test: api subdirectory is needed in PATH by pytest

Also, besides making "-o" groupable, use string lists to store
lists of options temporarily while processing all instances of
the "-o" group.

.gitignore

@@ -1,13 +1,16 @@
 *.5
+*.7
 *.8
 *.a
 *.d
 *.o
+*.orig
 *.pc
 *.pot
+*.rej
 *.so
 *.so.*
-*.swp
+*.sw*
 *~
 
 .export.sym
@@ -17,11 +20,11 @@
 Makefile
 make.tmpl
 
-configure.h
-version.h
-
 /autom4te.cache/
+/autoscan.log
 /config.log
 /config.status
+/configure.scan
 /cscope.out
+/tags
 /tmp/

Makefile.in

@@ -95,7 +95,7 @@ endif
 DISTCLEAN_TARGETS += cscope.out
 CLEAN_DIRS += autom4te.cache
 
-check check_system check_cluster check_local check_lvmetad unit: all
+check check_system check_cluster check_local check_lvmetad check_lvmpolld unit: all
 	$(MAKE) -C test $(@)
 
 conf.generate: tools
@@ -131,6 +131,9 @@ rpm: dist
 generate: conf.generate
 	$(MAKE) -C conf generate
 
+all_man:
+	$(MAKE) -C man all_man
+
 install_system_dirs:
 	$(INSTALL_DIR) $(DESTDIR)$(DEFAULT_SYS_DIR)
 	$(INSTALL_ROOT_DIR) $(DESTDIR)$(DEFAULT_ARCHIVE_DIR)
@@ -150,6 +153,9 @@ install_systemd_generators:
 install_systemd_units:
 	$(MAKE) -C scripts install_systemd_units
 
+install_all_man:
+	$(MAKE) -C man install_all_man
+
 ifeq ("@PYTHON_BINDINGS@", "yes")
 install_python_bindings:
 	$(MAKE) -C liblvm/python install_python_bindings
@@ -223,3 +229,12 @@ memcheck: test-programs
 ruby-test:
 	$(RUBY) report-generators/test/ts.rb
 endif
+
+ifneq ($(shell which ctags),)
+.PHONY: tags
+tags:
+	test -z "$(shell find $(top_srcdir) -type f -name '*.[ch]' -newer tags 2>/dev/null | head -1)" || $(RM) tags
+	test -f tags || find $(top_srcdir) -maxdepth 5 -type f -name '*.[ch]' -exec ctags -a '{}' +
+
+CLEAN_TARGETS += tags
+endif

VERSION

@@ -1 +1 @@
-2.02.124(2)-git (2015-07-03)
+2.02.134(2)-git (2015-10-30)

VERSION_DM

@@ -1 +1 @@
-1.02.101-git (2015-07-03)
+1.02.111-git (2015-10-30)

WHATS_NEW

@@ -1,3 +1,128 @@
+Version 2.02.134 - 
+====================================
+
+Version 2.02.133 - 30th October 2015
+====================================
+  Support repeated -o|--options for reporting commands.
+  Support -o- and -o# for reporting commands to remove and compact fields.
+  Fix missing PVs from pvs output if vgremove is run concurrently.
+  Remove unwanted error message when running pvs/vgs/lvs and vgremove at once.
+  Check newly created VG's metadata do not overlap in metadata ring buffer.
+  Check metadata area size is at least the minimum size defined for the format.
+  Thin pool targets uses low_water_mark from profile.
+  Dropping 'yet' from error of unsupported thick snapshot of snapshots.
+  Do not support unpartitioned DASD devices with CDL formatted with pvcreate.
+  For thins use flush for suspend only when volume size is reduced.
+  Enable code which detects the need of flush during suspend.
+  Ensure --use-policy will resize volume to fit below threshold.
+  Correct percentage evaluation when checking thin-pool over threshold.
+  Fix lvmcache to move PV from VG to orphans if VG is removed and lvmetad used.
+  Fix lvmcache to not cache even invalid info about PV which got removed.
+  Support checking of memlock daemon counter.
+  Allow all log levels to be used with the lvmetad -l option.
+  Add optional shutdown when idle support for lvmetad.
+  Fix missing in-sync progress info while lvconvert used with lvmpolld.
+  Add report/compact_output_cols to lvm.conf to define report cols to compact.
+  Do not change logging in lvm2 library when it's already set.
+  Check for enough space in thin-pool in command before creating new thin.
+  Make libblkid detect all copies of the same signature if use_blkid_wiping=1.
+  Fix vgimportclone with -n to not add number unnecessarily to base VG name.
+  Cleanup vgimportclone script and remove dependency on awk, grep, cut and tr.
+  Add vg_missing_pv_count report field to report number of missing PVs in a VG.
+  Properly identify internal LV holding sanlock locks within lv_role field.
+  Add metadata_devices and seg_metadata_le_ranges report fields for raid vols.
+  Fix lvm2-{activation,clvmd,cmirrord,monitor} service to exec before mounting.
+
+Version 2.02.132 - 22nd September 2015
+======================================
+  Fix lvmconf to set locking_type=2 if external locking library is requested.
+  Remove verbose message when rescanning an unchanged device. (2.02.119)
+  Add origin_uuid, mirror_log_uuid, move_pv_uuid, convert_lv_uuid report fields.
+  Add pool_lv_uuid, metadata_lv_uuid, data_lv_uuid reporting fields.
+  Fix PV label processing failure after pvcreate in lvm shell with lvmetad.
+
+Version 2.02.131 - 15th September 2015
+======================================
+  Rename 'make install_full_man' to install_all_man and add all_man target.
+  Fix vgimportclone cache_dir path name (2.02.115).
+  Swapping of LV identifiers handles more complex LVs.
+  Use passed list of PVS when allocating space in lvconvert --thinpool.
+  Disallow usage of --stripe and --stripesize when creating cache pool.
+  Warn user when caching raid or thin pool data LV.
+  When layering LV, move LV flags with segments.
+  Ignore persistent cache if configuration changed. (2.02.127)
+  Fix devices/filter to be applied before disk-accessing filters. (2.02.112)
+  Make tags only when requested via 'make tags'.
+  Configure supports --disable-dependency-tracking for one-time builds.
+  Fix usage of configure.h when building in srcdir != builddir.
+
+Version 2.02.130 - 5th September 2015
+=====================================
+  Fix use of uninitialized device status if reading outdated .cache record.
+  Restore support for --monitor option in lvcreate (2.02.112).
+  Read thin-pool data and metadata percent without flush.
+  Detect blocked thin-pool and avoid scanning their thin volumes.
+  Check if dm device is usable before checking its size (2.02.116).
+  Extend parsing of cache_check version in configure.
+  Make lvpoll error messages visible in lvmpolld's stderr and in syslog.
+  Add 'make install_full_man' to install all man pages regardless of config.
+
+Version 2.02.129 - 26th August 2015
+===================================
+  Drop error message when vgdisplay encounters an exported VG. (2.02.27)
+  Fix shared library generation to stop exporting internal functions.(2.02.120)
+  Accept --cachemode with lvconvert.
+  Fix and improve reporting properties of cache-pool.
+  Enable usage of --cachepolicy and --cachesetting with lvconvert.
+  Don't allow to reduce size of thin-pool metadata.
+  Fix debug buffer overflows in cmirrord logging.
+  Add --foreground and --help to cmirrord.
+
+Version 2.02.128 - 17th August 2015
+===================================
+  Allocation setting cache_pool_cachemode is replaced by cache_mode.
+  Don't attempt to close config file that couldn't be opened.
+  Check for valid cache mode in validation of cache segment.
+  Change internal interface handling cache mode and policy.
+  When no cache policy specified, prefer smq (if available) over mq.
+  Add demo cache-mq and cache-smq profiles.
+  Add cmd profilable allocation/cache_policy,cache_settings,cache_mode.
+  Require cache_check 0.5.4 for use of --clear-needs-check-flag.
+  Fix lvmetad udev rules to not override SYSTEMD_WANTS, add the service instead.
+
+Version 2.02.127 - 10th August 2015
+===================================
+  Do not init filters, locking, lvmetad, lvmpolld if command doesn't use it.
+  Order fields in struct cmd_context more logically.
+  Add lock_type to lvmcache VG summary and info structs.
+  Fix regression in cache causing some PVs to bypass filters (2.02.105).
+  Make configure --enable-realtime the default now.
+  Update .gitignore and configure.in files to reflect usage of current tree.
+
+Version 2.02.126 - 24th July 2015
+=================================
+  Fix long option hyphen removal. (2.02.122)
+  Fix clvmd freeze if client disappears without first releasing its locks.
+  Fix lvconvert segfaults while performing snapshots merge.
+  Ignore errors during detection if use_blkid_wiping=1 and --force is used.
+  Recognise DM_ABORT_ON_INTERNAL_ERRORS env var override in lvm logging fn.
+  Fix alloc segfault when extending LV with fewer stripes than in first seg.
+  Fix handling of cache policy name.
+  Set cache policy before with the first lvm2 cache pool metadata commit.
+  Fix detection of thin-pool overprovisioning (2.02.124).
+  Fix lvmpolld segfaults on 32 bit architectures.
+  Add lvmlockd lock_args validation to vg_validate.
+  Fix ignored --startstopservices option if running lvmconf with systemd.
+  Hide sanlock LVs when processing LVs in VG unless named or --all used.
+
+Version 2.02.125 - 7th July 2015
+================================
+  Fix getline memory usage in lvmpolld.
+  Add support --clear-needs-check-flag for cache_check of cache pool metadata.
+  Add lvmetactl for developer use only.
+  Rename global/lock_retries to lvmlockd_retries.
+  Replace --enable-lvmlockd by --enable-lockd-sanlock and --enable-lockd-dlm.
+
 Version 2.02.124 - 3rd July 2015
 ================================
   Move sending thin pool messages from resume to suspend phase.

WHATS_NEW_DM

@@ -1,3 +1,119 @@
+Version 1.02.111 - 
+====================================
+
+Version 1.02.110 - 30th October 2015
+====================================
+  Disable thin monitoring plugin when it fails too often (>10 times).
+  Fix/restore parsing of empty field '-' when processing dmeventd event.
+  Enhance dm_tree_node_size_changed() to recognize size reduction.
+  Support exit on idle for dmenventd (1 hour).
+  Add support to allow unmonitor device from plugin itself.
+  New design for thread co-operation in dmeventd.
+  Dmeventd read device status with 'noflush'.
+  Dmeventd closes control device when no device is monitored.
+  Thin plugin for dmeventd improved percentage usage.
+  Snapshot plugin for dmeventd improved percentage usage.
+  Add dm_hold_control_dev to allow holding of control device open.
+  Add dm_report_compact_given_fields to remove given empty fields from report.
+  Use libdm status parsing and local mem raid dmeventd plugin.
+  Use local mem pool and lock only lvm2 execution for mirror dmeventd plugin.
+  Lock protect only lvm2 execution for snapshot and thin dmeventd plugin.
+  Use local mempool for raid and mirror plugins.
+  Reworked thread initialization for dmeventd plugins.
+  Dmeventd handles snapshot overflow for now equally as invalid.
+  Convert dmeventd to use common logging macro system from libdm.
+  Return -ENOMEM when device registration fails instead of 0 (=success).
+  Enforce writethrough mode for cleaner policy.
+  Add support for recognition and deactivation of MD devices to blkdeactivate.
+  Move target status functions out of libdm-deptree.
+  Correct use of max_write_behind parameter when generating raid target line.
+  Fix dm-event systemd service to make sure it is executed before mounting.
+
+Version 1.02.109 - 22nd September 2016
+======================================
+  Update man pages for dmsetup and dmstats.
+  Improve help text for dmsetup.
+  Use --noflush and --nolockfs when removing device with --force.
+  Parse new Overflow status string for snapshot target.
+  Check dir path components are valid if using dm_create_dir, error out if not.
+  Fix /dev/mapper handling to remove dangling entries if symlinks are found.
+  Make it possible to use blank value as selection for string list report field.
+
+Version 1.02.108 - 15th September 2015
+======================================
+  Do not check for full thin pool when activating without messages (1.02.107).
+
+Version 1.02.107 - 5th September 2015
+=====================================
+  Parse thin-pool status with one single routine internally.
+  Add --histogram to select default histogram fields for list and report.
+  Add report fields for displaying latency histogram configuration and data.
+  Add dmstats --bounds to specify histogram boundaries for a new region.
+  Add dm_histogram_to_string() to format histogram data in string form.
+  Add public methods to libdm to access numerical histogram config and data.
+  Parse and store histogram data in dm_stats_list() and dm_stats_populate().
+  Add an argument to specify histogram bounds to dm_stats_create_region().
+  Add dm_histogram_bounds_from_{string,uint64_t}() to parse histogram bounds.
+  Add dm_histogram handle type to represent a latency histogram and its bounds.
+  Fix devmapper.pc pkgconfig file to not reference non-existent rt.pc file.
+  Reinstate dm_task_get_info@Base to libdevmapper exports. (1.02.106)
+
+Version 1.02.106 - 26th August 2015
+===================================
+  Add 'precise' column to statistics reports.
+  Add --precise switch to 'dmstats create' to request nanosecond counters.
+  Add precise argument to dm_stats_create_region().
+  Add support to libdm-stats for precise_timestamps
+
+Version 1.02.105 - 17th August 2015
+===================================
+  Fix 'dmstats list -o all' segfault.
+  Separate dmstats statistics fields from region information fields.
+  Add interval and interval_ns fields to dmstats reports.
+  Do not include internal glibc headers in libdm-timestamp.c (1.02.104)
+  Exit immediately if no device is supplied to dmsetup wipe_table.
+  Suppress dmsetup report headings when no data is output. (1.02.104)
+  Adjust dmsetup usage/help output selection to match command invoked.
+  Fix dmsetup -o all to select correct fields in splitname report.
+  Restructure internal dmsetup argument handling across all commands.
+  Add dm_report_is_empty() to indicate there is no data awaiting output.
+  Add more arg validation for dm_tree_node_add_cache_target().
+  Add --alldevices switch to replace use of --force for stats create / delete.
+
+Version 1.02.104 - 10th August 2015
+===================================
+  Add dmstats.8 man page
+  Add dmstats --segments switch to create one region per device segment.
+  Add dmstats --regionid, --allregions to specify a single / all stats regions.
+  Add dmstats --allprograms for stats commands that filter by program ID.
+  Add dmstats --auxdata and --programid args to specify aux data and program ID.
+  Add report stats sub-command to provide repeating stats reports.
+  Add clear, delete, list, and print stats sub-commands.
+  Add create stats sub-command and --start, --length, --areas and --areasize.
+  Recognize 'dmstats' as an alias for 'dmsetup stats' when run with this name.
+  Add a 'stats' command to dmsetup to configure, manage and report stats data.
+  Add statistics fields to dmsetup -o.
+  Add libdm-stats library to allow management of device-mapper statistics.
+  Add --nosuffix to suppress dmsetup unit suffixes in report output.
+  Add --units to control dmsetup report field output units.
+  Add support to redisplay column headings for repeating column reports.
+  Fix report header and row resource leaks.
+  Report timestamps of ioctls with dmsetup -vvv.
+  Recognize report field name variants without any underscores too.
+  Add dmsetup --interval and --count to repeat reports at specified intervals.
+  Add dm_timestamp functions to libdevmapper.
+  Recognise vg/lv name format in dmsetup.
+  Move size display code to libdevmapper as dm_size_to_string.
+
+Version 1.02.103 - 24th July 2015
+=================================
+  Introduce libdevmapper wrappers for all malloc-related functions.
+
+Version 1.02.102 - 7th July 2015
+================================
+  Include tool.h for default non-library use.
+  Introduce format macros with embedded % such as FMTu64.
+
 Version 1.02.101 - 3rd July 2015
 ================================
   Add experimental support to passing messages in suspend tree.

aclocal.m4

@@ -15,7 +15,7 @@ m4_ifndef([AC_CONFIG_MACRO_DIRS], [m4_defun([_AM_CONFIG_MACRO_DIRS], [])m4_defun
 # pkg.m4 - Macros to locate and utilise pkg-config.            -*- Autoconf -*-
 # serial 1 (pkg-config-0.24)
 # 
-# Copyright © 2004 Scott James Remnant <scott@netsplit.com>.
+# Copyright (c) 2004 Scott James Remnant <scott@netsplit.com>.
 #
 # This program is free software; you can redistribute it and/or modify
 # it under the terms of the GNU General Public License as published by

conf/.gitignore

@@ -0,0 +1,6 @@
+command_profile_template.profile
+example.conf
+lvmlocal.conf
+metadata_profile_template.profile
+configure.h
+lvm-version.h

conf/Makefile.in

@@ -1,5 +1,5 @@
 #
-# Copyright (C) 2004-2010 Red Hat, Inc. All rights reserved.
+# Copyright (C) 2004-2015 Red Hat, Inc. All rights reserved.
 #
 # This file is part of LVM2.
 #
@@ -20,7 +20,11 @@ CONFDEST=lvm.conf
 CONFLOCAL=lvmlocal.conf
 
 PROFILE_TEMPLATES=command_profile_template.profile metadata_profile_template.profile
-PROFILES=$(PROFILE_TEMPLATES) $(srcdir)/thin-generic.profile $(srcdir)/thin-performance.profile
+PROFILES=$(PROFILE_TEMPLATES) \
+	$(srcdir)/cache-mq.profile \
+	$(srcdir)/cache-smq.profile \
+	$(srcdir)/thin-generic.profile \
+	$(srcdir)/thin-performance.profile
 
 include $(top_builddir)/make.tmpl
 

conf/cache-mq.profile

@@ -0,0 +1,20 @@
+# Demo configuration 'mq' cache policy
+#
+# Note: This policy has been deprecated in favor of the smq policy
+# keyword "default" means, setting is left with kernel defaults.
+#
+
+allocation {
+	cache_pool_chunk_size = 64
+	cache_mode = "writethrough"
+	cache_policy = "mq"
+	cache_settings {
+		mq {
+			sequential_threshold = "default"	#  #nr_sequential_ios
+			random_threshold = "default"		#  #nr_random_ios
+			read_promote_adjustment = "default"
+			write_promote_adjustment = "default"
+			discard_promote_adjustment = "default"
+		}
+	}
+}

conf/cache-smq.profile

@@ -0,0 +1,14 @@
+# Demo configuration 'smq' cache policy
+#
+# The stochastic multi-queue (smq) policy addresses some of the problems
+# with the multiqueue (mq) policy and uses less memory.
+#
+
+allocation {
+	cache_pool_chunk_size = 64
+	cache_mode = "writethrough"
+	cache_policy = "smq"
+	cache_settings {
+	        # currently no settins for "smq" policy
+	}
+}

conf/lvmlocal.conf.in

@@ -24,28 +24,33 @@ local {
 
 	# Configuration option local/system_id.
 	# Defines the local system ID for lvmlocal mode.
-	# This is used when global/system_id_source is set
-	# to 'lvmlocal' in the main configuration file,
-	# e.g. lvm.conf.
-	# When used, it must be set to a unique value
-	# among all hosts sharing access to the storage,
+	# This is used when global/system_id_source is set to 'lvmlocal' in the
+	# main configuration file, e.g. lvm.conf. When used, it must be set to
+	# a unique value among all hosts sharing access to the storage,
 	# e.g. a host name.
-	# Example:
-	# Set no system ID.
+	# 
+	# Example
+	# Set no system ID:
 	# system_id = ""
-	# Example:
-	# Set the system_id to the string 'host1'.
+	# Set the system_id to a specific name:
 	# system_id = "host1"
+	# 
+	# This configuration option has an automatic default value.
 	# system_id = ""
 
 	# Configuration option local/extra_system_ids.
 	# A list of extra VG system IDs the local host can access.
-	# VGs with the system IDs listed here (in addition
-	# to the host's own system ID) can be fully accessed
-	# by the local host.  (These are system IDs that the
-	# host sees in VGs, not system IDs that identify the
-	# local host, which is determined by system_id_source.)
-	# Use this only after consulting 'man lvmsystemid'
-	# to be certain of correct usage and possible dangers.
+	# VGs with the system IDs listed here (in addition to the host's own
+	# system ID) can be fully accessed by the local host. (These are
+	# system IDs that the host sees in VGs, not system IDs that identify
+	# the local host, which is determined by system_id_source.)
+	# Use this only after consulting 'man lvmsystemid' to be certain of
+	# correct usage and possible dangers.
 	# This configuration option does not have a default value defined.
+
+	# Configuration option local/host_id.
+	# The lvmlockd sanlock host_id.
+	# This must be unique among all hosts, and must be between 1 and 2000.
+	# This configuration option has an automatic default value.
+	# host_id = 0
 }

configure.in

@@ -16,7 +16,7 @@ AC_PREREQ(2.61)
 dnl -- Process this file with autoconf to produce a configure script.
 AC_INIT
 AC_CONFIG_SRCDIR([lib/device/dev-cache.h])
-AC_CONFIG_HEADERS([lib/misc/configure.h])
+AC_CONFIG_HEADERS([include/configure.h])
 
 ################################################################################
 dnl -- Setup the directory where autoconf has auxilary files
@@ -40,6 +40,8 @@ case "$host_os" in
 		LVMETAD=no
 		LVMPOLLD=no
 		LVMLOCKD=no
+		LOCKDSANLOCK=no
+		LOCKDDLM=no
 		ODIRECT=yes
 		DM_IOCTLS=yes
 		SELINUX=yes
@@ -89,14 +91,19 @@ AC_PATH_TOOL(CSCOPE_CMD, cscope)
 dnl -- Check for header files.
 AC_HEADER_DIRENT
 AC_HEADER_MAJOR
+AC_HEADER_STDBOOL
 AC_HEADER_STDC
 AC_HEADER_SYS_WAIT
 AC_HEADER_TIME
 
-AC_CHECK_HEADERS([locale.h stddef.h syslog.h sys/file.h sys/time.h assert.h \
-  langinfo.h libgen.h signal.h sys/mman.h sys/resource.h sys/utsname.h \
-  sys/wait.h time.h], ,
-  [AC_MSG_ERROR(bailing out)])
+AC_CHECK_HEADERS([assert.h ctype.h dirent.h errno.h fcntl.h float.h \
+  getopt.h inttypes.h langinfo.h libgen.h limits.h locale.h paths.h \
+  signal.h stdarg.h stddef.h stdio.h stdlib.h string.h sys/file.h \
+  sys/ioctl.h syslog.h sys/mman.h sys/param.h sys/resource.h sys/stat.h \
+  sys/time.h sys/types.h sys/utsname.h sys/wait.h time.h \
+  unistd.h], , [AC_MSG_ERROR(bailing out)])
+
+AC_CHECK_HEADERS(termios.h sys/statvfs.h sys/timerfd.h)
 
 case "$host_os" in
 	linux*)
@@ -105,16 +112,13 @@ case "$host_os" in
 		AC_CHECK_HEADERS(machine/endian.h sys/disk.h,,AC_MSG_ERROR(bailing out)) ;;
 esac
 
-AC_CHECK_HEADERS([ctype.h dirent.h errno.h fcntl.h getopt.h inttypes.h limits.h \
-  stdarg.h stdio.h stdlib.h string.h sys/ioctl.h sys/param.h sys/stat.h \
-  sys/types.h unistd.h], , [AC_MSG_ERROR(bailing out)])
-AC_CHECK_HEADERS(termios.h sys/statvfs.h)
-
 ################################################################################
 dnl -- Check for typedefs, structures, and compiler characteristics.
 AC_C_CONST
 AC_C_INLINE
 AC_CHECK_MEMBERS([struct stat.st_rdev])
+AC_CHECK_TYPES([ptrdiff_t])
+AC_STRUCT_TM
 AC_TYPE_OFF_T
 AC_TYPE_PID_T
 AC_TYPE_SIGNAL
@@ -130,15 +134,13 @@ AC_TYPE_UINT8_T
 AC_TYPE_UINT16_T
 AC_TYPE_UINT32_T
 AC_TYPE_UINT64_T
-AC_CHECK_MEMBERS([struct stat.st_rdev])
-AC_STRUCT_TM
 
 ################################################################################
 dnl -- Check for functions
-AC_CHECK_FUNCS([ftruncate gethostname getpagesize \
-  gettimeofday memset mkdir mkfifo rmdir munmap nl_langinfo setenv setlocale \
-  strcasecmp strchr strcspn strspn strdup strncasecmp strerror strrchr \
-  strstr strtol strtoul uname], , [AC_MSG_ERROR(bailing out)])
+AC_CHECK_FUNCS([ftruncate gethostname getpagesize gettimeofday localtime_r \
+  memchr memset mkdir mkfifo munmap nl_langinfo realpath rmdir setenv \
+  setlocale strcasecmp strchr strcspn strdup strerror strncasecmp strndup \
+  strrchr strspn strstr strtol strtoul uname], , [AC_MSG_ERROR(bailing out)])
 AC_FUNC_ALLOCA
 AC_FUNC_CLOSEDIR_VOID
 AC_FUNC_CHOWN
@@ -146,12 +148,22 @@ AC_FUNC_FORK
 AC_FUNC_LSTAT
 AC_FUNC_MALLOC
 AC_FUNC_MEMCMP
+AC_FUNC_MKTIME
 AC_FUNC_MMAP
 AC_FUNC_REALLOC
 AC_FUNC_STAT
 AC_FUNC_STRTOD
 AC_FUNC_VPRINTF
 
+################################################################################
+dnl -- Disable dependency tracking
+AC_MSG_CHECKING(whether to enable dependency tracking)
+AC_ARG_ENABLE(dependency-tracking,
+	      AC_HELP_STRING([--disable-dependency-tracking],
+			     [speeds up one-time build.]),
+	      USE_TRACKING=$enableval, USE_TRACKING=yes)
+AC_MSG_RESULT($USE_TRACKING)
+
 ################################################################################
 dnl -- Enables statically-linked tools
 AC_MSG_CHECKING(whether to use static linking)
@@ -172,6 +184,9 @@ AC_SUBST(HAVE_FULL_RELRO)
 ################################################################################
 dnl -- Prefix is /usr by default, the exec_prefix default is setup later
 AC_PREFIX_DEFAULT(/usr)
+if test "$prefix" = NONE; then
+  datarootdir=${ac_default_prefix}/share
+fi
 
 ################################################################################
 dnl -- Setup the ownership of the files
@@ -558,6 +573,12 @@ case "$CACHE" in
  *) AC_MSG_ERROR([--with-cache parameter invalid]) ;;
 esac
 
+dnl -- cache_check needs-check flag
+AC_ARG_ENABLE(cache_check_needs_check,
+	      AC_HELP_STRING([--disable-cache_check_needs_check],
+			     [required if cache_check version is < 0.5]),
+	      CACHE_CHECK_NEEDS_CHECK=$enableval, CACHE_CHECK_NEEDS_CHECK=yes)
+
 # Test if necessary cache tools are available
 # if not - use plain defaults and warn user
 case "$CACHE" in
@@ -571,6 +592,28 @@ case "$CACHE" in
 			CACHE_CONFIGURE_WARN=y
 		fi
 	fi
+	if test "$CACHE_CHECK_NEEDS_CHECK" = yes; then
+		$CACHE_CHECK_CMD -V 2>/dev/null >conftest.tmp
+		read -r CACHE_CHECK_VSN < conftest.tmp
+		IFS=.- read -r CACHE_CHECK_VSN_MAJOR CACHE_CHECK_VSN_MINOR CACHE_CHECK_VSN_PATCH LEFTOVER < conftest.tmp
+		rm -f conftest.tmp
+
+		# Require version >= 0.5.4 for --clear-needs-check-flag
+		if test -z "$CACHE_CHECK_VSN_MAJOR" \
+			|| test -z "$CACHE_CHECK_VSN_MINOR" \
+			|| test -z "$CACHE_CHECK_VSN_PATCH"; then
+			AC_MSG_WARN([$CACHE_CHECK_CMD: Bad version "$CACHE_CHECK_VSN" found])
+			CACHE_CHECK_VERSION_WARN=y
+			CACHE_CHECK_NEEDS_CHECK=no
+		elif test "$CACHE_CHECK_VSN_MAJOR" -eq 0 ; then
+			if test "$CACHE_CHECK_VSN_MINOR" -lt 5 \
+				|| test "$CACHE_CHECK_VSN_MINOR" -eq 5 -a "$CACHE_CHECK_VSN_PATCH" -lt 4; then
+				AC_MSG_WARN([$CACHE_CHECK_CMD: Old version "$CACHE_CHECK_VSN" found])
+				CACHE_CHECK_VERSION_WARN=y
+				CACHE_CHECK_NEEDS_CHECK=no
+			fi
+		fi
+	fi
 	# Empty means a config way to ignore cache dumping
 	if test "$CACHE_DUMP_CMD" = "autodetect"; then
 		AC_PATH_TOOL(CACHE_DUMP_CMD, cache_dump)
@@ -598,6 +641,12 @@ case "$CACHE" in
 			CACHE_CONFIGURE_WARN=y
 		}
 	fi
+
+	AC_MSG_CHECKING([whether cache_check supports the needs-check flag])
+	AC_MSG_RESULT([$CACHE_CHECK_NEEDS_CHECK])
+	if test "$CACHE_CHECK_NEEDS_CHECK" = yes; then
+		AC_DEFINE([CACHE_CHECK_NEEDS_CHECK], 1, [Define to 1 if the external 'cache_check' tool requires the --clear-needs-check-flag option])
+	fi
 	;;
 esac
 
@@ -626,8 +675,8 @@ AC_MSG_RESULT($READLINE)
 dnl -- Disable realtime clock support
 AC_MSG_CHECKING(whether to enable realtime support)
 AC_ARG_ENABLE(realtime,
-	      AC_HELP_STRING([--enable-realtime], [enable realtime clock support]),
-	      REALTIME=$enableval)
+	      AC_HELP_STRING([--disable-realtime], [disable realtime clock support]),
+	      REALTIME=$enableval, REALTIME=yes)
 AC_MSG_RESULT($REALTIME)
 
 ################################################################################
@@ -1032,6 +1081,13 @@ if test "$TESTING" = yes; then
 	PKG_CHECK_MODULES(CUNIT, cunit >= 2.0)
 fi
 
+################################################################################
+dnl -- Set LVM2 testsuite data
+TESTSUITE_DATA='${datarootdir}/lvm2-testsuite'
+# double eval needed ${datarootdir} -> ${prefix}/share -> real path
+AC_DEFINE_UNQUOTED(TESTSUITE_DATA, ["$(eval echo $(eval echo $TESTSUITE_DATA))"], [Path to testsuite data])
+
+
 ################################################################################
 dnl -- Enable valgrind awareness of memory pools
 AC_MSG_CHECKING(whether to enable valgrind awareness of pools)
@@ -1077,6 +1133,96 @@ AC_MSG_RESULT($LVMETAD)
 
 BUILD_LVMETAD=$LVMETAD
 
+################################################################################
+dnl -- Build lvmpolld
+AC_MSG_CHECKING(whether to build lvmpolld)
+AC_ARG_ENABLE(lvmpolld,
+	      AC_HELP_STRING([--enable-lvmpolld],
+			     [enable the LVM Polling Daemon]),
+	      LVMPOLLD=$enableval)
+AC_MSG_RESULT($LVMPOLLD)
+
+BUILD_LVMPOLLD=$LVMPOLLD
+
+################################################################################
+dnl -- Build lockdsanlock
+AC_MSG_CHECKING(whether to build lockdsanlock)
+AC_ARG_ENABLE(lockd-sanlock,
+	      AC_HELP_STRING([--enable-lockd-sanlock],
+			     [enable the LVM lock daemon using sanlock]),
+	      LOCKDSANLOCK=$enableval)
+AC_MSG_RESULT($LOCKDSANLOCK)
+
+BUILD_LOCKDSANLOCK=$LOCKDSANLOCK
+
+if test "$BUILD_LOCKDSANLOCK" = yes; then
+	AC_DEFINE([LOCKDSANLOCK_SUPPORT], 1, [Define to 1 to include code that uses lvmlockd sanlock option.])
+fi
+
+################################################################################
+dnl -- Look for sanlock libraries
+if test "$BUILD_LOCKDSANLOCK" = yes; then
+	PKG_CHECK_MODULES(LOCKD_SANLOCK, libsanlock_client, [HAVE_LOCKD_SANLOCK=yes], $bailout)
+	BUILD_LVMLOCKD=yes
+fi
+
+################################################################################
+dnl -- Build lockddlm
+AC_MSG_CHECKING(whether to build lockddlm)
+AC_ARG_ENABLE(lockd-dlm,
+	      AC_HELP_STRING([--enable-lockd-dlm],
+			     [enable the LVM lock daemon using dlm]),
+	      LOCKDDLM=$enableval)
+AC_MSG_RESULT($LOCKDDLM)
+
+BUILD_LOCKDDLM=$LOCKDDLM
+
+if test "$BUILD_LOCKDDLM" = yes; then
+	AC_DEFINE([LOCKDDLM_SUPPORT], 1, [Define to 1 to include code that uses lvmlockd dlm option.])
+fi
+
+################################################################################
+dnl -- Look for dlm libraries
+if test "$BUILD_LOCKDDLM" = yes; then
+	PKG_CHECK_MODULES(LOCKD_DLM, libdlm, [HAVE_LOCKD_DLM=yes], $bailout)
+	BUILD_LVMLOCKD=yes
+fi
+
+################################################################################
+dnl -- Build lvmlockd
+
+AC_MSG_CHECKING(whether to build lvmlockd)
+AC_MSG_RESULT($BUILD_LVMLOCKD)
+
+if test "$BUILD_LVMLOCKD" = yes; then
+	AS_IF([test -n "$BUILD_LVMPOLLD"], [BUILD_LVMPOLLD=yes; AC_MSG_WARN([Enabling lvmpolld - required by lvmlockd.])])
+	AS_IF([test -n "$BUILD_LVMETAD"], [BUILD_LVMETAD=yes; AC_MSG_WARN([Enabling lvmetad - required by lvmlockd.])])
+	AC_MSG_CHECKING([defaults for use_lvmlockd])
+	AC_ARG_ENABLE(use_lvmlockd,
+		      AC_HELP_STRING([--disable-use-lvmlockd],
+				     [disable usage of LVM lock daemon]),
+		      [case ${enableval} in
+		       yes) DEFAULT_USE_LVMLOCKD=1 ;;
+		       *) DEFAULT_USE_LVMLOCKD=0 ;;
+		       esac], DEFAULT_USE_LVMLOCKD=1)
+	AC_MSG_RESULT($DEFAULT_USE_LVMLOCKD)
+	AC_DEFINE([LVMLOCKD_SUPPORT], 1, [Define to 1 to include code that uses lvmlockd.])
+
+	AC_ARG_WITH(lvmlockd-pidfile,
+		    AC_HELP_STRING([--with-lvmlockd-pidfile=PATH],
+				   [lvmlockd pidfile [PID_DIR/lvmlockd.pid]]),
+		    LVMLOCKD_PIDFILE=$withval,
+		    LVMLOCKD_PIDFILE="$DEFAULT_PID_DIR/lvmlockd.pid")
+	AC_DEFINE_UNQUOTED(LVMLOCKD_PIDFILE, ["$LVMLOCKD_PIDFILE"],
+			   [Path to lvmlockd pidfile.])
+else
+	DEFAULT_USE_LVMLOCKD=0
+fi
+AC_DEFINE_UNQUOTED(DEFAULT_USE_LVMLOCKD, [$DEFAULT_USE_LVMLOCKD],
+		   [Use lvmlockd by default.])
+
+################################################################################
+dnl -- Check lvmetad
 if test "$BUILD_LVMETAD" = yes; then
 	AC_MSG_CHECKING([defaults for use_lvmetad])
 	AC_ARG_ENABLE(use_lvmetad,
@@ -1103,16 +1249,7 @@ AC_DEFINE_UNQUOTED(DEFAULT_USE_LVMETAD, [$DEFAULT_USE_LVMETAD],
 		   [Use lvmetad by default.])
 
 ################################################################################
-dnl -- Build lvmpolld
-AC_MSG_CHECKING(whether to build lvmpolld)
-AC_ARG_ENABLE(lvmpolld,
-	      AC_HELP_STRING([--enable-lvmpolld],
-			     [enable the LVM Polling Daemon]),
-	      LVMPOLLD=$enableval)
-AC_MSG_RESULT($LVMPOLLD)
-
-BUILD_LVMPOLLD=$LVMPOLLD
-
+dnl -- Check lvmpolld
 if test "$BUILD_LVMPOLLD" = yes; then
 	AC_MSG_CHECKING([defaults for use_lvmpolld])
 	AC_ARG_ENABLE(use_lvmpolld,
@@ -1138,49 +1275,6 @@ fi
 AC_DEFINE_UNQUOTED(DEFAULT_USE_LVMPOLLD, [$DEFAULT_USE_LVMPOLLD],
 		   [Use lvmpolld by default.])
 
-################################################################################
-dnl -- Build lvmlockd
-AC_MSG_CHECKING(whether to build lvmlockd)
-AC_ARG_ENABLE(lvmlockd,
-	      AC_HELP_STRING([--enable-lvmlockd],
-			     [enable the LVM lock daemon]),
-	      LVMLOCKD=$enableval)
-AC_MSG_RESULT($LVMLOCKD)
-
-BUILD_LVMLOCKD=$LVMLOCKD
-
-if test "$BUILD_LVMLOCKD" = yes; then
-	AC_MSG_CHECKING([defaults for use_lvmlockd])
-	AC_ARG_ENABLE(use_lvmlockd,
-		      AC_HELP_STRING([--disable-use-lvmlockd],
-				     [disable usage of LVM lock daemon]),
-		      [case ${enableval} in
-		       yes) DEFAULT_USE_LVMLOCKD=1 ;;
-		       *) DEFAULT_USE_LVMLOCKD=0 ;;
-		       esac], DEFAULT_USE_LVMLOCKD=1)
-	AC_MSG_RESULT($DEFAULT_USE_LVMLOCKD)
-	AC_DEFINE([LVMLOCKD_SUPPORT], 1, [Define to 1 to include code that uses lvmlockd.])
-
-	AC_ARG_WITH(lvmlockd-pidfile,
-		    AC_HELP_STRING([--with-lvmlockd-pidfile=PATH],
-				   [lvmlockd pidfile [PID_DIR/lvmlockd.pid]]),
-		    LVMLOCKD_PIDFILE=$withval,
-		    LVMLOCKD_PIDFILE="$DEFAULT_PID_DIR/lvmlockd.pid")
-	AC_DEFINE_UNQUOTED(LVMLOCKD_PIDFILE, ["$LVMLOCKD_PIDFILE"],
-			   [Path to lvmlockd pidfile.])
-else
-	DEFAULT_USE_LVMLOCKD=0
-fi
-AC_DEFINE_UNQUOTED(DEFAULT_USE_LVMLOCKD, [$DEFAULT_USE_LVMLOCKD],
-		   [Use lvmlockd by default.])
-
-################################################################################
-dnl -- Look for sanlock and dlm libraries
-if test "$BUILD_LVMLOCKD" = yes; then
-	PKG_CHECK_MODULES(LOCKD_SANLOCK, libsanlock_client, [HAVE_LOCKD_SANLOCK=yes], $bailout)
-	PKG_CHECK_MODULES(LOCKD_DLM, libdlm, [HAVE_LOCKD_DLM=yes], $bailout)
-fi
-
 ################################################################################
 
 dnl -- Enable blkid wiping functionality
@@ -1452,6 +1546,10 @@ if [[ \( "$LVM1" = shared -o "$POOL" = shared -o "$CLUSTER" = shared \
 	AC_MSG_ERROR([Features cannot be 'shared' when building statically])
 fi
 
+################################################################################
+AC_CHECK_LIB(m, log10,
+	[M_LIBS="-lm"], hard_bailout)
+
 ################################################################################
 AC_CHECK_LIB([pthread], [pthread_mutex_lock],
 	[PTHREAD_LIBS="-lpthread"], hard_bailout)
@@ -1492,6 +1590,7 @@ if test "$REALTIME" = yes; then
 	if test "$HAVE_REALTIME" = yes; then
 		AC_DEFINE([HAVE_REALTIME], 1, [Define to 1 to include support for realtime clock.])
 		LIBS="-lrt $LIBS"
+		RT_LIB="-lrt"
 	else
 		AC_MSG_WARN(Disabling realtime clock)
 	fi
@@ -1642,6 +1741,19 @@ if test "$READLINE" = yes; then
 	AC_CHECK_HEADERS(readline/readline.h readline/history.h,,hard_bailout)
 fi
 
+if test "$BUILD_CMIRRORD" = yes; then
+	AC_CHECK_FUNCS(atexit,,hard_bailout)
+fi
+
+if test "$BUILD_LVMLOCKD" = yes; then
+	AC_CHECK_FUNCS(clock_gettime strtoull,,hard_bailout)
+fi
+
+if test "$BUILD_LVMPOLLD" = yes; then
+	AC_CHECK_FUNCS(strpbrk,,hard_bailout)
+	AC_FUNC_STRERROR_R
+fi
+
 if test "$CLVMD" != none; 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,,hard_bailout)
@@ -1684,12 +1796,10 @@ test "$lvm_exec_prefix" = NONE && lvm_exec_prefix=$ac_default_prefix
 LVM_PATH="$lvm_exec_prefix/sbin/lvm"
 AC_DEFINE_UNQUOTED(LVM_PATH, ["$LVM_PATH"], [Path to lvm binary.])
 
-if  test "$CLVMD" != none; then
-	clvmd_prefix=$ac_default_prefix
-	test "$prefix" != NONE && clvmd_prefix=$prefix
-	CLVMD_PATH="$clvmd_prefix/sbin/clvmd"
-	AC_DEFINE_UNQUOTED(CLVMD_PATH, ["$CLVMD_PATH"], [Path to clvmd binary.])
-fi
+clvmd_prefix=$ac_default_prefix
+test "$prefix" != NONE && clvmd_prefix=$prefix
+CLVMD_PATH="$clvmd_prefix/sbin/clvmd"
+AC_DEFINE_UNQUOTED(CLVMD_PATH, ["$CLVMD_PATH"], [Path to clvmd binary.])
 
