add read_ahead functions to library and dmsetup --readahead

(Not live yet.)

BUGS

@@ -1 +0,0 @@
-LVM2's device-mapper driver and ext3 are incompatible at the moment.

COPYING.LIB

@@ -1,14 +1,14 @@
+		  GNU LESSER GENERAL PUBLIC LICENSE
+		       Version 2.1, February 1999
 
-		  GNU LIBRARY GENERAL PUBLIC LICENSE
-		       Version 2, June 1991
-
- Copyright (C) 1991 Free Software Foundation, Inc.
- 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA
+ Copyright (C) 1991, 1999 Free Software Foundation, Inc.
+ 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
  Everyone is permitted to copy and distribute verbatim copies
  of this license document, but changing it is not allowed.
 
-[This is the first released version of the library GPL.  It is
- numbered 2 because it goes with version 2 of the ordinary GPL.]
+[This is the first released version of the Lesser GPL.  It also counts
+ as the successor of the GNU Library Public License, version 2, hence
+ the version number 2.1.]
 
 			    Preamble
 
@@ -17,97 +17,109 @@ freedom to share and change it.  By contrast, the GNU General Public
 Licenses are intended to guarantee your freedom to share and change
 free software--to make sure the software is free for all its users.
 
-  This license, the Library General Public License, applies to some
-specially designated Free Software Foundation software, and to any
-other libraries whose authors decide to use it.  You can use it for
-your libraries, too.
+  This license, the Lesser General Public License, applies to some
+specially designated software packages--typically libraries--of the
+Free Software Foundation and other authors who decide to use it.  You
+can use it too, but we suggest you first think carefully about whether
+this license or the ordinary General Public License is the better
+strategy to use in any particular case, based on the explanations below.
 
-  When we speak of free software, we are referring to freedom, not
-price.  Our General Public Licenses are designed to make sure that you
-have the freedom to distribute copies of free software (and charge for
-this service if you wish), that you receive source code or can get it
-if you want it, that you can change the software or use pieces of it
-in new free programs; and that you know you can do these things.
+  When we speak of free software, we are referring to freedom of use,
+not price.  Our General Public Licenses are designed to make sure that
+you have the freedom to distribute copies of free software (and charge
+for this service if you wish); that you receive source code or can get
+it if you want it; that you can change the software and use pieces of
+it in new free programs; and that you are informed that you can do
+these things.
 
   To protect your rights, we need to make restrictions that forbid
-anyone to deny you these rights or to ask you to surrender the rights.
-These restrictions translate to certain responsibilities for you if
-you distribute copies of the library, or if you modify it.
+distributors to deny you these rights or to ask you to surrender these
+rights.  These restrictions translate to certain responsibilities for
+you if you distribute copies of the library or if you modify it.
 
   For example, if you distribute copies of the library, whether gratis
 or for a fee, you must give the recipients all the rights that we gave
 you.  You must make sure that they, too, receive or can get the source
-code.  If you link a program with the library, you must provide
-complete object files to the recipients so that they can relink them
-with the library, after making changes to the library and recompiling
+code.  If you link other code with the library, you must provide
+complete object files to the recipients, so that they can relink them
+with the library after making changes to the library and recompiling
 it.  And you must show them these terms so they know their rights.
 
-  Our method of protecting your rights has two steps: (1) copyright
-the library, and (2) offer you this license which gives you legal
+  We protect your rights with a two-step method: (1) we copyright the
+library, and (2) we offer you this license, which gives you legal
 permission to copy, distribute and/or modify the library.
 
-  Also, for each distributor's protection, we want to make certain
-that everyone understands that there is no warranty for this free
-library.  If the library is modified by someone else and passed on, we
-want its recipients to know that what they have is not the original
-version, so that any problems introduced by others will not reflect on
-the original authors' reputations.
+  To protect each distributor, we want to make it very clear that
+there is no warranty for the free library.  Also, if the library is
+modified by someone else and passed on, the recipients should know
+that what they have is not the original version, so that the original
+author's reputation will not be affected by problems that might be
+introduced by others.
 
-  Finally, any free program is threatened constantly by software
-patents.  We wish to avoid the danger that companies distributing free
-software will individually obtain patent licenses, thus in effect
-transforming the program into proprietary software.  To prevent this,
-we have made it clear that any patent must be licensed for everyone's
-free use or not licensed at all.
+  Finally, software patents pose a constant threat to the existence of
+any free program.  We wish to make sure that a company cannot
+effectively restrict the users of a free program by obtaining a
+restrictive license from a patent holder.  Therefore, we insist that
+any patent license obtained for a version of the library must be
+consistent with the full freedom of use specified in this license.
 
-  Most GNU software, including some libraries, is covered by the ordinary
-GNU General Public License, which was designed for utility programs.  This
-license, the GNU Library General Public License, applies to certain
-designated libraries.  This license is quite different from the ordinary
-one; be sure to read it in full, and don't assume that anything in it is
-the same as in the ordinary license.
+  Most GNU software, including some libraries, is covered by the
+ordinary GNU General Public License.  This license, the GNU Lesser
+General Public License, applies to certain designated libraries, and
+is quite different from the ordinary General Public License.  We use
+this license for certain libraries in order to permit linking those
+libraries into non-free programs.
 
-  The reason we have a separate public license for some libraries is that
-they blur the distinction we usually make between modifying or adding to a
-program and simply using it.  Linking a program with a library, without
-changing the library, is in some sense simply using the library, and is
-analogous to running a utility program or application program.  However, in
-a textual and legal sense, the linked executable is a combined work, a
-derivative of the original library, and the ordinary General Public License
-treats it as such.
+  When a program is linked with a library, whether statically or using
+a shared library, the combination of the two is legally speaking a
+combined work, a derivative of the original library.  The ordinary
+General Public License therefore permits such linking only if the
+entire combination fits its criteria of freedom.  The Lesser General
+Public License permits more lax criteria for linking other code with
+the library.
 
-  Because of this blurred distinction, using the ordinary General
-Public License for libraries did not effectively promote software
-sharing, because most developers did not use the libraries.  We
-concluded that weaker conditions might promote sharing better.
+  We call this license the "Lesser" General Public License because it
+does Less to protect the user's freedom than the ordinary General
+Public License.  It also provides other free software developers Less
+of an advantage over competing non-free programs.  These disadvantages
+are the reason we use the ordinary General Public License for many
+libraries.  However, the Lesser license provides advantages in certain
+special circumstances.
 
-  However, unrestricted linking of non-free programs would deprive the
-users of those programs of all benefit from the free status of the
-libraries themselves.  This Library General Public License is intended to
-permit developers of non-free programs to use free libraries, while
-preserving your freedom as a user of such programs to change the free
-libraries that are incorporated in them.  (We have not seen how to achieve
-this as regards changes in header files, but we have achieved it as regards
-changes in the actual functions of the Library.)  The hope is that this
-will lead to faster development of free libraries.
+  For example, on rare occasions, there may be a special need to
+encourage the widest possible use of a certain library, so that it becomes
+a de-facto standard.  To achieve this, non-free programs must be
+allowed to use the library.  A more frequent case is that a free
+library does the same job as widely used non-free libraries.  In this
+case, there is little to gain by limiting the free library to free
+software only, so we use the Lesser General Public License.
+
+  In other cases, permission to use a particular library in non-free
+programs enables a greater number of people to use a large body of
+free software.  For example, permission to use the GNU C Library in
+non-free programs enables many more people to use the whole GNU
+operating system, as well as its variant, the GNU/Linux operating
+system.
+
+  Although the Lesser General Public License is Less protective of the
+users' freedom, it does ensure that the user of a program that is
+linked with the Library has the freedom and the wherewithal to run
+that program using a modified version of the Library.
 
   The precise terms and conditions for copying, distribution and
 modification follow.  Pay close attention to the difference between a
 "work based on the library" and a "work that uses the library".  The
-former contains code derived from the library, while the latter only
-works together with the library.
-
-  Note that it is possible for a library to be covered by the ordinary
-General Public License rather than by this special one.
+former contains code derived from the library, whereas the latter must
+be combined with the library in order to run.
 
-		  GNU LIBRARY GENERAL PUBLIC LICENSE
+		  GNU LESSER GENERAL PUBLIC LICENSE
    TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
 
-  0. This License Agreement applies to any software library which
-contains a notice placed by the copyright holder or other authorized
-party saying it may be distributed under the terms of this Library
-General Public License (also called "this License").  Each licensee is
-addressed as "you".
+  0. This License Agreement applies to any software library or other
+program which contains a notice placed by the copyright holder or
+other authorized party saying it may be distributed under the terms of
+this Lesser General Public License (also called "this License").
+Each licensee is addressed as "you".
 
   A "library" means a collection of software functions and/or data
 prepared so as to be conveniently linked with application programs
@@ -256,7 +268,7 @@ distribute the object code for the work under the terms of Section 6.
 Any executables containing that work also fall under Section 6,
 whether or not they are linked directly with the Library itself.
 
-  6. As an exception to the Sections above, you may also compile or
+  6. As an exception to the Sections above, you may also combine or
 link a "work that uses the Library" with the Library to produce a
 work containing portions of the Library, and distribute that work
 under terms of your choice, provided that the terms permit
@@ -283,23 +295,31 @@ of these things:
     Library will not necessarily be able to recompile the application
     to use the modified definitions.)
 
-    b) Accompany the work with a written offer, valid for at
+    b) Use a suitable shared library mechanism for linking with the
+    Library.  A suitable mechanism is one that (1) uses at run time a
+    copy of the library already present on the user's computer system,
+    rather than copying library functions into the executable, and (2)
+    will operate properly with a modified version of the library, if
+    the user installs one, as long as the modified version is
+    interface-compatible with the version that the work was made with.
+
+    c) Accompany the work with a written offer, valid for at
     least three years, to give the same user the materials
     specified in Subsection 6a, above, for a charge no more
     than the cost of performing this distribution.
 
-    c) If distribution of the work is made by offering access to copy
+    d) If distribution of the work is made by offering access to copy
     from a designated place, offer equivalent access to copy the above
     specified materials from the same place.
 
-    d) Verify that the user has already received a copy of these
+    e) Verify that the user has already received a copy of these
     materials or that you have already sent this user a copy.
 
   For an executable, the required form of the "work that uses the
 Library" must include any data and utility programs needed for
 reproducing the executable from it.  However, as a special exception,
-the source code distributed need not include anything that is normally
-distributed (in either source or binary form) with the major
+the materials to be distributed need not include anything that is
+normally distributed (in either source or binary form) with the major
 components (compiler, kernel, and so on) of the operating system on
 which the executable runs, unless that component itself accompanies
 the executable.
@@ -348,7 +368,7 @@ Library), the recipient automatically receives a license from the
 original licensor to copy, distribute, link with or modify the Library
 subject to these terms and conditions.  You may not impose any further
 restrictions on the recipients' exercise of the rights granted herein.
-You are not responsible for enforcing compliance by third parties to
+You are not responsible for enforcing compliance by third parties with
 this License.
 
   11. If, as a consequence of a court judgment or allegation of patent
@@ -391,7 +411,7 @@ excluded.  In such case, this License incorporates the limitation as if
 written in the body of this License.
 
   13. The Free Software Foundation may publish revised and/or new
-versions of the Library General Public License from time to time.
+versions of the Lesser General Public License from time to time.
 Such new versions will be similar in spirit to the present version,
 but may differ in detail to address new problems or concerns.
 
@@ -437,7 +457,7 @@ DAMAGES.
 
 		     END OF TERMS AND CONDITIONS
 
-     Appendix: How to Apply These Terms to Your New Libraries
+           How to Apply These Terms to Your New Libraries
 
   If you develop a new library, and you want it to be of the greatest
 possible use to the public, we recommend making it free software that
@@ -454,19 +474,18 @@ convey the exclusion of warranty; and each file should have at least the
     Copyright (C) <year>  <name of author>
 
     This library is free software; you can redistribute it and/or
-    modify it under the terms of the GNU Library General Public
+    modify it under the terms of the GNU Lesser General Public
     License as published by the Free Software Foundation; either
-    version 2 of the License, or (at your option) any later version.
+    version 2.1 of the License, or (at your option) any later version.
 
     This library is distributed in the hope that it will be useful,
     but WITHOUT ANY WARRANTY; without even the implied warranty of
     MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-    Library General Public License for more details.
+    Lesser General Public License for more details.
 
-    You should have received a copy of the GNU Library General Public
-    License along with this library; if not, write to the Free
-    Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
-    MA 02111-1307, USA
+    You should have received a copy of the GNU Lesser General Public
+    License along with this library; if not, write to the Free Software
+    Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
 
 Also add information on how to contact you by electronic and paper mail.
 
@@ -481,3 +500,5 @@ necessary.  Here is a sample; alter the names:
   Ty Coon, President of Vice
 
 That's all there is to it!
+
+

INSTALL

@@ -6,8 +6,8 @@ LVM2 installation
    Ensure the device-mapper has been installed on the machine.
 
    The device-mapper should be in the kernel (look for 'device-mapper'
-   messages in the kernel logs) and /usr/include/libdevmapper.h should 
-   be present.
+   messages in the kernel logs) and /usr/include/libdevmapper.h 
+   and libdevmapper.so should be present.
 
    The device-mapper is available from:
      ftp://ftp.sistina.com/pub/LVM2/device-mapper/
@@ -17,9 +17,15 @@ LVM2 installation
 
    Run the 'configure' script from the top directory.
 
-   If you do not have GNU readline (http://www.gnu.org/directory/readline.html)
-   installed use
-     ./configure --disable-readline
+   If you wish to use the built-in LVM2 shell and have GNU readline 
+   installed (http://www.gnu.org/directory/readline.html) use:
+     ./configure --enable-readline
+
+   If you don't want to include the LVM1 backwards-compatibility code use:
+     ./configure --with-lvm1=none 
+
+   To separate the LVM1 support into a shared library loaded by lvm.conf use:
+     ./configure --with-lvm1=shared
 
 
 3) Build and install LVM2.
@@ -31,6 +37,9 @@ LVM2 installation
 
    The tools will work fine without a configuration file being
    present, but you ought to review the example file in doc/example.conf.
-   For example, specifying the devices that LVM2 is to use should
+   For example, specifying the devices that LVM2 is to use can
    make the tools run more efficiently - and avoid scanning /dev/cdrom!
 
+Please also refer to the WHATS_NEW file and the manual pages for the 
+individual commands.
+

INTRO

@@ -1,18 +0,0 @@
-An introduction to LVM2
-=======================
-
-Background
-
-
-Compatibility with LVM1
-
-
-New features
-
-
-Missing features
-
-
-Future enhancements
-
-

Makefile.in

@@ -1,34 +1,73 @@
 #
-# Copyright (C) 2001 Sistina Software
+# Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.
+# Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved.
 #
-# This LVM library is free software; you can redistribute it and/or
-# modify it under the terms of the GNU Library General Public
-# License as published by the Free Software Foundation; either
-# version 2 of the License, or (at your option) any later version.
+# This file is part of LVM2.
 #
-# This LVM library is distributed in the hope that it will be useful,
-# but WITHOUT ANY WARRANTY; without even the implied warranty of
-# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-# Library General Public License for more details.
+# 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 Library General Public
-# License along with this LVM library; if not, write to the Free
-# Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
-# MA 02111-1307, USA
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software Foundation,
+# Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 
 srcdir = @srcdir@
 top_srcdir = @top_srcdir@
 VPATH = @srcdir@
 
-SUBDIRS = include man lib tools
+SUBDIRS = doc include man scripts
+
+ifeq ("@INTL@", "yes")
+  SUBDIRS += po
+endif
+
+SUBDIRS += lib tools daemons
+
+ifeq ("@DMEVENTD@", "yes")
+  SUBDIRS += dmeventd
+endif
 
 ifeq ($(MAKECMDGOALS),distclean)
-  SUBDIRS += test/mm test/device test/format1 test/regex test/filters
+  SUBDIRS += daemons/clvmd \
+	     dmeventd \
+	     lib/format1 \
+	     lib/format_pool \
+	     lib/locking \
+	     lib/mirror \
+	     lib/snapshot \
+	     po \
+	     test/mm test/device test/format1 test/regex test/filters
+  DISTCLEAN_TARGETS += lib/misc/configure.h
 endif
 
 include make.tmpl
 
+daemons: lib
 lib: include
-tools: include lib
+tools: lib
+dmeventd: tools
+po: tools daemons dmeventd
 
+ifeq ("@INTL@", "yes")
+lib.pofile: include.pofile
+tools.pofile: lib.pofile
+daemons.pofile: lib.pofile
+dmeventd.pofile: tools.pofile
+po.pofile: tools.pofile daemons.pofile dmeventd.pofile
+pofile: po.pofile
+endif
 
+ifneq ("@CFLOW_CMD@", "")
+tools.cflow: lib.cflow
+cflow: tools.cflow
+endif
+
+ifneq ("@CSCOPE_CMD@", "")
+cscope.out: tools
+	@CSCOPE_CMD@ -b -R
+all: cscope.out
+endif
+
+check: all
+	$(MAKE) -C test all

README

@@ -1,25 +1,23 @@
-This directory contains a beta release of LMV2, the new version of 
-the userland LVM tools designed for the new device-mapper for 
-the Linux kernel.
+This directory contains LVM2, the new version of the userland LVM
+tools designed for the new device-mapper for the Linux kernel.
 
 The device-mapper needs to be installed before compiling these LVM2 tools.
 
-For more information about LVM2 read the INTRO file.
+For more information about LVM2 read the WHATS_NEW file.
 Installation instructions are in INSTALL.
 
-This is beta-quality software, released for testing purposes only.
 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@tech.sistina.com:/data/cvs login
-  CVS password: cvs1
-  cvs -d :pserver:cvs@tech.sistina.com:/data/cvs checkout LVM2
+  cvs -d :pserver:cvs@sources.redhat.com:/cvs/lvm2 login
+  CVS password: cvs
+  cvs -d :pserver:cvs@sources.redhat.com:/cvs/lvm2 co LVM2
 
 Mailing list for discussion/bug reports etc.
-  lvm-devel@sistina.com
-  Subscribe from http://lists.sistina.com/mailman/listinfo/lvm-devel
+  linux-lvm@redhat.com
+  Subscribe from https://www.redhat.com/mailman/listinfo/linux-lvm
 

TODO

@@ -1,28 +0,0 @@
-before 2.0
------------
-
-vgexport
-vgimport
-snapshots
-pvmove
-device-mapper support for 2.5 kernel series
-review FIXMEs
-extra validation & full consistency checks in format1 with LVM1
-partial activation (aka VG quorum)
-error message review
-locking during metadata changes
-format2 with atomic transactions
-bidirectional format1/format2 migration tool
-persistent minors
-statistics target and tool support
-review tool exit codes for LVM1 compatibility
-
-before 2.1
-----------
-
-e2fsadm
-lvmsadc
-lvmsar
-pvdata
-vgsplit
-vgmknodes

VERSION

@@ -1 +1 @@
-1.95.07-cvs (2002-04-30)
+2.02.29-cvs (2007-08-24)

WHATS_NEW_DM

@@ -0,0 +1,313 @@
+Version 1.02.23 - 
+==================================
+  Add --readahead to dmsetup.
+  Add external read_ahead library functions and DM_READ_AHEAD_* definitions.
+  Fix double free in a libdevmapper-event error path.
+  Fix configure --with-dmeventd-path substitution.
+  Allow $DM_DEV_DIR envvar to override default of "/dev".
+  Create e.g., libdevmapper.so.1.02, in build dir alongside the .so file.
+  Avoid static link failure with some SELinux libraries.
+  Remove obsolete dmfs code from tree and update INSTALL.
+
+Version 1.02.22 - 21st August 2007
+==================================
+  Fix inconsistent licence notices: executables are GPLv2; libraries LGPLv2.1.
+  Update to use autoconf 2.61, while still supporting 2.57.
+  Avoid repeated dm_task free on some dm_event_get_registered_device errors.
+  Introduce log_sys_* macros from LVM2.
+  Export dm_fclose and dm_create_dir; remove libdm-file.h.
+  Don't log EROFS mkdir failures in _create_dir_recursive (for LVM2).
+  Add fclose wrapper dm_fclose that catches write failures (using ferror).
+
+Version 1.02.21 - 13th July 2007
+================================
+  Introduce _LOG_STDERR to send log_warn() messages to stderr not stdout.
+  Fix dmsetup -o devno string termination. (1.02.20) 
+
+Version 1.02.20 - 15th June 2007
+================================
+  Fix default dmsetup report buffering and add --unbuffered.
+  Add tree-based and dependency fields to dmsetup reports.
+
+Version 1.02.19 - 27th April 2007
+=================================
+  Standardise protective include file #defines.
+  Add regex functions to library.
+  Avoid trailing separator in reports when there are hidden sort fields.
+  Fix segfault in 'dmsetup status' without --showkeys against crypt target.
+  Deal with some more compiler warnings.
+  Introduce _add_field() and _is_same_field() to libdm-report.c.
+  Fix some libdevmapper-event and dmeventd memory leaks.
+  Remove unnecessary memset() return value checks.
+  Fix a few leaks in reporting error paths. [1.02.15+]
+
+Version 1.02.18 - 13th February 2007
+====================================
+  Improve dmeventd messaging protocol: drain pipe and tag messages.
+
+Version 1.02.17 - 29th January 2007
+===================================
+  Add recent reporting options to dmsetup man page.
+  Revise some report fields names.
+  Add dmsetup 'help' command and update usage text.
+  Use fixed-size fields in report interface and reorder.
+
+Version 1.02.16 - 25th January 2007
+===================================
+  Add some missing close() and fclose() return value checks.
+  Migrate dmsetup column-based output over to new libdevmapper report framework.
+  Add descriptions to reporting field definitions.
+  Add a dso-private variable to dmeventd dso interface.
+  Add dm_event_handler_[gs]et_timeout functions.
+  Streamline dm_report_field_* interface.
+  Add cmdline debug & version options to dmeventd.
+  Add DM_LIB_VERSION definition to configure.h.
+  Suppress 'Unrecognised field' error if report field is 'help'.
+  Add --separator and --sort to dmsetup (unused).
+  Make alignment flag optional when specifying report fields.
+
+Version 1.02.15 - 17th January 2007
+===================================
+  Add basic reporting functions to libdevmapper.
+  Fix a malloc error path in dmsetup message.
+  More libdevmapper-event interface changes and fixes.
+  Rename dm_saprintf() to dm_asprintf().
+  Report error if NULL pointer is supplied to dm_strdup_aux().
+  Reinstate dm_event_get_registered_device.
+
+Version 1.02.14 - 11th January 2007
+===================================
+  Add dm_saprintf().
+  Use CFLAGS when linking so mixed sparc builds can supply -m64.
+  Add dm_tree_use_no_flush_suspend().
+  Lots of dmevent changes including revised interface.
+  Export dm_basename().
+  Cope with a trailing space when comparing tables prior to possible reload.
+  Fix dmeventd to cope if monitored device disappears.
+
+Version 1.02.13 - 28 Nov 2006
+=============================
+  Update dmsetup man page (setgeometry & message).
+  Fix dmsetup free after getline with debug.
+  Suppress encryption key in 'dmsetup table' output unless --showkeys supplied.
+
+Version 1.02.12 - 13 Oct 2006
+=============================
+  Avoid deptree attempting to suspend a device that's already suspended.
+
+Version 1.02.11 -  12 Oct 2006
+==============================
+  Add suspend noflush support.
+  Add basic dmsetup loop support.
+  Switch dmsetup to use dm_malloc and dm_free.
+
+Version 1.02.10 - 19 Sep 2006
+=============================
+  Add dm_snprintf(), dm_split_words() and dm_split_lvm_name() to libdevmapper.
+  Reorder mm bounds_check code to reduce window for a dmeventd race.
+
+Version 1.02.09 - 15 Aug 2006
+=============================
+  Add --table argument to dmsetup for a one-line table.
+  Abort if errors are found during cmdline option processing.
+  Add lockfs indicator to debug output.
+
+Version 1.02.08 - 17 July 2006
+==============================
+  Append full patch to check in emails.
+  Avoid duplicate dmeventd subdir with 'make distclean'.
+  Update dmsetup man page.
+  Add --force to dmsetup remove* to load error target.
+  dmsetup remove_all also performs mknodes.
+  Don't suppress identical table reloads if permission changes.
+  Fix corelog segment line.
+  Suppress some compiler warnings.
+
+Version 1.02.07 - 11 May 2006
+=============================
+  Add DM_CORELOG flag to dm_tree_node_add_mirror_target().
+  Avoid a dmeventd compiler warning.
+
+Version 1.02.06 - 10 May 2006
+=============================
+  Move DEFS into configure.h.
+  Fix leaks in error paths found by coverity.
+  Remove dmsetup line buffer limitation.
+
+Version 1.02.05 - 19 Apr 2006
+=============================
+  Separate install_include target in makefiles.
+  Separate out DEFS from CFLAGS.
+  Support pkg-config.
+  Check for libsepol.
+
+Version 1.02.04 - 14 Apr 2006
+=============================
+  Bring dmsetup man page up-to-date.
+  Use name-based device refs if kernel doesn't support device number refs.
+  Fix memory leak (struct dm_ioctl) when struct dm_task is reused.
+  If _create_and_load_v4 fails part way through, revert the creation.
+  dmeventd thread/fifo fixes.
+  Add file & line to dm_strdup_aux().
+  Add setgeometry.
+
+Version 1.02.03 - 7 Feb 2006
+============================
+  Add exported functions to set uid, gid and mode.
+  Rename _log to dm_log and export.
+  Add dm_tree_skip_lockfs.
+  Fix dm_strdup debug definition.
+  Fix hash function to avoid using a negative array offset.
+  Don't inline _find in hash.c and tidy signed/unsigned etc.
+  Fix libdevmapper.h #endif.
+  Fix dmsetup version driver version.
+  Add sync, nosync and block_on_error mirror log parameters.
+  Add hweight32.
+  Fix dmeventd build.
+
+Version 1.02.02 - 2 Dec 2005
+============================
+  dmeventd added.
+  Export dm_task_update_nodes.
+  Use names instead of numbers in messages when ioctls fail.
+
+Version 1.02.01 - 23 Nov 2005
+=============================
+  Resume snapshot-origins last.
+  Drop leading zeros from dm_format_dev.
+  Suppress attempt to reload identical table.
+  Additional LVM- prefix matching for transitional period.
+
+Version 1.02.00 - 10 Nov 2005
+=============================
+  Added activation functions to library.
+  Added return macros.
+  Also suppress error if device doesn't exist with DM_DEVICE_STATUS.
+  Export dm_set_selinux_context().
+  Add dm_driver_version().
+  Added dependency tree functions to library.
+  Added hash, bitset, pool, dbg_malloc to library.
+  Added ls --tree to dmsetup.
+  Added dmsetup --nolockfs support for suspend/reload.
+
+Version 1.01.05 - 26 Sep 2005
+=============================
+  Resync list.h with LVM2.
+  Remember increased buffer size and use for subsequent calls.
+  On 'buffer full' condition, double buffer size and repeat ioctl.
+  Fix termination of getopt_long() option array.
+  Report 'buffer full' condition with v4 ioctl as well as with v1.
+
+Version 1.01.04 - 2 Aug 2005
+============================
+  Fix dmsetup ls -j and status --target with empty table.
+
+Version 1.01.03 - 13 Jun 2005
+=============================
+  Use matchpathcon mode parameter.
+  Fix configure script to re-enable selinux.
+
+Version 1.01.02 - 17 May 2005
+=============================
+  Call dm_lib_exit() and dm_lib_release() automatically now.
+  Add --target <target_type> filter to dmsetup table/status/ls.
+  Add --exec <command> to dmsetup ls.
+  Fix dmsetup getopt_long usage.
+
+Version 1.01.01 - 29 Mar 2005
+=============================
+  Update dmsetup man page.
+  Drop-in devmap_name replacement.
+  Add option to compile without ioctl for testing.
+  Fix DM_LIB_VERSION sed.
+
+Version 1.01.00 - 17 Jan 2005
+=============================
+  Add dm_task_no_open_count() to skip getting open_count.
+
+Version 1.00.21 - 7 Jan 2005
+============================
+  Fix /proc/devices parsing.
+
+Version 1.00.20 - 6 Jan 2005
+============================
+  Attempt to fix /dev/mapper/control transparently if it's wrong.
+  Configuration-time option for setting uid/gid/mode for /dev/mapper nodes.
+  Update kernel patches for 2.4.27/2.4.28-pre-4 (includes minor fixes).
+  Add --noheadings columns option for colon-separated dmsetup output.
+  Support device referencing by uuid or major/minor.
+  Warn if kernel data didn't fit in buffer.
+  Fix a printf.
+
+Version 1.00.19 - 3 July 2004
+=============================
+  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'
+  Fix library selinux linking.
+
+Version 1.00.16 - 16 Apr 2004
+=============================
+  Ignore error setting selinux file context if fs doesn't support it.
+
+Version 1.00.15 - 7 Apr 2004
+============================
+  Fix status overflow check in kernel patches.
+
+Version 1.00.14 - 6 Apr 2004
+============================
+  Fix static selinux build.
+
+Version 1.00.13 - 6 Apr 2004
+============================
+  Add some basic selinux support.
+
+Version 1.00.12 - 6 Apr 2004
+============================
+  Fix dmsetup.static install.
+
+Version 1.00.11 - 5 Apr 2004
+============================
+  configure --enable-static_link does static build in addition to dynamic.
+  Moved Makefile library targets definition into template.
+
+Version 1.00.10 - 2 Apr 2004
+============================
+  Fix DESTDIR handling.
+  Static build installs to dmsetup.static.
+  Basic support for internationalisation.
+  Minor Makefile tidy-ups/fixes.
+
+Version 1.00.09 - 31 Mar 2004
+=============================
+  Update copyright notices to Red Hat.
+  Move full mknodes functionality from dmsetup into libdevmapper.
+  Avoid sscanf %as for uClibc compatibility.
+  Cope if DM_LIST_VERSIONS is not defined.
+  Add DM_LIST_VERSIONS functionality to kernel patches.
+  Generate new kernel patches for 2.4.26-rc1.
+
+Version 1.00.08 - 27 Feb 2004
+=============================
+  Added 'dmsetup targets'.
+  Added event_nr support to 'dmsetup wait'.
+  Updated dmsetup man page.
+  Allow logging function to be reset to use internal one.
+  Bring log macros in line with LVM2 ones.
+  Added 'make install_static_lib' which installs libdevmapper.a.
+  Made configure/makefiles closer to LVM2 versions.
+  Fixed DESTDIR for make install/install_static_lib.
+  Updated README/INSTALL to reflect move to sources.redhat.com.
+  Updated autoconf files to 2003-06-17.
+

autoconf/install-sh

@@ -1,36 +1,62 @@
 #!/bin/sh
-#
 # install - install a program, script, or datafile
-# This comes from X11R5 (mit/util/scripts/install.sh).
+
+scriptversion=2006-10-14.15
+
+# This originates from X11R5 (mit/util/scripts/install.sh), which was
+# later released in X11R6 (xc/config/util/install.sh) with the
+# following copyright and license.
 #
-# Copyright 1991 by the Massachusetts Institute of Technology
+# Copyright (C) 1994 X Consortium
 #
-# Permission to use, copy, modify, distribute, and sell this software and its
-# documentation for any purpose is hereby granted without fee, provided that
-# the above copyright notice appear in all copies and that both that
-# copyright notice and this permission notice appear in supporting
-# documentation, and that the name of M.I.T. not be used in advertising or
-# publicity pertaining to distribution of the software without specific,
-# written prior permission.  M.I.T. makes no representations about the
-# suitability of this software for any purpose.  It is provided "as is"
-# without express or implied warranty.
+# Permission is hereby granted, free of charge, to any person obtaining a copy
+# of this software and associated documentation files (the "Software"), to
+# deal in the Software without restriction, including without limitation the
+# rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+# sell copies of the Software, and to permit persons to whom the Software is
+# furnished to do so, subject to the following conditions:
+#
+# The above copyright notice and this permission notice shall be included in
+# all copies or substantial portions of the Software.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE
+# X CONSORTIUM BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+# AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNEC-
+# TION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+#
+# Except as contained in this notice, the name of the X Consortium shall not
+# be used in advertising or otherwise to promote the sale, use or other deal-
+# ings in this Software without prior written authorization from the X Consor-
+# tium.
+#
+#
+# FSF changes to this file are in the public domain.
 #
 # Calling this script install-sh is preferred over install.sh, to prevent
 # `make' implicit rules from creating a file called install from it
 # when there is no Makefile.
 #
 # This script is compatible with the BSD install script, but was written
-# from scratch.  It can only install one file at a time, a restriction
-# shared with many OS's install programs.
+# from scratch.
 
+nl='
+'
+IFS=" ""	$nl"
 
 # set DOITPROG to echo to test this script
 
 # Don't use :- since 4.3BSD and earlier shells don't like it.
 doit="${DOITPROG-}"
+if test -z "$doit"; then
+  doit_exec=exec
+else
+  doit_exec=$doit
+fi
 
-
-# put in absolute paths if you don't have them in your path; or use env. vars.
+# Put in absolute file names if you don't have them in your path;
+# or use environment vars.
 
 mvprog="${MVPROG-mv}"
 cpprog="${CPPROG-cp}"
@@ -41,211 +67,441 @@ stripprog="${STRIPPROG-strip}"
 rmprog="${RMPROG-rm}"
 mkdirprog="${MKDIRPROG-mkdir}"
 
-transformbasename=""
-transform_arg=""
-instcmd="$mvprog"
-chmodcmd="$chmodprog 0755"
-chowncmd=""
-chgrpcmd=""
-stripcmd=""
+posix_glob=
+posix_mkdir=
+
+# Desired mode of installed file.
+mode=0755
+
+chmodcmd=$chmodprog
+chowncmd=
+chgrpcmd=
+stripcmd=
 rmcmd="$rmprog -f"
 mvcmd="$mvprog"
-src=""
-dst=""
-dir_arg=""
+src=
+dst=
+dir_arg=
+dstarg=
+no_target_directory=
 
-while [ x"$1" != x ]; do
-    case $1 in
-	-c) instcmd="$cpprog"
-	    shift
-	    continue;;
+usage="Usage: $0 [OPTION]... [-T] SRCFILE DSTFILE
+   or: $0 [OPTION]... SRCFILES... DIRECTORY
+   or: $0 [OPTION]... -t DIRECTORY SRCFILES...
+   or: $0 [OPTION]... -d DIRECTORIES...
 
-	-d) dir_arg=true
-	    shift
-	    continue;;
+In the 1st form, copy SRCFILE to DSTFILE.
+In the 2nd and 3rd, copy all SRCFILES to DIRECTORY.
+In the 4th, create DIRECTORIES.
 
-	-m) chmodcmd="$chmodprog $2"
-	    shift
-	    shift
-	    continue;;
+Options:
+-c         (ignored)
+-d         create directories instead of installing files.
+-g GROUP   $chgrpprog installed files to GROUP.
+-m MODE    $chmodprog installed files to MODE.
+-o USER    $chownprog installed files to USER.
+-s         $stripprog installed files.
+-t DIRECTORY  install into DIRECTORY.
+-T         report an error if DSTFILE is a directory.
+--help     display this help and exit.
+--version  display version info and exit.
 
-	-o) chowncmd="$chownprog $2"
-	    shift
-	    shift
-	    continue;;
+Environment variables override the default commands:
+  CHGRPPROG CHMODPROG CHOWNPROG CPPROG MKDIRPROG MVPROG RMPROG STRIPPROG
+"
 
-	-g) chgrpcmd="$chgrpprog $2"
-	    shift
-	    shift
-	    continue;;
+while test $# -ne 0; do
+  case $1 in
+    -c) shift
+        continue;;
 
-	-s) stripcmd="$stripprog"
-	    shift
-	    continue;;
+    -d) dir_arg=true
+        shift
+        continue;;
 
-	-t=*) transformarg=`echo $1 | sed 's/-t=//'`
-	    shift
-	    continue;;
+    -g) chgrpcmd="$chgrpprog $2"
+        shift
+        shift
+        continue;;
 
-	-b=*) transformbasename=`echo $1 | sed 's/-b=//'`
-	    shift
-	    continue;;
+    --help) echo "$usage"; exit $?;;
 
-	*)  if [ x"$src" = x ]
-	    then
-		src=$1
-	    else
-		# this colon is to work around a 386BSD /bin/sh bug
-		:
-		dst=$1
-	    fi
-	    shift
-	    continue;;
-    esac
-done
+    -m) mode=$2
+        shift
+        shift
+	case $mode in
+	  *' '* | *'	'* | *'
+'*	  | *'*'* | *'?'* | *'['*)
+	    echo "$0: invalid mode: $mode" >&2
+	    exit 1;;
+	esac
+        continue;;
 
-if [ x"$src" = x ]
-then
-	echo "install:	no input file specified"
-	exit 1
-else
-	true
-fi
+    -o) chowncmd="$chownprog $2"
+        shift
+        shift
+        continue;;
 
-if [ x"$dir_arg" != x ]; then
-	dst=$src
-	src=""
-	
-	if [ -d $dst ]; then
-		instcmd=:
-		chmodcmd=""
-	else
-		instcmd=mkdir
-	fi
-else
+    -s) stripcmd=$stripprog
+        shift
+        continue;;
 
-# Waiting for this to be detected by the "$instcmd $src $dsttmp" command
-# might cause directories to be created, which would be especially bad 
-# if $src (and thus $dsttmp) contains '*'.
-
-	if [ -f $src -o -d $src ]
-	then
-		true
-	else
-		echo "install:  $src does not exist"
-		exit 1
-	fi
-	
-	if [ x"$dst" = x ]
-	then
-		echo "install:	no destination specified"
-		exit 1
-	else
-		true
-	fi
-
-# If destination is a directory, append the input filename; if your system
-# does not like double slashes in filenames, you may need to add some logic
-
-	if [ -d $dst ]
-	then
-		dst="$dst"/`basename $src`
-	else
-		true
-	fi
-fi
-
-## this sed command emulates the dirname command
-dstdir=`echo $dst | sed -e 's,[^/]*$,,;s,/$,,;s,^$,.,'`
-
-# Make sure that the destination directory exists.
-#  this part is taken from Noah Friedman's mkinstalldirs script
-
-# Skip lots of stat calls in the usual case.
-if [ ! -d "$dstdir" ]; then
-defaultIFS='	
-'
-IFS="${IFS-${defaultIFS}}"
-
-oIFS="${IFS}"
-# Some sh's can't handle IFS=/ for some reason.
-IFS='%'
-set - `echo ${dstdir} | sed -e 's@/@%@g' -e 's@^%@/@'`
-IFS="${oIFS}"
-
-pathcomp=''
-
-while [ $# -ne 0 ] ; do
-	pathcomp="${pathcomp}${1}"
+    -t) dstarg=$2
 	shift
+	shift
+	continue;;
 
-	if [ ! -d "${pathcomp}" ] ;
-        then
-		$mkdirprog "${pathcomp}"
-	else
-		true
-	fi
+    -T) no_target_directory=true
+	shift
+	continue;;
 
-	pathcomp="${pathcomp}/"
+    --version) echo "$0 $scriptversion"; exit $?;;
+
+    --)	shift
+	break;;
+
+    -*)	echo "$0: invalid option: $1" >&2
+	exit 1;;
+
+    *)  break;;
+  esac
 done
+
+if test $# -ne 0 && test -z "$dir_arg$dstarg"; then
+  # When -d is used, all remaining arguments are directories to create.
+  # When -t is used, the destination is already specified.
+  # Otherwise, the last argument is the destination.  Remove it from $@.
+  for arg
+  do
+    if test -n "$dstarg"; then
+      # $@ is not empty: it contains at least $arg.
+      set fnord "$@" "$dstarg"
+      shift # fnord
+    fi
+    shift # arg
+    dstarg=$arg
+  done
 fi
 
-if [ x"$dir_arg" != x ]
-then
-	$doit $instcmd $dst &&
+if test $# -eq 0; then
+  if test -z "$dir_arg"; then
+    echo "$0: no input file specified." >&2
+    exit 1
+  fi
+  # It's OK to call `install-sh -d' without argument.
+  # This can happen when creating conditional directories.
+  exit 0
+fi
 
-	if [ x"$chowncmd" != x ]; then $doit $chowncmd $dst; else true ; fi &&
-	if [ x"$chgrpcmd" != x ]; then $doit $chgrpcmd $dst; else true ; fi &&
-	if [ x"$stripcmd" != x ]; then $doit $stripcmd $dst; else true ; fi &&
-	if [ x"$chmodcmd" != x ]; then $doit $chmodcmd $dst; else true ; fi
-else
+if test -z "$dir_arg"; then
+  trap '(exit $?); exit' 1 2 13 15
 
-# If we're going to rename the final executable, determine the name now.
+  # Set umask so as not to create temps with too-generous modes.
+  # However, 'strip' requires both read and write access to temps.
+  case $mode in
+    # Optimize common cases.
+    *644) cp_umask=133;;
+    *755) cp_umask=22;;
 
-	if [ x"$transformarg" = x ] 
-	then
-		dstfile=`basename $dst`
+    *[0-7])
+      if test -z "$stripcmd"; then
+	u_plus_rw=
+      else
+	u_plus_rw='% 200'
+      fi
+      cp_umask=`expr '(' 777 - $mode % 1000 ')' $u_plus_rw`;;
+    *)
+      if test -z "$stripcmd"; then
+	u_plus_rw=
+      else
+	u_plus_rw=,u+rw
+      fi
+      cp_umask=$mode$u_plus_rw;;
+  esac
+fi
+
+for src
+do
+  # Protect names starting with `-'.
+  case $src in
+    -*) src=./$src ;;
+  esac
+
+  if test -n "$dir_arg"; then
+    dst=$src
+    dstdir=$dst
+    test -d "$dstdir"
+    dstdir_status=$?
+  else
+
+    # Waiting for this to be detected by the "$cpprog $src $dsttmp" command
+    # might cause directories to be created, which would be especially bad
+    # if $src (and thus $dsttmp) contains '*'.
+    if test ! -f "$src" && test ! -d "$src"; then
+      echo "$0: $src does not exist." >&2
+      exit 1
+    fi
+
+    if test -z "$dstarg"; then
+      echo "$0: no destination specified." >&2
+      exit 1
+    fi
+
+    dst=$dstarg
+    # Protect names starting with `-'.
+    case $dst in
+      -*) dst=./$dst ;;
+    esac
+
+    # If destination is a directory, append the input filename; won't work
+    # if double slashes aren't ignored.
+    if test -d "$dst"; then
+      if test -n "$no_target_directory"; then
+	echo "$0: $dstarg: Is a directory" >&2
+	exit 1
+      fi
+      dstdir=$dst
+      dst=$dstdir/`basename "$src"`
+      dstdir_status=0
+    else
+      # Prefer dirname, but fall back on a substitute if dirname fails.
+      dstdir=`
+	(dirname "$dst") 2>/dev/null ||
+	expr X"$dst" : 'X\(.*[^/]\)//*[^/][^/]*/*$' \| \
+	     X"$dst" : 'X\(//\)[^/]' \| \
+	     X"$dst" : 'X\(//\)$' \| \
+	     X"$dst" : 'X\(/\)' \| . 2>/dev/null ||
+	echo X"$dst" |
+	    sed '/^X\(.*[^/]\)\/\/*[^/][^/]*\/*$/{
+		   s//\1/
+		   q
+		 }
+		 /^X\(\/\/\)[^/].*/{
+		   s//\1/
+		   q
+		 }
+		 /^X\(\/\/\)$/{
+		   s//\1/
+		   q
+		 }
+		 /^X\(\/\).*/{
+		   s//\1/
+		   q
+		 }
+		 s/.*/./; q'
+      `
+
+      test -d "$dstdir"
+      dstdir_status=$?
+    fi
+  fi
+
+  obsolete_mkdir_used=false
+
+  if test $dstdir_status != 0; then
+    case $posix_mkdir in
+      '')
+	# Create intermediate dirs using mode 755 as modified by the umask.
+	# This is like FreeBSD 'install' as of 1997-10-28.
+	umask=`umask`
+	case $stripcmd.$umask in
+	  # Optimize common cases.
+	  *[2367][2367]) mkdir_umask=$umask;;
+	  .*0[02][02] | .[02][02] | .[02]) mkdir_umask=22;;
+
+	  *[0-7])
+	    mkdir_umask=`expr $umask + 22 \
+	      - $umask % 100 % 40 + $umask % 20 \
+	      - $umask % 10 % 4 + $umask % 2
+	    `;;
+	  *) mkdir_umask=$umask,go-w;;
+	esac
+
+	# With -d, create the new directory with the user-specified mode.
+	# Otherwise, rely on $mkdir_umask.
+	if test -n "$dir_arg"; then
+	  mkdir_mode=-m$mode
 	else
-		dstfile=`basename $dst $transformbasename | 
-			sed $transformarg`$transformbasename
+	  mkdir_mode=
 	fi
 
-# don't allow the sed command to completely eliminate the filename
+	posix_mkdir=false
+	case $umask in
+	  *[123567][0-7][0-7])
+	    # POSIX mkdir -p sets u+wx bits regardless of umask, which
+	    # is incompatible with FreeBSD 'install' when (umask & 300) != 0.
+	    ;;
+	  *)
+	    tmpdir=${TMPDIR-/tmp}/ins$RANDOM-$$
+	    trap 'ret=$?; rmdir "$tmpdir/d" "$tmpdir" 2>/dev/null; exit $ret' 0
 
-	if [ x"$dstfile" = x ] 
-	then
-		dstfile=`basename $dst`
+	    if (umask $mkdir_umask &&
+		exec $mkdirprog $mkdir_mode -p -- "$tmpdir/d") >/dev/null 2>&1
+	    then
+	      if test -z "$dir_arg" || {
+		   # Check for POSIX incompatibilities with -m.
+		   # HP-UX 11.23 and IRIX 6.5 mkdir -m -p sets group- or
+		   # other-writeable bit of parent directory when it shouldn't.
+		   # FreeBSD 6.1 mkdir -m -p sets mode of existing directory.
+		   ls_ld_tmpdir=`ls -ld "$tmpdir"`
+		   case $ls_ld_tmpdir in
+		     d????-?r-*) different_mode=700;;
+		     d????-?--*) different_mode=755;;
+		     *) false;;
+		   esac &&
+		   $mkdirprog -m$different_mode -p -- "$tmpdir" && {
+		     ls_ld_tmpdir_1=`ls -ld "$tmpdir"`
+		     test "$ls_ld_tmpdir" = "$ls_ld_tmpdir_1"
+		   }
+		 }
+	      then posix_mkdir=:
+	      fi
+	      rmdir "$tmpdir/d" "$tmpdir"
+	    else
+	      # Remove any dirs left behind by ancient mkdir implementations.
+	      rmdir ./$mkdir_mode ./-p ./-- 2>/dev/null
+	    fi
+	    trap '' 0;;
+	esac;;
+    esac
+
+    if
+      $posix_mkdir && (
+	umask $mkdir_umask &&
+	$doit_exec $mkdirprog $mkdir_mode -p -- "$dstdir"
+      )
+    then :
+    else
+
+      # The umask is ridiculous, or mkdir does not conform to POSIX,
+      # or it failed possibly due to a race condition.  Create the
+      # directory the slow way, step by step, checking for races as we go.
+
+      case $dstdir in
+	/*) prefix=/ ;;
+	-*) prefix=./ ;;
+	*)  prefix= ;;
+      esac
+
+      case $posix_glob in
+        '')
+	  if (set -f) 2>/dev/null; then
+	    posix_glob=true
+	  else
+	    posix_glob=false
+	  fi ;;
+      esac
+
+      oIFS=$IFS
+      IFS=/
+      $posix_glob && set -f
+      set fnord $dstdir
+      shift
+      $posix_glob && set +f
+      IFS=$oIFS
+
+      prefixes=
+
+      for d
+      do
+	test -z "$d" && continue
+
+	prefix=$prefix$d
+	if test -d "$prefix"; then
+	  prefixes=
 	else
-		true
+	  if $posix_mkdir; then
+	    (umask=$mkdir_umask &&
+	     $doit_exec $mkdirprog $mkdir_mode -p -- "$dstdir") && break
+	    # Don't fail if two instances are running concurrently.
+	    test -d "$prefix" || exit 1
+	  else
+	    case $prefix in
+	      *\'*) qprefix=`echo "$prefix" | sed "s/'/'\\\\\\\\''/g"`;;
+	      *) qprefix=$prefix;;
+	    esac
+	    prefixes="$prefixes '$qprefix'"
+	  fi
 	fi
+	prefix=$prefix/
+      done
 
-# Make a temp file name in the proper directory.
+      if test -n "$prefixes"; then
+	# Don't fail if two instances are running concurrently.
+	(umask $mkdir_umask &&
+	 eval "\$doit_exec \$mkdirprog $prefixes") ||
+	  test -d "$dstdir" || exit 1
+	obsolete_mkdir_used=true
+      fi
+    fi
+  fi
 
-	dsttmp=$dstdir/#inst.$$#
+  if test -n "$dir_arg"; then
+    { test -z "$chowncmd" || $doit $chowncmd "$dst"; } &&
+    { test -z "$chgrpcmd" || $doit $chgrpcmd "$dst"; } &&
+    { test "$obsolete_mkdir_used$chowncmd$chgrpcmd" = false ||
+      test -z "$chmodcmd" || $doit $chmodcmd $mode "$dst"; } || exit 1
+  else
 
-# Move or copy the file name to the temp name
+    # Make a couple of temp file names in the proper directory.
+    dsttmp=$dstdir/_inst.$$_
+    rmtmp=$dstdir/_rm.$$_
 
-	$doit $instcmd $src $dsttmp &&
+    # Trap to clean up those temp files at exit.
+    trap 'ret=$?; rm -f "$dsttmp" "$rmtmp" && exit $ret' 0
 
-	trap "rm -f ${dsttmp}" 0 &&
+    # Copy the file name to the temp name.
+    (umask $cp_umask && $doit_exec $cpprog "$src" "$dsttmp") &&
 
-# and set any options; do chmod last to preserve setuid bits
+    # and set any options; do chmod last to preserve setuid bits.
+    #
+    # If any of these fail, we abort the whole thing.  If we want to
+    # ignore errors from any of these, just make sure not to ignore
+    # errors from the above "$doit $cpprog $src $dsttmp" command.
+    #
+    { test -z "$chowncmd" || $doit $chowncmd "$dsttmp"; } \
+      && { test -z "$chgrpcmd" || $doit $chgrpcmd "$dsttmp"; } \
+      && { test -z "$stripcmd" || $doit $stripcmd "$dsttmp"; } \
+      && { test -z "$chmodcmd" || $doit $chmodcmd $mode "$dsttmp"; } &&
 
-# If any of these fail, we abort the whole thing.  If we want to
-# ignore errors from any of these, just make sure not to ignore
-# errors from the above "$doit $instcmd $src $dsttmp" command.
+    # Now rename the file to the real destination.
+    { $doit $mvcmd -f "$dsttmp" "$dst" 2>/dev/null \
+      || {
+	   # The rename failed, perhaps because mv can't rename something else
+	   # to itself, or perhaps because mv is so ancient that it does not
+	   # support -f.
 
