Bug 38012

Alright, so I have bisected down my crash to bcd5023c961a44c7149936553b6929b2b233dd27, which alters the lid detection logic.

Reverting this change in 2.6.39 results in a kernel that is vaguely usable, however there is no output to the laptop's display.

Closing the lid works, it redetects displays and we are still good.

Reopening the lid causes the laptop to lock solid, exactly as it was on starting X.

I am now attempting to bisect where eDP handling for my laptop broke, joy.

On the bright side, bcd5023c961a44c7149936553b6929b2b233dd27 is confirmed to impact my system as well. (HP 8440p)

Zephaniah E. Loss-Cutler-Hull.

Alright, I have bisected this down to 9035a97a32836d0e456ddafaaf249a844e6e4b5e.

fe16d949b45036d9f80e20e07bde1ddacc930b10 is good.

452858338aec31c1f4414bf07f31663690479869 is good.

And 9035a97a is bad.

Now the catch, this is a merge commit, and I'm still trying to figure out how to untangle it to have something I can revert.

Suggestions are welcome.

(In reply to comment #2)
> Alright, I have bisected this down to 9035a97a32836d0e456ddafaaf249a844e6e4b5e.
> 
> fe16d949b45036d9f80e20e07bde1ddacc930b10 is good.
> 
> 452858338aec31c1f4414bf07f31663690479869 is good.
> 
> And 9035a97a is bad.
> 
> Now the catch, this is a merge commit, and I'm still trying to figure out how
> to untangle it to have something I can revert.
> 
> Suggestions are welcome.

So, knowing that 4528583 was the drm-intel-fixes branch, I have tested a version of 9035a97 with all the differences in drivers/gpu/drm/i915 between fe16d94 and 9035a97 reverted, effectively reverting all of the pure intel changes brought in by the merge.

The resulting kernel still had the bug.

At this point, I am stumped, and can not see how to bisect this problem any further.

Help?

Zephaniah E. Loss-Cutler-Hull.

And it's been bisected down on the xorg driver side as well.

0d26d950fdada1f59dc6cb31fe2f03004825f773, 'KMS: add fake EDID on eDP too' is the commit which causes kernels from 9035a97 and forward to do a hard lock when we try to bring the eDP live.

Due to some later changes a straight revert against 2.15.0 is not possible, but changing the line in question to:

if (is_panel(koutput->connector_type) && koutput->connector_type != DRM_MODE_CONNECTOR_eDP) {

allows 2.15.0 to work with 9035a97.

Of course, the 9035a97 merge has very little to do with the decision to treat it as a panel, so I am well and truly lost as to what the _right_ fix is.

In the mean time, keeping the xorg driver from treating eDP as a panel at the very least lets my system work again, even if it's the wrong solution in the long term.

Zephaniah E. Loss-Cutler-Hull.

After testing, we can't reproduce the issue with commit 9035a97a on our Capella test machine. We run Ubuntu 11.04 with the 2.14.0 driver.

*** Bug 39108 has been marked as a duplicate of this bug. ***

As discussed on bug 39108:


commit 212fa9868767637e8f430485eeb522c99e63fd16
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Wed Jul 13 21:11:14 2011 +0100

    Disable adding normal RTF modes for an eDP
    
    This is causing a hard hang with 2.6.39+, we don't know why so play safe
    and disable for the time being.
    
    References: https://bugs.freedesktop.org/show_bug.cgi?id=38012
    Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

as a temporary avoidance measure.

Jesse posted a patch for a related bug [bug 36888], please can you test https://bugs.freedesktop.org/attachment.cgi?id=48015

Also some major modesetting bugs were squashed in keithp/drm-intel-fixes

commit d74362c9e45689d8d7e3d4bcf6681c4358ef4f2e
Author: Keith Packard <keithp@keithp.com>
Date:   Thu Jul 28 14:47:14 2011 -0700

    drm/i915: Flush other plane register writes
    
    Writes to the plane control register are buffered in the chip until a
    write to the DSPADDR (pre-965) or DSPSURF (post-965) register occurs.
    
    This patch adds flushes in:
    
        intel_enable_plane
        gen6_init_clock_gating
        ivybridge_init_clock_gating
    
    Signed-off-by: Keith Packard <keithp@keithp.com>

commit 2704cf5fbd248871a745d210733c6319959d2b0c
Author: Jesse Barnes <jbarnes@virtuousgeek.org>
Date:   Thu Jul 28 11:52:45 2011 -0700

    drm/i915: flush plane control changes on ILK+ as well
    
    After writing to the plane control reg we need to write to the surface
    reg to trigger the double buffered register latch.  On previous
    chipsets, writing to DSPADDR was enough, but on ILK+ DSPSURF is the reg
    that triggers the double buffer latch.
    
    v2: write DSPADDR too to cover pre-965 chipsets
    v3: use flush_display_plane instead, that's what it's for
    v4: send the right patch
    
    Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
    Tested-by: Keith Packard <keithp@keithp.com>
    Reviewed-by: Keith Packard <keithp@keithp.com>
    Signed-off-by: Keith Packard <keithp@keithp.com>


please can you test those?

(In reply to comment #9)
> Also some major modesetting bugs were squashed in keithp/drm-intel-fixes
> 
> commit d74362c9e45689d8d7e3d4bcf6681c4358ef4f2e
> commit 2704cf5fbd248871a745d210733c6319959d2b0c

> please can you test those?

I can confirm that 3.0 with those two patches applied works perfectly with the 2.14.0-4ubuntu7.1 xorg driver which ships with Ubuntu 11.04.

So I think we can call this bug solved, and the 212fa9868767637e8f430485eeb522c99e63fd16 xorg driver commit mentioned above can be reverted.

Any chance that we can get those two patches marked for stable kernel backports?

Thanks!
Zephaniah E. Loss-Cutler-Hull.

(In reply to comment #10)
> So I think we can call this bug solved, and the
> 212fa9868767637e8f430485eeb522c99e63fd16 xorg driver commit mentioned above can
> be reverted.
> 
> Any chance that we can get those two patches marked for stable kernel
> backports?
> 
> Thanks!
> Zephaniah E. Loss-Cutler-Hull.

Or, not.

It worked once at home, and since then has been failing in the exact same manner, damn.

Sorry, still broken.

Zephaniah E. Loss-Cutler-Hull.

> It worked once at home, and since then has been failing in the exact same
> manner, damn.
> 
> Sorry, still broken.
> 
> Zephaniah E. Loss-Cutler-Hull.

Alright, after some poking, the patches are definitely an improvement which should get backported, but they are not the whole story.

I will try to get some debug logs tomorrow showing the modes being set.

The 3.0 kernel with the patches applied works for the laptop when no external display is connected.

It also works when my display at home is connected.

It locks up solidly when my display at work is connected, that is an Acer V193W.

Both are DVI monitors plugged into a docking station, the one at the office is 1440x900, the one at home is 1600x900.

The laptop is natively 1600x900, however I have caught it going into 1440x900 scaled by default when the work monitor is plugged in.

(In reply to comment #12)
> Alright, after some poking, the patches are definitely an improvement which
> should get backported, but they are not the whole story.

However drm-intel-next as of 2cfad3600e192f867c0e03905c90903f189c010a (other versions not tested) appears to be working perfectly.

I'll test more, but so far it's working in all the configurations I use the system in.

Zephaniah E. Loss-Cutler-Hull.

So, is it fixed now, right?