[PATCH] drm/ssd130x: Iterate over damage clips instead of using a merged rect

[PATCH] drm/ssd130x: Iterate over damage clips instead of using a merged rect

[Javier Martinez Canillas]

The drm_atomic_helper_damage_merged() helper merges all the damage clips
into one rectangle. If there are multiple damage clips that aren't close
to each other, the resulting rectangle could be quite big.

Instead of using that function helper, iterate over all the damage clips
and update them one by one.

Suggested-by: Jocelyn Falempe <jfalempe@redhat.com>
Signed-off-by: Javier Martinez Canillas <javierm@redhat.com>
---

 drivers/gpu/drm/solomon/ssd130x.c | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)

diff --git a/drivers/gpu/drm/solomon/ssd130x.c b/drivers/gpu/drm/solomon/ssd130x.c
index bc41a5ae810a..2428f1813a8d 100644
--- a/drivers/gpu/drm/solomon/ssd130x.c
+++ b/drivers/gpu/drm/solomon/ssd130x.c
@@ -578,21 +578,23 @@ static void ssd130x_primary_plane_helper_atomic_update(struct drm_plane *plane,
 	struct drm_plane_state *plane_state = drm_atomic_get_new_plane_state(state, plane);
 	struct drm_plane_state *old_plane_state = drm_atomic_get_old_plane_state(state, plane);
 	struct drm_shadow_plane_state *shadow_plane_state = to_drm_shadow_plane_state(plane_state);
+	struct drm_atomic_helper_damage_iter iter;
 	struct drm_device *drm = plane->dev;
-	struct drm_rect src_clip, dst_clip;
+	struct drm_rect dst_clip;
+	struct drm_rect damage;
 	int idx;
 
-	if (!drm_atomic_helper_damage_merged(old_plane_state, plane_state, &src_clip))
-		return;
-
 	dst_clip = plane_state->dst;
-	if (!drm_rect_intersect(&dst_clip, &src_clip))
-		return;
-
 	if (!drm_dev_enter(drm, &idx))
 		return;
 
-	ssd130x_fb_blit_rect(plane_state->fb, &shadow_plane_state->data[0], &dst_clip);
+	drm_atomic_helper_damage_iter_init(&iter, old_plane_state, plane_state);
+	drm_atomic_for_each_plane_damage(&iter, &damage) {
+		if (!drm_rect_intersect(&dst_clip, &damage))
+			continue;
+
+		ssd130x_fb_blit_rect(plane_state->fb, &shadow_plane_state->data[0], &damage);
+	}
 
 	drm_dev_exit(idx);
 }
-- 
2.37.3

Re: [PATCH] drm/ssd130x: Iterate over damage clips instead of using a merged rect

[Thomas Zimmermann]

[-- Attachment #1.1: Type: text/plain, Size: 2753 bytes --]

Hi

Am 30.09.22 um 10:01 schrieb Javier Martinez Canillas:
> The drm_atomic_helper_damage_merged() helper merges all the damage clips
> into one rectangle. If there are multiple damage clips that aren't close
> to each other, the resulting rectangle could be quite big.
> 
> Instead of using that function helper, iterate over all the damage clips
> and update them one by one.
> 
> Suggested-by: Jocelyn Falempe <jfalempe@redhat.com>
> Signed-off-by: Javier Martinez Canillas <javierm@redhat.com>
> ---
> 
>   drivers/gpu/drm/solomon/ssd130x.c | 18 ++++++++++--------
>   1 file changed, 10 insertions(+), 8 deletions(-)
> 
> diff --git a/drivers/gpu/drm/solomon/ssd130x.c b/drivers/gpu/drm/solomon/ssd130x.c
> index bc41a5ae810a..2428f1813a8d 100644
> --- a/drivers/gpu/drm/solomon/ssd130x.c
> +++ b/drivers/gpu/drm/solomon/ssd130x.c
> @@ -578,21 +578,23 @@ static void ssd130x_primary_plane_helper_atomic_update(struct drm_plane *plane,
>   	struct drm_plane_state *plane_state = drm_atomic_get_new_plane_state(state, plane);
>   	struct drm_plane_state *old_plane_state = drm_atomic_get_old_plane_state(state, plane);
>   	struct drm_shadow_plane_state *shadow_plane_state = to_drm_shadow_plane_state(plane_state);
> +	struct drm_atomic_helper_damage_iter iter;
>   	struct drm_device *drm = plane->dev;
> -	struct drm_rect src_clip, dst_clip;
> +	struct drm_rect dst_clip;
> +	struct drm_rect damage;
>   	int idx;
>   
> -	if (!drm_atomic_helper_damage_merged(old_plane_state, plane_state, &src_clip))
> -		return;
> -
>   	dst_clip = plane_state->dst;
> -	if (!drm_rect_intersect(&dst_clip, &src_clip))
> -		return;
> -
>   	if (!drm_dev_enter(drm, &idx))
>   		return;
>   
> -	ssd130x_fb_blit_rect(plane_state->fb, &shadow_plane_state->data[0], &dst_clip);
> +	drm_atomic_helper_damage_iter_init(&iter, old_plane_state, plane_state);
> +	drm_atomic_for_each_plane_damage(&iter, &damage) {
> +		if (!drm_rect_intersect(&dst_clip, &damage))
> +			continue;

