vendor libsecp256k1 so it doesn't depend on a shared library.

This commit is contained in:
fiatjaf
2024-12-26 17:05:17 -03:00
parent 95ddacb9f3
commit 46645ad4d6
74 changed files with 36414 additions and 4 deletions

23
justfile Normal file
View File

@@ -0,0 +1,23 @@
list:
@just --list
vendor-libsecp256k1:
#!/usr/bin/env fish
rm -r libsecp256k1
mkdir libsecp256k1
mkdir libsecp256k1/include
mkdir libsecp256k1/src
mkdir libsecp256k1/src/asm
mkdir libsecp256k1/src/modules
mkdir libsecp256k1/src/modules/extrakeys
mkdir libsecp256k1/src/modules/schnorrsig
wget https://api.github.com/repos/bitcoin-core/secp256k1/tarball/v0.6.0 -O libsecp256k1.tar.gz
tar -xvf libsecp256k1.tar.gz
rm libsecp256k1.tar.gz
cd bitcoin-core-secp256k1-*
for f in include/secp256k1.h include/secp256k1_ecdh.h include/secp256k1_ellswift.h include/secp256k1_extrakeys.h include/secp256k1_preallocated.h include/secp256k1_recovery.h include/secp256k1_schnorrsig.h src/asm/field_10x26_arm.s src/assumptions.h src/bench.c src/bench.h src/bench_ecmult.c src/bench_internal.c src/checkmem.h src/ecdsa.h src/ecdsa_impl.h src/eckey.h src/eckey_impl.h src/ecmult.h src/ecmult_compute_table.h src/ecmult_compute_table_impl.h src/ecmult_const.h src/ecmult_const_impl.h src/ecmult_gen.h src/ecmult_gen_compute_table.h src/ecmult_gen_compute_table_impl.h src/ecmult_gen_impl.h src/ecmult_impl.h src/field.h src/field_10x26.h src/field_10x26_impl.h src/field_5x52.h src/field_5x52_impl.h src/field_5x52_int128_impl.h src/field_impl.h src/group.h src/group_impl.h src/hash.h src/hash_impl.h src/hsort.h src/hsort_impl.h src/int128.h src/int128_impl.h src/int128_native.h src/int128_native_impl.h src/int128_struct.h src/int128_struct_impl.h src/modinv32.h src/modinv32_impl.h src/modinv64.h src/modinv64_impl.h src/modules/extrakeys/main_impl.h src/modules/schnorrsig/main_impl.h src/precompute_ecmult.c src/precompute_ecmult_gen.c src/precomputed_ecmult.c src/precomputed_ecmult.h src/precomputed_ecmult_gen.c src/precomputed_ecmult_gen.h src/scalar.h src/scalar_4x64.h src/scalar_4x64_impl.h src/scalar_8x32.h src/scalar_8x32_impl.h src/scalar_impl.h src/scalar_low.h src/scalar_low_impl.h src/scratch.h src/scratch_impl.h src/secp256k1.c src/selftest.h src/util.h
mv $f ../libsecp256k1/$f
end
cd ..
rm -r bitcoin-core-secp256k1-*

View File

@@ -0,0 +1,899 @@
#ifndef SECP256K1_H
#define SECP256K1_H
#ifdef __cplusplus
extern "C" {
#endif
#include <stddef.h>
/** Unless explicitly stated all pointer arguments must not be NULL.
*
* The following rules specify the order of arguments in API calls:
*
* 1. Context pointers go first, followed by output arguments, combined
* output/input arguments, and finally input-only arguments.
* 2. Array lengths always immediately follow the argument whose length
* they describe, even if this violates rule 1.
* 3. Within the OUT/OUTIN/IN groups, pointers to data that is typically generated
* later go first. This means: signatures, public nonces, secret nonces,
* messages, public keys, secret keys, tweaks.
* 4. Arguments that are not data pointers go last, from more complex to less
* complex: function pointers, algorithm names, messages, void pointers,
* counts, flags, booleans.
* 5. Opaque data pointers follow the function pointer they are to be passed to.
*/
/** Opaque data structure that holds context information
*
* The primary purpose of context objects is to store randomization data for
* enhanced protection against side-channel leakage. This protection is only
* effective if the context is randomized after its creation. See
* secp256k1_context_create for creation of contexts and
* secp256k1_context_randomize for randomization.
*
* A secondary purpose of context objects is to store pointers to callback
* functions that the library will call when certain error states arise. See
* secp256k1_context_set_error_callback as well as
* secp256k1_context_set_illegal_callback for details. Future library versions
* may use context objects for additional purposes.
*
* A constructed context can safely be used from multiple threads
* simultaneously, but API calls that take a non-const pointer to a context
* need exclusive access to it. In particular this is the case for
* secp256k1_context_destroy, secp256k1_context_preallocated_destroy,
* and secp256k1_context_randomize.
*
* Regarding randomization, either do it once at creation time (in which case
* you do not need any locking for the other calls), or use a read-write lock.
*/
typedef struct secp256k1_context_struct secp256k1_context;
/** Opaque data structure that holds a parsed and valid public key.
*
* The exact representation of data inside is implementation defined and not
* guaranteed to be portable between different platforms or versions. It is
* however guaranteed to be 64 bytes in size, and can be safely copied/moved.
* If you need to convert to a format suitable for storage or transmission,
* use secp256k1_ec_pubkey_serialize and secp256k1_ec_pubkey_parse. To
* compare keys, use secp256k1_ec_pubkey_cmp.
*/
typedef struct secp256k1_pubkey {
unsigned char data[64];
} secp256k1_pubkey;
/** Opaque data structure that holds a parsed ECDSA signature.
*
* The exact representation of data inside is implementation defined and not
* guaranteed to be portable between different platforms or versions. It is
* however guaranteed to be 64 bytes in size, and can be safely copied/moved.
* If you need to convert to a format suitable for storage, transmission, or
* comparison, use the secp256k1_ecdsa_signature_serialize_* and
* secp256k1_ecdsa_signature_parse_* functions.
*/
typedef struct secp256k1_ecdsa_signature {
unsigned char data[64];
} secp256k1_ecdsa_signature;
/** A pointer to a function to deterministically generate a nonce.
*
* Returns: 1 if a nonce was successfully generated. 0 will cause signing to fail.
* Out: nonce32: pointer to a 32-byte array to be filled by the function.
* In: msg32: the 32-byte message hash being verified (will not be NULL)
* key32: pointer to a 32-byte secret key (will not be NULL)
* algo16: pointer to a 16-byte array describing the signature
* algorithm (will be NULL for ECDSA for compatibility).
* data: Arbitrary data pointer that is passed through.
* attempt: how many iterations we have tried to find a nonce.
* This will almost always be 0, but different attempt values
* are required to result in a different nonce.
*
* Except for test cases, this function should compute some cryptographic hash of
* the message, the algorithm, the key and the attempt.
*/
typedef int (*secp256k1_nonce_function)(
unsigned char *nonce32,
const unsigned char *msg32,
const unsigned char *key32,
const unsigned char *algo16,
void *data,
unsigned int attempt
);
# if !defined(SECP256K1_GNUC_PREREQ)
# if defined(__GNUC__)&&defined(__GNUC_MINOR__)
# define SECP256K1_GNUC_PREREQ(_maj,_min) \
((__GNUC__<<16)+__GNUC_MINOR__>=((_maj)<<16)+(_min))
# else
# define SECP256K1_GNUC_PREREQ(_maj,_min) 0
# endif
# endif
/* When this header is used at build-time the SECP256K1_BUILD define needs to be set
* to correctly setup export attributes and nullness checks. This is normally done
* by secp256k1.c but to guard against this header being included before secp256k1.c
* has had a chance to set the define (e.g. via test harnesses that just includes
* secp256k1.c) we set SECP256K1_NO_BUILD when this header is processed without the
* BUILD define so this condition can be caught.
*/
#ifndef SECP256K1_BUILD
# define SECP256K1_NO_BUILD
#endif
/* Symbol visibility. */
#if defined(_WIN32)
/* GCC for Windows (e.g., MinGW) accepts the __declspec syntax
* for MSVC compatibility. A __declspec declaration implies (but is not
* exactly equivalent to) __attribute__ ((visibility("default"))), and so we
* actually want __declspec even on GCC, see "Microsoft Windows Function
* Attributes" in the GCC manual and the recommendations in
* https://gcc.gnu.org/wiki/Visibility. */
# if defined(SECP256K1_BUILD)
# if defined(DLL_EXPORT) || defined(SECP256K1_DLL_EXPORT)
/* Building libsecp256k1 as a DLL.
* 1. If using Libtool, it defines DLL_EXPORT automatically.
* 2. In other cases, SECP256K1_DLL_EXPORT must be defined. */
# define SECP256K1_API extern __declspec (dllexport)
# else
/* Building libsecp256k1 as a static library on Windows.
* No declspec is needed, and so we would want the non-Windows-specific
* logic below take care of this case. However, this may result in setting
* __attribute__ ((visibility("default"))), which is supposed to be a noop
* on Windows but may trigger warnings when compiling with -flto due to a
* bug in GCC, see
* https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116478 . */
# define SECP256K1_API extern
# endif
/* The user must define SECP256K1_STATIC when consuming libsecp256k1 as a static
* library on Windows. */
# elif !defined(SECP256K1_STATIC)
/* Consuming libsecp256k1 as a DLL. */
# define SECP256K1_API extern __declspec (dllimport)
# endif
#endif
#ifndef SECP256K1_API
/* All cases not captured by the Windows-specific logic. */
# if defined(__GNUC__) && (__GNUC__ >= 4) && defined(SECP256K1_BUILD)
/* Building libsecp256k1 using GCC or compatible. */
# define SECP256K1_API extern __attribute__ ((visibility ("default")))
# else
/* Fall back to standard C's extern. */
# define SECP256K1_API extern
# endif
#endif
/* Warning attributes
* NONNULL is not used if SECP256K1_BUILD is set to avoid the compiler optimizing out
* some paranoid null checks. */
# if defined(__GNUC__) && SECP256K1_GNUC_PREREQ(3, 4)
# define SECP256K1_WARN_UNUSED_RESULT __attribute__ ((__warn_unused_result__))
# else
# define SECP256K1_WARN_UNUSED_RESULT
# endif
# if !defined(SECP256K1_BUILD) && defined(__GNUC__) && SECP256K1_GNUC_PREREQ(3, 4)
# define SECP256K1_ARG_NONNULL(_x) __attribute__ ((__nonnull__(_x)))
# else
# define SECP256K1_ARG_NONNULL(_x)
# endif
/* Attribute for marking functions, types, and variables as deprecated */
#if !defined(SECP256K1_BUILD) && defined(__has_attribute)
# if __has_attribute(__deprecated__)
# define SECP256K1_DEPRECATED(_msg) __attribute__ ((__deprecated__(_msg)))
# else
# define SECP256K1_DEPRECATED(_msg)
# endif
#else
# define SECP256K1_DEPRECATED(_msg)
#endif
/* All flags' lower 8 bits indicate what they're for. Do not use directly. */
#define SECP256K1_FLAGS_TYPE_MASK ((1 << 8) - 1)
#define SECP256K1_FLAGS_TYPE_CONTEXT (1 << 0)
#define SECP256K1_FLAGS_TYPE_COMPRESSION (1 << 1)
/* The higher bits contain the actual data. Do not use directly. */
#define SECP256K1_FLAGS_BIT_CONTEXT_VERIFY (1 << 8)
#define SECP256K1_FLAGS_BIT_CONTEXT_SIGN (1 << 9)
#define SECP256K1_FLAGS_BIT_CONTEXT_DECLASSIFY (1 << 10)
#define SECP256K1_FLAGS_BIT_COMPRESSION (1 << 8)
/** Context flags to pass to secp256k1_context_create, secp256k1_context_preallocated_size, and
* secp256k1_context_preallocated_create. */
#define SECP256K1_CONTEXT_NONE (SECP256K1_FLAGS_TYPE_CONTEXT)
/** Deprecated context flags. These flags are treated equivalent to SECP256K1_CONTEXT_NONE. */
#define SECP256K1_CONTEXT_VERIFY (SECP256K1_FLAGS_TYPE_CONTEXT | SECP256K1_FLAGS_BIT_CONTEXT_VERIFY)
#define SECP256K1_CONTEXT_SIGN (SECP256K1_FLAGS_TYPE_CONTEXT | SECP256K1_FLAGS_BIT_CONTEXT_SIGN)
/* Testing flag. Do not use. */
#define SECP256K1_CONTEXT_DECLASSIFY (SECP256K1_FLAGS_TYPE_CONTEXT | SECP256K1_FLAGS_BIT_CONTEXT_DECLASSIFY)
/** Flag to pass to secp256k1_ec_pubkey_serialize. */
#define SECP256K1_EC_COMPRESSED (SECP256K1_FLAGS_TYPE_COMPRESSION | SECP256K1_FLAGS_BIT_COMPRESSION)
#define SECP256K1_EC_UNCOMPRESSED (SECP256K1_FLAGS_TYPE_COMPRESSION)
/** Prefix byte used to tag various encoded curvepoints for specific purposes */
#define SECP256K1_TAG_PUBKEY_EVEN 0x02
#define SECP256K1_TAG_PUBKEY_ODD 0x03
#define SECP256K1_TAG_PUBKEY_UNCOMPRESSED 0x04
#define SECP256K1_TAG_PUBKEY_HYBRID_EVEN 0x06
#define SECP256K1_TAG_PUBKEY_HYBRID_ODD 0x07
/** A built-in constant secp256k1 context object with static storage duration, to be
* used in conjunction with secp256k1_selftest.
*
* This context object offers *only limited functionality* , i.e., it cannot be used
* for API functions that perform computations involving secret keys, e.g., signing
* and public key generation. If this restriction applies to a specific API function,
* it is mentioned in its documentation. See secp256k1_context_create if you need a
* full context object that supports all functionality offered by the library.
*
* It is highly recommended to call secp256k1_selftest before using this context.
*/
SECP256K1_API const secp256k1_context *secp256k1_context_static;
/** Deprecated alias for secp256k1_context_static. */
SECP256K1_API const secp256k1_context *secp256k1_context_no_precomp
SECP256K1_DEPRECATED("Use secp256k1_context_static instead");
/** Perform basic self tests (to be used in conjunction with secp256k1_context_static)
*
* This function performs self tests that detect some serious usage errors and
* similar conditions, e.g., when the library is compiled for the wrong endianness.
* This is a last resort measure to be used in production. The performed tests are
* very rudimentary and are not intended as a replacement for running the test
* binaries.
*
* It is highly recommended to call this before using secp256k1_context_static.
* It is not necessary to call this function before using a context created with
* secp256k1_context_create (or secp256k1_context_preallocated_create), which will
* take care of performing the self tests.
*
* If the tests fail, this function will call the default error handler to abort the
* program (see secp256k1_context_set_error_callback).
*/
SECP256K1_API void secp256k1_selftest(void);
/** Create a secp256k1 context object (in dynamically allocated memory).
*
* This function uses malloc to allocate memory. It is guaranteed that malloc is
* called at most once for every call of this function. If you need to avoid dynamic
* memory allocation entirely, see secp256k1_context_static and the functions in
* secp256k1_preallocated.h.
*
* Returns: pointer to a newly created context object.
* In: flags: Always set to SECP256K1_CONTEXT_NONE (see below).
*
* The only valid non-deprecated flag in recent library versions is
* SECP256K1_CONTEXT_NONE, which will create a context sufficient for all functionality
* offered by the library. All other (deprecated) flags will be treated as equivalent
* to the SECP256K1_CONTEXT_NONE flag. Though the flags parameter primarily exists for
* historical reasons, future versions of the library may introduce new flags.
*
* If the context is intended to be used for API functions that perform computations
* involving secret keys, e.g., signing and public key generation, then it is highly
* recommended to call secp256k1_context_randomize on the context before calling
* those API functions. This will provide enhanced protection against side-channel
* leakage, see secp256k1_context_randomize for details.
*
* Do not create a new context object for each operation, as construction and
* randomization can take non-negligible time.
*/
SECP256K1_API secp256k1_context *secp256k1_context_create(
unsigned int flags
) SECP256K1_WARN_UNUSED_RESULT;
/** Copy a secp256k1 context object (into dynamically allocated memory).
*
* This function uses malloc to allocate memory. It is guaranteed that malloc is
* called at most once for every call of this function. If you need to avoid dynamic
* memory allocation entirely, see the functions in secp256k1_preallocated.h.
*
* Cloning secp256k1_context_static is not possible, and should not be emulated by
* the caller (e.g., using memcpy). Create a new context instead.
*
* Returns: pointer to a newly created context object.
* Args: ctx: pointer to a context to copy (not secp256k1_context_static).
*/
SECP256K1_API secp256k1_context *secp256k1_context_clone(
const secp256k1_context *ctx
) SECP256K1_ARG_NONNULL(1) SECP256K1_WARN_UNUSED_RESULT;
/** Destroy a secp256k1 context object (created in dynamically allocated memory).
*
* The context pointer may not be used afterwards.
*
* The context to destroy must have been created using secp256k1_context_create
* or secp256k1_context_clone. If the context has instead been created using
* secp256k1_context_preallocated_create or secp256k1_context_preallocated_clone, the
* behaviour is undefined. In that case, secp256k1_context_preallocated_destroy must
* be used instead.
*
* Args: ctx: pointer to a context to destroy, constructed using
* secp256k1_context_create or secp256k1_context_clone
* (i.e., not secp256k1_context_static).
*/
SECP256K1_API void secp256k1_context_destroy(
secp256k1_context *ctx
) SECP256K1_ARG_NONNULL(1);
/** Set a callback function to be called when an illegal argument is passed to
* an API call. It will only trigger for violations that are mentioned
* explicitly in the header.
*
* The philosophy is that these shouldn't be dealt with through a
* specific return value, as calling code should not have branches to deal with
* the case that this code itself is broken.
*
* On the other hand, during debug stage, one would want to be informed about
* such mistakes, and the default (crashing) may be inadvisable.
* When this callback is triggered, the API function called is guaranteed not
* to cause a crash, though its return value and output arguments are
* undefined.
*
* When this function has not been called (or called with fn==NULL), then the
* default handler will be used. The library provides a default handler which
* writes the message to stderr and calls abort. This default handler can be
* replaced at link time if the preprocessor macro
* USE_EXTERNAL_DEFAULT_CALLBACKS is defined, which is the case if the build
* has been configured with --enable-external-default-callbacks. Then the
* following two symbols must be provided to link against:
* - void secp256k1_default_illegal_callback_fn(const char *message, void *data);
* - void secp256k1_default_error_callback_fn(const char *message, void *data);
* The library can call these default handlers even before a proper callback data
* pointer could have been set using secp256k1_context_set_illegal_callback or
* secp256k1_context_set_error_callback, e.g., when the creation of a context
* fails. In this case, the corresponding default handler will be called with
* the data pointer argument set to NULL.
*
* Args: ctx: pointer to a context object.
* In: fun: pointer to a function to call when an illegal argument is
* passed to the API, taking a message and an opaque pointer.
* (NULL restores the default handler.)
* data: the opaque pointer to pass to fun above, must be NULL for the default handler.
*
* See also secp256k1_context_set_error_callback.
*/
SECP256K1_API void secp256k1_context_set_illegal_callback(
secp256k1_context *ctx,
void (*fun)(const char *message, void *data),
const void *data
) SECP256K1_ARG_NONNULL(1);
/** Set a callback function to be called when an internal consistency check
* fails.
*
* The default callback writes an error message to stderr and calls abort
* to abort the program.
*
* This can only trigger in case of a hardware failure, miscompilation,
* memory corruption, serious bug in the library, or other error would can
* otherwise result in undefined behaviour. It will not trigger due to mere
* incorrect usage of the API (see secp256k1_context_set_illegal_callback
* for that). After this callback returns, anything may happen, including
* crashing.
*
* Args: ctx: pointer to a context object.
* In: fun: pointer to a function to call when an internal error occurs,
* taking a message and an opaque pointer (NULL restores the
* default handler, see secp256k1_context_set_illegal_callback
* for details).
* data: the opaque pointer to pass to fun above, must be NULL for the default handler.
*
* See also secp256k1_context_set_illegal_callback.
*/
SECP256K1_API void secp256k1_context_set_error_callback(
secp256k1_context *ctx,
void (*fun)(const char *message, void *data),
const void *data
) SECP256K1_ARG_NONNULL(1);
/** Parse a variable-length public key into the pubkey object.
*
* Returns: 1 if the public key was fully valid.
* 0 if the public key could not be parsed or is invalid.
* Args: ctx: pointer to a context object.
* Out: pubkey: pointer to a pubkey object. If 1 is returned, it is set to a
* parsed version of input. If not, its value is undefined.
* In: input: pointer to a serialized public key
* inputlen: length of the array pointed to by input
*
* This function supports parsing compressed (33 bytes, header byte 0x02 or
* 0x03), uncompressed (65 bytes, header byte 0x04), or hybrid (65 bytes, header
* byte 0x06 or 0x07) format public keys.
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_ec_pubkey_parse(
const secp256k1_context *ctx,
secp256k1_pubkey *pubkey,
const unsigned char *input,
size_t inputlen
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3);
/** Serialize a pubkey object into a serialized byte sequence.
*
* Returns: 1 always.
* Args: ctx: pointer to a context object.
* Out: output: pointer to a 65-byte (if compressed==0) or 33-byte (if
* compressed==1) byte array to place the serialized key
* in.
* In/Out: outputlen: pointer to an integer which is initially set to the
* size of output, and is overwritten with the written
* size.
* In: pubkey: pointer to a secp256k1_pubkey containing an
* initialized public key.
* flags: SECP256K1_EC_COMPRESSED if serialization should be in
* compressed format, otherwise SECP256K1_EC_UNCOMPRESSED.
*/
SECP256K1_API int secp256k1_ec_pubkey_serialize(
const secp256k1_context *ctx,
unsigned char *output,
size_t *outputlen,
const secp256k1_pubkey *pubkey,
unsigned int flags
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3) SECP256K1_ARG_NONNULL(4);
/** Compare two public keys using lexicographic (of compressed serialization) order
*
* Returns: <0 if the first public key is less than the second
* >0 if the first public key is greater than the second
* 0 if the two public keys are equal
* Args: ctx: pointer to a context object
* In: pubkey1: first public key to compare
* pubkey2: second public key to compare
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_ec_pubkey_cmp(
const secp256k1_context *ctx,
const secp256k1_pubkey *pubkey1,
const secp256k1_pubkey *pubkey2
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3);
/** Sort public keys using lexicographic (of compressed serialization) order
*
* Returns: 0 if the arguments are invalid. 1 otherwise.
*
* Args: ctx: pointer to a context object
* In: pubkeys: array of pointers to pubkeys to sort
* n_pubkeys: number of elements in the pubkeys array
*/
SECP256K1_API int secp256k1_ec_pubkey_sort(
const secp256k1_context *ctx,
const secp256k1_pubkey **pubkeys,
size_t n_pubkeys
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2);
/** Parse an ECDSA signature in compact (64 bytes) format.
*
* Returns: 1 when the signature could be parsed, 0 otherwise.
* Args: ctx: pointer to a context object
* Out: sig: pointer to a signature object
* In: input64: pointer to the 64-byte array to parse
*
* The signature must consist of a 32-byte big endian R value, followed by a
* 32-byte big endian S value. If R or S fall outside of [0..order-1], the
* encoding is invalid. R and S with value 0 are allowed in the encoding.
*
* After the call, sig will always be initialized. If parsing failed or R or
* S are zero, the resulting sig value is guaranteed to fail verification for
* any message and public key.
*/
SECP256K1_API int secp256k1_ecdsa_signature_parse_compact(
const secp256k1_context *ctx,
secp256k1_ecdsa_signature *sig,
const unsigned char *input64
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3);
/** Parse a DER ECDSA signature.
*
* Returns: 1 when the signature could be parsed, 0 otherwise.
* Args: ctx: pointer to a context object
* Out: sig: pointer to a signature object
* In: input: pointer to the signature to be parsed
* inputlen: the length of the array pointed to be input
*
* This function will accept any valid DER encoded signature, even if the
* encoded numbers are out of range.
*
* After the call, sig will always be initialized. If parsing failed or the
* encoded numbers are out of range, signature verification with it is
* guaranteed to fail for every message and public key.
*/
SECP256K1_API int secp256k1_ecdsa_signature_parse_der(
const secp256k1_context *ctx,
secp256k1_ecdsa_signature *sig,
const unsigned char *input,
size_t inputlen
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3);
/** Serialize an ECDSA signature in DER format.
*
* Returns: 1 if enough space was available to serialize, 0 otherwise
* Args: ctx: pointer to a context object
* Out: output: pointer to an array to store the DER serialization
* In/Out: outputlen: pointer to a length integer. Initially, this integer
* should be set to the length of output. After the call
* it will be set to the length of the serialization (even
* if 0 was returned).
* In: sig: pointer to an initialized signature object
*/
SECP256K1_API int secp256k1_ecdsa_signature_serialize_der(
const secp256k1_context *ctx,
unsigned char *output,
size_t *outputlen,
const secp256k1_ecdsa_signature *sig
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3) SECP256K1_ARG_NONNULL(4);
/** Serialize an ECDSA signature in compact (64 byte) format.
*
* Returns: 1
* Args: ctx: pointer to a context object
* Out: output64: pointer to a 64-byte array to store the compact serialization
* In: sig: pointer to an initialized signature object
*
* See secp256k1_ecdsa_signature_parse_compact for details about the encoding.
*/
SECP256K1_API int secp256k1_ecdsa_signature_serialize_compact(
const secp256k1_context *ctx,
unsigned char *output64,
const secp256k1_ecdsa_signature *sig
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3);
/** Verify an ECDSA signature.
*
* Returns: 1: correct signature
* 0: incorrect or unparseable signature
* Args: ctx: pointer to a context object
* In: sig: the signature being verified.
* msghash32: the 32-byte message hash being verified.
* The verifier must make sure to apply a cryptographic
* hash function to the message by itself and not accept an
* msghash32 value directly. Otherwise, it would be easy to
* create a "valid" signature without knowledge of the
* secret key. See also
* https://bitcoin.stackexchange.com/a/81116/35586 for more
* background on this topic.
* pubkey: pointer to an initialized public key to verify with.
*
* To avoid accepting malleable signatures, only ECDSA signatures in lower-S
* form are accepted.
*
* If you need to accept ECDSA signatures from sources that do not obey this
* rule, apply secp256k1_ecdsa_signature_normalize to the signature prior to
* verification, but be aware that doing so results in malleable signatures.
*
* For details, see the comments for that function.
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_ecdsa_verify(
const secp256k1_context *ctx,
const secp256k1_ecdsa_signature *sig,
const unsigned char *msghash32,
const secp256k1_pubkey *pubkey
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3) SECP256K1_ARG_NONNULL(4);
/** Convert a signature to a normalized lower-S form.
*
* Returns: 1 if sigin was not normalized, 0 if it already was.
* Args: ctx: pointer to a context object
* Out: sigout: pointer to a signature to fill with the normalized form,
* or copy if the input was already normalized. (can be NULL if
* you're only interested in whether the input was already
* normalized).
* In: sigin: pointer to a signature to check/normalize (can be identical to sigout)
*
* With ECDSA a third-party can forge a second distinct signature of the same
* message, given a single initial signature, but without knowing the key. This
* is done by negating the S value modulo the order of the curve, 'flipping'
* the sign of the random point R which is not included in the signature.
*
* Forgery of the same message isn't universally problematic, but in systems
* where message malleability or uniqueness of signatures is important this can
* cause issues. This forgery can be blocked by all verifiers forcing signers
* to use a normalized form.
*
* The lower-S form reduces the size of signatures slightly on average when
* variable length encodings (such as DER) are used and is cheap to verify,
* making it a good choice. Security of always using lower-S is assured because
* anyone can trivially modify a signature after the fact to enforce this
* property anyway.
*
* The lower S value is always between 0x1 and
* 0x7FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF5D576E7357A4501DDFE92F46681B20A0,
* inclusive.
*
* No other forms of ECDSA malleability are known and none seem likely, but
* there is no formal proof that ECDSA, even with this additional restriction,
* is free of other malleability. Commonly used serialization schemes will also
* accept various non-unique encodings, so care should be taken when this
* property is required for an application.
*
* The secp256k1_ecdsa_sign function will by default create signatures in the
* lower-S form, and secp256k1_ecdsa_verify will not accept others. In case
* signatures come from a system that cannot enforce this property,
* secp256k1_ecdsa_signature_normalize must be called before verification.
*/
SECP256K1_API int secp256k1_ecdsa_signature_normalize(
const secp256k1_context *ctx,
secp256k1_ecdsa_signature *sigout,
const secp256k1_ecdsa_signature *sigin
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(3);
/** An implementation of RFC6979 (using HMAC-SHA256) as nonce generation function.
* If a data pointer is passed, it is assumed to be a pointer to 32 bytes of
* extra entropy.
*/
SECP256K1_API const secp256k1_nonce_function secp256k1_nonce_function_rfc6979;
/** A default safe nonce generation function (currently equal to secp256k1_nonce_function_rfc6979). */
SECP256K1_API const secp256k1_nonce_function secp256k1_nonce_function_default;
/** Create an ECDSA signature.
*
* Returns: 1: signature created
* 0: the nonce generation function failed, or the secret key was invalid.
* Args: ctx: pointer to a context object (not secp256k1_context_static).
* Out: sig: pointer to an array where the signature will be placed.
* In: msghash32: the 32-byte message hash being signed.
* seckey: pointer to a 32-byte secret key.
* noncefp: pointer to a nonce generation function. If NULL,
* secp256k1_nonce_function_default is used.
* ndata: pointer to arbitrary data used by the nonce generation function
* (can be NULL). If it is non-NULL and
* secp256k1_nonce_function_default is used, then ndata must be a
* pointer to 32-bytes of additional data.
*
* The created signature is always in lower-S form. See
* secp256k1_ecdsa_signature_normalize for more details.
*/
SECP256K1_API int secp256k1_ecdsa_sign(
const secp256k1_context *ctx,
secp256k1_ecdsa_signature *sig,
const unsigned char *msghash32,
const unsigned char *seckey,
secp256k1_nonce_function noncefp,
const void *ndata
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3) SECP256K1_ARG_NONNULL(4);
/** Verify an elliptic curve secret key.
*
* A secret key is valid if it is not 0 and less than the secp256k1 curve order
* when interpreted as an integer (most significant byte first). The
* probability of choosing a 32-byte string uniformly at random which is an
* invalid secret key is negligible. However, if it does happen it should
* be assumed that the randomness source is severely broken and there should
* be no retry.
*
* Returns: 1: secret key is valid
* 0: secret key is invalid
* Args: ctx: pointer to a context object.
* In: seckey: pointer to a 32-byte secret key.
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_ec_seckey_verify(
const secp256k1_context *ctx,
const unsigned char *seckey
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2);
/** Compute the public key for a secret key.
*
* Returns: 1: secret was valid, public key stores.
* 0: secret was invalid, try again.
* Args: ctx: pointer to a context object (not secp256k1_context_static).
* Out: pubkey: pointer to the created public key.
* In: seckey: pointer to a 32-byte secret key.
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_ec_pubkey_create(
const secp256k1_context *ctx,
secp256k1_pubkey *pubkey,
const unsigned char *seckey
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3);
/** Negates a secret key in place.
*
* Returns: 0 if the given secret key is invalid according to
* secp256k1_ec_seckey_verify. 1 otherwise
* Args: ctx: pointer to a context object
* In/Out: seckey: pointer to the 32-byte secret key to be negated. If the
* secret key is invalid according to
* secp256k1_ec_seckey_verify, this function returns 0 and
* seckey will be set to some unspecified value.
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_ec_seckey_negate(
const secp256k1_context *ctx,
unsigned char *seckey
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2);
/** Same as secp256k1_ec_seckey_negate, but DEPRECATED. Will be removed in
* future versions. */
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_ec_privkey_negate(
const secp256k1_context *ctx,
unsigned char *seckey
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2)
SECP256K1_DEPRECATED("Use secp256k1_ec_seckey_negate instead");
/** Negates a public key in place.
*
* Returns: 1 always
* Args: ctx: pointer to a context object
* In/Out: pubkey: pointer to the public key to be negated.
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_ec_pubkey_negate(
const secp256k1_context *ctx,
secp256k1_pubkey *pubkey
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2);
/** Tweak a secret key by adding tweak to it.
*
* Returns: 0 if the arguments are invalid or the resulting secret key would be
* invalid (only when the tweak is the negation of the secret key). 1
* otherwise.
* Args: ctx: pointer to a context object.
* In/Out: seckey: pointer to a 32-byte secret key. If the secret key is
* invalid according to secp256k1_ec_seckey_verify, this
* function returns 0. seckey will be set to some unspecified
* value if this function returns 0.
* In: tweak32: pointer to a 32-byte tweak, which must be valid according to
* secp256k1_ec_seckey_verify or 32 zero bytes. For uniformly
* random 32-byte tweaks, the chance of being invalid is
* negligible (around 1 in 2^128).
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_ec_seckey_tweak_add(
const secp256k1_context *ctx,
unsigned char *seckey,
const unsigned char *tweak32
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3);
/** Same as secp256k1_ec_seckey_tweak_add, but DEPRECATED. Will be removed in
* future versions. */
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_ec_privkey_tweak_add(
const secp256k1_context *ctx,
unsigned char *seckey,
const unsigned char *tweak32
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3)
SECP256K1_DEPRECATED("Use secp256k1_ec_seckey_tweak_add instead");
/** Tweak a public key by adding tweak times the generator to it.
*
* Returns: 0 if the arguments are invalid or the resulting public key would be
* invalid (only when the tweak is the negation of the corresponding
* secret key). 1 otherwise.
* Args: ctx: pointer to a context object.
* In/Out: pubkey: pointer to a public key object. pubkey will be set to an
* invalid value if this function returns 0.
* In: tweak32: pointer to a 32-byte tweak, which must be valid according to
* secp256k1_ec_seckey_verify or 32 zero bytes. For uniformly
* random 32-byte tweaks, the chance of being invalid is
* negligible (around 1 in 2^128).
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_ec_pubkey_tweak_add(
const secp256k1_context *ctx,
secp256k1_pubkey *pubkey,
const unsigned char *tweak32
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3);
/** Tweak a secret key by multiplying it by a tweak.
*
* Returns: 0 if the arguments are invalid. 1 otherwise.
* Args: ctx: pointer to a context object.
* In/Out: seckey: pointer to a 32-byte secret key. If the secret key is
* invalid according to secp256k1_ec_seckey_verify, this
* function returns 0. seckey will be set to some unspecified
* value if this function returns 0.
* In: tweak32: pointer to a 32-byte tweak. If the tweak is invalid according to
* secp256k1_ec_seckey_verify, this function returns 0. For
* uniformly random 32-byte arrays the chance of being invalid
* is negligible (around 1 in 2^128).
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_ec_seckey_tweak_mul(
const secp256k1_context *ctx,
unsigned char *seckey,
const unsigned char *tweak32
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3);
/** Same as secp256k1_ec_seckey_tweak_mul, but DEPRECATED. Will be removed in
* future versions. */
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_ec_privkey_tweak_mul(
const secp256k1_context *ctx,
unsigned char *seckey,
const unsigned char *tweak32
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3)
SECP256K1_DEPRECATED("Use secp256k1_ec_seckey_tweak_mul instead");
/** Tweak a public key by multiplying it by a tweak value.
*
* Returns: 0 if the arguments are invalid. 1 otherwise.
* Args: ctx: pointer to a context object.
* In/Out: pubkey: pointer to a public key object. pubkey will be set to an
* invalid value if this function returns 0.
* In: tweak32: pointer to a 32-byte tweak. If the tweak is invalid according to
* secp256k1_ec_seckey_verify, this function returns 0. For
* uniformly random 32-byte arrays the chance of being invalid
* is negligible (around 1 in 2^128).
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_ec_pubkey_tweak_mul(
const secp256k1_context *ctx,
secp256k1_pubkey *pubkey,
const unsigned char *tweak32
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3);
/** Randomizes the context to provide enhanced protection against side-channel leakage.
*
* Returns: 1: randomization successful
* 0: error
* Args: ctx: pointer to a context object (not secp256k1_context_static).
* In: seed32: pointer to a 32-byte random seed (NULL resets to initial state).
*
* While secp256k1 code is written and tested to be constant-time no matter what
* secret values are, it is possible that a compiler may output code which is not,
* and also that the CPU may not emit the same radio frequencies or draw the same
* amount of power for all values. Randomization of the context shields against
* side-channel observations which aim to exploit secret-dependent behaviour in
* certain computations which involve secret keys.
*
* It is highly recommended to call this function on contexts returned from
* secp256k1_context_create or secp256k1_context_clone (or from the corresponding
* functions in secp256k1_preallocated.h) before using these contexts to call API
* functions that perform computations involving secret keys, e.g., signing and
* public key generation. It is possible to call this function more than once on
* the same context, and doing so before every few computations involving secret
* keys is recommended as a defense-in-depth measure. Randomization of the static
* context secp256k1_context_static is not supported.
*
* Currently, the random seed is mainly used for blinding multiplications of a
* secret scalar with the elliptic curve base point. Multiplications of this
* kind are performed by exactly those API functions which are documented to
* require a context that is not secp256k1_context_static. As a rule of thumb,
* these are all functions which take a secret key (or a keypair) as an input.
* A notable exception to that rule is the ECDH module, which relies on a different
* kind of elliptic curve point multiplication and thus does not benefit from
* enhanced protection against side-channel leakage currently.
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_context_randomize(
secp256k1_context *ctx,
const unsigned char *seed32
) SECP256K1_ARG_NONNULL(1);
/** Add a number of public keys together.
*
* Returns: 1: the sum of the public keys is valid.
* 0: the sum of the public keys is not valid.
* Args: ctx: pointer to a context object.
* Out: out: pointer to a public key object for placing the resulting public key.
* In: ins: pointer to array of pointers to public keys.
* n: the number of public keys to add together (must be at least 1).
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_ec_pubkey_combine(
const secp256k1_context *ctx,
secp256k1_pubkey *out,
const secp256k1_pubkey * const *ins,
size_t n
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3);
/** Compute a tagged hash as defined in BIP-340.
*
* This is useful for creating a message hash and achieving domain separation
* through an application-specific tag. This function returns
* SHA256(SHA256(tag)||SHA256(tag)||msg). Therefore, tagged hash
* implementations optimized for a specific tag can precompute the SHA256 state
* after hashing the tag hashes.
*
* Returns: 1 always.
* Args: ctx: pointer to a context object
* Out: hash32: pointer to a 32-byte array to store the resulting hash
* In: tag: pointer to an array containing the tag
* taglen: length of the tag array
* msg: pointer to an array containing the message
* msglen: length of the message array
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_tagged_sha256(
const secp256k1_context *ctx,
unsigned char *hash32,
const unsigned char *tag,
size_t taglen,
const unsigned char *msg,
size_t msglen
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3) SECP256K1_ARG_NONNULL(5);
#ifdef __cplusplus
}
#endif
#endif /* SECP256K1_H */

View File

@@ -0,0 +1,63 @@
#ifndef SECP256K1_ECDH_H
#define SECP256K1_ECDH_H
#include "secp256k1.h"
#ifdef __cplusplus
extern "C" {
#endif
/** A pointer to a function that hashes an EC point to obtain an ECDH secret
*
* Returns: 1 if the point was successfully hashed.
* 0 will cause secp256k1_ecdh to fail and return 0.
* Other return values are not allowed, and the behaviour of
* secp256k1_ecdh is undefined for other return values.
* Out: output: pointer to an array to be filled by the function
* In: x32: pointer to a 32-byte x coordinate
* y32: pointer to a 32-byte y coordinate
* data: arbitrary data pointer that is passed through
*/
typedef int (*secp256k1_ecdh_hash_function)(
unsigned char *output,
const unsigned char *x32,
const unsigned char *y32,
void *data
);
/** An implementation of SHA256 hash function that applies to compressed public key.
* Populates the output parameter with 32 bytes. */
SECP256K1_API const secp256k1_ecdh_hash_function secp256k1_ecdh_hash_function_sha256;
/** A default ECDH hash function (currently equal to secp256k1_ecdh_hash_function_sha256).
* Populates the output parameter with 32 bytes. */
SECP256K1_API const secp256k1_ecdh_hash_function secp256k1_ecdh_hash_function_default;
/** Compute an EC Diffie-Hellman secret in constant time
*
* Returns: 1: exponentiation was successful
* 0: scalar was invalid (zero or overflow) or hashfp returned 0
* Args: ctx: pointer to a context object.
* Out: output: pointer to an array to be filled by hashfp.
* In: pubkey: pointer to a secp256k1_pubkey containing an initialized public key.
* seckey: a 32-byte scalar with which to multiply the point.
* hashfp: pointer to a hash function. If NULL,
* secp256k1_ecdh_hash_function_sha256 is used
* (in which case, 32 bytes will be written to output).
* data: arbitrary data pointer that is passed through to hashfp
* (can be NULL for secp256k1_ecdh_hash_function_sha256).
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_ecdh(
const secp256k1_context *ctx,
unsigned char *output,
const secp256k1_pubkey *pubkey,
const unsigned char *seckey,
secp256k1_ecdh_hash_function hashfp,
void *data
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3) SECP256K1_ARG_NONNULL(4);
#ifdef __cplusplus
}
#endif
#endif /* SECP256K1_ECDH_H */

View File

@@ -0,0 +1,200 @@
#ifndef SECP256K1_ELLSWIFT_H
#define SECP256K1_ELLSWIFT_H
#include "secp256k1.h"
#ifdef __cplusplus
extern "C" {
#endif
/* This module provides an implementation of ElligatorSwift as well as a
* version of x-only ECDH using it (including compatibility with BIP324).
*
* ElligatorSwift is described in https://eprint.iacr.org/2022/759 by
* Chavez-Saab, Rodriguez-Henriquez, and Tibouchi. It permits encoding
* uniformly chosen public keys as 64-byte arrays which are indistinguishable
* from uniformly random arrays.
*
* Let f be the function from pairs of field elements to point X coordinates,
* defined as follows (all operations modulo p = 2^256 - 2^32 - 977)
* f(u,t):
* - Let C = 0xa2d2ba93507f1df233770c2a797962cc61f6d15da14ecd47d8d27ae1cd5f852,
* a square root of -3.
* - If u=0, set u=1 instead.
* - If t=0, set t=1 instead.
* - If u^3 + t^2 + 7 = 0, multiply t by 2.
* - Let X = (u^3 + 7 - t^2) / (2 * t)
* - Let Y = (X + t) / (C * u)
* - Return the first in [u + 4 * Y^2, (-X/Y - u) / 2, (X/Y - u) / 2] that is an
* X coordinate on the curve (at least one of them is, for any u and t).
*
* Then an ElligatorSwift encoding of x consists of the 32-byte big-endian
* encodings of field elements u and t concatenated, where f(u,t) = x.
* The encoding algorithm is described in the paper, and effectively picks a
* uniformly random pair (u,t) among those which encode x.
*
* If the Y coordinate is relevant, it is given the same parity as t.
*
* Changes w.r.t. the paper:
* - The u=0, t=0, and u^3+t^2+7=0 conditions result in decoding to the point
* at infinity in the paper. Here they are remapped to finite points.
* - The paper uses an additional encoding bit for the parity of y. Here the
* parity of t is used (negating t does not affect the decoded x coordinate,
* so this is possible).
*
* For mathematical background about the scheme, see the doc/ellswift.md file.
*/
/** A pointer to a function used by secp256k1_ellswift_xdh to hash the shared X
* coordinate along with the encoded public keys to a uniform shared secret.
*
* Returns: 1 if a shared secret was successfully computed.
* 0 will cause secp256k1_ellswift_xdh to fail and return 0.
* Other return values are not allowed, and the behaviour of
* secp256k1_ellswift_xdh is undefined for other return values.
* Out: output: pointer to an array to be filled by the function
* In: x32: pointer to the 32-byte serialized X coordinate
* of the resulting shared point (will not be NULL)
* ell_a64: pointer to the 64-byte encoded public key of party A
* (will not be NULL)
* ell_b64: pointer to the 64-byte encoded public key of party B
* (will not be NULL)
* data: arbitrary data pointer that is passed through
*/
typedef int (*secp256k1_ellswift_xdh_hash_function)(
unsigned char *output,
const unsigned char *x32,
const unsigned char *ell_a64,
const unsigned char *ell_b64,
void *data
);
/** An implementation of an secp256k1_ellswift_xdh_hash_function which uses
* SHA256(prefix64 || ell_a64 || ell_b64 || x32), where prefix64 is the 64-byte
* array pointed to by data. */
SECP256K1_API const secp256k1_ellswift_xdh_hash_function secp256k1_ellswift_xdh_hash_function_prefix;
/** An implementation of an secp256k1_ellswift_xdh_hash_function compatible with
* BIP324. It returns H_tag(ell_a64 || ell_b64 || x32), where H_tag is the
* BIP340 tagged hash function with tag "bip324_ellswift_xonly_ecdh". Equivalent
* to secp256k1_ellswift_xdh_hash_function_prefix with prefix64 set to
* SHA256("bip324_ellswift_xonly_ecdh")||SHA256("bip324_ellswift_xonly_ecdh").
* The data argument is ignored. */
SECP256K1_API const secp256k1_ellswift_xdh_hash_function secp256k1_ellswift_xdh_hash_function_bip324;
/** Construct a 64-byte ElligatorSwift encoding of a given pubkey.
*
* Returns: 1 always.
* Args: ctx: pointer to a context object
* Out: ell64: pointer to a 64-byte array to be filled
* In: pubkey: pointer to a secp256k1_pubkey containing an
* initialized public key
* rnd32: pointer to 32 bytes of randomness
*
* It is recommended that rnd32 consists of 32 uniformly random bytes, not
* known to any adversary trying to detect whether public keys are being
* encoded, though 16 bytes of randomness (padded to an array of 32 bytes,
* e.g., with zeros) suffice to make the result indistinguishable from
* uniform. The randomness in rnd32 must not be a deterministic function of
* the pubkey (it can be derived from the private key, though).
*
* It is not guaranteed that the computed encoding is stable across versions
* of the library, even if all arguments to this function (including rnd32)
* are the same.
*
* This function runs in variable time.
*/
SECP256K1_API int secp256k1_ellswift_encode(
const secp256k1_context *ctx,
unsigned char *ell64,
const secp256k1_pubkey *pubkey,
const unsigned char *rnd32
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3) SECP256K1_ARG_NONNULL(4);
/** Decode a 64-bytes ElligatorSwift encoded public key.
*
* Returns: always 1
* Args: ctx: pointer to a context object
* Out: pubkey: pointer to a secp256k1_pubkey that will be filled
* In: ell64: pointer to a 64-byte array to decode
*
* This function runs in variable time.
*/
SECP256K1_API int secp256k1_ellswift_decode(
const secp256k1_context *ctx,
secp256k1_pubkey *pubkey,
const unsigned char *ell64
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3);
/** Compute an ElligatorSwift public key for a secret key.
*
* Returns: 1: secret was valid, public key was stored.
* 0: secret was invalid, try again.
* Args: ctx: pointer to a context object
* Out: ell64: pointer to a 64-byte array to receive the ElligatorSwift
* public key
* In: seckey32: pointer to a 32-byte secret key
* auxrnd32: (optional) pointer to 32 bytes of randomness
*
* Constant time in seckey and auxrnd32, but not in the resulting public key.
*
* It is recommended that auxrnd32 contains 32 uniformly random bytes, though
* it is optional (and does result in encodings that are indistinguishable from
* uniform even without any auxrnd32). It differs from the (mandatory) rnd32
* argument to secp256k1_ellswift_encode in this regard.
*
* This function can be used instead of calling secp256k1_ec_pubkey_create
* followed by secp256k1_ellswift_encode. It is safer, as it uses the secret
* key as entropy for the encoding (supplemented with auxrnd32, if provided).
*
* Like secp256k1_ellswift_encode, this function does not guarantee that the
* computed encoding is stable across versions of the library, even if all
* arguments (including auxrnd32) are the same.
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_ellswift_create(
const secp256k1_context *ctx,
unsigned char *ell64,
const unsigned char *seckey32,
const unsigned char *auxrnd32
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3);
/** Given a private key, and ElligatorSwift public keys sent in both directions,
* compute a shared secret using x-only Elliptic Curve Diffie-Hellman (ECDH).
*
* Returns: 1: shared secret was successfully computed
* 0: secret was invalid or hashfp returned 0
* Args: ctx: pointer to a context object.
* Out: output: pointer to an array to be filled by hashfp.
* In: ell_a64: pointer to the 64-byte encoded public key of party A
* (will not be NULL)
* ell_b64: pointer to the 64-byte encoded public key of party B
* (will not be NULL)
* seckey32: pointer to our 32-byte secret key
* party: boolean indicating which party we are: zero if we are
* party A, non-zero if we are party B. seckey32 must be
* the private key corresponding to that party's ell_?64.
* This correspondence is not checked.
* hashfp: pointer to a hash function.
* data: arbitrary data pointer passed through to hashfp.
*
* Constant time in seckey32.
*
* This function is more efficient than decoding the public keys, and performing
* ECDH on them.
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_ellswift_xdh(
const secp256k1_context *ctx,
unsigned char *output,
const unsigned char *ell_a64,
const unsigned char *ell_b64,
const unsigned char *seckey32,
int party,
secp256k1_ellswift_xdh_hash_function hashfp,
void *data
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3) SECP256K1_ARG_NONNULL(4) SECP256K1_ARG_NONNULL(5) SECP256K1_ARG_NONNULL(7);
#ifdef __cplusplus
}
#endif
#endif /* SECP256K1_ELLSWIFT_H */

