1.1 Getting Started

This manual documents the Libgcrypt library application programming interface (API). All functions and data types provided by the library are explained.

The reader is assumed to possess basic knowledge about applied cryptography.

This manual can be used in several ways. If read from the beginning to the end, it gives a good introduction into the library and how it can be used in an application. Forward references are included where necessary. Later on, the manual can be used as a reference manual to get just the information needed about any particular interface of the library. Experienced programmers might want to start looking at the examples at the end of the manual, and then only read up those parts of the interface which are unclear.

1.2 Features

Libgcrypt might have a couple of advantages over other libraries doing a similar job.

It’s Free Software

Anybody can use, modify, and redistribute it under the terms of the GNU Lesser General Public License (see GNU Lesser General Public License). Note, that some parts (which are in general not needed by applications) are subject to the terms of the GNU General Public License (see GNU General Public License); please see the README file of the distribution for of list of these parts.

It encapsulates the low level cryptography

Libgcrypt provides a high level interface to cryptographic building blocks using an extensible and flexible API.

1.3 Overview

The Libgcrypt library is fully thread-safe, where it makes sense to be thread-safe. Not thread-safe are some cryptographic functions that modify a certain context stored in handles. If the user really intents to use such functions from different threads on the same handle, he has to take care of the serialization of such functions himself. If not described otherwise, every function is thread-safe.

Libgcrypt depends on the library ‘libgpg-error’, which contains some common code used by other GnuPG components.

2.1 Header

All interfaces (data types and functions) of the library are defined in the header file gcrypt.h. You must include this in all source files using the library, either directly or through some other header file, like this:

#include <gcrypt.h>

The name space of Libgcrypt is gcry_* for function and type names and GCRY* for other symbols. In addition the same name prefixes with one prepended underscore are reserved for internal use and should never be used by an application. Note that Libgcrypt uses libgpg-error, which uses gpg_* as name space for function and type names and GPG_* for other symbols, including all the error codes.

Certain parts of gcrypt.h may be excluded by defining these macros:

GCRYPT_NO_MPI_MACROS

Do not define the shorthand macros mpi_* for gcry_mpi_*.

GCRYPT_NO_DEPRECATED

Do not include definitions for deprecated features. This is useful to make sure that no deprecated features are used.

2.2 Building sources

If you want to compile a source file including the ‘gcrypt.h’ header file, you must make sure that the compiler can find it in the directory hierarchy. This is accomplished by adding the path to the directory in which the header file is located to the compilers include file search path (via the -I option).

However, the path to the include file is determined at the time the source is configured. To solve this problem, Libgcrypt ships with a small helper program libgcrypt-config that knows the path to the include file and other configuration options. The options that need to be added to the compiler invocation at compile time are output by the --cflags option to libgcrypt-config. The following example shows how it can be used at the command line:

gcc -c foo.c `libgcrypt-config --cflags`

Adding the output of ‘libgcrypt-config --cflags’ to the compiler’s command line will ensure that the compiler can find the Libgcrypt header file.

A similar problem occurs when linking the program with the library. Again, the compiler has to find the library files. For this to work, the path to the library files has to be added to the library search path (via the -L option). For this, the option --libs to libgcrypt-config can be used. For convenience, this option also outputs all other options that are required to link the program with the Libgcrypt libraries (in particular, the ‘-lgcrypt’ option). The example shows how to link foo.o with the Libgcrypt library to a program foo.

gcc -o foo foo.o `libgcrypt-config --libs`

Of course you can also combine both examples to a single command by specifying both options to libgcrypt-config:

gcc -o foo foo.c `libgcrypt-config --cflags --libs`

2.3 Building sources using Automake

It is much easier if you use GNU Automake instead of writing your own Makefiles. If you do that, you do not have to worry about finding and invoking the libgcrypt-config script at all. Libgcrypt provides an extension to Automake that does all the work for you.

Macro: AM_PATH_LIBGCRYPT ([minimum-version], [action-if-found], [action-if-not-found]) ¶

Check whether Libgcrypt (at least version minimum-version, if given) exists on the host system. If it is found, execute action-if-found, otherwise do action-if-not-found, if given.

Additionally, the function defines LIBGCRYPT_CFLAGS to the flags needed for compilation of the program to find the gcrypt.h header file, and LIBGCRYPT_LIBS to the linker flags needed to link the program to the Libgcrypt library. If the used helper script does not match the target type you are building for a warning is printed and the string libgcrypt is appended to the variable gpg_config_script_warn.

This macro searches for libgcrypt-config along the PATH. If you are cross-compiling, it is useful to set the environment variable SYSROOT to the top directory of your target. The macro will then first look for the helper program in the bin directory below that top directory. An absolute directory name must be used for SYSROOT. Finally, if the configure command line option --with-libgcrypt-prefix is used, only its value is used for the top directory below which the helper script is expected.

You can use the defined Autoconf variables like this in your Makefile.am:

AM_CPPFLAGS = $(LIBGCRYPT_CFLAGS)
LDADD = $(LIBGCRYPT_LIBS)

2.4 Initializing the library

Before the library can be used, it must initialize itself. This is achieved by invoking the function gcry_check_version described below.

Also, it is often desirable to check that the version of Libgcrypt used is indeed one which fits all requirements. Even with binary compatibility, new features may have been introduced, but due to problem with the dynamic linker an old version may actually be used. So you may want to check that the version is okay right after program startup.

Function: const char * gcry_check_version (const char *req_version) ¶

The function gcry_check_version initializes some subsystems used by Libgcrypt and must be invoked before any other function in the library. See Multi-Threading.

Furthermore, this function returns the version number of the library. It can also verify that the version number is higher than a certain required version number req_version, if this value is not a null pointer.

Libgcrypt uses a concept known as secure memory, which is a region of memory set aside for storing sensitive data. Because such memory is a scarce resource, it needs to be setup in advanced to a fixed size. Further, most operating systems have special requirements on how that secure memory can be used. For example, it might be required to install an application as “setuid(root)” to allow allocating such memory. Libgcrypt requires a sequence of initialization steps to make sure that this works correctly. The following examples show the necessary steps.

If you don’t have a need for secure memory, for example if your application does not use secret keys or other confidential data or it runs in a controlled environment where key material floating around in memory is not a problem, you should initialize Libgcrypt this way:

  /* Version check should be the very first call because it
     makes sure that important subsystems are initialized.
     #define NEED_LIBGCRYPT_VERSION to the minimum required version. */
  if (!gcry_check_version (NEED_LIBGCRYPT_VERSION))
    {
      fprintf (stderr, "libgcrypt is too old (need %s, have %s)\n",
         NEED_LIBGCRYPT_VERSION, gcry_check_version (NULL));
      exit (2);
    }

  /* Disable secure memory.  */
  gcry_control (GCRYCTL_DISABLE_SECMEM, 0);

  /* ... If required, other initialization goes here.  */

  /* Tell Libgcrypt that initialization has completed. */
  gcry_control (GCRYCTL_INITIALIZATION_FINISHED, 0);

If you have to protect your keys or other information in memory against being swapped out to disk and to enable an automatic overwrite of used and freed memory, you need to initialize Libgcrypt this way:

  /* Version check should be the very first call because it
     makes sure that important subsystems are initialized.
     #define NEED_LIBGCRYPT_VERSION to the minimum required version. */
  if (!gcry_check_version (NEED_LIBGCRYPT_VERSION))
    {
      fprintf (stderr, "libgcrypt is too old (need %s, have %s)\n",
         NEED_LIBGCRYPT_VERSION, gcry_check_version (NULL));
      exit (2);
    }

  /* We don't want to see any warnings, e.g. because we have not yet
     parsed program options which might be used to suppress such
     warnings. */
  gcry_control (GCRYCTL_SUSPEND_SECMEM_WARN);

  /* ... If required, other initialization goes here.  Note that the
     process might still be running with increased privileges and that
     the secure memory has not been initialized.  */

  /* Allocate a pool of 16k secure memory.  This makes the secure memory
     available and also drops privileges where needed.  Note that by
     using functions like gcry_xmalloc_secure and gcry_mpi_snew Libgcrypt
     may expand the secure memory pool with memory which lacks the
     property of not being swapped out to disk.   */
  gcry_control (GCRYCTL_INIT_SECMEM, 16384, 0);

  /* It is now okay to let Libgcrypt complain when there was/is
     a problem with the secure memory. */
  gcry_control (GCRYCTL_RESUME_SECMEM_WARN);

  /* ... If required, other initialization goes here.  */

  /* Tell Libgcrypt that initialization has completed. */
  gcry_control (GCRYCTL_INITIALIZATION_FINISHED, 0);

It is important that these initialization steps are not done by a library but by the actual application. A library using Libgcrypt might want to check for finished initialization using:

  if (!gcry_control (GCRYCTL_INITIALIZATION_FINISHED_P))
    {
      fputs ("libgcrypt has not been initialized\n", stderr);
      abort ();
    }

Instead of terminating the process, the library may instead print a warning and try to initialize Libgcrypt itself. See also the section on multi-threading below for more pitfalls.

2.5 Multi-Threading

As mentioned earlier, the Libgcrypt library is thread-safe if you adhere to the following requirements:

• If you use pthread and your applications forks and does not directly call exec (even calling stdio functions), all kind of problems may occur. Future versions of Libgcrypt will try to cleanup using pthread_atfork but even that may lead to problems. This is a common problem with almost all applications using pthread and fork.
• The function gcry_check_version must be called before any other function in the library. To achieve this in multi-threaded programs, you must synchronize the memory with respect to other threads that also want to use Libgcrypt. For this, it is sufficient to call gcry_check_version before creating the other threads using Libgcrypt¹.
• Just like the function gpg_strerror, the function gcry_strerror is not thread safe. You have to use gpg_strerror_r instead.

2.6 How to enable the FIPS mode

Libgcrypt may be used in a FIPS 140-2 mode. Note, that this does not necessary mean that Libcgrypt is an appoved FIPS 140-2 module. Check the NIST database at http://csrc.nist.gov/groups/STM/cmvp/ to see what versions of Libgcrypt are approved.

Because FIPS 140 has certain restrictions on the use of cryptography which are not always wanted, Libgcrypt needs to be put into FIPS mode explicitly. Three alternative mechanisms are provided to switch Libgcrypt into this mode:

• If the file /proc/sys/crypto/fips_enabled exists and contains a numeric value other than 0, Libgcrypt is put into FIPS mode at initialization time. Obviously this works only on systems with a proc file system (i.e. GNU/Linux).
• If the file /etc/gcrypt/fips_enabled exists, Libgcrypt is put into FIPS mode at initialization time. Note that this filename is hardwired and does not depend on any configuration options.
• If the application requests FIPS mode using the control command GCRYCTL_FORCE_FIPS_MODE. This must be done prior to any initialization (i.e. before gcry_check_version).

In addition to the standard FIPS mode, Libgcrypt may also be put into an Enforced FIPS mode by writing a non-zero value into the file /etc/gcrypt/fips_enabled or by using the control command GCRYCTL_SET_ENFORCED_FIPS_FLAG before any other calls to libgcrypt. The Enforced FIPS mode helps to detect applications which don’t fulfill all requirements for using Libgcrypt in FIPS mode (see Description of the FIPS Mode).

Once Libgcrypt has been put into FIPS mode, it is not possible to switch back to standard mode without terminating the process first. If the logging verbosity level of Libgcrypt has been set to at least 2, the state transitions and the self-tests are logged.

2.7 How to disable hardware features

Libgcrypt makes use of certain hardware features. If the use of a feature is not desired it may be either be disabled by a program or globally using a configuration file. The currently supported features are

padlock-rng

padlock-aes

padlock-sha

padlock-mmul

intel-cpu

intel-fast-shld

intel-bmi2

intel-ssse3

intel-sse4.1

intel-pclmul

intel-aesni

intel-rdrand

intel-avx

intel-avx2

intel-fast-vpgather

intel-rdtsc

intel-shaext

intel-vaes-vpclmul

arm-neon

arm-aes

arm-sha1

arm-sha2

arm-pmull

ppc-vcrypto

ppc-arch_3_00

ppc-arch_2_07

s390x-msa

s390x-msa-4

s390x-msa-8

s390x-vx

To disable a feature for all processes using Libgcrypt 1.6 or newer, create the file /etc/gcrypt/hwf.deny and put each feature not to be used on a single line. Empty lines, white space, and lines prefixed with a hash mark are ignored. The file should be world readable.

To disable a feature specifically for a program that program must tell it Libgcrypt before before calling gcry_check_version. Example:²

  gcry_control (GCRYCTL_DISABLE_HWF, "intel-rdrand", NULL);

To print the list of active features you may use this command:

  mpicalc --print-config | grep ^hwflist: | tr : '\n' | tail -n +2

3.1 Controlling the library

Function: gcry_error_t gcry_control (enum gcry_ctl_cmds cmd, ...) ¶

This function can be used to influence the general behavior of Libgcrypt in several ways. Depending on cmd, more arguments can or have to be provided.

GCRYCTL_ENABLE_M_GUARD; Arguments: none

This command enables the built-in memory guard. It must not be used to activate the memory guard after the memory management has already been used; therefore it can ONLY be used before gcry_check_version. Note that the memory guard is NOT used when the user of the library has set his own memory management callbacks.

GCRYCTL_ENABLE_QUICK_RANDOM; Arguments: none

This command inhibits the use the very secure random quality level (GCRY_VERY_STRONG_RANDOM) and degrades all request down to GCRY_STRONG_RANDOM. In general this is not recommended. However, for some applications the extra quality random Libgcrypt tries to create is not justified and this option may help to get better performance. Please check with a crypto expert whether this option can be used for your application.

This option can only be used at initialization time.

GCRYCTL_DUMP_RANDOM_STATS; Arguments: none

This command dumps random number generator related statistics to the library’s logging stream.

GCRYCTL_DUMP_MEMORY_STATS; Arguments: none

This command dumps memory management related statistics to the library’s logging stream.

GCRYCTL_DUMP_SECMEM_STATS; Arguments: none

This command dumps secure memory management related statistics to the library’s logging stream.

GCRYCTL_DROP_PRIVS; Arguments: none

This command disables the use of secure memory and drops the privileges of the current process. This command has not much use; the suggested way to disable secure memory is to use GCRYCTL_DISABLE_SECMEM right after initialization.

GCRYCTL_DISABLE_SECMEM; Arguments: none

This command disables the use of secure memory. If this command is used in FIPS mode, FIPS mode will be disabled and the function gcry_fips_mode_active returns false. However, in Enforced FIPS mode this command has no effect at all.

Many applications do not require secure memory, so they should disable it right away. This command should be executed right after gcry_check_version.

GCRYCTL_DISABLE_LOCKED_SECMEM; Arguments: none

This command disables the use of the mlock call for secure memory. Disabling the use of mlock may for example be done if an encrypted swap space is in use. This command should be executed right after gcry_check_version. Note that by using functions like gcry_xmalloc_secure and gcry_mpi_snew Libgcrypt may expand the secure memory pool with memory which lacks the property of not being swapped out to disk (but will still be zeroed out on free).

GCRYCTL_DISABLE_PRIV_DROP; Arguments: none

This command sets a global flag to tell the secure memory subsystem that it shall not drop privileges after secure memory has been allocated. This command is commonly used right after gcry_check_version but may also be used right away at program startup. It won’t have an effect after the secure memory pool has been initialized. WARNING: A process running setuid(root) is a severe security risk. Processes making use of Libgcrypt or other complex code should drop these extra privileges as soon as possible. If this command has been used the caller is responsible for dropping the privileges.

GCRYCTL_INIT_SECMEM; Arguments: unsigned int nbytes

This command is used to allocate a pool of secure memory and thus enabling the use of secure memory. It also drops all extra privileges the process has (i.e. if it is run as setuid (root)). If the argument nbytes is 0, secure memory will be disabled. The minimum amount of secure memory allocated is currently 16384 bytes; you may thus use a value of 1 to request that default size.

GCRYCTL_AUTO_EXPAND_SECMEM; Arguments: unsigned int chunksize

This command enables on-the-fly expanding of the secure memory area. Note that by using functions like gcry_xmalloc_secure and gcry_mpi_snew will do this auto expanding anyway. The argument to this option is the suggested size for new secure memory areas. A larger size improves performance of all memory allocation and releasing functions. The given chunksize is rounded up to the next 32KiB. The drawback of auto expanding is that memory might be swapped out to disk; this can be fixed by configuring the system to use an encrypted swap space.

GCRYCTL_TERM_SECMEM; Arguments: none

This command zeroises the secure memory and destroys the handler. The secure memory pool may not be used anymore after running this command. If the secure memory pool as already been destroyed, this command has no effect. Applications might want to run this command from their exit handler to make sure that the secure memory gets properly destroyed. This command is not necessarily thread-safe but that should not be needed in cleanup code. It may be called from a signal handler.

GCRYCTL_DISABLE_SECMEM_WARN; Arguments: none

Disable warning messages about problems with the secure memory subsystem. This command should be run right after gcry_check_version.

GCRYCTL_SUSPEND_SECMEM_WARN; Arguments: none

Postpone warning messages from the secure memory subsystem. See the initialization example, on how to use it.

GCRYCTL_RESUME_SECMEM_WARN; Arguments: none

Resume warning messages from the secure memory subsystem. See the initialization example, on how to use it.

GCRYCTL_USE_SECURE_RNDPOOL; Arguments: none

This command tells the PRNG to store random numbers in secure memory. This command should be run right after gcry_check_version and not later than the command GCRYCTL_INIT_SECMEM. Note that in FIPS mode the secure memory is always used.

GCRYCTL_SET_RANDOM_SEED_FILE; Arguments: const char *filename

This command specifies the file, which is to be used as seed file for the PRNG. If the seed file is registered prior to initialization of the PRNG, the seed file’s content (if it exists and seems to be valid) is fed into the PRNG pool. After the seed file has been registered, the PRNG can be signalled to write out the PRNG pool’s content into the seed file with the following command.

GCRYCTL_UPDATE_RANDOM_SEED_FILE; Arguments: none

Write out the PRNG pool’s content into the registered seed file.

