v4l: Build error with et61x251, if V4L1_COMPAT is not selected

With current git tree I get an build error
for et61x251, if V4L1_COMPAT is not selected:

drivers/media/video/et61x251/et61x251_core.c: In et61x251_show_:
drivers/media/video/et61x251/et61x251_core.c:718: error: implicit declaration of to_video_device

Fix: add VIDEO_V4L1_COMPAT as a dependency for this driver.

Signed-off-by Andreas Herrmann <[email protected]>

---
 drivers/media/video/et61x251/Kconfig |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

(quoted diff) +0 -0

b/drivers/media/video/et61x251/Kconfig index dcc1a03..9143424 100644

drivers/media/video/et61x251/Kconfig +1 -1

@@ -1,6 +1,6 @@
 config USB_ET61X251
 	tristate "USB ET61X[12]51 PC Camera Controller support"
-	depends on VIDEO_V4L2
+	depends on VIDEO_V4L2 && VIDEO_V4L1_COMPAT
 	---help---
 	  Say Y here if you want support for cameras based on Etoms ET61X151
 	  or ET61X251 PC Camera Controllers.

Quoted 7 lines · On Thursday 13 September 2007 14:36:27 [email protected] wrote:

With current git tree I get an build error
for et61x251, if V4L1_COMPAT is not selected:

drivers/media/video/et61x251/et61x251_core.c: In et61x251_show_:
drivers/media/video/et61x251/et61x251_core.c:718: error: implicit declaration of to_video_device

Quoted 1 line

Fix: add VIDEO_V4L1_COMPAT as a dependency for this driver.

No, wait. The driver really is V4L2 only. If you can, please fix it by replacing to_video_device() with container_of() instead:

cam = video_get_drvdata(container_of(cd, struct video_device, class_dev));

This fixes a kernel build problem and
should make it into 2.6.23, I think.

Regards,

Andreas

--

Get rid of some v4l1 remainders to avoid kernel build errors if V4L1_COMPAT is not selected:

drivers/media/video/et61x251/et61x251_core.c: In et61x251_show_:
drivers/media/video/et61x251/et61x251_core.c:718: error: implicit
declaration of to_video_device

Fix as suggested by Luca Risolia <[email protected]>

Signed-off-by Andreas Herrmann <[email protected]>

---
 drivers/media/video/et61x251/et61x251_core.c |   24

++++++++++++++++-------- 1 files changed, 16 insertions(+), 8 deletions(-)

(quoted diff) +0 -0

b/drivers/media/video/et61x251/et61x251_core.c index 585bd1f..a3ee968 100644

drivers/media/video/et61x251/et61x251_core.c +8 -4

@@ -715,7 +715,8 @@ static ssize_t et61x251_show_reg(struct class_device*
 cd, char* buf) if (mutex_lock_interruptible(&et61x251_sysfs_lock))
 		return -ERESTARTSYS;
 
-	cam = video_get_drvdata(to_video_device(cd));
+	cam = video_get_drvdata(container_of(cd, struct video_device,
+					     class_dev));
 	if (!cam) {
 		mutex_unlock(&et61x251_sysfs_lock);
 		return -ENODEV;

@@ -739,7 +740,8 @@ et61x251_store_reg(struct class_device* cd, const char*
 buf, size_t len) if (mutex_lock_interruptible(&et61x251_sysfs_lock))
 		return -ERESTARTSYS;
 
-	cam = video_get_drvdata(to_video_device(cd));
+	cam = video_get_drvdata(container_of(cd, struct video_device,
+					     class_dev));
 	if (!cam) {
 		mutex_unlock(&et61x251_sysfs_lock);
 		return -ENODEV;

@@ -771,7 +773,8 @@ static ssize_t et61x251_show_val(struct class_device*
 cd, char* buf) if (mutex_lock_interruptible(&et61x251_sysfs_lock))
 		return -ERESTARTSYS;
 
-	cam = video_get_drvdata(to_video_device(cd));
+	cam = video_get_drvdata(container_of(cd, struct video_device,
+					     class_dev));
 	if (!cam) {
 		mutex_unlock(&et61x251_sysfs_lock);
 		return -ENODEV;

