1.. SPDX-License-Identifier: GPL-2.0
2
3======================================
4Introduction to HID report descriptors
5======================================
6
7This chapter is meant to give a broad overview of what HID report
8descriptors are, and of how a casual (non-kernel) programmer can deal
9with HID devices that are not working well with Linux.
10
11.. contents::
12    :local:
13    :depth: 2
14
15.. toctree::
16   :maxdepth: 2
17
18   hidreport-parsing
19
20
21Introduction
22============
23
24HID stands for Human Interface Device, and can be whatever device you
25are using to interact with a computer, be it a mouse, a touchpad, a
26tablet, a microphone.
27
28Many HID devices work out the box, even if their hardware is different.
29For example, mice can have any number of buttons; they may have a
30wheel; movement sensitivity differs between different models, and so
31on. Nonetheless, most of the time everything just works, without the
32need to have specialized code in the kernel for every mouse model
33developed since 1970.
34
35This is because modern HID devices do advertise their capabilities
36through the *HID report descriptor*, a fixed set of bytes describing
37exactly what *HID reports* may be sent between the device and the host
38and the meaning of each individual bit in those reports. For example,
39a HID Report Descriptor may specify that "in a report with ID 3 the
40bits from 8 to 15 is the delta x coordinate of a mouse".
41
42The HID report itself then merely carries the actual data values
43without any extra meta information. Note that HID reports may be sent
44from the device ("Input Reports", i.e. input events), to the device
45("Output Reports" to e.g. change LEDs) or used for device configuration
46("Feature reports"). A device may support one or more HID reports.
47
48The HID subsystem is in charge of parsing the HID report descriptors,
49and converts HID events into normal input device interfaces (see
50Documentation/hid/hid-transport.rst). Devices may misbehave because the
51HID report descriptor provided by the device is wrong, or because it
52needs to be dealt with in a special way, or because some special
53device or interaction mode is not handled by the default code.
54
55The format of HID report descriptors is described by two documents,
56available from the `USB Implementers Forum <https://www.usb.org/>`_
57`HID web page <https://www.usb.org/hid>`_ address:
58
59 * the `HID USB Device Class Definition
60   <https://www.usb.org/document-library/device-class-definition-hid-111>`_ (HID Spec from now on)
61 * the `HID Usage Tables <https://usb.org/document-library/hid-usage-tables-14>`_ (HUT from now on)
62
63The HID subsystem can deal with different transport drivers
64(USB, I2C, Bluetooth, etc.). See Documentation/hid/hid-transport.rst.
65
66Parsing HID report descriptors
67==============================
68
69The current list of HID devices can be found at ``/sys/bus/hid/devices/``.
70For each device, say ``/sys/bus/hid/devices/0003\:093A\:2510.0002/``,
71one can read the corresponding report descriptor::
72
73  $ hexdump -C /sys/bus/hid/devices/0003\:093A\:2510.0002/report_descriptor
74  00000000  05 01 09 02 a1 01 09 01  a1 00 05 09 19 01 29 03  |..............).|
75  00000010  15 00 25 01 75 01 95 03  81 02 75 05 95 01 81 01  |..%.u.....u.....|
76  00000020  05 01 09 30 09 31 09 38  15 81 25 7f 75 08 95 03  |...0.1.8..%.u...|
77  00000030  81 06 c0 c0                                       |....|
78  00000034
79
80Optional: the HID report descriptor can be read also by
81directly accessing the hidraw driver [#hidraw]_.
82
83The basic structure of HID report descriptors is defined in the HID
84spec, while HUT "defines constants that can be interpreted by an
85application to identify the purpose and meaning of a data field in a
86HID report". Each entry is defined by at least two bytes, where the
87first one defines what type of value is following and is described in
88the HID spec, while the second one carries the actual value and is
89described in the HUT.
90
91HID report descriptors can, in principle, be painstakingly parsed by
92hand, byte by byte.
93
94A short introduction on how to do this is sketched in
95Documentation/hid/hidreport-parsing.rst; you only need to understand it
96if you need to patch HID report descriptors.
