-This is a guide on how to build [[OpenAFS]] from source code. Note that [[OpenAFS]] pre-built binaries are available on the [[OpenAFS]] site and are available as prebuilt packages for many platforms. These instructions may be useful for you if you need to build [[OpenAFS]] from source.
+These are notes on how to build [[OpenAFS]] from source code. Note that [[OpenAFS]] pre-built binaries are available on the [[OpenAFS]] site and are available as prebuilt packages for many platforms. These instructions may be useful for you if you need to build [[OpenAFS]] from source.
### <a name="Requirements"></a> Requirements
-Tools
+#### <a name="Tools"></a> Tools
- cvs client
- autoconf
- automake
- perl 5.6
-- gcc (versions?)
+- gcc
- GNU make
- lex/yacc (flex/bison)
-Libraries
+#### <a name="Libraries"></a> Libraries
- libc
- kerberos, optional, but recommended
- ncurses
- kernel headers
+The Kerberos development libraries are required if you are going to build with Kerberos 5 support. The [[OpenAFS]] legacy kaserver is deprecated.
+
If you are building on Debian, you can get the required software with the following apt-get commands.
apt-get install cvs autoconf automake make gcc flex bison
### <a name="Getting the Source Code from CVS"></a> Getting the Source Code from CVS
-You can get development snapshots from the [[OpenAFS]] CVS repository. The CVS tree may not always have code which can currently be built. While every effort is made to keep the head of the tree buildable, you may at any time find yourself between commits and hence have a tree which does not build, or worse, causes more serious problems.
+You can get development snapshots from the [[OpenAFS]] CVS repository . The CVS tree may not always have code which can currently be built. While every effort is made to keep the head of the tree buildable, you may at any time find yourself between commits and hence have a tree which does not build, or worse, causes more serious problems.
First you need to run cvs login. This step is normally only done once. A ~/.cvspass file will be created for additional checkouts.
cvs -d :pserver:anonymous@cvs.openafs.org:/cvs login
password is anonymous
-To check out the 1.4 branch:
+Before doing a CVS check out, you'll need to decide which branch you want to check out. The trunk is for bleeding edge development and may not even build. The current stable series is openafs-stable-1\_4\_x and the current development work is going on the openafs-devel-1\_5\_x branch.
- cvs -d :pserver:anonymous@cvs.openafs.org:/cvs co -r openafs-stable-1_4_x openafs
+See the CVSWeb interface at <http://www.openafs.org/frameset/cgi-bin/cvsweb.cgi/openafs/> for the full list of available branches and tags.
-To check out the 1.5 development branch:
+To check out the stable branch:
- cvs -d :pserver:anonymous@cvs.openafs.org:/cvs co -r openafs-devel-1_5_x openafs
+ cvs -d :pserver:anonymous@cvs.openafs.org:/cvs checkout -r openafs-stable-1_4_x -d openafs-stable-1_4_x openafs
-See the CVSWeb interface at <http://www.openafs.org/frameset/cgi-bin/cvsweb.cgi/openafs/> to see a full list of available branches and tags
+This will create a source tree in a directory named openafs-stable-1\_4\_x.
-CVS snapshots do not include files generated by autoconf; You can run regen.sh at the top level to create these files. You will need to have autoconf and automake installed on your system.
+To check out the development branch:
-### <a name="Building the Binaries"></a> Building the Binaries
+ cvs -d :pserver:anonymous@cvs.openafs.org:/cvs checkout -r openafs-devel-1_5_x -d openafs-devel-1_5_x openafs
-#### <a name="Quick Answer"></a> Quick Answer
+This will create a source tree in a directory named openafs-devel-1\_5\_x.
-To build [[OpenAFS]] with Kerberos 5 support and the traditional paths:
+Next, step into your sandbox and run the regen.sh script to create the configure script.
- cd openafs
+ cd openafs-stable-1_4_x.
./regen.sh
- ./configure --enable-debug --enable-transarc-paths --with-krb5-conf=(full path to krb5-config script)
+
+### <a name="Building the Binaries"></a> Building the Binaries
+
+#### <a name="Modern Paths"></a> Modern Paths
+
+To build [[OpenAFS]] with Kerberos 5 support, and with a custom install path,
+
+ ./configure --prefix=/usr/local/openafs --with-krb5-conf=(full path to krb5-config script)
make
- make dest
+ sudo make install
-You will need the Kerberos development libraries to build Kerbose support.
+#### <a name="Transarc Paths"></a> Transarc Paths
-### <a name="Installing the Binaries"></a> Installing the Binaries
+By convention, [[OpenAFS]] server binaries and related files are located in /usr/afs, and client binaries and related files are located in /usr/vice. These are known as Transarc paths, so called because that is is the convention used by Transarc, the company that first commercialized AFS. To build with the Transarc paths, specify --enable-transarc-paths as a configure option.