 ################################################################################
 dnl -- dmeventd pidfile and executable path
@@ -1778,14 +1888,14 @@ test "$interface" != ioctl && AC_MSG_ERROR([--with-interface=ioctl required. fs
 AC_MSG_RESULT($interface)
 
 ################################################################################
-DM_LIB_VERSION="\"`cat "$srcdir"/VERSION_DM 2>/dev/null || echo Unknown`\""
-AC_DEFINE_UNQUOTED(DM_LIB_VERSION, $DM_LIB_VERSION, [Library version])
+read DM_LIB_VERSION < "$srcdir"/VERSION_DM 2>/dev/null || DM_LIB_VERSION=Unknown
+AC_DEFINE_UNQUOTED(DM_LIB_VERSION, "$DM_LIB_VERSION", [Library version])
 
 DM_LIB_PATCHLEVEL=`cat "$srcdir"/VERSION_DM | $AWK -F '[[-. ]]' '{printf "%s.%s.%s",$1,$2,$3}'`
 
-LVM_VERSION="\"`cat "$srcdir"/VERSION 2>/dev/null || echo Unknown`\""
+read VER < "$srcdir"/VERSION 2>/dev/null || VER=Unknown
 
-VER=`cat "$srcdir"/VERSION`
+LVM_VERSION=\"$VER\"
 LVM_RELEASE_DATE="\"`echo $VER | $SED 's/.* (//;s/).*//'`\""
 VER=`echo "$VER" | $AWK '{print $1}'`
 LVM_RELEASE="\"`echo "$VER" | $AWK -F '-' '{print $2}'`\""
@@ -1804,6 +1914,8 @@ AC_SUBST(BUILD_DMEVENTD)
 AC_SUBST(BUILD_LVMETAD)
 AC_SUBST(BUILD_LVMPOLLD)
 AC_SUBST(BUILD_LVMLOCKD)
+AC_SUBST(BUILD_LOCKDSANLOCK)
+AC_SUBST(BUILD_LOCKDDLM)
 AC_SUBST(CACHE)
 AC_SUBST(CFLAGS)
 AC_SUBST(CFLOW_CMD)
@@ -1851,7 +1963,6 @@ AC_SUBST(DLM_LIBS)
 AC_SUBST(DL_LIBS)
 AC_SUBST(DMEVENTD)
 AC_SUBST(DMEVENTD_PATH)
-AC_SUBST(DM_LIB_VERSION)
 AC_SUBST(DM_LIB_PATCHLEVEL)
 AC_SUBST(ELDFLAGS)
 AC_SUBST(FSADM)
@@ -1882,6 +1993,7 @@ AC_SUBST(OCF)
 AC_SUBST(OCFDIR)
 AC_SUBST(PKGCONFIG)
 AC_SUBST(POOL)
+AC_SUBST(M_LIBS)
 AC_SUBST(PTHREAD_LIBS)
 AC_SUBST(PYTHON)
 AC_SUBST(PYTHON_BINDINGS)
@@ -1890,6 +2002,7 @@ AC_SUBST(PYTHON_LIBDIRS)
 AC_SUBST(QUORUM_CFLAGS)
 AC_SUBST(QUORUM_LIBS)
 AC_SUBST(RAID)
+AC_SUBST(RT_LIB)
 AC_SUBST(READLINE_LIBS)
 AC_SUBST(REPLICATORS)
 AC_SUBST(SACKPT_CFLAGS)
@@ -1902,6 +2015,7 @@ AC_SUBST(SNAPSHOTS)
 AC_SUBST(STATICDIR)
 AC_SUBST(STATIC_LINK)
 AC_SUBST(TESTING)
+AC_SUBST(TESTSUITE_DATA)
 AC_SUBST(THIN)
 AC_SUBST(THIN_CHECK_CMD)
 AC_SUBST(THIN_DUMP_CMD)
@@ -1917,6 +2031,7 @@ AC_SUBST(UDEV_SYNC)
 AC_SUBST(UDEV_SYSTEMD_BACKGROUND_JOBS)
 AC_SUBST(UDEV_RULE_EXEC_DETECTION)
 AC_SUBST(UDEV_HAS_BUILTIN_BLKID)
+AC_SUBST(USE_TRACKING)
 AC_SUBST(VALGRIND_POOL)
 AC_SUBST(WRITE_INSTALL)
 AC_SUBST(DMEVENTD_PIDFILE)
@@ -1971,7 +2086,7 @@ lib/format_pool/Makefile
 lib/locking/Makefile
 lib/mirror/Makefile
 lib/replicator/Makefile
-lib/misc/lvm-version.h
+include/lvm-version.h
 lib/raid/Makefile
 lib/snapshot/Makefile
 lib/thin/Makefile
@@ -2033,9 +2148,3 @@ AS_IF([test -n "$CACHE_CONFIGURE_WARN"],
 
 AS_IF([test "$ODIRECT" != yes],
       [AC_MSG_WARN([O_DIRECT disabled: low-memory pvmove may lock up])])
-
-AS_IF([test "$BUILD_LVMLOCKD" == yes && test "$BUILD_LVMPOLLD" == no],
-      [AC_MSG_WARN([lvmlockd requires lvmpolld])])
-
-AS_IF([test "$BUILD_LVMLOCKD" == yes && test "$BUILD_LVMETAD" == no],
-      [AC_MSG_WARN([lvmlockd requires lvmetad])])

daemons/clvmd/.gitignore

@@ -0,0 +1 @@
+clvmd

daemons/clvmd/clvmd-command.c

@@ -323,6 +323,7 @@ void cmd_client_cleanup(struct local_client *client)
 	int lkid;
 	char *lockname;
 
+	DEBUGLOG("Client thread cleanup (%p)\n", client);
 	if (!client->bits.localsock.private)
 		return;
 
@@ -331,7 +332,7 @@ void cmd_client_cleanup(struct local_client *client)
 	dm_hash_iterate(v, lock_hash) {
 		lkid = (int)(long)dm_hash_get_data(lock_hash, v);
 		lockname = dm_hash_get_key(lock_hash, v);
-		DEBUGLOG("cleanup: Unlocking lock %s %x\n", lockname, lkid);
+		DEBUGLOG("Cleanup (%p): Unlocking lock %s %x\n", client, lockname, lkid);
 		(void) sync_unlock(lockname, lkid);
 	}
 
@@ -339,7 +340,6 @@ void cmd_client_cleanup(struct local_client *client)
 	client->bits.localsock.private = NULL;
 }
 
-
 static int restart_clvmd(void)
 {
 	const char **argv;

daemons/clvmd/clvmd-common.h

@@ -18,15 +18,10 @@
 #ifndef _LVM_CLVMD_COMMON_H
 #define _LVM_CLVMD_COMMON_H
 
-#include "configure.h"
-
 #define _REENTRANT
-#define _GNU_SOURCE
-#define _FILE_OFFSET_BITS 64
 
-#include "libdevmapper.h"
+#include "tool.h"
+
 #include "lvm-logging.h"
 
-#include <unistd.h>
-
 #endif

daemons/clvmd/clvmd-openais.c

@@ -243,7 +243,7 @@ static void openais_cpg_confchg_callback(cpg_handle_t handle,
 	struct node_info *ninfo;
 
 	DEBUGLOG("confchg callback. %" PRIsize_t " joined, "
-		 "%" PRIsize_t " left, %" PRIsize_t " members\n",
+		 FMTsize_t " left, %" PRIsize_t " members\n",
 		 joined_list_entries, left_list_entries, member_list_entries);
 
 	for (i=0; i<joined_list_entries; i++) {

daemons/clvmd/clvmd.c

@@ -154,7 +154,7 @@ static void usage(const char *prog, FILE *file)
 {
 	fprintf(file, "Usage: %s [options]\n"
 		"   -C       Sets debug level (from -d) on all clvmd instances clusterwide\n"
-		"   -d[n]    Set debug logging (0:none, 1:stderr (implies -f option), 2:syslog)\n"
+		"   -d[<n>]  Set debug logging (0:none, 1:stderr (implies -f option), 2:syslog)\n"
 		"   -E<uuid> Take this lock uuid as exclusively locked resource (for restart)\n"
 		"   -f       Don't fork, run in the foreground\n"
 		"   -h       Show this help information\n"
@@ -223,6 +223,7 @@ void debuglog(const char *fmt, ...)
 		fprintf(stderr, "CLVMD[%x]: %.15s ", (int)pthread_self(), ctime_r(&P, buf_ctime) + 4);
 		vfprintf(stderr, fmt, ap);
 		va_end(ap);
+		fflush(stderr);
 		break;
 	case DEBUG_SYSLOG:
 		if (!syslog_init) {
@@ -598,7 +599,9 @@ 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");
+	DEBUGLOG("Starting LVM thread\n");
+	DEBUGLOG("Main cluster socket fd %d (%p) with local socket %d (%p)\n",
+		 local_client_head.fd, &local_client_head, newfd->fd, newfd);
 
 	/* Don't let anyone else to do work until we are started */
 	pthread_create(&lvm_thread, &stack_attr, lvm_thread_fn, &lvm_params);
@@ -698,7 +701,7 @@ static int local_rendezvous_callback(struct local_client *thisfd, char *buf,
 		newfd->type = LOCAL_SOCK;
 		newfd->callback = local_sock_callback;
 		newfd->bits.localsock.all_success = 1;
-		DEBUGLOG("Got new connection on fd %d\n", newfd->fd);
+		DEBUGLOG("Got new connection on fd %d (%p)\n", newfd->fd, newfd);
 		*new_client = newfd;
 	}
 	return 1;
@@ -850,18 +853,48 @@ static void main_loop(int cmd_timeout)
 		struct local_client *thisfd;
 		struct timeval tv = { cmd_timeout, 0 };
 		int quorate = clops->is_quorate();
+		int client_count = 0;
+		int max_fd = 0;
+		struct local_client *lastfd = &local_client_head;
+		struct local_client *nextfd = local_client_head.next;
 
 		/* Wait on the cluster FD and all local sockets/pipes */
 		local_client_head.fd = clops->get_main_cluster_fd();
 		FD_ZERO(&in);
+
 		for (thisfd = &local_client_head; thisfd; thisfd = thisfd->next) {
+			client_count++;
+			max_fd = max(max_fd, thisfd->fd);
+		}
+
+		if (max_fd > FD_SETSIZE - 32) {
+			fprintf(stderr, "WARNING: There are too many connections to clvmd.  Investigate and take action now!\n");
+ 			fprintf(stderr, "WARNING: Your cluster may freeze up if the number of clvmd file descriptors (%d) exceeds %d.\n", max_fd + 1, FD_SETSIZE);
+		}
+
+		for (thisfd = &local_client_head; thisfd; thisfd = nextfd, nextfd = thisfd ? thisfd->next : NULL) {
+
+			if (thisfd->removeme && !cleanup_zombie(thisfd)) {
+				struct local_client *free_fd = thisfd;
+				lastfd->next = nextfd;
+				DEBUGLOG("removeme set for %p with %d monitored fds remaining\n", free_fd, client_count - 1);
+
+				/* Queue cleanup, this also frees the client struct */
+				add_to_lvmqueue(free_fd, NULL, 0, NULL);
+
+				continue;
+			}
+
+			lastfd = thisfd;
+
 			if (thisfd->removeme)
 				continue;
 
 			/* if the cluster is not quorate then don't listen for new requests */
 			if ((thisfd->type != LOCAL_RENDEZVOUS &&
 			     thisfd->type != LOCAL_SOCK) || quorate)
-				FD_SET(thisfd->fd, &in);
+				if (thisfd->fd < FD_SETSIZE)
+					FD_SET(thisfd->fd, &in);
 		}
 
 		select_status = select(FD_SETSIZE, &in, NULL, NULL, &tv);
@@ -877,31 +910,20 @@ static void main_loop(int cmd_timeout)
 		}
 
 		if (select_status > 0) {
-			struct local_client *lastfd = NULL;
 			char csid[MAX_CSID_LEN];
 			char buf[max_cluster_message];
 
 			for (thisfd = &local_client_head; thisfd; thisfd = thisfd->next) {
-				if (thisfd->removeme && !cleanup_zombie(thisfd)) {
-					struct local_client *free_fd = thisfd;
-					lastfd->next = thisfd->next;
-					DEBUGLOG("removeme set for fd %d\n", free_fd->fd);
-
-					/* Queue cleanup, this also frees the client struct */
-					add_to_lvmqueue(free_fd, NULL, 0, NULL);
-					break;
-				}
-
-				if (FD_ISSET(thisfd->fd, &in)) {
+				if (thisfd->fd < FD_SETSIZE && FD_ISSET(thisfd->fd, &in)) {
 					struct local_client *newfd = NULL;
 					int ret;
 
+					/* FIXME Remove from main thread in case it blocks! */
 					/* Do callback */
 					ret = thisfd->callback(thisfd, buf, sizeof(buf),
 							       csid, &newfd);
 					/* Ignore EAGAIN */
 					if (ret < 0 && (errno == EAGAIN || errno == EINTR)) {
-						lastfd = thisfd;
 						continue;
                                         }
 
@@ -917,17 +939,16 @@ static void main_loop(int cmd_timeout)
 						DEBUGLOG("ret == %d, errno = %d. removing client\n",
 							 ret, errno);
 						thisfd->removeme = 1;
-						break;
+						continue;
 					}
 
 					/* New client...simply add it to the list */
 					if (newfd) {
 						newfd->next = thisfd->next;
 						thisfd->next = newfd;
-						break;
+						thisfd = newfd;
 					}
 				}
-				lastfd = thisfd;
 			}
 		}
 
@@ -1420,7 +1441,7 @@ static int read_from_local_sock(struct local_client *thisfd)
 	thisfd->bits.localsock.in_progress = TRUE;
 	thisfd->bits.localsock.state = PRE_COMMAND;
 	thisfd->bits.localsock.cleanup_needed = 1;
-	DEBUGLOG("Creating pre&post thread\n");
+	DEBUGLOG("Creating pre&post thread for pipe fd %d (%p)\n", newfd->fd, newfd);
 	status = pthread_create(&thisfd->bits.localsock.threadid,
 				&stack_attr, pre_and_post_thread, thisfd);
 	DEBUGLOG("Created pre&post thread, state = %d\n", status);
@@ -1674,7 +1695,7 @@ static __attribute__ ((noreturn)) void *pre_and_post_thread(void *arg)
 	sigset_t ss;
 	int pipe_fd = client->bits.localsock.pipe;
 
-	DEBUGLOG("Pre&post thread (%p), pipe %d\n", client, pipe_fd);
+	DEBUGLOG("Pre&post thread (%p), pipe fd %d\n", client, pipe_fd);
 	pthread_mutex_lock(&client->bits.localsock.mutex);
 
 	/* Ignore SIGUSR1 (handled by master process) but enable
@@ -1694,7 +1715,7 @@ static __attribute__ ((noreturn)) void *pre_and_post_thread(void *arg)
 		if ((status = do_pre_command(client)))
 			client->bits.localsock.all_success = 0;
 
-		DEBUGLOG("Pre&post thread (%p) writes status %d down to pipe %d\n",
+		DEBUGLOG("Pre&post thread (%p) writes status %d down to pipe fd %d\n",
 			 client, status, pipe_fd);
 
 		/* Tell the parent process we have finished this bit */
@@ -1976,7 +1997,7 @@ static int process_work_item(struct lvm_thread_cmd *cmd)
 {
 	/* If msg is NULL then this is a cleanup request */
 	if (cmd->msg == NULL) {
-		DEBUGLOG("process_work_item: free fd %d\n", cmd->client->fd);
+		DEBUGLOG("process_work_item: free %p\n", cmd->client);
 		cmd_client_cleanup(cmd->client);
 		pthread_mutex_destroy(&cmd->client->bits.localsock.mutex);
 		pthread_cond_destroy(&cmd->client->bits.localsock.cond);

daemons/clvmd/lvm-functions.c

@@ -510,7 +510,7 @@ int do_lock_lv(unsigned char command, unsigned char lock_flags, char *resource)
 	DEBUGLOG("do_lock_lv: resource '%s', cmd = %s, flags = %s, critical_section = %d\n",
 		 resource, decode_locking_cmd(command), decode_flags(lock_flags), critical_section());
 
-	if (!cmd->config_initialized || config_files_changed(cmd)) {
+	if (!cmd->initialized.config || config_files_changed(cmd)) {
 		/* Reinitialise various settings inc. logging, filters */
 		if (do_refresh_cache()) {
 			log_error("Updated config file invalid. Aborting.");
@@ -899,7 +899,7 @@ int init_clvm(struct dm_hash_table *excl_uuid)
 	if (!get_initial_state(excl_uuid))
 		log_error("Cannot load initial lock states.");
 
-	if (!(cmd = create_toolcontext(1, NULL, 0, 1))) {
+	if (!(cmd = create_toolcontext(1, NULL, 0, 1, 1, 1))) {
 		log_error("Failed to allocate command context");
 		return 0;
 	}

daemons/cmirrord/.gitignore

@@ -0,0 +1 @@
+cmirrord

daemons/cmirrord/clogd.c

@@ -15,6 +15,7 @@
 #include "link_mon.h"
 #include "local.h"
 
+#include <getopt.h>
 #include <errno.h>
 #include <fcntl.h>
 #include <sys/socket.h>
@@ -32,14 +33,49 @@ static void daemonize(void);
 static void init_all(void);
 static void cleanup_all(void);
 
-int main(int argc __attribute__((unused)), char *argv[] __attribute__((unused)))
+static void usage (FILE *dest)
 {
-	daemonize();
+	fprintf (dest, "Usage: cmirrord [options]\n"
+		 "   -f, --foreground    stay in the foreground, log to the terminal\n"
+		 "   -h, --help          print this help\n");
+}
+
+int main(int argc, char *argv[])
+{
+	int foreground_mode = 0;
+	struct option longopts[] = {
+		{ "foreground", no_argument, NULL, 'f' },
+		{ "help"      , no_argument, NULL, 'h' },
+		{ 0, 0, 0, 0 }
+	};
+	int opt;
+
+	while ((opt = getopt_long (argc, argv, "fh", longopts, NULL)) != -1) {
+		switch (opt) {
+		case 'f':
+			foreground_mode = 1;
+			break;
+		case 'h':
+			usage (stdout);
+			exit (0);
+		default:
+			usage (stderr);
+			exit (2);
+		}
+	}
+	if (optind < argc) {
+		usage (stderr);
+		exit (2);
+	}
+
+	if (!foreground_mode)
+		daemonize();
 
 	init_all();
 
 	/* Parent can now exit, we're ready to handle requests */
-	kill(getppid(), SIGTERM);
+	if (!foreground_mode)
+		kill(getppid(), SIGTERM);
 
 	LOG_PRINT("Starting cmirrord:");
 	LOG_PRINT(" Built: "__DATE__" "__TIME__"\n");
@@ -209,6 +245,16 @@ static void daemonize(void)
 	}
 
 	LOG_OPEN("cmirrord", LOG_PID, LOG_DAEMON);
+}
+
+/*
+ * init_all
+ *
+ * Initialize modules.  Exit on failure.
+ */
+static void init_all(void)
+{
+	int r;
 
 	(void) dm_prepare_selinux_context(CMIRRORD_PIDFILE, S_IFREG);
 	if (dm_create_lockfile(CMIRRORD_PIDFILE) == 0)
@@ -227,16 +273,6 @@ static void daemonize(void)
 	signal(SIGUSR2, &sig_handler);
 	sigemptyset(&signal_mask);
 	signal_received = 0;
-}
-
-/*
- * init_all
- *
- * Initialize modules.  Exit on failure.
- */
-static void init_all(void)
-{
-	int r;
 
 	if ((r = init_local()) ||
 	    (r = init_cluster())) {

daemons/cmirrord/cluster.c

@@ -104,10 +104,11 @@ static SaVersionT version = { 'B', 1, 1 };
 #endif
 
 #define DEBUGGING_HISTORY 100
+#define DEBUGGING_BUFLEN 128
 #define LOG_SPRINT(cc, f, arg...) do {				\
 		cc->idx++;					\
 		cc->idx = cc->idx % DEBUGGING_HISTORY;		\
-		sprintf(cc->debugging[cc->idx], f, ## arg);	\
+		snprintf(cc->debugging[cc->idx], DEBUGGING_BUFLEN, f, ## arg); \
 	} while (0)
 
 static int log_resp_rec = 0;
@@ -150,7 +151,7 @@ struct clog_cpg {
 	uint32_t checkpoint_requesters[MAX_CHECKPOINT_REQUESTERS];
 	struct checkpoint_data *checkpoint_list;
 	int idx;
-	char debugging[DEBUGGING_HISTORY][128];
+	char debugging[DEBUGGING_HISTORY][DEBUGGING_BUFLEN];
 };
 
 static struct dm_list clog_cpg_list;
@@ -1294,7 +1295,9 @@ static void cpg_join_callback(struct clog_cpg *match,
 	uint32_t my_pid = (uint32_t)getpid();
 	uint32_t lowest = match->lowest_id;
 	struct clog_request *rq;
-	char dbuf[32] = { 0 };
+	char dbuf[64] = { 0 };
+	char *dbuf_p = dbuf;
+	size_t dbuf_rem = sizeof dbuf;
 
 	/* Assign my_cluster_id */
 	if ((my_cluster_id == 0xDEAD) && (joined->pid == my_pid))
@@ -1310,9 +1313,17 @@ static void cpg_join_callback(struct clog_cpg *match,
 	if (joined->nodeid == my_cluster_id)
 		goto out;
 
-	for (i = 0; i < member_list_entries - 1; i++)
-		sprintf(dbuf+strlen(dbuf), "%u-", member_list[i].nodeid);
-	sprintf(dbuf+strlen(dbuf), "(%u)", joined->nodeid);
+	for (i = 0; i < member_list_entries - 1; i++) {
+		int written = snprintf(dbuf_p, dbuf_rem, "%u-", member_list[i].nodeid);
+		if (written < 0) continue; /* impossible */
+		if ((unsigned)written >= dbuf_rem) {
+			dbuf_rem = 0;
+			break;
+		}
+		dbuf_rem -= written;
+		dbuf_p += written;
+	}
+	snprintf(dbuf_p, dbuf_rem, "(%u)", joined->nodeid);
 	LOG_COND(log_checkpoint, "[%s] Joining node, %u needs checkpoint [%s]",
 		 SHORT_UUID(match->name.value), joined->nodeid, dbuf);
 

daemons/cmirrord/functions.c

@@ -32,12 +32,13 @@
 #define LOG_OFFSET 2
 
 #define RESYNC_HISTORY 50
+#define RESYNC_BUFLEN 128
 //static char resync_history[RESYNC_HISTORY][128];
 //static int idx = 0;
 #define LOG_SPRINT(_lc, f, arg...) do {					\
 		lc->idx++;						\
 		lc->idx = lc->idx % RESYNC_HISTORY;			\
-		sprintf(lc->resync_history[lc->idx], f, ## arg);	\
+		snprintf(lc->resync_history[lc->idx], RESYNC_BUFLEN, f, ## arg); \
 	} while (0)
 
 struct log_header {
@@ -88,7 +89,7 @@ struct log_c {
 	size_t disk_size;       /* size of disk_buffer in bytes */
 	void *disk_buffer;      /* aligned memory for O_DIRECT */
 	int idx;
-	char resync_history[RESYNC_HISTORY][128];
+	char resync_history[RESYNC_HISTORY][RESYNC_BUFLEN];
 };
 
 struct mark_entry {
@@ -1444,7 +1445,7 @@ static int disk_status_info(struct log_c *lc, struct dm_ulog_request *rq)
 	char *data = (char *)rq->data;
 	struct stat statbuf;
 
-	if(fstat(lc->disk_fd, &statbuf)) {
+	if (fstat(lc->disk_fd, &statbuf)) {
 		rq->error = -errno;
 		return -errno;
 	}
@@ -1507,7 +1508,7 @@ static int disk_status_table(struct log_c *lc, struct dm_ulog_request *rq)
 	char *data = (char *)rq->data;
 	struct stat statbuf;
 
-	if(fstat(lc->disk_fd, &statbuf)) {
+	if (fstat(lc->disk_fd, &statbuf)) {
 		rq->error = -errno;
 		return -errno;
 	}

daemons/dmeventd/.gitignore

@@ -0,0 +1 @@
+dmeventd

daemons/dmeventd/libdevmapper-event.c

@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2005-2007 Red Hat, Inc. All rights reserved.
+ * Copyright (C) 2005-2015 Red Hat, Inc. All rights reserved.
  *
  * This file is part of the device-mapper userspace tools.
  *
@@ -12,23 +12,22 @@
  * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
+#include "dm-logging.h"
 #include "dmlib.h"
 #include "libdevmapper-event.h"
-//#include "libmultilog.h"
 #include "dmeventd.h"
 
-#include <errno.h>
 #include <fcntl.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
 #include <sys/file.h>
 #include <sys/types.h>
 #include <sys/stat.h>
-#include <unistd.h>
 #include <sys/wait.h>
 #include <arpa/inet.h>		/* for htonl, ntohl */
+#include <pthread.h>
+#include <syslog.h>
 
+static int _debug_level = 0;
+static int _use_syslog = 0;
 static int _sequence_nr = 0;
 
 struct dm_event_handler {
@@ -199,7 +198,7 @@ static int _check_message_id(struct dm_event_daemon_message *msg)
 	if ((sscanf(msg->data, "%d:%d", &pid, &seq_nr) != 2) ||
 	    (pid != getpid()) || (seq_nr != _sequence_nr)) {
 		log_error("Ignoring out-of-sequence reply from dmeventd. "
-			  "Expected %d:%d but received %s", getpid(),
+			  "Expected %d:%d but received %s.", getpid(),
 			  _sequence_nr, msg->data);
 		return 0;
 	}
@@ -234,7 +233,7 @@ static int _daemon_read(struct dm_event_fifos *fifos,
 			FD_SET(fifos->server, &fds);
 			ret = select(fifos->server + 1, &fds, NULL, NULL, &tval);
 			if (ret < 0 && errno != EINTR) {
-				log_error("Unable to read from event server");
+				log_error("Unable to read from event server.");
 				return 0;
 			}
 			if ((ret == 0) && (i > 4) && !bytes) {
@@ -300,7 +299,7 @@ static int _daemon_write(struct dm_event_fifos *fifos,
 		if (ret < 0) {
 			if (errno == EINTR)
 				continue;
-			log_error("Unable to talk to event daemon");
+			log_error("Unable to talk to event daemon.");
 			return 0;
 		}
 		if (ret == 0)
@@ -309,7 +308,7 @@ static int _daemon_write(struct dm_event_fifos *fifos,
 		if (ret < 0) {
 			if ((errno == EINTR) || (errno == EAGAIN))
 				continue;
-			log_error("Unable to talk to event daemon");
+			log_error("Unable to talk to event daemon.");
 			return 0;
 		}
 	}
@@ -321,7 +320,7 @@ static int _daemon_write(struct dm_event_fifos *fifos,
 			FD_SET(fifos->client, &fds);
 			ret = select(fifos->client + 1, NULL, &fds, NULL, NULL);
 			if ((ret < 0) && (errno != EINTR)) {
-				log_error("Unable to talk to event daemon");
+				log_error("Unable to talk to event daemon.");
 				return 0;
 			}
 		} while (ret < 1);
@@ -331,7 +330,7 @@ static int _daemon_write(struct dm_event_fifos *fifos,
 			if ((errno == EINTR) || (errno == EAGAIN))
 				continue;
 			else {
-				log_error("Unable to talk to event daemon");
+				log_error("Unable to talk to event daemon.");
 				return 0;
 			}
 		}
@@ -361,7 +360,7 @@ int daemon_talk(struct dm_event_fifos *fifos,
 			  getpid(), _sequence_nr,
 			  dso_name ? : "-", dev_name ? : "-", evmask, timeout)))
 	    < 0) {
-		log_error("_daemon_talk: message allocation failed");
+		log_error("_daemon_talk: message allocation failed.");
 		return -ENOMEM;
 	}
 	msg->cmd = cmd;
@@ -449,11 +448,11 @@ static int _start_daemon(char *dmeventd_path, struct dm_event_fifos *fifos)
 
 	else if (!pid) {
 		execvp(args[0], args);
-		log_error("Unable to exec dmeventd: %s", strerror(errno));
+		log_error("Unable to exec dmeventd: %s.", strerror(errno));
 		_exit(EXIT_FAILURE);
 	} else {
 		if (waitpid(pid, &status, 0) < 0)
-			log_error("Unable to start dmeventd: %s",
+			log_error("Unable to start dmeventd: %s.",
 				  strerror(errno));
 		else if (WEXITSTATUS(status))
 			log_error("Unable to start dmeventd.");
@@ -526,7 +525,7 @@ static struct dm_task *_get_device_info(const struct dm_event_handler *dmevh)
 	struct dm_info info;
 
 	if (!(dmt = dm_task_create(DM_DEVICE_INFO))) {
-		log_error("_get_device_info: dm_task creation for info failed");
+		log_error("_get_device_info: dm_task creation for info failed.");
 		return NULL;
 	}
 
@@ -544,17 +543,17 @@ static struct dm_task *_get_device_info(const struct dm_event_handler *dmevh)
 
 	/* FIXME Add name or uuid or devno to messages */
 	if (!dm_task_run(dmt)) {
-		log_error("_get_device_info: dm_task_run() failed");
+		log_error("_get_device_info: dm_task_run() failed.");
 		goto bad;
 	}
 
 	if (!dm_task_get_info(dmt, &info)) {
-		log_error("_get_device_info: failed to get info for device");
+		log_error("_get_device_info: failed to get info for device.");
 		goto bad;
 	}
 
 	if (!info.exists) {
-		log_error("_get_device_info: %s%s%s%.0d%s%.0d%s%s: device not found",
+		log_error("_get_device_info: %s%s%s%.0d%s%.0d%s%s: device not found.",
 			  dmevh->uuid ? : "",
 			  (!dmevh->uuid && dmevh->dev_name) ? dmevh->dev_name : "",
 			  (!dmevh->uuid && !dmevh->dev_name && dmevh->major > 0) ? "(" : "",
@@ -623,12 +622,12 @@ int dm_event_register_handler(const struct dm_event_handler *dmevh)
 	    !strstr(dmevh->dso, "libdevmapper-event-lvm2snapshot.so") &&
 	    !strstr(dmevh->dso, "libdevmapper-event-lvm2mirror.so") &&
 	    !strstr(dmevh->dso, "libdevmapper-event-lvm2raid.so"))
-		log_warn("WARNING: %s: dmeventd plugins are deprecated", dmevh->dso);
+		log_warn("WARNING: %s: dmeventd plugins are deprecated.", dmevh->dso);
 
 
 	if ((err = _do_event(DM_EVENT_CMD_REGISTER_FOR_EVENT, dmevh->dmeventd_path, &msg,
 			     dmevh->dso, uuid, dmevh->mask, dmevh->timeout)) < 0) {
-		log_error("%s: event registration failed: %s",
+		log_error("%s: event registration failed: %s.",
 			  dm_task_get_name(dmt),
 			  msg.data ? msg.data : strerror(-err));
 		ret = 0;
@@ -655,7 +654,7 @@ int dm_event_unregister_handler(const struct dm_event_handler *dmevh)
 
 	if ((err = _do_event(DM_EVENT_CMD_UNREGISTER_FOR_EVENT, dmevh->dmeventd_path, &msg,
 			    dmevh->dso, uuid, dmevh->mask, dmevh->timeout)) < 0) {
-		log_error("%s: event deregistration failed: %s",
+		log_error("%s: event deregistration failed: %s.",
 			  dm_task_get_name(dmt),
 			  msg.data ? msg.data : strerror(-err));
 		ret = 0;
@@ -828,6 +827,79 @@ int dm_event_get_version(struct dm_event_fifos *fifos, int *version) {
 	return 1;
 }
 
+void dm_event_log_set(int debug_level, int use_syslog)
+{
+	_debug_level = debug_level;
+	_use_syslog = use_syslog;
+}
+
+void dm_event_log(const char *subsys, int level, const char *file,
+		  int line, int dm_errno_or_class,
+		  const char *format, va_list ap)
+{
+	static pthread_mutex_t _log_mutex = PTHREAD_MUTEX_INITIALIZER;
+	static time_t start = 0;
+	const char *indent = "";
+	FILE *stream = stdout;
+	int prio = -1;
+	time_t now;
+
+	switch (level & ~(_LOG_STDERR | _LOG_ONCE)) {
+	case _LOG_DEBUG:
+		if (_debug_level < 3)
+			return;
+		prio = LOG_DEBUG;
+		indent = "      ";
+		break;
+	case _LOG_INFO:
+		if (_debug_level < 2)
+			return;
+		prio = LOG_INFO;
+		indent = "    ";
+		break;
+	case _LOG_NOTICE:
+		if (_debug_level < 1)
+			return;
+		prio = LOG_NOTICE;
+		indent = "  ";
+		break;
+	case _LOG_WARN:
+		prio = LOG_WARNING;
+		break;
+	case _LOG_ERR:
+		prio = LOG_ERR;
+		stream = stderr;
+		break;
+	default:
+		prio = LOG_CRIT;
+	}
+
+	/* Serialize to keep lines readable */
+	pthread_mutex_lock(&_log_mutex);
+
+	if (_use_syslog) {
+		vsyslog(prio, format, ap);
+	} else {
+		now = time(NULL);
+		if (!start)
+			start = now;
+		now -= start;
+		fprintf(stream, "[%2d:%02d] %8x:%-6s%s",
+			(int)now / 60, (int)now % 60,
+			// TODO: Maybe use shorter ID
+			// ((int)(pthread_self()) >> 6) & 0xffff,
+			(int)pthread_self(), subsys,
+			(_debug_level > 3) ? "" : indent);
+		if (_debug_level > 3)
+			fprintf(stream, "%28s:%4d %s", file, line, indent);
+		vfprintf(stream, _(format), ap);
+		fputc('\n', stream);
+		fflush(stream);
+	}
+
+	pthread_mutex_unlock(&_log_mutex);
+}
+
 #if 0				/* left out for now */
 
 static char *_skip_string(char *src, const int delimiter)
@@ -861,7 +933,7 @@ int dm_event_get_timeout(const char *device_path, uint32_t *timeout)
 			     0, 0))) {
 		char *p = _skip_string(msg.data, ' ');
 		if (!p) {
-			log_error("malformed reply from dmeventd '%s'\n",
+			log_error("Malformed reply from dmeventd '%s'.",
 				  msg.data);
 			dm_free(msg.data);
 			return -EIO;

daemons/dmeventd/libdevmapper-event.h

@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2005-2007 Red Hat, Inc. All rights reserved.
+ * Copyright (C) 2005-2015 Red Hat, Inc. All rights reserved.
  *
  * This file is part of the device-mapper userspace tools.
  *
@@ -105,6 +105,25 @@ int dm_event_get_registered_device(struct dm_event_handler *dmevh, int next);
 int dm_event_register_handler(const struct dm_event_handler *dmevh);
 int dm_event_unregister_handler(const struct dm_event_handler *dmevh);
 
+/* Set debug level for logging, and whether to log on stdout/stderr or syslog */
+void dm_event_log_set(int debug_level, int use_syslog);
+
+/* Log messages acroding to current debug level  */
+__attribute__((format(printf, 6, 0)))
+void dm_event_log(const char *subsys, int level, const char *file,
+		  int line, int dm_errno_or_class,
+		  const char *format, va_list ap);
+/* Macro to route print_log do dm_event_log() */
+#define DM_EVENT_LOG_FN(subsys) \
+void print_log(int level, const char *file, int line, int dm_errno_or_class,\
+	       const char *format, ...)\
+{\
+	va_list ap;\
+	va_start(ap, format);\
+	dm_event_log(subsys, level, file, line, dm_errno_or_class, format, ap);\
+	va_end(ap);\
+}
+
 /* Prototypes for DSO interface, see dmeventd.c, struct dso_data for
    detailed descriptions. */
 // FIXME  misuse of bitmask as enum

daemons/dmeventd/plugins/lvm2/dmeventd_lvm.c

@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2010 Red Hat, Inc. All rights reserved.
+ * Copyright (C) 2010-2015 Red Hat, Inc. All rights reserved.
  *
  * This file is part of LVM2.
  *
@@ -13,15 +13,11 @@
  */
 
 #include "lib.h"
-#include "log.h"
-
-#include "lvm2cmd.h"
 #include "dmeventd_lvm.h"
+#include "libdevmapper-event.h"
+#include "lvm2cmd.h"
 
 #include <pthread.h>
-#include <syslog.h>
-
-extern int dmeventd_debug;
 
 /*
  * register_device() is called first and performs initialisation.
@@ -36,48 +32,19 @@ static int _register_count = 0;
 static struct dm_pool *_mem_pool = NULL;
 static void *_lvm_handle = NULL;
 
+DM_EVENT_LOG_FN("lvm")
+
+static void _lvm2_print_log(int level, const char *file, int line,
+			    int dm_errno_or_class, const char *msg)
+{
+	print_log(level, file, line, dm_errno_or_class, "%s", msg);
+}
+
 /*
  * Currently only one event can be processed at a time.
  */
 static pthread_mutex_t _event_mutex = PTHREAD_MUTEX_INITIALIZER;
 
-/*
- * FIXME Do not pass things directly to syslog, rather use the existing logging
- * facilities to sort logging ... however that mechanism needs to be somehow
- * configurable and we don't have that option yet
- */
-static void _temporary_log_fn(int level,
-			      const char *file __attribute__((unused)),
-			      int line __attribute__((unused)),
-			      int dm_errno __attribute__((unused)),
-			      const char *message)
-{
-	level &= ~(_LOG_STDERR | _LOG_ONCE);
-
-	switch (level) {
-	case _LOG_DEBUG:
-		if (dmeventd_debug >= 3)
-			syslog(LOG_DEBUG, "%s", message);
-		break;
-	case _LOG_INFO:
-		if (dmeventd_debug >= 2)
-			syslog(LOG_INFO, "%s", message);
-		break;
-	case _LOG_NOTICE:
-		if (dmeventd_debug >= 1)
-			syslog(LOG_NOTICE, "%s", message);
-		break;
-	case _LOG_WARN:
-		syslog(LOG_WARNING, "%s", message);
-		break;
-	case _LOG_ERR:
-		syslog(LOG_ERR, "%s", message);
-		break;
-	default:
-		syslog(LOG_CRIT, "%s", message);
-	}
-}
-
 void dmeventd_lvm2_lock(void)
 {
 	pthread_mutex_lock(&_event_mutex);
@@ -94,24 +61,26 @@ int dmeventd_lvm2_init(void)
 
 	pthread_mutex_lock(&_register_mutex);
 
-	/*
-	 * Need some space for allocations.  1024 should be more
-	 * than enough for what we need (device mapper name splitting)
-	 */
-	if (!_mem_pool && !(_mem_pool = dm_pool_create("mirror_dso", 1024)))
-		goto out;
-
 	if (!_lvm_handle) {
-		if (!getenv("LVM_LOG_FILE_EPOCH"))
-			lvm2_log_fn(_temporary_log_fn);
-		if (!(_lvm_handle = lvm2_init())) {
-			dm_pool_destroy(_mem_pool);
-			_mem_pool = NULL;
+		lvm2_log_fn(_lvm2_print_log);
+
+		if (!(_lvm_handle = lvm2_init()))
+			goto out;
+
+		/*
+		 * Need some space for allocations.  1024 should be more
+		 * than enough for what we need (device mapper name splitting)
+		 */
+		if (!_mem_pool && !(_mem_pool = dm_pool_create("mirror_dso", 1024))) {
+			lvm2_exit(_lvm_handle);
+			_lvm_handle = NULL;
 			goto out;
 		}
+
 		lvm2_disable_dmeventd_monitoring(_lvm_handle);
 		/* FIXME Temporary: move to dmeventd core */
 		lvm2_run(_lvm_handle, "_memlock_inc");
+		log_debug("lvm plugin initilized.");
 	}
 
 	_register_count++;
@@ -127,11 +96,13 @@ void dmeventd_lvm2_exit(void)
 	pthread_mutex_lock(&_register_mutex);
 
 	if (!--_register_count) {
+		log_debug("lvm plugin shuting down.");
 		lvm2_run(_lvm_handle, "_memlock_dec");
 		dm_pool_destroy(_mem_pool);
 		_mem_pool = NULL;
 		lvm2_exit(_lvm_handle);
 		_lvm_handle = NULL;
+		log_debug("lvm plugin exited.");
 	}
 
 	pthread_mutex_unlock(&_register_mutex);
@@ -154,8 +125,8 @@ int dmeventd_lvm2_command(struct dm_pool *mem, char *buffer, size_t size,
 	int r;
 
 	if (!dm_split_lvm_name(mem, device, &vg, &lv, &layer)) {
-		syslog(LOG_ERR, "Unable to determine VG name from %s.\n",
-		       device);
+		log_error("Unable to determine VG name from %s.",
+			  device);
 		return 0;
 	}
 
@@ -169,7 +140,7 @@ int dmeventd_lvm2_command(struct dm_pool *mem, char *buffer, size_t size,
 	dm_pool_free(mem, vg);
 
 	if (r < 0) {
-		syslog(LOG_ERR, "Unable to form LVM command. (too long).\n");
+		log_error("Unable to form LVM command. (too long).");
 		return 0;
 	}
 

daemons/dmeventd/plugins/lvm2/dmeventd_lvm.h

@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2010 Red Hat, Inc. All rights reserved.
+ * Copyright (C) 2010-2015 Red Hat, Inc. All rights reserved.
  *
  * This file is part of LVM2.
  *
@@ -39,4 +39,36 @@ struct dm_pool *dmeventd_lvm2_pool(void);
 int dmeventd_lvm2_command(struct dm_pool *mem, char *buffer, size_t size,
 			  const char *cmd, const char *device);
 
+#define dmeventd_lvm2_run_with_lock(cmdline) \
+	({\
+		int rc;\
+		dmeventd_lvm2_lock();\
+		rc = dmeventd_lvm2_run(cmdline);\
+		dmeventd_lvm2_unlock();\
+		rc;\
+	})
+
+#define dmeventd_lvm2_init_with_pool(name, st) \
+	({\
+		struct dm_pool *mem;\
+		st = NULL;\
+		if (dmeventd_lvm2_init()) {\
+			if ((mem = dm_pool_create(name, 2048)) &&\
+			    (st = dm_pool_zalloc(mem, sizeof(*st))))\
+				st->mem = mem;\
+			else {\
+				if (mem)\
+					dm_pool_destroy(mem);\
+				dmeventd_lvm2_exit();\
+			}\
+		}\
+		st;\
+	})
+
+#define dmeventd_lvm2_exit_with_pool(pool) \
+	do {\
+		dm_pool_destroy(pool->mem);\
+		dmeventd_lvm2_exit();\
+	} while(0)
+
 #endif /* _DMEVENTD_LVMWRAP_H */

daemons/dmeventd/plugins/mirror/dmeventd_mirror.c

@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2005-2012 Red Hat, Inc. All rights reserved.
+ * Copyright (C) 2005-2015 Red Hat, Inc. All rights reserved.
  *
  * This file is part of LVM2.
  *
@@ -13,20 +13,24 @@
  */
 
 #include "lib.h"
-
 #include "libdevmapper-event.h"
 #include "dmeventd_lvm.h"
 #include "defaults.h"
 
-#include <syslog.h> /* FIXME Replace syslog with multilog */
-/* FIXME Missing openlog? */
-/* FIXME Replace most syslogs with log_error() style messages and add complete context. */
 /* FIXME Reformat to 80 char lines. */
 
 #define ME_IGNORE    0
 #define ME_INSYNC    1
 #define ME_FAILURE   2
 
+struct dso_state {
+	struct dm_pool *mem;
+	char cmd_lvscan[512];
+	char cmd_lvconvert[512];
+};
+
+DM_EVENT_LOG_FN("mirr")
+
 static int _process_status_code(const char status_code, const char *dev_name,
 				const char *dev_type, int r)
 {
@@ -39,18 +43,15 @@ static int _process_status_code(const char status_code, const char *dev_name,
 	 *    U => Unclassified failure (bug)
 	 */ 
 	if (status_code == 'F') {
-		syslog(LOG_ERR, "%s device %s flush failed.",
-		       dev_type, dev_name);
+		log_error("%s device %s flush failed.", dev_type, dev_name);
 		r = ME_FAILURE;
 	} else if (status_code == 'S')
-		syslog(LOG_ERR, "%s device %s sync failed.",
-		       dev_type, dev_name);
+		log_error("%s device %s sync failed.", dev_type, dev_name);
 	else if (status_code == 'R')
-		syslog(LOG_ERR, "%s device %s read failed.",
-		       dev_type, dev_name);
+		log_error("%s device %s read failed.", dev_type, dev_name);
 	else if (status_code != 'A') {
-		syslog(LOG_ERR, "%s device %s has failed (%c).",
-		       dev_type, dev_name, status_code);
+		log_error("%s device %s has failed (%c).",
+			  dev_type, dev_name, status_code);
 		r = ME_FAILURE;
 	}
 
@@ -125,62 +126,49 @@ out:
 
 out_parse:
 	dm_free(args);
-	syslog(LOG_ERR, "Unable to parse mirror status string.");
+	log_error("Unable to parse mirror status string.");
+
 	return ME_IGNORE;
 }
 
-static int _remove_failed_devices(const char *device)
+static int _remove_failed_devices(const char *cmd_lvscan, const char *cmd_lvconvert)
 {
 	int r;
-#define CMD_SIZE 256	/* FIXME Use system restriction */
-	char cmd_str[CMD_SIZE];
 
-	if (!dmeventd_lvm2_command(dmeventd_lvm2_pool(), cmd_str, sizeof(cmd_str),
-				   "lvscan --cache", device))
-		return -1;
-
-	r = dmeventd_lvm2_run(cmd_str);
-
-	if (!r)
-		syslog(LOG_INFO, "Re-scan of mirror device %s failed.", device);
-
-	if (!dmeventd_lvm2_command(dmeventd_lvm2_pool(), cmd_str, sizeof(cmd_str),
-				  "lvconvert --config devices{ignore_suspended_devices=1} "
-				  "--repair --use-policies", device))
-		return -ENAMETOOLONG; /* FIXME Replace with generic error return - reason for failure has already got logged */
+	if (!dmeventd_lvm2_run_with_lock(cmd_lvscan))
+		log_info("Re-scan of mirrored device failed.");
 
 	/* if repair goes OK, report success even if lvscan has failed */
-	r = dmeventd_lvm2_run(cmd_str);
+	r = dmeventd_lvm2_run_with_lock(cmd_lvconvert);
 
-	syslog(LOG_INFO, "Repair of mirrored device %s %s.", device,
-	       (r) ? "finished successfully" : "failed");
+	log_info("Repair of mirrored device %s.",
+		 (r) ? "finished successfully" : "failed");
 
-	return (r) ? 0 : -1;
+	return r;
 }
 
 void process_event(struct dm_task *dmt,
 		   enum dm_event_mask event __attribute__((unused)),
-		   void **unused __attribute__((unused)))
+		   void **user)
 {
+	struct dso_state *state = *user;
 	void *next = NULL;
 	uint64_t start, length;
 	char *target_type = NULL;
 	char *params;
 	const char *device = dm_task_get_name(dmt);
 
-	dmeventd_lvm2_lock();
-
 	do {
 		next = dm_get_next_target(dmt, next, &start, &length,
 					  &target_type, &params);
 
 		if (!target_type) {
-			syslog(LOG_INFO, "%s mapping lost.", device);
+			log_info("%s mapping lost.", device);
 			continue;
 		}
 
 		if (strcmp(target_type, "mirror")) {
-			syslog(LOG_INFO, "%s has unmirrored portion.", device);
+			log_info("%s has unmirrored portion.", device);
 			continue;
 		}
 
@@ -190,54 +178,75 @@ void process_event(struct dm_task *dmt,
 			   _part_ of the device is in sync
 			   Also, this is not an error
 			*/
-			syslog(LOG_NOTICE, "%s is now in-sync.", device);
+			log_notice("%s is now in-sync.", device);
 			break;
 		case ME_FAILURE:
-			syslog(LOG_ERR, "Device failure in %s.", device);
-			if (_remove_failed_devices(device))
+			log_error("Device failure in %s.", device);
+			if (!_remove_failed_devices(state->cmd_lvscan,
+						    state->cmd_lvconvert))
 				/* FIXME Why are all the error return codes unused? Get rid of them? */
-				syslog(LOG_ERR, "Failed to remove faulty devices in %s.",
-				       device);
+				log_error("Failed to remove faulty devices in %s.",
+					  device);
 			/* Should check before warning user that device is now linear
 			else
-				syslog(LOG_NOTICE, "%s is now a linear device.\n",
-					device);
+				log_notice("%s is now a linear device.",
+					   device);
 			*/
 			break;
 		case ME_IGNORE:
 			break;
 		default:
 			/* FIXME Provide value then! */
-			syslog(LOG_INFO, "Unknown event received.");
+			log_info("Unknown event received.");
 		}
 	} while (next);
-
-	dmeventd_lvm2_unlock();
 }
 
 int register_device(const char *device,
 		    const char *uuid __attribute__((unused)),
 		    int major __attribute__((unused)),
 		    int minor __attribute__((unused)),
-		    void **unused __attribute__((unused)))
+		    void **user)
 {
-	if (!dmeventd_lvm2_init())
-		return 0;
+	struct dso_state *state;
 
-	syslog(LOG_INFO, "Monitoring mirror device %s for events.", device);
+	if (!dmeventd_lvm2_init_with_pool("mirror_state", state))
+		goto_bad;
+
+	if (!dmeventd_lvm2_command(state->mem, state->cmd_lvscan, sizeof(state->cmd_lvscan),
+				   "lvscan --cache", device)) {
+		dmeventd_lvm2_exit_with_pool(state);
+		goto_bad;
+	}
+
+	if (!dmeventd_lvm2_command(state->mem, state->cmd_lvconvert, sizeof(state->cmd_lvconvert),
+				   "lvconvert --repair --use-policies", device)) {
+		dmeventd_lvm2_exit_with_pool(state);
+		goto_bad;
+	}
+
+	*user = state;
+
+	log_info("Monitoring mirror device %s for events.", device);
 
 	return 1;
+bad:
+	log_error("Failed to monitor mirror %s.", device);
+
+	return 0;
 }
 
 int unregister_device(const char *device,
 		      const char *uuid __attribute__((unused)),
 		      int major __attribute__((unused)),
 		      int minor __attribute__((unused)),
-		      void **unused __attribute__((unused)))
+		      void **user)
 {
-	syslog(LOG_INFO, "No longer monitoring mirror device %s for events.",
-	       device);
-	dmeventd_lvm2_exit();
+	struct dso_state *state = *user;
+
+	dmeventd_lvm2_exit_with_pool(state);
+	log_info("No longer monitoring mirror device %s for events.",
+		 device);
 
 	return 1;
 }

daemons/dmeventd/plugins/raid/dmeventd_raid.c

@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2005-2011 Red Hat, Inc. All rights reserved.
+ * Copyright (C) 2005-2015 Red Hat, Inc. All rights reserved.
  *
  * This file is part of LVM2.
  *
@@ -13,168 +13,133 @@
  */
 
 #include "lib.h"
-
-#include "libdevmapper-event.h"
 #include "dmeventd_lvm.h"
+#include "libdevmapper-event.h"
+
+struct dso_state {
+	struct dm_pool *mem;
+	char cmd_lvscan[512];
+	char cmd_lvconvert[512];
+	int failed;
+};
+
+DM_EVENT_LOG_FN("raid")
 
-#include <syslog.h> /* FIXME Replace syslog with multilog */
-/* FIXME Missing openlog? */
-/* FIXME Replace most syslogs with log_error() style messages and add complete context. */
 /* FIXME Reformat to 80 char lines. */
 
-/*
- * run_repair is a close copy to
- * plugins/mirror/dmeventd_mirror.c:_remove_failed_devices()
- */
-static int run_repair(const char *device)
+static int _process_raid_event(struct dso_state *state, char *params, const char *device)
 {
-	int r;
-#define CMD_SIZE 256	/* FIXME Use system restriction */
-	char cmd_str[CMD_SIZE];
+	struct dm_status_raid *status;
+	const char *d;
 
-	if (!dmeventd_lvm2_command(dmeventd_lvm2_pool(), cmd_str, sizeof(cmd_str),
-				  "lvscan --cache", device))
-		return -1;
-
-	r = dmeventd_lvm2_run(cmd_str);
-
-	if (!r)
-		syslog(LOG_INFO, "Re-scan of RAID device %s failed.", device);
-
-	if (!dmeventd_lvm2_command(dmeventd_lvm2_pool(), cmd_str, sizeof(cmd_str),
-				  "lvconvert --config devices{ignore_suspended_devices=1} "
-				  "--repair --use-policies", device))
-		return -1;
-
-	/* if repair goes OK, report success even if lvscan has failed */
-	r = dmeventd_lvm2_run(cmd_str);
-
-	if (!r)
-		syslog(LOG_INFO, "Repair of RAID device %s failed.", device);
-
-	return (r) ? 0 : -1;
-}
-
-static int _process_raid_event(char *params, const char *device)
-{
-	int i, n, failure = 0;
-	char *p, *a[4];
-	char *raid_type;
-	char *num_devices;
-	char *health_chars;
-	char *resync_ratio;
-
-	/*
-	 * RAID parms:     <raid_type> <#raid_disks> \
-	 *                 <health chars> <resync ratio>
-	 */
-	if (!dm_split_words(params, 4, 0, a)) {
-		syslog(LOG_ERR, "Failed to process status line for %s\n",
-		       device);
-		return -EINVAL;
-	}
-	raid_type = a[0];
-	num_devices = a[1];
-	health_chars = a[2];
-	resync_ratio = a[3];
-
-	if (!(n = atoi(num_devices))) {
-		syslog(LOG_ERR, "Failed to parse number of devices for %s: %s",
-		       device, num_devices);
-		return -EINVAL;
+	if (!dm_get_status_raid(state->mem, params, &status)) {
+		log_error("Failed to process status line for %s.", device);
+		return 0;
 	}
 
-	for (i = 0; i < n; i++) {
-		switch (health_chars[i]) {
-		case 'A':
-			/* Device is 'A'live and well */
-		case 'a':
-			/* Device is 'a'live, but not yet in-sync */
-			break;
-		case 'D':
-			syslog(LOG_ERR,
-			       "Device #%d of %s array, %s, has failed.",
-			       i, raid_type, device);
-			failure++;
-			break;
-		default:
-			/* Unhandled character returned from kernel */
-			break;
+	if ((d = strchr(status->dev_health, 'D'))) {
+		if (state->failed)
+			goto out; /* already reported */
+
+		log_error("Device #%d of %s array, %s, has failed.",
+			  (int)(d - status->dev_health),
+			  status->raid_type, device);
+
+		state->failed = 1;
+		if (!dmeventd_lvm2_run_with_lock(state->cmd_lvscan))
+			log_warn("WARNING: Re-scan of RAID device %s failed.", device);
+
+		/* if repair goes OK, report success even if lvscan has failed */
+		if (!dmeventd_lvm2_run_with_lock(state->cmd_lvconvert)) {
+			log_info("Repair of RAID device %s failed.", device);
+			dm_pool_free(state->mem, status);
+			return 0;
 		}
-		if (failure)
-			return run_repair(device);
+	} else {
+		state->failed = 0;
+		log_info("%s array, %s, is %s in-sync.",
+			 status->raid_type, device,
+			 (status->insync_regions == status->total_regions) ? "now" : "not");
 	}
+out:
+	dm_pool_free(state->mem, status);
 
-	p = strstr(resync_ratio, "/");
-	if (!p) {
-		syslog(LOG_ERR, "Failed to parse resync_ratio for %s: %s",
-		       device, resync_ratio);
-		return -EINVAL;
-	}
-	p[0] = '\0';
-	syslog(LOG_INFO, "%s array, %s, is %s in-sync.",
-	       raid_type, device, strcmp(resync_ratio, p+1) ? "not" : "now");
-
-	return 0;
+	return 1;
 }
 
 void process_event(struct dm_task *dmt,
 		   enum dm_event_mask event __attribute__((unused)),
-		   void **unused __attribute__((unused)))
+		   void **user)
 {
+	struct dso_state *state = *user;
 	void *next = NULL;
 	uint64_t start, length;
 	char *target_type = NULL;
 	char *params;
 	const char *device = dm_task_get_name(dmt);
 
