load and loadAndInit Subroutines
Loads a module into the current process.
int *load ( ModuleName, Flags, LibraryPath) char *ModuleName; uint Flags; char *LibraryPath;
int *loadAndInit ( ModuleName, Flags, LibraryPath) char *ModuleName; uint Flags; char *LibraryPath;
The load and loadAndInit subroutines load the specified module into the calling process's address space. A module can be a regular file or a member of an archive. When adding a new module to the address space of a 32-bit process, the load operation may cause the break value to change.
The load subroutine is not a preferred method to load C++ modules. Use loadAndInit subroutine instead. The loadAndInit subroutine uses the same interface as load but performs C++ initialization.
The exec subroutine is similar to the load subroutine, except that:
- The load subroutine does not replace the current program with a new one.
- The exec subroutine does not have an explicit library path parameter; it has only the LIBPATH and LD_LIBRARY_PATH environment variables. Also, these library path environment variables are ignored when the program using the exec subroutine has more privilege than the caller (for example, in the case of a set-UID program).
A large application can be split up into one or more modules in one of two ways that allow execution within the same process. The first way is to create each of the application's modules separately and use load to explicitly load a module when it is needed. The other way is to specify the relationship between the modules when they are created by defining imported and exported symbols.
Modules can import symbols from other modules. Whenever symbols are imported from one or more other modules, these modules are automatically loaded to resolve the symbol references if the required modules are not already loaded, and if the imported symbols are not specified as deferred imports. These modules can be archive members in libraries or individual files and can have either shared or private file characteristics that control how and where they are loaded.
Shared modules (typically members of a shared library archive) are loaded into the shared library region, when their access permissions allow sharing, that is, when they have read-other permission. Private modules, and shared modules without the required permissions for sharing, are loaded into the process private region.
When the loader resolves a symbol, it uses the file name recorded with that symbol to find the module that exports the symbol. If the file name contains any / (slash) characters, it is used directly and must name an appropriate file or archive member. However, if the file name is a base name (contains no / characters), the loader searches the directories specified in the default library path for a file (i.e. a module or an archive) with that base name.
The LibraryPath is a string containing one or more directory path names separated by colons. See the section Searching for Dependent Modules for information on library path searching.
When a process is executing under ptrace control, portions of the process's address space are recopied after the load processing completes. For a 32-bit process, the main program text (loaded in segment 1) and shared library modules (loaded in segment 13) are recopied. Any breakpoints or other modifications to these segments must be reinserted after the load call. For a 64-bit process, shared library modules are recopied after a load call. The debugger will be notified by setting the W_SLWTED flag in the status returned by wait, so that it can reinsert breakpoints.
When a process executing under ptrace control calls load, the debugger is notified by setting the W_SLWTED flag in the status returned by wait. Any modules newly loaded into the shared library segments will be copied to the process's private copy of these segments, so that they can be examined or modified by the debugger.
The load subroutine will call initialization routines (init routines) for the new module and any of its dependents if they were not already loaded.
Modules loaded by this subroutine are automatically unloaded when the process terminates or when the exec subroutine is executed. They are explicitly unloaded by calling the unload subroutine.
Searching for Dependent Modules
The load operation and the exec operation differ slightly in their dependent module search mechanism. When a module is added to the address space of a running process (the load operation), the rules outlined in the next section are used to find the named module. Note that dependency relationships may be loosely defined as a tree but recursive relationships between modules may also exist. The following components may used to create a complete library search path:
- If the L_LIBPATH_EXEC flag is set, the library search path used at exec-time.
- The value of the LibraryPath parameter if it is non-null. Note that a null string is a valid search path which refers to the current working directory. If the LibraryPath parameter is NULL, the value of the LIBPATH environment variable, or alternatively the LD_LIBRARY_PATH environment variable (if LIBPATH is not set), is used instead.
- The library search path contained in the loader section of the module being loaded (the ModuleName parameter).
- The library search path contained in the loader section of the module whose immediate dependents are being loaded. Note that this per-module information changes when searching for each module's immediate dependents.
To find the ModuleName module, components 1 and 2 are used. To find dependents, components 1, 2, 3 and 4 are used in order. Note that if any modules that are already part of the running process satisfy the dependency requirements of the newly loaded module(s), pre-existing modules are not loaded again.
