1 Overview of the V4L2 driver framework
2 =====================================
4 This text documents the various structures provided by the V4L2 framework and
11 The V4L2 drivers tend to be very complex due to the complexity of the
12 hardware: most devices have multiple ICs, export multiple device nodes in
13 /dev, and create also non-V4L2 devices such as DVB, ALSA, FB, I2C and input
16 Especially the fact that V4L2 drivers have to setup supporting ICs to
17 do audio/video muxing/encoding/decoding makes it more complex than most.
18 Usually these ICs are connected to the main bridge driver through one or
19 more I2C busses, but other busses can also be used. Such devices are
22 For a long time the framework was limited to the video_device struct for
23 creating V4L device nodes and video_buf for handling the video buffers
24 (note that this document does not discuss the video_buf framework).
26 This meant that all drivers had to do the setup of device instances and
27 connecting to sub-devices themselves. Some of this is quite complicated
28 to do right and many drivers never did do it correctly.
30 There is also a lot of common code that could never be refactored due to
31 the lack of a framework.
33 So this framework sets up the basic building blocks that all drivers
34 need and this same framework should make it much easier to refactor
35 common code into utility functions shared by all drivers.
41 All drivers have the following structure:
43 1) A struct for each device instance containing the device state.
45 2) A way of initializing and commanding sub-devices (if any).
47 3) Creating V4L2 device nodes (/dev/videoX, /dev/vbiX, /dev/radioX and
48 /dev/vtxX) and keeping track of device-node specific data.
50 4) Filehandle-specific structs containing per-filehandle data;
52 5) video buffer handling.
54 This is a rough schematic of how it all relates:
58 +-sub-device instances
62 \-filehandle instances
65 Structure of the framework
66 --------------------------
68 The framework closely resembles the driver structure: it has a v4l2_device
69 struct for the device instance data, a v4l2_subdev struct to refer to
70 sub-device instances, the video_device struct stores V4L2 device node data
71 and in the future a v4l2_fh struct will keep track of filehandle instances
72 (this is not yet implemented).
78 Each device instance is represented by a struct v4l2_device (v4l2-device.h).
79 Very simple devices can just allocate this struct, but most of the time you
80 would embed this struct inside a larger struct.
82 You must register the device instance:
84 v4l2_device_register(struct device *dev, struct v4l2_device *v4l2_dev);
86 Registration will initialize the v4l2_device struct and link dev->driver_data
87 to v4l2_dev. If v4l2_dev->name is empty then it will be set to a value derived
88 from dev (driver name followed by the bus_id, to be precise). If you set it
89 up before calling v4l2_device_register then it will be untouched. If dev is
90 NULL, then you *must* setup v4l2_dev->name before calling v4l2_device_register.
92 The first 'dev' argument is normally the struct device pointer of a pci_dev,
93 usb_device or platform_device. It is rare for dev to be NULL, but it happens
94 with ISA devices or when one device creates multiple PCI devices, thus making
95 it impossible to associate v4l2_dev with a particular parent.
99 v4l2_device_unregister(struct v4l2_device *v4l2_dev);
101 Unregistering will also automatically unregister all subdevs from the device.
103 Sometimes you need to iterate over all devices registered by a specific
104 driver. This is usually the case if multiple device drivers use the same
105 hardware. E.g. the ivtvfb driver is a framebuffer driver that uses the ivtv
106 hardware. The same is true for alsa drivers for example.
108 You can iterate over all registered devices as follows:
110 static int callback(struct device *dev, void *p)
112 struct v4l2_device *v4l2_dev = dev_get_drvdata(dev);
114 /* test if this device was inited */
115 if (v4l2_dev == NULL)
123 struct device_driver *drv;
126 /* Find driver 'ivtv' on the PCI bus.
127 pci_bus_type is a global. For USB busses use usb_bus_type. */
128 drv = driver_find("ivtv", &pci_bus_type);
129 /* iterate over all ivtv device instances */
130 err = driver_for_each_device(drv, NULL, p, callback);
135 Sometimes you need to keep a running counter of the device instance. This is
136 commonly used to map a device instance to an index of a module option array.
138 The recommended approach is as follows:
140 static atomic_t drv_instance = ATOMIC_INIT(0);
142 static int __devinit drv_probe(struct pci_dev *pdev,
143 const struct pci_device_id *pci_id)
146 state->instance = atomic_inc_return(&drv_instance) - 1;
153 Many drivers need to communicate with sub-devices. These devices can do all
154 sort of tasks, but most commonly they handle audio and/or video muxing,
155 encoding or decoding. For webcams common sub-devices are sensors and camera
158 Usually these are I2C devices, but not necessarily. In order to provide the
159 driver with a consistent interface to these sub-devices the v4l2_subdev struct
160 (v4l2-subdev.h) was created.
