Ask to be notified of system-wide events


#include <sys/procmgr.h>

int procmgr_event_notify (
              unsigned flags,
              const struct sigevent * event );


A bitwise OR of the type of events that you want to be notified of, or 0 to unarm the sigevent. The event types include:

For more information, see " Event types," below.

A pointer to a sigevent structure that specifies how you want to be notified.



Use the -l c option to qcc to link against this library. This library is usually included automatically.


The procmgr_event_notify() function requests that the process manager notify the caller of the system-wide events identified by the given flags.

This function lets you add a single notification request for the process. If you call it again, the new request replaces the previous one. If you want to your process to have more than one notification request, use procmgr_event_notify_add() and procmgr_event_notify_delete() instead.

To remove the current notification request, pass a value of 0 for flags.

Event types

The following event types are defined in <sys/procmgr.h>:

A process set a configuration string.
A process in session 1 died. This event is most useful for watching for the death of daemon processes that use procmgr_daemon() to put themselves in session 1 as well as close and redirect file descriptors. As a result of this closing and redirecting, the death of daemons is difficult to detect otherwise.

Notification is via the given event, so no information is provided as to which process died. Once you've received the event, you'll need to do something else to find out if processes you care about had died. You can do this by walking through the list of all processes, looking for specific process IDs or process names. If you don't find one, then it has died. The sample code below demonstrates how you can do this.

A resource manager added or removed an entry (i.e. mountpoint) to or from the pathname space. This is generally associated with resource manager calls to resmgr_attach() and resmgr_detach() . Terminating a resource manager process also generates this event if the mountpoints haven't been detached.
A process called sync() to synchronize the filesystems.
A process set a system configuration string.
A process changed the time of day by calling ClockTime() or clock_settime() .


-1 on error; any other value indicates success.


 * This demonstrates procmgr_event_notify() with the 
 * PROCMGR_DAEMON_DEATH flag. This flag allows you to 
 * be notified if any process in session 1 dies.  
 * Daemons are processes that do things that make
 * their death hard to detect (they become daemons by calling
 * procmgr_daemon()).  One of the things that happens is that
 * daemons end up in session 1.  Hence, the usefulness of the
 * When you are notified, you're not told who died.
 * It's up to you to know who should be running.  Once notified,
 * you could then walk through the list of which processes are
 * still running and see if all the expected processes are still
 * running. If you know the process id of the processes you
 * are watching out for then this is easiest.  If you don't know
 * the process id then your next option may be by process name.
 * The code below does a lookup by process name.

#include <devctl.h>
#include <dirent.h>
#include <errno.h>
#include <fcntl.h>
#include <libgen.h>
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <sys/neutrino.h>
#include <sys/procfs.h>
#include <sys/procmgr.h>

static int check_if_running( char *process );


struct dinfo_s {
    procfs_debuginfo    info;
    char                pathbuffer[PATH_MAX];

main( int argc, char **argv )
    char            *daemon_to_watch;
    int             chid, coid, rcvid;
    struct sigevent event;
    struct _pulse   msg;
    if (argc != 2) {
        printf( "use: %s process_to_watch_for\n", argv[0] );
        exit( EXIT_FAILURE );
    daemon_to_watch = argv[1];   /* the process to watch for */
    chid = ChannelCreate( 0 );
    coid = ConnectAttach( 0, 0, chid, _NTO_SIDE_CHANNEL, 0 );
                      DAEMON_DIED_CODE, 0 );

     * Ask to be notified via a pulse whenever a 
     * daemon process dies
    if (procmgr_event_notify( PROCMGR_EVENT_DAEMON_DEATH, 
                              &event ) == -1) {
        fprintf( stderr, "procmgr_event_notify() failed" );
        exit( EXIT_FAILURE );

    while (1) {
        rcvid = MsgReceive( chid, &msg, sizeof(msg), NULL );
        if (rcvid != 0) {
            /* not a pulse; could be an unexpected message or
               error */
            exit( EXIT_FAILURE );
        if (check_if_running( daemon_to_watch ) == 0)
            printf( "%s is no longer running\n", daemon_to_watch);
    return 0;

 * check_if_running - This will walk through all processes 
 * to see if this particular one is still running.
static int
check_if_running( char *process )
    DIR             *dirp;
    struct dirent   *dire;
    char            buffer[20];
    int             fd, status;
    pid_t           pid;
    struct dinfo_s  dinfo;
    if ((dirp = opendir( "/proc" )) == NULL) {
        perror( "Could not open '/proc'" );
        return -1;
    while (1) {
        if ((dire = readdir( dirp )) == NULL)
        if (isdigit( dire->d_name[0] )) {
            pid = strtoul( dire->d_name, NULL, 0 );
            sprintf( buffer, "/proc/%d/as", pid );
            if ((fd = open( buffer, O_RDONLY )) != NULL) {
                status = devctl( fd, DCMD_PROC_MAPDEBUG_BASE,
                                 &dinfo, sizeof(dinfo), NULL );
                if (status == EOK) {
                    if (!strcmp( process,
                                 basename( ) ))
                        closedir (dirp);

                        /* You should close fd to prevent memory
                           leaking */
                        return 1;
                } /* else some errors are expected, e.g. procnto
                     has no MAPDEBUG info and there is a timing
                     issue  with getting info on the process
                     that died, ignore errors */
                close( fd );
    closedir( dirp );
    return 0;


QNX Neutrino

Cancellation point No
Interrupt handler No
Signal handler Yes
Thread Yes

Last modified: 2013-12-23

comments powered by Disqus