1 Information for Netatalk Developers
2 ===================================
4 For basic installation instructions, see http://netatalk.sourceforge.net .
6 Netatalk is an implementation of "AFP over TCP".
7 DSI is a session layer used to carry AFP over TCP.
8 The complete stack looks like this:
16 -+---------------------------+- (kernel boundary)
18 +------------+--------------+
20 +------------+--------------+
22 +---------------------------+
24 +---------------------------+
28 The `configure' shell script attempts to guess correct values for
29 various system-dependent variables used during compilation. It uses
30 those values to create a `Makefile' in each directory of the package.
31 It may also create one or more `.h' files containing system-dependent
32 definitions. Finally, it creates a shell script `config.status' that
33 you can run in the future to recreate the current configuration, a file
34 `config.cache' that saves the results of its tests to speed up
35 reconfiguring, and a file `config.log' containing compiler output
36 (useful mainly for debugging `configure').
38 If you need to do unusual things to compile the package, please try
39 to figure out how `configure' could check whether to do them, and mail
40 diffs or instructions to the address given in the `README' so they can
41 be considered for the next release. If at some point `config.cache'
42 contains results you don't want to keep, you may remove or edit it.
44 The file `configure.in' is used to create `configure' by a program
45 called `autoconf'. You only need `configure.in' if you want to change
46 it or regenerate `configure' using a newer version of `autoconf'.
52 Libtool encapsulates the platform specific dependencies for the
53 creation of libraries. It determines if the local platform can support
54 shared libraries or if it only supports static libraries.
56 Documentation: http://www.gnu.org/software/libtool/
59 GNU m4 is an implementation of the Unix macro processor. It reads
60 stdin and copies to stdout expanding defined macros as it processes
63 Documentation: http://www.gnu.org/software/m4/
66 Autoconf is a package of m4 macros that produce shell scripts to
67 configure source code packages.
69 Documentation: http://www.gnu.org/software/autoconf/
72 Automake is a tool that generates 'Makefile.in' files.
74 Documentation: http://www.gnu.org/software/automake/
78 5. OpenSSL and/or Libgcrypt
79 The OpenSSL Project is a collaborative effort to develop a robust,
80 commercial-grade, full-featured, and Open Source toolkit implementing
81 the Secure Sockets Layer (SSL v2/v3) and Transport Layer Security (TLS
82 v1) protocols as well as a full-strength general purpose cryptography
84 This is required to enable DHX login support.
86 Get everything at http://www.openssl.org/
88 The Libgcrypt is a general purpose cryptographic library based on
90 This is required to enable DHX2 login support.
92 Get everything at http://directory.fsf.org/project/libgcrypt/
95 Wietse Venema's network logger, also known as TCPD or LOG_TCP. These
96 programs log the client host name of incoming telnet, ftp, rsh,
97 rlogin, finger etc. requests. Security options are: access control per
98 host, domain and/or service; detection of host name spoofing or host
99 address spoofing; booby traps to implement an early-warning system.
100 TCP Wrappers can be gotten at ftp://ftp.porcupine.org/pub/security/
101 Netatalk uses TCP Wrappers to authorize host access when using
102 afpovertcp. It should be noted that if DDP is in use, the connection
103 will still be allowed as TCP Wrappers do not impact DDP connections.
105 7. PAM (Pluggable Authentication Modules)
106 PAM provides a flexible mechanism for authenticating
107 users. PAM was invented by SUN Microsystems.
109 Author: Andrew Morgan <morgan@linux.kernel.org>
111 Linux-PAM is a suite of shared libraries that enable the local system
112 administrator to choose how applications authenticate users.
113 You can get the Linux PAM documentation and sources from
114 http://www.kernel.org/pub/linux/libs/pam/
115 Netatalk also supports other standard PAM implementations such as OpenPAM.
118 Berkeley DB is a programmatic toolkit that provides fast, reliable,
119 scalable, and mission-critical database support to software
120 developers. BDB can downloaded from
121 http://www.oracle.com/database/berkeley-db/index.html
122 Netatalk's CNID database uses the library and header files from BDB.
123 Currently, Netatalk supports BDB 4.6 and later.
125 Error checking and logging
126 ==========================
127 We wan't rigid error checking and concise log messages. This often leads
128 to signifant code bloat where the relevant function call is buried in error
129 checking and logging statements.
130 In order to alleviate error checking and code readability, we provide a set
131 of error checking macros in <atalk/errchk.h>. These macros compare the return
132 value of statements againt 0, NULL, -1 (and maybe more, check it out).
133 Every macro comes in four flavours: EC_CHECK, EC_CHECK_LOG, EC_CHECK_LOG_ERR
135 - EC_CHECK just checks the CHECK
136 - EC_CHECK_LOG additionally logs the stringified function call.
