ldd

- list dynamic dependencies of executable files or shared objects

Synopsis

ldd [-d | -r] [-c] [-D]  [-e envar] [-f] [-i] [-L] [-l] [-p]
    [-s] [-U | -u] [-v] [-w] filename...

Description

The ldd utility lists the dynamic dependencies of executable files or shared objects. ldd uses the runtime linker, ld.so.1, to generate the diagnostics. The runtime linker takes the object being inspected and prepares the object as would occur in a running process. By default, ldd triggers the loading of any lazy dependencies, and deferred dependencies.

ldd lists the path names of all shared objects that would be loaded when filename is loaded. ldd expects the shared objects that are being inspected to have execute permission. If a shared object does not have execute permission, ldd issues a warning before attempting to process the file.

ldd processes its input one file at a time. For each file, ldd performs one of the following:

• Lists the object dependencies if the dependencies exist.

• Succeeds quietly if dependencies do not exist.

• Prints an error message if processing fails.

The dynamic objects that are inspected by ldd are not executed. Therefore, ldd does not list any shared objects explicitly attached using dlopen(3C). To display all the objects in use by a process, or a core file, use pldd(1).

Options

ldd can also check the compatibility of filename with the shared objects filename uses. With the following options, ldd prints warnings for any unresolved symbol references that would occur when filename is loaded.

-d

Check immediate references.

-r

Check both immediate references and lazy references.

Only one of the options -d or -r can be specified during any single invocation of ldd.

immediate references are typically to data items used by the executable or shared object code. immediate references are also pointers to functions, and even calls to functions made from a position dependent shared object. lazy references are typically calls to global functions made from a position independent shared object, or calls to external functions made from an executable. For more information on these types of reference, see When Relocations Are Performed in the Linker and Libraries Guide. Object loading can also be affected by relocation processing. See Lazy Loading under USAGE for more details.

Some unresolved symbol references are not reported by default. These unresolved references can be reported with the following options. These options are only useful when combined with either the -d or the -r options.

-p

Expose any unresolved symbol errors to explicit parent and external references.

-w

Expose any unresolved weak symbol references.

A shared object can make reference to symbols that should be supplied by the caller of the shared object. These references can be explicitly classified when the shared object is created, as being available from a parent, or simply as being external. See the -M mapfile option of ld(1), and the PARENT and EXTERN symbol definition keywords. When examining a dynamic executable, a parent or external reference that can not be resolved is flagged as an error. However by default, when examining a shared object, a parent or external reference that can not be resolved is not flagged as an error. The -p option, when used with either the -d or -r options, causes any unresolved parent or external reference to be flagged as a relocation error.

Symbols that are used by relocations may be defined as weak references. By default, if a weak symbol reference can not be resolved, the relocation is ignored and a zero written to the relocation offset. The -w option, when used with either the -d or the -r options, causes any unresolved relocation against a weak symbol reference to be flagged as a relocation error.

ldd can also check dependency use. With each of the following options, ldd prints warnings for any unreferenced, or unused dependencies that are loaded when filename is loaded. Only when a symbol reference is bound to a dependency, is that dependency deemed used. These options are therefore only useful when symbol references are being checked. If the -r option is not in effect, the -d option is enabled.

A dependency that is defined by an object but is not bound to from that object is an unreferenced dependency. A dependency that is not bound to by any other object when filename is loaded is an unused object.

Dependencies can be located in default system locations, or in locations that must be specified by search paths. Search paths may be specified globally, such as the environment variable LD_LIBRARY_PATH. Search paths can also be defined in dynamic objects as runpaths. See the -R option to ld(1). Search paths that are not used to satisfy any dependencies cause unnecessary file system processing.

-U

Displays any unreferenced, or unused dependencies. If an unreferenced dependency is not bound to by other objects loaded with filename, the dependency is also flagged as unused. Cyclic dependencies that are not bound to from objects outside of the cycle are also deemed unreferenced.