-	if [ x"$chowncmd" != x ]; then $doit $chowncmd $dsttmp; else true;fi &&
-	if [ x"$chgrpcmd" != x ]; then $doit $chgrpcmd $dsttmp; else true;fi &&
-	if [ x"$stripcmd" != x ]; then $doit $stripcmd $dsttmp; else true;fi &&
-	if [ x"$chmodcmd" != x ]; then $doit $chmodcmd $dsttmp; else true;fi &&
+	   # Now remove or move aside any old file at destination location.
+	   # We try this two ways since rm can't unlink itself on some
+	   # systems and the destination file might be busy for other
+	   # reasons.  In this case, the final cleanup might fail but the new
+	   # file should still install successfully.
+	   {
+	     if test -f "$dst"; then
+	       $doit $rmcmd -f "$dst" 2>/dev/null \
+	       || { $doit $mvcmd -f "$dst" "$rmtmp" 2>/dev/null \
+		     && { $doit $rmcmd -f "$rmtmp" 2>/dev/null; :; }; }\
+	       || {
+		 echo "$0: cannot unlink or rename $dst" >&2
+		 (exit 1); exit 1
+	       }
+	     else
+	       :
+	     fi
+	   } &&
 
-# Now rename the file to the real destination.
+	   # Now rename the file to the real destination.
+	   $doit $mvcmd "$dsttmp" "$dst"
+	 }
+    } || exit 1
 
-	$doit $rmcmd -f $dstdir/$dstfile &&
-	$doit $mvcmd $dsttmp $dstdir/$dstfile 
+    trap '' 0
+  fi
+done
 
-fi &&
-
-
-exit 0
+# Local variables:
+# eval: (add-hook 'write-file-hooks 'time-stamp)
+# time-stamp-start: "scriptversion="
+# time-stamp-format: "%:y-%02m-%02d.%02H"
+# time-stamp-end: "$"
+# End:

configure.in

@@ -1,88 +1,405 @@
-################################################################################
 ##
-##    Copyright 1999-2000 Sistina Software, Inc.
+## Copyright (C) 2000-2004 Sistina Software, Inc. All rights reserved.
+## Copyright (C) 2004, 2007 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 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_INIT(lib/device/dev-cache.h)
+AC_PREREQ(2.57)
+################################################################################
+dnl -- Process this file with autoconf to produce a configure script.
+AC_INIT
+AC_CONFIG_SRCDIR([lib/device/dev-cache.h])
 
-dnl setup the directory where autoconf has auxilary files
-AC_CONFIG_AUX_DIR(autoconf) 
+################################################################################
+AC_CONFIG_HEADERS(lib/misc/configure.h)
 
-dnl Checks for programs.
+################################################################################
+dnl -- Setup the directory where autoconf has auxilary files
+AC_CONFIG_AUX_DIR(autoconf)
+
+################################################################################
+dnl -- Get system type
+AC_CANONICAL_TARGET([])
+
+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="$LDFLAGS -Wl,--export-dynamic"
+		LIB_SUFFIX="so"
+		DEVMAPPER=yes
+		ODIRECT=yes
+		SELINUX=yes
+		REALTIME=yes
+		CLUSTER=internal
+		FSADM=no ;;
+	darwin*)
+		CFLAGS="$CFLAGS -no-cpp-precomp -fno-common"
+		COPTIMISE_FLAG="-O2"
+		CLDFLAGS="$CLDFLAGS"
+		CLDWHOLEARCHIVE="-all_load"
+		CLDNOWHOLEARCHIVE=
+		LDDEPS="$LDDEPS"
+		LDFLAGS="$LDFLAGS"
+		LIB_SUFFIX="dylib"
+		DEVMAPPER=yes
+		ODIRECT=no
+		SELINUX=no
+		REALTIME=no
+		CLUSTER=none
+		FSADM=no ;;
+esac
+
+################################################################################
+dnl -- Checks for programs.
 AC_PROG_AWK
 AC_PROG_CC
 AC_PROG_INSTALL
 AC_PROG_LN_S
 AC_PROG_MAKE_SET
 AC_PROG_RANLIB
+AC_PATH_PROG(CFLOW_CMD, cflow)
+AC_PATH_PROG(CSCOPE_CMD, cscope)
 
-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_STRUCT_ST_RDEV
-AC_HEADER_TIME
+AC_TYPE_MODE_T
+AC_CHECK_MEMBERS([struct stat.st_rdev])
+AC_STRUCT_TM
 
-dnl -- prefix is /usr by default, the exec_prefix default is setup later
+################################################################################
+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
+
+################################################################################
+dnl -- Prefix is /usr by default, the exec_prefix default is setup later
 AC_PREFIX_DEFAULT(/usr)
 
-dnl -- setup the ownership of the files
+################################################################################
+dnl -- Parallel make jobs?
+AC_ARG_ENABLE(jobs, [  --enable-jobs=NUM       Number of jobs to run simultaneously], JOBS=-j$enableval, JOBS=-j2)
+
+################################################################################
+dnl -- Setup the ownership of the files
+AC_MSG_CHECKING(file owner)
+OWNER="root"
+
 AC_ARG_WITH(user,
   [  --with-user=USER        Set the owner of installed files ],
-  [ OWNER="$withval" ],
-  [ OWNER="root" ])
+  [ OWNER="$withval" ])
+AC_MSG_RESULT($OWNER)
 
-dnl -- setup the group ownership of the files
+if test x$OWNER != x; then
+	OWNER="-o $OWNER"
+fi
+
+################################################################################
+dnl -- Setup the group ownership of the files
+AC_MSG_CHECKING(group owner)
+GROUP="root"
 AC_ARG_WITH(group,
   [  --with-group=GROUP      Set the group owner of installed files ],
-  [ GROUP="$withval" ],
-  [ GROUP="root" ])
+  [ GROUP="$withval" ])
+AC_MSG_RESULT($GROUP)
 
-AC_ARG_ENABLE(jobs, [  --enable-jobs=NUM       Number of jobs to run simultaneously], JOBS=-j$enableval, JOBS=)
+if test x$GROUP != x; then
+	GROUP="-g $GROUP"
+fi
 
-dnl Enables staticly linked tools
-AC_ARG_ENABLE(static_link, [  --enable-static_link    Use this to link the tools to the liblvm library
+################################################################################
+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$LVM1_FALLBACK = xyes; then
+	AC_DEFINE([LVM1_FALLBACK], 1, [Define to 1 if 'lvm' should fall back to using LVM1 binaries if device-mapper is missing from the kernel])
+fi
+
+################################################################################
+dnl -- format1 inclusion type
+AC_MSG_CHECKING(whether to include support for lvm1 metadata)
+AC_ARG_WITH(lvm1,
+  [  --with-lvm1=TYPE        LVM1 metadata support: internal/shared/none
+                          [TYPE=internal] ],
+  [ LVM1="$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
+)
+fi;
+
+if test x$LVM1 = xinternal; then
+	AC_DEFINE([LVM1_INTERNAL], 1, [Define to 1 to include built-in support for LVM1 metadata.])
+fi
+
+################################################################################
+dnl -- format_pool inclusion type
+AC_MSG_CHECKING(whether to include support for GFS pool metadata)
+AC_ARG_WITH(pool,
+  [  --with-pool=TYPE        GFS pool read-only support: internal/shared/none
+                          [TYPE=internal] ],
+  [ POOL="$withval" ],
+  [ POOL="internal" ])
+AC_MSG_RESULT($POOL)
+
+if [[ "x$POOL" != xnone -a "x$POOL" != xinternal -a "x$POOL" != xshared ]];
+ then  AC_MSG_ERROR(
+--with-pool parameter invalid
+)
+fi;
+
+if test x$POOL = xinternal; then
+	AC_DEFINE([POOL_INTERNAL], 1, [Define to 1 to include built-in support for GFS pool metadata.])
+fi
+
+################################################################################
+dnl -- cluster_locking inclusion type
+AC_MSG_CHECKING(whether to include support for cluster locking)
+AC_ARG_WITH(cluster,
+  [  --with-cluster=TYPE     Cluster LVM locking support: internal/shared/none
+                          [TYPE=internal] ],
+  [ CLUSTER="$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$CLUSTER = xinternal; then
+	AC_DEFINE([CLUSTER_LOCKING_INTERNAL], 1, [Define to 1 to include built-in support for clustered LVM locking.])
+fi
+
+################################################################################
+dnl -- snapshots inclusion type
+AC_MSG_CHECKING(whether to include snapshots)
+AC_ARG_WITH(snapshots,
+  [  --with-snapshots=TYPE   Snapshot support: internal/shared/none
+                          [TYPE=internal] ],
+  [ SNAPSHOTS="$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$SNAPSHOTS = xinternal; then
+	AC_DEFINE([SNAPSHOT_INTERNAL], 1, [Define to 1 to include built-in support for snapshots.])
+fi
+
+################################################################################
+dnl -- mirrors inclusion type
+AC_MSG_CHECKING(whether to include mirrors)
+AC_ARG_WITH(mirrors,
+  [  --with-mirrors=TYPE     Mirror support: internal/shared/none
+                          [TYPE=internal] ],
+  [ MIRRORS="$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$MIRRORS = xinternal; then
+	AC_DEFINE([MIRRORED_INTERNAL], 1, [Define to 1 to include built-in support for mirrors.])
+fi
+
+################################################################################
+dnl -- Enables staticly-linked tools
+AC_MSG_CHECKING(whether to use static linking)
+AC_ARG_ENABLE(static_link, [  --enable-static_link    Use this to link the tools to their libraries
                           statically.  Default is dynamic linking],  STATIC_LINK=$enableval, STATIC_LINK=no)
+AC_MSG_RESULT($STATIC_LINK)
 
-dnl Disable readline
-AC_ARG_ENABLE(readline, [  --disable-readline      Disable readline support],  \
-READLINE=$enableval, READLINE=yes)
+################################################################################
+dnl -- Enable readline
+AC_MSG_CHECKING(whether to enable readline)
+AC_ARG_ENABLE(readline, [  --enable-readline       Enable readline support],
+READLINE=$enableval, READLINE=no)
+AC_MSG_RESULT($READLINE)
 
-dnl Mess with default exec_prefix
+if test x$READLINE = xyes; then
+	AC_DEFINE([READLINE_SUPPORT], 1, [Define to 1 to include the LVM readline shell.])
+fi
+
+################################################################################
+dnl -- Disable selinux
+AC_MSG_CHECKING(whether to enable selinux support)
+AC_ARG_ENABLE(selinux, [  --disable-selinux       Disable selinux support],
+SELINUX=$enableval)
+AC_MSG_RESULT($SELINUX)
+
+################################################################################
+dnl -- Disable realtime clock support
+AC_MSG_CHECKING(whether to enable realtime support)
+AC_ARG_ENABLE(realtime, [  --disable-realtime      Disable realtime clock support],
+REALTIME=$enableval)
+AC_MSG_RESULT($REALTIME)
+
+################################################################################
+dnl -- Build cluster LVM daemon
+AC_MSG_CHECKING(whether to build cluster LVM daemon)
+AC_ARG_WITH(clvmd,
+  [  --with-clvmd=TYPE       Build cluster LVM Daemon: cman/gulm/none/all
+                          [TYPE=none] ],
+  [ CLVMD="$withval" ],
+  [ CLVMD="none" ])
+if test x$CLVMD = xyes; then
+	CLVMD=all
+fi
+AC_MSG_RESULT($CLVMD)
+
+dnl -- If clvmd enabled without cluster locking, automagically include it
+if  test x$CLVMD != xnone && test x$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$DEBUG = xyes; then
+	COPTIMISE_FLAG=
+else
+	CSCOPE_CMD=
+fi
+
+################################################################################
+dnl -- Override optimisation
+AC_MSG_CHECKING(for C optimisation flag)
+AC_ARG_WITH(optimisation,
+  [  --with-optimisation=OPT C optimisation flag [OPT=-O2] ],
+  [ COPTIMISE_FLAG="$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
+	AC_DEFINE([DEVMAPPER_SUPPORT], 1, [Define to 1 to enable device-mapper interaction.])
+fi
+
+################################################################################
+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$ODIRECT = xyes; then
+	AC_DEFINE([O_DIRECT_SUPPORT], 1, [Define to 1 to enable O_DIRECT support.])
+fi
+
+################################################################################
+dnl -- Enable cmdlib
+AC_MSG_CHECKING(whether to compile liblvm2cmd.so)
+AC_ARG_ENABLE(cmdlib, [  --enable-cmdlib         Build shared command library],
+CMDLIB=$enableval, CMDLIB=no)
+AC_MSG_RESULT($CMDLIB)
+
+################################################################################
+dnl -- Enable fsadm
+AC_MSG_CHECKING(whether to build fsadm)
+AC_ARG_ENABLE(fsadm, [  --enable-fsadm          Enable fsadm],
+FSADM=$enableval)
+AC_MSG_RESULT($FSADM)
+
+################################################################################
+dnl -- enable dmeventd handling
+AC_MSG_CHECKING(whether to use dmeventd)
+AC_ARG_ENABLE(dmeventd, [  --enable-dmeventd       Enable the device-mapper event daemon],
+DMEVENTD=$enableval)
+AC_MSG_RESULT($DMEVENTD)
+
+dnl -- dmeventd currently requires internal mirror support
+if test x$DMEVENTD = xyes && test x$MIRRORS != xinternal; then
+AC_MSG_ERROR(
+--enable-dmeventd currently requires --with-mirrors=internal
+)
+fi
+
+if test x$DMEVENTD = xyes; then
+	AC_DEFINE([DMEVENTD], 1, [Define to 1 to enable the device-mapper event daemon.])
+fi
+################################################################################
+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(
@@ -95,11 +412,80 @@ 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 readline (Shamelessly copied from parted 1.4.17)
+################################################################################
+dnl -- Check for dlopen
+AC_CHECK_LIB(dl, dlopen, HAVE_LIBDL=yes, HAVE_LIBDL=no)
+
+if [[ "x$HAVE_LIBDL" = xyes ]]; then
+	AC_DEFINE([HAVE_LIBDL], 1, [Define to 1 if dynamic libraries are available.])
+	LIBS="-ldl $LIBS"
+else
+	HAVE_LIBDL=no
+fi
+
+################################################################################
+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$STATIC_LINK" = xyes ]];
+ then  AC_MSG_ERROR(
+Features cannot be 'shared' when building statically
+)
+fi
+
+################################################################################
+dnl -- Check for selinux
+if test x$SELINUX = xyes; then
+	AC_CHECK_LIB(sepol, sepol_check_context, HAVE_SEPOL=yes, HAVE_SEPOL=no)
+
+	if test x$HAVE_SEPOL = xyes; then
+		LIBS="-lsepol $LIBS"
+	fi
+
+	AC_CHECK_LIB(selinux, is_selinux_enabled, HAVE_SELINUX=yes, HAVE_SELINUX=no)
+
+	if test x$HAVE_SELINUX = xyes; then
+		AC_DEFINE([HAVE_SELINUX], 1, [Define to 1 to include support for selinux.])
+		LIBS="-lselinux $LIBS"
+	else
+		AC_MSG_WARN(Disabling selinux)
+	fi
+
+	# With --enable-static_link and selinux enabled, linking lvm.static
+	# fails on at least Debian unstable due to unsatisfied references
+	# to pthread_mutex_lock and _unlock.  See if we need -lpthread.
+	if test "$STATIC_LINK-$HAVE_SELINUX" = yes-yes; then
+		lvm_saved_libs=$LIBS
+		LIBS="$LIBS -static"
+		AC_SEARCH_LIBS([pthread_mutex_lock], [pthread],
+		  [test "$ac_cv_search_pthread_mutex_lock" = "none required" ||
+				LIB_PTHREAD=-lpthread])
+		LIBS=$lvm_saved_libs
+	fi
+fi
+
+################################################################################
+dnl -- Check for realtime clock support
+if test x$REALTIME = xyes; then
+	AC_CHECK_LIB(rt, clock_gettime, HAVE_REALTIME=yes, HAVE_REALTIME=no)
+
+	if test x$HAVE_REALTIME = xyes; then
+		AC_DEFINE([HAVE_REALTIME], 1, [Define to 1 to include support for realtime clock.])
+		LIBS="-lrt $LIBS"
+	else
+		AC_MSG_WARN(Disabling realtime clock)
+	fi
+fi
+
+################################################################################
+dnl -- Check for getopt
+AC_CHECK_HEADERS(getopt.h, AC_DEFINE([HAVE_GETOPTLONG], 1, [Define to 1 to if getopt_long is available.]))
+
+################################################################################
+dnl -- Check for readline (Shamelessly copied from parted 1.4.17)
 if test x$READLINE = xyes; then
 	AC_CHECK_LIB(readline, readline, ,
 		AC_MSG_ERROR(
@@ -110,39 +496,183 @@ 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, HAVE_RL_COMPLETION_MATCHES=yes,
-		HAVE_RL_COMPLETION_MATCHES=no)
+	AC_CHECK_FUNC(rl_completion_matches, AC_DEFINE([HAVE_RL_COMPLETION_MATCHES], 1, [Define to 1 if rl_completion_matches() is available.]))
 fi
 
+################################################################################
+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$INTL = xyes; then
+	INTL_PACKAGE="lvm2"
+	AC_PATH_PROG(MSGFMT, msgfmt)
+	if [[ "x$MSGFMT" == x ]];
+		then  AC_MSG_ERROR(
+		msgfmt not found in path $PATH
+		)
+	fi;
+
+	AC_ARG_WITH(localedir,
+  		    [  --with-localedir=DIR    Translation files in DIR [PREFIX/share/locale]],
+  		    [ LOCALEDIR="$withval" ],
+  		    [ LOCALEDIR='${prefix}/share/locale' ])
+fi
+
+################################################################################
+AC_ARG_WITH(confdir,
+	    [  --with-confdir=DIR      Configuration files in DIR [/etc]],
+  	    [ CONFDIR="$withval" ],
+ 	    [ CONFDIR='/etc' ])
+
+AC_ARG_WITH(staticdir,
+	    [  --with-staticdir=DIR    Static binary in DIR [EXEC_PREFIX/sbin]],
+  	    [ STATICDIR="$withval" ],
+ 	    [ STATICDIR='${exec_prefix}/sbin' ])
+
+AC_ARG_WITH(dmdir,
+	    [  --with-dmdir=DIR        Build against device-mapper source tree in DIR],
+  	    [ DMDIR="$withval" CPPFLAGS="$CPPFLAGS -I$DMDIR/include"],
+	    [ DMDIR= ])
+
+# Convert a relative dir name to absolute.
+case $DMDIR in
+  /*) ;;
+  *) DMDIR="`pwd`/$DMDIR" ;;
+esac
+
+################################################################################
+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 != xnone; then
+	AC_CHECK_HEADERS(mntent.h netdb.h netinet/in.h pthread.h search.h sys/mount.h sys/socket.h sys/uio.h sys/un.h utmpx.h,,AC_MSG_ERROR(bailing out))
+	AC_CHECK_FUNCS(dup2 getmntent memmove select socket,,AC_MSG_ERROR(bailing out))
+	AC_FUNC_GETMNTENT
+#	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
+
+AC_CHECK_HEADERS(libdevmapper.h,,AC_MSG_ERROR(bailing out))
+
+if test x$HAVE_SELINUX = xyes; then
+	AC_CHECK_HEADERS(selinux/selinux.h,,AC_MSG_ERROR(bailing out))
+fi
+
+################################################################################
+AC_PATH_PROG(MODPROBE_CMD, modprobe)
+
+if test x$MODPROBE_CMD != x; then
+	AC_DEFINE_UNQUOTED([MODPROBE_CMD], ["$MODPROBE_CMD"], [The path to 'modprobe', if available.])
+fi
+
+################################################################################
 if test "-f VERSION"; then
   LVM_VERSION="\"`cat VERSION`\""
 else
   LVM_VERSION="Unknown"
 fi
 
+################################################################################
 AC_SUBST(JOBS)
 AC_SUBST(STATIC_LINK)
-AC_SUBST(READLINE)
-AC_SUBST(HAVE_RL_COMPLETION_MATCHES)
+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(LDDEPS)
+AC_SUBST(LDFLAGS)
+AC_SUBST(LIB_SUFFIX)
 AC_SUBST(LIBS)
 AC_SUBST(LVM_VERSION)
-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                                                               \
-include/Makefile						 	\
-lib/Makefile							 	\
-man/Makefile							 	\
-tools/Makefile							 	\
-tools/version.h								\
-test/mm/Makefile							\
-test/device/Makefile							\
-test/format1/Makefile							\
-test/regex/Makefile                                                     \
-test/filters/Makefile                                                   \
-)
+AC_SUBST(LVM1_FALLBACK)
+AC_SUBST(DEBUG)
+AC_SUBST(DEVMAPPER)
+AC_SUBST(HAVE_LIBDL)
+AC_SUBST(HAVE_SELINUX)
+AC_SUBST(HAVE_REALTIME)
+AC_SUBST(CMDLIB)
+AC_SUBST(MSGFMT)
+AC_SUBST(LOCALEDIR)
+AC_SUBST(CONFDIR)
+AC_SUBST(STATICDIR)
+AC_SUBST(DMDIR)
+AC_SUBST(INTL_PACKAGE)
+AC_SUBST(INTL)
+AC_SUBST(CLVMD)
+AC_SUBST(CLUSTER)
+AC_SUBST(FSADM)
+AC_SUBST(DMEVENTD)
+AC_SUBST(CFLOW_CMD)
+AC_SUBST(CSCOPE_CMD)
+AC_SUBST([LIB_PTHREAD])
+
+################################################################################
+dnl -- First and last lines should not contain files to generate in order to
+dnl -- keep utility scripts running properly
+AC_CONFIG_FILES([
+Makefile
+make.tmpl
+daemons/Makefile
+daemons/clvmd/Makefile
+dmeventd/Makefile
+dmeventd/mirror/Makefile
+doc/Makefile
+include/Makefile
+lib/Makefile
+lib/format1/Makefile
+lib/format_pool/Makefile
+lib/locking/Makefile
+lib/mirror/Makefile
+lib/snapshot/Makefile
+test/Makefile
+man/Makefile
+po/Makefile
+scripts/Makefile
+tools/Makefile
+tools/version.h
+tools/fsadm/Makefile
+])
+AC_OUTPUT
+
+if test x$ODIRECT != xyes; then
+  AC_MSG_WARN(Warning: O_DIRECT disabled: low-memory pvmove may lock up)
+fi
+
+if test x$FSADM == xyes; then
+  AC_MSG_WARN(fsadm support is untested)
+fi
+
+if test x$DMEVENTD == xyes; then
+  AC_MSG_WARN(dmeventd 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 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@
+
+ifneq ("@CLVMD@", "none")
+  SUBDIRS = clvmd
+endif
+
+include $(top_srcdir)/make.tmpl
+

daemons/clvmd/Makefile.in

@@ -0,0 +1,98 @@
+#
+# 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
+
+srcdir = @srcdir@
+top_srcdir = @top_srcdir@
+VPATH = @srcdir@
+
+SOURCES = \
+	clvmd-command.c  \
+	clvmd.c          \
+	lvm-functions.c  \
+	refresh_clvmd.c
+
+ifeq ("@CLVMD@", "gulm")
+	GULM = yes
+endif
+
+ifeq ("@CLVMD@", "cman")
+	CMAN = yes
+endif
+
+ifeq ("@CLVMD@", "openais")
+	OPENAIS = yes
+	GULM = no
+	CMAN = no
+endif
+
+ifeq ("@CLVMD@", "all")
+	GULM = yes
+	CMAN = yes
+	OPENAIS = no
+endif
+
+ifeq ("@DEBUG@", "yes")
+	DEFS += -DDEBUG
+endif
+
+ifeq ("$(GULM)", "yes")
+	SOURCES += clvmd-gulm.c tcp-comms.c
+	LMLIBS += -lccs -lgulm
+	DEFS += -DUSE_GULM
+endif
+
+ifeq ("$(CMAN)", "yes")
+	SOURCES += clvmd-cman.c
+	LMLIBS += -ldlm -lcman
+	DEFS += -DUSE_CMAN
+endif
+
+ifeq ("$(OPENAIS)", "yes")
+	SOURCES += clvmd-openais.c
+	LMLIBS += -lSaLck -lcpg
+	DEFS += -DUSE_OPENAIS
+endif
+
+TARGETS = \
+	clvmd
+
+LVMLIBS = -llvm -lpthread
+
+ifeq ("@DMEVENTD@", "yes")
+	LVMLIBS += -ldevmapper-event
+endif
+ 
+LVMLIBS += -ldevmapper
+
+DEFS += -D_REENTRANT
+CFLAGS += -fno-strict-aliasing
+
+include $(top_srcdir)/make.tmpl
+
+INSTALL_TARGETS = \
+	install_clvmd
+
+clvmd: $(OBJECTS) $(top_srcdir)/lib/liblvm.a
+	$(CC) -o clvmd $(OBJECTS) $(CFLAGS) $(LDFLAGS) \
+		$(LVMLIBS) $(LMLIBS) $(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 +1,70 @@
+/*
+ * Copyright (C) 2002-2004 Sistina Software, Inc. All rights reserved.
+ * Copyright (C) 2004-2007 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 */
+#define CLVMD_FLAG_NODEERRS     4       /* Reply has errors in node-specific portion */
+
+/* Name of the local socket to communicate between libclvm and clvmd */
+//static const char CLVMD_SOCKNAME[]="/var/run/clvmd";
+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
+
+/* Misc functions */
+#define CLVMD_CMD_REFRESH	    40
+#define CLVMD_CMD_GET_CLUSTERNAME   41
+#define CLVMD_CMD_SET_DEBUG	    42
+#endif

daemons/clvmd/clvmd-cman.c

@@ -0,0 +1,514 @@
+/*
+ * 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 <libdlm.h>
+
+#include "clvmd-comms.h"
+#include "clvm.h"
+#include "log.h"
+#include "clvmd.h"
+#include "lvm-functions.h"
+
+#define LOCKSPACE_NAME "clvmd"
+
+static int num_nodes;
+static struct cman_node *nodes = NULL;
+static struct cman_node this_node;
+static int count_nodes; /* size of allocated nodes array */
+static int max_updown_nodes = 50;	/* Current size of the allocated array */
+/* Node up/down status, indexed by nodeid */
+static int *node_updown = NULL;
+static dlm_lshandle_t *lockspace;
+static cman_handle_t c_handle;
+
+static void count_clvmds_running(void);
+static void get_members(void);
+static int nodeid_from_csid(const char *csid);
+static int name_from_nodeid(int nodeid, char *name);
+static void event_callback(cman_handle_t handle, void *private, int reason, int arg);
+static void data_callback(cman_handle_t handle, void *private,
+			  char *buf, int len, uint8_t port, int nodeid);
+
+struct lock_wait {
+	pthread_cond_t cond;
+	pthread_mutex_t mutex;
+	struct dlm_lksb lksb;
+};
+
+static int _init_cluster(void)
+{
+	/* Open the cluster communication socket */
+	c_handle = cman_init(NULL);
+	if (!c_handle) {
+		syslog(LOG_ERR, "Can't open cluster manager socket: %m");
+		return -1;
+	}
+	DEBUGLOG("Connected to CMAN\n");
+
+	if (cman_start_recv_data(c_handle, data_callback, CLUSTER_PORT_CLVMD)) {
+		syslog(LOG_ERR, "Can't bind cluster socket: %m");
+		return -1;
+	}
+
+	if (cman_start_notification(c_handle, event_callback)) {
+		syslog(LOG_ERR, "Can't start cluster event listening");
+		return -1;
+	}
+
+	/* Get the cluster members list */
+	get_members();
+	count_clvmds_running();
+
+	DEBUGLOG("CMAN initialisation complete\n");
+
+	/* Create a lockspace for LV & VG locks to live in */
+	lockspace = dlm_create_lockspace(LOCKSPACE_NAME, 0600);
+	if (!lockspace) {
+		syslog(LOG_ERR, "Unable to create lockspace for CLVM: %m");
+		return -1;
+	}
+	dlm_ls_pthread_init(lockspace);
+	DEBUGLOG("DLM initialisation complete\n");
+	return 0;
+}
+
+static void _cluster_init_completed(void)
+{
+	clvmd_cluster_init_completed();
+}
+
+static int _get_main_cluster_fd()
+{
+	return cman_get_fd(c_handle);
+}
+
+static int _get_num_nodes()
+{
+	int i;
+	int nnodes = 0;
+
+	/* return number of ACTIVE nodes */
+	for (i=0; i<num_nodes; i++) {
+		if (nodes[i].cn_member && nodes[i].cn_nodeid)
+			nnodes++;
+	}
+	return nnodes;
+}
+
+/* send_message with the fd check removed */
+static int _cluster_send_message(const void *buf, int msglen, const char *csid,
+				 const char *errtext)
+{
+	int nodeid = 0;
+
+	if (csid)
+		memcpy(&nodeid, csid, CMAN_MAX_CSID_LEN);
+
+	if (cman_send_data(c_handle, buf, msglen, 0, CLUSTER_PORT_CLVMD, nodeid) <= 0)
+	{
+		log_error("%s", errtext);
+	}
+	return msglen;
+}
+
+static void _get_our_csid(char *csid)
+{
+	if (this_node.cn_nodeid == 0) {
+		cman_get_node(c_handle, 0, &this_node);
+	}
+	memcpy(csid, &this_node.cn_nodeid, CMAN_MAX_CSID_LEN);
+}
+
+/* Call a callback routine for each node is that known (down means not running a clvmd) */
+static int _cluster_do_node_callback(struct local_client *client,
+				     void (*callback) (struct local_client *,
+						       const char *,
+						       int))
+{
+	int i;
+	int somedown = 0;
+
+	for (i = 0; i < _get_num_nodes(); i++) {
+		if (nodes[i].cn_member && nodes[i].cn_nodeid) {
+			callback(client, (char *)&nodes[i].cn_nodeid, node_updown[nodes[i].cn_nodeid]);
+			if (!node_updown[nodes[i].cn_nodeid])
+				somedown = -1;
+		}
+	}
+	return somedown;
+}
+
+/* Process OOB messages from the cluster socket */
+static void event_callback(cman_handle_t handle, void *private, int reason, int arg)
+{
+	char namebuf[MAX_CLUSTER_MEMBER_NAME_LEN];
+
+	switch (reason) {
+        case CMAN_REASON_PORTCLOSED:
+		name_from_nodeid(arg, namebuf);
+		log_notice("clvmd on node %s has died\n", namebuf);
+		DEBUGLOG("Got port closed message, removing node %s\n", namebuf);
+
+		node_updown[arg] = 0;
+		break;
+
+	case CMAN_REASON_STATECHANGE:
+		DEBUGLOG("Got state change message, re-reading members list\n");
+		get_members();
+		break;
+
+#if defined(LIBCMAN_VERSION) && LIBCMAN_VERSION >= 2
+	case CMAN_REASON_PORTOPENED:
+		/* Ignore this, wait for startup message from clvmd itself */
+		break;
+
+	case CMAN_REASON_TRY_SHUTDOWN:
+		DEBUGLOG("Got try shutdown, sending OK\n");
+		cman_replyto_shutdown(c_handle, 1);
+		break;
+#endif
+	default:
+		/* ERROR */
+		DEBUGLOG("Got unknown event callback message: %d\n", reason);
+		break;
+	}
+}
+
+static struct local_client *cman_client;
+static int _cluster_fd_callback(struct local_client *fd, char *buf, int len,
+				const char *csid,
+				struct local_client **new_client)
+{
+
+	/* Save this for data_callback */
+	cman_client = fd;
+
+	/* We never return a new client */
+	*new_client = NULL;
+
+	return cman_dispatch(c_handle, 0);
+}
+
+
+static void data_callback(cman_handle_t handle, void *private,
+			  char *buf, int len, uint8_t port, int nodeid)
+{
+	/* Ignore looped back messages */
+	if (nodeid == this_node.cn_nodeid)
+		return;
+	process_message(cman_client, buf, len, (char *)&nodeid);
+}
+
+static void _add_up_node(const char *csid)
+{
+	/* It's up ! */
+	int nodeid = nodeid_from_csid(csid);
+
+	if (nodeid >= max_updown_nodes) {
+	        int new_size = nodeid + 10;
+		int *new_updown = realloc(node_updown, new_size);
+
+		if (new_updown) {
+			node_updown = new_updown;
+			max_updown_nodes = new_size;
+			DEBUGLOG("realloced more space for nodes. now %d\n",
+				 max_updown_nodes);
+		} else {
+			log_error
+				("Realloc failed. Node status for clvmd will be wrong. quitting\n");
+			exit(999);
+		}
+	}
+	node_updown[nodeid] = 1;
+	DEBUGLOG("Added new node %d to updown list\n", nodeid);
+}
+
+static void _cluster_closedown()
+{
+	unlock_all();
+	dlm_release_lockspace(LOCKSPACE_NAME, lockspace, 1);
+	cman_finish(c_handle);
+}
+
+static int is_listening(int nodeid)
+{
+	int status;
+
+	do {
+		status = cman_is_listening(c_handle, nodeid, CLUSTER_PORT_CLVMD);
+		if (status < 0 && errno == EBUSY) {	/* Don't busywait */
+			sleep(1);
+			errno = EBUSY;	/* In case sleep trashes it */
+		}
+	}
+	while (status < 0 && errno == EBUSY);
+
+	return status;
+}
+
+/* Populate the list of CLVMDs running.
+   called only at startup time */
+static void count_clvmds_running(void)
+{
+	int i;
+
+	for (i = 0; i < num_nodes; i++) {
+		node_updown[nodes[i].cn_nodeid] = is_listening(nodes[i].cn_nodeid);
+	}
+}
+
+/* Get a list of active cluster members */
+static void get_members()
+{
+	int retnodes;
+	int status;
+
+	num_nodes = cman_get_node_count(c_handle);
+	if (num_nodes == -1) {
+		log_error("Unable to get node count");
+		return;
+	}
+
+	/* 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 cman_node));
+		if (!nodes) {
+			log_error("Unable to allocate nodes array\n");
+			exit(5);
+		}
+	}
+
+	status = cman_get_nodes(c_handle, count_nodes, &retnodes, nodes);
+	if (status < 0) {
+		log_error("Unable to get node details");
+		exit(6);
+	}
+
+	if (node_updown == NULL) {
+		size_t buf_len = sizeof(int) * max(num_nodes, max_updown_nodes);
+		node_updown = malloc(buf_len);
+		if (node_updown)
+			memset(node_updown, 0, buf_len);
+	}
+}
+
+
+/* Convert a node name to a CSID */
+static int _csid_from_name(char *csid, const char *name)
+{
+	int i;
+
+	for (i = 0; i < num_nodes; i++) {
+		if (strcmp(name, nodes[i].cn_name) == 0) {
+			memcpy(csid, &nodes[i].cn_nodeid, CMAN_MAX_CSID_LEN);
+			return 0;
+		}
+	}
+	return -1;
+}
+
+/* Convert a CSID to a node name */
+static int _name_from_csid(const char *csid, char *name)
+{
+	int i;
+
+	for (i = 0; i < num_nodes; i++) {
+		if (memcmp(csid, &nodes[i].cn_nodeid, CMAN_MAX_CSID_LEN) == 0) {
+			strcpy(name, nodes[i].cn_name);
+			return 0;
+		}
+	}
+	/* Who?? */
+	strcpy(name, "Unknown");
+	return -1;
+}
+
+/* Convert a node ID to a node name */
+static int name_from_nodeid(int nodeid, char *name)
+{
+	int i;
+
+	for (i = 0; i < num_nodes; i++) {
+		if (nodeid == nodes[i].cn_nodeid) {
+			strcpy(name, nodes[i].cn_name);
+			return 0;
+		}
+	}
+	/* Who?? */
+	strcpy(name, "Unknown");
+	return -1;
+}
+
+/* Convert a CSID to a node ID */
+static int nodeid_from_csid(const char *csid)
+{
+        int nodeid;
+
+	memcpy(&nodeid, csid, CMAN_MAX_CSID_LEN);
+
+	return nodeid;
+}
+
+static int _is_quorate()
+{
+	return cman_is_quorate(c_handle);
+}
+
+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);
+}
+
+static 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;
+}
+
+static int _sync_unlock(const char *resource /* UNUSED */, int lockid)
+{
+	int status;
+	struct lock_wait lwait;
+
+	DEBUGLOG("sync_unlock: '%s' lkid:%x\n", resource, lockid);
+
+	pthread_cond_init(&lwait.cond, NULL);
+	pthread_mutex_init(&lwait.mutex, NULL);
+	pthread_mutex_lock(&lwait.mutex);
+
+	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;
+
+}
+
+static int _get_cluster_name(char *buf, int buflen)
+{
+	cman_cluster_t cluster_info;
+	int status;
+
+	status = cman_get_cluster(c_handle, &cluster_info);
+	if (!status) {
+		strncpy(buf, cluster_info.ci_name, buflen);
+	}
+	return status;
+}
+
+static struct cluster_ops _cluster_cman_ops = {
+	.cluster_init_completed   = _cluster_init_completed,
+	.cluster_send_message     = _cluster_send_message,
+	.name_from_csid           = _name_from_csid,
+	.csid_from_name           = _csid_from_name,
+	.get_num_nodes            = _get_num_nodes,
+	.cluster_fd_callback      = _cluster_fd_callback,
+	.get_main_cluster_fd      = _get_main_cluster_fd,
+	.cluster_do_node_callback = _cluster_do_node_callback,
+	.is_quorate               = _is_quorate,
+	.get_our_csid             = _get_our_csid,
+	.add_up_node              = _add_up_node,
+	.cluster_closedown        = _cluster_closedown,
+	.get_cluster_name         = _get_cluster_name,
+	.sync_lock                = _sync_lock,
+	.sync_unlock              = _sync_unlock,
+};
+
+struct cluster_ops *init_cman_cluster(void)
+{
+	if (!_init_cluster())
+		return &_cluster_cman_ops;
+	else
+		return NULL;
+}

daemons/clvmd/clvmd-command.c

@@ -0,0 +1,326 @@
+/*
+ * 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 <libdevmapper.h>
+#include <libdlm.h>
+
+#include "list.h"
+#include "locking.h"
+#include "log.h"
+#include "lvm-functions.h"
+#include "clvmd-comms.h"
+#include "clvm.h"
+#include "clvmd.h"
+
+extern debug_t debug;
+extern struct cluster_ops *clops;
+
+/* 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) {
+			char *new_buf;
+			buflen = arglen + 200;
+			new_buf = realloc(*buf, buflen);
+			if (new_buf == NULL) {
+				status = errno;
+				free (*buf);
+			}
+			*buf = new_buf;
+		}
+		if (*buf) {
+			uname(&nodeinfo);
+			*retlen = 1 + snprintf(*buf, buflen,
+					       "TEST from %s: %s v%s",
+					       nodeinfo.nodename, args,
+					       nodeinfo.release);
+		}
+		break;
+
+	case CLVMD_CMD_LOCK_VG:
+		lockname = &args[2];
+		/* Check to see if the VG is in use by LVM1 */
+		status = do_check_lvm1(lockname);
+		/* P_global causes a cache refresh */
+		if (strcmp(lockname, "P_global") == 0)
+		      do_refresh_cache();
+
+		break;
+
+	case CLVMD_CMD_LOCK_LV:
+		/* This is the biggie */
+		lock_cmd = args[0] & 0x3F;
+		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, "%s",
+					 get_last_lvm_error());
+			return EIO;
+		}
+		break;
+
+	case CLVMD_CMD_REFRESH:
+		do_refresh_cache();
+		break;
+
+	case CLVMD_CMD_SET_DEBUG:
+		debug = args[0];
+		break;
+
+	case CLVMD_CMD_GET_CLUSTERNAME:
+		status = clops->get_cluster_name(*buf, buflen);
+		if (!status)
+			*retlen = strlen(*buf)+1;
+		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, "%s", strerror(status));
+	}
+
+	return status;
+
+}
+
+static int lock_vg(struct local_client *client)
+{
+    struct dm_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 dm_hash_table *)client->bits.localsock.private;
+    }
+    else {
+	lock_hash = dm_hash_create(3);
+	if (!lock_hash)
+	    return ENOMEM;
+	client->bits.localsock.private = (void *)lock_hash;
+    }
+
+    lock_cmd = args[0] & 0x3F;
+    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)dm_hash_lookup(lock_hash, lockname);
+	if (lkid == 0)
+	    return EINVAL;
+
+	status = sync_unlock(lockname, lkid);
+	if (status)
+	    status = errno;
+	else
+	    dm_hash_remove(lock_hash, lockname);
+    }
+    else {
+	/* Read locks need to be PR; other modes get passed through */
+	if ((lock_cmd & LCK_TYPE_MASK) == LCK_READ) {
+	    lock_cmd &= ~LCK_TYPE_MASK;
+	    lock_cmd |= LCK_PREAD;
+	}
+	status = sync_lock(lockname, (int)lock_cmd, (lock_flags & LCK_NONBLOCK) ? LKF_NOQUEUE : 0, &lkid);
+	if (status)
+	    status = errno;
+	else
+	    dm_hash_insert(lock_hash, lockname, (void *)(long)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 *)(long)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;
+
+	case CLVMD_CMD_REFRESH:
+	case CLVMD_CMD_GET_CLUSTERNAME:
+	case CLVMD_CMD_SET_DEBUG:
+		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 dm_hash_node *v;
+	struct dm_hash_table *lock_hash =
+	    (struct dm_hash_table *)client->bits.localsock.private;
+
+	dm_hash_iterate(v, lock_hash) {
+		int lkid = (int)(long)dm_hash_get_data(lock_hash, v);
+		char *lockname = dm_hash_get_key(lock_hash, v);
+
+		DEBUGLOG("cleanup: Unlocking lock %s %x\n", lockname, lkid);
+		sync_unlock(lockname, lkid);
+	}
+
+	dm_hash_destroy(lock_hash);
+	client->bits.localsock.private = 0;
+    }
+}

daemons/clvmd/clvmd-comms.h

@@ -0,0 +1,97 @@
+/*
+ * 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;
+
+struct cluster_ops {
+	void (*cluster_init_completed) (void);
+
+	int (*cluster_send_message) (const void *buf, int msglen,
+				     const char *csid,
+				     const char *errtext);
+	int (*name_from_csid) (const char *csid, char *name);
+	int (*csid_from_name) (char *csid, const char *name);
+	int (*get_num_nodes) (void);
+	int (*cluster_fd_callback) (struct local_client *fd, char *buf, int len,
+				    const char *csid,
+				    struct local_client **new_client);
+	int (*get_main_cluster_fd) (void);	/* gets accept FD or cman cluster socket */
+	int (*cluster_do_node_callback) (struct local_client *client,
+					 void (*callback) (struct local_client *,
+							   const char *csid,
+							   int node_up));
+	int (*is_quorate) (void);
+
+	void (*get_our_csid) (char *csid);
+	void (*add_up_node) (const char *csid);
+	void (*reread_config) (void);
+	void (*cluster_closedown) (void);
+
+	int (*get_cluster_name)(char *buf, int buflen);
+
+	int (*sync_lock) (const char *resource, int mode,
+			  int flags, int *lockid);
+	int (*sync_unlock) (const char *resource, int lockid);
+
+};
+
+#ifdef USE_GULM
+#  include "tcp-comms.h"
+struct cluster_ops *init_gulm_cluster(void);
+#define MAX_CSID_LEN 			GULM_MAX_CSID_LEN
+#define MAX_CLUSTER_MEMBER_NAME_LEN	GULM_MAX_CLUSTER_MEMBER_NAME_LEN
+#endif
+
+#ifdef USE_CMAN
+#  include <netinet/in.h>
+#  include "libcman.h"
+#  define CMAN_MAX_CSID_LEN 4
+#  ifndef MAX_CSID_LEN
+#    define MAX_CSID_LEN CMAN_MAX_CSID_LEN
+#  endif
+#  undef MAX_CLUSTER_MEMBER_NAME_LEN
+#  define MAX_CLUSTER_MEMBER_NAME_LEN   CMAN_MAX_NODENAME_LEN
+#  define CMAN_MAX_CLUSTER_MESSAGE 1500
+#  define CLUSTER_PORT_CLVMD 11
+struct cluster_ops *init_cman_cluster(void);
+#endif
+
+#ifdef USE_OPENAIS
+#  include <openais/saAis.h>
+#  include <openais/totem/totem.h>
+#  define OPENAIS_CSID_LEN (sizeof(int))
+#  define OPENAIS_MAX_CLUSTER_MESSAGE         MESSAGE_SIZE_MAX
+#  define OPENAIS_MAX_CLUSTER_MEMBER_NAME_LEN SA_MAX_NAME_LENGTH
+#  ifndef MAX_CLUSTER_MEMBER_NAME_LEN
+#    define MAX_CLUSTER_MEMBER_NAME_LEN       SA_MAX_NAME_LENGTH
+#  endif
+#  ifndef CMAN_MAX_CLUSTER_MESSAGE
+#    define CMAN_MAX_CLUSTER_MESSAGE          MESSAGE_SIZE_MAX
+#  endif
+#  ifndef MAX_CSID_LEN
+#    define MAX_CSID_LEN sizeof(int)
+#  endif
+struct cluster_ops *init_openais_cluster(void);
+#endif
+
+
+#endif

daemons/clvmd/clvmd-gulm.h

@@ -0,0 +1,13 @@
+
+/* DLM constant that clvmd uses as a generic NONBLOCK lock flag */
+#define LKF_NOQUEUE 1
+
+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(const char *node, char *addr);
+extern void tcp_remove_client(const char *csid);
+extern int alloc_client(int fd, const char *csid, struct local_client **new_client);
+
+void gulm_add_up_node(const char *csid);
+int gulm_name_from_csid(const char *csid, char *name);