Multiple instances of the applications sharing the same random seed file can be started in parallel, in which case they will read out the same pool and then race for updating it (the last update overwrites earlier updates). They will differentiate only by the weak entropy that is added in read_seed_file based on the PID and clock, and up to 16 bytes of weak random non-blockingly. The consequence is that the output of these different instances is correlated to some extent. In a perfect attack scenario, the attacker can control (or at least guess) the PID and clock of the application, and drain the system’s entropy pool to reduce the "up to 16 bytes" above to 0. Then the dependencies of the initial states of the pools are completely known. Note that this is not an issue if random of GCRY_VERY_STRONG_RANDOM quality is requested as in this case enough extra entropy gets mixed. It is also not an issue when using Linux (rndlinux driver), because this one guarantees to read full 16 bytes from /dev/urandom and thus there is no way for an attacker without kernel access to control these 16 bytes.

GCRYCTL_CLOSE_RANDOM_DEVICE; Arguments: none

Try to close the random device. If on Unix system you call fork(), the child process does no call exec(), and you do not intend to use Libgcrypt in the child, it might be useful to use this control code to close the inherited file descriptors of the random device. If Libgcrypt is later used again by the child, the device will be re-opened. On non-Unix systems this control code is ignored.

GCRYCTL_SET_VERBOSITY; Arguments: int level

This command sets the verbosity of the logging. A level of 0 disables all extra logging whereas positive numbers enable more verbose logging. The level may be changed at any time but be aware that no memory synchronization is done so the effect of this command might not immediately show up in other threads. This command may even be used prior to gcry_check_version.

GCRYCTL_SET_DEBUG_FLAGS; Arguments: unsigned int flags

Set the debug flag bits as given by the argument. Be aware that no memory synchronization is done so the effect of this command might not immediately show up in other threads. The debug flags are not considered part of the API and thus may change without notice. As of now bit 0 enables debugging of cipher functions and bit 1 debugging of multi-precision-integers. This command may even be used prior to gcry_check_version.

GCRYCTL_CLEAR_DEBUG_FLAGS; Arguments: unsigned int flags

Set the debug flag bits as given by the argument. Be aware that that no memory synchronization is done so the effect of this command might not immediately show up in other threads. This command may even be used prior to gcry_check_version.

GCRYCTL_DISABLE_INTERNAL_LOCKING; Arguments: none

This command does nothing. It exists only for backward compatibility.

GCRYCTL_ANY_INITIALIZATION_P; Arguments: none

This command returns true if the library has been basically initialized. Such a basic initialization happens implicitly with many commands to get certain internal subsystems running. The common and suggested way to do this basic initialization is by calling gcry_check_version.

GCRYCTL_INITIALIZATION_FINISHED; Arguments: none

This command tells the library that the application has finished the initialization.

GCRYCTL_INITIALIZATION_FINISHED_P; Arguments: none

This command returns true if the command
GCRYCTL_INITIALIZATION_FINISHED has already been run.

GCRYCTL_SET_THREAD_CBS; Arguments: struct ath_ops *ath_ops

This command is obsolete since version 1.6.

GCRYCTL_FAST_POLL; Arguments: none

Run a fast random poll.

GCRYCTL_SET_RNDEGD_SOCKET; Arguments: const char *filename

This command may be used to override the default name of the EGD socket to connect to. It may be used only during initialization as it is not thread safe. Changing the socket name again is not supported. The function may return an error if the given filename is too long for a local socket name.

EGD is an alternative random gatherer, used only on systems lacking a proper random device.

GCRYCTL_PRINT_CONFIG; Arguments: FILE *stream

This command dumps information pertaining to the configuration of the library to the given stream. If NULL is given for stream, the log system is used. This command may be used before the initialization has been finished but not before a gcry_check_version. Note that the macro estream_t can be used instead of gpgrt_stream_t.

GCRYCTL_OPERATIONAL_P; Arguments: none

This command returns true if the library is in an operational state. This information makes only sense in FIPS mode. In contrast to other functions, this is a pure test function and won’t put the library into FIPS mode or change the internal state. This command may be used before the initialization has been finished but not before a gcry_check_version.

GCRYCTL_FIPS_MODE_P; Arguments: none

This command returns true if the library is in FIPS mode. Note, that this is no indication about the current state of the library. This command may be used before the initialization has been finished but not before a gcry_check_version. An application may use this command or the convenience macro below to check whether FIPS mode is actually active.

Function: int gcry_fips_mode_active (void) ¶

Returns true if the FIPS mode is active. Note that this is implemented as a macro.

GCRYCTL_FORCE_FIPS_MODE; Arguments: none

Running this command puts the library into FIPS mode. If the library is already in FIPS mode, a self-test is triggered and thus the library will be put into operational state. This command may be used before a call to gcry_check_version and that is actually the recommended way to let an application switch the library into FIPS mode. Note that Libgcrypt will reject an attempt to switch to fips mode during or after the initialization.

GCRYCTL_SET_ENFORCED_FIPS_FLAG; Arguments: none

Running this command sets the internal flag that puts the library into the enforced FIPS mode during the FIPS mode initialization. This command does not affect the library if the library is not put into the FIPS mode and it must be used before any other libgcrypt library calls that initialize the library such as gcry_check_version. Note that Libgcrypt will reject an attempt to switch to the enforced fips mode during or after the initialization.

GCRYCTL_SET_PREFERRED_RNG_TYPE; Arguments: int

These are advisory commands to select a certain random number generator. They are only advisory because libraries may not know what an application actually wants or vice versa. Thus Libgcrypt employs a priority check to select the actually used RNG. If an applications selects a lower priority RNG but a library requests a higher priority RNG Libgcrypt will switch to the higher priority RNG. Applications and libraries should use these control codes before gcry_check_version. The available generators are:

GCRY_RNG_TYPE_STANDARD

A conservative standard generator based on the “Continuously Seeded Pseudo Random Number Generator” designed by Peter Gutmann.

GCRY_RNG_TYPE_FIPS

A deterministic random number generator conforming to he document “NIST-Recommended Random Number Generator Based on ANSI X9.31 Appendix A.2.4 Using the 3-Key Triple DES and AES Algorithms” (2005-01-31). This implementation uses the AES variant.

GCRY_RNG_TYPE_SYSTEM

A wrapper around the system’s native RNG. On Unix system these are usually the /dev/random and /dev/urandom devices.

The default is GCRY_RNG_TYPE_STANDARD unless FIPS mode as been enabled; in which case GCRY_RNG_TYPE_FIPS is used and locked against further changes.

GCRYCTL_GET_CURRENT_RNG_TYPE; Arguments: int *

This command stores the type of the currently used RNG as an integer value at the provided address.

GCRYCTL_SELFTEST; Arguments: none

This may be used at anytime to have the library run all implemented self-tests. It works in standard and in FIPS mode. Returns 0 on success or an error code on failure.

GCRYCTL_DISABLE_HWF; Arguments: const char *name

Libgcrypt detects certain features of the CPU at startup time. For performance tests it is sometimes required not to use such a feature. This option may be used to disable a certain feature; i.e. Libgcrypt behaves as if this feature has not been detected. This call can be used several times to disable a set of features, or features may be given as a colon or comma delimited string. The special feature "all" can be used to disable all available features.

Note that the detection code might be run if the feature has been disabled. This command must be used at initialization time; i.e. before calling gcry_check_version.

GCRYCTL_REINIT_SYSCALL_CLAMP; Arguments: none

Libgcrypt wraps blocking system calls with two functions calls (“system call clamp”) to give user land threading libraries a hook for re-scheduling. This works by reading the system call clamp from Libgpg-error at initialization time. However sometimes Libgcrypt needs to be initialized before the user land threading systems and at that point the system call clamp has not been registered with Libgpg-error and in turn Libgcrypt would not use them. The control code can be used to tell Libgcrypt that a system call clamp has now been registered with Libgpg-error and advise Libgcrypt to read the clamp again. Obviously this control code may only be used before a second thread is started in a process.

3.2 Error Handling

Many functions in Libgcrypt can return an error if they fail. For this reason, the application should always catch the error condition and take appropriate measures, for example by releasing the resources and passing the error up to the caller, or by displaying a descriptive message to the user and cancelling the operation.

Some error values do not indicate a system error or an error in the operation, but the result of an operation that failed properly. For example, if you try to decrypt a tempered message, the decryption will fail. Another error value actually means that the end of a data buffer or list has been reached. The following descriptions explain for many error codes what they mean usually. Some error values have specific meanings if returned by a certain functions. Such cases are described in the documentation of those functions.

Libgcrypt uses the libgpg-error library. This allows to share the error codes with other components of the GnuPG system, and to pass error values transparently from the crypto engine, or some helper application of the crypto engine, to the user. This way no information is lost. As a consequence, Libgcrypt does not use its own identifiers for error codes, but uses those provided by libgpg-error. They usually start with GPG_ERR_.

However, Libgcrypt does provide aliases for the functions defined in libgpg-error, which might be preferred for name space consistency.

Most functions in Libgcrypt return an error code in the case of failure. For this reason, the application should always catch the error condition and take appropriate measures, for example by releasing the resources and passing the error up to the caller, or by displaying a descriptive message to the user and canceling the operation.

Some error values do not indicate a system error or an error in the operation, but the result of an operation that failed properly.

GnuPG components, including Libgcrypt, use an extra library named libgpg-error to provide a common error handling scheme. For more information on libgpg-error, see the according manual.

• Error Values
• Error Sources
• Error Codes
• Error Strings

{
  gcry_cipher_hd_t handle;
  gcry_error_t err = 0;

  err = gcry_cipher_open (&handle, GCRY_CIPHER_AES,
                          GCRY_CIPHER_MODE_CBC, 0);
  if (err)
    {
      fprintf (stderr, "Failure: %s/%s\n",
               gcry_strsource (err),
               gcry_strerror (err));
    }
}

4.1 Progress handler

It is often useful to retrieve some feedback while long running operations are performed.

Data type: gcry_handler_progress_t ¶

Progress handler functions have to be of the type gcry_handler_progress_t, which is defined as:

void (*gcry_handler_progress_t) (void *, const char *, int, int, int)

The following function may be used to register a handler function for this purpose.

Function: void gcry_set_progress_handler (gcry_handler_progress_t cb, void *cb_data) ¶

This function installs cb as the ‘Progress handler’ function. It may be used only during initialization. cb must be defined as follows:

void
my_progress_handler (void *cb_data, const char *what,
                     int printchar, int current, int total)
{
  /* Do something.  */
}

A description of the arguments of the progress handler function follows.

cb_data

The argument provided in the call to gcry_set_progress_handler.

what

A string identifying the type of the progress output. The following values for what are defined:

need_entropy

Not enough entropy is available. total holds the number of required bytes.

wait_dev_random

Waiting to re-open a random device. total gives the number of seconds until the next try.

primegen

Values for printchar:

\n

Prime generated.

!

Need to refresh the pool of prime numbers.

<, >

Number of bits adjusted.

^

Searching for a generator.

.

Fermat test on 10 candidates failed.

:

Restart with a new random value.

+

Rabin Miller test passed.

4.2 Allocation handler

It is possible to make Libgcrypt use special memory allocation functions instead of the built-in ones.

Memory allocation functions are of the following types:

Data type: gcry_handler_alloc_t ¶

This type is defined as: void *(*gcry_handler_alloc_t) (size_t n).

Data type: gcry_handler_secure_check_t ¶

This type is defined as: int *(*gcry_handler_secure_check_t) (const void *).

Data type: gcry_handler_realloc_t ¶

This type is defined as: void *(*gcry_handler_realloc_t) (void *p, size_t n).

Data type: gcry_handler_free_t ¶

This type is defined as: void *(*gcry_handler_free_t) (void *).

Special memory allocation functions can be installed with the following function:

Function: void gcry_set_allocation_handler (gcry_handler_alloc_t func_alloc, gcry_handler_alloc_t func_alloc_secure, gcry_handler_secure_check_t func_secure_check, gcry_handler_realloc_t func_realloc, gcry_handler_free_t func_free) ¶

Install the provided functions and use them instead of the built-in functions for doing memory allocation. Using this function is in general not recommended because the standard Libgcrypt allocation functions are guaranteed to zeroize memory if needed.

This function may be used only during initialization and may not be used in fips mode.

4.3 Error handler

The following functions may be used to register handler functions that are called by Libgcrypt in case certain error conditions occur. They may and should be registered prior to calling gcry_check_version.

Data type: gcry_handler_no_mem_t ¶

This type is defined as: int (*gcry_handler_no_mem_t) (void *, size_t, unsigned int)

Function: void gcry_set_outofcore_handler (gcry_handler_no_mem_t func_no_mem, void *cb_data) ¶

This function registers func_no_mem as ‘out-of-core handler’, which means that it will be called in the case of not having enough memory available. The handler is called with 3 arguments: The first one is the pointer cb_data as set with this function, the second is the requested memory size and the last being a flag. If bit 0 of the flag is set, secure memory has been requested. The handler should either return true to indicate that Libgcrypt should try again allocating memory or return false to let Libgcrypt use its default fatal error handler.

Data type: gcry_handler_error_t ¶

This type is defined as: void (*gcry_handler_error_t) (void *, int, const char *)

Function: void gcry_set_fatalerror_handler (gcry_handler_error_t func_error, void *cb_data) ¶

This function registers func_error as ‘error handler’, which means that it will be called in error conditions.

4.4 Logging handler

Data type: gcry_handler_log_t ¶

This type is defined as: void (*gcry_handler_log_t) (void *, int, const char *, va_list)

Function: void gcry_set_log_handler (gcry_handler_log_t func_log, void *cb_data) ¶

This function registers func_log as ‘logging handler’, which means that it will be called in case Libgcrypt wants to log a message. This function may and should be used prior to calling gcry_check_version.

5.1 Available ciphers

GCRY_CIPHER_NONE

This is not a real algorithm but used by some functions as error return. The value always evaluates to false.

GCRY_CIPHER_IDEA ¶

This is the IDEA algorithm.

GCRY_CIPHER_3DES ¶

Triple-DES with 3 Keys as EDE. The key size of this algorithm is 168 bits but you have to pass 192 bits because the most significant bits of each byte are ignored.

GCRY_CIPHER_CAST5 ¶

CAST128-5 block cipher algorithm. The key size is 128 bits.

GCRY_CIPHER_BLOWFISH ¶

The blowfish algorithm. The supported key sizes are 8 to 576 bits in 8 bit increments.

GCRY_CIPHER_SAFER_SK128

Reserved and not currently implemented.

GCRY_CIPHER_DES_SK

Reserved and not currently implemented.

GCRY_CIPHER_AES ¶

GCRY_CIPHER_AES128

GCRY_CIPHER_RIJNDAEL

GCRY_CIPHER_RIJNDAEL128

AES (Rijndael) with a 128 bit key.

GCRY_CIPHER_AES192

GCRY_CIPHER_RIJNDAEL192

AES (Rijndael) with a 192 bit key.

GCRY_CIPHER_AES256

GCRY_CIPHER_RIJNDAEL256

AES (Rijndael) with a 256 bit key.

GCRY_CIPHER_TWOFISH ¶

The Twofish algorithm with a 256 bit key.

GCRY_CIPHER_TWOFISH128

The Twofish algorithm with a 128 bit key.

GCRY_CIPHER_ARCFOUR ¶

An algorithm which is 100% compatible with RSA Inc.’s RC4 algorithm. Note that this is a stream cipher and must be used very carefully to avoid a couple of weaknesses.

GCRY_CIPHER_DES ¶

Standard DES with a 56 bit key. You need to pass 64 bit but the high bits of each byte are ignored. Note, that this is a weak algorithm which can be broken in reasonable time using a brute force approach.

GCRY_CIPHER_SERPENT128 ¶

GCRY_CIPHER_SERPENT192

GCRY_CIPHER_SERPENT256

The Serpent cipher from the AES contest.

GCRY_CIPHER_RFC2268_40 ¶

GCRY_CIPHER_RFC2268_128

Ron’s Cipher 2 in the 40 and 128 bit variants.

GCRY_CIPHER_SEED ¶

A 128 bit cipher as described by RFC4269.

GCRY_CIPHER_CAMELLIA128 ¶

GCRY_CIPHER_CAMELLIA192

GCRY_CIPHER_CAMELLIA256

The Camellia cipher by NTT. See http://info.isl.ntt.co.jp/crypt/eng/camellia/specifications.html.

GCRY_CIPHER_SALSA20 ¶

This is the Salsa20 stream cipher.

GCRY_CIPHER_SALSA20R12 ¶

This is the Salsa20/12 - reduced round version of Salsa20 stream cipher.

GCRY_CIPHER_GOST28147 ¶

The GOST 28147-89 cipher, defined in the respective GOST standard. Translation of this GOST into English is provided in the RFC-5830.

GCRY_CIPHER_GOST28147_MESH ¶

The GOST 28147-89 cipher, defined in the respective GOST standard. Translation of this GOST into English is provided in the RFC-5830. This cipher will use CryptoPro keymeshing as defined in RFC 4357 if it has to be used for the selected parameter set.

GCRY_CIPHER_CHACHA20 ¶

This is the ChaCha20 stream cipher.

GCRY_CIPHER_SM4 ¶

A 128 bit cipher by the State Cryptography Administration of China (SCA). See https://tools.ietf.org/html/draft-ribose-cfrg-sm4-10.

5.2 Available cipher modes

GCRY_CIPHER_MODE_NONE

No mode specified. This should not be used. The only exception is that if Libgcrypt is not used in FIPS mode and if any debug flag has been set, this mode may be used to bypass the actual encryption.

GCRY_CIPHER_MODE_ECB ¶

Electronic Codebook mode.

GCRY_CIPHER_MODE_CFB

GCRY_CIPHER_MODE_CFB8 ¶

Cipher Feedback mode. For GCRY_CIPHER_MODE_CFB the shift size equals the block size of the cipher (e.g. for AES it is CFB-128). For GCRY_CIPHER_MODE_CFB8 the shift size is 8 bit but that variant is not yet available.

GCRY_CIPHER_MODE_CBC ¶

Cipher Block Chaining mode.

GCRY_CIPHER_MODE_STREAM

Stream mode, only to be used with stream cipher algorithms.

