[Intel-gfx] [PATCH v2 1/2] drm/i915: Fix modeset handling during gpu reset.

[Ville Syrjälä]

On Mon, Apr 11, 2016 at 08:23:33AM +0200, Maarten Lankhorst wrote:
> This function would call drm_modeset_lock_all, while the suspend/resume
> functions already have their own locking. Fix this by factoring out
> __intel_display_resume, and calling the atomic helpers for duplicating
> atomic state and disabling all crtc's during suspend.
> 
> Signed-off-by: Maarten Lankhorst <maarten.lankhorst at linux.intel.com>
> Fixes: e2c8b8701e2d ("drm/i915: Use atomic helpers for suspend, v2.")
> Cc: drm-intel-fixes at lists.freedesktop.org
> ---
>  drivers/gpu/drm/i915/intel_display.c | 123 +++++++++++++++++++++++------------
>  1 file changed, 83 insertions(+), 40 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
> index feb7028341b8..2f76efc86417 100644
> --- a/drivers/gpu/drm/i915/intel_display.c
> +++ b/drivers/gpu/drm/i915/intel_display.c
> @@ -3126,8 +3126,44 @@ static void intel_update_primary_planes(struct drm_device *dev)
>  	}
>  }
>  
> +static int
> +__intel_display_resume(struct drm_device *dev,
> +		       struct drm_atomic_state *state,
> +		       bool *setup)
> +{
> +	struct drm_crtc_state *crtc_state;
> +	struct drm_crtc *crtc;
> +	int i;
> +
> +	if (!setup || !*setup) {
> +		if (setup)
> +			*setup = true;

This is getting nasty. How about dealing with -EDEADLK right after the
modeset_lock_all() in the caller, so that we don't have to be concerned
with getting called mutiple times here?

> +
> +		intel_modeset_setup_hw_state(dev);
> +		i915_redisable_vga(dev);
> +	}
> +
> +	if (!state)
> +		return 0;
> +
> +	for_each_crtc_in_state(state, crtc, crtc_state, i) {
> +		/*
> +		 * Force recalculation even if we restore
> +		 * current state. With fast modeset this may not result
> +		 * in a modeset when the state is compatible.
> +		 */
> +		crtc_state->mode_changed = true;
> +	}
> +
> +	return drm_atomic_commit(state);

To go along with my suggestion about dealing with -EDEADLK upfront,
we could then have a WARN_ON(ret == -EDEADLK) here to make sure we
didn't make a mess of things.

> +}
> +
>  void intel_prepare_reset(struct drm_device *dev)
>  {
> +	struct drm_atomic_state *state;
> +	struct drm_modeset_acquire_ctx *pctx;
> +	int ret;
> +
>  	/* no reset support for gen2 */
>  	if (IS_GEN2(dev))
>  		return;
> @@ -3137,16 +3173,39 @@ void intel_prepare_reset(struct drm_device *dev)
>  		return;
>  
>  	drm_modeset_lock_all(dev);
> +	pctx = dev->mode_config.acquire_ctx;

Still looks like "power context" to me.

I was wondering if we should have a private context here, but I guess
that would fall over with some mode_config.acquire_ctx usage in the
display init code?

