How to monitor an external process for events by i

Question

Is there any library which's got some function the allows one to monitor an external process for events by its pid_t? I mean, monitoring whether an external process has exited, or whether it has created one or more child processes (with fork), or whether it has become another executable image (via an exec or posix_spawn function family call) or whether a Unix signal was delivered to it.

EDIT

I need something that does not interfere with the execution of the program that is being monitored. So, I'm not supposed to use ptrace, since it stops the process which is being monitored when it emits some signal and it's necessary to resume the process whenever this happens.

Answer 1

There are few tools available which can collect information about a process while it is running.

I suggest you to use perf and systemTap.

https://perf.wiki.kernel.org/index.php/Main_Page

http://sourceware.org/systemtap/SystemTap_Beginners_Guide/index.html

Answer 2

If you can run as root, then you can use the netlink interface proc events:

http://bewareofgeek.livejournal.com/2945.html

I just compiled it cleanly on fedora 17 x86_64 and it gives me this:

[root@hip1 yotest]# ./proc
set mcast listen ok
fork: parent tid=2358 pid=2358 -> child tid=21007 pid=21007
exec: tid=21007 pid=21007
fork: parent tid=21007 pid=21007 -> child tid=21008 pid=21008
fork: parent tid=21007 pid=21007 -> child tid=21009 pid=21009
fork: parent tid=21007 pid=21007 -> child tid=21010 pid=21010
fork: parent tid=21007 pid=21007 -> child tid=21011 pid=21011
exec: tid=21010 pid=21010
exec: tid=21008 pid=21008
exec: tid=21011 pid=21011
exec: tid=21009 pid=21009
exit: tid=21008 pid=21008 exit_code=0
fork: parent tid=21010 pid=21010 -> child tid=21012 pid=21012
exit: tid=21009 pid=21009 exit_code=0
exec: tid=21012 pid=21012
exit: tid=21012 pid=21012 exit_code=0
exit: tid=21010 pid=21010 exit_code=0
exit: tid=21011 pid=21011 exit_code=0
exit: tid=21007 pid=21007 exit_code=0

You'll need to filter for the specific pids which are of interest to you, but you can easily do that in the switch statement on line 107.

For purposes of preservation:

#include <sys/socket.h>
#include <linux/netlink.h>
#include <linux/connector.h>
#include <linux/cn_proc.h>
#include <signal.h>
#include <errno.h>
#include <stdbool.h>
#include <unistd.h>
#include <string.h>
#include <stdlib.h>
#include <stdio.h>

/*
 * connect to netlink
 * returns netlink socket, or -1 on error
 */
static int nl_connect()
{
    int rc;
    int nl_sock;
    struct sockaddr_nl sa_nl;

    nl_sock = socket(PF_NETLINK, SOCK_DGRAM, NETLINK_CONNECTOR);
    if (nl_sock == -1) {
        perror("socket");
        return -1;
    }

    sa_nl.nl_family = AF_NETLINK;
    sa_nl.nl_groups = CN_IDX_PROC;
    sa_nl.nl_pid = getpid();

    rc = bind(nl_sock, (struct sockaddr *)&sa_nl, sizeof(sa_nl));
    if (rc == -1) {
        perror("bind");
        close(nl_sock);
        return -1;
    }

    return nl_sock;
}

/*
 * subscribe on proc events (process notifications)
 */
static int set_proc_ev_listen(int nl_sock, bool enable)
{
    int rc;
    struct __attribute__ ((aligned(NLMSG_ALIGNTO))) {
        struct nlmsghdr nl_hdr;
        struct __attribute__ ((__packed__)) {
            struct cn_msg cn_msg;
            enum proc_cn_mcast_op cn_mcast;
        };
    } nlcn_msg;

    memset(&nlcn_msg, 0, sizeof(nlcn_msg));
    nlcn_msg.nl_hdr.nlmsg_len = sizeof(nlcn_msg);
    nlcn_msg.nl_hdr.nlmsg_pid = getpid();
    nlcn_msg.nl_hdr.nlmsg_type = NLMSG_DONE;

