update pofile

DEBUGLOG() message now print threadid rather than PID which is more useful.

VERSION

@@ -1 +1 @@
-2.00.24-cvs (2004-09-16)
+2.01.08-cvs (2005-03-22)

WHATS_NEW

@@ -1,3 +1,168 @@
+Version 2.01.08 - 22nd March 2005
+=================================
+  Add clustered attribute so vgchange can identify clustered VGs w/o locking.
+  Improve detection of external changes affecting internal cache.
+  Add 'already in device cache' debug message.
+  Add -a to pvdisplay -C.
+  Avoid rmdir opendir error messsages when dir was already removed.
+  Tighten signal handlers.
+  Avoid some compiler warnings.
+  Additional rename failure error message.
+  read/write may be macros.
+  clvmd: don't take out lvm thread lock at startup, it only protects jobs list.
+
+Version 2.01.07 - 8th March 2005
+================================
+  Cope with new devices appearing by rescanning /dev if a uuid can't be found.
+  Remove DESTDIR from LVM_SHARED_PATH.
+  clvmd fixes: make FDs close-on-exec
+               gulm unlocks VG & orphan locks at startup in case they are stale
+               gulm now unlocks VG & orphan locks if client dies.
+
+Version 2.01.06 - 1st March 2005
+================================
+  Suppress 'open failed' error messages during scanning.
+  Option to suppress warnings of file descriptors left open.
+  Fix default value of metadatacopies in documentation (2->1).
+  Fix clvmd-gulm locking.
+  ./configure --enable-debug now enables debugging code in clvmd.
+  Fix clvmd-gulm node up/down code so it actually works.
+  clvmd-gulm now releases locks when shut down.
+
+Version 2.01.05 - 18th February 2005
+====================================
+  Static binary invokes dynamic binary if appropriate.
+  Make clvmd config check a little more tolerant.
+  gulm clvmd can now cope with >1 message arriving in a TCP message.
+
+Version 2.01.04 - 9th February 2005
+===================================
+  Add fixed offset to imported pool minor numbers.
+  Update binary pathnames in clvmd_init_rhel4.
+  lvm2cmd.so should skip the check for open fds.
+  Remove unused -f from pvmove.
+  Gulm clvmd doesn't report "connection refused" errors.
+  clvmd does a basic config file sanity check at startup.
+  Fix potential thread shutdown race in clvmd.
+
+Version 2.01.03 - 1st February 2005
+===================================
+  More 64-bit display/report fixes.
+  More informative startup mesg if can't create /etc/lvm.
+  Fix snapshot device size bug (since 2.01.01).
+  clvmd announces startup and cluster connection in syslog.
+  Gulm clvmd doesn't hang trying to talk to a rebooted node.
+  Gulm clvmd doesn't print cman error on startup.
+
+Version 2.01.02 - 21st January 2005
+===================================
+  Update clvmd_init_rhel4: use lvm.static and don't load dlm.
+  Fix some size_t printing.
+  Fix 64 bit xlate consts.
+  Split out pool sptype_names to avoid unused const.
+  Always fail if random id generation fails.
+  Recognise gnbd devices.
+  Fix clvmd startup bug introduced in cman/gulm amalgamation.
+  Improve reporting of node-specific locking errors.
+
+Version 2.01.01 - 19th January 2005
+===================================
+  Fix clvmd lv_info_by_lvid open_count.
+  Store snapshot and origin sizes separately.
+  Update vgcreate man page.
+
+Version 2.01.00 - 17th January 2005
+===================================
+  Fix vgscan metadata auto-correction.
+  Only ask libdevmapper for open_count when we need it.
+  Adjust RHEL4 clvmd init script priority.
+  Enable building of CMAN & GULM versions of clvmd into a single binary
+
+Version 2.00.33 - 7th January 2005
+==================================
+  pvcreate wipes first 4 sectors unless given --zero n.
+  gulm clvmd now uses new ccsd key names.
+  gulm clvmd now doesn't ignore the first node in cluster.conf
+  Improve clvmd failure message if it's already running.
+  Allow user to kill clvmd during initialisation.
+  Fix off-by-one error in cluster_locking that could cause read hangs.
+
+Version 2.00.32 - 22nd December 2004
+====================================
+  Drop static/dl restriction for now.
+  Fix an error fprintf.
+  Fix vgdisplay -s. Breaks (undocumented) lvs/pvs/vgs -s instead for now.
+  Fix device reference counting on re-opens.
+  Ignore sysfs symlinks when DT_UNKNOWN.
+  Add clvmd init script for RHEL4.
+  Skip devices that are too small to be PVs.
+  Fix pvchange -x segfault with lvm2-format orphan.
+  Cope with empty msdos partition tables.
+  Add CONTRIBUTORS file.
+
+Version 2.00.31 - 12th December 2004
+====================================
+  Reopen RO file descriptors RW if necessary.
+
+Version 2.00.30 - 10th December 2004
+====================================
+  Additional device-handling debug messages.
+  Additional verbosity level -vvvv includes line numbers and backtraces.
+  Verbose messages now go to stderr not stdout.
+  Close any stray file descriptors before starting.
+  Refine partitionable checks for certain device types.
+  Allow devices/types to override built-ins.
+  Fix lvreduce man page .i->.I
+  Fix vgsplit man page title.
+  Fix clvmd man makefile.
+  Extend dev_open logging.
+  Make clvmd_fix_conf.sh UNDOable.
+
+Version 2.00.29 - 27th November 2004
+====================================
+  xlate compilation fix.
+
+Version 2.00.28 - 27th November 2004
+====================================
+  Fix partition table & md signature detection.
+  Minor configure/makefile tidy.
+  Export version.h from tools for clvmd.
+
+Version 2.00.27 - 24th November 2004
+====================================
+  Trap large memory allocation requests.
+  Fix to partition table detection code.
+  Improve filter debug mesgs.
+  Make clvmd_fix_conf.sh UNDOable
+
+Version 2.00.26 - 23rd November 2004
+====================================
+  Improve pool debugging stats.
+  Detect partition table signature.
+  pvcreate wipes md superblocks. (With --uuid or --restorefile it prompts.)
+  Separate out md superblock detection code.
+  Prevent snapshot origin resizing.
+  Improve a vgremove error message.
+  Update some man pages.
+  Allow y/n with -ae args (exclusive activation).
+  Fixes to lvcreate vgname parsing.
+  Fix dm_name string size calculation.
+  Improve clvmd error reporting during startup.
+  Make clvmd cope with large gaps in node numbers IDs.
+  Make clvmd initialisation cope better with debugging output.
+  Tidy clvmd socket callbacks so all work happens outside main loop.
+  clvmd -V now displays lvm version too.
+  Add optional gulm build for clvmd
+
+Version 2.00.25 - 29th September 2004
+=====================================
+  Fix return code from rm_link for vgmknodes.
+  Make clvmd LV hash table thread-safe.
+  Fix clvmd locking so it will lock out multiple users on the same node.
+  Fix clvmd VG locking to it can cope with multiple VG locks.
+  Remove spurious trailing dot in lvreduce man page.
+  Fix vgremove locking.
+
 Version 2.00.24 - 16th September 2004
 =====================================
   Fix pool_empty so it really does empty the memory pool.

WHATS_NEW_DM

@@ -1,5 +1,25 @@
-Version 1.00.20 -
+Version 1.01.01 -
 =============================
+  Add option to compile without ioctl for testing.
+  Fix DM_LIB_VERSION sed.
+
+Version 1.01.00 - 17 Jan 2005
+=============================
+  Add dm_task_no_open_count() to skip getting open_count.
+
+Version 1.00.21 - 7 Jan 2005
+============================
+  Fix /proc/devices parsing.
+
+Version 1.00.20 - 6 Jan 2005
+============================
+  Attempt to fix /dev/mapper/control transparently if it's wrong.
+  Configuration-time option for setting uid/gid/mode for /dev/mapper nodes.
+  Update kernel patches for 2.4.27/2.4.28-pre-4 (includes minor fixes).
+  Add --noheadings columns option for colon-separated dmsetup output.
+  Support device referencing by uuid or major/minor.
+  Warn if kernel data didn't fit in buffer.
+  Fix a printf.
 
 Version 1.00.19 - 3 July 2004
 =============================

configure.in

@@ -280,12 +280,18 @@ AC_MSG_RESULT($SELINUX)
 ################################################################################
 dnl -- Build cluster LVM daemon
 AC_MSG_CHECKING(whether to build cluster LVM daemon)
-AC_ARG_WITH(clvmd, [  --with-clvmd            Build cluster LVM Daemon],
-CLVMD=$withval, CLVMD=no)
+AC_ARG_WITH(clvmd,
+  [  --with-clvmd=TYPE       Build cluster LVM Daemon: cman/gulm/none/all
+                          [TYPE=none] ],
+  [ CLVMD="$withval" ],
+  [ CLVMD="none" ])
+if test x$CLVMD = xyes; then
+	CLVMD=all
+fi
 AC_MSG_RESULT($CLVMD)
 
 dnl -- If clvmd enabled without cluster locking, automagically include it
-if  test x$CLVMD = xyes && test x$CLUSTER = xnone; then
+if  test x$CLVMD != xnone && test x$CLUSTER = xnone; then
 	CLUSTER=internal
 fi
 
@@ -383,7 +389,7 @@ fi
 dnl -- Check for dlopen
 AC_CHECK_LIB(dl, dlopen, HAVE_LIBDL=yes, HAVE_LIBDL=no)
 
-if [[ "x$HAVE_LIBDL" = xyes -a "x$STATIC_LINK" = xno ]]; then
+if [[ "x$HAVE_LIBDL" = xyes ]]; then
 	CFLAGS="$CFLAGS -DHAVE_LIBDL"
 	LIBS="-ldl $LIBS"
 else
@@ -475,7 +481,7 @@ if test x$READLINE = xyes; then
 	AC_CHECK_HEADERS(readline/readline.h readline/history.h,,AC_MSG_ERROR(bailing out))
 fi
 
-if test x$CLVMD = xyes; then
+if test x$CLVMD != xnone; then
 	AC_CHECK_HEADERS(mntent.h netdb.h netinet/in.h pthread.h search.h sys/mount.h sys/socket.h sys/uio.h sys/un.h utmpx.h,,AC_MSG_ERROR(bailing out))
 	AC_CHECK_FUNCS(dup2 getmntent memmove select socket,,AC_MSG_ERROR(bailing out))
 	AC_FUNC_GETMNTENT

daemons/Makefile.in

@@ -15,7 +15,7 @@ srcdir = @srcdir@
 top_srcdir = @top_srcdir@
 VPATH = @srcdir@
 
-ifeq ("@CLVMD@", "yes")
+ifneq ("@CLVMD@", "none")
   SUBDIRS = clvmd
 endif
 

daemons/clvmd/Makefile.in

@@ -16,26 +16,53 @@ top_srcdir = @top_srcdir@
 VPATH = @srcdir@
 
 SOURCES = \
-	clvmd-cman.c     \
 	clvmd-command.c  \
 	clvmd.c          \
-	libclvm.c        \
 	lvm-functions.c  \
 	system-lv.c
 
+ifeq ("@CLVMD@", "gulm")
+	GULM = yes
+endif
+
+ifeq ("@CLVMD@", "cman")
+	CMAN = yes
+endif
+
+ifeq ("@CLVMD@", "all")
+	GULM = yes
+	CMAN = yes
+endif
+
+ifeq ("@DEBUG@", "yes")
+	CFLAGS += -DDEBUG
+endif
+
+ifeq ("$(GULM)", "yes")
+	SOURCES += clvmd-gulm.c tcp-comms.c
+	LMLIBS += -lccs -lgulm
+	CFLAGS += -DUSE_GULM
+endif
+
+ifeq ("$(CMAN)", "yes")
+	SOURCES += clvmd-cman.c
+	LMLIBS += -ldlm
+	CFLAGS += -DUSE_CMAN
+endif
+
 TARGETS = \
 	clvmd
 
 include $(top_srcdir)/make.tmpl
 
 CFLAGS += -D_REENTRANT -fno-strict-aliasing
-LIBS += -ldevmapper -ldlm -llvm -lpthread
+LIBS += -ldevmapper -llvm -lpthread
 
 INSTALL_TARGETS = \
 	install_clvmd
 
 clvmd: $(OBJECTS) $(top_srcdir)/lib/liblvm.a
-	$(CC) -o clvmd $(OBJECTS) $(LDFLAGS) $(LVMLIBS) $(LIBS)
+	$(CC) -o clvmd $(OBJECTS) $(LDFLAGS) $(LVMLIBS) $(LMLIBS) $(LIBS)
 
 .PHONY: install_clvmd
 

daemons/clvmd/clvm.h

@@ -43,6 +43,7 @@ struct clvm_header {
 /* Flags */
 #define CLVMD_FLAG_LOCAL        1	/* Only do this on the local node */
 #define CLVMD_FLAG_SYSTEMLV     2	/* Data in system LV under my node name */
+#define CLVMD_FLAG_NODEERRS     4       /* Reply has errors in node-specific portion */
 
 /* Name of the local socket to communicate between libclvm and clvmd */
 //static const char CLVMD_SOCKNAME[]="/var/run/clvmd";

daemons/clvmd/clvmd-cman.c

@@ -55,7 +55,6 @@ static int max_updown_nodes = 50;	/* Current size of the allocated array */
 static int *node_updown = NULL;
 static dlm_lshandle_t *lockspace;
 
-static void sigusr1_handler(int sig);
 static void count_clvmds_running(void);
 static void get_members(void);
 static int nodeid_from_csid(char *csid);
@@ -67,7 +66,7 @@ struct lock_wait {
 	struct dlm_lksb lksb;
 };
 
-int init_cluster()
+static int _init_cluster(void)
 {
 	struct sockaddr_cl saddr;
 	int port = CLUSTER_PORT_CLVMD;
@@ -75,9 +74,11 @@ int init_cluster()
 	/* Open the cluster communication socket */
 	cluster_sock = socket(AF_CLUSTER, SOCK_DGRAM, CLPROTO_CLIENT);
 	if (cluster_sock == -1) {
-		perror("Can't open cluster socket");
+		/* Don't print an error here because we could be just probing for CMAN */
 		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 */
@@ -87,7 +88,7 @@ int init_cluster()
 	if (bind
 	    (cluster_sock, (struct sockaddr *) &saddr,
 	     sizeof(struct sockaddr_cl))) {
-		log_error("Can't bind cluster socket: %m");
+		syslog(LOG_ERR, "Can't bind cluster socket: %m");
 		return -1;
 	}
 
@@ -98,25 +99,31 @@ int init_cluster()
 	/* Create a lockspace for LV & VG locks to live in */
 	lockspace = dlm_create_lockspace(LOCKSPACE_NAME, 0600);
 	if (!lockspace) {
-		log_error("Unable to create lockspace for CLVM\n");
+		syslog(LOG_ERR, "Unable to create lockspace for CLVM: %m");
 		return -1;
 	}
 	dlm_ls_pthread_init(lockspace);
+
 	return 0;
 }
 
-int get_main_cluster_fd()
+static void _cluster_init_completed(void)
+{
+	clvmd_cluster_init_completed();
+}
+
+static int _get_main_cluster_fd()
 {
 	return cluster_sock;
 }
 
-int get_num_nodes()
+static int _get_num_nodes()
 {
 	return num_nodes;
 }
 
 /* send_message with the fd check removed */
-int cluster_send_message(void *buf, int msglen, char *csid, const char *errtext)
+static int _cluster_send_message(void *buf, int msglen, char *csid, const char *errtext)
 {
 	struct iovec iov[2];
 	struct msghdr msg;
@@ -136,7 +143,7 @@ int cluster_send_message(void *buf, int msglen, char *csid, const char *errtext)
 	if (csid) {
 		msg.msg_name = &saddr;
 		msg.msg_namelen = sizeof(saddr);
-		memcpy(&saddr.scl_nodeid, csid, MAX_CSID_LEN);
+		memcpy(&saddr.scl_nodeid, csid, CMAN_MAX_CSID_LEN);
 	} else {		/* Cluster broadcast */
 
 		msg.msg_name = NULL;
@@ -152,26 +159,26 @@ int cluster_send_message(void *buf, int msglen, char *csid, const char *errtext)
 	return len;
 }
 
-void get_our_csid(char *csid)
+static void _get_our_csid(char *csid)
 {
 	int i;
-	memset(csid, 0, MAX_CSID_LEN);
+	memset(csid, 0, CMAN_MAX_CSID_LEN);
 
 	for (i = 0; i < num_nodes; i++) {
 		if (nodes[i].us)
-			memcpy(csid, &nodes[i].node_id, MAX_CSID_LEN);
+			memcpy(csid, &nodes[i].node_id, CMAN_MAX_CSID_LEN);
 	}
 }
 
 /* Call a callback routine for each node that known (down mean not running a clvmd) */
-int cluster_do_node_callback(struct local_client *client,
+static int _cluster_do_node_callback(struct local_client *client,
 			     void (*callback) (struct local_client *, char *,
 					       int))
 {
 	int i;
 	int somedown = 0;
 
-	for (i = 0; i < get_num_nodes(); i++) {
+	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])
 			somedown = -1;
@@ -203,7 +210,7 @@ static void process_oob_msg(char *buf, int len, int nodeid)
 	}
 }
 
-int cluster_fd_callback(struct local_client *fd, char *buf, int len, char *csid,
+static int _cluster_fd_callback(struct local_client *client, char *buf, int len, char *csid,
 			struct local_client **new_client)
 {
 	struct iovec iov[2];
@@ -247,34 +254,39 @@ int cluster_fd_callback(struct local_client *fd, char *buf, int len, char *csid,
 		len = -1;
 		errno = EAGAIN;
 	}
-	memcpy(csid, &saddr.scl_nodeid, sizeof(saddr.scl_nodeid));
+	else {
+		memcpy(csid, &saddr.scl_nodeid, sizeof(saddr.scl_nodeid));
+		/* Send it back to clvmd */
+		process_message(client, buf, len, csid);
+	}
 	return len;
 }
 
-void add_up_node(char *csid)
+static void _add_up_node(char *csid)
 {
 	/* It's up ! */
 	int nodeid = nodeid_from_csid(csid);
 
 	if (nodeid >= max_updown_nodes) {
-		int *new_updown = realloc(node_updown, max_updown_nodes + 10);
+	        int new_size = nodeid + 10;
+		int *new_updown = realloc(node_updown, new_size);
 
 		if (new_updown) {
 			node_updown = new_updown;
-			max_updown_nodes += 10;
+			max_updown_nodes = new_size;
 			DEBUGLOG("realloced more space for nodes. now %d\n",
 				 max_updown_nodes);
 		} else {
 			log_error
-			    ("Realloc failed. Node status for clvmd will be wrong\n");
-			return;
+			    ("Realloc failed. Node status for clvmd will be wrong. quitting\n");
+			exit(999);
 		}
 	}
 	node_updown[nodeid] = 1;
 	DEBUGLOG("Added new node %d to updown list\n", nodeid);
 }
 
-void cluster_closedown()
+static void _cluster_closedown()
 {
 	unlock_all();
 	dlm_release_lockspace(LOCKSPACE_NAME, lockspace, 1);
@@ -303,7 +315,7 @@ static int is_listening(int nodeid)
 
 /* Populate the list of CLVMDs running.
    called only at startup time */
-void count_clvmds_running(void)
+static void count_clvmds_running(void)
 {
 	int i;
 
@@ -319,7 +331,7 @@ static void get_members()
 
 	num_nodes = ioctl(cluster_sock, SIOCCLUSTER_GETMEMBERS, 0);
 	if (num_nodes == -1) {
-		perror("get nodes");
+		log_error("Unable to get node count");
 	} else {
 	        /* Not enough room for new nodes list ? */
 	        if (num_nodes > count_nodes && nodes) {
@@ -331,16 +343,16 @@ static void get_members()
 		        count_nodes = num_nodes + 10; /* Overallocate a little */
 		        nodes = malloc(count_nodes * sizeof(struct cl_cluster_node));
 			if (!nodes) {
-			        perror("Unable to allocate nodes array\n");
+			        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) {
-		        perror("get node details");
+		        log_error("Unable to get node details");
 			exit(6);
 		}
 
@@ -362,13 +374,13 @@ static void get_members()
 }
 
 /* Convert a node name to a CSID */
-int csid_from_name(char *csid, char *name)
+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, MAX_CSID_LEN);
+			memcpy(csid, &nodes[i].node_id, CMAN_MAX_CSID_LEN);
 			return 0;
 		}
 	}
@@ -376,12 +388,12 @@ int csid_from_name(char *csid, char *name)
 }
 
 /* Convert a CSID to a node name */
-int name_from_csid(char *csid, char *name)
+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, MAX_CSID_LEN) == 0) {
+		if (memcmp(csid, &nodes[i].node_id, CMAN_MAX_CSID_LEN) == 0) {
 			strcpy(name, nodes[i].name);
 			return 0;
 		}
@@ -392,7 +404,7 @@ int name_from_csid(char *csid, char *name)
 }
 
 /* Convert a node ID to a node name */
-int name_from_nodeid(int nodeid, char *name)
+static int name_from_nodeid(int nodeid, char *name)
 {
 	int i;
 
@@ -412,12 +424,12 @@ static int nodeid_from_csid(char *csid)
 {
         int nodeid;
 
-	memcpy(&nodeid, csid, MAX_CSID_LEN);
+	memcpy(&nodeid, csid, CMAN_MAX_CSID_LEN);
 
 	return nodeid;
 }
 
-int is_quorate()
+static int _is_quorate()
 {
 	return ioctl(cluster_sock, SIOCCLUSTER_ISQUORATE, 0);
 }
@@ -431,7 +443,7 @@ static void sync_ast_routine(void *arg)
 	pthread_mutex_unlock(&lwait->mutex);
 }
 
-int sync_lock(const char *resource, int mode, int flags, int *lockid)
+static int _sync_lock(const char *resource, int mode, int flags, int *lockid)
 {
 	int status;
 	struct lock_wait lwait;
@@ -441,6 +453,7 @@ int sync_lock(const char *resource, int mode, int flags, int *lockid)
 		return -1;
 	}
 
+	DEBUGLOG("sync_lock: '%s' mode:%d flags=%d\n", resource,mode,flags);
 	/* Conversions need the lockid in the LKSB */
 	if (flags & LKF_CONVERT)
 		lwait.lksb.sb_lkid = *lockid;
@@ -466,17 +479,20 @@ int sync_lock(const char *resource, int mode, int flags, int *lockid)
 	*lockid = lwait.lksb.sb_lkid;
 
 	errno = lwait.lksb.sb_status;
+	DEBUGLOG("sync_lock: returning lkid %x\n", *lockid);
 	if (lwait.lksb.sb_status)
 		return -1;
 	else
 		return 0;
 }
 
-int sync_unlock(const char *resource /* UNUSED */, int lockid)
+static int _sync_unlock(const char *resource /* UNUSED */, int lockid)
 {
 	int status;
 	struct lock_wait lwait;
 
+	DEBUGLOG("sync_unlock: '%s' lkid:%x\n", resource, lockid);
+
 	pthread_cond_init(&lwait.cond, NULL);
 	pthread_mutex_init(&lwait.mutex, NULL);
 	pthread_mutex_lock(&lwait.mutex);
@@ -497,3 +513,28 @@ int sync_unlock(const char *resource /* UNUSED */, int lockid)
 		return 0;
 
 }
+
+static struct cluster_ops _cluster_cman_ops = {
+	.cluster_init_completed   = _cluster_init_completed,
+	.cluster_send_message     = _cluster_send_message,
+	.name_from_csid           = _name_from_csid,
+	.csid_from_name           = _csid_from_name,
+	.get_num_nodes            = _get_num_nodes,
+	.cluster_fd_callback      = _cluster_fd_callback,
+	.get_main_cluster_fd      = _get_main_cluster_fd,
+	.cluster_do_node_callback = _cluster_do_node_callback,
+	.is_quorate               = _is_quorate,
+	.get_our_csid             = _get_our_csid,
+	.add_up_node              = _add_up_node,
+	.cluster_closedown        = _cluster_closedown,
+	.sync_lock                = _sync_lock,
+	.sync_unlock              = _sync_unlock,
+};
+
+struct cluster_ops *init_cman_cluster(void)
+{
+	if (!_init_cluster())
+		return &_cluster_cman_ops;
+	else
+		return NULL;
+}

daemons/clvmd/clvmd-command.c

@@ -66,6 +66,7 @@
 #include <errno.h>
 
 #include "list.h"
+#include "hash.h"
 #include "locking.h"
 #include "log.h"
 #include "lvm-functions.h"
@@ -135,6 +136,61 @@ int do_command(struct local_client *client, struct clvm_header *msg, int msglen,
 
 }
 
+static int lock_vg(struct local_client *client)
+{
+    struct hash_table *lock_hash;
+    struct clvm_header *header =
+	(struct clvm_header *) client->bits.localsock.cmd;
+    unsigned char lock_cmd;
+    unsigned char lock_flags;
+    char *args = header->node + strlen(header->node) + 1;
+    int lkid;
+    int status = 0;
+    char *lockname;
+
+    /* Keep a track of VG locks in our own hash table. In current
+       practice there should only ever be more than two VGs locked
+       if a user tries to merge lots of them at once */
+    if (client->bits.localsock.private) {
+	lock_hash = (struct hash_table *)client->bits.localsock.private;
+    }
+    else {
+	lock_hash = hash_create(3);
+	if (!lock_hash)
+	    return ENOMEM;
+	client->bits.localsock.private = (void *)lock_hash;
+    }
+
+    lock_cmd = args[0];
+    lock_flags = args[1];
+    lockname = &args[2];
+    DEBUGLOG("doing PRE command LOCK_VG '%s' at %x (client=%p)\n", lockname, lock_cmd, client);
+
+    if (lock_cmd == LCK_UNLOCK) {
+
+	lkid = (int)(long)hash_lookup(lock_hash, lockname);
+	if (lkid == 0)
+	    return EINVAL;
+
+	status = sync_unlock(lockname, lkid);
+	if (status)
+	    status = errno;
+	else
+	    hash_remove(lock_hash, lockname);
+    }
+    else {
+
+	status = sync_lock(lockname, (int)lock_cmd, (int)lock_flags, &lkid);
+	if (status)
+	    status = errno;
+	else
+	    hash_insert(lock_hash, lockname, (void *)lkid);
+    }
+
+    return status;
+}
+
+
 /* Pre-command is a good place to get locks that are needed only for the duration
    of the commands around the cluster (don't forget to free them in post-command),
    and to sanity check the command arguments */
@@ -156,20 +212,7 @@ int do_pre_command(struct local_client *client)
 		break;
 
 	case CLVMD_CMD_LOCK_VG:
-		lock_cmd = args[0];
-		lock_flags = args[1];
-		lockname = &args[2];
-		DEBUGLOG("doing PRE command LOCK_VG %s at %x\n", lockname,
-			 lock_cmd);
-		if (lock_cmd == LCK_UNLOCK) {
-			hold_unlock(lockname);
-		} else {
-			status =
-			    hold_lock(lockname, (int) lock_cmd,
-				      (int) lock_flags);
-			if (status)
-				status = errno;
-		}
+       	        status = lock_vg(client);
 		break;
 
 	case CLVMD_CMD_LOCK_LV:
@@ -202,6 +245,7 @@ int do_post_command(struct local_client *client)
 	case CLVMD_CMD_TEST:
 		status =
 		    sync_unlock("CLVMD_TEST", (int) (long) client->bits.localsock.private);
+		client->bits.localsock.private = 0;
 		break;
 
 	case CLVMD_CMD_LOCK_VG:
@@ -217,3 +261,26 @@ int do_post_command(struct local_client *client)
 	}
 	return status;
 }
+
+
+/* Called when the client is about to be deleted */
+void cmd_client_cleanup(struct local_client *client)
+{
+    if (client->bits.localsock.private) {
+
+	struct hash_node *v;
+	struct hash_table *lock_hash =
+	    (struct hash_table *)client->bits.localsock.private;
+
+	hash_iterate(v, lock_hash) {
+		int lkid = (int)(long)hash_get_data(lock_hash, v);
+		char *lockname = hash_get_key(lock_hash, v);
+
+		DEBUGLOG("cleanup: Unlocking lock %s %x\n", lockname, lkid);
+		sync_unlock(lockname, lkid);
+	}
+
+	hash_destroy(lock_hash);
+	client->bits.localsock.private = 0;
+    }
+}

daemons/clvmd/clvmd-comms.h

@@ -22,33 +22,47 @@
 
 struct local_client;
 
-extern int cluster_send_message(void *buf, int msglen, char *csid,
+struct cluster_ops {
+	void (*cluster_init_completed) (void);
+
+	int (*cluster_send_message) (void *buf, int msglen, char *csid,
 				const char *errtext);
-extern int name_from_csid(char *csid, char *name);
-extern int csid_from_name(char *csid, char *name);
-extern int get_num_nodes(void);
-extern int cluster_fd_callback(struct local_client *fd, char *buf, int len,
+	int (*name_from_csid) (char *csid, char *name);
+	int (*csid_from_name) (char *csid, char *name);
+	int (*get_num_nodes) (void);
+	int (*cluster_fd_callback) (struct local_client *fd, char *buf, int len,
 			       char *csid, struct local_client **new_client);
-extern int init_cluster(void);
-extern int get_main_cluster_fd(void);	/* gets accept FD or cman cluster socket */
-extern int cluster_do_node_callback(struct local_client *client,
+	int (*get_main_cluster_fd) (void);	/* gets accept FD or cman cluster socket */
+	int (*cluster_do_node_callback) (struct local_client *client,
 				    void (*callback) (struct local_client *,
 						      char *csid, int node_up));
-extern int is_quorate(void);
+	int (*is_quorate) (void);
 
-extern void get_our_csid(char *csid);
-extern void add_up_node(char *csid);
-extern void cluster_closedown(void);
+	void (*get_our_csid) (char *csid);
+	void (*add_up_node) (char *csid);
+	void (*cluster_closedown) (void);
 
-extern int sync_lock(const char *resource, int mode, int flags, int *lockid);
-extern int sync_unlock(const char *resource, int lockid);
+	int (*sync_lock) (const char *resource, int mode, int flags, int *lockid);
+	int (*sync_unlock) (const char *resource, int lockid);
+
+};
 
 #ifdef USE_GULM
-#include "tcp-comms.h"
-#else
-/* cman */
-#include "cnxman-socket.h"
-#define MAX_CSID_LEN 4
+#  include "tcp-comms.h"
+struct cluster_ops *init_gulm_cluster(void);
+#define MAX_CSID_LEN 			GULM_MAX_CSID_LEN
+#define MAX_CLUSTER_MEMBER_NAME_LEN	GULM_MAX_CLUSTER_MEMBER_NAME_LEN
+#endif
+
+#ifdef USE_CMAN
+#  include "cnxman-socket.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
+struct cluster_ops *init_cman_cluster(void);
 #endif
 
 

daemons/clvmd/clvmd-gulm.c

@@ -2,6 +2,7 @@
 *******************************************************************************
 **
 **  Copyright (C) Sistina Software, Inc.  2002-2003  All rights reserved.
+**  Copyright (C) 2004-2005 Red Hat, Inc. All rights reserved.
 **
 *******************************************************************************
 ******************************************************************************/
@@ -46,6 +47,7 @@
 #include "log.h"
 #include "clvm.h"
 #include "clvmd-comms.h"
+#include "lvm-functions.h"
 #include "clvmd.h"
 #include "hash.h"
 #include "clvmd-gulm.h"
@@ -65,6 +67,7 @@ static uint8_t current_corestate;
 static int num_nodes;
 
 static char *cluster_name;
+static int in_shutdown = 0;
 
 static pthread_mutex_t lock_start_mutex;
 static volatile int lock_start_flag;
@@ -72,7 +75,7 @@ static volatile int lock_start_flag;
 struct node_info
 {
     enum {NODE_UNKNOWN, NODE_DOWN, NODE_UP, NODE_CLVMD} state;
-    char name[MAX_CLUSTER_MEMBER_NAME_LEN];
+    char name[GULM_MAX_CLUSTER_MEMBER_NAME_LEN];
 };
 
 struct lock_wait
@@ -88,6 +91,8 @@ static int read_from_core_sock(struct local_client *client, char *buf, int len,
 static int read_from_lock_sock(struct local_client *client, char *buf, int len, char *csid,
 			       struct local_client **new_client);
 static int get_all_cluster_nodes(void);
+static int _csid_from_name(char *csid, char *name);
+static void _cluster_closedown(void);
 
 /* In tcp-comms.c */
 extern struct hash_table *sock_hash;
@@ -112,6 +117,9 @@ static int add_internal_client(int fd, fd_callback_t callback)
     client->callback = callback;
     add_client(client);
 
+    /* Set Close-on-exec */
+    fcntl(fd, F_SETFD, 1);
+
     return 0;
 }
 
@@ -123,7 +131,7 @@ static lg_lockspace_callbacks_t lock_callbacks;
 static void badsig_handler(int sig)
 {
     DEBUGLOG("got sig %d\n", sig);
-    cluster_closedown();
+    _cluster_closedown();
     exit(0);
 }
 
@@ -135,15 +143,34 @@ static void sighup_handler(int sig)
     get_all_cluster_nodes();
 }
 
-int init_cluster()
+static int _init_cluster(void)
 {
     int status;
     int ccs_h;
+    int port = 0;
+    char *portstr;
 
     /* Get cluster name from CCS */
-    /* TODO: is this right? */
-    ccs_h = ccs_connect();
+    ccs_h = ccs_force_connect(NULL, 0);
+    if (ccs_h < 0)
+    {
+	syslog(LOG_ERR, "Cannot login in to CCSD server\n");
+	return -1;
+    }
+
     ccs_get(ccs_h, "//cluster/@name", &cluster_name);
+    DEBUGLOG("got cluster name %s\n", cluster_name);
+
+    if (!ccs_get(ccs_h, "//cluster/clvm/@port", &portstr))
+    {
+	port = atoi(portstr);
+	free(portstr);
+	DEBUGLOG("got port number %d\n", port);
+
+	if (port <= 0 && port >= 65536)
+	    port = 0;
+    }
+
     ccs_disconnect(ccs_h);
 
