No OneTemporary
Actions

Size

444 KB

Subscribers

None

View Options

This file is larger than 256 KB, so syntax highlighting was skipped.

	diff --git a/doc/gcrypt.texi b/doc/gcrypt.texi
	index 327c15dc..8eab85ed 100644
	--- a/doc/gcrypt.texi
	+++ b/doc/gcrypt.texi
	@@ -1,7028 +1,7019 @@
	\input texinfo @c -- Texinfo --
	@c %**start of header
	@setfilename gcrypt.info
	@include version.texi
	@settitle The Libgcrypt Reference Manual
	@c Unify some of the indices.
	@syncodeindex tp fn
	@syncodeindex pg fn
	@c %**end of header
	@copying
	This manual is for Libgcrypt version @value{VERSION} and was last
	updated @value{UPDATED}. Libgcrypt is GNU's library of cryptographic
	building blocks.

	@noindent
	Copyright @copyright{} 2000, 2002, 2003, 2004, 2006, 2007, 2008, 2009, 2011, 2012 Free Software Foundation, Inc. @*
	Copyright @copyright{} 2012, 2013, 2016, 2017 g10 Code GmbH

	@quotation
	Permission is granted to copy, distribute and/or modify this document
	under the terms of the GNU General Public License as published by the
	Free Software Foundation; either version 2 of the License, or (at your
	option) any later version. The text of the license can be found in the
	section entitled ``GNU General Public License''.
	@end quotation
	@end copying

	@dircategory GNU Libraries
	@direntry
	* libgcrypt: (gcrypt). Cryptographic function library.
	@end direntry

	@c A couple of macros with no effect on texinfo
	@c but used by the yat2m processor.
	@macro manpage {a}
	@end macro
	@macro mansect {a}
	@end macro
	@macro manpause
	@end macro
	@macro mancont
	@end macro

	@c
	@c Printing stuff taken from gcc.
	@c
	@macro gnupgtabopt{body}
	@code{\body\}
	@end macro


	@c
	@c Titlepage
	@c
	@setchapternewpage odd
	@titlepage
	@title The Libgcrypt Reference Manual
	@subtitle Version @value{VERSION}
	@subtitle @value{UPDATED}
	@author Werner Koch (@email{wk@@gnupg.org})
	@author Moritz Schulte (@email{mo@@g10code.com})

	@page
	@vskip 0pt plus 1filll
	@insertcopying
	@end titlepage

	@ifnothtml
	@summarycontents
	@contents
	@page
	@end ifnothtml


	@ifnottex
	@node Top
	@top The Libgcrypt Library
	@insertcopying
	@end ifnottex


	@menu
	* Introduction:: What is Libgcrypt.
	* Preparation:: What you should do before using the library.
	* Generalities:: General library functions and data types.
	* Handler Functions:: Working with handler functions.
	* Symmetric cryptography:: How to use symmetric cryptography.
	* Public Key cryptography:: How to use public key cryptography.
	* Hashing:: How to use hash algorithms.
	* Message Authentication Codes:: How to use MAC algorithms.
	* Key Derivation:: How to derive keys from strings
	* Random Numbers:: How to work with random numbers.
	* S-expressions:: How to manage S-expressions.
	* MPI library:: How to work with multi-precision-integers.
	* Prime numbers:: How to use the Prime number related functions.
	* Utilities:: Utility functions.
	* Tools:: Utility tools.
	* Configuration:: Configuration files and environment variables.
	* Architecture:: How Libgcrypt works internally.

	Appendices

	* Self-Tests:: Description of the self-tests.
	* FIPS Mode:: Description of the FIPS mode.
	* Library Copying:: The GNU Lesser General Public License
	says how you can copy and share Libgcrypt.
	* Copying:: The GNU General Public License says how you
	can copy and share some parts of Libgcrypt.

	Indices

	* Figures and Tables:: Index of figures and tables.
	* Concept Index:: Index of concepts and programs.
	* Function and Data Index:: Index of functions, variables and data types.

	@end menu

	@ifhtml
	@page
	@summarycontents
	@contents
	@end ifhtml


	@c **********************************************************
	@c ***************** Introduction *********************
	@c **********************************************************
	@node Introduction
	@chapter Introduction

	Libgcrypt is a library providing cryptographic building blocks.

	@menu
	* Getting Started:: How to use this manual.
	* Features:: A glance at Libgcrypt's features.
	* Overview:: Overview about the library.
	@end menu

	@node Getting Started
	@section Getting Started

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

	@noindent
	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.


	@node Features
	@section Features

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

	@table @asis
	@item It's Free Software
	Anybody can use, modify, and redistribute it under the terms of the GNU
	Lesser General Public License (@pxref{Library Copying}). Note, that
	some parts (which are in general not needed by applications) are subject
	to the terms of the GNU General Public License (@pxref{Copying}); please
	see the README file of the distribution for of list of these parts.

	@item It encapsulates the low level cryptography
	Libgcrypt provides a high level interface to cryptographic
	building blocks using an extensible and flexible API.

	@end table

	@node Overview
	@section Overview

	@noindent
	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.

	@c **********************************************************
	@c ***************** Preparation **********************
	@c **********************************************************
	@node Preparation
	@chapter Preparation

	To use Libgcrypt, you have to perform some changes to your
	sources and the build system. The necessary changes are small and
	explained in the following sections. At the end of this chapter, it
	is described how the library is initialized, and how the requirements
	of the library are verified.

	@menu
	* Header:: What header file you need to include.
	* Building sources:: How to build sources using the library.
	* Building sources using Automake:: How to build sources with the help of Automake.
	* Initializing the library:: How to initialize the library.
	* Multi-Threading:: How Libgcrypt can be used in a MT environment.
	* Enabling FIPS mode:: How to enable the FIPS mode.
	* Hardware features:: How to disable hardware features.
	@end menu


	@node Header
	@section Header

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

	@example
	#include <gcrypt.h>
	@end example

	The name space of Libgcrypt is @code{gcry_*} for function
	and type names and @code{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 @code{gpg_*} as
	name space for function and type names and @code{GPG_*} for other
	symbols, including all the error codes.

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

	@table @code
	@item GCRYPT_NO_MPI_MACROS
	Do not define the shorthand macros @code{mpi_} for @code{gcry_mpi_}.

	@item GCRYPT_NO_DEPRECATED
	Do not include definitions for deprecated features. This is useful to
	make sure that no deprecated features are used.
	@end table

	@node Building sources
	@section 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 @option{-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 @command{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 @option{--cflags} option to @command{libgcrypt-config}. The following
	example shows how it can be used at the command line:

	@example
	gcc -c foo.c `libgcrypt-config --cflags`
	@end example

	Adding the output of @samp{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 @option{-L} option). For this, the option @option{--libs} to
	@command{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 @samp{-lgcrypt}
	option). The example shows how to link @file{foo.o} with the Libgcrypt
	library to a program @command{foo}.

	@example
	gcc -o foo foo.o `libgcrypt-config --libs`
	@end example

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

	@example
	gcc -o foo foo.c `libgcrypt-config --cflags --libs`
	@end example

	@node Building sources using Automake
	@section 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 @command{libgcrypt-config} script at all.
	Libgcrypt provides an extension to Automake that does all
	the work for you.

	@c A simple macro for optional variables.
	@macro ovar{varname}
	@r{[}@var{\varname\}@r{]}
	@end macro
	@defmac AM_PATH_LIBGCRYPT (@ovar{minimum-version}, @ovar{action-if-found}, @ovar{action-if-not-found})
	Check whether Libgcrypt (at least version
	@var{minimum-version}, if given) exists on the host system. If it is
	found, execute @var{action-if-found}, otherwise do
	@var{action-if-not-found}, if given.

	Additionally, the function defines @code{LIBGCRYPT_CFLAGS} to the
	flags needed for compilation of the program to find the
	@file{gcrypt.h} header file, and @code{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 @code{libgcrypt} is appended to the
	variable @code{gpg_config_script_warn}.

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

	@end defmac

	You can use the defined Autoconf variables like this in your
	@file{Makefile.am}:

	@example
	AM_CPPFLAGS = $(LIBGCRYPT_CFLAGS)
	LDADD = $(LIBGCRYPT_LIBS)
	@end example

	@node Initializing the library
	@section Initializing the library

	Before the library can be used, it must initialize itself. This is
	achieved by invoking the function @code{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.

	@deftypefun {const char } gcry_check_version (const char @var{req_version})

	The function @code{gcry_check_version} initializes some subsystems used
	by Libgcrypt and must be invoked before any other function in the
	library.
	@xref{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 @var{req_version}, if this value is not a null
	pointer.
	@end deftypefun

	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:

	@example
	/* 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);
	@end example


	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:

	@example
	/* 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);
	@}

	@anchor{sample-use-suspend-secmem}
	/* 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);

	@anchor{sample-use-resume-secmem}
	/* 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);
	@end example

	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:

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

	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.



	@node Multi-Threading
	@section Multi-Threading

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

	@itemize @bullet
	@item
	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.


	@item
	The function @code{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
	@code{gcry_check_version} before creating the other threads using
	Libgcrypt@footnote{At least this is true for POSIX threads,
	as @code{pthread_create} is a function that synchronizes memory with
	respects to other threads. There are many functions which have this
	property, a complete list can be found in POSIX, IEEE Std 1003.1-2003,
	Base Definitions, Issue 6, in the definition of the term ``Memory
	Synchronization''. For other thread packages, more relaxed or more
	strict rules may apply.}.

	@item
	Just like the function @code{gpg_strerror}, the function
	@code{gcry_strerror} is not thread safe. You have to use
	@code{gpg_strerror_r} instead.

	@end itemize


	@node Enabling FIPS mode
	@section How to enable the FIPS mode
	@cindex FIPS mode
	@cindex FIPS 140

	@anchor{enabling fips mode}
	Libgcrypt may be used in a FIPS 140-3 mode. Note, that this does not
	necessary mean that Libcgrypt is an appoved FIPS 140-3 module. Check the
	NIST database at @url{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:

	@itemize
	@item
	If the file @file{/proc/sys/crypto/fips_enabled} exists and contains a
	numeric value other than @code{0}, Libgcrypt is put into FIPS mode at
	initialization time. Obviously this works only on systems with a
	@code{proc} file system (i.e. GNU/Linux).

	@item
	If the file @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.

	@item
	If the application requests FIPS mode using the control command
	@code{GCRYCTL_FORCE_FIPS_MODE}. This must be done prior to any
	initialization (i.e. before @code{gcry_check_version}).

	@end itemize

	@node Hardware features
	@section How to disable hardware features
	@cindex hardware features

	@anchor{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

	@table @code
	@item padlock-rng
	@item padlock-aes
	@item padlock-sha
	@item padlock-mmul
	@item intel-cpu
	@item intel-fast-shld
	@item intel-bmi2
	@item intel-ssse3
	@item intel-sse4.1
	@item intel-pclmul
	@item intel-aesni
	@item intel-rdrand
	@item intel-avx
	@item intel-avx2
	@item intel-fast-vpgather
	@item intel-rdtsc
	@item intel-shaext
	@item intel-vaes-vpclmul
	@item arm-neon
	@item arm-aes
	@item arm-sha1
	@item arm-sha2
	@item arm-pmull
	@item ppc-vcrypto
	@item ppc-arch_3_00
	@item ppc-arch_2_07
	@item s390x-msa
	@item s390x-msa-4
	@item s390x-msa-8
	@item s390x-vx
	@end table

	To disable a feature for all processes using Libgcrypt 1.6 or newer,
	create the file @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 @code{gcry_check_version}.
	Example:@footnote{NB. Libgcrypt uses the RDRAND feature only as one
	source of entropy. A CPU with a broken RDRAND will thus not
	compromise of the random number generator}

	@example
	gcry_control (GCRYCTL_DISABLE_HWF, "intel-rdrand", NULL);
	@end example

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

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


	@c **********************************************************
	@c ***************** General **************************
	@c **********************************************************
	@node Generalities
	@chapter Generalities

	@menu
	* Controlling the library:: Controlling Libgcrypt's behavior.
	* Error Handling:: Error codes and such.
	@end menu

	@node Controlling the library
	@section Controlling the library

	@deftypefun gcry_error_t gcry_control (enum gcry_ctl_cmds @var{cmd}, ...)

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

	@table @code
	@item 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
	@code{gcry_check_version}. Note that the memory guard is NOT used
	when the user of the library has set his own memory management
	callbacks.

	@item GCRYCTL_ENABLE_QUICK_RANDOM; Arguments: none
	This command inhibits the use the very secure random quality level
	(@code{GCRY_VERY_STRONG_RANDOM}) and degrades all request down to
	@code{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.


	@item GCRYCTL_DUMP_RANDOM_STATS; Arguments: none
	This command dumps random number generator related statistics to the
	library's logging stream.

	@item GCRYCTL_DUMP_MEMORY_STATS; Arguments: none
	This command dumps memory management related statistics to the library's
	logging stream.

	@item GCRYCTL_DUMP_SECMEM_STATS; Arguments: none
	This command dumps secure memory management related statistics to the
	library's logging stream.

	@item 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 @code{GCRYCTL_DISABLE_SECMEM} right
	after initialization.

	@item GCRYCTL_DISABLE_SECMEM; Arguments: none
	This command disables the use of secure memory. In 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
	@code{gcry_check_version}.

	@item 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
	@code{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).

	@item 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
	@code{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.

	@item 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
	@var{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.

	@item 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 @code{gcry_xmalloc_secure} and
	@code{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.

	@item 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.

	@item GCRYCTL_DISABLE_SECMEM_WARN; Arguments: none
	Disable warning messages about problems with the secure memory
	subsystem. This command should be run right after
	@code{gcry_check_version}.

	@item GCRYCTL_SUSPEND_SECMEM_WARN; Arguments: none
	Postpone warning messages from the secure memory subsystem.
	@xref{sample-use-suspend-secmem,,the initialization example}, on how to
	use it.

	@item GCRYCTL_RESUME_SECMEM_WARN; Arguments: none
	Resume warning messages from the secure memory subsystem.
	@xref{sample-use-resume-secmem,,the initialization example}, on how to
	use it.

	@item 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 @code{gcry_check_version} and not
	later than the command GCRYCTL_INIT_SECMEM. Note that in FIPS mode the
	secure memory is always used.

	@item 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.


	@item 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 @code{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.

	@item 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.

	@item 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 @code{gcry_check_version}.

	@item 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
	@code{gcry_check_version}.

	@item 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 @code{gcry_check_version}.

	@item GCRYCTL_DISABLE_INTERNAL_LOCKING; Arguments: none
	This command does nothing. It exists only for backward compatibility.

	@item 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.

	@item GCRYCTL_INITIALIZATION_FINISHED; Arguments: none
	This command tells the library that the application has finished the
	initialization.

	@item GCRYCTL_INITIALIZATION_FINISHED_P; Arguments: none
	This command returns true if the command@*
	GCRYCTL_INITIALIZATION_FINISHED has already been run.

	@item GCRYCTL_SET_THREAD_CBS; Arguments: struct ath_ops *ath_ops
	This command is obsolete since version 1.6.

	@item GCRYCTL_FAST_POLL; Arguments: none
	Run a fast random poll.

	@item 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.

	@item 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 @var{stream}, the log
	system is used. This command may be used before the initialization has
	been finished but not before a @code{gcry_check_version}. Note that
	the macro @code{estream_t} can be used instead of @code{gpgrt_stream_t}.

	@item 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 @code{gcry_check_version}.

	@item 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 @code{gcry_check_version}. An application may use this command or
	the convenience macro below to check whether FIPS mode is actually
	active.

	@deftypefun int gcry_fips_mode_active (void)

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



	@item 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 @code{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.

	@item GCRYCTL_SET_ENFORCED_FIPS_FLAG; Arguments: none
	This command is obsolete and has no effect; do not use it.

	@item 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
	@code{gcry_check_version}. The available generators are:
	@table @code
	@item GCRY_RNG_TYPE_STANDARD
	A conservative standard generator based on the ``Continuously Seeded
	Pseudo Random Number Generator'' designed by Peter Gutmann.
	@item 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.
	@item 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.
	@end table
	The default is @code{GCRY_RNG_TYPE_STANDARD} unless FIPS mode as been
	enabled; in which case @code{GCRY_RNG_TYPE_FIPS} is used and locked
	against further changes.

	@item 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.


	@item 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.

	@item 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 @code{gcry_check_version}.

	@item 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.


	@end table

	@end deftypefun

	@c **********************************************************
	@c ***************** Errors **************************
	@c **********************************************************
	@node Error Handling
	@section 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 @code{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
	@code{libgpg-error}. They usually start with @code{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.

	@menu
	* Error Values:: The error value and what it means.
	* Error Sources:: A list of important error sources.
	* Error Codes:: A list of important error codes.
	* Error Strings:: How to get a descriptive string from a value.
	@end menu


	@node Error Values
	@subsection Error Values
	@cindex error values
	@cindex error codes
	@cindex error sources

	@deftp {Data type} {gcry_err_code_t}
	The @code{gcry_err_code_t} type is an alias for the
	@code{libgpg-error} type @code{gpg_err_code_t}. The error code
	indicates the type of an error, or the reason why an operation failed.

	A list of important error codes can be found in the next section.
	@end deftp

	@deftp {Data type} {gcry_err_source_t}
	The @code{gcry_err_source_t} type is an alias for the
	@code{libgpg-error} type @code{gpg_err_source_t}. The error source
	has not a precisely defined meaning. Sometimes it is the place where
	the error happened, sometimes it is the place where an error was
	encoded into an error value. Usually the error source will give an
	indication to where to look for the problem. This is not always true,
	but it is attempted to achieve this goal.

	A list of important error sources can be found in the next section.
	@end deftp

	@deftp {Data type} {gcry_error_t}
	The @code{gcry_error_t} type is an alias for the @code{libgpg-error}
	type @code{gpg_error_t}. An error value like this has always two
	components, an error code and an error source. Both together form the
	error value.

	Thus, the error value can not be directly compared against an error
	code, but the accessor functions described below must be used.
	However, it is guaranteed that only 0 is used to indicate success
	(@code{GPG_ERR_NO_ERROR}), and that in this case all other parts of
	the error value are set to 0, too.

	Note that in Libgcrypt, the error source is used purely for
	diagnostic purposes. Only the error code should be checked to test
	for a certain outcome of a function. The manual only documents the
	error code part of an error value. The error source is left
	unspecified and might be anything.
	@end deftp

	@deftypefun {gcry_err_code_t} gcry_err_code (@w{gcry_error_t @var{err}})
	The static inline function @code{gcry_err_code} returns the
	@code{gcry_err_code_t} component of the error value @var{err}. This
	function must be used to extract the error code from an error value in
	order to compare it with the @code{GPG_ERR_*} error code macros.
	@end deftypefun

	@deftypefun {gcry_err_source_t} gcry_err_source (@w{gcry_error_t @var{err}})
	The static inline function @code{gcry_err_source} returns the
	@code{gcry_err_source_t} component of the error value @var{err}. This
	function must be used to extract the error source from an error value in
	order to compare it with the @code{GPG_ERR_SOURCE_*} error source macros.
	@end deftypefun

	@deftypefun {gcry_error_t} gcry_err_make (@w{gcry_err_source_t @var{source}}, @w{gcry_err_code_t @var{code}})
	The static inline function @code{gcry_err_make} returns the error
	value consisting of the error source @var{source} and the error code
	@var{code}.

	This function can be used in callback functions to construct an error
	value to return it to the library.
	@end deftypefun

	@deftypefun {gcry_error_t} gcry_error (@w{gcry_err_code_t @var{code}})
	The static inline function @code{gcry_error} returns the error value
	consisting of the default error source and the error code @var{code}.

	For @acronym{GCRY} applications, the default error source is
	@code{GPG_ERR_SOURCE_USER_1}. You can define
	@code{GCRY_ERR_SOURCE_DEFAULT} before including @file{gcrypt.h} to
	change this default.

	This function can be used in callback functions to construct an error
	value to return it to the library.
	@end deftypefun

	The @code{libgpg-error} library provides error codes for all system
	error numbers it knows about. If @var{err} is an unknown error
	number, the error code @code{GPG_ERR_UNKNOWN_ERRNO} is used. The
	following functions can be used to construct error values from system
	errno numbers.

	@deftypefun {gcry_error_t} gcry_err_make_from_errno (@w{gcry_err_source_t @var{source}}, @w{int @var{err}})
	The function @code{gcry_err_make_from_errno} is like
	@code{gcry_err_make}, but it takes a system error like @code{errno}
	instead of a @code{gcry_err_code_t} error code.
	@end deftypefun

	@deftypefun {gcry_error_t} gcry_error_from_errno (@w{int @var{err}})
	The function @code{gcry_error_from_errno} is like @code{gcry_error},
	but it takes a system error like @code{errno} instead of a
	@code{gcry_err_code_t} error code.
	@end deftypefun

	Sometimes you might want to map system error numbers to error codes
	directly, or map an error code representing a system error back to the
	system error number. The following functions can be used to do that.

	@deftypefun {gcry_err_code_t} gcry_err_code_from_errno (@w{int @var{err}})
	The function @code{gcry_err_code_from_errno} returns the error code
	for the system error @var{err}. If @var{err} is not a known system
	error, the function returns @code{GPG_ERR_UNKNOWN_ERRNO}.
	@end deftypefun

	@deftypefun {int} gcry_err_code_to_errno (@w{gcry_err_code_t @var{err}})
	The function @code{gcry_err_code_to_errno} returns the system error
	for the error code @var{err}. If @var{err} is not an error code
	representing a system error, or if this system error is not defined on
	this system, the function returns @code{0}.
	@end deftypefun


	@node Error Sources
	@subsection Error Sources
	@cindex error codes, list of

	The library @code{libgpg-error} defines an error source for every
	component of the GnuPG system. The error source part of an error
	value is not well defined. As such it is mainly useful to improve the
	diagnostic error message for the user.

	If the error code part of an error value is @code{0}, the whole error
	value will be @code{0}. In this case the error source part is of
	course @code{GPG_ERR_SOURCE_UNKNOWN}.

	The list of error sources that might occur in applications using
	@acronym{Libgcrypt} is:

	@table @code
	@item GPG_ERR_SOURCE_UNKNOWN
	The error source is not known. The value of this error source is
	@code{0}.

	@item GPG_ERR_SOURCE_GPGME
	The error source is @acronym{GPGME} itself.

	@item GPG_ERR_SOURCE_GPG
	The error source is GnuPG, which is the crypto engine used for the
	OpenPGP protocol.

	@item GPG_ERR_SOURCE_GPGSM
	The error source is GPGSM, which is the crypto engine used for the
	OpenPGP protocol.

	@item GPG_ERR_SOURCE_GCRYPT
	The error source is @code{libgcrypt}, which is used by crypto engines
	to perform cryptographic operations.

	@item GPG_ERR_SOURCE_GPGAGENT
	The error source is @command{gpg-agent}, which is used by crypto
	engines to perform operations with the secret key.

	@item GPG_ERR_SOURCE_PINENTRY
	The error source is @command{pinentry}, which is used by
	@command{gpg-agent} to query the passphrase to unlock a secret key.

	@item GPG_ERR_SOURCE_SCD
	The error source is the SmartCard Daemon, which is used by
	@command{gpg-agent} to delegate operations with the secret key to a
	SmartCard.

	@item GPG_ERR_SOURCE_KEYBOX
	The error source is @code{libkbx}, a library used by the crypto
	engines to manage local keyrings.

	@item GPG_ERR_SOURCE_USER_1
	@item GPG_ERR_SOURCE_USER_2
	@item GPG_ERR_SOURCE_USER_3
	@item GPG_ERR_SOURCE_USER_4
	These error sources are not used by any GnuPG component and can be
	used by other software. For example, applications using
	Libgcrypt can use them to mark error values coming from callback
	handlers. Thus @code{GPG_ERR_SOURCE_USER_1} is the default for errors
	created with @code{gcry_error} and @code{gcry_error_from_errno},
	unless you define @code{GCRY_ERR_SOURCE_DEFAULT} before including
	@file{gcrypt.h}.
	@end table


	@node Error Codes
	@subsection Error Codes
	@cindex error codes, list of

	The library @code{libgpg-error} defines many error values. The
	following list includes the most important error codes.

	@table @code
	@item GPG_ERR_EOF
	This value indicates the end of a list, buffer or file.

	@item GPG_ERR_NO_ERROR
	This value indicates success. The value of this error code is
	@code{0}. Also, it is guaranteed that an error value made from the
	error code @code{0} will be @code{0} itself (as a whole). This means
	that the error source information is lost for this error code,
	however, as this error code indicates that no error occurred, this is
	generally not a problem.

	@item GPG_ERR_GENERAL
	This value means that something went wrong, but either there is not
	enough information about the problem to return a more useful error
	value, or there is no separate error value for this type of problem.

	@item GPG_ERR_ENOMEM
	This value means that an out-of-memory condition occurred.

	@item GPG_ERR_E...
	System errors are mapped to GPG_ERR_EFOO where FOO is the symbol for
	the system error.

	@item GPG_ERR_INV_VALUE
	This value means that some user provided data was out of range.

	@item GPG_ERR_UNUSABLE_PUBKEY
	This value means that some recipients for a message were invalid.

	@item GPG_ERR_UNUSABLE_SECKEY
	This value means that some signers were invalid.

	@item GPG_ERR_NO_DATA
	This value means that data was expected where no data was found.

	@item GPG_ERR_CONFLICT
	This value means that a conflict of some sort occurred.

	@item GPG_ERR_NOT_IMPLEMENTED
	This value indicates that the specific function (or operation) is not
	implemented. This error should never happen. It can only occur if
	you use certain values or configuration options which do not work,
	but for which we think that they should work at some later time.

	@item GPG_ERR_DECRYPT_FAILED
	This value indicates that a decryption operation was unsuccessful.

	@item GPG_ERR_WRONG_KEY_USAGE
	This value indicates that a key is not used appropriately.

	@item GPG_ERR_NO_SECKEY
	This value indicates that no secret key for the user ID is available.

	@item GPG_ERR_UNSUPPORTED_ALGORITHM
	This value means a verification failed because the cryptographic
	algorithm is not supported by the crypto backend.

	@item GPG_ERR_BAD_SIGNATURE
	This value means a verification failed because the signature is bad.

	@item GPG_ERR_NO_PUBKEY
	This value means a verification failed because the public key is not
	available.

	@item GPG_ERR_NOT_OPERATIONAL
	This value means that the library is not yet in state which allows to
	use this function. This error code is in particular returned if
	Libgcrypt is operated in FIPS mode and the internal state of the
	library does not yet or not anymore allow the use of a service.

	This error code is only available with newer libgpg-error versions, thus
	you might see ``invalid error code'' when passing this to
	@code{gpg_strerror}. The numeric value of this error code is 176.

	@item GPG_ERR_USER_1
	@item GPG_ERR_USER_2
	@item ...
	@item GPG_ERR_USER_16
	These error codes are not used by any GnuPG component and can be
	freely used by other software. Applications using Libgcrypt
	might use them to mark specific errors returned by callback handlers
	if no suitable error codes (including the system errors) for these
	errors exist already.
	@end table


	@node Error Strings
	@subsection Error Strings
	@cindex error values, printing of
	@cindex error codes, printing of
	@cindex error sources, printing of
	@cindex error strings

	@deftypefun {const char *} gcry_strerror (@w{gcry_error_t @var{err}})
	The function @code{gcry_strerror} returns a pointer to a statically
	allocated string containing a description of the error code contained
	in the error value @var{err}. This string can be used to output a
	diagnostic message to the user.
	@end deftypefun


	@deftypefun {const char *} gcry_strsource (@w{gcry_error_t @var{err}})
	The function @code{gcry_strsource} returns a pointer to a statically
	allocated string containing a description of the error source
	contained in the error value @var{err}. This string can be used to
	output a diagnostic message to the user.
	@end deftypefun

	The following example illustrates the use of the functions described
	above:

	@example
	@{
	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));
	@}
	@}
	@end example

	@c **********************************************************
	@c ***************** General **************************
	@c **********************************************************
	@node Handler Functions
	@chapter Handler Functions

	Libgcrypt makes it possible to install so called `handler functions',
	which get called by Libgcrypt in case of certain events.

	@menu
	* Progress handler:: Using a progress handler function.
	* Allocation handler:: Using special memory allocation functions.
	* Error handler:: Using error handler functions.
	* Logging handler:: Using a special logging function.
	@end menu

	@node Progress handler
	@section Progress handler

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

	@deftp {Data type} gcry_handler_progress_t
	Progress handler functions have to be of the type
	@code{gcry_handler_progress_t}, which is defined as:

	@code{void (gcry_handler_progress_t) (void , const char *, int, int, int)}
	@end deftp

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

	@deftypefun void gcry_set_progress_handler (gcry_handler_progress_t @var{cb}, void *@var{cb_data})

	This function installs @var{cb} as the `Progress handler' function.
	It may be used only during initialization. @var{cb} must be defined
	as follows:

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

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

	@table @var
	@item cb_data
	The argument provided in the call to @code{gcry_set_progress_handler}.
	@item what
	A string identifying the type of the progress output. The following
	values for @var{what} are defined:

	@table @code
	@item need_entropy
	Not enough entropy is available. @var{total} holds the number of
	required bytes.

	@item wait_dev_random
	Waiting to re-open a random device. @var{total} gives the number of
	seconds until the next try.

	@item primegen
	Values for @var{printchar}:
	@table @code
	@item \n
	Prime generated.
	@item !
	Need to refresh the pool of prime numbers.
	@item <, >
	Number of bits adjusted.
	@item ^
	Searching for a generator.
	@item .
	Fermat test on 10 candidates failed.
	@item :
	Restart with a new random value.
	@item +
	Rabin Miller test passed.
	@end table

	@end table

	@end table
	@end deftypefun

	@node Allocation handler
	@section 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:
	@deftp {Data type} gcry_handler_alloc_t
	This type is defined as: @code{void (gcry_handler_alloc_t) (size_t n)}.
	@end deftp
	@deftp {Data type} gcry_handler_secure_check_t
	This type is defined as: @code{int (gcry_handler_secure_check_t) (const void *)}.
	@end deftp
	@deftp {Data type} gcry_handler_realloc_t
	This type is defined as: @code{void (gcry_handler_realloc_t) (void *p, size_t n)}.
	@end deftp
	@deftp {Data type} gcry_handler_free_t
	This type is defined as: @code{void (gcry_handler_free_t) (void *)}.
	@end deftp

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

	@deftypefun void gcry_set_allocation_handler (gcry_handler_alloc_t @var{func_alloc}, gcry_handler_alloc_t @var{func_alloc_secure}, gcry_handler_secure_check_t @var{func_secure_check}, gcry_handler_realloc_t @var{func_realloc}, gcry_handler_free_t @var{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.


	@end deftypefun

	@node Error handler
	@section 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 @code{gcry_check_version}.

	@deftp {Data type} gcry_handler_no_mem_t
	This type is defined as: @code{int (gcry_handler_no_mem_t) (void , size_t, unsigned int)}
	@end deftp
	@deftypefun void gcry_set_outofcore_handler (gcry_handler_no_mem_t @var{func_no_mem}, void *@var{cb_data})
	This function registers @var{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 @var{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.
	@end deftypefun

	@deftp {Data type} gcry_handler_error_t
	This type is defined as: @code{void (gcry_handler_error_t) (void , int, const char *)}
	@end deftp

	@deftypefun void gcry_set_fatalerror_handler (gcry_handler_error_t @var{func_error}, void *@var{cb_data})
	This function registers @var{func_error} as `error handler',
	which means that it will be called in error conditions.
	@end deftypefun

	@node Logging handler
	@section Logging handler

	@deftp {Data type} gcry_handler_log_t
	This type is defined as: @code{void (gcry_handler_log_t) (void , int, const char *, va_list)}
	@end deftp

	@deftypefun void gcry_set_log_handler (gcry_handler_log_t @var{func_log}, void *@var{cb_data})
	This function registers @var{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
	@code{gcry_check_version}.
	@end deftypefun

	@c **********************************************************
	@c ***************** Ciphers **************************
	@c **********************************************************
	@c @include cipher-ref.texi
	@node Symmetric cryptography
	@chapter Symmetric cryptography

	The cipher functions are used for symmetrical cryptography,
	i.e. cryptography using a shared key. The programming model follows
	an open/process/close paradigm and is in that similar to other
	building blocks provided by Libgcrypt.

	@menu
	* Available ciphers:: List of ciphers supported by the library.
	* Available cipher modes:: List of cipher modes supported by the library.
	* Working with cipher handles:: How to perform operations related to cipher handles.
	* General cipher functions:: General cipher functions independent of cipher handles.
	@end menu

	@node Available ciphers
	@section Available ciphers

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

	@item GCRY_CIPHER_IDEA
	@cindex IDEA
	This is the IDEA algorithm.

	@item GCRY_CIPHER_3DES
	@cindex 3DES
	@cindex Triple-DES
	@cindex DES-EDE
	@cindex Digital Encryption Standard
	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.

	@item GCRY_CIPHER_CAST5
	@cindex CAST5
	CAST128-5 block cipher algorithm. The key size is 128 bits.

	@item GCRY_CIPHER_BLOWFISH
	@cindex Blowfish
	The blowfish algorithm. The supported key sizes are 8 to 576 bits in
	8 bit increments.

	@item GCRY_CIPHER_SAFER_SK128
	Reserved and not currently implemented.

	@item GCRY_CIPHER_DES_SK
	Reserved and not currently implemented.

	@item GCRY_CIPHER_AES
	@itemx GCRY_CIPHER_AES128
	@itemx GCRY_CIPHER_RIJNDAEL
	@itemx GCRY_CIPHER_RIJNDAEL128
	@cindex Rijndael
	@cindex AES
	@cindex Advanced Encryption Standard
	AES (Rijndael) with a 128 bit key.

	@item GCRY_CIPHER_AES192
	@itemx GCRY_CIPHER_RIJNDAEL192
	AES (Rijndael) with a 192 bit key.

	@item GCRY_CIPHER_AES256
	@itemx GCRY_CIPHER_RIJNDAEL256
	AES (Rijndael) with a 256 bit key.

	@item GCRY_CIPHER_TWOFISH
	@cindex Twofish
	The Twofish algorithm with a 256 bit key.

	@item GCRY_CIPHER_TWOFISH128
	The Twofish algorithm with a 128 bit key.

	@item GCRY_CIPHER_ARCFOUR
	@cindex Arcfour
	@cindex RC4
	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.

	@item GCRY_CIPHER_DES
	@cindex 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.

	@item GCRY_CIPHER_SERPENT128
	@itemx GCRY_CIPHER_SERPENT192
	@itemx GCRY_CIPHER_SERPENT256
	@cindex Serpent
	The Serpent cipher from the AES contest.

	@item GCRY_CIPHER_RFC2268_40
	@itemx GCRY_CIPHER_RFC2268_128
	@cindex rfc-2268
	@cindex RC2
	Ron's Cipher 2 in the 40 and 128 bit variants.

	@item GCRY_CIPHER_SEED
	@cindex Seed (cipher)
	A 128 bit cipher as described by RFC4269.

	@item GCRY_CIPHER_CAMELLIA128
	@itemx GCRY_CIPHER_CAMELLIA192
	@itemx GCRY_CIPHER_CAMELLIA256
	@cindex Camellia
	The Camellia cipher by NTT. See
	@uref{http://info.isl.ntt.co.jp/@/crypt/@/eng/@/camellia/@/specifications.html}.

	@item GCRY_CIPHER_SALSA20
	@cindex Salsa20
	This is the Salsa20 stream cipher.

	@item GCRY_CIPHER_SALSA20R12
	@cindex Salsa20/12
	This is the Salsa20/12 - reduced round version of Salsa20 stream cipher.

	@item GCRY_CIPHER_GOST28147
	@cindex GOST 28147-89
	The GOST 28147-89 cipher, defined in the respective GOST standard.
	Translation of this GOST into English is provided in the RFC-5830.

	@item GCRY_CIPHER_GOST28147_MESH
	@cindex GOST 28147-89 CryptoPro keymeshing
	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.

	@item GCRY_CIPHER_CHACHA20
	@cindex ChaCha20
	This is the ChaCha20 stream cipher.

	@item GCRY_CIPHER_SM4
	@cindex SM4 (cipher)
	A 128 bit cipher by the State Cryptography Administration
	of China (SCA). See
	@uref{https://tools.ietf.org/html/draft-ribose-cfrg-sm4-10}.

	@end table

	@node Available cipher modes
	@section Available cipher modes

	@table @code
	@item 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.

	@item GCRY_CIPHER_MODE_ECB
	@cindex ECB, Electronic Codebook mode
	Electronic Codebook mode.

	@item GCRY_CIPHER_MODE_CFB
	@item GCRY_CIPHER_MODE_CFB8
	@cindex CFB, Cipher Feedback mode
	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.

	@item GCRY_CIPHER_MODE_CBC
	@cindex CBC, Cipher Block Chaining mode
	Cipher Block Chaining mode.

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

	@item GCRY_CIPHER_MODE_OFB
	@cindex OFB, Output Feedback mode
	Output Feedback mode.

	@item GCRY_CIPHER_MODE_CTR
	@cindex CTR, Counter mode
	Counter mode.

	@item GCRY_CIPHER_MODE_AESWRAP
	@cindex AES-Wrap mode
	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 @code{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.

	@item GCRY_CIPHER_MODE_CCM
	@cindex CCM, Counter with CBC-MAC mode
	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.

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

	@item GCRY_CIPHER_MODE_POLY1305
	@cindex Poly1305 based AEAD mode with ChaCha20
	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.

	@item GCRY_CIPHER_MODE_OCB
	@cindex OCB, OCB3
	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 @code{gcry_cipher_ctl} using the command
	@code{GCRYCTL_SET_TAGLEN} and the address of an @code{int} variable
	set to 12 (for 96 bit) or 8 (for 64 bit) provided for the
	@code{buffer} argument and @code{sizeof(int)} for @code{buflen}.

	Note that the use of @code{gcry_cipher_final} is required.

	@item GCRY_CIPHER_MODE_XTS
	@cindex XTS, XTS mode
	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 (@code{GCRY_CIPHER_AES256}). The 128-bit tweak value
	is feed to XTS mode as little-endian byte array using
	@code{gcry_cipher_setiv} function. When encrypting or decrypting,
	full-sized data unit buffers needs to be passed to
	@code{gcry_cipher_encrypt} or @code{gcry_cipher_decrypt}. The tweak
	value is automatically incremented after each call of
	@code{gcry_cipher_encrypt} and @code{gcry_cipher_decrypt}.
	Auto-increment allows avoiding need of setting IV between processing
	of sequential data units.

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

	@item GCRY_CIPHER_MODE_SIV
	@cindex SIV, SIV mode
	Synthetic Initialization Vector (SIV) is an Authenticated Encryption
	with Associated Data (AEAD) block cipher mode, which is specified in
	RFC-5297. This mode works with block ciphers with block size of 128
	bits and uses tag length of 128 bits. Depending on how it is used,
	SIV achieves either the goal of deterministic authenticated encryption
	or the goal of nonce-based, misuse-resistant authenticated encryption.

	The SIV mode requires doubling key-length, for example, using 512-bit
	key with AES-256 (@code{GCRY_CIPHER_AES256}). Multiple AD instances can
	be passed to SIV mode with separate calls to
	@code{gcry_cipher_authenticate}. Nonce may be passed either through
	@code{gcry_cipher_setiv} or in the last call to
	@code{gcry_cipher_authenticate}. Note that use of @code{gcry_cipher_setiv}
	blocks any further calls to @code{gcry_cipher_authenticate} as nonce needs
	to be the last AD element with the SIV mode. When encrypting or decrypting,
	full-sized plaintext or ciphertext needs to be passed to
	@code{gcry_cipher_encrypt} or @code{gcry_cipher_decrypt}. Decryption tag
	needs to be given to SIV mode before decryption using
	@code{gcry_cipher_set_decryption_tag}.

	@item GCRY_CIPHER_MODE_GCM_SIV
	@cindex GCM-SIV, GCM-SIV mode, AES-GCM-SIV
	This mode implements is GCM-SIV Authenticated Encryption with
	Associated Data (AEAD) block cipher mode specified in RFC-5297
	(AES-GCM-SIV: Nonce Misuse-Resistant Authenticated Encryption).
	This implementations works with block ciphers with block size of
	128 bits and uses tag length of 128 bits. Supported key lengths
	by the mode are 128 bits and 256 bits. GCM-SIV is specified as
	nonce misuse resistant, so that it does not fail catastrophically
	if a nonce is repeated.

	When encrypting or decrypting, full-sized plaintext or ciphertext
	needs to be passed to @code{gcry_cipher_encrypt} or
	@code{gcry_cipher_decrypt}. Decryption tag needs to be given to
	GCM-SIV mode before decryption using @code{gcry_cipher_set_decryption_tag}.

	@end table

	@node Working with cipher handles
	@section 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:

	@deftypefun gcry_error_t gcry_cipher_open (gcry_cipher_hd_t *@var{hd}, int @var{algo}, int @var{mode}, unsigned int @var{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 @var{algo}. See
	@ref{Available ciphers}, for a list of supported ciphers and the
	according constants.

	Besides using the constants directly, the function
	@code{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 @var{mode}. See
	@ref{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
	(@code{GCRY_CIPHER_MODE_STREAM}) only works with stream ciphers.
	Poly1305 AEAD mode (@code{GCRY_CIPHER_MODE_POLY1305}) only works with
	ChaCha20 stream cipher. The block cipher modes
	(@code{GCRY_CIPHER_MODE_ECB}, @code{GCRY_CIPHER_MODE_CBC},
	@code{GCRY_CIPHER_MODE_CFB}, @code{GCRY_CIPHER_MODE_OFB},
	@code{GCRY_CIPHER_MODE_CTR} and @code{GCRY_CIPHER_MODE_EAX}) will work
	with any block cipher algorithm. GCM mode
	(@code{GCRY_CIPHER_MODE_GCM}), CCM mode (@code{GCRY_CIPHER_MODE_CCM}),
	OCB mode (@code{GCRY_CIPHER_MODE_OCB}), XTS mode
	(@code{GCRY_CIPHER_MODE_XTS}), SIV mode
	(@code{GCRY_CIPHER_MODE_SIV}) and GCM-SIV mode
	(@code{GCRY_CIPHER_MODE_GCM_SIV}) will only work with block cipher
	algorithms which have the block size of 16 bytes.

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

	@table @code
	@item GCRY_CIPHER_SECURE
	Make sure that all operations are allocated in secure memory. This is
	useful when the key material is highly confidential.
	@item GCRY_CIPHER_ENABLE_SYNC
	@cindex sync mode (OpenPGP)
	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 @code{gcry_cipher_sync}.
	@item GCRY_CIPHER_CBC_CTS
	@cindex cipher text stealing
	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).
	@item GCRY_CIPHER_CBC_MAC
	@cindex 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.
	@end table
	@end deftypefun

	Use the following function to release an existing handle:

	@deftypefun void gcry_cipher_close (gcry_cipher_hd_t @var{h})

	This function releases the context created by @code{gcry_cipher_open}.
	It also zeroises all sensitive information associated with this cipher
	handle.
	@end deftypefun

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

	@deftypefun gcry_error_t gcry_cipher_setkey (gcry_cipher_hd_t @var{h}, const void *@var{k}, size_t @var{l})

	Set the key @var{k} used for encryption or decryption in the context
	denoted by the handle @var{h}. The length @var{l} (in bytes) of the
	key @var{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.

	@end deftypefun

	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:

	@deftypefun gcry_error_t gcry_cipher_setiv (gcry_cipher_hd_t @var{h}, const void *@var{k}, size_t @var{l})

	Set the initialization vector used for encryption or decryption. The
	vector is passed as the buffer @var{K} of length @var{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.

	@end deftypefun

	@deftypefun gcry_error_t gcry_cipher_setctr (gcry_cipher_hd_t @var{h}, const void *@var{c}, size_t @var{l})

	Set the counter vector used for encryption or decryption. The counter
	is passed as the buffer @var{c} of length @var{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).
	@end deftypefun

	@deftypefun gcry_error_t gcry_cipher_reset (gcry_cipher_hd_t @var{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.
	@end deftypefun

	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:

	@deftypefun gcry_error_t gcry_cipher_authenticate (gcry_cipher_hd_t @var{h}, const void *@var{abuf}, size_t @var{abuflen})

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

	@end deftypefun

	@deftypefun {gcry_error_t} gcry_cipher_gettag @
	(@w{gcry_cipher_hd_t @var{h}}, @
	@w{void *@var{tag}}, @w{size_t @var{taglen}})

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

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

	@end deftypefun

	@deftypefun {gcry_error_t} gcry_cipher_checktag @
	(@w{gcry_cipher_hd_t @var{h}}, @
	@w{const void *@var{tag}}, @w{size_t @var{taglen}})

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

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

	@end deftypefun

	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.

	@deftypefun gcry_error_t gcry_cipher_encrypt (gcry_cipher_hd_t @var{h}, unsigned char {out}, size_t @var{outsize}, const unsigned char @var{in}, size_t @var{inlen})

	@code{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 @var{h}. There are 2
	ways to use the function: If @var{in} is passed as @code{NULL} and
	@var{inlen} is @code{0}, in-place encryption of the data in @var{out} of
	length @var{outsize} takes place. With @var{in} being not @code{NULL},
	@var{inlen} bytes are encrypted to the buffer @var{out} which must have
	at least a size of @var{inlen}. @var{outsize} must be set to the
	allocated size of @var{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 @code{gcry_cipher_final} is used
	before the final data chunk is passed to this function.

	The function returns @code{0} on success or an error code.
	@end deftypefun


	@deftypefun gcry_error_t gcry_cipher_decrypt (gcry_cipher_hd_t @var{h}, unsigned char {out}, size_t @var{outsize}, const unsigned char @var{in}, size_t @var{inlen})

	@code{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 @var{h}. There are 2
	ways to use the function: If @var{in} is passed as @code{NULL} and
	@var{inlen} is @code{0}, in-place decryption of the data in @var{out} or
	length @var{outsize} takes place. With @var{in} being not @code{NULL},
	@var{inlen} bytes are decrypted to the buffer @var{out} which must have
	at least a size of @var{inlen}. @var{outsize} must be set to the
	allocated size of @var{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 @code{gcry_cipher_final} is used
	before the final data chunk is passed to this function.

	The function returns @code{0} on success or an error code.
	@end deftypefun


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

	@deftypefun gcry_error_t gcry_cipher_final (gcry_cipher_hd_t @var{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.
	@end deftypefun


	The SIV mode and the GCM-SIV mode requires decryption tag to be input
	before decryption. This is done with:

	@deftypefun gcry_error_t gcry_cipher_set_decryption_tag (gcry_cipher_hd_t @var{h}, const void *@var{tag}, size_t @var{taglen})

	Set decryption tag for SIV or GCM-SIV mode decryption. This is
	implemented as a macro.
	@end deftypefun


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

	@deftypefun gcry_error_t gcry_cipher_sync (gcry_cipher_hd_t @var{h})

	Perform the OpenPGP sync operation on context @var{h}. Note that this
	is a no-op unless the context was created with the flag
	@code{GCRY_CIPHER_ENABLE_SYNC}
	@end deftypefun

	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:

	@deftypefun gcry_error_t gcry_cipher_ctl (gcry_cipher_hd_t @var{h}, int @var{cmd}, void *@var{buffer}, size_t @var{buflen})

	@code{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 @var{cmd} and the passed context
	handle @var{h}. Please see the comments in the source code
	(@code{src/global.c}) for details.
	@end deftypefun

	@deftypefun gcry_error_t gcry_cipher_info (gcry_cipher_hd_t @var{h}, @
	int @var{what}, void @var{buffer}, size_t @var{nbytes})

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

	@c begin constants for gcry_cipher_info
	@table @code

	@item 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. @var{buffer} must be
	given as NULL. On success the result is stored @var{nbytes}. The
	taglen is returned in bytes.

	@end table
	@c end constants for gcry_cipher_info

	@end deftypefun

	@node General cipher functions
	@section 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.