97
98In practice you should not parse HID report descriptors by hand; rather,
99you should use an existing parser. Among all the available ones
100
101  * the online `USB Descriptor and Request Parser
102    <http://eleccelerator.com/usbdescreqparser/>`_;
103  * `hidrdd <https://github.com/abend0c1/hidrdd>`_,
104    that provides very detailed and somewhat verbose descriptions
105    (verbosity can be useful if you are not familiar with HID report
106    descriptors);
107  * `hid-tools <https://gitlab.freedesktop.org/libevdev/hid-tools>`_,
108    a complete utility set that allows, among other things,
109    to record and replay the raw HID reports and to debug
110    and replay HID devices.
111    It is being actively developed by the Linux HID subsystem maintainers.
112
113Parsing the mouse HID report descriptor with `hid-tools
114<https://gitlab.freedesktop.org/libevdev/hid-tools>`_ leads to
115(explanations interposed)::
116
117    $ ./hid-decode /sys/bus/hid/devices/0003\:093A\:2510.0002/report_descriptor
118    # device 0:0
119    # 0x05, 0x01,		     // Usage Page (Generic Desktop)	    0
120    # 0x09, 0x02,		     // Usage (Mouse)			    2
121    # 0xa1, 0x01,		     // Collection (Application)	    4
122    # 0x09, 0x01,		     // Usage (Pointer)		    	    6
123    # 0xa1, 0x00,		     // Collection (Physical)  	    	    8
124    # 0x05, 0x09, 		     //	Usage Page (Button)		   10
125
126what follows is a button ::
127
128    # 0x19, 0x01, 		     //	Usage Minimum (1)		   12
129    # 0x29, 0x03, 		     //	Usage Maximum (3)		   14
130
131first button is button number 1, last button is button number 3 ::
132
133    # 0x15, 0x00, 		     //	Logical Minimum (0)		   16
134    # 0x25, 0x01, 		     //	Logical Maximum (1)		   18
135
136each button can send values from 0 up to including 1
137(i.e. they are binary buttons) ::
138
139    # 0x75, 0x01, 		     //	Report Size (1) 		   20
140
141each button is sent as exactly one bit ::
142
143    # 0x95, 0x03, 		     //	Report Count (3)		   22
144
145and there are three of those bits (matching the three buttons) ::
146
147    # 0x81, 0x02, 		     //	Input (Data,Var,Abs)		   24
148
149it's actual Data (not constant padding), they represent
150a single variable (Var) and their values are Absolute (not relative);
151See HID spec Sec. 6.2.2.5 "Input, Output, and Feature Items" ::
152
153    # 0x75, 0x05, 		     //	Report Size (5) 		   26
154
155five additional padding bits, needed to reach a byte ::
156
157    # 0x95, 0x01, 		     //	Report Count (1)		   28
158
159those five bits are repeated only once ::
160
161    # 0x81, 0x01, 		     //	Input (Cnst,Arr,Abs)		   30
162
163and take Constant (Cnst) values i.e. they can be ignored. ::
164
165    # 0x05, 0x01,		     // Usage Page (Generic Desktop)       32
166    # 0x09, 0x30,		     // Usage (X)			   34
167    # 0x09, 0x31,		     // Usage (Y)			   36
168    # 0x09, 0x38,		     // Usage (Wheel) 		    	   38
169
170The mouse has also two physical positions (Usage (X), Usage (Y))
171and a wheel (Usage (Wheel)) ::
172
173    # 0x15, 0x81, 		     //	Logical Minimum (-127)  	   40
174    # 0x25, 0x7f, 		     //	Logical Maximum (127)		   42
175
176each of them can send values ranging from -127 up to including 127 ::
177
178    # 0x75, 0x08, 		     //	Report Size (8) 		   44
179
180which is represented by eight bits ::
181
182    # 0x95, 0x03, 		     //	Report Count (3)		   46
183
184and there are three of those eight bits, matching X, Y and Wheel. ::
185
186    # 0x81, 0x06,		     // Input (Data,Var,Rel)  	    	   48
187
188This time the data values are Relative (Rel), i.e. they represent
189the change from the previously sent report (event) ::
190
191    # 0xc0,			     // End Collection 		    	   50
192    # 0xc0,			     // End Collection  		   51
193    #
194    R: 52 05 01 09 02 a1 01 09 01 a1 00 05 09 19 01 29 03 15 00 25 01 75 01 95 03 81 02 75 05 95 01 81 01 05 01 09 30 09 31 09 38 15 81 25 7f 75 08 95 03 81 06 c0 c0
195    N: device 0:0
196    I: 3 0001 0001
197
198
199This Report Descriptor tells us that the mouse input will be
200transmitted using four bytes: the first one for the buttons (three
201bits used, five for padding), the last three for the mouse X, Y and
202wheel changes, respectively.