162 Each sub-device driver must have a v4l2_subdev struct. This struct can be
163 stand-alone for simple sub-devices or it might be embedded in a larger struct
164 if more state information needs to be stored. Usually there is a low-level
165 device struct (e.g. i2c_client) that contains the device data as setup
166 by the kernel. It is recommended to store that pointer in the private
167 data of v4l2_subdev using v4l2_set_subdevdata(). That makes it easy to go
168 from a v4l2_subdev to the actual low-level bus-specific device data.
170 You also need a way to go from the low-level struct to v4l2_subdev. For the
171 common i2c_client struct the i2c_set_clientdata() call is used to store a
172 v4l2_subdev pointer, for other busses you may have to use other methods.
174 From the bridge driver perspective you load the sub-device module and somehow
175 obtain the v4l2_subdev pointer. For i2c devices this is easy: you call
176 i2c_get_clientdata(). For other busses something similar needs to be done.
177 Helper functions exists for sub-devices on an I2C bus that do most of this
180 Each v4l2_subdev contains function pointers that sub-device drivers can
181 implement (or leave NULL if it is not applicable). Since sub-devices can do
182 so many different things and you do not want to end up with a huge ops struct
183 of which only a handful of ops are commonly implemented, the function pointers
184 are sorted according to category and each category has its own ops struct.
186 The top-level ops struct contains pointers to the category ops structs, which
187 may be NULL if the subdev driver does not support anything from that category.
191 struct v4l2_subdev_core_ops {
192 int (*g_chip_ident)(struct v4l2_subdev *sd, struct v4l2_dbg_chip_ident *chip);
193 int (*log_status)(struct v4l2_subdev *sd);
194 int (*init)(struct v4l2_subdev *sd, u32 val);
198 struct v4l2_subdev_tuner_ops {
202 struct v4l2_subdev_audio_ops {
206 struct v4l2_subdev_video_ops {
210 struct v4l2_subdev_ops {
211 const struct v4l2_subdev_core_ops *core;
212 const struct v4l2_subdev_tuner_ops *tuner;
213 const struct v4l2_subdev_audio_ops *audio;
214 const struct v4l2_subdev_video_ops *video;
217 The core ops are common to all subdevs, the other categories are implemented
218 depending on the sub-device. E.g. a video device is unlikely to support the
219 audio ops and vice versa.
221 This setup limits the number of function pointers while still making it easy
222 to add new ops and categories.
224 A sub-device driver initializes the v4l2_subdev struct using:
226 v4l2_subdev_init(sd, &ops);
228 Afterwards you need to initialize subdev->name with a unique name and set the
229 module owner. This is done for you if you use the i2c helper functions.
231 A device (bridge) driver needs to register the v4l2_subdev with the
234 int err = v4l2_device_register_subdev(v4l2_dev, sd);
236 This can fail if the subdev module disappeared before it could be registered.
237 After this function was called successfully the subdev->dev field points to
240 You can unregister a sub-device using:
242 v4l2_device_unregister_subdev(sd);
244 Afterwards the subdev module can be unloaded and sd->dev == NULL.
246 You can call an ops function either directly:
248 err = sd->ops->core->g_chip_ident(sd, &chip);
250 but it is better and easier to use this macro:
252 err = v4l2_subdev_call(sd, core, g_chip_ident, &chip);
254 The macro will to the right NULL pointer checks and returns -ENODEV if subdev
255 is NULL, -ENOIOCTLCMD if either subdev->core or subdev->core->g_chip_ident is
256 NULL, or the actual result of the subdev->ops->core->g_chip_ident ops.
258 It is also possible to call all or a subset of the sub-devices:
260 v4l2_device_call_all(v4l2_dev, 0, core, g_chip_ident, &chip);
262 Any subdev that does not support this ops is skipped and error results are
263 ignored. If you want to check for errors use this:
265 err = v4l2_device_call_until_err(v4l2_dev, 0, core, g_chip_ident, &chip);
267 Any error except -ENOIOCTLCMD will exit the loop with that error. If no
268 errors (except -ENOIOCTLCMD) occured, then 0 is returned.
