pre-release

Add vgid to struct physical_volume and pass with vg_name to some functions.

Makefile.in

@@ -24,8 +24,13 @@ endif
 
 SUBDIRS += lib tools daemons
 
+ifeq ("@DMEVENTD@", "yes")
+  SUBDIRS += dmeventd
+endif
+
 ifeq ($(MAKECMDGOALS),distclean)
   SUBDIRS += daemons/clvmd \
+	     dmeventd \
 	     lib/format1 \
 	     lib/format_pool \
 	     lib/locking \
@@ -40,13 +45,15 @@ include make.tmpl
 daemons: lib
 lib: include
 tools: lib
-po: tools daemons
+dmeventd: tools
+po: tools daemons dmeventd
 
 ifeq ("@INTL@", "yes")
 lib.pofile: include.pofile
 tools.pofile: lib.pofile
 daemons.pofile: lib.pofile
-po.pofile: tools.pofile daemons.pofile
+dmeventd.pofile: tools.pofile
+po.pofile: tools.pofile daemons.pofile dmeventd.pofile
 pofile: po.pofile
 endif
 

VERSION

@@ -1 +1 @@
-2.02.00-cvs (2005-10-16)
+2.02.03-cvs (2006-04-14)

WHATS_NEW

@@ -1,5 +1,56 @@
-Version 2.02.00 - 
+Version 2.02.03 - 14th April 2006
+=================================
+  vgrename accepts vgid and exported VG.
+  Add --partial to pvs.
+  When choosing between identically-named VGs, also consider creation_host.
+  Provide total log suppression with 2.
+  Fix vgexport/vgimport to set/reset PV exported flag so pv_attr is correct.
+  Add vgid to struct physical_volume and pass with vg_name to some functions.
+  If two or more VGs are found with the same name, use one that is not exported.
+  Whenever vgname is captured, also capture vgid and whether exported.
+  Remove an incorrect unlock_vg() from process_each_lv().
+  Update extent size information in vgchange and vgcreate man pages.
+  Introduce origin_from_cow() and lv_is_visible().
+  pvremove without -f now fails if there's no PV label.
+  Support lvconvert -s.
+  Suppress locking library load failure message if --ignorelockingfailure.
+  Propagate partial mode around cluster.
+  Fix archive file expiration.
+  Fix dmeventd build.
+  clvmd now uses libcman rather than cman ioctls.
+  clvmd will allow new cman to shutdown on request.
+
+Version 2.02.02 - 7th February 2006
 ===================================
+  Add %.so: %.a make template rule.
+  Switchover library building to use LIB_SUFFIX.
+  Only do lockfs filesystem sync when suspending snapshots.
+  Always print warning if activation is disabled.
+  vgreduce removes mirror images.
+  Add --mirrorsonly to vgreduce.
+  vgreduce replaces active LVs with error segment before removing them.
+  Set block_on_error parameter if available.
+  Add target_version.
+  Add details to format1 'Invalid LV in extent map' error message.
+  Fix lvscan snapshot full display.
+  Bring lvdisplay man page example into line.
+  Add mirror dmeventd library.
+  Add some activation logic to remove_mirror_images().
+  lvconvert can remove specified PVs from a mirror.
+  lvconvert turns an existing LV into a mirror.
+  Allow signed mirrors arguments.
+  Move create_mirror_log() into toollib.
+  Determine parallel PVs to avoid with ALLOC_NORMAL allocation.
+  Fix lv_empty.
+
+Version 2.02.01 - 23rd November 2005
+====================================
+  Fix lvdisplay cmdline to accept snapshots.
+  Fix open RO->RW promotion.
+  Fix missing vg_revert in lvcreate error path.
+
+Version 2.02.00 - 10th November 2005
+====================================
   Extend allocation areas to avoid overflow with contiguous with other PVs.
   Stop lvcreate attempting to wipe zero or error segments.
   Added new lvs table attributes.

WHATS_NEW_DM

@@ -1,4 +1,41 @@
-Version 1.02.00 - 
+Version 1.02.04 - 14 Apr 2006
+=============================
+  Bring dmsetup man page up-to-date.
+  Use name-based device refs if kernel doesn't support device number refs.
+  Fix memory leak (struct dm_ioctl) when struct dm_task is reused.
+  If _create_and_load_v4 fails part way through, revert the creation.
+  dmeventd thread/fifo fixes.
+  Add file & line to dm_strdup_aux().
+  Add setgeometry.
+
+Version 1.02.03 - 7 Feb 2006
+============================
+  Add exported functions to set uid, gid and mode.
+  Rename _log to dm_log and export.
+  Add dm_tree_skip_lockfs.
+  Fix dm_strdup debug definition.
+  Fix hash function to avoid using a negative array offset.
+  Don't inline _find in hash.c and tidy signed/unsigned etc.
+  Fix libdevmapper.h #endif.
+  Fix dmsetup version driver version.
+  Add sync, nosync and block_on_error mirror log parameters.
+  Add hweight32.
+  Fix dmeventd build.
+
+Version 1.02.02 - 2 Dec 2005
+============================
+  dmeventd added.
+  Export dm_task_update_nodes.
+  Use names instead of numbers in messages when ioctls fail.
+
+Version 1.02.01 - 23 Nov 2005
+=============================
+  Resume snapshot-origins last.
+  Drop leading zeros from dm_format_dev.
+  Suppress attempt to reload identical table.
+  Additional LVM- prefix matching for transitional period.
+
+Version 1.02.00 - 10 Nov 2005
 =============================
   Added activation functions to library.
   Added return macros.

configure.in

@@ -35,7 +35,7 @@ case "$host_os" in
 		CLDNOWHOLEARCHIVE="-Wl,-no-whole-archive"
 		LDDEPS="$LDDEPS .export.sym"
 		LDFLAGS="$LDFLAGS -Wl,--export-dynamic"
-		SOFLAG="-shared"
+		LIB_SUFFIX="so"
 		DEVMAPPER=yes
 		ODIRECT=yes
 		SELINUX=yes
@@ -49,7 +49,7 @@ case "$host_os" in
 		CLDNOWHOLEARCHIVE=
 		LDDEPS="$LDDEPS"
 		LDFLAGS="$LDFLAGS"
-		SOFLAG="-dynamiclib"
+		LIB_SUFFIX="dylib"
 		DEVMAPPER=yes
 		ODIRECT=no
 		SELINUX=no
@@ -355,6 +355,23 @@ AC_ARG_ENABLE(fsadm, [  --enable-fsadm          Enable fsadm],
 FSADM=$enableval)
 AC_MSG_RESULT($FSADM)
 
+################################################################################
+dnl -- enable dmeventd handling
+AC_MSG_CHECKING(whether to use dmeventd)
+AC_ARG_ENABLE(dmeventd, [  --enable-dmeventd       Enable the device-mapper event daemon],
+DMEVENTD=$enableval)
+AC_MSG_RESULT($DMEVENTD)
+
+dnl -- dmeventd currently requires internal mirror support
+if test x$DMEVENTD = xyes && test x$MIRRORS != xinternal; then
+AC_MSG_ERROR(
+--enable-dmeventd currently requires --with-mirrors=internal
+)
+fi
+
+if test x$DMEVENTD = xyes; then
+	CFLAGS="$CFLAGS -DDMEVENTD"
+fi
 ################################################################################
 dnl -- Mess with default exec_prefix
 if [[ "x$exec_prefix" = xNONE -a "x$prefix" = xNONE ]];
@@ -545,7 +562,7 @@ AC_SUBST(CLDWHOLEARCHIVE)
 AC_SUBST(CLDNOWHOLEARCHIVE)
 AC_SUBST(LDDEPS)
 AC_SUBST(LDFLAGS)
-AC_SUBST(SOFLAG)
+AC_SUBST(LIB_SUFFIX)
 AC_SUBST(LIBS)
 AC_SUBST(LVM_VERSION)
 AC_SUBST(LVM1_FALLBACK)
@@ -563,6 +580,7 @@ AC_SUBST(INTL)
 AC_SUBST(CLVMD)
 AC_SUBST(CLUSTER)
 AC_SUBST(FSADM)
+AC_SUBST(DMEVENTD)
 
 ################################################################################
 dnl -- First and last lines should not contain files to generate in order to 
@@ -572,6 +590,8 @@ Makefile								\
 make.tmpl								\
 daemons/Makefile							\
 daemons/clvmd/Makefile							\
+dmeventd/Makefile							\
+dmeventd/mirror/Makefile						\
 doc/Makefile								\
 include/Makefile						 	\
 lib/Makefile							 	\
@@ -599,3 +619,7 @@ fi
 if test x$FSADM == xyes; then
   AC_MSG_WARN(fsadm support is untested)
 fi
+
+if test x$DMEVENTD == xyes; then
+  AC_MSG_WARN(dmeventd support is untested)
+fi

daemons/clvmd/Makefile.in

@@ -46,17 +46,26 @@ endif
 
 ifeq ("$(CMAN)", "yes")
 	SOURCES += clvmd-cman.c
-	LMLIBS += -ldlm
+	LMLIBS += -ldlm -lcman
 	CFLAGS += -DUSE_CMAN
 endif
 
 TARGETS = \
 	clvmd
 
-include $(top_srcdir)/make.tmpl
+LVMLIBS = -llvm -lpthread
+
+ifeq ("@DMEVENTD@", "yes")
+	LVMLIBS += -ldevmapper-event
+endif
+ 
+ifeq ("@DEVMAPPER@", "yes")
+	LVMLIBS += -ldevmapper
+endif
 
 CFLAGS += -D_REENTRANT -fno-strict-aliasing
-LIBS += -ldevmapper -llvm -lpthread
+
+include $(top_srcdir)/make.tmpl
 
 INSTALL_TARGETS = \
 	install_clvmd

daemons/clvmd/clvmd-cman.c

@@ -46,19 +46,23 @@
 
 #define LOCKSPACE_NAME "clvmd"
 
-static int cluster_sock;
 static int num_nodes;
-static struct cl_cluster_node *nodes = NULL;
+static struct cman_node *nodes = NULL;
+static struct cman_node this_node;
 static int count_nodes; /* size of allocated nodes array */
 static int max_updown_nodes = 50;	/* Current size of the allocated array */
 /* Node up/down status, indexed by nodeid */
 static int *node_updown = NULL;
 static dlm_lshandle_t *lockspace;
+static cman_handle_t c_handle;
 
 static void count_clvmds_running(void);
 static void get_members(void);
 static int nodeid_from_csid(char *csid);
 static int name_from_nodeid(int nodeid, char *name);
+static void event_callback(cman_handle_t handle, void *private, int reason, int arg);
+static void data_callback(cman_handle_t handle, void *private,
+			  char *buf, int len, uint8_t port, int nodeid);
 
 struct lock_wait {
 	pthread_cond_t cond;
@@ -68,30 +72,23 @@ struct lock_wait {
 
 static int _init_cluster(void)
 {
-	struct sockaddr_cl saddr;
-	int port = CLUSTER_PORT_CLVMD;
-
 	/* Open the cluster communication socket */
-	cluster_sock = socket(AF_CLUSTER, SOCK_DGRAM, CLPROTO_CLIENT);
-	if (cluster_sock == -1) {
-		/* Don't print an error here because we could be just probing for CMAN */
+	c_handle = cman_init(NULL);
+	if (!c_handle) {
+		syslog(LOG_ERR, "Can't open cluster manager socket: %m");
 		return -1;
 	}
-	/* Set Close-on-exec */
-	fcntl(cluster_sock, F_SETFD, 1);
 
-	/* Bind to our port number on the cluster.
-	   Writes to this will block if the cluster loses quorum */
-	saddr.scl_family = AF_CLUSTER;
-	saddr.scl_port = port;
-
-	if (bind
-	    (cluster_sock, (struct sockaddr *) &saddr,
-	     sizeof(struct sockaddr_cl))) {
+	if (cman_start_recv_data(c_handle, data_callback, CLUSTER_PORT_CLVMD)) {
 		syslog(LOG_ERR, "Can't bind cluster socket: %m");
 		return -1;
 	}
 
+	if (cman_start_notification(c_handle, event_callback)) {
+		syslog(LOG_ERR, "Can't start cluster event listening");
+		return -1;
+	}
+
 	/* Get the cluster members list */
 	get_members();
 	count_clvmds_running();
@@ -114,63 +111,46 @@ static void _cluster_init_completed(void)
 
 static int _get_main_cluster_fd()
 {
-	return cluster_sock;
+	return cman_get_fd(c_handle);
 }
 
 static int _get_num_nodes()
 {
-	return num_nodes;
+	int i;
+	int nnodes = 0;
+
+	/* return number of ACTIVE nodes */
+	for (i=0; i<num_nodes; i++) {
+		if (nodes[i].cn_member)
+			nnodes++;
+	}
+	return nnodes;
 }
 
 /* send_message with the fd check removed */
 static int _cluster_send_message(void *buf, int msglen, char *csid, const char *errtext)
 {
-	struct iovec iov[2];
-	struct msghdr msg;
-	struct sockaddr_cl saddr;
-	int len = 0;
+	int nodeid = 0;
 
-	msg.msg_control = NULL;
-	msg.msg_controllen = 0;
-	msg.msg_iovlen = 1;
-	msg.msg_iov = iov;
-	msg.msg_flags = 0;
-	iov[0].iov_len = msglen;
-	iov[0].iov_base = buf;
+	if (csid)
+		memcpy(&nodeid, csid, CMAN_MAX_CSID_LEN);
 
-	saddr.scl_family = AF_CLUSTER;
-	saddr.scl_port = CLUSTER_PORT_CLVMD;
-	if (csid) {
-		msg.msg_name = &saddr;
-		msg.msg_namelen = sizeof(saddr);
-		memcpy(&saddr.scl_nodeid, csid, CMAN_MAX_CSID_LEN);
-	} else {		/* Cluster broadcast */
-
-		msg.msg_name = NULL;
-		msg.msg_namelen = 0;
-	}
-
-	do {
-		len = sendmsg(cluster_sock, &msg, 0);
-		if (len < 0 && errno != EAGAIN)
+	if (cman_send_data(c_handle, buf, msglen, 0, CLUSTER_PORT_CLVMD, nodeid) <= 0)
+	{
 			log_error(errtext);
-
-	} while (len == -1 && errno == EAGAIN);
-	return len;
+	}
+	return msglen;
 }
 
 static void _get_our_csid(char *csid)
 {
-	int i;
-	memset(csid, 0, CMAN_MAX_CSID_LEN);
-
-	for (i = 0; i < num_nodes; i++) {
-		if (nodes[i].us)
-			memcpy(csid, &nodes[i].node_id, CMAN_MAX_CSID_LEN);
+	if (this_node.cn_nodeid == 0) {
+		cman_get_node(c_handle, 0, &this_node);
 	}
+	memcpy(csid, &this_node.cn_nodeid, CMAN_MAX_CSID_LEN);
 }
 
-/* Call a callback routine for each node that known (down mean not running a clvmd) */
+/* Call a callback routine for each node is that known (down means not running a clvmd) */
 static int _cluster_do_node_callback(struct local_client *client,
 			     void (*callback) (struct local_client *, char *,
 					       int))
@@ -179,8 +159,8 @@ static int _cluster_do_node_callback(struct local_client *client,
 	int somedown = 0;
 
 	for (i = 0; i < _get_num_nodes(); i++) {
-		callback(client, (char *)&nodes[i].node_id, node_updown[nodes[i].node_id]);
-		if (!node_updown[nodes[i].node_id])
+		callback(client, (char *)&nodes[i].cn_nodeid, node_updown[nodes[i].cn_nodeid]);
+		if (!node_updown[nodes[i].cn_nodeid])
 			somedown = -1;
 	}
 	return somedown;
@@ -188,78 +168,63 @@ static int _cluster_do_node_callback(struct local_client *client,
 
 /* Process OOB message from the cluster socket,
    this currently just means that a node has stopped listening on our port */
-static void process_oob_msg(char *buf, int len, int nodeid)
+static void event_callback(cman_handle_t handle, void *private, int reason, int arg)
 {
-	char namebuf[256];
-	switch (buf[0]) {
-        case CLUSTER_OOB_MSG_PORTCLOSED:
-		name_from_nodeid(nodeid, namebuf);
-		log_notice("clvmd on node %s has died\n", namebuf);
-		DEBUGLOG("Got OOB message, removing node %s\n", namebuf);
+	char namebuf[MAX_CLUSTER_NAME_LEN];
 
-		node_updown[nodeid] = 0;
+	switch (reason) {
+        case CMAN_REASON_PORTCLOSED:
+		name_from_nodeid(arg, namebuf);
+		log_notice("clvmd on node %s has died\n", namebuf);
+		DEBUGLOG("Got port closed message, removing node %s\n", namebuf);
+
+		node_updown[arg] = 0;
 		break;
 
-	case CLUSTER_OOB_MSG_STATECHANGE:
-		DEBUGLOG("Got OOB message, Cluster state change\n");
+	case CMAN_REASON_STATECHANGE:
+		DEBUGLOG("Got state change message, re-reading members list\n");
 		get_members();
 		break;
+
+#if defined(LIBCMAN_VERSION) && LIBCMAN_VERSION >= 2
+	case CMAN_REASON_PORTOPENED:
+		/* Ignore this, wait for startup message from clvmd itself */
+		break;
+
+	case CMAN_REASON_TRY_SHUTDOWN:
+		DEBUGLOG("Got try shutdown, sending OK\n");
+		cman_replyto_shutdown(c_handle, 1);
+		break;
+#endif
 	default:
 		/* ERROR */
-		DEBUGLOG("Got unknown OOB message: %d\n", buf[0]);
+		DEBUGLOG("Got unknown event callback message: %d\n", reason);
+		break;
 	}
 }
 
-static int _cluster_fd_callback(struct local_client *client, char *buf, int len, char *csid,
+static struct local_client *cman_client;
+static int _cluster_fd_callback(struct local_client *fd, char *buf, int len, char *csid,
 			struct local_client **new_client)
 {
-	struct iovec iov[2];
-	struct msghdr msg;
-	struct sockaddr_cl saddr;
+
+	/* Save this for data_callback */
+	cman_client = fd;
 
 	/* We never return a new client */
 	*new_client = NULL;
 
-	msg.msg_control = NULL;
-	msg.msg_controllen = 0;
-	msg.msg_iovlen = 1;
-	msg.msg_iov = iov;
-	msg.msg_name = &saddr;
-	msg.msg_flags = 0;
-	msg.msg_namelen = sizeof(saddr);
-	iov[0].iov_len = len;
-	iov[0].iov_base = buf;
+	return cman_dispatch(c_handle, 0);
+}
 
-	len = recvmsg(cluster_sock, &msg, MSG_OOB | O_NONBLOCK);
-	if (len < 0 && errno == EAGAIN)
-		return len;
 
-	DEBUGLOG("Read on cluster socket, len = %d\n", len);
-
-	/* A real error */
-	if (len < 0) {
-		log_error("read error on cluster socket: %m");
-		return 0;
-	}
-
-	/* EOF - we have left the cluster */
-	if (len == 0)
-		return 0;
-
-	/* Is it OOB? probably a node gone down */
-	if (msg.msg_flags & MSG_OOB) {
-		process_oob_msg(iov[0].iov_base, len, saddr.scl_nodeid);
-
-		/* Tell the upper layer to ignore this message */
-		len = -1;
-		errno = EAGAIN;
-	}
-	else {
-		memcpy(csid, &saddr.scl_nodeid, sizeof(saddr.scl_nodeid));
-		/* Send it back to clvmd */
-		process_message(client, buf, len, csid);
-	}
-	return len;
+static void data_callback(cman_handle_t handle, void *private,
+			  char *buf, int len, uint8_t port, int nodeid)
+{
+	/* Ignore looped back messages */
+	if (nodeid == this_node.cn_nodeid)
+		return;
+	process_message(cman_client, buf, len, (char *)&nodeid);
 }
 
 static void _add_up_node(char *csid)
@@ -290,19 +255,15 @@ static void _cluster_closedown()
 {
 	unlock_all();
 	dlm_release_lockspace(LOCKSPACE_NAME, lockspace, 1);
-	close(cluster_sock);
+	cman_finish(c_handle);
 }
 
 static int is_listening(int nodeid)
 {
-	struct cl_listen_request rq;
 	int status;
 
-	rq.port = CLUSTER_PORT_CLVMD;
-	rq.nodeid = nodeid;
-
 	do {
-		status = ioctl(cluster_sock, SIOCCLUSTER_ISLISTENING, &rq);
+		status = cman_is_listening(c_handle, nodeid, CLUSTER_PORT_CLVMD);
 		if (status < 0 && errno == EBUSY) {	/* Don't busywait */
 			sleep(1);
 			errno = EBUSY;	/* In case sleep trashes it */
@@ -320,19 +281,22 @@ static void count_clvmds_running(void)
 	int i;
 
 	for (i = 0; i < num_nodes; i++) {
-		node_updown[nodes[i].node_id] = is_listening(nodes[i].node_id);
+		node_updown[nodes[i].cn_nodeid] = is_listening(nodes[i].cn_nodeid);
 	}
 }
 
 /* Get a list of active cluster members */
 static void get_members()
 {
-	struct cl_cluster_nodelist nodelist;
+	int retnodes;
+	int status;
 
-	num_nodes = ioctl(cluster_sock, SIOCCLUSTER_GETMEMBERS, 0);
+	num_nodes = cman_get_node_count(c_handle);
 	if (num_nodes == -1) {
 		log_error("Unable to get node count");
-	} else {
+		return;
+	}
+
 	        /* Not enough room for new nodes list ? */
 	        if (num_nodes > count_nodes && nodes) {
 			free(nodes);
@@ -341,28 +305,19 @@ static void get_members()
 
 		if (nodes == NULL) {
 		        count_nodes = num_nodes + 10; /* Overallocate a little */
-		        nodes = malloc(count_nodes * sizeof(struct cl_cluster_node));
+		nodes = malloc(count_nodes * sizeof(struct cman_node));
 			if (!nodes) {
 			        log_error("Unable to allocate nodes array\n");
 				exit(5);
 			}
 		}
-		nodelist.max_members = count_nodes;
-		nodelist.nodes = nodes;
 
-		num_nodes = ioctl(cluster_sock, SIOCCLUSTER_GETMEMBERS, &nodelist);
-		if (num_nodes <= 0) {
+	status = cman_get_nodes(c_handle, count_nodes, &retnodes, nodes);
+	if (status < 0) {
 		        log_error("Unable to get node details");
 			exit(6);
 		}
 
-		/* Sanity check struct */
-		if (nodes[0].size != sizeof(struct cl_cluster_node)) {
-			log_error
-			    ("sizeof(cl_cluster_node) does not match size returned from the kernel: aborting\n");
-			exit(10);
-		}
-
 		if (node_updown == NULL) {
 			node_updown =
 			    (int *) malloc(sizeof(int) *
@@ -371,7 +326,6 @@ static void get_members()
 			       sizeof(int) * max(num_nodes, max_updown_nodes));
 		}
 	}
-}
 
 /* Convert a node name to a CSID */
 static int _csid_from_name(char *csid, char *name)
@@ -379,8 +333,8 @@ static int _csid_from_name(char *csid, char *name)
 	int i;
 
 	for (i = 0; i < num_nodes; i++) {
-		if (strcmp(name, nodes[i].name) == 0) {
-			memcpy(csid, &nodes[i].node_id, CMAN_MAX_CSID_LEN);
+		if (strcmp(name, nodes[i].cn_name) == 0) {
+			memcpy(csid, &nodes[i].cn_nodeid, CMAN_MAX_CSID_LEN);
 			return 0;
 		}
 	}
@@ -393,8 +347,8 @@ static int _name_from_csid(char *csid, char *name)
 	int i;
 
 	for (i = 0; i < num_nodes; i++) {
-		if (memcmp(csid, &nodes[i].node_id, CMAN_MAX_CSID_LEN) == 0) {
-			strcpy(name, nodes[i].name);
+		if (memcmp(csid, &nodes[i].cn_nodeid, CMAN_MAX_CSID_LEN) == 0) {
+			strcpy(name, nodes[i].cn_name);
 			return 0;
 		}
 	}
@@ -409,8 +363,8 @@ static int name_from_nodeid(int nodeid, char *name)
 	int i;
 
 	for (i = 0; i < num_nodes; i++) {
-		if (nodeid == nodes[i].node_id) {
-			strcpy(name, nodes[i].name);
+		if (nodeid == nodes[i].cn_nodeid) {
+			strcpy(name, nodes[i].cn_name);
 			return 0;
 		}
 	}
@@ -431,7 +385,7 @@ static int nodeid_from_csid(char *csid)
 
 static int _is_quorate()
 {
-	return ioctl(cluster_sock, SIOCCLUSTER_ISQUORATE, 0);
+	return cman_is_quorate(c_handle);
 }
 
 static void sync_ast_routine(void *arg)

daemons/clvmd/clvmd-comms.h

@@ -56,15 +56,19 @@ struct cluster_ops *init_gulm_cluster(void);
 #endif
 
 #ifdef USE_CMAN
-#  include "cnxman-socket.h"
+#  include <netinet/in.h>
+#  include "libcman.h"
 #  define CMAN_MAX_CSID_LEN 4
 #  ifndef MAX_CSID_LEN
 #    define MAX_CSID_LEN CMAN_MAX_CSID_LEN
 #  endif
 #  undef MAX_CLUSTER_MEMBER_NAME_LEN
-#  define MAX_CLUSTER_MEMBER_NAME_LEN	CMAN_MAX_CLUSTER_MEMBER_NAME_LEN
+#  define MAX_CLUSTER_MEMBER_NAME_LEN   CMAN_MAX_NODENAME_LEN
+#  define CMAN_MAX_CLUSTER_MESSAGE 1500
+#  define CLUSTER_PORT_CLVMD 11
 struct cluster_ops *init_cman_cluster(void);
 #endif
 
 
+
 #endif

daemons/clvmd/clvmd.c

@@ -247,7 +247,7 @@ int main(int argc, char *argv[])
 	if ((clops = init_cman_cluster())) {
 		max_csid_len = CMAN_MAX_CSID_LEN;
 		max_cluster_message = CMAN_MAX_CLUSTER_MESSAGE;
-		max_cluster_member_name_len = CMAN_MAX_CLUSTER_MEMBER_NAME_LEN;
+		max_cluster_member_name_len = CMAN_MAX_NODENAME_LEN;
 		syslog(LOG_NOTICE, "Cluster LVM daemon started - connected to CMAN");
 	}
 #endif
@@ -509,6 +509,7 @@ static void main_loop(int local_sock, int cmd_timeout)
 		int quorate = clops->is_quorate();
 
 		/* Wait on the cluster FD and all local sockets/pipes */
+		local_client_head.fd = clops->get_main_cluster_fd();
 		FD_ZERO(&in);
 		for (thisfd = &local_client_head; thisfd != NULL;
 		     thisfd = thisfd->next) {

daemons/clvmd/lvm-functions.c

@@ -303,6 +303,9 @@ int do_lock_lv(unsigned char command, unsigned char lock_flags, char *resource)
 		}
 	}
 
+	if (lock_flags & LCK_PARTIAL_MODE)
+		init_partial(1);
+
 	switch (command) {
 	case LCK_LV_EXCLUSIVE:
 		status = do_activate_lv(resource, lock_flags, LKM_EXMODE);
@@ -331,6 +334,9 @@ int do_lock_lv(unsigned char command, unsigned char lock_flags, char *resource)
 		break;
 	}
 
+	if (lock_flags & LCK_PARTIAL_MODE)
+		init_partial(0);
+
 	/* clean the pool for another command */
 	dm_pool_empty(cmd->mem);
 

daemons/dmeventd/.exported_symbols

@@ -1,3 +1,5 @@
-process_event
-register_device
-unregister_device
+dm_event_register
+dm_event_unregister
+dm_event_get_registered_device
+dm_event_set_timeout
+dm_event_get_timeout

daemons/dmeventd/Makefile.in

@@ -1,14 +1,13 @@
 #
-# Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.
-# Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+# Copyright (C) 2005 Red Hat, Inc. All rights reserved.
 #
 # This file is part of the device-mapper userspace tools.
 #
 # This copyrighted material is made available to anyone wishing to use,
 # modify, copy, or redistribute it subject to the terms and conditions
-# of the GNU General Public License v.2.
+# of the GNU Lesser General Public License v.2.1.
 #
-# You should have received a copy of the GNU General Public License
+# You should have received a copy of the GNU Lesser General Public License
 # along with this program; if not, write to the Free Software Foundation,
 # Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 
@@ -16,36 +15,41 @@ srcdir = @srcdir@
 top_srcdir = @top_srcdir@
 VPATH = @srcdir@
 
-TARGETS = dmevent dmeventd
-INSTALL_TYPE = install_dynamic
+SOURCES = libdevmapper-event.c \
+	  dmeventd.c
 
-SOURCES = noop.c
-CLEAN_TARGETS = dmevent.o dmeventd.o
+LIB_STATIC = libdevmapper-event.a
 
 ifeq ("@LIB_SUFFIX@","dylib")
-  LIB_SHARED = libdmeventdnoop.dylib
+  LIB_SHARED = libdevmapper-event.dylib
 else
-  LIB_SHARED = libdmeventdnoop.so
+  LIB_SHARED = libdevmapper-event.so
 endif
- 
-LDFLAGS += -ldl -ldevmapper -lmultilog
 
 include ../make.tmpl
 
-libdmeventdnoop.so: noop.o
+CLDFLAGS += -ldl -ldevmapper -lpthread
 
-dmevent: dmevent.o $(interfacedir)/libdevmapper.$(LIB_SUFFIX) $(top_srcdir)/lib/event/libdmevent.$(LIB_SUFFIX)
-	$(CC) -o $@ dmevent.o $(LDFLAGS) \
-	      -L$(interfacedir) -L$(DESTDIR)/lib -L$(top_srcdir)/lib/event -L$(top_srcdir)/multilog $(LIBS)
+.PHONY: install_dynamic install_static
 
-dmeventd: dmeventd.o $(interfacedir)/libdevmapper.$(LIB_SUFFIX) $(top_srcdir)/lib/event/libdmevent.$(LIB_SUFFIX)
-	$(CC) -o $@ dmeventd.o $(LDFLAGS) \
-	      -L$(interfacedir) -L$(DESTDIR)/lib -L$(top_srcdir)/lib/event -L$(top_srcdir)/multilog -lpthread -ldmevent $(LIBS)
+INSTALL_TYPE = install_dynamic
+
+ifeq ("@STATIC_LINK@", "yes")
+  INSTALL_TYPE += install_static
+endif
 
 install: $(INSTALL_TYPE)
 
-.PHONY: install_dynamic
+install_dynamic: libdevmapper-event.$(LIB_SUFFIX)
+	$(INSTALL) -D $(OWNER) $(GROUP) -m 555 $(STRIP) $< \
+		$(libdir)/libdevmapper-event.$(LIB_SUFFIX).$(LIB_VERSION)
+	$(LN_S) -f libdevmapper-event.$(LIB_SUFFIX).$(LIB_VERSION) \
+		$(libdir)/libdevmapper-event.$(LIB_SUFFIX)
+	$(INSTALL) -D $(OWNER) $(GROUP) -m 444 libdevmapper-event.h \
+		$(includedir)/libdevmapper-event.h
 
-install_dynamic: dmeventd
-	$(INSTALL) -D $(OWNER) $(GROUP) -m 555 $(STRIP) dmeventd $(sbindir)/dmeventd
+install_static: libdevmapper-event.a
+	$(INSTALL) -D $(OWNER) $(GROUP) -m 555 $(STRIP) $< \
+		$(libdir)/libdevmapper-event.a.$(LIB_VERSION)
+	$(LN_S) -f libdevmapper-event.a.$(LIB_VERSION) $(libdir)/libdevmapper-event.a
 

daemons/dmeventd/dmevent.c

@@ -1,240 +0,0 @@
-/*
- * Copyright (C) 2005 Red Hat, Inc. All rights reserved.
- *
- * This file is part of the device-mapper userspace tools.
- *
- * This copyrighted material is made available to anyone wishing to use,
- * modify, copy, or redistribute it subject to the terms and conditions
- * of the GNU Lesser General Public License v.2.1.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with this program; if not, write to the Free Software Foundation,
- * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
- */
-
-#include "libdevmapper.h"
-#include "libdm-event.h"
-#include "libmultilog.h"
-
-#include <errno.h>
-#include <fcntl.h>
-#include <stdio.h>
-#include <stdint.h>
-#include <stdlib.h>
-#include <string.h>
-#include <sys/file.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <unistd.h>
-#include <dlfcn.h>
-
-static enum event_type events = ALL_ERRORS; /* All until we can distinguish. */
-static char default_dso_name[] = "noop";  /* default DSO is noop */
-static int default_reg = 1;		 /* default action is register */
-static uint32_t timeout;
-
-struct event_ops {
-int (*dm_register_for_event)(char *dso_name, char *device,
-			     enum event_type event_types);
-int (*dm_unregister_for_event)(char *dso_name, char *device,
-			       enum event_type event_types);
-int (*dm_get_registered_device)(char **dso_name, char **device,
-				enum event_type *event_types, int next);
-int (*dm_set_event_timeout)(char *device, uint32_t time);
-int (*dm_get_event_timeout)(char *device, uint32_t *time);
-};
-
-/* Display help. */
-static void print_usage(char *name)
-{
-	char *cmd = strrchr(name, '/');
-
-	cmd = cmd ? cmd + 1 : name;
-	printf("Usage::\n"
-	       "%s [options] <device>\n"
-	       "\n"
-	       "Options:\n"
-	       "  -d <dso>           Specify the DSO to use.\n"
-	       "  -h                 Print this usage.\n"
-	       "  -l                 List registered devices.\n"
-	       "  -r                 Register for event (default).\n"
-	       "  -t <timeout>       (un)register for timeout event.\n"
-	       "  -u                 Unregister for event.\n"
-	       "\n", cmd);
-}
-
-/* Parse command line arguments. */
-static int parse_argv(int argc, char **argv, char **dso_name_arg,
-		      char **device_arg, int *reg, int *list)
-{
-	int c;
-	const char *options = "d:hlrt:u";
-
-	while ((c = getopt(argc, argv, options)) != -1) {
-		switch (c) {
-		case 'd':
-			*dso_name_arg = optarg;
-			break;
-		case 'h':
-			print_usage(argv[0]);
-			exit(EXIT_SUCCESS);
-		case 'l':
-			*list = 1;
-			break;
-		case 'r':
-			*reg = 1;
-			break;
-		case 't':
-			events = TIMEOUT;
-			if (sscanf(optarg, "%"SCNu32, &timeout) != 1){
-				fprintf(stderr, "invalid timeout '%s'\n",
-					optarg);
-				timeout = 0;
-			}
-			break;
-		case 'u':
-			*reg = 0;
-			break;
-		default:
-			fprintf(stderr, "Unknown option '%c'.\n"
-				"Try '-h' for help.\n", c);
-
-			return 0;
-		}
-	}
-
-	if (optind >= argc) {
-		if (!*list) {
-			fprintf(stderr, "You need to specify a device.\n");
-			return 0;
-		}
-	} else 
-		*device_arg = argv[optind];
-
-	return 1;
-}
-
-
-static int lookup_symbol(void *dl, void **symbol, const char *name)
-{
-	if ((*symbol = dlsym(dl, name)))
-		return 1;
-
-	fprintf(stderr, "error looking up %s symbol: %s\n", name, dlerror());
-
-	return 0;
-}
-
-static int lookup_symbols(void *dl, struct event_ops *e)
-{
-	return lookup_symbol(dl, (void *) &e->dm_register_for_event,
-			     "dm_register_for_event") &&
-	       lookup_symbol(dl, (void *) &e->dm_unregister_for_event,
-			     "dm_unregister_for_event") &&
-	       lookup_symbol(dl, (void *) &e->dm_get_registered_device,
-			     "dm_get_registered_device") &&
-	       lookup_symbol(dl, (void *) &e->dm_set_event_timeout,
-			     "dm_set_event_timeout") &&
-	       lookup_symbol(dl, (void *) &e->dm_get_event_timeout,
-			     "dm_get_event_timeout");
-}
-
-int main(int argc, char **argv)
-{
-	void *dl;
-	struct event_ops e;
-	int list = 0, next = 0, ret, reg = default_reg;
-	char *device, *device_arg = NULL, *dso_name, *dso_name_arg = NULL;
-
-	if (!parse_argv(argc, argv, &dso_name_arg, &device_arg, &reg, &list))
-		exit(EXIT_FAILURE);
-
-	if (device_arg) {
-		if (!(device = strdup(device_arg)))
-			exit(EXIT_FAILURE);
-	} else
-		device = NULL;
-
-	if (dso_name_arg) {
-		if (!(dso_name = strdup(dso_name_arg)))
-			exit(EXIT_FAILURE);
-	} else {
-		if (!(dso_name = strdup(default_dso_name)))
-			exit(EXIT_FAILURE);
-	}
-
-	/* FIXME: use -v/-q options to set this */
-	multilog_add_type(standard, NULL);
-	multilog_init_verbose(standard, _LOG_DEBUG);
-
-	if (!(dl = dlopen("libdmevent.so", RTLD_NOW))){
-		fprintf(stderr, "Cannot dlopen libdmevent.so: %s\n", dlerror());
-		goto out;
-	}
-	if (!(lookup_symbols(dl, &e)))
-		goto out;
-	if (list) {
-		while (1) {
-			if ((ret= e.dm_get_registered_device(&dso_name,
-							     &device,
-							     &events, next)))
-				break;
-			printf("%s %s 0x%x", dso_name, device, events);
-			if (events & TIMEOUT){
-				if ((ret = e.dm_get_event_timeout(device,
-							  	  &timeout))) {
-					ret = EXIT_FAILURE;
-					goto out;
-				}
-				printf(" %"PRIu32"\n", timeout);
-			} else
-				printf("\n");
-			if (device_arg)
-				break;
-
-			next = 1;
-		}
-
-		ret = (ret && device_arg) ? EXIT_FAILURE : EXIT_SUCCESS;
-		goto out;
-	}
-
-	if ((ret = reg ? e.dm_register_for_event(dso_name, device, events) :
-			 e.dm_unregister_for_event(dso_name, device, events))) {
-		fprintf(stderr, "Failed to %sregister %s: %s\n",
-			reg ? "": "un", device, strerror(-ret));
-		ret = EXIT_FAILURE;
-	} else {
-		if (reg && (events & TIMEOUT) &&
-		    ((ret = e.dm_set_event_timeout(device, timeout)))){
-			fprintf(stderr, "Failed to set timeout for %s: %s\n",
-				device, strerror(-ret));
-			ret = EXIT_FAILURE;
-		} else {
-			printf("%s %sregistered successfully.\n",
-			       device, reg ? "" : "un");
-			ret = EXIT_SUCCESS;
-		}
-	}
-
-   out:
-	multilog_del_type(standard);
-
-	if (device)
-		free(device);
-	if (dso_name)
-		free(dso_name);
-
-	exit(ret);
-}
-
-/*
- * Overrides for Emacs so that we follow Linus's tabbing style.
- * Emacs will notice this stuff at the end of the file and automatically
- * adjust the settings for this buffer only.  This must remain at the end
- * of the file.
- * ---------------------------------------------------------------------------
- * Local variables:
- * c-file-style: "linux"
- * End:
- */