daemons/clvmd/clvmd-openais.c

@@ -0,0 +1,737 @@
+/******************************************************************************
+*******************************************************************************
+**
+**  Copyright (C) 2007 Red Hat, Inc. All rights reserved.
+**
+*******************************************************************************
+******************************************************************************/
+
+/* This provides the interface between clvmd and OpenAIS as the cluster
+ * and lock manager.
+ *
+ */
+
+#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 <libdevmapper.h>
+
+#include <openais/saAis.h>
+#include <openais/saLck.h>
+#include <openais/cpg.h>
+
+#include "list.h"
+#include "locking.h"
+#include "log.h"
+#include "clvm.h"
+#include "clvmd-comms.h"
+#include "lvm-functions.h"
+#include "clvmd.h"
+
+/* Timeout value for several openais calls */
+#define TIMEOUT 10
+
+static void lck_lock_callback(SaInvocationT invocation,
+			      SaLckLockStatusT lockStatus,
+			      SaAisErrorT error);
+static void lck_unlock_callback(SaInvocationT invocation,
+				SaAisErrorT error);
+static void cpg_deliver_callback (cpg_handle_t handle,
+				  struct cpg_name *groupName,
+				  uint32_t nodeid,
+				  uint32_t pid,
+				  void *msg,
+				  int msg_len);
+static void cpg_confchg_callback(cpg_handle_t handle,
+				 struct cpg_name *groupName,
+				 struct cpg_address *member_list, int member_list_entries,
+				 struct cpg_address *left_list, int left_list_entries,
+				 struct cpg_address *joined_list, int joined_list_entries);
+static void _cluster_closedown(void);
+
+/* Hash list of nodes in the cluster */
+static struct dm_hash_table *node_hash;
+
+/* For associating lock IDs & resource handles */
+static struct dm_hash_table *lock_hash;
+
+/* Number of active nodes */
+static int num_nodes;
+static unsigned int our_nodeid;
+
+static struct local_client *cluster_client;
+
+/* OpenAIS handles */
+static cpg_handle_t cpg_handle;
+static SaLckHandleT lck_handle;
+
+static struct cpg_name cpg_group_name;
+
+/* Openais callback structs */
+cpg_callbacks_t cpg_callbacks = {
+	.cpg_deliver_fn =            cpg_deliver_callback,
+	.cpg_confchg_fn =            cpg_confchg_callback,
+};
+
+SaLckCallbacksT lck_callbacks = {
+        .saLckLockGrantCallback      = lck_lock_callback,
+        .saLckResourceUnlockCallback = lck_unlock_callback
+};
+
+struct node_info
+{
+	enum {NODE_UNKNOWN, NODE_DOWN, NODE_UP, NODE_CLVMD} state;
+	int nodeid;
+};
+
+struct lock_info
+{
+	SaLckResourceHandleT res_handle;
+	SaLckLockIdT         lock_id;
+	SaNameT              lock_name;
+};
+
+struct lock_wait
+{
+	pthread_cond_t cond;
+	pthread_mutex_t mutex;
+	int status;
+};
+
+/* Set errno to something approximating the right value and return 0 or -1 */
+static int ais_to_errno(SaAisErrorT err)
+{
+	switch(err)
+	{
+	case SA_AIS_OK:
+		return 0;
+        case SA_AIS_ERR_LIBRARY:
+		errno = EINVAL;
+		break;
+        case SA_AIS_ERR_VERSION:
+		errno = EINVAL;
+		break;
+        case SA_AIS_ERR_INIT:
+		errno = EINVAL;
+		break;
+        case SA_AIS_ERR_TIMEOUT:
+		errno = ETIME;
+		break;
+        case SA_AIS_ERR_TRY_AGAIN:
+		errno = EAGAIN;
+		break;
+        case SA_AIS_ERR_INVALID_PARAM:
+		errno = EINVAL;
+		break;
+        case SA_AIS_ERR_NO_MEMORY:
+		errno = ENOMEM;
+		break;
+        case SA_AIS_ERR_BAD_HANDLE:
+		errno = EINVAL;
+		break;
+        case SA_AIS_ERR_BUSY:
+		errno = EBUSY;
+		break;
+        case SA_AIS_ERR_ACCESS:
+		errno = EPERM;
+		break;
+        case SA_AIS_ERR_NOT_EXIST:
+		errno = ENOENT;
+		break;
+        case SA_AIS_ERR_NAME_TOO_LONG:
+		errno = ENAMETOOLONG;
+		break;
+        case SA_AIS_ERR_EXIST:
+		errno = EEXIST;
+		break;
+        case SA_AIS_ERR_NO_SPACE:
+		errno = ENOSPC;
+		break;
+        case SA_AIS_ERR_INTERRUPT:
+		errno = EINTR;
+		break;
+	case SA_AIS_ERR_NAME_NOT_FOUND:
+		errno = ENOENT;
+		break;
+        case SA_AIS_ERR_NO_RESOURCES:
+		errno = ENOMEM;
+		break;
+        case SA_AIS_ERR_NOT_SUPPORTED:
+		errno = EOPNOTSUPP;
+		break;
+        case SA_AIS_ERR_BAD_OPERATION:
+		errno = EINVAL;
+		break;
+        case SA_AIS_ERR_FAILED_OPERATION:
+		errno = EIO;
+		break;
+        case SA_AIS_ERR_MESSAGE_ERROR:
+		errno = EIO;
+		break;
+        case SA_AIS_ERR_QUEUE_FULL:
+		errno = EXFULL;
+		break;
+        case SA_AIS_ERR_QUEUE_NOT_AVAILABLE:
+		errno = EINVAL;
+		break;
+        case SA_AIS_ERR_BAD_FLAGS:
+		errno = EINVAL;
+		break;
+        case SA_AIS_ERR_TOO_BIG:
+		errno = E2BIG;
+		break;
+        case SA_AIS_ERR_NO_SECTIONS:
+		errno = ENOMEM;
+		break;
+	default:
+		errno = EINVAL;
+		break;
+	}
+	return -1;
+}
+
+static char *print_csid(const char *csid)
+{
+	static char buf[128];
+	int id;
+
+	memcpy(&id, csid, sizeof(int));
+	sprintf(buf, "%d", id);
+	return buf;
+}
+
+static int add_internal_client(int fd, fd_callback_t callback)
+{
+	struct local_client *client;
+
+	DEBUGLOG("Add_internal_client, fd = %d\n", fd);
+
+	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);
+
+	/* Set Close-on-exec */
+	fcntl(fd, F_SETFD, 1);
+
+	return 0;
+}
+
+static void cpg_deliver_callback (cpg_handle_t handle,
+				  struct cpg_name *groupName,
+				  uint32_t nodeid,
+				  uint32_t pid,
+				  void *msg,
+				  int msg_len)
+{
+	int target_nodeid;
+
+	memcpy(&target_nodeid, msg, OPENAIS_CSID_LEN);
+
+	DEBUGLOG("Got message from nodeid %d for %d. len %d\n",
+		 nodeid, target_nodeid, msg_len-4);
+
+	if (target_nodeid == our_nodeid)
+		process_message(cluster_client, (char *)msg+OPENAIS_CSID_LEN,
+				msg_len-OPENAIS_CSID_LEN, (char*)&nodeid);
+}
+
+static void cpg_confchg_callback(cpg_handle_t handle,
+				 struct cpg_name *groupName,
+				 struct cpg_address *member_list, int member_list_entries,
+				 struct cpg_address *left_list, int left_list_entries,
+				 struct cpg_address *joined_list, int joined_list_entries)
+{
+	int i;
+	struct node_info *ninfo;
+
+	DEBUGLOG("confchg callback. %d joined, %d left, %d members\n",
+		 joined_list_entries, left_list_entries, member_list_entries);
+
+	for (i=0; i<joined_list_entries; i++) {
+		ninfo = dm_hash_lookup_binary(node_hash,
+					      (char *)&joined_list[i].nodeid,
+					      OPENAIS_CSID_LEN);
+		if (!ninfo) {
+			ninfo = malloc(sizeof(struct node_info));
+			if (!ninfo) {
+				break;
+			}
+			else {
+				ninfo->nodeid = joined_list[i].nodeid;
+				dm_hash_insert_binary(node_hash,
+						      (char *)&ninfo->nodeid,
+						      OPENAIS_CSID_LEN, ninfo);
+			}
+		}
+		ninfo->state = NODE_CLVMD;
+	}
+
+	for (i=0; i<left_list_entries; i++) {
+		ninfo = dm_hash_lookup_binary(node_hash,
+					      (char *)&left_list[i].nodeid,
+					      OPENAIS_CSID_LEN);
+		if (ninfo)
+			ninfo->state = NODE_DOWN;
+	}
+
+	num_nodes = joined_list_entries;
+}
+
+static void lck_lock_callback(SaInvocationT invocation,
+			      SaLckLockStatusT lockStatus,
+			      SaAisErrorT error)
+{
+	struct lock_wait *lwait = (struct lock_wait *)(long)invocation;
+
+	DEBUGLOG("lck_lock_callback, error = %d\n", error);
+
+	lwait->status = error;
+	pthread_mutex_lock(&lwait->mutex);
+	pthread_cond_signal(&lwait->cond);
+	pthread_mutex_unlock(&lwait->mutex);
+}
+
+static void lck_unlock_callback(SaInvocationT invocation,
+				SaAisErrorT error)
+{
+	struct lock_wait *lwait = (struct lock_wait *)(long)invocation;
+
+	DEBUGLOG("lck_unlock_callback\n");
+
+	lwait->status = SA_AIS_OK;
+	pthread_mutex_lock(&lwait->mutex);
+	pthread_cond_signal(&lwait->cond);
+	pthread_mutex_unlock(&lwait->mutex);
+}
+
+static int lck_dispatch(struct local_client *client, char *buf, int len,
+			const char *csid, struct local_client **new_client)
+{
+	*new_client = NULL;
+	saLckDispatch(lck_handle, SA_DISPATCH_ONE);
+	return 1;
+}
+
+static int _init_cluster(void)
+{
+	SaAisErrorT err;
+	SaVersionT  ver = { 'B', 1, 1 };
+	int select_fd;
+
+	node_hash = dm_hash_create(100);
+	lock_hash = dm_hash_create(10);
+
+	err = cpg_initialize(&cpg_handle,
+			     &cpg_callbacks);
+	if (err != SA_AIS_OK) {
+		syslog(LOG_ERR, "Cannot initialise OpenAIS CPG service: %d",
+		       err);
+		DEBUGLOG("Cannot initialise OpenAIS CPG service: %d", err);
+		return ais_to_errno(err);
+	}
+
+	err = saLckInitialize(&lck_handle,
+			      &lck_callbacks,
+			      &ver);
+	if (err != SA_AIS_OK) {
+		cpg_initialize(&cpg_handle, &cpg_callbacks);
+		syslog(LOG_ERR, "Cannot initialise OpenAIS lock service: %d",
+		       err);
+		DEBUGLOG("Cannot initialise OpenAIS lock service: %d\n\n", err);
+		return ais_to_errno(err);
+	}
+
+	/* Connect to the clvmd group */
+	strcpy((char *)cpg_group_name.value, "clvmd");
+	cpg_group_name.length = strlen((char *)cpg_group_name.value);
+	err = cpg_join(cpg_handle, &cpg_group_name);
+	if (err != SA_AIS_OK) {
+		cpg_finalize(cpg_handle);
+		saLckFinalize(lck_handle);
+		syslog(LOG_ERR, "Cannot join clvmd process group");
+		DEBUGLOG("Cannot join clvmd process group: %d\n", err);
+		return ais_to_errno(err);
+	}
+
+	err = cpg_local_get(cpg_handle,
+			    &our_nodeid);
+	if (err != SA_AIS_OK) {
+		cpg_finalize(cpg_handle);
+		saLckFinalize(lck_handle);
+		syslog(LOG_ERR, "Cannot get local node id\n");
+		return ais_to_errno(err);
+	}
+	DEBUGLOG("Our local node id is %d\n", our_nodeid);
+
+	saLckSelectionObjectGet(lck_handle, (SaSelectionObjectT *)&select_fd);
+	add_internal_client(select_fd, lck_dispatch);
+
+	DEBUGLOG("Connected to OpenAIS\n");
+
+	return 0;
+}
+
+static void _cluster_closedown(void)
+{
+	DEBUGLOG("cluster_closedown\n");
+	unlock_all();
+
+	saLckFinalize(lck_handle);
+	cpg_finalize(cpg_handle);
+}
+
+static void _get_our_csid(char *csid)
+{
+	memcpy(csid, &our_nodeid, sizeof(int));
+}
+
+/* OpenAIS doesn't really have nmode names so we
+   just use the node ID in hex instead */
+static int _csid_from_name(char *csid, const char *name)
+{
+	int nodeid;
+	struct node_info *ninfo;
+
+	if (sscanf(name, "%x", &nodeid) == 1) {
+		ninfo = dm_hash_lookup_binary(node_hash, csid, OPENAIS_CSID_LEN);
+		if (ninfo)
+			return nodeid;
+	}
+	return -1;
+}
+
+static int _name_from_csid(const char *csid, char *name)
+{
+	struct node_info *ninfo;
+
+	ninfo = dm_hash_lookup_binary(node_hash, csid, OPENAIS_CSID_LEN);
+	if (!ninfo)
+	{
+		sprintf(name, "UNKNOWN %s", print_csid(csid));
+		return -1;
+	}
+
+	sprintf(name, "%x", ninfo->nodeid);
+	return 0;
+}
+
+static int _get_num_nodes()
+{
+	DEBUGLOG("num_nodes = %d\n", num_nodes);
+	return num_nodes;
+}
+
+/* Node is now known to be running a clvmd */
+static void _add_up_node(const char *csid)
+{
+	struct node_info *ninfo;
+
+	ninfo = dm_hash_lookup_binary(node_hash, csid, OPENAIS_CSID_LEN);
+	if (!ninfo) {
+		DEBUGLOG("openais_add_up_node no node_hash entry for csid %s\n",
+			 print_csid(csid));
+		return;
+	}
+
+	DEBUGLOG("openais_add_up_node %d\n", ninfo->nodeid);
+
+	ninfo->state = NODE_CLVMD;
+
+	return;
+}
+
+/* Call a callback for each node, so the caller knows whether it's up or down */
+static int _cluster_do_node_callback(struct local_client *master_client,
+				     void (*callback)(struct local_client *,
+						      const char *csid, int node_up))
+{
+	struct dm_hash_node *hn;
+	struct node_info *ninfo;
+
+	dm_hash_iterate(hn, node_hash)
+	{
+		char csid[OPENAIS_CSID_LEN];
+
+		ninfo = dm_hash_get_data(node_hash, hn);
+		memcpy(csid, dm_hash_get_key(node_hash, hn), OPENAIS_CSID_LEN);
+
+		DEBUGLOG("down_callback. node %d, state = %d\n", ninfo->nodeid,
+			 ninfo->state);
+
+		if (ninfo->state != NODE_DOWN)
+			callback(master_client, csid, ninfo->state == NODE_CLVMD);
+	}
+	return 0;
+}
+
+/* Real locking */
+static int _lock_resource(char *resource, int mode, int flags, int *lockid)
+{
+	struct lock_wait lwait;
+	struct lock_info *linfo;
+	SaLckResourceHandleT res_handle;
+	SaAisErrorT err;
+	SaLckLockIdT lock_id;
+
+	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 OpenAIS LCK */
+	if (flags & LCK_NONBLOCK) flags = SA_LCK_LOCK_NO_QUEUE;
+
+	linfo = malloc(sizeof(struct lock_info));
+	if (!linfo)
+		return -1;
+
+	DEBUGLOG("lock_resource '%s', flags=%d, mode=%d\n", resource, flags, mode);
+
+	linfo->lock_name.length = strlen(resource)+1;
+	strcpy((char *)linfo->lock_name.value, resource);
+
+	err = saLckResourceOpen(lck_handle, &linfo->lock_name,
+				SA_LCK_RESOURCE_CREATE, TIMEOUT, &res_handle);
+	if (err != SA_AIS_OK)
+	{
+		DEBUGLOG("ResourceOpen returned %d\n", err);
+		free(linfo);
+		return ais_to_errno(err);
+	}
+
+	err = saLckResourceLockAsync(res_handle,
+				     (SaInvocationT)(long)&lwait,
+				     &lock_id,
+				     mode,
+				     flags,
+				     0);
+	if (err != SA_AIS_OK)
+	{
+		free(linfo);
+		saLckResourceClose(res_handle);
+		return ais_to_errno(err);
+	}
+
+	/* Wait for it to complete */
+	pthread_cond_wait(&lwait.cond, &lwait.mutex);
+	pthread_mutex_unlock(&lwait.mutex);
+
+	DEBUGLOG("lock_resource returning %d, lock_id=%llx\n", lwait.status,
+		 lock_id);
+
+	linfo->lock_id = lock_id;
+	linfo->res_handle = res_handle;
+
+	dm_hash_insert(lock_hash, resource, linfo);
+
+	return ais_to_errno(lwait.status);
+}
+
+
+static int _unlock_resource(char *resource, int lockid)
+{
+	struct lock_wait lwait;
+	SaAisErrorT err;
+	struct lock_info *linfo;
+
+	pthread_cond_init(&lwait.cond, NULL);
+	pthread_mutex_init(&lwait.mutex, NULL);
+	pthread_mutex_lock(&lwait.mutex);
+
+	DEBUGLOG("unlock_resource %s\n", resource);
+	linfo = dm_hash_lookup(lock_hash, resource);
+	if (!linfo)
+		return 0;
+
+	DEBUGLOG("unlock_resource: lockid: %llx\n", linfo->lock_id);
+	err = saLckResourceUnlockAsync((SaInvocationT)(long)&lwait, linfo->lock_id);
+	if (err != SA_AIS_OK)
+	{
+		DEBUGLOG("Unlock returned %d\n", err);
+		return ais_to_errno(err);
+	}
+
+	/* Wait for it to complete */
+	pthread_cond_wait(&lwait.cond, &lwait.mutex);
+	pthread_mutex_unlock(&lwait.mutex);
+
+	/* Release the resource */
+	dm_hash_remove(lock_hash, resource);
+	saLckResourceClose(linfo->res_handle);
+	free(linfo);
+
+	return ais_to_errno(lwait.status);
+}
+
+static 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, SA_LCK_EX_LOCK_MODE, flags, lockid);
+		if (status)
+			goto out;
+
+		/* If we can't get this lock too then bail out */
+		status = _lock_resource(lock2, SA_LCK_EX_LOCK_MODE, LCK_NONBLOCK,
+					lockid);
+		if (status == SA_LCK_LOCK_NOT_QUEUED)
+		{
+			_unlock_resource(lock1, *lockid);
+			status = -1;
+			errno = EAGAIN;
+		}
+		break;
+
+	case LCK_PREAD:
+	case LCK_READ:
+		status = _lock_resource(lock1, SA_LCK_PR_LOCK_MODE, flags, lockid);
+		if (status)
+			goto out;
+		_unlock_resource(lock2, *lockid);
+		break;
+
+	case LCK_WRITE:
+		status = _lock_resource(lock2, SA_LCK_EX_LOCK_MODE, flags, lockid);
+		if (status)
+			goto out;
+		_unlock_resource(lock1, *lockid);
+		break;
+
+	default:
+		status = -1;
+		errno = EINVAL;
+		break;
+	}
+out:
+	*lockid = mode;
+	return status;
+}
+
+static 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);
+
+	_unlock_resource(lock1, lockid);
+	_unlock_resource(lock2, lockid);
+
+	return status;
+}
+
+/* We are always quorate ! */
+static int _is_quorate()
+{
+	return 1;
+}
+
+static int _get_main_cluster_fd(void)
+{
+	int select_fd;
+
+	cpg_fd_get(cpg_handle, &select_fd);
+	return select_fd;
+}
+
+static int _cluster_fd_callback(struct local_client *fd, char *buf, int len,
+				const char *csid,
+				struct local_client **new_client)
+{
+	cluster_client = fd;
+	*new_client = NULL;
+	cpg_dispatch(cpg_handle, SA_DISPATCH_ONE);
+	return 1;
+}
+
+static int _cluster_send_message(const void *buf, int msglen, const char *csid,
+				 const char *errtext)
+{
+	struct iovec iov[2];
+	SaAisErrorT err;
+	int target_node;
+
+	if (csid)
+		memcpy(&target_node, csid, OPENAIS_CSID_LEN);
+	else
+		target_node = 0;
+
+	iov[0].iov_base = &target_node;
+	iov[0].iov_len = sizeof(int);
+	iov[1].iov_base = (char *)buf;
+	iov[1].iov_len = msglen;
+
+	err = cpg_mcast_joined(cpg_handle, CPG_TYPE_AGREED, iov, 2);
+	return ais_to_errno(err);
+}
+
+/* We don't have a cluster name to report here */
+static int _get_cluster_name(char *buf, int buflen)
+{
+	strncpy(buf, "OpenAIS", buflen);
+	return 0;
+}
+
+static struct cluster_ops _cluster_openais_ops = {
+	.cluster_init_completed   = NULL,
+	.cluster_send_message     = _cluster_send_message,
+	.name_from_csid           = _name_from_csid,
+	.csid_from_name           = _csid_from_name,
+	.get_num_nodes            = _get_num_nodes,
+	.cluster_fd_callback      = _cluster_fd_callback,
+	.get_main_cluster_fd      = _get_main_cluster_fd,
+	.cluster_do_node_callback = _cluster_do_node_callback,
+	.is_quorate               = _is_quorate,
+	.get_our_csid             = _get_our_csid,
+	.add_up_node              = _add_up_node,
+	.reread_config            = NULL,
+	.cluster_closedown        = _cluster_closedown,
+	.get_cluster_name         = _get_cluster_name,
+	.sync_lock                = _sync_lock,
+	.sync_unlock              = _sync_unlock,
+};
+
+struct cluster_ops *init_openais_cluster(void)
+{
+	if (!_init_cluster())
+		return &_cluster_openais_ops;
+	else
+		return NULL;
+}

daemons/clvmd/clvmd.h

@@ -0,0 +1,126 @@
+/*
+ * 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;
+};
+
+typedef enum {DEBUG_OFF, DEBUG_STDERR, DEBUG_SYSLOG} debug_t;
+
+/*
+ * 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,
+			      const 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;
+	uint8_t removeme;
+
+	union {
+		struct localsock_bits localsock;
+		struct pipe_bits pipe;
+		struct netsock_bits net;
+	} bits;
+};
+
+#define DEBUGLOG(fmt, args...) debuglog(fmt, ## args);
+
+#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);
+extern void process_message(struct local_client *client, const char *buf,
+			    int len, const char *csid);
+extern void debuglog(const char *fmt, ... )
+  __attribute__ ((format(printf, 1, 2)));
+
+int sync_lock(const char *resource, int mode, int flags, int *lockid);
+int sync_unlock(const char *resource, int lockid);
+
+#endif

daemons/clvmd/cnxman-socket.h

@@ -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 CMAN_MAX_CLUSTER_MESSAGE          1500
+#define CMAN_MAX_CLUSTER_MEMBER_NAME_LEN   255
+#define MAX_BARRIER_NAME_LEN           33
+#define MAX_SA_ADDR_LEN                12
+#define MAX_CLUSTER_NAME_LEN           16
+
+/* 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[CMAN_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/lvm-functions.c

@@ -0,0 +1,631 @@
+/*
+ * 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 <libdevmapper.h>
+#include <libdlm.h>
+
+#include "list.h"
+#include "lvm-types.h"
+#include "clvm.h"
+#include "clvmd-comms.h"
+#include "clvmd.h"
+#include "lvm-functions.h"
+
+/* LVM2 headers */
+#include "toolcontext.h"
+#include "lvmcache.h"
+#include "log.h"
+#include "activate.h"
+#include "locking.h"
+#include "defaults.h"
+
+static struct cmd_context *cmd = NULL;
+static struct dm_hash_table *lv_hash = NULL;
+static pthread_mutex_t lv_hash_lock;
+static pthread_mutex_t lvm_lock;
+static char last_error[1024];
+
+struct lv_info {
+	int lock_id;
+	int lock_mode;
+};
+
+char *get_last_lvm_error()
+{
+	return last_error;
+}
+
+/* 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 = dm_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 dm_hash_node *v;
+
+	pthread_mutex_lock(&lv_hash_lock);
+	dm_hash_iterate(v, lv_hash) {
+		struct lv_info *lvi = dm_hash_get_data(lv_hash, v);
+
+		sync_unlock(dm_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 = dm_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);
+			dm_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 = dm_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);
+		dm_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, unsigned char lock_flags, int mode)
+{
+	int oldmode;
+	int status;
+	int activate_lv;
+	int exclusive = 0;
+	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)) {
+		exclusive = 1;
+		mode = LKM_EXMODE;
+	}
+
+	/* Try to get the lock if it's a clustered volume group */
+	if (lock_flags & LCK_CLUSTER_VG) {
+		status = hold_lock(resource, mode, LKF_NOQUEUE);
+		if (status) {
+			/* Return an LVM-sensible error for this.
+			 * Forcing EIO makes the upper level return this text
+			 * rather than the strerror text for EAGAIN.
+			 */
+			if (errno == EAGAIN) {
+				sprintf(last_error, "Volume is busy on another node");
+				errno = EIO;
+			}
+			return errno;
+		}
+	}
+
+	/* If it's suspended then resume it */
+	if (!lv_info_by_lvid(cmd, resource, &lvi, 0, 0))
+		return EIO;
+
+	if (lvi.suspended)
+		if (!lv_resume(cmd, resource))
+			return EIO;
+
+	/* Now activate it */
+	if (!lv_activate(cmd, resource, exclusive))
+		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_resume_lv, 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, 0, 0))
+		return EIO;
+
+	if (lvi.exists) {
+		if (!lv_suspend_if_active(cmd, resource)) {
+			return EIO;
+		}
+	}
+	return 0;
+}
+
+static int do_deactivate_lv(char *resource, unsigned char lock_flags)
+{
+	int oldmode;
+	int status;
+
+	/* Is it open ? */
+	oldmode = get_current_lock(resource);
+	if (oldmode == -1 && (lock_flags & LCK_CLUSTER_VG)) {
+		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;
+
+	if (lock_flags & LCK_CLUSTER_VG) {
+		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 = %x\n",
+		 resource, command, lock_flags);
+
+	pthread_mutex_lock(&lvm_lock);
+	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.");
+			pthread_mutex_unlock(&lvm_lock);
+			return EINVAL;
+		}
+	}
+
+	if (lock_flags & LCK_PARTIAL_MODE)
+		init_partial(1);
+
+	if (lock_flags & LCK_MIRROR_NOSYNC_MODE)
+		init_mirror_in_sync(1);
+
+	if (!(lock_flags & LCK_DMEVENTD_MONITOR_MODE))
+		init_dmeventd_monitor(0);
+
+	switch (command) {
+	case LCK_LV_EXCLUSIVE:
+		status = do_activate_lv(resource, lock_flags, 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, lock_flags, LKM_CRMODE);
+		break;
+
+	case LCK_LV_DEACTIVATE:
+		status = do_deactivate_lv(resource, lock_flags);
+		break;
+
+	default:
+		DEBUGLOG("Invalid LV command 0x%x\n", command);
+		status = EINVAL;
+		break;
+	}
+
+	if (lock_flags & LCK_PARTIAL_MODE)
+		init_partial(0);
+
+	if (lock_flags & LCK_MIRROR_NOSYNC_MODE)
+		init_mirror_in_sync(0);
+
+	if (!(lock_flags & LCK_DMEVENTD_MONITOR_MODE))
+		init_dmeventd_monitor(DEFAULT_DMEVENTD_MONITOR);
+
+	/* clean the pool for another command */
+	dm_pool_empty(cmd->mem);
+	pthread_mutex_unlock(&lvm_lock);
+
+	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)
+{
+	int status;
+
+	/* 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;
+
+			pthread_mutex_lock(&lvm_lock);
+			status = lv_info_by_lvid(cmd, resource, &lvi, 0, 0);
+			pthread_mutex_unlock(&lvm_lock);
+			if (!status)
+				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(const char *vgname)
+{
+	int status;
+
+	status = check_lvm1_vg_inactive(cmd, vgname);
+
+	return status == 1 ? 0 : EBUSY;
+}
+
+int do_refresh_cache()
+{
+	int ret;
+	DEBUGLOG("Refreshing context\n");
+	log_notice("Refreshing context");
+
+	ret = refresh_toolcontext(cmd);
+	init_full_scan_done(0);
+	lvmcache_label_scan(cmd, 2);
+
+	return ret==1?0:-1;
+}
+
+
+/* Only called at gulm startup. Drop any leftover VG or P_orphan locks
+   that might be hanging around if we died for any reason
+*/
+static void drop_vg_locks()
+{
+	char vg[128];
+	char line[255];
+	FILE *vgs =
+	    popen
+	    ("lvm pvs --nolocking --noheadings -o vg_name", "r");
+
+	sync_unlock("P_orphans", LCK_EXCL);
+
+	if (!vgs)
+		return;
+
+	while (fgets(line, sizeof(line), vgs)) {
+		char *vgend;
+		char *vgstart;
+
+		if (line[strlen(line)-1] == '\n')
+			line[strlen(line)-1] = '\0';
+
+		vgstart = line + strspn(line, " ");
+		vgend = vgstart + strcspn(vgstart, " ");
+		*vgend = '\0';
+
+		if (strncmp(vgstart, "WARNING:", 8) == 0)
+			continue;
+
+		sprintf(vg, "V_%s", vgstart);
+		sync_unlock(vg, LCK_EXCL);
+
+	}
+	if (fclose(vgs))
+		DEBUGLOG("vgs fclose failed: %s\n", strerror(errno));
+}
+
+/*
+ * 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], vg_flags[25];
+	char uuid[65];
+	char line[255];
+	FILE *lvs =
+	    popen
+	    ("lvm lvs --nolocking --noheadings -o vg_uuid,lv_uuid,lv_attr,vg_attr",
+	     "r");
+
+	if (!lvs)
+		return NULL;
+
+	while (fgets(line, sizeof(line), lvs)) {
+	        if (sscanf(line, "%s %s %s %s\n", vg, lv, flags, vg_flags) == 4) {
+
+			/* States: s:suspended a:active S:dropped snapshot I:invalid snapshot */
+		        if (strlen(vg) == 38 &&                         /* is is a valid UUID ? */
+			    (flags[4] == 'a' || flags[4] == 's') &&	/* is it active or suspended? */
+			    vg_flags[5] == 'c') {			/* is it clustered ? */
+				/* 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);
+			}
+		}
+	}
+	if (fclose(lvs))
+		DEBUGLOG("lvs fclose failed: %s\n", strerror(errno));
+	return NULL;
+}
+
+static void lvm2_log_fn(int level, const char *file, int line,
+			const char *message)
+{
+
+	/* Send messages to the normal LVM2 logging system too,
+	   so we get debug output when it's asked for. 
+ 	   We need to NULL the function ptr otherwise it will just call
+	   back into here! */
+	init_log_fn(NULL);
+	print_log(level, file, line, "%s", message);
+	init_log_fn(lvm2_log_fn);
+
+	/*
+	 * Ignore non-error messages, but store the latest one for returning
+	 * to the user.
+	 */
+	if (level != _LOG_ERR && level != _LOG_FATAL)
+		return;
+
+	strncpy(last_error, message, sizeof(last_error));
+	last_error[sizeof(last_error)-1] = '\0';
+}
+
+/* This checks some basic cluster-LVM configuration stuff */
+static void check_config()
+{
+	int locking_type;
+
+	locking_type = find_config_tree_int(cmd, "global/locking_type", 1);
+
+	if (locking_type == 3) /* compiled-in cluster support */
+		return;
+
+	if (locking_type == 2) { /* External library, check name */
+		const char *libname;
+
+		libname = find_config_tree_str(cmd, "global/locking_library",
+					  "");
+		if (strstr(libname, "liblvm2clusterlock.so"))
+			return;
+
+		log_error("Incorrect LVM locking library specified in lvm.conf, cluster operations may not work.");
+		return;
+	}
+	log_error("locking_type not set correctly in lvm.conf, cluster operations will not work.");
+}
+
+void init_lvhash()
+{
+	/* Create hash table for keeping LV locks & status */
+	lv_hash = dm_hash_create(100);
+	pthread_mutex_init(&lv_hash_lock, NULL);
+	pthread_mutex_init(&lvm_lock, NULL);
+}
+
+/* Called to initialise the LVM context of the daemon */
+int init_lvm(int using_gulm)
+{
+	if (!(cmd = create_toolcontext(NULL, 0, 1))) {
+		log_error("Failed to allocate command context");
+		return 0;
+	}
+
+	/* Use LOG_DAEMON for syslog messages instead of LOG_USER */
+	init_syslog(LOG_DAEMON);
+	openlog("clvmd", LOG_PID, LOG_DAEMON);
+	init_debug(cmd->current_settings.debug);
+	init_verbose(cmd->current_settings.verbose + VERBOSE_BASE_LEVEL);
+	set_activation(cmd->current_settings.activation);
+
+	/* Check lvm.conf is setup for cluster-LVM */
+	check_config();
+
+	/* Remove any non-LV locks that may have been left around */
+	if (using_gulm)
+		drop_vg_locks();
+
+	get_initial_state();
+
+	/* Trap log messages so we can pass them back to the user */
+	init_log_fn(lvm2_log_fn);
+
+	return 1;
+}

daemons/clvmd/lvm-functions.h

@@ -0,0 +1,37 @@
+/*
+ * 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(const char *vgname);
+extern int do_refresh_cache(void);
+extern int init_lvm(int using_gulm);
+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);
+extern char *get_last_lvm_error(void);
+
+#endif

daemons/clvmd/refresh_clvmd.c

@@ -0,0 +1,363 @@
+/*
+ * Copyright (C) 2002-2004 Sistina Software, Inc. All rights reserved.
+ * Copyright (C) 2004-2007 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
+ */
+
+/*
+ * Tell all clvmds in a cluster to refresh their toolcontext
+ *
+ */
+
+#include <stddef.h>
+#include <sys/socket.h>
+#include <sys/un.h>
+#include <errno.h>
+#include <unistd.h>
+#include <libdevmapper.h>
+#include <stdint.h>
+#include <stdio.h>
+
+#include "clvm.h"
+#include "refresh_clvmd.h"
+
+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
+
+static int _clvmd_sock = -1;
+
+/* Open connection to the clvm 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) {
+		fprintf(stderr, "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;
+
+		fprintf(stderr, "connect() failed on local socket: %s\n",
+			  strerror(errno));
+		if (close(local_socket))
+			return -1;
+
+		errno = saved_errno;
+		return -1;
+	}
+
+	return local_socket;
+}
+
+/* Send a request and return the status */
+static int _send_request(const 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;
+		fprintf(stderr, "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;
+		fprintf(stderr, "Error reading data from clvmd: %s", strerror(errno));
+		return 0;
+	}
+
+	if (len == 0) {
+		fprintf(stderr, "EOF reading CLVMD");
+		errno = ENOTCONN;
+		return 0;
+	}
+
+	/* Allocate buffer */
+	buflen = len + outheader->arglen;
+	*retbuf = dm_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;
+
+		/* Only return an error here if there are no node-specific
+		   errors present in the message that might have more detail */
+		if (!(outheader->flags & CLVMD_FLAG_NODEERRS)) {
+			fprintf(stderr, "cluster request failed: %s\n", 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];
+	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
+	 */
+	*response = dm_malloc(sizeof(lvm_response_t) * num_responses +
+			    sizeof(int) * 2);
+	if (!*response) {
+		errno = ENOMEM;
+		status = 0;
+		goto out;
+	}
+
+	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;
+
+		memcpy(&rarray[i].status, inptr, sizeof(int));
+		inptr += sizeof(int);
+
+		rarray[i].response = dm_malloc(strlen(inptr) + 1);
+		if (rarray[i].response == NULL) {
+			/* Free up everything else and return error */
+			int j;
+			for (j = 0; j < i; j++)
+				dm_free(rarray[i].response);
+			free(*response);
+			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)
+		dm_free(retbuf);
+
+	return status;
+}
+
+/* Free reply array */
+static int _cluster_free_request(lvm_response_t * response, int num)
+{
+	int i;
+
+	for (i = 0; i < num; i++) {
+		dm_free(response[i].response);
+	}
+
+	dm_free(response);
+
+	return 1;
+}
+
+int refresh_clvmd()
+{
+	int num_responses;
+	char args[1]; // No args really.
+	lvm_response_t *response;
+	int saved_errno;
+	int status;
+	int i;
+
+	status = _cluster_request(CLVMD_CMD_REFRESH, "*", args, 0, &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) {
+			fprintf(stderr, "clvmd not running on node %s",
+				  response[i].node);
+			status = 0;
+			errno = response[i].status;
+		} else if (response[i].status) {
+			fprintf(stderr, "Error resetting node %s: %s",
+				  response[i].node,
+				  response[i].response[0] ?
+				  	response[i].response :
+				  	strerror(response[i].status));
+			status = 0;
+			errno = response[i].status;
+		}
+	}
+
+	saved_errno = errno;
+	_cluster_free_request(response, num_responses);
+	errno = saved_errno;
+
+	return status;
+}
+
+int debug_clvmd(int level, int clusterwide)
+{
+	int num_responses;
+	char args[1];
+	const char *nodes;
+	lvm_response_t *response;
+	int saved_errno;
+	int status;
+	int i;
+
+	args[0] = level;
+	if (clusterwide)
+		nodes = "*";
+	else
+		nodes = ".";
+
+	status = _cluster_request(CLVMD_CMD_SET_DEBUG, nodes, args, 1, &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) {
+			fprintf(stderr, "clvmd not running on node %s",
+				  response[i].node);
+			status = 0;
+			errno = response[i].status;
+		} else if (response[i].status) {
+			fprintf(stderr, "Error setting debug on node %s: %s",
+				  response[i].node,
+				  response[i].response[0] ?
+				  	response[i].response :
+				  	strerror(response[i].status));
+			status = 0;
+			errno = response[i].status;
+		}
+	}
+
+	saved_errno = errno;
+	_cluster_free_request(response, num_responses);
+	errno = saved_errno;
+
+	return status;
+}

daemons/clvmd/refresh_clvmd.h

@@ -0,0 +1,18 @@
+/*
+ * Copyright (C) 2007 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
+ */
+
+
+int refresh_clvmd(void);
+int debug_clvmd(int level, int clusterwide);
+

daemons/clvmd/tcp-comms.c

@@ -0,0 +1,501 @@
+/******************************************************************************
+*******************************************************************************
+**
+**  Copyright (C) Sistina Software, Inc.  2002-2003  All rights reserved.
+**  Copyright (C) 2004-2005 Red Hat, Inc. 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 <libdevmapper.h>
+
+#include "clvm.h"
+#include "clvmd-comms.h"
+#include "clvmd.h"
+#include "clvmd-gulm.h"
+
+#define DEFAULT_TCP_PORT 21064
+
+static int listen_fd = -1;
+static int tcp_port;
+struct dm_hash_table *sock_hash;
+
+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 */
+int init_comms(unsigned short port)
+{
+    struct sockaddr_in6 addr;
+
+    sock_hash = dm_hash_create(100);
+    tcp_port = port ? : DEFAULT_TCP_PORT;
+
+    listen_fd = socket(AF_INET6, SOCK_STREAM, 0);
+
+    if (listen_fd < 0)
+    {
+	return -1;
+    }
+    else
+    {
+	int one = 1;
+	setsockopt(listen_fd, SOL_SOCKET, SO_REUSEADDR, &one, sizeof(int));
+	setsockopt(listen_fd, SOL_SOCKET, SO_KEEPALIVE, &one, sizeof(int));
+    }
+
+    memset(&addr, 0, sizeof(addr)); // Bind to INADDR_ANY
+    addr.sin6_family = AF_INET6;
+    addr.sin6_port = htons(tcp_port);
+
+    if (bind(listen_fd, (struct sockaddr *)&addr, sizeof(addr)) < 0)
+    {
+	DEBUGLOG("Can't bind to port: %s\n", strerror(errno));
+	syslog(LOG_ERR, "Can't bind to port %d, is clvmd already running ?", tcp_port);
+	close(listen_fd);
+	return -1;
+    }
+
+    listen(listen_fd, 5);
+
+    /* Set Close-on-exec */
+    fcntl(listen_fd, F_SETFD, 1);
+
+    return 0;
+}
+
+void tcp_remove_client(const char *c_csid)
+{
+    struct local_client *client;
+    char csid[GULM_MAX_CSID_LEN];
+    unsigned int i;
+    memcpy(csid, c_csid, sizeof csid);
+    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 */
+    for (i = 0; i < 2; i++)
+    {
+	client = dm_hash_lookup_binary(sock_hash, csid, GULM_MAX_CSID_LEN);
+	if (client)
+	{
+	    dm_hash_remove_binary(sock_hash, csid, GULM_MAX_CSID_LEN);
+	    client->removeme = 1;
+	    close(client->fd);
+	}
+	/* Look for a mangled one too, on the 2nd iteration. */
+	csid[0] ^= 0x80;
+    }
+}
+
+int alloc_client(int fd, const char *c_csid, struct local_client **new_client)
+{
+    struct local_client *client;
+    char csid[GULM_MAX_CSID_LEN];
+    memcpy(csid, c_csid, sizeof csid);
+
+    DEBUGLOG("alloc_client %d csid = %s\n", fd, print_csid(csid));
+
+    /* 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 (dm_hash_lookup_binary(sock_hash, csid, GULM_MAX_CSID_LEN))
+    {
+	DEBUGLOG("alloc_client mangling CSID for second connection\n");
+	/* This is a duplicate connection but we can't close it because
+	   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 (dm_hash_lookup_binary(sock_hash, csid, GULM_MAX_CSID_LEN))
+        {
+	    DEBUGLOG("Multiple incoming connections from node\n");
+            syslog(LOG_ERR, " Bogus incoming connection from %d.%d.%d.%d\n", csid[0],csid[1],csid[2],csid[3]);
+
+	    free(client);
+            errno = ECONNREFUSED;
+            return -1;
+        }
+    }
+    dm_hash_insert_binary(sock_hash, csid, GULM_MAX_CSID_LEN, client);
+
+    return 0;
+}
+
+int get_main_gulm_cluster_fd()
+{
+    return listen_fd;
+}
+
+
+/* Read on main comms (listen) socket, accept it */
+int cluster_fd_gulm_callback(struct local_client *fd, char *buf, int len, const char *csid,
+			struct local_client **new_client)
+{
+    int newfd;
+    struct sockaddr_in6 addr;
+    socklen_t addrlen = sizeof(addr);
+    int status;
+    char name[GULM_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.
+    */
+    if (gulm_name_from_csid((char *)&addr.sin6_addr, name) < 0)
+    {
+	syslog(LOG_ERR, "Got connect from non-cluster node %s\n",
+	       print_csid((char *)&addr.sin6_addr));
+	DEBUGLOG("Got connect from non-cluster node %s\n",
+		 print_csid((char *)&addr.sin6_addr));
+	close(newfd);
+
+	errno = EAGAIN;
+	return -1;
+    }
+
+    status = alloc_client(newfd, (char *)&addr.sin6_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;
+}
+
+/* Try to get at least 'len' bytes from the socket */
+static int really_read(int fd, char *buf, int len)
+{
+	int got, offset;
+
+	got = offset = 0;
+
+	do {
+		got = read(fd, buf+offset, len-offset);
+		DEBUGLOG("really_read. got %d bytes\n", got);
+		offset += got;
+	} while (got > 0 && offset < len);
+
+	if (got < 0)
+		return got;
+	else
+		return offset;
+}
+
+
+static int read_from_tcpsock(struct local_client *client, char *buf, int len, char *csid,
+			     struct local_client **new_client)
+{
+    struct sockaddr_in6 addr;
+    socklen_t slen = sizeof(addr);
+    struct clvm_header *header = (struct clvm_header *)buf;
+    int status;
+    uint32_t arglen;
+
+    DEBUGLOG("read_from_tcpsock fd %d\n", client->fd);
+    *new_client = NULL;
+
+    /* Get "csid" */
+    getpeername(client->fd, (struct sockaddr *)&addr, &slen);
+    memcpy(csid, &addr.sin6_addr, GULM_MAX_CSID_LEN);
+
+    /* Read just the header first, then get the rest if there is any.
+     * Stream sockets, sigh.
+     */
+    status = really_read(client->fd, buf, sizeof(struct clvm_header));
+    if (status > 0)
+    {
+	    int status2;
+
+	    arglen = ntohl(header->arglen);
+
+	    /* Get the rest */
+	    if (arglen && arglen < GULM_MAX_CLUSTER_MESSAGE)
+	    {
+		    status2 = really_read(client->fd, buf+status, arglen);
+		    if (status2 > 0)
+			    status += status2;
+		    else
+			    status = status2;
+	    }
+    }
+
+    DEBUGLOG("read_from_tcpsock, status = %d(errno = %d)\n", status, errno);
+
+    /* 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[GULM_MAX_CSID_LEN];
+
+	memcpy(remcsid, csid, GULM_MAX_CSID_LEN);
+	close(client->fd);
+
+	/* If the csid was mangled, then make sure we remove the right entry */
+	if (client->bits.net.flags)
+	    remcsid[0] ^= 0x80;
+	dm_hash_remove_binary(sock_hash, remcsid, GULM_MAX_CSID_LEN);
+
+	/* Tell cluster manager layer */
+	add_down_node(remcsid);
+    }
+    else {
+	    gulm_add_up_node(csid);
+	    /* Send it back to clvmd */
+	    process_message(client, buf, status, csid);
+    }
+    return status;
+}
+
+int gulm_connect_csid(const char *csid, struct local_client **newclient)
+{
+    int fd;
+    struct sockaddr_in6 addr;
+    int status;
+    int one = 1;
+
+    DEBUGLOG("Connecting socket\n");
+    fd = socket(PF_INET6, SOCK_STREAM, 0);
+
+    if (fd < 0)
+    {
+	syslog(LOG_ERR, "Unable to create new socket: %m");
+	return -1;
+    }
+
+    addr.sin6_family = AF_INET6;
+    memcpy(&addr.sin6_addr, csid, GULM_MAX_CSID_LEN);
+    addr.sin6_port = htons(tcp_port);
+
+    DEBUGLOG("Connecting socket %d\n", fd);
+    if (connect(fd, (struct sockaddr *)&addr, sizeof(struct sockaddr_in6)) < 0)
+    {
+	/* "Connection refused" is "normal" because clvmd may not yet be running
+	 * on that node.
+	 */
+	if (errno != ECONNREFUSED)
+	{
+	    syslog(LOG_ERR, "Unable to connect to remote node: %m");
+	}
+	DEBUGLOG("Unable to connect to remote node: %s\n", strerror(errno));
+	close(fd);
+	return -1;
+    }
+
+    /* Set Close-on-exec */
+    fcntl(fd, F_SETFD, 1);
+    setsockopt(fd, SOL_SOCKET, SO_KEEPALIVE, &one, sizeof(int));
+
+    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 */
+    gulm_add_up_node(csid);
+    return status;
+}
+
+/* Send a message to a known CSID */
+static int tcp_send_message(void *buf, int msglen, const char *csid, const char *errtext)
+{
+    int status;
+    struct local_client *client;
+    char ourcsid[GULM_MAX_CSID_LEN];
+
+    assert(csid);
+
+    DEBUGLOG("tcp_send_message, csid = %s, msglen = %d\n", print_csid(csid), msglen);
+
+    /* Don't connect to ourself */
+    get_our_gulm_csid(ourcsid);
+    if (memcmp(csid, ourcsid, GULM_MAX_CSID_LEN) == 0)
+	return msglen;
+
+    client = dm_hash_lookup_binary(sock_hash, csid, GULM_MAX_CSID_LEN);
+    if (!client)
+    {
+	status = gulm_connect_csid(csid, &client);
+	if (status)
+	    return -1;
+    }
+    DEBUGLOG("tcp_send_message, fd = %d\n", client->fd);
+
+    return write(client->fd, buf, msglen);
+}
+
+
+int gulm_cluster_send_message(void *buf, int msglen, const 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[GULM_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;
+}
+
+/* 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)
+{
+	struct utsname info;
+
+	uname(&info);
+	get_ip_address(info.nodename, addr);
+
+	return 0;
+}
+
+/* Public version of above for those that don't care what protocol
+   we're using */
+void get_our_gulm_csid(char *csid)
+{
+    static char our_csid[GULM_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, GULM_MAX_CSID_LEN);
+}
+
+static void map_v4_to_v6(struct in_addr *ip4, struct in6_addr *ip6)
+{
+   ip6->s6_addr32[0] = 0;
+   ip6->s6_addr32[1] = 0;
+   ip6->s6_addr32[2] = htonl(0xffff);
+   ip6->s6_addr32[3] = ip4->s_addr;
+}
+
+/* Get someone else's IP address from DNS */
+int get_ip_address(const char *node, char *addr)
+{
+    struct hostent *he;
+
+    memset(addr, 0, GULM_MAX_CSID_LEN);
+
+    // TODO: what do we do about multi-homed hosts ???
+    // CCSs ip_interfaces solved this but some bugger removed it.
+
+    /* 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)
+    {
+	memcpy(addr, he->h_addr_list[0],
+	       he->h_length);
+    }
+    else
+    {
+	he = gethostbyname2(node, AF_INET);
+	if (!he)
+	    return -1;
+	map_v4_to_v6((struct in_addr *)he->h_addr_list[0], (struct in6_addr *)addr);
+    }
+
+    return 0;
+}
+
+char *print_csid(const char *csid)
+{
+    static char buf[128];
+    int *icsid = (int *)csid;
+
+    sprintf(buf, "[%x.%x.%x.%x]",
+	    icsid[0],icsid[1],icsid[2],icsid[3]);
+
+    return buf;
+}