203
204Indeed, for any event, the mouse will send a *report* of four bytes.
205We can check the values sent by resorting e.g. to the `hid-recorder`
206tool, from `hid-tools <https://gitlab.freedesktop.org/libevdev/hid-tools>`_:
207The sequence of bytes sent by clicking and releasing button 1, then button 2, then button 3 is::
208
209  $ sudo ./hid-recorder /dev/hidraw1
210
211  ....
212  output of hid-decode
213  ....
214
215  #  Button: 1  0  0 | # | X:	 0 | Y:    0 | Wheel:	 0
216  E: 000000.000000 4 01 00 00 00
217  #  Button: 0  0  0 | # | X:	 0 | Y:    0 | Wheel:	 0
218  E: 000000.183949 4 00 00 00 00
219  #  Button: 0  1  0 | # | X:	 0 | Y:    0 | Wheel:	 0
220  E: 000001.959698 4 02 00 00 00
221  #  Button: 0  0  0 | # | X:	 0 | Y:    0 | Wheel:	 0
222  E: 000002.103899 4 00 00 00 00
223  #  Button: 0  0  1 | # | X:	 0 | Y:    0 | Wheel:	 0
224  E: 000004.855799 4 04 00 00 00
225  #  Button: 0  0  0 | # | X:    0 | Y:    0 | Wheel:    0
226  E: 000005.103864 4 00 00 00 00
227
228This example shows that when button 2 is clicked,
229the bytes ``02 00 00 00`` are sent, and the immediately subsequent
230event (``00 00 00 00``) is the release of button 2 (no buttons are
231pressed, remember that the data values are *absolute*).
232
233If instead one clicks and holds button 1, then clicks and holds button
2342, releases button 1, and finally releases button 2, the reports are::
235
236  #  Button: 1  0  0 | # | X:    0 | Y:    0 | Wheel:    0
237  E: 000044.175830 4 01 00 00 00
238  #  Button: 1  1  0 | # | X:    0 | Y:    0 | Wheel:    0
239  E: 000045.975997 4 03 00 00 00
240  #  Button: 0  1  0 | # | X:    0 | Y:    0 | Wheel:    0
241  E: 000047.407930 4 02 00 00 00
242  #  Button: 0  0  0 | # | X:    0 | Y:    0 | Wheel:    0
243  E: 000049.199919 4 00 00 00 00
244
245where with ``03 00 00 00`` both buttons are pressed, and with the
246subsequent ``02 00 00 00`` button 1 is released while button 2 is still
247active.
248
249Output, Input and Feature Reports
250---------------------------------
251
252HID devices can have Input Reports, like in the mouse example, Output
253Reports, and Feature Reports. "Output" means that the information is
254sent to the device. For example, a joystick with force feedback will
255have some output; the led of a keyboard would need an output as well.
256"Input" means that data come from the device.
257
258"Feature"s are not meant to be consumed by the end user and define
259configuration options for the device. They can be queried from the host;
260when declared as *Volatile* they should be changed by the host.
261
262
263Collections, Report IDs and Evdev events
264========================================
265
266A single device can logically group data into different independent
267sets, called a *Collection*. Collections can be nested and there are
268different types of collections (see the HID spec 6.2.2.6
269"Collection, End Collection Items" for details).
270
271Different reports are identified by means of different *Report ID*
272fields, i.e. a number identifying the structure of the immediately
273following report.