-	dmeventd_lvm2_lock();
-
 	do {
 		next = dm_get_next_target(dmt, next, &start, &length,
 					  &target_type, &params);
 
 		if (!target_type) {
-			syslog(LOG_INFO, "%s mapping lost.", device);
+			log_info("%s mapping lost.", device);
 			continue;
 		}
 
 		if (strcmp(target_type, "raid")) {
-			syslog(LOG_INFO, "%s has non-raid portion.", device);
+			log_info("%s has non-raid portion.", device);
 			continue;
 		}
 
-		if (_process_raid_event(params, device))
-			syslog(LOG_ERR, "Failed to process event for %s",
-			       device);
+		if (!_process_raid_event(state, params, device))
+			log_error("Failed to process event for %s.",
+				  device);
 	} while (next);
-
-	dmeventd_lvm2_unlock();
 }
 
 int register_device(const char *device,
 		    const char *uuid __attribute__((unused)),
 		    int major __attribute__((unused)),
 		    int minor __attribute__((unused)),
-		    void **unused __attribute__((unused)))
+		    void **user)
 {
-	if (!dmeventd_lvm2_init())
-		return 0;
+	struct dso_state *state;
 
-	syslog(LOG_INFO, "Monitoring RAID device %s for events.", device);
+	if (!dmeventd_lvm2_init_with_pool("raid_state", state))
+		goto_bad;
+
+	if (!dmeventd_lvm2_command(state->mem, state->cmd_lvscan, sizeof(state->cmd_lvscan),
+				   "lvscan --cache", device) ||
+	    !dmeventd_lvm2_command(state->mem, state->cmd_lvconvert, sizeof(state->cmd_lvconvert),
+				   "lvconvert --config devices{ignore_suspended_devices=1} "
+				   "--repair --use-policies", device)) {
+		dmeventd_lvm2_exit_with_pool(state);
+		goto_bad;
+	}
+
+	*user = state;
+
+	log_info("Monitoring RAID device %s for events.", device);
 
 	return 1;
+bad:
+	log_error("Failed to monitor RAID %s.", device);
+
+	return 0;
 }
 
 int unregister_device(const char *device,
 		      const char *uuid __attribute__((unused)),
 		      int major __attribute__((unused)),
 		      int minor __attribute__((unused)),
-		      void **unused __attribute__((unused)))
+		      void **user)
 {
-	syslog(LOG_INFO, "No longer monitoring RAID device %s for events.",
-	       device);
-	dmeventd_lvm2_exit();
+	struct dso_state *state = *user;
+
+	dmeventd_lvm2_exit_with_pool(state);
+	log_info("No longer monitoring RAID device %s for events.",
+		 device);
 
 	return 1;
 }

daemons/dmeventd/plugins/snapshot/dmeventd_snapshot.c

@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2007-2011 Red Hat, Inc. All rights reserved.
+ * Copyright (C) 2007-2015 Red Hat, Inc. All rights reserved.
  *
  * This file is part of LVM2.
  *
@@ -13,31 +13,31 @@
  */
 
 #include "lib.h"
-
-#include "libdevmapper-event.h"
 #include "dmeventd_lvm.h"
+#include "libdevmapper-event.h"
 
 #include <sys/wait.h>
-#include <syslog.h> /* FIXME Replace syslog with multilog */
 #include <stdarg.h>
-/* FIXME Missing openlog? */
+#include <pthread.h>
 
 /* First warning when snapshot is 80% full. */
-#define WARNING_THRESH 80
+#define WARNING_THRESH	(DM_PERCENT_1 * 80)
 /* Run a check every 5%. */
-#define CHECK_STEP 5
+#define CHECK_STEP	(DM_PERCENT_1 *  5)
 /* Do not bother checking snapshots less than 50% full. */
-#define CHECK_MINIMUM 50
+#define CHECK_MINIMUM	(DM_PERCENT_1 * 50)
 
 #define UMOUNT_COMMAND "/bin/umount"
 
 struct dso_state {
 	struct dm_pool *mem;
-	int percent_check;
+	dm_percent_t percent_check;
 	uint64_t known_size;
-	char cmd_str[1024];
+	char cmd_lvextend[512];
 };
 
+DM_EVENT_LOG_FN("snap")
+
 static int _run(const char *cmd, ...)
 {
         va_list ap;
@@ -62,7 +62,7 @@ static int _run(const char *cmd, ...)
                 va_end(ap);
 
                 execvp(cmd, (char **)argv);
-                syslog(LOG_ERR, "Failed to execute %s: %s.\n", cmd, strerror(errno));
+                log_sys_error("exec", cmd);
                 exit(127);
         }
 
@@ -81,18 +81,56 @@ static int _run(const char *cmd, ...)
 
 static int _extend(const char *cmd)
 {
-	return dmeventd_lvm2_run(cmd);
+	log_debug("Extending snapshot via %s.", cmd);
+	return dmeventd_lvm2_run_with_lock(cmd);
 }
 
+#ifdef SNAPSHOT_REMOVE
+/* Remove invalid snapshot from dm-table */
+/* Experimental for now and not used by default */
+static int _remove(const char *uuid)
+{
+	int r = 1;
+	uint32_t cookie = 0;
+	struct dm_task *dmt;
+
+	if (!(dmt = dm_task_create(DM_DEVICE_REMOVE)))
+		return 0;
+
+	if (!dm_task_set_uuid(dmt, uuid)) {
+		r = 0;
+		goto_out;
+	}
+
+	dm_task_retry_remove(dmt);
+
+	if (!dm_task_set_cookie(dmt, &cookie, 0)) {
+		r = 0;
+		goto_out;
+	}
+
+	if (!dm_task_run(dmt)) {
+		r = 0;
+		goto_out;
+	}
+out:
+	dm_task_destroy(dmt);
+
+	return r;
+}
+#endif /* SNAPSHOT_REMOVE */
+
 static void _umount(const char *device, int major, int minor)
 {
 	FILE *mounts;
 	char buffer[4096];
 	char *words[3];
 	struct stat st;
+	const char procmounts[] = "/proc/mounts";
 
-	if (!(mounts = fopen("/proc/mounts", "r"))) {
-		syslog(LOG_ERR, "Could not read /proc/mounts. Not umounting %s.\n", device);
+	if (!(mounts = fopen(procmounts, "r"))) {
+		log_sys_error("fopen", procmounts);
+		log_error("Not umounting %s.", device);
 		return;
 	}
 
@@ -112,21 +150,22 @@ static void _umount(const char *device, int major, int minor)
 		if (S_ISBLK(st.st_mode) &&
 		    major(st.st_rdev) == major &&
 		    minor(st.st_rdev) == minor) {
-			syslog(LOG_ERR, "Unmounting invalid snapshot %s from %s.\n", device, words[1]);
-                        if (!_run(UMOUNT_COMMAND, "-fl", words[1], NULL))
-                                syslog(LOG_ERR, "Failed to umount snapshot %s from %s: %s.\n",
-                                       device, words[1], strerror(errno));
+			log_error("Unmounting invalid snapshot %s from %s.", device, words[1]);
+			if (!_run(UMOUNT_COMMAND, "-fl", words[1], NULL))
+				log_error("Failed to umount snapshot %s from %s: %s.",
+					  device, words[1], strerror(errno));
 		}
 	}
 
 	if (fclose(mounts))
-		syslog(LOG_ERR, "Failed to close /proc/mounts.\n");
+		log_sys_error("close", procmounts);
 }
 
 void process_event(struct dm_task *dmt,
 		   enum dm_event_mask event __attribute__((unused)),
-		   void **private)
+		   void **user)
 {
+	struct dso_state *state = *user;
 	void *next = NULL;
 	uint64_t start, length;
 	char *target_type = NULL;
@@ -134,28 +173,47 @@ void process_event(struct dm_task *dmt,
 	struct dm_status_snapshot *status = NULL;
 	const char *device = dm_task_get_name(dmt);
 	int percent;
-	struct dso_state *state = *private;
+	struct dm_info info;
 
 	/* No longer monitoring, waiting for remove */
 	if (!state->percent_check)
 		return;
 
-	dmeventd_lvm2_lock();
-
 	dm_get_next_target(dmt, next, &start, &length, &target_type, &params);
-	if (!target_type)
-		goto out;
+	if (!target_type || strcmp(target_type, "snapshot")) {
+		log_error("Target %s is not snapshot.", target_type);
+		return;
+	}
 
-	if (!dm_get_status_snapshot(state->mem, params, &status))
-		goto out;
+	if (!dm_get_status_snapshot(state->mem, params, &status)) {
+		log_error("Cannot parse snapshot %s state: %s.", device, params);
+		return;
+	}
 
-	if (status->invalid) {
-		struct dm_info info;
-		if (dm_task_get_info(dmt, &info)) {
-			dmeventd_lvm2_unlock();
+	/*
+	 * If the snapshot has been invalidated or we failed to parse
+	 * the status string. Report the full status string to syslog.
+	 */
+	if (status->invalid || status->overflow || !status->total_sectors) {
+		log_warn("WARNING: Snapshot %s changed state to: %s and should be removed.",
+			 device, params);
+		state->percent_check = 0;
+		if (dm_task_get_info(dmt, &info))
 			_umount(device, info.major, info.minor);
-			return;
-		} /* else; too bad, but this is best-effort thing... */
+#ifdef SNAPSHOT_REMOVE
+		/* Maybe configurable ? */
+		_remove(dm_task_get_uuid(dmt));
+#endif
+		pthread_kill(pthread_self(), SIGALRM);
+		goto out;
+	}
+
+	if (length <= (status->used_sectors - status->metadata_sectors)) {
+		/* TODO eventually recognize earlier when room is enough */
+		log_info("Dropping monitoring of fully provisioned snapshot %s.",
+			 device);
+		pthread_kill(pthread_self(), SIGALRM);
+		goto out;
 	}
 
 	/* Snapshot size had changed. Clear the threshold. */
@@ -164,69 +222,50 @@ void process_event(struct dm_task *dmt,
 		state->known_size = status->total_sectors;
 	}
 
-	/*
-	 * If the snapshot has been invalidated or we failed to parse
-	 * the status string. Report the full status string to syslog.
-	 */
-	if (status->invalid || !status->total_sectors) {
-		syslog(LOG_ERR, "Snapshot %s changed state to: %s\n", device, params);
-		state->percent_check = 0;
-		goto out;
-	}
-
-	percent = (int) (100 * status->used_sectors / status->total_sectors);
+	percent = dm_make_percent(status->used_sectors, status->total_sectors);
 	if (percent >= state->percent_check) {
 		/* Usage has raised more than CHECK_STEP since the last
 		   time. Run actions. */
 		state->percent_check = (percent / CHECK_STEP) * CHECK_STEP + CHECK_STEP;
 
 		if (percent >= WARNING_THRESH) /* Print a warning to syslog. */
-			syslog(LOG_WARNING, "Snapshot %s is now %i%% full.\n", device, percent);
-		/* Try to extend the snapshot, in accord with user-set policies */
-		if (!_extend(state->cmd_str))
-			syslog(LOG_ERR, "Failed to extend snapshot %s.\n", device);
-	}
+			log_warn("WARNING: Snapshot %s is now %.2f%% full.",
+				 device, dm_percent_to_float(percent));
 
+		/* Try to extend the snapshot, in accord with user-set policies */
+		if (!_extend(state->cmd_lvextend))
+			log_error("Failed to extend snapshot %s.", device);
+	}
 out:
-	if (status)
-		dm_pool_free(state->mem, status);
-	dmeventd_lvm2_unlock();
+	dm_pool_free(state->mem, status);
 }
 
 int register_device(const char *device,
 		    const char *uuid __attribute__((unused)),
 		    int major __attribute__((unused)),
 		    int minor __attribute__((unused)),
-		    void **private)
+		    void **user)
 {
-	struct dm_pool *statemem = NULL;
 	struct dso_state *state;
 
-	if (!dmeventd_lvm2_init())
-		goto out;
+	if (!dmeventd_lvm2_init_with_pool("snapshot_state", state))
+		goto_bad;
 
-	if (!(statemem = dm_pool_create("snapshot_state", 512)) ||
-	    !(state = dm_pool_zalloc(statemem, sizeof(*state))))
-		goto bad;
+	if (!dmeventd_lvm2_command(state->mem, state->cmd_lvextend,
+				   sizeof(state->cmd_lvextend),
+				   "lvextend --use-policies", device)) {
+		dmeventd_lvm2_exit_with_pool(state);
+		goto_bad;
+	}
 
-	if (!dmeventd_lvm2_command(statemem, state->cmd_str,
-				   sizeof(state->cmd_str),
-				   "lvextend --use-policies", device))
-		goto bad;
-
-	state->mem = statemem;
 	state->percent_check = CHECK_MINIMUM;
-	*private = state;
+	*user = state;
 
-	syslog(LOG_INFO, "Monitoring snapshot %s\n", device);
+	log_info("Monitoring snapshot %s.", device);
 
 	return 1;
 bad:
-	if (statemem)
-		dm_pool_destroy(statemem);
-	dmeventd_lvm2_exit();
-out:
-	syslog(LOG_ERR, "Failed to monitor snapshot %s.\n", device);
+	log_error("Failed to monitor snapshot %s.", device);
 
 	return 0;
 }
@@ -235,13 +274,12 @@ int unregister_device(const char *device,
 		      const char *uuid __attribute__((unused)),
 		      int major __attribute__((unused)),
 		      int minor __attribute__((unused)),
-		      void **private)
+		      void **user)
 {
-	struct dso_state *state = *private;
+	struct dso_state *state = *user;
 
-	syslog(LOG_INFO, "No longer monitoring snapshot %s\n", device);
-	dm_pool_destroy(state->mem);
-	dmeventd_lvm2_exit();
+	dmeventd_lvm2_exit_with_pool(state);
+	log_info("No longer monitoring snapshot %s.", device);
 
 	return 1;
 }

daemons/dmeventd/plugins/thin/dmeventd_thin.c

@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2011-2013 Red Hat, Inc. All rights reserved.
+ * Copyright (C) 2011-2015 Red Hat, Inc. All rights reserved.
  *
  * This file is part of LVM2.
  *
@@ -12,25 +12,33 @@
  * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
-#include "lib.h"
-
-#include "libdevmapper-event.h"
+#include "lib.h"	/* using here lvm log */
 #include "dmeventd_lvm.h"
+#include "libdevmapper-event.h"
 
 #include <sys/wait.h>
-#include <syslog.h> /* FIXME Replace syslog with multilog */
 #include <stdarg.h>
-/* FIXME Missing openlog? */
+#include <pthread.h>
 
-/* First warning when thin is 80% full. */
-#define WARNING_THRESH 80
+/* TODO - move this mountinfo code into library to be reusable */
+#ifdef __linux__
+#  include "kdev_t.h"
+#else
+#  define MAJOR(x) major((x))
+#  define MINOR(x) minor((x))
+#endif
+
+/* First warning when thin data or metadata is 80% full. */
+#define WARNING_THRESH	(DM_PERCENT_1 * 80)
 /* Run a check every 5%. */
-#define CHECK_STEP 5
-/* Do not bother checking thins less than 50% full. */
-#define CHECK_MINIMUM 50
+#define CHECK_STEP	(DM_PERCENT_1 *  5)
+/* Do not bother checking thin data or metadata is less than 50% full. */
+#define CHECK_MINIMUM	(DM_PERCENT_1 * 50)
 
 #define UMOUNT_COMMAND "/bin/umount"
 
+#define MAX_FAILS	(10)
+
 #define THIN_DEBUG 0
 
 struct dso_state {
@@ -39,18 +47,11 @@ struct dso_state {
 	int data_percent_check;
 	uint64_t known_metadata_size;
 	uint64_t known_data_size;
+	unsigned fails;
 	char cmd_str[1024];
 };
 
-
-/* TODO - move this mountinfo code into library to be reusable */
-#ifdef __linux__
-#  include "kdev_t.h"
-#else
-#  define MAJOR(x) major((x))
-#  define MINOR(x) minor((x))
-#  define MKDEV(x,y) makedev((x),(y))
-#endif
+DM_EVENT_LOG_FN("thin")
 
 /* Get dependencies for device, and try to find matching device */
 static int _has_deps(const char *name, int tp_major, int tp_minor, int *dev_minor)
@@ -93,8 +94,8 @@ static int _has_deps(const char *name, int tp_major, int tp_minor, int *dev_mino
 	{
 		char dev_name[PATH_MAX];
 		if (dm_device_get_name(major, minor, 0, dev_name, sizeof(dev_name)))
-			syslog(LOG_DEBUG, "Found %s (%u:%u) depends on %s",
-			       name, major, *dev_minor, dev_name);
+			log_debug("Found %s (%u:%u) depends on %s.",
+				  name, major, *dev_minor, dev_name);
 	}
 #endif
 	r = 1;
@@ -141,14 +142,6 @@ out:
 	return r;
 }
 
-static int _extend(struct dso_state *state)
-{
-#if THIN_DEBUG
-	syslog(LOG_INFO, "dmeventd executes: %s.\n", state->cmd_str);
-#endif
-	return dmeventd_lvm2_run(state->cmd_str);
-}
-
 static int _run(const char *cmd, ...)
 {
 	va_list ap;
@@ -168,12 +161,12 @@ static int _run(const char *cmd, ...)
 		argv = alloca(sizeof(const char *) * (argc + 1));
 
 		argv[0] = cmd;
-                va_start(ap, cmd);
+		va_start(ap, cmd);
 		while ((argv[++i] = va_arg(ap, const char *)));
 		va_end(ap);
 
 		execvp(cmd, (char **)argv);
-		syslog(LOG_ERR, "Failed to execute %s: %s.\n", cmd, strerror(errno));
+		log_sys_error("exec", cmd);
 		exit(127);
 	}
 
@@ -191,9 +184,9 @@ static int _run(const char *cmd, ...)
 }
 
 struct mountinfo_s {
+	const char *device;
 	struct dm_info info;
 	dm_bitset_t minors; /* Bitset for active thin pool minors */
-	const char *device;
 };
 
 static int _umount_device(char *buffer, unsigned major, unsigned minor,
@@ -202,11 +195,11 @@ static int _umount_device(char *buffer, unsigned major, unsigned minor,
 	struct mountinfo_s *data = cb_data;
 
 	if ((major == data->info.major) && dm_bit(data->minors, minor)) {
-		syslog(LOG_INFO, "Unmounting thin volume %s from %s.\n",
-		       data->device, target);
+		log_info("Unmounting thin volume %s from %s.",
+			 data->device, target);
 		if (!_run(UMOUNT_COMMAND, "-fl", target, NULL))
-			syslog(LOG_ERR, "Failed to umount thin %s from %s: %s.\n",
-			       data->device, target, strerror(errno));
+			log_error("Failed to umount thin %s from %s: %s.",
+				  data->device, target, strerror(errno));
 	}
 
 	return 1;
@@ -216,78 +209,94 @@ static int _umount_device(char *buffer, unsigned major, unsigned minor,
  * Find all thin pool users and try to umount them.
  * TODO: work with read-only thin pool support
  */
-static void _umount(struct dm_task *dmt, const char *device)
+static void _umount(struct dm_task *dmt)
 {
 	/* TODO: Convert to use hash to reduce memory usage */
 	static const size_t MINORS = (1U << 20); /* 20 bit */
-	struct mountinfo_s data = {
-		.device = device,
-	};
+	struct mountinfo_s data = { NULL };
 
 	if (!dm_task_get_info(dmt, &data.info))
 		return;
 
-	dmeventd_lvm2_unlock();
+	data.device = dm_task_get_name(dmt);
 
 	if (!(data.minors = dm_bitset_create(NULL, MINORS))) {
-		syslog(LOG_ERR, "Failed to allocate bitset. Not unmounting %s.\n", device);
+		log_error("Failed to allocate bitset. Not unmounting %s.", data.device);
 		goto out;
 	}
 
 	if (!_find_all_devs(data.minors, data.info.major, data.info.minor)) {
-		syslog(LOG_ERR, "Failed to detect mounted volumes for %s.\n", device);
+		log_error("Failed to detect mounted volumes for %s.", data.device);
 		goto out;
 	}
 
 	if (!dm_mountinfo_read(_umount_device, &data)) {
-		syslog(LOG_ERR, "Could not parse mountinfo file.\n");
+		log_error("Could not parse mountinfo file.");
 		goto out;
 	}
 
 out:
 	if (data.minors)
 		dm_bitset_destroy(data.minors);
-	dmeventd_lvm2_lock();
+}
+
+static void _use_policy(struct dm_task *dmt, struct dso_state *state)
+{
+#if THIN_DEBUG
+	log_info("dmeventd executes: %s.", state->cmd_str);
+#endif
+	if (!dmeventd_lvm2_run_with_lock(state->cmd_str)) {
+		log_error("Failed to extend thin pool %s.",
+			  dm_task_get_name(dmt));
+		_umount(dmt);
+		state->fails++;
+	} else
+		state->fails = 0;
 }
 
 void process_event(struct dm_task *dmt,
 		   enum dm_event_mask event __attribute__((unused)),
-		   void **private)
+		   void **user)
 {
 	const char *device = dm_task_get_name(dmt);
 	int percent;
-	struct dso_state *state = *private;
+	struct dso_state *state = *user;
 	struct dm_status_thin_pool *tps = NULL;
 	void *next = NULL;
 	uint64_t start, length;
 	char *target_type = NULL;
 	char *params;
+	int needs_policy = 0;
 
 #if 0
 	/* No longer monitoring, waiting for remove */
 	if (!state->meta_percent_check && !state->data_percent_check)
 		return;
 #endif
-	dmeventd_lvm2_lock();
+	if (event & DM_EVENT_DEVICE_ERROR) {
+		/* Error -> no need to check and do instant resize */
+		_use_policy(dmt, state);
+		goto out;
+	}
 
 	dm_get_next_target(dmt, next, &start, &length, &target_type, &params);
 
 	if (!target_type || (strcmp(target_type, "thin-pool") != 0)) {
-		syslog(LOG_ERR, "Invalid target type.\n");
+		log_error("Invalid target type.");
 		goto out;
 	}
 
 	if (!dm_get_status_thin_pool(state->mem, params, &tps)) {
-		syslog(LOG_ERR, "Failed to parse status.\n");
-		_umount(dmt, device);
+		log_error("Failed to parse status.");
+		_umount(dmt);
 		goto out;
 	}
 
 #if THIN_DEBUG
-	syslog(LOG_INFO, "%p: Got status %" PRIu64 " / %" PRIu64
-	       " %" PRIu64  " / %" PRIu64 ".\n", state,
-	       tps->used_metadata_blocks, tps->total_metadata_blocks,
-	       tps->used_data_blocks, tps->total_data_blocks);
+	log_debug("Thin pool status " FMTu64 "/" FMTu64 "  "
+		  FMTu64 "/" FMTu64 ".",
+		  tps->used_metadata_blocks, tps->total_metadata_blocks,
+		  tps->used_data_blocks, tps->total_data_blocks);
 #endif
 
 	/* Thin pool size had changed. Clear the threshold. */
@@ -301,7 +310,7 @@ void process_event(struct dm_task *dmt,
 		state->known_data_size = tps->total_data_blocks;
 	}
 
-	percent = 100 * tps->used_metadata_blocks / tps->total_metadata_blocks;
+	percent = dm_make_percent(tps->used_metadata_blocks, tps->total_metadata_blocks);
 	if (percent >= state->metadata_percent_check) {
 		/*
 		 * Usage has raised more than CHECK_STEP since the last
@@ -311,18 +320,12 @@ void process_event(struct dm_task *dmt,
 
 		/* FIXME: extension of metadata needs to be written! */
 		if (percent >= WARNING_THRESH) /* Print a warning to syslog. */
-			syslog(LOG_WARNING, "Thin metadata %s is now %i%% full.\n",
-			       device, percent);
-		 /* Try to extend the metadata, in accord with user-set policies */
-		if (!_extend(state)) {
-			syslog(LOG_ERR, "Failed to extend thin metadata %s.\n",
-			       device);
-			_umount(dmt, device);
-		}
-		/* FIXME: hmm READ-ONLY switch should happen in error path */
+			log_warn("WARNING: Thin pool %s metadata is now %.2f%% full.",
+				 device, dm_percent_to_float(percent));
+		needs_policy = 1;
 	}
 
-	percent = 100 * tps->used_data_blocks / tps->total_data_blocks;
+	percent = dm_make_percent(tps->used_data_blocks, tps->total_data_blocks);
 	if (percent >= state->data_percent_check) {
 		/*
 		 * Usage has raised more than CHECK_STEP since
@@ -331,56 +334,53 @@ void process_event(struct dm_task *dmt,
 		state->data_percent_check = (percent / CHECK_STEP) * CHECK_STEP + CHECK_STEP;
 
 		if (percent >= WARNING_THRESH) /* Print a warning to syslog. */
-			syslog(LOG_WARNING, "Thin %s is now %i%% full.\n", device, percent);
-		/* Try to extend the thin data, in accord with user-set policies */
-		if (!_extend(state)) {
-			syslog(LOG_ERR, "Failed to extend thin %s.\n", device);
-			state->data_percent_check = 0;
-			_umount(dmt, device);
-		}
-		/* FIXME: hmm READ-ONLY switch should happen in error path */
+			log_warn("WARNING: Thin pool %s data is now %.2f%% full.",
+				 device, dm_percent_to_float(percent));
+		needs_policy = 1;
 	}
+
+	if (needs_policy)
+		_use_policy(dmt, state);
 out:
 	if (tps)
 		dm_pool_free(state->mem, tps);
 
-	dmeventd_lvm2_unlock();
+	if (state->fails >= MAX_FAILS) {
+		log_warn("WARNING: Dropping monitoring of %s. "
+			 "lvm2 command fails too often (%u times in raw).",
+			 device, state->fails);
+		pthread_kill(pthread_self(), SIGALRM);
+	}
 }
 
 int register_device(const char *device,
 		    const char *uuid __attribute__((unused)),
 		    int major __attribute__((unused)),
 		    int minor __attribute__((unused)),
-		    void **private)
+		    void **user)
 {
-	struct dm_pool *statemem = NULL;
 	struct dso_state *state;
 
-	if (!dmeventd_lvm2_init())
-		goto bad;
+	if (!dmeventd_lvm2_init_with_pool("thin_pool_state", state))
+		goto_bad;
 
-	if (!(statemem = dm_pool_create("thin_pool_state", 2048)) ||
-	    !(state = dm_pool_zalloc(statemem, sizeof(*state))) ||
-	    !dmeventd_lvm2_command(statemem, state->cmd_str,
+	if (!dmeventd_lvm2_command(state->mem, state->cmd_str,
 				   sizeof(state->cmd_str),
 				   "lvextend --use-policies",
 				   device)) {
-		if (statemem)
-			dm_pool_destroy(statemem);
-		dmeventd_lvm2_exit();
-		goto bad;
+		dmeventd_lvm2_exit_with_pool(state);
+		goto_bad;
 	}
 
-	state->mem = statemem;
 	state->metadata_percent_check = CHECK_MINIMUM;
 	state->data_percent_check = CHECK_MINIMUM;
-	*private = state;
+	*user = state;
 
-	syslog(LOG_INFO, "Monitoring thin %s.\n", device);
+	log_info("Monitoring thin %s.", device);
 
 	return 1;
 bad:
-	syslog(LOG_ERR, "Failed to monitor thin %s.\n", device);
+	log_error("Failed to monitor thin %s.", device);
 
 	return 0;
 }
@@ -389,13 +389,12 @@ int unregister_device(const char *device,
 		      const char *uuid __attribute__((unused)),
 		      int major __attribute__((unused)),
 		      int minor __attribute__((unused)),
-		      void **private)
+		      void **user)
 {
-	struct dso_state *state = *private;
+	struct dso_state *state = *user;
 
-	syslog(LOG_INFO, "No longer monitoring thin %s.\n", device);
-	dm_pool_destroy(state->mem);
-	dmeventd_lvm2_exit();
+	dmeventd_lvm2_exit_with_pool(state);
+	log_info("No longer monitoring thin %s.", device);
 
 	return 1;
 }

daemons/lvmetad/.gitignore

@@ -0,0 +1,2 @@
+lvmetad
+lvmetactl

daemons/lvmetad/Makefile.in

@@ -18,7 +18,7 @@ top_builddir = @top_builddir@
 SOURCES = lvmetad-core.c
 SOURCES2 = testclient.c
 
-TARGETS = lvmetad
+TARGETS = lvmetad lvmetactl
 
 .PHONY: install_lvmetad
 
@@ -41,6 +41,12 @@ lvmetad: $(OBJECTS) $(top_builddir)/libdaemon/client/libdaemonclient.a \
 		    $(top_builddir)/libdaemon/server/libdaemonserver.a
 	$(CC) $(CFLAGS) $(LDFLAGS) -o $@ $(OBJECTS) $(LVMLIBS) $(LIBS)
 
+lvmetactl: lvmetactl.o $(top_builddir)/libdaemon/client/libdaemonclient.a \
+	$(top_builddir)/libdaemon/server/libdaemonserver.a
+	$(CC) $(CFLAGS) $(LDFLAGS) -o $@ lvmetactl.o $(LVMLIBS)
+
+CLEAN_TARGETS += lvmetactl.o
+
 # TODO: No idea. No idea how to test either.
 #ifneq ("$(CFLOW_CMD)", "")
 #CFLOW_SOURCES = $(addprefix $(srcdir)/, $(SOURCES))

daemons/lvmetad/lvmetactl.c

@@ -0,0 +1,208 @@
+/*
+ * Copyright (C) 2014 Red Hat, Inc.
+ *
+ * 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.
+ */
+
+#include "tool.h"
+
+#include "lvmetad-client.h"
+
+daemon_handle h;
+
+static void print_reply(daemon_reply reply)
+{
+	const char *a = daemon_reply_str(reply, "response", NULL);
+	const char *b = daemon_reply_str(reply, "status", NULL);
+	const char *c = daemon_reply_str(reply, "reason", NULL);
+
+	printf("response \"%s\" status \"%s\" reason \"%s\"\n",
+	       a ? a : "", b ? b : "", c ? c : "");
+}
+
+int main(int argc, char **argv)
+{
+	daemon_reply reply;
+	char *cmd;
+	char *uuid;
+	char *name;
+	int val;
+	int ver;
+
+	if (argc < 2) {
+		printf("lvmetactl dump\n");
+		printf("lvmetactl pv_list\n");
+		printf("lvmetactl vg_list\n");
+		printf("lvmetactl vg_lookup_name <name>\n");
+		printf("lvmetactl vg_lookup_uuid <uuid>\n");
+		printf("lvmetactl pv_lookup_uuid <uuid>\n");
+		printf("lvmetactl set_global_invalid 0|1\n");
+		printf("lvmetactl get_global_invalid\n");
+		printf("lvmetactl set_vg_version <uuid> <name> <version>\n");
+		printf("lvmetactl vg_lock_type <uuid>\n");
+		return -1;
+	}
+
+	cmd = argv[1];
+
+	h = lvmetad_open(NULL);
+
+	if (!strcmp(cmd, "dump")) {
+		reply = daemon_send_simple(h, "dump",
+					   "token = %s", "skip",
+					   NULL);
+		printf("%s\n", reply.buffer.mem);
+
+	} else if (!strcmp(cmd, "pv_list")) {
+		reply = daemon_send_simple(h, "pv_list",
+					   "token = %s", "skip",
+					   NULL);
+		printf("%s\n", reply.buffer.mem);
+
+	} else if (!strcmp(cmd, "vg_list")) {
+		reply = daemon_send_simple(h, "vg_list",
+					   "token = %s", "skip",
+					   NULL);
+		printf("%s\n", reply.buffer.mem);
+
+	} else if (!strcmp(cmd, "set_global_invalid")) {
+		if (argc < 3) {
+			printf("set_global_invalid 0|1\n");
+			return -1;
+		}
+		val = atoi(argv[2]);
+
+		reply = daemon_send_simple(h, "set_global_info",
+					   "global_invalid = %d", val,
+					   "token = %s", "skip",
+					   NULL);
+		print_reply(reply);
+
+	} else if (!strcmp(cmd, "get_global_invalid")) {
+		reply = daemon_send_simple(h, "get_global_info",
+					   "token = %s", "skip",
+					   NULL);
+		printf("%s\n", reply.buffer.mem);
+
+	} else if (!strcmp(cmd, "set_vg_version")) {
+		if (argc < 5) {
+			printf("set_vg_version <uuid> <name> <ver>\n");
+			return -1;
+		}
+		uuid = argv[2];
+		name = argv[3];
+		ver = atoi(argv[4]);
+
+		if ((strlen(uuid) == 1) && (uuid[0] == '-'))
+			uuid = NULL;
+		if ((strlen(name) == 1) && (name[0] == '-'))
+			name = NULL;
+
+		if (uuid && name) {
+			reply = daemon_send_simple(h, "set_vg_info",
+						   "uuid = %s", uuid,
+						   "name = %s", name,
+						   "version = %d", ver,
+						    "token = %s", "skip",
+						    NULL);
+		} else if (uuid) {
+			reply = daemon_send_simple(h, "set_vg_info",
+						   "uuid = %s", uuid,
+						   "version = %d", ver,
+						    "token = %s", "skip",
+						    NULL);
+		} else if (name) {
+			reply = daemon_send_simple(h, "set_vg_info",
+						   "name = %s", name,
+						   "version = %d", ver,
+						    "token = %s", "skip",
+						    NULL);
+		} else {
+			printf("name or uuid required\n");
+			return -1;
+		}
+
+		print_reply(reply);
+
+	} else if (!strcmp(cmd, "vg_lookup_name")) {
+		if (argc < 3) {
+			printf("vg_lookup_name <name>\n");
+			return -1;
+		}
+		name = argv[2];
+
+		reply = daemon_send_simple(h, "vg_lookup",
+					   "name = %s", name,
+					   "token = %s", "skip",
+					   NULL);
+		printf("%s\n", reply.buffer.mem);
+
+	} else if (!strcmp(cmd, "vg_lookup_uuid")) {
+		if (argc < 3) {
+			printf("vg_lookup_uuid <uuid>\n");
+			return -1;
+		}
+		uuid = argv[2];
+
+		reply = daemon_send_simple(h, "vg_lookup",
+					   "uuid = %s", uuid,
+					   "token = %s", "skip",
+					   NULL);
+		printf("%s\n", reply.buffer.mem);
+
+	} else if (!strcmp(cmd, "vg_lock_type")) {
+		struct dm_config_node *metadata;
+		const char *lock_type;
+
+		if (argc < 3) {
+			printf("vg_lock_type <uuid>\n");
+			return -1;
+		}
+		uuid = argv[2];
+
+		reply = daemon_send_simple(h, "vg_lookup",
+					   "uuid = %s", uuid,
+					   "token = %s", "skip",
+					   NULL);
+		/* printf("%s\n", reply.buffer.mem); */
+
+		metadata = dm_config_find_node(reply.cft->root, "metadata");
+		if (!metadata) {
+			printf("no metadata\n");
+			goto out;
+		}
+
+		lock_type = dm_config_find_str(metadata, "metadata/lock_type", NULL);
+		if (!lock_type) {
+			printf("no lock_type\n");
+			goto out;
+		}
+		printf("lock_type %s\n", lock_type);
+
+	} else if (!strcmp(cmd, "pv_lookup_uuid")) {
+		if (argc < 3) {
+			printf("pv_lookup_uuid <uuid>\n");
+			return -1;
+		}
+		uuid = argv[2];
+
+		reply = daemon_send_simple(h, "pv_lookup",
+					   "uuid = %s", uuid,
+					   "token = %s", "skip",
+					   NULL);
+		printf("%s\n", reply.buffer.mem);
+
+	} else {
+		printf("unknown command\n");
+		goto out_close;
+	}
+out:
+	daemon_reply_destroy(reply);
+out_close:
+	daemon_close(h);
+	return 0;
+}

daemons/lvmetad/lvmetad-core.c

@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2012 Red Hat, Inc.
+ * Copyright (C) 2012-2015 Red Hat, Inc.
  *
  * This file is part of LVM2.
  *
@@ -14,17 +14,18 @@
 
 #define _XOPEN_SOURCE 500  /* pthread */
 
-#include "configure.h"
+#define _REENTRANT
+
+#include "tool.h"
+
 #include "daemon-io.h"
-#include "config-util.h"
 #include "daemon-server.h"
 #include "daemon-log.h"
 #include "lvm-version.h"
 
 #include <assert.h>
+#include <errno.h>
 #include <pthread.h>
-#include <stdint.h>
-#include <unistd.h>
 
 #define LVMETAD_SOCKET DEFAULT_RUN_DIR "/lvmetad.socket"
 
@@ -123,6 +124,7 @@ struct vg_info {
 #define VGFL_INVALID 0x00000001
 
 typedef struct {
+	daemon_idle *idle;
 	log_state *log; /* convenience */
 	const char *log_config;
 
@@ -578,7 +580,7 @@ static void mark_outdated_pv(lvmetad_state *s, const char *vgid, const char *pvi
 		    !(cft_vgid = make_text_node(outdated_pvs, "vgid", dm_pool_strdup(outdated_pvs->mem, vgid),
 						outdated_pvs->root, NULL)))
 			abort();
-		if(!dm_hash_insert(s->vgid_to_outdated_pvs, cft_vgid->v->v.str, outdated_pvs))
+		if (!dm_hash_insert(s->vgid_to_outdated_pvs, cft_vgid->v->v.str, outdated_pvs))
 			abort();
 		DEBUGLOG(s, "created outdated_pvs list for VG %s", vgid);
 	}
@@ -999,7 +1001,8 @@ static response pv_gone(lvmetad_state *s, request r)
 
 	DEBUGLOG(s, "pv_gone (updated): %s / %" PRIu64, pvid, device);
 
-	pvmeta = dm_hash_lookup(s->pvid_to_pvmeta, pvid);
+	if (!(pvmeta = dm_hash_lookup(s->pvid_to_pvmeta, pvid)))
+		return reply_unknown("PVID does not exist");
 	vgid = dm_hash_lookup(s->pvid_to_vgid, pvid);
 
 	dm_hash_remove_binary(s->device_to_pvid, &device, sizeof(device));
@@ -1021,9 +1024,6 @@ static response pv_gone(lvmetad_state *s, request r)
 		dm_free(vgid);
 	}
 
-	if (!pvmeta)
-		return reply_unknown("PVID does not exist");
-
 	if (!alt_device)
 		dm_config_destroy(pvmeta);
 
@@ -1312,20 +1312,29 @@ static response set_vg_info(lvmetad_state *s, request r)
 {
 	struct dm_config_tree *vg;
 	struct vg_info *info;
-	const char *uuid = daemon_request_str(r, "uuid", NULL);
+	const char *name;
+	const char *uuid;
 	const int64_t new_version = daemon_request_int(r, "version", -1);
 	int64_t cache_version;
 
-	if (!uuid)
-		goto out;
-
 	if (new_version == -1)
 		goto out;
 
-	vg = dm_hash_lookup(s->vgid_to_metadata, uuid);
-	if (!vg)
+	if (!(uuid = daemon_request_str(r, "uuid", NULL)))
+		goto use_name;
+
+	if ((vg = dm_hash_lookup(s->vgid_to_metadata, uuid)))
+		goto vers;
+use_name:
+	if (!(name = daemon_request_str(r, "name", NULL)))
 		goto out;
 
+	if (!(uuid = dm_hash_lookup(s->vgname_to_vgid, name)))
+		goto out;
+
+	if (!(vg = dm_hash_lookup(s->vgid_to_metadata, uuid)))
+		goto out;
+vers:
 	if (!new_version)
 		goto inval;
 
@@ -1340,7 +1349,8 @@ inval:
 		if (!info)
 			goto bad;
 		memset(info, 0, sizeof(struct vg_info));
-		dm_hash_insert(s->vgid_to_info, uuid, (void*)info);
+		if (!dm_hash_insert(s->vgid_to_info, uuid, (void*)info))
+			goto bad;
 	}
 
 	info->external_version = new_version;
@@ -1584,6 +1594,9 @@ static int init(daemon_state *s)
 	/* if (ls->initial_registrations)
 	   _process_initial_registrations(ds->initial_registrations); */
 
+	if (ls->idle)
+		ls->idle->is_idle = 1;
+
 	return 1;
 }
 
@@ -1606,21 +1619,39 @@ static int fini(daemon_state *s)
 	return 1;
 }
 
+static int process_timeout_arg(const char *str, unsigned *max_timeouts)
+{
+	char *endptr;
+	unsigned long l;
+
+	errno = 0;
+	l = strtoul(str, &endptr, 10);
+	if (errno || *endptr || l >= UINT_MAX)
+		return 0;
+
+	*max_timeouts = (unsigned) l;
+
+	return 1;
+}
+
 static void usage(const char *prog, FILE *file)
 {
 	fprintf(file, "Usage:\n"
-		"%s [-V] [-h] [-f] [-l {all|wire|debug}] [-s path]\n\n"
+		"%s [-V] [-h] [-f] [-l level[,level ...]] [-s path] [-t secs]\n\n"
 		"   -V       Show version of lvmetad\n"
 		"   -h       Show this help information\n"
 		"   -f       Don't fork, run in the foreground\n"
-		"   -l       Logging message level (-l {all|wire|debug})\n"
+		"   -l       Logging message levels (all,fatal,error,warn,info,wire,debug)\n"
 		"   -p       Set path to the pidfile\n"
-		"   -s       Set path to the socket to listen on\n\n", prog);
+		"   -s       Set path to the socket to listen on\n"
+		"   -t       Time to wait in seconds before shutdown on idle (missing or 0 = inifinite)\n\n", prog);
 }
 
 int main(int argc, char *argv[])
 {
 	signed char opt;
+	struct timeval timeout;
+	daemon_idle di = { .ptimeout = &timeout };
 	lvmetad_state ls = { .log_config = "" };
 	daemon_state s = {
 		.daemon_fini = fini,
@@ -1635,7 +1666,7 @@ int main(int argc, char *argv[])
 	};
 
 	// use getopt_long
-	while ((opt = getopt(argc, argv, "?fhVl:p:s:")) != EOF) {
+	while ((opt = getopt(argc, argv, "?fhVl:p:s:t:")) != EOF) {
 		switch (opt) {
 		case 'h':
 			usage(argv[0], stdout);
@@ -1655,6 +1686,15 @@ int main(int argc, char *argv[])
 		case 's': // --socket
 			s.socket_path = optarg;
 			break;
+		case 't':
+			if (!process_timeout_arg(optarg, &di.max_timeouts)) {
+				fprintf(stderr, "Invalid value of timeout parameter.\n");
+				exit(EXIT_FAILURE);
+			}
+			/* 0 equals to wait indefinitely */
+			if (di.max_timeouts)
+				s.idle = ls.idle = &di;
+			break;
 		case 'V':
 			printf("lvmetad version: " LVM_VERSION "\n");
 			exit(1);

daemons/lvmetad/testclient.c

@@ -1,3 +1,18 @@
+/*
+ * Copyright (C) 2011-2014 Red Hat, Inc.
+ *
+ * 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
+ */
+#include "tool.h"
+
 #include "lvmetad-client.h"
 #include "label.h"
 #include "lvmcache.h"
@@ -109,7 +124,7 @@ int main(int argc, char **argv) {
 
 	if (argc > 1) {
 		int i;
-		struct cmd_context *cmd = create_toolcontext(0, NULL, 0, 0);
+		struct cmd_context *cmd = create_toolcontext(0, NULL, 0, 0, 1, 1);
 		for (i = 1; i < argc; ++i) {
 			const char *uuid = NULL;
 			scan(h, argv[i]);

daemons/lvmlockd/.gitignore

@@ -0,0 +1,2 @@
+lvmlockctl
+lvmlockd

daemons/lvmlockd/Makefile.in

@@ -15,10 +15,15 @@ srcdir = @srcdir@
 top_srcdir = @top_srcdir@
 top_builddir = @top_builddir@
 
-SOURCES = \
-	lvmlockd-core.c \
-	lvmlockd-sanlock.c \
-	lvmlockd-dlm.c
+SOURCES = lvmlockd-core.c
+
+ifeq ("@BUILD_LOCKDSANLOCK@", "yes")
+  SOURCES += lvmlockd-sanlock.c
+endif
+
+ifeq ("@BUILD_LOCKDDLM@", "yes")
+  SOURCES += lvmlockd-dlm.c
+endif
 
 TARGETS = lvmlockd lvmlockctl
 
@@ -29,7 +34,15 @@ include $(top_builddir)/make.tmpl
 INCLUDES += -I$(top_srcdir)/libdaemon/server
 LVMLIBS = -ldaemonserver $(LVMINTERNAL_LIBS) -ldevmapper
 
-LIBS += $(PTHREAD_LIBS) -ldlm_lt -lsanlock_client -lrt
+LIBS += $(PTHREAD_LIBS)
+
+ifeq ("@BUILD_LOCKDSANLOCK@", "yes")
+  LIBS += -lsanlock_client
+endif
+
+ifeq ("@BUILD_LOCKDDLM@", "yes")
+  LIBS += -ldlm_lt
+endif
 
 LDFLAGS += -L$(top_builddir)/libdaemon/server
 CLDFLAGS += -L$(top_builddir)/libdaemon/server

daemons/lvmlockd/lvmlockctl.c

@@ -8,49 +8,41 @@
  * of the GNU Lesser General Public License v.2.1.
  */
 
-#define _GNU_SOURCE
-#include "configure.h"
+#include "tool.h"
+
 #include "lvmlockd-client.h"
 
-#include <stdio.h>
-#include <stdint.h>
 #include <stddef.h>
-#include <stdlib.h>
-#include <unistd.h>
 #include <getopt.h>
-#include <string.h>
 #include <signal.h>
 #include <errno.h>
 #include <fcntl.h>
-#include <sys/types.h>
+#include <syslog.h>
 #include <sys/wait.h>
 #include <sys/socket.h>
 #include <sys/un.h>
 
-static int quit;
-static int info;
-static int dump;
-static int wait_opt;
-static int force_opt;
-static int gl_enable;
-static int gl_disable;
-static int stop_lockspaces;
-static char *able_vg_name;
+static int quit = 0;
+static int info = 0;
+static int dump = 0;
+static int wait_opt = 0;
+static int force_opt = 0;
+static int kill_vg = 0;
+static int drop_vg = 0;
+static int gl_enable = 0;
+static int gl_disable = 0;
+static int stop_lockspaces = 0;
+static char *arg_vg_name = NULL;
 
 #define DUMP_SOCKET_NAME "lvmlockd-dump.sock"
 #define DUMP_BUF_SIZE (1024 * 1024)
-static char dump_buf[DUMP_BUF_SIZE];
+static char dump_buf[DUMP_BUF_SIZE+1];
 static int dump_len;
 static struct sockaddr_un dump_addr;
 static socklen_t dump_addrlen;
 
 daemon_handle _lvmlockd;
 
-#define log_debug(fmt, args...) \
-do { \
-	printf(fmt "\n", ##args); \
-} while (0)
-
 #define log_error(fmt, args...) \
 do { \
 	printf(fmt "\n", ##args); \
@@ -108,6 +100,8 @@ static void find_client_info(uint32_t client_id, uint32_t *pid, char *cl_name)
 	}
 }
 
+static int first_ls = 1;
+
 static void format_info_ls(char *line)
 {
 	char ls_name[MAX_NAME+1] = { 0 };
@@ -120,7 +114,9 @@ static void format_info_ls(char *line)
 	sscanf(line, "info=ls ls_name=%s vg_name=%s vg_uuid=%s vg_sysid=%s vg_args=%s lm_type=%s",
 	       ls_name, vg_name, vg_uuid, vg_sysid, lock_args, lock_type);
 
-	printf("\n");
+	if (!first_ls)
+		printf("\n");
+	first_ls = 0;
 
 	printf("VG %s lock_type=%s %s\n", vg_name, lock_type, vg_uuid);
 
@@ -141,7 +137,7 @@ static void format_info_ls_action(char *line)
 
 	find_client_info(client_id, &pid, cl_name);
 
-	printf("OP %s pid %u (%s)", op, pid, cl_name);
+	printf("OP %s pid %u (%s)\n", op, pid, cl_name);
 }
 
 static void format_info_r(char *line, char *r_name_out, char *r_type_out)
@@ -155,21 +151,20 @@ static void format_info_r(char *line, char *r_name_out, char *r_type_out)
 	sscanf(line, "info=r name=%s type=%s mode=%s %s version=%u",
 	       r_name, r_type, mode, sh_count, &ver);
 
-	/* when mode is not un, wait and print each lk line */
+	strcpy(r_name_out, r_name);
+	strcpy(r_type_out, r_type);
 
-	if (strcmp(mode, "un")) {
-		strcpy(r_name_out, r_name);
-		strcpy(r_type_out, r_type);
+	/* when mode is not un, wait and print each lk line */
+	if (strcmp(mode, "un"))
 		return;
-	}
 
 	/* when mode is un, there will be no lk lines, so print now */
 
 	if (!strcmp(r_type, "gl")) {
-		printf("LK GL un ver %4u\n", ver);
+		printf("LK GL un ver %u\n", ver);
 
 	} else if (!strcmp(r_type, "vg")) {
-		printf("LK VG un ver %4u\n", ver);
+		printf("LK VG un ver %u\n", ver);
 
