2.00.25

agk - you may want to check this.

Makefile.in

@@ -22,24 +22,31 @@ ifeq ("@INTL@", "yes")
   SUBDIRS += po
 endif
 
-SUBDIRS += lib tools
+SUBDIRS += lib tools daemons
 
 ifeq ($(MAKECMDGOALS),distclean)
-  SUBDIRS += lib/format1 \
+  SUBDIRS += daemons/clvmd \
+	     lib/format1 \
+	     lib/format_pool \
+	     lib/locking \
+	     lib/mirror \
+	     lib/snapshot \
 	     po \
 	     test/mm test/device test/format1 test/regex test/filters
 endif
 
 include make.tmpl
 
+daemons: lib
 lib: include
 tools: lib
-po: lib tools
+po: tools daemons
 
 ifeq ("@INTL@", "yes")
 lib.pofile: include.pofile
 tools.pofile: lib.pofile
-po.pofile: lib.pofile tools.pofile
+daemons.pofile: lib.pofile
+po.pofile: tools.pofile daemons.pofile
 pofile: po.pofile
 endif
 

README

@@ -9,8 +9,8 @@ Installation instructions are in INSTALL.
 There is no warranty - see COPYING and COPYING.LIB.
 
 Tarballs are available from:
-  ftp://ftp.sistina.com/pub/LVM2/tools/
-  ftp://ftp.sistina.com/pub/LVM2/device-mapper/
+  ftp://sources.redhat.com/pub/lvm2/
+  ftp://sources.redhat.com/pub/dm/
 
 To access the CVS tree use:
   cvs -d :pserver:cvs@sources.redhat.com:/cvs/lvm2 login

VERSION

@@ -1 +1 @@
-2.00.15-cvs (2004-04-19)
+2.00.25-cvs (2004-09-29)

WHATS_NEW

@@ -1,3 +1,128 @@
+Version 2.00.25 - 29th September 2004
+=====================================
+  Fix return code from rm_link for vgmknodes.
+  Make clvmd LV hash table thread-safe.
+  Fix clvmd locking so it will lock out multiple users on the same node.
+  Fix clvmd VG locking to it can cope with multiple VG locks.
+  Remove spurious trailing dot in lvreduce man page.
+  Fix vgremove locking.
+
+Version 2.00.24 - 16th September 2004
+=====================================
+  Fix pool_empty so it really does empty the memory pool.
+  Rename old segtypes files to segtype.
+  Some fixes to memory debugging code.
+  Exclude internal commands formats & segtypes from install.
+
+Version 2.00.23 - 15th September 2004
+=====================================
+  Export dm name build & split functions.
+  Use O_NOATIME on devices if available.
+  Write log message when each segtype/format gets initialised.
+  New commands 'segtypes' and 'formats'.
+  Suppress pvmove abort message in test mode.
+  Improve pvcreate/remove device not found error message.
+  Allow pvmove to move data within the same PV.
+  Describe how pvmove works on man page.
+  Test for incompatible format/segtype combinations in lv_extend.
+  Fix lvchange example on man page.
+
+Version 2.00.22 - 3rd September 2004
+====================================
+  Fix /dev/vgname perms.
+  Restructure xlate.h.
+  Add clvmd man page.
+
+Version 2.00.21 - 19th August 2004
+==================================
+  Update cnxman-socket.h from cman.
+  Recognise iseries/vd devices.
+  Use 'make install_cluster' to install cluster extensions only.
+  Cope with DT_UNKNOWN in sysfs.
+  Fix extents_moved metadata size comment.
+  Remove duplicate line in pvremove help text.
+  Support variable mirror region size.
+  Support PE ranges in pvmove source PV.
+  Fixes to as-yet-unused LV segment splitting code.
+  Change alloc_areas to pe_ranges and allow suppression of availability checks.
+  Add dev_size column to pvs.
+  Add report columns for in-kernel device number.
+
+Version 2.00.20 - 3 July 2004
+=============================
+  More autoconf fixes.
+  Fix device number handling for 2.6 kernels.
+
+Version 2.00.19 - 29 June 2004
+==============================
+  Reduce severity of setlocale failure message.
+  Recognise argv[0] "initrd-lvm" (pld-linux).
+  Make -O2 configurable.
+  Added --disable-selinux to configure script.
+  LD_FLAGS->LDFLAGS & LD_DEPS->LDDEPS in configure script.
+  Add init_debug to clvmd.
+
+Version 2.00.18 - 24 June 2004
+==============================
+  Fix vgchange activation.
+  Add cluster support.
+
+Version 2.00.17 - 20 June 2004
+==============================
+  configure --enable-fsadm to try out fsadm.  fsadm is not tested yet.
+  Display all filtered devices, not just PVs, with pvs -a.
+  Fix sync_dir() when no / in filename
+  vgcfgbackup -f accepts template with %s for VG name.
+  Extend hash functions to handle non-null-terminated data.
+  Add local activation support.
+  Tidy relative paths in makefile includes.
+  fsadm support for fsck and resizing - needs testing.
+  Add read-only GFS pool support.
+  Add lvm2create_initrd script from http://poochiereds.net/svn/lvm2/
+  Fix rounding of large diplayed sizes.
+  Suppress decimal point when using units of sectors/bytes.
+  Additional kernel target checks before pvmove & snapshot creation.
+  Add i2o_block.
+
+Version 2.00.16 - 24 May 2004
+=============================
+  Set area_count within alloc_lv_segment.
+  Remove error labels from lvresize.
+  Fix a pvs error path.
+  xxchange -ae for exclusive activation.
+  Don't return non-zero status if there aren't any volume groups.
+  Add --alloc argument to tools.
+  Rename allocation policies to contiguous, normal, anywhere, inherit.
+  nextfree becomes normal; anywhere isn't implemented yet.
+  LV inherits allocation policy from VG. Defaults: LV - inherit; VG - normal
+  Additional status character added to vgs to indicate allocation policy.
+  Add reset_fn to external_locking.
+  Ensure presence of virtual targets before attempting activating.
+  Attempt to fix resizing of snapshot origins.
+  Restructure lvresize, bringing it closer to lvcreate.
+  A quick sanity check on vg_disk struct when read in.  More checks needed.
+  Only include visible LVs in active/open counts.
+  Add virtual segment types, zero and error.  A large sparse device can be
+constructed as a writeable snapshot of a large zero segment.
+  Add --type to lvcreate/resize.
+  Push lv_create & alloc policy up to tool level.
+  Fix pvdisplay return code.
+  Detect invalid LV names in arg lists.
+  Reporting uses line-at-a-time output.
+  lvm2 format sets unlimited_vols format flag.
+  Internal-only metadata flag support.
+  Basic checking for presence of device-mapper targets.
+  Separate out polldaemon.
+  Revise internal locking semantics.
+  Move find_pv_by_name to library.
+  Rename move->copy.
+  Add devices to segments report.
+  Begin separating out segment code. There's a lot of change here.
+  Compress any (obsolete) long LVM1 pvids encountered.
+  Support for tagged config files.
+  Don't abort operations if selinux present but disabled.
+  Fix typo in configure which left HAVE_LIBDL unset.
+
 Version 2.00.15 - 19 Apr 2004
 =============================
   configure --with-owner= --with-group= to avoid -o and -g args to 'install'

WHATS_NEW_DM

@@ -1,3 +1,19 @@
+Version 1.00.20 -
+=============================
+
+Version 1.00.19 - 3 July 2004
+=============================
+  More autoconf fixes.
+  Fix a dmsetup newline.
+  Fix device number handling for 2.6 kernels.
+
+Version 1.00.18 - 20 Jun 2004
+=============================
+  Fix a uuid free in libdm-iface.
+  Fix a targets string size calc in driver.
+  Add -c to dmsetup for column-based output.
+  Add target message-passing ioctl.
+
 Version 1.00.17 - 17 Apr 2004
 =============================
   configure --with-owner= --with-group= to avoid -o and -g args to 'install'

configure.in

@@ -1,29 +1,64 @@
 ##
-##    Copyright 1999-2000 Sistina Software, Inc.
+## Copyright (C) 2000-2004 Sistina Software, Inc. All rights reserved.
+## Copyright (C) 2004 Red Hat, Inc. All rights reserved.
 ##
-##    This is free software released under the GNU General Public License.
-##    There is no warranty for this software.  See the file COPYING for
-##    details.
+## This file is part of the LVM2.
 ##
-##    See the file CONTRIBUTORS for a list of contributors.
+## 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.
 ##
-##    This file is maintained by:
-##      AJ Lewis <lewis@sistina.com>
-## 
-##    File name: configure.in
-##
-##    Description: Input file for autoconf.  Generates the configure script 
-##                 that tries to keep everything nice and portable.  It also
-##                 simplifies distribution package building considerably.
+## 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
 ################################################################################
 
-dnl Process this file with autoconf to produce a configure script.
+AC_PREREQ(2.53)
+################################################################################
+dnl -- Process this file with autoconf to produce a configure script.
 AC_INIT(lib/device/dev-cache.h)
 
-dnl setup the directory where autoconf has auxilary files
+################################################################################
+dnl -- Setup the directory where autoconf has auxilary files
 AC_CONFIG_AUX_DIR(autoconf) 
 
-dnl Checks for programs.
+################################################################################
+dnl -- Get system type
+AC_CANONICAL_SYSTEM
+
+case "$host_os" in
+	linux*)
+		CFLAGS="$CFLAGS"
+		COPTIMISE_FLAG="-O2"
+		CLDFLAGS="$CLDFLAGS -Wl,--version-script,.export.sym"
+		CLDWHOLEARCHIVE="-Wl,-whole-archive"
+		CLDNOWHOLEARCHIVE="-Wl,-no-whole-archive"
+		LDDEPS="$LDDEPS .export.sym"
+		LDFLAGS="$LDFLAGS -Wl,--export-dynamic"
+		SOFLAG="-shared"
+		DEVMAPPER=yes
+		ODIRECT=yes
+		SELINUX=yes
+		CLUSTER=internal
+		FSADM=no ;;
+	darwin*)
+		CFLAGS="$CFLAGS -no-cpp-precomp -fno-common"
+		COPTIMISE_FLAG="-O2"
+		CLDFLAGS="$CLDFLAGS"
+		CLDWHOLEARCHIVE="-all_load"
+		CLDNOWHOLEARCHIVE=
+		LDDEPS="$LDDEPS"
+		LDFLAGS="$LDFLAGS"
+		SOFLAG="-dynamiclib"
+		DEVMAPPER=no
+		ODIRECT=no
+		SELINUX=no
+		CLUSTER=none
+		FSADM=no ;;
+esac
+
+################################################################################
+dnl -- Checks for programs.
 AC_PROG_AWK
 AC_PROG_CC
 AC_PROG_INSTALL
@@ -31,158 +66,304 @@ AC_PROG_LN_S
 AC_PROG_MAKE_SET
 AC_PROG_RANLIB
 
-dnl Checks for header files.
+################################################################################
+dnl -- Checks for header files.
 AC_HEADER_DIRENT
 AC_HEADER_STDC
-AC_CHECK_HEADERS(fcntl.h malloc.h sys/ioctl.h unistd.h)
+AC_HEADER_SYS_WAIT
+AC_HEADER_TIME
 
-dnl Checks for typedefs, structures, and compiler characteristics.
+AC_CHECK_HEADERS(fcntl.h limits.h locale.h stddef.h syslog.h sys/file.h sys/ioctl.h sys/param.h sys/time.h,,AC_MSG_ERROR(bailing out))
+AC_CHECK_HEADERS(assert.h ctype.h libgen.h signal.h stdio.h sys/mman.h sys/resource.h sys/stat.h sys/types.h sys/utsname.h sys/wait.h time.h,,AC_MSG_ERROR(bailing out))
+
+case "$host_os" in
+	linux*)
+		AC_CHECK_HEADERS(asm/byteorder.h linux/fs.h malloc.h,,AC_MSG_ERROR(bailing out)) ;;
+	darwin*)
+		AC_CHECK_HEADERS(machine/endian.h sys/disk.h,,AC_MSG_ERROR(bailing out)) ;;
+esac
+
+################################################################################
+dnl -- Checks for typedefs, structures, and compiler characteristics.
 AC_C_CONST
 AC_C_INLINE
 AC_TYPE_OFF_T
 AC_TYPE_PID_T
 AC_TYPE_SIZE_T
+AC_TYPE_MODE_T
 AC_STRUCT_ST_RDEV
-AC_HEADER_TIME
+AC_STRUCT_TM
 
-dnl Get system type
-AC_CANONICAL_SYSTEM
+################################################################################
+dnl -- Check for functions
+AC_CHECK_FUNCS(gethostname getpagesize memset munmap setlocale strcasecmp strchr strdup strncasecmp strerror strrchr strstr strtol strtoul,,AC_MSG_ERROR(bailing out))
+AC_FUNC_ALLOCA
+AC_FUNC_CLOSEDIR_VOID
+AC_FUNC_FORK
+AC_FUNC_LSTAT
+AC_FUNC_MALLOC
+AC_FUNC_MEMCMP
+AC_FUNC_MMAP
+AC_FUNC_STAT
+AC_FUNC_STRTOD
 
-case "$host_os" in
-	linux*)
-		CFLAGS= 
-		CLDFLAGS="-Wl,--version-script,.export.sym"
-		CLDWHOLEARCHIVE="-Wl,-whole-archive"
-		CLDNOWHOLEARCHIVE="-Wl,-no-whole-archive"
-		LD_DEPS=".export.sym"
-		LD_FLAGS="-Wl,--export-dynamic"
-		SOFLAG="-shared"
-		DEVMAPPER=yes
-		ODIRECT=yes ;;
-	darwin*)
-		CFLAGS="-no-cpp-precomp -fno-common"
-		CLDFLAGS=
-		CLDWHOLEARCHIVE="-all_load"
-		CLDNOWHOLEARCHIVE=
-		LD_DEPS=
-		LD_FLAGS=
-		SOFLAG="-dynamiclib"
-		DEVMAPPER=no
-		ODIRECT=no ;;
-esac
-
-dnl -- prefix is /usr by default, the exec_prefix default is setup later
+################################################################################
+dnl -- Prefix is /usr by default, the exec_prefix default is setup later
 AC_PREFIX_DEFAULT(/usr)
 
-OWNER="root"
-GROUP="root"
+################################################################################
+dnl -- Parallel make jobs?
+AC_ARG_ENABLE(jobs, [  --enable-jobs=NUM       Number of jobs to run simultaneously], JOBS=-j$enableval, JOBS=-j2)
+
+################################################################################
+dnl -- Setup the ownership of the files
+AC_MSG_CHECKING(file owner)
+OWNER="root"
 
-dnl -- setup the ownership of the files
 AC_ARG_WITH(user,
   [  --with-user=USER        Set the owner of installed files ],
   [ OWNER="$withval" ])
+AC_MSG_RESULT($OWNER)
 
 if test x$OWNER != x; then
 	OWNER="-o $OWNER"
 fi
 
-dnl -- setup the group ownership of the files
+################################################################################
+dnl -- Setup the group ownership of the files
+AC_MSG_CHECKING(group owner)
+GROUP="root"
 AC_ARG_WITH(group,
   [  --with-group=GROUP      Set the group owner of installed files ],
   [ GROUP="$withval" ])
+AC_MSG_RESULT($GROUP)
 
 if test x$GROUP != x; then
 	GROUP="-g $GROUP"
 fi
 
+################################################################################
 dnl -- LVM1 tool fallback option
+AC_MSG_CHECKING(whether to enable lvm1 fallback)
 AC_ARG_ENABLE(lvm1_fallback, [  --enable-lvm1_fallback  Use this to fall back and use LVM1 binaries if
                           device-mapper is missing from the kernel],  LVM1_FALLBACK=$enableval, LVM1_FALLBACK=no)
+AC_MSG_RESULT($LVM1_FALLBACK)
 
 if test x$LVM1_FALLBACK = xyes; then
 	CFLAGS="$CFLAGS -DLVM1_FALLBACK"
 fi
 
+################################################################################
 dnl -- format1 inclusion type
+AC_MSG_CHECKING(whether to include support for lvm1 metadata)
 AC_ARG_WITH(lvm1,
   [  --with-lvm1=TYPE        LVM1 metadata support: internal/shared/none
                           [TYPE=internal] ],
   [ LVM1="$withval" ],
   [ LVM1="internal" ])
+AC_MSG_RESULT($LVM1)
 
 if [[ "x$LVM1" != xnone -a "x$LVM1" != xinternal -a "x$LVM1" != xshared ]];
  then  AC_MSG_ERROR(
 --with-lvm1 parameter invalid
 )
- exit
 fi;
 
 if test x$LVM1 = xinternal; then
 	CFLAGS="$CFLAGS -DLVM1_INTERNAL"
 fi
 
-AC_ARG_ENABLE(jobs, [  --enable-jobs=NUM       Number of jobs to run simultaneously], JOBS=-j$enableval, JOBS=-j2)
+################################################################################
+dnl -- format_pool inclusion type
+AC_MSG_CHECKING(whether to include support for GFS pool metadata)
+AC_ARG_WITH(pool,
+  [  --with-pool=TYPE        GFS pool read-only support: internal/shared/none
+                          [TYPE=internal] ],
+  [ POOL="$withval" ],
+  [ POOL="internal" ])
+AC_MSG_RESULT($POOL)
 
-dnl Enables staticly-linked tools
+if [[ "x$POOL" != xnone -a "x$POOL" != xinternal -a "x$POOL" != xshared ]];
+ then  AC_MSG_ERROR(
+--with-pool parameter invalid
+)
+fi;
+
+if test x$POOL = xinternal; then
+	CFLAGS="$CFLAGS -DPOOL_INTERNAL"
+fi
+
+################################################################################
+dnl -- cluster_locking inclusion type
+AC_MSG_CHECKING(whether to include support for cluster locking)
+AC_ARG_WITH(cluster,
+  [  --with-cluster=TYPE     Cluster LVM locking support: internal/shared/none
+                          [TYPE=internal] ],
+  [ CLUSTER="$withval" ])
+AC_MSG_RESULT($CLUSTER)
+
+if [[ "x$CLUSTER" != xnone -a "x$CLUSTER" != xinternal -a "x$CLUSTER" != xshared ]];
+ then  AC_MSG_ERROR(
+--with-cluster parameter invalid
+)
+fi;
+
+if test x$CLUSTER = xinternal; then
+	CFLAGS="$CFLAGS -DCLUSTER_LOCKING_INTERNAL"
+fi
+
+################################################################################
+dnl -- snapshots inclusion type
+AC_MSG_CHECKING(whether to include snapshots)
+AC_ARG_WITH(snapshots,
+  [  --with-snapshots=TYPE   Snapshot support: internal/shared/none
+                          [TYPE=internal] ],
+  [ SNAPSHOTS="$withval" ],
+  [ SNAPSHOTS="internal" ])
+AC_MSG_RESULT($SNAPSHOTS)
+
+if [[ "x$SNAPSHOTS" != xnone -a "x$SNAPSHOTS" != xinternal -a "x$SNAPSHOTS" != xshared ]];
+ then  AC_MSG_ERROR(
+--with-snapshots parameter invalid
+)
+fi;
+
+if test x$SNAPSHOTS = xinternal; then
+	CFLAGS="$CFLAGS -DSNAPSHOT_INTERNAL"
+fi
+
+################################################################################
+dnl -- mirrors inclusion type
+AC_MSG_CHECKING(whether to include mirrors)
+AC_ARG_WITH(mirrors,
+  [  --with-mirrors=TYPE     Mirror support: internal/shared/none
+                          [TYPE=internal] ],
+  [ MIRRORS="$withval" ],
+  [ MIRRORS="internal" ])
+AC_MSG_RESULT($MIRRORS)
+
+if [[ "x$MIRRORS" != xnone -a "x$MIRRORS" != xinternal -a "x$MIRRORS" != xshared ]];
+ then  AC_MSG_ERROR(
+--with-mirrors parameter invalid
+)
+fi;
+
+if test x$MIRRORS = xinternal; then
+	CFLAGS="$CFLAGS -DMIRRORED_INTERNAL"
+fi
+
+################################################################################
+dnl -- Enables staticly-linked tools
+AC_MSG_CHECKING(whether to use static linking)
 AC_ARG_ENABLE(static_link, [  --enable-static_link    Use this to link the tools to their libraries
                           statically.  Default is dynamic linking],  STATIC_LINK=$enableval, STATIC_LINK=no)
+AC_MSG_RESULT($STATIC_LINK)
 
-dnl Enable readline
-AC_ARG_ENABLE(readline, [  --enable-readline       Enable readline support],  \
+################################################################################
+dnl -- Enable readline
+AC_MSG_CHECKING(whether to enable readline)
+AC_ARG_ENABLE(readline, [  --enable-readline       Enable readline support],
 READLINE=$enableval, READLINE=no)
+AC_MSG_RESULT($READLINE)
 
 if test x$READLINE = xyes; then
 	CFLAGS="$CFLAGS -DREADLINE_SUPPORT"
 fi
 
-echo $ac_n "checking whether to enable debugging""... $ac_c" 1>&6
-dnl Enable Debugging
-AC_ARG_ENABLE(debug,    [  --enable-debug          Enable debugging],  \
-DEBUG=yes, DEBUG=no)
-echo "$ac_t""$DEBUG" 1>&6
+################################################################################
+dnl -- Disable selinux
+AC_MSG_CHECKING(whether to enable selinux support)
+AC_ARG_ENABLE(selinux, [  --disable-selinux       Disable selinux support],
+SELINUX=$enableval)
+AC_MSG_RESULT($SELINUX)
 
-echo $ac_n "checking whether to enable device-mapper""... $ac_c" 1>&6
-dnl Disable devmapper
-AC_ARG_ENABLE(devmapper, [  --disable-devmapper     Disable device-mapper interaction],  \
-DEVMAPPER=no)
-echo "$ac_t""$DEVMAPPER" 1>&6
+################################################################################
+dnl -- Build cluster LVM daemon
+AC_MSG_CHECKING(whether to build cluster LVM daemon)
+AC_ARG_WITH(clvmd, [  --with-clvmd            Build cluster LVM Daemon],
+CLVMD=$withval, CLVMD=no)
+AC_MSG_RESULT($CLVMD)
+
+dnl -- If clvmd enabled without cluster locking, automagically include it
+if  test x$CLVMD = xyes && test x$CLUSTER = xnone; then
+	CLUSTER=internal
+fi
+
+################################################################################
+dnl -- Enable debugging
+AC_MSG_CHECKING(whether to enable debugging)
+AC_ARG_ENABLE(debug,    [  --enable-debug          Enable debugging],
+DEBUG=$enableval, DEBUG=no)
+AC_MSG_RESULT($DEBUG)
+
+dnl -- Normally turn off optimisation for debug builds
+if test x$DEBUG = xyes; then
+	COPTIMISE_FLAG=
+fi
+
+################################################################################
+dnl -- Override optimisation
+AC_MSG_CHECKING(for C optimisation flag)
+AC_ARG_WITH(optimisation,
+  [  --with-optimisation=OPT C optimisation flag [OPT=-O2] ],
+  [ COPTIMISE_FLAG="$withval" ])
+AC_MSG_RESULT($COPTIMISE_FLAG)
+
+################################################################################
+dnl -- Disable devmapper
+AC_MSG_CHECKING(whether to use device-mapper)
+AC_ARG_ENABLE(devmapper, [  --disable-devmapper     Disable device-mapper interaction],
+DEVMAPPER=$enableval)
+AC_MSG_RESULT($DEVMAPPER)
 
 if test x$DEVMAPPER = xyes; then
 	CFLAGS="$CFLAGS -DDEVMAPPER_SUPPORT"
 fi
 
-echo $ac_n "checking whether to enable O_DIRECT""... $ac_c" 1>&6
-dnl Disable O_DIRECT
-AC_ARG_ENABLE(o_direct, [  --disable-o_direct      Disable O_DIRECT],  \
-ODIRECT=no)
-echo "$ac_t""$ODIRECT" 1>&6
+################################################################################
+dnl -- Disable O_DIRECT
+AC_MSG_CHECKING(whether to enable O_DIRECT)
+AC_ARG_ENABLE(o_direct, [  --disable-o_direct      Disable O_DIRECT],
+ODIRECT=$enableval)
+AC_MSG_RESULT($ODIRECT)
 
 if test x$ODIRECT = xyes; then
 	CFLAGS="$CFLAGS -DO_DIRECT_SUPPORT"
 fi
 
-echo $ac_n "checking whether to compile liblvm2cmd.so""... $ac_c" 1>&6
-dnl Enable cmdlib
-AC_ARG_ENABLE(cmdlib, [  --enable-cmdlib         Build shared command library],  \
-CMDLIB=yes, CMDLIB=no)
-echo "$ac_t""$CMDLIB" 1>&6
+################################################################################
+dnl -- Enable cmdlib
+AC_MSG_CHECKING(whether to compile liblvm2cmd.so)
+AC_ARG_ENABLE(cmdlib, [  --enable-cmdlib         Build shared command library],
+CMDLIB=$enableval, CMDLIB=no)
+AC_MSG_RESULT($CMDLIB)
 
 if test x$CMDLIB = xyes; then
 	CFLAGS="$CFLAGS -DCMDLIB"
 fi
 
-dnl Mess with default exec_prefix
+################################################################################
+dnl -- Enable fsadm
+AC_MSG_CHECKING(whether to build fsadm)
+AC_ARG_ENABLE(fsadm, [  --enable-fsadm          Enable fsadm],
+FSADM=$enableval)
+AC_MSG_RESULT($FSADM)
+
+################################################################################
+dnl -- Mess with default exec_prefix
 if [[ "x$exec_prefix" = xNONE -a "x$prefix" = xNONE ]];
  then  exec_prefix="";
 fi;
 
-dnl Checks for library functions.
+################################################################################
+dnl -- Checks for library functions.
 AC_PROG_GCC_TRADITIONAL
 AC_TYPE_SIGNAL
 AC_FUNC_VPRINTF
-AC_CHECK_FUNCS(mkdir rmdir uname)
+AC_CHECK_FUNCS(mkdir rmdir uname,,AC_MSG_ERROR(bailing out))
 
-dnl check for termcap (Shamelessly copied from parted 1.4.17)
+################################################################################
+dnl -- Check for termcap (Shamelessly copied from parted 1.4.17)
 if test x$READLINE = xyes; then
 	AC_SEARCH_LIBS(tgetent, ncurses curses termcap termlib, ,
 		AC_MSG_ERROR(
@@ -195,40 +376,51 @@ Note: if you are using precompiled packages you will also need the development
 Note: (n)curses also seems to work as a substitute for termcap.  This was
   not found either - but you could try installing that as well.
 )
-	exit
 	)
 fi
 
-dnl Check for dlopen
+################################################################################
+dnl -- Check for dlopen
 AC_CHECK_LIB(dl, dlopen, HAVE_LIBDL=yes, HAVE_LIBDL=no)
 
-if [[ "x$HAVE_LIBDL" = xyes -a "xSTATIC_LINK" = xno ]]; then
+if [[ "x$HAVE_LIBDL" = xyes -a "x$STATIC_LINK" = xno ]]; then
 	CFLAGS="$CFLAGS -DHAVE_LIBDL"
 	LIBS="-ldl $LIBS"
 else
 	HAVE_LIBDL=no
 fi
 
-dnl Check for shared/static conflicts
-if [[ "x$LVM1" = xshared -a "x$STATIC_LINK" = xyes ]];
+################################################################################
+dnl -- Check for shared/static conflicts
+if [[ \( "x$LVM1" = xshared -o "x$POOL" = xshared -o "x$CLUSTER" = xshared \
+      -o "x$SNAPSHOTS" = xshared -o "x$MIRRORS" = xshared \
+      \) -a "x$STATIC_LINK" = xyes ]];
  then  AC_MSG_ERROR(
 Features cannot be 'shared' when building statically
 )
- exit
 fi
 
-dnl Check for is_selinux_enabled
-AC_CHECK_LIB(selinux, is_selinux_enabled, HAVE_SELINUX=yes, HAVE_SELINUX=no)
+################################################################################
+dnl -- Check for is_selinux_enabled
+if test x$SELINUX = xyes; then
+	AC_MSG_CHECKING(for is_selinux_enabled function)
+	AC_CHECK_LIB(selinux, is_selinux_enabled, HAVE_SELINUX=yes, HAVE_SELINUX=no)
+	AC_MSG_RESULT($HAVE_SELINUX)
 
-if test x$HAVE_SELINUX = xyes; then
-	CFLAGS="$CFLAGS -DHAVE_SELINUX"
-	LIBS="-lselinux $LIBS"
+	if test x$HAVE_SELINUX = xyes; then
+		CFLAGS="$CFLAGS -DHAVE_SELINUX"
+		LIBS="-lselinux $LIBS"
+	else
+		AC_MSG_WARN(Disabling selinux)
+	fi
 fi
 
-dnl Check for getopt
+################################################################################
+dnl -- Check for getopt
 AC_CHECK_HEADERS(getopt.h, CFLAGS="$CFLAGS -DHAVE_GETOPTLONG")
 
-dnl Check for readline (Shamelessly copied from parted 1.4.17)
+################################################################################
+dnl -- Check for readline (Shamelessly copied from parted 1.4.17)
 if test x$READLINE = xyes; then
 	AC_CHECK_LIB(readline, readline, ,
 		AC_MSG_ERROR(
@@ -239,17 +431,17 @@ support with --disable-readline or download and install readline from:
 Note: if you are using precompiled packages you will also need the development
 package as well (which may be called readline-devel or something similar).
 )
-		exit
 	)
 	AC_CHECK_FUNC(rl_completion_matches, CFLAGS="$CFLAGS -DHAVE_RL_COMPLETION_MATCHES")
 		
 fi
 
-echo $ac_n "checking whether to enable internationalisation""... $ac_c" 1>&6
-dnl Internationalisation stuff
-AC_ARG_ENABLE(nls, [  --enable-nls            Enable Native Language Support],\
-		INTL=yes, INTL=no)
-echo "$ac_t""$INTL" 1>&6
+################################################################################
+dnl -- Internationalisation stuff
+AC_MSG_CHECKING(whether to enable internationalisation)
+AC_ARG_ENABLE(nls, [  --enable-nls            Enable Native Language Support],
+		INTL=$enableval, INTL=no)
+AC_MSG_RESULT($INTL)
 
 if test x$INTL = xyes; then
 	INTL_PACKAGE="lvm2"
@@ -258,7 +450,6 @@ if test x$INTL = xyes; then
 		then  AC_MSG_ERROR(
 		msgfmt not found in path $PATH
 		)
-		exit  
 	fi;
 
 	AC_ARG_WITH(localedir,
@@ -267,6 +458,7 @@ if test x$INTL = xyes; then
   		    [ LOCALEDIR='${prefix}/share/locale' ])
 fi
 
+################################################################################
 AC_ARG_WITH(confdir,
 	    [  --with-confdir=DIR      Configuration files in DIR [/etc]],
   	    [ CONFDIR="$withval" ],
@@ -277,24 +469,69 @@ AC_ARG_WITH(staticdir,
   	    [ STATICDIR="$withval" ],
  	    [ STATICDIR='${exec_prefix}/sbin' ])
 
+################################################################################
+dnl -- Ensure additional headers required
+if test x$READLINE = xyes; then
+	AC_CHECK_HEADERS(readline/readline.h readline/history.h,,AC_MSG_ERROR(bailing out))
+fi
 
+if test x$CLVMD = xyes; then
+	AC_CHECK_HEADERS(mntent.h netdb.h netinet/in.h pthread.h search.h sys/mount.h sys/socket.h sys/uio.h sys/un.h utmpx.h,,AC_MSG_ERROR(bailing out))
+	AC_CHECK_FUNCS(dup2 getmntent memmove select socket,,AC_MSG_ERROR(bailing out))
+	AC_FUNC_GETMNTENT
+#	AC_FUNC_REALLOC
+	AC_FUNC_SELECT_ARGTYPES
+fi
+
+if test x$FSADM = xyes; then
+	AC_CHECK_HEADERS(fstab.h sys/mount.h sys/vfs.h,,AC_MSG_ERROR(bailing out))
+	AC_CHECK_FUNCS(memmove,,AC_MSG_ERROR(bailing out))
+fi
+
+if test x$CLUSTER != xnone; then
+	AC_CHECK_HEADERS(sys/socket.h sys/un.h,,AC_MSG_ERROR(bailing out))
+	AC_CHECK_FUNCS(socket,,AC_MSG_ERROR(bailing out))
+fi
+
+if test x$HAVE_LIBDL = xyes; then
+	AC_CHECK_HEADERS(dlfcn.h,,AC_MSG_ERROR(bailing out))
+fi
+
+if test x$INTL = xyes; then
+	AC_CHECK_HEADERS(libintl.h,,AC_MSG_ERROR(bailing out))
+fi
+
+if test x$DEVMAPPER = xyes; then
+	AC_CHECK_HEADERS(libdevmapper.h,,AC_MSG_ERROR(bailing out))
+fi
+
+if test x$HAVE_SELINUX = xyes; then
+	AC_CHECK_HEADERS(selinux/selinux.h,,AC_MSG_ERROR(bailing out))
+fi
+
+################################################################################
 if test "-f VERSION"; then
   LVM_VERSION="\"`cat VERSION`\""
 else
   LVM_VERSION="Unknown"
 fi
 
+################################################################################
 AC_SUBST(JOBS)
 AC_SUBST(STATIC_LINK)
 AC_SUBST(LVM1)
+AC_SUBST(POOL)
+AC_SUBST(SNAPSHOTS)
+AC_SUBST(MIRRORS)
 AC_SUBST(OWNER)
 AC_SUBST(GROUP)
 AC_SUBST(CFLAGS)
+AC_SUBST(COPTIMISE_FLAG)
 AC_SUBST(CLDFLAGS)
 AC_SUBST(CLDWHOLEARCHIVE)
 AC_SUBST(CLDNOWHOLEARCHIVE)
-AC_SUBST(LD_DEPS)
-AC_SUBST(LD_FLAGS)
+AC_SUBST(LDDEPS)
+AC_SUBST(LDFLAGS)
 AC_SUBST(SOFLAG)
 AC_SUBST(LIBS)
 AC_SUBST(LVM_VERSION)
@@ -310,20 +547,31 @@ AC_SUBST(CONFDIR)
 AC_SUBST(STATICDIR)
 AC_SUBST(INTL_PACKAGE)
 AC_SUBST(INTL)
+AC_SUBST(CLVMD)
+AC_SUBST(CLUSTER)
+AC_SUBST(FSADM)
 
-dnl First and last lines should not contain files to generate in order to 
-dnl keep utility scripts running properly
+################################################################################
+dnl -- First and last lines should not contain files to generate in order to 
+dnl -- keep utility scripts running properly
 AC_OUTPUT( 								\
 Makefile								\
-make.tmpl                                                               \
+make.tmpl								\
+daemons/Makefile							\
+daemons/clvmd/Makefile							\
 doc/Makefile								\
 include/Makefile						 	\
 lib/Makefile							 	\
 lib/format1/Makefile						 	\
+lib/format_pool/Makefile						\
+lib/locking/Makefile							\
+lib/mirror/Makefile							\
+lib/snapshot/Makefile							\
 man/Makefile							 	\
 po/Makefile								\
 tools/Makefile							 	\
 tools/version.h								\
+tools/fsadm/Makefile							\
 test/mm/Makefile							\
 test/device/Makefile							\
 test/format1/Makefile							\
@@ -332,8 +580,9 @@ test/filters/Makefile                                                   \
 )
 
 if test x$ODIRECT != xyes; then
-  echo
-  echo Warning: O_DIRECT disabled.
-  echo Use of pvmove may cause machine to lock up under low memory conditions.
-  echo
+  AC_MSG_WARN(Warning: O_DIRECT disabled: low-memory pvmove may lock up)
+fi
+
+if test x$FSADM == xyes; then
+  AC_MSG_WARN(fsadm support is untested)
 fi

daemons/Makefile.in

@@ -0,0 +1,23 @@
+#
+# Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+#
+# This file is part of the LVM2.
+#
+# This copyrighted material is made available to anyone wishing to use,
+# modify, copy, or redistribute it subject to the terms and conditions
+# of the GNU General Public License v.2.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software Foundation,
+# Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+
+srcdir = @srcdir@
+top_srcdir = @top_srcdir@
+VPATH = @srcdir@
+
+ifeq ("@CLVMD@", "yes")
+  SUBDIRS = clvmd
+endif
+
+include $(top_srcdir)/make.tmpl
+

daemons/clvmd/Makefile.in

@@ -0,0 +1,49 @@
+#
+# Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+#
+# This file is part of the LVM2.
+#
+# This copyrighted material is made available to anyone wishing to use,
+# modify, copy, or redistribute it subject to the terms and conditions
+# of the GNU General Public License v.2.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software Foundation,
+# Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+
+srcdir = @srcdir@
+top_srcdir = @top_srcdir@
+VPATH = @srcdir@
+
+SOURCES = \
+	clvmd-cman.c     \
+	clvmd-command.c  \
+	clvmd.c          \
+	libclvm.c        \
+	lvm-functions.c  \
+	system-lv.c
+
+TARGETS = \
+	clvmd
+
+include $(top_srcdir)/make.tmpl
+
+CFLAGS += -D_REENTRANT -fno-strict-aliasing
+LIBS += -ldevmapper -ldlm -llvm -lpthread
+
+INSTALL_TARGETS = \
+	install_clvmd
+
+clvmd: $(OBJECTS) $(top_srcdir)/lib/liblvm.a
+	$(CC) -o clvmd $(OBJECTS) $(LDFLAGS) $(LVMLIBS) $(LIBS)
+
+.PHONY: install_clvmd
+
+install_clvmd: $(TARGETS)
+	$(INSTALL) -D $(OWNER) $(GROUP) -m 555 $(STRIP) clvmd \
+		$(sbindir)/clvmd
+
+install: $(INSTALL_TARGETS)
+
+install_cluster: $(INSTALL_TARGETS)
+

daemons/clvmd/clvm.h

@@ -0,0 +1,65 @@
+/*
+ * Copyright (C) 2002-2004 Sistina Software, Inc. All rights reserved.
+ * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU General Public License v.2.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+/* Definitions for CLVMD server and clients */
+
+/*
+ * The protocol spoken over the cluster and across the local socket.
+ */
+
+#ifndef _CLVM_H
+#define _CLVM_H
+
+struct clvm_header {
+	uint8_t  cmd;	        /* See below */
+	uint8_t  flags;	        /* See below */
+	uint16_t xid;	        /* Transaction ID */
+	uint32_t clientid;	/* Only used in Daemon->Daemon comms */
+	int32_t  status;	/* For replies, whether request succeeded */
+	uint32_t arglen;	/* Length of argument below. 
+				   If >1500 then it will be passed 
+				   around the cluster in the system LV */
+	char node[1];		/* Actually a NUL-terminated string, node name.
+				   If this is empty then the command is 
+				   forwarded to all cluster nodes unless 
+				   FLAG_LOCAL is also set. */
+	char args[1];		/* Arguments for the command follow the 
+				   node name, This member is only
+				   valid if the node name is empty */
+} __attribute__ ((packed));
+
+/* Flags */
+#define CLVMD_FLAG_LOCAL        1	/* Only do this on the local node */
+#define CLVMD_FLAG_SYSTEMLV     2	/* Data in system LV under my node name */
+
+/* Name of the local socket to communicate between libclvm and clvmd */
+//static const char CLVMD_SOCKNAME[]="/var/run/clvmd";
+static const char CLVMD_SOCKNAME[] = "\0clvmd";
+
+/* Internal commands & replies */
+#define CLVMD_CMD_REPLY    1
+#define CLVMD_CMD_VERSION  2	/* Send version around cluster when we start */
+#define CLVMD_CMD_GOAWAY   3	/* Die if received this - we are running 
+				   an incompatible version */
+#define CLVMD_CMD_TEST     4	/* Just for mucking about */
+
+#define CLVMD_CMD_LOCK              30
+#define CLVMD_CMD_UNLOCK            31
+
+/* Lock/Unlock commands */
+#define CLVMD_CMD_LOCK_LV           50
+#define CLVMD_CMD_LOCK_VG           51
+
+#endif

daemons/clvmd/clvmd-cman.c

@@ -0,0 +1,503 @@
+/*
+ * Copyright (C) 2002-2004 Sistina Software, Inc. All rights reserved.
+ * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU General Public License v.2.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+/*
+ * CMAN communication layer for clvmd.
+ */
+
+#include <pthread.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <sys/socket.h>
+#include <sys/uio.h>
+#include <sys/un.h>
+#include <sys/time.h>
+#include <sys/ioctl.h>
+#include <sys/utsname.h>
+#include <syslog.h>
+#include <netinet/in.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <signal.h>
+#include <unistd.h>
+#include <fcntl.h>
+#include <getopt.h>
+#include <errno.h>
+
+#include "clvmd-comms.h"
+#include "clvm.h"
+#include "libdlm.h"
+#include "log.h"
+#include "clvmd.h"
+#include "lvm-functions.h"
+
+#define LOCKSPACE_NAME "clvmd"
+
+static int cluster_sock;
+static int num_nodes;
+static struct cl_cluster_node *nodes = NULL;
+static int count_nodes; /* size of allocated nodes array */
+static int max_updown_nodes = 50;	/* Current size of the allocated array */
+/* Node up/down status, indexed by nodeid */
+static int *node_updown = NULL;
+static dlm_lshandle_t *lockspace;
+
+static void sigusr1_handler(int sig);
+static void count_clvmds_running(void);
+static void get_members(void);
+static int nodeid_from_csid(char *csid);
+static int name_from_nodeid(int nodeid, char *name);
+
+struct lock_wait {
+	pthread_cond_t cond;
+	pthread_mutex_t mutex;
+	struct dlm_lksb lksb;
+};
+
+int init_cluster()
+{
+	struct sockaddr_cl saddr;
+	int port = CLUSTER_PORT_CLVMD;
+
+	/* Open the cluster communication socket */
+	cluster_sock = socket(AF_CLUSTER, SOCK_DGRAM, CLPROTO_CLIENT);
+	if (cluster_sock == -1) {
+		perror("Can't open cluster socket");
+		return -1;
+	}
+
+	/* Bind to our port number on the cluster.
+	   Writes to this will block if the cluster loses quorum */
+	saddr.scl_family = AF_CLUSTER;
+	saddr.scl_port = port;
+
+	if (bind
+	    (cluster_sock, (struct sockaddr *) &saddr,
+	     sizeof(struct sockaddr_cl))) {
+		log_error("Can't bind cluster socket: %m");
+		return -1;
+	}
+
+	/* Get the cluster members list */
+	get_members();
+	count_clvmds_running();
+
+	/* Create a lockspace for LV & VG locks to live in */
+	lockspace = dlm_create_lockspace(LOCKSPACE_NAME, 0600);
+	if (!lockspace) {
+		log_error("Unable to create lockspace for CLVM\n");
+		return -1;
+	}
+	dlm_ls_pthread_init(lockspace);
+	return 0;
+}
+
+int get_main_cluster_fd()
+{
+	return cluster_sock;
+}
+
+int get_num_nodes()
+{
+	return num_nodes;
+}
+
+/* send_message with the fd check removed */
+int cluster_send_message(void *buf, int msglen, char *csid, const char *errtext)
+{
+	struct iovec iov[2];
+	struct msghdr msg;
+	struct sockaddr_cl saddr;
+	int len = 0;
+
+	msg.msg_control = NULL;
+	msg.msg_controllen = 0;
+	msg.msg_iovlen = 1;
+	msg.msg_iov = iov;
+	msg.msg_flags = 0;
+	iov[0].iov_len = msglen;
+	iov[0].iov_base = buf;
+
+	saddr.scl_family = AF_CLUSTER;
+	saddr.scl_port = CLUSTER_PORT_CLVMD;
+	if (csid) {
+		msg.msg_name = &saddr;
+		msg.msg_namelen = sizeof(saddr);
+		memcpy(&saddr.scl_nodeid, csid, MAX_CSID_LEN);
+	} else {		/* Cluster broadcast */
+
+		msg.msg_name = NULL;
+		msg.msg_namelen = 0;
+	}
+
+	do {
+		len = sendmsg(cluster_sock, &msg, 0);
+		if (len < 0 && errno != EAGAIN)
+			log_error(errtext);
+
+	} while (len == -1 && errno == EAGAIN);
+	return len;
+}
+
+void get_our_csid(char *csid)
+{
+	int i;
+	memset(csid, 0, MAX_CSID_LEN);
+
+	for (i = 0; i < num_nodes; i++) {
+		if (nodes[i].us)
+			memcpy(csid, &nodes[i].node_id, MAX_CSID_LEN);
+	}
+}
+
+/* Call a callback routine for each node that known (down mean not running a clvmd) */
+int cluster_do_node_callback(struct local_client *client,
+			     void (*callback) (struct local_client *, char *,
+					       int))
+{
+	int i;
+	int somedown = 0;
+
+	for (i = 0; i < get_num_nodes(); i++) {
+		callback(client, (char *)&nodes[i].node_id, node_updown[nodes[i].node_id]);
+		if (!node_updown[nodes[i].node_id])
+			somedown = -1;
+	}
+	return somedown;
+}
+
+/* Process OOB message from the cluster socket,
+   this currently just means that a node has stopped listening on our port */
+static void process_oob_msg(char *buf, int len, int nodeid)
+{
+	char namebuf[256];
+	switch (buf[0]) {
+        case CLUSTER_OOB_MSG_PORTCLOSED:
+		name_from_nodeid(nodeid, namebuf);
+		log_notice("clvmd on node %s has died\n", namebuf);
+		DEBUGLOG("Got OOB message, removing node %s\n", namebuf);
+
+		node_updown[nodeid] = 0;
+		break;
+
+	case CLUSTER_OOB_MSG_STATECHANGE:
+		DEBUGLOG("Got OOB message, Cluster state change\n");
+		get_members();
+		break;
+	default:
+		/* ERROR */
+		DEBUGLOG("Got unknown OOB message: %d\n", buf[0]);
+	}
+}
+
+int cluster_fd_callback(struct local_client *fd, char *buf, int len, char *csid,
+			struct local_client **new_client)
+{
+	struct iovec iov[2];
+	struct msghdr msg;
+	struct sockaddr_cl saddr;
+
+	/* We never return a new client */
+	*new_client = NULL;
+
+	msg.msg_control = NULL;
+	msg.msg_controllen = 0;
+	msg.msg_iovlen = 1;
+	msg.msg_iov = iov;
+	msg.msg_name = &saddr;
+	msg.msg_flags = 0;
+	msg.msg_namelen = sizeof(saddr);
+	iov[0].iov_len = len;
+	iov[0].iov_base = buf;
+
+	len = recvmsg(cluster_sock, &msg, MSG_OOB | O_NONBLOCK);
+	if (len < 0 && errno == EAGAIN)
+		return len;
+
+	DEBUGLOG("Read on cluster socket, len = %d\n", len);
+
+	/* A real error */
+	if (len < 0) {
+		log_error("read error on cluster socket: %m");
+		return 0;
+	}
+
+	/* EOF - we have left the cluster */
+	if (len == 0)
+		return 0;
+
+	/* Is it OOB? probably a node gone down */
+	if (msg.msg_flags & MSG_OOB) {
+		process_oob_msg(iov[0].iov_base, len, saddr.scl_nodeid);
+
+		/* Tell the upper layer to ignore this message */
+		len = -1;
+		errno = EAGAIN;
+	}
+	memcpy(csid, &saddr.scl_nodeid, sizeof(saddr.scl_nodeid));
+	return len;
+}
+
+void add_up_node(char *csid)
+{
+	/* It's up ! */
+	int nodeid = nodeid_from_csid(csid);
+
+	if (nodeid >= max_updown_nodes) {
+		int *new_updown = realloc(node_updown, max_updown_nodes + 10);
+
+		if (new_updown) {
+			node_updown = new_updown;
+			max_updown_nodes += 10;
+			DEBUGLOG("realloced more space for nodes. now %d\n",
+				 max_updown_nodes);
+		} else {
+			log_error
+			    ("Realloc failed. Node status for clvmd will be wrong\n");
+			return;
+		}
+	}
+	node_updown[nodeid] = 1;
+	DEBUGLOG("Added new node %d to updown list\n", nodeid);
+}
+
+void cluster_closedown()
+{
+	unlock_all();
+	dlm_release_lockspace(LOCKSPACE_NAME, lockspace, 1);
+	close(cluster_sock);
+}
+
+static int is_listening(int nodeid)
+{
+	struct cl_listen_request rq;
+	int status;
+
+	rq.port = CLUSTER_PORT_CLVMD;
+	rq.nodeid = nodeid;
+
+	do {
+		status = ioctl(cluster_sock, SIOCCLUSTER_ISLISTENING, &rq);
+		if (status < 0 && errno == EBUSY) {	/* Don't busywait */
+			sleep(1);
+			errno = EBUSY;	/* In case sleep trashes it */
+		}
+	}
+	while (status < 0 && errno == EBUSY);
+
+	return status;
+}
+
+/* Populate the list of CLVMDs running.
+   called only at startup time */
+void count_clvmds_running(void)
+{
+	int i;
+
+	for (i = 0; i < num_nodes; i++) {
+		node_updown[nodes[i].node_id] = is_listening(nodes[i].node_id);
+	}
+}
+
+/* Get a list of active cluster members */
+static void get_members()
+{
+	struct cl_cluster_nodelist nodelist;
+
+	num_nodes = ioctl(cluster_sock, SIOCCLUSTER_GETMEMBERS, 0);
+	if (num_nodes == -1) {
+		perror("get nodes");
+	} else {
+	        /* Not enough room for new nodes list ? */
+	        if (num_nodes > count_nodes && nodes) {
+			free(nodes);
+			nodes = NULL;
+		}
+
+		if (nodes == NULL) {
+		        count_nodes = num_nodes + 10; /* Overallocate a little */
+		        nodes = malloc(count_nodes * sizeof(struct cl_cluster_node));
+			if (!nodes) {
+			        perror("Unable to allocate nodes array\n");
+				exit(5);
+			}
+		}
+		nodelist.max_members = count_nodes;
+		nodelist.nodes = nodes;
+
+		num_nodes = ioctl(cluster_sock, SIOCCLUSTER_GETMEMBERS, &nodelist);
+		if (num_nodes <= 0) {
+		        perror("get node details");
+			exit(6);
+		}
+
+		/* Sanity check struct */
+		if (nodes[0].size != sizeof(struct cl_cluster_node)) {
+			log_error
+			    ("sizeof(cl_cluster_node) does not match size returned from the kernel: aborting\n");
+			exit(10);
+		}
+
+		if (node_updown == NULL) {
+			node_updown =
+			    (int *) malloc(sizeof(int) *
+					   max(num_nodes, max_updown_nodes));
+			memset(node_updown, 0,
+			       sizeof(int) * max(num_nodes, max_updown_nodes));
+		}
+	}
+}
+
+/* Convert a node name to a CSID */
+int csid_from_name(char *csid, char *name)
+{
+	int i;
+
+	for (i = 0; i < num_nodes; i++) {
+		if (strcmp(name, nodes[i].name) == 0) {
+			memcpy(csid, &nodes[i].node_id, MAX_CSID_LEN);
+			return 0;
+		}
+	}
+	return -1;
+}
+
+/* Convert a CSID to a node name */
+int name_from_csid(char *csid, char *name)
+{
+	int i;
+
+	for (i = 0; i < num_nodes; i++) {
+		if (memcmp(csid, &nodes[i].node_id, MAX_CSID_LEN) == 0) {
+			strcpy(name, nodes[i].name);
+			return 0;
+		}
+	}
+	/* Who?? */
+	strcpy(name, "Unknown");
+	return -1;
+}
+
+/* Convert a node ID to a node name */
+int name_from_nodeid(int nodeid, char *name)
+{
+	int i;
+
+	for (i = 0; i < num_nodes; i++) {
+		if (nodeid == nodes[i].node_id) {
+			strcpy(name, nodes[i].name);
+			return 0;
+		}
+	}
+	/* Who?? */
+	strcpy(name, "Unknown");
+	return -1;
+}
+
+/* Convert a CSID to a node ID */
+static int nodeid_from_csid(char *csid)
+{
+        int nodeid;
+
+	memcpy(&nodeid, csid, MAX_CSID_LEN);
+
+	return nodeid;
+}
+
+int is_quorate()
+{
+	return ioctl(cluster_sock, SIOCCLUSTER_ISQUORATE, 0);
+}
+
+static void sync_ast_routine(void *arg)
+{
+	struct lock_wait *lwait = arg;
+
+	pthread_mutex_lock(&lwait->mutex);
+	pthread_cond_signal(&lwait->cond);
+	pthread_mutex_unlock(&lwait->mutex);
+}
+
+int sync_lock(const char *resource, int mode, int flags, int *lockid)
+{
+	int status;
+	struct lock_wait lwait;
+
+	if (!lockid) {
+		errno = EINVAL;
+		return -1;
+	}
+
+	DEBUGLOG("sync_lock: '%s' mode:%d flags=%d\n", resource,mode,flags);
+	/* Conversions need the lockid in the LKSB */
+	if (flags & LKF_CONVERT)
+		lwait.lksb.sb_lkid = *lockid;
+
+	pthread_cond_init(&lwait.cond, NULL);
+	pthread_mutex_init(&lwait.mutex, NULL);
+	pthread_mutex_lock(&lwait.mutex);
+
+	status = dlm_ls_lock(lockspace,
+			     mode,
+			     &lwait.lksb,
+			     flags,
+			     resource,
+			     strlen(resource),
+			     0, sync_ast_routine, &lwait, NULL, NULL);
+	if (status)
+		return status;
+
+	/* Wait for it to complete */
+	pthread_cond_wait(&lwait.cond, &lwait.mutex);
+	pthread_mutex_unlock(&lwait.mutex);
+
+	*lockid = lwait.lksb.sb_lkid;
+
+	errno = lwait.lksb.sb_status;
+	DEBUGLOG("sync_lock: returning lkid %x\n", *lockid);
+	if (lwait.lksb.sb_status)
+		return -1;
+	else
+		return 0;
+}
+
+int sync_unlock(const char *resource /* UNUSED */, int lockid)
+{
+	int status;
+	struct lock_wait lwait;
+
+	DEBUGLOG("sync_unlock: '%s' lkid:%x\n", resource, lockid);
+
+	pthread_cond_init(&lwait.cond, NULL);
+	pthread_mutex_init(&lwait.mutex, NULL);
+	pthread_mutex_lock(&lwait.mutex);
+
+	status = dlm_ls_unlock(lockspace, lockid, 0, &lwait.lksb, &lwait);
+
+	if (status)
+		return status;
+
+	/* Wait for it to complete */
+	pthread_cond_wait(&lwait.cond, &lwait.mutex);
+	pthread_mutex_unlock(&lwait.mutex);
+
+	errno = lwait.lksb.sb_status;
+	if (lwait.lksb.sb_status != EUNLOCK)
+		return -1;
+	else
+		return 0;
+
+}