    nlcn_msg.cn_msg.id.idx = CN_IDX_PROC;
    nlcn_msg.cn_msg.id.val = CN_VAL_PROC;
    nlcn_msg.cn_msg.len = sizeof(enum proc_cn_mcast_op);

    nlcn_msg.cn_mcast = enable ? PROC_CN_MCAST_LISTEN : PROC_CN_MCAST_IGNORE;

    rc = send(nl_sock, &nlcn_msg, sizeof(nlcn_msg), 0);
    if (rc == -1) {
        perror("netlink send");
        return -1;
    }

    return 0;
}

/*
 * handle a single process event
 */
static volatile bool need_exit = false;
static int handle_proc_ev(int nl_sock)
{
    int rc;
    struct __attribute__ ((aligned(NLMSG_ALIGNTO))) {
        struct nlmsghdr nl_hdr;
        struct __attribute__ ((__packed__)) {
            struct cn_msg cn_msg;
            struct proc_event proc_ev;
        };
    } nlcn_msg;

    while (!need_exit) {
        rc = recv(nl_sock, &nlcn_msg, sizeof(nlcn_msg), 0);
        if (rc == 0) {
            /* shutdown? */
            return 0;
        } else if (rc == -1) {
            if (errno == EINTR) continue;
            perror("netlink recv");
            return -1;
        }
        switch (nlcn_msg.proc_ev.what) {
            case PROC_EVENT_NONE:
                printf("set mcast listen ok\n");
                break;
            case PROC_EVENT_FORK:
                printf("fork: parent tid=%d pid=%d -> child tid=%d pid=%d\n",
                        nlcn_msg.proc_ev.event_data.fork.parent_pid,
                        nlcn_msg.proc_ev.event_data.fork.parent_tgid,
                        nlcn_msg.proc_ev.event_data.fork.child_pid,
                        nlcn_msg.proc_ev.event_data.fork.child_tgid);
                break;
            case PROC_EVENT_EXEC:
                printf("exec: tid=%d pid=%d\n",
                        nlcn_msg.proc_ev.event_data.exec.process_pid,
                        nlcn_msg.proc_ev.event_data.exec.process_tgid);
                break;
            case PROC_EVENT_UID:
                printf("uid change: tid=%d pid=%d from %d to %d\n",
                        nlcn_msg.proc_ev.event_data.id.process_pid,
                        nlcn_msg.proc_ev.event_data.id.process_tgid,
                        nlcn_msg.proc_ev.event_data.id.r.ruid,
                        nlcn_msg.proc_ev.event_data.id.e.euid);
                break;
            case PROC_EVENT_GID:
                printf("gid change: tid=%d pid=%d from %d to %d\n",
                        nlcn_msg.proc_ev.event_data.id.process_pid,
                        nlcn_msg.proc_ev.event_data.id.process_tgid,
                        nlcn_msg.proc_ev.event_data.id.r.rgid,
                        nlcn_msg.proc_ev.event_data.id.e.egid);
                break;
            case PROC_EVENT_EXIT:
                printf("exit: tid=%d pid=%d exit_code=%d\n",
                        nlcn_msg.proc_ev.event_data.exit.process_pid,
                        nlcn_msg.proc_ev.event_data.exit.process_tgid,
                        nlcn_msg.proc_ev.event_data.exit.exit_code);
                break;
            default:
                printf("unhandled proc event\n");
                break;
        }
    }

    return 0;
}

static void on_sigint(int unused)
{
    need_exit = true;
}

int main(int argc, const char *argv[])
{
    int nl_sock;
    int rc = EXIT_SUCCESS;

    signal(SIGINT, &on_sigint);
    siginterrupt(SIGINT, true);

    nl_sock = nl_connect();
    if (nl_sock == -1)
        exit(EXIT_FAILURE);

    rc = set_proc_ev_listen(nl_sock, true);
    if (rc == -1) {
        rc = EXIT_FAILURE;
        goto out;
    }

    rc = handle_proc_ev(nl_sock);
    if (rc == -1) {
        rc = EXIT_FAILURE;
        goto out;
    }

    set_proc_ev_listen(nl_sock, false);

out:
    close(nl_sock);
    exit(rc);
}

(gcc -o proc proc.c)