     /* Block locking until we are logged in */
@@ -155,7 +182,8 @@ int init_cluster()
     lock_hash = hash_create(10);
 
     /* Get all nodes from CCS */
-    get_all_cluster_nodes();
+    if (get_all_cluster_nodes())
+	return -1;
 
     /* Initialise GULM library */
     status = lg_initialize(&gulm_if, cluster_name, "clvmd");
@@ -174,7 +202,7 @@ int init_cluster()
     }
 
     /* Initialise the inter-node comms */
-    status = init_comms();
+    status = init_comms(port);
     if (status)
 	return status;
 
@@ -202,7 +230,7 @@ int init_cluster()
 	exit(status);
     }
 
-    /* Request a list of nodes, we can;t really do anything until
+    /* Request a list of nodes, we can't really do anything until
        this comes back */
     status = lg_core_nodelist(gulm_if);
     if (status)
@@ -221,12 +249,13 @@ int init_cluster()
     return 0;
 }
 
-void cluster_closedown()
+static void _cluster_closedown(void)
 {
     DEBUGLOG("cluster_closedown\n");
+    in_shutdown = 1;
+    unlock_all();
     lg_lock_logout(gulm_if);
     lg_core_logout(gulm_if);
-    lg_core_shutdown(gulm_if);
     lg_release(gulm_if);
 }
 
@@ -237,6 +266,7 @@ static void drop_expired_locks(char *nodename)
     struct utsname nodeinfo;
     uint8_t mask[GIO_KEY_SIZE];
 
+    DEBUGLOG("Dropping expired locks for %s\n", nodename?nodename:"(null)");
     memset(mask, 0xff, GIO_KEY_SIZE);
 
     if (!nodename)
@@ -316,11 +346,11 @@ static void set_node_state(struct node_info *ninfo, char *csid, uint8_t nodestat
 	     ninfo->name, ninfo->state, num_nodes);
 }
 
-static struct node_info *add_or_set_node(char *name, uint32_t ip, uint8_t state)
+static struct node_info *add_or_set_node(char *name, struct in6_addr *ip, uint8_t state)
 {
     struct node_info *ninfo;
 
-    ninfo = hash_lookup_binary(node_hash, (char *)&ip, MAX_CSID_LEN);
+    ninfo = hash_lookup_binary(node_hash, (char *)ip, GULM_MAX_CSID_LEN);
     if (!ninfo)
     {
 	/* If we can't find that node then re-read the config file in case it
@@ -329,7 +359,7 @@ static struct node_info *add_or_set_node(char *name, uint32_t ip, uint8_t state)
 	get_all_cluster_nodes();
 
 	/* Now try again */
-	ninfo = hash_lookup_binary(node_hash, (char *)&ip, MAX_CSID_LEN);
+	ninfo = hash_lookup_binary(node_hash, (char *)ip, GULM_MAX_CSID_LEN);
 	if (!ninfo)
 	{
 	    DEBUGLOG("Ignoring node %s, not part of the SAN cluster\n", name);
@@ -337,12 +367,17 @@ static struct node_info *add_or_set_node(char *name, uint32_t ip, uint8_t state)
 	}
     }
 
-    set_node_state(ninfo, (char *)&ip, state);
+    set_node_state(ninfo, (char *)ip, state);
 
     return ninfo;
 }
 
-static int core_nodelist(void *misc, lglcb_t type, char *name, uint32_t ip, uint8_t state)
+static void _get_our_csid(char *csid)
+{
+	get_our_gulm_csid(csid);
+}
+
+static int core_nodelist(void *misc, lglcb_t type, char *name, struct in6_addr *ip, uint8_t state)
 {
     DEBUGLOG("CORE nodelist\n");
 
@@ -354,7 +389,7 @@ static int core_nodelist(void *misc, lglcb_t type, char *name, uint32_t ip, uint
     {
 	if (type == lglcb_item)
 	{
-	    DEBUGLOG("Got nodelist, item: %s, %#x, %#x\n", name, ip, state);
+	    DEBUGLOG("Got nodelist, item: %s, %#x\n", name, state);
 
 	    add_or_set_node(name, ip, state);
 	}
@@ -362,14 +397,14 @@ static int core_nodelist(void *misc, lglcb_t type, char *name, uint32_t ip, uint
 	{
 	    if (type == lglcb_stop)
 	    {
-		char ourcsid[MAX_CSID_LEN];
+		char ourcsid[GULM_MAX_CSID_LEN];
 
 		DEBUGLOG("Got Nodelist, stop\n");
 		clvmd_cluster_init_completed();
 
 		/* Mark ourself as up */
-		get_our_csid(ourcsid);
-		add_up_node(ourcsid);
+		_get_our_csid(ourcsid);
+		gulm_add_up_node(ourcsid);
 	    }
 	    else
 	    {
@@ -381,24 +416,24 @@ static int core_nodelist(void *misc, lglcb_t type, char *name, uint32_t ip, uint
     return 0;
 }
 
-static int core_statechange(void *misc, uint8_t corestate, uint32_t masterip, char *mastername)
+static int core_statechange(void *misc, uint8_t corestate, uint8_t quorate, struct in6_addr *masterip, char *mastername)
 {
-    DEBUGLOG("CORE Got statechange  corestate:%#x masterip:%#x mastername:%s\n",
-	     corestate, masterip, mastername);
+    DEBUGLOG("CORE Got statechange  corestate:%#x mastername:%s\n",
+	     corestate, mastername);
 
     current_corestate = corestate;
     return 0;
 }
 
-static int core_nodechange(void *misc, char *nodename, uint32_t nodeip, uint8_t nodestate)
+static int core_nodechange(void *misc, char *nodename, struct in6_addr *nodeip, uint8_t nodestate)
 {
     struct node_info *ninfo;
 
-    DEBUGLOG("CORE node change, name=%s, ip=%x, state = %d\n", nodename, nodeip, nodestate);
+    DEBUGLOG("CORE node change, name=%s, state = %d\n", nodename, nodestate);
 
     /* If we don't get nodeip here, try a lookup by name */
     if (!nodeip)
-	csid_from_name((char *)&nodeip, nodename);
+	_csid_from_name((char *)nodeip, nodename);
     if (!nodeip)
 	return 0;
 
@@ -439,17 +474,23 @@ static int lock_login_reply(void *misc, uint32_t error, uint8_t which)
 	lock_start_flag = 0;
 	pthread_mutex_unlock(&lock_start_mutex);
     }
-
+    
     return 0;
 }
 
-static int lock_lock_state(void *misc, uint8_t *key, uint16_t keylen, uint8_t state, uint32_t flags, uint32_t error,
+static int lock_lock_state(void *misc, uint8_t *key, uint16_t keylen,
+			   uint64_t subid, uint64_t start, uint64_t stop,
+			   uint8_t state, uint32_t flags, uint32_t error,
 			   uint8_t *LVB, uint16_t LVBlen)
 {
     struct lock_wait *lwait;
 
     DEBUGLOG("LOCK lock state: %s, error = %d\n", key, error);
 
+    /* No waiting process to wake up when we are shutting down */
+    if (in_shutdown)
+	    return 0;
+
     lwait = hash_lookup(lock_hash, key);
     if (!lwait)
     {
@@ -519,19 +560,18 @@ int get_next_node_csid(void **context, char *csid)
 	return 0;
     }
 
-    memcpy(csid, hash_get_key(node_hash, *context), MAX_CSID_LEN);
+    memcpy(csid, hash_get_key(node_hash, *context), GULM_MAX_CSID_LEN);
     return 1;
 }
 
-int name_from_csid(char *csid, char *name)
+int gulm_name_from_csid(char *csid, char *name)
 {
     struct node_info *ninfo;
 
-    ninfo = hash_lookup_binary(node_hash, csid, MAX_CSID_LEN);
+    ninfo = hash_lookup_binary(node_hash, csid, GULM_MAX_CSID_LEN);
     if (!ninfo)
     {
-	sprintf(name, "UNKNOWN [%d.%d.%d.%d]",
-		csid[0], csid[1], csid[2], csid[3]);
+        sprintf(name, "UNKNOWN %s", print_csid(csid));
 	return -1;
     }
 
@@ -540,7 +580,7 @@ int name_from_csid(char *csid, char *name)
 }
 
 
-int csid_from_name(char *csid, char *name)
+static int _csid_from_name(char *csid, char *name)
 {
     struct hash_node *hn;
     struct node_info *ninfo;
@@ -550,28 +590,31 @@ int csid_from_name(char *csid, char *name)
 	ninfo = hash_get_data(node_hash, hn);
 	if (strcmp(ninfo->name, name) == 0)
 	{
-	    memcpy(csid, hash_get_key(node_hash, hn), MAX_CSID_LEN);
+	    memcpy(csid, hash_get_key(node_hash, hn), GULM_MAX_CSID_LEN);
 	    return 0;
 	}
     }
     return -1;
 }
 
-int get_num_nodes()
+static int _get_num_nodes()
 {
     DEBUGLOG("num_nodes = %d\n", num_nodes);
     return num_nodes;
 }
 
 /* Node is now known to be running a clvmd */
-void add_up_node(char *csid)
+void gulm_add_up_node(char *csid)
 {
     struct node_info *ninfo;
 
-    ninfo = hash_lookup_binary(node_hash, csid, MAX_CSID_LEN);
-    if (!ninfo)
+    ninfo = hash_lookup_binary(node_hash, csid, GULM_MAX_CSID_LEN);
+    if (!ninfo) {
+	    DEBUGLOG("gulm_add_up_node no node_hash entry for csid %s\n", print_csid(csid));
 	return;
+    }
 
+    DEBUGLOG("gulm_add_up_node %s\n", ninfo->name);
     ninfo->state = NODE_CLVMD;
     return;
 
@@ -581,7 +624,7 @@ void add_down_node(char *csid)
 {
     struct node_info *ninfo;
 
-    ninfo = hash_lookup_binary(node_hash, csid, MAX_CSID_LEN);
+    ninfo = hash_lookup_binary(node_hash, csid, GULM_MAX_CSID_LEN);
     if (!ninfo)
 	return;
 
@@ -589,30 +632,42 @@ void add_down_node(char *csid)
        running clvmd - gulm may set it DOWN quite soon */
     if (ninfo->state == NODE_CLVMD)
 	ninfo->state = NODE_UP;
+    drop_expired_locks(ninfo->name);
     return;
 
 }
 
 /* Call a callback for each node, so the caller knows whether it's up or down */
-int cluster_do_node_callback(struct local_client *master_client,
-			     void (*callback)(struct local_client *, char *csid, int node_up))
+static int _cluster_do_node_callback(struct local_client *master_client,
+				     void (*callback)(struct local_client *, char *csid, int node_up))
 {
     struct hash_node *hn;
     struct node_info *ninfo;
 
     hash_iterate(hn, node_hash)
     {
-	char csid[MAX_CSID_LEN];
+	char csid[GULM_MAX_CSID_LEN];
 	struct local_client *client;
 
 	ninfo = hash_get_data(node_hash, hn);
-	memcpy(csid, hash_get_key(node_hash, hn), MAX_CSID_LEN);
+	memcpy(csid, hash_get_key(node_hash, hn), GULM_MAX_CSID_LEN);
 
 	DEBUGLOG("down_callback. node %s, state = %d\n", ninfo->name, ninfo->state);
 
-	client = hash_lookup_binary(sock_hash, csid, MAX_CSID_LEN);
-	if (client)
-	    callback(master_client, csid, ninfo->state == NODE_CLVMD);
+	client = hash_lookup_binary(sock_hash, csid, GULM_MAX_CSID_LEN);
+	if (!client)
+	{
+	    /* If it's up but not connected, try to make contact */
+	    if (ninfo->state == NODE_UP)
+		    gulm_connect_csid(csid, &client);
+
+	    client = hash_lookup_binary(sock_hash, csid, GULM_MAX_CSID_LEN);
+
+	}
+	if (ninfo->state != NODE_DOWN)
+		callback(master_client, csid, ninfo->state == NODE_CLVMD);
+
+
     }
     return 0;
 }
@@ -623,15 +678,13 @@ static int gulm_to_errno(int gulm_ret)
     switch (gulm_ret)
     {
     case lg_err_TryFailed:
-	errno = EAGAIN;
-	break;
-
     case lg_err_AlreadyPend:
-	errno = EBUSY;
+	    errno = EAGAIN;
+	    break;
 
 	/* More?? */
     default:
-	errno = EINVAL;
+	    errno = EINVAL;
     }
 
     return gulm_ret ? -1 : 0;
@@ -661,6 +714,7 @@ static int _lock_resource(char *resource, int mode, int flags, int *lockid)
     DEBUGLOG("lock_resource '%s', flags=%d, mode=%d\n", resource, flags, mode);
 
     status = lg_lock_state_req(gulm_if, resource, strlen(resource)+1,
+			       0, 0, 0,
 			       mode, flags, NULL, 0);
     if (status)
     {
@@ -692,6 +746,7 @@ static int _unlock_resource(char *resource, int lockid)
 
     DEBUGLOG("unlock_resource %s\n", resource);
     status = lg_lock_state_req(gulm_if, resource, strlen(resource)+1,
+			       0, 0, 0,
 			       lg_lock_state_Unlock, 0, NULL, 0);
 
     if (status)
@@ -700,6 +755,11 @@ static int _unlock_resource(char *resource, int lockid)
 	return status;
     }
 
+    /* When we are shutting down, don't wait for unlocks
+       to be acknowledged, just do it. */
+    if (in_shutdown)
+	    return status;
+
     /* Wait for it to complete */
 
     pthread_cond_wait(&lwait.cond, &lwait.mutex);
@@ -720,7 +780,7 @@ static int _unlock_resource(char *resource, int lockid)
    To aid unlocking, we store the lock mode in the lockid (as GULM
    doesn't use this).
 */
-int sync_lock(const char *resource, int mode, int flags, int *lockid)
+static int _sync_lock(const char *resource, int mode, int flags, int *lockid)
 {
     int status;
     char lock1[strlen(resource)+3];
@@ -736,7 +796,7 @@ int sync_lock(const char *resource, int mode, int flags, int *lockid)
 	if (status)
 	    goto out;
 
-	/* If we can't get this lock then bail out */
+	/* If we can't get this lock too then bail out */
 	status = _lock_resource(lock2, lg_lock_state_Exclusive, LCK_NONBLOCK, lockid);
         if (status == lg_err_TryFailed)
         {
@@ -748,10 +808,16 @@ int sync_lock(const char *resource, int mode, int flags, int *lockid)
 
     case LCK_READ:
 	status = _lock_resource(lock1, lg_lock_state_Shared, flags, lockid);
+	if (status)
+		goto out;
+	status = _unlock_resource(lock2, *lockid);
 	break;
 
     case LCK_WRITE:
 	status = _lock_resource(lock2, lg_lock_state_Exclusive, flags, lockid);
+	if (status)
+		goto out;
+	status = _unlock_resource(lock1, *lockid);
 	break;
 
     default:
@@ -764,7 +830,7 @@ int sync_lock(const char *resource, int mode, int flags, int *lockid)
     return status;
 }
 
-int sync_unlock(const char *resource, int lockid)
+static int _sync_unlock(const char *resource, int lockid)
 {
     int status = 0;
     char lock1[strlen(resource)+3];
@@ -778,29 +844,14 @@ int sync_unlock(const char *resource, int lockid)
 	   lockid == LCK_READ ||
 	   lockid == LCK_WRITE);
 
-    switch (lockid)
-    {
-    case LCK_EXCL:
-	status = _unlock_resource(lock1, lockid);
-	if (status)
-	    goto out;
-	status = _unlock_resource(lock2, lockid);
-	break;
+    status = _unlock_resource(lock1, lockid);
+    if (!status)
+	    status = _unlock_resource(lock2, lockid);
 
-    case LCK_READ:
-	status = _unlock_resource(lock1, lockid);
-	break;
-
-    case LCK_WRITE:
-	status = _unlock_resource(lock2, lockid);
-	break;
-    }
-
- out:
     return status;
 }
 
-int is_quorate()
+static int _is_quorate()
 {
     if (current_corestate == lg_core_Slave ||
 	current_corestate == lg_core_Master ||
@@ -819,31 +870,43 @@ static int get_all_cluster_nodes()
     int ctree;
     char *nodename;
     int error;
+    int i;
 
     /* Open the config file */
-    ctree = ccs_connect();
-    if (ctree <= 0)
+    ctree = ccs_force_connect(NULL, 1);
+    if (ctree < 0)
     {
 	log_error("Error connecting to CCS");
 	return -1;
     }
 
-    error = ccs_get(ctree, "//nodes/node/@name", &nodename);
-    while (nodename)
+    for (i=1;;i++)
     {
-	char nodeip[MAX_CSID_LEN];
-	char *clvmflag;
+	char nodekey[256];
+	char nodeip[GULM_MAX_CSID_LEN];
+	int  clvmflag = 1;
+	char *clvmflagstr;
 	char key[256];
 
-	sprintf(key, "//nodes/node[@name=\"%s\"]/clvm", nodename);
-	ccs_get(ctree, key, &clvmflag);
+	sprintf(nodekey, "//cluster/clusternodes/clusternode[%d]/@name", i);
+	error = ccs_get(ctree, nodekey, &nodename);
+	if (error)
+	    break;
 
-	if ((get_ip_address(nodename, nodeip) == 0) && atoi(clvmflag))
+	sprintf(key, "//cluster/clusternodes/clusternode[@name=\"%s\"]/clvm", nodename);
+	if (!ccs_get(ctree, key, &clvmflagstr))
+	{
+	    clvmflag = atoi(clvmflagstr);
+	    free(clvmflagstr);
+	}
+
+	DEBUGLOG("Got node %s from ccs(clvmflag = %d)\n", nodename, clvmflag);
+	if ((get_ip_address(nodename, nodeip) == 0) && clvmflag)
 	{
 	    struct node_info *ninfo;
 
 	    /* If it's not in the list, then add it */
-	    ninfo = hash_lookup_binary(node_hash, nodeip, MAX_CSID_LEN);
+	    ninfo = hash_lookup_binary(node_hash, nodeip, GULM_MAX_CSID_LEN);
 	    if (!ninfo)
 	    {
 		ninfo = malloc(sizeof(struct node_info));
@@ -856,16 +919,14 @@ static int get_all_cluster_nodes()
 		strcpy(ninfo->name, nodename);
 
 		ninfo->state = NODE_DOWN;
-		hash_insert_binary(node_hash, nodeip, MAX_CSID_LEN, ninfo);
+		hash_insert_binary(node_hash, nodeip, GULM_MAX_CSID_LEN, ninfo);
 	    }
 	}
 	else
 	{
 	    DEBUGLOG("node %s has clvm disabled\n", nodename);
 	}
-	if (clvmflag) free(clvmflag);
 	free(nodename);
-	error = ccs_get(ctree, "//nodes/node/@name", &nodename);
     }
 
     /* Finished with config file */
@@ -874,7 +935,42 @@ static int get_all_cluster_nodes()
     return 0;
 }
 
-int gulm_fd(void)
+static int _get_main_cluster_fd(void)
 {
-    return lg_core_selector(gulm_if);
+	return get_main_gulm_cluster_fd();
+}
+
+static int _cluster_fd_callback(struct local_client *fd, char *buf, int len, char *csid, struct local_client **new_client)
+{
+	return cluster_fd_gulm_callback(fd, buf, len, csid, new_client);
+}
+
+static int _cluster_send_message(void *buf, int msglen, char *csid, const char *errtext)
+{
+	return gulm_cluster_send_message(buf, msglen, csid, errtext);
+}
+
+static struct cluster_ops _cluster_gulm_ops = {
+	.cluster_init_completed   = NULL,
+	.cluster_send_message     = _cluster_send_message,
+	.name_from_csid           = gulm_name_from_csid,
+	.csid_from_name           = _csid_from_name,
+	.get_num_nodes            = _get_num_nodes,
+	.cluster_fd_callback      = _cluster_fd_callback,
+	.get_main_cluster_fd      = _get_main_cluster_fd,
+	.cluster_do_node_callback = _cluster_do_node_callback,
+	.is_quorate               = _is_quorate,
+	.get_our_csid             = _get_our_csid,
+	.add_up_node              = gulm_add_up_node,
+	.cluster_closedown        = _cluster_closedown,
+	.sync_lock                = _sync_lock,
+	.sync_unlock              = _sync_unlock,
+};
+
+struct cluster_ops *init_gulm_cluster(void)
+{
+	if (!_init_cluster())
+		return &_cluster_gulm_ops;
+	else
+		return NULL;
 }

daemons/clvmd/clvmd-gulm.h

@@ -7,3 +7,6 @@ extern int gulm_fd(void);
 extern int get_ip_address(char *node, char *addr);
 extern void tcp_remove_client(char *csid);
 extern int alloc_client(int fd, char *csid, struct local_client **new_client);
+
+void gulm_add_up_node(char *csid);
+int gulm_name_from_csid(char *csid, char *name);

daemons/clvmd/clvmd.c

@@ -1,6 +1,6 @@
 /*
  * Copyright (C) 2002-2004 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.
  *
@@ -34,11 +34,13 @@
 #include <unistd.h>
 #include <fcntl.h>
 #include <getopt.h>
+#include <syslog.h>
 #include <errno.h>
 
 #include "clvmd-comms.h"
 #include "lvm-functions.h"
 #include "clvm.h"
+#include "version.h"
 #include "clvmd.h"
 #include "libdlm.h"
 #include "system-lv.h"
@@ -54,9 +56,9 @@
 
 /* The maximum size of a message that will fit into a packet. Anything bigger
    than this is sent via the system LV */
-#define MAX_INLINE_MESSAGE (MAX_CLUSTER_MESSAGE-sizeof(struct clvm_header))
+#define MAX_INLINE_MESSAGE (max_cluster_message-sizeof(struct clvm_header))
 
-#define ISLOCAL_CSID(c) (memcmp(c, our_csid, MAX_CSID_LEN) == 0)
+#define ISLOCAL_CSID(c) (memcmp(c, our_csid, max_csid_len) == 0)
 
 /* Head of the fd list. Also contains
    the cluster_socket details */
@@ -64,7 +66,12 @@ static struct local_client local_client_head;
 
 static unsigned short global_xid = 0;	/* Last transaction ID issued */
 
+static struct cluster_ops *clops = NULL;
+
 static char our_csid[MAX_CSID_LEN];
+static unsigned max_csid_len;
+static unsigned max_cluster_message;
+static unsigned max_cluster_member_name_len;
 
 /* Structure of items on the LVM thread list */
 struct lvm_thread_cmd {
@@ -81,7 +88,15 @@ static pthread_t lvm_thread;
 static pthread_mutex_t lvm_thread_mutex;
 static pthread_cond_t lvm_thread_cond;
 static struct list lvm_cmd_head;
-static int quit = 0;
+static volatile sig_atomic_t quit = 0;
+static int child_pipe[2];
+
+/* Reasons the daemon failed initialisation */
+#define DFAIL_INIT       1
+#define DFAIL_LOCAL_SOCK 2
+#define DFAIL_CLUSTER_IF 3
+#define DFAIL_MALLOC     4
+#define SUCCESS          0
 
 /* Prototypes for code further down */
 static void sigusr2_handler(int sig);
@@ -129,6 +144,18 @@ static void usage(char *prog, FILE *file)
 	fprintf(file, "\n");
 }
 
+/* Called to signal the parent how well we got on during initialisation */
+static void child_init_signal(int status)
+{
+        if (child_pipe[1]) {
+	        write(child_pipe[1], &status, sizeof(status));
+		close(child_pipe[1]);
+	}
+	if (status)
+	        exit(status);
+}
+
+
 int main(int argc, char *argv[])
 {
 	int local_sock;
@@ -138,6 +165,7 @@ int main(int argc, char *argv[])
 	int debug = 0;
 	int cmd_timeout = DEFAULT_CMD_TIMEOUT;
 	sigset_t ss;
+	int using_gulm = 0;
 
 	/* Deal with command-line arguments */
 	opterr = 0;
@@ -166,7 +194,8 @@ int main(int argc, char *argv[])
 			break;
 
 		case 'V':
-			printf("\nCluster LVM Daemon version %d.%d.%d\n\n",
+		        printf("Cluster LVM daemon version: %s\n", LVM_VERSION);
+			printf("Protocol version:           %d.%d.%d\n",
 			       CLVMD_MAJOR_VERSION, CLVMD_MINOR_VERSION,
 			       CLVMD_PATCH_VERSION);
 			exit(1);
@@ -188,14 +217,12 @@ int main(int argc, char *argv[])
 	   but the cluster is not ready yet */
 	local_sock = open_local_sock();
 	if (local_sock < 0)
-		exit(2);
+		child_init_signal(DFAIL_LOCAL_SOCK);
 
 	/* Set up signal handlers, USR1 is for cluster change notifications (in cman)
 	   USR2 causes child threads to exit.
 	   PIPE should be ignored */
 	signal(SIGUSR2, sigusr2_handler);
-	signal(SIGTERM, sigterm_handler);
-	signal(SIGINT, sigterm_handler);
 	signal(SIGPIPE, SIG_IGN);
 
 	/* Block SIGUSR2 in the main process */
@@ -210,26 +237,45 @@ int main(int argc, char *argv[])
 	init_lvhash();
 
 	/* Start the cluster interface */
-	if (init_cluster()) {
+#ifdef USE_CMAN
+	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;
+		syslog(LOG_NOTICE, "Cluster LVM daemon started - connected to CMAN");
+	}
+#endif
+#ifdef USE_GULM
+	if (!clops)
+		if ((clops = init_gulm_cluster())) {
+			max_csid_len = GULM_MAX_CSID_LEN;
+			max_cluster_message = GULM_MAX_CLUSTER_MESSAGE;
+			max_cluster_member_name_len = GULM_MAX_CLUSTER_MEMBER_NAME_LEN;
+			using_gulm = 1;
+			syslog(LOG_NOTICE, "Cluster LVM daemon started - connected to GULM");
+		}
+#endif
+
+	if (!clops) {
 		DEBUGLOG("Can't initialise cluster interface\n");
 		log_error("Can't initialise cluster interface\n");
-		exit(5);
+		child_init_signal(DFAIL_CLUSTER_IF);
 	}
 	DEBUGLOG("Cluster ready, doing some more initialisation\n");
 
 	/* Save our CSID */
 	uname(&nodeinfo);
-	get_our_csid(our_csid);
+	clops->get_our_csid(our_csid);
 
 	/* Initialise the FD list head */
-	local_client_head.fd = get_main_cluster_fd();
+	local_client_head.fd = clops->get_main_cluster_fd();
 	local_client_head.type = CLUSTER_MAIN_SOCK;
-	local_client_head.callback = cluster_fd_callback;
+	local_client_head.callback = clops->cluster_fd_callback;
 
 	/* Add the local socket to the list */
 	newfd = malloc(sizeof(struct local_client));
 	if (!newfd)
-		exit(2);
+	        child_init_signal(DFAIL_MALLOC);
 
 	newfd->fd = local_sock;
 	newfd->type = LOCAL_RENDEZVOUS;
@@ -240,17 +286,21 @@ int main(int argc, char *argv[])
 	/* This needs to be started after cluster initialisation
 	   as it may need to take out locks */
 	DEBUGLOG("starting LVM thread\n");
-	pthread_create(&lvm_thread, NULL, lvm_thread_fn, nodeinfo.nodename);
+	pthread_create(&lvm_thread, NULL, lvm_thread_fn, (void *)using_gulm);
 
-#ifndef USE_GULM
 	/* Tell the rest of the cluster our version number */
 	/* CMAN can do this immediately, gulm needs to wait until
 	   the core initialisation has finished and the node list
 	   has been gathered */
-	send_version_message();
-#endif
+	if (clops->cluster_init_completed)
+		clops->cluster_init_completed();
 
 	DEBUGLOG("clvmd ready for work\n");
+	child_init_signal(SUCCESS);
+
+	/* Try to shutdown neatly */
+	signal(SIGTERM, sigterm_handler);
+	signal(SIGINT, sigterm_handler);
 
 	/* Do some work */
 	main_loop(local_sock, cmd_timeout);
@@ -303,6 +353,7 @@ static int local_rendezvous_callback(struct local_client *thisfd, char *buf,
 		newfd->bits.localsock.threadid = 0;
 		newfd->bits.localsock.finished = 0;
 		newfd->bits.localsock.pipe_client = NULL;
+		newfd->bits.localsock.private = NULL;
 		newfd->bits.localsock.all_success = 1;
 		DEBUGLOG("Got new connection on fd %d\n", newfd->fd);
 		*new_client = newfd;
@@ -391,9 +442,9 @@ static void timedout_callback(struct local_client *client, char *csid,
 {
 	if (node_up) {
 		struct node_reply *reply;
-		char nodename[MAX_CLUSTER_MEMBER_NAME_LEN];
+		char nodename[max_cluster_member_name_len];
 
-		name_from_csid(csid, nodename);
+		clops->name_from_csid(csid, nodename);
 		DEBUGLOG("PJC: checking for a reply from %s\n", nodename);
 		pthread_mutex_lock(&client->bits.localsock.reply_mutex);
 
@@ -422,7 +473,7 @@ static void timedout_callback(struct local_client *client, char *csid,
 static void request_timed_out(struct local_client *client)
 {
 	DEBUGLOG("Request timed-out. padding\n");
-	cluster_do_node_callback(client, timedout_callback);
+	clops->cluster_do_node_callback(client, timedout_callback);
 
 	if (client->bits.localsock.num_replies !=
 	    client->bits.localsock.expected_replies) {
@@ -447,7 +498,7 @@ static void main_loop(int local_sock, int cmd_timeout)
 		int select_status;
 		struct local_client *thisfd;
 		struct timeval tv = { cmd_timeout, 0 };
-		int quorate = is_quorate();
+		int quorate = clops->is_quorate();
 
 		/* Wait on the cluster FD and all local sockets/pipes */
 		FD_ZERO(&in);
@@ -461,9 +512,8 @@ static void main_loop(int local_sock, int cmd_timeout)
 
 		if ((select_status = select(FD_SETSIZE, &in, NULL, NULL, &tv)) > 0) {
 			struct local_client *lastfd = NULL;
-			struct clvm_header *inheader;
 			char csid[MAX_CSID_LEN];
-			char buf[MAX_CLUSTER_MESSAGE];
+			char buf[max_cluster_message];
 
 			for (thisfd = &local_client_head; thisfd != NULL;
 			     thisfd = thisfd->next) {
@@ -490,13 +540,14 @@ static void main_loop(int local_sock, int cmd_timeout)
 						    type == CLUSTER_INTERNAL)
 							goto closedown;
 
-						DEBUGLOG
-						    ("ret == %d, errno = %d. removing client\n",
-						     ret, errno);
+						DEBUGLOG("ret == %d, errno = %d. removing client\n",
+							 ret, errno);
 						lastfd->next = thisfd->next;
 						free_fd = thisfd;
 						thisfd = lastfd;
-						free(free_fd);
+
+						/* Queue cleanup, this also frees the client struct */
+						add_to_lvmqueue(free_fd, NULL, 0, NULL);
 						break;
 					}
 
@@ -506,33 +557,6 @@ static void main_loop(int local_sock, int cmd_timeout)
 						thisfd->next = newfd;
 						break;
 					}
-
-					switch (thisfd->type) {
-					case CLUSTER_MAIN_SOCK:
-					case CLUSTER_DATA_SOCK:
-						inheader =
-						    (struct clvm_header *) buf;
-						ntoh_clvm(inheader);	/* Byteswap fields */
-						if (inheader->cmd ==
-						    CLVMD_CMD_REPLY)
-							    process_reply
-							    (inheader, ret,
-							     csid);
-						else
-							add_to_lvmqueue(thisfd,
-									inheader,
-									ret,
-									csid);
-						break;
-
-						/* All the work for these is done in the callback
-						   rightly or wrongly... */
-					case LOCAL_RENDEZVOUS:
-					case LOCAL_SOCK:
-					case THREAD_PIPE:
-					case CLUSTER_INTERNAL:
-						break;
-					}
 				}
 				lastfd = thisfd;
 			}
@@ -575,30 +599,69 @@ static void main_loop(int local_sock, int cmd_timeout)
 	}
 
       closedown:
-	cluster_closedown();
+	clops->cluster_closedown();
 	close(local_sock);
 }
 
-/* Fork into the background and detach from our parent process */
+/*
+ * Fork into the background and detach from our parent process.
+ * In the interests of user-friendliness we wait for the daemon
+ * to complete initialisation before returning its status
+ * the the user.
+ */
 static void be_daemon()
 {
-	pid_t pid;
+        pid_t pid;
+	int child_status;
 	int devnull = open("/dev/null", O_RDWR);
 	if (devnull == -1) {
 		perror("Can't open /dev/null");
 		exit(3);
 	}
 
+	pipe(child_pipe);
+
 	switch (pid = fork()) {
 	case -1:
 		perror("clvmd: can't fork");
 		exit(2);
 
-	case 0:		/* child */
+	case 0:		/* Child */
+	        close(child_pipe[0]);
 		break;
 
-	default:		/* Parent */
-		exit(0);
+	default:       /* Parent */
+		close(child_pipe[1]);
+		if (read(child_pipe[0], &child_status, sizeof(child_status)) !=
+		    sizeof(child_status)) {
+
+		        fprintf(stderr, "clvmd failed in initialisation\n");
+		        exit(DFAIL_INIT);
+		}
+		else {
+		        switch (child_status) {
+			case SUCCESS:
+			        break;
+			case DFAIL_INIT:
+			        fprintf(stderr, "clvmd failed in initialisation\n");
+				break;
+			case DFAIL_LOCAL_SOCK:
+			        fprintf(stderr, "clvmd could not create local socket\n");
+				fprintf(stderr, "Another clvmd is probably already running\n");
+				break;
+			case DFAIL_CLUSTER_IF:
+			        fprintf(stderr, "clvmd could not connect to cluster manager\n");
+				fprintf(stderr, "Consult syslog for more information\n");
+				break;
+			case DFAIL_MALLOC:
+			        fprintf(stderr, "clvmd failed, not enough memory\n");
+				break;
+			default:
+			        fprintf(stderr, "clvmd failed, error was %d\n", child_status);
+				break;
+			}
+			exit(child_status);
+		}
 	}
 
