1.. SPDX-License-Identifier: GPL-2.0 2 3V4L2 sub-devices 4---------------- 5 6Many drivers need to communicate with sub-devices. These devices can do all 7sort of tasks, but most commonly they handle audio and/or video muxing, 8encoding or decoding. For webcams common sub-devices are sensors and camera 9controllers. 10 11Usually these are I2C devices, but not necessarily. In order to provide the 12driver with a consistent interface to these sub-devices the 13:c:type:`v4l2_subdev` struct (v4l2-subdev.h) was created. 14 15Each sub-device driver must have a :c:type:`v4l2_subdev` struct. This struct 16can be stand-alone for simple sub-devices or it might be embedded in a larger 17struct if more state information needs to be stored. Usually there is a 18low-level device struct (e.g. ``i2c_client``) that contains the device data as 19setup by the kernel. It is recommended to store that pointer in the private 20data of :c:type:`v4l2_subdev` using :c:func:`v4l2_set_subdevdata`. That makes 21it easy to go from a :c:type:`v4l2_subdev` to the actual low-level bus-specific 22device data. 23 24You also need a way to go from the low-level struct to :c:type:`v4l2_subdev`. 25For the common i2c_client struct the i2c_set_clientdata() call is used to store 26a :c:type:`v4l2_subdev` pointer, for other buses you may have to use other 27methods. 28 29Bridges might also need to store per-subdev private data, such as a pointer to 30bridge-specific per-subdev private data. The :c:type:`v4l2_subdev` structure 31provides host private data for that purpose that can be accessed with 32:c:func:`v4l2_get_subdev_hostdata` and :c:func:`v4l2_set_subdev_hostdata`. 33 34From the bridge driver perspective, you load the sub-device module and somehow 35obtain the :c:type:`v4l2_subdev` pointer. For i2c devices this is easy: you call 36``i2c_get_clientdata()``. For other buses something similar needs to be done. 37Helper functions exist for sub-devices on an I2C bus that do most of this 38tricky work for you. 39 40Each :c:type:`v4l2_subdev` contains function pointers that sub-device drivers 41can implement (or leave ``NULL`` if it is not applicable). Since sub-devices can 42do so many different things and you do not want to end up with a huge ops struct 43of which only a handful of ops are commonly implemented, the function pointers 44are sorted according to category and each category has its own ops struct. 45 46The top-level ops struct contains pointers to the category ops structs, which 47may be NULL if the subdev driver does not support anything from that category. 48 49It looks like this: 50 51.. code-block:: c 52 53 struct v4l2_subdev_core_ops { 54 int (*log_status)(struct v4l2_subdev *sd); 55 int (*init)(struct v4l2_subdev *sd, u32 val); 56 ... 57 }; 58 59 struct v4l2_subdev_tuner_ops { 60 ... 61 }; 62 63 struct v4l2_subdev_audio_ops { 64 ... 65 }; 66 67 struct v4l2_subdev_video_ops { 68 ... 69 }; 70 71 struct v4l2_subdev_pad_ops { 72 ... 73 }; 74 75 struct v4l2_subdev_ops { 76 const struct v4l2_subdev_core_ops *core; 77 const struct v4l2_subdev_tuner_ops *tuner; 78 const struct v4l2_subdev_audio_ops *audio; 79 const struct v4l2_subdev_video_ops *video; 80 const struct v4l2_subdev_pad_ops *video; 81 }; 82 83The core ops are common to all subdevs, the other categories are implemented 84depending on the sub-device. E.g. a video device is unlikely to support the 85audio ops and vice versa. 86 87This setup limits the number of function pointers while still making it easy 88to add new ops and categories. 89 90A sub-device driver initializes the :c:type:`v4l2_subdev` struct using: 91 92 :c:func:`v4l2_subdev_init <v4l2_subdev_init>` 93 (:c:type:`sd <v4l2_subdev>`, &\ :c:type:`ops <v4l2_subdev_ops>`). 94 95 96Afterwards you need to initialize :c:type:`sd <v4l2_subdev>`->name with a 97unique name and set the module owner. This is done for you if you use the 98i2c helper functions. 99 100If integration with the media framework is needed, you must initialize the 101:c:type:`media_entity` struct embedded in the :c:type:`v4l2_subdev` struct 102(entity field) by calling :c:func:`media_entity_pads_init`, if the entity has 103pads: 104 105.. code-block:: c 106 107 struct media_pad *pads = &my_sd->pads; 108 int err; 109 110 err = media_entity_pads_init(&sd->entity, npads, pads); 111 112The pads array must have been previously initialized. There is no need to 113manually set the struct media_entity function and name fields, but the 114revision field must be initialized if needed. 