All of lore.kernel.org
 help / color / mirror / Atom feed
From: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
To: AngeloGioacchino Del Regno
	<angelogioacchino.delregno@collabora.com>,
	robdclark@gmail.com
Cc: freedreno@lists.freedesktop.org, airlied@linux.ie,
	linux-arm-msm@vger.kernel.org, konrad.dybcio@somainline.org,
	linux-kernel@vger.kernel.org, dri-devel@lists.freedesktop.org,
	jami.kettunen@somainline.org, maxime@cerno.tech,
	marijn.suijten@somainline.org, kernel@collabora.com,
	sean@poorly.run
Subject: Re: [PATCH] drm/msm: Initialize MDSS irq domain at probe time
Date: Mon, 29 Nov 2021 05:20:13 +0300	[thread overview]
Message-ID: <bf1540f1-4a9f-a9d6-d487-929107c487fd@linaro.org> (raw)
In-Reply-To: <20211125150947.354076-1-angelogioacchino.delregno@collabora.com>

Hi,

On 25/11/2021 18:09, AngeloGioacchino Del Regno wrote:
> Since commit 8f59ee9a570c ("drm/msm/dsi: Adjust probe order"), the
> DSI host gets initialized earlier, but this caused unability to probe
> the entire stack of components because they all depend on interrupts
> coming from the main `mdss` node (mdp5, or dpu1).
> 
> To fix this issue, also anticipate probing mdp5 or dpu1 by initializing
> them at msm_pdev_probe() time: this will make sure that we add the
> required interrupt controller mapping before dsi and/or other components
> try to initialize, finally satisfying the dependency.
> 
> While at it, also change the allocation of msm_drm_private to use the
> devm variant of kzalloc().
> 
> Fixes: 8f59ee9a570c ("drm/msm/dsi: Adjust probe order")
> Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>

I have been thinking about this. I do not feel that this is the correct 
approach. Currently DRM device exists only when all components are 
bound. If any of the subdevices is removed, corresponding component is 
delteted (and thus all components are unbound), the DRM device is taken 
down. This results in the state cleanup, userspace notifications, etc.

With your changes, DRM device will continue to exist even after one of 
subdevices is removed. This is not an expected behaviour, since 
subdrivers do not perform full cleanup, delegating that to DRM device 
takedown.

I suppose that proper solution would be to split msm_drv.c into into:
- generic components & drm code to be called from mdp4/mdp5/dpu driver 
(making mdp4, mdp5 or dpu1 the components master)

- bare mdss driver, taking care only about IRQs, OF devices population - 
calling proper mdss_init/mdss_destroy functions. Most probably we can 
drop this part altogether and just make md5_mdss.c/dpu_mdss.c proper 
platform drivers.