dst_clip will be overwritten here. So need to init it within the loop first.

> +
> +		ssd130x_fb_blit_rect(plane_state->fb, &shadow_plane_state->data[0], &damage);

In simpledrm, we adjust the destination address with dst_clip like this:

   iosys_map_incr(&dst, drm_fb_clip_offset(sdev->pitch, sdev->format, 
&dst_clip));

How does this work in ssd130x? You never use dst_clip to adjust to the 
changed location. Won't you have out-of-bounds writes on the device?

Best regards
Thomas


> +	}
>   
>   	drm_dev_exit(idx);
>   }

-- 
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Maxfeldstr. 5, 90409 Nürnberg, Germany
(HRB 36809, AG Nürnberg)
Geschäftsführer: Ivo Totev

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 840 bytes --]

Re: [PATCH] drm/ssd130x: Iterate over damage clips instead of using a merged rect

[Javier Martinez Canillas]

Hello Thomas,

Thanks a lot for your feedback.

On 9/30/22 10:26, Thomas Zimmermann wrote:
> Hi
> 
> Am 30.09.22 um 10:01 schrieb Javier Martinez Canillas:
>> The drm_atomic_helper_damage_merged() helper merges all the damage clips
>> into one rectangle. If there are multiple damage clips that aren't close
>> to each other, the resulting rectangle could be quite big.
>>
>> Instead of using that function helper, iterate over all the damage clips
>> and update them one by one.
>>
>> Suggested-by: Jocelyn Falempe <jfalempe@redhat.com>
>> Signed-off-by: Javier Martinez Canillas <javierm@redhat.com>
>> ---
>>
>>   drivers/gpu/drm/solomon/ssd130x.c | 18 ++++++++++--------
>>   1 file changed, 10 insertions(+), 8 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/solomon/ssd130x.c b/drivers/gpu/drm/solomon/ssd130x.c
>> index bc41a5ae810a..2428f1813a8d 100644
>> --- a/drivers/gpu/drm/solomon/ssd130x.c
>> +++ b/drivers/gpu/drm/solomon/ssd130x.c
>> @@ -578,21 +578,23 @@ static void ssd130x_primary_plane_helper_atomic_update(struct drm_plane *plane,
>>   	struct drm_plane_state *plane_state = drm_atomic_get_new_plane_state(state, plane);
>>   	struct drm_plane_state *old_plane_state = drm_atomic_get_old_plane_state(state, plane);
>>   	struct drm_shadow_plane_state *shadow_plane_state = to_drm_shadow_plane_state(plane_state);
>> +	struct drm_atomic_helper_damage_iter iter;
>>   	struct drm_device *drm = plane->dev;
>> -	struct drm_rect src_clip, dst_clip;
>> +	struct drm_rect dst_clip;
>> +	struct drm_rect damage;
>>   	int idx;
>>   
>> -	if (!drm_atomic_helper_damage_merged(old_plane_state, plane_state, &src_clip))
>> -		return;
>> -
>>   	dst_clip = plane_state->dst;
>> -	if (!drm_rect_intersect(&dst_clip, &src_clip))
>> -		return;
>> -
>>   	if (!drm_dev_enter(drm, &idx))
>>   		return;
>>   
>> -	ssd130x_fb_blit_rect(plane_state->fb, &shadow_plane_state->data[0], &dst_clip);
>> +	drm_atomic_helper_damage_iter_init(&iter, old_plane_state, plane_state);
>> +	drm_atomic_for_each_plane_damage(&iter, &damage) {
>> +		if (!drm_rect_intersect(&dst_clip, &damage))
>> +			continue;
> 
> dst_clip will be overwritten here. So need to init it within the loop first.
> 

Oh, indeed. I'll move it inside the loop. Thanks for catching this.

