1 <?xml version="1.0" encoding="UTF-8" ?>
2 <!DOCTYPE manualpage SYSTEM "./style/manualpage.dtd">
3 <?xml-stylesheet type="text/xsl" href="./style/manual.en.xsl"?>
4 <!-- $LastChangedRevision$ -->
7 Licensed to the Apache Software Foundation (ASF) under one or more
8 contributor license agreements. See the NOTICE file distributed with
9 this work for additional information regarding copyright ownership.
10 The ASF licenses this file to You under the Apache License, Version 2.0
11 (the "License"); you may not use this file except in compliance with
12 the License. You may obtain a copy of the License at
14 http://www.apache.org/licenses/LICENSE-2.0
16 Unless required by applicable law or agreed to in writing, software
17 distributed under the License is distributed on an "AS IS" BASIS,
18 WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
19 See the License for the specific language governing permissions and
20 limitations under the License.
23 <manualpage metafile="dso.xml.meta">
25 <title>Dynamic Shared Object (DSO) Support</title>
28 <p>The Apache HTTP Server is a modular program where the
29 administrator can choose the functionality to include in the
30 server by selecting a set of modules. The modules can be
31 statically compiled into the <program>httpd</program> binary when the
32 server is built. Alternatively, modules can be compiled as
33 Dynamic Shared Objects (DSOs) that exist separately from the
34 main <program>httpd</program> binary file. DSO modules may be
35 compiled at the time the server is built, or they may be
36 compiled and added at a later time using the Apache Extension
37 Tool (<program>apxs</program>).</p>
39 <p>This document describes how to use DSO modules as well as
40 the theory behind their use.</p>
44 <section id="implementation"><title>Implementation</title>
48 <module>mod_so</module>
51 <directive module="mod_so">LoadModule</directive>
55 <p>The DSO support for loading individual Apache modules is based
56 on a module named <module>mod_so</module> which must be statically
57 compiled into the Apache core. It is the only module besides
58 <module>core</module> which cannot be put into a DSO
59 itself. Practically all other distributed Apache modules can then
60 be placed into a DSO by individually enabling the DSO build for
61 them via <program>configure</program>'s
62 <code>--enable-<em>module</em>=shared</code> option as discussed
63 in the <a href="install.html">install documentation</a>. After a
64 module is compiled into a DSO named <code>mod_foo.so</code> you
65 can use <module>mod_so</module>'s <directive
66 module="mod_so">LoadModule</directive> command in your
67 <code>httpd.conf</code> file to load this module at server startup
70 <p>To simplify this creation of DSO files for Apache modules
71 (especially for third-party modules) a new support program
72 named <program>apxs</program> (<dfn>APache
73 eXtenSion</dfn>) is available. It can be used to build DSO based
74 modules <em>outside of</em> the Apache source tree. The idea is
75 simple: When installing Apache the <program>configure</program>'s
76 <code>make install</code> procedure installs the Apache C
77 header files and puts the platform-dependent compiler and
78 linker flags for building DSO files into the <program>apxs</program>
79 program. This way the user can use <program>apxs</program> to compile
80 his Apache module sources without the Apache distribution
81 source tree and without having to fiddle with the
82 platform-dependent compiler and linker flags for DSO
86 <section id="usage"><title>Usage Summary</title>
88 <p>To give you an overview of the DSO features of Apache 2.0,
89 here is a short and concise summary:</p>
93 Build and install a <em>distributed</em> Apache module, say
94 <code>mod_foo.c</code>, into its own DSO
95 <code>mod_foo.so</code>:
98 $ ./configure --prefix=/path/to/install --enable-foo=shared<br />
104 Build and install a <em>third-party</em> Apache module, say
105 <code>mod_foo.c</code>, into its own DSO
106 <code>mod_foo.so</code>:
109 $ ./configure --add-module=<var>module_type</var>:/path/to/3rdparty/mod_foo.c \<br />
111 --enable-foo=shared<br />
