Treat all devices at the container level as spares

[thirdparty/mdadm.git] / mdmon.c

diff --git a/mdmon.c b/mdmon.c
index 85f44bc27f0355000e6c1be614e7c4ba30bdf2bc..28b01398f715ed55c7675f51594747f8c3c50fc0 100644 (file)
--- a/mdmon.c
+++ b/mdmon.c
@@ -80,19 +80,23 @@ static struct superswitch *find_metadata_methods(char *vers)
 }
 
 
-static int make_pidfile(char *devname, int o_excl)
+int make_pidfile(char *devname, int o_excl)
 {
        char path[100];
        char pid[10];
        int fd;
+       int n;
+
        sprintf(path, "/var/run/mdadm/%s.pid", devname);
 
        fd = open(path, O_RDWR|O_CREAT|o_excl, 0600);
        if (fd < 0)
-               return -1;
+               return -errno;
        sprintf(pid, "%d\n", getpid());
-       write(fd, pid, strlen(pid));
+       n = write(fd, pid, strlen(pid));
        close(fd);
+       if (n < 0)
+               return -errno;
        return 0;
 }
 
@@ -136,9 +140,11 @@ void remove_pidfile(char *devname)
 
        sprintf(buf, "/var/run/mdadm/%s.pid", devname);
        unlink(buf);
+       sprintf(buf, "/var/run/mdadm/%s.sock", devname);
+       unlink(buf);
 }
 
-static int make_control_sock(char *devname)
+int make_control_sock(char *devname)
 {
        char path[100];
        int sfd;
@@ -164,6 +170,12 @@ static int make_control_sock(char *devname)
        return sfd;
 }
 
+int socket_hup_requested;
+static void hup(int sig)
+{
+       socket_hup_requested = 1;
+}
+
 static void wake_me(int sig)
 {
 
@@ -191,6 +203,7 @@ int main(int argc, char *argv[])
        struct sigaction act;
        int pfd[2];
        int status;
+       int ignore;
 
        if (argc != 2) {
                fprintf(stderr, "Usage: md-manage /device/name/for/container\n");
@@ -210,7 +223,10 @@ int main(int argc, char *argv[])
 
        /* Fork, and have the child tell us when they are ready */
        if (do_fork()) {
-               pipe(pfd);
+               if (pipe(pfd) != 0) {
+                       fprintf(stderr, "mdmon: failed to create pipe\n");
+                       exit(1);
+               }
                switch(fork()) {
                case -1:
                        fprintf(stderr, "mdmon: failed to fork: %s\n",
@@ -245,21 +261,27 @@ int main(int argc, char *argv[])
                                container->devname);
                        exit(3);
                } else {
+                       int err;
+
                        /* cleanup the old monitor, this one is taking over */
                        try_kill_monitor(container->devname);
-                       if (make_pidfile(container->devname, 0) < 0) {
+                       err = make_pidfile(container->devname, 0);
+                       if (err < 0) {
                                fprintf(stderr, "mdmon: %s Cannot create pidfile\n",
                                        container->devname);
-                               exit(3);
+                               if (err == -EROFS) {
+                                       /* FIXME implement a mechanism to
+                                        * prevent duplicate monitor instances
+                                        */
+                                       fprintf(stderr,
+                                               "mdmon: continuing on read-only file system\n");
+                               } else
+                                       exit(3);
                        }
                }
        }
 
        container->sock = make_control_sock(container->devname);
-       if (container->sock < 0) {
-               fprintf(stderr, "mdmon: Cannot create socket in /var/run/mdadm\n");
-               exit(3);
-       }
        container->arrays = NULL;
 
        mdi = sysfs_read(mdfd, container->devnum,
@@ -292,7 +314,7 @@ int main(int argc, char *argv[])
        container->devs = NULL;
        for (di = mdi->devs; di; di = di->next) {
                struct mdinfo *cd = malloc(sizeof(*cd));
-               cd = di;
+               *cd = *di;
                cd->next = container->devs;
                container->devs = cd;
        }
@@ -308,18 +330,20 @@ int main(int argc, char *argv[])
        /* Ok, this is close enough.  We can say goodbye to our parent now.
         */
        status = 0;
-       write(pfd[1], &status, sizeof(status));
+       if (write(pfd[1], &status, sizeof(status)) < 0)
+               fprintf(stderr, "mdmon: failed to notify our parent: %d\n",
+                       getppid());
        close(pfd[1]);
 
-       chdir("/");
+       ignore = chdir("/");
        setsid();
        close(0);
        open("/dev/null", O_RDWR);
        close(1);
-       dup(0);
+       ignore = dup(0);
 #ifndef DEBUG
        close(2);
-       dup(0);
+       ignore = dup(0);
 #endif
 
        mlockall(MCL_FUTURE);
@@ -329,15 +353,20 @@ int main(int argc, char *argv[])
         */
        sigemptyset(&set);
        sigaddset(&set, SIGUSR1);
+       sigaddset(&set, SIGHUP);
+       sigaddset(&set, SIGALRM);
        sigprocmask(SIG_BLOCK, &set, NULL);
        act.sa_handler = wake_me;
        act.sa_flags = 0;
        sigaction(SIGUSR1, &act, NULL);
+       sigaction(SIGALRM, &act, NULL);
+       act.sa_handler = hup;
+       sigaction(SIGHUP, &act, NULL);
        act.sa_handler = SIG_IGN;
        sigaction(SIGPIPE, &act, NULL);
 
        if (clone_monitor(container) < 0) {
-               fprintf(stderr, "md-manage: failed to start monitor process: %s\n",
+               fprintf(stderr, "mdmon: failed to start monitor process: %s\n",
                        strerror(errno));
                exit(2);
        }