>> +
>> +		ssd130x_fb_blit_rect(plane_state->fb, &shadow_plane_state->data[0], &damage);
> 
> In simpledrm, we adjust the destination address with dst_clip like this:
> 
>    iosys_map_incr(&dst, drm_fb_clip_offset(sdev->pitch, sdev->format, 
> &dst_clip));
> 
> How does this work in ssd130x? You never use dst_clip to adjust to the 
> changed location. Won't you have out-of-bounds writes on the device?
> 

Right, in ssd130x what I do is:

static int ssd130x_fb_blit_rect(struct drm_framebuffer *fb, const struct iosys_map *vmap,
				struct drm_rect *rect)
{
	struct iosys_map dst;
...
	u8 *buf = NULL;
...
	buf = kcalloc(dst_pitch, drm_rect_height(rect), GFP_KERNEL);
...
	iosys_map_set_vaddr(&dst, buf);
	drm_fb_xrgb8888_to_mono(&dst, &dst_pitch, vmap, fb, rect);
...
	ssd130x_update_rect(ssd130x, buf, rect);
}

I understand that's correct too?

-- 
Best regards,

Javier Martinez Canillas
Core Platforms
Red Hat

Re: [PATCH] drm/ssd130x: Iterate over damage clips instead of using a merged rect

[Thomas Zimmermann]

[-- Attachment #1.1: Type: text/plain, Size: 3936 bytes --]

Hi

Am 30.09.22 um 11:25 schrieb Javier Martinez Canillas:
> Hello Thomas,
> 
> Thanks a lot for your feedback.
> 
> On 9/30/22 10:26, Thomas Zimmermann wrote:
>> Hi
>>
>> Am 30.09.22 um 10:01 schrieb Javier Martinez Canillas:
>>> The drm_atomic_helper_damage_merged() helper merges all the damage clips
>>> into one rectangle. If there are multiple damage clips that aren't close
>>> to each other, the resulting rectangle could be quite big.
>>>
>>> Instead of using that function helper, iterate over all the damage clips
>>> and update them one by one.
>>>
>>> Suggested-by: Jocelyn Falempe <jfalempe@redhat.com>
>>> Signed-off-by: Javier Martinez Canillas <javierm@redhat.com>
>>> ---
>>>
>>>    drivers/gpu/drm/solomon/ssd130x.c | 18 ++++++++++--------
>>>    1 file changed, 10 insertions(+), 8 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/solomon/ssd130x.c b/drivers/gpu/drm/solomon/ssd130x.c
>>> index bc41a5ae810a..2428f1813a8d 100644
>>> --- a/drivers/gpu/drm/solomon/ssd130x.c
>>> +++ b/drivers/gpu/drm/solomon/ssd130x.c
>>> @@ -578,21 +578,23 @@ static void ssd130x_primary_plane_helper_atomic_update(struct drm_plane *plane,
>>>    	struct drm_plane_state *plane_state = drm_atomic_get_new_plane_state(state, plane);
>>>    	struct drm_plane_state *old_plane_state = drm_atomic_get_old_plane_state(state, plane);
>>>    	struct drm_shadow_plane_state *shadow_plane_state = to_drm_shadow_plane_state(plane_state);
>>> +	struct drm_atomic_helper_damage_iter iter;
>>>    	struct drm_device *drm = plane->dev;
>>> -	struct drm_rect src_clip, dst_clip;
>>> +	struct drm_rect dst_clip;
>>> +	struct drm_rect damage;
>>>    	int idx;
>>>    
>>> -	if (!drm_atomic_helper_damage_merged(old_plane_state, plane_state, &src_clip))
>>> -		return;
>>> -
>>>    	dst_clip = plane_state->dst;
>>> -	if (!drm_rect_intersect(&dst_clip, &src_clip))
>>> -		return;
>>> -
>>>    	if (!drm_dev_enter(drm, &idx))
>>>    		return;
>>>    
>>> -	ssd130x_fb_blit_rect(plane_state->fb, &shadow_plane_state->data[0], &dst_clip);
>>> +	drm_atomic_helper_damage_iter_init(&iter, old_plane_state, plane_state);
>>> +	drm_atomic_for_each_plane_damage(&iter, &damage) {
>>> +		if (!drm_rect_intersect(&dst_clip, &damage))
>>> +			continue;
>>
>> dst_clip will be overwritten here. So need to init it within the loop first.
>>
> 
> Oh, indeed. I'll move it inside the loop. Thanks for catching this.
> 
>>> +
>>> +		ssd130x_fb_blit_rect(plane_state->fb, &shadow_plane_state->data[0], &damage);
>>
>> In simpledrm, we adjust the destination address with dst_clip like this:
>>
>>     iosys_map_incr(&dst, drm_fb_clip_offset(sdev->pitch, sdev->format,
>> &dst_clip));
>>
>> How does this work in ssd130x? You never use dst_clip to adjust to the
>> changed location. Won't you have out-of-bounds writes on the device?
>>
> 
> Right, in ssd130x what I do is:
> 
> static int ssd130x_fb_blit_rect(struct drm_framebuffer *fb, const struct iosys_map *vmap,
> 				struct drm_rect *rect)
> {
> 	struct iosys_map dst;
> ...
> 	u8 *buf = NULL;
> ...
> 	buf = kcalloc(dst_pitch, drm_rect_height(rect), GFP_KERNEL);
> ...
> 	iosys_map_set_vaddr(&dst, buf);
> 	drm_fb_xrgb8888_to_mono(&dst, &dst_pitch, vmap, fb, rect);
> ...
> 	ssd130x_update_rect(ssd130x, buf, rect);
> }
> 
> I understand that's correct too?