daemons/dmeventd/dmeventd.h

@@ -0,0 +1,13 @@
+#ifndef __DMEVENTD_DOT_H__
+#define __DMEVENTD_DOT_H__
+
+#define EXIT_LOCKFILE_INUSE      2
+#define EXIT_DESC_CLOSE_FAILURE  3
+#define EXIT_OPEN_PID_FAILURE    4
+#define EXIT_FIFO_FAILURE        5
+#define EXIT_CHDIR_FAILURE       6
+
+void dmeventd(void)
+    __attribute((noreturn));
+
+#endif /* __DMEVENTD_DOT_H__ */

daemons/dmeventd/libdevmapper-event.c

@@ -0,0 +1,510 @@
+ /*
+ * Copyright (C) 2005 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of the device-mapper userspace tools.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include "lib.h"
+#include "libdevmapper-event.h"
+//#include "libmultilog.h"
+#include "dmeventd.h"
+
+#include <errno.h>
+#include <fcntl.h>
+#include <stdio.h>
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/file.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <unistd.h>
+#include <signal.h>
+#include <sys/wait.h>
+
+/* Set by any of the external fxns the first time one of them is called */
+/* FIXME Unused */
+// static int _logging = 0;
+
+/* Fetch a string off src and duplicate it into *dest. */
+/* FIXME: move to seperate module to share with the daemon. */
+static const char delimiter = ' ';
+static char *fetch_string(char **src)
+{
+	char *p, *ret;
+
+	if ((p = strchr(*src, delimiter)))
+		*p = 0;
+
+	if ((ret = dm_strdup(*src)))
+		*src += strlen(ret) + 1;
+
+	if (p)
+		*p = delimiter;
+
+	return ret;
+}
+
+/* Parse a device message from the daemon. */
+static int parse_message(struct dm_event_daemon_message *msg, char **dso_name,
+			 char **device, enum dm_event_type *events)
+{
+	char *p = msg->msg;
+
+	if ((*dso_name = fetch_string(&p)) &&
+	    (*device   = fetch_string(&p))) {
+		*events = atoi(p);
+
+		return 0;
+	}
+
+	return -ENOMEM;
+}
+
+/*
+ * daemon_read
+ * @fifos
+ * @msg
+ *
+ * Read message from daemon.
+ *
+ * Returns: 0 on failure, 1 on success
+ */
+static int daemon_read(struct dm_event_fifos *fifos, struct dm_event_daemon_message *msg)
+{
+	unsigned bytes = 0;
+	int ret = 0;
+	fd_set fds;
+
+	memset(msg, 0, sizeof(*msg));
+	while (bytes < sizeof(*msg)) {
+		do {
+			/* Watch daemon read FIFO for input. */
+			FD_ZERO(&fds);
+			FD_SET(fifos->server, &fds);
+			ret = select(fifos->server+1, &fds, NULL, NULL, NULL);
+			if (ret < 0 && errno != EINTR) {
+				/* FIXME Log error */
+				return 0;
+			}
+		} while (ret < 1);
+
+		ret = read(fifos->server, msg, sizeof(*msg) - bytes);
+		if (ret < 0) {
+			if ((errno == EINTR) || (errno == EAGAIN))
+				continue;
+			else {
+				/* FIXME Log error */
+				return 0;
+			}
+		}
+
+		bytes += ret;
+	}
+
+	return bytes == sizeof(*msg);
+}
+
+/* Write message to daemon. */
+static int daemon_write(struct dm_event_fifos *fifos, struct dm_event_daemon_message *msg)
+{
+	unsigned bytes = 0;
+	int ret = 0;
+	fd_set fds;
+
+	while (bytes < sizeof(*msg)) {
+		do {
+			/* Watch daemon write FIFO to be ready for output. */
+			FD_ZERO(&fds);
+			FD_SET(fifos->client, &fds);
+			ret = select(fifos->client +1, NULL, &fds, NULL, NULL);
+			if ((ret < 0) && (errno != EINTR)) {
+				/* FIXME Log error */
+				return 0;
+			}
+		} while (ret < 1);
+
+		ret = write(fifos->client, msg, sizeof(*msg) - bytes);
+		if (ret < 0) {
+			if ((errno == EINTR) || (errno == EAGAIN))
+				continue;
+			else {
+				/* fixme: log error */
+				return 0;
+			}
+		}
+
+		bytes += ret;
+	}
+
+	return bytes == sizeof(*msg);
+}
+
+static int daemon_talk(struct dm_event_fifos *fifos, struct dm_event_daemon_message *msg,
+		       int cmd, char *dso_name, char *device,
+		       enum dm_event_type events, uint32_t timeout)
+{
+	memset(msg, 0, sizeof(*msg));
+
+	/*
+	 * Set command and pack the arguments
+	 * into ASCII message string.
+	 */
+	msg->opcode.cmd = cmd;
+
+	if (sizeof(msg->msg) <= (unsigned) snprintf(msg->msg, sizeof(msg->msg),
+						    "%s %s %u %"PRIu32,
+						    dso_name ? dso_name : "",
+						    device ? device : "",
+						    events, timeout)) {
+		stack;
+		return -ENAMETOOLONG;
+	}
+
+	/*
+	 * Write command and message to and
+	 * read status return code from daemon.
+	 */
+	if (!daemon_write(fifos, msg)) {
+		stack;
+		return -EIO;
+	}
+
+	if (!daemon_read(fifos, msg)) {
+		stack;
+		return -EIO;
+	}
+
+	return msg->opcode.status;
+}
+
+static volatile sig_atomic_t daemon_running = 0;
+
+static void daemon_running_signal_handler(int sig)
+{
+	daemon_running = 1;
+}
+
+/*
+ * start_daemon
+ *
+ * This function forks off a process (dmeventd) that will handle
+ * the events.  A signal must be returned from the child to
+ * indicate when it is ready to handle requests.  The parent
+ * (this function) returns 1 if there is a daemon running. 
+ *
+ * Returns: 1 on success, 0 otherwise
+ */
+static int start_daemon(void)
+{
+	int pid, ret=0;
+	void *old_hand;
+	sigset_t set, oset;
+
+	/* Must be able to acquire signal */
+	old_hand = signal(SIGUSR1, &daemon_running_signal_handler);
+	if (old_hand == SIG_ERR) {
+		log_error("Unable to setup signal handler.");
+		return 0;
+	}
+
+	if (sigemptyset(&set) || sigaddset(&set, SIGUSR1)) {
+		log_error("Unable to fill signal set.");
+	} else if (sigprocmask(SIG_UNBLOCK, &set, &oset)) {
+		log_error("Can't unblock the potentially blocked signal SIGUSR1");
+	}
+	
+	pid = fork();
+
+	if (pid < 0)
+		log_error("Unable to fork.\n");
+	else if (pid) { /* parent waits for child to get ready for requests */
+		int status;
+
+		/* FIXME Better way to do this? */
+		while (!waitpid(pid, &status, WNOHANG) && !daemon_running)
+			sleep(1);
+
+		if (daemon_running) {
+			ret = 1;
+		} else {
+			switch (WEXITSTATUS(status)) {
+			case EXIT_LOCKFILE_INUSE:
+				/*
+				 * Note, this is ok... we still have daemon
+				 * that we can communicate with...
+				 */
+				log_print("Starting dmeventd failed: "
+					  "dmeventd already running.\n");
+				ret = 1;
+				break;
+			default:
+				log_error("Unable to start dmeventd.\n");
+				break;
+			}
+		}
+		/*
+		 * Sometimes, a single process may perform multiple calls
+		 * that result in a daemon starting and exiting.  If we
+		 * don't reset this, the second (or greater) time the daemon
+		 * is started will cause this logic not to work.
+		 */
+		daemon_running = 0;
+	} else {
+		signal(SIGUSR1, SIG_IGN); /* don't care about error */
+
+		/* dmeventd function is responsible for properly setting **
+		** itself up.  It must never return - only exit.  This is**
+		** why it is followed by an EXIT_FAILURE                 */
+		dmeventd();
+		exit(EXIT_FAILURE);
+	}
+
+	/* FIXME What if old_hand is SIG_ERR? */
+	if (signal(SIGUSR1, old_hand) == SIG_ERR)
+		log_error("Unable to reset signal handler.");
+
+	if (sigprocmask(SIG_SETMASK, &oset, NULL))
+		log_error("Unable to reset signal mask.");
+
+	return ret;
+}
+
+/* Initialize client. */
+static int init_client(struct dm_event_fifos *fifos)
+{
+	/* FIXME Is fifo the most suitable method? */
+	/* FIXME Why not share comms/daemon code with something else e.g. multipath? */
+
+	/* init fifos */
+	memset(fifos, 0, sizeof(*fifos));
+	fifos->client_path = DM_EVENT_FIFO_CLIENT;
+	fifos->server_path = DM_EVENT_FIFO_SERVER;
+
+	/* FIXME The server should be responsible for these, not the client. */
+	/* Create fifos */
+	if (((mkfifo(fifos->client_path, 0600) == -1) && errno != EEXIST) ||
+	    ((mkfifo(fifos->server_path, 0600) == -1) && errno != EEXIST)) {
+		log_error("%s: Failed to create a fifo.\n", __func__);
+		return 0;
+	}
+
+	/* FIXME Warn/abort if perms are wrong - not something to fix silently. */
+	/* If they were already there, make sure permissions are ok. */
+	if (chmod(fifos->client_path, 0600)) {
+		log_error("Unable to set correct file permissions on %s",
+			fifos->client_path);
+		return 0;
+	}
+
+	if (chmod(fifos->server_path, 0600)) {
+		log_error("Unable to set correct file permissions on %s",
+			fifos->server_path);
+		return 0;
+	}
+
+	/*
+	 * Open the fifo used to read from the daemon.
+	 * Allows daemon to create its write fifo...
+	 */
+	if ((fifos->server = open(fifos->server_path, O_RDWR)) < 0) {
+		log_error("%s: open server fifo %s\n",
+			__func__, fifos->server_path);
+		stack;
+		return 0;
+	}
+
+	/* Lock out anyone else trying to do communication with the daemon. */
+	/* FIXME Why failure not retry?  How do multiple processes communicate? */
+	if (flock(fifos->server, LOCK_EX) < 0){
+		log_error("%s: flock %s\n", __func__, fifos->server_path);
+		close(fifos->server);
+		return 0;
+	}
+
+	/* Anyone listening?  If not, errno will be ENXIO */
+	while ((fifos->client = open(fifos->client_path,
+				  O_WRONLY | O_NONBLOCK)) < 0) {
+		if (errno != ENXIO) {
+			log_error("%s: Can't open client fifo %s: %s\n",
+				  __func__, fifos->client_path, strerror(errno));
+			close(fifos->server);
+			stack;
+			return 0;
+		}
+		
+		/* FIXME Unnecessary if daemon was started before calling this */
+		if (!start_daemon()) {
+			stack;
+			return 0;
+		}
+	}
+	
+	return 1;
+}
+
+static void dtr_client(struct dm_event_fifos *fifos)
+{
+	if (flock(fifos->server, LOCK_UN))
+		log_error("flock unlock %s\n", fifos->server_path);
+
+	close(fifos->client);
+	close(fifos->server);
+}
+
+/* Check, if a block device exists. */
+static int device_exists(char *device)
+{
+	struct stat st_buf;
+	char path2[PATH_MAX];
+
+	if (!device)
+		return 0;
+
+	if (device[0] == '/') /* absolute path */
+		return !stat(device, &st_buf) && S_ISBLK(st_buf.st_mode);
+
+	if (PATH_MAX <= snprintf(path2, PATH_MAX, "%s/%s", dm_dir(), device))
+		return 0;
+
+	return !stat(path2, &st_buf) && S_ISBLK(st_buf.st_mode);
+}
+
+/* Handle the event (de)registration call and return negative error codes. */
+static int do_event(int cmd, struct dm_event_daemon_message *msg,
+		    char *dso_name, char *device, enum dm_event_type events,
+		    uint32_t timeout)
+{
+	int ret;
+	struct dm_event_fifos fifos;
+
+	/* FIXME Start the daemon here if it's not running e.g. exclusive lock file */
+	/* FIXME Move this to separate 'dm_event_register_handler' - if no daemon here, fail */
+	if (!init_client(&fifos)) {
+		stack;
+		return -ESRCH;
+	}
+
+	/* FIXME Use separate 'dm_event_register_handler' function to pass in dso? */
+	ret = daemon_talk(&fifos, msg, cmd, dso_name, device, events, timeout);
+
+	/* what is the opposite of init? */
+	dtr_client(&fifos);
+	
+	return ret;
+}
+
+/* FIXME remove dso_name - use handle instead */
+/* FIXME Use uuid not path! */
+/* External library interface. */
+int dm_event_register(char *dso_name, char *device_path,
+		      enum dm_event_type events)
+{
+	int ret;
+	struct dm_event_daemon_message msg;
+
+	if (!device_exists(device_path)) {
+		log_error("%s: device not found", device_path);
+		return 0;
+	}
+
+	if ((ret = do_event(DM_EVENT_CMD_REGISTER_FOR_EVENT, &msg,
+			    dso_name, device_path, events, 0)) < 0) {
+		log_error("%s: event registration failed: %s", device_path,
+			  strerror(-ret));
+		return 0;
+	}
+
+	return 1;
+}
+
+int dm_event_unregister(char *dso_name, char *device_path,
+			enum dm_event_type events)
+{
+	int ret;
+	struct dm_event_daemon_message msg;
+
+	if (!device_exists(device_path)) {
+		log_error("%s: device not found", device_path);
+		return 0;
+	}
+
+	if ((ret = do_event(DM_EVENT_CMD_UNREGISTER_FOR_EVENT, &msg,
+			    dso_name, device_path, events, 0)) < 0) {
+		log_error("%s: event deregistration failed: %s", device_path,
+			  strerror(-ret));
+		return 0;
+	}
+
+	return 1;
+}
+
+int dm_event_get_registered_device(char **dso_name, char **device_path,
+			     enum dm_event_type *events, int next)
+{
+	int ret;
+	char *dso_name_arg = NULL, *device_path_arg = NULL;
+	struct dm_event_daemon_message msg;
+
+	if (!(ret = do_event(next ? DM_EVENT_CMD_GET_NEXT_REGISTERED_DEVICE :
+				    DM_EVENT_CMD_GET_REGISTERED_DEVICE,
+			     &msg, *dso_name, *device_path, *events, 0)))
+		ret = parse_message(&msg, &dso_name_arg, &device_path_arg,
+				    events);
+
+	if (next){
+		if (*dso_name)
+			dm_free(*dso_name);
+		if (*device_path)
+			dm_free(*device_path);
+		*dso_name = dso_name_arg;
+		*device_path = device_path_arg;
+	} else {
+		if (!(*dso_name))
+			*dso_name = dso_name_arg;
+		if (!(*device_path))
+			*device_path = device_path_arg;
+	}
+
+	return ret;
+}
+
+int dm_event_set_timeout(char *device_path, uint32_t timeout)
+{
+	struct dm_event_daemon_message msg;
+
+	if (!device_exists(device_path))
+		return -ENODEV;
+	return do_event(DM_EVENT_CMD_SET_TIMEOUT, &msg,
+			NULL, device_path, 0, timeout);
+}
+
+int dm_event_get_timeout(char *device_path, uint32_t *timeout)
+{
+	int ret;
+	struct dm_event_daemon_message msg;
+
+	if (!device_exists(device_path))
+		return -ENODEV;
+	if (!(ret = do_event(DM_EVENT_CMD_GET_TIMEOUT, &msg, NULL, device_path, 0, 0)))
+		*timeout = atoi(msg.msg);
+	return ret;
+}
+/*
+ * Overrides for Emacs so that we follow Linus's tabbing style.
+ * Emacs will notice this stuff at the end of the file and automatically
+ * adjust the settings for this buffer only.  This must remain at the end
+ * of the file.
+ * ---------------------------------------------------------------------------
+ * Local variables:
+ * c-file-style: "linux"
+ * End:
+ */

daemons/dmeventd/libdevmapper-event.h

@@ -0,0 +1,108 @@
+/*
+ * Copyright (C) 2005 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of the device-mapper userspace tools.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+/*
+ * Note that this file is released only as part of a technology preview
+ * and its contents may change in future updates in ways that do not
+ * preserve compatibility.
+ */
+
+#ifndef LIB_DMEVENT_H
+#define LIB_DMEVENT_H
+
+#include <stdint.h>
+
+/* FIXME This stuff must be configurable. */
+
+#define	DM_EVENT_DAEMON		"/sbin/dmeventd"
+#define DM_EVENT_LOCKFILE	"/var/lock/dmeventd"
+#define	DM_EVENT_FIFO_CLIENT	"/var/run/dmeventd-client"
+#define	DM_EVENT_FIFO_SERVER	"/var/run/dmeventd-server"
+#define DM_EVENT_PIDFILE	"/var/run/dmeventd.pid"
+
+#define DM_EVENT_DEFAULT_TIMEOUT 10
+
+/* Commands for the daemon passed in the message below. */
+enum dm_event_command {
+	DM_EVENT_CMD_ACTIVE = 1,
+	DM_EVENT_CMD_REGISTER_FOR_EVENT,
+	DM_EVENT_CMD_UNREGISTER_FOR_EVENT,
+	DM_EVENT_CMD_GET_REGISTERED_DEVICE,
+	DM_EVENT_CMD_GET_NEXT_REGISTERED_DEVICE,
+	DM_EVENT_CMD_SET_TIMEOUT,
+	DM_EVENT_CMD_GET_TIMEOUT,
+};
+
+/* Message passed between client and daemon. */
+struct dm_event_daemon_message {
+	union {
+		unsigned int cmd;	/* FIXME Use fixed size. */
+		int	 status;	/* FIXME Use fixed size. */
+	} opcode;
+	char msg[252];		/* FIXME Why is this 252 ? */
+} __attribute__((packed));	/* FIXME Do this properly! */
+
+/* FIXME Is this meant to be exported?  I can't see where the interface uses it. */
+/* Fifos for client/daemon communication. */
+struct dm_event_fifos {
+	int client;
+	int server;
+	const char *client_path;
+	const char *server_path;
+};
+
+/* Event type definitions. */
+/* FIXME Use masks to separate the types and provide for extension. */
+enum dm_event_type {
+	DM_EVENT_SINGLE		= 0x01, /* Report multiple errors just once. */
+	DM_EVENT_MULTI		= 0x02, /* Report all of them. */
+
+	DM_EVENT_SECTOR_ERROR	= 0x04, /* Failure on a particular sector. */
+	DM_EVENT_DEVICE_ERROR	= 0x08, /* Device failure. */
+	DM_EVENT_PATH_ERROR	= 0x10, /* Failure on an io path. */
+	DM_EVENT_ADAPTOR_ERROR	= 0x20, /* Failure off a host adaptor. */
+
+	DM_EVENT_SYNC_STATUS	= 0x40, /* Mirror synchronization completed/failed. */
+	DM_EVENT_TIMEOUT	= 0x80, /* Timeout has occured */
+};
+
+/* FIXME Use a mask. */
+#define	DM_EVENT_ALL_ERRORS (DM_EVENT_SECTOR_ERROR | DM_EVENT_DEVICE_ERROR | \
+			     DM_EVENT_PATH_ERROR | DM_EVENT_ADAPTOR_ERROR)
+
+/* Prototypes for event lib interface. */
+
+/* FIXME Replace device with standard name/uuid/devno choice */
+/* Interface changes: 
+   First register a handler, passing in a unique ref for the device. */
+//  int dm_event_register_handler(const char *dso_name, const char *device);
+//  int dm_event_register(const char *dso_name, const char *name, const char *uuid, uint32_t major, uint32_t minor, enum dm_event_type events);
+/* Or (better?) add to task structure and use existing functions - run a task to register/unregister events - we may need to run task withe that with the new event mechanism anyway, then the dso calls just hook in.
+*/
+ 
+/* FIXME Missing consts? */
+int dm_event_register(char *dso_name, char *device, enum dm_event_type events);
+int dm_event_unregister(char *dso_name, char *device,
+			enum dm_event_type events);
+int dm_event_get_registered_device(char **dso_name, char **device,
+				   enum dm_event_type *events, int next);
+int dm_event_set_timeout(char *device, uint32_t timeout);
+int dm_event_get_timeout(char *device, uint32_t *timeout);
+
+/* Prototypes for DSO interface. */
+void process_event(const char *device, enum dm_event_type event);
+int register_device(const char *device);
+int unregister_device(const char *device);
+
+#endif

daemons/dmeventd/mktestdevices

@@ -1,12 +0,0 @@
-#!/bin/sh
-#
-# Create test devices for dmeventd
-#
-
-trap "rm -f /tmp/tmp.$$" 0 1 2 3 15
-
-echo "0 1024 zero" > /tmp/tmp.$$
-dmsetup create test /tmp/tmp.$$
-dmsetup create test1 /tmp/tmp.$$
-
-kill -15 $$

daemons/dmeventd/noop.c

@@ -1,39 +0,0 @@
-/*
- * Copyright (C) 2005 Red Hat, Inc. All rights reserved.
- *
- * This file is part of the device-mapper userspace tools.
- *
- * This copyrighted material is made available to anyone wishing to use,
- * modify, copy, or redistribute it subject to the terms and conditions
- * of the GNU Lesser General Public License v.2.1.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with this program; if not, write to the Free Software Foundation,
- * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
- */
-
-#include "libdm-event.h"
-#include "libmultilog.h"
-
-
-void process_event(char *device, enum event_type event)
-{
-	log_err("[%s] %s(%d) - Device: %s, Event %d\n",
-		__FILE__, __func__, __LINE__, device, event);
-}
-
-int register_device(char *device)
-{
-	log_err("[%s] %s(%d) - Device: %s\n",
-		__FILE__, __func__, __LINE__, device);
-
-	return 1;
-}
-
-int unregister_device(char *device)
-{
-	log_err("[%s] %s(%d) - Device: %s\n",
-		__FILE__, __func__, __LINE__, device);
-
-	return 1;
-}

daemons/dmeventd/plugins/Makefile.in

@@ -0,0 +1,22 @@
+#
+# Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.
+# Copyright (C) 2004-2005 Red Hat, Inc. All rights reserved.
+#
+# This file is part of the LVM2.
+#
+# This copyrighted material is made available to anyone wishing to use,
+# modify, copy, or redistribute it subject to the terms and conditions
+# of the GNU General Public License v.2.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software Foundation,
+# Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+
+srcdir = @srcdir@
+top_srcdir = @top_srcdir@
+VPATH = @srcdir@
+
+SUBDIRS += mirror
+
+include $(top_srcdir)/make.tmpl
+

daemons/dmeventd/plugins/mirror/.exported_symbols

@@ -0,0 +1,3 @@
+process_event
+register_device
+unregister_device

daemons/dmeventd/plugins/mirror/Makefile.in

@@ -0,0 +1,36 @@
+#
+# Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.
+# Copyright (C) 2004-2005 Red Hat, Inc. All rights reserved.
+#
+# This file is part of the LVM2.
+#
+# This copyrighted material is made available to anyone wishing to use,
+# modify, copy, or redistribute it subject to the terms and conditions
+# of the GNU General Public License v.2.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software Foundation,
+# Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+
+srcdir = @srcdir@
+top_srcdir = @top_srcdir@
+VPATH = @srcdir@
+
+INCLUDES += -I${top_srcdir}/tools
+CLDFLAGS += -L${top_srcdir}/tools -ldevmapper -llvm2cmd
+
+SOURCES = dmeventd_mirror.c
+
+ifeq ("@LIB_SUFFIX@","dylib")
+  LIB_SHARED = libdevmapper-event-lvm2mirror.dylib
+else
+  LIB_SHARED = libdevmapper-event-lvm2mirror.so
+endif
+
+include $(top_srcdir)/make.tmpl
+
+install: libdevmapper-event-lvm2mirror.$(LIB_SUFFIX)
+	$(INSTALL) -D $(OWNER) $(GROUP) -m 555 $(STRIP) $< \
+		$(libdir)/$<.$(LIB_VERSION)
+	$(LN_S) -f $<.$(LIB_VERSION) $(libdir)/$<
+

daemons/dmeventd/plugins/mirror/dmeventd_mirror.c

@@ -0,0 +1,246 @@
+/*
+ * Copyright (C) 2005 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include "libdevmapper.h"
+#include "libdevmapper-event.h"
+#include "lvm2cmd.h"
+#include "lvm-string.h"
+
+#include <errno.h>
+#include <signal.h>
+#include <string.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <pthread.h>
+#include <unistd.h>
+
+#include <syslog.h> /* FIXME Replace syslog with multilog */
+
+#define ME_IGNORE    0
+#define ME_INSYNC    1
+#define ME_FAILURE   2
+
+/* FIXME: We may need to lock around operations to these */
+static int register_count = 0;
+static struct dm_pool *mem_pool = NULL;
+
+static int _get_mirror_event(char *params)
+{
+	int i, rtn = ME_INSYNC;
+	int max_args = 30;  /* should support at least 8-way mirrors */
+	char *args[max_args];
+	char *dev_status_str;
+	char *log_status_str;
+	char *sync_str;
+	char *p;
+	int log_argc, num_devs, num_failures=0;
+
+	if (max_args <= split_words(params, max_args, args)) {
+		syslog(LOG_ERR, "Unable to split mirror parameters: Arg list too long");
+		return -E2BIG;
+	}
+
+	/*
+	 * Unused:  0 409600 mirror
+	 * Used  :  2 253:4 253:5 400/400 1 AA 3 cluster 253:3 A
+	*/
+	num_devs = atoi(args[0]);
+	dev_status_str = args[3 + num_devs];
+	log_argc = atoi(args[4 + num_devs]);
+	log_status_str = args[4 + num_devs + log_argc];
+	sync_str = args[1 + num_devs];
+
+	/* Check for bad mirror devices */
+	for (i = 0; i < num_devs; i++) {
+		if (dev_status_str[i] == 'D') {
+			syslog(LOG_ERR, "Mirror device, %s, has failed.\n", args[i+1]);
+			num_failures++;
+		}
+	}
+
+	/* Check for bad log device */
+	if (log_status_str[0] == 'D') {
+		syslog(LOG_ERR, "Log device, %s, has failed.\n",
+		       args[3 + num_devs + log_argc]);
+		num_failures++;
+	}
+
+	if (num_failures) {
+		rtn = ME_FAILURE;
+		goto out;
+	}
+
+	p = strstr(sync_str, "/");
+	if (p) {
+		p[0] = '\0';
+		if (strcmp(sync_str, p+1))
+			rtn = ME_IGNORE;
+		p[0] = '/';
+	} else {
+		/*
+		 * How the hell did we get this?
+		 * Might mean all our parameters are screwed.
+		 */
+		syslog(LOG_ERR, "Unable to parse sync string.");
+		rtn = ME_IGNORE;
+	}
+ out:
+	return rtn;
+}
+
+static void _temporary_log_fn(int level, const char *file, int line, const char *format)
+{
+	return;
+	syslog(LOG_DEBUG, "%s", format);
+}
+
+static int _remove_failed_devices(const char *device)
+{
+	int r;
+	void *handle;
+	int cmd_size = 256;	/* FIXME Use system restriction */
+	char cmd_str[cmd_size];
+	char *vg = NULL, *lv = NULL, *layer = NULL;
+
+	if (strlen(device) > 200)
+		return -ENAMETOOLONG;
+
+	if (!split_dm_name(mem_pool, device, &vg, &lv, &layer)) {
+		syslog(LOG_ERR, "Unable to determine VG name from %s",
+		       device);
+		return -ENOMEM;
+	}
+
+	/* FIXME Is any sanity-checking required on %s? */
+	if (cmd_size <= snprintf(cmd_str, cmd_size, "vgreduce --removemissing %s", vg)) {
+		/* this error should be caught above, but doesn't hurt to check again */
+		syslog(LOG_ERR, "Unable to form LVM command: Device name too long");
+		dm_pool_empty(mem_pool);  /* FIXME: not safe with multiple threads */
+		return -ENAMETOOLONG;
+	}
+
+	lvm2_log_fn(_temporary_log_fn);
+	handle = lvm2_init();
+	lvm2_log_level(handle, 1);
+	r = lvm2_run(handle, cmd_str);
+
+	dm_pool_empty(mem_pool);  /* FIXME: not safe with multiple threads */
+	return (r == 1)? 0: -1;
+}
+
+void process_event(const char *device, enum dm_event_type event)
+{
+	struct dm_task *dmt;
+	void *next = NULL;
+	uint64_t start, length;
+	char *target_type = NULL;
+	char *params;
+
+	/* FIXME Move inside libdevmapper */
+	if (!(dmt = dm_task_create(DM_DEVICE_STATUS))) {
+		syslog(LOG_ERR, "Unable to create dm_task.\n");
+		goto fail;
+	}
+
+	if (!dm_task_set_name(dmt, device)) {
+		syslog(LOG_ERR, "Unable to set device name.\n");
+		goto fail;
+	}
+
+	if (!dm_task_run(dmt)) {
+		syslog(LOG_ERR, "Unable to run task.\n");
+		goto fail;
+	}
+
+	do {
+		next = dm_get_next_target(dmt, next, &start, &length,
+					  &target_type, &params);
+
+		if (strcmp(target_type, "mirror")) {
+			syslog(LOG_INFO, "%s has unmirrored portion.\n", device);
+			continue;
+		}
+
+		switch(_get_mirror_event(params)) {
+		case ME_INSYNC:
+			/* FIXME: all we really know is that this
+			   _part_ of the device is in sync
+			   Also, this is not an error
+			*/
+			syslog(LOG_NOTICE, "%s is now in-sync\n", device);
+			break;
+		case ME_FAILURE:
+			syslog(LOG_ERR, "Device failure in %s\n", device);
+			if (_remove_failed_devices(device))
+				syslog(LOG_ERR, "Failed to remove faulty devices in %s\n",
+				       device);
+			/* Should check before warning user that device is now linear
+			else
+				syslog(LOG_NOTICE, "%s is now a linear device.\n",
+					device);
+			*/
+			break;
+		case ME_IGNORE:
+			break;
+		default:
+			syslog(LOG_INFO, "Unknown event received.\n");
+		}
+	} while (next);
+
+ fail:
+	if (dmt)
+		dm_task_destroy(dmt);
+}
+
+int register_device(const char *device)
+{
+	syslog(LOG_INFO, "Monitoring %s for events\n", device);
+
+	/*
+	 * Need some space for allocations.  1024 should be more
+	 * than enough for what we need (device mapper name splitting)
+	 */
+	if (!mem_pool)
+		mem_pool = dm_pool_create("mirror_dso", 1024);
+
+	if (!mem_pool)
+		return 0;
+
+	register_count++;
+
+        return 1;
+}
+
+int unregister_device(const char *device)
+{
+        syslog(LOG_INFO, "Stopped monitoring %s for events\n", device);
+
+	if (!(--register_count)) {
+		dm_pool_destroy(mem_pool);
+		mem_pool = NULL;
+	}
+
+        return 1;
+}
+
+/*
+ * Overrides for Emacs so that we follow Linus's tabbing style.
+ * Emacs will notice this stuff at the end of the file and automatically
+ * adjust the settings for this buffer only.  This must remain at the end
+ * of the file.
+ * ---------------------------------------------------------------------------
+ * Local variables:
+ * c-file-style: "linux"
+ * End:
+ */

dmeventd/Makefile.in

@@ -0,0 +1,22 @@
+#
+# Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.
+# Copyright (C) 2004-2005 Red Hat, Inc. All rights reserved.
+#
+# This file is part of the LVM2.
+#
+# This copyrighted material is made available to anyone wishing to use,
+# modify, copy, or redistribute it subject to the terms and conditions
+# of the GNU General Public License v.2.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software Foundation,
+# Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+
+srcdir = @srcdir@
+top_srcdir = @top_srcdir@
+VPATH = @srcdir@
+
+SUBDIRS += mirror
+
+include $(top_srcdir)/make.tmpl
+

dmeventd/mirror/.exported_symbols

@@ -0,0 +1,3 @@
+process_event
+register_device
+unregister_device

dmeventd/mirror/Makefile.in

@@ -0,0 +1,36 @@
+#
+# Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.
+# Copyright (C) 2004-2005 Red Hat, Inc. All rights reserved.
+#
+# This file is part of the LVM2.
+#
+# This copyrighted material is made available to anyone wishing to use,
+# modify, copy, or redistribute it subject to the terms and conditions
+# of the GNU General Public License v.2.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software Foundation,
+# Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+
+srcdir = @srcdir@
+top_srcdir = @top_srcdir@
+VPATH = @srcdir@
+
+INCLUDES += -I${top_srcdir}/tools
+CLDFLAGS += -L${top_srcdir}/tools -ldevmapper -llvm2cmd
+
+SOURCES = dmeventd_mirror.c
+
+ifeq ("@LIB_SUFFIX@","dylib")
+  LIB_SHARED = libdevmapper-event-lvm2mirror.dylib
+else
+  LIB_SHARED = libdevmapper-event-lvm2mirror.so
+endif
+
+include $(top_srcdir)/make.tmpl
+
+install: libdevmapper-event-lvm2mirror.$(LIB_SUFFIX)
+	$(INSTALL) -D $(OWNER) $(GROUP) -m 555 $(STRIP) $< \
+		$(libdir)/$<.$(LIB_VERSION)
+	$(LN_S) -f $<.$(LIB_VERSION) $(libdir)/$<
+

dmeventd/mirror/dmeventd_mirror.c

@@ -0,0 +1,246 @@
+/*
+ * Copyright (C) 2005 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include "libdevmapper.h"
+#include "libdevmapper-event.h"
+#include "lvm2cmd.h"
+#include "lvm-string.h"
+
+#include <errno.h>
+#include <signal.h>
+#include <string.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <pthread.h>
+#include <unistd.h>
+
+#include <syslog.h> /* FIXME Replace syslog with multilog */
+
+#define ME_IGNORE    0
+#define ME_INSYNC    1
+#define ME_FAILURE   2
+
+/* FIXME: We may need to lock around operations to these */
+static int register_count = 0;
+static struct dm_pool *mem_pool = NULL;
+
+static int _get_mirror_event(char *params)
+{
+	int i, rtn = ME_INSYNC;
+	int max_args = 30;  /* should support at least 8-way mirrors */
+	char *args[max_args];
+	char *dev_status_str;
+	char *log_status_str;
+	char *sync_str;
+	char *p;
+	int log_argc, num_devs, num_failures=0;
+
+	if (max_args <= split_words(params, max_args, args)) {
+		syslog(LOG_ERR, "Unable to split mirror parameters: Arg list too long");
+		return -E2BIG;
+	}
+
+	/*
+	 * Unused:  0 409600 mirror
+	 * Used  :  2 253:4 253:5 400/400 1 AA 3 cluster 253:3 A
+	*/
+	num_devs = atoi(args[0]);
+	dev_status_str = args[3 + num_devs];
+	log_argc = atoi(args[4 + num_devs]);
+	log_status_str = args[4 + num_devs + log_argc];
+	sync_str = args[1 + num_devs];
+
+	/* Check for bad mirror devices */
+	for (i = 0; i < num_devs; i++) {
+		if (dev_status_str[i] == 'D') {
+			syslog(LOG_ERR, "Mirror device, %s, has failed.\n", args[i+1]);
+			num_failures++;
+		}
+	}
+
+	/* Check for bad log device */
+	if (log_status_str[0] == 'D') {
+		syslog(LOG_ERR, "Log device, %s, has failed.\n",
+		       args[3 + num_devs + log_argc]);
+		num_failures++;
+	}
+
+	if (num_failures) {
+		rtn = ME_FAILURE;
+		goto out;
+	}
+
+	p = strstr(sync_str, "/");
+	if (p) {
+		p[0] = '\0';
+		if (strcmp(sync_str, p+1))
+			rtn = ME_IGNORE;
+		p[0] = '/';
+	} else {
+		/*
+		 * How the hell did we get this?
+		 * Might mean all our parameters are screwed.
+		 */
+		syslog(LOG_ERR, "Unable to parse sync string.");
+		rtn = ME_IGNORE;
+	}
+ out:
+	return rtn;
+}
+
+static void _temporary_log_fn(int level, const char *file, int line, const char *format)
+{
+	return;
+	syslog(LOG_DEBUG, "%s", format);
+}
+
+static int _remove_failed_devices(const char *device)
+{
+	int r;
+	void *handle;
+	int cmd_size = 256;	/* FIXME Use system restriction */
+	char cmd_str[cmd_size];
+	char *vg = NULL, *lv = NULL, *layer = NULL;
+
+	if (strlen(device) > 200)
+		return -ENAMETOOLONG;
+
+	if (!split_dm_name(mem_pool, device, &vg, &lv, &layer)) {
+		syslog(LOG_ERR, "Unable to determine VG name from %s",
+		       device);
+		return -ENOMEM;
+	}
+
+	/* FIXME Is any sanity-checking required on %s? */
+	if (cmd_size <= snprintf(cmd_str, cmd_size, "vgreduce --removemissing %s", vg)) {
+		/* this error should be caught above, but doesn't hurt to check again */
+		syslog(LOG_ERR, "Unable to form LVM command: Device name too long");
+		dm_pool_empty(mem_pool);  /* FIXME: not safe with multiple threads */
+		return -ENAMETOOLONG;
+	}
+
+	lvm2_log_fn(_temporary_log_fn);
+	handle = lvm2_init();
+	lvm2_log_level(handle, 1);
+	r = lvm2_run(handle, cmd_str);
+
+	dm_pool_empty(mem_pool);  /* FIXME: not safe with multiple threads */
+	return (r == 1)? 0: -1;
+}
+
+void process_event(const char *device, enum dm_event_type event)
+{
+	struct dm_task *dmt;
+	void *next = NULL;
+	uint64_t start, length;
+	char *target_type = NULL;
+	char *params;
+
+	/* FIXME Move inside libdevmapper */
+	if (!(dmt = dm_task_create(DM_DEVICE_STATUS))) {
+		syslog(LOG_ERR, "Unable to create dm_task.\n");
+		goto fail;
+	}
+
+	if (!dm_task_set_name(dmt, device)) {
+		syslog(LOG_ERR, "Unable to set device name.\n");
+		goto fail;
+	}
+
+	if (!dm_task_run(dmt)) {
+		syslog(LOG_ERR, "Unable to run task.\n");
+		goto fail;
+	}
+
+	do {
+		next = dm_get_next_target(dmt, next, &start, &length,
+					  &target_type, &params);
+
+		if (strcmp(target_type, "mirror")) {
+			syslog(LOG_INFO, "%s has unmirrored portion.\n", device);
+			continue;
+		}
+
+		switch(_get_mirror_event(params)) {
+		case ME_INSYNC:
+			/* FIXME: all we really know is that this
+			   _part_ of the device is in sync
+			   Also, this is not an error
+			*/
+			syslog(LOG_NOTICE, "%s is now in-sync\n", device);
+			break;
+		case ME_FAILURE:
+			syslog(LOG_ERR, "Device failure in %s\n", device);
+			if (_remove_failed_devices(device))
+				syslog(LOG_ERR, "Failed to remove faulty devices in %s\n",
+				       device);
+			/* Should check before warning user that device is now linear
+			else
+				syslog(LOG_NOTICE, "%s is now a linear device.\n",
+					device);
+			*/
+			break;
+		case ME_IGNORE:
+			break;
+		default:
+			syslog(LOG_INFO, "Unknown event received.\n");
+		}
+	} while (next);
+
+ fail:
+	if (dmt)
+		dm_task_destroy(dmt);
+}
+
+int register_device(const char *device)
+{
+	syslog(LOG_INFO, "Monitoring %s for events\n", device);
+
+	/*
+	 * Need some space for allocations.  1024 should be more
+	 * than enough for what we need (device mapper name splitting)
+	 */
+	if (!mem_pool)
+		mem_pool = dm_pool_create("mirror_dso", 1024);
+
+	if (!mem_pool)
+		return 0;
+
+	register_count++;
+
+        return 1;
+}
+
+int unregister_device(const char *device)
+{
+        syslog(LOG_INFO, "Stopped monitoring %s for events\n", device);
+
+	if (!(--register_count)) {
+		dm_pool_destroy(mem_pool);
+		mem_pool = NULL;
+	}
+
+        return 1;
+}
+
+/*
+ * Overrides for Emacs so that we follow Linus's tabbing style.
+ * Emacs will notice this stuff at the end of the file and automatically
+ * adjust the settings for this buffer only.  This must remain at the end
+ * of the file.
+ * ---------------------------------------------------------------------------
+ * Local variables:
+ * c-file-style: "linux"
+ * End:
+ */

doc/example.conf

@@ -289,4 +289,9 @@ activation {
     # dirs = [ "/etc/lvm/metadata", "/mnt/disk2/lvm/metadata2" ]
 #}
 