	@deftypefun gcry_error_t gcry_cipher_algo_info (int @var{algo}, int @var{what}, void @var{buffer}, size_t @var{nbytes})

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

	Here is a list of supported codes for @var{what}:

	@c begin constants for gcry_cipher_algo_info
	@table @code
	@item 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
	@var{nbytes}; @var{buffer} must be zero. Note that it is usually
	better to use the convenience function
	@code{gcry_cipher_get_algo_keylen}.

	@item GCRYCTL_GET_BLKLEN:
	Return the block length of the algorithm. The length is returned as a
	number of octets in @var{nbytes}; @var{buffer} must be zero. Note
	that it is usually better to use the convenience function
	@code{gcry_cipher_get_algo_blklen}.

	@item GCRYCTL_TEST_ALGO:
	Returns @code{0} when the specified algorithm is available for use.
	@var{buffer} and @var{nbytes} must be zero.

	@end table
	@c end constants for gcry_cipher_algo_info

	@end deftypefun
	@c end gcry_cipher_algo_info

	@deftypefun size_t gcry_cipher_get_algo_keylen (@var{algo})

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

	This is a convenience functions which should be preferred over
	@code{gcry_cipher_algo_info} because it allows for proper type
	checking.
	@end deftypefun
	@c end gcry_cipher_get_algo_keylen

	@deftypefun size_t gcry_cipher_get_algo_blklen (int @var{algo})

	This functions returns the block-length of the algorithm @var{algo}
	counted in octets. On error @code{0} is returned.

	This is a convenience functions which should be preferred over
	@code{gcry_cipher_algo_info} because it allows for proper type
	checking.
	@end deftypefun
	@c end gcry_cipher_get_algo_blklen


	@deftypefun {const char *} gcry_cipher_algo_name (int @var{algo})

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

	@deftypefun int gcry_cipher_map_name (const char *@var{name})

	@code{gcry_cipher_map_name} returns the algorithm identifier for the
	cipher algorithm described by the string @var{name}. If this algorithm
	is not available @code{0} is returned.
	@end deftypefun

	@deftypefun int gcry_cipher_mode_from_oid (const char *@var{string})

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


	@c **********************************************************
	@c ***************** Public Key ***********************
	@c **********************************************************
	@node Public Key cryptography
	@chapter Public Key cryptography

	Public key cryptography, also known as asymmetric cryptography, is an
	easy way for key management and to provide digital signatures.
	Libgcrypt provides two completely different interfaces to
	public key cryptography, this chapter explains the one based on
	S-expressions.

	@menu
	* Available algorithms:: Algorithms supported by the library.
	* Used S-expressions:: Introduction into the used S-expression.
	* Cryptographic Functions:: Functions for performing the cryptographic actions.
	* Dedicated ECC Functions:: Dedicated functions for elliptic curves.
	* General public-key related Functions:: General functions, not implementing any cryptography.
	@end menu

	@node Available algorithms
	@section Available algorithms

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

	@node Used S-expressions
	@section Used S-expressions

	Libgcrypt's API for asymmetric cryptography is based on data structures
	called S-expressions (see
	@uref{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.

	@noindent
	The following information are stored in S-expressions:

	@itemize
	@item keys

	@item plain text data

	@item encrypted data

	@item signatures

	@end itemize

	@noindent
	To describe how Libgcrypt expect keys, we use examples. Note that
	words in
	@ifnottex
	uppercase
	@end ifnottex
	@iftex
	italics
	@end iftex
	indicate parameters whereas lowercase words are literals.

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

	@menu
	* RSA key parameters:: Parameters used with an RSA key.
	* DSA key parameters:: Parameters used with a DSA key.
	* ECC key parameters:: Parameters used with ECC keys.
	@end menu

	@node RSA key parameters
	@subsection RSA key parameters

	@noindent
	An RSA private key is described by this S-expression:

	@example
	(private-key
	(rsa
	(n @var{n-mpi})
	(e @var{e-mpi})
	(d @var{d-mpi})
	(p @var{p-mpi})
	(q @var{q-mpi})
	(u @var{u-mpi})))
	@end example

	@noindent
	An RSA public key is described by this S-expression:

	@example
	(public-key
	(rsa
	(n @var{n-mpi})
	(e @var{e-mpi})))
	@end example


	@table @var
	@item n-mpi
	RSA public modulus @math{n}.
	@item e-mpi
	RSA public exponent @math{e}.
	@item d-mpi
	RSA secret exponent @math{d = e^{-1} \bmod (p-1)(q-1)}.
	@item p-mpi
	RSA secret prime @math{p}.
	@item q-mpi
	RSA secret prime @math{q} with @math{p < q}.
	@item u-mpi
	Multiplicative inverse @math{u = p^{-1} \bmod q}.
	@end table

	For signing and decryption the parameters @math{(p, q, u)} are optional
	but greatly improve the performance. Either all of these optional
	parameters must be given or none of them. They are mandatory for
	gcry_pk_testkey.

	Note that OpenSSL uses slighly different parameters: @math{q < p} and
	@math{u = q^{-1} \bmod p}. To use these parameters you will need to
	swap the values and recompute @math{u}. Here is example code to do this:

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




	@node DSA key parameters
	@subsection DSA key parameters

	@noindent
	A DSA private key is described by this S-expression:

	@example
	(private-key
	(dsa
	(p @var{p-mpi})
	(q @var{q-mpi})
	(g @var{g-mpi})
	(y @var{y-mpi})
	(x @var{x-mpi})))
	@end example

	@table @var
	@item p-mpi
	DSA prime @math{p}.
	@item q-mpi
	DSA group order @math{q} (which is a prime divisor of @math{p-1}).
	@item g-mpi
	DSA group generator @math{g}.
	@item y-mpi
	DSA public key value @math{y = g^x \bmod p}.
	@item x-mpi
	DSA secret exponent x.
	@end table

	The public key is similar with "private-key" replaced by "public-key"
	and no @var{x-mpi}.


	@node ECC key parameters
	@subsection ECC key parameters

	@anchor{ecc_keyparam}
	@noindent
	An ECC private key is described by this S-expression:

	@example
	(private-key
	(ecc
	(p @var{p-mpi})
	(a @var{a-mpi})
	(b @var{b-mpi})
	(g @var{g-point})
	(n @var{n-mpi})
	(q @var{q-point})
	(d @var{d-mpi})))
	@end example

	@table @var
	@item p-mpi
	Prime specifying the field @math{GF(p)}.
	@item a-mpi
	@itemx b-mpi
	The two coefficients of the Weierstrass equation @math{y^2 = x^3 + ax + b}
	@item g-point
	Base point @math{g}.
	@item n-mpi
	Order of @math{g}
	@item q-point
	The point representing the public key @math{Q = dG}.
	@item d-mpi
	The private key @math{d}
	@end table

	All point values are encoded in standard format; Libgcrypt does in
	general only support uncompressed points, thus the first byte needs to
	be @code{0x04}. However ``EdDSA'' describes its own compression
	scheme which is used by default; the non-standard first byte
	@code{0x40} may optionally be used to explicit flag the use of the
	algorithm’s native compression method.

	The public key is similar with "private-key" replaced by "public-key"
	and no @var{d-mpi}.

	If the domain parameters are well-known, the name of this curve may be
	used. For example

	@example
	(private-key
	(ecc
	(curve "NIST P-192")
	(q @var{q-point})
	(d @var{d-mpi})))
	@end example

	Note that @var{q-point} is optional for a private key. The
	@code{curve} parameter may be given in any case and is used to replace
	missing parameters.

	@noindent
	Currently implemented curves are:

	@table @code
	@item Curve25519
	@itemx X25519
	@itemx 1.3.6.1.4.1.3029.1.5.1
	@itemx 1.3.101.110
	The RFC-8410 255 bit curve, its RFC name, OpenPGP and RFC OIDs.

	@item X448
	@itemx 1.3.101.111
	The RFC-8410 448 bit curve and its RFC OID.

	@item Ed25519
	@itemx 1.3.6.1.4.1.11591.15.1
	@itemx 1.3.101.112
	The signing variant of the RFC-8410 255 bit curve, its OpenPGP and RFC OIDs.

	@item Ed448
	@itemx 1.3.101.113
	The signing variant of the RFC-8410 448 bit curve and its RFC OID.

	@item NIST P-192
	@itemx 1.2.840.10045.3.1.1
	@itemx nistp192
	@itemx prime192v1
	@itemx secp192r1
	The NIST 192 bit curve, its OID and aliases.

	@item NIST P-224
	@itemx 1.3.132.0.33
	@itemx nistp224
	@itemx secp224r1
	The NIST 224 bit curve, its OID and aliases.

	@item NIST P-256
	@itemx 1.2.840.10045.3.1.7
	@itemx nistp256
	@itemx prime256v1
	@itemx secp256r1
	The NIST 256 bit curve, its OID and aliases.

	@item NIST P-384
	@itemx 1.3.132.0.34
	@itemx nistp384
	@itemx secp384r1
	The NIST 384 bit curve, its OID and aliases.

	@item NIST P-521
	@itemx 1.3.132.0.35
	@itemx nistp521
	@itemx secp521r1
	The NIST 521 bit curve, its OID and aliases.

	@item brainpoolP160r1
	@itemx 1.3.36.3.3.2.8.1.1.1
	The Brainpool 160 bit curve and its OID.

	@item brainpoolP192r1
	@itemx 1.3.36.3.3.2.8.1.1.3
	The Brainpool 192 bit curve and its OID.

	@item brainpoolP224r1
	@itemx 1.3.36.3.3.2.8.1.1.5
	The Brainpool 224 bit curve and its OID.

	@item brainpoolP256r1
	@itemx 1.3.36.3.3.2.8.1.1.7
	The Brainpool 256 bit curve and its OID.

	@item brainpoolP320r1
	@itemx 1.3.36.3.3.2.8.1.1.9
	The Brainpool 320 bit curve and its OID.

	@item brainpoolP384r1
	@itemx 1.3.36.3.3.2.8.1.1.11
	The Brainpool 384 bit curve and its OID.

	@item brainpoolP512r1
	@itemx 1.3.36.3.3.2.8.1.1.13
	The Brainpool 512 bit curve and its OID.

	@end table
	As usual the OIDs may optionally be prefixed with the string @code{OID.}
	or @code{oid.}.


	@node Cryptographic Functions
	@section Cryptographic Functions

	@noindent
	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:

	@table @code

	@item comp
	@itemx nocomp
	@cindex comp
	@cindex nocomp
	If supported by the algorithm and curve the @code{comp} flag requests
	that points are returned in compact (compressed) representation. The
	@code{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
	@code{comp} is used with the ``EdDSA'' algorithm the key generation
	prefix the public key with a @code{0x40} byte.

	@item pkcs1
	@cindex PKCS1
	Use PKCS#1 block type 2 padding for encryption, block type 1 padding
	for signing.

	@item oaep
	@cindex OAEP
	Use RSA-OAEP padding for encryption.

	@item pss
	@cindex PSS
	Use RSA-PSS padding for signing.

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

	@item rfc6979
	@cindex RFC6979
	For DSA and ECDSA use a deterministic scheme for the k parameter.

	@item no-blinding
	@cindex 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.

	@item param
	@cindex 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.

	@item transient-key
	@cindex 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.

	@item no-keytest
	@cindex 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.

	@item use-x931
	@cindex X9.31
	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 @code{derive-parms} is given or Libgcrypt is
	in FIPS mode.

	@item use-fips186
	@cindex FIPS 186
	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
	@code{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.

	@item use-fips186-2
	@cindex FIPS 186-2
	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.

	@end table

	@noindent
	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:

	@deftypefun gcry_error_t gcry_pk_encrypt (@w{gcry_sexp_t *@var{r_ciph},} @w{gcry_sexp_t @var{data},} @w{gcry_sexp_t @var{pkey}})

	Obviously a public key must be provided for encryption. It is
	expected as an appropriate S-expression (see above) in @var{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.

	@noindent
	If you don't want to let Libgcrypt handle the padding, you must pass an
	appropriate MPI using this expression for @var{data}:

	@example
	(data
	(flags raw)
	(value @var{mpi}))
	@end example

	@noindent
	This has the same semantics as the old style MPI only way. @var{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 @var{data}:

	@example
	(data
	(flags pkcs1)
	(value @var{block}))
	@end example

	@noindent
	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 @var{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 @var{r_ciph}.
	The caller is responsible to release this value using
	@code{gcry_sexp_release}. In case of an error, an error code is
	returned and @var{r_ciph} will be set to @code{NULL}.

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

	@example
	(enc-val
	(rsa
	(a @var{a-mpi})))
	@end example

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

	@example
	(enc-val
	(elg
	(a @var{a-mpi})
	(b @var{b-mpi})))
	@end example

	@noindent
	Where @var{a-mpi} and @var{b-mpi} are MPIs with the result of the
	Elgamal encryption operation.
	@end deftypefun
	@c end gcry_pk_encrypt

	@deftypefun gcry_error_t gcry_pk_decrypt (@w{gcry_sexp_t *@var{r_plain},} @w{gcry_sexp_t @var{data},} @w{gcry_sexp_t @var{skey}})

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

	@example
	(enc-val
	(flags)
	(elg
	(a @var{a-mpi})
	(b @var{b-mpi})))
	@end example

	@noindent
	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:

	@example
	(flags @var{padding-method})
	@end example

	@noindent
	Currently @var{padding-method} is either @code{pkcs1} for PKCS#1 block
	type 2 padding, or @code{oaep} for RSA-OAEP padding.

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

	@example
	(value @var{plaintext})
	@end example
	@end deftypefun
	@c end gcry_pk_decrypt


	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:

	@deftypefun gcry_error_t gcry_pk_sign (@w{gcry_sexp_t *@var{r_sig},} @w{gcry_sexp_t @var{data},} @w{gcry_sexp_t @var{skey}})

	This function creates a digital signature for @var{data} using the
	private key @var{skey} and place it into the variable at the address of
	@var{r_sig}. @var{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:

	@example
	(data
	(flags pkcs1)
	(hash @var{hash-algo} @var{block}))
	@end example

	@noindent
	This example requests to sign the data in @var{block} after applying
	PKCS#1 block type 1 style padding. @var{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 @var{block} must
	match the size of that message digests; the function checks that this
	and other constraints are valid.

	@noindent
	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:

	@example
	(data
	(flags raw)
	(value @var{mpi}))
	@end example

	@noindent
	Here, the data to be signed is directly given as an @var{MPI}.

	@noindent
	For DSA the input data is expected in this format:

	@example
	(data
	(flags raw)
	(value @var{mpi}))
	@end example

	@noindent
	Here, the data to be signed is directly given as an @var{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.

	@example
	(data
	(flags rfc6979)
	(hash @var{hash-algo} @var{block}))
	@end example

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


	@noindent
	The signature is returned as a newly allocated S-expression in
	@var{r_sig} using this format for RSA:

	@example
	(sig-val
	(rsa
	(s @var{s-mpi})))
	@end example

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

	@example
	(sig-val
	(dsa
	(r @var{r-mpi})
	(s @var{s-mpi})))
	@end example

	Where @var{r-mpi} and @var{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:

	@example
	(data
	(flags eddsa)
	(hash-algo sha512)
	(value @var{message}))
	@end example

	Note that the @var{message} may be of any length; hashing is part of
	the algorithm. Using a large data block for @var{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 @var{message} is the
	@code{tbsCertificate} part and in CMS @var{message} is the
	@code{signedAttrs} part; see RFC-8410 and RFC-8419.


	@end deftypefun
	@c end gcry_pk_sign

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

	@deftypefun gcry_error_t gcry_pk_verify (@w{gcry_sexp_t @var{sig}}, @w{gcry_sexp_t @var{data}}, @w{gcry_sexp_t @var{pkey}})

	This is used to check whether the signature @var{sig} matches the
	@var{data}. The public key @var{pkey} must be provided to perform this
	verification. This function is similar in its parameters to
	@code{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 @code{gcry_pk_sign}, is passed to
	the function in @var{sig}.

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

	@end deftypefun
	@c end gcry_pk_verify

	Additionally, libgcrypt provides three functions for digital
	signatures. Those functions are useful when hashing computation
	should be closely combined with signature computation.

	@deftypefun gcry_error_t gcry_pk_hash_sign (@w{gcry_sexp_t @var{result},} @w{const char @var{data_tmpl},} @w{gcry_sexp_t @var{skey},} @w{gcry_md_hd_t @var{hd},} @w{gcry_ctx_t @var{ctx}})

	This function is a variant of @code{gcry_pk_sign} which takes as
	additional parameter @var{hd}, handle for hash, and an optional
	context @var{ctx}. @var{skey} is a private key in S-expression. The
	hash algorithm used by the handle needs to be enabled and input needs
	to be supplied beforehand. @var{data-tmpl} specifies a template to
	compose an S-expression to be signed. A template should include
	@code{"(hash %s %b)"} or @code{"(hash ALGONAME %b)"}. For the former
	case, "%s" is substituted by the string of algorithm of
	@code{gcry_md_get_algo (}@var{hd}@code{)} and when @code{gcry_md_read}
	is called, @code{ALGO=0} is used internally. For the latter case,
	hash algorithm by @code{ALGONAME} is used when @code{gcry_md_read} is
	called internally. The hash handle must not yet been finalized; the
	function takes a copy of the state and does a finalize on the copy.
	The last argument, @var{ctx}, may be used for supplying nonce
	externally. If no need, @var{ctx} should be passed as @code{NULL}.

	@end deftypefun
	@c end gcry_pk_hash_sign

	@deftypefun gcry_error_t gcry_pk_hash_verify (@w{gcry_sexp_t @var{sigval},} @w{const char *@var{data_tmpl}}, @w{gcry_sexp_t @var{pkey},} @w{gcry_md_hd_t @var{hd},} @w{gcry_ctx_t @var{ctx}})

	This function is a variant of @code{gcry_pk_verify} which takes as
	additional parameter @var{hd}, handle for hash, and an optional
	context @var{ctx}. @var{pkey} is a public key in S-expression. See
	@code{gcry_pk_hash_sign}, for the explanation of handle for hash,
	@var{data-tmpl} and @var{ctx}.

	@end deftypefun
	@c end gcry_pk_hash_verify

	@deftypefun gcry_error_t gcry_pk_random_override_new (@w{gcry_ctx_t @var{r_ctx},} @w{const unsigned char @var{p},} @w{size_t @var{len}})

	This function is used to allocate a new context for nonce, by memory
	area pointed to by @var{p} to @var{len} bytes. This context can be
	used when calling @code{gcry_pk_hash_sign} or
	@code{gcry_pk_hash_verify} to supply nonce externally, instead of
	generating internally.

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

	@end deftypefun
	@c end gcry_pk_random_override_new


	@node Dedicated ECC Functions
	@section Dedicated functions for elliptic curves.

	@noindent
	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:

	@table @code
	@item GCRY_ECC_CURVE25519
	@item GCRY_ECC_CURVE448
	@end table

	@deftypefun @w{unsigned int} gcry_ecc_get_algo_keylen (@w{int @var{curveid}});

	Returns the length in bytes of a point on the curve with the id
	@var{curveid}. 0 is returned for curves which have no assigned id.
	@end deftypefun


	@deftypefun gpg_error_t gcry_ecc_mul_point @
	(@w{int @var{curveid}}, @
	@w{unsigned char *@var{result}}, @
	@w{const unsigned char *@var{scalar}}, @
	@w{const unsigned char *@var{point}})

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


	@node General public-key related Functions
	@section General public-key related Functions

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

	@deftypefun {const char *} gcry_pk_algo_name (int @var{algo})

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

	@deftypefun int gcry_pk_map_name (const char *@var{name})

	Map the algorithm @var{name} to a public key algorithm Id. Returns 0 if
	the algorithm name is not known.
	@end deftypefun

	@deftypefun int gcry_pk_test_algo (int @var{algo})

	Return 0 if the public key algorithm @var{algo} is available for use.
	Note that this is implemented as a macro.
	@end deftypefun


	@deftypefun {unsigned int} gcry_pk_get_nbits (gcry_sexp_t @var{key})

	Return what is commonly referred as the key length for the given
	public or private in @var{key}.
	@end deftypefun

	@deftypefun {unsigned char } gcry_pk_get_keygrip (@w{gcry_sexp_t @var{key}}, @w{unsigned char @var{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. @var{array}
	must either provide space for 20 bytes or be @code{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 @var{array} is returned.
	@code{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 @var{key}.
	@end deftypefun

	@deftypefun gcry_error_t gcry_pk_testkey (gcry_sexp_t @var{key})

	Return zero if the private key @var{key} is `sane', an error code otherwise.
	Note that it is not possible to check the `saneness' of a public key.

	@end deftypefun


	@deftypefun gcry_error_t gcry_pk_algo_info (@w{int @var{algo}}, @w{int @var{what}}, @w{void @var{buffer}}, @w{size_t @var{nbytes}})

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

	@table @code
	@item GCRYCTL_TEST_ALGO:
	Return 0 if the specified algorithm is available for use.
	@var{buffer} must be @code{NULL}, @var{nbytes} may be passed as
	@code{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:

	@table @code
	@item GCRY_PK_USAGE_SIGN
	Algorithm is usable for signing.
	@item GCRY_PK_USAGE_ENCR
	Algorithm is usable for encryption.
	@end table

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

	@item 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.

	@item GCRYCTL_GET_ALGO_NPKEY
	Return the number of elements the public key for algorithm @var{algo}
	consist of. Return 0 for an unknown algorithm.

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

	@item GCRYCTL_GET_ALGO_NSIGN
	Return the number of elements a signature created with the algorithm
	@var{algo} consists of. Return 0 for an unknown algorithm or for an
	algorithm not capable of creating signatures.

	@item GCRYCTL_GET_ALGO_NENCR
	Return the number of elements a encrypted message created with the algorithm
	@var{algo} consists of. Return 0 for an unknown algorithm or for an
	algorithm not capable of encryption.
	@end table

	@noindent
	Please note that parameters not required should be passed as @code{NULL}.
	@end deftypefun
	@c end gcry_pk_algo_info


	@deftypefun gcry_error_t gcry_pk_ctl (@w{int @var{cmd}}, @w{void *@var{buffer}}, @w{size_t @var{buflen}})

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

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

	@end table
	@end deftypefun
	@c end gcry_pk_ctl

	@noindent
	Libgcrypt also provides a function to generate public key
	pairs:

	@deftypefun gcry_error_t gcry_pk_genkey (@w{gcry_sexp_t *@var{r_key}}, @w{gcry_sexp_t @var{parms}})

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

	@noindent
	Here is an example for @var{parms} to create an 2048 bit RSA key:

	@example
	(genkey
	(rsa
	(nbits 4:2048)))
	@end example

	@noindent
	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:

	@table @code
	@item 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 @code{4:} in the
	above example.

	@item curve @var{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 @code{nbits} has been
	given. The available names are listed with the description of the ECC
	public key parameters.

	@item rsa-use-e @var{value}
	This is only used with RSA to give a hint for the public exponent. The
	@var{value} will be used as a base to test for a usable exponent. Some
	values are special:

	@table @samp
	@item 0
	Use a secure and fast value. This is currently the number 41.
	@item 1
	Use a value as required by some crypto policies. This is currently
	the number 65537.
	@item 2
	Reserved
	@item > 2
	Use the given value.
	@end table

	@noindent
	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).


	@item qbits @var{n}
	This is only meanigful for DSA keys. If it is given the DSA key is
	generated with a Q parameyer of size @var{n} bits. If it is not given
	or zero Q is deduced from NBITS in this way:
	@table @samp
	@item 512 <= N <= 1024
	Q = 160
	@item N = 2048
	Q = 224
	@item N = 3072
	Q = 256
	@item N = 7680
	Q = 384
	@item N = 15360
	Q = 512
	@end table
	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.

	@item domain @var{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:

	@example
	(genkey
	(dsa
	(domain
	(p @var{p-mpi})
	(q @var{q-mpi})
	(g @var{q-mpi}))))
	@end example

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

	@item derive-parms @var{list}
	This is currently only implemented for RSA and DSA keys. It is not
	allowed to use this together with a @code{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.

	@example
	(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#))))
	@end example

	@example
	(genkey
	(dsa
	(nbits 4:1024)
	(derive-parms
	(seed @var{seed-mpi}))))
	@end example


	@item flags @var{flaglist}
	This is preferred way to define flags. @var{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.
	@example
	(genkey
	(ecc
	(flags transient-key)))
	@end example

	@item transient-key
	@itemx use-x931
	@itemx use-fips186
	@itemx use-fips186-2
	These are deprecated ways to set a flag with that name; see above for
	a description of each flag.


	@end table
	@c end table of parameters

	@noindent
	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.

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

	@example
	(key-data
	(public-key
	(elg
	(p @var{p-mpi})
	(g @var{g-mpi})
	(y @var{y-mpi})))
	(private-key
	(elg
	(p @var{p-mpi})
	(g @var{g-mpi})
	(y @var{y-mpi})
	(x @var{x-mpi})))
	(misc-key-info
	(pm1-factors @var{n1 n2 ... nn}))
	@end example

	@example
	(key-data
	(public-key
	(ecc
	(curve Ed25519)
	(flags eddsa)
	(q @var{q-value})))
	(private-key
	(ecc
	(curve Ed25519)
	(flags eddsa)
	(q @var{q-value})
	(d @var{d-value}))))
	@end example

	@noindent
	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. @var{n1 n2 ... nn} is a list
	of prime numbers used to composite @var{p-mpi}; this is in general not
	a very useful information and only available if the key generation
	algorithm provides them.
	@end deftypefun
	@c end gcry_pk_genkey


	@noindent
	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:

	@deftypefun gcry_error_t gcry_pubkey_get_sexp (@w{gcry_sexp_t *@var{r_sexp}}, @
	@w{int @var{mode}}, @w{gcry_ctx_t @var{ctx}})

	Return an S-expression representing the context @var{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 @var{r_sexp}; on error an error code is returned and NULL is stored
	at @var{r_sexp}. @var{mode} must be one of:

	@table @code
	@item 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.

	@item GCRY_PK_GET_PUBKEY
	Return the public key even if the context has the private key
	parameter.

	@item GCRY_PK_GET_SECKEY
	Return the private key or the error @code{GPG_ERR_NO_SECKEY} if it is
	not possible.
	@end table

	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.

	@end deftypefun
	@c end gcry_pubkey_get_sexp





	@c **********************************************************
	@c ***************** Hash Functions *******************
	@c **********************************************************
	@node Hashing
	@chapter Hashing

	Libgcrypt provides an easy and consistent to use interface for hashing.
	Hashing is buffered and several hash algorithms can be updated at once.
	It is possible to compute a HMAC using the same routines. The
	programming model follows an open/process/close paradigm and is in that
	similar to other building blocks provided by Libgcrypt.

	For convenience reasons, a few cyclic redundancy check value operations
	are also supported.

	@menu
	* Available hash algorithms:: List of hash algorithms supported by the library.
	* Working with hash algorithms:: List of functions related to hashing.
	@end menu

	@node Available hash algorithms
	@section Available hash algorithms

	@c begin table of hash algorithms
	@cindex SHA-1
	@cindex SHA-224, SHA-256, SHA-384, SHA-512, SHA-512/224, SHA-512/256
	@cindex SHA3-224, SHA3-256, SHA3-384, SHA3-512, SHAKE128, SHAKE256
	@cindex RIPE-MD-160
	@cindex MD2, MD4, MD5
	@cindex TIGER, TIGER1, TIGER2
	@cindex HAVAL
	@cindex SM3
	@cindex Whirlpool
	@cindex BLAKE2b-512, BLAKE2b-384, BLAKE2b-256, BLAKE2b-160
	@cindex BLAKE2s-256, BLAKE2s-224, BLAKE2s-160, BLAKE2s-128
	@cindex CRC32
	@table @code
	@item 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 @code{0}.

	@item 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.

	@item 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.

	@item 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.


	@item 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.

	@item 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.

	@item 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.

	@item GCRY_MD_TIGER1
	This is the TIGER variant as used by the NESSIE project. It uses the
	most commonly used output print order.

	@item GCRY_MD_TIGER2
	This is another variant of TIGER with a different padding scheme.


	@item 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.

	@item 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.

	@item GCRY_MD_SHA256
	This is the SHA-256 algorithm which yields a message digest of 32 bytes.
	See FIPS 180-2 for the specification.

	@item GCRY_MD_SHA384
	This is the SHA-384 algorithm which yields a message digest of 48 bytes.
	See FIPS 180-2 for the specification.

	@item GCRY_MD_SHA512
	This is the SHA-512 algorithm which yields a message digest of 64 bytes.
	See FIPS 180-2 for the specification.

	@item 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.

	@item 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.

	@item GCRY_MD_SHA3_224
	This is the SHA3-224 algorithm which yields a message digest of 28 bytes.
	See FIPS 202 for the specification.

	@item GCRY_MD_SHA3_256
	This is the SHA3-256 algorithm which yields a message digest of 32 bytes.
	See FIPS 202 for the specification.

	@item GCRY_MD_SHA3_384
	This is the SHA3-384 algorithm which yields a message digest of 48 bytes.
	See FIPS 202 for the specification.

	@item GCRY_MD_SHA3_512
	This is the SHA3-512 algorithm which yields a message digest of 64 bytes.
	See FIPS 202 for the specification.

	@item GCRY_MD_SHAKE128
	This is the SHAKE128 extendable-output function (XOF) algorithm with 128 bit
	security strength.
	See FIPS 202 for the specification.

	@item GCRY_MD_SHAKE256
	This is the SHAKE256 extendable-output function (XOF) algorithm with 256 bit
	security strength.
	See FIPS 202 for the specification.

	@item 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.

	@item 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.

	@item 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.

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

	@item GCRY_MD_GOSTR3411_94
	This is the hash algorithm described in GOST R 34.11-94 which yields a
	message digest of 32 bytes.

	@item 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.

	@item 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.

	@item GCRY_MD_BLAKE2B_512
	This is the BLAKE2b-512 algorithm which yields a message digest of 64 bytes.
	See RFC 7693 for the specification.

	@item GCRY_MD_BLAKE2B_384
	This is the BLAKE2b-384 algorithm which yields a message digest of 48 bytes.
	See RFC 7693 for the specification.

	@item GCRY_MD_BLAKE2B_256
	This is the BLAKE2b-256 algorithm which yields a message digest of 32 bytes.
	See RFC 7693 for the specification.

	@item GCRY_MD_BLAKE2B_160
	This is the BLAKE2b-160 algorithm which yields a message digest of 20 bytes.
	See RFC 7693 for the specification.

	@item GCRY_MD_BLAKE2S_256
	This is the BLAKE2s-256 algorithm which yields a message digest of 32 bytes.
	See RFC 7693 for the specification.

	@item GCRY_MD_BLAKE2S_224
	This is the BLAKE2s-224 algorithm which yields a message digest of 28 bytes.
	See RFC 7693 for the specification.

	@item GCRY_MD_BLAKE2S_160
	This is the BLAKE2s-160 algorithm which yields a message digest of 20 bytes.
	See RFC 7693 for the specification.

	@item GCRY_MD_BLAKE2S_128
	This is the BLAKE2s-128 algorithm which yields a message digest of 16 bytes.
	See RFC 7693 for the specification.

	@item GCRY_MD_SM3
	This is the SM3 algorithm which yields a message digest of 32 bytes.

	@end table
	@c end table of hash algorithms

	@node Working with hash algorithms
	@section Working with hash algorithms

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

	@deftypefun gcry_error_t gcry_md_open (gcry_md_hd_t *@var{hd}, int @var{algo}, unsigned int @var{flags})

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

	For a list of supported algorithms, see @ref{Available hash
	algorithms}.

	The flags allowed for @var{mode} are:

	@c begin table of hash flags
	@table @code
	@item GCRY_MD_FLAG_SECURE
	Allocate all buffers and the resulting digest in "secure memory". Use
	this is the hashed data is highly confidential.

	@item GCRY_MD_FLAG_HMAC
	@cindex 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
	@code{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 @ref{Working with cipher handles}.

	@item GCRY_MD_FLAG_BUGEMU1
	@cindex bug emulation
	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.


	@end table
	@c begin table of hash flags

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

	@end deftypefun
	@c end function gcry_md_open

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

	@deftypefun gcry_error_t gcry_md_enable (gcry_md_hd_t @var{h}, int @var{algo})

	Add the message digest algorithm @var{algo} to the digest object
	described by handle @var{h}. Duplicated enabling of algorithms is
	detected and ignored.
	@end deftypefun

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

	@deftypefun gcry_error_t gcry_md_setkey (gcry_md_hd_t @var{h}, const void *@var{key}, size_t @var{keylen})

	For use with the HMAC feature or BLAKE2 keyed hash, set the MAC key to
	the value of @var{key} of length @var{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.

	@end deftypefun


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

	@deftypefun void gcry_md_close (gcry_md_hd_t @var{h})

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


	@end deftypefun

	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:

	@deftypefun void gcry_md_reset (gcry_md_hd_t @var{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.
	@end deftypefun


	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:

	@deftypefun gcry_error_t gcry_md_copy (gcry_md_hd_t *@var{handle_dst}, gcry_md_hd_t @var{handle_src})

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


	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.

	@deftypefun void gcry_md_write (gcry_md_hd_t @var{h}, const void *@var{buffer}, size_t @var{length})

	Pass @var{length} bytes of the data in @var{buffer} to the digest object
	with handle @var{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.
	@end deftypefun

	@deftypefun void gcry_md_putc (gcry_md_hd_t @var{h}, int @var{c})

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

	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.

	@deftypefun void gcry_md_final (gcry_md_hd_t @var{h})

	Finalize the message digest calculation. This is not really needed
	because @code{gcry_md_read} and @code{gcry_md_extract} do this implicitly.
	After this has been done no further updates (by means of @code{gcry_md_write}
	or @code{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.
	@end deftypefun

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

	@deftypefun {unsigned char *} gcry_md_read (gcry_md_hd_t @var{h}, int @var{algo})

	@code{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 @code{gcry_md_close} or
	@code{gcry_md_reset} or it has been updated as a mitigation measure
	against timing attacks. @var{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 @code{NULL} if the requested algorithm has not
	been enabled.
	@end deftypefun

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

	@deftypefun gpg_err_code_t gcry_md_extract (gcry_md_hd_t @var{h}, @
	int @var{algo}, void *@var{buffer}, size_t @var{length})

	@code{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 @var{buffer} of the length @var{length}. Buffer will be fully
	populated with new output. @var{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.
	@end deftypefun

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

	@deftypefun gpg_err_code_t gcry_md_hash_buffers ( @
	@w{int @var{algo}}, @w{unsigned int @var{flags}}, @
	@w{void *@var{digest}}, @
	@w{const gcry_buffer_t *@var{iov}}, @w{int @var{iovcnt}} )

	@code{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 @var{iov} and @var{iovcnt}. @var{digest} must be
	allocated by the caller, large enough to hold the message digest
	yielded by the the specified algorithm @var{algo}. This required size
	may be obtained by using the function @code{gcry_md_get_algo_dlen}.

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

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

	On success the function returns 0 and stores the resulting hash or MAC
	at @var{digest}.
	@end deftypefun

	@deftypefun void gcry_md_hash_buffer (int @var{algo}, void @var{digest}, const void @var{buffer}, size_t @var{length});

	@code{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 @var{length} bytes at
	@var{buffer}. @var{digest} must be allocated by the caller, large
	enough to hold the message digest yielded by the the specified algorithm
	@var{algo}. This required size may be obtained by using the function
	@code{gcry_md_get_algo_dlen}.

	Note that in contrast to @code{gcry_md_hash_buffers} this function
	will abort the process if an unavailable algorithm is used.
	@end deftypefun

	@c ***********************************
	@c *** MD info functions *********
	@c ***********************************

	Hash algorithms are identified by internal algorithm numbers (see
	@code{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.

	@deftypefun {const char *} gcry_md_algo_name (int @var{algo})

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

	@deftypefun int gcry_md_map_name (const char *@var{name})

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

	@deftypefun gcry_error_t gcry_md_get_asnoid (int @var{algo}, void @var{buffer}, size_t @var{length})

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

	@end deftypefun


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

	@deftypefun gcry_error_t gcry_md_test_algo (int @var{algo})

	The macro returns 0 if the algorithm @var{algo} is available for use.
	@end deftypefun

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

	@deftypefun {unsigned int} gcry_md_get_algo_dlen (int @var{algo})

	Retrieve the length in bytes of the digest yielded by algorithm
	@var{algo}. This is often used prior to @code{gcry_md_read} to allocate
	sufficient memory for the digest.
	@end deftypefun


	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:

	@deftypefun int gcry_md_get_algo (gcry_md_hd_t @var{h})

	Retrieve the algorithm used with the handle @var{h}. Note that this
	does not work reliable if more than one algorithm is enabled in @var{h}.
	@end deftypefun

	The following macro might also be useful:

	@deftypefun int gcry_md_is_secure (gcry_md_hd_t @var{h})

	This function returns true when the digest object @var{h} is allocated
	in "secure memory"; i.e. @var{h} was created with the
	@code{GCRY_MD_FLAG_SECURE}.
	@end deftypefun

	@deftypefun int gcry_md_is_enabled (gcry_md_hd_t @var{h}, int @var{algo})

	This function returns true when the algorithm @var{algo} has been
	enabled for the digest object @var{h}.
	@end deftypefun



	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.

	@deftypefun void gcry_md_debug (gcry_md_hd_t @var{h}, const char *@var{suffix})

	Enable debugging for the digest object with handle @var{h}. This
	creates files named @file{dbgmd-<n>.<string>} while doing the
	actual hashing. @var{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 @code{gcry_md_write} or
	@code{gcry_md_putc}. If @code{NULL} is used for @var{suffix}, the
	debugging is stopped and the file closed. This is only rarely required
	because @code{gcry_md_close} implicitly stops debugging.
	@end deftypefun



	@c **********************************************************
	@c ***************** MAC Functions ********************
	@c **********************************************************
	@node Message Authentication Codes
	@chapter Message Authentication Codes

	Libgcrypt provides an easy and consistent to use interface for generating
	Message Authentication Codes (MAC). MAC generation is buffered and interface
	similar to the one used with hash algorithms. The programming model follows
	an open/process/close paradigm and is in that similar to other building blocks
	provided by Libgcrypt.

	@menu
	* Available MAC algorithms:: List of MAC algorithms supported by the library.
	* Working with MAC algorithms:: List of functions related to MAC algorithms.
	@end menu

	@node Available MAC algorithms
	@section Available MAC algorithms

	@c begin table of MAC algorithms
	@cindex HMAC-SHA-1
	@cindex HMAC-SHA-224, HMAC-SHA-256, HMAC-SHA-384, HMAC-SHA-512
	@cindex HMAC-SHA-512/224, HMAC-SHA-512/256
	@cindex HMAC-SHA3-224, HMAC-SHA3-256, HMAC-SHA3-384, HMAC-SHA3-512
	@cindex HMAC-RIPE-MD-160
	@cindex HMAC-MD2, HMAC-MD4, HMAC-MD5
	@cindex HMAC-TIGER1
	@cindex HMAC-SM3
	@cindex HMAC-Whirlpool
	@cindex HMAC-Stribog-256, HMAC-Stribog-512
	@cindex HMAC-GOSTR-3411-94
	@cindex HMAC-BLAKE2s, HMAC-BLAKE2b
	@table @code
	@item 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 @code{0}.

	@item GCRY_MAC_HMAC_SHA256
	This is keyed-hash message authentication code (HMAC) message authentication
	algorithm based on the SHA-256 hash algorithm.

	@item GCRY_MAC_HMAC_SHA224
	This is HMAC message authentication algorithm based on the SHA-224 hash
	algorithm.

	@item GCRY_MAC_HMAC_SHA512
	This is HMAC message authentication algorithm based on the SHA-512 hash
	algorithm.

	@item GCRY_MAC_HMAC_SHA384
	This is HMAC message authentication algorithm based on the SHA-384 hash
	algorithm.

	@item GCRY_MAC_HMAC_SHA3_256
	This is HMAC message authentication algorithm based on the SHA3-256 hash
	algorithm.

	@item GCRY_MAC_HMAC_SHA3_224
	This is HMAC message authentication algorithm based on the SHA3-224 hash
	algorithm.

	@item GCRY_MAC_HMAC_SHA3_512
	This is HMAC message authentication algorithm based on the SHA3-512 hash
	algorithm.

	@item GCRY_MAC_HMAC_SHA3_384
	This is HMAC message authentication algorithm based on the SHA3-384 hash
	algorithm.

	@item GCRY_MAC_HMAC_SHA512_224
	This is HMAC message authentication algorithm based on the SHA-512/224 hash
	algorithm.

	@item GCRY_MAC_HMAC_SHA512_256
	This is HMAC message authentication algorithm based on the SHA-512/256 hash
	algorithm.

	@item GCRY_MAC_HMAC_SHA1
	This is HMAC message authentication algorithm based on the SHA-1 hash
	algorithm.

	@item GCRY_MAC_HMAC_MD5
	This is HMAC message authentication algorithm based on the MD5 hash
	algorithm.

	@item GCRY_MAC_HMAC_MD4
	This is HMAC message authentication algorithm based on the MD4 hash
	algorithm.

	@item GCRY_MAC_HMAC_RMD160
	This is HMAC message authentication algorithm based on the RIPE-MD-160 hash
	algorithm.

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

	@item GCRY_MAC_HMAC_GOSTR3411_94
	This is HMAC message authentication algorithm based on the GOST R 34.11-94 hash
	algorithm.

	@item GCRY_MAC_HMAC_STRIBOG256
	This is HMAC message authentication algorithm based on the 256-bit hash
	algorithm described in GOST R 34.11-2012.

	@item GCRY_MAC_HMAC_STRIBOG512
	This is HMAC message authentication algorithm based on the 512-bit hash
	algorithm described in GOST R 34.11-2012.

	@item GCRY_MAC_HMAC_BLAKE2B_512
	This is HMAC message authentication algorithm based on the BLAKE2b-512 hash
	algorithm.

	@item GCRY_MAC_HMAC_BLAKE2B_384
	This is HMAC message authentication algorithm based on the BLAKE2b-384 hash
	algorithm.

	@item GCRY_MAC_HMAC_BLAKE2B_256
	This is HMAC message authentication algorithm based on the BLAKE2b-256 hash
	algorithm.

	@item GCRY_MAC_HMAC_BLAKE2B_160
	This is HMAC message authentication algorithm based on the BLAKE2b-160 hash
	algorithm.

	@item GCRY_MAC_HMAC_BLAKE2S_256
	This is HMAC message authentication algorithm based on the BLAKE2s-256 hash
	algorithm.

	@item GCRY_MAC_HMAC_BLAKE2S_224
	This is HMAC message authentication algorithm based on the BLAKE2s-224 hash
	algorithm.

	@item GCRY_MAC_HMAC_BLAKE2S_160
	This is HMAC message authentication algorithm based on the BLAKE2s-160 hash
	algorithm.

	@item GCRY_MAC_HMAC_BLAKE2S_128
	This is HMAC message authentication algorithm based on the BLAKE2s-128 hash
	algorithm.

	@item GCRY_MAC_HMAC_SM3
	This is HMAC message authentication algorithm based on the SM3 hash
	algorithm.

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

	@item GCRY_MAC_CMAC_3DES
	This is CMAC message authentication algorithm based on the three-key EDE
	Triple-DES block cipher algorithm.

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

	@item GCRY_MAC_CMAC_CAST5
	This is CMAC message authentication algorithm based on the CAST128-5
	block cipher algorithm.

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

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

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

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

	@item GCRY_MAC_CMAC_RFC2268
	This is CMAC message authentication algorithm based on the Ron's Cipher 2
	block cipher algorithm.

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

	@item GCRY_MAC_CMAC_GOST28147
	This is CMAC message authentication algorithm based on the GOST 28147-89
	block cipher algorithm.

	@item GCRY_MAC_CMAC_SM4
	This is CMAC message authentication algorithm based on the SM4
	block cipher algorithm.

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

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

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

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

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

	@item GCRY_MAC_POLY1305
	This is plain Poly1305 message authentication algorithm, used with
	one-time key.

	@item GCRY_MAC_POLY1305_AES
	This is Poly1305-AES message authentication algorithm, used with
	key and one-time nonce.

	@item GCRY_MAC_POLY1305_CAMELLIA
	This is Poly1305-Camellia message authentication algorithm, used with
	key and one-time nonce.

	@item GCRY_MAC_POLY1305_TWOFISH
	This is Poly1305-Twofish message authentication algorithm, used with
	key and one-time nonce.

	@item GCRY_MAC_POLY1305_SERPENT
	This is Poly1305-Serpent message authentication algorithm, used with
	key and one-time nonce.

	@item GCRY_MAC_POLY1305_SEED
	This is Poly1305-SEED message authentication algorithm, used with
	key and one-time nonce.

	@item GCRY_MAC_GOST28147_IMIT
	This is MAC construction defined in GOST 28147-89 (see RFC 5830 Section 8).

	@end table
	@c end table of MAC algorithms

	@node Working with MAC algorithms
	@section Working with MAC algorithms

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

	@deftypefun gcry_error_t gcry_mac_open (gcry_mac_hd_t *@var{hd}, int @var{algo}, unsigned int @var{flags}, gcry_ctx_t @var{ctx})

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

	For a list of supported algorithms, see @ref{Available MAC algorithms}.

	The flags allowed for @var{mode} are:

	@c begin table of MAC flags
	@table @code
	@item GCRY_MAC_FLAG_SECURE
	Allocate all buffers and the resulting MAC in "secure memory". Use this if the
	MAC data is highly confidential.

	@end table
	@c begin table of MAC flags

	@end deftypefun
	@c end function gcry_mac_open


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

	@deftypefun gcry_error_t gcry_mac_setkey (gcry_mac_hd_t @var{h}, const void *@var{key}, size_t @var{keylen})

	Set the MAC key to the value of @var{key} of length @var{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.
	@end deftypefun


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

	@deftypefun gcry_error_t gcry_mac_setiv (gcry_mac_hd_t @var{h}, const void *@var{iv}, size_t @var{ivlen})

	Set the IV to the value of @var{iv} of length @var{ivlen} bytes.
	@end deftypefun


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

	@deftypefun void gcry_mac_close (gcry_mac_hd_t @var{h})

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


	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:

	@deftypefun gcry_error_t gcry_mac_reset (gcry_mac_hd_t @var{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.
	@end deftypefun


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

	@deftypefun gcry_error_t gcry_mac_write (gcry_mac_hd_t @var{h}, const void *@var{buffer}, size_t @var{length})

	Pass @var{length} bytes of the data in @var{buffer} to the MAC object
	with handle @var{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.
	@end deftypefun

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

	@deftypefun gcry_error_t gcry_mac_read (gcry_mac_hd_t @var{h}, void @var{buffer}, size_t @var{length})

	@code{gcry_mac_read} returns the MAC after finalizing the calculation.
	Function copies the resulting MAC value to @var{buffer} of the length
	@var{length}. If @var{length} is larger than length of resulting MAC value,
	then length of MAC is returned through @var{length}.
	@end deftypefun

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

	@deftypefun gcry_error_t gcry_mac_verify (gcry_mac_hd_t @var{h}, void *@var{buffer}, size_t @var{length})

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


	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:

	@deftypefun {int} gcry_mac_get_algo (gcry_mac_hd_t @var{h})

	Retrieve the algorithm used with the handle @var{h}.
	@end deftypefun


	@c ***********************************
	@c *** MAC info functions ********
	@c ***********************************

	MAC algorithms are identified by internal algorithm numbers (see
	@code{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.

	@deftypefun {const char *} gcry_mac_algo_name (int @var{algo})

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

	@deftypefun int gcry_mac_map_name (const char *@var{name})

	Map the algorithm with @var{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.
	@end deftypefun


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

	@deftypefun gcry_error_t gcry_mac_test_algo (int @var{algo})

	The macro returns 0 if the MAC algorithm @var{algo} is available for use.
	@end deftypefun


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

	@deftypefun {unsigned int} gcry_mac_get_algo_maclen (int @var{algo})

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


	@deftypefun {unsigned int} gcry_mac_get_algo_keylen (@var{algo})

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



	@c *******************************************************
	@c ***************** KDF ***************************
	@c *******************************************************
	@node Key Derivation
	@chapter Key Derivation

	@acronym{Libgcypt} provides a general purpose function to derive keys
	from strings.