115 116A reference to the entity will be automatically acquired/released when the 117subdev device node (if any) is opened/closed. 118 119Don't forget to cleanup the media entity before the sub-device is destroyed: 120 121.. code-block:: c 122 123 media_entity_cleanup(&sd->entity); 124 125If a sub-device driver implements sink pads, the subdev driver may set the 126link_validate field in :c:type:`v4l2_subdev_pad_ops`to provide its own link 127validation function. For every link in the pipeline, the link_validate pad 128operation of the sink end of the link is called. In both cases the driver is 129still responsible for validating the correctness of the format configuration 130between sub-devices and video nodes. 131 132If link_validate op is not set, the default function 133:c:func:`v4l2_subdev_link_validate_default` is used instead. This function 134ensures that width, height and the media bus pixel code are equal on both source 135and sink of the link. Subdev drivers are also free to use this function to 136perform the checks mentioned above in addition to their own checks. 137 138Subdev registration 139~~~~~~~~~~~~~~~~~~~ 140 141There are currently two ways to register subdevices with the V4L2 core. The 142first (traditional) possibility is to have subdevices registered by bridge 143drivers. This can be done when the bridge driver has the complete information 144about subdevices connected to it and knows exactly when to register them. This 145is typically the case for internal subdevices, like video data processing units 146within SoCs or complex PCI(e) boards, camera sensors in USB cameras or connected 147to SoCs, which pass information about them to bridge drivers, usually in their 148platform data. 149 150There are however also situations where subdevices have to be registered 151asynchronously to bridge devices. An example of such a configuration is a Device 152Tree based system where information about subdevices is made available to the 153system independently from the bridge devices, e.g. when subdevices are defined 154in DT as I2C device nodes. The API used in this second case is described further 155below. 156 157Using one or the other registration method only affects the probing process, the 158run-time bridge-subdevice interaction is in both cases the same. 159 160In the **synchronous** case a device (bridge) driver needs to register the 161:c:type:`v4l2_subdev` with the v4l2_device: 162 163 :c:func:`v4l2_device_register_subdev <v4l2_device_register_subdev>` 164 (:c:type:`v4l2_dev <v4l2_device>`, :c:type:`sd <v4l2_subdev>`). 165 166This can fail if the subdev module disappeared before it could be registered. 167After this function was called successfully the subdev->dev field points to 168the :c:type:`v4l2_device`. 169 170If the v4l2_device parent device has a non-NULL mdev field, the sub-device 171entity will be automatically registered with the media device. 172 173You can unregister a sub-device using: 174 175 :c:func:`v4l2_device_unregister_subdev <v4l2_device_unregister_subdev>` 176 (:c:type:`sd <v4l2_subdev>`). 177 178 179Afterwards the subdev module can be unloaded and 180:c:type:`sd <v4l2_subdev>`->dev == ``NULL``. 181 182In the **asynchronous** case subdevice probing can be invoked independently of 183the bridge driver availability. The subdevice driver then has to verify whether 184all the requirements for a successful probing are satisfied. This can include a 185check for a master clock availability. If any of the conditions aren't satisfied 186the driver might decide to return ``-EPROBE_DEFER`` to request further reprobing 187attempts. Once all conditions are met the subdevice shall be registered using 188the :c:func:`v4l2_async_register_subdev` function. Unregistration is 189performed using the :c:func:`v4l2_async_unregister_subdev` call. Subdevices 190registered this way are stored in a global list of subdevices, ready to be 191picked up by bridge drivers. 192 193Bridge drivers in turn have to register a notifier object. This is 194performed using the :c:func:`v4l2_async_notifier_register` call. To 195unregister the notifier the driver has to call 196:c:func:`v4l2_async_notifier_unregister`. The former of the two functions 197takes two arguments: a pointer to struct :c:type:`v4l2_device` and a 198pointer to struct :c:type:`v4l2_async_notifier`. 