[Intel-gfx] 3.0 (or SNA?) regression: failed to train DP, aborting

[Andrew Lutomirski]

On Mon, Jul 25, 2011 at 12:54 AM, Keith Packard <keithp at keithp.com> wrote:
> On Sat, 23 Jul 2011 14:40:36 -0400, Andrew Lutomirski <luto at mit.edu> wrote:
>> I have a Q67 (DQ67SW board) attached to a Dell U2711 via DP.  In
>> previous kernels, the DP link has worked flawlessly.  I just booted
>> 3.0-final and simultaneously enabled SNA, and now when my screen goes
>> to sleep I don't get an image back until I power cycle the monitor.
>> dmesg says:
>>
>> [drm:intel_dp_complete_link_train] *ERROR* failed to train DP,
>> aborting
>
> Jesse put together a set of 7 display port patches around July 7 which
> were merged just after 3.0-rc6. If you try 3.0-rc6 and find that it
> works, you should be able to bisect that really quickly; there are only
> 13 patches post-rc6 in drivers/gpu/drm/i915:
>
> $ git bisect start v3.0 v3.0-rc6 -- drivers/gpu/drm/i915

The offending commit appears to be:

commit 885a50147f00a8a80108904bf58a18af357717f3
Author: Jesse Barnes <jbarnes at virtuousgeek.org>
Date:   Thu Jul 7 11:11:01 2011 -0700

    drm/i915/dp: remove DPMS mode tracking from DP

    We currently use this when a hot plug event is received, only checking
    the link status and re-training if we had previously configured a link.
    However if we want to preserve the DP configuration across both hot plug
    and DPMS events (which we do for userspace apps that don't respond to
    hot plug uevents), we need to unconditionally check the link and try to
    bring it up on hot plug.

    Signed-off-by: Jesse Barnes <jbarnes at virtuousgeek.org>
    Reviewed-by: Keith Packard <keithp at keithp.com>
    Signed-off-by: Keith Packard <keithp at keithp.com>

A debugging patch and its output are attached.

If I had to guess, though, it's a race: a hotplug event happens during
the intel_dp_dpms callback, confusing the code that's trying to train
the link.

--Andy