Driver Model


  |  +---------------------------------------------------------------+
S |  | 'gpm -t msc -m /var/run/lirc/lircm' or a well configured X    |
O |  | (see section 'Configuring lircmd') for details                |
F |  +--------------+------------------------------------------------+
T |                 |
W |   /var/run/lirc/lircm (named pipe (FIFO) ==> one connection)
A |                 |
R |     +-----------+-----------+   +-------------------------------+
E |     | mouse daemon (lircmd) |   | tools (irexec, irxevent, ...) |
  |     | configured with       |   | configured with ~/.lircrc     |
  |     | lircmd.conf           |   |                               |
  |     +-----------+-----------+   +-----------+-------------------+
  |                 |                           |
  |                 +-------------+-------------+
  |                               |
  |              /var/run/lirc/lircd (socket ==> multiple connections)
  |                               |
S |              +----------------+--------------------------+
O |              | decoder daemon (lircd), irrecord or mode2 |    TCP/IP
F |              | lircd is configured through lircd.conf    +---  port
T | User space   |                                           |     8765
W |              +------------------+------------------------+
A |                                 |            |
R |                                 |            |
E |                                 |            |
  |                                 |   /dev/uinput (Linux input layer)
  |                                 |
  |                                 |
  +---------------------------------+----------------------------------
  | Kernel space                    |              (character device
  |                                 |                  driver ==>
  |                    +------------+----------+    one connection)
  |                    |                       |
  |                /dev/lirc               /dev/ttySx
  |                    |                       |
  |       +------------+-------------+   +-----+---------------+
  |       |  LIRC device driver      |   | Linux serial driver |
  |       | (with ioctl-interface)   |   |                     |
  |       +------------+-------------+   +----------+----------+
  |                    |                            |
--+--------------------+----------------------------+------------------
  |                    |                            |
  |         +----------+------------+               |
  |         |                       |               |
  | +-------+----------------+ +----+-----+ +-------+-----------------+
H | | serial / parallel port | | TV cards | | Irman/RemoteMaster/etc. |
W | +------------------------+ +----------+ +-------------------------+
  |

Formats


  • /dev/lirc:

    highly depends on the mode selected with ioctls:

    LIRC_MODE_MODE2

    outputs packets containing an int value describing a IR signal

    • bits 0-23 contain the length of the pulse/space in microseconds
    • bits 31-24 can be:
      • 0: space
      • 1: pulse
      • 2: timeout (has to be enabled using LIRC_SET_REC_TIMEOUT ioctl and is not supported by all receivers)
    • all other values are reserved

    Lengths greater than or equal to 16 seconds are clamped to 0xffffff.

    References:

    • drivers/lirc_serial/lirc_serial.c
    • drivers/lirc_parallel/lirc_parallel.c
    • tools/mode2.c (dumps the output from the driver to stdout)

    LIRC_MODE_LIRCCODE

    outputs codes of configurable length in big endian byte order

  • /var/run/lirc/lircd:

    outputs strings containing all information about the remote and the pressed button.

    References:

    • daemons/lircd.c
    • tools/irw.c

  • /var/run/lirc/lircm:

    • MouseSystems
      5 byte packets:
      • byte 1: button information
      • byte 2: change on X axis
      • byte 3: change on Y axis
      • byte 4,5: 0
    • IMPS/2
      4 byte packets: check the source code for details
    • IntelliMouse
      4 byte packets: check the source code for details

    References:

    • daemons/lircmd.c

Writing TV card drivers using lirc_dev


The lirc_dev module is a helper and abstraction layer for other modules. It registers /dev/lirc device in a system (including support for devfs) and waits for plugin registration. After that it serves device requests (open, read, poll, ioctl, close) and if needed calls callback functions from plugin(s) to communicate with the physical device.

Plugins can be registered and unregistered many times. The current implementation allows two concurrent plugins, but can be easily changed by increasing the MAX_IRCTL_DEVICES definition. It also allows receiving of scan codes, which have more than 8 bits. Current limit for a scan code is 16*8 bits and also can be changed by increasing the BUFLEN definition.

For an API description see lirc_dev.h. The lirc_gpio module can be treated as examples of using this API.
This code contains many lines with debug messages (activated by debug option) and they will sustain until more tests will be performed.