	@deftypefun gpg_error_t gcry_kdf_derive ( @
	@w{const void *@var{passphrase}}, @w{size_t @var{passphraselen}}, @
	@w{int @var{algo}}, @w{int @var{subalgo}}, @
	@w{const void *@var{salt}}, @w{size_t @var{saltlen}}, @
	@w{unsigned long @var{iterations}}, @
	@w{size_t @var{keysize}}, @w{void *@var{keybuffer}} )


	Derive a key from a passphrase. @var{keysize} gives the requested
	size of the keys in octets. @var{keybuffer} is a caller provided
	buffer filled on success with the derived key. The input passphrase
	is taken from @var{passphrase} which is an arbitrary memory buffer of
	@var{passphraselen} octets. @var{algo} specifies the KDF algorithm to
	use; see below. @var{subalgo} specifies an algorithm used internally
	by the KDF algorithms; this is usually a hash algorithm but certain
	KDF algorithms may use it differently. @var{salt} is a salt of length
	@var{saltlen} octets, as needed by most KDF algorithms.
	@var{iterations} is a positive integer parameter to most KDFs.

	@noindent
	On success 0 is returned; on failure an error code.

	@noindent
	Currently supported KDFs (parameter @var{algo}):

	@table @code
	@item GCRY_KDF_SIMPLE_S2K
	The OpenPGP simple S2K algorithm (cf. RFC4880). Its use is strongly
	deprecated. @var{salt} and @var{iterations} are not needed and may be
	passed as @code{NULL}/@code{0}.

	@item GCRY_KDF_SALTED_S2K
	The OpenPGP salted S2K algorithm (cf. RFC4880). Usually not used.
	@var{iterations} is not needed and may be passed as @code{0}. @var{saltlen}
	must be given as 8.

	@item GCRY_KDF_ITERSALTED_S2K
	The OpenPGP iterated+salted S2K algorithm (cf. RFC4880). This is the
	default for most OpenPGP applications. @var{saltlen} must be given as
	8. Note that OpenPGP defines a special encoding of the
	@var{iterations}; however this function takes the plain decoded
	iteration count.

	@item GCRY_KDF_PBKDF2
	The PKCS#5 Passphrase Based Key Derivation Function number 2.

	@item GCRY_KDF_SCRYPT
	The SCRYPT Key Derivation Function. The subalgorithm is used to specify
	the CPU/memory cost parameter N, and the number of iterations
	is used for the parallelization parameter p. The block size is fixed
	at 8 in the current implementation.

	@end table
	@end deftypefun


	@c **********************************************************
	@c ***************** Random ***************************
	@c **********************************************************
	@node Random Numbers
	@chapter Random Numbers

	@menu
	* Quality of random numbers:: Libgcrypt uses different quality levels.
	* Retrieving random numbers:: How to retrieve random numbers.
	@end menu

	@node Quality of random numbers
	@section Quality of random numbers

	@acronym{Libgcypt} offers random numbers of different quality levels:

	@deftp {Data type} gcry_random_level_t
	The constants for the random quality levels are of this enum type.
	@end deftp

	@table @code
	@item GCRY_WEAK_RANDOM
	For all functions, except for @code{gcry_mpi_randomize}, this level maps
	to GCRY_STRONG_RANDOM. If you do not want this, consider using
	@code{gcry_create_nonce}.
	@item GCRY_STRONG_RANDOM
	Use this level for session keys and similar purposes.
	@item GCRY_VERY_STRONG_RANDOM
	Use this level for long term key material.
	@end table

	@node Retrieving random numbers
	@section Retrieving random numbers

	@deftypefun void gcry_randomize (unsigned char *@var{buffer}, size_t @var{length}, enum gcry_random_level @var{level})

	Fill @var{buffer} with @var{length} random bytes using a random quality
	as defined by @var{level}.
	@end deftypefun

	@deftypefun {void *} gcry_random_bytes (size_t @var{nbytes}, enum gcry_random_level @var{level})

	Convenience function to allocate a memory block consisting of
	@var{nbytes} fresh random bytes using a random quality as defined by
	@var{level}.
	@end deftypefun

	@deftypefun {void *} gcry_random_bytes_secure (size_t @var{nbytes}, enum gcry_random_level @var{level})

	Convenience function to allocate a memory block consisting of
	@var{nbytes} fresh random bytes using a random quality as defined by
	@var{level}. This function differs from @code{gcry_random_bytes} in
	that the returned buffer is allocated in a ``secure'' area of the
	memory.
	@end deftypefun

	@deftypefun void gcry_create_nonce (unsigned char *@var{buffer}, size_t @var{length})

	Fill @var{buffer} with @var{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.

	@end deftypefun



	@c **********************************************************
	@c ***************** S-Expressions *********************
	@c **********************************************************
	@node S-expressions
	@chapter S-expressions

	S-expressions are used by the public key functions to pass complex data
	structures around. These LISP like objects are used by some
	cryptographic protocols (cf. RFC-2692) and Libgcrypt provides functions
	to parse and construct them. For detailed information, see
	@cite{Ron Rivest, code and description of S-expressions,
	@uref{http://theory.lcs.mit.edu/~rivest/sexp.html}}.

	@menu
	* Data types for S-expressions:: Data types related with S-expressions.
	* Working with S-expressions:: How to work with S-expressions.
	@end menu

	@node Data types for S-expressions
	@section Data types for S-expressions

	@deftp {Data type} gcry_sexp_t
	The @code{gcry_sexp_t} type describes an object with the Libgcrypt internal
	representation of an S-expression.
	@end deftp

	@node Working with S-expressions
	@section Working with S-expressions

	@noindent
	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:


	@deftypefun gcry_error_t gcry_sexp_new (@w{gcry_sexp_t @var{r_sexp}}, @w{const void @var{buffer}}, @w{size_t @var{length}}, @w{int @var{autodetect}})

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

	@deftypefun gcry_error_t gcry_sexp_create (@w{gcry_sexp_t @var{r_sexp}}, @w{void @var{buffer}}, @w{size_t @var{length}}, @w{int @var{autodetect}}, @w{void (@var{freefnc})(void)})

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

	@deftypefun gcry_error_t gcry_sexp_sscan (@w{gcry_sexp_t @var{r_sexp}}, @w{size_t @var{erroff}}, @w{const char *@var{buffer}}, @w{size_t @var{length}})

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

	@deftypefun gcry_error_t gcry_sexp_build (@w{gcry_sexp_t @var{r_sexp}}, @w{size_t @var{erroff}}, @w{const char *@var{format}, ...})

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

	@table @samp
	@item %m
	The next argument is expected to be of type @code{gcry_mpi_t} and a copy of
	its value is inserted into the resulting S-expression. The MPI is
	stored as a signed integer.
	@item %M
	The next argument is expected to be of type @code{gcry_mpi_t} and a copy of
	its value is inserted into the resulting S-expression. The MPI is
	stored as an unsigned integer.
	@item %s
	The next argument is expected to be of type @code{char *} and that
	string is inserted into the resulting S-expression.
	@item %d
	The next argument is expected to be of type @code{int} and its value is
	inserted into the resulting S-expression.
	@item %u
	The next argument is expected to be of type @code{unsigned int} and
	its value is inserted into the resulting S-expression.
	@item %b
	The next argument is expected to be of type @code{int} directly
	followed by an argument of type @code{char *}. This represents a
	buffer of given length to be inserted into the resulting S-expression.
	@item %S
	The next argument is expected to be of type @code{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.

	@end table

	@noindent
	No other format characters are defined and would return an error. Note
	that the format character @samp{%%} does not exists, because a percent
	sign is not a valid character in an S-expression.
	@end deftypefun

	@deftypefun void gcry_sexp_release (@w{gcry_sexp_t @var{sexp}})

	Release the S-expression object @var{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.
	@end deftypefun


	@noindent
	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.

	@deftypefun size_t gcry_sexp_sprint (@w{gcry_sexp_t @var{sexp}}, @w{int @var{mode}}, @w{char *@var{buffer}}, @w{size_t @var{maxlength}})

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

	@noindent
	The following formats are supported:

	@table @code
	@item GCRYSEXP_FMT_DEFAULT
	Returns a convenient external S-expression representation.

	@item GCRYSEXP_FMT_CANON
	Return the S-expression in canonical format.

	@item GCRYSEXP_FMT_BASE64
	Not currently supported.

	@item GCRYSEXP_FMT_ADVANCED
	Returns the S-expression in advanced format.
	@end table
	@end deftypefun

	@deftypefun void gcry_sexp_dump (@w{gcry_sexp_t @var{sexp}})

	Dumps @var{sexp} in a format suitable for debugging to Libgcrypt's
	logging stream.
	@end deftypefun

	@noindent
	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.

	@deftypefun size_t gcry_sexp_canon_len (@w{const unsigned char @var{buffer}}, @w{size_t @var{length}}, @w{size_t @var{erroff}}, @w{int *@var{errcode}})

	Scan the canonical encoded @var{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 @var{length} is not 0, the maximum
	length to scan is given; this can be used for syntax checks of
	data passed from outside. @var{errcode} and @var{erroff} may both be
	passed as @code{NULL}.

	@end deftypefun


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

	@deftypefun gcry_sexp_t gcry_sexp_find_token (@w{const gcry_sexp_t @var{list}}, @w{const char *@var{token}}, @w{size_t @var{toklen}})

	Scan the S-expression for a sublist with a type (the car of the list)
	matching the string @var{token}. If @var{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 @code{NULL}
	when not found.
	@end deftypefun


	@deftypefun int gcry_sexp_length (@w{const gcry_sexp_t @var{list}})

	Return the length of the @var{list}. For a valid S-expression this
	should be at least 1.
	@end deftypefun


	@deftypefun gcry_sexp_t gcry_sexp_nth (@w{const gcry_sexp_t @var{list}}, @w{int @var{number}})

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

	@deftypefun gcry_sexp_t gcry_sexp_car (@w{const gcry_sexp_t @var{list}})

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

	@deftypefun gcry_sexp_t gcry_sexp_cdr (@w{const gcry_sexp_t @var{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 @code{NULL} on error.
	@end deftypefun


	@deftypefun {const char } gcry_sexp_nth_data (@w{const gcry_sexp_t @var{list}}, @w{int @var{number}}, @w{size_t @var{datalen}})

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

	@noindent
	Here is an example on how to extract and print the surname (Meier) from
	the S-expression @samp{(Name Otto Meier (address Burgplatz 3))}:

	@example
	size_t len;
	const char *name;

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

	@deftypefun {void } gcry_sexp_nth_buffer (@w{const gcry_sexp_t @var{list}}, @w{int @var{number}}, @w{size_t @var{rlength}})

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

	@noindent
	Here is an example on how to extract and print the CRC value from the
	S-expression @samp{(hash crc32 #23ed00d7)}:

	@example
	size_t len;
	char *value;

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

	@deftypefun {char *} gcry_sexp_nth_string (@w{gcry_sexp_t @var{list}}, @w{int @var{number}})

	This function is used to get and convert data from a @var{list}. The
	data is assumed to be a Nul terminated string. The caller must
	release this returned value using @code{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, @code{NULL} is returned.
	@end deftypefun

	@deftypefun gcry_mpi_t gcry_sexp_nth_mpi (@w{gcry_sexp_t @var{list}}, @w{int @var{number}}, @w{int @var{mpifmt}})

	This function is used to get and convert data from a @var{list}. This
	data is assumed to be an MPI stored in the format described by
	@var{mpifmt} and returned as a standard Libgcrypt MPI. The caller must
	release this returned value using @code{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, @code{NULL} is returned. If you use
	this function to parse results of a public key function, you most
	likely want to use @code{GCRYMPI_FMT_USG}.
	@end deftypefun

	@deftypefun gpg_error_t gcry_sexp_extract_param ( @
	@w{gcry_sexp_t @var{sexp}}, @
	@w{const char *@var{path}}, @
	@w{const char *@var{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:

	@table @samp
	@item +
	Switch to unsigned integer format (GCRYMPI_FMT_USG). This is the
	default mode.
	@item -
	Switch to standard signed format (GCRYMPI_FMT_STD).
	@item /
	Switch to opaque MPI format. The resulting MPIs may not be used for
	computations; see @code{gcry_mpi_get_opaque} for details.
	@item &
	Switch to buffer descriptor mode. See below for details.
	@item %s
	Switch to string mode. The expected argument is the address of a
	@code{char *} variable; the caller must release that value. If the
	parameter was marked optional and is not found, NULL is stored.
	@item %#s
	Switch to multi string mode. The expected argument is the address of a
	@code{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 @code{()} is
	inserted in place of that sublist.
	@item %u
	Switch to unsigned integer mode. The expected argument is address of
	a @code{unsigned int} variable.
	@item %lu
	Switch to unsigned long integer mode. The expected argument is address of
	a @code{unsigned long} variable.
	@item %d
	Switch to signed integer mode. The expected argument is address of
	a @code{int} variable.
	@item %ld
	Switch to signed long integer mode. The expected argument is address of
	a @code{long} variable.
	@item %zu
	Switch to size_t mode. The expected argument is address of
	a @code{size_t} variable.
	@item ?
	If immediately following a parameter letter (no white space allowed),
	that parameter is considered optional.
	@end table

	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 @code{gcry_mpi_t} variable is expected that must be set to
	@code{NULL} prior to invoking this function, and finally a @code{NULL}
	is expected. For example

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

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

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

	In buffer descriptor mode a pointer to a @code{gcry_buffer_t}
	descriptor is expected instead of a pointer to an MPI. The caller may
	use two different operation modes here: If the @var{data} field of the
	provided descriptor is @code{NULL}, the function allocates a new
	buffer and stores it at @var{data}; the other fields are set
	accordingly with @var{off} set to 0. If @var{data} is not
	@code{NULL}, the function assumes that the @var{data}, @var{size}, and
	@var{off} fields specify a buffer where to but the value of the
	respective parameter; on return the @var{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
	@var{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 @code{NULL}, and all passed buffers
	are either truncated if the caller supplied the buffer, or deallocated
	if the function allocated the buffer.
	@end deftypefun


	@c **********************************************************
	@c ***************** MPIs **** *********************
	@c **********************************************************
	@node MPI library
	@chapter MPI library

	@menu
	* Data types:: MPI related data types.
	* Basic functions:: First steps with MPI numbers.
	* MPI formats:: External representation of MPIs.
	* Calculations:: Performing MPI calculations.
	* Comparisons:: How to compare MPI values.
	* Bit manipulations:: How to access single bits of MPI values.
	* EC functions:: Elliptic curve related functions.
	* Miscellaneous:: Miscellaneous MPI functions.
	@end menu

	Public key cryptography is based on mathematics with large numbers. To
	implement the public key functions, a library for handling these large
	numbers is required. Because of the general usefulness of such a
	library, its interface is exposed by Libgcrypt.
	In the context of Libgcrypt and in most other applications, these large
	numbers are called MPIs (multi-precision-integers).

	@node Data types
	@section Data types

	@deftp {Data type} {gcry_mpi_t}
	This type represents an object to hold an MPI.
	@end deftp

	@deftp {Data type} {gcry_mpi_point_t}
	This type represents an object to hold a point for elliptic curve math.
	@end deftp

	@node Basic functions
	@section Basic functions

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

	@deftypefun gcry_mpi_t gcry_mpi_new (@w{unsigned int @var{nbits}})

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

	@deftypefun gcry_mpi_t gcry_mpi_snew (@w{unsigned int @var{nbits}})

	This is identical to @code{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.
	@end deftypefun

	@deftypefun gcry_mpi_t gcry_mpi_copy (@w{const gcry_mpi_t @var{a}})

	Create a new MPI as the exact copy of @var{a} but with the constant
	and immutable flags cleared.
	@end deftypefun


	@deftypefun void gcry_mpi_release (@w{gcry_mpi_t @var{a}})

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

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

	@deftypefun gcry_mpi_t gcry_mpi_set (@w{gcry_mpi_t @var{w}}, @w{const gcry_mpi_t @var{u}})

	Assign the value of @var{u} to @var{w} and return @var{w}. If
	@code{NULL} is passed for @var{w}, a new MPI is allocated, set to the
	value of @var{u} and returned.
	@end deftypefun

	@deftypefun gcry_mpi_t gcry_mpi_set_ui (@w{gcry_mpi_t @var{w}}, @w{unsigned long @var{u}})

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

	@deftypefun gcry_error_t gcry_mpi_get_ui (@w{unsigned int *@var{w}}, @w{gcry_mpi_t @var{u}})

	If @var{u} is not negative and small enough to be stored in an
	@code{unsigned int} variable, store its value at @var{w}. If the
	value does not fit or is negative return GPG_ERR_ERANGE and do not
	change the value stored at @var{w}. Note that this function returns
	an @code{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 @code{unsigned long}.
	@end deftypefun


	@deftypefun void gcry_mpi_swap (@w{gcry_mpi_t @var{a}}, @w{gcry_mpi_t @var{b}})

	Swap the values of @var{a} and @var{b}.
	@end deftypefun

	@deftypefun void gcry_mpi_snatch (@w{gcry_mpi_t @var{w}}, @
	@w{const gcry_mpi_t @var{u}})

	Set @var{u} into @var{w} and release @var{u}. If @var{w} is
	@code{NULL} only @var{u} will be released.
	@end deftypefun

	@deftypefun void gcry_mpi_neg (@w{gcry_mpi_t @var{w}}, @w{gcry_mpi_t @var{u}})

	Set the sign of @var{w} to the negative of @var{u}.
	@end deftypefun

	@deftypefun void gcry_mpi_abs (@w{gcry_mpi_t @var{w}})

	Clear the sign of @var{w}.
	@end deftypefun


	@node MPI formats
	@section MPI formats

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

	@deftypefun gcry_error_t gcry_mpi_scan (@w{gcry_mpi_t @var{r_mpi}}, @w{enum gcry_mpi_format @var{format}}, @w{const unsigned char @var{buffer}}, @w{size_t @var{buflen}}, @w{size_t *@var{nscanned}})

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

	@table @code
	@item GCRYMPI_FMT_STD
	2-complement stored without a length header. Note that
	@code{gcry_mpi_print} stores a @code{0} as a string of zero length.

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

	@item GCRYMPI_FMT_SSH
	As used in the Secure Shell protocol. This is @code{GCRYMPI_FMT_STD}
	with a 4 byte big endian header.

	@item 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, @var{buflen} must be zero.

	@item GCRYMPI_FMT_USG
	Simple unsigned integer.
	@end table

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


	@deftypefun gcry_error_t gcry_mpi_print (@w{enum gcry_mpi_format @var{format}}, @w{unsigned char @var{buffer}}, @w{size_t @var{buflen}}, @w{size_t @var{nwritten}}, @w{const gcry_mpi_t @var{a}})

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

	@deftypefun gcry_error_t gcry_mpi_aprint (@w{enum gcry_mpi_format @var{format}}, @w{unsigned char *@var{buffer}}, @w{size_t @var{nbytes}}, @w{const gcry_mpi_t @var{a}})

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

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

	@end deftypefun

	@deftypefun void gcry_mpi_dump (@w{const gcry_mpi_t @var{a}})

	Dump the value of @var{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 @code{NULL} for
	@var{a}.
	@end deftypefun


	@node Calculations
	@section Calculations

	@noindent
	Basic arithmetic operations:

	@deftypefun void gcry_mpi_add (@w{gcry_mpi_t @var{w}}, @w{gcry_mpi_t @var{u}}, @w{gcry_mpi_t @var{v}})

	@math{@var{w} = @var{u} + @var{v}}.
	@end deftypefun


	@deftypefun void gcry_mpi_add_ui (@w{gcry_mpi_t @var{w}}, @w{gcry_mpi_t @var{u}}, @w{unsigned long @var{v}})

	@math{@var{w} = @var{u} + @var{v}}. Note that @var{v} is an unsigned integer.
	@end deftypefun


	@deftypefun void gcry_mpi_addm (@w{gcry_mpi_t @var{w}}, @w{gcry_mpi_t @var{u}}, @w{gcry_mpi_t @var{v}}, @w{gcry_mpi_t @var{m}})

	@math{@var{w} = @var{u} + @var{v} \bmod @var{m}}.
	@end deftypefun

	@deftypefun void gcry_mpi_sub (@w{gcry_mpi_t @var{w}}, @w{gcry_mpi_t @var{u}}, @w{gcry_mpi_t @var{v}})

	@math{@var{w} = @var{u} - @var{v}}.
	@end deftypefun

	@deftypefun void gcry_mpi_sub_ui (@w{gcry_mpi_t @var{w}}, @w{gcry_mpi_t @var{u}}, @w{unsigned long @var{v}})

	@math{@var{w} = @var{u} - @var{v}}. @var{v} is an unsigned integer.
	@end deftypefun

	@deftypefun void gcry_mpi_subm (@w{gcry_mpi_t @var{w}}, @w{gcry_mpi_t @var{u}}, @w{gcry_mpi_t @var{v}}, @w{gcry_mpi_t @var{m}})

	@math{@var{w} = @var{u} - @var{v} \bmod @var{m}}.
	@end deftypefun

	@deftypefun void gcry_mpi_mul (@w{gcry_mpi_t @var{w}}, @w{gcry_mpi_t @var{u}}, @w{gcry_mpi_t @var{v}})

	@math{@var{w} = @var{u} * @var{v}}.
	@end deftypefun

	@deftypefun void gcry_mpi_mul_ui (@w{gcry_mpi_t @var{w}}, @w{gcry_mpi_t @var{u}}, @w{unsigned long @var{v}})

	@math{@var{w} = @var{u} * @var{v}}. @var{v} is an unsigned integer.
	@end deftypefun

	@deftypefun void gcry_mpi_mulm (@w{gcry_mpi_t @var{w}}, @w{gcry_mpi_t @var{u}}, @w{gcry_mpi_t @var{v}}, @w{gcry_mpi_t @var{m}})

	@math{@var{w} = @var{u} * @var{v} \bmod @var{m}}.
	@end deftypefun

	@deftypefun void gcry_mpi_mul_2exp (@w{gcry_mpi_t @var{w}}, @w{gcry_mpi_t @var{u}}, @w{unsigned long @var{e}})

	@c FIXME: I am in need for a real TeX{info} guru:
	@c I don't know why TeX can grok @var{e} here.
	@math{@var{w} = @var{u} * 2^e}.
	@end deftypefun

	@deftypefun void gcry_mpi_div (@w{gcry_mpi_t @var{q}}, @w{gcry_mpi_t @var{r}}, @w{gcry_mpi_t @var{dividend}}, @w{gcry_mpi_t @var{divisor}}, @w{int @var{round}})

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

	@deftypefun void gcry_mpi_mod (@w{gcry_mpi_t @var{r}}, @w{gcry_mpi_t @var{dividend}}, @w{gcry_mpi_t @var{divisor}})

	@math{@var{r} = @var{dividend} \bmod @var{divisor}}.
	@end deftypefun

	@deftypefun void gcry_mpi_powm (@w{gcry_mpi_t @var{w}}, @w{const gcry_mpi_t @var{b}}, @w{const gcry_mpi_t @var{e}}, @w{const gcry_mpi_t @var{m}})

	@c I don't know why TeX can grok @var{e} here.
	@math{@var{w} = @var{b}^e \bmod @var{m}}.
	@end deftypefun

	@deftypefun int gcry_mpi_gcd (@w{gcry_mpi_t @var{g}}, @w{gcry_mpi_t @var{a}}, @w{gcry_mpi_t @var{b}})

	Set @var{g} to the greatest common divisor of @var{a} and @var{b}.
	Return true if the @var{g} is 1.
	@end deftypefun

	@deftypefun int gcry_mpi_invm (@w{gcry_mpi_t @var{x}}, @w{gcry_mpi_t @var{a}}, @w{gcry_mpi_t @var{m}})

	Set @var{x} to the multiplicative inverse of @math{@var{a} \bmod @var{m}}.
	Return true if the inverse exists.
	@end deftypefun


	@node Comparisons
	@section Comparisons

	@noindent
	The next 2 functions are used to compare MPIs:


	@deftypefun int gcry_mpi_cmp (@w{const gcry_mpi_t @var{u}}, @w{const gcry_mpi_t @var{v}})

	Compare the multi-precision-integers number @var{u} and @var{v}
	returning 0 for equality, a positive value for @var{u} > @var{v} and a
	negative for @var{u} < @var{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.
	@end deftypefun

	@deftypefun int gcry_mpi_cmp_ui (@w{const gcry_mpi_t @var{u}}, @w{unsigned long @var{v}})

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

	@deftypefun int gcry_mpi_is_neg (@w{const gcry_mpi_t @var{a}})

	Return 1 if @var{a} is less than zero; return 0 if zero or positive.
	@end deftypefun


	@node Bit manipulations
	@section Bit manipulations

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

	@deftypefun {unsigned int} gcry_mpi_get_nbits (@w{gcry_mpi_t @var{a}})

	Return the number of bits required to represent @var{a}.
	@end deftypefun

	@deftypefun int gcry_mpi_test_bit (@w{gcry_mpi_t @var{a}}, @w{unsigned int @var{n}})

	Return true if bit number @var{n} (counting from 0) is set in @var{a}.
	@end deftypefun

	@deftypefun void gcry_mpi_set_bit (@w{gcry_mpi_t @var{a}}, @w{unsigned int @var{n}})

	Set bit number @var{n} in @var{a}.
	@end deftypefun

	@deftypefun void gcry_mpi_clear_bit (@w{gcry_mpi_t @var{a}}, @w{unsigned int @var{n}})

	Clear bit number @var{n} in @var{a}.
	@end deftypefun

	@deftypefun void gcry_mpi_set_highbit (@w{gcry_mpi_t @var{a}}, @w{unsigned int @var{n}})

	Set bit number @var{n} in @var{a} and clear all bits greater than @var{n}.
	@end deftypefun

	@deftypefun void gcry_mpi_clear_highbit (@w{gcry_mpi_t @var{a}}, @w{unsigned int @var{n}})

	Clear bit number @var{n} in @var{a} and all bits greater than @var{n}.
	@end deftypefun

	@deftypefun void gcry_mpi_rshift (@w{gcry_mpi_t @var{x}}, @w{gcry_mpi_t @var{a}}, @w{unsigned int @var{n}})

	Shift the value of @var{a} by @var{n} bits to the right and store the
	result in @var{x}.
	@end deftypefun

	@deftypefun void gcry_mpi_lshift (@w{gcry_mpi_t @var{x}}, @w{gcry_mpi_t @var{a}}, @w{unsigned int @var{n}})

	Shift the value of @var{a} by @var{n} bits to the left and store the
	result in @var{x}.
	@end deftypefun

	@node EC functions
	@section EC functions

	@noindent
	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.

	@deftypefun gcry_mpi_point_t gcry_mpi_point_new (@w{unsigned int @var{nbits}})

	Allocate a new point object, initialize it to 0, and allocate enough
	memory for a points of at least @var{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 @var{nbits} is usually the right thing to do.
	@end deftypefun

	@deftypefun void gcry_mpi_point_release (@w{gcry_mpi_point_t @var{point}})

	Release @var{point} and free all associated resources. Passing
	@code{NULL} is allowed and ignored.
	@end deftypefun

	@deftypefun gcry_mpi_point_t gcry_mpi_point_copy (@w{gcry_mpi_point_t @var{point}})

	Allocate and return a new point object and initialize it with
	@var{point}. If @var{point} is NULL the function is identical to
	@code{gcry_mpi_point_new(0)}.
	@end deftypefun

	@deftypefun void gcry_mpi_point_get (@w{gcry_mpi_t @var{x}}, @
	@w{gcry_mpi_t @var{y}}, @w{gcry_mpi_t @var{z}}, @
	@w{gcry_mpi_point_t @var{point}})

	Store the projective coordinates from @var{point} into the MPIs
	@var{x}, @var{y}, and @var{z}. If a coordinate is not required,
	@code{NULL} may be used for @var{x}, @var{y}, or @var{z}.
	@end deftypefun

	@deftypefun void gcry_mpi_point_snatch_get (@w{gcry_mpi_t @var{x}}, @
	@w{gcry_mpi_t @var{y}}, @w{gcry_mpi_t @var{z}}, @
	@w{gcry_mpi_point_t @var{point}})

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

	@deftypefun gcry_mpi_point_t gcry_mpi_point_set ( @
	@w{gcry_mpi_point_t @var{point}}, @
	@w{gcry_mpi_t @var{x}}, @w{gcry_mpi_t @var{y}}, @w{gcry_mpi_t @var{z}})

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

	@deftypefun gcry_mpi_point_t gcry_mpi_point_snatch_set ( @
	@w{gcry_mpi_point_t @var{point}}, @
	@w{gcry_mpi_t @var{x}}, @w{gcry_mpi_t @var{y}}, @w{gcry_mpi_t @var{z}})

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

	@anchor{gcry_mpi_ec_new}
	@deftypefun gpg_error_t gcry_mpi_ec_new (@w{gcry_ctx_t *@var{r_ctx}}, @
	@w{gcry_sexp_t @var{keyparam}}, @w{const char *@var{curvename}})

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

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

	On success the function returns 0 and stores the new context object at
	@var{r_ctx}; this object eventually needs to be released
	(@pxref{gcry_ctx_release}). On error the function stores @code{NULL} at
	@var{r_ctx} and returns an error code.
	@end deftypefun

	@deftypefun gcry_mpi_t gcry_mpi_ec_get_mpi ( @
	@w{const char *@var{name}}, @w{gcry_ctx_t @var{ctx}}, @w{int @var{copy}})

	Return the MPI with @var{name} from the context @var{ctx}. If not
	found @code{NULL} is returned. If the returned MPI may later be
	modified, it is suggested to pass @code{1} to @var{copy}, so that the
	function guarantees that a modifiable copy of the MPI is returned. If
	@code{0} is used for @var{copy}, this function may return a constant
	flagged MPI. In any case @code{gcry_mpi_release} needs to be called
	to release the result. For valid names @ref{ecc_keyparam}. If the
	public key @code{q} is requested but only the private key @code{d} is
	available, @code{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:
	@table @var
	@item q@@eddsa
	Return an EdDSA style compressed point. This is only supported for
	Twisted Edwards curves.
	@end table
	@end deftypefun

	@deftypefun gcry_mpi_point_t gcry_mpi_ec_get_point ( @
	@w{const char *@var{name}}, @w{gcry_ctx_t @var{ctx}}, @w{int @var{copy}})

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

	@deftypefun gpg_error_t gcry_mpi_ec_set_mpi ( @
	@w{const char *@var{name}}, @w{gcry_mpi_t @var{newvalue}}, @
	@w{gcry_ctx_t @var{ctx}})

	Store the MPI @var{newvalue} at @var{name} into the context @var{ctx}.
	On success @code{0} is returned; on error an error code. Valid names
	are the MPI parameters of an elliptic curve (@pxref{ecc_keyparam}).
	@end deftypefun

	@deftypefun gpg_error_t gcry_mpi_ec_set_point ( @
	@w{const char *@var{name}}, @w{gcry_mpi_point_t @var{newvalue}}, @
	@w{gcry_ctx_t @var{ctx}})

	Store the point @var{newvalue} at @var{name} into the context
	@var{ctx}. On success @code{0} is returned; on error an error code.
	Valid names are the point parameters of an elliptic curve
	(@pxref{ecc_keyparam}).
	@end deftypefun

	@deftypefun gpg_err_code_t gcry_mpi_ec_decode_point ( @
	@w{mpi_point_t @var{result}}, @w{gcry_mpi_t @var{value}}, @
	@w{gcry_ctx_t @var{ctx}})

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


	@deftypefun int gcry_mpi_ec_get_affine ( @
	@w{gcry_mpi_t @var{x}}, @w{gcry_mpi_t @var{y}}, @
	@w{gcry_mpi_point_t @var{point}}, @w{gcry_ctx_t @var{ctx}})

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

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

	@end deftypefun

	@deftypefun void gcry_mpi_ec_dup ( @
	@w{gcry_mpi_point_t @var{w}}, @w{gcry_mpi_point_t @var{u}}, @
	@w{gcry_ctx_t @var{ctx}})

	Double the point @var{u} of the elliptic curve described by @var{ctx}
	and store the result into @var{w}.
	@end deftypefun

	@deftypefun void gcry_mpi_ec_add ( @
	@w{gcry_mpi_point_t @var{w}}, @w{gcry_mpi_point_t @var{u}}, @
	@w{gcry_mpi_point_t @var{v}}, @w{gcry_ctx_t @var{ctx}})

	Add the points @var{u} and @var{v} of the elliptic curve described by
	@var{ctx} and store the result into @var{w}.
	@end deftypefun

	@deftypefun void gcry_mpi_ec_sub ( @
	@w{gcry_mpi_point_t @var{w}}, @w{gcry_mpi_point_t @var{u}}, @
	@w{gcry_mpi_point_t @var{v}}, @w{gcry_ctx_t @var{ctx}})

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

	@deftypefun void gcry_mpi_ec_mul ( @
	@w{gcry_mpi_point_t @var{w}}, @w{gcry_mpi_t @var{n}}, @
	@w{gcry_mpi_point_t @var{u}}, @w{gcry_ctx_t @var{ctx}})

	Multiply the point @var{u} of the elliptic curve described by
	@var{ctx} by @var{n} and store the result into @var{w}.
	@end deftypefun

	@deftypefun int gcry_mpi_ec_curve_point ( @
	@w{gcry_mpi_point_t @var{point}}, @w{gcry_ctx_t @var{ctx}})

	Return true if @var{point} is on the elliptic curve described by
	@var{ctx}.
	@end deftypefun


	@node Miscellaneous
	@section Miscellaneous

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

	@deftypefun gcry_mpi_t gcry_mpi_set_opaque (@w{gcry_mpi_t @var{a}}, @w{void *@var{p}}, @w{unsigned int @var{nbits}})

	Store @var{nbits} of the value @var{p} points to in @var{a} and mark
	@var{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
	@var{a}). Ownership of @var{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 @var{p} has been allocated in a way
	that @code{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.
	@end deftypefun

	@deftypefun gcry_mpi_t gcry_mpi_set_opaque_copy (@w{gcry_mpi_t @var{a}}, @w{const void *@var{p}}, @w{unsigned int @var{nbits}})

	Same as @code{gcry_mpi_set_opaque} but ownership of @var{p} is not
	taken instead a copy of @var{p} is used.
	@end deftypefun


	@deftypefun {void } gcry_mpi_get_opaque (@w{gcry_mpi_t @var{a}}, @w{unsigned int @var{nbits}})

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

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

	@table @code
	@item GCRYMPI_FLAG_SECURE
	Setting this flag converts @var{a} into an MPI stored in "secure
	memory". Clearing this flag is not allowed.
	@item 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.
	@item 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.
	@item 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 @code{GCRYMPI_CONST_ONE}, @code{GCRYMPI_CONST_TWO},
	@code{GCRYMPI_CONST_THREE}, @code{GCRYMPI_CONST_FOUR}, and
	@code{GCRYMPI_CONST_EIGHT}.
	@item GCRYMPI_FLAG_USER1
	@itemx GCRYMPI_FLAG_USER2
	@itemx GCRYMPI_FLAG_USER3
	@itemx GCRYMPI_FLAG_USER4
	These flags are reserved for use by the application.
	@end table

	@deftypefun void gcry_mpi_set_flag (@w{gcry_mpi_t @var{a}}, @
	@w{enum gcry_mpi_flag @var{flag}})

	Set the @var{flag} for the MPI @var{a}. The only allowed flags are
	@code{GCRYMPI_FLAG_SECURE}, @code{GCRYMPI_FLAG_IMMUTABLE}, and
	@code{GCRYMPI_FLAG_CONST}.
	@end deftypefun

	@deftypefun void gcry_mpi_clear_flag (@w{gcry_mpi_t @var{a}}, @
	@w{enum gcry_mpi_flag @var{flag}})

	Clear @var{flag} for the multi-precision-integers @var{a}. The only
	allowed flag is @code{GCRYMPI_FLAG_IMMUTABLE} but only if
	@code{GCRYMPI_FLAG_CONST} is not set. If @code{GCRYMPI_FLAG_CONST} is
	set, clearing @code{GCRYMPI_FLAG_IMMUTABLE} will simply be ignored.
	@end deftypefun
	o
	@deftypefun int gcry_mpi_get_flag (@w{gcry_mpi_t @var{a}}, @
	@w{enum gcry_mpi_flag @var{flag}})

	Return true if @var{flag} is set for @var{a}.
	@end deftypefun


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

	@deftypefun void gcry_mpi_randomize (@w{gcry_mpi_t @var{w}}, @w{unsigned int @var{nbits}}, @w{enum gcry_random_level @var{level}})

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

	@c **********************************************************
	@c ****************** Prime numbers *********************
	@c **********************************************************
	@node Prime numbers
	@chapter Prime numbers

	@menu
	* Generation:: Generation of new prime numbers.
	* Checking:: Checking if a given number is prime.
	@end menu

	@node Generation
	@section Generation

	@deftypefun gcry_error_t gcry_prime_generate (gcry_mpi_t @var{prime},unsigned int @var{prime_bits}, unsigned int @var{factor_bits}, gcry_mpi_t @var{factors}, gcry_prime_check_func_t @var{cb_func}, void @var{cb_arg}, gcry_random_level_t @var{random_level}, unsigned int @var{flags})

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

	@deftypefun gcry_error_t gcry_prime_group_generator (gcry_mpi_t @var{r_g}, gcry_mpi_t @var{prime}, gcry_mpi_t @var{factors}, gcry_mpi_t @var{start_g})

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

	@deftypefun void gcry_prime_release_factors (gcry_mpi_t *@var{factors})

	Convenience function to release the @var{factors} array.
	@end deftypefun

	@node Checking
	@section Checking

	@deftypefun gcry_error_t gcry_prime_check (gcry_mpi_t @var{p}, unsigned int @var{flags})

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

	@c **********************************************************
	@c ****************** Utilities *************************
	@c **********************************************************
	@node Utilities
	@chapter Utilities

	@menu
	* Memory allocation:: Functions related with memory allocation.
	* Context management:: Functions related with context management.
	* Buffer description:: A data type to describe buffers.
	* Config reporting:: How to return Libgcrypt's configuration.
	@end menu


	@node Memory allocation
	@section Memory allocation

	@deftypefun {void *} gcry_malloc (size_t @var{n})

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

	@deftypefun {void *} gcry_malloc_secure (size_t @var{n})
	Like @code{gcry_malloc}, but uses secure memory.
	@end deftypefun

	@deftypefun {void *} gcry_calloc (size_t @var{n}, size_t @var{m})

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

	@deftypefun {void *} gcry_calloc_secure (size_t @var{n}, size_t @var{m})
	Like @code{gcry_calloc}, but uses secure memory.
	@end deftypefun

	@deftypefun {void } gcry_realloc (void @var{p}, size_t @var{n})

	This function tries to resize the memory area pointed to by @var{p} to
	@var{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 @var{p} is secure memory or not,
	gcry_realloc tries to use secure memory as well.
	@end deftypefun

	@deftypefun void gcry_free (void *@var{p})
	Release the memory area pointed to by @var{p}.
	@end deftypefun


	@node Context management
	@section 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.

	@deftp {Data type} {gcry_ctx_t}
	This type is used to refer to the general purpose context object.
	@end deftp

	@anchor{gcry_ctx_release}
	@deftypefun void gcry_ctx_release (gcry_ctx_t @var{ctx})
	Release the context object @var{ctx} and all associated resources. A
	@code{NULL} passed as @var{ctx} is ignored.
	@end deftypefun

	@node Buffer description
	@section Buffer description

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

	@deftp {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:

	@table @code
	@item .size
	This is either 0 for no information available or indicates the
	allocated length of the buffer.
	@item .off
	This is the offset into the buffer.
	@item .len
	This is the valid length of the buffer starting at @code{.off}.
	@item .data
	This is the address of the buffer.
	@end table
	@end deftp

	@node Config reporting
	@section How to return Libgcrypt's configuration.

	Although @code{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:

	@deftypefun {char *} gcry_get_config @
	(@w{int @var{mode}}, @
	@w{const char *@var{what}})

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

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


	@c **********************************************************
	@c ******************* Tools **************************
	@c **********************************************************
	@node Tools
	@chapter Tools

	@menu
	* hmac256:: A standalone HMAC-SHA-256 implementation
	@end menu

	@manpage hmac256.1
	@node hmac256
	@section A HMAC-SHA-256 tool
	@ifset manverb
	.B hmac256
	\- Compute an HMAC-SHA-256 MAC
	@end ifset

	@mansect synopsis
	@ifset manverb
	.B hmac256
	.RB [ \-\-binary ]
	.I key
	.I [FILENAME]
	@end ifset

	@mansect description
	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.

	@noindent
	@command{hmac256} is commonly invoked as

	@example
	hmac256 "This is my key" foo.txt
	@end example

	@noindent
	This compute the MAC on the file @file{foo.txt} using the key given on
	the command line.

	@mansect options
	@noindent
	@command{hmac256} understands these options:

	@table @gnupgtabopt

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

	@item --version
	Print version of the program and exit.

	@end table

	@mansect see also
	@ifset isman
	@command{sha256sum}(1)
	@end ifset
	@manpause

	@c **********************************************************
	@c ************** Environment Variables ***************
	@c **********************************************************
	@node Configuration
	@chapter Configuration files and environment variables

	This chapter describes which files and environment variables can be
	used to change the behaviour of Libgcrypt.

	@noindent
	The environment variables considered by Libgcrypt are:

	@table @code

	@item GCRYPT_BARRETT
	@cindex GCRYPT_BARRETT
	By setting this variable to any value a different algorithm for
	modular reduction is used for ECC.

	@item GCRYPT_RNDUNIX_DBG
	@item GCRYPT_RNDUNIX_DBGALL
	@cindex GCRYPT_RNDUNIX_DBG
	@cindex GCRYPT_RNDUNIX_DBGALL
	These two environment variables are used to enable debug output for
	the rndunix entropy gatherer, which is used on systems lacking a
	/dev/random device. The value of @code{GCRYPT_RNDUNIX_DBG} is a file
	name or @code{-} for stdout. Debug output is the written to this
	file. By setting @code{GCRYPT_RNDUNIX_DBGALL} to any value the debug
	output will be more verbose.

	@item GCRYPT_RNDW32_NOPERF
	@cindex GCRYPT_RNDW32_NOPERF
	Setting this environment variable on Windows to any value disables
	the use of performance data (@code{HKEY_PERFORMANCE_DATA}) as source
	for entropy. On some older Windows systems this could help to speed
	up the creation of random numbers but also decreases the amount of
	data used to init the random number generator.

	@item GCRYPT_RNDW32_DBG
	@cindex GCRYPT_RNDW32_DBG
	Setting the value of this variable to a positive integer logs
	information about the Windows entropy gatherer using the standard log
	interface.


	@item HOME
	@cindex HOME
	This is used to locate the socket to connect to the EGD random
	daemon. The EGD can be used on system without a /dev/random to speed
	up the random number generator. It is not needed on the majority of
	today's operating systems and support for EGD requires the use of a
	configure option at build time.

	@end table

	@noindent
	The files which Libgcrypt uses to retrieve system information and the
	files which can be created by the user to modify Libgcrypt's behavior
	are:

	@table @file

	@item /etc/gcrypt/hwf.deny
	@cindex /etc/gcrypt/hwf.deny
	This file can be used to disable the use of hardware based
	optimizations, @pxref{hardware features}.


	@item /etc/gcrypt/random.conf
	@cindex /etc/gcrypt/random.conf
	This file can be used to globally change parameters of the random
	generator. The file is a simple text file where empty lines and
	lines with the first non white-space character being '#' are
	ignored. Supported options are

	@table @file
	@item disable-jent
	@cindex disable-jent
	Disable the use of the jitter based entropy generator.

	@item only-urandom
	@cindex only-urandom
	Always use the non-blocking /dev/urandom or the respective system call
	instead of the blocking /dev/random. If Libgcrypt is used early in
	the boot process of the system, this option should only be used if the
	system also supports the getrandom system call.

	@end table

	@item /etc/gcrypt/fips_enabled
	@itemx /proc/sys/crypto/fips_enabled
	@cindex /etc/gcrypt/fips_enabled
	@cindex fips_enabled
	On Linux these files are used to enable FIPS mode, @pxref{enabling fips mode}.

	@item /proc/cpuinfo
	@itemx /proc/self/auxv
	@cindex /proc/cpuinfo
	@cindex /proc/self/auxv
	On Linux running on the ARM architecture, these files are used to read
	hardware capabilities of the CPU.

	@end table


	@c **********************************************************
	@c *************** Architecure Overview ***************
	@c **********************************************************
	@node Architecture
	@chapter Architecture

	This chapter describes the internal architecture of Libgcrypt.

	Libgcrypt is a function library written in ISO C-90. Any compliant
	compiler should be able to build Libgcrypt as long as the target is
	either a POSIX platform or compatible to the API used by Windows NT.
	Provisions have been take so that the library can be directly used from
	C++ applications; however building with a C++ compiler is not supported.

	Building Libgcrypt is done by using the common @code{./configure && make}
	approach. The configure command is included in the source distribution
	and as a portable shell script it works on any Unix-alike system. The
	result of running the configure script are a C header file
	(@file{config.h}), customized Makefiles, the setup of symbolic links and
	a few other things. After that the make tool builds and optionally
	installs the library and the documentation. See the files
	@file{INSTALL} and @file{README} in the source distribution on how to do
	this.

	Libgcrypt is developed using a Subversion@footnote{A version control
	system available for many platforms} repository. Although all released
	versions are tagged in this repository, they should not be used to build
	production versions of Libgcrypt. Instead released tarballs should be
	used. These tarballs are available from several places with the master
	copy at @indicateurl{ftp://ftp.gnupg.org/gcrypt/libgcrypt/}.
	Announcements of new releases are posted to the
	@indicateurl{gnupg-announce@@gnupg.org} mailing list@footnote{See
	@url{http://www.gnupg.org/documentation/mailing-lists.en.html} for
	details.}.


	@float Figure,fig:subsystems
	@caption{Libgcrypt subsystems}
	@center @image{libgcrypt-modules, 150mm,,Libgcrypt subsystems}
	@end float

	Libgcrypt consists of several subsystems (@pxref{fig:subsystems}) and
	all these subsystems provide a public API; this includes the helper
	subsystems like the one for S-expressions. The API style depends on the
	subsystem; in general an open-use-close approach is implemented. The
	open returns a handle to a context used for all further operations on
	this handle, several functions may then be used on this handle and a
	final close function releases all resources associated with the handle.

	@menu
	* Public-Key Subsystem Architecture:: About public keys.
	* Symmetric Encryption Subsystem Architecture:: About standard ciphers.
	* Hashing and MACing Subsystem Architecture:: About hashing.
	* Multi-Precision-Integer Subsystem Architecture:: About big integers.
	* Prime-Number-Generator Subsystem Architecture:: About prime numbers.
	* Random-Number Subsystem Architecture:: About random stuff.
	@c * Helper Subsystems Architecture:: About other stuff.
	@end menu



	@node Public-Key Subsystem Architecture
	@section 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.
	@c see @ref{S-expression Subsystem Architecture}.

	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:

	@table @code

	@item gcry_pk_encrypt
	Encrypt data using a public key.

	@item gcry_pk_decrypt
	Decrypt data using a private key.

	@item gcry_pk_sign
	Sign data using a private key.

	@item gcry_pk_verify
	Verify that a signature matches the data.

	@item gcry_pk_testkey
	Perform a consistency over a public or private key.

	@item gcry_pk_genkey
	Create a new public/private key pair.

	@end table

	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 (@file{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 @math{y = x r^{e} \bmod n}
	is decrypted and the unblinded value @math{x' = y' r^{-1} \bmod n}
	returned. The blinding value @math{r} is a random value with the size
	of the modulus @math{n} and generated with @code{GCRY_WEAK_RANDOM}
	random level.