 	/* Detach ourself from the calling environment */
@@ -659,7 +722,7 @@ static int read_from_local_sock(struct local_client *thisfd)
 		if (thisfd->bits.localsock.threadid) {
 			DEBUGLOG("Waiting for child thread\n");
 			pthread_mutex_lock(&thisfd->bits.localsock.mutex);
-			thisfd->bits.localsock.state = POST_COMMAND;
+			thisfd->bits.localsock.state = PRE_COMMAND;
 			pthread_cond_signal(&thisfd->bits.localsock.cond);
 			pthread_mutex_unlock(&thisfd->bits.localsock.mutex);
 			pthread_kill(thisfd->bits.localsock.threadid, SIGUSR2);
@@ -717,6 +780,7 @@ static int read_from_local_sock(struct local_client *thisfd)
 		struct local_client *newfd;
 		char csid[MAX_CSID_LEN];
 		struct clvm_header *inheader;
+		int status;
 
 		inheader = (struct clvm_header *) buffer;
 
@@ -727,7 +791,7 @@ static int read_from_local_sock(struct local_client *thisfd)
 		if (thisfd->bits.localsock.in_progress) {
 			struct clvm_header reply;
 			reply.cmd = CLVMD_CMD_REPLY;
-			reply.status = -EBUSY;
+			reply.status = EBUSY;
 			reply.arglen = 0;
 			reply.flags = 0;
 			send_message(&reply, sizeof(reply), our_csid,
@@ -750,7 +814,7 @@ static int read_from_local_sock(struct local_client *thisfd)
 		if (!thisfd->bits.localsock.cmd) {
 			struct clvm_header reply;
 			reply.cmd = CLVMD_CMD_REPLY;
-			reply.status = -ENOMEM;
+			reply.status = ENOMEM;
 			reply.arglen = 0;
 			reply.flags = 0;
 			send_message(&reply, sizeof(reply), our_csid,
@@ -791,13 +855,13 @@ static int read_from_local_sock(struct local_client *thisfd)
 		}
 
 		/* Check the node name for validity */
-		if (inheader->node[0] && csid_from_name(csid, inheader->node)) {
+		if (inheader->node[0] && clops->csid_from_name(csid, inheader->node)) {
 			/* Error, node is not in the cluster */
 			struct clvm_header reply;
 			DEBUGLOG("Unknown node: '%s'\n", inheader->node);
 
 			reply.cmd = CLVMD_CMD_REPLY;
-			reply.status = -ENOENT;
+			reply.status = ENOENT;
 			reply.flags = 0;
 			reply.arglen = 0;
 			send_message(&reply, sizeof(reply), our_csid,
@@ -828,7 +892,7 @@ static int read_from_local_sock(struct local_client *thisfd)
 			close(comms_pipe[1]);
 
 			reply.cmd = CLVMD_CMD_REPLY;
-			reply.status = -ENOMEM;
+			reply.status = ENOMEM;
 			reply.arglen = 0;
 			reply.flags = 0;
 			send_message(&reply, sizeof(reply), our_csid,
@@ -863,8 +927,10 @@ static int read_from_local_sock(struct local_client *thisfd)
 		/* Run the pre routine */
 		thisfd->bits.localsock.in_progress = TRUE;
 		thisfd->bits.localsock.state = PRE_COMMAND;
-		pthread_create(&thisfd->bits.localsock.threadid, NULL,
+		DEBUGLOG("Creating pre&post thread\n");
+		status = pthread_create(&thisfd->bits.localsock.threadid, NULL,
 			       pre_and_post_thread, thisfd);
+		DEBUGLOG("Created pre&post thread, state = %d\n", status);
 	}
 	return len;
 }
@@ -921,7 +987,7 @@ static int distribute_command(struct local_client *thisfd)
 		/* if node is empty then do it on the whole cluster */
 		if (inheader->node[0] == '\0') {
 			thisfd->bits.localsock.expected_replies =
-			    get_num_nodes();
+			    clops->get_num_nodes();
 			thisfd->bits.localsock.num_replies = 0;
 			thisfd->bits.localsock.sent_time = time(NULL);
 			thisfd->bits.localsock.in_progress = TRUE;
@@ -942,7 +1008,7 @@ static int distribute_command(struct local_client *thisfd)
                         /* Do it on a single node */
 			char csid[MAX_CSID_LEN];
 
-			if (csid_from_name(csid, inheader->node)) {
+			if (clops->csid_from_name(csid, inheader->node)) {
 				/* This has already been checked so should not happen */
 				return 0;
 			} else {
@@ -952,7 +1018,7 @@ static int distribute_command(struct local_client *thisfd)
 				thisfd->bits.localsock.in_progress = TRUE;
 
 				/* Are we the requested node ?? */
-				if (memcmp(csid, our_csid, MAX_CSID_LEN) == 0) {
+				if (memcmp(csid, our_csid, max_csid_len) == 0) {
 					DEBUGLOG("Doing command on local node only\n");
 					add_to_lvmqueue(thisfd, inheader, len, NULL);
 				} else {
@@ -984,14 +1050,14 @@ void process_remote_command(struct clvm_header *msg, int msglen, int fd,
 			    char *csid)
 {
 	char *replyargs;
-	char nodename[MAX_CLUSTER_MEMBER_NAME_LEN];
+	char nodename[max_cluster_member_name_len];
 	int replylen = 0;
-	int buflen = MAX_CLUSTER_MESSAGE - sizeof(struct clvm_header) - 1;
+	int buflen = max_cluster_message - sizeof(struct clvm_header) - 1;
 	int status;
 	int msg_malloced = 0;
 
 	/* Get the node name as we /may/ need it later */
-	name_from_csid(csid, nodename);
+	clops->name_from_csid(csid, nodename);
 
 	DEBUGLOG("process_remote_command %d for clientid 0x%x on node %s\n",
 		 msg->cmd, msg->clientid, nodename);
@@ -1016,7 +1082,7 @@ void process_remote_command(struct clvm_header *msg, int msglen, int fd,
 
 				/* Return a failure response */
 				head.cmd = CLVMD_CMD_REPLY;
-				head.status = -EFBIG;
+				head.status = EFBIG;
 				head.flags = 0;
 				head.clientid = msg->clientid;
 				head.arglen = 0;
@@ -1033,7 +1099,7 @@ void process_remote_command(struct clvm_header *msg, int msglen, int fd,
 			     msg->arglen);
 			/* Return a failure response */
 			head.cmd = CLVMD_CMD_REPLY;
-			head.status = -ENOMEM;
+			head.status = ENOMEM;
 			head.flags = 0;
 			head.clientid = msg->clientid;
 			head.arglen = 0;
@@ -1057,7 +1123,7 @@ void process_remote_command(struct clvm_header *msg, int msglen, int fd,
 	if (msg->cmd == CLVMD_CMD_VERSION) {
 		int *version_nums = (int *) msg->args;
 		char node[256];
-		name_from_csid(csid, node);
+		clops->name_from_csid(csid, node);
 		DEBUGLOG("Remote node %s is version %d.%d.%d\n",
 			 node,
 			 ntohl(version_nums[0]),
@@ -1078,17 +1144,17 @@ void process_remote_command(struct clvm_header *msg, int msglen, int fd,
 			byebyemsg.flags = 0;
 			byebyemsg.arglen = 0;
 			byebyemsg.clientid = 0;
-			cluster_send_message(&byebyemsg, sizeof(byebyemsg),
+			clops->cluster_send_message(&byebyemsg, sizeof(byebyemsg),
 					     our_csid,
 					     "Error Sending GOAWAY message");
 		} else {
-			add_up_node(csid);
+			clops->add_up_node(csid);
 		}
 		return;
 	}
 
 	/* Allocate a default reply buffer */
-	replyargs = malloc(MAX_CLUSTER_MESSAGE - sizeof(struct clvm_header));
+	replyargs = malloc(max_cluster_message - sizeof(struct clvm_header));
 
 	if (replyargs != NULL) {
 		/* Run the command */
@@ -1096,7 +1162,7 @@ void process_remote_command(struct clvm_header *msg, int msglen, int fd,
 		    do_command(NULL, msg, msglen, &replyargs, buflen,
 			       &replylen);
 	} else {
-		status = -ENOMEM;
+		status = ENOMEM;
 	}
 
 	/* If it wasn't a reply, then reply */
@@ -1127,11 +1193,10 @@ void process_remote_command(struct clvm_header *msg, int msglen, int fd,
 
 				/* If System LV operation failed then report it as EFBIG but only do it
 				   if the data buffer has something in it. */
-				if (system_lv_write_data
-				    (aggreply,
-				     replylen + sizeof(struct clvm_header)) < 0
+				if (system_lv_write_data(aggreply,
+							 replylen + sizeof(struct clvm_header)) < 0
 				    && replylen > 0)
-					agghead->status = -EFBIG;
+					agghead->status = EFBIG;
 
 				send_message(agghead,
 					     sizeof(struct clvm_header), csid,
@@ -1147,7 +1212,7 @@ void process_remote_command(struct clvm_header *msg, int msglen, int fd,
 				agghead->node[0] = '\0';
 				send_message(aggreply,
 					     sizeof(struct clvm_header) +
-					     replylen + 2, csid, fd,
+					     replylen, csid, fd,
 					     "Error sending command reply");
 			}
 		} else {
@@ -1156,7 +1221,7 @@ void process_remote_command(struct clvm_header *msg, int msglen, int fd,
 			DEBUGLOG("Error attempting to realloc return buffer\n");
 			/* Return a failure response */
 			head.cmd = CLVMD_CMD_REPLY;
-			head.status = -ENOMEM;
+			head.status = ENOMEM;
 			head.flags = 0;
 			head.clientid = msg->clientid;
 			head.arglen = 0;
@@ -1188,13 +1253,13 @@ static void add_reply_to_list(struct local_client *client, int status,
 	reply = malloc(sizeof(struct node_reply));
 	if (reply) {
 		reply->status = status;
-		name_from_csid(csid, reply->node);
+		clops->name_from_csid(csid, reply->node);
 		DEBUGLOG("Reply from node %s: %d bytes\n", reply->node, len);
 
 		if (len > 0) {
 			reply->replymsg = malloc(len);
 			if (!reply->replymsg) {
-				reply->status = -ENOMEM;
+				reply->status = ENOMEM;
 			} else {
 				memcpy(reply->replymsg, buf, len);
 			}
@@ -1273,13 +1338,12 @@ static void *pre_and_post_thread(void *arg)
 
 		DEBUGLOG("Got post command condition...\n");
 
+		/* POST function must always run, even if the client aborts */
+		status = 0;
 		do_post_command(client);
 
 		write(pipe_fd, &status, sizeof(int));
 
-		if (client->bits.localsock.finished)
-			break;
-
 		DEBUGLOG("Waiting for next pre command\n");
 
 		pthread_mutex_lock(&client->bits.localsock.mutex);
@@ -1292,7 +1356,8 @@ static void *pre_and_post_thread(void *arg)
 		DEBUGLOG("Got pre command condition...\n");
 	}
 	DEBUGLOG("Subthread finished\n");
-	return (void *) 0;
+	pthread_exit((void *) 0);
+	return 0;
 }
 
 /* Process a command on the local node and store the result */
@@ -1300,8 +1365,8 @@ static int process_local_command(struct clvm_header *msg, int msglen,
 				 struct local_client *client,
 				 unsigned short xid)
 {
-	char *replybuf = malloc(MAX_CLUSTER_MESSAGE);
-	int buflen = MAX_CLUSTER_MESSAGE - sizeof(struct clvm_header) - 1;
+	char *replybuf = malloc(max_cluster_message);
+	int buflen = max_cluster_message - sizeof(struct clvm_header) - 1;
 	int replylen = 0;
 	int status;
 
@@ -1383,9 +1448,10 @@ static void send_local_reply(struct local_client *client, int status, int fd)
 	replybuf = malloc(message_len);
 
 	clientreply = (struct clvm_header *) replybuf;
-	clientreply->status = -status;
+	clientreply->status = status;
 	clientreply->cmd = CLVMD_CMD_REPLY;
 	clientreply->node[0] = '\0';
+	clientreply->flags = 0;
 
 	ptr = clientreply->args;
 
@@ -1397,6 +1463,9 @@ static void send_local_reply(struct local_client *client, int status, int fd)
 		strcpy(ptr, thisreply->node);
 		ptr += strlen(thisreply->node) + 1;
 
+		if (thisreply->status)
+			clientreply->flags |= CLVMD_FLAG_NODEERRS;
+
 		*(int *) ptr = thisreply->status;
 		ptr += sizeof(int);
 
@@ -1465,7 +1534,8 @@ static void send_version_message()
 	version_nums[1] = htonl(CLVMD_MINOR_VERSION);
 	version_nums[2] = htonl(CLVMD_PATCH_VERSION);
 
-	cluster_send_message(message, sizeof(message), NULL,
+	hton_clvm(msg);
+	clops->cluster_send_message(message, sizeof(message), NULL,
 			     "Error Sending version number");
 }
 
@@ -1477,8 +1547,8 @@ static int send_message(void *buf, int msglen, char *csid, int fd,
 
 	/* Send remote messages down the cluster socket */
 	if (csid == NULL || !ISLOCAL_CSID(csid)) {
-		hton_clvm((struct clvm_header *) buf);	/* Byte swap if necessary */
-		return cluster_send_message(buf, msglen, csid, errtext);
+		hton_clvm((struct clvm_header *) buf);
+		return clops->cluster_send_message(buf, msglen, csid, errtext);
 	} else {
 		int ptr = 0;
 
@@ -1498,6 +1568,14 @@ static int send_message(void *buf, int msglen, char *csid, int fd,
 
 static int process_work_item(struct lvm_thread_cmd *cmd)
 {
+
+	/* If msg is NULL then this is a cleanup request */
+	if (cmd->msg == NULL) {
+		cmd_client_cleanup(cmd->client);
+		free(cmd->client);
+		return 0;
+	}
+
 	if (!cmd->remote) {
 		DEBUGLOG("process_work_item: local\n");
 		process_local_command(cmd->msg, cmd->msglen, cmd->client,
@@ -1517,9 +1595,9 @@ static void *lvm_thread_fn(void *arg)
 {
 	struct list *cmdl, *tmp;
 	sigset_t ss;
+	int using_gulm = (int)arg;
 
 	DEBUGLOG("LVM thread function started\n");
-	pthread_mutex_lock(&lvm_thread_mutex);
 
 	/* Ignore SIGUSR1 & 2 */
 	sigemptyset(&ss);
@@ -1528,8 +1606,7 @@ static void *lvm_thread_fn(void *arg)
 	pthread_sigmask(SIG_BLOCK, &ss, NULL);
 
 	/* Initialise the interface to liblvm */
-	init_lvm();
-	pthread_mutex_unlock(&lvm_thread_mutex);
+	init_lvm(using_gulm);
 
 	/* Now wait for some actual work */
 	for (;;) {
@@ -1565,21 +1642,26 @@ static int add_to_lvmqueue(struct local_client *client, struct clvm_header *msg,
 
 	cmd = malloc(sizeof(struct lvm_thread_cmd));
 	if (!cmd)
-		return -ENOMEM;
+		return ENOMEM;
 
-	cmd->msg = malloc(msglen);
-	if (!cmd->msg) {
-		log_error("Unable to allocate buffer space\n");
-		free(cmd);
-		return -1;
+	if (msglen) {
+		cmd->msg = malloc(msglen);
+		if (!cmd->msg) {
+			log_error("Unable to allocate buffer space\n");
+			free(cmd);
+			return -1;
+		}
+		memcpy(cmd->msg, msg, msglen);
+	}
+	else {
+		cmd->msg = NULL;
 	}
-
 	cmd->client = client;
 	cmd->msglen = msglen;
 	cmd->xid = client->xid;
-	memcpy(cmd->msg, msg, msglen);
+
 	if (csid) {
-		memcpy(cmd->csid, csid, MAX_CSID_LEN);
+		memcpy(cmd->csid, csid, max_csid_len);
 		cmd->remote = 1;
 	} else {
 		cmd->remote = 0;
@@ -1610,6 +1692,8 @@ static int open_local_sock()
 		log_error("Can't create local socket: %m");
 		return -1;
 	}
+	/* Set Close-on-exec */
+	fcntl(local_socket, F_SETFD, 1);
 
 	memset(&sockaddr, 0, sizeof(sockaddr));
 	memcpy(sockaddr.sun_path, CLVMD_SOCKNAME, sizeof(CLVMD_SOCKNAME));
@@ -1630,11 +1714,24 @@ static int open_local_sock()
 	return local_socket;
 }
 
+void process_message(struct local_client *client, char *buf, int len, char *csid)
+{
+	struct clvm_header *inheader;
+
+	inheader = (struct clvm_header *) buf;
+	ntoh_clvm(inheader);	/* Byteswap fields */
+	if (inheader->cmd == CLVMD_CMD_REPLY)
+		process_reply(inheader, len, csid);
+	else
+		add_to_lvmqueue(client, inheader, len, csid);
+}
+
+
 static void check_all_callback(struct local_client *client, char *csid,
 			       int node_up)
 {
 	if (!node_up)
-		add_reply_to_list(client, -EHOSTDOWN, csid, "CLVMD not running",
+		add_reply_to_list(client, EHOSTDOWN, csid, "CLVMD not running",
 				  18);
 }
 
@@ -1644,7 +1741,7 @@ static void check_all_callback(struct local_client *client, char *csid,
 static int check_all_clvmds_running(struct local_client *client)
 {
 	DEBUGLOG("check_all_clvmds_running\n");
-	return cluster_do_node_callback(client, check_all_callback);
+	return clops->cluster_do_node_callback(client, check_all_callback);
 }
 
 /* Return a local_client struct given a client ID.
@@ -1681,7 +1778,6 @@ static void ntoh_clvm(struct clvm_header *hdr)
 static void sigusr2_handler(int sig)
 {
 	DEBUGLOG("SIGUSR2 received\n");
-	pthread_exit((void *) -1);
 	return;
 }
 
@@ -1691,3 +1787,14 @@ static void sigterm_handler(int sig)
 	quit = 1;
 	return;
 }
+
+int sync_lock(const char *resource, int mode, int flags, int *lockid)
+{
+	return clops->sync_lock(resource, mode, flags, lockid);
+}
+
+int sync_unlock(const char *resource, int lockid)
+{
+	return clops->sync_unlock(resource, lockid);
+}
+

daemons/clvmd/clvmd.h

@@ -95,7 +95,7 @@ struct local_client {
 };
 
 #ifdef DEBUG
-#define DEBUGLOG(fmt, args...) fprintf(stderr, "CLVMD[%d]: %ld ", getpid(), time(NULL) ); fprintf(stderr, fmt, ## args)
+#define DEBUGLOG(fmt, args...) fprintf(stderr, "CLVMD[%x]: %ld ", (int)pthread_self(), time(NULL) ); fprintf(stderr, fmt, ## args)
 #else
 #define DEBUGLOG(fmt, args...)
 #endif
@@ -111,9 +111,13 @@ extern int do_command(struct local_client *client, struct clvm_header *msg,
 /* Pre and post command routines are called only on the local node */
 extern int do_pre_command(struct local_client *client);
 extern int do_post_command(struct local_client *client);
-
+extern void cmd_client_cleanup(struct local_client *client);
 extern int add_client(struct local_client *new_client);
 
 extern void clvmd_cluster_init_completed(void);
+extern void process_message(struct local_client *client, char *buf, int len, char *csid);
+
+int sync_lock(const char *resource, int mode, int flags, int *lockid);
+int sync_unlock(const char *resource, int lockid);
 
 #endif

daemons/clvmd/cnxman-socket.h

@@ -63,8 +63,8 @@
 
 
 /* Maximum size of a cluster message */
-#define MAX_CLUSTER_MESSAGE          1500
-#define MAX_CLUSTER_MEMBER_NAME_LEN   255
+#define CMAN_MAX_CLUSTER_MESSAGE          1500
+#define CMAN_MAX_CLUSTER_MEMBER_NAME_LEN   255
 #define MAX_BARRIER_NAME_LEN           33
 #define MAX_SA_ADDR_LEN                12
 #define MAX_CLUSTER_NAME_LEN           16
@@ -105,6 +105,7 @@
 #define MSG_MULTICAST 0x080000	/* Message was sent to all nodes in the cluster
 				 */
 #define MSG_ALLINT    0x100000	/* Send out of all interfaces */
+#define MSG_REPLYEXP  0x200000	/* Reply is expected */
 
 typedef enum { NODESTATE_REMOTEMEMBER, NODESTATE_JOINING, NODESTATE_MEMBER,
 	       NODESTATE_DEAD } nodestate_t;
@@ -146,7 +147,7 @@ struct cl_cluster_node {
 	unsigned int leave_reason;
 	unsigned int incarnation;
 	nodestate_t state;
-	char name[MAX_CLUSTER_MEMBER_NAME_LEN];
+	char name[CMAN_MAX_CLUSTER_MEMBER_NAME_LEN];
 	unsigned char votes;
 };
 

daemons/clvmd/libclvm.c

@@ -1,446 +0,0 @@
-/*
- * Copyright (C) 1997-2004 Sistina Software, Inc. All rights reserved.
- * Copyright (C) 2004 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 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
- */
-
-/* library functions for Cluster LVM Daemon */
-
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <sys/socket.h>
-#include <sys/uio.h>
-#include <sys/un.h>
-#include <sys/time.h>
-#include <sys/ioctl.h>
-#include <sys/utsname.h>
-#include <syslog.h>
-#include <netinet/in.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <stddef.h>
-#include <signal.h>
-#include <unistd.h>
-#include <fcntl.h>
-#include <search.h>
-#include <errno.h>
-
-#include "clvm.h"
-#include "libclvm.h"
-
-/* CLVM in hex! */
-#define LVM_SIGNATURE 0x434C564D
-
-#define MAX_CLUSTER_MEMBER_NAME_LEN 255
-
-/* NOTE: the LVMD uses the socket FD as the client ID, this means
-   that any client that calls fork() will inherit the context of
-   it's parent. */
-static int clvmd_sock = -1;
-
-static int open_local_sock(void)
-{
-	int local_socket;
-	struct sockaddr_un sockaddr;
-
-	/* Open local socket */
-	local_socket = socket(PF_UNIX, SOCK_STREAM, 0);
-	if (local_socket < 0) {
-		perror("Can't create local socket");
-		return -1;
-	}
-
-	fcntl(local_socket, F_SETFD, !FD_CLOEXEC);
-
-	strcpy(sockaddr.sun_path, CLVMD_SOCKNAME);
-	sockaddr.sun_family = AF_UNIX;
-	if (connect
-	    (local_socket, (struct sockaddr *) &sockaddr, sizeof(sockaddr))) {
-		int saved_errno = errno;
-
-		close(local_socket);
-
-		errno = saved_errno;
-		return -1;
-	}
-	return local_socket;
-}
-
-/* Send a request and return the status */
-static int send_request(char *inbuf, int inlen, char **retbuf)
-{
-	char outbuf[PIPE_BUF];
-	struct clvm_header *outheader = (struct clvm_header *) outbuf;
-	int len;
-	int off;
-	fd_set fds;
-
-	FD_ZERO(&fds);
-	FD_SET(clvmd_sock, &fds);
-
-	/* Send it to CLVMD */
-	if (write(clvmd_sock, inbuf, inlen) != inlen) {
-		perror("Error writing to CLVMD");
-		return -1;
-	}
-
-	/* Get the response */
-	if ((len = read(clvmd_sock, outbuf, sizeof(struct clvm_header))) < 0) {
-		perror("Error reading CLVMD");
-		return -1;
-	}
-	if (len == 0) {
-		fprintf(stderr, "EOF reading CLVMD");
-		errno = ENOTCONN;
-		return -1;
-	}
-
-	/* Allocate buffer */
-	*retbuf = malloc(len + outheader->arglen);
-	if (!*retbuf) {
-		errno = ENOMEM;
-		return -1;
-	}
-
-	/* Copy the header */
-	memcpy(*retbuf, outbuf, len);
-	outheader = (struct clvm_header *) *retbuf;
-
-	/* Read the returned values */
-	off = 1;		/* we've already read the first byte */
-
-	while (off < outheader->arglen && len > 0) {
-		len = read(clvmd_sock, outheader->args + off, PIPE_BUF);
-		if (len > 0)
-			off += len;
-	}
-
-	/* Was it an error ? */
-	if (outheader->status < 0) {
-		errno = -outheader->status;
-		return -2;
-	}
-	return 0;
-}
-
-/* Build the structure header and parse-out wildcard node names */
-static void build_header(struct clvm_header *head, int cmd, const char *node,
-			 void *data, int len)
-{
-	head->cmd = cmd;
-	head->status = 0;
-	head->flags = 0;
-	head->clientid = 0;
-	head->arglen = len;
-	if (node) {
-		/* Allow a couple of special node names:
-		   "*" for all nodes,
-		   "." for the local node only
-		 */
-		if (strcmp(node, "*") == 0) {
-			head->node[0] = '\0';
-		} else if (strcmp(node, ".") == 0) {
-			head->node[0] = '\0';
-			head->flags = CLVMD_FLAG_LOCAL;
-		} else {
-			strcpy(head->node, node);
-		}
-	} else {
-		head->node[0] = '\0';
-	}
-}
-
-/* Send a message to a(or all) node(s) in the cluster */
-int lvm_cluster_write(char cmd, char *node, void *data, int len)
-{
-	char outbuf[sizeof(struct clvm_header) + len + strlen(node) + 1];
-	char *retbuf = NULL;
-	int status;
-	struct clvm_header *head = (struct clvm_header *) outbuf;
-
-	if (clvmd_sock == -1)
-		clvmd_sock = open_local_sock();
-	if (clvmd_sock == -1)
-		return -1;
-
-	build_header(head, cmd, node, data, len);
-	memcpy(head->node + strlen(head->node) + 1, data, len);
-
-	status =
-	    send_request(outbuf,
-			 sizeof(struct clvm_header) + strlen(head->node) + len,
-			 &retbuf);
-	if (retbuf)
-		free(retbuf);
-
-	return status;
-}
-
-/* API: Send a message to a(or all) node(s) in the cluster
-   and wait for replies */
-int lvm_cluster_request(char cmd, const char *node, void *data, int len,
-			lvm_response_t ** response, int *num)
-{
-	char outbuf[sizeof(struct clvm_header) + len + strlen(node) + 1];
-	int *outptr;
-	char *inptr;
-	char *retbuf = NULL;
-	int status;
-	int i;
-	int num_responses = 0;
-	struct clvm_header *head = (struct clvm_header *) outbuf;
-	lvm_response_t *rarray;
-
-	*num = 0;
-
-	if (clvmd_sock == -1)
-		clvmd_sock = open_local_sock();
-	if (clvmd_sock == -1)
-		return -1;
-
-	build_header(head, cmd, node, data, len);
-	memcpy(head->node + strlen(head->node) + 1, data, len);
-
-	status =
-	    send_request(outbuf,
-			 sizeof(struct clvm_header) + strlen(head->node) + len,
-			 &retbuf);
-	if (status == 0 || status == -2) {
-		/* Count the number of responses we got */
-		head = (struct clvm_header *) retbuf;
-		inptr = head->args;
-		while (inptr[0]) {
-			num_responses++;
-			inptr += strlen(inptr) + 1;
-			inptr += sizeof(int);
-			inptr += strlen(inptr) + 1;
-		}
-
-		/* Allocate response array. With an extra pair of INTs on the front to sanity
-		   check the pointer when we are given it back to free */
-		outptr =
-		    malloc(sizeof(lvm_response_t) * num_responses +
-			   sizeof(int) * 2);
-		if (!outptr) {
-			if (retbuf)
-				free(retbuf);
-			errno = ENOMEM;
-			return -1;
-		}
-
-		*response = (lvm_response_t *) (outptr + 2);
-		outptr[0] = LVM_SIGNATURE;
-		outptr[1] = num_responses;
-		rarray = *response;
-
-		/* Unpack the response into an lvm_response_t array */
-		inptr = head->args;
-		i = 0;
-		while (inptr[0]) {
-			strcpy(rarray[i].node, inptr);
-			inptr += strlen(inptr) + 1;
-
-			rarray[i].status = *(int *) inptr;
-			inptr += sizeof(int);
-
-			rarray[i].response = malloc(strlen(inptr) + 1);
-			if (rarray[i].response == NULL) {
-				/* Free up everything else and return error */
-				int j;
-				for (j = 0; j < i; j++)
-					free(rarray[i].response);
-				free(outptr);
-				errno = ENOMEM;
-				return -1;
-			}
-
-			strcpy(rarray[i].response, inptr);
-			rarray[i].len = strlen(inptr);
-			inptr += strlen(inptr) + 1;
-			i++;
-		}
-		*num = num_responses;
-		*response = rarray;
-	}
-
-	if (retbuf)
-		free(retbuf);
-	return status;
-}
-
-/* API: Free reply array */
-int lvm_cluster_free_request(lvm_response_t * response)
-{
-	int *ptr = (int *) response - 2;
-	int i;
-	int num;
-
-	/* Check it's ours to free */
-	if (response == NULL || *ptr != LVM_SIGNATURE) {
-		errno = EINVAL;
-		return -1;
-	}
-
-	num = ptr[1];
-	for (i = 0; i < num; i++) {
-		free(response[i].response);
-	}
-	free(ptr);
-
-	return 0;
-}
-
-/* These are a "higher-level" API providing black-box lock/unlock
-   functions for cluster LVM...maybe */
-
-/* Set by lock(), used by unlock() */
-static int num_responses;
-static lvm_response_t *response;
-
-int lvm_lock_for_cluster(char scope, char *name, int verbosity)
-{
-	int status;
-	int i;
-	char *args;
-	int len;
-
-	if (name) {
-		len = strlen(name) + 2;
-		args = alloca(len);
-		strcpy(args + 1, name);
-	} else {
-		len = 2;
-		args = alloca(len);
-		args[1] = '\0';
-	}
-	args[0] = scope;
-
-	status = lvm_cluster_request(CLVMD_CMD_LOCK,
-				     "", args, len, &response, &num_responses);
-
-	/* If any nodes were down then display them and return an error */
-	for (i = 0; i < num_responses; i++) {
-		if (response[i].status == -EHOSTDOWN) {
-			if (verbosity)
-				fprintf(stderr,
-					"clvmd not running on node %s\n",
-					response[i].node);
-			status = -1;
-		}
-	}
-
-	/* If there was an error then free the memory now as the caller won't
-	   want to do the unlock */
-	if (status) {
-		int saved_errno = errno;
-		lvm_cluster_free_request(response);
-		num_responses = 0;
-		errno = saved_errno;
-	}
-	return status;
-}
-
-int lvm_unlock_for_cluster(char scope, char *name, int verbosity)
-{
-	int status;
-	int i;
-	int len;
-	int failed;
-	int num_unlock_responses;
-	char *args;
-	lvm_response_t *unlock_response;
-
-	/* We failed - this should not have been called */
-	if (num_responses == 0)
-		return 0;
-
-	if (name) {
-		len = strlen(name) + 2;
-		args = alloca(len);
-		strcpy(args + 1, name);
-	} else {
-		len = 2;
-		args = alloca(len);
-		args[1] = '\0';
-	}
-	args[0] = scope;
-
-	/* See if it failed anywhere */
-	failed = 0;
-	for (i = 0; i < num_responses; i++) {
-		if (response[i].status != 0)
-			failed++;
-	}
-
-	/* If it failed on any nodes then we only unlock on
-	   the nodes that succeeded */
-	if (failed) {
-		for (i = 0; i < num_responses; i++) {
-			/* Unlock the ones that succeeded */
-			if (response[i].status == 0) {
-				status = lvm_cluster_request(CLVMD_CMD_UNLOCK,
-							     response[i].node,
-							     args, len,
-							     &unlock_response,
-							     &num_unlock_responses);
-				if (status) {
-					if (verbosity)
-						fprintf(stderr,
-							"cluster command to node %s failed: %s\n",
-							response[i].node,
-							strerror(errno));
-				} else if (unlock_response[0].status != 0) {
-					if (verbosity > 1)
-						fprintf(stderr,
-							"unlock on node %s failed: %s\n",
-							response[i].node,
-							strerror(unlock_response
-								 [0].status));
-				}
-				lvm_cluster_free_request(unlock_response);
-			} else {
-				if (verbosity)
-					fprintf(stderr,
-						"command on node %s failed: '%s' - will be left locked\n",
-						response[i].node,
-						strerror(response[i].status));
-			}
-		}
-	} else {
-		/* All OK, we can do a full cluster unlock */
-		status = lvm_cluster_request(CLVMD_CMD_UNLOCK,
-					     "",
-					     args, len,
-					     &unlock_response,
-					     &num_unlock_responses);
-		if (status) {
-			if (verbosity > 1)
-				fprintf(stderr, "cluster command failed: %s\n",
-					strerror(errno));
-		} else {
-			for (i = 0; i < num_unlock_responses; i++) {
-				if (unlock_response[i].status != 0) {
-					if (verbosity > 1)
-						fprintf(stderr,
-							"unlock on node %s failed: %s\n",
-							response[i].node,
-							strerror(unlock_response
-								 [0].status));
-				}
-			}
-		}
-		lvm_cluster_free_request(unlock_response);
-	}
-	lvm_cluster_free_request(response);
-
-	return 0;
-}

daemons/clvmd/libclvm.h

@@ -1,36 +0,0 @@
-/*
- * Copyright (C) 1997-2004 Sistina Software, Inc. All rights reserved.
- * Copyright (C) 2004 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 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
- */
-
-#ifndef _LIBCLVM_H
-#define _LIBCLVM_H
-
-typedef struct lvm_response {
-	char node[255];
-	char *response;
-	int status;
-	int len;
-
-} lvm_response_t;
-
-extern int lvm_cluster_request(char cmd, const char *node, void *data, int len,
-			       lvm_response_t ** response, int *num);
-extern int lvm_cluster_write(char cmd, char *node, void *data, int len);
-extern int lvm_cluster_free_request(lvm_response_t * response);
-
-/* The "high-level" API */
-extern int lvm_lock_for_cluster(char scope, char *name, int verbosity);
-extern int lvm_unlock_for_cluster(char scope, char *name, int verbosity);
-
-#endif

daemons/clvmd/lvm-functions.c

@@ -46,6 +46,7 @@
 
 static struct cmd_context *cmd = NULL;
 static struct hash_table *lv_hash = NULL;
+static pthread_mutex_t lv_hash_lock;
 
 struct lv_info {
 	int lock_id;
@@ -57,7 +58,9 @@ static int get_current_lock(char *resource)
 {
 	struct lv_info *lvi;
 
