s4-kdc: create a 'pac' private grouping library

[Samba.git] / howto4.txt

Samba4 developer howto
======================

tridge@samba.org, December 2004

A more up to date version of this howto can be found in the wiki 
at http://wiki.samba.org/index.php/Samba4/HOWTO.

This is a very basic document on how to setup a simple Samba4
server. This is aimed at developers who are already familiar with
Samba3 and wish to participate in Samba4 development. This is not
aimed at production use of Samba4.

.. contents::

Step 1: download Samba4
-----------------------

If you have downloaded the Samba4 code via a tarball released from the
samba.org website, Step 1 has already been completed for you.  For testing
with the version released in the tarball, you may continue on to Step 2.  Note
that the references below to the top-level directory named "samba4" will
instead be based on the name of the tarball downloaded (e.g.
"samba-4.0.0alpha3" for the tarball samba-4.0.0alpha3.tar.gz).

There are 2 methods of doing this:

  method 1:  "rsync -avz samba.org::ftp/unpacked/samba_4_0_test/ samba4"

  method 2:  "git clone git://git.samba.org/samba.git samba4; cd samba4 && git checkout -b v4-0-test origin/v4-0-test; cd .."

both methods will create a directory called "samba4" in the current
directory. If you don't have rsync or git then install one of them. 

Since only released versions of Samba contain a pregenerated configure script, 
you will have to generate it by hand::

 $ cd samba4/source
 $ ./autogen.sh

Note that the above rsync command will give you a checked out git
repository. So if you also have git you can update it to the latest
version at some future date using::

  $ cd samba4
  $ git pull origin v4-0-test

Step 2: compile Samba4
----------------------

Recommended optional development libraries:
- acl and xattr development libraries
- gnutls
- readline

Run this::

  $ cd samba4/source
  $ ./configure
  $ make


Step 2bis: recompile Samba4
---------------------------

This only applies for those who are recompiling Samba4 after updating the code
(using "rsync" or "git").

Due to some imperfections in our actual build system (hope that this changes
soon) it is recommended to perform this after the source upgrade:

  $ cd samba4/source
  $ make clean
  $ ./autogen.sh
  $ ./configure
  $ make idl_full
  $ make

Not all the steps are needed every time but doing so makes sure that you won't
have old compiled objects standing in the way and cause malfunctions.
It also makes sure that changes in the IDL files are correctly catched up.

Step 3: install Samba4
----------------------

Run this as a user who have permission to write to the install
directory (defaults to /usr/local/samba). Use --prefix option to
configure above to change this.

::
 
  # make install

Step 4: provision Samba4
------------------------

The "provision" step sets up a basic user database. Be warned that this
removes all preexisting database data (if any)!

It must be run as a user with permission to write to the install directory
(typically "root").

::

  # cd source
  # ./setup/provision --realm=YOUR.REALM --domain=YOURDOM \
  #  --adminpass=SOMEPASSWORD --server-role='domain controller'

'YOURDOM' is the NT4 style domain name. 'YOUR.REALM' is your kerberos
realm, which is typically your DNS domain name.

If you provisioned a more recent Samba4 system already you should be able to
use the procedures shown in "upgrading-samba4.txt" to upgrade it and keep all
data.

When you are using Samba3 at the moment you could try the experimental script
"upgrade_from_s3" under the "setup" directory of the source
distribution (it isn't included in binary distributions yet).

Step 5: Create a simple smb.conf
--------------------------------

The provisioning will create a very simple smb.conf with no shares by
default. You will need to update it to add at least one share. For
example::

  [test]
        path = /data/test
        read only = no

Step 6: starting Samba4
-----------------------

The simplest is to just run "samba", but as a developer you may find
the following more useful::

   # samba -i -M single

that means "start samba without messages in stdout, and running a
single process. That mode of operation makes debugging samba with gdb
particularly easy.

Note that now it is no longer necessary to have an instance of nmbd
from Samba 3 running.  If you are running any smbd or nmbd processes
they need to be stopped before starting samba from Samba 4.

Make sure you put the bin and sbin directories from your new install
in your $PATH. Make sure you run the right version!

Step 7: testing Samba4
----------------------

try this command::

  $ smbclient //localhost/test -Uadministrator%SOMEPASSWORD


NOTE about filesystem support
-----------------------------

To use the advanced features of Samba4 you need a filesystem that
supports both the "user" and "system" xattr namespaces.

If you run Linux with a 2.6 kernel and ext3 this means you need to
include the option "user_xattr" in your /etc/fstab. For example::

   /dev/hda3            /home                   ext3    user_xattr     1 1

You also need to compile your kernel with the XATTR and SECURITY
options for your filesystem. For ext3 that means you need::

   CONFIG_EXT3_FS_XATTR=y
   CONFIG_EXT3_FS_SECURITY=y

If you are running a Linux 2.6 kernel with CONFIG_IKCONFIG_PROC
defined you can check this with the following command::

   $ zgrep CONFIG_EXT3_FS /proc/config.gz

If you don't have a filesystem with xattr support, then you can
simulate it by using the option::

   posix:eadb = /usr/local/samba/eadb.tdb

that will place all extra file attributes (NT ACLs, DOS EAs, streams
etc), in that tdb. It is not efficient, and doesn't scale well, but at
least it gives you a choice when you don't have a modern filesystem.

Testing your filesystem
-----------------------

To test your filesystem support, install the 'attr' package and run
the following 4 commands as root::

  # touch test.txt
  # setfattr -n user.test -v test test.txt
  # setfattr -n security.test -v test2 test.txt
  # getfattr -d test.txt
  # getfattr -n security.test -d test.txt

You should see output like this::

  # file: test.txt
  user.test="test"
  
  # file: test.txt
  security.test="test2"

If you get any "Operation not supported" errors then it means your
kernel is not configured correctly, or your filesystem is not mounted
with the right options.

If you get any "Operation not permitted" errors then it probably means
you didn't try the test as root.

..
        vim: ft=rest