1 <?xml version="1.0" encoding="UTF-8"?>
2 <chapter id="configuration">
3 <title>Setting up Netatalk</title>
6 <title>File Services<indexterm>
7 <primary>File Services</primary>
9 <secondary>Netatalk's File Services</secondary>
12 <para>Netatalk supplies AFP<indexterm>
13 <primary>AFP</primary>
15 <secondary>Apple Filing Protocol</secondary>
16 </indexterm> services.</para>
19 <title>Setting up the AFP file server</title>
21 <para>AFP (the Apple Filing Protocol) is the protocol Apple Macintoshes
22 use for file services. The protocol has evolved over the years. The
23 latest changes to the protocol, called "AFP 3.3", were added with the
24 release of Snow Leopard<indexterm>
25 <primary>Snow Leopard</primary>
27 <secondary>Mac OS X 10.6</secondary>
28 </indexterm> (Mac OS X 10.6).</para>
30 <para>The afpd daemon offers the fileservices to Apple clients. The only
31 configuration file is <filename>afp.conf</filename>. It uses a ini style
32 configuration syntax.</para>
34 <para>Mac OS X 10.5 (Leopard) added support for Time Machine backups
35 over AFP. Two new functions ensure that backups are written to spinning
36 disk, not just in the server's cache. Different host operating systems
37 honour this cache flushing differently. To make a volume a Time Machine
38 target use the volume option "<option>time machine =
41 <para>Starting with Netatalk 2.1 UNIX symlinks<indexterm>
42 <primary>symlink</primary>
44 <secondary>UNIX symlink</secondary>
45 </indexterm> can be used on the server. Semantics are the same as for
46 eg NFS, ie they are not resolved on the server side but instead it's
47 completely up to the client to resolve them, resulting in links that
48 point somewhere inside the clients filesystem view.</para>
51 <title>afp.conf</title>
53 <para><filename>afp.conf</filename> is the configuration file used by
54 afpd to determine the behaviour and configuration of the AFP file
55 serverand the AFP volume that it provides.</para>
57 <para>The <filename>afp.conf</filename> is divided into several
58 sections:<variablelist>
63 <para>The global section defines general server options</para>
71 <para>The homes section defines user home volumes</para>
74 </variablelist>Any section not called <option>Global</option> or
75 <option>Homes</option> is interpreted as an AFP volume.</para>
77 <para>For sharing user homes by defining a <option>Homes</option>
78 section you must specify the option <option>basedir regex</option>
79 which can be a simple string with the path to the parent directory of
80 all user homes or a regular expression.</para>
84 <para><programlisting>[Homes]
86 </programlisting></para>
88 <para>Now any user logging into the AFP server will have a user volume
89 available whos path is <filename>/home/NAME</filename>.</para>
91 <para>A more complex setup would be a server with a large amount of
92 user homes which are split across eg two different
93 filesystems:<itemizedlist>
95 <para>/RAID1/homes</para>
99 <para>/RAID2/morehomes</para>
101 </itemizedlist>The following configuration is
102 required:<programlisting>[Homes]
103 basedir regex = /RAID./.*homes
104 </programlisting></para>
106 <para>If <option>basedir regex</option> contains symlink, set the
107 canonicalized absolute path. When <filename>/home</filename> links to
108 <filename>/usr/home</filename>: <programlisting>[Homes]
109 basedir regex = /usr/home</programlisting></para>
111 <para>For a more detailed explanation of the available options, please
112 refer to the <citerefentry>
113 <refentrytitle>afp.conf</refentrytitle>
115 <manvolnum>5</manvolnum>
116 </citerefentry> man page.</para>
120 <sect2 id="CNID-backends">
121 <title>CNID<indexterm>
122 <primary>CNID</primary>
124 <secondary>Catalog Node ID</secondary>
125 </indexterm> backends<indexterm>
126 <primary>Backend</primary>
128 <secondary>CNID backend</secondary>
131 <para>Unlike other protocols like SMB or NFS, the AFP protocol mostly
132 refers to files and directories by ID and not by a path (the IDs are
133 also called CNID, that means Catalog Node ID). A typical AFP request
134 uses a directory ID<indexterm>
135 <primary>DID</primary>
137 <secondary>Directory ID</secondary>
138 </indexterm> and a filename, something like <phrase>"server, please
139 open the file named 'Test' in the directory with id 167"</phrase>. For
140 example "Aliases" on the Mac basically work by ID (with a fallback to
141 the absolute path in more recent AFP clients. But this applies only to
142 Finder, not to applications).</para>
144 <para>Every file in an AFP volume has to have a unique file ID<indexterm>
145 <primary>FID</primary>
147 <secondary>File ID</secondary>
148 </indexterm>, IDs must, according to the specs, never be reused, and
149 IDs are 32 bit numbers (Directory IDs use the same ID pool). So, after
150 ~4 billion files/folders have been written to an AFP volume, the ID pool
151 is depleted and no new file can be written to the volume. No whining
154 <para>Netatalk needs to map IDs to files and folders in the host
155 filesystem. To achieve this, several different CNID backends<indexterm>
156 <primary>CNID backend</primary>
157 </indexterm> are available and can be choosed by the <option>cnid
158 scheme</option><indexterm>
159 <primary>cnidscheme</primary>
161 <secondary>specifying a CNID backend</secondary>
162 </indexterm> option in the <citerefentry>
163 <refentrytitle>afp.conf</refentrytitle>
165 <manvolnum>5</manvolnum>
166 </citerefentry> configuration file. A CNID backend is basically a
167 database storing ID <-> name mappings.</para>
169 <para>The CNID Databases are by default located in
170 <filename>/var/netatalk/CNID</filename>.</para>
172 <para>There is a command line utility called <command>dbd</command>
173 available which can be used to verify, repair and rebuild the CNID
177 <para>There are some CNID related things you should keep in mind when
178 working with netatalk:</para>
182 <para>Don't nest volumes<indexterm>
183 <primary>Nested volumes</primary>
188 <para>CNID backends are databases, so they turn afpd into a file
189 server/database mix.</para>
193 <para>If there's no more space on the filesystem left, the
194 database will get corrupted. You can work around this by either
195 using the <option>vol dbpath</option> option and put the database
196 files into another location or, if you use quotas, make sure the
197 CNID database folder is owned by a user/group without a
199 <primary>Quotas</primary>
201 <secondary>Disk usage quotas</secondary>
206 <para>Be careful with CNID databases for volumes that are mounted
207 via NFS. That is a pretty audacious decision to make anyway, but
208 putting a database there as well is really asking for trouble,
209 i.e. database corruption. Use the <option>vol dbpath</option>
210 directive to put the databases onto a local disk if you must use
212 <primary>NFS</primary>
214 <secondary>Network File System</secondary>
215 </indexterm> mounted volumes.</para>
221 <title>cdb<indexterm>
222 <primary>CDB</primary>
224 <secondary>"cdb" CNID backend</secondary>
227 <para>The "concurrent database" backend is based on Berkeley DB. With
228 this backend, several afpd daemons access the CNID database directly.