daemons/clvmd/clvmd-command.c

@@ -0,0 +1,285 @@
+/*
+ * Copyright (C) 2002-2004 Sistina Software, Inc. All rights reserved.
+ * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU General Public License v.2.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+/*
+
+  CLVMD Cluster LVM daemon command processor.
+
+  To add commands to the daemon simply add a processor in do_command and return
+  and messages back in buf and the length in *retlen. The initial value of
+  buflen is the maximum size of the buffer. if buf is not large enough then it
+  may be reallocated by the functions in here to a suitable size bearing in
+  mind that anything larger than the passed-in size will have to be returned
+  using the system LV and so performance will suffer.
+
+  The status return will be negated and passed back to the originating node.
+
+  pre- and post- command routines are called only on the local node. The
+  purpose is primarily to get and release locks, though the pre- routine should
+  also do any other local setups required by the command (if any) and can
+  return a failure code that prevents the command from being distributed around
+  the cluster
+
+  The pre- and post- routines are run in their own thread so can block as long
+  they like, do_command is run in the main clvmd thread so should not block for
+  too long. If the pre-command returns an error code (!=0) then the command
+  will not be propogated around the cluster but the post-command WILL be called
+
+  Also note that the pre and post routine are *always* called on the local
+  node, even if the command to be executed was only requested to run on a
+  remote node. It may peek inside the client structure to check the status of
+  the command.
+
+  The clients of the daemon must, naturally, understand the return messages and
+  codes.
+
+  Routines in here may only READ the values in the client structure passed in
+  apart from client->private which they are free to do what they like with.
+
+*/
+
+#include <pthread.h>
+#include <sys/types.h>
+#include <sys/utsname.h>
+#include <sys/ioctl.h>
+#include <sys/socket.h>
+#include <sys/stat.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdint.h>
+#include <fcntl.h>
+#include <string.h>
+#include <stddef.h>
+#include <unistd.h>
+#include <errno.h>
+
+#include "list.h"
+#include "hash.h"
+#include "locking.h"
+#include "log.h"
+#include "lvm-functions.h"
+#include "clvmd-comms.h"
+#include "clvm.h"
+#include "clvmd.h"
+#include "libdlm.h"
+
+/* This is where all the real work happens:
+   NOTE: client will be NULL when this is executed on a remote node */
+int do_command(struct local_client *client, struct clvm_header *msg, int msglen,
+	       char **buf, int buflen, int *retlen)
+{
+	char *args = msg->node + strlen(msg->node) + 1;
+	int arglen = msglen - sizeof(struct clvm_header) - strlen(msg->node);
+	int status = 0;
+	char *lockname;
+	struct utsname nodeinfo;
+	unsigned char lock_cmd;
+	unsigned char lock_flags;
+
+	/* Do the command */
+	switch (msg->cmd) {
+		/* Just a test message */
+	case CLVMD_CMD_TEST:
+		if (arglen > buflen) {
+			buflen = arglen + 200;
+			*buf = realloc(*buf, buflen);
+		}
+		uname(&nodeinfo);
+		*retlen = 1 + snprintf(*buf, buflen, "TEST from %s: %s v%s",
+				       nodeinfo.nodename, args,
+				       nodeinfo.release);
+		break;
+
+	case CLVMD_CMD_LOCK_VG:
+		/* Check to see if the VG is in use by LVM1 */
+		status = do_check_lvm1(&args[2]);
+		break;
+
+	case CLVMD_CMD_LOCK_LV:
+		/* This is the biggie */
+		lock_cmd = args[0];
+		lock_flags = args[1];
+		lockname = &args[2];
+		status = do_lock_lv(lock_cmd, lock_flags, lockname);
+		/* Replace EIO with something less scary */
+		if (status == EIO) {
+			*retlen =
+			    1 + snprintf(*buf, buflen,
+					 "Internal lvm error, check syslog");
+			return EIO;
+		}
+		break;
+
+	default:
+		/* Won't get here because command is validated in pre_command */
+		break;
+	}
+
+	/* Check the status of the command and return the error text */
+	if (status) {
+		*retlen = 1 + snprintf(*buf, buflen, strerror(status));
+	}
+
+	return status;
+
+}
+
+static int lock_vg(struct local_client *client)
+{
+    struct hash_table *lock_hash;
+    struct clvm_header *header =
+	(struct clvm_header *) client->bits.localsock.cmd;
+    unsigned char lock_cmd;
+    unsigned char lock_flags;
+    char *args = header->node + strlen(header->node) + 1;
+    int lkid;
+    int status = 0;
+    char *lockname;
+
+    /* Keep a track of VG locks in our own hash table. In current
+       practice there should only ever be more than two VGs locked
+       if a user tries to merge lots of them at once */
+    if (client->bits.localsock.private) {
+	lock_hash = (struct hash_table *)client->bits.localsock.private;
+    }
+    else {
+	lock_hash = hash_create(3);
+	if (!lock_hash)
+	    return ENOMEM;
+	client->bits.localsock.private = (void *)lock_hash;
+    }
+
+    lock_cmd = args[0];
+    lock_flags = args[1];
+    lockname = &args[2];
+    DEBUGLOG("doing PRE command LOCK_VG '%s' at %x (client=%p)\n", lockname, lock_cmd, client);
+
+    if (lock_cmd == LCK_UNLOCK) {
+
+	lkid = (int)(long)hash_lookup(lock_hash, lockname);
+	if (lkid == 0)
+	    return EINVAL;
+
+	status = sync_unlock(lockname, lkid);
+	if (status)
+	    status = errno;
+	else
+	    hash_remove(lock_hash, lockname);
+    }
+    else {
+
+	status = sync_lock(lockname, (int)lock_cmd, (int)lock_flags, &lkid);
+	if (status)
+	    status = errno;
+	else
+	    hash_insert(lock_hash, lockname, (void *)lkid);
+    }
+
+    return status;
+}
+
+
+/* Pre-command is a good place to get locks that are needed only for the duration
+   of the commands around the cluster (don't forget to free them in post-command),
+   and to sanity check the command arguments */
+int do_pre_command(struct local_client *client)
+{
+	struct clvm_header *header =
+	    (struct clvm_header *) client->bits.localsock.cmd;
+	unsigned char lock_cmd;
+	unsigned char lock_flags;
+	char *args = header->node + strlen(header->node) + 1;
+	int lockid;
+	int status = 0;
+	char *lockname;
+
+	switch (header->cmd) {
+	case CLVMD_CMD_TEST:
+		status = sync_lock("CLVMD_TEST", LKM_EXMODE, 0, &lockid);
+		client->bits.localsock.private = (void *) lockid;
+		break;
+
+	case CLVMD_CMD_LOCK_VG:
+       	        status = lock_vg(client);
+		break;
+
+	case CLVMD_CMD_LOCK_LV:
+		lock_cmd = args[0];
+		lock_flags = args[1];
+		lockname = &args[2];
+		status = pre_lock_lv(lock_cmd, lock_flags, lockname);
+		break;
+
+	default:
+		log_error("Unknown command %d received\n", header->cmd);
+		status = EINVAL;
+	}
+	return status;
+}
+
+/* Note that the post-command routine is called even if the pre-command or the real command
+   failed */
+int do_post_command(struct local_client *client)
+{
+	struct clvm_header *header =
+	    (struct clvm_header *) client->bits.localsock.cmd;
+	int status = 0;
+	unsigned char lock_cmd;
+	unsigned char lock_flags;
+	char *args = header->node + strlen(header->node) + 1;
+	char *lockname;
+
+	switch (header->cmd) {
+	case CLVMD_CMD_TEST:
+		status =
+		    sync_unlock("CLVMD_TEST", (int) (long) client->bits.localsock.private);
+		client->bits.localsock.private = 0;
+		break;
+
+	case CLVMD_CMD_LOCK_VG:
+		/* Nothing to do here */
+		break;
+
+	case CLVMD_CMD_LOCK_LV:
+		lock_cmd = args[0];
+		lock_flags = args[1];
+		lockname = &args[2];
+		status = post_lock_lv(lock_cmd, lock_flags, lockname);
+		break;
+	}
+	return status;
+}
+
+
+/* Called when the client is about to be deleted */
+void cmd_client_cleanup(struct local_client *client)
+{
+    if (client->bits.localsock.private) {
+
+	struct hash_node *v;
+	struct hash_table *lock_hash =
+	    (struct hash_table *)client->bits.localsock.private;
+
+	hash_iterate(v, lock_hash) {
+		int lkid = (int)(long)hash_get_data(lock_hash, v);
+
+		DEBUGLOG("cleanup: Unlocking lkid %x\n", lkid);
+		sync_unlock("DUMMY", lkid);
+	}
+
+	hash_destroy(lock_hash);
+	client->bits.localsock.private = 0;
+    }
+}

daemons/clvmd/clvmd-comms.h

@@ -0,0 +1,55 @@
+/*
+ * Copyright (C) 2002-2004 Sistina Software, Inc. All rights reserved.
+ * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU General Public License v.2.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+/*
+ * Abstraction layer for clvmd cluster communications
+ */
+
+#ifndef _CLVMD_COMMS_H
+#define _CLVMD_COMMS_H
+
+struct local_client;
+
+extern int cluster_send_message(void *buf, int msglen, char *csid,
+				const char *errtext);
+extern int name_from_csid(char *csid, char *name);
+extern int csid_from_name(char *csid, char *name);
+extern int get_num_nodes(void);
+extern int cluster_fd_callback(struct local_client *fd, char *buf, int len,
+			       char *csid, struct local_client **new_client);
+extern int init_cluster(void);
+extern int get_main_cluster_fd(void);	/* gets accept FD or cman cluster socket */
+extern int cluster_do_node_callback(struct local_client *client,
+				    void (*callback) (struct local_client *,
+						      char *csid, int node_up));
+extern int is_quorate(void);
+
+extern void get_our_csid(char *csid);
+extern void add_up_node(char *csid);
+extern void cluster_closedown(void);
+
+extern int sync_lock(const char *resource, int mode, int flags, int *lockid);
+extern int sync_unlock(const char *resource, int lockid);
+
+#ifdef USE_GULM
+#include "tcp-comms.h"
+#else
+/* cman */
+#include "cnxman-socket.h"
+#define MAX_CSID_LEN 4
+#endif
+
+
+#endif

daemons/clvmd/clvmd-gulm.c

@@ -0,0 +1,880 @@
+/******************************************************************************
+*******************************************************************************
+**
+**  Copyright (C) Sistina Software, Inc.  2002-2003  All rights reserved.
+**
+*******************************************************************************
+******************************************************************************/
+
+/* This provides the interface between clvmd and gulm as the cluster
+ * and lock manager.
+ *
+ * It also provides the "liblm" functions too as it's hard (and pointless)
+ * to seperate them out when using gulm.
+ *
+ * What it does /not/ provide is the communications between clvmd daemons
+ * on the cluster nodes. That is done in tcp-comms.c
+ */
+
+
+#include <pthread.h>
+#include <sys/types.h>
+#include <sys/utsname.h>
+#include <sys/ioctl.h>
+#include <sys/socket.h>
+#include <sys/stat.h>
+#include <sys/file.h>
+#include <sys/socket.h>
+#include <netinet/in.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdint.h>
+#include <signal.h>
+#include <fcntl.h>
+#include <string.h>
+#include <stddef.h>
+#include <stdint.h>
+#include <unistd.h>
+#include <errno.h>
+#include <utmpx.h>
+#include <syslog.h>
+#include <assert.h>
+
+#include "ccs.h"
+#include "list.h"
+#include "locking.h"
+#include "log.h"
+#include "clvm.h"
+#include "clvmd-comms.h"
+#include "clvmd.h"
+#include "hash.h"
+#include "clvmd-gulm.h"
+#include "libgulm.h"
+#include "hash.h"
+
+/* Hash list of nodes in the cluster */
+static struct hash_table *node_hash;
+
+/* hash list of outstanding lock requests */
+static struct hash_table *lock_hash;
+
+/* Copy of the current core state */
+static uint8_t current_corestate;
+
+/* Number of active nodes */
+static int num_nodes;
+
+static char *cluster_name;
+
+static pthread_mutex_t lock_start_mutex;
+static volatile int lock_start_flag;
+
+struct node_info
+{
+    enum {NODE_UNKNOWN, NODE_DOWN, NODE_UP, NODE_CLVMD} state;
+    char name[MAX_CLUSTER_MEMBER_NAME_LEN];
+};
+
+struct lock_wait
+{
+    pthread_cond_t cond;
+    pthread_mutex_t mutex;
+    int status;
+};
+
+/* Forward */
+static int read_from_core_sock(struct local_client *client, char *buf, int len, char *csid,
+			       struct local_client **new_client);
+static int read_from_lock_sock(struct local_client *client, char *buf, int len, char *csid,
+			       struct local_client **new_client);
+static int get_all_cluster_nodes(void);
+
+/* In tcp-comms.c */
+extern struct hash_table *sock_hash;
+
+static int add_internal_client(int fd, fd_callback_t callback)
+{
+    struct local_client *client;
+
+    DEBUGLOG("Add_internal_client, fd = %d\n", fd);
+
+    /* Add a GULM file descriptor it to the main loop */
+    client = malloc(sizeof(struct local_client));
+    if (!client)
+    {
+	DEBUGLOG("malloc failed\n");
+	return -1;
+    }
+
+    memset(client, 0, sizeof(struct local_client));
+    client->fd = fd;
+    client->type = CLUSTER_INTERNAL;
+    client->callback = callback;
+    add_client(client);
+
+    return 0;
+}
+
+/* Gulm library handle */
+static gulm_interface_p gulm_if;
+static lg_core_callbacks_t core_callbacks;
+static lg_lockspace_callbacks_t lock_callbacks;
+
+static void badsig_handler(int sig)
+{
+    DEBUGLOG("got sig %d\n", sig);
+    cluster_closedown();
+    exit(0);
+}
+
+static void sighup_handler(int sig)
+{
+    DEBUGLOG("got SIGHUP\n");
+
+    /* Re-read CCS node list */
+    get_all_cluster_nodes();
+}
+
+int init_cluster()
+{
+    int status;
+    int ccs_h;
+
+    /* Get cluster name from CCS */
+    /* TODO: is this right? */
+    ccs_h = ccs_connect();
+    ccs_get(ccs_h, "//cluster/@name", &cluster_name);
+    ccs_disconnect(ccs_h);
+
+    /* Block locking until we are logged in */
+    pthread_mutex_init(&lock_start_mutex, NULL);
+    pthread_mutex_lock(&lock_start_mutex);
+    lock_start_flag = 1;
+
+    node_hash = hash_create(100);
+    lock_hash = hash_create(10);
+
+    /* Get all nodes from CCS */
+    get_all_cluster_nodes();
+
+    /* Initialise GULM library */
+    status = lg_initialize(&gulm_if, cluster_name, "clvmd");
+    if (status)
+    {
+	DEBUGLOG("lg_initialize failed: %d\n", status);
+	return status;
+    }
+
+    /* Connect to core - we are not "important" :-) */
+    status = lg_core_login(gulm_if, 0);
+    if (status)
+    {
+	DEBUGLOG("lg_core_login failed: %d\n", status);
+	return status;
+    }
+
+    /* Initialise the inter-node comms */
+    status = init_comms();
+    if (status)
+	return status;
+
+    /* Add core FD to the list */
+    status = add_internal_client(lg_core_selector(gulm_if), read_from_core_sock);
+    if (status)
+    {
+	DEBUGLOG("can't allocate client space\n");
+	return status;
+    }
+
+    /* Connect to the lock server */
+    if (lg_lock_login(gulm_if, "CLVM"))
+    {
+	syslog(LOG_ERR, "Cannot login in to LOCK server\n");
+	DEBUGLOG("Cannot login in to LOCK server\n");
+	exit(88);
+    }
+
+    /* Add lockspace FD to the list */
+    status = add_internal_client(lg_lock_selector(gulm_if), read_from_lock_sock);
+    if (status)
+    {
+	DEBUGLOG("can't allocate client space\n");
+	exit(status);
+    }
+
+    /* Request a list of nodes, we can;t really do anything until
+       this comes back */
+    status = lg_core_nodelist(gulm_if);
+    if (status)
+    {
+	DEBUGLOG("lg_core_nodelist failed: %d\n", status);
+	return status;
+    }
+
+    /* So I can kill it without taking GULM down too */
+    signal(SIGINT, badsig_handler);
+    signal(SIGTERM, badsig_handler);
+
+    /* Re-read the node list on SIGHUP */
+    signal(SIGHUP, sighup_handler);
+
+    return 0;
+}
+
+void cluster_closedown()
+{
+    DEBUGLOG("cluster_closedown\n");
+    lg_lock_logout(gulm_if);
+    lg_core_logout(gulm_if);
+    lg_core_shutdown(gulm_if);
+    lg_release(gulm_if);
+}
+
+/* Expire locks for a named node, or us */
+#define GIO_KEY_SIZE 46
+static void drop_expired_locks(char *nodename)
+{
+    struct utsname nodeinfo;
+    uint8_t mask[GIO_KEY_SIZE];
+
+    memset(mask, 0xff, GIO_KEY_SIZE);
+
+    if (!nodename)
+    {
+	uname(&nodeinfo);
+	nodename = nodeinfo.nodename;
+    }
+
+    if (lg_lock_drop_exp(gulm_if, nodename, mask, GIO_KEY_SIZE))
+    {
+	DEBUGLOG("Error calling lg_lock_drop_exp()\n");
+    }
+}
+
+
+static int read_from_core_sock(struct local_client *client, char *buf, int len, char *csid,
+			       struct local_client **new_client)
+{
+    int status;
+
+    *new_client = NULL;
+    status = lg_core_handle_messages(gulm_if, &core_callbacks, NULL);
+    return status<0 ? status : 1;
+}
+
+static int read_from_lock_sock(struct local_client *client, char *buf, int len, char *csid,
+			       struct local_client **new_client)
+{
+    int status;
+
+    *new_client = NULL;
+    status = lg_lock_handle_messages(gulm_if, &lock_callbacks, NULL);
+    return status<0 ? status : 1;
+}
+
+
+/* CORE callback routines */
+static int core_login_reply(void *misc, uint64_t gen, uint32_t error, uint32_t rank, uint8_t corestate)
+{
+   DEBUGLOG("CORE Got a Login reply.  gen:%lld err:%d rank:%d corestate:%d\n",
+         gen, error, rank, corestate);
+
+   if (error)
+       exit(error);
+
+   current_corestate = corestate;
+   return 0;
+}
+
+static void set_node_state(struct node_info *ninfo, char *csid, uint8_t nodestate)
+{
+    if (nodestate == lg_core_Logged_in)
+    {
+	/* Don't clobber NODE_CLVMD state */
+	if (ninfo->state != NODE_CLVMD)
+	{
+	    if (ninfo->state == NODE_UNKNOWN ||
+		ninfo->state == NODE_DOWN)
+		num_nodes++;
+
+	    ninfo->state = NODE_UP;
+	}
+    }
+    else
+    {
+	if (nodestate == lg_core_Expired ||
+	    nodestate == lg_core_Fenced ||
+	    nodestate == lg_core_Logged_out)
+	{
+	    if (ninfo->state != NODE_DOWN)
+		num_nodes--;
+	    ninfo->state = NODE_DOWN;
+	    tcp_remove_client(csid);
+	}
+    }
+    DEBUGLOG("set_node_state, '%s' state = %d, num_nodes=%d\n",
+	     ninfo->name, ninfo->state, num_nodes);
+}
+
+static struct node_info *add_or_set_node(char *name, uint32_t ip, uint8_t state)
+{
+    struct node_info *ninfo;
+
+    ninfo = hash_lookup_binary(node_hash, (char *)&ip, MAX_CSID_LEN);
+    if (!ninfo)
+    {
+	/* If we can't find that node then re-read the config file in case it
+	   was added after we were started */
+	DEBUGLOG("Node %s not found, re-reading config file\n", name);
+	get_all_cluster_nodes();
+
+	/* Now try again */
+	ninfo = hash_lookup_binary(node_hash, (char *)&ip, MAX_CSID_LEN);
+	if (!ninfo)
+	{
+	    DEBUGLOG("Ignoring node %s, not part of the SAN cluster\n", name);
+	    return NULL;
+	}
+    }
+
+    set_node_state(ninfo, (char *)&ip, state);
+
+    return ninfo;
+}
+
+static int core_nodelist(void *misc, lglcb_t type, char *name, uint32_t ip, uint8_t state)
+{
+    DEBUGLOG("CORE nodelist\n");
+
+    if (type == lglcb_start)
+    {
+	DEBUGLOG("Got Nodelist, start\n");
+    }
+    else
+    {
+	if (type == lglcb_item)
+	{
+	    DEBUGLOG("Got nodelist, item: %s, %#x, %#x\n", name, ip, state);
+
+	    add_or_set_node(name, ip, state);
+	}
+	else
+	{
+	    if (type == lglcb_stop)
+	    {
+		char ourcsid[MAX_CSID_LEN];
+
+		DEBUGLOG("Got Nodelist, stop\n");
+		clvmd_cluster_init_completed();
+
+		/* Mark ourself as up */
+		get_our_csid(ourcsid);
+		add_up_node(ourcsid);
+	    }
+	    else
+	    {
+		DEBUGLOG("Unknown lglcb_t %#x\n", type);
+	    }
+	}
+    }
+
+    return 0;
+}
+
+static int core_statechange(void *misc, uint8_t corestate, uint32_t masterip, char *mastername)
+{
+    DEBUGLOG("CORE Got statechange  corestate:%#x masterip:%#x mastername:%s\n",
+	     corestate, masterip, mastername);
+
+    current_corestate = corestate;
+    return 0;
+}
+
+static int core_nodechange(void *misc, char *nodename, uint32_t nodeip, uint8_t nodestate)
+{
+    struct node_info *ninfo;
+
+    DEBUGLOG("CORE node change, name=%s, ip=%x, state = %d\n", nodename, nodeip, nodestate);
+
+    /* If we don't get nodeip here, try a lookup by name */
+    if (!nodeip)
+	csid_from_name((char *)&nodeip, nodename);
+    if (!nodeip)
+	return 0;
+
+    ninfo = add_or_set_node(nodename, nodeip, nodestate);
+    if (!ninfo)
+	return 0;
+
+    /* Check if we need to drop any expired locks */
+    if (ninfo->state == NODE_DOWN)
+    {
+	drop_expired_locks(nodename);
+    }
+
+    return 0;
+}
+static int core_error(void *misc, uint32_t err)
+{
+    DEBUGLOG("CORE error: %d\n", err);
+    // Not sure what happens here
+    return 0;
+}
+
+/* LOCK callback routines */
+static int lock_login_reply(void *misc, uint32_t error, uint8_t which)
+{
+    DEBUGLOG("LOCK Got a Login reply.  err:%d which:%d\n",
+	     error, which);
+
+    if (error)
+	exit(error);
+
+    /* Drop any expired locks for us that might be hanging around */
+    drop_expired_locks(NULL);
+
+    /* Enable locking operations in other threads */
+    if (lock_start_flag)
+    {
+	lock_start_flag = 0;
+	pthread_mutex_unlock(&lock_start_mutex);
+    }
+
+    return 0;
+}
+
+static int lock_lock_state(void *misc, uint8_t *key, uint16_t keylen, uint8_t state, uint32_t flags, uint32_t error,
+			   uint8_t *LVB, uint16_t LVBlen)
+{
+    struct lock_wait *lwait;
+
+    DEBUGLOG("LOCK lock state: %s, error = %d\n", key, error);
+
+    lwait = hash_lookup(lock_hash, key);
+    if (!lwait)
+    {
+	DEBUGLOG("Can't find hash entry for resource %s\n", key);
+	return 0;
+    }
+    lwait->status = error;
+    pthread_mutex_lock(&lwait->mutex);
+    pthread_cond_signal(&lwait->cond);
+    pthread_mutex_unlock(&lwait->mutex);
+
+    return 0;
+}
+static int lock_error(void *misc, uint32_t err)
+{
+    DEBUGLOG("LOCK error: %d\n", err);
+    // Not sure what happens here
+    return 0;
+}
+
+
+/* CORE callbacks */
+static lg_core_callbacks_t core_callbacks = {
+    .login_reply  = core_login_reply,
+    .nodelist     = core_nodelist,
+    .statechange  = core_statechange,
+    .nodechange   = core_nodechange,
+    .error        = core_error,
+};
+
+/* LOCK callbacks */
+static lg_lockspace_callbacks_t lock_callbacks = {
+    .login_reply   = lock_login_reply,
+    .lock_state    = lock_lock_state,
+    .error         = lock_error,
+};
+
+/* Allow tcp-comms to loop round the list of active nodes */
+int get_next_node_csid(void **context, char *csid)
+{
+    struct node_info *ninfo = NULL;
+
+    /* First node */
+    if (!*context)
+    {
+	*context = hash_get_first(node_hash);
+    }
+    else
+    {
+	*context = hash_get_next(node_hash, *context);
+    }
+    if (*context)
+	ninfo = hash_get_data(node_hash, *context);
+
+    /* Find a node that is UP */
+    while (*context && ninfo->state == NODE_DOWN)
+    {
+	*context = hash_get_next(node_hash, *context);
+	if (*context)
+	{
+	    ninfo = hash_get_data(node_hash, *context);
+	}
+    }
+
+    if (!*context || ninfo->state == NODE_DOWN)
+    {
+	return 0;
+    }
+
+    memcpy(csid, hash_get_key(node_hash, *context), MAX_CSID_LEN);
+    return 1;
+}
+
+int name_from_csid(char *csid, char *name)
+{
+    struct node_info *ninfo;
+
+    ninfo = hash_lookup_binary(node_hash, csid, MAX_CSID_LEN);
+    if (!ninfo)
+    {
+	sprintf(name, "UNKNOWN [%d.%d.%d.%d]",
+		csid[0], csid[1], csid[2], csid[3]);
+	return -1;
+    }
+
+    strcpy(name, ninfo->name);
+    return 0;
+}
+
+
+int csid_from_name(char *csid, char *name)
+{
+    struct hash_node *hn;
+    struct node_info *ninfo;
+
+    hash_iterate(hn, node_hash)
+    {
+	ninfo = hash_get_data(node_hash, hn);
+	if (strcmp(ninfo->name, name) == 0)
+	{
+	    memcpy(csid, hash_get_key(node_hash, hn), MAX_CSID_LEN);
+	    return 0;
+	}
+    }
+    return -1;
+}
+
+int get_num_nodes()
+{
+    DEBUGLOG("num_nodes = %d\n", num_nodes);
+    return num_nodes;
+}
+
+/* Node is now known to be running a clvmd */
+void add_up_node(char *csid)
+{
+    struct node_info *ninfo;
+
+    ninfo = hash_lookup_binary(node_hash, csid, MAX_CSID_LEN);
+    if (!ninfo)
+	return;
+
+    ninfo->state = NODE_CLVMD;
+    return;
+
+}
+/* Node is now known to be NOT running a clvmd */
+void add_down_node(char *csid)
+{
+    struct node_info *ninfo;
+
+    ninfo = hash_lookup_binary(node_hash, csid, MAX_CSID_LEN);
+    if (!ninfo)
+	return;
+
+    /* Only set it to UP if it was previously known to be
+       running clvmd - gulm may set it DOWN quite soon */
+    if (ninfo->state == NODE_CLVMD)
+	ninfo->state = NODE_UP;
+    return;
+
+}
+
+/* Call a callback for each node, so the caller knows whether it's up or down */
+int cluster_do_node_callback(struct local_client *master_client,
+			     void (*callback)(struct local_client *, char *csid, int node_up))
+{
+    struct hash_node *hn;
+    struct node_info *ninfo;
+
+    hash_iterate(hn, node_hash)
+    {
+	char csid[MAX_CSID_LEN];
+	struct local_client *client;
+
+	ninfo = hash_get_data(node_hash, hn);
+	memcpy(csid, hash_get_key(node_hash, hn), MAX_CSID_LEN);
+
+	DEBUGLOG("down_callback. node %s, state = %d\n", ninfo->name, ninfo->state);
+
+	client = hash_lookup_binary(sock_hash, csid, MAX_CSID_LEN);
+	if (client)
+	    callback(master_client, csid, ninfo->state == NODE_CLVMD);
+    }
+    return 0;
+}
+
+/* Convert gulm error codes to unix errno numbers */
+static int gulm_to_errno(int gulm_ret)
+{
+    switch (gulm_ret)
+    {
+    case lg_err_TryFailed:
+	errno = EAGAIN;
+	break;
+
+    case lg_err_AlreadyPend:
+	errno = EBUSY;
+
+	/* More?? */
+    default:
+	errno = EINVAL;
+    }
+
+    return gulm_ret ? -1 : 0;
+}
+
+/* Real locking */
+static int _lock_resource(char *resource, int mode, int flags, int *lockid)
+{
+    int status;
+    struct lock_wait lwait;
+
+    /* Wait until the lock module is ready */
+    if (lock_start_flag)
+    {
+	pthread_mutex_lock(&lock_start_mutex);
+	pthread_mutex_unlock(&lock_start_mutex);
+    }
+
+    pthread_cond_init(&lwait.cond, NULL);
+    pthread_mutex_init(&lwait.mutex, NULL);
+    pthread_mutex_lock(&lwait.mutex);
+
+    /* This needs to be converted from DLM/LVM2 value for GULM */
+    if (flags == LCK_NONBLOCK) flags = lg_lock_flag_Try;
+
+    hash_insert(lock_hash, resource, &lwait);
+    DEBUGLOG("lock_resource '%s', flags=%d, mode=%d\n", resource, flags, mode);
+
+    status = lg_lock_state_req(gulm_if, resource, strlen(resource)+1,
+			       mode, flags, NULL, 0);
+    if (status)
+    {
+	DEBUGLOG("lg_lock_state returned %d\n", status);
+	return status;
+    }
+
+    /* Wait for it to complete */
+    pthread_cond_wait(&lwait.cond, &lwait.mutex);
+    pthread_mutex_unlock(&lwait.mutex);
+
+    hash_remove(lock_hash, resource);
+    DEBUGLOG("lock-resource returning %d\n", lwait.status);
+
+    return gulm_to_errno(lwait.status);
+}
+
+
+static int _unlock_resource(char *resource, int lockid)
+{
+    int status;
+    struct lock_wait lwait;
+
+    pthread_cond_init(&lwait.cond, NULL);
+    pthread_mutex_init(&lwait.mutex, NULL);
+    pthread_mutex_lock(&lwait.mutex);
+
+    hash_insert(lock_hash, resource, &lwait);
+
+    DEBUGLOG("unlock_resource %s\n", resource);
+    status = lg_lock_state_req(gulm_if, resource, strlen(resource)+1,
+			       lg_lock_state_Unlock, 0, NULL, 0);
+
+    if (status)
+    {
+	DEBUGLOG("lg_lock_state(unlock) returned %d\n", status);
+	return status;
+    }
+
+    /* Wait for it to complete */
+
+    pthread_cond_wait(&lwait.cond, &lwait.mutex);
+    pthread_mutex_unlock(&lwait.mutex);
+
+    hash_remove(lock_hash, resource);
+
+    return gulm_to_errno(lwait.status);
+}
+
+
+/* These two locking functions MUST be called in a seperate thread from
+   the clvmd main loop because they expect to be woken up by it.
+
+   These are abstractions around the real locking functions (above)
+   as we need to emulate the DLM's EX/PW/CW interaction with GULM using
+   two locks.
+   To aid unlocking, we store the lock mode in the lockid (as GULM
+   doesn't use this).
+*/
+int sync_lock(const char *resource, int mode, int flags, int *lockid)
+{
+    int status;
+    char lock1[strlen(resource)+3];
+    char lock2[strlen(resource)+3];
+
+    snprintf(lock1, sizeof(lock1), "%s-1", resource);
+    snprintf(lock2, sizeof(lock2), "%s-2", resource);
+
+    switch (mode)
+    {
+    case LCK_EXCL:
+	status = _lock_resource(lock1, lg_lock_state_Exclusive, flags, lockid);
+	if (status)
+	    goto out;
+
+	/* If we can't get this lock then bail out */
+	status = _lock_resource(lock2, lg_lock_state_Exclusive, LCK_NONBLOCK, lockid);
+        if (status == lg_err_TryFailed)
+        {
+           _unlock_resource(lock1, *lockid);
+           status = -1;
+           errno = EAGAIN;
+        }
+	break;
+
+    case LCK_READ:
+	status = _lock_resource(lock1, lg_lock_state_Shared, flags, lockid);
+	break;
+
+    case LCK_WRITE:
+	status = _lock_resource(lock2, lg_lock_state_Exclusive, flags, lockid);
+	break;
+
+    default:
+	status = -1;
+	errno = EINVAL;
+	break;
+    }
+ out:
+    *lockid = mode;
+    return status;
+}
+
+int sync_unlock(const char *resource, int lockid)
+{
+    int status = 0;
+    char lock1[strlen(resource)+3];
+    char lock2[strlen(resource)+3];
+
+    snprintf(lock1, sizeof(lock1), "%s-1", resource);
+    snprintf(lock2, sizeof(lock2), "%s-2", resource);
+
+    /* The held lock mode is in the lock id */
+    assert(lockid == LCK_EXCL ||
+	   lockid == LCK_READ ||
+	   lockid == LCK_WRITE);
+
+    switch (lockid)
+    {
+    case LCK_EXCL:
+	status = _unlock_resource(lock1, lockid);
+	if (status)
+	    goto out;
+	status = _unlock_resource(lock2, lockid);
+	break;
+
+    case LCK_READ:
+	status = _unlock_resource(lock1, lockid);
+	break;
+
+    case LCK_WRITE:
+	status = _unlock_resource(lock2, lockid);
+	break;
+    }
+
+ out:
+    return status;
+}
+
+int is_quorate()
+{
+    if (current_corestate == lg_core_Slave ||
+	current_corestate == lg_core_Master ||
+	current_corestate == lg_core_Client)
+	return 1;
+    else
+	return 0;
+}
+
+/* Get all the cluster node names & IPs from CCS and
+   add them to our node list so we know who to talk to.
+   Called when we start up and if we get sent SIGHUP.
+*/
+static int get_all_cluster_nodes()
+{
+    int ctree;
+    char *nodename;
+    int error;
+
+    /* Open the config file */
+    ctree = ccs_connect();
+    if (ctree <= 0)
+    {
+	log_error("Error connecting to CCS");
+	return -1;
+    }
+
+    error = ccs_get(ctree, "//nodes/node/@name", &nodename);
+    while (nodename)
+    {
+	char nodeip[MAX_CSID_LEN];
+	char *clvmflag;
+	char key[256];
+
+	sprintf(key, "//nodes/node[@name=\"%s\"]/clvm", nodename);
+	ccs_get(ctree, key, &clvmflag);
+
+	if ((get_ip_address(nodename, nodeip) == 0) && atoi(clvmflag))
+	{
+	    struct node_info *ninfo;
+
+	    /* If it's not in the list, then add it */
+	    ninfo = hash_lookup_binary(node_hash, nodeip, MAX_CSID_LEN);
+	    if (!ninfo)
+	    {
+		ninfo = malloc(sizeof(struct node_info));
+		if (!ninfo)
+		{
+		    syslog(LOG_ERR, "Cannot alloc memory for node info\n");
+		    ccs_disconnect(ctree);
+		    return -1;
+		}
+		strcpy(ninfo->name, nodename);
+
+		ninfo->state = NODE_DOWN;
+		hash_insert_binary(node_hash, nodeip, MAX_CSID_LEN, ninfo);
+	    }
+	}
+	else
+	{
+	    DEBUGLOG("node %s has clvm disabled\n", nodename);
+	}
+	if (clvmflag) free(clvmflag);
+	free(nodename);
+	error = ccs_get(ctree, "//nodes/node/@name", &nodename);
+    }
+
+    /* Finished with config file */
+    ccs_disconnect(ctree);
+
+    return 0;
+}
+
+int gulm_fd(void)
+{
+    return lg_core_selector(gulm_if);
+}

daemons/clvmd/clvmd-gulm.h

@@ -0,0 +1,9 @@
+
+
+
+extern int get_next_node_csid(void **context, char *csid);
+extern void add_down_node(char *csid);
+extern int gulm_fd(void);
+extern int get_ip_address(char *node, char *addr);
+extern void tcp_remove_client(char *csid);
+extern int alloc_client(int fd, char *csid, struct local_client **new_client);

daemons/clvmd/clvmd.h

@@ -0,0 +1,119 @@
+/*
+ * Copyright (C) 2002-2004 Sistina Software, Inc. All rights reserved.
+ * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU General Public License v.2.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#ifndef _CLVMD_H
+#define _CLVMD_H
+
+#define CLVMD_MAJOR_VERSION 0
+#define CLVMD_MINOR_VERSION 2
+#define CLVMD_PATCH_VERSION 1
+
+/* Name of the cluster LVM admin lock */
+#define ADMIN_LOCK_NAME "CLVMD_ADMIN"
+
+/* Default time (in seconds) we will wait for all remote commands to execute
+   before declaring them dead */
+#define DEFAULT_CMD_TIMEOUT 60
+
+/* One of these for each reply we get from command execution on a node */
+struct node_reply {
+	char node[MAX_CLUSTER_MEMBER_NAME_LEN];
+	char *replymsg;
+	int status;
+	struct node_reply *next;
+};
+
+/*
+ * These exist for the use of local sockets only when we are
+ * collecting responses from all cluster nodes
+ */
+struct localsock_bits {
+	struct node_reply *replies;
+	int num_replies;
+	int expected_replies;
+	time_t sent_time;	/* So we can check for timeouts */
+	int in_progress;	/* Only execute one cmd at a time per client */
+	int sent_out;		/* Flag to indicate that a command was sent
+				   to remote nodes */
+	void *private;		/* Private area for command processor use */
+	void *cmd;		/* Whole command as passed down local socket */
+	int cmd_len;		/* Length of above */
+	int pipe;		/* Pipe to send PRE completion status down */
+	int finished;		/* Flag to tell subthread to exit */
+	int all_success;	/* Set to 0 if any node (or the pre_command)
+				   failed */
+	struct local_client *pipe_client;
+	pthread_t threadid;
+	enum { PRE_COMMAND, POST_COMMAND, QUIT } state;
+	pthread_mutex_t mutex;	/* Main thread and worker synchronisation */
+	pthread_cond_t cond;
+
+	pthread_mutex_t reply_mutex;	/* Protect reply structure */
+};
+
+/* Entries for PIPE clients */
+struct pipe_bits {
+	struct local_client *client;	/* Actual (localsock) client */
+	pthread_t threadid;		/* Our own copy of the thread id */
+};
+
+/* Entries for Network socket clients */
+struct netsock_bits {
+	void *private;
+	int flags;
+};
+
+typedef int (*fd_callback_t) (struct local_client * fd, char *buf, int len,
+			      char *csid, struct local_client ** new_client);
+
+/* One of these for each fd we are listening on */
+struct local_client {
+	int fd;
+	enum { CLUSTER_MAIN_SOCK, CLUSTER_DATA_SOCK, LOCAL_RENDEZVOUS,
+		    LOCAL_SOCK, THREAD_PIPE, CLUSTER_INTERNAL } type;
+	struct local_client *next;
+	unsigned short xid;
+	fd_callback_t callback;
+
+	union {
+		struct localsock_bits localsock;
+		struct pipe_bits pipe;
+		struct netsock_bits net;
+	} bits;
+};
+
+#ifdef DEBUG
+#define DEBUGLOG(fmt, args...) fprintf(stderr, "CLVMD[%d]: %ld ", getpid(), time(NULL) ); fprintf(stderr, fmt, ## args)
+#else
+#define DEBUGLOG(fmt, args...)
+#endif
+
+#ifndef max
+#define max(a,b) ((a)>(b)?(a):(b))
+#endif
+
+/* The real command processor is in clvmd-command.c */
+extern int do_command(struct local_client *client, struct clvm_header *msg,
+		      int msglen, char **buf, int buflen, int *retlen);
+
+/* Pre and post command routines are called only on the local node */
+extern int do_pre_command(struct local_client *client);
+extern int do_post_command(struct local_client *client);
+extern void cmd_client_cleanup(struct local_client *client);
+extern int add_client(struct local_client *new_client);
+
+extern void clvmd_cluster_init_completed(void);
+
+#endif

daemons/clvmd/cnxman-socket.h

@@ -0,0 +1,226 @@
+/******************************************************************************
+*******************************************************************************
+**
+**  Copyright (C) Sistina Software, Inc.  1997-2003  All rights reserved.
+**  Copyright (C) 2004 Red Hat, Inc.  All rights reserved.
+**
+**  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.
+**
+*******************************************************************************
+******************************************************************************/
+
+/* CMAN socket interface header,
+   may be include by user or kernel code */
+
+#ifndef __CNXMAN_SOCKET_H
+#define __CNXMAN_SOCKET_H
+
+/* A currently unused number. TIPC also uses this number and you're unlikely
+   to be using both.
+ */
+#define AF_CLUSTER 30
+#define PF_CLUSTER AF_CLUSTER
+
+/* Protocol(socket) types */
+#define CLPROTO_MASTER 2
+#define CLPROTO_CLIENT 3
+
+/* ioctls -- should register these properly */
+#define SIOCCLUSTER_NOTIFY            _IOW('x', 0x01, int)
+#define SIOCCLUSTER_REMOVENOTIFY      _IO( 'x', 0x02)
+#define SIOCCLUSTER_GETMEMBERS        _IOR('x', 0x03, struct cl_cluster_nodelist)
+#define SIOCCLUSTER_SETEXPECTED_VOTES _IOW('x', 0x04, int)
+#define SIOCCLUSTER_ISQUORATE         _IO( 'x', 0x05)
+#define SIOCCLUSTER_ISLISTENING       _IOW('x', 0x06, struct cl_listen_request)
+#define SIOCCLUSTER_GETALLMEMBERS     _IOR('x', 0x07, struct cl_cluster_nodelist)
+#define SIOCCLUSTER_SET_VOTES         _IOW('x', 0x08, int)
+#define SIOCCLUSTER_GET_VERSION       _IOR('x', 0x09, struct cl_version)
+#define SIOCCLUSTER_SET_VERSION       _IOW('x', 0x0a, struct cl_version)
+#define SIOCCLUSTER_ISACTIVE          _IO( 'x', 0x0b)
+#define SIOCCLUSTER_KILLNODE          _IOW('x', 0x0c, int)
+#define SIOCCLUSTER_GET_JOINCOUNT     _IO( 'x', 0x0d)
+#define SIOCCLUSTER_SERVICE_REGISTER  _IOW('x', 0x0e, char)
+#define SIOCCLUSTER_SERVICE_UNREGISTER _IO('x', 0x0f)
+#define SIOCCLUSTER_SERVICE_JOIN      _IO( 'x', 0x10)
+#define SIOCCLUSTER_SERVICE_LEAVE     _IO( 'x', 0x20)
+#define SIOCCLUSTER_SERVICE_SETSIGNAL _IOW('x', 0x30, int)
+#define SIOCCLUSTER_SERVICE_STARTDONE _IOW('x', 0x40, unsigned int)
+#define SIOCCLUSTER_SERVICE_GETEVENT  _IOR('x', 0x50, struct cl_service_event)
+#define SIOCCLUSTER_SERVICE_GETMEMBERS _IOR('x', 0x60, struct cl_cluster_nodelist)
+#define SIOCCLUSTER_SERVICE_GLOBALID  _IOR('x', 0x70, uint32_t)
+#define SIOCCLUSTER_SERVICE_SETLEVEL  _IOR('x', 0x80, int)
+#define SIOCCLUSTER_GETNODE	      _IOWR('x', 0x90, struct cl_cluster_node)
+#define SIOCCLUSTER_BARRIER           _IOW('x', 0x0a0, struct cl_barrier_info)
+
+/* These were setsockopts */
+#define SIOCCLUSTER_PASS_SOCKET       _IOW('x', 0x0b0, struct cl_passed_sock)
+#define SIOCCLUSTER_SET_NODENAME      _IOW('x', 0x0b1, char *)
+#define SIOCCLUSTER_SET_NODEID        _IOW('x', 0x0b2, int)
+#define SIOCCLUSTER_JOIN_CLUSTER      _IOW('x', 0x0b3, struct cl_join_cluster_info)
+#define SIOCCLUSTER_LEAVE_CLUSTER     _IOW('x', 0x0b4, int)
+
+
+/* Maximum size of a cluster message */
+#define MAX_CLUSTER_MESSAGE          1500
+#define MAX_CLUSTER_MEMBER_NAME_LEN   255
+#define MAX_BARRIER_NAME_LEN           33
+#define MAX_SA_ADDR_LEN                12
+#define MAX_CLUSTER_NAME_LEN           16
+
+/* Well-known cluster port numbers */
+#define CLUSTER_PORT_MEMBERSHIP  1	/* Mustn't block during cluster
+					 * transitions! */
+#define CLUSTER_PORT_SERVICES    2
+#define CLUSTER_PORT_SYSMAN      10	/* Remote execution daemon */
+#define CLUSTER_PORT_CLVMD       11	/* Cluster LVM daemon */
+#define CLUSTER_PORT_SLM         12	/* LVM SLM (simple lock manager) */
+
+/* Port numbers above this will be blocked when the cluster is inquorate or in
+ * transition */
+#define HIGH_PROTECTED_PORT      9
+
+/* Reasons for leaving the cluster */
+#define CLUSTER_LEAVEFLAG_DOWN     0	/* Normal shutdown */
+#define CLUSTER_LEAVEFLAG_KILLED   1
+#define CLUSTER_LEAVEFLAG_PANIC    2
+#define CLUSTER_LEAVEFLAG_REMOVED  3	/* This one can reduce quorum */
+#define CLUSTER_LEAVEFLAG_REJECTED 4	/* Not allowed into the cluster in the
+					 * first place */
+#define CLUSTER_LEAVEFLAG_INCONSISTENT 5	/* Our view of the cluster is
+						 * in a minority */
+#define CLUSTER_LEAVEFLAG_DEAD         6	/* Discovered to be dead */
+#define CLUSTER_LEAVEFLAG_FORCE     0x10	/* Forced by command-line */
+
+/* OOB messages sent to a local socket */
+#define CLUSTER_OOB_MSG_PORTCLOSED  1
+#define CLUSTER_OOB_MSG_STATECHANGE 2
+#define CLUSTER_OOB_MSG_SERVICEEVENT 3
+
+/* Sendmsg flags, these are above the normal sendmsg flags so they don't
+ * interfere */
+#define MSG_NOACK     0x010000	/* Don't need an ACK for this message */
+#define MSG_QUEUE     0x020000	/* Queue the message for sending later */
+#define MSG_MULTICAST 0x080000	/* Message was sent to all nodes in the cluster
+				 */
+#define MSG_ALLINT    0x100000	/* Send out of all interfaces */
+#define MSG_REPLYEXP  0x200000	/* Reply is expected */
+
+typedef enum { NODESTATE_REMOTEMEMBER, NODESTATE_JOINING, NODESTATE_MEMBER,
+	       NODESTATE_DEAD } nodestate_t;
+
+
+struct sockaddr_cl {
+	unsigned short scl_family;
+	unsigned char scl_flags;
+	unsigned char scl_port;
+	int           scl_nodeid;
+};
+
+/*
+ * This is how we pass the multicast & receive sockets into kernel space.
+ */
+struct cl_passed_sock {
+	int fd;			/* FD of master socket to do multicast on */
+	int number;		/* Socket number, to match up recvonly & bcast
+				 * sockets */
+        int multicast;          /* Is it multicast or receive ? */
+};
+
+/* Cluster configuration info passed when we join the cluster */
+struct cl_join_cluster_info {
+	unsigned char votes;
+	unsigned int expected_votes;
+	unsigned int two_node;
+	unsigned int config_version;
+
+        char cluster_name[17];
+};
+
+
+/* This is the structure, per node, returned from the membership ioctl */
+struct cl_cluster_node {
+	unsigned int size;
+	unsigned int node_id;
+	unsigned int us;
+	unsigned int leave_reason;
+	unsigned int incarnation;
+	nodestate_t state;
+	char name[MAX_CLUSTER_MEMBER_NAME_LEN];
+	unsigned char votes;
+};
+
+/* The struct passed to the membership ioctls */
+struct cl_cluster_nodelist {
+        uint32_t max_members;
+        struct cl_cluster_node *nodes;
+};
+
+/* Structure passed to SIOCCLUSTER_ISLISTENING */
+struct cl_listen_request {
+	unsigned char port;
+        int           nodeid;
+};
+
+/* A Cluster PORTCLOSED message - received by a local user as an OOB message */
+struct cl_portclosed_oob {
+	unsigned char cmd;	/* CLUSTER_OOB_MSG_PORTCLOSED */
+	unsigned char port;
+};
+
+/* Get all version numbers or set the config version */
+struct cl_version {
+	unsigned int major;
+	unsigned int minor;
+	unsigned int patch;
+	unsigned int config;
+};
+
+/* structure passed to barrier ioctls */
+struct cl_barrier_info {
+	char cmd;
+	char name[MAX_BARRIER_NAME_LEN];
+	unsigned int flags;
+	unsigned long arg;
+};
+
+typedef enum { SERVICE_EVENT_STOP, SERVICE_EVENT_START, SERVICE_EVENT_FINISH,
+		SERVICE_EVENT_LEAVEDONE } service_event_t;
+
+typedef enum { SERVICE_START_FAILED, SERVICE_START_JOIN, SERVICE_START_LEAVE }
+		service_start_t;
+
+struct cl_service_event {
+	service_event_t type;
+	service_start_t start_type;
+	unsigned int event_id;
+	unsigned int last_stop;
+	unsigned int last_start;
+	unsigned int last_finish;
+	unsigned int node_count;
+};
+
+
+/* Commands to the barrier ioctl */
+#define BARRIER_IOCTL_REGISTER 1
+#define BARRIER_IOCTL_CHANGE   2
+#define BARRIER_IOCTL_DELETE   3
+#define BARRIER_IOCTL_WAIT     4
+
+/* Attributes of a barrier - bitmask */
+#define BARRIER_ATTR_AUTODELETE 1
+#define BARRIER_ATTR_MULTISTEP  2
+#define BARRIER_ATTR_MANUAL     4
+#define BARRIER_ATTR_ENABLED    8
+#define BARRIER_ATTR_CALLBACK  16
+
+/* Attribute setting commands */
+#define BARRIER_SETATTR_AUTODELETE 1
+#define BARRIER_SETATTR_MULTISTEP  2
+#define BARRIER_SETATTR_ENABLED    3
+#define BARRIER_SETATTR_NODES      4
+#define BARRIER_SETATTR_CALLBACK   5
+#define BARRIER_SETATTR_TIMEOUT    6
+
+#endif

daemons/clvmd/libclvm.c

@@ -0,0 +1,446 @@
+/*
+ * Copyright (C) 1997-2004 Sistina Software, Inc. All rights reserved.
+ * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU General Public License v.2.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+/* library functions for Cluster LVM Daemon */
+
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <sys/socket.h>
+#include <sys/uio.h>
+#include <sys/un.h>
+#include <sys/time.h>
+#include <sys/ioctl.h>
+#include <sys/utsname.h>
+#include <syslog.h>
+#include <netinet/in.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <signal.h>
+#include <unistd.h>
+#include <fcntl.h>
+#include <search.h>
+#include <errno.h>
+
+#include "clvm.h"
+#include "libclvm.h"
+
+/* CLVM in hex! */
+#define LVM_SIGNATURE 0x434C564D
+
+#define MAX_CLUSTER_MEMBER_NAME_LEN 255
+
+/* NOTE: the LVMD uses the socket FD as the client ID, this means
+   that any client that calls fork() will inherit the context of
+   it's parent. */
+static int clvmd_sock = -1;
+
+static int open_local_sock(void)
+{
+	int local_socket;
+	struct sockaddr_un sockaddr;
+
+	/* Open local socket */
+	local_socket = socket(PF_UNIX, SOCK_STREAM, 0);
+	if (local_socket < 0) {
+		perror("Can't create local socket");
+		return -1;
+	}
+
+	fcntl(local_socket, F_SETFD, !FD_CLOEXEC);
+
+	strcpy(sockaddr.sun_path, CLVMD_SOCKNAME);
+	sockaddr.sun_family = AF_UNIX;
+	if (connect
+	    (local_socket, (struct sockaddr *) &sockaddr, sizeof(sockaddr))) {
+		int saved_errno = errno;
+
+		close(local_socket);
+
+		errno = saved_errno;
+		return -1;
+	}
+	return local_socket;
+}
+
+/* Send a request and return the status */
+static int send_request(char *inbuf, int inlen, char **retbuf)
+{
+	char outbuf[PIPE_BUF];
+	struct clvm_header *outheader = (struct clvm_header *) outbuf;
+	int len;
+	int off;
+	fd_set fds;
+
+	FD_ZERO(&fds);
+	FD_SET(clvmd_sock, &fds);
+
+	/* Send it to CLVMD */
+	if (write(clvmd_sock, inbuf, inlen) != inlen) {
+		perror("Error writing to CLVMD");
+		return -1;
+	}
+
+	/* Get the response */
+	if ((len = read(clvmd_sock, outbuf, sizeof(struct clvm_header))) < 0) {
+		perror("Error reading CLVMD");
+		return -1;
+	}
+	if (len == 0) {
+		fprintf(stderr, "EOF reading CLVMD");
+		errno = ENOTCONN;
+		return -1;
+	}
+
+	/* Allocate buffer */
+	*retbuf = malloc(len + outheader->arglen);
+	if (!*retbuf) {
+		errno = ENOMEM;
+		return -1;
+	}
+
+	/* Copy the header */
+	memcpy(*retbuf, outbuf, len);
+	outheader = (struct clvm_header *) *retbuf;
+
+	/* Read the returned values */
+	off = 1;		/* we've already read the first byte */
+
+	while (off < outheader->arglen && len > 0) {
+		len = read(clvmd_sock, outheader->args + off, PIPE_BUF);
+		if (len > 0)
+			off += len;
+	}
+
+	/* Was it an error ? */
+	if (outheader->status < 0) {
+		errno = -outheader->status;
+		return -2;
+	}
+	return 0;
+}
+
+/* Build the structure header and parse-out wildcard node names */
+static void build_header(struct clvm_header *head, int cmd, const char *node,
+			 void *data, int len)
+{
+	head->cmd = cmd;
+	head->status = 0;
+	head->flags = 0;
+	head->clientid = 0;
+	head->arglen = len;
+	if (node) {
+		/* Allow a couple of special node names:
+		   "*" for all nodes,
+		   "." for the local node only
+		 */
+		if (strcmp(node, "*") == 0) {
+			head->node[0] = '\0';
+		} else if (strcmp(node, ".") == 0) {
+			head->node[0] = '\0';
+			head->flags = CLVMD_FLAG_LOCAL;
+		} else {
+			strcpy(head->node, node);
+		}
+	} else {
+		head->node[0] = '\0';
+	}
+}
+
+/* Send a message to a(or all) node(s) in the cluster */
+int lvm_cluster_write(char cmd, char *node, void *data, int len)
+{
+	char outbuf[sizeof(struct clvm_header) + len + strlen(node) + 1];
+	char *retbuf = NULL;
+	int status;
+	struct clvm_header *head = (struct clvm_header *) outbuf;
+
+	if (clvmd_sock == -1)
+		clvmd_sock = open_local_sock();
+	if (clvmd_sock == -1)
+		return -1;
+
+	build_header(head, cmd, node, data, len);
+	memcpy(head->node + strlen(head->node) + 1, data, len);
+
+	status =
+	    send_request(outbuf,
+			 sizeof(struct clvm_header) + strlen(head->node) + len,
+			 &retbuf);
+	if (retbuf)
+		free(retbuf);
+
+	return status;
+}
+
+/* API: Send a message to a(or all) node(s) in the cluster
+   and wait for replies */
+int lvm_cluster_request(char cmd, const char *node, void *data, int len,
+			lvm_response_t ** response, int *num)
+{
+	char outbuf[sizeof(struct clvm_header) + len + strlen(node) + 1];
+	int *outptr;
+	char *inptr;
+	char *retbuf = NULL;
+	int status;
+	int i;
+	int num_responses = 0;
+	struct clvm_header *head = (struct clvm_header *) outbuf;
+	lvm_response_t *rarray;
+
+	*num = 0;
+
+	if (clvmd_sock == -1)
+		clvmd_sock = open_local_sock();
+	if (clvmd_sock == -1)
+		return -1;
+
+	build_header(head, cmd, node, data, len);
+	memcpy(head->node + strlen(head->node) + 1, data, len);
+
+	status =
+	    send_request(outbuf,
+			 sizeof(struct clvm_header) + strlen(head->node) + len,
+			 &retbuf);
+	if (status == 0 || status == -2) {
+		/* Count the number of responses we got */
+		head = (struct clvm_header *) retbuf;
+		inptr = head->args;
+		while (inptr[0]) {
+			num_responses++;
+			inptr += strlen(inptr) + 1;
+			inptr += sizeof(int);
+			inptr += strlen(inptr) + 1;
+		}
+
+		/* Allocate response array. With an extra pair of INTs on the front to sanity
+		   check the pointer when we are given it back to free */
+		outptr =
+		    malloc(sizeof(lvm_response_t) * num_responses +
+			   sizeof(int) * 2);
+		if (!outptr) {
+			if (retbuf)
+				free(retbuf);
+			errno = ENOMEM;
+			return -1;
+		}
+
+		*response = (lvm_response_t *) (outptr + 2);
+		outptr[0] = LVM_SIGNATURE;
+		outptr[1] = num_responses;
+		rarray = *response;
+
+		/* Unpack the response into an lvm_response_t array */
+		inptr = head->args;
+		i = 0;
+		while (inptr[0]) {
+			strcpy(rarray[i].node, inptr);
+			inptr += strlen(inptr) + 1;
+
+			rarray[i].status = *(int *) inptr;
+			inptr += sizeof(int);
+
+			rarray[i].response = malloc(strlen(inptr) + 1);
+			if (rarray[i].response == NULL) {
+				/* Free up everything else and return error */
+				int j;
+				for (j = 0; j < i; j++)
+					free(rarray[i].response);
+				free(outptr);
+				errno = ENOMEM;
+				return -1;
+			}
+
+			strcpy(rarray[i].response, inptr);
+			rarray[i].len = strlen(inptr);
+			inptr += strlen(inptr) + 1;
+			i++;
+		}
+		*num = num_responses;
+		*response = rarray;
+	}
+
+	if (retbuf)
+		free(retbuf);
+	return status;
+}
+
+/* API: Free reply array */
+int lvm_cluster_free_request(lvm_response_t * response)
+{
+	int *ptr = (int *) response - 2;
+	int i;
+	int num;
+
+	/* Check it's ours to free */
+	if (response == NULL || *ptr != LVM_SIGNATURE) {
+		errno = EINVAL;
+		return -1;
+	}
+
+	num = ptr[1];
+	for (i = 0; i < num; i++) {
+		free(response[i].response);
+	}
+	free(ptr);
+
+	return 0;
+}
+
+/* These are a "higher-level" API providing black-box lock/unlock
+   functions for cluster LVM...maybe */
+
+/* Set by lock(), used by unlock() */
+static int num_responses;
+static lvm_response_t *response;
+
+int lvm_lock_for_cluster(char scope, char *name, int verbosity)
+{
+	int status;
+	int i;
+	char *args;
+	int len;
+
+	if (name) {
+		len = strlen(name) + 2;
+		args = alloca(len);
+		strcpy(args + 1, name);
+	} else {
+		len = 2;
+		args = alloca(len);
+		args[1] = '\0';
+	}
+	args[0] = scope;
+
+	status = lvm_cluster_request(CLVMD_CMD_LOCK,
+				     "", args, len, &response, &num_responses);
+
+	/* If any nodes were down then display them and return an error */
+	for (i = 0; i < num_responses; i++) {
+		if (response[i].status == -EHOSTDOWN) {
+			if (verbosity)
+				fprintf(stderr,
+					"clvmd not running on node %s\n",
+					response[i].node);
+			status = -1;
+		}
+	}
+
+	/* If there was an error then free the memory now as the caller won't
+	   want to do the unlock */
+	if (status) {
+		int saved_errno = errno;
+		lvm_cluster_free_request(response);
+		num_responses = 0;
+		errno = saved_errno;
+	}
+	return status;
+}
+
+int lvm_unlock_for_cluster(char scope, char *name, int verbosity)
+{
+	int status;
+	int i;
+	int len;
+	int failed;
+	int num_unlock_responses;
+	char *args;
+	lvm_response_t *unlock_response;
+
+	/* We failed - this should not have been called */
+	if (num_responses == 0)
+		return 0;
+
+	if (name) {
+		len = strlen(name) + 2;
+		args = alloca(len);
+		strcpy(args + 1, name);
+	} else {
+		len = 2;
+		args = alloca(len);
+		args[1] = '\0';
+	}
+	args[0] = scope;
+
+	/* See if it failed anywhere */
+	failed = 0;
+	for (i = 0; i < num_responses; i++) {
+		if (response[i].status != 0)
+			failed++;
+	}
+
+	/* If it failed on any nodes then we only unlock on
+	   the nodes that succeeded */
+	if (failed) {
+		for (i = 0; i < num_responses; i++) {
+			/* Unlock the ones that succeeded */
+			if (response[i].status == 0) {
+				status = lvm_cluster_request(CLVMD_CMD_UNLOCK,
+							     response[i].node,
+							     args, len,
+							     &unlock_response,
+							     &num_unlock_responses);
+				if (status) {
+					if (verbosity)
+						fprintf(stderr,
+							"cluster command to node %s failed: %s\n",
+							response[i].node,
+							strerror(errno));
+				} else if (unlock_response[0].status != 0) {
+					if (verbosity > 1)
+						fprintf(stderr,
+							"unlock on node %s failed: %s\n",
+							response[i].node,
+							strerror(unlock_response
+								 [0].status));
+				}
+				lvm_cluster_free_request(unlock_response);
+			} else {
+				if (verbosity)
+					fprintf(stderr,
+						"command on node %s failed: '%s' - will be left locked\n",
+						response[i].node,
+						strerror(response[i].status));
+			}
+		}
+	} else {
+		/* All OK, we can do a full cluster unlock */
+		status = lvm_cluster_request(CLVMD_CMD_UNLOCK,
+					     "",
+					     args, len,
+					     &unlock_response,
+					     &num_unlock_responses);
+		if (status) {
+			if (verbosity > 1)
+				fprintf(stderr, "cluster command failed: %s\n",
+					strerror(errno));
+		} else {
+			for (i = 0; i < num_unlock_responses; i++) {
+				if (unlock_response[i].status != 0) {
+					if (verbosity > 1)
+						fprintf(stderr,
+							"unlock on node %s failed: %s\n",
+							response[i].node,
+							strerror(unlock_response
+								 [0].status));
+				}
+			}
+		}
+		lvm_cluster_free_request(unlock_response);
+	}
+	lvm_cluster_free_request(response);
+
+	return 0;
+}

