Merge branch 'master' into devel-3.1

Assemble.c

@@ -315,6 +315,9 @@ int Assemble(struct supertype *st, char *mddev,
 			}
 			/* It is worth looking inside this container.
 			 */
+			if (verbose > 0)
+				fprintf(stderr, Name ": looking in container %s\n",
+					devname);
 		next_member:
 			if (tmpdev->content)
 				content = tmpdev->content;
@@ -405,6 +408,9 @@ int Assemble(struct supertype *st, char *mddev,
 					fprintf(stderr, Name ": member %s in %s is already assembled\n",
 						content->text_version,
 						devname);
+			skip:
+				if (tmpdev->content)
+					goto next_member;
 				tst->ss->free_super(tst);
 				tst = NULL;
 				content = NULL;
@@ -412,6 +418,21 @@ int Assemble(struct supertype *st, char *mddev,
 					goto loop;
 				return 1;
 			}
+			if (ident->member && ident->member[0]) {
+				char *s = strchr(content->text_version+1, '/');
+				if (s == NULL) {
+					fprintf(stderr, Name ": badly formatted version: %s\n",
+						content->text_version);
+					goto skip;
+				}
+				if (strcmp(ident->member, s+1) != 0) {
+					if (report_missmatch)
+						fprintf(stderr,
+							Name ": skipping wrong member %s\n",
+							content->text_version);
+					goto skip;
+				}
+			}
 			st = tst; tst = NULL;
 			if (!auto_assem && tmpdev->next != NULL) {
 				fprintf(stderr, Name ": %s is a container, but is not "
@@ -420,6 +441,9 @@ int Assemble(struct supertype *st, char *mddev,
 				st->ss->free_super(st);
 				return 1;
 			}
+			if (verbose > 0)
+				fprintf(stderr, Name ": found match on member %s in %s\n",
+					content->text_version, devname);
 			break;
 		}
 		if (st == NULL)
@@ -565,6 +589,7 @@ int Assemble(struct supertype *st, char *mddev,
 #endif
 	/* Ok, no bad inconsistancy, we can try updating etc */
 	bitmap_done = 0;
+	content->update_private = NULL;
 	for (tmpdev = devlist; tmpdev; tmpdev=tmpdev->next) if (tmpdev->used == 1) {
 		char *devname = tmpdev->devname;
 		struct stat stb;
@@ -717,6 +742,8 @@ int Assemble(struct supertype *st, char *mddev,
 		}
 		devcnt++;
 	}
+	free(content->update_private);
+	content->update_private = NULL;
 
 	if (devcnt == 0) {
 		fprintf(stderr, Name ": no devices found for %s\n",

Detail.c

@@ -194,7 +194,12 @@ int Detail(char *dev, int brief, int export, int test, char *homehost)
 				st->ss->export_detail_super(st);
 		} else {
 			struct map_ent *mp, *map = NULL;
+			char nbuf[64];
 			mp = map_by_devnum(&map, fd2devnum(fd));
+			if (mp) {
+				__fname_from_uuid(mp->uuid, 0, nbuf, ':');
+				printf("MD_UUID=%s\n", nbuf+5);
+			}
 			if (mp && mp->path &&
 			    strncmp(mp->path, "/dev/md/", 8) == 0)
 				printf("MD_DEVNAME=%s\n", mp->path+8);
@@ -540,6 +545,7 @@ This is pretty boring
 		    1, avail, avail_disks))
 		rv = 2;
 
+	free(disks);
 out:
 	close(fd);
 	return rv;

Manage.c

@@ -140,7 +140,7 @@ static void remove_devices(int devnum, char *path)
 		strcpy(path2, path);
 		pe = path2 + strlen(path2);
 	} else
-		path = NULL;
+		path2 = path = NULL;
 	
 	for (part = 0; part < 16; part++) {
 		if (part) {
@@ -161,6 +161,7 @@ static void remove_devices(int devnum, char *path)
 				unlink(path2);
 		}
 	}
+	free(path2);
 }
 	