View File

@@ -0,0 +1,250 @@
#ifndef SECP256K1_EXTRAKEYS_H
#define SECP256K1_EXTRAKEYS_H
#include "secp256k1.h"
#ifdef __cplusplus
extern "C" {
#endif
/** Opaque data structure that holds a parsed and valid "x-only" public key.
* An x-only pubkey encodes a point whose Y coordinate is even. It is
* serialized using only its X coordinate (32 bytes). See BIP-340 for more
* information about x-only pubkeys.
*
* The exact representation of data inside is implementation defined and not
* guaranteed to be portable between different platforms or versions. It is
* however guaranteed to be 64 bytes in size, and can be safely copied/moved.
* If you need to convert to a format suitable for storage, transmission, use
* use secp256k1_xonly_pubkey_serialize and secp256k1_xonly_pubkey_parse. To
* compare keys, use secp256k1_xonly_pubkey_cmp.
*/
typedef struct secp256k1_xonly_pubkey {
unsigned char data[64];
} secp256k1_xonly_pubkey;
/** Opaque data structure that holds a keypair consisting of a secret and a
* public key.
*
* The exact representation of data inside is implementation defined and not
* guaranteed to be portable between different platforms or versions. It is
* however guaranteed to be 96 bytes in size, and can be safely copied/moved.
*/
typedef struct secp256k1_keypair {
unsigned char data[96];
} secp256k1_keypair;
/** Parse a 32-byte sequence into a xonly_pubkey object.
*
* Returns: 1 if the public key was fully valid.
* 0 if the public key could not be parsed or is invalid.
*
* Args: ctx: pointer to a context object.
* Out: pubkey: pointer to a pubkey object. If 1 is returned, it is set to a
* parsed version of input. If not, it's set to an invalid value.
* In: input32: pointer to a serialized xonly_pubkey.
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_xonly_pubkey_parse(
const secp256k1_context *ctx,
secp256k1_xonly_pubkey *pubkey,
const unsigned char *input32
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3);
/** Serialize an xonly_pubkey object into a 32-byte sequence.
*
* Returns: 1 always.
*
* Args: ctx: pointer to a context object.
* Out: output32: pointer to a 32-byte array to place the serialized key in.
* In: pubkey: pointer to a secp256k1_xonly_pubkey containing an initialized public key.
*/
SECP256K1_API int secp256k1_xonly_pubkey_serialize(
const secp256k1_context *ctx,
unsigned char *output32,
const secp256k1_xonly_pubkey *pubkey
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3);
/** Compare two x-only public keys using lexicographic order
*
* Returns: <0 if the first public key is less than the second
* >0 if the first public key is greater than the second
* 0 if the two public keys are equal
* Args: ctx: pointer to a context object.
* In: pubkey1: first public key to compare
* pubkey2: second public key to compare
*/
SECP256K1_API int secp256k1_xonly_pubkey_cmp(
const secp256k1_context *ctx,
const secp256k1_xonly_pubkey *pk1,
const secp256k1_xonly_pubkey *pk2
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3);
/** Converts a secp256k1_pubkey into a secp256k1_xonly_pubkey.
*
* Returns: 1 always.
*
* Args: ctx: pointer to a context object.
* Out: xonly_pubkey: pointer to an x-only public key object for placing the converted public key.
* pk_parity: Ignored if NULL. Otherwise, pointer to an integer that
* will be set to 1 if the point encoded by xonly_pubkey is
* the negation of the pubkey and set to 0 otherwise.
* In: pubkey: pointer to a public key that is converted.
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_xonly_pubkey_from_pubkey(
const secp256k1_context *ctx,
secp256k1_xonly_pubkey *xonly_pubkey,
int *pk_parity,
const secp256k1_pubkey *pubkey
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(4);
/** Tweak an x-only public key by adding the generator multiplied with tweak32
* to it.
*
* Note that the resulting point can not in general be represented by an x-only
* pubkey because it may have an odd Y coordinate. Instead, the output_pubkey
* is a normal secp256k1_pubkey.
*
* Returns: 0 if the arguments are invalid or the resulting public key would be
* invalid (only when the tweak is the negation of the corresponding
* secret key). 1 otherwise.
*
* Args: ctx: pointer to a context object.
* Out: output_pubkey: pointer to a public key to store the result. Will be set
* to an invalid value if this function returns 0.
* In: internal_pubkey: pointer to an x-only pubkey to apply the tweak to.
* tweak32: pointer to a 32-byte tweak, which must be valid
* according to secp256k1_ec_seckey_verify or 32 zero
* bytes. For uniformly random 32-byte tweaks, the chance of
* being invalid is negligible (around 1 in 2^128).
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_xonly_pubkey_tweak_add(
const secp256k1_context *ctx,
secp256k1_pubkey *output_pubkey,
const secp256k1_xonly_pubkey *internal_pubkey,
const unsigned char *tweak32
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3) SECP256K1_ARG_NONNULL(4);
/** Checks that a tweaked pubkey is the result of calling
* secp256k1_xonly_pubkey_tweak_add with internal_pubkey and tweak32.
*
* The tweaked pubkey is represented by its 32-byte x-only serialization and
* its pk_parity, which can both be obtained by converting the result of
* tweak_add to a secp256k1_xonly_pubkey.
*
* Note that this alone does _not_ verify that the tweaked pubkey is a
* commitment. If the tweak is not chosen in a specific way, the tweaked pubkey
* can easily be the result of a different internal_pubkey and tweak.
*
* Returns: 0 if the arguments are invalid or the tweaked pubkey is not the
* result of tweaking the internal_pubkey with tweak32. 1 otherwise.
* Args: ctx: pointer to a context object.
* In: tweaked_pubkey32: pointer to a serialized xonly_pubkey.
* tweaked_pk_parity: the parity of the tweaked pubkey (whose serialization
* is passed in as tweaked_pubkey32). This must match the
* pk_parity value that is returned when calling
* secp256k1_xonly_pubkey with the tweaked pubkey, or
* this function will fail.
* internal_pubkey: pointer to an x-only public key object to apply the tweak to.
* tweak32: pointer to a 32-byte tweak.
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_xonly_pubkey_tweak_add_check(
const secp256k1_context *ctx,
const unsigned char *tweaked_pubkey32,
int tweaked_pk_parity,
const secp256k1_xonly_pubkey *internal_pubkey,
const unsigned char *tweak32
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(4) SECP256K1_ARG_NONNULL(5);
/** Compute the keypair for a valid secret key.
*
* See the documentation of `secp256k1_ec_seckey_verify` for more information
* about the validity of secret keys.
*
* Returns: 1: secret key is valid
* 0: secret key is invalid
* Args: ctx: pointer to a context object (not secp256k1_context_static).
* Out: keypair: pointer to the created keypair.
* In: seckey: pointer to a 32-byte secret key.
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_keypair_create(
const secp256k1_context *ctx,
secp256k1_keypair *keypair,
const unsigned char *seckey
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3);
/** Get the secret key from a keypair.
*
* Returns: 1 always.
* Args: ctx: pointer to a context object.
* Out: seckey: pointer to a 32-byte buffer for the secret key.
* In: keypair: pointer to a keypair.
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_keypair_sec(
const secp256k1_context *ctx,
unsigned char *seckey,
const secp256k1_keypair *keypair
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3);
/** Get the public key from a keypair.
*
* Returns: 1 always.
* Args: ctx: pointer to a context object.
* Out: pubkey: pointer to a pubkey object, set to the keypair public key.
* In: keypair: pointer to a keypair.
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_keypair_pub(
const secp256k1_context *ctx,
secp256k1_pubkey *pubkey,
const secp256k1_keypair *keypair
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3);
/** Get the x-only public key from a keypair.
*
* This is the same as calling secp256k1_keypair_pub and then
* secp256k1_xonly_pubkey_from_pubkey.
*
* Returns: 1 always.
* Args: ctx: pointer to a context object.
* Out: pubkey: pointer to an xonly_pubkey object, set to the keypair
* public key after converting it to an xonly_pubkey.
* pk_parity: Ignored if NULL. Otherwise, pointer to an integer that will be set to the
* pk_parity argument of secp256k1_xonly_pubkey_from_pubkey.
* In: keypair: pointer to a keypair.
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_keypair_xonly_pub(
const secp256k1_context *ctx,
secp256k1_xonly_pubkey *pubkey,
int *pk_parity,
const secp256k1_keypair *keypair
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(4);
/** Tweak a keypair by adding tweak32 to the secret key and updating the public
* key accordingly.
*
* Calling this function and then secp256k1_keypair_pub results in the same
* public key as calling secp256k1_keypair_xonly_pub and then
* secp256k1_xonly_pubkey_tweak_add.
*
* Returns: 0 if the arguments are invalid or the resulting keypair would be
* invalid (only when the tweak is the negation of the keypair's
* secret key). 1 otherwise.
*
* Args: ctx: pointer to a context object.
* In/Out: keypair: pointer to a keypair to apply the tweak to. Will be set to
* an invalid value if this function returns 0.
* In: tweak32: pointer to a 32-byte tweak, which must be valid according to
* secp256k1_ec_seckey_verify or 32 zero bytes. For uniformly
* random 32-byte tweaks, the chance of being invalid is
* negligible (around 1 in 2^128).
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_keypair_xonly_tweak_add(
const secp256k1_context *ctx,
secp256k1_keypair *keypair,
const unsigned char *tweak32
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3);
#ifdef __cplusplus
}
#endif
#endif /* SECP256K1_EXTRAKEYS_H */

View File

@@ -0,0 +1,134 @@
#ifndef SECP256K1_PREALLOCATED_H
#define SECP256K1_PREALLOCATED_H
#include "secp256k1.h"
#ifdef __cplusplus
extern "C" {
#endif
/* The module provided by this header file is intended for settings in which it
* is not possible or desirable to rely on dynamic memory allocation. It provides
* functions for creating, cloning, and destroying secp256k1 context objects in a
* contiguous fixed-size block of memory provided by the caller.
*
* Context objects created by functions in this module can be used like contexts
* objects created by functions in secp256k1.h, i.e., they can be passed to any
* API function that expects a context object (see secp256k1.h for details). The
* only exception is that context objects created by functions in this module
* must be destroyed using secp256k1_context_preallocated_destroy (in this
* module) instead of secp256k1_context_destroy (in secp256k1.h).
*
* It is guaranteed that functions in this module will not call malloc or its
* friends realloc, calloc, and free.
*/
/** Determine the memory size of a secp256k1 context object to be created in
* caller-provided memory.
*
* The purpose of this function is to determine how much memory must be provided
* to secp256k1_context_preallocated_create.
*
* Returns: the required size of the caller-provided memory block
* In: flags: which parts of the context to initialize.
*/
SECP256K1_API size_t secp256k1_context_preallocated_size(
unsigned int flags
) SECP256K1_WARN_UNUSED_RESULT;
/** Create a secp256k1 context object in caller-provided memory.
*
* The caller must provide a pointer to a rewritable contiguous block of memory
* of size at least secp256k1_context_preallocated_size(flags) bytes, suitably
* aligned to hold an object of any type.
*
* The block of memory is exclusively owned by the created context object during
* the lifetime of this context object, which begins with the call to this
* function and ends when a call to secp256k1_context_preallocated_destroy
* (which destroys the context object again) returns. During the lifetime of the
* context object, the caller is obligated not to access this block of memory,
* i.e., the caller may not read or write the memory, e.g., by copying the memory
* contents to a different location or trying to create a second context object
* in the memory. In simpler words, the prealloc pointer (or any pointer derived
* from it) should not be used during the lifetime of the context object.
*
* Returns: pointer to newly created context object.
* In: prealloc: pointer to a rewritable contiguous block of memory of
* size at least secp256k1_context_preallocated_size(flags)
* bytes, as detailed above.
* flags: which parts of the context to initialize.
*
* See secp256k1_context_create (in secp256k1.h) for further details.
*
* See also secp256k1_context_randomize (in secp256k1.h)
* and secp256k1_context_preallocated_destroy.
*/
SECP256K1_API secp256k1_context *secp256k1_context_preallocated_create(
void *prealloc,
unsigned int flags
) SECP256K1_ARG_NONNULL(1) SECP256K1_WARN_UNUSED_RESULT;
/** Determine the memory size of a secp256k1 context object to be copied into
* caller-provided memory.
*
* Returns: the required size of the caller-provided memory block.
* In: ctx: pointer to a context to copy.
*/
SECP256K1_API size_t secp256k1_context_preallocated_clone_size(
const secp256k1_context *ctx
) SECP256K1_ARG_NONNULL(1) SECP256K1_WARN_UNUSED_RESULT;
/** Copy a secp256k1 context object into caller-provided memory.
*
* The caller must provide a pointer to a rewritable contiguous block of memory
* of size at least secp256k1_context_preallocated_size(flags) bytes, suitably
* aligned to hold an object of any type.
*
* The block of memory is exclusively owned by the created context object during
* the lifetime of this context object, see the description of
* secp256k1_context_preallocated_create for details.
*
* Cloning secp256k1_context_static is not possible, and should not be emulated by
* the caller (e.g., using memcpy). Create a new context instead.
*
* Returns: pointer to a newly created context object.
* Args: ctx: pointer to a context to copy (not secp256k1_context_static).
* In: prealloc: pointer to a rewritable contiguous block of memory of
* size at least secp256k1_context_preallocated_size(flags)
* bytes, as detailed above.
*/
SECP256K1_API secp256k1_context *secp256k1_context_preallocated_clone(
const secp256k1_context *ctx,
void *prealloc
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_WARN_UNUSED_RESULT;
/** Destroy a secp256k1 context object that has been created in
* caller-provided memory.
*
* The context pointer may not be used afterwards.
*
* The context to destroy must have been created using
* secp256k1_context_preallocated_create or secp256k1_context_preallocated_clone.
* If the context has instead been created using secp256k1_context_create or
* secp256k1_context_clone, the behaviour is undefined. In that case,
* secp256k1_context_destroy must be used instead.
*
* If required, it is the responsibility of the caller to deallocate the block
* of memory properly after this function returns, e.g., by calling free on the
* preallocated pointer given to secp256k1_context_preallocated_create or
* secp256k1_context_preallocated_clone.
*
* Args: ctx: pointer to a context to destroy, constructed using
* secp256k1_context_preallocated_create or
* secp256k1_context_preallocated_clone
* (i.e., not secp256k1_context_static).
*/
SECP256K1_API void secp256k1_context_preallocated_destroy(
secp256k1_context *ctx
) SECP256K1_ARG_NONNULL(1);
#ifdef __cplusplus
}
#endif
#endif /* SECP256K1_PREALLOCATED_H */

View File

@@ -0,0 +1,113 @@
#ifndef SECP256K1_RECOVERY_H
#define SECP256K1_RECOVERY_H
#include "secp256k1.h"
#ifdef __cplusplus
extern "C" {
#endif
/** Opaque data structure that holds a parsed ECDSA signature,
* supporting pubkey recovery.
*
* The exact representation of data inside is implementation defined and not
* guaranteed to be portable between different platforms or versions. It is
* however guaranteed to be 65 bytes in size, and can be safely copied/moved.
* If you need to convert to a format suitable for storage or transmission, use
* the secp256k1_ecdsa_signature_serialize_* and
* secp256k1_ecdsa_signature_parse_* functions.
*
* Furthermore, it is guaranteed that identical signatures (including their
* recoverability) will have identical representation, so they can be
* memcmp'ed.
*/
typedef struct secp256k1_ecdsa_recoverable_signature {
unsigned char data[65];
} secp256k1_ecdsa_recoverable_signature;
/** Parse a compact ECDSA signature (64 bytes + recovery id).
*
* Returns: 1 when the signature could be parsed, 0 otherwise
* Args: ctx: pointer to a context object
* Out: sig: pointer to a signature object
* In: input64: pointer to a 64-byte compact signature
* recid: the recovery id (0, 1, 2 or 3)
*/
SECP256K1_API int secp256k1_ecdsa_recoverable_signature_parse_compact(
const secp256k1_context *ctx,
secp256k1_ecdsa_recoverable_signature *sig,
const unsigned char *input64,
int recid
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3);
/** Convert a recoverable signature into a normal signature.
*
* Returns: 1
* Args: ctx: pointer to a context object.
* Out: sig: pointer to a normal signature.
* In: sigin: pointer to a recoverable signature.
*/
SECP256K1_API int secp256k1_ecdsa_recoverable_signature_convert(
const secp256k1_context *ctx,
secp256k1_ecdsa_signature *sig,
const secp256k1_ecdsa_recoverable_signature *sigin
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3);
/** Serialize an ECDSA signature in compact format (64 bytes + recovery id).
*
* Returns: 1
* Args: ctx: pointer to a context object.
* Out: output64: pointer to a 64-byte array of the compact signature.
* recid: pointer to an integer to hold the recovery id.
* In: sig: pointer to an initialized signature object.
*/
SECP256K1_API int secp256k1_ecdsa_recoverable_signature_serialize_compact(
const secp256k1_context *ctx,
unsigned char *output64,
int *recid,
const secp256k1_ecdsa_recoverable_signature *sig
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3) SECP256K1_ARG_NONNULL(4);
/** Create a recoverable ECDSA signature.
*
* Returns: 1: signature created
* 0: the nonce generation function failed, or the secret key was invalid.
* Args: ctx: pointer to a context object (not secp256k1_context_static).
* Out: sig: pointer to an array where the signature will be placed.
* In: msghash32: the 32-byte message hash being signed.
* seckey: pointer to a 32-byte secret key.
* noncefp: pointer to a nonce generation function. If NULL,
* secp256k1_nonce_function_default is used.
* ndata: pointer to arbitrary data used by the nonce generation function
* (can be NULL for secp256k1_nonce_function_default).
*/
SECP256K1_API int secp256k1_ecdsa_sign_recoverable(
const secp256k1_context *ctx,
secp256k1_ecdsa_recoverable_signature *sig,
const unsigned char *msghash32,
const unsigned char *seckey,
secp256k1_nonce_function noncefp,
const void *ndata
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3) SECP256K1_ARG_NONNULL(4);
/** Recover an ECDSA public key from a signature.
*
* Returns: 1: public key successfully recovered (which guarantees a correct signature).
* 0: otherwise.
* Args: ctx: pointer to a context object.
* Out: pubkey: pointer to the recovered public key.
* In: sig: pointer to initialized signature that supports pubkey recovery.
* msghash32: the 32-byte message hash assumed to be signed.
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_ecdsa_recover(
const secp256k1_context *ctx,
secp256k1_pubkey *pubkey,
const secp256k1_ecdsa_recoverable_signature *sig,
const unsigned char *msghash32
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3) SECP256K1_ARG_NONNULL(4);
#ifdef __cplusplus
}
#endif
#endif /* SECP256K1_RECOVERY_H */

View File

@@ -0,0 +1,190 @@
#ifndef SECP256K1_SCHNORRSIG_H
#define SECP256K1_SCHNORRSIG_H
#include "secp256k1.h"
#include "secp256k1_extrakeys.h"
#ifdef __cplusplus
extern "C" {
#endif
/** This module implements a variant of Schnorr signatures compliant with
* Bitcoin Improvement Proposal 340 "Schnorr Signatures for secp256k1"
* (https://github.com/bitcoin/bips/blob/master/bip-0340.mediawiki).
*/
/** A pointer to a function to deterministically generate a nonce.
*
* Same as secp256k1_nonce function with the exception of accepting an
* additional pubkey argument and not requiring an attempt argument. The pubkey
* argument can protect signature schemes with key-prefixed challenge hash
* inputs against reusing the nonce when signing with the wrong precomputed
* pubkey.
*
* Returns: 1 if a nonce was successfully generated. 0 will cause signing to
* return an error.
* Out: nonce32: pointer to a 32-byte array to be filled by the function
* In: msg: the message being verified. Is NULL if and only if msglen
* is 0.
* msglen: the length of the message
* key32: pointer to a 32-byte secret key (will not be NULL)
* xonly_pk32: the 32-byte serialized xonly pubkey corresponding to key32
* (will not be NULL)
* algo: pointer to an array describing the signature
* algorithm (will not be NULL)
* algolen: the length of the algo array
* data: arbitrary data pointer that is passed through
*
* Except for test cases, this function should compute some cryptographic hash of
* the message, the key, the pubkey, the algorithm description, and data.
*/
typedef int (*secp256k1_nonce_function_hardened)(
unsigned char *nonce32,
const unsigned char *msg,
size_t msglen,
const unsigned char *key32,
const unsigned char *xonly_pk32,
const unsigned char *algo,
size_t algolen,
void *data
);
/** An implementation of the nonce generation function as defined in Bitcoin
* Improvement Proposal 340 "Schnorr Signatures for secp256k1"
* (https://github.com/bitcoin/bips/blob/master/bip-0340.mediawiki).
*
* If a data pointer is passed, it is assumed to be a pointer to 32 bytes of
* auxiliary random data as defined in BIP-340. If the data pointer is NULL,
* the nonce derivation procedure follows BIP-340 by setting the auxiliary
* random data to zero. The algo argument must be non-NULL, otherwise the
* function will fail and return 0. The hash will be tagged with algo.
* Therefore, to create BIP-340 compliant signatures, algo must be set to
* "BIP0340/nonce" and algolen to 13.
*/
SECP256K1_API const secp256k1_nonce_function_hardened secp256k1_nonce_function_bip340;
/** Data structure that contains additional arguments for schnorrsig_sign_custom.
*
* A schnorrsig_extraparams structure object can be initialized correctly by
* setting it to SECP256K1_SCHNORRSIG_EXTRAPARAMS_INIT.
*
* Members:
* magic: set to SECP256K1_SCHNORRSIG_EXTRAPARAMS_MAGIC at initialization
* and has no other function than making sure the object is
* initialized.
* noncefp: pointer to a nonce generation function. If NULL,
* secp256k1_nonce_function_bip340 is used
* ndata: pointer to arbitrary data used by the nonce generation function
* (can be NULL). If it is non-NULL and
* secp256k1_nonce_function_bip340 is used, then ndata must be a
* pointer to 32-byte auxiliary randomness as per BIP-340.
*/
typedef struct secp256k1_schnorrsig_extraparams {
unsigned char magic[4];
secp256k1_nonce_function_hardened noncefp;
void *ndata;
} secp256k1_schnorrsig_extraparams;
#define SECP256K1_SCHNORRSIG_EXTRAPARAMS_MAGIC { 0xda, 0x6f, 0xb3, 0x8c }
#define SECP256K1_SCHNORRSIG_EXTRAPARAMS_INIT {\
SECP256K1_SCHNORRSIG_EXTRAPARAMS_MAGIC,\
NULL,\
NULL\
}
/** Create a Schnorr signature.
*
* Does _not_ strictly follow BIP-340 because it does not verify the resulting
* signature. Instead, you can manually use secp256k1_schnorrsig_verify and
* abort if it fails.
*
* This function only signs 32-byte messages. If you have messages of a
* different size (or the same size but without a context-specific tag
* prefix), it is recommended to create a 32-byte message hash with
* secp256k1_tagged_sha256 and then sign the hash. Tagged hashing allows
* providing an context-specific tag for domain separation. This prevents
* signatures from being valid in multiple contexts by accident.
*
* Returns 1 on success, 0 on failure.
* Args: ctx: pointer to a context object (not secp256k1_context_static).
* Out: sig64: pointer to a 64-byte array to store the serialized signature.
* In: msg32: the 32-byte message being signed.
* keypair: pointer to an initialized keypair.
* aux_rand32: 32 bytes of fresh randomness. While recommended to provide
* this, it is only supplemental to security and can be NULL. A
* NULL argument is treated the same as an all-zero one. See
* BIP-340 "Default Signing" for a full explanation of this
* argument and for guidance if randomness is expensive.
*/
SECP256K1_API int secp256k1_schnorrsig_sign32(
const secp256k1_context *ctx,
unsigned char *sig64,
const unsigned char *msg32,
const secp256k1_keypair *keypair,
const unsigned char *aux_rand32
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3) SECP256K1_ARG_NONNULL(4);
/** Same as secp256k1_schnorrsig_sign32, but DEPRECATED. Will be removed in
* future versions. */
SECP256K1_API int secp256k1_schnorrsig_sign(
const secp256k1_context *ctx,
unsigned char *sig64,
const unsigned char *msg32,
const secp256k1_keypair *keypair,
const unsigned char *aux_rand32
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3) SECP256K1_ARG_NONNULL(4)
SECP256K1_DEPRECATED("Use secp256k1_schnorrsig_sign32 instead");
/** Create a Schnorr signature with a more flexible API.
*
* Same arguments as secp256k1_schnorrsig_sign except that it allows signing
* variable length messages and accepts a pointer to an extraparams object that
* allows customizing signing by passing additional arguments.
*
* Equivalent to secp256k1_schnorrsig_sign32(..., aux_rand32) if msglen is 32
* and extraparams is initialized as follows:
* ```
* secp256k1_schnorrsig_extraparams extraparams = SECP256K1_SCHNORRSIG_EXTRAPARAMS_INIT;
* extraparams.ndata = (unsigned char*)aux_rand32;
* ```
*
* Returns 1 on success, 0 on failure.
* Args: ctx: pointer to a context object (not secp256k1_context_static).
* Out: sig64: pointer to a 64-byte array to store the serialized signature.
* In: msg: the message being signed. Can only be NULL if msglen is 0.
* msglen: length of the message.
* keypair: pointer to an initialized keypair.
* extraparams: pointer to an extraparams object (can be NULL).
*/
SECP256K1_API int secp256k1_schnorrsig_sign_custom(
const secp256k1_context *ctx,
unsigned char *sig64,
const unsigned char *msg,
size_t msglen,
const secp256k1_keypair *keypair,
secp256k1_schnorrsig_extraparams *extraparams
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(5);
/** Verify a Schnorr signature.
*
* Returns: 1: correct signature
* 0: incorrect signature
* Args: ctx: pointer to a context object.
* In: sig64: pointer to the 64-byte signature to verify.
* msg: the message being verified. Can only be NULL if msglen is 0.
* msglen: length of the message
* pubkey: pointer to an x-only public key to verify with
*/
SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_schnorrsig_verify(
const secp256k1_context *ctx,
const unsigned char *sig64,
const unsigned char *msg,
size_t msglen,
const secp256k1_xonly_pubkey *pubkey
) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(5);
#ifdef __cplusplus
}
#endif
#endif /* SECP256K1_SCHNORRSIG_H */

View File

@@ -0,0 +1,916 @@
@ vim: set tabstop=8 softtabstop=8 shiftwidth=8 noexpandtab syntax=armasm:
/***********************************************************************
* Copyright (c) 2014 Wladimir J. van der Laan *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
/*
ARM implementation of field_10x26 inner loops.
Note:
- To avoid unnecessary loads and make use of available registers, two
'passes' have every time been interleaved, with the odd passes accumulating c' and d'
which will be added to c and d respectively in the even passes
*/
.syntax unified
@ eabi attributes - see readelf -A
.eabi_attribute 24, 1 @ Tag_ABI_align_needed = 8-byte
.eabi_attribute 25, 1 @ Tag_ABI_align_preserved = 8-byte, except leaf SP
.text
@ Field constants
.set field_R0, 0x3d10
.set field_R1, 0x400
.set field_not_M, 0xfc000000 @ ~M = ~0x3ffffff
.align 2
.global secp256k1_fe_mul_inner
.type secp256k1_fe_mul_inner, %function
.hidden secp256k1_fe_mul_inner
@ Arguments:
@ r0 r Restrict: can overlap with a, not with b
@ r1 a
@ r2 b
@ Stack (total 4+10*4 = 44)
@ sp + #0 saved 'r' pointer
@ sp + #4 + 4*X t0,t1,t2,t3,t4,t5,t6,t7,u8,t9
secp256k1_fe_mul_inner:
stmfd sp!, {r4, r5, r6, r7, r8, r9, r10, r11, r14}
sub sp, sp, #48 @ frame=44 + alignment
str r0, [sp, #0] @ save result address, we need it only at the end
/******************************************
* Main computation code.
******************************************
Allocation:
r0,r14,r7,r8 scratch
r1 a (pointer)
r2 b (pointer)
r3:r4 c
r5:r6 d
r11:r12 c'
r9:r10 d'
Note: do not write to r[] here, it may overlap with a[]
*/
/* A - interleaved with B */
ldr r7, [r1, #0*4] @ a[0]
ldr r8, [r2, #9*4] @ b[9]
ldr r0, [r1, #1*4] @ a[1]
umull r5, r6, r7, r8 @ d = a[0] * b[9]
ldr r14, [r2, #8*4] @ b[8]
umull r9, r10, r0, r8 @ d' = a[1] * b[9]
ldr r7, [r1, #2*4] @ a[2]
umlal r5, r6, r0, r14 @ d += a[1] * b[8]
ldr r8, [r2, #7*4] @ b[7]
umlal r9, r10, r7, r14 @ d' += a[2] * b[8]
ldr r0, [r1, #3*4] @ a[3]
umlal r5, r6, r7, r8 @ d += a[2] * b[7]
ldr r14, [r2, #6*4] @ b[6]
umlal r9, r10, r0, r8 @ d' += a[3] * b[7]
ldr r7, [r1, #4*4] @ a[4]
umlal r5, r6, r0, r14 @ d += a[3] * b[6]
ldr r8, [r2, #5*4] @ b[5]
umlal r9, r10, r7, r14 @ d' += a[4] * b[6]
ldr r0, [r1, #5*4] @ a[5]
umlal r5, r6, r7, r8 @ d += a[4] * b[5]
ldr r14, [r2, #4*4] @ b[4]
umlal r9, r10, r0, r8 @ d' += a[5] * b[5]
ldr r7, [r1, #6*4] @ a[6]
umlal r5, r6, r0, r14 @ d += a[5] * b[4]
ldr r8, [r2, #3*4] @ b[3]
umlal r9, r10, r7, r14 @ d' += a[6] * b[4]
ldr r0, [r1, #7*4] @ a[7]
umlal r5, r6, r7, r8 @ d += a[6] * b[3]
ldr r14, [r2, #2*4] @ b[2]
umlal r9, r10, r0, r8 @ d' += a[7] * b[3]
ldr r7, [r1, #8*4] @ a[8]
umlal r5, r6, r0, r14 @ d += a[7] * b[2]
ldr r8, [r2, #1*4] @ b[1]
umlal r9, r10, r7, r14 @ d' += a[8] * b[2]
ldr r0, [r1, #9*4] @ a[9]
umlal r5, r6, r7, r8 @ d += a[8] * b[1]
ldr r14, [r2, #0*4] @ b[0]
umlal r9, r10, r0, r8 @ d' += a[9] * b[1]
ldr r7, [r1, #0*4] @ a[0]
umlal r5, r6, r0, r14 @ d += a[9] * b[0]
@ r7,r14 used in B
bic r0, r5, field_not_M @ t9 = d & M
str r0, [sp, #4 + 4*9]
mov r5, r5, lsr #26 @ d >>= 26
orr r5, r5, r6, asl #6
mov r6, r6, lsr #26
/* B */
umull r3, r4, r7, r14 @ c = a[0] * b[0]
adds r5, r5, r9 @ d += d'
adc r6, r6, r10
bic r0, r5, field_not_M @ u0 = d & M
mov r5, r5, lsr #26 @ d >>= 26
orr r5, r5, r6, asl #6
mov r6, r6, lsr #26
movw r14, field_R0 @ c += u0 * R0
umlal r3, r4, r0, r14
bic r14, r3, field_not_M @ t0 = c & M
str r14, [sp, #4 + 0*4]
mov r3, r3, lsr #26 @ c >>= 26
orr r3, r3, r4, asl #6
mov r4, r4, lsr #26
mov r14, field_R1 @ c += u0 * R1
umlal r3, r4, r0, r14
/* C - interleaved with D */
ldr r7, [r1, #0*4] @ a[0]
ldr r8, [r2, #2*4] @ b[2]
ldr r14, [r2, #1*4] @ b[1]
umull r11, r12, r7, r8 @ c' = a[0] * b[2]
ldr r0, [r1, #1*4] @ a[1]
umlal r3, r4, r7, r14 @ c += a[0] * b[1]
ldr r8, [r2, #0*4] @ b[0]
umlal r11, r12, r0, r14 @ c' += a[1] * b[1]
ldr r7, [r1, #2*4] @ a[2]
umlal r3, r4, r0, r8 @ c += a[1] * b[0]
ldr r14, [r2, #9*4] @ b[9]
umlal r11, r12, r7, r8 @ c' += a[2] * b[0]
ldr r0, [r1, #3*4] @ a[3]
umlal r5, r6, r7, r14 @ d += a[2] * b[9]
ldr r8, [r2, #8*4] @ b[8]
umull r9, r10, r0, r14 @ d' = a[3] * b[9]
ldr r7, [r1, #4*4] @ a[4]
umlal r5, r6, r0, r8 @ d += a[3] * b[8]
ldr r14, [r2, #7*4] @ b[7]
umlal r9, r10, r7, r8 @ d' += a[4] * b[8]
ldr r0, [r1, #5*4] @ a[5]
umlal r5, r6, r7, r14 @ d += a[4] * b[7]
ldr r8, [r2, #6*4] @ b[6]
umlal r9, r10, r0, r14 @ d' += a[5] * b[7]
ldr r7, [r1, #6*4] @ a[6]
umlal r5, r6, r0, r8 @ d += a[5] * b[6]
ldr r14, [r2, #5*4] @ b[5]
umlal r9, r10, r7, r8 @ d' += a[6] * b[6]
ldr r0, [r1, #7*4] @ a[7]
umlal r5, r6, r7, r14 @ d += a[6] * b[5]
ldr r8, [r2, #4*4] @ b[4]
umlal r9, r10, r0, r14 @ d' += a[7] * b[5]
ldr r7, [r1, #8*4] @ a[8]
umlal r5, r6, r0, r8 @ d += a[7] * b[4]
ldr r14, [r2, #3*4] @ b[3]
umlal r9, r10, r7, r8 @ d' += a[8] * b[4]
ldr r0, [r1, #9*4] @ a[9]
umlal r5, r6, r7, r14 @ d += a[8] * b[3]
ldr r8, [r2, #2*4] @ b[2]
umlal r9, r10, r0, r14 @ d' += a[9] * b[3]
umlal r5, r6, r0, r8 @ d += a[9] * b[2]
bic r0, r5, field_not_M @ u1 = d & M
mov r5, r5, lsr #26 @ d >>= 26
orr r5, r5, r6, asl #6
mov r6, r6, lsr #26
movw r14, field_R0 @ c += u1 * R0
umlal r3, r4, r0, r14
bic r14, r3, field_not_M @ t1 = c & M
str r14, [sp, #4 + 1*4]
mov r3, r3, lsr #26 @ c >>= 26
orr r3, r3, r4, asl #6
mov r4, r4, lsr #26
mov r14, field_R1 @ c += u1 * R1
umlal r3, r4, r0, r14
/* D */
adds r3, r3, r11 @ c += c'
adc r4, r4, r12
adds r5, r5, r9 @ d += d'
adc r6, r6, r10
bic r0, r5, field_not_M @ u2 = d & M
mov r5, r5, lsr #26 @ d >>= 26
orr r5, r5, r6, asl #6
mov r6, r6, lsr #26
movw r14, field_R0 @ c += u2 * R0
umlal r3, r4, r0, r14
bic r14, r3, field_not_M @ t2 = c & M
str r14, [sp, #4 + 2*4]
mov r3, r3, lsr #26 @ c >>= 26
orr r3, r3, r4, asl #6
mov r4, r4, lsr #26
mov r14, field_R1 @ c += u2 * R1
umlal r3, r4, r0, r14
/* E - interleaved with F */
ldr r7, [r1, #0*4] @ a[0]
ldr r8, [r2, #4*4] @ b[4]
umull r11, r12, r7, r8 @ c' = a[0] * b[4]
ldr r8, [r2, #3*4] @ b[3]
umlal r3, r4, r7, r8 @ c += a[0] * b[3]
ldr r7, [r1, #1*4] @ a[1]
umlal r11, r12, r7, r8 @ c' += a[1] * b[3]
ldr r8, [r2, #2*4] @ b[2]
umlal r3, r4, r7, r8 @ c += a[1] * b[2]
ldr r7, [r1, #2*4] @ a[2]
umlal r11, r12, r7, r8 @ c' += a[2] * b[2]
ldr r8, [r2, #1*4] @ b[1]
umlal r3, r4, r7, r8 @ c += a[2] * b[1]
ldr r7, [r1, #3*4] @ a[3]
umlal r11, r12, r7, r8 @ c' += a[3] * b[1]
ldr r8, [r2, #0*4] @ b[0]
umlal r3, r4, r7, r8 @ c += a[3] * b[0]
ldr r7, [r1, #4*4] @ a[4]
umlal r11, r12, r7, r8 @ c' += a[4] * b[0]
ldr r8, [r2, #9*4] @ b[9]
umlal r5, r6, r7, r8 @ d += a[4] * b[9]
ldr r7, [r1, #5*4] @ a[5]
umull r9, r10, r7, r8 @ d' = a[5] * b[9]
ldr r8, [r2, #8*4] @ b[8]
umlal r5, r6, r7, r8 @ d += a[5] * b[8]
ldr r7, [r1, #6*4] @ a[6]
umlal r9, r10, r7, r8 @ d' += a[6] * b[8]
ldr r8, [r2, #7*4] @ b[7]
umlal r5, r6, r7, r8 @ d += a[6] * b[7]
ldr r7, [r1, #7*4] @ a[7]
umlal r9, r10, r7, r8 @ d' += a[7] * b[7]
ldr r8, [r2, #6*4] @ b[6]
umlal r5, r6, r7, r8 @ d += a[7] * b[6]
ldr r7, [r1, #8*4] @ a[8]
umlal r9, r10, r7, r8 @ d' += a[8] * b[6]
ldr r8, [r2, #5*4] @ b[5]
umlal r5, r6, r7, r8 @ d += a[8] * b[5]
ldr r7, [r1, #9*4] @ a[9]
umlal r9, r10, r7, r8 @ d' += a[9] * b[5]
ldr r8, [r2, #4*4] @ b[4]
umlal r5, r6, r7, r8 @ d += a[9] * b[4]
bic r0, r5, field_not_M @ u3 = d & M
mov r5, r5, lsr #26 @ d >>= 26
orr r5, r5, r6, asl #6
mov r6, r6, lsr #26
movw r14, field_R0 @ c += u3 * R0
umlal r3, r4, r0, r14
bic r14, r3, field_not_M @ t3 = c & M
str r14, [sp, #4 + 3*4]
mov r3, r3, lsr #26 @ c >>= 26
orr r3, r3, r4, asl #6
mov r4, r4, lsr #26
mov r14, field_R1 @ c += u3 * R1
umlal r3, r4, r0, r14
/* F */
adds r3, r3, r11 @ c += c'
adc r4, r4, r12
adds r5, r5, r9 @ d += d'
adc r6, r6, r10
bic r0, r5, field_not_M @ u4 = d & M
mov r5, r5, lsr #26 @ d >>= 26
orr r5, r5, r6, asl #6
mov r6, r6, lsr #26
movw r14, field_R0 @ c += u4 * R0
umlal r3, r4, r0, r14
bic r14, r3, field_not_M @ t4 = c & M
str r14, [sp, #4 + 4*4]
mov r3, r3, lsr #26 @ c >>= 26
orr r3, r3, r4, asl #6
mov r4, r4, lsr #26
mov r14, field_R1 @ c += u4 * R1
umlal r3, r4, r0, r14
/* G - interleaved with H */
ldr r7, [r1, #0*4] @ a[0]
ldr r8, [r2, #6*4] @ b[6]
ldr r14, [r2, #5*4] @ b[5]
umull r11, r12, r7, r8 @ c' = a[0] * b[6]
ldr r0, [r1, #1*4] @ a[1]
umlal r3, r4, r7, r14 @ c += a[0] * b[5]
ldr r8, [r2, #4*4] @ b[4]
umlal r11, r12, r0, r14 @ c' += a[1] * b[5]
ldr r7, [r1, #2*4] @ a[2]
umlal r3, r4, r0, r8 @ c += a[1] * b[4]
ldr r14, [r2, #3*4] @ b[3]
umlal r11, r12, r7, r8 @ c' += a[2] * b[4]
ldr r0, [r1, #3*4] @ a[3]
umlal r3, r4, r7, r14 @ c += a[2] * b[3]
ldr r8, [r2, #2*4] @ b[2]
umlal r11, r12, r0, r14 @ c' += a[3] * b[3]
ldr r7, [r1, #4*4] @ a[4]
umlal r3, r4, r0, r8 @ c += a[3] * b[2]
ldr r14, [r2, #1*4] @ b[1]
umlal r11, r12, r7, r8 @ c' += a[4] * b[2]
ldr r0, [r1, #5*4] @ a[5]
umlal r3, r4, r7, r14 @ c += a[4] * b[1]
ldr r8, [r2, #0*4] @ b[0]
umlal r11, r12, r0, r14 @ c' += a[5] * b[1]
ldr r7, [r1, #6*4] @ a[6]
umlal r3, r4, r0, r8 @ c += a[5] * b[0]
ldr r14, [r2, #9*4] @ b[9]
umlal r11, r12, r7, r8 @ c' += a[6] * b[0]
ldr r0, [r1, #7*4] @ a[7]
umlal r5, r6, r7, r14 @ d += a[6] * b[9]
ldr r8, [r2, #8*4] @ b[8]
umull r9, r10, r0, r14 @ d' = a[7] * b[9]
ldr r7, [r1, #8*4] @ a[8]
umlal r5, r6, r0, r8 @ d += a[7] * b[8]
ldr r14, [r2, #7*4] @ b[7]
umlal r9, r10, r7, r8 @ d' += a[8] * b[8]
ldr r0, [r1, #9*4] @ a[9]
umlal r5, r6, r7, r14 @ d += a[8] * b[7]
ldr r8, [r2, #6*4] @ b[6]
umlal r9, r10, r0, r14 @ d' += a[9] * b[7]
umlal r5, r6, r0, r8 @ d += a[9] * b[6]
bic r0, r5, field_not_M @ u5 = d & M
mov r5, r5, lsr #26 @ d >>= 26
orr r5, r5, r6, asl #6
mov r6, r6, lsr #26
movw r14, field_R0 @ c += u5 * R0
umlal r3, r4, r0, r14
bic r14, r3, field_not_M @ t5 = c & M
str r14, [sp, #4 + 5*4]
mov r3, r3, lsr #26 @ c >>= 26
orr r3, r3, r4, asl #6
mov r4, r4, lsr #26
mov r14, field_R1 @ c += u5 * R1
umlal r3, r4, r0, r14
/* H */
adds r3, r3, r11 @ c += c'
adc r4, r4, r12
adds r5, r5, r9 @ d += d'
adc r6, r6, r10
bic r0, r5, field_not_M @ u6 = d & M
mov r5, r5, lsr #26 @ d >>= 26
orr r5, r5, r6, asl #6
mov r6, r6, lsr #26
movw r14, field_R0 @ c += u6 * R0
umlal r3, r4, r0, r14
bic r14, r3, field_not_M @ t6 = c & M
str r14, [sp, #4 + 6*4]
mov r3, r3, lsr #26 @ c >>= 26
orr r3, r3, r4, asl #6
mov r4, r4, lsr #26
mov r14, field_R1 @ c += u6 * R1
umlal r3, r4, r0, r14
/* I - interleaved with J */
ldr r8, [r2, #8*4] @ b[8]
ldr r7, [r1, #0*4] @ a[0]
ldr r14, [r2, #7*4] @ b[7]
umull r11, r12, r7, r8 @ c' = a[0] * b[8]
ldr r0, [r1, #1*4] @ a[1]
umlal r3, r4, r7, r14 @ c += a[0] * b[7]
ldr r8, [r2, #6*4] @ b[6]
umlal r11, r12, r0, r14 @ c' += a[1] * b[7]
ldr r7, [r1, #2*4] @ a[2]
umlal r3, r4, r0, r8 @ c += a[1] * b[6]
ldr r14, [r2, #5*4] @ b[5]
umlal r11, r12, r7, r8 @ c' += a[2] * b[6]
ldr r0, [r1, #3*4] @ a[3]
umlal r3, r4, r7, r14 @ c += a[2] * b[5]
ldr r8, [r2, #4*4] @ b[4]
umlal r11, r12, r0, r14 @ c' += a[3] * b[5]
ldr r7, [r1, #4*4] @ a[4]
umlal r3, r4, r0, r8 @ c += a[3] * b[4]
ldr r14, [r2, #3*4] @ b[3]
umlal r11, r12, r7, r8 @ c' += a[4] * b[4]
ldr r0, [r1, #5*4] @ a[5]
umlal r3, r4, r7, r14 @ c += a[4] * b[3]
ldr r8, [r2, #2*4] @ b[2]
umlal r11, r12, r0, r14 @ c' += a[5] * b[3]
ldr r7, [r1, #6*4] @ a[6]
umlal r3, r4, r0, r8 @ c += a[5] * b[2]
ldr r14, [r2, #1*4] @ b[1]
umlal r11, r12, r7, r8 @ c' += a[6] * b[2]
ldr r0, [r1, #7*4] @ a[7]
umlal r3, r4, r7, r14 @ c += a[6] * b[1]
ldr r8, [r2, #0*4] @ b[0]
umlal r11, r12, r0, r14 @ c' += a[7] * b[1]
ldr r7, [r1, #8*4] @ a[8]
umlal r3, r4, r0, r8 @ c += a[7] * b[0]
ldr r14, [r2, #9*4] @ b[9]
umlal r11, r12, r7, r8 @ c' += a[8] * b[0]
ldr r0, [r1, #9*4] @ a[9]
umlal r5, r6, r7, r14 @ d += a[8] * b[9]
ldr r8, [r2, #8*4] @ b[8]
umull r9, r10, r0, r14 @ d' = a[9] * b[9]
umlal r5, r6, r0, r8 @ d += a[9] * b[8]
bic r0, r5, field_not_M @ u7 = d & M
mov r5, r5, lsr #26 @ d >>= 26
orr r5, r5, r6, asl #6
mov r6, r6, lsr #26
movw r14, field_R0 @ c += u7 * R0
umlal r3, r4, r0, r14
bic r14, r3, field_not_M @ t7 = c & M
str r14, [sp, #4 + 7*4]
mov r3, r3, lsr #26 @ c >>= 26
orr r3, r3, r4, asl #6
mov r4, r4, lsr #26
mov r14, field_R1 @ c += u7 * R1
umlal r3, r4, r0, r14
/* J */
adds r3, r3, r11 @ c += c'
adc r4, r4, r12
adds r5, r5, r9 @ d += d'
adc r6, r6, r10
bic r0, r5, field_not_M @ u8 = d & M
str r0, [sp, #4 + 8*4]
mov r5, r5, lsr #26 @ d >>= 26
orr r5, r5, r6, asl #6
mov r6, r6, lsr #26
movw r14, field_R0 @ c += u8 * R0
umlal r3, r4, r0, r14
/******************************************
* compute and write back result
******************************************
Allocation:
r0 r
r3:r4 c
r5:r6 d
r7 t0
r8 t1
r9 t2
r11 u8
r12 t9
r1,r2,r10,r14 scratch
Note: do not read from a[] after here, it may overlap with r[]
*/
ldr r0, [sp, #0]
add r1, sp, #4 + 3*4 @ r[3..7] = t3..7, r11=u8, r12=t9
ldmia r1, {r2,r7,r8,r9,r10,r11,r12}
add r1, r0, #3*4
stmia r1, {r2,r7,r8,r9,r10}
bic r2, r3, field_not_M @ r[8] = c & M
str r2, [r0, #8*4]
mov r3, r3, lsr #26 @ c >>= 26
orr r3, r3, r4, asl #6
mov r4, r4, lsr #26
mov r14, field_R1 @ c += u8 * R1
umlal r3, r4, r11, r14
movw r14, field_R0 @ c += d * R0
umlal r3, r4, r5, r14
adds r3, r3, r12 @ c += t9
adc r4, r4, #0
add r1, sp, #4 + 0*4 @ r7,r8,r9 = t0,t1,t2
ldmia r1, {r7,r8,r9}
ubfx r2, r3, #0, #22 @ r[9] = c & (M >> 4)
str r2, [r0, #9*4]
mov r3, r3, lsr #22 @ c >>= 22
orr r3, r3, r4, asl #10
mov r4, r4, lsr #22
movw r14, field_R1 << 4 @ c += d * (R1 << 4)
umlal r3, r4, r5, r14
movw r14, field_R0 >> 4 @ d = c * (R0 >> 4) + t0 (64x64 multiply+add)
umull r5, r6, r3, r14 @ d = c.lo * (R0 >> 4)
adds r5, r5, r7 @ d.lo += t0
mla r6, r14, r4, r6 @ d.hi += c.hi * (R0 >> 4)
adc r6, r6, 0 @ d.hi += carry
bic r2, r5, field_not_M @ r[0] = d & M
str r2, [r0, #0*4]
mov r5, r5, lsr #26 @ d >>= 26
orr r5, r5, r6, asl #6
mov r6, r6, lsr #26
movw r14, field_R1 >> 4 @ d += c * (R1 >> 4) + t1 (64x64 multiply+add)
umull r1, r2, r3, r14 @ tmp = c.lo * (R1 >> 4)
adds r5, r5, r8 @ d.lo += t1
adc r6, r6, #0 @ d.hi += carry
adds r5, r5, r1 @ d.lo += tmp.lo
mla r2, r14, r4, r2 @ tmp.hi += c.hi * (R1 >> 4)
adc r6, r6, r2 @ d.hi += carry + tmp.hi
bic r2, r5, field_not_M @ r[1] = d & M
str r2, [r0, #1*4]
mov r5, r5, lsr #26 @ d >>= 26 (ignore hi)
orr r5, r5, r6, asl #6
add r5, r5, r9 @ d += t2
str r5, [r0, #2*4] @ r[2] = d
add sp, sp, #48
ldmfd sp!, {r4, r5, r6, r7, r8, r9, r10, r11, pc}
.size secp256k1_fe_mul_inner, .-secp256k1_fe_mul_inner
.align 2
.global secp256k1_fe_sqr_inner
.type secp256k1_fe_sqr_inner, %function
.hidden secp256k1_fe_sqr_inner
@ Arguments:
@ r0 r Can overlap with a
@ r1 a
@ Stack (total 4+10*4 = 44)
@ sp + #0 saved 'r' pointer
@ sp + #4 + 4*X t0,t1,t2,t3,t4,t5,t6,t7,u8,t9
secp256k1_fe_sqr_inner:
stmfd sp!, {r4, r5, r6, r7, r8, r9, r10, r11, r14}
sub sp, sp, #48 @ frame=44 + alignment
str r0, [sp, #0] @ save result address, we need it only at the end
/******************************************
* Main computation code.
******************************************
Allocation:
r0,r14,r2,r7,r8 scratch
r1 a (pointer)
r3:r4 c
r5:r6 d
r11:r12 c'
r9:r10 d'
Note: do not write to r[] here, it may overlap with a[]
*/
/* A interleaved with B */
ldr r0, [r1, #1*4] @ a[1]*2
ldr r7, [r1, #0*4] @ a[0]
mov r0, r0, asl #1
ldr r14, [r1, #9*4] @ a[9]
umull r3, r4, r7, r7 @ c = a[0] * a[0]
ldr r8, [r1, #8*4] @ a[8]
mov r7, r7, asl #1
umull r5, r6, r7, r14 @ d = a[0]*2 * a[9]
ldr r7, [r1, #2*4] @ a[2]*2
umull r9, r10, r0, r14 @ d' = a[1]*2 * a[9]
ldr r14, [r1, #7*4] @ a[7]
umlal r5, r6, r0, r8 @ d += a[1]*2 * a[8]
mov r7, r7, asl #1
ldr r0, [r1, #3*4] @ a[3]*2
umlal r9, r10, r7, r8 @ d' += a[2]*2 * a[8]
ldr r8, [r1, #6*4] @ a[6]
umlal r5, r6, r7, r14 @ d += a[2]*2 * a[7]
mov r0, r0, asl #1
ldr r7, [r1, #4*4] @ a[4]*2
umlal r9, r10, r0, r14 @ d' += a[3]*2 * a[7]
ldr r14, [r1, #5*4] @ a[5]
mov r7, r7, asl #1
umlal r5, r6, r0, r8 @ d += a[3]*2 * a[6]
umlal r9, r10, r7, r8 @ d' += a[4]*2 * a[6]
umlal r5, r6, r7, r14 @ d += a[4]*2 * a[5]
umlal r9, r10, r14, r14 @ d' += a[5] * a[5]
bic r0, r5, field_not_M @ t9 = d & M
str r0, [sp, #4 + 9*4]
mov r5, r5, lsr #26 @ d >>= 26
orr r5, r5, r6, asl #6
mov r6, r6, lsr #26
/* B */
adds r5, r5, r9 @ d += d'
adc r6, r6, r10
bic r0, r5, field_not_M @ u0 = d & M
mov r5, r5, lsr #26 @ d >>= 26
orr r5, r5, r6, asl #6
mov r6, r6, lsr #26
movw r14, field_R0 @ c += u0 * R0
umlal r3, r4, r0, r14
bic r14, r3, field_not_M @ t0 = c & M
str r14, [sp, #4 + 0*4]
mov r3, r3, lsr #26 @ c >>= 26
orr r3, r3, r4, asl #6
mov r4, r4, lsr #26
mov r14, field_R1 @ c += u0 * R1
umlal r3, r4, r0, r14
/* C interleaved with D */
ldr r0, [r1, #0*4] @ a[0]*2
ldr r14, [r1, #1*4] @ a[1]
mov r0, r0, asl #1
ldr r8, [r1, #2*4] @ a[2]
umlal r3, r4, r0, r14 @ c += a[0]*2 * a[1]
mov r7, r8, asl #1 @ a[2]*2
umull r11, r12, r14, r14 @ c' = a[1] * a[1]
ldr r14, [r1, #9*4] @ a[9]
umlal r11, r12, r0, r8 @ c' += a[0]*2 * a[2]
ldr r0, [r1, #3*4] @ a[3]*2
ldr r8, [r1, #8*4] @ a[8]
umlal r5, r6, r7, r14 @ d += a[2]*2 * a[9]
mov r0, r0, asl #1
ldr r7, [r1, #4*4] @ a[4]*2
umull r9, r10, r0, r14 @ d' = a[3]*2 * a[9]
ldr r14, [r1, #7*4] @ a[7]
umlal r5, r6, r0, r8 @ d += a[3]*2 * a[8]
mov r7, r7, asl #1
ldr r0, [r1, #5*4] @ a[5]*2
umlal r9, r10, r7, r8 @ d' += a[4]*2 * a[8]
ldr r8, [r1, #6*4] @ a[6]
mov r0, r0, asl #1
umlal r5, r6, r7, r14 @ d += a[4]*2 * a[7]
umlal r9, r10, r0, r14 @ d' += a[5]*2 * a[7]
umlal r5, r6, r0, r8 @ d += a[5]*2 * a[6]
umlal r9, r10, r8, r8 @ d' += a[6] * a[6]
bic r0, r5, field_not_M @ u1 = d & M
mov r5, r5, lsr #26 @ d >>= 26
orr r5, r5, r6, asl #6
mov r6, r6, lsr #26
movw r14, field_R0 @ c += u1 * R0
umlal r3, r4, r0, r14
bic r14, r3, field_not_M @ t1 = c & M
str r14, [sp, #4 + 1*4]
mov r3, r3, lsr #26 @ c >>= 26
orr r3, r3, r4, asl #6
mov r4, r4, lsr #26
mov r14, field_R1 @ c += u1 * R1
umlal r3, r4, r0, r14
/* D */
adds r3, r3, r11 @ c += c'
adc r4, r4, r12
adds r5, r5, r9 @ d += d'
adc r6, r6, r10
bic r0, r5, field_not_M @ u2 = d & M
mov r5, r5, lsr #26 @ d >>= 26
orr r5, r5, r6, asl #6
mov r6, r6, lsr #26
movw r14, field_R0 @ c += u2 * R0
umlal r3, r4, r0, r14
bic r14, r3, field_not_M @ t2 = c & M
str r14, [sp, #4 + 2*4]
mov r3, r3, lsr #26 @ c >>= 26
orr r3, r3, r4, asl #6
mov r4, r4, lsr #26
mov r14, field_R1 @ c += u2 * R1
umlal r3, r4, r0, r14
/* E interleaved with F */
ldr r7, [r1, #0*4] @ a[0]*2
ldr r0, [r1, #1*4] @ a[1]*2
ldr r14, [r1, #2*4] @ a[2]
mov r7, r7, asl #1
ldr r8, [r1, #3*4] @ a[3]
ldr r2, [r1, #4*4]
umlal r3, r4, r7, r8 @ c += a[0]*2 * a[3]
mov r0, r0, asl #1
umull r11, r12, r7, r2 @ c' = a[0]*2 * a[4]
mov r2, r2, asl #1 @ a[4]*2
umlal r11, r12, r0, r8 @ c' += a[1]*2 * a[3]
ldr r8, [r1, #9*4] @ a[9]
umlal r3, r4, r0, r14 @ c += a[1]*2 * a[2]
ldr r0, [r1, #5*4] @ a[5]*2
umlal r11, r12, r14, r14 @ c' += a[2] * a[2]
ldr r14, [r1, #8*4] @ a[8]
mov r0, r0, asl #1
umlal r5, r6, r2, r8 @ d += a[4]*2 * a[9]
ldr r7, [r1, #6*4] @ a[6]*2
umull r9, r10, r0, r8 @ d' = a[5]*2 * a[9]
mov r7, r7, asl #1
ldr r8, [r1, #7*4] @ a[7]
umlal r5, r6, r0, r14 @ d += a[5]*2 * a[8]
umlal r9, r10, r7, r14 @ d' += a[6]*2 * a[8]
umlal r5, r6, r7, r8 @ d += a[6]*2 * a[7]
umlal r9, r10, r8, r8 @ d' += a[7] * a[7]
bic r0, r5, field_not_M @ u3 = d & M
mov r5, r5, lsr #26 @ d >>= 26
orr r5, r5, r6, asl #6
mov r6, r6, lsr #26
movw r14, field_R0 @ c += u3 * R0
umlal r3, r4, r0, r14
bic r14, r3, field_not_M @ t3 = c & M
str r14, [sp, #4 + 3*4]
mov r3, r3, lsr #26 @ c >>= 26
orr r3, r3, r4, asl #6
mov r4, r4, lsr #26
mov r14, field_R1 @ c += u3 * R1
umlal r3, r4, r0, r14
/* F */
adds r3, r3, r11 @ c += c'
adc r4, r4, r12
adds r5, r5, r9 @ d += d'
adc r6, r6, r10
bic r0, r5, field_not_M @ u4 = d & M
mov r5, r5, lsr #26 @ d >>= 26
orr r5, r5, r6, asl #6
mov r6, r6, lsr #26
movw r14, field_R0 @ c += u4 * R0
umlal r3, r4, r0, r14
bic r14, r3, field_not_M @ t4 = c & M
str r14, [sp, #4 + 4*4]
mov r3, r3, lsr #26 @ c >>= 26
orr r3, r3, r4, asl #6
mov r4, r4, lsr #26
mov r14, field_R1 @ c += u4 * R1
umlal r3, r4, r0, r14
/* G interleaved with H */
ldr r7, [r1, #0*4] @ a[0]*2
ldr r0, [r1, #1*4] @ a[1]*2
mov r7, r7, asl #1
ldr r8, [r1, #5*4] @ a[5]
ldr r2, [r1, #6*4] @ a[6]
umlal r3, r4, r7, r8 @ c += a[0]*2 * a[5]
ldr r14, [r1, #4*4] @ a[4]
mov r0, r0, asl #1
umull r11, r12, r7, r2 @ c' = a[0]*2 * a[6]
ldr r7, [r1, #2*4] @ a[2]*2
umlal r11, r12, r0, r8 @ c' += a[1]*2 * a[5]
mov r7, r7, asl #1
ldr r8, [r1, #3*4] @ a[3]
umlal r3, r4, r0, r14 @ c += a[1]*2 * a[4]
mov r0, r2, asl #1 @ a[6]*2
umlal r11, r12, r7, r14 @ c' += a[2]*2 * a[4]
ldr r14, [r1, #9*4] @ a[9]
umlal r3, r4, r7, r8 @ c += a[2]*2 * a[3]
ldr r7, [r1, #7*4] @ a[7]*2
umlal r11, r12, r8, r8 @ c' += a[3] * a[3]
mov r7, r7, asl #1
ldr r8, [r1, #8*4] @ a[8]
umlal r5, r6, r0, r14 @ d += a[6]*2 * a[9]
umull r9, r10, r7, r14 @ d' = a[7]*2 * a[9]
umlal r5, r6, r7, r8 @ d += a[7]*2 * a[8]
umlal r9, r10, r8, r8 @ d' += a[8] * a[8]
bic r0, r5, field_not_M @ u5 = d & M
mov r5, r5, lsr #26 @ d >>= 26
orr r5, r5, r6, asl #6
mov r6, r6, lsr #26
movw r14, field_R0 @ c += u5 * R0
umlal r3, r4, r0, r14
bic r14, r3, field_not_M @ t5 = c & M
str r14, [sp, #4 + 5*4]
mov r3, r3, lsr #26 @ c >>= 26
orr r3, r3, r4, asl #6
mov r4, r4, lsr #26
mov r14, field_R1 @ c += u5 * R1
umlal r3, r4, r0, r14
/* H */
adds r3, r3, r11 @ c += c'
adc r4, r4, r12
adds r5, r5, r9 @ d += d'
adc r6, r6, r10
bic r0, r5, field_not_M @ u6 = d & M
mov r5, r5, lsr #26 @ d >>= 26
orr r5, r5, r6, asl #6
mov r6, r6, lsr #26
movw r14, field_R0 @ c += u6 * R0
umlal r3, r4, r0, r14
bic r14, r3, field_not_M @ t6 = c & M
str r14, [sp, #4 + 6*4]
mov r3, r3, lsr #26 @ c >>= 26
orr r3, r3, r4, asl #6
mov r4, r4, lsr #26
mov r14, field_R1 @ c += u6 * R1
umlal r3, r4, r0, r14
/* I interleaved with J */
ldr r7, [r1, #0*4] @ a[0]*2
ldr r0, [r1, #1*4] @ a[1]*2
mov r7, r7, asl #1
ldr r8, [r1, #7*4] @ a[7]
ldr r2, [r1, #8*4] @ a[8]
umlal r3, r4, r7, r8 @ c += a[0]*2 * a[7]
ldr r14, [r1, #6*4] @ a[6]
mov r0, r0, asl #1
umull r11, r12, r7, r2 @ c' = a[0]*2 * a[8]
ldr r7, [r1, #2*4] @ a[2]*2
umlal r11, r12, r0, r8 @ c' += a[1]*2 * a[7]
ldr r8, [r1, #5*4] @ a[5]
umlal r3, r4, r0, r14 @ c += a[1]*2 * a[6]
ldr r0, [r1, #3*4] @ a[3]*2
mov r7, r7, asl #1
umlal r11, r12, r7, r14 @ c' += a[2]*2 * a[6]
ldr r14, [r1, #4*4] @ a[4]
mov r0, r0, asl #1
umlal r3, r4, r7, r8 @ c += a[2]*2 * a[5]
mov r2, r2, asl #1 @ a[8]*2
umlal r11, r12, r0, r8 @ c' += a[3]*2 * a[5]
umlal r3, r4, r0, r14 @ c += a[3]*2 * a[4]
umlal r11, r12, r14, r14 @ c' += a[4] * a[4]
ldr r8, [r1, #9*4] @ a[9]
umlal r5, r6, r2, r8 @ d += a[8]*2 * a[9]
@ r8 will be used in J
bic r0, r5, field_not_M @ u7 = d & M
mov r5, r5, lsr #26 @ d >>= 26
orr r5, r5, r6, asl #6
mov r6, r6, lsr #26
movw r14, field_R0 @ c += u7 * R0
umlal r3, r4, r0, r14
bic r14, r3, field_not_M @ t7 = c & M
str r14, [sp, #4 + 7*4]
mov r3, r3, lsr #26 @ c >>= 26
orr r3, r3, r4, asl #6
mov r4, r4, lsr #26
mov r14, field_R1 @ c += u7 * R1
umlal r3, r4, r0, r14
/* J */
adds r3, r3, r11 @ c += c'
adc r4, r4, r12
umlal r5, r6, r8, r8 @ d += a[9] * a[9]
bic r0, r5, field_not_M @ u8 = d & M
str r0, [sp, #4 + 8*4]
mov r5, r5, lsr #26 @ d >>= 26
orr r5, r5, r6, asl #6
mov r6, r6, lsr #26
movw r14, field_R0 @ c += u8 * R0
umlal r3, r4, r0, r14
/******************************************
* compute and write back result
******************************************
Allocation:
r0 r
r3:r4 c
r5:r6 d
r7 t0
r8 t1
r9 t2
r11 u8
r12 t9
r1,r2,r10,r14 scratch
Note: do not read from a[] after here, it may overlap with r[]
*/
ldr r0, [sp, #0]
add r1, sp, #4 + 3*4 @ r[3..7] = t3..7, r11=u8, r12=t9
ldmia r1, {r2,r7,r8,r9,r10,r11,r12}
add r1, r0, #3*4
stmia r1, {r2,r7,r8,r9,r10}
bic r2, r3, field_not_M @ r[8] = c & M
str r2, [r0, #8*4]
mov r3, r3, lsr #26 @ c >>= 26
orr r3, r3, r4, asl #6
mov r4, r4, lsr #26
mov r14, field_R1 @ c += u8 * R1
umlal r3, r4, r11, r14
movw r14, field_R0 @ c += d * R0
umlal r3, r4, r5, r14
adds r3, r3, r12 @ c += t9
adc r4, r4, #0
add r1, sp, #4 + 0*4 @ r7,r8,r9 = t0,t1,t2
ldmia r1, {r7,r8,r9}
ubfx r2, r3, #0, #22 @ r[9] = c & (M >> 4)
str r2, [r0, #9*4]
mov r3, r3, lsr #22 @ c >>= 22
orr r3, r3, r4, asl #10
mov r4, r4, lsr #22
movw r14, field_R1 << 4 @ c += d * (R1 << 4)
umlal r3, r4, r5, r14
movw r14, field_R0 >> 4 @ d = c * (R0 >> 4) + t0 (64x64 multiply+add)
umull r5, r6, r3, r14 @ d = c.lo * (R0 >> 4)
adds r5, r5, r7 @ d.lo += t0
mla r6, r14, r4, r6 @ d.hi += c.hi * (R0 >> 4)
adc r6, r6, 0 @ d.hi += carry
bic r2, r5, field_not_M @ r[0] = d & M
str r2, [r0, #0*4]
mov r5, r5, lsr #26 @ d >>= 26
orr r5, r5, r6, asl #6
mov r6, r6, lsr #26
movw r14, field_R1 >> 4 @ d += c * (R1 >> 4) + t1 (64x64 multiply+add)
umull r1, r2, r3, r14 @ tmp = c.lo * (R1 >> 4)
adds r5, r5, r8 @ d.lo += t1
adc r6, r6, #0 @ d.hi += carry
adds r5, r5, r1 @ d.lo += tmp.lo
mla r2, r14, r4, r2 @ tmp.hi += c.hi * (R1 >> 4)
adc r6, r6, r2 @ d.hi += carry + tmp.hi
bic r2, r5, field_not_M @ r[1] = d & M
str r2, [r0, #1*4]
mov r5, r5, lsr #26 @ d >>= 26 (ignore hi)
orr r5, r5, r6, asl #6
add r5, r5, r9 @ d += t2
str r5, [r0, #2*4] @ r[2] = d
add sp, sp, #48
ldmfd sp!, {r4, r5, r6, r7, r8, r9, r10, r11, pc}
.size secp256k1_fe_sqr_inner, .-secp256k1_fe_sqr_inner
.section .note.GNU-stack,"",%progbits