 From what I understand about ssd130x, blit_rect looks correct up to the 
call to update_rect.  The values in the rect parameter are for the 
damage area of the plane. In update_rect, the destination coords x and y 
are also taken from rect. But they should come from dst_clip, which is 
the on-screen location. Does that make sense?

Best regards
Thomas

> 

-- 
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Maxfeldstr. 5, 90409 Nürnberg, Germany
(HRB 36809, AG Nürnberg)
Geschäftsführer: Ivo Totev

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 840 bytes --]

Re: [PATCH] drm/ssd130x: Iterate over damage clips instead of using a merged rect

[Javier Martinez Canillas]

Hello Thomas,

On 9/30/22 12:11, Thomas Zimmermann wrote:

[...]

>>
>>>> +
>>>> +		ssd130x_fb_blit_rect(plane_state->fb, &shadow_plane_state->data[0], &damage);
>>>
>>> In simpledrm, we adjust the destination address with dst_clip like this:
>>>
>>>     iosys_map_incr(&dst, drm_fb_clip_offset(sdev->pitch, sdev->format,
>>> &dst_clip));
>>>
>>> How does this work in ssd130x? You never use dst_clip to adjust to the
>>> changed location. Won't you have out-of-bounds writes on the device?
>>>
>>
>> Right, in ssd130x what I do is:
>>
>> static int ssd130x_fb_blit_rect(struct drm_framebuffer *fb, const struct iosys_map *vmap,
>> 				struct drm_rect *rect)
>> {
>> 	struct iosys_map dst;
>> ...
>> 	u8 *buf = NULL;
>> ...
>> 	buf = kcalloc(dst_pitch, drm_rect_height(rect), GFP_KERNEL);
>> ...
>> 	iosys_map_set_vaddr(&dst, buf);
>> 	drm_fb_xrgb8888_to_mono(&dst, &dst_pitch, vmap, fb, rect);
>> ...
>> 	ssd130x_update_rect(ssd130x, buf, rect);
>> }
>>
>> I understand that's correct too?
> 
>  From what I understand about ssd130x, blit_rect looks correct up to the 
> call to update_rect.  The values in the rect parameter are for the 
> damage area of the plane. In update_rect, the destination coords x and y 
> are also taken from rect. But they should come from dst_clip, which is 
> the on-screen location. Does that make sense?
> 

I believe you are correct. Then what I should do is to not pass the damage
area to ssd130x_fb_blit_rect() as the struct drm_rect argument but instead
the dst_clip as filled by drm_rect_intersect(). Does that sound correct ?

In other words, the following:

	drm_atomic_helper_damage_iter_init(&iter, old_plane_state, plane_state);
	drm_atomic_for_each_plane_damage(&iter, &damage) {
		dst_clip = plane_state->dst;
		if (!drm_rect_intersect(&dst_clip, &damage))
			continue;

		ssd130x_fb_blit_rect(plane_state->fb, &shadow_plane_state->data[0], &dst_clip);
	}

-- 
Best regards,

Javier Martinez Canillas
Core Platforms
Red Hat