	@cindex X9.31
	@cindex FIPS 186
	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.



	@node Symmetric Encryption Subsystem Architecture
	@section Symmetric Encryption Subsystem Architecture

	The interface to work with symmetric encryption algorithms is made up
	of functions from the @code{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
	(@file{cipher/cipher.c}) implements the modes and calls the core
	algorithm functions to process each block.

	The most important functions are:

	@table @code

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

	@item gcry_cipher_close
	Release an instance.

	@item gcry_cipher_setkey
	Set a key to be used for encryption or decryption.

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

	@item gcry_cipher_encrypt
	@itemx 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.

	@end table

	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.



	@node Hashing and MACing Subsystem Architecture
	@section Hashing and MACing Subsystem Architecture

	The interface to work with message digests and CRC algorithms is made
	up of functions from the @code{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:

	@table @code
	@item 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.

	@item gcry_md_enable
	Enable an additional algorithm for the instance.

	@item gcry_md_setkey
	Set the key for the MAC.

	@item 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.

	@item gcry_md_putc
	Buffered version of @code{gcry_md_write} implemented as a macro.

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

	@item gcry_md_close
	Release an instance

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

	@end table

	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.



	@node Multi-Precision-Integer Subsystem Architecture
	@section 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.

	@noindent
	Major features of Libgcrypt's multi-precision-integer code compared to
	GMP are:

	@itemize
	@item
	Avoidance of stack based allocations to allow protection against
	swapping out of sensitive data and for easy zeroing of sensitive
	intermediate results.

	@item
	Optional use of secure memory and tracking of its use so that results
	are also put into secure memory.

	@item
	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.

	@item
	Removal of unnecessary code to reduce complexity.

	@item
	Functions specialized for public key cryptography.

	@end itemize



	@node Prime-Number-Generator Subsystem Architecture
	@section 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.

	The generation of random prime numbers is based on the Lim and Lee
	algorithm to create practically save primes.@footnote{Chae Hoon Lim
	and Pil Joong Lee. A key recovery attack on discrete log-based schemes
	using a prime order subgroup. In Burton S. Kaliski Jr., editor,
	Advances in Cryptology: Crypto '97, pages 249-263, Berlin /
	Heidelberg / New York, 1997. Springer-Verlag. Described on page 260.}
	This algorithm creates a pool of smaller primes, select a few of them
	to create candidate primes of the form @math{2 * p_0 * p_1 * ... * p_n
	+ 1}, tests the candidate for primality and permutates the pool until
	a prime has been found. It is possible to clamp one of the small
	primes to a certain size to help DSA style algorithms. Because most
	of the small primes in the pool are not used for the resulting prime
	number, they are saved for later use (see @code{save_pool_prime} and
	@code{get_pool_prime} in @file{cipher/primegen.c}). The prime
	generator optionally supports the finding of an appropriate generator.

	@noindent
	The primality test works in three steps:

	@enumerate
	@item
	The standard sieve algorithm using the primes up to 4999 is used as a
	quick first check.

	@item
	A Fermat test filters out almost all non-primes.

	@item
	A 5 round Rabin-Miller test is finally used. The first round uses a
	witness of 2, whereas the next rounds use a random witness.

	@end enumerate

	To support the generation of RSA and DSA keys in FIPS mode according
	to X9.31 and FIPS 186-2, Libgcrypt implements two additional prime
	generation functions: @code{_gcry_derive_x931_prime} and
	@code{_gcry_generate_fips186_2_prime}. These functions are internal
	and not available through the public API.



	@node Random-Number Subsystem Architecture
	@section Random-Number Subsystem Architecture

	Libgcrypt provides 3 levels or random quality: The level
	@code{GCRY_VERY_STRONG_RANDOM} usually used for key generation, the
	level @code{GCRY_STRONG_RANDOM} for all other strong random
	requirements and the function @code{gcry_create_nonce} which is used
	for weaker usages like nonces. There is also a level
	@code{GCRY_WEAK_RANDOM} which in general maps to
	@code{GCRY_STRONG_RANDOM} except when used with the function
	@code{gcry_mpi_randomize}, where it randomizes an
	multi-precision-integer using the @code{gcry_create_nonce} function.

	@noindent
	There are two distinct random generators available:

	@itemize
	@item
	The Continuously Seeded Pseudo Random Number Generator (CSPRNG), which
	is based on the classic GnuPG derived big pool implementation.
	Implemented in @code{random/random-csprng.c} and used by default.
	@item
	A FIPS approved ANSI X9.31 PRNG using AES with a 128 bit key. Implemented in
	@code{random/random-fips.c} and used if Libgcrypt is in FIPS mode.
	@end itemize

	@noindent
	Both generators make use of so-called entropy gathering modules:

	@table @asis
	@item rndlinux
	Uses the operating system provided @file{/dev/random} and
	@file{/dev/urandom} devices. The @file{/dev/gcrypt/random.conf}
	config option @option{only-urandom} can be used to inhibit the use of
	the blocking @file{/dev/random} device.

	@item rndunix
	Runs several operating system commands to collect entropy from sources
	like virtual machine and process statistics. It is a kind of
	poor-man's @code{/dev/random} implementation. It is not available in
	FIPS mode.

	@item rndegd
	Uses the operating system provided Entropy Gathering Daemon (EGD).
	The EGD basically uses the same algorithms as rndunix does. However
	as a system daemon it keeps on running and thus can serve several
	processes requiring entropy input and does not waste collected entropy
	if the application does not need all the collected entropy. It is not
	available in FIPS mode.

	@item rndw32
	Targeted for the Microsoft Windows OS. It uses certain properties of
	that system and is the only gathering module available for that OS.

	@item rndhw
	Extra module to collect additional entropy by utilizing a hardware
	random number generator. As of now the supported hardware RNG is
	the Padlock engine of VIA (Centaur) CPUs and x86 CPUs with the RDRAND
	instruction. It is not available in FIPS mode.

	@item rndjent
	Extra module to collect additional entropy using a CPU jitter based
	approach. This is only used on X86 hardware where the RDTSC opcode is
	available. The @file{/dev/gcrypt/random.conf} config option
	@option{disable-jent} can be used to inhibit the use of this module.

	@end table


	@menu
	* CSPRNG Description:: Description of the CSPRNG.
	* FIPS PRNG Description:: Description of the FIPS X9.31 PRNG.
	@end menu


	@node CSPRNG Description
	@subsection Description of the CSPRNG

	This random number generator is loosely modelled after the one
	described in Peter Gutmann's paper: "Software Generation of
	Practically Strong Random Numbers".@footnote{Also described in chapter
	6 of his book "Cryptographic Security Architecture", New York, 2004,
	ISBN 0-387-95387-6.}

	A pool of 600 bytes is used and mixed using the core SHA-1 hash
	transform function. Several extra features are used to make the
	robust against a wide variety of attacks and to protect against
	failures of subsystems. The state of the generator may be saved to a
	file and initially seed form a file.

	Depending on how Libgcrypt was build the generator is able to select
	the best working entropy gathering module. It makes use of the slow
	and fast collection methods and requires the pool to initially seeded
	form the slow gatherer or a seed file. An entropy estimation is used
	to mix in enough data from the gather modules before returning the
	actual random output. Process fork detection and protection is
	implemented.

	@c FIXME: The design and implementation needs a more verbose description.

	The implementation of the nonce generator (for
	@code{gcry_create_nonce}) is a straightforward repeated hash design: A
	28 byte buffer is initially seeded with the PID and the time in
	seconds in the first 20 bytes and with 8 bytes of random taken from
	the @code{GCRY_STRONG_RANDOM} generator. Random numbers are then
	created by hashing all the 28 bytes with SHA-1 and saving that again
	in the first 20 bytes. The hash is also returned as result.


	@node FIPS PRNG Description
	@subsection Description of the FIPS X9.31 PRNG

	The core of this deterministic random number generator is implemented
	according to the document ``NIST-Recommended Random Number Generator
	Based on ANSI X9.31 Appendix A.2.4 Using the 3-Key Triple DES and AES
	Algorithms'', dated 2005-01-31. This implementation uses the AES
	variant.

	The generator is based on contexts to utilize the same core functions
	for all random levels as required by the high-level interface. All
	random generators return their data in 128 bit blocks. If the caller
	requests less bits, the extra bits are not used. The key for each
	generator is only set once at the first time a generator context is
	used. The seed value is set along with the key and again after 1000
	output blocks.

	On Unix like systems the @code{GCRY_VERY_STRONG_RANDOM} and
	@code{GCRY_STRONG_RANDOM} generators are keyed and seeded using the
	rndlinux module with the @file{/dev/random} device. Thus these
	generators may block until the OS kernel has collected enough entropy.
	When used with Microsoft Windows the rndw32 module is used instead.

	The generator used for @code{gcry_create_nonce} is keyed and seeded
	from the @code{GCRY_STRONG_RANDOM} generator. Thus is may also block
	if the @code{GCRY_STRONG_RANDOM} generator has not yet been used
	before and thus gets initialized on the first use by
	@code{gcry_create_nonce}. This special treatment is justified by the
	weaker requirements for a nonce generator and to save precious kernel
	entropy for use by the ``real'' random generators.

	A self-test facility uses a separate context to check the
	functionality of the core X9.31 functions using a known answers test.
	During runtime each output block is compared to the previous one to
	detect a stuck generator.

	The DT value for the generator is made up of the current time down to
	microseconds (if available) and a free running 64 bit counter. When
	used with the test context the DT value is taken from the context and
	incremented on each use.

	@c @node Helper Subsystems Architecture
	@c @section Helper Subsystems Architecture
	@c
	@c There are a few smaller subsystems which are mainly used internally by
	@c Libgcrypt but also available to applications.
	@c
	@c @menu
	@c * S-expression Subsystem Architecture:: Details about the S-expression architecture.
	@c * Memory Subsystem Architecture:: Details about the memory allocation architecture.
	@c * Miscellaneous Subsystems Architecture:: Details about other subsystems.
	@c @end menu
	@c
	@c @node S-expression Subsystem Architecture
	@c @subsection S-expression Subsystem Architecture
	@c
	@c Libgcrypt provides an interface to S-expression to create and parse
	@c them. To use an S-expression with Libgcrypt it needs first be
	@c converted into the internal representation used by Libgcrypt (the type
	@c @code{gcry_sexp_t}). The conversion functions support a large subset
	@c of the S-expression specification and further feature a printf like
	@c function to convert a list of big integers or other binary data into
	@c an S-expression.
	@c
	@c Libgcrypt currently implements S-expressions using a tagged linked
	@c list. However this is not exposed to an application and may be
	@c changed in future releases to reduce overhead when already working
	@c with canonically encoded S-expressions. Secure memory is supported by
	@c this S-expressions implementation.
	@c
	@c @node Memory Subsystem Architecture
	@c @subsection Memory Subsystem Architecture
	@c
	@c TBD.
	@c
	@c
	@c @node Miscellaneous Subsystems Architecture
	@c @subsection Miscellaneous Subsystems Architecture
	@c
	@c TBD.
	@c
	@c



	@c **********************************************************
	@c ***************** Appendices ***********************
	@c **********************************************************

	@c ********************************************
	@node Self-Tests
	@appendix Description of the Self-Tests

	In addition to the build time regression test suite, Libgcrypt
	implements self-tests to be performed at runtime. Which self-tests
	are actually used depends on the mode Libgcrypt is used in. In
	standard mode a limited set of self-tests is run at the time an
	algorithm is first used. Note that not all algorithms feature a
	self-test in standard mode. The @code{GCRYCTL_SELFTEST} control
	command may be used to run all implemented self-tests at any time;
	this will even run more tests than those run in FIPS mode.

	If any of the self-tests fails, the library immediately returns an
	error code to the caller. If Libgcrypt is in FIPS mode the self-tests
	will be performed within the ``Self-Test'' state and any failure puts
	the library into the ``Error'' state.

	@c --------------------------------
	@section Power-Up Tests

	Power-up tests are only performed if Libgcrypt is in FIPS mode.

	@subsection Symmetric Cipher Algorithm Power-Up Tests

	The following symmetric encryption algorithm tests are run during
	power-up:

	@table @asis
	@item 3DES
	To test the 3DES 3-key EDE encryption in ECB mode these tests are
	run:
	@enumerate
	@item
	A known answer test is run on a 64 bit test vector processed by 64
	rounds of Single-DES block encryption and decryption using a key
	changed with each round.
	@item
	A known answer test is run on a 64 bit test vector processed by 16
	rounds of 2-key and 3-key Triple-DES block encryption and decryptions
	using a key changed with each round.
	@item
	10 known answer tests using 3-key Triple-DES EDE encryption, comparing
	the ciphertext to the known value, then running a decryption and
	comparing it to the initial plaintext.
	@end enumerate
	(@code{cipher/des.c:selftest})

	@item AES-128
	A known answer tests is run using one test vector and one test
	key with AES in ECB mode. (@code{cipher/rijndael.c:selftest_basic_128})

	@item AES-192
	A known answer tests is run using one test vector and one test
	key with AES in ECB mode. (@code{cipher/rijndael.c:selftest_basic_192})

	@item AES-256
	A known answer tests is run using one test vector and one test key
	with AES in ECB mode. (@code{cipher/rijndael.c:selftest_basic_256})
	@end table

	@subsection Hash Algorithm Power-Up Tests

	The following hash algorithm tests are run during power-up:

	@table @asis
	@item SHA-1
	A known answer test using the string @code{"abc"} is run.
	(@code{cipher/@/sha1.c:@/selftests_sha1})
	@item SHA-224
	A known answer test using the string @code{"abc"} is run.
	(@code{cipher/@/sha256.c:@/selftests_sha224})
	@item SHA-256
	A known answer test using the string @code{"abc"} is run.
	(@code{cipher/@/sha256.c:@/selftests_sha256})
	@item SHA-384
	A known answer test using the string @code{"abc"} is run.
	(@code{cipher/@/sha512.c:@/selftests_sha384})
	@item SHA-512
	A known answer test using the string @code{"abc"} is run.
	(@code{cipher/@/sha512.c:@/selftests_sha512})
	@end table

	@subsection MAC Algorithm Power-Up Tests

	The following MAC algorithm tests are run during power-up:

	@table @asis
	@item HMAC SHA-1
	A known answer test using 9 byte of data and a 64 byte key is run.
	(@code{cipher/hmac-tests.c:selftests_sha1})
	@item HMAC SHA-224
	A known answer test using 28 byte of data and a 4 byte key is run.
	(@code{cipher/hmac-tests.c:selftests_sha224})
	@item HMAC SHA-256
	A known answer test using 28 byte of data and a 4 byte key is run.
	(@code{cipher/hmac-tests.c:selftests_sha256})
	@item HMAC SHA-384
	A known answer test using 28 byte of data and a 4 byte key is run.
	(@code{cipher/hmac-tests.c:selftests_sha384})
	@item HMAC SHA-512
	A known answer test using 28 byte of data and a 4 byte key is run.
	(@code{cipher/hmac-tests.c:selftests_sha512})
	@end table

	@subsection Random Number Power-Up Test

	The DRNG is tested during power-up this way:

	@enumerate
	@item
	Requesting one block of random using the public interface to check
	general working and the duplicated block detection.
	@item
	3 know answer tests using pre-defined keys, seed and initial DT
	values. For each test 3 blocks of 16 bytes are requested and compared
	to the expected result. The DT value is incremented for each block.
	@end enumerate

	@subsection Public Key Algorithm Power-Up Tests

	The public key algorithms are tested during power-up:

	@table @asis
	@item RSA
	A pre-defined 1024 bit RSA key is used and these tests are run
	in turn:
	@enumerate
	@item
	Conversion of S-expression to internal format.
	(@code{cipher/@/rsa.c:@/selftests_rsa})
	@item
	Private key consistency check.
	(@code{cipher/@/rsa.c:@/selftests_rsa})
	@item
	A pre-defined 20 byte value is signed with PKCS#1 padding for SHA-1.
	The result is verified using the public key against the original data
	and against modified data. (@code{cipher/@/rsa.c:@/selftest_sign_1024})
	@item
	A 1000 bit random value is encrypted and checked that it does not
	match the original random value. The encrypted result is then
	decrypted and checked that it matches the original random value.
	(@code{cipher/@/rsa.c:@/selftest_encr_1024})
	@end enumerate

	@item DSA
	A pre-defined 1024 bit DSA key is used and these tests are run in turn:
	@enumerate
	@item
	Conversion of S-expression to internal format.
	(@code{cipher/@/dsa.c:@/selftests_dsa})
	@item
	Private key consistency check.
	(@code{cipher/@/dsa.c:@/selftests_dsa})
	@item
	A pre-defined 20 byte value is signed with PKCS#1 padding for
	SHA-1. The result is verified using the public key against the
	original data and against modified data.
	(@code{cipher/@/dsa.c:@/selftest_sign_1024})
	@end enumerate
	@end table

	@subsection Integrity Power-Up Tests

	The integrity of the Libgcrypt is tested during power-up but only if
	checking has been enabled at build time. The check works by computing
	a HMAC SHA-256 checksum over the file used to load Libgcrypt into
	memory. That checksum is compared against a checksum stored in a file
	of the same name but with a single dot as a prefix and a suffix of
	@file{.hmac}.


	@subsection Critical Functions Power-Up Tests

	The 3DES weak key detection is tested during power-up by calling the
	detection function with keys taken from a table listening all weak
	keys. The table itself is protected using a SHA-1 hash.
	(@code{cipher/@/des.c:@/selftest})



	@c --------------------------------
	@section Conditional Tests

	The conditional tests are performed if a certain condition is met.
	This may occur at any time; the library does not necessary enter the
	``Self-Test'' state to run these tests but will transit to the
	``Error'' state if a test failed.

	@subsection Key-Pair Generation Tests

	After an asymmetric key-pair has been generated, Libgcrypt runs a
	pair-wise consistency tests on the generated key. On failure the
	generated key is not used, an error code is returned and, if in FIPS
	mode, the library is put into the ``Error'' state.

	@table @asis
	@item RSA
	The test uses a random number 64 bits less the size of the modulus as
	plaintext and runs an encryption and decryption operation in turn. The
	encrypted value is checked to not match the plaintext and the result
	of the decryption is checked to match the plaintext.

	A new random number of the same size is generated, signed and verified
	to test the correctness of the signing operation. As a second signing
	test, the signature is modified by incrementing its value and then
	verified with the expected result that the verification fails.
	(@code{cipher/@/rsa.c:@/test_keys})
	@item DSA
	The test uses a random number of the size of the Q parameter to create
	a signature and then checks that the signature verifies. As a second
	signing test, the data is modified by incrementing its value and then
	verified against the signature with the expected result that the
	verification fails. (@code{cipher/@/dsa.c:@/test_keys})
	@end table


	@subsection Software Load Tests

	No code is loaded at runtime.

	@subsection Manual Key Entry Tests

	A manual key entry feature is not implemented in Libgcrypt.


	@subsection Continuous RNG Tests

	The continuous random number test is only used in FIPS mode. The RNG
	generates blocks of 128 bit size; the first block generated per
	context is saved in the context and another block is generated to be
	returned to the caller. Each block is compared against the saved
	block and then stored in the context. If a duplicated block is
	detected an error is signaled and the library is put into the
	``Fatal-Error'' state.
	(@code{random/@/random-fips.c:@/x931_aes_driver})



	@c --------------------------------
	@section Application Requested Tests

	The application may requests tests at any time by means of the
	@code{GCRYCTL_SELFTEST} control command. Note that using these tests
	is not FIPS conform: Although Libgcrypt rejects all application
	requests for services while running self-tests, it does not ensure
	that no other operations of Libgcrypt are still being executed. Thus,
	in FIPS mode an application requesting self-tests needs to power-cycle
	Libgcrypt instead.

	When self-tests are requested, Libgcrypt runs all the tests it does
	during power-up as well as a few extra checks as described below.

	@subsection Symmetric Cipher Algorithm Tests

	The following symmetric encryption algorithm tests are run in addition
	to the power-up tests:

	@table @asis
	@item AES-128
	A known answer tests with test vectors taken from NIST SP800-38a and
	using the high level functions is run for block modes CFB and OFB.

	@end table

	@subsection Hash Algorithm Tests

	The following hash algorithm tests are run in addition to the
	power-up tests:

	@table @asis
	@item SHA-1
	@itemx SHA-224
	@itemx SHA-256
	@enumerate
	@item
	A known answer test using a 56 byte string is run.
	@item
	A known answer test using a string of one million letters "a" is run.
	@end enumerate
	(@code{cipher/@/sha1.c:@/selftests_sha1},
	@code{cipher/@/sha256.c:@/selftests_sha224},
	@code{cipher/@/sha256.c:@/selftests_sha256})
	@item SHA-384
	@item SHA-512
	@enumerate
	@item
	A known answer test using a 112 byte string is run.
	@item
	A known answer test using a string of one million letters "a" is run.
	@end enumerate
	(@code{cipher/@/sha512.c:@/selftests_sha384},
	@code{cipher/@/sha512.c:@/selftests_sha512})
	@end table

	@subsection MAC Algorithm Tests

	The following MAC algorithm tests are run in addition to the power-up
	tests:

	@table @asis
	@item HMAC SHA-1
	@enumerate
	@item
	A known answer test using 9 byte of data and a 20 byte key is run.
	@item
	A known answer test using 9 byte of data and a 100 byte key is run.
	@item
	A known answer test using 9 byte of data and a 49 byte key is run.
	@end enumerate
	(@code{cipher/hmac-tests.c:selftests_sha1})
	@item HMAC SHA-224
	@itemx HMAC SHA-256
	@itemx HMAC SHA-384
	@itemx HMAC SHA-512
	@enumerate
	@item
	A known answer test using 9 byte of data and a 20 byte key is run.
	@item
	A known answer test using 50 byte of data and a 20 byte key is run.
	@item
	A known answer test using 50 byte of data and a 26 byte key is run.
	@item
	A known answer test using 54 byte of data and a 131 byte key is run.
	@item
	A known answer test using 152 byte of data and a 131 byte key is run.
	@end enumerate
	(@code{cipher/@/hmac-tests.c:@/selftests_sha224},
	@code{cipher/@/hmac-tests.c:@/selftests_sha256},
	@code{cipher/@/hmac-tests.c:@/selftests_sha384},
	@code{cipher/@/hmac-tests.c:@/selftests_sha512})
	@end table


	@c ********************************************
	@node FIPS Mode
	@appendix Description of the FIPS Mode

	This appendix gives detailed information pertaining to the FIPS mode.
	In particular, the changes to the standard mode and the finite state
	machine are described. The self-tests required in this mode are
	described in the appendix on self-tests.

	@c -------------------------------
	@section Restrictions in FIPS Mode

	@noindent
	If Libgcrypt is used in FIPS mode these restrictions are effective:

	@itemize
	@item
	The cryptographic algorithms are restricted to this list:

	@table @asis
	@item GCRY_CIPHER_3DES
	3 key EDE Triple-DES symmetric encryption.
	@item GCRY_CIPHER_AES128
	AES 128 bit symmetric encryption.
	@item GCRY_CIPHER_AES192
	AES 192 bit symmetric encryption.
	@item GCRY_CIPHER_AES256
	AES 256 bit symmetric encryption.
	@item GCRY_MD_SHA1
	SHA-1 message digest.
	@item GCRY_MD_SHA224
	SHA-224 message digest.
	@item GCRY_MD_SHA256
	SHA-256 message digest.
	@item GCRY_MD_SHA384
	SHA-384 message digest.
	@item GCRY_MD_SHA512
	SHA-512 message digest.
	@item GCRY_MD_SHA1,GCRY_MD_FLAG_HMAC
	HMAC using a SHA-1 message digest.
	@item GCRY_MD_SHA224,GCRY_MD_FLAG_HMAC
	HMAC using a SHA-224 message digest.
	@item GCRY_MD_SHA256,GCRY_MD_FLAG_HMAC
	HMAC using a SHA-256 message digest.
	@item GCRY_MD_SHA384,GCRY_MD_FLAG_HMAC
	HMAC using a SHA-384 message digest.
	@item GCRY_MD_SHA512,GCRY_MD_FLAG_HMAC
	HMAC using a SHA-512 message digest.
	@item GCRY_PK_RSA
	RSA encryption and signing.
	@item GCRY_PK_DSA
	DSA signing.
	@end table

	Note that the CRC algorithms are not considered cryptographic algorithms
	and thus are in addition available.

	@item
	RSA key generation refuses to create a key with a keysize of
	less than 1024 bits.

	@item
	DSA key generation refuses to create a key with a keysize other
	than 1024 bits.

	@item
	The @code{transient-key} flag for RSA and DSA key generation is ignored.

	@item
	Support for the VIA Padlock engine is disabled.

	@item
	FIPS mode may only be used on systems with a /dev/random device.
	Switching into FIPS mode on other systems will fail at runtime.

	@item
	Saving and loading a random seed file is ignored.

	@item
	An X9.31 style random number generator is used in place of the
	large-pool-CSPRNG generator.

	@item
	The command @code{GCRYCTL_ENABLE_QUICK_RANDOM} is ignored.

	@item
	Message digest debugging is disabled.

	@item
	All debug output related to cryptographic data is suppressed.

	@item
	On-the-fly self-tests are not performed, instead self-tests are run
	before entering operational state.

	@item
	The function @code{gcry_set_allocation_handler} may not be used. In FIPS mode
	this function does not have any effect, because FIPS has a requirements for
	memory zeroization.

	@item
	The digest algorithm MD5 may not be used.

	@item
	In FIPS mode the command @code{GCRYCTL_DISABLE_SECMEM} is ignored.

	@item
	A handler set by @code{gcry_set_outofcore_handler} is ignored.
	@item
	A handler set by @code{gcry_set_fatalerror_handler} is ignored.

	@end itemize

	Note that when we speak about disabling FIPS mode, it merely means
	that the function @code{gcry_fips_mode_active} returns false; it does
	not mean that any non FIPS algorithms are allowed.

	@c ********************************************
	@section FIPS Finite State Machine

	The FIPS mode of libgcrypt implements a finite state machine (FSM) using
	8 states (@pxref{tbl:fips-states}) and checks at runtime that only valid
	transitions (@pxref{tbl:fips-state-transitions}) may happen.

	@float Figure,fig:fips-fsm
	@caption{FIPS mode state diagram}
	@center @image{fips-fsm,150mm,,FIPS FSM Diagram}
	@end float

	@float Table,tbl:fips-states
	@caption{FIPS mode states}
	@noindent
	States used by the FIPS FSM:
	@table @asis

	@item Power-Off
	Libgcrypt is not runtime linked to another application. This usually
	means that the library is not loaded into main memory. This state is
	documentation only.

	@item Power-On
	Libgcrypt is loaded into memory and API calls may be made. Compiler
	introduced constructor functions may be run. Note that Libgcrypt does
	not implement any arbitrary constructor functions to be called by the
	operating system

	@item Init
	The Libgcrypt initialization functions are performed and the library has
	not yet run any self-test.

	@item Self-Test
	Libgcrypt is performing self-tests.

	@item Operational
	Libgcrypt is in the operational state and all interfaces may be used.

	@item Error
	Libgrypt is in the error state. When calling any FIPS relevant
	interfaces they either return an error (@code{GPG_ERR_NOT_OPERATIONAL})
	or put Libgcrypt into the Fatal-Error state and won't return.

	@item Fatal-Error
	Libgcrypt is in a non-recoverable error state and
	will automatically transit into the Shutdown state.

	@item Shutdown
	Libgcrypt is about to be terminated and removed from the memory. The
	application may at this point still running cleanup handlers.

	@end table
	@end float


	@float Table,tbl:fips-state-transitions
	@caption{FIPS mode state transitions}
	@noindent
	The valid state transitions (@pxref{fig:fips-fsm}) are:
	@table @code
	@item 1
	Power-Off to Power-On is implicitly done by the OS loading Libgcrypt as
	a shared library and having it linked to an application.

	@item 2
	Power-On to Init is triggered by the application calling the
	Libgcrypt initialization function @code{gcry_check_version}.

	@item 3
	Init to Self-Test is either triggered by a dedicated API call or implicit
	by invoking a libgrypt service controlled by the FSM.

	@item 4
	Self-Test to Operational is triggered after all self-tests passed
	successfully.

	@item 5
	Operational to Shutdown is an artificial state without any direct action
	in Libgcrypt. When reaching the Shutdown state the library is
	deinitialized and can't return to any other state again.

	@item 6
	Shutdown to Power-off is the process of removing Libgcrypt from the
	computer's memory. For obvious reasons the Power-Off state can't be
	represented within Libgcrypt and thus this transition is for
	documentation only.

	@item 7
	Operational to Error is triggered if Libgcrypt detected an application
	error which can't be returned to the caller but still allows Libgcrypt
	to properly run. In the Error state all FIPS relevant interfaces return
	an error code.

	@item 8
	Error to Shutdown is similar to the Operational to Shutdown transition
	(5).

	@item 9
	Error to Fatal-Error is triggered if Libgrypt detects an fatal error
	while already being in Error state.

	@item 10
	Fatal-Error to Shutdown is automatically entered by Libgcrypt
	after having reported the error.

	@item 11
	Power-On to Shutdown is an artificial state to document that Libgcrypt
	has not ye been initialized but the process is about to terminate.

	@item 12
	Power-On to Fatal-Error will be triggered if certain Libgcrypt functions
	are used without having reached the Init state.

	@item 13
	Self-Test to Fatal-Error is triggered by severe errors in Libgcrypt while
	running self-tests.

	@item 14
	Self-Test to Error is triggered by a failed self-test.

	@item 15
	Operational to Fatal-Error is triggered if Libcrypt encountered a
	non-recoverable error.

	@item 16
	Operational to Self-Test is triggered if the application requested to run
	the self-tests again.

	@item 17
	Error to Self-Test is triggered if the application has requested to run
	self-tests to get to get back into operational state after an error.

	@item 18
	Init to Error is triggered by errors in the initialization code.

	@item 19
	Init to Fatal-Error is triggered by non-recoverable errors in the
	initialization code.

	@item 20
	Error to Error is triggered by errors while already in the Error
	state.


	@end table
	@end float

	@c ********************************************
	@section FIPS Miscellaneous Information

	Libgcrypt does not do any key management on itself; the application
	needs to care about it. Keys which are passed to Libgcrypt should be
	allocated in secure memory as available with the functions
	@code{gcry_malloc_secure} and @code{gcry_calloc_secure}. By calling
	@code{gcry_free} on this memory, the memory and thus the keys are
	overwritten with zero bytes before releasing the memory.

	For use with the random number generator, Libgcrypt generates 3
	internal keys which are stored in the encryption contexts used by the
	RNG. These keys are stored in secure memory for the lifetime of the
	process. Application are required to use @code{GCRYCTL_TERM_SECMEM}
	before process termination. This will zero out the entire secure
	memory and thus also the encryption contexts with these keys.



	@c **********************************************************
	@c *********** Appendices (license etc.) **************
	@c **********************************************************
	@include lgpl.texi

	@include gpl.texi

	@node Figures and Tables
	@unnumbered List of Figures and Tables

	@listoffloats Figure

	@listoffloats Table

	@node Concept Index
	@unnumbered Concept Index

	@printindex cp

	@node Function and Data Index
	@unnumbered Function and Data Index

	@printindex fn



	@bye

	-GCRYCTL_SET_RANDOM_DAEMON_SOCKET
	-GCRYCTL_USE_RANDOM_DAEMON
	-The random daemon is still a bit experimental, thus we do not document
	-them. Note that they should be used during initialization and that
	-these functions are not really thread safe.
	-
	-
	-
	-
	@c LocalWords: int HD
	diff --git a/random/Makefile.am b/random/Makefile.am
	index 434deaab..b4f0ae1f 100644
	--- a/random/Makefile.am
	+++ b/random/Makefile.am
	@@ -1,85 +1,80 @@
	# Makefile for cipher modules
	# Copyright (C) 2008 Free Software Foundation, Inc.
	#
	# This file is part of Libgcrypt.
	#
	# Libgcrypt is free software; you can redistribute it and/or modify
	# it under the terms of the GNU Lesser General Public License as
	# published by the Free Software Foundation; either version 2.1 of
	# the License, or (at your option) any later version.
	#
	# Libgcrypt is distributed in the hope that it will be useful,
	# but WITHOUT ANY WARRANTY; without even the implied warranty of
	# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
	# GNU Lesser General Public License for more details.
	#
	# You should have received a copy of the GNU Lesser General Public
	# License along with this program; if not, see <http://www.gnu.org/licenses/>.

	# Process this file with automake to produce Makefile.in

	# Need to include ../src in addition to top_srcdir because gcrypt.h is
	# a built header.
	AM_CPPFLAGS = -I../src -I$(top_srcdir)/src
	AM_CFLAGS = $(GPG_ERROR_CFLAGS)

	noinst_LTLIBRARIES = librandom.la

	GCRYPT_MODULES = @GCRYPT_RANDOM@

	librandom_la_DEPENDENCIES = $(GCRYPT_MODULES)
	librandom_la_LIBADD = $(GCRYPT_MODULES)

	librandom_la_SOURCES = \
	random.c random.h \
	rand-internal.h \
	random-csprng.c \
	random-drbg.c \
	random-system.c \
	rndjent.c \
	rndhw.c

	-if USE_RANDOM_DAEMON
	-librandom_la_SOURCES += random-daemon.c
	-endif USE_RANDOM_DAEMON
	-
	-
	EXTRA_librandom_la_SOURCES = \
	rndgetentropy.c \
	rndlinux.c \
	rndegd.c \
	rndunix.c \
	rndw32.c \
	rndw32ce.c \
	jitterentropy-gcd.c jitterentropy-gcd.h \
	jitterentropy-health.c jitterentropy-health.h \
	jitterentropy-noise.c jitterentropy-noise.h \
	jitterentropy-sha3.c jitterentropy-sha3.h \
	jitterentropy-timer.c jitterentropy-timer.h \
	jitterentropy-base.h \
	jitterentropy-base.c jitterentropy.h jitterentropy-base-user.h

	# The rndjent module needs to be compiled without optimization. */
	if ENABLE_O_FLAG_MUNGING
	o_flag_munging = sed -e 's/-O$[1-9sg][1-9sg]*$/-O0/g' -e 's/-Ofast/-O0/g'
	else
	o_flag_munging = cat
	endif

	rndjent.o: $(srcdir)/rndjent.c jitterentropy-base-user.h \
	$(srcdir)/jitterentropy-gcd.c $(srcdir)/jitterentropy-gcd.h \
	$(srcdir)/jitterentropy-health.c $(srcdir)/jitterentropy-health.h \
	$(srcdir)/jitterentropy-noise.c $(srcdir)/jitterentropy-noise.h \
	$(srcdir)/jitterentropy-sha3.c $(srcdir)/jitterentropy-sha3.h \
	$(srcdir)/jitterentropy-timer.c $(srcdir)/jitterentropy-timer.h \
	$(srcdir)/jitterentropy-base.c $(srcdir)/jitterentropy.h
	`echo $(COMPILE) -c $(srcdir)/rndjent.c \| $(o_flag_munging) `

	rndjent.lo: $(srcdir)/rndjent.c jitterentropy-base-user.h \
	$(srcdir)/jitterentropy-gcd.c $(srcdir)/jitterentropy-gcd.h \
	$(srcdir)/jitterentropy-health.c $(srcdir)/jitterentropy-health.h \
	$(srcdir)/jitterentropy-noise.c $(srcdir)/jitterentropy-noise.h \
	$(srcdir)/jitterentropy-sha3.c $(srcdir)/jitterentropy-sha3.h \
	$(srcdir)/jitterentropy-timer.c $(srcdir)/jitterentropy-timer.h \
	$(srcdir)/jitterentropy-base.c $(srcdir)/jitterentropy.h
	`echo $(LTCOMPILE) -c $(srcdir)/rndjent.c \| $(o_flag_munging) `
	diff --git a/random/rand-internal.h b/random/rand-internal.h
	index 1bfc7e4c..13d701c5 100644
	--- a/random/rand-internal.h
	+++ b/random/rand-internal.h
	@@ -1,154 +1,152 @@
	/* rand-internal.h - header to glue the random functions
	* Copyright (C) 1998, 2002 Free Software Foundation, Inc.
	*
	* This file is part of Libgcrypt.
	*
	* Libgcrypt is free software; you can redistribute it and/or modify
	* it under the terms of the GNU Lesser general Public License as
	* published by the Free Software Foundation; either version 2.1 of
	* the License, or (at your option) any later version.
	*
	* Libgcrypt is distributed in the hope that it will be useful,
	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
	* GNU Lesser General Public License for more details.
	*
	* You should have received a copy of the GNU Lesser General Public
	* License along with this program; if not, see <http://www.gnu.org/licenses/>.
	*/

	#ifndef G10_RAND_INTERNAL_H
	#define G10_RAND_INTERNAL_H

	#include "../src/cipher-proto.h"

	/* Constants used to define the origin of random added to the pool.
	The code is sensitive to the order of the values. */
	enum random_origins
	{
	RANDOM_ORIGIN_INIT = 0, /* Used only for initialization. */
	RANDOM_ORIGIN_EXTERNAL = 1, /* Added from an external source. */
	RANDOM_ORIGIN_FASTPOLL = 2, /* Fast random poll function. */
	RANDOM_ORIGIN_SLOWPOLL = 3, /* Slow poll function. */
	RANDOM_ORIGIN_EXTRAPOLL = 4 /* Used to mark an extra pool seed
	due to a GCRY_VERY_STRONG_RANDOM
	random request. */
	};

	#define RANDOM_CONF_DISABLE_JENT 1
	#define RANDOM_CONF_ONLY_URANDOM 2


	/-- random.c --/
	unsigned int _gcry_random_read_conf (void);
	void _gcry_random_progress (const char *what, int printchar,
	int current, int total);


	/-- random-csprng.c --/
	void _gcry_rngcsprng_initialize (int full);
	void _gcry_rngcsprng_close_fds (void);
	void _gcry_rngcsprng_dump_stats (void);
	void _gcry_rngcsprng_secure_alloc (void);
	void _gcry_rngcsprng_enable_quick_gen (void);
	-void _gcry_rngcsprng_set_daemon_socket (const char *socketname);
	-int _gcry_rngcsprng_use_daemon (int onoff);
	int _gcry_rngcsprng_is_faked (void);
	gcry_error_t _gcry_rngcsprng_add_bytes (const void *buf, size_t buflen,
	int quality);
	void *_gcry_rngcsprng_get_bytes (size_t nbytes,
	enum gcry_random_level level);
	void *_gcry_rngcsprng_get_bytes_secure (size_t nbytes,
	enum gcry_random_level level);
	void _gcry_rngcsprng_randomize (void *buffer, size_t length,
	enum gcry_random_level level);
	void _gcry_rngcsprng_set_seed_file (const char *name);
	void _gcry_rngcsprng_update_seed_file (void);
	void _gcry_rngcsprng_fast_poll (void);

	/-- random-drbg.c --/
	void _gcry_rngdrbg_inititialize (int full);
	void _gcry_rngdrbg_close_fds (void);
	void _gcry_rngdrbg_dump_stats (void);
	int _gcry_rngdrbg_is_faked (void);
	gcry_error_t _gcry_rngdrbg_add_bytes (const void *buf, size_t buflen,
	int quality);
	void _gcry_rngdrbg_randomize (void *buffer, size_t length,
	enum gcry_random_level level);
	gcry_error_t _gcry_rngdrbg_selftest (selftest_report_func_t report);

	/-- random-system.c --/
	void _gcry_rngsystem_initialize (int full);
	void _gcry_rngsystem_close_fds (void);
	void _gcry_rngsystem_dump_stats (void);
	int _gcry_rngsystem_is_faked (void);
	gcry_error_t _gcry_rngsystem_add_bytes (const void *buf, size_t buflen,
	int quality);
	void _gcry_rngsystem_randomize (void *buffer, size_t length,
	enum gcry_random_level level);



	/-- rndgetentropy.c --/
	int _gcry_rndgetentropy_gather_random (void (add) (const void , size_t,
	enum random_origins),
	enum random_origins origin,
	size_t length, int level);

	/-- rndlinux.c --/
	int _gcry_rndlinux_gather_random (void (add) (const void , size_t,
	enum random_origins),
	enum random_origins origin,
	size_t length, int level);

	/-- rndunix.c --/
	int _gcry_rndunix_gather_random (void (add) (const void , size_t,
	enum random_origins),
	enum random_origins origin,
	size_t length, int level);

	/-- rndegd.c --/
	int _gcry_rndegd_gather_random (void (add) (const void , size_t,
	enum random_origins),
	enum random_origins origin,
	size_t length, int level);
	int _gcry_rndegd_connect_socket (int nofail);

	/-- rndw32.c --/
	int _gcry_rndw32_gather_random (void (add) (const void , size_t,
	enum random_origins),
	enum random_origins origin,
	size_t length, int level);
	void _gcry_rndw32_gather_random_fast (void (add)(const void, size_t,
	enum random_origins),
	enum random_origins origin );

	/-- rndw32ce.c --/
	int _gcry_rndw32ce_gather_random (void (add) (const void , size_t,
	enum random_origins),
	enum random_origins origin,
	size_t length, int level);
	void _gcry_rndw32ce_gather_random_fast (void (add)(const void, size_t,
	enum random_origins),
	enum random_origins origin );

	/-- rndjent.c --/
	size_t _gcry_rndjent_poll (void (add)(const void,
	size_t, enum random_origins),
	enum random_origins origin,
	size_t length);
	void _gcry_rndjent_dump_stats (void);
	void _gcry_rndjent_fini (void);

	/-- rndhw.c --/
	int _gcry_rndhw_failed_p (void);
	void _gcry_rndhw_poll_fast (void (add)(const void, size_t,
	enum random_origins),
	enum random_origins origin);
	size_t _gcry_rndhw_poll_slow (void (add)(const void, size_t,
	enum random_origins),
	enum random_origins origin, size_t req_length);



	#endif /G10_RAND_INTERNAL_H/
	diff --git a/random/random-csprng.c b/random/random-csprng.c
	index f3ad48f3..aff93101 100644
	--- a/random/random-csprng.c
	+++ b/random/random-csprng.c
	@@ -1,1388 +1,1322 @@
	/* random-csprng.c - CSPRNG style random number generator (libgcrypt classic)
	* Copyright (C) 1998, 2000, 2001, 2002, 2003, 2004, 2005, 2006,
	* 2007, 2008, 2010, 2012 Free Software Foundation, Inc.
	*
	* This file is part of Libgcrypt.
	*
	* Libgcrypt is free software; you can redistribute it and/or modify
	* it under the terms of the GNU Lesser General Public License as
	* published by the Free Software Foundation; either version 2.1 of
	* the License, or (at your option) any later version.
	*
	* Libgcrypt is distributed in the hope that it will be useful,
	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
	* GNU Lesser General Public License for more details.
	*
	* You should have received a copy of the GNU Lesser General Public
	* License along with this program; if not, see <http://www.gnu.org/licenses/>.
	*/

	/*
	This random number generator is modelled after the one described in
	Peter Gutmann's 1998 Usenix Security Symposium paper: "Software
	Generation of Practically Strong Random Numbers". See also chapter
	6 in his book "Cryptographic Security Architecture", New York,
	2004, ISBN 0-387-95387-6.