daemons/clvmd/libclvm.h

@@ -0,0 +1,36 @@
+/*
+ * Copyright (C) 1997-2004 Sistina Software, Inc. All rights reserved.
+ * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU General Public License v.2.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#ifndef _LIBCLVM_H
+#define _LIBCLVM_H
+
+typedef struct lvm_response {
+	char node[255];
+	char *response;
+	int status;
+	int len;
+
+} lvm_response_t;
+
+extern int lvm_cluster_request(char cmd, const char *node, void *data, int len,
+			       lvm_response_t ** response, int *num);
+extern int lvm_cluster_write(char cmd, char *node, void *data, int len);
+extern int lvm_cluster_free_request(lvm_response_t * response);
+
+/* The "high-level" API */
+extern int lvm_lock_for_cluster(char scope, char *name, int verbosity);
+extern int lvm_unlock_for_cluster(char scope, char *name, int verbosity);
+
+#endif

daemons/clvmd/lvm-functions.c

@@ -0,0 +1,461 @@
+/*
+ * Copyright (C) 2002-2004 Sistina Software, Inc. All rights reserved.
+ * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU General Public License v.2.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include <pthread.h>
+#include <sys/types.h>
+#include <sys/utsname.h>
+#include <sys/ioctl.h>
+#include <sys/socket.h>
+#include <sys/stat.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdint.h>
+#include <fcntl.h>
+#include <string.h>
+#include <stddef.h>
+#include <stdint.h>
+#include <unistd.h>
+#include <errno.h>
+#include <syslog.h>
+#include <assert.h>
+
+#include "libdlm.h"
+#include "clvm.h"
+#include "clvmd-comms.h"
+#include "clvmd.h"
+#include "lvm-functions.h"
+
+/* LVM2 headers */
+#include "toolcontext.h"
+#include "log.h"
+#include "activate.h"
+#include "hash.h"
+#include "locking.h"
+
+static struct cmd_context *cmd = NULL;
+static struct hash_table *lv_hash = NULL;
+static pthread_mutex_t lv_hash_lock;
+
+struct lv_info {
+	int lock_id;
+	int lock_mode;
+};
+
+/* Return the mode a lock is currently held at (or -1 if not held) */
+static int get_current_lock(char *resource)
+{
+	struct lv_info *lvi;
+
+	pthread_mutex_lock(&lv_hash_lock);
+	lvi = hash_lookup(lv_hash, resource);
+	pthread_mutex_unlock(&lv_hash_lock);
+	if (lvi) {
+		return lvi->lock_mode;
+	} else {
+		return -1;
+	}
+}
+
+/* Called at shutdown to tidy the lockspace */
+void unlock_all()
+{
+	struct hash_node *v;
+
+	pthread_mutex_lock(&lv_hash_lock);
+	hash_iterate(v, lv_hash) {
+		struct lv_info *lvi = hash_get_data(lv_hash, v);
+
+		sync_unlock(hash_get_key(lv_hash, v), lvi->lock_id);
+	}
+	pthread_mutex_unlock(&lv_hash_lock);
+}
+
+/* Gets a real lock and keeps the info in the hash table */
+int hold_lock(char *resource, int mode, int flags)
+{
+	int status;
+	int saved_errno;
+	struct lv_info *lvi;
+
+	flags &= LKF_NOQUEUE;	/* Only LKF_NOQUEUE is valid here */
+
+	pthread_mutex_lock(&lv_hash_lock);
+	lvi = hash_lookup(lv_hash, resource);
+	pthread_mutex_unlock(&lv_hash_lock);
+	if (lvi) {
+		/* Already exists - convert it */
+		status =
+		    sync_lock(resource, mode, LKF_CONVERT | flags,
+			      &lvi->lock_id);
+		saved_errno = errno;
+		if (!status)
+			lvi->lock_mode = mode;
+
+		if (status) {
+			DEBUGLOG("hold_lock. convert to %d failed: %s\n", mode,
+				 strerror(errno));
+		}
+		errno = saved_errno;
+	} else {
+		lvi = malloc(sizeof(struct lv_info));
+		if (!lvi)
+			return -1;
+
+		lvi->lock_mode = mode;
+		status = sync_lock(resource, mode, flags, &lvi->lock_id);
+		saved_errno = errno;
+		if (status) {
+			free(lvi);
+			DEBUGLOG("hold_lock. lock at %d failed: %s\n", mode,
+				 strerror(errno));
+		} else {
+		        pthread_mutex_lock(&lv_hash_lock);
+			hash_insert(lv_hash, resource, lvi);
+			pthread_mutex_unlock(&lv_hash_lock);
+		}
+		errno = saved_errno;
+	}
+	return status;
+}
+
+/* Unlock and remove it from the hash table */
+int hold_unlock(char *resource)
+{
+	struct lv_info *lvi;
+	int status;
+	int saved_errno;
+
+	pthread_mutex_lock(&lv_hash_lock);
+	lvi = hash_lookup(lv_hash, resource);
+	pthread_mutex_unlock(&lv_hash_lock);
+	if (!lvi) {
+		DEBUGLOG("hold_unlock, lock not already held\n");
+		return 0;
+	}
+
+	status = sync_unlock(resource, lvi->lock_id);
+	saved_errno = errno;
+	if (!status) {
+	    	pthread_mutex_lock(&lv_hash_lock);
+		hash_remove(lv_hash, resource);
+		pthread_mutex_unlock(&lv_hash_lock);
+		free(lvi);
+	} else {
+		DEBUGLOG("hold_unlock. unlock failed(%d): %s\n", status,
+			 strerror(errno));
+	}
+
+	errno = saved_errno;
+	return status;
+}
+
+/* Watch the return codes here.
+   liblvm API functions return 1(true) for success, 0(false) for failure and don't set errno.
+   libdlm API functions return 0 for success, -1 for failure and do set errno.
+   These functions here return 0 for success or >0 for failure (where the retcode is errno)
+*/
+
+/* Activate LV exclusive or non-exclusive */
+static int do_activate_lv(char *resource, int mode)
+{
+	int oldmode;
+	int status;
+	int activate_lv;
+	struct lvinfo lvi;
+
+	/* Is it already open ? */
+	oldmode = get_current_lock(resource);
+	if (oldmode == mode) {
+		return 0;	/* Nothing to do */
+	}
+
+	/* Does the config file want us to activate this LV ? */
+	if (!lv_activation_filter(cmd, resource, &activate_lv))
+		return EIO;
+
+	if (!activate_lv)
+		return 0;	/* Success, we did nothing! */
+
+	/* Do we need to activate exclusively? */
+	if (activate_lv == 2)
+		mode = LKM_EXMODE;
+
+	/* OK, try to get the lock */
+	status = hold_lock(resource, mode, LKF_NOQUEUE);
+	if (status)
+		return errno;
+
+	/* If it's suspended then resume it */
+	if (!lv_info_by_lvid(cmd, resource, &lvi))
+		return EIO;
+
+	if (lvi.suspended)
+		if (!lv_resume(cmd, resource))
+			return EIO;
+
+	/* Now activate it */
+	if (!lv_activate(cmd, resource))
+		return EIO;
+
+	return 0;
+}
+
+/* Resume the LV if it was active */
+static int do_resume_lv(char *resource)
+{
+	int oldmode;
+
+	/* Is it open ? */
+	oldmode = get_current_lock(resource);
+	if (oldmode == -1) {
+		DEBUGLOG("do_deactivate_lock, lock not already held\n");
+		return 0;	/* We don't need to do anything */
+	}
+
+	if (!lv_resume_if_active(cmd, resource))
+		return EIO;
+
+	return 0;
+}
+
+/* Suspend the device if active */
+static int do_suspend_lv(char *resource)
+{
+	int oldmode;
+	struct lvinfo lvi;
+
+	/* Is it open ? */
+	oldmode = get_current_lock(resource);
+	if (oldmode == -1) {
+		DEBUGLOG("do_suspend_lv, lock held at %d\n", oldmode);
+		return 0; /* Not active, so it's OK */
+	}
+
+	/* Only suspend it if it exists */
+	if (!lv_info_by_lvid(cmd, resource, &lvi))
+		return EIO;
+
+	if (lvi.exists) {
+		if (!lv_suspend_if_active(cmd, resource)) {
+			return EIO;
+		}
+	}
+	return 0;
+}
+
+static int do_deactivate_lv(char *resource)
+{
+	int oldmode;
+	int status;
+
+	/* Is it open ? */
+	oldmode = get_current_lock(resource);
+	if (oldmode == -1) {
+		DEBUGLOG("do_deactivate_lock, lock not already held\n");
+		return 0;	/* We don't need to do anything */
+	}
+
+	if (!lv_deactivate(cmd, resource))
+		return EIO;
+
+	status = hold_unlock(resource);
+	if (status)
+		return errno;
+
+	return 0;
+}
+
+/* This is the LOCK_LV part that happens on all nodes in the cluster -
+   it is responsible for the interaction with device-mapper and LVM */
+int do_lock_lv(unsigned char command, unsigned char lock_flags, char *resource)
+{
+	int status = 0;
+
+	DEBUGLOG("do_lock_lv: resource '%s', cmd = 0x%x, flags = %d\n",
+		 resource, command, lock_flags);
+
+	if (!cmd->config_valid || config_files_changed(cmd)) {
+		/* Reinitialise various settings inc. logging, filters */
+		if (!refresh_toolcontext(cmd)) {
+			log_error("Updated config file invalid. Aborting.");
+			return EINVAL;
+		}
+	}
+
+	switch (command) {
+	case LCK_LV_EXCLUSIVE:
+		status = do_activate_lv(resource, LKM_EXMODE);
+		break;
+
+	case LCK_LV_SUSPEND:
+		status = do_suspend_lv(resource);
+		break;
+
+	case LCK_UNLOCK:
+	case LCK_LV_RESUME:	/* if active */
+		status = do_resume_lv(resource);
+		break;
+
+	case LCK_LV_ACTIVATE:
+		status = do_activate_lv(resource, LKM_CRMODE);
+		break;
+
+	case LCK_LV_DEACTIVATE:
+		status = do_deactivate_lv(resource);
+		break;
+
+	default:
+		DEBUGLOG("Invalid LV command 0x%x\n", command);
+		status = EINVAL;
+		break;
+	}
+
+	/* clean the pool for another command */
+	pool_empty(cmd->mem);
+
+	DEBUGLOG("Command return is %d\n", status);
+	return status;
+}
+
+/* Functions to do on the local node only BEFORE the cluster-wide stuff above happens */
+int pre_lock_lv(unsigned char command, unsigned char lock_flags, char *resource)
+{
+	/* Nearly all the stuff happens cluster-wide. Apart from SUSPEND. Here we get the
+	   lock out on this node (because we are the node modifying the metadata)
+	   before suspending cluster-wide.
+	 */
+	if (command == LCK_LV_SUSPEND) {
+		DEBUGLOG("pre_lock_lv: resource '%s', cmd = 0x%x, flags = %d\n",
+			 resource, command, lock_flags);
+
+		if (hold_lock(resource, LKM_PWMODE, LKF_NOQUEUE))
+			return errno;
+	}
+	return 0;
+}
+
+/* Functions to do on the local node only AFTER the cluster-wide stuff above happens */
+int post_lock_lv(unsigned char command, unsigned char lock_flags,
+		 char *resource)
+{
+	/* Opposite of above, done on resume after a metadata update */
+	if (command == LCK_LV_RESUME) {
+		int oldmode;
+
+		DEBUGLOG
+		    ("post_lock_lv: resource '%s', cmd = 0x%x, flags = %d\n",
+		     resource, command, lock_flags);
+
+		/* If the lock state is PW then restore it to what it was */
+		oldmode = get_current_lock(resource);
+		if (oldmode == LKM_PWMODE) {
+			struct lvinfo lvi;
+
+			if (!lv_info_by_lvid(cmd, resource, &lvi))
+				return EIO;
+
+			if (lvi.exists) {
+				if (hold_lock(resource, LKM_CRMODE, 0))
+					return errno;
+			} else {
+				if (hold_unlock(resource))
+					return errno;
+			}
+		}
+	}
+	return 0;
+}
+
+/* Check if a VG is un use by LVM1 so we don't stomp on it */
+int do_check_lvm1(char *vgname)
+{
+	int status;
+
+	status = check_lvm1_vg_inactive(cmd, vgname);
+
+	return status == 1 ? 0 : EBUSY;
+}
+
+/*
+ * Ideally, clvmd should be started before any LVs are active
+ * but this may not be the case...
+ * I suppose this also comes in handy if clvmd crashes, not that it would!
+ */
+static void *get_initial_state()
+{
+	char lv[64], vg[64], flags[25];
+	char uuid[65];
+	char line[255];
+	FILE *lvs =
+	    popen
+	    ("/sbin/lvm lvs --nolocking --noheadings -o vg_uuid,lv_uuid,lv_attr",
+	     "r");
+
+	if (!lvs)
+		return NULL;
+
+	while (fgets(line, sizeof(line), lvs)) {
+		if (sscanf(line, "%s %s %s\n", vg, lv, flags) == 3) {
+			/* States: s:suspended a:active S:dropped snapshot I:invalid snapshot */
+			if (flags[4] == 'a' || flags[4] == 's') {	/* is it active or suspended? */
+				/* Convert hyphen-separated UUIDs into one */
+				memcpy(&uuid[0], &vg[0], 6);
+				memcpy(&uuid[6], &vg[7], 4);
+				memcpy(&uuid[10], &vg[12], 4);
+				memcpy(&uuid[14], &vg[17], 4);
+				memcpy(&uuid[18], &vg[22], 4);
+				memcpy(&uuid[22], &vg[27], 4);
+				memcpy(&uuid[26], &vg[32], 6);
+				memcpy(&uuid[32], &lv[0], 6);
+				memcpy(&uuid[38], &lv[7], 4);
+				memcpy(&uuid[42], &lv[12], 4);
+				memcpy(&uuid[46], &lv[17], 4);
+				memcpy(&uuid[50], &lv[22], 4);
+				memcpy(&uuid[54], &lv[27], 4);
+				memcpy(&uuid[58], &lv[32], 6);
+				uuid[64] = '\0';
+
+				DEBUGLOG("getting initial lock for %s\n", uuid);
+				hold_lock(uuid, LKM_CRMODE, LKF_NOQUEUE);
+			}
+		}
+	}
+	fclose(lvs);
+	return NULL;
+}
+
+void init_lvhash()
+{
+	/* Create hash table for keeping LV locks & status */
+	lv_hash = hash_create(100);
+	pthread_mutex_init(&lv_hash_lock, NULL);
+}
+
+/* Called to initialise the LVM context of the daemon */
+int init_lvm(void)
+{
+	if (!(cmd = create_toolcontext(NULL))) {
+		log_error("Failed to allocate command context");
+		return 0;
+	}
+
+	/* Use LOG_DAEMON for syslog messages instead of LOG_USER */
+	init_syslog(LOG_DAEMON);
+	init_debug(_LOG_ERR);
+
+	get_initial_state();
+
+	return 1;
+}

daemons/clvmd/lvm-functions.h

@@ -0,0 +1,35 @@
+/*
+ * Copyright (C) 2002-2004 Sistina Software, Inc. All rights reserved.
+ * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU General Public License v.2.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+/* Functions in lvm-functions.c */
+
+#ifndef _LVM_FUNCTIONS_H
+#define _LVM_FUNCTIONS_H
+
+extern int pre_lock_lv(unsigned char lock_cmd, unsigned char lock_flags,
+		       char *resource);
+extern int do_lock_lv(unsigned char lock_cmd, unsigned char lock_flags,
+		      char *resource);
+extern int post_lock_lv(unsigned char lock_cmd, unsigned char lock_flags,
+			char *resource);
+extern int do_check_lvm1(char *vgname);
+extern int init_lvm(void);
+extern void init_lvhash(void);
+
+extern int hold_unlock(char *resource);
+extern int hold_lock(char *resource, int mode, int flags);
+extern void unlock_all(void);
+
+#endif

daemons/clvmd/system-lv.c

@@ -0,0 +1,369 @@
+/*
+ * Copyright (C) 2002-2004 Sistina Software, Inc. All rights reserved.
+ * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU General Public License v.2.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+/* Routines dealing with the System LV */
+
+#include <pthread.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <sys/socket.h>
+#include <sys/uio.h>
+#include <sys/un.h>
+#include <sys/time.h>
+#include <sys/ioctl.h>
+#include <sys/mount.h>
+#include <sys/utsname.h>
+#include <syslog.h>
+#include <netinet/in.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <signal.h>
+#include <unistd.h>
+#include <fcntl.h>
+#include <dirent.h>
+#include <errno.h>
+#include <mntent.h>
+
+#include "libdlm.h"
+#include "log.h"
+#include "list.h"
+#include "locking.h"
+#include "system-lv.h"
+#include "clvmd-comms.h"
+#ifdef HAVE_CCS
+#include "ccs.h"
+#endif
+
+#define SYSTEM_LV_FILESYSTEM "ext2"
+#define SYSTEM_LV_MOUNTPOINT "/tmp/.clvmd-XXXXXX"
+
+extern char *config_filename(void);
+
+static char system_lv_name[PATH_MAX] = { '\0' };
+static char mount_point[PATH_MAX] = { '\0' };
+static int mounted = 0;
+static int mounted_rw = 0;
+static int lockid;
+static const char *lock_name = "CLVM_SYSTEM_LV";
+
+/* Look in /proc/mounts or (as a last resort) /etc/mtab to
+   see if the system-lv is mounted. If it is mounted and we
+   think it's not then abort because we don't have the right
+   lock status and we don't know what other processes are doing with it.
+
+   Returns 1 for mounted, 0 for not mounted so it matches the condition
+   of the "mounted" static variable above.
+*/
+static int is_really_mounted(void)
+{
+	FILE *mountfile;
+	struct mntent *ment;
+
+	mountfile = setmntent("/proc/mounts", "r");
+	if (!mountfile) {
+		mountfile = setmntent("/etc/mtab", "r");
+		if (!mountfile) {
+			log_error("Unable to open /proc/mounts or /etc/mtab");
+			return -1;
+		}
+	}
+
+	/* Look for system LV name in the file */
+	do {
+		ment = getmntent(mountfile);
+		if (ment) {
+			if (strcmp(ment->mnt_fsname, system_lv_name) == 0) {
+				endmntent(mountfile);
+				return 1;
+			}
+		}
+	}
+	while (ment);
+
+	endmntent(mountfile);
+	return 0;
+}
+
+/* Get the system LV name from the config file */
+static int find_system_lv(void)
+{
+	if (system_lv_name[0] == '\0') {
+#ifdef HAVE_CCS
+		int error;
+		ccs_node_t *ctree;
+
+		/* Read the cluster config file */
+		/* Open the config file */
+		error = open_ccs_file(&ctree, "clvm.ccs");
+		if (error) {
+			perror("reading config file");
+			return -1;
+		}
+
+		strcpy(system_lv_name, find_ccs_str(ctree,
+						    "cluster/systemlv", '/',
+						    "/dev/vg/system_lv"));
+
+		/* Finished with config file */
+		close_ccs_file(ctree);
+#else
+		if (getenv("CLVMD_SYSTEM_LV"))
+			strcpy(system_lv_name, getenv("CLVMD_SYSTEM_LV"));
+		else
+			return -1;
+#endif
+	}
+
+	/* See if it has been mounted outside our control */
+	if (is_really_mounted() != mounted) {
+		log_error
+		    ("The system LV state has been mounted/umounted outside the control of clvmd\n"
+		     "it cannot not be used for cluster communications until this is fixed.\n");
+		return -1;
+	}
+	return 0;
+}
+
+/* No prizes */
+int system_lv_umount(void)
+{
+	if (!mounted)
+		return 0;
+
+	if (umount(mount_point) < 0) {
+		log_error("umount of system LV (%s) failed: %m\n",
+			  system_lv_name);
+		return -1;
+	}
+
+	sync_unlock(lock_name, lockid);
+	mounted = 0;
+
+	/* Remove the mount point */
+	rmdir(mount_point);
+
+	return 0;
+}
+
+int system_lv_mount(int readwrite)
+{
+	int status;
+	int saved_errno;
+	int fd;
+
+	if (find_system_lv()) {
+		errno = EBUSY;
+		return -1;
+	}
+
+	/* Is it already mounted suitably? */
+	if (mounted) {
+		if (!readwrite || (readwrite && mounted_rw)) {
+			return 0;
+		} else {
+			/* Mounted RO and we need RW */
+			if (system_lv_umount() < 0)
+				return -1;
+		}
+	}
+
+	/* Randomize the mount point */
+	strcpy(mount_point, SYSTEM_LV_MOUNTPOINT);
+	fd = mkstemp(mount_point);
+	if (fd < 0) {
+		log_error("mkstemp for system LV mount point failed: %m\n");
+		return -1;
+	}
+
+	/* Race condition here but there's no mkstemp for directories */
+	close(fd);
+	unlink(mount_point);
+	mkdir(mount_point, 0600);
+
+	/* Make sure we have a system-lv lock */
+	status =
+	    sync_lock(lock_name, (readwrite) ? LKM_EXMODE : LKM_CRMODE, 0,
+		      &lockid);
+	if (status < 0)
+		return -1;
+
+	/* Mount it */
+	if (mount(system_lv_name, mount_point, SYSTEM_LV_FILESYSTEM,
+		  MS_MGC_VAL | MS_NOSUID | MS_NODEV | MS_NOEXEC | MS_SYNCHRONOUS
+		  | (readwrite ? 0 : MS_RDONLY), NULL) < 0) {
+		/* mount(2) returns EINVAL if the volume has no FS on it. So, if we want to
+		   write to it we try to make a filesystem in it and retry the mount */
+		if (errno == EINVAL && readwrite) {
+			char cmd[256];
+
+			log_error("Attempting mkfs on system LV device %s\n",
+				  system_lv_name);
+			snprintf(cmd, sizeof(cmd), "/sbin/mkfs -t %s %s",
+				 SYSTEM_LV_FILESYSTEM, system_lv_name);
+			system(cmd);
+
+			if (mount
+			    (system_lv_name, mount_point, SYSTEM_LV_FILESYSTEM,
+			     MS_MGC_VAL | MS_NOSUID | MS_NODEV | MS_NOEXEC |
+			     MS_SYNCHRONOUS | (readwrite ? 0 : MS_RDONLY),
+			     NULL) == 0)
+				goto mounted;
+		}
+
+		saved_errno = errno;
+		log_error("mount of system LV (%s, %s, %s) failed: %m\n",
+			  system_lv_name, mount_point, SYSTEM_LV_FILESYSTEM);
+		sync_unlock(lock_name, lockid);
+		errno = saved_errno;
+		return -1;
+	}
+
+      mounted:
+/* Set the internal flags */
+	mounted = 1;
+	mounted_rw = readwrite;
+
+	return 0;
+}
+
+/* Erase *all* files in the root directory of the system LV.
+   This *MUST* be called with an appropriate lock held!
+   The LV is left mounted RW because it is assumed that the
+   caller wants to write something here after clearing some space */
+int system_lv_eraseall(void)
+{
+	DIR *dir;
+	struct dirent *ent;
+	char fname[PATH_MAX];
+
+	/* Must be mounted R/W */
+	system_lv_mount(1);
+
+	dir = opendir(mount_point);
+	if (!dir)
+		return -1;
+
+	while ((ent = readdir(dir))) {
+		struct stat st;
+		snprintf(fname, sizeof(fname), "%s/%s", mount_point,
+			 ent->d_name);
+
+		if (stat(fname, &st)) {
+			if (S_ISREG(st.st_mode))
+				unlink(fname);
+		}
+	}
+	closedir(dir);
+	return 0;
+}
+
+/* This is a "high-level" routine - it mounts the system LV, writes
+   the data into a file named after this node and then umounts the LV
+   again */
+int system_lv_write_data(char *data, ssize_t len)
+{
+	struct utsname nodeinfo;
+	char fname[PATH_MAX];
+	int outfile;
+	ssize_t thiswrite;
+	ssize_t written;
+
+	if (system_lv_mount(1))
+		return -1;
+
+	/* Build the file name we are goingto use. */
+	uname(&nodeinfo);
+	snprintf(fname, sizeof(fname), "%s/%s", mount_point, nodeinfo.nodename);
+
+	/* Open the file for output */
+	outfile = open(fname, O_RDWR | O_CREAT | O_TRUNC, 0600);
+	if (outfile < 0) {
+		int saved_errno = errno;
+		system_lv_umount();
+		errno = saved_errno;
+		return -1;
+	}
+
+	written = 0;
+	do {
+		thiswrite = write(outfile, data + written, len - written);
+		if (thiswrite > 0)
+			written += thiswrite;
+
+	} while (written < len && thiswrite > 0);
+
+	close(outfile);
+
+	system_lv_umount();
+	return (thiswrite < 0) ? -1 : 0;
+}
+
+/* This is a "high-level" routine - it mounts the system LV, reads
+   the data from a named file and then umounts the LV
+   again */
+int system_lv_read_data(char *fname_base, char *data, ssize_t *len)
+{
+	char fname[PATH_MAX];
+	int outfile;
+	struct stat st;
+	ssize_t filesize;
+	ssize_t thisread;
+	ssize_t readbytes;
+
+	if (system_lv_mount(0))
+		return -1;
+
+	/* Build the file name we are going to use. */
+	snprintf(fname, sizeof(fname), "%s/%s", mount_point, fname_base);
+
+	/* Get the file size and stuff. Actually we only need the file size but
+	   this will also check that the file exists */
+	if (stat(fname, &st) < 0) {
+		int saved_errno = errno;
+
+		log_error("stat of file %s on system LV failed: %m\n", fname);
+		system_lv_umount();
+		errno = saved_errno;
+		return -1;
+	}
+	filesize = st.st_size;
+
+	outfile = open(fname, O_RDONLY);
+	if (outfile < 0) {
+		int saved_errno = errno;
+
+		log_error("open of file %s on system LV failed: %m\n", fname);
+		system_lv_umount();
+		errno = saved_errno;
+		return -1;
+	}
+
+	readbytes = 0;
+	do {
+		thisread =
+		    read(outfile, data + readbytes, filesize - readbytes);
+		if (thisread > 0)
+			readbytes += thisread;
+
+	} while (readbytes < filesize && thisread > 0);
+
+	close(outfile);
+
+	system_lv_umount();
+
+	*len = readbytes;
+	return (thisread < 0) ? -1 : 0;
+}

daemons/clvmd/system-lv.h

@@ -0,0 +1,30 @@
+/*
+ * Copyright (C) 2002-2004 Sistina Software, Inc. All rights reserved.
+ * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU General Public License v.2.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#ifndef _CLVM_SYSTEM_LV_H
+#define _CLVM_SYSTEM_LV_H
+
+/* Prototypes for System-LV functions */
+
+/* "low-level" functions */
+extern int system_lv_umount(void);
+extern int system_lv_mount(int readwrite);
+extern int system_lv_eraseall(void);
+
+/* "high-level" functions */
+extern int system_lv_write_data(char *data, ssize_t len);
+extern int system_lv_read_data(char *fname_base, char *data, ssize_t *len);
+
+#endif

daemons/clvmd/tcp-comms.c

@@ -0,0 +1,480 @@
+/******************************************************************************
+*******************************************************************************
+**
+**  Copyright (C) Sistina Software, Inc.  2002-2003  All rights reserved.
+**
+*******************************************************************************
+******************************************************************************/
+
+/* This provides the inter-clvmd communications for a system without CMAN.
+   There is a listening TCP socket which accepts new connections in the
+   normal way.
+   It can also make outgoing connnections to the other clvmd nodes.
+*/
+
+
+#include <pthread.h>
+#include <sys/types.h>
+#include <sys/utsname.h>
+#include <sys/ioctl.h>
+#include <sys/socket.h>
+#include <sys/stat.h>
+#include <sys/socket.h>
+#include <netinet/in.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdint.h>
+#include <fcntl.h>
+#include <string.h>
+#include <stddef.h>
+#include <stdint.h>
+#include <unistd.h>
+#include <errno.h>
+#include <syslog.h>
+#include <netdb.h>
+#include <assert.h>
+
+#include "ccs.h"
+#include "clvm.h"
+#include "clvmd-comms.h"
+#include "clvmd.h"
+#include "clvmd-gulm.h"
+#include "hash.h"
+
+#define DEFAULT_TCP_PORT 21064
+
+static int listen_fd = -1;
+static int tcp_port;
+struct hash_table *sock_hash;
+
+static int get_tcp_port(int default_port);
+static int get_our_ip_address(char *addr, int *family);
+static int read_from_tcpsock(struct local_client *fd, char *buf, int len, char *csid,
+			     struct local_client **new_client);
+
+/* Called by init_cluster() to open up the listening socket */
+// TODO: IPv6 compat.
+int init_comms()
+{
+    struct sockaddr *addr = NULL;
+    struct sockaddr_in addr4;
+    struct sockaddr_in6 addr6;
+    int    addr_len;
+    int    family;
+    char   address[MAX_CSID_LEN];
+
+    sock_hash = hash_create(100);
+    tcp_port = get_tcp_port(DEFAULT_TCP_PORT);
+
+    /* Get IP address and IP type */
+    get_our_ip_address(address, &family);
+    if (family == AF_INET)
+    {
+	memcpy(&addr4.sin_addr, addr, sizeof(struct in_addr));
+	addr = (struct sockaddr *)&addr4;
+	addr4.sin_port = htons(tcp_port);
+	addr_len = sizeof(addr4);
+    }
+    else
+    {
+	memcpy(&addr6.sin6_addr, addr, sizeof(struct in6_addr));
+	addr = (struct sockaddr *)&addr6;
+	addr6.sin6_port = htons(tcp_port);
+	addr_len = sizeof(addr6);
+    }
+
+    listen_fd = socket(family, SOCK_STREAM, 0);
+
+    if (listen_fd < 0)
+    {
+	return -1;
+    }
+    else
+    {
+	int one = 1;
+	setsockopt(listen_fd, SOL_SOCKET, SO_REUSEADDR, &one, sizeof(int));
+    }
+
+    addr->sa_family = family;
+
+    if (bind(listen_fd, addr, addr_len) < 0)
+    {
+	DEBUGLOG("Can't bind to port\n");
+	syslog(LOG_ERR, "Can't bind to port %d, is clvmd already running ?", tcp_port);
+	close(listen_fd);
+	return -1;
+    }
+
+    listen(listen_fd, 5);
+
+    return 0;
+}
+
+void tcp_remove_client(char *csid)
+ {
+    struct local_client *client;
+    DEBUGLOG("tcp_remove_client\n");
+
+    /* Don't actually close the socket here - that's the
+       job of clvmd.c whch will do the job when it notices the
+       other end has gone. We just need to remove the client(s) from
+       the hash table so we don't try to use it for sending any more */
+    client = hash_lookup_binary(sock_hash, csid, MAX_CSID_LEN);
+    if (client)
+    {
+	hash_remove_binary(sock_hash, csid, MAX_CSID_LEN);
+    }
+
+    /* Look for a mangled one too */
+    csid[0] ^= 0x80;
+
+    client = hash_lookup_binary(sock_hash, csid, MAX_CSID_LEN);
+    if (client)
+    {
+	hash_remove_binary(sock_hash, csid, MAX_CSID_LEN);
+    }
+
+    /* Put it back as we found it */
+    csid[0] ^= 0x80;
+}
+
+int alloc_client(int fd, char *csid, struct local_client **new_client)
+{
+    struct local_client *client;
+
+    DEBUGLOG("alloc_client %d csid = [%d.%d.%d.%d]\n", fd,csid[0],csid[1],csid[2],csid[3]);
+
+    /* Create a local_client and return it */
+    client = malloc(sizeof(struct local_client));
+    if (!client)
+    {
+	DEBUGLOG("malloc failed\n");
+	return -1;
+    }
+
+    memset(client, 0, sizeof(struct local_client));
+    client->fd = fd;
+    client->type = CLUSTER_DATA_SOCK;
+    client->callback = read_from_tcpsock;
+    if (new_client)
+	*new_client = client;
+
+    /* Add to our list of node sockets */
+    if (hash_lookup_binary(sock_hash, csid, MAX_CSID_LEN))
+    {
+	DEBUGLOG("alloc_client mangling CSID for second connection\n");
+	/* This is a duplicate connection but we can't close it because
+	   the other end may already have started sending.
+	   So, we mangle the IP address and keep it, all sending will
+	   go out of the main FD
+	*/
+	csid[0] ^= 0x80;
+	client->bits.net.flags = 1; /* indicate mangled CSID */
+
+        /* If it still exists then kill the connection as we should only
+           ever have one incoming connection from each node */
+        if (hash_lookup_binary(sock_hash, csid, MAX_CSID_LEN))
+        {
+	    DEBUGLOG("Multiple incoming connections from node\n");
+            syslog(LOG_ERR, " Bogus incoming connection from %d.%d.%d.%d\n", csid[0],csid[1],csid[2],csid[3]);
+
+	    free(client);
+            errno = ECONNREFUSED;
+            return -1;
+        }
+    }
+    hash_insert_binary(sock_hash, csid, MAX_CSID_LEN, client);
+
+    return 0;
+}
+
+int get_main_cluster_fd()
+{
+    return listen_fd;
+}
+
+
+/* Read on main comms (listen) socket, accept it */
+int cluster_fd_callback(struct local_client *fd, char *buf, int len, char *csid,
+			struct local_client **new_client)
+{
+    int newfd;
+    struct sockaddr_in addr;
+    socklen_t addrlen = sizeof(addr);
+    int status;
+    char name[MAX_CLUSTER_MEMBER_NAME_LEN];
+
+    DEBUGLOG("cluster_fd_callback\n");
+    *new_client = NULL;
+    newfd = accept(listen_fd, (struct sockaddr *)&addr, &addrlen);
+
+    DEBUGLOG("cluster_fd_callback, newfd=%d (errno=%d)\n", newfd, errno);
+    if (!newfd)
+    {
+	syslog(LOG_ERR, "error in accept: %m");
+	errno = EAGAIN;
+	return -1; /* Don't return an error or clvmd will close the listening FD */
+    }
+
+    /* Check that the client is a member of the cluster
+       and reject if not.
+       // FIXME: IPv4 specific
+    */
+    if (name_from_csid((char *)&addr.sin_addr.s_addr, name) < 0)
+    {
+	char *ip = (char *)&addr.sin_addr.s_addr;
+	syslog(LOG_ERR, "Got connect from non-cluster node %d.%d.%d.%d\n",
+	       ip[0], ip[1], ip[2], ip[3]);
+	DEBUGLOG("Got connect from non-cluster node %d.%d.%d.%d\n",
+		 ip[0], ip[1], ip[2], ip[3]);
+	close(newfd);
+
+	errno = EAGAIN;
+	return -1;
+    }
+
+    status = alloc_client(newfd, (char *)&addr.sin_addr.s_addr, new_client);
+    if (status)
+    {
+	DEBUGLOG("cluster_fd_callback, alloc_client failed, status = %d\n", status);
+	close(newfd);
+	/* See above... */
+	errno = EAGAIN;
+	return -1;
+    }
+    DEBUGLOG("cluster_fd_callback, returning %d, %p\n", newfd, *new_client);
+    return newfd;
+}
+
+
+static int read_from_tcpsock(struct local_client *client, char *buf, int len, char *csid,
+			     struct local_client **new_client)
+{
+    struct sockaddr_in addr;
+    socklen_t slen = sizeof(addr);
+    int status;
+
+    DEBUGLOG("read_from_tcpsock fd %d\n", client->fd);
+    *new_client = NULL;
+
+    /* Get "csid" */
+    getpeername(client->fd, (struct sockaddr *)&addr, &slen);
+    memcpy(csid, &addr.sin_addr.s_addr, MAX_CSID_LEN);
+
+    status = read(client->fd, buf, len);
+
+    DEBUGLOG("read_from_tcpsock, status = %d(errno = %d)\n", status, errno);
+
+    /* Remove it from the hash table if there's an error, clvmd will
+       remove the socket from its lists and free the client struct */
+    if (status == 0 ||
+	(status < 0 && errno != EAGAIN && errno != EINTR))
+    {
+	char remcsid[MAX_CSID_LEN];
+
+	memcpy(remcsid, csid, MAX_CSID_LEN);
+	close(client->fd);
+
+	/* If the csid was mangled, then make sure we remove the right entry */
+	if (client->bits.net.flags)
+	    remcsid[0] ^= 0x80;
+	hash_remove_binary(sock_hash, remcsid, MAX_CSID_LEN);
+
+	/* Tell cluster manager layer */
+	add_down_node(remcsid);
+    }
+    return status;
+}
+
+static int connect_csid(char *csid, struct local_client **newclient)
+{
+    int fd;
+    struct sockaddr_in addr;
+    int status;
+
+    DEBUGLOG("Connecting socket\n");
+    fd = socket(PF_INET, SOCK_STREAM, 0);
+
+    if (fd < 0)
+    {
+	syslog(LOG_ERR, "Unable to create new socket: %m");
+	return -1;
+    }
+
+    addr.sin_family = AF_INET;
+    memcpy(&addr.sin_addr.s_addr, csid, MAX_CSID_LEN);
+    addr.sin_port = htons(tcp_port);
+
+    DEBUGLOG("Connecting socket %d\n", fd);
+    if (connect(fd, (struct sockaddr *)&addr, sizeof(struct sockaddr_in)) < 0)
+    {
+	syslog(LOG_ERR, "Unable to connect to remote node: %m");
+	DEBUGLOG("Unable to connect to remote node: %s\n", strerror(errno));
+	close(fd);
+	return -1;
+    }
+
+    status = alloc_client(fd, csid, newclient);
+    if (status)
+	close(fd);
+    else
+	add_client(*newclient);
+
+    /* If we can connect to it, it must be running a clvmd */
+    add_up_node(csid);
+    return status;
+}
+
+/* Send a message to a known CSID */
+static int tcp_send_message(void *buf, int msglen, unsigned char *csid, const char *errtext)
+{
+    int status;
+    struct local_client *client;
+    char ourcsid[MAX_CSID_LEN];
+
+    assert(csid);
+
+    DEBUGLOG("tcp_send_message, csid = [%d.%d.%d.%d], msglen = %d\n", csid[0],csid[1],csid[2],csid[3], msglen);
+
+    /* Don't connect to ourself */
+    get_our_csid(ourcsid);
+    if (memcmp(csid, ourcsid, MAX_CSID_LEN) == 0)
+	return msglen;
+
+    client = hash_lookup_binary(sock_hash, csid, MAX_CSID_LEN);
+    if (!client)
+    {
+	status = connect_csid(csid, &client);
+	if (status)
+	    return -1;
+    }
+    DEBUGLOG("tcp_send_message, fd = %d\n", client->fd);
+
+    return write(client->fd, buf, msglen);
+}
+
+
+int cluster_send_message(void *buf, int msglen, char *csid, const char *errtext)
+{
+    int status=0;
+
+    DEBUGLOG("cluster send message, csid = %p, msglen = %d\n", csid, msglen);
+
+    /* If csid is NULL then send to all known (not just connected) nodes */
+    if (!csid)
+    {
+	void *context = NULL;
+	char loop_csid[MAX_CSID_LEN];
+
+	/* Loop round all gulm-known nodes */
+	while (get_next_node_csid(&context, loop_csid))
+	{
+	    status = tcp_send_message(buf, msglen, loop_csid, errtext);
+	    if (status == 0 ||
+		(status < 0 && (errno == EAGAIN || errno == EINTR)))
+		break;
+	}
+    }
+    else
+    {
+
+	status = tcp_send_message(buf, msglen, csid, errtext);
+    }
+    return status;
+}
+
+static int get_tcp_port(int default_port)
+{
+    int ccs_handle;
+    int port = default_port;
+    char *portstr;
+
+    ccs_handle = ccs_connect();
+    if (ccs_handle)
+    {
+	return port;
+    }
+
+    if (!ccs_get(ccs_handle, "//clvm/@port", &portstr))
+    {
+	port = atoi(portstr);
+	free(portstr);
+
+	if (port <= 0 && port >= 65536)
+	    port = default_port;
+    }
+    ccs_disconnect(ccs_handle);
+
+    DEBUGLOG("Using port %d for communications\n", port);
+    return port;
+}
+
+/* To get our own IP address we get the locally bound address of the
+   socket that's talking to GULM in the assumption(eek) that it will
+   be on the "right" network in a multi-homed system */
+static int get_our_ip_address(char *addr, int *family)
+{
+    /* Use a sockaddr_in6 to make sure it's big enough */
+    struct sockaddr_in6 saddr;
+    int socklen = sizeof(saddr);
+
+    if (!getsockname(gulm_fd(), (struct sockaddr *)&saddr, &socklen))
+    {
+	if (saddr.sin6_family == AF_INET6)
+	{
+	    memcpy(addr, &saddr.sin6_addr, sizeof(saddr.sin6_addr));
+	}
+	else
+	{
+	    struct sockaddr_in *sin4 = (struct sockaddr_in *)&saddr;
+	    memcpy(addr, &sin4->sin_addr, sizeof(sin4->sin_addr));
+	}
+	return 0;
+    }
+    return -1;
+}
+
+/* Public version of above for those that don't care what protocol
+   we're using */
+void get_our_csid(char *csid)
+{
+    static char our_csid[MAX_CSID_LEN];
+    static int got_csid = 0;
+
+    if (!got_csid)
+    {
+	int family;
+
+	memset(our_csid, 0, sizeof(our_csid));
+	if (get_our_ip_address(our_csid, &family))
+	{
+	    got_csid = 1;
+	}
+    }
+    memcpy(csid, our_csid, MAX_CSID_LEN);
+}
+
+/* Get someone else's IP address from DNS */
+int get_ip_address(char *node, char *addr)
+{
+    struct hostent *he;
+
+    memset(addr, 0, MAX_CSID_LEN);
+
+    // TODO: what do we do about multi-homed hosts ???
+    // CCSs ip_interfaces solved this but some bugger removed it.
+
+    /* Try IPv6 first. The man page for gethostbyname implies that
+       it will lookup ip6 & ip4 names, but it seems not to */
+    he = gethostbyname2(node, AF_INET6);
+    if (!he)
+	he = gethostbyname2(node, AF_INET);
+    if (!he)
+	return -1;
+
+    /* For IPv4 address just use the lower 4 bytes */
+    memcpy(&addr, he->h_addr_list[0],
+	   he->h_length);
+
+    return 0;
+}

daemons/clvmd/tcp-comms.h

@@ -0,0 +1,7 @@
+#include <netinet/in.h>
+
+#define MAX_CLUSTER_MESSAGE 1600
+#define MAX_CSID_LEN sizeof(struct in6_addr)
+#define MAX_CLUSTER_MEMBER_NAME_LEN 128
+
+extern int init_comms(void);

doc/Makefile.in

@@ -18,7 +18,7 @@ VPATH = @srcdir@
 CONFSRC=example.conf
 CONFDEST=lvm.conf
 
-include ../make.tmpl
+include $(top_srcdir)/make.tmpl
 
 install:
 	@if [ ! -e $(confdir)/$(CONFDEST) ]; then \

include/.symlinks

@@ -1,4 +1,6 @@
+../daemons/clvmd/clvm.h
 ../lib/activate/activate.h
+../lib/activate/targets.h
 ../lib/cache/lvmcache.h
 ../lib/commands/errors.h
 ../lib/commands/toolcontext.h
@@ -20,12 +22,16 @@
 ../lib/filters/filter-sysfs.h
 ../lib/filters/filter.h
 ../lib/format1/format1.h
+../lib/format_pool/format_pool.h
 ../lib/format_text/format-text.h
+../lib/format_text/text_export.h
+../lib/format_text/text_import.h
 ../lib/label/label.h
 ../lib/locking/locking.h
 ../lib/log/log.h
 ../lib/metadata/lv_alloc.h
 ../lib/metadata/metadata.h
+../lib/metadata/segtype.h
 ../lib/mm/dbg_malloc.h
 ../lib/mm/memlock.h
 ../lib/mm/pool.h

include/Makefile.in

@@ -20,6 +20,8 @@ VPATH = @srcdir@
 
 LN_S = @LN_S@
 
+.PHONY: clean distclean all install pofile install_cluster
+
 all: .symlinks_created
 
 .symlinks_created: .symlinks
@@ -37,5 +39,5 @@ clean:
 
 install:
 
-.PHONY: clean distclean all install pofile
+install_cluster:
 

lib/Makefile.in

@@ -20,6 +20,18 @@ ifeq ("@LVM1@", "shared")
   SUBDIRS = format1
 endif
 
+ifeq ("@POOL@", "shared")
+  SUBDIRS += format_pool
+endif
+
+ifeq ("@SNAPSHOTS@", "shared")
+  SUBDIRS += snapshot
+endif
+
+ifeq ("@MIRRORS@", "shared")
+  SUBDIRS += mirror
+endif
+
 SOURCES =\
 	activate/activate.c \
 	cache/lvmcache.c \
@@ -33,6 +45,7 @@ SOURCES =\
 	device/dev-io.c \
 	device/device.c \
 	display/display.c \
+	error/errseg.c \
 	filters/filter-composite.c \
 	filters/filter-persistent.c \
 	filters/filter-regex.c \
@@ -57,6 +70,7 @@ SOURCES =\
 	metadata/metadata.c \
 	metadata/mirror.c \
 	metadata/pv_map.c \
+	metadata/segtype.c \
 	metadata/snapshot_manip.c \
 	misc/crc.c \
 	misc/lvm-file.c \
@@ -67,7 +81,9 @@ SOURCES =\
 	regex/parse_rx.c \
 	regex/ttree.c \
 	report/report.c \
-	uuid/uuid.c 
+	striped/striped.c \
+	uuid/uuid.c \
+	zero/zero.c
 
 ifeq ("@LVM1@", "internal")
   SOURCES +=\
@@ -80,6 +96,30 @@ ifeq ("@LVM1@", "internal")
 	format1/vg_number.c
 endif
 
+ifeq ("@POOL@", "internal")
+  SOURCES +=\
+	format_pool/disk_rep.c \
+	format_pool/format_pool.c \
+	format_pool/import_export.c \
+	format_pool/pool_label.c
+endif
+
+ifeq ("@CLUSTER@", "internal")
+  SOURCES += locking/cluster_locking.c
+endif
+
+ifeq ("@CLUSTER@", "shared")
+  SUBDIRS += locking
+endif
+
+ifeq ("@SNAPSHOTS@", "internal")
+  SOURCES += snapshot/snapshot.c
+endif
+
+ifeq ("@MIRRORS@", "internal")
+  SOURCES += mirror/mirrored.c
+endif
+
 ifeq ("@DEBUG@", "yes")
   SOURCES += mm/dbg_malloc.c
 endif
@@ -104,5 +144,5 @@ LIB_STATIC = liblvm.a
 
 $(SUBDIRS): $(LIB_STATIC)
 
-include ../make.tmpl
+include $(top_srcdir)/make.tmpl
 

lib/activate/activate.c

@@ -25,6 +25,7 @@
 #include "toolcontext.h"
 #include "dev_manager.h"
 #include "str_list.h"
+#include "config.h"
 
 #include <limits.h>
 #include <fcntl.h>
@@ -173,13 +174,13 @@ int activation(void)
 static int _passes_activation_filter(struct cmd_context *cmd,
 				     struct logical_volume *lv)
 {
-	struct config_node *cn;
+	const struct config_node *cn;
 	struct config_value *cv;
 	char *str;
 	char path[PATH_MAX];
 
 	if (!(cn = find_config_node(cmd->cft->root, "activation/volume_list"))) {
-		/* If no hosts tags defined, activate */
+		/* If no host tags defined, activate */
 		if (list_empty(&cmd->tags))
 			return 1;
 
@@ -341,7 +342,7 @@ static int _lv_info(const struct logical_volume *lv, int mknodes,
 	if (!activation())
 		return 0;
 
-	if (!(dm = dev_manager_create(lv->vg->name, lv->vg->cmd->cft))) {
+	if (!(dm = dev_manager_create(lv->vg->cmd, lv->vg->name))) {
 		stack;
 		return 0;
 	}
@@ -387,7 +388,7 @@ int lv_snapshot_percent(struct logical_volume *lv, float *percent)
 	if (!activation())
 		return 0;
 
-	if (!(dm = dev_manager_create(lv->vg->name, lv->vg->cmd->cft))) {
+	if (!(dm = dev_manager_create(lv->vg->cmd, lv->vg->name))) {
 		stack;
 		return 0;
 	}
@@ -419,7 +420,7 @@ int lv_mirror_percent(struct logical_volume *lv, int wait, float *percent,
 	if (!info.exists)
 		return 0;
 
-	if (!(dm = dev_manager_create(lv->vg->name, lv->vg->cmd->cft))) {
+	if (!(dm = dev_manager_create(lv->vg->cmd, lv->vg->name))) {
 		stack;
 		return 0;
 	}
@@ -462,7 +463,7 @@ static int _lv_activate_lv(struct logical_volume *lv)
 	int r;
 	struct dev_manager *dm;
 
-	if (!(dm = dev_manager_create(lv->vg->name, lv->vg->cmd->cft))) {
+	if (!(dm = dev_manager_create(lv->vg->cmd, lv->vg->name))) {
 		stack;
 		return 0;
 	}
@@ -479,7 +480,7 @@ static int _lv_deactivate(struct logical_volume *lv)
 	int r;
 	struct dev_manager *dm;
 
-	if (!(dm = dev_manager_create(lv->vg->name, lv->vg->cmd->cft))) {
+	if (!(dm = dev_manager_create(lv->vg->cmd, lv->vg->name))) {
 		stack;
 		return 0;
 	}
@@ -496,7 +497,7 @@ static int _lv_suspend_lv(struct logical_volume *lv)
 	int r;
 	struct dev_manager *dm;
 
-	if (!(dm = dev_manager_create(lv->vg->name, lv->vg->cmd->cft))) {
+	if (!(dm = dev_manager_create(lv->vg->cmd, lv->vg->name))) {
 		stack;
 		return 0;
 	}
@@ -509,7 +510,7 @@ static int _lv_suspend_lv(struct logical_volume *lv)
 }
 