229 Berkeley DB locking is used to synchronize access, if more than one
230 afpd process is active for a volume. The drawback is, that the crash
231 of a single afpd process might corrupt the database. cdb should only
232 be used when sharing home directories for a larger number of users
233 <emphasis>and</emphasis> it has been determined that a large number of
234 <command>cnid_dbd</command> processes is problematic.</para>
238 <title>dbd<indexterm>
239 <primary>DBD</primary>
241 <secondary>"dbd" CNID backend</secondary>
244 <para>Access to the CNID database is restricted to the cnid_dbd daemon
245 process. afpd processes communicate with the daemon for database reads
246 and updates. The probability for database corruption is practically
249 <para>This is the default backend since Netatalk 2.1.</para>
253 <title>tdb<indexterm>
254 <primary>tdb</primary>
256 <secondary>"tdb" CNID backend</secondary>
259 <para><abbrev>tdb</abbrev> is another persistent CNID database, it's
260 Samba's <emphasis>Trivial Database</emphasis>. It could be used
261 instead of <abbrev>cdb</abbrev> for user volumes.<important>
262 <para>Only ever use it for volumes that are
263 <emphasis>not</emphasis> shared and accessed by multiple clients
265 </important>This backend is also used internally (as in-memory CNID
266 database) as a fallback in case opening the primary database can't be
267 opened, because <abbrev>tdb</abbrev> can work as in-memory database.
268 This of course means upon restart the CNIDs are gone.</para>
272 <title>last<indexterm>
273 <primary>Last</primary>
275 <secondary>"last" CNID backend</secondary>
278 <para>The last backend is a in-memory tdb database. It is not
279 persistent. Starting with netatalk 3.0, it becomes the <emphasis> read
280 only mode</emphasis> automatically. This is useful e.g. for
285 <sect2 id="charsets">
286 <title>Charsets<indexterm>
287 <primary>Charset</primary>
289 <secondary>character set</secondary>
290 </indexterm>/Unicode<indexterm>
291 <primary>Unicode</primary>
297 <title>Why Unicode?</title>
299 <para>Internally, computers don't know anything about characters and
300 texts, they only know numbers. Therefore, each letter is assigned a
301 number. A character set, often referred to as
302 <emphasis>charset</emphasis> or
303 <emphasis>codepage</emphasis><indexterm>
304 <primary>Codepage</primary>
305 </indexterm>, defines the mappings between numbers and
308 <para>If two or more computer systems need to communicate with each
309 other, the have to use the same character set. In the 1960s the
311 <primary>ASCII</primary>
313 <secondary>American Standard Code for Information
314 Interchange</secondary>
315 </indexterm> (American Standard Code for Information Interchange)
316 character set was defined by the American Standards Association. The
317 original form of ASCII represented 128 characters, more than enough to
318 cover the English alphabet and numerals. Up to date, ASCII has been
319 the normative character scheme used by computers.</para>
321 <para>Later versions defined 256 characters to produce a more
322 international fluency and to include some slightly esoteric graphical
323 characters. Using this mode of encoding each character takes exactly
324 one byte. Obviously, 256 characters still wasn't enough to map all the
325 characters used in the various languages into one character
328 <para>As a result localized character sets were defined later, e.g the
329 ISO-8859 character sets. Most operating system vendors introduced
330 their own characters sets to satisfy their needs, e.g. IBM defined the
331 <emphasis>codepage 437 (DOSLatinUS)</emphasis>, Apple introduced the
332 <emphasis>MacRoman</emphasis><indexterm>
333 <primary>MacRoman</primary>
335 <secondary>MacRoman charset</secondary>
336 </indexterm> codepage and so on. The characters that were assigned
337 number larger than 127 were referred to as
338 <emphasis>extended</emphasis> characters. These character sets
339 conflict with another, as they use the same number for different
340 characters, or vice versa.</para>
342 <para>Almost all of those characters sets defined 256 characters,
343 where the first 128 (0-127) character mappings are identical to ASCII.
344 As a result, communication between systems using different codepages
345 was effectively limited to the ASCII charset.</para>
347 <para>To solve this problem new, larger character sets were defined.