	Note that the acronym CSPRNG stands for "Continuously Seeded
	PseudoRandom Number Generator" as used in Peter's implementation of
	the paper and not only for "Cryptographically Secure PseudoRandom
	Number Generator".
	*/


	#include <config.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <errno.h>
	#include <string.h>
	#include <sys/time.h>
	#include <sys/types.h>
	#include <sys/stat.h>
	#include <unistd.h>
	#include <fcntl.h>
	#include <time.h>
	#ifdef HAVE_GETHRTIME
	#include <sys/times.h>
	#endif
	#ifdef HAVE_GETTIMEOFDAY
	#include <sys/time.h>
	#endif
	#ifdef HAVE_GETRUSAGE
	#include <sys/resource.h>
	#endif
	#ifdef __MINGW32__
	#include <process.h>
	#endif
	#ifdef HAVE_W32_SYSTEM
	#include <windows.h>
	#endif
	#include "g10lib.h"
	#include "random.h"
	#include "rand-internal.h"
	#include "cipher.h" /* _gcry_sha1_hash_buffer */
	#include "../cipher/sha1.h" /* _gcry_sha1_mixblock */

	#ifndef RAND_MAX /* For SunOS. */
	#define RAND_MAX 32767
	#endif

	/* Check whether we can lock the seed file read write. */
	#if defined(HAVE_FCNTL) && defined(HAVE_FTRUNCATE) && !defined(HAVE_W32_SYSTEM)
	#define LOCK_SEED_FILE 1
	#else
	#define LOCK_SEED_FILE 0
	#endif

	/* Define the constant we use for transforming the pool at read-out. */
	#if SIZEOF_UNSIGNED_LONG == 8
	#define ADD_VALUE 0xa5a5a5a5a5a5a5a5
	#elif SIZEOF_UNSIGNED_LONG == 4
	#define ADD_VALUE 0xa5a5a5a5
	#else
	#error weird size for an unsigned long
	#endif

	/* Contstants pertaining to the hash pool. */
	#define BLOCKLEN 64 /* Hash this amount of bytes... */
	#define DIGESTLEN 20 /* ... into a digest of this length (sha-1). */
	/* POOLBLOCKS is the number of digests which make up the pool. */
	#define POOLBLOCKS 30
	/* POOLSIZE must be a multiple of the digest length to make the AND
	operations faster, the size should also be a multiple of unsigned
	long. */
	#define POOLSIZE (POOLBLOCKS*DIGESTLEN)
	#if (POOLSIZE % SIZEOF_UNSIGNED_LONG)
	#error Please make sure that poolsize is a multiple of unsigned long
	#endif
	#define POOLWORDS (POOLSIZE / SIZEOF_UNSIGNED_LONG)


	/* RNDPOOL is the pool we use to collect the entropy and to stir it
	up. Its allocated size is POOLSIZE+BLOCKLEN. Note that this is
	also an indication on whether the module has been fully
	initialized. */
	static unsigned char *rndpool;

	/* KEYPOOL is used as a scratch copy to read out random from RNDPOOL.
	Its allocated size is also POOLSIZE+BLOCKLEN. */
	static unsigned char *keypool;

	/* This is the offset into RNDPOOL where the next random bytes are to
	be mixed in. */
	static size_t pool_writepos;

	/* When reading data out of KEYPOOL, we start the read at different
	positions. This variable keeps track on where to read next. */
	static size_t pool_readpos;

	/* This flag is set to true as soon as the pool has been completely
	filled the first time. This may happen either by reading a seed
	file or by adding enough entropy. */
	static int pool_filled;

	/* This counter is used to track whether the initial seeding has been
	done with enough bytes from a reliable entropy source. */
	static size_t pool_filled_counter;

	/* If random of level GCRY_VERY_STRONG_RANDOM has been requested we
	have stricter requirements on what kind of entropy is in the pool.
	In particular POOL_FILLED is not sufficient. Thus we add some
	extra seeding and set this flag to true if the extra seeding has
	been done. */
	static int did_initial_extra_seeding;

	/* This variable is used to estimated the amount of fresh entropy
	available in RNDPOOL. */
	static int pool_balance;

	/* After a mixing operation this variable will be set to true and
	cleared if new entropy has been added or a remix is required for
	other reasons. */
	static int just_mixed;

	/* The name of the seed file or NULL if no seed file has been defined.
	The seed file needs to be registered at initialization time. We
	keep a malloced copy here. */
	static char *seed_file_name;

	/* If a seed file has been registered and maybe updated on exit this
	flag set. */
	static int allow_seed_file_update;

	/* Option flag set at initialiation time to force allocation of the
	pool in secure memory. */
	static int secure_alloc;

	/* This function pointer is set to the actual entropy gathering
	function during initialization. After initialization it is
	guaranteed to point to function. (On systems without a random
	gatherer module a dummy function is used).*/
	static int (slow_gather_fnc)(void ()(const void*, size_t,
	enum random_origins),
	enum random_origins, size_t, int);

	/* This function is set to the actual fast entropy gathering function
	during initialization. If it is NULL, no such function is
	available. */
	static void (fast_gather_fnc)(void ()(const void*, size_t,
	enum random_origins),
	enum random_origins);


	/* Option flag useful for debugging and the test suite. If set
	requests for very strong random are degraded to strong random. Not
	used by regular applications. */
	static int quick_test;

	/* This is the lock we use to protect all pool operations. */
	GPGRT_LOCK_DEFINE (pool_lock);

	/* This is a helper for assert calls. These calls are used to assert
	that functions are called in a locked state. It is not meant to be
	thread-safe but as a method to get aware of missing locks in the
	test suite. */
	static int pool_is_locked;


	/* We keep some counters in this structure for the sake of the
	_gcry_random_dump_stats () function. */
	static struct
	{
	unsigned long mixrnd;
	unsigned long mixkey;
	unsigned long slowpolls;
	unsigned long fastpolls;
	unsigned long getbytes1;
	unsigned long ngetbytes1;
	unsigned long getbytes2;
	unsigned long ngetbytes2;
	unsigned long addbytes;
	unsigned long naddbytes;
	} rndstats;



	-/* --- Stuff pertaining to the random daemon support. --- */
	-#ifdef USE_RANDOM_DAEMON
	-
	-/* If ALLOW_DAEMON is true, the module will try to use the random
	- daemon first. If the daemon has failed, this variable is set to
	- back to false and the code continues as normal. Note, we don't
	- test this flag in a locked state because a wrong value does not
	- harm and the trhead will find out itself that the daemon does not
	- work and set it (again) to false. */
	-static int allow_daemon;
	-
	-/* During initialization, the user may set a non-default socket name
	- for accessing the random daemon. If this value is NULL, the
	- default name will be used. */
	-static char *daemon_socket_name;
	-
	-#endif /USE_RANDOM_DAEMON/
	-
	-
	-
	/* --- Prototypes --- */
	static void read_pool (byte *buffer, size_t length, int level );
	static void add_randomness (const void *buffer, size_t length,
	enum random_origins origin);
	static void random_poll (void);
	static void do_fast_random_poll (void);
	static int (getfnc_gather_random (void))(void ()(const void*, size_t,
	enum random_origins),
	enum random_origins, size_t, int);
	static void (getfnc_fast_random_poll (void))(void ()(const void*, size_t,
	enum random_origins),
	enum random_origins);
	static void read_random_source (enum random_origins origin,
	size_t length, int level);



	/* --- Functions --- */


	/* Basic initialization which is required to initialize mutexes and
	such. It does not run a full initialization so that the filling of
	the random pool can be delayed until it is actually needed. We
	assume that this function is used before any concurrent access
	happens. */
	static void
	initialize_basics(void)
	{
	static int initialized;

	if (!initialized)
	{
	initialized = 1;

	#ifdef USE_RANDOM_DAEMON
	_gcry_daemon_initialize_basics ();
	#endif /USE_RANDOM_DAEMON/

	/* Make sure that we are still using the values we have
	traditionally used for the random levels. */
	gcry_assert (GCRY_WEAK_RANDOM == 0
	&& GCRY_STRONG_RANDOM == 1
	&& GCRY_VERY_STRONG_RANDOM == 2);
	}
	}

	/* Take the pool lock. */
	static void
	lock_pool (void)
	{
	int err;

	err = gpgrt_lock_lock (&pool_lock);
	if (err)
	log_fatal ("failed to acquire the pool lock: %s\n", gpg_strerror (err));
	pool_is_locked = 1;
	}

	/* Release the pool lock. */
	static void
	unlock_pool (void)
	{
	int err;

	pool_is_locked = 0;
	err = gpgrt_lock_unlock (&pool_lock);
	if (err)
	log_fatal ("failed to release the pool lock: %s\n", gpg_strerror (err));
	}


	/* Full initialization of this module. */
	static void
	initialize(void)
	{
	/* Although the basic initialization should have happened already,
	we call it here to make sure that all prerequisites are met. */
	initialize_basics ();

	/* Now we can look the pool and complete the initialization if
	necessary. */
	lock_pool ();
	if (!rndpool)
	{
	/* The data buffer is allocated somewhat larger, so that we can
	use this extra space (which is allocated in secure memory) as
	a temporary hash buffer */
	rndpool = (secure_alloc
	? xcalloc_secure (1, POOLSIZE + BLOCKLEN)
	: xcalloc (1, POOLSIZE + BLOCKLEN));
	keypool = (secure_alloc
	? xcalloc_secure (1, POOLSIZE + BLOCKLEN)
	: xcalloc (1, POOLSIZE + BLOCKLEN));

	/* Setup the slow entropy gathering function. The code requires
	that this function exists. */
	slow_gather_fnc = getfnc_gather_random ();

	/* Setup the fast entropy gathering function. */
	fast_gather_fnc = getfnc_fast_random_poll ();

	}
	unlock_pool ();
	}




	/* Initialize this random subsystem. If FULL is false, this function
	merely calls the initialize and does not do anything more. Doing
	this is not really required but when running in a threaded
	environment we might get a race condition otherwise. */
	void
	_gcry_rngcsprng_initialize (int full)
	{
	if (!full)
	initialize_basics ();
	else
	initialize ();
	}


	/* Try to close the FDs of the random gather module. This is
	currently only implemented for rndlinux. */
	void
	_gcry_rngcsprng_close_fds (void)
	{
	lock_pool ();
	#if USE_RNDGETENTROPY
	_gcry_rndgetentropy_gather_random (NULL, 0, 0, 0);
	#endif
	#if USE_RNDLINUX
	_gcry_rndlinux_gather_random (NULL, 0, 0, 0);
	#endif
	pool_writepos = 0;
	pool_readpos = 0;
	pool_filled = 0;
	pool_filled_counter = 0;
	did_initial_extra_seeding = 0;
	pool_balance = 0;
	just_mixed = 0;
	xfree (rndpool);
	xfree (keypool);
	rndpool = NULL;
	keypool = NULL;
	unlock_pool ();
	}


	void
	_gcry_rngcsprng_dump_stats (void)
	{
	/* In theory we would need to lock the stats here. However this
	function is usually called during cleanup and then we _might_ run
	into problems. */

	log_info ("random usage: poolsize=%d mixed=%lu polls=%lu/%lu added=%lu/%lu\n"
	" outmix=%lu getlvl1=%lu/%lu getlvl2=%lu/%lu%s\n",
	POOLSIZE, rndstats.mixrnd, rndstats.slowpolls, rndstats.fastpolls,
	rndstats.naddbytes, rndstats.addbytes,
	rndstats.mixkey, rndstats.ngetbytes1, rndstats.getbytes1,
	rndstats.ngetbytes2, rndstats.getbytes2,
	_gcry_rndhw_failed_p()? " (hwrng failed)":"");
	}


	/* This function should be called during initialization and before
	initialization of this module to place the random pools into secure
	memory. */
	void
	_gcry_rngcsprng_secure_alloc (void)
	{
	secure_alloc = 1;
	}


	/* This may be called before full initialization to degrade the
	quality of the RNG for the sake of a faster running test suite. */
	void
	_gcry_rngcsprng_enable_quick_gen (void)
	{
	quick_test = 1;
	}


	-void
	-_gcry_rngcsprng_set_daemon_socket (const char *socketname)
	-{
	-#ifdef USE_RANDOM_DAEMON
	- if (daemon_socket_name)
	- BUG ();
	-
	- daemon_socket_name = gcry_xstrdup (socketname);
	-#else /!USE_RANDOM_DAEMON/
	- (void)socketname;
	-#endif /!USE_RANDOM_DAEMON/
	-}
	-
	-/* With ONOFF set to 1, enable the use of the daemon. With ONOFF set
	- to 0, disable the use of the daemon. With ONOF set to -1, return
	- whether the daemon has been enabled. */
	-int
	-_gcry_rngcsprng_use_daemon (int onoff)
	-{
	-#ifdef USE_RANDOM_DAEMON
	- int last;
	-
	- /* This is not really thread safe. However it is expected that this
	- function is being called during initialization and at that point
	- we are for other reasons not really thread safe. We do not want
	- to lock it because we might eventually decide that this function
	- may even be called prior to gcry_check_version. */
	- last = allow_daemon;
	- if (onoff != -1)
	- allow_daemon = onoff;
	-
	- return last;
	-#else /!USE_RANDOM_DAEMON/
	- (void)onoff;
	- return 0;
	-#endif /!USE_RANDOM_DAEMON/
	-}
	-
	-
	/* This function returns true if no real RNG is available or the
	quality of the RNG has been degraded for test purposes. */
	int
	_gcry_rngcsprng_is_faked (void)
	{
	/* We need to initialize due to the runtime determination of
	available entropy gather modules. */
	initialize();
	return quick_test;
	}


	/* Add BUFLEN bytes from BUF to the internal random pool. QUALITY
	should be in the range of 0..100 to indicate the goodness of the
	entropy added, or -1 for goodness not known. */
	gcry_error_t
	_gcry_rngcsprng_add_bytes (const void *buf, size_t buflen, int quality)
	{
	size_t nbytes;
	const char *bufptr;

	if (quality == -1)
	quality = 35;
	else if (quality > 100)
	quality = 100;
	else if (quality < 0)
	quality = 0;

	if (!buf)
	return gpg_error (GPG_ERR_INV_ARG);

	if (!buflen \|\| quality < 10)
	return 0; /* Take a shortcut. */

	/* Because we don't increment the entropy estimation with FASTPOLL,
	we don't need to take lock that estimation while adding from an
	external source. This limited entropy estimation also means that
	we can't take QUALITY into account. */
	initialize_basics ();
	bufptr = buf;
	while (buflen)
	{
	nbytes = buflen > POOLSIZE? POOLSIZE : buflen;
	lock_pool ();
	if (rndpool)
	add_randomness (bufptr, nbytes, RANDOM_ORIGIN_EXTERNAL);
	unlock_pool ();
	bufptr += nbytes;
	buflen -= nbytes;
	}
	return 0;
	}


	/* Public function to fill the buffer with LENGTH bytes of
	cryptographically strong random bytes. Level GCRY_WEAK_RANDOM is
	not very strong, GCRY_STRONG_RANDOM is strong enough for most
	usage, GCRY_VERY_STRONG_RANDOM is good for key generation stuff but
	may be very slow. */
	void
	_gcry_rngcsprng_randomize (void *buffer, size_t length,
	enum gcry_random_level level)
	{
	unsigned char *p;

	/* Make sure we are initialized. */
	initialize ();

	/* Handle our hack used for regression tests of Libgcrypt. */
	if ( quick_test && level > GCRY_STRONG_RANDOM )
	level = GCRY_STRONG_RANDOM;

	/* Make sure the level is okay. */
	level &= 3;

	-#ifdef USE_RANDOM_DAEMON
	- if (allow_daemon
	- && !_gcry_daemon_randomize (daemon_socket_name, buffer, length, level))
	- return; /* The daemon succeeded. */
	- allow_daemon = 0; /* Daemon failed - switch off. */
	-#endif /USE_RANDOM_DAEMON/
	-
	/* Acquire the pool lock. */
	lock_pool ();

	/* Update the statistics. */
	if (level >= GCRY_VERY_STRONG_RANDOM)
	{
	rndstats.getbytes2 += length;
	rndstats.ngetbytes2++;
	}
	else
	{
	rndstats.getbytes1 += length;
	rndstats.ngetbytes1++;
	}

	/* Read the random into the provided buffer. */
	for (p = buffer; length > 0;)
	{
	size_t n;

	n = length > POOLSIZE? POOLSIZE : length;
	read_pool (p, n, level);
	length -= n;
	p += n;
	}

	/* Release the pool lock. */
	unlock_pool ();
	}




	/*
	* Mix the 600 byte pool. Note that the 64 byte scratch area directly
	* follows the pool. The numbers in the diagram give the number of
	* bytes.
	* <................600...............> <.64.>
	* pool \|------------------------------------\| \|------\|
	* <20><.24.> <20>
	* \| \| +-----+
	* +-----\|-------------------------------\|-+
	* +-------------------------------\|-\|-+
	* v v v
	* \|------\|
	* <hash>
	* +---------------------------------------+
	* v
	* <20>
	* pool' \|------------------------------------\|
	* <20><20><.24.>
	* +---\|-----\|---------------------------+
	* +-----\|---------------------------\|-+
	* +---------------------------\|-\|-+
	* v v v
	* \|------\|
	* <hash>
	* \|
	* +-----------------------------------+
	* v
	* <20>
	* pool'' \|------------------------------------\|
	* <20><20><20><.24.>
	* +---\|-----\|-----------------------+
	* +-----\|-----------------------\|-+
	* +-----------------------\|-\|-+
	* v v v
	*
	* and so on until we did this for all 30 blocks.
	*
	* To better protect against implementation errors in this code, we
	* xor a digest of the entire pool into the pool before mixing.
	*
	* Note: this function must only be called with a locked pool.
	*/
	static void
	mix_pool(unsigned char *pool)
	{
	static unsigned char failsafe_digest[DIGESTLEN];
	static int failsafe_digest_valid;

	unsigned char *hashbuf = pool + POOLSIZE;
	unsigned char p, pend;
	int i, n;
	SHA1_CONTEXT md;
	unsigned int nburn;

	#if DIGESTLEN != 20
	#error must have a digest length of 20 for SHA-1
	#endif

	gcry_assert (pool_is_locked);
	_gcry_sha1_mixblock_init (&md);

	/* pool_0 -> pool'. */
	pend = pool + POOLSIZE;
	memcpy (hashbuf, pend - DIGESTLEN, DIGESTLEN);
	memcpy (hashbuf+DIGESTLEN, pool, BLOCKLEN-DIGESTLEN);
	nburn = _gcry_sha1_mixblock (&md, hashbuf);
	memcpy (pool, hashbuf, DIGESTLEN);

	if (failsafe_digest_valid && pool == rndpool)
	{
	for (i=0; i < DIGESTLEN; i++)
	pool[i] ^= failsafe_digest[i];
	}

	/* Loop for the remaining iterations. */
	p = pool;
	for (n=1; n < POOLBLOCKS; n++)
	{
	if (p + BLOCKLEN < pend)
	memcpy (hashbuf, p, BLOCKLEN);
	else
	{
	unsigned char *pp = p;

	for (i=0; i < BLOCKLEN; i++ )
	{
	if ( pp >= pend )
	pp = pool;
	hashbuf[i] = *pp++;
	}
	}

	_gcry_sha1_mixblock (&md, hashbuf);
	p += DIGESTLEN;
	memcpy (p, hashbuf, DIGESTLEN);
	}

	/* Our hash implementation does only leave small parts (64 bytes)
	of the pool on the stack, so it is okay not to require secure
	memory here. Before we use this pool, it will be copied to the
	help buffer anyway. */
	if ( pool == rndpool)
	{
	_gcry_sha1_hash_buffer (failsafe_digest, pool, POOLSIZE);
	failsafe_digest_valid = 1;
	}

	_gcry_burn_stack (nburn);
	}


	void
	_gcry_rngcsprng_set_seed_file (const char *name)
	{
	if (seed_file_name)
	BUG ();
	seed_file_name = xstrdup (name);
	}



	/* Helper for my_open.
	* Return a malloced wide char string from an UTF-8 encoded input
	* string STRING. Caller must free this value. Returns NULL and sets
	* ERRNO on failure. Calling this function with STRING set to NULL is
	* not defined. */
	#ifdef HAVE_W32_SYSTEM
	static wchar_t *
	utf8_to_wchar (const char *string)
	{
	int n;
	size_t nbytes;
	wchar_t *result;

	n = MultiByteToWideChar (CP_UTF8, 0, string, -1, NULL, 0);
	if (n < 0)
	{
	gpg_err_set_errno (EINVAL);
	return NULL;
	}

	nbytes = (size_t)(n+1) * sizeof(*result);
	if (nbytes / sizeof(*result) != (n+1))
	{
	gpg_err_set_errno (ENOMEM);
	return NULL;
	}
	result = xtrymalloc (nbytes);
	if (!result)
	return NULL;

	n = MultiByteToWideChar (CP_UTF8, 0, string, -1, result, n);
	if (n < 0)
	{
	xfree (result);
	gpg_err_set_errno (EINVAL);
	result = NULL;
	}
	return result;
	}
	#endif /HAVE_W32_SYSTEM/


	/* Helper for my_open. */
	#ifdef HAVE_W32_SYSTEM
	static int
	any8bitchar (const char *string)
	{
	if (string)
	for ( ; *string; string++)
	if ((*string & 0x80))
	return 1;
	return 0;
	}
	#endif /HAVE_W32_SYSTEM/


	/* A wrapper around open to handle Unicode file names under Windows. */
	static int
	my_open (const char *name, int flags, unsigned int mode)
	{
	#ifdef HAVE_W32_SYSTEM
	if (any8bitchar (name))
	{
	wchar_t *wname;
	int ret;

	wname = utf8_to_wchar (name);
	if (!wname)
	return -1;
	ret = _wopen (wname, flags, mode);
	xfree (wname);
	return ret;
	}
	else
	return open (name, flags, mode);
	#else
	return open (name, flags, mode);
	#endif
	}


	/* Lock an open file identified by file descriptor FD and wait a
	reasonable time to succeed. With FOR_WRITE set to true a write
	lock will be taken. FNAME is used only for diagnostics. Returns 0
	on success or -1 on error. */
	static int
	lock_seed_file (int fd, const char *fname, int for_write)
	{
	#ifdef __GCC__
	#warning Check whether we can lock on Windows.
	#endif
	#if LOCK_SEED_FILE
	struct flock lck;
	struct timeval tv;
	int backoff=0;

	/* We take a lock on the entire file. */
	memset (&lck, 0, sizeof lck);
	lck.l_type = for_write? F_WRLCK : F_RDLCK;
	lck.l_whence = SEEK_SET;

	while (fcntl (fd, F_SETLK, &lck) == -1)
	{
	if (errno != EAGAIN && errno != EACCES)
	{
	log_info (_("can't lock `%s': %s\n"), fname, strerror (errno));
	return -1;
	}

	if (backoff > 2) /* Show the first message after ~2.25 seconds. */
	log_info( _("waiting for lock on `%s'...\n"), fname);

	tv.tv_sec = backoff;
	tv.tv_usec = 250000;
	select (0, NULL, NULL, NULL, &tv);
	if (backoff < 10)
	backoff++ ;
	}
	#else
	(void)fd;
	(void)fname;
	(void)for_write;
	#endif /!LOCK_SEED_FILE/
	return 0;
	}


	/* Read in a seed from the random_seed file and return true if this
	was successful.

	Note: Multiple instances of 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 32 bytes from a non-blocking entropy source. The
	consequence is that the output of these different instances is
	correlated to some extent. In the perfect 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 32 bytes" above to 0. Then the dependencies of the initial
	states of the pools are completely known. */
	static int
	read_seed_file (void)
	{
	int fd;
	struct stat sb;
	unsigned char buffer[POOLSIZE];
	int n;

	gcry_assert (pool_is_locked);

	if (!seed_file_name)
	return 0;

	#ifdef HAVE_DOSISH_SYSTEM
	fd = my_open (seed_file_name, O_RDONLY \| O_BINARY, 0);
	#else
	fd = my_open (seed_file_name, O_RDONLY, 0);
	#endif
	if( fd == -1 && errno == ENOENT)
	{
	allow_seed_file_update = 1;
	return 0;
	}

	if (fd == -1 )
	{
	log_info(_("can't open `%s': %s\n"), seed_file_name, strerror(errno) );
	return 0;
	}
	if (lock_seed_file (fd, seed_file_name, 0))
	{
	close (fd);
	return 0;
	}
	if (fstat( fd, &sb ) )
	{
	log_info(_("can't stat `%s': %s\n"), seed_file_name, strerror(errno) );
	close(fd);
	return 0;
	}
	if (!S_ISREG(sb.st_mode) )
	{
	log_info(_("`%s' is not a regular file - ignored\n"), seed_file_name );
	close(fd);
	return 0;
	}
	if (!sb.st_size )
	{
	log_info(_("note: random_seed file is empty\n") );
	close(fd);
	allow_seed_file_update = 1;
	return 0;
	}
	if (sb.st_size != POOLSIZE )
	{
	log_info(_("warning: invalid size of random_seed file - not used\n") );
	close(fd);
	return 0;
	}

	do
	{
	n = read( fd, buffer, POOLSIZE );
	}
	while (n == -1 && errno == EINTR );

	if (n != POOLSIZE)
	{
	log_fatal(_("can't read `%s': %s\n"), seed_file_name,strerror(errno) );
	close(fd);/NOTREACHED/
	return 0;
	}

	close(fd);

	add_randomness( buffer, POOLSIZE, RANDOM_ORIGIN_INIT );
	/* add some minor entropy to the pool now (this will also force a mixing) */
	{
	pid_t x = getpid();
	add_randomness( &x, sizeof(x), RANDOM_ORIGIN_INIT );
	}
	{
	time_t x = time(NULL);
	add_randomness( &x, sizeof(x), RANDOM_ORIGIN_INIT );
	}
	{
	clock_t x = clock();
	add_randomness( &x, sizeof(x), RANDOM_ORIGIN_INIT );
	}

	/* And read a few bytes from our entropy source. If we have the
	* Jitter RNG we can fast get a lot of entropy. Thus we read 1024
	* bits from that source.
	*
	* Without the Jitter RNG we keep the old method of reading only a
	* few bytes usually from /dev/urandom which won't block. */
	if (_gcry_rndjent_get_version (NULL))
	read_random_source (RANDOM_ORIGIN_INIT, 128, GCRY_STRONG_RANDOM);
	else
	read_random_source (RANDOM_ORIGIN_INIT, 32, GCRY_STRONG_RANDOM);

	allow_seed_file_update = 1;
	return 1;
	}


	void
	_gcry_rngcsprng_update_seed_file (void)
	{
	unsigned long sp, dp;
	int fd, i;

	/* We do only a basic initialization so that we can lock the pool.
	This is required to cope with the case that this function is
	called by some cleanup code at a point where the RNG has never
	been initialized. */
	initialize_basics ();
	lock_pool ();

	if ( !seed_file_name \|\| !rndpool \|\| !pool_filled )
	{
	unlock_pool ();
	return;
	}
	if ( !allow_seed_file_update )
	{
	unlock_pool ();
	log_info(_("note: random_seed file not updated\n"));
	return;
	}

	/* At this point we know that there is something in the pool and
	thus we can conclude that the pool has been fully initialized. */


	/* Copy the entropy pool to a scratch pool and mix both of them. */
	for (i=0,dp=(unsigned long)(void)keypool, sp=(unsigned long)(void)rndpool;
	i < POOLWORDS; i++, dp++, sp++ )
	{
	dp = sp + ADD_VALUE;
	}
	mix_pool(rndpool); rndstats.mixrnd++;
	mix_pool(keypool); rndstats.mixkey++;

	#if defined(HAVE_DOSISH_SYSTEM) \|\| defined(__CYGWIN__)
	fd = my_open (seed_file_name, O_WRONLY\|O_CREAT\|O_TRUNC\|O_BINARY,
	S_IRUSR\|S_IWUSR );
	#else
	# if LOCK_SEED_FILE
	fd = my_open (seed_file_name, O_WRONLY\|O_CREAT, S_IRUSR\|S_IWUSR );
	# else
	fd = my_open (seed_file_name, O_WRONLY\|O_CREAT\|O_TRUNC, S_IRUSR\|S_IWUSR );
	# endif
	#endif

	if (fd == -1 )
	log_info (_("can't create `%s': %s\n"), seed_file_name, strerror(errno) );
	else if (lock_seed_file (fd, seed_file_name, 1))
	{
	close (fd);
	}
	#if LOCK_SEED_FILE
	else if (ftruncate (fd, 0))
	{
	log_info(_("can't write `%s': %s\n"), seed_file_name, strerror(errno));
	close (fd);
	}
	#endif /LOCK_SEED_FILE/
	else
	{
	do
	{
	i = write (fd, keypool, POOLSIZE );
	}
	while (i == -1 && errno == EINTR);
	if (i != POOLSIZE)
	log_info (_("can't write `%s': %s\n"),seed_file_name, strerror(errno));
	if (close(fd))
	log_info (_("can't close `%s': %s\n"),seed_file_name, strerror(errno));
	}

	unlock_pool ();
	}


	/* Read random out of the pool. This function is the core of the
	public random functions. Note that Level GCRY_WEAK_RANDOM is not
	anymore handled special and in fact is an alias in the API for
	level GCRY_STRONG_RANDOM. Must be called with the pool already
	locked. */
	static void
	read_pool (byte *buffer, size_t length, int level)
	{
	int i;
	unsigned long sp, dp;
	/* The volatile is there to make sure the compiler does not optimize
	the code away in case the getpid function is badly attributed.
	Note that we keep a pid in a static variable as well as in a
	stack based one; the latter is to detect ill behaving thread
	libraries, ignoring the pool mutexes. */
	static volatile pid_t my_pid = (pid_t)(-1);
	volatile pid_t my_pid2;

	gcry_assert (pool_is_locked);

	retry:
	/* Get our own pid, so that we can detect a fork. */
	my_pid2 = getpid ();
	if (my_pid == (pid_t)(-1))
	my_pid = my_pid2;
	if ( my_pid != my_pid2 )
	{
	/* We detected a plain fork; i.e. we are now the child. Update
	the static pid and add some randomness. */
	pid_t x;

	my_pid = my_pid2;
	x = my_pid;
	add_randomness (&x, sizeof(x), RANDOM_ORIGIN_INIT);
	just_mixed = 0; /* Make sure it will get mixed. */
	}

	gcry_assert (pool_is_locked);

	/* Our code does not allow to extract more than POOLSIZE. Better
	check it here. */
	if (length > POOLSIZE)
	{
	log_bug("too many random bits requested\n");
	}

	if (!pool_filled)
	{
	if (read_seed_file() )
	pool_filled = 1;
	}

	/* For level 2 quality (key generation) we always make sure that the
	pool has been seeded enough initially. */
	if (level == GCRY_VERY_STRONG_RANDOM && !did_initial_extra_seeding)
	{
	size_t needed;

	pool_balance = 0;
	needed = length - pool_balance;
	if (needed < 16) /* At least 128 bits. */
	needed = 16;
	else if( needed > POOLSIZE )
	BUG ();
	read_random_source (RANDOM_ORIGIN_EXTRAPOLL, needed,
	GCRY_VERY_STRONG_RANDOM);
	pool_balance += needed;
	did_initial_extra_seeding = 1;
	}

	/* For level 2 make sure that there is enough random in the pool. */
	if (level == GCRY_VERY_STRONG_RANDOM && pool_balance < length)
	{
	size_t needed;

	if (pool_balance < 0)
	pool_balance = 0;
	needed = length - pool_balance;
	if (needed > POOLSIZE)
	BUG ();
	read_random_source (RANDOM_ORIGIN_EXTRAPOLL, needed,
	GCRY_VERY_STRONG_RANDOM);
	pool_balance += needed;
	}

	/* Make sure the pool is filled. */
	while (!pool_filled)
	random_poll();

	/* Always do a fast random poll (we have to use the unlocked version). */
	do_fast_random_poll();

	/* Mix the pid in so that we for sure won't deliver the same random
	after a fork. */
	{
	pid_t apid = my_pid;
	add_randomness (&apid, sizeof (apid), RANDOM_ORIGIN_INIT);
	}

	/* Mix the pool (if add_randomness() didn't it). */
	if (!just_mixed)
	{
	mix_pool(rndpool);
	rndstats.mixrnd++;
	}

	/* Create a new pool. */
	for(i=0,dp=(unsigned long)(void)keypool, sp=(unsigned long)(void)rndpool;
	i < POOLWORDS; i++, dp++, sp++ )
	dp = sp + ADD_VALUE;

	/* Mix both pools. */
	mix_pool(rndpool); rndstats.mixrnd++;
	mix_pool(keypool); rndstats.mixkey++;

	/* Read the requested data. We use a read pointer to read from a
	different position each time. */
	while (length--)
	{
	*buffer++ = keypool[pool_readpos++];
	if (pool_readpos >= POOLSIZE)
	pool_readpos = 0;
	pool_balance--;
	}

	if (pool_balance < 0)
	pool_balance = 0;

	/* Clear the keypool. */
	memset (keypool, 0, POOLSIZE);

	/* We need to detect whether a fork has happened. A fork might have
	an identical pool and thus the child and the parent could emit
	the very same random number. This test here is to detect forks
	in a multi-threaded process. It does not work with all thread
	implementations in particular not with pthreads. However it is
	good enough for GNU Pth. */
	if ( getpid () != my_pid2 )
	{
	pid_t x = getpid();
	add_randomness (&x, sizeof(x), RANDOM_ORIGIN_INIT);
	just_mixed = 0; /* Make sure it will get mixed. */
	my_pid = x; /* Also update the static pid. */
	goto retry;
	}
	}



	/* Add LENGTH bytes of randomness from buffer to the pool. ORIGIN is
	used to specify the randomness origin. This is one of the
	RANDOM_ORIGIN_* values. */
	static void
	add_randomness (const void *buffer, size_t length, enum random_origins origin)
	{
	const unsigned char *p = buffer;
	size_t count = 0;

	gcry_assert (pool_is_locked);

	rndstats.addbytes += length;
	rndstats.naddbytes++;
	while (length-- )
	{
	rndpool[pool_writepos++] ^= *p++;
	count++;
	if (pool_writepos >= POOLSIZE )
	{
	/* It is possible that we are invoked before the pool is
	filled using an unreliable origin of entropy, for example
	the fast random poll. To avoid flagging the pool as
	filled in this case, we track the initial filling state
	separately. See also the remarks about the seed file. */
	if (origin >= RANDOM_ORIGIN_SLOWPOLL && !pool_filled)
	{
	pool_filled_counter += count;
	count = 0;
	if (pool_filled_counter >= POOLSIZE)
	pool_filled = 1;
	}
	pool_writepos = 0;
	mix_pool(rndpool); rndstats.mixrnd++;
	just_mixed = !length;
	}
	}
	}



	static void
	random_poll()
	{
	rndstats.slowpolls++;
	read_random_source (RANDOM_ORIGIN_SLOWPOLL, POOLSIZE/5, GCRY_STRONG_RANDOM);
	}


	/* Runtime determination of the slow entropy gathering module. */
	static int (*
	getfnc_gather_random (void))(void ()(const void, size_t,
	enum random_origins),
	enum random_origins, size_t, int)
	{
	int (fnc)(void ()(const void*, size_t, enum random_origins),
	enum random_origins, size_t, int);

	#if USE_RNDGETENTROPY
	fnc = _gcry_rndgetentropy_gather_random;
	return fnc;
	#endif

	#if USE_RNDLINUX
	if ( !access (NAME_OF_DEV_RANDOM, R_OK)
	&& !access (NAME_OF_DEV_URANDOM, R_OK))
	{
	fnc = _gcry_rndlinux_gather_random;
	return fnc;
	}
	#endif

	#if USE_RNDEGD
	if ( _gcry_rndegd_connect_socket (1) != -1 )
	{
	fnc = _gcry_rndegd_gather_random;
	return fnc;
	}
	#endif

	#if USE_RNDUNIX
	fnc = _gcry_rndunix_gather_random;
	return fnc;
	#endif

	#if USE_RNDW32
	fnc = _gcry_rndw32_gather_random;
	return fnc;
	#endif

	#if USE_RNDW32CE
	fnc = _gcry_rndw32ce_gather_random;
	return fnc;
	#endif

	log_fatal (_("no entropy gathering module detected\n"));

	return NULL; /NOTREACHED/
	}

	/* Runtime determination of the fast entropy gathering function.
	(Currently a compile time method is used.) */
	static void (*
	getfnc_fast_random_poll (void))( void ()(const void, size_t,
	enum random_origins),
	enum random_origins)
	{
	#if USE_RNDW32
	return _gcry_rndw32_gather_random_fast;
	#endif
	#if USE_RNDW32CE
	return _gcry_rndw32ce_gather_random_fast;
	#endif
	return NULL;
	}



	static void
	do_fast_random_poll (void)
	{
	gcry_assert (pool_is_locked);

	rndstats.fastpolls++;

	if (fast_gather_fnc)
	fast_gather_fnc (add_randomness, RANDOM_ORIGIN_FASTPOLL);

	/* Continue with the generic functions. */
	#if HAVE_GETHRTIME
	{
	hrtime_t tv;
	tv = gethrtime();
	add_randomness( &tv, sizeof(tv), RANDOM_ORIGIN_FASTPOLL );
	}
	#elif HAVE_GETTIMEOFDAY
	{
	struct timeval tv;
	if( gettimeofday( &tv, NULL ) )
	BUG();
	add_randomness( &tv.tv_sec, sizeof(tv.tv_sec), RANDOM_ORIGIN_FASTPOLL );
	add_randomness( &tv.tv_usec, sizeof(tv.tv_usec), RANDOM_ORIGIN_FASTPOLL );
	}
	#elif HAVE_CLOCK_GETTIME
	{ struct timespec tv;
	if( clock_gettime( CLOCK_REALTIME, &tv ) == -1 )
	BUG();
	add_randomness( &tv.tv_sec, sizeof(tv.tv_sec), RANDOM_ORIGIN_FASTPOLL );
	add_randomness( &tv.tv_nsec, sizeof(tv.tv_nsec), RANDOM_ORIGIN_FASTPOLL );
	}
	#else /* use times */
	# ifndef HAVE_DOSISH_SYSTEM
	{ struct tms buf;
	times( &buf );
	add_randomness( &buf, sizeof buf, RANDOM_ORIGIN_FASTPOLL );
	}
	# endif
	#endif

	#ifdef HAVE_GETRUSAGE
	# ifdef RUSAGE_SELF
	{
	struct rusage buf;
	/* QNX/Neutrino does return ENOSYS - so we just ignore it and add
	whatever is in buf. In a chroot environment it might not work
	at all (i.e. because /proc/ is not accessible), so we better
	ignore all error codes and hope for the best. */
	getrusage (RUSAGE_SELF, &buf );
	add_randomness( &buf, sizeof buf, RANDOM_ORIGIN_FASTPOLL );
	memset( &buf, 0, sizeof buf );
	}
	# else /!RUSAGE_SELF/
	# ifdef __GCC__
	# warning There is no RUSAGE_SELF on this system
	# endif
	# endif /!RUSAGE_SELF/
	#endif /HAVE_GETRUSAGE/

	/* Time and clock are available on all systems - so we better do it
	just in case one of the above functions didn't work. */
	{
	time_t x = time(NULL);
	add_randomness( &x, sizeof(x), RANDOM_ORIGIN_FASTPOLL );
	}
	{
	clock_t x = clock();
	add_randomness( &x, sizeof(x), RANDOM_ORIGIN_FASTPOLL );
	}

	/* If the system features a fast hardware RNG, read some bytes from
	there. */
	_gcry_rndhw_poll_fast (add_randomness, RANDOM_ORIGIN_FASTPOLL);
	}


	/* The fast random pool function as called at some places in
	libgcrypt. This is merely a wrapper to make sure that this module
	is initialized and to lock the pool. Note, that this function is a
	NOP unless a random function has been used or _gcry_initialize (1)
	has been used. We use this hack so that the internal use of this
	function in cipher_open and md_open won't start filling up the
	random pool, even if no random will be required by the process. */
	void
	_gcry_rngcsprng_fast_poll (void)
	{
	initialize_basics ();

	lock_pool ();
	if (rndpool)
	{
	/* Yes, we are fully initialized. */
	do_fast_random_poll ();
	}
	unlock_pool ();
	}



	static void
	read_random_source (enum random_origins origin, size_t length, int level)
	{
	if ( !slow_gather_fnc )
	log_fatal ("Slow entropy gathering module not yet initialized\n");

	if (slow_gather_fnc (add_randomness, origin, length, level) < 0)
	log_fatal ("No way to gather entropy for the RNG\n");
	}
	diff --git a/random/random-daemon.c b/random/random-daemon.c
	deleted file mode 100644
	index 8ea4df28..00000000
	--- a/random/random-daemon.c
	+++ /dev/null
	@@ -1,336 +0,0 @@
	-/* random-daemon.c - Access to the external random daemon
	- * Copyright (C) 2006 Free Software Foundation, Inc.
	- *
	- * This file is part of Libgcrypt.
	- *
	- * Libgcrypt is free software; you can redistribute it and/or modify
	- * it under the terms of the GNU Lesser General Public License as
	- * published by the Free Software Foundation; either version 2.1 of
	- * the License, or (at your option) any later version.
	- *
	- * Libgcrypt is distributed in the hope that it will be useful,
	- * but WITHOUT ANY WARRANTY; without even the implied warranty of
	- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
	- * GNU Lesser General Public License for more details.
	- *
	- * You should have received a copy of the GNU Lesser General Public
	- * License along with this program; if not, write to the Free Software
	- * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA
	- */
	-
	-/*
	- The functions here are used by random.c to divert calls to an
	- external random number daemon. The actual daemon we use is
	- gcryptrnd. Such a daemon is useful to keep a persistent pool in
	- memory over invocations of a single application and to allow
	- prioritizing access to the actual entropy sources. The drawback is
	- that we need to use IPC (i.e. unix domain socket) to convey
	- sensitive data.
	- */
	-
	-#include <config.h>
	-#include <stdio.h>
	-#include <stdlib.h>
	-#include <assert.h>
	-#include <string.h>
	-#include <sys/socket.h>
	-#include <sys/un.h>
	-#include <errno.h>
	-#include <unistd.h>
	-
	-#include "g10lib.h"
	-#include "random.h"
	-
	-
	-
	-/* This is default socket name we use in case the provided socket name
	- is NULL. */
	-#define RANDOM_DAEMON_SOCKET "/var/run/libgcrypt/S.gcryptrnd"
	-
	-/* The lock serializing access to the daemon. */
	-GPGRT_LOCK_DEFINE (daemon_lock);
	-
	-/* The socket connected to the daemon. */
	-static int daemon_socket = -1;
	-
	-/* Creates a socket connected to the daemon. On success, store the
	- socket fd in SOCK. Returns error code. /
	-static gcry_error_t
	-connect_to_socket (const char socketname, int sock)
	-{
	- struct sockaddr_un *srvr_addr;
	- socklen_t addrlen;
	- gcry_error_t err;
	- int fd;
	- int rc;
	-
	- srvr_addr = NULL;
	-
	- /* Create a socket. */
	- fd = socket (AF_UNIX, SOCK_STREAM, 0);
	- if (fd == -1)
	- {
	- log_error ("can't create socket: %s\n", strerror (errno));
	- err = gcry_error_from_errno (errno);
	- goto out;
	- }
	-
	- /* Set up address. */
	- srvr_addr = gcry_malloc (sizeof *srvr_addr);
	- if (! srvr_addr)
	- {
	- log_error ("malloc failed: %s\n", strerror (errno));
	- err = gcry_error_from_errno (errno);
	- goto out;
	- }
	- memset (srvr_addr, 0, sizeof *srvr_addr);
	- srvr_addr->sun_family = AF_UNIX;
	- if (strlen (socketname) + 1 >= sizeof (srvr_addr->sun_path))
	- {
	- log_error ("socket name `%s' too long\n", socketname);
	- err = gcry_error (GPG_ERR_ENAMETOOLONG);
	- goto out;
	- }
	- strcpy (srvr_addr->sun_path, socketname);
	- addrlen = (offsetof (struct sockaddr_un, sun_path)
	- + strlen (srvr_addr->sun_path) + 1);
	-
	- /* Connect socket. */
	- rc = connect (fd, (struct sockaddr *) srvr_addr, addrlen);
	- if (rc == -1)
	- {
	- log_error ("error connecting socket `%s': %s\n",
	- srvr_addr->sun_path, strerror (errno));
	- err = gcry_error_from_errno (errno);
	- goto out;
	- }
	-
	- err = 0;
	-
	- out:
	-
	- gcry_free (srvr_addr);
	- if (err)
	- {
	- close (fd);
	- fd = -1;
	- }
	- *sock = fd;
	-
	- return err;
	-}
	-
	-
	-/* Initialize basics of this module. This should be viewed as a
	- constructor to prepare locking. */
	-void
	-_gcry_daemon_initialize_basics (void)
	-{
	- /* Not anymore required. */
	-}
	-
	-
	-
	-/* Send LENGTH bytes of BUFFER to file descriptor FD. Returns 0 on
	- success or another value on write error. */
	-static int
	-writen (int fd, const void *buffer, size_t length)
	-{
	- ssize_t n;
	-
	- while (length)
	- {
	- do
	- n = ath_write (fd, buffer, length);
	- while (n < 0 && errno == EINTR);
	- if (n < 0)
	- {
	- log_error ("write error: %s\n", strerror (errno));
	- return -1; /* write error */
	- }
	- length -= n;
	- buffer = (const char*)buffer + n;
	- }
	- return 0; /* Okay */
	-}
	-
	-static int
	-readn (int fd, void buf, size_t buflen, size_t ret_nread)
	-{
	- size_t nleft = buflen;
	- int nread;
	- char *p;
	-
	- p = buf;
	- while (nleft > 0)
	- {
	- nread = ath_read (fd, buf, nleft);
	- if (nread < 0)
	- {
	- if (nread == EINTR)
	- nread = 0;
	- else
	- return -1;
	- }
	- else if (!nread)
	- break; /* EOF */
	- nleft -= nread;
	- buf = (char*)buf + nread;
	- }
	- if (ret_nread)
	- *ret_nread = buflen - nleft;
	- return 0;
	-}
	-
	-/* This functions requests REQ_NBYTES from the daemon. If NONCE is
	- true, the data should be suited for a nonce. If NONCE is FALSE,
	- data of random level LEVEL will be generated. The retrieved random
	- data will be stored in BUFFER. Returns error code. */
	-static gcry_error_t
	-call_daemon (const char *socketname,
	- void *buffer, size_t req_nbytes, int nonce,
	- enum gcry_random_level level)
	-{
	- static int initialized;
	- unsigned char buf[255];
	- gcry_error_t err = 0;
	- size_t nbytes;
	- size_t nread;
	- int rc;
	-
	- if (!req_nbytes)
	- return 0;
	-
	- gpgrt_lock_lock (&daemon_lock);
	-
	- /* Open the socket if that has not been done. */
	- if (!initialized)
	- {
	- initialized = 1;
	- err = connect_to_socket (socketname ? socketname : RANDOM_DAEMON_SOCKET,
	- &daemon_socket);
	- if (err)
	- {
	- daemon_socket = -1;
	- log_info ("not using random daemon\n");
	- gpgrt_lock_unlock (&daemon_lock);
	- return err;
	- }
	- }
	-
	- /* Check that we have a valid socket descriptor. */
	- if ( daemon_socket == -1 )
	- {
	- gpgrt_lock_unlock (&daemon_lock);
	- return gcry_error (GPG_ERR_INTERNAL);
	- }
	-
	-
	- /* Do the real work. */
	-
	- do
	- {
	- /* Process in chunks. */
	- nbytes = req_nbytes > sizeof (buf) ? sizeof (buf) : req_nbytes;
	- req_nbytes -= nbytes;
	-
	- /* Construct request. */
	- buf[0] = 3;
	- if (nonce)
	- buf[1] = 10;
	- else if (level == GCRY_VERY_STRONG_RANDOM)
	- buf[1] = 12;
	- else if (level == GCRY_STRONG_RANDOM)
	- buf[1] = 11;
	- buf[2] = nbytes;
	-
	- /* Send request. */
	- rc = writen (daemon_socket, buf, 3);
	- if (rc == -1)
	- {
	- err = gcry_error_from_errno (errno);
	- break;
	- }
	-
	- /* Retrieve response. */
	-
	- rc = readn (daemon_socket, buf, 2, &nread);
	- if (rc == -1)
	- {
	- err = gcry_error_from_errno (errno);
	- log_error ("read error: %s\n", _gcry_strerror (err));
	- break;
	- }
	- if (nread && buf[0])
	- {
	- log_error ("random daemon returned error code %d\n", buf[0]);
	- err = gcry_error (GPG_ERR_INTERNAL); /* ? */
	- break;
	- }
	- if (nread != 2)
	- {
	- log_error ("response too small\n");
	- err = gcry_error (GPG_ERR_PROTOCOL_VIOLATION); /* ? */
	- break;
	- }
	-
	- /* if (1)/ / Do this in verbose mode? */
	- /* log_info ("received response with %d bytes of data\n", buf[1]);*/
	-
	- if (buf[1] < nbytes)
	- {
	- log_error ("error: server returned less bytes than requested\n");
	- err = gcry_error (GPG_ERR_PROTOCOL_VIOLATION); /* ? */
	- break;
	- }
	- else if (buf[1] > nbytes)
	- {
	- log_error ("warning: server returned more bytes than requested\n");
	- err = gcry_error (GPG_ERR_PROTOCOL_VIOLATION); /* ? */
	- break;
	- }
	-
	- assert (nbytes <= sizeof (buf));
	-
	- rc = readn (daemon_socket, buf, nbytes, &nread);
	- if (rc == -1)
	- {
	- err = gcry_error_from_errno (errno);
	- log_error ("read error: %s\n", _gcry_strerror (err));
	- break;
	- }
	-
	- if (nread != nbytes)
	- {
	- log_error ("too little random data read\n");
	- err = gcry_error (GPG_ERR_INTERNAL);
	- break;
	- }
	-
	- /* Successfuly read another chunk of data. */
	- memcpy (buffer, buf, nbytes);
	- buffer = ((char *) buffer) + nbytes;
	- }
	- while (req_nbytes);
	-
	- gpgrt_lock_unlock (&daemon_lock);
	-
	- return err;
	-}
	-
	-/* Internal function to fill BUFFER with LENGTH bytes of random. We
	- support GCRY_STRONG_RANDOM and GCRY_VERY_STRONG_RANDOM here.
	- Return 0 on success. */
	-int
	-_gcry_daemon_randomize (const char *socketname,
	- void *buffer, size_t length,
	- enum gcry_random_level level)
	-{
	- gcry_error_t err;
	-
	- err = call_daemon (socketname, buffer, length, 0, level);
	-
	- return err ? -1 : 0;
	-}
	-
	-/* END */
	diff --git a/random/random.c b/random/random.c
	index 9aab7893..c0435d7b 100644
	--- a/random/random.c
	+++ b/random/random.c
	@@ -1,584 +1,562 @@
	/* random.c - Random number switch
	* Copyright (C) 2003, 2006, 2008, 2012 Free Software Foundation, Inc.
	*
	* This file is part of Libgcrypt.
	*
	* Libgcrypt is free software; you can redistribute it and/or modify
	* it under the terms of the GNU Lesser General Public License as
	* published by the Free Software Foundation; either version 2.1 of
	* the License, or (at your option) any later version.
	*
	* Libgcrypt is distributed in the hope that it will be useful,
	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
	* GNU Lesser General Public License for more details.
	*
	* You should have received a copy of the GNU Lesser General Public
	* License along with this program; if not, see <http://www.gnu.org/licenses/>.
	*/

	/*
	This module switches between different implementations of random
	number generators and provides a few help functions.
	*/

	#include <config.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <errno.h>
	#include <time.h>
	#include <sys/types.h>
	#include <unistd.h>
	#ifdef HAVE_SYSLOG
	# include <syslog.h>
	#endif /HAVE_SYSLOG/
	#include <ctype.h>

	#include "g10lib.h"
	#include "random.h"
	#include "rand-internal.h"
	#include "cipher.h" /* For _gcry_sha1_hash_buffer(). */

	/* The name of a file used to globally configure the RNG. */
	#define RANDOM_CONF_FILE "/etc/gcrypt/random.conf"


	/* If not NULL a progress function called from certain places and the
	opaque value passed along. Registered by
	_gcry_register_random_progress (). */
	static void (progress_cb) (void ,const char*,int,int, int );
	static void *progress_cb_data;

	/* Flags indicating the requested RNG types. */
	static struct
	{
	int standard;
	int fips;
	int system;
	} rng_types;


	/* This is the lock we use to protect the buffer used by the nonce
	generation. */
	GPGRT_LOCK_DEFINE (nonce_buffer_lock);



	/* --- Functions --- */


	/* Used to register a progress callback. This needs to be called
	before any threads are created. */
	void
	_gcry_register_random_progress (void (cb)(void ,const char*,int,int,int),
	void *cb_data )
	{
	progress_cb = cb;
	progress_cb_data = cb_data;
	}


	/* This progress function is currently used by the random modules to
	give hints on how much more entropy is required. */
	void
	_gcry_random_progress (const char *what, int printchar, int current, int total)
	{
	if (progress_cb)
	progress_cb (progress_cb_data, what, printchar, current, total);
	}


	/* Read a file with configure options. The file is a simple text file
	* where empty lines and lines with the first non white-space
	* character being '#' are ignored. Supported configure options are:
	*
	* disable-jent - Disable the jitter based extra entropy generator.
	* This sets the RANDOM_CONF_DISABLE_JENT bit.
	* only-urandom - Always use /dev/urandom instead of /dev/random.
	* This sets the RANDOM_CONF_ONLY_URANDOM bit.
	*
	* The function returns a bit vector with flags read from the file.
	*/
	unsigned int
	_gcry_random_read_conf (void)
	{
	const char *fname = RANDOM_CONF_FILE;
	FILE *fp;
	char buffer[256];
	char p, pend;
	int lnr = 0;
	unsigned int result = 0;

	fp = fopen (fname, "r");
	if (!fp)
	return result;

	for (;;)
	{
	if (!fgets (buffer, sizeof buffer, fp))
	{
	if (!feof (fp))
	{
	#ifdef HAVE_SYSLOG
	syslog (LOG_USER\|LOG_WARNING,
	"Libgcrypt warning: error reading '%s', line %d",
	fname, lnr);
	#endif /HAVE_SYSLOG/
	}
	fclose (fp);
	return result;
	}
	lnr++;
	for (p=buffer; my_isascii (p) && isspace (p); p++)
	;
	pend = strchr (p, '\n');
	if (pend)
	*pend = 0;
	pend = p + (*p? (strlen (p)-1):0);
	for ( ;pend > p; pend--)
	if (my_isascii (pend) && isspace (pend))
	*pend = 0;
	if (!p \|\| p == '#')
	continue;

	if (!strcmp (p, "disable-jent"))
	result \|= RANDOM_CONF_DISABLE_JENT;
	else if (!strcmp (p, "only-urandom"))
	result \|= RANDOM_CONF_ONLY_URANDOM;
	else
	{
	#ifdef HAVE_SYSLOG
	syslog (LOG_USER\|LOG_WARNING,
	"Libgcrypt warning: unknown option in '%s', line %d",
	fname, lnr);
	#endif /HAVE_SYSLOG/
	}
	}
	}


	/* Set the preferred RNG type. This may be called at any time even
	before gcry_check_version. Thus we can't assume any thread system
	initialization. A type of 0 is used to indicate that any Libgcrypt
	initialization has been done.*/
	void
	_gcry_set_preferred_rng_type (int type)
	{
	static int any_init;

	if (!type)
	{
	any_init = 1;
	}
	else if (type == GCRY_RNG_TYPE_STANDARD)
	{
	rng_types.standard = 1;
	}
	else if (any_init)
	{
	/* After any initialization has been done we only allow to
	upgrade to the standard RNG (handled above). All other
	requests are ignored. The idea is that the application needs
	to declare a preference for a weaker RNG as soon as possible
	and before any library sets a preference. We assume that a
	library which uses Libgcrypt calls an init function very
	early. This way --- even if the library gets initialized
	early by the application --- it is unlikely that it can
	select a lower priority RNG.