 /*
- * These two functions return the number of LVs in the state,
+ * These two functions return the number of visible LVs in the state,
  * or -1 on error.
  */
 int lvs_in_vg_activated(struct volume_group *vg)
@@ -523,7 +524,8 @@ int lvs_in_vg_activated(struct volume_group *vg)
 
 	list_iterate(lvh, &vg->lvs) {
 		lv = list_item(lvh, struct lv_list)->lv;
-		count += (_lv_active(lv) == 1);
+		if (lv->status & VISIBLE_LV)
+			count += (_lv_active(lv) == 1);
 	}
 
 	return count;
@@ -540,7 +542,8 @@ int lvs_in_vg_opened(struct volume_group *vg)
 
 	list_iterate(lvh, &vg->lvs) {
 		lv = list_item(lvh, struct lv_list)->lv;
-		count += (_lv_open_count(lv) > 0);
+		if (lv->status & VISIBLE_LV)
+			count += (_lv_open_count(lv) > 0);
 	}
 
 	return count;

lib/activate/dev_manager.c

@@ -21,7 +21,11 @@
 #include "lvm-string.h"
 #include "fs.h"
 #include "defaults.h"
+#include "segtype.h"
+#include "display.h"
 #include "toolcontext.h"
+#include "targets.h"
+#include "config.h"
 
 #include <libdevmapper.h>
 #include <limits.h>
@@ -62,13 +66,8 @@ enum {
 	SUSPENDED = 4,
 	NOPROPAGATE = 5,
 	TOPLEVEL = 6,
-	REMOVE = 7
-};
-
-enum {
-	MIRR_DISABLED,
-	MIRR_RUNNING,
-	MIRR_COMPLETED
+	REMOVE = 7,
+	RESUME_IMMEDIATE = 8
 };
 
 typedef enum {
@@ -112,14 +111,14 @@ struct dl_list {
 };
 
 static const char *stripe_filler = NULL;
-static uint32_t mirror_region_size = 0;
 
 struct dev_manager {
 	struct pool *mem;
 
-	struct config_tree *cft;
+	struct cmd_context *cmd;
+
 	const char *stripe_filler;
-	uint32_t mirror_region_size;
+	void *target_state;
 	uint32_t pvmove_mirror_count;
 
 	char *vg_name;
@@ -166,89 +165,6 @@ static inline void _clear_flag(struct dev_layer *dl, int bit)
 	dl->flags &= ~(1 << bit);
 }
 
-/*
- * Device layer names are all of the form <vg>-<lv>-<layer>, any
- * other hyphens that appear in these names are quoted with yet
- * another hyphen.  The top layer of any device has no layer
- * name.  eg, vg0-lvol0.
- */
-static void _count_hyphens(const char *str, size_t *len, int *hyphens)
-{
-	const char *ptr;
-
-	for (ptr = str; *ptr; ptr++, (*len)++)
-		if (*ptr == '-')
-			(*hyphens)++;
-}
-
-/*
- * Copies a string, quoting hyphens with hyphens.
- */
-static void _quote_hyphens(char **out, const char *src)
-{
-	while (*src) {
-		if (*src == '-')
-			*(*out)++ = '-';
-
-		*(*out)++ = *src++;
-	}
-}
-
-/*
- * <vg>-<lv>-<layer> or if !layer just <vg>-<lv>.
- */
-static char *_build_name(struct pool *mem, const char *vg,
-			 const char *lv, const char *layer)
-{
-	size_t len = 0;
-	int hyphens = 0;
-	char *r, *out;
-
-	_count_hyphens(vg, &len, &hyphens);
-	_count_hyphens(lv, &len, &hyphens);
-
-	if (layer && *layer)
-		_count_hyphens(layer, &len, &hyphens);
-
-	len += hyphens + 2;
-
-	if (!(r = pool_alloc(mem, len))) {
-		stack;
-		return NULL;
-	}
-
-	out = r;
-	_quote_hyphens(&out, vg);
-	*out++ = '-';
-	_quote_hyphens(&out, lv);
-
-	if (layer && *layer) {
-		*out++ = '-';
-		_quote_hyphens(&out, layer);
-	}
-	*out = '\0';
-
-	return r;
-}
-
-/* Find start of LV component in hyphenated name */
-static char *_find_lv_name(char *vg)
-{
-	char *c = vg;
-
-	while (*c && *(c + 1)) {
-		if (*c == '-') {
-			if (*(c + 1) == '-')
-				c++;
-			else
-				return (c + 1);
-		}
-		c++;
-	}
-
-	return NULL;
-}
-
 static char *_build_dlid(struct pool *mem, const char *lvid, const char *layer)
 {
 	char *dlid;
@@ -336,7 +252,7 @@ static int _info_run(const char *name, const char *uuid, struct dm_info *info,
 static int _info(const char *name, const char *uuid, int mknodes,
 		 struct dm_info *info, struct pool *mem, char **uuid_out)
 {
-	if (!mknodes && uuid && *uuid && 
+	if (!mknodes && uuid && *uuid &&
 	    _info_run(NULL, uuid, info, 0, mem, uuid_out) && info->exists)
 		return 1;
 
@@ -427,18 +343,16 @@ static int _percent_run(struct dev_manager *dm, const char *name,
 	uint64_t start, length;
 	char *type = NULL;
 	char *params = NULL;
-	float percent2;
 	struct list *segh = &lv->segments;
 	struct lv_segment *seg = NULL;
+	struct segment_type *segtype;
 
-	uint64_t numerator, denominator;
 	uint64_t total_numerator = 0, total_denominator = 0;
 
 	*percent = -1;
 
 	if (!(dmt = _setup_task(name, uuid, event_nr,
-				wait ? DM_DEVICE_WAITEVENT : DM_DEVICE_STATUS)))
-	{
+				wait ? DM_DEVICE_WAITEVENT : DM_DEVICE_STATUS))) {
 		stack;
 		return 0;
 	}
@@ -471,40 +385,19 @@ static int _percent_run(struct dev_manager *dm, const char *name,
 		if (!type || !params || strcmp(type, target_type))
 			continue;
 
-		/* Mirror? */
-		if (!strcmp(type, "mirror")) {
-			log_debug("Mirror status: %s", params);
-			if (sscanf(params, "%*d %*x:%*x %*x:%*x %" PRIu64
-				   "/%" PRIu64, &numerator,
-				   &denominator) != 2) {
-				log_error("Failure parsing mirror status: %s",
-					  params);
-				goto out;
-			}
-			total_numerator += numerator;
-			total_denominator += denominator;
-
-			if (seg && (seg->status & PVMOVE))
-				seg->extents_moved = dm->mirror_region_size *
-				    numerator / lv->vg->extent_size;
+		if (!(segtype = get_segtype_from_string(dm->cmd, type)))
 			continue;
+
+		if (segtype->ops->target_percent &&
+		    !segtype->ops->target_percent(&dm->target_state, dm->mem,
+						  dm->cmd->cft, seg, params,
+						  &total_numerator,
+						  &total_denominator,
+						  percent)) {
+			stack;
+			goto out;
 		}
 
-		if (strcmp(type, "snapshot"))
-			continue;
-
-		/* Snapshot */
-		if (index(params, '/')) {
-			if (sscanf(params, "%" PRIu64 "/%" PRIu64,
-				   &numerator, &denominator) == 2) {
-				total_numerator += numerator;
-				total_denominator += denominator;
-			}
-			continue;
-		} else if (sscanf(params, "%f", &percent2) == 1) {
-			*percent += percent2;
-			*percent /= 2;
-		}
 	} while (next);
 
 	if (lv && (segh = list_next(&lv->segments, segh))) {
@@ -515,7 +408,7 @@ static int _percent_run(struct dev_manager *dm, const char *name,
 
 	if (total_denominator)
 		*percent = (float) total_numerator *100 / total_denominator;
-	else
+	else if (*percent < 0)
 		*percent = 100;
 
 	log_debug("LV percent: %f", *percent);
@@ -543,10 +436,16 @@ static int _percent(struct dev_manager *dm, const char *name, const char *uuid,
 	return 0;
 }
 
-static int _rename(struct dev_layer *dl, char *newname)
+static int _rename(struct dev_manager *dm, struct dev_layer *dl, char *newname)
 {
 	int r = 1;
 	struct dm_task *dmt;
+	char *vgname, *lvname, *layer;
+
+	if (!split_dm_name(dm->mem, dl->name, &vgname, &lvname, &layer)) {
+		log_error("Couldn't split up dm layer name %s", dl->name);
+		return 0;
+	}
 
 	log_verbose("Renaming %s to %s", dl->name, newname);
 
@@ -561,11 +460,13 @@ static int _rename(struct dev_layer *dl, char *newname)
 		goto out;
 	}
 
-	if (!(r = dm_task_run(dmt)))
+	if (!(r = dm_task_run(dmt))) {
 		log_error("Couldn't rename device '%s'.", dl->name);
+		goto out;
+	}
 
 	if (r && _get_flag(dl, VISIBLE))
-		fs_rename_lv(dl->lv, newname, _find_lv_name(dl->name));
+		fs_rename_lv(dl->lv, newname, lvname);
 
 	dl->name = newname;
 
@@ -574,6 +475,55 @@ static int _rename(struct dev_layer *dl, char *newname)
 	return r;
 }
 
+static int _suspend_or_resume(const char *name, action_t suspend)
+{
+	int r;
+	struct dm_task *dmt;
+	int sus = (suspend == SUSPEND) ? 1 : 0;
+	int task = sus ? DM_DEVICE_SUSPEND : DM_DEVICE_RESUME;
+
+	log_very_verbose("%s %s", sus ? "Suspending" : "Resuming", name);
+	if (!(dmt = _setup_task(name, NULL, 0, task))) {
+		stack;
+		return 0;
+	}
+
+	if (!(r = dm_task_run(dmt)))
+		log_error("Couldn't %s device '%s'", sus ? "suspend" : "resume",
+			  name);
+
+	dm_task_destroy(dmt);
+	return r;
+}
+
+static int _suspend(struct dev_layer *dl)
+{
+	if (!dl->info.exists || dl->info.suspended)
+		return 1;
+
+	if (!_suspend_or_resume(dl->name, SUSPEND)) {
+		stack;
+		return 0;
+	}
+
+	dl->info.suspended = 1;
+	return 1;
+}
+
+static int _resume(struct dev_layer *dl)
+{
+	if (!dl->info.exists || !dl->info.suspended)
+		return 1;
+
+	if (!_suspend_or_resume(dl->name, RESUME)) {
+		stack;
+		return 0;
+	}
+
+	dl->info.suspended = 0;
+	return 1;
+}
+
 static int _load(struct dev_manager *dm, struct dev_layer *dl, int task)
 {
 	int r = 1;
@@ -633,9 +583,9 @@ static int _load(struct dev_manager *dm, struct dev_layer *dl, int task)
 		log_error("Couldn't load device '%s'.", dl->name);
 		if ((dl->lv->minor >= 0 || dl->lv->major >= 0) &&
 		    _get_flag(dl, VISIBLE))
-			log_error("Perhaps the persistent device number "
-				  "%d:%d is already in use?",
-				  dl->lv->major, dl->lv->minor);
+			    log_error("Perhaps the persistent device number "
+				      "%d:%d is already in use?",
+				      dl->lv->major, dl->lv->minor);
 	}
 
 	if (!dm_task_get_info(dmt, &dl->info)) {
@@ -650,6 +600,13 @@ static int _load(struct dev_manager *dm, struct dev_layer *dl, int task)
 		goto out;
 	}
 
+	if (_get_flag(dl, RESUME_IMMEDIATE) && dl->info.suspended &&
+	    !_resume(dl)) {
+		stack;
+		r = 0;
+		goto out;
+	}
+
 	log_very_verbose("Activated %s %s %03u:%03u", dl->name,
 			 dl->dlid, dl->info.major, dl->info.minor);
 
@@ -696,55 +653,6 @@ static int _remove(struct dev_layer *dl)
 	return r;
 }
 
-static int _suspend_or_resume(const char *name, action_t suspend)
-{
-	int r;
-	struct dm_task *dmt;
-	int sus = (suspend == SUSPEND) ? 1 : 0;
-	int task = sus ? DM_DEVICE_SUSPEND : DM_DEVICE_RESUME;
-
-	log_very_verbose("%s %s", sus ? "Suspending" : "Resuming", name);
-	if (!(dmt = _setup_task(name, NULL, 0, task))) {
-		stack;
-		return 0;
-	}
-
-	if (!(r = dm_task_run(dmt)))
-		log_error("Couldn't %s device '%s'", sus ? "suspend" : "resume",
-			  name);
-
-	dm_task_destroy(dmt);
-	return r;
-}
-
-static int _suspend(struct dev_layer *dl)
-{
-	if (!dl->info.exists || dl->info.suspended)
-		return 1;
-
-	if (!_suspend_or_resume(dl->name, SUSPEND)) {
-		stack;
-		return 0;
-	}
-
-	dl->info.suspended = 1;
-	return 1;
-}
-
-static int _resume(struct dev_layer *dl)
-{
-	if (!dl->info.exists || !dl->info.suspended)
-		return 1;
-
-	if (!_suspend_or_resume(dl->name, RESUME)) {
-		stack;
-		return 0;
-	}
-
-	dl->info.suspended = 0;
-	return 1;
-}
-
 /*
  * The functions that populate the table in a dm_task as part of
  * a create/reload.
@@ -759,98 +667,26 @@ static int _emit_target_line(struct dev_manager *dm, struct dm_task *dmt,
 			     size_t paramsize)
 {
 	uint64_t esize = seg->lv->vg->extent_size;
-	uint32_t s, start_area = 0u, areas = seg->area_count;
-	int w = 0, tw = 0;
+	int w = 0;
 	const char *target = NULL;
-	const char *trailing_space;
-	int mirror_status;
-	struct dev_layer *dl;
-	char devbuf[10];
+	int r;
 
-	switch (seg->type) {
-	case SEG_SNAPSHOT:
+	if (!seg->segtype->ops->compose_target_line) {
 		log_error("_emit_target: Internal error: Can't handle "
-			  "SEG_SNAPSHOT");
+			  "segment type %s", seg->segtype->name);
 		return 0;
-		/* Target formats:
-		 *   linear [device offset]+
-		 *   striped #stripes stripe_size [device offset]+
-		 *   mirror  log_type #log_params [log_params]* 
-		 *           #mirrors [device offset]+
-		 */
-	case SEG_STRIPED:
-		if (areas == 1)
-			target = "linear";
-		else if (areas > 1) {
-			target = "striped";
-			if ((tw = lvm_snprintf(params, paramsize, "%u %u ",
-					       areas, seg->stripe_size)) < 0)
-				goto error;
-			w = tw;
-		} else {
-			log_error("_emit_target: Internal error: SEG_STRIPED "
-				  "with no stripes");
-			return 0;
-		}
-		break;
-	case SEG_MIRRORED:
-		mirror_status = MIRR_RUNNING;
-		if (seg->status & PVMOVE) {
-			if (seg->extents_moved == seg->area_len) {
-				mirror_status = MIRR_COMPLETED;
-				start_area = 1;
-			} else if (dm->pvmove_mirror_count++) {
-				mirror_status = MIRR_DISABLED;
-				areas = 1;
-			}
-		}
-		if (mirror_status != MIRR_RUNNING) {
-			target = "linear";
-			break;
-		}
-		target = "mirror";
-		if ((tw = lvm_snprintf(params, paramsize, "core 1 %u %u ",
-				       dm->mirror_region_size, areas)) < 0)
-			goto error;
-		w = tw;
-		break;
 	}
 
-	for (s = start_area; s < areas; s++, w += tw) {
-		trailing_space = (areas - s - 1) ? " " : "";
-		if ((seg->area[s].type == AREA_PV &&
-		     (!seg->area[s].u.pv.pv || !seg->area[s].u.pv.pv->dev)) ||
-		    (seg->area[s].type == AREA_LV && !seg->area[s].u.lv.lv))
-			tw = lvm_snprintf(params + w, paramsize - w,
-					  "%s 0%s", dm->stripe_filler,
-					  trailing_space);
-		else if (seg->area[s].type == AREA_PV)
-			tw = lvm_snprintf(params + w, paramsize - w,
-					  "%s %" PRIu64 "%s",
-					  dev_name(seg->area[s].u.pv.pv->dev),
-					  (seg->area[s].u.pv.pv->pe_start +
-					   (esize * seg->area[s].u.pv.pe)),
-					  trailing_space);
-		else {
-			if (!(dl = hash_lookup(dm->layers,
-					       seg->area[s].u.lv.lv->lvid.s))) {
-				log_error("device layer %s missing from hash",
-					  seg->area[s].u.lv.lv->lvid.s);
-				return 0;
-			}
-			if (!dm_format_dev(devbuf, sizeof(devbuf), dl->info.major, dl->info.minor)) {
-				log_error("Failed to format device number as dm target (%u,%u)",
-					  dl->info.major, dl->info.minor);
-				return 0;
-			}
-			tw = lvm_snprintf(params + w, paramsize - w,
-					  "%s %" PRIu64 "%s", devbuf,
-					  esize * seg->area[s].u.lv.le,
-					  trailing_space);
-		}
-
-		if (tw < 0)
-			goto error;
+	if ((r = seg->segtype->ops->compose_target_line(dm, dm->mem,
+							dm->cmd->cft,
+							&dm->target_state, seg,
+							params, paramsize,
+							&target, &w,
+							&dm->
+							pvmove_mirror_count)) <=
+	    0) {
+		stack;
+		return r;
 	}
 
 	log_debug("Adding target: %" PRIu64 " %" PRIu64 " %s %s",
@@ -863,11 +699,62 @@ static int _emit_target_line(struct dev_manager *dm, struct dm_task *dmt,
 	}
 
 	return 1;
+}
 
-      error:
-	log_debug("Insufficient space in params[%" PRIsize_t "] for target "
-		  "parameters.", paramsize);
-	return -1;
+int compose_areas_line(struct dev_manager *dm, struct lv_segment *seg,
+		       char *params, size_t paramsize, int *pos, int start_area,
+		       int areas)
+{
+	uint32_t s;
+	int tw = 0;
+	const char *trailing_space;
+	uint64_t esize = seg->lv->vg->extent_size;
+	struct dev_layer *dl;
+	char devbuf[10];
+
+	for (s = start_area; s < areas; s++, *pos += tw) {
+		trailing_space = (areas - s - 1) ? " " : "";
+		if ((seg->area[s].type == AREA_PV &&
+		     (!seg->area[s].u.pv.pv || !seg->area[s].u.pv.pv->dev)) ||
+		    (seg->area[s].type == AREA_LV && !seg->area[s].u.lv.lv))
+			tw = lvm_snprintf(params + *pos, paramsize - *pos,
+					  "%s 0%s", dm->stripe_filler,
+					  trailing_space);
+		else if (seg->area[s].type == AREA_PV)
+			tw = lvm_snprintf(params + *pos, paramsize - *pos,
+					  "%s %" PRIu64 "%s",
+					  dev_name(seg->area[s].u.pv.pv->dev),
+					  (seg->area[s].u.pv.pv->pe_start +
+					   (esize * seg->area[s].u.pv.pe)),
+					  trailing_space);
+		else {
+			if (!(dl = hash_lookup(dm->layers,
+					       seg->area[s].u.lv.lv->lvid.s))) {
+				log_error("device layer %s missing from hash",
+					  seg->area[s].u.lv.lv->lvid.s);
+				return 0;
+			}
+			if (!dm_format_dev
+			    (devbuf, sizeof(devbuf), dl->info.major,
+			     dl->info.minor)) {
+				log_error
+				    ("Failed to format device number as dm target (%u,%u)",
+				     dl->info.major, dl->info.minor);
+				return 0;
+			}
+			tw = lvm_snprintf(params + *pos, paramsize - *pos,
+					  "%s %" PRIu64 "%s", devbuf,
+					  esize * seg->area[s].u.lv.le,
+					  trailing_space);
+		}
+
+		if (tw < 0) {
+			stack;
+			return -1;
+		}
+	}
+
+	return 1;
 }
 
 static int _emit_target(struct dev_manager *dm, struct dm_task *dmt,
@@ -892,6 +779,9 @@ static int _emit_target(struct dev_manager *dm, struct dm_task *dmt,
 		if (ret >= 0)
 			return ret;
 
+		log_debug("Insufficient space in params[%" PRIsize_t
+			  "] for target parameters.", paramsize);
+
 		paramsize *= 2;
 	} while (paramsize < MAX_TARGET_PARAMSIZE);
 
@@ -1023,8 +913,8 @@ static int _populate_snapshot(struct dev_manager *dm,
 /*
  * dev_manager implementation.
  */
-struct dev_manager *dev_manager_create(const char *vg_name,
-				       struct config_tree *cft)
+struct dev_manager *dev_manager_create(struct cmd_context *cmd,
+				       const char *vg_name)
 {
 	struct pool *mem;
 	struct dev_manager *dm;
@@ -1039,22 +929,16 @@ struct dev_manager *dev_manager_create(const char *vg_name,
 		goto bad;
 	}
 
+	dm->cmd = cmd;
 	dm->mem = mem;
-	dm->cft = cft;
+
 	if (!stripe_filler) {
-		stripe_filler = find_config_str(cft->root,
+		stripe_filler = find_config_str(cmd->cft->root,
 						"activation/missing_stripe_filler",
 						DEFAULT_STRIPE_FILLER);
 	}
 	dm->stripe_filler = stripe_filler;
 
-	if (!mirror_region_size) {
-		mirror_region_size = 2 * find_config_int(cft->root,
-							 "activation/mirror_region_size",
-							 DEFAULT_MIRROR_REGION_SIZE);
-	}
-	dm->mirror_region_size = mirror_region_size;
-
 	if (!(dm->vg_name = pool_strdup(dm->mem, vg_name))) {
 		stack;
 		goto bad;
@@ -1070,6 +954,8 @@ struct dev_manager *dev_manager_create(const char *vg_name,
 	list_init(&dm->remove_list);
 	list_init(&dm->suspend_list);
 
+	dm->target_state = NULL;
+
 	return dm;
 
       bad:
@@ -1091,7 +977,7 @@ int dev_manager_info(struct dev_manager *dm, const struct logical_volume *lv,
 	/*
 	 * Build a name for the top layer.
 	 */
-	if (!(name = _build_name(dm->mem, lv->vg->name, lv->name, NULL))) {
+	if (!(name = build_dm_name(dm->mem, lv->vg->name, lv->name, NULL))) {
 		stack;
 		return 0;
 	}
@@ -1116,7 +1002,7 @@ int dev_manager_snapshot_percent(struct dev_manager *dm,
 	/*
 	 * Build a name for the top layer.
 	 */
-	if (!(name = _build_name(dm->mem, lv->vg->name, lv->name, NULL))) {
+	if (!(name = build_dm_name(dm->mem, lv->vg->name, lv->name, NULL))) {
 		stack;
 		return 0;
 	}
@@ -1148,7 +1034,7 @@ int dev_manager_mirror_percent(struct dev_manager *dm,
 	/*
 	 * Build a name for the top layer.
 	 */
-	if (!(name = _build_name(dm->mem, lv->vg->name, lv->name, NULL))) {
+	if (!(name = build_dm_name(dm->mem, lv->vg->name, lv->name, NULL))) {
 		stack;
 		return 0;
 	}
@@ -1212,7 +1098,7 @@ static struct dev_layer *_create_layer(struct dev_manager *dm,
 	char *name, *dlid;
 	struct dev_layer *dl;
 
-	if (!(name = _build_name(dm->mem, lv->vg->name, lv->name, layer))) {
+	if (!(name = build_dm_name(dm->mem, lv->vg->name, lv->name, layer))) {
 		stack;
 		return NULL;
 	}
@@ -1282,15 +1168,13 @@ static int _expand_vanilla(struct dev_manager *dm, struct logical_volume *lv,
 	/* Add dependencies for any LVs that segments refer to */
 	list_iterate(segh, &lv->segments) {
 		seg = list_item(segh, struct lv_segment);
-		if (seg->type != SEG_STRIPED && seg->type != SEG_MIRRORED)
-			continue;
 		for (s = 0; s < seg->area_count; s++) {
 			if (seg->area[s].type != AREA_LV)
 				continue;
 			if (!str_list_add(dm->mem, &dl->pre_create,
 					  _build_dlid(dm->mem,
-						      seg->area[s].u.lv.
-						      lv->lvid.s, NULL))) {
+						      seg->area[s].u.lv.lv->
+						      lvid.s, NULL))) {
 				stack;
 				return 0;
 			}
@@ -1311,7 +1195,7 @@ static int _expand_vanilla(struct dev_manager *dm, struct logical_volume *lv,
 	_clear_flag(dlr, VISIBLE);
 	_clear_flag(dlr, TOPLEVEL);
 	_set_flag(dlr, REMOVE);
-	
+
 	/* add the dependency on the real device */
 	if (!str_list_add(dm->mem, &dl->pre_create,
 			  pool_strdup(dm->mem, dlr->dlid))) {
@@ -1336,6 +1220,9 @@ static int _expand_origin_real(struct dev_manager *dm,
 	_clear_flag(dl, VISIBLE);
 	_clear_flag(dl, TOPLEVEL);
 
+	/* Size changes must take effect before tables using it are reloaded */
+	_set_flag(dl, RESUME_IMMEDIATE);
+
 	real_dlid = dl->dlid;
 
 	if (!(dl = _create_layer(dm, NULL, lv))) {
@@ -1656,11 +1543,11 @@ static int _create_rec(struct dev_manager *dm, struct dev_layer *dl)
 	if (dl->info.exists) {
 		if ((suffix = rindex(dl->dlid, '-')))
 			suffix++;
-		newname = _build_name(dm->mem, dm->vg_name, dl->lv->name,
-				      suffix);
+		newname = build_dm_name(dm->mem, dm->vg_name, dl->lv->name,
+					suffix);
 		if (strcmp(newname, dl->name)) {
 			if (!_suspend_parents(dm, dl) ||
-			    !_suspend(dl) || !_rename(dl, newname)) {
+			    !_suspend(dl) || !_rename(dm, dl, newname)) {
 				stack;
 				return 0;
 			}
@@ -1688,7 +1575,6 @@ static int _create_rec(struct dev_manager *dm, struct dev_layer *dl)
 	return 1;
 }
 
-
 static int _build_all_layers(struct dev_manager *dm, struct volume_group *vg)
 {
 	struct list *lvh;
@@ -1756,8 +1642,7 @@ static int _populate_pre_suspend_lists(struct dev_manager *dm)
 				continue;
 			}
 
-			if (!str_list_add(dm->mem, &dep->pre_create,
-					  dl->dlid)) {
+			if (!str_list_add(dm->mem, &dep->pre_create, dl->dlid)) {
 				stack;
 				return 0;
 			}
@@ -1773,8 +1658,7 @@ static int _populate_pre_suspend_lists(struct dev_manager *dm)
 				continue;
 			}
 
-			if (!str_list_add(dm->mem, &dep->pre_suspend,
-					  dl->dlid)) {
+			if (!str_list_add(dm->mem, &dep->pre_suspend, dl->dlid)) {
 				stack;
 				return 0;
 			}
@@ -2086,6 +1970,73 @@ static int _remove_suspended_lvs(struct dev_manager *dm,
 	return 1;
 }
 
+static int _targets_present(struct dev_manager *dm, struct list *lvs)
+{
+	struct logical_volume *lv;
+	struct list *lvh, *segh;
+	struct segment_type *segtype;
+	struct lv_segment *seg;
+	int snapshots = 0, mirrors = 0;
+
+	list_iterate(lvh, lvs) {
+		lv = list_item(lvh, struct lv_list)->lv;
+
+		if (!snapshots)
+			if (lv_is_cow(lv) || lv_is_origin(lv))
+				snapshots = 1;
+
+		if (!mirrors)
+			if (lv->status & PVMOVE)
+				mirrors = 1;
+
+		if (lv->status & VIRTUAL) {
+			list_iterate(segh, &lv->segments) {
+				seg = list_item(segh, struct lv_segment);
+				if (seg->segtype->ops->target_present &&
+				    !seg->segtype->ops->target_present()) {
+					log_error("Can't expand LV: %s target "
+						  "support missing "
+						  "from kernel?",
+						  seg->segtype->name);
+					return 0;
+				}
+			}
+		}
+	}
+
+	if (mirrors) {
+		if (!(segtype = get_segtype_from_string(dm->cmd, "mirror"))) {
+			log_error("Can't expand LV: Mirror support "
+				  "missing from tools?");
+			return 0;
+		}
+
+		if (!segtype->ops->target_present ||
+		    !segtype->ops->target_present()) {
+			log_error("Can't expand LV: Mirror support missing "
+				  "from kernel?");
+			return 0;
+		}
+	}
+
+	if (snapshots) {
+		if (!(segtype = get_segtype_from_string(dm->cmd, "snapshot"))) {
+			log_error("Can't expand LV: Snapshot support "
+				  "missing from tools?");
+			return 0;
+		}
+
+		if (!segtype->ops->target_present ||
+		    !segtype->ops->target_present()) {
+			log_error("Can't expand LV: Snapshot support missing "
+				  "from kernel?");
+			return 0;
+		}
+	}
+
+	return 1;
+}
+
 static int _fill_in_active_list(struct dev_manager *dm, struct volume_group *vg)
 {
 	char *dlid;
@@ -2168,6 +2119,12 @@ static int _action(struct dev_manager *dm, struct logical_volume *lv,
 		}
 	}
 
+	if (!_targets_present(dm, &dm->active_list) ||
+	    !_targets_present(dm, &dm->reload_list)) {
+		stack;
+		return 0;
+	}
+
 	if (!_execute(dm, lv->vg)) {
 		stack;
 		return 0;
@@ -2195,8 +2152,8 @@ int dev_manager_lv_mknodes(const struct logical_volume *lv)
 {
 	char *name;
 
-	if (!(name = _build_name(lv->vg->cmd->mem, lv->vg->name,
-				 lv->name, NULL))) {
+	if (!(name = build_dm_name(lv->vg->cmd->mem, lv->vg->name,
+				   lv->name, NULL))) {
 		stack;
 		return 0;
 	}

lib/activate/dev_manager.h

@@ -16,17 +16,16 @@
 #ifndef _LVM_DEV_MANAGER_H
 #define _LVM_DEV_MANAGER_H
 
-#include "metadata.h"
-#include "config.h"
-
+struct logical_volume;
+struct cmd_context;
 struct dev_manager;
 struct dm_info;
 
 /*
  * Constructor and destructor.
  */
-struct dev_manager *dev_manager_create(const char *vg_name,
-				       struct config_tree *cf);
+struct dev_manager *dev_manager_create(struct cmd_context *cmd,
+				       const char *vg_name);
 void dev_manager_destroy(struct dev_manager *dm);
 void dev_manager_exit(void);
 

lib/activate/fs.c

@@ -46,7 +46,7 @@ static int _mk_dir(const char *dev_dir, const char *vg_name)
 		return 1;
 
 	log_very_verbose("Creating directory %s", vg_path);
-	if (mkdir(vg_path, 0555)) {
+	if (mkdir(vg_path, 0777)) {
 		log_sys_error("mkdir", vg_path);
 		return 0;
 	}
@@ -65,10 +65,10 @@ static int _rm_dir(const char *dev_dir, const char *vg_name)
 		return 0;
 	}
 
-	log_very_verbose("Removing directory %s", vg_path);
-
-	if (is_empty_dir(vg_path))
+	if (is_empty_dir(vg_path)) {
+		log_very_verbose("Removing directory %s", vg_path);
 		rmdir(vg_path);
+	}
 
 	return 1;
 }
@@ -181,7 +181,7 @@ static int _mk_link(const char *dev_dir, const char *vg_name,
 
 #ifdef HAVE_SELINUX
         if (!set_selinux_context(lv_path)) {
-                log_sys_error("set_selinux_context", lv_path);
+                stack;
                 return 0;
         }
 #endif
@@ -202,9 +202,9 @@ static int _rm_link(const char *dev_dir, const char *vg_name,
 	}
 
 	if (lstat(lv_path, &buf) || !S_ISLNK(buf.st_mode)) {
-		if (errno != ENOENT)
-			log_error("%s not symbolic link - not removing",
-				  lv_path);
+		if (errno == ENOENT)
+			return 1;
+		log_error("%s not symbolic link - not removing", lv_path);
 		return 0;
 	}
 

lib/activate/targets.h

@@ -0,0 +1,25 @@
+/*
+ * Copyright (C) 2003-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU General Public License v.2.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#ifndef _LVM_TARGETS_H
+#define _LVM_TARGETS_H
+
+struct dev_manager;
+struct lv_segment;
+
+int compose_areas_line(struct dev_manager *dm, struct lv_segment *seg,                                 char *params, size_t paramsize, int *pos,
+		                        int start_area, int areas);
+
+#endif

lib/cache/lvmcache.h

@@ -20,7 +20,6 @@
 #include "dev-cache.h"
 #include "uuid.h"
 #include "label.h"
-#include "metadata.h"
 
 #define ORPHAN ""
 
@@ -30,6 +29,10 @@
 /* LVM specific per-volume info */
 /* Eventual replacement for struct physical_volume perhaps? */
 
+struct cmd_context;
+struct format_type;
+struct volume_group;
+
 struct lvmcache_vginfo {
 	struct list list;	/* Join these vginfos together */
 	struct list infos;	/* List head for lvmcache_infos */

lib/commands/toolcontext.c

@@ -33,6 +33,8 @@
 #include "display.h"
 #include "memlock.h"
 #include "str_list.h"
+#include "segtype.h"
+#include "lvmcache.h"
 
 #ifdef HAVE_LIBDL
 #include "sharedlib.h"
@@ -42,6 +44,10 @@
 #include "format1.h"
 #endif
 
+#ifdef POOL_INTERNAL
+#include "format_pool.h"
+#endif
+
 #include <locale.h>
 #include <sys/stat.h>
 #include <sys/utsname.h>
@@ -116,8 +122,7 @@ static void _init_logging(struct cmd_context *cmd)
 	    find_config_int(cmd->cft->root, "global/test", 0);
 
 	/* Settings for logging to file */
-	if (find_config_int(cmd->cft->root, "log/overwrite",
-			    DEFAULT_OVERWRITE))
+	if (find_config_int(cmd->cft->root, "log/overwrite", DEFAULT_OVERWRITE))
 		append = 0;
 
 	log_file = find_config_str(cmd->cft->root, "log/file", 0);
@@ -198,51 +203,260 @@ static int _process_config(struct cmd_context *cmd)
 	return 1;
 }
 
-/* Find and read config file */
-static int _init_config(struct cmd_context *cmd)
+static int _set_tag(struct cmd_context *cmd, const char *tag)
 {
-	struct stat info;
-	char config_file[PATH_MAX] = "";
+	log_very_verbose("Setting host tag: %s", pool_strdup(cmd->libmem, tag));
 
-	if (!(cmd->cft = create_config_tree())) {
-		stack;
-		return 0;
-	}
-
-	/* No config file if LVM_SYSTEM_DIR is empty */
-	if (!*cmd->sys_dir)
-		return 1;
-
-	if (lvm_snprintf(config_file, sizeof(config_file),
-			 "%s/lvm.conf", cmd->sys_dir) < 0) {
-		log_error("LVM_SYSTEM_DIR was too long");
-		destroy_config_tree(cmd->cft);
-		return 0;
-	}
-
-	/* Is there a config file? */
-	if (stat(config_file, &info) == -1) {
-		if (errno == ENOENT)
-			return 1;
-		log_sys_error("stat", config_file);
-		destroy_config_tree(cmd->cft);
-		return 0;
-	}
-
-	if (!read_config_file(cmd->cft, config_file)) {
-		log_error("Failed to load config file %s", config_file);
-		destroy_config_tree(cmd->cft);
+	if (!str_list_add(cmd->libmem, &cmd->tags, tag)) {
+		log_error("_set_tag: str_list_add %s failed", tag);
 		return 0;
 	}
 
 	return 1;
 }
 
-static int _init_dev_cache(struct cmd_context *cmd)
+static int _check_host_filters(struct cmd_context *cmd, struct config_node *hn,
+			       int *passes)
 {
 	struct config_node *cn;
 	struct config_value *cv;
 
+	*passes = 1;
+
+	for (cn = hn; cn; cn = cn->sib) {
+		if (!cn->v)
+			continue;
+		if (!strcmp(cn->key, "host_list")) {
+			*passes = 0;
+			if (cn->v->type == CFG_EMPTY_ARRAY)
+				continue;
+			for (cv = cn->v; cv; cv = cv->next) {
+				if (cv->type != CFG_STRING) {
+					log_error("Invalid hostname string "
+						  "for tag %s", cn->key);
+					return 0;
+				}
+				if (!strcmp(cv->v.str, cmd->hostname)) {
+					*passes = 1;
+					return 1;
+				}
+			}
+		}
+		if (!strcmp(cn->key, "host_filter")) {
+			log_error("host_filter not supported yet");
+			return 0;
+		}
+	}
+
+	return 1;
+}
+
+static int _init_tags(struct cmd_context *cmd, struct config_tree *cft)
+{
+	const struct config_node *tn, *cn;
+	const char *tag;
+	int passes;
+
+	if (!(tn = find_config_node(cft->root, "tags")) || !tn->child)
+		return 1;
+
+	/* NB hosttags 0 when already 1 intentionally does not delete the tag */
+	if (!cmd->hosttags && find_config_int(cft->root, "tags/hosttags",
+					      DEFAULT_HOSTTAGS)) {
+		/* FIXME Strip out invalid chars: only A-Za-z0-9_+.- */
+		if (!_set_tag(cmd, cmd->hostname)) {
+			stack;
+			return 0;
+		}
+		cmd->hosttags = 1;
+	}
+
+	for (cn = tn->child; cn; cn = cn->sib) {
+		if (cn->v)
+			continue;
+		tag = cn->key;
+		if (*tag == '@')
+			tag++;
+		if (!validate_name(tag)) {
+			log_error("Invalid tag in config file: %s", cn->key);
+			return 0;
+		}
+		if (cn->child) {
+			passes = 0;
+			if (!_check_host_filters(cmd, cn->child, &passes)) {
+				stack;
+				return 0;
+			}
+			if (!passes)
+				continue;
+		}
+		if (!_set_tag(cmd, tag)) {
+			stack;
+			return 0;
+		}
+	}
+
+	return 1;
+}
+
+static int _load_config_file(struct cmd_context *cmd, const char *tag)
+{
+	char config_file[PATH_MAX] = "";
+	const char *filler = "";
+	struct stat info;
+	struct config_tree_list *cfl;
+
+	if (*tag)
+		filler = "_";
+
+	if (lvm_snprintf(config_file, sizeof(config_file), "%s/lvm%s%s.conf",
+			 cmd->sys_dir, filler, tag) < 0) {
+		log_error("LVM_SYSTEM_DIR or tag was too long");
+		return 0;
+	}
+
+	if (!(cfl = pool_alloc(cmd->libmem, sizeof(*cfl)))) {
+		log_error("config_tree_list allocation failed");
+		return 0;
+	}
+
+	if (!(cfl->cft = create_config_tree(config_file))) {
+		log_error("config_tree allocation failed");
+		return 0;
+	}
+
+	/* Is there a config file? */
+	if (stat(config_file, &info) == -1) {
+		if (errno == ENOENT) {
+			list_add(&cmd->config_files, &cfl->list);
+			goto out;
+		}
+		log_sys_error("stat", config_file);
+		destroy_config_tree(cfl->cft);
+		return 0;
+	}
+
+	log_very_verbose("Loading config file: %s", config_file);
+	if (!read_config_file(cfl->cft)) {
+		log_error("Failed to load config file %s", config_file);
+		destroy_config_tree(cfl->cft);
+		return 0;
+	}
+
+	list_add(&cmd->config_files, &cfl->list);
+
+      out:
+	if (*tag)
+		_init_tags(cmd, cfl->cft);
+	else
+		/* Use temporary copy of lvm.conf while loading other files */
+		cmd->cft = cfl->cft;
+
+	return 1;
+}
+
+/* Find and read first config file */
+static int _init_lvm_conf(struct cmd_context *cmd)
+{
+	/* No config file if LVM_SYSTEM_DIR is empty */
+	if (!*cmd->sys_dir) {
+		if (!(cmd->cft = create_config_tree(NULL))) {
+			log_error("Failed to create config tree");
+			return 0;
+		}
+		return 1;
+	}
+
+	if (!_load_config_file(cmd, "")) {
+		stack;
+		return 0;
+	}
+
+	return 1;
+}
+
+/* Read any additional config files */
+static int _init_tag_configs(struct cmd_context *cmd)
+{
+	struct str_list *sl;
+
+	/* Tag list may grow while inside this loop */
+	list_iterate_items(sl, &cmd->tags) {
+		if (!_load_config_file(cmd, sl->str)) {
+			stack;
+			return 0;
+		}
+
+	}
+
+	return 1;
+}
+
+static int _merge_config_files(struct cmd_context *cmd)
+{
+	struct config_tree_list *cfl;
+
+	/* Replace temporary duplicate copy of lvm.conf */
+	if (cmd->cft->root) {
+		if (!(cmd->cft = create_config_tree(NULL))) {
+			log_error("Failed to create config tree");
+			return 0;
+		}
+	}
+
+	list_iterate_items(cfl, &cmd->config_files) {
+		/* Merge all config trees into cmd->cft using merge/tag rules */
+		if (!merge_config_tree(cmd, cmd->cft, cfl->cft)) {
+			stack;
+			return 0;
+		}
+	}
+
+	return 1;
+}
+
+static void _destroy_tags(struct cmd_context *cmd)
+{
+	struct list *slh, *slht;
+
+	list_iterate_safe(slh, slht, &cmd->tags) {
+		list_del(slh);
+	}
+}
+
+int config_files_changed(struct cmd_context *cmd)
+{
+	struct config_tree_list *cfl;
+
+	list_iterate_items(cfl, &cmd->config_files) {
+		if (config_file_changed(cfl->cft))
+			return 1;
+	}
+
+	return 0;
+}
+
+static void _destroy_tag_configs(struct cmd_context *cmd)
+{
+	struct config_tree_list *cfl;
+
+	if (cmd->cft && cmd->cft->root) {
+		destroy_config_tree(cmd->cft);
+		cmd->cft = NULL;
+	}
+
+	list_iterate_items(cfl, &cmd->config_files) {
+		destroy_config_tree(cfl->cft);
+	}
+
+	list_init(&cmd->config_files);
+}
+
+static int _init_dev_cache(struct cmd_context *cmd)
+{
+	const struct config_node *cn;
+	struct config_value *cv;
+
 	if (!dev_cache_init()) {
 		stack;
 		return 0;
@@ -281,7 +495,7 @@ static int _init_dev_cache(struct cmd_context *cmd)
 static struct dev_filter *_init_filter_components(struct cmd_context *cmd)
 {
 	unsigned nr_filt = 0;
-	struct config_node *cn;
+	const struct config_node *cn;
 	struct dev_filter *filters[MAX_FILTERS];
 
 	memset(filters, 0, sizeof(filters));
@@ -351,8 +565,8 @@ static int _init_filters(struct cmd_context *cmd)
 		return 0;
 	}
 
-	dev_cache =
-	    find_config_str(cmd->cft->root, "devices/cache", cache_file);
+	dev_cache = find_config_str(cmd->cft->root, "devices/cache",
+				    cache_file);
 	if (!(f4 = persistent_filter_create(f3, dev_cache))) {
 		log_error("Failed to create persistent device filter");
 		return 0;
@@ -384,7 +598,7 @@ static int _init_formats(struct cmd_context *cmd)
 	struct list *fmth;
 
 #ifdef HAVE_LIBDL
-	struct config_node *cn;
+	const struct config_node *cn;
 #endif
 
 	label_init();
@@ -396,6 +610,13 @@ static int _init_formats(struct cmd_context *cmd)
 	list_add(&cmd->formats, &fmt->list);
 #endif
 
+#ifdef POOL_INTERNAL
+	if (!(fmt = init_pool_format(cmd)))
+		return 0;
+	fmt->library = NULL;
+	list_add(&cmd->formats, &fmt->list);
+#endif
+
 #ifdef HAVE_LIBDL
 	/* Load any formats in shared libs */
 	if ((cn = find_config_node(cmd->cft->root, "global/format_libraries"))) {
@@ -454,6 +675,97 @@ static int _init_formats(struct cmd_context *cmd)
 	return 0;
 }
 
+static int _init_segtypes(struct cmd_context *cmd)
+{
+	struct segment_type *segtype;
+
+#ifdef HAVE_LIBDL
+	const struct config_node *cn;
+#endif
+
+	if (!(segtype = init_striped_segtype(cmd)))
+		return 0;
+	segtype->library = NULL;
+	list_add(&cmd->segtypes, &segtype->list);
+
+	if (!(segtype = init_zero_segtype(cmd)))
+		return 0;
+	segtype->library = NULL;
+	list_add(&cmd->segtypes, &segtype->list);
+
+	if (!(segtype = init_error_segtype(cmd)))
+		return 0;
+	segtype->library = NULL;
+	list_add(&cmd->segtypes, &segtype->list);
+
+#ifdef SNAPSHOT_INTERNAL
+	if (!(segtype = init_snapshot_segtype(cmd)))
+		return 0;
+	segtype->library = NULL;
+	list_add(&cmd->segtypes, &segtype->list);
+#endif
+
+#ifdef MIRRORED_INTERNAL
+	if (!(segtype = init_mirrored_segtype(cmd)))
+		return 0;
+	segtype->library = NULL;
+	list_add(&cmd->segtypes, &segtype->list);
+#endif
+
+#ifdef HAVE_LIBDL
+	/* Load any formats in shared libs */
+	if ((cn = find_config_node(cmd->cft->root, "global/segment_libraries"))) {
+
+		struct config_value *cv;
+		struct segment_type *(*init_segtype_fn) (struct cmd_context *);
+		void *lib;
+		struct list *sgtl, *tmp;
+		struct segment_type *segtype2;
+
+		for (cv = cn->v; cv; cv = cv->next) {
+			if (cv->type != CFG_STRING) {
+				log_error("Invalid string in config file: "
+					  "global/segment_libraries");
+				return 0;
+			}
+			if (!(lib = load_shared_library(cmd->cft, cv->v.str,
+							"segment type"))) {
+				stack;
+				return 0;
+			}
+
+			if (!(init_segtype_fn = dlsym(lib, "init_segtype"))) {
+				log_error("Shared library %s does not contain "
+					  "segment type functions", cv->v.str);
+				dlclose(lib);
+				return 0;
+			}
+
+			if (!(segtype = init_segtype_fn(cmd)))
+				return 0;
+			segtype->library = lib;
+			list_add(&cmd->segtypes, &segtype->list);
+
+			list_iterate_safe(sgtl, tmp, &cmd->segtypes) {
+				segtype2 = list_item(sgtl, struct segment_type);
+				if (!strcmp(segtype2->name, segtype->name)) {
+					log_error("Duplicate segment type %s: "
+						  "unloading shared library %s",
+						  segtype->name, cv->v.str);
+					list_del(&segtype->list);
+					segtype->ops->destroy(segtype);
+					dlclose(lib);
+					break;
+				}
+
+			}
+		}
+	}
+#endif
+
+	return 1;
+}
+
 static int _init_hostname(struct cmd_context *cmd)
 {
 	struct utsname uts;
@@ -476,105 +788,6 @@ static int _init_hostname(struct cmd_context *cmd)
 	return 1;
 }
 
-static int _set_tag(struct cmd_context *cmd, const char *tag)
-{
-	log_very_verbose("Setting host tag: %s", pool_strdup(cmd->libmem, tag));
-
-	if (!str_list_add(cmd->libmem, &cmd->tags, tag)) {
-		log_error("_init_tags: str_list_add %s failed", tag);
-		return 0;
-	}
-
-	return 1;
-}
-
-static int _check_host_filters(struct cmd_context *cmd, struct config_node *hn,
-			       int *passes)
-{
-	struct config_node *cn;
-	struct config_value *cv;
-
-	*passes = 1;
-
-	for (cn = hn; cn; cn = cn->sib) {
-		if (!cn->v)
-			continue;
-		if (!strcmp(cn->key, "host_list")) {
-			*passes = 0;
-			if (cn->v->type == CFG_EMPTY_ARRAY)
-				continue;
-			for (cv = cn->v; cv; cv = cv->next) {
-				if (cv->type != CFG_STRING) {
-					log_error("Invalid hostname string "
-						  "for tag %s", cn->key);
-					return 0;
-				}
-				if (!strcmp(cv->v.str, cmd->hostname)) {
-					*passes = 1;
-					return 1;
-				}
-			}
-		}
-		if (!strcmp(cn->key, "host_filter")) {
-			log_error("host_filter not supported yet");
-			return 0;
-		}
-	}
-
-	return 1;
-}
-	
-static int _init_tags(struct cmd_context *cmd)
-{
-	struct config_node *tn, *cn;
-	const char *tag;
-	int passes;
-
-	list_init(&cmd->tags);
-
-	if (!(tn = find_config_node(cmd->cft->root, "tags")) ||
-	    !tn->child) {
-		log_very_verbose("No tags defined in config file");
-		return 1;
-	}
-
-	if (find_config_int(cmd->cft->root, "tags/hosttags",
-			    DEFAULT_HOSTTAGS)) {
-		/* FIXME Strip out invalid chars: only A-Za-z0-9_+.- */
-		if (!_set_tag(cmd, cmd->hostname)) {
-			stack;
-			return 0;
-		}
-	}
-
-	for (cn = tn->child; cn; cn = cn->sib) {
-		if (cn->v)
-			continue;
-		tag = cn->key;
-		if (*tag == '@')
-			tag++;
-		if (!validate_name(tag)) {
-			log_error("Invalid tag in config file: %s", cn->key);
-			return 0;
-		}
-		if (cn->child) {
-			passes = 0;
-			if (!_check_host_filters(cmd, cn->child, &passes)) {
-				stack;
-				return 0;
-			}
-			if (!passes)
-				continue;
-		}
-		if (!_set_tag(cmd, tag)) {
-			stack;
-			return 0;
-		}
-	}
-
-	return 1;
-}
-
 /* Entry point */
 struct cmd_context *create_toolcontext(struct arg *the_args)
 {
@@ -585,7 +798,7 @@ struct cmd_context *create_toolcontext(struct arg *the_args)
 #endif
 
 	if (!setlocale(LC_ALL, ""))
-		log_error("setlocale failed");
+		log_very_verbose("setlocale failed");
 
 #ifdef INTL_PACKAGE
 	bindtextdomain(INTL_PACKAGE, LOCALEDIR);
@@ -599,7 +812,11 @@ struct cmd_context *create_toolcontext(struct arg *the_args)
 	}
 	memset(cmd, 0, sizeof(*cmd));
 	cmd->args = the_args;
+	cmd->hosttags = 0;
 	list_init(&cmd->formats);
+	list_init(&cmd->segtypes);
+	list_init(&cmd->tags);
+	list_init(&cmd->config_files);
 
 	strcpy(cmd->sys_dir, DEFAULT_SYS_DIR);
 
@@ -610,16 +827,28 @@ struct cmd_context *create_toolcontext(struct arg *the_args)
 	if (*cmd->sys_dir && !create_dir(cmd->sys_dir))
 		goto error;
 
-	if (!_init_config(cmd))
-		goto error;
-
-	_init_logging(cmd);
-
 	if (!(cmd->libmem = pool_create(4 * 1024))) {
 		log_error("Library memory pool creation failed");
 		return 0;
 	}
 
+	if (!_init_lvm_conf(cmd))
+		goto error;
+
+	_init_logging(cmd);
+
+	if (!_init_hostname(cmd))
+		goto error;
+
+	if (!_init_tags(cmd, cmd->cft))
+		goto error;
+
+	if (!_init_tag_configs(cmd))
+		goto error;
+
+	if (!_merge_config_files(cmd))
+		goto error;
+
 	if (!_process_config(cmd))
 		goto error;
 
@@ -639,14 +868,12 @@ struct cmd_context *create_toolcontext(struct arg *the_args)
 	if (!_init_formats(cmd))
 		goto error;
 
-	if (!_init_hostname(cmd))
-		goto error;
-
-	if (!_init_tags(cmd))
+	if (!_init_segtypes(cmd))
 		goto error;
 
 	cmd->current_settings = cmd->default_settings;
 
+	cmd->config_valid = 1;
 	return cmd;
 
       error:
@@ -654,16 +881,6 @@ struct cmd_context *create_toolcontext(struct arg *the_args)
 	return NULL;
 }
 
-int refresh_toolcontext(struct cmd_context *cmd)
-{
-	_init_logging(cmd);
-	_init_tags(cmd);
-
-	/* FIXME Reset filters and dev_cache */
-
-	return 1;
-}
-
 static void _destroy_formats(struct list *formats)
 {
 	struct list *fmtl, *tmp;
@@ -682,6 +899,83 @@ static void _destroy_formats(struct list *formats)
 	}
 }
 
+static void _destroy_segtypes(struct list *segtypes)
+{
+	struct list *sgtl, *tmp;
+	struct segment_type *segtype;
+	void *lib;
+
+	list_iterate_safe(sgtl, tmp, segtypes) {
+		segtype = list_item(sgtl, struct segment_type);
+		list_del(&segtype->list);
+		lib = segtype->library;
+		segtype->ops->destroy(segtype);
+#ifdef HAVE_LIBDL
+		if (lib)
+			dlclose(lib);
+#endif
+	}
+}
+
+int refresh_toolcontext(struct cmd_context *cmd)
+{
+	log_verbose("Reloading config files");
+
+	if (cmd->config_valid) {
+		if (cmd->dump_filter)
+			persistent_filter_dump(cmd->filter);
+	}
+
+	activation_exit();
+	lvmcache_destroy();
+	label_exit();
+	_destroy_segtypes(&cmd->segtypes);
+	_destroy_formats(&cmd->formats);
+	if (cmd->filter) {
+		cmd->filter->destroy(cmd->filter);
+		cmd->filter = NULL;
+	}
+	dev_cache_exit();
+	_destroy_tags(cmd);
+	_destroy_tag_configs(cmd);
+
+	cmd->config_valid = 0;
+
+	cmd->hosttags = 0;
+
+	if (!_init_lvm_conf(cmd))
+		return 0;
+
+	_init_logging(cmd);
+
+	if (!_init_tags(cmd, cmd->cft))
+		return 0;
+
+	if (!_init_tag_configs(cmd))
+		return 0;
+
+	if (!_merge_config_files(cmd))
+		return 0;
+
+	if (!_process_config(cmd))
+		return 0;
+
+	if (!_init_dev_cache(cmd))
+		return 0;
+
+	if (!_init_filters(cmd))
+		return 0;
+
+	if (!_init_formats(cmd))
+		return 0;
+
+	if (!_init_segtypes(cmd))
+		return 0;
+
+	cmd->config_valid = 1;
+	return 1;
+}
+
 void destroy_toolcontext(struct cmd_context *cmd)
 {
 	if (cmd->dump_filter)
@@ -690,11 +984,13 @@ void destroy_toolcontext(struct cmd_context *cmd)
 	activation_exit();
 	lvmcache_destroy();
 	label_exit();
+	_destroy_segtypes(&cmd->segtypes);
 	_destroy_formats(&cmd->formats);
 	cmd->filter->destroy(cmd->filter);
 	pool_destroy(cmd->mem);
 	dev_cache_exit();
-	destroy_config_tree(cmd->cft);
+	_destroy_tags(cmd);
+	_destroy_tag_configs(cmd);
 	pool_destroy(cmd->libmem);
 	dbg_free(cmd);
 

lib/commands/toolcontext.h

@@ -17,9 +17,7 @@
 #define _LVM_TOOLCONTEXT_H
 
 #include "dev-cache.h"
-#include "config.h"
 #include "pool.h"
-#include "metadata.h"
 
 #include <stdio.h>
 #include <limits.h>
@@ -47,6 +45,8 @@ struct config_info {
 	mode_t umask;
 };
 
+struct config_tree;
+
 /* FIXME Split into tool & library contexts */
 /* command-instance-related variables needed by library */
 struct cmd_context {
@@ -57,6 +57,7 @@ struct cmd_context {
 	struct format_type *fmt_backup;	/* Format to use for backups */
 
 	struct list formats;	/* Available formats */
+	struct list segtypes;	/* Available segment types */
 	const char *hostname;
 	const char *kernel_vsn;
 
