Message-ID: <5716E757.8070702@rock-chips.com>
Date: Wed, 20 Apr 2016 10:20:07 +0800
From: Mark yao <mark.yao@rock-chips.com>
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:31.0) Gecko/20100101 Thunderbird/31.8.0
MIME-Version: 1.0
To: John Keeping <john@metanate.com>
CC: David Airlie <airlied@linux.ie>, Heiko Stuebner <heiko@sntech.de>,
        dri-devel@lists.freedesktop.org, linux-arm-kernel@lists.infradead.org,
        linux-rockchip@lists.infradead.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH] drm/rockchip: get rid of rockchip_drm_crtc_mode_config
References: <1460948611-32259-1-git-send-email-mark.yao@rock-chips.com>	<20160418102538.0ed730ca.john@metanate.com>	<57159B37.9050806@rock-chips.com> <20160419100630.06518b63.john@metanate.com>
In-Reply-To: <20160419100630.06518b63.john@metanate.com>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: 8bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 4061
Lines: 116

On 2016年04月19日 17:06, John Keeping wrote:
> On Tue, 19 Apr 2016 10:43:03 +0800, Mark yao wrote:
>
>> On 2016年04月18日 17:25, John Keeping wrote:
>>> On Mon, 18 Apr 2016 11:03:31 +0800, Mark Yao wrote:
>>>   
>>>>> We need to take care of the vop status when use
>>>>> rockchip_drm_crtc_mode_config, if vop is disabled,
>>>>> the function would failed, that is terrible.
>>>>>
>>>>> Save connector type and output mode on drm_display_mode->private_flags
>>>>> at encoder mode_fixup, then we can configure the type and mode safely
>>>>> on crtc mode_set.
>>> Since Rockchip is atomic, shouldn't this be using atomic_check hooks and
>>> a subclassed crtc_state structure?
>>>   
>> I try to use atomic_check with crtc_state, but it seems not easy,
>> there are two value need transmit from connector to vop: connector type
>> and out_mode
>>
>> the connector type I think we can loop the atomic state to find the
>> connector type.
>> but the out_mode is a custom value, I can't find a generic way to
>> transmit it with atomic state.
>>
>> BTW, I think on atomic side, the drm_display_mode is under control by
>> atomic state,
>> and the mode->private_flags is not use by drm framework, I found i915
>> and gma500 also use
>> mode->private_flags to transmit custom value.
>>
>> So I think it's no problem using mode->private_flags.
> The documentation for drm_display_mode::private says:
>
>      It shouldn't be used by atomic drivers since they can store any
>      additional data by subclassing state structures.
>
> This applies to private_flags as well.
>
> I think this means that we should do something like the patch below
> (which isn't even compile tested and doesn't cover implementing
> drm_encoder_helper_funcs::atomic_check on all of the encoders):
>
> -- >8 --
> diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_drv.h b/drivers/gpu/drm/rockchip/rockchip_drm_drv.h
> index 00d17d71aa4c..9d65fa9188f4 100644
> --- a/drivers/gpu/drm/rockchip/rockchip_drm_drv.h
> +++ b/drivers/gpu/drm/rockchip/rockchip_drm_drv.h
> @@ -31,6 +31,14 @@
>   struct drm_device;
>   struct drm_connector;
>   
> +struct rockchip_crtc_state {
> +	struct drm_crtc_state base;
> +	int output_type;
> +	int output_mode;
> +};
> +
> +#define to_rockchip_crtc_state(s) container_of(x, struct rockchip_crtc_state, base)
> +
>   /*
>    * Rockchip drm private crtc funcs.
>    * @enable_vblank: enable crtc vblank irq.
> diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_vop.c b/drivers/gpu/drm/rockchip/rockchip_drm_vop.c
> index a619f120f801..5cdf3123e1eb 100644
> --- a/drivers/gpu/drm/rockchip/rockchip_drm_vop.c
> +++ b/drivers/gpu/drm/rockchip/rockchip_drm_vop.c
> @@ -1044,13 +1044,34 @@ static void vop_crtc_destroy(struct drm_crtc *crtc)
>   	drm_crtc_cleanup(crtc);
>   }
>   
> +static struct drm_crtc_state *vop_crtc_duplicate_state(struct drm_crtc *crtc)
> +{
> +	struct rockchip_crtc_state *rockchip_state;
> +
> +	rockchip_state = kzalloc(sizeof(*rockchip_state), GFP_KERNEL);
> +	if (!rockchip_state)
> +		return NULL;
> +
> +	__drm_atomic_helper_crtc_duplicate_state(crtc, &rockchip_state->base);
> +	return &rockchip_state->base;
> +}
> +
> +static void vop_crtc_destroy_state(struct drm_crtc *crtc,
> +				   struct drm_crtc_state *state)
> +{
> +	struct rockchip_crtc_state *s = to_rockchip_crtc_state(state);
> +
> +	__drm_atomic_helper_crtc_destroy_state(crtc, &s->base);
> +	kfree(s);
> +}
> +
>   static const struct drm_crtc_funcs vop_crtc_funcs = {
>   	.set_config = drm_atomic_helper_set_config,
>   	.page_flip = drm_atomic_helper_page_flip,
>   	.destroy = vop_crtc_destroy,
>   	.reset = drm_atomic_helper_crtc_reset,
> -	.atomic_duplicate_state = drm_atomic_helper_crtc_duplicate_state,
> -	.atomic_destroy_state = drm_atomic_helper_crtc_destroy_state,
> +	.atomic_duplicate_state = vop_crtc_duplicate_state,
> +	.atomic_destroy_state = vop_crtc_destroy_state,
>   };
>   
>   static bool vop_win_pending_is_complete(struct vop_win *vop_win)
>
>
>
Hi John

Good, I had implement it according to  your advise.

Thanks.

-- 
Ｍark Yao