348 To make room for more character mappings, these character sets use at
349 least 2 bytes to store a character. They are therefore referred to as
350 <emphasis>multibyte</emphasis> character sets.</para>
352 <para>One standardized multibyte charset encoding scheme is known as
353 <ulink url="http://www.unicode.org/">unicode</ulink>. A big advantage
354 of using a multibyte charset is that you only need one. There is no
355 need to make sure two computers use the same charset when they are
356 communicating.</para>
360 <title>character sets used by Apple</title>
362 <para>In the past, Apple clients used single-byte charsets to
363 communicate over the network. Over the years Apple defined a number of
364 codepages, western users will most likely be using the
365 <emphasis>MacRoman</emphasis> codepage.</para>
367 <para>Codepages defined by Apple include:</para>
371 <para>MacArabic, MacFarsi</para>
375 <para>MacCentralEurope</para>
379 <para>MacChineseSimple</para>
383 <para>MacChineseTraditional</para>
387 <para>MacCroation</para>
391 <para>MacCyrillic</para>
395 <para>MacDevanagari</para>
399 <para>MacGreek</para>
403 <para>MacHebrew</para>
407 <para>MacIcelandic</para>
411 <para>MacJapanese</para>
415 <para>MacKorean</para>
419 <para>MacRoman</para>
423 <para>MacRomanian</para>
431 <para>MacTurkish</para>
435 <para>Starting with Mac OS X and AFP3, <ulink
436 url="http://www.utf-8.com/">UTF-8</ulink> is used. UTF-8 encodes
437 Unicode characters in an ASCII compatible way, each Unicode character
438 is encoded into 1-6 ASCII characters. UTF-8 is therefore not really a
439 charset itself, it's an encoding of the Unicode charset.</para>
441 <para>To complicate things, Unicode defines several <emphasis> <ulink
442 url="http://www.unicode.org/reports/tr15/index.html">normalization</ulink>
443 </emphasis> forms. While <ulink
444 url="http://www.samba.org">samba</ulink><indexterm>
445 <primary>Samba</primary>
446 </indexterm> uses <emphasis>precomposed</emphasis><indexterm>
447 <primary>Precomposed</primary>
449 <secondary>Precomposed Unicode normalization</secondary>
450 </indexterm> Unicode, which most Unix tools prefer as well, Apple
451 decided to use the <emphasis>decomposed</emphasis><indexterm>
452 <primary>Decomposed</primary>
454 <secondary>Decomposed Unicode normalization</secondary>
455 </indexterm> normalization.</para>
457 <para>For example lets take the German character
458 '<keycode>ä</keycode>'. Using the precomposed normalization, Unicode
459 maps this character to 0xE4. In decomposed normalization, 'ä' is
460 actually mapped to two characters, 0x61 and 0x308. 0x61 is the mapping
461 for an 'a', 0x308 is the mapping for a <emphasis>COMBINING
462 DIAERESIS</emphasis>.</para>
464 <para>Netatalk refers to precomposed UTF-8 as
465 <emphasis>UTF8</emphasis><indexterm>
466 <primary>UTF8</primary>
468 <secondary>Netatalk's precomposed UTF-8 encoding</secondary>
469 </indexterm> and to decomposed UTF-8 as
470 <emphasis>UTF8-MAC</emphasis><indexterm>
471 <primary>UTF8-MAC</primary>
473 <secondary>Netatalk's decomposed UTF-8 encoding</secondary>
478 <title>afpd and character sets</title>
480 <para>To support new AFP 3.x and older AFP 2.x clients at the same
481 time, afpd needs to be able to convert between the various charsets
482 used. AFP 3.x clients always use UTF8-MAC, AFP 2.x clients use one of
483 the Apple codepages.</para>
485 <para>At the time of this writing, netatalk supports the following
486 Apple codepages:</para>
490 <para>MAC_CENTRALEUROPE</para>
494 <para>MAC_CHINESE_SIMP</para>
498 <para>MAC_CHINESE_TRAD</para>
502 <para>MAC_CYRILLIC</para>
506 <para>MAC_GREEK</para>
510 <para>MAC_HEBREW</para>
514 <para>MAC_JAPANESE</para>
518 <para>MAC_KOREAN</para>
522 <para>MAC_ROMAN</para>
526 <para>MAC_TURKISH</para>
530 <para>afpd handles three different character set options:</para>
534 <term>unix charset<indexterm>
535 <primary>unix charset</primary>
537 <secondary>afpd's unix charset setting</secondary>
541 <para>This is the codepage used internally by your operating
542 system. If not specified, it defaults to <option>UTF8</option>.
543 If <option>LOCALE</option> is specified and your system support
544 Unix locales, afpd tries to detect the codepage. afpd uses this
545 codepage to read its configuration files, so you can use
546 extended characters for volume names, login messages, etc. see
548 <refentrytitle>afp.conf</refentrytitle>
550 <manvolnum>5</manvolnum>
551 </citerefentry>.</para>
556 <term>mac charset<indexterm>
557 <primary>mac charset</primary>
559 <secondary>afpd's mac charset setting</secondary>
563 <para>As already mentioned, older Mac OS clients (up to AFP 2.2)
564 use codepages to communicate with afpd. However, there is no
565 support for negotiating the codepage used by the client in the
566 AFP protocol. If not specified otherwise, afpd assumes the
567 <emphasis>MacRoman</emphasis> codepage is used. In case you're
568 clients use another codepage, e.g.
569 <emphasis>MacCyrillic</emphasis>, you'll <emphasis
570 role="bold">have</emphasis> to explicitly configure this. see
572 <refentrytitle>afp.conf</refentrytitle>
574 <manvolnum>5</manvolnum>
575 </citerefentry>.</para>
580 <term>vol charset<indexterm>
581 <primary>vol charset</primary>
583 <secondary>afpd's vol charset setting</secondary>
587 <para>This defines the charset afpd should use for filenames on
588 disk. By default, it is the same as <option>unix
589 charset</option>. If you have <ulink
590 url="http://www.gnu.org/software/libiconv/">iconv</ulink><indexterm>
591 <primary>Iconv</primary>
593 <secondary>iconv encoding conversion engine</secondary>
594 </indexterm> installed, you can use any iconv provided charset
597 <para>afpd needs a way to preserve extended macintosh
598 characters, or characters illegal in unix filenames, when saving
599 files on a unix filesystem. Earlier versions used the the so
600 called CAP encoding<indexterm>
601 <primary>CAP encoding</primary>
603 <secondary>CAP style character encoding</secondary>
604 </indexterm>. An extended character (>0x7F) would be
605 converted to a :xx hex sequence, e.g. the Apple Logo (MacRoman:
606 0xF0) was saved as :f0. Some special characters will be
607 converted as to :xx notation as well. '/' will be encoded to
608 :2f, if <option>usedots</option> was not specified, a leading
609 dot '.' will be encoded as :2e.</para>
611 <para>Even though this version now uses <option>UTF8</option> as
612 the default encoding for filenames, '/' will be converted to
613 ':'. For western users another useful setting could be
614 <option>vol charset = ISO-8859-15</option>.</para>
616 <para>If a character cannot be converted from the <option>mac
617 charset</option> to the selected <option>vol charset</option>,
618 afpd will save it as a CAP encoded character. For AFP3 clients,
619 afpd will convert the UTF8 character to <option>mac
620 charset</option> first. If this conversion fails, you'll receive
621 a -50 error on the mac. <emphasis>Note</emphasis>: Whenever you
622 can, please stick with the default UTF8 volume format. see
624 <refentrytitle>afp.conf</refentrytitle>
626 <manvolnum>5</manvolnum>
627 </citerefentry>.</para>
634 <sect2 id="authentication">
635 <title>Authentication<indexterm>
636 <primary>Authentication</primary>
638 <secondary>between AFP client and server</secondary>
642 <title>AFP authentication basics</title>
644 <para>Apple chose a flexible model called "User Authentication
646 <primary>UAM</primary>
648 <secondary>User Authentication Module</secondary>