+	pthread_mutex_lock(&lv_hash_lock);
 	lvi = hash_lookup(lv_hash, resource);
+	pthread_mutex_unlock(&lv_hash_lock);
 	if (lvi) {
 		return lvi->lock_mode;
 	} else {
@@ -69,11 +72,14 @@ static int get_current_lock(char *resource)
 void unlock_all()
 {
 	struct hash_node *v;
+
+	pthread_mutex_lock(&lv_hash_lock);
 	hash_iterate(v, lv_hash) {
 		struct lv_info *lvi = hash_get_data(lv_hash, v);
 
 		sync_unlock(hash_get_key(lv_hash, v), lvi->lock_id);
 	}
+	pthread_mutex_unlock(&lv_hash_lock);
 }
 
 /* Gets a real lock and keeps the info in the hash table */
@@ -85,7 +91,9 @@ int hold_lock(char *resource, int mode, int flags)
 
 	flags &= LKF_NOQUEUE;	/* Only LKF_NOQUEUE is valid here */
 
+	pthread_mutex_lock(&lv_hash_lock);
 	lvi = hash_lookup(lv_hash, resource);
+	pthread_mutex_unlock(&lv_hash_lock);
 	if (lvi) {
 		/* Already exists - convert it */
 		status =
@@ -113,7 +121,9 @@ int hold_lock(char *resource, int mode, int flags)
 			DEBUGLOG("hold_lock. lock at %d failed: %s\n", mode,
 				 strerror(errno));
 		} else {
+		        pthread_mutex_lock(&lv_hash_lock);
 			hash_insert(lv_hash, resource, lvi);
+			pthread_mutex_unlock(&lv_hash_lock);
 		}
 		errno = saved_errno;
 	}
@@ -127,8 +137,9 @@ int hold_unlock(char *resource)
 	int status;
 	int saved_errno;
 
+	pthread_mutex_lock(&lv_hash_lock);
 	lvi = hash_lookup(lv_hash, resource);
-
+	pthread_mutex_unlock(&lv_hash_lock);
 	if (!lvi) {
 		DEBUGLOG("hold_unlock, lock not already held\n");
 		return 0;
@@ -137,7 +148,9 @@ int hold_unlock(char *resource)
 	status = sync_unlock(resource, lvi->lock_id);
 	saved_errno = errno;
 	if (!status) {
+	    	pthread_mutex_lock(&lv_hash_lock);
 		hash_remove(lv_hash, resource);
+		pthread_mutex_unlock(&lv_hash_lock);
 		free(lvi);
 	} else {
 		DEBUGLOG("hold_unlock. unlock failed(%d): %s\n", status,
@@ -185,7 +198,7 @@ static int do_activate_lv(char *resource, int mode)
 		return errno;
 
 	/* If it's suspended then resume it */
-	if (!lv_info_by_lvid(cmd, resource, &lvi))
+	if (!lv_info_by_lvid(cmd, resource, &lvi, 0))
 		return EIO;
 
 	if (lvi.suspended)
@@ -231,7 +244,7 @@ static int do_suspend_lv(char *resource)
 	}
 
 	/* Only suspend it if it exists */
-	if (!lv_info_by_lvid(cmd, resource, &lvi))
+	if (!lv_info_by_lvid(cmd, resource, &lvi, 0))
 		return EIO;
 
 	if (lvi.exists) {
@@ -350,7 +363,7 @@ int post_lock_lv(unsigned char command, unsigned char lock_flags,
 		if (oldmode == LKM_PWMODE) {
 			struct lvinfo lvi;
 
-			if (!lv_info_by_lvid(cmd, resource, &lvi))
+			if (!lv_info_by_lvid(cmd, resource, &lvi, 0))
 				return EIO;
 
 			if (lvi.exists) {
@@ -375,6 +388,44 @@ int do_check_lvm1(char *vgname)
 	return status == 1 ? 0 : EBUSY;
 }
 
+
+/* Only called at gulm startup. Drop any leftover VG or P_orphan locks
+   that might be hanging around if we died for any reason
+*/
+static void drop_vg_locks()
+{
+	char vg[128];
+	char line[255];
+	FILE *vgs =
+	    popen
+	    ("lvm pvs --nolocking --noheadings -o vg_name", "r");
+
+	sync_unlock("P_orphans", LCK_EXCL);
+
+	if (!vgs)
+		return;
+
+	while (fgets(line, sizeof(line), vgs)) {
+		char *vgend;
+		char *vgstart;
+
+		if (line[strlen(line)-1] == '\n')
+			line[strlen(line)-1] = '\0';
+
+		vgstart = line + strspn(line, " ");
+		vgend = vgstart + strcspn(vgstart, " ");
+		*vgend = '\0';
+
+		if (strncmp(vgstart, "WARNING:", 8) == 0)
+			continue;
+
+		sprintf(vg, "V_%s", vgstart);
+		sync_unlock(vg, LCK_EXCL);
+
+	}
+	fclose(vgs);
+}
+
 /*
  * Ideally, clvmd should be started before any LVs are active
  * but this may not be the case...
@@ -387,16 +438,18 @@ static void *get_initial_state()
 	char line[255];
 	FILE *lvs =
 	    popen
-	    ("/sbin/lvm lvs --nolocking --noheadings -o vg_uuid,lv_uuid,lv_attr",
+	    ("lvm lvs --nolocking --noheadings -o vg_uuid,lv_uuid,lv_attr",
 	     "r");
 
 	if (!lvs)
 		return NULL;
 
 	while (fgets(line, sizeof(line), lvs)) {
-		if (sscanf(line, "%s %s %s\n", vg, lv, flags) == 3) {
+	        if (sscanf(line, "%s %s %s\n", vg, lv, flags) == 3) {
+
 			/* States: s:suspended a:active S:dropped snapshot I:invalid snapshot */
-			if (flags[4] == 'a' || flags[4] == 's') {	/* is it active or suspended? */
+		        if (strlen(vg) == 38 &&                         /* is is a valid UUID ? */
+			    (flags[4] == 'a' || flags[4] == 's')) {	/* is it active or suspended? */
 				/* Convert hyphen-separated UUIDs into one */
 				memcpy(&uuid[0], &vg[0], 6);
 				memcpy(&uuid[6], &vg[7], 4);
@@ -423,14 +476,39 @@ static void *get_initial_state()
 	return NULL;
 }
 
+/* This checks some basic cluster-LVM configuration stuff */
+static void check_config()
+{
+	int locking_type;
+
+	locking_type = find_config_int(cmd->cft->root, "global/locking_type", 1);
+
+	if (locking_type == 3) /* compiled-in cluster support */
+		return;
+
+	if (locking_type == 2) { /* External library, check name */
+		const char *libname;
+
+		libname = find_config_str(cmd->cft->root, "global/locking_library",
+					  "");
+		if (strstr(libname, "liblvm2clusterlock.so"))
+			return;
+
+		log_error("Incorrect LVM locking library specified in lvm.conf, cluster operations may not work.");
+		return;
+	}
+	log_error("locking_type not set correctly in lvm.conf, cluster operations will not work.");
+}
+
 void init_lvhash()
 {
 	/* Create hash table for keeping LV locks & status */
 	lv_hash = hash_create(100);
+	pthread_mutex_init(&lv_hash_lock, NULL);
 }
 
 /* Called to initialise the LVM context of the daemon */
-int init_lvm(void)
+int init_lvm(int using_gulm)
 {
 	if (!(cmd = create_toolcontext(NULL))) {
 		log_error("Failed to allocate command context");
@@ -441,6 +519,13 @@ int init_lvm(void)
 	init_syslog(LOG_DAEMON);
 	init_debug(_LOG_ERR);
 
+	/* Check lvm.conf is setup for cluster-LVM */
+	check_config();
+
+	/* Remove any non-LV locks that may have been left around */
+	if (using_gulm)
+		drop_vg_locks();
+
 	get_initial_state();
 
 	return 1;

daemons/clvmd/lvm-functions.h

@@ -25,7 +25,7 @@ extern int do_lock_lv(unsigned char lock_cmd, unsigned char lock_flags,
 extern int post_lock_lv(unsigned char lock_cmd, unsigned char lock_flags,
 			char *resource);
 extern int do_check_lvm1(char *vgname);
-extern int init_lvm(void);
+extern int init_lvm(int using_gulm);
 extern void init_lvhash(void);
 
 extern int hold_unlock(char *resource);

daemons/clvmd/system-lv.c

@@ -42,7 +42,9 @@
 #include "list.h"
 #include "locking.h"
 #include "system-lv.h"
+#include "clvm.h"
 #include "clvmd-comms.h"
+#include "clvmd.h"
 #ifdef HAVE_CCS
 #include "ccs.h"
 #endif

daemons/clvmd/tcp-comms.c

@@ -2,6 +2,7 @@
 *******************************************************************************
 **
 **  Copyright (C) Sistina Software, Inc.  2002-2003  All rights reserved.
+**  Copyright (C) 2004-2005 Red Hat, Inc. All rights reserved.
 **
 *******************************************************************************
 ******************************************************************************/
@@ -34,7 +35,6 @@
 #include <netdb.h>
 #include <assert.h>
 
-#include "ccs.h"
 #include "clvm.h"
 #include "clvmd-comms.h"
 #include "clvmd.h"
@@ -47,43 +47,19 @@ static int listen_fd = -1;
 static int tcp_port;
 struct hash_table *sock_hash;
 
-static int get_tcp_port(int default_port);
 static int get_our_ip_address(char *addr, int *family);
 static int read_from_tcpsock(struct local_client *fd, char *buf, int len, char *csid,
 			     struct local_client **new_client);
 
 /* Called by init_cluster() to open up the listening socket */
-// TODO: IPv6 compat.
-int init_comms()
+int init_comms(unsigned short port)
 {
-    struct sockaddr *addr = NULL;
-    struct sockaddr_in addr4;
-    struct sockaddr_in6 addr6;
-    int    addr_len;
-    int    family;
-    char   address[MAX_CSID_LEN];
+    struct sockaddr_in6 addr;
 
     sock_hash = hash_create(100);
-    tcp_port = get_tcp_port(DEFAULT_TCP_PORT);
+    tcp_port = port ? port : DEFAULT_TCP_PORT;
 
-    /* Get IP address and IP type */
-    get_our_ip_address(address, &family);
-    if (family == AF_INET)
-    {
-	memcpy(&addr4.sin_addr, addr, sizeof(struct in_addr));
-	addr = (struct sockaddr *)&addr4;
-	addr4.sin_port = htons(tcp_port);
-	addr_len = sizeof(addr4);
-    }
-    else
-    {
-	memcpy(&addr6.sin6_addr, addr, sizeof(struct in6_addr));
-	addr = (struct sockaddr *)&addr6;
-	addr6.sin6_port = htons(tcp_port);
-	addr_len = sizeof(addr6);
-    }
-
-    listen_fd = socket(family, SOCK_STREAM, 0);
+    listen_fd = socket(AF_INET6, SOCK_STREAM, 0);
 
     if (listen_fd < 0)
     {
@@ -95,11 +71,13 @@ int init_comms()
 	setsockopt(listen_fd, SOL_SOCKET, SO_REUSEADDR, &one, sizeof(int));
     }
 
-    addr->sa_family = family;
+    memset(&addr, 0, sizeof(addr)); // Bind to INADDR_ANY
+    addr.sin6_family = AF_INET6;
+    addr.sin6_port = htons(tcp_port);
 
-    if (bind(listen_fd, addr, addr_len) < 0)
+    if (bind(listen_fd, (struct sockaddr *)&addr, sizeof(addr)) < 0)
     {
-	DEBUGLOG("Can't bind to port\n");
+	DEBUGLOG("Can't bind to port: %s\n", strerror(errno));
 	syslog(LOG_ERR, "Can't bind to port %d, is clvmd already running ?", tcp_port);
 	close(listen_fd);
 	return -1;
@@ -107,6 +85,9 @@ int init_comms()
 
     listen(listen_fd, 5);
 
+    /* Set Close-on-exec */
+    fcntl(listen_fd, F_SETFD, 1);
+
     return 0;
 }
 
@@ -119,19 +100,19 @@ void tcp_remove_client(char *csid)
        job of clvmd.c whch will do the job when it notices the
        other end has gone. We just need to remove the client(s) from
        the hash table so we don't try to use it for sending any more */
-    client = hash_lookup_binary(sock_hash, csid, MAX_CSID_LEN);
+    client = hash_lookup_binary(sock_hash, csid, GULM_MAX_CSID_LEN);
     if (client)
     {
-	hash_remove_binary(sock_hash, csid, MAX_CSID_LEN);
+	hash_remove_binary(sock_hash, csid, GULM_MAX_CSID_LEN);
     }
 
     /* Look for a mangled one too */
     csid[0] ^= 0x80;
 
-    client = hash_lookup_binary(sock_hash, csid, MAX_CSID_LEN);
+    client = hash_lookup_binary(sock_hash, csid, GULM_MAX_CSID_LEN);
     if (client)
     {
-	hash_remove_binary(sock_hash, csid, MAX_CSID_LEN);
+	hash_remove_binary(sock_hash, csid, GULM_MAX_CSID_LEN);
     }
 
     /* Put it back as we found it */
@@ -142,7 +123,7 @@ int alloc_client(int fd, char *csid, struct local_client **new_client)
 {
     struct local_client *client;
 
-    DEBUGLOG("alloc_client %d csid = [%d.%d.%d.%d]\n", fd,csid[0],csid[1],csid[2],csid[3]);
+    DEBUGLOG("alloc_client %d csid = %s\n", fd, print_csid(csid));
 
     /* Create a local_client and return it */
     client = malloc(sizeof(struct local_client));
@@ -160,7 +141,7 @@ int alloc_client(int fd, char *csid, struct local_client **new_client)
 	*new_client = client;
 
     /* Add to our list of node sockets */
-    if (hash_lookup_binary(sock_hash, csid, MAX_CSID_LEN))
+    if (hash_lookup_binary(sock_hash, csid, GULM_MAX_CSID_LEN))
     {
 	DEBUGLOG("alloc_client mangling CSID for second connection\n");
 	/* This is a duplicate connection but we can't close it because
@@ -173,7 +154,7 @@ int alloc_client(int fd, char *csid, struct local_client **new_client)
 
         /* If it still exists then kill the connection as we should only
            ever have one incoming connection from each node */
-        if (hash_lookup_binary(sock_hash, csid, MAX_CSID_LEN))
+        if (hash_lookup_binary(sock_hash, csid, GULM_MAX_CSID_LEN))
         {
 	    DEBUGLOG("Multiple incoming connections from node\n");
             syslog(LOG_ERR, " Bogus incoming connection from %d.%d.%d.%d\n", csid[0],csid[1],csid[2],csid[3]);
@@ -183,26 +164,26 @@ int alloc_client(int fd, char *csid, struct local_client **new_client)
             return -1;
         }
     }
-    hash_insert_binary(sock_hash, csid, MAX_CSID_LEN, client);
+    hash_insert_binary(sock_hash, csid, GULM_MAX_CSID_LEN, client);
 
     return 0;
 }
 
-int get_main_cluster_fd()
+int get_main_gulm_cluster_fd()
 {
     return listen_fd;
 }
 
 
 /* Read on main comms (listen) socket, accept it */
-int cluster_fd_callback(struct local_client *fd, char *buf, int len, char *csid,
+int cluster_fd_gulm_callback(struct local_client *fd, char *buf, int len, char *csid,
 			struct local_client **new_client)
 {
     int newfd;
-    struct sockaddr_in addr;
+    struct sockaddr_in6 addr;
     socklen_t addrlen = sizeof(addr);
     int status;
-    char name[MAX_CLUSTER_MEMBER_NAME_LEN];
+    char name[GULM_MAX_CLUSTER_MEMBER_NAME_LEN];
 
     DEBUGLOG("cluster_fd_callback\n");
     *new_client = NULL;
@@ -218,22 +199,20 @@ int cluster_fd_callback(struct local_client *fd, char *buf, int len, char *csid,
 
     /* Check that the client is a member of the cluster
        and reject if not.
-       // FIXME: IPv4 specific
     */
-    if (name_from_csid((char *)&addr.sin_addr.s_addr, name) < 0)
+    if (gulm_name_from_csid((char *)&addr.sin6_addr, name) < 0)
     {
-	char *ip = (char *)&addr.sin_addr.s_addr;
-	syslog(LOG_ERR, "Got connect from non-cluster node %d.%d.%d.%d\n",
-	       ip[0], ip[1], ip[2], ip[3]);
-	DEBUGLOG("Got connect from non-cluster node %d.%d.%d.%d\n",
-		 ip[0], ip[1], ip[2], ip[3]);
+	syslog(LOG_ERR, "Got connect from non-cluster node %s\n",
+	       print_csid((char *)&addr.sin6_addr));
+	DEBUGLOG("Got connect from non-cluster node %s\n",
+		 print_csid((char *)&addr.sin6_addr));
 	close(newfd);
 
 	errno = EAGAIN;
 	return -1;
     }
 
-    status = alloc_client(newfd, (char *)&addr.sin_addr.s_addr, new_client);
+    status = alloc_client(newfd, (char *)&addr.sin6_addr, new_client);
     if (status)
     {
 	DEBUGLOG("cluster_fd_callback, alloc_client failed, status = %d\n", status);
@@ -250,18 +229,39 @@ int cluster_fd_callback(struct local_client *fd, char *buf, int len, char *csid,
 static int read_from_tcpsock(struct local_client *client, char *buf, int len, char *csid,
 			     struct local_client **new_client)
 {
-    struct sockaddr_in addr;
+    struct sockaddr_in6 addr;
     socklen_t slen = sizeof(addr);
+    struct clvm_header *header = (struct clvm_header *)buf;
     int status;
+    uint32_t arglen;
 
     DEBUGLOG("read_from_tcpsock fd %d\n", client->fd);
     *new_client = NULL;
 
     /* Get "csid" */
     getpeername(client->fd, (struct sockaddr *)&addr, &slen);
-    memcpy(csid, &addr.sin_addr.s_addr, MAX_CSID_LEN);
+    memcpy(csid, &addr.sin6_addr, GULM_MAX_CSID_LEN);
 
-    status = read(client->fd, buf, len);
+    /* Read just the header first, then get the rest if there is any.
+     * Stream sockets, sigh.
+     */
+    status = read(client->fd, buf, sizeof(struct clvm_header));
+    if (status > 0)
+    {
+	    int status2;
+
+	    arglen = ntohl(header->arglen);
+
+	    /* Get the rest */
+	    if (arglen && arglen < GULM_MAX_CLUSTER_MESSAGE)
+	    {
+		    status2 = read(client->fd, buf+status, arglen);
+		    if (status2 > 0)
+			    status += status2;
+		    else
+			    status = status2;
+	    }
+    }
 
     DEBUGLOG("read_from_tcpsock, status = %d(errno = %d)\n", status, errno);
 
@@ -270,30 +270,35 @@ static int read_from_tcpsock(struct local_client *client, char *buf, int len, ch
     if (status == 0 ||
 	(status < 0 && errno != EAGAIN && errno != EINTR))
     {
-	char remcsid[MAX_CSID_LEN];
+	char remcsid[GULM_MAX_CSID_LEN];
 
-	memcpy(remcsid, csid, MAX_CSID_LEN);
+	memcpy(remcsid, csid, GULM_MAX_CSID_LEN);
 	close(client->fd);
 
 	/* If the csid was mangled, then make sure we remove the right entry */
 	if (client->bits.net.flags)
 	    remcsid[0] ^= 0x80;
-	hash_remove_binary(sock_hash, remcsid, MAX_CSID_LEN);
+	hash_remove_binary(sock_hash, remcsid, GULM_MAX_CSID_LEN);
 
 	/* Tell cluster manager layer */
 	add_down_node(remcsid);
     }
+    else {
+	    gulm_add_up_node(csid);
+	    /* Send it back to clvmd */
+	    process_message(client, buf, len, csid);
+    }
     return status;
 }
 
-static int connect_csid(char *csid, struct local_client **newclient)
+int gulm_connect_csid(char *csid, struct local_client **newclient)
 {
     int fd;
-    struct sockaddr_in addr;
+    struct sockaddr_in6 addr;
     int status;
 
     DEBUGLOG("Connecting socket\n");
-    fd = socket(PF_INET, SOCK_STREAM, 0);
+    fd = socket(PF_INET6, SOCK_STREAM, 0);
 
     if (fd < 0)
     {
@@ -301,19 +306,28 @@ static int connect_csid(char *csid, struct local_client **newclient)
 	return -1;
     }
 
-    addr.sin_family = AF_INET;
-    memcpy(&addr.sin_addr.s_addr, csid, MAX_CSID_LEN);
-    addr.sin_port = htons(tcp_port);
+    addr.sin6_family = AF_INET6;
+    memcpy(&addr.sin6_addr, csid, GULM_MAX_CSID_LEN);
+    addr.sin6_port = htons(tcp_port);
 
     DEBUGLOG("Connecting socket %d\n", fd);
-    if (connect(fd, (struct sockaddr *)&addr, sizeof(struct sockaddr_in)) < 0)
+    if (connect(fd, (struct sockaddr *)&addr, sizeof(struct sockaddr_in6)) < 0)
     {
-	syslog(LOG_ERR, "Unable to connect to remote node: %m");
+	/* "Connection refused" is "normal" because clvmd may not yet be running
+	 * on that node.
+	 */
+	if (errno != ECONNREFUSED)
+	{
+	    syslog(LOG_ERR, "Unable to connect to remote node: %m");
+	}
 	DEBUGLOG("Unable to connect to remote node: %s\n", strerror(errno));
 	close(fd);
 	return -1;
     }
 
+    /* Set Close-on-exec */
+    fcntl(fd, F_SETFD, 1);
+
     status = alloc_client(fd, csid, newclient);
     if (status)
 	close(fd);
@@ -321,7 +335,7 @@ static int connect_csid(char *csid, struct local_client **newclient)
 	add_client(*newclient);
 
     /* If we can connect to it, it must be running a clvmd */
-    add_up_node(csid);
+    gulm_add_up_node(csid);
     return status;
 }
 
@@ -330,21 +344,21 @@ static int tcp_send_message(void *buf, int msglen, unsigned char *csid, const ch
 {
     int status;
     struct local_client *client;
-    char ourcsid[MAX_CSID_LEN];
+    char ourcsid[GULM_MAX_CSID_LEN];
 
     assert(csid);
 
-    DEBUGLOG("tcp_send_message, csid = [%d.%d.%d.%d], msglen = %d\n", csid[0],csid[1],csid[2],csid[3], msglen);
+    DEBUGLOG("tcp_send_message, csid = %s, msglen = %d\n", print_csid(csid), msglen);
 
     /* Don't connect to ourself */
-    get_our_csid(ourcsid);
-    if (memcmp(csid, ourcsid, MAX_CSID_LEN) == 0)
+    get_our_gulm_csid(ourcsid);
+    if (memcmp(csid, ourcsid, GULM_MAX_CSID_LEN) == 0)
 	return msglen;
 
-    client = hash_lookup_binary(sock_hash, csid, MAX_CSID_LEN);
+    client = hash_lookup_binary(sock_hash, csid, GULM_MAX_CSID_LEN);
     if (!client)
     {
-	status = connect_csid(csid, &client);
+	status = gulm_connect_csid(csid, &client);
 	if (status)
 	    return -1;
     }
@@ -354,7 +368,7 @@ static int tcp_send_message(void *buf, int msglen, unsigned char *csid, const ch
 }
 
 
-int cluster_send_message(void *buf, int msglen, char *csid, const char *errtext)
+int gulm_cluster_send_message(void *buf, int msglen, char *csid, const char *errtext)
 {
     int status=0;
 
@@ -364,7 +378,7 @@ int cluster_send_message(void *buf, int msglen, char *csid, const char *errtext)
     if (!csid)
     {
 	void *context = NULL;
-	char loop_csid[MAX_CSID_LEN];
+	char loop_csid[GULM_MAX_CSID_LEN];
 
 	/* Loop round all gulm-known nodes */
 	while (get_next_node_csid(&context, loop_csid))
@@ -383,62 +397,24 @@ int cluster_send_message(void *buf, int msglen, char *csid, const char *errtext)
     return status;
 }
 
-static int get_tcp_port(int default_port)
-{
-    int ccs_handle;
-    int port = default_port;
-    char *portstr;
-
-    ccs_handle = ccs_connect();
-    if (ccs_handle)
-    {
-	return port;
-    }
-
-    if (!ccs_get(ccs_handle, "//clvm/@port", &portstr))
-    {
-	port = atoi(portstr);
-	free(portstr);
-
-	if (port <= 0 && port >= 65536)
-	    port = default_port;
-    }
-    ccs_disconnect(ccs_handle);
-
-    DEBUGLOG("Using port %d for communications\n", port);
-    return port;
-}
-
 /* To get our own IP address we get the locally bound address of the
    socket that's talking to GULM in the assumption(eek) that it will
    be on the "right" network in a multi-homed system */
 static int get_our_ip_address(char *addr, int *family)
 {
-    /* Use a sockaddr_in6 to make sure it's big enough */
-    struct sockaddr_in6 saddr;
-    int socklen = sizeof(saddr);
+	struct utsname info;
+
+	uname(&info);
+	get_ip_address(info.nodename, addr);
 
-    if (!getsockname(gulm_fd(), (struct sockaddr *)&saddr, &socklen))
-    {
-	if (saddr.sin6_family == AF_INET6)
-	{
-	    memcpy(addr, &saddr.sin6_addr, sizeof(saddr.sin6_addr));
-	}
-	else
-	{
-	    struct sockaddr_in *sin4 = (struct sockaddr_in *)&saddr;
-	    memcpy(addr, &sin4->sin_addr, sizeof(sin4->sin_addr));
-	}
 	return 0;
-    }
-    return -1;
 }
 
 /* Public version of above for those that don't care what protocol
    we're using */
-void get_our_csid(char *csid)
+void get_our_gulm_csid(char *csid)
 {
-    static char our_csid[MAX_CSID_LEN];
+    static char our_csid[GULM_MAX_CSID_LEN];
     static int got_csid = 0;
 
     if (!got_csid)
@@ -451,7 +427,15 @@ void get_our_csid(char *csid)
 	    got_csid = 1;
 	}
     }
-    memcpy(csid, our_csid, MAX_CSID_LEN);
+    memcpy(csid, our_csid, GULM_MAX_CSID_LEN);
+}
+
+static void map_v4_to_v6(struct in_addr *ip4, struct in6_addr *ip6)
+{
+   ip6->s6_addr32[0] = 0;
+   ip6->s6_addr32[1] = 0;
+   ip6->s6_addr32[2] = htonl(0xffff);
+   ip6->s6_addr32[3] = ip4->s_addr;
 }
 
 /* Get someone else's IP address from DNS */
@@ -459,7 +443,7 @@ int get_ip_address(char *node, char *addr)
 {
     struct hostent *he;
 
-    memset(addr, 0, MAX_CSID_LEN);
+    memset(addr, 0, GULM_MAX_CSID_LEN);
 
     // TODO: what do we do about multi-homed hosts ???
     // CCSs ip_interfaces solved this but some bugger removed it.
@@ -467,14 +451,29 @@ int get_ip_address(char *node, char *addr)
     /* Try IPv6 first. The man page for gethostbyname implies that
        it will lookup ip6 & ip4 names, but it seems not to */
     he = gethostbyname2(node, AF_INET6);
-    if (!he)
+    if (he)
+    {
+	memcpy(addr, he->h_addr_list[0],
+	       he->h_length);
+    }
+    else
+    {
 	he = gethostbyname2(node, AF_INET);
-    if (!he)
-	return -1;
-
-    /* For IPv4 address just use the lower 4 bytes */
-    memcpy(&addr, he->h_addr_list[0],
-	   he->h_length);
+	if (!he)
+	    return -1;
+	map_v4_to_v6((struct in_addr *)he->h_addr_list[0], (struct in6_addr *)addr);
+    }
 
     return 0;
 }
+
+char *print_csid(char *csid)
+{
+    static char buf[128];
+    int *icsid = (int *)csid;
+
+    sprintf(buf, "[%x.%x.%x.%x]",
+	    icsid[0],icsid[1],icsid[2],icsid[3]);
+
+    return buf;
+}

daemons/clvmd/tcp-comms.h

@@ -1,7 +1,13 @@
 #include <netinet/in.h>
 
-#define MAX_CLUSTER_MESSAGE 1600
-#define MAX_CSID_LEN sizeof(struct in6_addr)
-#define MAX_CLUSTER_MEMBER_NAME_LEN 128
+#define GULM_MAX_CLUSTER_MESSAGE 1600
+#define GULM_MAX_CSID_LEN sizeof(struct in6_addr)
+#define GULM_MAX_CLUSTER_MEMBER_NAME_LEN 128
 
-extern int init_comms(void);
+extern int init_comms(unsigned short);
+extern char *print_csid(char *);
+int get_main_gulm_cluster_fd(void);
+int cluster_fd_gulm_callback(struct local_client *fd, char *buf, int len, char *csid, struct local_client **new_client);
+int gulm_cluster_send_message(void *buf, int msglen, char *csid, const char *errtext);
+void get_our_gulm_csid(char *csid);
+int gulm_connect_csid(char *csid, struct local_client **newclient);

doc/example.conf

@@ -27,6 +27,12 @@ devices {
     # the device will be accepted or rejected (ignored).  Devices that
     # don't match any patterns are accepted.
 
+    # Be careful if there there are symbolic links or multiple filesystem 
+    # entries for the same device as each name is checked separately against
+    # the list of patterns.  The effect is that if any name matches any 'a'
+    # pattern, the device is accepted; otherwise if any name matches any 'r'
+    # pattern it is rejected; otherwise it is accepted.
+
     # Remember to run vgscan after you change this parameter to ensure 
     # that the cache file gets regenerated (see below).
 
@@ -252,11 +258,10 @@ activation {
 #
 # metadata {
     # Default number of copies of metadata to hold on each PV.  0, 1 or 2.
-    # It's best to leave this at 2.
-    # You might want to override it from the command line with 0 or 1 
+    # You might want to override it from the command line with 0 
     # when running pvcreate on new PVs which are to be added to large VGs.
 