> ---
>   drivers/gpu/drm/msm/msm_drv.c | 81 ++++++++++++++++-------------------
>   1 file changed, 38 insertions(+), 43 deletions(-)
> 
> diff --git a/drivers/gpu/drm/msm/msm_drv.c b/drivers/gpu/drm/msm/msm_drv.c
> index 7936e8d498dd..790acf4993c0 100644
> --- a/drivers/gpu/drm/msm/msm_drv.c
> +++ b/drivers/gpu/drm/msm/msm_drv.c
> @@ -512,45 +512,12 @@ static int msm_init_vram(struct drm_device *dev)
>   static int msm_drm_init(struct device *dev, const struct drm_driver *drv)
>   {
>   	struct platform_device *pdev = to_platform_device(dev);
> -	struct drm_device *ddev;
> -	struct msm_drm_private *priv;
> -	struct msm_kms *kms;
> -	struct msm_mdss *mdss;
> +	struct drm_device *ddev = platform_get_drvdata(pdev);
> +	struct msm_drm_private *priv = ddev->dev_private;
> +	struct msm_kms *kms = priv->kms;
> +	struct msm_mdss *mdss = priv->mdss;
>   	int ret, i;
>   
> -	ddev = drm_dev_alloc(drv, dev);
> -	if (IS_ERR(ddev)) {
> -		DRM_DEV_ERROR(dev, "failed to allocate drm_device\n");
> -		return PTR_ERR(ddev);
> -	}
> -
> -	platform_set_drvdata(pdev, ddev);
> -
> -	priv = kzalloc(sizeof(*priv), GFP_KERNEL);
> -	if (!priv) {
> -		ret = -ENOMEM;
> -		goto err_put_drm_dev;
> -	}
> -
> -	ddev->dev_private = priv;
> -	priv->dev = ddev;
> -
> -	switch (get_mdp_ver(pdev)) {
> -	case KMS_MDP5:
> -		ret = mdp5_mdss_init(ddev);
> -		break;
> -	case KMS_DPU:
> -		ret = dpu_mdss_init(ddev);
> -		break;
> -	default:
> -		ret = 0;
> -		break;
> -	}
> -	if (ret)
> -		goto err_free_priv;
> -
> -	mdss = priv->mdss;
> -
>   	priv->wq = alloc_ordered_workqueue("msm", 0);
>   	priv->hangcheck_period = DRM_MSM_HANGCHECK_DEFAULT_PERIOD;
>   
> @@ -685,11 +652,6 @@ static int msm_drm_init(struct device *dev, const struct drm_driver *drv)
>   err_destroy_mdss:
>   	if (mdss && mdss->funcs)
>   		mdss->funcs->destroy(ddev);
> -err_free_priv:
> -	kfree(priv);
> -err_put_drm_dev:
> -	drm_dev_put(ddev);
> -	platform_set_drvdata(pdev, NULL);
>   	return ret;
>   }
>   
> @@ -1382,12 +1344,42 @@ static const struct component_master_ops msm_drm_ops = {
>   static int msm_pdev_probe(struct platform_device *pdev)
>   {
>   	struct component_match *match = NULL;
> +	struct msm_drm_private *priv;
> +	struct drm_device *ddev;
>   	int ret;
>   
> +	priv = devm_kzalloc(&pdev->dev, sizeof(*priv), GFP_KERNEL);
> +	if (!priv)
> +		return -ENOMEM;
> +
> +	ddev = drm_dev_alloc(&msm_driver, &pdev->dev);
> +	if (IS_ERR(ddev)) {
> +		DRM_DEV_ERROR(&pdev->dev, "failed to allocate drm_device\n");
> +		return PTR_ERR(ddev);
> +	}
> +
> +	platform_set_drvdata(pdev, ddev);
> +	ddev->dev_private = priv;
> +	priv->dev = ddev;
> +
> +	switch (get_mdp_ver(pdev)) {
> +	case KMS_MDP5:
> +		ret = mdp5_mdss_init(ddev);
> +		break;
> +	case KMS_DPU:
> +		ret = dpu_mdss_init(ddev);
> +		break;
> +	default:
> +		ret = 0;
> +		break;
> +	}
> +	if (ret)
> +		goto err_put_drm_dev;
> +
>   	if (get_mdp_ver(pdev)) {
>   		ret = add_display_components(pdev, &match);
>   		if (ret)
> -			return ret;
> +			goto fail;
>   	}
>   
>   	ret = add_gpu_components(&pdev->dev, &match);
> @@ -1409,6 +1401,9 @@ static int msm_pdev_probe(struct platform_device *pdev)
>   
>   fail:
>   	of_platform_depopulate(&pdev->dev);
> +err_put_drm_dev:
> +	drm_dev_put(ddev);
> +	platform_set_drvdata(pdev, NULL);
>   	return ret;
>   }
>   
> 


-- 
With best wishes
Dmitry

WARNING: multiple messages have this Message-ID (diff)
From: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
To: AngeloGioacchino Del Regno 
	<angelogioacchino.delregno@collabora.com>,
	robdclark@gmail.com
Cc: sean@poorly.run, airlied@linux.ie, daniel@ffwll.ch,
	maxime@cerno.tech, linux-arm-msm@vger.kernel.org,
	dri-devel@lists.freedesktop.org, freedreno@lists.freedesktop.org,
	linux-kernel@vger.kernel.org, kernel@collabora.com,
	konrad.dybcio@somainline.org, marijn.suijten@somainline.org,
	jami.kettunen@somainline.org
Subject: Re: [PATCH] drm/msm: Initialize MDSS irq domain at probe time
Date: Mon, 29 Nov 2021 05:20:13 +0300	[thread overview]
Message-ID: <bf1540f1-4a9f-a9d6-d487-929107c487fd@linaro.org> (raw)
In-Reply-To: <20211125150947.354076-1-angelogioacchino.delregno@collabora.com>

Hi,

On 25/11/2021 18:09, AngeloGioacchino Del Regno wrote:
> Since commit 8f59ee9a570c ("drm/msm/dsi: Adjust probe order"), the
> DSI host gets initialized earlier, but this caused unability to probe
> the entire stack of components because they all depend on interrupts
> coming from the main `mdss` node (mdp5, or dpu1).
> 
> To fix this issue, also anticipate probing mdp5 or dpu1 by initializing
> them at msm_pdev_probe() time: this will make sure that we add the
> required interrupt controller mapping before dsi and/or other components
> try to initialize, finally satisfying the dependency.
> 
> While at it, also change the allocation of msm_drm_private to use the
> devm variant of kzalloc().
> 
> Fixes: 8f59ee9a570c ("drm/msm/dsi: Adjust probe order")
> Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>

I have been thinking about this. I do not feel that this is the correct 
approach. Currently DRM device exists only when all components are 
bound. If any of the subdevices is removed, corresponding component is 
delteted (and thus all components are unbound), the DRM device is taken 
down. This results in the state cleanup, userspace notifications, etc.