137 - EC_CHECK_LOG_ERR allows specifying the return value
138 - EC_CHECK_CUSTOM allows custom actions
139 The macros EC_CHECK* unconditionally jump to a cleanup label where the
140 neccessary cleanup can be done alongside controlling the return value.
141 EC_CHECK_CUSTOM doesn't do that, so an extra "goto EC_CLEANUP" may be
142 performed as appropiate.
145 - stat() without EC macro:
146 static int func(const char *name) {
149 if ((ret = stat(name, &some_struct_stat)) != 0) {
151 ret = -1; /* often needed to explicitly set the error indicating return value */
162 - stat() with EC macro:
163 static int func(const char *name) {
164 EC_INIT; /* expands to int ret = 0; */
166 char *uppername = NULL
167 EC_NULL(uppername = strdup(name));
168 EC_ZERO(strtoupper(uppername));
170 EC_ZERO(stat(uppername, &some_struct_stat)); /* expands to complete if block from above */
175 if (uppername) free(uppername);
179 A boileplate function template is:
196 The ini parser is taken from <http://ndevilla.free.fr/iniparser/>.
197 It has been slightly modified:
199 - "include" directive added
200 - iniparser_getstrdup() to complemnt iniparser_getstring(), it return allocated strings
201 which the caller must free as necessary
202 - the API has been modifed such that all iniparser_get* funcs take a section and a parameter
203 as sepereta args instead of one string of the form "section:parameter" in the original
206 CNID Database Daemons
207 =====================
209 The CNID database daemons cnid_metad and cnid_dbd are a implementation of
210 the netatalk CNID database support that attempts to put all functionality
211 into separate daemons.
212 There is one cnid_dbd daemon per netatalk volume. The underlying database
213 structure is based on Berkeley DB and the database format is the same
214 as in the cdb CNID backend, so this can be used as a drop-in replacement.
218 - No locking issues or leftover locks due to crashed afpd daemons any
219 more. Since there is only one thread of control accessing the
220 database, no locking is needed and changes appear atomic.
222 - Berkeley DB transactions are difficult to get right with several
223 processes attempting to access the CNID database simultanously. This
224 is much easier with a single process and the database can be made nearly
225 crashproof this way (at a performance cost).
227 - No problems with user permissions and access to underlying database
228 files, the cnid_dbd process runs under a configurable user
229 ID that normally also owns the underlying database
230 and can be contacted by whatever afpd daemon accesses a volume.
232 - If an afpd process crashes, the CNID database is unaffected. If the
233 process was making changes to the database at the time of the crash,
234 those changes will be rolled back entirely (transactions).
235 If the process was not using the database at the time of the crash,
236 no corrective action is necessary. In any case, database consistency
241 - Performance in an environment of processes sharing the database
242 (files) is potentially better for two reasons:
245 ii) r/o access to database pages is possible by more than one
246 process at once, r/w access is possible for nonoverlapping regions.
248 The current implementation of cnid_dbd uses unix domain sockets as
249 the IPC mechanism. While this is not the fastest possible method, it
250 is very portable and the cnid_dbd IPC mechanisms can be extended to
251 use faster IPC (like mmap) on architectures where it is
252 supported. As a ballpark figure, 20000 requests/replies to the cnid_dbd
253 daemon take about 0.6 seconds on a Pentium III 733 Mhz running Linux
254 Kernel 2.4.18 using unix domain sockets. The requests are "empty"
255 (no database lookups/changes), so this is just the IPC
258 I have not measured the effects of the advantages of simultanous
262 Installation and configuration
264 There are two executeables that will be built in etc/cnid_dbd and
265 installed into the systems binaries directories of netatalk
266 cnid_metad and cnid_dbd. cnid_metad should run all the
267 time with root permissions. It will be notified when an instance of
268 afpd starts up and will in turn make sure that a cnid_dbd daemon is
269 started for the volume that afpd wishes to access. The cnid_dbd daemon runs as
270 long as necessary and services any
271 other instances of afpd that access the volume. You can safely kill it
272 with SIGTERM, it will be restarted automatically by cnid_metad as soon
273 as the volume is accessed again.
275 cnid_dbd changes to the Berkeley DB directory on startup and sets
276 effective UID and GID to owner and group of that directory. Database and
277 supporting files should therefore be writeable by that user/group.
279 Current shortcomings:
281 - The parameter file parsing of db_param is very simpleminded. It is
282 easy to cause buffer overruns and the like.
283 Also, there is no support for blanks (or weird characters) in
284 filenames for the usock_file parameter.
286 - There is no protection against a malicious user connecting to the
287 cnid_dbd socket and changing the database.