+# Event daemon
+#
+#dmeventd {
+    # mirror_library = "libdevmapper-event-lvm2mirror.so"
+#}
 

lib/Makefile.in

@@ -133,6 +133,10 @@ ifeq ("@HAVE_LIBDL@", "yes")
 	misc/sharedlib.c
 endif
 
+ifeq ("@DMEVENTD@", "yes")
+  CLDFLAGS += -ldevmapper-event
+endif
+
 LIB_STATIC = liblvm.a
 
 $(SUBDIRS): $(LIB_STATIC)

lib/activate/activate.c

@@ -27,6 +27,7 @@
 #include "str_list.h"
 #include "config.h"
 #include "filter.h"
+#include "segtype.h"
 
 #include <limits.h>
 #include <fcntl.h>
@@ -75,6 +76,11 @@ int driver_version(char *version, size_t size)
 {
 	return 0;
 }
+int target_version(const char *target_name, uint32_t *maj,
+		   uint32_t *min, uint32_t *patchlevel)
+{
+	return 0;
+}
 int target_present(const char *target_name)
 {
 	return 0;
@@ -89,12 +95,12 @@ int lv_info_by_lvid(struct cmd_context *cmd, const char *lvid_s,
 {
 	return 0;
 }
-int lv_snapshot_percent(struct logical_volume *lv, float *percent)
+int lv_snapshot_percent(const struct logical_volume *lv, float *percent)
 {
 	return 0;
 }
-int lv_mirror_percent(struct logical_volume *lv, int wait, float *percent,
-		      uint32_t *event_nr)
+int lv_mirror_percent(struct cmd_context *cmd, struct logical_volume *lv,
+		      int wait, float *percent, uint32_t *event_nr)
 {
 	return 0;
 }
@@ -170,8 +176,8 @@ void set_activation(int act)
 		log_verbose("Activation enabled. Device-mapper kernel "
 			    "driver will be used.");
 	else
-		log_verbose("Activation disabled. No device-mapper "
-			    "interaction will be attempted.");
+		log_print("WARNING: Activation disabled. No device-mapper "
+			  "interaction will be attempted.");
 }
 
 int activation(void)
@@ -277,7 +283,8 @@ int driver_version(char *version, size_t size)
 	return dm_driver_version(version, size);
 }
 
-static int _target_present(const char *target_name)
+int target_version(const char *target_name, uint32_t *maj,
+		   uint32_t *min, uint32_t *patchlevel)
 {
 	int r = 0;
 	struct dm_task *dmt;
@@ -300,6 +307,9 @@ static int _target_present(const char *target_name)
 
 		if (!strcmp(target_name, target->name)) {
 			r = 1;
+			*maj = target->version[0];
+			*min = target->version[1];
+			*patchlevel = target->version[2];
 			goto out;
 		}
 
@@ -314,6 +324,7 @@ static int _target_present(const char *target_name)
 
 int target_present(const char *target_name, int use_modprobe)
 {
+	uint32_t maj, min, patchlevel;
 #ifdef MODPROBE_CMD
 	char module[128];
 #endif
@@ -323,7 +334,7 @@ int target_present(const char *target_name, int use_modprobe)
 
 #ifdef MODPROBE_CMD
 	if (use_modprobe) {
-		if (_target_present(target_name))
+		if (target_version(target_name, &maj, &min, &patchlevel))
 			return 1;
 
 		if (lvm_snprintf(module, sizeof(module), "dm-%s", target_name)
@@ -338,7 +349,7 @@ int target_present(const char *target_name, int use_modprobe)
 	}
 #endif
 
-	return _target_present(target_name);
+	return target_version(target_name, &maj, &min, &patchlevel);
 }
 
 /*
@@ -396,7 +407,7 @@ int lv_info_by_lvid(struct cmd_context *cmd, const char *lvid_s,
 /*
  * Returns 1 if percent set, else 0 on failure.
  */
-int lv_snapshot_percent(struct logical_volume *lv, float *percent)
+int lv_snapshot_percent(const struct logical_volume *lv, float *percent)
 {
 	int r;
 	struct dev_manager *dm;
@@ -416,8 +427,8 @@ int lv_snapshot_percent(struct logical_volume *lv, float *percent)
 }
 
 /* FIXME Merge with snapshot_percent */
-int lv_mirror_percent(struct cmd_context *cmd, struct logical_volume *lv, int wait, float *percent,
-		      uint32_t *event_nr)
+int lv_mirror_percent(struct cmd_context *cmd, struct logical_volume *lv,
+		      int wait, float *percent, uint32_t *event_nr)
 {
 	int r;
 	struct dev_manager *dm;
@@ -563,18 +574,53 @@ int lvs_in_vg_opened(struct volume_group *vg)
 	return count;
 }
 
+static int _register_dev_for_events(struct cmd_context *cmd,
+				    struct logical_volume *lv, int do_reg)
+{
+#ifdef DMEVENTD
+	struct list *tmp;
+	struct lv_segment *seg;
+	int (*reg) (struct dm_pool *mem, struct lv_segment *,
+		    struct config_tree *cft, int events);
+
+	list_iterate(tmp, &lv->segments) {
+		seg = list_item(tmp, struct lv_segment);
+
+		reg = NULL;
+
+		if (do_reg) {
+			if (seg->segtype->ops->target_register_events)
+				reg = seg->segtype->ops->target_register_events;
+		} else if (seg->segtype->ops->target_unregister_events)
+			reg = seg->segtype->ops->target_unregister_events;
+
+		if (reg)
+			/* FIXME specify events */
+			if (!reg(cmd->mem, seg, cmd->cft, 0)) {
+				stack;
+				return 0;
+			}
+	}
+
+#endif
+	return 1;
+}
+
 static int _lv_suspend(struct cmd_context *cmd, const char *lvid_s,
 		       int error_if_not_suspended)
 {
-	struct logical_volume *lv;
+	struct logical_volume *lv, *lv_pre;
 	struct lvinfo info;
 
 	if (!activation())
 		return 1;
 
+	if (!(lv = lv_from_lvid(cmd, lvid_s, 0)))
+		return_0;
+
 	/* Use precommitted metadata if present */
-	if (!(lv = lv_from_lvid(cmd, lvid_s, 1)))
-		return 0;
+	if (!(lv_pre = lv_from_lvid(cmd, lvid_s, 1)))
+		return_0;
 
 	if (test_mode()) {
 		_skip("Suspending '%s'.", lv->name);
@@ -588,13 +634,16 @@ static int _lv_suspend(struct cmd_context *cmd, const char *lvid_s,
 		return error_if_not_suspended ? 0 : 1;
 
 	/* If VG was precommitted, preload devices for the LV */
-	if ((lv->vg->status & PRECOMMITTED)) {
-		if (!_lv_preload(lv)) {
+	if ((lv_pre->vg->status & PRECOMMITTED)) {
+		if (!_lv_preload(lv_pre)) {
 			/* FIXME Revert preloading */
 			return_0;
 		}
 	}
 
+	if (!_register_dev_for_events(cmd, lv, 0))
+		stack;
+
 	memlock_inc();
 	if (!_lv_suspend_lv(lv)) {
 		memlock_dec();
@@ -645,6 +694,9 @@ static int _lv_resume(struct cmd_context *cmd, const char *lvid_s,
 	memlock_dec();
 	fs_unlock();
 
+	if (!_register_dev_for_events(cmd, lv, 1))
+		stack;
+
 	return 1;
 }
 
@@ -688,6 +740,9 @@ int lv_deactivate(struct cmd_context *cmd, const char *lvid_s)
 		return 0;
 	}
 
+	if (!_register_dev_for_events(cmd, lv, 0))
+		stack;
+
 	memlock_inc();
 	r = _lv_deactivate(lv);
 	memlock_dec();
@@ -758,6 +813,9 @@ static int _lv_activate(struct cmd_context *cmd, const char *lvid_s,
 	memlock_dec();
 	fs_unlock();
 
+	if (!_register_dev_for_events(cmd, lv, 1))
+		stack;
+
 	return r;
 }
 

lib/activate/activate.h

@@ -37,6 +37,8 @@ int library_version(char *version, size_t size);
 int lvm1_present(struct cmd_context *cmd);
 
 int target_present(const char *target_name, int use_modprobe);
+int target_version(const char *target_name, uint32_t *maj,
+                   uint32_t *min, uint32_t *patchlevel);
 
 void activation_exit(void);
 
@@ -68,9 +70,9 @@ int lv_activation_filter(struct cmd_context *cmd, const char *lvid_s,
 /*
  * Returns 1 if percent has been set, else 0.
  */
-int lv_snapshot_percent(struct logical_volume *lv, float *percent);
-int lv_mirror_percent(struct cmd_context *cmd, struct logical_volume *lv, int wait, float *percent,
-		      uint32_t *event_nr);
+int lv_snapshot_percent(const struct logical_volume *lv, float *percent);
+int lv_mirror_percent(struct cmd_context *cmd, struct logical_volume *lv,
+		      int wait, float *percent, uint32_t *event_nr);
 
 /*
  * Return number of LVs in the VG that are active.

lib/activate/dev_manager.c

@@ -432,7 +432,8 @@ void dev_manager_exit(void)
 }
 
 int dev_manager_snapshot_percent(struct dev_manager *dm,
-				 struct logical_volume *lv, float *percent)
+				 const struct logical_volume *lv,
+				 float *percent)
 {
 	char *name;
 	const char *dlid;
@@ -845,7 +846,6 @@ static int _add_new_lv_to_dtree(struct dev_manager *dm, struct dm_tree *dtree,
 	    dm_tree_node_get_context(dnode))
 		return 1;
 
-	/* FIXME How do we determine whether a pre-existing node need reloading or not? */
 	if (!(lvlayer = dm_pool_alloc(dm->mem, sizeof(*lvlayer)))) {
 		log_error("_add_new_lv_to_dtree: pool alloc failed for %s %s.", lv->name, layer);
 		return 0;
@@ -978,6 +978,8 @@ static int _tree_action(struct dev_manager *dm, struct logical_volume *lv, actio
 			goto_out;
 		break;
 	case SUSPEND:
+		if (!lv_is_origin(lv) && !lv_is_cow(lv))
+			dm_tree_skip_lockfs(root);
 		if (!dm_tree_suspend_children(root, dlid, ID_LEN + sizeof(UUID_PREFIX) - 1))
 			goto_out;
 		break;

lib/activate/dev_manager.h

@@ -41,7 +41,8 @@ int dev_manager_info(struct dm_pool *mem, const char *name,
 		     const struct logical_volume *lv,
 		     int mknodes, int with_open_count, struct dm_info *info);
 int dev_manager_snapshot_percent(struct dev_manager *dm,
-				 struct logical_volume *lv, float *percent);
+				 const struct logical_volume *lv,
+				 float *percent);
 int dev_manager_mirror_percent(struct dev_manager *dm,
 			       struct logical_volume *lv, int wait,
 			       float *percent, uint32_t *event_nr);

lib/cache/lvmcache.c

@@ -86,7 +86,8 @@ int vgs_locked(void)
 	return _vgs_locked;
 }
 
-struct lvmcache_vginfo *vginfo_from_vgname(const char *vgname)
+/* If vgid supplied, require a match. */
+struct lvmcache_vginfo *vginfo_from_vgname(const char *vgname, const char *vgid)
 {
 	struct lvmcache_vginfo *vginfo;
 
@@ -96,10 +97,16 @@ struct lvmcache_vginfo *vginfo_from_vgname(const char *vgname)
 	if (!(vginfo = dm_hash_lookup(_vgname_hash, vgname)))
 		return NULL;
 
+	if (vgid)
+		do 
+			if (!strncmp(vgid, vginfo->vgid, sizeof(vginfo->vgid)))
+				return vginfo;
+		while ((vginfo = vginfo->next));
+
 	return vginfo;
 }
 
-const struct format_type *fmt_from_vgname(const char *vgname)
+const struct format_type *fmt_from_vgname(const char *vgname, const char *vgid)
 {
 	struct lvmcache_vginfo *vginfo;
 	struct lvmcache_info *info;
@@ -107,8 +114,9 @@ const struct format_type *fmt_from_vgname(const char *vgname)
 	struct list *devh, *tmp;
 	struct list devs;
 	struct device_list *devl;
+	char vgid_found[ID_LEN + 1];
 
-	if (!(vginfo = vginfo_from_vgname(vgname)))
+	if (!(vginfo = vginfo_from_vgname(vgname, vgid)))
 		return NULL;
 
 	/* This function is normally called before reading metadata so
@@ -120,6 +128,8 @@ const struct format_type *fmt_from_vgname(const char *vgname)
 		list_add(&devs, &devl->list);
 	}
 
+	memcpy(vgid_found, vginfo->vgid, sizeof(vgid_found));
+
 	list_iterate_safe(devh, tmp, &devs) {
 		devl = list_item(devh, struct device_list);
 		label_read(devl->dev, &label);
@@ -127,6 +137,11 @@ const struct format_type *fmt_from_vgname(const char *vgname)
 		dm_free(devl);
 	}
 
+	/* If vginfo changed, caller needs to rescan */
+	if (!(vginfo = vginfo_from_vgname(vgname, vgid_found)) ||
+	    strncmp(vginfo->vgid, vgid_found, sizeof(vgid_found)))
+		return NULL;
+
 	return vginfo->fmt;
 }
 
@@ -148,6 +163,19 @@ struct lvmcache_vginfo *vginfo_from_vgid(const char *vgid)
 	return vginfo;
 }
 
+const char *vgname_from_vgid(struct dm_pool *mem, const char *vgid)
+{
+	struct lvmcache_vginfo *vginfo;
+
+	if ((vginfo = vginfo_from_vgid(vgid))) {
+		if (mem)
+			return dm_pool_strdup(mem, vginfo->vgname);
+		return vginfo->vgname;
+	}
+
+	return NULL;
+}
+
 struct lvmcache_info *info_from_pvid(const char *pvid)
 {
 	struct lvmcache_info *info;
@@ -232,10 +260,33 @@ int lvmcache_label_scan(struct cmd_context *cmd, int full_scan)
 	return r;
 }
 
+struct list *lvmcache_get_vgids(struct cmd_context *cmd, int full_scan)
+{
+	struct list *vgids;
+	struct lvmcache_vginfo *vginfo;
+
+	lvmcache_label_scan(cmd, full_scan);
+
+	if (!(vgids = str_list_create(cmd->mem))) {
+		log_error("vgids list allocation failed");
+		return NULL;
+	}
+
+	list_iterate_items(vginfo, &_vginfos) {
+		if (!str_list_add(cmd->mem, vgids, 
+				  dm_pool_strdup(cmd->mem, vginfo->vgid))) {
+			log_error("strlist allocation failed");
+			return NULL;
+		}
+	}
+
+	return vgids;
+}
+
 struct list *lvmcache_get_vgnames(struct cmd_context *cmd, int full_scan)
 {
 	struct list *vgnames;
-	struct lvmcache_vginfo *vgi;
+	struct lvmcache_vginfo *vginfo;
 
 	lvmcache_label_scan(cmd, full_scan);
 
@@ -244,9 +295,9 @@ struct list *lvmcache_get_vgnames(struct cmd_context *cmd, int full_scan)
 		return NULL;
 	}
 
-	list_iterate_items(vgi, &_vginfos) {
+	list_iterate_items(vginfo, &_vginfos) {
 		if (!str_list_add(cmd->mem, vgnames, 
-				  dm_pool_strdup(cmd->mem, vgi->vgname))) {
+				  dm_pool_strdup(cmd->mem, vginfo->vgname))) {
 			log_error("strlist allocation failed");
 			return NULL;
 		}
@@ -297,7 +348,7 @@ struct device *device_from_pvid(struct cmd_context *cmd, struct id *pvid)
 	return NULL;
 }
 
-static void _drop_vginfo(struct lvmcache_info *info)
+static int _drop_vginfo(struct lvmcache_info *info)
 {
 	if (!list_empty(&info->list)) {
 		list_del(&info->list);
@@ -306,8 +357,18 @@ static void _drop_vginfo(struct lvmcache_info *info)
 
 	if (info->vginfo && list_empty(&info->vginfo->infos)) {
 		dm_hash_remove(_vgname_hash, info->vginfo->vgname);
+		if (info->vginfo->next) {
+                	if (!dm_hash_insert(_vgname_hash, info->vginfo->vgname, info->vginfo->next)) {
+                        	log_error("vg hash re-insertion failed: %s",
+					  info->vginfo->vgname);
+                        	return 0;
+			}
+                }
+
 		if (info->vginfo->vgname)
 			dm_free(info->vginfo->vgname);
+		if (info->vginfo->creation_host)
+			dm_free(info->vginfo->creation_host);
 		if (*info->vginfo->vgid)
 			dm_hash_remove(_vgid_hash, info->vginfo->vgid);
 		list_del(&info->vginfo->list);
@@ -315,6 +376,8 @@ static void _drop_vginfo(struct lvmcache_info *info)
 	}
 
 	info->vginfo = NULL;
+
+	return 1;
 }
 
 /* Unused
@@ -356,8 +419,10 @@ static int _lvmcache_update_vgid(struct lvmcache_info *info, const char *vgid)
 
 	if (info->vginfo && *info->vginfo->vgid)
 		dm_hash_remove(_vgid_hash, info->vginfo->vgid);
-	if (!vgid)
+	if (!vgid) {
+		log_debug("lvmcache: %s: clearing VGID", dev_name(info->dev));
 		return 1;
+	}
 
 	strncpy(info->vginfo->vgid, vgid, sizeof(info->vginfo->vgid));
 	info->vginfo->vgid[sizeof(info->vginfo->vgid) - 1] = '\0';
@@ -367,19 +432,110 @@ static int _lvmcache_update_vgid(struct lvmcache_info *info, const char *vgid)
 		return 0;
 	}
 
+	log_debug("lvmcache: %s: setting %s VGID to %s", dev_name(info->dev),
+		  info->vginfo->vgname, info->vginfo->vgid);
+
 	return 1;
 }
 
-int lvmcache_update_vgname(struct lvmcache_info *info, const char *vgname)
+static int _insert_vginfo(struct lvmcache_vginfo *new_vginfo, const char *vgid,
+			  uint32_t vgstatus, const char *creation_host,
+			  struct lvmcache_vginfo *primary_vginfo)
 {
-	struct lvmcache_vginfo *vginfo;
+	struct lvmcache_vginfo *last_vginfo = primary_vginfo;
+	char uuid_primary[64], uuid_new[64];
+	int use_new = 0;
+	
+	/* Pre-existing VG takes precedence. Unexported VG takes precedence. */
+	if (primary_vginfo) {
+		if (!id_write_format((struct id *)vgid, uuid_new, sizeof(uuid_new)))
+			return_0;
+
+		if (!id_write_format((struct id *)&primary_vginfo->vgid, uuid_primary,
+				     sizeof(uuid_primary)))
+			return_0;
+
+		/*
+		 * If   Primary not exported, new exported => keep
+		 * Else Primary exported, new not exported => change
+		 * Else Primary has hostname for this machine => keep
+		 * Else Primary has no hostname, new has one => change
+		 * Else New has hostname for this machine => change
+		 * Else Keep primary.
+		 */
+		if (!(primary_vginfo->status & EXPORTED_VG) &&
+		    (vgstatus & EXPORTED_VG))
+			log_error("WARNING: Duplicate VG name %s: "
+				  "Existing %s takes precedence over "
+				  "exported %s", new_vginfo->vgname,
+				  uuid_primary, uuid_new);
+		else if ((primary_vginfo->status & EXPORTED_VG) &&
+			   !(vgstatus & EXPORTED_VG)) {
+			log_error("WARNING: Duplicate VG name %s: "
+				  "%s takes precedence over exported %s",
+				  new_vginfo->vgname, uuid_new,
+				  uuid_primary);
+			use_new = 1;
+		} else if (primary_vginfo->creation_host &&
+			   !strcmp(primary_vginfo->creation_host,
+				   primary_vginfo->fmt->cmd->hostname))
+			log_error("WARNING: Duplicate VG name %s: "
+				  "Existing %s (created here) takes precedence "
+				  "over %s", new_vginfo->vgname, uuid_primary,
+				  uuid_new);
+		else if (!primary_vginfo->creation_host && creation_host) {
+			log_error("WARNING: Duplicate VG name %s: "
+				  "%s (with creation_host) takes precedence over %s",
+				  new_vginfo->vgname, uuid_new,
+				  uuid_primary);
+			use_new = 1;
+		} else if (creation_host &&
+			   !strcmp(creation_host,
+				   primary_vginfo->fmt->cmd->hostname)) {
+			log_error("WARNING: Duplicate VG name %s: "
+				  "%s (created here) takes precedence over %s",
+				  new_vginfo->vgname, uuid_new,
+				  uuid_primary);
+			use_new = 1;
+		}
+
+		if (!use_new) {
+			while (last_vginfo->next)
+				last_vginfo = last_vginfo->next;
+			last_vginfo->next = new_vginfo;
+			return 1;
+		}
+
+		dm_hash_remove(_vgname_hash, primary_vginfo->vgname);
+	}
+
+	if (!dm_hash_insert(_vgname_hash, new_vginfo->vgname, new_vginfo)) {
+		log_error("cache_update: vg hash insertion failed: %s",
+		  	new_vginfo->vgname);
+		return 0;
+	}
+
+	if (primary_vginfo)
+		new_vginfo->next = primary_vginfo;
+
+	return 1;
+}
+
+static int _lvmcache_update_vgname(struct lvmcache_info *info,
+				   const char *vgname, const char *vgid,
+				   uint32_t vgstatus, const char *creation_host)
+{
+	struct lvmcache_vginfo *vginfo, *primary_vginfo;
+	// struct lvmcache_vginfo  *old_vginfo, *next;
 
 	/* If vgname is NULL and we don't already have a vgname, 
 	 * assume ORPHAN - we want every entry to have a vginfo
 	 * attached for scanning reasons.
 	 */
-	if (!vgname && !info->vginfo)
+	if (!vgname && !info->vginfo) {
 		vgname = ORPHAN;
+		vgid = ORPHAN;
+	}
 
 	if (!vgname || (info->vginfo && !strcmp(info->vginfo->vgname, vgname)))
 		return 1;
@@ -388,7 +544,43 @@ int lvmcache_update_vgname(struct lvmcache_info *info, const char *vgname)
 	_drop_vginfo(info);
 
 	/* Get existing vginfo or create new one */
-	if (!(vginfo = vginfo_from_vgname(vgname))) {
+	if (!(vginfo = vginfo_from_vgname(vgname, vgid))) {
+/*** FIXME - vginfo ends up duplicated instead of renamed.
+		// Renaming?  This lookup fails.
+		if ((vginfo = vginfo_from_vgid(vgid))) {
+			next = vginfo->next;
+			old_vginfo = vginfo_from_vgname(vginfo->vgname, NULL);
+			if (old_vginfo == vginfo) {
+				dm_hash_remove(_vgname_hash, old_vginfo->vgname);
+				if (old_vginfo->next) {
+					if (!dm_hash_insert(_vgname_hash, old_vginfo->vgname, old_vginfo->next)) {
+                        			log_error("vg hash re-insertion failed: %s",
+							  old_vginfo->vgname);
+                        			return 0;
+					}
+				}
+			} else do {
+				if (old_vginfo->next == vginfo) {
+					old_vginfo->next = vginfo->next;
+					break;
+				}
+			} while ((old_vginfo = old_vginfo->next));
+			vginfo->next = NULL;
+
+			dm_free(vginfo->vgname);
+			if (!(vginfo->vgname = dm_strdup(vgname))) {
+				log_error("cache vgname alloc failed for %s", vgname);
+				return 0;
+			}
+
+			// Rename so can assume new name does not already exist 
+			if (!dm_hash_insert(_vgname_hash, vginfo->vgname, vginfo->next)) {
+				log_error("vg hash re-insertion failed: %s",
+					  vginfo->vgname);
+                      		return 0;
+			}
+		} else {
+***/
 		if (!(vginfo = dm_malloc(sizeof(*vginfo)))) {
 			log_error("lvmcache_update_vgname: list alloc failed");
 			return 0;
@@ -400,9 +592,9 @@ int lvmcache_update_vgname(struct lvmcache_info *info, const char *vgname)
 			return 0;
 		}
 		list_init(&vginfo->infos);
-		if (!dm_hash_insert(_vgname_hash, vginfo->vgname, vginfo)) {
-			log_error("cache_update: vg hash insertion failed: %s",
-				  vginfo->vgname);
+			primary_vginfo = vginfo_from_vgname(vgname, NULL);
+		if (!_insert_vginfo(vginfo, vgid, vgstatus, creation_host,
+				    primary_vginfo)) {
 			dm_free(vginfo->vgname);
 			dm_free(vginfo);
 			return 0;
@@ -412,6 +604,9 @@ int lvmcache_update_vgname(struct lvmcache_info *info, const char *vgname)
 			list_add(&_vginfos, &vginfo->list);
 		else
 			list_add_h(&_vginfos, &vginfo->list);
+/***
+		}
+***/
 	}
 
 	info->vginfo = vginfo;
@@ -420,8 +615,59 @@ int lvmcache_update_vgname(struct lvmcache_info *info, const char *vgname)
 	/* FIXME Check consistency of list! */
 	vginfo->fmt = info->fmt;
 
-	log_debug("lvmcache: %s now %s%s", dev_name(info->dev),
-		  *vgname ? "in VG " : "orphaned", vgname);
+	log_debug("lvmcache: %s: now %s%s%s%s%s", dev_name(info->dev),
+		  *vgname ? "in VG " : "orphaned", vgname,
+		  vginfo->vgid[0] ? " (" : "",
+		  vginfo->vgid[0] ? vginfo->vgid : "",
+		  vginfo->vgid[0] ? ")" : "");
+
+	return 1;
+}
+
+static int _lvmcache_update_vgstatus(struct lvmcache_info *info, uint32_t vgstatus,
+				     const char *creation_host)
+{
+	if (!info || !info->vginfo)
+		return 1;
+
+	if ((info->vginfo->status & EXPORTED_VG) != (vgstatus & EXPORTED_VG))
+		log_debug("lvmcache: %s: VG %s %s exported",
+			  dev_name(info->dev), info->vginfo->vgname,
+			  vgstatus & EXPORTED_VG ? "now" : "no longer");
+
+	info->vginfo->status = vgstatus;
+
+	if (!creation_host)
+		return 1;
+
+	if (info->vginfo->creation_host && !strcmp(creation_host,
+						   info->vginfo->creation_host))
+		return 1;
+
+	if (info->vginfo->creation_host)
+		dm_free(info->vginfo->creation_host);
+
+	if (!(info->vginfo->creation_host = dm_strdup(creation_host))) {
+		log_error("cache creation host alloc failed for %s",
+			  creation_host);
+		return 0;
+	}
+
+	log_debug("lvmcache: %s: VG %s: Set creation host to %s.",
+		  dev_name(info->dev), info->vginfo->vgname, creation_host);
+
+	return 1;
+}
+
+int lvmcache_update_vgname_and_id(struct lvmcache_info *info,
+				  const char *vgname, const char *vgid,
+				  uint32_t vgstatus, const char *creation_host)
+{
+	if (!_lvmcache_update_vgname(info, vgname, vgid, vgstatus,
+				     creation_host) ||
+	    !_lvmcache_update_vgid(info, vgid) ||
+	    !_lvmcache_update_vgstatus(info, vgstatus, creation_host))
+		return_0;
 
 	return 1;
 }
@@ -431,20 +677,17 @@ int lvmcache_update_vg(struct volume_group *vg)
 	struct pv_list *pvl;
 	struct lvmcache_info *info;
 	char pvid_s[ID_LEN + 1];
-	int vgid_updated = 0;
 
 	pvid_s[sizeof(pvid_s) - 1] = '\0';
 
 	list_iterate_items(pvl, &vg->pvs) {
 		strncpy(pvid_s, (char *) &pvl->pv->id, sizeof(pvid_s) - 1);
 		/* FIXME Could pvl->pv->dev->pvid ever be different? */
-		if ((info = info_from_pvid(pvid_s))) {
-			lvmcache_update_vgname(info, vg->name);
-			if (!vgid_updated) {
-				_lvmcache_update_vgid(info, (char *) &vg->id);
-				vgid_updated = 1;
-			}
-		}
+		if ((info = info_from_pvid(pvid_s)) &&
+		    !lvmcache_update_vgname_and_id(info, vg->name,
+						   (char *) &vg->id,
+						   vg->status, NULL))
+			return_0;
 	}
 
 	return 1;
@@ -452,7 +695,8 @@ int lvmcache_update_vg(struct volume_group *vg)
 
 struct lvmcache_info *lvmcache_add(struct labeller *labeller, const char *pvid,
 				   struct device *dev,
-				   const char *vgname, const char *vgid)
+				   const char *vgname, const char *vgid,
+				   uint32_t vgstatus)
 {
 	struct label *label;
 	struct lvmcache_info *existing, *info;
@@ -548,7 +792,7 @@ struct lvmcache_info *lvmcache_add(struct labeller *labeller, const char *pvid,
 		return NULL;
 	}
 
-	if (!lvmcache_update_vgname(info, vgname)) {
+	if (!lvmcache_update_vgname_and_id(info, vgname, vgid, vgstatus, NULL)) {
 		if (!existing) {
 			dm_hash_remove(_pvid_hash, pvid_s);
 			strcpy(info->dev->pvid, "");
@@ -558,10 +802,6 @@ struct lvmcache_info *lvmcache_add(struct labeller *labeller, const char *pvid,
 		return NULL;
 	}
 
-	if (!_lvmcache_update_vgid(info, vgid))
-		/* Non-critical */
-		stack;
-
 	return info;
 }
 
@@ -576,9 +816,16 @@ static void _lvmcache_destroy_entry(struct lvmcache_info *info)
 
 static void _lvmcache_destroy_vgnamelist(struct lvmcache_vginfo *vginfo)
 {
-	if (vginfo->vgname)
-		dm_free(vginfo->vgname);
-	dm_free(vginfo);
+	struct lvmcache_vginfo *next;
+
+	do {
+		next = vginfo->next;
+		if (vginfo->vgname)
+			dm_free(vginfo->vgname);
+		if (vginfo->creation_host)
+			dm_free(vginfo->creation_host);
+		dm_free(vginfo);
+	} while ((vginfo = next));
 }
 
 static void _lvmcache_destroy_lockname(int present)

lib/cache/lvmcache.h

@@ -33,15 +33,20 @@ struct cmd_context;
 struct format_type;
 struct volume_group;
 
+/* One per VG */
 struct lvmcache_vginfo {
 	struct list list;	/* Join these vginfos together */
 	struct list infos;	/* List head for lvmcache_infos */
 	const struct format_type *fmt;
 	char *vgname;		/* "" == orphan */
+	uint32_t status;
 	char vgid[ID_LEN + 1];
 	char _padding[7];
+	struct lvmcache_vginfo *next; /* Another VG with same name? */
+	char *creation_host;
 };
 
+/* One per device */
 struct lvmcache_info {
 	struct list list;	/* Join VG members together */
 	struct list mdas;	/* list head for metadata areas */
@@ -64,21 +69,26 @@ int lvmcache_label_scan(struct cmd_context *cmd, int full_scan);
 /* Add/delete a device */
 struct lvmcache_info *lvmcache_add(struct labeller *labeller, const char *pvid,
 				   struct device *dev,
-				   const char *vgname, const char *vgid);
+				   const char *vgname, const char *vgid,
+				   uint32_t vgstatus);
 void lvmcache_del(struct lvmcache_info *info);
 
 /* Update things */
-int lvmcache_update_vgname(struct lvmcache_info *info, const char *vgname);
+int lvmcache_update_vgname_and_id(struct lvmcache_info *info,
+				  const char *vgname, const char *vgid,
+				  uint32_t vgstatus, const char *hostname);
 int lvmcache_update_vg(struct volume_group *vg);
 
 void lvmcache_lock_vgname(const char *vgname, int read_only);
 void lvmcache_unlock_vgname(const char *vgname);
 
 /* Queries */
-const struct format_type *fmt_from_vgname(const char *vgname);
-struct lvmcache_vginfo *vginfo_from_vgname(const char *vgname);
+const struct format_type *fmt_from_vgname(const char *vgname, const char *vgid);
+struct lvmcache_vginfo *vginfo_from_vgname(const char *vgname,
+					   const char *vgid);
 struct lvmcache_vginfo *vginfo_from_vgid(const char *vgid);
 struct lvmcache_info *info_from_pvid(const char *pvid);
+const char *vgname_from_vgid(struct dm_pool *mem, const char *vgid);
 struct device *device_from_pvid(struct cmd_context *cmd, struct id *pvid);
 int vgs_locked(void);
 int vgname_is_locked(const char *vgname);
@@ -87,4 +97,8 @@ int vgname_is_locked(const char *vgname);
 /* Set full_scan to 1 to reread every filtered device label */
 struct list *lvmcache_get_vgnames(struct cmd_context *cmd, int full_scan);
 
+/* Returns list of struct str_lists containing pool-allocated copy of vgids */
+/* Set full_scan to 1 to reread every filtered device label */
+struct list *lvmcache_get_vgids(struct cmd_context *cmd, int full_scan);
+
 #endif

lib/commands/toolcontext.c

@@ -657,7 +657,7 @@ static int _init_formats(struct cmd_context *cmd)
 				return 0;
 			}
 			if (!(lib = load_shared_library(cmd->cft, cv->v.str,
-							"format"))) {
+							"format", 0))) {
 				stack;
 				return 0;
 			}
@@ -753,7 +753,7 @@ static int _init_segtypes(struct cmd_context *cmd)
 				return 0;
 			}
 			if (!(lib = load_shared_library(cmd->cft, cv->v.str,
-							"segment type"))) {
+							"segment type", 0))) {
 				stack;
 				return 0;
 			}

lib/config/defaults.h

@@ -34,6 +34,8 @@
 #define DEFAULT_LOCK_DIR "/var/lock/lvm"
 #define DEFAULT_LOCKING_LIB "lvm2_locking.so"
 
