drm_do_probe_ddc_edid ENXIO check too aggressive?

[Daniel Drake]

On Wed, Dec 18, 2013 at 2:18 PM, Daniel Drake <drake at endlessm.com> wrote:
> Yes, this looks very similar to the approach I tried earlier. I guess
> the patch was written for the same reasons as well.
> Sean, any objections to me taking your patch and sending it upstream?
>
> http://git.chromium.org/gitweb/?p=chromiumos/third_party/kernel-next.git;a=commitdiff;h=77aed71acf4b84e722ead24c820d8ad059121e5b

Seems easier said than done. The patch looks innocent but it touches
on other stuff that has changed a lot.

For example, it adds cancel_delayed_work_sync() in the hdmi_poweroff()
path. With my naive backport to vanilla-3.8, hdmi_poweroff() gets
called within the context of that work item, and a work item trying to
cancel_delayed_work_sync() on itself hangs. Reproduced just by
unplugging the display once.

I see that things have changed substantially in the ChromiumOS tree,
for example this patch probably has some effect:

commit 29ae0c6096395a96a597453271946fc7d8442b6e
Author: Sean Paul <seanpaul at chromium.org>
Date:   Thu Jun 20 17:24:12 2013 -0400

    drm/exynos: Consolidate suspend/resume in drm_drv


As this job seems bigger than anticipated and I don't have any Exynos
hardware that can boot mainline, I think I might have to stop here,
and hope that ChromiumOS guys upstream all this soon?

Daniel