View File

@@ -0,0 +1,87 @@
/***********************************************************************
* Copyright (c) 2020 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_ASSUMPTIONS_H
#define SECP256K1_ASSUMPTIONS_H
#include <limits.h>
#include "util.h"
#if defined(SECP256K1_INT128_NATIVE)
#include "int128_native.h"
#endif
/* This library, like most software, relies on a number of compiler implementation defined (but not undefined)
behaviours. Although the behaviours we require are essentially universal we test them specifically here to
reduce the odds of experiencing an unwelcome surprise.
*/
#if defined(__has_attribute)
# if __has_attribute(__unavailable__)
__attribute__((__unavailable__("Don't call this function. It only exists because STATIC_ASSERT cannot be used outside a function.")))
# endif
#endif
static void secp256k1_assumption_checker(void) {
/* Bytes are 8 bits. */
STATIC_ASSERT(CHAR_BIT == 8);
/* No integer promotion for uint32_t. This ensures that we can multiply uintXX_t values where XX >= 32
without signed overflow, which would be undefined behaviour. */
STATIC_ASSERT(UINT_MAX <= UINT32_MAX);
/* Conversions from unsigned to signed outside of the bounds of the signed type are
implementation-defined. Verify that they function as reinterpreting the lower
bits of the input in two's complement notation. Do this for conversions:
- from uint(N)_t to int(N)_t with negative result
- from uint(2N)_t to int(N)_t with negative result
- from int(2N)_t to int(N)_t with negative result
- from int(2N)_t to int(N)_t with positive result */
/* To int8_t. */
STATIC_ASSERT(((int8_t)(uint8_t)0xAB == (int8_t)-(int8_t)0x55));
STATIC_ASSERT((int8_t)(uint16_t)0xABCD == (int8_t)-(int8_t)0x33);
STATIC_ASSERT((int8_t)(int16_t)(uint16_t)0xCDEF == (int8_t)(uint8_t)0xEF);
STATIC_ASSERT((int8_t)(int16_t)(uint16_t)0x9234 == (int8_t)(uint8_t)0x34);
/* To int16_t. */
STATIC_ASSERT((int16_t)(uint16_t)0xBCDE == (int16_t)-(int16_t)0x4322);
STATIC_ASSERT((int16_t)(uint32_t)0xA1B2C3D4 == (int16_t)-(int16_t)0x3C2C);
STATIC_ASSERT((int16_t)(int32_t)(uint32_t)0xC1D2E3F4 == (int16_t)(uint16_t)0xE3F4);
STATIC_ASSERT((int16_t)(int32_t)(uint32_t)0x92345678 == (int16_t)(uint16_t)0x5678);
/* To int32_t. */
STATIC_ASSERT((int32_t)(uint32_t)0xB2C3D4E5 == (int32_t)-(int32_t)0x4D3C2B1B);
STATIC_ASSERT((int32_t)(uint64_t)0xA123B456C789D012ULL == (int32_t)-(int32_t)0x38762FEE);
STATIC_ASSERT((int32_t)(int64_t)(uint64_t)0xC1D2E3F4A5B6C7D8ULL == (int32_t)(uint32_t)0xA5B6C7D8);
STATIC_ASSERT((int32_t)(int64_t)(uint64_t)0xABCDEF0123456789ULL == (int32_t)(uint32_t)0x23456789);
/* To int64_t. */
STATIC_ASSERT((int64_t)(uint64_t)0xB123C456D789E012ULL == (int64_t)-(int64_t)0x4EDC3BA928761FEEULL);
#if defined(SECP256K1_INT128_NATIVE)
STATIC_ASSERT((int64_t)(((uint128_t)0xA1234567B8901234ULL << 64) + 0xC5678901D2345678ULL) == (int64_t)-(int64_t)0x3A9876FE2DCBA988ULL);
STATIC_ASSERT(((int64_t)(int128_t)(((uint128_t)0xB1C2D3E4F5A6B7C8ULL << 64) + 0xD9E0F1A2B3C4D5E6ULL)) == (int64_t)(uint64_t)0xD9E0F1A2B3C4D5E6ULL);
STATIC_ASSERT(((int64_t)(int128_t)(((uint128_t)0xABCDEF0123456789ULL << 64) + 0x0123456789ABCDEFULL)) == (int64_t)(uint64_t)0x0123456789ABCDEFULL);
/* To int128_t. */
STATIC_ASSERT((int128_t)(((uint128_t)0xB1234567C8901234ULL << 64) + 0xD5678901E2345678ULL) == (int128_t)(-(int128_t)0x8E1648B3F50E80DCULL * 0x8E1648B3F50E80DDULL + 0x5EA688D5482F9464ULL));
#endif
/* Right shift on negative signed values is implementation defined. Verify that it
acts as a right shift in two's complement with sign extension (i.e duplicating
the top bit into newly added bits). */
STATIC_ASSERT((((int8_t)0xE8) >> 2) == (int8_t)(uint8_t)0xFA);
STATIC_ASSERT((((int16_t)0xE9AC) >> 4) == (int16_t)(uint16_t)0xFE9A);
STATIC_ASSERT((((int32_t)0x937C918A) >> 9) == (int32_t)(uint32_t)0xFFC9BE48);
STATIC_ASSERT((((int64_t)0xA8B72231DF9CF4B9ULL) >> 19) == (int64_t)(uint64_t)0xFFFFF516E4463BF3ULL);
#if defined(SECP256K1_INT128_NATIVE)
STATIC_ASSERT((((int128_t)(((uint128_t)0xCD833A65684A0DBCULL << 64) + 0xB349312F71EA7637ULL)) >> 39) == (int128_t)(((uint128_t)0xFFFFFFFFFF9B0674ULL << 64) + 0xCAD0941B79669262ULL));
#endif
/* This function is not supposed to be called. */
VERIFY_CHECK(0);
}
#endif /* SECP256K1_ASSUMPTIONS_H */

279
libsecp256k1/src/bench.c Normal file
View File

@@ -0,0 +1,279 @@
/***********************************************************************
* Copyright (c) 2014 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#include <stdio.h>
#include <string.h>
#include "../include/secp256k1.h"
#include "util.h"
#include "bench.h"
static void help(int default_iters) {
printf("Benchmarks the following algorithms:\n");
printf(" - ECDSA signing/verification\n");
#ifdef ENABLE_MODULE_ECDH
printf(" - ECDH key exchange (optional module)\n");
#endif
#ifdef ENABLE_MODULE_RECOVERY
printf(" - Public key recovery (optional module)\n");
#endif
#ifdef ENABLE_MODULE_SCHNORRSIG
printf(" - Schnorr signatures (optional module)\n");
#endif
printf("\n");
printf("The default number of iterations for each benchmark is %d. This can be\n", default_iters);
printf("customized using the SECP256K1_BENCH_ITERS environment variable.\n");
printf("\n");
printf("Usage: ./bench [args]\n");
printf("By default, all benchmarks will be run.\n");
printf("args:\n");
printf(" help : display this help and exit\n");
printf(" ecdsa : all ECDSA algorithms--sign, verify, recovery (if enabled)\n");
printf(" ecdsa_sign : ECDSA siging algorithm\n");
printf(" ecdsa_verify : ECDSA verification algorithm\n");
printf(" ec : all EC public key algorithms (keygen)\n");
printf(" ec_keygen : EC public key generation\n");
#ifdef ENABLE_MODULE_RECOVERY
printf(" ecdsa_recover : ECDSA public key recovery algorithm\n");
#endif
#ifdef ENABLE_MODULE_ECDH
printf(" ecdh : ECDH key exchange algorithm\n");
#endif
#ifdef ENABLE_MODULE_SCHNORRSIG
printf(" schnorrsig : all Schnorr signature algorithms (sign, verify)\n");
printf(" schnorrsig_sign : Schnorr sigining algorithm\n");
printf(" schnorrsig_verify : Schnorr verification algorithm\n");
#endif
#ifdef ENABLE_MODULE_ELLSWIFT
printf(" ellswift : all ElligatorSwift benchmarks (encode, decode, keygen, ecdh)\n");
printf(" ellswift_encode : ElligatorSwift encoding\n");
printf(" ellswift_decode : ElligatorSwift decoding\n");
printf(" ellswift_keygen : ElligatorSwift key generation\n");
printf(" ellswift_ecdh : ECDH on ElligatorSwift keys\n");
#endif
printf("\n");
}
typedef struct {
secp256k1_context *ctx;
unsigned char msg[32];
unsigned char key[32];
unsigned char sig[72];
size_t siglen;
unsigned char pubkey[33];
size_t pubkeylen;
} bench_data;
static void bench_verify(void* arg, int iters) {
int i;
bench_data* data = (bench_data*)arg;
for (i = 0; i < iters; i++) {
secp256k1_pubkey pubkey;
secp256k1_ecdsa_signature sig;
data->sig[data->siglen - 1] ^= (i & 0xFF);
data->sig[data->siglen - 2] ^= ((i >> 8) & 0xFF);
data->sig[data->siglen - 3] ^= ((i >> 16) & 0xFF);
CHECK(secp256k1_ec_pubkey_parse(data->ctx, &pubkey, data->pubkey, data->pubkeylen) == 1);
CHECK(secp256k1_ecdsa_signature_parse_der(data->ctx, &sig, data->sig, data->siglen) == 1);
CHECK(secp256k1_ecdsa_verify(data->ctx, &sig, data->msg, &pubkey) == (i == 0));
data->sig[data->siglen - 1] ^= (i & 0xFF);
data->sig[data->siglen - 2] ^= ((i >> 8) & 0xFF);
data->sig[data->siglen - 3] ^= ((i >> 16) & 0xFF);
}
}
static void bench_sign_setup(void* arg) {
int i;
bench_data *data = (bench_data*)arg;
for (i = 0; i < 32; i++) {
data->msg[i] = i + 1;
}
for (i = 0; i < 32; i++) {
data->key[i] = i + 65;
}
}
static void bench_sign_run(void* arg, int iters) {
int i;
bench_data *data = (bench_data*)arg;
unsigned char sig[74];
for (i = 0; i < iters; i++) {
size_t siglen = 74;
int j;
secp256k1_ecdsa_signature signature;
CHECK(secp256k1_ecdsa_sign(data->ctx, &signature, data->msg, data->key, NULL, NULL));
CHECK(secp256k1_ecdsa_signature_serialize_der(data->ctx, sig, &siglen, &signature));
for (j = 0; j < 32; j++) {
data->msg[j] = sig[j];
data->key[j] = sig[j + 32];
}
}
}
static void bench_keygen_setup(void* arg) {
int i;
bench_data *data = (bench_data*)arg;
for (i = 0; i < 32; i++) {
data->key[i] = i + 65;
}
}
static void bench_keygen_run(void *arg, int iters) {
int i;
bench_data *data = (bench_data*)arg;
for (i = 0; i < iters; i++) {
unsigned char pub33[33];
size_t len = 33;
secp256k1_pubkey pubkey;
CHECK(secp256k1_ec_pubkey_create(data->ctx, &pubkey, data->key));
CHECK(secp256k1_ec_pubkey_serialize(data->ctx, pub33, &len, &pubkey, SECP256K1_EC_COMPRESSED));
memcpy(data->key, pub33 + 1, 32);
}
}
#ifdef ENABLE_MODULE_ECDH
# include "modules/ecdh/bench_impl.h"
#endif
#ifdef ENABLE_MODULE_RECOVERY
# include "modules/recovery/bench_impl.h"
#endif
#ifdef ENABLE_MODULE_SCHNORRSIG
# include "modules/schnorrsig/bench_impl.h"
#endif
#ifdef ENABLE_MODULE_ELLSWIFT
# include "modules/ellswift/bench_impl.h"
#endif
int main(int argc, char** argv) {
int i;
secp256k1_pubkey pubkey;
secp256k1_ecdsa_signature sig;
bench_data data;
int d = argc == 1;
int default_iters = 20000;
int iters = get_iters(default_iters);
/* Check for invalid user arguments */
char* valid_args[] = {"ecdsa", "verify", "ecdsa_verify", "sign", "ecdsa_sign", "ecdh", "recover",
"ecdsa_recover", "schnorrsig", "schnorrsig_verify", "schnorrsig_sign", "ec",
"keygen", "ec_keygen", "ellswift", "encode", "ellswift_encode", "decode",
"ellswift_decode", "ellswift_keygen", "ellswift_ecdh"};
size_t valid_args_size = sizeof(valid_args)/sizeof(valid_args[0]);
int invalid_args = have_invalid_args(argc, argv, valid_args, valid_args_size);
if (argc > 1) {
if (have_flag(argc, argv, "-h")
|| have_flag(argc, argv, "--help")
|| have_flag(argc, argv, "help")) {
help(default_iters);
return 0;
} else if (invalid_args) {
fprintf(stderr, "./bench: unrecognized argument.\n\n");
help(default_iters);
return 1;
}
}
/* Check if the user tries to benchmark optional module without building it */
#ifndef ENABLE_MODULE_ECDH
if (have_flag(argc, argv, "ecdh")) {
fprintf(stderr, "./bench: ECDH module not enabled.\n");
fprintf(stderr, "Use ./configure --enable-module-ecdh.\n\n");
return 1;
}
#endif
#ifndef ENABLE_MODULE_RECOVERY
if (have_flag(argc, argv, "recover") || have_flag(argc, argv, "ecdsa_recover")) {
fprintf(stderr, "./bench: Public key recovery module not enabled.\n");
fprintf(stderr, "Use ./configure --enable-module-recovery.\n\n");
return 1;
}
#endif
#ifndef ENABLE_MODULE_SCHNORRSIG
if (have_flag(argc, argv, "schnorrsig") || have_flag(argc, argv, "schnorrsig_sign") || have_flag(argc, argv, "schnorrsig_verify")) {
fprintf(stderr, "./bench: Schnorr signatures module not enabled.\n");
fprintf(stderr, "Use ./configure --enable-module-schnorrsig.\n\n");
return 1;
}
#endif
#ifndef ENABLE_MODULE_ELLSWIFT
if (have_flag(argc, argv, "ellswift") || have_flag(argc, argv, "ellswift_encode") || have_flag(argc, argv, "ellswift_decode") ||
have_flag(argc, argv, "encode") || have_flag(argc, argv, "decode") || have_flag(argc, argv, "ellswift_keygen") ||
have_flag(argc, argv, "ellswift_ecdh")) {
fprintf(stderr, "./bench: ElligatorSwift module not enabled.\n");
fprintf(stderr, "Use ./configure --enable-module-ellswift.\n\n");
return 1;
}
#endif
/* ECDSA benchmark */
data.ctx = secp256k1_context_create(SECP256K1_CONTEXT_NONE);
for (i = 0; i < 32; i++) {
data.msg[i] = 1 + i;
}
for (i = 0; i < 32; i++) {
data.key[i] = 33 + i;
}
data.siglen = 72;
CHECK(secp256k1_ecdsa_sign(data.ctx, &sig, data.msg, data.key, NULL, NULL));
CHECK(secp256k1_ecdsa_signature_serialize_der(data.ctx, data.sig, &data.siglen, &sig));
CHECK(secp256k1_ec_pubkey_create(data.ctx, &pubkey, data.key));
data.pubkeylen = 33;
CHECK(secp256k1_ec_pubkey_serialize(data.ctx, data.pubkey, &data.pubkeylen, &pubkey, SECP256K1_EC_COMPRESSED) == 1);
print_output_table_header_row();
if (d || have_flag(argc, argv, "ecdsa") || have_flag(argc, argv, "verify") || have_flag(argc, argv, "ecdsa_verify")) run_benchmark("ecdsa_verify", bench_verify, NULL, NULL, &data, 10, iters);
if (d || have_flag(argc, argv, "ecdsa") || have_flag(argc, argv, "sign") || have_flag(argc, argv, "ecdsa_sign")) run_benchmark("ecdsa_sign", bench_sign_run, bench_sign_setup, NULL, &data, 10, iters);
if (d || have_flag(argc, argv, "ec") || have_flag(argc, argv, "keygen") || have_flag(argc, argv, "ec_keygen")) run_benchmark("ec_keygen", bench_keygen_run, bench_keygen_setup, NULL, &data, 10, iters);
secp256k1_context_destroy(data.ctx);
#ifdef ENABLE_MODULE_ECDH
/* ECDH benchmarks */
run_ecdh_bench(iters, argc, argv);
#endif
#ifdef ENABLE_MODULE_RECOVERY
/* ECDSA recovery benchmarks */
run_recovery_bench(iters, argc, argv);
#endif
#ifdef ENABLE_MODULE_SCHNORRSIG
/* Schnorr signature benchmarks */
run_schnorrsig_bench(iters, argc, argv);
#endif
#ifdef ENABLE_MODULE_ELLSWIFT
/* ElligatorSwift benchmarks */
run_ellswift_bench(iters, argc, argv);
#endif
return 0;
}

188
libsecp256k1/src/bench.h Normal file
View File

@@ -0,0 +1,188 @@
/***********************************************************************
* Copyright (c) 2014 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_BENCH_H
#define SECP256K1_BENCH_H
#include <stdlib.h>
#include <stdint.h>
#include <stdio.h>
#include <string.h>
#if (defined(_MSC_VER) && _MSC_VER >= 1900)
# include <time.h>
#else
# include <sys/time.h>
#endif
static int64_t gettime_i64(void) {
#if (defined(_MSC_VER) && _MSC_VER >= 1900)
/* C11 way to get wallclock time */
struct timespec tv;
if (!timespec_get(&tv, TIME_UTC)) {
fputs("timespec_get failed!", stderr);
exit(1);
}
return (int64_t)tv.tv_nsec / 1000 + (int64_t)tv.tv_sec * 1000000LL;
#else
struct timeval tv;
gettimeofday(&tv, NULL);
return (int64_t)tv.tv_usec + (int64_t)tv.tv_sec * 1000000LL;
#endif
}
#define FP_EXP (6)
#define FP_MULT (1000000LL)
/* Format fixed point number. */
static void print_number(const int64_t x) {
int64_t x_abs, y;
int c, i, rounding, g; /* g = integer part size, c = fractional part size */
size_t ptr;
char buffer[30];
if (x == INT64_MIN) {
/* Prevent UB. */
printf("ERR");
return;
}
x_abs = x < 0 ? -x : x;
/* Determine how many decimals we want to show (more than FP_EXP makes no
* sense). */
y = x_abs;
c = 0;
while (y > 0LL && y < 100LL * FP_MULT && c < FP_EXP) {
y *= 10LL;
c++;
}
/* Round to 'c' decimals. */
y = x_abs;
rounding = 0;
for (i = c; i < FP_EXP; ++i) {
rounding = (y % 10) >= 5;
y /= 10;
}
y += rounding;
/* Format and print the number. */
ptr = sizeof(buffer) - 1;
buffer[ptr] = 0;
g = 0;
if (c != 0) { /* non zero fractional part */
for (i = 0; i < c; ++i) {
buffer[--ptr] = '0' + (y % 10);
y /= 10;
}
} else if (c == 0) { /* fractional part is 0 */
buffer[--ptr] = '0';
}
buffer[--ptr] = '.';
do {
buffer[--ptr] = '0' + (y % 10);
y /= 10;
g++;
} while (y != 0);
if (x < 0) {
buffer[--ptr] = '-';
g++;
}
printf("%5.*s", g, &buffer[ptr]); /* Prints integer part */
printf("%-*s", FP_EXP, &buffer[ptr + g]); /* Prints fractional part */
}
static void run_benchmark(char *name, void (*benchmark)(void*, int), void (*setup)(void*), void (*teardown)(void*, int), void* data, int count, int iter) {
int i;
int64_t min = INT64_MAX;
int64_t sum = 0;
int64_t max = 0;
for (i = 0; i < count; i++) {
int64_t begin, total;
if (setup != NULL) {
setup(data);
}
begin = gettime_i64();
benchmark(data, iter);
total = gettime_i64() - begin;
if (teardown != NULL) {
teardown(data, iter);
}
if (total < min) {
min = total;
}
if (total > max) {
max = total;
}
sum += total;
}
/* ',' is used as a column delimiter */
printf("%-30s, ", name);
print_number(min * FP_MULT / iter);
printf(" , ");
print_number(((sum * FP_MULT) / count) / iter);
printf(" , ");
print_number(max * FP_MULT / iter);
printf("\n");
}
static int have_flag(int argc, char** argv, char *flag) {
char** argm = argv + argc;
argv++;
while (argv != argm) {
if (strcmp(*argv, flag) == 0) {
return 1;
}
argv++;
}
return 0;
}
/* takes an array containing the arguments that the user is allowed to enter on the command-line
returns:
- 1 if the user entered an invalid argument
- 0 if all the user entered arguments are valid */
static int have_invalid_args(int argc, char** argv, char** valid_args, size_t n) {
size_t i;
int found_valid;
char** argm = argv + argc;
argv++;
while (argv != argm) {
found_valid = 0;
for (i = 0; i < n; i++) {
if (strcmp(*argv, valid_args[i]) == 0) {
found_valid = 1; /* user entered a valid arg from the list */
break;
}
}
if (found_valid == 0) {
return 1; /* invalid arg found */
}
argv++;
}
return 0;
}
static int get_iters(int default_iters) {
char* env = getenv("SECP256K1_BENCH_ITERS");
if (env) {
return strtol(env, NULL, 0);
} else {
return default_iters;
}
}
static void print_output_table_header_row(void) {
char* bench_str = "Benchmark"; /* left justified */
char* min_str = " Min(us) "; /* center alignment */
char* avg_str = " Avg(us) ";
char* max_str = " Max(us) ";
printf("%-30s,%-15s,%-15s,%-15s\n", bench_str, min_str, avg_str, max_str);
printf("\n");
}
#endif /* SECP256K1_BENCH_H */

View File

@@ -0,0 +1,367 @@
/***********************************************************************
* Copyright (c) 2017 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#include <stdio.h>
#include "secp256k1.c"
#include "../include/secp256k1.h"
#include "util.h"
#include "hash_impl.h"
#include "field_impl.h"
#include "group_impl.h"
#include "scalar_impl.h"
#include "ecmult_impl.h"
#include "bench.h"
#define POINTS 32768
static void help(char **argv) {
printf("Benchmark EC multiplication algorithms\n");
printf("\n");
printf("Usage: %s <help|pippenger_wnaf|strauss_wnaf|simple>\n", argv[0]);
printf("The output shows the number of multiplied and summed points right after the\n");
printf("function name. The letter 'g' indicates that one of the points is the generator.\n");
printf("The benchmarks are divided by the number of points.\n");
printf("\n");
printf("default (ecmult_multi): picks pippenger_wnaf or strauss_wnaf depending on the\n");
printf(" batch size\n");
printf("pippenger_wnaf: for all batch sizes\n");
printf("strauss_wnaf: for all batch sizes\n");
printf("simple: multiply and sum each point individually\n");
}
typedef struct {
/* Setup once in advance */
secp256k1_context* ctx;
secp256k1_scratch_space* scratch;
secp256k1_scalar* scalars;
secp256k1_ge* pubkeys;
secp256k1_gej* pubkeys_gej;
secp256k1_scalar* seckeys;
secp256k1_gej* expected_output;
secp256k1_ecmult_multi_func ecmult_multi;
/* Changes per benchmark */
size_t count;
int includes_g;
/* Changes per benchmark iteration, used to pick different scalars and pubkeys
* in each run. */
size_t offset1;
size_t offset2;
/* Benchmark output. */
secp256k1_gej* output;
} bench_data;
/* Hashes x into [0, POINTS) twice and store the result in offset1 and offset2. */
static void hash_into_offset(bench_data* data, size_t x) {
data->offset1 = (x * 0x537b7f6f + 0x8f66a481) % POINTS;
data->offset2 = (x * 0x7f6f537b + 0x6a1a8f49) % POINTS;
}
/* Check correctness of the benchmark by computing
* sum(outputs) ?= (sum(scalars_gen) + sum(seckeys)*sum(scalars))*G */
static void bench_ecmult_teardown_helper(bench_data* data, size_t* seckey_offset, size_t* scalar_offset, size_t* scalar_gen_offset, int iters) {
int i;
secp256k1_gej sum_output, tmp;
secp256k1_scalar sum_scalars;
secp256k1_gej_set_infinity(&sum_output);
secp256k1_scalar_set_int(&sum_scalars, 0);
for (i = 0; i < iters; ++i) {
secp256k1_gej_add_var(&sum_output, &sum_output, &data->output[i], NULL);
if (scalar_gen_offset != NULL) {
secp256k1_scalar_add(&sum_scalars, &sum_scalars, &data->scalars[(*scalar_gen_offset+i) % POINTS]);
}
if (seckey_offset != NULL) {
secp256k1_scalar s = data->seckeys[(*seckey_offset+i) % POINTS];
secp256k1_scalar_mul(&s, &s, &data->scalars[(*scalar_offset+i) % POINTS]);
secp256k1_scalar_add(&sum_scalars, &sum_scalars, &s);
}
}
secp256k1_ecmult_gen(&data->ctx->ecmult_gen_ctx, &tmp, &sum_scalars);
CHECK(secp256k1_gej_eq_var(&tmp, &sum_output));
}
static void bench_ecmult_setup(void* arg) {
bench_data* data = (bench_data*)arg;
/* Re-randomize offset to ensure that we're using different scalars and
* group elements in each run. */
hash_into_offset(data, data->offset1);
}
static void bench_ecmult_gen(void* arg, int iters) {
bench_data* data = (bench_data*)arg;
int i;
for (i = 0; i < iters; ++i) {
secp256k1_ecmult_gen(&data->ctx->ecmult_gen_ctx, &data->output[i], &data->scalars[(data->offset1+i) % POINTS]);
}
}
static void bench_ecmult_gen_teardown(void* arg, int iters) {
bench_data* data = (bench_data*)arg;
bench_ecmult_teardown_helper(data, NULL, NULL, &data->offset1, iters);
}
static void bench_ecmult_const(void* arg, int iters) {
bench_data* data = (bench_data*)arg;
int i;
for (i = 0; i < iters; ++i) {
secp256k1_ecmult_const(&data->output[i], &data->pubkeys[(data->offset1+i) % POINTS], &data->scalars[(data->offset2+i) % POINTS]);
}
}
static void bench_ecmult_const_teardown(void* arg, int iters) {
bench_data* data = (bench_data*)arg;
bench_ecmult_teardown_helper(data, &data->offset1, &data->offset2, NULL, iters);
}
static void bench_ecmult_1p(void* arg, int iters) {
bench_data* data = (bench_data*)arg;
int i;
for (i = 0; i < iters; ++i) {
secp256k1_ecmult(&data->output[i], &data->pubkeys_gej[(data->offset1+i) % POINTS], &data->scalars[(data->offset2+i) % POINTS], NULL);
}
}
static void bench_ecmult_1p_teardown(void* arg, int iters) {
bench_data* data = (bench_data*)arg;
bench_ecmult_teardown_helper(data, &data->offset1, &data->offset2, NULL, iters);
}
static void bench_ecmult_0p_g(void* arg, int iters) {
bench_data* data = (bench_data*)arg;
int i;
for (i = 0; i < iters; ++i) {
secp256k1_ecmult(&data->output[i], NULL, &secp256k1_scalar_zero, &data->scalars[(data->offset1+i) % POINTS]);
}
}
static void bench_ecmult_0p_g_teardown(void* arg, int iters) {
bench_data* data = (bench_data*)arg;
bench_ecmult_teardown_helper(data, NULL, NULL, &data->offset1, iters);
}
static void bench_ecmult_1p_g(void* arg, int iters) {
bench_data* data = (bench_data*)arg;
int i;
for (i = 0; i < iters/2; ++i) {
secp256k1_ecmult(&data->output[i], &data->pubkeys_gej[(data->offset1+i) % POINTS], &data->scalars[(data->offset2+i) % POINTS], &data->scalars[(data->offset1+i) % POINTS]);
}
}
static void bench_ecmult_1p_g_teardown(void* arg, int iters) {
bench_data* data = (bench_data*)arg;
bench_ecmult_teardown_helper(data, &data->offset1, &data->offset2, &data->offset1, iters/2);
}
static void run_ecmult_bench(bench_data* data, int iters) {
char str[32];
sprintf(str, "ecmult_gen");
run_benchmark(str, bench_ecmult_gen, bench_ecmult_setup, bench_ecmult_gen_teardown, data, 10, iters);
sprintf(str, "ecmult_const");
run_benchmark(str, bench_ecmult_const, bench_ecmult_setup, bench_ecmult_const_teardown, data, 10, iters);
/* ecmult with non generator point */
sprintf(str, "ecmult_1p");
run_benchmark(str, bench_ecmult_1p, bench_ecmult_setup, bench_ecmult_1p_teardown, data, 10, iters);
/* ecmult with generator point */
sprintf(str, "ecmult_0p_g");
run_benchmark(str, bench_ecmult_0p_g, bench_ecmult_setup, bench_ecmult_0p_g_teardown, data, 10, iters);
/* ecmult with generator and non-generator point. The reported time is per point. */
sprintf(str, "ecmult_1p_g");
run_benchmark(str, bench_ecmult_1p_g, bench_ecmult_setup, bench_ecmult_1p_g_teardown, data, 10, 2*iters);
}
static int bench_ecmult_multi_callback(secp256k1_scalar* sc, secp256k1_ge* ge, size_t idx, void* arg) {
bench_data* data = (bench_data*)arg;
if (data->includes_g) ++idx;
if (idx == 0) {
*sc = data->scalars[data->offset1];
*ge = secp256k1_ge_const_g;
} else {
*sc = data->scalars[(data->offset1 + idx) % POINTS];
*ge = data->pubkeys[(data->offset2 + idx - 1) % POINTS];
}
return 1;
}
static void bench_ecmult_multi(void* arg, int iters) {
bench_data* data = (bench_data*)arg;
int includes_g = data->includes_g;
int iter;
int count = data->count;
iters = iters / data->count;
for (iter = 0; iter < iters; ++iter) {
data->ecmult_multi(&data->ctx->error_callback, data->scratch, &data->output[iter], data->includes_g ? &data->scalars[data->offset1] : NULL, bench_ecmult_multi_callback, arg, count - includes_g);
data->offset1 = (data->offset1 + count) % POINTS;
data->offset2 = (data->offset2 + count - 1) % POINTS;
}
}
static void bench_ecmult_multi_setup(void* arg) {
bench_data* data = (bench_data*)arg;
hash_into_offset(data, data->count);
}
static void bench_ecmult_multi_teardown(void* arg, int iters) {
bench_data* data = (bench_data*)arg;
int iter;
iters = iters / data->count;
/* Verify the results in teardown, to avoid doing comparisons while benchmarking. */
for (iter = 0; iter < iters; ++iter) {
secp256k1_gej tmp;
secp256k1_gej_add_var(&tmp, &data->output[iter], &data->expected_output[iter], NULL);
CHECK(secp256k1_gej_is_infinity(&tmp));
}
}
static void generate_scalar(uint32_t num, secp256k1_scalar* scalar) {
secp256k1_sha256 sha256;
unsigned char c[10] = {'e', 'c', 'm', 'u', 'l', 't', 0, 0, 0, 0};
unsigned char buf[32];
int overflow = 0;
c[6] = num;
c[7] = num >> 8;
c[8] = num >> 16;
c[9] = num >> 24;
secp256k1_sha256_initialize(&sha256);
secp256k1_sha256_write(&sha256, c, sizeof(c));
secp256k1_sha256_finalize(&sha256, buf);
secp256k1_scalar_set_b32(scalar, buf, &overflow);
CHECK(!overflow);
}
static void run_ecmult_multi_bench(bench_data* data, size_t count, int includes_g, int num_iters) {
char str[32];
size_t iters = 1 + num_iters / count;
size_t iter;
data->count = count;
data->includes_g = includes_g;
/* Compute (the negation of) the expected results directly. */
hash_into_offset(data, data->count);
for (iter = 0; iter < iters; ++iter) {
secp256k1_scalar tmp;
secp256k1_scalar total = data->scalars[(data->offset1++) % POINTS];
size_t i = 0;
for (i = 0; i + 1 < count; ++i) {
secp256k1_scalar_mul(&tmp, &data->seckeys[(data->offset2++) % POINTS], &data->scalars[(data->offset1++) % POINTS]);
secp256k1_scalar_add(&total, &total, &tmp);
}
secp256k1_scalar_negate(&total, &total);
secp256k1_ecmult(&data->expected_output[iter], NULL, &secp256k1_scalar_zero, &total);
}
/* Run the benchmark. */
if (includes_g) {
sprintf(str, "ecmult_multi_%ip_g", (int)count - 1);
} else {
sprintf(str, "ecmult_multi_%ip", (int)count);
}
run_benchmark(str, bench_ecmult_multi, bench_ecmult_multi_setup, bench_ecmult_multi_teardown, data, 10, count * iters);
}
int main(int argc, char **argv) {
bench_data data;
int i, p;
size_t scratch_size;
int iters = get_iters(10000);
data.ecmult_multi = secp256k1_ecmult_multi_var;
if (argc > 1) {
if(have_flag(argc, argv, "-h")
|| have_flag(argc, argv, "--help")
|| have_flag(argc, argv, "help")) {
help(argv);
return 0;
} else if(have_flag(argc, argv, "pippenger_wnaf")) {
printf("Using pippenger_wnaf:\n");
data.ecmult_multi = secp256k1_ecmult_pippenger_batch_single;
} else if(have_flag(argc, argv, "strauss_wnaf")) {
printf("Using strauss_wnaf:\n");
data.ecmult_multi = secp256k1_ecmult_strauss_batch_single;
} else if(have_flag(argc, argv, "simple")) {
printf("Using simple algorithm:\n");
} else {
fprintf(stderr, "%s: unrecognized argument '%s'.\n\n", argv[0], argv[1]);
help(argv);
return 1;
}
}
data.ctx = secp256k1_context_create(SECP256K1_CONTEXT_NONE);
scratch_size = secp256k1_strauss_scratch_size(POINTS) + STRAUSS_SCRATCH_OBJECTS*16;
if (!have_flag(argc, argv, "simple")) {
data.scratch = secp256k1_scratch_space_create(data.ctx, scratch_size);
} else {
data.scratch = NULL;
}
/* Allocate stuff */
data.scalars = malloc(sizeof(secp256k1_scalar) * POINTS);
data.seckeys = malloc(sizeof(secp256k1_scalar) * POINTS);
data.pubkeys = malloc(sizeof(secp256k1_ge) * POINTS);
data.pubkeys_gej = malloc(sizeof(secp256k1_gej) * POINTS);
data.expected_output = malloc(sizeof(secp256k1_gej) * (iters + 1));
data.output = malloc(sizeof(secp256k1_gej) * (iters + 1));
/* Generate a set of scalars, and private/public keypairs. */
secp256k1_gej_set_ge(&data.pubkeys_gej[0], &secp256k1_ge_const_g);
secp256k1_scalar_set_int(&data.seckeys[0], 1);
for (i = 0; i < POINTS; ++i) {
generate_scalar(i, &data.scalars[i]);
if (i) {
secp256k1_gej_double_var(&data.pubkeys_gej[i], &data.pubkeys_gej[i - 1], NULL);
secp256k1_scalar_add(&data.seckeys[i], &data.seckeys[i - 1], &data.seckeys[i - 1]);
}
}
secp256k1_ge_set_all_gej_var(data.pubkeys, data.pubkeys_gej, POINTS);
print_output_table_header_row();
/* Initialize offset1 and offset2 */
hash_into_offset(&data, 0);
run_ecmult_bench(&data, iters);
for (i = 1; i <= 8; ++i) {
run_ecmult_multi_bench(&data, i, 1, iters);
}
/* This is disabled with low count of iterations because the loop runs 77 times even with iters=1
* and the higher it goes the longer the computation takes(more points)
* So we don't run this benchmark with low iterations to prevent slow down */
if (iters > 2) {
for (p = 0; p <= 11; ++p) {
for (i = 9; i <= 16; ++i) {
run_ecmult_multi_bench(&data, i << p, 1, iters);
}
}
}
if (data.scratch != NULL) {
secp256k1_scratch_space_destroy(data.ctx, data.scratch);
}
secp256k1_context_destroy(data.ctx);
free(data.scalars);
free(data.pubkeys);
free(data.pubkeys_gej);
free(data.seckeys);
free(data.output);
free(data.expected_output);
return(0);
}

