start fleshing out socket code, ping monitor to see if it is alive

From: Dan Williams <dan.j.williams@intel.com>

Signed-off-by: Dan Williams <dan.j.williams@intel.com>

Makefile

@@ -79,7 +79,7 @@ SRCS =  mdadm.c config.c mdstat.c  ReadMe.c util.c Manage.c Assemble.c Build.c \
 
 MON_OBJS = mdmon.o monitor.o managemon.o util.o mdstat.o sysfs.o config.o \
 	Kill.o sg_io.o dlink.o ReadMe.o super0.o super1.o super-intel.o \
-	super-ddf.o sha1.o crc32.o
+	super-ddf.o sha1.o crc32.o msg.o
 
 
 STATICSRC = pwgr.c
@@ -125,6 +125,7 @@ mdadm.O2 : $(SRCS) mdadm.h
 
 mdmon : $(MON_OBJS)
 	$(CC) $(LDFLAGS) -o mdmon $(MON_OBJS) $(LDLIBS)
+msg.o: msg.c msg.h
 
 test_stripe : restripe.c mdadm.h
 	$(CC) $(CXFLAGS) $(LDFLAGS) -o test_stripe -DMAIN restripe.c

managemon.c

@@ -74,6 +74,7 @@
 #endif
 #include	"mdadm.h"
 #include	"mdmon.h"
+#include	"msg.h"
 #include	<sys/socket.h>
 
 
@@ -302,12 +303,35 @@ void manage(struct mdstat_ent *mdstat, struct active_array *aa,
 void read_sock(int pfd)
 {
 	int fd;
+	struct md_message msg;
+	int terminate = 0;
+	long fl;
+	int tmo = 3; /* 3 second timeout before hanging up the socket */
 
-	// FIXME set non-blocking
 	fd = accept(pfd, NULL, NULL);
 	if (fd < 0)
 		return;
-	// FIXME do something useful
+
+	fl = fcntl(fd, F_GETFL, 0);
+	fl |= O_NONBLOCK;
+	fcntl(fd, F_SETFL, fl);
+
+	do {
+		msg.buf = NULL;
+
+		/* read and validate the message */
+		if (receive_message(fd, &msg, tmo) == 0) {
+			// FIXME: handle message contents
+			ack(fd, msg.seq, tmo);
+		} else {
+			terminate = 1;
+			nack(fd, -1, tmo);
+		}
+
+		if (msg.buf)
+			free(msg.buf);
+	} while (!terminate);
+
 	close(fd);
 }
 void do_manager(struct supertype *container)

mdmon.c

@@ -34,6 +34,7 @@
 #include	<errno.h>
 #include	<string.h>
 #include	<fcntl.h>
+#include	<signal.h>
 
 #include	<sched.h>
 
@@ -76,14 +77,14 @@ static struct superswitch *find_metadata_methods(char *vers)
 }
 
 
-static int make_pidfile(char *devname)
+static int make_pidfile(char *devname, int o_excl)
 {
 	char path[100];
 	char pid[10];
 	int fd;
 	sprintf(path, "/var/run/mdadm/%s.pid", devname);
 
-	fd = open(path, O_RDWR|O_CREAT|O_EXCL, 0600);
+	fd = open(path, O_RDWR|O_CREAT|o_excl, 0600);
 	if (fd < 0)
 		return -1;
 	sprintf(pid, "%d\n", getpid());
@@ -92,6 +93,40 @@ static int make_pidfile(char *devname)
 	return 0;
 }
 
+static void try_kill_monitor(char *devname)
+{
+	char buf[100];
+	int fd;
+	pid_t pid;
+
+	sprintf(buf, "/var/run/mdadm/%s.pid", devname);
+	fd = open(buf, O_RDONLY);
+	if (fd < 0)
+		return;
+
+	if (read(fd, buf, sizeof(buf)) < 0) {
+		close(fd);
+		return;
+	}
+
+	close(fd);
+	pid = strtoul(buf, NULL, 10);
+
+	/* kill this process if it is mdmon */
+	sprintf(buf, "/proc/%lu/cmdline", (unsigned long) pid);
+	fd = open(buf, O_RDONLY);
+	if (fd < 0)
+		return;
+
+	if (read(fd, buf, sizeof(buf)) < 0) {
+		close(fd);
+		return;
+	}
+
+	if (strstr(buf, "mdmon") != NULL)
+		kill(pid, SIGTERM);
+}
+
 static int make_control_sock(char *devname)
 {
 	char path[100];
@@ -150,10 +185,20 @@ int main(int argc, char *argv[])
 	/* If this fails, we hope it already exists */
 	mkdir("/var/run/mdadm", 0600);
 	/* pid file lives in /var/run/mdadm/mdXX.pid */
-	if (make_pidfile(container->devname) < 0) {
-		fprintf(stderr, "md-manage: %s already managed\n",
-			container->devname);
-		exit(3);
+	if (make_pidfile(container->devname, O_EXCL) < 0) {
+		if (ping_monitor(container->devname) == 0) {
+			fprintf(stderr, "mdmon: %s already managed\n",
+				container->devname);
+			exit(3);
+		} else {
+			/* cleanup the old monitor, this one is taking over */
+			try_kill_monitor(container->devname);
+			if (make_pidfile(container->devname, 0) < 0) {
+				fprintf(stderr, "mdmon: %s Cannot create pidfile\n",
+					container->devname);
+				exit(3);
+			}
+		}
 	}
 
 	container->sock = make_control_sock(container->devname);

msg.h

@@ -17,6 +17,9 @@
  * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
  */
 
+#ifndef _MSG_H
+#define _MSG_H
+
 struct mdinfo;
 struct md_message {
 	int seq;
@@ -57,3 +60,4 @@ extern int connect_monitor(char *devname);
 extern int ping_monitor(char *devname);
 extern int send_remove_device(int fd, dev_t rdev, int seq, int tmo);
 
+#endif