-    # pvmetadatacopies = 2
+    # pvmetadatacopies = 1
 
     # Approximate default size of on-disk metadata areas in sectors.
     # You should increase this if you have large volume groups or

doc/tagging.txt

@@ -0,0 +1,165 @@
+Tagging aims
+============
+  1) Ability to attach an unordered list of tags to LVM metadata objects.
+  2) Ability to add or remove tags easily.
+  3) Ability to select LVM objects for processing according to presence/absence
+     of specific tags.
+  4) Ability to control through the config file which VGs/LVs are activated 
+     on different machines using names or tags.
+  5) Ability to overlay settings from different config files e.g. override
+     some settings in a global config file locally.
+
+Clarifications
+==============
+  1) Tag character set: A-Za-z0-9_+.- 
+     Can't start with hyphen & max length is 128 (NAME_LEN).
+  2) LVM object types that can be tagged:
+       VG, LV, LV segment
+       PV - tags are stored in VG metadata so disappear when PV becomes orphaned
+     Snapshots can't be tagged, but their origin may be.
+  3) A tag can be used in place of any command line LVM object reference that
+     accepts (a) a list of objects; or (b) a single object as long as the
+     tag expands to a single object.  This is not supported everywhere yet.
+     Duplicate arguments in a list after argument expansion may get removed 
+     retaining the first copy of each argument.
+  4) Wherever there may be ambiguity of argument type, a tag must be prefixed 
+     by '@'; elsewhere an '@' prefix is optional.
+  5) LVM1 objects cannot be tagged, as the disk format doesn't support it.
+  6) Tags can be added or removed with --addtag or --deltag.
+
+Config file Extensions
+======================
+  To define host tags in config file:
+
+  tags {
+  	# Set a tag with the hostname
+	hosttags = 1
+
+	tag1 { }
+
+  	tag2 {
+		# If no exact match, tag is not set.
+		host_list = [ "hostname", "dbase" ]
+	}
+  }
+
+Activation config file example
+==============================
+  activation {
+      volume_list = [ "vg1/lvol0", "@database" ]
+  }
+
+  Matches against vgname, vgname/lvname or @tag set in *metadata*.
+  @* matches exactly against *any* tag set on the host.
+  The VG or LV only gets activated if a metadata tag matches.
+  The default if there is no match is not to activate.
+  If volume_list is not present and any tags are defined on the host 
+  then it only activates if a host tag matches a metadata tag.
+  If volume_list is not present and no tags are defined on the host 
+  then it does activate.
+
+Multiple config files
+=====================
+  (a) lvm.conf
+  (b) lvm_<host_tag>.conf
+
+  At startup, load lvm.conf.
+  Process tag settings.
+  If any host tags were defined, load lvm_tag.conf for each tag, if present.
+
+  When searching for a specific config file entry, search order is (b)
+  then (a), stopping at the first match.  
+  Within (b) use reverse order tags got set, so file for last tag set is
+  searched first.
+  New tags set in (b) *do* trigger additional config file loads. 
+
+Usage Examples
+==============
+  1) Simple activation control via metadata with static config files
+
+  lvm.conf:  (Identical on every machine - global settings)
+    tags {
+      hostname_tags = 1
+    }
+
+  From any machine in the cluster, add db1 to the list of machines that
+  activate vg1/lvol2:
+
+  lvchange --tag @db1 vg1/lvol2
+  (followed by lvchange -ay to actually activate it)
+
+
+  2) Multiple hosts.  
+
+    Activate vg1 only on the database hosts, db1 and db2.
+    Activate vg2 only on the fileserver host fs1.
+    Activate nothing initially on the fileserver backup host fsb1, but be
+    prepared for it to take over from fs1.
+
+  Option (i) - centralised admin, static configuration replicated between hosts  
+    # Add @database tag to vg1's metadata
+    vgchange --tag @database vg1
+
+    # Add @fileserver tag to vg2's metadata
+    vgchange --tag @fileserver vg2
+
+    lvm.conf:  (Identical on every machine)
+      tags {
+        database {
+          host_list = [ "db1", "db2" ]
+        }
+        fileserver {
+	  host_list = [ "fs1" ]
+        }
+        fileserverbackup {
+          host_list = [ "fsb1" ]
+        }
+      }
+
+      activation {
+        # Only activate if host has a tag that matches a metadata tag
+        volume_list = [ "@*" ]
+      }
+  
+  In the event of the fileserver host going down, vg2 can be brought up
+  on fsb1 by running *on any node* 'vgchange --tag @fileserverbackup vg2'
+  followed by 'vgchange -ay vg2'
+  
+  
+  Option (ii) - localised admin & configuation
+  (i.e. each host holds *locally* which classes of volumes to activate)
+    # Add @database tag to vg1's metadata
+    vgchange --tag @database vg1
+  
+    # Add @fileserver tag to vg2's metadata
+    vgchange --tag @fileserver vg2
+  
+    lvm.conf:  (Identical on every machine - global settings)
+      tags {
+        hosttags = 1
+      }
+  
+    lvm_db1.conf: (only needs to be on db1 - could be symlink to lvm_db.conf)
+      activation {
+        volume_list = [ "@database" ]
+      }
+  
+    lvm_db2.conf: (only needs to be on db2 - could be symlink to lvm_db.conf)
+      activation {
+        volume_list = [ "@database" ]
+      }
+  
+    lvm_fs1.conf: (only needs to be on fs1 - could be symlink to lvm_fs.conf)
+      activation {
+        volume_list = [ "@fileserver" ]
+      }
+  
+    If fileserver goes down, to bring a spare machine fsb1 in as fileserver,
+    create lvm_fsb1.conf on fsb1 (or symlink to lvm_fs.conf):
+
+      activation {
+        volume_list = [ "@fileserver" ]
+      }
+
+    and run 'vgchange -ay vg2' or 'vgchange -ay @fileserver'
+

include/.symlinks

@@ -47,3 +47,4 @@
 ../lib/report/report.h
 ../lib/uuid/uuid.h
 ../po/pogen.h
+../tools/version.h

lib/Makefile.in

@@ -43,6 +43,7 @@ SOURCES =\
 	datastruct/str_list.c \
 	device/dev-cache.c \
 	device/dev-io.c \
+	device/dev-md.c \
 	device/device.c \
 	display/display.c \
 	error/errseg.c \
@@ -75,6 +76,7 @@ SOURCES =\
 	misc/crc.c \
 	misc/lvm-file.c \
 	misc/lvm-string.c \
+	mm/dbg_malloc.c \
 	mm/memlock.c \
 	mm/pool.c \
 	regex/matcher.c \
@@ -120,10 +122,6 @@ ifeq ("@MIRRORS@", "internal")
   SOURCES += mirror/mirrored.c
 endif
 
-ifeq ("@DEBUG@", "yes")
-  SOURCES += mm/dbg_malloc.c
-endif
-
 ifeq ("@DEVMAPPER@", "yes")
   SOURCES +=\
 	activate/dev_manager.c \

lib/activate/activate.c

@@ -78,12 +78,13 @@ int target_present(const char *target_name)
 {
 	return 0;
 }
-int lv_info(const struct logical_volume *lv, struct lvinfo *info)
+int lv_info(const struct logical_volume *lv, struct lvinfo *info,
+	    int with_open_count)
 {
 	return 0;
 }
 int lv_info_by_lvid(struct cmd_context *cmd, const char *lvid_s,
-		    struct lvinfo *info)
+		    struct lvinfo *info, int with_open_count)
 {
 	return 0;
 }
@@ -333,7 +334,7 @@ int target_present(const char *target_name)
  * Returns 1 if info structure populated, else 0 on failure.
  */
 static int _lv_info(const struct logical_volume *lv, int mknodes,
-		    struct lvinfo *info)
+		    struct lvinfo *info, int with_open_count)
 {
 	int r;
 	struct dev_manager *dm;
@@ -347,7 +348,7 @@ static int _lv_info(const struct logical_volume *lv, int mknodes,
 		return 0;
 	}
 
-	if (!(r = dev_manager_info(dm, lv, mknodes, &dminfo)))
+	if (!(r = dev_manager_info(dm, lv, mknodes, with_open_count, &dminfo)))
 		stack;
 
 	info->exists = dminfo.exists;
@@ -361,20 +362,21 @@ static int _lv_info(const struct logical_volume *lv, int mknodes,
 	return r;
 }
 
-int lv_info(const struct logical_volume *lv, struct lvinfo *info)
+int lv_info(const struct logical_volume *lv, struct lvinfo *info,
+	    int with_open_count)
 {
-	return _lv_info(lv, 0, info);
+	return _lv_info(lv, 0, info, with_open_count);
 }
 
 int lv_info_by_lvid(struct cmd_context *cmd, const char *lvid_s,
-		    struct lvinfo *info)
+		    struct lvinfo *info, int with_open_count)
 {
 	struct logical_volume *lv;
 
 	if (!(lv = lv_from_lvid(cmd, lvid_s)))
 		return 0;
 
-	return _lv_info(lv, 0, info);
+	return _lv_info(lv, 0, info, with_open_count);
 }
 
 /*
@@ -412,7 +414,7 @@ int lv_mirror_percent(struct logical_volume *lv, int wait, float *percent,
 	if (!activation())
 		return 0;
 
-	if (!lv_info(lv, &info)) {
+	if (!lv_info(lv, &info, 0)) {
 		stack;
 		return 0;
 	}
@@ -437,7 +439,7 @@ static int _lv_active(struct logical_volume *lv)
 {
 	struct lvinfo info;
 
-	if (!lv_info(lv, &info)) {
+	if (!lv_info(lv, &info, 0)) {
 		stack;
 		return -1;
 	}
@@ -449,7 +451,7 @@ static int _lv_open_count(struct logical_volume *lv)
 {
 	struct lvinfo info;
 
-	if (!lv_info(lv, &info)) {
+	if (!lv_info(lv, &info, 1)) {
 		stack;
 		return -1;
 	}
@@ -566,7 +568,7 @@ static int _lv_suspend(struct cmd_context *cmd, const char *lvid_s,
 		return 1;
 	}
 
-	if (!lv_info(lv, &info)) {
+	if (!lv_info(lv, &info, 0)) {
 		stack;
 		return 0;
 	}
@@ -612,7 +614,7 @@ static int _lv_resume(struct cmd_context *cmd, const char *lvid_s,
 		return 1;
 	}
 
-	if (!lv_info(lv, &info)) {
+	if (!lv_info(lv, &info, 0)) {
 		stack;
 		return 0;
 	}
@@ -657,7 +659,7 @@ int lv_deactivate(struct cmd_context *cmd, const char *lvid_s)
 		return 1;
 	}
 
-	if (!lv_info(lv, &info)) {
+	if (!lv_info(lv, &info, 1)) {
 		stack;
 		return 0;
 	}
@@ -726,7 +728,7 @@ static int _lv_activate(struct cmd_context *cmd, const char *lvid_s, int filter)
 		return 1;
 	}
 
-	if (!lv_info(lv, &info)) {
+	if (!lv_info(lv, &info, 0)) {
 		stack;
 		return 0;
 	}
@@ -765,7 +767,7 @@ int lv_mknodes(struct cmd_context *cmd, const struct logical_volume *lv)
 		return r;
 	}
 
-	if (!_lv_info(lv, 1, &info)) {
+	if (!_lv_info(lv, 1, &info, 0)) {
 		stack;
 		return 0;
 	}

lib/activate/activate.h

@@ -55,9 +55,10 @@ int lv_mknodes(struct cmd_context *cmd, const struct logical_volume *lv);
 /*
  * Returns 1 if info structure has been populated, else 0.
  */
-int lv_info(const struct logical_volume *lv, struct lvinfo *info);
+int lv_info(const struct logical_volume *lv, struct lvinfo *info,
+	    int with_open_count);
 int lv_info_by_lvid(struct cmd_context *cmd, const char *lvid_s,
-		    struct lvinfo *info);
+		    struct lvinfo *info, int with_open_count);
 
 /*
  * Returns 1 if activate_lv has been set: 1 = activate; 0 = don't.

lib/activate/dev_manager.c

@@ -211,7 +211,8 @@ static struct dm_task *_setup_task(const char *name, const char *uuid,
 }
 
 static int _info_run(const char *name, const char *uuid, struct dm_info *info,
-		     int mknodes, struct pool *mem, char **uuid_out)
+		     int mknodes, int with_open_count, struct pool *mem,
+		     char **uuid_out)
 {
 	int r = 0;
 	struct dm_task *dmt;
@@ -225,6 +226,10 @@ static int _info_run(const char *name, const char *uuid, struct dm_info *info,
 		return 0;
 	}
 
+	if (!with_open_count)
+		if (!dm_task_no_open_count(dmt))
+			log_error("Failed to disable open_count");
+
 	if (!dm_task_run(dmt)) {
 		stack;
 		goto out;
@@ -250,14 +255,17 @@ static int _info_run(const char *name, const char *uuid, struct dm_info *info,
 }
 
 static int _info(const char *name, const char *uuid, int mknodes,
-		 struct dm_info *info, struct pool *mem, char **uuid_out)
+		 int with_open_count, struct dm_info *info,
+		 struct pool *mem, char **uuid_out)
 {
 	if (!mknodes && uuid && *uuid &&
-	    _info_run(NULL, uuid, info, 0, mem, uuid_out) && info->exists)
+	    _info_run(NULL, uuid, info, 0, with_open_count, mem, uuid_out) &&
+	    	      info->exists)
 		return 1;
 
 	if (name)
-		return _info_run(name, NULL, info, mknodes, mem, uuid_out);
+		return _info_run(name, NULL, info, mknodes, with_open_count,
+				 mem, uuid_out);
 
 	return 0;
 }
@@ -279,6 +287,9 @@ static int _status_run(const char *name, const char *uuid,
 		return 0;
 	}
 
+	if (!dm_task_no_open_count(dmt))
+		log_error("Failed to disable open_count");
+
 	if (!dm_task_run(dmt)) {
 		stack;
 		goto out;
@@ -357,6 +368,9 @@ static int _percent_run(struct dev_manager *dm, const char *name,
 		return 0;
 	}
 
+	if (!dm_task_no_open_count(dmt))
+		log_error("Failed to disable open_count");
+
 	if (!dm_task_run(dmt)) {
 		stack;
 		goto out;
@@ -460,6 +474,9 @@ static int _rename(struct dev_manager *dm, struct dev_layer *dl, char *newname)
 		goto out;
 	}
 
+	if (!dm_task_no_open_count(dmt))
+		log_error("Failed to disable open_count");
+
 	if (!(r = dm_task_run(dmt))) {
 		log_error("Couldn't rename device '%s'.", dl->name);
 		goto out;
@@ -488,6 +505,9 @@ static int _suspend_or_resume(const char *name, action_t suspend)
 		return 0;
 	}
 
+	if (!dm_task_no_open_count(dmt))
+		log_error("Failed to disable open_count");
+
 	if (!(r = dm_task_run(dmt)))
 		log_error("Couldn't %s device '%s'", sus ? "suspend" : "resume",
 			  name);
@@ -579,6 +599,9 @@ static int _load(struct dev_manager *dm, struct dev_layer *dl, int task)
 			log_very_verbose("Activating %s read-only", dl->name);
 	}
 
+	if (!dm_task_no_open_count(dmt))
+		log_error("Failed to disable open_count");
+
 	if (!(r = dm_task_run(dmt))) {
 		log_error("Couldn't load device '%s'.", dl->name);
 		if ((dl->lv->minor >= 0 || dl->lv->major >= 0) &&
@@ -635,6 +658,9 @@ static int _remove(struct dev_layer *dl)
 		return 0;
 	}
 
+	if (!dm_task_no_open_count(dmt))
+		log_error("Failed to disable open_count");
+
 	/* Suppress error message if it's still in use - we'll log it later */
 	log_suppress(1);
 
@@ -852,6 +878,7 @@ static int _populate_snapshot(struct dev_manager *dm,
 	struct snapshot *s;
 	struct dev_layer *dlo, *dlc;
 	char devbufo[10], devbufc[10];
+	uint64_t size;
 
 	if (!(s = find_cow(dl->lv))) {
 		log_error("Couldn't find snapshot for '%s'.", dl->lv->name);
@@ -899,10 +926,10 @@ static int _populate_snapshot(struct dev_manager *dm,
 		return 0;
 	}
 
-	log_debug("Adding target: 0 %" PRIu64 " snapshot %s",
-		  s->origin->size, params);
-	if (!dm_task_add_target
-	    (dmt, UINT64_C(0), s->origin->size, "snapshot", params)) {
+	size = (uint64_t) s->le_count * s->origin->vg->extent_size;
+
+	log_debug("Adding target: 0 %" PRIu64 " snapshot %s", size, params);
+	if (!dm_task_add_target(dmt, UINT64_C(0), size, "snapshot", params)) {
 		stack;
 		return 0;
 	}
@@ -919,7 +946,7 @@ struct dev_manager *dev_manager_create(struct cmd_context *cmd,
 	struct pool *mem;
 	struct dev_manager *dm;
 
-	if (!(mem = pool_create(16 * 1024))) {
+	if (!(mem = pool_create("dev_manager", 16 * 1024))) {
 		stack;
 		return NULL;
 	}
@@ -970,7 +997,7 @@ void dev_manager_destroy(struct dev_manager *dm)
 }
 
 int dev_manager_info(struct dev_manager *dm, const struct logical_volume *lv,
-		     int mknodes, struct dm_info *info)
+		     int mknodes, int with_open_count, struct dm_info *info)
 {
 	char *name;
 
@@ -986,7 +1013,8 @@ int dev_manager_info(struct dev_manager *dm, const struct logical_volume *lv,
 	 * Try and get some info on this device.
 	 */
 	log_debug("Getting device info for %s", name);
-	if (!_info(name, lv->lvid.s, mknodes, info, NULL, NULL)) {
+	if (!_info(name, lv->lvid.s, mknodes, with_open_count, info, NULL,
+		   NULL)) {
 		stack;
 		return 0;
 	}
@@ -1065,7 +1093,7 @@ static struct dev_layer *_create_dev(struct dev_manager *dm, char *name,
 	dl->name = name;
 
 	log_debug("Getting device info for %s", dl->name);
-	if (!_info(dl->name, dlid, 0, &dl->info, dm->mem, &uuid)) {
+	if (!_info(dl->name, dlid, 0, 0, &dl->info, dm->mem, &uuid)) {
 		stack;
 		return NULL;
 	}

lib/activate/dev_manager.h

@@ -36,7 +36,7 @@ void dev_manager_exit(void);
  * unsuspended until the snapshot is also created.)
  */
 int dev_manager_info(struct dev_manager *dm, const struct logical_volume *lv,
-		     int mknodes, struct dm_info *info);
+		     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);
 int dev_manager_mirror_percent(struct dev_manager *dm,

lib/activate/fs.c

@@ -65,10 +65,10 @@ static int _rm_dir(const char *dev_dir, const char *vg_name)
 		return 0;
 	}
 
-	log_very_verbose("Removing directory %s", vg_path);
-
-	if (is_empty_dir(vg_path))
+	if (dir_exists(vg_path) && is_empty_dir(vg_path)) {
+		log_very_verbose("Removing directory %s", vg_path);
 		rmdir(vg_path);
+	}
 
 	return 1;
 }
@@ -202,9 +202,9 @@ static int _rm_link(const char *dev_dir, const char *vg_name,
 	}
 
 	if (lstat(lv_path, &buf) || !S_ISLNK(buf.st_mode)) {
-		if (errno != ENOENT)
-			log_error("%s not symbolic link - not removing",
-				  lv_path);
+		if (errno == ENOENT)
+			return 1;
+		log_error("%s not symbolic link - not removing", lv_path);
 		return 0;
 	}
 

lib/cache/lvmcache.c

@@ -104,10 +104,30 @@ struct lvmcache_vginfo *vginfo_from_vgname(const char *vgname)
 const struct format_type *fmt_from_vgname(const char *vgname)
 {
 	struct lvmcache_vginfo *vginfo;
+	struct label *label;
+	struct list *ih, *devh, *tmp;
+	struct list devs;
+	struct device_list *devl;
 
 	if (!(vginfo = vginfo_from_vgname(vgname)))
 		return NULL;
 
+	/* This function is normally called before reading metadata so
+ 	 * we check cached labels here. Unfortunately vginfo is volatile. */
+	list_init(&devs);
+	list_iterate(ih, &vginfo->infos) {
+		devl = malloc(sizeof(*devl));
+		devl->dev = list_item(ih, struct lvmcache_info)->dev;
+		list_add(&devs, &devl->list);
+	}
+
+	list_iterate_safe(devh, tmp, &devs) {
+		devl = list_item(devh, struct device_list);
+		label_read(devl->dev, &label);
+		list_del(&devl->list);
+		free(devl);
+	}
+
 	return vginfo->fmt;
 }
 
@@ -188,7 +208,7 @@ int lvmcache_label_scan(struct cmd_context *cmd, int full_scan)
 		goto out;
 	}
 
-	if (!(iter = dev_iter_create(cmd->filter))) {
+	if (!(iter = dev_iter_create(cmd->filter, (full_scan == 2) ? 1: 0))) {
 		log_error("dev_iter creation failed");
 		goto out;
 	}
@@ -266,7 +286,7 @@ struct device *device_from_pvid(struct cmd_context *cmd, struct id *pvid)
 	if (memlock())
 		return NULL;
 
-	lvmcache_label_scan(cmd, 1);
+	lvmcache_label_scan(cmd, 2);
 
 	/* Try again */
 	if ((info = info_from_pvid((char *) pvid))) {
@@ -403,6 +423,9 @@ 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);
+
 	return 1;
 }
 
@@ -551,6 +574,8 @@ static void _lvmcache_destroy_lockname(int present)
 
 void lvmcache_destroy(void)
 {
+	log_verbose("Wiping internal VG cache");
+
 	_has_scanned = 0;
 
 	if (_vgid_hash) {

lib/cache/lvmcache.h

@@ -56,7 +56,8 @@ struct lvmcache_info {
 int lvmcache_init(void);
 void lvmcache_destroy(void);
 
-/* Set full_scan to 1 to reread every filtered device label */
+/* Set full_scan to 1 to reread every filtered device label or
+ * 2 to rescan /dev for new devices */
 int lvmcache_label_scan(struct cmd_context *cmd, int full_scan);
 
 /* Add/delete a device */

lib/commands/toolcontext.c

@@ -181,6 +181,12 @@ static int _process_config(struct cmd_context *cmd)
 		return 0;
 	}
 
+	if (*cmd->proc_dir && !dir_exists(cmd->proc_dir)) {
+		log_error("Warning: proc dir %s not found - some checks will be bypassed",
+			  cmd->proc_dir);
+		cmd->proc_dir[0] = '\0';
+	}
+
 	/* activation? */
 	cmd->default_settings.activation = find_config_int(cmd->cft->root,
 							   "global/activation",
@@ -519,8 +525,8 @@ static struct dev_filter *_init_filter_components(struct cmd_context *cmd)
 
 	/* regex filter. Optional. */
 	if (!(cn = find_config_node(cmd->cft->root, "devices/filter")))
-		log_debug("devices/filter not found in config file: no regex "
-			  "filter installed");
+		log_very_verbose("devices/filter not found in config file: "
+				 "no regex filter installed");
 
 	else if (!(filters[nr_filt++] = regex_filter_create(cn->v))) {
 		log_error("Failed to create regex device filter");
@@ -537,6 +543,7 @@ static struct dev_filter *_init_filter_components(struct cmd_context *cmd)
 	/* md component filter. Optional, non-critical. */
 	if (find_config_bool(cmd->cft->root, "devices/md_component_detection",
 			     DEFAULT_MD_COMPONENT_DETECTION)) {
+		init_md_filtering(1);
 		if ((filters[nr_filt] = md_filter_create()))
 			nr_filt++;
 	}
@@ -824,10 +831,15 @@ struct cmd_context *create_toolcontext(struct arg *the_args)
 		goto error;
 
 	/* Create system directory if it doesn't already exist */
-	if (*cmd->sys_dir && !create_dir(cmd->sys_dir))
+	if (*cmd->sys_dir && !create_dir(cmd->sys_dir)) {
+		log_error("Failed to create LVM2 system dir for metadata backups, config "
+			  "files and internal cache.");
+		log_error("Set environment variable LVM_SYSTEM_DIR to alternative location "
+			  "or empty string.");
 		goto error;
+	}
 
-	if (!(cmd->libmem = pool_create(4 * 1024))) {
+	if (!(cmd->libmem = pool_create("library", 4 * 1024))) {
 		log_error("Library memory pool creation failed");
 		return 0;
 	}
@@ -858,7 +870,7 @@ struct cmd_context *create_toolcontext(struct arg *the_args)
 	if (!_init_filters(cmd))
 		goto error;
 
-	if (!(cmd->mem = pool_create(4 * 1024))) {
+	if (!(cmd->mem = pool_create("command", 4 * 1024))) {
 		log_error("Command memory pool creation failed");
 		return 0;
 	}

lib/config/config.c

@@ -99,7 +99,7 @@ static int _tok_match(const char *str, const char *b, const char *e)
 struct config_tree *create_config_tree(const char *filename)
 {
 	struct cs *c;
-	struct pool *mem = pool_create(10 * 1024);
+	struct pool *mem = pool_create("config", 10 * 1024);
 
 	if (!mem) {
 		stack;

lib/device/dev-cache.c

@@ -21,6 +21,7 @@
 #include "lvm-types.h"
 #include "btree.h"
 #include "filter.h"
+#include "filter-persistent.h"
 
 #include <unistd.h>
 #include <sys/param.h>
@@ -82,6 +83,7 @@ struct device *dev_create_file(const char *filename, struct device *dev,
 	dev->dev = 0;
 	dev->fd = -1;
 	dev->open_count = 0;
+	dev->block_size = -1;
 	memset(dev->pvid, 0, sizeof(dev->pvid));
 	list_init(&dev->open_list);
 
@@ -101,6 +103,7 @@ static struct device *_dev_create(dev_t d)
 	dev->dev = d;
 	dev->fd = -1;
 	dev->open_count = 0;
+	dev->block_size = -1;
 	dev->end = UINT64_C(0);
 	memset(dev->pvid, 0, sizeof(dev->pvid));
 	list_init(&dev->open_list);
@@ -184,7 +187,7 @@ static int _add_alias(struct device *dev, const char *path)
 	/* Is name already there? */
 	list_iterate(ah, &dev->aliases) {
 		if (!strcmp(list_item(ah, struct str_list)->str, path)) {
-			stack;
+			log_debug("%s: Already in device cache", path);
 			return 1;
 		}
 	}
@@ -353,11 +356,11 @@ static int _insert(const char *path, int rec)
 	return r;
 }
 
-static void _full_scan(void)
+static void _full_scan(int dev_scan)
 {
 	struct list *dh;
 
-	if (_cache.has_scanned)
+	if (_cache.has_scanned && !dev_scan)
 		return;
 
 	list_iterate(dh, &_cache.dirs) {
@@ -366,6 +369,7 @@ static void _full_scan(void)
 	};
 
 	_cache.has_scanned = 1;
+	init_full_scan_done(1);
 }
 
 int dev_cache_has_scanned(void)
@@ -377,17 +381,16 @@ void dev_cache_scan(int do_scan)
 {
 	if (!do_scan)
 		_cache.has_scanned = 1;
-	else {
-		_cache.has_scanned = 0;
-		_full_scan();
-	}
+	else
+		_full_scan(1);
 }
 
 int dev_cache_init(void)
 {
 	_cache.names = NULL;
+	_cache.has_scanned = 0;
 
-	if (!(_cache.mem = pool_create(10 * 1024))) {
+	if (!(_cache.mem = pool_create("dev_cache", 10 * 1024))) {
 		stack;
 		return 0;
 	}
@@ -538,7 +541,7 @@ struct device *dev_cache_get(const char *name, struct dev_filter *f)
 	return (d && (!f || f->passes_filter(f, d))) ? d : NULL;
 }
 
-struct dev_iter *dev_iter_create(struct dev_filter *f)
+struct dev_iter *dev_iter_create(struct dev_filter *f, int dev_scan)
 {
 	struct dev_iter *di = dbg_malloc(sizeof(*di));
 
@@ -547,7 +550,14 @@ struct dev_iter *dev_iter_create(struct dev_filter *f)
 		return NULL;
 	}
 
-	_full_scan();
+
+	if (dev_scan) {
+		/* Flag gets reset between each command */
+		if (!full_scan_done())
+			persistent_filter_wipe(f); /* Calls _full_scan(1) */
+	} else
+		_full_scan(0);
+
 	di->current = btree_first(_cache.devices);
 	di->filter = f;
 

lib/device/dev-cache.h

@@ -45,7 +45,7 @@ struct device *dev_cache_get(const char *name, struct dev_filter *f);
  * Object for iterating through the cache.
  */
 struct dev_iter;
-struct dev_iter *dev_iter_create(struct dev_filter *f);
+struct dev_iter *dev_iter_create(struct dev_filter *f, int dev_scan);
 void dev_iter_destroy(struct dev_iter *iter);
 struct device *dev_iter_get(struct dev_iter *iter);
 

lib/device/dev-io.c

@@ -81,7 +81,9 @@ static int _io(struct device_area *where, void *buffer, int should_write)
 	}
 
 	if (lseek(fd, (off_t) where->start, SEEK_SET) < 0) {
-		log_sys_error("lseek", dev_name(where->dev));
+		log_error("%s: lseek %" PRIu64 " failed: %s",
+			  dev_name(where->dev), (uint64_t) where->start,
+			  strerror(errno));
 		return 0;
 	}
 
@@ -92,6 +94,14 @@ static int _io(struct device_area *where, void *buffer, int should_write)
 			    read(fd, buffer, (size_t) where->size - total);
 		while ((n < 0) && ((errno == EINTR) || (errno == EAGAIN)));
 
+		if (n < 0)
+			log_error("%s: %s failed after %" PRIu64 " of %" PRIu64
+				  " at %" PRIu64 ": %s", dev_name(where->dev),
+				  should_write ? "write" : "read",
+				  (uint64_t) total,
+				  (uint64_t) where->size,
+				  (uint64_t) where->start, strerror(errno));
+
 		if (n <= 0)
 			break;
 
@@ -114,14 +124,18 @@ static int _io(struct device_area *where, void *buffer, int should_write)
  */
 static int _get_block_size(struct device *dev, unsigned int *size)
 {
-	int s;
+	const char *name = dev_name(dev);
 
-	if (ioctl(dev_fd(dev), BLKBSZGET, &s) < 0) {
-		log_sys_error("ioctl BLKBSZGET", dev_name(dev));
-		return 0;
+	if ((dev->block_size == -1)) {
+		if (ioctl(dev_fd(dev), BLKBSZGET, &dev->block_size) < 0) {
+			log_sys_error("ioctl BLKBSZGET", name);
+			return 0;
+		}
+		log_debug("%s: block size is %u bytes", name, dev->block_size);
 	}
 
-	*size = (unsigned int) s;
+	*size = (unsigned int) dev->block_size;
+
 	return 1;
 }
 
@@ -217,7 +231,6 @@ int dev_get_size(const struct device *dev, uint64_t *size)
 	int fd;
 	const char *name = dev_name(dev);
 
-	log_very_verbose("Getting size of %s", name);
 	if ((fd = open(name, O_RDONLY)) < 0) {
 		log_sys_error("open", name);
 		return 0;
@@ -225,22 +238,27 @@ int dev_get_size(const struct device *dev, uint64_t *size)
 
 	if (ioctl(fd, BLKGETSIZE64, size) < 0) {
 		log_sys_error("ioctl BLKGETSIZE64", name);
-		close(fd);
+		if (close(fd))
+			log_sys_error("close", name);
 		return 0;
 	}
 
 	*size >>= BLKSIZE_SHIFT;	/* Convert to sectors */
-	close(fd);
+	if (close(fd))
+		log_sys_error("close", name);
+
+	log_very_verbose("%s: size is %" PRIu64 " sectors", name, *size);
+
 	return 1;
 }
 
+/* FIXME Unused
 int dev_get_sectsize(struct device *dev, uint32_t *size)
 {
 	int fd;
 	int s;
 	const char *name = dev_name(dev);
 
-	log_very_verbose("Getting size of %s", name);
 	if ((fd = open(name, O_RDONLY)) < 0) {
 		log_sys_error("open", name);
 		return 0;
@@ -254,8 +272,12 @@ int dev_get_sectsize(struct device *dev, uint32_t *size)
 
 	close(fd);
 	*size = (uint32_t) s;
+
+	log_very_verbose("%s: sector size is %" PRIu32 " bytes", name, *size);
+
 	return 1;
 }
+*/
 
 void dev_flush(struct device *dev)
 {
@@ -274,8 +296,20 @@ int dev_open_flags(struct device *dev, int flags, int direct, int quiet)
 	const char *name;
 
 	if (dev->fd >= 0) {
-		dev->open_count++;
-		return 1;
+		if ((dev->flags & DEV_OPENED_RW) ||
+		    ((flags & O_ACCMODE) != O_RDWR)) {
+			dev->open_count++;
+			return 1;
+		}
+
+		if (dev->open_count) {
+			/* FIXME Ensure we never get here */
+			log_debug("WARNING: %s already opened read-only", 
+				  dev_name(dev));
+			dev->open_count++;
+		}
+
+		dev_close_immediate(dev);
 	}
 
 	if (memlock())
@@ -307,18 +341,25 @@ int dev_open_flags(struct device *dev, int flags, int direct, int quiet)
 #endif
 
 	if ((dev->fd = open(name, flags, 0777)) < 0) {
-		log_sys_error("open", name);
+		if (quiet)
+			log_sys_debug("open", name);
+		else
+			log_sys_error("open", name);
 		return 0;
 	}
 
-	dev->open_count = 1;
+	dev->open_count++;
 	dev->flags &= ~DEV_ACCESSED_W;
+	if ((flags & O_ACCMODE) == O_RDWR)
+		dev->flags |= DEV_OPENED_RW;
+	else
+		dev->flags &= ~DEV_OPENED_RW;
 
 	if (!(dev->flags & DEV_REGULAR) &&
 	    ((fstat(dev->fd, &buf) < 0) || (buf.st_rdev != dev->dev))) {
 		log_error("%s: fstat failed: Has device name changed?", name);
-		dev_close(dev);
-		dev->fd = -1;
+		dev_close_immediate(dev);
+		dev->open_count = 0;
 		return 0;
 	}
 
@@ -327,12 +368,13 @@ int dev_open_flags(struct device *dev, int flags, int direct, int quiet)
 		dev_flush(dev);
 #endif
 
-	if ((flags & O_CREAT) && !(flags & O_TRUNC)) {
+	if ((flags & O_CREAT) && !(flags & O_TRUNC))
 		dev->end = lseek(dev->fd, (off_t) 0, SEEK_END);
-	}
 
 	list_add(&_open_devices, &dev->open_list);
-	log_debug("Opened %s", dev_name(dev));
+
+	log_debug("Opened %s %s", dev_name(dev),
+		  dev->flags & DEV_OPENED_RW ? "RW" : "RO");
 
 	return 1;
 }
@@ -360,6 +402,7 @@ static void _close(struct device *dev)
 	if (close(dev->fd))
 		log_sys_error("close", dev_name(dev));
 	dev->fd = -1;
+	dev->block_size = -1;
 	list_del(&dev->open_list);
 
 	log_debug("Closed %s", dev_name(dev));
@@ -385,8 +428,11 @@ static int _dev_close(struct device *dev, int immediate)
 		dev_flush(dev);
 #endif
 
+	if (dev->open_count > 0)
+		dev->open_count--;
+
 	/* FIXME lookup device in cache to get vgname and see if it's locked? */
-	if (--dev->open_count < 1 && (immediate || !vgs_locked()))
+	if (immediate || (dev->open_count < 1 && !vgs_locked()))
 		_close(dev);
 
 	return 1;
@@ -418,8 +464,10 @@ int dev_read(struct device *dev, uint64_t offset, size_t len, void *buffer)
 {
 	struct device_area where;
 
-	if (!dev->open_count)
+	if (!dev->open_count) {
+		stack;
 		return 0;
+	}
 
 	where.dev = dev;
 	where.start = offset;
@@ -437,8 +485,10 @@ int dev_append(struct device *dev, size_t len, void *buffer)
 {
 	int r;
 
-	if (!dev->open_count)
+	if (!dev->open_count) {
+		stack;
 		return 0;
+	}
 
 	r = dev_write(dev, dev->end, len, buffer);
 	dev->end += (uint64_t) len;
@@ -453,8 +503,10 @@ int dev_write(struct device *dev, uint64_t offset, size_t len, void *buffer)
 {
 	struct device_area where;
 