	This scheme helps to ensure that existing unmodified
	applications (e.g. gpg2), which don't known about the new RNG
	selection system, will continue to use the standard RNG and
	not be tricked by some library to use a lower priority RNG.
	There are some loopholes here but at least most GnuPG stuff
	should be save because it calls src_c{gcry_control
	(GCRYCTL_SUSPEND_SECMEM_WARN);} quite early and thus inhibits
	switching to a low priority RNG.
	*/
	}
	else if (type == GCRY_RNG_TYPE_FIPS)
	{
	rng_types.fips = 1;
	}
	else if (type == GCRY_RNG_TYPE_SYSTEM)
	{
	rng_types.system = 1;
	}
	}


	/* Initialize this random subsystem. If FULL is false, this function
	merely calls the basic initialization of the module and does not do
	anything more. Doing this is not really required but when running
	in a threaded environment we might get a race condition
	otherwise. */
	void
	_gcry_random_initialize (int full)
	{
	if (fips_mode ())
	_gcry_rngdrbg_inititialize (full);
	else if (rng_types.standard)
	_gcry_rngcsprng_initialize (full);
	else if (rng_types.fips)
	_gcry_rngdrbg_inititialize (full);
	else if (rng_types.system)
	_gcry_rngsystem_initialize (full);
	else
	_gcry_rngcsprng_initialize (full);
	}


	/* If possible close file descriptors used by the RNG. */
	void
	_gcry_random_close_fds (void)
	{
	/* Note that we can't do that directly because each random system
	has its own lock functions which need to be used for accessing
	the entropy gatherer. */

	if (fips_mode ())
	_gcry_rngdrbg_close_fds ();
	else if (rng_types.standard)
	_gcry_rngcsprng_close_fds ();
	else if (rng_types.fips)
	_gcry_rngdrbg_close_fds ();
	else if (rng_types.system)
	_gcry_rngsystem_close_fds ();
	else
	_gcry_rngcsprng_close_fds ();
	}


	/* Return the current RNG type. IGNORE_FIPS_MODE is a flag used to
	skip the test for FIPS. This is useful, so that we are able to
	return the type of the RNG even before we have setup FIPS mode
	(note that FIPS mode is enabled by default until it is switched off
	by the initialization). This is mostly useful for the regression
	test. */
	int
	_gcry_get_rng_type (int ignore_fips_mode)
	{
	if (!ignore_fips_mode && fips_mode ())
	return GCRY_RNG_TYPE_FIPS;
	else if (rng_types.standard)
	return GCRY_RNG_TYPE_STANDARD;
	else if (rng_types.fips)
	return GCRY_RNG_TYPE_FIPS;
	else if (rng_types.system)
	return GCRY_RNG_TYPE_SYSTEM;
	else
	return GCRY_RNG_TYPE_STANDARD;
	}


	void
	_gcry_random_dump_stats (void)
	{
	if (fips_mode ())
	_gcry_rngdrbg_dump_stats ();
	else
	_gcry_rngcsprng_dump_stats ();
	_gcry_rndjent_dump_stats ();
	}


	/* This function should be called during initialization and before
	initialization of this module to place the random pools into secure
	memory. */
	void
	_gcry_secure_random_alloc (void)
	{
	if (fips_mode ())
	; /* Not used; the FIPS RNG is always in secure mode. */
	else
	_gcry_rngcsprng_secure_alloc ();
	}


	/* This may be called before full initialization to degrade the
	quality of the RNG for the sake of a faster running test suite. */
	void
	_gcry_enable_quick_random_gen (void)
	{
	if (fips_mode ())
	; /* Not used. */
	else
	_gcry_rngcsprng_enable_quick_gen ();
	}


	-void
	-_gcry_set_random_daemon_socket (const char *socketname)
	-{
	- if (fips_mode ())
	- ; /* Not used. */
	- else
	- _gcry_rngcsprng_set_daemon_socket (socketname);
	-}
	-
	-/* With ONOFF set to 1, enable the use of the daemon. With ONOFF set
	- to 0, disable the use of the daemon. With ONOF set to -1, return
	- whether the daemon has been enabled. */
	-int
	-_gcry_use_random_daemon (int onoff)
	-{
	- if (fips_mode ())
	- return 0; /* Never enabled in fips mode. */
	- else
	- return _gcry_rngcsprng_use_daemon (onoff);
	-}
	-
	-
	/* This function returns true if no real RNG is available or the
	quality of the RNG has been degraded for test purposes. */
	int
	_gcry_random_is_faked (void)
	{
	if (fips_mode ())
	return _gcry_rngdrbg_is_faked ();
	else
	return _gcry_rngcsprng_is_faked ();
	}


	/* Add BUFLEN bytes from BUF to the internal random pool. QUALITY
	should be in the range of 0..100 to indicate the goodness of the
	entropy added, or -1 for goodness not known. */
	gcry_err_code_t
	_gcry_random_add_bytes (const void *buf, size_t buflen, int quality)
	{
	if (fips_mode ())
	return 0; /* No need for this in fips mode. */
	else if (rng_types.standard)
	return gpg_err_code (_gcry_rngcsprng_add_bytes (buf, buflen, quality));
	else if (rng_types.fips)
	return 0;
	else if (rng_types.system)
	return 0;
	else /* default */
	return gpg_err_code (_gcry_rngcsprng_add_bytes (buf, buflen, quality));
	}


	/* Helper function. */
	static void
	do_randomize (void *buffer, size_t length, enum gcry_random_level level)
	{
	if (fips_mode ())
	_gcry_rngdrbg_randomize (buffer, length, level);
	else if (rng_types.standard)
	_gcry_rngcsprng_randomize (buffer, length, level);
	else if (rng_types.fips)
	_gcry_rngdrbg_randomize (buffer, length, level);
	else if (rng_types.system)
	_gcry_rngsystem_randomize (buffer, length, level);
	else /* default */
	_gcry_rngcsprng_randomize (buffer, length, level);
	}

	/* The public function to return random data of the quality LEVEL.
	Returns a pointer to a newly allocated and randomized buffer of
	LEVEL and NBYTES length. Caller must free the buffer. */
	void *
	_gcry_random_bytes (size_t nbytes, enum gcry_random_level level)
	{
	void *buffer;

	buffer = xmalloc (nbytes);
	do_randomize (buffer, nbytes, level);
	return buffer;
	}


	/* The public function to return random data of the quality LEVEL;
	this version of the function returns the random in a buffer allocated
	in secure memory. Caller must free the buffer. */
	void *
	_gcry_random_bytes_secure (size_t nbytes, enum gcry_random_level level)
	{
	void *buffer;

	/* Historical note (1.3.0--1.4.1): The buffer was only allocated
	in secure memory if the pool in random-csprng.c was also set to
	use secure memory. */
	buffer = xmalloc_secure (nbytes);
	do_randomize (buffer, nbytes, level);
	return buffer;
	}


	/* Public function to fill the buffer with LENGTH bytes of
	cryptographically strong random bytes. Level GCRY_WEAK_RANDOM is
	not very strong, GCRY_STRONG_RANDOM is strong enough for most
	usage, GCRY_VERY_STRONG_RANDOM is good for key generation stuff but
	may be very slow. */
	void
	_gcry_randomize (void *buffer, size_t length, enum gcry_random_level level)
	{
	do_randomize (buffer, length, level);
	}


	/* This function may be used to specify the file to be used as a seed
	file for the PRNG. This function should be called prior to the
	initialization of the random module. NAME may not be NULL. */
	void
	_gcry_set_random_seed_file (const char *name)
	{
	if (fips_mode ())
	; /* No need for this in fips mode. */
	else if (rng_types.standard)
	_gcry_rngcsprng_set_seed_file (name);
	else if (rng_types.fips)
	;
	else if (rng_types.system)
	;
	else /* default */
	_gcry_rngcsprng_set_seed_file (name);
	}


	/* If a seed file has been setup, this function may be used to write
	back the random numbers entropy pool. */
	void
	_gcry_update_random_seed_file (void)
	{
	if (fips_mode ())
	; /* No need for this in fips mode. */
	else if (rng_types.standard)
	_gcry_rngcsprng_update_seed_file ();
	else if (rng_types.fips)
	;
	else if (rng_types.system)
	;
	else /* default */
	_gcry_rngcsprng_update_seed_file ();
	}



	/* The fast random pool function as called at some places in
	libgcrypt. This is merely a wrapper to make sure that this module
	is initialized and to lock the pool. Note, that this function is a
	NOP unless a random function has been used or _gcry_initialize (1)
	has been used. We use this hack so that the internal use of this
	function in cipher_open and md_open won't start filling up the
	random pool, even if no random will be required by the process. */
	void
	_gcry_fast_random_poll (void)
	{
	if (fips_mode ())
	; /* No need for this in fips mode. */
	else if (rng_types.standard)
	_gcry_rngcsprng_fast_poll ();
	else if (rng_types.fips)
	;
	else if (rng_types.system)
	;
	else /* default */
	_gcry_rngcsprng_fast_poll ();
	}



	/* Create an unpredicable nonce of LENGTH bytes in BUFFER. */
	void
	_gcry_create_nonce (void *buffer, size_t length)
	{
	static unsigned char nonce_buffer[20+8];
	static int nonce_buffer_initialized = 0;
	static volatile pid_t my_pid; /* The volatile is there to make sure the
	compiler does not optimize the code away
	in case the getpid function is badly
	attributed. */
	volatile pid_t apid;
	unsigned char *p;
	size_t n;
	int err;

	/* First check whether we shall use the FIPS nonce generator. This
	is only done in FIPS mode, in all other modes, we use our own
	nonce generator which is seeded by the RNG actual in use. */
	if (fips_mode ())
	{
	_gcry_rngdrbg_randomize (buffer, length, GCRY_WEAK_RANDOM);
	return;
	}

	/* This is the nonce generator, which formerly lived in
	random-csprng.c. It is now used by all RNG types except when in
	FIPS mode (not that this means it is also used if the FIPS RNG
	has been selected but we are not in fips mode). */

	/* Make sure we are initialized. */
	_gcry_random_initialize (1);

	/* Acquire the nonce buffer lock. */
	err = gpgrt_lock_lock (&nonce_buffer_lock);
	if (err)
	log_fatal ("failed to acquire the nonce buffer lock: %s\n",
	gpg_strerror (err));

	apid = getpid ();
	/* The first time initialize our buffer. */
	if (!nonce_buffer_initialized)
	{
	time_t atime = time (NULL);
	pid_t xpid = apid;

	my_pid = apid;

	if ((sizeof apid + sizeof atime) > sizeof nonce_buffer)
	BUG ();

	/* Initialize the first 20 bytes with a reasonable value so that
	a failure of gcry_randomize won't affect us too much. Don't
	care about the uninitialized remaining bytes. */
	p = nonce_buffer;
	memcpy (p, &xpid, sizeof xpid);
	p += sizeof xpid;
	memcpy (p, &atime, sizeof atime);

	/* Initialize the never changing private part of 64 bits. */
	_gcry_randomize (nonce_buffer+20, 8, GCRY_WEAK_RANDOM);

	nonce_buffer_initialized = 1;
	}
	else if ( my_pid != apid )
	{
	/* We forked. Need to reseed the buffer - doing this for the
	private part should be sufficient. */
	do_randomize (nonce_buffer+20, 8, GCRY_WEAK_RANDOM);
	/* Update the pid so that we won't run into here again and
	again. */
	my_pid = apid;
	}

	/* Create the nonce by hashing the entire buffer, returning the hash
	and updating the first 20 bytes of the buffer with this hash. */
	for (p = buffer; length > 0; length -= n, p += n)
	{
	_gcry_sha1_hash_buffer (nonce_buffer,
	nonce_buffer, sizeof nonce_buffer);
	n = length > 20? 20 : length;
	memcpy (p, nonce_buffer, n);
	}

	/* Release the nonce buffer lock. */
	err = gpgrt_lock_unlock (&nonce_buffer_lock);
	if (err)
	log_fatal ("failed to release the nonce buffer lock: %s\n",
	gpg_strerror (err));
	}


	/* Run the self-tests for the RNG. This is currently only implemented
	for the FIPS generator. */
	gpg_error_t
	_gcry_random_selftest (selftest_report_func_t report)
	{
	if (fips_mode ())
	return _gcry_rngdrbg_selftest (report);
	else
	return 0; /* No selftests yet. */
	}
	diff --git a/random/random.h b/random/random.h
	index c00b9ed5..caac4b64 100644
	--- a/random/random.h
	+++ b/random/random.h
	@@ -1,79 +1,77 @@
	/* random.h - random functions
	* Copyright (C) 1998, 2002, 2006 Free Software Foundation, Inc.
	*
	* This file is part of Libgcrypt.
	*
	* Libgcrypt is free software; you can redistribute it and/or modify
	* it under the terms of the GNU Lesser General Public License as
	* published by the Free Software Foundation; either version 2.1 of
	* the License, or (at your option) any later version.
	*
	* Libgcrypt is distributed in the hope that it will be useful,
	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
	* GNU Lesser General Public License for more details.
	*
	* You should have received a copy of the GNU Lesser General Public
	* License along with this program; if not, write to the Free Software
	* Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA
	*/
	#ifndef G10_RANDOM_H
	#define G10_RANDOM_H

	#include "types.h"
	#include "../src/gcrypt-testapi.h" /* struct gcry_drbg_test_vector */

	/-- random.c --/
	void _gcry_register_random_progress (void (cb)(void ,const char*,int,int,int),
	void *cb_data );

	void _gcry_set_preferred_rng_type (int type);
	void _gcry_random_initialize (int full);
	void _gcry_random_close_fds (void);
	int _gcry_get_rng_type (int ignore_fips_mode);
	void _gcry_random_dump_stats(void);
	void _gcry_secure_random_alloc(void);
	void _gcry_enable_quick_random_gen (void);
	int _gcry_random_is_faked(void);
	-void _gcry_set_random_daemon_socket (const char *socketname);
	-int _gcry_use_random_daemon (int onoff);
	void _gcry_set_random_seed_file (const char *name);
	void _gcry_update_random_seed_file (void);

	void _gcry_fast_random_poll( void );

	gcry_err_code_t _gcry_random_init_external_test (void **r_context,
	unsigned int flags,
	const void *key,
	size_t keylen,
	const void *seed,
	size_t seedlen,
	const void *dt,
	size_t dtlen);
	gcry_err_code_t _gcry_random_run_external_test (void *context,
	char *buffer, size_t buflen);
	void _gcry_random_deinit_external_test (void *context);

	/-- random-drbg.c --/
	gpg_err_code_t _gcry_rngdrbg_reinit (const char *flagstr,
	gcry_buffer_t *pers, int npers);
	gpg_err_code_t _gcry_rngdrbg_cavs_test (struct gcry_drbg_test_vector *t,
	unsigned char *buf);
	gpg_err_code_t _gcry_rngdrbg_healthcheck_one (struct gcry_drbg_test_vector *t);

	/-- rndegd.c --/
	gpg_error_t _gcry_rndegd_set_socket_name (const char *name);

	/-- rndjent.c --/
	unsigned int _gcry_rndjent_get_version (int *r_active);


	/-- random-daemon.c (only used from random.c) --/
	#ifdef USE_RANDOM_DAEMON
	void _gcry_daemon_initialize_basics (void);
	int _gcry_daemon_randomize (const char *socketname,
	void *buffer, size_t length,
	enum gcry_random_level level);
	#endif /USE_RANDOM_DAEMON/

	#endif /G10_RANDOM_H/
	diff --git a/src/global.c b/src/global.c
	index 1ff736d7..58873372 100644
	--- a/src/global.c
	+++ b/src/global.c
	@@ -1,1379 +1,1374 @@
	/* global.c - global control functions
	* Copyright (C) 1998, 1999, 2000, 2001, 2002, 2003
	* 2004, 2005, 2006, 2008, 2011,
	* 2012 Free Software Foundation, Inc.
	* Copyright (C) 2013, 2014, 2017 g10 Code GmbH
	*
	* This file is part of Libgcrypt.
	*
	* Libgcrypt is free software; you can redistribute it and/or modify
	* it under the terms of the GNU Lesser general Public License as
	* published by the Free Software Foundation; either version 2.1 of
	* the License, or (at your option) any later version.
	*
	* Libgcrypt is distributed in the hope that it will be useful,
	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
	* GNU Lesser General Public License for more details.
	*
	* You should have received a copy of the GNU Lesser General Public
	* License along with this program; if not, see <http://www.gnu.org/licenses/>.
	*/

	#include <config.h>

	#include <stdio.h>
	#include <stdlib.h>
	#include <string.h>
	#include <stdarg.h>
	#include <ctype.h>
	#include <limits.h>
	#include <errno.h>
	#include <unistd.h>
	#ifdef HAVE_SYSLOG
	# include <syslog.h>
	#endif /HAVE_SYSLOG/

	#include "g10lib.h"
	#include "gcrypt-testapi.h"
	#include "cipher.h"
	#include "stdmem.h" /* our own memory allocator */
	#include "secmem.h" /* our own secmem allocator */




	/****************
	* flag bits: 0 : general cipher debug
	* 1 : general MPI debug
	*/
	static unsigned int debug_flags;

	/* gcry_control (GCRYCTL_SET_FIPS_MODE), sets this flag so that the
	initialization code switched fips mode on. */
	static int force_fips_mode;

	/* Controlled by global_init(). */
	int _gcry_global_any_init_done;

	/*
	* Functions called before and after blocking syscalls.
	* Initialized by global_init and used via
	* _gcry_pre_syscall and _gcry_post_syscall.
	*/
	static void (*pre_syscall_func)(void);
	static void (*post_syscall_func)(void);


	/* Memory management. */

	static gcry_handler_alloc_t alloc_func;
	static gcry_handler_alloc_t alloc_secure_func;
	static gcry_handler_secure_check_t is_secure_func;
	static gcry_handler_realloc_t realloc_func;
	static gcry_handler_free_t free_func;
	static gcry_handler_no_mem_t outofcore_handler;
	static void *outofcore_handler_value;
	static int no_secure_memory;

	/* Prototypes. */
	static gpg_err_code_t external_lock_test (int cmd);




	/* This is our handmade constructor. It gets called by any function
	likely to be called at startup. The suggested way for an
	application to make sure that this has been called is by using
	gcry_check_version. */
	static void
	global_init (void)
	{
	gcry_error_t err = 0;

	if (_gcry_global_any_init_done)
	return;
	_gcry_global_any_init_done = 1;

	/* Tell the random module that we have seen an init call. */
	_gcry_set_preferred_rng_type (0);

	/* Get the system call clamp functions. */
	if (!pre_syscall_func)
	gpgrt_get_syscall_clamp (&pre_syscall_func, &post_syscall_func);

	/* See whether the system is in FIPS mode. This needs to come as
	early as possible but after ATH has been initialized. */
	_gcry_initialize_fips_mode (force_fips_mode);

	/* Before we do any other initialization we need to test available
	hardware features. */
	_gcry_detect_hw_features ();

	/* Initialize the modules - this is mainly allocating some memory and
	creating mutexes. */
	err = _gcry_cipher_init ();
	if (err)
	goto fail;
	err = _gcry_md_init ();
	if (err)
	goto fail;
	err = _gcry_mac_init ();
	if (err)
	goto fail;
	err = _gcry_pk_init ();
	if (err)
	goto fail;
	err = _gcry_primegen_init ();
	if (err)
	goto fail;
	err = _gcry_secmem_module_init ();
	if (err)
	goto fail;
	err = _gcry_mpi_init ();
	if (err)
	goto fail;

	return;

	fail:
	BUG ();
	}


	/* This function is called by the macro fips_is_operational and makes
	sure that the minimal initialization has been done. This is far
	from a perfect solution and hides problems with an improper
	initialization but at least in single-threaded mode it should work
	reliable.

	The reason we need this is that a lot of applications don't use
	Libgcrypt properly by not running any initialization code at all.
	They just call a Libgcrypt function and that is all what they want.
	Now with the FIPS mode, that has the side effect of entering FIPS
	mode (for security reasons, FIPS mode is the default if no
	initialization has been done) and bailing out immediately because
	the FSM is in the wrong state. If we always run the init code,
	Libgcrypt can test for FIPS mode and at least if not in FIPS mode,
	it will behave as before. Note that this on-the-fly initialization
	is only done for the cryptographic functions subject to FIPS mode
	and thus not all API calls will do such an initialization. */
	int
	_gcry_global_is_operational (void)
	{
	if (!_gcry_global_any_init_done)
	{
	#ifdef HAVE_SYSLOG
	syslog (LOG_USER\|LOG_WARNING, "Libgcrypt warning: "
	"missing initialization - please fix the application");
	#endif /HAVE_SYSLOG/
	global_init ();
	}
	return _gcry_fips_is_operational ();
	}




	/* Version number parsing. */

	/* This function parses the first portion of the version number S and
	stores it in *NUMBER. On success, this function returns a pointer
	into S starting with the first character, which is not part of the
	initial number portion; on failure, NULL is returned. */
	static const char*
	parse_version_number( const char s, int number )
	{
	int val = 0;

	if( *s == '0' && isdigit(s[1]) )
	return NULL; /* leading zeros are not allowed */
	for ( ; isdigit(*s); s++ ) {
	val *= 10;
	val += *s - '0';
	}
	*number = val;
	return val < 0? NULL : s;
	}

	/* This function breaks up the complete string-representation of the
	version number S, which is of the following struture: <major
	number>.<minor number>.<micro number><patch level>. The major,
	minor and micro number components will be stored in MAJOR, MINOR
	and *MICRO.

	On success, the last component, the patch level, will be returned;
	in failure, NULL will be returned. */

	static const char *
	parse_version_string( const char s, int major, int minor, int micro )
	{
	s = parse_version_number( s, major );
	if( !s \|\| *s != '.' )
	return NULL;
	s++;
	s = parse_version_number( s, minor );
	if( !s \|\| *s != '.' )
	return NULL;
	s++;
	s = parse_version_number( s, micro );
	if( !s )
	return NULL;
	return s; /* patchlevel */
	}

	/* If REQ_VERSION is non-NULL, check that the version of the library
	is at minimum the requested one. Returns the string representation
	of the library version if the condition is satisfied; return NULL
	if the requested version is newer than that of the library.

	If a NULL is passed to this function, no check is done, but the
	string representation of the library is simply returned. */
	const char *
	_gcry_check_version (const char *req_version)
	{
	const char *ver = VERSION;
	int my_major, my_minor, my_micro;
	int rq_major, rq_minor, rq_micro;
	const char *my_plvl;

	if (req_version && req_version[0] == 1 && req_version[1] == 1)
	return _gcry_compat_identification ();

	/* Initialize library. */
	global_init ();

	if ( !req_version )
	/* Caller wants our version number. */
	return ver;

	/* Parse own version number. */
	my_plvl = parse_version_string( ver, &my_major, &my_minor, &my_micro );
	if ( !my_plvl )
	/* very strange our own version is bogus. Shouldn't we use
	assert() here and bail out in case this happens? -mo. */
	return NULL;

	/* Parse requested version number. */
	if (!parse_version_string (req_version, &rq_major, &rq_minor, &rq_micro))
	return NULL; /* req version string is invalid, this can happen. */

	/* Compare version numbers. */
	if ( my_major > rq_major
	\|\| (my_major == rq_major && my_minor > rq_minor)
	\|\| (my_major == rq_major && my_minor == rq_minor
	&& my_micro > rq_micro)
	\|\| (my_major == rq_major && my_minor == rq_minor
	&& my_micro == rq_micro))
	{
	return ver;
	}

	return NULL;
	}


	static void
	print_config (const char *what, gpgrt_stream_t fp)
	{
	int i;
	const char *s;

	if (!what \|\| !strcmp (what, "version"))
	{
	gpgrt_fprintf (fp, "version:%s:%x:%s:%x:\n",
	VERSION, GCRYPT_VERSION_NUMBER,
	GPGRT_VERSION, GPGRT_VERSION_NUMBER);
	}
	if (!what \|\| !strcmp (what, "cc"))
	{
	gpgrt_fprintf (fp, "cc:%d:%s:\n",
	#if GPGRT_VERSION_NUMBER >= 0x011b00 /* 1.27 */
	GPGRT_GCC_VERSION
	#else
	_GPG_ERR_GCC_VERSION /* Due to a bug in gpg-error.h. */
	#endif
	,
	#ifdef __clang__
	"clang:" __VERSION__
	#elif __GNUC__
	"gcc:" __VERSION__
	#else
	":"
	#endif
	);
	}

	if (!what \|\| !strcmp (what, "ciphers"))
	gpgrt_fprintf (fp, "ciphers:%s:\n", LIBGCRYPT_CIPHERS);
	if (!what \|\| !strcmp (what, "pubkeys"))
	gpgrt_fprintf (fp, "pubkeys:%s:\n", LIBGCRYPT_PUBKEY_CIPHERS);
	if (!what \|\| !strcmp (what, "digests"))
	gpgrt_fprintf (fp, "digests:%s:\n", LIBGCRYPT_DIGESTS);

	if (!what \|\| !strcmp (what, "rnd-mod"))
	{
	gpgrt_fprintf (fp, "rnd-mod:"
	#if USE_RNDEGD
	"egd:"
	#endif
	#if USE_RNDGETENTROPY
	"getentropy:"
	#endif
	#if USE_RNDLINUX
	"linux:"
	#endif
	#if USE_RNDUNIX
	"unix:"
	#endif
	#if USE_RNDW32
	"w32:"
	#endif
	"\n");
	}

	if (!what \|\| !strcmp (what, "cpu-arch"))
	{
	gpgrt_fprintf (fp, "cpu-arch:"
	#if defined(HAVE_CPU_ARCH_X86)
	"x86"
	#elif defined(HAVE_CPU_ARCH_ALPHA)
	"alpha"
	#elif defined(HAVE_CPU_ARCH_SPARC)
	"sparc"
	#elif defined(HAVE_CPU_ARCH_MIPS)
	"mips"
	#elif defined(HAVE_CPU_ARCH_M68K)
	"m68k"
	#elif defined(HAVE_CPU_ARCH_PPC)
	"ppc"
	#elif defined(HAVE_CPU_ARCH_ARM)
	"arm"
	#endif
	":\n");
	}

	if (!what \|\| !strcmp (what, "mpi-asm"))
	gpgrt_fprintf (fp, "mpi-asm:%s:\n", _gcry_mpi_get_hw_config ());

	if (!what \|\| !strcmp (what, "hwflist"))
	{
	unsigned int hwfeatures, afeature;

	hwfeatures = _gcry_get_hw_features ();
	gpgrt_fprintf (fp, "hwflist:");
	for (i=0; (s = _gcry_enum_hw_features (i, &afeature)); i++)
	if ((hwfeatures & afeature))
	gpgrt_fprintf (fp, "%s:", s);
	gpgrt_fprintf (fp, "\n");
	}

	if (!what \|\| !strcmp (what, "fips-mode"))
	{
	/* We use y/n instead of 1/0 for the stupid reason that
	* Emacsen's compile error parser would accidentally flag that
	* line when printed during "make check" as an error. The
	* second field is obsolete and thus empty (used to be used for
	* a so-called enforced-fips-mode). The third field has an
	* option static string describing the module versions; this is
	* an optional configure option. */
	gpgrt_fprintf (fp, "fips-mode:%c::%s:\n",
	fips_mode ()? 'y':'n',
	#ifdef FIPS_MODULE_VERSION
	fips_mode () ? FIPS_MODULE_VERSION : ""
	#else
	""
	#endif /* FIPS_MODULE_VERSION */
	);
	}

	if (!what \|\| !strcmp (what, "rng-type"))
	{
	/* The currently used RNG type. */
	unsigned int jver;
	int active;

	i = _gcry_get_rng_type (0);
	switch (i)
	{
	case GCRY_RNG_TYPE_STANDARD: s = "standard"; break;
	case GCRY_RNG_TYPE_FIPS: s = "fips"; break;
	case GCRY_RNG_TYPE_SYSTEM: s = "system"; break;
	default: BUG ();
	}
	jver = _gcry_rndjent_get_version (&active);
	gpgrt_fprintf (fp, "rng-type:%s:%d:%u:%d:\n", s, i, jver, active);
	}

	if (!what \|\| !strcmp (what, "compliance"))
	{
	/* Right now we have no certification for 1.9 so we return an
	* empty string. As soon as this version has been approved for
	* VS-Nfd we will put the string "de-vs" into the second
	* field. If further specifications are required they are added
	* as parameters to that field. Other certifications will go
	* into field 3 and so on.
	* field 1: keyword "compliance"
	* field 2: German VS-Nfd is marked with "de-vs"
	* field 3: reserved for FIPS.
	*/
	gpgrt_fprintf (fp, "compliance:%s::\n", "");
	}
	}


	/* With a MODE of 0 return a malloced string with configured features.
	* In that case a WHAT of NULL returns everything in the same way
	* GCRYCTL_PRINT_CONFIG would do. With a specific WHAT string only
	* the requested feature is returned (w/o the trailing LF. On error
	* NULL is returned. */
	char *
	_gcry_get_config (int mode, const char *what)
	{
	gpgrt_stream_t fp;
	int save_errno;
	void *data;
	char *p;

	if (mode)
	{
	gpg_err_set_errno (EINVAL);
	return NULL;
	}

	fp = gpgrt_fopenmem (0, "w+b,samethread");
	if (!fp)
	return NULL;

	print_config (what, fp);

	if (!what)
	{
	/* Null-terminate bulk output. */
	gpgrt_fwrite ("\0", 1, 1, fp);
	}

	if (gpgrt_ferror (fp))
	{
	save_errno = errno;
	gpgrt_fclose (fp);
	gpg_err_set_errno (save_errno);
	return NULL;
	}

	gpgrt_rewind (fp);
	if (gpgrt_fclose_snatch (fp, &data, NULL))
	{
	save_errno = errno;
	gpgrt_fclose (fp);
	gpg_err_set_errno (save_errno);
	return NULL;
	}

	if (!data)
	{
	/* Nothing was printed (unknown value for WHAT). This is okay,
	* so clear ERRNO to indicate this. */
	gpg_err_set_errno (0);
	return NULL;
	}

	/* Strip trailing LF. */
	if (what && (p = strchr (data, '\n')))
	*p = 0;

	return data;
	}




	#if _GCRY_GCC_VERSION >= 40200
	# pragma GCC diagnostic push
	# pragma GCC diagnostic ignored "-Wswitch"
	#endif

	/* Command dispatcher function, acting as general control
	function. */
	gcry_err_code_t
	_gcry_vcontrol (enum gcry_ctl_cmds cmd, va_list arg_ptr)
	{
	static int init_finished = 0;
	gcry_err_code_t rc = 0;

	switch (cmd)
	{
	case GCRYCTL_ENABLE_M_GUARD:
	_gcry_private_enable_m_guard ();
	break;

	case GCRYCTL_ENABLE_QUICK_RANDOM:
	_gcry_set_preferred_rng_type (0);
	_gcry_enable_quick_random_gen ();
	break;

	case GCRYCTL_FAKED_RANDOM_P:
	/* Return an error if the RNG is faked one (e.g. enabled by
	ENABLE_QUICK_RANDOM. */
	if (_gcry_random_is_faked ())
	rc = GPG_ERR_GENERAL; /* Use as TRUE value. */
	break;

	case GCRYCTL_DUMP_RANDOM_STATS:
	_gcry_random_dump_stats ();
	break;

	case GCRYCTL_DUMP_MEMORY_STATS:
	/m_print_stats("[fixme: prefix]");/
	break;

	case GCRYCTL_DUMP_SECMEM_STATS:
	_gcry_secmem_dump_stats (0);
	break;

	case GCRYCTL_DROP_PRIVS:
	global_init ();
	_gcry_secmem_init (0);
	break;

	case GCRYCTL_DISABLE_SECMEM:
	global_init ();
	no_secure_memory = 1;
	break;

	case GCRYCTL_INIT_SECMEM:
	global_init ();
	_gcry_secmem_init (va_arg (arg_ptr, unsigned int));
	if ((_gcry_secmem_get_flags () & GCRY_SECMEM_FLAG_NOT_LOCKED))
	rc = GPG_ERR_GENERAL;
	break;

	case GCRYCTL_TERM_SECMEM:
	global_init ();
	_gcry_secmem_term ();
	break;

	case GCRYCTL_DISABLE_SECMEM_WARN:
	_gcry_set_preferred_rng_type (0);
	_gcry_secmem_set_flags ((_gcry_secmem_get_flags ()
	\| GCRY_SECMEM_FLAG_NO_WARNING));
	break;

	case GCRYCTL_SUSPEND_SECMEM_WARN:
	_gcry_set_preferred_rng_type (0);
	_gcry_secmem_set_flags ((_gcry_secmem_get_flags ()
	\| GCRY_SECMEM_FLAG_SUSPEND_WARNING));
	break;

	case GCRYCTL_RESUME_SECMEM_WARN:
	_gcry_set_preferred_rng_type (0);
	_gcry_secmem_set_flags ((_gcry_secmem_get_flags ()
	& ~GCRY_SECMEM_FLAG_SUSPEND_WARNING));
	break;

	case GCRYCTL_AUTO_EXPAND_SECMEM:
	_gcry_secmem_set_auto_expand (va_arg (arg_ptr, unsigned int));
	break;

	case GCRYCTL_USE_SECURE_RNDPOOL:
	global_init ();
	_gcry_secure_random_alloc (); /* Put random number into secure memory. */
	break;

	case GCRYCTL_SET_RANDOM_SEED_FILE:
	_gcry_set_preferred_rng_type (0);
	_gcry_set_random_seed_file (va_arg (arg_ptr, const char *));
	break;

	case GCRYCTL_UPDATE_RANDOM_SEED_FILE:
	_gcry_set_preferred_rng_type (0);
	if ( fips_is_operational () )
	_gcry_update_random_seed_file ();
	break;

	case GCRYCTL_SET_VERBOSITY:
	_gcry_set_preferred_rng_type (0);
	_gcry_set_log_verbosity (va_arg (arg_ptr, int));
	break;

	case GCRYCTL_SET_DEBUG_FLAGS:
	debug_flags \|= va_arg (arg_ptr, unsigned int);
	break;

	case GCRYCTL_CLEAR_DEBUG_FLAGS:
	debug_flags &= ~va_arg (arg_ptr, unsigned int);
	break;

	case GCRYCTL_DISABLE_INTERNAL_LOCKING:
	/* Not used anymore. */
	global_init ();
	break;

	case GCRYCTL_ANY_INITIALIZATION_P:
	if (_gcry_global_any_init_done)
	rc = GPG_ERR_GENERAL;
	break;

	case GCRYCTL_INITIALIZATION_FINISHED_P:
	if (init_finished)
	rc = GPG_ERR_GENERAL; /* Yes. */
	break;

	case GCRYCTL_INITIALIZATION_FINISHED:
	/* This is a hook which should be used by an application after
	all initialization has been done and right before any threads
	are started. It is not really needed but the only way to be
	really sure that all initialization for thread-safety has
	been done. */
	if (! init_finished)
	{
	global_init ();
	/* Do only a basic random initialization, i.e. init the
	mutexes. */
	_gcry_random_initialize (0);
	init_finished = 1;
	/* Force us into operational state if in FIPS mode. */
	(void)fips_is_operational ();
	}
	break;

	case GCRYCTL_SET_THREAD_CBS:
	/* This is now a dummy call. We used to install our own thread
	library here. */
	_gcry_set_preferred_rng_type (0);
	global_init ();
	break;

	case GCRYCTL_FAST_POLL:
	_gcry_set_preferred_rng_type (0);
	/* We need to do make sure that the random pool is really
	initialized so that the poll function is not a NOP. */
	_gcry_random_initialize (1);

	if ( fips_is_operational () )
	_gcry_fast_random_poll ();
	break;

	case GCRYCTL_SET_RNDEGD_SOCKET:
	#if USE_RNDEGD
	_gcry_set_preferred_rng_type (0);
	rc = _gcry_rndegd_set_socket_name (va_arg (arg_ptr, const char *));
	#else
	rc = GPG_ERR_NOT_SUPPORTED;
	#endif
	break;

	case GCRYCTL_SET_RANDOM_DAEMON_SOCKET:
	- _gcry_set_preferred_rng_type (0);
	- _gcry_set_random_daemon_socket (va_arg (arg_ptr, const char *));
	+ rc = GPG_ERR_NOT_SUPPORTED;
	break;

	case GCRYCTL_USE_RANDOM_DAEMON:
	- /* We need to do make sure that the random pool is really
	- initialized so that the poll function is not a NOP. */
	- _gcry_set_preferred_rng_type (0);
	- _gcry_random_initialize (1);
	- _gcry_use_random_daemon (!! va_arg (arg_ptr, int));
	+ rc = GPG_ERR_NOT_SUPPORTED;
	break;

	case GCRYCTL_CLOSE_RANDOM_DEVICE:
	_gcry_random_close_fds ();
	break;

	/* This command dumps information pertaining to the
	configuration of libgcrypt to the given stream. It may be
	used before the initialization has been finished but not
	before a gcry_version_check. See also gcry_get_config. */
	case GCRYCTL_PRINT_CONFIG:
	{
	FILE fp = va_arg (arg_ptr, FILE );
	char *tmpstr;
	_gcry_set_preferred_rng_type (0);
	tmpstr = _gcry_get_config (0, NULL);
	if (tmpstr)
	{
	if (fp)
	fputs (tmpstr, fp);
	else
	log_info ("%s", tmpstr);
	xfree (tmpstr);
	}
	}
	break;

	case GCRYCTL_OPERATIONAL_P:
	/* Returns true if the library is in an operational state. This
	is always true for non-fips mode. */
	_gcry_set_preferred_rng_type (0);
	if (_gcry_fips_test_operational ())
	rc = GPG_ERR_GENERAL; /* Used as TRUE value */
	break;

	case GCRYCTL_FIPS_MODE_P:
	if (fips_mode ())
	rc = GPG_ERR_GENERAL; /* Used as TRUE value */
	break;

	case GCRYCTL_FORCE_FIPS_MODE:
	/* Performing this command puts the library into fips mode. If
	the library has already been initialized into fips mode, a
	selftest is triggered. It is not possible to put the libraty
	into fips mode after having passed the initialization. */
	_gcry_set_preferred_rng_type (0);
	if (!_gcry_global_any_init_done)
	{
	/* Not yet initialized at all. Set a flag so that we are put
	into fips mode during initialization. */
	force_fips_mode = 1;
	}
	else
	{
	/* Already initialized. If we are already operational we
	run a selftest. If not we use the is_operational call to
	force us into operational state if possible. */
	if (_gcry_fips_test_error_or_operational ())
	_gcry_fips_run_selftests (1);
	if (_gcry_fips_is_operational ())
	rc = GPG_ERR_GENERAL; /* Used as TRUE value */
	}
	break;

	case GCRYCTL_SELFTEST:
	/* Run a selftest. This works in fips mode as well as in
	standard mode. In contrast to the power-up tests, we use an
	extended version of the selftests. Returns 0 on success or an
	error code. */
	global_init ();
	rc = _gcry_fips_run_selftests (1);
	break;

	case GCRYCTL_FIPS_SERVICE_INDICATOR:
	/* Get FIPS Service Indicator for a given algorithm and optional mode.
	* Returns GPG_ERR_NO_ERROR if algorithm is allowed or
	* GPG_ERR_NOT_SUPPORTED otherwise */
	rc = _gcry_fips_indicator (arg_ptr);
	break;

	case PRIV_CTL_INIT_EXTRNG_TEST: /* Init external random test. */
	rc = GPG_ERR_NOT_SUPPORTED;
	break;
	case PRIV_CTL_RUN_EXTRNG_TEST: /* Run external DRBG test. */
	{
	struct gcry_drbg_test_vector *test =
	va_arg (arg_ptr, struct gcry_drbg_test_vector *);
	unsigned char buf = va_arg (arg_ptr, unsigned char );

	if (buf)
	rc = _gcry_rngdrbg_cavs_test (test, buf);
	else
	rc = _gcry_rngdrbg_healthcheck_one (test);
	}
	break;
	case PRIV_CTL_DEINIT_EXTRNG_TEST: /* Deinit external random test. */
	rc = GPG_ERR_NOT_SUPPORTED;
	break;
	case PRIV_CTL_EXTERNAL_LOCK_TEST: /* Run external lock test */
	rc = external_lock_test (va_arg (arg_ptr, int));
	break;
	case PRIV_CTL_DUMP_SECMEM_STATS:
	_gcry_secmem_dump_stats (1);
	break;

	case GCRYCTL_DISABLE_HWF:
	{
	const char name = va_arg (arg_ptr, const char );
	rc = _gcry_disable_hw_feature (name);
	}
	break;

	case GCRYCTL_SET_ENFORCED_FIPS_FLAG:
	/* Obsolete - ignore */
	break;

	case GCRYCTL_SET_PREFERRED_RNG_TYPE:
	/* This may be called before gcry_check_version. */
	{
	int i = va_arg (arg_ptr, int);
	/* Note that we may not pass 0 to _gcry_set_preferred_rng_type. */
	if (i > 0)
	_gcry_set_preferred_rng_type (i);
	}
	break;

	case GCRYCTL_GET_CURRENT_RNG_TYPE:
	{
	int ip = va_arg (arg_ptr, int);
	if (ip)
	*ip = _gcry_get_rng_type (!_gcry_global_any_init_done);
	}
	break;

	case GCRYCTL_DISABLE_LOCKED_SECMEM:
	_gcry_set_preferred_rng_type (0);
	_gcry_secmem_set_flags ((_gcry_secmem_get_flags ()
	\| GCRY_SECMEM_FLAG_NO_MLOCK));
	break;

	case GCRYCTL_DISABLE_PRIV_DROP:
	_gcry_set_preferred_rng_type (0);
	_gcry_secmem_set_flags ((_gcry_secmem_get_flags ()
	\| GCRY_SECMEM_FLAG_NO_PRIV_DROP));
	break;

	case GCRYCTL_INACTIVATE_FIPS_FLAG:
	case GCRYCTL_REACTIVATE_FIPS_FLAG:
	rc = GPG_ERR_NOT_IMPLEMENTED;
	break;

	case GCRYCTL_DRBG_REINIT:
	{
	const char flagstr = va_arg (arg_ptr, const char );
	gcry_buffer_t pers = va_arg (arg_ptr, gcry_buffer_t );
	int npers = va_arg (arg_ptr, int);
	if (va_arg (arg_ptr, void *) \|\| npers < 0)
	rc = GPG_ERR_INV_ARG;
	else if (_gcry_get_rng_type (!_gcry_global_any_init_done)
	!= GCRY_RNG_TYPE_FIPS)
	rc = GPG_ERR_NOT_SUPPORTED;
	else
	rc = _gcry_rngdrbg_reinit (flagstr, pers, npers);
	}
	break;

	case GCRYCTL_REINIT_SYSCALL_CLAMP:
	if (!pre_syscall_func)
	gpgrt_get_syscall_clamp (&pre_syscall_func, &post_syscall_func);
	break;

	default:
	_gcry_set_preferred_rng_type (0);
	rc = GPG_ERR_INV_OP;
	}

	return rc;
	}

	#if _GCRY_GCC_VERSION >= 40200
	# pragma GCC diagnostic pop
	#endif


	/* Set custom allocation handlers. This is in general not useful
	* because the libgcrypt allocation functions are guaranteed to
	* provide proper allocation handlers which zeroize memory if needed.
	* NOTE: All 5 functions should be set. */
	void
	_gcry_set_allocation_handler (gcry_handler_alloc_t new_alloc_func,
	gcry_handler_alloc_t new_alloc_secure_func,
	gcry_handler_secure_check_t new_is_secure_func,
	gcry_handler_realloc_t new_realloc_func,
	gcry_handler_free_t new_free_func)
	{
	global_init ();

	if (fips_mode ())
	{
	/* In FIPS mode, we can not use custom allocation handlers because
	* fips requires explicit zeroization and we can not guarantee that
	* with custom free functions (and we can not do it transparently as
	* in free we do not know the zize). */
	return;
	}

	alloc_func = new_alloc_func;
	alloc_secure_func = new_alloc_secure_func;
	is_secure_func = new_is_secure_func;
	realloc_func = new_realloc_func;
	free_func = new_free_func;
	}