And some info on netlink:

excerpt: http://www.linuxjournal.com/article/7356

Netlink is asynchronous because, as with any other socket API, it provides a socket queue to smooth the burst of messages. The system call for sending a netlink message queues the message to the receiver's netlink queue and then invokes the receiver's reception handler. The receiver, within the reception handler's context, can decide whether to process the message immediately or leave the message in the queue and process it later in a different context. Unlike netlink, system calls require synchronous processing. Therefore, if we use a system call to pass a message from user space to the kernel, the kernel scheduling granularity may be affected if the time to process that message is long.

There's also this interesting announcement for nltrace made recently that you might find interesting as well! http://lists.infradead.org/pipermail/libnl/2013-April/000993.html

Answer 3

Use "pidof" system command from the procps library . Very simple and easy to use. If it returns something then the process is running or vice versa.

Answer 4

Run the target binary using a preload library that catches fork(). As long as all child processes also use the preload library, you'll see all local child processes, no matter how executed.

Here is an example implementation.

First, the forkmonitor.h header file. It defines the messages passed from the preload library, to the monitoring process:

#ifndef   FORKMONITOR_H
#define   FORKMONITOR_H

#define   FORKMONITOR_ENVNAME "FORKMONITOR_SOCKET"

#ifndef   UNIX_PATH_MAX
#define   UNIX_PATH_MAX 108
#endif

#define TYPE_EXEC       1   /* When a binary is executed */
#define TYPE_DONE       2   /* exit() or return from main() */
#define TYPE_FORK       3
#define TYPE_VFORK      4
#define TYPE_EXIT       5   /* _exit() or _Exit() */
#define TYPE_ABORT      6   /* abort() */

struct message {
    pid_t          pid;     /* Process ID */
    pid_t          ppid;    /* Parent process ID */
    pid_t          sid;     /* Session ID */
    pid_t          pgid;    /* Process group ID */
    uid_t          uid;     /* Real user ID */
    gid_t          gid;     /* Real group ID */
    uid_t          euid;    /* Effective user ID */
    gid_t          egid;    /* Effective group ID */
    unsigned short len;     /* Length of data[] */
    unsigned char  type;    /* One of the TYPE_ constants */
    char           data[0]; /* Optional payload, possibly longer */
};

#endif /* FORKMONITOR_H */

The FORKMONITOR_SOCKET environment variable (named by the FORKMONITOR_ENVNAME macro above) specifies the Unix domain datagram socket addess to the monitoring process. If not defined or empty, no monitoring messages are sent.

Here is the library itself, libforkmonitor.c. Note that I simplified the code quite a bit, leaving out multithreaded initialization (since it's rare for a library to call any of the intercepted functions, and even rarer to do it from multiple threads). It would be better to use atomic built-ins (__sync_bool_compare_and_swap()) to update the function pointer, and atomic getter (__sync_fetch_and_or(,0)) to retrieve the function pointer, to avoid any issues with wonky libraries. (This is quite safe for multithreaded programs, as the pointers will only be modified prior to main() is executed.)