270 The second argument to both calls is a group ID. If 0, then all subdevs are
271 called. If non-zero, then only those whose group ID match that value will
272 be called. Before a bridge driver registers a subdev it can set sd->grp_id
273 to whatever value it wants (it's 0 by default). This value is owned by the
274 bridge driver and the sub-device driver will never modify or use it.
276 The group ID gives the bridge driver more control how callbacks are called.
277 For example, there may be multiple audio chips on a board, each capable of
278 changing the volume. But usually only one will actually be used when the
279 user want to change the volume. You can set the group ID for that subdev to
280 e.g. AUDIO_CONTROLLER and specify that as the group ID value when calling
281 v4l2_device_call_all(). That ensures that it will only go to the subdev
284 The advantage of using v4l2_subdev is that it is a generic struct and does
285 not contain any knowledge about the underlying hardware. So a driver might
286 contain several subdevs that use an I2C bus, but also a subdev that is
287 controlled through GPIO pins. This distinction is only relevant when setting
288 up the device, but once the subdev is registered it is completely transparent.
291 I2C sub-device drivers
292 ----------------------
294 Since these drivers are so common, special helper functions are available to
295 ease the use of these drivers (v4l2-common.h).
297 The recommended method of adding v4l2_subdev support to an I2C driver is to
298 embed the v4l2_subdev struct into the state struct that is created for each
299 I2C device instance. Very simple devices have no state struct and in that case
300 you can just create a v4l2_subdev directly.
302 A typical state struct would look like this (where 'chipname' is replaced by
303 the name of the chip):
305 struct chipname_state {
306 struct v4l2_subdev sd;
307 ... /* additional state fields */
310 Initialize the v4l2_subdev struct as follows:
312 v4l2_i2c_subdev_init(&state->sd, client, subdev_ops);
314 This function will fill in all the fields of v4l2_subdev and ensure that the
315 v4l2_subdev and i2c_client both point to one another.
317 You should also add a helper inline function to go from a v4l2_subdev pointer
318 to a chipname_state struct:
320 static inline struct chipname_state *to_state(struct v4l2_subdev *sd)
322 return container_of(sd, struct chipname_state, sd);
325 Use this to go from the v4l2_subdev struct to the i2c_client struct:
327 struct i2c_client *client = v4l2_get_subdevdata(sd);
329 And this to go from an i2c_client to a v4l2_subdev struct:
331 struct v4l2_subdev *sd = i2c_get_clientdata(client);
333 Finally you need to make a command function to make driver->command()
334 call the right subdev_ops functions:
336 static int subdev_command(struct i2c_client *client, unsigned cmd, void *arg)
338 return v4l2_subdev_command(i2c_get_clientdata(client), cmd, arg);
341 If driver->command is never used then you can leave this out. Eventually the
342 driver->command usage should be removed from v4l.
344 Make sure to call v4l2_device_unregister_subdev(sd) when the remove() callback
345 is called. This will unregister the sub-device from the bridge driver. It is
346 safe to call this even if the sub-device was never registered.
348 You need to do this because when the bridge driver destroys the i2c adapter
349 the remove() callbacks are called of the i2c devices on that adapter.
350 After that the corresponding v4l2_subdev structures are invalid, so they
351 have to be unregistered first. Calling v4l2_device_unregister_subdev(sd)
352 from the remove() callback ensures that this is always done correctly.
355 The bridge driver also has some helper functions it can use:
357 struct v4l2_subdev *sd = v4l2_i2c_new_subdev(adapter, "module_foo", "chipid", 0x36);
359 This loads the given module (can be NULL if no module needs to be loaded) and
360 calls i2c_new_device() with the given i2c_adapter and chip/address arguments.
361 If all goes well, then it registers the subdev with the v4l2_device. It gets
362 the v4l2_device by calling i2c_get_adapdata(adapter), so you should make sure
363 to call i2c_set_adapdata(adapter, v4l2_device) when you setup the i2c_adapter
366 You can also use v4l2_i2c_new_probed_subdev() which is very similar to
367 v4l2_i2c_new_subdev(), except that it has an array of possible I2C addresses
368 that it should probe. Internally it calls i2c_new_probed_device().
370 Both functions return NULL if something went wrong.
372 Note that the chipid you pass to v4l2_i2c_new_(probed_)subdev() is usually
373 the same as the module name. It allows you to specify a chip variant, e.g.
374 "saa7114" or "saa7115". In general though the i2c driver autodetects this.
375 The use of chipid is something that needs to be looked at more closely at a
376 later date. It differs between i2c drivers and as such can be confusing.
