Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | $Id: input-programming.txt,v 1.4 2001/05/04 09:47:14 vojtech Exp $ |
| 2 | |
| 3 | Programming input drivers |
| 4 | ~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 5 | |
| 6 | 1. Creating an input device driver |
| 7 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 8 | |
| 9 | 1.0 The simplest example |
| 10 | ~~~~~~~~~~~~~~~~~~~~~~~~ |
| 11 | |
| 12 | Here comes a very simple example of an input device driver. The device has |
| 13 | just one button and the button is accessible at i/o port BUTTON_PORT. When |
| 14 | pressed or released a BUTTON_IRQ happens. The driver could look like: |
| 15 | |
| 16 | #include <linux/input.h> |
| 17 | #include <linux/module.h> |
| 18 | #include <linux/init.h> |
| 19 | |
| 20 | #include <asm/irq.h> |
| 21 | #include <asm/io.h> |
| 22 | |
| 23 | static void button_interrupt(int irq, void *dummy, struct pt_regs *fp) |
| 24 | { |
| 25 | input_report_key(&button_dev, BTN_1, inb(BUTTON_PORT) & 1); |
| 26 | input_sync(&button_dev); |
| 27 | } |
| 28 | |
| 29 | static int __init button_init(void) |
| 30 | { |
| 31 | if (request_irq(BUTTON_IRQ, button_interrupt, 0, "button", NULL)) { |
| 32 | printk(KERN_ERR "button.c: Can't allocate irq %d\n", button_irq); |
| 33 | return -EBUSY; |
| 34 | } |
| 35 | |
| 36 | button_dev.evbit[0] = BIT(EV_KEY); |
| 37 | button_dev.keybit[LONG(BTN_0)] = BIT(BTN_0); |
| 38 | |
| 39 | input_register_device(&button_dev); |
| 40 | } |
| 41 | |
| 42 | static void __exit button_exit(void) |
| 43 | { |
| 44 | input_unregister_device(&button_dev); |
| 45 | free_irq(BUTTON_IRQ, button_interrupt); |
| 46 | } |
| 47 | |
| 48 | module_init(button_init); |
| 49 | module_exit(button_exit); |
| 50 | |
| 51 | 1.1 What the example does |
| 52 | ~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 53 | |
| 54 | First it has to include the <linux/input.h> file, which interfaces to the |
| 55 | input subsystem. This provides all the definitions needed. |
| 56 | |
| 57 | In the _init function, which is called either upon module load or when |
| 58 | booting the kernel, it grabs the required resources (it should also check |
| 59 | for the presence of the device). |
| 60 | |
| 61 | Then it sets the input bitfields. This way the device driver tells the other |
| 62 | parts of the input systems what it is - what events can be generated or |
| 63 | accepted by this input device. Our example device can only generate EV_KEY type |
| 64 | events, and from those only BTN_0 event code. Thus we only set these two |
| 65 | bits. We could have used |
| 66 | |
| 67 | set_bit(EV_KEY, button_dev.evbit); |
| 68 | set_bit(BTN_0, button_dev.keybit); |
| 69 | |
| 70 | as well, but with more than single bits the first approach tends to be |
| 71 | shorter. |
| 72 | |
| 73 | Then the example driver registers the input device structure by calling |
| 74 | |
| 75 | input_register_device(&button_dev); |
| 76 | |
| 77 | This adds the button_dev structure to linked lists of the input driver and |
| 78 | calls device handler modules _connect functions to tell them a new input |
| 79 | device has appeared. Because the _connect functions may call kmalloc(, |
| 80 | GFP_KERNEL), which can sleep, input_register_device() must not be called |
| 81 | from an interrupt or with a spinlock held. |
| 82 | |
| 83 | While in use, the only used function of the driver is |
| 84 | |
| 85 | button_interrupt() |
| 86 | |
| 87 | which upon every interrupt from the button checks its state and reports it |
| 88 | via the |
| 89 | |
| 90 | input_report_key() |
| 91 | |
| 92 | call to the input system. There is no need to check whether the interrupt |
| 93 | routine isn't reporting two same value events (press, press for example) to |
| 94 | the input system, because the input_report_* functions check that |
| 95 | themselves. |
| 96 | |
| 97 | Then there is the |
| 98 | |
| 99 | input_sync() |
| 100 | |
| 101 | call to tell those who receive the events that we've sent a complete report. |
| 102 | This doesn't seem important in the one button case, but is quite important |
| 103 | for for example mouse movement, where you don't want the X and Y values |