@@ -803,7 +806,8 @@ et61x251_store_val(struct class_device* cd, const char*
 buf, size_t len) if (mutex_lock_interruptible(&et61x251_sysfs_lock))
 		return -ERESTARTSYS;
 
-	cam = video_get_drvdata(to_video_device(cd));
+	cam = video_get_drvdata(container_of(cd, struct video_device,
+					     class_dev));
 	if (!cam) {
 		mutex_unlock(&et61x251_sysfs_lock);
 		return -ENODEV;

@@ -839,7 +843,8 @@ static ssize_t et61x251_show_i2c_reg(struct

class_device* cd, char* buf) if
(mutex_lock_interruptible(&et61x251_sysfs_lock))

drivers/media/video/et61x251/et61x251_core.c +4 -2

 		return -ERESTARTSYS;
 
-	cam = video_get_drvdata(to_video_device(cd));
+	cam = video_get_drvdata(container_of(cd, struct video_device,
+					     class_dev));
 	if (!cam) {
 		mutex_unlock(&et61x251_sysfs_lock);
 		return -ENODEV;

@@ -865,7 +870,8 @@ et61x251_store_i2c_reg(struct class_device* cd, const
 char* buf, size_t len) if (mutex_lock_interruptible(&et61x251_sysfs_lock))
 		return -ERESTARTSYS;
 
-	cam = video_get_drvdata(to_video_device(cd));
+	cam = video_get_drvdata(container_of(cd, struct video_device,
+					     class_dev));
 	if (!cam) {
 		mutex_unlock(&et61x251_sysfs_lock);
 		return -ENODEV;

@@ -897,7 +903,8 @@ static ssize_t et61x251_show_i2c_val(struct

class_device* cd, char* buf) if
(mutex_lock_interruptible(&et61x251_sysfs_lock))

drivers/media/video/et61x251/et61x251_core.c +4 -2

 		return -ERESTARTSYS;
 
-	cam = video_get_drvdata(to_video_device(cd));
+	cam = video_get_drvdata(container_of(cd, struct video_device,
+					     class_dev));
 	if (!cam) {
 		mutex_unlock(&et61x251_sysfs_lock);
 		return -ENODEV;

@@ -934,7 +941,8 @@ et61x251_store_i2c_val(struct class_device* cd, const
 char* buf, size_t len) if (mutex_lock_interruptible(&et61x251_sysfs_lock))
 		return -ERESTARTSYS;
 
-	cam = video_get_drvdata(to_video_device(cd));
+	cam = video_get_drvdata(container_of(cd, struct video_device,
+					     class_dev));
 	if (!cam) {
 		mutex_unlock(&et61x251_sysfs_lock);
 		return -ENODEV;

Hacked-by Luca Risolia <[email protected]>

Quoted 18 lines · On Friday 14 September 2007 00:27:17 Andreas Herrmann wrote:

This fixes a kernel build problem and
should make it into 2.6.23, I think.

Regards,

Andreas

--

Get rid of some v4l1 remainders to avoid kernel build errors if V4L1_COMPAT is not selected:

drivers/media/video/et61x251/et61x251_core.c: In et61x251_show_:
drivers/media/video/et61x251/et61x251_core.c:718: error: implicit
declaration of to_video_device

Fix as suggested by Luca Risolia <[email protected]>

Quoted 20 lines · Luca Risolia

Hacked-by Luca Risolia <[email protected]>

Quoted 17 lines · On Friday 14 September 2007 00:27:17 Andreas Herrmann wrote:

This fixes a kernel build problem and
should make it into 2.6.23, I think.

Regards,

Andreas

--

Get rid of some v4l1 remainders to avoid kernel build errors if V4L1_COMPAT is not selected:

drivers/media/video/et61x251/et61x251_core.c: In et61x251_show_:
drivers/media/video/et61x251/et61x251_core.c:718: error: implicit
declaration of to_video_device

Quoted 1 line

Quoted 1 line

Fix as suggested by Luca Risolia <[email protected]>

This patch is really ugly.

Why can't the "to_video_device()" macro be used? Just move it to a place where it's usable! IOW, what's wrong with the *much* simpler patch below?