GCRY_CIPHER_MODE_OFB ¶

Output Feedback mode.

GCRY_CIPHER_MODE_CTR ¶

Counter mode.

GCRY_CIPHER_MODE_AESWRAP ¶

This mode is used to implement the AES-Wrap algorithm according to RFC-3394. It may be used with any 128 bit block length algorithm, however the specs require one of the 3 AES algorithms. These special conditions apply: If gcry_cipher_setiv has not been used the standard IV is used; if it has been used the lower 64 bit of the IV are used as the Alternative Initial Value. On encryption the provided output buffer must be 64 bit (8 byte) larger than the input buffer; in-place encryption is still allowed. On decryption the output buffer may be specified 64 bit (8 byte) shorter than then input buffer. As per specs the input length must be at least 128 bits and the length must be a multiple of 64 bits.

GCRY_CIPHER_MODE_CCM ¶

Counter with CBC-MAC mode is an Authenticated Encryption with Associated Data (AEAD) block cipher mode, which is specified in ’NIST Special Publication 800-38C’ and RFC 3610.

GCRY_CIPHER_MODE_GCM ¶

Galois/Counter Mode (GCM) is an Authenticated Encryption with Associated Data (AEAD) block cipher mode, which is specified in ’NIST Special Publication 800-38D’.

GCRY_CIPHER_MODE_POLY1305 ¶

This mode implements the Poly1305 Authenticated Encryption with Associated Data (AEAD) mode according to RFC-8439. This mode can be used with ChaCha20 stream cipher.

GCRY_CIPHER_MODE_OCB ¶

OCB is an Authenticated Encryption with Associated Data (AEAD) block cipher mode, which is specified in RFC-7253. Supported tag lengths are 128, 96, and 64 bit with the default being 128 bit. To switch to a different tag length gcry_cipher_ctl using the command GCRYCTL_SET_TAGLEN and the address of an int variable set to 12 (for 96 bit) or 8 (for 64 bit) provided for the buffer argument and sizeof(int) for buflen.

Note that the use of gcry_cipher_final is required.

GCRY_CIPHER_MODE_XTS ¶

XEX-based tweaked-codebook mode with ciphertext stealing (XTS) mode is used to implement the AES-XTS as specified in IEEE 1619 Standard Architecture for Encrypted Shared Storage Media and NIST SP800-38E.

The XTS mode requires doubling key-length, for example, using 512-bit key with AES-256 (GCRY_CIPHER_AES256). The 128-bit tweak value is feed to XTS mode as little-endian byte array using gcry_cipher_setiv function. When encrypting or decrypting, full-sized data unit buffers needs to be passed to gcry_cipher_encrypt or gcry_cipher_decrypt. The tweak value is automatically incremented after each call of gcry_cipher_encrypt and gcry_cipher_decrypt. Auto-increment allows avoiding need of setting IV between processing of sequential data units.

GCRY_CIPHER_MODE_EAX ¶

EAX is an Authenticated Encryption with Associated Data (AEAD) block cipher mode by Bellare, Rogaway, and Wagner (see http://web.cs.ucdavis.edu/~rogaway/papers/eax.html).

5.3 Working with cipher handles

To use a cipher algorithm, you must first allocate an according handle. This is to be done using the open function:

Function: gcry_error_t gcry_cipher_open (gcry_cipher_hd_t *hd, int algo, int mode, unsigned int flags) ¶

This function creates the context handle required for most of the other cipher functions and returns a handle to it in ‘hd’. In case of an error, an according error code is returned.

The ID of algorithm to use must be specified via algo. See Available ciphers, for a list of supported ciphers and the according constants.

Besides using the constants directly, the function gcry_cipher_map_name may be used to convert the textual name of an algorithm into the according numeric ID.

The cipher mode to use must be specified via mode. See Available cipher modes, for a list of supported cipher modes and the according constants. Note that some modes are incompatible with some algorithms - in particular, stream mode (GCRY_CIPHER_MODE_STREAM) only works with stream ciphers. Poly1305 AEAD mode (GCRY_CIPHER_MODE_POLY1305) only works with ChaCha20 stream cipher. The block cipher modes (GCRY_CIPHER_MODE_ECB, GCRY_CIPHER_MODE_CBC, GCRY_CIPHER_MODE_CFB, GCRY_CIPHER_MODE_OFB, GCRY_CIPHER_MODE_CTR and GCRY_CIPHER_MODE_EAX) will work with any block cipher algorithm. GCM mode (GCRY_CIPHER_MODE_GCM), CCM mode (GCRY_CIPHER_MODE_CCM), OCB mode (GCRY_CIPHER_MODE_OCB), and XTS mode (GCRY_CIPHER_MODE_XTS) will only work with block cipher algorithms which have the block size of 16 bytes.

The third argument flags can either be passed as 0 or as the bit-wise OR of the following constants.

GCRY_CIPHER_SECURE

Make sure that all operations are allocated in secure memory. This is useful when the key material is highly confidential.

GCRY_CIPHER_ENABLE_SYNC ¶

This flag enables the CFB sync mode, which is a special feature of Libgcrypt’s CFB mode implementation to allow for OpenPGP’s CFB variant. See gcry_cipher_sync.

GCRY_CIPHER_CBC_CTS ¶

Enable cipher text stealing (CTS) for the CBC mode. Cannot be used simultaneous as GCRY_CIPHER_CBC_MAC. CTS mode makes it possible to transform data of almost arbitrary size (only limitation is that it must be greater than the algorithm’s block size).

GCRY_CIPHER_CBC_MAC ¶

Compute CBC-MAC keyed checksums. This is the same as CBC mode, but only output the last block. Cannot be used simultaneous as GCRY_CIPHER_CBC_CTS.

Use the following function to release an existing handle:

Function: void gcry_cipher_close (gcry_cipher_hd_t h) ¶

This function releases the context created by gcry_cipher_open. It also zeroises all sensitive information associated with this cipher handle.

In order to use a handle for performing cryptographic operations, a ‘key’ has to be set first:

Function: gcry_error_t gcry_cipher_setkey (gcry_cipher_hd_t h, const void *k, size_t l) ¶

Set the key k used for encryption or decryption in the context denoted by the handle h. The length l (in bytes) of the key k must match the required length of the algorithm set for this context or be in the allowed range for algorithms with variable key size. The function checks this and returns an error if there is a problem. A caller should always check for an error.

Most crypto modes requires an initialization vector (IV), which usually is a non-secret random string acting as a kind of salt value. The CTR mode requires a counter, which is also similar to a salt value. To set the IV or CTR, use these functions:

Function: gcry_error_t gcry_cipher_setiv (gcry_cipher_hd_t h, const void *k, size_t l) ¶

Set the initialization vector used for encryption or decryption. The vector is passed as the buffer K of length l bytes and copied to internal data structures. The function checks that the IV matches the requirement of the selected algorithm and mode.

This function is also used by AEAD modes and with Salsa20 and ChaCha20 stream ciphers to set or update the required nonce. In these cases it needs to be called after setting the key.

Function: gcry_error_t gcry_cipher_setctr (gcry_cipher_hd_t h, const void *c, size_t l) ¶

Set the counter vector used for encryption or decryption. The counter is passed as the buffer c of length l bytes and copied to internal data structures. The function checks that the counter matches the requirement of the selected algorithm (i.e., it must be the same size as the block size).

Function: gcry_error_t gcry_cipher_reset (gcry_cipher_hd_t h) ¶

Set the given handle’s context back to the state it had after the last call to gcry_cipher_setkey and clear the initialization vector.

Note that gcry_cipher_reset is implemented as a macro.

Authenticated Encryption with Associated Data (AEAD) block cipher modes require the handling of the authentication tag and the additional authenticated data, which can be done by using the following functions:

Function: gcry_error_t gcry_cipher_authenticate (gcry_cipher_hd_t h, const void *abuf, size_t abuflen) ¶

Process the buffer abuf of length abuflen as the additional authenticated data (AAD) for AEAD cipher modes.

Function: gcry_error_t gcry_cipher_gettag (gcry_cipher_hd_t h, void *tag, size_t taglen) ¶

This function is used to read the authentication tag after encryption. The function finalizes and outputs the authentication tag to the buffer tag of length taglen bytes.

Depending on the used mode certain restrictions for taglen are enforced: For GCM taglen must be at least 16 or one of the allowed truncated lengths (4, 8, 12, 13, 14, or 15).

Function: gcry_error_t gcry_cipher_checktag (gcry_cipher_hd_t h, const void *tag, size_t taglen) ¶

Check the authentication tag after decryption. The authentication tag is passed as the buffer tag of length taglen bytes and compared to internal authentication tag computed during decryption. Error code GPG_ERR_CHECKSUM is returned if the authentication tag in the buffer tag does not match the authentication tag calculated during decryption.

Depending on the used mode certain restrictions for taglen are enforced: For GCM taglen must either be 16 or one of the allowed truncated lengths (4, 8, 12, 13, 14, or 15).

The actual encryption and decryption is done by using one of the following functions. They may be used as often as required to process all the data.

Function: gcry_error_t gcry_cipher_encrypt (gcry_cipher_hd_t h, unsigned char *out, size_t outsize, const unsigned char *in, size_t inlen) ¶

gcry_cipher_encrypt is used to encrypt the data. This function can either work in place or with two buffers. It uses the cipher context already setup and described by the handle h. There are 2 ways to use the function: If in is passed as NULL and inlen is 0, in-place encryption of the data in out of length outsize takes place. With in being not NULL, inlen bytes are encrypted to the buffer out which must have at least a size of inlen. outsize must be set to the allocated size of out, so that the function can check that there is sufficient space. Note that overlapping buffers are not allowed.

Depending on the selected algorithms and encryption mode, the length of the buffers must be a multiple of the block size.

Some encryption modes require that gcry_cipher_final is used before the final data chunk is passed to this function.

The function returns 0 on success or an error code.

Function: gcry_error_t gcry_cipher_decrypt (gcry_cipher_hd_t h, unsigned char *out, size_t outsize, const unsigned char *in, size_t inlen) ¶

gcry_cipher_decrypt is used to decrypt the data. This function can either work in place or with two buffers. It uses the cipher context already setup and described by the handle h. There are 2 ways to use the function: If in is passed as NULL and inlen is 0, in-place decryption of the data in out or length outsize takes place. With in being not NULL, inlen bytes are decrypted to the buffer out which must have at least a size of inlen. outsize must be set to the allocated size of out, so that the function can check that there is sufficient space. Note that overlapping buffers are not allowed.

Depending on the selected algorithms and encryption mode, the length of the buffers must be a multiple of the block size.

Some encryption modes require that gcry_cipher_final is used before the final data chunk is passed to this function.

The function returns 0 on success or an error code.

The OCB mode features integrated padding and must thus be told about the end of the input data. This is done with:

Function: gcry_error_t gcry_cipher_final (gcry_cipher_hd_t h) ¶

Set a flag in the context to tell the encrypt and decrypt functions that their next call will provide the last chunk of data. Only the first call to this function has an effect and only for modes which support it. Checking the error is in general not necessary. This is implemented as a macro.

OpenPGP (as defined in RFC-4880) requires a special sync operation in some places. The following function is used for this:

Function: gcry_error_t gcry_cipher_sync (gcry_cipher_hd_t h) ¶

Perform the OpenPGP sync operation on context h. Note that this is a no-op unless the context was created with the flag GCRY_CIPHER_ENABLE_SYNC

Some of the described functions are implemented as macros utilizing a catch-all control function. This control function is rarely used directly but there is nothing which would inhibit it:

Function: gcry_error_t gcry_cipher_ctl (gcry_cipher_hd_t h, int cmd, void *buffer, size_t buflen) ¶

gcry_cipher_ctl controls various aspects of the cipher module and specific cipher contexts. Usually some more specialized functions or macros are used for this purpose. The semantics of the function and its parameters depends on the the command cmd and the passed context handle h. Please see the comments in the source code (src/global.c) for details.

Function: gcry_error_t gcry_cipher_info (gcry_cipher_hd_t h, int what, void *buffer, size_t *nbytes) ¶

gcry_cipher_info is used to retrieve various information about a cipher context or the cipher module in general.

GCRYCTL_GET_TAGLEN:

Return the length of the tag for an AE algorithm mode. An error is returned for modes which do not support a tag. buffer must be given as NULL. On success the result is stored nbytes. The taglen is returned in bytes.

5.4 General cipher functions

To work with the algorithms, several functions are available to map algorithm names to the internal identifiers, as well as ways to retrieve information about an algorithm or the current cipher context.

Function: gcry_error_t gcry_cipher_algo_info (int algo, int what, void *buffer, size_t *nbytes) ¶

This function is used to retrieve information on a specific algorithm. You pass the cipher algorithm ID as algo and the type of information requested as what. The result is either returned as the return code of the function or copied to the provided buffer whose allocated length must be available in an integer variable with the address passed in nbytes. This variable will also receive the actual used length of the buffer.

Here is a list of supported codes for what:

GCRYCTL_GET_KEYLEN:

Return the length of the key. If the algorithm supports multiple key lengths, the maximum supported value is returned. The length is returned as number of octets (bytes) and not as number of bits in nbytes; buffer must be zero. Note that it is usually better to use the convenience function gcry_cipher_get_algo_keylen.

GCRYCTL_GET_BLKLEN:

Return the block length of the algorithm. The length is returned as a number of octets in nbytes; buffer must be zero. Note that it is usually better to use the convenience function gcry_cipher_get_algo_blklen.

GCRYCTL_TEST_ALGO:

Returns 0 when the specified algorithm is available for use. buffer and nbytes must be zero.

Function: size_t gcry_cipher_get_algo_keylen (algo) ¶

This function returns length of the key for algorithm algo. If the algorithm supports multiple key lengths, the maximum supported key length is returned. On error 0 is returned. The key length is returned as number of octets.

This is a convenience functions which should be preferred over gcry_cipher_algo_info because it allows for proper type checking.

Function: size_t gcry_cipher_get_algo_blklen (int algo) ¶

This functions returns the block-length of the algorithm algo counted in octets. On error 0 is returned.

This is a convenience functions which should be preferred over gcry_cipher_algo_info because it allows for proper type checking.

Function: const char * gcry_cipher_algo_name (int algo) ¶

gcry_cipher_algo_name returns a string with the name of the cipher algorithm algo. If the algorithm is not known or another error occurred, the string "?" is returned. This function should not be used to test for the availability of an algorithm.

Function: int gcry_cipher_map_name (const char *name) ¶

gcry_cipher_map_name returns the algorithm identifier for the cipher algorithm described by the string name. If this algorithm is not available 0 is returned.

Function: int gcry_cipher_mode_from_oid (const char *string) ¶

Return the cipher mode associated with an ASN.1 object identifier. The object identifier is expected to be in the IETF-style dotted decimal notation. The function returns 0 for an unknown object identifier or when no mode is associated with it.

6.1 Available algorithms

Libgcrypt supports the RSA (Rivest-Shamir-Adleman) algorithms as well as DSA (Digital Signature Algorithm), Elgamal, ECDSA, ECDH, and EdDSA.

6.2 Used S-expressions

Libgcrypt’s API for asymmetric cryptography is based on data structures called S-expressions (see http://people.csail.mit.edu/rivest/sexp.html) and does not work with contexts/handles as most of the other building blocks of Libgcrypt do.

The following information are stored in S-expressions:

• keys
• plain text data
• encrypted data
• signatures

To describe how Libgcrypt expect keys, we use examples. Note that words in uppercase indicate parameters whereas lowercase words are literals.

Note that all MPI (multi-precision-integers) values are expected to be in GCRYMPI_FMT_USG format. An easy way to create S-expressions is by using gcry_sexp_build which allows to pass a string with printf-like escapes to insert MPI values.

• RSA key parameters
• DSA key parameters
• ECC key parameters

(private-key
  (rsa
    (n n-mpi)
    (e e-mpi)
    (d d-mpi)
    (p p-mpi)
    (q q-mpi)
    (u u-mpi)))

(public-key
  (rsa
    (n n-mpi)
    (e e-mpi)))

  if (gcry_mpi_cmp (p, q) > 0)
    {
      gcry_mpi_swap (p, q);
      gcry_mpi_invm (u, p, q);
    }

(private-key
  (dsa
    (p p-mpi)
    (q q-mpi)
    (g g-mpi)
    (y y-mpi)
    (x x-mpi)))

(private-key
  (ecc
    (p p-mpi)
    (a a-mpi)
    (b b-mpi)
    (g g-point)
    (n n-mpi)
    (q q-point)
    (d d-mpi)))

(private-key
  (ecc
    (curve "NIST P-192")
    (q q-point)
    (d d-mpi)))

6.3 Cryptographic Functions

Some functions operating on S-expressions support ‘flags’ to influence the operation. These flags have to be listed in a sub-S-expression named ‘flags’. Flag names are case-sensitive. The following flags are known:

comp ¶

nocomp

If supported by the algorithm and curve the comp flag requests that points are returned in compact (compressed) representation. The nocomp flag requests that points are returned with full coordinates. The default depends on the the algorithm and curve. The compact representation requires a small overhead before a point can be used but halves the size of a to be conveyed public key. If comp is used with the “EdDSA” algorithm the key generation prefix the public key with a 0x40 byte.

pkcs1 ¶

Use PKCS#1 block type 2 padding for encryption, block type 1 padding for signing.

oaep ¶

Use RSA-OAEP padding for encryption.

pss ¶

Use RSA-PSS padding for signing.

eddsa ¶

Use the EdDSA scheme signing instead of the default ECDSA algorithm. Note that the EdDSA uses a special form of the public key.

rfc6979 ¶

For DSA and ECDSA use a deterministic scheme for the k parameter.

no-blinding ¶

Do not use a technique called ‘blinding’, which is used by default in order to prevent leaking of secret information. Blinding is only implemented by RSA, but it might be implemented by other algorithms in the future as well, when necessary.

param ¶

For ECC key generation also return the domain parameters. For ECC signing and verification override default parameters by provided domain parameters of the public or private key.

transient-key ¶

This flag is only meaningful for RSA, DSA, and ECC key generation. If given the key is created using a faster and a somewhat less secure random number generator. This flag may be used for keys which are only used for a short time or per-message and do not require full cryptographic strength.

no-keytest ¶

This flag skips internal failsafe tests to assert that a generated key is properly working. It currently has an effect only for standard ECC key generation. It is mostly useful along with transient-key to achieve fastest ECC key generation.

use-x931 ¶

Force the use of the ANSI X9.31 key generation algorithm instead of the default algorithm. This flag is only meaningful for RSA key generation and usually not required. Note that this algorithm is implicitly used if either derive-parms is given or Libgcrypt is in FIPS mode.

use-fips186 ¶

Force the use of the FIPS 186 key generation algorithm instead of the default algorithm. This flag is only meaningful for DSA and usually not required. Note that this algorithm is implicitly used if either derive-parms is given or Libgcrypt is in FIPS mode. As of now FIPS 186-2 is implemented; after the approval of FIPS 186-3 the code will be changed to implement 186-3.

use-fips186-2 ¶

Force the use of the FIPS 186-2 key generation algorithm instead of the default algorithm. This algorithm is slightly different from FIPS 186-3 and allows only 1024 bit keys. This flag is only meaningful for DSA and only required for FIPS testing backward compatibility.