649 </indexterm> (UAMs) for authentication purposes between AFP client
650 and server. An AFP client initially connecting to an AFP server will
651 ask for the list of UAMs which the server provides, and will choose
652 the one with strongest encryption that the client supports.</para>
654 <para>Several UAMs have been developed by Apple over the time, some by
655 3rd-party developers.</para>
659 <title>UAMs supported by Netatalk</title>
661 <para>Netatalk supports the following ones by default:</para>
665 <para>"No User Authent"<indexterm>
666 <primary>No User Authent</primary>
668 <secondary>"No User Authent" UAM (guest access)</secondary>
669 </indexterm> UAM (guest access without authentication)</para>
673 <para>"Cleartxt Passwrd"<indexterm>
674 <primary>Cleartxt Passwrd</primary>
676 <secondary>"Cleartxt Passwrd" UAM</secondary>
677 </indexterm> UAM (no password encryption)</para>
681 <para>"Randnum exchange"<indexterm>
682 <primary>Randnum exchange</primary>
684 <secondary>"Randnum exchange" UAM</secondary>
685 </indexterm>/"2-Way Randnum exchange"<indexterm>
686 <primary>2-Way Randnum exchange</primary>
688 <secondary>"2-Way Randnum exchange" UAM</secondary>
689 </indexterm> UAMs (weak password encryption, separate password
694 <para>"DHCAST128"<indexterm>
695 <primary>DHCAST128</primary>
697 <secondary>"DHCAST128" UAM</secondary>
698 </indexterm> UAM (stronger password encryption)</para>
702 <para>"DHX2"<indexterm>
703 <primary>DHX2</primary>
705 <secondary>"DHX2" UAM</secondary>
706 </indexterm> UAM (successor of DHCAST128)</para>
710 <para>There exist other optional UAMs as well:</para>
714 <para>"PGPuam 1.0"<indexterm>
715 <primary>PGPuam 1.0</primary>
717 <secondary>"PGPuam 1.0" UAM</secondary>
718 </indexterm><indexterm>
719 <primary>uams_pgp.so</primary>
721 <secondary>"PGPuam 1.0" UAM</secondary>
722 </indexterm> UAM (PGP-based authentication for pre-Mac OS X
723 clients. You'll also need the <ulink
724 url="http://www.vmeng.com/vinnie/papers/pgpuam.html">PGPuam
725 client</ulink> to let this work)</para>
727 <para>You'll have to add <filename>"--enable-pgp-uam"</filename>
728 to your configure switches to have this UAM available.</para>
732 <para>"Kerberos IV"<indexterm>
733 <primary>Kerberos IV</primary>
735 <secondary>"Kerberos IV" UAM</secondary>
736 </indexterm><indexterm>
737 <primary>uams_krb4.so</primary>
739 <secondary>"Kerberos IV" UAM</secondary>
740 </indexterm>/"AFS Kerberos"<indexterm>
741 <primary>AFS Kerberos</primary>
743 <secondary>"AFS Kerberos" UAM (Kerberos IV)</secondary>
744 </indexterm> UAMs (suitable to use <ulink
745 url="http://web.mit.edu/macdev/KfM/Common/Documentation/faq.html">Kerberos
746 v4 based authentication</ulink> and AFS file servers)</para>
748 <para>Use <filename>"--enable-krb4-uam"</filename> at compile time
749 to activate the build of this UAM.</para>
753 <para>"Client Krb v2"<indexterm>
754 <primary>Client Krb v2</primary>
756 <secondary>"Client Krb v2" UAM (Kerberos V)</secondary>
757 </indexterm> UAM (Kerberos V, suitable for "Single Sign On"
758 Scenarios with OS X clients -- see below)</para>
760 <para><filename>"--enable-krbV-uam"</filename> will provide you
761 with the ability to use this UAM.</para>
765 <para>You can configure which UAMs should be activated by defining
766 "<option>uam list</option>" in <option>Global</option> section.
767 <command>afpd</command> will log which UAMs it's using and if problems
768 occur while activating them in either
769 <filename>netatalk.log</filename> or syslog at startup time.
771 <refentrytitle>asip-status.pl</refentrytitle>
773 <manvolnum>1</manvolnum>
774 </citerefentry> can be used to query the available UAMs of AFP
775 servers as well.</para>
777 <para>Having a specific UAM available at the server does not
778 automatically mean that a client can use it. Client-side support is
779 also necessary. For older Macintoshes running Mac OS < X DHCAST128
780 support exists since AppleShare client 3.8.x.</para>
782 <para>On OS X, there exist some client-side techniques to make the
783 AFP-client more verbose, so one can have a look what's happening while
784 negotiating the UAMs to use. Compare with this <ulink
785 url="http://article.gmane.org/gmane.network.netatalk.devel/7383/">hint</ulink>.</para>
789 <title>Which UAMs to activate?</title>
791 <para>It depends primarily on your needs and on the kind of Mac OS
792 versions you have to support. Basically one should try to use
793 DHCAST128 and DHX2 where possible because of its strength of password
798 <para>Unless you really have to supply guest access to your
799 server's volumes ensure that you disable "No User Authent" since
800 it might lead accidentally to unauthorized access. In case you
801 must enable guest access take care that you enforce this on a per
802 volume base using the access controls.</para>
806 <para>The "ClearTxt Passwrd" UAM is as bad as it sounds since
807 passwords go unencrypted over the wire. Try to avoid it at both
808 the server's side as well as on the client's. Note: If you want to
809 provide Mac OS 8/9 clients with NetBoot-services then you need
810 uams_cleartext.so since the AFP-client integrated into the Mac's
811 firmware can only deal with this basic form of
812 authentication.</para>
816 <para>Since "Randnum exchange"/"2-Way Randnum exchange" uses only
817 56 bit DES for encryption it should be avoided as well. Another
818 disadvantage is the fact that the passwords have to be stored in
819 cleartext on the server and that it doesn't integrate into both
820 PAM scenarios or classic /etc/shadow (you have to administrate
821 passwords separately by using the <citerefentry>
822 <refentrytitle>afppasswd</refentrytitle>
824 <manvolnum>1</manvolnum>
825 </citerefentry> utility, if clients should use these
830 <para>"DHCAST128" or "DHX2" should be a good compromise for most
831 people since it combines stronger encryption with PAM
836 <para>Using the Kerberos V<indexterm>
837 <primary>Kerberos V</primary>
839 <secondary>"Client Krb v2" UAM</secondary>
840 </indexterm> ("Client Krb v2") UAM, it's possible to implement
841 real single sign on scenarios using Kerberos tickets. The password
842 is not sent over the network. Instead, the user password is used
843 to decrypt a service ticket for the appleshare server. The service
844 ticket contains an encryption key for the client and some
845 encrypted data (which only the appleshare server can decrypt). The
846 encrypted portion of the service ticket is sent to the server and
847 used to authenticate the user. Because of the way that the afpd
848 service principal detection is implemented, this authentication
849 method is vulnerable to man-in-the-middle attacks.</para>
853 <para>For a more detailed overview over the technical implications of
854 the different UAMs, please have a look at Apple's <ulink
855 url="http://developer.apple.com/library/mac/#documentation/Networking/Conceptual/AFP/AFPSecurity/AFPSecurity.html#//apple_ref/doc/uid/TP40000854-CH232-SW1">File
856 Server Security</ulink> pages.</para>
860 <title>Using different authentication sources with specific
863 <para>Some UAMs provide the ability to use different authentication
864 "backends", namely <filename>uams_cleartext.so</filename>,
865 <filename>uams_dhx.so</filename> and
866 <filename>uams_dhx2.so</filename>. They can use either classic Unix
867 passwords from <filename>/etc/passwd</filename>
868 (<filename>/etc/shadow</filename>) or PAM if the system supports that.