 	} else if (!strcmp(r_type, "lv")) {
 		printf("LK LV un %s\n", r_name);
@@ -197,10 +192,10 @@ static void format_info_lk(char *line, char *r_name, char *r_type)
 	find_client_info(client_id, &pid, cl_name);
 
 	if (!strcmp(r_type, "gl")) {
-		printf("LK GL %s ver %4u pid %u (%s)\n", mode, ver, pid, cl_name);
+		printf("LK GL %s ver %u pid %u (%s)\n", mode, ver, pid, cl_name);
 
 	} else if (!strcmp(r_type, "vg")) {
-		printf("LK VG %s ver %4u pid %u (%s)\n", mode, ver, pid, cl_name);
+		printf("LK VG %s ver %u pid %u (%s)\n", mode, ver, pid, cl_name);
 
 	} else if (!strcmp(r_type, "lv")) {
 		printf("LK LV %s %s\n", mode, r_name);
@@ -233,28 +228,25 @@ static void format_info_r_action(char *line, char *r_name, char *r_type)
 	find_client_info(client_id, &pid, cl_name);
 
 	if (strcmp(op, "lock")) {
-		printf("OP %s pid %u (%s)", op, pid, cl_name);
+		printf("OP %s pid %u (%s)\n", op, pid, cl_name);
 		return;
 	}
 
 	if (!strcmp(r_type, "gl")) {
-		printf("LW GL %s ver %4u pid %u (%s)\n", mode, 0, pid, cl_name);
+		printf("LW GL %s ver %u pid %u (%s)\n", mode, 0, pid, cl_name);
 
 	} else if (!strcmp(r_type, "vg")) {
-		printf("LW VG %s ver %4u pid %u (%s)\n", mode, 0, pid, cl_name);
+		printf("LW VG %s ver %u pid %u (%s)\n", mode, 0, pid, cl_name);
 
 	} else if (!strcmp(r_type, "lv")) {
 		printf("LW LV %s %s\n", mode, r_name);
 	}
 }
 
-static void format_info_line(char *line)
+static void format_info_line(char *line, char *r_name, char *r_type)
 {
-	char r_name[MAX_NAME+1];
-	char r_type[MAX_NAME+1];
-
 	if (!strncmp(line, "info=structs ", strlen("info=structs "))) {
-		printf("%s\n", line);
+		/* only print this in the raw info dump */
 
 	} else if (!strncmp(line, "info=client ", strlen("info=client "))) {
 		save_client_info(line);
@@ -266,8 +258,13 @@ static void format_info_line(char *line)
 		format_info_ls_action(line);
 
 	} else if (!strncmp(line, "info=r ", strlen("info=r "))) {
-		memset(r_name, 0, sizeof(r_name));
-		memset(r_type, 0, sizeof(r_type));
+		/*
+		 * r_name/r_type are reset when a new resource is found.
+		 * They are reused for the lock and action lines that
+		 * follow a resource line.
+		 */
+		memset(r_name, 0, MAX_NAME+1);
+		memset(r_type, 0, MAX_NAME+1);
 		format_info_r(line, r_name, r_type);
 
 	} else if (!strncmp(line, "info=lk ", strlen("info=lk "))) {
@@ -285,6 +282,8 @@ static void format_info_line(char *line)
 static void format_info(void)
 {
 	char line[MAX_LINE];
+	char r_name[MAX_NAME+1];
+	char r_type[MAX_NAME+1];
 	int i, j;
 
 	j = 0;
@@ -294,7 +293,7 @@ static void format_info(void)
 		line[j++] = dump_buf[i];
 
 		if ((line[j-1] == '\n') || (line[j-1] == '\0')) {
-			format_info_line(line);
+			format_info_line(line, r_name, r_type);
 			j = 0;
 			memset(line, 0, sizeof(line));
 		}
@@ -327,8 +326,6 @@ static daemon_reply _lvmlockd_send(const char *req_name, ...)
 static int _lvmlockd_result(daemon_reply reply, int *result)
 {
 	int reply_result;
-	const char *reply_flags;
-	const char *lock_type;
 
 	if (reply.error) {
 		log_error("lvmlockd_result reply error %d", reply.error);
@@ -346,14 +343,8 @@ static int _lvmlockd_result(daemon_reply reply, int *result)
 		return 0;
 	}
 
-	/* The lock_type that lvmlockd used for locking. */
-	lock_type = daemon_reply_str(reply, "lock_type", "none");
-
 	*result = reply_result;
 
-	reply_flags = daemon_reply_str(reply, "result_flags", NULL);
-
-	log_debug("lvmlockd_result %d %s lm %s", reply_result, reply_flags, lock_type);
 	return 1;
 }
 
@@ -387,8 +378,12 @@ static int setup_dump_socket(void)
 	dump_addrlen = sizeof(sa_family_t) + strlen(dump_addr.sun_path+1) + 1;
 
 	rv = bind(s, (struct sockaddr *) &dump_addr, dump_addrlen);
-	if (rv < 0)
+	if (rv < 0) {
+		rv = -errno;
+		if (!close(s))
+			log_error("failed to close dump socket");
 		return rv;
+	}
 
 	return s;
 }
@@ -398,6 +393,7 @@ static int do_dump(const char *req_name)
 	daemon_reply reply;
 	int result;
 	int fd, rv = 0;
+	int count = 0;
 
 	fd = setup_dump_socket();
 	if (fd < 0) {
@@ -428,13 +424,18 @@ static int do_dump(const char *req_name)
 
 	memset(dump_buf, 0, sizeof(dump_buf));
 
-	rv = recvfrom(fd, dump_buf, dump_len, MSG_WAITALL,
+retry:
+	rv = recvfrom(fd, dump_buf + count, dump_len - count, MSG_WAITALL,
 		      (struct sockaddr *)&dump_addr, &dump_addrlen);
 	if (rv < 0) {
 		log_error("recvfrom error %d %d", rv, errno);
 		rv = -errno;
 		goto out;
 	}
+	count += rv;
+
+	if (count < dump_len)
+		goto retry;
 
 	rv = 0;
 	if ((info && dump) || !strcmp(req_name, "dump"))
@@ -442,7 +443,8 @@ static int do_dump(const char *req_name)
 	else
 		format_info();
 out:
-	close(fd);
+	if (close(fd))
+		log_error("failed to close dump socket %d", fd);
 	return rv;
 }
 
@@ -453,9 +455,9 @@ static int do_able(const char *req_name)
 	int rv;
 
 	reply = _lvmlockd_send(req_name,
-				"cmd = %s", "lvmlock",
+				"cmd = %s", "lvmlockctl",
 				"pid = %d", getpid(),
-				"vg_name = %s", able_vg_name,
+				"vg_name = %s", arg_vg_name,
 				NULL);
 
 	if (!_lvmlockd_result(reply, &result)) {
@@ -484,7 +486,7 @@ static int do_stop_lockspaces(void)
 		strcat(opts, "force ");
 
 	reply = _lvmlockd_send("stop_all",
-				"cmd = %s", "lvmlock",
+				"cmd = %s", "lvmlockctl",
 				"pid = %d", getpid(),
 				"opts = %s", opts[0] ? opts : "none",
 				NULL);
@@ -500,6 +502,87 @@ static int do_stop_lockspaces(void)
 	return rv;
 }
 
+static int do_kill(void)
+{
+	daemon_reply reply;
+	int result;
+	int rv;
+
+	syslog(LOG_EMERG, "Lost access to sanlock lease storage in VG %s.", arg_vg_name);
+	/* These two lines explain the manual alternative to the FIXME below. */
+	syslog(LOG_EMERG, "Immediately deactivate LVs in VG %s.", arg_vg_name);
+	syslog(LOG_EMERG, "Once VG is unused, run lvmlockctl --drop %s.", arg_vg_name);
+
+	/*
+	 * It may not be strictly necessary to notify lvmlockd of the kill, but
+	 * lvmlockd can use this information to avoid attempting any new lock
+	 * requests in the VG (which would fail anyway), and can return an
+	 * error indicating that the VG has been killed.
+	 */
+
+	reply = _lvmlockd_send("kill_vg",
+				"cmd = %s", "lvmlockctl",
+				"pid = %d", getpid(),
+				"vg_name = %s", arg_vg_name,
+				NULL);
+
+	if (!_lvmlockd_result(reply, &result)) {
+		log_error("lvmlockd result %d", result);
+		rv = result;
+	} else {
+		rv = 0;
+	}
+
+	daemon_reply_destroy(reply);
+
+	/*
+	 * FIXME: here is where we should implement a strong form of
+	 * blkdeactivate, and if it completes successfully, automatically call
+	 * do_drop() afterward.  (The drop step may not always be necessary
+	 * if the lvm commands run while shutting things down release all the
+	 * leases.)
+	 *
+	 * run_strong_blkdeactivate();
+	 * do_drop();
+	 */
+
+	return rv;
+}
+
+static int do_drop(void)
+{
+	daemon_reply reply;
+	int result;
+	int rv;
+
+	syslog(LOG_WARNING, "Dropping locks for VG %s.", arg_vg_name);
+
+	/*
+	 * Check for misuse by looking for any active LVs in the VG
+	 * and refusing this operation if found?  One possible way
+	 * to kill LVs (e.g. if fs cannot be unmounted) is to suspend
+	 * them, or replace them with the error target.  In that
+	 * case the LV will still appear to be active, but it is
+	 * safe to release the lock.
+	 */
+
+	reply = _lvmlockd_send("drop_vg",
+				"cmd = %s", "lvmlockctl",
+				"pid = %d", getpid(),
+				"vg_name = %s", arg_vg_name,
+				NULL);
+
+	if (!_lvmlockd_result(reply, &result)) {
+		log_error("lvmlockd result %d", result);
+		rv = result;
+	} else {
+		rv = 0;
+	}
+
+	daemon_reply_destroy(reply);
+	return rv;
+}
+
 static void print_usage(void)
 {
 	printf("lvmlockctl options\n");
@@ -516,12 +599,16 @@ static void print_usage(void)
 	printf("      Wait option for other commands.\n");
 	printf("--force | -f 0|1>\n");
 	printf("      Force option for other commands.\n");
+	printf("--kill | -k <vgname>\n");
+	printf("      Kill access to the VG when sanlock cannot renew lease.\n");
+	printf("--drop | -r <vgname>\n");
+	printf("      Clear locks for the VG when it is unused after kill (-k).\n");
+	printf("--gl-enable | -E <vgname>\n");
+	printf("      Tell lvmlockd to enable the global lock in a sanlock VG.\n");
+	printf("--gl-disable | -D <vgname>\n");
+	printf("      Tell lvmlockd to disable the global lock in a sanlock VG.\n");
 	printf("--stop-lockspaces | -S\n");
 	printf("      Stop all lockspaces.\n");
-	printf("--gl-enable <vg_name>\n");
-	printf("      Tell lvmlockd to enable the global lock in a sanlock vg.\n");
-	printf("--gl-disable <vg_name>\n");
-	printf("      Tell lvmlockd to disable the global lock in a sanlock vg.\n");
 }
 
 static int read_options(int argc, char *argv[])
@@ -536,6 +623,8 @@ static int read_options(int argc, char *argv[])
 		{"dump",            no_argument,       0,  'd' },
 		{"wait",            required_argument, 0,  'w' },
 		{"force",           required_argument, 0,  'f' },
+		{"kill",            required_argument, 0,  'k' },
+		{"drop",            required_argument, 0,  'r' },
 		{"gl-enable",       required_argument, 0,  'E' },
 		{"gl-disable",      required_argument, 0,  'D' },
 		{"stop-lockspaces", no_argument,       0,  'S' },
@@ -548,7 +637,7 @@ static int read_options(int argc, char *argv[])
 	}
 
 	while (1) {
-		c = getopt_long(argc, argv, "hqidE:D:w:S", long_options, &option_index);
+		c = getopt_long(argc, argv, "hqidE:D:w:k:r:S", long_options, &option_index);
 		if (c == -1)
 			break;
 
@@ -572,13 +661,21 @@ static int read_options(int argc, char *argv[])
 		case 'w':
 			wait_opt = atoi(optarg);
 			break;
+		case 'k':
+			kill_vg = 1;
+			arg_vg_name = strdup(optarg);
+			break;
+		case 'r':
+			drop_vg = 1;
+			arg_vg_name = strdup(optarg);
+			break;
 		case 'E':
 			gl_enable = 1;
-			able_vg_name = strdup(optarg);
+			arg_vg_name = strdup(optarg);
 			break;
 		case 'D':
 			gl_disable = 1;
-			able_vg_name = strdup(optarg);
+			arg_vg_name = strdup(optarg);
 			break;
 		case 'S':
 			stop_lockspaces = 1;
@@ -604,7 +701,7 @@ int main(int argc, char **argv)
 	_lvmlockd = lvmlockd_open(NULL);
 
 	if (_lvmlockd.socket_fd < 0 || _lvmlockd.error) {
-		log_error("lvmlockd open error %d", _lvmlockd.error);
+		log_error("Cannot connect to lvmlockd.");
 		return -1;
 	}
 
@@ -623,12 +720,24 @@ int main(int argc, char **argv)
 		goto out;
 	}
 
+	if (kill_vg) {
+		rv = do_kill();
+		goto out;
+	}
+
+	if (drop_vg) {
+		rv = do_drop();
+		goto out;
+	}
+
 	if (gl_enable) {
+		syslog(LOG_INFO, "Enabling global lock in VG %s.", arg_vg_name);
 		rv = do_able("enable_gl");
 		goto out;
 	}
 
 	if (gl_disable) {
+		syslog(LOG_INFO, "Disabling global lock in VG %s.", arg_vg_name);
 		rv = do_able("disable_gl");
 		goto out;
 	}

daemons/lvmlockd/lvmlockd-client.h

@@ -45,5 +45,8 @@ static inline void lvmlockd_close(daemon_handle h)
 #define EMANAGER  214
 #define EPREPARE  215
 #define ELOCKD    216
+#define EVGKILLED 217 /* sanlock lost access to leases and VG is killed. */
+#define ELOCKIO   218 /* sanlock io errors during lock op, may be transient. */
+#define EREMOVED  219
 
 #endif	/* _LVM_LVMLOCKD_CLIENT_H */

daemons/lvmlockd/lvmlockd-dlm.c

@@ -10,28 +10,9 @@
 
 #define _XOPEN_SOURCE 500  /* pthread */
 #define _ISOC99_SOURCE
-#define _GNU_SOURCE
 
-#include <assert.h>
-#include <pthread.h>
-#include <stdint.h>
-#include <stddef.h>
-#include <stdlib.h>
-#include <unistd.h>
-#include <stdio.h>
-#include <poll.h>
-#include <errno.h>
-#include <string.h>
-#include <endian.h>
-#include <fcntl.h>
-#include <byteswap.h>
-#include <syslog.h>
-#include <dirent.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <sys/socket.h>
+#include "tool.h"
 
-#include "configure.h"
 #include "daemon-server.h"
 #include "daemon-log.h"
 #include "xlate.h"
@@ -45,6 +26,17 @@
  */
 #include "libdlm.h"
 
+#include <pthread.h>
+#include <stddef.h>
+#include <poll.h>
+#include <errno.h>
+#include <endian.h>
+#include <fcntl.h>
+#include <byteswap.h>
+#include <syslog.h>
+#include <dirent.h>
+#include <sys/socket.h>
+
 struct lm_dlm {
 	dlm_lshandle_t *dh;
 };
@@ -75,6 +67,8 @@ int lm_data_size_dlm(void)
 #define VG_LOCK_ARGS_MINOR 0
 #define VG_LOCK_ARGS_PATCH 0
 
+static int dlm_has_lvb_bug;
+
 static int cluster_name_from_args(char *vg_args, char *clustername)
 {
 	return last_string_from_args(vg_args, clustername);
@@ -104,6 +98,7 @@ static int check_args_version(char *vg_args)
 
 static int read_cluster_name(char *clustername)
 {
+	static const char close_error_msg[] = "read_cluster_name: close_error %d";
 	char *n;
 	int fd;
 	int rv;
@@ -119,24 +114,26 @@ static int read_cluster_name(char *clustername)
 		return fd;
 	}
 
-	rv = read(fd, clustername, MAX_ARGS - 1);
+	rv = read(fd, clustername, MAX_ARGS);
 	if (rv < 0) {
 		log_error("read_cluster_name: cluster name read error %d, check dlm_controld", fd);
-		close(fd);
+		if (close(fd))
+			log_error(close_error_msg, fd);
 		return rv;
 	}
 
 	n = strstr(clustername, "\n");
 	if (n)
 		*n = '\0';
-	close(fd);
+	if (close(fd))
+		log_error(close_error_msg, fd);
 	return 0;
 }
 
 int lm_init_vg_dlm(char *ls_name, char *vg_name, uint32_t flags, char *vg_args)
 {
-	char clustername[MAX_ARGS];
-	char lock_args_version[MAX_ARGS];
+	char clustername[MAX_ARGS+1];
+	char lock_args_version[MAX_ARGS+1];
 	int rv;
 
 	memset(clustername, 0, sizeof(clustername));
@@ -163,8 +160,9 @@ int lm_init_vg_dlm(char *ls_name, char *vg_name, uint32_t flags, char *vg_args)
 
 int lm_prepare_lockspace_dlm(struct lockspace *ls)
 {
-	char sys_clustername[MAX_ARGS];
-	char arg_clustername[MAX_ARGS];
+	char sys_clustername[MAX_ARGS+1];
+	char arg_clustername[MAX_ARGS+1];
+	uint32_t major = 0, minor = 0, patch = 0;
 	struct lm_dlm *lmd;
 	int rv;
 
@@ -175,6 +173,17 @@ int lm_prepare_lockspace_dlm(struct lockspace *ls)
 	if (rv < 0)
 		return -EMANAGER;
 
+	rv = dlm_kernel_version(&major, &minor, &patch);
+	if (rv < 0) {
+		log_error("prepare_lockspace_dlm kernel_version not detected %d", rv);
+		dlm_has_lvb_bug = 1;
+	}
+
+	if ((major == 6) && (minor == 0) && (patch == 1)) {
+		log_debug("dlm kernel version %u.%u.%u has lvb bug", major, minor, patch);
+		dlm_has_lvb_bug = 1;
+	}
+
 	if (!ls->vg_args[0]) {
 		/* global lockspace has no vg args */
 		goto skip_args;
@@ -251,12 +260,6 @@ int lm_rem_lockspace_dlm(struct lockspace *ls, int free_vg)
  out:
 	free(lmd);
 	ls->lm_data = NULL;
-
-	if (!strcmp(ls->name, gl_lsname_dlm)) {
-		gl_running_dlm = 0;
-		gl_auto_dlm = 0;
-	}
-
 	return 0;
 }
 
@@ -340,7 +343,7 @@ static int to_dlm_mode(int ld_mode)
 }
 
 static int lm_adopt_dlm(struct lockspace *ls, struct resource *r, int ld_mode,
-			uint32_t *r_version)
+			struct val_blk *vb_out)
 {
 	struct lm_dlm *lmd = (struct lm_dlm *)ls->lm_data;
 	struct rd_dlm *rdd = (struct rd_dlm *)r->lm_data;
@@ -349,7 +352,7 @@ static int lm_adopt_dlm(struct lockspace *ls, struct resource *r, int ld_mode,
 	int mode;
 	int rv;
 
-	*r_version = 0;
+	memset(vb_out, 0, sizeof(struct val_blk));
 
 	if (!r->lm_init) {
 		rv = lm_add_resource_dlm(ls, r, 0);
@@ -391,7 +394,7 @@ static int lm_adopt_dlm(struct lockspace *ls, struct resource *r, int ld_mode,
 			  (void *)1, (void *)1, (void *)1,
 			  NULL, NULL);
 
-	if (rv == -EAGAIN) {
+	if (rv == -1 && errno == -EAGAIN) {
 		log_debug("S %s R %s adopt_dlm adopt mode %d try other mode",
 			  ls->name, r->name, ld_mode);
 		rv = -EUCLEAN;
@@ -428,14 +431,13 @@ static int lm_adopt_dlm(struct lockspace *ls, struct resource *r, int ld_mode,
  */
 
 int lm_lock_dlm(struct lockspace *ls, struct resource *r, int ld_mode,
-		uint32_t *r_version, int adopt)
+		struct val_blk *vb_out, int adopt)
 {
 	struct lm_dlm *lmd = (struct lm_dlm *)ls->lm_data;
 	struct rd_dlm *rdd = (struct rd_dlm *)r->lm_data;
 	struct dlm_lksb *lksb;
 	struct val_blk vb;
 	uint32_t flags = 0;
-	uint16_t vb_version;
 	int mode;
 	int rv;
 
@@ -443,7 +445,7 @@ int lm_lock_dlm(struct lockspace *ls, struct resource *r, int ld_mode,
 		/* When adopting, we don't follow the normal method
 		   of acquiring a NL lock then converting it to the
 		   desired mode. */
-		return lm_adopt_dlm(ls, r, ld_mode, r_version);
+		return lm_adopt_dlm(ls, r, ld_mode, vb_out);
 	}
 
 	if (!r->lm_init) {
@@ -471,19 +473,37 @@ int lm_lock_dlm(struct lockspace *ls, struct resource *r, int ld_mode,
 	log_debug("S %s R %s lock_dlm", ls->name, r->name);
 
 	if (daemon_test) {
-		*r_version = 0;
+		memset(vb_out, 0, sizeof(struct val_blk));
 		return 0;
 	}
 
+	/*
+	 * The dlm lvb bug means that converting NL->EX will not return 
+	 * the latest lvb, so we have to convert NL->PR->EX to reread it.
+	 */
+	if (dlm_has_lvb_bug && (ld_mode == LD_LK_EX)) {
+		rv = dlm_ls_lock_wait(lmd->dh, LKM_PRMODE, lksb, flags,
+				      r->name, strlen(r->name),
+				      0, NULL, NULL, NULL);
+		if (rv == -1) {
+			log_debug("S %s R %s lock_dlm acquire mode PR for %d rv %d",
+				  ls->name, r->name, mode, rv);
+			goto lockrv;
+		}
+
+		/* Fall through to request EX. */
+	}
+
 	rv = dlm_ls_lock_wait(lmd->dh, mode, lksb, flags,
 			      r->name, strlen(r->name),
 			      0, NULL, NULL, NULL);
-	if (rv == -EAGAIN) {
-		log_error("S %s R %s lock_dlm mode %d rv EAGAIN", ls->name, r->name, mode);
+lockrv:
+	if (rv == -1 && errno == EAGAIN) {
+		log_debug("S %s R %s lock_dlm acquire mode %d rv EAGAIN", ls->name, r->name, mode);
 		return -EAGAIN;
 	}
 	if (rv < 0) {
-		log_error("S %s R %s lock_dlm error %d", ls->name, r->name, rv);
+		log_error("S %s R %s lock_dlm acquire error %d errno %d", ls->name, r->name, rv, errno);
 		return rv;
 	}
 
@@ -491,28 +511,22 @@ int lm_lock_dlm(struct lockspace *ls, struct resource *r, int ld_mode,
 		if (lksb->sb_flags & DLM_SBF_VALNOTVALID) {
 			log_debug("S %s R %s lock_dlm VALNOTVALID", ls->name, r->name);
 			memset(rdd->vb, 0, sizeof(struct val_blk));
-			*r_version = 0;
+			memset(vb_out, 0, sizeof(struct val_blk));
 			goto out;
 		}
 
+		/*
+		 * 'vb' contains disk endian values, not host endian.
+		 * It is copied directly to rdd->vb which is also kept
+		 * in disk endian form.
+		 * vb_out is returned to the caller in host endian form.
+		 */
 		memcpy(&vb, lksb->sb_lvbptr, sizeof(struct val_blk));
-		vb_version = le16_to_cpu(vb.version);
+		memcpy(rdd->vb, &vb, sizeof(vb));
 
-		if (vb_version && ((vb_version & 0xFF00) > (VAL_BLK_VERSION & 0xFF00))) {
-			log_error("S %s R %s lock_dlm ignore vb_version %x",
-				  ls->name, r->name, vb_version);
-			*r_version = 0;
-			free(rdd->vb);
-			rdd->vb = NULL;
-			lksb->sb_lvbptr = NULL;
-			goto out;
-		}
-
-		*r_version = le32_to_cpu(vb.r_version);
-		memcpy(rdd->vb, &vb, sizeof(vb)); /* rdd->vb saved as le */
-
-		log_debug("S %s R %s lock_dlm get r_version %u",
-			  ls->name, r->name, *r_version);
+		vb_out->version = le16_to_cpu(vb.version);
+		vb_out->flags = le16_to_cpu(vb.flags);
+		vb_out->r_version = le32_to_cpu(vb.r_version);
 	}
 out:
 	return 0;
@@ -556,7 +570,7 @@ int lm_convert_dlm(struct lockspace *ls, struct resource *r,
 	rv = dlm_ls_lock_wait(lmd->dh, mode, lksb, flags,
 			      r->name, strlen(r->name),
 			      0, NULL, NULL, NULL);
-	if (rv == -EAGAIN) {
+	if (rv == -1 && errno == EAGAIN) {
 		/* FIXME: When does this happen?  Should something different be done? */
 		log_error("S %s R %s convert_dlm mode %d rv EAGAIN", ls->name, r->name, mode);
 		return -EAGAIN;
@@ -568,17 +582,17 @@ int lm_convert_dlm(struct lockspace *ls, struct resource *r,
 }
 
 int lm_unlock_dlm(struct lockspace *ls, struct resource *r,
-		  uint32_t r_version, uint32_t lmuf_flags)
+		  uint32_t r_version, uint32_t lmu_flags)
 {
 	struct lm_dlm *lmd = (struct lm_dlm *)ls->lm_data;
 	struct rd_dlm *rdd = (struct rd_dlm *)r->lm_data;
 	struct dlm_lksb *lksb = &rdd->lksb;
+	struct val_blk vb_prev;
+	struct val_blk vb_next;
 	uint32_t flags = 0;
+	int new_vb = 0;
 	int rv;
 
-	log_debug("S %s R %s unlock_dlm r_version %u flags %x",
-		  ls->name, r->name, r_version, lmuf_flags);
-
 	/*
 	 * Do not set PERSISTENT, because we don't need an orphan
 	 * NL lock to protect anything.
@@ -586,19 +600,46 @@ int lm_unlock_dlm(struct lockspace *ls, struct resource *r,
 
 	flags |= LKF_CONVERT;
 
-	if (rdd->vb && r_version && (r->mode == LD_LK_EX)) {
-		if (!rdd->vb->version) {
-			/* first time vb has been written */
-			rdd->vb->version = cpu_to_le16(VAL_BLK_VERSION);
-		}
-		if (r_version)
-			rdd->vb->r_version = cpu_to_le32(r_version);
-		memcpy(lksb->sb_lvbptr, rdd->vb, sizeof(struct val_blk));
+	if (rdd->vb && (r->mode == LD_LK_EX)) {
 
-		log_debug("S %s R %s unlock_dlm set r_version %u",
-			  ls->name, r->name, r_version);
+		/* vb_prev and vb_next are in disk endian form */
+		memcpy(&vb_prev, rdd->vb, sizeof(struct val_blk));
+		memcpy(&vb_next, rdd->vb, sizeof(struct val_blk));
+
+		if (!vb_prev.version) {
+			vb_next.version = cpu_to_le16(VAL_BLK_VERSION);
+			new_vb = 1;
+		}
+
+		if ((lmu_flags & LMUF_FREE_VG) && (r->type == LD_RT_VG)) {
+			vb_next.flags = cpu_to_le16(VBF_REMOVED);
+			new_vb = 1;
+		}
+
+		if (r_version) {
+			vb_next.r_version = cpu_to_le32(r_version);
+			new_vb = 1;
+		}
+
+		if (new_vb) {
+			memcpy(rdd->vb, &vb_next, sizeof(struct val_blk));
+			memcpy(lksb->sb_lvbptr, &vb_next, sizeof(struct val_blk));
+
+			log_debug("S %s R %s unlock_dlm vb old %x %x %u new %x %x %u",
+				  ls->name, r->name,
+				  le16_to_cpu(vb_prev.version),
+				  le16_to_cpu(vb_prev.flags),
+				  le32_to_cpu(vb_prev.r_version),
+				  le16_to_cpu(vb_next.version),
+				  le16_to_cpu(vb_next.flags),
+				  le32_to_cpu(vb_next.r_version));
+		} else {
+			log_debug("S %s R %s unlock_dlm vb unchanged", ls->name, r->name);
+		}
 
 		flags |= LKF_VALBLK;
+	} else {
+		log_debug("S %s R %s unlock_dlm", ls->name, r->name);
 	}
 
 	if (daemon_test)
@@ -621,8 +662,65 @@ int lm_unlock_dlm(struct lockspace *ls, struct resource *r,
 
 #define DLM_LOCKSPACES_PATH "/sys/kernel/config/dlm/cluster/spaces"
 
+/*
+ * FIXME: this should be implemented differently.
+ * It's not nice to use an aspect of the dlm clustering
+ * implementation, which could change.  It would be
+ * better to do something like use a special lock in the
+ * lockspace that was held PR by all nodes, and then an
+ * EX request on it could check if it's started (and
+ * possibly also notify others to stop it automatically).
+ * Or, possibly an enhancement to libdlm that would give
+ * info about lockspace members.
+ *
+ * (We could let the VG be removed while others still
+ * have the lockspace running, which largely works, but
+ * introduces problems if another VG with the same name is
+ * recreated while others still have the lockspace running
+ * for the previous VG.  We'd also want a way to clean up
+ * the stale lockspaces on the others eventually.)
+ */
+
+int lm_hosts_dlm(struct lockspace *ls, int notify)
+{
+	static const char closedir_err_msg[] = "lm_hosts_dlm: closedir failed";
+	char ls_nodes_path[PATH_MAX];
+	struct dirent *de;
+	DIR *ls_dir;
+	int count = 0;
+
+	memset(ls_nodes_path, 0, sizeof(ls_nodes_path));
+	snprintf(ls_nodes_path, PATH_MAX-1, "%s/%s/nodes",
+		 DLM_LOCKSPACES_PATH, ls->name);
+
+	if (!(ls_dir = opendir(ls_nodes_path)))
+		return -ECONNREFUSED;
+
+	while ((de = readdir(ls_dir))) {
+		if (de->d_name[0] == '.')
+			continue;
+		count++;
+	}
+
+	if (closedir(ls_dir))
+		log_error(closedir_err_msg);
+
+	if (!count) {
+		log_error("lm_hosts_dlm found no nodes in %s", ls_nodes_path);
+		return 0;
+	}
+
+	/*
+	 * Assume that a count of one node represents ourself,
+	 * and any value over one represents other nodes.
+	 */
+
+	return count - 1;
+}
+
 int lm_get_lockspaces_dlm(struct list_head *ls_rejoin)
 {
+	static const char closedir_err_msg[] = "lm_get_lockspace_dlm: closedir failed";
 	struct lockspace *ls;
 	struct dirent *de;
 	DIR *ls_dir;
@@ -638,7 +736,8 @@ int lm_get_lockspaces_dlm(struct list_head *ls_rejoin)
 			continue;
 
 		if (!(ls = alloc_lockspace())) {
-			closedir(ls_dir);
+			if (closedir(ls_dir))
+				log_error(closedir_err_msg);
 			return -ENOMEM;
 		}
 
@@ -648,13 +747,14 @@ int lm_get_lockspaces_dlm(struct list_head *ls_rejoin)
 		list_add_tail(&ls->list, ls_rejoin);
 	}
 
-	closedir(ls_dir);
+	if (closedir(ls_dir))
+		log_error(closedir_err_msg);
 	return 0;
 }
 
 int lm_is_running_dlm(void)
 {
-	char sys_clustername[MAX_ARGS];
+	char sys_clustername[MAX_ARGS+1];
 	int rv;
 
 	memset(sys_clustername, 0, sizeof(sys_clustername));
@@ -664,3 +764,4 @@ int lm_is_running_dlm(void)
 		return 0;
 	return 1;
 }
+

daemons/lvmlockd/lvmlockd-internal.h

@@ -50,7 +50,9 @@ enum {
 	LD_OP_RENAME_FINAL,
 	LD_OP_RUNNING_LM,
 	LD_OP_FIND_FREE_LOCK,
-	LD_OP_FORGET_VG_NAME,
+	LD_OP_KILL_VG,
+	LD_OP_DROP_VG,
+	LD_OP_BUSY,
 };
 
 /* resource types */
@@ -83,11 +85,12 @@ struct client {
 	unsigned int recv : 1;
 	unsigned int dead : 1;
 	unsigned int poll_ignore : 1;
+	unsigned int lock_ops : 1;
 	char name[MAX_NAME+1];
 };
 
 #define LD_AF_PERSISTENT           0x00000001
-#define LD_AF_UNUSED               0x00000002 /* use me */
+#define LD_AF_NO_CLIENT            0x00000002
 #define LD_AF_UNLOCK_CANCEL        0x00000004
 #define LD_AF_NEXT_VERSION         0x00000008
 #define LD_AF_WAIT                 0x00000010
@@ -98,9 +101,10 @@ struct client {
 #define LD_AF_SEARCH_LS            0x00000200
 #define LD_AF_WAIT_STARTING        0x00001000
 #define LD_AF_DUP_GL_LS            0x00002000
-#define LD_AF_INACTIVE_LS          0x00004000
-#define LD_AF_ADD_LS_ERROR         0x00008000
 #define LD_AF_ADOPT                0x00010000
+#define LD_AF_WARN_GL_REMOVED	   0x00020000
+#define LD_AF_LV_LOCK              0x00040000
+#define LD_AF_LV_UNLOCK            0x00080000
 
 /*
  * Number of times to repeat a lock request after
@@ -127,8 +131,8 @@ struct action {
 	char vg_name[MAX_NAME+1];
 	char lv_name[MAX_NAME+1];
 	char lv_uuid[MAX_NAME+1];
-	char vg_args[MAX_ARGS];
-	char lv_args[MAX_ARGS];
+	char vg_args[MAX_ARGS+1];
+	char lv_args[MAX_ARGS+1];
 	char vg_sysid[MAX_NAME+1];
 };
 
@@ -139,13 +143,14 @@ struct resource {
 	int8_t mode;
 	unsigned int sh_count;		/* number of sh locks on locks list */
 	uint32_t version;
+	uint32_t last_client_id;	/* last client_id to lock or unlock resource */
 	unsigned int lm_init : 1;	/* lm_data is initialized */
 	unsigned int adopt : 1;		/* temp flag in remove_inactive_lvs */
 	unsigned int version_zero_valid : 1;
+	unsigned int use_vb : 1;
 	struct list_head locks;
 	struct list_head actions;
-	struct val_blk *vb;
-	char lv_args[MAX_ARGS];
+	char lv_args[MAX_ARGS+1];
 	char lm_data[0];		/* lock manager specific data */
 };
 
@@ -164,7 +169,7 @@ struct lockspace {
 	char name[MAX_NAME+1];
 	char vg_name[MAX_NAME+1];
 	char vg_uuid[64];
-	char vg_args[MAX_ARGS];		/* lock manager specific args */
+	char vg_args[MAX_ARGS+1];	/* lock manager specific args */
 	char vg_sysid[MAX_NAME+1];
 	int8_t lm_type;			/* lock manager: LM_DLM, LM_SANLOCK */
 	void *lm_data;
@@ -182,13 +187,20 @@ struct lockspace {
 	unsigned int thread_done : 1;
 	unsigned int sanlock_gl_enabled: 1;
 	unsigned int sanlock_gl_dup: 1;
+	unsigned int free_vg: 1;
+	unsigned int kill_vg: 1;
+	unsigned int drop_vg: 1;
 
 	struct list_head actions;	/* new client actions */
 	struct list_head resources;	/* resource/lock state for gl/vg/lv */
 };
 
+/* val_blk version */
 #define VAL_BLK_VERSION 0x0101
 
+/* val_blk flags */
+#define VBF_REMOVED 0x0001
+
 struct val_blk {
 	uint16_t version;
 	uint16_t flags;
@@ -198,49 +210,6 @@ struct val_blk {
 /* lm_unlock flags */
 #define LMUF_FREE_VG 0x00000001
 
-struct lockspace *alloc_lockspace(void);
-int lockspaces_empty(void);
-int last_string_from_args(char *args_in, char *last);
-int version_from_args(char *args, unsigned int *major, unsigned int *minor, unsigned int *patch);
-
-int lm_init_vg_dlm(char *ls_name, char *vg_name, uint32_t flags, char *vg_args);
-int lm_prepare_lockspace_dlm(struct lockspace *ls);
-int lm_add_lockspace_dlm(struct lockspace *ls, int adopt);
-int lm_rem_lockspace_dlm(struct lockspace *ls, int free_vg);
-int lm_lock_dlm(struct lockspace *ls, struct resource *r, int ld_mode,
-		uint32_t *r_version, int adopt);
-int lm_convert_dlm(struct lockspace *ls, struct resource *r,
-		   int ld_mode, uint32_t r_version);
-int lm_unlock_dlm(struct lockspace *ls, struct resource *r,
-		  uint32_t r_version, uint32_t lmu_flags);
-int lm_rem_resource_dlm(struct lockspace *ls, struct resource *r);
-int lm_get_lockspaces_dlm(struct list_head *ls_rejoin);
-int lm_data_size_dlm(void);
-int lm_is_running_dlm(void);
-
-int lm_init_vg_sanlock(char *ls_name, char *vg_name, uint32_t flags, char *vg_args);
-int lm_init_lv_sanlock(char *ls_name, char *vg_name, char *lv_name, char *vg_args, char *lv_args, uint64_t free_offset);
-int lm_free_lv_sanlock(struct lockspace *ls, struct resource *r);
-int lm_rename_vg_sanlock(char *ls_name, char *vg_name, uint32_t flags, char *vg_args);
-int lm_prepare_lockspace_sanlock(struct lockspace *ls);
-int lm_add_lockspace_sanlock(struct lockspace *ls, int adopt);
-int lm_rem_lockspace_sanlock(struct lockspace *ls, int free_vg);
-int lm_lock_sanlock(struct lockspace *ls, struct resource *r, int ld_mode,
-		    uint32_t *r_version, int *retry, int adopt);
-int lm_convert_sanlock(struct lockspace *ls, struct resource *r,
-		       int ld_mode, uint32_t r_version);
-int lm_unlock_sanlock(struct lockspace *ls, struct resource *r,
-		      uint32_t r_version, uint32_t lmu_flags);
-int lm_able_gl_sanlock(struct lockspace *ls, int enable);
-int lm_ex_disable_gl_sanlock(struct lockspace *ls);
-int lm_hosts_sanlock(struct lockspace *ls, int notify);
-int lm_rem_resource_sanlock(struct lockspace *ls, struct resource *r);
-int lm_gl_is_enabled(struct lockspace *ls);
-int lm_get_lockspaces_sanlock(struct list_head *ls_rejoin);
-int lm_data_size_sanlock(void);
-int lm_is_running_sanlock(void);
-int lm_find_free_lock_sanlock(struct lockspace *ls, uint64_t *free_offset);
-
 #define container_of(ptr, type, member) ({                      \
 	const typeof( ((type *)0)->member ) *__mptr = (ptr);    \
 	(type *)( (char *)__mptr - offsetof(type,member) );})
@@ -351,13 +320,9 @@ static inline int list_empty(const struct list_head *head)
 EXTERN int gl_type_static;
 EXTERN int gl_use_dlm;
 EXTERN int gl_use_sanlock;
-EXTERN pthread_mutex_t gl_type_mutex;
-
 EXTERN char gl_lsname_dlm[MAX_NAME+1];
 EXTERN char gl_lsname_sanlock[MAX_NAME+1];
-
-EXTERN int gl_running_dlm;
-EXTERN int gl_auto_dlm;
+EXTERN int global_dlm_lockspace_exists;
 
 EXTERN int daemon_test; /* run as much as possible without a live lock manager */
 EXTERN int daemon_debug;
@@ -365,9 +330,258 @@ EXTERN int daemon_host_id;
 EXTERN const char *daemon_host_id_file;
 EXTERN int sanlock_io_timeout;
 
+/*
+ * This flag is set to 1 if we see multiple vgs with the global
+ * lock enabled.  While this is set, we return a special flag
+ * with the vg lock result indicating to the lvm command that
+ * there is a duplicate gl in the vg which should be resolved.
+ * While this is set, find_lockspace_name has the side job of
+ * counting the number of lockspaces with enabled gl's so that
+ * this can be set back to zero when the duplicates are disabled.
+ */
+EXTERN int sanlock_gl_dup;
+
 void log_level(int level, const char *fmt, ...)  __attribute__((format(printf, 2, 3)));
 #define log_debug(fmt, args...) log_level(LOG_DEBUG, fmt, ##args)
 #define log_error(fmt, args...) log_level(LOG_ERR, fmt, ##args)
 #define log_warn(fmt, args...) log_level(LOG_WARNING, fmt, ##args)
 
+struct lockspace *alloc_lockspace(void);
+int lockspaces_empty(void);
+int last_string_from_args(char *args_in, char *last);
+int version_from_args(char *args, unsigned int *major, unsigned int *minor, unsigned int *patch);
+
+
+#ifdef LOCKDDLM_SUPPORT
+
+int lm_init_vg_dlm(char *ls_name, char *vg_name, uint32_t flags, char *vg_args);
+int lm_prepare_lockspace_dlm(struct lockspace *ls);
+int lm_add_lockspace_dlm(struct lockspace *ls, int adopt);
+int lm_rem_lockspace_dlm(struct lockspace *ls, int free_vg);
+int lm_lock_dlm(struct lockspace *ls, struct resource *r, int ld_mode,
+		struct val_blk *vb_out, int adopt);
+int lm_convert_dlm(struct lockspace *ls, struct resource *r,
+		   int ld_mode, uint32_t r_version);
+int lm_unlock_dlm(struct lockspace *ls, struct resource *r,
+		  uint32_t r_version, uint32_t lmu_flags);
+int lm_rem_resource_dlm(struct lockspace *ls, struct resource *r);
+int lm_get_lockspaces_dlm(struct list_head *ls_rejoin);
+int lm_data_size_dlm(void);
+int lm_is_running_dlm(void);
+int lm_hosts_dlm(struct lockspace *ls, int notify);
+
+static inline int lm_support_dlm(void)
+{
+	return 1;
+}
+
+#else
+
+static inline int lm_init_vg_dlm(char *ls_name, char *vg_name, uint32_t flags, char *vg_args)
+{
+	return -1;
+}
+
+static inline int lm_prepare_lockspace_dlm(struct lockspace *ls)
+{
+	return -1;
+}
+
+static inline int lm_add_lockspace_dlm(struct lockspace *ls, int adopt)
+{
+	return -1;
+}
+
+static inline int lm_rem_lockspace_dlm(struct lockspace *ls, int free_vg)
+{
+	return -1;
+}
+
+static inline int lm_lock_dlm(struct lockspace *ls, struct resource *r, int ld_mode,
+		struct val_blk *vb_out, int adopt)
+{
+	return -1;
+}
+
+static inline int lm_convert_dlm(struct lockspace *ls, struct resource *r,
+		   int ld_mode, uint32_t r_version)
+{
+	return -1;
+}
+
+static inline int lm_unlock_dlm(struct lockspace *ls, struct resource *r,
+		  uint32_t r_version, uint32_t lmu_flags)
+{
+	return -1;
+}
+
+static inline int lm_rem_resource_dlm(struct lockspace *ls, struct resource *r)
+{
+	return -1;
+}
+
+static inline int lm_get_lockspaces_dlm(struct list_head *ls_rejoin)
+{
+	return -1;
+}
+
+static inline int lm_data_size_dlm(void)
+{
+	return -1;
+}
+
+static inline int lm_is_running_dlm(void)
+{
+	return 0;
+}
+
+static inline int lm_support_dlm(void)
+{
+	return 0;
+}
+
+static inline int lm_hosts_dlm(struct lockspace *ls, int notify)
+{
+	return 0;
+}
+
+#endif /* dlm support */
+
+#ifdef LOCKDSANLOCK_SUPPORT
+
+int lm_init_vg_sanlock(char *ls_name, char *vg_name, uint32_t flags, char *vg_args);
+int lm_init_lv_sanlock(char *ls_name, char *vg_name, char *lv_name, char *vg_args, char *lv_args, uint64_t free_offset);
+int lm_free_lv_sanlock(struct lockspace *ls, struct resource *r);
+int lm_rename_vg_sanlock(char *ls_name, char *vg_name, uint32_t flags, char *vg_args);
+int lm_prepare_lockspace_sanlock(struct lockspace *ls);
+int lm_add_lockspace_sanlock(struct lockspace *ls, int adopt);
+int lm_rem_lockspace_sanlock(struct lockspace *ls, int free_vg);
+int lm_lock_sanlock(struct lockspace *ls, struct resource *r, int ld_mode,
+		    struct val_blk *vb_out, int *retry, int adopt);
+int lm_convert_sanlock(struct lockspace *ls, struct resource *r,
+		       int ld_mode, uint32_t r_version);
+int lm_unlock_sanlock(struct lockspace *ls, struct resource *r,
+		      uint32_t r_version, uint32_t lmu_flags);
+int lm_able_gl_sanlock(struct lockspace *ls, int enable);
+int lm_ex_disable_gl_sanlock(struct lockspace *ls);
+int lm_hosts_sanlock(struct lockspace *ls, int notify);
+int lm_rem_resource_sanlock(struct lockspace *ls, struct resource *r);
+int lm_gl_is_enabled(struct lockspace *ls);
+int lm_get_lockspaces_sanlock(struct list_head *ls_rejoin);
+int lm_data_size_sanlock(void);
+int lm_is_running_sanlock(void);
+int lm_find_free_lock_sanlock(struct lockspace *ls, uint64_t *free_offset);
+
+static inline int lm_support_sanlock(void)
+{
+	return 1;
+}
+
+#else
+
+static inline int lm_init_vg_sanlock(char *ls_name, char *vg_name, uint32_t flags, char *vg_args)
+{
+	return -1;
+}
+
+static inline int lm_init_lv_sanlock(char *ls_name, char *vg_name, char *lv_name, char *vg_args, char *lv_args, uint64_t free_offset)
+{
+	return -1;
+}
+
+static inline int lm_free_lv_sanlock(struct lockspace *ls, struct resource *r)
+{
+	return -1;
+}
+
+static inline int lm_rename_vg_sanlock(char *ls_name, char *vg_name, uint32_t flags, char *vg_args)
+{
+	return -1;
+}
+
+static inline int lm_prepare_lockspace_sanlock(struct lockspace *ls)
+{
+	return -1;
+}
+
+static inline int lm_add_lockspace_sanlock(struct lockspace *ls, int adopt)
+{
+	return -1;
+}
+
+static inline int lm_rem_lockspace_sanlock(struct lockspace *ls, int free_vg)
+{
+	return -1;
+}
+
+static inline int lm_lock_sanlock(struct lockspace *ls, struct resource *r, int ld_mode,
+		    struct val_blk *vb_out, int *retry, int adopt)
+{
+	return -1;
+}
+
+static inline int lm_convert_sanlock(struct lockspace *ls, struct resource *r,
+		       int ld_mode, uint32_t r_version)
+{
+	return -1;
+}
+
+static inline int lm_unlock_sanlock(struct lockspace *ls, struct resource *r,
+		      uint32_t r_version, uint32_t lmu_flags)
+{
+	return -1;
+}
+
+static inline int lm_able_gl_sanlock(struct lockspace *ls, int enable)
+{
+	return -1;
+}
+
+static inline int lm_ex_disable_gl_sanlock(struct lockspace *ls)
+{
+	return -1;
+}
+
+static inline int lm_hosts_sanlock(struct lockspace *ls, int notify)
+{
+	return -1;
+}
+
+static inline int lm_rem_resource_sanlock(struct lockspace *ls, struct resource *r)
+{
+	return -1;
+}
+
+static inline int lm_gl_is_enabled(struct lockspace *ls)
+{
+	return -1;
+}
+
+static inline int lm_get_lockspaces_sanlock(struct list_head *ls_rejoin)
+{
+	return -1;
+}
+
+static inline int lm_data_size_sanlock(void)
+{
+	return -1;
+}
+
+static inline int lm_is_running_sanlock(void)
+{
+	return 0;
+}
+
+static inline int lm_find_free_lock_sanlock(struct lockspace *ls, uint64_t *free_offset)
+{
+	return -1;
+}
+
+static inline int lm_support_sanlock(void)
+{
+	return 0;
+}
+
+#endif /* sanlock support */
+
 #endif	/* _LVM_LVMLOCKD_INTERNAL_H */

daemons/lvmlockd/lvmlockd-sanlock.c

@@ -10,23 +10,9 @@
 
 #define _XOPEN_SOURCE 500  /* pthread */
 #define _ISOC99_SOURCE
-#define _GNU_SOURCE
 
-#include <assert.h>
-#include <pthread.h>
-#include <stdint.h>
-#include <stddef.h>
-#include <stdlib.h>
-#include <unistd.h>
-#include <stdio.h>
-#include <poll.h>
-#include <errno.h>
-#include <string.h>
-#include <syslog.h>
-#include <sys/types.h>
-#include <sys/socket.h>
+#include "tool.h"
 