daemons/clvmd/tcp-comms.h

@@ -0,0 +1,13 @@
+#include <netinet/in.h>
+
+#define GULM_MAX_CLUSTER_MESSAGE 1600
+#define GULM_MAX_CSID_LEN sizeof(struct in6_addr)
+#define GULM_MAX_CLUSTER_MEMBER_NAME_LEN 128
+
+extern int init_comms(unsigned short);
+extern char *print_csid(const char *);
+int get_main_gulm_cluster_fd(void);
+int cluster_fd_gulm_callback(struct local_client *fd, char *buf, int len, const char *csid, struct local_client **new_client);
+int gulm_cluster_send_message(void *buf, int msglen, const char *csid, const char *errtext);
+void get_our_gulm_csid(char *csid);
+int gulm_connect_csid(const char *csid, struct local_client **newclient);

daemons/dmeventd/.exported_symbols

@@ -0,0 +1,19 @@
+dm_event_handler_create
+dm_event_handler_destroy
+dm_event_handler_set_dso
+dm_event_handler_set_dev_name
+dm_event_handler_set_uuid
+dm_event_handler_set_major
+dm_event_handler_set_minor
+dm_event_handler_set_event_mask
+dm_event_handler_get_dso
+dm_event_handler_get_devname
+dm_event_handler_get_uuid
+dm_event_handler_get_major
+dm_event_handler_get_minor
+dm_event_handler_get_event_mask
+dm_event_register_handler
+dm_event_unregister_handler
+dm_event_get_registered_device
+dm_event_handler_set_timeout
+dm_event_handler_get_timeout

daemons/dmeventd/Makefile.in

@@ -0,0 +1,83 @@
+#
+# Copyright (C) 2005-2007 Red Hat, Inc. All rights reserved.
+#
+# This file is part of the device-mapper userspace tools.
+#
+# This copyrighted material is made available to anyone wishing to use,
+# modify, copy, or redistribute it subject to the terms and conditions
+# of the GNU Lesser General Public License v.2.1.
+#
+# You should have received a copy of the GNU Lesser General Public License
+# along with this program; if not, write to the Free Software Foundation,
+# Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+
+srcdir = @srcdir@
+top_srcdir = @top_srcdir@
+VPATH = @srcdir@
+
+SOURCES = libdevmapper-event.c
+
+LIB_STATIC = libdevmapper-event.a
+
+ifeq ("@LIB_SUFFIX@","dylib")
+  LIB_SHARED = libdevmapper-event.dylib
+else
+  LIB_SHARED = libdevmapper-event.so
+endif
+
+TARGETS = dmeventd
+CLEAN_TARGETS = dmeventd.o
+
+include ../make.tmpl
+
+LDFLAGS += -ldl -ldevmapper -lpthread
+CLDFLAGS += -ldl -ldevmapper -lpthread
+
+dmeventd: $(LIB_SHARED) dmeventd.o
+	$(CC) -o $@ dmeventd.o $(CFLAGS) $(LDFLAGS) \
+	-L. -ldevmapper-event $(LIBS) -rdynamic
+
+.PHONY: install_dynamic install_static install_include \
+	install_pkgconfig install_dmeventd
+
+INSTALL_TYPE = install_dynamic
+
+ifeq ("@STATIC_LINK@", "yes")
+  INSTALL_TYPE += install_static
+endif
+
+ifeq ("@PKGCONFIG@", "yes")
+  INSTALL_TYPE += install_pkgconfig
+endif
+
+install: $(INSTALL_TYPE) install_include install_dmeventd
+
+install_include:
+	$(INSTALL) -D $(OWNER) $(GROUP) -m 444 libdevmapper-event.h \
+		$(includedir)/libdevmapper-event.h
+
+install_dynamic: libdevmapper-event.$(LIB_SUFFIX)
+	$(INSTALL) -D $(OWNER) $(GROUP) -m 555 $(STRIP) $< \
+		$(libdir)/libdevmapper-event.$(LIB_SUFFIX).$(LIB_VERSION)
+	$(LN_S) -f libdevmapper-event.$(LIB_SUFFIX).$(LIB_VERSION) \
+		$(libdir)/libdevmapper-event.$(LIB_SUFFIX)
+
+install_dmeventd: dmeventd
+	$(INSTALL) -D $(OWNER) $(GROUP) -m 555 $(STRIP) $< $(sbindir)/$<
+
+install_pkgconfig:
+	$(INSTALL) -D $(OWNER) $(GROUP) -m 444 libdevmapper-event.pc \
+		$(usrlibdir)/pkgconfig/devmapper-event.pc
+
+install_static: libdevmapper-event.a
+	$(INSTALL) -D $(OWNER) $(GROUP) -m 555 $(STRIP) $< \
+		$(libdir)/libdevmapper-event.a.$(LIB_VERSION)
+	$(LN_S) -f libdevmapper-event.a.$(LIB_VERSION) $(libdir)/libdevmapper-event.a
+
+.PHONY: distclean_lib distclean
+
+distclean_lib: 
+	$(RM) libdevmapper-event.pc
+
+distclean: distclean_lib
+

daemons/dmeventd/dmeventd.h

@@ -0,0 +1,66 @@
+/*
+ * Copyright (C) 2005-2007 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of the device-mapper userspace tools.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#ifndef __DMEVENTD_DOT_H__
+#define __DMEVENTD_DOT_H__
+
+/* FIXME This stuff must be configurable. */
+
+#define	DM_EVENT_DAEMON		"/sbin/dmeventd"
+#define DM_EVENT_LOCKFILE	"/var/lock/dmeventd"
+#define	DM_EVENT_FIFO_CLIENT	"/var/run/dmeventd-client"
+#define	DM_EVENT_FIFO_SERVER	"/var/run/dmeventd-server"
+#define DM_EVENT_PIDFILE	"/var/run/dmeventd.pid"
+
+#define DM_EVENT_DEFAULT_TIMEOUT 10
+
+/* Commands for the daemon passed in the message below. */
+enum dm_event_command {
+	DM_EVENT_CMD_ACTIVE = 1,
+	DM_EVENT_CMD_REGISTER_FOR_EVENT,
+	DM_EVENT_CMD_UNREGISTER_FOR_EVENT,
+	DM_EVENT_CMD_GET_REGISTERED_DEVICE,
+	DM_EVENT_CMD_GET_NEXT_REGISTERED_DEVICE,
+	DM_EVENT_CMD_SET_TIMEOUT,
+	DM_EVENT_CMD_GET_TIMEOUT,
+	DM_EVENT_CMD_HELLO,
+};
+
+/* Message passed between client and daemon. */
+struct dm_event_daemon_message {
+	uint32_t cmd;
+	uint32_t size;
+	char *data;
+};
+
+/* FIXME Is this meant to be exported?  I can't see where the
+   interface uses it. */
+/* Fifos for client/daemon communication. */
+struct dm_event_fifos {
+	int client;
+	int server;
+	const char *client_path;
+	const char *server_path;
+};
+
+/*      EXIT_SUCCESS             0 -- stdlib.h */
+/*      EXIT_FAILURE             1 -- stdlib.h */
+#define EXIT_LOCKFILE_INUSE      2
+#define EXIT_DESC_CLOSE_FAILURE  3
+#define EXIT_DESC_OPEN_FAILURE   4
+#define EXIT_OPEN_PID_FAILURE    5
+#define EXIT_FIFO_FAILURE        6
+#define EXIT_CHDIR_FAILURE       7
+
+#endif /* __DMEVENTD_DOT_H__ */

daemons/dmeventd/libdevmapper-event.c

@@ -0,0 +1,809 @@
+/*
+ * Copyright (C) 2005-2007 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of the device-mapper userspace tools.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include "lib.h"
+#include "libdevmapper-event.h"
+//#include "libmultilog.h"
+#include "dmeventd.h"
+
+#include <errno.h>
+#include <fcntl.h>
+#include <stdio.h>
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/file.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <unistd.h>
+#include <sys/wait.h>
+#include <arpa/inet.h>		/* for htonl, ntohl */
+
+static int _sequence_nr = 0;
+
+struct dm_event_handler {
+	char *dso;
+
+	char *dev_name;
+
+	char *uuid;
+	int major;
+	int minor;
+	uint32_t timeout;
+
+	enum dm_event_mask mask;
+};
+
+static void _dm_event_handler_clear_dev_info(struct dm_event_handler *dmevh)
+{
+	if (dmevh->dev_name)
+		dm_free(dmevh->dev_name);
+	if (dmevh->uuid)
+		dm_free(dmevh->uuid);
+	dmevh->dev_name = dmevh->uuid = NULL;
+	dmevh->major = dmevh->minor = 0;
+}
+
+struct dm_event_handler *dm_event_handler_create(void)
+{
+	struct dm_event_handler *dmevh = NULL;
+
+	if (!(dmevh = dm_malloc(sizeof(*dmevh))))
+		return NULL;
+
+	dmevh->dso = dmevh->dev_name = dmevh->uuid = NULL;
+	dmevh->major = dmevh->minor = 0;
+	dmevh->mask = 0;
+	dmevh->timeout = 0;
+
+	return dmevh;
+}
+
+void dm_event_handler_destroy(struct dm_event_handler *dmevh)
+{
+	_dm_event_handler_clear_dev_info(dmevh);
+	if (dmevh->dso)
+		dm_free(dmevh->dso);
+	dm_free(dmevh);
+}
+
+int dm_event_handler_set_dso(struct dm_event_handler *dmevh, const char *path)
+{
+	if (!path) /* noop */
+		return 0;
+	if (dmevh->dso)
+		dm_free(dmevh->dso);
+
+	dmevh->dso = dm_strdup(path);
+	if (!dmevh->dso)
+		return -ENOMEM;
+
+	return 0;
+}
+
+int dm_event_handler_set_dev_name(struct dm_event_handler *dmevh, const char *dev_name)
+{
+	if (!dev_name)
+		return 0;
+
+	_dm_event_handler_clear_dev_info(dmevh);
+
+	dmevh->dev_name = dm_strdup(dev_name);
+	if (!dmevh->dev_name)
+		return -ENOMEM;
+	return 0;
+}
+
+int dm_event_handler_set_uuid(struct dm_event_handler *dmevh, const char *uuid)
+{
+	if (!uuid)
+		return 0;
+
+	_dm_event_handler_clear_dev_info(dmevh);
+
+	dmevh->uuid = dm_strdup(uuid);
+	if (!dmevh->dev_name)
+		return -ENOMEM;
+	return 0;
+}
+
+void dm_event_handler_set_major(struct dm_event_handler *dmevh, int major)
+{
+	int minor = dmevh->minor;
+
+	_dm_event_handler_clear_dev_info(dmevh);
+
+	dmevh->major = major;
+	dmevh->minor = minor;
+}
+
+void dm_event_handler_set_minor(struct dm_event_handler *dmevh, int minor)
+{
+	int major = dmevh->major;
+
+	_dm_event_handler_clear_dev_info(dmevh);
+
+	dmevh->major = major;
+	dmevh->minor = minor;
+}
+
+void dm_event_handler_set_event_mask(struct dm_event_handler *dmevh,
+				     enum dm_event_mask evmask)
+{
+	dmevh->mask = evmask;
+}
+
+void dm_event_handler_set_timeout(struct dm_event_handler *dmevh, int timeout)
+{
+	dmevh->timeout = timeout;
+}
+
+const char *dm_event_handler_get_dso(const struct dm_event_handler *dmevh)
+{
+	return dmevh->dso;
+}
+
+const char *dm_event_handler_get_dev_name(const struct dm_event_handler *dmevh)
+{
+	return dmevh->dev_name;
+}
+
+const char *dm_event_handler_get_uuid(const struct dm_event_handler *dmevh)
+{
+	return dmevh->uuid;
+}
+
+int dm_event_handler_get_major(const struct dm_event_handler *dmevh)
+{
+	return dmevh->major;
+}
+
+int dm_event_handler_get_minor(const struct dm_event_handler *dmevh)
+{
+	return dmevh->minor;
+}
+
+int dm_event_handler_get_timeout(const struct dm_event_handler *dmevh)
+{
+	return dmevh->timeout;
+}
+
+enum dm_event_mask dm_event_handler_get_event_mask(const struct dm_event_handler *dmevh)
+{
+	return dmevh->mask;
+}
+
+static int _check_message_id(struct dm_event_daemon_message *msg)
+{
+	int pid, seq_nr;
+
+	if ((sscanf(msg->data, "%d:%d", &pid, &seq_nr) != 2) ||
+	    (pid != getpid()) || (seq_nr != _sequence_nr)) {
+		log_error("Ignoring out-of-sequence reply from dmeventd. "
+			  "Expected %d:%d but received %s", getpid(),
+			  _sequence_nr, msg->data);
+		return 0;
+	}
+
+	return 1;
+}
+
+/*
+ * daemon_read
+ * @fifos
+ * @msg
+ *
+ * Read message from daemon.
+ *
+ * Returns: 0 on failure, 1 on success
+ */
+static int _daemon_read(struct dm_event_fifos *fifos,
+			struct dm_event_daemon_message *msg)
+{
+	unsigned bytes = 0;
+	int ret, i;
+	fd_set fds;
+	struct timeval tval = { 0, 0 };
+	size_t size = 2 * sizeof(uint32_t);	/* status + size */
+	char *buf = alloca(size);
+	int header = 1;
+
+	while (bytes < size) {
+		for (i = 0, ret = 0; (i < 20) && (ret < 1); i++) {
+			/* Watch daemon read FIFO for input. */
+			FD_ZERO(&fds);
+			FD_SET(fifos->server, &fds);
+			tval.tv_sec = 1;
+			ret = select(fifos->server + 1, &fds, NULL, NULL,
+				     &tval);
+			if (ret < 0 && errno != EINTR) {
+				log_error("Unable to read from event server");
+				return 0;
+			}
+		}
+		if (ret < 1) {
+			log_error("Unable to read from event server.");
+			return 0;
+		}
+
+		ret = read(fifos->server, buf + bytes, size);
+		if (ret < 0) {
+			if ((errno == EINTR) || (errno == EAGAIN))
+				continue;
+			else {
+				log_error("Unable to read from event server.");
+				return 0;
+			}
+		}
+
+		bytes += ret;
+		if (bytes == 2 * sizeof(uint32_t) && header) {
+			msg->cmd = ntohl(*((uint32_t *)buf));
+			msg->size = ntohl(*((uint32_t *)buf + 1));
+			buf = msg->data = dm_malloc(msg->size);
+			size = msg->size;
+			bytes = 0;
+			header = 0;
+		}
+	}
+
+	if (bytes != size) {
+		if (msg->data)
+			dm_free(msg->data);
+		msg->data = NULL;
+	}
+
+	return bytes == size;
+}
+
+/* Write message to daemon. */
+static int _daemon_write(struct dm_event_fifos *fifos,
+			 struct dm_event_daemon_message *msg)
+{
+	unsigned bytes = 0;
+	int ret = 0;
+	fd_set fds;
+
+	size_t size = 2 * sizeof(uint32_t) + msg->size;
+	char *buf = alloca(size);
+	char drainbuf[128];
+	struct timeval tval = { 0, 0 };
+
+	*((uint32_t *)buf) = htonl(msg->cmd);
+	*((uint32_t *)buf + 1) = htonl(msg->size);
+	memcpy(buf + 2 * sizeof(uint32_t), msg->data, msg->size);
+
+	/* drain the answer fifo */
+	while (1) {
+		FD_ZERO(&fds);
+		FD_SET(fifos->server, &fds);
+		tval.tv_usec = 100;
+		ret = select(fifos->server + 1, &fds, NULL, NULL, &tval);
+		if ((ret < 0) && (errno != EINTR)) {
+			log_error("Unable to talk to event daemon");
+			return 0;
+		}
+		if (ret == 0)
+			break;
+		read(fifos->server, drainbuf, 127);
+	}
+
+	while (bytes < size) {
+		do {
+			/* Watch daemon write FIFO to be ready for output. */
+			FD_ZERO(&fds);
+			FD_SET(fifos->client, &fds);
+			ret = select(fifos->client + 1, NULL, &fds, NULL, NULL);
+			if ((ret < 0) && (errno != EINTR)) {
+				log_error("Unable to talk to event daemon");
+				return 0;
+			}
+		} while (ret < 1);
+
+		ret = write(fifos->client, ((char *) buf) + bytes,
+			    size - bytes);
+		if (ret < 0) {
+			if ((errno == EINTR) || (errno == EAGAIN))
+				continue;
+			else {
+				log_error("Unable to talk to event daemon");
+				return 0;
+			}
+		}
+
+		bytes += ret;
+	}
+
+	return bytes == size;
+}
+
+static int _daemon_talk(struct dm_event_fifos *fifos,
+			struct dm_event_daemon_message *msg, int cmd,
+			const char *dso_name, const char *dev_name,
+			enum dm_event_mask evmask, uint32_t timeout)
+{
+	const char *dso = dso_name ? dso_name : "";
+	const char *dev = dev_name ? dev_name : "";
+	const char *fmt = "%d:%d %s %s %u %" PRIu32;
+	int msg_size;
+	memset(msg, 0, sizeof(*msg));
+
+	/*
+	 * Set command and pack the arguments
+	 * into ASCII message string.
+	 */
+	msg->cmd = cmd;
+	if (cmd == DM_EVENT_CMD_HELLO)
+		fmt = "%d:%d HELLO";
+	if ((msg_size = dm_asprintf(&(msg->data), fmt, getpid(), _sequence_nr,
+				    dso, dev, evmask, timeout)) < 0) {
+		log_error("_daemon_talk: message allocation failed");
+		return -ENOMEM;
+	}
+	msg->size = msg_size;
+
+	/*
+	 * Write command and message to and
+	 * read status return code from daemon.
+	 */
+	if (!_daemon_write(fifos, msg)) {
+		stack;
+		dm_free(msg->data);
+		msg->data = 0;
+		return -EIO;
+	}
+
+	do {
+
+		if (msg->data)
+			dm_free(msg->data);
+		msg->data = 0;
+
+		if (!_daemon_read(fifos, msg)) {
+			stack;
+			return -EIO;
+		}
+	} while (!_check_message_id(msg));
+
+	_sequence_nr++;
+
+	return (int32_t) msg->cmd;
+}
+
+/*
+ * start_daemon
+ *
+ * This function forks off a process (dmeventd) that will handle
+ * the events.  I am currently test opening one of the fifos to
+ * ensure that the daemon is running and listening...  I thought
+ * this would be less expensive than fork/exec'ing every time.
+ * Perhaps there is an even quicker/better way (no, checking the
+ * lock file is _not_ a better way).
+ *
+ * Returns: 1 on success, 0 otherwise
+ */
+static int _start_daemon(struct dm_event_fifos *fifos)
+{
+	int pid, ret = 0;
+	int status;
+	struct stat statbuf;
+
+	if (stat(fifos->client_path, &statbuf))
+		goto start_server;
+
+	if (!S_ISFIFO(statbuf.st_mode)) {
+		log_error("%s is not a fifo.", fifos->client_path);
+		return 0;
+	}
+
+	/* Anyone listening?  If not, errno will be ENXIO */
+	fifos->client = open(fifos->client_path, O_WRONLY | O_NONBLOCK);
+	if (fifos->client >= 0) {
+		/* server is running and listening */
+
+		close(fifos->client);
+		return 1;
+	} else if (errno != ENXIO) {
+		/* problem */
+
+		log_error("%s: Can't open client fifo %s: %s",
+			  __func__, fifos->client_path, strerror(errno));
+		stack;
+		return 0;
+	}
+
+      start_server:
+	/* server is not running */
+	pid = fork();
+
+	if (pid < 0)
+		log_error("Unable to fork.");
+
+	else if (!pid) {
+		execvp(DMEVENTD_PATH, NULL);
+		exit(EXIT_FAILURE);
+	} else {
+		if (waitpid(pid, &status, 0) < 0)
+			log_error("Unable to start dmeventd: %s",
+				  strerror(errno));
+		else if (WEXITSTATUS(status))
+			log_error("Unable to start dmeventd.");
+		else
+			ret = 1;
+	}
+
+	return ret;
+}
+
+/* Initialize client. */
+static int _init_client(struct dm_event_fifos *fifos)
+{
+	/* FIXME? Is fifo the most suitable method? Why not share
+	   comms/daemon code with something else e.g. multipath? */
+
+	/* init fifos */
+	memset(fifos, 0, sizeof(*fifos));
+	fifos->client_path = DM_EVENT_FIFO_CLIENT;
+	fifos->server_path = DM_EVENT_FIFO_SERVER;
+
+	if (!_start_daemon(fifos)) {
+		stack;
+		return 0;
+	}
+
+	/* Open the fifo used to read from the daemon. */
+	if ((fifos->server = open(fifos->server_path, O_RDWR)) < 0) {
+		log_error("%s: open server fifo %s",
+			  __func__, fifos->server_path);
+		stack;
+		return 0;
+	}
+
+	/* Lock out anyone else trying to do communication with the daemon. */
+	if (flock(fifos->server, LOCK_EX) < 0) {
+		log_error("%s: flock %s", __func__, fifos->server_path);
+		close(fifos->server);
+		return 0;
+	}
+
+/*	if ((fifos->client = open(fifos->client_path, O_WRONLY | O_NONBLOCK)) < 0) {*/
+	if ((fifos->client = open(fifos->client_path, O_RDWR | O_NONBLOCK)) < 0) {
+		log_error("%s: Can't open client fifo %s: %s",
+			  __func__, fifos->client_path, strerror(errno));
+		close(fifos->server);
+		stack;
+		return 0;
+	}
+
+	return 1;
+}
+
+static void _dtr_client(struct dm_event_fifos *fifos)
+{
+	if (flock(fifos->server, LOCK_UN))
+		log_error("flock unlock %s", fifos->server_path);
+
+	close(fifos->client);
+	close(fifos->server);
+}
+
+/* Get uuid of a device */
+static struct dm_task *_get_device_info(const struct dm_event_handler *dmevh)
+{
+	struct dm_task *dmt;
+	struct dm_info info;
+
+	if (!(dmt = dm_task_create(DM_DEVICE_INFO))) {
+		log_error("_get_device_info: dm_task creation for info failed");
+		return NULL;
+	}
+
+	if (dmevh->uuid)
+		dm_task_set_uuid(dmt, dmevh->uuid);
+	else if (dmevh->dev_name)
+		dm_task_set_name(dmt, dmevh->dev_name);
+	else if (dmevh->major && dmevh->minor) {
+		dm_task_set_major(dmt, dmevh->major);
+		dm_task_set_minor(dmt, dmevh->minor);
+        }
+
+	/* FIXME Add name or uuid or devno to messages */
+	if (!dm_task_run(dmt)) {
+		log_error("_get_device_info: dm_task_run() failed");
+		goto failed;
+	}
+
+	if (!dm_task_get_info(dmt, &info)) {
+		log_error("_get_device_info: failed to get info for device");
+		goto failed;
+	}
+
+	if (!info.exists) {
+		log_error("_get_device_info: device not found");
+		goto failed;
+	}
+
+	return dmt;
+
+failed:
+	dm_task_destroy(dmt);
+	return NULL;
+}
+
+/* Handle the event (de)registration call and return negative error codes. */
+static int _do_event(int cmd, struct dm_event_daemon_message *msg,
+		     const char *dso_name, const char *dev_name,
+		     enum dm_event_mask evmask, uint32_t timeout)
+{
+	int ret;
+	struct dm_event_fifos fifos;
+
+	if (!_init_client(&fifos)) {
+		stack;
+		return -ESRCH;
+	}
+
+	ret = _daemon_talk(&fifos, msg, DM_EVENT_CMD_HELLO, 0, 0, 0, 0);
+
+	if (msg->data)
+		dm_free(msg->data);
+	msg->data = 0;
+
+	if (!ret)
+		ret = _daemon_talk(&fifos, msg, cmd, dso_name, dev_name, evmask, timeout);
+
+	/* what is the opposite of init? */
+	_dtr_client(&fifos);
+
+	return ret;
+}
+
+/* External library interface. */
+int dm_event_register_handler(const struct dm_event_handler *dmevh)
+{
+	int ret = 1, err;
+	const char *uuid;
+	struct dm_task *dmt;
+	struct dm_event_daemon_message msg = { 0, 0, NULL };
+
+	if (!(dmt = _get_device_info(dmevh))) {
+		stack;
+		return 0;
+	}
+
+	uuid = dm_task_get_uuid(dmt);
+
+	if ((err = _do_event(DM_EVENT_CMD_REGISTER_FOR_EVENT, &msg,
+			     dmevh->dso, uuid, dmevh->mask, dmevh->timeout)) < 0) {
+		log_error("%s: event registration failed: %s",
+			  dm_task_get_name(dmt),
+			  msg.data ? msg.data : strerror(-err));
+		ret = 0;
+	}
+
+	if (msg.data)
+		dm_free(msg.data);
+
+	dm_task_destroy(dmt);
+
+	return ret;
+}
+
+int dm_event_unregister_handler(const struct dm_event_handler *dmevh)
+{
+	int ret = 1, err;
+	const char *uuid;
+	struct dm_task *dmt;
+	struct dm_event_daemon_message msg = { 0, 0, NULL };
+
+	if (!(dmt = _get_device_info(dmevh))) {
+		stack;
+		return 0;
+	}
+
+	uuid = dm_task_get_uuid(dmt);
+
+	if ((err = _do_event(DM_EVENT_CMD_UNREGISTER_FOR_EVENT, &msg,
+			    dmevh->dso, uuid, dmevh->mask, dmevh->timeout)) < 0) {
+		log_error("%s: event deregistration failed: %s",
+			  dm_task_get_name(dmt),
+			  msg.data ? msg.data : strerror(-err));
+		ret = 0;
+	}
+
+	if (msg.data)
+		dm_free(msg.data);
+
+	dm_task_destroy(dmt);
+
+	return ret;
+}
+
+/* Fetch a string off src and duplicate it into *dest. */
+/* FIXME: move to separate module to share with the daemon. */
+static char *_fetch_string(char **src, const int delimiter)
+{
+	char *p, *ret;
+
+	if ((p = strchr(*src, delimiter)))
+		*p = 0;
+
+	if ((ret = dm_strdup(*src)))
+		*src += strlen(ret) + 1;
+
+	if (p)
+		*p = delimiter;
+
+	return ret;
+}
+
+/* Parse a device message from the daemon. */
+static int _parse_message(struct dm_event_daemon_message *msg, char **dso_name,
+			 char **uuid, enum dm_event_mask *evmask)
+{
+	char *id = NULL;
+	char *p = msg->data;
+
+	if ((id = _fetch_string(&p, ' ')) &&
+	    (*dso_name = _fetch_string(&p, ' ')) &&
+	    (*uuid = _fetch_string(&p, ' '))) {
+		*evmask = atoi(p);
+
+		dm_free(id);
+		return 0;
+	}
+
+	if (id)
+		dm_free(id);
+	return -ENOMEM;
+}
+
+/*
+ * Returns 0 if handler found; error (-ENOMEM, -ENOENT) otherwise.
+ */
+int dm_event_get_registered_device(struct dm_event_handler *dmevh, int next)
+{
+	int ret = 0;
+	const char *uuid = NULL;
+	char *reply_dso = NULL, *reply_uuid = NULL;
+	enum dm_event_mask reply_mask = 0;
+	struct dm_task *dmt = NULL;
+	struct dm_event_daemon_message msg = { 0, 0, NULL };
+
+	if (!(dmt = _get_device_info(dmevh))) {
+		stack;
+		return 0;
+	}
+
+	uuid = dm_task_get_uuid(dmt);
+
+	if (!(ret = _do_event(next ? DM_EVENT_CMD_GET_NEXT_REGISTERED_DEVICE :
+			     DM_EVENT_CMD_GET_REGISTERED_DEVICE,
+			      &msg, dmevh->dso, uuid, dmevh->mask, 0))) {
+		/* FIXME this will probably horribly break if we get
+		   ill-formatted reply */
+		ret = _parse_message(&msg, &reply_dso, &reply_uuid, &reply_mask);
+	} else {
+		ret = -ENOENT;
+		goto fail;
+	}
+
+	dm_task_destroy(dmt);
+	dmt = NULL;
+
+	if (msg.data) {
+		dm_free(msg.data);
+		msg.data = NULL;
+	}
+
+	_dm_event_handler_clear_dev_info(dmevh);
+	dmevh->uuid = dm_strdup(reply_uuid);
+	if (!dmevh->uuid) {
+		ret = -ENOMEM;
+		goto fail;
+	}
+
+	if (!(dmt = _get_device_info(dmevh))) {
+		ret = -ENXIO; /* dmeventd probably gave us bogus uuid back */
+		goto fail;
+	}
+
+	dm_event_handler_set_dso(dmevh, reply_dso);
+	dm_event_handler_set_event_mask(dmevh, reply_mask);
+
+	if (reply_dso) {
+		dm_free(reply_dso);
+		reply_dso = NULL;
+	}
+
+	if (reply_uuid) {
+		dm_free(reply_uuid);
+		reply_uuid = NULL;
+	}
+
+	dmevh->dev_name = dm_strdup(dm_task_get_name(dmt));
+	if (!dmevh->dev_name) {
+		ret = -ENOMEM;
+		goto fail;
+	}
+
+	struct dm_info info;
+	if (!dm_task_get_info(dmt, &info)) {
+		ret = -1;
+		goto fail;
+	}
+
+	dmevh->major = info.major;
+	dmevh->minor = info.minor;
+
+	dm_task_destroy(dmt);
+
+	return ret;
+
+ fail:
+	if (msg.data)
+		dm_free(msg.data);
+	if (reply_dso)
+		dm_free(reply_dso);
+	if (reply_uuid)
+		dm_free(reply_uuid);
+	_dm_event_handler_clear_dev_info(dmevh);
+	if (dmt)
+		dm_task_destroy(dmt);
+	return ret;
+}
+
+#if 0				/* left out for now */
+
+static char *_skip_string(char *src, const int delimiter)
+{
+	src = srtchr(src, delimiter);
+	if (src && *(src + 1))
+		return src + 1;
+	return NULL;
+}
+
+int dm_event_set_timeout(const char *device_path, uint32_t timeout)
+{
+	struct dm_event_daemon_message msg = { 0, 0, NULL };
+
+	if (!device_exists(device_path))
+		return -ENODEV;
+
+	return _do_event(DM_EVENT_CMD_SET_TIMEOUT, &msg,
+			 NULL, device_path, 0, timeout);
+}
+
+int dm_event_get_timeout(const char *device_path, uint32_t *timeout)
+{
+	int ret;
+	struct dm_event_daemon_message msg = { 0, 0, NULL };
+
+	if (!device_exists(device_path))
+		return -ENODEV;
+	if (!(ret = _do_event(DM_EVENT_CMD_GET_TIMEOUT, &msg, NULL, device_path,
+			     0, 0))) {
+		char *p = _skip_string(msg.data, ' ');
+		if (!p) {
+			log_error("malformed reply from dmeventd '%s'\n",
+				  msg.data);
+			return -EIO;
+		}
+		*timeout = atoi(p);
+	}
+	if (msg.data)
+		dm_free(msg.data);
+	return ret;
+}
+#endif

daemons/dmeventd/libdevmapper-event.h

@@ -0,0 +1,106 @@
+/*
+ * Copyright (C) 2005-2007 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of the device-mapper userspace tools.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+/*
+ * Note that this file is released only as part of a technology preview
+ * and its contents may change in future updates in ways that do not
+ * preserve compatibility.
+ */
+
+#ifndef LIB_DMEVENT_H
+#define LIB_DMEVENT_H
+
+#include <stdint.h>
+
+/*
+ * Event library interface.
+ */
+
+enum dm_event_mask {
+	DM_EVENT_SETTINGS_MASK  = 0x0000FF,
+	DM_EVENT_SINGLE		= 0x000001, /* Report multiple errors just once. */
+	DM_EVENT_MULTI		= 0x000002, /* Report all of them. */
+
+	DM_EVENT_ERROR_MASK     = 0x00FF00,
+	DM_EVENT_SECTOR_ERROR	= 0x000100, /* Failure on a particular sector. */
+	DM_EVENT_DEVICE_ERROR	= 0x000200, /* Device failure. */
+	DM_EVENT_PATH_ERROR	= 0x000400, /* Failure on an io path. */
+	DM_EVENT_ADAPTOR_ERROR	= 0x000800, /* Failure of a host adaptor. */
+
+	DM_EVENT_STATUS_MASK    = 0xFF0000,
+	DM_EVENT_SYNC_STATUS	= 0x010000, /* Mirror synchronization completed/failed. */
+	DM_EVENT_TIMEOUT	= 0x020000, /* Timeout has occured */
+
+	DM_EVENT_REGISTRATION_PENDING = 0x1000000, /* Monitor thread is setting-up/shutting-down */
+};
+
+#define DM_EVENT_ALL_ERRORS DM_EVENT_ERROR_MASK
+
+struct dm_event_handler;
+
+struct dm_event_handler *dm_event_handler_create(void);
+void dm_event_handler_destroy(struct dm_event_handler *dmevh);
+
+/*
+ * Path of shared library to handle events.
+ *
+ * All of dso, device_name and uuid strings are duplicated, you do not
+ * need to keep the pointers valid after the call succeeds. Thes may
+ * return -ENOMEM though.
+ */
+int dm_event_handler_set_dso(struct dm_event_handler *dmevh, const char *path);
+
+/*
+ * Identify the device to monitor by exactly one of device_name, uuid or
+ * device number. String arguments are duplicated, see above.
+ */
+int dm_event_handler_set_dev_name(struct dm_event_handler *dmevh, const char *device_name);
+
+int dm_event_handler_set_uuid(struct dm_event_handler *dmevh, const char *uuid);
+
+void dm_event_handler_set_major(struct dm_event_handler *dmevh, int major);
+void dm_event_handler_set_minor(struct dm_event_handler *dmevh, int minor);
+void dm_event_handler_set_timeout(struct dm_event_handler *dmevh, int timeout);
+
+/*
+ * Specify mask for events to monitor.
+ */
+void dm_event_handler_set_event_mask(struct dm_event_handler *dmevh,
+				     enum dm_event_mask evmask);
+
+const char *dm_event_handler_get_dso(const struct dm_event_handler *dmevh);
+const char *dm_event_handler_get_dev_name(const struct dm_event_handler *dmevh);
+const char *dm_event_handler_get_uuid(const struct dm_event_handler *dmevh);
+int dm_event_handler_get_major(const struct dm_event_handler *dmevh);
+int dm_event_handler_get_minor(const struct dm_event_handler *dmevh);
+int dm_event_handler_get_timeout(const struct dm_event_handler *dmevh);
+enum dm_event_mask dm_event_handler_get_event_mask(const struct dm_event_handler *dmevh);
+
+/* FIXME Review interface (what about this next thing?) */
+int dm_event_get_registered_device(struct dm_event_handler *dmevh, int next);
+
+/*
+ * Initiate monitoring using dmeventd.
+ */
+int dm_event_register_handler(const struct dm_event_handler *dmevh);
+int dm_event_unregister_handler(const struct dm_event_handler *dmevh);
+
+/* Prototypes for DSO interface, see dmeventd.c, struct dso_data for
+   detailed descriptions. */
+void process_event(struct dm_task *dmt, enum dm_event_mask evmask, void **user);
+int register_device(const char *device_name, const char *uuid, int major, int minor, void **user);
+int unregister_device(const char *device_name, const char *uuid, int major,
+		      int minor, void **user);
+
+#endif

daemons/dmeventd/libdevmapper-event.pc.in

@@ -0,0 +1,12 @@
+prefix=@prefix@
+exec_prefix=@exec_prefix@
+libdir=@libdir@
+includedir=@includedir@
+
+Name: devmapper-event
+Description: device-mapper event library
+Version: @DM_LIB_VERSION@
+Requires: devmapper
+Cflags: -I${includedir}
+Libs: -L${libdir} -ldevmapper-event
+Libs.private: -lpthread -ldl

daemons/dmeventd/plugins/Makefile.in

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

daemons/dmeventd/plugins/mirror/.exported_symbols

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

daemons/dmeventd/plugins/mirror/Makefile.in

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

daemons/dmeventd/plugins/mirror/dmeventd_mirror.c

@@ -0,0 +1,281 @@
+/*
+ * Copyright (C) 2005 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include "lvm2cmd.h"
+
+#include <libdevmapper.h>
+#include <libdevmapper-event.h>
+#include <errno.h>
+#include <signal.h>
+#include <string.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <pthread.h>
+#include <unistd.h>
+
+#include <syslog.h> /* FIXME Replace syslog with multilog */
+/* FIXME Missing openlog? */
+
+#define ME_IGNORE    0
+#define ME_INSYNC    1
+#define ME_FAILURE   2
+
+/*
+ * register_device() is called first and performs initialisation.
+ * Only one device may be registered or unregistered at a time.
+ */
+static pthread_mutex_t _register_mutex = PTHREAD_MUTEX_INITIALIZER;
+
+/*
+ * Number of active registrations.
+ */
+static int _register_count = 0;
+
+static struct dm_pool *_mem_pool = NULL;
+static void *_lvm_handle = NULL;
+
+/*
+ * Currently only one event can be processed at a time.
+ */
+static pthread_mutex_t _event_mutex = PTHREAD_MUTEX_INITIALIZER;
+
+static int _get_mirror_event(char *params)
+{
+	int i, r = ME_INSYNC;
+	char **args = NULL;
+	char *dev_status_str;
+	char *log_status_str;
+	char *sync_str;
+	char *p = NULL;
+	int log_argc, num_devs;
+
+	/*
+	 * dm core parms:	     0 409600 mirror
+	 * Mirror core parms:	     2 253:4 253:5 400/400
+	 * New-style failure params: 1 AA
+	 * New-style log params:     3 cluster 253:3 A
+	 *			 or  3 disk 253:3 A
+	 *			 or  1 core
+	 */
+
+	/* number of devices */
+	if (!dm_split_words(params, 1, 0, &p))
+		goto out_parse;
+
+	if (!(num_devs = atoi(p)))
+		goto out_parse;
+	p += strlen(p) + 1;
+
+	/* devices names + "400/400" + "1 AA" + 1 or 3 log parms + NULL */
+	args = dm_malloc((num_devs + 7) * sizeof(char *));
+	if (!args || dm_split_words(p, num_devs + 7, 0, args) < num_devs + 5)
+		goto out_parse;
+
+	dev_status_str = args[2 + num_devs];
+	log_argc = atoi(args[3 + num_devs]);
+	log_status_str = args[3 + num_devs + log_argc];
+	sync_str = args[num_devs];
+
+	/* Check for bad mirror devices */
+	for (i = 0; i < num_devs; i++)
+		if (dev_status_str[i] == 'D') {
+			syslog(LOG_ERR, "Mirror device, %s, has failed.\n", args[i]);
+			r = ME_FAILURE;
+		}
+
+	/* Check for bad disk log device */
+	if (log_argc > 1 && log_status_str[0] == 'D') {
+		syslog(LOG_ERR, "Log device, %s, has failed.\n",
+		       args[2 + num_devs + log_argc]);
+		r = ME_FAILURE;
+	}
+
+	if (r == ME_FAILURE)
+		goto out;
+
+	p = strstr(sync_str, "/");
+	if (p) {
+		p[0] = '\0';
+		if (strcmp(sync_str, p+1))
+			r = ME_IGNORE;
+		p[0] = '/';
+	} else
+		goto out_parse;
+
+out:
+	if (args)
+		dm_free(args);
+	return r;
+	
+out_parse:
+	if (args)
+		dm_free(args);
+	syslog(LOG_ERR, "Unable to parse mirror status string.");
+	return ME_IGNORE;
+}
+
+static void _temporary_log_fn(int level, const char *file,
+			      int line, const char *format)
+{
+	if (!strncmp(format, "WARNING: ", 9) && (level < 5))
+		syslog(LOG_CRIT, "%s", format);
+	else
+		syslog(LOG_DEBUG, "%s", format);
+}
+
+static int _remove_failed_devices(const char *device)
+{
+	int r;
+#define CMD_SIZE 256	/* FIXME Use system restriction */
+	char cmd_str[CMD_SIZE];
+	char *vg = NULL, *lv = NULL, *layer = NULL;
+
+	if (strlen(device) > 200)  /* FIXME Use real restriction */
+		return -ENAMETOOLONG;	/* FIXME These return code distinctions are not used so remove them! */
+
+	if (!dm_split_lvm_name(_mem_pool, device, &vg, &lv, &layer)) {
+		syslog(LOG_ERR, "Unable to determine VG name from %s",
+		       device);
+		return -ENOMEM;	/* FIXME Replace with generic error return - reason for failure has already got logged */
+	}
+
+	/* FIXME Is any sanity-checking required on %s? */
+	if (CMD_SIZE <= snprintf(cmd_str, CMD_SIZE, "vgreduce --config devices{ignore_suspended_devices=1} --removemissing %s", vg)) {
+		/* this error should be caught above, but doesn't hurt to check again */
+		syslog(LOG_ERR, "Unable to form LVM command: Device name too long");
+		dm_pool_empty(_mem_pool);  /* FIXME: not safe with multiple threads */
+		return -ENAMETOOLONG; /* FIXME Replace with generic error return - reason for failure has already got logged */
+	}
+
+	r = lvm2_run(_lvm_handle, cmd_str);
+
+	dm_pool_empty(_mem_pool);  /* FIXME: not safe with multiple threads */
+	return (r == 1) ? 0 : -1;
+}
+
+void process_event(struct dm_task *dmt, enum dm_event_mask event,
+		   void **unused __attribute((unused)))
+{
+	void *next = NULL;
+	uint64_t start, length;
+	char *target_type = NULL;
+	char *params;
+	const char *device = dm_task_get_name(dmt);
+
+	if (pthread_mutex_trylock(&_event_mutex)) {
+		syslog(LOG_NOTICE, "Another thread is handling an event.  Waiting...");
+		pthread_mutex_lock(&_event_mutex);
+	}
+	do {
+		next = dm_get_next_target(dmt, next, &start, &length,
+					  &target_type, &params);
+
+		if (!target_type) {
+			syslog(LOG_INFO, "%s mapping lost.\n", device);
+			continue;
+		}
+
+		if (strcmp(target_type, "mirror")) {
+			syslog(LOG_INFO, "%s has unmirrored portion.\n", device);
+			continue;
+		}
+
+		switch(_get_mirror_event(params)) {
+		case ME_INSYNC:
+			/* FIXME: all we really know is that this
+			   _part_ of the device is in sync
+			   Also, this is not an error
+			*/
+			syslog(LOG_NOTICE, "%s is now in-sync\n", device);
+			break;
+		case ME_FAILURE:
+			syslog(LOG_ERR, "Device failure in %s\n", device);
+			if (_remove_failed_devices(device))
+				/* FIXME Why are all the error return codes unused? Get rid of them? */
+				syslog(LOG_ERR, "Failed to remove faulty devices in %s\n",
+				       device);
+			/* Should check before warning user that device is now linear
+			else
+				syslog(LOG_NOTICE, "%s is now a linear device.\n",
+					device);
+			*/
+			break;
+		case ME_IGNORE:
+			break;
+		default:
+			/* FIXME Provide value then! */
+			syslog(LOG_INFO, "Unknown event received.\n");
+		}
+	} while (next);
+
+	pthread_mutex_unlock(&_event_mutex);
+}
+
+int register_device(const char *device, const char *uuid, int major, int minor,
+		   void **unused __attribute((unused)))
+{
+	int r = 0;
+
+	pthread_mutex_lock(&_register_mutex);
+
+	syslog(LOG_INFO, "Monitoring mirror device %s for events\n", device);
+
+	/*
+	 * Need some space for allocations.  1024 should be more
+	 * than enough for what we need (device mapper name splitting)
+	 */
+	if (!_mem_pool && !(_mem_pool = dm_pool_create("mirror_dso", 1024)))
+		goto out;
+
+	if (!_lvm_handle) {
+		lvm2_log_fn(_temporary_log_fn);
+		if (!(_lvm_handle = lvm2_init())) {
+			dm_pool_destroy(_mem_pool);
+			_mem_pool = NULL;
+			goto out;
+		}
+		lvm2_log_level(_lvm_handle, LVM2_LOG_SUPPRESS);
+		/* FIXME Temporary: move to dmeventd core */
+		lvm2_run(_lvm_handle, "_memlock_inc");
+	}
+
+	_register_count++;
+	r = 1;
+
+out:
+	pthread_mutex_unlock(&_register_mutex);
+
+	return r;
+}
+
+int unregister_device(const char *device, const char *uuid, int major, int minor,
+		   void **unused __attribute((unused)))
+{
+	pthread_mutex_lock(&_register_mutex);
+
+	syslog(LOG_INFO, "No longer monitoring mirror device %s for events\n",
+	       device);
+
+	if (!--_register_count) {
+		dm_pool_destroy(_mem_pool);
+		_mem_pool = NULL;
+		lvm2_run(_lvm_handle, "_memlock_dec");
+		lvm2_exit(_lvm_handle);
+		_lvm_handle = NULL;
+	}
+
+	pthread_mutex_unlock(&_register_mutex);
+
+	return 1;
+}

debian/README.Debian

@@ -0,0 +1,6 @@
+LVM2 requires the device-mapper kernel module (dm-mod).  This is
+available as a kernel patch for 2.4 (in kernel-patch-device-mapper), and
+is distributed with linux 2.5 and above.  The LVM1 kernel module (lvm-mod)
+will not work with lvm2 packages.  dm-mod and lvm-mod may both be loaded
+in the kernel at the same time with no problems.  Without dm-mod, this
+package is pretty useless.

debian/changelog

@@ -1,3 +1,56 @@
+lvm2 (1.95.15-1) unstable; urgency=low
+  
+  * New upstream release.
+  * Remove undocumented manpage symlinks.
+  * Update description to be more informative.  (Closes: #173499)
+  * Add kernel-patch-device-mapper suggestion.
+  * Update standards version.
+
+ -- Andres Salomon <dilinger@mp3revolution.net>  Sun, 16 Feb 2002 04:21:26 -0400
+
+lvm2 (1.95.11-1) unstable; urgency=low
+
+  * New upstream release.  (Closes: #171436)
+  * Removed TODO and INTRO from debian/docs; added WHATS_NEW.
+  * Remove vgcfgrestore.8 undocumented symlink.
+  * Added a README.Debian, mentioning the device-mapper kernel module
+    requirement that lvm2 has.  (Closes: #171674, #163020)
+  * Get rid of debian/conffiles (debhelper's smart enough to figure that out).
+  * debian/copyright fix to appease lintian.
+  * Fix typo in tools/commands.h that caused /usr/sbin/; to be created.
+
+ -- Andres Salomon <dilinger@mp3revolution.net>  Mon,  9 Dec 2002 02:51:02 -0400
+
+lvm2 (1.95.10-2) unstable; urgency=low
+
+  * Fix software raid problems by ensuring lvm init script runs after
+    raidtools init script.  (Closes: #152569)
+
+ -- Andres Salomon <dilinger@mp3revolution.net>  Tue,  3 Sep 2002 04:05:43 -0400
+
+lvm2 (1.95.10-1) unstable; urgency=low
+
+  * New upstream release (Beta 3.2).
+  * Change all references to /dev/device-mapper/control to
+    /dev/mapper/control.
+
+ -- Andres Salomon <dilinger@mp3revolution.net>  Sun,  1 Sep 2002 18:55:12 -0400
+
+lvm2 (0.95.05-3) unstable; urgency=low
+
+  * Get rid of awk dependency in init script.  (Closes: #146257)
+
+ -- Andres Salomon <dilinger@mp3revolution.net>  Sun, 12 May 2002 04:39:06 -0500
+
+lvm2 (0.95.05-2) unstable; urgency=low
+
+  * Use ${shlibs:Depends} in Depends.
+  * Get rid of postinst/postrm scripts, use debhelper's init script instead.
+  * Add Conflicts against lvm10, lvm-common.
+  * Fix endian issues on big-endian machines.
+
+ -- Andres Salomon <dilinger@mp3revolution.net>  Thu,  2 May 2002 23:53:53 -0500
+
 lvm2 (0.95.05-1) unstable; urgency=low
 
   * New release (Beta2).

debian/conffiles

@@ -1,2 +0,0 @@
-/etc/lvm/lvm.conf
-/etc/init.d/lvm2

debian/control

@@ -2,18 +2,23 @@ Source: lvm2
 Section: admin
 Priority: optional
 Maintainer: Andres Salomon <dilinger@mp3revolution.net>
-Build-Depends: debhelper (>> 3.0.0), libdevmapper-dev, libreadline4-dev
-Standards-Version: 3.5.2
+Build-Depends: debhelper (>> 3.0.0), libdevmapper-dev (>= 0.96.04), libreadline4-dev
+Standards-Version: 3.5.8.0
 
 Package: lvm2
 Architecture: any
-Depends: libdevmapper0
+Depends: ${shlibs:Depends}
+Conflicts: lvm10, lvm-common
 Replaces: lvm10, lvm-common
 Provides: lvm-binaries
-Suggests: dmsetup
+Suggests: dmsetup, kernel-patch-device-mapper
 Description: The Linux Logical Volume Manager
  This is LVM2, the rewrite of The Linux Logical Volume Manager.  LVM
  supports enterprise level volume management of disk and disk subsystems
  by grouping arbitrary disks into volume groups. The total capacity of
  volume groups can be allocated to logical volumes, which are accessed as
  regular block devices.
+ .
+ LVM2 is currently stable, but has some unimplemented features (most notably,
+ pvmove and e2fsadm).  It is not yet recommended for production use.  It is
+ backwards-compatible with LVM1 (lvm10), and requires Linux kernel 2.4.

debian/copyright

@@ -3,7 +3,7 @@ Wed, 20 Feb 2002 03:17:25 -0500.
 