377 To see which chip variants are supported you can look in the i2c driver code
378 for the i2c_device_id table. This lists all the possibilities.
384 The actual device nodes in the /dev directory are created using the
385 video_device struct (v4l2-dev.h). This struct can either be allocated
386 dynamically or embedded in a larger struct.
388 To allocate it dynamically use:
390 struct video_device *vdev = video_device_alloc();
395 vdev->release = video_device_release;
397 If you embed it in a larger struct, then you must set the release()
398 callback to your own function:
400 struct video_device *vdev = &my_vdev->vdev;
402 vdev->release = my_vdev_release;
404 The release callback must be set and it is called when the last user
405 of the video device exits.
407 The default video_device_release() callback just calls kfree to free the
410 You should also set these fields:
412 - v4l2_dev: set to the v4l2_device parent device.
413 - name: set to something descriptive and unique.
414 - fops: set to the v4l2_file_operations struct.
415 - ioctl_ops: if you use the v4l2_ioctl_ops to simplify ioctl maintenance
416 (highly recommended to use this and it might become compulsory in the
417 future!), then set this to your v4l2_ioctl_ops struct.
418 - parent: you only set this if v4l2_device was registered with NULL as
419 the parent device struct. This only happens in cases where one hardware
420 device has multiple PCI devices that all share the same v4l2_device core.
422 The cx88 driver is an example of this: one core v4l2_device struct, but
423 it is used by both an raw video PCI device (cx8800) and a MPEG PCI device
424 (cx8802). Since the v4l2_device cannot be associated with a particular
425 PCI device it is setup without a parent device. But when the struct
426 video_device is setup you do know which parent PCI device to use.
428 If you use v4l2_ioctl_ops, then you should set either .unlocked_ioctl or
429 .ioctl to video_ioctl2 in your v4l2_file_operations struct.
431 The v4l2_file_operations struct is a subset of file_operations. The main
432 difference is that the inode argument is omitted since it is never used.
435 video_device registration
436 -------------------------
438 Next you register the video device: this will create the character device
441 err = video_register_device(vdev, VFL_TYPE_GRABBER, -1);
443 video_device_release(vdev); /* or kfree(my_vdev); */
447 Which device is registered depends on the type argument. The following
450 VFL_TYPE_GRABBER: videoX for video input/output devices
451 VFL_TYPE_VBI: vbiX for vertical blank data (i.e. closed captions, teletext)
452 VFL_TYPE_RADIO: radioX for radio tuners
453 VFL_TYPE_VTX: vtxX for teletext devices (deprecated, don't use)
455 The last argument gives you a certain amount of control over the device
456 kernel number used (i.e. the X in videoX). Normally you will pass -1 to
457 let the v4l2 framework pick the first free number. But if a driver creates
458 many devices, then it can be useful to have different video devices in
459 separate ranges. For example, video capture devices start at 0, video
460 output devices start at 16.
462 So you can use the last argument to specify a minimum kernel number and
463 the v4l2 framework will try to pick the first free number that is equal
464 or higher to what you passed. If that fails, then it will just pick the
467 Whenever a device node is created some attributes are also created for you.
468 If you look in /sys/class/video4linux you see the devices. Go into e.g.
469 video0 and you will see 'name' and 'index' attributes. The 'name' attribute
470 is the 'name' field of the video_device struct. The 'index' attribute is
471 a device node index that can be assigned by the driver, or that is calculated
474 If you call video_register_device(), then the index is just increased by
475 1 for each device node you register. The first video device node you register
476 always starts off with 0.
478 Alternatively you can call video_register_device_index() which is identical
479 to video_register_device(), but with an extra index argument. Here you can
480 pass a specific index value (between 0 and 31) that should be used.
482 Users can setup udev rules that utilize the index attribute to make fancy
483 device names (e.g. 'mpegX' for MPEG video capture device nodes).
485 After the device was successfully registered, then you can use these fields:
487 - vfl_type: the device type passed to video_register_device.
488 - minor: the assigned device minor number.
489 - num: the device kernel number (i.e. the X in videoX).
490 - index: the device index number (calculated or set explicitly using
491 video_register_device_index).
493 If the registration failed, then you need to call video_device_release()
494 to free the allocated video_device struct, or free your own struct if the
495 video_device was embedded in it. The vdev->release() callback will never
496 be called if the registration failed, nor should you ever attempt to
497 unregister the device if the registration failed.