+
+There are a couple of side effects that you need need to be aware of when building with the --enable-transarc-paths mode. First of all, the typical make install target does not work in this mode. Instead the 'make dest' target is used to build a directory of the binaries to be copied to the target system. Secondly, the packaging targets are not executed, so for example the redhat spec file is not generated to build the rpms.
+
+To build with Keberos support (recommended), you'll need to have the Keberos development libraries, and if available for your platform, the krb5-config. You will need the full path the the krb5-config script. For example
+
+ $ which krb5-config
+ /usr/bin/krb5-config
+
+To build [[OpenAFS]] with Kerberos 5 support and the Transarc path conventions:
+
+ ./configure --enable-transarc-paths --with-krb5-conf=/usr/bin/krb5-config
+ make
+ make dest
-After make dest, the binaries are placed into a dest directory, under a platform directory, such as i386\_linux26.
+If all goes well, then the binaries are located in a platform sub-directory, the name of which is platform specific, for example 'i386\_linux26/dest'.
+
+The 'make install' command does not work with Transarc paths. You will have to manually copy the binaries into place after running make dest. For more information, see the Quick Start Guide for Unix on the [[OpenAFS]] documentation page.
# cp -r i386_linux26/dest/root.client/usr/vice/etc/modload /usr/vice/etc
# cp i386_linux26/dest/root.client/usr/vice/etc/afsd /usr/vice/etc
# cp -r i386_linux26/dest/lib /usr/afsws
# cp -r i386_linux26/dest/root.server/usr/afs/* /usr/afs
-### <a name="Initial testing"></a> Initial testing
+### <a name="Running the Test Suite"></a> Running the Test Suite
+
+[[OpenAFS]] includes a suite of basic test scripts in the src/tests directory. The tests directory also contains a utility called afs-newcell.pl to create a test cell on a single host. You will need to already have a kerberos server running with an AFS principal and an admin principal. You should also have a separate partition mounted as /vicepa for the test volumes. See the src/tests/afs-newcell.pl for details.
+
+The paths to the binaries and AFS configuration is are written to src/tests/OpenAFS/Dirpaths.pm when the tests are built. This means you must run 'make all' in src/tests before attempting to use afs-newcell.pl. The afs-newcell.pl program must be run as root.
+
+ cd src/tests
+ make all
+ sudo perl afs-newcell.pl
+
+The afs-newcell.pl program will prompt for the required cell configuration unless started with the --batch options. You will need to specify the name of your new cell, the host name of this machine, the target partition id (a for /vicepa), and the username of the AFS administrator (which is admin by default). You will also need to provide the type of Kerberos server to be used, the name of the Kerberos realm (which can be different than the AFS cell name), and the location of the keytab file that contains the AFS server encryption key and the admin's encryption key. Finally, you may provide any server options for the AFS database and fileservers.
+
+ Server options:
+ What server name should be used? [host.domain.com]
+ What cellname should be used? [testcell]
+ What vice partition? [a]
+ What administrator username? [admin]
+
+ Kerberos options:
+ Which Kerberos is to be used? [MIT]
+ What Kerberos realm? [TESTCELL]
+ What keytab file? [/usr/afs/etc/krb5.keytab]
+
+ Database Server options:
+ ptserver options: []
+ vlserver options: []
+
+ Fileserver options:
+ Use DAFS fileserver (requires DAFS build option)? (yes/no) [no]
+ fileserver options: []
+ volserver options: []
+ salvageserver options: []
+ salvager options: []
+
+The parameters are optionally saved to a script called run-afs-newcell.sh, in case you need to re-run the setup. Also, the Kerberos parameters for the new cell are saved in the file run-tests.conf. If all goes well, the servers should be running and the client loaded. You should be able to see your new cell at /afs/testcell (or whatever you called your cell). Your cell will have a few, empty test volumes that you should be able to see with a vos listvldb.
+
+Run the run-tests script to run the AFS test suite. The program can be run as an ordinary user. The keytab file specified when afs-newcell was run will be used to authenticate the admin user for the tests.
+
+ ./run-tests -all
+
+The output will look something like this,
+
+ Authenticating to cell testcell.
+ ...
+ Running creat1
+ Running mkdir1
+ Running mkdir2
+ ....
+ -----------------------------------------------------------
+ Failed test(s) were: write-large write-ro ...
-- [[MichaelMeffie]] - 09 Oct 2007
git://git.openafs.org / openafs-wiki.git / blobdiff