 It was downloaded from http://www.sistina.com/products_lvm.htm
 
-Upstream Author(s): LVM Development Team
+Upstream Author: LVM Development Team
 
 Copyright (c) 2001-2002 LVM Development Team
 

debian/docs

@@ -1,8 +1,5 @@
 BUGS
-INTRO
 README
-TODO
 VERSION
-doc/example.conf
-doc/pvmove_outline.txt
-doc/testing.txt
+WHATS_NEW
+doc/*

debian/init.d

@@ -14,10 +14,10 @@ modprobe dm-mod >/dev/null 2>&1
 
 # Create necessary files in /dev for device-mapper
 create_devfiles() {
-	DIR="/dev/device-mapper"
+	DIR="/dev/mapper"
 	FILE="$DIR/control"
-	major=$(awk '$2 ~ /^misc$/ {print $1}' /proc/devices)
-	minor=$(awk "\$2 ~ /^$DM_NAME\$/ {print \$1}" /proc/misc)
+	major=$(grep "[0-9] misc$" /proc/devices | sed 's/[ ]\+misc//')
+	minor=$(grep "[0-9] device-mapper$" /proc/misc | sed 's/[ ]\+device-mapper//')
 
 	if test ! -d $DIR; then
 		mkdir --mode=755 $DIR >/dev/null 2>&1
@@ -56,8 +56,7 @@ case "$1" in
 	echo "$NAME."
 	;;
   *)
-	N=/etc/init.d/$NAME
-	echo "Usage: $N {start|stop|restart|force-reload}" >&2
+	echo "Usage: /etc/init.d/$NAME {start|stop|restart|force-reload}" >&2
 	exit 1
 	;;
 esac

debian/postinst

@@ -1,49 +0,0 @@
-#! /bin/sh
-# postinst script for lvm2
-#
-# see: dh_installdeb(1)
-
-set -e
-
-# summary of how this script can be called:
-#        * <postinst> `configure' <most-recently-configured-version>
-#        * <old-postinst> `abort-upgrade' <new version>
-#        * <conflictor's-postinst> `abort-remove' `in-favour' <package>
-#          <new-version>
-#        * <deconfigured's-postinst> `abort-deconfigure' `in-favour'
-#          <failed-install-package> <version> `removing'
-#          <conflicting-package> <version>
-# for details, see http://www.debian.org/doc/debian-policy/ or
-# the debian-policy package
-#
-# quoting from the policy:
-#     Any necessary prompting should almost always be confined to the
-#     post-installation script, and should be protected with a conditional
-#     so that unnecessary prompting doesn't happen if a package's
-#     installation fails and the `postinst' is called with `abort-upgrade',
-#     `abort-remove' or `abort-deconfigure'.
-
-case "$1" in
-    configure)
-    	update-rc.d lvm2 start 25 S . start 50 0 6 . >/dev/null
-	/etc/init.d/lvm2 start
-    ;;
-
-    abort-upgrade|abort-remove|abort-deconfigure)
-
-    ;;
-
-    *)
-        echo "postinst called with unknown argument \`$1'" >&2
-        exit 1
-    ;;
-esac
-
-# dh_installdeb will replace this with shell code automatically
-# generated by other debhelper scripts.
-
-#DEBHELPER#
-
-exit 0
-
-

debian/postrm

@@ -1,39 +0,0 @@
-#! /bin/sh
-# postrm script for lvm2
-#
-# see: dh_installdeb(1)
-
-set -e
-
-# summary of how this script can be called:
-#        * <postrm> `remove'
-#        * <postrm> `purge'
-#        * <old-postrm> `upgrade' <new-version>
-#        * <new-postrm> `failed-upgrade' <old-version>
-#        * <new-postrm> `abort-install'
-#        * <new-postrm> `abort-install' <old-version>
-#        * <new-postrm> `abort-upgrade' <old-version>
-#        * <disappearer's-postrm> `disappear' <r>overwrit>r> <new-version>
-# for details, see http://www.debian.org/doc/debian-policy/ or
-# the debian-policy package
-
-
-case "$1" in
-	purge)
-		rm -f /etc/init.d/lvm2
-		update-rc.d lvm2 remove >/dev/null
-	;;
-	remove|upgrade|failed-upgrade|abort-install|abort-upgrade|disappear)
-	;;
-
-	*)
-		echo "postrm called with unknown argument \`$1'" >&2
-		exit 1
-esac
-
-# dh_installdeb will replace this with shell code automatically
-# generated by other debhelper scripts.
-
-#DEBHELPER#
-
-exit 0

debian/rules

@@ -98,11 +98,10 @@ binary-arch: build install
 #	dh_installemacsen -a
 #	dh_installpam -a
 #	dh_installmime -a
-	dh_installinit -n
+	dh_installinit --update-rcd-params="start 26 S . start 50 0 6 ."
 	dh_installcron
 	dh_installman
 	dh_installinfo 
-	dh_undocumented
 	dh_installchangelogs
 	dh_strip
 	dh_link

debian/undocumented

@@ -1,14 +0,0 @@
-e2fsadm.8
-lvmdiskscan.8
-lvmsadc.8
-lvmsar.8
-lvresize.8
-pvdata.8
-pvmove.8
-pvresize.8
-version.8
-vgcfgrestore.8
-vgexport.8
-vgimport.8
-vgmknodes.8
-vgsplit.8

dmeventd/Makefile.in

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

dmeventd/mirror/.exported_symbols

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

dmeventd/mirror/Makefile.in

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

dmeventd/mirror/dmeventd_mirror.c

@@ -0,0 +1,281 @@
+/*
+ * Copyright (C) 2005 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include "lvm2cmd.h"
+
+#include <libdevmapper.h>
+#include <libdevmapper-event.h>
+#include <errno.h>
+#include <signal.h>
+#include <string.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <pthread.h>
+#include <unistd.h>
+
+#include <syslog.h> /* FIXME Replace syslog with multilog */
+/* FIXME Missing openlog? */
+
+#define ME_IGNORE    0
+#define ME_INSYNC    1
+#define ME_FAILURE   2
+
+/*
+ * register_device() is called first and performs initialisation.
+ * Only one device may be registered or unregistered at a time.
+ */
+static pthread_mutex_t _register_mutex = PTHREAD_MUTEX_INITIALIZER;
+
+/*
+ * Number of active registrations.
+ */
+static int _register_count = 0;
+
+static struct dm_pool *_mem_pool = NULL;
+static void *_lvm_handle = NULL;
+
+/*
+ * Currently only one event can be processed at a time.
+ */
+static pthread_mutex_t _event_mutex = PTHREAD_MUTEX_INITIALIZER;
+
+static int _get_mirror_event(char *params)
+{
+	int i, r = ME_INSYNC;
+	char **args = NULL;
+	char *dev_status_str;
+	char *log_status_str;
+	char *sync_str;
+	char *p = NULL;
+	int log_argc, num_devs;
+
+	/*
+	 * dm core parms:	     0 409600 mirror
+	 * Mirror core parms:	     2 253:4 253:5 400/400
+	 * New-style failure params: 1 AA
+	 * New-style log params:     3 cluster 253:3 A
+	 *			 or  3 disk 253:3 A
+	 *			 or  1 core
+	 */
+
+	/* number of devices */
+	if (!dm_split_words(params, 1, 0, &p))
+		goto out_parse;
+
+	if (!(num_devs = atoi(p)))
+		goto out_parse;
+	p += strlen(p) + 1;
+
+	/* devices names + "400/400" + "1 AA" + 1 or 3 log parms + NULL */
+	args = dm_malloc((num_devs + 7) * sizeof(char *));
+	if (!args || dm_split_words(p, num_devs + 7, 0, args) < num_devs + 5)
+		goto out_parse;
+
+	dev_status_str = args[2 + num_devs];
+	log_argc = atoi(args[3 + num_devs]);
+	log_status_str = args[3 + num_devs + log_argc];
+	sync_str = args[num_devs];
+
+	/* Check for bad mirror devices */
+	for (i = 0; i < num_devs; i++)
+		if (dev_status_str[i] == 'D') {
+			syslog(LOG_ERR, "Mirror device, %s, has failed.\n", args[i]);
+			r = ME_FAILURE;
+		}
+
+	/* Check for bad disk log device */
+	if (log_argc > 1 && log_status_str[0] == 'D') {
+		syslog(LOG_ERR, "Log device, %s, has failed.\n",
+		       args[2 + num_devs + log_argc]);
+		r = ME_FAILURE;
+	}
+
+	if (r == ME_FAILURE)
+		goto out;
+
+	p = strstr(sync_str, "/");
+	if (p) {
+		p[0] = '\0';
+		if (strcmp(sync_str, p+1))
+			r = ME_IGNORE;
+		p[0] = '/';
+	} else
+		goto out_parse;
+
+out:
+	if (args)
+		dm_free(args);
+	return r;
+	
+out_parse:
+	if (args)
+		dm_free(args);
+	syslog(LOG_ERR, "Unable to parse mirror status string.");
+	return ME_IGNORE;
+}
+
+static void _temporary_log_fn(int level, const char *file,
+			      int line, const char *format)
+{
+	if (!strncmp(format, "WARNING: ", 9) && (level < 5))
+		syslog(LOG_CRIT, "%s", format);
+	else
+		syslog(LOG_DEBUG, "%s", format);
+}
+
+static int _remove_failed_devices(const char *device)
+{
+	int r;
+#define CMD_SIZE 256	/* FIXME Use system restriction */
+	char cmd_str[CMD_SIZE];
+	char *vg = NULL, *lv = NULL, *layer = NULL;
+
+	if (strlen(device) > 200)  /* FIXME Use real restriction */
+		return -ENAMETOOLONG;	/* FIXME These return code distinctions are not used so remove them! */
+
+	if (!dm_split_lvm_name(_mem_pool, device, &vg, &lv, &layer)) {
+		syslog(LOG_ERR, "Unable to determine VG name from %s",
+		       device);
+		return -ENOMEM;	/* FIXME Replace with generic error return - reason for failure has already got logged */
+	}
+
+	/* FIXME Is any sanity-checking required on %s? */
+	if (CMD_SIZE <= snprintf(cmd_str, CMD_SIZE, "vgreduce --config devices{ignore_suspended_devices=1} --removemissing %s", vg)) {
+		/* this error should be caught above, but doesn't hurt to check again */
+		syslog(LOG_ERR, "Unable to form LVM command: Device name too long");
+		dm_pool_empty(_mem_pool);  /* FIXME: not safe with multiple threads */
+		return -ENAMETOOLONG; /* FIXME Replace with generic error return - reason for failure has already got logged */
+	}
+
+	r = lvm2_run(_lvm_handle, cmd_str);
+
+	dm_pool_empty(_mem_pool);  /* FIXME: not safe with multiple threads */
+	return (r == 1) ? 0 : -1;
+}
+
+void process_event(struct dm_task *dmt, enum dm_event_mask event,
+		   void **unused __attribute((unused)))
+{
+	void *next = NULL;
+	uint64_t start, length;
+	char *target_type = NULL;
+	char *params;
+	const char *device = dm_task_get_name(dmt);
+
+	if (pthread_mutex_trylock(&_event_mutex)) {
+		syslog(LOG_NOTICE, "Another thread is handling an event.  Waiting...");
+		pthread_mutex_lock(&_event_mutex);
+	}
+	do {
+		next = dm_get_next_target(dmt, next, &start, &length,
+					  &target_type, &params);
+
+		if (!target_type) {
+			syslog(LOG_INFO, "%s mapping lost.\n", device);
+			continue;
+		}
+
+		if (strcmp(target_type, "mirror")) {
+			syslog(LOG_INFO, "%s has unmirrored portion.\n", device);
+			continue;
+		}
+
+		switch(_get_mirror_event(params)) {
+		case ME_INSYNC:
+			/* FIXME: all we really know is that this
+			   _part_ of the device is in sync
+			   Also, this is not an error
+			*/
+			syslog(LOG_NOTICE, "%s is now in-sync\n", device);
+			break;
+		case ME_FAILURE:
+			syslog(LOG_ERR, "Device failure in %s\n", device);
+			if (_remove_failed_devices(device))
+				/* FIXME Why are all the error return codes unused? Get rid of them? */
+				syslog(LOG_ERR, "Failed to remove faulty devices in %s\n",
+				       device);
+			/* Should check before warning user that device is now linear
+			else
+				syslog(LOG_NOTICE, "%s is now a linear device.\n",
+					device);
+			*/
+			break;
+		case ME_IGNORE:
+			break;
+		default:
+			/* FIXME Provide value then! */
+			syslog(LOG_INFO, "Unknown event received.\n");
+		}
+	} while (next);
+
+	pthread_mutex_unlock(&_event_mutex);
+}
+
+int register_device(const char *device, const char *uuid, int major, int minor,
+		   void **unused __attribute((unused)))
+{
+	int r = 0;
+
+	pthread_mutex_lock(&_register_mutex);
+
+	syslog(LOG_INFO, "Monitoring mirror device %s for events\n", device);
+
+	/*
+	 * Need some space for allocations.  1024 should be more
+	 * than enough for what we need (device mapper name splitting)
+	 */
+	if (!_mem_pool && !(_mem_pool = dm_pool_create("mirror_dso", 1024)))
+		goto out;
+
+	if (!_lvm_handle) {
+		lvm2_log_fn(_temporary_log_fn);
+		if (!(_lvm_handle = lvm2_init())) {
+			dm_pool_destroy(_mem_pool);
+			_mem_pool = NULL;
+			goto out;
+		}
+		lvm2_log_level(_lvm_handle, LVM2_LOG_SUPPRESS);
+		/* FIXME Temporary: move to dmeventd core */
+		lvm2_run(_lvm_handle, "_memlock_inc");
+	}
+
+	_register_count++;
+	r = 1;
+
+out:
+	pthread_mutex_unlock(&_register_mutex);
+
+	return r;
+}
+
+int unregister_device(const char *device, const char *uuid, int major, int minor,
+		   void **unused __attribute((unused)))
+{
+	pthread_mutex_lock(&_register_mutex);
+
+	syslog(LOG_INFO, "No longer monitoring mirror device %s for events\n",
+	       device);
+
+	if (!--_register_count) {
+		dm_pool_destroy(_mem_pool);
+		_mem_pool = NULL;
+		lvm2_run(_lvm_handle, "_memlock_dec");
+		lvm2_exit(_lvm_handle);
+		_lvm_handle = NULL;
+	}
+
+	pthread_mutex_unlock(&_register_mutex);
+
+	return 1;
+}

doc/Makefile.in

@@ -0,0 +1,29 @@
+#
+# 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
+
+srcdir = @srcdir@
+top_srcdir = @top_srcdir@
+VPATH = @srcdir@
+
+CONFSRC=example.conf
+CONFDEST=lvm.conf
+
+include $(top_srcdir)/make.tmpl
+
+install:
+	@if [ ! -e $(confdir)/$(CONFDEST) ]; then \
+		echo "Installing $(CONFSRC) as $(confdir)/$(CONFDEST)"; \
+		@INSTALL@ -D $(OWNER) $(GROUP) -m 644 $(CONFSRC) \
+			$(confdir)/$(CONFDEST); \
+	fi
+

doc/example.conf

@@ -1,74 +1,130 @@
-# This is an example configuration file for the LVM2 system.  It
-# contains the default settings that would be used if there was no
+# This is an example configuration file for the LVM2 system.
+# It contains the default settings that would be used if there was no
 # /etc/lvm/lvm.conf file.
-# Refer to 'man lvm.conf' for further information.
+#
+# Refer to 'man lvm.conf' for further information including the file layout.
+#
+# To put this file in a different directory and override /etc/lvm set
+# the environment variable LVM_SYSTEM_DIR before running the tools.
 
 
-# This section allows the user to configure which block devices should
+# This section allows you to configure which block devices should
 # be used by the LVM system.
 devices {
 
-    # where do you want your volume groups to appear ?
+    # Where do you want your volume groups to appear ?
     dir = "/dev"
 
     # An array of directories that contain the device nodes you wish
     # to use with LVM2.
-    scan = "/dev"
+    scan = [ "/dev" ]
 
-    # A very important option, that allows you to tune the LVM2 system
-    # to just look at a restricted set of devices that you're
-    # interested in.
+    # If several entries in the scanned directories correspond to the
+    # same block device and the tools need to display a name for device,
+    # all the pathnames are matched against each item in the following
+    # list of regular expressions in turn and the first match is used.
+    preferred_names = [ ]
 
+    # preferred_names = [ "^/dev/mpath/", "^/dev/[hs]d" ]
+
+    # A filter that tells LVM2 to only use a restricted set of devices.
     # The filter consists of an array of regular expressions.  These
     # expressions can be delimited by a character of your choice, and
     # prefixed with either an 'a' (for accept) or 'r' (for reject).
-    # ATM you cannot use anchors (^ or $) in your regular expression.
+    # The first expression found to match a device name determines if
+    # the device will be accepted or rejected (ignored).  Devices that
+    # don't match any patterns are accepted.
+
+    # Be careful if there there are symbolic links or multiple filesystem 
+    # entries for the same device as each name is checked separately against
+    # the list of patterns.  The effect is that if any name matches any 'a'
+    # pattern, the device is accepted; otherwise if any name matches any 'r'
+    # pattern it is rejected; otherwise it is accepted.
+
+    # Don't have more than one filter line active at once: only one gets used.
+
+    # Run vgscan after you change this parameter to ensure that
+    # the cache file gets regenerated (see below).
+    # If it doesn't do what you expect, check the output of 'vgscan -vvvv'.
 
-    # Remember to run vgscan after you change this parameter.
 
     # By default we accept every block device:
-    filter = "a/.*/"
+    filter = [ "a/.*/" ]
+
+    # Exclude the cdrom drive
+    # filter = [ "r|/dev/cdrom|" ]
 
     # When testing I like to work with just loopback devices:
-    # filter = ["a/loop/", "r/.*/"]
+    # filter = [ "a/loop/", "r/.*/" ]
 
     # Or maybe all loops and ide drives except hdc:
-    # filter =["a|loop|", "r|/dev/hdc|", "a|/dev/ide|", "r|.*|"]
+    # filter =[ "a|loop|", "r|/dev/hdc|", "a|/dev/ide|", "r|.*|" ]
 
-    # The results of all the filtering are cached on disk to avoid
-    # rescanning dud devices (which can take a very long time).  By
-    # default this cache file is hidden in the /etc/lvm directory, it
-    # is human readable to aid filter debugging.
-    cache = "/etc/lvm/.cache"
+    # Use anchors if you want to be really specific
+    # filter = [ "a|^/dev/hda8$|", "r/.*/" ]
+
+    # The results of the filtering are cached on disk to avoid
+    # rescanning dud devices (which can take a very long time).
+    # By default this cache is stored in the /etc/lvm/cache directory
+    # in a file called '.cache'.
+    # It is safe to delete the contents: the tools regenerate it.
+    # (The old setting 'cache' is still respected if neither of
+    # these new ones is present.)
+    cache_dir = "/etc/lvm/cache"
+    cache_file_prefix = ""
 
     # You can turn off writing this cache file by setting this to 0.
     write_cache_state = 1
+
+    # Advanced settings.
+
+    # List of pairs of additional acceptable block device types found 
+    # in /proc/devices with maximum (non-zero) number of partitions.
+    # types = [ "fd", 16 ]
+
+    # If sysfs is mounted (2.6 kernels) restrict device scanning to 
+    # the block devices it believes are valid.
+    # 1 enables; 0 disables.
+    sysfs_scan = 1	
+
+    # By default, LVM2 will ignore devices used as components of
+    # software RAID (md) devices by looking for md superblocks.
+    # 1 enables; 0 disables.
+    md_component_detection = 1
+
+    # If, while scanning the system for PVs, LVM2 encounters a device-mapper
+    # device that has its I/O suspended, it waits for it to become accessible.
+    # Set this to 1 to skip such devices.  This should only be needed
+    # in recovery situations.
+    ignore_suspended_devices = 0
 }
 
-# A section that allows the user to configure the nature of the
+# This section that allows you to configure the nature of the
 # information that LVM2 reports.
 log {
 
-    # Where should the log of error and debug messages go ?  By
-    # default there is no log.
-    #file = "/var/log/lvm2.log"
-
-    # Should we overwrite the last log.  By default we append.
-    overwrite = 0
-
-    # There are 9 log levels, with 9 being the most verbose.
-    level = 3
-    
-    # Controls the messages sent to stdout or stderr while running 
-    # LVM2.  There are three levels of verbosity, 3 being the most 
-    # verbose.
+    # Controls the messages sent to stdout or stderr.
+    # There are three levels of verbosity, 3 being the most verbose.
     verbose = 0
 
     # Should we send log messages through syslog?
     # 1 is yes; 0 is no.
     syslog = 1
 
-    # Choose format of output messages
+    # Should we log error and debug messages to a file?
+    # By default there is no log file.
+    #file = "/var/log/lvm2.log"
+
+    # Should we overwrite the log file each time the program is run?
+    # By default we append.
+    overwrite = 0
+
+    # What level of log messages should we send to the log file and/or syslog?
+    # There are 6 syslog-like log levels currently in use - 2 to 7 inclusive.
+    # 7 is the most verbose (LOG_DEBUG).
+    level = 0
+    
+    # Format of output messages
     # Whether or not (1 or 0) to indent messages according to their severity
     indent = 1
 
@@ -76,13 +132,18 @@ log {
     command_names = 0
 
     # A prefix to use before the message text (but after the command name,
-    # if selected)
+    # if selected).  Default is two spaces, so you can see/grep the severity
+    # of each message.
     prefix = "  "
 
-    # To make the messages look similar to the original LVM use:
+    # To make the messages look similar to the original LVM tools use:
     #   indent = 0
     #   command_names = 1
     #   prefix = " -- "
+
+    # Set this if you want log messages during activation.
+    # Don't use this in low memory situations (can deadlock).
+    # activation = 0
 }
 
 # Configuration of metadata backups and archiving.  In LVM2 when we
@@ -93,19 +154,20 @@ backup {
 
     # Should we maintain a backup of the current metadata configuration ?
     # Use 1 for Yes; 0 for No.
-    # Think very hard before turning this off.
+    # Think very hard before turning this off!
     backup = 1
 
     # Where shall we keep it ?
+    # Remember to back up this directory regularly!
     backup_dir = "/etc/lvm/backup"
 
-
     # Should we maintain an archive of old metadata configurations.
     # Use 1 for Yes; 0 for No.
     # On by default.  Think very hard before turning this off.
     archive = 1
 
     # Where should archived files go ?
+    # Remember to back up this directory regularly!
     archive_dir = "/etc/lvm/archive"
     
     # What is the minimum number of archive files you wish to keep ?
@@ -115,20 +177,15 @@ backup {
     retain_days = 30
 }
 
-# Settings for the running LVM2 in shell mode.
+# Settings for the running LVM2 in shell (readline) mode.
 shell {
 
     # Number of lines of history to store in ~/.lvm_history
     history_size = 100
 }
 
-# Metadata settings
-metadata {
-    # List of directories holding copies of text format metadata
-    dirs = [ "/etc/lvm/metadata" ]
-}
 
-# Miscellaneous global settings
+# Miscellaneous global LVM2 settings
 global {
     
     # The file creation mask for any files and directories created.
@@ -143,10 +200,194 @@ global {
     # command.  Defaults to off.
     test = 0
 
-    # Default metadata format commands use - "lvm1" (default) or "text"
-    format = "lvm1"
+    # Default value for --units argument
+    units = "h"
+
+    # Whether or not to communicate with the kernel device-mapper.
+    # Set to 0 if you want to use the tools to manipulate LVM metadata 
+    # without activating any logical volumes.
+    # If the device-mapper kernel driver is not present in your kernel
+    # setting this to 0 should suppress the error messages.
+    activation = 1
+
+    # If we can't communicate with device-mapper, should we try running 
+    # the LVM1 tools?
+    # This option only applies to 2.4 kernels and is provided to help you
+    # switch between device-mapper kernels and LVM1 kernels.
+    # The LVM1 tools need to be installed with .lvm1 suffices
+    # e.g. vgscan.lvm1 and they will stop working after you start using
+    # the new lvm2 on-disk metadata format.
+    # The default value is set when the tools are built.
+    # fallback_to_lvm1 = 0
+
+    # The default metadata format that commands should use - "lvm1" or "lvm2".
+    # The command line override is -M1 or -M2.
+    # Defaults to "lvm1" if compiled in, else "lvm2".
+    # format = "lvm1"
 
     # Location of proc filesystem
     proc = "/proc"
+
+    # Type of locking to use. Defaults to local file-based locking (1).
+    # Turn locking off by setting to 0 (dangerous: risks metadata corruption
+    # if LVM2 commands get run concurrently).
+    # Type 2 uses the external shared library locking_library.
+    # Type 3 uses built-in clustered locking.
+    locking_type = 1
+
+    # If using external locking (type 2) and initialisation fails,
+    # with this set to 1 an attempt will be made to use the built-in
+    # clustered locking.
+    # If you are using a customised locking_library you should set this to 0.
+    fallback_to_clustered_locking = 1
+
+    # If an attempt to initialise type 2 or type 3 locking failed, perhaps
+    # because cluster components such as clvmd are not running, with this set
+    # to 1 an attempt will be made to use local file-based locking (type 1).
+    # If this succeeds, only commands against local volume groups will proceed.
+    # Volume Groups marked as clustered will be ignored.
+    fallback_to_local_locking = 1
+
+    # Local non-LV directory that holds file-based locks while commands are
+    # in progress.  A directory like /tmp that may get wiped on reboot is OK.
+    locking_dir = "/var/lock/lvm"
+
+    # Other entries can go here to allow you to load shared libraries
+    # e.g. if support for LVM1 metadata was compiled as a shared library use
+    #   format_libraries = "liblvm2format1.so" 
+    # Full pathnames can be given.
+
+    # Search this directory first for shared libraries.
+    #   library_dir = "/lib"
+
+    # The external locking library to load if locking_type is set to 2.
+    #   locking_library = "liblvm2clusterlock.so"
 }
 
+activation {
+    # Device used in place of missing stripes if activating incomplete volume.
+    # For now, you need to set this up yourself first (e.g. with 'dmsetup')
+    # For example, you could make it return I/O errors using the 'error' 
+    # target or make it return zeros.
+    missing_stripe_filler = "/dev/ioerror"
+
+    # How much stack (in KB) to reserve for use while devices suspended
+    reserved_stack = 256
+
+    # How much memory (in KB) to reserve for use while devices suspended
+    reserved_memory = 8192
+
+    # Nice value used while devices suspended
+    process_priority = -18
+
+    # If volume_list is defined, each LV is only activated if there is a
+    # match against the list.
+    #   "vgname" and "vgname/lvname" are matched exactly.
+    #   "@tag" matches any tag set in the LV or VG.
+    #   "@*" matches if any tag defined on the host is also set in the LV or VG
+    #
+    # volume_list = [ "vg1", "vg2/lvol1", "@tag1", "@*" ]
+
+    # Size (in KB) of each copy operation when mirroring
+    mirror_region_size = 512
+
+    # Setting to use when there is no readahead value stored in the metadata.
+    #
+    # "none" - Disable readahead.
+    # "auto" - Use default value chosen by kernel.
+    readahead = "auto"
+
+    # 'mirror_image_fault_policy' and 'mirror_log_fault_policy' define
+    # how a device failure affecting a mirror is handled.
+    # A mirror is composed of mirror images (copies) and a log.
+    # A disk log ensures that a mirror does not need to be re-synced
+    # (all copies made the same) every time a machine reboots or crashes.
+    #
+    # In the event of a failure, the specified policy will be used to
+    # determine what happens:
+    #
+    # "remove" - Simply remove the faulty device and run without it.  If
+    #            the log device fails, the mirror would convert to using
+    #            an in-memory log.  This means the mirror will not
+    #            remember its sync status across crashes/reboots and
+    #            the entire mirror will be re-synced.  If a
+    #            mirror image fails, the mirror will convert to a
+    #            non-mirrored device if there is only one remaining good
+    #            copy.
+    #
+    # "allocate" - Remove the faulty device and try to allocate space on
+    #            a new device to be a replacement for the failed device.
+    #            Using this policy for the log is fast and maintains the
+    #            ability to remember sync state through crashes/reboots.
+    #            Using this policy for a mirror device is slow, as it
+    #            requires the mirror to resynchronize the devices, but it
+    #            will preserve the mirror characteristic of the device.
+    #            This policy acts like "remove" if no suitable device and
+    #            space can be allocated for the replacement.
+    #            Currently this is not implemented properly and behaves
+    #            similarly to:
+    #
+    # "allocate_anywhere" - Operates like "allocate", but it does not
+    #            require that the new space being allocated be on a
+    #            device is not part of the mirror.  For a log device
+    #            failure, this could mean that the log is allocated on
+    #            the same device as a mirror device.  For a mirror
+    #            device, this could mean that the mirror device is
+    #            allocated on the same device as another mirror device.
+    #            This policy would not be wise for mirror devices
+    #            because it would break the redundant nature of the
+    #            mirror.  This policy acts like "remove" if no suitable
+    #            device and space can be allocated for the replacement.
+
+    mirror_log_fault_policy = "allocate"
+    mirror_device_fault_policy = "remove"
+}
+
+
+####################
+# Advanced section #
+####################
+
+# Metadata settings
+#
+# metadata {
+    # Default number of copies of metadata to hold on each PV.  0, 1 or 2.
+    # You might want to override it from the command line with 0 
+    # when running pvcreate on new PVs which are to be added to large VGs.
+
+    # pvmetadatacopies = 1
+
+    # Approximate default size of on-disk metadata areas in sectors.
+    # You should increase this if you have large volume groups or
+    # you want to retain a large on-disk history of your metadata changes.
+
+    # pvmetadatasize = 255
+
+    # List of directories holding live copies of text format metadata.
+    # These directories must not be on logical volumes!
+    # It's possible to use LVM2 with a couple of directories here,
+    # preferably on different (non-LV) filesystems, and with no other 
+    # on-disk metadata (pvmetadatacopies = 0). Or this can be in
+    # addition to on-disk metadata areas.
+    # The feature was originally added to simplify testing and is not
+    # supported under low memory situations - the machine could lock up.
+    #
+    # Never edit any files in these directories by hand unless you
+    # you are absolutely sure you know what you are doing! Use
+    # the supplied toolset to make changes (e.g. vgcfgrestore).
+
+    # dirs = [ "/etc/lvm/metadata", "/mnt/disk2/lvm/metadata2" ]
+#}
+
+# Event daemon
+#
+# dmeventd {
+    # mirror_library is the library used when monitoring a mirror device.
+    #
+    # "libdevmapper-event-lvm2mirror.so" attempts to recover from failures.
+    # It removes failed devices from a volume group and reconfigures a
+    # mirror as necessary.
+    #
+    # mirror_library = "libdevmapper-event-lvm2mirror.so"
+#}
+

doc/example_cmdlib.c

@@ -0,0 +1,47 @@
+/*
+ * 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 "lvm2cmd.h"
+
+/* All output gets passed to this function line-by-line */
+void test_log_fn(int level, const char *file, int line, const char *format)
+{
+	/* Extract and process output here rather than printing it */
+
+	if (level != 4)
+		return;
+
+	printf("%s\n", format);
+	return;
+}
+
+int main(int argc, char **argv)
+{
+	void *handle;
+	int r;
+
+	lvm2_log_fn(test_log_fn);
+
+	handle = lvm2_init();
+
+	lvm2_log_level(handle, 1);
+	r = lvm2_run(handle, "vgs --noheadings vg1");
+
+	/* More commands here */
+
+	lvm2_exit(handle);
+
+	return r;
+}
+

doc/tagging.txt

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

doc/testing.txt

@@ -22,8 +22,8 @@ LVM2 that is running the LV's on my development box.
    }
 
 
-   The important this to note is the devices section which makes sure that
-   only the loopback devices are considered for LVM2 operations.
+   The important thing to note is the devices section which makes sure
+   that only the loopback devices are considered for LVM2 operations.
 
 4) When you want to use this test setup just set the environment
    variable LVM_SYSTEM_DIR to point to your config directory
@@ -39,8 +39,3 @@ LVM2 that is running the LV's on my development box.
 7) Test away.  Make sure that you are explicit about which lvm
    executable you want to execute (eg, ./lvm if you are in
    LVM2/tools).
-
-   
-
-
-

include/.symlinks

@@ -1,33 +1,52 @@
+../daemons/clvmd/clvm.h
 ../lib/activate/activate.h
+../lib/activate/targets.h
+../lib/cache/lvmcache.h
 ../lib/commands/errors.h
 ../lib/commands/toolcontext.h
 ../lib/config/config.h
 ../lib/config/defaults.h
-../lib/datastruct/bitset.h
 ../lib/datastruct/btree.h
-../lib/datastruct/hash.h
 ../lib/datastruct/list.h
 ../lib/datastruct/lvm-types.h
+../lib/datastruct/str_list.h
 ../lib/device/dev-cache.h
 ../lib/device/device.h
 ../lib/display/display.h
 ../lib/filters/filter-composite.h
+../lib/filters/filter-md.h
 ../lib/filters/filter-persistent.h
 ../lib/filters/filter-regex.h
+../lib/filters/filter-sysfs.h
 ../lib/filters/filter.h
 ../lib/format1/format1.h
-../lib/format1/lvm1_label.h
+../lib/format_pool/format_pool.h
+../lib/format_text/archiver.h
 ../lib/format_text/format-text.h
+../lib/format_text/text_export.h
+../lib/format_text/text_import.h
 ../lib/label/label.h
-../lib/label/uuid-map.h
 ../lib/locking/locking.h
 ../lib/log/log.h
+../lib/metadata/lv_alloc.h
 ../lib/metadata/metadata.h
-../lib/mm/dbg_malloc.h
-../lib/mm/pool.h
+../lib/metadata/metadata-exported.h
+../lib/metadata/pv_alloc.h
+../lib/metadata/segtype.h
+../lib/mm/memlock.h
 ../lib/mm/xlate.h
+../lib/misc/configure.h
+../lib/misc/crc.h
+../lib/misc/intl.h
+../lib/misc/util.h
+../lib/misc/last-path-component.h
+../lib/misc/lib.h
+../lib/misc/lvm-exec.h
 ../lib/misc/lvm-file.h
 ../lib/misc/lvm-string.h
-../lib/regex/matcher.h
+../lib/misc/lvm-wrappers.h
+../lib/misc/sharedlib.h
+../lib/report/report.h
 ../lib/uuid/uuid.h
-../lib/vgcache/vgcache.h
+../po/pogen.h
+../tools/version.h

include/Makefile.in

@@ -1,20 +1,16 @@
 #
-# Copyright (C) 2001 Sistina Software
+# Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.
+# Copyright (C) 2004 Red Hat, Inc. All rights reserved.
 #
-# This LVM library is free software; you can redistribute it and/or
-# modify it under the terms of the GNU Library General Public
-# License as published by the Free Software Foundation; either
-# version 2 of the License, or (at your option) any later version.
+# This file is part of LVM2.
 #
-# This LVM library is distributed in the hope that it will be useful,
-# but WITHOUT ANY WARRANTY; without even the implied warranty of
-# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-# Library General Public License for more details.
+# 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 Library General Public
-# License along with this LVM library; if not, write to the Free
-# Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
-# MA 02111-1307, USA
+# 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
 
 SHELL = /bin/sh
 
@@ -24,6 +20,8 @@ VPATH = @srcdir@
 
 LN_S = @LN_S@
 
+.PHONY: clean distclean all install pofile install_cluster
+
 all: .symlinks_created
 
 .symlinks_created: .symlinks
@@ -35,9 +33,13 @@ distclean:
 	find . -maxdepth 1 -type l -exec $(RM) \{\} \;
 	$(RM) Makefile .include_symlinks .symlinks_created
 
+pofile: all
+
 clean:
 
 install:
 
-.PHONY: clean distclean all install
+install_cluster:
+
+cflow:
 

lib/Makefile.in

@@ -1,44 +1,67 @@
 #
-# Copyright (C) 2001 Sistina Software (UK) Limited
+# Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.
+# Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved.
 #
-# This file is released under the GPL.
+# 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
 
 srcdir = @srcdir@
 top_srcdir = @top_srcdir@
 VPATH = @srcdir@
 
-SOURCES=\
+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 \
-	activate/dev_manager.c \
-	activate/fs.c \
+	cache/lvmcache.c \
+	commands/toolcontext.c \
 	config/config.c \
-	datastruct/bitset.c \
 	datastruct/btree.c \
-	datastruct/hash.c \
+	datastruct/list.c \
+	datastruct/str_list.c \
 	device/dev-cache.c \
 	device/dev-io.c \
+	device/dev-md.c \
 	device/device.c \
 	display/display.c \
+	error/errseg.c \
 	filters/filter-composite.c \
 	filters/filter-persistent.c \
 	filters/filter-regex.c \
+	filters/filter-sysfs.c \
+	filters/filter-md.c \
 	filters/filter.c \
-	format1/disk-rep.c \
-	format1/format1.c \
-	format1/import-export.c \
-	format1/import-extents.c \
-	format1/layout.c \
-	format1/lvm1_label.c \
-	format1/vg_number.c \
 	format_text/archive.c \
+	format_text/archiver.c \
 	format_text/export.c \
 	format_text/flags.c \
 	format_text/format-text.c \
 	format_text/import.c \
+	format_text/import_vsn1.c \
+	format_text/tags.c \
+	format_text/text_label.c \
 	label/label.c \
-	label/uuid-map.c \
-	locking/external_locking.c \
 	locking/file_locking.c \
 	locking/locking.c \
 	locking/no_locking.c \
@@ -46,23 +69,87 @@ SOURCES=\
 	metadata/lv_manip.c \
 	metadata/merge.c \
 	metadata/metadata.c \
+	metadata/mirror.c \
+	metadata/pv_manip.c \
 	metadata/pv_map.c \
+	metadata/segtype.c \
 	metadata/snapshot_manip.c \
+	misc/crc.c \
+	misc/lvm-exec.c \
 	misc/lvm-file.c \
-	mm/dbg_malloc.c \
-	mm/pool.c \
-	regex/matcher.c \
-	regex/parse_rx.c \
-	regex/ttree.c \
+	misc/lvm-string.c \
+	misc/lvm-wrappers.c \
+	misc/timestamp.c \
+	misc/util.c \
+	mm/memlock.c \
+	report/report.c \
+	striped/striped.c \
 	uuid/uuid.c \
-	vgcache/vgcache.c
+	zero/zero.c
 
-TARGETS=liblvm.a
+ifeq ("@LVM1@", "internal")
+  SOURCES +=\
+	format1/disk-rep.c \
+	format1/format1.c \
+	format1/import-export.c \
+	format1/import-extents.c \
+	format1/layout.c \
+	format1/lvm1-label.c \
+	format1/vg_number.c
+endif
 
-include ../make.tmpl
+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
 
-liblvm.a: $(OBJECTS)
-	$(RM) $@
-	$(AR) r $@ $(OBJECTS)
-	$(RANLIB) $@
+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 ("@DEVMAPPER@", "yes")
+  SOURCES +=\
+	activate/dev_manager.c \
+	activate/fs.c
+endif
+
+ifeq ("@HAVE_LIBDL@", "yes")
+  SOURCES +=\
+	locking/external_locking.c \
+	misc/sharedlib.c
+endif
+
+ifeq ("@DMEVENTD@", "yes")
+  CLDFLAGS += -ldevmapper-event
+endif
+
+LIB_STATIC = liblvm.a
+
+$(SUBDIRS): $(LIB_STATIC)
+
+CLEAN_TARGETS += liblvm.cflow
+
+include $(top_srcdir)/make.tmpl
+
+liblvm.cflow: $(SOURCES)
+	set -e; (echo -n "SOURCES += "; \
+		 echo $(SOURCES) | \
+		 sed "s/^/ /;s/ / $(top_srcdir)\/lib\//g;s/$$//"; \
+		 ) > $@
+
+cflow: liblvm.cflow

lib/activate/activate.h

@@ -1,42 +1,105 @@
 /*
- * Copyright (C) 2001 Sistina Software (UK) Limited.
+ * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved.
  *
- * This file is released under the LGPL.
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
 #ifndef LVM_ACTIVATE_H
 #define LVM_ACTIVATE_H
 
-#include <libdevmapper.h>
+#include "metadata-exported.h"
+
+struct lvinfo {
+	int exists;
+	int suspended;
+	unsigned int open_count;
+	int major;
+	int minor;
+	int read_only;
+	int live_table;
+	int inactive_table;
+	uint32_t read_ahead;
+};
+
+void set_activation(int activation);
+int activation(void);
 
 int driver_version(char *version, size_t size);
 int library_version(char *version, size_t size);
+int lvm1_present(struct cmd_context *cmd);
+
+int target_present(const char *target_name, int use_modprobe);
+int target_version(const char *target_name, uint32_t *maj,
+                   uint32_t *min, uint32_t *patchlevel);
+int list_segment_modules(struct dm_pool *mem, const struct lv_segment *seg,
+			 struct list *modules);
+int list_lv_modules(struct dm_pool *mem, const struct logical_volume *lv,
+		    struct list *modules);
+
+void activation_release(void);
+void activation_exit(void);
+
+int lv_suspend(struct cmd_context *cmd, const char *lvid_s);
+int lv_suspend_if_active(struct cmd_context *cmd, const char *lvid_s);
+int lv_resume(struct cmd_context *cmd, const char *lvid_s);
+int lv_resume_if_active(struct cmd_context *cmd, const char *lvid_s);
+int lv_activate(struct cmd_context *cmd, const char *lvid_s, int exclusive);
+int lv_activate_with_filter(struct cmd_context *cmd, const char *lvid_s,
+			    int exclusive);
+int lv_deactivate(struct cmd_context *cmd, const char *lvid_s);
+
+int lv_mknodes(struct cmd_context *cmd, const struct logical_volume *lv);
 
 /*
  * Returns 1 if info structure has been populated, else 0.
  */
-int lv_info(struct logical_volume *lv, struct dm_info *info);
+int lv_info(struct cmd_context *cmd, const struct logical_volume *lv, struct lvinfo *info,
+	    int with_open_count, int with_read_ahead);
+int lv_info_by_lvid(struct cmd_context *cmd, const char *lvid_s,
+		    struct lvinfo *info, int with_open_count, int with_read_ahead);
 
 /*
- * These should eventually use config file
- * to determine whether or not to activate
+ * Returns 1 if activate_lv has been set: 1 = activate; 0 = don't.
  */
-int lv_suspend_if_active(struct cmd_context *cmd, const char *lvid_s);
-int lv_resume_if_active(struct cmd_context *cmd, const char *lvid_s);
-int lv_activate(struct cmd_context *cmd, const char *lvid_s);
-int lv_deactivate(struct cmd_context *cmd, const char *lvid_s);
+int lv_activation_filter(struct cmd_context *cmd, const char *lvid_s,
+			 int *activate_lv);
 
 /*
- * FIXME:
- * I don't like the *lvs_in_vg* function names.
+ * Returns 1 if percent has been set, else 0.
  */
+int lv_snapshot_percent(const struct logical_volume *lv, float *percent);
+int lv_mirror_percent(struct cmd_context *cmd, struct logical_volume *lv,
+		      int wait, float *percent, uint32_t *event_nr);
 
 /*
  * Return number of LVs in the VG that are active.
  */
 int lvs_in_vg_activated(struct volume_group *vg);
-int lvs_in_vg_opened(struct volume_group *vg);
+int lvs_in_vg_activated_by_uuid_only(struct volume_group *vg);
+int lvs_in_vg_opened(const struct volume_group *vg);
 
-int lv_setup_cow_store(struct logical_volume *lv);
+
+int monitor_dev_for_events(struct cmd_context *cmd,
+			    struct logical_volume *lv, int do_reg);
+
+/*
+ * Returns 1 if PV has a dependency tree that uses anything in VG.
+ */
+int pv_uses_vg(struct physical_volume *pv,
+	       struct volume_group *vg);
+
+/*
+ * Returns 1 if mapped device is not suspended.
+ */
+int device_is_usable(dev_t dev);
 
 #endif

lib/activate/dev_manager.h

@@ -1,23 +1,36 @@
 /*
- * Copyright (C) 2002 Sistina Software (UK) Limited.
+ * Copyright (C) 2002-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004-2006 Red Hat, Inc. All rights reserved.
  *
- * This file is released under the LGPL.
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
 #ifndef _LVM_DEV_MANAGER_H
 #define _LVM_DEV_MANAGER_H
 
-#include "metadata.h"
-
-#include <libdevmapper.h>
-
+struct logical_volume;
+struct volume_group;
+struct cmd_context;
 struct dev_manager;
+struct dm_info;
+struct device;
 
 /*
  * Constructor and destructor.
  */
-struct dev_manager *dev_manager_create(const char *vg_name);
+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_release(void);
+void dev_manager_exit(void);
 
 /*
  * The device handler is responsible for creating all the layered
@@ -25,16 +38,31 @@ void dev_manager_destroy(struct dev_manager *dm);
  * (eg, an origin is created before its snapshot, but is not
  * unsuspended until the snapshot is also created.)
  */
-int dev_manager_info(struct dev_manager *dm, struct logical_volume *lv,
-		     struct dm_info *info);
-int dev_manager_suspend(struct dev_manager *dm, struct logical_volume *lv);
+int dev_manager_info(struct dm_pool *mem, const char *name,
+		     const struct logical_volume *lv,
+		     int mknodes, int with_open_count, int with_read_ahead,
+		     struct dm_info *info, uint32_t *read_ahead);
+int dev_manager_snapshot_percent(struct dev_manager *dm,
+				 const struct logical_volume *lv,
+				 float *percent);
+int dev_manager_mirror_percent(struct dev_manager *dm,
+			       struct logical_volume *lv, int wait,
+			       float *percent, uint32_t *event_nr);
+int dev_manager_suspend(struct dev_manager *dm, struct logical_volume *lv,
+			int lockfs);
 int dev_manager_activate(struct dev_manager *dm, struct logical_volume *lv);
+int dev_manager_preload(struct dev_manager *dm, struct logical_volume *lv);
 int dev_manager_deactivate(struct dev_manager *dm, struct logical_volume *lv);
 
+int dev_manager_lv_mknodes(const struct logical_volume *lv);
+int dev_manager_lv_rmnodes(const struct logical_volume *lv);
 
 /*
  * Put the desired changes into effect.
  */
 int dev_manager_execute(struct dev_manager *dm);
 
+int dev_manager_device_uses_vg(struct device *dev,
+			       struct volume_group *vg);
+
 #endif

lib/activate/fs.c

@@ -1,31 +1,37 @@
 /*
- * Copyright (C) 2001 Sistina Software (UK) Limited.
+ * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved.
  *
- * This file is released under the LGPL.
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
+#include "lib.h"
 #include "fs.h"
-#include "log.h"
 #include "toolcontext.h"
 #include "lvm-string.h"
 #include "lvm-file.h"
+#include "memlock.h"
 
-#include <sys/types.h>
 #include <sys/stat.h>
 #include <fcntl.h>
 #include <unistd.h>
 #include <limits.h>
-#include <stdio.h>
-#include <string.h>
+#include <dirent.h>
 
-#include <libdevmapper.h>
-
-static int _mk_dir(struct volume_group *vg)
+static int _mk_dir(const char *dev_dir, const char *vg_name)
 {
 	char vg_path[PATH_MAX];
 
-	if (lvm_snprintf(vg_path, sizeof(vg_path), "%s%s",
-			 vg->cmd->dev_dir, vg->name) == -1) {
+	if (dm_snprintf(vg_path, sizeof(vg_path), "%s%s",
+			 dev_dir, vg_name) == -1) {
 		log_error("Couldn't construct name of volume "
 			  "group directory.");
 		return 0;
@@ -35,7 +41,7 @@ static int _mk_dir(struct volume_group *vg)
 		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;
 	}
@@ -43,51 +49,119 @@ static int _mk_dir(struct volume_group *vg)
 	return 1;
 }
 
-static int _rm_dir(struct volume_group *vg)
+static int _rm_dir(const char *dev_dir, const char *vg_name)
 {
 	char vg_path[PATH_MAX];
 