Warning: Due to the used kernel API it requires kernel 2.2.4 or higher.
Any suggestions and questions are welcome. Artur Lipowski


Writing Applications for LIRC


As LIRC is able to both receive and send IR commands there are two possible types of applications. Programs that send IR commands like xrc and irsend or programs that receive commands like irexec, irxevent and irpty. Both types of applications will have to connect to the lircd daemon using the socket interface usually located in /var/run/lirc/lircd. Communication on the socket uses human readable format. The end of a line is indicated by a newline character.

Whenever lircd receives a IR signal it will broadcast the following string to each client:

  <code> <repeat count> <button name> <remote control name>

code is a 64-bit encoding (in hexadecimal representation) of the IR signal. It's usage in applications is deprecated and should be ignored. The repeat count shows how long the user has been holding down a button. The counter will start at 0 and increment each time a new IR signal has been received. The button name and remote control name are defined in the lircd config file. Their purpose should be quite self-explanatory. They must not contain any whitespace.
The only other situation when lircd broadcasts to all clients is when it receives the SIGHUP signal and successfully re-reads its config file. Then it will send a SIGHUP packet to its clients indicating that its configuration might have changed. This feature is e.g. used in xrc to rebuild the list of supported remote controls each time lircd's configuration changes. The format of the packet will be explained later.

Applications that want to send out IR commands can use the following commands:

  SEND_ONCE <remote control name> <button name> [<repeat count>]
  SEND_START <remote control name> <button name>
  SEND_STOP <remote control name> <button name>

The SEND_ONCE directive tells lircd to send the IR signal associated with the given remote control and button name, and then repeat it repeat count times. repeat count is a decimal number between 0 and repeat_max. The latter can be given as a command line argument to lircd, and defaults to 600. If repeat count is not specified or is less than the minimum number of repeats for the selected remote control, the minimum value will be used. SEND_START tells lircd to start repeating the given button until it receives a SEND_STOP command. However, the number of repeats is limited to repeat_max. lircd won't accept any new send commands while it is repeating.

lircd also understands the following commands:

  VERSION
  LIST [<remote control name>]

The response to the VERSION command will be a packet containing lircd's version.
The LIST command without further arguments can be used to get a list of all remote controls known to lircd. If a name of a supported remote control is given as argument all buttons of the given remote control are listed in the reply packet. Have a look at xrc for an example how this can be used.

There still remains to explain the format of lircd's reply packets. Here is a formal description of the packets:

  BEGIN
  <command>
  [SUCCESS|ERROR]
  [DATA
  n
  n lines of data]
  END

The protocol guarantees that broadcasted messages won't interfere with reply packets. But broadcasts may appear at any point between packets. command is the command lircd is currently processing. Its an exact copy of the command the client application has sent. The only exception are SIGHUP packages where command is substituted with SIGHUP. Note that SIGHUP packages may appear just after you have sent a command to lircd, so you have to make sure you don't confuse them with replies. SIGHUP packages come without any further data while each reply to a command contains either SUCCESS or ERROR indicating the result of processing the command. In case of an error the following data is a message explaining the problem. This message can be used to create an error message for the user.
If the command was successful, data is only sent for the commands that return some information. Note that a packet containing 0 lines of data can be a valid reply.


The lirc_client library


If you only want to make your application receive IR commands and if you don't want to mess with all the protocol stuff you can use the lirc_client library that comes with LIRC since version 0.6.0. With the help of this library your program can look as simple as this:


/****************************************************************************
 ** irexec.c ****************************************************************
 ****************************************************************************
 *
 * irexec  - execute programs according to the pressed remote control buttons
 *
 * Copyright (C) 1998 Trent Piepho <xyzzy@u.washington.edu>
 * Copyright (C) 1998 Christoph Bartelmus <lirc@bartelmus.de>
 *
 */

#ifdef HAVE_CONFIG_H
# include <config.h>
#endif

#include <errno.h>
#include <unistd.h>
#include <stdarg.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include "lirc_client.h"

char *progname;

int main(int argc, char *argv[])
{
        struct lirc_config *config;

        progname=argv[0];
        if(argc>2)
        {
                fprintf(stderr,"Usage: %s <config file>\n",progname);
                exit(EXIT_FAILURE);
        }
        if(lirc_init("irexec",1)==-1) exit(EXIT_FAILURE);

        if(lirc_readconfig(argc==2 ? argv[1]:NULL,&config,NULL)==0)
        {
                char *code;
                char *c;
                int ret;

                while(lirc_nextcode(&code)==0)
                {
                        if(code==NULL) continue;
                        while((ret=lirc_code2char(config,code,&c))==0 &&
                              c!=NULL)
                        {
#ifdef DEBUG
                                printf("Execing command \"%s\"\n",c);
#endif
                                system(c);
                        }
                        free(code);
                        if(ret==-1) break;
                }
                lirc_freeconfig(config);
        }

        lirc_deinit();
        exit(EXIT_SUCCESS);
}

Before anything else you have to include the header file for the lirc_client library. This is done with