View File

@@ -0,0 +1,436 @@
/***********************************************************************
* Copyright (c) 2014-2015 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#include <stdio.h>
#include "secp256k1.c"
#include "../include/secp256k1.h"
#include "assumptions.h"
#include "util.h"
#include "hash_impl.h"
#include "field_impl.h"
#include "group_impl.h"
#include "scalar_impl.h"
#include "ecmult_impl.h"
#include "bench.h"
static void help(int default_iters) {
printf("Benchmarks various internal routines.\n");
printf("\n");
printf("The default number of iterations for each benchmark is %d. This can be\n", default_iters);
printf("customized using the SECP256K1_BENCH_ITERS environment variable.\n");
printf("\n");
printf("Usage: ./bench_internal [args]\n");
printf("By default, all benchmarks will be run.\n");
printf("args:\n");
printf(" help : display this help and exit\n");
printf(" scalar : all scalar operations (add, half, inverse, mul, negate, split)\n");
printf(" field : all field operations (half, inverse, issquare, mul, normalize, sqr, sqrt)\n");
printf(" group : all group operations (add, double, to_affine)\n");
printf(" ecmult : all point multiplication operations (ecmult_wnaf) \n");
printf(" hash : all hash algorithms (hmac, rng6979, sha256)\n");
printf(" context : all context object operations (context_create)\n");
printf("\n");
}
typedef struct {
secp256k1_scalar scalar[2];
secp256k1_fe fe[4];
secp256k1_ge ge[2];
secp256k1_gej gej[2];
unsigned char data[64];
int wnaf[256];
} bench_inv;
static void bench_setup(void* arg) {
bench_inv *data = (bench_inv*)arg;
static const unsigned char init[4][32] = {
/* Initializer for scalar[0], fe[0], first half of data, the X coordinate of ge[0],
and the (implied affine) X coordinate of gej[0]. */
{
0x02, 0x03, 0x05, 0x07, 0x0b, 0x0d, 0x11, 0x13,
0x17, 0x1d, 0x1f, 0x25, 0x29, 0x2b, 0x2f, 0x35,
0x3b, 0x3d, 0x43, 0x47, 0x49, 0x4f, 0x53, 0x59,
0x61, 0x65, 0x67, 0x6b, 0x6d, 0x71, 0x7f, 0x83
},
/* Initializer for scalar[1], fe[1], first half of data, the X coordinate of ge[1],
and the (implied affine) X coordinate of gej[1]. */
{
0x82, 0x83, 0x85, 0x87, 0x8b, 0x8d, 0x81, 0x83,
0x97, 0xad, 0xaf, 0xb5, 0xb9, 0xbb, 0xbf, 0xc5,
0xdb, 0xdd, 0xe3, 0xe7, 0xe9, 0xef, 0xf3, 0xf9,
0x11, 0x15, 0x17, 0x1b, 0x1d, 0xb1, 0xbf, 0xd3
},
/* Initializer for fe[2] and the Z coordinate of gej[0]. */
{
0x3d, 0x2d, 0xef, 0xf4, 0x25, 0x98, 0x4f, 0x5d,
0xe2, 0xca, 0x5f, 0x41, 0x3f, 0x3f, 0xce, 0x44,
0xaa, 0x2c, 0x53, 0x8a, 0xc6, 0x59, 0x1f, 0x38,
0x38, 0x23, 0xe4, 0x11, 0x27, 0xc6, 0xa0, 0xe7
},
/* Initializer for fe[3] and the Z coordinate of gej[1]. */
{
0xbd, 0x21, 0xa5, 0xe1, 0x13, 0x50, 0x73, 0x2e,
0x52, 0x98, 0xc8, 0x9e, 0xab, 0x00, 0xa2, 0x68,
0x43, 0xf5, 0xd7, 0x49, 0x80, 0x72, 0xa7, 0xf3,
0xd7, 0x60, 0xe6, 0xab, 0x90, 0x92, 0xdf, 0xc5
}
};
secp256k1_scalar_set_b32(&data->scalar[0], init[0], NULL);
secp256k1_scalar_set_b32(&data->scalar[1], init[1], NULL);
secp256k1_fe_set_b32_limit(&data->fe[0], init[0]);
secp256k1_fe_set_b32_limit(&data->fe[1], init[1]);
secp256k1_fe_set_b32_limit(&data->fe[2], init[2]);
secp256k1_fe_set_b32_limit(&data->fe[3], init[3]);
CHECK(secp256k1_ge_set_xo_var(&data->ge[0], &data->fe[0], 0));
CHECK(secp256k1_ge_set_xo_var(&data->ge[1], &data->fe[1], 1));
secp256k1_gej_set_ge(&data->gej[0], &data->ge[0]);
secp256k1_gej_rescale(&data->gej[0], &data->fe[2]);
secp256k1_gej_set_ge(&data->gej[1], &data->ge[1]);
secp256k1_gej_rescale(&data->gej[1], &data->fe[3]);
memcpy(data->data, init[0], 32);
memcpy(data->data + 32, init[1], 32);
}
static void bench_scalar_add(void* arg, int iters) {
int i, j = 0;
bench_inv *data = (bench_inv*)arg;
for (i = 0; i < iters; i++) {
j += secp256k1_scalar_add(&data->scalar[0], &data->scalar[0], &data->scalar[1]);
}
CHECK(j <= iters);
}
static void bench_scalar_negate(void* arg, int iters) {
int i;
bench_inv *data = (bench_inv*)arg;
for (i = 0; i < iters; i++) {
secp256k1_scalar_negate(&data->scalar[0], &data->scalar[0]);
}
}
static void bench_scalar_half(void* arg, int iters) {
int i;
bench_inv *data = (bench_inv*)arg;
secp256k1_scalar s = data->scalar[0];
for (i = 0; i < iters; i++) {
secp256k1_scalar_half(&s, &s);
}
data->scalar[0] = s;
}
static void bench_scalar_mul(void* arg, int iters) {
int i;
bench_inv *data = (bench_inv*)arg;
for (i = 0; i < iters; i++) {
secp256k1_scalar_mul(&data->scalar[0], &data->scalar[0], &data->scalar[1]);
}
}
static void bench_scalar_split(void* arg, int iters) {
int i, j = 0;
bench_inv *data = (bench_inv*)arg;
secp256k1_scalar tmp;
for (i = 0; i < iters; i++) {
secp256k1_scalar_split_lambda(&tmp, &data->scalar[1], &data->scalar[0]);
j += secp256k1_scalar_add(&data->scalar[0], &tmp, &data->scalar[1]);
}
CHECK(j <= iters);
}
static void bench_scalar_inverse(void* arg, int iters) {
int i, j = 0;
bench_inv *data = (bench_inv*)arg;
for (i = 0; i < iters; i++) {
secp256k1_scalar_inverse(&data->scalar[0], &data->scalar[0]);
j += secp256k1_scalar_add(&data->scalar[0], &data->scalar[0], &data->scalar[1]);
}
CHECK(j <= iters);
}
static void bench_scalar_inverse_var(void* arg, int iters) {
int i, j = 0;
bench_inv *data = (bench_inv*)arg;
for (i = 0; i < iters; i++) {
secp256k1_scalar_inverse_var(&data->scalar[0], &data->scalar[0]);
j += secp256k1_scalar_add(&data->scalar[0], &data->scalar[0], &data->scalar[1]);
}
CHECK(j <= iters);
}
static void bench_field_half(void* arg, int iters) {
int i;
bench_inv *data = (bench_inv*)arg;
for (i = 0; i < iters; i++) {
secp256k1_fe_half(&data->fe[0]);
}
}
static void bench_field_normalize(void* arg, int iters) {
int i;
bench_inv *data = (bench_inv*)arg;
for (i = 0; i < iters; i++) {
secp256k1_fe_normalize(&data->fe[0]);
}
}
static void bench_field_normalize_weak(void* arg, int iters) {
int i;
bench_inv *data = (bench_inv*)arg;
for (i = 0; i < iters; i++) {
secp256k1_fe_normalize_weak(&data->fe[0]);
}
}
static void bench_field_mul(void* arg, int iters) {
int i;
bench_inv *data = (bench_inv*)arg;
for (i = 0; i < iters; i++) {
secp256k1_fe_mul(&data->fe[0], &data->fe[0], &data->fe[1]);
}
}
static void bench_field_sqr(void* arg, int iters) {
int i;
bench_inv *data = (bench_inv*)arg;
for (i = 0; i < iters; i++) {
secp256k1_fe_sqr(&data->fe[0], &data->fe[0]);
}
}
static void bench_field_inverse(void* arg, int iters) {
int i;
bench_inv *data = (bench_inv*)arg;
for (i = 0; i < iters; i++) {
secp256k1_fe_inv(&data->fe[0], &data->fe[0]);
secp256k1_fe_add(&data->fe[0], &data->fe[1]);
}
}
static void bench_field_inverse_var(void* arg, int iters) {
int i;
bench_inv *data = (bench_inv*)arg;
for (i = 0; i < iters; i++) {
secp256k1_fe_inv_var(&data->fe[0], &data->fe[0]);
secp256k1_fe_add(&data->fe[0], &data->fe[1]);
}
}
static void bench_field_sqrt(void* arg, int iters) {
int i, j = 0;
bench_inv *data = (bench_inv*)arg;
secp256k1_fe t;
for (i = 0; i < iters; i++) {
t = data->fe[0];
j += secp256k1_fe_sqrt(&data->fe[0], &t);
secp256k1_fe_add(&data->fe[0], &data->fe[1]);
}
CHECK(j <= iters);
}
static void bench_field_is_square_var(void* arg, int iters) {
int i, j = 0;
bench_inv *data = (bench_inv*)arg;
secp256k1_fe t = data->fe[0];
for (i = 0; i < iters; i++) {
j += secp256k1_fe_is_square_var(&t);
secp256k1_fe_add(&t, &data->fe[1]);
secp256k1_fe_normalize_var(&t);
}
CHECK(j <= iters);
}
static void bench_group_double_var(void* arg, int iters) {
int i;
bench_inv *data = (bench_inv*)arg;
for (i = 0; i < iters; i++) {
secp256k1_gej_double_var(&data->gej[0], &data->gej[0], NULL);
}
}
static void bench_group_add_var(void* arg, int iters) {
int i;
bench_inv *data = (bench_inv*)arg;
for (i = 0; i < iters; i++) {
secp256k1_gej_add_var(&data->gej[0], &data->gej[0], &data->gej[1], NULL);
}
}
static void bench_group_add_affine(void* arg, int iters) {
int i;
bench_inv *data = (bench_inv*)arg;
for (i = 0; i < iters; i++) {
secp256k1_gej_add_ge(&data->gej[0], &data->gej[0], &data->ge[1]);
}
}
static void bench_group_add_affine_var(void* arg, int iters) {
int i;
bench_inv *data = (bench_inv*)arg;
for (i = 0; i < iters; i++) {
secp256k1_gej_add_ge_var(&data->gej[0], &data->gej[0], &data->ge[1], NULL);
}
}
static void bench_group_add_zinv_var(void* arg, int iters) {
int i;
bench_inv *data = (bench_inv*)arg;
for (i = 0; i < iters; i++) {
secp256k1_gej_add_zinv_var(&data->gej[0], &data->gej[0], &data->ge[1], &data->gej[0].y);
}
}
static void bench_group_to_affine_var(void* arg, int iters) {
int i;
bench_inv *data = (bench_inv*)arg;
for (i = 0; i < iters; ++i) {
secp256k1_ge_set_gej_var(&data->ge[1], &data->gej[0]);
/* Use the output affine X/Y coordinates to vary the input X/Y/Z coordinates.
Note that the resulting coordinates will generally not correspond to a point
on the curve, but this is not a problem for the code being benchmarked here.
Adding and normalizing have less overhead than EC operations (which could
guarantee the point remains on the curve). */
secp256k1_fe_add(&data->gej[0].x, &data->ge[1].y);
secp256k1_fe_add(&data->gej[0].y, &data->fe[2]);
secp256k1_fe_add(&data->gej[0].z, &data->ge[1].x);
secp256k1_fe_normalize_var(&data->gej[0].x);
secp256k1_fe_normalize_var(&data->gej[0].y);
secp256k1_fe_normalize_var(&data->gej[0].z);
}
}
static void bench_ecmult_wnaf(void* arg, int iters) {
int i, bits = 0, overflow = 0;
bench_inv *data = (bench_inv*)arg;
for (i = 0; i < iters; i++) {
bits += secp256k1_ecmult_wnaf(data->wnaf, 256, &data->scalar[0], WINDOW_A);
overflow += secp256k1_scalar_add(&data->scalar[0], &data->scalar[0], &data->scalar[1]);
}
CHECK(overflow >= 0);
CHECK(bits <= 256*iters);
}
static void bench_sha256(void* arg, int iters) {
int i;
bench_inv *data = (bench_inv*)arg;
secp256k1_sha256 sha;
for (i = 0; i < iters; i++) {
secp256k1_sha256_initialize(&sha);
secp256k1_sha256_write(&sha, data->data, 32);
secp256k1_sha256_finalize(&sha, data->data);
}
}
static void bench_hmac_sha256(void* arg, int iters) {
int i;
bench_inv *data = (bench_inv*)arg;
secp256k1_hmac_sha256 hmac;
for (i = 0; i < iters; i++) {
secp256k1_hmac_sha256_initialize(&hmac, data->data, 32);
secp256k1_hmac_sha256_write(&hmac, data->data, 32);
secp256k1_hmac_sha256_finalize(&hmac, data->data);
}
}
static void bench_rfc6979_hmac_sha256(void* arg, int iters) {
int i;
bench_inv *data = (bench_inv*)arg;
secp256k1_rfc6979_hmac_sha256 rng;
for (i = 0; i < iters; i++) {
secp256k1_rfc6979_hmac_sha256_initialize(&rng, data->data, 64);
secp256k1_rfc6979_hmac_sha256_generate(&rng, data->data, 32);
}
}
static void bench_context(void* arg, int iters) {
int i;
(void)arg;
for (i = 0; i < iters; i++) {
secp256k1_context_destroy(secp256k1_context_create(SECP256K1_CONTEXT_NONE));
}
}
int main(int argc, char **argv) {
bench_inv data;
int default_iters = 20000;
int iters = get_iters(default_iters);
int d = argc == 1; /* default */
if (argc > 1) {
if (have_flag(argc, argv, "-h")
|| have_flag(argc, argv, "--help")
|| have_flag(argc, argv, "help")) {
help(default_iters);
return 0;
}
}
print_output_table_header_row();
if (d || have_flag(argc, argv, "scalar") || have_flag(argc, argv, "half")) run_benchmark("scalar_half", bench_scalar_half, bench_setup, NULL, &data, 10, iters*100);
if (d || have_flag(argc, argv, "scalar") || have_flag(argc, argv, "add")) run_benchmark("scalar_add", bench_scalar_add, bench_setup, NULL, &data, 10, iters*100);
if (d || have_flag(argc, argv, "scalar") || have_flag(argc, argv, "negate")) run_benchmark("scalar_negate", bench_scalar_negate, bench_setup, NULL, &data, 10, iters*100);
if (d || have_flag(argc, argv, "scalar") || have_flag(argc, argv, "mul")) run_benchmark("scalar_mul", bench_scalar_mul, bench_setup, NULL, &data, 10, iters*10);
if (d || have_flag(argc, argv, "scalar") || have_flag(argc, argv, "split")) run_benchmark("scalar_split", bench_scalar_split, bench_setup, NULL, &data, 10, iters);
if (d || have_flag(argc, argv, "scalar") || have_flag(argc, argv, "inverse")) run_benchmark("scalar_inverse", bench_scalar_inverse, bench_setup, NULL, &data, 10, iters);
if (d || have_flag(argc, argv, "scalar") || have_flag(argc, argv, "inverse")) run_benchmark("scalar_inverse_var", bench_scalar_inverse_var, bench_setup, NULL, &data, 10, iters);
if (d || have_flag(argc, argv, "field") || have_flag(argc, argv, "half")) run_benchmark("field_half", bench_field_half, bench_setup, NULL, &data, 10, iters*100);
if (d || have_flag(argc, argv, "field") || have_flag(argc, argv, "normalize")) run_benchmark("field_normalize", bench_field_normalize, bench_setup, NULL, &data, 10, iters*100);
if (d || have_flag(argc, argv, "field") || have_flag(argc, argv, "normalize")) run_benchmark("field_normalize_weak", bench_field_normalize_weak, bench_setup, NULL, &data, 10, iters*100);
if (d || have_flag(argc, argv, "field") || have_flag(argc, argv, "sqr")) run_benchmark("field_sqr", bench_field_sqr, bench_setup, NULL, &data, 10, iters*10);
if (d || have_flag(argc, argv, "field") || have_flag(argc, argv, "mul")) run_benchmark("field_mul", bench_field_mul, bench_setup, NULL, &data, 10, iters*10);
if (d || have_flag(argc, argv, "field") || have_flag(argc, argv, "inverse")) run_benchmark("field_inverse", bench_field_inverse, bench_setup, NULL, &data, 10, iters);
if (d || have_flag(argc, argv, "field") || have_flag(argc, argv, "inverse")) run_benchmark("field_inverse_var", bench_field_inverse_var, bench_setup, NULL, &data, 10, iters);
if (d || have_flag(argc, argv, "field") || have_flag(argc, argv, "issquare")) run_benchmark("field_is_square_var", bench_field_is_square_var, bench_setup, NULL, &data, 10, iters);
if (d || have_flag(argc, argv, "field") || have_flag(argc, argv, "sqrt")) run_benchmark("field_sqrt", bench_field_sqrt, bench_setup, NULL, &data, 10, iters);
if (d || have_flag(argc, argv, "group") || have_flag(argc, argv, "double")) run_benchmark("group_double_var", bench_group_double_var, bench_setup, NULL, &data, 10, iters*10);
if (d || have_flag(argc, argv, "group") || have_flag(argc, argv, "add")) run_benchmark("group_add_var", bench_group_add_var, bench_setup, NULL, &data, 10, iters*10);
if (d || have_flag(argc, argv, "group") || have_flag(argc, argv, "add")) run_benchmark("group_add_affine", bench_group_add_affine, bench_setup, NULL, &data, 10, iters*10);
if (d || have_flag(argc, argv, "group") || have_flag(argc, argv, "add")) run_benchmark("group_add_affine_var", bench_group_add_affine_var, bench_setup, NULL, &data, 10, iters*10);
if (d || have_flag(argc, argv, "group") || have_flag(argc, argv, "add")) run_benchmark("group_add_zinv_var", bench_group_add_zinv_var, bench_setup, NULL, &data, 10, iters*10);
if (d || have_flag(argc, argv, "group") || have_flag(argc, argv, "to_affine")) run_benchmark("group_to_affine_var", bench_group_to_affine_var, bench_setup, NULL, &data, 10, iters);
if (d || have_flag(argc, argv, "ecmult") || have_flag(argc, argv, "wnaf")) run_benchmark("ecmult_wnaf", bench_ecmult_wnaf, bench_setup, NULL, &data, 10, iters);
if (d || have_flag(argc, argv, "hash") || have_flag(argc, argv, "sha256")) run_benchmark("hash_sha256", bench_sha256, bench_setup, NULL, &data, 10, iters);
if (d || have_flag(argc, argv, "hash") || have_flag(argc, argv, "hmac")) run_benchmark("hash_hmac_sha256", bench_hmac_sha256, bench_setup, NULL, &data, 10, iters);
if (d || have_flag(argc, argv, "hash") || have_flag(argc, argv, "rng6979")) run_benchmark("hash_rfc6979_hmac_sha256", bench_rfc6979_hmac_sha256, bench_setup, NULL, &data, 10, iters);
if (d || have_flag(argc, argv, "context")) run_benchmark("context_create", bench_context, bench_setup, NULL, &data, 10, iters);
return 0;
}

102
libsecp256k1/src/checkmem.h Normal file
View File

@@ -0,0 +1,102 @@
/***********************************************************************
* Copyright (c) 2022 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
/* The code here is inspired by Kris Kwiatkowski's approach in
* https://github.com/kriskwiatkowski/pqc/blob/main/src/common/ct_check.h
* to provide a general interface for memory-checking mechanisms, primarily
* for constant-time checking.
*/
/* These macros are defined by this header file:
*
* - SECP256K1_CHECKMEM_ENABLED:
* - 1 if memory-checking integration is available, 0 otherwise.
* This is just a compile-time macro. Use the next macro to check it is actually
* available at runtime.
* - SECP256K1_CHECKMEM_RUNNING():
* - Acts like a function call, returning 1 if memory checking is available
* at runtime.
* - SECP256K1_CHECKMEM_CHECK(p, len):
* - Assert or otherwise fail in case the len-byte memory block pointed to by p is
* not considered entirely defined.
* - SECP256K1_CHECKMEM_CHECK_VERIFY(p, len):
* - Like SECP256K1_CHECKMEM_CHECK, but only works in VERIFY mode.
* - SECP256K1_CHECKMEM_UNDEFINE(p, len):
* - marks the len-byte memory block pointed to by p as undefined data (secret data,
* in the context of constant-time checking).
* - SECP256K1_CHECKMEM_DEFINE(p, len):
* - marks the len-byte memory pointed to by p as defined data (public data, in the
* context of constant-time checking).
* - SECP256K1_CHECKMEM_MSAN_DEFINE(p, len):
* - Like SECP256K1_CHECKMEM_DEFINE, but applies only to memory_sanitizer.
*
*/
#ifndef SECP256K1_CHECKMEM_H
#define SECP256K1_CHECKMEM_H
/* Define a statement-like macro that ignores the arguments. */
#define SECP256K1_CHECKMEM_NOOP(p, len) do { (void)(p); (void)(len); } while(0)
/* If compiling under msan, map the SECP256K1_CHECKMEM_* functionality to msan.
* Choose this preferentially, even when VALGRIND is defined, as msan-compiled
* binaries can't be run under valgrind anyway. */
#if defined(__has_feature)
# if __has_feature(memory_sanitizer)
# include <sanitizer/msan_interface.h>
# define SECP256K1_CHECKMEM_ENABLED 1
# define SECP256K1_CHECKMEM_UNDEFINE(p, len) __msan_allocated_memory((p), (len))
# define SECP256K1_CHECKMEM_DEFINE(p, len) __msan_unpoison((p), (len))
# define SECP256K1_CHECKMEM_MSAN_DEFINE(p, len) __msan_unpoison((p), (len))
# define SECP256K1_CHECKMEM_CHECK(p, len) __msan_check_mem_is_initialized((p), (len))
# define SECP256K1_CHECKMEM_RUNNING() (1)
# endif
#endif
#if !defined SECP256K1_CHECKMEM_MSAN_DEFINE
# define SECP256K1_CHECKMEM_MSAN_DEFINE(p, len) SECP256K1_CHECKMEM_NOOP((p), (len))
#endif
/* If valgrind integration is desired (through the VALGRIND define), implement the
* SECP256K1_CHECKMEM_* macros using valgrind. */
#if !defined SECP256K1_CHECKMEM_ENABLED
# if defined VALGRIND
# include <stddef.h>
# if defined(__clang__) && defined(__APPLE__)
# pragma clang diagnostic push
# pragma clang diagnostic ignored "-Wreserved-identifier"
# endif
# include <valgrind/memcheck.h>
# if defined(__clang__) && defined(__APPLE__)
# pragma clang diagnostic pop
# endif
# define SECP256K1_CHECKMEM_ENABLED 1
# define SECP256K1_CHECKMEM_UNDEFINE(p, len) VALGRIND_MAKE_MEM_UNDEFINED((p), (len))
# define SECP256K1_CHECKMEM_DEFINE(p, len) VALGRIND_MAKE_MEM_DEFINED((p), (len))
# define SECP256K1_CHECKMEM_CHECK(p, len) VALGRIND_CHECK_MEM_IS_DEFINED((p), (len))
/* VALGRIND_MAKE_MEM_DEFINED returns 0 iff not running on memcheck.
* This is more precise than the RUNNING_ON_VALGRIND macro, which
* checks for valgrind in general instead of memcheck specifically. */
# define SECP256K1_CHECKMEM_RUNNING() (VALGRIND_MAKE_MEM_DEFINED(NULL, 0) != 0)
# endif
#endif
/* As a fall-back, map these macros to dummy statements. */
#if !defined SECP256K1_CHECKMEM_ENABLED
# define SECP256K1_CHECKMEM_ENABLED 0
# define SECP256K1_CHECKMEM_UNDEFINE(p, len) SECP256K1_CHECKMEM_NOOP((p), (len))
# define SECP256K1_CHECKMEM_DEFINE(p, len) SECP256K1_CHECKMEM_NOOP((p), (len))
# define SECP256K1_CHECKMEM_CHECK(p, len) SECP256K1_CHECKMEM_NOOP((p), (len))
# define SECP256K1_CHECKMEM_RUNNING() (0)
#endif
#if defined VERIFY
#define SECP256K1_CHECKMEM_CHECK_VERIFY(p, len) SECP256K1_CHECKMEM_CHECK((p), (len))
#else
#define SECP256K1_CHECKMEM_CHECK_VERIFY(p, len) SECP256K1_CHECKMEM_NOOP((p), (len))
#endif
#endif /* SECP256K1_CHECKMEM_H */

21
libsecp256k1/src/ecdsa.h Normal file
View File

@@ -0,0 +1,21 @@
/***********************************************************************
* Copyright (c) 2013, 2014 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_ECDSA_H
#define SECP256K1_ECDSA_H
#include <stddef.h>
#include "scalar.h"
#include "group.h"
#include "ecmult.h"
static int secp256k1_ecdsa_sig_parse(secp256k1_scalar *r, secp256k1_scalar *s, const unsigned char *sig, size_t size);
static int secp256k1_ecdsa_sig_serialize(unsigned char *sig, size_t *size, const secp256k1_scalar *r, const secp256k1_scalar *s);
static int secp256k1_ecdsa_sig_verify(const secp256k1_scalar* r, const secp256k1_scalar* s, const secp256k1_ge *pubkey, const secp256k1_scalar *message);
static int secp256k1_ecdsa_sig_sign(const secp256k1_ecmult_gen_context *ctx, secp256k1_scalar* r, secp256k1_scalar* s, const secp256k1_scalar *seckey, const secp256k1_scalar *message, const secp256k1_scalar *nonce, int *recid);
#endif /* SECP256K1_ECDSA_H */

View File

@@ -0,0 +1,304 @@
/***********************************************************************
* Copyright (c) 2013-2015 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_ECDSA_IMPL_H
#define SECP256K1_ECDSA_IMPL_H
#include "scalar.h"
#include "field.h"
#include "group.h"
#include "ecmult.h"
#include "ecmult_gen.h"
#include "ecdsa.h"
/** Group order for secp256k1 defined as 'n' in "Standards for Efficient Cryptography" (SEC2) 2.7.1
* $ sage -c 'load("secp256k1_params.sage"); print(hex(N))'
* 0xfffffffffffffffffffffffffffffffebaaedce6af48a03bbfd25e8cd0364141
*/
static const secp256k1_fe secp256k1_ecdsa_const_order_as_fe = SECP256K1_FE_CONST(
0xFFFFFFFFUL, 0xFFFFFFFFUL, 0xFFFFFFFFUL, 0xFFFFFFFEUL,
0xBAAEDCE6UL, 0xAF48A03BUL, 0xBFD25E8CUL, 0xD0364141UL
);
/** Difference between field and order, values 'p' and 'n' values defined in
* "Standards for Efficient Cryptography" (SEC2) 2.7.1.
* $ sage -c 'load("secp256k1_params.sage"); print(hex(P-N))'
* 0x14551231950b75fc4402da1722fc9baee
*/
static const secp256k1_fe secp256k1_ecdsa_const_p_minus_order = SECP256K1_FE_CONST(
0, 0, 0, 1, 0x45512319UL, 0x50B75FC4UL, 0x402DA172UL, 0x2FC9BAEEUL
);
static int secp256k1_der_read_len(size_t *len, const unsigned char **sigp, const unsigned char *sigend) {
size_t lenleft;
unsigned char b1;
VERIFY_CHECK(len != NULL);
*len = 0;
if (*sigp >= sigend) {
return 0;
}
b1 = *((*sigp)++);
if (b1 == 0xFF) {
/* X.690-0207 8.1.3.5.c the value 0xFF shall not be used. */
return 0;
}
if ((b1 & 0x80) == 0) {
/* X.690-0207 8.1.3.4 short form length octets */
*len = b1;
return 1;
}
if (b1 == 0x80) {
/* Indefinite length is not allowed in DER. */
return 0;
}
/* X.690-207 8.1.3.5 long form length octets */
lenleft = b1 & 0x7F; /* lenleft is at least 1 */
if (lenleft > (size_t)(sigend - *sigp)) {
return 0;
}
if (**sigp == 0) {
/* Not the shortest possible length encoding. */
return 0;
}
if (lenleft > sizeof(size_t)) {
/* The resulting length would exceed the range of a size_t, so
* it is certainly longer than the passed array size. */
return 0;
}
while (lenleft > 0) {
*len = (*len << 8) | **sigp;
(*sigp)++;
lenleft--;
}
if (*len > (size_t)(sigend - *sigp)) {
/* Result exceeds the length of the passed array.
(Checking this is the responsibility of the caller but it
can't hurt do it here, too.) */
return 0;
}
if (*len < 128) {
/* Not the shortest possible length encoding. */
return 0;
}
return 1;
}
static int secp256k1_der_parse_integer(secp256k1_scalar *r, const unsigned char **sig, const unsigned char *sigend) {
int overflow = 0;
unsigned char ra[32] = {0};
size_t rlen;
if (*sig == sigend || **sig != 0x02) {
/* Not a primitive integer (X.690-0207 8.3.1). */
return 0;
}
(*sig)++;
if (secp256k1_der_read_len(&rlen, sig, sigend) == 0) {
return 0;
}
if (rlen == 0 || rlen > (size_t)(sigend - *sig)) {
/* Exceeds bounds or not at least length 1 (X.690-0207 8.3.1). */
return 0;
}
if (**sig == 0x00 && rlen > 1 && (((*sig)[1]) & 0x80) == 0x00) {
/* Excessive 0x00 padding. */
return 0;
}
if (**sig == 0xFF && rlen > 1 && (((*sig)[1]) & 0x80) == 0x80) {
/* Excessive 0xFF padding. */
return 0;
}
if ((**sig & 0x80) == 0x80) {
/* Negative. */
overflow = 1;
}
/* There is at most one leading zero byte:
* if there were two leading zero bytes, we would have failed and returned 0
* because of excessive 0x00 padding already. */
if (rlen > 0 && **sig == 0) {
/* Skip leading zero byte */
rlen--;
(*sig)++;
}
if (rlen > 32) {
overflow = 1;
}
if (!overflow) {
if (rlen) memcpy(ra + 32 - rlen, *sig, rlen);
secp256k1_scalar_set_b32(r, ra, &overflow);
}
if (overflow) {
secp256k1_scalar_set_int(r, 0);
}
(*sig) += rlen;
return 1;
}
static int secp256k1_ecdsa_sig_parse(secp256k1_scalar *rr, secp256k1_scalar *rs, const unsigned char *sig, size_t size) {
const unsigned char *sigend = sig + size;
size_t rlen;
if (sig == sigend || *(sig++) != 0x30) {
/* The encoding doesn't start with a constructed sequence (X.690-0207 8.9.1). */
return 0;
}
if (secp256k1_der_read_len(&rlen, &sig, sigend) == 0) {
return 0;
}
if (rlen != (size_t)(sigend - sig)) {
/* Tuple exceeds bounds or garage after tuple. */
return 0;
}
if (!secp256k1_der_parse_integer(rr, &sig, sigend)) {
return 0;
}
if (!secp256k1_der_parse_integer(rs, &sig, sigend)) {
return 0;
}
if (sig != sigend) {
/* Trailing garbage inside tuple. */
return 0;
}
return 1;
}
static int secp256k1_ecdsa_sig_serialize(unsigned char *sig, size_t *size, const secp256k1_scalar* ar, const secp256k1_scalar* as) {
unsigned char r[33] = {0}, s[33] = {0};
unsigned char *rp = r, *sp = s;
size_t lenR = 33, lenS = 33;
secp256k1_scalar_get_b32(&r[1], ar);
secp256k1_scalar_get_b32(&s[1], as);
while (lenR > 1 && rp[0] == 0 && rp[1] < 0x80) { lenR--; rp++; }
while (lenS > 1 && sp[0] == 0 && sp[1] < 0x80) { lenS--; sp++; }
if (*size < 6+lenS+lenR) {
*size = 6 + lenS + lenR;
return 0;
}
*size = 6 + lenS + lenR;
sig[0] = 0x30;
sig[1] = 4 + lenS + lenR;
sig[2] = 0x02;
sig[3] = lenR;
memcpy(sig+4, rp, lenR);
sig[4+lenR] = 0x02;
sig[5+lenR] = lenS;
memcpy(sig+lenR+6, sp, lenS);
return 1;
}
static int secp256k1_ecdsa_sig_verify(const secp256k1_scalar *sigr, const secp256k1_scalar *sigs, const secp256k1_ge *pubkey, const secp256k1_scalar *message) {
unsigned char c[32];
secp256k1_scalar sn, u1, u2;
#if !defined(EXHAUSTIVE_TEST_ORDER)
secp256k1_fe xr;
#endif
secp256k1_gej pubkeyj;
secp256k1_gej pr;
if (secp256k1_scalar_is_zero(sigr) || secp256k1_scalar_is_zero(sigs)) {
return 0;
}
secp256k1_scalar_inverse_var(&sn, sigs);
secp256k1_scalar_mul(&u1, &sn, message);
secp256k1_scalar_mul(&u2, &sn, sigr);
secp256k1_gej_set_ge(&pubkeyj, pubkey);
secp256k1_ecmult(&pr, &pubkeyj, &u2, &u1);
if (secp256k1_gej_is_infinity(&pr)) {
return 0;
}
#if defined(EXHAUSTIVE_TEST_ORDER)
{
secp256k1_scalar computed_r;
secp256k1_ge pr_ge;
secp256k1_ge_set_gej(&pr_ge, &pr);
secp256k1_fe_normalize(&pr_ge.x);
secp256k1_fe_get_b32(c, &pr_ge.x);
secp256k1_scalar_set_b32(&computed_r, c, NULL);
return secp256k1_scalar_eq(sigr, &computed_r);
}
#else
secp256k1_scalar_get_b32(c, sigr);
/* we can ignore the fe_set_b32_limit return value, because we know the input is in range */
(void)secp256k1_fe_set_b32_limit(&xr, c);
/** We now have the recomputed R point in pr, and its claimed x coordinate (modulo n)
* in xr. Naively, we would extract the x coordinate from pr (requiring a inversion modulo p),
* compute the remainder modulo n, and compare it to xr. However:
*
* xr == X(pr) mod n
* <=> exists h. (xr + h * n < p && xr + h * n == X(pr))
* [Since 2 * n > p, h can only be 0 or 1]
* <=> (xr == X(pr)) || (xr + n < p && xr + n == X(pr))
* [In Jacobian coordinates, X(pr) is pr.x / pr.z^2 mod p]
* <=> (xr == pr.x / pr.z^2 mod p) || (xr + n < p && xr + n == pr.x / pr.z^2 mod p)
* [Multiplying both sides of the equations by pr.z^2 mod p]
* <=> (xr * pr.z^2 mod p == pr.x) || (xr + n < p && (xr + n) * pr.z^2 mod p == pr.x)
*
* Thus, we can avoid the inversion, but we have to check both cases separately.
* secp256k1_gej_eq_x implements the (xr * pr.z^2 mod p == pr.x) test.
*/
if (secp256k1_gej_eq_x_var(&xr, &pr)) {
/* xr * pr.z^2 mod p == pr.x, so the signature is valid. */
return 1;
}
if (secp256k1_fe_cmp_var(&xr, &secp256k1_ecdsa_const_p_minus_order) >= 0) {
/* xr + n >= p, so we can skip testing the second case. */
return 0;
}
secp256k1_fe_add(&xr, &secp256k1_ecdsa_const_order_as_fe);
if (secp256k1_gej_eq_x_var(&xr, &pr)) {
/* (xr + n) * pr.z^2 mod p == pr.x, so the signature is valid. */
return 1;
}
return 0;
#endif
}
static int secp256k1_ecdsa_sig_sign(const secp256k1_ecmult_gen_context *ctx, secp256k1_scalar *sigr, secp256k1_scalar *sigs, const secp256k1_scalar *seckey, const secp256k1_scalar *message, const secp256k1_scalar *nonce, int *recid) {
unsigned char b[32];
secp256k1_gej rp;
secp256k1_ge r;
secp256k1_scalar n;
int overflow = 0;
int high;
secp256k1_ecmult_gen(ctx, &rp, nonce);
secp256k1_ge_set_gej(&r, &rp);
secp256k1_fe_normalize(&r.x);
secp256k1_fe_normalize(&r.y);
secp256k1_fe_get_b32(b, &r.x);
secp256k1_scalar_set_b32(sigr, b, &overflow);
if (recid) {
/* The overflow condition is cryptographically unreachable as hitting it requires finding the discrete log
* of some P where P.x >= order, and only 1 in about 2^127 points meet this criteria.
*/
*recid = (overflow << 1) | secp256k1_fe_is_odd(&r.y);
}
secp256k1_scalar_mul(&n, sigr, seckey);
secp256k1_scalar_add(&n, &n, message);
secp256k1_scalar_inverse(sigs, nonce);
secp256k1_scalar_mul(sigs, sigs, &n);
secp256k1_scalar_clear(&n);
secp256k1_gej_clear(&rp);
secp256k1_ge_clear(&r);
high = secp256k1_scalar_is_high(sigs);
secp256k1_scalar_cond_negate(sigs, high);
if (recid) {
*recid ^= high;
}
/* P.x = order is on the curve, so technically sig->r could end up being zero, which would be an invalid signature.
* This is cryptographically unreachable as hitting it requires finding the discrete log of P.x = N.
*/
return (int)(!secp256k1_scalar_is_zero(sigr)) & (int)(!secp256k1_scalar_is_zero(sigs));
}
#endif /* SECP256K1_ECDSA_IMPL_H */

25
libsecp256k1/src/eckey.h Normal file
View File

@@ -0,0 +1,25 @@
/***********************************************************************
* Copyright (c) 2013, 2014 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_ECKEY_H
#define SECP256K1_ECKEY_H
#include <stddef.h>
#include "group.h"
#include "scalar.h"
#include "ecmult.h"
#include "ecmult_gen.h"
static int secp256k1_eckey_pubkey_parse(secp256k1_ge *elem, const unsigned char *pub, size_t size);
static int secp256k1_eckey_pubkey_serialize(secp256k1_ge *elem, unsigned char *pub, size_t *size, int compressed);
static int secp256k1_eckey_privkey_tweak_add(secp256k1_scalar *key, const secp256k1_scalar *tweak);
static int secp256k1_eckey_pubkey_tweak_add(secp256k1_ge *key, const secp256k1_scalar *tweak);
static int secp256k1_eckey_privkey_tweak_mul(secp256k1_scalar *key, const secp256k1_scalar *tweak);
static int secp256k1_eckey_pubkey_tweak_mul(secp256k1_ge *key, const secp256k1_scalar *tweak);
#endif /* SECP256K1_ECKEY_H */

View File

@@ -0,0 +1,92 @@
/***********************************************************************
* Copyright (c) 2013, 2014 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_ECKEY_IMPL_H
#define SECP256K1_ECKEY_IMPL_H
#include "eckey.h"
#include "scalar.h"
#include "field.h"
#include "group.h"
#include "ecmult_gen.h"
static int secp256k1_eckey_pubkey_parse(secp256k1_ge *elem, const unsigned char *pub, size_t size) {
if (size == 33 && (pub[0] == SECP256K1_TAG_PUBKEY_EVEN || pub[0] == SECP256K1_TAG_PUBKEY_ODD)) {
secp256k1_fe x;
return secp256k1_fe_set_b32_limit(&x, pub+1) && secp256k1_ge_set_xo_var(elem, &x, pub[0] == SECP256K1_TAG_PUBKEY_ODD);
} else if (size == 65 && (pub[0] == SECP256K1_TAG_PUBKEY_UNCOMPRESSED || pub[0] == SECP256K1_TAG_PUBKEY_HYBRID_EVEN || pub[0] == SECP256K1_TAG_PUBKEY_HYBRID_ODD)) {
secp256k1_fe x, y;
if (!secp256k1_fe_set_b32_limit(&x, pub+1) || !secp256k1_fe_set_b32_limit(&y, pub+33)) {
return 0;
}
secp256k1_ge_set_xy(elem, &x, &y);
if ((pub[0] == SECP256K1_TAG_PUBKEY_HYBRID_EVEN || pub[0] == SECP256K1_TAG_PUBKEY_HYBRID_ODD) &&
secp256k1_fe_is_odd(&y) != (pub[0] == SECP256K1_TAG_PUBKEY_HYBRID_ODD)) {
return 0;
}
return secp256k1_ge_is_valid_var(elem);
} else {
return 0;
}
}
static int secp256k1_eckey_pubkey_serialize(secp256k1_ge *elem, unsigned char *pub, size_t *size, int compressed) {
if (secp256k1_ge_is_infinity(elem)) {
return 0;
}
secp256k1_fe_normalize_var(&elem->x);
secp256k1_fe_normalize_var(&elem->y);
secp256k1_fe_get_b32(&pub[1], &elem->x);
if (compressed) {
*size = 33;
pub[0] = secp256k1_fe_is_odd(&elem->y) ? SECP256K1_TAG_PUBKEY_ODD : SECP256K1_TAG_PUBKEY_EVEN;
} else {
*size = 65;
pub[0] = SECP256K1_TAG_PUBKEY_UNCOMPRESSED;
secp256k1_fe_get_b32(&pub[33], &elem->y);
}
return 1;
}
static int secp256k1_eckey_privkey_tweak_add(secp256k1_scalar *key, const secp256k1_scalar *tweak) {
secp256k1_scalar_add(key, key, tweak);
return !secp256k1_scalar_is_zero(key);
}
static int secp256k1_eckey_pubkey_tweak_add(secp256k1_ge *key, const secp256k1_scalar *tweak) {
secp256k1_gej pt;
secp256k1_gej_set_ge(&pt, key);
secp256k1_ecmult(&pt, &pt, &secp256k1_scalar_one, tweak);
if (secp256k1_gej_is_infinity(&pt)) {
return 0;
}
secp256k1_ge_set_gej(key, &pt);
return 1;
}
static int secp256k1_eckey_privkey_tweak_mul(secp256k1_scalar *key, const secp256k1_scalar *tweak) {
int ret;
ret = !secp256k1_scalar_is_zero(tweak);
secp256k1_scalar_mul(key, key, tweak);
return ret;
}
static int secp256k1_eckey_pubkey_tweak_mul(secp256k1_ge *key, const secp256k1_scalar *tweak) {
secp256k1_gej pt;
if (secp256k1_scalar_is_zero(tweak)) {
return 0;
}
secp256k1_gej_set_ge(&pt, key);
secp256k1_ecmult(&pt, &pt, tweak, &secp256k1_scalar_zero);
secp256k1_ge_set_gej(key, &pt);
return 1;
}
#endif /* SECP256K1_ECKEY_IMPL_H */

61
libsecp256k1/src/ecmult.h Normal file
View File

@@ -0,0 +1,61 @@
/***********************************************************************
* Copyright (c) 2013, 2014, 2017 Pieter Wuille, Andrew Poelstra *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_ECMULT_H
#define SECP256K1_ECMULT_H
#include "group.h"
#include "scalar.h"
#include "scratch.h"
#ifndef ECMULT_WINDOW_SIZE
# define ECMULT_WINDOW_SIZE 15
# ifdef DEBUG_CONFIG
# pragma message DEBUG_CONFIG_MSG("ECMULT_WINDOW_SIZE undefined, assuming default value")
# endif
#endif
#ifdef DEBUG_CONFIG
# pragma message DEBUG_CONFIG_DEF(ECMULT_WINDOW_SIZE)
#endif
/* No one will ever need more than a window size of 24. The code might
* be correct for larger values of ECMULT_WINDOW_SIZE but this is not
* tested.
*
* The following limitations are known, and there are probably more:
* If WINDOW_G > 27 and size_t has 32 bits, then the code is incorrect
* because the size of the memory object that we allocate (in bytes)
* will not fit in a size_t.
* If WINDOW_G > 31 and int has 32 bits, then the code is incorrect
* because certain expressions will overflow.
*/
#if ECMULT_WINDOW_SIZE < 2 || ECMULT_WINDOW_SIZE > 24
# error Set ECMULT_WINDOW_SIZE to an integer in range [2..24].
#endif
/** The number of entries a table with precomputed multiples needs to have. */
#define ECMULT_TABLE_SIZE(w) (1L << ((w)-2))
/** Double multiply: R = na*A + ng*G */
static void secp256k1_ecmult(secp256k1_gej *r, const secp256k1_gej *a, const secp256k1_scalar *na, const secp256k1_scalar *ng);
typedef int (secp256k1_ecmult_multi_callback)(secp256k1_scalar *sc, secp256k1_ge *pt, size_t idx, void *data);
/**
* Multi-multiply: R = inp_g_sc * G + sum_i ni * Ai.
* Chooses the right algorithm for a given number of points and scratch space
* size. Resets and overwrites the given scratch space. If the points do not
* fit in the scratch space the algorithm is repeatedly run with batches of
* points. If no scratch space is given then a simple algorithm is used that
* simply multiplies the points with the corresponding scalars and adds them up.
* Returns: 1 on success (including when inp_g_sc is NULL and n is 0)
* 0 if there is not enough scratch space for a single point or
* callback returns 0
*/
static int secp256k1_ecmult_multi_var(const secp256k1_callback* error_callback, secp256k1_scratch *scratch, secp256k1_gej *r, const secp256k1_scalar *inp_g_sc, secp256k1_ecmult_multi_callback cb, void *cbdata, size_t n);
#endif /* SECP256K1_ECMULT_H */

View File

@@ -0,0 +1,16 @@
/*****************************************************************************************************
* Copyright (c) 2013, 2014, 2017, 2021 Pieter Wuille, Andrew Poelstra, Jonas Nick, Russell O'Connor *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php. *
*****************************************************************************************************/
#ifndef SECP256K1_ECMULT_COMPUTE_TABLE_H
#define SECP256K1_ECMULT_COMPUTE_TABLE_H
/* Construct table of all odd multiples of gen in range 1..(2**(window_g-1)-1). */
static void secp256k1_ecmult_compute_table(secp256k1_ge_storage* table, int window_g, const secp256k1_gej* gen);
/* Like secp256k1_ecmult_compute_table, but one for both gen and gen*2^128. */
static void secp256k1_ecmult_compute_two_tables(secp256k1_ge_storage* table, secp256k1_ge_storage* table_128, int window_g, const secp256k1_ge* gen);
#endif /* SECP256K1_ECMULT_COMPUTE_TABLE_H */

View File

@@ -0,0 +1,49 @@
/*****************************************************************************************************
* Copyright (c) 2013, 2014, 2017, 2021 Pieter Wuille, Andrew Poelstra, Jonas Nick, Russell O'Connor *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php. *
*****************************************************************************************************/
#ifndef SECP256K1_ECMULT_COMPUTE_TABLE_IMPL_H
#define SECP256K1_ECMULT_COMPUTE_TABLE_IMPL_H
#include "ecmult_compute_table.h"
#include "group_impl.h"
#include "field_impl.h"
#include "ecmult.h"
#include "util.h"
static void secp256k1_ecmult_compute_table(secp256k1_ge_storage* table, int window_g, const secp256k1_gej* gen) {
secp256k1_gej gj;
secp256k1_ge ge, dgen;
int j;
gj = *gen;
secp256k1_ge_set_gej_var(&ge, &gj);
secp256k1_ge_to_storage(&table[0], &ge);
secp256k1_gej_double_var(&gj, gen, NULL);
secp256k1_ge_set_gej_var(&dgen, &gj);
for (j = 1; j < ECMULT_TABLE_SIZE(window_g); ++j) {
secp256k1_gej_set_ge(&gj, &ge);
secp256k1_gej_add_ge_var(&gj, &gj, &dgen, NULL);
secp256k1_ge_set_gej_var(&ge, &gj);
secp256k1_ge_to_storage(&table[j], &ge);
}
}
/* Like secp256k1_ecmult_compute_table, but one for both gen and gen*2^128. */
static void secp256k1_ecmult_compute_two_tables(secp256k1_ge_storage* table, secp256k1_ge_storage* table_128, int window_g, const secp256k1_ge* gen) {
secp256k1_gej gj;
int i;
secp256k1_gej_set_ge(&gj, gen);
secp256k1_ecmult_compute_table(table, window_g, &gj);
for (i = 0; i < 128; ++i) {
secp256k1_gej_double_var(&gj, &gj, NULL);
}
secp256k1_ecmult_compute_table(table_128, window_g, &gj);
}
#endif /* SECP256K1_ECMULT_COMPUTE_TABLE_IMPL_H */

View File

@@ -0,0 +1,38 @@
/***********************************************************************
* Copyright (c) 2015 Andrew Poelstra *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_ECMULT_CONST_H
#define SECP256K1_ECMULT_CONST_H
#include "scalar.h"
#include "group.h"
/**
* Multiply: R = q*A (in constant-time for q)
*/
static void secp256k1_ecmult_const(secp256k1_gej *r, const secp256k1_ge *a, const secp256k1_scalar *q);
/**
* Same as secp256k1_ecmult_const, but takes in an x coordinate of the base point
* only, specified as fraction n/d (numerator/denominator). Only the x coordinate of the result is
* returned.
*
* If known_on_curve is 0, a verification is performed that n/d is a valid X
* coordinate, and 0 is returned if not. Otherwise, 1 is returned.
*
* d being NULL is interpreted as d=1. If non-NULL, d must not be zero. q must not be zero.
*
* Constant time in the value of q, but not any other inputs.
*/
static int secp256k1_ecmult_const_xonly(
secp256k1_fe *r,
const secp256k1_fe *n,
const secp256k1_fe *d,
const secp256k1_scalar *q,
int known_on_curve
);
#endif /* SECP256K1_ECMULT_CONST_H */

