[PATCH] [media] ov2640: add support for async device registration

[PATCH] [media] ov2640: add support for async device registration

[Josh Wu]

Move the clock detection code to the beginning of the probe().
If we meet any error in the clock detecting, then defer the probe.

Signed-off-by: Josh Wu <josh.wu@atmel.com>
---
 drivers/media/i2c/soc_camera/ov2640.c |   43 +++++++++++++++++++++------------
 1 file changed, 28 insertions(+), 15 deletions(-)

diff --git a/drivers/media/i2c/soc_camera/ov2640.c b/drivers/media/i2c/soc_camera/ov2640.c
index 6c6b1c3..fb9b6e9 100644
--- a/drivers/media/i2c/soc_camera/ov2640.c
+++ b/drivers/media/i2c/soc_camera/ov2640.c
@@ -22,6 +22,7 @@
 #include <linux/videodev2.h>
 
 #include <media/soc_camera.h>
+#include <media/v4l2-async.h>
 #include <media/v4l2-clk.h>
 #include <media/v4l2-subdev.h>
 #include <media/v4l2-ctrls.h>
@@ -1069,6 +1070,7 @@ static int ov2640_probe(struct i2c_client *client,
 	struct ov2640_priv	*priv;
 	struct soc_camera_subdev_desc *ssdd = soc_camera_i2c_to_desc(client);
 	struct i2c_adapter	*adapter = to_i2c_adapter(client->dev.parent);
+	struct v4l2_clk		*clk;
 	int			ret;
 
 	if (!ssdd) {
@@ -1083,13 +1085,20 @@ static int ov2640_probe(struct i2c_client *client,
 		return -EIO;
 	}
 
+	clk = v4l2_clk_get(&client->dev, "mclk");
+	if (IS_ERR(clk))
+		return -EPROBE_DEFER;
+
 	priv = devm_kzalloc(&client->dev, sizeof(struct ov2640_priv), GFP_KERNEL);
 	if (!priv) {
 		dev_err(&adapter->dev,
 			"Failed to allocate memory for private data!\n");
-		return -ENOMEM;
+		ret = -ENOMEM;
+		goto err_kzalloc;
 	}
 
+	priv->clk = clk;
+
 	v4l2_i2c_subdev_init(&priv->subdev, client, &ov2640_subdev_ops);
 	v4l2_ctrl_handler_init(&priv->hdl, 2);
 	v4l2_ctrl_new_std(&priv->hdl, &ov2640_ctrl_ops,
@@ -1097,23 +1106,26 @@ static int ov2640_probe(struct i2c_client *client,
 	v4l2_ctrl_new_std(&priv->hdl, &ov2640_ctrl_ops,
 			V4L2_CID_HFLIP, 0, 1, 1, 0);
 	priv->subdev.ctrl_handler = &priv->hdl;
-	if (priv->hdl.error)
-		return priv->hdl.error;
-
-	priv->clk = v4l2_clk_get(&client->dev, "mclk");
-	if (IS_ERR(priv->clk)) {
-		ret = PTR_ERR(priv->clk);
-		goto eclkget;
+	if (priv->hdl.error) {
+		ret = priv->hdl.error;
+		goto err_kzalloc;
 	}
 
 	ret = ov2640_video_probe(client);
-	if (ret) {
-		v4l2_clk_put(priv->clk);
-eclkget:
-		v4l2_ctrl_handler_free(&priv->hdl);
-	} else {
-		dev_info(&adapter->dev, "OV2640 Probed\n");
-	}
+	if (ret)
+		goto err_probe;
+
+	ret = v4l2_async_register_subdev(&priv->subdev);
+	if (ret)
+		goto err_probe;
+
+	dev_info(&adapter->dev, "OV2640 Probed\n");
+	return 0;
+
+err_probe:
+	v4l2_ctrl_handler_free(&priv->hdl);
+err_kzalloc:
+	v4l2_clk_put(clk);
 
 	return ret;
 }
@@ -1122,6 +1134,7 @@ static int ov2640_remove(struct i2c_client *client)
 {
 	struct ov2640_priv       *priv = to_ov2640(client);
 
+	v4l2_async_unregister_subdev(&priv->subdev);
 	v4l2_clk_put(priv->clk);
 	v4l2_device_unregister_subdev(&priv->subdev);
 	v4l2_ctrl_handler_free(&priv->hdl);
-- 
1.7.9.5

Re: [PATCH] [media] ov2640: add support for async device registration

[Sylwester Nawrocki]

Hi Josh,

On 03/14/2014 11:12 AM, Josh Wu wrote:
> +	clk = v4l2_clk_get(&client->dev, "mclk");
> +	if (IS_ERR(clk))
> +		return -EPROBE_DEFER;

You should instead make it:

		return PTR_ERR(clk);

But you will need this patch for that to work:
http://git.kernel.org/cgit/linux/kernel/git/next/linux-next.git/commit/drivers/clk/clk.c?id=a34cd4666f3da84228a82f70c94b8d9b692034ea

With this patch there is no need to overwrite any returned error
value with EPROBE_DEFER.

> @@ -1097,23 +1106,26 @@ static int ov2640_probe(struct i2c_client *client,
>   	v4l2_ctrl_new_std(&priv->hdl,&ov2640_ctrl_ops,
>   			V4L2_CID_HFLIP, 0, 1, 1, 0);
>   	priv->subdev.ctrl_handler =&priv->hdl;
> -	if (priv->hdl.error)
> -		return priv->hdl.error;
> -
> -	priv->clk = v4l2_clk_get(&client->dev, "mclk");
> -	if (IS_ERR(priv->clk)) {
> -		ret = PTR_ERR(priv->clk);
> -		goto eclkget;

--
Regards,
Sylwester

Re: [PATCH] [media] ov2640: add support for async device registration

[Josh Wu]

Hi, Sylwester

Thanks for your review.