-#include "configure.h"
 #include "daemon-server.h"
 #include "daemon-log.h"
 #include "xlate.h"
@@ -39,53 +25,109 @@
 #include "sanlock_admin.h"
 #include "sanlock_resource.h"
 
+#include <pthread.h>
+#include <stddef.h>
+#include <poll.h>
+#include <errno.h>
+#include <syslog.h>
+#include <sys/socket.h>
+
 /*
- * If access to the pv containing the vg's leases is lost, sanlock cannot renew
- * the leases we have acquired for locked LVs.  This means that we could soon
- * loose the lease to another host which could activate our LV exclusively.  We
- * do not want to get to the point of two hosts having the same LV active
- * exclusively (it obviously violates the purpose of LV locks.)
- *
- * The default method of preventing this problem is for lvmlockd to do nothing,
- * which produces a safe but potentially inconvenient result.  Doing nothing
- * leads to our LV leases not being released, which leads to sanlock using the
- * local watchdog to reset us before another host can acquire our lock.  It
- * would often be preferrable to avoid the abrupt hard reset from the watchdog.
- *
- * There are other options to avoid being reset by our watchdog.  If we can
- * quickly stop using the LVs in question and release the locks for them, then
- * we could avoid a reset (there's a certain grace period of about 40 seconds
- * in which we can attempt this.)  To do this, we can tell sanlock to run a
- * specific program when it has lost access to our leases.  We could use this
- * program to:
- *
- * 1. Deactivate all lvs in the effected vg.  If all the leases are
- * deactivated, then our LV locks would be released and sanlock would no longer
- * use the watchdog to reset us.  If file systems are mounted on the active
- * lvs, then deactivating them would fail, so this option would be of limited
- * usefulness.
- *
- * 2. Option 1 could be extended to kill pids using the fs on the lv, unmount
- * the fs, and deactivate the lv.  This is probably out of scope for lvm
- * directly, and would likely need the help of another system service.
- *
- * 3. Use dmsetup suspend to block access to lvs in the effected vg.  If this
- * was successful, the local host could no longer write to the lvs, we could
- * safely release the LV locks, and sanlock would no longer reset us.  At this
- * point, with suspended lvs, the host would be in a fairly hobbled state, and
- * would almost certainly need a manual, forcible reset.
- *
- * 4. Option 3 could be extended to monitor the lost storage, and if it is
- * reconnected, the leases could be reacquired, and the suspended lvs resumed
- * (reacquiring leases will fail if another host has acquired them since they
- * were released.)  This complexity of this option, combined with the fact that
- * the error conditions are often not as simple as storage being lost and then
- * later connecting, will result in this option being too unreliable.
- *
- * Add a config option that we could use to select a different behavior than
- * the default.  Then implement one of the simpler options as a proof of
- * concept, which could be extended if needed.
- */
+-------------------------------------------------------------------------------
+For each VG, lvmlockd creates a sanlock lockspace that holds the leases for
+that VG.  There's a lease for the VG lock, and there's a lease for each active
+LV.  sanlock maintains (reads/writes) these leases, which exist on storage.
+That storage is a hidden LV within the VG: /dev/vg/lvmlock.  lvmlockd gives the
+path of this internal LV to sanlock, which then reads/writes the leases on it.
+
+# lvs -a cc -o+uuid
+  LV        VG   Attr       LSize   LV UUID
+  lv1       cc   -wi-a-----   2.00g 7xoDtu-yvNM-iwQx-C94t-BbYs-UzBl-o8hAIa
+  lv2       cc   -wi-a----- 100.00g exxNPX-wZdO-uCNy-yiGa-aJGT-JKVl-arfcYT
+  [lvmlock] cc   -wi-ao---- 256.00m iLpDel-hR0T-hJ3u-rnVo-PcDh-mcjt-sF9egM
+
+# sanlock status
+s lvm_cc:1:/dev/mapper/cc-lvmlock:0
+r lvm_cc:exxNPX-wZdO-uCNy-yiGa-aJGT-JKVl-arfcYT:/dev/mapper/cc-lvmlock:71303168:13 p 26099
+r lvm_cc:7xoDtu-yvNM-iwQx-C94t-BbYs-UzBl-o8hAIa:/dev/mapper/cc-lvmlock:70254592:3 p 26099
+
+This shows that sanlock is maintaining leases on /dev/mapper/cc-lvmlock.
+
+sanlock acquires a lockspace lease when the lockspace is joined, i.e. when the
+VG is started by 'vgchange --lock-start cc'.  This lockspace lease exists at
+/dev/mapper/cc-lvmlock offset 0, and sanlock regularly writes to it to maintain
+ownership of it.  Joining the lockspace (by acquiring the lockspace lease in
+it) then allows standard resource leases to be acquired in the lockspace for
+whatever the application wants.  lvmlockd uses resource leases for the VG lock
+and LV locks.
+
+sanlock acquires a resource lease for each actual lock that lvm commands use.
+Above, there are two LV locks that are held because the two LVs are active.
+These are on /dev/mapper/cc-lvmlock at offsets 71303168 and 70254592.  sanlock
+does not write to these resource leases except when acquiring and releasing
+them (e.g. lvchange -ay/-an).  The renewal of the lockspace lease maintains
+ownership of all the resource leases in the lockspace.
+
+If the host loses access to the disk that the sanlock lv lives on, then sanlock
+can no longer renew its lockspace lease.  The lockspace lease will eventually
+expire, at which point the host will lose ownership of it, and of all resource
+leases it holds in the lockspace.  Eventually, other hosts will be able to
+acquire those leases.  sanlock ensures that another host will not be able to
+acquire one of the expired leases until the current host has quit using it.
+
+It is important that the host "quit using" the leases it is holding if the
+sanlock storage is lost and they begin expiring.  If the host cannot quit using
+the leases and release them within a limited time, then sanlock will use the
+local watchdog to forcibly reset the host before any other host can acquire
+them.  This is severe, but preferable to possibly corrupting the data protected
+by the lease.  It ensures that two nodes will not be using the same lease at
+once.  For LV leases, that means that another host will not be able to activate
+the LV while another host still has it active.
+
+sanlock notifies the application that it cannot renew the lockspace lease.  The
+application needs to quit using all leases in the lockspace and release them as
+quickly as possible.  In the initial version, lvmlockd ignored this
+notification, so sanlock would eventually reach the point where it would use
+the local watchdog to reset the host.  However, it's better to attempt a
+response.  If that response succeeds, the host can avoid being reset.  If the
+response fails, then sanlock will eventually reset the host as the last resort.
+sanlock gives the application about 40 seconds to complete its response and
+release its leases before resetting the host.
+
+An application can specify the path and args of a program that sanlock should
+run to notify it if the lockspace lease cannot be renewed.  This program should
+carry out the application's response to the expiring leases: attempt to quit
+using the leases and then release them.  lvmlockd gives this command to sanlock
+for each VG when that VG is started: 'lvmlockctl --kill vg_name'
+
+If sanlock loses access to lease storage in that VG, it runs lvmlockctl --kill,
+which:
+
+1. Uses syslog to explain what is happening.
+
+2. Notifies lvmlockd that the VG is being killed, so lvmlockd can
+   immediatley return an error for this condition if any new lock
+   requests are made.  (This step would not be strictly necessary.)
+
+3. Attempts to quit using the VG.  This is not yet implemented, but
+   will eventually use blkdeactivate on the VG (or a more forceful
+   equivalent.)
+
+4. If step 3 was successful at terminating all use of the VG, then
+   lvmlockd is told to release all the leases for the VG.  If this
+   is all done without about 40 seconds, the host can avoid being
+   reset.
+
+Until steps 3 and 4 are fully implemented, manual steps can be substituted.
+This is primarily for testing since the problem needs to be noticed and
+responded to in a very short time.  The manual alternative to step 3 is to kill
+any processes using file systems on LV's in the VG, unmount all file systems on
+the LVs, and deactivate all the LVs.  Once this is done, the manual alternative
+to step 4 is to run 'lvmlockctl --drop vg_name', which tells lvmlockd to
+release all the leases for the VG.
+-------------------------------------------------------------------------------
+*/
+
 
 /*
  * Each lockspace thread has its own sanlock daemon connection.
@@ -174,7 +216,7 @@ static int lock_lv_name_from_args(char *vg_args, char *lock_lv_name)
 
 static int lock_lv_offset_from_args(char *lv_args, uint64_t *lock_lv_offset)
 {
-	char offset_str[MAX_ARGS];
+	char offset_str[MAX_ARGS+1];
 	int rv;
 
 	memset(offset_str, 0, sizeof(offset_str));
@@ -243,7 +285,8 @@ static int read_host_id_file(void)
 			break;
 		}
 	}
-	fclose(file);
+	if (fclose(file))
+		log_error("failed to close host id file %s", daemon_host_id_file);
 out:
 	log_debug("host_id %d from %s", host_id, daemon_host_id_file);
 	return host_id;
@@ -262,8 +305,8 @@ int lm_init_vg_sanlock(char *ls_name, char *vg_name, uint32_t flags, char *vg_ar
 	struct sanlk_lockspace ss;
 	struct sanlk_resourced rd;
 	struct sanlk_disk disk;
-	char lock_lv_name[MAX_ARGS];
-	char lock_args_version[MAX_ARGS];
+	char lock_lv_name[MAX_ARGS+1];
+	char lock_args_version[MAX_ARGS+1];
 	const char *gl_name = NULL;
 	uint32_t daemon_version;
 	uint32_t daemon_proto;
@@ -291,7 +334,7 @@ int lm_init_vg_sanlock(char *ls_name, char *vg_name, uint32_t flags, char *vg_ar
 	if (strlen(lock_lv_name) + strlen(lock_args_version) + 2 > MAX_ARGS)
 		return -EARGS;
 
-	snprintf(disk.path, SANLK_PATH_LEN, "/dev/mapper/%s-%s", vg_name, lock_lv_name);
+	snprintf(disk.path, SANLK_PATH_LEN-1, "/dev/mapper/%s-%s", vg_name, lock_lv_name);
 
 	log_debug("S %s init_vg_san path %s", ls_name, disk.path);
 
@@ -429,8 +472,8 @@ int lm_init_lv_sanlock(char *ls_name, char *vg_name, char *lv_name,
 		       char *vg_args, char *lv_args, uint64_t free_offset)
 {
 	struct sanlk_resourced rd;
-	char lock_lv_name[MAX_ARGS];
-	char lock_args_version[MAX_ARGS];
+	char lock_lv_name[MAX_ARGS+1];
+	char lock_args_version[MAX_ARGS+1];
 	uint64_t offset;
 	int align_size;
 	int rv;
@@ -451,7 +494,7 @@ int lm_init_lv_sanlock(char *ls_name, char *vg_name, char *lv_name,
 
 	strncpy(rd.rs.lockspace_name, ls_name, SANLK_NAME_LEN);
 	rd.rs.num_disks = 1;
-	snprintf(rd.rs.disks[0].path, SANLK_PATH_LEN, "/dev/mapper/%s-%s", vg_name, lock_lv_name);
+	snprintf(rd.rs.disks[0].path, SANLK_PATH_LEN-1, "/dev/mapper/%s-%s", vg_name, lock_lv_name);
 
 	align_size = sanlock_align(&rd.rs.disks[0]);
 	if (align_size <= 0) {
@@ -535,7 +578,7 @@ int lm_rename_vg_sanlock(char *ls_name, char *vg_name, uint32_t flags, char *vg_
 	struct sanlk_lockspace ss;
 	struct sanlk_resourced rd;
 	struct sanlk_disk disk;
-	char lock_lv_name[MAX_ARGS];
+	char lock_lv_name[MAX_ARGS+1];
 	uint64_t offset;
 	uint32_t io_timeout;
 	int align_size;
@@ -556,7 +599,7 @@ int lm_rename_vg_sanlock(char *ls_name, char *vg_name, uint32_t flags, char *vg_
 		return rv;
 	}
 
-	snprintf(disk.path, SANLK_PATH_LEN, "/dev/mapper/%s-%s", vg_name, lock_lv_name);
+	snprintf(disk.path, SANLK_PATH_LEN-1, "/dev/mapper/%s-%s", vg_name, lock_lv_name);
 
 	log_debug("S %s rename_vg_san path %s", ls_name, disk.path);
 
@@ -736,7 +779,7 @@ int lm_ex_disable_gl_sanlock(struct lockspace *ls)
 	strncpy(rd2.rs.name, R_NAME_GL_DISABLED, SANLK_NAME_LEN);
 
 	rd1.rs.num_disks = 1;
-	strncpy(rd1.rs.disks[0].path, lms->ss.host_id_disk.path, SANLK_PATH_LEN);
+	strncpy(rd1.rs.disks[0].path, lms->ss.host_id_disk.path, SANLK_PATH_LEN-1);
 	rd1.rs.disks[0].offset = lms->align_size * GL_LOCK_BEGIN;
 
 	rv = sanlock_acquire(lms->sock, -1, 0, 1, &rs1, NULL);
@@ -794,7 +837,7 @@ int lm_able_gl_sanlock(struct lockspace *ls, int enable)
 	strncpy(rd.rs.name, gl_name, SANLK_NAME_LEN);
 
 	rd.rs.num_disks = 1;
-	strncpy(rd.rs.disks[0].path, lms->ss.host_id_disk.path, SANLK_PATH_LEN);
+	strncpy(rd.rs.disks[0].path, lms->ss.host_id_disk.path, SANLK_PATH_LEN-1);
 	rd.rs.disks[0].offset = lms->align_size * GL_LOCK_BEGIN;
 
 	rv = sanlock_write_resource(&rd.rs, 0, 0, 0);
@@ -807,8 +850,6 @@ int lm_able_gl_sanlock(struct lockspace *ls, int enable)
 	log_debug("S %s able_gl %s", ls->name, gl_name);
 
 	ls->sanlock_gl_enabled = enable;
-	if (ls->sanlock_gl_dup && !enable)
-		ls->sanlock_gl_dup = 0;
 
 	if (enable)
 		strncpy(gl_lsname_sanlock, ls->name, MAX_NAME);
@@ -833,7 +874,7 @@ static int gl_is_enabled(struct lockspace *ls, struct lm_sanlock *lms)
 	/* leave rs.name empty, it is what we're checking */
 
 	rd.rs.num_disks = 1;
-	strncpy(rd.rs.disks[0].path, lms->ss.host_id_disk.path, SANLK_PATH_LEN);
+	strncpy(rd.rs.disks[0].path, lms->ss.host_id_disk.path, SANLK_PATH_LEN-1);
 
 	offset = lms->align_size * GL_LOCK_BEGIN;
 	rd.rs.disks[0].offset = offset;
@@ -891,7 +932,7 @@ int lm_find_free_lock_sanlock(struct lockspace *ls, uint64_t *free_offset)
 
 	strncpy(rd.rs.lockspace_name, ls->name, SANLK_NAME_LEN);
 	rd.rs.num_disks = 1;
-	strncpy(rd.rs.disks[0].path, lms->ss.host_id_disk.path, SANLK_PATH_LEN);
+	strncpy(rd.rs.disks[0].path, lms->ss.host_id_disk.path, SANLK_PATH_LEN-1);
 
 	offset = lms->align_size * LV_LOCK_BEGIN;
 
@@ -966,15 +1007,27 @@ int lm_prepare_lockspace_sanlock(struct lockspace *ls)
 {
 	struct stat st;
 	struct lm_sanlock *lms = NULL;
-	char lock_lv_name[MAX_ARGS];
+	char lock_lv_name[MAX_ARGS+1];
 	char lsname[SANLK_NAME_LEN + 1];
 	char disk_path[SANLK_PATH_LEN];
+	char killpath[SANLK_PATH_LEN];
+	char killargs[SANLK_PATH_LEN];
 	int gl_found;
 	int ret, rv;
 
 	memset(disk_path, 0, sizeof(disk_path));
 	memset(lock_lv_name, 0, sizeof(lock_lv_name));
 
+	/*
+	 * Construct the path to lvmlockctl by using the path to the lvm binary
+	 * and appending "lockctl" to get /path/to/lvmlockctl.
+	 */
+	memset(killpath, 0, sizeof(killpath));
+	snprintf(killpath, SANLK_PATH_LEN - 1, "%slockctl", LVM_PATH);
+
+	memset(killargs, 0, sizeof(killargs));
+	snprintf(killargs, SANLK_PATH_LEN - 1, "--kill %s", ls->vg_name);
+
 	rv = check_args_version(ls->vg_args, VG_LOCK_ARGS_MAJOR);
 	if (rv < 0) {
 		ret = -EARGS;
@@ -989,7 +1042,7 @@ int lm_prepare_lockspace_sanlock(struct lockspace *ls)
 		goto fail;
 	}
 
-	snprintf(disk_path, SANLK_PATH_LEN, "/dev/mapper/%s-%s",
+	snprintf(disk_path, SANLK_PATH_LEN-1, "/dev/mapper/%s-%s",
 		 ls->vg_name, lock_lv_name);
 
 	/*
@@ -1041,7 +1094,7 @@ int lm_prepare_lockspace_sanlock(struct lockspace *ls)
 	memcpy(lms->ss.name, lsname, SANLK_NAME_LEN);
 	lms->ss.host_id_disk.offset = 0;
 	lms->ss.host_id = ls->host_id;
-	strncpy(lms->ss.host_id_disk.path, disk_path, SANLK_PATH_LEN);
+	strncpy(lms->ss.host_id_disk.path, disk_path, SANLK_PATH_LEN-1);
 
 	if (daemon_test) {
 		if (!gl_lsname_sanlock[0]) {
@@ -1059,6 +1112,15 @@ int lm_prepare_lockspace_sanlock(struct lockspace *ls)
 		goto fail;
 	}
 
+	log_debug("set killpath to %s %s", killpath, killargs);
+
+	rv = sanlock_killpath(lms->sock, 0, killpath, killargs);
+	if (rv < 0) {
+		log_error("S %s killpath error %d", lsname, rv);
+		ret = -EMANAGER;
+		goto fail;
+	}
+
 	rv = sanlock_restrict(lms->sock, SANLK_RESTRICT_SIGKILL);
 	if (rv < 0) {
 		log_error("S %s restrict error %d", lsname, rv);
@@ -1145,7 +1207,8 @@ out:
 	return 0;
 
 fail:
-	close(lms->sock);
+	if (close(lms->sock))
+		log_error("failed to close sanlock daemon socket connection");
 	free(lms);
 	ls->lm_data = NULL;
 	return rv;
@@ -1181,7 +1244,8 @@ int lm_rem_lockspace_sanlock(struct lockspace *ls, int free_vg)
 		}
 	}
 out:
-	close(lms->sock);
+	if (close(lms->sock))
+		log_error("failed to close sanlock daemon socket connection");
 
 	free(lms);
 	ls->lm_data = NULL;
@@ -1236,7 +1300,7 @@ int lm_rem_resource_sanlock(struct lockspace *ls, struct resource *r)
 }
 
 int lm_lock_sanlock(struct lockspace *ls, struct resource *r, int ld_mode,
-		    uint32_t *r_version, int *retry, int adopt)
+		    struct val_blk *vb_out, int *retry, int adopt)
 {
 	struct lm_sanlock *lms = (struct lm_sanlock *)ls->lm_data;
 	struct rd_sanlock *rds = (struct rd_sanlock *)r->lm_data;
@@ -1244,7 +1308,6 @@ int lm_lock_sanlock(struct lockspace *ls, struct resource *r, int ld_mode,
 	uint64_t lock_lv_offset;
 	uint32_t flags = 0;
 	struct val_blk vb;
-	uint16_t vb_version;
 	int added = 0;
 	int rv;
 
@@ -1258,6 +1321,14 @@ int lm_lock_sanlock(struct lockspace *ls, struct resource *r, int ld_mode,
 
 	rs = &rds->rs;
 
+	/*
+	 * While there are duplicate global locks, keep checking
+	 * to see if any have been disabled.
+	 */
+	if (sanlock_gl_dup && ls->sanlock_gl_enabled &&
+	    (r->type == LD_RT_GL || r->type == LD_RT_VG))
+		ls->sanlock_gl_enabled = gl_is_enabled(ls, ls->lm_data);
+
 	if (r->type == LD_RT_LV) {
 		/*
 		 * The lv may have been removed and recreated with a new lease
@@ -1312,7 +1383,7 @@ int lm_lock_sanlock(struct lockspace *ls, struct resource *r, int ld_mode,
 		  (unsigned long long)rs->disks[0].offset);
 
 	if (daemon_test) {
-		*r_version = 0;
+		memset(vb_out, 0, sizeof(struct val_blk));
 		return 0;
 	}
 
@@ -1321,6 +1392,15 @@ int lm_lock_sanlock(struct lockspace *ls, struct resource *r, int ld_mode,
 	if (adopt)
 		flags |= SANLK_ACQUIRE_ORPHAN_ONLY;
 
+#ifdef SANLOCK_HAS_ACQUIRE_OWNER_NOWAIT
+	/*
+	 * Don't block waiting for a failed lease to expire since it causes
+	 * sanlock_acquire to block for a long time, which would prevent this
+	 * thread from processing other lock requests.
+	 */
+	flags |= SANLK_ACQUIRE_OWNER_NOWAIT;
+#endif
+
 	rv = sanlock_acquire(lms->sock, -1, flags, 1, &rs, NULL);
 
 	if (rv == -EAGAIN) {
@@ -1391,15 +1471,30 @@ int lm_lock_sanlock(struct lockspace *ls, struct resource *r, int ld_mode,
 		return -EAGAIN;
 	}
 
+#ifdef SANLOCK_HAS_ACQUIRE_OWNER_NOWAIT
+	if (rv == SANLK_ACQUIRE_OWNED_RETRY) {
+		/*
+		 * The lock is held by a failed host, and will eventually
+		 * expire.  If we retry we'll eventually acquire the lock
+		 * (or find someone else has acquired it).  The EAGAIN retry
+		 * attempts for SH locks above would not be sufficient for
+		 * the length of expiration time.  We could add a longer
+		 * retry time here to cover the full expiration time and block
+		 * the activation command for that long.  For now just return
+		 * the standard error indicating that another host still owns
+		 * the lease.  FIXME: return a different error number so the
+		 * command can print an different error indicating that the
+		 * owner of the lease is in the process of expiring?
+		 */
+		log_debug("S %s R %s lock_san acquire mode %d rv %d", ls->name, r->name, ld_mode, rv);
+		*retry = 0;
+		return -EAGAIN;
+	}
+#endif
 	if (rv < 0) {
 		log_error("S %s R %s lock_san acquire error %d",
 			  ls->name, r->name, rv);
 
-		if (added) {
-			lm_rem_resource_sanlock(ls, r);
-			return rv;
-		}
-
 		/* if the gl has been disabled, remove and free the gl resource */
 		if ((rv == SANLK_LEADER_RESOURCE) && (r->type == LD_RT_GL)) {
 			if (!lm_gl_is_enabled(ls)) {
@@ -1411,6 +1506,22 @@ int lm_lock_sanlock(struct lockspace *ls, struct resource *r, int ld_mode,
 			}
 		}
 
+		if (added)
+			lm_rem_resource_sanlock(ls, r);
+
+		/* sanlock gets i/o errors trying to read/write the leases. */
+		if (rv == -EIO)
+			rv = -ELOCKIO;
+
+		/*
+		 * The sanlock lockspace can disappear if the lease storage fails,
+		 * the delta lease renewals fail, the lockspace enters recovery,
+		 * lvmlockd holds no leases in the lockspace, so sanlock can
+		 * stop and free the lockspace.
+		 */
+		if (rv == -ENOSPC)
+			rv = -ELOCKIO;
+
 		return rv;
 	}
 
@@ -1418,26 +1529,23 @@ int lm_lock_sanlock(struct lockspace *ls, struct resource *r, int ld_mode,
 		rv = sanlock_get_lvb(0, rs, (char *)&vb, sizeof(vb));
 		if (rv < 0) {
 			log_error("S %s R %s lock_san get_lvb error %d", ls->name, r->name, rv);
-			*r_version = 0;
+			memset(rds->vb, 0, sizeof(struct val_blk));
+			memset(vb_out, 0, sizeof(struct val_blk));
 			goto out;
 		}
 
-		vb_version = le16_to_cpu(vb.version);
+		/*
+		 * 'vb' contains disk endian values, not host endian.
+		 * It is copied directly to rrs->vb which is also kept
+		 * in disk endian form.
+		 * vb_out is returned to the caller in host endian form.
+		 */
 
-		if (vb_version && ((vb_version & 0xFF00) > (VAL_BLK_VERSION & 0xFF00))) {
-			log_error("S %s R %s lock_san ignore vb_version %x",
-				  ls->name, r->name, vb_version);
-			*r_version = 0;
-			free(rds->vb);
-			rds->vb = NULL;
-			goto out;
-		}
+		memcpy(rds->vb, &vb, sizeof(vb));
 
-		*r_version = le32_to_cpu(vb.r_version);
-		memcpy(rds->vb, &vb, sizeof(vb)); /* rds->vb saved as le */
-
-		log_debug("S %s R %s lock_san get r_version %u",
-			  ls->name, r->name, *r_version);
+		vb_out->version = le16_to_cpu(vb.version);
+		vb_out->flags = le16_to_cpu(vb.flags);
+		vb_out->r_version = le32_to_cpu(vb.r_version);
 	}
 out:
 	return rv;
@@ -1592,9 +1700,11 @@ int lm_unlock_sanlock(struct lockspace *ls, struct resource *r,
 	}
 
 	rv = sanlock_release(lms->sock, -1, 0, 1, &rs);
-	if (rv < 0) {
+	if (rv < 0)
 		log_error("S %s R %s unlock_san release error %d", ls->name, r->name, rv);
-	}
+
+	if (rv == -EIO)
+		rv = -ELOCKIO;
 
 	return rv;
 }

daemons/lvmpolld/.gitignore

@@ -0,0 +1 @@
+lvmpolld

daemons/lvmpolld/lvmpolld-common.h

@@ -18,19 +18,14 @@
 #ifndef _LVM_LVMPOLLD_COMMON_H
 #define _LVM_LVMPOLLD_COMMON_H
 
-#include "configure.h"
-
 #define _REENTRANT
-#define _GNU_SOURCE
-#define _FILE_OFFSET_BITS 64
 
-#include "libdevmapper.h"
+#include "tool.h"
 
 #include "lvmpolld-cmd-utils.h"
 #include "lvmpolld-protocol.h"
 
 #include <assert.h>
 #include <errno.h>
-#include <unistd.h>
 
 #endif	/* _LVM_LVMPOLLD_COMMON_H */

daemons/lvmpolld/lvmpolld-core.c

@@ -83,6 +83,12 @@ static int _init(struct daemon_state *s)
 	struct lvmpolld_state *ls = s->private;
 	ls->log = s->log;
 
+	/*
+	 * log warnings to stderr by default. Otherwise we would miss any lvpoll
+	 * error messages in default configuration
+	 */
+	daemon_log_enable(ls->log, DAEMON_LOG_OUTLET_STDERR, DAEMON_LOG_WARN,  1);
+
 	if (!daemon_log_parse(ls->log, DAEMON_LOG_OUTLET_STDERR, ls->log_config, 1))
 		return 0;
 
@@ -284,7 +290,7 @@ static int poll_for_output(struct lvmpolld_lv *pdlv, struct lvmpolld_thread_data
 				 "caught input data in STDERR");
 
 			assert(read_single_line(data, 1)); /* may block indef. anyway */
-			INFO(pdlv->ls, "%s: PID %d: %s: '%s'", LVM2_LOG_PREFIX,
+			WARN(pdlv->ls, "%s: PID %d: %s: '%s'", LVM2_LOG_PREFIX,
 			     pdlv->cmd_pid, "STDERR", data->line);
 		} else if (fds[1].revents) {
 			if (fds[1].revents & POLLHUP)
@@ -327,15 +333,19 @@ static int poll_for_output(struct lvmpolld_lv *pdlv, struct lvmpolld_thread_data
 	if (fds[1].fd >= 0)
 		while (read_single_line(data, 1)) {
 			assert(r > 0);
-			INFO(pdlv->ls, "%s: PID %d: %s: %s", LVM2_LOG_PREFIX, pdlv->cmd_pid, "STDERR", data->line);
+			WARN(pdlv->ls, "%s: PID %d: %s: %s", LVM2_LOG_PREFIX, pdlv->cmd_pid, "STDERR", data->line);
 		}
 
 	if (WIFEXITED(ch_stat)) {
-		INFO(pdlv->ls, "%s: %s (PID %d) %s (%d)", PD_LOG_PREFIX,
-		     "lvm2 cmd", pdlv->cmd_pid, "exited with", WEXITSTATUS(ch_stat));
 		cmd_state.retcode = WEXITSTATUS(ch_stat);
+		if (cmd_state.retcode)
+			ERROR(pdlv->ls, "%s: %s (PID %d) %s (retcode: %d)", PD_LOG_PREFIX,
+			     "lvm2 cmd", pdlv->cmd_pid, "failed", cmd_state.retcode);
+		else
+			INFO(pdlv->ls, "%s: %s (PID %d) %s", PD_LOG_PREFIX,
+			     "lvm2 cmd", pdlv->cmd_pid, "finished successfully");
 	} else if (WIFSIGNALED(ch_stat)) {
-		WARN(pdlv->ls, "%s: %s (PID %d) %s (%d)", PD_LOG_PREFIX,
+		ERROR(pdlv->ls, "%s: %s (PID %d) %s (%d)", PD_LOG_PREFIX,
 		     "lvm2 cmd", pdlv->cmd_pid, "got terminated by signal",
 		     WTERMSIG(ch_stat));
 		cmd_state.signal = WTERMSIG(ch_stat);
@@ -529,7 +539,7 @@ static response progress_info(client_handle h, struct lvmpolld_state *ls, reques
 		if (st.polling_finished)
 			r = daemon_reply_simple(LVMPD_RESP_FINISHED,
 						"reason = %s", st.cmd_state.signal ? LVMPD_REAS_SIGNAL : LVMPD_REAS_RETCODE,
-						LVMPD_PARM_VALUE " = %d", st.cmd_state.signal ?: st.cmd_state.retcode,
+						LVMPD_PARM_VALUE " = %d", (int64_t)(st.cmd_state.signal ?: st.cmd_state.retcode),
 						NULL);
 		else
 			r = daemon_reply_simple(LVMPD_RESP_IN_PROGRESS, NULL);
@@ -566,6 +576,8 @@ static struct lvmpolld_lv *construct_pdlv(request req, struct lvmpolld_state *ls
 		return NULL;
 	}
 
+	pdlv->cmdargv = cmdargv;
+
 	cmdenvp = cmdenvp_ctr(pdlv);
 	if (!cmdenvp) {
 		pdlv_destroy(pdlv);
@@ -573,7 +585,6 @@ static struct lvmpolld_lv *construct_pdlv(request req, struct lvmpolld_state *ls
 		return NULL;
 	}
 
-	pdlv->cmdargv = cmdargv;
 	pdlv->cmdenvp = cmdenvp;
 
 	return pdlv;
@@ -584,12 +595,16 @@ static int spawn_detached_thread(struct lvmpolld_lv *pdlv)
 	int r;
 	pthread_attr_t attr;
 
-	pthread_attr_init(&attr);
-	pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_DETACHED);
+	if (pthread_attr_init(&attr) != 0)
+		return 0;
+
+	if (pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_DETACHED) != 0)
+		return 0;
 
 	r = pthread_create(&pdlv->tid, &attr, fork_and_poll, (void *)pdlv);
 
-	pthread_attr_destroy(&attr);
+	if (pthread_attr_destroy(&attr) != 0)
+		return 0;
 
 	return !r;
 }

daemons/lvmpolld/lvmpolld-data-utils.c

@@ -325,6 +325,7 @@ struct lvmpolld_thread_data *lvmpolld_thread_data_constructor(struct lvmpolld_lv
 
 	data->pdlv = NULL;
 	data->line = NULL;
+	data->line_size = 0;
 	data->fout = data->ferr = NULL;
 	data->outpipe[0] = data->outpipe[1] = data->errpipe[0] = data->errpipe[1] = -1;
 
@@ -365,7 +366,8 @@ void lvmpolld_thread_data_destroy(void *thread_private)
 		pdst_unlock(data->pdlv->pdst);
 	}
 
-	dm_free(data->line);
+	/* may get reallocated in getline(). dm_free must not be used */
+	free(data->line);
 
 	if (data->fout && !fclose(data->fout))
 		data->outpipe[0] = -1;
@@ -374,16 +376,16 @@ void lvmpolld_thread_data_destroy(void *thread_private)
 		data->errpipe[0] = -1;
 
 	if (data->outpipe[0] >= 0)
-		close(data->outpipe[0]);
+		(void) close(data->outpipe[0]);
 
 	if (data->outpipe[1] >= 0)
-		close(data->outpipe[1]);
+		(void) close(data->outpipe[1]);
 
 	if (data->errpipe[0] >= 0)
-		close(data->errpipe[0]);
+		(void) close(data->errpipe[0]);
 
 	if (data->errpipe[1] >= 0)
-		close(data->errpipe[1]);
+		(void) close(data->errpipe[1]);
 
 	dm_free(data);
 }

include/.symlinks.in

@@ -45,7 +45,6 @@
 @top_srcdir@/lib/metadata/vg.h
 @top_srcdir@/lib/mm/memlock.h
 @top_srcdir@/lib/mm/xlate.h
-@top_builddir@/lib/misc/configure.h
 @top_srcdir@/lib/misc/crc.h
 @top_srcdir@/lib/misc/intl.h
 @top_srcdir@/lib/misc/last-path-component.h
@@ -56,7 +55,6 @@
 @top_srcdir@/lib/misc/lvm-globals.h
 @top_srcdir@/lib/misc/lvm-signal.h
 @top_srcdir@/lib/misc/lvm-string.h
-@top_builddir@/lib/misc/lvm-version.h
 @top_srcdir@/lib/misc/lvm-percent.h
 @top_srcdir@/lib/misc/lvm-wrappers.h
 @top_srcdir@/lib/misc/sharedlib.h
@@ -76,3 +74,4 @@
 @top_srcdir@/libdm/misc/kdev_t.h
 @top_srcdir@/po/pogen.h
 @top_srcdir@/tools/lvm2cmd.h
+@top_srcdir@/tools/tool.h

include/Makefile.in

@@ -1,6 +1,6 @@
 #
 # Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.
-# Copyright (C) 2004-2010 Red Hat, Inc. All rights reserved.
+# Copyright (C) 2004-2015 Red Hat, Inc. All rights reserved.
 #
 # This file is part of LVM2.
 #
@@ -35,5 +35,5 @@ device-mapper: all
 
 cflow: all
 
-DISTCLEAN_TARGETS += .symlinks
+DISTCLEAN_TARGETS += .symlinks configure.h lvm-version.h
 CLEAN_TARGETS += $(LINKS) .include_symlinks .symlinks_created

include/configure.h.in

@@ -1,4 +1,4 @@
-/* lib/misc/configure.h.in.  Generated from configure.in by autoheader.  */
+/* include/configure.h.in.  Generated from configure.in by autoheader.  */
 
 /* Define to 1 to use libblkid detection of signatures when wiping. */
 #undef BLKID_WIPING_SUPPORT
@@ -6,6 +6,10 @@
 /* The path to 'cache_check', if available. */
 #undef CACHE_CHECK_CMD
 
+/* Define to 1 if the external 'cache_check' tool requires the
+   --clear-needs-check-flag option */
+#undef CACHE_CHECK_NEEDS_CHECK
+
 /* The path to 'cache_dump', if available. */
 #undef CACHE_DUMP_CMD
 
@@ -141,6 +145,9 @@
 /* Library version */
 #undef DM_LIB_VERSION
 
+/* Define to 1 if you have the `alarm' function. */
+#undef HAVE_ALARM
+
 /* Define to 1 if you have `alloca', as a function or macro. */
 #undef HAVE_ALLOCA
 
@@ -157,12 +164,18 @@
 /* Define to 1 if you have the <assert.h> header file. */
 #undef HAVE_ASSERT_H
 
+/* Define to 1 if you have the `atexit' function. */
+#undef HAVE_ATEXIT
+
 /* Define to 1 if canonicalize_file_name is available. */
 #undef HAVE_CANONICALIZE_FILE_NAME
 
 /* Define to 1 if your system has a working `chown' function. */
 #undef HAVE_CHOWN
 
+/* Define to 1 if you have the `clock_gettime' function. */
+#undef HAVE_CLOCK_GETTIME
+
 /* Define to 1 if you have the <corosync/cmap.h> header file. */
 #undef HAVE_COROSYNC_CMAP_H
 
@@ -172,6 +185,10 @@
 /* Define to 1 if you have the <ctype.h> header file. */
 #undef HAVE_CTYPE_H
 
+/* Define to 1 if you have the declaration of `strerror_r', and to 0 if you
+   don't. */
+#undef HAVE_DECL_STRERROR_R
+
 /* Define to 1 if you have the <dirent.h> header file. */
 #undef HAVE_DIRENT_H
 
@@ -190,6 +207,9 @@
 /* Define to 1 if you have the <fcntl.h> header file. */
 #undef HAVE_FCNTL_H
 
+/* Define to 1 if you have the <float.h> header file. */
+#undef HAVE_FLOAT_H
+
 /* Define to 1 if you have the `fork' function. */
 #undef HAVE_FORK
 
@@ -247,6 +267,9 @@
 /* Define to 1 if you have the <locale.h> header file. */
 #undef HAVE_LOCALE_H
 
+/* Define to 1 if you have the `localtime_r' function. */
+#undef HAVE_LOCALTIME_R
+
 /* Define to 1 if `lstat' has the bug that it succeeds when given the
    zero-length file name argument. */
 #undef HAVE_LSTAT_EMPTY_STRING_BUG
@@ -261,6 +284,9 @@
 /* Define to 1 if you have the <malloc.h> header file. */
 #undef HAVE_MALLOC_H
 
+/* Define to 1 if you have the `memchr' function. */
+#undef HAVE_MEMCHR
+
 /* Define to 1 if you have the `memmove' function. */
 #undef HAVE_MEMMOVE
 
@@ -297,9 +323,15 @@
 /* Define to 1 if you have the `nl_langinfo' function. */
 #undef HAVE_NL_LANGINFO
 
+/* Define to 1 if you have the <paths.h> header file. */
+#undef HAVE_PATHS_H
+
 /* Define to 1 if you have the <pthread.h> header file. */
 #undef HAVE_PTHREAD_H
 
+/* Define to 1 if the system has the type `ptrdiff_t'. */
+#undef HAVE_PTRDIFF_T
+
 /* Define to 1 if you have the <readline/history.h> header file. */
 #undef HAVE_READLINE_HISTORY_H
 
@@ -310,6 +342,9 @@
    and to 0 otherwise. */
 #undef HAVE_REALLOC
 
+/* Define to 1 if you have the `realpath' function. */
+#undef HAVE_REALPATH
+
 /* Define to 1 to include support for realtime clock. */
 #undef HAVE_REALTIME
 
@@ -359,6 +394,9 @@
 /* Define to 1 if you have the <stdarg.h> header file. */
 #undef HAVE_STDARG_H
 
+/* Define to 1 if stdbool.h conforms to C99. */
+#undef HAVE_STDBOOL_H
+
 /* Define to 1 if you have the <stddef.h> header file. */
 #undef HAVE_STDDEF_H
 
@@ -386,6 +424,9 @@
 /* Define to 1 if you have the `strerror' function. */
 #undef HAVE_STRERROR
 
+/* Define to 1 if you have the `strerror_r' function. */
+#undef HAVE_STRERROR_R
+
 /* Define to 1 if you have the <strings.h> header file. */
 #undef HAVE_STRINGS_H
 
@@ -395,6 +436,12 @@
 /* Define to 1 if you have the `strncasecmp' function. */
 #undef HAVE_STRNCASECMP
 
+/* Define to 1 if you have the `strndup' function. */
+#undef HAVE_STRNDUP
+
+/* Define to 1 if you have the `strpbrk' function. */
+#undef HAVE_STRPBRK
+
 /* Define to 1 if you have the `strrchr' function. */
 #undef HAVE_STRRCHR
 
@@ -410,6 +457,9 @@
 /* Define to 1 if you have the `strtoul' function. */
 #undef HAVE_STRTOUL
 
+/* Define to 1 if you have the `strtoull' function. */
+#undef HAVE_STRTOULL
+
 /* Define to 1 if `st_rdev' is a member of `struct stat'. */
 #undef HAVE_STRUCT_STAT_ST_RDEV
 
@@ -463,6 +513,9 @@
 /* Define to 1 if you have the <sys/stat.h> header file. */
 #undef HAVE_SYS_STAT_H
 
+/* Define to 1 if you have the <sys/timerfd.h> header file. */
+#undef HAVE_SYS_TIMERFD_H
+
 /* Define to 1 if you have the <sys/time.h> header file. */
 #undef HAVE_SYS_TIME_H
 
@@ -514,12 +567,21 @@
 /* Define to 1 if `vfork' works. */
 #undef HAVE_WORKING_VFORK
 
+/* Define to 1 if the system has the type `_Bool'. */
+#undef HAVE__BOOL
+
 /* Internalization package */
 #undef INTL_PACKAGE
 
 /* Locale-dependent data */
 #undef LOCALEDIR
 
+/* Define to 1 to include code that uses lvmlockd dlm option. */
+#undef LOCKDDLM_SUPPORT
+
+/* Define to 1 to include code that uses lvmlockd sanlock option. */
+#undef LOCKDSANLOCK_SUPPORT
+
 /* Define to 1 if `lstat' dereferences a symlink specified with a trailing
    slash. */
 #undef LSTAT_FOLLOWS_SLASHED_SYMLINK
@@ -625,6 +687,12 @@
 /* Define to 1 if you have the ANSI C header files. */
 #undef STDC_HEADERS
 
+/* Define to 1 if strerror_r returns char *. */
+#undef STRERROR_R_CHAR_P
+
+/* Path to testsuite data */
+#undef TESTSUITE_DATA
+
 /* The path to 'thin_check', if available. */
 #undef THIN_CHECK_CMD
 

lib/Makefile.in

@@ -62,6 +62,7 @@ SOURCES =\
 	device/dev-swap.c \
 	device/dev-type.c \
 	device/dev-luks.c \
+	device/dev-dasd.c \
 	display/display.c \
 	error/errseg.c \
 	unknown/unknown.c \
@@ -122,11 +123,6 @@ SOURCES =\
 	uuid/uuid.c \
 	zero/zero.c
 
-ifeq ("@HAVE_REALTIME@", "yes")
-  SOURCES +=\
-	misc/timestamp.c
-endif
-
 ifeq ("@LVM1@", "internal")
   SOURCES +=\
 	format1/disk-rep.c \
@@ -235,4 +231,4 @@ CFLAGS += $(BLKID_CFLAGS) $(UDEV_CFLAGS) $(VALGRIND_CFLAGS)
 
 $(SUBDIRS): $(LIB_STATIC)
 
-DISTCLEAN_TARGETS += misc/configure.h misc/lvm-version.h
+CLEAN_TARGETS += misc/configure.h misc/lvm-version.h

lib/activate/activate.c

@@ -180,7 +180,7 @@ int lv_passes_auto_activation_filter(struct cmd_context *cmd, struct logical_vol
 {
 	const struct dm_config_node *cn;
 
-	if (!(cn = find_config_tree_node(cmd, activation_auto_activation_volume_list_CFG, NULL))) {
+	if (!(cn = find_config_tree_array(cmd, activation_auto_activation_volume_list_CFG, NULL))) {
 		log_verbose("activation/auto_activation_volume_list configuration setting "
 			    "not defined: All logical volumes will be auto-activated.");
 		return 1;
@@ -467,7 +467,7 @@ static int _passes_activation_filter(struct cmd_context *cmd,
 {
 	const struct dm_config_node *cn;
 
-	if (!(cn = find_config_tree_node(cmd, activation_volume_list_CFG, NULL))) {
+	if (!(cn = find_config_tree_array(cmd, activation_volume_list_CFG, NULL))) {
 		log_verbose("activation/volume_list configuration setting "
 			    "not defined: Checking only host tags for %s/%s",
 			    lv->vg->name, lv->name);
@@ -496,7 +496,7 @@ static int _passes_readonly_filter(struct cmd_context *cmd,
 {
 	const struct dm_config_node *cn;
 
-	if (!(cn = find_config_tree_node(cmd, activation_read_only_volume_list_CFG, NULL)))
+	if (!(cn = find_config_tree_array(cmd, activation_read_only_volume_list_CFG, NULL)))
 		return 0;
 
 	return _lv_passes_volumes_filter(cmd, lv, cn, activation_read_only_volume_list_CFG);

lib/activate/dev_manager.c

@@ -143,23 +143,23 @@ static int _get_segment_status_from_target_params(const char *target_name,
 		return 0;
 	}
 
-	if (!strcmp(segtype->name, "cache")) {
+	if (segtype_is_cache(segtype)) {
 		if (!dm_get_status_cache(seg_status->mem, params, &(seg_status->cache)))
 			return_0;
 		seg_status->type = SEG_STATUS_CACHE;
-	} else if (!strcmp(segtype->name, "raid")) {
+	} else if (segtype_is_raid(segtype)) {
 		if (!dm_get_status_raid(seg_status->mem, params, &seg_status->raid))
 			return_0;
 		seg_status->type = SEG_STATUS_RAID;
-	} else if (!strcmp(segtype->name, "thin")) {
+	} else if (segtype_is_thin_volume(segtype)) {
 		if (!dm_get_status_thin(seg_status->mem, params, &seg_status->thin))
 			return_0;
 		seg_status->type = SEG_STATUS_THIN;
-	} else if (!strcmp(segtype->name, "thin-pool")) {
+	} else if (segtype_is_thin_pool(segtype)) {
 		if (!dm_get_status_thin_pool(seg_status->mem, params, &seg_status->thin_pool))
 			return_0;
 		seg_status->type = SEG_STATUS_THIN_POOL;
-	} else if (!strcmp(segtype->name, "snapshot")) {
+	} else if (segtype_is_snapshot(segtype)) {
 		if (!dm_get_status_snapshot(seg_status->mem, params, &seg_status->snapshot))
 			return_0;
 		seg_status->type = SEG_STATUS_SNAPSHOT;
@@ -518,6 +518,73 @@ out:
 	return r;
 }
 
+static int _ignore_unusable_thins(struct device *dev)
+{
+	/* TODO make function for thin testing */
+	struct dm_pool *mem;
+	struct dm_status_thin_pool *status;
+	struct dm_task *dmt = NULL;
+	void *next = NULL;
+	uint64_t start, length;
+	char *target_type = NULL;
+	char *params;
+	int minor, major;
+	int r = 0;
+
+	if (!(mem = dm_pool_create("unusable_thins", 128)))
+		return_0;
+
+	if (!(dmt = dm_task_create(DM_DEVICE_TABLE)))
+		goto_out;
+	if (!dm_task_no_open_count(dmt))
+		goto_out;
+	if (!dm_task_set_major_minor(dmt, MAJOR(dev->dev), MINOR(dev->dev), 1))
+		goto_out;
+	if (!dm_task_run(dmt)) {
+		log_error("Failed to get state of mapped device.");
+		goto out;
+	}
+	dm_get_next_target(dmt, next, &start, &length, &target_type, &params);
+	if (sscanf(params, "%d:%d", &minor, &major) != 2) {
+		log_error("Failed to get thin-pool major:minor for thin device %d:%d.",
+			  (int)MAJOR(dev->dev), (int)MINOR(dev->dev));
+		goto out;
+	}
+	dm_task_destroy(dmt);
+
+	if (!(dmt = dm_task_create(DM_DEVICE_STATUS)))
+		goto_out;
+	if (!dm_task_no_flush(dmt))
+		log_warn("Can't set no_flush.");
+	if (!dm_task_no_open_count(dmt))
+		goto_out;
+	if (!dm_task_set_major_minor(dmt, minor, major, 1))
+		goto_out;
+	if (!dm_task_run(dmt)) {
+		log_error("Failed to get state of mapped device.");
+		goto out;
+	}
+
+	dm_get_next_target(dmt, next, &start, &length, &target_type, &params);
+	if (!dm_get_status_thin_pool(mem, params, &status))
+		return_0;
+
+	if (status->read_only || status->out_of_data_space) {
+		log_warn("WARNING: %s: Thin's thin-pool needs inspection.",
+			 dev_name(dev));
+		goto out;
+	}
+
+	r = 1;
+out:
+	if (dmt)
+		dm_task_destroy(dmt);
+
+	dm_pool_destroy(mem);
+
+        return r;
+}
+
 /*
  * device_is_usable
  * @dev
@@ -646,6 +713,13 @@ int device_is_usable(struct device *dev, struct dev_usable_check_params check)
 			goto out;
 		}
 
+		/* TODO: extend check struct ? */
+		if (target_type && !strcmp(target_type, "thin") &&
+		    !_ignore_unusable_thins(dev)) {
+			log_debug_activation("%s: %s device %s not usable.", dev_name(dev), target_type, name);
+			goto out;
+		}
+
 		if (target_type && strcmp(target_type, "error"))
 			only_error_target = 0;
 	} while (next);
@@ -971,6 +1045,11 @@ static int _percent_run(struct dev_manager *dm, const char *name,
 				wait ? DM_DEVICE_WAITEVENT : DM_DEVICE_STATUS, 0, 0, 0)))
 		return_0;
 