118 Configure Apache for <em>later installation</em> of shared
122 $ ./configure --enable-so<br />
128 Build and install a <em>third-party</em> Apache module, say
129 <code>mod_foo.c</code>, into its own DSO
130 <code>mod_foo.so</code> <em>outside of</em> the Apache
131 source tree using <program>apxs</program>:
134 $ cd /path/to/3rdparty<br />
135 $ apxs -c mod_foo.c<br />
136 $ apxs -i -a -n foo mod_foo.la
141 <p>In all cases, once the shared module is compiled, you must
142 use a <directive module="mod_so">LoadModule</directive>
143 directive in <code>httpd.conf</code> to tell Apache to activate
147 <section id="background"><title>Background</title>
149 <p>On modern Unix derivatives there exists a nifty mechanism
150 usually called dynamic linking/loading of <em>Dynamic Shared
151 Objects</em> (DSO) which provides a way to build a piece of
152 program code in a special format for loading it at run-time
153 into the address space of an executable program.</p>
155 <p>This loading can usually be done in two ways: Automatically
156 by a system program called <code>ld.so</code> when an
157 executable program is started or manually from within the
158 executing program via a programmatic system interface to the
159 Unix loader through the system calls
160 <code>dlopen()/dlsym()</code>.</p>
162 <p>In the first way the DSO's are usually called <em>shared
163 libraries</em> or <em>DSO libraries</em> and named
164 <code>libfoo.so</code> or <code>libfoo.so.1.2</code>. They
165 reside in a system directory (usually <code>/usr/lib</code>)
166 and the link to the executable program is established at
167 build-time by specifying <code>-lfoo</code> to the linker
168 command. This hard-codes library references into the executable
169 program file so that at start-time the Unix loader is able to
170 locate <code>libfoo.so</code> in <code>/usr/lib</code>, in
171 paths hard-coded via linker-options like <code>-R</code> or in
172 paths configured via the environment variable
173 <code>LD_LIBRARY_PATH</code>. It then resolves any (yet
174 unresolved) symbols in the executable program which are
175 available in the DSO.</p>
177 <p>Symbols in the executable program are usually not referenced
178 by the DSO (because it's a reusable library of general code)
179 and hence no further resolving has to be done. The executable
180 program has no need to do anything on its own to use the
181 symbols from the DSO because the complete resolving is done by
182 the Unix loader. (In fact, the code to invoke
183 <code>ld.so</code> is part of the run-time startup code which
184 is linked into every executable program which has been bound
185 non-static). The advantage of dynamic loading of common library
186 code is obvious: the library code needs to be stored only once,
187 in a system library like <code>libc.so</code>, saving disk
188 space for every program.</p>
190 <p>In the second way the DSO's are usually called <em>shared
191 objects</em> or <em>DSO files</em> and can be named with an
192 arbitrary extension (although the canonical name is
193 <code>foo.so</code>). These files usually stay inside a
194 program-specific directory and there is no automatically
195 established link to the executable program where they are used.
196 Instead the executable program manually loads the DSO at
197 run-time into its address space via <code>dlopen()</code>. At
198 this time no resolving of symbols from the DSO for the
199 executable program is done. But instead the Unix loader
200 automatically resolves any (yet unresolved) symbols in the DSO
201 from the set of symbols exported by the executable program and
202 its already loaded DSO libraries (especially all symbols from
203 the ubiquitous <code>libc.so</code>). This way the DSO gets
204 knowledge of the executable program's symbol set as if it had
205 been statically linked with it in the first place.</p>
207 <p>Finally, to take advantage of the DSO's API the executable
208 program has to resolve particular symbols from the DSO via
209 <code>dlsym()</code> for later use inside dispatch tables
210 <em>etc.</em> In other words: The executable program has to
211 manually resolve every symbol it needs to be able to use it.