 
@@ -649,6 +650,7 @@ int Manage_subdevs(char *devname, int fd,
 						disc.state |= (1<<MD_DISK_SYNC);
 						break;
 					}
+				free(used);
 			}
 			if (dv->writemostly == 1)
 				disc.state |= (1 << MD_DISK_WRITEMOSTLY);

Monitor.c

@@ -33,14 +33,6 @@
 static void alert(char *event, char *dev, char *disc, char *mailaddr, char *mailfrom,
 		  char *cmd, int dosyslog);
 
-static char *percentalerts[] = {
-	"RebuildStarted",
-	"Rebuild20",
-	"Rebuild40",
-	"Rebuild60",
-	"Rebuild80",
-};
-
 /* The largest number of disks current arrays can manage is 384
  * This really should be dynamically, but that will have to wait
  * At least it isn't MD_SB_DISKS.
@@ -49,7 +41,7 @@ static char *percentalerts[] = {
 int Monitor(mddev_dev_t devlist,
 	    char *mailaddr, char *alert_cmd,
 	    int period, int daemonise, int scan, int oneshot,
-	    int dosyslog, int test, char* pidfile)
+	    int dosyslog, int test, char* pidfile, int increments)
 {
 	/*
 	 * Every few seconds, scan every md device looking for changes
@@ -77,8 +69,8 @@ int Monitor(mddev_dev_t devlist,
 	 *      An active device had a reverse transition
 	 *    RebuildStarted
 	 *      percent went from -1 to +ve
-	 *    Rebuild20 Rebuild40 Rebuild60 Rebuild80
+	 *    RebuildNN
-	 *      percent went from below to not-below that number
+	 *      percent went from below to not-below NN%
 	 *    DeviceDisappeared
 	 *      Couldn't access a device which was previously visible
 	 *
@@ -311,9 +303,17 @@ int Monitor(mddev_dev_t devlist,
 			if (mse &&
 			    st->percent >= 0 &&
 			    mse->percent >= 0 &&
-			    (mse->percent / 20) > (st->percent / 20))
+			    (mse->percent / increments) > (st->percent / increments)) {
-				alert(percentalerts[mse->percent/20],
+				char percentalert[15]; // "RebuildNN" (10 chars) or "RebuildStarted" (15 chars)
+
+				if((mse->percent / increments) == 0)
+					snprintf(percentalert, sizeof(percentalert), "RebuildStarted");
+				else
+					snprintf(percentalert, sizeof(percentalert), "Rebuild%02d", mse->percent);
+
+				alert(percentalert,
 				      dev, NULL, mailaddr, mailfrom, alert_cmd, dosyslog);
+			}
 
 			if (mse &&
 			    mse->percent == -1 &&

ReadMe.c

@@ -176,6 +176,7 @@ struct option long_options[] = {
     {"mail",      1, 0, 'm'},
     {"program",   1, 0, 'p'},
     {"alert",     1, 0, 'p'},
+    {"increment", 1, 0, 'r'},
     {"delay",     1, 0, 'd'},
     {"daemonise", 0, 0, 'f'},
     {"daemonize", 0, 0, 'f'},
@@ -496,6 +497,7 @@ char Help_monitor[] =
 "  --mail=       -m   : Address to mail alerts of failure to\n"
 "  --program=    -p   : Program to run when an event is detected\n"
 "  --alert=           : same as --program\n"
+"  --increment=  -r   : Report RebuildNN events in the given increment. default=20\n"
 "  --delay=      -d   : seconds of delay between polling state. default=60\n"
 "  --config=     -c   : specify a different config file\n"
 "  --scan        -s   : find mail-address/program in config file\n"

managemon.c

@@ -680,6 +680,12 @@ void do_manager(struct supertype *container)
 			read_sock(container);
 
 			if (container->sock < 0 || socket_hup_requested) {
+				/* If this fails, we hope it already exists
+				 * pid file lives in /var/run/mdadm/mdXX.pid
+				 */
+				mkdir("/var", 0600);
+				mkdir("/var/run", 0600);
+				mkdir("/var/run/mdadm", 0600);
 				close(container->sock);
 				container->sock = make_control_sock(container->devname);
 				make_pidfile(container->devname, 0);

mdadm.8

@@ -309,7 +309,7 @@ says to get a list of array devices from
 .BR /proc/mdstat .
 