869 <filename>uams_cleartext.so</filename> can be symlinked to either
870 <filename>uams_passwd.so</filename> or
871 <filename>uams_pam.so</filename>, <filename>uams_dhx.so</filename> to
872 <filename>uams_dhx_passwd.so</filename> or
873 <filename>uams_dhx_pam.so</filename> and
874 <filename>uams_dhx2.so</filename> to
875 <filename>uams_dhx2_passwd.so</filename> or
876 <filename>uams_dhx2_pam.so</filename>.</para>
878 <para>So, if it looks like this in Netatalk's UAMs folder (per default
879 <filename>/etc/netatalk/uams/</filename>):<programlisting>uams_clrtxt.so -> uams_pam.so
880 uams_dhx.so -> uams_dhx_pam.so
881 uams_dhx2.so -> uams_dhx2_pam.so</programlisting> then you're using PAM,
882 otherwise classic Unix passwords. The main advantage of using PAM is
883 that one can integrate Netatalk in centralized authentication
884 scenarios, eg. via LDAP, NIS and the like. Please always keep in mind
885 that the protection of your user's login credentials in such scenarios
886 also depends on the strength of encryption that the UAM in question
887 supplies. So think about eliminating weak UAMs like "ClearTxt Passwrd"
888 and "Randnum exchange" completely from your network.</para>
892 <title>Netatalk UAM overview table</title>
894 <para>A small overview of the most common used UAMs.</para>
896 <table orient="land">
897 <title>Netatalk UAM overview</title>
899 <tgroup align="center" cols="7">
900 <colspec colname="col1" colnum="1" colwidth="0.5*" />
902 <colspec colname="uam_guest" colnum="2" colwidth="1*" />
904 <colspec colname="uam_clrtxt" colnum="3" colwidth="1*" />
906 <colspec colname="uam_randnum" colnum="4" colwidth="1*" />
908 <colspec colname="uam_dhx" colnum="5" colwidth="1*" />
910 <colspec colname="uam_dhx2" colnum="6" colwidth="1*" />
912 <colspec colname="uam_gss" colnum="7" colwidth="1*" />
916 <entry align="center" rotate="0" valign="middle">UAM</entry>
918 <entry>No User Authent<indexterm>
919 <primary>uams_guest.so</primary>
921 <secondary>"No User Authent" UAM (guest
925 <entry>Cleartxt Passwrd<indexterm>
926 <primary>uams_cleartxt.so</primary>
928 <secondary>"Cleartxt Passwrd" UAM</secondary>
931 <entry>(2-Way) Randnum exchange<indexterm>
932 <primary>uams_randnum.so</primary>
934 <secondary>"(2-Way) Randnum exchange" UAM</secondary>
937 <entry>DHCAST128<indexterm>
938 <primary>uams_dhx.so</primary>
940 <secondary>"DHCAST128" UAM</secondary>
943 <entry>DHX2<indexterm>
944 <primary>uams_dhx2.so</primary>
946 <secondary>"DHX2" UAM</secondary>
949 <entry>Client Krb v2<indexterm>
950 <primary>uams_gss.so</primary>
952 <secondary>"Client Krb v2" UAM (Kerberos V)</secondary>
957 <entry align="center" rotate="0" valign="middle">pssword
960 <entry>guest access</entry>
962 <entry>max. 8 characters</entry>
964 <entry>max. 8 characters</entry>
966 <entry>max. 64 characters</entry>
968 <entry>max. 256 characters</entry>
970 <entry>Kerberos tickets</entry>
974 <entry align="center" rotate="0" valign="middle">Client
977 <entry>built-in into all Mac OS versions</entry>
979 <entry>built-in in all Mac OS versions except 10.0. Has to be
980 activated explicitly in recent Mac OS X versions</entry>
982 <entry>built-in into almost all Mac OS versions</entry>
984 <entry>built-in since AppleShare client 3.8.4, available as a
985 plug-in for 3.8.3, integrated in Mac OS X' AFP client</entry>
987 <entry>built-in since Mac OS X 10.2</entry>
989 <entry>built-in since Mac OS X 10.2</entry>
993 <entry align="center" rotate="0"
994 valign="middle">Encryption</entry>
996 <entry>Enables guest access without authentication between
997 client and server.</entry>
999 <entry>Password will be sent in cleartext over the wire. Just
1000 as bad as it sounds, therefore avoid at all if possible (note:
1001 providing NetBoot services requires the ClearTxt UAM)</entry>
1003 <entry>8-byte random numbers are sent over the wire,
1004 comparable with DES, 56 bits. Vulnerable to offline dictionary
1005 attack. Requires passwords in clear on the server.</entry>
1007 <entry>Password will be encrypted with 128 bit SSL, user will
1008 be authenticated against the server but not vice versa.