+	/* No freeze on overfilled thin-pool, read existing slightly outdated data */
+	if (lv && lv_is_thin_pool(lv) &&
+	    !dm_task_no_flush(dmt))
+		log_warn("Can't set no_flush flag."); /* Non fatal */
+
 	if (!dm_task_run(dmt))
 		goto_out;
 
@@ -1038,7 +1117,7 @@ static int _percent_run(struct dev_manager *dm, const char *name,
 			goto_out;
 	}
 
-	log_debug_activation("LV percent: %f", dm_percent_to_float(*overall_percent));
+	log_debug_activation("LV percent: %.2f", dm_percent_to_float(*overall_percent));
 	r = 1;
 
       out:
@@ -1918,7 +1997,6 @@ struct pool_cb_data {
 	int skip_zero;  /* to skip zeroed device header (check first 64B) */
 	int exec;       /* which binary to call */
 	int opts;
-	const char *defaults;
 	const char *global;
 };
 
@@ -1926,7 +2004,6 @@ static int _pool_callback(struct dm_tree_node *node,
 			  dm_node_callback_t type, void *cb_data)
 {
 	int ret, status, fd;
-	char *split;
 	const struct dm_config_node *cn;
 	const struct dm_config_value *cv;
 	const struct pool_cb_data *data = cb_data;
@@ -1941,23 +2018,19 @@ static int _pool_callback(struct dm_tree_node *node,
 	if (!*argv[0])
 		return 1; /* Checking disabled */
 
-	if ((cn = find_config_tree_node(mlv->vg->cmd, data->opts, NULL))) {
-		for (cv = cn->v; cv && args < 16; cv = cv->next) {
-			if (cv->type != DM_CFG_STRING) {
-				log_error("Invalid string in config file: "
-					  "global/%s_check_options",
-					  data->global);
-				return 0;
-			}
-			argv[++args] = cv->v.str;
-		}
-	} else {
-		/* Use default options (no support for options with spaces) */
-		if (!(split = dm_pool_strdup(data->dm->mem, data->defaults))) {
-			log_error("Failed to duplicate defaults.");
+	if (!(cn = find_config_tree_array(mlv->vg->cmd, data->opts, NULL))) {
+		log_error(INTERNAL_ERROR "Unable to find configuration for pool check options.");
+		return 0;
+	}
+
+	for (cv = cn->v; cv && args < 16; cv = cv->next) {
+		if (cv->type != DM_CFG_STRING) {
+			log_error("Invalid string in config file: "
+				  "global/%s_check_options",
+				  data->global);
 			return 0;
 		}
-		args = dm_split_words(split, 16, 0, (char**) argv + 1);
+		argv[++args] = cv->v.str;
 	}
 
 	if (args == 16) {
@@ -2048,14 +2121,12 @@ static int _pool_register_callback(struct dev_manager *dm,
 		data->skip_zero = 1;
 		data->exec = global_thin_check_executable_CFG;
 		data->opts = global_thin_check_options_CFG;
-		data->defaults = DEFAULT_THIN_CHECK_OPTION1 " " DEFAULT_THIN_CHECK_OPTION2;
 		data->global = "thin";
 	} else if (lv_is_cache(lv)) { /* cache pool */
 		data->pool_lv = first_seg(lv)->pool_lv;
 		data->skip_zero = dm->activation;
 		data->exec = global_cache_check_executable_CFG;
 		data->opts = global_cache_check_options_CFG;
-		data->defaults = DEFAULT_CACHE_CHECK_OPTION1;
 		data->global = "cache";
 	} else {
 		log_error(INTERNAL_ERROR "Registering unsupported pool callback.");
@@ -3206,7 +3277,7 @@ static int _tree_action(struct dev_manager *dm, const struct logical_volume *lv,
 		break;
 	case SUSPEND:
 		dm_tree_skip_lockfs(root);
-		if (!dm->flush_required && lv_is_mirror(lv) && !lv_is_pvmove(lv))
+		if (!dm->flush_required && !lv_is_pvmove(lv))
 			dm_tree_use_no_flush_suspend(root);
 		/* Fall through */
 	case SUSPEND_WITH_LOCKFS:
@@ -3225,7 +3296,14 @@ static int _tree_action(struct dev_manager *dm, const struct logical_volume *lv,
 		if (!dm_tree_preload_children(root, dlid, DLID_SIZE))
 			goto_out;
 
-		if (dm_tree_node_size_changed(root))
+		if ((dm_tree_node_size_changed(root) < 0))
+			dm->flush_required = 1;
+
+		/* Currently keep the code require flush for any
+		 * non 'thin pool/volume, mirror' or with any size change */
+		if (!lv_is_thin_volume(lv) &&
+		    !lv_is_thin_pool(lv) &&
+		    (!lv_is_mirror(lv) || dm_tree_node_size_changed(root)))
 			dm->flush_required = 1;
 
 		if (action == ACTIVATE) {

lib/cache/lvmcache.c

@@ -56,6 +56,7 @@ struct lvmcache_vginfo {
 	char _padding[7];
 	struct lvmcache_vginfo *next; /* Another VG with same name? */
 	char *creation_host;
+	char *lock_type;
 	uint32_t mda_checksum;
 	size_t mda_size;
 	size_t vgmetadata_size;
@@ -1447,7 +1448,7 @@ static int _lvmcache_update_vgname(struct lvmcache_info *info,
 }
 
 static int _lvmcache_update_vgstatus(struct lvmcache_info *info, uint32_t vgstatus,
-				     const char *creation_host)
+				     const char *creation_host, const char *lock_type)
 {
 	if (!info || !info->vginfo)
 		return 1;
@@ -1460,11 +1461,11 @@ static int _lvmcache_update_vgstatus(struct lvmcache_info *info, uint32_t vgstat
 	info->vginfo->status = vgstatus;
 
 	if (!creation_host)
-		return 1;
+		goto set_lock_type;
 
 	if (info->vginfo->creation_host && !strcmp(creation_host,
 						   info->vginfo->creation_host))
-		return 1;
+		goto set_lock_type;
 
 	if (info->vginfo->creation_host)
 		dm_free(info->vginfo->creation_host);
@@ -1478,6 +1479,24 @@ static int _lvmcache_update_vgstatus(struct lvmcache_info *info, uint32_t vgstat
 	log_debug_cache("lvmcache: %s: VG %s: Set creation host to %s.",
 			dev_name(info->dev), info->vginfo->vgname, creation_host);
 
+set_lock_type:
+
+	if (!lock_type)
+		goto out;
+
+	if (info->vginfo->lock_type && !strcmp(lock_type, info->vginfo->lock_type))
+		goto out;
+
+	if (info->vginfo->lock_type)
+		dm_free(info->vginfo->lock_type);
+
+	if (!(info->vginfo->lock_type = dm_strdup(lock_type))) {
+		log_error("cache creation host alloc failed for %s",
+			  lock_type);
+		return 0;
+	}
+
+out:
 	return 1;
 }
 
@@ -1523,10 +1542,6 @@ int lvmcache_update_vgname_and_id(struct lvmcache_info *info, struct lvmcache_vg
 		vgid = vgname;
 	}
 
-	/* When using lvmetad, the PV could not have become orphaned. */
-	if (lvmetad_active() && is_orphan_vg(vgname) && info->vginfo)
-		return 1;
-
 	/* If PV without mdas is already in a real VG, don't make it orphan */
 	if (is_orphan_vg(vgname) && info->vginfo &&
 	    mdas_empty_or_ignored(&info->mdas) &&
@@ -1546,7 +1561,7 @@ int lvmcache_update_vgname_and_id(struct lvmcache_info *info, struct lvmcache_vg
 	if (!_lvmcache_update_vgname(info, vgname, vgid, vgsummary->vgstatus,
 				     vgsummary->creation_host, info->fmt) ||
 	    !_lvmcache_update_vgid(info, info->vginfo, vgid) ||
-	    !_lvmcache_update_vgstatus(info, vgsummary->vgstatus, vgsummary->creation_host) ||
+	    !_lvmcache_update_vgstatus(info, vgsummary->vgstatus, vgsummary->creation_host, vgsummary->lock_type) ||
 	    !_lvmcache_update_vg_mda_info(info, vgsummary->mda_checksum, vgsummary->mda_size))
 		return_0;
 
@@ -1561,7 +1576,8 @@ int lvmcache_update_vg(struct volume_group *vg, unsigned precommitted)
 	struct lvmcache_vgsummary vgsummary = {
 		.vgname = vg->name,
 		.vgstatus = vg->status,
-		.vgid = vg->id
+		.vgid = vg->id,
+		.lock_type = vg->lock_type
 	};
 
 	pvid_s[sizeof(pvid_s) - 1] = '\0';
@@ -1851,19 +1867,17 @@ struct lvmcache_info *lvmcache_add(struct labeller *labeller, const char *pvid,
 			 * device already exists?  Things don't seem to work
 			 * if we do that for some reason.
 			 */
-			log_verbose("Found same device %s with same pvid %s",
-				    dev_name(existing->dev), pvid_s);
+			log_debug_cache("Found same device %s with same pvid %s",
+					dev_name(existing->dev), pvid_s);
 		}
 
 		/*
-		 * FIXME: when could this ever happen?
-		 * If this does happen, identify when/why here, and
-		 * if not, remove this code.
+		 * This happens when running pvcreate on an existing PV.
 		 */
 		if (strcmp(pvid_s, existing->dev->pvid))  {
-			log_warn("Replacing dev %s pvid %s with dev %s pvid %s",
-				 dev_name(existing->dev), existing->dev->pvid,
-				 dev_name(dev), pvid_s);
+			log_verbose("Replacing dev %s pvid %s with dev %s pvid %s",
+				    dev_name(existing->dev), existing->dev->pvid,
+				    dev_name(dev), pvid_s);
 		}
 
 		/*
@@ -2327,3 +2341,40 @@ int lvmcache_lookup_mda(struct lvmcache_vgsummary *vgsummary)
 
 	return 0;
 }
+
+int lvmcache_contains_lock_type_sanlock(struct cmd_context *cmd)
+{
+	struct lvmcache_vginfo *vginfo;
+
+	dm_list_iterate_items(vginfo, &_vginfos) {
+		if (vginfo->lock_type && !strcmp(vginfo->lock_type, "sanlock"))
+			return 1;
+	}
+
+	return 0;
+}
+
+void lvmcache_get_max_name_lengths(struct cmd_context *cmd,
+				   unsigned *pv_max_name_len,
+				   unsigned *vg_max_name_len)
+{
+	struct lvmcache_vginfo *vginfo;
+	struct lvmcache_info *info;
+	unsigned len;
+
+	*vg_max_name_len = 0;
+	*pv_max_name_len = 0;
+
+	dm_list_iterate_items(vginfo, &_vginfos) {
+		len = strlen(vginfo->vgname);
+		if (*vg_max_name_len < len)
+			*vg_max_name_len = len;
+
+		dm_list_iterate_items(info, &vginfo->infos) {
+			len = strlen(dev_name(info->dev));
+			if (*pv_max_name_len < len)
+				*pv_max_name_len = len;
+		}
+	}
+}
+

lib/cache/lvmcache.h

@@ -39,11 +39,23 @@ struct disk_locn;
 
 struct lvmcache_vginfo;
 
+/*
+ * vgsummary represents a summary of the VG that is read
+ * without a lock.  The info does not come through vg_read(),
+ * but through reading mdas.  It provides information about
+ * the VG that is needed to lock the VG and then read it fully
+ * with vg_read(), after which the VG summary should be checked
+ * against the full VG metadata to verify it was correct (since
+ * it was read without a lock.)
+ *
+ * Once read, vgsummary information is saved in lvmcache_vginfo.
+ */
 struct lvmcache_vgsummary {
 	const char *vgname;
 	struct id vgid;
 	uint64_t vgstatus;
 	char *creation_host;
+	const char *lock_type;
 	uint32_t mda_checksum;
 	size_t mda_size;
 };
@@ -176,4 +188,10 @@ int lvmcache_found_duplicate_pvs(void);
 
 void lvmcache_set_preferred_duplicates(const char *vgid);
 
+int lvmcache_contains_lock_type_sanlock(struct cmd_context *cmd);
+
+void lvmcache_get_max_name_lengths(struct cmd_context *cmd,
+			unsigned *pv_max_name_len, unsigned *vg_max_name_len);
+
+
 #endif

lib/cache/lvmetad.c

@@ -37,6 +37,70 @@ static struct cmd_context *_lvmetad_cmd = NULL;
 
 static struct volume_group *lvmetad_pvscan_vg(struct cmd_context *cmd, struct volume_group *vg);
 
+static int _log_debug_inequality(const char *name, struct dm_config_node *a, struct dm_config_node *b)
+{
+	int result = 0;
+	int final_result = 0;
+
+	if (a->v && b->v) {
+		result = compare_value(a->v, b->v);
+		if (result) {
+			struct dm_config_value *av = a->v;
+			struct dm_config_value *bv = b->v;
+
+			if (!strcmp(a->key, b->key)) {
+				if (a->v->type == DM_CFG_STRING && b->v->type == DM_CFG_STRING)
+					log_debug_lvmetad("VG %s metadata inequality at %s / %s: %s / %s",
+							  name, a->key, b->key, av->v.str, bv->v.str);
+				else if (a->v->type == DM_CFG_INT && b->v->type == DM_CFG_INT)
+					log_debug_lvmetad("VG %s metadata inequality at %s / %s: " FMTi64 " / " FMTi64,
+							  name, a->key, b->key, av->v.i, bv->v.i);
+				else
+					log_debug_lvmetad("VG %s metadata inequality at %s / %s: type %d / type %d",
+							  name, a->key, b->key, av->type, bv->type);
+			} else {
+				log_debug_lvmetad("VG %s metadata inequality at %s / %s", name, a->key, b->key);
+			}
+			final_result = result;
+		}
+	}
+
+	if (a->v && !b->v) {
+		log_debug_lvmetad("VG %s metadata inequality at %s / %s", name, a->key, b->key);
+		final_result = 1;
+	}
+
+	if (!a->v && b->v) {
+		log_debug_lvmetad("VG %s metadata inequality at %s / %s", name, a->key, b->key);
+		final_result = -1;
+	}
+
+	if (a->child && b->child) {
+		result = _log_debug_inequality(name, a->child, b->child);
+		if (result)
+			final_result = result;
+	}
+
+	if (a->sib && b->sib) {
+		result = _log_debug_inequality(name, a->sib, b->sib);
+		if (result)
+			final_result = result;
+	}
+	
+
+	if (a->sib && !b->sib) {
+		log_debug_lvmetad("VG %s metadata inequality at %s / %s", name, a->key, b->key);
+		final_result = 1;
+	}
+
+	if (!a->sib && b->sib) {
+		log_debug_lvmetad("VG %s metadata inequality at %s / %s", name, a->key, b->key);
+		final_result = -1;
+	}
+
+	return final_result;
+}
+
 void lvmetad_disconnect(void)
 {
 	if (_lvmetad_connected)
@@ -148,7 +212,7 @@ static int _lvmetad_pvscan_all_devs(struct cmd_context *cmd, activation_handler
 static daemon_reply _lvmetad_send(const char *id, ...)
 {
 	va_list ap;
-	daemon_reply repl;
+	daemon_reply repl = { 0 };
 	daemon_request req;
 	unsigned num_rescans = 0;
 	unsigned total_usecs_waited = 0;
@@ -158,8 +222,10 @@ static daemon_reply _lvmetad_send(const char *id, ...)
 retry:
 	req = daemon_request_make(id);
 
-	if (_lvmetad_token)
-		daemon_request_extend(req, "token = %s", _lvmetad_token, NULL);
+	if (_lvmetad_token && !daemon_request_extend(req, "token = %s", _lvmetad_token, NULL)) {
+		repl.error = ENOMEM;
+		return repl;
+	}
 
 	va_start(ap, id);
 	daemon_request_extend_v(req, ap);
@@ -432,6 +498,7 @@ struct volume_group *lvmetad_vg_lookup(struct cmd_context *cmd, const char *vgna
 	struct format_type *fmt;
 	struct dm_config_node *pvcn;
 	struct pv_list *pvl;
+	int rescan = 0;
 
 	if (!lvmetad_active())
 		return NULL;
@@ -490,16 +557,56 @@ struct volume_group *lvmetad_vg_lookup(struct cmd_context *cmd, const char *vgna
 		if (!(vg = import_vg_from_lvmetad_config_tree(reply.cft, fid)))
 			goto_out;
 
+		/*
+		 * Read the VG from disk, ignoring the lvmetad copy in these
+		 * cases:
+		 *
+		 * 1. The host is not using lvmlockd, but is reading lockd VGs
+		 * using the --shared option.  The shared option is meant to
+		 * let hosts not running lvmlockd look at lockd VGs, like the
+		 * foreign option allows hosts to look at foreign VGs.  When
+		 * --foreign is used, the code forces a rescan since the local
+		 * lvmetad cache of foreign VGs is likely stale.  Similarly,
+		 * for --shared, have the code reading the shared VGs below
+		 * not use the cached copy from lvmetad but to rescan the VG.
+		 *
+		 * 2. The host failed to acquire the VG lock from lvmlockd for
+		 * the lockd VG.  In this case, the usual mechanisms for
+		 * updating the lvmetad copy of the VG have been missed.  Since
+		 * we don't know if the cached copy is valid, assume it's not.
+		 *
+		 * 3. lvmetad has returned the "vg_invalid" flag, which is the
+		 * usual mechanism used by lvmlockd/lvmetad to cause a host to
+		 * reread a VG from disk that has been modified from another
+		 * host.
+		 */
+
+		if (is_lockd_type(vg->lock_type) && cmd->include_shared_vgs) {
+			log_debug_lvmetad("Rescan VG %s because including shared", vgname);
+			rescan = 1;
+		} else if (is_lockd_type(vg->lock_type) && cmd->lockd_vg_rescan) {
+			log_debug_lvmetad("Rescan VG %s because no lvmlockd lock is held", vgname);
+			rescan = 1;
+		} else if (dm_config_find_node(reply.cft->root, "vg_invalid")) {
+			log_debug_lvmetad("Rescan VG %s because lvmetad returned invalid", vgname);
+			rescan = 1;
+		}
+
 		/*
 		 * locking may have detected a newer vg version and
 		 * invalidated the cached vg.
 		 */
-		if (dm_config_find_node(reply.cft->root, "vg_invalid")) {
+		if (rescan) {
 			log_debug_lvmetad("Update invalid lvmetad cache for VG %s", vgname);
 			vg2 = lvmetad_pvscan_vg(cmd, vg);
 			release_vg(vg);
 			vg = vg2;
-			fid = vg->fid;
+			if (!vg) {
+				log_debug_lvmetad("VG %s from lvmetad not found during rescan.", vgname);
+				fid = NULL;
+				goto out;
+			} else
+				fid = vg->fid;
 		}
 
 		dm_list_iterate_items(pvl, &vg->pvs) {
@@ -950,6 +1057,51 @@ int lvmetad_pv_found(const struct id *pvid, struct device *dev, const struct for
 	     daemon_reply_int(reply, "seqno_after", -1) != daemon_reply_int(reply, "seqno_before", -1)))
 		log_warn("WARNING: Inconsistent metadata found for VG %s", vg->name);
 
+	/*
+	 * pvscan --cache does not perform any lvmlockd locking, and
+	 * pvscan --cache -aay skips autoactivation in lockd VGs.
+	 *
+	 * pvscan --cache populates lvmetad with VG metadata from disk.
+	 * No lvmlockd locking is needed.  It is expected that lockd VG
+	 * metadata that is read by pvscan and populated in lvmetad may
+	 * be immediately stale due to changes to the VG from other hosts
+	 * during or after this pvscan.  This is normal and not a problem.
+	 * When a subsequent lvm command uses the VG, it will lock the VG
+	 * with lvmlockd, read the VG from lvmetad, and update the cached
+	 * copy from disk if necessary.
+	 *
+	 * pvscan --cache -aay does not activate LVs in lockd VGs because
+	 * activation requires locking, and a lock-start operation is needed
+	 * on a lockd VG before any locking can be performed in it.
+	 *
+	 * An equivalent of pvscan --cache -aay for lockd VGs is:
+	 * 1. pvscan --cache
+	 * 2. vgchange --lock-start
+	 * 3. vgchange -aay -S 'locktype=sanlock || locktype=dlm'
+	 *
+	 * [We could eventually add support for autoactivating lockd VGs
+	 * using pvscan by incorporating the lock start step (which can
+	 * take a long time), but there may be a better option than
+	 * continuing to overload pvscan.]
+	 * 
+	 * Stages of starting a lockd VG:
+	 *
+	 * . pvscan --cache populates lockd VGs in lvmetad without locks,
+	 *   and this initial cached copy may quickly become stale.
+	 *
+	 * . vgchange --lock-start VG reads the VG without the VG lock
+	 *   because no locks are available until the locking is started.
+	 *   It only uses the VG name and lock_type from the VG metadata,
+	 *   and then only uses it to start the VG lockspace in lvmlockd.
+	 *
+	 * . Further lvm commands, e.g. activation, can then lock the VG
+	 *   with lvmlockd and use current VG metdata.
+	 */
+	if (handler && vg && is_lockd_type(vg->lock_type)) {
+		log_debug_lvmetad("Skip pvscan activation for lockd type VG %s", vg->name);
+		handler = NULL;
+	}
+
 	if (result && handler) {
 		status = daemon_reply_str(reply, "status", "<missing>");
 		vgname = daemon_reply_str(reply, "vgname", "<missing>");
@@ -1017,7 +1169,8 @@ static int _lvmetad_pvscan_single(struct metadata_area *mda, void *baton)
 	struct _lvmetad_pvscan_baton *b = baton;
 	struct volume_group *this;
 
-	this = mda_is_ignored(mda) ? NULL : mda->ops->vg_read(b->fid, "", mda, NULL, NULL, 1);
+	if (!(this = mda_is_ignored(mda) ? NULL : mda->ops->vg_read(b->fid, "", mda, NULL, NULL, 1)))
+		return 1;
 
 	/* FIXME Also ensure contents match etc. */
 	if (!b->vg || this->seqno > b->vg->seqno)
@@ -1033,6 +1186,8 @@ static int _lvmetad_pvscan_single(struct metadata_area *mda, void *baton)
  * due to something like an lvcreate from another host.
  * This is limited to changes that only affect the vg (not global state like
  * orphan PVs), so we only need to reread mdas on the vg's existing pvs.
+ * But, a previous PV in the VG may have been removed since we last read
+ * the VG, and that PV may have been reused for another VG.
  */
 
 static struct volume_group *lvmetad_pvscan_vg(struct cmd_context *cmd, struct volume_group *vg)
@@ -1045,13 +1200,17 @@ static struct volume_group *lvmetad_pvscan_vg(struct cmd_context *cmd, struct vo
 	struct format_instance *fid;
 	struct format_instance_ctx fic = { .type = 0 };
 	struct _lvmetad_pvscan_baton baton;
+	struct device *save_dev = NULL;
 
 	dm_list_iterate_items(pvl, &vg->pvs) {
 		/* missing pv */
 		if (!pvl->pv->dev)
 			continue;
 
-		info = lvmcache_info_from_pvid((const char *)&pvl->pv->id, 0);
+		if (!(info = lvmcache_info_from_pvid((const char *)&pvl->pv->id, 0))) {
+			log_error("Failed to find cached info for PV %s.", pv_dev_name(pvl->pv));
+			return NULL;
+		}
 
 		baton.vg = NULL;
 		baton.fid = lvmcache_fmt(info)->ops->create_instance(lvmcache_fmt(info), &fic);
@@ -1068,9 +1227,25 @@ static struct volume_group *lvmetad_pvscan_vg(struct cmd_context *cmd, struct vo
 
 		lvmcache_foreach_mda(info, _lvmetad_pvscan_single, &baton);
 
+		/*
+		 * The PV may have been removed from the VG by another host
+		 * since we last read the VG.
+		 */
 		if (!baton.vg) {
+			log_debug_lvmetad("Did not find VG %s in scan of PV %s", vg->name, dev_name(pvl->pv->dev));
 			lvmcache_fmt(info)->ops->destroy_instance(baton.fid);
-			return NULL;
+			continue;
+		}
+
+		/*
+		 * The PV may have been removed from the VG and used for a
+		 * different VG since we last read the VG.
+		 */
+		if (strcmp(baton.vg->name, vg->name)) {
+			log_debug_lvmetad("Did not find VG %s in scan of PV %s which is now VG %s",
+					  vg->name, dev_name(pvl->pv->dev), baton.vg->name);
+			release_vg(baton.vg);
+			continue;
 		}
 
 		if (!(vgmeta = export_vg_to_config_tree(baton.vg))) {
@@ -1081,9 +1256,12 @@ static struct volume_group *lvmetad_pvscan_vg(struct cmd_context *cmd, struct vo
 
 		if (!vgmeta_ret) {
 			vgmeta_ret = vgmeta;
+			save_dev = pvl->pv->dev;
 		} else {
-			if (!compare_config(vgmeta_ret->root, vgmeta->root)) {
-				log_error("VG metadata comparison failed");
+			if (compare_config(vgmeta_ret->root, vgmeta->root)) {
+				log_error("VG %s metadata comparison failed for device %s vs %s",
+					  vg->name, dev_name(pvl->pv->dev), save_dev ? dev_name(save_dev) : "none");
+				_log_debug_inequality(vg->name, vgmeta_ret->root, vgmeta->root);
 				dm_config_destroy(vgmeta);
 				dm_config_destroy(vgmeta_ret);
 				release_vg(baton.vg);
@@ -1153,7 +1331,7 @@ int lvmetad_pvscan_single(struct cmd_context *cmd, struct device *dev,
 			log_warn("WARNING: Ignoring obsolete format of metadata (%s) on device %s when using lvmetad",
 				  baton.fid->fmt->name, dev_name(dev));
 		else
-			log_error("WARNING: Ignoring obsolete format of metadata (%s) on device %s when using lvmetad",
+			log_error("Ignoring obsolete format of metadata (%s) on device %s when using lvmetad.",
 				  baton.fid->fmt->name, dev_name(dev));
 		lvmcache_fmt(info)->ops->destroy_instance(baton.fid);
 
@@ -1350,7 +1528,8 @@ static int _lvmetad_get_pv_cache_list(struct cmd_context *cmd, struct dm_list *p
 /*
  * Opening the device RDWR should trigger a udev db update.
  * FIXME: is there a better way to update the udev db than
- * doing an open/close of the device?
+ * doing an open/close of the device? - For example writing
+ * "change" to /sys/block/<device>/uevent?
  */
 static void _update_pv_in_udev(struct cmd_context *cmd, dev_t devt)
 {
@@ -1364,9 +1543,13 @@ static void _update_pv_in_udev(struct cmd_context *cmd, dev_t devt)
 		return;
 	}
 
-	if (!dev_open(dev))
+	if (!dev_open(dev)) {
+		stack;
 		return;
-	dev_close(dev);
+	}
+
+	if (!dev_close(dev))
+		stack;
 }
 
 /*
@@ -1500,7 +1683,7 @@ void lvmetad_validate_global_cache(struct cmd_context *cmd, int force)
 		return;
 	}
 
-	if (!lvmetad_used())
+	if (!lvmetad_active())
 		return;
 
 	log_debug_lvmetad("Validating global lvmetad cache");

lib/cache_segtype/cache.c

@@ -25,6 +25,11 @@
 #include "lv_alloc.h"
 #include "defaults.h"
 
+static const char _cache_module[] = "cache";
+
+/* TODO: using static field here, maybe should be a part of segment_type */
+static unsigned _feature_mask;
+
 #define SEG_LOG_ERROR(t, p...) \
         log_error(t " segment %s of logical volume %s.", ## p,	\
                   dm_config_parent_name(sn), seg->lv->name), 0;
@@ -66,20 +71,16 @@ static int _cache_pool_text_import(struct lv_segment *seg,
 	if (dm_config_has_node(sn, "cache_mode")) {
 		if (!(str = dm_config_find_str(sn, "cache_mode", NULL)))
 			return SEG_LOG_ERROR("cache_mode must be a string in");
-		if (!set_cache_pool_feature(&seg->feature_flags, str))
+		if (!cache_set_mode(seg, str))
 			return SEG_LOG_ERROR("Unknown cache_mode in");
-	} else
-		/* When missed in metadata, it's an old stuff - use writethrough */
-		seg->feature_flags |= DM_CACHE_FEATURE_WRITETHROUGH;
+	}
 
 	if (dm_config_has_node(sn, "policy")) {
 		if (!(str = dm_config_find_str(sn, "policy", NULL)))
 			return SEG_LOG_ERROR("policy must be a string in");
 		if (!(seg->policy_name = dm_pool_strdup(mem, str)))
 			return SEG_LOG_ERROR("Failed to duplicate policy in");
-	} else
-		/* Cannot use 'just' default, so pick one */
-		seg->policy_name = DEFAULT_CACHE_POOL_POLICY; /* FIXME make configurable */
+	}
 
 	/*
 	 * Read in policy args:
@@ -99,6 +100,9 @@ static int _cache_pool_text_import(struct lv_segment *seg,
 	 *   If the policy is not present, default policy is used.
 	 */
 	if ((sn = dm_config_find_node(sn, "policy_settings"))) {
+		if (!seg->policy_name)
+			return SEG_LOG_ERROR("policy_settings must have a policy_name in");
+
 		if (sn->v)
 			return SEG_LOG_ERROR("policy_settings must be a section in");
 
@@ -127,24 +131,33 @@ static int _cache_pool_text_export(const struct lv_segment *seg,
 {
 	const char *cache_mode;
 
-	if (!(cache_mode = get_cache_pool_cachemode_name(seg)))
-		return_0;
-
 	outf(f, "data = \"%s\"", seg_lv(seg, 0)->name);
 	outf(f, "metadata = \"%s\"", seg->metadata_lv->name);
 	outf(f, "chunk_size = %" PRIu32, seg->chunk_size);
-	outf(f, "cache_mode = \"%s\"", cache_mode);
 
-	if (seg->policy_name)
+	/*
+	 * Cache pool used by a cache LV holds data. Not ideal,
+	 * but not worth to break backward compatibility, by shifting
+	 * content to cache segment
+	 */
+	if (cache_mode_is_set(seg)) {
+		if (!(cache_mode = get_cache_mode_name(seg)))
+			return_0;
+		outf(f, "cache_mode = \"%s\"", cache_mode);
+	}
+
+	if (seg->policy_name) {
 		outf(f, "policy = \"%s\"", seg->policy_name);
 
-	if (seg->policy_settings) {
-		if (strcmp(seg->policy_settings->key, "policy_settings")) {
-			log_error(INTERNAL_ERROR "Incorrect policy_settings tree, %s.",
-				  seg->policy_settings->key);
-			return 0;
+		if (seg->policy_settings) {
+			if (strcmp(seg->policy_settings->key, "policy_settings")) {
+				log_error(INTERNAL_ERROR "Incorrect policy_settings tree, %s.",
+					  seg->policy_settings->key);
+				return 0;
+			}
+			if (seg->policy_settings->child)
+				out_config_node(f, seg->policy_settings);
 		}
-		out_config_node(f, seg->policy_settings);
 	}
 
 	return 1;
@@ -157,12 +170,29 @@ static void _destroy(struct segment_type *segtype)
 
 #ifdef DEVMAPPER_SUPPORT
 static int _target_present(struct cmd_context *cmd,
-				const struct lv_segment *seg __attribute__((unused)),
-				unsigned *attributes __attribute__((unused)))
+			   const struct lv_segment *seg __attribute__((unused)),
+			   unsigned *attributes __attribute__((unused)))
 {
-	uint32_t maj, min, patchlevel;
+	/* List of features with their kernel target version */
+	static const struct feature {
+		uint32_t maj;
+		uint32_t min;
+		unsigned cache_feature;
+		const char feature[12];
+		const char module[12]; /* check dm-%s */
+	} _features[] = {
+		{ 1, 3, CACHE_FEATURE_POLICY_MQ, "policy_mq", "cache-mq" },
+		{ 1, 8, CACHE_FEATURE_POLICY_SMQ, "policy_smq", "cache-smq" },
+	};
+	static const char _lvmconf[] = "global/cache_disabled_features";
+	static unsigned _attrs = 0;
 	static int _cache_checked = 0;
 	static int _cache_present = 0;
+	uint32_t maj, min, patchlevel;
+	unsigned i;
+	const struct dm_config_node *cn;
+	const struct dm_config_value *cv;
+	const char *str;
 
 	if (!_cache_checked) {
 		_cache_present = target_present(cmd, "cache", 1);
@@ -176,11 +206,53 @@ static int _target_present(struct cmd_context *cmd,
 
 		if ((maj < 1) ||
 		    ((maj == 1) && (min < 3))) {
-			log_error("The cache kernel module is version %u.%u.%u."
-				  "  Version 1.3.0+ is required.",
+			_cache_present = 0;
+			log_error("The cache kernel module is version %u.%u.%u. "
+				  "Version 1.3.0+ is required.",
 				  maj, min, patchlevel);
 			return 0;
 		}
+
+
+		for (i = 0; i < DM_ARRAY_SIZE(_features); ++i) {
+			if (((maj > _features[i].maj) ||
+			     (maj == _features[i].maj && min >= _features[i].min)) &&
+			    (!_features[i].module[0] || module_present(cmd, _features[i].module)))
+				_attrs |= _features[i].cache_feature;
+			else
+				log_very_verbose("Target %s does not support %s.",
+						 _cache_module, _features[i].feature);
+		}
+	}
+
+	if (attributes) {
+		if (!_feature_mask) {
+			/* Support runtime lvm.conf changes, N.B. avoid 32 feature */
+			if ((cn = find_config_tree_array(cmd, global_cache_disabled_features_CFG, NULL))) {
+				for (cv = cn->v; cv; cv = cv->next) {
+					if (cv->type != DM_CFG_STRING) {
+						log_error("Ignoring invalid string in config file %s.",
+							  _lvmconf);
+						continue;
+					}
+					str = cv->v.str;
+					if (!*str)
+						continue;
+					for (i = 0; i < DM_ARRAY_SIZE(_features); ++i)
+						if (strcasecmp(str, _features[i].feature) == 0)
+							_feature_mask |= _features[i].cache_feature;
+				}
+			}
+
+			_feature_mask = ~_feature_mask;
+
+			for (i = 0; i < DM_ARRAY_SIZE(_features); ++i)
+				if ((_attrs & _features[i].cache_feature) &&
+				    !(_feature_mask & _features[i].cache_feature))
+					log_very_verbose("Target %s %s support disabled by %s",
+							 _cache_module, _features[i].feature, _lvmconf);
+		}
+		*attributes = _attrs & _feature_mask;
 	}
 
 	return _cache_present;
@@ -306,7 +378,9 @@ static int _cache_add_target_line(struct dev_manager *dm,
 					   metadata_uuid,
 					   data_uuid,
 					   origin_uuid,
-					   seg->cleaner_policy ? "cleaner" : cache_pool_seg->policy_name,
+					   seg->cleaner_policy ? "cleaner" :
+						   /* undefined policy name -> likely an old "mq" */
+						   cache_pool_seg->policy_name ? : "mq",
 					   seg->cleaner_policy ? NULL : cache_pool_seg->policy_settings,
 					   cache_pool_seg->chunk_size))
 		return_0;
@@ -346,7 +420,7 @@ int init_cache_segtypes(struct cmd_context *cmd,
 		return 0;
 	}
 
-	segtype->name = "cache-pool";
+	segtype->name = SEG_TYPE_NAME_CACHE_POOL;
 	segtype->flags = SEG_CACHE_POOL | SEG_CANNOT_BE_ZEROED | SEG_ONLY_EXCLUSIVE;
 	segtype->ops = &_cache_pool_ops;
 
@@ -360,7 +434,7 @@ int init_cache_segtypes(struct cmd_context *cmd,
 		return 0;
 	}
 
-	segtype->name = "cache";
+	segtype->name = SEG_TYPE_NAME_CACHE;
 	segtype->flags = SEG_CACHE | SEG_ONLY_EXCLUSIVE;
 	segtype->ops = &_cache_ops;
 
@@ -368,5 +442,8 @@ int init_cache_segtypes(struct cmd_context *cmd,
 		return_0;
 	log_very_verbose("Initialised segtype: %s", segtype->name);
 
+	/* Reset mask for recalc */
+	_feature_mask = 0;
+
 	return 1;
 }

lib/commands/toolcontext.c

@@ -245,8 +245,10 @@ static int _parse_debug_classes(struct cmd_context *cmd)
 	const struct dm_config_value *cv;
 	int debug_classes = 0;
 
-	if (!(cn = find_config_tree_node(cmd, log_debug_classes_CFG, NULL)))
-		return DEFAULT_LOGGED_DEBUG_CLASSES;
+	if (!(cn = find_config_tree_array(cmd, log_debug_classes_CFG, NULL))) {
+		log_error(INTERNAL_ERROR "Unable to find configuration for log/debug_classes.");
+		return -1;
+	}
 
 	for (cv = cn->v; cv; cv = cv->next) {
 		if (cv->type != DM_CFG_STRING) {
@@ -360,7 +362,8 @@ static void _init_logging(struct cmd_context *cmd)
 
 	/* Tell device-mapper about our logging */
 #ifdef DEVMAPPER_SUPPORT
-	dm_log_with_errno_init(print_log);
+	if (!dm_log_is_non_default())
+		dm_log_with_errno_init(print_log);
 #endif
 	reset_log_duplicated();
 	reset_lvm_errno(1);
@@ -530,8 +533,6 @@ static int _process_config(struct cmd_context *cmd)
 	const struct dm_config_node *cn;
 	const struct dm_config_value *cv;
 	int64_t pv_min_kb;
-	const char *lvmetad_socket;
-	const char *lvmpolld_socket;
 	int udev_disabled = 0;
 	char sysfs_dir[PATH_MAX];
 
@@ -652,7 +653,7 @@ static int _process_config(struct cmd_context *cmd)
 		}
 	}
 
-	if ((cn = find_config_tree_node(cmd, activation_mlock_filter_CFG, NULL)))
+	if ((cn = find_config_tree_array(cmd, activation_mlock_filter_CFG, NULL)))
 		for (cv = cn->v; cv; cv = cv->next) 
 			if ((cv->type != DM_CFG_STRING) || !cv->v.str[0]) 
 				log_error("Ignoring invalid activation/mlock_filter entry in config file");
@@ -674,41 +675,9 @@ static int _process_config(struct cmd_context *cmd)
 	init_detect_internal_vg_cache_corruption
 		(find_config_tree_bool(cmd, global_detect_internal_vg_cache_corruption_CFG, NULL));
 
-	lvmetad_disconnect();
-	lvmpolld_disconnect();
-
-	lvmetad_socket = getenv("LVM_LVMETAD_SOCKET");
-	if (!lvmetad_socket)
-		lvmetad_socket = DEFAULT_RUN_DIR "/lvmetad.socket";
-
-	/* TODO?
-		lvmetad_socket = find_config_tree_str(cmd, "lvmetad/socket_path",
-						      DEFAULT_RUN_DIR "/lvmetad.socket");
-	*/
-	lvmetad_set_socket(lvmetad_socket);
-	cn = find_config_tree_node(cmd, devices_global_filter_CFG, NULL);
-	lvmetad_set_token(cn ? cn->v : NULL);
-
-	if (find_config_tree_int(cmd, global_locking_type_CFG, NULL) == 3 &&
-	    find_config_tree_bool(cmd, global_use_lvmetad_CFG, NULL)) {
-		log_warn("WARNING: configuration setting use_lvmetad overridden to 0 due to locking_type 3. "
-			 "Clustered environment not supported by lvmetad yet.");
-		lvmetad_set_active(NULL, 0);
-	} else
-		lvmetad_set_active(NULL, find_config_tree_bool(cmd, global_use_lvmetad_CFG, NULL));
-
-	lvmetad_init(cmd);
-
 	if (!_init_system_id(cmd))
 		return_0;
 
-	lvmpolld_socket = getenv("LVM_LVMPOLLD_SOCKET");
-	if (!lvmpolld_socket)
-		lvmpolld_socket = DEFAULT_RUN_DIR "/lvmpolld.socket";
-	lvmpolld_set_socket(lvmpolld_socket);
-
-	lvmpolld_set_active(find_config_tree_bool(cmd, global_use_lvmpolld_CFG, NULL));
-
 	return 1;
 }
 
@@ -1015,15 +984,9 @@ static int _init_dev_cache(struct cmd_context *cmd)
 
 	init_obtain_device_list_from_udev(device_list_from_udev);
 
-	if (!(cn = find_config_tree_node(cmd, devices_scan_CFG, NULL))) {
-		if (!dev_cache_add_dir("/dev")) {
-			log_error("Failed to add /dev to internal "
-				  "device cache");
-			return 0;
-		}
-		log_verbose("device/scan not in config file: "
-			    "Defaulting to /dev");
-		return 1;
+	if (!(cn = find_config_tree_array(cmd, devices_scan_CFG, NULL))) {
+		log_error(INTERNAL_ERROR "Unable to find configuration for devices/scan.");
+		return_0;
 	}
 
 	for (cv = cn->v; cv; cv = cv->next) {
@@ -1061,7 +1024,7 @@ static int _init_dev_cache(struct cmd_context *cmd)
 		}
 	}
 
-	if (!(cn = find_config_tree_node(cmd, devices_loopfiles_CFG, NULL)))
+	if (!(cn = find_config_tree_array(cmd, devices_loopfiles_CFG, NULL)))
 		return 1;
 
 	for (cv = cn->v; cv; cv = cv->next) {
@@ -1107,7 +1070,7 @@ static struct dev_filter *_init_lvmetad_filter_chain(struct cmd_context *cmd)
 			nr_filt++;
 	}
 
-	/* regex filter. Optional. */
+	/* global regex filter. Optional. */
 	if ((cn = find_config_tree_node(cmd, devices_global_filter_CFG, NULL))) {
 		if (!(filters[nr_filt] = regex_filter_create(cn->v))) {
 			log_error("Failed to create global regex device filter");
@@ -1116,6 +1079,17 @@ static struct dev_filter *_init_lvmetad_filter_chain(struct cmd_context *cmd)
 		nr_filt++;
 	}
 
+	/* regex filter. Optional. */
+	if (!lvmetad_used()) {
+		if ((cn = find_config_tree_node(cmd, devices_filter_CFG, NULL))) {
+			if (!(filters[nr_filt] = regex_filter_create(cn->v))) {
+				log_error("Failed to create regex device filter");
+				goto bad;
+			}
+			nr_filt++;
+		}
+	}
+
 	/* device type filter. Required. */
 	if (!(filters[nr_filt] = lvm_type_filter_create(cmd->dev_types))) {
 		log_error("Failed to create lvm type filter");
@@ -1183,30 +1157,33 @@ bad:
  *     md component filter -> fw raid filter
  *
  *   - cmd->filter - the filter chain used for lvmetad responses:
- *     persistent filter -> usable device filter(FILTER_MODE_POST_LVMETAD) ->
- *     regex filter
+ *     persistent filter -> regex_filter -> usable device filter(FILTER_MODE_POST_LVMETAD)
  *
  *   - cmd->full_filter - the filter chain used for all the remaining situations:
- *     lvmetad_filter -> filter
+ *     cmd->lvmetad_filter -> cmd->filter
  *
- * If lvmetad isnot used, there's just one filter chain:
+ * If lvmetad is not used, there's just one filter chain:
  *
  *   - cmd->filter == cmd->full_filter:
- *     persistent filter -> regex filter -> sysfs filter ->
- *     global regex filter -> type filter ->
- *     usable device filter(FILTER_MODE_NO_LVMETAD) ->
- *     mpath component filter -> partitioned filter ->
- *     md component filter -> fw raid filter
+ *     persistent filter -> sysfs filter -> global regex filter ->
+ *     regex_filter -> type filter -> usable device filter(FILTER_MODE_NO_LVMETAD) ->
+ *     mpath component filter -> partitioned filter -> md component filter -> fw raid filter
  *
  */
-static int _init_filters(struct cmd_context *cmd, unsigned load_persistent_cache)
+int init_filters(struct cmd_context *cmd, unsigned load_persistent_cache)
 {
 	const char *dev_cache;
 	struct dev_filter *filter = NULL, *filter_components[2] = {0};
+	int nr_filt;
 	struct stat st;
 	const struct dm_config_node *cn;
 	struct timespec ts, cts;
 
+	if (!cmd->initialized.connections) {
+		log_error(INTERNAL_ERROR "connections must be initialized before filters");
+		return 0;
+	}
+
 	cmd->dump_filter = 0;
 
 	cmd->lvmetad_filter = _init_lvmetad_filter_chain(cmd);
@@ -1226,26 +1203,26 @@ static int _init_filters(struct cmd_context *cmd, unsigned load_persistent_cache
 	 */
 	/* filter component 0 */
 	if (lvmetad_used()) {
-		if (!(filter_components[0] = usable_filter_create(cmd->dev_types, FILTER_MODE_POST_LVMETAD))) {
+		nr_filt = 0;
+		if ((cn = find_config_tree_array(cmd, devices_filter_CFG, NULL))) {
+			if (!(filter_components[nr_filt] = regex_filter_create(cn->v))) {
+				log_verbose("Failed to create regex device filter.");
+				goto bad;
+			}
+			nr_filt++;
+		}
+		if (!(filter_components[nr_filt] = usable_filter_create(cmd->dev_types, FILTER_MODE_POST_LVMETAD))) {
 			log_verbose("Failed to create usable device filter.");
 			goto bad;
 		}
+		nr_filt++;
+		if (!(filter = composite_filter_create(nr_filt, 0, filter_components)))
+			goto_bad;
 	} else {
-		filter_components[0] = cmd->lvmetad_filter;
+		filter = cmd->lvmetad_filter;
 		cmd->lvmetad_filter = NULL;
 	}
 
-	/* filter component 1 */
-	if ((cn = find_config_tree_node(cmd, devices_filter_CFG, NULL))) {
-		if (!(filter_components[1] = regex_filter_create(cn->v)))
-			goto_bad;
-		/* we have two filter components - create composite filter */
-		if (!(filter = composite_filter_create(2, 0, filter_components)))
-			goto_bad;
-	} else
-		/* we have only one filter component - no need to create composite filter */
-		filter = filter_components[0];
-
 	if (!(dev_cache = find_config_tree_str(cmd, devices_cache_CFG, NULL)))
 		goto_bad;
 
@@ -1257,9 +1234,12 @@ static int _init_filters(struct cmd_context *cmd, unsigned load_persistent_cache
 	cmd->filter = filter;
 
 	if (lvmetad_used()) {
-		filter_components[0] = cmd->lvmetad_filter;
-		filter_components[1] = cmd->filter;
-		if (!(cmd->full_filter = composite_filter_create(2, 0, filter_components)))
+		nr_filt = 0;
+		filter_components[nr_filt] = cmd->lvmetad_filter;
+		nr_filt++;
+		filter_components[nr_filt] = cmd->filter;
+		nr_filt++;
+		if (!(cmd->full_filter = composite_filter_create(nr_filt, 0, filter_components)))
 			goto_bad;
 	} else
 		cmd->full_filter = filter;
@@ -1287,6 +1267,7 @@ static int _init_filters(struct cmd_context *cmd, unsigned load_persistent_cache
 				    dev_cache);
 	}
 
+	cmd->initialized.filters = 1;
 	return 1;
 bad:
 	if (!filter) {
@@ -1310,6 +1291,7 @@ bad:
 	if (cmd->lvmetad_filter)
 		cmd->lvmetad_filter->destroy(cmd->lvmetad_filter);
 
+	cmd->initialized.filters = 0;
 	return 0;
 }
 
@@ -1353,7 +1335,7 @@ static int _init_formats(struct cmd_context *cmd)
 #ifdef HAVE_LIBDL
 	/* Load any formats in shared libs if not static */
 	if (!is_static() &&
-	    (cn = find_config_tree_node(cmd, global_format_libraries_CFG, NULL))) {
+	    (cn = find_config_tree_array(cmd, global_format_libraries_CFG, NULL))) {
 
 		const struct dm_config_value *cv;
 		struct format_type *(*init_format_fn) (struct cmd_context *);
@@ -1519,7 +1501,7 @@ static int _init_segtypes(struct cmd_context *cmd)
 #ifdef HAVE_LIBDL
 	/* Load any formats in shared libs unless static */
 	if (!is_static() &&
-	    (cn = find_config_tree_node(cmd, global_segment_libraries_CFG, NULL))) {
+	    (cn = find_config_tree_array(cmd, global_segment_libraries_CFG, NULL))) {
 
 		const struct dm_config_value *cv;
 		int (*init_multiple_segtypes_fn) (struct cmd_context *,
@@ -1683,11 +1665,80 @@ static int _reopen_stream(FILE *stream, int fd, const char *mode, const char *na
 	return 1;
 }
 