274Whenever a Report ID is needed it is transmitted as the first byte of
275any report. A device with only one supported HID report (like the mouse
276example above) may omit the report ID.
277
278Consider the following HID report descriptor::
279
280  05 01 09 02 A1 01 85 01 05 09 19 01 29 05 15 00
281  25 01 95 05 75 01 81 02 95 01 75 03 81 01 05 01
282  09 30 09 31 16 00 F8 26 FF 07 75 0C 95 02 81 06
283  09 38 15 80 25 7F 75 08 95 01 81 06 05 0C 0A 38
284  02 15 80 25 7F 75 08 95 01 81 06 C0 05 01 09 02
285  A1 01 85 02 05 09 19 01 29 05 15 00 25 01 95 05
286  75 01 81 02 95 01 75 03 81 01 05 01 09 30 09 31
287  16 00 F8 26 FF 07 75 0C 95 02 81 06 09 38 15 80
288  25 7F 75 08 95 01 81 06 05 0C 0A 38 02 15 80 25
289  7F 75 08 95 01 81 06 C0 05 01 09 07 A1 01 85 05
290  05 07 15 00 25 01 09 29 09 3E 09 4B 09 4E 09 E3
291  09 E8 09 E8 09 E8 75 01 95 08 81 02 95 00 81 01
292  C0 05 0C 09 01 A1 01 85 06 15 00 25 01 75 01 95
293  01 09 3F 81 06 09 3F 81 06 09 3F 81 06 09 3F 81
294  06 09 3F 81 06 09 3F 81 06 09 3F 81 06 09 3F 81
295  06 C0 05 0C 09 01 A1 01 85 03 09 05 15 00 26 FF
296  00 75 08 95 02 B1 02 C0
297
298After parsing it (try to parse it on your own using the suggested
299tools!) one can see that the device presents two ``Mouse`` Application
300Collections (with reports identified by Reports IDs 1 and 2,
301respectively), a ``Keypad`` Application Collection (whose report is
302identified by the Report ID 5) and two ``Consumer Controls`` Application
303Collections, (with Report IDs 6 and 3, respectively). Note, however,
304that a device can have different Report IDs for the same Application
305Collection.
306
307The data sent will begin with the Report ID byte, and will be followed
308by the corresponding information. For example, the data transmitted for
309the last consumer control::
310
311  0x05, 0x0C,        // Usage Page (Consumer)
312  0x09, 0x01,        // Usage (Consumer Control)
313  0xA1, 0x01,        // Collection (Application)
314  0x85, 0x03,        //   Report ID (3)
315  0x09, 0x05,        //   Usage (Headphone)
316  0x15, 0x00,        //   Logical Minimum (0)
317  0x26, 0xFF, 0x00,  //   Logical Maximum (255)
318  0x75, 0x08,        //   Report Size (8)
319  0x95, 0x02,        //   Report Count (2)
320  0xB1, 0x02,        //   Feature (Data,Var,Abs,No Wrap,Linear,Preferred State,No Null Position,Non-volatile)
321  0xC0,              // End Collection
322
323will be of three bytes: the first for the Report ID (3), the next two
324for the headphone, with two (``Report Count (2)``) bytes
325(``Report Size (8)``), each ranging from 0 (``Logical Minimum (0)``)
326to 255 (``Logical Maximum (255)``).
327
328All the Input data sent by the device should be translated into
329corresponding Evdev events, so that the remaining part of the stack can
330know what is going on, e.g. the bit for the first button translates into
331the ``EV_KEY/BTN_LEFT`` evdev event and relative X movement translates
332into the ``EV_REL/REL_X`` evdev event".
