Incremental: lock against multiple concurrent additions to an array.

In two devices are added via -I to one array at the same time, mdadm
can get badly confused.

Signed-off-by: NeilBrown <neilb@suse.de>

Build.c

@@ -62,6 +62,7 @@ int Build(char *mddev, int chunk, int level, int layout,
 	int mdfd;
 	char chosen_name[1024];
 	int uuid[4] = {0,0,0,0};
+	struct map_ent *map = NULL;
 
 	/* scan all devices, make sure they really are block devices */
 	for (dv = devlist; dv; dv=dv->next) {
@@ -116,12 +117,16 @@ int Build(char *mddev, int chunk, int level, int layout,
 		}
 
 	/* We need to create the device.  It can have no name. */
+	map_lock(&map);
 	mdfd = create_mddev(mddev, NULL, autof, LOCAL,
 			    chosen_name);
-	if (mdfd < 0)
+	if (mdfd < 0) {
+		map_unlock(&map);
 		return 1;
+	}
 
-	map_update(NULL, fd2devnum(mdfd), "none", uuid, chosen_name);
+	map_update(&map, fd2devnum(mdfd), "none", uuid, chosen_name);
+	map_unlock(&map);
 
 	vers = md_get_version(mdfd);
 

Create.c

@@ -80,6 +80,7 @@ int Create(struct supertype *st, char *mddev,
 	int did_default = 0;
 	unsigned long safe_mode_delay = 0;
 	char chosen_name[1024];
+	struct map_ent *map = NULL;
 
 	int major_num = BITMAP_MAJOR_HI;
 
@@ -422,6 +423,7 @@ int Create(struct supertype *st, char *mddev,
 	}
 
 	/* We need to create the device */
+	map_lock(&map);
 	mdfd = create_mddev(mddev, name, autof, LOCAL, chosen_name);
 	if (mdfd < 0)
 		return 1;
@@ -561,8 +563,9 @@ int Create(struct supertype *st, char *mddev,
 				" %s metadata\n", info.text_version);
 	}
 
-	map_update(NULL, fd2devnum(mdfd), info.text_version,
+	map_update(&map, fd2devnum(mdfd), info.text_version,
 		   info.uuid, chosen_name);
+	map_unlock(&map);
 
 	if (bitmap_file && vers < 9003) {
 		major_num = BITMAP_MAJOR_HOSTENDIAN;

Incremental.c

@@ -88,6 +88,7 @@ int Incremental(char *devname, int verbose, int runstop,
 	int active_disks;
 	int trustworthy = FOREIGN;
 	char *name_to_use;
+	mdu_array_info_t ainf;
 
 	struct createinfo *ci = conf_get_create_info();
 
@@ -243,8 +244,9 @@ int Incremental(char *devname, int verbose, int runstop,
 		return Incremental_container(st, devname, verbose, runstop,
 					     autof, trustworthy);
 	}
-	/* 4/ Check is array exists.
+	/* 4/ Check if array exists.
 	 */
+	map_lock(&map);
 	mp = map_by_uuid(&map, info.uuid);
 	if (mp)
 		mdfd = open_mddev(mp->path, 0);
@@ -298,6 +300,10 @@ int Incremental(char *devname, int verbose, int runstop,
 		}
 		info.array.working_disks = 1;
 		sysfs_free(sra);
+		/* 6/ Make sure /var/run/mdadm.map contains this array. */
+		map_update(&map, fd2devnum(mdfd),
+			   info.text_version,
+			   info.uuid, chosen_name);
 	} else {
 	/* 5b/ if it does */
 	/* - check one drive in array to make sure metadata is a reasonably */
@@ -366,16 +372,13 @@ int Incremental(char *devname, int verbose, int runstop,
 			info.array.working_disks ++;
 			
 	}
-	/* 6/ Make sure /var/run/mdadm.map contains this array. */
-	map_update(&map, fd2devnum(mdfd),
-		   info.text_version,
-		   info.uuid, chosen_name);
 