 .TP
-.B \-e ", " \-\-metadata=
+.BR \-e ", " \-\-metadata=
 Declare the style of RAID metadata (superblock) to be used.  The
 default is 0.90 for
 .BR \-\-create ,
@@ -1261,6 +1261,12 @@ reduce this as the kernel alerts
 .I mdadm
 immediately when there is any change.
 
+.TP
+.BR \-r ", " \-\-increment
+Give a percentage increment.
+.I mdadm
+will generate RebuildNN events with the given percentage increment.
+
 .TP
 .BR \-f ", " \-\-daemonise
 Tell
@@ -1859,8 +1865,10 @@ An md array started reconstruction. (syslog priority: Warning)
 .BI Rebuild NN
 Where
 .I NN
-is 20, 40, 60, or 80, this indicates that rebuild has passed that many
+is a two-digit number (ie. 05, 48). This indicates that rebuild
-percentage of the total. (syslog priority: Warning)
+has passed that many percent of the total. The events are generated
+with fixed increment since 0. Increment size may be specified with
+a commandline option (default is 20). (syslog priority: Warning)
 
 .TP
 .B RebuildFinished

mdadm.c

@@ -91,6 +91,7 @@ int main(int argc, char *argv[])
 	int require_homehost = 1;
 	char *mailaddr = NULL;
 	char *program = NULL;
+	int increments = 20;
 	int delay = 0;
 	int daemonise = 0;
 	char *pidfile = NULL;
@@ -714,6 +715,14 @@ int main(int argc, char *argv[])
 				program = optarg;
 			continue;
 
+		case O(MONITOR,'r'): /* rebuild increments */
+			increments = atoi(optarg);
+			if (increments>99 || increments<1) {
+				fprintf(stderr, Name ": please specify positive integer between 1 and 99 as rebuild increments.\n");
+				exit(2);
+			}
+			continue;
+
 		case O(MONITOR,'d'): /* delay in seconds */
 		case O(GROW, 'd'):
 		case O(BUILD,'d'): /* delay for bitmap updates */
@@ -1270,11 +1279,18 @@ int main(int argc, char *argv[])
 					struct mdstat_ent *ms = mdstat_read(0, 1);
 					struct mdstat_ent *e;
 					struct map_ent *map = NULL;
+					int members;
 					int v = verbose>1?0:verbose+1;
 
+					for (members = 0; members <= 1; members++) {
 					for (e=ms ; e ; e=e->next) {
 						char *name;
 						struct map_ent *me;
+						int member = e->metadata_version &&
+							strncmp(e->metadata_version,
+								"external:/", 10) == 0;
+						if (members != member)
+							continue;
 						me = map_by_devnum(&map, e->devnum);
 						if (me && me->path
 						    && strcmp(me->path, "/unknown") != 0)
@@ -1292,9 +1308,10 @@ int main(int argc, char *argv[])
 								     export, test,
 								     homehost);
 						else
-							rv |= WaitClean(name, v);
+							rv |= WaitClean(name, -1, v);
 						put_md_name(name);
 					}
+					}
 					free_mdstat(ms);
 				} else	if (devmode == 'S' && scan) {
 					/* apply --stop to all devices in /proc/mdstat */
@@ -1353,7 +1370,7 @@ int main(int argc, char *argv[])
 				case 'W':
 					rv |= Wait(dv->devname); continue;
 				case Waitclean:
-					rv |= WaitClean(dv->devname, verbose-quiet); continue;
+					rv |= WaitClean(dv->devname, -1, verbose-quiet); continue;
 				}
 				mdfd = open_mddev(dv->devname, 1);
 				if (mdfd>=0) {
@@ -1393,7 +1410,7 @@ int main(int argc, char *argv[])
 		}
 		rv= Monitor(devlist, mailaddr, program,
 			    delay?delay:60, daemonise, scan, oneshot,
-			    dosyslog, test, pidfile);
+			    dosyslog, test, pidfile, increments);
 		break;
 