-	if (!dev->open_count)
+	if (!dev->open_count) {
+		stack;
 		return 0;
+	}
 
 	where.dev = dev;
 	where.start = offset;
@@ -501,6 +553,5 @@ int dev_zero(struct device *dev, uint64_t offset, size_t len)
 	if (!dev_close(dev))
 		stack;
 
-	/* FIXME: Always display error */
 	return (len == 0);
 }

lib/device/dev-md.c

@@ -0,0 +1,69 @@
+/*
+ * Copyright (C) 2004 Luca Berra
+ * Copyright (C) 2004 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 "lib.h"
+#include "metadata.h"
+#include "xlate.h"
+
+/* Lifted from <linux/raid/md_p.h> because of difficulty including it */
+
+#define MD_SB_MAGIC 0xa92b4efc
+#define MD_RESERVED_BYTES (64 * 1024)
+#define MD_RESERVED_SECTORS (MD_RESERVED_BYTES / 512)
+#define MD_NEW_SIZE_SECTORS(x) ((x & ~(MD_RESERVED_SECTORS - 1)) \
+				- MD_RESERVED_SECTORS)
+
+/*
+ * Returns -1 on error
+ */
+int dev_is_md(struct device *dev, uint64_t *sb)
+{
+	int ret = 0;
+
+#ifdef linux
+
+	uint64_t size, sb_offset;
+	uint32_t md_magic;
+
+	if (!dev_get_size(dev, &size)) {
+		stack;
+		return -1;
+	}
+
+	if (size < MD_RESERVED_SECTORS * 2)
+		return 0;
+
+	if (!dev_open(dev)) {
+		stack;
+		return -1;
+	}
+
+	sb_offset = MD_NEW_SIZE_SECTORS(size) << SECTOR_SHIFT;
+
+	/* Check if it is an md component device. */
+	if (dev_read(dev, sb_offset, sizeof(uint32_t), &md_magic) &&
+	    (md_magic == xlate32(MD_SB_MAGIC))) {
+		if (sb)
+			*sb = sb_offset;
+		ret = 1;
+	}
+
+	if (!dev_close(dev))
+		stack;
+
+#endif
+	return ret;
+}
+

lib/device/device.c

@@ -13,6 +13,93 @@
  * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
+#include "lib.h"
+#include "lvm-types.h"
+#include "device.h"
+#include "metadata.h"
+#include "filter.h"
+#include "xlate.h"
+
+/* See linux/genhd.h and fs/partitions/msdos */
+
+#define PART_MAGIC 0xAA55
+#define PART_MAGIC_OFFSET UINT64_C(0x1FE)
+#define PART_OFFSET UINT64_C(0x1BE)
+
+struct partition {
+        uint8_t boot_ind;
+        uint8_t head;
+        uint8_t sector;
+        uint8_t cyl;
+        uint8_t sys_ind;	/* partition type */
+        uint8_t end_head;
+        uint8_t end_sector;
+        uint8_t end_cyl;
+        uint32_t start_sect;
+        uint32_t nr_sects;
+} __attribute__((packed));
+
+static int _is_partitionable(struct device *dev)
+{
+	int parts = max_partitions(MAJOR(dev->dev));
+
+	if ((parts <= 1) || (MINOR(dev->dev) % parts))
+		return 0;
+
+	return 1;
+}
+
+static int _has_partition_table(struct device *dev)
+{
+	int ret = 0;
+	unsigned p;
+	uint8_t buf[SECTOR_SIZE];
+	uint16_t *part_magic;
+	struct partition *part;
+
+	if (!dev_open(dev)) {
+		stack;
+		return -1;
+	}
+
+	if (!dev_read(dev, 0, sizeof(buf), &buf)) {
+		stack;
+		goto out;
+	}
+
+	/* FIXME Check for other types of partition table too */
+
+	/* Check for msdos partition table */
+	part_magic = (uint16_t *)(buf + PART_MAGIC_OFFSET);
+	if ((*part_magic == xlate16(PART_MAGIC))) {
+		part = (struct partition *) (buf + PART_OFFSET);
+		for (p = 0; p < 4; p++, part++) {
+			/* Table is invalid if boot indicator not 0 or 0x80 */
+			if ((part->boot_ind & 0x7f)) {
+				ret = 0;
+				break;
+			}
+			/* Must have at least one non-empty partition */
+			if (part->nr_sects)
+				ret = 1;
+		}
+	}
+
+      out:
+	if (!dev_close(dev))
+		stack;
+
+	return ret;
+}
+
+int is_partitioned_dev(struct device *dev)
+{
+	if (!_is_partitionable(dev))
+		return 0;
+
+	return _has_partition_table(dev);
+}
+
 #if 0
 #include <sys/stat.h>
 #include <sys/mman.h>
@@ -27,24 +114,13 @@
 #include <linux/major.h>
 #include <linux/genhd.h>
 
-#include "dbg_malloc.h"
-#include "log.h"
-#include "dev-cache.h"
-#include "metadata.h"
-#include "device.h"
-
 int _get_partition_type(struct dev_filter *filter, struct device *d);
 
-#define MINOR_PART(dm, d) (MINOR((d)->dev) % dev_max_partitions(dm, (d)->dev))
+#define MINOR_PART(dev) (MINOR((dev)->dev) % max_partitions(MINOR((dev)->dev)))
 
-int is_whole_disk(struct dev_filter *filter, struct device *d)
+int is_extended_partition(struct device *d)
 {
-	return (MINOR_PART(dm, d)) ? 0 : 1;
-}
-
-int is_extended_partition(struct dev_mgr *dm, struct device *d)
-{
-	return (MINOR_PART(dm, d) > 4) ? 1 : 0;
+	return (MINOR_PART(d) > 4) ? 1 : 0;
 }
 
 struct device *dev_primary(struct dev_mgr *dm, struct device *d)

lib/device/device.h

@@ -22,6 +22,7 @@
 #define DEV_ACCESSED_W		0x00000001	/* Device written to? */
 #define DEV_REGULAR		0x00000002	/* Regular file? */
 #define DEV_ALLOCED		0x00000004	/* dbg_malloc used */
+#define DEV_OPENED_RW		0x00000008	/* Opened RW */
 
 /*
  * All devices in LVM will be represented by one of these.
@@ -34,6 +35,7 @@ struct device {
 	/* private */
 	int fd;
 	int open_count;
+	int block_size;
 	uint32_t flags;
 	uint64_t end;
 	struct list open_list;
@@ -89,15 +91,14 @@ static inline const char *dev_name(const struct device *dev)
 /* Return a valid device name from the alias list; NULL otherwise */
 const char *dev_name_confirmed(struct device *dev, int quiet);
 
+/* Does device contain md superblock?  If so, where? */
+int dev_is_md(struct device *dev, uint64_t *sb);
+
 /* FIXME Check partition type if appropriate */
 
 #define is_lvm_partition(a) 1
+/* int is_lvm_partition(const char *name); */
 
-/*
-static inline int is_lvm_partition(const char *name)
-{
-	return 1;
-}
-*/
+int is_partitioned_dev(struct device *dev);
 
 #endif

lib/display/display.c

@@ -258,7 +258,7 @@ void pvdisplay_full(struct cmd_context *cmd, struct physical_volume *pv,
 		log_print("PV Size               %s" " / not usable %s",	/*  [LVM: %s]", */
 			  size,
 			  display_size(cmd, (pv->size -
-					     pv->pe_count * pv->pe_size),
+				       (uint64_t) pv->pe_count * pv->pe_size),
 				       SIZE_SHORT));
 
 	} else
@@ -317,7 +317,7 @@ void lvdisplay_colons(struct logical_volume *lv)
 {
 	int inkernel;
 	struct lvinfo info;
-	inkernel = lv_info(lv, &info) && info.exists;
+	inkernel = lv_info(lv, &info, 1) && info.exists;
 
 	log_print("%s%s/%s:%s:%d:%d:-1:%d:%" PRIu64 ":%d:-1:%d:%d:%d:%d",
 		  lv->vg->cmd->dev_dir,
@@ -348,7 +348,7 @@ int lvdisplay_full(struct cmd_context *cmd, struct logical_volume *lv,
 		return 0;
 	}
 
-	inkernel = lv_info(lv, &info) && info.exists;
+	inkernel = lv_info(lv, &info, 1) && info.exists;
 
 	log_print("--- Logical volume ---");
 

lib/filters/filter-md.c

@@ -28,42 +28,24 @@
 
 static int _ignore_md(struct dev_filter *f, struct device *dev)
 {
-	uint64_t size, sector;
-	uint32_t md_magic;
-
-	if (!dev_get_size(dev, &size)) {
-		stack;
-		return 0;
-	}
-
-	if (size < MD_RESERVED_SECTORS * 2)
-		/*
-		 * We could ignore it since it is obviously too
-		 * small, but that's not our job.
-		 */
+	int ret;
+	
+	if (!md_filtering())
 		return 1;
+	
+	ret = dev_is_md(dev, NULL);
 
-	if (!dev_open(dev)) {
-		stack;
+	if (ret == 1) {
+		log_debug("%s: Skipping md component device", dev_name(dev));
 		return 0;
 	}
 
-	sector = MD_NEW_SIZE_SECTORS(size);
-
-	/* Check if it is an md component device. */
-	if (dev_read(dev, sector << SECTOR_SHIFT, sizeof(uint32_t), &md_magic)) {
-		if (md_magic == MD_SB_MAGIC) {
-			log_debug("%s: Skipping md component device",
-				  dev_name(dev));
-			if (!dev_close(dev))
-				stack;
-			return 0;
-		}
+	if (ret < 0) {
+		log_debug("%s: Skipping: error in md component detection",
+			  dev_name(dev));
+		return 0;
 	}
 
-	if (!dev_close(dev))
-		stack;
-
 	return 1;
 }
 

lib/filters/filter-persistent.c

@@ -53,7 +53,9 @@ int persistent_filter_wipe(struct dev_filter *f)
 {
 	struct pfilter *pf = (struct pfilter *) f->private;
 
+	log_verbose("Wiping cache of LVM-capable devices");
 	hash_wipe(pf->devices);
+
 	/* Trigger complete device scan */
 	dev_cache_scan(1);
 
@@ -212,13 +214,10 @@ static int _lookup_p(struct dev_filter *f, struct device *dev)
 			sl = list_item(ah, struct str_list);
 			hash_insert(pf->devices, sl->str, l);
 		}
-	}
+	} else if (l == PF_BAD_DEVICE)
+			log_debug("%s: Skipping (cached)", dev_name(dev));
 
-	if (l == PF_BAD_DEVICE) {
-		log_debug("%s: Skipping (cached)", dev_name(dev));
-		return 0;
-	} else
-		return 1;
+	return (l == PF_BAD_DEVICE) ? 0 : 1;
 }
 
 static void _destroy(struct dev_filter *f)

lib/filters/filter-regex.c

@@ -101,7 +101,7 @@ static int _build_matcher(struct rfilter *rf, struct config_value *val)
 	unsigned count = 0;
 	int i, r = 0;
 
-	if (!(scratch = pool_create(1024))) {
+	if (!(scratch = pool_create("filter matcher", 1024))) {
 		stack;
 		return 0;
 	}
@@ -186,6 +186,9 @@ static int _accept_p(struct dev_filter *f, struct device *dev)
 		first = 0;
 	}
 
+	if (rejected)
+		log_debug("%s: Skipping (regex)", dev_name(dev));
+
 	/*
 	 * pass everything that doesn't match
 	 * anything.
@@ -201,7 +204,7 @@ static void _destroy(struct dev_filter *f)
 
 struct dev_filter *regex_filter_create(struct config_value *patterns)
 {
-	struct pool *mem = pool_create(10 * 1024);
+	struct pool *mem = pool_create("filter regex", 10 * 1024);
 	struct rfilter *rf;
 	struct dev_filter *f;
 

lib/filters/filter-sysfs.c

@@ -172,7 +172,7 @@ static int _read_devs(struct dev_set *ds, const char *dir)
 	unsigned char dtype;
 	struct stat info;
 	char path[PATH_MAX];
-	dev_t dev;
+	dev_t dev = { 0 };
 	int r = 1;
 
         if (!(dr = opendir(dir))) {
@@ -194,8 +194,10 @@ static int _read_devs(struct dev_set *ds, const char *dir)
 		dtype = d->d_type;
 
 		if (dtype == DT_UNKNOWN) {
-			if (stat(path, &info) >= 0) {
-				if (S_ISDIR(info.st_mode))
+			if (lstat(path, &info) >= 0) {
+				if (S_ISLNK(info.st_mode))
+					dtype = DT_LNK;
+				else if (S_ISDIR(info.st_mode))
 					dtype = DT_DIR;
 				else if (S_ISREG(info.st_mode))
 					dtype = DT_REG;
@@ -246,7 +248,11 @@ static int _accept_p(struct dev_filter *f, struct device *dev)
 	if (ds->initialised != 1)
 		return 1;
 
-	return _set_lookup(ds, dev->dev);
+	if (!_set_lookup(ds, dev->dev)) {
+		log_debug("%s: Skipping (sysfs)", dev_name(dev));
+		return 0;
+	} else
+		return 1;
 }
 
 static void _destroy(struct dev_filter *f)
@@ -265,7 +271,7 @@ struct dev_filter *sysfs_filter_create(const char *proc)
 	if (!_locate_sysfs_blocks(proc, sys_block, sizeof(sys_block)))
 		return NULL;
 
-	if (!(mem = pool_create(256))) {
+	if (!(mem = pool_create("sysfs", 256))) {
 		log_error("sysfs pool creation failed");
 		return NULL;
 	}

lib/filters/filter.c

@@ -18,6 +18,7 @@
 #include "filter.h"
 #include "lvm-string.h"
 #include "config.h"
+#include "metadata.h"
 
 #include <dirent.h>
 #include <unistd.h>
@@ -27,6 +28,10 @@
 
 #define NUMBER_OF_MAJORS 4096
 
+/* FIXME Make this sparse */
+/* 0 means LVM won't use this major number. */
+static int _max_partitions_by_major[NUMBER_OF_MAJORS];
+
 typedef struct {
 	const char *name;
 	const int max_partitions;
@@ -39,12 +44,19 @@ int md_major(void)
 	return _md_major;
 }
 
-/* This list can be supplemented with devices/types in the config file */
+/*
+ * Devices are only checked for partition tables if their minor number
+ * is a multiple of the number corresponding to their type below
+ * i.e. this gives the granularity of whole-device minor numbers.
+ * Use 1 if the device is not partitionable.
+ *
+ * The list can be supplemented with devices/types in the config file.
+ */
 static const device_info_t device_info[] = {
-	{"ide", 16},		/* IDE disk */
+	{"ide", 64},		/* IDE disk */
 	{"sd", 16},		/* SCSI disk */
-	{"md", 16},		/* Multiple Disk driver (SoftRAID) */
-	{"loop", 16},		/* Loop device */
+	{"md", 1},		/* Multiple Disk driver (SoftRAID) */
+	{"loop", 1},		/* Loop device */
 	{"dasd", 4},		/* DASD disk (IBM S/390, zSeries) */
 	{"dac960", 8},		/* DAC960 */
 	{"nbd", 16},		/* Network Block Device */
@@ -53,38 +65,60 @@ static const device_info_t device_info[] = {
 	{"ubd", 16},		/* User-mode virtual block device */
 	{"ataraid", 16},	/* ATA Raid */
 	{"drbd", 16},		/* Distributed Replicated Block Device */
+	{"emcpower", 16},	/* EMC Powerpath */
 	{"power2", 16},		/* EMC Powerpath */
 	{"i2o_block", 16},	/* i2o Block Disk */
 	{"iseries/vd", 8},	/* iSeries disks */
+	{"gnbd", 1},		/* Network block device */
 	{NULL, 0}
 };
 
 static int _passes_lvm_type_device_filter(struct dev_filter *f,
 					  struct device *dev)
 {
-	int fd;
 	const char *name = dev_name(dev);
+	int ret = 0;
+	uint64_t size;
 
 	/* Is this a recognised device type? */
-	if (!(((int *) f->private)[MAJOR(dev->dev)])) {
+	if (!_max_partitions_by_major[MAJOR(dev->dev)]) {
 		log_debug("%s: Skipping: Unrecognised LVM device type %"
 			  PRIu64, name, (uint64_t) MAJOR(dev->dev));
 		return 0;
 	}
 
 	/* Check it's accessible */
-	if ((fd = open(name, O_RDONLY)) < 0) {
-		log_debug("%s: Skipping: open failed: %s", name,
-			  strerror(errno));
+	if (!dev_open_flags(dev, O_RDONLY, 0, 1)) {
+		log_debug("%s: Skipping: open failed", name);
 		return 0;
 	}
+	
+	/* Check it's not too small */
+	if (!dev_get_size(dev, &size)) {
+		log_debug("%s: Skipping: dev_get_size failed", name);
+		goto out;
+	}
 
-	close(fd);
+	if (size < PV_MIN_SIZE) {
+		log_debug("%s: Skipping: Too small to hold a PV", name);
+		goto out;
+	}
 
-	return 1;
+	if (is_partitioned_dev(dev)) {
+		log_debug("%s: Skipping: Partition table signature found",
+			  name);
+		goto out;
+	}
+
+	ret = 1;
+
+      out:
+	dev_close(dev);
+
+	return ret;
 }
 
-static int *_scan_proc_dev(const char *proc, const struct config_node *cn)
+static int _scan_proc_dev(const char *proc, const struct config_node *cn)
 {
 	char line[80];
 	char proc_devices[PATH_MAX];
@@ -94,36 +128,31 @@ static int *_scan_proc_dev(const char *proc, const struct config_node *cn)
 	int blocksection = 0;
 	size_t dev_len = 0;
 	struct config_value *cv;
-	int *max_partitions_by_major;
 	char *name;
 
-	/* FIXME Make this sparse */
-	if (!(max_partitions_by_major =
-	      dbg_malloc(sizeof(int) * NUMBER_OF_MAJORS))) {
-		log_error("Filter failed to allocate max_partitions_by_major");
-		return NULL;
-	}
 
 	if (!*proc) {
 		log_verbose("No proc filesystem found: using all block device "
 			    "types");
 		for (i = 0; i < NUMBER_OF_MAJORS; i++)
-			max_partitions_by_major[i] = 1;
-		return max_partitions_by_major;
+			_max_partitions_by_major[i] = 1;
+		return 1;
 	}
 
+	/* All types unrecognised initially */
+	memset(_max_partitions_by_major, 0, sizeof(int) * NUMBER_OF_MAJORS);
+
 	if (lvm_snprintf(proc_devices, sizeof(proc_devices),
 			 "%s/devices", proc) < 0) {
 		log_error("Failed to create /proc/devices string");
-		return NULL;
+		return 0;
 	}
 
 	if (!(pd = fopen(proc_devices, "r"))) {
 		log_sys_error("fopen", proc_devices);
-		return NULL;
+		return 0;
 	}
 
-	memset(max_partitions_by_major, 0, sizeof(int) * NUMBER_OF_MAJORS);
 	while (fgets(line, 80, pd) != NULL) {
 		i = 0;
 		while (line[i] == ' ' && line[i] != '\0')
@@ -158,13 +187,13 @@ static int *_scan_proc_dev(const char *proc, const struct config_node *cn)
 			if (dev_len <= strlen(line + i) &&
 			    !strncmp(device_info[j].name, line + i, dev_len) &&
 			    (line_maj < NUMBER_OF_MAJORS)) {
-				max_partitions_by_major[line_maj] =
+				_max_partitions_by_major[line_maj] =
 				    device_info[j].max_partitions;
 				break;
 			}
 		}
 
-		if (max_partitions_by_major[line_maj] || !cn)
+		if (!cn)
 			continue;
 
 		/* Check devices/types for local variations */
@@ -172,7 +201,7 @@ static int *_scan_proc_dev(const char *proc, const struct config_node *cn)
 			if (cv->type != CFG_STRING) {
 				log_error("Expecting string in devices/types "
 					  "in config file");
-				return NULL;
+				return 0;
 			}
 			dev_len = strlen(cv->v.str);
 			name = cv->v.str;
@@ -181,24 +210,29 @@ static int *_scan_proc_dev(const char *proc, const struct config_node *cn)
 				log_error("Max partition count missing for %s "
 					  "in devices/types in config file",
 					  name);
-				return NULL;
+				return 0;
 			}
 			if (!cv->v.i) {
 				log_error("Zero partition count invalid for "
 					  "%s in devices/types in config file",
 					  name);
-				return NULL;
+				return 0;
 			}
 			if (dev_len <= strlen(line + i) &&
 			    !strncmp(name, line + i, dev_len) &&
 			    (line_maj < NUMBER_OF_MAJORS)) {
-				max_partitions_by_major[line_maj] = cv->v.i;
+				_max_partitions_by_major[line_maj] = cv->v.i;
 				break;
 			}
 		}
 	}
 	fclose(pd);
-	return max_partitions_by_major;
+	return 1;
+}
+
+int max_partitions(int major)
+{
+	return _max_partitions_by_major[major];
 }
 
 struct dev_filter *lvm_type_filter_create(const char *proc,
@@ -213,8 +247,9 @@ struct dev_filter *lvm_type_filter_create(const char *proc,
 
 	f->passes_filter = _passes_lvm_type_device_filter;
 	f->destroy = lvm_type_filter_destroy;
+	f->private = NULL;
 
-	if (!(f->private = _scan_proc_dev(proc, cn))) {
+	if (!_scan_proc_dev(proc, cn)) {
 		stack;
 		return NULL;
 	}
@@ -224,7 +259,6 @@ struct dev_filter *lvm_type_filter_create(const char *proc,
 
 void lvm_type_filter_destroy(struct dev_filter *f)
 {
-	dbg_free(f->private);
 	dbg_free(f);
 	return;
 }

lib/filters/filter.h

@@ -37,4 +37,6 @@ void lvm_type_filter_destroy(struct dev_filter *f);
 
 int md_major(void);
 
+int max_partitions(int major);
+
 #endif

lib/format1/disk-rep.c

@@ -482,7 +482,7 @@ int read_pvs_in_vg(const struct format_type *fmt, const char *vg_name,
 		/* vgcache_del(vg_name); */
 	}
 
-	if (!(iter = dev_iter_create(filter))) {
+	if (!(iter = dev_iter_create(filter, 1))) {
 		log_error("read_pvs_in_vg: dev_iter_create failed");
 		return 0;
 	}

lib/format1/format1.c

@@ -183,7 +183,7 @@ static struct volume_group *_vg_read(struct format_instance *fid,
 				     const char *vg_name,
 				     struct metadata_area *mda)
 {
-	struct pool *mem = pool_create(1024 * 10);
+	struct pool *mem = pool_create("lvm1 vg_read", 1024 * 10);
 	struct list pvs;
 	struct volume_group *vg = NULL;
 	list_init(&pvs);
@@ -276,7 +276,7 @@ static int _flatten_vg(struct format_instance *fid, struct pool *mem,
 static int _vg_write(struct format_instance *fid, struct volume_group *vg,
 		     struct metadata_area *mda)
 {
-	struct pool *mem = pool_create(1024 * 10);
+	struct pool *mem = pool_create("lvm1 vg_write", 1024 * 10);
 	struct list pvds;
 	int r = 0;
 
@@ -299,7 +299,7 @@ static int _vg_write(struct format_instance *fid, struct volume_group *vg,
 static int _pv_read(const struct format_type *fmt, const char *pv_name,
 		    struct physical_volume *pv, struct list *mdas)
 {
-	struct pool *mem = pool_create(1024);
+	struct pool *mem = pool_create("lvm1 pv_read", 1024);
 	struct disk_list *dl;
 	struct device *dev;
 	int r = 0;
@@ -421,7 +421,7 @@ static int _pv_write(const struct format_type *fmt, struct physical_volume *pv,
 	pv->pe_size = pv->pe_count = 0;
 	pv->pe_start = PE_ALIGN;
 
-	if (!(mem = pool_create(1024))) {
+	if (!(mem = pool_create("lvm1 pv_write", 1024))) {
 		stack;
 		return 0;
 	}

lib/format1/import-export.c

@@ -646,7 +646,7 @@ int import_snapshots(struct pool *mem, struct volume_group *vg,
 				continue;
 
 			/* insert the snapshot */
-			if (!vg_add_snapshot(org, cow, 1, NULL,
+			if (!vg_add_snapshot(org, cow, 1, NULL, org->le_count, 
 					     lvd->lv_chunk_size)) {
 				log_err("Couldn't add snapshot.");
 				return 0;

lib/format1/import-extents.c

@@ -344,7 +344,7 @@ int import_extents(struct cmd_context *cmd, struct volume_group *vg,
 		   struct list *pvds)
 {
 	int r = 0;
-	struct pool *scratch = pool_create(10 * 1024);
+	struct pool *scratch = pool_create("lvm1 import_extents", 10 * 1024);
 	struct hash_table *maps;
 
 	if (!scratch) {

lib/format1/vg_number.c

@@ -30,7 +30,7 @@ int get_free_vg_number(struct format_instance *fid, struct dev_filter *filter,
 	struct list *pvh;
 	struct list all_pvs;
 	struct disk_list *dl;
-	struct pool *mem = pool_create(10 * 1024);
+	struct pool *mem = pool_create("lvm1 vg_number", 10 * 1024);
 	int numbers[MAX_VG], i, r = 0;
 
 	list_init(&all_pvs);

lib/format_pool/disk_rep.c

@@ -259,7 +259,7 @@ static int _read_vg_pds(const struct format_type *fmt, struct pool *mem,
 
 	/* FIXME: maybe should return a different error in memory
 	 * allocation failure */
-	if (!(tmpmem = pool_create(512))) {
+	if (!(tmpmem = pool_create("pool read_vg", 512))) {
 		stack;
 		return 0;
 	}

lib/format_pool/disk_rep.h

@@ -20,6 +20,8 @@
 #include "metadata.h"
 #include "pool.h"
 
+#define MINOR_OFFSET 65536
+
 /* From NSP.cf */
 #define NSPMajorVersion	4
 #define NSPMinorVersion	1
@@ -66,29 +68,6 @@ struct pool_list;
 struct user_subpool;
 struct user_device;
 
-/* This must be kept up to date with sistina/pool/module/pool_sptypes.h */
-
-/*  Generic Labels  */
-#define SPTYPE_DATA                (0x00000000)
-
-/*  GFS specific labels  */
-#define SPTYPE_GFS_DATA            (0x68011670)
-#define SPTYPE_GFS_JOURNAL         (0x69011670)
-
-struct sptype_name {
-	const char *name;
-	uint32_t label;
-};
-
-static const struct sptype_name sptype_names[] = {
-	{"data",	SPTYPE_DATA},
-
-	{"gfs_data",	SPTYPE_GFS_DATA},
-	{"gfs_journal",	SPTYPE_GFS_JOURNAL},
-
-	{"", 0x0}		/*  This must be the last flag.  */
-};
-
 struct pool_disk {
 	uint64_t pl_magic;	/* Pool magic number */
 	uint64_t pl_pool_id;	/* Unique pool identifier */

lib/format_pool/format_pool.c

@@ -181,7 +181,7 @@ static struct volume_group *_vg_read(struct format_instance *fid,
 				     const char *vg_name,
 				     struct metadata_area *mda)
 {
-	struct pool *mem = pool_create(1024);
+	struct pool *mem = pool_create("pool vg_read", 1024);
 	struct list pds;
 	struct volume_group *vg = NULL;
 
@@ -227,7 +227,7 @@ static int _pv_setup(const struct format_type *fmt,
 static int _pv_read(const struct format_type *fmt, const char *pv_name,
 		    struct physical_volume *pv, struct list *mdas)
 {
-	struct pool *mem = pool_create(1024);
+	struct pool *mem = pool_create("pool pv_read", 1024);
 	struct pool_list *pl;
 	struct device *dev;
 	int r = 0;

lib/format_pool/import_export.c

@@ -19,6 +19,7 @@
 #include "metadata.h"
 #include "lvmcache.h"
 #include "disk_rep.h"
+#include "sptype_names.h"
 #include "lv_alloc.h"
 #include "str_list.h"
 #include "display.h"
@@ -107,7 +108,7 @@ int import_pool_lvs(struct volume_group *vg, struct pool *mem, struct list *pls)
 		/* for pool a minor of 0 is dynamic */
 		if (pl->pd.pl_minor) {
 			lv->status |= FIXED_MINOR;
-			lv->minor = pl->pd.pl_minor;
+			lv->minor = pl->pd.pl_minor + MINOR_OFFSET;
 		} else {
 			lv->minor = -1;
 		}

lib/format_pool/sptype_names.h

@@ -0,0 +1,42 @@
+/*
+ * Copyright (C) 1997-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004-2005 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 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
+ */
+
+#ifndef SPTYPE_NAMES_H
+#define SPTYPE_NAMES_H
+
+/* This must be kept up to date with sistina/pool/module/pool_sptypes.h */
+
+/*  Generic Labels  */
+#define SPTYPE_DATA                (0x00000000)
+
+/*  GFS specific labels  */
+#define SPTYPE_GFS_DATA            (0x68011670)
+#define SPTYPE_GFS_JOURNAL         (0x69011670)
+
+struct sptype_name {
+	const char *name;
+	uint32_t label;
+};
+
+static const struct sptype_name sptype_names[] = {
+	{"data",	SPTYPE_DATA},
+
+	{"gfs_data",	SPTYPE_GFS_DATA},
+	{"gfs_journal",	SPTYPE_GFS_JOURNAL},
+
+	{"", 0x0}		/*  This must be the last flag.  */
+};
+
+#endif

lib/format_text/archive.c

@@ -53,7 +53,7 @@ struct archive_file {
 	struct list list;
 
 	char *path;
-	int index;
+	uint32_t index;
 };
 
 /*
@@ -132,7 +132,8 @@ static char *_join(struct pool *mem, const char *dir, const char *name)
 static struct list *_scan_archive(struct pool *mem,
 				  const char *vgname, const char *dir)
 {
-	int i, count, ix;
+	int i, count;
+	uint32_t ix;
 	char vgname_found[64], *path;
 	struct dirent **dirent;
 	struct archive_file *af;
@@ -240,7 +241,7 @@ int archive_vg(struct volume_group *vg,
 	       uint32_t retain_days, uint32_t min_archive)
 {
 	int i, fd, renamed = 0;
-	unsigned int ix = 0;
+	uint32_t ix = 0;
 	struct archive_file *last;
 	FILE *fp = NULL;
 	char temp_file[PATH_MAX], archive_name[PATH_MAX];
@@ -285,7 +286,7 @@ int archive_vg(struct volume_group *vg,
 
 	for (i = 0; i < 10; i++) {
 		if (lvm_snprintf(archive_name, sizeof(archive_name),
-				 "%s/%s_%05d.vg", dir, vg->name, ix) < 0) {
+				 "%s/%s_%05u.vg", dir, vg->name, ix) < 0) {
 			log_error("Archive file name too long.");
 			return 0;
 		}

lib/format_text/export.c

@@ -513,7 +513,7 @@ static int _print_snapshot(struct formatter *f, struct snapshot *snap,
 	}
 
 	seg.le = 0;
-	seg.len = snap->origin->le_count;
+	seg.len = snap->le_count;
 	seg.origin = snap->origin;
 	seg.cow = snap->cow;
 	seg.chunk_size = snap->chunk_size;
@@ -652,7 +652,7 @@ static int _build_pv_names(struct formatter *f, struct volume_group *vg)
 	struct physical_volume *pv;
 	char buffer[32], *name;
 
-	if (!(f->mem = pool_create(512))) {
+	if (!(f->mem = pool_create("text pv_names", 512))) {
 		stack;
 		goto bad;
 	}

lib/format_text/format-text.c

@@ -88,8 +88,11 @@ static int _lv_setup(struct format_instance *fid, struct logical_volume *lv)
 	}
 */
 
-	if (!*lv->lvid.s)
-		lvid_create(&lv->lvid, &lv->vg->id);
+	if (!*lv->lvid.s && !lvid_create(&lv->lvid, &lv->vg->id)) {
+		log_error("Random lvid creation failed for %s/%s.",
+			  lv->vg->name, lv->name);
+		return 0;
+	}
 
 	return 1;
 }
@@ -185,6 +188,7 @@ static struct raw_locn *_find_vg_rlocn(struct device_area *dev_area,
 	size_t len;
 	char vgnamebuf[NAME_LEN + 2];
 	struct raw_locn *rlocn;
+	struct lvmcache_info *info;
 
 	rlocn = mdah->raw_locns;
 
@@ -193,7 +197,7 @@ static struct raw_locn *_find_vg_rlocn(struct device_area *dev_area,
 		if (!dev_read(dev_area->dev, dev_area->start + rlocn->offset,
 			      sizeof(vgnamebuf), vgnamebuf)) {
 			stack;
-			return NULL;
+			goto error;
 		}
 		if (!strncmp(vgnamebuf, vgname, len = strlen(vgname)) &&
 		    (isspace(vgnamebuf[len]) || vgnamebuf[len] == '{')) {
@@ -202,6 +206,10 @@ static struct raw_locn *_find_vg_rlocn(struct device_area *dev_area,
 		rlocn++;
 	}
 
+      error:
+	if ((info = info_from_pvid(dev_area->dev->pvid)))
+		lvmcache_update_vgname(info, ORPHAN);
+
 	return NULL;
 }
 
@@ -261,7 +269,7 @@ static struct volume_group *_vg_read_raw_area(struct format_instance *fid,
 	}
 
 	if (!(rlocn = _vg_posn(fid, area, vgname))) {
-		stack;
+		log_debug("VG %s not found on %s", vgname, dev_name(area->dev));
 		goto out;
 	}
 
@@ -1180,9 +1188,9 @@ static int _pv_read(const struct format_type *fmt, const char *pv_name,
 		return 1;
 	}
 