+#define DEFAULT_DMEVENTD_MIRROR_LIB "libdevmapper-event-lvm2mirror.so"
+
 #define DEFAULT_UMASK 0077
 
 #ifdef LVM1_FALLBACK

lib/device/dev-io.c

@@ -418,7 +418,6 @@ int dev_open_flags(struct device *dev, int flags, int direct, int quiet)
 	    ((fstat(dev->fd, &buf) < 0) || (buf.st_rdev != dev->dev))) {
 		log_error("%s: fstat failed: Has device name changed?", name);
 		dev_close_immediate(dev);
-		dev->open_count = 0;
 		return 0;
 	}
 
@@ -509,11 +508,9 @@ static int _dev_close(struct device *dev, int immediate)
 	if (dev->open_count > 0)
 		dev->open_count--;
 
-	if (immediate && dev->open_count) {
+	if (immediate && dev->open_count)
 		log_debug("%s: Immediate close attempt while still referenced",
 			  dev_name(dev));
-		dev->open_count = 0;
-	}
 
 	/* Close unless device is known to belong to a locked VG */
 	if (immediate ||

lib/format1/disk-rep.c

@@ -223,11 +223,11 @@ static int _read_lvd(struct device *dev, uint64_t pos, struct lv_disk *disk)
 	return 1;
 }
 
-static int _read_vgd(struct disk_list *data)
+int read_vgd(struct device *dev, struct vg_disk *vgd, struct pv_disk *pvd)
 {
-	struct vg_disk *vgd = &data->vgd;
-	uint64_t pos = data->pvd.vg_on_disk.base;
-	if (!dev_read(data->dev, pos, sizeof(*vgd), vgd))
+	uint64_t pos = pvd->vg_on_disk.base;
+
+	if (!dev_read(dev, pos, sizeof(*vgd), vgd))
 		fail;
 
 	_xlate_vgd(vgd);
@@ -319,13 +319,31 @@ static int _read_extents(struct disk_list *data)
 	return 1;
 }
 
+static void __update_lvmcache(const struct format_type *fmt,
+			      struct disk_list *dl,
+			      struct device *dev, const char *vgid,
+			      int exported)
+{
+	struct lvmcache_info *info;
+
+	if (!(info = lvmcache_add(fmt->labeller, dl->pvd.pv_uuid, dev,
+				  dl->pvd.vg_name, vgid,
+				  exported ? EXPORTED_VG : 0))) {
+		stack;
+		return;
+	}
+
+	info->device_size = xlate32(dl->pvd.pv_size) << SECTOR_SHIFT;
+	list_init(&info->mdas);
+	info->status &= ~CACHE_INVALID;
+}
+
 static struct disk_list *__read_disk(const struct format_type *fmt,
 				     struct device *dev, struct dm_pool *mem,
 				     const char *vg_name)
 {
-	struct disk_list *dl = dm_pool_alloc(mem, sizeof(*dl));
+	struct disk_list *dl = dm_pool_zalloc(mem, sizeof(*dl));
 	const char *name = dev_name(dev);
-	struct lvmcache_info *info;
 
 	if (!dl) {
 		stack;
@@ -342,41 +360,32 @@ static struct disk_list *__read_disk(const struct format_type *fmt,
 		goto bad;
 	}
 
-	if (!(info = lvmcache_add(fmt->labeller, dl->pvd.pv_uuid, dev,
-				  dl->pvd.vg_name, NULL)))
-		stack;
-	else {
-		info->device_size = xlate32(dl->pvd.pv_size) << SECTOR_SHIFT;
-		list_init(&info->mdas);
-		info->status &= ~CACHE_INVALID;
-	}
-
 	/*
 	 * is it an orphan ?
 	 */
 	if (!*dl->pvd.vg_name) {
 		log_very_verbose("%s is not a member of any format1 VG", name);
 
-		/* Update VG cache */
-		/* vgcache_add(dl->pvd.vg_name, NULL, dev, fmt); */
-
+		__update_lvmcache(fmt, dl, dev, NULL, 0);
 		return (vg_name) ? NULL : dl;
 	}
 
-	if (!_read_vgd(dl)) {
+	if (!read_vgd(dl->dev, &dl->vgd, &dl->pvd)) {
 		log_error("Failed to read VG data from PV (%s)", name);
+		__update_lvmcache(fmt, dl, dev, NULL, 0);
 		goto bad;
 	}
 
-	/* Update VG cache with what we found */
-	/* vgcache_add(dl->pvd.vg_name, dl->vgd.vg_uuid, dev, fmt); */
-
 	if (vg_name && strcmp(vg_name, dl->pvd.vg_name)) {
 		log_very_verbose("%s is not a member of the VG %s",
 				 name, vg_name);
+		__update_lvmcache(fmt, dl, dev, NULL, 0);
 		goto bad;
 	}
 
+	__update_lvmcache(fmt, dl, dev, dl->vgd.vg_uuid,
+			  dl->vgd.vg_status & VG_EXPORTED);
+
 	if (!_read_uuids(dl)) {
 		log_error("Failed to read PV uuid list from %s", name);
 		goto bad;
@@ -461,7 +470,7 @@ int read_pvs_in_vg(const struct format_type *fmt, const char *vg_name,
 	struct lvmcache_info *info;
 
 	/* Fast path if we already saw this VG and cached the list of PVs */
-	if (vg_name && (vginfo = vginfo_from_vgname(vg_name)) &&
+	if (vg_name && (vginfo = vginfo_from_vgname(vg_name, NULL)) &&
 	    vginfo->infos.n) {
 		list_iterate_items(info, &vginfo->infos) {
 			dev = info->dev;

lib/format1/disk-rep.h

@@ -204,7 +204,8 @@ int write_disks(const struct format_type *fmt, struct list *pvds);
  */
 int import_pv(struct dm_pool *mem, struct device *dev,
 	      struct volume_group *vg,
-	      struct physical_volume *pv, struct pv_disk *pvd);
+	      struct physical_volume *pv, struct pv_disk *pvd,
+	      struct vg_disk *vgd);
 int export_pv(struct cmd_context *cmd, struct dm_pool *mem,
 	      struct volume_group *vg,
 	      struct pv_disk *pvd, struct physical_volume *pv);
@@ -237,6 +238,7 @@ void export_numbers(struct list *pvds, struct volume_group *vg);
 
 void export_pv_act(struct list *pvds);
 int munge_pvd(struct device *dev, struct pv_disk *pvd);
+int read_vgd(struct device *dev, struct vg_disk *vgd, struct pv_disk *pvd);
 
 /* blech */
 int get_free_vg_number(struct format_instance *fid, struct dev_filter *filter,

lib/format1/format1.c

@@ -312,7 +312,7 @@ static int _pv_read(const struct format_type *fmt, const char *pv_name,
 		goto out;
 	}
 
-	if (!import_pv(fmt->cmd->mem, dl->dev, NULL, pv, &dl->pvd)) {
+	if (!import_pv(fmt->cmd->mem, dl->dev, NULL, pv, &dl->pvd, &dl->vgd)) {
 		stack;
 		goto out;
 	}
@@ -396,7 +396,7 @@ static int _pv_write(const struct format_type *fmt, struct physical_volume *pv,
 	struct lvmcache_info *info;
 
 	if (!(info = lvmcache_add(fmt->labeller, (char *) &pv->id, pv->dev,
-				  pv->vg_name, NULL))) {
+				  pv->vg_name, NULL, 0))) {
 		stack;
 		return 0;
 	}
@@ -502,6 +502,7 @@ static struct metadata_area_ops _metadata_format1_ops = {
 
 static struct format_instance *_create_instance(const struct format_type *fmt,
 						const char *vgname,
+						const char *vgid,
 						void *private)
 {
 	struct format_instance *fid;

lib/format1/import-export.c

@@ -49,7 +49,8 @@ static char *_create_lv_name(struct dm_pool *mem, const char *full_name)
 
 int import_pv(struct dm_pool *mem, struct device *dev,
 	      struct volume_group *vg,
-	      struct physical_volume *pv, struct pv_disk *pvd)
+	      struct physical_volume *pv, struct pv_disk *pvd,
+	      struct vg_disk *vgd)
 {
 	memset(pv, 0, sizeof(*pv));
 	memcpy(&pv->id, pvd->pv_uuid, ID_LEN);
@@ -60,6 +61,8 @@ int import_pv(struct dm_pool *mem, struct device *dev,
 		return 0;
 	}
 
+	memcpy(&pv->vgid, vgd->vg_uuid, sizeof(vg->id));
+
 	/* Store system_id from first PV if PV belongs to a VG */
 	if (vg && !*vg->system_id)
 		strncpy(vg->system_id, pvd->system_id, NAME_LEN);
@@ -426,7 +429,7 @@ int import_pvs(const struct format_type *fmt, struct dm_pool *mem,
 			return 0;
 		}
 
-		if (!import_pv(mem, dl->dev, vg, pvl->pv, &dl->pvd)) {
+		if (!import_pv(mem, dl->dev, vg, pvl->pv, &dl->pvd, &dl->vgd)) {
 			stack;
 			return 0;
 		}

lib/format1/import-extents.c

@@ -140,7 +140,12 @@ static int _fill_maps(struct dm_hash_table *maps, struct volume_group *vg,
 				lvm = lvms[lv_num];
 
 				if (!lvm) {
-					log_err("invalid lv in extent map");
+					log_error("Invalid LV in extent map "
+						  "(PV %s, PE %" PRIu32
+						  ", LV %" PRIu32
+						  ", LE %" PRIu32 ")",
+						  dev_name(pv->dev), i,
+						  lv_num, e[i].le_num);
 					return 0;
 				}
 

lib/format1/lvm1-label.c

@@ -58,11 +58,20 @@ static int _read(struct labeller *l, struct device *dev, char *buf,
 		 struct label **label)
 {
 	struct pv_disk *pvd = (struct pv_disk *) buf;
+	struct vg_disk vgd;
 	struct lvmcache_info *info;
+	const char *vgid = NULL;
+	int exported = 0;
 
 	munge_pvd(dev, pvd);
 
-	if (!(info = lvmcache_add(l, pvd->pv_uuid, dev, pvd->vg_name, NULL))) {
+	if (*pvd->vg_name && read_vgd(dev, &vgd, pvd)) {
+		vgid = vgd.vg_uuid;
+		exported = pvd->pv_status & VG_EXPORTED;
+	}
+
+	if (!(info = lvmcache_add(l, pvd->pv_uuid, dev, pvd->vg_name, vgid,
+				  exported))) {
 		stack;
 		return 0;
 	}

lib/format_pool/disk_rep.c

@@ -98,7 +98,7 @@ int read_pool_label(struct pool_list *pl, struct labeller *l,
 	log_debug("Calculated uuid %s for %s", uuid, pd->pl_pool_name);
 
 	if (!(info = lvmcache_add(l, (char *) &pvid, dev, pd->pl_pool_name,
-				  (char *) &vgid))) {
+				  (char *) &vgid, 0))) {
 		stack;
 		return 0;
 	}
@@ -314,7 +314,7 @@ int read_pool_pds(const struct format_type *fmt, const char *vg_name,
 		/*
 		 * If the cache scanning doesn't work, this will never work
 		 */
-		if (vg_name && (vginfo = vginfo_from_vgname(vg_name)) &&
+		if (vg_name && (vginfo = vginfo_from_vgname(vg_name, NULL)) &&
 		    vginfo->infos.n) {
 
 			if (_read_vg_pds(fmt, mem, vginfo, pdhead, &totaldevs)) {

lib/format_pool/format_pool.c

@@ -268,6 +268,7 @@ static struct metadata_area_ops _metadata_format_pool_ops = {
 
 static struct format_instance *_create_instance(const struct format_type *fmt,
 						const char *vgname,
+						const char *vgid,
 						void *private)
 {
 	struct format_instance *fid;

lib/format_pool/import_export.c

@@ -165,6 +165,7 @@ int import_pool_pv(const struct format_type *fmt, struct dm_pool *mem,
 		log_error("Unable to duplicate vg_name string");
 		return 0;
 	}
+	memcpy(&pv->vgid, &vg->id, sizeof(vg->id));
 	pv->status = 0;
 	pv->size = pd->pl_blocks;
 	pv->pe_size = POOL_PE_SIZE;

lib/format_text/archive.c

@@ -207,8 +207,8 @@ static void _remove_expired(struct list *archives, uint32_t archives_size,
 	/* Convert retain_days into the time after which we must retain */
 	retain_time = time(NULL) - (time_t) retain_days *SECS_PER_DAY;
 
-	/* Assume list is ordered oldest first (by index) */
-	list_iterate_items(bf, archives) {
+	/* Assume list is ordered newest first (by index) */
+	list_iterate_back_items(bf, archives) {
 		/* Get the mtime of the file and unlink if too old */
 		if (stat(bf->path, &sb)) {
 			log_sys_error("stat", bf->path);
@@ -311,7 +311,7 @@ static void _display_archive(struct cmd_context *cmd, struct archive_file *af)
 
 	if (!(context = create_text_context(cmd, af->path, NULL)) ||
 	    !(tf = cmd->fmt_backup->ops->create_instance(cmd->fmt_backup, NULL,
-							 context))) {
+							 NULL, context))) {
 		log_error("Couldn't create text instance object.");
 		return;
 	}

lib/format_text/archiver.c

@@ -257,7 +257,7 @@ struct volume_group *backup_read_vg(struct cmd_context *cmd,
 	if (!(context = create_text_context(cmd, file,
 					    cmd->cmd_line)) ||
 	    !(tf = cmd->fmt_backup->ops->create_instance(cmd->fmt_backup, NULL,
-							 context))) {
+							 NULL, context))) {
 		log_error("Couldn't create text format object.");
 		return NULL;
 	}
@@ -286,7 +286,7 @@ int backup_restore_vg(struct cmd_context *cmd, struct volume_group *vg)
 
 	/* Attempt to write out using currently active format */
 	if (!(vg->fid = cmd->fmt->ops->create_instance(cmd->fmt, vg->name,
-						       NULL))) {
+						       NULL, NULL))) {
 		log_error("Failed to allocate format instance");
 		return 0;
 	}
@@ -365,7 +365,7 @@ int backup_to_file(const char *file, const char *desc, struct volume_group *vg)
 
 	if (!(context = create_text_context(cmd, file, desc)) ||
 	    !(tf = cmd->fmt_backup->ops->create_instance(cmd->fmt_backup, NULL,
-							 context))) {
+							 NULL, context))) {
 		log_error("Couldn't create backup object.");
 		return 0;
 	}

lib/format_text/format-text.c

@@ -41,6 +41,7 @@
 
 static struct format_instance *_create_text_instance(const struct format_type
 						     *fmt, const char *vgname,
+						     const char *vgid,
 						     void *context);
 
 struct text_fid_context {
@@ -219,7 +220,7 @@ static struct raw_locn *_find_vg_rlocn(struct device_area *dev_area,
 
       error:
 	if ((info = info_from_pvid(dev_area->dev->pvid)))
-		lvmcache_update_vgname(info, ORPHAN);
+		lvmcache_update_vgname_and_id(info, ORPHAN, ORPHAN, 0, NULL);
 
 	return NULL;
 }
@@ -858,9 +859,11 @@ static int _scan_file(const struct format_type *fmt)
 				}
 
 				/* FIXME stat file to see if it's changed */
-				fid = _create_text_instance(fmt, NULL, NULL);
+				fid = _create_text_instance(fmt, NULL, NULL,
+							    NULL);
 				if ((vg = _vg_read_file_name(fid, vgname,
 							     path)))
+					/* FIXME Store creation host in vg */
 					lvmcache_update_vg(vg);
 			}
 
@@ -871,65 +874,92 @@ static int _scan_file(const struct format_type *fmt)
 	return 1;
 }
 
-int vgname_from_mda(const struct format_type *fmt, struct device_area *dev_area,
-		    char *buf, uint32_t size)
+const char *vgname_from_mda(const struct format_type *fmt,
+			    struct device_area *dev_area, struct id *vgid,
+			    uint32_t *vgstatus, char **creation_host)
 {
 	struct raw_locn *rlocn;
 	struct mda_header *mdah;
-	unsigned int len;
-	int r = 0;
+	uint32_t wrap = 0;
+	const char *vgname = NULL;
+	unsigned int len = 0;
+	char buf[NAME_LEN + 1];
 
 	if (!dev_open(dev_area->dev)) {
 		stack;
-		return 0;
+		return NULL;
 	}
 
-	if (!(mdah = _raw_read_mda_header(fmt, dev_area))) {
-		stack;
+	if (!(mdah = _raw_read_mda_header(fmt, dev_area)))
+		goto_out;
+
+	/* FIXME Cope with returning a list */
+	rlocn = mdah->raw_locns;
+
+	/* Do quick check for a vgname */
+	if (!dev_read(dev_area->dev, dev_area->start + rlocn->offset,
+		      NAME_LEN, buf))
+		goto_out;
+
+	while (buf[len] && !isspace(buf[len]) && buf[len] != '{' &&
+	       len < (NAME_LEN - 1))
+		len++;
+
+	buf[len] = '\0';
+
+	/* Ignore this entry if the characters aren't permissible */
+	if (!validate_name(buf))
+		goto_out;
+
+	/* We found a VG - now check the metadata */
+	if (rlocn->offset + rlocn->size > mdah->size)
+		wrap = (uint32_t) ((rlocn->offset + rlocn->size) - mdah->size);
+
+	if (wrap > rlocn->offset) {
+		log_error("%s: metadata too large for circular buffer",
+			  dev_name(dev_area->dev));
 		goto out;
 	}
 
-	rlocn = mdah->raw_locns;
+	/* FIXME 64-bit */
+	if (!(vgname = text_vgname_import(fmt, dev_area->dev,
+					  (off_t) (dev_area->start +
+						   rlocn->offset),
+					  (uint32_t) (rlocn->size - wrap),
+					  (off_t) (dev_area->start +
+						   MDA_HEADER_SIZE),
+					  wrap, calc_crc, rlocn->checksum,
+					  vgid, vgstatus, creation_host)))
+		goto_out;
 
-	while (rlocn->offset) {
-		if (!dev_read(dev_area->dev, dev_area->start + rlocn->offset,
-			      size, buf)) {
-			stack;
-			goto out;
-		}
-		len = 0;
-		while (buf[len] && !isspace(buf[len]) && buf[len] != '{' &&
-		       len < (size - 1))
-			len++;
-		buf[len] = '\0';
-
-		/* Ignore this entry if the characters aren't permissible */
-		if (!validate_name(buf)) {
-			stack;
-			goto out;
-		}
-
-		r = 1;
-		break;
-
-		/* FIXME Cope with returning a list */
-		rlocn++;
+	/* Ignore this entry if the characters aren't permissible */
+	if (!validate_name(vgname)) {
+		stack;
+		vgname = NULL;
+		goto out;
 	}
 
+	log_debug("%s: Found metadata at %" PRIu64 " size %" PRIu64
+		  " for %s (%s)", 
+		  dev_name(dev_area->dev), dev_area->start + rlocn->offset,
+		  rlocn->size, vgname, vgid->uuid);
+
       out:
 	if (!dev_close(dev_area->dev))
 		stack;
 
-	return r;
+	return vgname;
 }
 
 static int _scan_raw(const struct format_type *fmt)
 {
 	struct raw_list *rl;
 	struct list *raw_list;
-	char vgnamebuf[NAME_LEN + 2];
+	const char *vgname;
 	struct volume_group *vg;
 	struct format_instance fid;
+	struct id vgid;
+	uint32_t vgstatus;
 
 	raw_list = &((struct mda_lists *) fmt->private)->raws;
 
@@ -938,9 +968,9 @@ static int _scan_raw(const struct format_type *fmt)
 
 	list_iterate_items(rl, raw_list) {
 		/* FIXME We're reading mdah twice here... */
-		if (vgname_from_mda(fmt, &rl->dev_area, vgnamebuf,
-				    sizeof(vgnamebuf))) {
-			if ((vg = _vg_read_raw_area(&fid, vgnamebuf,
+		if ((vgname = vgname_from_mda(fmt, &rl->dev_area, &vgid, &vgstatus,
+					      NULL))) {
+			if ((vg = _vg_read_raw_area(&fid, vgname,
 						    &rl->dev_area, 0)))
 				lvmcache_update_vg(vg);
 		}
@@ -1109,7 +1139,7 @@ static int _pv_write(const struct format_type *fmt, struct physical_volume *pv,
 	/* FIXME Test mode don't update cache? */
 
 	if (!(info = lvmcache_add(fmt->labeller, (char *) &pv->id, pv->dev,
-				  ORPHAN, NULL))) {
+				  ORPHAN, NULL, 0))) {
 		stack;
 		return 0;
 	}
@@ -1245,7 +1275,7 @@ static int _pv_read(const struct format_type *fmt, const char *pv_name,
 	/* Have we already cached vgname? */
 	if (info->vginfo && info->vginfo->vgname && *info->vginfo->vgname &&
 	    get_pv_from_vg_by_id(info->fmt, info->vginfo->vgname,
-				 info->dev->pvid, pv)) {
+				 info->vginfo->vgid, info->dev->pvid, pv)) {
 		return 1;
 	}
 
@@ -1256,6 +1286,7 @@ static int _pv_read(const struct format_type *fmt, const char *pv_name,
 		if (info->vginfo && info->vginfo->vgname &&
 		    *info->vginfo->vgname &&
 		    get_pv_from_vg_by_id(info->fmt, info->vginfo->vgname,
+					 info->vginfo->vgid,
 					 info->dev->pvid, pv)) {
 			return 1;
 		}
@@ -1450,6 +1481,7 @@ static int _pv_setup(const struct format_type *fmt,
 /* NULL vgname means use only the supplied context e.g. an archive file */
 static struct format_instance *_create_text_instance(const struct format_type
 						     *fmt, const char *vgname,
+						     const char *vgid,
 						     void *context)
 {
 	struct format_instance *fid;
@@ -1535,7 +1567,7 @@ static struct format_instance *_create_text_instance(const struct format_type
 
 		/* Scan PVs in VG for any further MDAs */
 		lvmcache_label_scan(fmt->cmd, 0);
-		if (!(vginfo = vginfo_from_vgname(vgname))) {
+		if (!(vginfo = vginfo_from_vgname(vgname, vgid))) {
 			stack;
 			goto out;
 		}
@@ -1569,7 +1601,6 @@ static struct format_instance *_create_text_instance(const struct format_type
 
       out:
 	return fid;
-
 }
 
 void *create_text_context(struct cmd_context *cmd, const char *path,

lib/format_text/format-text.h

@@ -54,7 +54,8 @@ int add_mda(const struct format_type *fmt, struct dm_pool *mem, struct list *mda
 	    struct device *dev, uint64_t start, uint64_t size);
 void del_mdas(struct list *mdas);
 
-int vgname_from_mda(const struct format_type *fmt, struct device_area *dev_area,
-		    char *buf, uint32_t size);
+const char *vgname_from_mda(const struct format_type *fmt,
+			    struct device_area *dev_area, struct id *vgid,
+			    uint32_t *vgstatus, char **creation_host);
 
 #endif

lib/format_text/import-export.h

@@ -47,6 +47,10 @@ struct text_vg_version_ops {
 					 struct config_tree * cf);
 	void (*read_desc) (struct dm_pool * mem, struct config_tree * cf,
 			   time_t *when, char **desc);
+	const char *(*read_vgname) (const struct format_type *fmt,
+				    struct config_tree *cft,
+				    struct id *vgid, uint32_t *vgstatus,
+				    char **creation_host);
 };
 
 struct text_vg_version_ops *text_vg_vsn1_init(void);
@@ -70,5 +74,12 @@ struct volume_group *text_vg_import_fd(struct format_instance *fid,
 				       checksum_fn_t checksum_fn,
 				       uint32_t checksum,
 				       time_t *when, char **desc);
+const char *text_vgname_import(const struct format_type *fmt,
+			       struct device *dev,
+                               off_t offset, uint32_t size,
+                               off_t offset2, uint32_t size2,
+                               checksum_fn_t checksum_fn, uint32_t checksum,
+                               struct id *vgid, uint32_t *vgstatus,
+			       char **creation_host);
 
 #endif

lib/format_text/import.c

@@ -23,6 +23,53 @@
 /* FIXME Use tidier inclusion method */
 static struct text_vg_version_ops *(_text_vsn_list[2]);
 
+const char *text_vgname_import(const struct format_type *fmt,
+			       struct device *dev,
+			       off_t offset, uint32_t size,
+			       off_t offset2, uint32_t size2,
+			       checksum_fn_t checksum_fn, uint32_t checksum,
+			       struct id *vgid, uint32_t *vgstatus,
+			       char **creation_host)
+{
+	struct config_tree *cft;
+	struct text_vg_version_ops **vsn;
+	const char *vgname;
+
+	static int _initialised = 0;
+
+	if (!_initialised) {
+		_text_vsn_list[0] = text_vg_vsn1_init();
+		_text_vsn_list[1] = NULL;
+		_initialised = 1;
+	}
+
+	if (!(cft = create_config_tree(NULL)))
+		goto_out;
+
+	if ((!dev && !read_config_file(cft)) ||
+	    (dev && !read_config_fd(cft, dev, offset, size,
+				    offset2, size2, checksum_fn, checksum)))
+		goto_out;
+
+	/* 
+	 * Find a set of version functions that can read this file
+	 */
+	for (vsn = &_text_vsn_list[0]; *vsn; vsn++) {
+		if (!(*vsn)->check_version(cft))
+			continue;
+
+		if (!(vgname = (*vsn)->read_vgname(fmt, cft, vgid, vgstatus,
+						   creation_host)))
+			goto_out;
+
+		break;
+	}
+
+      out:
+	destroy_config_tree(cft);
+	return vgname;
+}
+
 struct volume_group *text_vg_import_fd(struct format_instance *fid,
 				       const char *file,
 				       struct device *dev,

lib/format_text/import_vsn1.c

@@ -167,6 +167,8 @@ static int _read_pv(struct format_instance *fid, struct dm_pool *mem,
 		return 0;
 	}
 
+	memcpy(&pv->vgid, &vg->id, sizeof(vg->id));
+
 	if (!(cn = find_config_node(pvn, "status"))) {
 		log_error("Couldn't find status flags for physical volume.");
 		return 0;
@@ -798,10 +800,59 @@ static void _read_desc(struct dm_pool *mem,
 	*when = u;
 }
 
+static const char *_read_vgname(const struct format_type *fmt,
+				struct config_tree *cft, struct id *vgid,
+				uint32_t *vgstatus, char **creation_host)
+{
+	struct config_node *vgn, *cn;
+	struct dm_pool *mem = fmt->cmd->mem;
+	char *vgname;
+	int old_suppress;
+
+	old_suppress = log_suppress(2);
+	*creation_host = dm_pool_strdup(mem,
+					find_config_str(cft->root,
+						        "creation_host", ""));
+	log_suppress(old_suppress);
+
+	/* skip any top-level values */
+	for (vgn = cft->root; (vgn && vgn->v); vgn = vgn->sib) ;
+
+	if (!vgn) {
+		log_error("Couldn't find volume group in file.");
+		return 0;
+	}
+
+	if (!(vgname = dm_pool_strdup(mem, vgn->key)))
+		return_0;
+
+	vgn = vgn->child;
+
+	if (!_read_id(vgid, vgn, "id")) {
+		log_error("Couldn't read uuid for volume group %s.", vgname);
+		return 0;
+	}
+
+	if (!(cn = find_config_node(vgn, "status"))) {
+		log_error("Couldn't find status flags for volume group %s.",
+			  vgname);
+		return 0;
+	}
+
+	if (!(read_flags(vgstatus, VG_FLAGS, cn->v))) {
+		log_error("Couldn't read status flags for volume group %s.",
+			  vgname);
+		return 0;
+	}
+
+	return vgname;
+}
+
 static struct text_vg_version_ops _vsn1_ops = {
 	check_version:_check_version,
 	read_vg:_read_vg,
-	read_desc:_read_desc
+	read_desc:_read_desc,
+	read_vgname:_read_vgname
 };
 
 struct text_vg_version_ops *text_vg_vsn1_init(void)

lib/format_text/text_label.c

@@ -195,13 +195,16 @@ static int _read(struct labeller *l, struct device *dev, char *buf,
 	struct disk_locn *dlocn_xl;
 	uint64_t offset;
 	struct metadata_area *mda;
-	char vgnamebuf[NAME_LEN + 2];
+	struct id vgid;
 	struct mda_context *mdac;
+	const char *vgname;
+	uint32_t vgstatus;
+	char *creation_host;
 
 	pvhdr = (struct pv_header *) ((void *) buf + xlate32(lh->offset_xl));
 
-	if (!(info = lvmcache_add(l, pvhdr->pv_uuid, dev, NULL, NULL)))
-		return 0;
+	if (!(info = lvmcache_add(l, pvhdr->pv_uuid, dev, NULL, NULL, 0)))
+		return_0;
 	*label = info->label;
 
 	info->device_size = xlate64(pvhdr->device_size_xl);
@@ -232,10 +235,12 @@ static int _read(struct labeller *l, struct device *dev, char *buf,
 
 	list_iterate_items(mda, &info->mdas) {
 		mdac = (struct mda_context *) mda->metadata_locn;
-		if (vgname_from_mda(info->fmt, &mdac->area, vgnamebuf,
-				    sizeof(vgnamebuf))) {
-			lvmcache_update_vgname(info, vgnamebuf);
-		}
+		if ((vgname = vgname_from_mda(info->fmt, &mdac->area, 
+					      &vgid, &vgstatus, &creation_host)) &&
+		    !lvmcache_update_vgname_and_id(info, vgname,
+						   (char *) &vgid, vgstatus,
+						   creation_host))
+			return_0;
 	}
 
 	info->status &= ~CACHE_INVALID;

lib/label/label.c

@@ -117,15 +117,6 @@ static struct labeller *_find_labeller(struct device *dev, char *buf,
 	int found = 0;
 	char readbuf[LABEL_SCAN_SIZE];
 
-	if (!dev_open(dev)) {
-		stack;
-
-		if ((info = info_from_pvid(dev->pvid)))
-			lvmcache_update_vgname(info, ORPHAN);
-
-		return NULL;
-	}
-
 	if (!dev_read(dev, UINT64_C(0), LABEL_SCAN_SIZE, readbuf)) {
 		log_debug("%s: Failed to read label area", dev_name(dev));
 		goto out;
@@ -184,13 +175,11 @@ static struct labeller *_find_labeller(struct device *dev, char *buf,
       out:
 	if (!found) {
 		if ((info = info_from_pvid(dev->pvid)))
-			lvmcache_update_vgname(info, ORPHAN);
+			lvmcache_update_vgname_and_id(info, ORPHAN, ORPHAN,
+						      0, NULL);
 		log_very_verbose("%s: No label detected", dev_name(dev));
 	}
 
-	if (!dev_close(dev))
-		stack;
-
 	return r;
 }
 
@@ -272,16 +261,29 @@ int label_read(struct device *dev, struct label **result)
 	char buf[LABEL_SIZE];
 	struct labeller *l;
 	uint64_t sector;
-	int r;
+	struct lvmcache_info *info;
+	int r = 0;
 
-	if (!(l = _find_labeller(dev, buf, &sector))) {
+	if (!dev_open(dev)) {
 		stack;
-		return 0;
+
+		if ((info = info_from_pvid(dev->pvid)))
+			lvmcache_update_vgname_and_id(info, ORPHAN, ORPHAN,
+						      0, NULL);
+
+		goto out;
 	}
 
+	if (!(l = _find_labeller(dev, buf, &sector)))
+		goto_out;
+
 	if ((r = (l->ops->read)(l, dev, buf, result)) && result && *result)
 		(*result)->sector = sector;
 
+      out:
+	if (!dev_close(dev))
+		stack;
+
 	return r;
 }
 
@@ -335,18 +337,35 @@ int label_write(struct device *dev, struct label *label)
 	return r;
 }
 
+/* Unused */
 int label_verify(struct device *dev)
 {
 	struct labeller *l;
 	char buf[LABEL_SIZE];
 	uint64_t sector;
+	struct lvmcache_info *info;
+	int r = 0;
 
-	if (!(l = _find_labeller(dev, buf, &sector))) {
+	if (!dev_open(dev)) {
 		stack;
-		return 0;
+
+		if ((info = info_from_pvid(dev->pvid)))
+			lvmcache_update_vgname_and_id(info, ORPHAN, ORPHAN,
+						      0, NULL);
+
+		goto out;
 	}
 
-	return ((l->ops->verify) ? l->ops->verify(l, buf, sector) : 1);
+	if (!(l = _find_labeller(dev, buf, &sector)))
+		goto_out;
+
+	r = l->ops->verify ? l->ops->verify(l, buf, sector) : 1;
+
+      out:
+	if (!dev_close(dev))
+		stack;
+
+	return r;
 }
 
 void label_destroy(struct label *label)

lib/locking/cluster_locking.c

@@ -330,6 +330,9 @@ static int _lock_for_cluster(unsigned char cmd, unsigned int flags, char *name)
 	args[0] = flags & 0x7F; /* Maskoff lock flags */
 	args[1] = flags & 0xC0; /* Bitmap flags */
 
+	if (partial_mode())
+		args[1] |= LCK_PARTIAL_MODE;
+
 	/*
 	 * VG locks are just that: locks, and have no side effects
 	 * so we only need to do them on the local node because all

lib/locking/external_locking.c

@@ -72,7 +72,7 @@ int init_external_locking(struct locking_type *locking, struct config_tree *cft)
 	libname = find_config_str(cft->root, "global/locking_library",
 				  DEFAULT_LOCKING_LIB);
 
-	if (!(_locking_lib = load_shared_library(cft, libname, "locking"))) {
+	if (!(_locking_lib = load_shared_library(cft, libname, "locking", 1))) {
 		stack;
 		return 0;
 	}

lib/locking/locking.h

@@ -67,6 +67,11 @@ int check_lvm1_vg_inactive(struct cmd_context *cmd, const char *vgname);
 #define LCK_LOCAL	0x00000040	/* Don't propagate to other nodes */
 #define LCK_CLUSTER_VG	0x00000080	/* VG is clustered */
 
+/*
+ * Additional lock bits for cluster communication
+ */
+#define LCK_PARTIAL_MODE	0x00000001	/* Running in partial mode */
+
 /*
  * Common combinations
  */

lib/log/log.c

@@ -90,9 +90,13 @@ void init_syslog(int facility)
 	_syslog = 1;
 }
 
-void log_suppress(int suppress)
+int log_suppress(int suppress)
 {
+	int old_suppress = _log_suppress;
+
 	_log_suppress = suppress;
+
+	return old_suppress;
 }
 
 void release_log_memory(void)
@@ -253,6 +257,9 @@ void print_log(int level, const char *file, int line, const char *format, ...)
 	const char *message;
 	const char *trformat;		/* Translated format string */
 
+	if (_log_suppress == 2)
+		return;
+
 	trformat = _(format);
 
 	if (_lvm2_log_fn) {

lib/log/log.h

@@ -86,8 +86,9 @@ int ignorelockingfailure(void);
 int lockingfailed(void);
 int security_level(void);
 
-/* Suppress messages to stdout/stderr */
-void log_suppress(int suppress);
+/* Suppress messages to stdout/stderr (1) or everywhere (2) */
+/* Returns previous setting */
+int log_suppress(int suppress);
 
 /* Suppress messages to syslog */
 void syslog_suppress(int suppress);

lib/metadata/lv_alloc.h

@@ -52,7 +52,8 @@ struct alloc_handle *allocate_extents(struct volume_group *vg,
                                       uint32_t mirrored_pe,
                                       uint32_t status,
                                       struct list *allocatable_pvs,
-                                      alloc_policy_t alloc);
+				      alloc_policy_t alloc,
+				      struct list *parallel_areas);
 
 int lv_add_segment(struct alloc_handle *ah,
 		   uint32_t first_area, uint32_t num_areas,
@@ -83,4 +84,7 @@ int lv_add_more_mirrored_areas(struct logical_volume *lv,
 
 void alloc_destroy(struct alloc_handle *ah);
 
+struct list *build_parallel_areas_from_lv(struct cmd_context *cmd,
+					  struct logical_volume *lv);
+
 #endif

lib/metadata/lv_manip.c

@@ -1,6 +1,6 @@
 /*
  * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.
- * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ * Copyright (C) 2004-2005 Red Hat, Inc. All rights reserved.
  *
  * This file is part of LVM2.
  *
@@ -24,6 +24,18 @@
 #include "display.h"
 #include "segtype.h"
 
+/*
+ * PVs used by a segment of an LV
+ */
+struct seg_pvs {
+	struct list list;
+
+	struct list pvs;	/* struct pv_list */
+
+	uint32_t le;
+	uint32_t len;
+};
+
 /*
  * Find first unused LV number.
  */
@@ -342,13 +354,16 @@ static int _lv_reduce(struct logical_volume *lv, uint32_t extents, int delete)
 }
 
 /*
- * Empty an LV
+ * Empty an LV.
  */
 int lv_empty(struct logical_volume *lv)
 {
-	return _lv_reduce(lv, 0, lv->le_count);
+	return _lv_reduce(lv, lv->le_count, 0);
 }
 
+/*
+ * Remove given number of extents from LV.
+ */
 int lv_reduce(struct logical_volume *lv, uint32_t extents)
 {
 	return _lv_reduce(lv, extents, 1);
@@ -391,6 +406,10 @@ struct alloc_handle {
 	uint32_t log_count;		/* Number of parallel 1-extent logs */
 	uint32_t total_area_len;	/* Total number of parallel extents */
 