That "to_video_device()" macro has absolutely _nothing_ to do with CONFIG_VIDEO_V4L1_COMPAT, as far as I can tell!

Linus
---

include/media/v4l2-dev.h +2 -0

index d62847f..17f8f3a 100644
@@ -337,6 +337,9 @@ void *priv;
 	struct class_device class_dev; /* sysfs */
 };
 
+/* Class-dev to video-device */
+#define to_video_device(cd) container_of(cd, struct video_device,

class_dev) +
/* Version 2 functions */
extern int video_register_device(struct video_device *vfd, int type, int
nr); void video_unregister_device(struct video_device *);

include/media/v4l2-dev.h +0 -2

@@ -354,11 +357,9 @@ extern int video_usercopy(struct inode *inode, struct
 file *file, int (*func)(struct inode *inode, struct file *file,
 				      unsigned int cmd, void *arg));
 
-
 #ifdef CONFIG_VIDEO_V4L1_COMPAT
 #include <linux/mm.h>
 
-#define to_video_device(cd) container_of(cd, struct video_device,

class_dev) static inline int __must_check
video_device_create_file(struct video_device *vfd,
struct class_device_attribute *attr)

Quoted 4 lines · On Friday 14 September 2007 02:09:01 Linus Torvalds wrote:

Quoted 3 lines · Luca Risolia

Hacked-by Luca Risolia <[email protected]>

On Friday 14 September 2007 00:27:17 Andreas Herrmann wrote:

Quoted 8 lines

Quoted 4 lines

Quoted 4 lines

This fixes a kernel build problem and
should make it into 2.6.23, I think.

This patch is really ugly.

Quoted 2 lines

Why can't the "to_video_device()" macro be used? Just move it to a place where it's usable! IOW, what's wrong with the *much* simpler patch below?

There's nothing wtong in my opinion. I do not know the exact reason why Mauro moved "to_video_device()" into CONFIG_VIDEO_V4L1_COMPAT. Pheraps he can give more details about this change.

Quoted 30 lines

That "to_video_device()" macro has absolutely _nothing_ to do with CONFIG_VIDEO_V4L1_COMPAT, as far as I can tell!

Linus
---

include/media/v4l2-dev.h +2 -0

index d62847f..17f8f3a 100644
@@ -337,6 +337,9 @@ void *priv;
 	struct class_device class_dev; /* sysfs */
 };
 
+/* Class-dev to video-device */
+#define to_video_device(cd) container_of(cd, struct video_device,

class_dev) +
/* Version 2 functions */
extern int video_register_device(struct video_device *vfd, int type, int
nr); void video_unregister_device(struct video_device *);

include/media/v4l2-dev.h +0 -2

@@ -354,11 +357,9 @@ extern int video_usercopy(struct inode *inode, struct
 file *file, int (*func)(struct inode *inode, struct file *file,
 				      unsigned int cmd, void *arg));
 
-
 #ifdef CONFIG_VIDEO_V4L1_COMPAT
 #include <linux/mm.h>
 
-#define to_video_device(cd) container_of(cd, struct video_device,

Quoted 3 lines

class_dev) static inline int __must_check
video_device_create_file(struct video_device *vfd,
struct class_device_attribute *attr)

Best regards
Luca Risolia

Quoted 1 line

Quoted 1 line

This patch is really ugly.

Well, yes. I should have known as I converted so many occurences of to_video_device to container_of in my second patch.

Quoted 6 lines

Quoted 2 lines

Why can't the "to_video_device()" macro be used? Just move it to a place where it's usable! IOW, what's wrong with the *much* simpler patch below?

There's nothing wtong in my opinion. I do not know the exact reason why Mauro moved "to_video_device()" into CONFIG_VIDEO_V4L1_COMPAT. Pheraps he can give more details about this change.