 	case GROW:

mdadm.h

@@ -153,6 +153,11 @@ struct mdinfo {
 	int			cache_size; /* size of raid456 stripe cache*/
 	int			mismatch_cnt;
 	char			text_version[50];
+	void 			*update_private; /* for passing metadata-format
+						  * specific update data
+						  * between successive calls to
+						  * update_super()
+						  */
 
 	int container_member; /* for assembling external-metatdata arrays
 			       * This is to be used internally by metadata
@@ -749,11 +754,11 @@ extern int Examine(mddev_dev_t devlist, int brief, int export, int scan,
 extern int Monitor(mddev_dev_t devlist,
 		   char *mailaddr, char *alert_cmd,
 		   int period, int daemonise, int scan, int oneshot,
-		   int dosyslog, int test, char *pidfile);
+		   int dosyslog, int test, char *pidfile, int increments);
 
 extern int Kill(char *dev, int force, int quiet, int noexcl);
 extern int Wait(char *dev);
-extern int WaitClean(char *dev, int verbose);
+extern int WaitClean(char *dev, int sock, int verbose);
 
 extern int Incremental(char *devname, int verbose, int runstop,
 		       struct supertype *st, char *homehost, int require_homehost,
@@ -813,6 +818,7 @@ extern void uuid_from_super(int uuid[4], mdp_super_t *super);
 extern const int uuid_match_any[4];
 extern int same_uuid(int a[4], int b[4], int swapuuid);
 extern void copy_uuid(void *a, int b[4], int swapuuid);
+extern char *__fname_from_uuid(int id[4], int swap, char *buf, char sep);
 extern char *fname_from_uuid(struct supertype *st,
 			     struct mdinfo *info, char *buf, char sep);
 extern unsigned long calc_csum(void *super, int bytes);

mdmon.c

@@ -113,6 +113,14 @@ static struct superswitch *find_metadata_methods(char *vers)
 	return NULL;
 }
 
+static int test_pidfile(char *devname)
+{
+	char path[100];
+	struct stat st;
+
+	sprintf(path, "/var/run/mdadm/%s.pid", devname);
+	return stat(path, &st);
+}
 
 int make_pidfile(char *devname, int o_excl)
 {
@@ -149,27 +157,30 @@ int is_container_member(struct mdstat_ent *mdstat, char *container)
 	return 1;
 }
 
-void remove_pidfile(char *devname);
+pid_t devname2mdmon(char *devname)
-static void try_kill_monitor(char *devname)
+{
+	char buf[100];
+	pid_t pid = -1;
+	int fd;
+
+	sprintf(buf, "/var/run/mdadm/%s.pid", devname);
+	fd = open(buf, O_RDONLY|O_NOATIME);
+	if (fd < 0)
+		return -1;
+
+	if (read(fd, buf, sizeof(buf)) > 0)
+		sscanf(buf, "%d\n", &pid);
+	close(fd);
+
+	return pid;
+}
+
+static void try_kill_monitor(pid_t pid, char *devname, int sock)
 {
 	char buf[100];
 	int fd;
-	pid_t pid;
 	struct mdstat_ent *mdstat;
 