1009 Therefor weak against man-in-the-middle attacks.</entry>
1011 <entry>Password will be encrypted using libgcrypt with CAST
1012 128 in CBC mode. User will be authenticated against the server
1013 but not vice versa. Therefor weak against man-in-the-middle
1016 <entry>Password is not sent over the network. Due to the
1017 service principal detection method, this authentication method
1018 is vulnerable to man-in-the-middle attacks.</entry>
1022 <entry align="center" rotate="0" valign="middle">Server
1025 <entry align="center" valign="middle">uams_guest.so</entry>
1027 <entry align="center" valign="middle">uams_cleartxt.so</entry>
1029 <entry align="center" valign="middle">uams_randnum.so</entry>
1031 <entry align="center" valign="middle">uams_dhx.so</entry>
1033 <entry align="center" valign="middle">uams_dhx2.so</entry>
1035 <entry align="center" valign="middle">uams_gss.so</entry>
1039 <entry align="center" rotate="0" valign="middle">Password
1040 storage method</entry>
1042 <entry align="center" valign="middle">None</entry>
1044 <entry align="center" valign="middle">Either /etc/passwd
1045 (/etc/shadow) or PAM</entry>
1047 <entry align="center" valign="middle">Passwords stored in
1048 clear text in a separate text file</entry>
1050 <entry align="center" valign="middle">Either /etc/passwd
1051 (/etc/shadow) or PAM</entry>
1053 <entry align="center" valign="middle">Either /etc/passwd
1054 (/etc/shadow) or PAM</entry>
1056 <entry align="center" valign="middle">At the Kerberos Key
1057 Distribution Center*</entry>
1063 <para>* Have a look at this <ulink
1064 url="http://cryptnet.net/fdp/admin/kerby-infra/en/kerby-infra.html">Kerberos
1065 overview</ulink></para>
1068 <sect3 id="sshtunnel">
1069 <title>SSH tunneling</title>
1071 <para>Tunneling and all sort of VPN stuff has nothing to do with AFP
1072 authentication and UAMs in general. But since Apple introduced an
1073 option called "Allow Secure Connections Using SSH" and many people
1074 tend to confuse both things, we'll speak about that here too.</para>
1076 <sect4 id="manualsshtunnel">
1077 <title>Manually tunneling an AFP session</title>
1079 <para>This works since the first AFP servers that spoke "AFP over
1080 TCP" appeared in networks. One simply tunnels the remote server's
1081 AFP port to a local port different than 548 and connects locally to
1082 this port afterwards. On OS X this can be done by</para>
1084 <programlisting>ssh -l $USER $SERVER -L 10548:127.0.0.1:548 sleep 3000</programlisting>
1086 <para>After establishing the tunnel one will use
1087 <filename>"afp://127.0.0.1:10548"</filename> in the "Connect to
1088 server" dialog. All AFP traffic including the initial connection
1089 attempts will be sent encrypted over the wire since the local AFP
1090 client will connect to the Mac's local port 10548 which will be
1091 forwarded to the remote server's AFP port (we used the default 548)
1094 <para>These sorts of tunnels are an ideal solution if you've to
1095 access an AFP server providing weak authentications mechanisms
1096 through the Internet without having the ability to use a "real" VPN.
1097 Note that you can let <command>ssh</command> compress the data by
1098 using its "-C" switch and that the tunnel endpoints can be different
1099 from both AFP client and server (compare with the SSH documentation
1100 for details).</para>
1103 <sect4 id="autosshtunnel">
1104 <title>Automatically establishing a tunneled AFP connection</title>
1106 <para>From Mac OS X 10.2 to 10.4, Apple added an "Allow Secure
1107 Connections Using SSH" checkbox to the "Connect to Server" dialog.
1108 The idea behind: When the server signals that it can be contacted by
1109 SSH then Mac OS X' AFP client tries to establish the tunnel and
1110 automagically sends all AFP traffic through it.</para>
1112 <para>But it took until the release of Mac OS X 10.3 that this
1113 feature worked the first time... partly. In case, the SSH tunnel
1114 can't be established the AFP client <emphasis
1115 role="strong">silently</emphasis> fell back to an unencrypted AFP
1116 connection attempt.</para>
1118 <para>Netatalk's afpd will report that it is capable of handling SSH
1119 tunneled AFP requests, when both "<option>advertise ssh</option>"
1120 and "<option>fqdn</option>" options are set in
1121 <option>Global</option> section (double check with <citerefentry>
1122 <refentrytitle>asip-status.pl</refentrytitle>
1124 <manvolnum>1</manvolnum>
1125 </citerefentry> after you restarted afpd when you made changes to
1126 the settings). But there are a couple of reasons why you don't want
1127 to use this option at all:</para>
1131 <para>Tunneling TCP over TCP (as SSH does) is not the best idea.
1132 There exist better solutions like VPNs based on the IP
1137 <para>Since this SSH kludge isn't a normal UAM that integrates
1138 directly into the AFP authentication mechanisms but instead uses
1139 a single flag signalling clients whether they can <emphasis
1140 role="strong">try</emphasis> to establish a tunnel or not, it
1141 makes life harder to see what's happening when things go
1146 <para>You cannot control which machines are logged on by
1147 Netatalk tools like a <command>macusers</command> since all
1148 connection attempts seem to be made from localhost.</para>
1152 <para>On the other side you've to limit access to afpd to
1153 localhost only (TCP wrappers) when you want to ensure that all
1154 AFP sessions are SSH encrypted or...</para>
1158 <para>...when you're using 10.2 - 10.3.3 then you get the
1159 opposite of what you'd expect: potentially unencrypted AFP
1160 communication (including logon credentials) on the network
1161 without a single notification that establishing the tunnel
1162 failed. Apple fixed that not until Mac OS X 10.3.4.</para>
1166 <para>Encrypting all AFP sessions via SSH can lead to a
1167 significantly higher load on the Netatalk server</para>
1175 <title>ACL Support<indexterm>
1176 <primary>ACLs</primary>
1177 </indexterm></title>
1179 <para>ACL support for AFP is implemented for ZFS ACLs on Solaris and
1180 derived platforms and for POSIX 1e ACLs on Linux.</para>
1183 <title>Configuration</title>
1185 <para>For a basic mode of operation there's nothing to configure.