333
334Events
335======
336
337In Linux, one ``/dev/input/event*`` is created for each ``Application
338Collection``. Going back to the mouse example, and repeating the
339sequence where one clicks and holds button 1, then clicks and holds
340button 2, releases button 1, and finally releases button 2, one gets::
341
342  $ sudo libinput record /dev/input/event1
343  # libinput record
344  version: 1
345  ndevices: 1
346  libinput:
347    version: "1.23.0"
348    git: "unknown"
349  system:
350    os: "opensuse-tumbleweed:20230619"
351    kernel: "6.3.7-1-default"
352    dmi: "dmi:bvnHP:bvrU77Ver.01.05.00:bd03/24/2022:br5.0:efr20.29:svnHP:pnHPEliteBook64514inchG9NotebookPC:pvr:rvnHP:rn89D2:rvrKBCVersion14.1D.00:cvnHP:ct10:cvr:sku5Y3J1EA#ABZ:"
353  devices:
354  - node: /dev/input/event1
355    evdev:
356      # Name: PixArt HP USB Optical Mouse
357      # ID: bus 0x3 vendor 0x3f0 product 0x94a version 0x111
358      # Supported Events:
359      # Event type 0 (EV_SYN)
360      # Event type 1 (EV_KEY)
361      #   Event code 272 (BTN_LEFT)
362      #   Event code 273 (BTN_RIGHT)
363      #   Event code 274 (BTN_MIDDLE)
364      # Event type 2 (EV_REL)
365      #   Event code 0 (REL_X)
366      #   Event code 1 (REL_Y)
367      #   Event code 8 (REL_WHEEL)
368      #   Event code 11 (REL_WHEEL_HI_RES)
369      # Event type 4 (EV_MSC)
370      #   Event code 4 (MSC_SCAN)
371      # Properties:
372      name: "PixArt HP USB Optical Mouse"
373      id: [3, 1008, 2378, 273]
374      codes:
375  	0: [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15] # EV_SYN
376  	1: [272, 273, 274] # EV_KEY
377  	2: [0, 1, 8, 11] # EV_REL
378  	4: [4] # EV_MSC
379      properties: []
380    hid: [
381      0x05, 0x01, 0x09, 0x02, 0xa1, 0x01, 0x09, 0x01, 0xa1, 0x00, 0x05, 0x09, 0x19, 0x01, 0x29, 0x03,
382      0x15, 0x00, 0x25, 0x01, 0x95, 0x08, 0x75, 0x01, 0x81, 0x02, 0x05, 0x01, 0x09, 0x30, 0x09, 0x31,
383      0x09, 0x38, 0x15, 0x81, 0x25, 0x7f, 0x75, 0x08, 0x95, 0x03, 0x81, 0x06, 0xc0, 0xc0
384    ]
385    udev:
386      properties:
387      - ID_INPUT=1
388      - ID_INPUT_MOUSE=1
389      - LIBINPUT_DEVICE_GROUP=3/3f0/94a:usb-0000:05:00.3-2
390    quirks:
391    events:
392    # Current time is 12:31:56
393    - evdev:
394      - [  0,	   0,	4,   4,      30] # EV_MSC / MSC_SCAN		     30 (obfuscated)
395      - [  0,	   0,	1, 272,       1] # EV_KEY / BTN_LEFT		      1
396      - [  0,	   0,	0,   0,       0] # ------------ SYN_REPORT (0) ---------- +0ms
397    - evdev:
398      - [  1, 207892,	4,   4,      30] # EV_MSC / MSC_SCAN		     30 (obfuscated)
399      - [  1, 207892,	1, 273,       1] # EV_KEY / BTN_RIGHT		      1
400      - [  1, 207892,	0,   0,       0] # ------------ SYN_REPORT (0) ---------- +1207ms
401    - evdev:
402      - [  2, 367823,	4,   4,      30] # EV_MSC / MSC_SCAN		     30 (obfuscated)
403      - [  2, 367823,	1, 272,       0] # EV_KEY / BTN_LEFT		      0
404      - [  2, 367823,	0,   0,       0] # ------------ SYN_REPORT (0) ---------- +1160ms
405    # Current time is 12:32:00
406    - evdev:
407      - [  3, 247617,	4,   4,      30] # EV_MSC / MSC_SCAN		     30 (obfuscated)
408      - [  3, 247617,	1, 273,       0] # EV_KEY / BTN_RIGHT		      0
409      - [  3, 247617,   0,   0,       0] # ------------ SYN_REPORT (0) ---------- +880ms
410
411Note: if ``libinput record`` is not available on your system try using
412``evemu-record``.
