Return-Path: Delivered-To: apache-cvs-archive@hyperreal.org Received: (qmail 10247 invoked by uid 6000); 13 Apr 1998 15:40:21 -0000 Received: (qmail 10239 invoked by alias); 13 Apr 1998 15:40:19 -0000 Delivered-To: apache-1.3-cvs@hyperreal.org Received: (qmail 10236 invoked by uid 168); 13 Apr 1998 15:40:19 -0000 Date: 13 Apr 1998 15:40:19 -0000 Message-ID: <19980413154019.10235.qmail@hyperreal.org> From: rse@hyperreal.org To: apache-1.3-cvs@hyperreal.org Subject: cvs commit: apache-1.3/src README.DSO Sender: apache-cvs-owner@apache.org Precedence: bulk Reply-To: new-httpd@apache.org rse 98/04/13 08:40:18 Added: src README.DSO Log: Here it comes: My first cut for an overview README.DSO file describing the relevant aspects of dynamic shared object (DSO) support in Apache 1.3. My hope is that this information makes the DSO stuff more clear to those not very familiar with it. And I hope my written down information is correct... ;-) Revision Changes Path 1.1 apache-1.3/src/README.DSO Index: README.DSO =================================================================== Apache 1.3 Dynamic Shared Object (DSO) support ============================================== Ralf S. Engelschall, April 1998 Background ---------- On modern Unix derivates there exists a nifty mechanism usually named Dynamic Shared Object (DSO) which provides a way to build a piece of program code in a special format to be able to load it under run-time into the address space of an executable program. This loading can usually be done in two ways: Automatically by a system program named ld.so when the Unix loader has to start an executable program or manually from within the executing program via a pragmatic system interface to the Unix loader through the system calls dlopen()/dlsym(). In the first way the DSO's are usually called "shared libraries" or "DSO libraries" and named libfoo.so or libfoo.so.1.2. They stay inside a system directory (usually /usr/lib) and the link to the executable program is established under link-time by specifying -lfoo to the linker command. This hardcodes library references into the executable program file therewith under start-time the Unix loader is able to lookup libfoo.so from /usr/lib or from paths configured via the environment variable LD_LIBRARY_PATH. It then resolves any (still unresolved) symbols in the executable program which are defined and exported in the DSO. Symbols in the executable program are usually not used inside the DSO (because its a reuseable library of general code) and hence no resolving this way has to be done. The executable program has no to do anything to be able to use the symbols from the DSO because the complete resolving is done by the Unix loader. In the second way the DSO's are usually called "shared objects" or "DSO files" and can be arbitrarily named (although the canonical name is foo.so). These files usually stay inside a program-specific directory and there is no automatically established link to the executable program where they are used. Instead the executable program under run-time manually loads the DSO into his address space via dlopen(). At this time no resolving of symbols from the DSO for the executable program is done. But instead the Unix loader automatically resolves any (still unresolved) symbols in the DSO which are defined and exported in the executable program. This way the DSO gets knowledge of the executable program as it would have been statically linked to it under program link-time. Finally to make the DSO accessible to the executable program it resolves particular symbols from the DSO via dlsym() for later use inside dispatch tables, etc. In other words: The executable program has no to manually resolve anything to be able to use it. Although this DSO mechanism sounds straight foreward there is at least one difficult step here: The resolving of symbols from the executable program for the DSO when using a DSO to extend a program (the second way). Why? Because this resolving is against the library design (where the library has no knowledge of any program it is used for) and is neither available under all platforms nor standardized. In practice only global symbols from the executable program are available to the DSO which are explicitly marked as exported. And forcing this exportation of global symbols is the main problem one has to solve when using DSO for extending a program under run-time. Practical Usage --------------- The shared library approach is the typical one, because this is the way the DSO mechanism was designed for, hence it is used for mostly all types of libraries the operating system provides. On the other hand using shared objects for extending a program is not used by a lot of programs. As of 1998 there are only a few software package available which use the DSO mechanism to actually extend their functionality under run-time: Perl 5 (via it's XS mechanism and the DynaLoader module), GIMP, Netscape Server, etc. But Apache 1.3 now is also one of these, because Apache already uses a module concept to extend its functionality and really uses a dispatch-list-based approach to link these modules into the Apache core functionality. So, Apache is really predestinated for using DSO to load it's modules under run-time. The idea now is to provide two optional features for Apache 1.3: To compile and place the Apache core program into a DSO library for shared usage and to compile and place Apache modules into DSO files for explicit loading under run-time. Implementation -------------- To place the complete Apache core program into a DSO library the rule SHARED_CORE has to be enabled via APACI's --enable-rule=SHARED_CORE option (see ../INSTALL file) or by changing the Rule command in src/Configuration.tmpl to "Rule SHARED_CORE=yes" (see ./INSTALL file) the Apache core code then is placed into a DSO library named libhttpd.so. Because one cannot link a DSO against static libraries, an additional executable program named libhttpd.ep is created which both ties those static code and provides a stub for the main() function. Finally the httpd executable program itself is replaced by a bootstrapping code which automatically makes sure the