+	struct physical_volume *mirrored_pv;	/* FIXME Remove this */
+	uint32_t mirrored_pe;			/* FIXME Remove this */
+	struct list *parallel_areas;	/* PVs to avoid */
+
 	struct alloced_area log_area;	/* Extent used for log */
 	struct list alloced_areas[0];	/* Lists of areas in each stripe */
 };
@@ -404,7 +423,9 @@ static struct alloc_handle *_alloc_init(struct dm_pool *mem,
 					uint32_t mirrors,
 					uint32_t stripes,
 					uint32_t log_count,
-					struct physical_volume *mirrored_pv)
+					struct physical_volume *mirrored_pv,
+					uint32_t mirrored_pe,
+					struct list *parallel_areas)
 {
 	struct alloc_handle *ah;
 	uint32_t s, area_count;
@@ -458,6 +479,10 @@ static struct alloc_handle *_alloc_init(struct dm_pool *mem,
 	for (s = 0; s < ah->area_count; s++)
 		list_init(&ah->alloced_areas[s]);
 
+	ah->mirrored_pv = mirrored_pv;
+	ah->mirrored_pe = mirrored_pe;
+	ah->parallel_areas = parallel_areas;
+
 	return ah;
 }
 
@@ -646,7 +671,6 @@ static int _check_contiguous(struct lv_segment *prev_lvseg,
 /*
  * Choose sets of parallel areas to use, respecting any constraints.
  */
-/* FIXME Also accept existing areas new space must be parallel to */
 static int _find_parallel_space(struct alloc_handle *ah, alloc_policy_t alloc,
 				struct list *pvms, struct pv_area **areas,
 				uint32_t areas_size, unsigned can_split,
@@ -655,10 +679,15 @@ static int _find_parallel_space(struct alloc_handle *ah, alloc_policy_t alloc,
 {
 	struct pv_map *pvm;
 	struct pv_area *pva;
+	struct pv_list *pvl;
 	unsigned already_found_one = 0;
 	unsigned contiguous = 0, contiguous_count = 0;
 	unsigned ix;
 	unsigned ix_offset = 0;	/* Offset for non-contiguous allocations */
+	uint32_t max_parallel;	/* Maximum extents to allocate */
+	uint32_t next_le;
+	struct seg_pvs *spvs;
+	struct list *parallel_pvs;
 
 	/* FIXME Do calculations on free extent counts before selecting space */
 	/* FIXME Select log PV appropriately if there isn't one yet */
@@ -676,6 +705,25 @@ static int _find_parallel_space(struct alloc_handle *ah, alloc_policy_t alloc,
 	do {
 		ix = 0;
 
+		parallel_pvs = NULL;
+		max_parallel = needed;
+
+		/*
+		 * If there are existing parallel PVs, avoid them and reduce
+		 * the maximum we can allocate in one go accordingly.
+		 */
+		if (ah->parallel_areas) {
+			list_iterate_items(spvs, ah->parallel_areas) {
+				next_le = (prev_lvseg ? prev_lvseg->le + prev_lvseg->len : 0) + *allocated;
+				if (next_le >= spvs->le) {
+					if (next_le + max_parallel > spvs->le + spvs->len)
+						max_parallel = (spvs->le + spvs->len - next_le) * ah->area_multiple;
+					parallel_pvs = &spvs->pvs;
+					break;
+				}
+			}
+		}
+
 		/*
 		 * Put the smallest area of each PV that is at least the
 		 * size we need into areas array.  If there isn't one
@@ -686,10 +734,18 @@ static int _find_parallel_space(struct alloc_handle *ah, alloc_policy_t alloc,
 			if (list_empty(&pvm->areas))
 				continue;	/* Next PV */
 
-			/* Don't allocate onto the log pv */
-			if ((alloc != ALLOC_ANYWHERE) && ah->log_count &&
-			    (pvm->pv == ah->log_area.pv))
-				continue;	/* Next PV */
+			if (alloc != ALLOC_ANYWHERE) {
+				/* Don't allocate onto the log pv */
+				if (ah->log_count &&
+				    pvm->pv == ah->log_area.pv)
+					continue;	/* Next PV */
+
+				/* Avoid PVs used by existing parallel areas */
+				if (parallel_pvs)
+					list_iterate_items(pvl, parallel_pvs)
+						if (pvm->pv == pvl->pv)
+							goto next_pv;
+			}
 
 			already_found_one = 0;
 			/* First area in each list is the largest */
@@ -700,17 +756,17 @@ static int _find_parallel_space(struct alloc_handle *ah, alloc_policy_t alloc,
 							      pvm->pv,
 							      pva, areas)) {
 						contiguous_count++;
-						break; /* Next PV */
+						goto next_pv;
 					}
 					continue;
 				}
 
 				/* Is it big enough on its own? */
-				if ((pva->count < needed - *allocated) &&
+				if ((pva->count < max_parallel - *allocated) &&
 				    ((!can_split && !ah->log_count) ||
 				     (already_found_one &&
 				      !(alloc == ALLOC_ANYWHERE))))
-					break;	/* Next PV */
+					goto next_pv;
 
 				if (!already_found_one ||
 				    alloc == ALLOC_ANYWHERE) {
@@ -720,8 +776,9 @@ static int _find_parallel_space(struct alloc_handle *ah, alloc_policy_t alloc,
 
 				areas[ix + ix_offset - 1] = pva;
 
-				break;	/* Next PV */
+				goto next_pv;
 			}
+		next_pv:
 			if (ix >= areas_size)
 				break;
 		}
@@ -743,7 +800,7 @@ static int _find_parallel_space(struct alloc_handle *ah, alloc_policy_t alloc,
 
 		/* First time around, use smallest area as log_area */
 		/* FIXME decide which PV to use at top of function instead */
-		if (!_alloc_parallel_area(ah, needed, areas,
+		if (!_alloc_parallel_area(ah, max_parallel, areas,
 					  allocated,
 					  (ah->log_count && !ah->log_area.len) ?
 						*(areas + ix_offset + ix - 1) :
@@ -764,13 +821,11 @@ static int _find_parallel_space(struct alloc_handle *ah, alloc_policy_t alloc,
  */
 static int _allocate(struct alloc_handle *ah,
 		     struct volume_group *vg,
-			   struct logical_volume *lv, uint32_t status,
-			   uint32_t new_extents,
-			   struct list *allocatable_pvs,
-			   uint32_t stripes, uint32_t mirrors,
-			   struct segment_type *segtype,
-			   struct physical_volume *mirrored_pv,
-			   uint32_t mirrored_pe)
+		     struct logical_volume *lv, uint32_t status,
+		     uint32_t new_extents,
+		     struct list *allocatable_pvs,
+		     uint32_t stripes, uint32_t mirrors,
+		     struct segment_type *segtype)
 {
 	struct pv_area **areas;
 	uint32_t allocated = lv ? lv->le_count : 0;
@@ -781,12 +836,12 @@ static int _allocate(struct alloc_handle *ah,
 	struct list *pvms;
 	uint32_t areas_size;
 
-	if (allocated >= new_extents) {
+	if (allocated >= new_extents && !ah->log_count) {
 		log_error("_allocate called with no work to do!");
 		return 1;
 	}
 
-	if (mirrored_pv || (ah->alloc == ALLOC_CONTIGUOUS))
+	if (ah->mirrored_pv || (ah->alloc == ALLOC_CONTIGUOUS))
 		can_split = 0;
 
 	if (lv && !list_empty(&lv->segments))
@@ -906,7 +961,8 @@ struct alloc_handle *allocate_extents(struct volume_group *vg,
 				      uint32_t mirrored_pe,
 				      uint32_t status,
 				      struct list *allocatable_pvs,
-				      alloc_policy_t alloc)
+				      alloc_policy_t alloc,
+				      struct list *parallel_areas)
 {
 	struct alloc_handle *ah;
 
@@ -929,15 +985,15 @@ struct alloc_handle *allocate_extents(struct volume_group *vg,
 		alloc = vg->alloc;
 
 	if (!(ah = _alloc_init(vg->cmd->mem, segtype, alloc, mirrors,
-			       stripes, log_count, mirrored_pv))) {
+			       stripes, log_count, mirrored_pv,
+			       mirrored_pe, parallel_areas))) {
 		stack;
 		return NULL;
 	}
 
 	if (!segtype_is_virtual(segtype) &&
 	    !_allocate(ah, vg, lv, status, (lv ? lv->le_count : 0) + extents,
-		       allocatable_pvs,
-		       stripes, mirrors, segtype, mirrored_pv, mirrored_pe)) {
+		       allocatable_pvs, stripes, mirrors, segtype)) {
 		stack;
 		alloc_destroy(ah);
 		return NULL;
@@ -1098,8 +1154,7 @@ int lv_add_more_mirrored_areas(struct logical_volume *lv,
 		return 0;
 	}
 
-	list_iterate_items(seg, &lv->segments)
-		break;
+	seg = first_seg(lv);
 
 	old_area_count = seg->area_count;
 	new_area_count = old_area_count + num_extra_areas;
@@ -1139,7 +1194,7 @@ int lv_extend(struct logical_volume *lv,
 
 	if (!(ah = allocate_extents(lv->vg, lv, segtype, stripes, mirrors, 0,
 				    extents, mirrored_pv, mirrored_pe, status,
-				    allocatable_pvs, alloc))) {
+				    allocatable_pvs, alloc, NULL))) {
 		stack;
 		return 0;
 	}
@@ -1272,3 +1327,116 @@ struct logical_volume *lv_create_empty(struct format_instance *fi,
 
 	return lv;
 }
+
+/* Recursively process each PV used by part of an LV */
+static int _for_each_pv(struct cmd_context *cmd, struct logical_volume *lv,
+			uint32_t le, uint32_t len,
+			int (*fn)(struct cmd_context *cmd, struct pv_segment *peg, struct seg_pvs *spvs),
+			struct seg_pvs *spvs)
+{
+	struct lv_segment *seg;
+	uint32_t s;
+	uint32_t remaining_seg_len, area_len, area_multiple;
+
+	if (!(seg = find_seg_by_le(lv, le))) {
+		log_error("Failed to find segment for %s extent %" PRIu32,
+			  lv->name, le);
+		return 0;
+	}
+
+	/* Remaining logical length of segment */
+	remaining_seg_len = seg->len - (le - seg->le);
+
+	if (len > remaining_seg_len)
+		remaining_seg_len = len;
+
+	if (spvs->len > remaining_seg_len)
+		spvs->len = remaining_seg_len;
+
+	area_multiple = segtype_is_striped(seg->segtype) ? seg->area_count : 1;
+	area_len = remaining_seg_len / area_multiple;
+
+	for (s = 0; s < seg->area_count; s++) {
+		if (seg_type(seg, s) == AREA_LV) {
+			if (!_for_each_pv(cmd, seg_lv(seg, s),
+					  seg_le(seg, s) + (le - seg->le) / area_multiple,
+					  area_len, fn, spvs)) {
+				stack;
+				return 0;
+			}
+		} else if (seg_type(seg, s) == AREA_PV) {
+			if (!fn(cmd, seg_pvseg(seg, s), spvs)) {
+				stack;
+				return 0;
+			}
+		}
+	}
+
+	return 1;
+}
+
+static int _add_pvs(struct cmd_context *cmd, struct pv_segment *peg, struct seg_pvs *spvs)
+{
+	struct pv_list *pvl;
+
+	/* FIXME Don't add again if it's already on the list! */
+
+	if (!(pvl = dm_pool_alloc(cmd->mem, sizeof(*pvl)))) {
+		log_error("pv_list allocation failed");
+		return 0;
+	}
+
+	pvl->pv = peg->pv;
+
+	/* FIXME Use ordered list to facilitate comparison */
+	list_add(&spvs->pvs, &pvl->list);
+
+	/* FIXME Add mirror logs, snapshot cow LVs etc. */
+
+	return 1;
+}
+
+/*
+ * Construct list of segments of LVs showing which PVs they use.
+ */
+struct list *build_parallel_areas_from_lv(struct cmd_context *cmd,
+					  struct logical_volume *lv)
+{
+	struct list *parallel_areas;
+	struct seg_pvs *spvs;
+	uint32_t current_le = 0;
+
+	if (!(parallel_areas = dm_pool_alloc(cmd->mem, sizeof(*parallel_areas)))) {
+		log_error("parallel_areas allocation failed");
+		return NULL;
+	}
+
+	list_init(parallel_areas);
+
+	do {
+		if (!(spvs = dm_pool_zalloc(cmd->mem, sizeof(*spvs)))) {
+			log_error("allocation failed");
+			return NULL;
+		}
+
+		list_init(&spvs->pvs);
+
+		spvs->le = current_le;
+		spvs->len = lv->le_count - current_le;
+
+		list_add(parallel_areas, &spvs->list);
+
+		/* Find next segment end */
+		/* FIXME Unnecessary nesting! */
+		if (!_for_each_pv(cmd, lv, current_le, lv->le_count, _add_pvs, spvs)) {
+			stack;
+			return NULL;
+		}
+
+		current_le = spvs->le + spvs->len;
+	} while (current_le < lv->le_count);
+
+	/* FIXME Merge adjacent segments with identical PV lists (avoids need for contiguous allocation attempts between successful allocations) */
+
+	return parallel_areas;
+}

lib/metadata/metadata.c

@@ -71,6 +71,8 @@ static int _add_pv_to_vg(struct format_instance *fid, struct volume_group *vg,
 		return 0;
 	}
 
+	memcpy(&pv->vgid, &vg->id, sizeof(vg->id));
+
 	/* Units of 512-byte sectors */
 	pv->pe_size = vg->extent_size;
 
@@ -143,13 +145,14 @@ static int _copy_pv(struct physical_volume *pv_to,
 }
 
 int get_pv_from_vg_by_id(const struct format_type *fmt, const char *vg_name,
-			 const char *id, struct physical_volume *pv)
+			 const char *vgid, const char *pvid,
+			 struct physical_volume *pv)
 {
 	struct volume_group *vg;
 	struct pv_list *pvl;
 	int consistent = 0;
 
-	if (!(vg = vg_read(fmt->cmd, vg_name, &consistent))) {
+	if (!(vg = vg_read(fmt->cmd, vg_name, vgid, &consistent))) {
 		log_error("get_pv_from_vg_by_id: vg_read failed to read VG %s",
 			  vg_name);
 		return 0;
@@ -160,7 +163,7 @@ int get_pv_from_vg_by_id(const struct format_type *fmt, const char *vg_name,
 			  vg_name);
 
 	list_iterate_items(pvl, &vg->pvs) {
-		if (id_equal(&pvl->pv->id, (const struct id *) id)) {
+		if (id_equal(&pvl->pv->id, (const struct id *) pvid)) {
 			if (!_copy_pv(pv, pvl->pv)) {
 				stack;
 				return 0;
@@ -239,7 +242,7 @@ struct volume_group *vg_create(struct cmd_context *cmd, const char *vg_name,
 	/* is this vg name already in use ? */
 	old_partial = partial_mode();
 	init_partial(1);
-	if (vg_read(cmd, vg_name, &consistent)) {
+	if (vg_read(cmd, vg_name, NULL, &consistent)) {
 		log_err("A volume group called '%s' already exists.", vg_name);
 		goto bad;
 	}
@@ -286,7 +289,7 @@ struct volume_group *vg_create(struct cmd_context *cmd, const char *vg_name,
 	list_init(&vg->tags);
 
 	if (!(vg->fid = cmd->fmt->ops->create_instance(cmd->fmt, vg_name,
-						       NULL))) {
+						       NULL, NULL))) {
 		log_error("Failed to create format instance");
 		goto bad;
 	}
@@ -496,7 +499,7 @@ struct physical_volume *pv_create(const struct format_type *fmt,
 				  uint64_t pvmetadatasize, struct list *mdas)
 {
 	struct dm_pool *mem = fmt->cmd->mem;
-	struct physical_volume *pv = dm_pool_alloc(mem, sizeof(*pv));
+	struct physical_volume *pv = dm_pool_zalloc(mem, sizeof(*pv));
 
 	if (!pv) {
 		stack;
@@ -862,7 +865,7 @@ static struct volume_group *_vg_read_orphans(struct cmd_context *cmd)
 	struct volume_group *vg;
 	struct physical_volume *pv;
 
-	if (!(vginfo = vginfo_from_vgname(ORPHAN))) {
+	if (!(vginfo = vginfo_from_vgname(ORPHAN, NULL))) {
 		stack;
 		return NULL;
 	}
@@ -907,6 +910,7 @@ static struct volume_group *_vg_read_orphans(struct cmd_context *cmd)
  */
 static struct volume_group *_vg_read(struct cmd_context *cmd,
 				     const char *vgname,
+				     const char *vgid,
 				     int *consistent, int precommitted)
 {
 	struct format_instance *fid;
@@ -928,15 +932,15 @@ static struct volume_group *_vg_read(struct cmd_context *cmd,
 
 	/* Find the vgname in the cache */
 	/* If it's not there we must do full scan to be completely sure */
-	if (!(fmt = fmt_from_vgname(vgname))) {
+	if (!(fmt = fmt_from_vgname(vgname, vgid))) {
 		lvmcache_label_scan(cmd, 0);
-		if (!(fmt = fmt_from_vgname(vgname))) {
+		if (!(fmt = fmt_from_vgname(vgname, vgid))) {
 			if (memlock()) {
 				stack;
 				return NULL;
 			}
 			lvmcache_label_scan(cmd, 2);
-			if (!(fmt = fmt_from_vgname(vgname))) {
+			if (!(fmt = fmt_from_vgname(vgname, vgid))) {
 				stack;
 				return NULL;
 			}
@@ -947,7 +951,7 @@ static struct volume_group *_vg_read(struct cmd_context *cmd,
 		use_precommitted = 0;
 
 	/* create format instance with appropriate metadata area */
-	if (!(fid = fmt->ops->create_instance(fmt, vgname, NULL))) {
+	if (!(fid = fmt->ops->create_instance(fmt, vgname, vgid, NULL))) {
 		log_error("Failed to create format instance");
 		return NULL;
 	}
@@ -978,7 +982,7 @@ static struct volume_group *_vg_read(struct cmd_context *cmd,
 		inconsistent = 0;
 
 		lvmcache_label_scan(cmd, 2);
-		if (!(fmt = fmt_from_vgname(vgname))) {
+		if (!(fmt = fmt_from_vgname(vgname, vgid))) {
 			stack;
 			return NULL;
 		}
@@ -987,7 +991,7 @@ static struct volume_group *_vg_read(struct cmd_context *cmd,
 			use_precommitted = 0;
 
 		/* create format instance with appropriate metadata area */
-		if (!(fid = fmt->ops->create_instance(fmt, vgname, NULL))) {
+		if (!(fid = fmt->ops->create_instance(fmt, vgname, vgid, NULL))) {
 			log_error("Failed to create format instance");
 			return NULL;
 		}
@@ -1069,12 +1073,12 @@ static struct volume_group *_vg_read(struct cmd_context *cmd,
 }
 
 struct volume_group *vg_read(struct cmd_context *cmd, const char *vgname,
-			     int *consistent)
+			     const char *vgid, int *consistent)
 {
 	struct volume_group *vg;
 	struct lv_list *lvl;
 
-	if (!(vg = _vg_read(cmd, vgname, consistent, 0)))
+	if (!(vg = _vg_read(cmd, vgname, vgid, consistent, 0)))
 		return NULL;
 
 	if (!check_pv_segments(vg)) {
@@ -1112,7 +1116,7 @@ static struct volume_group *_vg_read_by_vgid(struct cmd_context *cmd,
 	/* Is corresponding vgname already cached? */
 	if ((vginfo = vginfo_from_vgid(vgid)) &&
 	    vginfo->vgname && *vginfo->vgname) {
-		if ((vg = _vg_read(cmd, vginfo->vgname,
+		if ((vg = _vg_read(cmd, vginfo->vgname, vgid,
 				   &consistent, precommitted)) &&
 		    !strncmp(vg->id.uuid, vgid, ID_LEN)) {
 			if (!consistent) {
@@ -1143,7 +1147,7 @@ static struct volume_group *_vg_read_by_vgid(struct cmd_context *cmd,
 		if (!vgname || !*vgname)
 			continue;	// FIXME Unnecessary? 
 		consistent = 0;
-		if ((vg = _vg_read(cmd, vgname, &consistent,
+		if ((vg = _vg_read(cmd, vgname, vgid, &consistent,
 				   precommitted)) &&
 		    !strncmp(vg->id.uuid, vgid, ID_LEN)) {
 			if (!consistent) {
@@ -1245,13 +1249,18 @@ struct list *get_vgs(struct cmd_context *cmd, int full_scan)
 	return lvmcache_get_vgnames(cmd, full_scan);
 }
 
+struct list *get_vgids(struct cmd_context *cmd, int full_scan)
+{
+	return lvmcache_get_vgids(cmd, full_scan);
+}
+
 struct list *get_pvs(struct cmd_context *cmd)
 {
 	struct str_list *strl;
 	struct list *results;
-	const char *vgname;
+	const char *vgname, *vgid;
 	struct list *pvh, *tmp;
-	struct list *vgnames;
+	struct list *vgids;
 	struct volume_group *vg;
 	int consistent = 0;
 	int old_partial;
@@ -1267,7 +1276,7 @@ struct list *get_pvs(struct cmd_context *cmd)
 	list_init(results);
 
 	/* Get list of VGs */
-	if (!(vgnames = get_vgs(cmd, 0))) {
+	if (!(vgids = get_vgids(cmd, 0))) {
 		log_error("get_pvs: get_vgs failed");
 		return NULL;
 	}
@@ -1278,12 +1287,16 @@ struct list *get_pvs(struct cmd_context *cmd)
 	old_pvmove = pvmove_mode();
 	init_partial(1);
 	init_pvmove(1);
-	list_iterate_items(strl, vgnames) {
-		vgname = strl->str;
-		if (!vgname)
+	list_iterate_items(strl, vgids) {
+		vgid = strl->str;
+		if (!vgid)
 			continue;	/* FIXME Unnecessary? */
 		consistent = 0;
-		if (!(vg = vg_read(cmd, vgname, &consistent))) {
+		if (!(vgname = vgname_from_vgid(NULL, vgid))) {
+			stack;
+			continue;
+		}
+		if (!(vg = vg_read(cmd, vgname, vgid, &consistent))) {
 			stack;
 			continue;
 		}

lib/metadata/metadata.h

@@ -122,6 +122,7 @@ struct physical_volume {
 	struct device *dev;
 	const struct format_type *fmt;
 	const char *vg_name;
+	struct id vgid;
 
 	uint32_t status;
 	uint64_t size;
@@ -382,6 +383,7 @@ struct format_handler {
 	 */
 	struct format_instance *(*create_instance) (const struct format_type *
 						    fmt, const char *vgname,
+						    const char *vgid,
 						    void *context);
 
 	/*
@@ -403,7 +405,7 @@ int vg_write(struct volume_group *vg);
 int vg_commit(struct volume_group *vg);
 int vg_revert(struct volume_group *vg);
 struct volume_group *vg_read(struct cmd_context *cmd, const char *vg_name,
-			     int *consistent);
+			     const char *vgid, int *consistent);
 struct physical_volume *pv_read(struct cmd_context *cmd, const char *pv_name,
 				struct list *mdas, uint64_t *label_sector,
 				int warnings);
@@ -411,6 +413,7 @@ struct list *get_pvs(struct cmd_context *cmd);
 
 /* Set full_scan to 1 to re-read every (filtered) device label */
 struct list *get_vgs(struct cmd_context *cmd, int full_scan);
+struct list *get_vgids(struct cmd_context *cmd, int full_scan);
 
 int pv_write(struct cmd_context *cmd, struct physical_volume *pv,
 	     struct list *mdas, int64_t label_sector);
@@ -478,7 +481,8 @@ struct pv_list *find_pv_in_vg(struct volume_group *vg, const char *pv_name);
 struct physical_volume *find_pv_in_vg_by_uuid(struct volume_group *vg,
 					      struct id *id);
 int get_pv_from_vg_by_id(const struct format_type *fmt, const char *vg_name,
-			 const char *id, struct physical_volume *pv);
+			 const char *vgid, const char *pvid,
+			 struct physical_volume *pv);
 
 /* Find an LV within a given VG */
 struct lv_list *find_lv_in_vg(struct volume_group *vg, const char *lv_name);
@@ -535,11 +539,16 @@ int lv_split_segment(struct logical_volume *lv, uint32_t le);
  */
 int lv_is_origin(const struct logical_volume *lv);
 int lv_is_cow(const struct logical_volume *lv);
+int lv_is_visible(const struct logical_volume *lv);
 
 int pv_is_in_vg(struct volume_group *vg, struct physical_volume *pv);
 
+/* Given a cow LV, return return the snapshot lv_segment that uses it */
 struct lv_segment *find_cow(const struct logical_volume *lv);
 
+/* Given a cow LV, return its origin */
+struct logical_volume *origin_from_cow(const struct logical_volume *lv);
+
 int vg_add_snapshot(struct format_instance *fid, const char *name,
 		    struct logical_volume *origin, struct logical_volume *cow,
 		    union lvid *lvid, uint32_t extent_count,
@@ -561,8 +570,14 @@ int create_mirror_layers(struct alloc_handle *ah,
 			 uint32_t status,
 			 uint32_t region_size,
 			 struct logical_volume *log_lv);
-int remove_mirror_images(struct lv_segment *mirrored_seg, uint32_t num_mirrors);
-int remove_all_mirror_images(struct logical_volume *lv);
+int add_mirror_layers(struct alloc_handle *ah,
+		      uint32_t num_mirrors,
+		      uint32_t existing_mirrors,
+		      struct logical_volume *lv,
+		      struct segment_type *segtype);
+
+int remove_mirror_images(struct lv_segment *mirrored_seg, uint32_t num_mirrors,
+			 struct list *removable_pvs, int remove_log);
 /*
  * Given mirror image or mirror log segment, find corresponding mirror segment 
  */

lib/metadata/mirror.c

@@ -1,6 +1,6 @@
 /*
  * Copyright (C) 2003-2004 Sistina Software, Inc. All rights reserved.  
- * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ * Copyright (C) 2004-2005 Red Hat, Inc. All rights reserved.
  *
  * This file is part of LVM2.
  *
@@ -21,6 +21,7 @@
 #include "activate.h"
 #include "lv_alloc.h"
 #include "lvm-string.h"
+#include "locking.h"	/* FIXME Should not be used in this file */
 
 struct lv_segment *find_mirror_seg(struct lv_segment *seg)
 {
@@ -62,6 +63,9 @@ static void _move_lv_segments(struct logical_volume *lv_to, struct logical_volum
 
 	list_init(&lv_from->segments);
 
+	lv_to->le_count = lv_from->le_count;
+	lv_to->size = lv_from->size;
+
 	lv_from->le_count = 0;
 	lv_from->size = 0;
 }
@@ -69,65 +73,163 @@ static void _move_lv_segments(struct logical_volume *lv_to, struct logical_volum
 /*
  * Reduce mirrored_seg to num_mirrors images.
  */
-int remove_mirror_images(struct lv_segment *mirrored_seg, uint32_t num_mirrors)
+int remove_mirror_images(struct lv_segment *mirrored_seg, uint32_t num_mirrors,
+			 struct list *removable_pvs, int remove_log)
 {
 	uint32_t m;
+	uint32_t s, s1;
+	struct logical_volume *sub_lv;
+	struct logical_volume *log_lv = NULL;
+	struct logical_volume *lv1 = NULL;
+	struct physical_volume *pv;
+	struct lv_segment *seg;
+	struct lv_segment_area area;
+	int all_pvs_removable, pv_found;
+	struct pv_list *pvl;
+	uint32_t old_area_count = mirrored_seg->area_count;
+	uint32_t new_area_count = mirrored_seg->area_count;
+
+	log_very_verbose("Reducing mirror set from %" PRIu32 " to %"
+			 PRIu32 " image(s)%s.",
+			 old_area_count, num_mirrors,
+			 remove_log ? " and no log volume" : "");
+
+	/* Move removable_pvs to end of array */
+	if (removable_pvs) {
+		for (s = 0; s < mirrored_seg->area_count; s++) {
+			all_pvs_removable = 1;
+			sub_lv = seg_lv(mirrored_seg, s);
+			list_iterate_items(seg, &sub_lv->segments) {
+				for (s1 = 0; s1 < seg->area_count; s1++) {
+					if (seg_type(seg, s1) != AREA_PV)
+						/* FIXME Recurse for AREA_LV */
+						continue;
+
+					pv = seg_pv(seg, s1);
+
+					pv_found = 0;
+					list_iterate_items(pvl, removable_pvs) {
+						if (pv->dev->dev == pvl->pv->dev->dev) {
+							pv_found = 1;
+							break;
+						}
+					}
+					if (!pv_found) {
+						all_pvs_removable = 0;
+						break;
+					}
+				}
+				if (!all_pvs_removable)
+					break;
+			}
+			if (all_pvs_removable) {
+				/* Swap segment to end */
+				new_area_count--;
+				area = mirrored_seg->areas[new_area_count];
+				mirrored_seg->areas[new_area_count] = mirrored_seg->areas[s];
+				mirrored_seg->areas[s] = area;
+			}
+			/* Found enough matches? */
+			if (new_area_count == num_mirrors)
+				break;
+		}
+		if (new_area_count == mirrored_seg->area_count) {
+			log_error("No mirror images found using specified PVs.");
+			return 0;
+		}
+	}
 
 	for (m = num_mirrors; m < mirrored_seg->area_count; m++) {
+		seg_lv(mirrored_seg, m)->status &= ~MIRROR_IMAGE;
+		seg_lv(mirrored_seg, m)->status |= VISIBLE_LV;
+	}
+
+	mirrored_seg->area_count = num_mirrors;
+
+	/* If no more mirrors, remove mirror layer */
+	if (num_mirrors == 1) {
+		lv1 = seg_lv(mirrored_seg, 0);
+		_move_lv_segments(mirrored_seg->lv, lv1);
+		mirrored_seg->lv->status &= ~MIRRORED;
+		remove_log = 1;
+	}
+
+	if (remove_log) {
+		log_lv = mirrored_seg->log_lv;
+		mirrored_seg->log_lv = NULL;
+	}
+
+	/*
+	 * To successfully remove these unwanted LVs we need to
+	 * remove the LVs from the mirror set, commit that metadata
+	 * then deactivate and remove them fully.
+	 */
+
+	/* FIXME lv1 has no segments here so shouldn't be written to disk! */
+
+	if (!vg_write(mirrored_seg->lv->vg)) {
+		log_error("intermediate VG write failed.");
+		return 0;
+	}
+
+	if (!suspend_lv(mirrored_seg->lv->vg->cmd, mirrored_seg->lv)) {
+		log_error("Failed to lock %s", mirrored_seg->lv->name);
+		vg_revert(mirrored_seg->lv->vg);
+		return 0;
+	}
+
+	if (!vg_commit(mirrored_seg->lv->vg)) {
+		resume_lv(mirrored_seg->lv->vg->cmd, mirrored_seg->lv);
+		return 0;
+	}
+
+	log_very_verbose("Updating \"%s\" in kernel", mirrored_seg->lv->name);
+
+	if (!resume_lv(mirrored_seg->lv->vg->cmd, mirrored_seg->lv)) {
+		log_error("Problem reactivating %s", mirrored_seg->lv->name);
+		return 0;
+	}
+
+	/* Delete the 'orphan' LVs */
+	for (m = num_mirrors; m < old_area_count; m++) {
+		if (!deactivate_lv(mirrored_seg->lv->vg->cmd, seg_lv(mirrored_seg, m))) {
+			stack;
+			return 0;
+		}
+
 		if (!lv_remove(seg_lv(mirrored_seg, m))) {
 			stack;
 			return 0;
 		}
 	}
 
-	mirrored_seg->area_count = num_mirrors;
+	if (lv1) {
+		if (!deactivate_lv(mirrored_seg->lv->vg->cmd, lv1)) {
+			stack;
+			return 0;
+		}
+
+		if (!lv_remove(lv1)) {
+			stack;
+			return 0;
+		}
+	}
+
+	if (log_lv) {
+		if (!deactivate_lv(mirrored_seg->lv->vg->cmd, log_lv)) {
+			stack;
+			return 0;
+		}
+
+		if (!lv_remove(log_lv)) {
+			stack;
+			return 0;
+		}
+	}
 
 	return 1;
 }
 
-int remove_all_mirror_images(struct logical_volume *lv)
-{
-	struct lv_segment *seg;
-	struct logical_volume *lv1;
-
-	seg = first_seg(lv);
-
-	if (!remove_mirror_images(seg, 1)) {
-		stack;
-		return 0;
-	}
-
-	if (seg->log_lv && !lv_remove(seg->log_lv)) {
-		stack;
-		return 0;
-	}
-
-	lv1 = seg_lv(seg, 0);
-
-	_move_lv_segments(lv, lv1);
-
-	if (!lv_remove(lv1)) {
-		stack;
-		return 0;
-	}
-
-	lv->status &= ~MIRRORED;
-
-	return 1;
-}
-
-/*
- * Add mirror images to an existing mirror
- */
-/* FIXME
-int add_mirror_images(struct alloc_handle *ah,
-		      uint32_t first_area,
-		      uint32_t num_areas,
-		      struct logical_volume *lv)
-{
-}
-*/
-
 static int _create_layers_for_mirror(struct alloc_handle *ah,
 				     uint32_t first_area,
 				     uint32_t num_mirrors,
@@ -161,7 +263,10 @@ static int _create_layers_for_mirror(struct alloc_handle *ah,
 			return 0;
 		}
 
-		if (!lv_add_segment(ah, m, 1, img_lvs[m],
+		if (m < first_area)
+			continue;
+
+		if (!lv_add_segment(ah, m - first_area, 1, img_lvs[m],
 				    get_segtype_from_string(lv->vg->cmd,
 							    "striped"),
 				    0, NULL, 0, 0, 0, NULL)) {
@@ -219,6 +324,30 @@ int create_mirror_layers(struct alloc_handle *ah,
 	return 1;
 }
 
+int add_mirror_layers(struct alloc_handle *ah,
+		      uint32_t num_mirrors,
+		      uint32_t existing_mirrors,
+		      struct logical_volume *lv,
+		      struct segment_type *segtype)
+{
+	struct logical_volume **img_lvs;
+
+	if (!(img_lvs = alloca(sizeof(*img_lvs) * num_mirrors))) {
+		log_error("img_lvs allocation failed. "
+			  "Remove new LV and retry.");
+		return 0;
+	}
+
+	if (!_create_layers_for_mirror(ah, 0, num_mirrors,
+				       lv, segtype,
+				       img_lvs)) {
+		stack;
+		return 0;
+	}
+
+	return lv_add_more_mirrored_areas(lv, img_lvs, num_mirrors, 0);
+}
+
 /* 
  * Replace any LV segments on given PV with temporary mirror.
  * Returns list of LVs changed.

lib/metadata/segtype.h

@@ -78,6 +78,12 @@ struct segtype_handler {
 			       uint64_t *total_denominator, float *percent);
 	int (*target_present) (void);
 	void (*destroy) (const struct segment_type * segtype);
+	int (*target_register_events) (struct dm_pool *mem,
+				       struct lv_segment *seg,
+				       struct config_tree *cft, int events);
+	int (*target_unregister_events) (struct dm_pool *mem,
+					 struct lv_segment *seg,
+					 struct config_tree *cft, int events);
 };
 
 struct segment_type *get_segtype_from_string(struct cmd_context *cmd,

lib/metadata/snapshot_manip.c

@@ -28,12 +28,26 @@ int lv_is_cow(const struct logical_volume *lv)
 	return lv->snapshot ? 1 : 0;
 }
 
+int lv_is_visible(const struct logical_volume *lv)
+{
+	if (lv_is_cow(lv))
+		return lv_is_visible(find_cow(lv)->lv);
+
+	return lv->status & VISIBLE_LV ? 1 : 0;
+}
+
 /* Given a cow LV, return the snapshot lv_segment that uses it */
 struct lv_segment *find_cow(const struct logical_volume *lv)
 {
 	return lv->snapshot;
 }
 
+/* Given a cow LV, return its origin */
+struct logical_volume *origin_from_cow(const struct logical_volume *lv)
+{
+	return lv->snapshot->origin;
+}
+
 int vg_add_snapshot(struct format_instance *fid, const char *name,
 		    struct logical_volume *origin,
 		    struct logical_volume *cow, union lvid *lvid,

lib/mirror/mirrored.c

@@ -25,6 +25,13 @@
 #include "lvm-string.h"
 #include "targets.h"
 #include "activate.h"
+#include "sharedlib.h"
+
+#ifdef DMEVENTD
+#  include <libdevmapper-event.h>
+#endif
+
+static int _block_on_error_available = 0;
 
 enum {
 	MIRR_DISABLED,
@@ -221,6 +228,7 @@ static int _add_log(struct dev_manager *dm, struct lv_segment *seg,
 {
 	unsigned clustered = 0;
 	char *log_dlid = NULL;
+	uint32_t log_flags = 0;
 
 	/*
 	 * Use clustered mirror log for non-exclusive activation 
@@ -237,8 +245,10 @@ static int _add_log(struct dev_manager *dm, struct lv_segment *seg,
 		return 0;
 	}
 
-	/* FIXME Add sync parm? */
-	return dm_tree_node_add_mirror_target_log(node, region_size, clustered, log_dlid, area_count);
+	if (_block_on_error_available && !(seg->status & PVMOVE))
+		log_flags |= DM_BLOCK_ON_ERROR;
+
+	return dm_tree_node_add_mirror_target_log(node, region_size, clustered, log_dlid, area_count, log_flags);
 }
 
 static int _add_target_line(struct dev_manager *dm, struct dm_pool *mem,
@@ -316,15 +326,115 @@ static int _target_present(void)
 {
 	static int checked = 0;
 	static int present = 0;
+	uint32_t maj, min, patchlevel;
+	unsigned maj2, min2, patchlevel2;
+        char vsn[80];
 
-	if (!checked)
+	if (!checked) {
 		present = target_present("mirror", 1);
 