-	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);
-
 	/* first rule of survival... don't off yourself */
 	if (pid == getpid())
 		return;
@@ -194,10 +205,9 @@ static void try_kill_monitor(char *devname)
 	for ( ; mdstat; mdstat = mdstat->next)
 		if (is_container_member(mdstat, devname)) {
 			sprintf(buf, "/dev/%s", mdstat->dev);
-			WaitClean(buf, 0);
+			WaitClean(buf, sock, 0);
 		}
 	free_mdstat(mdstat);
-	remove_pidfile(devname);
 }
 
 void remove_pidfile(char *devname)
@@ -355,9 +365,34 @@ int mdmon(char *devname, int devnum, int scan, char *switchroot)
 	int pfd[2];
 	int status;
 	int ignore;
+	pid_t victim = -1;
+	int victim_sock = -1;
 
 	dprintf("starting mdmon for %s in %s\n",
 		devname, switchroot ? : "/");
+
+	/* try to spawn mdmon instances from the target file system */
+	if (switchroot && strcmp(switchroot, "/") != 0) {
+		char path[1024];
+		pid_t pid;
+
+		sprintf(path, "%s/sbin/mdmon", switchroot);
+		switch (fork()) {
+		case 0:
+			execl(path, "mdmon", devname, NULL);
+			exit(1);
+		case -1:
+			return 1;
+		default:
+			pid = wait(&status);
+			if (pid > -1 && WIFEXITED(status) &&
+			    WEXITSTATUS(status) == 0)
+				return 0;
+			else
+				return 1;
+		}
+	}
+
 	mdfd = open_dev(devnum);
 	if (mdfd < 0) {
 		fprintf(stderr, "mdmon: %s: %s\n", devname,
@@ -400,6 +435,7 @@ int mdmon(char *devname, int devnum, int scan, char *switchroot)
 	container->devname = devname;
 	container->arrays = NULL;
 	container->subarray[0] = 0;
+	container->sock = -1;
 
 	if (!container->devname) {
 		fprintf(stderr, "mdmon: failed to allocate container name string\n");
@@ -464,12 +500,10 @@ int mdmon(char *devname, int devnum, int scan, char *switchroot)
 
 	if (switchroot) {
 		/* we assume we assume that /sys /proc /dev are available in
-		 * the new root (see nash:setuproot)
+		 * the new root
-		 *
-		 * kill any monitors in the current namespace and change
-		 * to the new one
 		 */
-		try_kill_monitor(container->devname);
+		victim = devname2mdmon(container->devname);
+		victim_sock = connect_monitor(container->devname);
 		if (chroot(switchroot) != 0) {
 			fprintf(stderr, "mdmon: failed to chroot to '%s': %s\n",
 				switchroot, strerror(errno));
@@ -477,40 +511,15 @@ int mdmon(char *devname, int devnum, int scan, char *switchroot)
 		}
 	}
 
-	/* If this fails, we hope it already exists 
-	 * pid file lives in /var/run/mdadm/mdXX.pid
-	 */
-	mkdir("/var", 0600);
-	mkdir("/var/run", 0600);
-	mkdir("/var/run/mdadm", 0600);
 	ignore = chdir("/");
-	if (make_pidfile(container->devname, O_EXCL) < 0) {
+	if (victim < 0 && test_pidfile(container->devname) == 0) {
 		if (ping_monitor(container->devname) == 0) {
 			fprintf(stderr, "mdmon: %s already managed\n",
 				container->devname);
 			exit(3);
-		} else {
+		} else if (victim < 0)
-			int err;
+			victim = devname2mdmon(container->devname);
-
-			/* cleanup the old monitor, this one is taking over */
-			try_kill_monitor(container->devname);
-			err = make_pidfile(container->devname, 0);
-			if (err < 0) {
-				fprintf(stderr, "mdmon: %s Cannot create pidfile\n",
-					container->devname);
-				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->ss->load_super(container, mdfd, devname)) {
 		fprintf(stderr, "mdmon: Cannot load metadata for %s\n",
 			devname);
@@ -536,7 +545,7 @@ int mdmon(char *devname, int devnum, int scan, char *switchroot)
 	ignore = dup(0);
 #endif
 