BTW, just a few V4L2 drivers and videodev seem to use this construct:
video/usbvision/usbvision-video.c: container_of(cd, struct video_device, class_dev);

video/sn9c102/sn9c102_core.c: cam = video_get_drvdata(container_of(cd, struct video_device,
video/sn9c102/sn9c102_core.c- class_dev));

video/videodev.c: struct video_device *vfd = container_of(cd, struct video_device,
video/videodev.c- class_dev);

And then their are drivers with other variants of container_of, e.g.:
video/pvrusb2/pvrusb2-v4l2.c: vp = container_of(chp,struct pvr2_v4l2,channel);

  video/bt8xx/bttv-vbi.c: struct bttv_buffer *buf = container_of(vb,struct bttv_buffer,vb);
   ...

I think, there should be a to_video_device macro for usage in v4l2.
An most probable for the other container_of stuff (when more there is more than one occurence of a particular construct), drivers should provide their own macro, e.g. to_video_buffer() or so.

That's what other subsystems do. It is more self-explanatory than a direct usage of container_of.

So why not apply (my first patch ... oh no, of course that's rubbish ;-) Linus' below patch for 2.6.23 to fix the compilation for that particular driver. And to work on the conversion of container_of() stuff to meaningful macros for the next kernel release?

From the above, you can clearly see that et62x251 and s9c102 provides an

From my POV, a driver that is creating its own userspace API is not

Quoted 36 lines

Quoted 1 line

Quoted 1 line

This patch is really ugly.

Well, yes. I should have known as I converted so many occurences of to_video_device to container_of in my second patch.

Quoted 7 lines

Quoted 3 lines

Why can't the "to_video_device()" macro be used? Just move it to a place where it's usable! IOW, what's wrong with the *much* simpler patch below?

There's nothing wtong in my opinion. I do not know the exact reason why Mauro moved "to_video_device()" into CONFIG_VIDEO_V4L1_COMPAT. Pheraps he can give more details about this change.

BTW, just a few V4L2 drivers and videodev seem to use this construct:
video/usbvision/usbvision-video.c: container_of(cd, struct
video_device, class_dev);

video/sn9c102/sn9c102_core.c: cam =
video_get_drvdata(container_of(cd, struct video_device, video/sn9c102/sn9c102_core.c-
class_dev));

video/videodev.c: struct video_device *vfd = container_of(cd,
struct video_device, video/videodev.c-

           class_dev);

And then their are drivers with other variants of container_of, e.g.:
video/pvrusb2/pvrusb2-v4l2.c: vp = container_of(chp,struct
pvr2_v4l2,channel); video/bt8xx/bttv-vbi.c: struct bttv_buffer *buf = container_of(vb,struct bttv_buffer,vb); ...

I think, there should be a to_video_device macro for usage in v4l2. An most probable for the other container_of stuff (when more there is more than one occurence of a particular construct), drivers should provide their own macro, e.g. to_video_buffer() or so.

Quoted 7 lines

That's what other subsystems do. It is more self-explanatory than a direct usage of container_of.

So why not apply (my first patch ... oh no, of course that's rubbish ;-) Linus' below patch for 2.6.23 to fix the compilation for that particular driver. And to work on the conversion of container_of() stuff to meaningful macros for the next kernel release?

The to_video_device and the container_of (cd, struct video_device, class_dev) (as you noticed at the above drivers) are used to provide procfs interface.

On videodev, there's the v4l2 core stuff, used on all V4L drivers. It allows some control to the V4L devices (basically, see/change the modprobe loading parameters).

The other drivers that uses to_video_device (or the container_of alternative) to create other userspace interfaces, specific to each driver and not documented at V4L2 API:

bttv-driver.c:static CLASS_DEVICE_ATTR(card, S_IRUGO, show_card, NULL); et61x251_core.c:static CLASS_DEVICE_ATTR(reg, S_IRUGO | S_IWUSR, et61x251_core.c:static CLASS_DEVICE_ATTR(val, S_IRUGO | S_IWUSR, et61x251_core.c:static CLASS_DEVICE_ATTR(i2c_reg, S_IRUGO | S_IWUSR, et61x251_core.c:static CLASS_DEVICE_ATTR(i2c_val, S_IRUGO | S_IWUSR, ov511.c:static CLASS_DEVICE_ATTR(custom_id, S_IRUGO, show_custom_id, NULL); ov511.c:static CLASS_DEVICE_ATTR(model, S_IRUGO, show_model, NULL); ov511.c:static CLASS_DEVICE_ATTR(bridge, S_IRUGO, show_bridge, NULL); ov511.c:static CLASS_DEVICE_ATTR(sensor, S_IRUGO, show_sensor, NULL); ov511.c:static CLASS_DEVICE_ATTR(brightness, S_IRUGO, show_brightness, NULL); ov511.c:static CLASS_DEVICE_ATTR(saturation, S_IRUGO, show_saturation, NULL); ov511.c:static CLASS_DEVICE_ATTR(contrast, S_IRUGO, show_contrast, NULL); ov511.c:static CLASS_DEVICE_ATTR(hue, S_IRUGO, show_hue, NULL);
ov511.c:static CLASS_DEVICE_ATTR(exposure, S_IRUGO, show_exposure, NULL); pwc-if.c:static CLASS_DEVICE_ATTR(pan_tilt, S_IRUGO | S_IWUSR, show_pan_tilt, pwc-if.c:static CLASS_DEVICE_ATTR(button, S_IRUGO | S_IWUSR, show_snapshot_button_status, sn9c102_core.c:static CLASS_DEVICE_ATTR(reg, S_IRUGO | S_IWUSR,
sn9c102_core.c:static CLASS_DEVICE_ATTR(val, S_IRUGO | S_IWUSR, sn9c102_core.c:static CLASS_DEVICE_ATTR(i2c_reg, S_IRUGO | S_IWUSR, sn9c102_core.c:static CLASS_DEVICE_ATTR(i2c_val, S_IRUGO | S_IWUSR, sn9c102_core.c:static CLASS_DEVICE_ATTR(green, S_IWUGO, NULL, sn9c102_store_green); sn9c102_core.c:static CLASS_DEVICE_ATTR(blue, S_IWUGO, NULL, sn9c102_store_blue); sn9c102_core.c:static CLASS_DEVICE_ATTR(red, S_IWUGO, NULL, sn9c102_store_red); sn9c102_core.c:static CLASS_DEVICE_ATTR(frame_header, S_IRUGO, stv680.c:static CLASS_DEVICE_ATTR(name, S_IRUGO, show_##name, NULL); usbvision-video.c:static CLASS_DEVICE_ATTR(version, S_IRUGO, show_version, NULL); usbvision-video.c:static CLASS_DEVICE_ATTR(model, S_IRUGO, show_model, NULL); usbvision-video.c:static CLASS_DEVICE_ATTR(hue, S_IRUGO, show_hue, NULL); usbvision-video.c:static CLASS_DEVICE_ATTR(contrast, S_IRUGO, show_contrast, NULL); usbvision-video.c:static CLASS_DEVICE_ATTR(brightness, S_IRUGO, show_brightness, NULL); usbvision-video.c:static CLASS_DEVICE_ATTR(saturation, S_IRUGO, show_saturation, NULL); usbvision-video.c:static CLASS_DEVICE_ATTR(streaming, S_IRUGO, show_streaming, NULL); usbvision-video.c:static CLASS_DEVICE_ATTR(compression, S_IRUGO, show_compression, NULL); usbvision-video.c:static CLASS_DEVICE_ATTR(bridge, S_IRUGO, show_device_bridge, NULL);

From the above, you can clearly see that et62x251 and s9c102 provides an interface to directly change a register at the device. The other drivers allows reading/changing a few controls directly via /proc (this is also possible, using V4L2 interface). This is not recommended, since V4L2 API should be the proper way to control the devices.

From my POV, a driver that is creating its own userspace API is not fully compliant with V4L2 API.

Cheers,
Mauro

through ioctl()? It's not as immediate and safe as controlling the device registers through /sysfs (not /proc). However, the sysfs interface in those drivers appeared before V4L2 had its own ioctls and we agreed to keep and export the interface to the only users selecting CONFIG_VIDEO_ADV_DEBUG (ok, this is actually valid for the sn9c102, I'll submit a patch for the et61x251 in the future).

Quoted 4 lines

From my POV, a driver that is creating its own userspace API is not fully compliant with V4L2 API.

Isn't Video4Linux2 ioctl() supersedeing sysfs in this case?