This option also displays any unused search paths.

-u

Displays any unused objects.

Only one of the options -U or -u can be specified during any single invocation of ldd, although -U is a superset of -u. Objects that are found to be unreferenced, or unused when using the -r option, should be removed as dependencies. These objects provide no references, but result in unnecessary overhead when filename is loaded. When using the -d option, any objects that are found to be unreferenced, or unused are not immediately required when filename is loaded. These objects are candidates for lazy loading. See Lazy Loading under USAGE for more details.

The removal of unused dependencies reduces runtime-linking overhead. The removal of unreferenced dependencies reduces runtime-linking overhead to a lesser degree. However, the removal of unreferenced dependencies guards against a dependency being unused when combined with different objects, or as the other object dependencies evolve.

The removal of unused search paths can reduce the work required to locate dependencies. This can be significant when accessing files from a file server over a network. Note, a search path can be encoded within an object to satisfy the requirements of dlopen(3C). This search path might not be required to obtain the dependencies of this object, and hence will look unused to ldd.

The following additional options are supported:

-c

Disables any configuration file use. Configuration files can be employed to alter default search paths, and provide alternative object dependencies. See crle(1).

-D

Skip deferred dependency loading. By default, ldd forces the processing of both lazy dependencies and deferred dependencies. See also the -L option. During normal process execution, deferred dependencies are only loaded when the first runtime binding to a deferred reference is made. When using the -D option, the use of the -d or -r options do not trigger the loading of any deferred dependencies. See the -z deferred option of ld(1).

-e envar

Sets the environment variable envar.

This option is useful for experimenting with environment variables that are recognized by the runtime linker that can adversely affect ldd, for example, LD_PRELOAD.

This option is also useful for extracting additional information solely from the object under inspection, for example, LD_DEBUG. See ld.so.1(1) and lari(1).

-f

Forces ldd to check for an executable file that is not secure. When ldd is invoked by a superuser, by default ldd does not process any executable that is not secure. An executable is not considered secure if the interpreter that the executable specifies does not reside under /lib or /usr/lib. An executable is also not considered secure if the interpreter cannot be determined. See Security under USAGE.

-i

Displays the order of execution of initialization sections. The order that is discovered can be affected by use of the -d or -r options. See Initialization Order under USAGE.

-L

Enables lazy loading. By default, ldd forces the processing of both lazy dependencies and deferred dependencies. See also the -D option. During normal process execution, lazy loading is the default mode of operation. In this case, any lazy dependencies, or filters, are only loaded into the process when reference is made to a symbol that is defined within the lazy object. The -d or -r options, together with the -L option, can be used to inspect the dependencies, and their order of loading as would occur in a running process. See the -z lazyload option of ld(1).

-l

Forces the immediate processing of any filters so that all filtees, and their dependencies, are listed. The immediate processing of filters is now the default mode of operation for ldd. However, under this default any auxiliary filtees that cannot be found are silently ignored. Under the -l option, missing auxiliary filtees generate an error message.

-s

Displays the search path used to locate shared object dependencies.

-v

Displays all dependency relationships incurred when processing filename. This option also displays any dependency version requirements. See pvs(1).

Usage

Security

A superuser should use the -f option only if the executable to be examined is known to be trustworthy. The use of -f on an untrustworthy executable while superuser can compromise system security. If an executables trustworthyness is unknown, a superuser should temporarily become a regular user. Then invoke ldd as this regular user.

Untrustworthy objects can be safely examined with dump(1), elfdump(1), elfedit(1), and with mdb(1), as long as the :r subcommand is not used. In addition, a non-superuser can use either the :r subcommand of mdb, or truss(1) to examine an untrustworthy executable without too much risk of compromise. To minimize risk when using ldd, mdb :r, or truss on an untrustworthy executable, use the UID "nobody".