+		/*
+		 * block_on_error available with mirror target >= 1.1
+		 * or with 1.0 in RHEL4U3 driver >= 4.5
+		 */
+		/* FIXME Move this into libdevmapper */
+
+		if (target_version("mirror", &maj, &min, &patchlevel) &&
+		    maj == 1 && 
+		    (min >= 1 || 
+		     (min == 0 && driver_version(vsn, sizeof(vsn)) &&
+		      sscanf(vsn, "%u.%u.%u", &maj2, &min2, &patchlevel2) == 3 &&
+		      maj2 == 4 && min2 == 5 && patchlevel2 == 0)))	/* RHEL4U3 */
+			_block_on_error_available = 1;
+	}
+
 	checked = 1;
 
 	return present;
 }
-#endif
+
+#ifdef DMEVENTD
+static int _setup_registration(struct dm_pool *mem, struct config_tree *cft,
+			       char **dso)
+{
+	char *path;
+	const char *libpath;
+
+	if (!(path = dm_pool_alloc(mem, PATH_MAX))) {
+		log_error("Failed to allocate dmeventd library path.");
+		return 0;
+	}
+
+	libpath = find_config_str(cft->root, "dmeventd/mirror_library",
+				  DEFAULT_DMEVENTD_MIRROR_LIB);
+
+	get_shared_library_path(cft, libpath, path, PATH_MAX);
+
+	*dso = path;
+
+	return 1;
+}
+
+/* FIXME This gets run while suspended and performs banned operations. */
+/* FIXME Merge these two functions */
+static int _target_register_events(struct dm_pool *mem,
+				   struct lv_segment *seg,
+				   struct config_tree *cft, int events)
+{
+	char *dso, *name;
+	struct logical_volume *lv;
+	struct volume_group *vg;
+
+	lv = seg->lv;
+	vg = lv->vg;
+
+	if (!_setup_registration(mem, cft, &dso)) {
+		stack;
+		return 0;
+	}
+
+	if (!(name = build_dm_name(mem, vg->name, lv->name, NULL)))
+		return_0;
+
+	/* FIXME Save a returned handle here so we can unregister it later */
+	if (!dm_event_register(dso, name, DM_EVENT_ALL_ERRORS))
+		return_0;
+
+	log_info("Registered %s for events", name);
+
+	return 1;
+}
+
+static int _target_unregister_events(struct dm_pool *mem,
+				     struct lv_segment *seg,
+				     struct config_tree *cft, int events)
+{
+	char *dso;
+	char *name;
+	struct logical_volume *lv;
+	struct volume_group *vg;
+
+	lv = seg->lv;
+	vg = lv->vg;
+
+	/* FIXME Remove this and use handle to avoid config file race */
+	if (!_setup_registration(mem, cft, &dso))
+		return_0;
+
+	if (!(name = build_dm_name(mem, vg->name, lv->name, NULL)))
+		return_0;
+
+	/* FIXME Use handle returned by registration function instead of dso */
+	if (!dm_event_unregister(dso, name, DM_EVENT_ALL_ERRORS))
+		return_0;
+
+	log_info("Unregistered %s for events", name);
+
+	return 1;
+}
+
+#endif /* DMEVENTD */
+#endif /* DEVMAPPER_SUPPORT */
 
 static void _destroy(const struct segment_type *segtype)
 {
@@ -341,6 +451,10 @@ static struct segtype_handler _mirrored_ops = {
 	add_target_line:_add_target_line,
 	target_percent:_target_percent,
 	target_present:_target_present,
+#ifdef DMEVENTD
+	target_register_events:_target_register_events,
+	target_unregister_events:_target_unregister_events,
+#endif
 #endif
 	destroy:_destroy,
 };

lib/misc/sharedlib.c

@@ -22,8 +22,8 @@
 #include <sys/stat.h>
 #include <dlfcn.h>
 
-static void _get_library_path(struct config_tree *cft, const char *libname,
-			      char *path, int path_len)
+void get_shared_library_path(struct config_tree *cft, const char *libname,
+			     char *path, int path_len)
 {
 	struct stat info;
 	const char *lib_dir;
@@ -38,17 +38,23 @@ static void _get_library_path(struct config_tree *cft, const char *libname,
 }
 
 void *load_shared_library(struct config_tree *cft, const char *libname,
-			  const char *desc)
+			  const char *desc, int silent)
 {
 	char path[PATH_MAX];
 	void *library;
 
-	_get_library_path(cft, libname, path, sizeof(path));
+	get_shared_library_path(cft, libname, path, sizeof(path));
 
 	log_very_verbose("Opening shared %s library %s", desc, path);
 
-	if (!(library = dlopen(path, RTLD_LAZY)))
-		log_error("Unable to open external %s library %s", desc, path);
+	if (!(library = dlopen(path, RTLD_LAZY))) {
+		if (silent && ignorelockingfailure())
+			log_verbose("Unable to open external %s library %s",
+				    desc, path);
+		else
+			log_error("Unable to open external %s library %s",
+				  desc, path);
+	}
 
 	return library;
 }

lib/misc/sharedlib.h

@@ -16,5 +16,7 @@
 #include "config.h"
 #include <dlfcn.h>
 
+void get_shared_library_path(struct config_tree *cft, const char *libname,
+			     char *path, int path_len);
 void *load_shared_library(struct config_tree *cf, const char *libname,
-			  const char *what);
+			  const char *what, int silent);

lib/report/report.c

@@ -324,7 +324,6 @@ static int _lvstatus_disp(struct report_handle *rh, struct field *field,
 	const struct logical_volume *lv = (const struct logical_volume *) data;
 	struct lvinfo info;
 	char *repstr;
-	struct lv_segment *snap_seg;
 	float snap_percent;
 
 	if (!(repstr = dm_pool_zalloc(rh->mem, 7))) {
@@ -344,7 +343,7 @@ static int _lvstatus_disp(struct report_handle *rh, struct field *field,
 		repstr[0] = 'v';
 	else if (lv_is_origin(lv))
 		repstr[0] = 'o';
-	else if (find_cow(lv))
+	else if (lv_is_cow(lv))
 		repstr[0] = 's';
 	else
 		repstr[0] = '-';
@@ -377,8 +376,8 @@ static int _lvstatus_disp(struct report_handle *rh, struct field *field,
 			repstr[4] = 'd';	/* Inactive without table */
 
 		/* Snapshot dropped? */
-		if (info.live_table && (snap_seg = find_cow(lv)) &&
-		    (!lv_snapshot_percent(snap_seg->cow, &snap_percent) ||
+		if (info.live_table && lv_is_cow(lv) &&
+		    (!lv_snapshot_percent(lv, &snap_percent) ||
 		     snap_percent < 0 || snap_percent >= 100)) {
 			repstr[0] = toupper(repstr[0]);
 			if (info.suspended)
@@ -491,10 +490,9 @@ static int _origin_disp(struct report_handle *rh, struct field *field,
 			const void *data)
 {
 	const struct logical_volume *lv = (const struct logical_volume *) data;
-	struct lv_segment *snap_seg;
 
-	if ((snap_seg = find_cow(lv)))
-		return _string_disp(rh, field, &snap_seg->origin->name);
+	if (lv_is_cow(lv))
+		return _string_disp(rh, field, &origin_from_cow(lv)->name);
 
 	field->report_string = "";
 	field->sort_value = (const void *) field->report_string;
@@ -527,8 +525,7 @@ static int _lvname_disp(struct report_handle *rh, struct field *field,
 	char *repstr;
 	size_t len;
 
-	/* FIXME Remove need for snapshot special case */
-	if (lv->status & VISIBLE_LV || lv_is_cow(lv)) {
+	if (lv_is_visible(lv)) {
 		repstr = lv->name;
 		return _string_disp(rh, field, &repstr);
 	}
@@ -667,11 +664,10 @@ static int _chunksize_disp(struct report_handle *rh, struct field *field,
 			   const void *data)
 {
 	const struct lv_segment *seg = (const struct lv_segment *) data;
-	struct lv_segment *snap_seg;
 	uint64_t size;
 
-	if ((snap_seg = find_cow(seg->lv)))
-		size = (uint64_t) snap_seg->chunk_size;
+	if (lv_is_cow(seg->lv))
+		size = (uint64_t) find_cow(seg->lv)->chunk_size;
 	else
 		size = 0;
 
@@ -840,7 +836,6 @@ static int _snpercent_disp(struct report_handle *rh, struct field *field,
 			   const void *data)
 {
 	const struct logical_volume *lv = (const struct logical_volume *) data;
-	struct lv_segment *snap_seg;
 	struct lvinfo info;
 	float snap_percent;
 	uint64_t *sortval;
@@ -851,16 +846,15 @@ static int _snpercent_disp(struct report_handle *rh, struct field *field,
 		return 0;
 	}
 
-	if (!(snap_seg = find_cow(lv)) ||
-	    (lv_info(lv->vg->cmd, snap_seg->cow, &info, 0) && !info.exists)) {
+	if (!lv_is_cow(lv) ||
+	    (lv_info(lv->vg->cmd, lv, &info, 0) && !info.exists)) {
 		field->report_string = "";
 		*sortval = UINT64_C(0);
 		field->sort_value = sortval;
 		return 1;
 	}
 
-	if (!lv_snapshot_percent(snap_seg->cow, &snap_percent)
-	    || snap_percent < 0) {
+	if (!lv_snapshot_percent(lv, &snap_percent) || snap_percent < 0) {
 		field->report_string = "100.00";
 		*sortval = UINT64_C(100);
 		field->sort_value = sortval;

libdm/.exported_symbols

@@ -2,6 +2,7 @@ dm_lib_release
 dm_lib_exit
 dm_driver_version
 dm_get_library_version
+dm_log
 dm_log_init
 dm_log_init_verbose
 dm_task_create
@@ -22,9 +23,14 @@ dm_task_set_major
 dm_task_set_minor
 dm_task_set_sector
 dm_task_set_message
+dm_task_set_uid
+dm_task_set_gid
+dm_task_set_mode
+dm_task_suppress_identical_reload
 dm_task_add_target
 dm_task_no_open_count
 dm_task_skip_lockfs
+dm_task_update_nodes
 dm_task_run
 dm_get_next_target
 dm_set_dev_dir
@@ -58,14 +64,16 @@ dm_tree_node_add_striped_target
 dm_tree_node_add_mirror_target
 dm_tree_node_add_mirror_target_log
 dm_tree_node_add_target_area
+dm_tree_skip_lockfs
 dm_is_dm_major
 dm_mknodes
 dm_malloc_aux
-dm_strdup
+dm_malloc_aux_debug
+dm_strdup_aux
 dm_free_aux
 dm_realloc_aux
-dm_dump_memory
-dm_bounds_check
+dm_dump_memory_debug
+dm_bounds_check_debug
 dm_pool_create
 dm_pool_destroy
 dm_pool_alloc
@@ -100,3 +108,4 @@ dm_hash_get_data
 dm_hash_get_first
 dm_hash_get_next
 dm_set_selinux_context
+dm_task_set_geometry

libdm/Makefile.in

@@ -17,14 +17,6 @@ top_srcdir = @top_srcdir@
 VPATH = @srcdir@
 interface = @interface@
 
-ifeq ("@DMEVENTD@", "yes")
-   SUBDIRS += event
-endif
-
-ifeq ($(MAKECMDGOALS),distclean)
-  SUBDIRS += event
-endif
-
 SOURCES =\
 	datastruct/bitset.c \
 	datastruct/hash.c \

libdm/datastruct/bitset.c

@@ -57,7 +57,7 @@ void dm_bit_union(dm_bitset_t out, dm_bitset_t in1, dm_bitset_t in2)
  */
 static inline int _test_word(uint32_t test, int bit)
 {
-	while (bit < DM_BITS_PER_INT) {
+	while (bit < (int) DM_BITS_PER_INT) {
 		if (test & (0x1 << bit))
 			return bit;
 		bit++;
@@ -73,7 +73,10 @@ int dm_bit_get_next(dm_bitset_t bs, int last_bit)
 
 	last_bit++;		/* otherwise we'll return the same bit again */
 
-	while (last_bit < bs[0]) {
+	/*
+	 * bs[0] holds number of bits
+	 */
+	while (last_bit < (int) bs[0]) {
 		word = last_bit >> INT_SHIFT;
 		test = bs[word + 1];
 		bit = last_bit & (DM_BITS_PER_INT - 1);

libdm/datastruct/hash.c

@@ -18,13 +18,13 @@
 struct dm_hash_node {
 	struct dm_hash_node *next;
 	void *data;
-	int keylen;
+	unsigned keylen;
 	char key[0];
 };
 
 struct dm_hash_table {
-	int num_nodes;
-	int num_slots;
+	unsigned num_nodes;
+	unsigned num_slots;
 	struct dm_hash_node **slots;
 };
 
@@ -56,7 +56,7 @@ static unsigned char _nums[] = {
 	209
 };
 
-static struct dm_hash_node *_create_node(const char *str, int len)
+static struct dm_hash_node *_create_node(const char *str, unsigned len)
 {
 	struct dm_hash_node *n = dm_malloc(sizeof(*n) + len);
 
@@ -68,13 +68,14 @@ static struct dm_hash_node *_create_node(const char *str, int len)
 	return n;
 }
 
-static unsigned _hash(const char *str, uint32_t len)
+static unsigned long _hash(const unsigned char *str, unsigned len)
 {
-	unsigned long h = 0, g, i;
+	unsigned long h = 0, g;
+	unsigned i;
 
 	for (i = 0; i < len; i++) {
 		h <<= 4;
-		h += _nums[(int) *str++];
+		h += _nums[*str++];
 		g = h & ((unsigned long) 0xf << 16u);
 		if (g) {
 			h ^= g >> 16u;
@@ -120,7 +121,7 @@ struct dm_hash_table *dm_hash_create(unsigned size_hint)
 static void _free_nodes(struct dm_hash_table *t)
 {
 	struct dm_hash_node *c, *n;
-	int i;
+	unsigned i;
 
 	for (i = 0; i < t->num_slots; i++)
 		for (c = t->slots[i]; c; c = n) {
@@ -136,8 +137,8 @@ void dm_hash_destroy(struct dm_hash_table *t)
 	dm_free(t);
 }
 
-static inline struct dm_hash_node **_find(struct dm_hash_table *t, const char *key,
-				       uint32_t len)
+static struct dm_hash_node **_find(struct dm_hash_table *t, const char *key,
+				   uint32_t len)
 {
 	unsigned h = _hash(key, len) & (t->num_slots - 1);
 	struct dm_hash_node **c;
@@ -153,11 +154,12 @@ void *dm_hash_lookup_binary(struct dm_hash_table *t, const char *key,
 			 uint32_t len)
 {
 	struct dm_hash_node **c = _find(t, key, len);
+
 	return *c ? (*c)->data : 0;
 }
 
 int dm_hash_insert_binary(struct dm_hash_table *t, const char *key,
-		       uint32_t len, void *data)
+			  uint32_t len, void *data)
 {
 	struct dm_hash_node **c = _find(t, key, len);
 
@@ -214,7 +216,7 @@ unsigned dm_hash_get_num_entries(struct dm_hash_table *t)
 void dm_hash_iter(struct dm_hash_table *t, dm_hash_iterate_fn f)
 {
 	struct dm_hash_node *c;
-	int i;
+	unsigned i;
 
 	for (i = 0; i < t->num_slots; i++)
 		for (c = t->slots[i]; c; c = c->next)
@@ -225,7 +227,7 @@ void dm_hash_wipe(struct dm_hash_table *t)
 {
 	_free_nodes(t);
 	memset(t->slots, 0, sizeof(struct dm_hash_node *) * t->num_slots);
-	t->num_nodes = 0;
+	t->num_nodes = 0u;
 }
 
 char *dm_hash_get_key(struct dm_hash_table *t, struct dm_hash_node *n)
@@ -241,7 +243,7 @@ void *dm_hash_get_data(struct dm_hash_table *t, struct dm_hash_node *n)
 static struct dm_hash_node *_next_slot(struct dm_hash_table *t, unsigned s)
 {
 	struct dm_hash_node *c = NULL;
-	int i;
+	unsigned i;
 
 	for (i = s; i < t->num_slots && !c; i++)
 		c = t->slots[i];
@@ -257,5 +259,6 @@ struct dm_hash_node *dm_hash_get_first(struct dm_hash_table *t)
 struct dm_hash_node *dm_hash_get_next(struct dm_hash_table *t, struct dm_hash_node *n)
 {
 	unsigned h = _hash(n->key, n->keylen) & (t->num_slots - 1);
+
 	return n->next ? n->next : _next_slot(t, h + 1);
 }

libdm/ioctl/libdm-compat.h

@@ -116,6 +116,7 @@ static struct cmd_data _cmd_data_v1[] = {
 	{ "mknodes",	0,			{4, 0, 0} },
 	{ "versions",	0,			{4, 1, 0} },
 	{ "message",	0,			{4, 2, 0} },
+	{ "setgeometry",0,			{4, 6, 0} },
 };
 /* *INDENT-ON* */
 

libdm/ioctl/libdm-iface.c

@@ -59,7 +59,9 @@
 #define NUMBER_OF_MAJORS 4096
 
 /* dm major version no for running kernel */
-static int _dm_version = DM_VERSION_MAJOR;
+static unsigned _dm_version = DM_VERSION_MAJOR;
+static unsigned _dm_version_minor = 0;
+static unsigned _dm_version_patchlevel = 0;
 static int _log_suppress = 0;
 
 static dm_bitset_t _dm_bitset = NULL;
@@ -103,6 +105,9 @@ static struct cmd_data _cmd_data_v4[] = {
 #ifdef DM_TARGET_MSG
 	{"message",	DM_TARGET_MSG,		{4, 2, 0}},
 #endif
+#ifdef DM_DEV_SET_GEOMETRY
+	{"setgeometry",	DM_DEV_SET_GEOMETRY,	{4, 6, 0}},
+#endif
 };
 /* *INDENT-ON* */
 
@@ -236,10 +241,10 @@ static int _create_control(const char *control, uint32_t major, uint32_t minor)
 	}
 
 #ifdef HAVE_SELINUX
-        if (!set_selinux_context(control, S_IFCHR)) {
-                stack;
-                return 0;
-        }
+	if (!dm_set_selinux_context(control, S_IFCHR)) {
+		stack;
+		return 0;
+	}
 #endif
 
 	return 1;
@@ -672,11 +677,13 @@ static int _dm_task_run_v1(struct dm_task *dmt)
 #ifdef DM_IOCTLS
 	else if (ioctl(_control_fd, command, dmi) < 0) {
 		if (_log_suppress)
-			log_verbose("device-mapper ioctl cmd %d failed: %s",
-				    _IOC_NR(command), strerror(errno));
+			log_verbose("device-mapper: %s ioctl failed: %s", 
+				    _cmd_data_v1[dmt->type].name,
+				    strerror(errno));
 		else
-			log_error("device-mapper ioctl cmd %d failed: %s",
-				  _IOC_NR(command), strerror(errno));
+			log_error("device-mapper: %s ioctl failed: %s",
+				  _cmd_data_v1[dmt->type].name,
+				  strerror(errno));
 		goto bad;
 	}
 #else /* Userspace alternative for testing */
@@ -740,7 +747,7 @@ static int _dm_task_run_v1(struct dm_task *dmt)
 
 int dm_task_get_driver_version(struct dm_task *dmt, char *version, size_t size)
 {
-	unsigned int *v;
+	unsigned *v;
 
 #ifdef DM_COMPAT
 	if (_dm_version == 1)
@@ -754,6 +761,9 @@ int dm_task_get_driver_version(struct dm_task *dmt, char *version, size_t size)
 
 	v = dmt->dmi.v4->version;
 	snprintf(version, size, "%u.%u.%u", v[0], v[1], v[2]);
+	_dm_version_minor = v[1];
+	_dm_version_patchlevel = v[2];
+
 	return 1;
 }
 
@@ -799,7 +809,7 @@ int dm_check_version(void)
 	if (!_dm_compat)
 		goto bad;
 
-	log_verbose("device-mapper ioctl protocol version %d failed. "
+	log_verbose("device-mapper ioctl protocol version %u failed. "
 		    "Trying protocol version 1.", _dm_version);
 	_dm_version = 1;
 	if (_check_version(dmversion, sizeof(dmversion), 0)) {
@@ -875,7 +885,7 @@ int dm_format_dev(char *buf, int bufsize, uint32_t dev_major,
 	if (bufsize < 8)
 		return 0;
 
-	r = snprintf(buf, bufsize, "%03u:%03u", dev_major, dev_minor);
+	r = snprintf(buf, (size_t) bufsize, "%u:%u", dev_major, dev_minor);
 	if (r < 0 || r > bufsize - 1)
 		return 0;
 
@@ -966,6 +976,12 @@ int dm_task_set_ro(struct dm_task *dmt)
 	return 1;
 }
 
+int dm_task_suppress_identical_reload(struct dm_task *dmt)
+{
+	dmt->suppress_identical_reload = 1;
+	return 1;
+}
+
 int dm_task_set_newname(struct dm_task *dmt, const char *newname)
 {
 	if (!(dmt->newname = dm_strdup(newname))) {
@@ -993,6 +1009,23 @@ int dm_task_set_sector(struct dm_task *dmt, uint64_t sector)
 	return 1;
 }
 
+int dm_task_set_geometry(struct dm_task *dmt, const char *cylinders, const char *heads, const char *sectors, const char *start)
+{
+	size_t len = strlen(cylinders) + 1 + strlen(heads) + 1 + strlen(sectors) + 1 + strlen(start) + 1;
+
+	if (!(dmt->geometry = dm_malloc(len))) {
+		log_error("dm_task_set_geometry: dm_malloc failed");
+		return 0;
+	}
+
+	if (sprintf(dmt->geometry, "%s %s %s %s", cylinders, heads, sectors, start) < 0) {
+		log_error("dm_task_set_geometry: sprintf failed");
+		return 0;
+	}
+
+	return 1;
+}
+
 int dm_task_no_open_count(struct dm_task *dmt)
 {
 	dmt->no_open_count = 1;
@@ -1020,7 +1053,8 @@ struct target *create_target(uint64_t start, uint64_t len, const char *type,
 	struct target *t = dm_malloc(sizeof(*t));
 
 	if (!t) {
-		log_error("create_target: malloc(%d) failed", sizeof(*t));
+		log_error("create_target: malloc(%" PRIsize_t ") failed",
+			  sizeof(*t));
 		return NULL;
 	}
 
@@ -1086,6 +1120,40 @@ static void *_add_target(struct target *t, void *out, void *end)
 	return out;
 }
 
+static int _lookup_dev_name(uint64_t dev, char *buf, size_t len)
+{
+	struct dm_names *names;
+	unsigned next = 0;
+	struct dm_task *dmt;
+	int r = 0;
+ 
+	if (!(dmt = dm_task_create(DM_DEVICE_LIST)))
+		return 0;
+ 
+	if (!dm_task_run(dmt))
+		goto out;
+
+	if (!(names = dm_task_get_names(dmt)))
+		goto out;
+ 
+	if (!names->dev)
+		goto out;
+ 
+	do {
+		names = (void *) names + next;
+		if (names->dev == dev) {
+			strncpy(buf, names->name, len);
+			r = 1;
+			break;
+		}
+		next = names->next;
+	} while (next);
+
+      out:
+	dm_task_destroy(dmt);
+	return r;
+}
+
 static struct dm_ioctl *_flatten(struct dm_task *dmt, unsigned repeat_count)
 {
 	const size_t min_size = 16 * 1024;
@@ -1114,11 +1182,26 @@ static struct dm_ioctl *_flatten(struct dm_task *dmt, unsigned repeat_count)
 		return NULL;
 	}
 
+	if (count && dmt->geometry) {
+		log_error("targets and geometry are incompatible");
+		return NULL;
+	}
+
 	if (dmt->newname && (dmt->sector || dmt->message)) {
 		log_error("message and newname are incompatible");
 		return NULL;
 	}
 
+	if (dmt->newname && dmt->geometry) {
+		log_error("geometry and newname are incompatible");
+		return NULL;
+	}
+
+	if (dmt->geometry && (dmt->sector || dmt->message)) {
+		log_error("geometry and message are incompatible");
+		return NULL;
+	}
+
 	if (dmt->sector && !dmt->message) {
 		log_error("message is required with sector");
 		return NULL;
@@ -1130,6 +1213,9 @@ static struct dm_ioctl *_flatten(struct dm_task *dmt, unsigned repeat_count)
 	if (dmt->message)
 		len += sizeof(struct dm_target_msg) + strlen(dmt->message) + 1;
 
+	if (dmt->geometry)
+		len += strlen(dmt->geometry) + 1;
+
 	/*
 	 * Give len a minimum size so that we have space to store
 	 * dependencies or status information.
@@ -1155,16 +1241,6 @@ static struct dm_ioctl *_flatten(struct dm_task *dmt, unsigned repeat_count)
 	dmi->data_size = len;
 	dmi->data_start = sizeof(struct dm_ioctl);
 
-	if (dmt->dev_name)
-		strncpy(dmi->name, dmt->dev_name, sizeof(dmi->name));
-
-	if (dmt->type == DM_DEVICE_SUSPEND)
-		dmi->flags |= DM_SUSPEND_FLAG;
-	if (dmt->read_only)
-		dmi->flags |= DM_READONLY_FLAG;
-	if (dmt->skip_lockfs)
-		dmi->flags |= DM_SKIP_LOCKFS_FLAG;
-
 	if (dmt->minor >= 0) {
 		if (dmt->major <= 0) {
 			log_error("Missing major number for persistent device.");
@@ -1174,9 +1250,31 @@ static struct dm_ioctl *_flatten(struct dm_task *dmt, unsigned repeat_count)
 		dmi->dev = MKDEV(dmt->major, dmt->minor);
 	}
 
+	/* Does driver support device number referencing? */
+	if (_dm_version_minor < 3 && !dmt->dev_name && !dmt->uuid && dmi->dev) {
+		if (!_lookup_dev_name(dmi->dev, dmi->name, sizeof(dmi->name))) {
+			log_error("Unable to find name for device (%" PRIu32
+				  ":%" PRIu32 ")", dmt->major, dmt->minor);
+			goto bad;
+		}
+		log_verbose("device (%" PRIu32 ":%" PRIu32 ") is %s "
+			    "for compatibility with old kernel",
+			    dmt->major, dmt->minor, dmi->name);
+	}
+
+	if (dmt->dev_name)
+		strncpy(dmi->name, dmt->dev_name, sizeof(dmi->name));
+
 	if (dmt->uuid)
 		strncpy(dmi->uuid, dmt->uuid, sizeof(dmi->uuid));
 
+	if (dmt->type == DM_DEVICE_SUSPEND)
+		dmi->flags |= DM_SUSPEND_FLAG;
+	if (dmt->read_only)
+		dmi->flags |= DM_READONLY_FLAG;
+	if (dmt->skip_lockfs)
+		dmi->flags |= DM_SKIP_LOCKFS_FLAG;
+
 	dmi->target_count = count;
 	dmi->event_nr = dmt->event_nr;
 
@@ -1196,6 +1294,9 @@ static struct dm_ioctl *_flatten(struct dm_task *dmt, unsigned repeat_count)
 		strcpy(tmsg->message, dmt->message);
 	}
 
+	if (dmt->geometry)
+		strcpy(b, dmt->geometry);
+
 	return dmi;
 
       bad:
@@ -1302,6 +1403,9 @@ static int _create_and_load_v4(struct dm_task *dmt)
 
 	task->major = dmt->major;
 	task->minor = dmt->minor;
+	task->uid = dmt->uid;
+	task->gid = dmt->gid;
+	task->mode = dmt->mode;
 
 	r = dm_task_run(task);
 	dm_task_destroy(task);
@@ -1330,7 +1434,7 @@ static int _create_and_load_v4(struct dm_task *dmt)
 	task->tail = NULL;
 	dm_task_destroy(task);
 	if (!r)
-		return r;
+		goto revert;
 
 	/* Use the original structure last so the info will be correct */
 	dmt->type = DM_DEVICE_RESUME;
@@ -1339,6 +1443,82 @@ static int _create_and_load_v4(struct dm_task *dmt)
 
 	r = dm_task_run(dmt);
 
+	if (r)
+		return r;
+
+      revert:
+ 	dmt->type = DM_DEVICE_REMOVE;
+	dm_free(dmt->uuid);
+	dmt->uuid = NULL;
+
+	if (!dm_task_run(dmt))
+		log_error("Failed to revert device creation.");
+
+	return r;
+}
+
+static int _reload_with_suppression_v4(struct dm_task *dmt)
+{
+	struct dm_task *task;
+	struct target *t1, *t2;
+	int matches = 1;
+	int r;
+
+	/* New task to get existing table information */
+	if (!(task = dm_task_create(DM_DEVICE_TABLE))) {
+		log_error("Failed to create device-mapper task struct");
+		return 0;
+	}
+
+	/* Copy across relevant fields */
+	if (dmt->dev_name && !dm_task_set_name(task, dmt->dev_name)) {
+		dm_task_destroy(task);
+		return 0;
+	}
+
+	if (dmt->uuid && !dm_task_set_uuid(task, dmt->uuid)) {
+		dm_task_destroy(task);
+		return 0;
+	}
+
+	task->major = dmt->major;
+	task->minor = dmt->minor;
+
+	r = dm_task_run(task);
+
+	if (!r) {
+		dm_task_destroy(task);
+		return r;
+	}
+
+	t1 = dmt->head;
+	t2 = task->head;
+
+	while (t1 && t2) {
+		if ((t1->start != t2->start) ||
+		    (t1->length != t2->length) ||
+		    (strcmp(t1->type, t2->type)) ||
+		    (strcmp(t1->params, t2->params))) {
+			matches = 0;
+			break;
+		}
+		t1 = t1->next;
+		t2 = t2->next;
+	}
+	
+	if (matches && !t1 && !t2) {
+		dmt->dmi.v4 = task->dmi.v4;
+		task->dmi.v4 = NULL;
+		dm_task_destroy(task);
+		return 1;
+	}
+
+	dm_task_destroy(task);
+
+	/* Now do the original reload */
+	dmt->suppress_identical_reload = 0;
+	r = dm_task_run(dmt);
+
 	return r;
 }
 
@@ -1361,7 +1541,7 @@ static struct dm_ioctl *_do_dm_ioctl(struct dm_task *dmt, unsigned command,
 	if (dmt->no_open_count)
 		dmi->flags |= DM_SKIP_BDGET_FLAG;
 
-	log_debug("dm %s %s %s%s%s %s%0.0d%s%0.0d%s"
+	log_debug("dm %s %s %s%s%s %s%.0d%s%.0d%s"
 		  "%s%c %.0llu %s [%u]",
 		  _cmd_data_v4[dmt->type].name,
 		  dmi->name, dmi->uuid, dmt->newname ? " " : "",
@@ -1383,13 +1563,15 @@ static struct dm_ioctl *_do_dm_ioctl(struct dm_task *dmt, unsigned command,
 			dmi->flags &= ~DM_EXISTS_FLAG;	/* FIXME */
 		else {
 			if (_log_suppress)
-				log_verbose("device-mapper ioctl "
-					    "cmd %d failed: %s",
-					    _IOC_NR(command), strerror(errno));
+				log_verbose("device-mapper: %s ioctl "
+					    "failed: %s",
+				    	    _cmd_data_v4[dmt->type].name,
+					    strerror(errno));
 			else
-				log_error("device-mapper ioctl "
-					  "cmd %d failed: %s",
-					  _IOC_NR(command), strerror(errno));
+				log_error("device-mapper: %s ioctl "
+					  "failed: %s",
+				    	   _cmd_data_v4[dmt->type].name,
+					  strerror(errno));
 			dm_free(dmi);
 			return NULL;
 		}
@@ -1399,6 +1581,11 @@ static struct dm_ioctl *_do_dm_ioctl(struct dm_task *dmt, unsigned command,
 	return dmi;
 }
 
+void dm_task_update_nodes(void)
+{
+	update_devs();
+}
+
 int dm_task_run(struct dm_task *dmt)
 {
 	struct dm_ioctl *dmi;
@@ -1426,6 +1613,9 @@ int dm_task_run(struct dm_task *dmt)
 	    !dmt->uuid && dmt->major <= 0)
 		return _mknodes_v4(dmt);
 
+	if ((dmt->type == DM_DEVICE_RELOAD) && dmt->suppress_identical_reload)
+		return _reload_with_suppression_v4(dmt);
+
 	if (!_open_control())
 		return 0;
 
@@ -1484,6 +1674,9 @@ repeat_ioctl:
 		break;
 	}
 
+	/* Was structure reused? */
+	if (dmt->dmi.v4)
+		dm_free(dmt->dmi.v4);
 	dmt->dmi.v4 = dmi;
 	return 1;
 

libdm/ioctl/libdm-targets.h

@@ -50,9 +50,11 @@ struct dm_task {
 	} dmi;
 	char *newname;
 	char *message;
+	char *geometry;
 	uint64_t sector;
 	int no_open_count;
 	int skip_lockfs;
+	int suppress_identical_reload;
 
 	char *uuid;
 };

libdm/libdevmapper.h

@@ -42,7 +42,8 @@
  */
 
 typedef void (*dm_log_fn) (int level, const char *file, int line,
-			   const char *f, ...);
+			   const char *f, ...)
+    __attribute__ ((format(printf, 4, 5)));
 
 /*
  * The library user may wish to register their own
@@ -79,7 +80,9 @@ enum {
 
 	DM_DEVICE_LIST_VERSIONS,
 	
-	DM_DEVICE_TARGET_MSG
+	DM_DEVICE_TARGET_MSG,
+
+	DM_DEVICE_SET_GEOMETRY
 };
 
 struct dm_task;
@@ -140,11 +143,16 @@ int dm_task_set_ro(struct dm_task *dmt);
 int dm_task_set_newname(struct dm_task *dmt, const char *newname);
 int dm_task_set_minor(struct dm_task *dmt, int minor);
 int dm_task_set_major(struct dm_task *dmt, int major);
+int dm_task_set_uid(struct dm_task *dmt, uid_t uid);
+int dm_task_set_gid(struct dm_task *dmt, gid_t gid);
+int dm_task_set_mode(struct dm_task *dmt, mode_t mode);
 int dm_task_set_event_nr(struct dm_task *dmt, uint32_t event_nr);
+int dm_task_set_geometry(struct dm_task *dmt, const char *cylinders, const char *heads, const char *sectors, const char *start);
 int dm_task_set_message(struct dm_task *dmt, const char *message);
 int dm_task_set_sector(struct dm_task *dmt, uint64_t sector);
 int dm_task_no_open_count(struct dm_task *dmt);
 int dm_task_skip_lockfs(struct dm_task *dmt);
+int dm_task_suppress_identical_reload(struct dm_task *dmt);
 
 /*
  * Use these to prepare for a create or reload.
@@ -168,6 +176,12 @@ void *dm_get_next_target(struct dm_task *dmt,
  */
 int dm_task_run(struct dm_task *dmt);
 
+/*
+ * Call this to make or remove the device nodes associated with previously
+ * issued commands.
+ */
+void dm_task_update_nodes(void);
+
 /*
  * Configure the device-mapper directory
  */
@@ -292,6 +306,13 @@ int dm_tree_suspend_children(struct dm_tree_node *dnode,
 				   const char *uuid_prefix,
 				   size_t uuid_prefix_len);
 
+/*
+ * Skip the filesystem sync when suspending.
+ * Does nothing with other functions.
+ * Use this when no snapshots are involved.
+ */ 
+void dm_tree_skip_lockfs(struct dm_tree_node *dnode);
+
 /*
  * Is the uuid prefix present in the tree?
  * Only returns 0 if every node was checked successfully.
@@ -324,11 +345,18 @@ int dm_tree_node_add_striped_target(struct dm_tree_node *node,
 				       uint32_t stripe_size);
 int dm_tree_node_add_mirror_target(struct dm_tree_node *node,
 				      uint64_t size);
+ 
+/* Mirror log flags */
+#define DM_NOSYNC		0x00000001	/* Known already in sync */
+#define DM_FORCESYNC		0x00000002	/* Force resync */
+#define DM_BLOCK_ON_ERROR	0x00000004	/* On error, suspend I/O */
+
 int dm_tree_node_add_mirror_target_log(struct dm_tree_node *node,
 					  uint32_t region_size,
 					  unsigned clustered,
 					  const char *log_uuid,
-					  unsigned area_count);
+					  unsigned area_count,
+					  uint32_t flags);
 int dm_tree_node_add_target_area(struct dm_tree_node *node,
 				    const char *dev_name,
 				    const char *dlid,
@@ -343,29 +371,34 @@ int dm_tree_node_add_target_area(struct dm_tree_node *node,
  *******************/
 
 void *dm_malloc_aux(size_t s, const char *file, int line);
-#define dm_malloc(s) dm_malloc_aux((s), __FILE__, __LINE__)
-
-char *dm_strdup(const char *str);
+void *dm_malloc_aux_debug(size_t s, const char *file, int line);
+char *dm_strdup_aux(const char *str, const char *file, int line);
+void dm_free_aux(void *p);
+void *dm_realloc_aux(void *p, unsigned int s, const char *file, int line);
+int dm_dump_memory_debug(void);
+void dm_bounds_check_debug(void);
 
 #ifdef DEBUG_MEM
 
-void dm_free_aux(void *p);
-void *dm_realloc_aux(void *p, unsigned int s, const char *file, int line);
-int dm_dump_memory(void);
-void dm_bounds_check(void);
-
+#  define dm_malloc(s) dm_malloc_aux_debug((s), __FILE__, __LINE__)
+#  define dm_strdup(s) dm_strdup_aux((s), __FILE__, __LINE__)
 #  define dm_free(p) dm_free_aux(p)
 #  define dm_realloc(p, s) dm_realloc_aux(p, s, __FILE__, __LINE__)
+#  define dm_dump_memory() dm_dump_memory_debug()
+#  define dm_bounds_check() dm_bounds_check_debug()
 
 #else
 
+#  define dm_malloc(s) dm_malloc_aux((s), __FILE__, __LINE__)
+#  define dm_strdup(s) strdup(s)
 #  define dm_free(p) free(p)
 #  define dm_realloc(p, s) realloc(p, s)
-#  define dm_dump_memory()
-#  define dm_bounds_check()
+#  define dm_dump_memory() {}
+#  define dm_bounds_check() {}
 