#define  _POSIX_C_SOURCE 200809L
#define  _GNU_SOURCE
#include <stdlib.h>
#include <unistd.h>
#include <sys/types.h>
#include <sys/socket.h>
#include <sys/stat.h>
#include <sys/un.h>
#include <dlfcn.h>
#include <limits.h>
#include <string.h>
#include <stdio.h>
#include <errno.h>
#include "forkmonitor.h"

static pid_t (*actual_fork)(void)  = NULL;
static pid_t (*actual_vfork)(void) = NULL;
static void  (*actual_abort)(void) = NULL;
static void  (*actual__exit)(int)  = NULL;
static void  (*actual__Exit)(int)  = NULL;
static int     commfd = -1;

#define MINIMUM_COMMFD  31

static void notify(const int type, struct message *const msg, const size_t extra)
{
    const int    saved_errno = errno;

    msg->pid  = getpid();
    msg->ppid = getppid();
    msg->sid  = getsid(0);
    msg->pgid = getpgrp();
    msg->uid  = getuid();
    msg->gid  = getgid();
    msg->euid = geteuid();
    msg->egid = getegid();
    msg->len  = extra;
    msg->type = type;

    /* Since we don't have any method of dealing with send() errors
     * or partial send()s, we just fire one off and hope for the best. */
    send(commfd, msg, sizeof (struct message) + extra, MSG_EOR | MSG_NOSIGNAL);

    errno = saved_errno;
}

void libforkmonitor_init(void) __attribute__((constructor));
void libforkmonitor_init(void)
{
    const int saved_errno = errno;
    int       result;

    /* Save the actual fork() call pointer. */
    if (!actual_fork)
        *(void **)&actual_fork = dlsym(RTLD_NEXT, "fork");

    /* Save the actual vfork() call pointer. */
    if (!actual_vfork)
        *(void **)&actual_vfork = dlsym(RTLD_NEXT, "vfork");

    /* Save the actual abort() call pointer. */
    if (!actual_abort)
        *(void **)&actual_abort = dlsym(RTLD_NEXT, "abort");

    /* Save the actual _exit() call pointer. */
    if (!actual__exit)
        *(void **)&actual__exit = dlsym(RTLD_NEXT, "_exit");
    if (!actual__exit)
        *(void **)&actual__exit = dlsym(RTLD_NEXT, "_Exit");

    /* Save the actual abort() call pointer. */
    if (!actual__Exit)
        *(void **)&actual__Exit = dlsym(RTLD_NEXT, "_Exit");
    if (!actual__Exit)
        *(void **)&actual__Exit = dlsym(RTLD_NEXT, "_exit");

    /* Open an Unix domain datagram socket to the observer. */
    if (commfd == -1) {
        const char *address;

        /* Connect to where? */
        address = getenv(FORKMONITOR_ENVNAME);
        if (address && *address) {
            struct sockaddr_un addr;

            memset(&addr, 0, sizeof addr);
            addr.sun_family = AF_UNIX;
            strncpy(addr.sun_path, address, sizeof addr.sun_path - 1);

            /* Create and bind the socket. */
            commfd = socket(AF_UNIX, SOCK_DGRAM, 0);
            if (commfd != -1) {
                if (connect(commfd, (const struct sockaddr *)&addr, sizeof (addr)) == -1) {
                    /* Failed. Close the socket. */
                    do {
                        result = close(commfd);
                    } while (result == -1 && errno == EINTR);
                    commfd = -1;
                }
            }

            /* Move commfd to a high descriptor, to avoid complications. */
            if (commfd != -1 && commfd < MINIMUM_COMMFD) {
                const int newfd = MINIMUM_COMMFD;
                do {
                    result = dup2(commfd, newfd);
                } while (result == -1 && errno == EINTR);
                if (!result) {
                    do {
                        result = close(commfd);
                    } while (result == -1 && errno == EINTR);
                    commfd = newfd;
                }
            }
        }
    }

    /* Send an init message, listing the executable path. */
    if (commfd != -1) {
        size_t          len = 128;
        struct message *msg = NULL;

        while (1) {
            ssize_t n;

            free(msg);
            msg = malloc(sizeof (struct message) + len);
            if (!msg) {
                len = 0;
                break;
            }

            n = readlink("/proc/self/exe", msg->data, len);
            if (n > (ssize_t)0 && (size_t)n < len) {
                msg->data[n] = '\0';
                len = n + 1;
                break;
            }

            len = (3 * len) / 2;
            if (len >= 65536U) {
                free(msg);
                msg = NULL;
                len = 0;
                break;
            }
        }

        if (len > 0) {
            /* INIT message with executable name */
            notify(TYPE_EXEC, msg, len);
            free(msg);
        } else {
            /* INIT message without executable name */
            struct message msg2;
            notify(TYPE_EXEC, &msg2, sizeof msg2);
        }
    }