#include <lirc/lirc_client.h>

Note that our example differs in this point because it was taken directly from the lirc-0.6.0 source that comes with its own lirc_client.h but we have to use the one that is already installed on the system.

The next step is to initialize the library code with lirc_init(). This function connects to lircd and does some internal init stuff.

int lirc_init(char *prog,int verbose);

The first argument to this function is the string users will have to provide as prog token in their .lircrc config files. If the second argument is non-zero error messages will be printed to stderr. Otherwise no error messages will ever be displayed. This function returns the file descriptor of the socket that is connected to lircd or -1 if an error occurred.

The example continues by reading a config file. This is done by the lirc_readconfig() function:

int lirc_readconfig(char *file,struct lirc_config **config,
                    int (check)(char *s));

If you want to load the default config file you should pass NULL as first argument. If you want to load some other config file the file argument should contain the complete path to the file. Your program should give the user the possibility to use an other than the default config file. You should also be able to load multiple config files by calling this function several times.
The config argument is used to pass the pointer to the config file data structures back to your application. You will need it for calls to the lirc_code2char() function. The last argument is a call-back function that can be used to do syntax checks with the config strings. The library code will call the call-back function for all config strings where the prog token in the config file matches the prog string you provided with the lirc_init() function. If there is an error in the config string the call-back function should return -1, otherwise 0. If you don't need to do any syntax checks you can pass NULL here. The function returns -1 if an error occurred, 0 otherwise.

The lirc_nextcode() function blocks until there is something available on the lircd socket. This way it can be used in the main loop of your program like in our example.

int lirc_nextcode(char **code);

If an error occurs (usually this means that the socket has been closed by the daemon) this function returns -1. Otherwise it returns 0 and code points to the next string available in the data stream. This string has to be freed by your application using the free(3) function. If no complete string is available code will be NULL.
If you use some GUI-toolkit for your program then you probably won't be able to use this function in your program's main loop because this is already handled by the GUI-toolkit. In this situation you should use the call-back abilities of the toolkit that will notify you whenever there is some input available from a file descriptor (you get the file descriptor from the lirc_init() function). E.g. you can use the gdk_input_add()/gdk_input_remove() functions with gtk or the QSocketNotifier class with Qt. If you don't have such functionality in your toolkit or can't use it for some reason you can still use SIGIO signals for this purpose. Check the documentation for your GUI-toolkit and signal(2) for further information.
Please note that using call-backs you still have to use some kind of while loop to read strings from the socket because several strings may be available in the data stream and you will only get a notification for the first one. This poses a problem for us because lirc_nextcode() blocks until there is something available from the socket which is not what we need here. You can solve this problem by setting the O_NONBLOCK flag for the socket using the fcntl(2) function. Have a look at the current xirw code that is available from the LIRC homepage for an implementation example.

To get the config string that the user has provided in the config file in response to a button press you use the following function:

int lirc_code2char(struct lirc_config *config,char *code,char **string);

config is a pointer to the config file data structure that you can get with lirc_readconfig() and code is the code transmitted to your application on the lircd socket. If an action should be taken string will point to the config string the user has provided in the config file. The user might want to take several actions on a button press so you have to execute this function until string is NULL, which means that no more actions shall be taken, or an error occurs. The function returns -1 if an error occurred, 0 otherwise.

In our example there are only two clean-up functions to be explained.

void lirc_freeconfig(struct lirc_config *config);

This functions frees the data structures associated with config.

int lirc_deinit();

lirc_deinit() closes the connection to lircd and does some internal clean-up stuff.