With your changes, DRM device will continue to exist even after one of 
subdevices is removed. This is not an expected behaviour, since 
subdrivers do not perform full cleanup, delegating that to DRM device 
takedown.

I suppose that proper solution would be to split msm_drv.c into into:
- generic components & drm code to be called from mdp4/mdp5/dpu driver 
(making mdp4, mdp5 or dpu1 the components master)

- bare mdss driver, taking care only about IRQs, OF devices population - 
calling proper mdss_init/mdss_destroy functions. Most probably we can 
drop this part altogether and just make md5_mdss.c/dpu_mdss.c proper 
platform drivers.

> ---
>   drivers/gpu/drm/msm/msm_drv.c | 81 ++++++++++++++++-------------------
>   1 file changed, 38 insertions(+), 43 deletions(-)
> 
> diff --git a/drivers/gpu/drm/msm/msm_drv.c b/drivers/gpu/drm/msm/msm_drv.c
> index 7936e8d498dd..790acf4993c0 100644
> --- a/drivers/gpu/drm/msm/msm_drv.c
> +++ b/drivers/gpu/drm/msm/msm_drv.c
> @@ -512,45 +512,12 @@ static int msm_init_vram(struct drm_device *dev)
>   static int msm_drm_init(struct device *dev, const struct drm_driver *drv)
>   {
>   	struct platform_device *pdev = to_platform_device(dev);
> -	struct drm_device *ddev;
> -	struct msm_drm_private *priv;
> -	struct msm_kms *kms;
> -	struct msm_mdss *mdss;
> +	struct drm_device *ddev = platform_get_drvdata(pdev);
> +	struct msm_drm_private *priv = ddev->dev_private;
> +	struct msm_kms *kms = priv->kms;
> +	struct msm_mdss *mdss = priv->mdss;
>   	int ret, i;
>   
> -	ddev = drm_dev_alloc(drv, dev);
> -	if (IS_ERR(ddev)) {
> -		DRM_DEV_ERROR(dev, "failed to allocate drm_device\n");
> -		return PTR_ERR(ddev);
> -	}
> -
> -	platform_set_drvdata(pdev, ddev);
> -
> -	priv = kzalloc(sizeof(*priv), GFP_KERNEL);
> -	if (!priv) {
> -		ret = -ENOMEM;
> -		goto err_put_drm_dev;
> -	}
> -
> -	ddev->dev_private = priv;
> -	priv->dev = ddev;
> -
> -	switch (get_mdp_ver(pdev)) {
> -	case KMS_MDP5:
> -		ret = mdp5_mdss_init(ddev);
> -		break;
> -	case KMS_DPU:
> -		ret = dpu_mdss_init(ddev);
> -		break;
> -	default:
> -		ret = 0;
> -		break;
> -	}
> -	if (ret)
> -		goto err_free_priv;
> -
> -	mdss = priv->mdss;
> -
>   	priv->wq = alloc_ordered_workqueue("msm", 0);
>   	priv->hangcheck_period = DRM_MSM_HANGCHECK_DEFAULT_PERIOD;
>   
> @@ -685,11 +652,6 @@ static int msm_drm_init(struct device *dev, const struct drm_driver *drv)
>   err_destroy_mdss:
>   	if (mdss && mdss->funcs)
>   		mdss->funcs->destroy(ddev);
> -err_free_priv:
> -	kfree(priv);
> -err_put_drm_dev:
> -	drm_dev_put(ddev);
> -	platform_set_drvdata(pdev, NULL);
>   	return ret;
>   }
>   
> @@ -1382,12 +1344,42 @@ static const struct component_master_ops msm_drm_ops = {
>   static int msm_pdev_probe(struct platform_device *pdev)
>   {
>   	struct component_match *match = NULL;
> +	struct msm_drm_private *priv;
> +	struct drm_device *ddev;
>   	int ret;
>   
> +	priv = devm_kzalloc(&pdev->dev, sizeof(*priv), GFP_KERNEL);
> +	if (!priv)
> +		return -ENOMEM;
> +
> +	ddev = drm_dev_alloc(&msm_driver, &pdev->dev);
> +	if (IS_ERR(ddev)) {
> +		DRM_DEV_ERROR(&pdev->dev, "failed to allocate drm_device\n");
> +		return PTR_ERR(ddev);
> +	}
> +
> +	platform_set_drvdata(pdev, ddev);
> +	ddev->dev_private = priv;
> +	priv->dev = ddev;
> +
> +	switch (get_mdp_ver(pdev)) {
> +	case KMS_MDP5:
> +		ret = mdp5_mdss_init(ddev);
> +		break;
> +	case KMS_DPU:
> +		ret = dpu_mdss_init(ddev);
> +		break;
> +	default:
> +		ret = 0;
> +		break;
> +	}
> +	if (ret)
> +		goto err_put_drm_dev;
> +
>   	if (get_mdp_ver(pdev)) {
>   		ret = add_display_components(pdev, &match);
>   		if (ret)
> -			return ret;
> +			goto fail;
>   	}
>   
>   	ret = add_gpu_components(&pdev->dev, &match);
> @@ -1409,6 +1401,9 @@ static int msm_pdev_probe(struct platform_device *pdev)
>   
>   fail:
>   	of_platform_depopulate(&pdev->dev);
> +err_put_drm_dev:
> +	drm_dev_put(ddev);
> +	platform_set_drvdata(pdev, NULL);
>   	return ret;
>   }
>   
> 