    /* Restore errno. */
    errno = saved_errno;
}

void libforkmonitor_done(void) __attribute__((destructor));
void libforkmonitor_done(void)
{
    const int saved_errno = errno;
    int       result;

    /* Send an exit message, no data. */
    if (commfd != -1) {
        struct message msg;
        notify(TYPE_DONE, &msg, sizeof msg);
    }

    /* If commfd is open, close it. */
    if (commfd != -1) {
        do {
            result = close(commfd);
        } while (result == -1 && errno == EINTR);
    }

    /* Restore errno. */
    errno = saved_errno;
}

/*
 * Hooked C library functions.
*/

pid_t fork(void)
{
    pid_t result;

    if (!actual_fork) {
        const int saved_errno = errno;

        *(void **)&actual_fork = dlsym(RTLD_NEXT, "fork");
        if (!actual_fork) {
            errno = EAGAIN;
            return (pid_t)-1;
        }

        errno = saved_errno;
    }

    result = actual_fork();
    if (!result && commfd != -1) {
        struct message msg;
        notify(TYPE_FORK, &msg, sizeof msg);
    }

    return result;
}

pid_t vfork(void)
{
    pid_t result;

    if (!actual_vfork) {
        const int saved_errno = errno;

        *(void **)&actual_vfork = dlsym(RTLD_NEXT, "vfork");
        if (!actual_vfork) {
            errno = EAGAIN;
            return (pid_t)-1;
        }

        errno = saved_errno;
    }

    result = actual_vfork();
    if (!result && commfd != -1) {
        struct message msg;
        notify(TYPE_VFORK, &msg, sizeof msg);
    }

    return result;
}

void _exit(const int code)
{
    if (!actual__exit) {
        const int saved_errno = errno;
        *(void **)&actual__exit = dlsym(RTLD_NEXT, "_exit");
        if (!actual__exit)
            *(void **)&actual__exit = dlsym(RTLD_NEXT, "_Exit");
        errno = saved_errno;
    }

    if (commfd != -1) {
        struct {
            struct message  msg;
            int             extra;
        } data;

        memcpy(&data.msg.data[0], &code, sizeof code);
        notify(TYPE_EXIT, &(data.msg), sizeof (struct message) + sizeof (int));
    }

    if (actual__exit)
        actual__exit(code);

    exit(code);
}

void _Exit(const int code)
{
    if (!actual__Exit) {
        const int saved_errno = errno;
        *(void **)&actual__Exit = dlsym(RTLD_NEXT, "_Exit");
        if (!actual__Exit)
            *(void **)&actual__Exit = dlsym(RTLD_NEXT, "_exit");
        errno = saved_errno;
    }

    if (commfd != -1) {
        struct {
            struct message  msg;
            int             extra;
        } data;

        memcpy(&data.msg.data[0], &code, sizeof code);
        notify(TYPE_EXIT, &(data.msg), sizeof (struct message) + sizeof (int));
    }

    if (actual__Exit)
        actual__Exit(code);

    exit(code);
}

void abort(void)
{
    if (!actual_abort) {
        const int saved_errno = errno;
        *(void **)&actual_abort = dlsym(RTLD_NEXT, "abort");
        errno = saved_errno;
    }

    if (commfd != -1) {
        struct message msg;
        notify(TYPE_ABORT, &msg, sizeof msg);
    }

    actual_abort();
    exit(127);
}

The libforkmonitor_init() function is called automatically by the runtime linker before the process main() is called, and libforkmonitor_done() is called when the process returns from main() or calls exit().