@@ -68,12 +69,15 @@ struct cmd_context {
 	struct dev_filter *filter;
 	int dump_filter;	/* Dump filter when exiting? */
 
+	struct list config_files;
+	int config_valid;
 	struct config_tree *cft;
 	struct config_info default_settings;
 	struct config_info current_settings;
 
 	/* List of defined tags */
 	struct list tags;
+	int hosttags;
 
 	char sys_dir[PATH_MAX];
 	char dev_dir[PATH_MAX];
@@ -83,5 +87,6 @@ struct cmd_context {
 struct cmd_context *create_toolcontext(struct arg *the_args);
 void destroy_toolcontext(struct cmd_context *cmd);
 int refresh_toolcontext(struct cmd_context *cmd);
+int config_files_changed(struct cmd_context *cmd);
 
 #endif

lib/config/config.c

@@ -18,6 +18,8 @@
 #include "crc.h"
 #include "pool.h"
 #include "device.h"
+#include "str_list.h"
+#include "toolcontext.h"
 
 #include <sys/stat.h>
 #include <sys/mman.h>
@@ -56,6 +58,7 @@ struct cs {
 	struct pool *mem;
 	time_t timestamp;
 	char *filename;
+	int exists;
 };
 
 static void _get_token(struct parser *p, int tok_prev);
@@ -93,7 +96,7 @@ static int _tok_match(const char *str, const char *b, const char *e)
 /*
  * public interface
  */
-struct config_tree *create_config_tree(void)
+struct config_tree *create_config_tree(const char *filename)
 {
 	struct cs *c;
 	struct pool *mem = pool_create(10 * 1024);
@@ -103,7 +106,7 @@ struct config_tree *create_config_tree(void)
 		return 0;
 	}
 
-	if (!(c = pool_alloc(mem, sizeof(*c)))) {
+	if (!(c = pool_zalloc(mem, sizeof(*c)))) {
 		stack;
 		pool_destroy(mem);
 		return 0;
@@ -112,7 +115,9 @@ struct config_tree *create_config_tree(void)
 	c->mem = mem;
 	c->cft.root = (struct config_node *) NULL;
 	c->timestamp = 0;
-	c->filename = NULL;
+	c->exists = 0;
+	if (filename)
+		c->filename = pool_strdup(c->mem, filename);
 	return &c->cft;
 }
 
@@ -204,29 +209,33 @@ int read_config_fd(struct config_tree *cft, struct device *dev,
 	return r;
 }
 
-int read_config_file(struct config_tree *cft, const char *file)
+int read_config_file(struct config_tree *cft)
 {
 	struct cs *c = (struct cs *) cft;
 	struct stat info;
 	struct device *dev;
 	int r = 1;
 
-	if (stat(file, &info)) {
-		log_sys_error("stat", file);
+	if (stat(c->filename, &info)) {
+		log_sys_error("stat", c->filename);
+		c->exists = 0;
 		return 0;
 	}
 
 	if (!S_ISREG(info.st_mode)) {
-		log_error("%s is not a regular file", file);
+		log_error("%s is not a regular file", c->filename);
+		c->exists = 0;
 		return 0;
 	}
 
+	c->exists = 1;
+
 	if (info.st_size == 0) {
-		log_verbose("%s is empty", file);
+		log_verbose("%s is empty", c->filename);
 		return 1;
 	}
 
-	if (!(dev = dev_create_file(file, NULL, NULL))) {
+	if (!(dev = dev_create_file(c->filename, NULL, NULL))) {
 		stack;
 		return 0;
 	}
@@ -242,7 +251,6 @@ int read_config_file(struct config_tree *cft, const char *file)
 	dev_close(dev);
 
 	c->timestamp = info.st_mtime;
-	c->filename = pool_strdup(c->mem, file);
 
 	return r;
 }
@@ -255,74 +263,43 @@ time_t config_file_timestamp(struct config_tree *cft)
 }
 
 /*
- * Returns 1 if config file reloaded
+ * Return 1 if config files ought to be reloaded
  */
-int reload_config_file(struct config_tree **cft)
+int config_file_changed(struct config_tree *cft)
 {
-	struct config_tree *new_cft;
-	struct cs *c = (struct cs *) *cft;
-	struct cs *new_cs;
+	struct cs *c = (struct cs *) cft;
 	struct stat info;
-	struct device *dev;
-	int r;
 
 	if (!c->filename)
 		return 0;
 
 	if (stat(c->filename, &info) == -1) {
-		if (errno == ENOENT)
-			return 1;
+		/* Ignore a deleted config file: still use original data */
+		if (errno == ENOENT) {
+			if (!c->exists)
+				return 0;
+			log_very_verbose("Config file %s has disappeared!",
+					 c->filename);
+			goto reload;
+		}
 		log_sys_error("stat", c->filename);
-		log_error("Failed to reload configuration file");
+		log_error("Failed to reload configuration files");
 		return 0;
 	}
 
 	if (!S_ISREG(info.st_mode)) {
 		log_error("Configuration file %s is not a regular file",
 			  c->filename);
-		return 0;
+		goto reload;
 	}
 
 	/* Unchanged? */
 	if (c->timestamp == info.st_mtime)
 		return 0;
 
-	log_verbose("Detected config file change: Reloading %s", c->filename);
-
-	if (info.st_size == 0) {
-		log_verbose("Config file reload: %s is empty", c->filename);
-		return 0;
-	}
-
-	if (!(new_cft = create_config_tree())) {
-		log_error("Allocation of new config_tree failed");
-		return 0;
-	}
-
-	if (!(dev = dev_create_file(c->filename, NULL, NULL))) {
-		stack;
-		return 0;
-	}
-
-	if (!dev_open_flags(dev, O_RDONLY, 0, 0)) {
-		stack;
-		return 0;
-	}
-
-	r = read_config_fd(new_cft, dev, 0, (size_t) info.st_size,
-			   0, 0, (checksum_fn_t) NULL, 0);
-
-	dev_close(dev);
-
-	if (r) {
-		new_cs = (struct cs *) new_cft;
-		new_cs->filename = pool_strdup(new_cs->mem, c->filename);
-		new_cs->timestamp = info.st_mtime;
-		destroy_config_tree(*cft);
-		*cft = new_cft;
-	}
-
-	return r;
+      reload:
+	log_verbose("Detected config file change to %s", c->filename);
+	return 1;
 }
 
 static void _write_value(FILE *fp, struct config_value *v)
@@ -739,7 +716,8 @@ static char *_dup_tok(struct parser *p)
 /*
  * utility functions
  */
-struct config_node *find_config_node(struct config_node *cn, const char *path)
+struct config_node *find_config_node(const struct config_node *cn,
+				     const char *path)
 {
 	const char *e;
 
@@ -767,13 +745,13 @@ struct config_node *find_config_node(struct config_node *cn, const char *path)
 		path = e;
 	}
 
-	return cn;
+	return (struct config_node *) cn;
 }
 
-const char *find_config_str(struct config_node *cn,
+const char *find_config_str(const struct config_node *cn,
 			    const char *path, const char *fail)
 {
-	struct config_node *n = find_config_node(cn, path);
+	const struct config_node *n = find_config_node(cn, path);
 
 	if (n && n->v->type == CFG_STRING) {
 		if (*n->v->v.str)
@@ -787,9 +765,9 @@ const char *find_config_str(struct config_node *cn,
 	return fail;
 }
 
-int find_config_int(struct config_node *cn, const char *path, int fail)
+int find_config_int(const struct config_node *cn, const char *path, int fail)
 {
-	struct config_node *n = find_config_node(cn, path);
+	const struct config_node *n = find_config_node(cn, path);
 
 	if (n && n->v->type == CFG_INT) {
 		log_very_verbose("Setting %s to %d", path, n->v->v.i);
@@ -801,9 +779,10 @@ int find_config_int(struct config_node *cn, const char *path, int fail)
 	return fail;
 }
 
-float find_config_float(struct config_node *cn, const char *path, float fail)
+float find_config_float(const struct config_node *cn, const char *path,
+			float fail)
 {
-	struct config_node *n = find_config_node(cn, path);
+	const struct config_node *n = find_config_node(cn, path);
 
 	if (n && n->v->type == CFG_FLOAT) {
 		log_very_verbose("Setting %s to %f", path, n->v->v.r);
@@ -843,9 +822,9 @@ static int _str_to_bool(const char *str, int fail)
 	return fail;
 }
 
-int find_config_bool(struct config_node *cn, const char *path, int fail)
+int find_config_bool(const struct config_node *cn, const char *path, int fail)
 {
-	struct config_node *n = find_config_node(cn, path);
+	const struct config_node *n = find_config_node(cn, path);
 	struct config_value *v;
 
 	if (!n)
@@ -864,10 +843,10 @@ int find_config_bool(struct config_node *cn, const char *path, int fail)
 	return fail;
 }
 
-int get_config_uint32(struct config_node *cn, const char *path,
+int get_config_uint32(const struct config_node *cn, const char *path,
 		      uint32_t *result)
 {
-	struct config_node *n;
+	const struct config_node *n;
 
 	n = find_config_node(cn, path);
 
@@ -878,10 +857,10 @@ int get_config_uint32(struct config_node *cn, const char *path,
 	return 1;
 }
 
-int get_config_uint64(struct config_node *cn, const char *path,
+int get_config_uint64(const struct config_node *cn, const char *path,
 		      uint64_t *result)
 {
-	struct config_node *n;
+	const struct config_node *n;
 
 	n = find_config_node(cn, path);
 
@@ -893,9 +872,10 @@ int get_config_uint64(struct config_node *cn, const char *path,
 	return 1;
 }
 
-int get_config_str(struct config_node *cn, const char *path, char **result)
+int get_config_str(const struct config_node *cn, const char *path,
+		   char **result)
 {
-	struct config_node *n;
+	const struct config_node *n;
 
 	n = find_config_node(cn, path);
 
@@ -905,3 +885,115 @@ int get_config_str(struct config_node *cn, const char *path, char **result)
 	*result = n->v->v.str;
 	return 1;
 }
+
+/* Insert cn2 after cn1 */
+static void _insert_config_node(struct config_node **cn1,
+				struct config_node *cn2)
+{
+	if (!*cn1) {
+		*cn1 = cn2;
+		cn2->sib = NULL;
+	} else {
+		cn2->sib = (*cn1)->sib;
+		(*cn1)->sib = cn2;
+	}
+}
+
+/*
+ * Merge section cn2 into section cn1 (which has the same name)
+ * overwriting any existing cn1 nodes with matching names.
+ */
+static void _merge_section(struct config_node *cn1, struct config_node *cn2)
+{
+	struct config_node *cn, *nextn, *oldn;
+	struct config_value *cv;
+
+	for (cn = cn2->child; cn; cn = nextn) {
+		nextn = cn->sib;
+
+		/* Skip "tags" */
+		if (!strcmp(cn->key, "tags"))
+			continue;
+
+		/* Subsection? */
+		if (!cn->v)
+			/* Ignore - we don't have any of these yet */
+			continue;
+		/* Not already present? */
+		if (!(oldn = find_config_node(cn1->child, cn->key))) {
+			_insert_config_node(&cn1->child, cn);
+			continue;
+		}
+		/* Merge certain value lists */
+		if ((!strcmp(cn1->key, "activation") &&
+		     !strcmp(cn->key, "volume_list")) ||
+		    (!strcmp(cn1->key, "devices") &&
+		     (!strcmp(cn->key, "filter") || !strcmp(cn->key, "types")))) {
+			cv = cn->v;
+			while (cv->next)
+				cv = cv->next;
+			cv->next = oldn->v;
+		}
+
+		/* Replace values */
+		oldn->v = cn->v;
+	}
+}
+
+static int _match_host_tags(struct list *tags, struct config_node *tn)
+{
+	struct config_value *tv;
+	const char *str;
+
+	for (tv = tn->v; tv; tv = tv->next) {
+		if (tv->type != CFG_STRING)
+			continue;
+		str = tv->v.str;
+		if (*str == '@')
+			str++;
+		if (!*str)
+			continue;
+		if (str_list_match_item(tags, str))
+			return 1;
+	}
+
+	return 0;
+}
+
+/* Destructively merge a new config tree into an existing one */
+int merge_config_tree(struct cmd_context *cmd, struct config_tree *cft,
+		      struct config_tree *newdata)
+{
+	struct config_node *root = cft->root;
+	struct config_node *cn, *nextn, *oldn, *tn, *cn2;
+
+	for (cn = newdata->root; cn; cn = nextn) {
+		nextn = cn->sib;
+		/* Ignore tags section */
+		if (!strcmp(cn->key, "tags"))
+			continue;
+		/* If there's a tags node, skip if host tags don't match */
+		if ((tn = find_config_node(cn->child, "tags"))) {
+			if (!_match_host_tags(&cmd->tags, tn))
+				continue;
+		}
+		if (!(oldn = find_config_node(root, cn->key))) {
+			_insert_config_node(&cft->root, cn);
+			/* Remove any "tags" nodes */
+			for (cn2 = cn->child; cn2; cn2 = cn2->sib) {
+				if (!strcmp(cn2->key, "tags")) {
+					cn->child = cn2->sib;
+					continue;
+				}
+				if (cn2->sib && !strcmp(cn2->sib->key, "tags")) {
+					cn2->sib = cn2->sib->sib;
+					continue;
+				}
+			}
+			continue;
+		}
+		_merge_section(oldn, cn);
+	}
+
+	return 1;
+}

lib/config/config.h

@@ -16,7 +16,8 @@
 #ifndef _LVM_CONFIG_H
 #define _LVM_CONFIG_H
 
-#include "device.h"
+struct device;
+struct cmd_context;
 
 enum {
 	CFG_STRING,
@@ -45,42 +46,51 @@ struct config_tree {
 	struct config_node *root;
 };
 
-struct config_tree *create_config_tree(void);
-void destroy_config_tree(struct config_tree *cf);
+struct config_tree_list {
+	struct list list;
+	struct config_tree *cft;
+};
+
+struct config_tree *create_config_tree(const char *filename);
+void destroy_config_tree(struct config_tree *cft);
 
 typedef uint32_t (*checksum_fn_t) (uint32_t initial, void *buf, uint32_t size);
 
-int read_config_fd(struct config_tree *cf, struct device *dev,
+int read_config_fd(struct config_tree *cft, struct device *dev,
 		   off_t offset, size_t size, off_t offset2, size_t size2,
 		   checksum_fn_t checksum_fn, uint32_t checksum);
 
-int read_config_file(struct config_tree *cf, const char *file);
-int write_config_file(struct config_tree *cf, const char *file);
-int reload_config_file(struct config_tree **cf);
-time_t config_file_timestamp(struct config_tree *cf);
+int read_config_file(struct config_tree *cft);
+int write_config_file(struct config_tree *cft, const char *file);
+time_t config_file_timestamp(struct config_tree *cft);
+int config_file_changed(struct config_tree *cft);
+int merge_config_tree(struct cmd_context *cmd, struct config_tree *cft,
+		      struct config_tree *newdata);
 
-struct config_node *find_config_node(struct config_node *cn,
+struct config_node *find_config_node(const struct config_node *cn,
 				     const char *path);
 
-const char *find_config_str(struct config_node *cn, const char *path,
+const char *find_config_str(const struct config_node *cn, const char *path,
 			    const char *fail);
 
-int find_config_int(struct config_node *cn, const char *path, int fail);
+int find_config_int(const struct config_node *cn, const char *path, int fail);
 
-float find_config_float(struct config_node *cn, const char *path, float fail);
+float find_config_float(const struct config_node *cn, const char *path,
+			float fail);
 
 /*
  * Understands (0, ~0), (y, n), (yes, no), (on,
  * off), (true, false).
  */
-int find_config_bool(struct config_node *cn, const char *path, int fail);
+int find_config_bool(const struct config_node *cn, const char *path, int fail);
 
-int get_config_uint32(struct config_node *cn, const char *path,
+int get_config_uint32(const struct config_node *cn, const char *path,
 		      uint32_t *result);
 
-int get_config_uint64(struct config_node *cn, const char *path,
+int get_config_uint64(const struct config_node *cn, const char *path,
 		      uint64_t *result);
 
-int get_config_str(struct config_node *cn, const char *path, char **result);
+int get_config_str(const struct config_node *cn, const char *path,
+		   char **result);
 
 #endif

lib/config/defaults.h

@@ -91,14 +91,14 @@
 #define DEFAULT_REP_HEADINGS 1
 #define DEFAULT_REP_SEPARATOR " "
 
-#define DEFAULT_LVS_COLS "lv_name,vg_name,lv_attr,lv_size,origin,snap_percent,move_pv,move_percent"
+#define DEFAULT_LVS_COLS "lv_name,vg_name,lv_attr,lv_size,origin,snap_percent,move_pv,copy_percent"
 #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"
 #define DEFAULT_SEGS_COLS "lv_name,vg_name,lv_attr,stripes,segtype,seg_size"
 
-#define DEFAULT_LVS_COLS_VERB "lv_name,vg_name,seg_count,lv_attr,lv_size,lv_major,lv_minor,origin,snap_percent,move_pv,move_percent,lv_uuid"
+#define DEFAULT_LVS_COLS_VERB "lv_name,vg_name,seg_count,lv_attr,lv_size,lv_major,lv_minor,lv_kernel_major,lv_kernel_minor,origin,snap_percent,move_pv,copy_percent,lv_uuid"
 #define DEFAULT_VGS_COLS_VERB "vg_name,vg_attr,vg_extent_size,pv_count,lv_count,snap_count,vg_size,vg_free,vg_uuid"
-#define DEFAULT_PVS_COLS_VERB "pv_name,vg_name,pv_fmt,pv_attr,pv_size,pv_free,pv_uuid"
+#define DEFAULT_PVS_COLS_VERB "pv_name,vg_name,pv_fmt,pv_attr,pv_size,pv_free,dev_size,pv_uuid"
 #define DEFAULT_SEGS_COLS_VERB "lv_name,vg_name,lv_attr,seg_start,seg_size,stripes,segtype,stripesize,chunksize"
 
 #define DEFAULT_LVS_SORT "vg_name,lv_name"

lib/datastruct/hash.c

@@ -19,7 +19,8 @@
 struct hash_node {
 	struct hash_node *next;
 	void *data;
-	char key[1];
+	int keylen;
+	char key[0];
 };
 
 struct hash_table {
@@ -56,22 +57,23 @@ static unsigned char _nums[] = {
 	209
 };
 
-static struct hash_node *_create_node(const char *str)
+static struct hash_node *_create_node(const char *str, int len)
 {
-	/* remember sizeof(n) includes an extra char from key[1],
-	   so not adding 1 to the strlen as you would expect */
-	struct hash_node *n = dbg_malloc(sizeof(*n) + strlen(str));
+	struct hash_node *n = dbg_malloc(sizeof(*n) + len);
 
-	if (n)
-		strcpy(n->key, str);
+	if (n) {
+		memcpy(n->key, str, len);
+		n->keylen = len;
+	}
 
 	return n;
 }
 
-static unsigned _hash(const char *str)
+static unsigned _hash(const char *str, uint32_t len)
 {
-	unsigned long h = 0, g;
-	while (*str) {
+	unsigned long h = 0, g, i;
+
+	for (i = 0; i < len; i++) {
 		h <<= 4;
 		h += _nums[(int) *str++];
 		g = h & ((unsigned long) 0xf << 16u);
@@ -80,6 +82,7 @@ static unsigned _hash(const char *str)
 			h ^= g >> 5u;
 		}
 	}
+
 	return h;
 }
 
@@ -134,32 +137,35 @@ void hash_destroy(struct hash_table *t)
 	dbg_free(t);
 }
 
-static struct hash_node **_find(struct hash_table *t, const char *key)
+static inline struct hash_node **_find(struct hash_table *t, const char *key,
+				       uint32_t len)
 {
-	unsigned h = _hash(key) & (t->num_slots - 1);
+	unsigned h = _hash(key, len) & (t->num_slots - 1);
 	struct hash_node **c;
 
 	for (c = &t->slots[h]; *c; c = &((*c)->next))
-		if (!strcmp(key, (*c)->key))
+		if (!memcmp(key, (*c)->key, len))
 			break;
 
 	return c;
 }
 
-void *hash_lookup(struct hash_table *t, const char *key)
+void *hash_lookup_binary(struct hash_table *t, const char *key,
+			 uint32_t len)
 {
-	struct hash_node **c = _find(t, key);
+	struct hash_node **c = _find(t, key, len);
 	return *c ? (*c)->data : 0;
 }
 
-int hash_insert(struct hash_table *t, const char *key, void *data)
+int hash_insert_binary(struct hash_table *t, const char *key,
+		       uint32_t len, void *data)
 {
-	struct hash_node **c = _find(t, key);
+	struct hash_node **c = _find(t, key, len);
 
 	if (*c)
 		(*c)->data = data;
 	else {
-		struct hash_node *n = _create_node(key);
+		struct hash_node *n = _create_node(key, len);
 
 		if (!n)
 			return 0;
@@ -173,9 +179,10 @@ int hash_insert(struct hash_table *t, const char *key, void *data)
 	return 1;
 }
 
-void hash_remove(struct hash_table *t, const char *key)
+void hash_remove_binary(struct hash_table *t, const char *key,
+			uint32_t len)
 {
-	struct hash_node **c = _find(t, key);
+	struct hash_node **c = _find(t, key, len);
 
 	if (*c) {
 		struct hash_node *old = *c;
@@ -185,6 +192,21 @@ void hash_remove(struct hash_table *t, const char *key)
 	}
 }
 
+void *hash_lookup(struct hash_table *t, const char *key)
+{
+	return hash_lookup_binary(t, key, strlen(key) + 1);
+}
+
+int hash_insert(struct hash_table *t, const char *key, void *data)
+{
+	return hash_insert_binary(t, key, strlen(key) + 1, data);
+}
+
+void hash_remove(struct hash_table *t, const char *key)
+{
+	hash_remove_binary(t, key, strlen(key) + 1);
+}
+
 unsigned hash_get_num_entries(struct hash_table *t)
 {
 	return t->num_nodes;
@@ -235,6 +257,6 @@ struct hash_node *hash_get_first(struct hash_table *t)
 
 struct hash_node *hash_get_next(struct hash_table *t, struct hash_node *n)
 {
-	unsigned h = _hash(n->key) & (t->num_slots - 1);
+	unsigned h = _hash(n->key, n->keylen) & (t->num_slots - 1);
 	return n->next ? n->next : _next_slot(t, h + 1);
 }

lib/datastruct/hash.h

@@ -26,10 +26,14 @@ void hash_destroy(struct hash_table *t);
 void hash_wipe(struct hash_table *t);
 
 void *hash_lookup(struct hash_table *t, const char *key);
-void *hash_lookup_fixed(struct hash_table *t, const char *key, uint32_t len);
 int hash_insert(struct hash_table *t, const char *key, void *data);
 void hash_remove(struct hash_table *t, const char *key);
 
+void *hash_lookup_binary(struct hash_table *t, const char *key, uint32_t len);
+int hash_insert_binary(struct hash_table *t, const char *key, uint32_t len,
+		       void *data);
+void hash_remove_binary(struct hash_table *t, const char *key, uint32_t len);
+
 unsigned hash_get_num_entries(struct hash_table *t);
 void hash_iter(struct hash_table *t, iterate_fn f);
 

lib/device/dev-io.c

@@ -212,7 +212,7 @@ static int _aligned_io(struct device_area *where, void *buffer,
  * Public functions
  *---------------------------------------------------------------*/
 
-int dev_get_size(struct device *dev, uint64_t *size)
+int dev_get_size(const struct device *dev, uint64_t *size)
 {
 	int fd;
 	const char *name = dev_name(dev);
@@ -300,6 +300,12 @@ int dev_open_flags(struct device *dev, int flags, int direct, int quiet)
 		flags |= O_DIRECT;
 #endif
 
+#ifdef O_NOATIME
+	/* Don't update atime on device inodes */
+	if (!(dev->flags & DEV_REGULAR))
+		flags |= O_NOATIME;
+#endif
+
 	if ((dev->fd = open(name, flags, 0777)) < 0) {
 		log_sys_error("open", name);
 		return 0;

lib/device/device.h

@@ -55,7 +55,7 @@ struct device_area {
 /*
  * All io should use these routines.
  */
-int dev_get_size(struct device *dev, uint64_t *size);
+int dev_get_size(const struct device *dev, uint64_t *size);
 int dev_get_sectsize(struct device *dev, uint32_t *size);
 
 /* Use quiet version if device number could change e.g. when opening LV */

lib/display/display.c

@@ -18,6 +18,7 @@
 #include "display.h"
 #include "activate.h"
 #include "toolcontext.h"
+#include "segtype.h"
 
 #define SIZE_BUF 128
 
@@ -26,23 +27,13 @@ static struct {
 	const char *str;
 } _policies[] = {
 	{
-	ALLOC_NEXT_FREE, "next free"}, {
 	ALLOC_CONTIGUOUS, "contiguous"}, {
-	ALLOC_DEFAULT, "next free (default)"}
-};
-
-static struct {
-	segment_type_t segtype;
-	const char *str;
-} _segtypes[] = {
-	{
-	SEG_STRIPED, "striped"}, {
-	SEG_MIRRORED, "mirror"}, {
-	SEG_SNAPSHOT, "snapshot"}
+	ALLOC_NORMAL, "normal"}, {
+	ALLOC_ANYWHERE, "anywhere"}, {
+	ALLOC_INHERIT, "inherit"}
 };
 
 static int _num_policies = sizeof(_policies) / sizeof(*_policies);
-static int _num_segtypes = sizeof(_segtypes) / sizeof(*_segtypes);
 
 uint64_t units_to_bytes(const char *units, char *unit_type)
 {
@@ -124,17 +115,6 @@ const char *get_alloc_string(alloc_policy_t alloc)
 	return NULL;
 }
 
-const char *get_segtype_string(segment_type_t segtype)
-{
-	int i;
-
-	for (i = 0; i < _num_segtypes; i++)
-		if (_segtypes[i].segtype == segtype)
-			return _segtypes[i].str;
-
-	return "unknown";
-}
-
 alloc_policy_t get_alloc_from_string(const char *str)
 {
 	int i;
@@ -143,26 +123,19 @@ alloc_policy_t get_alloc_from_string(const char *str)
 		if (!strcmp(_policies[i].str, str))
 			return _policies[i].alloc;
 
-	log_error("Unrecognised allocation policy - using default");
-	return ALLOC_DEFAULT;
-}
-
-segment_type_t get_segtype_from_string(const char *str)
-{
-	int i;
-
-	for (i = 0; i < _num_segtypes; i++)
-		if (!strcmp(_segtypes[i].str, str))
-			return _segtypes[i].segtype;
-
-	log_error("Unrecognised segment type - using default (striped)");
-	return SEG_STRIPED;
+	/* Special case for old metadata */
+	if(!strcmp("next free", str))
+		return ALLOC_NORMAL;
+
+	log_error("Unrecognised allocation policy %s", str);
+	return ALLOC_INVALID;
 }
 
+/* Size supplied in sectors */
 const char *display_size(struct cmd_context *cmd, uint64_t size, size_len_t sl)
 {
 	int s;
-	int suffix = 1;
+	int suffix = 1, precision;
 	uint64_t byte = UINT64_C(0);
 	uint64_t units = UINT64_C(1024);
 	char *size_buf = NULL;
@@ -197,8 +170,9 @@ const char *display_size(struct cmd_context *cmd, uint64_t size, size_len_t sl)
 
 	if (s < 8) {
 		byte = cmd->current_settings.unit_factor;
-		size *= UINT64_C(1024);
+		size *= UINT64_C(512);
 	} else {
+		size /= 2;
 		suffix = 1;
 		if (cmd->current_settings.unit_type == 'H')
 			units = UINT64_C(1000);
@@ -210,8 +184,18 @@ const char *display_size(struct cmd_context *cmd, uint64_t size, size_len_t sl)
 			s++, byte /= units;
 	}
 
-	snprintf(size_buf, SIZE_BUF - 1, "%.2f%s", (float) size / byte,
-		 suffix ? size_str[s][sl] : "");
+	/* FIXME Make precision configurable */
+	switch(toupper((int) cmd->current_settings.unit_type)) {
+	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[s][sl] : "");
 
 	return size_buf;
 }
@@ -264,18 +248,18 @@ void pvdisplay_full(struct cmd_context *cmd, struct physical_volume *pv,
 	log_print("VG Name               %s%s", pv->vg_name,
 		  pv->status & EXPORTED_VG ? " (exported)" : "");
 
-	size = display_size(cmd, (uint64_t) pv->size / 2, SIZE_SHORT);
+	size = display_size(cmd, (uint64_t) pv->size, SIZE_SHORT);
 	if (pv->pe_size && pv->pe_count) {
 
 /******** FIXME display LVM on-disk data size
-		size2 = display_size(pv->size / 2, SIZE_SHORT);
+		size2 = display_size(pv->size, SIZE_SHORT);
 ********/
 
 		log_print("PV Size               %s" " / not usable %s",	/*  [LVM: %s]", */
-			  size, display_size(cmd,
-					     (pv->size -
-					      pv->pe_count * pv->pe_size) / 2,
-					     SIZE_SHORT));
+			  size,
+			  display_size(cmd, (pv->size -
+					     pv->pe_count * pv->pe_size),
+				       SIZE_SHORT));
 
 	} else
 		log_print("PV Size               %s", size);
@@ -386,8 +370,7 @@ int lvdisplay_full(struct cmd_context *cmd, struct logical_volume *lv,
 			snap_active = lv_snapshot_percent(snap->cow,
 							  &snap_percent);
 			if (!snap_active || snap_percent < 0 ||
-			    snap_percent >= 100)
-				snap_active = 0;
+			    snap_percent >= 100) snap_active = 0;
 			log_print("                       %s%s/%s [%s]",
 				  lv->vg->cmd->dev_dir, lv->vg->name,
 				  snap->cow->name,
@@ -419,7 +402,7 @@ int lvdisplay_full(struct cmd_context *cmd, struct logical_volume *lv,
 
 	log_print("LV Size                %s",
 		  display_size(cmd,
-			       snap ? snap->origin->size / 2 : lv->size / 2,
+			       snap ? snap->origin->size : lv->size,
 			       SIZE_SHORT));
 
 	log_print("Current LE             %u",
@@ -440,11 +423,11 @@ int lvdisplay_full(struct cmd_context *cmd, struct logical_volume *lv,
 			snap_percent = 100;
 
 		log_print("Snapshot chunk size    %s",
-			  display_size(cmd, (uint64_t) snap->chunk_size / 2,
+			  display_size(cmd, (uint64_t) snap->chunk_size,
 				       SIZE_SHORT));
 
 /*
-	size = display_size(lv->size / 2, SIZE_SHORT);
+	size = display_size(lv->size, SIZE_SHORT);
 	sscanf(size, "%f", &fsize);
 	fused = fsize * snap_percent / 100;
 */
@@ -478,7 +461,7 @@ int lvdisplay_full(struct cmd_context *cmd, struct logical_volume *lv,
 	return 0;
 }
 
-static void _display_stripe(struct lv_segment *seg, uint32_t s, const char *pre)
+void display_stripe(const struct lv_segment *seg, uint32_t s, const char *pre)
 {
 	switch (seg->area[s].type) {
 	case AREA_PV:
@@ -506,52 +489,18 @@ static void _display_stripe(struct lv_segment *seg, uint32_t s, const char *pre)
 
 int lvdisplay_segments(struct logical_volume *lv)
 {
-	uint32_t s;
-	struct list *segh;
 	struct lv_segment *seg;
 
 	log_print("--- Segments ---");
 
-	list_iterate(segh, &lv->segments) {
-		seg = list_item(segh, struct lv_segment);
-
+	list_iterate_items(seg, &lv->segments) {
 		log_print("Logical extent %u to %u:",
 			  seg->le, seg->le + seg->len - 1);
 
-		if (seg->type == SEG_STRIPED && seg->area_count == 1)
-			log_print("  Type\t\tlinear");
-		else
-			log_print("  Type\t\t%s",
-				  get_segtype_string(seg->type));
+		log_print("  Type\t\t%s", seg->segtype->ops->name(seg));
 
-		switch (seg->type) {
-		case SEG_STRIPED:
-			if (seg->area_count == 1)
-				_display_stripe(seg, 0, "  ");
-			else {
-				log_print("  Stripes\t\t%u", seg->area_count);
-				log_print("  Stripe size\t\t%u KB",
-					  seg->stripe_size / 2);
-
-				for (s = 0; s < seg->area_count; s++) {
-					log_print("  Stripe %d:", s);
-					_display_stripe(seg, s, "    ");
-				}
-			}
-			log_print(" ");
-			break;
-		case SEG_SNAPSHOT:
-			break;
-		case SEG_MIRRORED:
-			log_print("  Mirrors\t\t%u", seg->area_count);
-			log_print("  Mirror size\t\t%u", seg->area_len);
-			log_print("  Mirror original:");
-			_display_stripe(seg, 0, "    ");
-			log_print("  Mirror destination:");
-			_display_stripe(seg, 1, "    ");
-			log_print(" ");
-			break;
-		}
+		if (seg->segtype->ops->display)
+			seg->segtype->ops->display(seg);
 	}
 
 	log_print(" ");
@@ -605,7 +554,7 @@ void vgdisplay_full(struct volume_group *vg)
 	log_print("Open LV               %u", lvs_in_vg_opened(vg));
 /****** FIXME Max LV Size
       log_print ( "MAX LV Size           %s",
-               ( s1 = display_size ( LVM_LV_SIZE_MAX(vg) / 2, SIZE_SHORT)));
+               ( s1 = display_size ( LVM_LV_SIZE_MAX(vg), SIZE_SHORT)));
       free ( s1);
 *********/
 	log_print("Max PV                %u", vg->max_pv);
@@ -614,32 +563,25 @@ void vgdisplay_full(struct volume_group *vg)
 
 	log_print("VG Size               %s",
 		  display_size(vg->cmd,
-			       (uint64_t) vg->extent_count * (vg->extent_size /
-							      2), SIZE_SHORT));
+			       (uint64_t) vg->extent_count * vg->extent_size,
+			       SIZE_SHORT));
 
 	log_print("PE Size               %s",
-		  display_size(vg->cmd, (uint64_t) vg->extent_size / 2,
+		  display_size(vg->cmd, (uint64_t) vg->extent_size,
 			       SIZE_SHORT));
 
 	log_print("Total PE              %u", vg->extent_count);
 
 	log_print("Alloc PE / Size       %u / %s",
-		  vg->extent_count - vg->free_count, display_size(vg->cmd,
-								  ((uint64_t)
-								   vg->
-								   extent_count
-								   -
-								   vg->
-								   free_count) *
-								  (vg->
-								   extent_size /
-								   2),
-								  SIZE_SHORT));
+		  vg->extent_count - vg->free_count,
+		  display_size(vg->cmd,
+			       ((uint64_t) vg->extent_count - vg->free_count) *
+			       vg->extent_size, SIZE_SHORT));
 
 	log_print("Free  PE / Size       %u / %s", vg->free_count,
 		  display_size(vg->cmd,
-			       (uint64_t) vg->free_count * (vg->extent_size /
-							    2), SIZE_SHORT));
+			       (uint64_t) vg->free_count * vg->extent_size,
+			       SIZE_SHORT));
 
 	if (!id_write_format(&vg->id, uuid, sizeof(uuid))) {
 		stack;
@@ -708,15 +650,34 @@ void vgdisplay_short(struct volume_group *vg)
 {
 	log_print("\"%s\" %-9s [%-9s used / %s free]", vg->name,
 /********* FIXME if "open" print "/used" else print "/idle"???  ******/
-		  display_size(vg->cmd, (uint64_t) vg->extent_count *
-			       vg->extent_size / 2, SIZE_SHORT),
+		  display_size(vg->cmd,
+			       (uint64_t) vg->extent_count * vg->extent_size,
+			       SIZE_SHORT),
 		  display_size(vg->cmd,
 			       ((uint64_t) vg->extent_count -
-				vg->free_count) * vg->extent_size / 2,
-			       SIZE_SHORT), display_size(vg->cmd,
-							 (uint64_t) vg->
-							 free_count *
-							 vg->extent_size / 2,
-							 SIZE_SHORT));
+				vg->free_count) * vg->extent_size,
+			       SIZE_SHORT),
+		  display_size(vg->cmd,
+			       (uint64_t) vg->free_count * vg->extent_size,
+			       SIZE_SHORT));
 	return;
 }
+
+void display_formats(struct cmd_context *cmd)
+{
+	struct format_type *fmt;
+
+	list_iterate_items(fmt, &cmd->formats) {
+		log_print("%s", fmt->name);
+	}
+}
+
+void display_segtypes(struct cmd_context *cmd)
+{
+	struct segment_type *segtype;
+
+	list_iterate_items(segtype, &cmd->segtypes) {
+		log_print("%s", segtype->name);
+	}
+}
+

lib/display/display.h

@@ -27,6 +27,7 @@ uint64_t units_to_bytes(const char *units, char *unit_type);
 /* Specify size in KB */
 const char *display_size(struct cmd_context *cmd, uint64_t size, size_len_t sl);
 char *display_uuid(char *uuidstr);
+void display_stripe(const struct lv_segment *seg, uint32_t s, const char *pre);
 
 void pvdisplay_colons(struct physical_volume *pv);
 void pvdisplay_full(struct cmd_context *cmd, struct physical_volume *pv,
@@ -44,16 +45,13 @@ void vgdisplay_full(struct volume_group *vg);
 void vgdisplay_colons(struct volume_group *vg);
 void vgdisplay_short(struct volume_group *vg);
 
+void display_formats(struct cmd_context *cmd);
+void display_segtypes(struct cmd_context *cmd);
+
 /*
  * Allocation policy display conversion routines.
  */
 const char *get_alloc_string(alloc_policy_t alloc);
 alloc_policy_t get_alloc_from_string(const char *str);
 
-/*
- * Segment type display conversion routines.
- */
-segment_type_t get_segtype_from_string(const char *str);
-const char *get_segtype_string(segment_type_t segtype);
-
 #endif

lib/error/errseg.c

@@ -0,0 +1,103 @@
+/*
+ * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU General Public License v.2.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include "lib.h"
+#include "pool.h"
+#include "list.h"
+#include "toolcontext.h"
+#include "segtype.h"
+#include "display.h"
+#include "text_export.h"
+#include "text_import.h"
+#include "config.h"
+#include "str_list.h"
+#include "targets.h"
+#include "lvm-string.h"
+#include "activate.h"
+
+static const char *_name(const struct lv_segment *seg)
+{
+	return seg->segtype->name;
+}
+
+static int _merge_segments(struct lv_segment *seg1, struct lv_segment *seg2)
+{
+	seg1->len += seg2->len;
+	seg1->area_len += seg2->area_len;
+
+	return 1;
+}
+
+#ifdef DEVMAPPER_SUPPORT
+static int _compose_target_line(struct dev_manager *dm, struct pool *mem,
+				struct config_tree *cft, void **target_state,
+				struct lv_segment *seg, char *params,
+				size_t paramsize, const char **target, int *pos,
+				uint32_t *pvmove_mirror_count)
+{
+	/*   error */
+
+	*target = "error";
+	*params = '\0';
+
+	return 1;
+}
+
+static int _target_present(void)
+{
+	static int checked = 0;
+	static int present = 0;
+
+	if (!checked)
+		present = target_present("error");
+
+	checked = 1;
+	return present;
+}
+#endif
+
+static void _destroy(const struct segment_type *segtype)
+{
+	dbg_free((void *) segtype);
+}
+
+static struct segtype_handler _error_ops = {
+	name:_name,
+	merge_segments:_merge_segments,
+#ifdef DEVMAPPER_SUPPORT
+	compose_target_line:_compose_target_line,
+	target_present:_target_present,
+#endif
+	destroy:_destroy,
+};
+
+struct segment_type *init_error_segtype(struct cmd_context *cmd)
+{
+	struct segment_type *segtype = dbg_malloc(sizeof(*segtype));
+
+	if (!segtype) {
+		stack;
+		return NULL;
+	}
+
+	segtype->cmd = cmd;
+	segtype->ops = &_error_ops;
+	segtype->name = "error";
+	segtype->private = NULL;
+	segtype->flags = SEG_CAN_SPLIT | SEG_VIRTUAL;
+
+	log_very_verbose("Initialised segtype: %s", segtype->name);
+
+	return segtype;
+}

lib/filters/filter-composite.c

@@ -48,7 +48,7 @@ static void _destroy(struct dev_filter *f)
 
 struct dev_filter *composite_filter_create(int n, struct dev_filter **filters)
 {
-	struct dev_filter **filters_copy, *cf;
+	struct dev_filter **filters_copy, *cft;
 
 	if (!filters) {
 		stack;
@@ -63,15 +63,15 @@ struct dev_filter *composite_filter_create(int n, struct dev_filter **filters)
 	memcpy(filters_copy, filters, sizeof(*filters) * n);
 	filters_copy[n] = NULL;
 
-	if (!(cf = dbg_malloc(sizeof(*cf)))) {
+	if (!(cft = dbg_malloc(sizeof(*cft)))) {
 		log_error("compsoite filters allocation failed");
 		dbg_free(filters_copy);
 		return NULL;
 	}
 
-	cf->passes_filter = _and_p;
-	cf->destroy = _destroy;
-	cf->private = filters_copy;
+	cft->passes_filter = _and_p;
+	cft->destroy = _destroy;
+	cft->private = filters_copy;
 
-	return cf;
+	return cft;
 }

lib/filters/filter-persistent.c

@@ -63,7 +63,7 @@ int persistent_filter_wipe(struct dev_filter *f)
 static int _read_array(struct pfilter *pf, struct config_tree *cft,
 		       const char *path, void *data)
 {
-	struct config_node *cn;
+	const struct config_node *cn;
 	struct config_value *cv;
 
 	if (!(cn = find_config_node(cft->root, path))) {
@@ -99,12 +99,12 @@ int persistent_filter_load(struct dev_filter *f)
 	int r = 0;
 	struct config_tree *cft;
 
-	if (!(cft = create_config_tree())) {
+	if (!(cft = create_config_tree(pf->file))) {
 		stack;
 		return 0;
 	}
 
-	if (!read_config_file(cft, pf->file)) {
+	if (!read_config_file(cft)) {
 		stack;
 		goto out;
 	}

lib/filters/filter-sysfs.c

@@ -169,6 +169,8 @@ static int _read_devs(struct dev_set *ds, const char *dir)
 {
         struct dirent *d;
         DIR *dr;
+	unsigned char dtype;
+	struct stat info;
 	char path[PATH_MAX];
 	dev_t dev;
 	int r = 1;
@@ -189,19 +191,29 @@ static int _read_devs(struct dev_set *ds, const char *dir)
 			continue;
 		}
 
-		if (d->d_type == DT_DIR) {
+		dtype = d->d_type;
+
+		if (dtype == DT_UNKNOWN) {
+			if (stat(path, &info) >= 0) {
+				if (S_ISDIR(info.st_mode))
+					dtype = DT_DIR;
+				else if (S_ISREG(info.st_mode))
+					dtype = DT_REG;
+			}
+		}
+
+		if (dtype == DT_DIR) {
 			if (!_read_devs(ds, path)) {
 				r = 0;
 				break;
 			}
 		}
 
-		if ((d->d_type == DT_REG && !strcmp(d->d_name, "dev")))
+		if ((dtype == DT_REG && !strcmp(d->d_name, "dev")))
 			if (!_read_dev(path, &dev) || !_set_insert(ds, dev)) {
 				r = 0;
 				break;
 			}
-
 	}
 
         if (closedir(dr))

lib/filters/filter.c

@@ -25,7 +25,7 @@
 #include <fcntl.h>
 #include <limits.h>
 
-#define NUMBER_OF_MAJORS 256
+#define NUMBER_OF_MAJORS 4096
 
 typedef struct {
 	const char *name;
@@ -54,6 +54,8 @@ static const device_info_t device_info[] = {
 	{"ataraid", 16},	/* ATA Raid */
 	{"drbd", 16},		/* Distributed Replicated Block Device */
 	{"power2", 16},		/* EMC Powerpath */
+	{"i2o_block", 16},	/* i2o Block Disk */
+	{"iseries/vd", 8},	/* iSeries disks */
 	{NULL, 0}
 };
 
@@ -82,7 +84,7 @@ static int _passes_lvm_type_device_filter(struct dev_filter *f,
 	return 1;
 }
 
-static int *_scan_proc_dev(const char *proc, struct config_node *cn)
+static int *_scan_proc_dev(const char *proc, const struct config_node *cn)
 {
 	char line[80];
 	char proc_devices[PATH_MAX];
@@ -95,6 +97,7 @@ static int *_scan_proc_dev(const char *proc, struct config_node *cn)
 	int *max_partitions_by_major;
 	char *name;
 
+	/* FIXME Make this sparse */
 	if (!(max_partitions_by_major =
 	      dbg_malloc(sizeof(int) * NUMBER_OF_MAJORS))) {
 		log_error("Filter failed to allocate max_partitions_by_major");
@@ -199,7 +202,7 @@ static int *_scan_proc_dev(const char *proc, struct config_node *cn)
 }
 
 struct dev_filter *lvm_type_filter_create(const char *proc,
-					  struct config_node *cn)
+					  const struct config_node *cn)
 {
 	struct dev_filter *f;
 

lib/filters/filter.h

@@ -21,7 +21,9 @@
 #include <sys/stat.h>
 
 #ifdef linux
-#  include <linux/kdev_t.h>
+#  define MAJOR(dev)	((dev & 0xfff00) >> 8)
+#  define MINOR(dev)	((dev & 0xff) | ((dev >> 12) & 0xfff00))
+#  define MKDEV(ma,mi)	((mi & 0xff) | (ma << 8) | ((mi & ~0xff) << 12))
 #else
 #  define MAJOR(x) major((x))
 #  define MINOR(x) minor((x))
@@ -29,7 +31,7 @@
 #endif
 
 struct dev_filter *lvm_type_filter_create(const char *proc,
-					  struct config_node *cn);
+					  const struct config_node *cn);
 
 void lvm_type_filter_destroy(struct dev_filter *f);
 

lib/format1/Makefile.in

@@ -27,9 +27,7 @@ SOURCES =\
 
 LIB_SHARED = liblvm2format1.so
 
-include ../../make.tmpl
-
-.PHONY: install
+include $(top_srcdir)/make.tmpl
 
 install: liblvm2format1.so
 	$(INSTALL) -D $(OWNER) $(GROUP) -m 555 $(STRIP) $< \

lib/format1/disk-rep.c

@@ -117,6 +117,7 @@ static void _xlate_extents(struct pe_disk *extents, uint32_t count)
 static int _munge_formats(struct pv_disk *pvd)
 {
 	uint32_t pe_start;
+	int b, e;
 
 	switch (pvd->version) {
 	case 1:
@@ -134,17 +135,54 @@ static int _munge_formats(struct pv_disk *pvd)
 		return 0;
 	}
 
+        /* UUID too long? */
+        if (pvd->pv_uuid[ID_LEN]) {
+		/* Retain ID_LEN chars from end */
+                for (e = ID_LEN; e < sizeof(pvd->pv_uuid); e++) {
+                        if (!pvd->pv_uuid[e]) {
+                                e--;
+                                break;
+                        }
+                }
+		for (b = 0; b < ID_LEN; b++) {
+			pvd->pv_uuid[b] = pvd->pv_uuid[++e - ID_LEN];
+			/* FIXME Remove all invalid chars */
+			if (pvd->pv_uuid[b] == '/')
+				pvd->pv_uuid[b] = '#';
+		}
+		memset(&pvd->pv_uuid[ID_LEN], 0, sizeof(pvd->pv_uuid) - ID_LEN);
+        }
+
+	/* If UUID is missing, create one */
+	if (pvd->pv_uuid[0] == '\0')
+		uuid_from_num(pvd->pv_uuid, pvd->pv_number);
+
 	return 1;
 }
 
-static int _read_pvd(struct device *dev, struct pv_disk *pvd)
+/* 
+ * If exported, remove "PV_EXP" from end of VG name 
+ */
+static void _munge_exported_vg(struct pv_disk *pvd)
 {
-	if (!dev_read(dev, UINT64_C(0), sizeof(*pvd), pvd)) {
-		log_very_verbose("Failed to read PV data from %s",
-				 dev_name(dev));
-		return 0;
-	}
+	int l;
+	size_t s;
 
+	/* Return if PV not in a VG */
+	if ((!*pvd->vg_name))
+		return;
+	/* FIXME also check vgd->status & VG_EXPORTED? */
+
+	l = strlen(pvd->vg_name);
+	s = sizeof(EXPORTED_TAG);
+	if (!strncmp(pvd->vg_name + l - s + 1, EXPORTED_TAG, s)) {
+		pvd->vg_name[l - s + 1] = '\0';
+                pvd->pv_status |= VG_EXPORTED;
+        }
+}
+
+int munge_pvd(struct device *dev, struct pv_disk *pvd)
+{
 	_xlate_pvd(pvd);
 
 	if (pvd->id[0] != 'H' || pvd->id[1] != 'M') {
@@ -159,13 +197,23 @@ static int _read_pvd(struct device *dev, struct pv_disk *pvd)
 		return 0;
 	}
 
-	/* If UUID is missing, create one */
-	if (pvd->pv_uuid[0] == '\0')
-		uuid_from_num(pvd->pv_uuid, pvd->pv_number);
+	/* If VG is exported, set VG name back to the real name */
+	_munge_exported_vg(pvd);
 
 	return 1;
 }
 
+static int _read_pvd(struct device *dev, struct pv_disk *pvd)
+{
+	if (!dev_read(dev, UINT64_C(0), sizeof(*pvd), pvd)) {
+		log_very_verbose("Failed to read PV data from %s",
+				 dev_name(dev));
+		return 0;
+	}
+
+	return munge_pvd(dev, pvd);
+}
+
 static int _read_lvd(struct device *dev, uint64_t pos, struct lv_disk *disk)
 {
 	if (!dev_read(dev, pos, sizeof(*disk), disk))
@@ -185,6 +233,9 @@ static int _read_vgd(struct disk_list *data)
 
 	_xlate_vgd(vgd);
 
+	if ((vgd->lv_max > MAX_LV) || (vgd->pv_max > MAX_PV))
+		fail;
+		
 	/* If UUID is missing, create one */
 	if (vgd->vg_uuid[0] == '\0')
 		uuid_from_num(vgd->vg_uuid, vgd->vg_number);
@@ -269,26 +320,6 @@ static int _read_extents(struct disk_list *data)
 	return 1;
 }
 
-/* 
- * If exported, remove "PV_EXP" from end of VG name 
- */
-void munge_exported_vg(struct pv_disk *pvd)
-{
-	int l;
-	size_t s;
-
-	/* Return if PV not in a VG */
-	if ((!*pvd->vg_name))
-		return;
-
-	l = strlen(pvd->vg_name);
-	s = sizeof(EXPORTED_TAG);
-	if (!strncmp(pvd->vg_name + l - s + 1, EXPORTED_TAG, s)) {
-		pvd->vg_name[l - s + 1] = '\0'; 
-		pvd->pv_status |= VG_EXPORTED;
-	}
-}
-
 static struct disk_list *__read_disk(const struct format_type *fmt,
 				     struct device *dev, struct pool *mem,
 				     const char *vg_name)
@@ -312,9 +343,6 @@ static struct disk_list *__read_disk(const struct format_type *fmt,
 		goto bad;
 	}
 
-	/* If VG is exported, set VG name back to the real name */
-	munge_exported_vg(&dl->pvd);
-
 	if (!(info = lvmcache_add(fmt->labeller, dl->pvd.pv_uuid, dev,
 				  dl->pvd.vg_name, NULL)))
 		stack;

lib/format1/disk-rep.h

@@ -216,7 +216,7 @@ int export_vg(struct vg_disk *vgd, struct volume_group *vg);
 
 int import_lv(struct pool *mem, struct logical_volume *lv, struct lv_disk *lvd);
 
-int import_extents(struct pool *mem, struct volume_group *vg,
+int import_extents(struct cmd_context *cmd, struct volume_group *vg,
 		   struct list *pvds);
 int export_extents(struct disk_list *dl, uint32_t lv_num,
 		   struct logical_volume *lv, struct physical_volume *pv);
@@ -237,7 +237,7 @@ int export_uuids(struct disk_list *dl, struct volume_group *vg);
 void export_numbers(struct list *pvds, struct volume_group *vg);
 
 void export_pv_act(struct list *pvds);
-void munge_exported_vg(struct pv_disk *pvd);
+int munge_pvd(struct device *dev, struct pv_disk *pvd);
 
 /* blech */
 int get_free_vg_number(struct format_instance *fid, struct dev_filter *filter,

lib/format1/format1.c

@@ -24,6 +24,7 @@
 #include "lvmcache.h"
 #include "lvm1-label.h"
 #include "format1.h"
+#include "segtype.h"
 
 #define FMT_LVM1_NAME "lvm1"
 
@@ -164,7 +165,7 @@ static struct volume_group *_build_vg(struct format_instance *fid,
 	if (!import_lvs(mem, vg, pvs))
 		goto bad;
 
-	if (!import_extents(mem, vg, pvs))
+	if (!import_extents(fid->fmt->cmd, vg, pvs))
 		goto bad;
 
 	if (!import_snapshots(mem, vg, pvs))
@@ -345,7 +346,7 @@ static int _pv_setup(const struct format_type *fmt,
 		pv->size--;
 	if (pv->size > MAX_PV_SIZE) {
 		log_error("Physical volumes cannot be bigger than %s",
-			  display_size(fmt->cmd, (uint64_t) MAX_PV_SIZE / 2,
+			  display_size(fmt->cmd, (uint64_t) MAX_PV_SIZE,
 				       SIZE_SHORT));
 		return 0;
 	}
@@ -386,7 +387,7 @@ static int _lv_setup(struct format_instance *fid, struct logical_volume *lv)
 	}
 	if (lv->size > max_size) {
 		log_error("logical volumes cannot be larger than %s",
-			  display_size(fid->fmt->cmd, max_size / 2,
+			  display_size(fid->fmt->cmd, max_size,
 				       SIZE_SHORT));
 		return 0;
 	}
@@ -468,21 +469,18 @@ static int _vg_setup(struct format_instance *fid, struct volume_group *vg)
 
 	if (vg->extent_size > MAX_PE_SIZE || vg->extent_size < MIN_PE_SIZE) {
 		log_error("Extent size must be between %s and %s",
-			  display_size(fid->fmt->cmd, (uint64_t) MIN_PE_SIZE
-				       / 2,
-				       SIZE_SHORT), display_size(fid->fmt->cmd,
-								 (uint64_t)
-								 MAX_PE_SIZE
-								 / 2,
-								 SIZE_SHORT));
+			  display_size(fid->fmt->cmd, (uint64_t) MIN_PE_SIZE,
+				       SIZE_SHORT),
+			  display_size(fid->fmt->cmd, (uint64_t) MAX_PE_SIZE,
+				       SIZE_SHORT));
 
 		return 0;
 	}
 
 	if (vg->extent_size % MIN_PE_SIZE) {
 		log_error("Extent size must be multiple of %s",
-			  display_size(fid->fmt->cmd,
-				       (uint64_t) MIN_PE_SIZE / 2, SIZE_SHORT));
+			  display_size(fid->fmt->cmd, (uint64_t) MIN_PE_SIZE,
+				       SIZE_SHORT));
 		return 0;
 	}
 
@@ -495,6 +493,17 @@ static int _vg_setup(struct format_instance *fid, struct volume_group *vg)
 	return 1;
 }
 
+static int _segtype_supported (struct format_instance *fid, 
+			       struct segment_type *segtype)
+{
+	if (!(segtype->flags & SEG_FORMAT1_SUPPORT)) {
+		stack;
+		return 0;
+	}
+
+	return 1;
+}
+
 static struct metadata_area_ops _metadata_format1_ops = {
 	vg_read:_vg_read,
 	vg_write:_vg_write,
@@ -545,6 +554,7 @@ static struct format_handler _format1_ops = {
 	pv_write:_pv_write,
 	lv_setup:_lv_setup,
 	vg_setup:_vg_setup,
+	segtype_supported:_segtype_supported,
 	create_instance:_create_instance,
 	destroy_instance:_destroy_instance,
 	destroy:_destroy,
@@ -581,5 +591,7 @@ struct format_type *init_format(struct cmd_context *cmd)
 		return NULL;
 	}
 