1186 Netatalk reads ACLs on the fly and calculates effective permissions
1187 which are then send to the AFP client via the so called
1189 <primary>UARights</primary>
1190 </indexterm> permission bits. On a Mac, the Finder uses these bits
1191 to adjust permission in Finder windows. For example folder whos UNIX
1192 mode would only result in in read-only permissions for a user will not
1193 be displayed with a read-only icon and the user will be able to write
1194 to the folder given the folder has an ACL giving the user write
1197 <para>By default, the effective permission of the authenticated user
1198 are only mapped to the mentioned UARights<indexterm>
1199 <primary>UARights</primary>
1200 </indexterm>permission structure, not the UNIX mode. You can adjust
1201 this behaviour with the configuration option <link
1202 linkend="map_acls">map acls</link>.</para>
1204 <para>However, neither in Finder "Get Info" windows nor in Terminal
1205 will you be able to see the ACLs, that's a result of how ACLs in OS X
1206 are designed. If you want to be able to display ACLs on the client,
1207 things get more involved as you must then setup both client and server
1208 to be part on a authentication domain (directory service, eg LDAP,
1209 OpenDirectory). The reason is, that in OS X ACLs are bound to UUIDs,
1210 not just uid's or gid's. Therefor afpd must be able to map every
1211 filesystem uid and gid to a UUID so that it can return the server side
1212 ACLs which are bound to UNIX uid and gid mapped to OS X UUIDs.</para>
1214 <para>Netatalk can query a directory server using LDAP queries. Either
1215 the directory server already provides an UUID attribute for user and
1216 groups (Active Directory, Open Directory) or you reuse an unused
1217 attribute (or add a new one) to you directory server (eg
1220 <para>In detail:</para>
1224 <para>For Solaris/ZFS: ZFS Volumes</para>
1226 <para>You should configure a ZFS ACL know for any volume you want
1227 to use with Netatalk:</para>
1229 <screen>aclinherit = passthrough
1230 aclmode = passthrough</screen>
1232 <para>For an explanation of what this knob does and how to apply
1233 it, check your hosts ZFS documentation (eg man zfs).</para>
1237 <para>Authentication Domain</para>
1239 <para>Your server and the clients must be part of a security
1240 association where identity data is coming from a common source.
1241 ACLs in Darwin are based on UUIDs and so is the ACL specification
1242 in AFP 3.2. Therefor your source of identity data has to provide
1243 an attribute for every user and group where a UUID is stored as a
1244 ASCII string. In other words:</para>
1248 <para>you need an Open Directory Server or an LDAP server
1249 where you store UUIDs in some attribute</para>
1253 <para>your clients must be configured to use this
1258 <para>your server should be configured to use this server via
1259 nsswitch and PAM</para>
1263 <para>configure Netatalk via the special <link
1264 linkend="acl_options">LDAP options for ACLs</link> in <link
1265 linkend="afp.conf.5">afp.conf</link> so that Netatalk is able
1266 to retrieve the UUID for users and groups via LDAP search
1275 <title>OS X ACLs</title>
1277 <para>With Access Control Lists (ACLs) Mac OS X offers a powerful
1278 extension of the traditional UNIX permissions model. An ACL is an
1279 ordered list of Access Control Entries (ACEs) explicitly granting or
1280 denying a set of permissions to a given user or group.</para>
1282 <para>Unlike UNIX permissions, which are bound to user or group IDs,
1283 ACLs are tied to UUIDs. For this reason accessing an object's ACL
1284 requires server and client to use a common directory service which
1285 translates between UUIDs and user/group IDs.</para>
1287 <para>ACLs and UNIX permissions interact in a rather simple way. As
1288 ACLs are optional UNIX permissions act as a default mechanism for
1289 access control. Changing an objects's UNIX permissions will leave it's
1290 ACL intact and modifying an ACL will never change the object's UNIX
1291 permissions. While doing access checks, OS X first examines an
1292 object's ACL evaluating ACEs in order until all requested rights have
1293 been granted, a requested right has been explicitly denied by an ACE
1294 or the end of the list has been reached. In case there is no ACL or
1295 the permissions granted by the ACL are not sufficient to fulfill the
1296 request, OS X next evaluates the object's UNIX permissions. Therefore
1297 ACLs always have precedence over UNIX permissions.</para>
1301 <title>ZFS ACLs</title>
1303 <para>ZFS ACLs closely match OS X ACLs. Both offer mostly identical
1304 fine grained permissions and inheritance settings.</para>
1308 <title>POSIX ACLs</title>
1311 <title>Overview</title>
1313 <para>Compared to OS X or NFSv4 ACLs, Posix ACLs represent a
1314 different, less versatile approach to overcome the limitations of
1315 the traditional UNIX permissions. Implementations are based on the
1316 withdrawn Posix 1003.1e standard.</para>
1318 <para>The standard defines two types of ACLs. Files and directories
1319 can have access ACLs which are consulted for access checks.
1320 Directories can also have default ACLs irrelevant to access checks.
1321 When a new object is created inside a directory with a default ACL,
1322 the default ACL is applied to the new object as it's access ACL.
1323 Subdirectories inherit default ACLs from their parent. There are no
1324 further mechanisms of inheritance control. </para>
1326 <para>Architectural differences between Posix ACLs and OS X ACLs
1327 especially involve:</para>
1329 <para><itemizedlist>
1331 <para>No fine-granular permissions model. Like UNIX
1332 permissions Posix ACLs only differentiate between read, write
1333 and execute permissions.</para>
1337 <para>Entries within an ACL are unordered.</para>
1341 <para>Posix ACLs can only grant rights. There is no way to
1342 explicitly deny rights by an entry.</para>
1346 <para>UNIX permissions are integrated into an ACL as special
1349 </itemizedlist></para>
1351 <para>Posix 1003.1e defines 6 different types of ACL entries. The
1352 first three types are used to integrate standard UNIX permissions.