The libforkmonitor_init() opens an Unix domain datagram socket to the monitoring process, and sends its credentials and the path to the current executable. Every child process (as long as the preload library is still loaded) executes this after they're loaded, so there is no need to catch exec*() or posix_spawn*() or 'popen()` etc. functions at all.

The C library functions fork() and vfork() are intercepted. These interceptions are needed to catch the cases where the original program forks to create slave processes without executing any other binary. (At least GNU C library uses fork() internally, so these will catch popen(), posix_spawn(), etc. too.)

Additionally, C library functions _exit(), _Exit(), and abort() are intercepted too. I added these because some binaries, especially Dash, like to use _exit(), and I thought it would be nice to catch all forms of normal exits. (Death due to signals is not detected, however; and if a binary executes another binary, you'll only get the new EXEC message. Note the process and parent process ID's.)

Here is a simple monitoring program, forkmonitor.c:

#define  _POSIX_C_SOURCE 200809L
#include <unistd.h>
#include <stdlib.h>
#include <sys/socket.h>
#include <sys/un.h>
#include <signal.h>
#include <pwd.h>
#include <grp.h>
#include <string.h>
#include <stdio.h>
#include <errno.h>
#include "forkmonitor.h"

static volatile sig_atomic_t  done = 0;

static void done_handler(const int signum)
{
    if (!done)
        done = signum;
}

static int catch_done(const int signum)
{
    struct sigaction  act;

    sigemptyset(&act.sa_mask);
    act.sa_handler = done_handler;
    act.sa_flags = 0;

    if (sigaction(signum, &act, NULL) == -1)
        return errno;

    return 0;
}

static const char *username(const uid_t uid)
{
    static char    buffer[128];
    struct passwd *pw;

    pw = getpwuid(uid);
    if (!pw)
        return NULL;

    strncpy(buffer, pw->pw_name, sizeof buffer - 1);
    buffer[sizeof buffer - 1] = '\0';

    return (const char *)buffer;
}

static const char *groupname(const gid_t gid)
{
    static char   buffer[128];
    struct group *gr;

    gr = getgrgid(gid);
    if (!gr)
        return NULL;

    strncpy(buffer, gr->gr_name, sizeof buffer - 1);
    buffer[sizeof buffer - 1] = '\0';

    return (const char *)buffer;
}

int main(int argc, char *argv[])
{
    const size_t    msglen = 65536;
    struct message *msg;
    int             socketfd, result;
    const char     *user, *group;

    if (catch_done(SIGINT) || catch_done(SIGQUIT) || catch_done(SIGHUP) ||
        catch_done(SIGTERM) || catch_done(SIGPIPE)) {
        fprintf(stderr, "Cannot set signal handlers: %s.\n", strerror(errno));
        return 1;
    }

    if (argc != 2 || !strcmp(argv[1], "-h") || !strcmp(argv[1], "--help")) {
        fprintf(stderr, "\n");
        fprintf(stderr, "Usage: %s [ -h | --help ]\n", argv[0]);
        fprintf(stderr, "       %s MONITOR-SOCKET-PATH\n", argv[0]);
        fprintf(stderr, "\n");
        fprintf(stderr, "This program outputs events reported by libforkmonitor\n");
        fprintf(stderr, "to Unix domain datagram sockets at MONITOR-SOCKET-PATH.\n");
        fprintf(stderr, "\n");
        return 0;
    }

    msg = malloc(msglen);
    if (!msg) {
        fprintf(stderr, "Out of memory.\n");
        return 1;
    }

    socketfd = socket(AF_UNIX, SOCK_DGRAM, 0);
    if (socketfd == -1) {
        fprintf(stderr, "Cannot create an Unix domain datagram socket: %s.\n", strerror(errno));
        return 1;
    }