 #endif
 
+
 /*
  * The pool allocator is useful when you are going to allocate
  * lots of memory, use the memory for a bit, and then free the
@@ -498,6 +531,17 @@ int dm_bit_get_next(dm_bitset_t bs, int last_bit);
 #define dm_bit_copy(bs1, bs2) \
    memcpy(bs1 + 1, bs2 + 1, ((*bs1 / DM_BITS_PER_INT) + 1) * sizeof(int))
 
+/* Returns number of set bits */
+static inline unsigned hweight32(uint32_t i)
+{
+	unsigned r = (i & 0x55555555) + ((i >> 1) & 0x55555555);
+
+	r =    (r & 0x33333333) + ((r >>  2) & 0x33333333);
+	r =    (r & 0x0F0F0F0F) + ((r >>  4) & 0x0F0F0F0F);
+	r =    (r & 0x00FF00FF) + ((r >>  8) & 0x00FF00FF);
+	return (r & 0x0000FFFF) + ((r >> 16) & 0x0000FFFF);
+}
+
 /****************
  * hash functions
  ****************/
@@ -532,9 +576,9 @@ struct dm_hash_node *dm_hash_get_next(struct dm_hash_table *t, struct dm_hash_no
 	for (v = dm_hash_get_first(h); v; \
 	     v = dm_hash_get_next(h, v))
 
-#endif				/* LIB_DEVICE_MAPPER_H */
-
 /*********
  * selinux
  *********/
 int dm_set_selinux_context(const char *path, mode_t mode);
+
+#endif				/* LIB_DEVICE_MAPPER_H */

libdm/libdm-common.c

@@ -61,14 +61,14 @@ static void _default_log(int level, const char *file, int line,
 		fprintf(stdout, "\n");
 }
 
-dm_log_fn _log = _default_log;
+dm_log_fn dm_log = _default_log;
 
 void dm_log_init(dm_log_fn fn)
 {
 	if (fn)
-		_log = fn;
+		dm_log = fn;
 	else
-		_log = _default_log;
+		dm_log = _default_log;
 }
 
 void dm_log_init_verbose(int level)
@@ -96,7 +96,8 @@ struct dm_task *dm_task_create(int type)
 	struct dm_task *dmt = dm_malloc(sizeof(*dmt));
 
 	if (!dmt) {
-		log_error("dm_task_create: malloc(%d) failed", sizeof(*dmt));
+		log_error("dm_task_create: malloc(%" PRIsize_t ") failed",
+			  sizeof(*dmt));
 		return NULL;
 	}
 
@@ -180,6 +181,27 @@ int dm_task_set_minor(struct dm_task *dmt, int minor)
 	return 1;
 }
 
+int dm_task_set_uid(struct dm_task *dmt, uid_t uid)
+{
+	dmt->uid = uid;
+
+	return 1;
+}
+
+int dm_task_set_gid(struct dm_task *dmt, gid_t gid)
+{
+	dmt->gid = gid;
+
+	return 1;
+}
+
+int dm_task_set_mode(struct dm_task *dmt, mode_t mode)
+{
+	dmt->mode = mode;
+
+	return 1;
+}
+
 int dm_task_add_target(struct dm_task *dmt, uint64_t start, uint64_t size,
 		       const char *ttype, const char *params)
 {
@@ -488,3 +510,4 @@ out:
 	dm_task_destroy(dmt);
 	return r;
 }
+

libdm/libdm-deptree.c

@@ -25,6 +25,9 @@
 
 #define MAX_TARGET_PARAMSIZE 500000
 
+/* FIXME Fix interface so this is used only by LVM */
+#define UUID_PREFIX "LVM-"
+
 /* Supported segment types */
 enum {
 	SEG_ERROR, 
@@ -82,6 +85,8 @@ struct load_segment {
 	uint32_t region_size;		/* Mirror */
 	unsigned clustered;		/* Mirror */
 	unsigned mirror_area_count;	/* Mirror */
+	uint32_t flags;			/* Mirror log */
+	char *uuid;			/* Clustered mirror log */
 };
 
 /* Per-device properties */
@@ -112,6 +117,8 @@ struct dm_tree_node {
         struct list uses;       	/* Nodes this node uses */
         struct list used_by;    	/* Nodes that use this node */
 
+	int activation_priority;	/* 0 gets activated first */
+
 	void *context;			/* External supplied context */
 
 	struct load_properties props;	/* For creation/table (re)load */
@@ -122,6 +129,7 @@ struct dm_tree {
 	struct dm_hash_table *devs;
 	struct dm_hash_table *uuids;
 	struct dm_tree_node root;
+	int skip_lockfs;		/* 1 skips lockfs (for non-snapshots) */
 };
 
 /* FIXME Consider exporting this */
@@ -134,7 +142,7 @@ static int _dm_snprintf(char *buf, size_t bufsize, const char *format, ...)
         n = vsnprintf(buf, bufsize, format, ap);
         va_end(ap);
 
-        if (n < 0 || (n > bufsize - 1))
+        if (n < 0 || (n > (int) bufsize - 1))
                 return -1;
 
         return n;
@@ -153,6 +161,7 @@ struct dm_tree *dm_tree_create(void)
 	dtree->root.dtree = dtree;
 	list_init(&dtree->root.uses);
 	list_init(&dtree->root.used_by);
+	dtree->skip_lockfs = 0;
 
 	if (!(dtree->mem = dm_pool_create("dtree", 1024))) {
 		log_error("dtree pool creation failed");
@@ -310,6 +319,7 @@ static struct dm_tree_node *_create_dm_tree_node(struct dm_tree *dtree,
 	node->uuid = uuid;
 	node->info = *info;
 	node->context = context;
+	node->activation_priority = 0;
 
 	list_init(&node->uses);
 	list_init(&node->used_by);
@@ -348,8 +358,15 @@ static struct dm_tree_node *_find_dm_tree_node(struct dm_tree *dtree,
 static struct dm_tree_node *_find_dm_tree_node_by_uuid(struct dm_tree *dtree,
 						       const char *uuid)
 {
-	/* FIXME Do we need to cope with missing LVM- prefix too? */
-	return dm_hash_lookup(dtree->uuids, uuid);
+	struct dm_tree_node *node;
+
+	if ((node = dm_hash_lookup(dtree->uuids, uuid)))
+		return node;
+
+	if (strncmp(uuid, UUID_PREFIX, sizeof(UUID_PREFIX) - 1))
+		return NULL;
+
+	return dm_hash_lookup(dtree->uuids, uuid + sizeof(UUID_PREFIX) - 1);
 }
 
 static int _deps(struct dm_task **dmt, struct dm_pool *mem, uint32_t major, uint32_t minor,
@@ -653,13 +670,13 @@ static int _uuid_prefix_matches(const char *uuid, const char *uuid_prefix, size_
 	if (uuid_prefix_len <= 4)
 		return 0;
 
-	if (!strncmp(uuid, "LVM-", 4))
+	if (!strncmp(uuid, UUID_PREFIX, sizeof(UUID_PREFIX) - 1))
 		return 0;
 
-	if (strncmp(uuid_prefix, "LVM-", 4))
+	if (strncmp(uuid_prefix, UUID_PREFIX, sizeof(UUID_PREFIX) - 1))
 		return 0;
 
-	if (!strncmp(uuid, uuid_prefix + 4, uuid_prefix_len - 4))
+	if (!strncmp(uuid, uuid_prefix + sizeof(UUID_PREFIX) - 1, uuid_prefix_len - (sizeof(UUID_PREFIX) - 1)))
 		return 1;
 
 	return 0;
@@ -886,12 +903,13 @@ static int _resume_node(const char *name, uint32_t major, uint32_t minor,
 }
 
 static int _suspend_node(const char *name, uint32_t major, uint32_t minor,
-			 struct dm_info *newinfo)
+			 int skip_lockfs, struct dm_info *newinfo)
 {
 	struct dm_task *dmt;
 	int r;
 
-	log_verbose("Suspending %s (%" PRIu32 ":%" PRIu32 ")", name, major, minor);
+	log_verbose("Suspending %s (%" PRIu32 ":%" PRIu32 ")%s", name, major,
+		    minor, skip_lockfs ? "" : " with filesystem sync.");
 
 	if (!(dmt = dm_task_create(DM_DEVICE_SUSPEND))) {
 		log_error("Suspend dm_task creation failed for %s", name);
@@ -907,6 +925,9 @@ static int _suspend_node(const char *name, uint32_t major, uint32_t minor,
 	if (!dm_task_no_open_count(dmt))
 		log_error("Failed to disable open_count");
 
+	if (skip_lockfs && !dm_task_skip_lockfs(dmt))
+		log_error("Failed to set skip_lockfs flag.");
+
 	if ((r = dm_task_run(dmt)))
 		r = dm_task_get_info(dmt, newinfo);
 
@@ -965,6 +986,11 @@ int dm_tree_deactivate_children(struct dm_tree_node *dnode,
 	return 1;
 }
 
+void dm_tree_skip_lockfs(struct dm_tree_node *dnode)
+{
+	dnode->dtree->skip_lockfs = 1;
+}
+
 int dm_tree_suspend_children(struct dm_tree_node *dnode,
 				   const char *uuid_prefix,
 				   size_t uuid_prefix_len)
@@ -1005,7 +1031,8 @@ int dm_tree_suspend_children(struct dm_tree_node *dnode,
 		    !info.exists)
 			continue;
 
-		if (!_suspend_node(name, info.major, info.minor, &newinfo)) {
+		if (!_suspend_node(name, info.major, info.minor,
+				   child->dtree->skip_lockfs, &newinfo)) {
 			log_error("Unable to suspend %s (%" PRIu32
 				  ":%" PRIu32 ")", name, info.major,
 				  info.minor);
@@ -1026,7 +1053,7 @@ int dm_tree_suspend_children(struct dm_tree_node *dnode,
 		}
 
 		/* Ignore if it doesn't belong to this VG */
-		if (uuid_prefix && strncmp(uuid, uuid_prefix, uuid_prefix_len))
+		if (!_uuid_prefix_matches(uuid, uuid_prefix, uuid_prefix_len))
 			continue;
 
 		if (dm_tree_node_num_children(child, 0))
@@ -1045,6 +1072,7 @@ int dm_tree_activate_children(struct dm_tree_node *dnode,
 	struct dm_info newinfo;
 	const char *name;
 	const char *uuid;
+	int priority;
 
 	/* Activate children first */
 	while ((child = dm_tree_next_child(&handle, dnode, 0))) {
@@ -1058,36 +1086,53 @@ int dm_tree_activate_children(struct dm_tree_node *dnode,
 
 		if (dm_tree_node_num_children(child, 0))
 			dm_tree_activate_children(child, uuid_prefix, uuid_prefix_len);
+	}
 
-		if (!(name = dm_tree_node_get_name(child))) {
-			stack;
-			continue;
-		}
+	handle = NULL;
 
-		/* Rename? */
-		if (child->props.new_name) {
-			if (!_rename_node(name, child->props.new_name, child->info.major, child->info.minor)) {
-				log_error("Failed to rename %s (%" PRIu32
-					  ":%" PRIu32 ") to %s", name, child->info.major,
-					  child->info.minor, child->props.new_name);
-				return 0;
+	for (priority = 0; priority < 2; priority++) {
+		while ((child = dm_tree_next_child(&handle, dnode, 0))) {
+			if (!(uuid = dm_tree_node_get_uuid(child))) {
+				stack;
+				continue;
 			}
-			child->name = child->props.new_name;
-			child->props.new_name = NULL;
+
+			if (!_uuid_prefix_matches(uuid, uuid_prefix, uuid_prefix_len))
+				continue;
+
+			if (priority != child->activation_priority)
+				continue;
+
+			if (!(name = dm_tree_node_get_name(child))) {
+				stack;
+				continue;
+			}
+
+			/* Rename? */
+			if (child->props.new_name) {
+				if (!_rename_node(name, child->props.new_name, child->info.major, child->info.minor)) {
+					log_error("Failed to rename %s (%" PRIu32
+						  ":%" PRIu32 ") to %s", name, child->info.major,
+						  child->info.minor, child->props.new_name);
+					return 0;
+				}
+				child->name = child->props.new_name;
+				child->props.new_name = NULL;
+			}
+
+			if (!child->info.inactive_table && !child->info.suspended)
+				continue;
+
+			if (!_resume_node(name, child->info.major, child->info.minor, &newinfo)) {
+				log_error("Unable to resume %s (%" PRIu32
+					  ":%" PRIu32 ")", name, child->info.major,
+					  child->info.minor);
+				continue;
+			}
+
+			/* Update cached info */
+			child->info = newinfo;
 		}
-
-		if (!child->info.inactive_table && !child->info.suspended)
-			continue;
-
-		if (!_resume_node(name, child->info.major, child->info.minor, &newinfo)) {
-			log_error("Unable to resume %s (%" PRIu32
-				  ":%" PRIu32 ")", name, child->info.major,
-				  child->info.minor);
-			continue;
-		}
-
-		/* Update cached info */
-		child->info = newinfo;
 	}
 
 	handle = NULL;
@@ -1180,10 +1225,12 @@ static int _emit_areas_line(struct dm_task *dmt, struct load_segment *seg, char
 
 static int _emit_segment_line(struct dm_task *dmt, struct load_segment *seg, uint64_t *seg_start, char *params, size_t paramsize)
 {
+	unsigned log_parm_count;
         int pos = 0;
 	int tw;
         int r;
 	char originbuf[10], cowbuf[10], logbuf[10];
+	const char *logtype;
 
 	switch(seg->type) {
 	case SEG_ERROR:
@@ -1192,33 +1239,84 @@ static int _emit_segment_line(struct dm_task *dmt, struct load_segment *seg, uin
 	case SEG_LINEAR:
 		break;
 	case SEG_MIRRORED:
+		log_parm_count = 1;	/* Region size */
+		log_parm_count += hweight32(seg->flags);	/* [no]sync, block_on_error etc. */
+
 		if (seg->clustered) {
-			if ((tw = _dm_snprintf(params + pos, paramsize - pos, "clustered ")) < 0) {
+			if (seg->uuid)
+				log_parm_count++;	/* uuid */
+			if ((tw = _dm_snprintf(params + pos, paramsize - pos, "clustered_")) < 0) {
                         	stack;	/* Out of space */
                         	return -1;
                 	}
 			pos += tw;
 		}
-		if (!seg->log) {
-			if ((tw = _dm_snprintf(params + pos, paramsize - pos, "core 1 ")) < 0) {
-                        	stack;	/* Out of space */
-                        	return -1;
-                	}
-			pos += tw;
-		} else {
+
+		if (!seg->log)
+			logtype = "core";
+		else {
+			logtype = "disk";
+			log_parm_count++;
 			if (!_build_dev_string(logbuf, sizeof(logbuf), seg->log))
 				return_0;
-			if ((tw = _dm_snprintf(params + pos, paramsize - pos, "disk 2 %s ", logbuf)) < 0) {
+		}
+
+		if ((tw = _dm_snprintf(params + pos, paramsize - pos, "%s %u ", logtype, log_parm_count)) < 0) {
+                        stack;	/* Out of space */
+                        return -1;
+                }
+		pos += tw;
+
+		if (seg->log) {
+			if ((tw = _dm_snprintf(params + pos, paramsize - pos, "%s ", logbuf)) < 0) {
                         	stack;	/* Out of space */
                         	return -1;
                 	}
 			pos += tw;
 		}
-		if ((tw = _dm_snprintf(params + pos, paramsize - pos, "%u %u ", seg->region_size, seg->mirror_area_count)) < 0) {
+
+		if ((tw = _dm_snprintf(params + pos, paramsize - pos, "%u ", seg->region_size)) < 0) {
                        	stack; /* Out of space */
                        	return -1;
                	}
 		pos += tw;
+
+		if (seg->clustered && seg->uuid) {
+			if ((tw = _dm_snprintf(params + pos, paramsize - pos, "%s ", seg->uuid)) < 0) {
+				stack;  /* Out of space */
+				return -1;
+			}
+			pos += tw;
+		}
+
+		if ((seg->flags & DM_NOSYNC)) {
+			if ((tw = _dm_snprintf(params + pos, paramsize - pos, "nosync ")) < 0) {
+                       		stack; /* Out of space */
+                       		return -1;
+               		}
+			pos += tw;
+		} else if ((seg->flags & DM_FORCESYNC)) {
+			if ((tw = _dm_snprintf(params + pos, paramsize - pos, "sync ")) < 0) {
+                       		stack; /* Out of space */
+                       		return -1;
+               		}
+			pos += tw;
+		}
+
+		if ((seg->flags & DM_BLOCK_ON_ERROR)) {
+			if ((tw = _dm_snprintf(params + pos, paramsize - pos, "block_on_error ")) < 0) {
+                       		stack; /* Out of space */
+                       		return -1;
+               		}
+			pos += tw;
+		}
+
+		if ((tw = _dm_snprintf(params + pos, paramsize - pos, "%u ", seg->mirror_area_count)) < 0) {
+                       	stack; /* Out of space */
+                       	return -1;
+               	}
+		pos += tw;
+
 		break;
 	case SEG_SNAPSHOT:
 		if (!_build_dev_string(originbuf, sizeof(originbuf), seg->origin))
@@ -1343,8 +1441,15 @@ static int _load_node(struct dm_tree_node *dnode)
 		if (!_emit_segment(dmt, seg, &seg_start))
 			goto_out;
 
-	if ((r = dm_task_run(dmt)))
+	if (!dm_task_suppress_identical_reload(dmt))
+		log_error("Failed to suppress reload of identical tables.");
+
+	if ((r = dm_task_run(dmt))) {
 		r = dm_task_get_info(dmt, &dnode->info);
+		if (r && !dnode->info.inactive_table)
+			log_verbose("Suppressed %s identical table reload.",
+				    dnode->name);
+	}
 
 	dnode->props.segment_count = 0;
 
@@ -1352,7 +1457,6 @@ out:
 	dm_task_destroy(dmt);
 
 	return r;
-
 }
 
 int dm_tree_preload_children(struct dm_tree_node *dnode,
@@ -1371,8 +1475,8 @@ int dm_tree_preload_children(struct dm_tree_node *dnode,
 			continue;
 
 		/* Ignore if it doesn't belong to this VG */
-		if (uuid_prefix && child->info.exists &&
-		    strncmp(child->uuid, uuid_prefix, uuid_prefix_len))
+		if (child->info.exists &&
+		    !_uuid_prefix_matches(child->uuid, uuid_prefix, uuid_prefix_len))
 			continue;
 
 		if (dm_tree_node_num_children(child, 0))
@@ -1402,6 +1506,9 @@ int dm_tree_preload_children(struct dm_tree_node *dnode,
 		if (!dm_tree_node_num_children(child, 1))
 			continue;
 
+		if (!child->info.inactive_table && !child->info.suspended)
+			continue;
+
 		if (!_resume_node(name, child->info.major, child->info.minor, &newinfo)) {
 			log_error("Unable to resume %s (%" PRIu32
 				  ":%" PRIu32 ")", name, child->info.major,
@@ -1435,7 +1542,7 @@ int dm_tree_children_use_uuid(struct dm_tree_node *dnode,
 			return 1;
 		}
 
-		if (!strncmp(uuid, uuid_prefix, uuid_prefix_len))
+		if (_uuid_prefix_matches(uuid, uuid_prefix, uuid_prefix_len))
 			return 1;
 
 		if (dm_tree_node_num_children(child, 0))
@@ -1492,6 +1599,9 @@ int dm_tree_node_add_snapshot_origin_target(struct dm_tree_node *dnode,
 	if (!_link_tree_nodes(dnode, origin_node))
 		return_0;
 
+	/* Resume snapshot origins after new snapshots */
+	dnode->activation_priority = 1;
+
 	return 1;
 }
 
@@ -1577,7 +1687,8 @@ int dm_tree_node_add_mirror_target_log(struct dm_tree_node *node,
 					  uint32_t region_size,
 					  unsigned clustered, 
 					  const char *log_uuid,
-					  unsigned area_count)
+					  unsigned area_count,
+					  uint32_t flags)
 {
 	struct dm_tree_node *log_node = NULL;
 	struct load_segment *seg;
@@ -1594,14 +1705,21 @@ int dm_tree_node_add_mirror_target_log(struct dm_tree_node *node,
 			log_error("Couldn't find mirror log uuid %s.", log_uuid);
 			return 0;
 		}
+
 		if (!_link_tree_nodes(node, log_node))
 			return_0;
+
+		if (!(seg->uuid = dm_pool_strdup(node->dtree->mem, log_uuid))) {
+			log_error("log uuid pool_strdup failed");
+			return 0;
+		}
 	}
 
 	seg->log = log_node;
 	seg->region_size = region_size;
 	seg->clustered = clustered;
 	seg->mirror_area_count = area_count;
+	seg->flags = flags;
 
 	return 1;
 }

libdm/libdm-event.h

@@ -1,95 +0,0 @@
-/*
- * Copyright (C) 2005 Red Hat, Inc. All rights reserved.
- *
- * This file is part of the device-mapper userspace tools.
- *
- * This copyrighted material is made available to anyone wishing to use,
- * modify, copy, or redistribute it subject to the terms and conditions
- * of the GNU Lesser General Public License v.2.1.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with this program; if not, write to the Free Software Foundation,
- * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
- */
-
-#ifndef LIB_DMEVENT_H
-#define LIB_DMEVENT_H
-
-#include "list.h"
-
-#include <stdint.h>
-
-#define	DAEMON		"/sbin/dmeventd"
-#define LOCKFILE	"/var/lock/dmeventd"
-#define	FIFO_CLIENT	"/var/run/dmeventd-client"
-#define	FIFO_SERVER	"/var/run/dmeventd-server"
-#define PIDFILE		"/var/run/dmeventd.pid"
-
-#define DEFAULT_TIMEOUT 10
-/* Commands for the daemon passed in the message below. */
-enum dmeventd_command {
-	CMD_ACTIVE = 1,
-	CMD_REGISTER_FOR_EVENT,
-	CMD_UNREGISTER_FOR_EVENT,
-	CMD_GET_REGISTERED_DEVICE,
-	CMD_GET_NEXT_REGISTERED_DEVICE,
-	CMD_SET_TIMEOUT,
-	CMD_GET_TIMEOUT,
-};
-
-/* Message passed between client and daemon. */
-struct daemon_message {
-	union {
-		unsigned int cmd;
-		int	 status;
-	} opcode;
-	char msg[252];
-} __attribute__((packed));
-
-/* Fifos for client/daemon communication. */
-struct fifos {
-	int client;
-	int server;
-	const char *client_path;
-	const char *server_path;
-};
-
-/* Event type definitions. */
-enum event_type {
-	SINGLE		= 0x01, /* Report multiple errors just once. */
-	MULTI		= 0x02, /* Report all of them. */
-	SECTOR_ERROR	= 0x04, /* Failure on a particular sector. */
-	DEVICE_ERROR	= 0x08, /* Device failure. */
-	PATH_ERROR	= 0x10, /* Failure on an io path. */
-	ADAPTOR_ERROR	= 0x20, /* Failure off a host adaptor. */
-	SYNC_STATUS	= 0x40, /* Mirror synchronization completed/failed. */
-	TIMEOUT		= 0x80, /* Timeout has occured */
-};
-#define	ALL_ERRORS (SECTOR_ERROR | DEVICE_ERROR | PATH_ERROR | ADAPTOR_ERROR)
-
-/* Prototypes for event lib interface. */
-int dm_register_for_event(char *dso_name, char *device, enum event_type events);
-int dm_unregister_for_event(char *dso_name, char *device,
-			    enum event_type events);
-int dm_get_registered_device(char **dso_name, char **device,
-			     enum event_type *events, int next);
-int dm_set_event_timeout(char *device, uint32_t timeout);
-int dm_get_event_timeout(char *device, uint32_t *timeout);
-
-/* Prototypes for DSO interface. */
-void process_event(char *device, enum event_type event);
-int register_device(char *device);
-int unregister_device(char *device);
-
-#endif
-
-/*
- * Overrides for Emacs so that we follow Linus's tabbing style.
- * Emacs will notice this stuff at the end of the file and automatically
- * adjust the settings for this buffer only.  This must remain at the end
- * of the file.
- * ---------------------------------------------------------------------------
- * Local variables:
- * c-file-style: "linux"
- * End:
- */

libdm/libdm-file.c

@@ -20,10 +20,6 @@
 #include <fcntl.h>
 #include <dirent.h>
 
-#ifdef linux
-#  include <malloc.h>
-#endif
-
 static int _create_dir_recursive(const char *dir)
 {
 	char *orig, *s;
@@ -31,7 +27,7 @@ static int _create_dir_recursive(const char *dir)
 
 	log_verbose("Creating directory \"%s\"", dir);
 	/* Create parent directories */
-	orig = s = strdup(dir);
+	orig = s = dm_strdup(dir);
 	while ((s = strchr(s, '/')) != NULL) {
 		*s = '\0';
 		if (*orig) {
@@ -39,13 +35,13 @@ static int _create_dir_recursive(const char *dir)
 			if (rc < 0 && errno != EEXIST) {
 				log_error("%s: mkdir failed: %s", orig,
 					  strerror(errno));
-				free(orig);
+				dm_free(orig);
 				return 0;
 			}
 		}
 		*s++ = '/';
 	}
-	free(orig);
+	dm_free(orig);
 
 	/* Create final directory */
 	rc = mkdir(dir, 0777);

libdm/mm/dbg_malloc.c

@@ -18,9 +18,9 @@
 #include <assert.h>
 #include <stdarg.h>
 
-char *dm_strdup(const char *str)
+char *dm_strdup_aux(const char *str, const char *file, int line)
 {
-	char *ret = dm_malloc(strlen(str) + 1);
+	char *ret = dm_malloc_aux_debug(strlen(str) + 1, file, line);
 
 	if (ret)
 		strcpy(ret, str);
@@ -28,8 +28,6 @@ char *dm_strdup(const char *str)
 	return ret;
 }
 
-#ifdef DEBUG_MEM
-
 struct memblock {
 	struct memblock *prev, *next;	/* All allocated blocks are linked */
 	size_t length;		/* Size of the requested block */
@@ -51,7 +49,7 @@ static struct {
 static struct memblock *_head = 0;
 static struct memblock *_tail = 0;
 
-void *dm_malloc_aux(size_t s, const char *file, int line)
+void *dm_malloc_aux_debug(size_t s, const char *file, int line)
 {
 	struct memblock *nb;
 	size_t tsize = s + sizeof(*nb) + sizeof(unsigned long);
@@ -72,9 +70,7 @@ void *dm_malloc_aux(size_t s, const char *file, int line)
 	nb->file = file;
 	nb->line = line;
 
-#ifdef BOUNDS_CHECK
 	dm_bounds_check();
-#endif
 
 	/* setup fields */
 	nb->magic = nb + 1;
@@ -125,9 +121,7 @@ void dm_free_aux(void *p)
 	if (!p)
 		return;
 
-#ifdef BOUNDS_CHECK
 	dm_bounds_check();
-#endif
 
 	/* sanity check */
 	assert(mb->magic == p);
@@ -171,7 +165,7 @@ void *dm_realloc_aux(void *p, unsigned int s, const char *file, int line)
 	void *r;
 	struct memblock *mb = ((struct memblock *) p) - 1;
 
-	r = dm_malloc_aux(s, file, line);
+	r = dm_malloc_aux_debug(s, file, line);
 
 	if (p) {
 		memcpy(r, p, mb->length);
@@ -181,7 +175,7 @@ void *dm_realloc_aux(void *p, unsigned int s, const char *file, int line)
 	return r;
 }
 
-int dm_dump_memory(void)
+int dm_dump_memory_debug(void)
 {
 	unsigned long tot = 0;
 	struct memblock *mb;
@@ -204,9 +198,9 @@ int dm_dump_memory(void)
 		}
 		str[sizeof(str) - 1] = '\0';
 
-		_log(_LOG_INFO, mb->file, mb->line,
-		     "block %d at %p, size %" PRIsize_t "\t [%s]",
-		     mb->id, mb->magic, mb->length, str);
+		dm_log(_LOG_INFO, mb->file, mb->line,
+		       "block %d at %p, size %" PRIsize_t "\t [%s]",
+		       mb->id, mb->magic, mb->length, str);
 		tot += mb->length;
 	}
 
@@ -216,7 +210,7 @@ int dm_dump_memory(void)
 	return 1;
 }
 
-void dm_bounds_check(void)
+void dm_bounds_check_debug(void)
 {
 	struct memblock *mb = _head;
 	while (mb) {
@@ -230,8 +224,6 @@ void dm_bounds_check(void)
 	}
 }
 
-#else
-
 void *dm_malloc_aux(size_t s, const char *file, int line)
 {
 	if (s > 50000000) {
@@ -242,5 +234,3 @@ void *dm_malloc_aux(size_t s, const char *file, int line)
 
 	return malloc(s);
 }
-
-#endif

libdm/mm/pool-fast.c

@@ -85,7 +85,7 @@ void *dm_pool_alloc_aligned(struct dm_pool *p, size_t s, unsigned alignment)
 	/* have we got room ? */
 	if (!c || (c->begin > c->end) || (c->end - c->begin < s)) {
 		/* allocate new chunk */
-		int needed = s + alignment + sizeof(struct chunk);
+		size_t needed = s + alignment + sizeof(struct chunk);
 		c = _new_chunk(p, (needed > p->chunk_size) ?
 			       needed : p->chunk_size);
 

make.tmpl.in

@@ -27,11 +27,9 @@ LIBS = @LIBS@
 CFLAGS += @DEFS@
 CFLAGS += @CFLAGS@
 CLDFLAGS += @CLDFLAGS@
-CLDWHOLEARCHIVE += @CLDWHOLEARCHIVE@
-CLDNOWHOLEARCHIVE += @CLDNOWHOLEARCHIVE@
 LDDEPS += @LDDEPS@
 LDFLAGS += @LDFLAGS@
-SOFLAG += @SOFLAG@
+LIB_SUFFIX = @LIB_SUFFIX@
 
 # Setup directory variables
 prefix = @prefix@
@@ -56,7 +54,7 @@ ifndef MAKEFLAGS
 MAKEFLAGS = @JOBS@
 endif
 
-SUFFIXES = .c .d .o .so .a .po .pot .mo
+SUFFIXES = .c .d .o .so .a .po .pot .mo .dylib
 
 CFLAGS += -fPIC -Wall -Wundef -Wshadow -Wcast-align -Wwrite-strings -Wmissing-prototypes -Wmissing-declarations -Wnested-externs -Winline
 
@@ -151,9 +149,21 @@ $(TARGETS): $(OBJECTS)
 %.so: %.o
 	$(CC) -c $(INCLUDES) $(CFLAGS) $< -o $@
 
+ifeq ("@LIB_SUFFIX@","so")
 $(LIB_SHARED): $(OBJECTS) $(LDDEPS)
-	$(CC) $(SOFLAG) -Wl,-soname,$(notdir $@).$(LIB_VERSION) \
-	$(CLDFLAGS) $(OBJECTS) -o $@
+	$(CC) -shared -Wl,-soname,$(notdir $@).$(LIB_VERSION) \
+	$(CLDFLAGS) $(OBJECTS) $(LIBS) -o $@
+endif
+
+ifeq ("@LIB_SUFFIX@","dylib")
+$(LIB_SHARED): $(OBJECTS) $(LDDEPS)
+	$(CC) -dynamiclib -dylib_current_version,$(LIB_VERSION) \
+	$(CLDFLAGS) $(OBJECTS) $(LIBS) -o $@
+endif
+
+%.so: %.a
+	$(CC) -shared -Wl,-soname,$(notdir $@).$(LIB_VERSION) \
+	$(CLDFLAGS) $(LIBS) -o $@ @CLDWHOLEARCHIVE@ $< @CLDNOWHOLEARCHIVE@
 
 $(LIB_STATIC): $(OBJECTS)
 	$(RM) $@

man/dmsetup.8.in

@@ -1,4 +1,4 @@
-.TH DMSETUP 8 "Sep 17 2003" "Linux" "MAINTENTANCE COMMANDS"
+.TH DMSETUP 8 "Apr 06 2006" "Linux" "MAINTENTANCE COMMANDS"
 .SH NAME
 dmsetup \- low level logical volume management
 .SH SYNOPSIS
@@ -12,7 +12,7 @@ dmsetup \- low level logical volume management
 .B dmsetup remove_all
 .br
 .B dmsetup suspend
-.I device_name
+.I [--nolockfs] device_name
 .br
 .B dmsetup resume
 .I device_name
@@ -29,7 +29,7 @@ dmsetup \- low level logical volume management
 .B dmsetup rename
 .I device_name new_name
 .br
-.B dmsetup ls [--target target_type] [--exec command]
+.B dmsetup ls [--target target_type] [--exec command] [--tree [-o options]]
 .br
 .B dmsetup info 
 .I [device_name]
@@ -147,10 +147,17 @@ Outputs some brief information about the device in the form:
 .IP \fBls
 .I [--target target_type]
 .I [--exec command]
+.I [--tree [-o options]]
 .br
 List device names.  Optionally only list devices that have at least
 one target of the specified type.  Optionally execute a command for
 each device.  The device name is appended to the supplied command.
+--tree displays dependencies between devices as a tree.
+It accepts a comma-separate list of options.
+Some specify the information displayed against each node:
+device/nodevice; active, open, rw, uuid.
+Others specify how the tree is displayed:
+ascii, utf, vt100; compact, inverted, notrunc.
 .IP \fBload|reload
 .I device_name [table_file]
 .br
@@ -183,11 +190,14 @@ Outputs status information for each of the device's targets.
 With --target, only information relating to the specified target type
 is displayed.
 .IP \fBsuspend
+.I [--nolockfs]
 .I device_name
 .br
 Suspends a device.  Any I/O that has already been mapped by the device
 but has not yet completed will be flushed.  Any further I/O to that
 device will be postponed for as long as the device is suspended.
+If there's a filesystem on the device which supports the operation, 
+an attempt will be made to sync it first unless --nolockfs is specified.
 .IP \fBtable
 .I [--target target_type]
 .I [device_name]

man/lvdisplay.8

@@ -43,8 +43,8 @@ The values are:
 Display the mapping of logical extents to physical volumes and
 physical extents.
 .SH Examples
-"lvdisplay -v /dev/vg00/lvol2" shows attributes of that logical volume
-and its mapping of logical to physical extents. In case snapshot
+"lvdisplay -v /dev/vg00/lvol2" shows attributes of that logical volume.
+If snapshot
 logical volumes have been created for this original logical volume,
 this command shows a list of all snapshot logical volumes and their
 status (active or inactive) as well.

man/vgchange.8

@@ -17,6 +17,8 @@ vgchange \- change attributes of a volume group
 .RB [ \-l | \-\-logicalvolume
 .IR MaxLogicalVolumes ]
 .RB [ \-P | \-\-partial]
+.RB [ \-s | \-\-physicalextentsize
+.IR PhysicalExtentSize [ \fBkKmMgGtT\fR ]]
 .RB [ -t | \-\-test]
 .RB [ \-v | \-\-verbose]
 .RB [ \-\-version ]
@@ -59,6 +61,27 @@ exclusively because they can only be used on one node at once.
 Changes the maximum logical volume number of an existing inactive
 volume group.
 .TP
+.BR \-s ", " \-\-physicalextentsize " " \fIPhysicalExtentSize\fR[\fBkKmMgGtT\fR]
+Changes the physical extent size on physical volumes of this volume group.
+A size suffix (k for kilobytes up to t for terabytes) is optional, megabytes
+is the default if no suffix is present.
+The default is 4 MB and it must be at least 1 KB and a power of 2.
+ 
+Before increasing the physical extent size, you might need to use lvresize,
+pvresize and/or pvmove so that everything fits.  For example, every
+contiguous range of extents used in a logical volume must start and 
+end on an extent boundary.
+ 
+If the volume group metadata uses lvm1 format, extents can vary in size from
+8KB to 16GB and there is a limit of 65534 extents in each logical volume.  The
+default of 4 MB leads to a maximum logical volume size of around 256GB.
+ 
+If the volume group metadata uses lvm2 format those restrictions do not apply,
+but having a large number of extents will slow down the tools but have no
+impact on I/O performance to the logical volume.  The smallest PE is 1KB.
+ 
+The 2.4 kernel has a limitation of 2TB per block device.
+.TP
 .BR \-x ", " \-\-resizeable { y | n }
 Enables or disables the extension/reduction of this volume group
 with/by physical volumes.

man/vgcreate.8

@@ -59,17 +59,22 @@ as described in \fBpvcreate(8)\fP.
 .BR \-s ", " \-\-physicalextentsize " " \fIPhysicalExtentSize\fR[\fBkKmMgGtT\fR]
 Sets the physical extent size on physical volumes of this volume group.
 A size suffix (k for kilobytes up to t for terabytes) is optional, megabytes
-is the default if no suffix is present.  Values can be from 8 KB to 16 GB in
-powers of 2.  The default is 4 MB.
+is the default if no suffix is present.  
+The default is 4 MB and it must be at least 1 KB and a power of 2.
+
 Once this value has been set, it is difficult to change it without recreating
 the volume group which would involve backing up and restoring data on any 
-logical volumes.
-If the volume group metadata uses lvm1 format, there is a limit of 65534
-extents in each logical volume, so the default of 4 MB leads to a maximum 
-logical volume size of around 256GB.
-If the volume group metadata uses lvm2 format there is no such restriction,
-although having a large number of extents will slow down
-the tools but have no impact on I/O performance to the logical volume.
+logical volumes.  However, if no extents need moving for the new
+value to apply, it can be altered using vgchange \-s.
+
+If the volume group metadata uses lvm1 format, extents can vary in size from
+8KB to 16GB and there is a limit of 65534 extents in each logical volume.  The
+default of 4 MB leads to a maximum logical volume size of around 256GB.  
+
+If the volume group metadata uses lvm2 format those restrictions do not apply,
+but having a large number of extents will slow down the tools but have no
+impact on I/O performance to the logical volume.  The smallest PE is 1KB.
+
 The 2.4 kernel has a limitation of 2TB per block device.
 .SH EXAMPLES
 To create a volume group named

tools/Makefile.in

@@ -87,6 +87,10 @@ ifeq ("@CMDLIB@", "yes")
 	INSTALL_TARGETS += $(INSTALL_CMDLIB_TARGETS)
 endif
 