503 When the video device nodes have to be removed, either during the unload
504 of the driver or because the USB device was disconnected, then you should
507 video_unregister_device(vdev);
509 This will remove the device nodes from sysfs (causing udev to remove them
512 After video_unregister_device() returns no new opens can be done.
514 However, in the case of USB devices some application might still have one
515 of these device nodes open. You should block all new accesses to read,
516 write, poll, etc. except possibly for certain ioctl operations like
519 When the last user of the video device node exits, then the vdev->release()
520 callback is called and you can do the final cleanup there.
523 video_device helper functions
524 -----------------------------
526 There are a few useful helper functions:
528 You can set/get driver private data in the video_device struct using:
530 void *video_get_drvdata(struct video_device *vdev);
531 void video_set_drvdata(struct video_device *vdev, void *data);
533 Note that you can safely call video_set_drvdata() before calling
534 video_register_device().
538 struct video_device *video_devdata(struct file *file);
540 returns the video_device belonging to the file struct.
542 The final helper function combines video_get_drvdata with
545 void *video_drvdata(struct file *file);
547 You can go from a video_device struct to the v4l2_device struct using:
549 struct v4l2_device *v4l2_dev = vdev->v4l2_dev;
551 video buffer helper functions
552 -----------------------------
554 The v4l2 core API provides a standard method for dealing with video
555 buffers. Those methods allow a driver to implement read(), mmap() and
556 overlay() on a consistent way.
558 There are currently methods for using video buffers on devices that
559 supports DMA with scatter/gather method (videobuf-dma-sg), DMA with
560 linear access (videobuf-dma-contig), and vmalloced buffers, mostly
561 used on USB drivers (videobuf-vmalloc).
563 Any driver using videobuf should provide operations (callbacks) for
566 ops->buf_setup - calculates the size of the video buffers and avoid they
567 to waste more than some maximum limit of RAM;
568 ops->buf_prepare - fills the video buffer structs and calls
569 videobuf_iolock() to alloc and prepare mmaped memory;
570 ops->buf_queue - advices the driver that another buffer were
571 requested (by read() or by QBUF);
572 ops->buf_release - frees any buffer that were allocated.
574 In order to use it, the driver need to have a code (generally called at
575 interrupt context) that will properly handle the buffer request lists,
576 announcing that a new buffer were filled.
578 The irq handling code should handle the videobuf task lists, in order
579 to advice videobuf that a new frame were filled, in order to honor to a
580 request. The code is generally like this one:
581 if (list_empty(&dma_q->active))
584 buf = list_entry(dma_q->active.next, struct vbuffer, vb.queue);
586 if (!waitqueue_active(&buf->vb.done))
589 /* Some logic to handle the buf may be needed here */
591 list_del(&buf->vb.queue);
592 do_gettimeofday(&buf->vb.ts);
593 wake_up(&buf->vb.done);
595 Those are the videobuffer functions used on drivers, implemented on
598 - Videobuf init functions
599 videobuf_queue_sg_init()
600 Initializes the videobuf infrastructure. This function should be
601 called before any other videobuf function on drivers that uses DMA
602 Scatter/Gather buffers.
604 videobuf_queue_dma_contig_init
605 Initializes the videobuf infrastructure. This function should be
606 called before any other videobuf function on drivers that need DMA
609 videobuf_queue_vmalloc_init()
610 Initializes the videobuf infrastructure. This function should be
611 called before any other videobuf function on USB (and other drivers)
612 that need a vmalloced type of videobuf.
615 Prepares the videobuf memory for the proper method (read, mmap, overlay).
617 - videobuf_queue_is_busy()
618 Checks if a videobuf is streaming.
620 - videobuf_queue_cancel()
621 Stops video handling.
623 - videobuf_mmap_free()
627 Stops video handling, ends mmap and frees mmap and other buffers.
629 - V4L2 api functions. Those functions correspond to VIDIOC_foo ioctls:
630 videobuf_reqbufs(), videobuf_querybuf(), videobuf_qbuf(),
631 videobuf_dqbuf(), videobuf_streamon(), videobuf_streamoff().
633 - V4L1 api function (corresponds to VIDIOCMBUF ioctl):
635 This function is used to provide backward compatibility with V4L1
638 - Some help functions for read()/poll() operations:
639 videobuf_read_stream()
640 For continuous stream read()
643 videobuf_poll_stream()
644 polling help function
646 The better way to understand it is to take a look at vivi driver. One
647 of the main reasons for vivi is to be a videobuf usage example. the
648 vivi_thread_tick() does the task that the IRQ callback would do on PCI
649 drivers (or the irq callback on USB).