    {
        struct sockaddr_un  addr;
        size_t              len;

        if (argv[1])
            len = strlen(argv[1]);
        else
            len = 0;
        if (len < 1 || len >= UNIX_PATH_MAX) {
            fprintf(stderr, "%s: Path is too long (max. %d characters)\n", argv[1], UNIX_PATH_MAX - 1);
            return 1;
        }

        memset(&addr, 0, sizeof addr);
        addr.sun_family = AF_UNIX;
        memcpy(addr.sun_path, argv[1], len + 1); /* Include '\0' at end */

        if (bind(socketfd, (struct sockaddr *)&addr, sizeof (addr)) == -1) {
            fprintf(stderr, "Cannot bind to %s: %s.\n", argv[1], strerror(errno));
            return 1;
        }
    }

    printf("Waiting for connections.\n");
    printf("\n");

    /* Infinite loop. */
    while (!done) {
        ssize_t  n;

        n = recv(socketfd, msg, msglen, 0);
        if (n == -1) {
            const char *const errmsg = strerror(errno);
            fprintf(stderr, "%s.\n", errmsg);
            fflush(stderr);
            break;
        }

        if (msglen < sizeof (struct message)) {
            fprintf(stderr, "Received a partial message; discarded.\n");
            fflush(stderr);
            continue;
        }

        switch (msg->type) {
        case TYPE_EXEC:
            printf("Received an EXEC message:\n");
            break;
        case TYPE_DONE:
            printf("Received a DONE message:\n");
            break;
        case TYPE_FORK:
            printf("Received a FORK message:\n");
            break;
        case TYPE_VFORK:
            printf("Received a VFORK message:\n");
            break;
        case TYPE_EXIT:
            printf("Received an EXIT message:\n");
            break;
        case TYPE_ABORT:
            printf("Received an ABORT message:\n");
            break;
        default:
            printf("Received an UNKNOWN message:\n");
            break;
        }

        if (msg->type == TYPE_EXEC && (size_t)n > sizeof (struct message)) {
            if (*((char *)msg + n - 1) == '\0')
                printf("\tExecutable:        '%s'\n", (char *)msg + sizeof (struct message));
        }

        printf("\tProcess ID:         %d\n", (int)msg->pid);
        printf("\tParent process ID:  %d\n", (int)msg->ppid);
        printf("\tSession ID:         %d\n", (int)msg->sid);
        printf("\tProcess group ID:   %d\n", (int)msg->pgid);

        user = username(msg->uid);
        if (user)
            printf("\tReal user:         '%s' (%d)\n", user, (int)msg->uid);
        else
            printf("\tReal user:          %d\n", (int)msg->uid);

        group = groupname(msg->gid);
        if (group)
            printf("\tReal group:        '%s' (%d)\n", group, (int)msg->gid);
        else
            printf("\tReal group:         %d\n", (int)msg->gid);

        user = username(msg->euid);
        if (user)
            printf("\tEffective user:    '%s' (%d)\n", user, (int)msg->euid);
        else
            printf("\tEffective user:     %d\n", (int)msg->euid);

        group = groupname(msg->egid);
        if (group)
            printf("\tEffective group:   '%s' (%d)\n", group, (int)msg->egid);
        else
            printf("\tEffective group:    %d\n", (int)msg->egid);

        printf("\n");
        fflush(stdout);
    }

    do {
        result = close(socketfd);
    } while (result == -1 && errno == EINTR);

    unlink(argv[1]);

    return 0;
}

It takes a single command-line parameter, the Unix domain socket address. It should be an absolute file system path.

You can stop the monitoring program via INT (Ctrl+C), HUP, QUIT, and TERM signals.

Compile the library using

gcc -W -Wall -O3 -fpic -fPIC -c libforkmonitor.c
gcc -shared -Wl,-soname,libforkmonitor.so libforkmonitor.o -ldl -o libforkmonitor.so

and the monitor program using

gcc -W -Wall -O3 forkmonitor.c -o forkmonitor

In one terminal window, start the forkmonitor first:

./forkmonitor "$PWD/commsocket"

In another terminal window, in the same directory, run the monitored command, automatically preloading the libforkmonitor.so library and specifying the socket for the monitor:

env "LD_PRELOAD=$PWD/libforkmonitor.so" "FORKMONITOR_SOCKET=$PWD/commsocket" command args...