View File

@@ -0,0 +1,399 @@
/***********************************************************************
* Copyright (c) 2015, 2022 Pieter Wuille, Andrew Poelstra *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_ECMULT_CONST_IMPL_H
#define SECP256K1_ECMULT_CONST_IMPL_H
#include "scalar.h"
#include "group.h"
#include "ecmult_const.h"
#include "ecmult_impl.h"
#if defined(EXHAUSTIVE_TEST_ORDER)
/* We need 2^ECMULT_CONST_GROUP_SIZE - 1 to be less than EXHAUSTIVE_TEST_ORDER, because
* the tables cannot have infinities in them (this breaks the effective-affine technique's
* z-ratio tracking) */
# if EXHAUSTIVE_TEST_ORDER == 199
# define ECMULT_CONST_GROUP_SIZE 4
# elif EXHAUSTIVE_TEST_ORDER == 13
# define ECMULT_CONST_GROUP_SIZE 3
# elif EXHAUSTIVE_TEST_ORDER == 7
# define ECMULT_CONST_GROUP_SIZE 2
# else
# error "Unknown EXHAUSTIVE_TEST_ORDER"
# endif
#else
/* Group size 4 or 5 appears optimal. */
# define ECMULT_CONST_GROUP_SIZE 5
#endif
#define ECMULT_CONST_TABLE_SIZE (1L << (ECMULT_CONST_GROUP_SIZE - 1))
#define ECMULT_CONST_GROUPS ((129 + ECMULT_CONST_GROUP_SIZE - 1) / ECMULT_CONST_GROUP_SIZE)
#define ECMULT_CONST_BITS (ECMULT_CONST_GROUPS * ECMULT_CONST_GROUP_SIZE)
/** Fill a table 'pre' with precomputed odd multiples of a.
*
* The resulting point set is brought to a single constant Z denominator, stores the X and Y
* coordinates as ge points in pre, and stores the global Z in globalz.
*
* 'pre' must be an array of size ECMULT_CONST_TABLE_SIZE.
*/
static void secp256k1_ecmult_const_odd_multiples_table_globalz(secp256k1_ge *pre, secp256k1_fe *globalz, const secp256k1_gej *a) {
secp256k1_fe zr[ECMULT_CONST_TABLE_SIZE];
secp256k1_ecmult_odd_multiples_table(ECMULT_CONST_TABLE_SIZE, pre, zr, globalz, a);
secp256k1_ge_table_set_globalz(ECMULT_CONST_TABLE_SIZE, pre, zr);
}
/* Given a table 'pre' with odd multiples of a point, put in r the signed-bit multiplication of n with that point.
*
* For example, if ECMULT_CONST_GROUP_SIZE is 4, then pre is expected to contain 8 entries:
* [1*P, 3*P, 5*P, 7*P, 9*P, 11*P, 13*P, 15*P]. n is then expected to be a 4-bit integer (range 0-15), and its
* bits are interpreted as signs of powers of two to look up.
*
* For example, if n=4, which is 0100 in binary, which is interpreted as [- + - -], so the looked up value is
* [ -(2^3) + (2^2) - (2^1) - (2^0) ]*P = -7*P. Every valid n translates to an odd number in range [-15,15],
* which means we just need to look up one of the precomputed values, and optionally negate it.
*/
#define ECMULT_CONST_TABLE_GET_GE(r,pre,n) do { \
unsigned int m = 0; \
/* If the top bit of n is 0, we want the negation. */ \
volatile unsigned int negative = ((n) >> (ECMULT_CONST_GROUP_SIZE - 1)) ^ 1; \
/* Let n[i] be the i-th bit of n, then the index is
* sum(cnot(n[i]) * 2^i, i=0..l-2)
* where cnot(b) = b if n[l-1] = 1 and 1 - b otherwise.
* For example, if n = 4, in binary 0100, the index is 3, in binary 011.
*
* Proof:
* Let
* x = sum((2*n[i] - 1)*2^i, i=0..l-1)
* = 2*sum(n[i] * 2^i, i=0..l-1) - 2^l + 1
* be the value represented by n.
* The index is (x - 1)/2 if x > 0 and -(x + 1)/2 otherwise.
* Case x > 0:
* n[l-1] = 1
* index = sum(n[i] * 2^i, i=0..l-1) - 2^(l-1)
* = sum(n[i] * 2^i, i=0..l-2)
* Case x <= 0:
* n[l-1] = 0
* index = -(2*sum(n[i] * 2^i, i=0..l-1) - 2^l + 2)/2
* = 2^(l-1) - 1 - sum(n[i] * 2^i, i=0..l-1)
* = sum((1 - n[i]) * 2^i, i=0..l-2)
*/ \
unsigned int index = ((unsigned int)(-negative) ^ n) & ((1U << (ECMULT_CONST_GROUP_SIZE - 1)) - 1U); \
secp256k1_fe neg_y; \
VERIFY_CHECK((n) < (1U << ECMULT_CONST_GROUP_SIZE)); \
VERIFY_CHECK(index < (1U << (ECMULT_CONST_GROUP_SIZE - 1))); \
/* Unconditionally set r->x = (pre)[m].x. r->y = (pre)[m].y. because it's either the correct one
* or will get replaced in the later iterations, this is needed to make sure `r` is initialized. */ \
(r)->x = (pre)[m].x; \
(r)->y = (pre)[m].y; \
for (m = 1; m < ECMULT_CONST_TABLE_SIZE; m++) { \
/* This loop is used to avoid secret data in array indices. See
* the comment in ecmult_gen_impl.h for rationale. */ \
secp256k1_fe_cmov(&(r)->x, &(pre)[m].x, m == index); \
secp256k1_fe_cmov(&(r)->y, &(pre)[m].y, m == index); \
} \
(r)->infinity = 0; \
secp256k1_fe_negate(&neg_y, &(r)->y, 1); \
secp256k1_fe_cmov(&(r)->y, &neg_y, negative); \
} while(0)
/* For K as defined in the comment of secp256k1_ecmult_const, we have several precomputed
* formulas/constants.
* - in exhaustive test mode, we give an explicit expression to compute it at compile time: */
#ifdef EXHAUSTIVE_TEST_ORDER
static const secp256k1_scalar secp256k1_ecmult_const_K = ((SECP256K1_SCALAR_CONST(0, 0, 0, (1U << (ECMULT_CONST_BITS - 128)) - 2U, 0, 0, 0, 0) + EXHAUSTIVE_TEST_ORDER - 1U) * (1U + EXHAUSTIVE_TEST_LAMBDA)) % EXHAUSTIVE_TEST_ORDER;
/* - for the real secp256k1 group we have constants for various ECMULT_CONST_BITS values. */
#elif ECMULT_CONST_BITS == 129
/* For GROUP_SIZE = 1,3. */
static const secp256k1_scalar secp256k1_ecmult_const_K = SECP256K1_SCALAR_CONST(0xac9c52b3ul, 0x3fa3cf1ful, 0x5ad9e3fdul, 0x77ed9ba4ul, 0xa880b9fcul, 0x8ec739c2ul, 0xe0cfc810ul, 0xb51283ceul);
#elif ECMULT_CONST_BITS == 130
/* For GROUP_SIZE = 2,5. */
static const secp256k1_scalar secp256k1_ecmult_const_K = SECP256K1_SCALAR_CONST(0xa4e88a7dul, 0xcb13034eul, 0xc2bdd6bful, 0x7c118d6bul, 0x589ae848ul, 0x26ba29e4ul, 0xb5c2c1dcul, 0xde9798d9ul);
#elif ECMULT_CONST_BITS == 132
/* For GROUP_SIZE = 4,6 */
static const secp256k1_scalar secp256k1_ecmult_const_K = SECP256K1_SCALAR_CONST(0x76b1d93dul, 0x0fae3c6bul, 0x3215874bul, 0x94e93813ul, 0x7937fe0dul, 0xb66bcaaful, 0xb3749ca5ul, 0xd7b6171bul);
#else
# error "Unknown ECMULT_CONST_BITS"
#endif
static void secp256k1_ecmult_const(secp256k1_gej *r, const secp256k1_ge *a, const secp256k1_scalar *q) {
/* The approach below combines the signed-digit logic from Mike Hamburg's
* "Fast and compact elliptic-curve cryptography" (https://eprint.iacr.org/2012/309)
* Section 3.3, with the GLV endomorphism.
*
* The idea there is to interpret the bits of a scalar as signs (1 = +, 0 = -), and compute a
* point multiplication in that fashion. Let v be an n-bit non-negative integer (0 <= v < 2^n),
* and v[i] its i'th bit (so v = sum(v[i] * 2^i, i=0..n-1)). Then define:
*
* C_l(v, A) = sum((2*v[i] - 1) * 2^i*A, i=0..l-1)
*
* Then it holds that C_l(v, A) = sum((2*v[i] - 1) * 2^i*A, i=0..l-1)
* = (2*sum(v[i] * 2^i, i=0..l-1) + 1 - 2^l) * A
* = (2*v + 1 - 2^l) * A
*
* Thus, one can compute q*A as C_256((q + 2^256 - 1) / 2, A). This is the basis for the
* paper's signed-digit multi-comb algorithm for multiplication using a precomputed table.
*
* It is appealing to try to combine this with the GLV optimization: the idea that a scalar
* s can be written as s1 + lambda*s2, where lambda is a curve-specific constant such that
* lambda*A is easy to compute, and where s1 and s2 are small. In particular we have the
* secp256k1_scalar_split_lambda function which performs such a split with the resulting s1
* and s2 in range (-2^128, 2^128) mod n. This does work, but is uninteresting:
*
* To compute q*A:
* - Let s1, s2 = split_lambda(q)
* - Let R1 = C_256((s1 + 2^256 - 1) / 2, A)
* - Let R2 = C_256((s2 + 2^256 - 1) / 2, lambda*A)
* - Return R1 + R2
*
* The issue is that while s1 and s2 are small-range numbers, (s1 + 2^256 - 1) / 2 (mod n)
* and (s2 + 2^256 - 1) / 2 (mod n) are not, undoing the benefit of the splitting.
*
* To make it work, we want to modify the input scalar q first, before splitting, and then only
* add a 2^128 offset of the split results (so that they end up in the single 129-bit range
* [0,2^129]). A slightly smaller offset would work due to the bounds on the split, but we pick
* 2^128 for simplicity. Let s be the scalar fed to split_lambda, and f(q) the function to
* compute it from q:
*
* To compute q*A:
* - Compute s = f(q)
* - Let s1, s2 = split_lambda(s)
* - Let v1 = s1 + 2^128 (mod n)
* - Let v2 = s2 + 2^128 (mod n)
* - Let R1 = C_l(v1, A)
* - Let R2 = C_l(v2, lambda*A)
* - Return R1 + R2
*
* l will thus need to be at least 129, but we may overshoot by a few bits (see
* further), so keep it as a variable.
*
* To solve for s, we reason:
* q*A = R1 + R2
* <=> q*A = C_l(s1 + 2^128, A) + C_l(s2 + 2^128, lambda*A)
* <=> q*A = (2*(s1 + 2^128) + 1 - 2^l) * A + (2*(s2 + 2^128) + 1 - 2^l) * lambda*A
* <=> q*A = (2*(s1 + s2*lambda) + (2^129 + 1 - 2^l) * (1 + lambda)) * A
* <=> q = 2*(s1 + s2*lambda) + (2^129 + 1 - 2^l) * (1 + lambda) (mod n)
* <=> q = 2*s + (2^129 + 1 - 2^l) * (1 + lambda) (mod n)
* <=> s = (q + (2^l - 2^129 - 1) * (1 + lambda)) / 2 (mod n)
* <=> f(q) = (q + K) / 2 (mod n)
* where K = (2^l - 2^129 - 1)*(1 + lambda) (mod n)
*
* We will process the computation of C_l(v1, A) and C_l(v2, lambda*A) in groups of
* ECMULT_CONST_GROUP_SIZE, so we set l to the smallest multiple of ECMULT_CONST_GROUP_SIZE
* that is not less than 129; this equals ECMULT_CONST_BITS.
*/
/* The offset to add to s1 and s2 to make them non-negative. Equal to 2^128. */
static const secp256k1_scalar S_OFFSET = SECP256K1_SCALAR_CONST(0, 0, 0, 1, 0, 0, 0, 0);
secp256k1_scalar s, v1, v2;
secp256k1_ge pre_a[ECMULT_CONST_TABLE_SIZE];
secp256k1_ge pre_a_lam[ECMULT_CONST_TABLE_SIZE];
secp256k1_fe global_z;
int group, i;
/* We're allowed to be non-constant time in the point, and the code below (in particular,
* secp256k1_ecmult_const_odd_multiples_table_globalz) cannot deal with infinity in a
* constant-time manner anyway. */
if (secp256k1_ge_is_infinity(a)) {
secp256k1_gej_set_infinity(r);
return;
}
/* Compute v1 and v2. */
secp256k1_scalar_add(&s, q, &secp256k1_ecmult_const_K);
secp256k1_scalar_half(&s, &s);
secp256k1_scalar_split_lambda(&v1, &v2, &s);
secp256k1_scalar_add(&v1, &v1, &S_OFFSET);
secp256k1_scalar_add(&v2, &v2, &S_OFFSET);
#ifdef VERIFY
/* Verify that v1 and v2 are in range [0, 2^129-1]. */
for (i = 129; i < 256; ++i) {
VERIFY_CHECK(secp256k1_scalar_get_bits_limb32(&v1, i, 1) == 0);
VERIFY_CHECK(secp256k1_scalar_get_bits_limb32(&v2, i, 1) == 0);
}
#endif
/* Calculate odd multiples of A and A*lambda.
* All multiples are brought to the same Z 'denominator', which is stored
* in global_z. Due to secp256k1' isomorphism we can do all operations pretending
* that the Z coordinate was 1, use affine addition formulae, and correct
* the Z coordinate of the result once at the end.
*/
secp256k1_gej_set_ge(r, a);
secp256k1_ecmult_const_odd_multiples_table_globalz(pre_a, &global_z, r);
for (i = 0; i < ECMULT_CONST_TABLE_SIZE; i++) {
secp256k1_ge_mul_lambda(&pre_a_lam[i], &pre_a[i]);
}
/* Next, we compute r = C_l(v1, A) + C_l(v2, lambda*A).
*
* We proceed in groups of ECMULT_CONST_GROUP_SIZE bits, operating on that many bits
* at a time, from high in v1, v2 to low. Call these bits1 (from v1) and bits2 (from v2).
*
* Now note that ECMULT_CONST_TABLE_GET_GE(&t, pre_a, bits1) loads into t a point equal
* to C_{ECMULT_CONST_GROUP_SIZE}(bits1, A), and analogously for pre_lam_a / bits2.
* This means that all we need to do is add these looked up values together, multiplied
* by 2^(ECMULT_GROUP_SIZE * group).
*/
for (group = ECMULT_CONST_GROUPS - 1; group >= 0; --group) {
/* Using the _var get_bits function is ok here, since it's only variable in offset and count, not in the scalar. */
unsigned int bits1 = secp256k1_scalar_get_bits_var(&v1, group * ECMULT_CONST_GROUP_SIZE, ECMULT_CONST_GROUP_SIZE);
unsigned int bits2 = secp256k1_scalar_get_bits_var(&v2, group * ECMULT_CONST_GROUP_SIZE, ECMULT_CONST_GROUP_SIZE);
secp256k1_ge t;
int j;
ECMULT_CONST_TABLE_GET_GE(&t, pre_a, bits1);
if (group == ECMULT_CONST_GROUPS - 1) {
/* Directly set r in the first iteration. */
secp256k1_gej_set_ge(r, &t);
} else {
/* Shift the result so far up. */
for (j = 0; j < ECMULT_CONST_GROUP_SIZE; ++j) {
secp256k1_gej_double(r, r);
}
secp256k1_gej_add_ge(r, r, &t);
}
ECMULT_CONST_TABLE_GET_GE(&t, pre_a_lam, bits2);
secp256k1_gej_add_ge(r, r, &t);
}
/* Map the result back to the secp256k1 curve from the isomorphic curve. */
secp256k1_fe_mul(&r->z, &r->z, &global_z);
}
static int secp256k1_ecmult_const_xonly(secp256k1_fe* r, const secp256k1_fe *n, const secp256k1_fe *d, const secp256k1_scalar *q, int known_on_curve) {
/* This algorithm is a generalization of Peter Dettman's technique for
* avoiding the square root in a random-basepoint x-only multiplication
* on a Weierstrass curve:
* https://mailarchive.ietf.org/arch/msg/cfrg/7DyYY6gg32wDgHAhgSb6XxMDlJA/
*
*
* === Background: the effective affine technique ===
*
* Let phi_u be the isomorphism that maps (x, y) on secp256k1 curve y^2 = x^3 + 7 to
* x' = u^2*x, y' = u^3*y on curve y'^2 = x'^3 + u^6*7. This new curve has the same order as
* the original (it is isomorphic), but moreover, has the same addition/doubling formulas, as
* the curve b=7 coefficient does not appear in those formulas (or at least does not appear in
* the formulas implemented in this codebase, both affine and Jacobian). See also Example 9.5.2
* in https://www.math.auckland.ac.nz/~sgal018/crypto-book/ch9.pdf.
*
* This means any linear combination of secp256k1 points can be computed by applying phi_u
* (with non-zero u) on all input points (including the generator, if used), computing the
* linear combination on the isomorphic curve (using the same group laws), and then applying
* phi_u^{-1} to get back to secp256k1.
*
* Switching to Jacobian coordinates, note that phi_u applied to (X, Y, Z) is simply
* (X, Y, Z/u). Thus, if we want to compute (X1, Y1, Z) + (X2, Y2, Z), with identical Z
* coordinates, we can use phi_Z to transform it to (X1, Y1, 1) + (X2, Y2, 1) on an isomorphic
* curve where the affine addition formula can be used instead.
* If (X3, Y3, Z3) = (X1, Y1) + (X2, Y2) on that curve, then our answer on secp256k1 is
* (X3, Y3, Z3*Z).
*
* This is the effective affine technique: if we have a linear combination of group elements
* to compute, and all those group elements have the same Z coordinate, we can simply pretend
* that all those Z coordinates are 1, perform the computation that way, and then multiply the
* original Z coordinate back in.
*
* The technique works on any a=0 short Weierstrass curve. It is possible to generalize it to
* other curves too, but there the isomorphic curves will have different 'a' coefficients,
* which typically does affect the group laws.
*
*
* === Avoiding the square root for x-only point multiplication ===
*
* In this function, we want to compute the X coordinate of q*(n/d, y), for
* y = sqrt((n/d)^3 + 7). Its negation would also be a valid Y coordinate, but by convention
* we pick whatever sqrt returns (which we assume to be a deterministic function).
*
* Let g = y^2*d^3 = n^3 + 7*d^3. This also means y = sqrt(g/d^3).
* Further let v = sqrt(d*g), which must exist as d*g = y^2*d^4 = (y*d^2)^2.
*
* The input point (n/d, y) also has Jacobian coordinates:
*
* (n/d, y, 1)
* = (n/d * v^2, y * v^3, v)
* = (n/d * d*g, y * sqrt(d^3*g^3), v)
* = (n/d * d*g, sqrt(y^2 * d^3*g^3), v)
* = (n*g, sqrt(g/d^3 * d^3*g^3), v)
* = (n*g, sqrt(g^4), v)
* = (n*g, g^2, v)
*
* It is easy to verify that both (n*g, g^2, v) and its negation (n*g, -g^2, v) have affine X
* coordinate n/d, and this holds even when the square root function doesn't have a
* deterministic sign. We choose the (n*g, g^2, v) version.
*
* Now switch to the effective affine curve using phi_v, where the input point has coordinates
* (n*g, g^2). Compute (X, Y, Z) = q * (n*g, g^2) there.
*
* Back on secp256k1, that means q * (n*g, g^2, v) = (X, Y, v*Z). This last point has affine X
* coordinate X / (v^2*Z^2) = X / (d*g*Z^2). Determining the affine Y coordinate would involve
* a square root, but as long as we only care about the resulting X coordinate, no square root
* is needed anywhere in this computation.
*/
secp256k1_fe g, i;
secp256k1_ge p;
secp256k1_gej rj;
/* Compute g = (n^3 + B*d^3). */
secp256k1_fe_sqr(&g, n);
secp256k1_fe_mul(&g, &g, n);
if (d) {
secp256k1_fe b;
VERIFY_CHECK(!secp256k1_fe_normalizes_to_zero(d));
secp256k1_fe_sqr(&b, d);
VERIFY_CHECK(SECP256K1_B <= 8); /* magnitude of b will be <= 8 after the next call */
secp256k1_fe_mul_int(&b, SECP256K1_B);
secp256k1_fe_mul(&b, &b, d);
secp256k1_fe_add(&g, &b);
if (!known_on_curve) {
/* We need to determine whether (n/d)^3 + 7 is square.
*
* is_square((n/d)^3 + 7)
* <=> is_square(((n/d)^3 + 7) * d^4)
* <=> is_square((n^3 + 7*d^3) * d)
* <=> is_square(g * d)
*/
secp256k1_fe c;
secp256k1_fe_mul(&c, &g, d);
if (!secp256k1_fe_is_square_var(&c)) return 0;
}
} else {
secp256k1_fe_add_int(&g, SECP256K1_B);
if (!known_on_curve) {
/* g at this point equals x^3 + 7. Test if it is square. */
if (!secp256k1_fe_is_square_var(&g)) return 0;
}
}
/* Compute base point P = (n*g, g^2), the effective affine version of (n*g, g^2, v), which has
* corresponding affine X coordinate n/d. */
secp256k1_fe_mul(&p.x, &g, n);
secp256k1_fe_sqr(&p.y, &g);
p.infinity = 0;
/* Perform x-only EC multiplication of P with q. */
VERIFY_CHECK(!secp256k1_scalar_is_zero(q));
secp256k1_ecmult_const(&rj, &p, q);
VERIFY_CHECK(!secp256k1_gej_is_infinity(&rj));
/* The resulting (X, Y, Z) point on the effective-affine isomorphic curve corresponds to
* (X, Y, Z*v) on the secp256k1 curve. The affine version of that has X coordinate
* (X / (Z^2*d*g)). */
secp256k1_fe_sqr(&i, &rj.z);
secp256k1_fe_mul(&i, &i, &g);
if (d) secp256k1_fe_mul(&i, &i, d);
secp256k1_fe_inv(&i, &i);
secp256k1_fe_mul(r, &rj.x, &i);
return 1;
}
#endif /* SECP256K1_ECMULT_CONST_IMPL_H */

View File

@@ -0,0 +1,143 @@
/***********************************************************************
* Copyright (c) Pieter Wuille, Peter Dettman *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_ECMULT_GEN_H
#define SECP256K1_ECMULT_GEN_H
#include "scalar.h"
#include "group.h"
/* Configuration parameters for the signed-digit multi-comb algorithm:
*
* - COMB_BLOCKS is the number of blocks the input is split into. Each
* has a corresponding table.
* - COMB_TEETH is the number of bits simultaneously covered by one table.
* - COMB_RANGE is the number of bits in supported scalars. For production
* purposes, only 256 is reasonable, but smaller numbers are supported for
* exhaustive test mode.
*
* The comb's spacing (COMB_SPACING), or the distance between the teeth,
* is defined as ceil(COMB_RANGE / (COMB_BLOCKS * COMB_TEETH)). Each block covers
* COMB_SPACING * COMB_TEETH consecutive bits in the input.
*
* The size of the precomputed table is COMB_BLOCKS * (1 << (COMB_TEETH - 1))
* secp256k1_ge_storages.
*
* The number of point additions equals COMB_BLOCKS * COMB_SPACING. Each point
* addition involves a cmov from (1 << (COMB_TEETH - 1)) table entries and a
* conditional negation.
*
* The number of point doublings is COMB_SPACING - 1. */
#if defined(EXHAUSTIVE_TEST_ORDER)
/* We need to control these values for exhaustive tests because
* the table cannot have infinities in them (secp256k1_ge_storage
* doesn't support infinities) */
# undef COMB_BLOCKS
# undef COMB_TEETH
# if EXHAUSTIVE_TEST_ORDER == 7
# define COMB_RANGE 3
# define COMB_BLOCKS 1
# define COMB_TEETH 2
# elif EXHAUSTIVE_TEST_ORDER == 13
# define COMB_RANGE 4
# define COMB_BLOCKS 1
# define COMB_TEETH 2
# elif EXHAUSTIVE_TEST_ORDER == 199
# define COMB_RANGE 8
# define COMB_BLOCKS 2
# define COMB_TEETH 3
# else
# error "Unknown exhaustive test order"
# endif
# if (COMB_RANGE >= 32) || ((EXHAUSTIVE_TEST_ORDER >> (COMB_RANGE - 1)) != 1)
# error "COMB_RANGE != ceil(log2(EXHAUSTIVE_TEST_ORDER+1))"
# endif
#else /* !defined(EXHAUSTIVE_TEST_ORDER) */
# define COMB_RANGE 256
#endif /* defined(EXHAUSTIVE_TEST_ORDER) */
/* Use (11, 6) as default configuration, which results in a 22 kB table. */
#ifndef COMB_BLOCKS
# define COMB_BLOCKS 11
# ifdef DEBUG_CONFIG
# pragma message DEBUG_CONFIG_MSG("COMB_BLOCKS undefined, assuming default value")
# endif
#endif
#ifndef COMB_TEETH
# define COMB_TEETH 6
# ifdef DEBUG_CONFIG
# pragma message DEBUG_CONFIG_MSG("COMB_TEETH undefined, assuming default value")
# endif
#endif
/* Use ceil(COMB_RANGE / (COMB_BLOCKS * COMB_TEETH)) as COMB_SPACING. */
#define COMB_SPACING CEIL_DIV(COMB_RANGE, COMB_BLOCKS * COMB_TEETH)
/* Range checks on the parameters. */
/* The remaining COMB_* parameters are derived values, don't modify these. */
/* - The number of bits covered by all the blocks; must be at least COMB_RANGE. */
#define COMB_BITS (COMB_BLOCKS * COMB_TEETH * COMB_SPACING)
/* - The number of entries per table. */
#define COMB_POINTS (1 << (COMB_TEETH - 1))
/* Sanity checks. */
#if !(1 <= COMB_BLOCKS && COMB_BLOCKS <= 256)
# error "COMB_BLOCKS must be in the range [1, 256]"
#endif
#if !(1 <= COMB_TEETH && COMB_TEETH <= 8)
# error "COMB_TEETH must be in the range [1, 8]"
#endif
#if COMB_BITS < COMB_RANGE
# error "COMB_BLOCKS * COMB_TEETH * COMB_SPACING is too low"
#endif
/* These last 2 checks are not strictly required, but prevent gratuitously inefficient
* configurations. Note that they compare with 256 rather than COMB_RANGE, so they do
* permit somewhat excessive values for the exhaustive test case, where testing with
* suboptimal parameters may be desirable. */
#if (COMB_BLOCKS - 1) * COMB_TEETH * COMB_SPACING >= 256
# error "COMB_BLOCKS can be reduced"
#endif
#if COMB_BLOCKS * (COMB_TEETH - 1) * COMB_SPACING >= 256
# error "COMB_TEETH can be reduced"
#endif
#ifdef DEBUG_CONFIG
# pragma message DEBUG_CONFIG_DEF(COMB_RANGE)
# pragma message DEBUG_CONFIG_DEF(COMB_BLOCKS)
# pragma message DEBUG_CONFIG_DEF(COMB_TEETH)
# pragma message DEBUG_CONFIG_DEF(COMB_SPACING)
#endif
typedef struct {
/* Whether the context has been built. */
int built;
/* Values chosen such that
*
* n*G == comb(n + scalar_offset, G/2) + ge_offset.
*
* This expression lets us use scalar blinding and optimize the comb precomputation. See
* ecmult_gen_impl.h for more details. */
secp256k1_scalar scalar_offset;
secp256k1_ge ge_offset;
/* Factor used for projective blinding. This value is used to rescale the Z
* coordinate of the first table lookup. */
secp256k1_fe proj_blind;
} secp256k1_ecmult_gen_context;
static void secp256k1_ecmult_gen_context_build(secp256k1_ecmult_gen_context* ctx);
static void secp256k1_ecmult_gen_context_clear(secp256k1_ecmult_gen_context* ctx);
/** Multiply with the generator: R = a*G */
static void secp256k1_ecmult_gen(const secp256k1_ecmult_gen_context* ctx, secp256k1_gej *r, const secp256k1_scalar *a);
static void secp256k1_ecmult_gen_blind(secp256k1_ecmult_gen_context *ctx, const unsigned char *seed32);
#endif /* SECP256K1_ECMULT_GEN_H */

View File

@@ -0,0 +1,14 @@
/***********************************************************************
* Copyright (c) Pieter Wuille, Gregory Maxwell *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_ECMULT_GEN_COMPUTE_TABLE_H
#define SECP256K1_ECMULT_GEN_COMPUTE_TABLE_H
#include "ecmult_gen.h"
static void secp256k1_ecmult_gen_compute_table(secp256k1_ge_storage* table, const secp256k1_ge* gen, int blocks, int teeth, int spacing);
#endif /* SECP256K1_ECMULT_GEN_COMPUTE_TABLE_H */

View File

@@ -0,0 +1,108 @@
/***********************************************************************
* Copyright (c) Pieter Wuille, Gregory Maxwell, Peter Dettman *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_ECMULT_GEN_COMPUTE_TABLE_IMPL_H
#define SECP256K1_ECMULT_GEN_COMPUTE_TABLE_IMPL_H
#include "ecmult_gen_compute_table.h"
#include "group_impl.h"
#include "field_impl.h"
#include "scalar_impl.h"
#include "ecmult_gen.h"
#include "util.h"
static void secp256k1_ecmult_gen_compute_table(secp256k1_ge_storage* table, const secp256k1_ge* gen, int blocks, int teeth, int spacing) {
size_t points = ((size_t)1) << (teeth - 1);
size_t points_total = points * blocks;
secp256k1_ge* prec = checked_malloc(&default_error_callback, points_total * sizeof(*prec));
secp256k1_gej* ds = checked_malloc(&default_error_callback, teeth * sizeof(*ds));
secp256k1_gej* vs = checked_malloc(&default_error_callback, points_total * sizeof(*vs));
secp256k1_gej u;
size_t vs_pos = 0;
secp256k1_scalar half;
int block, i;
VERIFY_CHECK(points_total > 0);
/* u is the running power of two times gen we're working with, initially gen/2. */
secp256k1_scalar_half(&half, &secp256k1_scalar_one);
secp256k1_gej_set_infinity(&u);
for (i = 255; i >= 0; --i) {
/* Use a very simple multiplication ladder to avoid dependency on ecmult. */
secp256k1_gej_double_var(&u, &u, NULL);
if (secp256k1_scalar_get_bits_limb32(&half, i, 1)) {
secp256k1_gej_add_ge_var(&u, &u, gen, NULL);
}
}
#ifdef VERIFY
{
/* Verify that u*2 = gen. */
secp256k1_gej double_u;
secp256k1_gej_double_var(&double_u, &u, NULL);
VERIFY_CHECK(secp256k1_gej_eq_ge_var(&double_u, gen));
}
#endif
for (block = 0; block < blocks; ++block) {
int tooth;
/* Here u = 2^(block*teeth*spacing) * gen/2. */
secp256k1_gej sum;
secp256k1_gej_set_infinity(&sum);
for (tooth = 0; tooth < teeth; ++tooth) {
/* Here u = 2^((block*teeth + tooth)*spacing) * gen/2. */
/* Make sum = sum(2^((block*teeth + t)*spacing), t=0..tooth) * gen/2. */
secp256k1_gej_add_var(&sum, &sum, &u, NULL);
/* Make u = 2^((block*teeth + tooth)*spacing + 1) * gen/2. */
secp256k1_gej_double_var(&u, &u, NULL);
/* Make ds[tooth] = u = 2^((block*teeth + tooth)*spacing + 1) * gen/2. */
ds[tooth] = u;
/* Make u = 2^((block*teeth + tooth + 1)*spacing) * gen/2, unless at the end. */
if (block + tooth != blocks + teeth - 2) {
int bit_off;
for (bit_off = 1; bit_off < spacing; ++bit_off) {
secp256k1_gej_double_var(&u, &u, NULL);
}
}
}
/* Now u = 2^((block*teeth + teeth)*spacing) * gen/2
* = 2^((block+1)*teeth*spacing) * gen/2 */
/* Next, compute the table entries for block number block in Jacobian coordinates.
* The entries will occupy vs[block*points + i] for i=0..points-1.
* We start by computing the first (i=0) value corresponding to all summed
* powers of two times G being negative. */
secp256k1_gej_neg(&vs[vs_pos++], &sum);
/* And then teeth-1 times "double" the range of i values for which the table
* is computed: in each iteration, double the table by taking an existing
* table entry and adding ds[tooth]. */
for (tooth = 0; tooth < teeth - 1; ++tooth) {
size_t stride = ((size_t)1) << tooth;
size_t index;
for (index = 0; index < stride; ++index, ++vs_pos) {
secp256k1_gej_add_var(&vs[vs_pos], &vs[vs_pos - stride], &ds[tooth], NULL);
}
}
}
VERIFY_CHECK(vs_pos == points_total);
/* Convert all points simultaneously from secp256k1_gej to secp256k1_ge. */
secp256k1_ge_set_all_gej_var(prec, vs, points_total);
/* Convert all points from secp256k1_ge to secp256k1_ge_storage output. */
for (block = 0; block < blocks; ++block) {
size_t index;
for (index = 0; index < points; ++index) {
VERIFY_CHECK(!secp256k1_ge_is_infinity(&prec[block * points + index]));
secp256k1_ge_to_storage(&table[block * points + index], &prec[block * points + index]);
}
}
/* Free memory. */
free(vs);
free(ds);
free(prec);
}
#endif /* SECP256K1_ECMULT_GEN_COMPUTE_TABLE_IMPL_H */

View File

@@ -0,0 +1,341 @@
/***********************************************************************
* Copyright (c) Pieter Wuille, Gregory Maxwell, Peter Dettman *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_ECMULT_GEN_IMPL_H
#define SECP256K1_ECMULT_GEN_IMPL_H
#include "util.h"
#include "scalar.h"
#include "group.h"
#include "ecmult_gen.h"
#include "hash_impl.h"
#include "precomputed_ecmult_gen.h"
static void secp256k1_ecmult_gen_context_build(secp256k1_ecmult_gen_context *ctx) {
secp256k1_ecmult_gen_blind(ctx, NULL);
ctx->built = 1;
}
static int secp256k1_ecmult_gen_context_is_built(const secp256k1_ecmult_gen_context* ctx) {
return ctx->built;
}
static void secp256k1_ecmult_gen_context_clear(secp256k1_ecmult_gen_context *ctx) {
ctx->built = 0;
secp256k1_scalar_clear(&ctx->scalar_offset);
secp256k1_ge_clear(&ctx->ge_offset);
secp256k1_fe_clear(&ctx->proj_blind);
}
/* Compute the scalar (2^COMB_BITS - 1) / 2, the difference between the gn argument to
* secp256k1_ecmult_gen, and the scalar whose encoding the table lookup bits are drawn
* from (before applying blinding). */
static void secp256k1_ecmult_gen_scalar_diff(secp256k1_scalar* diff) {
int i;
/* Compute scalar -1/2. */
secp256k1_scalar neghalf;
secp256k1_scalar_half(&neghalf, &secp256k1_scalar_one);
secp256k1_scalar_negate(&neghalf, &neghalf);
/* Compute offset = 2^(COMB_BITS - 1). */
*diff = secp256k1_scalar_one;
for (i = 0; i < COMB_BITS - 1; ++i) {
secp256k1_scalar_add(diff, diff, diff);
}
/* The result is the sum 2^(COMB_BITS - 1) + (-1/2). */
secp256k1_scalar_add(diff, diff, &neghalf);
}
static void secp256k1_ecmult_gen(const secp256k1_ecmult_gen_context *ctx, secp256k1_gej *r, const secp256k1_scalar *gn) {
uint32_t comb_off;
secp256k1_ge add;
secp256k1_fe neg;
secp256k1_ge_storage adds;
secp256k1_scalar d;
/* Array of uint32_t values large enough to store COMB_BITS bits. Only the bottom
* 8 are ever nonzero, but having the zero padding at the end if COMB_BITS>256
* avoids the need to deal with out-of-bounds reads from a scalar. */
uint32_t recoded[(COMB_BITS + 31) >> 5] = {0};
int first = 1, i;
memset(&adds, 0, sizeof(adds));
/* We want to compute R = gn*G.
*
* To blind the scalar used in the computation, we rewrite this to be
* R = (gn - b)*G + b*G, with a blinding value b determined by the context.
*
* The multiplication (gn-b)*G will be performed using a signed-digit multi-comb (see Section
* 3.3 of "Fast and compact elliptic-curve cryptography" by Mike Hamburg,
* https://eprint.iacr.org/2012/309).
*
* Let comb(s, P) = sum((2*s[i]-1)*2^i*P for i=0..COMB_BITS-1), where s[i] is the i'th bit of
* the binary representation of scalar s. So the s[i] values determine whether -2^i*P (s[i]=0)
* or +2^i*P (s[i]=1) are added together. COMB_BITS is at least 256, so all bits of s are
* covered. By manipulating:
*
* comb(s, P) = sum((2*s[i]-1)*2^i*P for i=0..COMB_BITS-1)
* <=> comb(s, P) = sum((2*s[i]-1)*2^i for i=0..COMB_BITS-1) * P
* <=> comb(s, P) = (2*sum(s[i]*2^i for i=0..COMB_BITS-1) - sum(2^i for i=0..COMB_BITS-1)) * P
* <=> comb(s, P) = (2*s - (2^COMB_BITS - 1)) * P
*
* If we wanted to compute (gn-b)*G as comb(s, G), it would need to hold that
*
* (gn - b) * G = (2*s - (2^COMB_BITS - 1)) * G
* <=> s = (gn - b + (2^COMB_BITS - 1))/2 (mod order)
*
* We use an alternative here that avoids the modular division by two: instead we compute
* (gn-b)*G as comb(d, G/2). For that to hold it must be the case that
*
* (gn - b) * G = (2*d - (2^COMB_BITS - 1)) * (G/2)
* <=> d = gn - b + (2^COMB_BITS - 1)/2 (mod order)
*
* Adding precomputation, our final equations become:
*
* ctx->scalar_offset = (2^COMB_BITS - 1)/2 - b (mod order)
* ctx->ge_offset = b*G
* d = gn + ctx->scalar_offset (mod order)
* R = comb(d, G/2) + ctx->ge_offset
*
* comb(d, G/2) function is then computed by summing + or - 2^(i-1)*G, for i=0..COMB_BITS-1,
* depending on the value of the bits d[i] of the binary representation of scalar d.
*/
/* Compute the scalar d = (gn + ctx->scalar_offset). */
secp256k1_scalar_add(&d, &ctx->scalar_offset, gn);
/* Convert to recoded array. */
for (i = 0; i < 8 && i < ((COMB_BITS + 31) >> 5); ++i) {
recoded[i] = secp256k1_scalar_get_bits_limb32(&d, 32 * i, 32);
}
secp256k1_scalar_clear(&d);
/* In secp256k1_ecmult_gen_prec_table we have precomputed sums of the
* (2*d[i]-1) * 2^(i-1) * G points, for various combinations of i positions.
* We rewrite our equation in terms of these table entries.
*
* Let mask(b) = sum(2^((b*COMB_TEETH + t)*COMB_SPACING) for t=0..COMB_TEETH-1),
* with b ranging from 0 to COMB_BLOCKS-1. So for example with COMB_BLOCKS=11,
* COMB_TEETH=6, COMB_SPACING=4, we would have:
* mask(0) = 2^0 + 2^4 + 2^8 + 2^12 + 2^16 + 2^20,
* mask(1) = 2^24 + 2^28 + 2^32 + 2^36 + 2^40 + 2^44,
* mask(2) = 2^48 + 2^52 + 2^56 + 2^60 + 2^64 + 2^68,
* ...
* mask(10) = 2^240 + 2^244 + 2^248 + 2^252 + 2^256 + 2^260
*
* We will split up the bits d[i] using these masks. Specifically, each mask is
* used COMB_SPACING times, with different shifts:
*
* d = (d & mask(0)<<0) + (d & mask(1)<<0) + ... + (d & mask(COMB_BLOCKS-1)<<0) +
* (d & mask(0)<<1) + (d & mask(1)<<1) + ... + (d & mask(COMB_BLOCKS-1)<<1) +
* ...
* (d & mask(0)<<(COMB_SPACING-1)) + ...
*
* Now define table(b, m) = (m - mask(b)/2) * G, and we will precompute these values for
* b=0..COMB_BLOCKS-1, and for all values m which (d & mask(b)) can take (so m can take on
* 2^COMB_TEETH distinct values).
*
* If m=(d & mask(b)), then table(b, m) is the sum of 2^i * (2*d[i]-1) * G/2, with i
* iterating over the set bits in mask(b). In our example, table(2, 2^48 + 2^56 + 2^68)
* would equal (2^48 - 2^52 + 2^56 - 2^60 - 2^64 + 2^68) * G/2.
*
* With that, we can rewrite comb(d, G/2) as:
*
* 2^0 * (table(0, d>>0 & mask(0)) + ... + table(COMB_BLOCKS-1, d>>0 & mask(COMP_BLOCKS-1)))
* + 2^1 * (table(0, d>>1 & mask(0)) + ... + table(COMB_BLOCKS-1, d>>1 & mask(COMP_BLOCKS-1)))
* + 2^2 * (table(0, d>>2 & mask(0)) + ... + table(COMB_BLOCKS-1, d>>2 & mask(COMP_BLOCKS-1)))
* + ...
* + 2^(COMB_SPACING-1) * (table(0, d>>(COMB_SPACING-1) & mask(0)) + ...)
*
* Or more generically as
*
* sum(2^i * sum(table(b, d>>i & mask(b)), b=0..COMB_BLOCKS-1), i=0..COMB_SPACING-1)
*
* This is implemented using an outer loop that runs in reverse order over the lines of this
* equation, which in each iteration runs an inner loop that adds the terms of that line and
* then doubles the result before proceeding to the next line.
*
* In pseudocode:
* c = infinity
* for comb_off in range(COMB_SPACING - 1, -1, -1):
* for block in range(COMB_BLOCKS):
* c += table(block, (d >> comb_off) & mask(block))
* if comb_off > 0:
* c = 2*c
* return c
*
* This computes c = comb(d, G/2), and thus finally R = c + ctx->ge_offset. Note that it would
* be possible to apply an initial offset instead of a final offset (moving ge_offset to take
* the place of infinity above), but the chosen approach allows using (in a future improvement)
* an incomplete addition formula for most of the multiplication.
*
* The last question is how to implement the table(b, m) function. For any value of b,
* m=(d & mask(b)) can only take on at most 2^COMB_TEETH possible values (the last one may have
* fewer as there mask(b) may exceed the curve order). So we could create COMB_BLOCK tables
* which contain a value for each such m value.
*
* Now note that if m=(d & mask(b)), then flipping the relevant bits of m results in negating
* the result of table(b, m). This is because table(b,m XOR mask(b)) = table(b, mask(b) - m) =
* (mask(b) - m - mask(b)/2)*G = (-m + mask(b)/2)*G = -(m - mask(b)/2)*G = -table(b, m).
* Because of this it suffices to only store the first half of the m values for every b. If an
* entry from the second half is needed, we look up its bit-flipped version instead, and negate
* it.
*
* secp256k1_ecmult_gen_prec_table[b][index] stores the table(b, m) entries. Index
* is the relevant mask(b) bits of m packed together without gaps. */
/* Outer loop: iterate over comb_off from COMB_SPACING - 1 down to 0. */
comb_off = COMB_SPACING - 1;
while (1) {
uint32_t block;
uint32_t bit_pos = comb_off;
/* Inner loop: for each block, add table entries to the result. */
for (block = 0; block < COMB_BLOCKS; ++block) {
/* Gather the mask(block)-selected bits of d into bits. They're packed:
* bits[tooth] = d[(block*COMB_TEETH + tooth)*COMB_SPACING + comb_off]. */
uint32_t bits = 0, sign, abs, index, tooth;
/* Instead of reading individual bits here to construct the bits variable,
* build up the result by xoring rotated reads together. In every iteration,
* one additional bit is made correct, starting at the bottom. The bits
* above that contain junk. This reduces leakage by avoiding computations
* on variables that can have only a low number of possible values (e.g.,
* just two values when reading a single bit into a variable.) See:
* https://www.usenix.org/system/files/conference/usenixsecurity18/sec18-alam.pdf
*/
for (tooth = 0; tooth < COMB_TEETH; ++tooth) {
/* Construct bitdata s.t. the bottom bit is the bit we'd like to read.
*
* We could just set bitdata = recoded[bit_pos >> 5] >> (bit_pos & 0x1f)
* but this would simply discard the bits that fall off at the bottom,
* and thus, for example, bitdata could still have only two values if we
* happen to shift by exactly 31 positions. We use a rotation instead,
* which ensures that bitdata doesn't loose entropy. This relies on the
* rotation being atomic, i.e., the compiler emitting an actual rot
* instruction. */
uint32_t bitdata = secp256k1_rotr32(recoded[bit_pos >> 5], bit_pos & 0x1f);
/* Clear the bit at position tooth, but sssh, don't tell clang. */
uint32_t volatile vmask = ~(1 << tooth);
bits &= vmask;
/* Write the bit into position tooth (and junk into higher bits). */
bits ^= bitdata << tooth;
bit_pos += COMB_SPACING;
}
/* If the top bit of bits is 1, flip them all (corresponding to looking up
* the negated table value), and remember to negate the result in sign. */
sign = (bits >> (COMB_TEETH - 1)) & 1;
abs = (bits ^ -sign) & (COMB_POINTS - 1);
VERIFY_CHECK(sign == 0 || sign == 1);
VERIFY_CHECK(abs < COMB_POINTS);
/** This uses a conditional move to avoid any secret data in array indexes.
* _Any_ use of secret indexes has been demonstrated to result in timing
* sidechannels, even when the cache-line access patterns are uniform.
* See also:
* "A word of warning", CHES 2013 Rump Session, by Daniel J. Bernstein and Peter Schwabe
* (https://cryptojedi.org/peter/data/chesrump-20130822.pdf) and
* "Cache Attacks and Countermeasures: the Case of AES", RSA 2006,
* by Dag Arne Osvik, Adi Shamir, and Eran Tromer
* (https://www.tau.ac.il/~tromer/papers/cache.pdf)
*/
for (index = 0; index < COMB_POINTS; ++index) {
secp256k1_ge_storage_cmov(&adds, &secp256k1_ecmult_gen_prec_table[block][index], index == abs);
}
/* Set add=adds or add=-adds, in constant time, based on sign. */
secp256k1_ge_from_storage(&add, &adds);
secp256k1_fe_negate(&neg, &add.y, 1);
secp256k1_fe_cmov(&add.y, &neg, sign);
/* Add the looked up and conditionally negated value to r. */
if (EXPECT(first, 0)) {
/* If this is the first table lookup, we can skip addition. */
secp256k1_gej_set_ge(r, &add);
/* Give the entry a random Z coordinate to blind intermediary results. */
secp256k1_gej_rescale(r, &ctx->proj_blind);
first = 0;
} else {
secp256k1_gej_add_ge(r, r, &add);
}
}
/* Double the result, except in the last iteration. */
if (comb_off-- == 0) break;
secp256k1_gej_double(r, r);
}
/* Correct for the scalar_offset added at the start (ge_offset = b*G, while b was
* subtracted from the input scalar gn). */
secp256k1_gej_add_ge(r, r, &ctx->ge_offset);
/* Cleanup. */
secp256k1_fe_clear(&neg);
secp256k1_ge_clear(&add);
secp256k1_memclear(&adds, sizeof(adds));
secp256k1_memclear(&recoded, sizeof(recoded));
}
/* Setup blinding values for secp256k1_ecmult_gen. */
static void secp256k1_ecmult_gen_blind(secp256k1_ecmult_gen_context *ctx, const unsigned char *seed32) {
secp256k1_scalar b;
secp256k1_scalar diff;
secp256k1_gej gb;
secp256k1_fe f;
unsigned char nonce32[32];
secp256k1_rfc6979_hmac_sha256 rng;
unsigned char keydata[64];
/* Compute the (2^COMB_BITS - 1)/2 term once. */
secp256k1_ecmult_gen_scalar_diff(&diff);
if (seed32 == NULL) {
/* When seed is NULL, reset the final point and blinding value. */
secp256k1_ge_neg(&ctx->ge_offset, &secp256k1_ge_const_g);
secp256k1_scalar_add(&ctx->scalar_offset, &secp256k1_scalar_one, &diff);
ctx->proj_blind = secp256k1_fe_one;
return;
}
/* The prior blinding value (if not reset) is chained forward by including it in the hash. */
secp256k1_scalar_get_b32(keydata, &ctx->scalar_offset);
/** Using a CSPRNG allows a failure free interface, avoids needing large amounts of random data,
* and guards against weak or adversarial seeds. This is a simpler and safer interface than
* asking the caller for blinding values directly and expecting them to retry on failure.
*/
VERIFY_CHECK(seed32 != NULL);
memcpy(keydata + 32, seed32, 32);
secp256k1_rfc6979_hmac_sha256_initialize(&rng, keydata, 64);
secp256k1_memclear(keydata, sizeof(keydata));
/* Compute projective blinding factor (cannot be 0). */
secp256k1_rfc6979_hmac_sha256_generate(&rng, nonce32, 32);
secp256k1_fe_set_b32_mod(&f, nonce32);
secp256k1_fe_cmov(&f, &secp256k1_fe_one, secp256k1_fe_normalizes_to_zero(&f));
ctx->proj_blind = f;
/* For a random blinding value b, set scalar_offset=diff-b, ge_offset=bG */
secp256k1_rfc6979_hmac_sha256_generate(&rng, nonce32, 32);
secp256k1_scalar_set_b32(&b, nonce32, NULL);
/* The blinding value cannot be zero, as that would mean ge_offset = infinity,
* which secp256k1_gej_add_ge cannot handle. */
secp256k1_scalar_cmov(&b, &secp256k1_scalar_one, secp256k1_scalar_is_zero(&b));
secp256k1_rfc6979_hmac_sha256_finalize(&rng);
secp256k1_ecmult_gen(ctx, &gb, &b);
secp256k1_scalar_negate(&b, &b);
secp256k1_scalar_add(&ctx->scalar_offset, &b, &diff);
secp256k1_ge_set_gej(&ctx->ge_offset, &gb);
/* Clean up. */
secp256k1_memclear(nonce32, sizeof(nonce32));
secp256k1_scalar_clear(&b);
secp256k1_gej_clear(&gb);
secp256k1_fe_clear(&f);
secp256k1_rfc6979_hmac_sha256_clear(&rng);
}
#endif /* SECP256K1_ECMULT_GEN_IMPL_H */

