Re: [PATCH 6/7] drm/bridge: Introduce early_enable and late disable

[Aradhya Bhatia]

Hi Maxime,

Thank you for reviewing the patches!

On 16/05/24 13:52, Maxime Ripard wrote:
> On Sat, May 11, 2024 at 09:00:50PM +0530, Aradhya Bhatia wrote:
>> diff --git a/include/drm/drm_bridge.h b/include/drm/drm_bridge.h
>> index 4baca0d9107b..40f93230abb2 100644
>> --- a/include/drm/drm_bridge.h
>> +++ b/include/drm/drm_bridge.h
>> @@ -206,6 +206,20 @@ struct drm_bridge_funcs {
>>  	 */
>>  	void (*post_disable)(struct drm_bridge *bridge);
>>  
>> +	/**
>> +	 * @late_disable:
>> +	 *
>> +	 * This callback should disable the bridge. It is called right after the
>> +	 * preceding element in the display pipe is disabled. If the preceding
>> +	 * element is a bridge this means it's called after that bridge's
>> +	 * @atomic_post_disable. If the preceding element is a &drm_crtc it's
>> +	 * called right after the crtc's &drm_crtc_helper_funcs.atomic_disable
>> +	 * hook.
>> +	 *
>> +	 * The @late_disable callback is optional.
>> +	 */
>> +	void (*late_disable)(struct drm_bridge *bridge);
>> +
>>  	/**
>>  	 * @mode_set:
>>  	 *
>> @@ -235,6 +249,26 @@ struct drm_bridge_funcs {
>>  	void (*mode_set)(struct drm_bridge *bridge,
>>  			 const struct drm_display_mode *mode,
>>  			 const struct drm_display_mode *adjusted_mode);
>> +
>> +	/**
>> +	 * @early_enable:
>> +	 *
>> +	 * This callback should enable the bridge. It is called right before
>> +	 * the preceding element in the display pipe is enabled. If the
>> +	 * preceding element is a bridge this means it's called before that
>> +	 * bridge's @early_enable. If the preceding element is a &drm_crtc it's
>> +	 * called right before the crtc's &drm_crtc_helper_funcs.atomic_enable
>> +	 * hook.
>> +	 *
>> +	 * The display pipe (i.e. clocks and timing signals) feeding this bridge
>> +	 * will not yet be running when this callback is called. The bridge can
>> +	 * enable the display link feeding the next bridge in the chain (if
>> +	 * there is one) when this callback is called.
>> +	 *
>> +	 * The @early_enable callback is optional.
>> +	 */
>> +	void (*early_enable)(struct drm_bridge *bridge);
>> +
> 
> You don't need the legacy option here, just go straight for the atomic one.

Got it! I can remove these in v2.

> 
>>  	/**
>>  	 * @pre_enable:
>>  	 *
>> @@ -285,6 +319,26 @@ struct drm_bridge_funcs {
>>  	 */
>>  	void (*enable)(struct drm_bridge *bridge);
>>  
>> +	/**
>> +	 * @atomic_early_enable:
>> +	 *
>> +	 * This callback should enable the bridge. It is called right before
>> +	 * the preceding element in the display pipe is enabled. If the
>> +	 * preceding element is a bridge this means it's called before that
>> +	 * bridge's @atomic_early_enable. If the preceding element is a
>> +	 * &drm_crtc it's called right before the crtc's
>> +	 * &drm_crtc_helper_funcs.atomic_enable hook.
>> +	 *
>> +	 * The display pipe (i.e. clocks and timing signals) feeding this bridge
>> +	 * will not yet be running when this callback is called. The bridge can
>> +	 * enable the display link feeding the next bridge in the chain (if
>> +	 * there is one) when this callback is called.
>> +	 *
>> +	 * The @early_enable callback is optional.
>> +	 */
>> +	void (*atomic_early_enable)(struct drm_bridge *bridge,
>> +				    struct drm_bridge_state *old_bridge_state);
>> +
>>  	/**
>>  	 * @atomic_pre_enable:
>>  	 *
>> @@ -361,6 +415,21 @@ struct drm_bridge_funcs {
>>  	void (*atomic_post_disable)(struct drm_bridge *bridge,
>>  				    struct drm_bridge_state *old_bridge_state);
>>  
>> +	/**
>> +	 * @atomic_late_disable:
>> +	 *
>> +	 * This callback should disable the bridge. It is called right after the
>> +	 * preceding element in the display pipe is disabled. If the preceding
>> +	 * element is a bridge this means it's called after that bridge's
>> +	 * @atomic_late_disable. If the preceding element is a &drm_crtc it's
>> +	 * called right after the crtc's &drm_crtc_helper_funcs.atomic_disable
>> +	 * hook.
>> +	 *
>> +	 * The @atomic_late_disable callback is optional.
>> +	 */
>> +	void (*atomic_late_disable)(struct drm_bridge *bridge,
>> +				    struct drm_bridge_state *old_bridge_state);
>> +
> 
> But more importantly, I don't quite get the use case you're trying to
> solve here.
> 
> If I got the rest of your series, the Cadence DSI bridge needs to be
> powered up before its source is started. You can't use atomic_enable or
> atomic_pre_enable because it would start the source before the DSI
> bridge. Is that correct?
> 

That's right. I cannot use bridge_atomic_pre_enable /
bridge_atomic_enable here. But that's because my source is CRTC, which
gets enabled via crtc_atomic_enable.


> If it is, then how is it different from what
> drm_atomic_bridge_chain_pre_enable is doing? The assumption there is
> that it starts enabling bridges last to first, to it should be enabled
> before anything starts.
> 
> The whole bridge enabling order code starts to be a bit of a mess, so it
> would be great if you could list all the order variations we have
> currently, and why none work for cdns-dsi.
> 

Of course! I can elaborate on the order.

Without my patches (and given there isn't any bridge setting the
"pre_enable_prev_first" flag) the order of enable for any single display
chain, looks like this -

	crtc_enable
	
	bridge[n]_pre_enable
	---
	bridge[1]_pre_enable

	encoder_enable

	bridge[1]_enable
	---
	bridge[n]_enable

The tidss enables at the crtc_enable level, and hence is the first
entity with stream on. cdns-dsi doesn't stand a chance with
bridge_atmoic_pre_enable / bridge_atmoic_enable hooks. And there is no
bridge call happening before crtc currently.

That's where the early_enable APIs come into picture. They are being
called before the crtc is even enabled, helping cdns-dsi to be enabled
before tidss. The order then looks like -


	bridge[1]_early_enable
	---
	bridge[n]_early_enable

	crtc_enable
	---
	(the order is same from here on)


Regards
Aradhya