On 3/15/2014 5:17 AM, Sylwester Nawrocki wrote:
> Hi Josh,
>
> On 03/14/2014 11:12 AM, Josh Wu wrote:
>> +    clk = v4l2_clk_get(&client->dev, "mclk");
>> +    if (IS_ERR(clk))
>> +        return -EPROBE_DEFER;
>
> You should instead make it:
>
>         return PTR_ERR(clk);
>
> But you will need this patch for that to work:
> http://git.kernel.org/cgit/linux/kernel/git/next/linux-next.git/commit/drivers/clk/clk.c?id=a34cd4666f3da84228a82f70c94b8d9b692034ea 
>
>
> With this patch there is no need to overwrite any returned error
> value with EPROBE_DEFER.

Thanks for the information. I will use this in v2 version.

Best Regards,
Josh Wu

>
>> @@ -1097,23 +1106,26 @@ static int ov2640_probe(struct i2c_client 
>> *client,
>>       v4l2_ctrl_new_std(&priv->hdl,&ov2640_ctrl_ops,
>>               V4L2_CID_HFLIP, 0, 1, 1, 0);
>>       priv->subdev.ctrl_handler =&priv->hdl;
>> -    if (priv->hdl.error)
>> -        return priv->hdl.error;
>> -
>> -    priv->clk = v4l2_clk_get(&client->dev, "mclk");
>> -    if (IS_ERR(priv->clk)) {
>> -        ret = PTR_ERR(priv->clk);
>> -        goto eclkget;
>
> -- 
> Regards,
> Sylwester

Re: [PATCH] [media] ov2640: add support for async device registration

[Sylwester Nawrocki]

Hi Josh,

On 19/03/14 10:17, Josh Wu wrote:
> On 3/15/2014 5:17 AM, Sylwester Nawrocki wrote:
>> > On 03/14/2014 11:12 AM, Josh Wu wrote:
>>> >> +    clk = v4l2_clk_get(&client->dev, "mclk");
>>> >> +    if (IS_ERR(clk))
>>> >> +        return -EPROBE_DEFER;
>> >
>> > You should instead make it:
>> >
>> >         return PTR_ERR(clk);
>> >
>> > But you will need this patch for that to work:
>> > http://git.kernel.org/cgit/linux/kernel/git/next/linux-next.git/commit/drivers/clk/clk.c?id=a34cd4666f3da84228a82f70c94b8d9b692034ea 
>> >
>> >
>> > With this patch there is no need to overwrite any returned error
>> > value with EPROBE_DEFER.
>
> Thanks for the information. I will use this in v2 version.

Oops, I missed somehow that it's v4l2_clk_get(), rather than
clk_get(). So it seems it will not work when you return PTR_ERR(clk),
since v4l2_clk_get() returns -ENODEV when clock is not found.
I think we should modify v4l2_clk_get() so it returns EPROBE_DEFER
rather than ENODEV on error. I anticipate v4l2_clk_get() might be
using clk_get() internally in future, and the v4l2 clk look up
will be used as a fallback only. So sensor drivers should just
do something like:

 clk = v4l2_clk_get(...);
 if (IS_ERR(clk))
	return PTR_ERR(clk);

--
Regards,
Sylwester

Re: [PATCH] [media] ov2640: add support for async device registration

[Josh Wu]

Hi, Sylwester

On 3/20/2014 10:44 PM, Sylwester Nawrocki wrote:
> Hi Josh,
>
> On 19/03/14 10:17, Josh Wu wrote:
>> On 3/15/2014 5:17 AM, Sylwester Nawrocki wrote:
>>>> On 03/14/2014 11:12 AM, Josh Wu wrote:
>>>>>> +    clk = v4l2_clk_get(&client->dev, "mclk");
>>>>>> +    if (IS_ERR(clk))
>>>>>> +        return -EPROBE_DEFER;
>>>> You should instead make it:
>>>>
>>>>          return PTR_ERR(clk);
>>>>
>>>> But you will need this patch for that to work:
>>>> http://git.kernel.org/cgit/linux/kernel/git/next/linux-next.git/commit/drivers/clk/clk.c?id=a34cd4666f3da84228a82f70c94b8d9b692034ea
>>>>
>>>>
>>>> With this patch there is no need to overwrite any returned error
>>>> value with EPROBE_DEFER.
>> Thanks for the information. I will use this in v2 version.
> Oops, I missed somehow that it's v4l2_clk_get(), rather than
> clk_get(). So it seems it will not work when you return PTR_ERR(clk),
> since v4l2_clk_get() returns -ENODEV when clock is not found.

right, I missed that. So this version is still valid one. The v2 that I 
already sent should be dropped.

> I think we should modify v4l2_clk_get() so it returns EPROBE_DEFER
> rather than ENODEV on error. I anticipate v4l2_clk_get() might be
> using clk_get() internally in future, and the v4l2 clk look up
> will be used as a fallback only. So sensor drivers should just
> do something like:
>
>   clk = v4l2_clk_get(...);
>   if (IS_ERR(clk))
> 	return PTR_ERR(clk);

I noticed that there are some driver like ov772x, ov9640 and etc, are 
not supported the async probe yet.
If we return the EPROBE_DEFER for all v4l2_clk_get(), that will cause 
the no-async probe function work incorrectly.
So IMO we can do your above suggestion after all sensors support async 
probe.

Thanks.
>
> --
> Regards,
> Sylwester

Best Regards,
Josh Wu