-	if (lvm_snprintf(vg_path, sizeof(vg_path), "%s%s",
-			 vg->cmd->dev_dir, vg->name) == -1) {
+	if (dm_snprintf(vg_path, sizeof(vg_path), "%s%s",
+			 dev_dir, vg_name) == -1) {
 		log_error("Couldn't construct name of volume "
 			  "group directory.");
 		return 0;
 	}
 
-	log_very_verbose("Removing directory %s", vg_path);
-
-	if (is_empty_dir(vg_path))
+	if (dir_exists(vg_path) && is_empty_dir(vg_path)) {
+		log_very_verbose("Removing directory %s", vg_path);
 		rmdir(vg_path);
+	}
 
 	return 1;
 }
 
-static int _mk_link(struct logical_volume *lv, const char *dev)
+static void _rm_blks(const char *dir)
 {
-	char lv_path[PATH_MAX], link_path[PATH_MAX];
+	const char *name;
+	char path[PATH_MAX];
+	struct dirent *dirent;
+	struct stat buf;
+	DIR *d;
+
+	if (!(d = opendir(dir))) {
+		log_sys_error("opendir", dir);
+		return;
+	}
+
+	while ((dirent = readdir(d))) {
+		name = dirent->d_name;
+
+		if (!strcmp(name, ".") || !strcmp(name, ".."))
+			continue;
+
+		if (dm_snprintf(path, sizeof(path), "%s/%s", dir, name) == -1) {
+			log_error("Couldn't create path for %s", name);
+			continue;
+		}
+
+		if (!lstat(path, &buf)) {
+			if (!S_ISBLK(buf.st_mode))
+				continue;
+			log_very_verbose("Removing %s", path);
+			if (unlink(path) < 0)
+				log_sys_error("unlink", path);
+		}
+	}
+}
+
+static int _mk_link(const char *dev_dir, const char *vg_name,
+		    const char *lv_name, const char *dev)
+{
+	char lv_path[PATH_MAX], link_path[PATH_MAX], lvm1_group_path[PATH_MAX];
+	char vg_path[PATH_MAX];
 	struct stat buf;
 
-	if (lvm_snprintf(lv_path, sizeof(lv_path), "%s%s/%s",
-			 lv->vg->cmd->dev_dir, lv->vg->name, lv->name) == -1) {
-		log_error("Couldn't create source pathname for "
-			  "logical volume link %s", lv->name);
+	if (dm_snprintf(vg_path, sizeof(vg_path), "%s%s",
+			 dev_dir, vg_name) == -1) {
+		log_error("Couldn't create path for volume group dir %s",
+			  vg_name);
 		return 0;
 	}
 
-	if (lvm_snprintf(link_path, sizeof(link_path), "%s/%s",
+	if (dm_snprintf(lv_path, sizeof(lv_path), "%s/%s", vg_path,
+			 lv_name) == -1) {
+		log_error("Couldn't create source pathname for "
+			  "logical volume link %s", lv_name);
+		return 0;
+	}
+
+	if (dm_snprintf(link_path, sizeof(link_path), "%s/%s",
 			 dm_dir(), dev) == -1) {
 		log_error("Couldn't create destination pathname for "
-			  "logical volume link for %s", lv->name);
+			  "logical volume link for %s", lv_name);
 		return 0;
 	}
 
+	if (dm_snprintf(lvm1_group_path, sizeof(lvm1_group_path), "%s/group",
+			 vg_path) == -1) {
+		log_error("Couldn't create pathname for LVM1 group file for %s",
+			  vg_name);
+		return 0;
+	}
+
+	/* To reach this point, the VG must have been locked.
+	 * As locking fails if the VG is active under LVM1, it's 
+	 * now safe to remove any LVM1 devices we find here
+	 * (as well as any existing LVM2 symlink). */
+	if (!lstat(lvm1_group_path, &buf)) {
+		if (!S_ISCHR(buf.st_mode)) {
+			log_error("Non-LVM1 character device found at %s",
+				  lvm1_group_path);
+		} else {
+			_rm_blks(vg_path);
+
+			log_very_verbose("Removing %s", lvm1_group_path);
+			if (unlink(lvm1_group_path) < 0)
+				log_sys_error("unlink", lvm1_group_path);
+		}
+	}
+
 	if (!lstat(lv_path, &buf)) {
-		if (!S_ISLNK(buf.st_mode)) {
+		if (!S_ISLNK(buf.st_mode) && !S_ISBLK(buf.st_mode)) {
 			log_error("Symbolic link %s not created: file exists",
 				  link_path);
 			return 0;
 		}
 
+		log_very_verbose("Removing %s", lv_path);
 		if (unlink(lv_path) < 0) {
 			log_sys_error("unlink", lv_path);
 			return 0;
@@ -100,26 +174,36 @@ static int _mk_link(struct logical_volume *lv, const char *dev)
 		return 0;
 	}
 
+#ifdef HAVE_SELINUX
+        if (!dm_set_selinux_context(lv_path, S_IFLNK)) {
+                stack;
+                return 0;
+        }
+#endif
+
 	return 1;
 }
 
-static int _rm_link(struct logical_volume *lv, const char *lv_name)
+static int _rm_link(const char *dev_dir, const char *vg_name,
+		    const char *lv_name)
 {
 	struct stat buf;
 	char lv_path[PATH_MAX];
 
-	if (lvm_snprintf(lv_path, sizeof(lv_path), "%s%s/%s",
-			 lv->vg->cmd->dev_dir, lv->vg->name, lv_name) == -1) {
+	if (dm_snprintf(lv_path, sizeof(lv_path), "%s%s/%s",
+			 dev_dir, vg_name, lv_name) == -1) {
 		log_error("Couldn't determine link pathname.");
 		return 0;
 	}
 
-	log_very_verbose("Removing link %s", lv_path);
 	if (lstat(lv_path, &buf) || !S_ISLNK(buf.st_mode)) {
+		if (errno == ENOENT)
+			return 1;
 		log_error("%s not symbolic link - not removing", lv_path);
 		return 0;
 	}
 
+	log_very_verbose("Removing link %s", lv_path);
 	if (unlink(lv_path) < 0) {
 		log_sys_error("unlink", lv_path);
 		return 0;
@@ -128,35 +212,148 @@ static int _rm_link(struct logical_volume *lv, const char *lv_name)
 	return 1;
 }
 
-int fs_add_lv(struct logical_volume *lv, const char *dev)
+typedef enum {
+	FS_ADD,
+	FS_DEL,
+	FS_RENAME
+} fs_op_t;
+
+static int _do_fs_op(fs_op_t type, const char *dev_dir, const char *vg_name,
+		     const char *lv_name, const char *dev,
+		     const char *old_lv_name)
 {
-	if (!_mk_dir(lv->vg) || !_mk_link(lv, dev)) {
-		stack;
-		return 0;
+	switch (type) {
+	case FS_ADD:
+		if (!_mk_dir(dev_dir, vg_name) ||
+		    !_mk_link(dev_dir, vg_name, lv_name, dev)) {
+			stack;
+			return 0;
+		}
+		break;
+	case FS_DEL:
+		if (!_rm_link(dev_dir, vg_name, lv_name) ||
+		    !_rm_dir(dev_dir, vg_name)) {
+			stack;
+			return 0;
+		}
+		break;
+		/* FIXME Use rename() */
+	case FS_RENAME:
+		if (old_lv_name && !_rm_link(dev_dir, vg_name, old_lv_name))
+			stack;
+
+		if (!_mk_link(dev_dir, vg_name, lv_name, dev))
+			stack;
 	}
 
 	return 1;
 }
 
-int fs_del_lv(struct logical_volume *lv)
+static LIST_INIT(_fs_ops);
+
+struct fs_op_parms {
+	struct list list;
+	fs_op_t type;
+	char *dev_dir;
+	char *vg_name;
+	char *lv_name;
+	char *dev;
+	char *old_lv_name;
+	char names[0];
+};
+
+static void _store_str(char **pos, char **ptr, const char *str)
 {
-	if (!_rm_link(lv, lv->name) || !_rm_dir(lv->vg)) {
-		stack;
+	strcpy(*pos, str);
+	*ptr = *pos;
+	*pos += strlen(*ptr) + 1;
+}
+
+static int _stack_fs_op(fs_op_t type, const char *dev_dir, const char *vg_name,
+			const char *lv_name, const char *dev,
+			const char *old_lv_name)
+{
+	struct fs_op_parms *fsp;
+	size_t len = strlen(dev_dir) + strlen(vg_name) + strlen(lv_name) +
+	    strlen(dev) + strlen(old_lv_name) + 5;
+	char *pos;
+
+	if (!(fsp = dm_malloc(sizeof(*fsp) + len))) {
+		log_error("No space to stack fs operation");
 		return 0;
 	}
 
+	pos = fsp->names;
+	fsp->type = type;
+
+	_store_str(&pos, &fsp->dev_dir, dev_dir);
+	_store_str(&pos, &fsp->vg_name, vg_name);
+	_store_str(&pos, &fsp->lv_name, lv_name);
+	_store_str(&pos, &fsp->dev, dev);
+	_store_str(&pos, &fsp->old_lv_name, old_lv_name);
+
+	list_add(&_fs_ops, &fsp->list);
+
 	return 1;
 }
 
-/* FIXME Use rename() */
+static void _pop_fs_ops(void)
+{
+	struct list *fsph, *fspht;
+	struct fs_op_parms *fsp;
+
+	list_iterate_safe(fsph, fspht, &_fs_ops) {
+		fsp = list_item(fsph, struct fs_op_parms);
+		_do_fs_op(fsp->type, fsp->dev_dir, fsp->vg_name, fsp->lv_name,
+			  fsp->dev, fsp->old_lv_name);
+		list_del(&fsp->list);
+		dm_free(fsp);
+	}
+}
+
+static int _fs_op(fs_op_t type, const char *dev_dir, const char *vg_name,
+		  const char *lv_name, const char *dev, const char *old_lv_name)
+{
+	if (memlock()) {
+		if (!_stack_fs_op(type, dev_dir, vg_name, lv_name, dev,
+				  old_lv_name)) {
+			stack;
+			return 0;
+		}
+		return 1;
+	}
+
+	return _do_fs_op(type, dev_dir, vg_name, lv_name, dev, old_lv_name);
+}
+
+int fs_add_lv(const struct logical_volume *lv, const char *dev)
+{
+	return _fs_op(FS_ADD, lv->vg->cmd->dev_dir, lv->vg->name, lv->name,
+		      dev, "");
+}
+
+int fs_del_lv(const struct logical_volume *lv)
+{
+	return _fs_op(FS_DEL, lv->vg->cmd->dev_dir, lv->vg->name, lv->name,
+		      "", "");
+}
+
+int fs_del_lv_byname(const char *dev_dir, const char *vg_name, const char *lv_name)
+{
+	return _fs_op(FS_DEL, dev_dir, vg_name, lv_name, "", "");
+}
+
 int fs_rename_lv(struct logical_volume *lv,
 		 const char *dev, const char *old_name)
 {
-	if (old_name && !_rm_link(lv, old_name))
-		stack;
-
-	if (!_mk_link(lv, dev))
-		stack;
-
-	return 1;
+	return _fs_op(FS_RENAME, lv->vg->cmd->dev_dir, lv->vg->name, lv->name,
+		      dev, old_name);
+}
+
+void fs_unlock(void)
+{
+	if (!memlock()) {
+		dm_lib_release();
+		_pop_fs_ops();
+	}
 }

lib/activate/fs.h

@@ -1,7 +1,16 @@
 /*
- * Copyright (C) 2001 Sistina Software (UK) Limited.
+ * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved.
  *
- * This file is released under the LGPL.
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
 #ifndef _LVM_FS_H
@@ -14,10 +23,11 @@
  * up the volume group directory in /dev and the
  * symbolic links to the dm device.
  */
-int fs_add_lv(struct logical_volume *lv, const char *dev);
-int fs_del_lv(struct logical_volume *lv);
+int fs_add_lv(const struct logical_volume *lv, const char *dev);
+int fs_del_lv(const struct logical_volume *lv);
+int fs_del_lv_byname(const char *dev_dir, const char *vg_name, const char *lv_name);
 int fs_rename_lv(struct logical_volume *lv,
 		 const char *dev, const char *old_name);
-
+void fs_unlock(void);
 
 #endif

lib/activate/targets.h

@@ -0,0 +1,34 @@
+/*
+ * Copyright (C) 2003-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004-2006 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#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);
+
+int add_areas_line(struct dev_manager *dm, struct lv_segment *seg,
+                   struct dm_tree_node *node, uint32_t start_area, uint32_t areas);
+
+int build_dev_string(struct dev_manager *dm, char *dlid, char *devbuf,
+		     size_t bufsize, const char *desc);
+
+char *build_dlid(struct dev_manager *dm, const char *lvid, const char *layer);
+
+#endif

lib/cache/lvmcache.c

@@ -0,0 +1,900 @@
+/*
+ * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include "lib.h"
+#include "lvmcache.h"
+#include "toolcontext.h"
+#include "dev-cache.h"
+#include "metadata.h"
+#include "filter.h"
+#include "memlock.h"
+#include "str_list.h"
+
+static struct dm_hash_table *_pvid_hash = NULL;
+static struct dm_hash_table *_vgid_hash = NULL;
+static struct dm_hash_table *_vgname_hash = NULL;
+static struct dm_hash_table *_lock_hash = NULL;
+static struct list _vginfos;
+static int _has_scanned = 0;
+static int _vgs_locked = 0;
+
+int lvmcache_init(void)
+{
+	list_init(&_vginfos);
+
+	if (!(_vgname_hash = dm_hash_create(128)))
+		return 0;
+
+	if (!(_vgid_hash = dm_hash_create(128)))
+		return 0;
+
+	if (!(_pvid_hash = dm_hash_create(128)))
+		return 0;
+
+	if (!(_lock_hash = dm_hash_create(128)))
+		return 0;
+
+	return 1;
+}
+
+void lvmcache_lock_vgname(const char *vgname, int read_only __attribute((unused)))
+{
+	if (!_lock_hash && !lvmcache_init()) {
+		log_error("Internal cache initialisation failed");
+		return;
+	}
+
+	if (!dm_hash_insert(_lock_hash, vgname, (void *) 1))
+		log_error("Cache locking failure for %s", vgname);
+
+	_vgs_locked++;
+}
+
+int vgname_is_locked(const char *vgname)
+{
+	if (!_lock_hash)
+		return 0;
+
+	return dm_hash_lookup(_lock_hash, vgname) ? 1 : 0;
+}
+
+void lvmcache_unlock_vgname(const char *vgname)
+{
+	/* FIXME: Clear all CACHE_LOCKED flags in this vg */
+	dm_hash_remove(_lock_hash, vgname);
+
+	/* FIXME Do this per-VG */
+	if (!--_vgs_locked)
+		dev_close_all();
+}
+
+int vgs_locked(void)
+{
+	return _vgs_locked;
+}
+
+/* If vgid supplied, require a match. */
+struct lvmcache_vginfo *vginfo_from_vgname(const char *vgname, const char *vgid)
+{
+	struct lvmcache_vginfo *vginfo;
+
+	if (!_vgname_hash)
+		return NULL;
+
+	if (!(vginfo = dm_hash_lookup(_vgname_hash, vgname)))
+		return NULL;
+
+	if (vgid)
+		do 
+			if (!strncmp(vgid, vginfo->vgid, ID_LEN))
+				return vginfo;
+		while ((vginfo = vginfo->next));
+
+	return vginfo;
+}
+
+const struct format_type *fmt_from_vgname(const char *vgname, const char *vgid)
+{
+	struct lvmcache_vginfo *vginfo;
+	struct lvmcache_info *info;
+	struct label *label;
+	struct list *devh, *tmp;
+	struct list devs;
+	struct device_list *devl;
+	char vgid_found[ID_LEN + 1] __attribute((aligned(8)));
+
+	if (!(vginfo = vginfo_from_vgname(vgname, vgid)))
+		return NULL;
+
+	/* This function is normally called before reading metadata so
+ 	 * we check cached labels here. Unfortunately vginfo is volatile. */
+	list_init(&devs);
+	list_iterate_items(info, &vginfo->infos) {
+		if (!(devl = dm_malloc(sizeof(*devl)))) {
+			log_error("device_list element allocation failed");
+			return NULL;
+		}
+		devl->dev = info->dev;
+		list_add(&devs, &devl->list);
+	}
+
+	memcpy(vgid_found, vginfo->vgid, sizeof(vgid_found));
+
+	list_iterate_safe(devh, tmp, &devs) {
+		devl = list_item(devh, struct device_list);
+		label_read(devl->dev, &label, UINT64_C(0));
+		list_del(&devl->list);
+		dm_free(devl);
+	}
+
+	/* If vginfo changed, caller needs to rescan */
+	if (!(vginfo = vginfo_from_vgname(vgname, vgid_found)) ||
+	    strncmp(vginfo->vgid, vgid_found, ID_LEN))
+		return NULL;
+
+	return vginfo->fmt;
+}
+
+struct lvmcache_vginfo *vginfo_from_vgid(const char *vgid)
+{
+	struct lvmcache_vginfo *vginfo;
+	char id[ID_LEN + 1] __attribute((aligned(8)));
+
+	if (!_vgid_hash || !vgid)
+		return NULL;
+
+	/* vgid not necessarily NULL-terminated */
+	strncpy(&id[0], vgid, ID_LEN);
+	id[ID_LEN] = '\0';
+
+	if (!(vginfo = dm_hash_lookup(_vgid_hash, id)))
+		return NULL;
+
+	return vginfo;
+}
+
+const char *vgname_from_vgid(struct dm_pool *mem, const char *vgid)
+{
+	struct lvmcache_vginfo *vginfo;
+	const char *vgname = NULL;
+
+	if (!*vgid)
+		vgname = ORPHAN;
+
+	if ((vginfo = vginfo_from_vgid(vgid)))
+		vgname = vginfo->vgname;
+
+	if (mem && vgname)
+		return dm_pool_strdup(mem, vgname);
+
+	return vgname;
+}
+
+struct lvmcache_info *info_from_pvid(const char *pvid)
+{
+	struct lvmcache_info *info;
+	char id[ID_LEN + 1] __attribute((aligned(8)));
+
+	if (!_pvid_hash || !pvid)
+		return NULL;
+
+	strncpy(&id[0], pvid, ID_LEN);
+	id[ID_LEN] = '\0';
+
+	if (!(info = dm_hash_lookup(_pvid_hash, id)))
+		return NULL;
+
+	return info;
+}
+
+static void _rescan_entry(struct lvmcache_info *info)
+{
+	struct label *label;
+
+	if (info->status & CACHE_INVALID)
+		label_read(info->dev, &label, UINT64_C(0));
+}
+
+static int _scan_invalid(void)
+{
+	dm_hash_iter(_pvid_hash, (dm_hash_iterate_fn) _rescan_entry);
+
+	return 1;
+}
+
+int lvmcache_label_scan(struct cmd_context *cmd, int full_scan)
+{
+	struct label *label;
+	struct dev_iter *iter;
+	struct device *dev;
+	struct format_type *fmt;
+
+	static int _scanning_in_progress = 0;
+	int r = 0;
+
+	/* Avoid recursion when a PVID can't be found! */
+	if (_scanning_in_progress)
+		return 0;
+
+	_scanning_in_progress = 1;
+
+	if (!_vgname_hash && !lvmcache_init()) {
+		log_error("Internal cache initialisation failed");
+		goto out;
+	}
+
+	if (_has_scanned && !full_scan) {
+		r = _scan_invalid();
+		goto out;
+	}
+
+	if (!(iter = dev_iter_create(cmd->filter, (full_scan == 2) ? 1 : 0))) {
+		log_error("dev_iter creation failed");
+		goto out;
+	}
+
+	while ((dev = dev_iter_get(iter)))
+		label_read(dev, &label, UINT64_C(0));
+
+	dev_iter_destroy(iter);
+
+	_has_scanned = 1;
+
+	/* Perform any format-specific scanning e.g. text files */
+	list_iterate_items(fmt, &cmd->formats) {
+		if (fmt->ops->scan && !fmt->ops->scan(fmt))
+			goto out;
+	}
+
+	r = 1;
+
+      out:
+	_scanning_in_progress = 0;
+
+	return r;
+}
+
+struct list *lvmcache_get_vgids(struct cmd_context *cmd, int full_scan)
+{
+	struct list *vgids;
+	struct lvmcache_vginfo *vginfo;
+
+	lvmcache_label_scan(cmd, full_scan);
+
+	if (!(vgids = str_list_create(cmd->mem))) {
+		log_error("vgids list allocation failed");
+		return NULL;
+	}
+
+	list_iterate_items(vginfo, &_vginfos) {
+		if (!str_list_add(cmd->mem, vgids, 
+				  dm_pool_strdup(cmd->mem, vginfo->vgid))) {
+			log_error("strlist allocation failed");
+			return NULL;
+		}
+	}
+
+	return vgids;
+}
+
+struct list *lvmcache_get_vgnames(struct cmd_context *cmd, int full_scan)
+{
+	struct list *vgnames;
+	struct lvmcache_vginfo *vginfo;
+
+	lvmcache_label_scan(cmd, full_scan);
+
+	if (!(vgnames = str_list_create(cmd->mem))) {
+		log_error("vgnames list allocation failed");
+		return NULL;
+	}
+
+	list_iterate_items(vginfo, &_vginfos) {
+		if (!str_list_add(cmd->mem, vgnames, 
+				  dm_pool_strdup(cmd->mem, vginfo->vgname))) {
+			log_error("strlist allocation failed");
+			return NULL;
+		}
+	}
+
+	return vgnames;
+}
+
+struct list *lvmcache_get_pvids(struct cmd_context *cmd, const char *vgname,
+				const char *vgid)
+{
+	struct list *pvids;
+	struct lvmcache_vginfo *vginfo;
+	struct lvmcache_info *info;
+
+	if (!(pvids = str_list_create(cmd->mem))) {
+		log_error("pvids list allocation failed");
+		return NULL;
+	}
+
+	if (!(vginfo = vginfo_from_vgname(vgname, vgid)))
+		return pvids;
+
+	list_iterate_items(info, &vginfo->infos) {
+		if (!str_list_add(cmd->mem, pvids, 
+				  dm_pool_strdup(cmd->mem, info->dev->pvid))) {
+			log_error("strlist allocation failed");
+			return NULL;
+		}
+	}
+
+	return pvids;
+}
+
+struct device *device_from_pvid(struct cmd_context *cmd, struct id *pvid)
+{
+	struct label *label;
+	struct lvmcache_info *info;
+
+	/* Already cached ? */
+	if ((info = info_from_pvid((char *) pvid))) {
+		if (label_read(info->dev, &label, UINT64_C(0))) {
+			info = (struct lvmcache_info *) label->info;
+			if (id_equal(pvid, (struct id *) &info->dev->pvid))
+				return info->dev;
+		}
+	}
+
+	lvmcache_label_scan(cmd, 0);
+
+	/* Try again */
+	if ((info = info_from_pvid((char *) pvid))) {
+		if (label_read(info->dev, &label, UINT64_C(0))) {
+			info = (struct lvmcache_info *) label->info;
+			if (id_equal(pvid, (struct id *) &info->dev->pvid))
+				return info->dev;
+		}
+	}
+
+	if (memlock())
+		return NULL;
+
+	lvmcache_label_scan(cmd, 2);
+
+	/* Try again */
+	if ((info = info_from_pvid((char *) pvid))) {
+		if (label_read(info->dev, &label, UINT64_C(0))) {
+			info = (struct lvmcache_info *) label->info;
+			if (id_equal(pvid, (struct id *) &info->dev->pvid))
+				return info->dev;
+		}
+	}
+
+	return NULL;
+}
+
+static int _drop_vginfo(struct lvmcache_info *info)
+{
+	if (!list_empty(&info->list)) {
+		list_del(&info->list);
+		list_init(&info->list);
+	}
+
+	if (info->vginfo && list_empty(&info->vginfo->infos)) {
+		dm_hash_remove(_vgname_hash, info->vginfo->vgname);
+		if (info->vginfo->next) {
+                	if (!dm_hash_insert(_vgname_hash, info->vginfo->vgname, info->vginfo->next)) {
+                        	log_error("vg hash re-insertion failed: %s",
+					  info->vginfo->vgname);
+                        	return 0;
+			}
+                }
+
+		if (info->vginfo->vgname)
+			dm_free(info->vginfo->vgname);
+		if (info->vginfo->creation_host)
+			dm_free(info->vginfo->creation_host);
+		if (*info->vginfo->vgid)
+			dm_hash_remove(_vgid_hash, info->vginfo->vgid);
+		list_del(&info->vginfo->list);
+		dm_free(info->vginfo);
+	}
+
+	info->vginfo = NULL;
+
+	return 1;
+}
+
+/* Unused
+void lvmcache_del(struct lvmcache_info *info)
+{
+	if (info->dev->pvid[0] && _pvid_hash)
+		dm_hash_remove(_pvid_hash, info->dev->pvid);
+
+	_drop_vginfo(info);
+
+	info->label->labeller->ops->destroy_label(info->label->labeller,
+						info->label); 
+	dm_free(info);
+
+	return;
+} */
+
+static int _lvmcache_update_pvid(struct lvmcache_info *info, const char *pvid)
+{
+	if (!strcmp(info->dev->pvid, pvid))
+		return 1;
+	if (*info->dev->pvid) {
+		dm_hash_remove(_pvid_hash, info->dev->pvid);
+	}
+	strncpy(info->dev->pvid, pvid, sizeof(info->dev->pvid));
+	if (!dm_hash_insert(_pvid_hash, pvid, info)) {
+		log_error("_lvmcache_update: pvid insertion failed: %s", pvid);
+		return 0;
+	}
+
+	return 1;
+}
+
+static int _lvmcache_update_vgid(struct lvmcache_info *info, const char *vgid)
+{
+	if (!vgid || !info->vginfo ||
+	    !strncmp(info->vginfo->vgid, vgid, ID_LEN))
+		return 1;
+
+	if (info->vginfo && *info->vginfo->vgid)
+		dm_hash_remove(_vgid_hash, info->vginfo->vgid);
+	if (!vgid) {
+		log_debug("lvmcache: %s: clearing VGID", dev_name(info->dev));
+		return 1;
+	}
+
+	strncpy(info->vginfo->vgid, vgid, ID_LEN);
+	info->vginfo->vgid[ID_LEN] = '\0';
+	if (!dm_hash_insert(_vgid_hash, info->vginfo->vgid, info->vginfo)) {
+		log_error("_lvmcache_update: vgid hash insertion failed: %s",
+			  info->vginfo->vgid);
+		return 0;
+	}
+
+	log_debug("lvmcache: %s: setting %s VGID to %s", dev_name(info->dev),
+		  info->vginfo->vgname, info->vginfo->vgid);
+
+	return 1;
+}
+
+static int _insert_vginfo(struct lvmcache_vginfo *new_vginfo, const char *vgid,
+			  uint32_t vgstatus, const char *creation_host,
+			  struct lvmcache_vginfo *primary_vginfo)
+{
+	struct lvmcache_vginfo *last_vginfo = primary_vginfo;
+	char uuid_primary[64] __attribute((aligned(8)));
+	char uuid_new[64] __attribute((aligned(8)));
+	int use_new = 0;
+	
+	/* Pre-existing VG takes precedence. Unexported VG takes precedence. */
+	if (primary_vginfo) {
+		if (!id_write_format((const struct id *)vgid, uuid_new, sizeof(uuid_new)))
+			return_0;
+
+		if (!id_write_format((const struct id *)&primary_vginfo->vgid, uuid_primary,
+				     sizeof(uuid_primary)))
+			return_0;
+
+		/*
+		 * If   Primary not exported, new exported => keep
+		 * Else Primary exported, new not exported => change
+		 * Else Primary has hostname for this machine => keep
+		 * Else Primary has no hostname, new has one => change
+		 * Else New has hostname for this machine => change
+		 * Else Keep primary.
+		 */
+		if (!(primary_vginfo->status & EXPORTED_VG) &&
+		    (vgstatus & EXPORTED_VG))
+			log_error("WARNING: Duplicate VG name %s: "
+				  "Existing %s takes precedence over "
+				  "exported %s", new_vginfo->vgname,
+				  uuid_primary, uuid_new);
+		else if ((primary_vginfo->status & EXPORTED_VG) &&
+			   !(vgstatus & EXPORTED_VG)) {
+			log_error("WARNING: Duplicate VG name %s: "
+				  "%s takes precedence over exported %s",
+				  new_vginfo->vgname, uuid_new,
+				  uuid_primary);
+			use_new = 1;
+		} else if (primary_vginfo->creation_host &&
+			   !strcmp(primary_vginfo->creation_host,
+				   primary_vginfo->fmt->cmd->hostname))
+			log_error("WARNING: Duplicate VG name %s: "
+				  "Existing %s (created here) takes precedence "
+				  "over %s", new_vginfo->vgname, uuid_primary,
+				  uuid_new);
+		else if (!primary_vginfo->creation_host && creation_host) {
+			log_error("WARNING: Duplicate VG name %s: "
+				  "%s (with creation_host) takes precedence over %s",
+				  new_vginfo->vgname, uuid_new,
+				  uuid_primary);
+			use_new = 1;
+		} else if (creation_host &&
+			   !strcmp(creation_host,
+				   primary_vginfo->fmt->cmd->hostname)) {
+			log_error("WARNING: Duplicate VG name %s: "
+				  "%s (created here) takes precedence over %s",
+				  new_vginfo->vgname, uuid_new,
+				  uuid_primary);
+			use_new = 1;
+		}
+
+		if (!use_new) {
+			while (last_vginfo->next)
+				last_vginfo = last_vginfo->next;
+			last_vginfo->next = new_vginfo;
+			return 1;
+		}
+
+		dm_hash_remove(_vgname_hash, primary_vginfo->vgname);
+	}
+
+	if (!dm_hash_insert(_vgname_hash, new_vginfo->vgname, new_vginfo)) {
+		log_error("cache_update: vg hash insertion failed: %s",
+		  	new_vginfo->vgname);
+		return 0;
+	}
+
+	if (primary_vginfo)
+		new_vginfo->next = primary_vginfo;
+
+	return 1;
+}
+
+static int _lvmcache_update_vgname(struct lvmcache_info *info,
+				   const char *vgname, const char *vgid,
+				   uint32_t vgstatus, const char *creation_host)
+{
+	struct lvmcache_vginfo *vginfo, *primary_vginfo;
+	// struct lvmcache_vginfo  *old_vginfo, *next;
+
+	/* If vgname is NULL and we don't already have a vgname, 
+	 * assume ORPHAN - we want every entry to have a vginfo
+	 * attached for scanning reasons.
+	 */
+	if (!vgname && !info->vginfo) {
+		vgname = ORPHAN;
+		vgid = ORPHAN;
+	}
+
+	if (!vgname || (info->vginfo && !strcmp(info->vginfo->vgname, vgname)))
+		return 1;
+
+	/* Remove existing vginfo entry */
+	_drop_vginfo(info);
+
+	/* Get existing vginfo or create new one */
+	if (!(vginfo = vginfo_from_vgname(vgname, vgid))) {
+/*** FIXME - vginfo ends up duplicated instead of renamed.
+		// Renaming?  This lookup fails.
+		if ((vginfo = vginfo_from_vgid(vgid))) {
+			next = vginfo->next;
+			old_vginfo = vginfo_from_vgname(vginfo->vgname, NULL);
+			if (old_vginfo == vginfo) {
+				dm_hash_remove(_vgname_hash, old_vginfo->vgname);
+				if (old_vginfo->next) {
+					if (!dm_hash_insert(_vgname_hash, old_vginfo->vgname, old_vginfo->next)) {
+                        			log_error("vg hash re-insertion failed: %s",
+							  old_vginfo->vgname);
+                        			return 0;
+					}
+				}
+			} else do {
+				if (old_vginfo->next == vginfo) {
+					old_vginfo->next = vginfo->next;
+					break;
+				}
+			} while ((old_vginfo = old_vginfo->next));
+			vginfo->next = NULL;
+
+			dm_free(vginfo->vgname);
+			if (!(vginfo->vgname = dm_strdup(vgname))) {
+				log_error("cache vgname alloc failed for %s", vgname);
+				return 0;
+			}
+
+			// Rename so can assume new name does not already exist 
+			if (!dm_hash_insert(_vgname_hash, vginfo->vgname, vginfo->next)) {
+				log_error("vg hash re-insertion failed: %s",
+					  vginfo->vgname);
+                      		return 0;
+			}
+		} else {
+***/
+		if (!(vginfo = dm_malloc(sizeof(*vginfo)))) {
+			log_error("lvmcache_update_vgname: list alloc failed");
+			return 0;
+		}
+		memset(vginfo, 0, sizeof(*vginfo));
+		if (!(vginfo->vgname = dm_strdup(vgname))) {
+			dm_free(vginfo);
+			log_error("cache vgname alloc failed for %s", vgname);
+			return 0;
+		}
+		list_init(&vginfo->infos);
+			primary_vginfo = vginfo_from_vgname(vgname, NULL);
+		if (!_insert_vginfo(vginfo, vgid, vgstatus, creation_host,
+				    primary_vginfo)) {
+			dm_free(vginfo->vgname);
+			dm_free(vginfo);
+			return 0;
+		}
+		/* Ensure orphans appear last on list_iterate */
+		if (is_orphan_vg(vgname))
+			list_add(&_vginfos, &vginfo->list);
+		else
+			list_add_h(&_vginfos, &vginfo->list);
+/***
+		}
+***/
+	}
+
+	info->vginfo = vginfo;
+	list_add(&vginfo->infos, &info->list);
+
+	/* FIXME Check consistency of list! */
+	vginfo->fmt = info->fmt;
+
+	log_debug("lvmcache: %s: now %s%s%s%s%s", dev_name(info->dev),
+		  !is_orphan_vg(vgname) ? "in VG " : "orphaned", vgname,
+		  vginfo->vgid[0] ? " (" : "",
+		  vginfo->vgid[0] ? vginfo->vgid : "",
+		  vginfo->vgid[0] ? ")" : "");
+
+	return 1;
+}
+
+static int _lvmcache_update_vgstatus(struct lvmcache_info *info, uint32_t vgstatus,
+				     const char *creation_host)
+{
+	if (!info || !info->vginfo)
+		return 1;
+
+	if ((info->vginfo->status & EXPORTED_VG) != (vgstatus & EXPORTED_VG))
+		log_debug("lvmcache: %s: VG %s %s exported",
+			  dev_name(info->dev), info->vginfo->vgname,
+			  vgstatus & EXPORTED_VG ? "now" : "no longer");
+
+	info->vginfo->status = vgstatus;
+
+	if (!creation_host)
+		return 1;
+
+	if (info->vginfo->creation_host && !strcmp(creation_host,
+						   info->vginfo->creation_host))
+		return 1;
+
+	if (info->vginfo->creation_host)
+		dm_free(info->vginfo->creation_host);
+
+	if (!(info->vginfo->creation_host = dm_strdup(creation_host))) {
+		log_error("cache creation host alloc failed for %s",
+			  creation_host);
+		return 0;
+	}
+
+	log_debug("lvmcache: %s: VG %s: Set creation host to %s.",
+		  dev_name(info->dev), info->vginfo->vgname, creation_host);
+
+	return 1;
+}
+
+int lvmcache_update_vgname_and_id(struct lvmcache_info *info,
+				  const char *vgname, const char *vgid,
+				  uint32_t vgstatus, const char *creation_host)
+{
+	if (!_lvmcache_update_vgname(info, vgname, vgid, vgstatus,
+				     creation_host) ||
+	    !_lvmcache_update_vgid(info, vgid) ||
+	    !_lvmcache_update_vgstatus(info, vgstatus, creation_host))
+		return_0;
+
+	return 1;
+}
+
+int lvmcache_update_vg(struct volume_group *vg)
+{
+	struct pv_list *pvl;
+	struct lvmcache_info *info;
+	char pvid_s[ID_LEN + 1] __attribute((aligned(8)));
+
+	pvid_s[sizeof(pvid_s) - 1] = '\0';
+
+	list_iterate_items(pvl, &vg->pvs) {
+		strncpy(pvid_s, (char *) &pvl->pv->id, sizeof(pvid_s) - 1);
+		/* FIXME Could pvl->pv->dev->pvid ever be different? */
+		if ((info = info_from_pvid(pvid_s)) &&
+		    !lvmcache_update_vgname_and_id(info, vg->name,
+						   (char *) &vg->id,
+						   vg->status, NULL))
+			return_0;
+	}
+
+	return 1;
+}
+
+struct lvmcache_info *lvmcache_add(struct labeller *labeller, const char *pvid,
+				   struct device *dev,
+				   const char *vgname, const char *vgid,
+				   uint32_t vgstatus)
+{
+	struct label *label;
+	struct lvmcache_info *existing, *info;
+	char pvid_s[ID_LEN + 1] __attribute((aligned(8)));
+
+	if (!_vgname_hash && !lvmcache_init()) {
+		log_error("Internal cache initialisation failed");
+		return NULL;
+	}
+
+	strncpy(pvid_s, pvid, sizeof(pvid_s));
+	pvid_s[sizeof(pvid_s) - 1] = '\0';
+
+	if (!(existing = info_from_pvid(pvid_s)) &&
+	    !(existing = info_from_pvid(dev->pvid))) {
+		if (!(label = label_create(labeller))) {
+			stack;
+			return NULL;
+		}
+		if (!(info = dm_malloc(sizeof(*info)))) {
+			log_error("lvmcache_info allocation failed");
+			label_destroy(label);
+			return NULL;
+		}
+		memset(info, 0, sizeof(*info));
+
+		label->info = info;
+		info->label = label;
+		list_init(&info->list);
+		info->dev = dev;
+	} else {
+		if (existing->dev != dev) {
+			/* Is the existing entry a duplicate pvid e.g. md ? */
+			if (MAJOR(existing->dev->dev) == md_major() &&
+			    MAJOR(dev->dev) != md_major()) {
+				log_very_verbose("Ignoring duplicate PV %s on "
+						 "%s - using md %s",
+						 pvid, dev_name(dev),
+						 dev_name(existing->dev));
+				return NULL;
+			} else if (dm_is_dm_major(MAJOR(existing->dev->dev)) &&
+				   !dm_is_dm_major(MAJOR(dev->dev))) {
+				log_very_verbose("Ignoring duplicate PV %s on "
+						 "%s - using dm %s",
+						 pvid, dev_name(dev),
+						 dev_name(existing->dev));
+				return NULL;
+			} else if (MAJOR(existing->dev->dev) != md_major() &&
+				   MAJOR(dev->dev) == md_major())
+				log_very_verbose("Duplicate PV %s on %s - "
+						 "using md %s", pvid,
+						 dev_name(existing->dev),
+						 dev_name(dev));
+			else if (!dm_is_dm_major(MAJOR(existing->dev->dev)) &&
+				 dm_is_dm_major(MAJOR(dev->dev)))
+				log_very_verbose("Duplicate PV %s on %s - "
+						 "using dm %s", pvid,
+						 dev_name(existing->dev),
+						 dev_name(dev));
+			/* FIXME If both dm, check dependencies */
+			//else if (dm_is_dm_major(MAJOR(existing->dev->dev)) &&
+				 //dm_is_dm_major(MAJOR(dev->dev)))
+				 //
+			else
+				log_error("Found duplicate PV %s: using %s not "
+					  "%s", pvid, dev_name(dev),
+					  dev_name(existing->dev));
+		}
+		/* Switch over to new preferred device */
+		existing->dev = dev;
+		info = existing;
+		/* Has labeller changed? */
+		if (info->label->labeller != labeller) {
+			label_destroy(info->label);
+			if (!(info->label = label_create(labeller))) {
+				/* FIXME leaves info without label! */
+				stack;
+				return NULL;
+			}
+			info->label->info = info;
+		}
+		label = info->label;
+	}
+
+	info->fmt = (const struct format_type *) labeller->private;
+	info->status |= CACHE_INVALID;
+
+	if (!_lvmcache_update_pvid(info, pvid_s)) {
+		if (!existing) {
+			dm_free(info);
+			label_destroy(label);
+		}
+		return NULL;
+	}
+
+	if (!lvmcache_update_vgname_and_id(info, vgname, vgid, vgstatus, NULL)) {
+		if (!existing) {
+			dm_hash_remove(_pvid_hash, pvid_s);
+			strcpy(info->dev->pvid, "");
+			dm_free(info);
+			label_destroy(label);
+		}
+		return NULL;
+	}
+
+	return info;
+}
+
+static void _lvmcache_destroy_entry(struct lvmcache_info *info)
+{
+	if (!list_empty(&info->list))
+		list_del(&info->list);
+	strcpy(info->dev->pvid, "");
+	label_destroy(info->label);
+	dm_free(info);
+}
+
+static void _lvmcache_destroy_vgnamelist(struct lvmcache_vginfo *vginfo)
+{
+	struct lvmcache_vginfo *next;
+
+	do {
+		next = vginfo->next;
+		if (vginfo->vgname)
+			dm_free(vginfo->vgname);
+		if (vginfo->creation_host)
+			dm_free(vginfo->creation_host);
+		dm_free(vginfo);
+	} while ((vginfo = next));
+}
+
+static void _lvmcache_destroy_lockname(int present __attribute((unused)))
+{
+	/* Nothing to do */
+}
+
+void lvmcache_destroy(void)
+{
+	log_verbose("Wiping internal VG cache");
+
+	_has_scanned = 0;
+
+	if (_vgid_hash) {
+		dm_hash_destroy(_vgid_hash);
+		_vgid_hash = NULL;
+	}
+
+	if (_pvid_hash) {
+		dm_hash_iter(_pvid_hash, (dm_hash_iterate_fn) _lvmcache_destroy_entry);
+		dm_hash_destroy(_pvid_hash);
+		_pvid_hash = NULL;
+	}
+
+	if (_vgname_hash) {
+		dm_hash_iter(_vgname_hash,
+			  (dm_hash_iterate_fn) _lvmcache_destroy_vgnamelist);
+		dm_hash_destroy(_vgname_hash);
+		_vgname_hash = NULL;
+	}
+
+	if (_lock_hash) {
+		dm_hash_iter(_lock_hash, (dm_hash_iterate_fn) _lvmcache_destroy_lockname);
+		dm_hash_destroy(_lock_hash);
+		_lock_hash = NULL;
+	}
+
+	list_init(&_vginfos);
+}

lib/cache/lvmcache.h

@@ -0,0 +1,107 @@
+/*
+ * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004-2006 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#ifndef _LVM_CACHE_H
+#define _LVM_CACHE_H
+
+#include "dev-cache.h"
+#include "uuid.h"
+#include "label.h"
+
+#define ORPHAN ""
+
+#define CACHE_INVALID	0x00000001
+#define CACHE_LOCKED	0x00000002
+
+/* LVM specific per-volume info */
+/* Eventual replacement for struct physical_volume perhaps? */
+
+struct cmd_context;
+struct format_type;
+struct volume_group;
+
+/* One per VG */
+struct lvmcache_vginfo {
+	struct list list;	/* Join these vginfos together */
+	struct list infos;	/* List head for lvmcache_infos */
+	const struct format_type *fmt;
+	char *vgname;		/* "" == orphan */
+	uint32_t status;
+	char vgid[ID_LEN + 1];
+	char _padding[7];
+	struct lvmcache_vginfo *next; /* Another VG with same name? */
+	char *creation_host;
+};
+
+/* One per device */
+struct lvmcache_info {
+	struct list list;	/* Join VG members together */
+	struct list mdas;	/* list head for metadata areas */
+	struct list das;	/* list head for data areas */
+	struct lvmcache_vginfo *vginfo;	/* NULL == unknown */
+	struct label *label;
+	const struct format_type *fmt;
+	struct device *dev;
+	uint64_t device_size;	/* Bytes */
+	uint32_t status;
+};
+
+int lvmcache_init(void);
+void lvmcache_destroy(void);
+
+/* Set full_scan to 1 to reread every filtered device label or
+ * 2 to rescan /dev for new devices */
+int lvmcache_label_scan(struct cmd_context *cmd, int full_scan);
+
+/* Add/delete a device */
+struct lvmcache_info *lvmcache_add(struct labeller *labeller, const char *pvid,
+				   struct device *dev,
+				   const char *vgname, const char *vgid,
+				   uint32_t vgstatus);
+void lvmcache_del(struct lvmcache_info *info);
+
+/* Update things */
+int lvmcache_update_vgname_and_id(struct lvmcache_info *info,
+				  const char *vgname, const char *vgid,
+				  uint32_t vgstatus, const char *hostname);
+int lvmcache_update_vg(struct volume_group *vg);
+
+void lvmcache_lock_vgname(const char *vgname, int read_only);
+void lvmcache_unlock_vgname(const char *vgname);
+
+/* Queries */
+const struct format_type *fmt_from_vgname(const char *vgname, const char *vgid);
+struct lvmcache_vginfo *vginfo_from_vgname(const char *vgname,
+					   const char *vgid);
+struct lvmcache_vginfo *vginfo_from_vgid(const char *vgid);
+struct lvmcache_info *info_from_pvid(const char *pvid);
+const char *vgname_from_vgid(struct dm_pool *mem, const char *vgid);
+struct device *device_from_pvid(struct cmd_context *cmd, struct id *pvid);
+int vgs_locked(void);
+int vgname_is_locked(const char *vgname);
+
+/* Returns list of struct str_lists containing pool-allocated copy of vgnames */
+/* Set full_scan to 1 to reread every filtered device label */
+struct list *lvmcache_get_vgnames(struct cmd_context *cmd, int full_scan);
+
+/* Returns list of struct str_lists containing pool-allocated copy of vgids */
+/* Set full_scan to 1 to reread every filtered device label */
+struct list *lvmcache_get_vgids(struct cmd_context *cmd, int full_scan);
+
+/* Returns list of struct str_lists containing pool-allocated copy of pvids */
+struct list *lvmcache_get_pvids(struct cmd_context *cmd, const char *vgname,
+				const char *vgid);
+
+#endif

lib/commands/errors.h

@@ -1,16 +1,24 @@
 /*
- * Copyright (C) 2001 Sistina Software (UK) Limited.
+ * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
  *
- * This file is released under the LGPL.
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
 #ifndef _LVM_ERRORS_H
 #define _LVM_ERRORS_H
 
-#define EINVALID_CMD_LINE	1
-#define ENO_SUCH_CMD		3
-#define ECMD_PROCESSED		4
+#define ECMD_PROCESSED		1
+#define ENO_SUCH_CMD		2
+#define EINVALID_CMD_LINE	3
 #define ECMD_FAILED		5
 
 #endif
-

lib/commands/toolcontext.h

@@ -1,37 +1,98 @@
 /*
- * Copyright (C) 2001 Sistina Software (UK) Limited.
+ * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved.
  *
- * This file is released under the LGPL.
+ * This file is part of LVM2.
  *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
 #ifndef _LVM_TOOLCONTEXT_H
 #define _LVM_TOOLCONTEXT_H
 
 #include "dev-cache.h"
-#include "config.h"
-#include "pool.h"
-#include "metadata.h"
 
-/* command-instance-related variables needed by library */
-struct cmd_context {
-	/* format handler allocates all objects from here */
-	struct pool *mem;
+#include <stdio.h>
+#include <limits.h>
 
-	struct format_type *fmt;	/* Current format to use by default */
-
-	/* FIXME Move into dynamic list */
-	struct format_type *fmt1;	/* Format1 */
-	struct format_type *fmtt;	/* Format_text */
-
-	char *cmd_line;
-	char *dev_dir;
-	struct dev_filter *filter;
-	struct config_file *cf;
-
-	struct command *command;
-	struct uuid_map *um;
-	struct arg *args;
+/*
+ * Config options that can be changed while commands are processed
+ */
+struct config_info {
+	int debug;
+	int verbose;
+	int test;
+	int syslog;
+	int activation;
+	int suffix;
+	int archive;		/* should we archive ? */
+	int backup;		/* should we backup ? */
+	int read_ahead;		/* DM_READ_AHEAD_NONE or _AUTO */
+	const char *msg_prefix;
+	struct format_type *fmt;
+	uint64_t unit_factor;
+	int cmd_name;		/* Show command name? */
+	mode_t umask;
+	char unit_type;
+	char _padding[1];
 };
 