199 200Before registering the notifier, bridge drivers must do two things: 201first, the notifier must be initialized using the 202:c:func:`v4l2_async_notifier_init`. Second, bridge drivers can then 203begin to form a list of subdevice descriptors that the bridge device 204needs for its operation. Subdevice descriptors are added to the notifier 205using the :c:func:`v4l2_async_notifier_add_subdev` call. This function 206takes two arguments: a pointer to struct :c:type:`v4l2_async_notifier`, 207and a pointer to the subdevice descripter, which is of type struct 208:c:type:`v4l2_async_subdev`. 209 210The V4L2 core will then use these descriptors to match asynchronously 211registered subdevices to them. If a match is detected the ``.bound()`` 212notifier callback is called. After all subdevices have been located the 213.complete() callback is called. When a subdevice is removed from the 214system the .unbind() method is called. All three callbacks are optional. 215 216Calling subdev operations 217~~~~~~~~~~~~~~~~~~~~~~~~~ 218 219The advantage of using :c:type:`v4l2_subdev` is that it is a generic struct and 220does not contain any knowledge about the underlying hardware. So a driver might 221contain several subdevs that use an I2C bus, but also a subdev that is 222controlled through GPIO pins. This distinction is only relevant when setting 223up the device, but once the subdev is registered it is completely transparent. 224 225Once te subdev has been registered you can call an ops function either 226directly: 227 228.. code-block:: c 229 230 err = sd->ops->core->g_std(sd, &norm); 231 232but it is better and easier to use this macro: 233 234.. code-block:: c 235 236 err = v4l2_subdev_call(sd, core, g_std, &norm); 237 238The macro will do the right ``NULL`` pointer checks and returns ``-ENODEV`` 239if :c:type:`sd <v4l2_subdev>` is ``NULL``, ``-ENOIOCTLCMD`` if either 240:c:type:`sd <v4l2_subdev>`->core or :c:type:`sd <v4l2_subdev>`->core->g_std is ``NULL``, or the actual result of the 241:c:type:`sd <v4l2_subdev>`->ops->core->g_std ops. 242 243It is also possible to call all or a subset of the sub-devices: 244 245.. code-block:: c 246 247 v4l2_device_call_all(v4l2_dev, 0, core, g_std, &norm); 248 249Any subdev that does not support this ops is skipped and error results are 250ignored. If you want to check for errors use this: 251 252.. code-block:: c 253 254 err = v4l2_device_call_until_err(v4l2_dev, 0, core, g_std, &norm); 255 256Any error except ``-ENOIOCTLCMD`` will exit the loop with that error. If no 257errors (except ``-ENOIOCTLCMD``) occurred, then 0 is returned. 258 259The second argument to both calls is a group ID. If 0, then all subdevs are 260called. If non-zero, then only those whose group ID match that value will 261be called. Before a bridge driver registers a subdev it can set 262:c:type:`sd <v4l2_subdev>`->grp_id to whatever value it wants (it's 0 by 263default). This value is owned by the bridge driver and the sub-device driver 264will never modify or use it. 265 266The group ID gives the bridge driver more control how callbacks are called. 267For example, there may be multiple audio chips on a board, each capable of 268changing the volume. But usually only one will actually be used when the 269user want to change the volume. You can set the group ID for that subdev to 270e.g. AUDIO_CONTROLLER and specify that as the group ID value when calling 271``v4l2_device_call_all()``. That ensures that it will only go to the subdev 272that needs it. 273 274If the sub-device needs to notify its v4l2_device parent of an event, then 275it can call ``v4l2_subdev_notify(sd, notification, arg)``. This macro checks 276whether there is a ``notify()`` callback defined and returns ``-ENODEV`` if not. 277Otherwise the result of the ``notify()`` call is returned. 278 279V4L2 sub-device userspace API 280----------------------------- 281 282Bridge drivers traditionally expose one or multiple video nodes to userspace, 283and control subdevices through the :c:type:`v4l2_subdev_ops` operations in 284response to video node operations. This hides the complexity of the underlying 285hardware from applications. For complex devices, finer-grained control of the 286device than what the video nodes offer may be required. In those cases, bridge 287drivers that implement :ref:`the media controller API <media_controller>` may 288opt for making the subdevice operations directly accessible from userpace. 