Note that because this uses the LD_PRELOAD and FORKMONITOR_SOCKET environment variables, child processes are ignored if their parent modifies the environment (removing the two environment variables), and when executing setuid or setgid binaries. This limitation can be avoided by eliminating the environment variables, and hardcoding them.

The run-time linker will not preload libraries for setuid or setgid binaries, unless the library is in one of the standard library directories, and also marked setgid.

Adding the library name to /etc/ld.so.preload will preload the library for all binaries, but you probably should add a mechanism into libforkmonitor_init() which limits the monitoring to desired binaries and/or a specified real user (as effective user changes when running a setuid binary).

For example, when I run

env "LD_PRELOAD=$PWD/libforkmonitor.so" "FORKMONITOR_SOCKET=$PWD/commsocket" sh -c 'date ; ls -laF'

the monitoring output is (anonymized):

Received an EXEC message:
Executable:        'bin/dash'
Process ID:         11403
Parent process ID:  9265
Session ID:         9265
Process group ID:   11403
Real user:         'username' (1000)
Real group:        'username' (1000)
Effective user:    'username' (1000)
Effective group:   'username' (1000)

Received a FORK message:
Process ID:         11404
Parent process ID:  11403
Session ID:         9265
Process group ID:   11403
Real user:         'username' (1000)
Real group:        'username' (1000)
Effective user:    'username' (1000)
Effective group:   'username' (1000)

Received an EXEC message:
Executable:        'bin/date'
Process ID:         11404
Parent process ID:  11403
Session ID:         9265
Process group ID:   11403
Real user:         'username' (1000)
Real group:        'username' (1000)
Effective user:    'username' (1000)
Effective group:   'username' (1000)

Received a DONE message:
Process ID:         11404
Parent process ID:  11403
Session ID:         9265
Process group ID:   11403
Real user:         'username' (1000)
Real group:        'username' (1000)
Effective user:    'username' (1000)
Effective group:   'username' (1000)

Received a FORK message:
Process ID:         11405
Parent process ID:  11403
Session ID:         9265
Process group ID:   11403
Real user:         'username' (1000)
Real group:        'username' (1000)
Effective user:    'username' (1000)
Effective group:   'username' (1000)

Received an EXEC message:
Executable:        'bin/ls'
Process ID:         11405
Parent process ID:  11403
Session ID:         9265
Process group ID:   11403
Real user:         'username' (1000)
Real group:        'username' (1000)
Effective user:    'username' (1000)
Effective group:   'username' (1000)

Received a DONE message:
Process ID:         11405
Parent process ID:  11403
Session ID:         9265
Process group ID:   11403
Real user:         'username' (1000)
Real group:        'username' (1000)
Effective user:    'username' (1000)
Effective group:   'username' (1000)

Received an EXIT message:
Process ID:         11403
Parent process ID:  9265
Session ID:         9265
Process group ID:   11403
Real user:         'username' (1000)
Real group:        'username' (1000)
Effective user:    'username' (1000)
Effective group:   'username' (1000)

This is a very lightweight process tree monitoring solution. Other than process startup, exit, and calling one of the intercepted functions (fork(), vfork(), _exit(), _Exit(), abort()), program execution will not be affected at all. Because the library is so lightweight, even those affected will only be affected by a very, very small amount; probably not enough to measure reliably.

It is obviously possible to intercept other functions, and/or use two-way communication, "pausing" the execution of the intercepted function until the monitoring application responds.

There are some pitfalls overall, especially related to setuid/setgid processes, and processes that generate a new environment (omitting the LD_PRELOAD and FORKMONITOR_SOCKET environment variables), but those can be worked around if superuser privileges are available.

Hope you find this informative. Questions?