View File

@@ -0,0 +1,853 @@
/******************************************************************************
* Copyright (c) 2013, 2014, 2017 Pieter Wuille, Andrew Poelstra, Jonas Nick *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php. *
******************************************************************************/
#ifndef SECP256K1_ECMULT_IMPL_H
#define SECP256K1_ECMULT_IMPL_H
#include <string.h>
#include <stdint.h>
#include "util.h"
#include "group.h"
#include "scalar.h"
#include "ecmult.h"
#include "precomputed_ecmult.h"
#if defined(EXHAUSTIVE_TEST_ORDER)
/* We need to lower these values for exhaustive tests because
* the tables cannot have infinities in them (this breaks the
* affine-isomorphism stuff which tracks z-ratios) */
# if EXHAUSTIVE_TEST_ORDER > 128
# define WINDOW_A 5
# elif EXHAUSTIVE_TEST_ORDER > 8
# define WINDOW_A 4
# else
# define WINDOW_A 2
# endif
#else
/* optimal for 128-bit and 256-bit exponents. */
# define WINDOW_A 5
/** Larger values for ECMULT_WINDOW_SIZE result in possibly better
* performance at the cost of an exponentially larger precomputed
* table. The exact table size is
* (1 << (WINDOW_G - 2)) * sizeof(secp256k1_ge_storage) bytes,
* where sizeof(secp256k1_ge_storage) is typically 64 bytes but can
* be larger due to platform-specific padding and alignment.
* Two tables of this size are used (due to the endomorphism
* optimization).
*/
#endif
#define WNAF_BITS 128
#define WNAF_SIZE_BITS(bits, w) CEIL_DIV(bits, w)
#define WNAF_SIZE(w) WNAF_SIZE_BITS(WNAF_BITS, w)
/* The number of objects allocated on the scratch space for ecmult_multi algorithms */
#define PIPPENGER_SCRATCH_OBJECTS 6
#define STRAUSS_SCRATCH_OBJECTS 5
#define PIPPENGER_MAX_BUCKET_WINDOW 12
/* Minimum number of points for which pippenger_wnaf is faster than strauss wnaf */
#define ECMULT_PIPPENGER_THRESHOLD 88
#define ECMULT_MAX_POINTS_PER_BATCH 5000000
/** Fill a table 'pre_a' with precomputed odd multiples of a.
* pre_a will contain [1*a,3*a,...,(2*n-1)*a], so it needs space for n group elements.
* zr needs space for n field elements.
*
* Although pre_a is an array of _ge rather than _gej, it actually represents elements
* in Jacobian coordinates with their z coordinates omitted. The omitted z-coordinates
* can be recovered using z and zr. Using the notation z(b) to represent the omitted
* z coordinate of b:
* - z(pre_a[n-1]) = 'z'
* - z(pre_a[i-1]) = z(pre_a[i]) / zr[i] for n > i > 0
*
* Lastly the zr[0] value, which isn't used above, is set so that:
* - a.z = z(pre_a[0]) / zr[0]
*/
static void secp256k1_ecmult_odd_multiples_table(int n, secp256k1_ge *pre_a, secp256k1_fe *zr, secp256k1_fe *z, const secp256k1_gej *a) {
secp256k1_gej d, ai;
secp256k1_ge d_ge;
int i;
VERIFY_CHECK(!a->infinity);
secp256k1_gej_double_var(&d, a, NULL);
/*
* Perform the additions using an isomorphic curve Y^2 = X^3 + 7*C^6 where C := d.z.
* The isomorphism, phi, maps a secp256k1 point (x, y) to the point (x*C^2, y*C^3) on the other curve.
* In Jacobian coordinates phi maps (x, y, z) to (x*C^2, y*C^3, z) or, equivalently to (x, y, z/C).
*
* phi(x, y, z) = (x*C^2, y*C^3, z) = (x, y, z/C)
* d_ge := phi(d) = (d.x, d.y, 1)
* ai := phi(a) = (a.x*C^2, a.y*C^3, a.z)
*
* The group addition functions work correctly on these isomorphic curves.
* In particular phi(d) is easy to represent in affine coordinates under this isomorphism.
* This lets us use the faster secp256k1_gej_add_ge_var group addition function that we wouldn't be able to use otherwise.
*/
secp256k1_ge_set_xy(&d_ge, &d.x, &d.y);
secp256k1_ge_set_gej_zinv(&pre_a[0], a, &d.z);
secp256k1_gej_set_ge(&ai, &pre_a[0]);
ai.z = a->z;
/* pre_a[0] is the point (a.x*C^2, a.y*C^3, a.z*C) which is equivalent to a.
* Set zr[0] to C, which is the ratio between the omitted z(pre_a[0]) value and a.z.
*/
zr[0] = d.z;
for (i = 1; i < n; i++) {
secp256k1_gej_add_ge_var(&ai, &ai, &d_ge, &zr[i]);
secp256k1_ge_set_xy(&pre_a[i], &ai.x, &ai.y);
}
/* Multiply the last z-coordinate by C to undo the isomorphism.
* Since the z-coordinates of the pre_a values are implied by the zr array of z-coordinate ratios,
* undoing the isomorphism here undoes the isomorphism for all pre_a values.
*/
secp256k1_fe_mul(z, &ai.z, &d.z);
}
SECP256K1_INLINE static void secp256k1_ecmult_table_verify(int n, int w) {
(void)n;
(void)w;
VERIFY_CHECK(((n) & 1) == 1);
VERIFY_CHECK((n) >= -((1 << ((w)-1)) - 1));
VERIFY_CHECK((n) <= ((1 << ((w)-1)) - 1));
}
SECP256K1_INLINE static void secp256k1_ecmult_table_get_ge(secp256k1_ge *r, const secp256k1_ge *pre, int n, int w) {
secp256k1_ecmult_table_verify(n,w);
if (n > 0) {
*r = pre[(n-1)/2];
} else {
*r = pre[(-n-1)/2];
secp256k1_fe_negate(&(r->y), &(r->y), 1);
}
}
SECP256K1_INLINE static void secp256k1_ecmult_table_get_ge_lambda(secp256k1_ge *r, const secp256k1_ge *pre, const secp256k1_fe *x, int n, int w) {
secp256k1_ecmult_table_verify(n,w);
if (n > 0) {
secp256k1_ge_set_xy(r, &x[(n-1)/2], &pre[(n-1)/2].y);
} else {
secp256k1_ge_set_xy(r, &x[(-n-1)/2], &pre[(-n-1)/2].y);
secp256k1_fe_negate(&(r->y), &(r->y), 1);
}
}
SECP256K1_INLINE static void secp256k1_ecmult_table_get_ge_storage(secp256k1_ge *r, const secp256k1_ge_storage *pre, int n, int w) {
secp256k1_ecmult_table_verify(n,w);
if (n > 0) {
secp256k1_ge_from_storage(r, &pre[(n-1)/2]);
} else {
secp256k1_ge_from_storage(r, &pre[(-n-1)/2]);
secp256k1_fe_negate(&(r->y), &(r->y), 1);
}
}
/** Convert a number to WNAF notation. The number becomes represented by sum(2^i * wnaf[i], i=0..bits),
* with the following guarantees:
* - each wnaf[i] is either 0, or an odd integer between -(1<<(w-1) - 1) and (1<<(w-1) - 1)
* - two non-zero entries in wnaf are separated by at least w-1 zeroes.
* - the number of set values in wnaf is returned. This number is at most 256, and at most one more
* than the number of bits in the (absolute value) of the input.
*/
static int secp256k1_ecmult_wnaf(int *wnaf, int len, const secp256k1_scalar *a, int w) {
secp256k1_scalar s;
int last_set_bit = -1;
int bit = 0;
int sign = 1;
int carry = 0;
VERIFY_CHECK(wnaf != NULL);
VERIFY_CHECK(0 <= len && len <= 256);
VERIFY_CHECK(a != NULL);
VERIFY_CHECK(2 <= w && w <= 31);
for (bit = 0; bit < len; bit++) {
wnaf[bit] = 0;
}
s = *a;
if (secp256k1_scalar_get_bits_limb32(&s, 255, 1)) {
secp256k1_scalar_negate(&s, &s);
sign = -1;
}
bit = 0;
while (bit < len) {
int now;
int word;
if (secp256k1_scalar_get_bits_limb32(&s, bit, 1) == (unsigned int)carry) {
bit++;
continue;
}
now = w;
if (now > len - bit) {
now = len - bit;
}
word = secp256k1_scalar_get_bits_var(&s, bit, now) + carry;
carry = (word >> (w-1)) & 1;
word -= carry << w;
wnaf[bit] = sign * word;
last_set_bit = bit;
bit += now;
}
#ifdef VERIFY
{
int verify_bit = bit;
VERIFY_CHECK(carry == 0);
while (verify_bit < 256) {
VERIFY_CHECK(secp256k1_scalar_get_bits_limb32(&s, verify_bit, 1) == 0);
verify_bit++;
}
}
#endif
return last_set_bit + 1;
}
struct secp256k1_strauss_point_state {
int wnaf_na_1[129];
int wnaf_na_lam[129];
int bits_na_1;
int bits_na_lam;
};
struct secp256k1_strauss_state {
/* aux is used to hold z-ratios, and then used to hold pre_a[i].x * BETA values. */
secp256k1_fe* aux;
secp256k1_ge* pre_a;
struct secp256k1_strauss_point_state* ps;
};
static void secp256k1_ecmult_strauss_wnaf(const struct secp256k1_strauss_state *state, secp256k1_gej *r, size_t num, const secp256k1_gej *a, const secp256k1_scalar *na, const secp256k1_scalar *ng) {
secp256k1_ge tmpa;
secp256k1_fe Z;
/* Split G factors. */
secp256k1_scalar ng_1, ng_128;
int wnaf_ng_1[129];
int bits_ng_1 = 0;
int wnaf_ng_128[129];
int bits_ng_128 = 0;
int i;
int bits = 0;
size_t np;
size_t no = 0;
secp256k1_fe_set_int(&Z, 1);
for (np = 0; np < num; ++np) {
secp256k1_gej tmp;
secp256k1_scalar na_1, na_lam;
if (secp256k1_scalar_is_zero(&na[np]) || secp256k1_gej_is_infinity(&a[np])) {
continue;
}
/* split na into na_1 and na_lam (where na = na_1 + na_lam*lambda, and na_1 and na_lam are ~128 bit) */
secp256k1_scalar_split_lambda(&na_1, &na_lam, &na[np]);
/* build wnaf representation for na_1 and na_lam. */
state->ps[no].bits_na_1 = secp256k1_ecmult_wnaf(state->ps[no].wnaf_na_1, 129, &na_1, WINDOW_A);
state->ps[no].bits_na_lam = secp256k1_ecmult_wnaf(state->ps[no].wnaf_na_lam, 129, &na_lam, WINDOW_A);
VERIFY_CHECK(state->ps[no].bits_na_1 <= 129);
VERIFY_CHECK(state->ps[no].bits_na_lam <= 129);
if (state->ps[no].bits_na_1 > bits) {
bits = state->ps[no].bits_na_1;
}
if (state->ps[no].bits_na_lam > bits) {
bits = state->ps[no].bits_na_lam;
}
/* Calculate odd multiples of a.
* All multiples are brought to the same Z 'denominator', which is stored
* in Z. Due to secp256k1' isomorphism we can do all operations pretending
* that the Z coordinate was 1, use affine addition formulae, and correct
* the Z coordinate of the result once at the end.
* The exception is the precomputed G table points, which are actually
* affine. Compared to the base used for other points, they have a Z ratio
* of 1/Z, so we can use secp256k1_gej_add_zinv_var, which uses the same
* isomorphism to efficiently add with a known Z inverse.
*/
tmp = a[np];
if (no) {
secp256k1_gej_rescale(&tmp, &Z);
}
secp256k1_ecmult_odd_multiples_table(ECMULT_TABLE_SIZE(WINDOW_A), state->pre_a + no * ECMULT_TABLE_SIZE(WINDOW_A), state->aux + no * ECMULT_TABLE_SIZE(WINDOW_A), &Z, &tmp);
if (no) secp256k1_fe_mul(state->aux + no * ECMULT_TABLE_SIZE(WINDOW_A), state->aux + no * ECMULT_TABLE_SIZE(WINDOW_A), &(a[np].z));
++no;
}
/* Bring them to the same Z denominator. */
if (no) {
secp256k1_ge_table_set_globalz(ECMULT_TABLE_SIZE(WINDOW_A) * no, state->pre_a, state->aux);
}
for (np = 0; np < no; ++np) {
for (i = 0; i < ECMULT_TABLE_SIZE(WINDOW_A); i++) {
secp256k1_fe_mul(&state->aux[np * ECMULT_TABLE_SIZE(WINDOW_A) + i], &state->pre_a[np * ECMULT_TABLE_SIZE(WINDOW_A) + i].x, &secp256k1_const_beta);
}
}
if (ng) {
/* split ng into ng_1 and ng_128 (where gn = gn_1 + gn_128*2^128, and gn_1 and gn_128 are ~128 bit) */
secp256k1_scalar_split_128(&ng_1, &ng_128, ng);
/* Build wnaf representation for ng_1 and ng_128 */
bits_ng_1 = secp256k1_ecmult_wnaf(wnaf_ng_1, 129, &ng_1, WINDOW_G);
bits_ng_128 = secp256k1_ecmult_wnaf(wnaf_ng_128, 129, &ng_128, WINDOW_G);
if (bits_ng_1 > bits) {
bits = bits_ng_1;
}
if (bits_ng_128 > bits) {
bits = bits_ng_128;
}
}
secp256k1_gej_set_infinity(r);
for (i = bits - 1; i >= 0; i--) {
int n;
secp256k1_gej_double_var(r, r, NULL);
for (np = 0; np < no; ++np) {
if (i < state->ps[np].bits_na_1 && (n = state->ps[np].wnaf_na_1[i])) {
secp256k1_ecmult_table_get_ge(&tmpa, state->pre_a + np * ECMULT_TABLE_SIZE(WINDOW_A), n, WINDOW_A);
secp256k1_gej_add_ge_var(r, r, &tmpa, NULL);
}
if (i < state->ps[np].bits_na_lam && (n = state->ps[np].wnaf_na_lam[i])) {
secp256k1_ecmult_table_get_ge_lambda(&tmpa, state->pre_a + np * ECMULT_TABLE_SIZE(WINDOW_A), state->aux + np * ECMULT_TABLE_SIZE(WINDOW_A), n, WINDOW_A);
secp256k1_gej_add_ge_var(r, r, &tmpa, NULL);
}
}
if (i < bits_ng_1 && (n = wnaf_ng_1[i])) {
secp256k1_ecmult_table_get_ge_storage(&tmpa, secp256k1_pre_g, n, WINDOW_G);
secp256k1_gej_add_zinv_var(r, r, &tmpa, &Z);
}
if (i < bits_ng_128 && (n = wnaf_ng_128[i])) {
secp256k1_ecmult_table_get_ge_storage(&tmpa, secp256k1_pre_g_128, n, WINDOW_G);
secp256k1_gej_add_zinv_var(r, r, &tmpa, &Z);
}
}
if (!r->infinity) {
secp256k1_fe_mul(&r->z, &r->z, &Z);
}
}
static void secp256k1_ecmult(secp256k1_gej *r, const secp256k1_gej *a, const secp256k1_scalar *na, const secp256k1_scalar *ng) {
secp256k1_fe aux[ECMULT_TABLE_SIZE(WINDOW_A)];
secp256k1_ge pre_a[ECMULT_TABLE_SIZE(WINDOW_A)];
struct secp256k1_strauss_point_state ps[1];
struct secp256k1_strauss_state state;
state.aux = aux;
state.pre_a = pre_a;
state.ps = ps;
secp256k1_ecmult_strauss_wnaf(&state, r, 1, a, na, ng);
}
static size_t secp256k1_strauss_scratch_size(size_t n_points) {
static const size_t point_size = (sizeof(secp256k1_ge) + sizeof(secp256k1_fe)) * ECMULT_TABLE_SIZE(WINDOW_A) + sizeof(struct secp256k1_strauss_point_state) + sizeof(secp256k1_gej) + sizeof(secp256k1_scalar);
return n_points*point_size;
}
static int secp256k1_ecmult_strauss_batch(const secp256k1_callback* error_callback, secp256k1_scratch *scratch, secp256k1_gej *r, const secp256k1_scalar *inp_g_sc, secp256k1_ecmult_multi_callback cb, void *cbdata, size_t n_points, size_t cb_offset) {
secp256k1_gej* points;
secp256k1_scalar* scalars;
struct secp256k1_strauss_state state;
size_t i;
const size_t scratch_checkpoint = secp256k1_scratch_checkpoint(error_callback, scratch);
secp256k1_gej_set_infinity(r);
if (inp_g_sc == NULL && n_points == 0) {
return 1;
}
/* We allocate STRAUSS_SCRATCH_OBJECTS objects on the scratch space. If these
* allocations change, make sure to update the STRAUSS_SCRATCH_OBJECTS
* constant and strauss_scratch_size accordingly. */
points = (secp256k1_gej*)secp256k1_scratch_alloc(error_callback, scratch, n_points * sizeof(secp256k1_gej));
scalars = (secp256k1_scalar*)secp256k1_scratch_alloc(error_callback, scratch, n_points * sizeof(secp256k1_scalar));
state.aux = (secp256k1_fe*)secp256k1_scratch_alloc(error_callback, scratch, n_points * ECMULT_TABLE_SIZE(WINDOW_A) * sizeof(secp256k1_fe));
state.pre_a = (secp256k1_ge*)secp256k1_scratch_alloc(error_callback, scratch, n_points * ECMULT_TABLE_SIZE(WINDOW_A) * sizeof(secp256k1_ge));
state.ps = (struct secp256k1_strauss_point_state*)secp256k1_scratch_alloc(error_callback, scratch, n_points * sizeof(struct secp256k1_strauss_point_state));
if (points == NULL || scalars == NULL || state.aux == NULL || state.pre_a == NULL || state.ps == NULL) {
secp256k1_scratch_apply_checkpoint(error_callback, scratch, scratch_checkpoint);
return 0;
}
for (i = 0; i < n_points; i++) {
secp256k1_ge point;
if (!cb(&scalars[i], &point, i+cb_offset, cbdata)) {
secp256k1_scratch_apply_checkpoint(error_callback, scratch, scratch_checkpoint);
return 0;
}
secp256k1_gej_set_ge(&points[i], &point);
}
secp256k1_ecmult_strauss_wnaf(&state, r, n_points, points, scalars, inp_g_sc);
secp256k1_scratch_apply_checkpoint(error_callback, scratch, scratch_checkpoint);
return 1;
}
/* Wrapper for secp256k1_ecmult_multi_func interface */
static int secp256k1_ecmult_strauss_batch_single(const secp256k1_callback* error_callback, secp256k1_scratch *scratch, secp256k1_gej *r, const secp256k1_scalar *inp_g_sc, secp256k1_ecmult_multi_callback cb, void *cbdata, size_t n) {
return secp256k1_ecmult_strauss_batch(error_callback, scratch, r, inp_g_sc, cb, cbdata, n, 0);
}
static size_t secp256k1_strauss_max_points(const secp256k1_callback* error_callback, secp256k1_scratch *scratch) {
return secp256k1_scratch_max_allocation(error_callback, scratch, STRAUSS_SCRATCH_OBJECTS) / secp256k1_strauss_scratch_size(1);
}
/** Convert a number to WNAF notation.
* The number becomes represented by sum(2^{wi} * wnaf[i], i=0..WNAF_SIZE(w)+1) - return_val.
* It has the following guarantees:
* - each wnaf[i] is either 0 or an odd integer between -(1 << w) and (1 << w)
* - the number of words set is always WNAF_SIZE(w)
* - the returned skew is 0 or 1
*/
static int secp256k1_wnaf_fixed(int *wnaf, const secp256k1_scalar *s, int w) {
int skew = 0;
int pos;
int max_pos;
int last_w;
const secp256k1_scalar *work = s;
if (secp256k1_scalar_is_zero(s)) {
for (pos = 0; pos < WNAF_SIZE(w); pos++) {
wnaf[pos] = 0;
}
return 0;
}
if (secp256k1_scalar_is_even(s)) {
skew = 1;
}
wnaf[0] = secp256k1_scalar_get_bits_var(work, 0, w) + skew;
/* Compute last window size. Relevant when window size doesn't divide the
* number of bits in the scalar */
last_w = WNAF_BITS - (WNAF_SIZE(w) - 1) * w;
/* Store the position of the first nonzero word in max_pos to allow
* skipping leading zeros when calculating the wnaf. */
for (pos = WNAF_SIZE(w) - 1; pos > 0; pos--) {
int val = secp256k1_scalar_get_bits_var(work, pos * w, pos == WNAF_SIZE(w)-1 ? last_w : w);
if(val != 0) {
break;
}
wnaf[pos] = 0;
}
max_pos = pos;
pos = 1;
while (pos <= max_pos) {
int val = secp256k1_scalar_get_bits_var(work, pos * w, pos == WNAF_SIZE(w)-1 ? last_w : w);
if ((val & 1) == 0) {
wnaf[pos - 1] -= (1 << w);
wnaf[pos] = (val + 1);
} else {
wnaf[pos] = val;
}
/* Set a coefficient to zero if it is 1 or -1 and the proceeding digit
* is strictly negative or strictly positive respectively. Only change
* coefficients at previous positions because above code assumes that
* wnaf[pos - 1] is odd.
*/
if (pos >= 2 && ((wnaf[pos - 1] == 1 && wnaf[pos - 2] < 0) || (wnaf[pos - 1] == -1 && wnaf[pos - 2] > 0))) {
if (wnaf[pos - 1] == 1) {
wnaf[pos - 2] += 1 << w;
} else {
wnaf[pos - 2] -= 1 << w;
}
wnaf[pos - 1] = 0;
}
++pos;
}
return skew;
}
struct secp256k1_pippenger_point_state {
int skew_na;
size_t input_pos;
};
struct secp256k1_pippenger_state {
int *wnaf_na;
struct secp256k1_pippenger_point_state* ps;
};
/*
* pippenger_wnaf computes the result of a multi-point multiplication as
* follows: The scalars are brought into wnaf with n_wnaf elements each. Then
* for every i < n_wnaf, first each point is added to a "bucket" corresponding
* to the point's wnaf[i]. Second, the buckets are added together such that
* r += 1*bucket[0] + 3*bucket[1] + 5*bucket[2] + ...
*/
static int secp256k1_ecmult_pippenger_wnaf(secp256k1_gej *buckets, int bucket_window, struct secp256k1_pippenger_state *state, secp256k1_gej *r, const secp256k1_scalar *sc, const secp256k1_ge *pt, size_t num) {
size_t n_wnaf = WNAF_SIZE(bucket_window+1);
size_t np;
size_t no = 0;
int i;
int j;
for (np = 0; np < num; ++np) {
if (secp256k1_scalar_is_zero(&sc[np]) || secp256k1_ge_is_infinity(&pt[np])) {
continue;
}
state->ps[no].input_pos = np;
state->ps[no].skew_na = secp256k1_wnaf_fixed(&state->wnaf_na[no*n_wnaf], &sc[np], bucket_window+1);
no++;
}
secp256k1_gej_set_infinity(r);
if (no == 0) {
return 1;
}
for (i = n_wnaf - 1; i >= 0; i--) {
secp256k1_gej running_sum;
for(j = 0; j < ECMULT_TABLE_SIZE(bucket_window+2); j++) {
secp256k1_gej_set_infinity(&buckets[j]);
}
for (np = 0; np < no; ++np) {
int n = state->wnaf_na[np*n_wnaf + i];
struct secp256k1_pippenger_point_state point_state = state->ps[np];
secp256k1_ge tmp;
int idx;
if (i == 0) {
/* correct for wnaf skew */
int skew = point_state.skew_na;
if (skew) {
secp256k1_ge_neg(&tmp, &pt[point_state.input_pos]);
secp256k1_gej_add_ge_var(&buckets[0], &buckets[0], &tmp, NULL);
}
}
if (n > 0) {
idx = (n - 1)/2;
secp256k1_gej_add_ge_var(&buckets[idx], &buckets[idx], &pt[point_state.input_pos], NULL);
} else if (n < 0) {
idx = -(n + 1)/2;
secp256k1_ge_neg(&tmp, &pt[point_state.input_pos]);
secp256k1_gej_add_ge_var(&buckets[idx], &buckets[idx], &tmp, NULL);
}
}
for(j = 0; j < bucket_window; j++) {
secp256k1_gej_double_var(r, r, NULL);
}
secp256k1_gej_set_infinity(&running_sum);
/* Accumulate the sum: bucket[0] + 3*bucket[1] + 5*bucket[2] + 7*bucket[3] + ...
* = bucket[0] + bucket[1] + bucket[2] + bucket[3] + ...
* + 2 * (bucket[1] + 2*bucket[2] + 3*bucket[3] + ...)
* using an intermediate running sum:
* running_sum = bucket[0] + bucket[1] + bucket[2] + ...
*
* The doubling is done implicitly by deferring the final window doubling (of 'r').
*/
for(j = ECMULT_TABLE_SIZE(bucket_window+2) - 1; j > 0; j--) {
secp256k1_gej_add_var(&running_sum, &running_sum, &buckets[j], NULL);
secp256k1_gej_add_var(r, r, &running_sum, NULL);
}
secp256k1_gej_add_var(&running_sum, &running_sum, &buckets[0], NULL);
secp256k1_gej_double_var(r, r, NULL);
secp256k1_gej_add_var(r, r, &running_sum, NULL);
}
return 1;
}
/**
* Returns optimal bucket_window (number of bits of a scalar represented by a
* set of buckets) for a given number of points.
*/
static int secp256k1_pippenger_bucket_window(size_t n) {
if (n <= 1) {
return 1;
} else if (n <= 4) {
return 2;
} else if (n <= 20) {
return 3;
} else if (n <= 57) {
return 4;
} else if (n <= 136) {
return 5;
} else if (n <= 235) {
return 6;
} else if (n <= 1260) {
return 7;
} else if (n <= 4420) {
return 9;
} else if (n <= 7880) {
return 10;
} else if (n <= 16050) {
return 11;
} else {
return PIPPENGER_MAX_BUCKET_WINDOW;
}
}
/**
* Returns the maximum optimal number of points for a bucket_window.
*/
static size_t secp256k1_pippenger_bucket_window_inv(int bucket_window) {
switch(bucket_window) {
case 1: return 1;
case 2: return 4;
case 3: return 20;
case 4: return 57;
case 5: return 136;
case 6: return 235;
case 7: return 1260;
case 8: return 1260;
case 9: return 4420;
case 10: return 7880;
case 11: return 16050;
case PIPPENGER_MAX_BUCKET_WINDOW: return SIZE_MAX;
}
return 0;
}
SECP256K1_INLINE static void secp256k1_ecmult_endo_split(secp256k1_scalar *s1, secp256k1_scalar *s2, secp256k1_ge *p1, secp256k1_ge *p2) {
secp256k1_scalar tmp = *s1;
secp256k1_scalar_split_lambda(s1, s2, &tmp);
secp256k1_ge_mul_lambda(p2, p1);
if (secp256k1_scalar_is_high(s1)) {
secp256k1_scalar_negate(s1, s1);
secp256k1_ge_neg(p1, p1);
}
if (secp256k1_scalar_is_high(s2)) {
secp256k1_scalar_negate(s2, s2);
secp256k1_ge_neg(p2, p2);
}
}
/**
* Returns the scratch size required for a given number of points (excluding
* base point G) without considering alignment.
*/
static size_t secp256k1_pippenger_scratch_size(size_t n_points, int bucket_window) {
size_t entries = 2*n_points + 2;
size_t entry_size = sizeof(secp256k1_ge) + sizeof(secp256k1_scalar) + sizeof(struct secp256k1_pippenger_point_state) + (WNAF_SIZE(bucket_window+1)+1)*sizeof(int);
return (sizeof(secp256k1_gej) << bucket_window) + sizeof(struct secp256k1_pippenger_state) + entries * entry_size;
}
static int secp256k1_ecmult_pippenger_batch(const secp256k1_callback* error_callback, secp256k1_scratch *scratch, secp256k1_gej *r, const secp256k1_scalar *inp_g_sc, secp256k1_ecmult_multi_callback cb, void *cbdata, size_t n_points, size_t cb_offset) {
const size_t scratch_checkpoint = secp256k1_scratch_checkpoint(error_callback, scratch);
/* Use 2(n+1) with the endomorphism, when calculating batch
* sizes. The reason for +1 is that we add the G scalar to the list of
* other scalars. */
size_t entries = 2*n_points + 2;
secp256k1_ge *points;
secp256k1_scalar *scalars;
secp256k1_gej *buckets;
struct secp256k1_pippenger_state *state_space;
size_t idx = 0;
size_t point_idx = 0;
int bucket_window;
secp256k1_gej_set_infinity(r);
if (inp_g_sc == NULL && n_points == 0) {
return 1;
}
bucket_window = secp256k1_pippenger_bucket_window(n_points);
/* We allocate PIPPENGER_SCRATCH_OBJECTS objects on the scratch space. If
* these allocations change, make sure to update the
* PIPPENGER_SCRATCH_OBJECTS constant and pippenger_scratch_size
* accordingly. */
points = (secp256k1_ge *) secp256k1_scratch_alloc(error_callback, scratch, entries * sizeof(*points));
scalars = (secp256k1_scalar *) secp256k1_scratch_alloc(error_callback, scratch, entries * sizeof(*scalars));
state_space = (struct secp256k1_pippenger_state *) secp256k1_scratch_alloc(error_callback, scratch, sizeof(*state_space));
if (points == NULL || scalars == NULL || state_space == NULL) {
secp256k1_scratch_apply_checkpoint(error_callback, scratch, scratch_checkpoint);
return 0;
}
state_space->ps = (struct secp256k1_pippenger_point_state *) secp256k1_scratch_alloc(error_callback, scratch, entries * sizeof(*state_space->ps));
state_space->wnaf_na = (int *) secp256k1_scratch_alloc(error_callback, scratch, entries*(WNAF_SIZE(bucket_window+1)) * sizeof(int));
buckets = (secp256k1_gej *) secp256k1_scratch_alloc(error_callback, scratch, ((size_t)1 << bucket_window) * sizeof(*buckets));
if (state_space->ps == NULL || state_space->wnaf_na == NULL || buckets == NULL) {
secp256k1_scratch_apply_checkpoint(error_callback, scratch, scratch_checkpoint);
return 0;
}
if (inp_g_sc != NULL) {
scalars[0] = *inp_g_sc;
points[0] = secp256k1_ge_const_g;
idx++;
secp256k1_ecmult_endo_split(&scalars[0], &scalars[1], &points[0], &points[1]);
idx++;
}
while (point_idx < n_points) {
if (!cb(&scalars[idx], &points[idx], point_idx + cb_offset, cbdata)) {
secp256k1_scratch_apply_checkpoint(error_callback, scratch, scratch_checkpoint);
return 0;
}
idx++;
secp256k1_ecmult_endo_split(&scalars[idx - 1], &scalars[idx], &points[idx - 1], &points[idx]);
idx++;
point_idx++;
}
secp256k1_ecmult_pippenger_wnaf(buckets, bucket_window, state_space, r, scalars, points, idx);
secp256k1_scratch_apply_checkpoint(error_callback, scratch, scratch_checkpoint);
return 1;
}
/* Wrapper for secp256k1_ecmult_multi_func interface */
static int secp256k1_ecmult_pippenger_batch_single(const secp256k1_callback* error_callback, secp256k1_scratch *scratch, secp256k1_gej *r, const secp256k1_scalar *inp_g_sc, secp256k1_ecmult_multi_callback cb, void *cbdata, size_t n) {
return secp256k1_ecmult_pippenger_batch(error_callback, scratch, r, inp_g_sc, cb, cbdata, n, 0);
}
/**
* Returns the maximum number of points in addition to G that can be used with
* a given scratch space. The function ensures that fewer points may also be
* used.
*/
static size_t secp256k1_pippenger_max_points(const secp256k1_callback* error_callback, secp256k1_scratch *scratch) {
size_t max_alloc = secp256k1_scratch_max_allocation(error_callback, scratch, PIPPENGER_SCRATCH_OBJECTS);
int bucket_window;
size_t res = 0;
for (bucket_window = 1; bucket_window <= PIPPENGER_MAX_BUCKET_WINDOW; bucket_window++) {
size_t n_points;
size_t max_points = secp256k1_pippenger_bucket_window_inv(bucket_window);
size_t space_for_points;
size_t space_overhead;
size_t entry_size = sizeof(secp256k1_ge) + sizeof(secp256k1_scalar) + sizeof(struct secp256k1_pippenger_point_state) + (WNAF_SIZE(bucket_window+1)+1)*sizeof(int);
entry_size = 2*entry_size;
space_overhead = (sizeof(secp256k1_gej) << bucket_window) + entry_size + sizeof(struct secp256k1_pippenger_state);
if (space_overhead > max_alloc) {
break;
}
space_for_points = max_alloc - space_overhead;
n_points = space_for_points/entry_size;
n_points = n_points > max_points ? max_points : n_points;
if (n_points > res) {
res = n_points;
}
if (n_points < max_points) {
/* A larger bucket_window may support even more points. But if we
* would choose that then the caller couldn't safely use any number
* smaller than what this function returns */
break;
}
}
return res;
}
/* Computes ecmult_multi by simply multiplying and adding each point. Does not
* require a scratch space */
static int secp256k1_ecmult_multi_simple_var(secp256k1_gej *r, const secp256k1_scalar *inp_g_sc, secp256k1_ecmult_multi_callback cb, void *cbdata, size_t n_points) {
size_t point_idx;
secp256k1_gej tmpj;
secp256k1_gej_set_infinity(r);
secp256k1_gej_set_infinity(&tmpj);
/* r = inp_g_sc*G */
secp256k1_ecmult(r, &tmpj, &secp256k1_scalar_zero, inp_g_sc);
for (point_idx = 0; point_idx < n_points; point_idx++) {
secp256k1_ge point;
secp256k1_gej pointj;
secp256k1_scalar scalar;
if (!cb(&scalar, &point, point_idx, cbdata)) {
return 0;
}
/* r += scalar*point */
secp256k1_gej_set_ge(&pointj, &point);
secp256k1_ecmult(&tmpj, &pointj, &scalar, NULL);
secp256k1_gej_add_var(r, r, &tmpj, NULL);
}
return 1;
}
/* Compute the number of batches and the batch size given the maximum batch size and the
* total number of points */
static int secp256k1_ecmult_multi_batch_size_helper(size_t *n_batches, size_t *n_batch_points, size_t max_n_batch_points, size_t n) {
if (max_n_batch_points == 0) {
return 0;
}
if (max_n_batch_points > ECMULT_MAX_POINTS_PER_BATCH) {
max_n_batch_points = ECMULT_MAX_POINTS_PER_BATCH;
}
if (n == 0) {
*n_batches = 0;
*n_batch_points = 0;
return 1;
}
/* Compute ceil(n/max_n_batch_points) and ceil(n/n_batches) */
*n_batches = CEIL_DIV(n, max_n_batch_points);
*n_batch_points = CEIL_DIV(n, *n_batches);
return 1;
}
typedef int (*secp256k1_ecmult_multi_func)(const secp256k1_callback* error_callback, secp256k1_scratch*, secp256k1_gej*, const secp256k1_scalar*, secp256k1_ecmult_multi_callback cb, void*, size_t);
static int secp256k1_ecmult_multi_var(const secp256k1_callback* error_callback, secp256k1_scratch *scratch, secp256k1_gej *r, const secp256k1_scalar *inp_g_sc, secp256k1_ecmult_multi_callback cb, void *cbdata, size_t n) {
size_t i;
int (*f)(const secp256k1_callback* error_callback, secp256k1_scratch*, secp256k1_gej*, const secp256k1_scalar*, secp256k1_ecmult_multi_callback cb, void*, size_t, size_t);
size_t n_batches;
size_t n_batch_points;
secp256k1_gej_set_infinity(r);
if (inp_g_sc == NULL && n == 0) {
return 1;
} else if (n == 0) {
secp256k1_ecmult(r, r, &secp256k1_scalar_zero, inp_g_sc);
return 1;
}
if (scratch == NULL) {
return secp256k1_ecmult_multi_simple_var(r, inp_g_sc, cb, cbdata, n);
}
/* Compute the batch sizes for Pippenger's algorithm given a scratch space. If it's greater than
* a threshold use Pippenger's algorithm. Otherwise use Strauss' algorithm.
* As a first step check if there's enough space for Pippenger's algo (which requires less space
* than Strauss' algo) and if not, use the simple algorithm. */
if (!secp256k1_ecmult_multi_batch_size_helper(&n_batches, &n_batch_points, secp256k1_pippenger_max_points(error_callback, scratch), n)) {
return secp256k1_ecmult_multi_simple_var(r, inp_g_sc, cb, cbdata, n);
}
if (n_batch_points >= ECMULT_PIPPENGER_THRESHOLD) {
f = secp256k1_ecmult_pippenger_batch;
} else {
if (!secp256k1_ecmult_multi_batch_size_helper(&n_batches, &n_batch_points, secp256k1_strauss_max_points(error_callback, scratch), n)) {
return secp256k1_ecmult_multi_simple_var(r, inp_g_sc, cb, cbdata, n);
}
f = secp256k1_ecmult_strauss_batch;
}
for(i = 0; i < n_batches; i++) {
size_t nbp = n < n_batch_points ? n : n_batch_points;
size_t offset = n_batch_points*i;
secp256k1_gej tmp;
if (!f(error_callback, scratch, &tmp, i == 0 ? inp_g_sc : NULL, cb, cbdata, nbp, offset)) {
return 0;
}
secp256k1_gej_add_var(r, r, &tmp, NULL);
n -= nbp;
}
return 1;
}
#endif /* SECP256K1_ECMULT_IMPL_H */

350
libsecp256k1/src/field.h Normal file
View File

@@ -0,0 +1,350 @@
/***********************************************************************
* Copyright (c) 2013, 2014 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_FIELD_H
#define SECP256K1_FIELD_H
#include "util.h"
/* This file defines the generic interface for working with secp256k1_fe
* objects, which represent field elements (integers modulo 2^256 - 2^32 - 977).
*
* The actual definition of the secp256k1_fe type depends on the chosen field
* implementation; see the field_5x52.h and field_10x26.h files for details.
*
* All secp256k1_fe objects have implicit properties that determine what
* operations are permitted on it. These are purely a function of what
* secp256k1_fe_ operations are applied on it, generally (implicitly) fixed at
* compile time, and do not depend on the chosen field implementation. Despite
* that, what these properties actually entail for the field representation
* values depends on the chosen field implementation. These properties are:
* - magnitude: an integer in [0,32]
* - normalized: 0 or 1; normalized=1 implies magnitude <= 1.
*
* In VERIFY mode, they are materialized explicitly as fields in the struct,
* allowing run-time verification of these properties. In that case, the field
* implementation also provides a secp256k1_fe_verify routine to verify that
* these fields match the run-time value and perform internal consistency
* checks. */
#ifdef VERIFY
# define SECP256K1_FE_VERIFY_FIELDS \
int magnitude; \
int normalized;
#else
# define SECP256K1_FE_VERIFY_FIELDS
#endif
#if defined(SECP256K1_WIDEMUL_INT128)
#include "field_5x52.h"
#elif defined(SECP256K1_WIDEMUL_INT64)
#include "field_10x26.h"
#else
#error "Please select wide multiplication implementation"
#endif
#ifdef VERIFY
/* Magnitude and normalized value for constants. */
#define SECP256K1_FE_VERIFY_CONST(d7, d6, d5, d4, d3, d2, d1, d0) \
/* Magnitude is 0 for constant 0; 1 otherwise. */ \
, (((d7) | (d6) | (d5) | (d4) | (d3) | (d2) | (d1) | (d0)) != 0) \
/* Normalized is 1 unless sum(d_i<<(32*i) for i=0..7) exceeds field modulus. */ \
, (!(((d7) & (d6) & (d5) & (d4) & (d3) & (d2)) == 0xfffffffful && ((d1) == 0xfffffffful || ((d1) == 0xfffffffe && (d0 >= 0xfffffc2f)))))
#else
#define SECP256K1_FE_VERIFY_CONST(d7, d6, d5, d4, d3, d2, d1, d0)
#endif
/** This expands to an initializer for a secp256k1_fe valued sum((i*32) * d_i, i=0..7) mod p.
*
* It has magnitude 1, unless d_i are all 0, in which case the magnitude is 0.
* It is normalized, unless sum(2^(i*32) * d_i, i=0..7) >= p.
*
* SECP256K1_FE_CONST_INNER is provided by the implementation.
*/
#define SECP256K1_FE_CONST(d7, d6, d5, d4, d3, d2, d1, d0) {SECP256K1_FE_CONST_INNER((d7), (d6), (d5), (d4), (d3), (d2), (d1), (d0)) SECP256K1_FE_VERIFY_CONST((d7), (d6), (d5), (d4), (d3), (d2), (d1), (d0)) }
static const secp256k1_fe secp256k1_fe_one = SECP256K1_FE_CONST(0, 0, 0, 0, 0, 0, 0, 1);
static const secp256k1_fe secp256k1_const_beta = SECP256K1_FE_CONST(
0x7ae96a2bul, 0x657c0710ul, 0x6e64479eul, 0xac3434e9ul,
0x9cf04975ul, 0x12f58995ul, 0xc1396c28ul, 0x719501eeul
);
#ifndef VERIFY
/* In non-VERIFY mode, we #define the fe operations to be identical to their
* internal field implementation, to avoid the potential overhead of a
* function call (even though presumably inlinable). */
# define secp256k1_fe_normalize secp256k1_fe_impl_normalize
# define secp256k1_fe_normalize_weak secp256k1_fe_impl_normalize_weak
# define secp256k1_fe_normalize_var secp256k1_fe_impl_normalize_var
# define secp256k1_fe_normalizes_to_zero secp256k1_fe_impl_normalizes_to_zero
# define secp256k1_fe_normalizes_to_zero_var secp256k1_fe_impl_normalizes_to_zero_var
# define secp256k1_fe_set_int secp256k1_fe_impl_set_int
# define secp256k1_fe_is_zero secp256k1_fe_impl_is_zero
# define secp256k1_fe_is_odd secp256k1_fe_impl_is_odd
# define secp256k1_fe_cmp_var secp256k1_fe_impl_cmp_var
# define secp256k1_fe_set_b32_mod secp256k1_fe_impl_set_b32_mod
# define secp256k1_fe_set_b32_limit secp256k1_fe_impl_set_b32_limit
# define secp256k1_fe_get_b32 secp256k1_fe_impl_get_b32
# define secp256k1_fe_negate_unchecked secp256k1_fe_impl_negate_unchecked
# define secp256k1_fe_mul_int_unchecked secp256k1_fe_impl_mul_int_unchecked
# define secp256k1_fe_add secp256k1_fe_impl_add
# define secp256k1_fe_mul secp256k1_fe_impl_mul
# define secp256k1_fe_sqr secp256k1_fe_impl_sqr
# define secp256k1_fe_cmov secp256k1_fe_impl_cmov
# define secp256k1_fe_to_storage secp256k1_fe_impl_to_storage
# define secp256k1_fe_from_storage secp256k1_fe_impl_from_storage
# define secp256k1_fe_inv secp256k1_fe_impl_inv
# define secp256k1_fe_inv_var secp256k1_fe_impl_inv_var
# define secp256k1_fe_get_bounds secp256k1_fe_impl_get_bounds
# define secp256k1_fe_half secp256k1_fe_impl_half
# define secp256k1_fe_add_int secp256k1_fe_impl_add_int
# define secp256k1_fe_is_square_var secp256k1_fe_impl_is_square_var
#endif /* !defined(VERIFY) */
/** Normalize a field element.
*
* On input, r must be a valid field element.
* On output, r represents the same value but has normalized=1 and magnitude=1.
*/
static void secp256k1_fe_normalize(secp256k1_fe *r);
/** Give a field element magnitude 1.
*
* On input, r must be a valid field element.
* On output, r represents the same value but has magnitude=1. Normalized is unchanged.
*/
static void secp256k1_fe_normalize_weak(secp256k1_fe *r);
/** Normalize a field element, without constant-time guarantee.
*
* Identical in behavior to secp256k1_fe_normalize, but not constant time in r.
*/
static void secp256k1_fe_normalize_var(secp256k1_fe *r);
/** Determine whether r represents field element 0.
*
* On input, r must be a valid field element.
* Returns whether r = 0 (mod p).
*/
static int secp256k1_fe_normalizes_to_zero(const secp256k1_fe *r);
/** Determine whether r represents field element 0, without constant-time guarantee.
*
* Identical in behavior to secp256k1_normalizes_to_zero, but not constant time in r.
*/
static int secp256k1_fe_normalizes_to_zero_var(const secp256k1_fe *r);
/** Set a field element to an integer in range [0,0x7FFF].
*
* On input, r does not need to be initialized, a must be in [0,0x7FFF].
* On output, r represents value a, is normalized and has magnitude (a!=0).
*/
static void secp256k1_fe_set_int(secp256k1_fe *r, int a);
/** Clear a field element to prevent leaking sensitive information. */
static void secp256k1_fe_clear(secp256k1_fe *a);
/** Determine whether a represents field element 0.
*
* On input, a must be a valid normalized field element.
* Returns whether a = 0 (mod p).
*
* This behaves identical to secp256k1_normalizes_to_zero{,_var}, but requires
* normalized input (and is much faster).
*/
static int secp256k1_fe_is_zero(const secp256k1_fe *a);
/** Determine whether a (mod p) is odd.
*
* On input, a must be a valid normalized field element.
* Returns (int(a) mod p) & 1.
*/
static int secp256k1_fe_is_odd(const secp256k1_fe *a);
/** Determine whether two field elements are equal.
*
* On input, a and b must be valid field elements with magnitudes not exceeding
* 1 and 31, respectively.
* Returns a = b (mod p).
*/
static int secp256k1_fe_equal(const secp256k1_fe *a, const secp256k1_fe *b);
/** Compare the values represented by 2 field elements, without constant-time guarantee.
*
* On input, a and b must be valid normalized field elements.
* Returns 1 if a > b, -1 if a < b, and 0 if a = b (comparisons are done as integers
* in range 0..p-1).
*/
static int secp256k1_fe_cmp_var(const secp256k1_fe *a, const secp256k1_fe *b);
/** Set a field element equal to the element represented by a provided 32-byte big endian value
* interpreted modulo p.
*
* On input, r does not need to be initialized. a must be a pointer to an initialized 32-byte array.
* On output, r = a (mod p). It will have magnitude 1, and not be normalized.
*/
static void secp256k1_fe_set_b32_mod(secp256k1_fe *r, const unsigned char *a);
/** Set a field element equal to a provided 32-byte big endian value, checking for overflow.
*
* On input, r does not need to be initialized. a must be a pointer to an initialized 32-byte array.
* On output, r = a if (a < p), it will be normalized with magnitude 1, and 1 is returned.
* If a >= p, 0 is returned, and r will be made invalid (and must not be used without overwriting).
*/
static int secp256k1_fe_set_b32_limit(secp256k1_fe *r, const unsigned char *a);
/** Convert a field element to 32-byte big endian byte array.
* On input, a must be a valid normalized field element, and r a pointer to a 32-byte array.
* On output, r = a (mod p).
*/
static void secp256k1_fe_get_b32(unsigned char *r, const secp256k1_fe *a);
/** Negate a field element.
*
* On input, r does not need to be initialized. a must be a valid field element with
* magnitude not exceeding m. m must be an integer constant expression in [0,31].
* Performs {r = -a}.
* On output, r will not be normalized, and will have magnitude m+1.
*/
#define secp256k1_fe_negate(r, a, m) ASSERT_INT_CONST_AND_DO(m, secp256k1_fe_negate_unchecked(r, a, m))
/** Like secp256k1_fe_negate_unchecked but m is not checked to be an integer constant expression.
*
* Should not be called directly outside of tests.
*/
static void secp256k1_fe_negate_unchecked(secp256k1_fe *r, const secp256k1_fe *a, int m);
/** Add a small integer to a field element.
*
* Performs {r += a}. The magnitude of r increases by 1, and normalized is cleared.
* a must be in range [0,0x7FFF].
*/
static void secp256k1_fe_add_int(secp256k1_fe *r, int a);
/** Multiply a field element with a small integer.
*
* On input, r must be a valid field element. a must be an integer constant expression in [0,32].
* The magnitude of r times a must not exceed 32.
* Performs {r *= a}.
* On output, r's magnitude is multiplied by a, and r will not be normalized.
*/
#define secp256k1_fe_mul_int(r, a) ASSERT_INT_CONST_AND_DO(a, secp256k1_fe_mul_int_unchecked(r, a))
/** Like secp256k1_fe_mul_int but a is not checked to be an integer constant expression.
*
* Should not be called directly outside of tests.
*/
static void secp256k1_fe_mul_int_unchecked(secp256k1_fe *r, int a);
/** Increment a field element by another.
*
* On input, r and a must be valid field elements, not necessarily normalized.
* The sum of their magnitudes must not exceed 32.
* Performs {r += a}.
* On output, r will not be normalized, and will have magnitude incremented by a's.
*/
static void secp256k1_fe_add(secp256k1_fe *r, const secp256k1_fe *a);
/** Multiply two field elements.
*
* On input, a and b must be valid field elements; r does not need to be initialized.
* r and a may point to the same object, but neither may point to the object pointed
* to by b. The magnitudes of a and b must not exceed 8.
* Performs {r = a * b}
* On output, r will have magnitude 1, but won't be normalized.
*/
static void secp256k1_fe_mul(secp256k1_fe *r, const secp256k1_fe *a, const secp256k1_fe * SECP256K1_RESTRICT b);
/** Square a field element.
*
* On input, a must be a valid field element; r does not need to be initialized. The magnitude
* of a must not exceed 8.
* Performs {r = a**2}
* On output, r will have magnitude 1, but won't be normalized.
*/
static void secp256k1_fe_sqr(secp256k1_fe *r, const secp256k1_fe *a);
/** Compute a square root of a field element.
*
* On input, a must be a valid field element with magnitude<=8; r need not be initialized.
* If sqrt(a) exists, performs {r = sqrt(a)} and returns 1.
* Otherwise, sqrt(-a) exists. The function performs {r = sqrt(-a)} and returns 0.
* The resulting value represented by r will be a square itself.
* Variables r and a must not point to the same object.
* On output, r will have magnitude 1 but will not be normalized.
*/
static int secp256k1_fe_sqrt(secp256k1_fe * SECP256K1_RESTRICT r, const secp256k1_fe * SECP256K1_RESTRICT a);
/** Compute the modular inverse of a field element.
*
* On input, a must be a valid field element; r need not be initialized.
* Performs {r = a**(p-2)} (which maps 0 to 0, and every other element to its
* inverse).
* On output, r will have magnitude (a.magnitude != 0) and be normalized.
*/
static void secp256k1_fe_inv(secp256k1_fe *r, const secp256k1_fe *a);
/** Compute the modular inverse of a field element, without constant-time guarantee.
*
* Behaves identically to secp256k1_fe_inv, but is not constant-time in a.
*/
static void secp256k1_fe_inv_var(secp256k1_fe *r, const secp256k1_fe *a);
/** Convert a field element to secp256k1_fe_storage.
*
* On input, a must be a valid normalized field element.
* Performs {r = a}.
*/
static void secp256k1_fe_to_storage(secp256k1_fe_storage *r, const secp256k1_fe *a);
/** Convert a field element back from secp256k1_fe_storage.
*
* On input, r need not be initialized.
* Performs {r = a}.
* On output, r will be normalized and will have magnitude 1.
*/
static void secp256k1_fe_from_storage(secp256k1_fe *r, const secp256k1_fe_storage *a);
/** If flag is true, set *r equal to *a; otherwise leave it. Constant-time. Both *r and *a must be initialized.*/
static void secp256k1_fe_storage_cmov(secp256k1_fe_storage *r, const secp256k1_fe_storage *a, int flag);
/** Conditionally move a field element in constant time.
*
* On input, both r and a must be valid field elements. Flag must be 0 or 1.
* Performs {r = flag ? a : r}.
*
* On output, r's magnitude will be the maximum of both input magnitudes.
* It will be normalized if and only if both inputs were normalized.
*/
static void secp256k1_fe_cmov(secp256k1_fe *r, const secp256k1_fe *a, int flag);
/** Halve the value of a field element modulo the field prime in constant-time.
*
* On input, r must be a valid field element.
* On output, r will be normalized and have magnitude floor(m/2) + 1 where m is
* the magnitude of r on input.
*/
static void secp256k1_fe_half(secp256k1_fe *r);
/** Sets r to a field element with magnitude m, normalized if (and only if) m==0.
* The value is chosen so that it is likely to trigger edge cases related to
* internal overflows. */
static void secp256k1_fe_get_bounds(secp256k1_fe *r, int m);
/** Determine whether a is a square (modulo p).
*
* On input, a must be a valid field element.
*/
static int secp256k1_fe_is_square_var(const secp256k1_fe *a);
/** Check invariants on a field element (no-op unless VERIFY is enabled). */
static void secp256k1_fe_verify(const secp256k1_fe *a);
#define SECP256K1_FE_VERIFY(a) secp256k1_fe_verify(a)
/** Check that magnitude of a is at most m (no-op unless VERIFY is enabled). */
static void secp256k1_fe_verify_magnitude(const secp256k1_fe *a, int m);
#define SECP256K1_FE_VERIFY_MAGNITUDE(a, m) secp256k1_fe_verify_magnitude(a, m)
#endif /* SECP256K1_FIELD_H */

View File

@@ -0,0 +1,57 @@
/***********************************************************************
* Copyright (c) 2013, 2014 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_FIELD_REPR_H
#define SECP256K1_FIELD_REPR_H
#include <stdint.h>
/** This field implementation represents the value as 10 uint32_t limbs in base
* 2^26. */
typedef struct {
/* A field element f represents the sum(i=0..9, f.n[i] << (i*26)) mod p,
* where p is the field modulus, 2^256 - 2^32 - 977.
*
* The individual limbs f.n[i] can exceed 2^26; the field's magnitude roughly
* corresponds to how much excess is allowed. The value
* sum(i=0..9, f.n[i] << (i*26)) may exceed p, unless the field element is
* normalized. */
uint32_t n[10];
/*
* Magnitude m requires:
* n[i] <= 2 * m * (2^26 - 1) for i=0..8
* n[9] <= 2 * m * (2^22 - 1)
*
* Normalized requires:
* n[i] <= (2^26 - 1) for i=0..8
* sum(i=0..9, n[i] << (i*26)) < p
* (together these imply n[9] <= 2^22 - 1)
*/
SECP256K1_FE_VERIFY_FIELDS
} secp256k1_fe;
/* Unpacks a constant into a overlapping multi-limbed FE element. */
#define SECP256K1_FE_CONST_INNER(d7, d6, d5, d4, d3, d2, d1, d0) { \
(d0) & 0x3FFFFFFUL, \
(((uint32_t)d0) >> 26) | (((uint32_t)(d1) & 0xFFFFFUL) << 6), \
(((uint32_t)d1) >> 20) | (((uint32_t)(d2) & 0x3FFFUL) << 12), \
(((uint32_t)d2) >> 14) | (((uint32_t)(d3) & 0xFFUL) << 18), \
(((uint32_t)d3) >> 8) | (((uint32_t)(d4) & 0x3UL) << 24), \
(((uint32_t)d4) >> 2) & 0x3FFFFFFUL, \
(((uint32_t)d4) >> 28) | (((uint32_t)(d5) & 0x3FFFFFUL) << 4), \
(((uint32_t)d5) >> 22) | (((uint32_t)(d6) & 0xFFFFUL) << 10), \
(((uint32_t)d6) >> 16) | (((uint32_t)(d7) & 0x3FFUL) << 16), \
(((uint32_t)d7) >> 10) \
}
typedef struct {
uint32_t n[8];
} secp256k1_fe_storage;
#define SECP256K1_FE_STORAGE_CONST(d7, d6, d5, d4, d3, d2, d1, d0) {{ (d0), (d1), (d2), (d3), (d4), (d5), (d6), (d7) }}
#define SECP256K1_FE_STORAGE_CONST_GET(d) d.n[7], d.n[6], d.n[5], d.n[4],d.n[3], d.n[2], d.n[1], d.n[0]
#endif /* SECP256K1_FIELD_REPR_H */

File diff suppressed because it is too large Load Diff

View File

@@ -0,0 +1,62 @@
/***********************************************************************
* Copyright (c) 2013, 2014 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_FIELD_REPR_H
#define SECP256K1_FIELD_REPR_H
#include <stdint.h>
/** This field implementation represents the value as 5 uint64_t limbs in base
* 2^52. */
typedef struct {
/* A field element f represents the sum(i=0..4, f.n[i] << (i*52)) mod p,
* where p is the field modulus, 2^256 - 2^32 - 977.
*
* The individual limbs f.n[i] can exceed 2^52; the field's magnitude roughly
* corresponds to how much excess is allowed. The value
* sum(i=0..4, f.n[i] << (i*52)) may exceed p, unless the field element is
* normalized. */
uint64_t n[5];
/*
* Magnitude m requires:
* n[i] <= 2 * m * (2^52 - 1) for i=0..3
* n[4] <= 2 * m * (2^48 - 1)
*
* Normalized requires:
* n[i] <= (2^52 - 1) for i=0..3
* sum(i=0..4, n[i] << (i*52)) < p
* (together these imply n[4] <= 2^48 - 1)
*/
SECP256K1_FE_VERIFY_FIELDS
} secp256k1_fe;
/* Unpacks a constant into a overlapping multi-limbed FE element. */
#define SECP256K1_FE_CONST_INNER(d7, d6, d5, d4, d3, d2, d1, d0) { \
(d0) | (((uint64_t)(d1) & 0xFFFFFUL) << 32), \
((uint64_t)(d1) >> 20) | (((uint64_t)(d2)) << 12) | (((uint64_t)(d3) & 0xFFUL) << 44), \
((uint64_t)(d3) >> 8) | (((uint64_t)(d4) & 0xFFFFFFFUL) << 24), \
((uint64_t)(d4) >> 28) | (((uint64_t)(d5)) << 4) | (((uint64_t)(d6) & 0xFFFFUL) << 36), \
((uint64_t)(d6) >> 16) | (((uint64_t)(d7)) << 16) \
}
typedef struct {
uint64_t n[4];
} secp256k1_fe_storage;
#define SECP256K1_FE_STORAGE_CONST(d7, d6, d5, d4, d3, d2, d1, d0) {{ \
(d0) | (((uint64_t)(d1)) << 32), \
(d2) | (((uint64_t)(d3)) << 32), \
(d4) | (((uint64_t)(d5)) << 32), \
(d6) | (((uint64_t)(d7)) << 32) \
}}
#define SECP256K1_FE_STORAGE_CONST_GET(d) \
(uint32_t)(d.n[3] >> 32), (uint32_t)d.n[3], \
(uint32_t)(d.n[2] >> 32), (uint32_t)d.n[2], \
(uint32_t)(d.n[1] >> 32), (uint32_t)d.n[1], \
(uint32_t)(d.n[0] >> 32), (uint32_t)d.n[0]
#endif /* SECP256K1_FIELD_REPR_H */