Unix loader is able to load and start libhttpd.ep by providing the LD_LIBRARY_PATH to libhttpd.so. The DSO support for loading Apache modules is implemented completely different: Here a module named mod_so.c is used which has to be statically compiled into the Apache core. It is the only module besides http_core.c which cannot be put into a DSO itself (bootstrapping!). Mostly all other distributed Apache modules then can be placed into a DSO by individually enabling the DSO build for them via APACI's --enable-shared option (see ../INSTALL file) or by changing the `AddModule' command in src/Configuration.tmpl into a `SharedModule' command (see ./INSTALL file). After a module is placed into a DSO named mod_foo.so you can use mod_so's `LoadModule' command in your httpd.conf file to load this module at server startup or restart. To simplify this creation of DSO files for Apache modules (especially for third-party ones) a new support program named `apxs' is available. I can be used to build DSO based modules _outside_ the Apache source tree. The idea is simple: When installing Apache the APACI "make install" procedure installs the Apache C header files and puts the platform-dependend compiler and linker flags for building DSO files into the `apxs' program. This way the user can use `apxs' to compile it's Apache module sources without the Apache distribution source tree and without having to fiddle with the platform-dependend compiler and linker flags for DSO support. Usage Summary ------------- To give you an overview of the DSO features of Apache 1.3, here is a short and concrete summary: 1. Placing the Apache core code (all the stuff which usually forms the httpd binary) into a DSO libhttpd.so, an executable program libhttpd.ep and a bootstrapping executable program httpd: o Build and install via APACI (preferred): $ ./configure --prefix=/path/to/install --enable-rule=SHARED_CORE ... $ make install o Build and install manually: - Edit src/Configuration: << "Rule SHARED_CORE=default" >> "Rule SHARED_CORE=yes" << "EXTRA_CFLAGS= " >> "EXTRA_CFLAGS= -DSHARED_CORE_DIR=\"/path/to/install/libexec\" $ make $ cp src/libhttpd.so* /path/to/install/libexec/ $ cp src/libhttpd.ep /path/to/install/libexec/ $ cp src/httpd /path/to/install/bin/ 2. Build and install a distributed Apache module, say mod_foo.c, into its own DSO mod_foo.so: o Build and install via APACI (preferred): $ ./configure --prefix=/path/to/install --enable-shared=foo $ make install o Build and install manually: - Edit src/Configuration: << "AddModule modules/xxxx/mod_foo.o" >> "SharedModule modules/xxxx/mod_foo.so" $ make $ cp src/xxxx/mod_foo.so /path/to/install/libexec - Edit /path/to/install/etc/httpd.conf >> "LoadModule foo_module /path/to/install/libexec/mod_foo.so" 3. Build and install a third-party Apache module, say mod_foo.c, into its own DSO mod_foo.so o Build and install via APACI (preferred): $ ./configure --add-module=/path/to/3rdparty/mod_foo.c --enable-shared=foo $ make install o Build and install manually: $ cp /path/to/3rdparty/mod_foo.c /path/to/apache-1.3/src/modules/extra/ - Edit src/Configuration: >> "SharedModule modules/extra/mod_foo.so" $ make $ cp src/xxxx/mod_foo.so /path/to/install/libexec - Edit /path/to/install/etc/httpd.conf >> "LoadModule foo_module /path/to/install/libexec/mod_foo.so" 4. Build and install a third-party Apache module, say mod_foo.c, into its own DSO mod_foo.so _outside_ the Apache source tree: o Build and install via APXS: $ cd /path/to/3rdparty $ apxs -c mod_foo.c $ apxs -i -a -n foo mod_foo.so Advantages & Disadvantages -------------------------- The above DSO based features of Apache 1.3 have the following advantages (+) and disadvantages (-): + The server package is more flexible under run-time because the actual used server process can be assembled under run-time via LoadModule httpd.conf configuration commands instead of Configuration AddModule commands under build-time. For instance this way one is able to run different server instances (standard & SSL version, minimalistic & powered up version [mod_perl, PHP3], etc.) with only one Apache installation. + The server package can be easily extended with third-party modules even after installation. This is at least a great benefit for vendor package maintainers who can create a Apache core package and additional packages containing extensions like PHP3, mod_perl, mod_fastcgi, etc. + Easier Apache module prototyping because with the DSO/APXS couple you can both works outside the Apache source tree and only need an `apxs -i' command followed by a `apachectl restart' to bring a new version of your currently developed module into the running Apache server. - The DSO mechanism cannot be used on any platform because not all operating systems support this mechanism. - The server is approximately 20% slower at startup time because of the symbol resolving overhead the Unix loader now has to do. - The server is approximately 5% slower at execution time under some platforms because position independed code (PIC) sometimes needs complicated assembler tricks for relative addressing which are not necessarily as fast as absolute addressing. - Because DSO modules cannot be linked against other DSO-based libraries (ld -lfoo) you cannot use the DSO mechanism for all types of modules. Or in other words, modules compiled as DSO files are restricted to only use symbols from the Apache core, from the C library (libc) or from static library archives (libfoo.a) containing position independend code. The only chance to use other code is to either make sure the Apache core itself already contains a reference to it or loading the code yourself via dlopen. - Because under some platforms like SVR4 there is no way to force the linker to export the global symbols when linking the Apache httpd executable program. This way these aren't available to modules built as DSO. The only chance here is to use the SHARED_CORE feature because this way the global symbols are forced to be exported. As a consequence the Apache src/Configure script automatically forced SHARED_CORE under those platforms when DSO should be used. Ralf S. Engelschall rse@engelschall.com www.engelschall.com