-	/* Perform full scan and try again */
-	if (!memlock()) {
-		lvmcache_label_scan(fmt->cmd, 1);
+	/* Perform full scan (just the first time) and try again */
+	if (!memlock() && !full_scan_done()) {
+		lvmcache_label_scan(fmt->cmd, 2);
 
 		if (info->vginfo && info->vginfo->vgname &&
 		    *info->vginfo->vgname &&

lib/format_text/import_vsn1.c

@@ -31,7 +31,7 @@ typedef int (*section_fn) (struct format_instance * fid, struct pool * mem,
 			   struct hash_table * pv_hash);
 
 #define _read_int32(root, path, result) \
-	get_config_uint32(root, path, result)
+	get_config_uint32(root, path, (uint32_t *) result)
 
 #define _read_uint32(root, path, result) \
 	get_config_uint32(root, path, result)

lib/label/label.c

@@ -117,12 +117,17 @@ static struct labeller *_find_labeller(struct device *dev, char *buf,
 	struct labeller_i *li;
 	struct labeller *r = NULL;
 	struct label_header *lh;
+	struct lvmcache_info *info;
 	uint64_t sector;
 	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;
 	}
 
@@ -182,10 +187,13 @@ static struct labeller *_find_labeller(struct device *dev, char *buf,
 		}
 	}
 
-	if (!found)
-		log_very_verbose("%s: No label detected", dev_name(dev));
-
       out:
+	if (!found) {
+		if ((info = info_from_pvid(dev->pvid)))
+			lvmcache_update_vgname(info, ORPHAN);
+		log_very_verbose("%s: No label detected", dev_name(dev));
+	}
+
 	if (!dev_close(dev))
 		stack;
 
@@ -279,7 +287,7 @@ int label_read(struct device *dev, struct label **result)
 		return 0;
 	}
 
-	if ((r = l->ops->read(l, dev, buf, result)) && result && *result)
+	if ((r = (l->ops->read)(l, dev, buf, result)) && result && *result)
 		(*result)->sector = sector;
 
 	return r;
@@ -309,7 +317,7 @@ int label_write(struct device *dev, struct label *label)
 	lh->sector_xl = xlate64(label->sector);
 	lh->offset_xl = xlate32(sizeof(*lh));
 
-	if (!label->labeller->ops->write(label, buf)) {
+	if (!(label->labeller->ops->write)(label, buf)) {
 		stack;
 		return 0;
 	}

lib/locking/cluster_locking.c

@@ -139,8 +139,7 @@ static int _send_request(char *inbuf, int inlen, char **retbuf)
 
 	/* Read the returned values */
 	off = 1;		/* we've already read the first byte */
-
-	while (off < outheader->arglen && len > 0) {
+	while (off <= outheader->arglen && len > 0) {
 		len = read(_clvmd_sock, outheader->args + off,
 			   buflen - off - offsetof(struct clvm_header, args));
 		if (len > 0)
@@ -148,10 +147,16 @@ static int _send_request(char *inbuf, int inlen, char **retbuf)
 	}
 
 	/* Was it an error ? */
-	if (outheader->status < 0) {
-		errno = -outheader->status;
-		log_error("cluster send request failed: %s", strerror(errno));
-		return 0;
+	if (outheader->status != 0) {
+		errno = outheader->status;
+
+		/* Only return an error here if there are no node-specific
+		   errors present in the message that might have more detail */
+		if (!(outheader->flags & CLVMD_FLAG_NODEERRS)) {
+			log_error("cluster request failed: %s", strerror(errno));
+			return 0;
+		}
+
 	}
 
 	return 1;
@@ -329,7 +334,7 @@ static int _lock_for_cluster(unsigned char cmd, unsigned int flags, char *name)
 	 * VG locks are just that: locks, and have no side effects
 	 * so we only need to do them on the local node because all
 	 * locks are cluster-wide.
-	 * Also, if the lock is exclusive it makes no sense to try to 
+	 * Also, if the lock is exclusive it makes no sense to try to
 	 * acquire it on all nodes, so just do that on the local node too.
 	 */
 	if (cmd == CLVMD_CMD_LOCK_VG ||
@@ -342,10 +347,11 @@ static int _lock_for_cluster(unsigned char cmd, unsigned int flags, char *name)
 
 	/* If any nodes were down then display them and return an error */
 	for (i = 0; i < num_responses; i++) {
-		if (response[i].status == -EHOSTDOWN) {
+		if (response[i].status == EHOSTDOWN) {
 			log_error("clvmd not running on node %s",
 				  response[i].node);
 			status = 0;
+			errno = response[i].status;
 		} else if (response[i].status) {
 			log_error("Error locking on node %s: %s",
 				  response[i].node,
@@ -353,6 +359,7 @@ static int _lock_for_cluster(unsigned char cmd, unsigned int flags, char *name)
 				  	response[i].response :
 				  	strerror(response[i].status));
 			status = 0;
+			errno = response[i].status;
 		}
 	}
 
@@ -439,7 +446,7 @@ int init_cluster_locking(struct locking_type *locking, struct config_tree *cft)
 	locking->lock_resource = _lock_resource;
 	locking->fin_locking = _locking_end;
 	locking->reset_locking = _reset_locking;
-	locking->flags = LCK_PRE_MEMLOCK;
+	locking->flags = LCK_PRE_MEMLOCK | LCK_CLUSTERED;
 
 	_clvmd_sock = _open_local_sock();
 	if (_clvmd_sock == -1)
@@ -456,6 +463,7 @@ int locking_init(int type, struct config_tree *cf, uint32_t *flags)
 
 	/* Ask LVM to lock memory before calling us */
 	*flags |= LCK_PRE_MEMLOCK;
+	*flags |= LCK_CLUSTERED;
 
 	return 1;
 }

lib/locking/file_locking.c

@@ -41,7 +41,7 @@ static char _lock_dir[NAME_LEN];
 
 static sig_t _oldhandler;
 static sigset_t _fullsigset, _intsigset;
-static int _handler_installed;
+static volatile sig_atomic_t _handler_installed;
 
 static int _release_lock(const char *file, int unlock)
 {
@@ -95,38 +95,40 @@ static void _reset_file_locking(void)
 static void _remove_ctrl_c_handler()
 {
 	siginterrupt(SIGINT, 0);
-	if (!_handler_installed || _oldhandler == SIG_ERR)
+	if (!_handler_installed)
 		return;
 
+	_handler_installed = 0;
+
 	sigprocmask(SIG_SETMASK, &_fullsigset, NULL);
 	if (signal(SIGINT, _oldhandler) == SIG_ERR)
 		log_sys_error("signal", "_remove_ctrl_c_handler");
-
-	_handler_installed = 0;
 }
 
 static void _trap_ctrl_c(int sig)
 {
 	_remove_ctrl_c_handler();
 	log_error("CTRL-c detected: giving up waiting for lock");
-	return;
 }
 
 static void _install_ctrl_c_handler()
 {
-	if ((_oldhandler = signal(SIGINT, _trap_ctrl_c)) == SIG_ERR)
+	_handler_installed = 1;
+
+	if ((_oldhandler = signal(SIGINT, _trap_ctrl_c)) == SIG_ERR) {
+		_handler_installed = 0;
 		return;
+	}
 
 	sigprocmask(SIG_SETMASK, &_intsigset, NULL);
 	siginterrupt(SIGINT, 1);
-
-	_handler_installed = 1;
 }
 
 static int _lock_file(const char *file, int flags)
 {
 	int operation;
 	int r = 1;
+	int old_errno;
 
 	struct lock_list *ll;
 	struct stat buf1, buf2;
@@ -176,10 +178,12 @@ static int _lock_file(const char *file, int flags)
 			_install_ctrl_c_handler();
 
 		r = flock(ll->lf, operation);
+		old_errno = errno;
 		if (!(flags & LCK_NONBLOCK))
 			_remove_ctrl_c_handler();
 
 		if (r) {
+			errno = old_errno;
 			log_sys_error("flock", ll->res);
 			goto err;
 		}

lib/locking/locking.c

@@ -124,6 +124,8 @@ static inline void _update_vg_lock_count(int flags)
  */
 int init_locking(int type, struct config_tree *cft)
 {
+	init_lockingfailed(0);
+
 	switch (type) {
 	case 0:
 		init_no_locking(&_locking, cft);
@@ -165,6 +167,7 @@ int init_locking(int type, struct config_tree *cft)
 	log_verbose("Locking disabled - only read operations permitted.");
 
 	init_no_locking(&_locking, cft);
+	init_lockingfailed(1);
 
 	return 1;
 }
@@ -321,3 +324,9 @@ int vg_write_lock_held(void)
 {
 	return _vg_write_lock_held;
 }
+
+int locking_is_clustered(void)
+{
+	return (_locking.flags & LCK_CLUSTERED) ? 1 : 0;
+}
+

lib/locking/locking.h

@@ -20,6 +20,7 @@ int init_locking(int type, struct config_tree *cf);
 void fin_locking(void);
 void reset_locking(void);
 int vg_write_lock_held(void);
+int locking_is_clustered(void);
 
 /*
  * LCK_VG:

lib/locking/locking_types.h

@@ -23,6 +23,7 @@ typedef void (*fin_lock_fn) (void);
 typedef void (*reset_lock_fn) (void);
 
 #define LCK_PRE_MEMLOCK	0x00000001	/* Is memlock() needed before calls? */
+#define LCK_CLUSTERED	0x00000002
 
 struct locking_type {
 	uint32_t flags;

lib/log/log.c

@@ -28,7 +28,9 @@ static struct str_list _log_dev_alias;
 static int _verbose_level = VERBOSE_BASE_LEVEL;
 static int _test = 0;
 static int _partial = 0;
+static int _md_filtering = 0;
 static int _pvmove = 0;
+static int _full_scan_done = 0;	/* Restrict to one full scan during each cmd */
 static int _debug_level = 0;
 static int _syslog = 0;
 static int _log_to_file = 0;
@@ -38,6 +40,7 @@ static int _indent = 1;
 static int _log_cmd_name = 0;
 static int _log_suppress = 0;
 static int _ignorelockingfailure = 0;
+static int _lockingfailed = 0;
 static int _security_level = SECURITY_LEVEL;
 static char _cmd_name[30] = "";
 static char _msg_prefix[30] = "  ";
@@ -138,16 +141,31 @@ void init_partial(int level)
 	_partial = level;
 }
 
+void init_md_filtering(int level)
+{
+	_md_filtering = level;
+}
+
 void init_pvmove(int level)
 {
 	_pvmove = level;
 }
 
+void init_full_scan_done(int level)
+{
+	_full_scan_done = level;
+}
+
 void init_ignorelockingfailure(int level)
 {
 	_ignorelockingfailure = level;
 }
 
+void init_lockingfailed(int level)
+{
+	_lockingfailed = level;
+}
+
 void init_security_level(int level)
 {
 	_security_level = level;
@@ -187,11 +205,26 @@ int partial_mode()
 	return _partial;
 }
 
+int md_filtering()
+{
+	return _md_filtering;
+}
+
 int pvmove_mode()
 {
 	return _pvmove;
 }
 
+int full_scan_done()
+{
+	return _full_scan_done;
+}
+
+int lockingfailed()
+{
+	return _lockingfailed;
+}
+
 int ignorelockingfailure()
 {
 	return _ignorelockingfailure;
@@ -215,7 +248,7 @@ int debug_level()
 void print_log(int level, const char *file, int line, const char *format, ...)
 {
 	va_list ap;
-	char buf[1024], buf2[4096];
+	char buf[1024], buf2[4096], locn[4096];
 	int bufused, n;
 	const char *message;
 	const char *trformat;		/* Translated format string */
@@ -243,36 +276,46 @@ void print_log(int level, const char *file, int line, const char *format, ...)
 
       log_it:
 	if (!_log_suppress) {
+		if (_verbose_level > _LOG_DEBUG)
+			lvm_snprintf(locn, sizeof(locn), "#%s:%d ",
+				     file, line);
+		else
+			locn[0] = '\0';
+
 		va_start(ap, format);
 		switch (level) {
 		case _LOG_DEBUG:
-			if (!strcmp("<backtrace>", format))
+			if (!strcmp("<backtrace>", format) &&
+			    _verbose_level <= _LOG_DEBUG)
 				break;
 			if (_verbose_level >= _LOG_DEBUG) {
-				printf("%s%s", _cmd_name, _msg_prefix);
+				fprintf(stderr, "%s%s%s", locn, _cmd_name,
+					_msg_prefix);
 				if (_indent)
-					printf("      ");
-				vprintf(trformat, ap);
-				putchar('\n');
+					fprintf(stderr, "      ");
+				vfprintf(stderr, trformat, ap);
+				fputc('\n', stderr);
 			}
 			break;
 
 		case _LOG_INFO:
 			if (_verbose_level >= _LOG_INFO) {
-				printf("%s%s", _cmd_name, _msg_prefix);
+				fprintf(stderr, "%s%s%s", locn, _cmd_name,
+					_msg_prefix);
 				if (_indent)
-					printf("    ");
-				vprintf(trformat, ap);
-				putchar('\n');
+					fprintf(stderr, "    ");
+				vfprintf(stderr, trformat, ap);
+				fputc('\n', stderr);
 			}
 			break;
 		case _LOG_NOTICE:
 			if (_verbose_level >= _LOG_NOTICE) {
-				printf("%s%s", _cmd_name, _msg_prefix);
+				fprintf(stderr, "%s%s%s", locn, _cmd_name,
+					_msg_prefix);
 				if (_indent)
-					printf("  ");
-				vprintf(trformat, ap);
-				putchar('\n');
+					fprintf(stderr, "  ");
+				vfprintf(stderr, trformat, ap);
+				fputc('\n', stderr);
 			}
 			break;
 		case _LOG_WARN:
@@ -284,7 +327,8 @@ void print_log(int level, const char *file, int line, const char *format, ...)
 			break;
 		case _LOG_ERR:
 			if (_verbose_level >= _LOG_ERR) {
-				fprintf(stderr, "%s%s", _cmd_name, _msg_prefix);
+				fprintf(stderr, "%s%s%s", locn, _cmd_name,
+					_msg_prefix);
 				vfprintf(stderr, trformat, ap);
 				fputc('\n', stderr);
 			}
@@ -292,7 +336,8 @@ void print_log(int level, const char *file, int line, const char *format, ...)
 		case _LOG_FATAL:
 		default:
 			if (_verbose_level >= _LOG_FATAL) {
-				fprintf(stderr, "%s%s", _cmd_name, _msg_prefix);
+				fprintf(stderr, "%s%s%s", locn, _cmd_name,
+					_msg_prefix);
 				vfprintf(stderr, trformat, ap);
 				fputc('\n', stderr);
 			}

lib/log/log.h

@@ -63,21 +63,27 @@ void fin_syslog(void);
 void init_verbose(int level);
 void init_test(int level);
 void init_partial(int level);
+void init_md_filtering(int level);
 void init_pvmove(int level);
+void init_full_scan_done(int level);
 void init_debug(int level);
 void init_cmd_name(int status);
 void init_msg_prefix(const char *prefix);
 void init_indent(int indent);
 void init_ignorelockingfailure(int level);
+void init_lockingfailed(int level);
 void init_security_level(int level);
 
 void set_cmd_name(const char *cmd_name);
 
 int test_mode(void);
 int partial_mode(void);
+int md_filtering(void);
 int pvmove_mode(void);
+int full_scan_done(void);
 int debug_level(void);
 int ignorelockingfailure(void);
+int lockingfailed(void);
 int security_level(void);
 
 /* Suppress messages to stdout/stderr */

lib/metadata/lv_manip.c

@@ -458,7 +458,7 @@ static int _allocate(struct volume_group *vg, struct logical_volume *lv,
 	if (segtype->flags & SEG_VIRTUAL)
 		return _alloc_virtual(lv, allocated, segtype);
 
-	if (!(scratch = pool_create(1024))) {
+	if (!(scratch = pool_create("allocation", 1024))) {
 		stack;
 		return 0;
 	}

lib/metadata/metadata.c

@@ -267,10 +267,13 @@ struct physical_volume *pv_create(const struct format_type *fmt,
 		return NULL;
 	}
 
-	if (!id)
-		id_create(&pv->id);
-	else
+	if (id)
 		memcpy(&pv->id, id, sizeof(*id));
+	else if (!id_create(&pv->id)) {
+		log_error("Failed to create random uuid for %s.",
+			  dev_name(dev));
+		return NULL;
+	}
 
 	pv->dev = dev;
 
@@ -651,7 +654,7 @@ struct volume_group *vg_read(struct cmd_context *cmd, const char *vgname,
 				stack;
 				return NULL;
 			}
-			lvmcache_label_scan(cmd, 1);
+			lvmcache_label_scan(cmd, 2);
 			if (!(fmt = fmt_from_vgname(vgname))) {
 				stack;
 				return NULL;
@@ -684,10 +687,46 @@ struct volume_group *vg_read(struct cmd_context *cmd, const char *vgname,
 		}
 	}
 
-	/* Failed to find VG */
+	/* Failed to find VG where we expected it - full scan and retry */
 	if (!correct_vg) {
-		stack;
-		return NULL;
+		inconsistent = 0;
+
+		lvmcache_label_scan(cmd, 2);
+		if (!(fmt = fmt_from_vgname(vgname))) {
+			stack;
+			return NULL;
+		}
+
+		/* create format instance with appropriate metadata area */
+		if (!(fid = fmt->ops->create_instance(fmt, vgname, NULL))) {
+			log_error("Failed to create format instance");
+			return NULL;
+		}
+
+		/* Ensure contents of all metadata areas match - else recover */
+		list_iterate(mdah, &fid->metadata_areas) {
+			mda = list_item(mdah, struct metadata_area);
+			if (!(vg = mda->ops->vg_read(fid, vgname, mda))) {
+				inconsistent = 1;
+				continue;
+			}
+			if (!correct_vg) {
+				correct_vg = vg;
+				continue;
+			}
+			/* FIXME Also ensure contents same - checksums same? */
+			if (correct_vg->seqno != vg->seqno) {
+				inconsistent = 1;
+				if (vg->seqno > correct_vg->seqno)
+					correct_vg = vg;
+			}
+		}
+
+		/* Give up looking */
+		if (!correct_vg) {
+			stack;
+			return NULL;
+		}
 	}
 
 	lvmcache_update_vg(correct_vg);
@@ -711,6 +750,11 @@ struct volume_group *vg_read(struct cmd_context *cmd, const char *vgname,
 			log_error("Automatic metadata correction failed");
 			return NULL;
 		}
+		if (!vg_commit(correct_vg)) {
+			log_error("Automatic metadata correction commit "
+				  "failed");
+			return NULL;
+		}
 	}
 
 	if ((correct_vg->status & PVMOVE) && !pvmove_mode()) {
@@ -760,7 +804,7 @@ struct volume_group *vg_read_by_vgid(struct cmd_context *cmd, const char *vgid)
 	 *       allowed to do a full scan here any more. */
 
 	// The slow way - full scan required to cope with vgrename 
-	if (!(vgnames = get_vgs(cmd, 1))) {
+	if (!(vgnames = get_vgs(cmd, 2))) {
 		log_error("vg_read_by_vgid: get_vgs failed");
 		return NULL;
 	}

lib/metadata/metadata.h

@@ -255,6 +255,7 @@ struct snapshot {
 
 	int persistent;		/* boolean */
 	uint32_t chunk_size;	/* in 512 byte sectors */
+	uint32_t le_count;
 
 	struct logical_volume *origin;
 	struct logical_volume *cow;
@@ -496,9 +497,9 @@ struct snapshot *find_cow(const struct logical_volume *lv);
 struct snapshot *find_origin(const struct logical_volume *lv);
 struct list *find_snapshots(const struct logical_volume *lv);
 
-int vg_add_snapshot(struct logical_volume *origin,
-		    struct logical_volume *cow,
-		    int persistent, struct id *id, uint32_t chunk_size);
+int vg_add_snapshot(struct logical_volume *origin, struct logical_volume *cow,
+		    int persistent, struct id *id, uint32_t extent_count,
+		    uint32_t chunk_size);
 
 int vg_remove_snapshot(struct volume_group *vg, struct logical_volume *cow);
 

lib/metadata/snapshot_manip.c

@@ -104,9 +104,9 @@ struct list *find_snapshots(const struct logical_volume *lv)
 	return snaplist;
 }
 
-int vg_add_snapshot(struct logical_volume *origin,
-		    struct logical_volume *cow,
-		    int persistent, struct id *id, uint32_t chunk_size)
+int vg_add_snapshot(struct logical_volume *origin, struct logical_volume *cow,
+		    int persistent, struct id *id, uint32_t extent_count,
+		    uint32_t chunk_size)
 {
 	struct snapshot *s;
 	struct snapshot_list *sl;
@@ -127,13 +127,15 @@ int vg_add_snapshot(struct logical_volume *origin,
 
 	s->persistent = persistent;
 	s->chunk_size = chunk_size;
+	s->le_count = extent_count;
 	s->origin = origin;
 	s->cow = cow;
 
 	if (id)
 		s->id = *id;
 	else if (!id_create(&s->id)) {
-		log_error("Snapshot UUID creation failed");
+		log_error("Random UUID creation failed for snapshot %s.",
+			  cow->name);
 		return 0;
 	}
 

lib/misc/lvm-file.c

@@ -86,7 +86,11 @@ int lvm_rename(const char *old, const char *new)
 {
 	struct stat buf;
 
-	link(old, new);
+	if (link(old, new)) {
+		log_error("%s: rename to %s failed: %s", old, new,
+			  strerror(errno));
+		return 0;
+	}
 
 	if (stat(old, &buf)) {
 		log_sys_error("stat", old);

lib/misc/lvm-string.c

@@ -134,17 +134,19 @@ static void _quote_hyphens(char **out, const char *src)
 char *build_dm_name(struct pool *mem, const char *vg,
 		    const char *lv, const char *layer)
 {
-	size_t len = 0;
-	int hyphens = 0;
+	size_t len = 1;
+	int hyphens = 1;
 	char *r, *out;
 
 	_count_hyphens(vg, &len, &hyphens);
 	_count_hyphens(lv, &len, &hyphens);
 
-	if (layer && *layer)
+	if (layer && *layer) {
 		_count_hyphens(layer, &len, &hyphens);
+		hyphens++;
+	}
 
-	len += hyphens + 2;
+	len += hyphens;
 
 	if (!(r = pool_alloc(mem, len))) {
 		stack;

lib/mm/dbg_malloc.c

@@ -19,6 +19,8 @@
 
 #include <stdarg.h>
 
+#ifdef DEBUG_MEM
+
 struct memblock {
 	struct memblock *prev, *next;	/* All allocated blocks are linked */
 	size_t length;		/* Size of the requested block */
@@ -46,13 +48,14 @@ void *malloc_aux(size_t s, const char *file, int line)
 	size_t tsize = s + sizeof(*nb) + sizeof(unsigned long);
 
 	if (s > 50000000) {
-		log_error("Huge memory allocation (size %" PRIuPTR
-			  ") rejected - bug?", s);
+		log_error("Huge memory allocation (size %" PRIsize_t
+			  ") rejected - metadata corruption?", s);
 		return 0;
 	}
 
 	if (!(nb = malloc(tsize))) {
-		log_error("couldn't allocate any memory, size = %" PRIuPTR, s);
+		log_error("couldn't allocate any memory, size = %" PRIsize_t,
+			  s);
 		return 0;
 	}
 
@@ -166,7 +169,6 @@ void *realloc_aux(void *p, unsigned int s, const char *file, int line)
 	return r;
 }
 
-#ifdef DEBUG_MEM
 int dump_memory(void)
 {
 	unsigned long tot = 0;
@@ -189,7 +191,7 @@ int dump_memory(void)
 		str[sizeof(str) - 1] = '\0';
 
 		print_log(_LOG_INFO, mb->file, mb->line,
-			  "block %d at %p, size %" PRIdPTR "\t [%s]",
+			  "block %d at %p, size %" PRIsize_t "\t [%s]",
 			  mb->id, mb->magic, mb->length, str);
 		tot += mb->length;
 	}
@@ -213,10 +215,18 @@ void bounds_check(void)
 		mb = mb->next;
 	}
 }
-#endif
 
-/*
- * Local variables:
- * c-file-style: "linux"
- * End:
- */
+#else
+
+void *malloc_aux(size_t s, const char *file, int line)
+{
+	if (s > 50000000) {
+		log_error("Huge memory allocation (size %" PRIsize_t
+			  ") rejected - metadata corruption?", s);
+		return 0;
+	}
+
+	return malloc(s);
+}
+
+#endif

lib/mm/dbg_malloc.h

@@ -20,22 +20,26 @@
 #include <stdlib.h>
 #include <string.h>
 
-#ifdef DEBUG_MEM
 void *malloc_aux(size_t s, const char *file, int line);
+#  define dbg_malloc(s) malloc_aux((s), __FILE__, __LINE__)
+
+#ifdef DEBUG_MEM
+
 void free_aux(void *p);
 void *realloc_aux(void *p, unsigned int s, const char *file, int line);
 int dump_memory(void);
 void bounds_check(void);
 
-#  define dbg_malloc(s) malloc_aux((s), __FILE__, __LINE__)
 #  define dbg_free(p) free_aux(p)
 #  define dbg_realloc(p, s) realloc_aux(p, s, __FILE__, __LINE__)
+
 #else
-#  define dbg_malloc(s) malloc(s)
+
 #  define dbg_free(p) free(p)
 #  define dbg_realloc(p, s) realloc(p, s)
 #  define dump_memory()
 #  define bounds_check()
+
 #endif
 
 static inline char *dbg_strdup(const char *str)

lib/mm/pool-debug.c

@@ -22,38 +22,64 @@ struct block {
 	void *data;
 };
 
+typedef struct {
+	unsigned block_serialno;	/* Non-decreasing serialno of block */
+	unsigned blocks_allocated;	/* Current number of blocks allocated */
+	unsigned blocks_max;	/* Max no of concurrently-allocated blocks */
+	unsigned int bytes, maxbytes;
+} pool_stats;
+
 struct pool {
+	const char *name;
+
 	int begun;
 	struct block *object;
 
 	struct block *blocks;
 	struct block *tail;
+
+	pool_stats stats;
 };
 
 /* by default things come out aligned for doubles */
 #define DEFAULT_ALIGNMENT __alignof__ (double)
 
-struct pool *pool_create(size_t chunk_hint)
+struct pool *pool_create(const char *name, size_t chunk_hint)
 {
 	struct pool *mem = dbg_malloc(sizeof(*mem));
 
 	if (!mem) {
-		log_error("Couldn't create memory pool (size %" PRIsize_t ")",
-			  sizeof(*mem));
+		log_error("Couldn't create memory pool %s (size %"
+			  PRIsize_t ")", name, sizeof(*mem));
 		return NULL;
 	}
 
+	mem->name = name;
 	mem->begun = 0;
 	mem->object = 0;
 	mem->blocks = mem->tail = NULL;
+
+	mem->stats.block_serialno = 0;
+	mem->stats.blocks_allocated = 0;
+	mem->stats.blocks_max = 0;
+	mem->stats.bytes = 0;
+	mem->stats.maxbytes = 0;
+
+#ifdef DEBUG_POOL
+	log_debug("Created mempool %s", name);
+#endif
+
 	return mem;
 }
 
-static void _free_blocks(struct block *b)
+static void _free_blocks(struct pool *p, struct block *b)
 {
 	struct block *n;
 
 	while (b) {
+		p->stats.bytes -= b->size;
+		p->stats.blocks_allocated--;
+
 		n = b->next;
 		dbg_free(b->data);
 		dbg_free(b);
@@ -61,9 +87,22 @@ static void _free_blocks(struct block *b)
 	}
 }
 
+static void _pool_stats(struct pool *p, const char *action)
+{
+#ifdef DEBUG_POOL
+	log_debug("%s mempool %s: %u/%u bytes, %u/%u blocks, "
+		  "%u allocations)", action, p->name, p->stats.bytes,
+		  p->stats.maxbytes, p->stats.blocks_allocated,
+		  p->stats.blocks_max, p->stats.block_serialno);
+#else
+	;
+#endif
+}
+
 void pool_destroy(struct pool *p)
 {
-	_free_blocks(p->blocks);
+	_pool_stats(p, "Destroying");
+	_free_blocks(p, p->blocks);
 	dbg_free(p);
 }
 
@@ -79,6 +118,15 @@ static void _append_block(struct pool *p, struct block *b)
 		p->tail = b;
 	} else
 		p->blocks = p->tail = b;
+
+	p->stats.block_serialno++;
+	p->stats.blocks_allocated++;
+	if (p->stats.blocks_allocated > p->stats.blocks_max)
+		p->stats.blocks_max = p->stats.blocks_allocated;
+
+	p->stats.bytes += b->size;
+	if (p->stats.bytes > p->stats.maxbytes)
+		p->stats.maxbytes = p->stats.bytes;
 }
 
 static struct block *_new_block(size_t s, unsigned alignment)
@@ -121,12 +169,14 @@ void *pool_alloc_aligned(struct pool *p, size_t s, unsigned alignment)
 		return NULL;
 
 	_append_block(p, b);
+
 	return b->data;
 }
 
 void pool_empty(struct pool *p)
 {
-	_free_blocks(p->blocks);
+	_pool_stats(p, "Emptying");
+	_free_blocks(p, p->blocks);
 	p->blocks = p->tail = NULL;
 }
 
@@ -134,6 +184,8 @@ void pool_free(struct pool *p, void *ptr)
 {
 	struct block *b, *prev = NULL;
 
+	_pool_stats(p, "Freeing (before)");
+
 	for (b = p->blocks; b; b = b->next) {
 		if (b->data == ptr)
 			break;
@@ -147,13 +199,15 @@ void pool_free(struct pool *p, void *ptr)
 	 */
 	assert(b);
 
-	_free_blocks(b);
+	_free_blocks(p, b);
 
 	if (prev) {
 		p->tail = prev;
 		prev->next = NULL;
 	} else
 		p->blocks = p->tail = NULL;
+
+	_pool_stats(p, "Freeing (after)");
 }
 
 int pool_begin_object(struct pool *p, size_t init_size)

lib/mm/pool-fast.c

@@ -36,14 +36,14 @@ struct chunk *_new_chunk(struct pool *p, size_t s);
 /* by default things come out aligned for doubles */
 #define DEFAULT_ALIGNMENT __alignof__ (double)
 
-struct pool *pool_create(size_t chunk_hint)
+struct pool *pool_create(const char *name, size_t chunk_hint)
 {
 	size_t new_size = 1024;
 	struct pool *p = dbg_malloc(sizeof(*p));
 
 	if (!p) {
-		log_error("Couldn't create memory pool (size %" PRIsize_t ")",
-			  sizeof(*p));
+		log_error("Couldn't create memory pool %s (size %"
+			  PRIsize_t ")", name, sizeof(*p));
 		return 0;
 	}
 	memset(p, 0, sizeof(*p));

lib/mm/pool.h

@@ -55,7 +55,7 @@
 struct pool;
 
 /* constructor and destructor */
-struct pool *pool_create(size_t chunk_hint);
+struct pool *pool_create(const char *name, size_t chunk_hint);
 void pool_destroy(struct pool *p);
 
 /* simple allocation/free routines */

lib/mm/xlate.h

@@ -27,14 +27,14 @@
 		       ((x) & 0xff000000U) >> 24 | \
 		       ((x) & 0x0000ff00U) << 8  | \
 		       ((x) & 0x00ff0000U) >> 8)
-#  define bswap_64(x) (((x) & 0x00000000000000ffU) << 56 | \
-		       ((x) & 0xff00000000000000U) >> 56 | \
-		       ((x) & 0x000000000000ff00U) << 40 | \
-		       ((x) & 0x00ff000000000000U) >> 40 | \
-		       ((x) & 0x0000000000ff0000U) << 24 | \
-		       ((x) & 0x0000ff0000000000U) >> 24 | \
-		       ((x) & 0x00000000ff000000U) << 8 | \
-		       ((x) & 0x000000ff00000000U) >> 8)
+#  define bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \
+		       ((x) & 0xff00000000000000ULL) >> 56 | \
+		       ((x) & 0x000000000000ff00ULL) << 40 | \
+		       ((x) & 0x00ff000000000000ULL) >> 40 | \
+		       ((x) & 0x0000000000ff0000ULL) << 24 | \
+		       ((x) & 0x0000ff0000000000ULL) >> 24 | \
+		       ((x) & 0x00000000ff000000ULL) << 8 | \
+		       ((x) & 0x000000ff00000000ULL) >> 8)
 #endif
 
 #if BYTE_ORDER == LITTLE_ENDIAN

lib/regex/matcher.c

@@ -275,7 +275,7 @@ struct matcher *matcher_create(struct pool *mem, const char **patterns,
 	int i;
 	size_t len = 0;
 	struct rx_node *rx;
-	struct pool *scratch = pool_create(10 * 1024);
+	struct pool *scratch = pool_create("regex matcher", 10 * 1024);
 	struct matcher *m;
 
 	if (!scratch) {

lib/report/report.c

@@ -295,7 +295,7 @@ static int _lvkmaj_disp(struct report_handle *rh, struct field *field,
 	struct lvinfo info;
 	uint64_t minusone = UINT64_C(-1);
 
-	if (lv_info(lv, &info) && info.exists)
+	if (lv_info(lv, &info, 0) && info.exists)
 		return _int_disp(rh, field, &info.major);
 	else
 		return _int_disp(rh, field, &minusone);
@@ -310,7 +310,7 @@ static int _lvkmin_disp(struct report_handle *rh, struct field *field,
 	struct lvinfo info;
 	uint64_t minusone = UINT64_C(-1);
 
-	if (lv_info(lv, &info) && info.exists)
+	if (lv_info(lv, &info, 0) && info.exists)
 		return _int_disp(rh, field, &info.minor);
 	else
 		return _int_disp(rh, field, &minusone);
@@ -362,7 +362,7 @@ static int _lvstatus_disp(struct report_handle *rh, struct field *field,
 	else
 		repstr[3] = '-';
 