I encourage you to use autoconf and automake for your projects. To check for the lirc_client library all you have to insert into your configure.ac (or configure.in) file is the following:

dnl Check for LIRC client support
dnl This really is not worth making a separate file for it.

have_lirc=yes
AC_REQUIRE_CPP
AC_CHECK_LIB(lirc_client,lirc_init,
  AC_CHECK_HEADER(lirc/lirc_client.h,true,have_lirc=no),have_lirc=no)

if test "$have_lirc" = "yes"; then
dnl  AC_DEFINE(HAVE_LIRC);
  true;
else
  AC_MSG_ERROR([*** LIRC client support not available ***]);
fi

There is also a more complex m4 macro in the contrib directory of the current LIRC distribution if you plan to add LIRC support to your application without using the lirc_client library.

While developing LIRC applications you might find a emulator for lircd useful. With this emulator you don't need a remote control to generate LIRC events. That way you can develop LIRC applications even if you don't have a LIRC compatible device yourself.


lircrcd protocol


lircrcd syntactically uses the same protocol as lircd described in the last section. It supports the following commands:

IDENT ident

Each program connecting to lircrcd identifies itself using this program. ident is the string that is used in the prog token inside the .lircrc file.

CODE code

When the client receives the code string from lircd it will send it to lircrcd and will receive back the applicable config string from the .lircrc config file. It should resend the CODE command until nothing is returned back which means that nothing (more) should happen in response to code. This command is used each time the lirc_code2char() function is called by a client.

GETMODE

lircrcd will return the current mode string.


Note for packagers


If you want to make a binary package for lirc (.deb, .rpm, ...), there are a couple of goodies you can take advantage of:

--with-driver=all
Builds support for as many devices as possible into a single lircd binary.
--with-driver=userspace
The same as the all driver, but no kernel modules will be built.
--enable-sandboxed
Prevents any changes out of the installation directory on "make install":
  • will not create device nodes on /dev/
  • will not run depmod
DESTDIR
You should use DESTDIR and not prefix to install into the sandbox, modules will be installed to /lib/ otherwise.

Finally, you should consider installing the doc/lirc.hwdb file. This one is a parseable list of LIRC supported devices, which is useful for LIRC configuration applications.
You should also have a look at the below HAL integration description.
For more information, refer to the next section.


Hardware Abstraction Layer (HAL) integration


If your system support the FreeDesktop Hardware Abstraction Layer (HAL), you will want to install the file contrib/hal/20-ircontrol-lirc.fdi in ${datarootdir}/hal/fdi/information/20thirdparty

This file can be generated at will using contrib/hal/gen-hal-fdi.pl.
Installing 20-ircontrol-lirc.fdi allow applications developers to detect USB remotes easily by searching for HAL devices with info.capabilities=input.remote_control

This is a work in progress. We will also provide more information in the future, such as the driver/manufacturer/device name(s), and possibly event reporting.


Note for configuration application developers


If you want to make a configuration application, lirc provides a parseable list of LIRC supported devices.

This file is generated at compilation time, along with its HTML equivalent and is available as doc/lirc.hwdb in the source tree. It should also be installed by the binary packages of your prefered distribution (bug report otherwise!).

The format is:
[remote controls type]
description;driver;lirc driver;HW_DEFAULT;lircd_conf;


Known bugs


  • If you use the lirc_serial or lirc_parallel driver regularly to transmit infra-red signals you might notice that your system clock will slow down. During transmit the driver turns off interrupts and hence some clock interrupts might get lost causing system clock inaccuracy. Unfortunately in order to ensure a good signal timing interrupts have to be disabled. Currently no work-around is known for this problem except using a program like netdate to synchronize your system clock regularly.

  • The lirc_serial and lirc_parallel drivers measure the time between interrupts on the serial resp. parallel port to get a pulse and space representation of the incoming infra-red signal. If interrupts are disabled by the CPU for a rather long time (>100 µs, which happens often e.g. during heavy IDE disk activity) some interrupts might get lost and the incoming data stream becomes disturbed. In this case decoding of the infra-red signal will fail. This is the downside of the really simple receiver circuits and can't be addressed in software except keeping the time where interrupts are disabled to a minimum.

    If you are using an IDE system you might want to try calling hdparm -u1 -d1 for all of your drives. This enables DMA for the drive and allows the driver to unmask other interrupts during handling of a disk interrupt. But be aware that this can be dangerous for some (buggy) IDE chipsets. Consult the hdparm man page for further information.



[LIRC homepage]
The LIRC Manual, last update: 10-June-2014