| 104 | to be interpreted separately, because that'd result in a different movement. |
| 105 | |
| 106 | 1.2 dev->open() and dev->close() |
| 107 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 108 | |
| 109 | In case the driver has to repeatedly poll the device, because it doesn't |
| 110 | have an interrupt coming from it and the polling is too expensive to be done |
| 111 | all the time, or if the device uses a valuable resource (eg. interrupt), it |
| 112 | can use the open and close callback to know when it can stop polling or |
| 113 | release the interrupt and when it must resume polling or grab the interrupt |
| 114 | again. To do that, we would add this to our example driver: |
| 115 | |
| 116 | int button_used = 0; |
| 117 | |
| 118 | static int button_open(struct input_dev *dev) |
| 119 | { |
| 120 | if (button_used++) |
| 121 | return 0; |
| 122 | |
| 123 | if (request_irq(BUTTON_IRQ, button_interrupt, 0, "button", NULL)) { |
| 124 | printk(KERN_ERR "button.c: Can't allocate irq %d\n", button_irq); |
| 125 | button_used--; |
| 126 | return -EBUSY; |
| 127 | } |
| 128 | |
| 129 | return 0; |
| 130 | } |
| 131 | |
| 132 | static void button_close(struct input_dev *dev) |
| 133 | { |
| 134 | if (!--button_used) |
| 135 | free_irq(IRQ_AMIGA_VERTB, button_interrupt); |
| 136 | } |
| 137 | |
| 138 | static int __init button_init(void) |
| 139 | { |
| 140 | ... |
| 141 | button_dev.open = button_open; |
| 142 | button_dev.close = button_close; |
| 143 | ... |
| 144 | } |
| 145 | |
| 146 | Note the button_used variable - we have to track how many times the open |
| 147 | function was called to know when exactly our device stops being used. |
| 148 | |
| 149 | The open() callback should return a 0 in case of success or any nonzero value |
| 150 | in case of failure. The close() callback (which is void) must always succeed. |
| 151 | |
| 152 | 1.3 Basic event types |
| 153 | ~~~~~~~~~~~~~~~~~~~~~ |
| 154 | |
| 155 | The most simple event type is EV_KEY, which is used for keys and buttons. |
| 156 | It's reported to the input system via: |
| 157 | |
| 158 | input_report_key(struct input_dev *dev, int code, int value) |
| 159 | |
| 160 | See linux/input.h for the allowable values of code (from 0 to KEY_MAX). |
| 161 | Value is interpreted as a truth value, ie any nonzero value means key |
| 162 | pressed, zero value means key released. The input code generates events only |
| 163 | in case the value is different from before. |
| 164 | |
| 165 | In addition to EV_KEY, there are two more basic event types: EV_REL and |
| 166 | EV_ABS. They are used for relative and absolute values supplied by the |
| 167 | device. A relative value may be for example a mouse movement in the X axis. |
| 168 | The mouse reports it as a relative difference from the last position, |
| 169 | because it doesn't have any absolute coordinate system to work in. Absolute |
| 170 | events are namely for joysticks and digitizers - devices that do work in an |
| 171 | absolute coordinate systems. |
| 172 | |
| 173 | Having the device report EV_REL buttons is as simple as with EV_KEY, simply |
| 174 | set the corresponding bits and call the |
| 175 | |
| 176 | input_report_rel(struct input_dev *dev, int code, int value) |
| 177 | |
| 178 | function. Events are generated only for nonzero value. |
| 179 | |
| 180 | However EV_ABS requires a little special care. Before calling |
| 181 | input_register_device, you have to fill additional fields in the input_dev |
| 182 | struct for each absolute axis your device has. If our button device had also |
| 183 | the ABS_X axis: |
| 184 | |
| 185 | button_dev.absmin[ABS_X] = 0; |
| 186 | button_dev.absmax[ABS_X] = 255; |
| 187 | button_dev.absfuzz[ABS_X] = 4; |
| 188 | button_dev.absflat[ABS_X] = 8; |
| 189 | |
| 190 | This setting would be appropriate for a joystick X axis, with the minimum of |
| 191 | 0, maximum of 255 (which the joystick *must* be able to reach, no problem if |
| 192 | it sometimes reports more, but it must be able to always reach the min and |
| 193 | max values), with noise in the data up to +- 4, and with a center flat |
| 194 | position of size 8. |
| 195 | |