+	log_very_verbose("Initialised format: %s", fmt->name);
+
 	return fmt;
 }

lib/format1/import-export.c

@@ -25,6 +25,7 @@
 #include "lvm-string.h"
 #include "filter.h"
 #include "toolcontext.h"
+#include "segtype.h"
 
 #include <time.h>
 
@@ -236,6 +237,7 @@ int import_vg(struct pool *mem,
 	vg->free_count = vgd->pe_total - vgd->pe_allocated;
 	vg->max_lv = vgd->lv_max;
 	vg->max_pv = vgd->pv_max;
+	vg->alloc = ALLOC_NORMAL;
 
 	if (partial)
 		vg->status |= PARTIAL_VG;
@@ -315,7 +317,7 @@ int import_lv(struct pool *mem, struct logical_volume *lv, struct lv_disk *lvd)
 	if (lvd->lv_allocation & LV_CONTIGUOUS)
 		lv->alloc = ALLOC_CONTIGUOUS;
 	else
-		lv->alloc = ALLOC_NEXT_FREE;
+		lv->alloc = ALLOC_NORMAL;
 
 	lv->read_ahead = lvd->lv_read_ahead;
 	lv->size = lvd->lv_size;
@@ -380,9 +382,10 @@ int export_extents(struct disk_list *dl, uint32_t lv_num,
 		seg = list_item(segh, struct lv_segment);
 
 		for (s = 0; s < seg->area_count; s++) {
-			if (seg->type != SEG_STRIPED) {
-				log_error("Non-striped segment type in LV %s: "
-					  "unsupported by format1", lv->name);
+			if (!(seg->segtype->flags & SEG_FORMAT1_SUPPORT)) {
+				log_error("Segment type %s in LV %s: "
+					  "unsupported by format1",
+					  seg->segtype->name, lv->name);
 				return 0;
 			}
 			if (seg->area[s].type != AREA_PV) {

lib/format1/import-extents.c

@@ -19,6 +19,8 @@
 #include "pool.h"
 #include "disk-rep.h"
 #include "lv_alloc.h"
+#include "display.h"
+#include "segtype.h"
 
 /*
  * After much thought I have decided it is easier,
@@ -201,21 +203,24 @@ static int _check_maps_are_complete(struct hash_table *maps)
 	return 1;
 }
 
-static int _read_linear(struct pool *mem, struct lv_map *lvm)
+static int _read_linear(struct cmd_context *cmd, struct lv_map *lvm)
 {
 	uint32_t le = 0;
 	struct lv_segment *seg;
 
 	while (le < lvm->lv->le_count) {
-		seg = alloc_lv_segment(mem, 1);
+		seg = alloc_lv_segment(cmd->mem, 1);
 
 		seg->lv = lvm->lv;
-		seg->type = SEG_STRIPED;
+		if (!(seg->segtype = get_segtype_from_string(cmd, "striped"))) {
+			stack;
+			return 0;
+		}
+
 		seg->le = le;
 		seg->len = 0;
 		seg->area_len = 0;
 		seg->stripe_size = 0;
-		seg->area_count = 1;
 
 		seg->area[0].type = AREA_PV;
 		seg->area[0].u.pv.pv = lvm->map[le].pv;
@@ -251,13 +256,12 @@ static int _check_stripe(struct lv_map *lvm, struct lv_segment *seg,
 		if ((lvm->map[le + st * len].pv != seg->area[st].u.pv.pv) ||
 		    (seg->area[st].u.pv.pv &&
 		     lvm->map[le + st * len].pe !=
-		     seg->area[st].u.pv.pe + seg->len))
-			return 0;
+		     seg->area[st].u.pv.pe + seg->len)) return 0;
 
 	return 1;
 }
 
-static int _read_stripes(struct pool *mem, struct lv_map *lvm)
+static int _read_stripes(struct cmd_context *cmd, struct lv_map *lvm)
 {
 	uint32_t st, le = 0, len;
 	struct lv_segment *seg;
@@ -273,15 +277,17 @@ static int _read_stripes(struct pool *mem, struct lv_map *lvm)
 	len = lvm->lv->le_count / lvm->stripes;
 
 	while (le < len) {
-		if (!(seg = alloc_lv_segment(mem, lvm->stripes))) {
+		if (!(seg = alloc_lv_segment(cmd->mem, lvm->stripes))) {
 			stack;
 			return 0;
 		}
 
 		seg->lv = lvm->lv;
-		seg->type = SEG_STRIPED;
+		if (!(seg->segtype = get_segtype_from_string(cmd, "striped"))) {
+			stack;
+			return 0;
+		}
 		seg->stripe_size = lvm->stripe_size;
-		seg->area_count = lvm->stripes;
 		seg->le = seg->area_count * le;
 		seg->len = 1;
 		seg->area_len = 1;
@@ -312,20 +318,20 @@ static int _read_stripes(struct pool *mem, struct lv_map *lvm)
 	return 1;
 }
 
-static int _build_segments(struct pool *mem, struct lv_map *lvm)
+static int _build_segments(struct cmd_context *cmd, struct lv_map *lvm)
 {
-	return (lvm->stripes > 1 ? _read_stripes(mem, lvm) :
-		_read_linear(mem, lvm));
+	return (lvm->stripes > 1 ? _read_stripes(cmd, lvm) :
+		_read_linear(cmd, lvm));
 }
 
-static int _build_all_segments(struct pool *mem, struct hash_table *maps)
+static int _build_all_segments(struct cmd_context *cmd, struct hash_table *maps)
 {
 	struct hash_node *n;
 	struct lv_map *lvm;
 
 	for (n = hash_get_first(maps); n; n = hash_get_next(maps, n)) {
 		lvm = (struct lv_map *) hash_get_data(maps, n);
-		if (!_build_segments(mem, lvm)) {
+		if (!_build_segments(cmd, lvm)) {
 			stack;
 			return 0;
 		}
@@ -334,7 +340,8 @@ static int _build_all_segments(struct pool *mem, struct hash_table *maps)
 	return 1;
 }
 
-int import_extents(struct pool *mem, struct volume_group *vg, struct list *pvds)
+int import_extents(struct cmd_context *cmd, struct volume_group *vg,
+		   struct list *pvds)
 {
 	int r = 0;
 	struct pool *scratch = pool_create(10 * 1024);
@@ -360,7 +367,7 @@ int import_extents(struct pool *mem, struct volume_group *vg, struct list *pvds)
 		goto out;
 	}
 
-	if (!_build_all_segments(mem, maps)) {
+	if (!_build_all_segments(cmd, maps)) {
 		log_err("Couldn't build extent segments.");
 		goto out;
 	}

lib/format1/lvm1-label.c

@@ -60,7 +60,8 @@ static int _read(struct labeller *l, struct device *dev, char *buf,
 	struct pv_disk *pvd = (struct pv_disk *) buf;
 	struct lvmcache_info *info;
 
-	munge_exported_vg(pvd);
+	munge_pvd(dev, pvd);
+
 	if (!(info = lvmcache_add(l, pvd->pv_uuid, dev, pvd->vg_name, NULL))) {
 		stack;
 		return 0;

lib/format_pool/.exported_symbols

@@ -0,0 +1 @@
+init_format

lib/format_pool/Makefile.in

@@ -0,0 +1,35 @@
+#
+# Copyright (C) 2003-2004 Sistina Software, Inc. All rights reserved.
+# Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+#
+# This file is part of the LVM2.
+#
+# This copyrighted material is made available to anyone wishing to use,
+# modify, copy, or redistribute it subject to the terms and conditions
+# of the GNU General Public License v.2.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software Foundation,
+# Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+
+srcdir = @srcdir@
+top_srcdir = @top_srcdir@
+VPATH = @srcdir@
+
+SOURCES =\
+	disk_rep.c \
+	format_pool.c \
+	import_export.c \
+	pool_label.c
+
+LIB_SHARED = liblvm2formatpool.so
+
+include $(top_srcdir)/make.tmpl
+
+install: liblvm2formatpool.so
+	$(INSTALL) -D $(OWNER) $(GROUP) -m 555 $(STRIP) $< \
+		$(libdir)/liblvm2formatpool.so.$(LIB_VERSION)
+	$(LN_S) -f liblvm2formatpool.so.$(LIB_VERSION) \
+		$(libdir)/liblvm2formatpool.so
+
+

lib/format_pool/disk_rep.c

@@ -0,0 +1,385 @@
+/*
+ * Copyright (C) 1997-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU General Public License v.2.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include "lib.h"
+#include "pool.h"
+#include "label.h"
+#include "metadata.h"
+#include "lvmcache.h"
+#include "filter.h"
+#include "list.h"
+#include "xlate.h"
+
+#include "disk_rep.h"
+
+/* FIXME: memcpy might not be portable */
+#define CPIN_8(x, y, z) {memcpy((x), (y), (z));}
+#define CPOUT_8(x, y, z) {memcpy((y), (x), (z));}
+#define CPIN_16(x, y) {(x) = xlate16_be((y));}
+#define CPOUT_16(x, y) {(y) = xlate16_be((x));}
+#define CPIN_32(x, y) {(x) = xlate32_be((y));}
+#define CPOUT_32(x, y) {(y) = xlate32_be((x));}
+#define CPIN_64(x, y) {(x) = xlate64_be((y));}
+#define CPOUT_64(x, y) {(y) = xlate64_be((x));}
+
+static int __read_pool_disk(const struct format_type *fmt, struct device *dev,
+			    struct pool *mem, struct pool_list *pl,
+			    const char *vg_name)
+{
+	char buf[512];
+
+	/* FIXME: Need to check the cache here first */
+	if (!dev_read(dev, UINT64_C(0), 512, buf)) {
+		log_very_verbose("Failed to read PV data from %s",
+				 dev_name(dev));
+		return 0;
+	}
+
+	if (!read_pool_label(pl, fmt->labeller, dev, buf, NULL)) {
+		stack;
+		return 0;
+	}
+
+	return 1;
+}
+
+static void _add_pl_to_list(struct list *head, struct pool_list *data)
+{
+	struct list *pvdh;
+	struct pool_list *pl;
+
+	list_iterate(pvdh, head) {
+		pl = list_item(pvdh, struct pool_list);
+
+		if (id_equal(&data->pv_uuid, &pl->pv_uuid)) {
+			char uuid[ID_LEN + 7];
+
+			id_write_format(&pl->pv_uuid, uuid, ID_LEN + 7);
+
+			if (MAJOR(data->dev->dev) != md_major()) {
+				log_very_verbose("Ignoring duplicate PV %s on "
+						 "%s", uuid,
+						 dev_name(data->dev));
+				return;
+			}
+			log_very_verbose("Duplicate PV %s - using md %s",
+					 uuid, dev_name(data->dev));
+			list_del(pvdh);
+			break;
+		}
+	}
+	list_add(head, &data->list);
+}
+
+int read_pool_label(struct pool_list *pl, struct labeller *l,
+		    struct device *dev, char *buf, struct label **label)
+{
+	struct lvmcache_info *info;
+	struct id pvid;
+	struct id vgid;
+	char uuid[ID_LEN + 7];
+	struct pool_disk *pd = &pl->pd;
+
+	pool_label_in(pd, buf);
+
+	get_pool_pv_uuid(&pvid, pd);
+	id_write_format(&pvid, uuid, ID_LEN + 7);
+	log_debug("Calculated uuid %s for %s", uuid, dev_name(dev));
+
+	get_pool_vg_uuid(&vgid, pd);
+	id_write_format(&vgid, uuid, ID_LEN + 7);
+	log_debug("Calculated uuid %s for %s", uuid, pd->pl_pool_name);
+
+	if (!(info = lvmcache_add(l, (char *) &pvid, dev, pd->pl_pool_name,
+				  (char *) &vgid))) {
+		stack;
+		return 0;
+	}
+	if (label)
+		*label = info->label;
+
+	info->device_size = xlate32_be(pd->pl_blocks) << SECTOR_SHIFT;
+	list_init(&info->mdas);
+
+	info->status &= ~CACHE_INVALID;
+
+	pl->dev = dev;
+	pl->pv = NULL;
+	memcpy(&pl->pv_uuid, &pvid, sizeof(pvid));
+
+	return 1;
+}
+
+/**
+ * pool_label_out - copies a pool_label_t into a char buffer
+ * @pl: ptr to a pool_label_t struct
+ * @buf: ptr to raw space where label info will be copied
+ *
+ * This function is important because it takes care of all of
+ * the endian issues when copying to disk.  This way, when
+ * machines of different architectures are used, they will
+ * be able to interpret ondisk labels correctly.  Always use
+ * this function before writing to disk.
+ */
+void pool_label_out(struct pool_disk *pl, char *buf)
+{
+	struct pool_disk *bufpl = (struct pool_disk *) buf;
+
+	CPOUT_64(pl->pl_magic, bufpl->pl_magic);
+	CPOUT_64(pl->pl_pool_id, bufpl->pl_pool_id);
+	CPOUT_8(pl->pl_pool_name, bufpl->pl_pool_name, POOL_NAME_SIZE);
+	CPOUT_32(pl->pl_version, bufpl->pl_version);
+	CPOUT_32(pl->pl_subpools, bufpl->pl_subpools);
+	CPOUT_32(pl->pl_sp_id, bufpl->pl_sp_id);
+	CPOUT_32(pl->pl_sp_devs, bufpl->pl_sp_devs);
+	CPOUT_32(pl->pl_sp_devid, bufpl->pl_sp_devid);
+	CPOUT_32(pl->pl_sp_type, bufpl->pl_sp_type);
+	CPOUT_64(pl->pl_blocks, bufpl->pl_blocks);
+	CPOUT_32(pl->pl_striping, bufpl->pl_striping);
+	CPOUT_32(pl->pl_sp_dmepdevs, bufpl->pl_sp_dmepdevs);
+	CPOUT_32(pl->pl_sp_dmepid, bufpl->pl_sp_dmepid);
+	CPOUT_32(pl->pl_sp_weight, bufpl->pl_sp_weight);
+	CPOUT_32(pl->pl_minor, bufpl->pl_minor);
+	CPOUT_32(pl->pl_padding, bufpl->pl_padding);
+	CPOUT_8(pl->pl_reserve, bufpl->pl_reserve, 184);
+}
+
+/**
+ * pool_label_in - copies a char buffer into a pool_label_t
+ * @pl: ptr to a pool_label_t struct
+ * @buf: ptr to raw space where label info is copied from
+ *
+ * This function is important because it takes care of all of
+ * the endian issues when information from disk is about to be
+ * used.  This way, when machines of different architectures
+ * are used, they will be able to interpret ondisk labels
+ * correctly.  Always use this function before using labels that
+ * were read from disk.
+ */
+void pool_label_in(struct pool_disk *pl, char *buf)
+{
+	struct pool_disk *bufpl = (struct pool_disk *) buf;
+
+	CPIN_64(pl->pl_magic, bufpl->pl_magic);
+	CPIN_64(pl->pl_pool_id, bufpl->pl_pool_id);
+	CPIN_8(pl->pl_pool_name, bufpl->pl_pool_name, POOL_NAME_SIZE);
+	CPIN_32(pl->pl_version, bufpl->pl_version);
+	CPIN_32(pl->pl_subpools, bufpl->pl_subpools);
+	CPIN_32(pl->pl_sp_id, bufpl->pl_sp_id);
+	CPIN_32(pl->pl_sp_devs, bufpl->pl_sp_devs);
+	CPIN_32(pl->pl_sp_devid, bufpl->pl_sp_devid);
+	CPIN_32(pl->pl_sp_type, bufpl->pl_sp_type);
+	CPIN_64(pl->pl_blocks, bufpl->pl_blocks);
+	CPIN_32(pl->pl_striping, bufpl->pl_striping);
+	CPIN_32(pl->pl_sp_dmepdevs, bufpl->pl_sp_dmepdevs);
+	CPIN_32(pl->pl_sp_dmepid, bufpl->pl_sp_dmepid);
+	CPIN_32(pl->pl_sp_weight, bufpl->pl_sp_weight);
+	CPIN_32(pl->pl_minor, bufpl->pl_minor);
+	CPIN_32(pl->pl_padding, bufpl->pl_padding);
+	CPIN_8(pl->pl_reserve, bufpl->pl_reserve, 184);
+}
+
+static char _calc_char(unsigned int id)
+{
+	/*
+	 * [0-9A-Za-z!#] - 64 printable chars (6-bits)
+	 */
+
+	if (id < 10)
+		return id + 48;
+	if (id < 36)
+		return (id - 10) + 65;
+	if (id < 62)
+		return (id - 36) + 97;
+	if (id == 62)
+		return '!';
+	if (id == 63)
+		return '#';
+
+	return '%';
+}
+
+void get_pool_uuid(char *uuid, uint64_t poolid, uint32_t spid, uint32_t devid)
+{
+	int i;
+	unsigned shifter = 0x003F;
+
+	assert(ID_LEN == 32);
+	memset(uuid, 0, ID_LEN);
+	strcat(uuid, "POOL0000000000");
+
+	/* We grab the entire 64 bits (+2 that get shifted in) */
+	for (i = 13; i < 24; i++) {
+		uuid[i] = _calc_char(((unsigned) poolid) & shifter);
+		poolid = poolid >> 6;
+	}
+
+	/* We grab the entire 32 bits (+4 that get shifted in) */
+	for (i = 24; i < 30; i++) {
+		uuid[i] = _calc_char((unsigned) (spid & shifter));
+		spid = spid >> 6;
+	}
+
+	/*
+	 * Since we can only have 128 devices, we only worry about the
+	 * last 12 bits
+	 */
+	for (i = 30; i < 32; i++) {
+		uuid[i] = _calc_char((unsigned) (devid & shifter));
+		devid = devid >> 6;
+	}
+
+}
+
+static int _read_vg_pds(const struct format_type *fmt, struct pool *mem,
+			struct lvmcache_vginfo *vginfo, struct list *head,
+			uint32_t *devcount)
+{
+
+	struct list *vgih = NULL;
+	struct device *dev;
+	struct pool_list *pl = NULL;
+	struct pool *tmpmem = NULL;
+
+	uint32_t sp_count = 0;
+	uint32_t *sp_devs = NULL;
+	int i;
+
+	/* FIXME: maybe should return a different error in memory
+	 * allocation failure */
+	if (!(tmpmem = pool_create(512))) {
+		stack;
+		return 0;
+	}
+
+	list_iterate(vgih, &vginfo->infos) {
+		dev = list_item(vgih, struct lvmcache_info)->dev;
+		if (dev &&
+		    !(pl = read_pool_disk(fmt, dev, mem, vginfo->vgname)))
+			    break;
+		/*
+		 * We need to keep track of the total expected number
+		 * of devices per subpool
+		 */
+		if (!sp_count) {
+			sp_count = pl->pd.pl_subpools;
+			if (!(sp_devs =
+			      pool_zalloc(tmpmem,
+					  sizeof(uint32_t) * sp_count))) {
+				log_error("Unable to allocate %d 32-bit uints",
+					  sp_count);
+				pool_destroy(tmpmem);
+				return 0;
+			}
+		}
+		/*
+		 * watch out for a pool label with a different subpool
+		 * count than the original - give up if it does
+		 */
+		if (sp_count != pl->pd.pl_subpools)
+			break;
+
+		_add_pl_to_list(head, pl);
+
+		if (sp_count > pl->pd.pl_sp_id && sp_devs[pl->pd.pl_sp_id] == 0)
+			sp_devs[pl->pd.pl_sp_id] = pl->pd.pl_sp_devs;
+	}
+
+	*devcount = 0;
+	for (i = 0; i < sp_count; i++) {
+		*devcount += sp_devs[i];
+	}
+
+	pool_destroy(tmpmem);
+
+	if (pl && *pl->pd.pl_pool_name)
+		return 1;
+
+	return 0;
+
+}
+
+int read_pool_pds(const struct format_type *fmt, const char *vg_name,
+		  struct pool *mem, struct list *pdhead)
+{
+	struct lvmcache_vginfo *vginfo;
+	uint32_t totaldevs;
+	int full_scan = -1;
+
+	do {
+		/*
+		 * If the cache scanning doesn't work, this will never work
+		 */
+		if (vg_name && (vginfo = vginfo_from_vgname(vg_name)) &&
+		    vginfo->infos.n) {
+
+			if (_read_vg_pds(fmt, mem, vginfo, pdhead, &totaldevs)) {
+				/*
+				 * If we found all the devices we were
+				 * expecting, return success
+				 */
+				if (list_size(pdhead) == totaldevs)
+					return 1;
+
+				/*
+				 * accept partial pool if we've done a full
+				 * rescan of the cache
+				 */
+				if (full_scan > 0)
+					return 1;
+			}
+		}
+		/* Failed */
+		list_init(pdhead);
+
+		full_scan++;
+		if (full_scan > 1) {
+			log_debug("No devices for vg %s found in cache",
+				  vg_name);
+			return 0;
+		}
+		lvmcache_label_scan(fmt->cmd, full_scan);
+
+	} while (1);
+
+}
+
+struct pool_list *read_pool_disk(const struct format_type *fmt,
+				 struct device *dev, struct pool *mem,
+				 const char *vg_name)
+{
+	struct pool_list *pl;
+
+	if (!dev_open(dev)) {
+		stack;
+		return NULL;
+	}
+
+	if (!(pl = pool_zalloc(mem, sizeof(*pl)))) {
+		log_error("Unable to allocate pool list structure");
+		return 0;
+	}
+
+	if (!__read_pool_disk(fmt, dev, mem, pl, vg_name)) {
+		stack;
+		return NULL;
+	}
+
+	if (!dev_close(dev))
+		stack;
+
+	return pl;
+
+}

lib/format_pool/disk_rep.h

@@ -0,0 +1,178 @@
+/*
+ * Copyright (C) 1997-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU General Public License v.2.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#ifndef DISK_REP_FORMAT_POOL_H
+#define DISK_REP_FORMAT_POOL_H
+
+#include "label.h"
+#include "metadata.h"
+#include "pool.h"
+
+/* From NSP.cf */
+#define NSPMajorVersion	4
+#define NSPMinorVersion	1
+#define NSPUpdateLevel	3
+
+/* From pool_std.h */
+#define POOL_NAME_SIZE          (256)
+#define POOL_MAGIC 		0x011670
+#define POOL_MAJOR              (121)
+#define POOL_MAX_DEVICES 	128
+
+/* When checking for version matching, the first two numbers **
+** are important for metadata formats, a.k.a pool labels.   **
+** All the numbers are important when checking if the user  **
+** space tools match up with the kernel module............. */
+#define POOL_VERSION		(NSPMajorVersion << 16 | \
+				 NSPMinorVersion <<  8 | \
+				 NSPUpdateLevel)
+
+/* Pool label is at the head of every pool disk partition */
+#define SIZEOF_POOL_LABEL       (8192)
+
+/* in sectors */
+#define POOL_PE_SIZE     (SIZEOF_POOL_LABEL >> SECTOR_SHIFT)
+#define POOL_PE_START    (SIZEOF_POOL_LABEL >> SECTOR_SHIFT)
+
+/* Helper fxns */
+#define get_pool_vg_uuid(id, pd) do { get_pool_uuid((char *)(id), \
+                                                    (pd)->pl_pool_id, 0, 0); \
+                                    } while(0)
+#define get_pool_pv_uuid(id, pd) do { get_pool_uuid((char *)(id), \
+                                                    (pd)->pl_pool_id, \
+                                                    (pd)->pl_sp_id, \
+                                                    (pd)->pl_sp_devid); \
+                                    } while(0)
+#define get_pool_lv_uuid(id, pd) do { get_pool_uuid((char *)&(id)[0], \
+                                                    (pd)->pl_pool_id, 0, 0); \
+                                      get_pool_uuid((char*)&(id)[1], \
+                                                    (pd)->pl_pool_id, 0, 0); \
+                                    } while(0)
+
+struct pool_disk;
+struct pool_list;
+struct user_subpool;
+struct user_device;
+
+/* This must be kept up to date with sistina/pool/module/pool_sptypes.h */
+
+/*  Generic Labels  */
+#define SPTYPE_DATA                (0x00000000)
+
+/*  GFS specific labels  */
+#define SPTYPE_GFS_DATA            (0x68011670)
+#define SPTYPE_GFS_JOURNAL         (0x69011670)
+
+struct sptype_name {
+	const char *name;
+	uint32_t label;
+};
+
+static const struct sptype_name sptype_names[] = {
+	{"data",	SPTYPE_DATA},
+
+	{"gfs_data",	SPTYPE_GFS_DATA},
+	{"gfs_journal",	SPTYPE_GFS_JOURNAL},
+
+	{"", 0x0}		/*  This must be the last flag.  */
+};
+
+struct pool_disk {
+	uint64_t pl_magic;	/* Pool magic number */
+	uint64_t pl_pool_id;	/* Unique pool identifier */
+	char pl_pool_name[POOL_NAME_SIZE];	/* Name of pool */
+	uint32_t pl_version;	/* Pool version */
+	uint32_t pl_subpools;	/* Number of subpools in this pool */
+	uint32_t pl_sp_id;	/* Subpool number within pool */
+	uint32_t pl_sp_devs;	/* Number of data partitions in this subpool */
+	uint32_t pl_sp_devid;	/* Partition number within subpool */
+	uint32_t pl_sp_type;	/* Partition type */
+	uint64_t pl_blocks;	/* Number of blocks in this partition */
+	uint32_t pl_striping;	/* Striping size within subpool */
+	/*
+	 * If the number of DMEP devices is zero, then the next field **
+	 * ** (pl_sp_dmepid) becomes the subpool ID for redirection.  In **
+	 * ** other words, if this subpool does not have the capability  **
+	 * ** to do DMEP, then it must specify which subpool will do it  **
+	 * ** in it's place
+	 */
+
+	/*
+	 * While the next 3 field are no longer used, they must stay to keep **
+	 * ** backward compatibility...........................................
+	 */
+	uint32_t pl_sp_dmepdevs;/* Number of dmep devices in this subpool */
+	uint32_t pl_sp_dmepid;	/* Dmep device number within subpool */
+	uint32_t pl_sp_weight;	/* if dmep dev, pref to using it */
+
+	uint32_t pl_minor;	/* the pool minor number */
+	uint32_t pl_padding;	/* reminder - think about alignment */
+
+	/*
+	 * Even though we're zeroing out 8k at the front of the disk before
+	 * writing the label, putting this in
+	 */
+	char pl_reserve[184];	/* bump the structure size out to 512 bytes */
+};
+
+struct pool_list {
+	struct list list;
+	struct pool_disk pd;
+	struct physical_volume *pv;
+	struct id pv_uuid;
+	struct device *dev;
+};
+
+struct user_subpool {
+	uint32_t initialized;
+	uint32_t id;
+	uint32_t striping;
+	uint32_t num_devs;
+	uint32_t type;
+	uint32_t dummy;
+	struct user_device *devs;
+};
+
+struct user_device {
+	uint32_t initialized;
+	uint32_t sp_id;
+	uint32_t devid;
+	uint32_t dummy;
+	uint64_t blocks;
+	struct physical_volume *pv;
+};
+
+int read_pool_label(struct pool_list *pl, struct labeller *l,
+		    struct device *dev, char *buf, struct label **label);
+void pool_label_out(struct pool_disk *pl, char *buf);
+void pool_label_in(struct pool_disk *pl, char *buf);
+void get_pool_uuid(char *uuid, uint64_t poolid, uint32_t spid, uint32_t devid);
+int import_pool_vg(struct volume_group *vg, struct pool *mem, struct list *pls);
+int import_pool_lvs(struct volume_group *vg, struct pool *mem,
+		    struct list *pls);
+int import_pool_pvs(const struct format_type *fmt, struct volume_group *vg,
+		    struct list *pvs, struct pool *mem, struct list *pls);
+int import_pool_pv(const struct format_type *fmt, struct pool *mem,
+		   struct volume_group *vg, struct physical_volume *pv,
+		   struct pool_list *pl);
+int import_pool_segments(struct list *lvs, struct pool *mem,
+			 struct user_subpool *usp, int sp_count);
+int read_pool_pds(const struct format_type *fmt, const char *vgname,
+		  struct pool *mem, struct list *head);
+struct pool_list *read_pool_disk(const struct format_type *fmt,
+				 struct device *dev, struct pool *mem,
+				 const char *vg_name);
+
+#endif				/* DISK_REP_POOL_FORMAT_H */

lib/format_pool/format_pool.c

@@ -0,0 +1,363 @@
+/*
+ * Copyright (C) 1997-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU General Public License v.2.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include "lib.h"
+#include "pool.h"
+#include "label.h"
+#include "metadata.h"
+#include "hash.h"
+#include "limits.h"
+#include "list.h"
+#include "display.h"
+#include "toolcontext.h"
+#include "lvmcache.h"
+#include "disk_rep.h"
+#include "format_pool.h"
+#include "pool_label.h"
+
+#define FMT_POOL_NAME "pool"
+
+/* Must be called after pvs are imported */
+static struct user_subpool *_build_usp(struct list *pls, struct pool *mem,
+				       int *sps)
+{
+
+	struct list *plhs;
+	struct pool_list *pl;
+	struct user_subpool *usp = NULL, *cur_sp = NULL;
+	struct user_device *cur_dev = NULL;
+
+	/*
+	 * FIXME: Need to do some checks here - I'm tempted to add a
+	 * user_pool structure and build the entire thing to check against.
+	 */
+	list_iterate(plhs, pls) {
+		pl = list_item(plhs, struct pool_list);
+
+		*sps = pl->pd.pl_subpools;
+		if (!usp && (!(usp = pool_zalloc(mem, sizeof(*usp) * (*sps))))) {
+			log_error("Unable to allocate %d subpool structures",
+				  *sps);
+			return 0;
+		}
+
+		if (cur_sp != &usp[pl->pd.pl_sp_id]) {
+			cur_sp = &usp[pl->pd.pl_sp_id];
+
+			cur_sp->id = pl->pd.pl_sp_id;
+			cur_sp->striping = pl->pd.pl_striping;
+			cur_sp->num_devs = pl->pd.pl_sp_devs;
+			cur_sp->type = pl->pd.pl_sp_type;
+			cur_sp->initialized = 1;
+		}
+
+		if (!cur_sp->devs &&
+		    (!(cur_sp->devs =
+		       pool_zalloc(mem,
+				   sizeof(*usp->devs) * pl->pd.pl_sp_devs)))) {
+
+			log_error("Unable to allocate %d pool_device "
+				  "structures", pl->pd.pl_sp_devs);
+			return 0;
+		}
+		cur_dev = &cur_sp->devs[pl->pd.pl_sp_devid];
+		cur_dev->sp_id = cur_sp->id;
+		cur_dev->devid = pl->pd.pl_sp_id;
+		cur_dev->blocks = pl->pd.pl_blocks;
+		cur_dev->pv = pl->pv;
+		cur_dev->initialized = 1;
+
+	}
+
+	return usp;
+}
+
+static int _check_usp(char *vgname, struct user_subpool *usp, int sp_count)
+{
+	int i, j;
+
+	for (i = 0; i < sp_count; i++) {
+		if (!usp[i].initialized) {
+			log_error("Missing subpool %d in pool %s", i, vgname);
+			return 0;
+		}
+		for (j = 0; j < usp[i].num_devs; j++) {
+			if (!usp[i].devs[j].initialized) {
+				log_error("Missing device %d for subpool %d"
+					  " in pool %s", j, i, vgname);
+				return 0;
+			}
+
+		}
+	}
+
+	return 1;
+}
+
+static struct volume_group *_build_vg_from_pds(struct format_instance
+					       *fid, struct pool *mem,
+					       struct list *pds)
+{
+	struct pool *smem = fid->fmt->cmd->mem;
+	struct volume_group *vg = NULL;
+	struct user_subpool *usp = NULL;
+	int sp_count;
+
+	if (!(vg = pool_zalloc(smem, sizeof(*vg)))) {
+		log_error("Unable to allocate volume group structure");
+		return NULL;
+	}
+
+	vg->cmd = fid->fmt->cmd;
+	vg->fid = fid;
+	vg->name = NULL;
+	vg->status = 0;
+	vg->extent_count = 0;
+	vg->pv_count = 0;
+	vg->lv_count = 0;
+	vg->snapshot_count = 0;
+	vg->seqno = 1;
+	vg->system_id = NULL;
+	list_init(&vg->pvs);
+	list_init(&vg->lvs);
+	list_init(&vg->snapshots);
+	list_init(&vg->tags);
+
+	if (!import_pool_vg(vg, smem, pds)) {
+		stack;
+		return NULL;
+	}
+
+	if (!import_pool_pvs(fid->fmt, vg, &vg->pvs, smem, pds)) {
+		stack;
+		return NULL;
+	}
+
+	if (!import_pool_lvs(vg, smem, pds)) {
+		stack;
+		return NULL;
+	}
+
+	/*
+	 * I need an intermediate subpool structure that contains all the
+	 * relevant info for this.  Then i can iterate through the subpool
+	 * structures for checking, and create the segments
+	 */
+	if (!(usp = _build_usp(pds, mem, &sp_count))) {
+		stack;
+		return NULL;
+	}
+
+	/*
+	 * check the subpool structures - we can't handle partial VGs in
+	 * the pool format, so this will error out if we're missing PVs
+	 */
+	if (!_check_usp(vg->name, usp, sp_count)) {
+		stack;
+		return NULL;
+	}
+
+	if (!import_pool_segments(&vg->lvs, smem, usp, sp_count)) {
+		stack;
+		return NULL;
+	}
+
+	return vg;
+}
+
+static struct volume_group *_vg_read(struct format_instance *fid,
+				     const char *vg_name,
+				     struct metadata_area *mda)
+{
+	struct pool *mem = pool_create(1024);
+	struct list pds;
+	struct volume_group *vg = NULL;
+
+	list_init(&pds);
+
+	/* We can safely ignore the mda passed in */
+
+	if (!mem) {
+		stack;
+		return NULL;
+	}
+
+	/* Strip dev_dir if present */
+	vg_name = strip_dir(vg_name, fid->fmt->cmd->dev_dir);
+
+	/* Read all the pvs in the vg */
+	if (!read_pool_pds(fid->fmt, vg_name, mem, &pds)) {
+		stack;
+		goto out;
+	}
+
+	/* Do the rest of the vg stuff */
+	if (!(vg = _build_vg_from_pds(fid, mem, &pds))) {
+		stack;
+		goto out;
+	}
+
+      out:
+	pool_destroy(mem);
+	return vg;
+}
+
+static int _pv_setup(const struct format_type *fmt,
+		     uint64_t pe_start, uint32_t extent_count,
+		     uint32_t extent_size,
+		     int pvmetadatacopies,
+		     uint64_t pvmetadatasize, struct list *mdas,
+		     struct physical_volume *pv, struct volume_group *vg)
+{
+	return 1;
+}
+
+static int _pv_read(const struct format_type *fmt, const char *pv_name,
+		    struct physical_volume *pv, struct list *mdas)
+{
+	struct pool *mem = pool_create(1024);
+	struct pool_list *pl;
+	struct device *dev;
+	int r = 0;
+
+	log_very_verbose("Reading physical volume data %s from disk", pv_name);
+
+	if (!mem) {
+		stack;
+		return 0;
+	}
+
+	if (!(dev = dev_cache_get(pv_name, fmt->cmd->filter))) {
+		stack;
+		goto out;
+	}
+
+	/*
+	 * I need to read the disk and populate a pv structure here
+	 * I'll probably need to abstract some of this later for the
+	 * vg_read code
+	 */
+	if (!(pl = read_pool_disk(fmt, dev, mem, NULL))) {
+		stack;
+		goto out;
+	}
+
+	if (!import_pool_pv(fmt, fmt->cmd->mem, NULL, pv, pl)) {
+		stack;
+		goto out;
+	}
+
+	pv->fmt = fmt;
+
+	r = 1;
+
+      out:
+	pool_destroy(mem);
+	return r;
+}
+
+/* *INDENT-OFF* */
+static struct metadata_area_ops _metadata_format_pool_ops = {
+	vg_read:_vg_read,
+};
+/* *INDENT-ON* */
+
+static struct format_instance *_create_instance(const struct format_type *fmt,
+						const char *vgname,
+						void *private)
+{
+	struct format_instance *fid;
+	struct metadata_area *mda;
+
+	if (!(fid = pool_zalloc(fmt->cmd->mem, sizeof(*fid)))) {
+		log_error("Unable to allocate format instance structure for "
+			  "pool format");
+		return NULL;
+	}
+
+	fid->fmt = fmt;
+	list_init(&fid->metadata_areas);
+
+	/* Define a NULL metadata area */
+	if (!(mda = pool_zalloc(fmt->cmd->mem, sizeof(*mda)))) {
+		log_error("Unable to allocate metadata area structure "
+			  "for pool format");
+		pool_free(fmt->cmd->mem, fid);
+		return NULL;
+	}
+
+	mda->ops = &_metadata_format_pool_ops;
+	mda->metadata_locn = NULL;
+	list_add(&fid->metadata_areas, &mda->list);
+
+	return fid;
+}
+
+static void _destroy_instance(struct format_instance *fid)
+{
+	return;
+}
+
+static void _destroy(const struct format_type *fmt)
+{
+	dbg_free((void *) fmt);
+}
+
+/* *INDENT-OFF* */
+static struct format_handler _format_pool_ops = {
+	pv_read:_pv_read,
+	pv_setup:_pv_setup,
+	create_instance:_create_instance,
+	destroy_instance:_destroy_instance,
+	destroy:_destroy,
+};
+/* *INDENT-ON */
+
+#ifdef POOL_INTERNAL
+struct format_type *init_pool_format(struct cmd_context *cmd)
+#else				/* Shared */
+struct format_type *init_format(struct cmd_context *cmd);
+struct format_type *init_format(struct cmd_context *cmd)
+#endif
+{
+	struct format_type *fmt = dbg_malloc(sizeof(*fmt));
+
+	if (!fmt) {
+		log_error("Unable to allocate format type structure for pool "
+			  "format");
+		return NULL;
+	}
+
+	fmt->cmd = cmd;
+	fmt->ops = &_format_pool_ops;
+	fmt->name = FMT_POOL_NAME;
+	fmt->alias = NULL;
+	fmt->features = 0;
+	fmt->private = NULL;
+
+	if (!(fmt->labeller = pool_labeller_create(fmt))) {
+		log_error("Couldn't create pool label handler.");
+		return NULL;
+	}
+
+	if (!(label_register_handler(FMT_POOL_NAME, fmt->labeller))) {
+		log_error("Couldn't register pool label handler.");
+		return NULL;
+	}
+
+	log_very_verbose("Initialised format: %s", fmt->name);
+
+	return fmt;
+}

lib/format_pool/format_pool.h

@@ -0,0 +1,25 @@
+/*
+ * Copyright (C) 2003-2004 Sistina Software, Inc. All rights reserved.
+ * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU General Public License v.2.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#ifndef _LVM_FORMAT_POOL_H
+#define _LVM_FORMAT_POOL_H
+
+#include "metadata.h"
+
+#ifdef POOL_INTERNAL
+struct format_type *init_pool_format(struct cmd_context *cmd);
+#endif
+
+#endif

lib/format_pool/import_export.c

@@ -0,0 +1,309 @@
+/*
+ * Copyright (C) 1997-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU General Public License v.2.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include "lib.h"
+#include "pool.h"
+#include "label.h"
+#include "metadata.h"
+#include "lvmcache.h"
+#include "disk_rep.h"
+#include "lv_alloc.h"
+#include "str_list.h"
+#include "display.h"
+#include "segtype.h"
+
+/* This file contains only imports at the moment... */
+
+int import_pool_vg(struct volume_group *vg, struct pool *mem, struct list *pls)
+{
+	struct list *plhs;
+	struct pool_list *pl;
+
+	list_iterate(plhs, pls) {
+		pl = list_item(plhs, struct pool_list);
+
+		vg->extent_count +=
+		    ((pl->pd.pl_blocks) / POOL_PE_SIZE);
+
+		vg->pv_count++;
+
+		if (vg->name)
+			continue;
+
+		vg->name = pool_strdup(mem, pl->pd.pl_pool_name);
+		get_pool_vg_uuid(&vg->id, &pl->pd);
+		vg->extent_size = POOL_PE_SIZE;
+		vg->status |= LVM_READ | LVM_WRITE | CLUSTERED | SHARED;
+		vg->free_count = 0;
+		vg->max_lv = 1;
+		vg->max_pv = POOL_MAX_DEVICES;
+		vg->alloc = ALLOC_NORMAL;
+		vg->lv_count = 0;
+	}
+
+	return 1;
+}
+
+int import_pool_lvs(struct volume_group *vg, struct pool *mem, struct list *pls)
+{
+	struct pool_list *pl;
+	struct list *plhs;
+	struct lv_list *lvl = pool_zalloc(mem, sizeof(*lvl));
+	struct logical_volume *lv;
+
+	if (!lvl) {
+		log_error("Unable to allocate lv list structure");
+		return 0;
+	}
+
+	if (!(lvl->lv = pool_zalloc(mem, sizeof(*lvl->lv)))) {
+		log_error("Unable to allocate logical volume structure");
+		return 0;
+	}
+
+	lv = lvl->lv;
+	lv->status = 0;
+	lv->vg = vg;
+	lv->alloc = ALLOC_NORMAL;
+	lv->size = 0;
+	lv->name = NULL;
+	lv->le_count = 0;
+	lv->read_ahead = 0;
+	list_init(&lv->segments);
+	list_init(&lv->tags);
+
+	list_iterate(plhs, pls) {
+		pl = list_item(plhs, struct pool_list);
+
+		lv->size += pl->pd.pl_blocks;
+
+		if (lv->name)
+			continue;
+
+		if (!(lv->name = pool_strdup(mem, pl->pd.pl_pool_name))) {
+			stack;
+			return 0;
+		}
+
+		get_pool_lv_uuid(lv->lvid.id, &pl->pd);
+		log_debug("Calculated lv uuid for lv %s: %s", lv->name,
+			  lv->lvid.s);
+
+		lv->status |= VISIBLE_LV | LVM_READ | LVM_WRITE;
+		lv->major = POOL_MAJOR;
+
+		/* for pool a minor of 0 is dynamic */
+		if (pl->pd.pl_minor) {
+			lv->status |= FIXED_MINOR;
+			lv->minor = pl->pd.pl_minor;
+		} else {
+			lv->minor = -1;
+		}
+		list_init(&lv->segments);
+		list_init(&lv->tags);
+	}
+
+	lv->le_count = lv->size / POOL_PE_SIZE;
+	lvl->lv = lv;
+	list_add(&vg->lvs, &lvl->list);
+	vg->lv_count++;
+
+	return 1;
+}
+
+int import_pool_pvs(const struct format_type *fmt, struct volume_group *vg,
+		    struct list *pvs, struct pool *mem, struct list *pls)
+{
+	struct pv_list *pvl;
+	struct pool_list *pl;
+	struct list *plhs;
+
+	list_iterate(plhs, pls) {
+		pl = list_item(plhs, struct pool_list);
+
+		if (!(pvl = pool_zalloc(mem, sizeof(*pvl)))) {
+			log_error("Unable to allocate pv list structure");
+			return 0;
+		}
+		if (!(pvl->pv = pool_zalloc(mem, sizeof(*pvl->pv)))) {
+			log_error("Unable to allocate pv structure");
+			return 0;
+		}
+		if (!import_pool_pv(fmt, mem, vg, pvl->pv, pl)) {
+			return 0;
+		}
+		pl->pv = pvl->pv;
+		pvl->mdas = NULL;
+		pvl->pe_ranges = NULL;
+		list_add(pvs, &pvl->list);
+	}
+
+	return 1;
+}
+
+int import_pool_pv(const struct format_type *fmt, struct pool *mem,
+		   struct volume_group *vg, struct physical_volume *pv,
+		   struct pool_list *pl)
+{
+	struct pool_disk *pd = &pl->pd;
+
+	memset(pv, 0, sizeof(*pv));
+
+	get_pool_pv_uuid(&pv->id, pd);
+	pv->fmt = fmt;
+
+	pv->dev = pl->dev;
+	if (!(pv->vg_name = pool_strdup(mem, pd->pl_pool_name))) {
+		log_error("Unable to duplicate vg_name string");
+		return 0;
+	}
+	pv->status = 0;
+	pv->size = pd->pl_blocks;
+	pv->pe_size = POOL_PE_SIZE;
+	pv->pe_start = POOL_PE_START;
+	pv->pe_count = pv->size / POOL_PE_SIZE;
+	pv->pe_alloc_count = pv->pe_count;
+
+	list_init(&pv->tags);
+
+	return 1;
+}
+
+static const char *_cvt_sptype(uint32_t sptype)
+{
+	int i;
+	for (i = 0; sptype_names[i].name[0]; i++) {
+		if (sptype == sptype_names[i].label) {
+			break;
+		}
+	}
+	log_debug("Found sptype %X and converted it to %s",
+		  sptype, sptype_names[i].name);
+	return sptype_names[i].name;
+}
+
+static int _add_stripe_seg(struct pool *mem,
+			   struct user_subpool *usp, struct logical_volume *lv,
+			   uint32_t *le_cur)
+{
+	struct lv_segment *seg;
+	int j;
+
+	if (!(seg = alloc_lv_segment(mem, usp->num_devs))) {
+		log_error("Unable to allocate striped lv_segment structure");
+		return 0;
+	}
+	if(usp->striping & (usp->striping - 1)) {
+		log_error("Stripe size must be a power of 2");
+		return 0;
+	}
+	seg->stripe_size = usp->striping;
+	seg->status |= 0;
+	seg->le += *le_cur;
+
+	/* add the subpool type to the segment tag list */
+	str_list_add(mem, &seg->tags, _cvt_sptype(usp->type));
+
+	for (j = 0; j < usp->num_devs; j++) {
+		if (!(seg->segtype = get_segtype_from_string(lv->vg->cmd,
+							     "striped"))) {
+			stack;
+			return 0;
+		}
+
+		seg->area_len = (usp->devs[j].blocks) / POOL_PE_SIZE;
+		seg->len += seg->area_len;
+		*le_cur += seg->area_len;
+		seg->lv = lv;
+
+		seg->area[j].type = AREA_PV;
+		seg->area[j].u.pv.pv = usp->devs[j].pv;
+		seg->area[j].u.pv.pe = 0;
+	}
+	list_add(&lv->segments, &seg->list);
+	return 1;
+}
+
+static int _add_linear_seg(struct pool *mem,
+			   struct user_subpool *usp, struct logical_volume *lv,
+			   uint32_t *le_cur)
+{
+	struct lv_segment *seg;
+	int j;
+
+	for (j = 0; j < usp->num_devs; j++) {
+		/* linear segments only have 1 data area */
+		if (!(seg = alloc_lv_segment(mem, 1))) {
+			log_error("Unable to allocate linear lv_segment "
+				  "structure");
+			return 0;
+		}
+		seg->stripe_size = usp->striping;
+		seg->le += *le_cur;
+		seg->chunk_size = POOL_PE_SIZE;
+		seg->status |= 0;
+		if (!(seg->segtype = get_segtype_from_string(lv->vg->cmd,
+							     "striped"))) {
+			stack;
+			return 0;
+		}
+		/* add the subpool type to the segment tag list */
+		str_list_add(mem, &seg->tags, _cvt_sptype(usp->type));
+
+		seg->lv = lv;
+
+		seg->area_len = (usp->devs[j].blocks) / POOL_PE_SIZE;
+		seg->len = seg->area_len;
+		*le_cur += seg->len;
+		seg->area[0].type = AREA_PV;
+		seg->area[0].u.pv.pv = usp->devs[j].pv;
+		seg->area[0].u.pv.pe = 0;
+		list_add(&lv->segments, &seg->list);
+	}
+	return 1;
+}
+
+int import_pool_segments(struct list *lvs, struct pool *mem,
+			 struct user_subpool *usp, int subpools)
+{
+
+	struct list *lvhs;
+	struct lv_list *lvl;
+	struct logical_volume *lv;
+	uint32_t le_cur = 0;
+	int i;
+
+	list_iterate(lvhs, lvs) {
+		lvl = list_item(lvhs, struct lv_list);
+
+		lv = lvl->lv;
+		for (i = 0; i < subpools; i++) {
+			if (usp[i].striping) {
+				if (!_add_stripe_seg(mem, &usp[i], lv, &le_cur)) {
+					stack;
+					return 0;
+				}
+			} else {
+				if (!_add_linear_seg(mem, &usp[i], lv, &le_cur)) {
+					stack;
+					return 0;
+				}
+			}
+		}
+	}
+
+	return 1;
+
+}

lib/format_pool/pool_label.c

@@ -0,0 +1,108 @@
+/*
+ * Copyright (C) 1997-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU General Public License v.2.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include "lib.h"
+#include "pool.h"
+#include "label.h"
+#include "metadata.h"
+#include "xlate.h"
+#include "disk_rep.h"
+#include "pool_label.h"
+
+#include <sys/stat.h>
+#include <fcntl.h>
+
+static void _not_supported(const char *op)
+{
+	log_error("The '%s' operation is not supported for the pool labeller.",
+		  op);
+}
+
+static int _can_handle(struct labeller *l, char *buf, uint64_t sector)
+{
+
+	struct pool_disk pd;
+
+	/*
+	 * POOL label must always be in first sector
+	 */
+	if (sector)
+		return 0;
+
+	pool_label_in(&pd, buf);
+
+	/* can ignore 8 rightmost bits for ondisk format check */
+	if ((pd.pl_magic == POOL_MAGIC) &&
+	    (pd.pl_version >> 8 == POOL_VERSION >> 8))
+		return 1;
+
+	return 0;
+}
+
+static int _write(struct label *label, char *buf)
+{
+	_not_supported("write");
+	return 0;
+}
+
+static int _read(struct labeller *l, struct device *dev, char *buf,
+		 struct label **label)
+{
+	struct pool_list pl;
+
+	return read_pool_label(&pl, l, dev, buf, label);
+}
+
+static int _initialise_label(struct labeller *l, struct label *label)
+{
+	strcpy(label->type, "POOL");
+
+	return 1;
+}
+
+static void _destroy_label(struct labeller *l, struct label *label)
+{
+	return;
+}
+
+static void _destroy(struct labeller *l)
+{
+	dbg_free(l);
+}
+
+struct label_ops _pool_ops = {
+      can_handle:_can_handle,
+      write:_write,
+      read:_read,
+      verify:_can_handle,
+      initialise_label:_initialise_label,
+      destroy_label:_destroy_label,
+      destroy:_destroy
+};
+
+struct labeller *pool_labeller_create(struct format_type *fmt)
+{
+	struct labeller *l;
+
+	if (!(l = dbg_malloc(sizeof(*l)))) {
+		log_error("Couldn't allocate labeller object.");
+		return NULL;
+	}
+
+	l->ops = &_pool_ops;
+	l->private = (const void *) fmt;
+
+	return l;
+}

lib/format_pool/pool_label.h

@@ -0,0 +1,23 @@
+/*
+ * Copyright (C) 1997-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU General Public License v.2.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#ifndef _LVM_POOL_LABEL_H
+#define _LVM_POOL_LABEL_H
+
+#include "metadata.h"
+
+struct labeller *pool_labeller_create(struct format_type *fmt);
+
+#endif

lib/format_text/archive.c

@@ -380,4 +380,3 @@ int backup_list(struct cmd_context *cmd, const char *dir, const char *vgname)
 
 	return 1;
 }
-

lib/format_text/export.c

@@ -20,6 +20,8 @@
 #include "pool.h"
 #include "display.h"
 #include "lvm-string.h"
+#include "segtype.h"
+#include "text_export.h"
 
 #include <stdarg.h>
 #include <time.h>
@@ -74,14 +76,6 @@ static void _init(void)
 /*
  * Formatting functions.
  */
-static int _out_size(struct formatter *f, uint64_t size, const char *fmt, ...)
-    __attribute__ ((format(printf, 3, 4)));
-
-static int _out_hint(struct formatter *f, const char *fmt, ...)
-    __attribute__ ((format(printf, 2, 3)));
-
-static int _out(struct formatter *f, const char *fmt, ...)
-    __attribute__ ((format(printf, 2, 3)));
 
 #define MAX_INDENT 5
 static void _inc_indent(struct formatter *f)
@@ -206,7 +200,7 @@ static int _sectors_to_units(uint64_t sectors, char *buffer, size_t s)
  * Appends a comment giving a size in more easily
  * readable form (eg, 4M instead of 8096).
  */
-static int _out_size(struct formatter *f, uint64_t size, const char *fmt, ...)
+int out_size(struct formatter *f, uint64_t size, const char *fmt, ...)
 {
 	char buffer[64];
 	va_list ap;
@@ -226,7 +220,7 @@ static int _out_size(struct formatter *f, uint64_t size, const char *fmt, ...)
  * Appends a comment indicating that the line is
  * only a hint.
  */
-static int _out_hint(struct formatter *f, const char *fmt, ...)
+int out_hint(struct formatter *f, const char *fmt, ...)
 {
 	va_list ap;
 	int r;
@@ -241,7 +235,7 @@ static int _out_hint(struct formatter *f, const char *fmt, ...)
 /*
  * The normal output function.
  */
-static int _out(struct formatter *f, const char *fmt, ...)
+int out_text(struct formatter *f, const char *fmt, ...)
 {
 	va_list ap;
 	int r;
@@ -253,8 +247,6 @@ static int _out(struct formatter *f, const char *fmt, ...)
 	return r;
 }
 
-#define _outf(args...) do {if (!_out(args)) {stack; return 0;}} while (0)
-
 static int _print_header(struct formatter *f,
 			 struct volume_group *vg, const char *desc)
 {
@@ -262,17 +254,17 @@ static int _print_header(struct formatter *f,
 
 	t = time(NULL);
 
-	_outf(f, "# Generated by LVM2: %s", ctime(&t));
-	_outf(f, CONTENTS_FIELD " = \"" CONTENTS_VALUE "\"");
-	_outf(f, FORMAT_VERSION_FIELD " = %d", FORMAT_VERSION_VALUE);
+	outf(f, "# Generated by LVM2: %s", ctime(&t));
+	outf(f, CONTENTS_FIELD " = \"" CONTENTS_VALUE "\"");
+	outf(f, FORMAT_VERSION_FIELD " = %d", FORMAT_VERSION_VALUE);
 	f->nl(f);
 
-	_outf(f, "description = \"%s\"", desc);
+	outf(f, "description = \"%s\"", desc);
 	f->nl(f);
-	_outf(f, "creation_host = \"%s\"\t# %s %s %s %s %s", _utsname.nodename,
-	      _utsname.sysname, _utsname.nodename, _utsname.release,
-	      _utsname.version, _utsname.machine);
-	_outf(f, "creation_time = %lu\t# %s", t, ctime(&t));
+	outf(f, "creation_host = \"%s\"\t# %s %s %s %s %s", _utsname.nodename,
+	     _utsname.sysname, _utsname.nodename, _utsname.release,
+	     _utsname.version, _utsname.machine);
+	outf(f, "creation_time = %lu\t# %s", t, ctime(&t));
 
 	return 1;
 }
@@ -286,34 +278,41 @@ static int _print_vg(struct formatter *f, struct volume_group *vg)
 		return 0;
 	}
 
-	_outf(f, "id = \"%s\"", buffer);
+	outf(f, "id = \"%s\"", buffer);
 
-	_outf(f, "seqno = %u", vg->seqno);
+	outf(f, "seqno = %u", vg->seqno);
 
 	if (!print_flags(vg->status, VG_FLAGS, buffer, sizeof(buffer))) {
 		stack;
 		return 0;
 	}
-	_outf(f, "status = %s", buffer);
+	outf(f, "status = %s", buffer);
 
 	if (!list_empty(&vg->tags)) {
 		if (!print_tags(&vg->tags, buffer, sizeof(buffer))) {
 			stack;
 			return 0;
 		}
-		_outf(f, "tags = %s", buffer);
+		outf(f, "tags = %s", buffer);
 	}
 
 	if (vg->system_id && *vg->system_id)
-		_outf(f, "system_id = \"%s\"", vg->system_id);
+		outf(f, "system_id = \"%s\"", vg->system_id);
 
