[Intel-gfx] LVDS output not working (anymore)

Chris Wilson chris at chris-wilson.co.uk
Wed Feb 9 14:59:40 CET 2011


On Wed, 09 Feb 2011 14:40:11 +0100, Oliver Seitz <info at vtnd.de> wrote:
> 
> >> I'm using intel D945GSEJT boards, connected to industrial LVDS panels
> >> without EDID or the like.
> 
> >> Now, on debian 6.0, LVDS output does not show up in xrandr's output list
> >> anymore. I understand that there have been some recent changes that are
> >> not fully applied to debian 6.0. Is there a way to activate the LVDS output?
> >
> > Please tell us more! Perhaps a debug log or two? A description of the
> > physical encoders and connectors?
> 
> I'm happy to provide all information that may be needed, but as I'm new 
> to helping in driver debugging, I do not know what may be helpful...
> 
> The board employs a 945GSE chip which is already equipped with LVDS 
> output. http://ark.intel.com/chipset.aspx?familyID=35553
> The LVDS output is avaiable on the board via header connector. 
> http://download.intel.com/support/motherboards/desktop/d945gsejt/sb/e77911001.pdf
> (This document chooses not to mention the type of the connector, though. 
> It is a JST-SHDR-30)
> 
> I have only connected channel A clock and data, and the power lines to a 
> NEC TFT display. EDID/DDC lines are open.
> 
> This setup worked on some old ubuntu (can't remember the version) 
> out-of-the-box. When KMS was introduced, it stopped working, as KMS did 
> not activate LVDS, probably due to the missing EDID info. "nomodeset" 
> did the trick since then, as X saw the LVDS option and xrandr was able 
> to activate it:

Thanks! I just needed to confirm that it was indeed an LVDS header on the
board as opposed to an SDVO LVDS.

As you have ascertained KMS ignores any LVDS that does not have any
modes available (either in the EDID or store in the VBT) and is otherwise
not known to the BIOS. In order to evade this bit of logic we would need
to parse the video= kernel parameter inside intel_lvds_init() and so setup
the user supplied fixed mode.

Patch in progress.
-Chris

> P.S.: I hope it is the correct style to send the logs as attachments 
> untarred and un-gzipped. I apologize if it is not.

Perfect.

-- 
Chris Wilson, Intel Open Source Technology Centre



More information about the Intel-gfx mailing list