+static int _init_lvmetad(struct cmd_context *cmd)
+{
+	const struct dm_config_node *cn;
+	const char *lvmetad_socket;
+
+	lvmetad_disconnect();
+
+	lvmetad_socket = getenv("LVM_LVMETAD_SOCKET");
+	if (!lvmetad_socket)
+		lvmetad_socket = DEFAULT_RUN_DIR "/lvmetad.socket";
+
+	/* TODO?
+		lvmetad_socket = find_config_tree_str(cmd, "lvmetad/socket_path",
+						      DEFAULT_RUN_DIR "/lvmetad.socket");
+	*/
+
+	lvmetad_set_socket(lvmetad_socket);
+	cn = find_config_tree_array(cmd, devices_global_filter_CFG, NULL);
+	lvmetad_set_token(cn ? cn->v : NULL);
+
+	if (find_config_tree_int(cmd, global_locking_type_CFG, NULL) == 3 &&
+	    find_config_tree_bool(cmd, global_use_lvmetad_CFG, NULL)) {
+		log_warn("WARNING: configuration setting use_lvmetad overridden to 0 due to locking_type 3. "
+			 "Clustered environment not supported by lvmetad yet.");
+		lvmetad_set_active(NULL, 0);
+	} else
+		lvmetad_set_active(NULL, find_config_tree_bool(cmd, global_use_lvmetad_CFG, NULL));
+
+	lvmetad_init(cmd);
+	return 1;
+}
+
+static int _init_lvmpolld(struct cmd_context *cmd)
+{
+	const char *lvmpolld_socket;
+
+	lvmpolld_disconnect();
+
+	lvmpolld_socket = getenv("LVM_LVMPOLLD_SOCKET");
+	if (!lvmpolld_socket)
+		lvmpolld_socket = DEFAULT_RUN_DIR "/lvmpolld.socket";
+	lvmpolld_set_socket(lvmpolld_socket);
+
+	lvmpolld_set_active(find_config_tree_bool(cmd, global_use_lvmpolld_CFG, NULL));
+	return 1;
+}
+
+int init_connections(struct cmd_context *cmd)
+{
+
+	if (!_init_lvmetad(cmd)) {
+		log_error("Failed to initialize lvmetad connection.");
+		goto bad;
+	}
+
+	if (!_init_lvmpolld(cmd)) {
+		log_error("Failed to initialize lvmpolld connection.");
+		goto bad;
+	}
+
+	cmd->initialized.connections = 1;
+	return 1;
+bad:
+	cmd->initialized.connections = 0;
+	return 0;
+}
+
 /* Entry point */
 struct cmd_context *create_toolcontext(unsigned is_long_lived,
 				       const char *system_dir,
 				       unsigned set_buffering,
-				       unsigned threaded)
+				       unsigned threaded,
+				       unsigned set_connections,
+				       unsigned set_filters)
 {
 	struct cmd_context *cmd;
 	FILE *new_stream;
@@ -1825,15 +1876,12 @@ struct cmd_context *create_toolcontext(unsigned is_long_lived,
 		goto_out;
 
 	if (!(cmd->dev_types = create_dev_types(cmd->proc_dir,
-						find_config_tree_node(cmd, devices_types_CFG, NULL))))
+						find_config_tree_array(cmd, devices_types_CFG, NULL))))
 		goto_out;
 
 	if (!_init_dev_cache(cmd))
 		goto_out;
 
-	if (!_init_filters(cmd, 1))
-		goto_out;
-
 	memlock_init(cmd);
 
 	if (!_init_formats(cmd))
@@ -1852,12 +1900,18 @@ struct cmd_context *create_toolcontext(unsigned is_long_lived,
 
 	_init_globals(cmd);
 
+	if (set_connections && !init_connections(cmd))
+		return_0;
+
+	if (set_filters && !init_filters(cmd, 1))
+		goto_out;
+
 	cmd->default_settings.cache_vgmetadata = 1;
 	cmd->current_settings = cmd->default_settings;
 
-	cmd->config_initialized = 1;
+	cmd->initialized.config = 1;
 out:
-	if (!cmd->config_initialized) {
+	if (!cmd->initialized.config) {
 		destroy_toolcontext(cmd);
 		cmd = NULL;
 	}
@@ -1929,14 +1983,19 @@ static void _destroy_filters(struct cmd_context *cmd)
 		cmd->full_filter->destroy(cmd->full_filter);
 		cmd->lvmetad_filter = cmd->filter = cmd->full_filter = NULL;
 	}
+	cmd->initialized.filters = 0;
 }
 
 int refresh_filters(struct cmd_context *cmd)
 {
 	int r, saved_ignore_suspended_devices = ignore_suspended_devices();
 
+	if (!cmd->initialized.filters)
+		/* if filters not initialized, there's nothing to refresh */
+		return 1;
+
 	_destroy_filters(cmd);
-	if (!(r = _init_filters(cmd, 0)))
+	if (!(r = init_filters(cmd, 0)))
                 stack;
 
 	/*
@@ -1965,7 +2024,6 @@ int refresh_toolcontext(struct cmd_context *cmd)
 	label_exit();
 	_destroy_segtypes(&cmd->segtypes);
 	_destroy_formats(cmd, &cmd->formats);
-	_destroy_filters(cmd);
 
 	if (!dev_cache_exit())
 		stack;
@@ -1983,7 +2041,7 @@ int refresh_toolcontext(struct cmd_context *cmd)
 
 	_destroy_config(cmd);
 
-	cmd->config_initialized = 0;
+	cmd->initialized.config = 0;
 
 	cmd->hosttags = 0;
 
@@ -2040,15 +2098,12 @@ int refresh_toolcontext(struct cmd_context *cmd)
 		return_0;
 
 	if (!(cmd->dev_types = create_dev_types(cmd->proc_dir,
-						find_config_tree_node(cmd, devices_types_CFG, NULL))))
+						find_config_tree_array(cmd, devices_types_CFG, NULL))))
 		return_0;
 
 	if (!_init_dev_cache(cmd))
 		return_0;
 
-	if (!_init_filters(cmd, 0))
-		return_0;
-
 	if (!_init_formats(cmd))
 		return_0;
 
@@ -2061,7 +2116,13 @@ int refresh_toolcontext(struct cmd_context *cmd)
 	if (!_init_backup(cmd))
 		return_0;
 
-	cmd->config_initialized = 1;
+	cmd->initialized.config = 1;
+
+	if (cmd->initialized.connections && !init_connections(cmd))
+		return_0;
+
+	if (!refresh_filters(cmd))
+		return_0;
 
 	reset_lvm_errno(1);
 	return 1;

lib/commands/toolcontext.h

@@ -60,29 +60,59 @@ struct config_tree_list {
 	struct dm_config_tree *cft;
 };
 
+struct cmd_context_initialized_parts {
+	unsigned config:1; /* used to reinitialize config if previous init was not successful */
+	unsigned filters:1;
+	unsigned connections:1;
+};
+
 /* FIXME Split into tool & library contexts */
 /* command-instance-related variables needed by library */
 struct cmd_context {
-	struct dm_pool *libmem;	/* For permanent config data */
-	struct dm_pool *mem;	/* Transient: Cleared between each command */
+	/*
+	 * Memory handlers.
+	 */
+	struct dm_pool *libmem;			/* for permanent config data */
+	struct dm_pool *mem;			/* transient: cleared between each command */
 
-	const struct format_type *fmt;	/* Current format to use by default */
-	struct format_type *fmt_backup;	/* Format to use for backups */
-
-	struct dm_list formats;	/* Available formats */
-	struct dm_list segtypes;	/* Available segment types */
-	const char *system_id;
-	const char *hostname;
-	const char *kernel_vsn;
-
-	unsigned rand_seed;
-	char *linebuffer;
+	/*
+	 * Command line and arguments.
+	 */
 	const char *cmd_line;
 	struct command *command;
 	char **argv;
 	struct arg_values *arg_values;
 	struct dm_list arg_value_groups;
-	unsigned is_long_lived:1;	/* Optimises persistent_filter handling */
+
+	/*
+	 * Format handlers.
+	 */
+	const struct format_type *fmt;		/* current format to use by default */
+	struct format_type *fmt_backup;		/* format to use for backups */
+	struct dm_list formats;			/* available formats */
+	struct dm_list segtypes;		/* available segment types */
+
+	/*
+	 * Machine and system identification.
+	 */
+	const char *system_id;
+	const char *hostname;
+	const char *kernel_vsn;
+
+	/*
+	 * Device identification.
+	 */
+	struct dev_types *dev_types;		/* recognized extra device types. */
+
+	/*
+	 * Initialization state.
+	 */
+	struct cmd_context_initialized_parts initialized;
+
+	/*
+	 * Switches.
+	 */
+	unsigned is_long_lived:1;		/* optimises persistent_filter handling */
 	unsigned handles_missing_pvs:1;
 	unsigned handles_unknown_segments:1;
 	unsigned use_linear_target:1;
@@ -93,73 +123,73 @@ struct cmd_context {
 	unsigned report_binary_values_as_numeric:1;
 	unsigned metadata_read_only:1;
 	unsigned ignore_clustered_vgs:1;
-	unsigned threaded:1;		/* Set if running within a thread e.g. clvmd */
-
-	const char *time_format;
-
-	unsigned independent_metadata_areas:1;	/* Active formats have MDAs outside PVs */
+	unsigned threaded:1;			/* set if running within a thread e.g. clvmd */
+	unsigned independent_metadata_areas:1;	/* active formats have MDAs outside PVs */
 	unsigned unknown_system_id:1;
-	unsigned include_foreign_vgs:1; /* report/display cmds can reveal foreign VGs */
-	unsigned include_shared_vgs:1;  /* report/display cmds can reveal lockd VGs */
-	unsigned include_active_foreign_vgs:1; /* cmd should process foreign VGs with active LVs */
-	unsigned vg_read_print_access_error:1; /* print access errors from vg_read */
+	unsigned include_foreign_vgs:1;		/* report/display cmds can reveal foreign VGs */
+	unsigned include_shared_vgs:1;		/* report/display cmds can reveal lockd VGs */
+	unsigned include_active_foreign_vgs:1;	/* cmd should process foreign VGs with active LVs */
+	unsigned vg_read_print_access_error:1;	/* print access errors from vg_read */
 	unsigned lockd_gl_disable:1;
 	unsigned lockd_vg_disable:1;
 	unsigned lockd_lv_disable:1;
+	unsigned lockd_gl_removed:1;
+	unsigned lockd_vg_rescan:1;
 	unsigned lockd_vg_default_sh:1;
-
-	struct dev_types *dev_types;
+	unsigned lockd_vg_enforce_sh:1;
 
 	/*
-	 * Use of filters depends on whether lvmetad is used or not:
-	 *
-	 *   - if lvmetad is used:
-	 *   	- cmd->lvmetad_filter used when scanning devices for lvmetad
-	 *   	- cmd->filter used when processing lvmetad responses
-	 *   	- cmd->full_filter used for remaining situations
-	 *
-	 *   - if lvmetad is not used:
-	 *   	- cmd->lvmetad_filter is NULL
-	 *   	- cmd->filter == cmd->full_filter used for all situations
-	 *
+	 * Filtering.
 	 */
-	struct dev_filter *lvmetad_filter;
-	struct dev_filter *filter;
-	struct dev_filter *full_filter;
-	int dump_filter;	/* Dump filter when exiting? */
+	struct dev_filter *lvmetad_filter;	/* pre-lvmetad filter chain */
+	struct dev_filter *filter;		/* post-lvmetad filter chain */
+	struct dev_filter *full_filter;		/* lvmetad_filter + filter */
+	int dump_filter;			/* Dump filter when exiting? */
 
-	struct dm_list config_files; /* master lvm config + any existing tag configs */
-	struct profile_params *profile_params; /* profile handling params including loaded profile configs */
-	struct dm_config_tree *cft; /* the whole cascade: CONFIG_STRING -> CONFIG_PROFILE -> CONFIG_FILE/CONFIG_MERGED_FILES */
-	int config_initialized; /* used to reinitialize config if previous init was not successful */
-
-	struct dm_hash_table *cft_def_hash; /* config definition hash used for validity check (item type + item recognized) */
-
-	/* selected settings with original default/configured value which can be changed during cmd processing */
-	struct config_info default_settings;
-	/* may contain changed values compared to default_settings */
-	struct config_info current_settings;
+	/*
+	 * Configuration.
+	 */
+	struct dm_list config_files; 		/* master lvm config + any existing tag configs */
+	struct profile_params *profile_params;	/* profile handling params including loaded profile configs */
+	struct dm_config_tree *cft;		/* the whole cascade: CONFIG_STRING -> CONFIG_PROFILE -> CONFIG_FILE/CONFIG_MERGED_FILES */
+	struct dm_hash_table *cft_def_hash;	/* config definition hash used for validity check (item type + item recognized) */
+	struct config_info default_settings;	/* selected settings with original default/configured value which can be changed during cmd processing */
+	struct config_info current_settings; 	/* may contain changed values compared to default_settings */
 
+	/*
+	 * Archives and backups.
+	 */
 	struct archive_params *archive_params;
 	struct backup_params *backup_params;
 	const char *stripe_filler;
 
-	/* List of defined tags */
-	struct dm_list tags;
-	const char *report_list_item_separator;
+	/*
+	 * Host tags.
+	 */
+	struct dm_list tags;			/* list of defined tags */
 	int hosttags;
 
-	/* Locking */
-	const char *lock_gl_mode; /* gl mode, from --lock-gl */
-	const char *lock_vg_mode; /* vg mode, from --lock-vg */
-	const char *lock_lv_mode; /* lv mode, from --lock-lv */
-
-	const char *lib_dir;		/* Cache value global/library_dir */
+	/*
+	 * Paths.
+	 */
+	const char *lib_dir;			/* cache value global/library_dir */
 	char system_dir[PATH_MAX];
 	char dev_dir[PATH_MAX];
 	char proc_dir[PATH_MAX];
-	char display_buffer[NAME_LEN * 10];	/* Ring buffer for upto 10 longest vg/lv names */
-	unsigned display_lvname_idx;		/* Index to ring buffer */
+
+	/*
+	 * Buffers.
+	 */
+	char display_buffer[NAME_LEN * 10];	/* ring buffer for upto 10 longest vg/lv names */
+	unsigned display_lvname_idx;		/* index to ring buffer */
+	char *linebuffer;
+
+	/*
+	 * Others - unsorted.
+	 */
+	const char *report_list_item_separator;
+	const char *time_format;
+	unsigned rand_seed;
 };
 
 /*
@@ -169,13 +199,17 @@ struct cmd_context {
 struct cmd_context *create_toolcontext(unsigned is_long_lived,
 				       const char *system_dir,
 				       unsigned set_buffering,
-				       unsigned threaded);
+				       unsigned threaded,
+				       unsigned set_connections,
+				       unsigned set_filters);
 void destroy_toolcontext(struct cmd_context *cmd);
 int refresh_toolcontext(struct cmd_context *cmd);
 int refresh_filters(struct cmd_context *cmd);
 int process_profilable_config(struct cmd_context *cmd);
 int config_files_changed(struct cmd_context *cmd);
 int init_lvmcache_orphans(struct cmd_context *cmd);
+int init_filters(struct cmd_context *cmd, unsigned load_persistent_cache);
+int init_connections(struct cmd_context *cmd);
 
 struct format_type *get_format_by_name(struct cmd_context *cmd, const char *format);
 

lib/config/config.c

@@ -581,8 +581,11 @@ int config_file_read(struct dm_config_tree *cft)
 		if (!(cf->dev = dev_create_file(filename, NULL, NULL, 1)))
 			return_0;
 
-		if (!dev_open_readonly_buffered(cf->dev))
+		if (!dev_open_readonly_buffered(cf->dev)) {
+			dev_destroy_file(cf->dev);
+			cf->dev = NULL;
 			return_0;
+		}
 	}
 
 	r = config_file_read_fd(cft, cf->dev, 0, (size_t) info.st_size, 0, 0,
@@ -666,14 +669,18 @@ static void _log_type_error(const char *path, cfg_def_type_t actual,
 					     actual_type_name, expected_type_name);
 }
 
-static struct dm_config_value *_get_def_array_values(struct dm_config_tree *cft,
+static struct dm_config_value *_get_def_array_values(struct cmd_context *cmd,
+						     struct dm_config_tree *cft,
 						     const cfg_def_item_t *def,
 						     uint32_t format_flags)
 {
+	const char *def_enc_value;
 	char *enc_value, *token, *p, *r;
 	struct dm_config_value *array = NULL, *v = NULL, *oldv = NULL;
 
-	if (!def->default_value.v_CFG_TYPE_STRING) {
+	def_enc_value = cfg_def_get_default_value(cmd, def, CFG_TYPE_ARRAY, NULL);
+
+	if (!def_enc_value) {
 		if (!(array = dm_config_create_value(cft))) {
 			log_error("Failed to create default empty array for %s.", def->name);
 			return NULL;
@@ -683,7 +690,7 @@ static struct dm_config_value *_get_def_array_values(struct dm_config_tree *cft,
 		return array;
 	}
 
-	if (!(p = token = enc_value = dm_strdup(def->default_value.v_CFG_TYPE_STRING))) {
+	if (!(p = token = enc_value = dm_strdup(def_enc_value))) {
 		log_error("_get_def_array_values: dm_strdup failed");
 		return NULL;
 	}
@@ -844,7 +851,7 @@ static int _check_value_differs_from_default(struct cft_check_handle *handle,
 	}
 
 	if (!v_def && (def->type & CFG_TYPE_ARRAY)) {
-		if (!(v_def_array = v_def_iter = _get_def_array_values(handle->cft, def, 0)))
+		if (!(v_def_array = v_def_iter = _get_def_array_values(handle->cmd, handle->cft, def, 0)))
 			return_0;
 		do {
 			/* iterate over each element of the array and check its value */
@@ -1036,9 +1043,14 @@ static int _config_def_check_tree(struct cft_check_handle *handle,
 				  size_t buf_size, struct dm_config_node *root)
 {
 	struct dm_config_node *cn;
+	cfg_def_item_t *def;
 	int valid, r = 1;
 	size_t len;
 
+	def = cfg_def_get_item_p(root->id);
+	if (def->flags & CFG_SECTION_NO_CHECK)
+		return 1;
+
 	for (cn = root->child; cn; cn = cn->sib) {
 		if ((valid = _config_def_check_node(handle, vp, pvp, rp, prp,
 						    buf_size, cn)) && !cn->v) {
@@ -1360,6 +1372,106 @@ int find_config_tree_bool(struct cmd_context *cmd, int id, struct profile *profi
 	return b;
 }
 
+static struct dm_config_node *_get_array_def_node(struct cmd_context *cmd,
+						  cfg_def_item_t *def,
+						  struct profile *profile)
+{
+	struct dm_config_node *cn;
+
+	if (def->flags & CFG_DEFAULT_UNDEFINED)
+		return NULL;
+
+	if (!(cn = dm_config_create_node(cmd->cft, def->name))) {
+		log_error("Failed to create default array node for %s.", def->name);
+		return NULL;
+	}
+
+	if (!(cn->v = _get_def_array_values(cmd, cmd->cft, def, 0))) {
+		dm_pool_free(cmd->cft->mem, cn);
+		return_NULL;
+	}
+
+	return cn;
+}
+
+struct _config_array_out_handle {
+	struct dm_pool *mem;
+	char *str;
+};
+
+static int _config_array_line(const struct dm_config_node *cn, const char *line, void *baton)
+{
+	struct _config_array_out_handle *handle = (struct _config_array_out_handle *) baton;
+
+	if (!(handle->str = dm_pool_strdup(handle->mem, line))) {
+		log_error("_config_array_line: dm_pool_strdup failed");
+		return 0;
+	}
+
+	return 1;
+}
+
+static void _log_array_value_used(struct dm_pool *mem, const struct dm_config_node *cn,
+				  const char *path, int default_used)
+{
+	struct _config_array_out_handle out_handle = { 0 };
+	struct dm_config_node_out_spec out_spec = { 0 };
+	uint32_t old_format_flags;
+
+	out_handle.mem = mem;
+	out_spec.line_fn = _config_array_line;
+
+	old_format_flags = dm_config_value_get_format_flags(cn->v);
+	dm_config_value_set_format_flags(cn->v,
+		DM_CONFIG_VALUE_FMT_COMMON_EXTRA_SPACES |
+		DM_CONFIG_VALUE_FMT_COMMON_ARRAY);
+
+	if (!dm_config_write_one_node_out(cn, &out_spec, &out_handle)) {
+		log_error("_log_array_value_used: failed to write node value");
+		out_handle.mem = NULL;
+	}
+
+	if (default_used)
+		log_very_verbose("%s not found in config: defaulting to %s",
+				 path, out_handle.mem ? out_handle.str : "<unknown>");
+	else
+		log_very_verbose("Setting %s to %s",
+				 path, out_handle.mem ? out_handle.str : "<unknown>");
+
+	if (out_handle.mem)
+		dm_pool_free(out_handle.mem, out_handle.str);
+	dm_config_value_set_format_flags(cn->v, old_format_flags);
+}
+
+const struct dm_config_node *find_config_tree_array(struct cmd_context *cmd, int id, struct profile *profile)
+{
+	cfg_def_item_t *item = cfg_def_get_item_p(id);
+	char path[CFG_PATH_MAX_LEN];
+	int profile_applied;
+	const struct dm_config_node *cn = NULL, *cn_def = NULL;
+	profile_applied = _apply_local_profile(cmd, profile);
+	_cfg_def_make_path(path, sizeof(path), item->id, item, 0);
+
+	if (!(item->type & CFG_TYPE_ARRAY))
+		log_error(INTERNAL_ERROR "%s cfg tree element not declared as array.", path);
+
+	if (_config_disabled(cmd, item, path) ||
+	    !(cn = find_config_tree_node(cmd, id, profile)))
+		cn_def = _get_array_def_node(cmd, item, profile);
+
+	if (cn)
+		_log_array_value_used(cmd->cft->mem, cn, path, 0);
+	else if (cn_def) {
+		_log_array_value_used(cmd->cft->mem, cn_def, path, 1);
+		cn = cn_def;
+	}
+
+	if (profile_applied)
+		remove_config_tree_by_source(cmd, profile->source);
+
+	return cn;
+}
+
 /* Insert cn2 after cn1 */
 static void _insert_config_node(struct dm_config_node **cn1,
 				struct dm_config_node *cn2)
@@ -1558,14 +1670,9 @@ static int _out_prefix_fn(const struct dm_config_node *cn, const char *line, voi
 	char path[CFG_PATH_MAX_LEN];
 	char commentline[MAX_COMMENT_LINE+1];
 
-	if (cn->id < 0)
+	if (cn->id <= 0)
 		return 1;
 
-	if (!cn->id) {
-		log_error(INTERNAL_ERROR "Configuration node %s has invalid id.", cn->key);
-		return 0;
-	}
-
 	if (out->tree_spec->type == CFG_DEF_TREE_LIST)
 		return 1;
 
@@ -1587,6 +1694,11 @@ static int _out_prefix_fn(const struct dm_config_node *cn, const char *line, voi
 		if (cfg_def->comment) {
 			int pos = 0;
 			while (_copy_one_line(cfg_def->comment, commentline, &pos, strlen(cfg_def->comment))) {
+				if ((commentline[0] == '#') && (strlen(commentline) == 1)) {
+					if (!out->tree_spec->withspaces)
+						continue;
+					commentline[0] = '\0';
+				}
 				fprintf(out->fp, "%s# %s\n", line, commentline);
 				/* withsummary prints only the first comment line. */
 				if (!out->tree_spec->withcomments)
@@ -1609,6 +1721,9 @@ static int _out_prefix_fn(const struct dm_config_node *cn, const char *line, voi
 		if (cfg_def->flags & CFG_DEFAULT_UNDEFINED)
 			fprintf(out->fp, "%s# This configuration %s does not have a default value defined.\n", line, node_type_name);
 
+		if (cfg_def->flags & CFG_DEFAULT_COMMENTED)
+			fprintf(out->fp, "%s# This configuration %s has an automatic default value.\n", line, node_type_name);
+
 		if ((out->tree_spec->type == CFG_DEF_TREE_FULL) &&
 		    (out->tree_spec->check_status[cn->id] & CFG_USED))
 			fprintf(out->fp, "%s# Value defined in existing configuration has been used for this setting.\n", line);
@@ -1826,7 +1941,7 @@ static struct dm_config_node *_add_def_node(struct dm_config_tree *cft,
 		if (spec->withspaces)
 			format_flags |= DM_CONFIG_VALUE_FMT_COMMON_EXTRA_SPACES;
 		format_flags |= DM_CONFIG_VALUE_FMT_COMMON_ARRAY;
-		cn->v = _get_def_array_values(cft, def, format_flags);
+		cn->v = _get_def_array_values(spec->cmd, cft, def, format_flags);
 	}
 
 	cn->child = NULL;

lib/config/config.h

@@ -50,7 +50,7 @@ struct profile_params {
 	struct dm_list profiles;                 /* list of profiles which are loaded already and which are ready for use */
 };
 
-#define CFG_PATH_MAX_LEN 64
+#define CFG_PATH_MAX_LEN 128
 
 /*
  * Structures used for definition of a configuration tree.
@@ -121,6 +121,8 @@ typedef union {
 #define CFG_DISABLED		0x200
 /* whether to print integers in octal form (prefixed by "0") */
 #define CFG_FORMAT_INT_OCTAL	0x400
+/* whether to disable checks for the whole config section subtree */
+#define CFG_SECTION_NO_CHECK	0x800
 
 /* configuration definition item structure */
 typedef struct cfg_def_item {
@@ -272,6 +274,7 @@ int find_config_tree_int(struct cmd_context *cmd, int id, struct profile *profil
 int64_t find_config_tree_int64(struct cmd_context *cmd, int id, struct profile *profile);
 float find_config_tree_float(struct cmd_context *cmd, int id, struct profile *profile);
 int find_config_tree_bool(struct cmd_context *cmd, int id, struct profile *profile);
+const struct dm_config_node *find_config_tree_array(struct cmd_context *cmd, int id, struct profile *profile);
 
 /*
  * Functions for configuration settings for which the default
@@ -293,5 +296,7 @@ int get_default_allocation_thin_pool_chunk_size_CFG(struct cmd_context *cmd, str
 #define get_default_unconfigured_allocation_thin_pool_chunk_size_CFG NULL
 int get_default_allocation_cache_pool_chunk_size_CFG(struct cmd_context *cmd, struct profile *profile);
 #define get_default_unconfigured_allocation_cache_pool_chunk_size_CFG NULL
+const char *get_default_allocation_cache_policy_CFG(struct cmd_context *cmd, struct profile *profile);
+#define get_default_unconfigured_allocation_cache_policy_CFG NULL
 
 #endif

lib/config/defaults.h

@@ -51,7 +51,7 @@
 #define DEFAULT_FALLBACK_TO_LOCAL_LOCKING 1
 #define DEFAULT_FALLBACK_TO_CLUSTERED_LOCKING 1
 #define DEFAULT_WAIT_FOR_LOCKS 1
-#define DEFAULT_LOCK_RETRIES 3
+#define DEFAULT_LVMLOCKD_LOCK_RETRIES 3
 #define DEFAULT_PRIORITISE_WRITE_LOCKS 1
 #define DEFAULT_USE_MLOCKALL 0
 #define DEFAULT_METADATA_READ_ONLY 0
@@ -63,6 +63,8 @@
 #define DEFAULT_MIRROR_LOG_FAULT_POLICY "allocate"
 #define DEFAULT_MIRROR_IMAGE_FAULT_POLICY "remove"
 #define DEFAULT_MIRROR_MAX_IMAGES 8 /* limited by kernel DM_KCOPYD_MAX_REGIONS */
+#define DEFAULT_RAID_MAX_IMAGES 8
+
 #define DEFAULT_RAID_FAULT_POLICY "warn"
 
 #define DEFAULT_DMEVENTD_RAID_LIB "libdevmapper-event-lvm2raid.so"
@@ -88,8 +90,8 @@
 #  define DEFAULT_THIN_CHECK_OPTIONS_CONFIG "#S" DEFAULT_THIN_CHECK_OPTION1
 #endif
 
-#define DEFAULT_THIN_REPAIR_OPTIONS ""
-#define DEFAULT_THIN_REPAIR_OPTIONS_CONFIG NULL
+#define DEFAULT_THIN_REPAIR_OPTION1 ""
+#define DEFAULT_THIN_REPAIR_OPTIONS_CONFIG "#S" DEFAULT_THIN_REPAIR_OPTION1
 #define DEFAULT_THIN_POOL_METADATA_REQUIRE_SEPARATE_PVS 0
 #define DEFAULT_THIN_POOL_MAX_METADATA_SIZE (16 * 1024 * 1024)  /* KB */
 #define DEFAULT_THIN_POOL_MIN_METADATA_SIZE 2048  /* KB */
@@ -101,16 +103,24 @@
 #define DEFAULT_THIN_POOL_ZERO 1
 #define DEFAULT_POOL_METADATA_SPARE 1 /* thin + cache */
 
-#define DEFAULT_CACHE_CHECK_OPTION1 "-q"
-#define DEFAULT_CACHE_CHECK_OPTIONS_CONFIG "#S" DEFAULT_CACHE_CHECK_OPTION1
-#define DEFAULT_CACHE_REPAIR_OPTIONS ""
-#define DEFAULT_CACHE_REPAIR_OPTIONS_CONFIG NULL
+#ifdef CACHE_CHECK_NEEDS_CHECK
+#  define DEFAULT_CACHE_CHECK_OPTION1 "-q"
+#  define DEFAULT_CACHE_CHECK_OPTION2 "--clear-needs-check-flag"
+#  define DEFAULT_CACHE_CHECK_OPTIONS_CONFIG "#S" DEFAULT_CACHE_CHECK_OPTION1 "#S" DEFAULT_CACHE_CHECK_OPTION2
+#else
+#  define DEFAULT_CACHE_CHECK_OPTION1 "-q"
+#  define DEFAULT_CACHE_CHECK_OPTION2 ""
+#  define DEFAULT_CACHE_CHECK_OPTIONS_CONFIG "#S" DEFAULT_CACHE_CHECK_OPTION1
+#endif
+
+#define DEFAULT_CACHE_REPAIR_OPTION1 ""
+#define DEFAULT_CACHE_REPAIR_OPTIONS_CONFIG "#S" DEFAULT_CACHE_REPAIR_OPTION1
 #define DEFAULT_CACHE_POOL_METADATA_REQUIRE_SEPARATE_PVS 0
 #define DEFAULT_CACHE_POOL_CHUNK_SIZE 64 /* KB */
 #define DEFAULT_CACHE_POOL_MIN_METADATA_SIZE 2048  /* KB */
 #define DEFAULT_CACHE_POOL_MAX_METADATA_SIZE (16 * 1024 * 1024)  /* KB */
-#define DEFAULT_CACHE_POOL_CACHEMODE "writethrough"
-#define DEFAULT_CACHE_POOL_POLICY "mq"
+#define DEFAULT_CACHE_POLICY "mq"
+#define DEFAULT_CACHE_MODE "writethrough"
 
 #define DEFAULT_UMASK 0077
 
@@ -144,11 +154,6 @@
 #  define DEFAULT_LOG_FACILITY LOG_USER
 #endif
 
-#define DEFAULT_LOGGED_DEBUG_CLASSES (LOG_CLASS_MEM | LOG_CLASS_DEVS | \
-    LOG_CLASS_ACTIVATION | LOG_CLASS_ALLOC | LOG_CLASS_LVMETAD | \
-    LOG_CLASS_METADATA | LOG_CLASS_CACHE | LOG_CLASS_LOCKING | \
-    LOG_CLASS_LVMPOLLD)
-
 #define DEFAULT_SYSLOG 1
 #define DEFAULT_VERBOSE 0
 #define DEFAULT_SILENT 0
@@ -196,6 +201,8 @@
 #define DEFAULT_REP_LIST_ITEM_SEPARATOR ","
 #define DEFAULT_TIME_FORMAT "%Y-%m-%d %T %z"
 
+#define DEFAULT_COMPACT_OUTPUT_COLS ""
+
 #define DEFAULT_LVS_COLS "lv_name,vg_name,lv_attr,lv_size,pool_lv,origin,data_percent,metadata_percent,move_pv,mirror_log,copy_percent,convert_lv"
 #define DEFAULT_VGS_COLS "vg_name,pv_count,lv_count,snap_count,vg_attr,vg_size,vg_free"
 #define DEFAULT_PVS_COLS "pv_name,vg_name,pv_fmt,pv_attr,pv_size,pv_free"

lib/datastruct/str_list.c

@@ -16,11 +16,14 @@
 #include "lib.h"
 #include "str_list.h"
 
+#include <ctype.h>
+
 struct dm_list *str_list_create(struct dm_pool *mem)
 {
 	struct dm_list *sl;
 
-	if (!(sl = dm_pool_alloc(mem, sizeof(struct dm_list)))) {
+	if (!(sl = mem ? dm_pool_alloc(mem, sizeof(struct dm_list))
+		       : dm_malloc(sizeof(struct dm_list)))) {
 		log_errno(ENOMEM, "str_list allocation failed");
 		return NULL;
 	}
@@ -37,7 +40,8 @@ static int _str_list_add_no_dup_check(struct dm_pool *mem, struct dm_list *sll,
 	if (!str)
 		return_0;
 
-	if (!(sln = dm_pool_alloc(mem, sizeof(*sln))))
+	if (!(sln = mem ? dm_pool_alloc(mem, sizeof(*sln))
+			: dm_malloc(sizeof(*sln))))
 		return_0;
 
 	sln->str = str;
@@ -158,3 +162,101 @@ int str_list_lists_equal(const struct dm_list *sll, const struct dm_list *sll2)
 
 	return 1;
 }
+
+char *str_list_to_str(struct dm_pool *mem, const struct dm_list *list,
+		      const char *delim)
+{
+	size_t delim_len = strlen(delim);
+	unsigned list_size = dm_list_size(list);
+	struct dm_str_list *sl;
+	char *str, *p;
+	size_t len = 0;
+	unsigned i = 0;
+
+	dm_list_iterate_items(sl, list)
+		len += strlen(sl->str);
+	if (list_size > 1)
+		len += ((list_size - 1) * delim_len);
+
+	str = mem ? dm_pool_alloc(mem, len+1) : dm_malloc(len+1);
+	if (!str) {
+		log_error("str_list_to_str: string allocation failed.");
+		return NULL;
+	}
+	str[len] = '\0';
+	p = str;
+
+	dm_list_iterate_items(sl, list) {
+		len = strlen(sl->str);
+		memcpy(p, sl->str, len);
+		p += len;
+
+		if (++i != list_size) {
+			memcpy(p, delim, delim_len);
+			p += delim_len;
+		}
+	}
+
+	return str;
+}
+
+struct dm_list *str_to_str_list(struct dm_pool *mem, const char *str,
+				const char *delim, int ignore_multiple_delim)
+{
+	size_t delim_len = strlen(delim);
+	struct dm_list *list;
+	const char *p1, *p2, *next;
+	char *str_item;
+	size_t len;
+
+	if (!(list = str_list_create(mem))) {
+		log_error("str_to_str_list: string list allocation failed.");
+		return NULL;
+	}
+
+	p1 = p2 = str;
+	while (*p1) {
+		if (!(p2 = strstr(p1, delim)))
+			next = p2 = str + strlen(str);
+		else
+			next = p2 + delim_len;
+
+		len = p2 - p1;
+		str_item = mem ? dm_pool_alloc(mem, len+1) : dm_malloc(len+1);
+		if (!str_item) {
+			log_error("str_to_str_list: string list item allocation failed.");
+			goto bad;
+		}
+		memcpy(str_item, p1, len);
+		str_item[len] = '\0';
+
+		if (!str_list_add_no_dup_check(mem, list, str_item))
+			goto_bad;
+
+		if (ignore_multiple_delim) {
+			while (!strncmp(next, delim, delim_len))
+				next += delim_len;
+		}
+
+		p1 = next;
+	}
+
+	return list;
+bad:
+	if (mem)
+		dm_pool_free(mem, list);
+	return NULL;
+}
+
+void str_list_destroy(struct dm_list *list, int deallocate_strings)
+{
+	struct dm_str_list *sl, *tmp_sl;
+
+	dm_list_iterate_items_safe(sl, tmp_sl, list) {
+		dm_list_del(&sl->list);
+		if (deallocate_strings)
+			dm_free((char *)sl->str);
+		dm_free(sl);
+	}
+	dm_free(list);
+}

lib/datastruct/str_list.h

@@ -30,5 +30,9 @@ int str_list_match_list(const struct dm_list *sll, const struct dm_list *sll2, c
 int str_list_lists_equal(const struct dm_list *sll, const struct dm_list *sll2);
 int str_list_dup(struct dm_pool *mem, struct dm_list *sllnew,
 		 const struct dm_list *sllold);
+char *str_list_to_str(struct dm_pool *mem, const struct dm_list *list, const char *delim);
+struct dm_list *str_to_str_list(struct dm_pool *mem, const char *str, const char *delim, int ignore_multiple_delim);
+/* Only for lists which were *not* allocated from the mem pool! */
+void str_list_destroy(struct dm_list *list, int deallocate_strings);
 
 #endif

lib/device/dev-cache.c

@@ -71,6 +71,16 @@ static void _dev_init(struct device *dev, int max_error_count)
 	dm_list_init(&dev->open_list);
 }
 
+void dev_destroy_file(struct device *dev)
+{
+	if (!(dev->flags & DEV_ALLOCED))
+		return;
+
+	dm_free((void *) dm_list_item(dev->aliases.n, struct dm_str_list)->str);
+	dm_free(dev->aliases.n);
+	dm_free(dev);
+}
+
 struct device *dev_create_file(const char *filename, struct device *dev,
 			       struct dm_str_list *alias, int use_malloc)
 {
@@ -331,11 +341,13 @@ static int _add_alias(struct device *dev, const char *path)
 	if (!dm_list_empty(&dev->aliases)) {
 		oldpath = dm_list_item(dev->aliases.n, struct dm_str_list)->str;
 		prefer_old = _compare_paths(path, oldpath);
-		log_debug_devs("%s: Aliased to %s in device cache%s",
-			       path, oldpath, prefer_old ? "" : " (preferred name)");
+		log_debug_devs("%s: Aliased to %s in device cache%s (%d:%d)",
+			       path, oldpath, prefer_old ? "" : " (preferred name)",
+			       (int) MAJOR(dev->dev), (int) MINOR(dev->dev));
 
 	} else
-		log_debug_devs("%s: Added to device cache", path);
+		log_debug_devs("%s: Added to device cache (%d:%d)", path,
+			       (int) MAJOR(dev->dev), (int) MINOR(dev->dev));
 
 	if (prefer_old)
 		dm_list_add(&dev->aliases, &sl->list);
@@ -681,10 +693,12 @@ static int _init_preferred_names(struct cmd_context *cmd)
 
 	_cache.preferred_names_matcher = NULL;
 
-	if (!(cn = find_config_tree_node(cmd, devices_preferred_names_CFG, NULL)) ||
+	if (!(cn = find_config_tree_array(cmd, devices_preferred_names_CFG, NULL)) ||
 	    cn->v->type == DM_CFG_EMPTY_ARRAY) {
-		log_very_verbose("devices/preferred_names not found in config file: "
-				 "using built-in preferences");
+		log_very_verbose("devices/preferred_names %s: "
+				 "using built-in preferences",
+				 cn && cn->v->type == DM_CFG_EMPTY_ARRAY ? "is empty"
+									 : "not found in config");
 		return 1;
 	}
 
@@ -934,6 +948,7 @@ struct device *dev_cache_get(const char *name, struct dev_filter *f)
 {
 	struct stat buf;
 	struct device *d = (struct device *) dm_hash_lookup(_cache.names, name);
+	int info_available = 0;
 
 	if (d && (d->flags & DEV_REGULAR))
 		return d;
@@ -943,8 +958,9 @@ struct device *dev_cache_get(const char *name, struct dev_filter *f)
 		if (d)
 			dm_hash_remove(_cache.names, name);
 		log_sys_very_verbose("stat", name);
-		return NULL;
-	}
+		d = NULL;
+	} else
+		info_available = 1;
 
 	if (d && (buf.st_rdev != d->dev)) {
 		dm_hash_remove(_cache.names, name);
@@ -952,7 +968,7 @@ struct device *dev_cache_get(const char *name, struct dev_filter *f)
 	}
 
 	if (!d) {
-		_insert(name, &buf, 0, obtain_device_list_from_udev());
+		_insert(name, info_available ? &buf : NULL, 0, obtain_device_list_from_udev());
 		d = (struct device *) dm_hash_lookup(_cache.names, name);
 		if (!d) {
 			_full_scan(0);

lib/device/dev-dasd.c

@@ -0,0 +1,111 @@
+/*
+ * Copyright (C) 2015 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 "dev-type.h"
+#include <sys/ioctl.h>
+
+#ifdef __linux__
+
+/*
+ * Interface taken from kernel header arch/s390/include/uapi/asm/dasd.h
+ */
+
+/* 
+ * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com>
+ * Copyright IBM Corp. 1999, 2000
+ * EMC Symmetrix ioctl Copyright EMC Corporation, 2008
+ * Author.........: Nigel Hislop <hislop_nigel@emc.com>
+ */
+
+#define DASD_IOCTL_LETTER 'D'
+#define DASD_API_VERSION 6
+
+/* 
+ * struct dasd_information2_t
+ * represents any data about the device, which is visible to userspace.
+ *  including foramt and featueres.
+ */
+typedef struct dasd_information2_t {
+	unsigned int devno;		/* S/390 devno */
+	unsigned int real_devno;	/* for aliases */
+	unsigned int schid;		/* S/390 subchannel identifier */
+	unsigned int cu_type  : 16;	/* from SenseID */
+	unsigned int cu_model :  8;	/* from SenseID */
+	unsigned int dev_type : 16;	/* from SenseID */
+	unsigned int dev_model : 8;	/* from SenseID */
+	unsigned int open_count;
+	unsigned int req_queue_len;
+	unsigned int chanq_len;		/* length of chanq */
+	char type[4];			/* from discipline.name, 'none' for unknown */
+	unsigned int status;		/* current device level */
+	unsigned int label_block;	/* where to find the VOLSER */
+	unsigned int FBA_layout;	/* fixed block size (like AIXVOL) */
+	unsigned int characteristics_size;
+	unsigned int confdata_size;
+	char characteristics[64];	/* from read_device_characteristics */
+	char configuration_data[256];	/* from read_configuration_data */
+	unsigned int format;		/* format info like formatted/cdl/ldl/... */
+	unsigned int features;		/* dasd features like 'ro',... */
+	unsigned int reserved0;		/* reserved for further use ,... */
+	unsigned int reserved1;		/* reserved for further use ,... */
+	unsigned int reserved2;		/* reserved for further use ,... */
+	unsigned int reserved3;		/* reserved for further use ,... */
+	unsigned int reserved4;		/* reserved for further use ,... */
+	unsigned int reserved5;		/* reserved for further use ,... */
+	unsigned int reserved6;		/* reserved for further use ,... */
+	unsigned int reserved7;		/* reserved for further use ,... */
+} dasd_information2_t;
+
+#define DASD_FORMAT_CDL  2
+
+/* Get information on a dasd device (enhanced) */
+#define BIODASDINFO2   _IOR(DASD_IOCTL_LETTER,3,dasd_information2_t)
+
+/*
+ * End of included interface.
+ */
+
+int dasd_is_cdl_formatted(struct device *dev)
+{
+	int ret = 0;
+	dasd_information2_t dasd_info2;
+
+	if (!dev_open_readonly(dev))
+		return_0;
+
+	if (ioctl(dev->fd, BIODASDINFO2, &dasd_info2)) {
+		log_sys_error("ioctl BIODASDINFO2", dev_name(dev));
+		goto out;
+	}
+
+	if (dasd_info2.format == DASD_FORMAT_CDL)
+		ret = 1;
+
+out:
+	if (!dev_close(dev))
+		stack;
+
+	return ret;
+}
+
+#else
+
+int dasd_is_cdl_formatted(struct device *dev)
+{
+	return 0;
+}
+
+#endif

lib/device/dev-io.c

@@ -586,12 +586,8 @@ static void _close(struct device *dev)
 
 	log_debug_devs("Closed %s", dev_name(dev));
 
-	if (dev->flags & DEV_ALLOCED) {
-		dm_free((void *) dm_list_item(dev->aliases.n, struct dm_str_list)->
-			 str);
-		dm_free(dev->aliases.n);
-		dm_free(dev);
-	}
+	if (dev->flags & DEV_ALLOCED)
+		dev_destroy_file(dev);
 }
 
 static int _dev_close(struct device *dev, int immediate)

lib/device/dev-type.c

@@ -363,7 +363,7 @@ static int _native_dev_is_partitioned(struct dev_types *dt, struct device *dev)
 		return 0;
 
 	/* Unpartitioned DASD devices are not supported. */
-	if (MAJOR(dev->dev) == dt->dasd_major)
+	if ((MAJOR(dev->dev) == dt->dasd_major) && dasd_is_cdl_formatted(dev))
 		return 1;
 
 	if (!dev_open_readonly_quiet(dev)) {
@@ -528,12 +528,14 @@ static inline int _type_in_flag_list(const char *type, uint32_t flag_list)
 		((flag_list & TYPE_DM_SNAPSHOT_COW) && !strcmp(type, "DM_snapshot_cow")));
 }
 
+#define MSG_FAILED_SIG_OFFSET "Failed to get offset of the %s signature on %s."
+#define MSG_FAILED_SIG_LENGTH "Failed to get length of the %s signature on %s."
+#define MSG_WIPING_SKIPPED " Wiping skipped."
+
 static int _blkid_wipe(blkid_probe probe, struct device *dev, const char *name,
 		       uint32_t types_to_exclude, uint32_t types_no_prompt,
 		       int yes, force_t force)
 {
-	static const char _msg_failed_offset[] = "Failed to get offset of the %s signature on %s.";
-	static const char _msg_failed_length[] = "Failed to get length of the %s signature on %s.";
 	static const char _msg_wiping[] = "Wiping %s signature on %s.";
 	const char *offset = NULL, *type = NULL, *magic = NULL,
 		   *usage = NULL, *label = NULL, *uuid = NULL;
@@ -544,21 +546,41 @@ static int _blkid_wipe(blkid_probe probe, struct device *dev, const char *name,
 		if (_type_in_flag_list(type, types_to_exclude))
 			return 2;
 		if (blkid_probe_lookup_value(probe, "SBMAGIC_OFFSET", &offset, NULL)) {
-			log_error(_msg_failed_offset, type, name);
-			return 0;
+			if (force < DONT_PROMPT) {
+				log_error(MSG_FAILED_SIG_OFFSET, type, name);
+				return 0;
+			} else {
+				log_error("WARNING: " MSG_FAILED_SIG_OFFSET MSG_WIPING_SKIPPED, type, name);
+				return 2;
+			}
 		}
 		if (blkid_probe_lookup_value(probe, "SBMAGIC", &magic, &len)) {
-			log_error(_msg_failed_length, type, name);
-			return 0;
+			if (force < DONT_PROMPT) {
+				log_error(MSG_FAILED_SIG_LENGTH, type, name);
+				return 0;
+			} else {
+				log_warn("WARNING: " MSG_FAILED_SIG_LENGTH MSG_WIPING_SKIPPED, type, name);
+				return 2;
+			}
 		}
 	} else if (!blkid_probe_lookup_value(probe, "PTTYPE", &type, NULL)) {
 		if (blkid_probe_lookup_value(probe, "PTMAGIC_OFFSET", &offset, NULL)) {
-			log_error(_msg_failed_offset, type, name);
-			return 0;
+			if (force < DONT_PROMPT) {
+				log_error(MSG_FAILED_SIG_OFFSET, type, name);
+				return 0;
+			} else {
+				log_warn("WARNING: " MSG_FAILED_SIG_OFFSET MSG_WIPING_SKIPPED, type, name);
+				return 2;
+			}
 		}
 		if (blkid_probe_lookup_value(probe, "PTMAGIC", &magic, &len)) {
-			log_error(_msg_failed_length, type, name);
-			return 0;
+			if (force < DONT_PROMPT) {
+				log_error(MSG_FAILED_SIG_LENGTH, type, name);
+				return 0;
+			} else {
+				log_warn("WARNING: " MSG_FAILED_SIG_LENGTH MSG_WIPING_SKIPPED, type, name);
+				return 2;
+			}
 		}
 		usage = "partition table";
 	} else
@@ -629,8 +651,13 @@ static int _wipe_known_signatures_with_blkid(struct device *dev, const char *nam
 						 BLKID_SUBLKS_BADCSUM);
 
 	while (!blkid_do_probe(probe)) {
-		if ((r_wipe = _blkid_wipe(probe, dev, name, types_to_exclude, types_no_prompt, yes, force)) == 1)
+		if ((r_wipe = _blkid_wipe(probe, dev, name, types_to_exclude, types_no_prompt, yes, force)) == 1) {
 			(*wiped)++;
+			if (blkid_probe_step_back(probe)) {
+				log_error("Failed to step back blkid probe to check just wiped signature.");
+				goto out;
+			}
+		}
 		/* do not count excluded types */
 		if (r_wipe != 2)
 			found++;
@@ -712,13 +739,20 @@ int wipe_known_signatures(struct cmd_context *cmd, struct device *dev,
 			  uint32_t types_no_prompt, int yes, force_t force,
 			  int *wiped)
 {
+	int blkid_wiping_enabled = find_config_tree_bool(cmd, allocation_use_blkid_wiping_CFG, NULL);
+
 #ifdef BLKID_WIPING_SUPPORT
-	if (find_config_tree_bool(cmd, allocation_use_blkid_wiping_CFG, NULL))
+	if (blkid_wiping_enabled)
 		return _wipe_known_signatures_with_blkid(dev, name,
 							 types_to_exclude,
 							 types_no_prompt,
 							 yes, force, wiped);
 #endif
+	if (blkid_wiping_enabled) {
+		log_warn("allocation/use_blkid_wiping=1 configuration setting is set "
+			 "while LVM is not compiled with blkid wiping support.");
+		log_warn("Falling back to native LVM signature detection.");
+	}
 	return _wipe_known_signatures_with_lvm(dev, name,
 					       types_to_exclude,
 					       types_no_prompt,

lib/device/dev-type.h

@@ -59,6 +59,7 @@ int major_is_scsi_device(struct dev_types *dt, int major);
 int dev_is_md(struct device *dev, uint64_t *sb);
 int dev_is_swap(struct device *dev, uint64_t *signature);
 int dev_is_luks(struct device *dev, uint64_t *signature);
+int dasd_is_cdl_formatted(struct device *dev);
 