View File

@@ -0,0 +1,522 @@
/***********************************************************************
* Copyright (c) 2013, 2014 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_FIELD_REPR_IMPL_H
#define SECP256K1_FIELD_REPR_IMPL_H
#include "checkmem.h"
#include "util.h"
#include "field.h"
#include "modinv64_impl.h"
#include "field_5x52_int128_impl.h"
#ifdef VERIFY
static void secp256k1_fe_impl_verify(const secp256k1_fe *a) {
const uint64_t *d = a->n;
int m = a->normalized ? 1 : 2 * a->magnitude;
/* secp256k1 'p' value defined in "Standards for Efficient Cryptography" (SEC2) 2.7.1. */
VERIFY_CHECK(d[0] <= 0xFFFFFFFFFFFFFULL * m);
VERIFY_CHECK(d[1] <= 0xFFFFFFFFFFFFFULL * m);
VERIFY_CHECK(d[2] <= 0xFFFFFFFFFFFFFULL * m);
VERIFY_CHECK(d[3] <= 0xFFFFFFFFFFFFFULL * m);
VERIFY_CHECK(d[4] <= 0x0FFFFFFFFFFFFULL * m);
if (a->normalized) {
if ((d[4] == 0x0FFFFFFFFFFFFULL) && ((d[3] & d[2] & d[1]) == 0xFFFFFFFFFFFFFULL)) {
VERIFY_CHECK(d[0] < 0xFFFFEFFFFFC2FULL);
}
}
}
#endif
static void secp256k1_fe_impl_get_bounds(secp256k1_fe *r, int m) {
r->n[0] = 0xFFFFFFFFFFFFFULL * 2 * m;
r->n[1] = 0xFFFFFFFFFFFFFULL * 2 * m;
r->n[2] = 0xFFFFFFFFFFFFFULL * 2 * m;
r->n[3] = 0xFFFFFFFFFFFFFULL * 2 * m;
r->n[4] = 0x0FFFFFFFFFFFFULL * 2 * m;
}
static void secp256k1_fe_impl_normalize(secp256k1_fe *r) {
uint64_t t0 = r->n[0], t1 = r->n[1], t2 = r->n[2], t3 = r->n[3], t4 = r->n[4];
/* Reduce t4 at the start so there will be at most a single carry from the first pass */
uint64_t m;
uint64_t x = t4 >> 48; t4 &= 0x0FFFFFFFFFFFFULL;
/* The first pass ensures the magnitude is 1, ... */
t0 += x * 0x1000003D1ULL;
t1 += (t0 >> 52); t0 &= 0xFFFFFFFFFFFFFULL;
t2 += (t1 >> 52); t1 &= 0xFFFFFFFFFFFFFULL; m = t1;
t3 += (t2 >> 52); t2 &= 0xFFFFFFFFFFFFFULL; m &= t2;
t4 += (t3 >> 52); t3 &= 0xFFFFFFFFFFFFFULL; m &= t3;
/* ... except for a possible carry at bit 48 of t4 (i.e. bit 256 of the field element) */
VERIFY_CHECK(t4 >> 49 == 0);
/* At most a single final reduction is needed; check if the value is >= the field characteristic */
x = (t4 >> 48) | ((t4 == 0x0FFFFFFFFFFFFULL) & (m == 0xFFFFFFFFFFFFFULL)
& (t0 >= 0xFFFFEFFFFFC2FULL));
/* Apply the final reduction (for constant-time behaviour, we do it always) */
t0 += x * 0x1000003D1ULL;
t1 += (t0 >> 52); t0 &= 0xFFFFFFFFFFFFFULL;
t2 += (t1 >> 52); t1 &= 0xFFFFFFFFFFFFFULL;
t3 += (t2 >> 52); t2 &= 0xFFFFFFFFFFFFFULL;
t4 += (t3 >> 52); t3 &= 0xFFFFFFFFFFFFFULL;
/* If t4 didn't carry to bit 48 already, then it should have after any final reduction */
VERIFY_CHECK(t4 >> 48 == x);
/* Mask off the possible multiple of 2^256 from the final reduction */
t4 &= 0x0FFFFFFFFFFFFULL;
r->n[0] = t0; r->n[1] = t1; r->n[2] = t2; r->n[3] = t3; r->n[4] = t4;
}
static void secp256k1_fe_impl_normalize_weak(secp256k1_fe *r) {
uint64_t t0 = r->n[0], t1 = r->n[1], t2 = r->n[2], t3 = r->n[3], t4 = r->n[4];
/* Reduce t4 at the start so there will be at most a single carry from the first pass */
uint64_t x = t4 >> 48; t4 &= 0x0FFFFFFFFFFFFULL;
/* The first pass ensures the magnitude is 1, ... */
t0 += x * 0x1000003D1ULL;
t1 += (t0 >> 52); t0 &= 0xFFFFFFFFFFFFFULL;
t2 += (t1 >> 52); t1 &= 0xFFFFFFFFFFFFFULL;
t3 += (t2 >> 52); t2 &= 0xFFFFFFFFFFFFFULL;
t4 += (t3 >> 52); t3 &= 0xFFFFFFFFFFFFFULL;
/* ... except for a possible carry at bit 48 of t4 (i.e. bit 256 of the field element) */
VERIFY_CHECK(t4 >> 49 == 0);
r->n[0] = t0; r->n[1] = t1; r->n[2] = t2; r->n[3] = t3; r->n[4] = t4;
}
static void secp256k1_fe_impl_normalize_var(secp256k1_fe *r) {
uint64_t t0 = r->n[0], t1 = r->n[1], t2 = r->n[2], t3 = r->n[3], t4 = r->n[4];
/* Reduce t4 at the start so there will be at most a single carry from the first pass */
uint64_t m;
uint64_t x = t4 >> 48; t4 &= 0x0FFFFFFFFFFFFULL;
/* The first pass ensures the magnitude is 1, ... */
t0 += x * 0x1000003D1ULL;
t1 += (t0 >> 52); t0 &= 0xFFFFFFFFFFFFFULL;
t2 += (t1 >> 52); t1 &= 0xFFFFFFFFFFFFFULL; m = t1;
t3 += (t2 >> 52); t2 &= 0xFFFFFFFFFFFFFULL; m &= t2;
t4 += (t3 >> 52); t3 &= 0xFFFFFFFFFFFFFULL; m &= t3;
/* ... except for a possible carry at bit 48 of t4 (i.e. bit 256 of the field element) */
VERIFY_CHECK(t4 >> 49 == 0);
/* At most a single final reduction is needed; check if the value is >= the field characteristic */
x = (t4 >> 48) | ((t4 == 0x0FFFFFFFFFFFFULL) & (m == 0xFFFFFFFFFFFFFULL)
& (t0 >= 0xFFFFEFFFFFC2FULL));
if (x) {
t0 += 0x1000003D1ULL;
t1 += (t0 >> 52); t0 &= 0xFFFFFFFFFFFFFULL;
t2 += (t1 >> 52); t1 &= 0xFFFFFFFFFFFFFULL;
t3 += (t2 >> 52); t2 &= 0xFFFFFFFFFFFFFULL;
t4 += (t3 >> 52); t3 &= 0xFFFFFFFFFFFFFULL;
/* If t4 didn't carry to bit 48 already, then it should have after any final reduction */
VERIFY_CHECK(t4 >> 48 == x);
/* Mask off the possible multiple of 2^256 from the final reduction */
t4 &= 0x0FFFFFFFFFFFFULL;
}
r->n[0] = t0; r->n[1] = t1; r->n[2] = t2; r->n[3] = t3; r->n[4] = t4;
}
static int secp256k1_fe_impl_normalizes_to_zero(const secp256k1_fe *r) {
uint64_t t0 = r->n[0], t1 = r->n[1], t2 = r->n[2], t3 = r->n[3], t4 = r->n[4];
/* z0 tracks a possible raw value of 0, z1 tracks a possible raw value of P */
uint64_t z0, z1;
/* Reduce t4 at the start so there will be at most a single carry from the first pass */
uint64_t x = t4 >> 48; t4 &= 0x0FFFFFFFFFFFFULL;
/* The first pass ensures the magnitude is 1, ... */
t0 += x * 0x1000003D1ULL;
t1 += (t0 >> 52); t0 &= 0xFFFFFFFFFFFFFULL; z0 = t0; z1 = t0 ^ 0x1000003D0ULL;
t2 += (t1 >> 52); t1 &= 0xFFFFFFFFFFFFFULL; z0 |= t1; z1 &= t1;
t3 += (t2 >> 52); t2 &= 0xFFFFFFFFFFFFFULL; z0 |= t2; z1 &= t2;
t4 += (t3 >> 52); t3 &= 0xFFFFFFFFFFFFFULL; z0 |= t3; z1 &= t3;
z0 |= t4; z1 &= t4 ^ 0xF000000000000ULL;
/* ... except for a possible carry at bit 48 of t4 (i.e. bit 256 of the field element) */
VERIFY_CHECK(t4 >> 49 == 0);
return (z0 == 0) | (z1 == 0xFFFFFFFFFFFFFULL);
}
static int secp256k1_fe_impl_normalizes_to_zero_var(const secp256k1_fe *r) {
uint64_t t0, t1, t2, t3, t4;
uint64_t z0, z1;
uint64_t x;
t0 = r->n[0];
t4 = r->n[4];
/* Reduce t4 at the start so there will be at most a single carry from the first pass */
x = t4 >> 48;
/* The first pass ensures the magnitude is 1, ... */
t0 += x * 0x1000003D1ULL;
/* z0 tracks a possible raw value of 0, z1 tracks a possible raw value of P */
z0 = t0 & 0xFFFFFFFFFFFFFULL;
z1 = z0 ^ 0x1000003D0ULL;
/* Fast return path should catch the majority of cases */
if ((z0 != 0ULL) & (z1 != 0xFFFFFFFFFFFFFULL)) {
return 0;
}
t1 = r->n[1];
t2 = r->n[2];
t3 = r->n[3];
t4 &= 0x0FFFFFFFFFFFFULL;
t1 += (t0 >> 52);
t2 += (t1 >> 52); t1 &= 0xFFFFFFFFFFFFFULL; z0 |= t1; z1 &= t1;
t3 += (t2 >> 52); t2 &= 0xFFFFFFFFFFFFFULL; z0 |= t2; z1 &= t2;
t4 += (t3 >> 52); t3 &= 0xFFFFFFFFFFFFFULL; z0 |= t3; z1 &= t3;
z0 |= t4; z1 &= t4 ^ 0xF000000000000ULL;
/* ... except for a possible carry at bit 48 of t4 (i.e. bit 256 of the field element) */
VERIFY_CHECK(t4 >> 49 == 0);
return (z0 == 0) | (z1 == 0xFFFFFFFFFFFFFULL);
}
SECP256K1_INLINE static void secp256k1_fe_impl_set_int(secp256k1_fe *r, int a) {
r->n[0] = a;
r->n[1] = r->n[2] = r->n[3] = r->n[4] = 0;
}
SECP256K1_INLINE static int secp256k1_fe_impl_is_zero(const secp256k1_fe *a) {
const uint64_t *t = a->n;
return (t[0] | t[1] | t[2] | t[3] | t[4]) == 0;
}
SECP256K1_INLINE static int secp256k1_fe_impl_is_odd(const secp256k1_fe *a) {
return a->n[0] & 1;
}
static int secp256k1_fe_impl_cmp_var(const secp256k1_fe *a, const secp256k1_fe *b) {
int i;
for (i = 4; i >= 0; i--) {
if (a->n[i] > b->n[i]) {
return 1;
}
if (a->n[i] < b->n[i]) {
return -1;
}
}
return 0;
}
static void secp256k1_fe_impl_set_b32_mod(secp256k1_fe *r, const unsigned char *a) {
r->n[0] = (uint64_t)a[31]
| ((uint64_t)a[30] << 8)
| ((uint64_t)a[29] << 16)
| ((uint64_t)a[28] << 24)
| ((uint64_t)a[27] << 32)
| ((uint64_t)a[26] << 40)
| ((uint64_t)(a[25] & 0xF) << 48);
r->n[1] = (uint64_t)((a[25] >> 4) & 0xF)
| ((uint64_t)a[24] << 4)
| ((uint64_t)a[23] << 12)
| ((uint64_t)a[22] << 20)
| ((uint64_t)a[21] << 28)
| ((uint64_t)a[20] << 36)
| ((uint64_t)a[19] << 44);
r->n[2] = (uint64_t)a[18]
| ((uint64_t)a[17] << 8)
| ((uint64_t)a[16] << 16)
| ((uint64_t)a[15] << 24)
| ((uint64_t)a[14] << 32)
| ((uint64_t)a[13] << 40)
| ((uint64_t)(a[12] & 0xF) << 48);
r->n[3] = (uint64_t)((a[12] >> 4) & 0xF)
| ((uint64_t)a[11] << 4)
| ((uint64_t)a[10] << 12)
| ((uint64_t)a[9] << 20)
| ((uint64_t)a[8] << 28)
| ((uint64_t)a[7] << 36)
| ((uint64_t)a[6] << 44);
r->n[4] = (uint64_t)a[5]
| ((uint64_t)a[4] << 8)
| ((uint64_t)a[3] << 16)
| ((uint64_t)a[2] << 24)
| ((uint64_t)a[1] << 32)
| ((uint64_t)a[0] << 40);
}
static int secp256k1_fe_impl_set_b32_limit(secp256k1_fe *r, const unsigned char *a) {
secp256k1_fe_impl_set_b32_mod(r, a);
return !((r->n[4] == 0x0FFFFFFFFFFFFULL) & ((r->n[3] & r->n[2] & r->n[1]) == 0xFFFFFFFFFFFFFULL) & (r->n[0] >= 0xFFFFEFFFFFC2FULL));
}
/** Convert a field element to a 32-byte big endian value. Requires the input to be normalized */
static void secp256k1_fe_impl_get_b32(unsigned char *r, const secp256k1_fe *a) {
r[0] = (a->n[4] >> 40) & 0xFF;
r[1] = (a->n[4] >> 32) & 0xFF;
r[2] = (a->n[4] >> 24) & 0xFF;
r[3] = (a->n[4] >> 16) & 0xFF;
r[4] = (a->n[4] >> 8) & 0xFF;
r[5] = a->n[4] & 0xFF;
r[6] = (a->n[3] >> 44) & 0xFF;
r[7] = (a->n[3] >> 36) & 0xFF;
r[8] = (a->n[3] >> 28) & 0xFF;
r[9] = (a->n[3] >> 20) & 0xFF;
r[10] = (a->n[3] >> 12) & 0xFF;
r[11] = (a->n[3] >> 4) & 0xFF;
r[12] = ((a->n[2] >> 48) & 0xF) | ((a->n[3] & 0xF) << 4);
r[13] = (a->n[2] >> 40) & 0xFF;
r[14] = (a->n[2] >> 32) & 0xFF;
r[15] = (a->n[2] >> 24) & 0xFF;
r[16] = (a->n[2] >> 16) & 0xFF;
r[17] = (a->n[2] >> 8) & 0xFF;
r[18] = a->n[2] & 0xFF;
r[19] = (a->n[1] >> 44) & 0xFF;
r[20] = (a->n[1] >> 36) & 0xFF;
r[21] = (a->n[1] >> 28) & 0xFF;
r[22] = (a->n[1] >> 20) & 0xFF;
r[23] = (a->n[1] >> 12) & 0xFF;
r[24] = (a->n[1] >> 4) & 0xFF;
r[25] = ((a->n[0] >> 48) & 0xF) | ((a->n[1] & 0xF) << 4);
r[26] = (a->n[0] >> 40) & 0xFF;
r[27] = (a->n[0] >> 32) & 0xFF;
r[28] = (a->n[0] >> 24) & 0xFF;
r[29] = (a->n[0] >> 16) & 0xFF;
r[30] = (a->n[0] >> 8) & 0xFF;
r[31] = a->n[0] & 0xFF;
}
SECP256K1_INLINE static void secp256k1_fe_impl_negate_unchecked(secp256k1_fe *r, const secp256k1_fe *a, int m) {
/* For all legal values of m (0..31), the following properties hold: */
VERIFY_CHECK(0xFFFFEFFFFFC2FULL * 2 * (m + 1) >= 0xFFFFFFFFFFFFFULL * 2 * m);
VERIFY_CHECK(0xFFFFFFFFFFFFFULL * 2 * (m + 1) >= 0xFFFFFFFFFFFFFULL * 2 * m);
VERIFY_CHECK(0x0FFFFFFFFFFFFULL * 2 * (m + 1) >= 0x0FFFFFFFFFFFFULL * 2 * m);
/* Due to the properties above, the left hand in the subtractions below is never less than
* the right hand. */
r->n[0] = 0xFFFFEFFFFFC2FULL * 2 * (m + 1) - a->n[0];
r->n[1] = 0xFFFFFFFFFFFFFULL * 2 * (m + 1) - a->n[1];
r->n[2] = 0xFFFFFFFFFFFFFULL * 2 * (m + 1) - a->n[2];
r->n[3] = 0xFFFFFFFFFFFFFULL * 2 * (m + 1) - a->n[3];
r->n[4] = 0x0FFFFFFFFFFFFULL * 2 * (m + 1) - a->n[4];
}
SECP256K1_INLINE static void secp256k1_fe_impl_mul_int_unchecked(secp256k1_fe *r, int a) {
r->n[0] *= a;
r->n[1] *= a;
r->n[2] *= a;
r->n[3] *= a;
r->n[4] *= a;
}
SECP256K1_INLINE static void secp256k1_fe_impl_add_int(secp256k1_fe *r, int a) {
r->n[0] += a;
}
SECP256K1_INLINE static void secp256k1_fe_impl_add(secp256k1_fe *r, const secp256k1_fe *a) {
r->n[0] += a->n[0];
r->n[1] += a->n[1];
r->n[2] += a->n[2];
r->n[3] += a->n[3];
r->n[4] += a->n[4];
}
SECP256K1_INLINE static void secp256k1_fe_impl_mul(secp256k1_fe *r, const secp256k1_fe *a, const secp256k1_fe * SECP256K1_RESTRICT b) {
secp256k1_fe_mul_inner(r->n, a->n, b->n);
}
SECP256K1_INLINE static void secp256k1_fe_impl_sqr(secp256k1_fe *r, const secp256k1_fe *a) {
secp256k1_fe_sqr_inner(r->n, a->n);
}
SECP256K1_INLINE static void secp256k1_fe_impl_cmov(secp256k1_fe *r, const secp256k1_fe *a, int flag) {
uint64_t mask0, mask1;
volatile int vflag = flag;
SECP256K1_CHECKMEM_CHECK_VERIFY(r->n, sizeof(r->n));
mask0 = vflag + ~((uint64_t)0);
mask1 = ~mask0;
r->n[0] = (r->n[0] & mask0) | (a->n[0] & mask1);
r->n[1] = (r->n[1] & mask0) | (a->n[1] & mask1);
r->n[2] = (r->n[2] & mask0) | (a->n[2] & mask1);
r->n[3] = (r->n[3] & mask0) | (a->n[3] & mask1);
r->n[4] = (r->n[4] & mask0) | (a->n[4] & mask1);
}
static SECP256K1_INLINE void secp256k1_fe_impl_half(secp256k1_fe *r) {
uint64_t t0 = r->n[0], t1 = r->n[1], t2 = r->n[2], t3 = r->n[3], t4 = r->n[4];
uint64_t one = (uint64_t)1;
uint64_t mask = -(t0 & one) >> 12;
/* Bounds analysis (over the rationals).
*
* Let m = r->magnitude
* C = 0xFFFFFFFFFFFFFULL * 2
* D = 0x0FFFFFFFFFFFFULL * 2
*
* Initial bounds: t0..t3 <= C * m
* t4 <= D * m
*/
t0 += 0xFFFFEFFFFFC2FULL & mask;
t1 += mask;
t2 += mask;
t3 += mask;
t4 += mask >> 4;
VERIFY_CHECK((t0 & one) == 0);
/* t0..t3: added <= C/2
* t4: added <= D/2
*
* Current bounds: t0..t3 <= C * (m + 1/2)
* t4 <= D * (m + 1/2)
*/
r->n[0] = (t0 >> 1) + ((t1 & one) << 51);
r->n[1] = (t1 >> 1) + ((t2 & one) << 51);
r->n[2] = (t2 >> 1) + ((t3 & one) << 51);
r->n[3] = (t3 >> 1) + ((t4 & one) << 51);
r->n[4] = (t4 >> 1);
/* t0..t3: shifted right and added <= C/4 + 1/2
* t4: shifted right
*
* Current bounds: t0..t3 <= C * (m/2 + 1/2)
* t4 <= D * (m/2 + 1/4)
*
* Therefore the output magnitude (M) has to be set such that:
* t0..t3: C * M >= C * (m/2 + 1/2)
* t4: D * M >= D * (m/2 + 1/4)
*
* It suffices for all limbs that, for any input magnitude m:
* M >= m/2 + 1/2
*
* and since we want the smallest such integer value for M:
* M == floor(m/2) + 1
*/
}
static SECP256K1_INLINE void secp256k1_fe_storage_cmov(secp256k1_fe_storage *r, const secp256k1_fe_storage *a, int flag) {
uint64_t mask0, mask1;
volatile int vflag = flag;
SECP256K1_CHECKMEM_CHECK_VERIFY(r->n, sizeof(r->n));
mask0 = vflag + ~((uint64_t)0);
mask1 = ~mask0;
r->n[0] = (r->n[0] & mask0) | (a->n[0] & mask1);
r->n[1] = (r->n[1] & mask0) | (a->n[1] & mask1);
r->n[2] = (r->n[2] & mask0) | (a->n[2] & mask1);
r->n[3] = (r->n[3] & mask0) | (a->n[3] & mask1);
}
static void secp256k1_fe_impl_to_storage(secp256k1_fe_storage *r, const secp256k1_fe *a) {
r->n[0] = a->n[0] | a->n[1] << 52;
r->n[1] = a->n[1] >> 12 | a->n[2] << 40;
r->n[2] = a->n[2] >> 24 | a->n[3] << 28;
r->n[3] = a->n[3] >> 36 | a->n[4] << 16;
}
static SECP256K1_INLINE void secp256k1_fe_impl_from_storage(secp256k1_fe *r, const secp256k1_fe_storage *a) {
r->n[0] = a->n[0] & 0xFFFFFFFFFFFFFULL;
r->n[1] = a->n[0] >> 52 | ((a->n[1] << 12) & 0xFFFFFFFFFFFFFULL);
r->n[2] = a->n[1] >> 40 | ((a->n[2] << 24) & 0xFFFFFFFFFFFFFULL);
r->n[3] = a->n[2] >> 28 | ((a->n[3] << 36) & 0xFFFFFFFFFFFFFULL);
r->n[4] = a->n[3] >> 16;
}
static void secp256k1_fe_from_signed62(secp256k1_fe *r, const secp256k1_modinv64_signed62 *a) {
const uint64_t M52 = UINT64_MAX >> 12;
const uint64_t a0 = a->v[0], a1 = a->v[1], a2 = a->v[2], a3 = a->v[3], a4 = a->v[4];
/* The output from secp256k1_modinv64{_var} should be normalized to range [0,modulus), and
* have limbs in [0,2^62). The modulus is < 2^256, so the top limb must be below 2^(256-62*4).
*/
VERIFY_CHECK(a0 >> 62 == 0);
VERIFY_CHECK(a1 >> 62 == 0);
VERIFY_CHECK(a2 >> 62 == 0);
VERIFY_CHECK(a3 >> 62 == 0);
VERIFY_CHECK(a4 >> 8 == 0);
r->n[0] = a0 & M52;
r->n[1] = (a0 >> 52 | a1 << 10) & M52;
r->n[2] = (a1 >> 42 | a2 << 20) & M52;
r->n[3] = (a2 >> 32 | a3 << 30) & M52;
r->n[4] = (a3 >> 22 | a4 << 40);
}
static void secp256k1_fe_to_signed62(secp256k1_modinv64_signed62 *r, const secp256k1_fe *a) {
const uint64_t M62 = UINT64_MAX >> 2;
const uint64_t a0 = a->n[0], a1 = a->n[1], a2 = a->n[2], a3 = a->n[3], a4 = a->n[4];
r->v[0] = (a0 | a1 << 52) & M62;
r->v[1] = (a1 >> 10 | a2 << 42) & M62;
r->v[2] = (a2 >> 20 | a3 << 32) & M62;
r->v[3] = (a3 >> 30 | a4 << 22) & M62;
r->v[4] = a4 >> 40;
}
static const secp256k1_modinv64_modinfo secp256k1_const_modinfo_fe = {
{{-0x1000003D1LL, 0, 0, 0, 256}},
0x27C7F6E22DDACACFLL
};
static void secp256k1_fe_impl_inv(secp256k1_fe *r, const secp256k1_fe *x) {
secp256k1_fe tmp = *x;
secp256k1_modinv64_signed62 s;
secp256k1_fe_normalize(&tmp);
secp256k1_fe_to_signed62(&s, &tmp);
secp256k1_modinv64(&s, &secp256k1_const_modinfo_fe);
secp256k1_fe_from_signed62(r, &s);
}
static void secp256k1_fe_impl_inv_var(secp256k1_fe *r, const secp256k1_fe *x) {
secp256k1_fe tmp = *x;
secp256k1_modinv64_signed62 s;
secp256k1_fe_normalize_var(&tmp);
secp256k1_fe_to_signed62(&s, &tmp);
secp256k1_modinv64_var(&s, &secp256k1_const_modinfo_fe);
secp256k1_fe_from_signed62(r, &s);
}
static int secp256k1_fe_impl_is_square_var(const secp256k1_fe *x) {
secp256k1_fe tmp;
secp256k1_modinv64_signed62 s;
int jac, ret;
tmp = *x;
secp256k1_fe_normalize_var(&tmp);
/* secp256k1_jacobi64_maybe_var cannot deal with input 0. */
if (secp256k1_fe_is_zero(&tmp)) return 1;
secp256k1_fe_to_signed62(&s, &tmp);
jac = secp256k1_jacobi64_maybe_var(&s, &secp256k1_const_modinfo_fe);
if (jac == 0) {
/* secp256k1_jacobi64_maybe_var failed to compute the Jacobi symbol. Fall back
* to computing a square root. This should be extremely rare with random
* input (except in VERIFY mode, where a lower iteration count is used). */
secp256k1_fe dummy;
ret = secp256k1_fe_sqrt(&dummy, &tmp);
} else {
ret = jac >= 0;
}
return ret;
}
#endif /* SECP256K1_FIELD_REPR_IMPL_H */

View File

@@ -0,0 +1,274 @@
/***********************************************************************
* Copyright (c) 2013, 2014 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_FIELD_INNER5X52_IMPL_H
#define SECP256K1_FIELD_INNER5X52_IMPL_H
#include <stdint.h>
#include "int128.h"
#include "util.h"
#define VERIFY_BITS(x, n) VERIFY_CHECK(((x) >> (n)) == 0)
#define VERIFY_BITS_128(x, n) VERIFY_CHECK(secp256k1_u128_check_bits((x), (n)))
SECP256K1_INLINE static void secp256k1_fe_mul_inner(uint64_t *r, const uint64_t *a, const uint64_t * SECP256K1_RESTRICT b) {
secp256k1_uint128 c, d;
uint64_t t3, t4, tx, u0;
uint64_t a0 = a[0], a1 = a[1], a2 = a[2], a3 = a[3], a4 = a[4];
const uint64_t M = 0xFFFFFFFFFFFFFULL, R = 0x1000003D10ULL;
VERIFY_BITS(a[0], 56);
VERIFY_BITS(a[1], 56);
VERIFY_BITS(a[2], 56);
VERIFY_BITS(a[3], 56);
VERIFY_BITS(a[4], 52);
VERIFY_BITS(b[0], 56);
VERIFY_BITS(b[1], 56);
VERIFY_BITS(b[2], 56);
VERIFY_BITS(b[3], 56);
VERIFY_BITS(b[4], 52);
VERIFY_CHECK(r != b);
VERIFY_CHECK(a != b);
/* [... a b c] is a shorthand for ... + a<<104 + b<<52 + c<<0 mod n.
* for 0 <= x <= 4, px is a shorthand for sum(a[i]*b[x-i], i=0..x).
* for 4 <= x <= 8, px is a shorthand for sum(a[i]*b[x-i], i=(x-4)..4)
* Note that [x 0 0 0 0 0] = [x*R].
*/
secp256k1_u128_mul(&d, a0, b[3]);
secp256k1_u128_accum_mul(&d, a1, b[2]);
secp256k1_u128_accum_mul(&d, a2, b[1]);
secp256k1_u128_accum_mul(&d, a3, b[0]);
VERIFY_BITS_128(&d, 114);
/* [d 0 0 0] = [p3 0 0 0] */
secp256k1_u128_mul(&c, a4, b[4]);
VERIFY_BITS_128(&c, 112);
/* [c 0 0 0 0 d 0 0 0] = [p8 0 0 0 0 p3 0 0 0] */
secp256k1_u128_accum_mul(&d, R, secp256k1_u128_to_u64(&c)); secp256k1_u128_rshift(&c, 64);
VERIFY_BITS_128(&d, 115);
VERIFY_BITS_128(&c, 48);
/* [(c<<12) 0 0 0 0 0 d 0 0 0] = [p8 0 0 0 0 p3 0 0 0] */
t3 = secp256k1_u128_to_u64(&d) & M; secp256k1_u128_rshift(&d, 52);
VERIFY_BITS(t3, 52);
VERIFY_BITS_128(&d, 63);
/* [(c<<12) 0 0 0 0 d t3 0 0 0] = [p8 0 0 0 0 p3 0 0 0] */
secp256k1_u128_accum_mul(&d, a0, b[4]);
secp256k1_u128_accum_mul(&d, a1, b[3]);
secp256k1_u128_accum_mul(&d, a2, b[2]);
secp256k1_u128_accum_mul(&d, a3, b[1]);
secp256k1_u128_accum_mul(&d, a4, b[0]);
VERIFY_BITS_128(&d, 115);
/* [(c<<12) 0 0 0 0 d t3 0 0 0] = [p8 0 0 0 p4 p3 0 0 0] */
secp256k1_u128_accum_mul(&d, R << 12, secp256k1_u128_to_u64(&c));
VERIFY_BITS_128(&d, 116);
/* [d t3 0 0 0] = [p8 0 0 0 p4 p3 0 0 0] */
t4 = secp256k1_u128_to_u64(&d) & M; secp256k1_u128_rshift(&d, 52);
VERIFY_BITS(t4, 52);
VERIFY_BITS_128(&d, 64);
/* [d t4 t3 0 0 0] = [p8 0 0 0 p4 p3 0 0 0] */
tx = (t4 >> 48); t4 &= (M >> 4);
VERIFY_BITS(tx, 4);
VERIFY_BITS(t4, 48);
/* [d t4+(tx<<48) t3 0 0 0] = [p8 0 0 0 p4 p3 0 0 0] */
secp256k1_u128_mul(&c, a0, b[0]);
VERIFY_BITS_128(&c, 112);
/* [d t4+(tx<<48) t3 0 0 c] = [p8 0 0 0 p4 p3 0 0 p0] */
secp256k1_u128_accum_mul(&d, a1, b[4]);
secp256k1_u128_accum_mul(&d, a2, b[3]);
secp256k1_u128_accum_mul(&d, a3, b[2]);
secp256k1_u128_accum_mul(&d, a4, b[1]);
VERIFY_BITS_128(&d, 114);
/* [d t4+(tx<<48) t3 0 0 c] = [p8 0 0 p5 p4 p3 0 0 p0] */
u0 = secp256k1_u128_to_u64(&d) & M; secp256k1_u128_rshift(&d, 52);
VERIFY_BITS(u0, 52);
VERIFY_BITS_128(&d, 62);
/* [d u0 t4+(tx<<48) t3 0 0 c] = [p8 0 0 p5 p4 p3 0 0 p0] */
/* [d 0 t4+(tx<<48)+(u0<<52) t3 0 0 c] = [p8 0 0 p5 p4 p3 0 0 p0] */
u0 = (u0 << 4) | tx;
VERIFY_BITS(u0, 56);
/* [d 0 t4+(u0<<48) t3 0 0 c] = [p8 0 0 p5 p4 p3 0 0 p0] */
secp256k1_u128_accum_mul(&c, u0, R >> 4);
VERIFY_BITS_128(&c, 113);
/* [d 0 t4 t3 0 0 c] = [p8 0 0 p5 p4 p3 0 0 p0] */
r[0] = secp256k1_u128_to_u64(&c) & M; secp256k1_u128_rshift(&c, 52);
VERIFY_BITS(r[0], 52);
VERIFY_BITS_128(&c, 61);
/* [d 0 t4 t3 0 c r0] = [p8 0 0 p5 p4 p3 0 0 p0] */
secp256k1_u128_accum_mul(&c, a0, b[1]);
secp256k1_u128_accum_mul(&c, a1, b[0]);
VERIFY_BITS_128(&c, 114);
/* [d 0 t4 t3 0 c r0] = [p8 0 0 p5 p4 p3 0 p1 p0] */
secp256k1_u128_accum_mul(&d, a2, b[4]);
secp256k1_u128_accum_mul(&d, a3, b[3]);
secp256k1_u128_accum_mul(&d, a4, b[2]);
VERIFY_BITS_128(&d, 114);
/* [d 0 t4 t3 0 c r0] = [p8 0 p6 p5 p4 p3 0 p1 p0] */
secp256k1_u128_accum_mul(&c, secp256k1_u128_to_u64(&d) & M, R); secp256k1_u128_rshift(&d, 52);
VERIFY_BITS_128(&c, 115);
VERIFY_BITS_128(&d, 62);
/* [d 0 0 t4 t3 0 c r0] = [p8 0 p6 p5 p4 p3 0 p1 p0] */
r[1] = secp256k1_u128_to_u64(&c) & M; secp256k1_u128_rshift(&c, 52);
VERIFY_BITS(r[1], 52);
VERIFY_BITS_128(&c, 63);
/* [d 0 0 t4 t3 c r1 r0] = [p8 0 p6 p5 p4 p3 0 p1 p0] */
secp256k1_u128_accum_mul(&c, a0, b[2]);
secp256k1_u128_accum_mul(&c, a1, b[1]);
secp256k1_u128_accum_mul(&c, a2, b[0]);
VERIFY_BITS_128(&c, 114);
/* [d 0 0 t4 t3 c r1 r0] = [p8 0 p6 p5 p4 p3 p2 p1 p0] */
secp256k1_u128_accum_mul(&d, a3, b[4]);
secp256k1_u128_accum_mul(&d, a4, b[3]);
VERIFY_BITS_128(&d, 114);
/* [d 0 0 t4 t3 c t1 r0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */
secp256k1_u128_accum_mul(&c, R, secp256k1_u128_to_u64(&d)); secp256k1_u128_rshift(&d, 64);
VERIFY_BITS_128(&c, 115);
VERIFY_BITS_128(&d, 50);
/* [(d<<12) 0 0 0 t4 t3 c r1 r0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */
r[2] = secp256k1_u128_to_u64(&c) & M; secp256k1_u128_rshift(&c, 52);
VERIFY_BITS(r[2], 52);
VERIFY_BITS_128(&c, 63);
/* [(d<<12) 0 0 0 t4 t3+c r2 r1 r0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */
secp256k1_u128_accum_mul(&c, R << 12, secp256k1_u128_to_u64(&d));
secp256k1_u128_accum_u64(&c, t3);
VERIFY_BITS_128(&c, 100);
/* [t4 c r2 r1 r0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */
r[3] = secp256k1_u128_to_u64(&c) & M; secp256k1_u128_rshift(&c, 52);
VERIFY_BITS(r[3], 52);
VERIFY_BITS_128(&c, 48);
/* [t4+c r3 r2 r1 r0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */
r[4] = secp256k1_u128_to_u64(&c) + t4;
VERIFY_BITS(r[4], 49);
/* [r4 r3 r2 r1 r0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */
}
SECP256K1_INLINE static void secp256k1_fe_sqr_inner(uint64_t *r, const uint64_t *a) {
secp256k1_uint128 c, d;
uint64_t a0 = a[0], a1 = a[1], a2 = a[2], a3 = a[3], a4 = a[4];
uint64_t t3, t4, tx, u0;
const uint64_t M = 0xFFFFFFFFFFFFFULL, R = 0x1000003D10ULL;
VERIFY_BITS(a[0], 56);
VERIFY_BITS(a[1], 56);
VERIFY_BITS(a[2], 56);
VERIFY_BITS(a[3], 56);
VERIFY_BITS(a[4], 52);
/** [... a b c] is a shorthand for ... + a<<104 + b<<52 + c<<0 mod n.
* px is a shorthand for sum(a[i]*a[x-i], i=0..x).
* Note that [x 0 0 0 0 0] = [x*R].
*/
secp256k1_u128_mul(&d, a0*2, a3);
secp256k1_u128_accum_mul(&d, a1*2, a2);
VERIFY_BITS_128(&d, 114);
/* [d 0 0 0] = [p3 0 0 0] */
secp256k1_u128_mul(&c, a4, a4);
VERIFY_BITS_128(&c, 112);
/* [c 0 0 0 0 d 0 0 0] = [p8 0 0 0 0 p3 0 0 0] */
secp256k1_u128_accum_mul(&d, R, secp256k1_u128_to_u64(&c)); secp256k1_u128_rshift(&c, 64);
VERIFY_BITS_128(&d, 115);
VERIFY_BITS_128(&c, 48);
/* [(c<<12) 0 0 0 0 0 d 0 0 0] = [p8 0 0 0 0 p3 0 0 0] */
t3 = secp256k1_u128_to_u64(&d) & M; secp256k1_u128_rshift(&d, 52);
VERIFY_BITS(t3, 52);
VERIFY_BITS_128(&d, 63);
/* [(c<<12) 0 0 0 0 d t3 0 0 0] = [p8 0 0 0 0 p3 0 0 0] */
a4 *= 2;
secp256k1_u128_accum_mul(&d, a0, a4);
secp256k1_u128_accum_mul(&d, a1*2, a3);
secp256k1_u128_accum_mul(&d, a2, a2);
VERIFY_BITS_128(&d, 115);
/* [(c<<12) 0 0 0 0 d t3 0 0 0] = [p8 0 0 0 p4 p3 0 0 0] */
secp256k1_u128_accum_mul(&d, R << 12, secp256k1_u128_to_u64(&c));
VERIFY_BITS_128(&d, 116);
/* [d t3 0 0 0] = [p8 0 0 0 p4 p3 0 0 0] */
t4 = secp256k1_u128_to_u64(&d) & M; secp256k1_u128_rshift(&d, 52);
VERIFY_BITS(t4, 52);
VERIFY_BITS_128(&d, 64);
/* [d t4 t3 0 0 0] = [p8 0 0 0 p4 p3 0 0 0] */
tx = (t4 >> 48); t4 &= (M >> 4);
VERIFY_BITS(tx, 4);
VERIFY_BITS(t4, 48);
/* [d t4+(tx<<48) t3 0 0 0] = [p8 0 0 0 p4 p3 0 0 0] */
secp256k1_u128_mul(&c, a0, a0);
VERIFY_BITS_128(&c, 112);
/* [d t4+(tx<<48) t3 0 0 c] = [p8 0 0 0 p4 p3 0 0 p0] */
secp256k1_u128_accum_mul(&d, a1, a4);
secp256k1_u128_accum_mul(&d, a2*2, a3);
VERIFY_BITS_128(&d, 114);
/* [d t4+(tx<<48) t3 0 0 c] = [p8 0 0 p5 p4 p3 0 0 p0] */
u0 = secp256k1_u128_to_u64(&d) & M; secp256k1_u128_rshift(&d, 52);
VERIFY_BITS(u0, 52);
VERIFY_BITS_128(&d, 62);
/* [d u0 t4+(tx<<48) t3 0 0 c] = [p8 0 0 p5 p4 p3 0 0 p0] */
/* [d 0 t4+(tx<<48)+(u0<<52) t3 0 0 c] = [p8 0 0 p5 p4 p3 0 0 p0] */
u0 = (u0 << 4) | tx;
VERIFY_BITS(u0, 56);
/* [d 0 t4+(u0<<48) t3 0 0 c] = [p8 0 0 p5 p4 p3 0 0 p0] */
secp256k1_u128_accum_mul(&c, u0, R >> 4);
VERIFY_BITS_128(&c, 113);
/* [d 0 t4 t3 0 0 c] = [p8 0 0 p5 p4 p3 0 0 p0] */
r[0] = secp256k1_u128_to_u64(&c) & M; secp256k1_u128_rshift(&c, 52);
VERIFY_BITS(r[0], 52);
VERIFY_BITS_128(&c, 61);
/* [d 0 t4 t3 0 c r0] = [p8 0 0 p5 p4 p3 0 0 p0] */
a0 *= 2;
secp256k1_u128_accum_mul(&c, a0, a1);
VERIFY_BITS_128(&c, 114);
/* [d 0 t4 t3 0 c r0] = [p8 0 0 p5 p4 p3 0 p1 p0] */
secp256k1_u128_accum_mul(&d, a2, a4);
secp256k1_u128_accum_mul(&d, a3, a3);
VERIFY_BITS_128(&d, 114);
/* [d 0 t4 t3 0 c r0] = [p8 0 p6 p5 p4 p3 0 p1 p0] */
secp256k1_u128_accum_mul(&c, secp256k1_u128_to_u64(&d) & M, R); secp256k1_u128_rshift(&d, 52);
VERIFY_BITS_128(&c, 115);
VERIFY_BITS_128(&d, 62);
/* [d 0 0 t4 t3 0 c r0] = [p8 0 p6 p5 p4 p3 0 p1 p0] */
r[1] = secp256k1_u128_to_u64(&c) & M; secp256k1_u128_rshift(&c, 52);
VERIFY_BITS(r[1], 52);
VERIFY_BITS_128(&c, 63);
/* [d 0 0 t4 t3 c r1 r0] = [p8 0 p6 p5 p4 p3 0 p1 p0] */
secp256k1_u128_accum_mul(&c, a0, a2);
secp256k1_u128_accum_mul(&c, a1, a1);
VERIFY_BITS_128(&c, 114);
/* [d 0 0 t4 t3 c r1 r0] = [p8 0 p6 p5 p4 p3 p2 p1 p0] */
secp256k1_u128_accum_mul(&d, a3, a4);
VERIFY_BITS_128(&d, 114);
/* [d 0 0 t4 t3 c r1 r0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */
secp256k1_u128_accum_mul(&c, R, secp256k1_u128_to_u64(&d)); secp256k1_u128_rshift(&d, 64);
VERIFY_BITS_128(&c, 115);
VERIFY_BITS_128(&d, 50);
/* [(d<<12) 0 0 0 t4 t3 c r1 r0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */
r[2] = secp256k1_u128_to_u64(&c) & M; secp256k1_u128_rshift(&c, 52);
VERIFY_BITS(r[2], 52);
VERIFY_BITS_128(&c, 63);
/* [(d<<12) 0 0 0 t4 t3+c r2 r1 r0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */
secp256k1_u128_accum_mul(&c, R << 12, secp256k1_u128_to_u64(&d));
secp256k1_u128_accum_u64(&c, t3);
VERIFY_BITS_128(&c, 100);
/* [t4 c r2 r1 r0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */
r[3] = secp256k1_u128_to_u64(&c) & M; secp256k1_u128_rshift(&c, 52);
VERIFY_BITS(r[3], 52);
VERIFY_BITS_128(&c, 48);
/* [t4+c r3 r2 r1 r0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */
r[4] = secp256k1_u128_to_u64(&c) + t4;
VERIFY_BITS(r[4], 49);
/* [r4 r3 r2 r1 r0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */
}
#endif /* SECP256K1_FIELD_INNER5X52_IMPL_H */

View File

@@ -0,0 +1,457 @@
/***********************************************************************
* Copyright (c) 2013, 2014 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_FIELD_IMPL_H
#define SECP256K1_FIELD_IMPL_H
#include "field.h"
#include "util.h"
#if defined(SECP256K1_WIDEMUL_INT128)
#include "field_5x52_impl.h"
#elif defined(SECP256K1_WIDEMUL_INT64)
#include "field_10x26_impl.h"
#else
#error "Please select wide multiplication implementation"
#endif
SECP256K1_INLINE static void secp256k1_fe_clear(secp256k1_fe *a) {
secp256k1_memclear(a, sizeof(secp256k1_fe));
}
SECP256K1_INLINE static int secp256k1_fe_equal(const secp256k1_fe *a, const secp256k1_fe *b) {
secp256k1_fe na;
SECP256K1_FE_VERIFY(a);
SECP256K1_FE_VERIFY(b);
SECP256K1_FE_VERIFY_MAGNITUDE(a, 1);
SECP256K1_FE_VERIFY_MAGNITUDE(b, 31);
secp256k1_fe_negate(&na, a, 1);
secp256k1_fe_add(&na, b);
return secp256k1_fe_normalizes_to_zero(&na);
}
static int secp256k1_fe_sqrt(secp256k1_fe * SECP256K1_RESTRICT r, const secp256k1_fe * SECP256K1_RESTRICT a) {
/** Given that p is congruent to 3 mod 4, we can compute the square root of
* a mod p as the (p+1)/4'th power of a.
*
* As (p+1)/4 is an even number, it will have the same result for a and for
* (-a). Only one of these two numbers actually has a square root however,
* so we test at the end by squaring and comparing to the input.
* Also because (p+1)/4 is an even number, the computed square root is
* itself always a square (a ** ((p+1)/4) is the square of a ** ((p+1)/8)).
*/
secp256k1_fe x2, x3, x6, x9, x11, x22, x44, x88, x176, x220, x223, t1;
int j, ret;
VERIFY_CHECK(r != a);
SECP256K1_FE_VERIFY(a);
SECP256K1_FE_VERIFY_MAGNITUDE(a, 8);
/** The binary representation of (p + 1)/4 has 3 blocks of 1s, with lengths in
* { 2, 22, 223 }. Use an addition chain to calculate 2^n - 1 for each block:
* 1, [2], 3, 6, 9, 11, [22], 44, 88, 176, 220, [223]
*/
secp256k1_fe_sqr(&x2, a);
secp256k1_fe_mul(&x2, &x2, a);
secp256k1_fe_sqr(&x3, &x2);
secp256k1_fe_mul(&x3, &x3, a);
x6 = x3;
for (j=0; j<3; j++) {
secp256k1_fe_sqr(&x6, &x6);
}
secp256k1_fe_mul(&x6, &x6, &x3);
x9 = x6;
for (j=0; j<3; j++) {
secp256k1_fe_sqr(&x9, &x9);
}
secp256k1_fe_mul(&x9, &x9, &x3);
x11 = x9;
for (j=0; j<2; j++) {
secp256k1_fe_sqr(&x11, &x11);
}
secp256k1_fe_mul(&x11, &x11, &x2);
x22 = x11;
for (j=0; j<11; j++) {
secp256k1_fe_sqr(&x22, &x22);
}
secp256k1_fe_mul(&x22, &x22, &x11);
x44 = x22;
for (j=0; j<22; j++) {
secp256k1_fe_sqr(&x44, &x44);
}
secp256k1_fe_mul(&x44, &x44, &x22);
x88 = x44;
for (j=0; j<44; j++) {
secp256k1_fe_sqr(&x88, &x88);
}
secp256k1_fe_mul(&x88, &x88, &x44);
x176 = x88;
for (j=0; j<88; j++) {
secp256k1_fe_sqr(&x176, &x176);
}
secp256k1_fe_mul(&x176, &x176, &x88);
x220 = x176;
for (j=0; j<44; j++) {
secp256k1_fe_sqr(&x220, &x220);
}
secp256k1_fe_mul(&x220, &x220, &x44);
x223 = x220;
for (j=0; j<3; j++) {
secp256k1_fe_sqr(&x223, &x223);
}
secp256k1_fe_mul(&x223, &x223, &x3);
/* The final result is then assembled using a sliding window over the blocks. */
t1 = x223;
for (j=0; j<23; j++) {
secp256k1_fe_sqr(&t1, &t1);
}
secp256k1_fe_mul(&t1, &t1, &x22);
for (j=0; j<6; j++) {
secp256k1_fe_sqr(&t1, &t1);
}
secp256k1_fe_mul(&t1, &t1, &x2);
secp256k1_fe_sqr(&t1, &t1);
secp256k1_fe_sqr(r, &t1);
/* Check that a square root was actually calculated */
secp256k1_fe_sqr(&t1, r);
ret = secp256k1_fe_equal(&t1, a);
#ifdef VERIFY
if (!ret) {
secp256k1_fe_negate(&t1, &t1, 1);
secp256k1_fe_normalize_var(&t1);
VERIFY_CHECK(secp256k1_fe_equal(&t1, a));
}
#endif
return ret;
}
#ifndef VERIFY
static void secp256k1_fe_verify(const secp256k1_fe *a) { (void)a; }
static void secp256k1_fe_verify_magnitude(const secp256k1_fe *a, int m) { (void)a; (void)m; }
#else
static void secp256k1_fe_impl_verify(const secp256k1_fe *a);
static void secp256k1_fe_verify(const secp256k1_fe *a) {
/* Magnitude between 0 and 32. */
SECP256K1_FE_VERIFY_MAGNITUDE(a, 32);
/* Normalized is 0 or 1. */
VERIFY_CHECK((a->normalized == 0) || (a->normalized == 1));
/* If normalized, magnitude must be 0 or 1. */
if (a->normalized) SECP256K1_FE_VERIFY_MAGNITUDE(a, 1);
/* Invoke implementation-specific checks. */
secp256k1_fe_impl_verify(a);
}
static void secp256k1_fe_verify_magnitude(const secp256k1_fe *a, int m) {
VERIFY_CHECK(m >= 0);
VERIFY_CHECK(m <= 32);
VERIFY_CHECK(a->magnitude <= m);
}
static void secp256k1_fe_impl_normalize(secp256k1_fe *r);
SECP256K1_INLINE static void secp256k1_fe_normalize(secp256k1_fe *r) {
SECP256K1_FE_VERIFY(r);
secp256k1_fe_impl_normalize(r);
r->magnitude = 1;
r->normalized = 1;
SECP256K1_FE_VERIFY(r);
}
static void secp256k1_fe_impl_normalize_weak(secp256k1_fe *r);
SECP256K1_INLINE static void secp256k1_fe_normalize_weak(secp256k1_fe *r) {
SECP256K1_FE_VERIFY(r);
secp256k1_fe_impl_normalize_weak(r);
r->magnitude = 1;
SECP256K1_FE_VERIFY(r);
}
static void secp256k1_fe_impl_normalize_var(secp256k1_fe *r);
SECP256K1_INLINE static void secp256k1_fe_normalize_var(secp256k1_fe *r) {
SECP256K1_FE_VERIFY(r);
secp256k1_fe_impl_normalize_var(r);
r->magnitude = 1;
r->normalized = 1;
SECP256K1_FE_VERIFY(r);
}
static int secp256k1_fe_impl_normalizes_to_zero(const secp256k1_fe *r);
SECP256K1_INLINE static int secp256k1_fe_normalizes_to_zero(const secp256k1_fe *r) {
SECP256K1_FE_VERIFY(r);
return secp256k1_fe_impl_normalizes_to_zero(r);
}
static int secp256k1_fe_impl_normalizes_to_zero_var(const secp256k1_fe *r);
SECP256K1_INLINE static int secp256k1_fe_normalizes_to_zero_var(const secp256k1_fe *r) {
SECP256K1_FE_VERIFY(r);
return secp256k1_fe_impl_normalizes_to_zero_var(r);
}
static void secp256k1_fe_impl_set_int(secp256k1_fe *r, int a);
SECP256K1_INLINE static void secp256k1_fe_set_int(secp256k1_fe *r, int a) {
VERIFY_CHECK(0 <= a && a <= 0x7FFF);
secp256k1_fe_impl_set_int(r, a);
r->magnitude = (a != 0);
r->normalized = 1;
SECP256K1_FE_VERIFY(r);
}
static void secp256k1_fe_impl_add_int(secp256k1_fe *r, int a);
SECP256K1_INLINE static void secp256k1_fe_add_int(secp256k1_fe *r, int a) {
VERIFY_CHECK(0 <= a && a <= 0x7FFF);
SECP256K1_FE_VERIFY(r);
secp256k1_fe_impl_add_int(r, a);
r->magnitude += 1;
r->normalized = 0;
SECP256K1_FE_VERIFY(r);
}
static int secp256k1_fe_impl_is_zero(const secp256k1_fe *a);
SECP256K1_INLINE static int secp256k1_fe_is_zero(const secp256k1_fe *a) {
SECP256K1_FE_VERIFY(a);
VERIFY_CHECK(a->normalized);
return secp256k1_fe_impl_is_zero(a);
}
static int secp256k1_fe_impl_is_odd(const secp256k1_fe *a);
SECP256K1_INLINE static int secp256k1_fe_is_odd(const secp256k1_fe *a) {
SECP256K1_FE_VERIFY(a);
VERIFY_CHECK(a->normalized);
return secp256k1_fe_impl_is_odd(a);
}
static int secp256k1_fe_impl_cmp_var(const secp256k1_fe *a, const secp256k1_fe *b);
SECP256K1_INLINE static int secp256k1_fe_cmp_var(const secp256k1_fe *a, const secp256k1_fe *b) {
SECP256K1_FE_VERIFY(a);
SECP256K1_FE_VERIFY(b);
VERIFY_CHECK(a->normalized);
VERIFY_CHECK(b->normalized);
return secp256k1_fe_impl_cmp_var(a, b);
}
static void secp256k1_fe_impl_set_b32_mod(secp256k1_fe *r, const unsigned char *a);
SECP256K1_INLINE static void secp256k1_fe_set_b32_mod(secp256k1_fe *r, const unsigned char *a) {
secp256k1_fe_impl_set_b32_mod(r, a);
r->magnitude = 1;
r->normalized = 0;
SECP256K1_FE_VERIFY(r);
}
static int secp256k1_fe_impl_set_b32_limit(secp256k1_fe *r, const unsigned char *a);
SECP256K1_INLINE static int secp256k1_fe_set_b32_limit(secp256k1_fe *r, const unsigned char *a) {
if (secp256k1_fe_impl_set_b32_limit(r, a)) {
r->magnitude = 1;
r->normalized = 1;
SECP256K1_FE_VERIFY(r);
return 1;
} else {
/* Mark the output field element as invalid. */
r->magnitude = -1;
return 0;
}
}
static void secp256k1_fe_impl_get_b32(unsigned char *r, const secp256k1_fe *a);
SECP256K1_INLINE static void secp256k1_fe_get_b32(unsigned char *r, const secp256k1_fe *a) {
SECP256K1_FE_VERIFY(a);
VERIFY_CHECK(a->normalized);
secp256k1_fe_impl_get_b32(r, a);
}
static void secp256k1_fe_impl_negate_unchecked(secp256k1_fe *r, const secp256k1_fe *a, int m);
SECP256K1_INLINE static void secp256k1_fe_negate_unchecked(secp256k1_fe *r, const secp256k1_fe *a, int m) {
SECP256K1_FE_VERIFY(a);
VERIFY_CHECK(m >= 0 && m <= 31);
SECP256K1_FE_VERIFY_MAGNITUDE(a, m);
secp256k1_fe_impl_negate_unchecked(r, a, m);
r->magnitude = m + 1;
r->normalized = 0;
SECP256K1_FE_VERIFY(r);
}
static void secp256k1_fe_impl_mul_int_unchecked(secp256k1_fe *r, int a);
SECP256K1_INLINE static void secp256k1_fe_mul_int_unchecked(secp256k1_fe *r, int a) {
SECP256K1_FE_VERIFY(r);
VERIFY_CHECK(a >= 0 && a <= 32);
VERIFY_CHECK(a*r->magnitude <= 32);
secp256k1_fe_impl_mul_int_unchecked(r, a);
r->magnitude *= a;
r->normalized = 0;
SECP256K1_FE_VERIFY(r);
}
static void secp256k1_fe_impl_add(secp256k1_fe *r, const secp256k1_fe *a);
SECP256K1_INLINE static void secp256k1_fe_add(secp256k1_fe *r, const secp256k1_fe *a) {
SECP256K1_FE_VERIFY(r);
SECP256K1_FE_VERIFY(a);
VERIFY_CHECK(r->magnitude + a->magnitude <= 32);
secp256k1_fe_impl_add(r, a);
r->magnitude += a->magnitude;
r->normalized = 0;
SECP256K1_FE_VERIFY(r);
}
static void secp256k1_fe_impl_mul(secp256k1_fe *r, const secp256k1_fe *a, const secp256k1_fe * SECP256K1_RESTRICT b);
SECP256K1_INLINE static void secp256k1_fe_mul(secp256k1_fe *r, const secp256k1_fe *a, const secp256k1_fe * SECP256K1_RESTRICT b) {
SECP256K1_FE_VERIFY(a);
SECP256K1_FE_VERIFY(b);
SECP256K1_FE_VERIFY_MAGNITUDE(a, 8);
SECP256K1_FE_VERIFY_MAGNITUDE(b, 8);
VERIFY_CHECK(r != b);
VERIFY_CHECK(a != b);
secp256k1_fe_impl_mul(r, a, b);
r->magnitude = 1;
r->normalized = 0;
SECP256K1_FE_VERIFY(r);
}
static void secp256k1_fe_impl_sqr(secp256k1_fe *r, const secp256k1_fe *a);
SECP256K1_INLINE static void secp256k1_fe_sqr(secp256k1_fe *r, const secp256k1_fe *a) {
SECP256K1_FE_VERIFY(a);
SECP256K1_FE_VERIFY_MAGNITUDE(a, 8);
secp256k1_fe_impl_sqr(r, a);
r->magnitude = 1;
r->normalized = 0;
SECP256K1_FE_VERIFY(r);
}
static void secp256k1_fe_impl_cmov(secp256k1_fe *r, const secp256k1_fe *a, int flag);
SECP256K1_INLINE static void secp256k1_fe_cmov(secp256k1_fe *r, const secp256k1_fe *a, int flag) {
VERIFY_CHECK(flag == 0 || flag == 1);
SECP256K1_FE_VERIFY(a);
SECP256K1_FE_VERIFY(r);
secp256k1_fe_impl_cmov(r, a, flag);
if (a->magnitude > r->magnitude) r->magnitude = a->magnitude;
if (!a->normalized) r->normalized = 0;
SECP256K1_FE_VERIFY(r);
}
static void secp256k1_fe_impl_to_storage(secp256k1_fe_storage *r, const secp256k1_fe *a);
SECP256K1_INLINE static void secp256k1_fe_to_storage(secp256k1_fe_storage *r, const secp256k1_fe *a) {
SECP256K1_FE_VERIFY(a);
VERIFY_CHECK(a->normalized);
secp256k1_fe_impl_to_storage(r, a);
}
static void secp256k1_fe_impl_from_storage(secp256k1_fe *r, const secp256k1_fe_storage *a);
SECP256K1_INLINE static void secp256k1_fe_from_storage(secp256k1_fe *r, const secp256k1_fe_storage *a) {
secp256k1_fe_impl_from_storage(r, a);
r->magnitude = 1;
r->normalized = 1;
SECP256K1_FE_VERIFY(r);
}
static void secp256k1_fe_impl_inv(secp256k1_fe *r, const secp256k1_fe *x);
SECP256K1_INLINE static void secp256k1_fe_inv(secp256k1_fe *r, const secp256k1_fe *x) {
int input_is_zero = secp256k1_fe_normalizes_to_zero(x);
SECP256K1_FE_VERIFY(x);
secp256k1_fe_impl_inv(r, x);
r->magnitude = x->magnitude > 0;
r->normalized = 1;
VERIFY_CHECK(secp256k1_fe_normalizes_to_zero(r) == input_is_zero);
SECP256K1_FE_VERIFY(r);
}
static void secp256k1_fe_impl_inv_var(secp256k1_fe *r, const secp256k1_fe *x);
SECP256K1_INLINE static void secp256k1_fe_inv_var(secp256k1_fe *r, const secp256k1_fe *x) {
int input_is_zero = secp256k1_fe_normalizes_to_zero(x);
SECP256K1_FE_VERIFY(x);
secp256k1_fe_impl_inv_var(r, x);
r->magnitude = x->magnitude > 0;
r->normalized = 1;
VERIFY_CHECK(secp256k1_fe_normalizes_to_zero(r) == input_is_zero);
SECP256K1_FE_VERIFY(r);
}
static int secp256k1_fe_impl_is_square_var(const secp256k1_fe *x);
SECP256K1_INLINE static int secp256k1_fe_is_square_var(const secp256k1_fe *x) {
int ret;
secp256k1_fe tmp = *x, sqrt;
SECP256K1_FE_VERIFY(x);
ret = secp256k1_fe_impl_is_square_var(x);
secp256k1_fe_normalize_weak(&tmp);
VERIFY_CHECK(ret == secp256k1_fe_sqrt(&sqrt, &tmp));
return ret;
}
static void secp256k1_fe_impl_get_bounds(secp256k1_fe* r, int m);
SECP256K1_INLINE static void secp256k1_fe_get_bounds(secp256k1_fe* r, int m) {
VERIFY_CHECK(m >= 0);
VERIFY_CHECK(m <= 32);
secp256k1_fe_impl_get_bounds(r, m);
r->magnitude = m;
r->normalized = (m == 0);
SECP256K1_FE_VERIFY(r);
}
static void secp256k1_fe_impl_half(secp256k1_fe *r);
SECP256K1_INLINE static void secp256k1_fe_half(secp256k1_fe *r) {
SECP256K1_FE_VERIFY(r);
SECP256K1_FE_VERIFY_MAGNITUDE(r, 31);
secp256k1_fe_impl_half(r);
r->magnitude = (r->magnitude >> 1) + 1;
r->normalized = 0;
SECP256K1_FE_VERIFY(r);
}
#endif /* defined(VERIFY) */
#endif /* SECP256K1_FIELD_IMPL_H */