Now that we know the key basics, we can carry on and explain how to encrypt and decrypt data. In almost all cases the data is a random session key which is in turn used for the actual encryption of the real data. There are 2 functions to do this:

Function: gcry_error_t gcry_pk_encrypt (gcry_sexp_t *r_ciph, gcry_sexp_t data, gcry_sexp_t pkey) ¶

Obviously a public key must be provided for encryption. It is expected as an appropriate S-expression (see above) in pkey. The data to be encrypted can either be in the simple old format, which is a very simple S-expression consisting only of one MPI, or it may be a more complex S-expression which also allows to specify flags for operation, like e.g. padding rules.

If you don’t want to let Libgcrypt handle the padding, you must pass an appropriate MPI using this expression for data:

(data
  (flags raw)
  (value mpi))

This has the same semantics as the old style MPI only way. MPI is the actual data, already padded appropriate for your protocol. Most RSA based systems however use PKCS#1 padding and so you can use this S-expression for data:

(data
  (flags pkcs1)
  (value block))

Here, the "flags" list has the "pkcs1" flag which let the function know that it should provide PKCS#1 block type 2 padding. The actual data to be encrypted is passed as a string of octets in block. The function checks that this data actually can be used with the given key, does the padding and encrypts it.

If the function could successfully perform the encryption, the return value will be 0 and a new S-expression with the encrypted result is allocated and assigned to the variable at the address of r_ciph. The caller is responsible to release this value using gcry_sexp_release. In case of an error, an error code is returned and r_ciph will be set to NULL.

The returned S-expression has this format when used with RSA:

(enc-val
  (rsa
    (a a-mpi)))

Where a-mpi is an MPI with the result of the RSA operation. When using the Elgamal algorithm, the return value will have this format:

(enc-val
  (elg
    (a a-mpi)
    (b b-mpi)))

Where a-mpi and b-mpi are MPIs with the result of the Elgamal encryption operation.

Function: gcry_error_t gcry_pk_decrypt (gcry_sexp_t *r_plain, gcry_sexp_t data, gcry_sexp_t skey) ¶

Obviously a private key must be provided for decryption. It is expected as an appropriate S-expression (see above) in skey. The data to be decrypted must match the format of the result as returned by gcry_pk_encrypt, but should be enlarged with a flags element:

(enc-val
  (flags)
  (elg
    (a a-mpi)
    (b b-mpi)))

This function does not remove padding from the data by default. To let Libgcrypt remove padding, give a hint in ‘flags’ telling which padding method was used when encrypting:

(flags padding-method)

Currently padding-method is either pkcs1 for PKCS#1 block type 2 padding, or oaep for RSA-OAEP padding.

The function returns 0 on success or an error code. The variable at the address of r_plain will be set to NULL on error or receive the decrypted value on success. The format of r_plain is a simple S-expression part (i.e. not a valid one) with just one MPI if there was no flags element in data; if at least an empty flags is passed in data, the format is:

(value plaintext)

Another operation commonly performed using public key cryptography is signing data. In some sense this is even more important than encryption because digital signatures are an important instrument for key management. Libgcrypt supports digital signatures using 2 functions, similar to the encryption functions:

Function: gcry_error_t gcry_pk_sign (gcry_sexp_t *r_sig, gcry_sexp_t data, gcry_sexp_t skey) ¶

This function creates a digital signature for data using the private key skey and place it into the variable at the address of r_sig. data may either be the simple old style S-expression with just one MPI or a modern and more versatile S-expression which allows to let Libgcrypt handle padding:

 (data
  (flags pkcs1)
  (hash hash-algo block))

This example requests to sign the data in block after applying PKCS#1 block type 1 style padding. hash-algo is a string with the hash algorithm to be encoded into the signature, this may be any hash algorithm name as supported by Libgcrypt. Most likely, this will be "sha256" or "sha1". It is obvious that the length of block must match the size of that message digests; the function checks that this and other constraints are valid.

If PKCS#1 padding is not required (because the caller does already provide a padded value), either the old format or better the following format should be used:

(data
  (flags raw)
  (value mpi))

Here, the data to be signed is directly given as an MPI.

For DSA the input data is expected in this format:

(data
  (flags raw)
  (value mpi))

Here, the data to be signed is directly given as an MPI. It is expect that this MPI is the the hash value. For the standard DSA using a MPI is not a problem in regard to leading zeroes because the hash value is directly used as an MPI. For better standard conformance it would be better to explicit use a memory string (like with pkcs1) but that is currently not supported. However, for deterministic DSA as specified in RFC6979 this can’t be used. Instead the following input is expected.

(data
  (flags rfc6979)
  (hash hash-algo block))

Note that the provided hash-algo is used for the internal HMAC; it should match the hash-algo used to create block.

The signature is returned as a newly allocated S-expression in r_sig using this format for RSA:

(sig-val
  (rsa
    (s s-mpi)))

Where s-mpi is the result of the RSA sign operation. For DSA the S-expression returned is:

(sig-val
  (dsa
    (r r-mpi)
    (s s-mpi)))

Where r-mpi and s-mpi are the result of the DSA sign operation.

For Elgamal signing (which is slow, yields large numbers and probably is not as secure as the other algorithms), the same format is used with "elg" replacing "dsa"; for ECDSA signing, the same format is used with "ecdsa" replacing "dsa".

For the EdDSA algorithm (cf. Ed25515) the required input parameters are:

(data
  (flags eddsa)
  (hash-algo sha512)
  (value message))

Note that the message may be of any length; hashing is part of the algorithm. Using a large data block for message is in general not suggested; in that case the used protocol should better require that a hash of the message is used as input to the EdDSA algorithm. Note that for X.509 certificates message is the tbsCertificate part and in CMS message is the signedAttrs part; see RFC-8410 and RFC-8419.

The operation most commonly used is definitely the verification of a signature. Libgcrypt provides this function:

Function: gcry_error_t gcry_pk_verify (gcry_sexp_t sig, gcry_sexp_t data, gcry_sexp_t pkey) ¶

This is used to check whether the signature sig matches the data. The public key pkey must be provided to perform this verification. This function is similar in its parameters to gcry_pk_sign with the exceptions that the public key is used instead of the private key and that no signature is created but a signature, in a format as created by gcry_pk_sign, is passed to the function in sig.

The result is 0 for success (i.e. the data matches the signature), or an error code where the most relevant code is GCRY_ERR_BAD_SIGNATURE to indicate that the signature does not match the provided data.

6.4 Dedicated functions for elliptic curves.

The S-expression based interface is for certain operations on elliptic curves not optimal. Thus a few special functions are implemented to support common operations on curves with one of these assigned curve ids:

GCRY_ECC_CURVE25519

GCRY_ECC_CURVE448

Function: unsigned int gcry_ecc_get_algo_keylen (int curveid); ¶

Returns the length in bytes of a point on the curve with the id curveid. 0 is returned for curves which have no assigned id.

Function: gpg_error_t gcry_ecc_mul_point (int curveid, unsigned char *result, const unsigned char *scalar, const unsigned char *point) ¶

This function computes the scalar multiplication on the Montgomery form of the curve with id curveid. If point is NULL the base point of the curve is used. The caller needs to provide a large enough buffer for result and a valid scalar and point.

6.5 General public-key related Functions

A couple of utility functions are available to retrieve the length of the key, map algorithm identifiers and perform sanity checks:

Function: const char * gcry_pk_algo_name (int algo) ¶

Map the public key algorithm id algo to a string representation of the algorithm name. For unknown algorithms this functions returns the string "?". This function should not be used to test for the availability of an algorithm.

Function: int gcry_pk_map_name (const char *name) ¶

Map the algorithm name to a public key algorithm Id. Returns 0 if the algorithm name is not known.

Function: int gcry_pk_test_algo (int algo) ¶

Return 0 if the public key algorithm algo is available for use. Note that this is implemented as a macro.

Function: unsigned int gcry_pk_get_nbits (gcry_sexp_t key) ¶

Return what is commonly referred as the key length for the given public or private in key.

Function: unsigned char * gcry_pk_get_keygrip (gcry_sexp_t key, unsigned char *array) ¶

Return the so called "keygrip" which is the SHA-1 hash of the public key parameters expressed in a way depended on the algorithm. array must either provide space for 20 bytes or be NULL. In the latter case a newly allocated array of that size is returned. On success a pointer to the newly allocated space or to array is returned. NULL is returned to indicate an error which is most likely an unknown algorithm or one where a "keygrip" has not yet been defined. The function accepts public or secret keys in key.

Function: gcry_error_t gcry_pk_testkey (gcry_sexp_t key) ¶

Return zero if the private key key is ‘sane’, an error code otherwise. Note that it is not possible to check the ‘saneness’ of a public key.

Function: gcry_error_t gcry_pk_algo_info (int algo, int what, void *buffer, size_t *nbytes) ¶

Depending on the value of what return various information about the public key algorithm with the id algo. Note that the function returns -1 on error and the actual error code must be retrieved using the function gcry_errno. The currently defined values for what are:

GCRYCTL_TEST_ALGO:

Return 0 if the specified algorithm is available for use. buffer must be NULL, nbytes may be passed as NULL or point to a variable with the required usage of the algorithm. This may be 0 for "don’t care" or the bit-wise OR of these flags:

GCRY_PK_USAGE_SIGN

Algorithm is usable for signing.

GCRY_PK_USAGE_ENCR

Algorithm is usable for encryption.

Unless you need to test for the allowed usage, it is in general better to use the macro gcry_pk_test_algo instead.

GCRYCTL_GET_ALGO_USAGE:

Return the usage flags for the given algorithm. An invalid algorithm return 0. Disabled algorithms are ignored here because we want to know whether the algorithm is at all capable of a certain usage.

GCRYCTL_GET_ALGO_NPKEY

Return the number of elements the public key for algorithm algo consist of. Return 0 for an unknown algorithm.

GCRYCTL_GET_ALGO_NSKEY

Return the number of elements the private key for algorithm algo consist of. Note that this value is always larger than that of the public key. Return 0 for an unknown algorithm.

GCRYCTL_GET_ALGO_NSIGN

Return the number of elements a signature created with the algorithm algo consists of. Return 0 for an unknown algorithm or for an algorithm not capable of creating signatures.

GCRYCTL_GET_ALGO_NENCR

Return the number of elements a encrypted message created with the algorithm algo consists of. Return 0 for an unknown algorithm or for an algorithm not capable of encryption.

Please note that parameters not required should be passed as NULL.

Function: gcry_error_t gcry_pk_ctl (int cmd, void *buffer, size_t buflen) ¶

This is a general purpose function to perform certain control operations. cmd controls what is to be done. The return value is 0 for success or an error code. Currently supported values for cmd are:

GCRYCTL_DISABLE_ALGO

Disable the algorithm given as an algorithm id in buffer. buffer must point to an int variable with the algorithm id and buflen must have the value sizeof (int). This function is not thread safe and should thus be used before any other threads are started.

Libgcrypt also provides a function to generate public key pairs:

Function: gcry_error_t gcry_pk_genkey (gcry_sexp_t *r_key, gcry_sexp_t parms) ¶

This function create a new public key pair using information given in the S-expression parms and stores the private and the public key in one new S-expression at the address given by r_key. In case of an error, r_key is set to NULL. The return code is 0 for success or an error code otherwise.

Here is an example for parms to create an 2048 bit RSA key:

(genkey
  (rsa
    (nbits 4:2048)))

To create an Elgamal key, substitute "elg" for "rsa" and to create a DSA key use "dsa". Valid ranges for the key length depend on the algorithms; all commonly used key lengths are supported. Currently supported parameters are:

nbits

This is always required to specify the length of the key. The argument is a string with a number in C-notation. The value should be a multiple of 8. Note that the S-expression syntax requires that a number is prefixed with its string length; thus the 4: in the above example.

curve name

For ECC a named curve may be used instead of giving the number of requested bits. This allows to request a specific curve to override a default selection Libgcrypt would have taken if nbits has been given. The available names are listed with the description of the ECC public key parameters.

rsa-use-e value

This is only used with RSA to give a hint for the public exponent. The value will be used as a base to test for a usable exponent. Some values are special:

‘0’

Use a secure and fast value. This is currently the number 41.

‘1’

Use a value as required by some crypto policies. This is currently the number 65537.

‘2’

Reserved

‘> 2’

Use the given value.

If this parameter is not used, Libgcrypt uses for historic reasons 65537. Note that the value must fit into a 32 bit unsigned variable and that the usual C prefixes are considered (e.g. 017 gives 15).

qbits n

This is only meanigful for DSA keys. If it is given the DSA key is generated with a Q parameyer of size n bits. If it is not given or zero Q is deduced from NBITS in this way:

‘512 <= N <= 1024’

Q = 160

‘N = 2048’

Q = 224

‘N = 3072’

Q = 256

‘N = 7680’

Q = 384

‘N = 15360’

Q = 512

Note that in this case only the values for N, as given in the table, are allowed. When specifying Q all values of N in the range 512 to 15680 are valid as long as they are multiples of 8.

domain list

This is only meaningful for DLP algorithms. If specified keys are generated with domain parameters taken from this list. The exact format of this parameter depends on the actual algorithm. It is currently only implemented for DSA using this format:

(genkey
  (dsa
    (domain
      (p p-mpi)
      (q q-mpi)
      (g q-mpi))))

nbits and qbits may not be specified because they are derived from the domain parameters.

derive-parms list

This is currently only implemented for RSA and DSA keys. It is not allowed to use this together with a domain specification. If given, it is used to derive the keys using the given parameters.

If given for an RSA key the X9.31 key generation algorithm is used even if libgcrypt is not in FIPS mode. If given for a DSA key, the FIPS 186 algorithm is used even if libgcrypt is not in FIPS mode.

(genkey
  (rsa
    (nbits 4:1024)
    (rsa-use-e 1:3)
    (derive-parms
      (Xp1 #1A1916DDB29B4EB7EB6732E128#)
      (Xp2 #192E8AAC41C576C822D93EA433#)
      (Xp  #D8CD81F035EC57EFE822955149D3BFF70C53520D
            769D6D76646C7A792E16EBD89FE6FC5B605A6493
            39DFC925A86A4C6D150B71B9EEA02D68885F5009
            B98BD984#)
      (Xq1 #1A5CF72EE770DE50CB09ACCEA9#)
      (Xq2 #134E4CAA16D2350A21D775C404#)
      (Xq  #CC1092495D867E64065DEE3E7955F2EBC7D47A2D
            7C9953388F97DDDC3E1CA19C35CA659EDC2FC325
            6D29C2627479C086A699A49C4C9CEE7EF7BD1B34
            321DE34A#))))

(genkey
  (dsa
    (nbits 4:1024)
    (derive-parms
      (seed seed-mpi))))

flags flaglist

This is preferred way to define flags. flaglist may contain any number of flags. See above for a specification of these flags.

Here is an example on how to create a key using curve Ed25519 with the ECDSA signature algorithm. Note that the use of ECDSA with that curve is in general not recommended.

(genkey
  (ecc
    (flags transient-key)))

transient-key

use-x931

use-fips186

use-fips186-2

These are deprecated ways to set a flag with that name; see above for a description of each flag.

The key pair is returned in a format depending on the algorithm. Both private and public keys are returned in one container and may be accompanied by some miscellaneous information.

Here are two examples; the first for Elgamal and the second for elliptic curve key generation:

(key-data
  (public-key
    (elg
      (p p-mpi)
      (g g-mpi)
      (y y-mpi)))
  (private-key
    (elg
      (p p-mpi)
      (g g-mpi)
      (y y-mpi)
      (x x-mpi)))
  (misc-key-info
    (pm1-factors n1 n2 ... nn))

(key-data
  (public-key
    (ecc
      (curve Ed25519)
      (flags eddsa)
      (q q-value)))
  (private-key
    (ecc
      (curve Ed25519)
      (flags eddsa)
      (q q-value)
      (d d-value))))

As you can see, some of the information is duplicated, but this provides an easy way to extract either the public or the private key. Note that the order of the elements is not defined, e.g. the private key may be stored before the public key. n1 n2 ... nn is a list of prime numbers used to composite p-mpi; this is in general not a very useful information and only available if the key generation algorithm provides them.

Future versions of Libgcrypt will have extended versions of the public key interface which will take an additional context to allow for pre-computations, special operations, and other optimization. As a first step a new function is introduced to help using the ECC algorithms in new ways:

Function: gcry_error_t gcry_pubkey_get_sexp (gcry_sexp_t *r_sexp, int mode, gcry_ctx_t ctx) ¶

Return an S-expression representing the context ctx. Depending on the state of that context, the S-expression may either be a public key, a private key or any other object used with public key operations. On success 0 is returned and a new S-expression is stored at r_sexp; on error an error code is returned and NULL is stored at r_sexp. mode must be one of:

0

Decide what to return depending on the context. For example if the private key parameter is available a private key is returned, if not a public key is returned.

GCRY_PK_GET_PUBKEY

Return the public key even if the context has the private key parameter.

GCRY_PK_GET_SECKEY

Return the private key or the error GPG_ERR_NO_SECKEY if it is not possible.