-	mlockall(MCL_FUTURE);
+	mlockall(MCL_CURRENT | MCL_FUTURE);
 
 	if (clone_monitor(container) < 0) {
 		fprintf(stderr, "mdmon: failed to start monitor process: %s\n",
@@ -544,6 +553,10 @@ int mdmon(char *devname, int devnum, int scan, char *switchroot)
 		exit(2);
 	}
 
+	if (victim > -1) {
+		try_kill_monitor(victim, container->devname, victim_sock);
+		close(victim_sock);
+	}
 	do_manager(container);
 
 	exit(0);

mdopen.c

@@ -43,7 +43,7 @@ void make_parts(char *dev, int cnt)
 	int odig = odig; /* quiet gcc -Os unitialized warning */
 	int i;
 	int nlen = strlen(dev) + 20;
-	char *name = malloc(nlen);
+	char *name;
 	int dig = isdigit(dev[strlen(dev)-1]);
 	char orig[1024];
 	char sym[1024];
@@ -52,6 +52,7 @@ void make_parts(char *dev, int cnt)
 	if (cnt==0) cnt=4;
 	if (lstat(dev, &stb)!= 0)
 		return;
+
 	if (S_ISLNK(stb.st_mode)) {
 		int len = readlink(dev, orig, sizeof(orig));
 		if (len < 0 || len > 1000)
@@ -63,6 +64,7 @@ void make_parts(char *dev, int cnt)
 		minor_num = minor(stb.st_rdev);
 	} else
 		   return;
+	name = malloc(nlen);
 	for (i=1; i <= cnt ; i++) {
 		struct stat stb2;
 		snprintf(name, nlen, "%s%s%d", dev, dig?"p":"", i);
@@ -92,6 +94,7 @@ void make_parts(char *dev, int cnt)
 		if (err == 0 && stat(name, &stb2) == 0)
 			add_dev(name, &stb2, 0, NULL);
 	}
+	free(name);
 }
 
 
@@ -156,7 +159,6 @@ int create_mddev(char *dev, char *name, int autof, int trustworthy,
 
 
 	if (dev) {
-		
 		if (strncmp(dev, "/dev/md/", 8) == 0) {
 			strcpy(cname, dev+8);
 		} else if (strncmp(dev, "/dev/", 5) == 0) {
@@ -307,7 +309,7 @@ int create_mddev(char *dev, char *name, int autof, int trustworthy,
 		}
 	}
 
-	if (dev)
+	if (dev && dev[0] == '/')
 		strcpy(chosen, dev);
 	else if (cname[0] == 0)
 		strcpy(chosen, devname);

msg.c

@@ -177,10 +177,8 @@ int connect_monitor(char *devname)
 	return sfd;
 }
 
-/* give the monitor a chance to update the metadata */
+int fping_monitor(int sfd)
-int ping_monitor(char *devname)
 {
-	int sfd = connect_monitor(devname);
 	int err = 0;
 
 	if (sfd < 0)
@@ -194,6 +192,16 @@ int ping_monitor(char *devname)
 	if (!err && wait_reply(sfd, 20) != 0)
 		err = -1;
 
+	return err;
+}
+
+
+/* give the monitor a chance to update the metadata */
+int ping_monitor(char *devname)
+{
+	int sfd = connect_monitor(devname);
+	int err = fping_monitor(sfd);
+
 	close(sfd);
 	return err;
 }

msg.h

@@ -27,6 +27,7 @@ extern int ack(int fd, int tmo);
 extern int wait_reply(int fd, int tmo);
 extern int connect_monitor(char *devname);
 extern int ping_monitor(char *devname);
+extern int fping_monitor(int sock);
 extern int ping_manager(char *devname);
 
 #define MSG_MAX_LEN (4*1024*1024)

super-ddf.c

@@ -1589,13 +1589,8 @@ static int init_super_ddf(struct supertype *st,
 	struct phys_disk *pd;
 	struct virtual_disk *vd;
 
