Subject: Re: [PATCH] drm: atmel-hlcdc: use a default gamma ramp if none is
 specified
To: Boris Brezillon <boris.brezillon@free-electrons.com>
Cc: linux-kernel@vger.kernel.org, David Airlie <airlied@linux.ie>,
        dri-devel@lists.freedesktop.org,
        Alexandre Belloni <alexandre.belloni@free-electrons.com>,
        Nicolas Ferre <nicolas.ferre@microchip.com>,
        Daniel Vetter <daniel.vetter@ffwll.ch>
References: <1499074930-10187-1-git-send-email-peda@axentia.se>
 <20170703133102.5788d2f2@bbrezillon>
 <b8ed51f9-89b5-fbe4-309e-3c9e7f4a89ed@axentia.se>
 <20170703140235.53329a09@bbrezillon>
From: Peter Rosin <peda@axentia.se>
Organization: Axentia Technologies AB
Message-ID: <ce26dca6-c9d6-8e6c-4d25-4ba4cc83de96@axentia.se>
Date: Mon, 3 Jul 2017 22:59:36 +0200
User-Agent: Mozilla/5.0 (Windows NT 10.0; WOW64; rv:52.0) Gecko/20100101
 Thunderbird/52.2.1
MIME-Version: 1.0
In-Reply-To: <20170703140235.53329a09@bbrezillon>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
SpamDiagnosticOutput: 1:99
SpamDiagnosticMetadata: NSPM
X-MS-Exchange-CrossTenant-OriginalArrivalTime: 03 Jul 2017 20:59:39.9719 (UTC)
X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted
X-MS-Exchange-Transport-CrossTenantHeadersStamped: DB6PR0202MB2552
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org
Content-Length: 3198
Lines: 78

On 2017-07-03 14:02, Boris Brezillon wrote:
> On Mon, 3 Jul 2017 13:53:28 +0200
> Peter Rosin <peda@axentia.se> wrote:
> 
>> On 2017-07-03 13:31, Boris Brezillon wrote:
>>> On Mon,  3 Jul 2017 11:42:10 +0200
>>> Peter Rosin <peda@axentia.se> wrote:
>>>   
>>>> At init and if the gamma_lut property is ever removed, the clut
>>>> registers must be programmed with a default gamma ramp instead of
>>>> being left in some unknown state.
>>>>
>>>> Fixes: 364a7bf574eb ("drm: atmel-hlcdc: add support for 8-bit color lookup table mode")
>>>> Signed-off-by: Peter Rosin <peda@axentia.se>
>>>> ---
>>>>  drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_plane.c | 17 ++++++++++++++++-
>>>>  1 file changed, 16 insertions(+), 1 deletion(-)
>>>>
>>>> diff --git a/drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_plane.c b/drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_plane.c
>>>> index b5bd9b0..0ccd93c 100644
>>>> --- a/drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_plane.c
>>>> +++ b/drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_plane.c
>>>> @@ -429,6 +429,14 @@ static void atmel_hlcdc_plane_update_format(struct atmel_hlcdc_plane *plane,
>>>>  				    ATMEL_HLCDC_LAYER_FORMAT_CFG, cfg);
>>>>  }
>>>>  
>>>> +static void atmel_hlcdc_default_gamma_ramp(struct atmel_hlcdc_layer *layer)
>>>> +{
>>>> +	int idx;
>>>> +
>>>> +	for (idx = 0; idx < ATMEL_HLCDC_CLUT_SIZE; idx++)
>>>> +		atmel_hlcdc_layer_write_clut(layer, idx, idx * 0x10101);
>>>> +}
>>>> +
>>>>  static void atmel_hlcdc_plane_update_clut(struct atmel_hlcdc_plane *plane)
>>>>  {
>>>>  	struct drm_crtc *crtc = plane->base.crtc;
>>>> @@ -438,9 +446,14 @@ static void atmel_hlcdc_plane_update_clut(struct atmel_hlcdc_plane *plane)
>>>>  	if (!crtc || !crtc->state)
>>>>  		return;
>>>>  
>>>> -	if (!crtc->state->color_mgmt_changed || !crtc->state->gamma_lut)
>>>> +	if (!crtc->state->color_mgmt_changed)
>>>>  		return;
>>>>  
>>>> +	if (!crtc->state->gamma_lut) {
>>>> +		atmel_hlcdc_default_gamma_ramp(&plane->layer);  
>>>
>>> Hm, I'd prefer to have state->gamma_lut properly initialized in
>>> atmel_hlcdc_crtc_reset(), this way you don't have to do that in the
>>> update path.  
>>
>> The gamma_lut property can be removed, so you have to handle it here anyway. No?
> 
> Hm, what do you mean by removed? AFAICT, a property, once attached to
> a DRM object, exists until the DRM object is destroyed. The data
> attached to this property (here, the gamma_lut array) can be NULL, but
> once you have allocated the data container, it will be duplicated (and
> possibly updated) every time an atomic operation is triggered.

By remove I meant someone somehow triggering a call like this:

	drm_atomic_crtc_set_property(crtc, ...->gamma_lut_property, 0);

By the looks of it, that is not happening? But I'm not sure of that, and
even if it's not happening today, that may change...

> By initializing this field in crtc->reset(), you enforce the
> default/reset state, which IIUC, is what you want here.

Agreed, if it is not possible to remove/clear out the gamma_lut property,
then the only thing needed is to initialize the hw clut registers to the
linear ramp somewhere early.

But is it indeed a fact that there is no way to clear out the gamma_lut prop?

Cheers,
peda