For each colon-separated portion of the aggregate search specification, if the base name is not found the search continues. Additionally, if the needed file is not an archive member, the search will continue past a file having the wrong object mode. If an archive member is needed, searching stops when the first match of the file name is found. If the file is not of the proper form, or in the case of an archive that does not contain the required archive member, or does not export a definition of a required symbol, an error occurs. The library path search is not performed when either a relative or an absolute path name is specified for a dependent module.
The library search path stored within the module is specified at link-edit time.
The load subroutine may cause the calling process to fail if the module specified has a very long chain of dependencies, (for example, lib1.a, which depends on lib2.a, which depends on lib3.a, etc). This is because the loader processes such relationships recursively on a fixed-size stack. This limitation is exposed only when processing a dependency chain that has over one thousand elements.
|ModuleName||Points to the name of the module to be loaded. The module name
consists of a path name, and, an optional member name. If the path
name contains at least on / character, the name is used directly,
and no directory searches are performed to locate the file. If the
path name contains no / characters, it is treated as a base name,
and should be in one of the directories listed in the library path.
The library path is either the value of the LibraryPath parameter if not a null value, or the value of the LIBPATH environment variable (if set; otherwise, LD_LIBRARY_PATH environment variable, if set) or the library path used at process exec time (if the L_LIBPATH_EXEC is set). If no library path is provided, the module should be in the current directory.
The ModuleName parameter may explicitly name an archive member. The syntax is pathname(member) where pathname follows the rules specified in the previous paragraph, and member is the name of a specific archive member. The parentheses are a required portion of the specification and no intervening spaces are allowed. If an archive member is named, the L_LOADMEMBER flag must be added to the Flags parameter. Otherwise, the entire ModuleName parameter is treated as an explicit filename.
|Flags||Modifies the behavior of the load and the loadAndInit services
as follows (see the ldr.h file). If no special behavior is
required, set the value of the flags parameter to 0 (zero). For compatibility,
a value of 1 (one) may also be specified.
|LibraryPath|| Points to a character string that specifies the default library
If the LibraryPath parameter is NULL, the LIBPATH environment variable is used, if set; otherwise, the LD_LIBRARY_PATH environment variable is used.
The library path is used to locate dependent modules that are specified as basenames (that is, their pathname components do not contain a / (slash) character.
Note the difference between setting the LibraryPath parameter to null, and having the LibraryPath parameter point to a null string (" "). A null string is a valid library path which consists of a single directory: the current directory.
Upon successful completion, the load and loadAndInit subroutines return the pointer to function for the entry point of the module. If the module has no entry point, the address of the data section of the module is returned.
If the load and loadAndInit subroutines fail, a null pointer is returned, the module is not loaded, and errno global variable is set to indicate the error. The load and loadAndInit subroutines fail if one or more of the following are true of a module to be explicitly or automatically loaded:
|EACCES||Indicates the file is not an ordinary file, or the mode of the program file denies execution permission, or search permission is denied on a component of the path prefix.|
|EINVAL||Indicates the file or archive member has a valid magic number in its header, but the header is damaged or is incorrect for the machine on which the file is to be run.|
|ELOOP||Indicates too many symbolic links were encountered in translating the path name.|
|ENOEXEC||Indicates an error occurred when loading or resolving symbols for the specified module. This can be due to an attempt to load a module with an invalid XCOFF header, a failure to resolve symbols that were not defined as deferred imports or several other load time related problems. The loadquery subroutine can be used to return more information about the load failure. If runtime linking is used, the load and the loadAndInit subroutines will fail if the runtime linker could not resolve some symbols. In this case, errno will be set to ENOEXEC, but the loadquery subroutine will not return any additional information.|
|ENOMEM||Indicates the program requires more memory than is allowed by the system-imposed maximum.|
|ETXTBSY||Indicates the file is currently open for writing by some process.|
|ENAMETOOLONG||Indicates a component of a path name exceeded 255 characters, or an entire path name exceeded 1023 characters.|
|ENOENT||Indicates a component of the path prefix does not exist, or the path name is a null value. For the dlopen subroutine, RTLD_MEMBER is not used when trying to open a member within the archive file.|
|ENOTDIR||Indicates a component of the path prefix is not a directory.|
|ESTALE||Indicates the process root or current directory is located in a virtual file system that has been unmounted.|