	/****************
	* Set an optional handler which is called in case the xmalloc functions
	* ran out of memory. This handler may do one of these things:
	* o free some memory and return true, so that the xmalloc function
	* tries again.
	* o Do whatever it like and return false, so that the xmalloc functions
	* use the default fatal error handler.
	* o Terminate the program and don't return.
	*
	* The handler function is called with 3 arguments: The opaque value set with
	* this function, the requested memory size, and a flag with these bits
	* currently defined:
	* bit 0 set = secure memory has been requested.
	*/
	void
	_gcry_set_outofcore_handler (int (f)(void, size_t, unsigned int), void *value)
	{
	global_init ();

	if (fips_mode () )
	{
	log_info ("out of core handler ignored in FIPS mode\n");
	return;
	}

	outofcore_handler = f;
	outofcore_handler_value = value;
	}

	/* Return the no_secure_memory flag. */
	static int
	get_no_secure_memory (void)
	{
	if (!no_secure_memory)
	return 0;
	if (fips_mode ())
	{
	no_secure_memory = 0;
	return 0;
	}
	return no_secure_memory;
	}


	static gcry_err_code_t
	do_malloc (size_t n, unsigned int flags, void **mem)
	{
	gcry_err_code_t err = 0;
	void *m;

	if ((flags & GCRY_ALLOC_FLAG_SECURE) && !get_no_secure_memory ())
	{
	if (alloc_secure_func)
	m = (*alloc_secure_func) (n);
	else
	m = _gcry_private_malloc_secure (n, !!(flags & GCRY_ALLOC_FLAG_XHINT));
	}
	else
	{
	if (alloc_func)
	m = (*alloc_func) (n);
	else
	m = _gcry_private_malloc (n);
	}

	if (!m)
	{
	/* Make sure that ERRNO has been set in case a user supplied
	memory handler didn't it correctly. */
	if (!errno)
	gpg_err_set_errno (ENOMEM);
	err = gpg_err_code_from_errno (errno);
	}
	else
	*mem = m;

	return err;
	}

	void *
	_gcry_malloc (size_t n)
	{
	void *mem = NULL;

	do_malloc (n, 0, &mem);

	return mem;
	}

	static void *
	_gcry_malloc_secure_core (size_t n, int xhint)
	{
	void *mem = NULL;

	do_malloc (n, (GCRY_ALLOC_FLAG_SECURE \| (xhint? GCRY_ALLOC_FLAG_XHINT:0)),
	&mem);

	return mem;
	}

	void *
	_gcry_malloc_secure (size_t n)
	{
	return _gcry_malloc_secure_core (n, 0);
	}

	int
	_gcry_is_secure (const void *a)
	{
	if (get_no_secure_memory ())
	return 0;
	if (is_secure_func)
	return is_secure_func (a) ;
	return _gcry_private_is_secure (a);
	}

	void
	_gcry_check_heap( const void *a )
	{
	(void)a;

	/* FIXME: implement this*/
	#if 0
	if( some_handler )
	some_handler(a)
	else
	_gcry_private_check_heap(a)
	#endif
	}

	static void *
	_gcry_realloc_core (void *a, size_t n, int xhint)
	{
	void *p;

	/* To avoid problems with non-standard realloc implementations and
	our own secmem_realloc, we divert to malloc and free here. */
	if (!a)
	return _gcry_malloc (n);
	if (!n)
	{
	xfree (a);
	return NULL;
	}

	if (realloc_func)
	p = realloc_func (a, n);
	else
	p = _gcry_private_realloc (a, n, xhint);
	if (!p && !errno)
	gpg_err_set_errno (ENOMEM);
	return p;
	}


	void *
	_gcry_realloc (void *a, size_t n)
	{
	return _gcry_realloc_core (a, n, 0);
	}


	void
	_gcry_free (void *p)
	{
	int save_errno;

	if (!p)
	return;

	/* In case ERRNO is set we better save it so that the free machinery
	may not accidentally change ERRNO. We restore it only if it was
	already set to comply with the usual C semantic for ERRNO. */
	save_errno = errno;
	if (free_func)
	free_func (p);
	else
	_gcry_private_free (p);

	if (save_errno && save_errno != errno)
	gpg_err_set_errno (save_errno);
	}

	void *
	_gcry_calloc (size_t n, size_t m)
	{
	size_t bytes;
	void *p;

	bytes = n * m; /* size_t is unsigned so the behavior on overflow is
	defined. */
	if (m && bytes / m != n)
	{
	gpg_err_set_errno (ENOMEM);
	return NULL;
	}

	p = _gcry_malloc (bytes);
	if (p)
	memset (p, 0, bytes);
	return p;
	}

	void *
	_gcry_calloc_secure (size_t n, size_t m)
	{
	size_t bytes;
	void *p;

	bytes = n * m; /* size_t is unsigned so the behavior on overflow is
	defined. */
	if (m && bytes / m != n)
	{
	gpg_err_set_errno (ENOMEM);
	return NULL;
	}

	p = _gcry_malloc_secure (bytes);
	if (p)
	memset (p, 0, bytes);
	return p;
	}


	static char *
	_gcry_strdup_core (const char *string, int xhint)
	{
	char *string_cp = NULL;
	size_t string_n = 0;

	string_n = strlen (string);

	if (_gcry_is_secure (string))
	string_cp = _gcry_malloc_secure_core (string_n + 1, xhint);
	else
	string_cp = _gcry_malloc (string_n + 1);

	if (string_cp)
	strcpy (string_cp, string);

	return string_cp;
	}

	/* Create and return a copy of the null-terminated string STRING. If
	* it is contained in secure memory, the copy will be contained in
	* secure memory as well. In an out-of-memory condition, NULL is
	* returned. */
	char *
	_gcry_strdup (const char *string)
	{
	return _gcry_strdup_core (string, 0);
	}

	void *
	_gcry_xmalloc( size_t n )
	{
	void *p;

	while ( !(p = _gcry_malloc( n )) )
	{
	if ( fips_mode ()
	\|\| !outofcore_handler
	\|\| !outofcore_handler (outofcore_handler_value, n, 0) )
	{
	_gcry_fatal_error (gpg_err_code_from_errno (errno), NULL);
	}
	}
	return p;
	}

	void *
	_gcry_xrealloc( void *a, size_t n )
	{
	void *p;

	while (!(p = _gcry_realloc_core (a, n, 1)))
	{
	if ( fips_mode ()
	\|\| !outofcore_handler
	\|\| !outofcore_handler (outofcore_handler_value, n,
	_gcry_is_secure(a)? 3:2))
	{
	_gcry_fatal_error (gpg_err_code_from_errno (errno), NULL );
	}
	}
	return p;
	}

	void *
	_gcry_xmalloc_secure( size_t n )
	{
	void *p;

	while (!(p = _gcry_malloc_secure_core (n, 1)))
	{
	if ( fips_mode ()
	\|\| !outofcore_handler
	\|\| !outofcore_handler (outofcore_handler_value, n, 1) )
	{
	_gcry_fatal_error (gpg_err_code_from_errno (errno),
	_("out of core in secure memory"));
	}
	}
	return p;
	}


	void *
	_gcry_xcalloc( size_t n, size_t m )
	{
	size_t nbytes;
	void *p;

	nbytes = n * m;
	if (m && nbytes / m != n)
	{
	gpg_err_set_errno (ENOMEM);
	_gcry_fatal_error(gpg_err_code_from_errno (errno), NULL );
	}

	p = _gcry_xmalloc ( nbytes );
	memset ( p, 0, nbytes );
	return p;
	}

	void *
	_gcry_xcalloc_secure( size_t n, size_t m )
	{
	size_t nbytes;
	void *p;

	nbytes = n * m;
	if (m && nbytes / m != n)
	{
	gpg_err_set_errno (ENOMEM);
	_gcry_fatal_error(gpg_err_code_from_errno (errno), NULL );
	}

	p = _gcry_xmalloc_secure ( nbytes );
	memset ( p, 0, nbytes );
	return p;
	}

	char *
	_gcry_xstrdup (const char *string)
	{
	char *p;

	while ( !(p = _gcry_strdup_core (string, 1)) )
	{
	size_t n = strlen (string);
	int is_sec = !!_gcry_is_secure (string);

	if (fips_mode ()
	\|\| !outofcore_handler
	\|\| !outofcore_handler (outofcore_handler_value, n, is_sec) )
	{
	_gcry_fatal_error (gpg_err_code_from_errno (errno),
	is_sec? _("out of core in secure memory"):NULL);
	}
	}

	return p;
	}


	/* Used before blocking system calls. */
	void
	_gcry_pre_syscall (void)
	{
	if (pre_syscall_func)
	pre_syscall_func ();
	}


	/* Used after blocking system calls. */
	void
	_gcry_post_syscall (void)
	{
	if (post_syscall_func)
	post_syscall_func ();
	}


	int
	_gcry_get_debug_flag (unsigned int mask)
	{
	if ( fips_mode () )
	return 0;
	return (debug_flags & mask);
	}



	/* It is often useful to get some feedback of long running operations.
	This function may be used to register a handler for this.
	The callback function CB is used as:

	void cb (void opaque, const char what, int printchar,
	int current, int total);

	Where WHAT is a string identifying the the type of the progress
	output, PRINTCHAR the character usually printed, CURRENT the amount
	of progress currently done and TOTAL the expected amount of
	progress. A value of 0 for TOTAL indicates that there is no
	estimation available.

	Defined values for WHAT:

	"need_entropy" X 0 number-of-bytes-required
	When running low on entropy
	"primegen" '\n' 0 0
	Prime generated
	'!'
	Need to refresh the prime pool
	'<','>'
	Number of bits adjusted
	'^'
	Looking for a generator
	'.'
	Fermat tests on 10 candidates failed
	':'
	Restart with a new random value
	'+'
	Rabin Miller test passed
	"pk_elg" '+','-','.','\n' 0 0
	Only used in debugging mode.
	"pk_dsa"
	Only used in debugging mode.
	*/
	void
	_gcry_set_progress_handler (void (cb)(void ,const char*,int, int, int),
	void *cb_data)
	{
	#if USE_DSA
	_gcry_register_pk_dsa_progress (cb, cb_data);
	#endif
	#if USE_ELGAMAL
	_gcry_register_pk_elg_progress (cb, cb_data);
	#endif
	_gcry_register_primegen_progress (cb, cb_data);
	_gcry_register_random_progress (cb, cb_data);
	}



	/* This is a helper for the regression test suite to test Libgcrypt's locks.
	It works using a one test lock with CMD controlling what to do:

	30111 - Allocate and init lock
	30112 - Take lock
	30113 - Release lock
	30114 - Destroy lock.

	This function is used by tests/t-lock.c - it is not part of the
	public API!
	*/
	static gpg_err_code_t
	external_lock_test (int cmd)
	{
	GPGRT_LOCK_DEFINE (testlock);
	gpg_err_code_t rc = 0;

	switch (cmd)
	{
	case 30111: /* Init Lock. */
	rc = gpgrt_lock_init (&testlock);
	break;

	case 30112: /* Take Lock. */
	rc = gpgrt_lock_lock (&testlock);
	break;

	case 30113: /* Release Lock. */
	rc = gpgrt_lock_unlock (&testlock);
	break;

	case 30114: /* Destroy Lock. */
	rc = gpgrt_lock_destroy (&testlock);
	break;

	default:
	rc = GPG_ERR_INV_OP;
	break;
	}

	return rc;
	}
	diff --git a/tests/benchmark.c b/tests/benchmark.c
	index 2e4efc14..b35b996c 100644
	--- a/tests/benchmark.c
	+++ b/tests/benchmark.c
	@@ -1,2065 +1,2056 @@
	/* benchmark.c - for libgcrypt
	* Copyright (C) 2002, 2004, 2005, 2006, 2008 Free Software Foundation, Inc.
	*
	* This file is part of Libgcrypt.
	*
	* Libgcrypt is free software; you can redistribute it and/or modify
	* it under the terms of the GNU Lesser general Public License as
	* published by the Free Software Foundation; either version 2.1 of
	* the License, or (at your option) any later version.
	*
	* Libgcrypt is distributed in the hope that it will be useful,
	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
	* GNU Lesser General Public License for more details.
	*
	* You should have received a copy of the GNU Lesser General Public
	* License along with this program; if not, see <http://www.gnu.org/licenses/>.
	*/

	#ifdef HAVE_CONFIG_H
	#include <config.h>
	#endif
	#include <stdio.h>
	#include <stdlib.h>
	#include <stdarg.h>

	#ifdef _GCRYPT_IN_LIBGCRYPT
	# include "../src/gcrypt-int.h"
	# include "../compat/libcompat.h"
	#else
	# include <gcrypt.h>
	#endif

	#include "stopwatch.h"


	#define PGM "benchmark"
	#include "t-common.h"

	/* Do encryption tests with large buffers (100 KiB). */
	static int large_buffers;

	/* Do encryption tests with huge buffers (256 MiB). */
	static int huge_buffers;

	/* Number of cipher repetitions. */
	static int cipher_repetitions;

	/* Number of hash repetitions. */
	static int hash_repetitions;

	/* Number of hash repetitions. */
	static int mac_repetitions;

	/* Alignment of the buffers. */
	static int buffer_alignment;

	/* Whether to include the keysetup in the cipher timings. */
	static int cipher_with_keysetup;

	/* Whether fips mode was active at startup. */
	static int in_fips_mode;

	/* Whether we are running as part of the regression test suite. */
	static int in_regression_test;

	/* Whether --progress is in use. */
	static int with_progress;

	/* Runtime flag to switch to a different progress output. */
	static int single_char_progress;


	static const char sample_private_dsa_key_1024[] =
	"(private-key\n"
	" (dsa\n"
	" (p #00A126202D592214C5A8F6016E2C3F4256052ACB1CB17D88E64B1293FAF08F5E4685"
	"03E6F68366B326A56284370EB2103E92D8346A163E44A08FDC422AC8E9E44268557A"
	"853539A6AF39353A59CE5E78FD98B57D0F3E3A7EBC8A256AC9A775BA59689F3004BF"
	"C3035730C4C0C51626C5D7F5852637EC589BB29DAB46C161572E4B#)\n"
	" (q #00DEB5A296421887179ECA1762884DE2AF8185AFC5#)\n"
	" (g #3958B34AE7747194ECBD312F8FEE8CBE3918E94DF9FD11E2912E56318F33BDC38622"
	"B18DDFF393074BCA8BAACF50DF27AEE529F3E8AEECE55C398DAB3A5E04C2EA142312"
	"FACA2FE7F0A88884F8DAC3979EE67598F9A383B2A2325F035C796F352A5C3CDF2CB3"
	"85AD24EC52A6E55247E1BB37D260F79E617D2A4446415B6AD79A#)\n"
	" (y #519E9FE9AB0545A6724E74603B7B04E48DC1437E0284A11EA605A7BA8AB1CF354FD4"
	"ECC93880AC293391C69B558AD84E7AAFA88F11D028CF3A378F241D6B056A90C588F6"
	"66F68D27262B4DA84657D15057D371BCEC1F6504032507D5B881E45FC93A1B973155"
	"D91C57219D090C3ACD75E7C2B9F1176A208AC03D6C12AC28A271#)\n"
	" (x #4186F8A58C5DF46C5BCFC7006BEEBF05E93C0CA7#)\n"
	"))\n";

	static const char sample_public_dsa_key_1024[] =
	"(public-key\n"
	" (dsa\n"
	" (p #00A126202D592214C5A8F6016E2C3F4256052ACB1CB17D88E64B1293FAF08F5E4685"
	"03E6F68366B326A56284370EB2103E92D8346A163E44A08FDC422AC8E9E44268557A"
	"853539A6AF39353A59CE5E78FD98B57D0F3E3A7EBC8A256AC9A775BA59689F3004BF"
	"C3035730C4C0C51626C5D7F5852637EC589BB29DAB46C161572E4B#)\n"
	" (q #00DEB5A296421887179ECA1762884DE2AF8185AFC5#)\n"
	" (g #3958B34AE7747194ECBD312F8FEE8CBE3918E94DF9FD11E2912E56318F33BDC38622"
	"B18DDFF393074BCA8BAACF50DF27AEE529F3E8AEECE55C398DAB3A5E04C2EA142312"
	"FACA2FE7F0A88884F8DAC3979EE67598F9A383B2A2325F035C796F352A5C3CDF2CB3"
	"85AD24EC52A6E55247E1BB37D260F79E617D2A4446415B6AD79A#)\n"
	" (y #519E9FE9AB0545A6724E74603B7B04E48DC1437E0284A11EA605A7BA8AB1CF354FD4"
	"ECC93880AC293391C69B558AD84E7AAFA88F11D028CF3A378F241D6B056A90C588F6"
	"66F68D27262B4DA84657D15057D371BCEC1F6504032507D5B881E45FC93A1B973155"
	"D91C57219D090C3ACD75E7C2B9F1176A208AC03D6C12AC28A271#)\n"
	"))\n";


	static const char sample_private_dsa_key_2048[] =
	"(private-key\n"
	" (dsa\n"
	" (p #00B54636673962B64F7DC23C71ACEF6E7331796F607560B194DFCC0CA370E858A365"
	"A413152FB6EB8C664BD171AC316FE5B381CD084D07377571599880A068EF1382D85C"
	"308B4E9DEAC12D66DE5C4A826EBEB5ED94A62E7301E18927E890589A2F230272A150"
	"C118BC3DC2965AE0D05BE4F65C6137B2BA7EDABB192C3070D202C10AA3F534574970"
	"71454DB8A73DDB6511A5BA98EF1450FD90DE5BAAFC9FD3AC22EBEA612DD075BB7405"
	"D56866D125E33982C046808F7CEBA8E5C0B9F19A6FE451461660A1CBA9EF68891179"
	"0256A573D3B8F35A5C7A0C6C31F2DB90E25A26845252AD9E485EF2D339E7B5890CD4"
	"2F9C9F315ED409171EC35CA04CC06B275577B3#)\n"
	" (q #00DA67989167FDAC4AE3DF9247A716859A30C0CF9C5A6DBA01EABA3481#)\n"
	" (g #48E35DA584A089D05142AA63603FDB00D131B07A0781E2D5A8F9614D2B33D3E40A78"
	"98A9E10CDBB612CF093F95A3E10D09566726F2C12823836B2D9CD974BB695665F3B3"
	"5D219A9724B87F380BD5207EDA0AE38C79E8F18122C3F76E4CEB0ABED3250914987F"
	"B30D4B9E19C04C28A5D4F45560AF586F6A1B41751EAD90AE7F044F4E2A4A50C1F508"
	"4FC202463F478F678B9A19392F0D2961C5391C546EF365368BB46410C9C1CEE96E9F"
	"0C953570C2ED06328B11C90E86E57CAA7FA5ABAA278E22A4C8C08E16EE59F484EC44"
	"2CF55535BAA2C6BEA8833A555372BEFE1E665D3C7DAEF58061D5136331EF4EB61BC3"
	"6EE4425A553AF8885FEA15A88135BE133520#)\n"
	" (y #66E0D1A69D663466F8FEF2B7C0878DAC93C36A2FB2C05E0306A53B926021D4B92A1C"
	"2FA6860061E88E78CBBBA49B0E12700F07DBF86F72CEB2927EDAC0C7E3969C3A47BB"
	"4E0AE93D8BB3313E93CC7A72DFEEE442EFBC81B3B2AEC9D8DCBE21220FB760201D79"
	"328C41C773866587A44B6954767D022A88072900E964089D9B17133603056C985C4F"
	"8A0B648F297F8D2C3CB43E4371DC6002B5B12CCC085BDB2CFC5074A0587566187EE3"
	"E11A2A459BD94726248BB8D6CC62938E11E284C2C183576FBB51749EB238C4360923"
	"79C08CE1C8CD77EB57404CE9B4744395ACF721487450BADE3220576F2F816248B0A7"
	"14A264330AECCB24DE2A1107847B23490897#)\n"
	" (x #477BD14676E22563C5ABA68025CEBA2A48D485F5B2D4AD4C0EBBD6D0#)\n"
	"))\n";


	static const char sample_public_dsa_key_2048[] =
	"(public-key\n"
	" (dsa\n"
	" (p #00B54636673962B64F7DC23C71ACEF6E7331796F607560B194DFCC0CA370E858A365"
	"A413152FB6EB8C664BD171AC316FE5B381CD084D07377571599880A068EF1382D85C"
	"308B4E9DEAC12D66DE5C4A826EBEB5ED94A62E7301E18927E890589A2F230272A150"
	"C118BC3DC2965AE0D05BE4F65C6137B2BA7EDABB192C3070D202C10AA3F534574970"
	"71454DB8A73DDB6511A5BA98EF1450FD90DE5BAAFC9FD3AC22EBEA612DD075BB7405"
	"D56866D125E33982C046808F7CEBA8E5C0B9F19A6FE451461660A1CBA9EF68891179"
	"0256A573D3B8F35A5C7A0C6C31F2DB90E25A26845252AD9E485EF2D339E7B5890CD4"
	"2F9C9F315ED409171EC35CA04CC06B275577B3#)\n"
	" (q #00DA67989167FDAC4AE3DF9247A716859A30C0CF9C5A6DBA01EABA3481#)\n"
	" (g #48E35DA584A089D05142AA63603FDB00D131B07A0781E2D5A8F9614D2B33D3E40A78"
	"98A9E10CDBB612CF093F95A3E10D09566726F2C12823836B2D9CD974BB695665F3B3"
	"5D219A9724B87F380BD5207EDA0AE38C79E8F18122C3F76E4CEB0ABED3250914987F"
	"B30D4B9E19C04C28A5D4F45560AF586F6A1B41751EAD90AE7F044F4E2A4A50C1F508"
	"4FC202463F478F678B9A19392F0D2961C5391C546EF365368BB46410C9C1CEE96E9F"
	"0C953570C2ED06328B11C90E86E57CAA7FA5ABAA278E22A4C8C08E16EE59F484EC44"
	"2CF55535BAA2C6BEA8833A555372BEFE1E665D3C7DAEF58061D5136331EF4EB61BC3"
	"6EE4425A553AF8885FEA15A88135BE133520#)\n"
	" (y #66E0D1A69D663466F8FEF2B7C0878DAC93C36A2FB2C05E0306A53B926021D4B92A1C"
	"2FA6860061E88E78CBBBA49B0E12700F07DBF86F72CEB2927EDAC0C7E3969C3A47BB"
	"4E0AE93D8BB3313E93CC7A72DFEEE442EFBC81B3B2AEC9D8DCBE21220FB760201D79"
	"328C41C773866587A44B6954767D022A88072900E964089D9B17133603056C985C4F"
	"8A0B648F297F8D2C3CB43E4371DC6002B5B12CCC085BDB2CFC5074A0587566187EE3"
	"E11A2A459BD94726248BB8D6CC62938E11E284C2C183576FBB51749EB238C4360923"
	"79C08CE1C8CD77EB57404CE9B4744395ACF721487450BADE3220576F2F816248B0A7"
	"14A264330AECCB24DE2A1107847B23490897#)\n"
	"))\n";


	static const char sample_private_dsa_key_3072[] =
	"(private-key\n"
	" (dsa\n"
	" (p #00BA73E148AEA5E8B64878AF5BE712B8302B9671C5F3EEB7722A9D0D9868D048C938"
	"877C91C335C7819292E69C7D34264F1578E32EC2DA8408DF75D0EB76E0D3030B84B5"
	"62D8EF93AB53BAB6B8A5DE464F5CA87AEA43BDCF0FB0B7815AA3114CFC84FD916A83"
	"B3D5FD78390189332232E9D037D215313FD002FF46C048B66703F87FAE092AAA0988"
	"AC745336EBE672A01DEDBD52395783579B67CF3AE1D6F1602CCCB12154FA0E00AE46"
	"0D9B289CF709194625BCB919B11038DEFC50ADBBA20C3F320078E4E9529B4F6848E2"
	"AB5E6278DB961FE226F2EEBD201E071C48C5BEF98B4D9BEE42C1C7102D893EBF8902"
	"D7A91266340AFD6CE1D09E52282FFF5B97EAFA3886A3FCF84FF76D1E06538D0D8E60"
	"B3332145785E07D29A5965382DE3470D1D888447FA9C00A2373378FC3FA7B9F7D17E"
	"95A6A5AE1397BE46D976EF2C96E89913AC4A09351CA661BF6F67E30407DA846946C7"
	"62D9BAA6B77825097D3E7B886456BB32E3E74516BF3FD93D71B257AA8F723E01CE33"
	"8015353D3778B02B892AF7#)\n"
	" (q #00BFF3F3CC18FA018A5B8155A8695E1E4939660D5E4759322C39D50F3B93E5F68B#)\n"
	" (g #6CCFD8219F5FCE8EF2BEF3262929787140847E38674B1EF8DB20255E212CB6330EC4"
	"DFE8A26AB7ECC5760DEB9BBF59A2B2821D510F1868172222867558B8D204E889C474"
	"7CA30FBF9D8CF41AE5D5BD845174641101593849FF333E6C93A6550931B2B9D56B98"
	"9CAB01729D9D736FA6D24A74D2DDE1E9E648D141473E443DD6BBF0B3CAB64F9FE4FC"
	"134B2EB57437789F75C744DF1FA67FA8A64603E5441BC7ECE29E00BDF262BDC81E8C"
	"7330A18A412DE38E7546D342B89A0AF675A89E6BEF00540EB107A2FE74EA402B0D89"
	"F5C02918DEEEAF8B8737AC866B09B50810AB8D8668834A1B9E1E53866E2B0A926FAB"
	"120A0CDE5B3715FFFE6ACD1AB73588DCC1EC4CE9392FE57F8D1D35811200CB07A0E6"
	"374E2C4B0AEB7E3D077B8545C0E438DCC0F1AE81E186930E99EBC5B91B77E92803E0"
	"21602887851A4FFDB3A7896AC655A0901218C121C5CBB0931E7D5EAC243F37711B5F"
	"D5A62B1B38A83F03D8F6703D8B98DF367FC8A76990335F62173A5391836F0F2413EC"
	"4997AF9EB55C6660B01A#)\n"
	" (y #2320B22434C5DB832B4EC267CC52E78DD5CCFA911E8F0804E7E7F32B186B2D4167AE"
	"4AA6869822E76400492D6A193B0535322C72B0B7AA4A87E33044FDC84BE24C64A053"
	"A37655EE9EABDCDC1FDF63F3F1C677CEB41595DF7DEFE9178D85A3D621B4E4775492"
	"8C0A58D2458D06F9562E4DE2FE6129A64063A99E88E54485B97484A28188C4D33F15"
	"DDC903B6CEA0135E3E3D27B4EA39319696305CE93D7BA7BE00367DBE3AAF43491E71"
	"CBF254744A5567F5D70090D6139E0C990239627B3A1C5B20B6F9F6374B8D8D8A8997"
	"437265BE1E3B4810D4B09254400DE287A0DFFBAEF339E48D422B1D41A37E642BC026"
	"73314701C8FA9792845C129351A87A945A03E6C895860E51D6FB8B7340A94D1A8A7B"
	"FA85AC83B4B14E73AB86CB96C236C8BFB0978B61B2367A7FE4F7891070F56C78D5DD"
	"F5576BFE5BE4F333A4E2664E79528B3294907AADD63F4F2E7AA8147B928D8CD69765"
	"3DB98C4297CB678046ED55C0DBE60BF7142C594603E4D705DC3D17270F9F086EC561"
	"2703D518D8D49FF0EBE6#)\n"
	" (x #00A9FFFC88E67D6F7B810E291C050BAFEA7FC4A75E8D2F16CFED3416FD77607232#)\n"
	"))\n";

	static const char sample_public_dsa_key_3072[] =
	"(public-key\n"
	" (dsa\n"
	" (p #00BA73E148AEA5E8B64878AF5BE712B8302B9671C5F3EEB7722A9D0D9868D048C938"
	"877C91C335C7819292E69C7D34264F1578E32EC2DA8408DF75D0EB76E0D3030B84B5"
	"62D8EF93AB53BAB6B8A5DE464F5CA87AEA43BDCF0FB0B7815AA3114CFC84FD916A83"
	"B3D5FD78390189332232E9D037D215313FD002FF46C048B66703F87FAE092AAA0988"
	"AC745336EBE672A01DEDBD52395783579B67CF3AE1D6F1602CCCB12154FA0E00AE46"
	"0D9B289CF709194625BCB919B11038DEFC50ADBBA20C3F320078E4E9529B4F6848E2"
	"AB5E6278DB961FE226F2EEBD201E071C48C5BEF98B4D9BEE42C1C7102D893EBF8902"
	"D7A91266340AFD6CE1D09E52282FFF5B97EAFA3886A3FCF84FF76D1E06538D0D8E60"
	"B3332145785E07D29A5965382DE3470D1D888447FA9C00A2373378FC3FA7B9F7D17E"
	"95A6A5AE1397BE46D976EF2C96E89913AC4A09351CA661BF6F67E30407DA846946C7"
	"62D9BAA6B77825097D3E7B886456BB32E3E74516BF3FD93D71B257AA8F723E01CE33"
	"8015353D3778B02B892AF7#)\n"
	" (q #00BFF3F3CC18FA018A5B8155A8695E1E4939660D5E4759322C39D50F3B93E5F68B#)\n"
	" (g #6CCFD8219F5FCE8EF2BEF3262929787140847E38674B1EF8DB20255E212CB6330EC4"
	"DFE8A26AB7ECC5760DEB9BBF59A2B2821D510F1868172222867558B8D204E889C474"
	"7CA30FBF9D8CF41AE5D5BD845174641101593849FF333E6C93A6550931B2B9D56B98"
	"9CAB01729D9D736FA6D24A74D2DDE1E9E648D141473E443DD6BBF0B3CAB64F9FE4FC"
	"134B2EB57437789F75C744DF1FA67FA8A64603E5441BC7ECE29E00BDF262BDC81E8C"
	"7330A18A412DE38E7546D342B89A0AF675A89E6BEF00540EB107A2FE74EA402B0D89"
	"F5C02918DEEEAF8B8737AC866B09B50810AB8D8668834A1B9E1E53866E2B0A926FAB"
	"120A0CDE5B3715FFFE6ACD1AB73588DCC1EC4CE9392FE57F8D1D35811200CB07A0E6"
	"374E2C4B0AEB7E3D077B8545C0E438DCC0F1AE81E186930E99EBC5B91B77E92803E0"
	"21602887851A4FFDB3A7896AC655A0901218C121C5CBB0931E7D5EAC243F37711B5F"
	"D5A62B1B38A83F03D8F6703D8B98DF367FC8A76990335F62173A5391836F0F2413EC"
	"4997AF9EB55C6660B01A#)\n"
	" (y #2320B22434C5DB832B4EC267CC52E78DD5CCFA911E8F0804E7E7F32B186B2D4167AE"
	"4AA6869822E76400492D6A193B0535322C72B0B7AA4A87E33044FDC84BE24C64A053"
	"A37655EE9EABDCDC1FDF63F3F1C677CEB41595DF7DEFE9178D85A3D621B4E4775492"
	"8C0A58D2458D06F9562E4DE2FE6129A64063A99E88E54485B97484A28188C4D33F15"
	"DDC903B6CEA0135E3E3D27B4EA39319696305CE93D7BA7BE00367DBE3AAF43491E71"
	"CBF254744A5567F5D70090D6139E0C990239627B3A1C5B20B6F9F6374B8D8D8A8997"
	"437265BE1E3B4810D4B09254400DE287A0DFFBAEF339E48D422B1D41A37E642BC026"
	"73314701C8FA9792845C129351A87A945A03E6C895860E51D6FB8B7340A94D1A8A7B"
	"FA85AC83B4B14E73AB86CB96C236C8BFB0978B61B2367A7FE4F7891070F56C78D5DD"
	"F5576BFE5BE4F333A4E2664E79528B3294907AADD63F4F2E7AA8147B928D8CD69765"
	"3DB98C4297CB678046ED55C0DBE60BF7142C594603E4D705DC3D17270F9F086EC561"
	"2703D518D8D49FF0EBE6#)\n"
	"))\n";


	static const char sample_public_elg_key_1024[] =
	"(public-key"
	" (elg"
	" (p #00F7CC7C08AF096B620C545C9353B1140D698FF8BE2D97A3515C17C7F8DABCDB8FB6"
	"64A46416C90C530C18DF5ABB6C1DDE3AE2FA9DDC9CE40DF644CDE2E759F6DE43F31A"
	"EEEBC136A460B3E4B0A8F99326A335145B19F4C81B13804894B7D2A30F78A8A7D7F4"
	"52B83836FDB0DE90BE327FB5E5318757BEF5FE0FC3A5461CBEA0D3#)"
	" (g #06#)"
	" (y #36B38FB63E3340A0DD8A0468E9FAA512A32DA010BF7110201D0A3DF1B8FEA0E16F3C"
	"80374584E554804B96EAA8C270FE531F75D0DBD81BA65640EDB1F76D46C27D2925B7"
	"3EC3B295CDAEEF242904A84D74FB2879425F82D4C5B59BB49A992F85D574168DED85"
	"D227600BBEF7AF0B8F0DEB785528370E4C4B3E4D65C536122A5A#)"
	" ))";
	static const char sample_private_elg_key_1024[] =
	"(private-key"
	" (elg"
	" (p #00F7CC7C08AF096B620C545C9353B1140D698FF8BE2D97A3515C17C7F8DABCDB8FB6"
	"64A46416C90C530C18DF5ABB6C1DDE3AE2FA9DDC9CE40DF644CDE2E759F6DE43F31A"
	"EEEBC136A460B3E4B0A8F99326A335145B19F4C81B13804894B7D2A30F78A8A7D7F4"
	"52B83836FDB0DE90BE327FB5E5318757BEF5FE0FC3A5461CBEA0D3#)"
	" (g #06#)"
	" (y #36B38FB63E3340A0DD8A0468E9FAA512A32DA010BF7110201D0A3DF1B8FEA0E16F3C"
	"80374584E554804B96EAA8C270FE531F75D0DBD81BA65640EDB1F76D46C27D2925B7"
	"3EC3B295CDAEEF242904A84D74FB2879425F82D4C5B59BB49A992F85D574168DED85"
	"D227600BBEF7AF0B8F0DEB785528370E4C4B3E4D65C536122A5A#)"
	" (x #03656C6186FCD27D4A4B1F5010DC0D2AE7833B501E423FCD51DE5EB6D80DACFE#)"
	" ))";


	static const char sample_public_elg_key_2048[] =
	"(public-key"
	" (elg"
	" (p #00BE5A2BB4E562D7B644E3D01321CB818DBA27295C339FC2C47EAE9823225EE1E7B6"
	"38C5DE300E931080E09CC89A18C9D180C16559FEF0D89D6A09534BB86489CCCEE30D"
	"C18E007A8726BB99F2B2D90D2694597757B120CD2435C0098AD1B74C20004C25BA97"
	"73EAA4FBEC594EE17F8B25867EEB0F9F857C751116ADED68ADA2A1E9F9F4F40D18F0"
	"EC1221CA6A746FC5F4CDA2B8B5D0AB83834564ACF6FDBB1AB01D4BFBD1E2C0108FF5"
	"5FB3190C6D6DA4D95EA683EFA44935CFBC0BF5C6118ACC3768AEA9A98D06024841B8"
	"D07C234289D22A5E3948F199C397AA991C59A55BEA0C01E91902E039116946FEA135"
	"768011AF6B622C5AF366EF0196FC4EAEAA8127#)"
	" (g #07#)"
	" (y #5AFF87BC23D8B97AA62897A5C1CDFFA86C59F39EDBD6012B6F333CE23D872009B8C8"
	"D1E220E18CFCADFE0AA16346BA2EA132472FFEC746D11C6E758896052313BB501210"
	"2389C683A25A3464E9B35A192BAE0A3BB99C973126F7560D968C4A754901DC967354"
	"D61A90ACD56D90DCC4337AFB71FAE3FD18C60EB0D6DD173877DF5DB5199C4931FE4E"
	"5046F814422580E1162798406FC6554781142DBB7922D4B5B37A111F23761636090F"
	"6212681E133365191CF15753AE737F17943ED4B7506DE0A85C3B6D63227F9D65ADF8"
	"2C3DF0676C8F43B5B1C07D9AD4E6D0C812401D7DA7B9484DBA8CD3B73B19A95EB237"
	"D493E092AEA2371AA904009C8960B0969D12#)"
	" ))";
	static const char sample_private_elg_key_2048[] =
	"(private-key"
	" (elg"
	" (p #00BE5A2BB4E562D7B644E3D01321CB818DBA27295C339FC2C47EAE9823225EE1E7B6"
	"38C5DE300E931080E09CC89A18C9D180C16559FEF0D89D6A09534BB86489CCCEE30D"
	"C18E007A8726BB99F2B2D90D2694597757B120CD2435C0098AD1B74C20004C25BA97"
	"73EAA4FBEC594EE17F8B25867EEB0F9F857C751116ADED68ADA2A1E9F9F4F40D18F0"
	"EC1221CA6A746FC5F4CDA2B8B5D0AB83834564ACF6FDBB1AB01D4BFBD1E2C0108FF5"
	"5FB3190C6D6DA4D95EA683EFA44935CFBC0BF5C6118ACC3768AEA9A98D06024841B8"
	"D07C234289D22A5E3948F199C397AA991C59A55BEA0C01E91902E039116946FEA135"
	"768011AF6B622C5AF366EF0196FC4EAEAA8127#)"
	" (g #07#)"
	" (y #5AFF87BC23D8B97AA62897A5C1CDFFA86C59F39EDBD6012B6F333CE23D872009B8C8"
	"D1E220E18CFCADFE0AA16346BA2EA132472FFEC746D11C6E758896052313BB501210"
	"2389C683A25A3464E9B35A192BAE0A3BB99C973126F7560D968C4A754901DC967354"
	"D61A90ACD56D90DCC4337AFB71FAE3FD18C60EB0D6DD173877DF5DB5199C4931FE4E"
	"5046F814422580E1162798406FC6554781142DBB7922D4B5B37A111F23761636090F"
	"6212681E133365191CF15753AE737F17943ED4B7506DE0A85C3B6D63227F9D65ADF8"
	"2C3DF0676C8F43B5B1C07D9AD4E6D0C812401D7DA7B9484DBA8CD3B73B19A95EB237"
	"D493E092AEA2371AA904009C8960B0969D12#)"
	" (x #0628C3903972C55BDC1BC4223075616D3F3BA57D55532DDB40CB14CF72070E0D28BF"
	"D0402B9088D25ED8FC#)"
	" ))";

	static const char sample_public_elg_key_3072[] =
	"(public-key"
	" (elg"
	" (p #008EAA3497AFE3706E1A57FFA52E68C64C500731B58EBAFEB51C4A20AB15BA57FA72"
	"BA1510A4703D5AA6F05DB67E4A776F92AD08800577DC686D00B793167A5D79C997E0"
	"5B9A9E5974B4B68B4D71ED8EC37F2F45235D901997D72915643F058E712AA18275A2"
	"C6F9F7C2B9B7CD1E814D215F12A840800B546AEF2A2E6C077CDD1A322738FFD36DB2"
	"FA5420B5848EED870BC1A6CF55040AE8D2A5945F11AE2BCBE107B41A59EFDBD3B05C"
	"F4C876C02C9AEAE22CD4C86806A415302936E4C1E5AA59DBBCCD2F83C20941A29888"
	"A70ADB94D3B8A6489C46BF2C5219CD9FD2341EA21D4E68A4ECC468FD09D215FE96D4"
	"7AEA12FD22B2456D2CC13672FC7E9772A365C68668157C51E46966B6A1831C429BA0"
	"D513519713C49C13C5FC7C14BE0A117627B204C4478D0A93C6B57929E448C9B65BF2"
	"390E04BC5940320C0262FC1A221E7C796493432239A6F12BC62C5CF32E8ADBC1730C"
	"84C6E6E6BD95AF62835941F3F344AF46BFE5A8F629D5FA699FE37EF8B8C6A2484E42"
	"D226206FDF7D1FB93A5457#)"
	" (g #0B#)"
	" (y #18E734FF645AE169079AEAFC78772371089AD3088627ECF77034AFBDF33ADF594AAF"
	"3288F6979E0DB59CE3D2F0FEE031DFF187F1E4549D3C79668794CB19C14481ECDE2D"
	"D50861AB674F87A011D50D35F28E424D0D2353850899C2CDD0CC8FDBFC5A0CA395F0"
	"E605D46CBDD140DBEF426EBD638C9ADD83C195C45CE84ED2D2B21B87800C783A4F79"
	"12226FEFBDA01C66B254534A51765AF09687275AA80C5DFBA143A6262E47C547D7E2"
	"289413F8C5C56AED3FA7E5DF5526958E2294FE318AF590C0E720029C202563E6E686"
	"9EC810F39A859262FB6047C1D418CAA9047A00BDB127B44B69CF6BC8E6B3709B4C23"
	"79783C5F8457EFE23EDA6FF00D1DDCC29268FC4A6C18577BE2B7004089CBB824027A"
	"A53C86B51DB054CC83B4F50C8923E2E9431F0A77D741237226CC68591083A2E40171"
	"5C7B74100BB74003E2264F8B44A0B0BC5404C44218ABE65C04AA573877506CE4F48C"
	"9E3F8AD1CD8DD9F285DD015C2FC5DEBCFA5779AD87F0BBC62E9EC6246021AB450DB9"
	"4DDDEFAFD2C7C66E235D#)"
	" ))";
	static const char sample_private_elg_key_3072[] =
	"(private-key"
	" (elg"
	" (p #008EAA3497AFE3706E1A57FFA52E68C64C500731B58EBAFEB51C4A20AB15BA57FA72"
	"BA1510A4703D5AA6F05DB67E4A776F92AD08800577DC686D00B793167A5D79C997E0"
	"5B9A9E5974B4B68B4D71ED8EC37F2F45235D901997D72915643F058E712AA18275A2"
	"C6F9F7C2B9B7CD1E814D215F12A840800B546AEF2A2E6C077CDD1A322738FFD36DB2"
	"FA5420B5848EED870BC1A6CF55040AE8D2A5945F11AE2BCBE107B41A59EFDBD3B05C"
	"F4C876C02C9AEAE22CD4C86806A415302936E4C1E5AA59DBBCCD2F83C20941A29888"
	"A70ADB94D3B8A6489C46BF2C5219CD9FD2341EA21D4E68A4ECC468FD09D215FE96D4"
	"7AEA12FD22B2456D2CC13672FC7E9772A365C68668157C51E46966B6A1831C429BA0"
	"D513519713C49C13C5FC7C14BE0A117627B204C4478D0A93C6B57929E448C9B65BF2"
	"390E04BC5940320C0262FC1A221E7C796493432239A6F12BC62C5CF32E8ADBC1730C"
	"84C6E6E6BD95AF62835941F3F344AF46BFE5A8F629D5FA699FE37EF8B8C6A2484E42"
	"D226206FDF7D1FB93A5457#)"
	" (g #0B#)"
	" (y #18E734FF645AE169079AEAFC78772371089AD3088627ECF77034AFBDF33ADF594AAF"
	"3288F6979E0DB59CE3D2F0FEE031DFF187F1E4549D3C79668794CB19C14481ECDE2D"
	"D50861AB674F87A011D50D35F28E424D0D2353850899C2CDD0CC8FDBFC5A0CA395F0"
	"E605D46CBDD140DBEF426EBD638C9ADD83C195C45CE84ED2D2B21B87800C783A4F79"
	"12226FEFBDA01C66B254534A51765AF09687275AA80C5DFBA143A6262E47C547D7E2"
	"289413F8C5C56AED3FA7E5DF5526958E2294FE318AF590C0E720029C202563E6E686"
	"9EC810F39A859262FB6047C1D418CAA9047A00BDB127B44B69CF6BC8E6B3709B4C23"
	"79783C5F8457EFE23EDA6FF00D1DDCC29268FC4A6C18577BE2B7004089CBB824027A"
	"A53C86B51DB054CC83B4F50C8923E2E9431F0A77D741237226CC68591083A2E40171"
	"5C7B74100BB74003E2264F8B44A0B0BC5404C44218ABE65C04AA573877506CE4F48C"
	"9E3F8AD1CD8DD9F285DD015C2FC5DEBCFA5779AD87F0BBC62E9EC6246021AB450DB9"
	"4DDDEFAFD2C7C66E235D#)"
	" (x #03A73F0389E470AAC831B039F8AA0C4EBD3A47DD083E32EEA08E4911236CD597C272"
	"9823D47A51C8535DA52FE6DAB3E8D1C20D#)"
	" ))";


	#define BUG() do {fprintf ( stderr, "Ooops at %s:%d\n", __FILE__ , __LINE__ );\
	exit(2);} while(0)


	static void
	show_sexp (const char *prefix, gcry_sexp_t a)
	{
	char *buf;
	size_t size;

	fputs (prefix, stderr);
	size = gcry_sexp_sprint (a, GCRYSEXP_FMT_ADVANCED, NULL, 0);
	buf = malloc (size);
	if (!buf)
	die ("out of core\n");

	gcry_sexp_sprint (a, GCRYSEXP_FMT_ADVANCED, buf, size);
	fprintf (stderr, "%.*s", (int)size, buf);
	}


	static void
	progress_cb (void cb_data, const char what, int printchar,
	int current, int total)
	{
	(void)cb_data;

	if (single_char_progress)
	{
	fputc (printchar, stdout);
	fflush (stderr);
	}
	else
	{
	fprintf (stderr, PGM ": progress (%s %c %d %d)\n",
	what, printchar, current, total);
	fflush (stderr);
	}
	}


	static void
	random_bench (int very_strong)
	{
	char buf[128];
	int i;

	printf ("%-10s", "random");

	if (!very_strong)
	{
	start_timer ();
	for (i=0; i < 100; i++)
	gcry_randomize (buf, sizeof buf, GCRY_STRONG_RANDOM);
	stop_timer ();
	printf (" %s", elapsed_time (1));
	}

	start_timer ();
	for (i=0; i < 100; i++)
	gcry_randomize (buf, 8,
	very_strong? GCRY_VERY_STRONG_RANDOM:GCRY_STRONG_RANDOM);
	stop_timer ();
	printf (" %s", elapsed_time (1));

	putchar ('\n');
	if (verbose)
	xgcry_control ((GCRYCTL_DUMP_RANDOM_STATS));
	}



	static void
	md_bench ( const char *algoname )
	{
	int algo;
	gcry_md_hd_t hd;
	int i, j, repcount;
	char buf_base[1000+15];
	size_t bufsize = 1000;
	char *buf;
	char *largebuf_base;
	char *largebuf;
	char digest[512/8];
	gcry_error_t err = GPG_ERR_NO_ERROR;

	if (!algoname)
	{
	for (i=1; i < 400; i++)
	if (in_fips_mode && i == GCRY_MD_MD5)
	; /* Don't use MD5 in fips mode. */
	else if ( !gcry_md_test_algo (i) )
	md_bench (gcry_md_algo_name (i));
	return;
	}

	buf = buf_base + ((16 - ((size_t)buf_base & 0x0f)) % buffer_alignment);

	algo = gcry_md_map_name (algoname);
	if (!algo)
	{
	fprintf (stderr, PGM ": invalid hash algorithm `%s'\n", algoname);
	exit (1);
	}

	err = gcry_md_open (&hd, algo, 0);
	if (err)
	{
	fprintf (stderr, PGM ": error opening hash algorithm `%s'\n", algoname);
	exit (1);
	}

	for (i=0; i < bufsize; i++)
	buf[i] = i;

	printf ("%-12s", gcry_md_algo_name (algo));

	start_timer ();
	for (repcount=0; repcount < hash_repetitions; repcount++)
	for (i=0; i < 1000; i++)
	gcry_md_write (hd, buf, bufsize);
	gcry_md_final (hd);
	stop_timer ();
	printf (" %s", elapsed_time (1));
	fflush (stdout);

	gcry_md_reset (hd);
	start_timer ();
	for (repcount=0; repcount < hash_repetitions; repcount++)
	for (i=0; i < 10000; i++)
	gcry_md_write (hd, buf, bufsize/10);
	gcry_md_final (hd);
	stop_timer ();
	printf (" %s", elapsed_time (1));
	fflush (stdout);

	gcry_md_reset (hd);
	start_timer ();
	for (repcount=0; repcount < hash_repetitions; repcount++)
	for (i=0; i < 1000000; i++)
	gcry_md_write (hd, buf, 1);
	gcry_md_final (hd);
	stop_timer ();
	printf (" %s", elapsed_time (1));
	fflush (stdout);

	start_timer ();
	for (repcount=0; repcount < hash_repetitions; repcount++)
	for (i=0; i < 1000; i++)
	for (j=0; j < bufsize; j++)
	gcry_md_putc (hd, buf[j]);
	gcry_md_final (hd);
	stop_timer ();
	printf (" %s", elapsed_time (1));
	fflush (stdout);

	gcry_md_close (hd);