+ifeq ("@DMEVENTD@", "yes")
+	LVMLIBS += -ldevmapper-event -lpthread
+endif
+
 ifeq ("@DEVMAPPER@", "yes")
 	LVMLIBS += -ldevmapper
 endif
@@ -107,8 +111,6 @@ liblvm2cmd.a: $(top_srcdir)/lib/liblvm.a $(OBJECTS)
 	$(AR) rs $@ $(OBJECTS)
 
 liblvm2cmd.so: liblvm2cmd.a $(LDDEPS)
-	$(CC) -o liblvm2cmd.so $(SOFLAG) $(CLDFLAGS) \
-		$(CLDWHOLEARCHIVE) liblvm2cmd.a $(CLDNOWHOLEARCHIVE)
 
 .commands: commands.h cmdnames.h Makefile
 	$(CC) -E -P cmdnames.h 2> /dev/null | \

tools/args.h

@@ -44,6 +44,7 @@ arg(minor_ARG, '\0', "minor", minor_arg)
 arg(type_ARG, '\0', "type", segtype_arg)
 arg(alloc_ARG, '\0', "alloc", alloc_arg)
 arg(separator_ARG, '\0', "separator", string_arg)
+arg(mirrorsonly_ARG, '\0', "mirrorsonly", NULL)
 
 /* Allow some variations */
 arg(resizable_ARG, '\0', "resizable", yes_no_arg)
@@ -85,7 +86,7 @@ arg(size_ARG, 'L', "size", size_mb_arg)
 arg(logicalextent_ARG, 'L', "logicalextent", int_arg_with_sign)
 arg(persistent_ARG, 'M', "persistent", yes_no_arg)
 arg(major_ARG, 'j', "major", major_arg)
-arg(mirrors_ARG, 'm', "mirrors", int_arg)
+arg(mirrors_ARG, 'm', "mirrors", int_arg_with_sign)
 arg(metadatatype_ARG, 'M', "metadatatype", metadatatype_arg)
 arg(maps_ARG, 'm', "maps", NULL)
 arg(name_ARG, 'n', "name", string_arg)

tools/commands.h

@@ -80,16 +80,26 @@ xx(lvchange,
 
 xx(lvconvert,
    "Change logical volume layout",
-   "lvconvert " "\n"
+   "lvconvert "
+   "[-m|--mirrors Mirrors]\n"
    "\t[--alloc AllocationPolicy]\n"
    "\t[-d|--debug]\n"
    "\t[-h|-?|--help]\n"
-   "\t[-m|--mirrors Mirrors]\n"
    "\t[-v|--verbose]\n"
    "\t[--version]" "\n"
-   "\tLogicalVolume[Path] [PhysicalVolume[Path]...]\n",
+   "\tLogicalVolume[Path] [PhysicalVolume[Path]...]\n\n"
 
-   alloc_ARG, mirrors_ARG, test_ARG)
+   "lvconvert "
+   "[-s|--snapshot]\n"
+   "\t[-c|--chunksize]\n"
+   "\t[-d|--debug]\n"
+   "\t[-h|-?|--help]\n"
+   "\t[-v|--verbose]\n"
+   "\t[-Z|--zero {y|n}]\n"
+   "\t[--version]" "\n"
+   "\tOriginalLogicalVolume[Path] SnapshotLogicalVolume[Path]\n",
+
+   alloc_ARG, chunksize_ARG, mirrors_ARG, snapshot_ARG, test_ARG, zero_ARG)
 
 xx(lvcreate,
    "Create a logical volume",
@@ -526,6 +536,7 @@ xx(pvs,
    "\t[--nosuffix]\n"
    "\t[-o|--options [+]Field[,Field]]\n"
    "\t[-O|--sort [+|-]key1[,[+|-]key2[,...]]]\n"
+   "\t[-P|--partial] " "\n"
    "\t[--segments]\n"
    "\t[--separator Separator]\n"
    "\t[--unbuffered]\n"
@@ -535,8 +546,8 @@ xx(pvs,
    "\t[PhysicalVolume [PhysicalVolume...]]\n",
    
    aligned_ARG, all_ARG, ignorelockingfailure_ARG, noheadings_ARG,
-   nolocking_ARG, nosuffix_ARG, options_ARG, segments_ARG, separator_ARG,
-   sort_ARG, unbuffered_ARG, units_ARG)
+   nolocking_ARG, nosuffix_ARG, options_ARG, partial_ARG, segments_ARG,
+   separator_ARG, sort_ARG, unbuffered_ARG, units_ARG)
 
 xx(pvscan,
    "List all physical volumes",
@@ -769,6 +780,7 @@ xx(vgreduce,
    "\t[-A|--autobackup y|n]\n"
    "\t[-d|--debug]\n"
    "\t[-h|--help]\n"
+   "\t[--mirrorsonly]\n"
    "\t[--removemissing]\n"
    "\t[-t|--test]\n"
    "\t[-v|--verbose]\n"
@@ -776,7 +788,7 @@ xx(vgreduce,
    "\tVolumeGroupName\n"
    "\t[PhysicalVolumePath...]\n",
 
-   all_ARG, autobackup_ARG, removemissing_ARG, test_ARG)
+   all_ARG, autobackup_ARG, mirrorsonly_ARG, removemissing_ARG, test_ARG)
 
 xx(vgremove,
    "Remove volume group(s)",

tools/dmsetup.c

@@ -86,8 +86,10 @@ enum {
 	READ_ONLY = 0,
 	COLS_ARG,
 	EXEC_ARG,
+	GID_ARG,
 	MAJOR_ARG,
 	MINOR_ARG,
+	MODE_ARG,
 	NOHEADINGS_ARG,
 	NOLOCKFS_ARG,
 	NOOPENCOUNT_ARG,
@@ -95,6 +97,7 @@ enum {
 	OPTIONS_ARG,
 	TARGET_ARG,
 	TREE_ARG,
+	UID_ARG,
 	UUID_ARG,
 	VERBOSE_ARG,
 	VERSION_ARG,
@@ -390,6 +393,15 @@ static int _create(int argc, char **argv, void *data)
 	if (_switches[MINOR_ARG] && !dm_task_set_minor(dmt, _values[MINOR_ARG]))
 		goto out;
 
+	if (_switches[UID_ARG] && !dm_task_set_uid(dmt, _values[UID_ARG]))
+		goto out;
+
+	if (_switches[GID_ARG] && !dm_task_set_gid(dmt, _values[GID_ARG]))
+		goto out;
+
+	if (_switches[MODE_ARG] && !dm_task_set_mode(dmt, _values[MODE_ARG]))
+		goto out;
+
 	if (_switches[NOOPENCOUNT_ARG] && !dm_task_no_open_count(dmt))
 		goto out;
 
@@ -438,7 +450,8 @@ static int _rename(int argc, char **argv, void *data)
 
 static int _message(int argc, char **argv, void *data)
 {
-	int r = 0, sz = 1, i;
+	int r = 0, i;
+	size_t sz = 1;
 	struct dm_task *dmt;
 	char *str;
 
@@ -455,7 +468,7 @@ static int _message(int argc, char **argv, void *data)
 		argv++;
 	}
 
-	if (!dm_task_set_sector(dmt, atoll(argv[1])))
+	if (!dm_task_set_sector(dmt, (uint64_t) atoll(argv[1])))
 		goto out;
 
 	argc -= 2;
@@ -495,6 +508,39 @@ static int _message(int argc, char **argv, void *data)
 	return r;
 }
 
+static int _setgeometry(int argc, char **argv, void *data)
+{
+	int r = 0;
+	struct dm_task *dmt;
+
+	if (!(dmt = dm_task_create(DM_DEVICE_SET_GEOMETRY)))
+		return 0;
+
+	if (_switches[UUID_ARG] || _switches[MAJOR_ARG]) {
+		if (!_set_task_device(dmt, NULL, 0))
+			goto out;
+	} else {
+		if (!_set_task_device(dmt, argv[1], 0))
+			goto out;
+		argc--;
+		argv++;
+	}
+
+	if (!dm_task_set_geometry(dmt, argv[1], argv[2], argv[3], argv[4]))
+		goto out;
+
+	/* run the task */
+	if (!dm_task_run(dmt))
+		goto out;
+
+	r = 1;
+
+      out:
+	dm_task_destroy(dmt);
+
+	return r;
+}
+
 static int _version(int argc, char **argv, void *data)
 {
 	char version[80];
@@ -502,7 +548,7 @@ static int _version(int argc, char **argv, void *data)
 	if (dm_get_library_version(version, sizeof(version)))
 		printf("Library version:   %s\n", version);
 
-	if (dm_driver_version(version, sizeof(version)))
+	if (!dm_driver_version(version, sizeof(version)))
 		return 0;
 
 	printf("Driver version:    %s\n", version);
@@ -580,7 +626,7 @@ static int _wait(int argc, char **argv, void *data)
 	}
 
 	return _simple(DM_DEVICE_WAITEVENT, name,
-		       (argc > 1) ? atoi(argv[argc - 1]) : 0, 1);
+		       (argc > 1) ? (uint32_t) atoi(argv[argc - 1]) : 0, 1);
 }
 
 static int _process_all(int argc, char **argv,
@@ -622,7 +668,7 @@ static int _process_all(int argc, char **argv,
 	return r;
 }
 
-static void _display_dev(struct dm_task *dmt, char *name)
+static void _display_dev(struct dm_task *dmt, const char *name)
 {
 	struct dm_info info;
 
@@ -635,7 +681,7 @@ static int _mknodes(int argc, char **argv, void *data)
 	return dm_mknodes(argc > 1 ? argv[1] : NULL);
 }
 
-static int _exec_command(char *name)
+static int _exec_command(const char *name)
 {
 	int n;
 	static char path[PATH_MAX];
@@ -651,7 +697,7 @@ static int _exec_command(char *name)
 		return 0;
 
 	n = snprintf(path, sizeof(path), "%s/%s", dm_dir(), name);
-	if (n < 0 || n > sizeof(path) - 1)
+	if (n < 0 || n > (int) sizeof(path) - 1)
 		return 0;
 
 	if (!argc) {
@@ -704,7 +750,7 @@ static int _status(int argc, char **argv, void *data)
 	char *params;
 	int cmd;
 	struct dm_names *names = (struct dm_names *) data;
-	char *name = NULL;
+	const char *name = NULL;
 	int matched = 0;
 	int ls_only = 0;
 	struct dm_info info;
@@ -742,7 +788,7 @@ static int _status(int argc, char **argv, void *data)
 		goto out;
 
 	if (!name)
-		name = (char *) dm_task_get_name(dmt);
+		name = dm_task_get_name(dmt);
 
 	/* Fetch targets and print 'em */
 	do {
@@ -1292,6 +1338,7 @@ struct command {
 
 static struct command _commands[] = {
 	{"create", "<dev_name> [-j|--major <major> -m|--minor <minor>]\n"
+	  "\t                  [-U|--uid <uid>] [-G|--gid <gid>] [-M|--mode <octal_mode>]\n"
 	  "\t                  [-u|uuid <uuid>] [--notable] [<table_file>]",
 	 1, 2, _create},
 	{"remove", "<device>", 0, 1, _remove},
@@ -1303,7 +1350,7 @@ static struct command _commands[] = {
 	{"reload", "<device> [<table_file>]", 0, 2, _load},
 	{"rename", "<device> <new_name>", 1, 2, _rename},
 	{"message", "<device> <sector> <message>", 2, -1, _message},
-	{"ls", "[--target <target_type>] [--exec <command>] [--tree]", 0, 0, _ls},
+	{"ls", "[--target <target_type>] [--exec <command>] [--tree [-o options]]", 0, 0, _ls},
 	{"info", "[<device>]", 0, 1, _info},
 	{"deps", "[<device>]", 0, 1, _deps},
 	{"status", "[<device>] [--target <target_type>]", 0, 1, _status},
@@ -1312,6 +1359,7 @@ static struct command _commands[] = {
 	{"mknodes", "[<device>]", 0, 1, _mknodes},
 	{"targets", "", 0, 0, _targets},
 	{"version", "", 0, 0, _version},
+	{"setgeometry", "<device> <cyl> <head> <sect> <start>", 5, 5, _setgeometry},
 	{NULL, NULL, 0, 0, NULL}
 };
 
@@ -1326,7 +1374,9 @@ static void _usage(FILE *out)
 		fprintf(out, "\t%s %s\n", _commands[i].name, _commands[i].help);
 	fprintf(out, "\n<device> may be device name or -u <uuid> or "
 		     "-j <major> -m <minor>\n");
-	fprintf(out, "Table_file contents may be supplied on stdin.\n\n");
+	fprintf(out, "Table_file contents may be supplied on stdin.\n");
+	fprintf(out, "Tree options are: ascii, utf, vt100; compact, inverted, notrunc;\n"
+		     "                  [no]device, active, open, rw and uuid.\n\n");
 	return;
 }
 
@@ -1418,8 +1468,10 @@ static int _process_switches(int *argc, char ***argv)
 		{"readonly", 0, &ind, READ_ONLY},
 		{"columns", 0, &ind, COLS_ARG},
 		{"exec", 1, &ind, EXEC_ARG},
+		{"gid", 1, &ind, GID_ARG},
 		{"major", 1, &ind, MAJOR_ARG},
 		{"minor", 1, &ind, MINOR_ARG},
+		{"mode", 1, &ind, MODE_ARG},
 		{"noheadings", 0, &ind, NOHEADINGS_ARG},
 		{"nolockfs", 0, &ind, NOLOCKFS_ARG},
 		{"noopencount", 0, &ind, NOOPENCOUNT_ARG},
@@ -1427,6 +1479,7 @@ static int _process_switches(int *argc, char ***argv)
 		{"options", 1, &ind, OPTIONS_ARG},
 		{"target", 1, &ind, TARGET_ARG},
 		{"tree", 0, &ind, TREE_ARG},
+		{"uid", 1, &ind, UID_ARG},
 		{"uuid", 1, &ind, UUID_ARG},
 		{"verbose", 1, &ind, VERBOSE_ARG},
 		{"version", 0, &ind, VERSION_ARG},
@@ -1478,7 +1531,7 @@ static int _process_switches(int *argc, char ***argv)
 
 	optarg = 0;
 	optind = OPTIND_INIT;
-	while ((ind = -1, c = GETOPTLONG_FN(*argc, *argv, "cCj:m:no:ru:v",
+	while ((ind = -1, c = GETOPTLONG_FN(*argc, *argv, "cCGj:m:Mno:ru:Uv",
 					    long_options, NULL)) != -1) {
 		if (c == 'c' || c == 'C' || ind == COLS_ARG)
 			_switches[COLS_ARG]++;
@@ -1504,6 +1557,19 @@ static int _process_switches(int *argc, char ***argv)
 			_switches[UUID_ARG]++;
 			_uuid = optarg;
 		}
+		if (c == 'G' || ind == GID_ARG) {
+			_switches[GID_ARG]++;
+			_values[GID_ARG] = atoi(optarg);
+		}
+		if (c == 'U' || ind == UID_ARG) {
+			_switches[UID_ARG]++;
+			_values[UID_ARG] = atoi(optarg);
+		}
+		if (c == 'M' || ind == MODE_ARG) {
+			_switches[MODE_ARG]++;
+			/* FIXME Accept modes as per chmod */
+			_values[MODE_ARG] = (int) strtol(optarg, NULL, 8);
+		}
 		if ((ind == EXEC_ARG)) {
 			_switches[EXEC_ARG]++;
 			_command = optarg;

tools/lvconvert.c

@@ -13,10 +13,23 @@
  */
 
 #include "tools.h"
+#include "lv_alloc.h"
 
 struct lvconvert_params {
+	int snapshot;
+	int zero;
+
+	const char *origin;
 	const char *lv_name;
+	const char *vg_name;
+
+	uint32_t chunk_size;
+	uint32_t region_size;
+
 	uint32_t mirrors;
+	sign_t mirrors_sign;
+
+	struct segment_type *segtype;
 
 	alloc_policy_t alloc;
 
@@ -25,30 +38,164 @@ struct lvconvert_params {
 	struct list *pvh;
 };
 
+static int _read_name_params(struct lvconvert_params *lp,
+			     struct cmd_context *cmd, int *pargc, char ***pargv)
+{
+	char *ptr;
+	const char *vg_name = NULL;
+
+	if (lp->snapshot) {
+		if (!*pargc) {
+			log_error("Please specify a logical volume to act as "
+				  "the snapshot origin.");
+			return 0;
+		}
+
+		lp->origin = *pargv[0];
+		(*pargv)++, (*pargc)--;
+		if (!(lp->vg_name = extract_vgname(cmd, lp->origin))) {
+			log_error("The origin name should include the "
+				  "volume group.");
+			return 0;
+		}
+
+		/* Strip the volume group from the origin */
+		if ((ptr = strrchr(lp->origin, (int) '/')))
+			lp->origin = ptr + 1;
+	}
+
+	if (!*pargc) {
+		log_error("Please provide logical volume path");
+		return 0;
+	}
+
+	lp->lv_name = (*pargv)[0];
+	(*pargv)++, (*pargc)--;
+
+	if (strchr(lp->lv_name, '/') &&
+	    (vg_name = extract_vgname(cmd, lp->lv_name)) &&
+	    lp->vg_name && strcmp(vg_name, lp->vg_name)) {
+		log_error("Please use a single volume group name "
+			  "(\"%s\" or \"%s\")", vg_name, lp->vg_name);
+		return 0;
+	}
+
+	if (!lp->vg_name)
+		lp->vg_name = vg_name;
+
+	if (!validate_name(lp->vg_name)) {
+		log_error("Please provide a valid volume group name");
+		return 0;
+	}
+
+	if ((ptr = strrchr(lp->lv_name, '/')))
+		lp->lv_name = ptr + 1;
+
+	if (!apply_lvname_restrictions(lp->lv_name))
+		return_0;
+
+	return 1;
+}
+
 static int _read_params(struct lvconvert_params *lp, struct cmd_context *cmd,
 			int argc, char **argv)
 {
 	memset(lp, 0, sizeof(*lp));
 
+	if (arg_count(cmd, mirrors_ARG) + arg_count(cmd, snapshot_ARG) != 1) {
+		log_error("Exactly one of --mirrors or --snapshot arguments "
+			  "required.");
+		return 0;
+	}
+
+	if (arg_count(cmd, snapshot_ARG))
+		lp->snapshot = 1;
+
+	if (arg_count(cmd, mirrors_ARG)) {
+		lp->mirrors = arg_uint_value(cmd, mirrors_ARG, 0);
+		lp->mirrors_sign = arg_sign_value(cmd, mirrors_ARG, 0);
+	}
+
 	lp->alloc = ALLOC_INHERIT;
 	if (arg_count(cmd, alloc_ARG))
 		lp->alloc = (alloc_policy_t) arg_uint_value(cmd, alloc_ARG,
 							    lp->alloc);
 
-	if (!arg_count(cmd, mirrors_ARG)) {
-		log_error("--mirrors argument required");
+	if (lp->snapshot) {
+		if (arg_count(cmd, regionsize_ARG)) {
+			log_error("--regionsize is only available with mirrors");
+			return 0;
+		}
+
+		if (arg_sign_value(cmd, chunksize_ARG, 0) == SIGN_MINUS) {
+			log_error("Negative chunk size is invalid");
+			return 0;
+		}
+		lp->chunk_size = 2 * arg_uint_value(cmd, chunksize_ARG, 8);
+		if (lp->chunk_size < 8 || lp->chunk_size > 1024 ||
+		    (lp->chunk_size & (lp->chunk_size - 1))) {
+			log_error("Chunk size must be a power of 2 in the "
+				  "range 4K to 512K");
+			return 0;
+		}
+		log_verbose("Setting chunksize to %d sectors.", lp->chunk_size);
+
+		if (!(lp->segtype = get_segtype_from_string(cmd, "snapshot")))
+			return_0;
+
+		lp->zero = strcmp(arg_str_value(cmd, zero_ARG,
+						(lp->segtype->flags &
+						 SEG_CANNOT_BE_ZEROED) ?
+						"n" : "y"), "n");
+
+	} else {	/* Mirrors */
+		if (arg_count(cmd, chunksize_ARG)) {
+			log_error("--chunksize is only available with "
+				  "snapshots");
+			return 0;
+		}
+
+		if (arg_count(cmd, zero_ARG)) {
+			log_error("--zero is only available with snapshots");
+			return 0;
+		}
+
+		/*
+	 	 * --regionsize is only valid if converting an LV into a mirror.
+	 	 * Checked when we know the state of the LV being converted.
+	 	 */
+		if (arg_count(cmd, regionsize_ARG)) {
+			if (arg_sign_value(cmd, regionsize_ARG, 0) ==
+				    SIGN_MINUS) {
+				log_error("Negative regionsize is invalid");
+				return 0;
+			}
+			lp->region_size = 2 * arg_uint_value(cmd,
+							     regionsize_ARG, 0);
+		} else
+			lp->region_size = 2 * find_config_int(cmd->cft->root,
+						"activation/mirror_region_size",
+						DEFAULT_MIRROR_REGION_SIZE);
+	
+		if (lp->region_size & (lp->region_size - 1)) {
+			log_error("Region size (%" PRIu32
+				  ") must be a power of 2", lp->region_size);
+			return 0;
+		}
+
+		if (!(lp->segtype = get_segtype_from_string(cmd, "striped")))
+			return_0;
+	}
+
+	if (activation() && lp->segtype->ops->target_present &&
+	    !lp->segtype->ops->target_present()) {
+		log_error("%s: Required device-mapper target(s) not "
+			  "detected in your kernel", lp->segtype->name);
 		return 0;
 	}
 
-	lp->mirrors = arg_uint_value(cmd, mirrors_ARG, 0) + 1;
-
-	if (!argc) {
-		log_error("Please give logical volume path");
-		return 0;
-	}
-
-	lp->lv_name = argv[0];
-	argv++, argc--;
+	if (!_read_name_params(lp, cmd, &argc, &argv))
+		return_0;
 
 	lp->pv_count = argc;
 	lp->pvs = argv;
@@ -61,8 +208,32 @@ static int lvconvert_mirrors(struct cmd_context * cmd, struct logical_volume * l
 {
 	struct lv_segment *seg;
 	uint32_t existing_mirrors;
-	// struct alloc_handle *ah = NULL;
-	// struct logical_volume *log_lv;
+	struct alloc_handle *ah = NULL;
+	struct logical_volume *log_lv;
+	struct list *parallel_areas;
+
+	seg = first_seg(lv);
+	existing_mirrors = seg->area_count;
+
+	/* Adjust required number of mirrors */
+	if (lp->mirrors_sign == SIGN_PLUS)
+		lp->mirrors = existing_mirrors + lp->mirrors;
+	else if (lp->mirrors_sign == SIGN_MINUS) {
+		if (lp->mirrors >= existing_mirrors) {
+			log_error("Logical volume %s only has %" PRIu32 " mirrors.",
+				  lv->name, existing_mirrors);
+			return 0;
+		}
+		lp->mirrors = existing_mirrors - lp->mirrors;
+	} else
+		lp->mirrors += 1;
+
+	if (arg_count(cmd, regionsize_ARG) && (lv->status & MIRRORED) &&
+	    (lp->region_size != seg->region_size)) {
+		log_error("Mirror log region size cannot be changed on "
+			  "an existing mirror.");
+		return 0;
+	}
 
 	if ((lp->mirrors == 1)) {
 		if (!(lv->status & MIRRORED)) {
@@ -70,11 +241,10 @@ static int lvconvert_mirrors(struct cmd_context * cmd, struct logical_volume * l
 				  lv->name);
 			return 1;
 		}
-		/* FIXME If allocatable_pvs supplied only remove those */
-		if (!remove_all_mirror_images(lv)) {
-			stack;
-			return 0;
-		}
+
+		if (!remove_mirror_images(seg, 1,
+					  lp->pv_count ? lp->pvh : NULL, 1))
+			return_0;
 	} else {		/* mirrors > 1 */
 		if ((lv->status & MIRRORED)) {
 			if (list_size(&lv->segments) != 1) {
@@ -82,8 +252,6 @@ static int lvconvert_mirrors(struct cmd_context * cmd, struct logical_volume * l
 					  "mirror segments.", lv->name);
 				return 0;
 			}
-			seg = first_seg(lv);
-			existing_mirrors = seg->area_count;
 			if (lp->mirrors == existing_mirrors) {
 				log_error("Logical volume %s already has %"
 					  PRIu32 " mirror(s).", lv->name,
@@ -100,27 +268,58 @@ static int lvconvert_mirrors(struct cmd_context * cmd, struct logical_volume * l
 					  "supported yet.");
 				return 0;
 			} else {
-				if (!remove_mirror_images(seg, lp->mirrors)) {
-					stack;
+				/* Reduce number of mirrors */
+				if (!remove_mirror_images(seg, lp->mirrors,
+							  lp->pv_count ?
+							  lp->pvh : NULL, 0))
+					return_0;
+			}
+		} else {
+			/* Make existing LV into mirror set */
+			/* FIXME Share code with lvcreate */
+
+			/* FIXME Why is this restriction here?  Fix it! */
+			list_iterate_items(seg, &lv->segments) {
+				if (seg_is_striped(seg) && seg->area_count > 1) {
+					log_error("Mirrors of striped volumes are not yet supported.");
 					return 0;
 				}
 			}
-		} else {
-			/* FIXME Share code with lvcreate */
-			/* region size, log_name, create log_lv, zero it */
-			// Allocate (mirrors) new areas & log - replace mirrored_pv with mirrored_lv
-			// Restructure as mirror - add existing param to create_mirror_layers
-			log_error("Adding mirror images is not supported yet.");
-			return 0;
+
+			if (!(parallel_areas = build_parallel_areas_from_lv(cmd, lv)))
+				return_0;
+
+			if (!(ah = allocate_extents(lv->vg, NULL, lp->segtype,
+						    1, lp->mirrors - 1, 1,
+						    lv->le_count * (lp->mirrors - 1),
+						    NULL, 0, 0, lp->pvh,
+						    lp->alloc,
+						    parallel_areas)))
+				return_0;
+
+			lp->region_size = adjusted_mirror_region_size(lv->vg->extent_size,
+								      lv->le_count,
+								      lp->region_size);
+
+			if (!(log_lv = create_mirror_log(cmd, lv->vg, ah,
+							 lp->alloc,
+							 lv->name))) {
+				log_error("Failed to create mirror log.");
+				return 0;
+			}
+
+			if (!create_mirror_layers(ah, 1, lp->mirrors, lv,
+						  lp->segtype, 0,
+						  lp->region_size,
+						  log_lv))
+				return_0;
 		}
 	}
 
 	log_very_verbose("Updating logical volume \"%s\" on disk(s)", lv->name);
 
-	if (!vg_write(lv->vg)) {
-		stack;
-		return 0;
-	}
+	if (!vg_write(lv->vg))
+		return_0;
 
 	backup(lv->vg);
 
@@ -147,6 +346,68 @@ static int lvconvert_mirrors(struct cmd_context * cmd, struct logical_volume * l
 	return 1;
 }
 
+static int lvconvert_snapshot(struct cmd_context *cmd,
+			      struct logical_volume *lv,
+			      struct lvconvert_params *lp)
+{
+	struct logical_volume *org;
+
+	if (!(org = find_lv(lv->vg, lp->origin))) {
+		log_error("Couldn't find origin volume '%s'.", lp->origin);
+		return 0;
+	}
+
+	if (org->status & (LOCKED|PVMOVE) || lv_is_cow(org)) {
+		log_error("Unable to create a snapshot of a %s LV.",
+			  org->status & LOCKED ? "locked" :
+			  org->status & PVMOVE ? "pvmove" : "snapshot");
+		return 0;
+	}
+
+	if (!lp->zero)
+		log_error("WARNING: \"%s\" not zeroed", lv->name);
+	else if (!zero_lv(cmd, lv)) {
+			log_error("Aborting. Failed to wipe snapshot "
+				  "exception store.");
+			return 0;
+	}
+
+	if (!deactivate_lv(cmd, lv)) {
+		log_error("Couldn't deactivate LV %s.", lv->name);
+		return 0;
+	}
+
+	if (!vg_add_snapshot(lv->vg->fid, NULL, org, lv, NULL, org->le_count,
+			     lp->chunk_size)) {
+		log_error("Couldn't create snapshot.");
+		return 0;
+	}
+
+	/* store vg on disk(s) */
+	if (!vg_write(lv->vg))
+		return_0;
+
+	backup(lv->vg);
+
+	if (!suspend_lv(cmd, org)) {
+		log_error("Failed to suspend origin %s", org->name);
+		vg_revert(lv->vg);
+		return 0;
+	}
+
+	if (!vg_commit(lv->vg))
+		return_0;
+
+	if (!resume_lv(cmd, org)) {
+		log_error("Problem reactivating origin %s", org->name);
+		return 0;
+	}
+
+	log_print("Logical volume %s converted to snapshot.", lv->name);
+
+	return 1;
+}
+
 static int lvconvert_single(struct cmd_context *cmd, struct logical_volume *lv,
 			    void *handle)
 {
@@ -179,6 +440,11 @@ static int lvconvert_single(struct cmd_context *cmd, struct logical_volume *lv,
 			return ECMD_FAILED;
 		if (!lvconvert_mirrors(cmd, lv, lp))
 			return ECMD_FAILED;
+	} else if (lp->snapshot) {
+		if (!archive(lv->vg))
+			return ECMD_FAILED;
+		if (!lvconvert_snapshot(cmd, lv, lp))
+			return ECMD_FAILED;
 	}
 
 	return ECMD_PROCESSED;
@@ -186,8 +452,6 @@ static int lvconvert_single(struct cmd_context *cmd, struct logical_volume *lv,
 
 int lvconvert(struct cmd_context * cmd, int argc, char **argv)
 {
-	const char *vg_name;
-	char *st;
 	int consistent = 1;
 	struct volume_group *vg;
 	struct lv_list *lvl;
@@ -199,41 +463,31 @@ int lvconvert(struct cmd_context * cmd, int argc, char **argv)
 		return EINVALID_CMD_LINE;
 	}
 
-	vg_name = extract_vgname(cmd, lp.lv_name);
+	log_verbose("Checking for existing volume group \"%s\"", lp.vg_name);
 
-	if (!validate_name(vg_name)) {
-		log_error("Please provide a valid volume group name");
-		return EINVALID_CMD_LINE;
-	}
-
-	if ((st = strrchr(lp.lv_name, '/')))
-		lp.lv_name = st + 1;
-
-	log_verbose("Checking for existing volume group \"%s\"", vg_name);
-
-	if (!lock_vol(cmd, vg_name, LCK_VG_WRITE)) {
-		log_error("Can't get lock for %s", vg_name);
+	if (!lock_vol(cmd, lp.vg_name, LCK_VG_WRITE)) {
+		log_error("Can't get lock for %s", lp.vg_name);
 		return ECMD_FAILED;
 	}
 
-	if (!(vg = vg_read(cmd, vg_name, &consistent))) {
-		log_error("Volume group \"%s\" doesn't exist", vg_name);
+	if (!(vg = vg_read(cmd, lp.vg_name, NULL, &consistent))) {
+		log_error("Volume group \"%s\" doesn't exist", lp.vg_name);
 		goto error;
 	}
 
 	if (vg->status & EXPORTED_VG) {
-		log_error("Volume group \"%s\" is exported", vg_name);
+		log_error("Volume group \"%s\" is exported", lp.vg_name);
 		goto error;
 	}
 
 	if (!(vg->status & LVM_WRITE)) {
-		log_error("Volume group \"%s\" is read-only", vg_name);
+		log_error("Volume group \"%s\" is read-only", lp.vg_name);
 		goto error;
 	}
 
 	if (!(lvl = find_lv_in_vg(vg, lp.lv_name))) {
 		log_error("Logical volume \"%s\" not found in "
-			  "volume group \"%s\"", lp.lv_name, vg_name);
+			  "volume group \"%s\"", lp.lv_name, lp.vg_name);
 		goto error;
 	}
 
@@ -249,6 +503,6 @@ int lvconvert(struct cmd_context * cmd, int argc, char **argv)
 	ret = lvconvert_single(cmd, lvl->lv, &lp);
 
 error:
-	unlock_vg(cmd, vg_name);
+	unlock_vg(cmd, lp.vg_name);
 	return ret;
 }

tools/lvcreate.c

@@ -1,6 +1,6 @@
 /*
  * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved. 
- * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ * Copyright (C) 2004-2005 Red Hat, Inc. All rights reserved.
  *
  * This file is part of LVM2.
  *
@@ -291,6 +291,10 @@ static int _read_params(struct lvcreate_params *lp, struct cmd_context *cmd,
 		lp->mirrors = arg_uint_value(cmd, mirrors_ARG, 0) + 1;
 		if (lp->mirrors == 1)
 			log_print("Redundant mirrors argument: default is 0");
+		if (arg_sign_value(cmd, mirrors_ARG, 0) == SIGN_MINUS) {
+			log_error("Mirrors argument may not be negative");
+			return 0;
+		}
 	}
 
 	if (lp->snapshot) {
@@ -432,16 +436,15 @@ static int _lvcreate(struct cmd_context *cmd, struct lvcreate_params *lp)
 	const char *tag;
 	int consistent = 1;
 	struct alloc_handle *ah = NULL;
-	char *log_name, lv_name_buf[128];
+	char lv_name_buf[128];
 	const char *lv_name;
-	size_t len;
 
 	status |= lp->permission | VISIBLE_LV;
 
 	/* does VG exist? */
 	log_verbose("Finding volume group \"%s\"", lp->vg_name);
 
-	if (!(vg = vg_read(cmd, lp->vg_name, &consistent))) {
+	if (!(vg = vg_read(cmd, lp->vg_name, NULL, &consistent))) {
 		log_error("Volume group \"%s\" doesn't exist", lp->vg_name);
 		return 0;
 	}
@@ -584,71 +587,25 @@ static int _lvcreate(struct cmd_context *cmd, struct lvcreate_params *lp)
 	}
 
 	if (lp->mirrors > 1) {
+		/* FIXME Calculate how many extents needed for the log */
+
+		if (!(ah = allocate_extents(vg, NULL, lp->segtype, lp->stripes,
+					    lp->mirrors, 1, lp->extents,
+					    NULL, 0, 0, pvh, lp->alloc,
+					    NULL))) {
+			stack;
+			return 0;
+		}
+
 		lp->region_size = adjusted_mirror_region_size(vg->extent_size,
 							      lp->extents,
 							      lp->region_size);
 
-		/* FIXME Calculate how many extents needed for the log */
-
-		len = strlen(lv_name) + 32;
-        	if (!(log_name = alloca(len)) ||
-		    !(generate_log_name_format(vg, lv_name, log_name, len))) {
-                	log_error("log_name allocation failed. "
-	                          "Remove new LV and retry.");
-                	return 0;
-        	}
-
-		if (!(log_lv = lv_create_empty(vg->fid, log_name, NULL,
-				VISIBLE_LV | LVM_READ | LVM_WRITE,
-				lp->alloc, 0, vg))) {
-			stack;
+		if (!(log_lv = create_mirror_log(cmd, vg, ah, lp->alloc,
+						 lv_name))) {
+			log_error("Failed to create mirror log.");
 			return 0;
 		}
-
-		if (!(ah = allocate_extents(vg, NULL, lp->segtype, lp->stripes,
-					    lp->mirrors, 1, lp->extents,
-					    NULL, 0, 0, pvh, lp->alloc))) {
-			stack;
-			return 0;
-		}
-
-		if (!lv_add_log_segment(ah, log_lv)) {
-			stack;
-			goto error;
-		}
-
-		/* store mirror log on disk(s) */
-		if (!vg_write(vg)) {
-			stack;
-			goto error;
-		}
-
-		backup(vg);
-
-		if (!vg_commit(vg)) {
-			stack;
-			goto error;
-		}
-
-		if (!activate_lv(cmd, log_lv)) {
-			log_error("Aborting. Failed to activate mirror log. "
-				  "Remove new LVs and retry.");
-			goto error;
-		}
-
-		if (activation() && !zero_lv(cmd, log_lv)) {
-			log_error("Aborting. Failed to wipe mirror log. "
-				  "Remove new LV and retry.");
-			goto error;
-		}
-
-		if (!deactivate_lv(cmd, log_lv)) {
-			log_error("Aborting. Failed to deactivate mirror log. "
-				  "Remove new LV and retry.");
-			goto error;
-		}
-
-		log_lv->status &= ~VISIBLE_LV;
 	}
 
 	if (!(lv = lv_create_empty(vg->fid, lv_name ? lv_name : "lvol%d", NULL,
@@ -761,6 +718,7 @@ static int _lvcreate(struct cmd_context *cmd, struct lvcreate_params *lp)
 
 		if (!suspend_lv(cmd, org)) {
 			log_error("Failed to suspend origin %s", org->name);
+			vg_revert(vg);
 			return 0;
 		}
 

tools/lvdisplay.c

@@ -18,7 +18,7 @@
 static int _lvdisplay_single(struct cmd_context *cmd, struct logical_volume *lv,
 			     void *handle)
 {
-	if (!arg_count(cmd, all_ARG) && !(lv->status & VISIBLE_LV))
+	if (!arg_count(cmd, all_ARG) && !lv_is_visible(lv))
 		return ECMD_PROCESSED;
 
 	if (arg_count(cmd, colon_ARG))

tools/lvremove.c

@@ -83,7 +83,7 @@ static int lvremove_single(struct cmd_context *cmd, struct logical_volume *lv,
 	}
 
 	if (lv_is_cow(lv)) {
-		origin = find_cow(lv)->origin;
+		origin = origin_from_cow(lv);
 		log_verbose("Removing snapshot %s", lv->name);
 		if (!vg_remove_snapshot(lv)) {
 			stack;