As of now this function supports only certain ECC operations because a context object is right now only defined for ECC. Over time this function will be extended to cover more algorithms.

7.1 Available hash algorithms

GCRY_MD_NONE

This is not a real algorithm but used by some functions as an error return value. This constant is guaranteed to have the value 0.

GCRY_MD_SHA1

This is the SHA-1 algorithm which yields a message digest of 20 bytes. Note that SHA-1 begins to show some weaknesses and it is suggested to fade out its use if strong cryptographic properties are required.

GCRY_MD_RMD160

This is the 160 bit version of the RIPE message digest (RIPE-MD-160). Like SHA-1 it also yields a digest of 20 bytes. This algorithm share a lot of design properties with SHA-1 and thus it is advisable not to use it for new protocols.

GCRY_MD_MD5

This is the well known MD5 algorithm, which yields a message digest of 16 bytes. Note that the MD5 algorithm has severe weaknesses, for example it is easy to compute two messages yielding the same hash (collision attack). The use of this algorithm is only justified for non-cryptographic application.

GCRY_MD_MD4

This is the MD4 algorithm, which yields a message digest of 16 bytes. This algorithm has severe weaknesses and should not be used.

GCRY_MD_MD2

This is an reserved identifier for MD-2; there is no implementation yet. This algorithm has severe weaknesses and should not be used.

GCRY_MD_TIGER

This is the TIGER/192 algorithm which yields a message digest of 24 bytes. Actually this is a variant of TIGER with a different output print order as used by GnuPG up to version 1.3.2.

GCRY_MD_TIGER1

This is the TIGER variant as used by the NESSIE project. It uses the most commonly used output print order.

GCRY_MD_TIGER2

This is another variant of TIGER with a different padding scheme.

GCRY_MD_HAVAL

This is an reserved value for the HAVAL algorithm with 5 passes and 160 bit. It yields a message digest of 20 bytes. Note that there is no implementation yet available.

GCRY_MD_SHA224

This is the SHA-224 algorithm which yields a message digest of 28 bytes. See Change Notice 1 for FIPS 180-2 for the specification.

GCRY_MD_SHA256

This is the SHA-256 algorithm which yields a message digest of 32 bytes. See FIPS 180-2 for the specification.

GCRY_MD_SHA384

This is the SHA-384 algorithm which yields a message digest of 48 bytes. See FIPS 180-2 for the specification.

GCRY_MD_SHA512

This is the SHA-512 algorithm which yields a message digest of 64 bytes. See FIPS 180-2 for the specification.

GCRY_MD_SHA512_224

This is the SHA-512/224 algorithm which yields a message digest of 28 bytes. See FIPS 180-4 for the specification.

GCRY_MD_SHA512_256

This is the SHA-512/256 algorithm which yields a message digest of 32 bytes. See FIPS 180-4 for the specification.

GCRY_MD_SHA3_224

This is the SHA3-224 algorithm which yields a message digest of 28 bytes. See FIPS 202 for the specification.

GCRY_MD_SHA3_256

This is the SHA3-256 algorithm which yields a message digest of 32 bytes. See FIPS 202 for the specification.

GCRY_MD_SHA3_384

This is the SHA3-384 algorithm which yields a message digest of 48 bytes. See FIPS 202 for the specification.

GCRY_MD_SHA3_512

This is the SHA3-512 algorithm which yields a message digest of 64 bytes. See FIPS 202 for the specification.

GCRY_MD_SHAKE128

This is the SHAKE128 extendable-output function (XOF) algorithm with 128 bit security strength. See FIPS 202 for the specification.

GCRY_MD_SHAKE256

This is the SHAKE256 extendable-output function (XOF) algorithm with 256 bit security strength. See FIPS 202 for the specification.

GCRY_MD_CRC32

This is the ISO 3309 and ITU-T V.42 cyclic redundancy check. It yields an output of 4 bytes. Note that this is not a hash algorithm in the cryptographic sense.

GCRY_MD_CRC32_RFC1510

This is the above cyclic redundancy check function, as modified by RFC 1510. It yields an output of 4 bytes. Note that this is not a hash algorithm in the cryptographic sense.

GCRY_MD_CRC24_RFC2440

This is the OpenPGP cyclic redundancy check function. It yields an output of 3 bytes. Note that this is not a hash algorithm in the cryptographic sense.

GCRY_MD_WHIRLPOOL

This is the Whirlpool algorithm which yields a message digest of 64 bytes.

GCRY_MD_GOSTR3411_94

This is the hash algorithm described in GOST R 34.11-94 which yields a message digest of 32 bytes.

GCRY_MD_STRIBOG256

This is the 256-bit version of hash algorithm described in GOST R 34.11-2012 which yields a message digest of 32 bytes.

GCRY_MD_STRIBOG512

This is the 512-bit version of hash algorithm described in GOST R 34.11-2012 which yields a message digest of 64 bytes.

GCRY_MD_BLAKE2B_512

This is the BLAKE2b-512 algorithm which yields a message digest of 64 bytes. See RFC 7693 for the specification.

GCRY_MD_BLAKE2B_384

This is the BLAKE2b-384 algorithm which yields a message digest of 48 bytes. See RFC 7693 for the specification.

GCRY_MD_BLAKE2B_256

This is the BLAKE2b-256 algorithm which yields a message digest of 32 bytes. See RFC 7693 for the specification.

GCRY_MD_BLAKE2B_160

This is the BLAKE2b-160 algorithm which yields a message digest of 20 bytes. See RFC 7693 for the specification.

GCRY_MD_BLAKE2S_256

This is the BLAKE2s-256 algorithm which yields a message digest of 32 bytes. See RFC 7693 for the specification.

GCRY_MD_BLAKE2S_224

This is the BLAKE2s-224 algorithm which yields a message digest of 28 bytes. See RFC 7693 for the specification.

GCRY_MD_BLAKE2S_160

This is the BLAKE2s-160 algorithm which yields a message digest of 20 bytes. See RFC 7693 for the specification.

GCRY_MD_BLAKE2S_128

This is the BLAKE2s-128 algorithm which yields a message digest of 16 bytes. See RFC 7693 for the specification.

GCRY_MD_SM3

This is the SM3 algorithm which yields a message digest of 32 bytes.

7.2 Working with hash algorithms

To use most of these function it is necessary to create a context; this is done using:

Function: gcry_error_t gcry_md_open (gcry_md_hd_t *hd, int algo, unsigned int flags) ¶

Create a message digest object for algorithm algo. flags may be given as an bitwise OR of constants described below. algo may be given as 0 if the algorithms to use are later set using gcry_md_enable. hd is guaranteed to either receive a valid handle or NULL.

For a list of supported algorithms, see Available hash algorithms.

The flags allowed for mode are:

GCRY_MD_FLAG_SECURE

Allocate all buffers and the resulting digest in "secure memory". Use this is the hashed data is highly confidential.

GCRY_MD_FLAG_HMAC ¶

Turn the algorithm into a HMAC message authentication algorithm. This only works if just one algorithm is enabled for the handle and that algorithm is not an extendable-output function. Note that the function gcry_md_setkey must be used to set the MAC key. The size of the MAC is equal to the message digest of the underlying hash algorithm. If you want CBC message authentication codes based on a cipher, see Working with cipher handles.

GCRY_MD_FLAG_BUGEMU1 ¶

Versions of Libgcrypt before 1.6.0 had a bug in the Whirlpool code which led to a wrong result for certain input sizes and write patterns. Using this flag emulates that bug. This may for example be useful for applications which use Whirlpool as part of their key generation. It is strongly suggested to use this flag only if really needed and if possible to the data should be re-processed using the regular Whirlpool algorithm.

Note that this flag works for the entire hash context. If needed arises it may be used to enable bug emulation for other hash algorithms. Thus you should not use this flag for a multi-algorithm hash context.

You may use the function gcry_md_is_enabled to later check whether an algorithm has been enabled.

If you want to calculate several hash algorithms at the same time, you have to use the following function right after the gcry_md_open:

Function: gcry_error_t gcry_md_enable (gcry_md_hd_t h, int algo) ¶

Add the message digest algorithm algo to the digest object described by handle h. Duplicated enabling of algorithms is detected and ignored.

If the flag GCRY_MD_FLAG_HMAC was used, the key for the MAC must be set using the function:

Function: gcry_error_t gcry_md_setkey (gcry_md_hd_t h, const void *key, size_t keylen) ¶

For use with the HMAC feature or BLAKE2 keyed hash, set the MAC key to the value of key of length keylen bytes. For HMAC, there is no restriction on the length of the key. For keyed BLAKE2b hash, length of the key must be in the range 1 to 64 bytes. For keyed BLAKE2s hash, length of the key must be in the range 1 to 32 bytes.

After you are done with the hash calculation, you should release the resources by using:

Function: void gcry_md_close (gcry_md_hd_t h) ¶

Release all resources of hash context h. h should not be used after a call to this function. A NULL passed as h is ignored. The function also zeroises all sensitive information associated with this handle.

Often you have to do several hash operations using the same algorithm. To avoid the overhead of creating and releasing context, a reset function is provided:

Function: void gcry_md_reset (gcry_md_hd_t h) ¶

Reset the current context to its initial state. This is effectively identical to a close followed by an open and enabling all currently active algorithms.

Often it is necessary to start hashing some data and then continue to hash different data. To avoid hashing the same data several times (which might not even be possible if the data is received from a pipe), a snapshot of the current hash context can be taken and turned into a new context:

Function: gcry_error_t gcry_md_copy (gcry_md_hd_t *handle_dst, gcry_md_hd_t handle_src) ¶

Create a new digest object as an exact copy of the object described by handle handle_src and store it in handle_dst. The context is not reset and you can continue to hash data using this context and independently using the original context.

Now that we have prepared everything to calculate hashes, it is time to see how it is actually done. There are two ways for this, one to update the hash with a block of memory and one macro to update the hash by just one character. Both methods can be used on the same hash context.

Function: void gcry_md_write (gcry_md_hd_t h, const void *buffer, size_t length) ¶

Pass length bytes of the data in buffer to the digest object with handle h to update the digest values. This function should be used for large blocks of data. If this function is used after the context has been finalized, it will keep on pushing the data through the algorithm specific transform function and change the context; however the results are not meaningful and this feature is only available to mitigate timing attacks.

Function: void gcry_md_putc (gcry_md_hd_t h, int c) ¶

Pass the byte in c to the digest object with handle h to update the digest value. This is an efficient function, implemented as a macro to buffer the data before an actual update.

The semantics of the hash functions do not provide for reading out intermediate message digests because the calculation must be finalized first. This finalization may for example include the number of bytes hashed in the message digest or some padding.

Function: void gcry_md_final (gcry_md_hd_t h) ¶

Finalize the message digest calculation. This is not really needed because gcry_md_read and gcry_md_extract do this implicitly. After this has been done no further updates (by means of gcry_md_write or gcry_md_putc should be done; However, to mitigate timing attacks it is sometimes useful to keep on updating the context after having stored away the actual digest. Only the first call to this function has an effect. It is implemented as a macro.

The way to read out the calculated message digest is by using the function:

Function: unsigned char * gcry_md_read (gcry_md_hd_t h, int algo) ¶

gcry_md_read returns the message digest after finalizing the calculation. This function may be used as often as required but it will always return the same value for one handle. The returned message digest is allocated within the message context and therefore valid until the handle is released or reset-ed (using gcry_md_close or gcry_md_reset or it has been updated as a mitigation measure against timing attacks. algo may be given as 0 to return the only enabled message digest or it may specify one of the enabled algorithms. The function does return NULL if the requested algorithm has not been enabled.

The way to read output of extendable-output function is by using the function:

Function: gpg_err_code_t gcry_md_extract (gcry_md_hd_t h, int algo, void *buffer, size_t length) ¶

gcry_mac_read returns output from extendable-output function. This function may be used as often as required to generate more output byte stream from the algorithm. Function extracts the new output bytes to buffer of the length length. Buffer will be fully populated with new output. algo may be given as 0 to return the only enabled message digest or it may specify one of the enabled algorithms. The function does return non-zero value if the requested algorithm has not been enabled.

Because it is often necessary to get the message digest of blocks of memory, two fast convenience function are available for this task:

Function: gpg_err_code_t gcry_md_hash_buffers ( int algo, unsigned int flags, void *digest, const gcry_buffer_t *iov, int iovcnt ) ¶

gcry_md_hash_buffers is a shortcut function to calculate a message digest from several buffers. This function does not require a context and immediately returns the message digest of the data described by iov and iovcnt. digest must be allocated by the caller, large enough to hold the message digest yielded by the the specified algorithm algo. This required size may be obtained by using the function gcry_md_get_algo_dlen.

iov is an array of buffer descriptions with iovcnt items. The caller should zero out the structures in this array and for each array item set the fields .data to the address of the data to be hashed, .len to number of bytes to be hashed. If .off is also set, the data is taken starting at .off bytes from the begin of the buffer. The field .size is not used.

The only supported flag value for flags is GCRY_MD_FLAG_HMAC which turns this function into a HMAC function; the first item in iov is then used as the key.

On success the function returns 0 and stores the resulting hash or MAC at digest.

Function: void gcry_md_hash_buffer (int algo, void *digest, const void *buffer, size_t length); ¶

gcry_md_hash_buffer is a shortcut function to calculate a message digest of a buffer. This function does not require a context and immediately returns the message digest of the length bytes at buffer. digest must be allocated by the caller, large enough to hold the message digest yielded by the the specified algorithm algo. This required size may be obtained by using the function gcry_md_get_algo_dlen.

Note that in contrast to gcry_md_hash_buffers this function will abort the process if an unavailable algorithm is used.

Hash algorithms are identified by internal algorithm numbers (see gcry_md_open for a list). However, in most applications they are used by names, so two functions are available to map between string representations and hash algorithm identifiers.

Function: const char * gcry_md_algo_name (int algo) ¶

Map the digest algorithm id algo to a string representation of the algorithm name. For unknown algorithms this function returns the string "?". This function should not be used to test for the availability of an algorithm.

Function: int gcry_md_map_name (const char *name) ¶

Map the algorithm with name to a digest algorithm identifier. Returns 0 if the algorithm name is not known. Names representing ASN.1 object identifiers are recognized if the IETF dotted format is used and the OID is prefixed with either "oid." or "OID.". For a list of supported OIDs, see the source code at cipher/md.c. This function should not be used to test for the availability of an algorithm.

Function: gcry_error_t gcry_md_get_asnoid (int algo, void *buffer, size_t *length) ¶

Return an DER encoded ASN.1 OID for the algorithm algo in the user allocated buffer. length must point to variable with the available size of buffer and receives after return the actual size of the returned OID. The returned error code may be GPG_ERR_TOO_SHORT if the provided buffer is to short to receive the OID; it is possible to call the function with NULL for buffer to have it only return the required size. The function returns 0 on success.

To test whether an algorithm is actually available for use, the following macro should be used:

Function: gcry_error_t gcry_md_test_algo (int algo) ¶

The macro returns 0 if the algorithm algo is available for use.

If the length of a message digest is not known, it can be retrieved using the following function:

Function: unsigned int gcry_md_get_algo_dlen (int algo) ¶

Retrieve the length in bytes of the digest yielded by algorithm algo. This is often used prior to gcry_md_read to allocate sufficient memory for the digest.

In some situations it might be hard to remember the algorithm used for the ongoing hashing. The following function might be used to get that information:

Function: int gcry_md_get_algo (gcry_md_hd_t h) ¶

Retrieve the algorithm used with the handle h. Note that this does not work reliable if more than one algorithm is enabled in h.

The following macro might also be useful:

Function: int gcry_md_is_secure (gcry_md_hd_t h) ¶

This function returns true when the digest object h is allocated in "secure memory"; i.e. h was created with the GCRY_MD_FLAG_SECURE.

Function: int gcry_md_is_enabled (gcry_md_hd_t h, int algo) ¶

This function returns true when the algorithm algo has been enabled for the digest object h.

Tracking bugs related to hashing is often a cumbersome task which requires to add a lot of printf statements into the code. Libgcrypt provides an easy way to avoid this. The actual data hashed can be written to files on request.

Function: void gcry_md_debug (gcry_md_hd_t h, const char *suffix) ¶

Enable debugging for the digest object with handle h. This creates files named dbgmd-<n>.<string> while doing the actual hashing. suffix is the string part in the filename. The number is a counter incremented for each new hashing. The data in the file is the raw data as passed to gcry_md_write or gcry_md_putc. If NULL is used for suffix, the debugging is stopped and the file closed. This is only rarely required because gcry_md_close implicitly stops debugging.

8.1 Available MAC algorithms

GCRY_MAC_NONE

This is not a real algorithm but used by some functions as an error return value. This constant is guaranteed to have the value 0.

GCRY_MAC_HMAC_SHA256

This is keyed-hash message authentication code (HMAC) message authentication algorithm based on the SHA-256 hash algorithm.

GCRY_MAC_HMAC_SHA224

This is HMAC message authentication algorithm based on the SHA-224 hash algorithm.

GCRY_MAC_HMAC_SHA512

This is HMAC message authentication algorithm based on the SHA-512 hash algorithm.

GCRY_MAC_HMAC_SHA384

This is HMAC message authentication algorithm based on the SHA-384 hash algorithm.

GCRY_MAC_HMAC_SHA3_256

This is HMAC message authentication algorithm based on the SHA3-256 hash algorithm.

GCRY_MAC_HMAC_SHA3_224

This is HMAC message authentication algorithm based on the SHA3-224 hash algorithm.

GCRY_MAC_HMAC_SHA3_512

This is HMAC message authentication algorithm based on the SHA3-512 hash algorithm.

GCRY_MAC_HMAC_SHA3_384

This is HMAC message authentication algorithm based on the SHA3-384 hash algorithm.

GCRY_MAC_HMAC_SHA512_224

This is HMAC message authentication algorithm based on the SHA-512/224 hash algorithm.

GCRY_MAC_HMAC_SHA512_256

This is HMAC message authentication algorithm based on the SHA-512/256 hash algorithm.

GCRY_MAC_HMAC_SHA1

This is HMAC message authentication algorithm based on the SHA-1 hash algorithm.