413
414When something does not work
415============================
416
417There can be a number of reasons why a device does not behave
418correctly. For example
419
420* The HID report descriptor provided by the HID device may be wrong
421  because e.g.
422
423  * it does not follow the standard, so that the kernel
424    will not able to make sense of the HID report descriptor;
425  * the HID report descriptor *does not match* what is actually
426    sent by the device (this can be verified by reading the raw HID
427    data);
428* the HID report descriptor may need some "quirks" (see later on).
429
430As a consequence, a ``/dev/input/event*`` may not be created
431for each Application Collection, and/or the events
432there may not match what you would expect.
433
434
435Quirks
436------
437
438There are some known peculiarities of HID devices that the kernel
439knows how to fix - these are called the HID quirks and a list of those
440is available in `include/linux/hid.h`.
441
442Should this be the case, it should be enough to add the required quirk
443in the kernel, for the HID device at hand. This can be done in the file
444`drivers/hid/hid-quirks.c`. How to do it should be relatively
445straightforward after looking into the file.
446
447The list of currently defined quirks, from `include/linux/hid.h`, is
448
449.. kernel-doc:: include/linux/hid.h
450   :doc: HID quirks
451
452Quirks for USB devices can be specified while loading the usbhid module,
453see ``modinfo usbhid``, although the proper fix should go into
454hid-quirks.c and **be submitted upstream**.
455See Documentation/process/submitting-patches.rst for guidelines on how
456to submit a patch. Quirks for other busses need to go into hid-quirks.c.
457
458Fixing HID report descriptors
459-----------------------------
460
461Should you need to patch HID report descriptors the easiest way is to
462resort to eBPF, as described in Documentation/hid/hid-bpf.rst.
463
464Basically, you can change any byte of the original HID report
465descriptor. The examples in samples/hid should be a good starting point
466for your code, see e.g. `samples/hid/hid_mouse.bpf.c`::
467
468  SEC("fmod_ret/hid_bpf_rdesc_fixup")
469  int BPF_PROG(hid_rdesc_fixup, struct hid_bpf_ctx *hctx)
470  {
471    ....
472       data[39] = 0x31;
473       data[41] = 0x30;
474    return 0;
475  }
476
477Of course this can be also done within the kernel source code, see e.g.
478`drivers/hid/hid-aureal.c` or `drivers/hid/hid-samsung.c` for a slightly
479more complex file.
480
481Check Documentation/hid/hidreport-parsing.rst if you need any help
482navigating the HID manuals and understanding the exact meaning of
483the HID report descriptor hex numbers.
484
485Whatever solution you come up with, please remember to **submit the
486fix to the HID maintainers**, so that it can be directly integrated in
487the kernel and that particular HID device will start working for
488everyone else. See Documentation/process/submitting-patches.rst for
489guidelines on how to do this.
490
491
492Modifying the transmitted data on the fly
493-----------------------------------------
494
495Using eBPF it is also possible to modify the data exchanged with the
496device. See again the examples in `samples/hid`.
497
498Again, **please post your fix**, so that it can be integrated in the
499kernel!
500
501Writing a specialized driver
502----------------------------
503
504This should really be your last resort.
505
506
507.. rubric:: Footnotes
508
509.. [#hidraw] read hidraw: see Documentation/hid/hidraw.rst and
510  file `samples/hidraw/hid-example.c` for an example.
511  The output of ``hid-example`` would be, for the same mouse::
512
513    $ sudo ./hid-example
514    Report Descriptor Size: 52
515    Report Descriptor:
516    5 1 9 2 a1 1 9 1 a1 0 5 9 19 1 29 3 15 0 25 1 75 1 95 3 81 2 75 5 95 1 81 1 5 1 9 30 9 31 9 38 15 81 25 7f 75 8 95 3 81 6 c0 c0
517
518    Raw Name: PixArt USB Optical Mouse
519    Raw Phys: usb-0000:05:00.4-2.3/input0
520    Raw Info:
521            bustype: 3 (USB)
522            vendor: 0x093a
523            product: 0x2510
524    ...
525