-	if (lv_info(lv, &info) && info.exists) {
+	if (lv_info(lv, &info, 1) && info.exists) {
 		if (info.suspended)
 			repstr[4] = 's';	/* Suspended */
 		else
@@ -573,7 +573,7 @@ static int _vgsize_disp(struct report_handle *rh, struct field *field,
 	const struct volume_group *vg = (const struct volume_group *) data;
 	uint64_t size;
 
-	size = vg->extent_count * vg->extent_size;
+	size = (uint64_t) vg->extent_count * vg->extent_size;
 
 	return _size64_disp(rh, field, &size);
 }
@@ -584,7 +584,7 @@ static int _segstart_disp(struct report_handle *rh, struct field *field,
 	const struct lv_segment *seg = (const struct lv_segment *) data;
 	uint64_t start;
 
-	start = seg->le * seg->lv->vg->extent_size;
+	start = (uint64_t) seg->le * seg->lv->vg->extent_size;
 
 	return _size64_disp(rh, field, &start);
 }
@@ -595,7 +595,7 @@ static int _segsize_disp(struct report_handle *rh, struct field *field,
 	const struct lv_segment *seg = (const struct lv_segment *) data;
 	uint64_t size;
 
-	size = seg->len * seg->lv->vg->extent_size;
+	size = (uint64_t) seg->len * seg->lv->vg->extent_size;
 
 	return _size64_disp(rh, field, &size);
 }
@@ -610,7 +610,7 @@ static int _pvused_disp(struct report_handle *rh, struct field *field,
 	if (!pv->pe_count)
 		used = 0LL;
 	else
-		used = pv->pe_alloc_count * pv->pe_size;
+		used = (uint64_t) pv->pe_alloc_count * pv->pe_size;
 
 	return _size64_disp(rh, field, &used);
 }
@@ -625,7 +625,7 @@ static int _pvfree_disp(struct report_handle *rh, struct field *field,
 	if (!pv->pe_count)
 		freespace = pv->size;
 	else
-		freespace = (pv->pe_count - pv->pe_alloc_count) * pv->pe_size;
+		freespace = (uint64_t) (pv->pe_count - pv->pe_alloc_count) * pv->pe_size;
 
 	return _size64_disp(rh, field, &freespace);
 }
@@ -640,7 +640,7 @@ static int _pvsize_disp(struct report_handle *rh, struct field *field,
 	if (!pv->pe_count)
 		size = pv->size;
 	else
-		size = pv->pe_count * pv->pe_size;
+		size = (uint64_t) pv->pe_count * pv->pe_size;
 
 	return _size64_disp(rh, field, &size);
 }
@@ -663,7 +663,7 @@ static int _vgfree_disp(struct report_handle *rh, struct field *field,
 	const struct volume_group *vg = (const struct volume_group *) data;
 	uint64_t freespace;
 
-	freespace = vg->free_count * vg->extent_size;
+	freespace = (uint64_t) vg->free_count * vg->extent_size;
 
 	return _size64_disp(rh, field, &freespace);
 }
@@ -774,7 +774,7 @@ static int _snpercent_disp(struct report_handle *rh, struct field *field,
 	}
 
 	if (!(snap = find_cow(lv)) ||
-	    (lv_info(snap->cow, &info) && !info.exists)) {
+	    (lv_info(snap->cow, &info, 0) && !info.exists)) {
 		field->report_string = "";
 		*sortval = UINT64_C(0);
 		field->sort_value = sortval;
@@ -1081,7 +1081,7 @@ void *report_init(struct cmd_context *cmd, const char *format, const char *keys,
 		rh->field_prefix = "";
 	}
 
-	if (!(rh->mem = pool_create(10 * 1024))) {
+	if (!(rh->mem = pool_create("report", 10 * 1024))) {
 		log_error("Allocation of memory pool for report failed");
 		return NULL;
 	}

lib/snapshot/snapshot.c

@@ -70,7 +70,8 @@ static int _text_import(struct lv_segment *seg, const struct config_node *sn,
 		return 0;
 	}
 
-	if (!vg_add_snapshot(org, cow, 1, &seg->lv->lvid.id[1], chunk_size)) {
+	if (!vg_add_snapshot(org, cow, 1, &seg->lv->lvid.id[1], seg->len,
+			     chunk_size)) {
 		stack;
 		return 0;
 	}

lib/uuid/uuid.c

@@ -29,9 +29,7 @@ static unsigned char _inverse_c[256];
 int lvid_create(union lvid *lvid, struct id *vgid)
 {
 	memcpy(lvid->id, vgid, sizeof(*lvid->id));
-	id_create(&lvid->id[1]);
-
-	return 1;
+	return id_create(&lvid->id[1]);
 }
 
 void uuid_from_num(char *uuid, uint32_t num)
@@ -83,16 +81,18 @@ int id_create(struct id *id)
 
 	memset(id->uuid, 0, len);
 	if ((randomfile = open("/dev/urandom", O_RDONLY)) < 0) {
-		log_sys_error("open", "id_create");
+		log_sys_error("open", "id_create: /dev/urandom");
 		return 0;
 	}
 
 	if (read(randomfile, id->uuid, len) != len) {
-		log_sys_error("read", "id_create");
-		close(randomfile);
+		log_sys_error("read", "id_create: /dev/urandom");
+		if (close(randomfile))
+			stack;
 		return 0;
 	}
-	close(randomfile);
+	if (close(randomfile))
+		stack;
 
         /*
          * Skip out the last 2 chars in randomized creation for LVM1

libdm/.exported_symbols

@@ -20,6 +20,7 @@ dm_task_set_minor
 dm_task_set_sector
 dm_task_set_message
 dm_task_add_target
+dm_task_no_open_count
 dm_get_next_target
 dm_task_run
 dm_set_dev_dir

libdm/Makefile.in

@@ -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 the device-mapper userspace tools.
 #
@@ -17,12 +17,20 @@ top_srcdir = @top_srcdir@
 VPATH = @srcdir@
 interface = @interface@
 
-SOURCES = libdm-common.c $(interface)/libdm-iface.c
+SOURCES = libdm-common.c libdm-file.c $(interface)/libdm-iface.c
 
 INCLUDES = -I$(interface)
 
 LIB_STATIC = $(interface)/libdevmapper.a
-LIB_SHARED = $(interface)/libdevmapper.so
+
+ifeq ("@LIB_SUFFIX@","dylib")
+  LIB_SHARED = $(interface)/libdevmapper.dylib
+else
+  LIB_SHARED = $(interface)/libdevmapper.so
+endif
+
+CFLAGS += -DDEVICE_UID=@DEVICE_UID@ -DDEVICE_GID=@DEVICE_GID@ \
+	  -DDEVICE_MODE=@DEVICE_MODE@
 
 include ../make.tmpl
 
@@ -38,7 +46,8 @@ endif
 install: $(INSTALL_TYPE)
 
 install_dynamic: install_@interface@
-	$(LN_S) -f libdevmapper.so.$(LIB_VERSION) $(libdir)/libdevmapper.so
+	$(LN_S) -f libdevmapper.$(LIB_SUFFIX).$(LIB_VERSION) \
+		$(libdir)/libdevmapper.$(LIB_SUFFIX)
 	$(INSTALL) -D $(OWNER) $(GROUP) -m 444 libdevmapper.h \
 		$(includedir)/libdevmapper.h
 
@@ -47,13 +56,13 @@ install_static: install_@interface@_static
 	$(INSTALL) -D $(OWNER) $(GROUP) -m 444 libdevmapper.h \
 		$(includedir)/libdevmapper.h
 
-install_fs: fs/libdevmapper.so
+install_fs: fs/libdevmapper.$(LIB_SUFFIX)
 	$(INSTALL) -D $(OWNER) $(GROUP) -m 555 $(STRIP) $< \
-		$(libdir)/libdevmapper.so.$(LIB_VERSION)
+		$(libdir)/libdevmapper.$(LIB_SUFFIX).$(LIB_VERSION)
 
-install_ioctl: ioctl/libdevmapper.so
+install_ioctl: ioctl/libdevmapper.$(LIB_SUFFIX)
 	$(INSTALL) -D $(OWNER) $(GROUP) -m 555 $(STRIP) $< \
-		$(libdir)/libdevmapper.so.$(LIB_VERSION)
+		$(libdir)/libdevmapper.$(LIB_SUFFIX).$(LIB_VERSION)
 
 install_ioctl_static: ioctl/libdevmapper.a
 	$(INSTALL) -D $(OWNER) $(GROUP) -m 555 $(STRIP) $< \

libdm/ioctl/libdm-iface.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 the device-mapper userspace tools.
  *
@@ -13,35 +13,31 @@
  * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
+#include "lib.h"
 #include "libdm-targets.h"
 #include "libdm-common.h"
-#include "log.h"
+#include "libdm-file.h"
 
 #ifdef DM_COMPAT
 #  include "libdm-compat.h"
 #endif
 
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-#include <unistd.h>
 #include <fcntl.h>
 #include <dirent.h>
-#include <errno.h>
-#include <sys/stat.h>
 #include <sys/ioctl.h>
 #include <limits.h>
 
 #ifdef linux
 #  include "kdev_t.h"
 #  include <linux/limits.h>
-#  include <linux/dm-ioctl.h>
 #else
 #  define MAJOR(x) major((x))
 #  define MINOR(x) minor((x))
 #  define MKDEV(x,y) makedev((x),(y))
 #endif
 
+#include <linux/dm-ioctl.h>
+
 /*
  * Ensure build compatibility.  
  * The hard-coded versions here are the highest present 
@@ -53,6 +49,13 @@
 #error The version of dm-ioctl.h included is incompatible.
 #endif
 
+/* FIXME This should be exported in device-mapper.h */
+#define DM_NAME "device-mapper"
+
+#define PROC_MISC "/proc/misc"
+#define PROC_DEVICES "/proc/devices"
+#define MISC_NAME "misc"
+
 /* dm major version no for running kernel */
 static int _dm_version = DM_VERSION_MAJOR;
 static int _log_suppress = 0;
@@ -114,22 +117,147 @@ static void *_align(void *ptr, unsigned int a)
 	return (void *) (((unsigned long) ptr + agn) & ~agn);
 }
 
+static int _get_proc_number(const char *file, const char *name,
+			    uint32_t *number)
+{
+	FILE *fl;
+	char nm[256];
+	int c;
+
+	if (!(fl = fopen(file, "r"))) {
+		log_error("%s: fopen failed: %s", file, strerror(errno));
+		return 0;
+	}
+
+	while (!feof(fl)) {
+		if (fscanf(fl, "%d %255s\n", number, &nm[0]) == 2) {
+			if (!strcmp(name, nm)) {
+				fclose(fl);
+				return 1;
+			}
+		} else do {
+			c = fgetc(fl);
+		} while (c != EOF && c != '\n');
+	}
+	fclose(fl);
+
+	log_error("%s: No entry for %s found", file, name);
+	return 0;
+}
+
+static int _control_device_number(uint32_t *major, uint32_t *minor)
+{
+	if (!_get_proc_number(PROC_DEVICES, MISC_NAME, major) ||
+	    !_get_proc_number(PROC_MISC, DM_NAME, minor)) {
+		*major = 0;
+		return 0;
+	}
+
+	return 1;
+}
+
+/*
+ * Returns 1 if exists; 0 if it doesn't; -1 if it's wrong
+ */
+static int _control_exists(const char *control, uint32_t major, uint32_t minor)
+{
+	struct stat buf;
+
+	if (stat(control, &buf) < 0) {
+		if (errno != ENOENT)
+			log_error("%s: stat failed: %s", control,
+				  strerror(errno));
+		return 0;
+	}
+
+	if (!S_ISCHR(buf.st_mode)) {
+		log_verbose("%s: Wrong inode type", control);
+		if (!unlink(control))
+			return 0;
+		log_error("%s: unlink failed: %s", control,
+			  strerror(errno));
+		return -1;
+	}
+
+	if (major && buf.st_rdev != MKDEV(major, minor)) {
+		log_verbose("%s: Wrong device number: (%u, %u) instead of "
+			    "(%u, %u)", control,
+			    MAJOR(buf.st_mode), MINOR(buf.st_mode),
+			    major, minor);
+		if (!unlink(control))
+			return 0;
+		log_error("%s: unlink failed: %s", control,
+			  strerror(errno));
+		return -1;
+	}
+
+	return 1;
+}
+
+static int _create_control(const char *control, uint32_t major, uint32_t minor)
+{
+	int ret;
+	mode_t old_umask;
+
+	if (!major)
+		return 0;
+
+	old_umask = umask(0022);
+	ret = create_dir(dm_dir());
+	umask(old_umask);
+
+	if (!ret)
+		return 0;
+
+	log_verbose("Creating device %s (%u, %u)", control, major, minor);
+
+	if (mknod(control, S_IFCHR | S_IRUSR | S_IWUSR,
+		  MKDEV(major, minor)) < 0)  {
+		log_error("%s: mknod failed: %s", control, strerror(errno));
+		return 0;
+	}
+
+#ifdef HAVE_SELINUX
+        if (!set_selinux_context(control)) {
+                stack;
+                return 0;
+        }
+#endif
+
+	return 1;
+}
+
 static int _open_control(void)
 {
+#ifdef DM_IOCTLS
 	char control[PATH_MAX];
+	uint32_t major = 0, minor;
 
 	if (_control_fd != -1)
 		return 1;
 
 	snprintf(control, sizeof(control), "%s/control", dm_dir());
 
+	if (!_control_device_number(&major, &minor))
+		log_error("Is device-mapper driver missing from kernel?");
+
+	if (!_control_exists(control, major, minor) &&
+	    !_create_control(control, major, minor))
+		goto error;
+
 	if ((_control_fd = open(control, O_RDWR)) < 0) {
 		log_error("%s: open failed: %s", control, strerror(errno));
-		log_error("Is device-mapper driver missing from kernel?");
-		return 0;
+		goto error;
 	}
 
 	return 1;
+
+error:
+	log_error("Failure to communicate with kernel device-mapper driver.");
+	return 0;
+#else
+	return 1;
+#endif
 }
 
 void dm_task_destroy(struct dm_task *dmt)
@@ -485,7 +613,9 @@ static int _dm_task_run_v1(struct dm_task *dmt)
 	if (dmt->type == DM_DEVICE_LIST) {
 		if (!_dm_names_v1(dmi))
 			goto bad;
-	} else if (ioctl(_control_fd, command, dmi) < 0) {
+	} 
+#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));
@@ -494,10 +624,17 @@ static int _dm_task_run_v1(struct dm_task *dmt)
 				  _IOC_NR(command), strerror(errno));
 		goto bad;
 	}
+#else /* Userspace alternative for testing */
+#endif
+
+	if (dmi->flags & DM_BUFFER_FULL_FLAG)
+		/* FIXME Increase buffer size and retry operation (if query) */
+		log_error("Warning: libdevmapper buffer too small for data");
 
 	switch (dmt->type) {
 	case DM_DEVICE_CREATE:
-		add_dev_node(dmt->dev_name, MAJOR(dmi->dev), MINOR(dmi->dev));
+		add_dev_node(dmt->dev_name, MAJOR(dmi->dev), MINOR(dmi->dev),
+			     dmt->uid, dmt->gid, dmt->mode);
 		break;
 
 	case DM_DEVICE_REMOVE:
@@ -511,7 +648,8 @@ static int _dm_task_run_v1(struct dm_task *dmt)
 	case DM_DEVICE_MKNODES:
 		if (dmi->flags & DM_EXISTS_FLAG)
 			add_dev_node(dmt->dev_name, MAJOR(dmi->dev),
-				     MINOR(dmi->dev));
+				     MINOR(dmi->dev),
+				     dmt->uid, dmt->gid, dmt->mode);
 		else
 			rm_dev_node(dmt->dev_name);
 		break;
@@ -800,6 +938,13 @@ int dm_task_set_sector(struct dm_task *dmt, uint64_t sector)
 	return 1;
 }
 
+int dm_task_no_open_count(struct dm_task *dmt)
+{
+	dmt->no_open_count = 1;
+
+	return 1;
+}
+
 int dm_task_set_event_nr(struct dm_task *dmt, uint32_t event_nr)
 {
 	dmt->event_nr = event_nr;
@@ -1152,7 +1297,8 @@ int dm_task_run(struct dm_task *dmt)
 	if (dmt->type == DM_DEVICE_CREATE && dmt->head)
 		return _create_and_load_v4(dmt);
 
-	if (dmt->type == DM_DEVICE_MKNODES && !dmt->dev_name)
+	if (dmt->type == DM_DEVICE_MKNODES && !dmt->dev_name &&
+	    !dmt->uuid && dmt->major <= 0)
 		return _mknodes_v4(dmt);
 
 	if (!_open_control())
@@ -1168,43 +1314,58 @@ int dm_task_run(struct dm_task *dmt)
 		dmi->flags |= DM_STATUS_TABLE_FLAG;
 
 	dmi->flags |= DM_EXISTS_FLAG;	/* FIXME */
-	log_debug("dm %s %s %s %s%.0llu %s", _cmd_data_v4[dmt->type].name,
+
+	if (dmt->no_open_count)
+		dmi->flags |= DM_SKIP_BDGET_FLAG;
+
+	log_debug("dm %s %s %s %s%c %.0llu %s", _cmd_data_v4[dmt->type].name,
 		  dmi->name, dmi->uuid, dmt->newname ? dmt->newname : "",
+		  dmt->no_open_count ? 'N' : 'O',
 		  dmt->sector, dmt->message ? dmt->message : "");
+#ifdef DM_IOCTLS
 	if (ioctl(_control_fd, command, dmi) < 0) {
 		if (errno == ENXIO && ((dmt->type == DM_DEVICE_INFO) ||
-				       (dmt->type == DM_DEVICE_MKNODES))) {
+				       (dmt->type == DM_DEVICE_MKNODES)))
 			dmi->flags &= ~DM_EXISTS_FLAG;	/* FIXME */
-			goto ignore_error;
+		else {
+			if (_log_suppress)
+				log_verbose("device-mapper ioctl "
+					    "cmd %d failed: %s",
+					    _IOC_NR(command), strerror(errno));
+			else
+				log_error("device-mapper ioctl "
+					  "cmd %d failed: %s",
+					  _IOC_NR(command), strerror(errno));
+			goto bad;
 		}
-		if (_log_suppress)
-			log_verbose("device-mapper ioctl cmd %d failed: %s",
-				    _IOC_NR(command), strerror(errno));
-		else
-			log_error("device-mapper ioctl cmd %d failed: %s",
-				  _IOC_NR(command), strerror(errno));
-		goto bad;
 	}
+#else /* Userspace alternative for testing */
+#endif
 
-      ignore_error:
 	switch (dmt->type) {
 	case DM_DEVICE_CREATE:
-		add_dev_node(dmt->dev_name, MAJOR(dmi->dev), MINOR(dmi->dev));
+		add_dev_node(dmt->dev_name, MAJOR(dmi->dev), MINOR(dmi->dev),
+			     dmt->uid, dmt->gid, dmt->mode);
 		break;
 
 	case DM_DEVICE_REMOVE:
-		rm_dev_node(dmt->dev_name);
+		/* FIXME Kernel needs to fill in dmi->name */
+		if (dmt->dev_name)
+			rm_dev_node(dmt->dev_name);
 		break;
 
 	case DM_DEVICE_RENAME:
-		rename_dev_node(dmt->dev_name, dmt->newname);
+		/* FIXME Kernel needs to fill in dmi->name */
+		if (dmt->dev_name)
+			rename_dev_node(dmt->dev_name, dmt->newname);
 		break;
 
 	case DM_DEVICE_MKNODES:
 		if (dmi->flags & DM_EXISTS_FLAG)
-			add_dev_node(dmt->dev_name, MAJOR(dmi->dev),
-				     MINOR(dmi->dev));
-		else
+			add_dev_node(dmi->name, MAJOR(dmi->dev),
+				     MINOR(dmi->dev),
+				     dmt->uid, dmt->gid, dmt->mode);
+		else if (dmt->dev_name)
 			rm_dev_node(dmt->dev_name);
 		break;
 

libdm/ioctl/libdm-targets.h

@@ -17,6 +17,7 @@
 #define LIB_DMTARGETS_H
 
 #include <inttypes.h>
+#include <sys/types.h>
 
 struct dm_ioctl;
 struct dm_ioctl_v1;
@@ -40,6 +41,9 @@ struct dm_task {
 	uint32_t event_nr;
 	int major;
 	int minor;
+	uid_t uid;
+	gid_t gid;
+	mode_t mode;
 	union {
 		struct dm_ioctl *v4;
 		struct dm_ioctl_v1 *v1;
@@ -47,6 +51,7 @@ struct dm_task {
 	char *newname;
 	char *message;
 	uint64_t sector;
+	int no_open_count;
 
 	char *uuid;
 };

libdm/libdevmapper.h

@@ -134,6 +134,7 @@ int dm_task_set_major(struct dm_task *dmt, int major);
 int dm_task_set_event_nr(struct dm_task *dmt, uint32_t event_nr);
 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);
 
 /*
  * Use these to prepare for a create or reload.

libdm/libdm-common.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 the device-mapper userspace tools.
  *
@@ -13,20 +13,15 @@
  * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
+#include "lib.h"
 #include "libdm-targets.h"
 #include "libdm-common.h"
 #include "list.h"
-#include "log.h"
 #include "kdev_t.h"
 
-#include <stdio.h>
-#include <stdlib.h>
 #include <stdarg.h>
-#include <string.h>
 #include <sys/param.h>
-#include <sys/stat.h>
-#include <unistd.h>
-#include <errno.h>
+
 #include <linux/dm-ioctl.h>
 
 #ifdef HAVE_SELINUX
@@ -63,7 +58,7 @@ static void _default_log(int level, const char *file, int line,
 	if (level < _LOG_WARN)
 		fprintf(stderr, "\n");
 	else
-		printf("\n");
+		fprintf(stdout, "\n");
 }
 
 dm_log_fn _log = _default_log;
@@ -100,19 +95,22 @@ struct dm_task *dm_task_create(int type)
 {
 	struct dm_task *dmt = malloc(sizeof(*dmt));
 
-	if (!dm_check_version())
-		return NULL;
-
 	if (!dmt) {
 		log_error("dm_task_create: malloc(%d) failed", sizeof(*dmt));
 		return NULL;
 	}
 
+	if (!dm_check_version())
+		return NULL;
+
 	memset(dmt, 0, sizeof(*dmt));
 
 	dmt->type = type;
 	dmt->minor = -1;
 	dmt->major = -1;
+	dmt->uid = DEVICE_UID;
+	dmt->gid = DEVICE_GID;
+	dmt->mode = DEVICE_MODE;
 
 	return dmt;
 }
@@ -202,7 +200,7 @@ int dm_task_add_target(struct dm_task *dmt, uint64_t start, uint64_t size,
 }
 
 #ifdef HAVE_SELINUX
-static int _set_selinux_context(const char *path)
+int set_selinux_context(const char *path)
 {
 	security_context_t scontext;
 
@@ -226,11 +224,13 @@ static int _set_selinux_context(const char *path)
 }
 #endif
 
-static int _add_dev_node(const char *dev_name, uint32_t major, uint32_t minor)
+static int _add_dev_node(const char *dev_name, uint32_t major, uint32_t minor,
+			 uid_t uid, gid_t gid, mode_t mode)
 {
 	char path[PATH_MAX];
 	struct stat info;
 	dev_t dev = MKDEV(major, minor);
+	mode_t old_mask;
 
 	_build_dev_path(path, sizeof(path), dev_name);
 
@@ -241,6 +241,7 @@ static int _add_dev_node(const char *dev_name, uint32_t major, uint32_t minor)
 			return 0;
 		}
 
+		/* If right inode already exists we don't touch uid etc. */
 		if (info.st_rdev == dev)
 			return 1;
 
@@ -251,12 +252,20 @@ static int _add_dev_node(const char *dev_name, uint32_t major, uint32_t minor)
 		}
 	}
 
-	if (mknod(path, S_IFBLK | S_IRUSR | S_IWUSR | S_IRGRP, dev) < 0) {
+	old_mask = umask(0);
+	if (mknod(path, S_IFBLK | mode, dev) < 0) {
 		log_error("Unable to make device node for '%s'", dev_name);
 		return 0;
 	}
+	umask(old_mask);
+
+	if (chown(path, uid, gid) < 0) {
+		log_error("%s: chown failed: %s", path, strerror(errno));
+		return 0;
+	}
+
 #ifdef HAVE_SELINUX
-	if (!_set_selinux_context(path))
+	if (!set_selinux_context(path))
 		return 0;
 #endif
 
@@ -324,11 +333,12 @@ typedef enum {
 } node_op_t;
 
 static int _do_node_op(node_op_t type, const char *dev_name, uint32_t major,
-		       uint32_t minor, const char *old_name)
+		       uint32_t minor, uid_t uid, gid_t gid, mode_t mode,
+		       const char *old_name)
 {
 	switch (type) {
 	case NODE_ADD:
-		return _add_dev_node(dev_name, major, minor);
+		return _add_dev_node(dev_name, major, minor, uid, gid, mode);
 	case NODE_DEL:
 		return _rm_dev_node(dev_name);
 	case NODE_RENAME:
@@ -346,6 +356,9 @@ struct node_op_parms {
 	char *dev_name;
 	uint32_t major;
 	uint32_t minor;
+	uid_t uid;
+	gid_t gid;
+	mode_t mode;
 	char *old_name;
 	char names[0];
 };
@@ -358,7 +371,8 @@ static void _store_str(char **pos, char **ptr, const char *str)
 }
 
 static int _stack_node_op(node_op_t type, const char *dev_name, uint32_t major,
-			  uint32_t minor, const char *old_name)
+			  uint32_t minor, uid_t uid, gid_t gid, mode_t mode,
+			  const char *old_name)
 {
 	struct node_op_parms *nop;
 	size_t len = strlen(dev_name) + strlen(old_name) + 2;
@@ -373,6 +387,9 @@ static int _stack_node_op(node_op_t type, const char *dev_name, uint32_t major,
 	nop->type = type;
 	nop->major = major;
 	nop->minor = minor;
+	nop->uid = uid;
+	nop->gid = gid;
+	nop->mode = mode;
 
 	_store_str(&pos, &nop->dev_name, dev_name);
 	_store_str(&pos, &nop->old_name, old_name);
@@ -390,25 +407,27 @@ static void _pop_node_ops(void)
 	list_iterate_safe(noph, nopht, &_node_ops) {
 		nop = list_item(noph, struct node_op_parms);
 		_do_node_op(nop->type, nop->dev_name, nop->major, nop->minor,
-			    nop->old_name);
+			    nop->uid, nop->gid, nop->mode, nop->old_name);
 		list_del(&nop->list);
 		free(nop);
 	}
 }
 
-int add_dev_node(const char *dev_name, uint32_t major, uint32_t minor)
+int add_dev_node(const char *dev_name, uint32_t major, uint32_t minor,
+		 uid_t uid, gid_t gid, mode_t mode)
 {
-	return _stack_node_op(NODE_ADD, dev_name, major, minor, "");
+	return _stack_node_op(NODE_ADD, dev_name, major, minor, uid, gid, mode,
+			      "");
 }
 
 int rename_dev_node(const char *old_name, const char *new_name)
 {
-	return _stack_node_op(NODE_RENAME, new_name, 0, 0, old_name);
+	return _stack_node_op(NODE_RENAME, new_name, 0, 0, 0, 0, 0, old_name);
 }
 
 int rm_dev_node(const char *dev_name)
 {
-	return _stack_node_op(NODE_DEL, dev_name, 0, 0, "");
+	return _stack_node_op(NODE_DEL, dev_name, 0, 0, 0, 0, 0, "");
 }
 
 void update_devs(void)

libdm/libdm-common.h.in

@@ -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 the device-mapper userspace tools.
  *
@@ -22,11 +22,14 @@ struct target *create_target(uint64_t start,
 			     uint64_t len,
 			     const char *type, const char *params);
 
-int add_dev_node(const char *dev_name, uint32_t minor, uint32_t major);
+int add_dev_node(const char *dev_name, uint32_t minor, uint32_t major,
+		 uid_t uid, gid_t gid, mode_t mode);
 int rm_dev_node(const char *dev_name);
 int rename_dev_node(const char *old_name, const char *new_name);
 void update_devs(void);
 
+int set_selinux_context(const char *path);
+
 #define DM_LIB_VERSION @DM_LIB_VERSION@
 
 #endif

libdm/libdm-file.c

@@ -0,0 +1,76 @@
+/*
+ * 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 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.
+ *
+ * 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
+ */
+
+#include "lib.h"
+#include "libdm-file.h"
+
+#include <sys/file.h>
+#include <fcntl.h>
+#include <dirent.h>
+
+#ifdef linux
+#  include <malloc.h>
+#endif
+
+static int _create_dir_recursive(const char *dir)
+{
+	char *orig, *s;
+	int rc;
+
+	log_verbose("Creating directory \"%s\"", dir);
+	/* Create parent directories */
+	orig = s = strdup(dir);
+	while ((s = strchr(s, '/')) != NULL) {
+		*s = '\0';
+		if (*orig) {
+			rc = mkdir(orig, 0777);
+			if (rc < 0 && errno != EEXIST) {
+				log_error("%s: mkdir failed: %s", orig,
+					  strerror(errno));
+				free(orig);
+				return 0;
+			}
+		}
+		*s++ = '/';
+	}
+	free(orig);
+
+	/* Create final directory */
+	rc = mkdir(dir, 0777);
+	if (rc < 0 && errno != EEXIST) {
+		log_error("%s: mkdir failed: %s", orig,
+			  strerror(errno));
+		return 0;
+	}
+	return 1;
+}
+
+int create_dir(const char *dir)
+{
+	struct stat info;
+
+	if (!*dir)
+		return 1;
+
+	if (stat(dir, &info) < 0)
+		return _create_dir_recursive(dir);
+
+	if (S_ISDIR(info.st_mode))
+		return 1;
+
+	log_error("Directory \"%s\" not found", dir);
+	return 0;
+}
+

libdm/libdm-file.h

@@ -0,0 +1,25 @@
+/*
+ * 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 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.
+ *
+ * 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
+ */
+
+#ifndef LIB_DMFILE_H
+#define LIB_DMFILE_H
+
+/*
+ * Create directory (recursively) if necessary.  Return 1
+ * if directory was successfully created (or already exists), else 0.
+ */
+int create_dir(const char *dir);
+
+#endif

man/Makefile.in

@@ -30,7 +30,7 @@ MAN8DIR=${mandir}/man8
 
 include $(top_srcdir)/make.tmpl
 
-ifeq ("@CLVMD@", "yes")
+ifneq ("@CLVMD@", "none")
   install: install_cluster
 endif
 

man/lvchange.8

@@ -4,26 +4,34 @@ lvchange \- change attributes of a logical volume
 .SH SYNOPSIS
 .B lvchange
 [\-\-addtag Tag]
-[\-A/\-\-autobackup y/n] [\-a/\-\-available y/n]
+[\-A/\-\-autobackup y/n] [\-a/\-\-available y/n/ey/en/ly/ln]
+[\-\-alloc AllocationPolicy]
 [\-C/\-\-contiguous y/n] [\-d/\-\-debug] [\-\-deltag Tag]
 [\-h/\-?/\-\-help]
 [\-\-ignorelockingfailure]
 [\-M/\-\-persistent y/n] [\-\-minor minor]
 [\-P/\-\-partial y/n]
 [\-p/\-\-permission r/w] [\-r/\-\-readahead ReadAheadSectors]
+[\-\-refresh]
 [\-t/\-\-test]
 [\-v/\-\-verbose] LogicalVolumePath [LogicalVolumePath...]
 .SH DESCRIPTION
-lvchange allows you to change the attributes of a logical volume.
+lvchange allows you to change the attributes of a logical volume
+including making them known to the kernel ready for use.
 .SH OPTIONS
 See \fBlvm\fP for common options.
 .TP
-.I \-a, \-\-available y/n
+.I \-a, \-\-available y/n/ey/en/ly/ln
 Controls the availability of the logical volumes for use.
-This is (among others) useful for changing the logical volume's name
-( see
-.B lvrename(8)
-) safely.
+Communicates with the kernel device-mapper driver via
+libdevmapper to activate (-ay) or deactivate (-an) the 
+logical volumes.
+.IP
+If clustered locking is enabled, -ae will activate exclusively
+on one node and -aly will activate only on the local node.
+To deactivate only on the local node use -aln.
+Logical volumes with single-host snapshots are always activated 
+exclusively because they can only be used on one node at once.
 .TP
 .I \-C, \-\-contiguous y/n
 Tries to set or reset the contiguous allocation policy for
@@ -42,11 +50,18 @@ Change access permission to read-only or read/write.
 .TP
 .I \-r, \-\-readahead ReadAheadSectors
 Change read ahead sector count per logical between 2 and 120.
-Not used by device-mapper.
+For compatability with LVM1 only. Ignored by LVM2.
+.TP
+.I \-\-refresh
+If the logical volume is active, reload its metadata.
+This is not necessary in normal operation, but may be useful
+if something has gone wrong or if you're doing clustering 
+manually without a clustered lock manager.
 .SH Examples
-"lvchange -p r vg00/lvol1" changes the permission on 
+"lvchange -pr vg00/lvol1" changes the permission on 
 volume lvol1 in volume group vg00 to be read-only.
 
 .SH SEE ALSO
 .BR lvm (8), 
-.BR lvcreate (8)
+.BR lvcreate (8),
+.BR vgchange (8)

man/lvcreate.8

@@ -4,6 +4,7 @@ lvcreate \- create a logical volume in an existing volume group
 .SH SYNOPSIS
 .B lvcreate
 [\-\-addtag Tag]
+[\-\-alloc AllocationPolicy]
 [\-A/\-\-autobackup y/n] [\-C/\-\-contiguous y/n] [\-d/\-\-debug]
 [\-h/\-?/\-\-help]
 [\-i/\-\-stripes Stripes [\-I/\-\-stripesize StripeSize]]