-- 
With best wishes
Dmitry

  parent reply	other threads:[~2021-11-29  2:20 UTC|newest]

Thread overview: 30+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-11-25 15:09 [PATCH] drm/msm: Initialize MDSS irq domain at probe time AngeloGioacchino Del Regno
2021-11-25 15:09 ` AngeloGioacchino Del Regno
2021-11-25 15:23 ` Dmitry Baryshkov
2021-11-25 15:23   ` Dmitry Baryshkov
2021-11-25 22:39 ` Dmitry Baryshkov
2021-11-25 22:39   ` Dmitry Baryshkov
2021-11-26  0:06 ` Dmitry Baryshkov
2021-11-26  0:06   ` Dmitry Baryshkov
2021-11-26  9:26   ` AngeloGioacchino Del Regno
2021-11-26  9:26     ` AngeloGioacchino Del Regno
2021-11-26 21:12     ` Dmitry Baryshkov
2021-11-26 21:12       ` Dmitry Baryshkov
2021-11-26 16:08   ` AngeloGioacchino Del Regno
2021-11-26 16:08     ` AngeloGioacchino Del Regno
2021-11-29  2:20 ` Dmitry Baryshkov [this message]
2021-11-29  2:20   ` Dmitry Baryshkov
2021-11-29 14:14   ` AngeloGioacchino Del Regno
2021-11-29 14:14     ` AngeloGioacchino Del Regno
2021-11-29 14:53     ` Dmitry Baryshkov
2021-11-29 14:53       ` Dmitry Baryshkov
2021-11-29 14:56       ` AngeloGioacchino Del Regno
2021-11-29 14:56         ` AngeloGioacchino Del Regno
2021-12-01 10:52 [PATCH v3 2/2] " AngeloGioacchino Del Regno
2021-12-01 20:20 ` [PATCH] " Dmitry Baryshkov
2021-12-01 20:20   ` Dmitry Baryshkov
2021-12-03 10:43   ` AngeloGioacchino Del Regno
2021-12-03 10:43     ` AngeloGioacchino Del Regno
2021-12-03 13:14     ` Dmitry Baryshkov
2021-12-03 13:14       ` Dmitry Baryshkov
2021-12-03 13:17       ` AngeloGioacchino Del Regno
2021-12-03 13:17         ` AngeloGioacchino Del Regno

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=bf1540f1-4a9f-a9d6-d487-929107c487fd@linaro.org \
    --to=dmitry.baryshkov@linaro.org \
    --cc=airlied@linux.ie \
    --cc=angelogioacchino.delregno@collabora.com \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=freedreno@lists.freedesktop.org \
    --cc=jami.kettunen@somainline.org \
    --cc=kernel@collabora.com \
    --cc=konrad.dybcio@somainline.org \
    --cc=linux-arm-msm@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=marijn.suijten@somainline.org \
    --cc=maxime@cerno.tech \
    --cc=robdclark@gmail.com \
    --cc=sean@poorly.run \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.