212 The advantage of such a mechanism is that optional program
213 parts need not be loaded (and thus do not spend memory) until
214 they are needed by the program in question. When required,
215 these program parts can be loaded dynamically to extend the
216 base program's functionality.</p>
218 <p>Although this DSO mechanism sounds straightforward there is
219 at least one difficult step here: The resolving of symbols from
220 the executable program for the DSO when using a DSO to extend a
221 program (the second way). Why? Because "reverse resolving" DSO
222 symbols from the executable program's symbol set is against the
223 library design (where the library has no knowledge about the
224 programs it is used by) and is neither available under all
225 platforms nor standardized. In practice the executable
226 program's global symbols are often not re-exported and thus not
227 available for use in a DSO. Finding a way to force the linker
228 to export all global symbols is the main problem one has to
229 solve when using DSO for extending a program at run-time.</p>
231 <p>The shared library approach is the typical one, because it
232 is what the DSO mechanism was designed for, hence it is used
233 for nearly all types of libraries the operating system
234 provides. On the other hand using shared objects for extending
235 a program is not used by a lot of programs.</p>
237 <p>As of 1998 there are only a few software packages available
238 which use the DSO mechanism to actually extend their
239 functionality at run-time: Perl 5 (via its XS mechanism and the
240 DynaLoader module), Netscape Server, <em>etc.</em> Starting
241 with version 1.3, Apache joined the crew, because Apache
242 already uses a module concept to extend its functionality and
243 internally uses a dispatch-list-based approach to link external
244 modules into the Apache core functionality. So, Apache is
245 really predestined for using DSO to load its modules at
249 <section id="advantages"><title>Advantages and Disadvantages</title>
251 <p>The above DSO based features have the following
255 <li>The server package is more flexible at run-time because
256 the actual server process can be assembled at run-time via
257 <directive module="mod_so">LoadModule</directive>
258 <code>httpd.conf</code> configuration commands instead of
259 <program>configure</program> options at build-time. For instance
260 this way one is able to run different server instances
261 (standard & SSL version, minimalistic & powered up
262 version [mod_perl, PHP3], <em>etc.</em>) with only one Apache
265 <li>The server package can be easily extended with
266 third-party modules even after installation. This is at least
267 a great benefit for vendor package maintainers who can create
268 a Apache core package and additional packages containing
269 extensions like PHP3, mod_perl, mod_fastcgi,
272 <li>Easier Apache module prototyping because with the
273 DSO/<program>apxs</program> pair you can both work outside the
274 Apache source tree and only need an <code>apxs -i</code>
275 command followed by an <code>apachectl restart</code> to
276 bring a new version of your currently developed module into
277 the running Apache server.</li>
280 <p>DSO has the following disadvantages:</p>
283 <li>The DSO mechanism cannot be used on every platform
284 because not all operating systems support dynamic loading of
285 code into the address space of a program.</li>
287 <li>The server is approximately 20% slower at startup time
288 because of the symbol resolving overhead the Unix loader now
291 <li>The server is approximately 5% slower at execution time
292 under some platforms because position independent code (PIC)
293 sometimes needs complicated assembler tricks for relative
294 addressing which are not necessarily as fast as absolute
297 <li>Because DSO modules cannot be linked against other
298 DSO-based libraries (<code>ld -lfoo</code>) on all platforms
299 (for instance a.out-based platforms usually don't provide
300 this functionality while ELF-based platforms do) you cannot
301 use the DSO mechanism for all types of modules. Or in other
302 words, modules compiled as DSO files are restricted to only
303 use symbols from the Apache core, from the C library
304 (<code>libc</code>) and all other dynamic or static libraries
305 used by the Apache core, or from static library archives
306 (<code>libfoo.a</code>) containing position independent code.
307 The only chances to use other code is to either make sure the
308 Apache core itself already contains a reference to it or
309 loading the code yourself via <code>dlopen()</code>.</li>