GCRY_MAC_HMAC_MD5

This is HMAC message authentication algorithm based on the MD5 hash algorithm.

GCRY_MAC_HMAC_MD4

This is HMAC message authentication algorithm based on the MD4 hash algorithm.

GCRY_MAC_HMAC_RMD160

This is HMAC message authentication algorithm based on the RIPE-MD-160 hash algorithm.

GCRY_MAC_HMAC_WHIRLPOOL

This is HMAC message authentication algorithm based on the WHIRLPOOL hash algorithm.

GCRY_MAC_HMAC_GOSTR3411_94

This is HMAC message authentication algorithm based on the GOST R 34.11-94 hash algorithm.

GCRY_MAC_HMAC_STRIBOG256

This is HMAC message authentication algorithm based on the 256-bit hash algorithm described in GOST R 34.11-2012.

GCRY_MAC_HMAC_STRIBOG512

This is HMAC message authentication algorithm based on the 512-bit hash algorithm described in GOST R 34.11-2012.

GCRY_MAC_HMAC_BLAKE2B_512

This is HMAC message authentication algorithm based on the BLAKE2b-512 hash algorithm.

GCRY_MAC_HMAC_BLAKE2B_384

This is HMAC message authentication algorithm based on the BLAKE2b-384 hash algorithm.

GCRY_MAC_HMAC_BLAKE2B_256

This is HMAC message authentication algorithm based on the BLAKE2b-256 hash algorithm.

GCRY_MAC_HMAC_BLAKE2B_160

This is HMAC message authentication algorithm based on the BLAKE2b-160 hash algorithm.

GCRY_MAC_HMAC_BLAKE2S_256

This is HMAC message authentication algorithm based on the BLAKE2s-256 hash algorithm.

GCRY_MAC_HMAC_BLAKE2S_224

This is HMAC message authentication algorithm based on the BLAKE2s-224 hash algorithm.

GCRY_MAC_HMAC_BLAKE2S_160

This is HMAC message authentication algorithm based on the BLAKE2s-160 hash algorithm.

GCRY_MAC_HMAC_BLAKE2S_128

This is HMAC message authentication algorithm based on the BLAKE2s-128 hash algorithm.

GCRY_MAC_HMAC_SM3

This is HMAC message authentication algorithm based on the SM3 hash algorithm.

GCRY_MAC_CMAC_AES

This is CMAC (Cipher-based MAC) message authentication algorithm based on the AES block cipher algorithm.

GCRY_MAC_CMAC_3DES

This is CMAC message authentication algorithm based on the three-key EDE Triple-DES block cipher algorithm.

GCRY_MAC_CMAC_CAMELLIA

This is CMAC message authentication algorithm based on the Camellia block cipher algorithm.

GCRY_MAC_CMAC_CAST5

This is CMAC message authentication algorithm based on the CAST128-5 block cipher algorithm.

GCRY_MAC_CMAC_BLOWFISH

This is CMAC message authentication algorithm based on the Blowfish block cipher algorithm.

GCRY_MAC_CMAC_TWOFISH

This is CMAC message authentication algorithm based on the Twofish block cipher algorithm.

GCRY_MAC_CMAC_SERPENT

This is CMAC message authentication algorithm based on the Serpent block cipher algorithm.

GCRY_MAC_CMAC_SEED

This is CMAC message authentication algorithm based on the SEED block cipher algorithm.

GCRY_MAC_CMAC_RFC2268

This is CMAC message authentication algorithm based on the Ron’s Cipher 2 block cipher algorithm.

GCRY_MAC_CMAC_IDEA

This is CMAC message authentication algorithm based on the IDEA block cipher algorithm.

GCRY_MAC_CMAC_GOST28147

This is CMAC message authentication algorithm based on the GOST 28147-89 block cipher algorithm.

GCRY_MAC_CMAC_SM4

This is CMAC message authentication algorithm based on the SM4 block cipher algorithm.

GCRY_MAC_GMAC_AES

This is GMAC (GCM mode based MAC) message authentication algorithm based on the AES block cipher algorithm.

GCRY_MAC_GMAC_CAMELLIA

This is GMAC message authentication algorithm based on the Camellia block cipher algorithm.

GCRY_MAC_GMAC_TWOFISH

This is GMAC message authentication algorithm based on the Twofish block cipher algorithm.

GCRY_MAC_GMAC_SERPENT

This is GMAC message authentication algorithm based on the Serpent block cipher algorithm.

GCRY_MAC_GMAC_SEED

This is GMAC message authentication algorithm based on the SEED block cipher algorithm.

GCRY_MAC_POLY1305

This is plain Poly1305 message authentication algorithm, used with one-time key.

GCRY_MAC_POLY1305_AES

This is Poly1305-AES message authentication algorithm, used with key and one-time nonce.

GCRY_MAC_POLY1305_CAMELLIA

This is Poly1305-Camellia message authentication algorithm, used with key and one-time nonce.

GCRY_MAC_POLY1305_TWOFISH

This is Poly1305-Twofish message authentication algorithm, used with key and one-time nonce.

GCRY_MAC_POLY1305_SERPENT

This is Poly1305-Serpent message authentication algorithm, used with key and one-time nonce.

GCRY_MAC_POLY1305_SEED

This is Poly1305-SEED message authentication algorithm, used with key and one-time nonce.

GCRY_MAC_GOST28147_IMIT

This is MAC construction defined in GOST 28147-89 (see RFC 5830 Section 8).

8.2 Working with MAC algorithms

To use most of these function it is necessary to create a context; this is done using:

Function: gcry_error_t gcry_mac_open (gcry_mac_hd_t *hd, int algo, unsigned int flags, gcry_ctx_t ctx) ¶

Create a MAC object for algorithm algo. flags may be given as an bitwise OR of constants described below. hd is guaranteed to either receive a valid handle or NULL. ctx is context object to associate MAC object with. ctx maybe set to NULL.

For a list of supported algorithms, see Available MAC algorithms.

The flags allowed for mode are:

GCRY_MAC_FLAG_SECURE

Allocate all buffers and the resulting MAC in "secure memory". Use this if the MAC data is highly confidential.

In order to use a handle for performing MAC algorithm operations, a ‘key’ has to be set first:

Function: gcry_error_t gcry_mac_setkey (gcry_mac_hd_t h, const void *key, size_t keylen) ¶

Set the MAC key to the value of key of length keylen bytes. With HMAC algorithms, there is no restriction on the length of the key. With CMAC algorithms, the length of the key is restricted to those supported by the underlying block cipher.

GMAC algorithms and Poly1305-with-cipher algorithms need initialization vector to be set, which can be performed with function:

Function: gcry_error_t gcry_mac_setiv (gcry_mac_hd_t h, const void *iv, size_t ivlen) ¶

Set the IV to the value of iv of length ivlen bytes.

After you are done with the MAC calculation, you should release the resources by using:

Function: void gcry_mac_close (gcry_mac_hd_t h) ¶

Release all resources of MAC context h. h should not be used after a call to this function. A NULL passed as h is ignored. The function also clears all sensitive information associated with this handle.

Often you have to do several MAC operations using the same algorithm. To avoid the overhead of creating and releasing context, a reset function is provided:

Function: gcry_error_t gcry_mac_reset (gcry_mac_hd_t h) ¶

Reset the current context to its initial state. This is effectively identical to a close followed by an open and setting same key.

Note that gcry_mac_reset is implemented as a macro.

Now that we have prepared everything to calculate MAC, it is time to see how it is actually done.

Function: gcry_error_t gcry_mac_write (gcry_mac_hd_t h, const void *buffer, size_t length) ¶

Pass length bytes of the data in buffer to the MAC object with handle h to update the MAC values. If this function is used after the context has been finalized, it will keep on pushing the data through the algorithm specific transform function and thereby change the context; however the results are not meaningful and this feature is only available to mitigate timing attacks.

The way to read out the calculated MAC is by using the function:

Function: gcry_error_t gcry_mac_read (gcry_mac_hd_t h, void *buffer, size_t *length) ¶

gcry_mac_read returns the MAC after finalizing the calculation. Function copies the resulting MAC value to buffer of the length length. If length is larger than length of resulting MAC value, then length of MAC is returned through length.

To compare existing MAC value with recalculated MAC, one is to use the function:

Function: gcry_error_t gcry_mac_verify (gcry_mac_hd_t h, void *buffer, size_t length) ¶

gcry_mac_verify finalizes MAC calculation and compares result with length bytes of data in buffer. Error code GPG_ERR_CHECKSUM is returned if the MAC value in the buffer buffer does not match the MAC calculated in object h.

In some situations it might be hard to remember the algorithm used for the MAC calculation. The following function might be used to get that information:

Function: int gcry_mac_get_algo (gcry_mac_hd_t h) ¶

Retrieve the algorithm used with the handle h.

MAC algorithms are identified by internal algorithm numbers (see gcry_mac_open for a list). However, in most applications they are used by names, so two functions are available to map between string representations and MAC algorithm identifiers.

Function: const char * gcry_mac_algo_name (int algo) ¶

Map the MAC algorithm id algo to a string representation of the algorithm name. For unknown algorithms this function returns the string "?". This function should not be used to test for the availability of an algorithm.

Function: int gcry_mac_map_name (const char *name) ¶

Map the algorithm with name to a MAC algorithm identifier. Returns 0 if the algorithm name is not known. This function should not be used to test for the availability of an algorithm.

To test whether an algorithm is actually available for use, the following macro should be used:

Function: gcry_error_t gcry_mac_test_algo (int algo) ¶

The macro returns 0 if the MAC algorithm algo is available for use.

If the length of a message digest is not known, it can be retrieved using the following function:

Function: unsigned int gcry_mac_get_algo_maclen (int algo) ¶

Retrieve the length in bytes of the MAC yielded by algorithm algo. This is often used prior to gcry_mac_read to allocate sufficient memory for the MAC value. On error 0 is returned.

Function: unsigned int gcry_mac_get_algo_keylen (algo) ¶

This function returns length of the key for MAC algorithm algo. If the algorithm supports multiple key lengths, the default supported key length is returned. On error 0 is returned. The key length is returned as number of octets.

10.1 Quality of random numbers

Libgcypt offers random numbers of different quality levels:

Data type: gcry_random_level_t ¶

The constants for the random quality levels are of this enum type.

GCRY_WEAK_RANDOM

For all functions, except for gcry_mpi_randomize, this level maps to GCRY_STRONG_RANDOM. If you do not want this, consider using gcry_create_nonce.

GCRY_STRONG_RANDOM

Use this level for session keys and similar purposes.

GCRY_VERY_STRONG_RANDOM

Use this level for long term key material.

10.2 Retrieving random numbers

Function: void gcry_randomize (unsigned char *buffer, size_t length, enum gcry_random_level level) ¶

Fill buffer with length random bytes using a random quality as defined by level.

Function: void * gcry_random_bytes (size_t nbytes, enum gcry_random_level level) ¶

Convenience function to allocate a memory block consisting of nbytes fresh random bytes using a random quality as defined by level.

Function: void * gcry_random_bytes_secure (size_t nbytes, enum gcry_random_level level) ¶

Convenience function to allocate a memory block consisting of nbytes fresh random bytes using a random quality as defined by level. This function differs from gcry_random_bytes in that the returned buffer is allocated in a “secure” area of the memory.

Function: void gcry_create_nonce (unsigned char *buffer, size_t length) ¶

Fill buffer with length unpredictable bytes. This is commonly called a nonce and may also be used for initialization vectors and padding. This is an extra function nearly independent of the other random function for 3 reasons: It better protects the regular random generator’s internal state, provides better performance and does not drain the precious entropy pool.

11.1 Data types for S-expressions

Data type: gcry_sexp_t ¶

The gcry_sexp_t type describes an object with the Libgcrypt internal representation of an S-expression.

11.2 Working with S-expressions

There are several functions to create an Libgcrypt S-expression object from its external representation or from a string template. There is also a function to convert the internal representation back into one of the external formats:

Function: gcry_error_t gcry_sexp_new (gcry_sexp_t *r_sexp, const void *buffer, size_t length, int autodetect) ¶

This is the generic function to create an new S-expression object from its external representation in buffer of length bytes. On success the result is stored at the address given by r_sexp. With autodetect set to 0, the data in buffer is expected to be in canonized format, with autodetect set to 1 the parses any of the defined external formats. If buffer does not hold a valid S-expression an error code is returned and r_sexp set to NULL. Note that the caller is responsible for releasing the newly allocated S-expression using gcry_sexp_release.

Function: gcry_error_t gcry_sexp_create (gcry_sexp_t *r_sexp, void *buffer, size_t length, int autodetect, void (*freefnc)(void*)) ¶

This function is identical to gcry_sexp_new but has an extra argument freefnc, which, when not set to NULL, is expected to be a function to release the buffer; most likely the standard free function is used for this argument. This has the effect of transferring the ownership of buffer to the created object in r_sexp. The advantage of using this function is that Libgcrypt might decide to directly use the provided buffer and thus avoid extra copying.

Function: gcry_error_t gcry_sexp_sscan (gcry_sexp_t *r_sexp, size_t *erroff, const char *buffer, size_t length) ¶

This is another variant of the above functions. It behaves nearly identical but provides an erroff argument which will receive the offset into the buffer where the parsing stopped on error.

Function: gcry_error_t gcry_sexp_build (gcry_sexp_t *r_sexp, size_t *erroff, const char *format, ...) ¶

This function creates an internal S-expression from the string template format and stores it at the address of r_sexp. If there is a parsing error, the function returns an appropriate error code and stores the offset into format where the parsing stopped in erroff. The function supports a couple of printf-like formatting characters and expects arguments for some of these escape sequences right after format. The following format characters are defined:

‘%m’

The next argument is expected to be of type gcry_mpi_t and a copy of its value is inserted into the resulting S-expression. The MPI is stored as a signed integer.

‘%M’

The next argument is expected to be of type gcry_mpi_t and a copy of its value is inserted into the resulting S-expression. The MPI is stored as an unsigned integer.

‘%s’

The next argument is expected to be of type char * and that string is inserted into the resulting S-expression.

‘%d’

The next argument is expected to be of type int and its value is inserted into the resulting S-expression.

‘%u’

The next argument is expected to be of type unsigned int and its value is inserted into the resulting S-expression.

‘%b’

The next argument is expected to be of type int directly followed by an argument of type char *. This represents a buffer of given length to be inserted into the resulting S-expression.

‘%S’

The next argument is expected to be of type gcry_sexp_t and a copy of that S-expression is embedded in the resulting S-expression. The argument needs to be a regular S-expression, starting with a parenthesis.

No other format characters are defined and would return an error. Note that the format character ‘%%’ does not exists, because a percent sign is not a valid character in an S-expression.

Function: void gcry_sexp_release (gcry_sexp_t sexp) ¶

Release the S-expression object sexp. If the S-expression is stored in secure memory it explicitly zeroises that memory; note that this is done in addition to the zeroisation always done when freeing secure memory.

The next 2 functions are used to convert the internal representation back into a regular external S-expression format and to show the structure for debugging.

Function: size_t gcry_sexp_sprint (gcry_sexp_t sexp, int mode, char *buffer, size_t maxlength) ¶

Copies the S-expression object sexp into buffer using the format specified in mode. maxlength must be set to the allocated length of buffer. The function returns the actual length of valid bytes put into buffer or 0 if the provided buffer is too short. Passing NULL for buffer returns the required length for buffer. For convenience reasons an extra byte with value 0 is appended to the buffer.

The following formats are supported:

GCRYSEXP_FMT_DEFAULT

Returns a convenient external S-expression representation.

GCRYSEXP_FMT_CANON

Return the S-expression in canonical format.

GCRYSEXP_FMT_BASE64

Not currently supported.

GCRYSEXP_FMT_ADVANCED

Returns the S-expression in advanced format.

Function: void gcry_sexp_dump (gcry_sexp_t sexp) ¶

Dumps sexp in a format suitable for debugging to Libgcrypt’s logging stream.

Often canonical encoding is used in the external representation. The following function can be used to check for valid encoding and to learn the length of the S-expression.

Function: size_t gcry_sexp_canon_len (const unsigned char *buffer, size_t length, size_t *erroff, int *errcode) ¶

Scan the canonical encoded buffer with implicit length values and return the actual length this S-expression uses. For a valid S-expression it should never return 0. If length is not 0, the maximum length to scan is given; this can be used for syntax checks of data passed from outside. errcode and erroff may both be passed as NULL.

There are functions to parse S-expressions and retrieve elements:

Function: gcry_sexp_t gcry_sexp_find_token (const gcry_sexp_t list, const char *token, size_t toklen) ¶

Scan the S-expression for a sublist with a type (the car of the list) matching the string token. If toklen is not 0, the token is assumed to be raw memory of this length. The function returns a newly allocated S-expression consisting of the found sublist or NULL when not found.

Function: int gcry_sexp_length (const gcry_sexp_t list) ¶

Return the length of the list. For a valid S-expression this should be at least 1.

Function: gcry_sexp_t gcry_sexp_nth (const gcry_sexp_t list, int number) ¶

Create and return a new S-expression from the element with index number in list. Note that the first element has the index 0. If there is no such element, NULL is returned.

Function: gcry_sexp_t gcry_sexp_car (const gcry_sexp_t list) ¶

Create and return a new S-expression from the first element in list; this is called the "type" and should always exist per S-expression specification and in general be a string. NULL is returned in case of a problem.

Function: gcry_sexp_t gcry_sexp_cdr (const gcry_sexp_t list) ¶

Create and return a new list form all elements except for the first one. Note that this function may return an invalid S-expression because it is not guaranteed, that the type exists and is a string. However, for parsing a complex S-expression it might be useful for intermediate lists. Returns NULL on error.

Function: const char * gcry_sexp_nth_data (const gcry_sexp_t list, int number, size_t *datalen) ¶

This function is used to get data from a list. A pointer to the actual data with index number is returned and the length of this data will be stored to datalen. If there is no data at the given index or the index represents another list, NULL is returned. Caution: The returned pointer is valid as long as list is not modified or released.