-	if (!info) {
-		st->sb = NULL;
-		return 0;
-	}
 	if (st->sb)
-		return init_super_ddf_bvd(st, info, size, name, homehost,
+		return init_super_ddf_bvd(st, info, size, name, homehost, uuid);
-					  uuid);
 
 	if (posix_memalign((void**)&ddf, 512, sizeof(*ddf)) != 0) {
 		fprintf(stderr, Name ": %s could not allocate superblock\n", __func__);
@@ -1604,6 +1599,12 @@ static int init_super_ddf(struct supertype *st,
 	memset(ddf, 0, sizeof(*ddf));
 	ddf->dlist = NULL; /* no physical disks yet */
 	ddf->conflist = NULL; /* No virtual disks yet */
+	st->sb = ddf;
+
+	if (info == NULL) {
+		/* zeroing superblock */
+		return 0;
+	}
 
 	/* At least 32MB *must* be reserved for the ddf.  So let's just
 	 * start 32MB from the end, and put the primary header there.
@@ -2971,12 +2972,22 @@ static struct mdinfo *container_content_ddf(struct supertype *st)
 	return rest;
 }
 
-static int store_zero_ddf(struct supertype *st, int fd)
+static int store_super_ddf(struct supertype *st, int fd)
 {
+	struct ddf_super *ddf = st->sb;
 	unsigned long long dsize;
 	void *buf;
 	int rc;
 
+	if (!ddf)
+		return 1;
+
+	/* ->dlist and ->conflist will be set for updates, currently not
+	 * supported
+	 */
+	if (ddf->dlist || ddf->conflist)
+		return 1;
+
 	if (!get_dev_size(fd, NULL, &dsize))
 		return 1;
 
@@ -3627,7 +3638,7 @@ struct superswitch super_ddf = {
 
 	.load_super	= load_super_ddf,
 	.init_super	= init_super_ddf,
-	.store_super	= store_zero_ddf,
+	.store_super	= store_super_ddf,
 	.free_super	= free_super_ddf,
 	.match_metadata_desc = match_metadata_desc_ddf,
 	.container_content = container_content_ddf,

sysfs.c

@@ -792,7 +792,7 @@ int sysfs_unique_holder(int devnum, long rdev)
 static char *clean_states[] = {
 	"clear", "inactive", "readonly", "read-auto", "clean", NULL };
 
-int WaitClean(char *dev, int verbose)
+int WaitClean(char *dev, int sock, int verbose)
 {
 	int fd;
 	struct mdinfo *mdi;
@@ -868,7 +868,8 @@ int WaitClean(char *dev, int verbose)
 		}
 		if (rv < 0)
 			rv = 1;
-		else if (ping_monitor(mdi->text_version) == 0) {
+		else if (fping_monitor(sock) == 0 ||
+			 ping_monitor(mdi->text_version) == 0) {
 			/* we need to ping to close the window between array
 			 * state transitioning to clean and the metadata being
 			 * marked clean

tests/09imsm-assemble

@@ -0,0 +1,46 @@
+# validate the prodigal member disk scenario i.e. a former container
+# member is returned after having been rebuilt on another system
+num_disks=4
+size=$((10*1024))
+mdadm -CR $container -e imsm -n $num_disks $dev0 $dev1 $dev2 $dev3
+mdadm -CR $member $dev0 $dev2 -n 2 -l 1 -z $size
+mdadm --wait $member
+mdadm -Ss
+
+# make dev0 and dev1 a new rebuild family
+mdadm -A $container $dev0 $dev1
+mdadm -I $container
+mdadm --wait ${member}_0
+mdadm -Ss
+
+# make dev2 and dev3 a new rebuild family
+mdadm -A $container $dev2 $dev3
+mdadm -I $container
+mdadm --wait ${member}_0
+mdadm -Ss
+
+# reassemble and make sure one of the families falls out
+mdadm -A $container $dev0 $dev1 $dev2 $dev3
+mdadm -I $container
+testdev ${member}_0 1 $size 1
+if mdadm --remove $container $dev0 ; then
+    # the dev[23] family won
+    imsm_check_removal $container $dev1
+    imsm_check_hold $container $dev2
+    imsm_check_hold $container $dev3
+else
+    # the dev[01] family won
+    imsm_check_hold $container $dev1
+    imsm_check_removal $container $dev2
+    imsm_check_removal $container $dev3
+fi
+mdadm -Ss
+
+# reassemble with a new id for the dev[23] family
+mdadm -A $container $dev0 $dev1
+mdadm -I $container
+mdadm -A ${container}2 $dev2 $dev3 --update=uuid
+mdadm -I ${container}2
+
+testdev ${member}_0 1 $size 1
+testdev ${member}_1 1 $size 1