+struct config_tree;
+struct archive_params;
+struct backup_params;
+
+/* FIXME Split into tool & library contexts */
+/* command-instance-related variables needed by library */
+struct cmd_context {
+	struct dm_pool *libmem;	/* For permanent config data */
+	struct dm_pool *mem;	/* Transient: Cleared between each command */
+
+	const struct format_type *fmt;	/* Current format to use by default */
+	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;
+
+	char *cmd_line;
+	struct command *command;
+	struct arg *args;
+	char **argv;
+	unsigned is_static;	/* Static binary? */
+	unsigned is_long_lived;	/* Optimises persistent_filter handling */
+
+	struct dev_filter *filter;
+	int dump_filter;	/* Dump filter when exiting? */
+
+	struct list config_files;
+	int config_valid;
+	struct config_tree *cft;
+	struct config_tree *cft_override;
+	struct config_info default_settings;
+	struct config_info current_settings;
+
+	struct archive_params *archive_params;
+	struct backup_params *backup_params;
+
+	/* List of defined tags */
+	struct list tags;
+	int hosttags;
+
+	char sys_dir[PATH_MAX];
+	char dev_dir[PATH_MAX];
+	char proc_dir[PATH_MAX];
+};
+
+struct cmd_context *create_toolcontext(struct arg *the_args, unsigned is_static, unsigned is_long_lived);
+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.h

@@ -1,72 +1,113 @@
 /*
- * Copyright (C) 2001 Sistina Software (UK) Limited.
+ * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved.
  *
- * This file is released under the LGPL.
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
 #ifndef _LVM_CONFIG_H
 #define _LVM_CONFIG_H
 
-#include <inttypes.h>
+#include "lvm-types.h"
+
+struct device;
+struct cmd_context;
 
 enum {
-        CFG_STRING,
-        CFG_FLOAT,
-        CFG_INT,
+	CFG_STRING,
+	CFG_FLOAT,
+	CFG_INT,
+	CFG_EMPTY_ARRAY
 };
 
 struct config_value {
-        int type;
-        union {
-                int i;
-                float r;
-                char *str;
-        } v;
-        struct config_value *next; /* for arrays */
+	int type;
+	union {
+		int64_t i;
+		float r;
+		char *str;
+	} v;
+	struct config_value *next;	/* for arrays */
 };
 
 struct config_node {
-        char *key;
-        struct config_node *sib, *child;
-        struct config_value *v;
+	char *key;
+	struct config_node *sib, *child;
+	struct config_value *v;
 };
 
-struct config_file {
-        struct config_node *root;
+struct config_tree {
+	struct config_node *root;
 };
 
-struct config_file *create_config_file(void);
-void destroy_config_file(struct config_file *cf);
+struct config_tree_list {
+	struct list list;
+	struct config_tree *cft;
+};
 
-int read_config(struct config_file *cf, const char *file);
-int write_config(struct config_file *cf, const char *file);
+struct config_tree *create_config_tree(const char *filename, int keep_open);
+struct config_tree *create_config_tree_from_string(struct cmd_context *cmd,
+						   const char *config_settings);
+void destroy_config_tree(struct config_tree *cft);
 
-struct config_node *find_config_node(struct config_node *cn,
-				     const char *path, char seperator);
+typedef uint32_t (*checksum_fn_t) (uint32_t initial, const void *buf, uint32_t size);
 
-const char *find_config_str(struct config_node *cn,
-			    const char *path, char sep, const char *fail);
+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 find_config_int(struct config_node *cn, const char *path,
-		    char sep, int fail);
+int read_config_file(struct config_tree *cft);
+int write_config_file(struct config_tree *cft, const char *file,
+		      int argc, char **argv);
+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);
 
-float find_config_float(struct config_node *cn, const char *path,
-			char sep, float fail);
+struct config_node *find_config_node(const 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(const struct config_node *cn, const char *path, int fail);
+float find_config_float(const struct config_node *cn, const char *path,
+			float fail);
+
+/*
+ * These versions check an override tree, if present, first.
+ */
+struct config_node *find_config_tree_node(struct cmd_context *cmd,
+					  const char *path);
+const char *find_config_tree_str(struct cmd_context *cmd,
+				 const char *path, const char *fail);
+int find_config_tree_int(struct cmd_context *cmd, const char *path,
+			 int fail);
+float find_config_tree_float(struct cmd_context *cmd, 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,
-		    char sep, int fail);
+int find_config_bool(const struct config_node *cn, const char *path, int fail);
+int find_config_tree_bool(struct cmd_context *cmd, const char *path, int fail);
 
+int get_config_uint32(const struct config_node *cn, const char *path,
+		      uint32_t *result);
 
+int get_config_uint64(const struct config_node *cn, const char *path,
+		      uint64_t *result);
 
-int get_config_uint32(struct config_node *cn, const char *path,
-		      char sep, uint32_t *result);
+int get_config_str(const struct config_node *cn, const char *path,
+		   char **result);
 
-int get_config_uint64(struct config_node *cn, const char *path,
-		      char sep, uint64_t *result);
+unsigned maybe_config_section(const char *str, unsigned len);
 
 #endif
-

lib/config/defaults.h

@@ -1,36 +1,124 @@
 /*
- * Copyright (C) 2001 Sistina Software (UK) Limited.
+ * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved.
  *
- * This file is released under the GPL.
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
 #ifndef _LVM_DEFAULTS_H
 #define _LVM_DEFAULTS_H
 
-
-#define DEFAULT_SYS_DIR "/etc/lvm"
-
 #define DEFAULT_ARCHIVE_ENABLED 1
 #define DEFAULT_BACKUP_ENABLED 1
 
 #define DEFAULT_ARCHIVE_SUBDIR "archive"
 #define DEFAULT_BACKUP_SUBDIR "backup"
+#define DEFAULT_CACHE_SUBDIR "cache"
+#define DEFAULT_CACHE_FILE_PREFIX ""
 
 #define DEFAULT_ARCHIVE_DAYS 30
 #define DEFAULT_ARCHIVE_NUMBER 10
 
+#define DEFAULT_SYS_DIR "/etc/lvm"
 #define DEFAULT_DEV_DIR "/dev"
 #define DEFAULT_PROC_DIR "/proc"
+#define DEFAULT_SYSFS_SCAN 1
+#define DEFAULT_MD_COMPONENT_DETECTION 1
+#define DEFAULT_IGNORE_SUSPENDED_DEVICES 1
 
 #define DEFAULT_LOCK_DIR "/var/lock/lvm"
+#define DEFAULT_LOCKING_LIB "liblvm2clusterlock.so"
+#define DEFAULT_FALLBACK_TO_LOCAL_LOCKING 1
+#define DEFAULT_FALLBACK_TO_CLUSTERED_LOCKING 1
+
+#define DEFAULT_MIRRORLOG "disk"
+#define DEFAULT_MIRROR_LOG_FAULT_POLICY "allocate"
+#define DEFAULT_MIRROR_DEV_FAULT_POLICY "remove"
+#define DEFAULT_DMEVENTD_MIRROR_LIB "libdevmapper-event-lvm2mirror.so"
+#define DEFAULT_DMEVENTD_MONITOR 1
 
 #define DEFAULT_UMASK 0077
 
-#define DEFAULT_FORMAT "lvm1"
-
-#ifdef READLINE_SUPPORT
-  #define DEFAULT_MAX_HISTORY 100
+#ifdef LVM1_FALLBACK
+#  define DEFAULT_FALLBACK_TO_LVM1 1
+#else
+#  define DEFAULT_FALLBACK_TO_LVM1 0
 #endif
 
+#ifdef LVM1_SUPPORT
+#  define DEFAULT_FORMAT "lvm1"
+#else
+#  define DEFAULT_FORMAT "lvm2"
+#endif
 
-#endif /* _LVM_DEFAULTS_H */
+#define DEFAULT_STRIPESIZE 64	/* KB */
+#define DEFAULT_PVMETADATASIZE 255
+#define DEFAULT_PVMETADATACOPIES 1
+#define DEFAULT_LABELSECTOR UINT64_C(1)
+#define DEFAULT_READ_AHEAD "auto"
+
+#define DEFAULT_MSG_PREFIX "  "
+#define DEFAULT_CMD_NAME 0
+#define DEFAULT_OVERWRITE 0
+
+#ifndef DEFAULT_LOG_FACILITY
+#  define DEFAULT_LOG_FACILITY LOG_USER
+#endif
+
+#define DEFAULT_SYSLOG 1
+#define DEFAULT_VERBOSE 0
+#define DEFAULT_LOGLEVEL 0
+#define DEFAULT_INDENT 1
+#define DEFAULT_UNITS "h"
+#define DEFAULT_SUFFIX 1
+#define DEFAULT_HOSTTAGS 0
+
+#ifdef DEVMAPPER_SUPPORT
+#  define DEFAULT_ACTIVATION 1
+#  define DEFAULT_RESERVED_MEMORY 8192
+#  define DEFAULT_RESERVED_STACK 256
+#  define DEFAULT_PROCESS_PRIORITY -18
+#else
+#  define DEFAULT_ACTIVATION 0
+#endif
+
+#define DEFAULT_STRIPE_FILLER "/dev/ioerror"
+#define DEFAULT_MIRROR_REGION_SIZE 512	/* KB */
+#define DEFAULT_INTERVAL 15
+
+#ifdef READLINE_SUPPORT
+#  define DEFAULT_MAX_HISTORY 100
+#endif
+
+#define DEFAULT_REP_ALIGNED 1
+#define DEFAULT_REP_BUFFERED 1
+#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,mirror_log,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_PVSEGS_COLS "pv_name,vg_name,pv_fmt,pv_attr,pv_size,pv_free,pvseg_start,pvseg_size"
+
+#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,mirror_log,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,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_PVSEGS_COLS_VERB "pv_name,vg_name,pv_fmt,pv_attr,pv_size,pv_free,pvseg_start,pvseg_size"
+
+#define DEFAULT_LVS_SORT "vg_name,lv_name"
+#define DEFAULT_VGS_SORT "vg_name"
+#define DEFAULT_PVS_SORT "pv_name"
+#define DEFAULT_SEGS_SORT "vg_name,lv_name,seg_start"
+#define DEFAULT_PVSEGS_SORT "pv_name,pvseg_start"
+
+#endif				/* _LVM_DEFAULTS_H */

lib/datastruct/bitset.c

@@ -1,79 +0,0 @@
-/*
- * Copyright (C) 2001 Sistina Software (UK) Limited.
- *
- * This file is released under the LGPL.
- */
-
-#include "bitset.h"
-#include "dbg_malloc.h"
-
-#include <stdlib.h>
-
-/* FIXME: calculate this. */
-#define INT_SHIFT 5
-
-bitset_t bitset_create(struct pool * mem, unsigned num_bits)
-{
-	int n = (num_bits / BITS_PER_INT) + 2;
-	int size = sizeof(int) * n;
-	unsigned *bs = pool_zalloc(mem, size);
-
-	if (!bs)
-		return NULL;
-
-	*bs = num_bits;
-	return bs;
-}
-
-void bitset_destroy(bitset_t bs)
-{
-	dbg_free(bs);
-}
-
-void bit_union(bitset_t out, bitset_t in1, bitset_t in2)
-{
-	int i;
-	for (i = (in1[0] / BITS_PER_INT) + 1; i; i--)
-		out[i] = in1[i] | in2[i];
-}
-
-/*
- * FIXME: slow
- */
-static inline int _test_word(uint32_t test, int bit)
-{
-	while (bit < BITS_PER_INT) {
-		if (test & (0x1 << bit))
-			return bit;
-		bit++;
-	}
-
-	return -1;
-}
-
-int bit_get_next(bitset_t bs, int last_bit)
-{
-	int bit, word;
-	uint32_t test;
-
-	last_bit++;		/* otherwise we'll return the same bit again */
-
-	while (last_bit < bs[0]) {
-		word = last_bit >> INT_SHIFT;
-		test = bs[word + 1];
-		bit = last_bit & (BITS_PER_INT - 1);
-
-		if ((bit = _test_word(test, bit)) >= 0)
-			return (word * BITS_PER_INT) + bit;
-
-		last_bit = last_bit - (last_bit & (BITS_PER_INT - 1)) +
-		    BITS_PER_INT;
-	}
-
-	return -1;
-}
-
-int bit_get_first(bitset_t bs)
-{
-	return bit_get_next(bs, -1);
-}

lib/datastruct/btree.c

@@ -1,11 +1,20 @@
 /*
- * Copyright (C) 2001 Sistina Software (UK) Limited.
+ * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved.
  *
- * This file is released under the LGPL.
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
+#include "lib.h"
 #include "btree.h"
-#include "log.h"
 
 struct node {
 	uint32_t key;
@@ -15,13 +24,13 @@ struct node {
 };
 
 struct btree {
-	struct pool *mem;
+	struct dm_pool *mem;
 	struct node *root;
 };
 
-struct btree *btree_create(struct pool *mem)
+struct btree *btree_create(struct dm_pool *mem)
 {
-	struct btree *t = pool_alloc(mem, sizeof(*t));
+	struct btree *t = dm_pool_alloc(mem, sizeof(*t));
 
 	if (t) {
 		t->mem = mem;
@@ -46,7 +55,8 @@ static uint32_t _shuffle(uint32_t k)
 #endif
 }
 
-struct node **_lookup(struct node **c, uint32_t key, struct node **p)
+static struct node **_lookup(struct node *const *c, uint32_t key,
+			     struct node **p)
 {
 	*p = NULL;
 	while (*c) {
@@ -61,10 +71,10 @@ struct node **_lookup(struct node **c, uint32_t key, struct node **p)
 			c = &(*c)->r;
 	}
 
-	return c;
+	return (struct node **)c;
 }
 
-void *btree_lookup(struct btree *t, uint32_t k)
+void *btree_lookup(const struct btree *t, uint32_t k)
 {
 	uint32_t key = _shuffle(k);
 	struct node *p, **c = _lookup(&t->root, key, &p);
@@ -77,7 +87,7 @@ int btree_insert(struct btree *t, uint32_t k, void *data)
 	struct node *p, **c = _lookup(&t->root, key, &p), *n;
 
 	if (!*c) {
-		if (!(n = pool_alloc(t->mem, sizeof(*n)))) {
+		if (!(n = dm_pool_alloc(t->mem, sizeof(*n)))) {
 			stack;
 			return 0;
 		}
@@ -93,19 +103,19 @@ int btree_insert(struct btree *t, uint32_t k, void *data)
 	return 1;
 }
 
-void *btree_get_data(struct btree_iter *it)
+void *btree_get_data(const struct btree_iter *it)
 {
 	return ((struct node *) it)->data;
 }
 
-static inline struct node *_left(struct node *n)
+static struct node *_left(struct node *n)
 {
 	while (n->l)
 		n = n->l;
 	return n;
 }
 
-struct btree_iter *btree_first(struct btree *t)
+struct btree_iter *btree_first(const struct btree *t)
 {
 	if (!t->root)
 		return NULL;
@@ -113,7 +123,7 @@ struct btree_iter *btree_first(struct btree *t)
 	return (struct btree_iter *) _left(t->root);
 }
 
-struct btree_iter *btree_next(struct btree_iter *it)
+struct btree_iter *btree_next(const struct btree_iter *it)
 {
 	struct node *n = (struct node *) it;
 	uint32_t k = n->key;

lib/datastruct/btree.h

@@ -1,26 +1,32 @@
 /*
- * Copyright (C) 2001 Sistina Software (UK) Limited.
+ * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004-2005 Red Hat, Inc. All rights reserved.
  *
- * This file is released under the GPL.
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
 #ifndef _LVM_BTREE_H
 #define _LVM_BTREE_H
 
-#include "lvm-types.h"
-#include "pool.h"
-
 struct btree;
 
-struct btree *btree_create(struct pool *mem);
+struct btree *btree_create(struct dm_pool *mem);
 
-void *btree_lookup(struct btree *t, uint32_t k);
+void *btree_lookup(const struct btree *t, uint32_t k);
 int btree_insert(struct btree *t, uint32_t k, void *data);
 
 struct btree_iter;
-void *btree_get_data(struct btree_iter *it);
+void *btree_get_data(const struct btree_iter *it);
 
-struct btree_iter *btree_first(struct btree *t);
-struct btree_iter *btree_next(struct btree_iter *it);
+struct btree_iter *btree_first(const struct btree *t);
+struct btree_iter *btree_next(const struct btree_iter *it);
 
 #endif

lib/datastruct/hash.c

@@ -1,232 +0,0 @@
-/*
- * Copyright (C) 2001 Sistina Software
- *
- * This file is released under the LGPL.
- */
-
-#include "dbg_malloc.h"
-#include "hash.h"
-#include "log.h"
-
-struct hash_node {
-	struct hash_node *next;
-	void *data;
-	char key[1];
-};
-
-struct hash_table {
-	int num_nodes;
-	int num_slots;
-	struct hash_node **slots;
-};
-
-/* Permutation of the Integers 0 through 255 */
-static unsigned char _nums[] = {
-	1, 14, 110, 25, 97, 174, 132, 119, 138, 170, 125, 118, 27, 233, 140, 51,
-	87, 197, 177, 107, 234, 169, 56, 68, 30, 7, 173, 73, 188, 40, 36, 65,
-	49, 213, 104, 190, 57, 211, 148, 223, 48, 115, 15, 2, 67, 186, 210, 28,
-	12, 181, 103, 70, 22, 58, 75, 78, 183, 167, 238, 157, 124, 147, 172,
-	    144,
-	176, 161, 141, 86, 60, 66, 128, 83, 156, 241, 79, 46, 168, 198, 41, 254,
-	178, 85, 253, 237, 250, 154, 133, 88, 35, 206, 95, 116, 252, 192, 54,
-	    221,
-	102, 218, 255, 240, 82, 106, 158, 201, 61, 3, 89, 9, 42, 155, 159, 93,
-	166, 80, 50, 34, 175, 195, 100, 99, 26, 150, 16, 145, 4, 33, 8, 189,
-	121, 64, 77, 72, 208, 245, 130, 122, 143, 55, 105, 134, 29, 164, 185,
-	    194,
-	193, 239, 101, 242, 5, 171, 126, 11, 74, 59, 137, 228, 108, 191, 232,
-	    139,
-	6, 24, 81, 20, 127, 17, 91, 92, 251, 151, 225, 207, 21, 98, 113, 112,
-	84, 226, 18, 214, 199, 187, 13, 32, 94, 220, 224, 212, 247, 204, 196,
-	    43,
-	249, 236, 45, 244, 111, 182, 153, 136, 129, 90, 217, 202, 19, 165, 231,
-	    71,
-	230, 142, 96, 227, 62, 179, 246, 114, 162, 53, 160, 215, 205, 180, 47,
-	    109,
-	44, 38, 31, 149, 135, 0, 216, 52, 63, 23, 37, 69, 39, 117, 146, 184,
-	163, 200, 222, 235, 248, 243, 219, 10, 152, 131, 123, 229, 203, 76, 120,
-	    209
-};
-
-static struct hash_node *_create_node(const char *str)
-{
-	/* 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));
-
-	if (n)
-		strcpy(n->key, str);
-
-	return n;
-}
-
-static unsigned _hash(const char *str)
-{
-	unsigned long int h = 0, g;
-	while (*str) {
-		h <<= 4;
-		h += _nums[(int) *str++];
-		g = h & ((unsigned long) 0xf << 16u);
-		if (g) {
-			h ^= g >> 16u;
-			h ^= g >> 5u;
-		}
-	}
-	return h;
-}
-
-struct hash_table *hash_create(unsigned size_hint)
-{
-	size_t len;
-	unsigned new_size = 16u;
-	struct hash_table *hc = dbg_malloc(sizeof(*hc));
-
-	if (!hc) {
-		stack;
-		return 0;
-	}
-
-	memset(hc, 0, sizeof(*hc));
-
-	/* round size hint up to a power of two */
-	while (new_size < size_hint)
-		new_size = new_size << 1;
-
-	hc->num_slots = new_size;
-	len = sizeof(*(hc->slots)) * new_size;
-	if (!(hc->slots = dbg_malloc(len))) {
-		stack;
-		goto bad;
-	}
-	memset(hc->slots, 0, len);
-	return hc;
-
-      bad:
-	dbg_free(hc->slots);
-	dbg_free(hc);
-	return 0;
-}
-
-static void _free_nodes(struct hash_table *t)
-{
-	struct hash_node *c, *n;
-	int i;
-
-	for (i = 0; i < t->num_slots; i++)
-		for (c = t->slots[i]; c; c = n) {
-			n = c->next;
-			dbg_free(c);
-		}
-}
-
-void hash_destroy(struct hash_table *t)
-{
-	_free_nodes(t);
-	dbg_free(t->slots);
-	dbg_free(t);
-}
-
-static inline struct hash_node **_find(struct hash_table *t, const char *key)
-{
-	unsigned h = _hash(key) & (t->num_slots - 1);
-	struct hash_node **c;
-
-	for (c = &t->slots[h]; *c; c = &((*c)->next))
-		if (!strcmp(key, (*c)->key))
-			break;
-
-	return c;
-}
-
-void *hash_lookup(struct hash_table *t, const char *key)
-{
-	struct hash_node **c = _find(t, key);
-	return *c ? (*c)->data : 0;
-}
-
-int hash_insert(struct hash_table *t, const char *key, void *data)
-{
-	struct hash_node **c = _find(t, key);
-
-	if (*c)
-		(*c)->data = data;
-	else {
-		struct hash_node *n = _create_node(key);
-
-		if (!n)
-			return 0;
-
-		n->data = data;
-		n->next = 0;
-		*c = n;
-		t->num_nodes++;
-	}
-
-	return 1;
-}
-
-void hash_remove(struct hash_table *t, const char *key)
-{
-	struct hash_node **c = _find(t, key);
-
-	if (*c) {
-		struct hash_node *old = *c;
-		*c = (*c)->next;
-		dbg_free(old);
-		t->num_nodes--;
-	}
-}
-
-unsigned hash_get_num_entries(struct hash_table *t)
-{
-	return t->num_nodes;
-}
-
-void hash_iter(struct hash_table *t, iterate_fn f)
-{
-	struct hash_node *c;
-	int i;
-
-	for (i = 0; i < t->num_slots; i++)
-		for (c = t->slots[i]; c; c = c->next)
-			f(c->data);
-}
-
-void hash_wipe(struct hash_table *t)
-{
-	_free_nodes(t);
-	memset(t->slots, 0, sizeof(struct hash_node *) * t->num_slots);
-	t->num_nodes = 0;
-}
-
-char *hash_get_key(struct hash_table *t, struct hash_node *n)
-{
-	return n->key;
-}
-
-void *hash_get_data(struct hash_table *t, struct hash_node *n)
-{
-	return n->data;
-}
-
-static struct hash_node *_next_slot(struct hash_table *t, unsigned int s)
-{
-	struct hash_node *c = NULL;
-	int i;
-
-	for (i = s; i < t->num_slots && !c; i++)
-		c = t->slots[i];
-
-	return c;
-}
-
-struct hash_node *hash_get_first(struct hash_table *t)
-{
-	return _next_slot(t, 0);
-}
-
-struct hash_node *hash_get_next(struct hash_table *t, struct hash_node *n)
-{
-	unsigned int h = _hash(n->key) & (t->num_slots - 1);
-	return n->next ? n->next : _next_slot(t, h + 1);
-}

lib/datastruct/hash.h

@@ -1,36 +0,0 @@
-/*
- * Copyright (C) 2001 Sistina Software
- *
- * This file is released under the GPL.
- */
-
-#ifndef _LVM_HASH_H
-#define _LVM_HASH_H
-
-struct hash_table;
-struct hash_node;
-
-typedef void (*iterate_fn)(void *data);
-
-struct hash_table *hash_create(unsigned size_hint);
-void hash_destroy(struct hash_table *t);
-void hash_wipe(struct hash_table *t);
-
-void *hash_lookup(struct hash_table *t, const char *key);
-int hash_insert(struct hash_table *t, const char *key, void *data);
-void hash_remove(struct hash_table *t, const char *key);
-
-unsigned hash_get_num_entries(struct hash_table *t);
-void hash_iter(struct hash_table *t, iterate_fn f);
-
-char *hash_get_key(struct hash_table *t, struct hash_node *n);
-void *hash_get_data(struct hash_table *t, struct hash_node *n);
-struct hash_node *hash_get_first(struct hash_table *t);
-struct hash_node *hash_get_next(struct hash_table *t, struct hash_node *n);
-
-#define hash_iterate(v, h) \
-	for (v = hash_get_first(h); v; \
-	     v = hash_get_next(h, v))
-
-#endif
-

lib/datastruct/list.c

@@ -0,0 +1,136 @@
+/*
+ * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.
+ * Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include "lib.h"
+
+/*
+ * Initialise a list before use.
+ * The list head's next and previous pointers point back to itself.
+ */
+void list_init(struct list *head)
+{
+	head->n = head->p = head;
+}
+
+/*
+ * Insert an element before 'head'.
+ * If 'head' is the list head, this adds an element to the end of the list.
+ */
+void list_add(struct list *head, struct list *elem)
+{
+	assert(head->n);
+
+	elem->n = head;
+	elem->p = head->p;
+
+	head->p->n = elem;
+	head->p = elem;
+}
+
+/*
+ * Insert an element after 'head'.
+ * If 'head' is the list head, this adds an element to the front of the list.
+ */
+void list_add_h(struct list *head, struct list *elem)
+{
+	assert(head->n);
+
+	elem->n = head->n;
+	elem->p = head;
+
+	head->n->p = elem;
+	head->n = elem;
+}
+
+/*
+ * Delete an element from its list.
+ * Note that this doesn't change the element itself - it may still be safe
+ * to follow its pointers.
+ */
+void list_del(struct list *elem)
+{
+	elem->n->p = elem->p;
+	elem->p->n = elem->n;
+}
+
+/*
+ * Is the list empty?
+ */
+int list_empty(const struct list *head)
+{
+	return head->n == head;
+}
+
+/*
+ * Is this the first element of the list?
+ */
+int list_start(const struct list *head, const struct list *elem)
+{
+	return elem->p == head;
+}
+
+/*
+ * Is this the last element of the list?
+ */
+int list_end(const struct list *head, const struct list *elem)
+{
+	return elem->n == head;
+}
+
+/*
+ * Return first element of the list or NULL if empty
+ */
+struct list *list_first(const struct list *head)
+{
+	return (list_empty(head) ? NULL : head->n);
+}
+
+/*
+ * Return last element of the list or NULL if empty
+ */
+struct list *list_last(const struct list *head)
+{
+	return (list_empty(head) ? NULL : head->p);
+}
+
+/*
+ * Return the previous element of the list, or NULL if we've reached the start.
+ */
+struct list *list_prev(const struct list *head, const struct list *elem)
+{
+	return (list_start(head, elem) ? NULL : elem->p);
+}
+
+/*
+ * Return the next element of the list, or NULL if we've reached the end.
+ */
+struct list *list_next(const struct list *head, const struct list *elem)
+{
+	return (list_end(head, elem) ? NULL : elem->n);
+}
+
+/*
+ * Return the number of elements in a list by walking it.
+ */
+unsigned int list_size(const struct list *head)
+{
+	unsigned int s = 0;
+	const struct list *v;
+
+	list_iterate(v, head)
+	    s++;
+
+	return s;
+}

lib/datastruct/list.h

@@ -1,7 +1,16 @@
 /*
- * Copyright (C) 2001 Sistina Software
+ * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.
+ * Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved.
  *
- * This file is released under the LGPL.
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
 #ifndef _LVM_LIST_H
@@ -9,64 +18,186 @@
 
 #include <assert.h>
 
+/*
+ * A list consists of a list head plus elements.
+ * Each element has 'next' and 'previous' pointers.
+ * The list head's pointers point to the first and the last element.
+ */
+
 struct list {
 	struct list *n, *p;
 };
 
-static inline void list_init(struct list *head) {
-	head->n = head->p = head;
-}
+/*
+ * Initialise a list before use.
+ * The list head's next and previous pointers point back to itself.
+ */
+#define LIST_INIT(name)	struct list name = { &(name), &(name) }
+void list_init(struct list *head);
 
-static inline void list_add(struct list *head, struct list *elem) {
-	assert(head->n);
+/*
+ * Insert an element before 'head'.
+ * If 'head' is the list head, this adds an element to the end of the list.
+ */
+void list_add(struct list *head, struct list *elem);
 
-	elem->n = head;
-	elem->p = head->p;
+/*
+ * Insert an element after 'head'.
+ * If 'head' is the list head, this adds an element to the front of the list.
+ */
+void list_add_h(struct list *head, struct list *elem);
 
-	head->p->n = elem;
-	head->p = elem;
-}
+/*
+ * Delete an element from its list.
+ * Note that this doesn't change the element itself - it may still be safe
+ * to follow its pointers.
+ */
+void list_del(struct list *elem);
 
-static inline void list_add_h(struct list *head, struct list *elem) {
-	assert(head->n);
+/*
+ * Is the list empty?
+ */
+int list_empty(const struct list *head);
 
-	elem->n = head->n;
-	elem->p = head;
+/*
+ * Is this the first element of the list?
+ */
+int list_start(const struct list *head, const struct list *elem);
 
-	head->n->p = elem;
-	head->n = elem;
-}
+/*
+ * Is this the last element of the list?
+ */
+int list_end(const struct list *head, const struct list *elem);
 
-static inline void list_del(struct list *elem) {
-	elem->n->p = elem->p;
-	elem->p->n = elem->n;
-}
+/*
+ * Return first element of the list or NULL if empty
+ */
+struct list *list_first(const struct list *head);
 
-static inline int list_empty(struct list *head) {
-	return head->n == head;
-}
+/*
+ * Return last element of the list or NULL if empty
+ */
+struct list *list_last(const struct list *head);
 
+/*
+ * Return the previous element of the list, or NULL if we've reached the start.
+ */
+struct list *list_prev(const struct list *head, const struct list *elem);
+
+/*
+ * Return the next element of the list, or NULL if we've reached the end.
+ */
+struct list *list_next(const struct list *head, const struct list *elem);
+
+/*
+ * Given the address v of an instance of 'struct list' called 'head' 
+ * contained in a structure of type t, return the containing structure.
+ */
+#define list_struct_base(v, t, head) \
+    ((t *)((uintptr_t)(v) - (uintptr_t)&((t *) 0)->head))
+
+/*
+ * Given the address v of an instance of 'struct list list' contained in
+ * a structure of type t, return the containing structure.
+ */
+#define list_item(v, t) list_struct_base((v), t, list)
+
+/*
+ * Given the address v of one known element e in a known structure of type t,
+ * return another element f.
+ */
+#define struct_field(v, t, e, f) \
+    (((t *)((uintptr_t)(v) - (uintptr_t)&((t *) 0)->e))->f)
+
+/*
+ * Given the address v of a known element e in a known structure of type t,
+ * return the list head 'list'
+ */
+#define list_head(v, t, e) struct_field(v, t, e, list)
+
+/*
+ * Set v to each element of a list in turn.
+ */
 #define list_iterate(v, head) \
 	for (v = (head)->n; v != head; v = v->n)
 
+/*
+ * Set v to each element in a list in turn, starting from the element 
+ * in front of 'start'.
+ * You can use this to 'unwind' a list_iterate and back out actions on
+ * already-processed elements.
+ * If 'start' is 'head' it walks the list backwards.
+ */
+#define list_uniterate(v, head, start) \
+	for (v = (start)->p; v != head; v = v->p)
+
+/*
+ * A safe way to walk a list and delete and free some elements along
+ * the way.
+ * t must be defined as a temporary variable of the same type as v.
+ */
 #define list_iterate_safe(v, t, head) \
 	for (v = (head)->n, t = v->n; v != head; v = t, t = v->n)
 
-static inline int list_size(struct list *head) {
-	int s = 0;
-	struct list *v;
+/*
+ * Walk a list, setting 'v' in turn to the containing structure of each item.
+ * The containing structure should be the same type as 'v'.
+ * The 'struct list' variable within the containing structure is 'field'.
+ */
+#define list_iterate_items_gen(v, head, field) \
+	for (v = list_struct_base((head)->n, typeof(*v), field); \
+	     &v->field != (head); \
+	     v = list_struct_base(v->field.n, typeof(*v), field))
 
-	list_iterate(v, head)
-		s++;
+/*
+ * Walk a list, setting 'v' in turn to the containing structure of each item.
+ * The containing structure should be the same type as 'v'.
+ * The list should be 'struct list list' within the containing structure.
+ */
+#define list_iterate_items(v, head) list_iterate_items_gen(v, (head), list)
 
-	return s;
-}
+/*
+ * Walk a list, setting 'v' in turn to the containing structure of each item.
+ * The containing structure should be the same type as 'v'.
+ * The 'struct list' variable within the containing structure is 'field'.
+ * t must be defined as a temporary variable of the same type as v.
+ */
+#define list_iterate_items_gen_safe(v, t, head, field) \
+	for (v = list_struct_base((head)->n, typeof(*v), field), \
+	     t = list_struct_base(v->field.n, typeof(*v), field); \
+	     &v->field != (head); \
+	     v = t, t = list_struct_base(v->field.n, typeof(*v), field))
+/*
+ * Walk a list, setting 'v' in turn to the containing structure of each item.
+ * The containing structure should be the same type as 'v'.
+ * The list should be 'struct list list' within the containing structure.
+ * t must be defined as a temporary variable of the same type as v.
+ */
+#define list_iterate_items_safe(v, t, head) \
+	list_iterate_items_gen_safe(v, t, (head), list)
 
-#define list_item(v, t) \
-    ((t *)((uintptr_t)(v) - (uintptr_t)&((t *) 0)->list))
+/*
+ * Walk a list backwards, setting 'v' in turn to the containing structure 
+ * of each item.
+ * The containing structure should be the same type as 'v'.
+ * The 'struct list' variable within the containing structure is 'field'.
+ */
+#define list_iterate_back_items_gen(v, head, field) \
+	for (v = list_struct_base((head)->p, typeof(*v), field); \
+	     &v->field != (head); \
+	     v = list_struct_base(v->field.p, typeof(*v), field))
 
-/* Given a known element in a known structure, locate the struct list */
-#define list_head(v, t, e) \
-    (((t *)((uintptr_t)(v) - (uintptr_t)&((t *) 0)->e))->list)
+/*
+ * Walk a list backwards, setting 'v' in turn to the containing structure 
+ * of each item.
+ * The containing structure should be the same type as 'v'.
+ * The list should be 'struct list list' within the containing structure.
+ */
+#define list_iterate_back_items(v, head) list_iterate_back_items_gen(v, (head), list)
+
+/*
+ * Return the number of elements in a list by walking it.
+ */
+unsigned int list_size(const struct list *head);
 
 #endif

lib/datastruct/lvm-types.h

@@ -1,7 +1,16 @@
 /*
- * Copyright (C) 2001 Sistina Software (UK) Limited.
+ * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved.
  *
- * This file is released under the GPL.
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
 #ifndef _LVM_TYPES_H
@@ -12,9 +21,13 @@
 #include <sys/types.h>
 #include <inttypes.h>
 
+/* Define some portable printing types */
+#define PRIsize_t "zu"
+#define PRIptrdiff_t "td"
+
 struct str_list {
 	struct list list;
-	char *str;
+	const char *str;
 };
 
 #endif

lib/datastruct/str_list.c

@@ -0,0 +1,129 @@
+/*
+ * Copyright (C) 2003-2004 Sistina Software, Inc. All rights reserved.
+ * Copyright (C) 2004-2005 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include "lib.h"
+#include "str_list.h"
+
+struct list *str_list_create(struct dm_pool *mem)
+{
+	struct list *sl;
+
+	if (!(sl = dm_pool_alloc(mem, sizeof(struct list)))) {
+		stack;
+		return NULL;
+	}
+
+	list_init(sl);
+
+	return sl;
+}
+
+int str_list_add(struct dm_pool *mem, struct list *sll, const char *str)
+{
+	struct str_list *sln;
+
+	if (!str) {
+		stack;
+		return 0;
+	}
+
+	/* Already in list? */
+	if (str_list_match_item(sll, str))
+		return 1;
+
+	if (!(sln = dm_pool_alloc(mem, sizeof(*sln)))) {
+		stack;
+		return 0;
+	}
+
+	sln->str = str;
+	list_add(sll, &sln->list);
+
+	return 1;
+}
+
+int str_list_del(struct list *sll, const char *str)
+{
+	struct list *slh, *slht;
+
+	list_iterate_safe(slh, slht, sll) {
+		if (!strcmp(str, list_item(slh, struct str_list)->str))
+			 list_del(slh);
+	}
+
+	return 1;
+}
+
+int str_list_dup(struct dm_pool *mem, struct list *sllnew,
+		 const struct list *sllold)
+{
+	struct str_list *sl;
+
+	list_init(sllnew);
+
+	list_iterate_items(sl, sllold) {
+		if (!str_list_add(mem, sllnew, dm_pool_strdup(mem, sl->str))) {
+			stack;
+			return 0;
+		}
+	}
+
+	return 1;
+}
+
+/*
+ * Is item on list?
+ */
+int str_list_match_item(const struct list *sll, const char *str)
+{
+	struct str_list *sl;
+
+	list_iterate_items(sl, sll)
+	    if (!strcmp(str, sl->str))
+		return 1;
+
+	return 0;
+}
+
+/*
+ * Is at least one item on both lists?
+ */
+int str_list_match_list(const struct list *sll, const struct list *sll2)
+{
+	struct str_list *sl;
+
+	list_iterate_items(sl, sll)
+	    if (str_list_match_item(sll2, sl->str))
+		return 1;
+
+	return 0;
+}
+
+/*
+ * Do both lists contain the same set of items?
+ */
+int str_list_lists_equal(const struct list *sll, const struct list *sll2)
+{
+	struct str_list *sl;
+
+	if (list_size(sll) != list_size(sll2))
+		return 0;
+
+	list_iterate_items(sl, sll)
+	    if (!str_list_match_item(sll2, sl->str))
+		return 0;
+
+	return 1;
+}

lib/datastruct/str_list.h

@@ -0,0 +1,28 @@
+/*
+ * Copyright (C) 2003-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004-2005 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#ifndef _LVM_STR_LIST_H
+#define _LVM_STR_LIST_H
+
+struct list *str_list_create(struct dm_pool *mem);
+int str_list_add(struct dm_pool *mem, struct list *sll, const char *str);
+int str_list_del(struct list *sll, const char *str);
+int str_list_match_item(const struct list *sll, const char *str);
+int str_list_match_list(const struct list *sll, const struct list *sll2);
+int str_list_lists_equal(const struct list *sll, const struct list *sll2);
+int str_list_dup(struct dm_pool *mem, struct list *sllnew,
+		 const struct list *sllold);
+
+#endif

lib/device/dev-cache.c

@@ -1,30 +1,29 @@
 /*
- * Copyright (C) 2001 Sistina Software (UK) Limited.
+ * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved.
  *
- * This file is released under the LGPL.
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
+#include "lib.h"
 #include "dev-cache.h"
-#include "log.h"
-#include "pool.h"
-#include "hash.h"
-#include "list.h"
 #include "lvm-types.h"
 #include "btree.h"
-#include "dbg_malloc.h"
+#include "filter.h"
+#include "filter-persistent.h"
+#include "toolcontext.h"
 
-#include <stdlib.h>
-#include <sys/types.h>
-#include <sys/stat.h>
 #include <unistd.h>
 #include <sys/param.h>
 #include <dirent.h>
-#include <linux/kdev_t.h>
-
-/*
- * FIXME: really need to seperate names from the devices since
- * multiple names can point to the same device.
- */
 
 struct dev_iter {
 	struct btree_iter *current;
@@ -37,50 +36,238 @@ struct dir_list {
 };
 
 static struct {
-	struct pool *mem;
-	struct hash_table *names;
+	struct dm_pool *mem;
+	struct dm_hash_table *names;
 	struct btree *devices;
+	struct dm_regex *preferred_names_matcher;
 
 	int has_scanned;
 	struct list dirs;
+	struct list files;
 
 } _cache;
 
-#define _alloc(x) pool_alloc(_cache.mem, (x))
-#define _free(x) pool_free(_cache.mem, (x))
+#define _alloc(x) dm_pool_zalloc(_cache.mem, (x))
+#define _free(x) dm_pool_free(_cache.mem, (x))
+#define _strdup(x) dm_pool_strdup(_cache.mem, (x))
 
 static int _insert(const char *path, int rec);
 
-static struct device *_create_dev(dev_t d)
+struct device *dev_create_file(const char *filename, struct device *dev,
+			       struct str_list *alias, int use_malloc)
+{
+	int allocate = !dev;
+
+	if (allocate) {
+		if (use_malloc) {
+			if (!(dev = dm_malloc(sizeof(*dev)))) {
+				log_error("struct device allocation failed");
+				return NULL;
+			}
+			if (!(alias = dm_malloc(sizeof(*alias)))) {
+				log_error("struct str_list allocation failed");
+				dm_free(dev);
+				return NULL;
+			}
+			if (!(alias->str = dm_strdup(filename))) {
+				log_error("filename strdup failed");
+				dm_free(dev);
+				dm_free(alias);
+				return NULL;
+			}
+			dev->flags = DEV_ALLOCED;
+		} else {
+			if (!(dev = _alloc(sizeof(*dev)))) {
+				log_error("struct device allocation failed");
+				return NULL;
+			}
+			if (!(alias = _alloc(sizeof(*alias)))) {
+				log_error("struct str_list allocation failed");
+				_free(dev);
+				return NULL;
+			}
+			if (!(alias->str = _strdup(filename))) {
+				log_error("filename strdup failed");
+				return NULL;
+			}
+		}
+	} else if (!(alias->str = dm_strdup(filename))) {
+		log_error("filename strdup failed");
+		return NULL;
+	}
+
+	dev->flags |= DEV_REGULAR;
+	list_init(&dev->aliases);
+	list_add(&dev->aliases, &alias->list);
+	dev->end = UINT64_C(0);
+	dev->dev = 0;
+	dev->fd = -1;
+	dev->open_count = 0;
+	dev->block_size = -1;
+	memset(dev->pvid, 0, sizeof(dev->pvid));
+	list_init(&dev->open_list);
+
+	return dev;
+}
+
+static struct device *_dev_create(dev_t d)
 {
 	struct device *dev;
 
 	if (!(dev = _alloc(sizeof(*dev)))) {
-		stack;
+		log_error("struct device allocation failed");
 		return NULL;
 	}
-
+	dev->flags = 0;
 	list_init(&dev->aliases);
 	dev->dev = d;
 	dev->fd = -1;
+	dev->open_count = 0;
+	dev->block_size = -1;
+	dev->end = UINT64_C(0);
+	memset(dev->pvid, 0, sizeof(dev->pvid));
+	list_init(&dev->open_list);
+
 	return dev;
 }
 
+void dev_set_preferred_name(struct str_list *sl, struct device *dev)
+{
+	/*
+	 * Don't interfere with ordering specified in config file.
+	 */
+	if (_cache.preferred_names_matcher)
+		return;
+
+	log_debug("%s: New preferred name", sl->str);
+	list_del(&sl->list);
+	list_add_h(&dev->aliases, &sl->list);
+}
+
+/* Return 1 if we prefer path1 else return 0 */
+static int _compare_paths(const char *path0, const char *path1)
+{
+	int slash0 = 0, slash1 = 0;
+	int m0, m1;
+	const char *p;
+	char p0[PATH_MAX], p1[PATH_MAX];
+	char *s0, *s1;
+	struct stat stat0, stat1;
+
+	/*
+	 * FIXME Better to compare patterns one-at-a-time against all names.
+	 */
+	if (_cache.preferred_names_matcher) {
+		m0 = dm_regex_match(_cache.preferred_names_matcher, path0);
+		m1 = dm_regex_match(_cache.preferred_names_matcher, path1);
+
+		if (m0 != m1) {
+			if (m0 < 0)
+				return 1;
+			if (m1 < 0)
+				return 0;
+			if (m0 < m1)
+				return 1;
+			if (m1 < m0)
+				return 0;
+		}
+	}
+
+	/*
+	 * Built-in rules.
+	 */
+
+	/* Return the path with fewer slashes */
+	for (p = path0; p++; p = (const char *) strchr(p, '/'))
+		slash0++;
+
+	for (p = path1; p++; p = (const char *) strchr(p, '/'))
+		slash1++;
+
+	if (slash0 < slash1)
+		return 0;
+	if (slash1 < slash0)
+		return 1;
+
+	strncpy(p0, path0, PATH_MAX);
+	strncpy(p1, path1, PATH_MAX);
+	s0 = &p0[0] + 1;
+	s1 = &p1[0] + 1;
+
+	/* We prefer symlinks - they exist for a reason!
+	 * So we prefer a shorter path before the first symlink in the name.
+	 * FIXME Configuration option to invert this? */
+	while (s0) {
+		s0 = strchr(s0, '/');
+		s1 = strchr(s1, '/');
+		if (s0) {
+			*s0 = '\0';
+			*s1 = '\0';
+		}
+		if (lstat(p0, &stat0)) {
+			log_sys_very_verbose("lstat", p0);
+			return 1;
+		}
+		if (lstat(p1, &stat1)) {
+			log_sys_very_verbose("lstat", p1);
+			return 0;
+		}
+		if (S_ISLNK(stat0.st_mode) && !S_ISLNK(stat1.st_mode))
+			return 0;
+		if (!S_ISLNK(stat0.st_mode) && S_ISLNK(stat1.st_mode))
+			return 1;
+		if (s0) {
+			*s0++ = '/';
+			*s1++ = '/';
+		}
+	}
+
+	/* ASCII comparison */
+	if (strcmp(path0, path1) < 0)
+		return 0;
+	else
+		return 1;
+}
+
 static int _add_alias(struct device *dev, const char *path)
 {
 	struct str_list *sl = _alloc(sizeof(*sl));
+	struct str_list *strl;
+	const char *oldpath;
+	int prefer_old = 1;
 
 	if (!sl) {
 		stack;
 		return 0;
 	}
 
-	if (!(sl->str = pool_strdup(_cache.mem, path))) {
+	/* Is name already there? */
+	list_iterate_items(strl, &dev->aliases) {
+		if (!strcmp(strl->str, path)) {
+			log_debug("%s: Already in device cache", path);
+			return 1;
+		}
+	}
+
+	if (!(sl->str = dm_pool_strdup(_cache.mem, path))) {
 		stack;
 		return 0;
 	}
 
-	list_add(&dev->aliases, &sl->list);
+	if (!list_empty(&dev->aliases)) {
+		oldpath = list_item(dev->aliases.n, struct str_list)->str;
+		prefer_old = _compare_paths(path, oldpath);
+		log_debug("%s: Aliased to %s in device cache%s",
+			  path, oldpath, prefer_old ? "" : " (preferred name)");
+
+	} else
+		log_debug("%s: Added to device cache", path);
+
+	if (prefer_old)
+		list_add(&dev->aliases, &sl->list);
+	else
+		list_add_h(&dev->aliases, &sl->list);
+
 	return 1;
 }
 
@@ -91,28 +278,44 @@ static int _add_alias(struct device *dev, const char *path)
 static int _insert_dev(const char *path, dev_t d)
 {
 	struct device *dev;
+	static dev_t loopfile_count = 0;
+	int loopfile = 0;
+
+	/* Generate pretend device numbers for loopfiles */
+	if (!d) {
+		if (dm_hash_lookup(_cache.names, path))
+			return 1;
+		d = ++loopfile_count;
+		loopfile = 1;
+	}
 
 	/* is this device already registered ? */
-	if (!(dev = (struct device *) btree_lookup(_cache.devices, d))) {
+	if (!(dev = (struct device *) btree_lookup(_cache.devices,
+						   (uint32_t) d))) {
 		/* create new device */
-		if (!(dev = _create_dev(d))) {
+		if (loopfile) {
+			if (!(dev = dev_create_file(path, NULL, NULL, 0))) {
+				stack;
+				return 0;
+			}
+		} else if (!(dev = _dev_create(d))) {
 			stack;
 			return 0;
 		}
 
-		if (!(btree_insert(_cache.devices, d, dev))) {
+		if (!(btree_insert(_cache.devices, (uint32_t) d, dev))) {
 			log_err("Couldn't insert device into binary tree.");
 			_free(dev);
 			return 0;
 		}
 	}
 
-	if (!_add_alias(dev, path)) {
+	if (!loopfile && !_add_alias(dev, path)) {
 		log_err("Couldn't add alias to dev cache.");
 		return 0;
 	}
 
-	if (!hash_insert(_cache.names, path, dev)) {
+	if (!dm_hash_insert(_cache.names, path, dev)) {
 		log_err("Couldn't add name to hash in dev cache.");
 		return 0;
 	}
@@ -122,8 +325,8 @@ static int _insert_dev(const char *path, dev_t d)
 
 static char *_join(const char *dir, const char *name)
 {
-	int len = strlen(dir) + strlen(name) + 2;
-	char *r = dbg_malloc(len);
+	size_t len = strlen(dir) + strlen(name) + 2;
+	char *r = dm_malloc(len);
 	if (r)
 		snprintf(r, len, "%s/%s", dir, name);
 
@@ -173,7 +376,7 @@ static int _insert_dir(const char *dir)
 
 			_collapse_slashes(path);
 			r &= _insert(path, 1);
-			dbg_free(path);
+			dm_free(path);
 
 			free(dirent[n]);
 		}
@@ -183,6 +386,28 @@ static int _insert_dir(const char *dir)
 	return r;
 }
 