Here is an example on how to extract and print the surname (Meier) from the S-expression ‘(Name Otto Meier (address Burgplatz 3))’:

size_t len;
const char *name;

name = gcry_sexp_nth_data (list, 2, &len);
printf ("my name is %.*s\n", (int)len, name);

Function: void * gcry_sexp_nth_buffer (const gcry_sexp_t list, int number, size_t *rlength) ¶

This function is used to get data from a list. A malloced buffer with the actual data at list index number is returned and the length of this buffer will be stored to rlength. If there is no data at the given index or the index represents another list, NULL is returned. The caller must release the result using gcry_free.

Here is an example on how to extract and print the CRC value from the S-expression ‘(hash crc32 #23ed00d7)’:

size_t len;
char *value;

value = gcry_sexp_nth_buffer (list, 2, &len);
if (value)
  fwrite (value, len, 1, stdout);
gcry_free (value);

Function: char * gcry_sexp_nth_string (gcry_sexp_t list, int number) ¶

This function is used to get and convert data from a list. The data is assumed to be a Nul terminated string. The caller must release this returned value using gcry_free. If there is no data at the given index, the index represents a list or the value can’t be converted to a string, NULL is returned.

Function: gcry_mpi_t gcry_sexp_nth_mpi (gcry_sexp_t list, int number, int mpifmt) ¶

This function is used to get and convert data from a list. This data is assumed to be an MPI stored in the format described by mpifmt and returned as a standard Libgcrypt MPI. The caller must release this returned value using gcry_mpi_release. If there is no data at the given index, the index represents a list or the value can’t be converted to an MPI, NULL is returned. If you use this function to parse results of a public key function, you most likely want to use GCRYMPI_FMT_USG.

Function: gpg_error_t gcry_sexp_extract_param ( gcry_sexp_t sexp, const char *path, const char *list, ...) ¶

Extract parameters from an S-expression using a list of parameter names. The names of these parameters are specified in LIST. White space between the parameter names are ignored. Some special characters and character sequences may be given to control the conversion:

‘+’

Switch to unsigned integer format (GCRYMPI_FMT_USG). This is the default mode.

‘-’

Switch to standard signed format (GCRYMPI_FMT_STD).

‘/’

Switch to opaque MPI format. The resulting MPIs may not be used for computations; see gcry_mpi_get_opaque for details.

‘&’

Switch to buffer descriptor mode. See below for details.

‘%s’

Switch to string mode. The expected argument is the address of a char * variable; the caller must release that value. If the parameter was marked optional and is not found, NULL is stored.

‘%#s’

Switch to multi string mode. The expected argument is the address of a char * variable; the caller must release that value. If the parameter was marked optional and is not found, NULL is stored. A multi string takes all values, assumes they are strings and concatenates them using a space as delimiter. In case a value is actually another list this is not further parsed but a () is inserted in place of that sublist.

‘%u’

Switch to unsigned integer mode. The expected argument is address of a unsigned int variable.

‘%lu’

Switch to unsigned long integer mode. The expected argument is address of a unsigned long variable.

‘%d’

Switch to signed integer mode. The expected argument is address of a int variable.

‘%ld’

Switch to signed long integer mode. The expected argument is address of a long variable.

‘%zu’

Switch to size_t mode. The expected argument is address of a size_t variable.

‘?’

If immediately following a parameter letter (no white space allowed), that parameter is considered optional.

In general parameter names are single letters. To use a string for a parameter name, enclose the name in single quotes.

Unless in buffer descriptor mode for each parameter name a pointer to an gcry_mpi_t variable is expected that must be set to NULL prior to invoking this function, and finally a NULL is expected. For example

  gcry_sexp_extract_param (key, NULL, "n/x+e d-'foo'",
                           &mpi_n, &mpi_x, &mpi_e, &mpi_d, &mpi_foo, NULL)

stores the parameter ’n’ from key as an unsigned MPI into mpi_n, the parameter ’x’ as an opaque MPI into mpi_x, the parameters ’e’ and ’d’ again as an unsigned MPI into mpi_e and mpi_d and finally the parameter ’foo’ as a signed MPI into mpi_foo.

path is an optional string used to locate a token. The exclamation mark separated tokens are used via gcry_sexp_find_token to find a start point inside the S-expression.

In buffer descriptor mode a pointer to a gcry_buffer_t descriptor is expected instead of a pointer to an MPI. The caller may use two different operation modes here: If the data field of the provided descriptor is NULL, the function allocates a new buffer and stores it at data; the other fields are set accordingly with off set to 0. If data is not NULL, the function assumes that the data, size, and off fields specify a buffer where to but the value of the respective parameter; on return the len field receives the number of bytes copied to that buffer; in case the buffer is too small, the function immediately returns with an error code (and len is set to 0).

The function returns 0 on success. On error an error code is returned, all passed MPIs that might have been allocated up to this point are deallocated and set to NULL, and all passed buffers are either truncated if the caller supplied the buffer, or deallocated if the function allocated the buffer.

12.1 Data types

Data type: gcry_mpi_t ¶

This type represents an object to hold an MPI.

Data type: gcry_mpi_point_t ¶

This type represents an object to hold a point for elliptic curve math.

12.2 Basic functions

To work with MPIs, storage must be allocated and released for the numbers. This can be done with one of these functions:

Function: gcry_mpi_t gcry_mpi_new (unsigned int nbits) ¶

Allocate a new MPI object, initialize it to 0 and initially allocate enough memory for a number of at least nbits. This pre-allocation is only a small performance issue and not actually necessary because Libgcrypt automatically re-allocates the required memory.

Function: gcry_mpi_t gcry_mpi_snew (unsigned int nbits) ¶

This is identical to gcry_mpi_new but allocates the MPI in the so called "secure memory" which in turn will take care that all derived values will also be stored in this "secure memory". Use this for highly confidential data like private key parameters.

Function: gcry_mpi_t gcry_mpi_copy (const gcry_mpi_t a) ¶

Create a new MPI as the exact copy of a but with the constant and immutable flags cleared.

Function: void gcry_mpi_release (gcry_mpi_t a) ¶

Release the MPI a and free all associated resources. Passing NULL is allowed and ignored. When a MPI stored in the "secure memory" is released, that memory gets wiped out immediately.

The simplest operations are used to assign a new value to an MPI:

Function: gcry_mpi_t gcry_mpi_set (gcry_mpi_t w, const gcry_mpi_t u) ¶

Assign the value of u to w and return w. If NULL is passed for w, a new MPI is allocated, set to the value of u and returned.

Function: gcry_mpi_t gcry_mpi_set_ui (gcry_mpi_t w, unsigned long u) ¶

Assign the value of u to w and return w. If NULL is passed for w, a new MPI is allocated, set to the value of u and returned. This function takes an unsigned int as type for u and thus it is only possible to set w to small values (usually up to the word size of the CPU).

Function: gcry_error_t gcry_mpi_get_ui (unsigned int *w, gcry_mpi_t u) ¶

If u is not negative and small enough to be stored in an unsigned int variable, store its value at w. If the value does not fit or is negative return GPG_ERR_ERANGE and do not change the value stored at w. Note that this function returns an unsigned int so that this value can immediately be used with the bit test functions. This is in contrast to the other "_ui" functions which allow for values up to an unsigned long.

Function: void gcry_mpi_swap (gcry_mpi_t a, gcry_mpi_t b) ¶

Swap the values of a and b.

Function: void gcry_mpi_snatch (gcry_mpi_t w, const gcry_mpi_t u) ¶

Set u into w and release u. If w is NULL only u will be released.

Function: void gcry_mpi_neg (gcry_mpi_t w, gcry_mpi_t u) ¶

Set the sign of w to the negative of u.

Function: void gcry_mpi_abs (gcry_mpi_t w) ¶

Clear the sign of w.

12.3 MPI formats

The following functions are used to convert between an external representation of an MPI and the internal one of Libgcrypt.

Function: gcry_error_t gcry_mpi_scan (gcry_mpi_t *r_mpi, enum gcry_mpi_format format, const unsigned char *buffer, size_t buflen, size_t *nscanned) ¶

Convert the external representation of an integer stored in buffer with a length of buflen into a newly created MPI returned which will be stored at the address of r_mpi. For certain formats the length argument is not required and should be passed as 0. A buflen larger than 16 MiByte will be rejected. After a successful operation the variable nscanned receives the number of bytes actually scanned unless nscanned was given as NULL. format describes the format of the MPI as stored in buffer:

GCRYMPI_FMT_STD

2-complement stored without a length header. Note that gcry_mpi_print stores a 0 as a string of zero length.

GCRYMPI_FMT_PGP

As used by OpenPGP (only defined as unsigned). This is basically GCRYMPI_FMT_STD with a 2 byte big endian length header. A length header indicating a length of more than 16384 is not allowed.

GCRYMPI_FMT_SSH

As used in the Secure Shell protocol. This is GCRYMPI_FMT_STD with a 4 byte big endian header.

GCRYMPI_FMT_HEX

Stored as a string with each byte of the MPI encoded as 2 hex digits. Negative numbers are prefix with a minus sign and in addition the high bit is always zero to make clear that an explicit sign ist used. When using this format, buflen must be zero.

GCRYMPI_FMT_USG

Simple unsigned integer.

Note that all of the above formats store the integer in big-endian format (MSB first).

Function: gcry_error_t gcry_mpi_print (enum gcry_mpi_format format, unsigned char *buffer, size_t buflen, size_t *nwritten, const gcry_mpi_t a) ¶

Convert the MPI a into an external representation described by format (see above) and store it in the provided buffer which has a usable length of at least the buflen bytes. If nwritten is not NULL, it will receive the number of bytes actually stored in buffer after a successful operation.

Function: gcry_error_t gcry_mpi_aprint (enum gcry_mpi_format format, unsigned char **buffer, size_t *nbytes, const gcry_mpi_t a) ¶

Convert the MPI a into an external representation described by format (see above) and store it in a newly allocated buffer which address will be stored in the variable buffer points to. The number of bytes stored in this buffer will be stored in the variable nbytes points to, unless nbytes is NULL.

Even if nbytes is zero, the function allocates at least one byte and store a zero there. Thus with formats GCRYMPI_FMT_STD and GCRYMPI_FMT_USG the caller may safely set a returned length of 0 to 1 to represent a zero as a 1 byte string.

Function: void gcry_mpi_dump (const gcry_mpi_t a) ¶

Dump the value of a in a format suitable for debugging to Libgcrypt’s logging stream. Note that one leading space but no trailing space or linefeed will be printed. It is okay to pass NULL for a.

12.4 Calculations

Basic arithmetic operations:

Function: void gcry_mpi_add (gcry_mpi_t w, gcry_mpi_t u, gcry_mpi_t v) ¶

w = u + v.

Function: void gcry_mpi_add_ui (gcry_mpi_t w, gcry_mpi_t u, unsigned long v) ¶

w = u + v. Note that v is an unsigned integer.

Function: void gcry_mpi_addm (gcry_mpi_t w, gcry_mpi_t u, gcry_mpi_t v, gcry_mpi_t m) ¶

w = u + v \bmod m.

Function: void gcry_mpi_sub (gcry_mpi_t w, gcry_mpi_t u, gcry_mpi_t v) ¶

w = u - v.

Function: void gcry_mpi_sub_ui (gcry_mpi_t w, gcry_mpi_t u, unsigned long v) ¶

w = u - v. v is an unsigned integer.

Function: void gcry_mpi_subm (gcry_mpi_t w, gcry_mpi_t u, gcry_mpi_t v, gcry_mpi_t m) ¶

w = u - v \bmod m.

Function: void gcry_mpi_mul (gcry_mpi_t w, gcry_mpi_t u, gcry_mpi_t v) ¶

w = u * v.

Function: void gcry_mpi_mul_ui (gcry_mpi_t w, gcry_mpi_t u, unsigned long v) ¶

w = u * v. v is an unsigned integer.

Function: void gcry_mpi_mulm (gcry_mpi_t w, gcry_mpi_t u, gcry_mpi_t v, gcry_mpi_t m) ¶

w = u * v \bmod m.

Function: void gcry_mpi_mul_2exp (gcry_mpi_t w, gcry_mpi_t u, unsigned long e) ¶

w = u * 2^e.

Function: void gcry_mpi_div (gcry_mpi_t q, gcry_mpi_t r, gcry_mpi_t dividend, gcry_mpi_t divisor, int round) ¶

q = dividend / divisor, r = dividend \bmod divisor. q and r may be passed as NULL. round is either negative for floored division (rounds towards the next lower integer) or zero for truncated division (rounds towards zero).

Function: void gcry_mpi_mod (gcry_mpi_t r, gcry_mpi_t dividend, gcry_mpi_t divisor) ¶

r = dividend \bmod divisor.

Function: void gcry_mpi_powm (gcry_mpi_t w, const gcry_mpi_t b, const gcry_mpi_t e, const gcry_mpi_t m) ¶

w = b^e \bmod m.

Function: int gcry_mpi_gcd (gcry_mpi_t g, gcry_mpi_t a, gcry_mpi_t b) ¶

Set g to the greatest common divisor of a and b. Return true if the g is 1.

Function: int gcry_mpi_invm (gcry_mpi_t x, gcry_mpi_t a, gcry_mpi_t m) ¶

Set x to the multiplicative inverse of a \bmod m. Return true if the inverse exists.

12.5 Comparisons

The next 2 functions are used to compare MPIs:

Function: int gcry_mpi_cmp (const gcry_mpi_t u, const gcry_mpi_t v) ¶

Compare the multi-precision-integers number u and v returning 0 for equality, a positive value for u > v and a negative for u < v. If both numbers are opaque values (cf, gcry_mpi_set_opaque) the comparison is done by checking the bit sizes using memcmp. If only one number is an opaque value, the opaque value is less than the other number.

Function: int gcry_mpi_cmp_ui (const gcry_mpi_t u, unsigned long v) ¶

Compare the multi-precision-integers number u with the unsigned integer v returning 0 for equality, a positive value for u > v and a negative for u < v.

Function: int gcry_mpi_is_neg (const gcry_mpi_t a) ¶

Return 1 if a is less than zero; return 0 if zero or positive.

12.6 Bit manipulations

There are a couple of functions to get information on arbitrary bits in an MPI and to set or clear them:

Function: unsigned int gcry_mpi_get_nbits (gcry_mpi_t a) ¶

Return the number of bits required to represent a.

Function: int gcry_mpi_test_bit (gcry_mpi_t a, unsigned int n) ¶

Return true if bit number n (counting from 0) is set in a.

Function: void gcry_mpi_set_bit (gcry_mpi_t a, unsigned int n) ¶

Set bit number n in a.

Function: void gcry_mpi_clear_bit (gcry_mpi_t a, unsigned int n) ¶

Clear bit number n in a.

Function: void gcry_mpi_set_highbit (gcry_mpi_t a, unsigned int n) ¶

Set bit number n in a and clear all bits greater than n.

Function: void gcry_mpi_clear_highbit (gcry_mpi_t a, unsigned int n) ¶

Clear bit number n in a and all bits greater than n.

Function: void gcry_mpi_rshift (gcry_mpi_t x, gcry_mpi_t a, unsigned int n) ¶

Shift the value of a by n bits to the right and store the result in x.

Function: void gcry_mpi_lshift (gcry_mpi_t x, gcry_mpi_t a, unsigned int n) ¶

Shift the value of a by n bits to the left and store the result in x.

12.7 EC functions

Libgcrypt provides an API to access low level functions used by its elliptic curve implementation. These functions allow to implement elliptic curve methods for which no explicit support is available.

Function: gcry_mpi_point_t gcry_mpi_point_new (unsigned int nbits) ¶

Allocate a new point object, initialize it to 0, and allocate enough memory for a points of at least nbits. This pre-allocation yields only a small performance win and is not really necessary because Libgcrypt automatically re-allocates the required memory. Using 0 for nbits is usually the right thing to do.

Function: void gcry_mpi_point_release (gcry_mpi_point_t point) ¶

Release point and free all associated resources. Passing NULL is allowed and ignored.

Function: gcry_mpi_point_t gcry_mpi_point_copy (gcry_mpi_point_t point) ¶

Allocate and return a new point object and initialize it with point. If point is NULL the function is identical to gcry_mpi_point_new(0).

Function: void gcry_mpi_point_get (gcry_mpi_t x, gcry_mpi_t y, gcry_mpi_t z, gcry_mpi_point_t point) ¶

Store the projective coordinates from point into the MPIs x, y, and z. If a coordinate is not required, NULL may be used for x, y, or z.

Function: void gcry_mpi_point_snatch_get (gcry_mpi_t x, gcry_mpi_t y, gcry_mpi_t z, gcry_mpi_point_t point) ¶

Store the projective coordinates from point into the MPIs x, y, and z. If a coordinate is not required, NULL may be used for x, y, or z. The object point is then released. Using this function instead of gcry_mpi_point_get and gcry_mpi_point_release has the advantage of avoiding some extra memory allocations and copies.

Function: gcry_mpi_point_t gcry_mpi_point_set ( gcry_mpi_point_t point, gcry_mpi_t x, gcry_mpi_t y, gcry_mpi_t z) ¶

Store the projective coordinates from x, y, and z into point. If a coordinate is given as NULL, the value 0 is used. If NULL is used for point a new point object is allocated and returned. Returns point or the newly allocated point object.

Function: gcry_mpi_point_t gcry_mpi_point_snatch_set ( gcry_mpi_point_t point, gcry_mpi_t x, gcry_mpi_t y, gcry_mpi_t z) ¶

Store the projective coordinates from x, y, and z into point. If a coordinate is given as NULL, the value 0 is used. If NULL is used for point a new point object is allocated and returned. The MPIs x, y, and z are released. Using this function instead of gcry_mpi_point_set and 3 calls to gcry_mpi_release has the advantage of avoiding some extra memory allocations and copies. Returns point or the newly allocated point object.

Function: gpg_error_t gcry_mpi_ec_new (gcry_ctx_t *r_ctx, gcry_sexp_t keyparam, const char *curvename) ¶

Allocate a new context for elliptic curve operations. If keyparam is given it specifies the parameters of the curve (see ecc_keyparam). If curvename is given in addition to keyparam and the key parameters do not include a named curve reference, the string curvename is used to fill in missing parameters. If only curvename is given, the context is initialized for this named curve.