 	/* 7/ Is there enough devices to possibly start the array? */
 	/* 7a/ if not, finish with success. */
 	if (info.array.level == LEVEL_CONTAINER) {
 		/* Try to assemble within the container */
 		close(mdfd);
+		map_unlock(&map);
 		sysfs_uevent(&info, "change");
 		if (verbose >= 0)
 			fprintf(stderr, Name
@@ -394,6 +397,7 @@ int Incremental(char *devname, int verbose, int runstop,
 			fprintf(stderr, Name
 			     ": %s attached to %s, not enough to start (%d).\n",
 				devname, chosen_name, active_disks);
+		map_unlock(&map);
 		close(mdfd);
 		return 0;
 	}
@@ -404,18 +408,18 @@ int Incremental(char *devname, int verbose, int runstop,
 	/*             are enough, */
 	/*   + add any bitmap file  */
 	/*   + start the array (auto-readonly). */
-{
-	mdu_array_info_t ainf;
 
 	if (ioctl(mdfd, GET_ARRAY_INFO, &ainf) == 0) {
 		if (verbose >= 0)
 			fprintf(stderr, Name
 			   ": %s attached to %s which is already active.\n",
 				devname, chosen_name);
-		close (mdfd);
+		close(mdfd);
+		map_unlock(&map);
 		return 0;
 	}
-}
+
+	map_unlock(&map);
 	if (runstop > 0 || active_disks >= info.array.working_disks) {
 		struct mdinfo *sra;
 		/* Let's try to start it */
@@ -749,13 +753,16 @@ int Incremental_container(struct supertype *st, char *devname, int verbose,
 
 	struct mdinfo *list = st->ss->container_content(st);
 	struct mdinfo *ra;
+	struct map_ent *map = NULL;
+
+	map_lock(&map);
 
 	for (ra = list ; ra ; ra = ra->next) {
 		struct mdinfo *dev, *sra;
 		int mdfd;
 		char chosen_name[1024];
 		int working = 0, preexist = 0;
-		struct map_ent *mp, *map = NULL;
+		struct map_ent *mp;
 		struct mddev_ident_s *match = NULL;
 
 		mp = map_by_uuid(&map, ra->uuid);
@@ -869,5 +876,6 @@ int Incremental_container(struct supertype *st, char *devname, int verbose,
 			   ra->uuid, chosen_name);
 		close(mdfd);
 	}
+	map_unlock(&map);
 	return 0;
 }

mapfile.c

@@ -86,6 +86,43 @@ int map_write(struct map_ent *mel)
 			      "/var/run/mdadm.map") == 0;
 }
 
+
+static int lfd = -1;
+static int lsubdir = 0;
+int map_lock(struct map_ent **melp)
+{
+	if (lfd < 0) {
+		lfd = open("/var/run/mdadm/map.lock", O_CREAT|O_RDWR, 0600);
+		if (lfd < 0) {
+			lfd = open("/var/run/mdadm.map.lock", O_CREAT|O_RDWR, 0600);
+			lsubdir = 0;
+		} else
+			lsubdir = 1;
+		if (lfd < 0)
+			return -1;
+		if (lockf(lfd, F_LOCK, 0) != 0) {
+			close(lfd);
+			lfd = -1;
+			return -1;
+		}
+	}
+	if (*melp)
+		map_free(*melp);
+	map_read(melp);
+	return 0;
+}
+
+void map_unlock(struct map_ent **melp)
+{
+	if (lfd >= 0)
+		close(lfd);
+	if (lsubdir)
+		unlink("/var/run/mdadm/map.lock");
+	else
+		unlink("/var/run/mdadm.map.lock");
+	lfd = -1;
+}
+
 void map_add(struct map_ent **melp,
 	    int devnum, char *metadata, int uuid[4], char *path)
 {

mdadm.h

@@ -325,6 +325,8 @@ extern void map_delete(struct map_ent **mapp, int devnum);
 extern void map_free(struct map_ent *map);
 extern void map_add(struct map_ent **melp,
 		    int devnum, char *metadata, int uuid[4], char *path);
+extern int map_lock(struct map_ent **melp);
+extern void map_unlock(struct map_ent **melp);
 
 /* various details can be requested */
 #define	GET_LEVEL	1