Lazy Loading

Lazy loading can be applied directly by specified lazy dependencies. See the -z lazyload option of ld(1). Lazy loading can also be applied indirectly through filters. See the -f option and -F option of ld(1). Objects that employ lazy loading techniques can experience variations in ldd output due to the options used. If an object expresses all its dependencies as lazy, the default operation of ldd lists all dependencies in the order in which the dependencies are recorded in that object:

example% ldd main 
        libelf.so.1 =>   /lib/libelf.so.1
        libnsl.so.1 =>   /lib/libnsl.so.1
        libc.so.1 =>     /lib/libc.so.1

The lazy loading behavior that occurs when this object is used at runtime can be enabled using the -L option. In this mode, lazy dependencies are loaded when reference is made to a symbol that is defined within the lazy object. Therefore, combining the -L option with use of the -d and -r options reveals the dependencies that are needed to satisfy the immediate, and lazy references respectively:

example% ldd -L main 
example% ldd -d main 
        libc.so.1 =>    /lib/libc.so.1 
example% ldd -r main 
        libc.so.1 =>    /lib/libc.so.1
        libelf.so.1 =>  /lib/libelf.so.1

Notice that in this example, the order of the dependencies that are listed is not the same as displayed from ldd with no options. Even with the -r option, the lazy reference to dependencies might not occur in the same order as would occur in a running program.

Observing lazy loading can also reveal objects that are not required to satisfy any references. These objects, in this example, libnsl.so.1, are candidates for removal from the link-line used to build the object being inspected.

Initialization Order

Objects that do not explicitly define their required dependencies might observe variations in the initialization section order displayed by ldd due to the options used. For example, a simple application might reveal:

example% ldd -i main 
        libA.so.1 =>   ./libA.so.1
        libc.so.1 =>   /lib/libc.so.1
        libB.so.1 =>   ./libB.so.1

    init object=./libB.so.1
    init object=./libA.so.1
    init object=/lib/libc.so.1

whereas, when relocations are applied, the initialization section order is:

example% ldd -ir main 
        .........

    init object=/lib/libc.so.1
    init object=./libB.so.1
    init object=./libA.so.1

In this case, libB.so.1 makes reference to a function in /usr/lib/libc.so.1. However, libB.so.1 has no explicit dependency on this library. Only after a relocation is discovered is a dependency then established. This implicit dependency affects the initialization section order.

Typically, the initialization section order established when an application is executed, is equivalent to ldd with the -d option. The optimum order can be obtained if all objects fully define their dependencies. Use of the ld(1) options -z defs and -z ignore when building dynamic objects is recommended.

Cyclic dependencies can result when one or more dynamic objects reference each other. Cyclic dependencies should be avoided, as a unique initialization sort order for these dependencies can not be established.

Users that prefer a more static analysis of object files can inspect dependencies using tools such as dump(1) and elfdump(1).

Files

/usr/lib/lddstub

Fake 32–bit executable loaded to check the dependencies of shared objects.

/usr/lib/64/lddstub

Fake 64–bit executable loaded to check the dependencies of shared objects.

Attributes

See attributes(5) for descriptions of the following attributes:

See Also

crle(1), dump(1), elfdump(1), elfedit(1), lari(1), ld(1), ld.so.1(1), mdb(1), pldd(1), pvs(1), truss(1), dlopen(3C), attributes(5)

Linker and Libraries Guide

Diagnostics

ldd prints the record of shared object path names to stdout. The optional list of symbol resolution problems is printed to stderr. If filename is not an executable file or a shared object, or if filename cannot be opened for reading, a non-zero exit status is returned.

Notes

Use of the -d or -r option with shared objects can give misleading results. ldd does a worst case analysis of the shared objects. However, in practice, the symbols reported as unresolved might be resolved by the executable file referencing the shared object. The runtime linkers preloading mechanism can be employed to add dependencies to the object being inspected. See LD_PRELOAD.

ldd uses the same algorithm as the runtime linker to locate shared objects.