1353 They form a minimal ACL, their presence is mandatory and only one
1354 entry of each type is allowed within an ACL.</para>
1356 <para><itemizedlist>
1358 <para>ACL_USER_OBJ: the owner's access rights.</para>
1362 <para>ACL_GROUP_OBJ: the owning group's access rights.</para>
1366 <para>ACL_OTHER: everybody's access rights.</para>
1368 </itemizedlist></para>
1370 <para>The remaining entry types expand the traditional permissions
1373 <para><itemizedlist>
1375 <para>ACL_USER: grants access rights to a certain user.</para>
1379 <para>ACL_GROUP: grants access rights to a certain
1384 <para>ACL_MASK: limits the maximum access rights which can be
1385 granted by entries of type ACL_GROUP_OBJ, ACL_USER and
1386 ACL_GROUP. As the name suggests, this entry acts as a mask.
1387 Only one ACL_MASK entry is allowed per ACL. If an ACL contains
1388 ACL_USER or ACL_GROUP entries, an ACL_MASK entry must be
1389 present too, otherwise it is optional.</para>
1391 </itemizedlist></para>
1393 <para>In order to maintain compatibility with applications not aware
1394 of ACLs, Posix 1003.1e changes the semantics of system calls and
1395 utilities which retrieve or manipulate an objects UNIX permissions.
1396 In case an object only has a minimal ACL, the group permissions bits
1397 of the UNIX permissions correspond to the value of the ACL_GROUP_OBJ
1400 <para>However, if the ACL also contains an ACL_MASK entry, the
1401 behavior of those system calls and utilities is different. The group
1402 permissions bits of the UNIX permissions correspond to the value of
1403 the ACL_MASK entry, i. e. calling "chmod g-w" will not only revoke
1404 write access for the group, but for all entities which have been
1405 granted write access by ACL_USER or ACL_GROUP entries.</para>
1409 <title>Mapping POSIX ACLs to OS X ACLs</title>
1411 <para>When a client wants to read an object's ACL, afpd maps it's
1412 Posix ACL onto an equivalent OS X ACL. Writing an object's ACL
1413 requires afpd to map an OS X ACL onto a Posix ACL. Due to
1414 architectural restrictions of Posix ACLs, it is usually impossible
1415 to find an exact mapping so that the result of the mapping process
1416 will be an approximation of the original ACL's semantic.</para>
1418 <para><itemizedlist>
1420 <para>afpd silently discard entries which deny a set of
1421 permissions because they they can't be represented within the
1422 Posix architecture. </para>
1426 <para>As entries within Posix ACLs are unordered, it is
1427 impossible to preserve order.</para>
1431 <para>Inheritance control is subject to severe limitations as
1434 <para>Entries with the only_inherit flag set will only
1435 become part of the directory's default ACL.</para>
1439 <para>Entries with at least one of the flags
1440 file_inherit, directory_inherit or limit_inherit set,
1441 will become part of the directory's access and default
1442 ACL, but the restrictions they impose on inheritance
1443 will be ignored.</para>
1445 </itemizedlist></para>
1449 <para>The lack of a fine-granular permission model on the
1450 Posix side will normally result in an increase of granted
1453 </itemizedlist></para>
1455 <para>As OS X clients aren't aware of the Posix 1003.1e specific
1456 relationship between UNIX permissions and ACL_MASK, afpd does not
1457 expose this feature to the client to avoid compatibility issues and
1458 handles *unix permissions and ACLs the same way as Apple's reference
1459 implementation of AFP does. When an object's UNIX permissions are
1460 requested, afpd calculates proper group rights and returns the
1461 result together with the owner's and everybody's access rights to
1462 the caller via "permissions" and "ua_permissions" members of the
1463 FPUnixPrivs structure (see Apple Filing Protocol Reference, page
1464 181). Changing an object's permissions, afpd always updates
1465 ACL_USER_OBJ, ACL_GROUP_OBJ and ACL_OTHERS. If an ACL_MASK entry is
1466 present too, afpd recalculates it's value so that the new group
1467 rights become effective and existing entries of type ACL_USER or
1468 ACL_GROUP stay intact.</para>
1474 <title>Filesystem Change Events<indexterm>
1475 <primary>FCE</primary>
1476 </indexterm></title>
1478 <para>Netatalk includes a nifty filesystem change event mechanism where
1479 afpd processes notfiy interested listeners about certain filesystem
1480 event by UDP network datagrams.</para>
1482 <para>For the format of the UDP packets and for an example C application
1483 that demonstrates how to use these in a listener, take a look at the
1484 Netatalk sourcefile <filename>bin/misc/fce.c</filename>.</para>
1486 <para>The currently supported FCE events are<itemizedlist>
1488 <para>file modification (fmod)</para>
1492 <para>file deletion (fdel)</para>
1496 <para>directory deletion (ddel)</para>
1500 <para>file creation (fcre)</para>
1504 <para>directory deletion (ddel)</para>
1506 </itemizedlist></para>
1508 <para>For details on the available simple configuration options take a
1509 look at <filename><link
1510 linkend="fceconf">afp.conf</link></filename>.</para>
1515 <title>Starting and stopping Netatalk</title>
1517 <para>The Netatalk distribution comes with several operating system
1518 specific startup script templates that are tailored according to the
1519 options given to the "configure" script before compiling. Currently,
1520 templates are provided for RedHat (sysv style), RedHat (systemd style),
1521 SUSE (sysv style), SUSE (systemd style), Gentoo, NetBSD, Debian and
1522 Solaris. You can select to install the generated startup script(s)
1524 <primary>Startscript</primary>
1526 <secondary>startup script</secondary>
1527 </indexterm> by specifying a system type to "configure". To
1528 automatically install startup scripts give one of the available
1529 <option>--with-init-style</option> option to "configure".</para>
1531 <para>Since new releases of Linux distributions appear all the time and
1532 the startup procedure for the other systems mentioned above might change
1533 as well, it is probably a good idea to not blindly install a startup
1534 script but to look at it first to see if it will work on your system. If
1535 you use Netatalk as part of a fixed setup, like a Linux distribution, an
1536 RPM or a BSD package, things will probably have been arranged properly for
1537 you. The following therefore applies mostly for people who have compiled
1538 Netatalk themselves.</para>
1540 <para>The following daemon need to be started by whatever startup script
1541 mechanism is used:</para>
1545 <para>netatalk<indexterm>
1546 <primary>netatalk</primary>
1551 <para>Additionally, make sure that the configuration file
1552 <filename>afp.conf</filename> is in the right place.</para>