289 290Device nodes named ``v4l-subdev``\ *X* can be created in ``/dev`` to access 291sub-devices directly. If a sub-device supports direct userspace configuration 292it must set the ``V4L2_SUBDEV_FL_HAS_DEVNODE`` flag before being registered. 293 294After registering sub-devices, the :c:type:`v4l2_device` driver can create 295device nodes for all registered sub-devices marked with 296``V4L2_SUBDEV_FL_HAS_DEVNODE`` by calling 297:c:func:`v4l2_device_register_subdev_nodes`. Those device nodes will be 298automatically removed when sub-devices are unregistered. 299 300The device node handles a subset of the V4L2 API. 301 302``VIDIOC_QUERYCTRL``, 303``VIDIOC_QUERYMENU``, 304``VIDIOC_G_CTRL``, 305``VIDIOC_S_CTRL``, 306``VIDIOC_G_EXT_CTRLS``, 307``VIDIOC_S_EXT_CTRLS`` and 308``VIDIOC_TRY_EXT_CTRLS``: 309 310 The controls ioctls are identical to the ones defined in V4L2. They 311 behave identically, with the only exception that they deal only with 312 controls implemented in the sub-device. Depending on the driver, those 313 controls can be also be accessed through one (or several) V4L2 device 314 nodes. 315 316``VIDIOC_DQEVENT``, 317``VIDIOC_SUBSCRIBE_EVENT`` and 318``VIDIOC_UNSUBSCRIBE_EVENT`` 319 320 The events ioctls are identical to the ones defined in V4L2. They 321 behave identically, with the only exception that they deal only with 322 events generated by the sub-device. Depending on the driver, those 323 events can also be reported by one (or several) V4L2 device nodes. 324 325 Sub-device drivers that want to use events need to set the 326 ``V4L2_SUBDEV_FL_HAS_EVENTS`` :c:type:`v4l2_subdev`.flags before registering 327 the sub-device. After registration events can be queued as usual on the 328 :c:type:`v4l2_subdev`.devnode device node. 329 330 To properly support events, the ``poll()`` file operation is also 331 implemented. 332 333Private ioctls 334 335 All ioctls not in the above list are passed directly to the sub-device 336 driver through the core::ioctl operation. 337 338Read-only sub-device userspace API 339---------------------------------- 340 341Bridge drivers that control their connected subdevices through direct calls to 342the kernel API realized by :c:type:`v4l2_subdev_ops` structure do not usually 343want userspace to be able to change the same parameters through the subdevice 344device node and thus do not usually register any. 345 346It is sometimes useful to report to userspace the current subdevice 347configuration through a read-only API, that does not permit applications to 348change to the device parameters but allows interfacing to the subdevice device 349node to inspect them. 350 351For instance, to implement cameras based on computational photography, userspace 352needs to know the detailed camera sensor configuration (in terms of skipping, 353binning, cropping and scaling) for each supported output resolution. To support 354such use cases, bridge drivers may expose the subdevice operations to userspace 355through a read-only API. 356 357To create a read-only device node for all the subdevices registered with the 358``V4L2_SUBDEV_FL_HAS_DEVNODE`` set, the :c:type:`v4l2_device` driver should call 359:c:func:`v4l2_device_register_ro_subdev_nodes`. 360 361Access to the following ioctls for userspace applications is restricted on 362sub-device device nodes registered with 363:c:func:`v4l2_device_register_ro_subdev_nodes`. 364 365``VIDIOC_SUBDEV_S_FMT``, 366``VIDIOC_SUBDEV_S_CROP``, 367``VIDIOC_SUBDEV_S_SELECTION``: 368 369 These ioctls are only allowed on a read-only subdevice device node 370 for the :ref:`V4L2_SUBDEV_FORMAT_TRY <v4l2-subdev-format-whence>` 371 formats and selection rectangles. 372 373``VIDIOC_SUBDEV_S_FRAME_INTERVAL``, 374``VIDIOC_SUBDEV_S_DV_TIMINGS``, 375``VIDIOC_SUBDEV_S_STD``: 376 377 These ioctls are not allowed on a read-only subdevice node. 378 379In case the ioctl is not allowed, or the format to modify is set to 380``V4L2_SUBDEV_FORMAT_ACTIVE``, the core returns a negative error code and 381the errno variable is set to ``-EPERM``. 382 383I2C sub-device drivers 384---------------------- 385 386Since these drivers are so common, special helper functions are available to 387ease the use of these drivers (``v4l2-common.h``). 