| 196 | If you don't need absfuzz and absflat, you can set them to zero, which mean |
| 197 | that the thing is precise and always returns to exactly the center position |
| 198 | (if it has any). |
| 199 | |
| 200 | 1.4 The void *private field |
| 201 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 202 | |
| 203 | This field in the input structure can be used to point to any private data |
| 204 | structures in the input device driver, in case the driver handles more than |
| 205 | one device. You'll need it in the open and close callbacks. |
| 206 | |
| 207 | 1.5 NBITS(), LONG(), BIT() |
| 208 | ~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 209 | |
| 210 | These three macros from input.h help some bitfield computations: |
| 211 | |
| 212 | NBITS(x) - returns the length of a bitfield array in longs for x bits |
| 213 | LONG(x) - returns the index in the array in longs for bit x |
| 214 | BIT(x) - returns the index in a long for bit x |
| 215 | |
| 216 | 1.6 The number, id* and name fields |
| 217 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 218 | |
| 219 | The dev->number is assigned by the input system to the input device when it |
| 220 | is registered. It has no use except for identifying the device to the user |
| 221 | in system messages. |
| 222 | |
| 223 | The dev->name should be set before registering the input device by the input |
| 224 | device driver. It's a string like 'Generic button device' containing a |
| 225 | user friendly name of the device. |
| 226 | |
| 227 | The id* fields contain the bus ID (PCI, USB, ...), vendor ID and device ID |
| 228 | of the device. The bus IDs are defined in input.h. The vendor and device ids |
| 229 | are defined in pci_ids.h, usb_ids.h and similar include files. These fields |
| 230 | should be set by the input device driver before registering it. |
| 231 | |
| 232 | The idtype field can be used for specific information for the input device |
| 233 | driver. |
| 234 | |
| 235 | The id and name fields can be passed to userland via the evdev interface. |
| 236 | |
| 237 | 1.7 The keycode, keycodemax, keycodesize fields |
| 238 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 239 | |
| 240 | These two fields will be used for any input devices that report their data |
| 241 | as scancodes. If not all scancodes can be known by autodetection, they may |
| 242 | need to be set by userland utilities. The keycode array then is an array |
| 243 | used to map from scancodes to input system keycodes. The keycode max will |
| 244 | contain the size of the array and keycodesize the size of each entry in it |
| 245 | (in bytes). |
| 246 | |
| 247 | 1.8 Key autorepeat |
| 248 | ~~~~~~~~~~~~~~~~~~ |
| 249 | |
| 250 | ... is simple. It is handled by the input.c module. Hardware autorepeat is |
| 251 | not used, because it's not present in many devices and even where it is |
| 252 | present, it is broken sometimes (at keyboards: Toshiba notebooks). To enable |
| 253 | autorepeat for your device, just set EV_REP in dev->evbit. All will be |
| 254 | handled by the input system. |
| 255 | |
| 256 | 1.9 Other event types, handling output events |
| 257 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 258 | |
| 259 | The other event types up to now are: |
| 260 | |
| 261 | EV_LED - used for the keyboard LEDs. |
| 262 | EV_SND - used for keyboard beeps. |
| 263 | |
| 264 | They are very similar to for example key events, but they go in the other |
| 265 | direction - from the system to the input device driver. If your input device |
| 266 | driver can handle these events, it has to set the respective bits in evbit, |
| 267 | *and* also the callback routine: |
| 268 | |
| 269 | button_dev.event = button_event; |
| 270 | |
| 271 | int button_event(struct input_dev *dev, unsigned int type, unsigned int code, int value); |
| 272 | { |
| 273 | if (type == EV_SND && code == SND_BELL) { |
| 274 | outb(value, BUTTON_BELL); |
| 275 | return 0; |
| 276 | } |
| 277 | return -1; |
| 278 | } |
| 279 | |
| 280 | This callback routine can be called from an interrupt or a BH (although that |
| 281 | isn't a rule), and thus must not sleep, and must not take too long to finish. |