tests/10ddf-create

@@ -55,8 +55,8 @@ mdadm -Ss
 mdadm -Asc /var/tmp/mdadm.conf
 check nosync  # This failed once. The raid5 was resyncing.
 
-mdadm -Dbs > /tmp/mdadm.conf
+mdadm -Dbs | sort > /tmp/mdadm.conf
-diff /tmp/mdadm.conf /var/tmp/mdadm.conf
+sort /var/tmp/mdadm.conf | diff /tmp/mdadm.conf -
 mdadm -Ss
 
 # and now assemble fully incrementally.
@@ -70,7 +70,7 @@ do
 done
 check nosync
 
-mdadm -Dbs > /tmp/mdadm.conf
+mdadm -Dbs | sort > /tmp/mdadm.conf
-diff /tmp/mdadm.conf /var/tmp/mdadm.conf
+sort /var/tmp/mdadm.conf | diff /tmp/mdadm.conf -
 mdadm -Ss
 rm /tmp/mdadm.conf /var/tmp/mdadm.conf

tests/env-09imsm-assemble

@@ -0,0 +1,32 @@
+imsm_check_hold() {
+   if mdadm --remove $1 $2; then
+       echo "$2 removal from $1 should have been blocked" >&2
+       cat /proc/mdstat >&2
+       mdadm -E $2
+       exit 1
+   fi
+}
+
+imsm_check_removal() {
+   if ! mdadm --remove $1 $2 ; then
+       echo "$2 removal from $1 should have succeeded" >&2
+       cat /proc/mdstat >&2
+       mdadm -E $2
+       exit 1
+   fi
+}
+
+setup_env() {
+	export IMSM_DEVNAME_AS_SERIAL=1
+	export IMSM_TEST_OROM=1
+	container=/dev/md/container
+	member=/dev/md/vol0
+}
+
+reset_env() {
+	unset IMSM_DEVNAME_AS_SERIAL
+	unset IMSM_TEST_OROM
+	unset imsm_check
+	unset container
+	unset member
+}

util.c

@@ -336,17 +336,15 @@ void copy_uuid(void *a, int b[4], int swapuuid)
 		memcpy(a, b, 16);
 }
 
-char *fname_from_uuid(struct supertype *st, struct mdinfo *info, char *buf, char sep)
+char *__fname_from_uuid(int id[4], int swap, char *buf, char sep)
 {
 	int i, j;
-	int id;
 	char uuid[16];
 	char *c = buf;
 	strcpy(c, "UUID-");
 	c += strlen(c);
-	copy_uuid(uuid, info->uuid, st->ss->swapuuid);
+	copy_uuid(uuid, id, swap);
 	for (i = 0; i < 4; i++) {
-		id = uuid[i];
 		if (i)
 			*c++ = sep;
 		for (j = 3; j >= 0; j--) {
@@ -355,6 +353,12 @@ char *fname_from_uuid(struct supertype *st, struct mdinfo *info, char *buf, char
 		}
 	}
 	return buf;
+
+}
+
+char *fname_from_uuid(struct supertype *st, struct mdinfo *info, char *buf, char sep)
+{
+	return __fname_from_uuid(info->uuid, st->ss->swapuuid, buf, sep);
 }
 
 #ifndef MDASSEMBLE