+static int _insert_file(const char *path)
+{
+	struct stat info;
+
+	if (stat(path, &info) < 0) {
+		log_sys_very_verbose("stat", path);
+		return 0;
+	}
+
+	if (!S_ISREG(info.st_mode)) {
+		log_debug("%s: Not a regular file", path);
+		return 0;
+	}
+
+	if (!_insert_dev(path, 0)) {
+		stack;
+		return 0;
+	}
+
+	return 1;
+}
+
 static int _insert(const char *path, int rec)
 {
 	struct stat info;
@@ -194,6 +419,17 @@ static int _insert(const char *path, int rec)
 	}
 
 	if (S_ISDIR(info.st_mode)) {	/* add a directory */
+		/* check it's not a symbolic link */
+		if (lstat(path, &info) < 0) {
+			log_sys_very_verbose("lstat", path);
+			return 0;
+		}
+
+		if (S_ISLNK(info.st_mode)) {
+			log_debug("%s: Symbolic link to directory", path);
+			return 0;
+		}
+
 		if (rec)
 			r = _insert_dir(path);
 
@@ -214,33 +450,107 @@ static int _insert(const char *path, int rec)
 	return r;
 }
 
-static void _full_scan(void)
+static void _full_scan(int dev_scan)
 {
-	struct list *dh;
+	struct dir_list *dl;
 
-	if (_cache.has_scanned)
+	if (_cache.has_scanned && !dev_scan)
 		return;
 
-	list_iterate(dh, &_cache.dirs) {
-		struct dir_list *dl = list_item(dh, struct dir_list);
+	list_iterate_items(dl, &_cache.dirs)
 		_insert_dir(dl->dir);
-	};
+
+	list_iterate_items(dl, &_cache.files)
+		_insert_file(dl->dir);
 
 	_cache.has_scanned = 1;
+	init_full_scan_done(1);
 }
 
-int dev_cache_init(void)
+int dev_cache_has_scanned(void)
+{
+	return _cache.has_scanned;
+}
+
+void dev_cache_scan(int do_scan)
+{
+	if (!do_scan)
+		_cache.has_scanned = 1;
+	else
+		_full_scan(1);
+}
+
+static int _init_preferred_names(struct cmd_context *cmd)
+{
+	const struct config_node *cn;
+	struct config_value *v;
+	struct dm_pool *scratch = NULL;
+	char **regex;
+	unsigned count = 0;
+	int i, r = 0;
+
+	_cache.preferred_names_matcher = NULL;
+
+	if (!(cn = find_config_tree_node(cmd, "devices/preferred_names")) ||
+	    cn->v->type == CFG_EMPTY_ARRAY) {
+		log_very_verbose("devices/preferred_names not found in config file: "
+				 "using built-in preferences");
+		return 1;
+	}
+
+	for (v = cn->v; v; v = v->next) {
+		if (v->type != CFG_STRING) {
+			log_error("preferred_names patterns must be enclosed in quotes");
+			return 0;
+		}       
+
+		count++;
+	}
+
+	if (!(scratch = dm_pool_create("preferred device name matcher", 1024)))
+		return_0;
+
+	if (!(regex = dm_pool_alloc(scratch, sizeof(*regex) * count))) {
+		log_error("Failed to allocate preferred device name "
+			  "pattern list.");
+		goto out;
+	}
+
+	for (v = cn->v, i = count - 1; v; v = v->next, i--) {
+		if (!(regex[i] = dm_pool_strdup(scratch, v->v.str))) {
+			log_error("Failed to allocate a preferred device name "
+				  "pattern.");
+			goto out;
+		}
+	}
+
+	if (!(_cache.preferred_names_matcher =
+		dm_regex_create(_cache.mem,(const char **) regex, count))) {
+		log_error("Preferred device name pattern matcher creation failed.");
+		goto out;
+	}
+
+	r = 1;
+
+out:
+	dm_pool_destroy(scratch);
+
+	return r;
+}
+
+int dev_cache_init(struct cmd_context *cmd)
 {
 	_cache.names = NULL;
+	_cache.has_scanned = 0;
 
-	if (!(_cache.mem = pool_create(10 * 1024))) {
+	if (!(_cache.mem = dm_pool_create("dev_cache", 10 * 1024))) {
 		stack;
 		return 0;
 	}
 
-	if (!(_cache.names = hash_create(128))) {
+	if (!(_cache.names = dm_hash_create(128))) {
 		stack;
-		pool_destroy(_cache.mem);
+		dm_pool_destroy(_cache.mem);
 		_cache.mem = 0;
 		return 0;
 	}
@@ -251,6 +561,10 @@ int dev_cache_init(void)
 	}
 
 	list_init(&_cache.dirs);
+	list_init(&_cache.files);
+
+	if (!_init_preferred_names(cmd))
+		goto_bad;
 
 	return 1;
 
@@ -259,24 +573,39 @@ int dev_cache_init(void)
 	return 0;
 }
 
-void _check_closed(struct device *dev)
+static void _check_closed(struct device *dev)
 {
 	if (dev->fd >= 0)
 		log_err("Device '%s' has been left open.", dev_name(dev));
 }
 
-static inline void _check_for_open_devices(void)
+static void _check_for_open_devices(void)
 {
-	hash_iter(_cache.names, (iterate_fn) _check_closed);
+	dm_hash_iter(_cache.names, (dm_hash_iterate_fn) _check_closed);
 }
 
 void dev_cache_exit(void)
 {
-	_check_for_open_devices();
-
-	pool_destroy(_cache.mem);
 	if (_cache.names)
-		hash_destroy(_cache.names);
+		_check_for_open_devices();
+
+	if (_cache.preferred_names_matcher)
+		_cache.preferred_names_matcher = NULL;
+
+	if (_cache.mem) {
+		dm_pool_destroy(_cache.mem);
+		_cache.mem = NULL;
+	}
+
+	if (_cache.names) {
+		dm_hash_destroy(_cache.names);
+		_cache.names = NULL;
+	}
+
+	_cache.devices = NULL;
+	_cache.has_scanned = 0;
+	list_init(&_cache.dirs);
+	list_init(&_cache.files);
 }
 
 int dev_cache_add_dir(const char *path)
@@ -295,33 +624,75 @@ int dev_cache_add_dir(const char *path)
 		return 1;
 	}
 
-	if (!(dl = _alloc(sizeof(*dl) + strlen(path) + 1)))
+	if (!(dl = _alloc(sizeof(*dl) + strlen(path) + 1))) {
+		log_error("dir_list allocation failed");
 		return 0;
+	}
 
 	strcpy(dl->dir, path);
 	list_add(&_cache.dirs, &dl->list);
 	return 1;
 }
 
+int dev_cache_add_loopfile(const char *path)
+{
+	struct dir_list *dl;
+	struct stat st;
+
+	if (stat(path, &st)) {
+		log_error("Ignoring %s: %s", path, strerror(errno));
+		/* But don't fail */
+		return 1;
+	}
+
+	if (!S_ISREG(st.st_mode)) {
+		log_error("Ignoring %s: Not a regular file", path);
+		return 1;
+	}
+
+	if (!(dl = _alloc(sizeof(*dl) + strlen(path) + 1))) {
+		log_error("dir_list allocation failed for file");
+		return 0;
+	}
+
+	strcpy(dl->dir, path);
+	list_add(&_cache.files, &dl->list);
+	return 1;
+}
+
 /* Check cached device name is still valid before returning it */
 /* This should be a rare occurrence */
+/* set quiet if the cache is expected to be out-of-date */
 /* FIXME Make rest of code pass/cache struct device instead of dev_name */
-const char *dev_name_confirmed(struct device *dev)
+const char *dev_name_confirmed(struct device *dev, int quiet)
 {
 	struct stat buf;
-	char *name;
+	const char *name;
 	int r;
 
+	if ((dev->flags & DEV_REGULAR))
+		return dev_name(dev);
+
 	while ((r = stat(name = list_item(dev->aliases.n,
 					  struct str_list)->str, &buf)) ||
 	       (buf.st_rdev != dev->dev)) {
-		if (r < 0)
-			log_sys_error("stat", name);
-		log_error("Path %s no longer valid for device(%d,%d)",
-			  name, (int) MAJOR(dev->dev), (int) MINOR(dev->dev));
+		if (r < 0) {
+			if (quiet)
+				log_sys_debug("stat", name);
+			else
+				log_sys_error("stat", name);
+		}
+		if (quiet)
+			log_debug("Path %s no longer valid for device(%d,%d)",
+				  name, (int) MAJOR(dev->dev),
+				  (int) MINOR(dev->dev));
+		else
+			log_error("Path %s no longer valid for device(%d,%d)",
+				  name, (int) MAJOR(dev->dev),
+				  (int) MINOR(dev->dev));
 
 		/* Remove the incorrect hash entry */
-		hash_remove(_cache.names, name);
+		dm_hash_remove(_cache.names, name);
 
 		/* Leave list alone if there isn't an alternative name */
 		/* so dev_name will always find something to return. */
@@ -344,30 +715,46 @@ const char *dev_name_confirmed(struct device *dev)
 struct device *dev_cache_get(const char *name, struct dev_filter *f)
 {
 	struct stat buf;
-	struct device *d = (struct device *) hash_lookup(_cache.names, name);
+	struct device *d = (struct device *) dm_hash_lookup(_cache.names, name);
+
+	if (d && (d->flags & DEV_REGULAR))
+		return d;
 
 	/* If the entry's wrong, remove it */
 	if (d && (stat(name, &buf) || (buf.st_rdev != d->dev))) {
-		hash_remove(_cache.names, name);
+		dm_hash_remove(_cache.names, name);
 		d = NULL;
 	}
 
 	if (!d) {
 		_insert(name, 0);
-		d = (struct device *) hash_lookup(_cache.names, name);
+		d = (struct device *) dm_hash_lookup(_cache.names, name);
+		if (!d) {
+			_full_scan(0);
+			d = (struct device *) dm_hash_lookup(_cache.names, name);
+		}
 	}
 
-	return (d && (!f || f->passes_filter(f, d))) ? d : NULL;
+	return (d && (!f || (d->flags & DEV_REGULAR) ||
+		      f->passes_filter(f, d))) ? d : NULL;
 }
 
-struct dev_iter *dev_iter_create(struct dev_filter *f)
+struct dev_iter *dev_iter_create(struct dev_filter *f, int dev_scan)
 {
-	struct dev_iter *di = dbg_malloc(sizeof(*di));
+	struct dev_iter *di = dm_malloc(sizeof(*di));
 
-	if (!di)
+	if (!di) {
+		log_error("dev_iter allocation failed");
 		return NULL;
+	}
+
+	if (dev_scan && !trust_cache()) {
+		/* Flag gets reset between each command */
+		if (!full_scan_done())
+			persistent_filter_wipe(f); /* Calls _full_scan(1) */
+	} else
+		_full_scan(0);
 
-	_full_scan();
 	di->current = btree_first(_cache.devices);
 	di->filter = f;
 
@@ -376,10 +763,10 @@ struct dev_iter *dev_iter_create(struct dev_filter *f)
 
 void dev_iter_destroy(struct dev_iter *iter)
 {
-	dbg_free(iter);
+	dm_free(iter);
 }
 
-static inline struct device *_iter_next(struct dev_iter *iter)
+static struct device *_iter_next(struct dev_iter *iter)
 {
 	struct device *d = btree_get_data(iter->current);
 	iter->current = btree_next(iter->current);
@@ -390,10 +777,21 @@ struct device *dev_iter_get(struct dev_iter *iter)
 {
 	while (iter->current) {
 		struct device *d = _iter_next(iter);
-		if (!iter->filter ||
+		if (!iter->filter || (d->flags & DEV_REGULAR) ||
 		    iter->filter->passes_filter(iter->filter, d))
 			return d;
 	}
 
 	return NULL;
 }
+
+int dev_fd(struct device *dev)
+{
+	return dev->fd;
+}
+
+const char *dev_name(const struct device *dev)
+{
+	return (dev) ? list_item(dev->aliases.n, struct str_list)->str :
+	    "unknown device";
+}

lib/device/dev-cache.h

@@ -1,41 +1,54 @@
 /*
- * Copyright (C) 2001 Sistina Software (UK) Limited.
+ * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved.
  *
- * This file is released under the GPL.
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
 #ifndef _LVM_DEV_CACHE_H
 #define _LVM_DEV_CACHE_H
 
-#include <sys/types.h>
-#include "lvm-types.h"
 #include "device.h"
 
 /*
  * predicate for devices.
  */
 struct dev_filter {
-	int (*passes_filter)(struct dev_filter *f, struct device *dev);
-	void (*destroy)(struct dev_filter *f);
+	int (*passes_filter) (struct dev_filter * f, struct device * dev);
+	void (*destroy) (struct dev_filter * f);
 	void *private;
 };
 
-
 /*
  * The global device cache.
  */
-int dev_cache_init(void);
+struct cmd_context;
+int dev_cache_init(struct cmd_context *cmd);
 void dev_cache_exit(void);
 
+/* Trigger(1) or avoid(0) a scan */
+void dev_cache_scan(int do_scan);
+int dev_cache_has_scanned(void);
+
 int dev_cache_add_dir(const char *path);
+int dev_cache_add_loopfile(const char *path);
 struct device *dev_cache_get(const char *name, struct dev_filter *f);
 
+void dev_set_preferred_name(struct str_list *sl, struct device *dev);
 
 /*
  * Object for iterating through the cache.
  */
 struct dev_iter;
-struct dev_iter *dev_iter_create(struct dev_filter *f);
+struct dev_iter *dev_iter_create(struct dev_filter *f, int dev_scan);
 void dev_iter_destroy(struct dev_iter *iter);
 struct device *dev_iter_get(struct dev_iter *iter);
 

lib/device/dev-io.c

@@ -1,51 +1,290 @@
 /*
- * Copyright (C) 2001 Sistina Software
+ * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.
+ * Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved.
  *
- * This file is released under the LGPL.
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
-#include "device.h"
+#include "lib.h"
 #include "lvm-types.h"
-#include "log.h"
+#include "device.h"
+#include "metadata.h"
+#include "lvmcache.h"
+#include "memlock.h"
+#include "locking.h"
 
-#include <sys/types.h>
+#include <limits.h>
 #include <sys/stat.h>
 #include <fcntl.h>
 #include <unistd.h>
 #include <sys/ioctl.h>
-#include <linux/fs.h>		// UGH!!! for BLKSSZGET
 
-int dev_get_size(struct device *dev, uint64_t * size)
+#ifdef linux
+#  define u64 uint64_t		/* Missing without __KERNEL__ */
+#  undef WNOHANG		/* Avoid redefinition */
+#  undef WUNTRACED		/* Avoid redefinition */
+#  include <linux/fs.h>		/* For block ioctl definitions */
+#  define BLKSIZE_SHIFT SECTOR_SHIFT
+#  ifndef BLKGETSIZE64		/* fs.h out-of-date */
+#    define BLKGETSIZE64 _IOR(0x12, 114, size_t)
+#  endif /* BLKGETSIZE64 */
+#else
+#  include <sys/disk.h>
+#  define BLKBSZGET DKIOCGETBLOCKSIZE
+#  define BLKSSZGET DKIOCGETBLOCKSIZE
+#  define BLKGETSIZE64 DKIOCGETBLOCKCOUNT
+#  define BLKFLSBUF DKIOCSYNCHRONIZECACHE
+#  define BLKSIZE_SHIFT 0
+#endif
+
+#ifdef O_DIRECT_SUPPORT
+#  ifndef O_DIRECT
+#    error O_DIRECT support configured but O_DIRECT definition not found in headers
+#  endif
+#endif
+
+static LIST_INIT(_open_devices);
+
+/*-----------------------------------------------------------------
+ * The standard io loop that keeps submitting an io until it's
+ * all gone.
+ *---------------------------------------------------------------*/
+static int _io(struct device_area *where, void *buffer, int should_write)
+{
+	int fd = dev_fd(where->dev);
+	ssize_t n = 0;
+	size_t total = 0;
+
+	if (fd < 0) {
+		log_error("Attempt to read an unopened device (%s).",
+			  dev_name(where->dev));
+		return 0;
+	}
+
+	/*
+	 * Skip all writes in test mode.
+	 */
+	if (should_write && test_mode())
+		return 1;
+
+	if (where->size > SSIZE_MAX) {
+		log_error("Read size too large: %" PRIu64, where->size);
+		return 0;
+	}
+
+	if (lseek(fd, (off_t) where->start, SEEK_SET) < 0) {
+		log_error("%s: lseek %" PRIu64 " failed: %s",
+			  dev_name(where->dev), (uint64_t) where->start,
+			  strerror(errno));
+		return 0;
+	}
+
+	while (total < (size_t) where->size) {
+		do
+			n = should_write ?
+			    write(fd, buffer, (size_t) where->size - total) :
+			    read(fd, buffer, (size_t) where->size - total);
+		while ((n < 0) && ((errno == EINTR) || (errno == EAGAIN)));
+
+		if (n < 0)
+			log_error("%s: %s failed after %" PRIu64 " of %" PRIu64
+				  " at %" PRIu64 ": %s", dev_name(where->dev),
+				  should_write ? "write" : "read",
+				  (uint64_t) total,
+				  (uint64_t) where->size,
+				  (uint64_t) where->start, strerror(errno));
+
+		if (n <= 0)
+			break;
+
+		total += n;
+		buffer += n;
+	}
+
+	return (total == (size_t) where->size);
+}
+
+/*-----------------------------------------------------------------
+ * LVM2 uses O_DIRECT when performing metadata io, which requires
+ * block size aligned accesses.  If any io is not aligned we have
+ * to perform the io via a bounce buffer, obviously this is quite
+ * inefficient.
+ *---------------------------------------------------------------*/
+
+/*
+ * Get the sector size from an _open_ device.
+ */
+static int _get_block_size(struct device *dev, unsigned int *size)
 {
-	int fd;
-	long s;
 	const char *name = dev_name(dev);
 
-	log_very_verbose("Getting size of %s", name);
+	if ((dev->block_size == -1)) {
+		if (ioctl(dev_fd(dev), BLKBSZGET, &dev->block_size) < 0) {
+			log_sys_error("ioctl BLKBSZGET", name);
+			return 0;
+		}
+		log_debug("%s: block size is %u bytes", name, dev->block_size);
+	}
+
+	*size = (unsigned int) dev->block_size;
+
+	return 1;
+}
+
+/*
+ * Widens a region to be an aligned region.
+ */
+static void _widen_region(unsigned int block_size, struct device_area *region,
+			  struct device_area *result)
+{
+	uint64_t mask = block_size - 1, delta;
+	memcpy(result, region, sizeof(*result));
+
+	/* adjust the start */
+	delta = result->start & mask;
+	if (delta) {
+		result->start -= delta;
+		result->size += delta;
+	}
+
+	/* adjust the end */
+	delta = (result->start + result->size) & mask;
+	if (delta)
+		result->size += block_size - delta;
+}
+
+static int _aligned_io(struct device_area *where, void *buffer,
+		       int should_write)
+{
+	void *bounce;
+	unsigned int block_size = 0;
+	uintptr_t mask;
+	struct device_area widened;
+
+	if (!(where->dev->flags & DEV_REGULAR) &&
+	    !_get_block_size(where->dev, &block_size)) {
+		stack;
+		return 0;
+	}
+
+	if (!block_size)
+		block_size = lvm_getpagesize();
+
+	_widen_region(block_size, where, &widened);
+
+	/* Do we need to use a bounce buffer? */
+	mask = block_size - 1;
+	if (!memcmp(where, &widened, sizeof(widened)) &&
+	    !((uintptr_t) buffer & mask))
+		return _io(where, buffer, should_write);
+
+	/* Allocate a bounce buffer with an extra block */
+	if (!(bounce = alloca((size_t) widened.size + block_size))) {
+		log_error("Bounce buffer alloca failed");
+		return 0;
+	}
+
+	/*
+	 * Realign start of bounce buffer (using the extra sector)
+	 */
+	if (((uintptr_t) bounce) & mask)
+		bounce = (void *) ((((uintptr_t) bounce) + mask) & ~mask);
+
+	/* channel the io through the bounce buffer */
+	if (!_io(&widened, bounce, 0)) {
+		if (!should_write) {
+			stack;
+			return 0;
+		}
+		/* FIXME pre-extend the file */
+		memset(bounce, '\n', widened.size);
+	}
+
+	if (should_write) {
+		memcpy(bounce + (where->start - widened.start), buffer,
+		       (size_t) where->size);
+
+		/* ... then we write */
+		return _io(&widened, bounce, 1);
+	}
+
+	memcpy(buffer, bounce + (where->start - widened.start),
+	       (size_t) where->size);
+
+	return 1;
+}
+
+static int _dev_get_size_file(const struct device *dev, uint64_t *size)
+{
+	const char *name = dev_name(dev);
+	struct stat info;
+
+	if (stat(name, &info)) {
+		log_sys_error("stat", name);
+		return 0;
+	}
+
+	*size = info.st_size;
+	*size >>= SECTOR_SHIFT;	/* Convert to sectors */
+
+	log_very_verbose("%s: size is %" PRIu64 " sectors", name, *size);
+
+	return 1;
+}
+
+static int _dev_get_size_dev(const struct device *dev, uint64_t *size)
+{
+	int fd;
+	const char *name = dev_name(dev);
+
 	if ((fd = open(name, O_RDONLY)) < 0) {
 		log_sys_error("open", name);
 		return 0;
 	}
 
-	/* FIXME: add 64 bit ioctl */
-	if (ioctl(fd, BLKGETSIZE, &s) < 0) {
-		log_sys_error("ioctl BLKGETSIZE", name);
-		close(fd);
+	if (ioctl(fd, BLKGETSIZE64, size) < 0) {
+		log_sys_error("ioctl BLKGETSIZE64", name);
+		if (close(fd))
+			log_sys_error("close", name);
 		return 0;
 	}
 
-	close(fd);
-	*size = (uint64_t) s;
+	*size >>= BLKSIZE_SHIFT;	/* Convert to sectors */
+	if (close(fd))
+		log_sys_error("close", name);
+
+	log_very_verbose("%s: size is %" PRIu64 " sectors", name, *size);
+
 	return 1;
 }
 
-int dev_get_sectsize(struct device *dev, uint32_t * size)
+/*-----------------------------------------------------------------
+ * Public functions
+ *---------------------------------------------------------------*/
+
+int dev_get_size(const struct device *dev, uint64_t *size)
+{
+	if ((dev->flags & DEV_REGULAR))
+		return _dev_get_size_file(dev, size);
+	else
+		return _dev_get_size_dev(dev, size);
+}
+
+/* FIXME Unused
+int dev_get_sectsize(struct device *dev, uint32_t *size)
 {
 	int fd;
 	int s;
 	const char *name = dev_name(dev);
 
-	log_very_verbose("Getting size of %s", name);
 	if ((fd = open(name, O_RDONLY)) < 0) {
 		log_sys_error("open", name);
 		return 0;
@@ -53,197 +292,388 @@ int dev_get_sectsize(struct device *dev, uint32_t * size)
 
 	if (ioctl(fd, BLKSSZGET, &s) < 0) {
 		log_sys_error("ioctl BLKSSZGET", name);
-		close(fd);
+		if (close(fd))
+			log_sys_error("close", name);
 		return 0;
 	}
 
-	close(fd);
+	if (close(fd))
+		log_sys_error("close", name);
+
 	*size = (uint32_t) s;
+
+	log_very_verbose("%s: sector size is %" PRIu32 " bytes", name, *size);
+
 	return 1;
 }
+*/
 
-int dev_open(struct device *dev, int flags)
+void dev_flush(struct device *dev)
+{
+	if (!(dev->flags & DEV_REGULAR) && ioctl(dev->fd, BLKFLSBUF, 0) >= 0)
+		return;
+
+	if (fsync(dev->fd) >= 0)
+		return;
+
+	sync();
+}
+
+int dev_open_flags(struct device *dev, int flags, int direct, int quiet)
 {
 	struct stat buf;
-	const char *name = dev_name_confirmed(dev);
+	const char *name;
+	int need_excl = 0, need_rw = 0;
 
-	if (!name) {
+	if ((flags & O_ACCMODE) == O_RDWR)
+		need_rw = 1;
+
+	if ((flags & O_EXCL))
+		need_excl = 1;
+
+	if (dev->fd >= 0) {
+		if (((dev->flags & DEV_OPENED_RW) || !need_rw) &&
+		    ((dev->flags & DEV_OPENED_EXCL) || !need_excl)) {
+			dev->open_count++;
+			return 1;
+		}
+
+		if (dev->open_count && !need_excl) {
+			/* FIXME Ensure we never get here */
+			log_debug("WARNING: %s already opened read-only", 
+				  dev_name(dev));
+			dev->open_count++;
+		}
+
+		dev_close_immediate(dev);
+	}
+
+	if (memlock())
+		log_error("WARNING: dev_open(%s) called while suspended",
+			  dev_name(dev));
+
+	if (dev->flags & DEV_REGULAR)
+		name = dev_name(dev);
+	else if (!(name = dev_name_confirmed(dev, quiet))) {
 		stack;
 		return 0;
 	}
 
-	if (dev->fd >= 0) {
-		log_error("Device '%s' has already been opened", name);
+	if (!(dev->flags & DEV_REGULAR)) {
+		if (stat(name, &buf) < 0) {
+			log_sys_error("%s: stat failed", name);
+			return 0;
+		}
+		if (buf.st_rdev != dev->dev) {
+			log_error("%s: device changed", name);
+			return 0;
+		}
+	}
+
+#ifdef O_DIRECT_SUPPORT
+	if (direct) {
+		if (!(dev->flags & DEV_O_DIRECT_TESTED))
+			dev->flags |= DEV_O_DIRECT;
+
+		if ((dev->flags & DEV_O_DIRECT))
+			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) {
+#ifdef O_DIRECT_SUPPORT
+		if (direct && !(dev->flags & DEV_O_DIRECT_TESTED)) {
+			flags &= ~O_DIRECT;
+			if ((dev->fd = open(name, flags, 0777)) >= 0) {
+				dev->flags &= ~DEV_O_DIRECT;
+				log_debug("%s: Not using O_DIRECT", name);
+				goto opened;
+			}
+		}
+#endif
+		if (quiet)
+			log_sys_debug("open", name);
+		else
+			log_sys_error("open", name);
 		return 0;
 	}
 
-	if ((stat(name, &buf) < 0) || (buf.st_rdev != dev->dev)) {
-		log_error("%s: stat failed: Has device name changed?", name);
-		return 0;
-	}
+#ifdef O_DIRECT_SUPPORT
+      opened:
+	if (direct)
+		dev->flags |= DEV_O_DIRECT_TESTED;
+#endif
+	dev->open_count++;
+	dev->flags &= ~DEV_ACCESSED_W;
 
-	if ((dev->fd = open(name, flags)) < 0) {
-		log_sys_error("open", name);
-		return 0;
-	}
+	if (need_rw)
+		dev->flags |= DEV_OPENED_RW;
+	else
+		dev->flags &= ~DEV_OPENED_RW;
 
-	if ((fstat(dev->fd, &buf) < 0) || (buf.st_rdev != dev->dev)) {
+	if (need_excl)
+		dev->flags |= DEV_OPENED_EXCL;
+	else
+		dev->flags &= ~DEV_OPENED_EXCL;
+
+	if (!(dev->flags & DEV_REGULAR) &&
+	    ((fstat(dev->fd, &buf) < 0) || (buf.st_rdev != dev->dev))) {
 		log_error("%s: fstat failed: Has device name changed?", name);
-		dev_close(dev);
+		dev_close_immediate(dev);
 		return 0;
 	}
 
-	dev->flags = 0;
+#ifndef O_DIRECT_SUPPORT
+	if (!(dev->flags & DEV_REGULAR))
+		dev_flush(dev);
+#endif
+
+	if ((flags & O_CREAT) && !(flags & O_TRUNC))
+		dev->end = lseek(dev->fd, (off_t) 0, SEEK_END);
+
+	list_add(&_open_devices, &dev->open_list);
+
+	log_debug("Opened %s %s%s%s", dev_name(dev),
+		  dev->flags & DEV_OPENED_RW ? "RW" : "RO",
+		  dev->flags & DEV_OPENED_EXCL ? " O_EXCL" : "",
+		  dev->flags & DEV_O_DIRECT ? " O_DIRECT" : "");
 
 	return 1;
 }
 
-static void _flush(int fd)
+int dev_open_quiet(struct device *dev)
 {
-	ioctl(fd, BLKFLSBUF, 0);
+	int flags;
+
+	flags = vg_write_lock_held() ? O_RDWR : O_RDONLY;
+
+	return dev_open_flags(dev, flags, 1, 1);
 }
 
-int dev_close(struct device *dev)
+int dev_open(struct device *dev)
 {
+	int flags;
+
+	flags = vg_write_lock_held() ? O_RDWR : O_RDONLY;
+
+	return dev_open_flags(dev, flags, 1, 0);
+}
+
+int dev_test_excl(struct device *dev)
+{
+	int flags;
+	int r;
+
+	flags = vg_write_lock_held() ? O_RDWR : O_RDONLY;
+	flags |= O_EXCL;
+
+	r = dev_open_flags(dev, flags, 1, 1);
+	if (r)
+		dev_close_immediate(dev);
+
+	return r;
+}
+
+static void _close(struct device *dev)
+{
+	if (close(dev->fd))
+		log_sys_error("close", dev_name(dev));
+	dev->fd = -1;
+	dev->block_size = -1;
+	list_del(&dev->open_list);
+
+	log_debug("Closed %s", dev_name(dev));
+
+	if (dev->flags & DEV_ALLOCED) {
+		dm_free((void *) list_item(dev->aliases.n, struct str_list)->
+			 str);
+		dm_free(dev->aliases.n);
+		dm_free(dev);
+	}
+}
+
+static int _dev_close(struct device *dev, int immediate)
+{
+	struct lvmcache_info *info;
+
 	if (dev->fd < 0) {
 		log_error("Attempt to close device '%s' "
 			  "which is not open.", dev_name(dev));
 		return 0;
 	}
 
+#ifndef O_DIRECT_SUPPORT
 	if (dev->flags & DEV_ACCESSED_W)
-		_flush(dev->fd);
+		dev_flush(dev);
+#endif
 
-	if (close(dev->fd))
-		log_sys_error("close", dev_name(dev));
+	if (dev->open_count > 0)
+		dev->open_count--;
 
-	dev->fd = -1;
+	if (immediate && dev->open_count)
+		log_debug("%s: Immediate close attempt while still referenced",
+			  dev_name(dev));
+
+	/* Close unless device is known to belong to a locked VG */
+	if (immediate ||
+	    (dev->open_count < 1 && 
+	     (!(info = info_from_pvid(dev->pvid)) ||
+	      !info->vginfo ||
+	      !vgname_is_locked(info->vginfo->vgname))))
+		_close(dev);
 
 	return 1;
 }
 
+int dev_close(struct device *dev)
+{
+	return _dev_close(dev, 0);
+}
+
+int dev_close_immediate(struct device *dev)
+{
+	return _dev_close(dev, 1);
+}
+
+void dev_close_all(void)
+{
+	struct list *doh, *doht;
+	struct device *dev;
+
+	list_iterate_safe(doh, doht, &_open_devices) {
+		dev = list_struct_base(doh, struct device, open_list);
+		if (dev->open_count < 1)
+			_close(dev);
+	}
+}
+
+int dev_read(struct device *dev, uint64_t offset, size_t len, void *buffer)
+{
+	struct device_area where;
+
+	if (!dev->open_count) {
+		stack;
+		return 0;
+	}
+
+	where.dev = dev;
+	where.start = offset;
+	where.size = len;
+
+	return _aligned_io(&where, buffer, 0);
+}
+
 /*
- *  FIXME: factor common code out.
+ * Read from 'dev' into 'buf', possibly in 2 distinct regions, denoted
+ * by (offset,len) and (offset2,len2).  Thus, the total size of
+ * 'buf' should be len+len2.
  */
-int _read(int fd, void *buf, size_t count)
+int dev_read_circular(struct device *dev, uint64_t offset, size_t len,
+		      uint64_t offset2, size_t len2, void *buf)
 {
-	size_t n = 0;
-	int tot = 0;
-
-	while (tot < count) {
-		do
-			n = read(fd, buf, count - tot);
-		while ((n < 0) && ((errno == EINTR) || (errno == EAGAIN)));
-
-		if (n <= 0)
-			return tot ? tot : n;
-
-		tot += n;
-		buf += n;
+	if (!dev_read(dev, offset, len, buf)) {
+		log_error("Read from %s failed", dev_name(dev));
+		return 0;
 	}
 
-	return tot;
+	/*
+	 * The second region is optional, and allows for
+	 * a circular buffer on the device.
+	 */
+	if (!len2)
+		return 1;
+
+	if (!dev_read(dev, offset2, len2, buf + len)) {
+		log_error("Circular read from %s failed",
+			  dev_name(dev));
+		return 0;
+	}
+
+	return 1;
 }
 
-int64_t dev_read(struct device * dev, uint64_t offset,
-		 int64_t len, void *buffer)
+/* FIXME If O_DIRECT can't extend file, dev_extend first; dev_truncate after.
+ *       But fails if concurrent processes writing
+ */
+
+/* FIXME pre-extend the file */
+int dev_append(struct device *dev, size_t len, void *buffer)
 {
-	const char *name = dev_name(dev);
-	int fd = dev->fd;
+	int r;
 
-	if (fd < 0) {
-		log_err("Attempt to read an unopened device (%s).", name);
+	if (!dev->open_count) {
+		stack;
 		return 0;
 	}
 
-	if (lseek(fd, offset, SEEK_SET) < 0) {
-		log_sys_error("lseek", name);
-		return 0;
-	}
+	r = dev_write(dev, dev->end, len, buffer);
+	dev->end += (uint64_t) len;
 
-	return _read(fd, buffer, len);
+#ifndef O_DIRECT_SUPPORT
+	dev_flush(dev);
+#endif
+	return r;
 }
 
-int _write(int fd, const void *buf, size_t count)
+int dev_write(struct device *dev, uint64_t offset, size_t len, void *buffer)
 {
-	size_t n = 0;
-	int tot = 0;
+	struct device_area where;
 
-	/* Skip all writes */
-	if (test_mode())
-		return count;
-
-	while (tot < count) {
-		do
-			n = write(fd, buf, count - tot);
-		while ((n < 0) && ((errno == EINTR) || (errno == EAGAIN)));
-
-		if (n <= 0)
-			return tot ? tot : n;
-
-		tot += n;
-		buf += n;
-	}
-
-	return tot;
-}
-
-int64_t dev_write(struct device * dev, uint64_t offset,
-		  int64_t len, void *buffer)
-{
-	const char *name = dev_name(dev);
-	int fd = dev->fd;
-
-	if (fd < 0) {
-		log_error("Attempt to write to unopened device %s", name);
+	if (!dev->open_count) {
+		stack;
 		return 0;
 	}
 
-	if (lseek(fd, offset, SEEK_SET) < 0) {
-		log_sys_error("lseek", name);
-		return 0;
-	}
+	where.dev = dev;
+	where.start = offset;
+	where.size = len;
 
 	dev->flags |= DEV_ACCESSED_W;
 
-	return _write(fd, buffer, len);
+	return _aligned_io(&where, buffer, 1);
 }
 
-int dev_zero(struct device *dev, uint64_t offset, int64_t len)
+int dev_set(struct device *dev, uint64_t offset, size_t len, int value)
 {
-	int64_t r, s;
-	char buffer[4096];
-	const char *name = dev_name(dev);
-	int fd = dev->fd;
+	size_t s;
+	char buffer[4096] __attribute((aligned(8)));
 
-	if (fd < 0) {
-		log_error("Attempt to zero part of an unopened device %s",
-			  name);
+	if (!dev_open(dev)) {
+		stack;
 		return 0;
 	}
 
-	if (lseek(fd, offset, SEEK_SET) < 0) {
-		log_sys_error("lseek", name);
-		return 0;
-	}
+	if ((offset % SECTOR_SIZE) || (len % SECTOR_SIZE))
+		log_debug("Wiping %s at %" PRIu64 " length %" PRIsize_t,
+			  dev_name(dev), offset, len);
+	else
+		log_debug("Wiping %s at sector %" PRIu64 " length %" PRIsize_t
+			  " sectors", dev_name(dev), offset >> SECTOR_SHIFT,
+			  len >> SECTOR_SHIFT);
 
-	memset(buffer, 0, sizeof(buffer));
+	memset(buffer, value, sizeof(buffer));
 	while (1) {
 		s = len > sizeof(buffer) ? sizeof(buffer) : len;
-		r = _write(fd, buffer, s);
-
-		if (r <= 0)
+		if (!dev_write(dev, offset, s, buffer))
 			break;
 
-		len -= r;
-		if (!len) {
-			r = 1;
+		len -= s;
+		if (!len)
 			break;
-		}
+
+		offset += s;
 	}
 
 	dev->flags |= DEV_ACCESSED_W;
 
-	/* FIXME: Always display error */
+	if (!dev_close(dev))
+		stack;
+
 	return (len == 0);
 }

lib/device/dev-md.c

@@ -0,0 +1,126 @@
+/*
+ * Copyright (C) 2004 Luca Berra
+ * Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include "lib.h"
+#include "metadata.h"
+#include "xlate.h"
+
+#ifdef linux
+
+/* Lifted from <linux/raid/md_p.h> because of difficulty including it */
+
+#define MD_SB_MAGIC 0xa92b4efc
+#define MD_RESERVED_BYTES (64 * 1024ULL)
+#define MD_RESERVED_SECTORS (MD_RESERVED_BYTES / 512)
+#define MD_NEW_SIZE_SECTORS(x) ((x & ~(MD_RESERVED_SECTORS - 1)) \
+				- MD_RESERVED_SECTORS)
+
+static int _dev_has_md_magic(struct device *dev, uint64_t sb_offset)
+{
+	uint32_t md_magic;
+
+	/* Version 1 is little endian; version 0.90.0 is machine endian */
+	if (dev_read(dev, sb_offset, sizeof(uint32_t), &md_magic) &&
+	    ((md_magic == xlate32(MD_SB_MAGIC)) ||
+	     (md_magic == MD_SB_MAGIC)))
+		return 1;
+
+	return 0;
+}
+
+/*
+ * Calculate the position of the superblock.
+ * It is always aligned to a 4K boundary and
+ * depending on minor_version, it can be:
+ * 0: At least 8K, but less than 12K, from end of device
+ * 1: At start of device
+ * 2: 4K from start of device.
+ */
+static uint64_t _v1_sb_offset(uint64_t size, unsigned minor_version)
+{
+	uint64_t sb_offset;
+
+	switch(minor_version) {
+	case 0:
+		sb_offset = (size - 8 * 2) & ~(4 * 2 - 1ULL);
+		break;
+	case 1:
+		sb_offset = 0;
+		break;
+	case 2:
+		sb_offset = 4 * 2;
+		break;
+	}
+	sb_offset <<= SECTOR_SHIFT;
+
+	return sb_offset;
+}
+
+/*
+ * Returns -1 on error
+ */
+int dev_is_md(struct device *dev, uint64_t *sb)
+{
+	int ret = 1;
+	unsigned minor = 0;
+	uint64_t size, sb_offset;
+
+	if (!dev_get_size(dev, &size)) {
+		stack;
+		return -1;
+	}
+
+	if (size < MD_RESERVED_SECTORS * 2)
+		return 0;
+
+	if (!dev_open(dev)) {
+		stack;
+		return -1;
+	}
+
+	/* Check if it is an md component device. */
+	/* Version 0.90.0 */
+	sb_offset = MD_NEW_SIZE_SECTORS(size) << SECTOR_SHIFT;
+	if (_dev_has_md_magic(dev, sb_offset))
+		goto out;
+
+	/* Version 1, try v1.0 -> v1.2 */
+	do {
+		sb_offset = _v1_sb_offset(size, minor);
+		if (_dev_has_md_magic(dev, sb_offset))
+			goto out;
+	} while (++minor <= 2);
+
+	ret = 0;
+
+out:
+	if (!dev_close(dev))
+		stack;
+
+	if (ret && sb)
+		*sb = sb_offset;
+
+	return ret;
+}
+
+#else
+
+int dev_is_md(struct device *dev __attribute((unused)),
+	      uint64_t *sb __attribute((unused)))
+{
+	return 0;
+}
+
+#endif

lib/device/device.c

@@ -1,56 +1,126 @@
 /*
- * Copyright (C) 2001 Sistina Software
+ * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.
+ * Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved.
  *
- * This LVM library is free software; you can redistribute it and/or
- * modify it under the terms of the GNU Library General Public
- * License as published by the Free Software Foundation; either
- * version 2 of the License, or (at your option) any later version.
+ * This file is part of LVM2.
  *
- * This LVM library is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * Library General Public License for more details.
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
  *
- * You should have received a copy of the GNU Library General Public
- * License along with this LVM library; if not, write to the Free
- * Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
- * MA 02111-1307, USA
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
-#include <sys/types.h>
+#include "lib.h"
+#include "lvm-types.h"
+#include "device.h"
+#include "metadata.h"
+#include "filter.h"
+#include "xlate.h"
+
+/* See linux/genhd.h and fs/partitions/msdos */
+
+#define PART_MAGIC 0xAA55
+#define PART_MAGIC_OFFSET UINT64_C(0x1FE)
+#define PART_OFFSET UINT64_C(0x1BE)
+
+struct partition {
+        uint8_t boot_ind;
+        uint8_t head;
+        uint8_t sector;
+        uint8_t cyl;
+        uint8_t sys_ind;	/* partition type */
+        uint8_t end_head;
+        uint8_t end_sector;
+        uint8_t end_cyl;
+        uint32_t start_sect;
+        uint32_t nr_sects;
+} __attribute__((packed));
+
+static int _is_partitionable(struct device *dev)
+{
+	int parts = max_partitions(MAJOR(dev->dev));
+
+	if ((parts <= 1) || (MINOR(dev->dev) % parts))
+		return 0;
+
+	return 1;
+}
+
+static int _has_partition_table(struct device *dev)
+{
+	int ret = 0;
+	unsigned p;
+	uint16_t buf[SECTOR_SIZE/sizeof(uint16_t)];
+	uint16_t *part_magic;
+	struct partition *part;
+
+	if (!dev_open(dev)) {
+		stack;
+		return -1;
+	}
+
+	if (!dev_read(dev, UINT64_C(0), sizeof(buf), &buf)) {
+		stack;
+		goto out;
+	}
+
+	/* FIXME Check for other types of partition table too */
+
+	/* Check for msdos partition table */
+	part_magic = buf + PART_MAGIC_OFFSET/sizeof(buf[0]);
+	if ((*part_magic == xlate16(PART_MAGIC))) {
+		part = (struct partition *) (buf + PART_OFFSET/sizeof(buf[0]));
+		for (p = 0; p < 4; p++, part++) {
+			/* Table is invalid if boot indicator not 0 or 0x80 */
+			if ((part->boot_ind & 0x7f)) {
+				ret = 0;
+				break;
+			}
+			/* Must have at least one non-empty partition */
+			if (part->nr_sects)
+				ret = 1;
+		}
+	}
+
+      out:
+	if (!dev_close(dev))
+		stack;
+
+	return ret;
+}
+
+int is_partitioned_dev(struct device *dev)
+{
+	if (!_is_partitionable(dev))
+		return 0;
+
+	return _has_partition_table(dev);
+}
+
+#if 0
 #include <sys/stat.h>
 #include <sys/mman.h>
 #include <stdio.h>
 #include <unistd.h>
 #include <fcntl.h>
 #include <ctype.h>
-#include <string.h>
+
 #include <errno.h>
 #include <sys/ioctl.h>
-
-#include "dbg_malloc.h"
-#include "log.h"
-#include "dev-cache.h"
-#include "metadata.h"
-#include "device.h"
-
 #include <linux/fs.h>
 #include <linux/major.h>
 #include <linux/genhd.h>
 
-#if 0
 int _get_partition_type(struct dev_filter *filter, struct device *d);
 
-#define MINOR_PART(dm, d) (MINOR((d)->dev) % dev_max_partitions(dm, (d)->dev))
+#define MINOR_PART(dev) (MINOR((dev)->dev) % max_partitions(MINOR((dev)->dev)))
 
-int is_whole_disk(struct dev_filter *filter, struct device *d)
+int is_extended_partition(struct device *d)
 {
-	return (MINOR_PART(dm, d)) ? 0 : 1;
-}
-
-int is_extended_partition(struct dev_mgr *dm, struct device *d)
-{
-	return (MINOR_PART(dm, d) > 4) ? 1 : 0;
+	return (MINOR_PART(d) > 4) ? 1 : 0;
 }
 
 struct device *dev_primary(struct dev_mgr *dm, struct device *d)
@@ -132,7 +202,7 @@ int _get_partition_type(struct dev_mgr *dm, struct device *d)
 		return 0;
 	}
 
-	if (!(buffer = dbg_malloc(SECTOR_SIZE))) {
+	if (!(buffer = dm_malloc(SECTOR_SIZE))) {
 		log_error("Failed to allocate partition table buffer");
 		return 0;
 	}

lib/device/device.h

@@ -1,28 +1,51 @@
 /*
- * Copyright (C) 2001 Sistina Software (UK) Limited.
+ * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.  
+ * Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved.
  *
- * This file is released under the GPL.
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
 #ifndef _LVM_DEVICE_H
 #define _LVM_DEVICE_H
 
-#include "lvm-types.h"
-#include "list.h"
+#include "uuid.h"
+
+#include <fcntl.h>
 
 #define DEV_ACCESSED_W		0x00000001	/* Device written to? */
+#define DEV_REGULAR		0x00000002	/* Regular file? */
+#define DEV_ALLOCED		0x00000004	/* dm_malloc used */
+#define DEV_OPENED_RW		0x00000008	/* Opened RW */
+#define DEV_OPENED_EXCL		0x00000010	/* Opened EXCL */
+#define DEV_O_DIRECT		0x00000020	/* Use O_DIRECT */
+#define DEV_O_DIRECT_TESTED	0x00000040	/* DEV_O_DIRECT is reliable */
 
 /*
  * All devices in LVM will be represented by one of these.
  * pointer comparisons are valid.
  */
 struct device {
-	struct list aliases; /* struct str_list from lvm-types.h */
+	struct list aliases;	/* struct str_list from lvm-types.h */
 	dev_t dev;
 
 	/* private */
 	int fd;
+	int open_count;
+	int block_size;
 	uint32_t flags;
+	uint64_t end;
+	struct list open_list;
+
+	char pvid[ID_LEN + 1];
+	char _padding[7];
 };
 
 struct device_list {
@@ -30,33 +53,52 @@ struct device_list {
 	struct device *dev;
 };
 
+struct device_area {
+	struct device *dev;
+	uint64_t start;		/* Bytes */
+	uint64_t size;		/* Bytes */
+};
+
 /*
  * 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);
 
-int dev_open(struct device *dev, int flags);
+/* Use quiet version if device number could change e.g. when opening LV */
+int dev_open(struct device *dev);
+int dev_open_quiet(struct device *dev);
+int dev_open_flags(struct device *dev, int flags, int direct, int quiet);
 int dev_close(struct device *dev);
+int dev_close_immediate(struct device *dev);
+void dev_close_all(void);
+int dev_test_excl(struct device *dev);
 
-int64_t dev_read(struct device *dev,
-		 uint64_t offset, int64_t len, void *buffer);
-int64_t dev_write(struct device *dev,
-		  uint64_t offset, int64_t len, void *buffer);
-int dev_zero(struct device *dev, uint64_t offset, int64_t len);
+int dev_fd(struct device *dev);
+const char *dev_name(const struct device *dev);
 
+int dev_read(struct device *dev, uint64_t offset, size_t len, void *buffer);
+int dev_read_circular(struct device *dev, uint64_t offset, size_t len,
+		      uint64_t offset2, size_t len2, void *buf);
+int dev_write(struct device *dev, uint64_t offset, size_t len, void *buffer);
+int dev_append(struct device *dev, size_t len, void *buffer);
+int dev_set(struct device *dev, uint64_t offset, size_t len, int value);
+void dev_flush(struct device *dev);
 
-static inline const char *dev_name(struct device *dev) {
-	return (dev) ? list_item(dev->aliases.n, struct str_list)->str :
-		       "unknown device";
-}
+struct device *dev_create_file(const char *filename, struct device *dev,
+			       struct str_list *alias, int use_malloc);
 
 /* Return a valid device name from the alias list; NULL otherwise */
-const char *dev_name_confirmed(struct device *dev);
+const char *dev_name_confirmed(struct device *dev, int quiet);
 
-static inline int is_lvm_partition(const char *name) {
-	return 1;
-}
+/* Does device contain md superblock?  If so, where? */
+int dev_is_md(struct device *dev, uint64_t *sb);
+
+/* FIXME Check partition type if appropriate */
+
+#define is_lvm_partition(a) 1
+/* int is_lvm_partition(const char *name); */
+
+int is_partitioned_dev(struct device *dev);
 
 #endif
-