> +
>  	/*
>  	 * Disabling the crtcs gracefully seems nicer. Also the
>  	 * g33 docs say we should at least disable all the planes.
>  	 */
> -	intel_display_suspend(dev);
> +
> +	state = drm_atomic_helper_duplicate_state(dev, pctx);
> +	if (IS_ERR(state)) {
> +		ret = PTR_ERR(state);
> +		state = NULL;
> +		DRM_ERROR("Duplicating state failed with %i\n", ret);
> +		goto err;
> +	}
> +
> +	ret = drm_atomic_helper_disable_all(dev, pctx);
> +	if (ret) {
> +		DRM_ERROR("Suspending crtc's failed with %i\n", ret);
> +		goto err;
> +	}
> +
> +	to_i915(dev)->modeset_restore_state = state;
> +	return;
> +
> +err:
> +	drm_atomic_state_free(state);
>  }
>  
>  void intel_finish_reset(struct drm_device *dev)
>  {
>  	struct drm_i915_private *dev_priv = to_i915(dev);
> +	struct drm_atomic_state *state = dev_priv->modeset_restore_state;
> +	int ret;
>  
>  	/*
>  	 * Flips in the rings will be nuked by the reset,
> @@ -3172,23 +3231,29 @@ void intel_finish_reset(struct drm_device *dev)
>  		 */
>  		intel_update_primary_planes(dev);
>  		return;
> -	}
> +	} else {
> +		/*
> +		 * The display has been reset as well,
> +		 * so need a full re-initialization.
> +		 */
> +		intel_runtime_pm_disable_interrupts(dev_priv);
> +		intel_runtime_pm_enable_interrupts(dev_priv);
>  
> -	/*
> -	 * The display has been reset as well,
> -	 * so need a full re-initialization.
> -	 */
> -	intel_runtime_pm_disable_interrupts(dev_priv);
> -	intel_runtime_pm_enable_interrupts(dev_priv);
> +		intel_modeset_init_hw(dev);
>  
> -	intel_modeset_init_hw(dev);
> +		spin_lock_irq(&dev_priv->irq_lock);
> +		if (dev_priv->display.hpd_irq_setup)
> +			dev_priv->display.hpd_irq_setup(dev);
> +		spin_unlock_irq(&dev_priv->irq_lock);
> +	}

This seems like an unrelated change. More to do with the second patch I
guess, though I'm still wondering if we can't just run through the
whole thing every time?

>  
> -	spin_lock_irq(&dev_priv->irq_lock);
> -	if (dev_priv->display.hpd_irq_setup)
> -		dev_priv->display.hpd_irq_setup(dev);
> -	spin_unlock_irq(&dev_priv->irq_lock);
> +	dev_priv->modeset_restore_state = NULL;
> +	if (state)
> +		state->acquire_ctx = dev->mode_config.acquire_ctx;
>  
> -	intel_display_resume(dev);
> +	ret = __intel_display_resume(dev, state, NULL);
> +	if (ret)
> +		DRM_ERROR("Restoring old state failed with %i\n", ret);
>  
>  	intel_hpd_init(dev_priv);
>  
> @@ -15885,6 +15950,8 @@ void intel_display_resume(struct drm_device *dev)
>  	bool setup = false;
>  
>  	dev_priv->modeset_restore_state = NULL;
> +	if (state)
> +		state->acquire_ctx = &ctx;
>  
>  	/*
>  	 * This is a cludge because with real atomic modeset mode_config.mutex
> @@ -15897,32 +15964,8 @@ void intel_display_resume(struct drm_device *dev)
>  
>  retry:
>  	ret = drm_modeset_lock_all_ctx(dev, &ctx);
> -
> -	if (ret == 0 && !setup) {
> -		setup = true;
> -
> -		intel_modeset_setup_hw_state(dev);
> -		i915_redisable_vga(dev);
> -	}
> -
> -	if (ret == 0 && state) {
> -		struct drm_crtc_state *crtc_state;
> -		struct drm_crtc *crtc;
> -		int i;
> -
> -		state->acquire_ctx = &ctx;
> -
> -		for_each_crtc_in_state(state, crtc, crtc_state, i) {
> -			/*
> -			 * Force recalculation even if we restore
> -			 * current state. With fast modeset this may not result
> -			 * in a modeset when the state is compatible.
> -			 */
> -			crtc_state->mode_changed = true;
> -		}
> -
> -		ret = drm_atomic_commit(state);
> -	}
> +	if (!ret)
> +		ret = __intel_display_resume(dev, state, &setup);
>  
>  	if (ret == -EDEADLK) {
>  		drm_modeset_backoff(&ctx);
> -- 
> 2.1.0

-- 
Ville Syrjälä
Intel OTC