Corrupted display on XO1.5 when suspended with screen on

[James Cameron]

On Tue, Jun 25, 2013 at 12:18:43AM -0500, Sebastian Silva wrote:
> El 21/06/13 13:56, John Watlington escribió:
> >This is likely due to the video controller being disabled before
> >the DCON driver has completely finished loading the image into the
> >DCON.
> >>Our image is based on 11.3.1 shipping firmware olpc-firmware-q3c06 (but my testing machine has q3c07 and exhibits the same malfunction).
> >>
> >>(...)
> >
> >We did some work to fix the DCON driver in newer releases.
> >
> So what component / package etc should I be looking to backport?

The kernel, specifically kernel patches to the DCON driver or related
components.

11.3.1 was based on kernel 2.6.35.13_xo1.5-20120508.1139.olpc.eb0c7a8
which corresponds to branch olpc-2.6.35 of the olpc-kernel repository:

http://dev.laptop.org/git/olpc-kernel/commit/?h=olpc-2.6.35&id=eb0c7a8

Development then switched to the branch x86-3.3, and this is where any
DCON or display driver fixes after 11.3.1 will be found.

I've had a quick look and these viafb fixes look relevant:

http://dev.laptop.org/git/olpc-kernel/commit/?h=x86-3.3&id=3ef9f18dfe5f97dbae96b8b7658590b6dbf81cea
http://dev.laptop.org/git/olpc-kernel/commit/?h=x86-3.3&id=d5caefcdf160b0b24153bba89b111419da2f92c3

Other than these viafb fixes, there have been no DCON driver fixes in
x86-3.3.  I recall some fixes in the ARM laptops though.  Perhaps they
were not backported.

As kernel fixing can be costly for you, let me take you back a few
steps in diagnosis ... can you confirm that the problem also affects
the original OLPC OS 11.3.1 and not only your own build of 11.3.1?

Does the problem only occur with automatic power management enabled?

Is there a way to induce the problem quickly?

It reminds me of #11231 and #10038.  You might see if the reproducers
in those tickets cause a problem.

Also, although we understand you cannot go to deployment phase with
anything later than 11.3.1 because of non-technical reasons, the
analysis of this problem could really benefit from trying to reproduce
it with 12.1.0, 13.1.0, or 13.2.0.  If the problem does not reproduce
in 12.1.0 for instance, it means the fix may be in that interval.  It
would also help to exclude hardware as a cause.

-- 
James Cameron
http://quozl.linux.org.au/