 /* Signature wiping. */
 #define TYPE_LVM1_MEMBER	0x001

lib/device/device.h

@@ -123,6 +123,7 @@ void dev_flush(struct device *dev);
 
 struct device *dev_create_file(const char *filename, struct device *dev,
 			       struct dm_str_list *alias, int use_malloc);
+void dev_destroy_file(struct device *dev);
 
 /* Return a valid device name from the alias list; NULL otherwise */
 const char *dev_name_confirmed(struct device *dev, int quiet);

lib/display/display.c

@@ -24,10 +24,6 @@
 
 #include <stdarg.h>
 
-#define SIZE_BUF 128
-
-typedef enum { SIZE_LONG = 0, SIZE_SHORT = 1, SIZE_UNIT = 2 } size_len_t;
-
 static const struct {
 	alloc_policy_t alloc;
 	const char str[14]; /* must be changed when size extends 13 chars */
@@ -146,164 +142,30 @@ const char *display_lvname(const struct logical_volume *lv)
 	return name;
 }
 
-#define BASE_UNKNOWN 0
-#define BASE_SHARED 1
-#define BASE_1024 8
-#define BASE_1000 15
-#define BASE_SPECIAL 21
-#define NUM_UNIT_PREFIXES 6
-#define NUM_SPECIAL 3
-
 /* Size supplied in sectors */
 static const char *_display_size(const struct cmd_context *cmd,
-				 uint64_t size, size_len_t sl)
+				 uint64_t size, dm_size_suffix_t suffix_type)
 {
-	unsigned base = BASE_UNKNOWN;
-	unsigned s;
-	int suffix, precision;
-	uint64_t byte = UINT64_C(0);
-	uint64_t units = UINT64_C(1024);
-	char *size_buf = NULL;
-	const char * const size_str[][3] = {
-		/* BASE_UNKNOWN */
-		{"         ", "   ", " "},	/* [0] */
-
-		/* BASE_SHARED - Used if cmd->si_unit_consistency = 0 */
-		{" Exabyte", " EB", "E"},	/* [1] */
-		{" Petabyte", " PB", "P"},	/* [2] */
-		{" Terabyte", " TB", "T"},	/* [3] */
-		{" Gigabyte", " GB", "G"},	/* [4] */
-		{" Megabyte", " MB", "M"},	/* [5] */
-		{" Kilobyte", " KB", "K"},	/* [6] */
-		{" Byte    ", " B", "B"},	/* [7] */
-
-		/* BASE_1024 - Used if cmd->si_unit_consistency = 1 */
-		{" Exbibyte", " EiB", "e"},	/* [8] */
-		{" Pebibyte", " PiB", "p"},	/* [9] */
-		{" Tebibyte", " TiB", "t"},	/* [10] */
-		{" Gibibyte", " GiB", "g"},	/* [11] */
-		{" Mebibyte", " MiB", "m"},	/* [12] */
-		{" Kibibyte", " KiB", "k"},	/* [13] */
-		{" Byte    ", " B", "b"},	/* [14] */
-
-		/* BASE_1000 - Used if cmd->si_unit_consistency = 1 */
-		{" Exabyte",  " EB", "E"},	/* [15] */
-		{" Petabyte", " PB", "P"},	/* [16] */
-		{" Terabyte", " TB", "T"},	/* [17] */
-		{" Gigabyte", " GB", "G"},	/* [18] */
-		{" Megabyte", " MB", "M"},	/* [19] */
-		{" Kilobyte", " kB", "K"},	/* [20] */
-
-		/* BASE_SPECIAL */
-		{" Byte    ", " B ", "B"},	/* [21] (shared with BASE_1000) */
-		{" Units   ", " Un", "U"},	/* [22] */
-		{" Sectors ", " Se", "S"},	/* [23] */
-	};
-
-	if (!(size_buf = dm_pool_alloc(cmd->mem, SIZE_BUF))) {
-		log_error("no memory for size display buffer");
-		return "";
-	}
-
-	suffix = cmd->current_settings.suffix;
-
-	if (!cmd->si_unit_consistency) {
-		/* Case-independent match */
-		for (s = 0; s < NUM_UNIT_PREFIXES; s++)
-			if (toupper((int) cmd->current_settings.unit_type) ==
-			    *size_str[BASE_SHARED + s][2]) {
-				base = BASE_SHARED;
-				break;
-			}
-	} else {
-		/* Case-dependent match for powers of 1000 */
-		for (s = 0; s < NUM_UNIT_PREFIXES; s++)
-			if (cmd->current_settings.unit_type ==
-			    *size_str[BASE_1000 + s][2]) {
-				base = BASE_1000;
-				break;
-			}
-
-		/* Case-dependent match for powers of 1024 */
-		if (base == BASE_UNKNOWN)
-			for (s = 0; s < NUM_UNIT_PREFIXES; s++)
-			if (cmd->current_settings.unit_type ==
-			    *size_str[BASE_1024 + s][2]) {
-				base = BASE_1024;
-				break;
-			}
-	}
-
-	if (base == BASE_UNKNOWN)
-		/* Check for special units - s, b or u */
-		for (s = 0; s < NUM_SPECIAL; s++)
-			if (toupper((int) cmd->current_settings.unit_type) ==
-			    *size_str[BASE_SPECIAL + s][2]) {
-				base = BASE_SPECIAL;
-				break;
-			}
-
-	if (size == UINT64_C(0)) {
-		if (base == BASE_UNKNOWN)
-			s = 0;
-		sprintf(size_buf, "0%s", suffix ? size_str[base + s][sl] : "");
-		return size_buf;
-	}
-
-	size *= UINT64_C(512);
-
-	if (base != BASE_UNKNOWN)
-		byte = cmd->current_settings.unit_factor;
-	else {
-		/* Human-readable style */
-		if (cmd->current_settings.unit_type == 'H') {
-			units = UINT64_C(1000);
-			base = BASE_1000;
-		} else {
-			units = UINT64_C(1024);
-			base = BASE_1024;
-		}
-
-		if (!cmd->si_unit_consistency)
-			base = BASE_SHARED;
-
-		byte = units * units * units * units * units * units;
-
-		for (s = 0; s < NUM_UNIT_PREFIXES && size < byte; s++)
-			byte /= units;
-
-		suffix = 1;
-	}
-
-	/* FIXME Make precision configurable */
-	switch (toupper(*size_str[base + s][SIZE_UNIT])) {
-	case 'B':
-	case 'S':
-		precision = 0;
-		break;
-	default:
-		precision = 2;
-	}
-
-	snprintf(size_buf, SIZE_BUF - 1, "%.*f%s", precision,
-		 (double) size / byte, suffix ? size_str[base + s][sl] : "");
-
-	return size_buf;
+	return dm_size_to_string(cmd->mem, size, cmd->current_settings.unit_type,
+				 cmd->si_unit_consistency, 
+				 cmd->current_settings.unit_factor,
+				 cmd->current_settings.suffix,
+				 suffix_type);
 }
 
 const char *display_size_long(const struct cmd_context *cmd, uint64_t size)
 {
-	return _display_size(cmd, size, SIZE_LONG);
+	return _display_size(cmd, size, DM_SIZE_LONG);
 }
 
 const char *display_size_units(const struct cmd_context *cmd, uint64_t size)
 {
-	return _display_size(cmd, size, SIZE_UNIT);
+	return _display_size(cmd, size, DM_SIZE_UNIT);
 }
 
 const char *display_size(const struct cmd_context *cmd, uint64_t size)
 {
-	return _display_size(cmd, size, SIZE_SHORT);
+	return _display_size(cmd, size, DM_SIZE_SHORT);
 }
 
 void pvdisplay_colons(const struct physical_volume *pv)

lib/error/errseg.c

@@ -95,7 +95,7 @@ struct segment_type *init_error_segtype(struct cmd_context *cmd)
 		return_NULL;
 
 	segtype->ops = &_error_ops;
-	segtype->name = "error";
+	segtype->name = SEG_TYPE_NAME_ERROR;
 	segtype->flags = SEG_CAN_SPLIT | SEG_VIRTUAL | SEG_CANNOT_BE_ZEROED;
 
 	log_very_verbose("Initialised segtype: %s", segtype->name);

lib/filters/filter-fwraid.c

@@ -61,6 +61,8 @@ static int _dev_is_fwraid(struct device *dev)
 	return 0;
 }
 
+#define MSG_SKIPPING "%s: Skipping firmware RAID component device"
+
 static int _ignore_fwraid(struct dev_filter *f __attribute__((unused)),
 			   struct device *dev)
 {
@@ -72,8 +74,11 @@ static int _ignore_fwraid(struct dev_filter *f __attribute__((unused)),
 	ret = _dev_is_fwraid(dev);
 
 	if (ret == 1) {
-		log_debug_devs("%s: Skipping firmware RAID component device [%s:%p]",
-				dev_name(dev), dev_ext_name(dev), dev->ext.handle);
+		if (dev->ext.src == DEV_EXT_NONE)
+			log_debug_devs(MSG_SKIPPING, dev_name(dev));
+		else
+			log_debug_devs(MSG_SKIPPING " [%s:%p]", dev_name(dev),
+					dev_ext_name(dev), dev->ext.handle);
 		return 0;
 	}
 

lib/filters/filter-md.c

@@ -18,6 +18,8 @@
 
 #ifdef __linux__
 
+#define MSG_SKIPPING "%s: Skipping md component device"
+
 static int _ignore_md(struct dev_filter *f __attribute__((unused)),
 		      struct device *dev)
 {
@@ -29,8 +31,11 @@ static int _ignore_md(struct dev_filter *f __attribute__((unused)),
 	ret = dev_is_md(dev, NULL);
 
 	if (ret == 1) {
-		log_debug_devs("%s: Skipping md component device [%s:%p]",
-				dev_name(dev), dev_ext_name(dev), dev->ext.handle);
+		if (dev->ext.src == DEV_EXT_NONE)
+			log_debug_devs(MSG_SKIPPING, dev_name(dev));
+		else
+			log_debug_devs(MSG_SKIPPING " [%s:%p]", dev_name(dev),
+					dev_ext_name(dev), dev->ext.handle);
 		return 0;
 	}
 

lib/filters/filter-mpath.c

@@ -244,11 +244,16 @@ static int _dev_is_mpath(struct dev_filter *f, struct device *dev)
 	return 0;
 }
 
+#define MSG_SKIPPING "%s: Skipping mpath component device"
+
 static int _ignore_mpath(struct dev_filter *f, struct device *dev)
 {
 	if (_dev_is_mpath(f, dev) == 1) {
-		log_debug_devs("%s: Skipping mpath component device [%s:%p]",
-				dev_name(dev), dev_ext_name(dev), dev->ext.handle);
+		if (dev->ext.src == DEV_EXT_NONE)
+			log_debug_devs(MSG_SKIPPING, dev_name(dev));
+		else
+			log_debug_devs(MSG_SKIPPING " [%s:%p]", dev_name(dev),
+					dev_ext_name(dev), dev->ext.handle);
 		return 0;
 	}
 

lib/filters/filter-partitioned.c

@@ -16,13 +16,18 @@
 #include "lib.h"
 #include "filter.h"
 
+#define MSG_SKIPPING "%s: Skipping: Partition table signature found"
+
 static int _passes_partitioned_filter(struct dev_filter *f, struct device *dev)
 {
 	struct dev_types *dt = (struct dev_types *) f->private;
 
 	if (dev_is_partitioned(dt, dev)) {
-		log_debug_devs("%s: Skipping: Partition table signature found [%s:%p]",
-			       dev_name(dev), dev_ext_name(dev), dev->ext.handle);
+		if (dev->ext.src == DEV_EXT_NONE)
+			log_debug_devs(MSG_SKIPPING, dev_name(dev));
+		else
+			log_debug_devs(MSG_SKIPPING " [%s:%p]", dev_name(dev),
+					dev_ext_name(dev), dev->ext.handle);
 		return 0;
 	}
 

lib/filters/filter-usable.c

@@ -29,22 +29,19 @@ static int _native_check_pv_min_size(struct device *dev)
 
 	/* Check it's accessible */
 	if (!dev_open_readonly_quiet(dev)) {
-		log_debug_devs("%s: Skipping: open failed [%s:%p]",
-				dev_name(dev), dev_ext_name(dev), dev->ext.handle);
+		log_debug_devs("%s: Skipping: open failed", dev_name(dev));
 		return 0;
 	}
 
 	/* Check it's not too small */
 	if (!dev_get_size(dev, &size)) {
-		log_debug_devs("%s: Skipping: dev_get_size failed [%s:%p]",
-				dev_name(dev), dev_ext_name(dev), dev->ext.handle);
+		log_debug_devs("%s: Skipping: dev_get_size failed", dev_name(dev));
 		goto out;
 	}
 
 	if (size < pv_min_size()) {
-		log_debug_devs("%s: Skipping: %s [%s:%p]", dev_name(dev),
-				_too_small_to_hold_pv_msg,
-				dev_ext_name(dev), dev->ext.handle);
+		log_debug_devs("%s: Skipping: %s", dev_name(dev),
+				_too_small_to_hold_pv_msg);
 		goto out;
 	}
 
@@ -115,26 +112,11 @@ static int _passes_usable_filter(struct dev_filter *f, struct device *dev)
 {
 	filter_mode_t mode = *((filter_mode_t *) f->private);
 	struct dev_usable_check_params ucp = {0};
-	int r;
-
-	/* check if the device is not too small to hold a PV */
-	switch (mode) {
-		case FILTER_MODE_NO_LVMETAD:
-			/* fall through */
-		case FILTER_MODE_PRE_LVMETAD:
-			if (!_check_pv_min_size(dev))
-				return 0;
-			break;
-		case FILTER_MODE_POST_LVMETAD:
-			/* nothing to do here */
-			break;
-	}
+	int r = 1;
 
 	/* further checks are done on dm devices only */
-	if (!dm_is_dm_major(MAJOR(dev->dev)))
-		return 1;
-
-	switch (mode) {
+	if (dm_is_dm_major(MAJOR(dev->dev))) {
+		switch (mode) {
 		case FILTER_MODE_NO_LVMETAD:
 			ucp.check_empty = 1;
 			ucp.check_blocked = 1;
@@ -163,10 +145,25 @@ static int _passes_usable_filter(struct dev_filter *f, struct device *dev)
 			ucp.check_error_target = 0;
 			ucp.check_reserved = 0;
 			break;
+		}
+
+		if (!(r = device_is_usable(dev, ucp)))
+			log_debug_devs("%s: Skipping unusable device.", dev_name(dev));
 	}
 
-	if (!(r = device_is_usable(dev, ucp)))
-		log_debug_devs("%s: Skipping unusable device", dev_name(dev));
+	if (r) {
+		/* check if the device is not too small to hold a PV */
+		switch (mode) {
+		case FILTER_MODE_NO_LVMETAD:
+			/* fall through */
+		case FILTER_MODE_PRE_LVMETAD:
+			r = _check_pv_min_size(dev);
+			break;
+		case FILTER_MODE_POST_LVMETAD:
+			/* nothing to do here */
+			break;
+		}
+	}
 
 	return r;
 }

lib/format1/import-export.c

@@ -164,7 +164,7 @@ int export_pv(struct cmd_context *cmd, struct dm_pool *mem __attribute__((unused
 	/* Is VG already exported or being exported? */
 	if (vg && vg_is_exported(vg)) {
 		/* Does system_id need setting? */
-		if (!*vg->lvm1_system_id ||
+		if (!vg->lvm1_system_id || !*vg->lvm1_system_id ||
 		    strncmp(vg->lvm1_system_id, EXPORTED_TAG,
 			    sizeof(EXPORTED_TAG) - 1)) {
 			if (!generate_lvm1_system_id(cmd, (char *)pvd->system_id, EXPORTED_TAG))

lib/format1/import-extents.c

@@ -219,7 +219,7 @@ static int _read_linear(struct cmd_context *cmd, struct lv_map *lvm)
 	struct lv_segment *seg;
 	struct segment_type *segtype;
 
-	if (!(segtype = get_segtype_from_string(cmd, "striped")))
+	if (!(segtype = get_segtype_from_string(cmd, SEG_TYPE_NAME_STRIPED)))
 		return_0;
 
 	while (le < lvm->lv->le_count) {
@@ -281,7 +281,7 @@ static int _read_stripes(struct cmd_context *cmd, struct lv_map *lvm)
 
 	total_area_len = lvm->lv->le_count / lvm->stripes;
 
-	if (!(segtype = get_segtype_from_string(cmd, "striped")))
+	if (!(segtype = get_segtype_from_string(cmd, SEG_TYPE_NAME_STRIPED)))
 		return_0;
 
 	while (first_area_le < total_area_len) {

lib/format_pool/import_export.c

@@ -188,8 +188,7 @@ static int _add_stripe_seg(struct dm_pool *mem,
 
 	area_len = (usp->devs[0].blocks) / POOL_PE_SIZE;
 
-	if (!(segtype = get_segtype_from_string(lv->vg->cmd,
-						     "striped")))
+	if (!(segtype = get_segtype_from_string(lv->vg->cmd, SEG_TYPE_NAME_STRIPED)))
 		return_0;
 
 	if (!(seg = alloc_lv_segment(segtype, lv, *le_cur,
@@ -226,7 +225,7 @@ static int _add_linear_seg(struct dm_pool *mem,
 	unsigned j;
 	uint32_t area_len;
 
-	if (!(segtype = get_segtype_from_string(lv->vg->cmd, "striped")))
+	if (!(segtype = get_segtype_from_string(lv->vg->cmd, SEG_TYPE_NAME_STRIPED)))
 		return_0;
 
 	for (j = 0; j < usp->num_devs; j++) {

lib/format_text/format-text.c

@@ -453,8 +453,9 @@ static struct raw_locn *_find_vg_rlocn(struct device_area *dev_area,
 				   "not match expected name %s.", vgname);
 
       bad:
-	if ((info = lvmcache_info_from_pvid(dev_area->dev->pvid, 0)))
-		lvmcache_update_vgname_and_id(info, &vgsummary_orphan);
+	if ((info = lvmcache_info_from_pvid(dev_area->dev->pvid, 0)) &&
+	    !lvmcache_update_vgname_and_id(info, &vgsummary_orphan))
+		stack;
 
 	return NULL;
 }
@@ -653,7 +654,7 @@ static int _vg_write_raw(struct format_instance *fid, struct volume_group *vg,
 
 	if ((new_wrap && old_wrap) ||
 	    (rlocn && (new_wrap || old_wrap) && (new_end > rlocn->offset)) ||
-	    (mdac->rlocn.size >= mdah->size)) {
+	    (MDA_HEADER_SIZE + (rlocn ? rlocn->size : 0) + mdac->rlocn.size >= mdah->size)) {
 		log_error("VG %s metadata too large for circular buffer",
 			  vg->name);
 		goto out;
@@ -1338,6 +1339,7 @@ static int _text_pv_write(const struct format_type *fmt, struct physical_volume
 
 	label = lvmcache_get_label(info);
 	label->sector = pv->label_sector;
+	label->dev = pv->dev;
 
 	lvmcache_update_pv(info, pv, fmt);
 
@@ -2144,7 +2146,6 @@ static int _text_pv_add_metadata_area(const struct format_type *fmt,
 					goto bad;
 			}
 			/* Otherwise, give up and take any usable space. */
-			/* FIXME: We should probably check for some minimum MDA size here. */
 			else
 				mda_size = limit - mda_start;
 
@@ -2237,10 +2238,16 @@ static int _text_pv_add_metadata_area(const struct format_type *fmt,
 	if (limit_applied)
 		log_very_verbose("Using limited metadata area size on %s "
 				 "with value %" PRIu64 " (limited by %s of "
-				 "%" PRIu64 ").", pv_dev_name(pv),
+				 FMTu64 ").", pv_dev_name(pv),
 				  mda_size, limit_name, limit);
 
 	if (mda_size) {
+		if (mda_size < MDA_SIZE_MIN) {
+			log_error("Metadata area size too small: %" PRIu64" bytes. "
+				  "It must be at least %u bytes.", mda_size, MDA_SIZE_MIN);
+			goto bad;
+		}
+
 		/* Wipe metadata area with zeroes. */
 		if (!dev_set((struct device *) pv->dev, mda_start,
 			(size_t) ((mda_size > wipe_size) ?
@@ -2491,7 +2498,7 @@ struct format_type *create_text_format(struct cmd_context *cmd)
 		goto bad;
 	}
 
-	if ((cn = find_config_tree_node(cmd, metadata_dirs_CFG, NULL))) {
+	if ((cn = find_config_tree_array(cmd, metadata_dirs_CFG, NULL))) {
 		for (cv = cn->v; cv; cv = cv->next) {
 			if (cv->type != DM_CFG_STRING) {
 				log_error("Invalid string in config file: "

lib/format_text/import_vsn1.c

@@ -375,7 +375,7 @@ static int _read_segment(struct logical_volume *lv, const struct dm_config_node
 		return 0;
 	}
 
-	segtype_str = "striped";
+	segtype_str = SEG_TYPE_NAME_STRIPED;
 
 	if (!dm_config_get_str(sn_child, "type", &segtype_str)) {
 		log_error("Segment type must be a string.");
@@ -600,6 +600,25 @@ static int _read_lvnames(struct format_instance *fid __attribute__((unused)),
 		return 0;
 	}
 
+	/*
+	 * The LV lock_args string is generated in lvmlockd, and the content
+	 * depends on the lock_type.
+	 *
+	 * lock_type dlm does not use LV lock_args, so the LV lock_args field
+	 * is just set to "dlm".
+	 *
+	 * lock_type sanlock uses the LV lock_args field to save the
+	 * location on disk of that LV's sanlock lock.  The disk name is
+	 * specified in the VG lock_args.  The lock_args string begins
+	 * with a version number, e.g. 1.0.0, followed by a colon, followed
+	 * by a number.  The number is the offset on disk where sanlock is
+	 * told to find the LV's lock.
+	 * e.g. lock_args = 1.0.0:70254592
+	 * means that the lock is located at offset 70254592.
+	 *
+	 * The lvmlockd code for each specific lock manager also validates
+	 * the lock_args before using it to access the lock manager.
+	 */
 	if (dm_config_get_str(lvn, "lock_args", &str)) {
 		if (!(lv->lock_args = dm_pool_strdup(mem, str)))
 			return_0;
@@ -828,6 +847,27 @@ static struct volume_group *_read_vg(struct format_instance *fid,
 			goto bad;
 	}
 
+	/*
+	 * The VG lock_args string is generated in lvmlockd, and the content
+	 * depends on the lock_type.  lvmlockd begins the lock_args string
+	 * with a version number, e.g. 1.0.0, followed by a colon, followed
+	 * by a string that depends on the lock manager.  The string after
+	 * the colon is information needed to use the lock manager for the VG.
+	 *
+	 * For sanlock, the string is the name of the internal LV used to store
+	 * sanlock locks.  lvmlockd needs to know where the locks are located
+	 * so it can pass that location to sanlock which needs to access the locks.
+	 * e.g. lock_args = 1.0.0:lvmlock
+	 * means that the locks are located on the the LV "lvmlock".
+	 *
+	 * For dlm, the string is the dlm cluster name.  lvmlockd needs to use
+	 * a dlm lockspace in this cluster to use the VG.
+	 * e.g. lock_args = 1.0.0:foo
+	 * means that the host needs to be a member of the cluster "foo".
+	 *
+	 * The lvmlockd code for each specific lock manager also validates
+	 * the lock_args before using it to access the lock manager.
+	 */
 	if (dm_config_get_str(vgn, "lock_args", &str)) {
 		if (!(vg->lock_args = dm_pool_strdup(vg->vgmem, str)))
 			goto bad;
@@ -991,6 +1031,11 @@ static void _read_desc(struct dm_pool *mem,
 	*when = u;
 }
 
+/*
+ * It would be more accurate to call this _read_vgsummary().
+ * It is used to read vgsummary information about a VG
+ * before locking and reading the VG via vg_read().
+ */
 static int _read_vgname(const struct format_type *fmt, const struct dm_config_tree *cft, 
 			struct lvmcache_vgsummary *vgsummary)
 {
@@ -1027,6 +1072,8 @@ static int _read_vgname(const struct format_type *fmt, const struct dm_config_tr
 		return 0;
 	}
 
+	dm_config_get_str(vgn, "lock_type", &vgsummary->lock_type);
+
 	return 1;
 }
 

lib/freeseg/freeseg.c

@@ -33,7 +33,7 @@ struct segment_type *init_free_segtype(struct cmd_context *cmd)
 		return_NULL;
 
 	segtype->ops = &_freeseg_ops;
-	segtype->name = "free";
+	segtype->name = SEG_TYPE_NAME_FREE;
 	segtype->flags = SEG_VIRTUAL | SEG_CANNOT_BE_ZEROED;
 
 	log_very_verbose("Initialised segtype: %s", segtype->name);

lib/label/label.c

@@ -105,7 +105,8 @@ static void _update_lvmcache_orphan(struct lvmcache_info *info)
 
         memcpy(&vgsummary_orphan.vgid, lvmcache_fmt(info)->orphan_vg_name, strlen(lvmcache_fmt(info)->orphan_vg_name));
 
-	lvmcache_update_vgname_and_id(info, &vgsummary_orphan);
+	if (!lvmcache_update_vgname_and_id(info, &vgsummary_orphan))
+		stack;
 }
 
 static struct labeller *_find_labeller(struct device *dev, char *buf,

lib/locking/locking.c

@@ -119,8 +119,9 @@ int init_locking(int type, struct cmd_context *cmd, int suppress_messages)
 	switch (type) {
 	case 0:
 		init_no_locking(&_locking, cmd, suppress_messages);
-		log_warn("WARNING: Locking disabled. Be careful! "
-			  "This could corrupt your metadata.");
+		log_warn_suppress(suppress_messages,
+			"WARNING: Locking disabled. Be careful! "
+			"This could corrupt your metadata.");
 		return 1;
 
 	case 1:

lib/locking/lvmlockd.h

@@ -17,8 +17,7 @@
 #define LOCKD_SANLOCK_LV_NAME "lvmlock"
 
 /* lockd_gl flags */
-#define LDGL_SKIP_CACHE_VALIDATE  0x00000001
-#define LDGL_UPDATE_NAMES         0x00000002
+#define LDGL_UPDATE_NAMES         0x00000001
 
 /* lockd_lv flags */
 #define LDLV_MODE_NO_SH           0x00000001
@@ -27,10 +26,8 @@
 /* lvmlockd result flags */
 #define LD_RF_NO_LOCKSPACES     0x00000001
 #define LD_RF_NO_GL_LS          0x00000002
-#define LD_RF_LOCAL_LS          0x00000004
+#define LD_RF_WARN_GL_REMOVED   0x00000004
 #define LD_RF_DUP_GL_LS         0x00000008
-#define LD_RF_INACTIVE_LS       0x00000010
-#define LD_RF_ADD_LS_ERROR      0x00000020
 
 /* lockd_state flags */
 #define LDST_EX			0x00000001
@@ -54,8 +51,8 @@ void lvmlockd_disconnect(void);
 
 /* vgcreate/vgremove use init/free */
 
-int lockd_init_vg(struct cmd_context *cmd, struct volume_group *vg, const char *lock_type);
-int lockd_free_vg_before(struct cmd_context *cmd, struct volume_group *vg);
+int lockd_init_vg(struct cmd_context *cmd, struct volume_group *vg, const char *lock_type, int lv_lock_count);
+int lockd_free_vg_before(struct cmd_context *cmd, struct volume_group *vg, int changing);
 void lockd_free_vg_final(struct cmd_context *cmd, struct volume_group *vg);
 
 /* vgrename */
@@ -65,7 +62,7 @@ int lockd_rename_vg_final(struct cmd_context *cmd, struct volume_group *vg, int
 
 /* start and stop the lockspace for a vg */
 
-int lockd_start_vg(struct cmd_context *cmd, struct volume_group *vg);
+int lockd_start_vg(struct cmd_context *cmd, struct volume_group *vg, int start_init);
 int lockd_stop_vg(struct cmd_context *cmd, struct volume_group *vg);
 int lockd_start_wait(struct cmd_context *cmd);
 
@@ -92,7 +89,7 @@ int lockd_init_lv_args(struct cmd_context *cmd, struct volume_group *vg,
 int lockd_free_lv(struct cmd_context *cmd, struct volume_group *vg,
 		  const char *lv_name, struct id *lv_id, const char *lock_args);
 
-const char *lockd_running_lock_type(struct cmd_context *cmd);
+const char *lockd_running_lock_type(struct cmd_context *cmd, int *found_multiple);
 
 int handle_sanlock_lv(struct cmd_context *cmd, struct volume_group *vg);
 
@@ -125,12 +122,12 @@ static inline int lvmlockd_use(void)
 	return 0;
 }
 
-static inline int lockd_init_vg(struct cmd_context *cmd, struct volume_group *vg, const char *lock_type)
+static inline int lockd_init_vg(struct cmd_context *cmd, struct volume_group *vg, const char *lock_type, int lv_lock_count)
 {
 	return 1;
 }
 
-static inline int lockd_free_vg_before(struct cmd_context *cmd, struct volume_group *vg)
+static inline int lockd_free_vg_before(struct cmd_context *cmd, struct volume_group *vg, int changing)
 {
 	return 1;
 }
@@ -150,7 +147,7 @@ static inline int lockd_rename_vg_final(struct cmd_context *cmd, struct volume_g
 	return 1;
 }
 
-static inline int lockd_start_vg(struct cmd_context *cmd, struct volume_group *vg)
+static inline int lockd_start_vg(struct cmd_context *cmd, struct volume_group *vg, int start_init)
 {
 	return 0;
 }
@@ -167,6 +164,14 @@ static inline int lockd_start_wait(struct cmd_context *cmd)
 
 static inline int lockd_gl_create(struct cmd_context *cmd, const char *def_mode, const char *vg_lock_type)
 {
+	/*
+	 * When lvm is built without lvmlockd support, creating a VG with
+	 * a shared lock type should fail.
+	 */
+	if (is_lockd_type(vg_lock_type)) {
+		log_error("Using a shared lock type requires lvmlockd.");
+		return 0;
+	}
 	return 1;
 }
 
@@ -218,8 +223,9 @@ static inline int lockd_free_lv(struct cmd_context *cmd, struct volume_group *vg
 	return 1;
 }
 
-static inline const char *lockd_running_lock_type(struct cmd_context *cmd)
+static inline const char *lockd_running_lock_type(struct cmd_context *cmd, int *found_multiple)
 {
+	log_error("Using a shared lock type requires lvmlockd.");
 	return NULL;
 }
 

lib/locking/no_locking.c

@@ -19,8 +19,6 @@
 #include "lvm-string.h"
 #include "activate.h"
 
-#include <signal.h>
-
 /*
  * No locking
  */

lib/log/log.c

@@ -36,7 +36,7 @@ static int _indent = 1;
 static int _log_suppress = 0;
 static char _msg_prefix[30] = "  ";
 static int _already_logging = 0;
-static int _abort_on_internal_errors = 0;
+static int _abort_on_internal_errors_config = 0;
 
 static lvm2_log_fn_t _lvm2_log_fn = NULL;
 
@@ -49,10 +49,7 @@ static size_t _lvm_errmsg_len = 0;
 
 void init_log_fn(lvm2_log_fn_t log_fn)
 {
-	if (log_fn)
-		_lvm2_log_fn = log_fn;
-	else
-		_lvm2_log_fn = NULL;
+	_lvm2_log_fn = log_fn;
 }
 
 /*
@@ -69,7 +66,7 @@ void init_log_file(const char *log_file, int append)
 	static const char statfile[] = "/proc/self/stat";
 	const char *env;
 	int pid;
-	long long starttime;
+	unsigned long long starttime;
 	FILE *st;
 	int i = 0;
 
@@ -92,9 +89,6 @@ void init_log_file(const char *log_file, int append)
 			   "%llu", &pid, &starttime) != 2) {
 			log_warn("WARNING: Cannot parse content of %s.", statfile);
 		} else {
-			if (fclose(st))
-				log_sys_debug("fclose", statfile);
-
 			if (dm_snprintf(_log_file_path, sizeof(_log_file_path),
 					"%s_%s_%d_%lld", log_file, env, pid, starttime) < 0) {
 				log_warn("WARNING: Debug log file path is too long for epoch.");
@@ -104,7 +98,11 @@ void init_log_file(const char *log_file, int append)
 				append = 1; /* force */
 			}
 		}
+
+		if (st && fclose(st))
+			log_sys_debug("fclose", statfile);
 	}
+
 no_epoch:
 	if (!(_log_file = fopen(log_file, append ? "a" : "w"))) {
 		log_sys_error("fopen", log_file);
@@ -217,9 +215,10 @@ void init_indent(int indent)
 	_indent = indent;
 }
 
+/* If present, environment setting will override this. */
 void init_abort_on_internal_errors(int fatal)
 {
-	_abort_on_internal_errors = fatal;
+	_abort_on_internal_errors_config = fatal;
 }
 
 void reset_lvm_errno(int store_errmsg)
@@ -276,10 +275,24 @@ void print_log(int level, const char *file, int line, int dm_errno_or_class,
 	size_t msglen;
 	const char *indent_spaces = "";
 	FILE *stream;
+	static int _abort_on_internal_errors_env_present = -1;
+	static int _abort_on_internal_errors_env = 0;
+	char *env_str;
 
 	level &= ~(_LOG_STDERR|_LOG_ONCE);
 
-	if (_abort_on_internal_errors &&
+	if (_abort_on_internal_errors_env_present < 0) {
+		if ((env_str = getenv("DM_ABORT_ON_INTERNAL_ERRORS"))) {
+			_abort_on_internal_errors_env_present = 1;
+			/* Set when env DM_ABORT_ON_INTERNAL_ERRORS is not "0" */
+			_abort_on_internal_errors_env = strcmp(env_str, "0");
+		} else
+			_abort_on_internal_errors_env_present = 0;
+	}
+
+	/* Use value from environment if present, otherwise use value from config. */
+	if (((_abort_on_internal_errors_env_present && _abort_on_internal_errors_env) ||
+	     (!_abort_on_internal_errors_env_present && _abort_on_internal_errors_config)) &&
 	    !strncmp(format, INTERNAL_ERROR, sizeof(INTERNAL_ERROR) - 1)) {
 		fatal_internal_error = 1;
 		/* Internal errors triggering abort cannot be suppressed. */

lib/log/log.h

@@ -37,8 +37,6 @@
  *
  */
 
-#include <stdio.h>		/* FILE */
-#include <string.h>		/* strerror() */
 #include <errno.h>
 
 #define EUNCLASSIFIED -1	/* Generic error code */

lib/log/lvm-logging.h

@@ -16,9 +16,9 @@
 #ifndef _LVM_LOGGING_H
 #define _LVM_LOGGING_H
 
+__attribute__ ((format(printf, 5, 6)))
 void print_log(int level, const char *file, int line, int dm_errno_or_class,
-	       const char *format, ...)
-    __attribute__ ((format(printf, 5, 6)));
+	       const char *format, ...);
 
 #define LOG_LINE(l, x...) \
     print_log(l, __FILE__, __LINE__ , 0, ## x)

lib/lvmpolld/lvmpolld-client.c

@@ -106,6 +106,7 @@ static void _explain_error_codes(int retcode)
 	/* lvmpolld specific return codes */
 	case LVMPD_RET_DUP_FAILED:
 		log_error("lvmpolld failed to duplicate file descriptors.");
+		/* fall through */
 	case LVMPD_RET_EXC_FAILED:
 		log_error("lvmpolld failed to exec() lvm binary.");
 		break;
@@ -146,7 +147,7 @@ static struct progress_info _request_progress_info(const char *uuid, unsigned ab
 	}
 
 	if (abort_polling &&
-	    !daemon_request_extend(req, LVMPD_PARM_ABORT " = %d", abort_polling, NULL)) {
+	    !daemon_request_extend(req, LVMPD_PARM_ABORT " = %d", (int64_t)abort_polling, NULL)) {
 		log_error("Failed to create " LVMPD_REQ_PROGRESS " request.");
 		goto out_req;
 	}
@@ -227,14 +228,14 @@ static int _process_poll_init(const struct cmd_context *cmd, const char *poll_ty
 	}
 
 	if (parms->aborting &&
-	    !(daemon_request_extend(req, LVMPD_PARM_ABORT " = %d", parms->aborting, NULL))) {
+	    !(daemon_request_extend(req, LVMPD_PARM_ABORT " = %d", (int64_t)(parms->aborting), NULL))) {
 		log_error("Failed to create %s request." , poll_type);
 		goto out_req;
 	}
 
 	if (cmd->handles_missing_pvs &&
 	    !(daemon_request_extend(req, LVMPD_PARM_HANDLE_MISSING_PVS " = %d",
-				    cmd->handles_missing_pvs, NULL))) {
+				    (int64_t)(cmd->handles_missing_pvs), NULL))) {
 		log_error("Failed to create %s request." , poll_type);
 		goto out_req;
 	}

lib/metadata/cache_manip.c

@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2014 Red Hat, Inc. All rights reserved.
+ * Copyright (C) 2014-2015 Red Hat, Inc. All rights reserved.
  *
  * This file is part of LVM2.
  *
@@ -29,7 +29,17 @@
 #define DM_HINT_OVERHEAD_PER_BLOCK	8  /* bytes */
 #define DM_MAX_HINT_WIDTH		(4+16)  /* bytes.  FIXME Configurable? */
 
-const char *get_cache_pool_cachemode_name(const struct lv_segment *seg)
+int cache_mode_is_set(const struct lv_segment *seg)
+{
+	if (seg_is_cache(seg))
+		seg = first_seg(seg->pool_lv);
+
+	return (seg->feature_flags & (DM_CACHE_FEATURE_WRITEBACK |
+				      DM_CACHE_FEATURE_WRITETHROUGH |
+				      DM_CACHE_FEATURE_PASSTHROUGH)) ? 1 : 0;
+}
+
+const char *get_cache_mode_name(const struct lv_segment *seg)
 {
 	if (seg->feature_flags & DM_CACHE_FEATURE_WRITEBACK)
 		return "writeback";
@@ -46,22 +56,68 @@ const char *get_cache_pool_cachemode_name(const struct lv_segment *seg)
 	return NULL;
 }
 
-int set_cache_pool_feature(uint64_t *feature_flags, const char *str)
+int cache_set_mode(struct lv_segment *seg, const char *str)
 {
+	struct cmd_context *cmd = seg->lv->vg->cmd;
+	int id;
+	uint64_t mode;
+
+	if (!str && !seg_is_cache(seg))
+		return 1;			/* Defaults only for cache */
+
+	if (seg_is_cache(seg))
+		seg = first_seg(seg->pool_lv);
+
+	if (!str) {
+		if (cache_mode_is_set(seg))
+			return 1;               /* Default already set in cache pool */
+
+		id = allocation_cache_mode_CFG;
+
+		/* If present, check backward compatible settings */
+		if (!find_config_node(cmd, cmd->cft, id) &&
+		    find_config_node(cmd, cmd->cft, allocation_cache_pool_cachemode_CFG))
+			id = allocation_cache_pool_cachemode_CFG;
+
+		str = find_config_tree_str(cmd, id, NULL);
+	}
+
 	if (!strcmp(str, "writeback"))
-		*feature_flags |= DM_CACHE_FEATURE_WRITEBACK;
+		mode = DM_CACHE_FEATURE_WRITEBACK;
 	else if (!strcmp(str, "writethrough"))
-		*feature_flags |= DM_CACHE_FEATURE_WRITETHROUGH;
-	else if (!strcmp(str, "passhrough"))
-		*feature_flags |= DM_CACHE_FEATURE_PASSTHROUGH;
+		mode = DM_CACHE_FEATURE_WRITETHROUGH;
+	else if (!strcmp(str, "passthrough"))
+		mode = DM_CACHE_FEATURE_PASSTHROUGH;
 	else {
-		log_error("Cache pool feature \"%s\" is unknown.", str);
+		log_error("Cannot set unknown cache mode \"%s\".", str);
 		return 0;
 	}
 
+	seg->feature_flags &= ~(DM_CACHE_FEATURE_WRITEBACK |
+				DM_CACHE_FEATURE_WRITETHROUGH |
+				DM_CACHE_FEATURE_PASSTHROUGH);
+	seg->feature_flags |= mode;
+
 	return 1;
 }
 
+/*
+ * At least warn a user if certain cache stacks may present some problems
+ */
+void cache_check_for_warns(const struct lv_segment *seg)
+{
+	struct logical_volume *origin_lv = seg_lv(seg, 0);
+
+	if (lv_is_raid(origin_lv) &&
+	    first_seg(seg->pool_lv)->feature_flags & DM_CACHE_FEATURE_WRITEBACK)
+		log_warn("WARNING: Data redundancy is lost with writeback "
+			 "caching of raid logical volume!");
+
+	if (lv_is_thin_pool_data(seg->lv))
+		log_warn("WARNING: Cached thin pool's data cannot be currently "
+			 "resized and require manual uncache before resize!");
+}
+
 int update_cache_pool_params(const struct segment_type *segtype,
 			     struct volume_group *vg, unsigned attr,
 			     int passed_args, uint32_t pool_data_extents,
@@ -205,7 +261,7 @@ struct logical_volume *lv_cache_create(struct logical_volume *pool_lv,
 	if (lv_is_thin_pool(cache_lv))
 		cache_lv = seg_lv(first_seg(cache_lv), 0); /* cache _tdata */
 
-	if (!(segtype = get_segtype_from_string(cmd, "cache")))
+	if (!(segtype = get_segtype_from_string(cmd, SEG_TYPE_NAME_CACHE)))
 		return_NULL;
 
 	if (!insert_layer_for_lv(cmd, cache_lv, CACHE, "_corig"))
@@ -322,7 +378,7 @@ int lv_cache_remove(struct logical_volume *cache_lv)
 		dirty_blocks = status->cache->dirty_blocks;
 		dm_pool_destroy(status->mem);
 		if (dirty_blocks) {
-			log_print_unless_silent("%" PRIu64 " blocks must still be flushed.",
+			log_print_unless_silent(FMTu64 " blocks must still be flushed.",
 						dirty_blocks);
 			sleep(1);
 		}
@@ -351,7 +407,7 @@ int lv_cache_remove(struct logical_volume *cache_lv)
 
 	/* Replace 'error' with 'cache' segtype */
 	cache_seg = first_seg(corigin_lv);
-	if (!(cache_seg->segtype = get_segtype_from_string(corigin_lv->vg->cmd, "cache")))
+	if (!(cache_seg->segtype = get_segtype_from_string(corigin_lv->vg->cmd, SEG_TYPE_NAME_CACHE)))
 		return_0;
 
 	if (!(cache_seg->areas = dm_pool_zalloc(cache_lv->vg->vgmem, sizeof(*cache_seg->areas))))
@@ -395,36 +451,104 @@ int lv_is_cache_origin(const struct logical_volume *lv)
 	return seg && lv_is_cache(seg->lv) && !lv_is_pending_delete(seg->lv) && (seg_lv(seg, 0) == lv);
 }
 
-int lv_cache_setpolicy(struct logical_volume *lv, struct dm_config_tree *policy)
+static const char *_get_default_cache_policy(struct cmd_context *cmd)
 {
-	struct lv_segment *seg = first_seg(lv);
-	const char *name;
-	struct dm_config_node *cn;
-	struct dm_config_tree *old = NULL, *new = NULL, *tmp = NULL;
-	int r = 0;
+	const struct segment_type *segtype = get_segtype_from_string(cmd, SEG_TYPE_NAME_CACHE);
+	unsigned attr = ~0;
+        const char *def = NULL;
 
-	if (lv_is_cache(lv))
-		seg = first_seg(seg->pool_lv);
-
-	if (seg->policy_settings) {
-		if (!(old = dm_config_create()))
-			goto_out;
-		if (!(new = dm_config_create()))
-			goto_out;
-		new->root = policy->root;
-		old->root = seg->policy_settings;
-		new->cascade = old;
-		if (!(tmp = policy = dm_config_flatten(new)))
-			goto_out;
+	if (!segtype ||
+	    !segtype->ops->target_present ||
+	    !segtype->ops->target_present(cmd, NULL, &attr)) {
+		log_warn("WARNING: Cannot detect default cache policy, using \""
+			 DEFAULT_CACHE_POLICY "\".");
+		return DEFAULT_CACHE_POLICY;
 	}
 
-	if ((cn = dm_config_find_node(policy->root, "policy_settings")) &&
-	    !(seg->policy_settings = dm_config_clone_node_with_mem(lv->vg->vgmem, cn, 0)))
-		goto_out;
+	if (attr & CACHE_FEATURE_POLICY_SMQ)
+		def = "smq";
+	else if (attr & CACHE_FEATURE_POLICY_MQ)
+		def = "mq";
+	else {
+		log_error("Default cache policy is not available.");
+		return NULL;
+	}
 
-	if ((name = dm_config_find_str(policy->root, "policy", NULL)) &&
-	    !(seg->policy_name = dm_pool_strdup(lv->vg->vgmem, name)))
-		goto_out;
+	log_debug_metadata("Detected default cache_policy \"%s\".", def);
+
+	return def;
+}
+
+int cache_set_policy(struct lv_segment *seg, const char *name,
+		     const struct dm_config_tree *settings)
+{
+	struct dm_config_node *cn;
+	const struct dm_config_node *cns;
+	struct dm_config_tree *old = NULL, *new = NULL, *tmp = NULL;
+	int r = 0;
+	const int passed_seg_is_cache = seg_is_cache(seg);
+
+	if (passed_seg_is_cache)
+		seg = first_seg(seg->pool_lv);
+
+	if (name) {
+		if (!(seg->policy_name = dm_pool_strdup(seg->lv->vg->vgmem, name))) {
+			log_error("Failed to duplicate policy name.");
+			return 0;
+		}
+	} else if (!seg->policy_name && passed_seg_is_cache) {
+		if (!(seg->policy_name = find_config_tree_str(seg->lv->vg->cmd, allocation_cache_policy_CFG, NULL)) &&
+		    !(seg->policy_name = _get_default_cache_policy(seg->lv->vg->cmd)))
+			return_0;
+	}
+
+	if (settings) {
+		if (!seg->policy_name) {
+			log_error(INTERNAL_ERROR "Can't set policy settings without policy name.");
+			return 0;
+		}
+
+		if (seg->policy_settings) {
+			if (!(old = dm_config_create()))
+				goto_out;
+			if (!(new = dm_config_create()))
+				goto_out;
+			new->root = settings->root;
+			old->root = seg->policy_settings;
+			new->cascade = old;
+			if (!(tmp = dm_config_flatten(new)))
+				goto_out;
+		}
+
+		if ((cn = dm_config_find_node((tmp) ? tmp->root : settings->root, "policy_settings")) &&
+		    !(seg->policy_settings = dm_config_clone_node_with_mem(seg->lv->vg->vgmem, cn, 0)))
+			goto_out;
+	} else if (passed_seg_is_cache && /* Look for command's profile cache_policies */
+		   (cns = find_config_tree_node(seg->lv->vg->cmd, allocation_cache_settings_CFG_SECTION, NULL))) {
+		/* Try to find our section for given policy */
+		for (cn = cns->child; cn; cn = cn->sib) {
+			/* Only matching section names */
+			if (cn->v || strcmp(cn->key, seg->policy_name) != 0)
+				continue;
+
+			if (!cn->child)
+				break;
+
+			if (!(new = dm_config_create()))
+				goto_out;
+
+			if (!(new->root = dm_config_clone_node_with_mem(new->mem,
+									cn->child, 1)))
+				goto_out;
+
+			if (!(seg->policy_settings = dm_config_create_node(new, "policy_settings")))
+				goto_out;
+
+			seg->policy_settings->child = new->root;
+
+			break; /* Only first match counts */
+		}
+	}
 
 restart: /* remove any 'default" nodes */
 	cn = seg->policy_settings ? seg->policy_settings->child : NULL;
@@ -439,12 +563,13 @@ restart: /* remove any 'default" nodes */
 	r = 1;
 
 out:
-	if (old)
-		dm_config_destroy(old);
-	if (new)
-		dm_config_destroy(new);
 	if (tmp)
 		dm_config_destroy(tmp);
+	if (new)
+		dm_config_destroy(new);
+	if (old)
+		dm_config_destroy(old);
+
 	return r;
 }