL, and a lot of trial/error and guesswork.

If there is some kind of bridge or external logic that has magic power hooks not in the code that it was developed from then you are basically out of luck

Note You need to log in before you can comment on or make changes to this bug.