388 389The recommended method of adding :c:type:`v4l2_subdev` support to an I2C driver 390is to embed the :c:type:`v4l2_subdev` struct into the state struct that is 391created for each I2C device instance. Very simple devices have no state 392struct and in that case you can just create a :c:type:`v4l2_subdev` directly. 393 394A typical state struct would look like this (where 'chipname' is replaced by 395the name of the chip): 396 397.. code-block:: c 398 399 struct chipname_state { 400 struct v4l2_subdev sd; 401 ... /* additional state fields */ 402 }; 403 404Initialize the :c:type:`v4l2_subdev` struct as follows: 405 406.. code-block:: c 407 408 v4l2_i2c_subdev_init(&state->sd, client, subdev_ops); 409 410This function will fill in all the fields of :c:type:`v4l2_subdev` ensure that 411the :c:type:`v4l2_subdev` and i2c_client both point to one another. 412 413You should also add a helper inline function to go from a :c:type:`v4l2_subdev` 414pointer to a chipname_state struct: 415 416.. code-block:: c 417 418 static inline struct chipname_state *to_state(struct v4l2_subdev *sd) 419 { 420 return container_of(sd, struct chipname_state, sd); 421 } 422 423Use this to go from the :c:type:`v4l2_subdev` struct to the ``i2c_client`` 424struct: 425 426.. code-block:: c 427 428 struct i2c_client *client = v4l2_get_subdevdata(sd); 429 430And this to go from an ``i2c_client`` to a :c:type:`v4l2_subdev` struct: 431 432.. code-block:: c 433 434 struct v4l2_subdev *sd = i2c_get_clientdata(client); 435 436Make sure to call 437:c:func:`v4l2_device_unregister_subdev`\ (:c:type:`sd <v4l2_subdev>`) 438when the ``remove()`` callback is called. This will unregister the sub-device 439from the bridge driver. It is safe to call this even if the sub-device was 440never registered. 441 442You need to do this because when the bridge driver destroys the i2c adapter 443the ``remove()`` callbacks are called of the i2c devices on that adapter. 444After that the corresponding v4l2_subdev structures are invalid, so they 445have to be unregistered first. Calling 446:c:func:`v4l2_device_unregister_subdev`\ (:c:type:`sd <v4l2_subdev>`) 447from the ``remove()`` callback ensures that this is always done correctly. 448 449 450The bridge driver also has some helper functions it can use: 451 452.. code-block:: c 453 454 struct v4l2_subdev *sd = v4l2_i2c_new_subdev(v4l2_dev, adapter, 455 "module_foo", "chipid", 0x36, NULL); 456 457This loads the given module (can be ``NULL`` if no module needs to be loaded) 458and calls :c:func:`i2c_new_client_device` with the given ``i2c_adapter`` and 459chip/address arguments. If all goes well, then it registers the subdev with 460the v4l2_device. 461 462You can also use the last argument of :c:func:`v4l2_i2c_new_subdev` to pass 463an array of possible I2C addresses that it should probe. These probe addresses 464are only used if the previous argument is 0. A non-zero argument means that you 465know the exact i2c address so in that case no probing will take place. 466 467Both functions return ``NULL`` if something went wrong. 468 469Note that the chipid you pass to :c:func:`v4l2_i2c_new_subdev` is usually 470the same as the module name. It allows you to specify a chip variant, e.g. 471"saa7114" or "saa7115". In general though the i2c driver autodetects this. 472The use of chipid is something that needs to be looked at more closely at a 473later date. It differs between i2c drivers and as such can be confusing. 474To see which chip variants are supported you can look in the i2c driver code 475for the i2c_device_id table. This lists all the possibilities. 476 477There are one more helper function: 478 479:c:func:`v4l2_i2c_new_subdev_board` uses an :c:type:`i2c_board_info` struct 480which is passed to the i2c driver and replaces the irq, platform_data and addr 481arguments. 482 483If the subdev supports the s_config core ops, then that op is called with 484the irq and platform_data arguments after the subdev was setup. 485 486The :c:func:`v4l2_i2c_new_subdev` function will call 487:c:func:`v4l2_i2c_new_subdev_board`, internally filling a 488:c:type:`i2c_board_info` structure using the ``client_type`` and the 489``addr`` to fill it. 490 491V4L2 sub-device functions and data structures 492--------------------------------------------- 493 494.. kernel-doc:: include/media/v4l2-subdev.h 495