	/* Now 100 hash operations on 10000 bytes using the fast function.
	We initialize the buffer so that all memory pages are committed
	and we have repeatable values. */
	if (gcry_md_get_algo_dlen (algo) > sizeof digest)
	die ("digest buffer too short\n");

	if (gcry_md_get_algo_dlen (algo))
	{
	largebuf_base = malloc (10000+15);
	if (!largebuf_base)
	die ("out of core\n");
	largebuf = (largebuf_base
	+ ((16 - ((size_t)largebuf_base & 0x0f)) % buffer_alignment));

	for (i=0; i < 10000; i++)
	largebuf[i] = i;
	start_timer ();
	for (repcount=0; repcount < hash_repetitions; repcount++)
	for (i=0; i < 100; i++)
	gcry_md_hash_buffer (algo, digest, largebuf, 10000);
	stop_timer ();
	printf (" %s", elapsed_time (1));
	free (largebuf_base);
	}

	putchar ('\n');
	fflush (stdout);
	}



	static void
	mac_bench ( const char *algoname )
	{
	int algo;
	gcry_mac_hd_t hd;
	int step, pos, j, i, repcount;
	char buf_base[1000+15];
	size_t bufsize = 1000;
	char *buf;
	char mac[3][512];
	char key[512];
	unsigned int maclen, keylen;
	size_t macoutlen;
	gcry_error_t err = GPG_ERR_NO_ERROR;

	if (!algoname)
	{
	for (i=1; i < 600; i++)
	if (in_fips_mode && i == GCRY_MAC_HMAC_MD5)
	; /* Don't use MD5 in fips mode. */
	else if ( !gcry_mac_test_algo (i) )
	mac_bench (gcry_mac_algo_name (i));
	return;
	}

	buf = buf_base + ((16 - ((size_t)buf_base & 0x0f)) % buffer_alignment);

	algo = gcry_mac_map_name (algoname);
	if (!algo)
	{
	fprintf (stderr, PGM ": invalid MAC algorithm `%s'\n", algoname);
	exit (1);
	}

	maclen = gcry_mac_get_algo_maclen (algo);
	if (maclen > sizeof(mac))
	maclen = sizeof(mac);

	keylen = gcry_mac_get_algo_keylen (algo);
	if (keylen == 0)
	keylen = 32;
	if (keylen > sizeof(key))
	keylen = sizeof(key);
	for (i=0; i < keylen; i++)
	key[i] = (keylen - i) ^ 0x54;

	err = gcry_mac_open (&hd, algo, 0, NULL);
	if (err)
	{
	fprintf (stderr, PGM ": error opening mac algorithm `%s': %s\n", algoname,
	gpg_strerror (err));
	exit (1);
	}

	err = gcry_mac_setkey (hd, key, keylen);
	if (err)
	{
	fprintf (stderr, PGM ": error setting key for mac algorithm `%s': %s\n",
	algoname, gpg_strerror (err));
	exit (1);
	}

	for (i=0; i < bufsize; i++)
	buf[i] = i;

	if (algo >= GCRY_MAC_POLY1305_AES && algo <= GCRY_MAC_POLY1305_SEED)
	{
	static const char iv[16] = { 1, 2, 3, 4, };
	err = gcry_mac_setiv(hd, iv, sizeof(iv));
	if (err)
	{
	fprintf (stderr, PGM ": error setting nonce for mac algorithm `%s': %s\n",
	algoname, gpg_strerror (err));
	exit (1);
	}
	}

	printf ("%-20s", gcry_mac_algo_name (algo));

	start_timer ();
	for (repcount=0; repcount < mac_repetitions; repcount++)
	for (i=0; i < 1000; i++)
	gcry_mac_write (hd, buf, bufsize);
	macoutlen = maclen;
	gcry_mac_read (hd, mac[0], &macoutlen);
	stop_timer ();
	printf (" %s", elapsed_time (1));
	fflush (stdout);

	gcry_mac_reset (hd);
	start_timer ();
	for (repcount=0; repcount < mac_repetitions; repcount++)
	for (i=0; i < 1000; i++)
	for (step=bufsize/10, pos=0, j=0; j < 10; j++, pos+=step)
	gcry_mac_write (hd, &buf[pos], step);
	macoutlen = maclen;
	gcry_mac_read (hd, mac[1], &macoutlen);
	stop_timer ();
	printf (" %s", elapsed_time (1));
	fflush (stdout);

	gcry_mac_reset (hd);
	start_timer ();
	for (repcount=0; repcount < mac_repetitions; repcount++)
	for (i=0; i < 1000; i++)
	for (step=bufsize/100, pos=0, j=0; j < 100; j++, pos+=step)
	gcry_mac_write (hd, &buf[pos], step);
	macoutlen = maclen;
	gcry_mac_read (hd, mac[2], &macoutlen);
	stop_timer ();
	printf (" %s", elapsed_time (1));
	fflush (stdout);

	gcry_mac_close (hd);

	for (i=1; i < 3; i++)
	{
	if (memcmp(mac[i-1], mac[i], maclen))
	{
	fprintf (stderr, PGM ": mac mismatch with algorithm `%s'\n",
	algoname);
	exit(1);
	}
	}

	putchar ('\n');
	fflush (stdout);
	}


	static void ccm_aead_init(gcry_cipher_hd_t hd, size_t buflen, int authlen)
	{
	const int _L = 4;
	const int noncelen = 15 - _L;
	char nonce[noncelen];
	u64 params[3];
	gcry_error_t err = GPG_ERR_NO_ERROR;

	memset (nonce, 0x33, noncelen);

	err = gcry_cipher_setiv (hd, nonce, noncelen);
	if (err)
	{
	fprintf (stderr, "gcry_cipher_setiv failed: %s\n",
	gpg_strerror (err));
	gcry_cipher_close (hd);
	exit (1);
	}

	params[0] = buflen; /* encryptedlen */
	params[1] = 0; /* aadlen */
	params[2] = authlen; /* authtaglen */
	err = gcry_cipher_ctl (hd, GCRYCTL_SET_CCM_LENGTHS, params, sizeof(params));
	if (err)
	{
	fprintf (stderr, "gcry_cipher_setiv failed: %s\n",
	gpg_strerror (err));
	gcry_cipher_close (hd);
	exit (1);
	}
	}


	static gcry_error_t
	cipher_encrypt (gcry_cipher_hd_t h, char *out, size_t outsize,
	const char *in, size_t inlen, size_t max_inlen)
	{
	gcry_error_t ret;

	while (inlen)
	{
	size_t currlen = inlen;

	if (currlen > max_inlen)
	currlen = max_inlen;

	ret = gcry_cipher_encrypt(h, out, outsize, in, currlen);
	if (ret)
	return ret;

	out += currlen;
	in += currlen;
	outsize -= currlen;
	inlen -= currlen;
	}

	return 0;
	}


	static gcry_error_t
	cipher_decrypt (gcry_cipher_hd_t h, char *out, size_t outsize,
	const char *in, size_t inlen, size_t max_inlen)
	{
	gcry_error_t ret;

	while (inlen)
	{
	size_t currlen = inlen;

	if (currlen > max_inlen)
	currlen = max_inlen;

	ret = gcry_cipher_decrypt(h, out, outsize, in, currlen);
	if (ret)
	return ret;

	out += currlen;
	in += currlen;
	outsize -= currlen;
	inlen -= currlen;
	}

	return 0;
	}


	static void
	cipher_bench ( const char *algoname )
	{
	static int header_printed;
	int algo;
	gcry_cipher_hd_t hd;
	int i;
	int keylen, blklen;
	char key[128];
	char outbuf, buf;
	char raw_outbuf, raw_buf;
	size_t allocated_buflen, buflen;
	int repetitions;
	static const struct {
	int mode;
	const char *name;
	int blocked;
	unsigned int max_inlen;
	void (* const aead_init)(gcry_cipher_hd_t hd, size_t buflen, int authlen);
	int req_blocksize;
	int authlen;
	int noncelen;
	int doublekey;
	} modes[] = {
	{ GCRY_CIPHER_MODE_ECB, " ECB/Stream", 1, 0xffffffffU },
	{ GCRY_CIPHER_MODE_CBC, " CBC/Poly1305", 1, 0xffffffffU },
	{ GCRY_CIPHER_MODE_CFB, " CFB", 0, 0xffffffffU },
	{ GCRY_CIPHER_MODE_OFB, " OFB", 0, 0xffffffffU },
	{ GCRY_CIPHER_MODE_CTR, " CTR", 0, 0xffffffffU },
	{ GCRY_CIPHER_MODE_XTS, " XTS", 0, 16 << 20,
	NULL, GCRY_XTS_BLOCK_LEN, 0, 0, 1 },
	{ GCRY_CIPHER_MODE_CCM, " CCM", 0, 0xffffffffU,
	ccm_aead_init, GCRY_CCM_BLOCK_LEN, 8, },
	{ GCRY_CIPHER_MODE_GCM, " GCM", 0, 0xffffffffU,
	NULL, GCRY_GCM_BLOCK_LEN, GCRY_GCM_BLOCK_LEN },
	{ GCRY_CIPHER_MODE_OCB, " OCB", 1, 0xffffffffU,
	NULL, 16, 16, 15 },
	{ GCRY_CIPHER_MODE_EAX, " EAX", 0, 0xffffffffU,
	NULL, 0, 8, 8 },
	{ GCRY_CIPHER_MODE_STREAM, "", 0, 0xffffffffU },
	{ GCRY_CIPHER_MODE_POLY1305, "", 0, 0xffffffffU,
	NULL, 1, 16, 12 },
	{0}
	};
	int modeidx;
	gcry_error_t err = GPG_ERR_NO_ERROR;

	if (!algoname)
	{
	for (i=1; i < 400; i++)
	if ( !gcry_cipher_test_algo (i) )
	cipher_bench (gcry_cipher_algo_name (i));
	return;
	}

	if (huge_buffers)
	{
	allocated_buflen = 256 * 1024 * 1024;
	repetitions = 4;
	}
	else if (large_buffers)
	{
	allocated_buflen = 1024 * 100;
	repetitions = 10;
	}
	else
	{
	allocated_buflen = 1024;
	repetitions = 1000;
	}
	repetitions *= cipher_repetitions;

	raw_buf = gcry_xcalloc (allocated_buflen+15, 1);
	buf = (raw_buf
	+ ((16 - ((size_t)raw_buf & 0x0f)) % buffer_alignment));
	outbuf = raw_outbuf = gcry_xmalloc (allocated_buflen+15);
	outbuf = (raw_outbuf
	+ ((16 - ((size_t)raw_outbuf & 0x0f)) % buffer_alignment));

	if (!header_printed)
	{
	if (cipher_repetitions != 1)
	printf ("Running each test %d times.\n", cipher_repetitions);
	printf ("%-12s", "");
	for (modeidx=0; modes[modeidx].mode; modeidx++)
	if (*modes[modeidx].name)
	printf (" %-15s", modes[modeidx].name );
	putchar ('\n');
	printf ("%-12s", "");
	for (modeidx=0; modes[modeidx].mode; modeidx++)
	if (*modes[modeidx].name)
	printf (" ---------------" );
	putchar ('\n');
	header_printed = 1;
	}

	algo = gcry_cipher_map_name (algoname);
	if (!algo)
	{
	fprintf (stderr, PGM ": invalid cipher algorithm `%s'\n", algoname);
	exit (1);
	}

	keylen = gcry_cipher_get_algo_keylen (algo);
	if (!keylen)
	{
	fprintf (stderr, PGM ": failed to get key length for algorithm `%s'\n",
	algoname);
	exit (1);
	}
	if ( keylen * 2 > sizeof key )
	{
	fprintf (stderr, PGM ": algo %d, keylength problem (%d)\n",
	algo, keylen );
	exit (1);
	}
	for (i=0; i < keylen * 2; i++)
	key[i] = i + (clock () & 0xff);

	blklen = gcry_cipher_get_algo_blklen (algo);
	if (!blklen)
	{
	fprintf (stderr, PGM ": failed to get block length for algorithm `%s'\n",
	algoname);
	exit (1);
	}

	printf ("%-12s", gcry_cipher_algo_name (algo));
	fflush (stdout);

	for (modeidx=0; modes[modeidx].mode; modeidx++)
	{
	size_t modekeylen = keylen * (!!modes[modeidx].doublekey + 1);
	int is_stream = modes[modeidx].mode == GCRY_CIPHER_MODE_STREAM
	\|\| modes[modeidx].mode == GCRY_CIPHER_MODE_POLY1305;

	if ((blklen > 1 && is_stream) \|\| (blklen == 1 && !is_stream))
	continue;

	if (modes[modeidx].mode == GCRY_CIPHER_MODE_POLY1305
	&& algo != GCRY_CIPHER_CHACHA20)
	continue;

	/* GCM is not available in FIPS mode */
	if (in_fips_mode && modes[modeidx].mode == GCRY_CIPHER_MODE_GCM)
	continue;

	if (modes[modeidx].req_blocksize > 0
	&& blklen != modes[modeidx].req_blocksize)
	{
	printf (" %7s %7s", "-", "-" );
	continue;
	}

	for (i=0; i < sizeof buf; i++)
	buf[i] = i;

	err = gcry_cipher_open (&hd, algo, modes[modeidx].mode, 0);
	if (err)
	{
	fprintf (stderr, PGM ": error opening cipher `%s'\n", algoname);
	exit (1);
	}

	if (!cipher_with_keysetup)
	{
	err = gcry_cipher_setkey (hd, key, modekeylen);
	if (err)
	{
	fprintf (stderr, "gcry_cipher_setkey failed: %s\n",
	gpg_strerror (err));
	gcry_cipher_close (hd);
	exit (1);
	}
	}

	buflen = allocated_buflen;
	if (modes[modeidx].blocked)
	buflen = (buflen / blklen) * blklen;

	start_timer ();
	for (i=err=0; !err && i < repetitions; i++)
	{
	if (cipher_with_keysetup)
	{
	err = gcry_cipher_setkey (hd, key, modekeylen);
	if (err)
	{
	fprintf (stderr, "gcry_cipher_setkey failed: %s\n",
	gpg_strerror (err));
	gcry_cipher_close (hd);
	exit (1);
	}
	}

	if (modes[modeidx].noncelen)
	{
	char nonce[100];
	size_t noncelen;

	noncelen = modes[modeidx].noncelen;
	if (noncelen > sizeof nonce)
	noncelen = sizeof nonce;
	memset (nonce, 42, noncelen);
	err = gcry_cipher_setiv (hd, nonce, noncelen);
	if (err)
	{
	fprintf (stderr, "gcry_cipher_setiv failed: %s\n",
	gpg_strerror (err));
	gcry_cipher_close (hd);
	exit (1);
	}
	}

	if (modes[modeidx].aead_init)
	{
	(*modes[modeidx].aead_init) (hd, buflen, modes[modeidx].authlen);
	gcry_cipher_final (hd);
	err = cipher_encrypt (hd, outbuf, buflen, buf, buflen,
	modes[modeidx].max_inlen);
	if (err)
	break;
	err = gcry_cipher_gettag (hd, outbuf, modes[modeidx].authlen);
	}
	else
	{
	err = cipher_encrypt (hd, outbuf, buflen, buf, buflen,
	modes[modeidx].max_inlen);
	}
	}
	stop_timer ();

	printf (" %s", elapsed_time (1));
	fflush (stdout);
	gcry_cipher_close (hd);
	if (err)
	{
	fprintf (stderr, "gcry_cipher_encrypt failed: %s\n",
	gpg_strerror (err) );
	exit (1);
	}

	err = gcry_cipher_open (&hd, algo, modes[modeidx].mode, 0);
	if (err)
	{
	fprintf (stderr, PGM ": error opening cipher `%s'/n", algoname);
	exit (1);
	}

	if (!cipher_with_keysetup)
	{
	err = gcry_cipher_setkey (hd, key, modekeylen);
	if (err)
	{
	fprintf (stderr, "gcry_cipher_setkey failed: %s\n",
	gpg_strerror (err));
	gcry_cipher_close (hd);
	exit (1);
	}
	}

	start_timer ();
	for (i=err=0; !err && i < repetitions; i++)
	{
	if (cipher_with_keysetup)
	{
	err = gcry_cipher_setkey (hd, key, modekeylen);
	if (err)
	{
	fprintf (stderr, "gcry_cipher_setkey failed: %s\n",
	gpg_strerror (err));
	gcry_cipher_close (hd);
	exit (1);
	}
	}

	if (modes[modeidx].noncelen)
	{
	char nonce[100];
	size_t noncelen;

	noncelen = modes[modeidx].noncelen;
	if (noncelen > sizeof nonce)
	noncelen = sizeof nonce;
	memset (nonce, 42, noncelen);
	err = gcry_cipher_setiv (hd, nonce, noncelen);
	if (err)
	{
	fprintf (stderr, "gcry_cipher_setiv failed: %s\n",
	gpg_strerror (err));
	gcry_cipher_close (hd);
	exit (1);
	}
	}

	if (modes[modeidx].aead_init)
	{
	(*modes[modeidx].aead_init) (hd, buflen, modes[modeidx].authlen);
	gcry_cipher_final (hd);
	err = cipher_decrypt (hd, outbuf, buflen, buf, buflen,
	modes[modeidx].max_inlen);
	if (err)
	break;
	err = gcry_cipher_checktag (hd, outbuf, modes[modeidx].authlen);
	if (gpg_err_code (err) == GPG_ERR_CHECKSUM)
	err = 0;
	}
	else
	{
	gcry_cipher_final (hd);
	err = cipher_decrypt (hd, outbuf, buflen, buf, buflen,
	modes[modeidx].max_inlen);
	}
	}
	stop_timer ();
	printf (" %s", elapsed_time (1));
	fflush (stdout);
	gcry_cipher_close (hd);
	if (err)
	{
	fprintf (stderr, "gcry_cipher_decrypt failed: %s\n",
	gpg_strerror (err) );
	exit (1);
	}
	}

	putchar ('\n');
	gcry_free (raw_buf);
	gcry_free (raw_outbuf);
	}



	static void
	rsa_bench (int iterations, int print_header, int no_blinding)
	{
	gpg_error_t err;
	int p_sizes[] = { 1024, 2048, 3072, 4096 };
	int testno;

	if (print_header)
	printf ("Algorithm generate %4dpriv %4dpublic\n"
	"------------------------------------------------\n",
	iterations, iterations );
	for (testno=0; testno < DIM (p_sizes); testno++)
	{
	gcry_sexp_t key_spec, key_pair, pub_key, sec_key;
	gcry_mpi_t x;
	gcry_sexp_t data;
	gcry_sexp_t sig = NULL;
	int count;
	unsigned nbits = p_sizes[testno];

	printf ("RSA %3d bit ", nbits);
	fflush (stdout);

	if (in_fips_mode && nbits < 2048)
	{
	puts ("[skipped in fips mode]");
	continue;
	}

	err = gcry_sexp_build (&key_spec, NULL,
	gcry_fips_mode_active ()
	? "(genkey (RSA (nbits %d)))"
	: "(genkey (RSA (nbits %d)(transient-key)))",
	nbits);
	if (err)
	die ("creating S-expression failed: %s\n", gcry_strerror (err));

	start_timer ();
	err = gcry_pk_genkey (&key_pair, key_spec);
	if (err)
	die ("creating %d bit RSA key failed: %s\n",
	nbits, gcry_strerror (err));

	pub_key = gcry_sexp_find_token (key_pair, "public-key", 0);
	if (! pub_key)
	die ("public part missing in key\n");
	sec_key = gcry_sexp_find_token (key_pair, "private-key", 0);
	if (! sec_key)
	die ("private part missing in key\n");
	gcry_sexp_release (key_pair);
	gcry_sexp_release (key_spec);

	stop_timer ();
	printf (" %s", elapsed_time (1));
	fflush (stdout);

	x = gcry_mpi_new (nbits);
	gcry_mpi_randomize (x, nbits-8, GCRY_WEAK_RANDOM);
	err = gcry_sexp_build (&data, NULL,
	"(data (flags raw) (value %m))", x);
	gcry_mpi_release (x);
	if (err)
	die ("converting data failed: %s\n", gcry_strerror (err));

	start_timer ();
	for (count=0; count < iterations; count++)
	{
	gcry_sexp_release (sig);
	err = gcry_pk_sign (&sig, data, sec_key);
	if (err)
	die ("signing failed (%d): %s\n", count, gpg_strerror (err));
	}
	stop_timer ();
	printf (" %s", elapsed_time (1));
	fflush (stdout);

	start_timer ();
	for (count=0; count < iterations; count++)
	{
	err = gcry_pk_verify (sig, data, pub_key);
	if (err)
	{
	putchar ('\n');
	show_sexp ("seckey:\n", sec_key);
	show_sexp ("data:\n", data);
	show_sexp ("sig:\n", sig);
	die ("verify failed (%d): %s\n", count, gpg_strerror (err));
	}
	}
	stop_timer ();
	printf (" %s", elapsed_time (1));

	if (no_blinding)
	{
	fflush (stdout);
	x = gcry_mpi_new (nbits);
	gcry_mpi_randomize (x, nbits-8, GCRY_WEAK_RANDOM);
	err = gcry_sexp_build (&data, NULL,
	"(data (flags no-blinding) (value %m))", x);
	gcry_mpi_release (x);
	if (err)
	die ("converting data failed: %s\n", gcry_strerror (err));

	start_timer ();
	for (count=0; count < iterations; count++)
	{
	gcry_sexp_release (sig);
	err = gcry_pk_sign (&sig, data, sec_key);
	if (err)
	die ("signing failed (%d): %s\n", count, gpg_strerror (err));
	}
	stop_timer ();
	printf (" %s", elapsed_time (1));
	fflush (stdout);
	}

	putchar ('\n');
	fflush (stdout);

	gcry_sexp_release (sig);
	gcry_sexp_release (data);
	gcry_sexp_release (sec_key);
	gcry_sexp_release (pub_key);
	}
	}


	static void
	elg_bench (int iterations, int print_header)
	{
	gpg_error_t err;
	gcry_sexp_t pub_key[3], sec_key[3];
	int p_sizes[3] = { 1024, 2048, 3072 };
	gcry_sexp_t data = NULL;
	gcry_sexp_t enc = NULL;
	gcry_sexp_t plain = NULL;
	int i, j;

	err = gcry_sexp_sscan (pub_key+0, NULL, sample_public_elg_key_1024,
	strlen (sample_public_elg_key_1024));
	if (!err)
	err = gcry_sexp_sscan (sec_key+0, NULL, sample_private_elg_key_1024,
	strlen (sample_private_elg_key_1024));
	if (!err)
	err = gcry_sexp_sscan (pub_key+1, NULL, sample_public_elg_key_2048,
	strlen (sample_public_elg_key_2048));
	if (!err)
	err = gcry_sexp_sscan (sec_key+1, NULL, sample_private_elg_key_2048,
	strlen (sample_private_elg_key_2048));
	if (!err)
	err = gcry_sexp_sscan (pub_key+2, NULL, sample_public_elg_key_3072,
	strlen (sample_public_elg_key_3072));
	if (!err)
	err = gcry_sexp_sscan (sec_key+2, NULL, sample_private_elg_key_3072,
	strlen (sample_private_elg_key_3072));
	if (err)
	{
	fprintf (stderr, PGM ": converting sample keys failed: %s\n",
	gcry_strerror (err));
	exit (1);
	}

	if (print_header)
	printf ("Algorithm generate %4dpriv %4dpublic\n"
	"------------------------------------------------\n",
	iterations, iterations );
	for (i=0; i < DIM (p_sizes); i++)
	{
	char timerbuf1[100];

	{
	gcry_mpi_t x = gcry_mpi_new (p_sizes[i]);
	gcry_mpi_randomize (x, p_sizes[i] - 16, GCRY_WEAK_RANDOM);
	err = gcry_sexp_build (&data, NULL, "(data (flags raw) (value %m))", x);
	gcry_mpi_release (x);
	}
	if (err)
	{
	fprintf (stderr, PGM ": converting data failed: %s\n",
	gcry_strerror (err));
	exit (1);
	}

	printf ("ELG %d bit -", p_sizes[i]);
	fflush (stdout);

	start_timer ();
	for (j=0; j < iterations; j++)
	{
	gcry_sexp_release (enc);
	err = gcry_pk_encrypt (&enc, data, pub_key[i]);
	if (err)
	{
	putchar ('\n');
	fprintf (stderr, PGM ": encrypt failed: %s\n",
	gpg_strerror (err));
	exit (1);
	}
	}
	stop_timer ();
	snprintf (timerbuf1, sizeof timerbuf1, " %s", elapsed_time (1));
	fflush (stdout);

	start_timer ();
	for (j=0; j < iterations; j++)
	{
	gcry_sexp_release (plain);
	err = gcry_pk_decrypt (&plain, enc, sec_key[i]);
	if (err)
	{
	putchar ('\n');
	fprintf (stderr, PGM ": decrypt failed: %s\n",
	gpg_strerror (err));
	exit (1);
	}
	}
	stop_timer ();

	printf (" %s %s\n", elapsed_time (1), timerbuf1);
	fflush (stdout);

	gcry_sexp_release (plain);
	plain = NULL;
	gcry_sexp_release (enc);
	enc = NULL;
	gcry_sexp_release (data);
	data = NULL;
	}

	for (i=0; i < DIM (p_sizes); i++)
	{
	gcry_sexp_release (sec_key[i]);
	gcry_sexp_release (pub_key[i]);
	}
	}


	static void
	dsa_bench (int iterations, int print_header)
	{
	gpg_error_t err;
	gcry_sexp_t pub_key[3], sec_key[3];
	int p_sizes[3] = { 1024, 2048, 3072 };
	int q_sizes[3] = { 160, 224, 256 };
	gcry_sexp_t data;
	gcry_sexp_t sig = NULL;
	int i, j;

	err = gcry_sexp_sscan (pub_key+0, NULL, sample_public_dsa_key_1024,
	strlen (sample_public_dsa_key_1024));
	if (!err)
	err = gcry_sexp_sscan (sec_key+0, NULL, sample_private_dsa_key_1024,
	strlen (sample_private_dsa_key_1024));
	if (!err)
	err = gcry_sexp_sscan (pub_key+1, NULL, sample_public_dsa_key_2048,
	strlen (sample_public_dsa_key_2048));
	if (!err)
	err = gcry_sexp_sscan (sec_key+1, NULL, sample_private_dsa_key_2048,
	strlen (sample_private_dsa_key_2048));
	if (!err)
	err = gcry_sexp_sscan (pub_key+2, NULL, sample_public_dsa_key_3072,
	strlen (sample_public_dsa_key_3072));
	if (!err)
	err = gcry_sexp_sscan (sec_key+2, NULL, sample_private_dsa_key_3072,
	strlen (sample_private_dsa_key_3072));
	if (err)
	{
	fprintf (stderr, PGM ": converting sample keys failed: %s\n",
	gcry_strerror (err));
	exit (1);
	}

	if (print_header)
	printf ("Algorithm generate %4dpriv %4dpublic\n"
	"------------------------------------------------\n",
	iterations, iterations );
	for (i=0; i < DIM (q_sizes); i++)
	{
	gcry_mpi_t x;

	x = gcry_mpi_new (q_sizes[i]);
	gcry_mpi_randomize (x, q_sizes[i], GCRY_WEAK_RANDOM);
	err = gcry_sexp_build (&data, NULL, "(data (flags raw) (value %m))", x);
	gcry_mpi_release (x);
	if (err)
	{
	fprintf (stderr, PGM ": converting data failed: %s\n",
	gcry_strerror (err));
	exit (1);
	}

	printf ("DSA %d/%d -", p_sizes[i], q_sizes[i]);
	fflush (stdout);

	if (in_fips_mode && !(p_sizes[i] == 2048 \|\| p_sizes[i] == 3072))
	{
	puts ("[skipped in fips mode]");
	goto next;
	}

	start_timer ();
	for (j=0; j < iterations; j++)
	{
	gcry_sexp_release (sig);
	err = gcry_pk_sign (&sig, data, sec_key[i]);
	if (err)
	{
	putchar ('\n');
	fprintf (stderr, PGM ": signing failed: %s\n",
	gpg_strerror (err));
	exit (1);
	}
	}
	stop_timer ();
	printf (" %s", elapsed_time (1));
	fflush (stdout);

	start_timer ();
	for (j=0; j < iterations; j++)
	{
	err = gcry_pk_verify (sig, data, pub_key[i]);
	if (err)
	{
	putchar ('\n');
	fprintf (stderr, PGM ": verify failed: %s\n",
	gpg_strerror (err));
	exit (1);
	}
	}
	stop_timer ();
	printf (" %s\n", elapsed_time (1));
	fflush (stdout);

	next:
	gcry_sexp_release (sig);
	gcry_sexp_release (data);
	sig = NULL;
	}


	for (i=0; i < DIM (q_sizes); i++)
	{
	gcry_sexp_release (sec_key[i]);
	gcry_sexp_release (pub_key[i]);
	}
	}


	static void
	ecc_bench (int iterations, int print_header)
	{
	#if USE_ECC
	gpg_error_t err;
	const char *p_sizes[] = { "192", "224", "256", "384", "521", "Ed25519", "Ed448",
	"gost256", "gost512" };
	int testno;

	if (print_header)
	printf ("Algorithm generate %4dpriv %4dpublic\n"
	"------------------------------------------------\n",
	iterations, iterations );
	for (testno=0; testno < DIM (p_sizes); testno++)
	{
	gcry_sexp_t key_spec, key_pair, pub_key, sec_key;
	gcry_mpi_t x;
	gcry_sexp_t data;
	gcry_sexp_t sig = NULL;
	int count;
	int p_size;
	int is_ed25519;
	int is_ed448;
	int is_gost;

	is_ed25519 = !strcmp (p_sizes[testno], "Ed25519");
	is_ed448 = !strcmp (p_sizes[testno], "Ed448");
	is_gost = !strncmp (p_sizes[testno], "gost", 4);

	/* Only P-{224,256,384,521} are allowed in fips mode */
	if (gcry_fips_mode_active()
	&& (is_ed25519 \|\| is_ed448 \|\| is_gost
	\|\| !strcmp (p_sizes[testno], "192")))
	continue;

	if (is_ed25519)
	{
	p_size = 256;
	printf ("EdDSA Ed25519 ");
	fflush (stdout);
	}
	else if (is_ed448)
	{
	p_size = 448;
	printf ("EdDSA Ed448 ");
	fflush (stdout);
	}
	else if (is_gost)
	{
	p_size = atoi (p_sizes[testno] + 4);
	printf ("GOST %3d bit ", p_size);
	fflush (stdout);
	}
	else
	{
	p_size = atoi (p_sizes[testno]);
	printf ("ECDSA %3d bit ", p_size);
	}
	fflush (stdout);

	if (is_ed25519)
	err = gcry_sexp_build (&key_spec, NULL,
	"(genkey (ecdsa (curve \"Ed25519\")"
	"(flags eddsa)))");
	else if (is_ed448)
	err = gcry_sexp_build (&key_spec, NULL,
	"(genkey (ecdsa (curve \"Ed448\")"
	"(flags eddsa)))");
	else if (is_gost)
	err = gcry_sexp_build (&key_spec, NULL,
	"(genkey (ecdsa (curve %s)))",
	p_size == 256 ? "GOST2001-test" : "GOST2012-512-test");
	else
	err = gcry_sexp_build (&key_spec, NULL,
	"(genkey (ECDSA (nbits %d)))", p_size);
	if (err)
	die ("creating S-expression failed: %s\n", gcry_strerror (err));

	start_timer ();
	err = gcry_pk_genkey (&key_pair, key_spec);
	if (err)
	die ("creating %d bit ECC key failed: %s\n",
	p_size, gcry_strerror (err));
	if (verbose > 2)
	show_sexp ("ECC key:\n", key_pair);

	pub_key = gcry_sexp_find_token (key_pair, "public-key", 0);
	if (! pub_key)
	die ("public part missing in key\n");
	sec_key = gcry_sexp_find_token (key_pair, "private-key", 0);
	if (! sec_key)
	die ("private part missing in key\n");
	gcry_sexp_release (key_pair);
	gcry_sexp_release (key_spec);

	stop_timer ();
	printf (" %s", elapsed_time (1));
	fflush (stdout);

	x = gcry_mpi_new (p_size);
	gcry_mpi_randomize (x, p_size, GCRY_WEAK_RANDOM);
	if (is_ed25519)
	err = gcry_sexp_build (&data, NULL,
	"(data (flags eddsa)(hash-algo sha512)"
	" (value %m))", x);
	else if (is_ed448)
	err = gcry_sexp_build (&data, NULL,
	"(data (flags eddsa)(hash-algo shake256)"
	" (value %m))", x);
	else if (is_gost)
	err = gcry_sexp_build (&data, NULL, "(data (flags gost) (value %m))", x);
	else
	err = gcry_sexp_build (&data, NULL, "(data (flags raw) (value %m))", x);
	gcry_mpi_release (x);

	if (err)
	die ("converting data failed: %s\n", gcry_strerror (err));

	start_timer ();
	for (count=0; count < iterations; count++)
	{
	gcry_sexp_release (sig);
	err = gcry_pk_sign (&sig, data, sec_key);
	if (err)
	{
	if (verbose)
	{
	putc ('\n', stderr);
	show_sexp ("signing key:\n", sec_key);
	show_sexp ("signed data:\n", data);
	}
	die ("signing failed: %s\n", gpg_strerror (err));
	}
	}
	stop_timer ();
	printf (" %s", elapsed_time (1));
	fflush (stdout);

	start_timer ();
	for (count=0; count < iterations; count++)
	{
	err = gcry_pk_verify (sig, data, pub_key);
	if (err)
	{
	putchar ('\n');
	show_sexp ("seckey:\n", sec_key);
	show_sexp ("data:\n", data);
	show_sexp ("sig:\n", sig);
	die ("verify failed: %s\n", gpg_strerror (err));
	}
	}
	stop_timer ();
	printf (" %s\n", elapsed_time (1));
	fflush (stdout);

	gcry_sexp_release (sig);
	gcry_sexp_release (data);
	gcry_sexp_release (sec_key);
	gcry_sexp_release (pub_key);
	}
	#endif /USE_ECC/
	}



	static void
	do_powm ( const char n_str, const char e_str, const char *m_str)
	{
	gcry_mpi_t e, n, msg, cip;
	gcry_error_t err;
	int i;

	err = gcry_mpi_scan (&n, GCRYMPI_FMT_HEX, n_str, 0, 0);
	if (err) BUG ();
	err = gcry_mpi_scan (&e, GCRYMPI_FMT_HEX, e_str, 0, 0);
	if (err) BUG ();
	err = gcry_mpi_scan (&msg, GCRYMPI_FMT_HEX, m_str, 0, 0);
	if (err) BUG ();

	cip = gcry_mpi_new (0);

	start_timer ();
	for (i=0; i < 1000; i++)
	gcry_mpi_powm (cip, msg, e, n);
	stop_timer ();
	printf (" %s", elapsed_time (1)); fflush (stdout);
	/* { */
	/* char buf; /

	/* if (gcry_mpi_aprint (GCRYMPI_FMT_HEX, (void*)&buf, NULL, cip)) /
	/* BUG (); */
	/* printf ("result: %s\n", buf); */
	/* gcry_free (buf); */
	/* } */
	gcry_mpi_release (cip);
	gcry_mpi_release (msg);
	gcry_mpi_release (n);
	gcry_mpi_release (e);
	}


	static void
	mpi_bench (void)
	{
	printf ("%-10s", "powm"); fflush (stdout);

	do_powm (
	"20A94417D4D5EF2B2DA99165C7DC87DADB3979B72961AF90D09D59BA24CB9A10166FDCCC9C659F2B9626EC23F3FA425F564A072BA941B03FA81767CC289E4",
	"29",
	"B870187A323F1ECD5B8A0B4249507335A1C4CE8394F38FD76B08C78A42C58F6EA136ACF90DFE8603697B1694A3D81114D6117AC1811979C51C4DD013D52F8"
	);
	do_powm (
	"20A94417D4D5EF2B2DA99165C7DC87DADB3979B72961AF90D09D59BA24CB9A10166FDCCC9C659F2B9626EC23F3FA425F564A072BA941B03FA81767CC289E41071F0246879A442658FBD18C1771571E7073EEEB2160BA0CBFB3404D627069A6CFBD53867AD2D9D40231648000787B5C84176B4336144644AE71A403CA40716",
	"29",
	"B870187A323F1ECD5B8A0B4249507335A1C4CE8394F38FD76B08C78A42C58F6EA136ACF90DFE8603697B1694A3D81114D6117AC1811979C51C4DD013D52F8FC4EE4BB446B83E48ABED7DB81CBF5E81DE4759E8D68AC985846D999F96B0D8A80E5C69D272C766AB8A23B40D50A4FA889FBC2BD2624222D8EB297F4BAEF8593847"
	);
	do_powm (
	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
	"29",
	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
	);

	putchar ('\n');


	}


	static void
	prime_bench (void)
	{
	gpg_error_t err;
	int i;
	gcry_mpi_t prime;
	int old_prog = single_char_progress;

	single_char_progress = 1;
	if (!with_progress)
	printf ("%-10s", "prime");
	fflush (stdout);
	start_timer ();
	for (i=0; i < 10; i++)
	{
	if (with_progress)
	fputs ("primegen ", stdout);
	err = gcry_prime_generate (&prime,
	1024, 0,
	NULL,
	NULL, NULL,
	GCRY_WEAK_RANDOM,
	GCRY_PRIME_FLAG_SECRET);
	if (with_progress)
	{
	fputc ('\n', stdout);
	fflush (stdout);
	}
	if (err)
	{
	fprintf (stderr, PGM ": error creating prime: %s\n",
	gpg_strerror (err));
	exit (1);
	}
	gcry_mpi_release (prime);
	}
	stop_timer ();
	if (with_progress)
	printf ("%-10s", "prime");
	printf (" %s\n", elapsed_time (1)); fflush (stdout);

	single_char_progress = old_prog;
	}


	int
	main( int argc, char **argv )
	{
	int last_argc = -1;
	int no_blinding = 0;
	- int use_random_daemon = 0;
	int use_secmem = 0;
	int pk_count = 100;

	buffer_alignment = 1;

	if (argc)
	{ argc--; argv++; }

	/* We skip this test if we are running under the test suite (no args
	and srcdir defined) and GCRYPT_NO_BENCHMARKS is set. */
	if (!argc && getenv ("srcdir") && getenv ("GCRYPT_NO_BENCHMARKS"))
	exit (77);

	if (getenv ("GCRYPT_IN_REGRESSION_TEST"))
	{
	in_regression_test = 1;
	pk_count = 10;
	}

	while (argc && last_argc != argc )
	{
	last_argc = argc;
	if (!strcmp (*argv, "--"))
	{
	argc--; argv++;
	break;
	}
	else if (!strcmp (*argv, "--help"))
	{
	fputs ("usage: benchmark "
	"[md\|mac\|cipher\|random\|mpi\|rsa\|dsa\|ecc\|prime [algonames]]\n",
	stdout);
	exit (0);
	}
	else if (!strcmp (*argv, "--verbose"))
	{
	verbose++;
	argc--; argv++;
	}
	else if (!strcmp (*argv, "--debug"))
	{
	verbose += 2;
	debug++;
	argc--; argv++;
	}
	- else if (!strcmp (*argv, "--use-random-daemon"))
	- {
	- use_random_daemon = 1;
	- argc--; argv++;
	- }
	else if (!strcmp (*argv, "--use-secmem"))
	{
	use_secmem = 1;
	argc--; argv++;
	}
	else if (!strcmp (*argv, "--prefer-standard-rng"))
	{
	/* This is anyway the default, but we may want to use it for
	debugging. */
	xgcry_control ((GCRYCTL_SET_PREFERRED_RNG_TYPE, GCRY_RNG_TYPE_STANDARD));
	argc--; argv++;
	}
	else if (!strcmp (*argv, "--prefer-fips-rng"))
	{
	xgcry_control ((GCRYCTL_SET_PREFERRED_RNG_TYPE, GCRY_RNG_TYPE_FIPS));
	argc--; argv++;
	}
	else if (!strcmp (*argv, "--prefer-system-rng"))
	{
	xgcry_control ((GCRYCTL_SET_PREFERRED_RNG_TYPE, GCRY_RNG_TYPE_SYSTEM));
	argc--; argv++;
	}
	else if (!strcmp (*argv, "--no-blinding"))
	{
	no_blinding = 1;
	argc--; argv++;
	}
	else if (!strcmp (*argv, "--large-buffers"))
	{
	large_buffers = 1;
	argc--; argv++;
	}
	else if (!strcmp (*argv, "--huge-buffers"))
	{
	huge_buffers = 1;
	argc--; argv++;
	}
	else if (!strcmp (*argv, "--cipher-repetitions"))
	{
	argc--; argv++;
	if (argc)
	{
	cipher_repetitions = atoi(*argv);
	argc--; argv++;
	}
	}
	else if (!strcmp (*argv, "--cipher-with-keysetup"))
	{
	cipher_with_keysetup = 1;
	argc--; argv++;
	}
	else if (!strcmp (*argv, "--hash-repetitions"))
	{
	argc--; argv++;
	if (argc)
	{
	hash_repetitions = atoi(*argv);
	argc--; argv++;
	}
	}
	else if (!strcmp (*argv, "--mac-repetitions"))
	{
	argc--; argv++;
	if (argc)
	{
	mac_repetitions = atoi(*argv);
	argc--; argv++;
	}
	}
	else if (!strcmp (*argv, "--pk-count"))
	{
	argc--; argv++;
	if (argc)
	{
	pk_count = atoi(*argv);
	argc--; argv++;
	}
	}
	else if (!strcmp (*argv, "--alignment"))
	{
	argc--; argv++;
	if (argc)
	{
	buffer_alignment = atoi(*argv);
	argc--; argv++;
	}
	}
	else if (!strcmp (*argv, "--disable-hwf"))
	{
	argc--; argv++;
	if (argc)
	{
	if (gcry_control (GCRYCTL_DISABLE_HWF, *argv, NULL))
	fprintf (stderr, PGM ": unknown hardware feature `%s'"
	" - option ignored\n", *argv);
	argc--; argv++;
	}
	}
	else if (!strcmp (*argv, "--fips"))
	{
	argc--; argv++;
	/* This command needs to be called before gcry_check_version. */
	xgcry_control ((GCRYCTL_FORCE_FIPS_MODE, 0));
	}
	else if (!strcmp (*argv, "--progress"))
	{
	argc--; argv++;
	with_progress = 1;
	}
	}

	if (buffer_alignment < 1 \|\| buffer_alignment > 16)
	die ("value for --alignment must be in the range 1 to 16\n");

	xgcry_control ((GCRYCTL_SET_VERBOSITY, (int)verbose));

	if (!gcry_check_version (GCRYPT_VERSION))
	{
	fprintf (stderr, PGM ": version mismatch; pgm=%s, library=%s\n",
	GCRYPT_VERSION, gcry_check_version (NULL));
	exit (1);
	}

	if (debug)
	xgcry_control ((GCRYCTL_SET_DEBUG_FLAGS, 1u , 0));

	if (gcry_fips_mode_active ())
	in_fips_mode = 1;
	else if (!use_secmem)
	xgcry_control ((GCRYCTL_DISABLE_SECMEM, 0));

	- if (use_random_daemon)
	- xgcry_control ((GCRYCTL_USE_RANDOM_DAEMON, 1));
	-
	if (with_progress)
	gcry_set_progress_handler (progress_cb, NULL);

	xgcry_control ((GCRYCTL_INITIALIZATION_FINISHED, 0));

	if (cipher_repetitions < 1)
	cipher_repetitions = 1;
	if (hash_repetitions < 1)
	hash_repetitions = 1;
	if (mac_repetitions < 1)
	mac_repetitions = 1;

	if (in_regression_test)
	fputs ("Note: " PGM " running in quick regression test mode.\n", stdout);

	if ( !argc )
	{
	xgcry_control ((GCRYCTL_ENABLE_QUICK_RANDOM, 0));
	md_bench (NULL);
	putchar ('\n');
	mac_bench (NULL);
	putchar ('\n');
	cipher_bench (NULL);
	putchar ('\n');
	rsa_bench (pk_count, 1, no_blinding);
	elg_bench (pk_count, 0);
	dsa_bench (pk_count, 0);
	ecc_bench (pk_count, 0);
	putchar ('\n');
	mpi_bench ();
	putchar ('\n');
	random_bench (0);
	}
	else if ( !strcmp (argv, "random") \|\| !strcmp (argv, "strongrandom"))
	{
	if (argc == 1)
	random_bench ((**argv == 's'));
	else if (argc == 2)
	{
	xgcry_control ((GCRYCTL_SET_RANDOM_SEED_FILE, argv[1]));
	random_bench ((**argv == 's'));
	xgcry_control ((GCRYCTL_UPDATE_RANDOM_SEED_FILE));
	}
	else
	fputs ("usage: benchmark [strong]random [seedfile]\n", stdout);
	}
	else if ( !strcmp (*argv, "md"))
	{
	if (argc == 1)
	md_bench (NULL);
	else
	for (argc--, argv++; argc; argc--, argv++)
	md_bench ( *argv );
	}
	else if ( !strcmp (*argv, "mac"))
	{
	if (argc == 1)
	mac_bench (NULL);
	else
	for (argc--, argv++; argc; argc--, argv++)
	mac_bench ( *argv );
	}
	else if ( !strcmp (*argv, "cipher"))
	{
	if (argc == 1)
	cipher_bench (NULL);
	else
	for (argc--, argv++; argc; argc--, argv++)
	cipher_bench ( *argv );
	}
	else if ( !strcmp (*argv, "mpi"))
	{
	mpi_bench ();
	}
	else if ( !strcmp (*argv, "pubkey"))
	{
	xgcry_control ((GCRYCTL_ENABLE_QUICK_RANDOM, 0));
	rsa_bench (pk_count, 1, no_blinding);
	elg_bench (pk_count, 0);
	dsa_bench (pk_count, 0);
	ecc_bench (pk_count, 0);
	}
	else if ( !strcmp (*argv, "rsa"))
	{
	xgcry_control ((GCRYCTL_ENABLE_QUICK_RANDOM, 0));
	rsa_bench (pk_count, 1, no_blinding);
	}
	else if ( !strcmp (*argv, "elg"))
	{
	xgcry_control ((GCRYCTL_ENABLE_QUICK_RANDOM, 0));
	elg_bench (pk_count, 1);
	}
	else if ( !strcmp (*argv, "dsa"))
	{
	xgcry_control ((GCRYCTL_ENABLE_QUICK_RANDOM, 0));
	dsa_bench (pk_count, 1);
	}
	else if ( !strcmp (*argv, "ecc"))
	{
	xgcry_control ((GCRYCTL_ENABLE_QUICK_RANDOM, 0));
	ecc_bench (pk_count, 1);
	}
	else if ( !strcmp (*argv, "prime"))
	{
	xgcry_control ((GCRYCTL_ENABLE_QUICK_RANDOM, 0));
	prime_bench ();
	}
	else
	{
	fprintf (stderr, PGM ": bad arguments\n");
	return 1;
	}


	if (in_fips_mode && !gcry_fips_mode_active ())
	fprintf (stderr, PGM ": FIPS mode is not anymore active\n");

	return 0;
	}

File Metadata

Mime Type: text/x-diff
Expires: Mon, Dec 23, 1:58 PM (1 d, 7 h)
Storage Engine: local-disk
Storage Format: Raw Data
Storage Handle: 0c/53/436b6ebc04bd3d45f365e4a9f8f7

No OneTemporaryActions

View Options

File Metadata

Event Timeline

No OneTemporary
Actions