[Intel-gfx] [PATCH] drm/i915: Revert ea0000f0 "Roll out the helper nonblock tracking"

[Daniel Vetter]

On Tue, Dec 20, 2016 at 04:27:42PM +0200, Jani Nikula wrote:
> On Sun, 27 Nov 2016, Jim Rees <rees at umich.edu> wrote:
> > Commit ea0000f0 has been causing multiple problems for i915 users. See for
> > example:
> >
> > https://bugs.freedesktop.org/show_bug.cgi?id=96781
> > https://bugs.freedesktop.org/show_bug.cgi?id=97529
> > https://bugzilla.redhat.com/show_bug.cgi?id=1385228
> > https://forums.opensuse.org/showthread.php/520969-drm-915-Resetting-chip-after-gpu-hang
> >
> > Those last two are not confirmed but look like the same problem.
> >
> > This has only been lightly tested and probably shouldn't go upstream yet.
> >
> > Signed-off-by: Jim Rees <rees at umich.edu>
> 
> Daniel, any comments on this one?

So root-cause is that vblank waits die, and from one of the bug reports it
sounds like drm-tip has that fixed. For that we need a reverse bisect to
find the bugfix so that we can backport it.

But vblank timeout shouldn't result in oopses, the helpers are meant to
carry on in such a case (although with some horribly long timeouts). From
the backtraces I think I know what's going on and how to paper over that,
typing a patch right now.
-Daniel

> 
> BR,
> Jani.
> 
> 
> > ---
> >  drivers/gpu/drm/i915/intel_display.c | 10 ----------
> >  1 file changed, 10 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
> > index 81c1149..c31c9e4 100644
> > --- a/drivers/gpu/drm/i915/intel_display.c
> > +++ b/drivers/gpu/drm/i915/intel_display.c
> > @@ -14346,8 +14346,6 @@ static void intel_atomic_commit_tail(struct drm_atomic_state *state)
> >  		WARN_ON(ret);
> >  	}
> >  
> > -	drm_atomic_helper_wait_for_dependencies(state);
> > -
> >  	if (intel_state->modeset) {
> >  		memcpy(dev_priv->min_pixclk, intel_state->min_pixclk,
> >  		       sizeof(intel_state->min_pixclk));
> > @@ -14470,8 +14468,6 @@ static void intel_atomic_commit_tail(struct drm_atomic_state *state)
> >  	if (intel_state->modeset && intel_can_enable_sagv(state))
> >  		intel_enable_sagv(dev_priv);
> >  
> > -	drm_atomic_helper_commit_hw_done(state);
> > -
> >  	if (intel_state->modeset)
> >  		intel_display_power_put(dev_priv, POWER_DOMAIN_MODESET);
> >  
> > @@ -14479,8 +14475,6 @@ static void intel_atomic_commit_tail(struct drm_atomic_state *state)
> >  	drm_atomic_helper_cleanup_planes(dev, state);
> >  	mutex_unlock(&dev->struct_mutex);
> >  
> > -	drm_atomic_helper_commit_cleanup_done(state);
> > -
> >  	drm_atomic_state_free(state);
> >  
> >  	/* As one of the primary mmio accessors, KMS has a high likelihood
> > @@ -14546,10 +14540,6 @@ static int intel_atomic_commit(struct drm_device *dev,
> >  		return -EINVAL;
> >  	}
> >  
> > -	ret = drm_atomic_helper_setup_commit(state, nonblock);
> > -	if (ret)
> > -		return ret;
> > -
> >  	INIT_WORK(&state->commit_work, intel_atomic_commit_work);
> >  
> >  	ret = intel_atomic_prepare_commit(dev, state, nonblock);
> 
> -- 
> Jani Nikula, Intel Open Source Technology Center
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/intel-gfx

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch