Skip to contentSkip to conversation

about open/release and vidioc_g_input/vidioc_s_input functions

question Alexey Klimov 3 participants 7 messages Mar 23, 2009 – Mar 27, 2009 Media (V4L/DVB)

lore
Share & tools
lore
#0 lore raw

Hello, all

After last convertion of radio drivers to use v4l2_device we have such code in many radio drivers:
(it's radio-terratec.c for example)

...
static int terratec_open(struct file *file)

{
        return 0;
}

static int terratec_release(struct file *file)

{
        return 0;
}

...

and

...
static int vidioc_g_input(struct file *filp, void *priv, unsigned int *i)

{
        *i = 0;
        return 0;
}

static int vidioc_s_input(struct file *filp, void *priv, unsigned int i)

{
        return i ? -EINVAL : 0;
}

...

Such code used in many radio-drivers as i understand.

Is it good to place this empty and almost empty functions in:
(here i see two variants)

1) In header file that be in linux/drivers/media/radio/ directory. Later, we can move some generic/or repeating code in this header.

2) In any v4l header. What header may contain this ?

?

For what ? Well, as i understand we can decrease amount of lines and provide this simple generic functions. It's like video_device_release_empty function behaviour. Maybe not only radio drivers can use such vidioc_g_input and vidioc_s_input.

Is it worth ?

Signature · 1 line
Best regards, Klimov Alexey
#1 lore raw
Quoted 44 lines · On Tuesday 24 March 2009 00:14:07 Alexey Klimov wrote:

Hello, all

After last convertion of radio drivers to use v4l2_device we have such code in many radio drivers:
(it's radio-terratec.c for example)

...
static int terratec_open(struct file *file)

{
        return 0;
}

static int terratec_release(struct file *file)

{
        return 0;
}

...

and

...
static int vidioc_g_input(struct file *filp, void *priv, unsigned int *i)

{
        *i = 0;
        return 0;
}

static int vidioc_s_input(struct file *filp, void *priv, unsigned int i)

{
        return i ? -EINVAL : 0;
}

...

Such code used in many radio-drivers as i understand.

Is it good to place this empty and almost empty functions in:
(here i see two variants)

1) In header file that be in linux/drivers/media/radio/ directory. Later, we can move some generic/or repeating code in this header.

2) In any v4l header. What header may contain this ?

Quoted 8 lines

?

For what ? Well, as i understand we can decrease amount of lines and provide this simple generic functions. It's like video_device_release_empty function behaviour. Maybe not only radio drivers can use such vidioc_g_input and vidioc_s_input.

Is it worth ?

I don't think it is worth doing this for g/s_input. I think it is useful to have them here: it makes it very clear that there is just a single input and the overhead in both lines and actual bytes is minimal.

But for the empty open and release functions you could easily handle that in v4l2-dev.c: if you leave the open and release callbacks to NULL, then v4l2_open and v4l2_release can just return 0. That would be nice.

Regards,

Hans

Signature · 1 line
Hans Verkuil - video4linux developer - sponsored by TANDBERG
#2 lore raw
Quoted 52 lines · On Tuesday 24 March 2009 08:06:39 Hans Verkuil wrote:
Quoted 51 lines · On Tuesday 24 March 2009 00:14:07 Alexey Klimov wrote:

Hello, all

After last convertion of radio drivers to use v4l2_device we have such code in many radio drivers:
(it's radio-terratec.c for example)

...
static int terratec_open(struct file *file)

{
        return 0;
}

static int terratec_release(struct file *file)

{
        return 0;
}

...

and

...
static int vidioc_g_input(struct file *filp, void *priv, unsigned int *i)

{
        *i = 0;
        return 0;
}

static int vidioc_s_input(struct file *filp, void *priv, unsigned int i)

{
        return i ? -EINVAL : 0;
}

...

Such code used in many radio-drivers as i understand.

Is it good to place this empty and almost empty functions in:
(here i see two variants)

1) In header file that be in linux/drivers/media/radio/ directory. Later, we can move some generic/or repeating code in this header.

2) In any v4l header. What header may contain this ?

?

For what ? Well, as i understand we can decrease amount of lines and provide this simple generic functions. It's like video_device_release_empty function behaviour. Maybe not only radio drivers can use such vidioc_g_input and vidioc_s_input.

Quoted 5 lines
Quoted 1 line

Is it worth ?

I don't think it is worth doing this for g/s_input. I think it is useful to have them here: it makes it very clear that there is just a single input and the overhead in both lines and actual bytes is minimal.

What about making that the default NULL g_input/s_input callbacks handling in video_ioctl2 ? Devices with a single input wouldn't need to define g_input/s_input callbacks at all.

Quoted 3 lines

But for the empty open and release functions you could easily handle that in v4l2-dev.c: if you leave the open and release callbacks to NULL, then v4l2_open and v4l2_release can just return 0. That would be nice.

Regards,

Laurent Pinchart

#3 lore raw
In reply to Hans Verkuil

Hello, Hans

Quoted 40 lines · Hans Verkuil
Quoted 26 lines · On Tuesday 24 March 2009 00:14:07 Alexey Klimov wrote:

Hello, all

...
static int terratec_open(struct file *file)

{
        return 0;
}

static int terratec_release(struct file *file)

{
        return 0;
}

...

...

Such code used in many radio-drivers as i understand.

Is it good to place this empty and almost empty functions in:
(here i see two variants)

1) In header file that be in linux/drivers/media/radio/ directory. Later, we can move some generic/or repeating code in this header.

2) In any v4l header. What header may contain this ?

Quoted 8 lines

?

For what ? Well, as i understand we can decrease amount of lines and provide this simple generic functions. It's like video_device_release_empty function behaviour. Maybe not only radio drivers can use such vidioc_g_input and vidioc_s_input.

Is it worth ?

I don't think it is worth doing this for g/s_input. I think it is useful to have them here: it makes it very clear that there is just a single input and the overhead in both lines and actual bytes is minimal.

Quoted 8 lines

But for the empty open and release functions you could easily handle that in v4l2-dev.c: if you leave the open and release callbacks to NULL, then v4l2_open and v4l2_release can just return 0. That would be nice.

Regards,

Hans

May i ask help with this ?
Hans, should it be looks like:

linux/drivers/media/radio/radio-terratec.c +0 -12
diff -r 56cf0f1772f7 linux/drivers/media/radio/radio-terratec.c
@@ -333,20 +333,8 @@
 	return a->index ? -EINVAL : 0;
 }
 
-static int terratec_open(struct file *file)
-{
-	return 0;
-}
-
-static int terratec_release(struct file *file)
-{
-	return 0;
-}
-
 static const struct v4l2_file_operations terratec_fops = {
 	.owner		= THIS_MODULE,
-	.open           = terratec_open,
-	.release        = terratec_release,
 	.ioctl		= video_ioctl2,
 };
 
linux/drivers/media/video/v4l2-dev.c +10 -2
diff -r 56cf0f1772f7 linux/drivers/media/video/v4l2-dev.c
@@ -264,7 +264,10 @@
 	/* and increase the device refcount */
 	video_get(vdev);
 	mutex_unlock(&videodev_lock);
-	ret = vdev->fops->open(filp);
+	if (vdev->fops->open == NULL)
+		ret = 0;
+	else
+		ret = vdev->fops->open(filp);
 	/* decrease the refcount in case of an error */
 	if (ret)
 		video_put(vdev);
@@ -275,7 +278,12 @@
 static int v4l2_release(struct inode *inode, struct file *filp)
 {
 	struct video_device *vdev = video_devdata(filp);
-	int ret = vdev->fops->release(filp);
+	int ret;
+
+	if (vdev->fops->release == NULL)
+		ret = 0;
+	else
+		ret = vdev->fops->release(filp);
 
 	/* decrease the refcount unconditionally since the release()
 	   return value is ignored. */

?

Or in v4l2_open function i can check if vdev->fops->open == NULL before video_get(vdev); (increasing the device refcount), and if it's NULL then unlock_mutex and return 0 ?
And the same in v4l2_release - just return 0 in the begining of function in case vdev->fops->release == NULL ?

What approach is better ?

Signature · 1 line
Best regards, Klimov Alexey
#4 lore raw
Quoted 112 lines · On Friday 27 March 2009 17:44:05 Alexey Klimov wrote:

Hello, Hans

Quoted 41 lines · Hans Verkuil
Quoted 26 lines · On Tuesday 24 March 2009 00:14:07 Alexey Klimov wrote:

Hello, all

...
static int terratec_open(struct file *file)

{
        return 0;
}

static int terratec_release(struct file *file)

{
        return 0;
}

...

...

Such code used in many radio-drivers as i understand.

Is it good to place this empty and almost empty functions in:
(here i see two variants)

1) In header file that be in linux/drivers/media/radio/ directory. Later, we can move some generic/or repeating code in this header.

2) In any v4l header. What header may contain this ?

Quoted 8 lines

?

For what ? Well, as i understand we can decrease amount of lines and provide this simple generic functions. It's like video_device_release_empty function behaviour. Maybe not only radio drivers can use such vidioc_g_input and vidioc_s_input.

Is it worth ?

I don't think it is worth doing this for g/s_input. I think it is useful to have them here: it makes it very clear that there is just a single input and the overhead in both lines and actual bytes is minimal.

Quoted 8 lines

But for the empty open and release functions you could easily handle that in v4l2-dev.c: if you leave the open and release callbacks to NULL, then v4l2_open and v4l2_release can just return 0. That would be nice.

Regards,

Hans

May i ask help with this ?
Hans, should it be looks like:

diff -r 56cf0f1772f7 linux/drivers/media/radio/radio-terratec.c
--- a/linux/drivers/media/radio/radio-terratec.c Mon Mar 23 19:18:34 2009
-0300 +++ b/linux/drivers/media/radio/radio-terratec.c Fri Mar 27
19:32:38 2009 +0300 @@ -333,20 +333,8 @@

linux/drivers/media/radio/radio-terratec.c +0 -12
 	return a->index ? -EINVAL : 0;
 }
 
-static int terratec_open(struct file *file)
-{
-	return 0;
-}
-
-static int terratec_release(struct file *file)
-{
-	return 0;
-}
-
 static const struct v4l2_file_operations terratec_fops = {
 	.owner		= THIS_MODULE,
-	.open           = terratec_open,
-	.release        = terratec_release,
 	.ioctl		= video_ioctl2,
 };

diff -r 56cf0f1772f7 linux/drivers/media/video/v4l2-dev.c

linux/drivers/media/video/v4l2-dev.c +10 -2
@@ -264,7 +264,10 @@
 	/* and increase the device refcount */
 	video_get(vdev);
 	mutex_unlock(&videodev_lock);
-	ret = vdev->fops->open(filp);
+	if (vdev->fops->open == NULL)
+		ret = 0;
+	else
+		ret = vdev->fops->open(filp);
 	/* decrease the refcount in case of an error */
 	if (ret)
 		video_put(vdev);
@@ -275,7 +278,12 @@
 static int v4l2_release(struct inode *inode, struct file *filp)
 {
 	struct video_device *vdev = video_devdata(filp);
-	int ret = vdev->fops->release(filp);
+	int ret;
+
+	if (vdev->fops->release == NULL)
+		ret = 0;
+	else
+		ret = vdev->fops->release(filp);
 
 	/* decrease the refcount unconditionally since the release()
 	   return value is ignored. */

?

Quoted 7 lines

Or in v4l2_open function i can check if vdev->fops->open == NULL before video_get(vdev); (increasing the device refcount), and if it's NULL then unlock_mutex and return 0 ?
And the same in v4l2_release - just return 0 in the begining of function in case vdev->fops->release == NULL ?

What approach is better ?

This is simpler:

diff -r 2e0c6ff1bda3 linux/drivers/media/video/v4l2-dev.c
--- a/linux/drivers/media/video/v4l2-dev.c Mon Mar 23 19:01:18 2009
+0100
+++ b/linux/drivers/media/video/v4l2-dev.c Fri Mar 27 17:47:51 2009
+0100
@@ -250,7 +250,7 @@
static int v4l2_open(struct inode *inode, struct file *filp)

 {
        struct video_device *vdev;

- int ret;
+ int ret = 0;

        /* Check if the video device is available */
        mutex_lock(&videodev_lock);
@@ -264,7 +264,9 @@
        /* and increase the device refcount */
        video_get(vdev);
        mutex_unlock(&videodev_lock);

- ret = vdev->fops->open(filp);
+ if (vdev->fops->open)
+ ret = vdev->fops->open(filp);
+

        /* decrease the refcount in case of an error */
        if (ret)
                video_put(vdev);
@@ -275,7 +277,10 @@
 static int v4l2_release(struct inode *inode, struct file *filp)
 {
        struct video_device *vdev = video_devdata(filp);

- int ret = vdev->fops->release(filp);
+ int ret = 0;
+
+ if (vdev->fops->release)
+ ret = vdev->fops->release(filp);

/* decrease the refcount unconditionally since the release()
return value is ignored. */

Regards,

Hans

Signature · 1 line
Hans Verkuil - video4linux developer - sponsored by TANDBERG
#5 lore raw
Quoted 150 lines · Hans Verkuil
Quoted 112 lines · On Friday 27 March 2009 17:44:05 Alexey Klimov wrote:

Hello, Hans

Quoted 41 lines · Hans Verkuil
Quoted 26 lines · On Tuesday 24 March 2009 00:14:07 Alexey Klimov wrote:

Hello, all

...
static int terratec_open(struct file *file)

{
        return 0;
}

static int terratec_release(struct file *file)

{
        return 0;
}

...

...

Such code used in many radio-drivers as i understand.

Is it good to place this empty and almost empty functions in:
(here i see two variants)

1) In header file that be in linux/drivers/media/radio/ directory. Later, we can move some generic/or repeating code in this header.

2) In any v4l header. What header may contain this ?

Quoted 8 lines

?

For what ? Well, as i understand we can decrease amount of lines and provide this simple generic functions. It's like video_device_release_empty function behaviour. Maybe not only radio drivers can use such vidioc_g_input and vidioc_s_input.

Is it worth ?

I don't think it is worth doing this for g/s_input. I think it is useful to have them here: it makes it very clear that there is just a single input and the overhead in both lines and actual bytes is minimal.

Quoted 8 lines

But for the empty open and release functions you could easily handle that in v4l2-dev.c: if you leave the open and release callbacks to NULL, then v4l2_open and v4l2_release can just return 0. That would be nice.

Regards,

Hans

May i ask help with this ?
Hans, should it be looks like:

diff -r 56cf0f1772f7 linux/drivers/media/radio/radio-terratec.c
--- a/linux/drivers/media/radio/radio-terratec.c Mon Mar 23 19:18:34 2009
-0300 +++ b/linux/drivers/media/radio/radio-terratec.c Fri Mar 27
19:32:38 2009 +0300 @@ -333,20 +333,8 @@

      return a->index ? -EINVAL : 0;

}

linux/drivers/media/radio/radio-terratec.c +0 -12
-static int terratec_open(struct file *file)
-{
-     return 0;
-}
-
-static int terratec_release(struct file *file)
-{
-     return 0;
-}
-
 static const struct v4l2_file_operations terratec_fops = {
      .owner          = THIS_MODULE,
-     .open           = terratec_open,
-     .release        = terratec_release,
      .ioctl          = video_ioctl2,
 };

diff -r 56cf0f1772f7 linux/drivers/media/video/v4l2-dev.c

linux/drivers/media/video/v4l2-dev.c Fri Mar 27 19:32:38 2009 +0300 renamed from linux/drivers/media/video/v4l2-dev.c Mon Mar 23 19:18:34 2009 -0300 +10 -2
@@ -264,7 +264,10 @@
      /* and increase the device refcount */
      video_get(vdev);
      mutex_unlock(&videodev_lock);
-     ret = vdev->fops->open(filp);
+     if (vdev->fops->open == NULL)
+             ret = 0;
+     else
+             ret = vdev->fops->open(filp);
      /* decrease the refcount in case of an error */
      if (ret)
              video_put(vdev);
@@ -275,7 +278,12 @@
 static int v4l2_release(struct inode *inode, struct file *filp)
 {
      struct video_device *vdev = video_devdata(filp);
-     int ret = vdev->fops->release(filp);
+     int ret;
+
+     if (vdev->fops->release == NULL)
+             ret = 0;
+     else
+             ret = vdev->fops->release(filp);
 
      /* decrease the refcount unconditionally since the release()
         return value is ignored. */

?

Quoted 7 lines

Or in v4l2_open function i can check if vdev->fops->open == NULL before video_get(vdev); (increasing the device refcount), and if it's NULL then unlock_mutex and return 0 ?
And the same in v4l2_release - just return 0 in the begining of function in case vdev->fops->release == NULL ?

What approach is better ?

This is simpler:

diff -r 2e0c6ff1bda3 linux/drivers/media/video/v4l2-dev.c
--- a/linux/drivers/media/video/v4l2-dev.c Mon Mar 23 19:01:18 2009
+0100
+++ b/linux/drivers/media/video/v4l2-dev.c Fri Mar 27 17:47:51 2009
+0100

linux/drivers/media/video/v4l2-dev.c +4 -2
@@ -250,7 +250,7 @@
 static int v4l2_open(struct inode *inode, struct file *filp)
 {
       struct video_device *vdev;
-       int ret;
+       int ret = 0;
 
       /* Check if the video device is available */
       mutex_lock(&videodev_lock);
@@ -264,7 +264,9 @@
       /* and increase the device refcount */
       video_get(vdev);
       mutex_unlock(&videodev_lock);
-       ret = vdev->fops->open(filp);
+       if (vdev->fops->open)
+               ret = vdev->fops->open(filp);
+
       /* decrease the refcount in case of an error */
       if (ret)
               video_put(vdev);
@@ -275,7 +277,10 @@
 static int v4l2_release(struct inode *inode, struct file *filp)

linux/drivers/media/video/v4l2-dev.c line 286

Quoted 10 lines
linux/drivers/media/video/v4l2-dev.c +4 -1
 {
       struct video_device *vdev = video_devdata(filp);
-       int ret = vdev->fops->release(filp);
+       int ret = 0;
+
+       if (vdev->fops->release)
+               ret = vdev->fops->release(filp);
 
       /* decrease the refcount unconditionally since the release()
          return value is ignored. */

Looks like you already did right patch ;-)
I don't know what to do, should i repost this like patch ?

Signature · 1 line
Best regards, Klimov Alexey
#6 lore raw
Quoted 159 lines · On Friday 27 March 2009 18:34:01 Alexey Klimov wrote:
Quoted 158 lines · Hans Verkuil
Quoted 112 lines · On Friday 27 March 2009 17:44:05 Alexey Klimov wrote:

Hello, Hans

Quoted 41 lines · Hans Verkuil
Quoted 26 lines · On Tuesday 24 March 2009 00:14:07 Alexey Klimov wrote:

Hello, all

...
static int terratec_open(struct file *file)

{
        return 0;
}

static int terratec_release(struct file *file)

{
        return 0;
}

...

...

Such code used in many radio-drivers as i understand.

Is it good to place this empty and almost empty functions in:
(here i see two variants)

1) In header file that be in linux/drivers/media/radio/ directory. Later, we can move some generic/or repeating code in this header.

2) In any v4l header. What header may contain this ?

Quoted 8 lines

?

For what ? Well, as i understand we can decrease amount of lines and provide this simple generic functions. It's like video_device_release_empty function behaviour. Maybe not only radio drivers can use such vidioc_g_input and vidioc_s_input.

Is it worth ?

I don't think it is worth doing this for g/s_input. I think it is useful to have them here: it makes it very clear that there is just a single input and the overhead in both lines and actual bytes is minimal.

Quoted 8 lines

But for the empty open and release functions you could easily handle that in v4l2-dev.c: if you leave the open and release callbacks to NULL, then v4l2_open and v4l2_release can just return 0. That would be nice.

Regards,

Hans

May i ask help with this ?
Hans, should it be looks like:

diff -r 56cf0f1772f7 linux/drivers/media/radio/radio-terratec.c
--- a/linux/drivers/media/radio/radio-terratec.c Mon Mar 23
19:18:34 2009 -0300 +++ b/linux/drivers/media/radio/radio-terratec.c

     Fri Mar 27 19:32:38 2009 +0300 @@ -333,20 +333,8 @@
      return a->index ? -EINVAL : 0;

}

linux/drivers/media/radio/radio-terratec.c +0 -12
-static int terratec_open(struct file *file)
-{
-     return 0;
-}
-
-static int terratec_release(struct file *file)
-{
-     return 0;
-}
-
 static const struct v4l2_file_operations terratec_fops = {
      .owner          = THIS_MODULE,
-     .open           = terratec_open,
-     .release        = terratec_release,
      .ioctl          = video_ioctl2,
 };

diff -r 56cf0f1772f7 linux/drivers/media/video/v4l2-dev.c
--- a/linux/drivers/media/video/v4l2-dev.c Mon Mar 23 19:18:34 2009
-0300 +++ b/linux/drivers/media/video/v4l2-dev.c Fri Mar 27
19:32:38 2009 +0300 @@ -264,7 +264,10 @@

      /* and increase the device refcount */
      video_get(vdev);
      mutex_unlock(&videodev_lock);
linux/drivers/media/video/v4l2-dev.c +10 -2
-     ret = vdev->fops->open(filp);
+     if (vdev->fops->open == NULL)
+             ret = 0;
+     else
+             ret = vdev->fops->open(filp);
      /* decrease the refcount in case of an error */
      if (ret)
              video_put(vdev);
@@ -275,7 +278,12 @@
 static int v4l2_release(struct inode *inode, struct file *filp)
 {
      struct video_device *vdev = video_devdata(filp);
-     int ret = vdev->fops->release(filp);
+     int ret;
+
+     if (vdev->fops->release == NULL)
+             ret = 0;
+     else
+             ret = vdev->fops->release(filp);
 
      /* decrease the refcount unconditionally since the release()
         return value is ignored. */

?

Quoted 7 lines

Or in v4l2_open function i can check if vdev->fops->open == NULL before video_get(vdev); (increasing the device refcount), and if it's NULL then unlock_mutex and return 0 ?
And the same in v4l2_release - just return 0 in the begining of function in case vdev->fops->release == NULL ?

What approach is better ?

This is simpler:

diff -r 2e0c6ff1bda3 linux/drivers/media/video/v4l2-dev.c
--- a/linux/drivers/media/video/v4l2-dev.c Mon Mar 23 19:01:18
2009 +0100
+++ b/linux/drivers/media/video/v4l2-dev.c Fri Mar 27 17:47:51
2009 +0100

linux/drivers/media/video/v4l2-dev.c +8 -3
@@ -250,7 +250,7 @@
 static int v4l2_open(struct inode *inode, struct file *filp)
 {
       struct video_device *vdev;
-       int ret;
+       int ret = 0;
 
       /* Check if the video device is available */
       mutex_lock(&videodev_lock);
@@ -264,7 +264,9 @@
       /* and increase the device refcount */
       video_get(vdev);
       mutex_unlock(&videodev_lock);
-       ret = vdev->fops->open(filp);
+       if (vdev->fops->open)
+               ret = vdev->fops->open(filp);
+
       /* decrease the refcount in case of an error */
       if (ret)
               video_put(vdev);
@@ -275,7 +277,10 @@
 static int v4l2_release(struct inode *inode, struct file *filp)
 {
       struct video_device *vdev = video_devdata(filp);
-       int ret = vdev->fops->release(filp);
+       int ret = 0;
+
+       if (vdev->fops->release)
+               ret = vdev->fops->release(filp);
 
Quoted 5 lines

linux/drivers/media/video/v4l2-dev.c line 289

Quoted 2 lines
linux/drivers/media/video/v4l2-dev.c +0 -0
       /* decrease the refcount unconditionally since the release()
          return value is ignored. */

Looks like you already did right patch ;-)
I don't know what to do, should i repost this like patch ?

Just turn it into a patch series with this as the first patch and fixing the radio drivers that can use it.

Here's my SoB for this one:

Regards,

Hans

Signature · 1 line
Hans Verkuil - video4linux developer - sponsored by TANDBERG