-	if (!_out_size(f, (uint64_t) vg->extent_size, "extent_size = %u",
-		       vg->extent_size)) {
+	if (!out_size(f, (uint64_t) vg->extent_size, "extent_size = %u",
+		      vg->extent_size)) {
 		stack;
 		return 0;
 	}
-	_outf(f, "max_lv = %u", vg->max_lv);
-	_outf(f, "max_pv = %u", vg->max_pv);
+	outf(f, "max_lv = %u", vg->max_lv);
+	outf(f, "max_pv = %u", vg->max_pv);
+
+	/* Default policy is NORMAL; INHERIT is meaningless */
+	if (vg->alloc != ALLOC_NORMAL && vg->alloc != ALLOC_INHERIT) {
+		f->nl(f);
+		outf(f, "allocation_policy = \"%s\"",
+		     get_alloc_string(vg->alloc));
+	}
 
 	return 1;
 }
@@ -336,7 +335,7 @@ static int _print_pvs(struct formatter *f, struct volume_group *vg)
 	char buffer[4096];
 	const char *name;
 
-	_outf(f, "physical_volumes {");
+	outf(f, "physical_volumes {");
 	_inc_indent(f);
 
 	list_iterate(pvh, &vg->pvs) {
@@ -348,7 +347,7 @@ static int _print_pvs(struct formatter *f, struct volume_group *vg)
 		}
 
 		f->nl(f);
-		_outf(f, "%s {", name);
+		outf(f, "%s {", name);
 		_inc_indent(f);
 
 		if (!id_write_format(&pv->id, buffer, sizeof(buffer))) {
@@ -356,8 +355,8 @@ static int _print_pvs(struct formatter *f, struct volume_group *vg)
 			return 0;
 		}
 
-		_outf(f, "id = \"%s\"", buffer);
-		if (!_out_hint(f, "device = \"%s\"", dev_name(pv->dev))) {
+		outf(f, "id = \"%s\"", buffer);
+		if (!out_hint(f, "device = \"%s\"", dev_name(pv->dev))) {
 			stack;
 			return 0;
 		}
@@ -367,119 +366,106 @@ static int _print_pvs(struct formatter *f, struct volume_group *vg)
 			stack;
 			return 0;
 		}
-		_outf(f, "status = %s", buffer);
+		outf(f, "status = %s", buffer);
 
 		if (!list_empty(&pv->tags)) {
 			if (!print_tags(&pv->tags, buffer, sizeof(buffer))) {
 				stack;
 				return 0;
 			}
-			_outf(f, "tags = %s", buffer);
+			outf(f, "tags = %s", buffer);
 		}
 
-		_outf(f, "pe_start = %" PRIu64, pv->pe_start);
-		if (!_out_size(f, vg->extent_size * (uint64_t) pv->pe_count,
-			       "pe_count = %u", pv->pe_count)) {
+		outf(f, "pe_start = %" PRIu64, pv->pe_start);
+		if (!out_size(f, vg->extent_size * (uint64_t) pv->pe_count,
+			      "pe_count = %u", pv->pe_count)) {
 			stack;
 			return 0;
 		}
 
 		_dec_indent(f);
-		_outf(f, "}");
+		outf(f, "}");
 	}
 
 	_dec_indent(f);
-	_outf(f, "}");
+	outf(f, "}");
 	return 1;
 }
 
 static int _print_segment(struct formatter *f, struct volume_group *vg,
 			  int count, struct lv_segment *seg)
 {
-	unsigned int s;
-	const char *name;
-	const char *type;
 	char buffer[4096];
 
-	_outf(f, "segment%u {", count);
+	outf(f, "segment%u {", count);
 	_inc_indent(f);
 
-	_outf(f, "start_extent = %u", seg->le);
-	if (!_out_size(f, (uint64_t) seg->len * vg->extent_size,
-		       "extent_count = %u", seg->len)) {
+	outf(f, "start_extent = %u", seg->le);
+	if (!out_size(f, (uint64_t) seg->len * vg->extent_size,
+		      "extent_count = %u", seg->len)) {
 		stack;
 		return 0;
 	}
 
 	f->nl(f);
-	_outf(f, "type = \"%s\"", get_segtype_string(seg->type));
+	outf(f, "type = \"%s\"", seg->segtype->name);
 
 	if (!list_empty(&seg->tags)) {
 		if (!print_tags(&seg->tags, buffer, sizeof(buffer))) {
 			stack;
 			return 0;
 		}
-		_outf(f, "tags = %s", buffer);
+		outf(f, "tags = %s", buffer);
 	}
 
-	switch (seg->type) {
-	case SEG_SNAPSHOT:
-		_outf(f, "chunk_size = %u", seg->chunk_size);
-		_outf(f, "origin = \"%s\"", seg->origin->name);
-		_outf(f, "cow_store = \"%s\"", seg->cow->name);
-		break;
-
-	case SEG_MIRRORED:
-	case SEG_STRIPED:
-		type = (seg->type == SEG_MIRRORED) ? "mirror" : "stripe";
-		_outf(f, "%s_count = %u%s", type, seg->area_count,
-		      (seg->area_count == 1) ? "\t# linear" : "");
-
-		if ((seg->type == SEG_MIRRORED) && (seg->status & PVMOVE))
-			_out_size(f, (uint64_t) seg->extents_moved,
-				  "extents_moved = %u", seg->extents_moved);
-
-		if ((seg->type == SEG_STRIPED) && (seg->area_count > 1))
-			_out_size(f, (uint64_t) seg->stripe_size,
-				  "stripe_size = %u", seg->stripe_size);
-
-		f->nl(f);
-
-		_outf(f, "%ss = [", type);
-		_inc_indent(f);
-
-		for (s = 0; s < seg->area_count; s++) {
-			switch (seg->area[s].type) {
-			case AREA_PV:
-				if (!(name = _get_pv_name(f, seg->
-							  area[s].u.pv.pv))) {
-					stack;
-					return 0;
-				}
-
-				_outf(f, "\"%s\", %u%s", name,
-				      seg->area[s].u.pv.pe,
-				      (s == seg->area_count - 1) ? "" : ",");
-				break;
-			case AREA_LV:
-				_outf(f, "\"%s\", %u%s",
-				      seg->area[s].u.lv.lv->name,
-				      seg->area[s].u.lv.le,
-				      (s == seg->area_count - 1) ? "" : ",");
-			}
-		}
-
-		_dec_indent(f);
-		_outf(f, "]");
-		break;
+	if (seg->segtype->ops->text_export &&
+	    !seg->segtype->ops->text_export(seg, f)) {
+		stack;
+		return 0;
 	}
 
 	_dec_indent(f);
-	_outf(f, "}");
+	outf(f, "}");
 
 	return 1;
 }
 
+int out_areas(struct formatter *f, const struct lv_segment *seg,
+	      const char *type)
+{
+	const char *name;
+	unsigned int s;
+
+	f->nl(f);
+
+	outf(f, "%ss = [", type);
+	_inc_indent(f);
+
+	for (s = 0; s < seg->area_count; s++) {
+		switch (seg->area[s].type) {
+		case AREA_PV:
+			if (!(name = _get_pv_name(f, seg->area[s].u.pv.pv))) {
+				stack;
+				return 0;
+			}
+
+			outf(f, "\"%s\", %u%s", name,
+			     seg->area[s].u.pv.pe,
+			     (s == seg->area_count - 1) ? "" : ",");
+			break;
+		case AREA_LV:
+			outf(f, "\"%s\", %u%s",
+			     seg->area[s].u.lv.lv->name,
+			     seg->area[s].u.lv.le,
+			     (s == seg->area_count - 1) ? "" : ",");
+		}
+	}
+
+	_dec_indent(f);
+	outf(f, "]");
+	return 1;
+}
+
 static int _count_segments(struct logical_volume *lv)
 {
 	int r = 0;
@@ -499,7 +485,7 @@ static int _print_snapshot(struct formatter *f, struct snapshot *snap,
 
 	f->nl(f);
 
-	_outf(f, "snapshot%u {", count);
+	outf(f, "snapshot%u {", count);
 	_inc_indent(f);
 
 	if (!id_write_format(&snap->id, buffer, sizeof(buffer))) {
@@ -507,25 +493,39 @@ static int _print_snapshot(struct formatter *f, struct snapshot *snap,
 		return 0;
 	}
 
-	_outf(f, "id = \"%s\"", buffer);
-	if (!print_flags(LVM_READ | LVM_WRITE | VISIBLE_LV, LV_FLAGS,
-			 buffer, sizeof(buffer))) {
+	outf(f, "id = \"%s\"", buffer);
+
+	seg.status = LVM_READ | LVM_WRITE | VISIBLE_LV;
+	if (!print_flags(seg.status, LV_FLAGS, buffer, sizeof(buffer))) {
 		stack;
 		return 0;
 	}
 
-	_outf(f, "status = %s", buffer);
-	_outf(f, "segment_count = 1");
+	outf(f, "status = %s", buffer);
+	outf(f, "segment_count = 1");
 
 	f->nl(f);
 
-	seg.type = SEG_SNAPSHOT;
+	if (!(seg.segtype = get_segtype_from_string(snap->origin->vg->cmd,
+						    "snapshot"))) {
+		stack;
+		return 0;
+	}
+
 	seg.le = 0;
 	seg.len = snap->origin->le_count;
 	seg.origin = snap->origin;
 	seg.cow = snap->cow;
 	seg.chunk_size = snap->chunk_size;
 
+	/* FIXME Dummy values */
+	list_init(&seg.list);
+	seg.lv = snap->cow;
+	seg.stripe_size = 0;
+	seg.area_count = 0;
+	seg.area_len = 0;
+	seg.extents_copied = 0;
+
 	/* Can't tag a snapshot independently of its origin */
 	list_init(&seg.tags);
 
@@ -535,7 +535,7 @@ static int _print_snapshot(struct formatter *f, struct snapshot *snap,
 	}
 
 	_dec_indent(f);
-	_outf(f, "}");
+	outf(f, "}");
 
 	return 1;
 }
@@ -572,14 +572,14 @@ static int _print_lvs(struct formatter *f, struct volume_group *vg)
 	if (list_empty(&vg->lvs))
 		return 1;
 
-	_outf(f, "logical_volumes {");
+	outf(f, "logical_volumes {");
 	_inc_indent(f);
 
 	list_iterate(lvh, &vg->lvs) {
 		lv = list_item(lvh, struct lv_list)->lv;
 
 		f->nl(f);
-		_outf(f, "%s {", lv->name);
+		outf(f, "%s {", lv->name);
 		_inc_indent(f);
 
 		/* FIXME: Write full lvid */
@@ -588,32 +588,33 @@ static int _print_lvs(struct formatter *f, struct volume_group *vg)
 			return 0;
 		}
 
-		_outf(f, "id = \"%s\"", buffer);
+		outf(f, "id = \"%s\"", buffer);
 
 		if (!print_flags(lv->status, LV_FLAGS, buffer, sizeof(buffer))) {
 			stack;
 			return 0;
 		}
-		_outf(f, "status = %s", buffer);
+		outf(f, "status = %s", buffer);
 
 		if (!list_empty(&lv->tags)) {
 			if (!print_tags(&lv->tags, buffer, sizeof(buffer))) {
 				stack;
 				return 0;
 			}
-			_outf(f, "tags = %s", buffer);
+			outf(f, "tags = %s", buffer);
 		}
 
-		if (lv->alloc != ALLOC_DEFAULT)
-			_outf(f, "allocation_policy = \"%s\"",
-			      get_alloc_string(lv->alloc));
+		if (lv->alloc != ALLOC_INHERIT)
+			outf(f, "allocation_policy = \"%s\"",
+			     get_alloc_string(lv->alloc));
+
 		if (lv->read_ahead)
-			_outf(f, "read_ahead = %u", lv->read_ahead);
+			outf(f, "read_ahead = %u", lv->read_ahead);
 		if (lv->major >= 0)
-			_outf(f, "major = %d", lv->major);
+			outf(f, "major = %d", lv->major);
 		if (lv->minor >= 0)
-			_outf(f, "minor = %d", lv->minor);
-		_outf(f, "segment_count = %u", _count_segments(lv));
+			outf(f, "minor = %d", lv->minor);
+		outf(f, "segment_count = %u", _count_segments(lv));
 		f->nl(f);
 
 		seg_count = 1;
@@ -625,7 +626,7 @@ static int _print_lvs(struct formatter *f, struct volume_group *vg)
 		}
 
 		_dec_indent(f);
-		_outf(f, "}");
+		outf(f, "}");
 	}
 
 	if (!_print_snapshots(f, vg)) {
@@ -634,7 +635,7 @@ static int _print_lvs(struct formatter *f, struct volume_group *vg)
 	}
 
 	_dec_indent(f);
-	_outf(f, "}");
+	outf(f, "}");
 
 	return 1;
 }
@@ -707,7 +708,7 @@ static int _text_vg_export(struct formatter *f,
 	if (f->header && !_print_header(f, vg, desc))
 		fail;
 
-	if (!_out(f, "%s {", vg->name))
+	if (!out_text(f, "%s {", vg->name))
 		fail;
 
 	_inc_indent(f);
@@ -724,7 +725,7 @@ static int _text_vg_export(struct formatter *f,
 		fail;
 
 	_dec_indent(f);
-	if (!_out(f, "}"))
+	if (!out_text(f, "}"))
 		fail;
 
 	if (!f->header && !_print_header(f, vg, desc))
@@ -804,4 +805,4 @@ int text_vg_export_raw(struct volume_group *vg, const char *desc, char *buf,
 	return r;
 }
 
-#undef _outf
+#undef outf

lib/format_text/flags.c

@@ -52,6 +52,8 @@ static struct flag _lv_flags[] = {
 	{VISIBLE_LV, "VISIBLE"},
 	{PVMOVE, "PVMOVE"},
 	{LOCKED, "LOCKED"},
+	{MIRRORED, NULL},
+	{VIRTUAL, NULL},
 	{0, NULL}
 };
 
@@ -92,18 +94,21 @@ int print_flags(uint32_t status, int type, char *buffer, size_t size)
 
 	for (f = 0; flags[f].mask; f++) {
 		if (status & flags[f].mask) {
+			status &= ~flags[f].mask;
+
+			/* Internal-only flag? */
+			if (!flags[f].description)
+				continue;
+
 			if (!first) {
 				if (!emit_to_buffer(&buffer, &size, ", "))
 					return 0;
-
 			} else
 				first = 0;
-
+	
 			if (!emit_to_buffer(&buffer, &size, "\"%s\"",
-					    flags[f].description))
+			    flags[f].description))
 				return 0;
-
-			status &= ~flags[f].mask;
 		}
 	}
 

lib/format_text/format-text.c

@@ -30,6 +30,7 @@
 #include "xlate.h"
 #include "label.h"
 #include "memlock.h"
+#include "lvmcache.h"
 
 #include <unistd.h>
 #include <sys/file.h>
@@ -80,7 +81,7 @@ static int _lv_setup(struct format_instance *fid, struct logical_volume *lv)
 	uint64_t max_size = UINT_MAX;
 
 	if (lv->size > max_size) {
-		char *dummy = display_size(max_size / 2, SIZE_SHORT);
+		char *dummy = display_size(max_size, SIZE_SHORT);
 		log_error("logical volumes cannot be larger than %s", dummy);
 		dbg_free(dummy);
 		return 0;
@@ -928,8 +929,7 @@ static int _mda_setup(const struct format_type *fmt,
 	/* FIXME If creating new mdas, wipe them! */
 	if (mda_size1) {
 		if (!add_mda(fmt, fmt->cmd->mem, mdas, pv->dev, start1,
-			     mda_size1))
-			return 0;
+			     mda_size1)) return 0;
 
 		if (!dev_zero((struct device *) pv->dev, start1,
 			      (size_t) (mda_size1 >
@@ -976,8 +976,7 @@ static int _mda_setup(const struct format_type *fmt,
 
 	if (mda_size2) {
 		if (!add_mda(fmt, fmt->cmd->mem, mdas, pv->dev, start2,
-			     mda_size2))
-			return 0;
+			     mda_size2)) return 0;
 		if (!dev_zero(pv->dev, start2,
 			      (size_t) (mda_size1 >
 					wipe_size ? wipe_size : mda_size1))) {
@@ -1066,8 +1065,7 @@ static int _pv_write(const struct format_type *fmt, struct physical_volume *pv,
 		}
 	}
 	if (!add_da
-	    (fmt, NULL, &info->das, pv->pe_start << SECTOR_SHIFT,
-	     UINT64_C(0))) {
+	    (fmt, NULL, &info->das, pv->pe_start << SECTOR_SHIFT, UINT64_C(0))) {
 		stack;
 		return 0;
 	}
@@ -1139,8 +1137,7 @@ static int _add_raw(struct list *raw_list, struct device_area *dev_area)
 		rl = list_item(rlh, struct raw_list);
 		/* FIXME Check size/overlap consistency too */
 		if (rl->dev_area.dev == dev_area->dev &&
-		    rl->dev_area.start == dev_area->start)
-			return 1;
+		    rl->dev_area.start == dev_area->start) return 1;
 	}
 
 	if (!(rl = dbg_malloc(sizeof(struct raw_list)))) {
@@ -1334,11 +1331,10 @@ static int _pv_setup(const struct format_type *fmt,
 					    list_item(mdash,
 						      struct metadata_area);
 					if (mda2->ops !=
-					    &_metadata_text_raw_ops)
-						continue;
+					    &_metadata_text_raw_ops) continue;
 					mdac2 =
-					    (struct mda_context *) mda2->
-					    metadata_locn;
+					    (struct mda_context *)
+					    mda2->metadata_locn;
 					if (!memcmp
 					    (&mdac2->area, &mdac->area,
 					     sizeof(mdac->area))) {
@@ -1356,8 +1352,7 @@ static int _pv_setup(const struct format_type *fmt,
 				}
 
 				if (!(mdac_new = pool_alloc(fmt->cmd->mem,
-							    sizeof(*mdac_new))))
-				{
+							    sizeof(*mdac_new)))) {
 					stack;
 					return 0;
 				}
@@ -1486,8 +1481,7 @@ static struct format_instance *_create_text_instance(const struct format_type
 				}
 
 				if (!(mdac_new = pool_alloc(fmt->cmd->mem,
-							    sizeof(*mdac_new))))
-				{
+							    sizeof(*mdac_new)))) {
 					stack;
 					return NULL;
 				}
@@ -1649,7 +1643,7 @@ struct format_type *create_text_format(struct cmd_context *cmd)
 	fmt->ops = &_text_handler;
 	fmt->name = FMT_TEXT_NAME;
 	fmt->alias = FMT_TEXT_ALIAS;
-	fmt->features = FMT_SEGMENTS | FMT_MDAS | FMT_TAGS;
+	fmt->features = FMT_SEGMENTS | FMT_MDAS | FMT_TAGS | FMT_UNLIMITED_VOLS;
 
 	if (!(mda_lists = dbg_malloc(sizeof(struct mda_lists)))) {
 		log_error("Failed to allocate dir_list");
@@ -1688,14 +1682,15 @@ struct format_type *create_text_format(struct cmd_context *cmd)
 		}
 	}
 
-	if (!(cn = find_config_node(cmd->cft->root, "metadata/disk_areas")))
-		return fmt;
-
-	for (cn = cn->child; cn; cn = cn->sib) {
-		if (!_get_config_disk_area(cmd, cn, &mda_lists->raws))
-			goto err;
+	if ((cn = find_config_node(cmd->cft->root, "metadata/disk_areas"))) {
+		for (cn = cn->child; cn; cn = cn->sib) {
+			if (!_get_config_disk_area(cmd, cn, &mda_lists->raws))
+				goto err;
+		}
 	}
 
+	log_very_verbose("Initialised format: %s", fmt->name);
+
 	return fmt;
 
       err:

lib/format_text/import.c

@@ -49,12 +49,12 @@ struct volume_group *text_vg_import_fd(struct format_instance *fid,
 	*desc = NULL;
 	*when = 0;
 
-	if (!(cft = create_config_tree())) {
+	if (!(cft = create_config_tree(file))) {
 		stack;
 		goto out;
 	}
 
-	if ((!dev && !read_config_file(cft, file)) ||
+	if ((!dev && !read_config_file(cft)) ||
 	    (dev && !read_config_fd(cft, dev, offset, size,
 				    offset2, size2, checksum_fn, checksum))) {
 		log_error("Couldn't read volume group metadata.");

lib/format_text/import_vsn1.c

@@ -22,6 +22,8 @@
 #include "toolcontext.h"
 #include "lvmcache.h"
 #include "lv_alloc.h"
+#include "segtype.h"
+#include "text_import.h"
 
 typedef int (*section_fn) (struct format_instance * fid, struct pool * mem,
 			   struct volume_group * vg, struct config_node * pvn,
@@ -234,17 +236,13 @@ static int _read_segment(struct pool *mem, struct volume_group *vg,
 			 struct logical_volume *lv, struct config_node *sn,
 			 struct hash_table *pv_hash)
 {
-	unsigned int s;
-	uint32_t area_count = 0;
+	uint32_t area_count = 0u;
 	struct lv_segment *seg;
 	struct config_node *cn;
 	struct config_value *cv;
-	const char *seg_name = sn->key;
 	uint32_t start_extent, extent_count;
-	uint32_t chunk_size, extents_moved = 0u, seg_status = 0u;
-	const char *org_name, *cow_name;
-	struct logical_volume *org, *cow, *lv1;
-	segment_type_t segtype;
+	struct segment_type *segtype;
+	const char *segtype_str;
 
 	if (!(sn = sn->child)) {
 		log_error("Empty segment section.");
@@ -263,40 +261,26 @@ static int _read_segment(struct pool *mem, struct volume_group *vg,
 		return 0;
 	}
 
-	segtype = SEG_STRIPED;	/* Default */
+	segtype_str = "striped";
+
 	if ((cn = find_config_node(sn, "type"))) {
 		cv = cn->v;
 		if (!cv || !cv->v.str) {
 			log_error("Segment type must be a string.");
 			return 0;
 		}
-		segtype = get_segtype_from_string(cv->v.str);
+		segtype_str = cv->v.str;
 	}
 
-	if (segtype == SEG_STRIPED) {
-		if (!_read_int32(sn, "stripe_count", &area_count)) {
-			log_error("Couldn't read 'stripe_count' for "
-				  "segment '%s'.", sn->key);
-			return 0;
-		}
+	if (!(segtype = get_segtype_from_string(vg->cmd, segtype_str))) {
+		stack;
+		return 0;
 	}
 
-	if (segtype == SEG_MIRRORED) {
-		if (!_read_int32(sn, "mirror_count", &area_count)) {
-			log_error("Couldn't read 'mirror_count' for "
-				  "segment '%s'.", sn->key);
-			return 0;
-		}
-
-		if (find_config_node(sn, "extents_moved")) {
-			if (_read_uint32(sn, "extents_moved", &extents_moved))
-				seg_status |= PVMOVE;
-			else {
-				log_error("Couldn't read 'extents_moved' for "
-					  "segment '%s'.", sn->key);
-				return 0;
-			}
-		}
+	if (segtype->ops->text_import_area_count &&
+	    !segtype->ops->text_import_area_count(sn, &area_count)) {
+		stack;
+		return 0;
 	}
 
 	if (!(seg = alloc_lv_segment(mem, area_count))) {
@@ -308,9 +292,15 @@ static int _read_segment(struct pool *mem, struct volume_group *vg,
 	seg->le = start_extent;
 	seg->len = extent_count;
 	seg->area_len = extent_count;
-	seg->type = segtype;
-	seg->status = seg_status;
-	seg->extents_moved = extents_moved;
+	seg->status = 0u;
+	seg->segtype = segtype;
+	seg->extents_copied = 0u;
+
+	if (seg->segtype->ops->text_import &&
+	    !seg->segtype->ops->text_import(seg, sn, pv_hash)) {
+		stack;
+		return 0;
+	}
 
 	/* Optional tags */
 	if ((cn = find_config_node(sn, "tags")) &&
@@ -320,145 +310,89 @@ static int _read_segment(struct pool *mem, struct volume_group *vg,
 		return 0;
 	}
 
-	switch (segtype) {
-	case SEG_SNAPSHOT:
-		lv->status |= SNAPSHOT;
-
-		if (!_read_uint32(sn, "chunk_size", &chunk_size)) {
-			log_error("Couldn't read chunk size for snapshot.");
-			return 0;
-		}
-
-		log_suppress(1);
-
-		if (!(cow_name = find_config_str(sn, "cow_store", NULL))) {
-			log_suppress(0);
-			log_error("Snapshot cow storage not specified.");
-			return 0;
-		}
-
-		if (!(org_name = find_config_str(sn, "origin", NULL))) {
-			log_suppress(0);
-			log_error("Snapshot origin not specified.");
-			return 0;
-		}
-
-		log_suppress(0);
-
-		if (!(cow = find_lv(vg, cow_name))) {
-			log_error("Unknown logical volume specified for "
-				  "snapshot cow store.");
-			return 0;
-		}
-
-		if (!(org = find_lv(vg, org_name))) {
-			log_error("Unknown logical volume specified for "
-				  "snapshot origin.");
-			return 0;
-		}
-
-		if (!vg_add_snapshot(org, cow, 1, &lv->lvid.id[1], chunk_size)) {
-			stack;
-			return 0;
-		}
-		break;
-
-	case SEG_STRIPED:
-		if ((area_count != 1) &&
-		    !_read_int32(sn, "stripe_size", &seg->stripe_size)) {
-			log_error("Couldn't read stripe_size for segment '%s'.",
-				  sn->key);
-			return 0;
-		}
-
-		if (!(cn = find_config_node(sn, "stripes"))) {
-			log_error("Couldn't find stripes array for segment "
-				  "'%s'.", sn->key);
-			return 0;
-		}
-
-		seg->area_len /= area_count;
-
-	case SEG_MIRRORED:
-		seg->area_count = area_count;
-
-		if (!seg->area_count) {
-			log_error("Zero areas not allowed for segment '%s'",
-				  sn->key);
-			return 0;
-		}
-
-		if ((seg->type == SEG_MIRRORED) &&
-		    !(cn = find_config_node(sn, "mirrors"))) {
-			log_error("Couldn't find mirrors array for segment "
-				  "'%s'.", sn->key);
-			return 0;
-		}
-
-		for (cv = cn->v, s = 0; cv && s < seg->area_count;
-		     s++, cv = cv->next) {
-
-			/* first we read the pv */
-			const char *bad = "Badly formed areas array for "
-			    "segment '%s'.";
-			struct physical_volume *pv;
-
-			if (cv->type != CFG_STRING) {
-				log_error(bad, sn->key);
-				return 0;
-			}
-
-			if (!cv->next) {
-				log_error(bad, sn->key);
-				return 0;
-			}
-
-			if (cv->next->type != CFG_INT) {
-				log_error(bad, sn->key);
-				return 0;
-			}
-
-			/* FIXME Cope if LV not yet read in */
-			if ((pv = hash_lookup(pv_hash, cv->v.str))) {
-				seg->area[s].type = AREA_PV;
-				seg->area[s].u.pv.pv = pv;
-				seg->area[s].u.pv.pe = cv->next->v.i;
-				/*
-				 * Adjust extent counts in the pv and vg.
-				 */
-				pv->pe_alloc_count += seg->area_len;
-				vg->free_count -= seg->area_len;
-
-			} else if ((lv1 = find_lv(vg, cv->v.str))) {
-				seg->area[s].type = AREA_LV;
-				seg->area[s].u.lv.lv = lv1;
-				seg->area[s].u.lv.le = cv->next->v.i;
-			} else {
-				log_error("Couldn't find volume '%s' "
-					  "for segment '%s'.",
-					  cv->v.str ? cv->v.str : "NULL",
-					  seg_name);
-				return 0;
-			}
-
-			cv = cv->next;
-		}
-
-		/*
-		 * Check we read the correct number of stripes.
-		 */
-		if (cv || (s < seg->area_count)) {
-			log_error("Incorrect number of areas in area array "
-				  "for segment '%s'.", seg_name);
-			return 0;
-		}
-
-	}
-
 	/*
 	 * Insert into correct part of segment list.
 	 */
 	_insert_segment(lv, seg);
+
+	if (seg->segtype->flags & SEG_AREAS_MIRRORED)
+		lv->status |= MIRRORED;
+
+	if (seg->segtype->flags & SEG_VIRTUAL)
+		lv->status |= VIRTUAL;
+
+	return 1;
+}
+
+int text_import_areas(struct lv_segment *seg, const struct config_node *sn,
+		      const struct config_node *cn, struct hash_table *pv_hash)
+{
+	unsigned int s;
+	struct config_value *cv;
+	struct logical_volume *lv1;
+	const char *seg_name = sn->key;
+
+	if (!seg->area_count) {
+		log_error("Zero areas not allowed for segment '%s'", sn->key);
+		return 0;
+	}
+
+	for (cv = cn->v, s = 0; cv && s < seg->area_count; s++, cv = cv->next) {
+
+		/* first we read the pv */
+		const char *bad = "Badly formed areas array for "
+		    "segment '%s'.";
+		struct physical_volume *pv;
+
+		if (cv->type != CFG_STRING) {
+			log_error(bad, sn->key);
+			return 0;
+		}
+
+		if (!cv->next) {
+			log_error(bad, sn->key);
+			return 0;
+		}
+
+		if (cv->next->type != CFG_INT) {
+			log_error(bad, sn->key);
+			return 0;
+		}
+
+		/* FIXME Cope if LV not yet read in */
+		if ((pv = hash_lookup(pv_hash, cv->v.str))) {
+			seg->area[s].type = AREA_PV;
+			seg->area[s].u.pv.pv = pv;
+			seg->area[s].u.pv.pe = cv->next->v.i;
+			/*
+			 * Adjust extent counts in the pv and vg.
+			 */
+			pv->pe_alloc_count += seg->area_len;
+			seg->lv->vg->free_count -= seg->area_len;
+
+		} else if ((lv1 = find_lv(seg->lv->vg, cv->v.str))) {
+			seg->area[s].type = AREA_LV;
+			seg->area[s].u.lv.lv = lv1;
+			seg->area[s].u.lv.le = cv->next->v.i;
+		} else {
+			log_error("Couldn't find volume '%s' "
+				  "for segment '%s'.",
+				  cv->v.str ? cv->v.str : "NULL", seg_name);
+			return 0;
+		}
+
+		cv = cv->next;
+	}
+
+	/*
+	 * Check we read the correct number of stripes.
+	 */
+	if (cv || (s < seg->area_count)) {
+		log_error("Incorrect number of areas in area array "
+			  "for segment '%s'.", seg_name);
+		return 0;
+	}
+
 	return 1;
 }
 
@@ -555,7 +489,7 @@ static int _read_lvnames(struct format_instance *fid, struct pool *mem,
 		return 0;
 	}
 
-	lv->alloc = ALLOC_DEFAULT;
+	lv->alloc = ALLOC_INHERIT;
 	if ((cn = find_config_node(lvn, "allocation_policy"))) {
 		struct config_value *cv = cn->v;
 		if (!cv || !cv->v.str) {
@@ -564,6 +498,10 @@ static int _read_lvnames(struct format_instance *fid, struct pool *mem,
 		}
 
 		lv->alloc = get_alloc_from_string(cv->v.str);
+		if (lv->alloc == ALLOC_INVALID) {
+			stack;
+			return 0;
+		}
 	}
 
 	/* read_ahead defaults to 0 */
@@ -765,6 +703,21 @@ static struct volume_group *_read_vg(struct format_instance *fid,
 		goto bad;
 	}
 
+	vg->alloc = ALLOC_NORMAL;
+	if ((cn = find_config_node(vgn, "allocation_policy"))) {
+		struct config_value *cv = cn->v;
+		if (!cv || !cv->v.str) {
+			log_error("allocation_policy must be a string.");
+			return 0;
+		}
+
+		vg->alloc = get_alloc_from_string(cv->v.str);
+		if (vg->alloc == ALLOC_INVALID) {
+			stack;
+			return 0;
+		}
+	}
+
 	/*
 	 * The pv hash memoises the pv section names -> pv
 	 * structures.

lib/format_text/text_export.h

@@ -0,0 +1,36 @@
+/*
+ * Copyright (C) 2003-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU General Public License v.2.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#ifndef _LVM_TEXT_EXPORT_H
+#define _LVM_TEXT_EXPORT_H
+
+#define outf(args...) do {if (!out_text(args)) {stack; return 0;}} while (0)
+
+struct formatter;
+struct lv_segment;
+
+int out_size(struct formatter *f, uint64_t size, const char *fmt, ...)
+    __attribute__ ((format(printf, 3, 4)));
+
+int out_hint(struct formatter *f, const char *fmt, ...)
+    __attribute__ ((format(printf, 2, 3)));
+
+int out_text(struct formatter *f, const char *fmt, ...)
+    __attribute__ ((format(printf, 2, 3)));
+
+int out_areas(struct formatter *f, const struct lv_segment *seg,
+	      const char *type);
+
+#endif

lib/format_text/text_import.h

@@ -0,0 +1,25 @@
+/*
+ * Copyright (C) 2003-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU General Public License v.2.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#ifndef _LVM_TEXT_IMPORT_H
+#define _LVM_TEXT_IMPORT_H
+
+struct lv_segment;
+struct config_node;
+
+int text_import_areas(struct lv_segment *seg, const struct config_node *sn,
+		      const struct config_node *cn, struct hash_table *pv_hash);
+
+#endif

lib/format_text/text_label.c

@@ -18,6 +18,7 @@
 #include "layout.h"
 #include "label.h"
 #include "xlate.h"
+#include "lvmcache.h"
 
 #include <sys/stat.h>
 #include <fcntl.h>

lib/label/label.c

@@ -19,6 +19,7 @@
 #include "crc.h"
 #include "xlate.h"
 #include "lvmcache.h"
+#include "metadata.h"
 
 #include <sys/stat.h>
 #include <fcntl.h>
@@ -78,6 +79,8 @@ void label_exit(void)
 		li->l->ops->destroy(li->l);
 		_free_li(li);
 	}
+
+	list_init(&_labellers);
 }
 
 int label_register_handler(const char *name, struct labeller *handler)

lib/label/label.h

@@ -16,7 +16,6 @@
 #ifndef _LVM_LABEL_H
 #define _LVM_LABEL_H
 
-#include "lvmcache.h"
 #include "uuid.h"
 #include "device.h"
 
@@ -25,6 +24,8 @@
 #define LABEL_SCAN_SECTORS 4L
 #define LABEL_SCAN_SIZE (LABEL_SCAN_SECTORS << SECTOR_SHIFT)
 
+struct labeller;
+
 /* On disk - 32 bytes */
 struct label_header {
 	uint8_t id[8];		/* LABELONE */

lib/locking/.exported_symbols

@@ -0,0 +1,4 @@
+locking_init
+locking_end
+lock_resource
+reset_locking

lib/locking/Makefile.in

@@ -0,0 +1,30 @@
+#
+# Copyright (C) 2003-2004 Sistina Software, Inc. All rights reserved.
+# Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+#
+# This file is part of the LVM2.
+#
+# This copyrighted material is made available to anyone wishing to use,
+# modify, copy, or redistribute it subject to the terms and conditions
+# of the GNU General Public License v.2.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software Foundation,
+# Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+
+srcdir = @srcdir@
+top_srcdir = @top_srcdir@
+VPATH = @srcdir@
+
+SOURCES = cluster_locking.c
+
+LIB_SHARED = liblvm2clusterlock.so
+
+include $(top_srcdir)/make.tmpl
+
+install install_cluster: liblvm2clusterlock.so
+	$(INSTALL) -D $(OWNER) $(GROUP) -m 555 $(STRIP) $< \
+		$(libdir)/liblvm2clusterlock.so.$(LIB_VERSION)
+	$(LN_S) -f liblvm2clusterlock.so.$(LIB_VERSION) \
+		$(libdir)/liblvm2clusterlock.so
+

lib/locking/cluster_locking.c

@@ -0,0 +1,462 @@
+/*
+ * Copyright (C) 2002-2004 Sistina Software, Inc. All rights reserved.
+ * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU General Public License v.2.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+/*
+ * Locking functions for LVM.
+ * The main purpose of this part of the library is to serialise LVM
+ * management operations across a cluster.
+ */
+
+#include "lib.h"
+#include "clvm.h"
+#include "lvm-string.h"
+#include "locking.h"
+#include "locking_types.h"
+
+#include <stddef.h>
+#include <sys/socket.h>
+#include <sys/un.h>
+#include <unistd.h>
+
+#ifndef CLUSTER_LOCKING_INTERNAL
+int lock_resource(struct cmd_context *cmd, const char *resource, int flags);
+void locking_end(void);
+int locking_init(int type, struct config_tree *cf, uint32_t *flags);
+#endif
+
+typedef struct lvm_response {
+	char node[255];
+	char *response;
+	int status;
+	int len;
+} lvm_response_t;
+
+/*
+ * This gets stuck at the start of memory we allocate so we
+ * can sanity-check it at deallocation time
+ */
+#define LVM_SIGNATURE 0x434C564D
+
+/*
+ * NOTE: the LVMD uses the socket FD as the client ID, this means
+ * that any client that calls fork() will inherit the context of
+ * it's parent.
+ */
+static int _clvmd_sock = -1;
+
+/* FIXME Install SIGPIPE handler? */
+
+/* Open connection to the Cluster Manager daemon */
+static int _open_local_sock(void)
+{
+	int local_socket;
+	struct sockaddr_un sockaddr;
+
+	/* Open local socket */
+	if ((local_socket = socket(PF_UNIX, SOCK_STREAM, 0)) < 0) {
+		log_error("Local socket creation failed: %s", strerror(errno));
+		return -1;
+	}
+
+	memset(&sockaddr, 0, sizeof(sockaddr));
+	memcpy(sockaddr.sun_path, CLVMD_SOCKNAME, sizeof(CLVMD_SOCKNAME));
+
+	sockaddr.sun_family = AF_UNIX;
+
+	if (connect(local_socket,(struct sockaddr *) &sockaddr,
+		    sizeof(sockaddr))) {
+		int saved_errno = errno;
+
+		log_error("connect() failed on local socket: %s",
+			  strerror(errno));
+		if (close(local_socket))
+			stack;
+
+		errno = saved_errno;
+		return -1;
+	}
+
+	return local_socket;
+}
+
+/* Send a request and return the status */
+static int _send_request(char *inbuf, int inlen, char **retbuf)
+{
+	char outbuf[PIPE_BUF];
+	struct clvm_header *outheader = (struct clvm_header *) outbuf;
+	int len;
+	int off;
+	int buflen;
+	int err;
+
+	/* Send it to CLVMD */
+ rewrite:
+	if ( (err = write(_clvmd_sock, inbuf, inlen)) != inlen) {
+	        if (err == -1 && errno == EINTR)
+		        goto rewrite;
+		log_error("Error writing data to clvmd: %s", strerror(errno));
+		return 0;
+	}
+
+	/* Get the response */
+ reread:
+	if ((len = read(_clvmd_sock, outbuf, sizeof(struct clvm_header))) < 0) {
+	        if (errno == EINTR)
+		        goto reread;
+		log_error("Error reading data from clvmd: %s", strerror(errno));
+		return 0;
+	}
+
+	if (len == 0) {
+		log_error("EOF reading CLVMD");
+		errno = ENOTCONN;
+		return 0;
+	}
+
+	/* Allocate buffer */
+	buflen = len + outheader->arglen;
+	*retbuf = dbg_malloc(buflen);
+	if (!*retbuf) {
+		errno = ENOMEM;
+		return 0;
+	}
+
+	/* Copy the header */
+	memcpy(*retbuf, outbuf, len);
+	outheader = (struct clvm_header *) *retbuf;
+
+	/* Read the returned values */
+	off = 1;		/* we've already read the first byte */
+
+	while (off < outheader->arglen && len > 0) {
+		len = read(_clvmd_sock, outheader->args + off,
+			   buflen - off - offsetof(struct clvm_header, args));
+		if (len > 0)
+			off += len;
+	}
+
+	/* Was it an error ? */
+	if (outheader->status < 0) {
+		errno = -outheader->status;
+		log_error("cluster send request failed: %s", strerror(errno));
+		return 0;
+	}
+
+	return 1;
+}
+
+/* Build the structure header and parse-out wildcard node names */
+static void _build_header(struct clvm_header *head, int cmd, const char *node,
+			  int len)
+{
+	head->cmd = cmd;
+	head->status = 0;
+	head->flags = 0;
+	head->clientid = 0;
+	head->arglen = len;
+
+	if (node) {
+		/*
+		 * Allow a couple of special node names:
+		 * "*" for all nodes,
+		 * "." for the local node only
+		 */
+		if (strcmp(node, "*") == 0) {
+			head->node[0] = '\0';
+		} else if (strcmp(node, ".") == 0) {
+			head->node[0] = '\0';
+			head->flags = CLVMD_FLAG_LOCAL;
+		} else
+			strcpy(head->node, node);
+	} else
+		head->node[0] = '\0';
+}
+
+/*
+ * Send a message to a(or all) node(s) in the cluster and wait for replies
+ */
+static int _cluster_request(char cmd, const char *node, void *data, int len,
+			   lvm_response_t ** response, int *num)
+{
+	char outbuf[sizeof(struct clvm_header) + len + strlen(node) + 1];
+	int *outptr;
+	char *inptr;
+	char *retbuf = NULL;
+	int status;
+	int i;
+	int num_responses = 0;
+	struct clvm_header *head = (struct clvm_header *) outbuf;
+	lvm_response_t *rarray;
+
+	*num = 0;
+
+	if (_clvmd_sock == -1)
+		_clvmd_sock = _open_local_sock();
+
+	if (_clvmd_sock == -1)
+		return 0;
+
+	_build_header(head, cmd, node, len);
+	memcpy(head->node + strlen(head->node) + 1, data, len);
+
+	status = _send_request(outbuf, sizeof(struct clvm_header) +
+			      strlen(head->node) + len, &retbuf);
+	if (!status)
+		goto out;
+
+	/* Count the number of responses we got */
+	head = (struct clvm_header *) retbuf;
+	inptr = head->args;
+	while (inptr[0]) {
+		num_responses++;
+		inptr += strlen(inptr) + 1;
+		inptr += sizeof(int);
+		inptr += strlen(inptr) + 1;
+	}
+
+	/*
+	 * Allocate response array.
+	 * With an extra pair of INTs on the front to sanity
+	 * check the pointer when we are given it back to free
+	 */
+	outptr = dbg_malloc(sizeof(lvm_response_t) * num_responses +
+			    sizeof(int) * 2);
+	if (!outptr) {
+		errno = ENOMEM;
+		status = 0;
+		goto out;
+	}
+
+	*response = (lvm_response_t *) (outptr + 2);
+	outptr[0] = LVM_SIGNATURE;
+	outptr[1] = num_responses;
+	rarray = *response;
+
+	/* Unpack the response into an lvm_response_t array */
+	inptr = head->args;
+	i = 0;
+	while (inptr[0]) {
+		strcpy(rarray[i].node, inptr);
+		inptr += strlen(inptr) + 1;
+
+		rarray[i].status = *(int *) inptr;
+		inptr += sizeof(int);
+
+		rarray[i].response = dbg_malloc(strlen(inptr) + 1);
+		if (rarray[i].response == NULL) {
+			/* Free up everything else and return error */
+			int j;
+			for (j = 0; j < i; j++)
+				dbg_free(rarray[i].response);
+			free(outptr);
+			errno = ENOMEM;
+			status = -1;
+			goto out;
+		}
+
+		strcpy(rarray[i].response, inptr);
+		rarray[i].len = strlen(inptr);
+		inptr += strlen(inptr) + 1;
+		i++;
+	}
+	*num = num_responses;
+	*response = rarray;
+
+      out:
+	if (retbuf)
+		dbg_free(retbuf);
+
+	return status;
+}
+
+/* Free reply array */
+static int _cluster_free_request(lvm_response_t * response)
+{
+	int *ptr = (int *) response - 2;
+	int i;
+	int num;
+
+	/* Check it's ours to free */
+	if (response == NULL || *ptr != LVM_SIGNATURE) {
+		errno = EINVAL;
+		return 0;
+	}
+
+	num = ptr[1];
+
+	for (i = 0; i < num; i++) {
+		dbg_free(response[i].response);
+	}
+
+	dbg_free(ptr);
+
+	return 1;
+}
+
+static int _lock_for_cluster(unsigned char cmd, unsigned int flags, char *name)
+{
+	int status;
+	int i;
+	char *args;
+	const char *node = "";
+	int len;
+	int saved_errno = errno;
+	lvm_response_t *response = NULL;
+	int num_responses;
+
+	assert(name);
+
+	len = strlen(name) + 3;
+	args = alloca(len);
+	strcpy(args + 2, name);
+
+	args[0] = flags & 0xBF; /* Maskoff LOCAL flag */
+	args[1] = 0;		/* Not used now */
+
+	/*
+	 * VG locks are just that: locks, and have no side effects
+	 * so we only need to do them on the local node because all
+	 * locks are cluster-wide.
+	 * Also, if the lock is exclusive it makes no sense to try to 
+	 * acquire it on all nodes, so just do that on the local node too.
+	 */
+	if (cmd == CLVMD_CMD_LOCK_VG ||
+	    (flags & LCK_TYPE_MASK) == LCK_EXCL ||
+	    (flags & LCK_LOCAL))
+		node = ".";
+
+	status = _cluster_request(cmd, node, args, len,
+				  &response, &num_responses);
+
+	/* If any nodes were down then display them and return an error */
+	for (i = 0; i < num_responses; i++) {
+		if (response[i].status == -EHOSTDOWN) {
+			log_error("clvmd not running on node %s",
+				  response[i].node);
+			status = 0;
+		} else if (response[i].status) {
+			log_error("Error locking on node %s: %s",
+				  response[i].node,
+				  response[i].response[0] ?
+				  	response[i].response :
+				  	strerror(response[i].status));
+			status = 0;
+		}
+	}
+
+	saved_errno = errno;
+	_cluster_free_request(response);
+	errno = saved_errno;
+
+	return status;
+}
+
+/* API entry point for LVM */
+#ifdef CLUSTER_LOCKING_INTERNAL
+static int _lock_resource(struct cmd_context *cmd, const char *resource,
+			  int flags)
+#else
+int lock_resource(struct cmd_context *cmd, const char *resource, int flags)
+#endif
+{
+	char lockname[PATH_MAX];
+	int cluster_cmd = 0;
+
+	assert(strlen(resource) < sizeof(lockname));
+
+	switch (flags & LCK_SCOPE_MASK) {
+	case LCK_VG:
+		/* If the VG name is empty then lock the unused PVs */
+		if (!resource || !*resource)
+			lvm_snprintf(lockname, sizeof(lockname), "P_orphans");
+		else
+			lvm_snprintf(lockname, sizeof(lockname), "V_%s",
+				     resource);
+
+		cluster_cmd = CLVMD_CMD_LOCK_VG;
+		flags &= LCK_TYPE_MASK;
+		break;
+
+	case LCK_LV:
+		cluster_cmd = CLVMD_CMD_LOCK_LV;
+		strcpy(lockname, resource);
+		flags &= 0xffdf;	/* Mask off HOLD flag */
+		break;
+
+	default:
+		log_error("Unrecognised lock scope: %d",
+			  flags & LCK_SCOPE_MASK);
+		return 0;
+	}
+
+	/* Send a message to the cluster manager */
+	log_very_verbose("Locking %s at 0x%x", lockname, flags);
+
+	return _lock_for_cluster(cluster_cmd, flags, lockname);
+}
+
+#ifdef CLUSTER_LOCKING_INTERNAL
+static void _locking_end(void)
+#else
+void locking_end(void)
+#endif
+{
+	if (_clvmd_sock != -1 && close(_clvmd_sock))
+		stack;
+
+	_clvmd_sock = -1;
+}
+
+#ifdef CLUSTER_LOCKING_INTERNAL
+static void _reset_locking(void)
+#else
+void reset_locking(void)
+#endif
+{
+	if (close(_clvmd_sock))
+		stack;
+
+	_clvmd_sock = _open_local_sock();
+	if (_clvmd_sock == -1)
+	        stack;
+}
+
+#ifdef CLUSTER_LOCKING_INTERNAL
+int init_cluster_locking(struct locking_type *locking, struct config_tree *cft)
+{
+	locking->lock_resource = _lock_resource;
+	locking->fin_locking = _locking_end;
+	locking->reset_locking = _reset_locking;
+	locking->flags = LCK_PRE_MEMLOCK;
+
+	_clvmd_sock = _open_local_sock();
+	if (_clvmd_sock == -1)
+		return 0;
+
+	return 1;
+}
+#else
+int locking_init(int type, struct config_tree *cf, uint32_t *flags)
+{
+	_clvmd_sock = _open_local_sock();
+	if (_clvmd_sock == -1)
+		return 0;
+
+	/* Ask LVM to lock memory before calling us */
+	*flags |= LCK_PRE_MEMLOCK;
+
+	return 1;
+}
+#endif

lib/locking/external_locking.c

@@ -19,6 +19,7 @@
 #include "sharedlib.h"
 
 static void *_locking_lib = NULL;
+static void (*_reset_fn) (void) = NULL;
 static void (*_end_fn) (void) = NULL;
 static int (*_lock_fn) (struct cmd_context * cmd, const char *resource,
 			int flags) = NULL;
@@ -45,6 +46,13 @@ static void _fin_external_locking(void)
 	_init_fn = NULL;
 	_end_fn = NULL;
 	_lock_fn = NULL;
+	_reset_fn = NULL;
+}
+
+static void _reset_external_locking(void)
+{
+	if (_reset_fn)
+		_reset_fn();
 }
 
 int init_external_locking(struct locking_type *locking, struct config_tree *cft)
@@ -58,6 +66,7 @@ int init_external_locking(struct locking_type *locking, struct config_tree *cft)
 
 	locking->lock_resource = _lock_resource;
 	locking->fin_locking = _fin_external_locking;
+	locking->reset_locking = _reset_external_locking;
 	locking->flags = 0;
 
 	libname = find_config_str(cft->root, "global/locking_library",
@@ -71,6 +80,7 @@ int init_external_locking(struct locking_type *locking, struct config_tree *cft)
 	/* Get the functions we need */
 	if (!(_init_fn = dlsym(_locking_lib, "locking_init")) ||
 	    !(_lock_fn = dlsym(_locking_lib, "lock_resource")) ||
+	    !(_reset_fn = dlsym(_locking_lib, "reset_locking")) ||
 	    !(_end_fn = dlsym(_locking_lib, "locking_end"))) {
 		log_error("Shared library %s does not contain locking "
 			  "functions", libname);

lib/locking/file_locking.c

@@ -232,9 +232,14 @@ static int _file_lock_resource(struct cmd_context *cmd, const char *resource,
 			if (!lv_resume_if_active(cmd, resource))
 				return 0;
 			break;
+		case LCK_NULL:
+			log_debug("Locking LV %s (NL)", resource);
+			if (!lv_deactivate(cmd, resource))
+				return 0;
+			break;
 		case LCK_READ:
 			log_debug("Locking LV %s (R)", resource);
-			if (!lv_activate(cmd, resource))
+			if (!lv_activate_with_filter(cmd, resource))
 				return 0;
 			break;
 		case LCK_WRITE:
@@ -244,7 +249,7 @@ static int _file_lock_resource(struct cmd_context *cmd, const char *resource,
 			break;
 		case LCK_EXCL:
 			log_debug("Locking LV %s (EX)", resource);
-			if (!lv_deactivate(cmd, resource))
+			if (!lv_activate_with_filter(cmd, resource))
 				return 0;
 			break;
 		default:

lib/locking/locking.c

@@ -145,6 +145,14 @@ int init_locking(int type, struct config_tree *cft)
 		return 1;
 #endif
 
+#ifdef CLUSTER_LOCKING_INTERNAL
+	case 3:
+		if (!init_cluster_locking(&_locking, cft))
+			break;
+		log_very_verbose("Cluster locking enabled.");
+		return 1;
+#endif
+
 	default:
 		log_error("Unknown locking type requested.");
 		return 0;
@@ -252,32 +260,54 @@ int lock_vol(struct cmd_context *cmd, const char *vol, int flags)
 }
 
 /* Unlock list of LVs */
-int unlock_lvs(struct cmd_context *cmd, struct list *lvs)
+int resume_lvs(struct cmd_context *cmd, struct list *lvs)
 {
 	struct list *lvh;
 	struct logical_volume *lv;
 
 	list_iterate(lvh, lvs) {
 		lv = list_item(lvh, struct lv_list)->lv;
-		unlock_lv(cmd, lv->lvid.s);
+		resume_lv(cmd, lv->lvid.s);
 	}
 
 	return 1;
 }
 
 /* Lock a list of LVs */
-int lock_lvs(struct cmd_context *cmd, struct list *lvs, int flags)
+int suspend_lvs(struct cmd_context *cmd, struct list *lvs)
 {
 	struct list *lvh;
 	struct logical_volume *lv;
 
 	list_iterate(lvh, lvs) {
 		lv = list_item(lvh, struct lv_list)->lv;
-		if (!lock_vol(cmd, lv->lvid.s, flags)) {
+		if (!suspend_lv(cmd, lv->lvid.s)) {
 			log_error("Failed to suspend %s", lv->name);
 			list_uniterate(lvh, lvs, lvh) {
 				lv = list_item(lvh, struct lv_list)->lv;
-				unlock_lv(cmd, lv->lvid.s);
+				resume_lv(cmd, lv->lvid.s);
+			}
+
+			return 0;
+		}
+	}
+
+	return 1;
+}
+
+/* Lock a list of LVs */
+int activate_lvs_excl(struct cmd_context *cmd, struct list *lvs)
+{
+	struct list *lvh;
+	struct logical_volume *lv;
+
+	list_iterate(lvh, lvs) {
+		lv = list_item(lvh, struct lv_list)->lv;
+		if (!activate_lv_excl(cmd, lv->lvid.s)) {
+			log_error("Failed to activate %s", lv->name);
+			list_uniterate(lvh, lvs, lvh) {
+				lv = list_item(lvh, struct lv_list)->lv;
+				activate_lv(cmd, lv->lvid.s);
 			}
 
 			return 0;

lib/locking/locking.h

@@ -13,7 +13,6 @@
  * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
-#include "metadata.h"
 #include "uuid.h"
 #include "config.h"
 
@@ -64,6 +63,7 @@ int check_lvm1_vg_inactive(struct cmd_context *cmd, const char *vgname);
  */
 #define LCK_NONBLOCK	0x00000010	/* Don't block waiting for lock? */
 #define LCK_HOLD	0x00000020	/* Hold lock when lock_vol returns? */
+#define LCK_LOCAL	0x00000040	/* Don't propagate to other nodes */
 
 /*
  * Common combinations
@@ -72,16 +72,28 @@ int check_lvm1_vg_inactive(struct cmd_context *cmd, const char *vgname);
 #define LCK_VG_WRITE		(LCK_VG | LCK_WRITE | LCK_HOLD)
 #define LCK_VG_UNLOCK		(LCK_VG | LCK_UNLOCK)
 
-#define LCK_LV_SUSPEND		(LCK_LV | LCK_WRITE)
+#define LCK_LV_EXCLUSIVE	(LCK_LV | LCK_EXCL | LCK_NONBLOCK)
+#define LCK_LV_SUSPEND		(LCK_LV | LCK_WRITE | LCK_NONBLOCK)
 #define LCK_LV_RESUME		(LCK_LV | LCK_UNLOCK | LCK_NONBLOCK)
-#define LCK_LV_UNLOCK		(LCK_LV | LCK_UNLOCK)
-#define LCK_LV_ACTIVATE		(LCK_LV | LCK_READ)
-#define LCK_LV_DEACTIVATE	(LCK_LV | LCK_EXCL)
+#define LCK_LV_ACTIVATE		(LCK_LV | LCK_READ | LCK_NONBLOCK)
+#define LCK_LV_DEACTIVATE	(LCK_LV | LCK_NULL | LCK_NONBLOCK)
 