If a parameter specifying a point (e.g. g or q) is not found, the parser looks for a non-encoded point by appending .x, .y, and .z to the parameter name and looking them all up to create a point. A parameter with the suffix .z is optional and defaults to 1.

On success the function returns 0 and stores the new context object at r_ctx; this object eventually needs to be released (see gcry_ctx_release). On error the function stores NULL at r_ctx and returns an error code.

Function: gcry_mpi_t gcry_mpi_ec_get_mpi ( const char *name, gcry_ctx_t ctx, int copy) ¶

Return the MPI with name from the context ctx. If not found NULL is returned. If the returned MPI may later be modified, it is suggested to pass 1 to copy, so that the function guarantees that a modifiable copy of the MPI is returned. If 0 is used for copy, this function may return a constant flagged MPI. In any case gcry_mpi_release needs to be called to release the result. For valid names ecc_keyparam. If the public key q is requested but only the private key d is available, q will be recomputed on the fly. If a point parameter is requested it is returned as an uncompressed encoded point unless these special names are used:

q@eddsa

Return an EdDSA style compressed point. This is only supported for Twisted Edwards curves.

Function: gcry_mpi_point_t gcry_mpi_ec_get_point ( const char *name, gcry_ctx_t ctx, int copy) ¶

Return the point with name from the context ctx. If not found NULL is returned. If the returned MPI may later be modified, it is suggested to pass 1 to copy, so that the function guarantees that a modifiable copy of the MPI is returned. If 0 is used for copy, this function may return a constant flagged point. In any case gcry_mpi_point_release needs to be called to release the result. If the public key q is requested but only the private key d is available, q will be recomputed on the fly.

Function: gpg_error_t gcry_mpi_ec_set_mpi ( const char *name, gcry_mpi_t newvalue, gcry_ctx_t ctx) ¶

Store the MPI newvalue at name into the context ctx. On success 0 is returned; on error an error code. Valid names are the MPI parameters of an elliptic curve (see ecc_keyparam).

Function: gpg_error_t gcry_mpi_ec_set_point ( const char *name, gcry_mpi_point_t newvalue, gcry_ctx_t ctx) ¶

Store the point newvalue at name into the context ctx. On success 0 is returned; on error an error code. Valid names are the point parameters of an elliptic curve (see ecc_keyparam).

Function: gpg_err_code_t gcry_mpi_ec_decode_point ( mpi_point_t result, gcry_mpi_t value, gcry_ctx_t ctx) ¶

Decode the point given as an MPI in value and store at result. To decide which encoding is used the function takes a context ctx which can be created with gcry_mpi_ec_new. If NULL is given for the context the function assumes a 0x04 prefixed uncompressed encoding. On error an error code is returned and result might be changed.

Function: int gcry_mpi_ec_get_affine ( gcry_mpi_t x, gcry_mpi_t y, gcry_mpi_point_t point, gcry_ctx_t ctx) ¶

Compute the affine coordinates from the projective coordinates in point and store them into x and y. If one coordinate is not required, NULL may be passed to x or y. ctx is the context object which has been created using gcry_mpi_ec_new. Returns 0 on success or not 0 if point is at infinity.

Note that you can use gcry_mpi_ec_set_point with the value GCRYMPI_CONST_ONE for z to convert affine coordinates back into projective coordinates.

Function: void gcry_mpi_ec_dup ( gcry_mpi_point_t w, gcry_mpi_point_t u, gcry_ctx_t ctx) ¶

Double the point u of the elliptic curve described by ctx and store the result into w.

Function: void gcry_mpi_ec_add ( gcry_mpi_point_t w, gcry_mpi_point_t u, gcry_mpi_point_t v, gcry_ctx_t ctx) ¶

Add the points u and v of the elliptic curve described by ctx and store the result into w.

Function: void gcry_mpi_ec_sub ( gcry_mpi_point_t w, gcry_mpi_point_t u, gcry_mpi_point_t v, gcry_ctx_t ctx) ¶

Subtracts the point v from the point u of the elliptic curve described by ctx and store the result into w. Only Twisted Edwards curves are supported for now.

Function: void gcry_mpi_ec_mul ( gcry_mpi_point_t w, gcry_mpi_t n, gcry_mpi_point_t u, gcry_ctx_t ctx) ¶

Multiply the point u of the elliptic curve described by ctx by n and store the result into w.

Function: int gcry_mpi_ec_curve_point ( gcry_mpi_point_t point, gcry_ctx_t ctx) ¶

Return true if point is on the elliptic curve described by ctx.

12.8 Miscellaneous

An MPI data type is allowed to be “misused” to store an arbitrary value. Two functions implement this kludge:

Function: gcry_mpi_t gcry_mpi_set_opaque (gcry_mpi_t a, void *p, unsigned int nbits) ¶

Store nbits of the value p points to in a and mark a as an opaque value (i.e. an value that can’t be used for any math calculation and is only used to store an arbitrary bit pattern in a). Ownership of p is taken by this function and thus the user may not use dereference the passed value anymore. It is required that them memory referenced by p has been allocated in a way that gcry_free is able to release it.

WARNING: Never use an opaque MPI for actual math operations. The only valid functions are gcry_mpi_get_opaque and gcry_mpi_release. Use gcry_mpi_scan to convert a string of arbitrary bytes into an MPI.

Function: gcry_mpi_t gcry_mpi_set_opaque_copy (gcry_mpi_t a, const void *p, unsigned int nbits) ¶

Same as gcry_mpi_set_opaque but ownership of p is not taken instead a copy of p is used.

Function: void * gcry_mpi_get_opaque (gcry_mpi_t a, unsigned int *nbits) ¶

Return a pointer to an opaque value stored in a and return its size in nbits. Note that the returned pointer is still owned by a and that the function should never be used for an non-opaque MPI.

Each MPI has an associated set of flags for special purposes. The currently defined flags are:

GCRYMPI_FLAG_SECURE

Setting this flag converts a into an MPI stored in "secure memory". Clearing this flag is not allowed.

GCRYMPI_FLAG_OPAQUE

This is an internal flag, indicating the an opaque valuue and not an integer is stored. This is an read-only flag; it may not be set or cleared.

GCRYMPI_FLAG_IMMUTABLE

If this flag is set, the MPI is marked as immutable. Setting or changing the value of that MPI is ignored and an error message is logged. The flag is sometimes useful for debugging.

GCRYMPI_FLAG_CONST

If this flag is set, the MPI is marked as a constant and as immutable Setting or changing the value of that MPI is ignored and an error message is logged. Such an MPI will never be deallocated and may thus be used without copying. Note that using gcry_mpi_copy will return a copy of that constant with this and the immutable flag cleared. A few commonly used constants are pre-defined and accessible using the macros GCRYMPI_CONST_ONE, GCRYMPI_CONST_TWO, GCRYMPI_CONST_THREE, GCRYMPI_CONST_FOUR, and GCRYMPI_CONST_EIGHT.

GCRYMPI_FLAG_USER1

GCRYMPI_FLAG_USER2

GCRYMPI_FLAG_USER3

GCRYMPI_FLAG_USER4

These flags are reserved for use by the application.

Function: void gcry_mpi_set_flag (gcry_mpi_t a, enum gcry_mpi_flag flag) ¶

Set the flag for the MPI a. The only allowed flags are GCRYMPI_FLAG_SECURE, GCRYMPI_FLAG_IMMUTABLE, and GCRYMPI_FLAG_CONST.

Function: void gcry_mpi_clear_flag (gcry_mpi_t a, enum gcry_mpi_flag flag) ¶

Clear flag for the multi-precision-integers a. The only allowed flag is GCRYMPI_FLAG_IMMUTABLE but only if GCRYMPI_FLAG_CONST is not set. If GCRYMPI_FLAG_CONST is set, clearing GCRYMPI_FLAG_IMMUTABLE will simply be ignored.

o

Function: int gcry_mpi_get_flag (gcry_mpi_t a, enum gcry_mpi_flag flag) ¶

Return true if flag is set for a.

To put a random value into an MPI, the following convenience function may be used:

Function: void gcry_mpi_randomize (gcry_mpi_t w, unsigned int nbits, enum gcry_random_level level) ¶

Set the multi-precision-integers w to a random non-negative number of nbits, using random data quality of level level. In case nbits is not a multiple of a byte, nbits is rounded up to the next byte boundary. When using a level of GCRY_WEAK_RANDOM this function makes use of gcry_create_nonce.

13.1 Generation

Function: gcry_error_t gcry_prime_generate (gcry_mpi_t *prime,unsigned int prime_bits, unsigned int factor_bits, gcry_mpi_t **factors, gcry_prime_check_func_t cb_func, void *cb_arg, gcry_random_level_t random_level, unsigned int flags) ¶

Generate a new prime number of prime_bits bits and store it in prime. If factor_bits is non-zero, one of the prime factors of (prime - 1) / 2 must be factor_bits bits long. If factors is non-zero, allocate a new, NULL-terminated array holding the prime factors and store it in factors. flags might be used to influence the prime number generation process.

Function: gcry_error_t gcry_prime_group_generator (gcry_mpi_t *r_g, gcry_mpi_t prime, gcry_mpi_t *factors, gcry_mpi_t start_g) ¶

Find a generator for prime where the factorization of (prime-1) is in the NULL terminated array factors. Return the generator as a newly allocated MPI in r_g. If start_g is not NULL, use this as the start for the search.

Function: void gcry_prime_release_factors (gcry_mpi_t *factors) ¶

Convenience function to release the factors array.

13.2 Checking

Function: gcry_error_t gcry_prime_check (gcry_mpi_t p, unsigned int flags) ¶

Check whether the number p is prime. Returns zero in case p is indeed a prime, returns GPG_ERR_NO_PRIME in case p is not a prime and a different error code in case something went horribly wrong.

14.1 Memory allocation

Function: void * gcry_malloc (size_t n) ¶

This function tries to allocate n bytes of memory. On success it returns a pointer to the memory area, in an out-of-core condition, it returns NULL.

Function: void * gcry_malloc_secure (size_t n) ¶

Like gcry_malloc, but uses secure memory.

Function: void * gcry_calloc (size_t n, size_t m) ¶

This function allocates a cleared block of memory (i.e. initialized with zero bytes) long enough to contain a vector of n elements, each of size m bytes. On success it returns a pointer to the memory block; in an out-of-core condition, it returns NULL.

Function: void * gcry_calloc_secure (size_t n, size_t m) ¶

Like gcry_calloc, but uses secure memory.

Function: void * gcry_realloc (void *p, size_t n) ¶

This function tries to resize the memory area pointed to by p to n bytes. On success it returns a pointer to the new memory area, in an out-of-core condition, it returns NULL. Depending on whether the memory pointed to by p is secure memory or not, gcry_realloc tries to use secure memory as well.

Function: void gcry_free (void *p) ¶

Release the memory area pointed to by p.

14.2 Context management

Some function make use of a context object. As of now there are only a few math functions. However, future versions of Libgcrypt may make more use of this context object.

Data type: gcry_ctx_t ¶

This type is used to refer to the general purpose context object.

Function: void gcry_ctx_release (gcry_ctx_t ctx) ¶

Release the context object ctx and all associated resources. A NULL passed as ctx is ignored.

14.3 Buffer description

To help hashing non-contiguous areas of memory a general purpose data type is defined:

Data type: gcry_buffer_t ¶

This type is a structure to describe a buffer. The user should make sure that this structure is initialized to zero. The available fields of this structure are:

.size

This is either 0 for no information available or indicates the allocated length of the buffer.

.off

This is the offset into the buffer.

.len

This is the valid length of the buffer starting at .off.

.data

This is the address of the buffer.

14.4 How to return Libgcrypt’s configuration.

Although GCRYCTL_PRINT_CONFIG can be used to print configuration options, it is sometimes necessary to check them in a program. This can be accomplished by using this function:

Function: char * gcry_get_config (int mode, const char *what) ¶

This function returns a malloced string with colon delimited configure options. With a value of 0 for mode this string resembles the output of GCRYCTL_PRINT_CONFIG. However, if what is not NULL, only the line where the first field (e.g. "cpu-arch") matches what is returned.

Other values than 0 for mode are not defined. The caller shall free the string using gcry_free. On error NULL is returned and ERRNO is set; if a value for WHAT is unknow ERRNO will be set to 0.

15.1 A HMAC-SHA-256 tool

This is a standalone HMAC-SHA-256 implementation used to compute an HMAC-SHA-256 message authentication code. The tool has originally been developed as a second implementation for Libgcrypt to allow comparing against the primary implementation and to be used for internal consistency checks. It should not be used for sensitive data because no mechanisms to clear the stack etc are used.

The code has been written in a highly portable manner and requires only a few standard definitions to be provided in a config.h file.

hmac256 is commonly invoked as

hmac256 "This is my key" foo.txt

This compute the MAC on the file foo.txt using the key given on the command line.

hmac256 understands these options:

--binary

Print the MAC as a binary string. The default is to print the MAC encoded has lower case hex digits.

--version

Print version of the program and exit.

17.1 Public-Key Architecture

Because public key cryptography is almost always used to process small amounts of data (hash values or session keys), the interface is not implemented using the open-use-close paradigm, but with single self-contained functions. Due to the wide variety of parameters required by different algorithms S-expressions, as flexible way to convey these parameters, are used. There is a set of helper functions to work with these S-expressions.

Aside of functions to register new algorithms, map algorithms names to algorithms identifiers and to lookup properties of a key, the following main functions are available:

gcry_pk_encrypt

Encrypt data using a public key.

gcry_pk_decrypt

Decrypt data using a private key.

gcry_pk_sign

Sign data using a private key.

gcry_pk_verify

Verify that a signature matches the data.

gcry_pk_testkey

Perform a consistency over a public or private key.

gcry_pk_genkey

Create a new public/private key pair.

All these functions lookup the module implementing the algorithm and pass the actual work to that module. The parsing of the S-expression input and the construction of S-expression for the return values is done by the high level code (cipher/pubkey.c). Thus the internal interface between the algorithm modules and the high level functions passes data in a custom format.

By default Libgcrypt uses a blinding technique for RSA decryption to mitigate real world timing attacks over a network: Instead of using the RSA decryption directly, a blinded value y = x r^{e} \bmod n is decrypted and the unblinded value x' = y' r^{-1} \bmod n returned. The blinding value r is a random value with the size of the modulus n and generated with GCRY_WEAK_RANDOM random level.

The algorithm used for RSA and DSA key generation depends on whether Libgcrypt is operated in standard or in FIPS mode. In standard mode an algorithm based on the Lim-Lee prime number generator is used. In FIPS mode RSA keys are generated as specified in ANSI X9.31 (1998) and DSA keys as specified in FIPS 186-2.

17.2 Symmetric Encryption Subsystem Architecture

The interface to work with symmetric encryption algorithms is made up of functions from the gcry_cipher_ name space. The implementation follows the open-use-close paradigm and uses registered algorithm modules for the actual work. Unless a module implements optimized cipher mode implementations, the high level code (cipher/cipher.c) implements the modes and calls the core algorithm functions to process each block.

The most important functions are:

gcry_cipher_open

Create a new instance to encrypt or decrypt using a specified algorithm and mode.

gcry_cipher_close

Release an instance.

gcry_cipher_setkey

Set a key to be used for encryption or decryption.

gcry_cipher_setiv

Set an initialization vector to be used for encryption or decryption.

gcry_cipher_encrypt

gcry_cipher_decrypt

Encrypt or decrypt data. These functions may be called with arbitrary amounts of data and as often as needed to encrypt or decrypt all data.

There is no strict alignment requirements for data, but the best performance can be archived if data is aligned to cacheline boundary.

There are also functions to query properties of algorithms or context, like block length, key length, map names or to enable features like padding methods.

17.3 Hashing and MACing Subsystem Architecture

The interface to work with message digests and CRC algorithms is made up of functions from the gcry_md_ name space. The implementation follows the open-use-close paradigm and uses registered algorithm modules for the actual work. Although CRC algorithms are not considered cryptographic hash algorithms, they share enough properties so that it makes sense to handle them in the same way. It is possible to use several algorithms at once with one context and thus compute them all on the same data.

The most important functions are:

gcry_md_open

Create a new message digest instance and optionally enable one algorithm. A flag may be used to turn the message digest algorithm into a HMAC algorithm.

gcry_md_enable

Enable an additional algorithm for the instance.

gcry_md_setkey

Set the key for the MAC.

gcry_md_write

Pass more data for computing the message digest to an instance.

There is no strict alignment requirements for data, but the best performance can be archived if data is aligned to cacheline boundary.

gcry_md_putc

Buffered version of gcry_md_write implemented as a macro.

gcry_md_read

Finalize the computation of the message digest or HMAC and return the result.

gcry_md_close

Release an instance

gcry_md_hash_buffer

Convenience function to directly compute a message digest over a memory buffer without the need to create an instance first.

There are also functions to query properties of algorithms or the instance, like enabled algorithms, digest length, map algorithm names. it is also possible to reset an instance or to copy the current state of an instance at any time. Debug functions to write the hashed data to files are available as well.

17.4 Multi-Precision-Integer Subsystem Architecture

The implementation of Libgcrypt’s big integer computation code is based on an old release of GNU Multi-Precision Library (GMP). The decision not to use the GMP library directly was due to stalled development at that time and due to security requirements which could not be provided by the code in GMP. As GMP does, Libgcrypt provides high performance assembler implementations of low level code for several CPUS to gain much better performance than with a generic C implementation.

Major features of Libgcrypt’s multi-precision-integer code compared to GMP are:

• Avoidance of stack based allocations to allow protection against swapping out of sensitive data and for easy zeroing of sensitive intermediate results.
• Optional use of secure memory and tracking of its use so that results are also put into secure memory.
• MPIs are identified by a handle (implemented as a pointer) to give better control over allocations and to augment them with extra properties like opaque data.
• Removal of unnecessary code to reduce complexity.
• Functions specialized for public key cryptography.

17.5 Prime-Number-Generator Subsystem Architecture

Libgcrypt provides an interface to its prime number generator. These functions make use of the internal prime number generator which is required for the generation for public key key pairs. The plain prime checking function is exported as well.