-#define unlock_lv(cmd, vol)	lock_vol(cmd, vol, LCK_LV_UNLOCK)
 #define unlock_vg(cmd, vol)	lock_vol(cmd, vol, LCK_VG_UNLOCK)
 
-/* Process list of LVs */
-int lock_lvs(struct cmd_context *cmd, struct list *lvs, int flags);
-int unlock_lvs(struct cmd_context *cmd, struct list *lvs);
+#define resume_lv(cmd, vol)	lock_vol(cmd, vol, LCK_LV_RESUME)
+#define suspend_lv(cmd, vol)	lock_vol(cmd, vol, LCK_LV_SUSPEND | LCK_HOLD)
+#define deactivate_lv(cmd, vol)	lock_vol(cmd, vol, LCK_LV_DEACTIVATE)
+#define activate_lv(cmd, vol)	lock_vol(cmd, vol, LCK_LV_ACTIVATE | LCK_HOLD)
+#define activate_lv_excl(cmd, vol)	\
+				lock_vol(cmd, vol, LCK_LV_EXCLUSIVE | LCK_HOLD)
+#define activate_lv_local(cmd, vol)	\
+	lock_vol(cmd, vol, LCK_LV_ACTIVATE | LCK_HOLD | LCK_LOCAL)
+#define deactivate_lv_local(cmd, vol)	\
+	lock_vol(cmd, vol, LCK_LV_DEACTIVATE | LCK_LOCAL)
+
+/* Process list of LVs */
+int suspend_lvs(struct cmd_context *cmd, struct list *lvs);
+int resume_lvs(struct cmd_context *cmd, struct list *lvs);
+int activate_lvs_excl(struct cmd_context *cmd, struct list *lvs);
+
 

lib/locking/locking_types.h

@@ -40,3 +40,4 @@ int init_no_locking(struct locking_type *locking, struct config_tree *cf);
 int init_file_locking(struct locking_type *locking, struct config_tree *cf);
 
 int init_external_locking(struct locking_type *locking, struct config_tree *cf);
+int init_cluster_locking(struct locking_type *locking, struct config_tree *cf);

lib/locking/no_locking.c

@@ -53,14 +53,16 @@ static int _no_lock_resource(struct cmd_context *cmd, const char *resource,
 		break;
 	case LCK_LV:
 		switch (flags & LCK_TYPE_MASK) {
+		case LCK_NULL:
+			return lv_deactivate(cmd, resource);
 		case LCK_UNLOCK:
 			return lv_resume_if_active(cmd, resource);
 		case LCK_READ:
-			return lv_activate(cmd, resource);
+			return lv_activate_with_filter(cmd, resource);
 		case LCK_WRITE:
 			return lv_suspend_if_active(cmd, resource);
 		case LCK_EXCL:
-			return lv_deactivate(cmd, resource);
+			return lv_activate_with_filter(cmd, resource);
 		default:
 			break;
 		}

lib/log/log.c

@@ -38,6 +38,7 @@ static int _indent = 1;
 static int _log_cmd_name = 0;
 static int _log_suppress = 0;
 static int _ignorelockingfailure = 0;
+static int _security_level = SECURITY_LEVEL;
 static char _cmd_name[30] = "";
 static char _msg_prefix[30] = "  ";
 static int _already_logging = 0;
@@ -147,6 +148,11 @@ void init_ignorelockingfailure(int level)
 	_ignorelockingfailure = level;
 }
 
+void init_security_level(int level)
+{
+	_security_level = level;
+}
+
 void init_cmd_name(int status)
 {
 	_log_cmd_name = status;
@@ -191,6 +197,11 @@ int ignorelockingfailure()
 	return _ignorelockingfailure;
 }
 
+int security_level()
+{
+	return _security_level;
+}
+
 void init_debug(int level)
 {
 	_debug_level = level;

lib/log/log.h

@@ -49,6 +49,7 @@
 #define _LOG_FATAL 2
 
 #define VERBOSE_BASE_LEVEL _LOG_WARN
+#define SECURITY_LEVEL 0
 
 void init_log_file(const char *log_file, int append);
 void init_log_direct(const char *log_file, int append);
@@ -68,6 +69,7 @@ void init_cmd_name(int status);
 void init_msg_prefix(const char *prefix);
 void init_indent(int indent);
 void init_ignorelockingfailure(int level);
+void init_security_level(int level);
 
 void set_cmd_name(const char *cmd_name);
 
@@ -76,6 +78,7 @@ int partial_mode(void);
 int pvmove_mode(void);
 int debug_level(void);
 int ignorelockingfailure(void);
+int security_level(void);
 
 /* Suppress messages to stdout/stderr */
 void log_suppress(int suppress);

lib/metadata/lv_manip.c

@@ -20,6 +20,8 @@
 #include "lvm-string.h"
 #include "toolcontext.h"
 #include "lv_alloc.h"
+#include "display.h"
+#include "segtype.h"
 
 /*
  * These functions adjust the pe counts in pv's
@@ -69,6 +71,7 @@ struct lv_segment *alloc_lv_segment(struct pool *mem, uint32_t num_areas)
 		return NULL;
 	}
 
+	seg->area_count = num_areas;
 	list_init(&seg->tags);
 
 	return seg;
@@ -76,6 +79,7 @@ struct lv_segment *alloc_lv_segment(struct pool *mem, uint32_t num_areas)
 
 static int _alloc_parallel_area(struct logical_volume *lv, uint32_t area_count,
 				uint32_t stripe_size,
+				struct segment_type *segtype,
 				struct pv_area **areas, uint32_t *ix)
 {
 	uint32_t count, area_len, smallest;
@@ -83,7 +87,9 @@ static int _alloc_parallel_area(struct logical_volume *lv, uint32_t area_count,
 	struct lv_segment *seg;
 	int striped = 0;
 
-	striped = 1;
+	/* Striped or mirrored? */
+	if (segtype->flags & SEG_AREAS_STRIPED)
+		striped = 1;
 
 	count = lv->le_count - *ix;
 	area_len = count / (striped ? area_count : 1);
@@ -98,13 +104,12 @@ static int _alloc_parallel_area(struct logical_volume *lv, uint32_t area_count,
 	}
 
 	seg->lv = lv;
-	seg->type = SEG_STRIPED;
+	seg->segtype = segtype;
 	seg->le = *ix;
 	seg->len = area_len * (striped ? area_count : 1);
 	seg->area_len = area_len;
-	seg->area_count = area_count;
 	seg->stripe_size = stripe_size;
-	seg->extents_moved = 0u;
+	seg->extents_copied = 0u;
 
 	for (s = 0; s < area_count; s++) {
 		struct pv_area *pva = areas[s];
@@ -116,6 +121,10 @@ static int _alloc_parallel_area(struct logical_volume *lv, uint32_t area_count,
 
 	list_add(&lv->segments, &seg->list);
 	*ix += seg->len;
+
+	if (!striped)
+		lv->status |= MIRRORED;
+
 	return 1;
 }
 
@@ -135,7 +144,8 @@ static int _comp_area(const void *l, const void *r)
 
 static int _alloc_parallel(struct logical_volume *lv,
 			   struct list *pvms, uint32_t allocated,
-			   uint32_t stripes, uint32_t stripe_size)
+			   uint32_t stripes, uint32_t stripe_size,
+			   uint32_t mirrors, struct segment_type *segtype)
 {
 	int r = 0;
 	struct list *pvmh;
@@ -145,7 +155,15 @@ static int _alloc_parallel(struct logical_volume *lv,
 	size_t len;
 	uint32_t area_count;
 
-	area_count = stripes;
+	if (stripes > 1 && mirrors > 1) {
+		log_error("striped mirrors are not supported yet");
+		return 0;
+	}
+
+	if (stripes > 1)
+		area_count = stripes;
+	else
+		area_count = mirrors;
 
 	list_iterate(pvmh, pvms)
 	    pv_count++;
@@ -179,8 +197,8 @@ static int _alloc_parallel(struct logical_volume *lv,
 		/* sort the areas so we allocate from the biggest */
 		qsort(areas, ix, sizeof(*areas), _comp_area);
 
-		if (!_alloc_parallel_area(lv, area_count, stripe_size, areas,
-					  &allocated)) {
+		if (!_alloc_parallel_area(lv, area_count, stripe_size, segtype,
+					  areas, &allocated)) {
 			stack;
 			goto out;
 		}
@@ -215,12 +233,14 @@ static int _alloc_linear_area(struct logical_volume *lv, uint32_t *ix,
 	}
 
 	seg->lv = lv;
-	seg->type = SEG_STRIPED;
+	if (!(seg->segtype = get_segtype_from_string(lv->vg->cmd, "striped"))) {
+		stack;
+		return 0;
+	}
 	seg->le = *ix;
 	seg->len = count;
 	seg->area_len = count;
 	seg->stripe_size = 0;
-	seg->area_count = 1;
 	seg->area[0].type = AREA_PV;
 	seg->area[0].u.pv.pv = map->pvl->pv;
 	seg->area[0].u.pv.pe = pva->start;
@@ -234,6 +254,7 @@ static int _alloc_linear_area(struct logical_volume *lv, uint32_t *ix,
 
 static int _alloc_mirrored_area(struct logical_volume *lv, uint32_t *ix,
 				struct pv_map *map, struct pv_area *pva,
+				struct segment_type *segtype,
 				struct physical_volume *mirrored_pv,
 				uint32_t mirrored_pe)
 {
@@ -251,14 +272,13 @@ static int _alloc_mirrored_area(struct logical_volume *lv, uint32_t *ix,
 	}
 
 	seg->lv = lv;
-	seg->type = SEG_MIRRORED;
-	seg->status = 0u;
+	seg->segtype = segtype;
 	seg->le = *ix;
+	seg->status = 0u;
 	seg->len = count;
 	seg->area_len = count;
 	seg->stripe_size = 0;
-	seg->area_count = 2;
-	seg->extents_moved = 0u;
+	seg->extents_copied = 0u;
 	/* FIXME Remove AREA_PV restriction here? */
 	seg->area[0].type = AREA_PV;
 	seg->area[0].u.pv.pv = mirrored_pv;
@@ -321,6 +341,7 @@ static int _alloc_contiguous(struct logical_volume *lv,
 /* FIXME Contiguous depends on *segment* (i.e. stripe) not LV */
 static int _alloc_mirrored(struct logical_volume *lv,
 			   struct list *pvms, uint32_t allocated,
+			   struct segment_type *segtype,
 			   struct physical_volume *mirrored_pv,
 			   uint32_t mirrored_pe)
 {
@@ -343,7 +364,7 @@ static int _alloc_mirrored(struct logical_volume *lv,
 			continue;
 		}
 
-		if (!_alloc_mirrored_area(lv, &allocated, pvm, pva,
+		if (!_alloc_mirrored_area(lv, &allocated, pvm, pva, segtype,
 					  mirrored_pv, mirrored_pe)) {
 			stack;
 			return 0;
@@ -395,12 +416,37 @@ static int _alloc_next_free(struct logical_volume *lv,
 	return 1;
 }
 
+static int _alloc_virtual(struct logical_volume *lv,
+			  uint32_t allocated, struct segment_type *segtype)
+{
+	struct lv_segment *seg;
+
+	if (!(seg = alloc_lv_segment(lv->vg->cmd->mem, 0))) {
+		log_err("Couldn't allocate new zero segment.");
+		return 0;
+	}
+
+	seg->lv = lv;
+	seg->segtype = segtype;
+	seg->status = 0u;
+	seg->le = allocated;
+	seg->len = lv->le_count - allocated;
+	seg->area_len = seg->len;
+	seg->stripe_size = 0;
+	seg->extents_copied = 0u;
+	list_add(&lv->segments, &seg->list);
+	lv->status |= VIRTUAL;
+
+	return 1;
+}
+
 /*
  * Chooses a correct allocation policy.
  */
 static int _allocate(struct volume_group *vg, struct logical_volume *lv,
 		     struct list *allocatable_pvs, uint32_t allocated,
-		     uint32_t stripes, uint32_t stripe_size,
+		     alloc_policy_t alloc, struct segment_type *segtype,
+		     uint32_t stripes, uint32_t stripe_size, uint32_t mirrors,
 		     struct physical_volume *mirrored_pv, uint32_t mirrored_pe,
 		     uint32_t status)
 {
@@ -409,32 +455,40 @@ static int _allocate(struct volume_group *vg, struct logical_volume *lv,
 	struct list *pvms, *old_tail = lv->segments.p, *segh;
 	struct lv_segment *seg;
 
+	if (segtype->flags & SEG_VIRTUAL)
+		return _alloc_virtual(lv, allocated, segtype);
+
 	if (!(scratch = pool_create(1024))) {
 		stack;
 		return 0;
 	}
 
+	if (alloc == ALLOC_INHERIT)
+		alloc = vg->alloc;
+
 	/*
 	 * Build the sets of available areas on the pv's.
 	 */
 	if (!(pvms = create_pv_maps(scratch, vg, allocatable_pvs)))
 		goto out;
 
-	if (stripes > 1)
-		r = _alloc_parallel(lv, pvms, allocated, stripes, stripe_size);
+	if (stripes > 1 || mirrors > 1)
+		r = _alloc_parallel(lv, pvms, allocated, stripes, stripe_size,
+				    mirrors, segtype);
 
 	else if (mirrored_pv)
-		r = _alloc_mirrored(lv, pvms, allocated, mirrored_pv,
+		r = _alloc_mirrored(lv, pvms, allocated, segtype, mirrored_pv,
 				    mirrored_pe);
-	else if (lv->alloc == ALLOC_CONTIGUOUS)
+
+	else if (alloc == ALLOC_CONTIGUOUS)
 		r = _alloc_contiguous(lv, pvms, allocated);
 
-	else if (lv->alloc == ALLOC_NEXT_FREE || lv->alloc == ALLOC_DEFAULT)
+	else if (alloc == ALLOC_NORMAL || alloc == ALLOC_ANYWHERE)
 		r = _alloc_next_free(lv, pvms, allocated);
 
 	else {
-		log_error("Unknown allocation policy: "
-			  "unable to setup logical volume.");
+		log_error("Unrecognised allocation policy: "
+			  "unable to set up logical volume.");
 		goto out;
 	}
 
@@ -554,57 +608,52 @@ struct logical_volume *lv_create_empty(struct format_instance *fi,
 	return lv;
 }
 
-struct logical_volume *lv_create(struct format_instance *fi,
-				 const char *name,
-				 uint32_t status,
-				 alloc_policy_t alloc,
-				 uint32_t stripes,
-				 uint32_t stripe_size,
-				 uint32_t extents,
-				 struct volume_group *vg,
-				 struct list *allocatable_pvs)
+int lv_extend(struct format_instance *fid,
+	      struct logical_volume *lv,
+	      struct segment_type *segtype,
+	      uint32_t stripes, uint32_t stripe_size,
+	      uint32_t mirrors, uint32_t extents,
+	      struct physical_volume *mirrored_pv, uint32_t mirrored_pe,
+	      uint32_t status, struct list *allocatable_pvs,
+	      alloc_policy_t alloc)
 {
-	struct logical_volume *lv;
+	uint32_t old_le_count = lv->le_count;
+	uint64_t old_size = lv->size;
 
-	if (!extents) {
-		log_error("Unable to create logical volume %s with no extents",
-			  name);
-		return NULL;
+	lv->le_count += extents;
+	lv->size += (uint64_t) extents *lv->vg->extent_size;
+
+	if (fid->fmt->ops->segtype_supported &&
+	    !fid->fmt->ops->segtype_supported(fid, segtype)) {
+		log_error("Metadata format (%s) does not support required "
+			  "LV segment type (%s).", fid->fmt->name,
+			  segtype->name);
+		log_error("Consider changing the metadata format by running "
+			  "vgconvert.");
+		return 0;
 	}
 
-	if (vg->free_count < extents) {
-		log_error("Insufficient free extents (%u) in volume group %s: "
-			  "%u required", vg->free_count, vg->name, extents);
-		return NULL;
-	}
-
-	if (stripes > list_size(allocatable_pvs)) {
-		log_error("Number of stripes (%u) must not exceed "
-			  "number of physical volumes (%d)", stripes,
-			  list_size(allocatable_pvs));
-		return NULL;
-	}
-
-	if (!(lv = lv_create_empty(fi, name, "lvol%d", status, alloc, vg))) {
+	if (!_allocate(lv->vg, lv, allocatable_pvs, old_le_count, alloc,
+		       segtype, stripes, stripe_size, mirrors, mirrored_pv,
+		       mirrored_pe, status)) {
+		lv->le_count = old_le_count;
+		lv->size = old_size;
 		stack;
-		return NULL;
+		return 0;
 	}
 
-	lv->size = (uint64_t) extents *vg->extent_size;
-	lv->le_count = extents;
+	if ((segtype->flags & SEG_CAN_SPLIT) && !lv_merge_segments(lv)) {
+		log_err("Couldn't merge segments after extending "
+			"logical volume.");
+		return 0;
+	}
 
-	if (!_allocate(vg, lv, allocatable_pvs, 0u, stripes, stripe_size,
-		       NULL, 0u, 0u)) {
+	if (fid->fmt->ops->lv_setup && !fid->fmt->ops->lv_setup(fid, lv)) {
 		stack;
-		return NULL;
+		return 0;
 	}
 
-	if (fi->fmt->ops->lv_setup && !fi->fmt->ops->lv_setup(fi, lv)) {
-		stack;
-		return NULL;
-	}
-
-	return lv;
+	return 1;
 }
 
 int lv_reduce(struct format_instance *fi,
@@ -628,7 +677,7 @@ int lv_reduce(struct format_instance *fi,
 			/* reduce this segment */
 			_put_extents(seg);
 			seg->len -= count;
-			striped = (seg->type == SEG_STRIPED);
+			striped = seg->segtype->flags & SEG_AREAS_STRIPED;
 			/* Caller must ensure exact divisibility */
 			if (striped && (count % seg->area_count)) {
 				log_error("Segment extent reduction %" PRIu32
@@ -655,68 +704,6 @@ int lv_reduce(struct format_instance *fi,
 	return 1;
 }
 
-int lv_extend(struct format_instance *fi,
-	      struct logical_volume *lv,
-	      uint32_t stripes, uint32_t stripe_size,
-	      uint32_t extents, struct list *allocatable_pvs)
-{
-	uint32_t old_le_count = lv->le_count;
-	uint64_t old_size = lv->size;
-
-	lv->le_count += extents;
-	lv->size += (uint64_t) extents *lv->vg->extent_size;
-
-	if (!_allocate(lv->vg, lv, allocatable_pvs, old_le_count,
-		       stripes, stripe_size, NULL, 0u, 0u)) {
-		lv->le_count = old_le_count;
-		lv->size = old_size;
-		stack;
-		return 0;
-	}
-
-	if (!lv_merge_segments(lv)) {
-		log_err("Couldn't merge segments after extending "
-			"logical volume.");
-		return 0;
-	}
-
-	if (fi->fmt->ops->lv_setup && !fi->fmt->ops->lv_setup(fi, lv)) {
-		stack;
-		return 0;
-	}
-
-	return 1;
-}
-
-int lv_extend_mirror(struct format_instance *fid,
-		     struct logical_volume *lv,
-		     struct physical_volume *mirrored_pv,
-		     uint32_t mirrored_pe,
-		     uint32_t extents, struct list *allocatable_pvs,
-		     uint32_t status)
-{
-	uint32_t old_le_count = lv->le_count;
-	uint64_t old_size = lv->size;
-
-	lv->le_count += extents;
-	lv->size += (uint64_t) extents *lv->vg->extent_size;
-
-	if (!_allocate(lv->vg, lv, allocatable_pvs, old_le_count,
-		       1, extents, mirrored_pv, mirrored_pe, status)) {
-		lv->le_count = old_le_count;
-		lv->size = old_size;
-		stack;
-		return 0;
-	}
-
-	if (fid->fmt->ops->lv_setup && !fid->fmt->ops->lv_setup(fid, lv)) {
-		stack;
-		return 0;
-	}
-
-	return 1;
-}
-
 int lv_remove(struct volume_group *vg, struct logical_volume *lv)
 {
 	struct list *segh;

lib/metadata/merge.c

@@ -18,60 +18,20 @@
 #include "toolcontext.h"
 #include "lv_alloc.h"
 #include "str_list.h"
-
-/*
- * Test whether two segments could be merged by the current merging code
- */
-static int _segments_compatible(struct lv_segment *first,
-				struct lv_segment *second)
-{
-	uint32_t width;
-	unsigned s;
-
-	/* FIXME Relax the seg type restriction */
-	if (!first || !second ||
-	    (first->type != SEG_STRIPED) || (second->type != first->type) ||
-	    (first->area_count != second->area_count) ||
-	    (first->stripe_size != second->stripe_size))
-		return 0;
-
-	for (s = 0; s < first->area_count; s++) {
-
-		/* FIXME Relax this to first area type != second area type */
-		/*       plus the additional AREA_LV checks needed */
-		if ((first->area[s].type != AREA_PV) ||
-		    (second->area[s].type != AREA_PV))
-			return 0;
-
-		width = first->area_len;
-
-		if ((first->area[s].u.pv.pv != second->area[s].u.pv.pv) ||
-		    (first->area[s].u.pv.pe + width != second->area[s].u.pv.pe))
-			return 0;
-	}
-
-	if (!str_list_lists_equal(&first->tags, &second->tags))
-		return 0;
-
-	return 1;
-}
+#include "segtype.h"
 
 /*
  * Attempt to merge two adjacent segments.
- * Currently only supports SEG_STRIPED on AREA_PV.
+ * Currently only supports striped segments on AREA_PV.
  * Returns success if successful, in which case 'first' 
  * gets adjusted to contain both areas.
  */
 static int _merge(struct lv_segment *first, struct lv_segment *second)
 {
+	if (!first || !second || first->segtype != second->segtype ||
+	    !first->segtype->ops->merge_segments) return 0;
 
-	if (!_segments_compatible(first, second))
-		return 0;
-
-	first->len += second->len;
-	first->area_len += second->area_len;
-
-	return 1;
+	return first->segtype->ops->merge_segments(first, second);
 }
 
 int lv_merge_segments(struct logical_volume *lv)
@@ -79,6 +39,9 @@ int lv_merge_segments(struct logical_volume *lv)
 	struct list *segh, *t;
 	struct lv_segment *current, *prev = NULL;
 
+	if (lv->status & LOCKED || lv->status & PVMOVE)
+		return 1;
+
 	list_iterate_safe(segh, t, &lv->segments) {
 		current = list_item(segh, struct lv_segment);
 
@@ -117,6 +80,7 @@ int lv_check_segments(struct logical_volume *lv)
 
 /*
  * Split the supplied segment at the supplied logical extent
+ * NB Use LE numbering that works across stripes PV1: 0,2,4 PV2: 1,3,5 etc.
  */
 static int _lv_split_segment(struct logical_volume *lv, struct lv_segment *seg,
 			     uint32_t le)
@@ -125,16 +89,16 @@ static int _lv_split_segment(struct logical_volume *lv, struct lv_segment *seg,
 	struct lv_segment *split_seg;
 	uint32_t s;
 	uint32_t offset = le - seg->le;
+	uint32_t area_offset;
 
-	if (seg->type == SEG_SNAPSHOT) {
-		log_error("Unable to split the snapshot segment at LE %" PRIu32
-			  " in LV %s", le, lv->name);
+	if (!(seg->segtype->flags & SEG_CAN_SPLIT)) {
+		log_error("Unable to split the %s segment at LE %" PRIu32
+			  " in LV %s", seg->segtype->name, le, lv->name);
 		return 0;
 	}
 
 	/* Clone the existing segment */
-	if (!(split_seg = alloc_lv_segment(lv->vg->cmd->mem,
-					   seg->area_count))) {
+	if (!(split_seg = alloc_lv_segment(lv->vg->cmd->mem, seg->area_count))) {
 		log_error("Couldn't allocate new LV segment.");
 		return 0;
 	}
@@ -148,8 +112,9 @@ static int _lv_split_segment(struct logical_volume *lv, struct lv_segment *seg,
 	}
 
 	/* In case of a striped segment, the offset has to be / stripes */
-	if (seg->type == SEG_STRIPED)
-		offset /= seg->area_count;
+	area_offset = offset;
+	if (seg->segtype->flags & SEG_AREAS_STRIPED)
+		area_offset /= seg->area_count;
 
 	/* Adjust the PV mapping */
 	for (s = 0; s < seg->area_count; s++) {
@@ -157,12 +122,19 @@ static int _lv_split_segment(struct logical_volume *lv, struct lv_segment *seg,
 		switch (seg->area[s].type) {
 		case AREA_LV:
 			split_seg->area[s].u.lv.le =
-			    seg->area[s].u.lv.le + offset;
+			    seg->area[s].u.lv.le + area_offset;
+			log_debug("Split %s:%u[%u] at %u: %s LE %u", lv->name,
+				  seg->le, s, le, seg->area[s].u.lv.lv->name,
+				  split_seg->area[s].u.lv.le);
 			break;
 
 		case AREA_PV:
 			split_seg->area[s].u.pv.pe =
-			    seg->area[s].u.pv.pe + offset;
+			    seg->area[s].u.pv.pe + area_offset;
+			log_debug("Split %s:%u[%u] at %u: %s PE %u", lv->name,
+				  seg->le, s, le,
+				  dev_name(seg->area[s].u.pv.pv->dev),
+				  split_seg->area[s].u.pv.pe);
 			break;
 
 		default:
@@ -172,8 +144,13 @@ static int _lv_split_segment(struct logical_volume *lv, struct lv_segment *seg,
 		}
 	}
 
-	split_seg->area_len = seg->area_len - offset;
-	seg->area_len = offset;
+	split_seg->area_len -= area_offset;
+	seg->area_len = area_offset;
+
+	split_seg->len -= offset;
+	seg->len = offset;
+
+	split_seg->le = seg->le + seg->len;
 
 	/* Add split off segment to the list _after_ the original one */
 	list_add_h(&seg->list, &split_seg->list);
@@ -205,4 +182,3 @@ int lv_split_segment(struct logical_volume *lv, uint32_t le)
 
 	return 1;
 }
-

lib/metadata/metadata.c

@@ -39,7 +39,7 @@ static int _add_pv_to_vg(struct format_instance *fid, struct volume_group *vg,
 	}
 
 	list_init(&mdas);
-	if (!(pv = pv_read(fid->fmt->cmd, pv_name, &mdas, NULL))) {
+	if (!(pv = pv_read(fid->fmt->cmd, pv_name, &mdas, NULL, 1))) {
 		log_error("%s not identified as an existing physical volume",
 			  pv_name);
 		return 0;
@@ -162,7 +162,8 @@ const char *strip_dir(const char *vg_name, const char *dev_dir)
 
 struct volume_group *vg_create(struct cmd_context *cmd, const char *vg_name,
 			       uint32_t extent_size, uint32_t max_pv,
-			       uint32_t max_lv, int pv_count, char **pv_names)
+			       uint32_t max_lv, alloc_policy_t alloc,
+			       int pv_count, char **pv_names)
 {
 	struct volume_group *vg;
 	struct pool *mem = cmd->mem;
@@ -211,6 +212,8 @@ struct volume_group *vg_create(struct cmd_context *cmd, const char *vg_name,
 	vg->max_lv = max_lv;
 	vg->max_pv = max_pv;
 
+	vg->alloc = alloc;
+
 	vg->pv_count = 0;
 	list_init(&vg->pvs);
 
@@ -418,6 +421,24 @@ struct physical_volume *find_pv(struct volume_group *vg, struct device *dev)
 	return NULL;
 }
 
+struct physical_volume *find_pv_by_name(struct cmd_context *cmd,
+					const char *pv_name)
+{
+	struct physical_volume *pv;
+
+	if (!(pv = pv_read(cmd, pv_name, NULL, NULL, 1))) {
+		log_error("Physical volume %s not found", pv_name);
+		return NULL;
+	}
+
+	if (!pv->vg_name[0]) {
+		log_error("Physical volume %s not in a volume group", pv_name);
+		return NULL;
+	}
+
+	return pv;
+}
+
 /* Find segment at a given logical extent in an LV */
 struct lv_segment *find_seg_by_le(struct logical_volume *lv, uint32_t le)
 {
@@ -584,7 +605,7 @@ static struct volume_group *_vg_read_orphans(struct cmd_context *cmd)
 
 	list_iterate(ih, &vginfo->infos) {
 		dev = list_item(ih, struct lvmcache_info)->dev;
-		if (!(pv = pv_read(cmd, dev_name(dev), NULL, NULL))) {
+		if (!(pv = pv_read(cmd, dev_name(dev), NULL, NULL, 1))) {
 			continue;
 		}
 		if (!(pvl = pool_zalloc(cmd->mem, sizeof(*pvl)))) {
@@ -793,7 +814,8 @@ struct logical_volume *lv_from_lvid(struct cmd_context *cmd, const char *lvid_s)
 
 /* FIXME Use label functions instead of PV functions */
 struct physical_volume *pv_read(struct cmd_context *cmd, const char *pv_name,
-				struct list *mdas, uint64_t *label_sector)
+				struct list *mdas, uint64_t *label_sector,
+				int warnings)
 {
 	struct physical_volume *pv;
 	struct label *label;
@@ -806,7 +828,9 @@ struct physical_volume *pv_read(struct cmd_context *cmd, const char *pv_name,
 	}
 
 	if (!(label_read(dev, &label))) {
-		log_error("No physical volume label read from %s", pv_name);
+		if (warnings)
+			log_error("No physical volume label read from %s",
+				  pv_name);
 		return 0;
 	}
 

lib/metadata/metadata.h

@@ -53,6 +53,8 @@
 #define SNAPSHOT		0x00001000	/* LV - tmp internal use only */
 #define PVMOVE			0x00002000	/* VG LV SEG */
 #define LOCKED			0x00004000	/* LV */
+#define MIRRORED		0x00008000	/* LV - internal use only */
+#define VIRTUAL			0x00010000	/* LV - internal use only */
 
 #define LVM_READ              	0x00000100	/* LV VG */
 #define LVM_WRITE             	0x00000200	/* LV VG */
@@ -66,18 +68,14 @@
 #define FMT_UNLIMITED_VOLS	0x00000008	/* Unlimited PVs/LVs? */
 #define FMT_RESTRICTED_LVIDS	0x00000010	/* LVID <= 255 */
 #define FMT_ORPHAN_ALLOCATABLE	0x00000020	/* Orphan PV allocatable? */
-  
-typedef enum {
-	ALLOC_DEFAULT,
-	ALLOC_NEXT_FREE,
-	ALLOC_CONTIGUOUS
-} alloc_policy_t;
 
 typedef enum {
-	SEG_STRIPED,
-	SEG_SNAPSHOT,
-	SEG_MIRRORED
-} segment_type_t;
+	ALLOC_INVALID,
+	ALLOC_INHERIT,
+	ALLOC_CONTIGUOUS,
+	ALLOC_NORMAL,
+	ALLOC_ANYWHERE
+} alloc_policy_t;
 
 typedef enum {
 	AREA_PV,
@@ -171,6 +169,7 @@ struct volume_group {
 	char *system_id;
 
 	uint32_t status;
+	alloc_policy_t alloc;
 
 	uint32_t extent_size;
 	uint32_t extent_count;
@@ -194,11 +193,12 @@ struct volume_group {
 	struct list tags;
 };
 
+struct segment_type;
 struct lv_segment {
 	struct list list;
 	struct logical_volume *lv;
 
-	segment_type_t type;
+	struct segment_type *segtype;
 	uint32_t le;
 	uint32_t len;
 
@@ -211,7 +211,7 @@ struct lv_segment {
 	struct logical_volume *origin;
 	struct logical_volume *cow;
 	uint32_t chunk_size;
-	uint32_t extents_moved;
+	uint32_t extents_copied;
 
 	struct list tags;
 
@@ -265,7 +265,7 @@ struct name_list {
 	char *name;
 };
 
-struct alloc_area {
+struct pe_range {
 	struct list list;
 	uint32_t start;		/* PEs */
 	uint32_t count;		/* PEs */
@@ -275,7 +275,7 @@ struct pv_list {
 	struct list list;
 	struct physical_volume *pv;
 	struct list *mdas;	/* Metadata areas */
-	struct list *alloc_areas;	/* Areas we may allocate from */
+	struct list *pe_ranges;	/* Ranges of PEs e.g. for allocation */
 };
 
 struct lv_list {
@@ -341,6 +341,12 @@ struct format_handler {
 	 */
 	int (*vg_setup) (struct format_instance * fi, struct volume_group * vg);
 
+	/*
+	 * Check whether particular segment type is supported.
+	 */
+	int (*segtype_supported) (struct format_instance *fid,
+				  struct segment_type *segtype);
+
 	/*
 	 * Create format instance with a particular metadata area
 	 */
@@ -369,7 +375,8 @@ struct volume_group *vg_read(struct cmd_context *cmd, const char *vg_name,
 			     int *consistent);
 struct volume_group *vg_read_by_vgid(struct cmd_context *cmd, const char *vgid);
 struct physical_volume *pv_read(struct cmd_context *cmd, const char *pv_name,
-				struct list *mdas, uint64_t *label_sector);
+				struct list *mdas, uint64_t *label_sector,
+				int warnings);
 struct list *get_pvs(struct cmd_context *cmd);
 
 /* Set full_scan to 1 to re-read every (filtered) device label */
@@ -392,27 +399,15 @@ struct physical_volume *pv_create(const struct format_type *fmt,
 
 struct volume_group *vg_create(struct cmd_context *cmd, const char *name,
 			       uint32_t extent_size, uint32_t max_pv,
-			       uint32_t max_lv, int pv_count, char **pv_names);
+			       uint32_t max_lv, alloc_policy_t alloc,
+			       int pv_count, char **pv_names);
 int vg_remove(struct volume_group *vg);
 int vg_rename(struct cmd_context *cmd, struct volume_group *vg,
 	      const char *new_name);
 int vg_extend(struct format_instance *fi, struct volume_group *vg,
 	      int pv_count, char **pv_names);
 
-/*
- * Create a new LV within a given volume group.
- *
- */
-struct logical_volume *lv_create(struct format_instance *fi,
-				 const char *name,
-				 uint32_t status,
-				 alloc_policy_t alloc,
-				 uint32_t stripes,
-				 uint32_t stripe_size,
-				 uint32_t extents,
-				 struct volume_group *vg,
-				 struct list *allocatable_pvs);
-
+/* Manipulate LVs */
 struct logical_volume *lv_create_empty(struct format_instance *fi,
 				       const char *name,
 				       const char *name_format,
@@ -420,22 +415,17 @@ struct logical_volume *lv_create_empty(struct format_instance *fi,
 				       alloc_policy_t alloc,
 				       struct volume_group *vg);
 
-/* Manipulate LVs */
 int lv_reduce(struct format_instance *fi,
 	      struct logical_volume *lv, uint32_t extents);
 
-int lv_extend(struct format_instance *fi,
+int lv_extend(struct format_instance *fid,
 	      struct logical_volume *lv,
-	      uint32_t stripes,
-	      uint32_t stripe_size,
-	      uint32_t extents, struct list *allocatable_pvs);
-
-int lv_extend_mirror(struct format_instance *fid,
-		     struct logical_volume *lv,
-		     struct physical_volume *mirrored_pv,
-		     uint32_t mirrored_pe,
-		     uint32_t extents, struct list *allocatable_pvs,
-		     uint32_t status);
+	      struct segment_type *segtype,
+	      uint32_t stripes, uint32_t stripe_size,
+	      uint32_t mirrors, uint32_t extents,
+	      struct physical_volume *mirrored_pv, uint32_t mirrored_pe,
+	      uint32_t status, struct list *allocatable_pvs,
+	      alloc_policy_t alloc);
 
 /* lv must be part of vg->lvs */
 int lv_remove(struct volume_group *vg, struct logical_volume *lv);
@@ -466,6 +456,8 @@ struct logical_volume *lv_from_lvid(struct cmd_context *cmd,
 /* FIXME Merge these functions with ones above */
 struct physical_volume *find_pv(struct volume_group *vg, struct device *dev);
 struct logical_volume *find_lv(struct volume_group *vg, const char *lv_name);
+struct physical_volume *find_pv_by_name(struct cmd_context *cmd,
+					const char *pv_name);
 
 /* Find LV segment containing given LE */
 struct lv_segment *find_seg_by_le(struct logical_volume *lv, uint32_t le);
@@ -515,18 +507,21 @@ int vg_remove_snapshot(struct volume_group *vg, struct logical_volume *cow);
  */
 int insert_pvmove_mirrors(struct cmd_context *cmd,
 			  struct logical_volume *lv_mirr,
-			  struct physical_volume *pv,
+			  struct list *source_pvl,
 			  struct logical_volume *lv,
 			  struct list *allocatable_pvs,
 			  struct list *lvs_changed);
 int remove_pvmove_mirrors(struct volume_group *vg,
 			  struct logical_volume *lv_mirr);
 struct logical_volume *find_pvmove_lv(struct volume_group *vg,
-				      struct device *dev);
-struct physical_volume *get_pvmove_pv_from_lv(struct logical_volume *lv);
-struct physical_volume *get_pvmove_pv_from_lv_mirr(struct logical_volume
-						   *lv_mirr);
-float pvmove_percent(struct logical_volume *lv_mirr);
+				      struct device *dev, uint32_t lv_type);
+struct logical_volume *find_pvmove_lv_from_pvname(struct cmd_context *cmd,
+						  struct volume_group *vg,
+						  const char *name,
+						  uint32_t lv_type);
+const char *get_pvmove_pvname_from_lv(struct logical_volume *lv);
+const char *get_pvmove_pvname_from_lv_mirr(struct logical_volume *lv_mirr);
+float copy_percent(struct logical_volume *lv_mirr);
 struct list *lvs_using_lv(struct cmd_context *cmd, struct volume_group *vg,
 			  struct logical_volume *lv);
 

lib/metadata/mirror.c

@@ -16,6 +16,9 @@
 #include "lib.h"
 #include "metadata.h"
 #include "toolcontext.h"
+#include "segtype.h"
+#include "display.h"
+#include "activate.h"
 
 /* 
  * Replace any LV segments on given PV with temporary mirror.
@@ -23,7 +26,7 @@
  */
 int insert_pvmove_mirrors(struct cmd_context *cmd,
 			  struct logical_volume *lv_mirr,
-			  struct physical_volume *pv,
+			  struct list *source_pvl,
 			  struct logical_volume *lv,
 			  struct list *allocatable_pvs,
 			  struct list *lvs_changed)
@@ -31,43 +34,135 @@ int insert_pvmove_mirrors(struct cmd_context *cmd,
 	struct list *segh;
 	struct lv_segment *seg;
 	struct lv_list *lvl;
+	struct pv_list *pvl;
 	int lv_used = 0;
 	uint32_t s, start_le, extent_count = 0u;
+	struct segment_type *segtype;
+	struct pe_range *per;
+	uint32_t pe_start, pe_end, per_end, stripe_multiplier;
 
+	/* Only 1 PV may feature in source_pvl */
+	pvl = list_item(source_pvl->n, struct pv_list);
+
+	if (!(segtype = get_segtype_from_string(lv->vg->cmd, "mirror"))) {
+		stack;
+		return 0;
+	}
+
+        if (activation() && segtype->ops->target_present &&
+            !segtype->ops->target_present()) {
+                log_error("%s: Required device-mapper target(s) not "
+                          "detected in your kernel", segtype->name);
+                return 0;
+        }
+
+	/* Split LV segments to match PE ranges */
 	list_iterate(segh, &lv->segments) {
 		seg = list_item(segh, struct lv_segment);
 		for (s = 0; s < seg->area_count; s++) {
 			if (seg->area[s].type != AREA_PV ||
-			    seg->area[s].u.pv.pv->dev != pv->dev)
+			    seg->area[s].u.pv.pv->dev != pvl->pv->dev)
 				continue;
 
-			if (!lv_used) {
-				if (!(lvl = pool_alloc(cmd->mem, sizeof(*lvl)))) {
-					log_error("lv_list alloc failed");
+			/* Do these PEs need moving? */
+			list_iterate_items(per, pvl->pe_ranges) {
+				pe_start = seg->area[s].u.pv.pe;
+				pe_end = pe_start + seg->area_len - 1;
+				per_end = per->start + per->count - 1;
+
+				/* No overlap? */
+				if ((pe_end < per->start) ||
+				    (pe_start > per_end))
+					continue;
+
+				if (seg->segtype->flags & SEG_AREAS_STRIPED)
+					stripe_multiplier = seg->area_count;
+				else
+					stripe_multiplier = 1;
+
+				if ((per->start != pe_start &&
+				     per->start > pe_start) &&
+				    !lv_split_segment(lv, seg->le +
+						      (per->start - pe_start) *
+						      stripe_multiplier)) {
+					stack;
+					return 0;
+				}
+
+				if ((per_end != pe_end &&
+				     per_end < pe_end) &&
+				    !lv_split_segment(lv, seg->le +
+						      (per_end - pe_start + 1) *
+						      stripe_multiplier)) {
+					stack;
 					return 0;
 				}
-				lvl->lv = lv;
-				list_add(lvs_changed, &lvl->list);
-				lv_used = 1;
 			}
+		}
+	}
 
-			start_le = lv_mirr->le_count;
-			if (!lv_extend_mirror(lv->vg->fid, lv_mirr,
-					      seg->area[s].u.pv.pv,
-					      seg->area[s].u.pv.pe,
-					      seg->area_len, allocatable_pvs,
-					      PVMOVE)) {
-				log_error("Allocation for temporary "
-					  "pvmove LV failed");
-				return 0;
+	/* Work through all segments on the supplied PV */
+	list_iterate(segh, &lv->segments) {
+		seg = list_item(segh, struct lv_segment);
+		for (s = 0; s < seg->area_count; s++) {
+			if (seg->area[s].type != AREA_PV ||
+			    seg->area[s].u.pv.pv->dev != pvl->pv->dev)
+				continue;
+
+			pe_start = seg->area[s].u.pv.pe;
+
+			/* Do these PEs need moving? */
+			list_iterate_items(per, pvl->pe_ranges) {
+				per_end = per->start + per->count - 1;
+
+				if ((pe_start < per->start) ||
+				    (pe_start > per_end))
+					continue;
+
+				log_debug("Matched PE range %u-%u against "
+					  "%s %u len %u", per->start, per_end,
+					  dev_name(seg->area[s].u.pv.pv->dev),
+					  seg->area[s].u.pv.pe, seg->area_len);
+
+				/* First time, add LV to list of LVs affected */
+				if (!lv_used) {
+					if (!(lvl = pool_alloc(cmd->mem, sizeof(*lvl)))) {
+						log_error("lv_list alloc failed");
+						return 0;
+					}
+					lvl->lv = lv;
+					list_add(lvs_changed, &lvl->list);
+					lv_used = 1;
+				}
+	
+				log_very_verbose("Moving %s:%u-%u of %s/%s",
+						 dev_name(pvl->pv->dev),
+						 seg->area[s].u.pv.pe,
+						 seg->area[s].u.pv.pe +
+						     seg->area_len - 1,
+						 lv->vg->name, lv->name);
+
+				start_le = lv_mirr->le_count;
+				if (!lv_extend(lv->vg->fid, lv_mirr, segtype, 1,
+				       	seg->area_len, 0u, seg->area_len,
+				       	seg->area[s].u.pv.pv,
+				       	seg->area[s].u.pv.pe,
+				       	PVMOVE, allocatable_pvs,
+				       	lv->alloc)) {
+					log_error("Unable to allocate "
+						  "temporary LV for pvmove.");
+					return 0;
+				}
+				seg->area[s].type = AREA_LV;
+				seg->area[s].u.lv.lv = lv_mirr;
+				seg->area[s].u.lv.le = start_le;
+	
+				extent_count += seg->area_len;
+	
+				lv->status |= LOCKED;
+
+				break;
 			}
-			seg->area[s].type = AREA_LV;
-			seg->area[s].u.lv.lv = lv_mirr;
-			seg->area[s].u.lv.le = start_le;
-
-			extent_count += seg->area_len;
-
-			lv->status |= LOCKED;
 		}
 	}
 
@@ -77,6 +172,7 @@ int insert_pvmove_mirrors(struct cmd_context *cmd,
 	return 1;
 }
 
+/* Remove a temporary mirror */
 int remove_pvmove_mirrors(struct volume_group *vg,
 			  struct logical_volume *lv_mirr)
 {
@@ -85,11 +181,13 @@ int remove_pvmove_mirrors(struct volume_group *vg,
 	struct lv_segment *seg, *mir_seg;
 	uint32_t s, c;
 
+	/* Loop through all LVs except the temporary mirror */
 	list_iterate(lvh, &vg->lvs) {
 		lv1 = list_item(lvh, struct lv_list)->lv;
 		if (lv1 == lv_mirr)
 			continue;
 
+		/* Find all segments that point at the temporary mirror */
 		list_iterate(segh, &lv1->segments) {
 			seg = list_item(segh, struct lv_segment);
 			for (s = 0; s < seg->area_count; s++) {
@@ -97,14 +195,19 @@ int remove_pvmove_mirrors(struct volume_group *vg,
 				    seg->area[s].u.lv.lv != lv_mirr)
 					continue;
 
+				/* Find the mirror segment pointed at */
 				if (!(mir_seg = find_seg_by_le(lv_mirr,
-							       seg->area[s].u.
-							       lv.le))) {
+							       seg->area[s].
+							       u.lv.le))) {
+					/* FIXME Error message */
 					log_error("No segment found with LE");
 					return 0;
 				}
 
-				if (mir_seg->type != SEG_MIRRORED ||
+				/* Check the segment params are compatible */
+				/* FIXME Improve error mesg & remove restrcn */
+				if ((!(mir_seg->segtype->flags
+				       & SEG_AREAS_MIRRORED)) ||
 				    !(mir_seg->status & PVMOVE) ||
 				    mir_seg->le != seg->area[s].u.lv.le ||
 				    mir_seg->area_count != 2 ||
@@ -113,7 +216,11 @@ int remove_pvmove_mirrors(struct volume_group *vg,
 					return 0;
 				}
 
-				if (mir_seg->extents_moved == mir_seg->area_len)
+				/* Replace original segment with newly-mirrored
+				 * area (or original if reverting)
+				 */
+				if (mir_seg->extents_copied == 
+				        mir_seg->area_len)
 					c = 1;
 				else
 					c = 0;
@@ -122,36 +229,47 @@ int remove_pvmove_mirrors(struct volume_group *vg,
 				seg->area[s].u.pv.pv = mir_seg->area[c].u.pv.pv;
 				seg->area[s].u.pv.pe = mir_seg->area[c].u.pv.pe;
 
-				mir_seg->type = SEG_STRIPED;
+				/* Replace mirror with old area */
+				if (!
+				    (mir_seg->segtype =
+				     get_segtype_from_string(vg->cmd,
+							     "striped"))) {
+					log_error("Missing striped segtype");
+					return 0;
+				}
 				mir_seg->area_count = 1;
 
+				/* FIXME Assumes only one pvmove at a time! */
 				lv1->status &= ~LOCKED;
 			}
 		}
+		if (!lv_merge_segments(lv1))
+			stack;
+
 	}
 
+
 	return 1;
 }
 
-struct physical_volume *get_pvmove_pv_from_lv_mirr(struct logical_volume
-						   *lv_mirr)
+const char *get_pvmove_pvname_from_lv_mirr(struct logical_volume *lv_mirr)
 {
 	struct list *segh;
 	struct lv_segment *seg;
 
 	list_iterate(segh, &lv_mirr->segments) {
 		seg = list_item(segh, struct lv_segment);
-		if (seg->type != SEG_MIRRORED)
+		if (!(seg->segtype->flags & SEG_AREAS_MIRRORED))
 			continue;
 		if (seg->area[0].type != AREA_PV)
 			continue;
-		return seg->area[0].u.pv.pv;
+		return dev_name(seg->area[0].u.pv.pv->dev);
 	}
 
 	return NULL;
 }
 
-struct physical_volume *get_pvmove_pv_from_lv(struct logical_volume *lv)
+const char *get_pvmove_pvname_from_lv(struct logical_volume *lv)
 {
 	struct list *segh;
 	struct lv_segment *seg;
@@ -162,7 +280,7 @@ struct physical_volume *get_pvmove_pv_from_lv(struct logical_volume *lv)
 		for (s = 0; s < seg->area_count; s++) {
 			if (seg->area[s].type != AREA_LV)
 				continue;
-			return get_pvmove_pv_from_lv_mirr(seg->area[s].u.lv.lv);
+			return get_pvmove_pvname_from_lv_mirr(seg->area[s].u.lv.lv);
 		}
 	}
 
@@ -170,7 +288,8 @@ struct physical_volume *get_pvmove_pv_from_lv(struct logical_volume *lv)
 }
 
 struct logical_volume *find_pvmove_lv(struct volume_group *vg,
-				      struct device *dev)
+				      struct device *dev,
+				      uint32_t lv_type)
 {
 	struct list *lvh, *segh;
 	struct logical_volume *lv;
@@ -180,9 +299,10 @@ struct logical_volume *find_pvmove_lv(struct volume_group *vg,
 	list_iterate(lvh, &vg->lvs) {
 		lv = list_item(lvh, struct lv_list)->lv;
 
-		if (!(lv->status & PVMOVE))
+		if (!(lv->status & lv_type))
 			continue;
 
+		/* Check segment origins point to pvname */
 		list_iterate(segh, &lv->segments) {
 			seg = list_item(segh, struct lv_segment);
 			if (seg->area[0].type != AREA_PV)
@@ -196,6 +316,21 @@ struct logical_volume *find_pvmove_lv(struct volume_group *vg,
 	return NULL;
 }
 
+struct logical_volume *find_pvmove_lv_from_pvname(struct cmd_context *cmd,
+					 	  struct volume_group *vg,
+				      		  const char *name,
+				      		  uint32_t lv_type)
+{
+	struct physical_volume *pv;
+
+	if (!(pv = find_pv_by_name(cmd, name))) {
+		stack;
+		return NULL;
+	}
+
+	return find_pvmove_lv(vg, pv->dev, lv_type);
+}
+
 struct list *lvs_using_lv(struct cmd_context *cmd, struct volume_group *vg,
 			  struct logical_volume *lv)
 {
@@ -218,6 +353,7 @@ struct list *lvs_using_lv(struct cmd_context *cmd, struct volume_group *vg,
 		if (lv1 == lv)
 			continue;
 
+		/* Find whether any segment points at the supplied LV */
 		list_iterate(segh, &lv1->segments) {
 			seg = list_item(segh, struct lv_segment);
 			for (s = 0; s < seg->area_count; s++) {
@@ -240,7 +376,7 @@ struct list *lvs_using_lv(struct cmd_context *cmd, struct volume_group *vg,
 	return lvs;
 }
 
-float pvmove_percent(struct logical_volume *lv_mirr)
+float copy_percent(struct logical_volume *lv_mirr)
 {
 	uint32_t numerator = 0u, denominator = 0u;
 	struct list *segh;
@@ -248,11 +384,13 @@ float pvmove_percent(struct logical_volume *lv_mirr)
 
 	list_iterate(segh, &lv_mirr->segments) {
 		seg = list_item(segh, struct lv_segment);
-		if (!(seg->status & PVMOVE))
-			continue;
 
-		numerator += seg->extents_moved;
 		denominator += seg->area_len;
+
+		if (seg->segtype->flags & SEG_AREAS_MIRRORED)
+			numerator += seg->extents_copied;
+		else
+			numerator += seg->area_len;
 	}
 
 	return denominator ? (float) numerator *100 / denominator : 100.0;

lib/metadata/pv_map.c

@@ -206,13 +206,13 @@ static int _create_areas(struct pool *mem, struct pv_map *pvm, uint32_t start,
 static int _create_allocatable_areas(struct pool *mem, struct pv_map *pvm)
 {
 	struct list *alloc_areas, *aah;
-	struct alloc_area *aa;
+	struct pe_range *aa;
 
-	alloc_areas = pvm->pvl->alloc_areas;
+	alloc_areas = pvm->pvl->pe_ranges;
 
 	if (alloc_areas) {
 		list_iterate(aah, alloc_areas) {
-			aa = list_item(aah, struct alloc_area);
+			aa = list_item(aah, struct pe_range);
 			if (!_create_areas(mem, pvm, aa->start, aa->count)) {
 				stack;
 				return 0;

lib/metadata/segtype.c

@@ -0,0 +1,32 @@
+/*
+ * Copyright (C) 2003-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU General Public License v.2.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include "lib.h"
+#include "toolcontext.h"
+#include "segtype.h"
+
+struct segment_type *get_segtype_from_string(struct cmd_context *cmd,
+					     const char *str)
+{
+	struct segment_type *segtype;
+
+	list_iterate_items(segtype, &cmd->segtypes) {
+		if (!strcmp(segtype->name, str))
+			return segtype;
+	}
+
+	log_error("Unrecognised segment type %s", str);
+	return NULL;
+}

lib/metadata/segtype.h

@@ -0,0 +1,92 @@
+/*
+ * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU General Public License v.2.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#ifndef _SEGTYPES_H
+#define _SEGTYPES_H
+
+struct segtype_handler;
+struct cmd_context;
+struct config_tree;
+struct lv_segment;
+struct formatter;
+struct config_node;
+struct hash_table;
+struct dev_manager;
+
+/* Feature flags */
+#define SEG_CAN_SPLIT		0x00000001
+#define SEG_AREAS_STRIPED	0x00000002
+#define SEG_AREAS_MIRRORED	0x00000004
+#define SEG_SNAPSHOT		0x00000008
+#define SEG_FORMAT1_SUPPORT	0x00000010
+#define SEG_VIRTUAL		0x00000020
+
+struct segment_type {
+	struct list list;
+	struct cmd_context *cmd;
+	uint32_t flags;
+	struct segtype_handler *ops;
+	const char *name;
+	void *library;
+	void *private;
+};
+
+struct segtype_handler {
+	const char *(*name) (const struct lv_segment * seg);
+	void (*display) (const struct lv_segment * seg);
+	int (*text_export) (const struct lv_segment * seg,
+			    struct formatter * f);
+	int (*text_import_area_count) (struct config_node * sn,
+				       uint32_t *area_count);
+	int (*text_import) (struct lv_segment * seg,
+			    const struct config_node * sn,
+			    struct hash_table * pv_hash);
+	int (*merge_segments) (struct lv_segment * seg1,
+			       struct lv_segment * seg2);
+	int (*compose_target_line) (struct dev_manager * dm, struct pool * mem,
+				    struct config_tree * cft,
+				    void **target_state,
+				    struct lv_segment * seg, char *params,
+				    size_t paramsize, const char **target,
+				    int *pos, uint32_t *pvmove_mirror_count);
+	int (*target_percent) (void **target_state, struct pool * mem,
+			       struct config_tree * cft,
+			       struct lv_segment * seg, char *params,
+			       uint64_t *total_numerator,
+			       uint64_t *total_denominator, float *percent);
+	int (*target_present) (void);
+	void (*destroy) (const struct segment_type * segtype);
+};
+
+struct segment_type *get_segtype_from_string(struct cmd_context *cmd,
+					     const char *str);
+
+struct segment_type *init_striped_segtype(struct cmd_context *cmd);
+struct segment_type *init_zero_segtype(struct cmd_context *cmd);
+struct segment_type *init_error_segtype(struct cmd_context *cmd);
+
+#ifdef SNAPSHOT_INTERNAL
+struct segment_type *init_snapshot_segtype(struct cmd_context *cmd);
+#endif
+
+#ifdef MIRRORED_INTERNAL
+struct segment_type *init_mirrored_segtype(struct cmd_context *cmd);
+#endif
+
+#ifdef CRYPT_INTERNAL
+struct segment_type *init_crypt_segtype(struct cmd_context *cmd);
+#endif
+
+#endif

lib/mirror/.exported_symbols

@@ -0,0 +1 @@
+init_segtype