212
libsecp256k1/src/group.h Normal file
View File

@@ -0,0 +1,212 @@
/***********************************************************************
* Copyright (c) 2013, 2014 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_GROUP_H
#define SECP256K1_GROUP_H
#include "field.h"
/** A group element in affine coordinates on the secp256k1 curve,
* or occasionally on an isomorphic curve of the form y^2 = x^3 + 7*t^6.
* Note: For exhaustive test mode, secp256k1 is replaced by a small subgroup of a different curve.
*/
typedef struct {
secp256k1_fe x;
secp256k1_fe y;
int infinity; /* whether this represents the point at infinity */
} secp256k1_ge;
#define SECP256K1_GE_CONST(a, b, c, d, e, f, g, h, i, j, k, l, m, n, o, p) {SECP256K1_FE_CONST((a),(b),(c),(d),(e),(f),(g),(h)), SECP256K1_FE_CONST((i),(j),(k),(l),(m),(n),(o),(p)), 0}
#define SECP256K1_GE_CONST_INFINITY {SECP256K1_FE_CONST(0, 0, 0, 0, 0, 0, 0, 0), SECP256K1_FE_CONST(0, 0, 0, 0, 0, 0, 0, 0), 1}
/** A group element of the secp256k1 curve, in jacobian coordinates.
* Note: For exhastive test mode, secp256k1 is replaced by a small subgroup of a different curve.
*/
typedef struct {
secp256k1_fe x; /* actual X: x/z^2 */
secp256k1_fe y; /* actual Y: y/z^3 */
secp256k1_fe z;
int infinity; /* whether this represents the point at infinity */
} secp256k1_gej;
#define SECP256K1_GEJ_CONST(a, b, c, d, e, f, g, h, i, j, k, l, m, n, o, p) {SECP256K1_FE_CONST((a),(b),(c),(d),(e),(f),(g),(h)), SECP256K1_FE_CONST((i),(j),(k),(l),(m),(n),(o),(p)), SECP256K1_FE_CONST(0, 0, 0, 0, 0, 0, 0, 1), 0}
#define SECP256K1_GEJ_CONST_INFINITY {SECP256K1_FE_CONST(0, 0, 0, 0, 0, 0, 0, 0), SECP256K1_FE_CONST(0, 0, 0, 0, 0, 0, 0, 0), SECP256K1_FE_CONST(0, 0, 0, 0, 0, 0, 0, 0), 1}
typedef struct {
secp256k1_fe_storage x;
secp256k1_fe_storage y;
} secp256k1_ge_storage;
#define SECP256K1_GE_STORAGE_CONST(a, b, c, d, e, f, g, h, i, j, k, l, m, n, o, p) {SECP256K1_FE_STORAGE_CONST((a),(b),(c),(d),(e),(f),(g),(h)), SECP256K1_FE_STORAGE_CONST((i),(j),(k),(l),(m),(n),(o),(p))}
#define SECP256K1_GE_STORAGE_CONST_GET(t) SECP256K1_FE_STORAGE_CONST_GET(t.x), SECP256K1_FE_STORAGE_CONST_GET(t.y)
/** Maximum allowed magnitudes for group element coordinates
* in affine (x, y) and jacobian (x, y, z) representation. */
#define SECP256K1_GE_X_MAGNITUDE_MAX 4
#define SECP256K1_GE_Y_MAGNITUDE_MAX 3
#define SECP256K1_GEJ_X_MAGNITUDE_MAX 4
#define SECP256K1_GEJ_Y_MAGNITUDE_MAX 4
#define SECP256K1_GEJ_Z_MAGNITUDE_MAX 1
/** Set a group element equal to the point with given X and Y coordinates */
static void secp256k1_ge_set_xy(secp256k1_ge *r, const secp256k1_fe *x, const secp256k1_fe *y);
/** Set a group element (affine) equal to the point with the given X coordinate, and given oddness
* for Y. Return value indicates whether the result is valid. */
static int secp256k1_ge_set_xo_var(secp256k1_ge *r, const secp256k1_fe *x, int odd);
/** Determine whether x is a valid X coordinate on the curve. */
static int secp256k1_ge_x_on_curve_var(const secp256k1_fe *x);
/** Determine whether fraction xn/xd is a valid X coordinate on the curve (xd != 0). */
static int secp256k1_ge_x_frac_on_curve_var(const secp256k1_fe *xn, const secp256k1_fe *xd);
/** Check whether a group element is the point at infinity. */
static int secp256k1_ge_is_infinity(const secp256k1_ge *a);
/** Check whether a group element is valid (i.e., on the curve). */
static int secp256k1_ge_is_valid_var(const secp256k1_ge *a);
/** Set r equal to the inverse of a (i.e., mirrored around the X axis) */
static void secp256k1_ge_neg(secp256k1_ge *r, const secp256k1_ge *a);
/** Set a group element equal to another which is given in jacobian coordinates. Constant time. */
static void secp256k1_ge_set_gej(secp256k1_ge *r, secp256k1_gej *a);
/** Set a group element equal to another which is given in jacobian coordinates. */
static void secp256k1_ge_set_gej_var(secp256k1_ge *r, secp256k1_gej *a);
/** Set a batch of group elements equal to the inputs given in jacobian coordinates */
static void secp256k1_ge_set_all_gej_var(secp256k1_ge *r, const secp256k1_gej *a, size_t len);
/** Bring a batch of inputs to the same global z "denominator", based on ratios between
* (omitted) z coordinates of adjacent elements.
*
* Although the elements a[i] are _ge rather than _gej, they actually represent elements
* in Jacobian coordinates with their z coordinates omitted.
*
* Using the notation z(b) to represent the omitted z coordinate of b, the array zr of
* z coordinate ratios must satisfy zr[i] == z(a[i]) / z(a[i-1]) for 0 < 'i' < len.
* The zr[0] value is unused.
*
* This function adjusts the coordinates of 'a' in place so that for all 'i', z(a[i]) == z(a[len-1]).
* In other words, the initial value of z(a[len-1]) becomes the global z "denominator". Only the
* a[i].x and a[i].y coordinates are explicitly modified; the adjustment of the omitted z coordinate is
* implicit.
*
* The coordinates of the final element a[len-1] are not changed.
*/
static void secp256k1_ge_table_set_globalz(size_t len, secp256k1_ge *a, const secp256k1_fe *zr);
/** Check two group elements (affine) for equality in variable time. */
static int secp256k1_ge_eq_var(const secp256k1_ge *a, const secp256k1_ge *b);
/** Set a group element (affine) equal to the point at infinity. */
static void secp256k1_ge_set_infinity(secp256k1_ge *r);
/** Set a group element (jacobian) equal to the point at infinity. */
static void secp256k1_gej_set_infinity(secp256k1_gej *r);
/** Set a group element (jacobian) equal to another which is given in affine coordinates. */
static void secp256k1_gej_set_ge(secp256k1_gej *r, const secp256k1_ge *a);
/** Check two group elements (jacobian) for equality in variable time. */
static int secp256k1_gej_eq_var(const secp256k1_gej *a, const secp256k1_gej *b);
/** Check two group elements (jacobian and affine) for equality in variable time. */
static int secp256k1_gej_eq_ge_var(const secp256k1_gej *a, const secp256k1_ge *b);
/** Compare the X coordinate of a group element (jacobian).
* The magnitude of the group element's X coordinate must not exceed 31. */
static int secp256k1_gej_eq_x_var(const secp256k1_fe *x, const secp256k1_gej *a);
/** Set r equal to the inverse of a (i.e., mirrored around the X axis) */
static void secp256k1_gej_neg(secp256k1_gej *r, const secp256k1_gej *a);
/** Check whether a group element is the point at infinity. */
static int secp256k1_gej_is_infinity(const secp256k1_gej *a);
/** Set r equal to the double of a. Constant time. */
static void secp256k1_gej_double(secp256k1_gej *r, const secp256k1_gej *a);
/** Set r equal to the double of a. If rzr is not-NULL this sets *rzr such that r->z == a->z * *rzr (where infinity means an implicit z = 0). */
static void secp256k1_gej_double_var(secp256k1_gej *r, const secp256k1_gej *a, secp256k1_fe *rzr);
/** Set r equal to the sum of a and b. If rzr is non-NULL this sets *rzr such that r->z == a->z * *rzr (a cannot be infinity in that case). */
static void secp256k1_gej_add_var(secp256k1_gej *r, const secp256k1_gej *a, const secp256k1_gej *b, secp256k1_fe *rzr);
/** Set r equal to the sum of a and b (with b given in affine coordinates, and not infinity). */
static void secp256k1_gej_add_ge(secp256k1_gej *r, const secp256k1_gej *a, const secp256k1_ge *b);
/** Set r equal to the sum of a and b (with b given in affine coordinates). This is more efficient
than secp256k1_gej_add_var. It is identical to secp256k1_gej_add_ge but without constant-time
guarantee, and b is allowed to be infinity. If rzr is non-NULL this sets *rzr such that r->z == a->z * *rzr (a cannot be infinity in that case). */
static void secp256k1_gej_add_ge_var(secp256k1_gej *r, const secp256k1_gej *a, const secp256k1_ge *b, secp256k1_fe *rzr);
/** Set r equal to the sum of a and b (with the inverse of b's Z coordinate passed as bzinv). */
static void secp256k1_gej_add_zinv_var(secp256k1_gej *r, const secp256k1_gej *a, const secp256k1_ge *b, const secp256k1_fe *bzinv);
/** Set r to be equal to lambda times a, where lambda is chosen in a way such that this is very fast. */
static void secp256k1_ge_mul_lambda(secp256k1_ge *r, const secp256k1_ge *a);
/** Clear a secp256k1_gej to prevent leaking sensitive information. */
static void secp256k1_gej_clear(secp256k1_gej *r);
/** Clear a secp256k1_ge to prevent leaking sensitive information. */
static void secp256k1_ge_clear(secp256k1_ge *r);
/** Convert a group element to the storage type. */
static void secp256k1_ge_to_storage(secp256k1_ge_storage *r, const secp256k1_ge *a);
/** Convert a group element back from the storage type. */
static void secp256k1_ge_from_storage(secp256k1_ge *r, const secp256k1_ge_storage *a);
/** If flag is true, set *r equal to *a; otherwise leave it. Constant-time. Both *r and *a must be initialized.*/
static void secp256k1_gej_cmov(secp256k1_gej *r, const secp256k1_gej *a, int flag);
/** If flag is true, set *r equal to *a; otherwise leave it. Constant-time. Both *r and *a must be initialized.*/
static void secp256k1_ge_storage_cmov(secp256k1_ge_storage *r, const secp256k1_ge_storage *a, int flag);
/** Rescale a jacobian point by b which must be non-zero. Constant-time. */
static void secp256k1_gej_rescale(secp256k1_gej *r, const secp256k1_fe *b);
/** Convert a group element that is not infinity to a 64-byte array. The output
* array is platform-dependent. */
static void secp256k1_ge_to_bytes(unsigned char *buf, const secp256k1_ge *a);
/** Convert a 64-byte array into group element. This function assumes that the
* provided buffer correctly encodes a group element. */
static void secp256k1_ge_from_bytes(secp256k1_ge *r, const unsigned char *buf);
/** Convert a group element (that is allowed to be infinity) to a 64-byte
* array. The output array is platform-dependent. */
static void secp256k1_ge_to_bytes_ext(unsigned char *data, const secp256k1_ge *ge);
/** Convert a 64-byte array into a group element. This function assumes that the
* provided buffer is the output of secp256k1_ge_to_bytes_ext. */
static void secp256k1_ge_from_bytes_ext(secp256k1_ge *ge, const unsigned char *data);
/** Determine if a point (which is assumed to be on the curve) is in the correct (sub)group of the curve.
*
* In normal mode, the used group is secp256k1, which has cofactor=1 meaning that every point on the curve is in the
* group, and this function returns always true.
*
* When compiling in exhaustive test mode, a slightly different curve equation is used, leading to a group with a
* (very) small subgroup, and that subgroup is what is used for all cryptographic operations. In that mode, this
* function checks whether a point that is on the curve is in fact also in that subgroup.
*/
static int secp256k1_ge_is_in_correct_subgroup(const secp256k1_ge* ge);
/** Check invariants on an affine group element (no-op unless VERIFY is enabled). */
static void secp256k1_ge_verify(const secp256k1_ge *a);
#define SECP256K1_GE_VERIFY(a) secp256k1_ge_verify(a)
/** Check invariants on a Jacobian group element (no-op unless VERIFY is enabled). */
static void secp256k1_gej_verify(const secp256k1_gej *a);
#define SECP256K1_GEJ_VERIFY(a) secp256k1_gej_verify(a)
#endif /* SECP256K1_GROUP_H */

View File

@@ -0,0 +1,974 @@
/***********************************************************************
* Copyright (c) 2013, 2014 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_GROUP_IMPL_H
#define SECP256K1_GROUP_IMPL_H
#include <string.h>
#include "field.h"
#include "group.h"
#include "util.h"
/* Begin of section generated by sage/gen_exhaustive_groups.sage. */
#define SECP256K1_G_ORDER_7 SECP256K1_GE_CONST(\
0x66625d13, 0x317ffe44, 0x63d32cff, 0x1ca02b9b,\
0xe5c6d070, 0x50b4b05e, 0x81cc30db, 0xf5166f0a,\
0x1e60e897, 0xa7c00c7c, 0x2df53eb6, 0x98274ff4,\
0x64252f42, 0x8ca44e17, 0x3b25418c, 0xff4ab0cf\
)
#define SECP256K1_G_ORDER_13 SECP256K1_GE_CONST(\
0xa2482ff8, 0x4bf34edf, 0xa51262fd, 0xe57921db,\
0xe0dd2cb7, 0xa5914790, 0xbc71631f, 0xc09704fb,\
0x942536cb, 0xa3e49492, 0x3a701cc3, 0xee3e443f,\
0xdf182aa9, 0x15b8aa6a, 0x166d3b19, 0xba84b045\
)
#define SECP256K1_G_ORDER_199 SECP256K1_GE_CONST(\
0x7fb07b5c, 0xd07c3bda, 0x553902e2, 0x7a87ea2c,\
0x35108a7f, 0x051f41e5, 0xb76abad5, 0x1f2703ad,\
0x0a251539, 0x5b4c4438, 0x952a634f, 0xac10dd4d,\
0x6d6f4745, 0x98990c27, 0x3a4f3116, 0xd32ff969\
)
/** Generator for secp256k1, value 'g' defined in
* "Standards for Efficient Cryptography" (SEC2) 2.7.1.
*/
#define SECP256K1_G SECP256K1_GE_CONST(\
0x79be667e, 0xf9dcbbac, 0x55a06295, 0xce870b07,\
0x029bfcdb, 0x2dce28d9, 0x59f2815b, 0x16f81798,\
0x483ada77, 0x26a3c465, 0x5da4fbfc, 0x0e1108a8,\
0xfd17b448, 0xa6855419, 0x9c47d08f, 0xfb10d4b8\
)
/* These exhaustive group test orders and generators are chosen such that:
* - The field size is equal to that of secp256k1, so field code is the same.
* - The curve equation is of the form y^2=x^3+B for some small constant B.
* - The subgroup has a generator 2*P, where P.x is as small as possible.
* - The subgroup has size less than 1000 to permit exhaustive testing.
* - The subgroup admits an endomorphism of the form lambda*(x,y) == (beta*x,y).
*/
#if defined(EXHAUSTIVE_TEST_ORDER)
# if EXHAUSTIVE_TEST_ORDER == 7
static const secp256k1_ge secp256k1_ge_const_g = SECP256K1_G_ORDER_7;
#define SECP256K1_B 6
# elif EXHAUSTIVE_TEST_ORDER == 13
static const secp256k1_ge secp256k1_ge_const_g = SECP256K1_G_ORDER_13;
#define SECP256K1_B 2
# elif EXHAUSTIVE_TEST_ORDER == 199
static const secp256k1_ge secp256k1_ge_const_g = SECP256K1_G_ORDER_199;
#define SECP256K1_B 4
# else
# error No known generator for the specified exhaustive test group order.
# endif
#else
static const secp256k1_ge secp256k1_ge_const_g = SECP256K1_G;
#define SECP256K1_B 7
#endif
/* End of section generated by sage/gen_exhaustive_groups.sage. */
static void secp256k1_ge_verify(const secp256k1_ge *a) {
SECP256K1_FE_VERIFY(&a->x);
SECP256K1_FE_VERIFY(&a->y);
SECP256K1_FE_VERIFY_MAGNITUDE(&a->x, SECP256K1_GE_X_MAGNITUDE_MAX);
SECP256K1_FE_VERIFY_MAGNITUDE(&a->y, SECP256K1_GE_Y_MAGNITUDE_MAX);
VERIFY_CHECK(a->infinity == 0 || a->infinity == 1);
(void)a;
}
static void secp256k1_gej_verify(const secp256k1_gej *a) {
SECP256K1_FE_VERIFY(&a->x);
SECP256K1_FE_VERIFY(&a->y);
SECP256K1_FE_VERIFY(&a->z);
SECP256K1_FE_VERIFY_MAGNITUDE(&a->x, SECP256K1_GEJ_X_MAGNITUDE_MAX);
SECP256K1_FE_VERIFY_MAGNITUDE(&a->y, SECP256K1_GEJ_Y_MAGNITUDE_MAX);
SECP256K1_FE_VERIFY_MAGNITUDE(&a->z, SECP256K1_GEJ_Z_MAGNITUDE_MAX);
VERIFY_CHECK(a->infinity == 0 || a->infinity == 1);
(void)a;
}
/* Set r to the affine coordinates of Jacobian point (a.x, a.y, 1/zi). */
static void secp256k1_ge_set_gej_zinv(secp256k1_ge *r, const secp256k1_gej *a, const secp256k1_fe *zi) {
secp256k1_fe zi2;
secp256k1_fe zi3;
SECP256K1_GEJ_VERIFY(a);
SECP256K1_FE_VERIFY(zi);
VERIFY_CHECK(!a->infinity);
secp256k1_fe_sqr(&zi2, zi);
secp256k1_fe_mul(&zi3, &zi2, zi);
secp256k1_fe_mul(&r->x, &a->x, &zi2);
secp256k1_fe_mul(&r->y, &a->y, &zi3);
r->infinity = a->infinity;
SECP256K1_GE_VERIFY(r);
}
/* Set r to the affine coordinates of Jacobian point (a.x, a.y, 1/zi). */
static void secp256k1_ge_set_ge_zinv(secp256k1_ge *r, const secp256k1_ge *a, const secp256k1_fe *zi) {
secp256k1_fe zi2;
secp256k1_fe zi3;
SECP256K1_GE_VERIFY(a);
SECP256K1_FE_VERIFY(zi);
VERIFY_CHECK(!a->infinity);
secp256k1_fe_sqr(&zi2, zi);
secp256k1_fe_mul(&zi3, &zi2, zi);
secp256k1_fe_mul(&r->x, &a->x, &zi2);
secp256k1_fe_mul(&r->y, &a->y, &zi3);
r->infinity = a->infinity;
SECP256K1_GE_VERIFY(r);
}
static void secp256k1_ge_set_xy(secp256k1_ge *r, const secp256k1_fe *x, const secp256k1_fe *y) {
SECP256K1_FE_VERIFY(x);
SECP256K1_FE_VERIFY(y);
r->infinity = 0;
r->x = *x;
r->y = *y;
SECP256K1_GE_VERIFY(r);
}
static int secp256k1_ge_is_infinity(const secp256k1_ge *a) {
SECP256K1_GE_VERIFY(a);
return a->infinity;
}
static void secp256k1_ge_neg(secp256k1_ge *r, const secp256k1_ge *a) {
SECP256K1_GE_VERIFY(a);
*r = *a;
secp256k1_fe_normalize_weak(&r->y);
secp256k1_fe_negate(&r->y, &r->y, 1);
SECP256K1_GE_VERIFY(r);
}
static void secp256k1_ge_set_gej(secp256k1_ge *r, secp256k1_gej *a) {
secp256k1_fe z2, z3;
SECP256K1_GEJ_VERIFY(a);
r->infinity = a->infinity;
secp256k1_fe_inv(&a->z, &a->z);
secp256k1_fe_sqr(&z2, &a->z);
secp256k1_fe_mul(&z3, &a->z, &z2);
secp256k1_fe_mul(&a->x, &a->x, &z2);
secp256k1_fe_mul(&a->y, &a->y, &z3);
secp256k1_fe_set_int(&a->z, 1);
r->x = a->x;
r->y = a->y;
SECP256K1_GEJ_VERIFY(a);
SECP256K1_GE_VERIFY(r);
}
static void secp256k1_ge_set_gej_var(secp256k1_ge *r, secp256k1_gej *a) {
secp256k1_fe z2, z3;
SECP256K1_GEJ_VERIFY(a);
if (secp256k1_gej_is_infinity(a)) {
secp256k1_ge_set_infinity(r);
return;
}
r->infinity = 0;
secp256k1_fe_inv_var(&a->z, &a->z);
secp256k1_fe_sqr(&z2, &a->z);
secp256k1_fe_mul(&z3, &a->z, &z2);
secp256k1_fe_mul(&a->x, &a->x, &z2);
secp256k1_fe_mul(&a->y, &a->y, &z3);
secp256k1_fe_set_int(&a->z, 1);
secp256k1_ge_set_xy(r, &a->x, &a->y);
SECP256K1_GEJ_VERIFY(a);
SECP256K1_GE_VERIFY(r);
}
static void secp256k1_ge_set_all_gej_var(secp256k1_ge *r, const secp256k1_gej *a, size_t len) {
secp256k1_fe u;
size_t i;
size_t last_i = SIZE_MAX;
#ifdef VERIFY
for (i = 0; i < len; i++) {
SECP256K1_GEJ_VERIFY(&a[i]);
}
#endif
for (i = 0; i < len; i++) {
if (a[i].infinity) {
secp256k1_ge_set_infinity(&r[i]);
} else {
/* Use destination's x coordinates as scratch space */
if (last_i == SIZE_MAX) {
r[i].x = a[i].z;
} else {
secp256k1_fe_mul(&r[i].x, &r[last_i].x, &a[i].z);
}
last_i = i;
}
}
if (last_i == SIZE_MAX) {
return;
}
secp256k1_fe_inv_var(&u, &r[last_i].x);
i = last_i;
while (i > 0) {
i--;
if (!a[i].infinity) {
secp256k1_fe_mul(&r[last_i].x, &r[i].x, &u);
secp256k1_fe_mul(&u, &u, &a[last_i].z);
last_i = i;
}
}
VERIFY_CHECK(!a[last_i].infinity);
r[last_i].x = u;
for (i = 0; i < len; i++) {
if (!a[i].infinity) {
secp256k1_ge_set_gej_zinv(&r[i], &a[i], &r[i].x);
}
}
#ifdef VERIFY
for (i = 0; i < len; i++) {
SECP256K1_GE_VERIFY(&r[i]);
}
#endif
}
static void secp256k1_ge_table_set_globalz(size_t len, secp256k1_ge *a, const secp256k1_fe *zr) {
size_t i;
secp256k1_fe zs;
#ifdef VERIFY
for (i = 0; i < len; i++) {
SECP256K1_GE_VERIFY(&a[i]);
SECP256K1_FE_VERIFY(&zr[i]);
}
#endif
if (len > 0) {
i = len - 1;
/* Ensure all y values are in weak normal form for fast negation of points */
secp256k1_fe_normalize_weak(&a[i].y);
zs = zr[i];
/* Work our way backwards, using the z-ratios to scale the x/y values. */
while (i > 0) {
if (i != len - 1) {
secp256k1_fe_mul(&zs, &zs, &zr[i]);
}
i--;
secp256k1_ge_set_ge_zinv(&a[i], &a[i], &zs);
}
}
#ifdef VERIFY
for (i = 0; i < len; i++) {
SECP256K1_GE_VERIFY(&a[i]);
}
#endif
}
static void secp256k1_gej_set_infinity(secp256k1_gej *r) {
r->infinity = 1;
secp256k1_fe_set_int(&r->x, 0);
secp256k1_fe_set_int(&r->y, 0);
secp256k1_fe_set_int(&r->z, 0);
SECP256K1_GEJ_VERIFY(r);
}
static void secp256k1_ge_set_infinity(secp256k1_ge *r) {
r->infinity = 1;
secp256k1_fe_set_int(&r->x, 0);
secp256k1_fe_set_int(&r->y, 0);
SECP256K1_GE_VERIFY(r);
}
static void secp256k1_gej_clear(secp256k1_gej *r) {
secp256k1_memclear(r, sizeof(secp256k1_gej));
}
static void secp256k1_ge_clear(secp256k1_ge *r) {
secp256k1_memclear(r, sizeof(secp256k1_ge));
}
static int secp256k1_ge_set_xo_var(secp256k1_ge *r, const secp256k1_fe *x, int odd) {
secp256k1_fe x2, x3;
int ret;
SECP256K1_FE_VERIFY(x);
r->x = *x;
secp256k1_fe_sqr(&x2, x);
secp256k1_fe_mul(&x3, x, &x2);
r->infinity = 0;
secp256k1_fe_add_int(&x3, SECP256K1_B);
ret = secp256k1_fe_sqrt(&r->y, &x3);
secp256k1_fe_normalize_var(&r->y);
if (secp256k1_fe_is_odd(&r->y) != odd) {
secp256k1_fe_negate(&r->y, &r->y, 1);
}
SECP256K1_GE_VERIFY(r);
return ret;
}
static void secp256k1_gej_set_ge(secp256k1_gej *r, const secp256k1_ge *a) {
SECP256K1_GE_VERIFY(a);
r->infinity = a->infinity;
r->x = a->x;
r->y = a->y;
secp256k1_fe_set_int(&r->z, 1);
SECP256K1_GEJ_VERIFY(r);
}
static int secp256k1_gej_eq_var(const secp256k1_gej *a, const secp256k1_gej *b) {
secp256k1_gej tmp;
SECP256K1_GEJ_VERIFY(b);
SECP256K1_GEJ_VERIFY(a);
secp256k1_gej_neg(&tmp, a);
secp256k1_gej_add_var(&tmp, &tmp, b, NULL);
return secp256k1_gej_is_infinity(&tmp);
}
static int secp256k1_gej_eq_ge_var(const secp256k1_gej *a, const secp256k1_ge *b) {
secp256k1_gej tmp;
SECP256K1_GEJ_VERIFY(a);
SECP256K1_GE_VERIFY(b);
secp256k1_gej_neg(&tmp, a);
secp256k1_gej_add_ge_var(&tmp, &tmp, b, NULL);
return secp256k1_gej_is_infinity(&tmp);
}
static int secp256k1_ge_eq_var(const secp256k1_ge *a, const secp256k1_ge *b) {
secp256k1_fe tmp;
SECP256K1_GE_VERIFY(a);
SECP256K1_GE_VERIFY(b);
if (a->infinity != b->infinity) return 0;
if (a->infinity) return 1;
tmp = a->x;
secp256k1_fe_normalize_weak(&tmp);
if (!secp256k1_fe_equal(&tmp, &b->x)) return 0;
tmp = a->y;
secp256k1_fe_normalize_weak(&tmp);
if (!secp256k1_fe_equal(&tmp, &b->y)) return 0;
return 1;
}
static int secp256k1_gej_eq_x_var(const secp256k1_fe *x, const secp256k1_gej *a) {
secp256k1_fe r;
SECP256K1_FE_VERIFY(x);
SECP256K1_GEJ_VERIFY(a);
VERIFY_CHECK(!a->infinity);
secp256k1_fe_sqr(&r, &a->z); secp256k1_fe_mul(&r, &r, x);
return secp256k1_fe_equal(&r, &a->x);
}
static void secp256k1_gej_neg(secp256k1_gej *r, const secp256k1_gej *a) {
SECP256K1_GEJ_VERIFY(a);
r->infinity = a->infinity;
r->x = a->x;
r->y = a->y;
r->z = a->z;
secp256k1_fe_normalize_weak(&r->y);
secp256k1_fe_negate(&r->y, &r->y, 1);
SECP256K1_GEJ_VERIFY(r);
}
static int secp256k1_gej_is_infinity(const secp256k1_gej *a) {
SECP256K1_GEJ_VERIFY(a);
return a->infinity;
}
static int secp256k1_ge_is_valid_var(const secp256k1_ge *a) {
secp256k1_fe y2, x3;
SECP256K1_GE_VERIFY(a);
if (a->infinity) {
return 0;
}
/* y^2 = x^3 + 7 */
secp256k1_fe_sqr(&y2, &a->y);
secp256k1_fe_sqr(&x3, &a->x); secp256k1_fe_mul(&x3, &x3, &a->x);
secp256k1_fe_add_int(&x3, SECP256K1_B);
return secp256k1_fe_equal(&y2, &x3);
}
static SECP256K1_INLINE void secp256k1_gej_double(secp256k1_gej *r, const secp256k1_gej *a) {
/* Operations: 3 mul, 4 sqr, 8 add/half/mul_int/negate */
secp256k1_fe l, s, t;
SECP256K1_GEJ_VERIFY(a);
r->infinity = a->infinity;
/* Formula used:
* L = (3/2) * X1^2
* S = Y1^2
* T = -X1*S
* X3 = L^2 + 2*T
* Y3 = -(L*(X3 + T) + S^2)
* Z3 = Y1*Z1
*/
secp256k1_fe_mul(&r->z, &a->z, &a->y); /* Z3 = Y1*Z1 (1) */
secp256k1_fe_sqr(&s, &a->y); /* S = Y1^2 (1) */
secp256k1_fe_sqr(&l, &a->x); /* L = X1^2 (1) */
secp256k1_fe_mul_int(&l, 3); /* L = 3*X1^2 (3) */
secp256k1_fe_half(&l); /* L = 3/2*X1^2 (2) */
secp256k1_fe_negate(&t, &s, 1); /* T = -S (2) */
secp256k1_fe_mul(&t, &t, &a->x); /* T = -X1*S (1) */
secp256k1_fe_sqr(&r->x, &l); /* X3 = L^2 (1) */
secp256k1_fe_add(&r->x, &t); /* X3 = L^2 + T (2) */
secp256k1_fe_add(&r->x, &t); /* X3 = L^2 + 2*T (3) */
secp256k1_fe_sqr(&s, &s); /* S' = S^2 (1) */
secp256k1_fe_add(&t, &r->x); /* T' = X3 + T (4) */
secp256k1_fe_mul(&r->y, &t, &l); /* Y3 = L*(X3 + T) (1) */
secp256k1_fe_add(&r->y, &s); /* Y3 = L*(X3 + T) + S^2 (2) */
secp256k1_fe_negate(&r->y, &r->y, 2); /* Y3 = -(L*(X3 + T) + S^2) (3) */
SECP256K1_GEJ_VERIFY(r);
}
static void secp256k1_gej_double_var(secp256k1_gej *r, const secp256k1_gej *a, secp256k1_fe *rzr) {
SECP256K1_GEJ_VERIFY(a);
/** For secp256k1, 2Q is infinity if and only if Q is infinity. This is because if 2Q = infinity,
* Q must equal -Q, or that Q.y == -(Q.y), or Q.y is 0. For a point on y^2 = x^3 + 7 to have
* y=0, x^3 must be -7 mod p. However, -7 has no cube root mod p.
*
* Having said this, if this function receives a point on a sextic twist, e.g. by
* a fault attack, it is possible for y to be 0. This happens for y^2 = x^3 + 6,
* since -6 does have a cube root mod p. For this point, this function will not set
* the infinity flag even though the point doubles to infinity, and the result
* point will be gibberish (z = 0 but infinity = 0).
*/
if (a->infinity) {
secp256k1_gej_set_infinity(r);
if (rzr != NULL) {
secp256k1_fe_set_int(rzr, 1);
}
return;
}
if (rzr != NULL) {
*rzr = a->y;
secp256k1_fe_normalize_weak(rzr);
}
secp256k1_gej_double(r, a);
SECP256K1_GEJ_VERIFY(r);
}
static void secp256k1_gej_add_var(secp256k1_gej *r, const secp256k1_gej *a, const secp256k1_gej *b, secp256k1_fe *rzr) {
/* 12 mul, 4 sqr, 11 add/negate/normalizes_to_zero (ignoring special cases) */
secp256k1_fe z22, z12, u1, u2, s1, s2, h, i, h2, h3, t;
SECP256K1_GEJ_VERIFY(a);
SECP256K1_GEJ_VERIFY(b);
if (a->infinity) {
VERIFY_CHECK(rzr == NULL);
*r = *b;
return;
}
if (b->infinity) {
if (rzr != NULL) {
secp256k1_fe_set_int(rzr, 1);
}
*r = *a;
return;
}
secp256k1_fe_sqr(&z22, &b->z);
secp256k1_fe_sqr(&z12, &a->z);
secp256k1_fe_mul(&u1, &a->x, &z22);
secp256k1_fe_mul(&u2, &b->x, &z12);
secp256k1_fe_mul(&s1, &a->y, &z22); secp256k1_fe_mul(&s1, &s1, &b->z);
secp256k1_fe_mul(&s2, &b->y, &z12); secp256k1_fe_mul(&s2, &s2, &a->z);
secp256k1_fe_negate(&h, &u1, 1); secp256k1_fe_add(&h, &u2);
secp256k1_fe_negate(&i, &s2, 1); secp256k1_fe_add(&i, &s1);
if (secp256k1_fe_normalizes_to_zero_var(&h)) {
if (secp256k1_fe_normalizes_to_zero_var(&i)) {
secp256k1_gej_double_var(r, a, rzr);
} else {
if (rzr != NULL) {
secp256k1_fe_set_int(rzr, 0);
}
secp256k1_gej_set_infinity(r);
}
return;
}
r->infinity = 0;
secp256k1_fe_mul(&t, &h, &b->z);
if (rzr != NULL) {
*rzr = t;
}
secp256k1_fe_mul(&r->z, &a->z, &t);
secp256k1_fe_sqr(&h2, &h);
secp256k1_fe_negate(&h2, &h2, 1);
secp256k1_fe_mul(&h3, &h2, &h);
secp256k1_fe_mul(&t, &u1, &h2);
secp256k1_fe_sqr(&r->x, &i);
secp256k1_fe_add(&r->x, &h3);
secp256k1_fe_add(&r->x, &t);
secp256k1_fe_add(&r->x, &t);
secp256k1_fe_add(&t, &r->x);
secp256k1_fe_mul(&r->y, &t, &i);
secp256k1_fe_mul(&h3, &h3, &s1);
secp256k1_fe_add(&r->y, &h3);
SECP256K1_GEJ_VERIFY(r);
}
static void secp256k1_gej_add_ge_var(secp256k1_gej *r, const secp256k1_gej *a, const secp256k1_ge *b, secp256k1_fe *rzr) {
/* Operations: 8 mul, 3 sqr, 11 add/negate/normalizes_to_zero (ignoring special cases) */
secp256k1_fe z12, u1, u2, s1, s2, h, i, h2, h3, t;
SECP256K1_GEJ_VERIFY(a);
SECP256K1_GE_VERIFY(b);
if (a->infinity) {
VERIFY_CHECK(rzr == NULL);
secp256k1_gej_set_ge(r, b);
return;
}
if (b->infinity) {
if (rzr != NULL) {
secp256k1_fe_set_int(rzr, 1);
}
*r = *a;
return;
}
secp256k1_fe_sqr(&z12, &a->z);
u1 = a->x;
secp256k1_fe_mul(&u2, &b->x, &z12);
s1 = a->y;
secp256k1_fe_mul(&s2, &b->y, &z12); secp256k1_fe_mul(&s2, &s2, &a->z);
secp256k1_fe_negate(&h, &u1, SECP256K1_GEJ_X_MAGNITUDE_MAX); secp256k1_fe_add(&h, &u2);
secp256k1_fe_negate(&i, &s2, 1); secp256k1_fe_add(&i, &s1);
if (secp256k1_fe_normalizes_to_zero_var(&h)) {
if (secp256k1_fe_normalizes_to_zero_var(&i)) {
secp256k1_gej_double_var(r, a, rzr);
} else {
if (rzr != NULL) {
secp256k1_fe_set_int(rzr, 0);
}
secp256k1_gej_set_infinity(r);
}
return;
}
r->infinity = 0;
if (rzr != NULL) {
*rzr = h;
}
secp256k1_fe_mul(&r->z, &a->z, &h);
secp256k1_fe_sqr(&h2, &h);
secp256k1_fe_negate(&h2, &h2, 1);
secp256k1_fe_mul(&h3, &h2, &h);
secp256k1_fe_mul(&t, &u1, &h2);
secp256k1_fe_sqr(&r->x, &i);
secp256k1_fe_add(&r->x, &h3);
secp256k1_fe_add(&r->x, &t);
secp256k1_fe_add(&r->x, &t);
secp256k1_fe_add(&t, &r->x);
secp256k1_fe_mul(&r->y, &t, &i);
secp256k1_fe_mul(&h3, &h3, &s1);
secp256k1_fe_add(&r->y, &h3);
SECP256K1_GEJ_VERIFY(r);
if (rzr != NULL) SECP256K1_FE_VERIFY(rzr);
}
static void secp256k1_gej_add_zinv_var(secp256k1_gej *r, const secp256k1_gej *a, const secp256k1_ge *b, const secp256k1_fe *bzinv) {
/* Operations: 9 mul, 3 sqr, 11 add/negate/normalizes_to_zero (ignoring special cases) */
secp256k1_fe az, z12, u1, u2, s1, s2, h, i, h2, h3, t;
SECP256K1_GEJ_VERIFY(a);
SECP256K1_GE_VERIFY(b);
SECP256K1_FE_VERIFY(bzinv);
if (a->infinity) {
secp256k1_fe bzinv2, bzinv3;
r->infinity = b->infinity;
secp256k1_fe_sqr(&bzinv2, bzinv);
secp256k1_fe_mul(&bzinv3, &bzinv2, bzinv);
secp256k1_fe_mul(&r->x, &b->x, &bzinv2);
secp256k1_fe_mul(&r->y, &b->y, &bzinv3);
secp256k1_fe_set_int(&r->z, 1);
SECP256K1_GEJ_VERIFY(r);
return;
}
if (b->infinity) {
*r = *a;
return;
}
/** We need to calculate (rx,ry,rz) = (ax,ay,az) + (bx,by,1/bzinv). Due to
* secp256k1's isomorphism we can multiply the Z coordinates on both sides
* by bzinv, and get: (rx,ry,rz*bzinv) = (ax,ay,az*bzinv) + (bx,by,1).
* This means that (rx,ry,rz) can be calculated as
* (ax,ay,az*bzinv) + (bx,by,1), when not applying the bzinv factor to rz.
* The variable az below holds the modified Z coordinate for a, which is used
* for the computation of rx and ry, but not for rz.
*/
secp256k1_fe_mul(&az, &a->z, bzinv);
secp256k1_fe_sqr(&z12, &az);
u1 = a->x;
secp256k1_fe_mul(&u2, &b->x, &z12);
s1 = a->y;
secp256k1_fe_mul(&s2, &b->y, &z12); secp256k1_fe_mul(&s2, &s2, &az);
secp256k1_fe_negate(&h, &u1, SECP256K1_GEJ_X_MAGNITUDE_MAX); secp256k1_fe_add(&h, &u2);
secp256k1_fe_negate(&i, &s2, 1); secp256k1_fe_add(&i, &s1);
if (secp256k1_fe_normalizes_to_zero_var(&h)) {
if (secp256k1_fe_normalizes_to_zero_var(&i)) {
secp256k1_gej_double_var(r, a, NULL);
} else {
secp256k1_gej_set_infinity(r);
}
return;
}
r->infinity = 0;
secp256k1_fe_mul(&r->z, &a->z, &h);
secp256k1_fe_sqr(&h2, &h);
secp256k1_fe_negate(&h2, &h2, 1);
secp256k1_fe_mul(&h3, &h2, &h);
secp256k1_fe_mul(&t, &u1, &h2);
secp256k1_fe_sqr(&r->x, &i);
secp256k1_fe_add(&r->x, &h3);
secp256k1_fe_add(&r->x, &t);
secp256k1_fe_add(&r->x, &t);
secp256k1_fe_add(&t, &r->x);
secp256k1_fe_mul(&r->y, &t, &i);
secp256k1_fe_mul(&h3, &h3, &s1);
secp256k1_fe_add(&r->y, &h3);
SECP256K1_GEJ_VERIFY(r);
}
static void secp256k1_gej_add_ge(secp256k1_gej *r, const secp256k1_gej *a, const secp256k1_ge *b) {
/* Operations: 7 mul, 5 sqr, 21 add/cmov/half/mul_int/negate/normalizes_to_zero */
secp256k1_fe zz, u1, u2, s1, s2, t, tt, m, n, q, rr;
secp256k1_fe m_alt, rr_alt;
int degenerate;
SECP256K1_GEJ_VERIFY(a);
SECP256K1_GE_VERIFY(b);
VERIFY_CHECK(!b->infinity);
/* In:
* Eric Brier and Marc Joye, Weierstrass Elliptic Curves and Side-Channel Attacks.
* In D. Naccache and P. Paillier, Eds., Public Key Cryptography, vol. 2274 of Lecture Notes in Computer Science, pages 335-345. Springer-Verlag, 2002.
* we find as solution for a unified addition/doubling formula:
* lambda = ((x1 + x2)^2 - x1 * x2 + a) / (y1 + y2), with a = 0 for secp256k1's curve equation.
* x3 = lambda^2 - (x1 + x2)
* 2*y3 = lambda * (x1 + x2 - 2 * x3) - (y1 + y2).
*
* Substituting x_i = Xi / Zi^2 and yi = Yi / Zi^3, for i=1,2,3, gives:
* U1 = X1*Z2^2, U2 = X2*Z1^2
* S1 = Y1*Z2^3, S2 = Y2*Z1^3
* Z = Z1*Z2
* T = U1+U2
* M = S1+S2
* Q = -T*M^2
* R = T^2-U1*U2
* X3 = R^2+Q
* Y3 = -(R*(2*X3+Q)+M^4)/2
* Z3 = M*Z
* (Note that the paper uses xi = Xi / Zi and yi = Yi / Zi instead.)
*
* This formula has the benefit of being the same for both addition
* of distinct points and doubling. However, it breaks down in the
* case that either point is infinity, or that y1 = -y2. We handle
* these cases in the following ways:
*
* - If b is infinity we simply bail by means of a VERIFY_CHECK.
*
* - If a is infinity, we detect this, and at the end of the
* computation replace the result (which will be meaningless,
* but we compute to be constant-time) with b.x : b.y : 1.
*
* - If a = -b, we have y1 = -y2, which is a degenerate case.
* But here the answer is infinity, so we simply set the
* infinity flag of the result, overriding the computed values
* without even needing to cmov.
*
* - If y1 = -y2 but x1 != x2, which does occur thanks to certain
* properties of our curve (specifically, 1 has nontrivial cube
* roots in our field, and the curve equation has no x coefficient)
* then the answer is not infinity but also not given by the above
* equation. In this case, we cmov in place an alternate expression
* for lambda. Specifically (y1 - y2)/(x1 - x2). Where both these
* expressions for lambda are defined, they are equal, and can be
* obtained from each other by multiplication by (y1 + y2)/(y1 + y2)
* then substitution of x^3 + 7 for y^2 (using the curve equation).
* For all pairs of nonzero points (a, b) at least one is defined,
* so this covers everything.
*/
secp256k1_fe_sqr(&zz, &a->z); /* z = Z1^2 */
u1 = a->x; /* u1 = U1 = X1*Z2^2 (GEJ_X_M) */
secp256k1_fe_mul(&u2, &b->x, &zz); /* u2 = U2 = X2*Z1^2 (1) */
s1 = a->y; /* s1 = S1 = Y1*Z2^3 (GEJ_Y_M) */
secp256k1_fe_mul(&s2, &b->y, &zz); /* s2 = Y2*Z1^2 (1) */
secp256k1_fe_mul(&s2, &s2, &a->z); /* s2 = S2 = Y2*Z1^3 (1) */
t = u1; secp256k1_fe_add(&t, &u2); /* t = T = U1+U2 (GEJ_X_M+1) */
m = s1; secp256k1_fe_add(&m, &s2); /* m = M = S1+S2 (GEJ_Y_M+1) */
secp256k1_fe_sqr(&rr, &t); /* rr = T^2 (1) */
secp256k1_fe_negate(&m_alt, &u2, 1); /* Malt = -X2*Z1^2 (2) */
secp256k1_fe_mul(&tt, &u1, &m_alt); /* tt = -U1*U2 (1) */
secp256k1_fe_add(&rr, &tt); /* rr = R = T^2-U1*U2 (2) */
/* If lambda = R/M = R/0 we have a problem (except in the "trivial"
* case that Z = z1z2 = 0, and this is special-cased later on). */
degenerate = secp256k1_fe_normalizes_to_zero(&m);
/* This only occurs when y1 == -y2 and x1^3 == x2^3, but x1 != x2.
* This means either x1 == beta*x2 or beta*x1 == x2, where beta is
* a nontrivial cube root of one. In either case, an alternate
* non-indeterminate expression for lambda is (y1 - y2)/(x1 - x2),
* so we set R/M equal to this. */
rr_alt = s1;
secp256k1_fe_mul_int(&rr_alt, 2); /* rr_alt = Y1*Z2^3 - Y2*Z1^3 (GEJ_Y_M*2) */
secp256k1_fe_add(&m_alt, &u1); /* Malt = X1*Z2^2 - X2*Z1^2 (GEJ_X_M+2) */
secp256k1_fe_cmov(&rr_alt, &rr, !degenerate); /* rr_alt (GEJ_Y_M*2) */
secp256k1_fe_cmov(&m_alt, &m, !degenerate); /* m_alt (GEJ_X_M+2) */
/* Now Ralt / Malt = lambda and is guaranteed not to be Ralt / 0.
* From here on out Ralt and Malt represent the numerator
* and denominator of lambda; R and M represent the explicit
* expressions x1^2 + x2^2 + x1x2 and y1 + y2. */
secp256k1_fe_sqr(&n, &m_alt); /* n = Malt^2 (1) */
secp256k1_fe_negate(&q, &t,
SECP256K1_GEJ_X_MAGNITUDE_MAX + 1); /* q = -T (GEJ_X_M+2) */
secp256k1_fe_mul(&q, &q, &n); /* q = Q = -T*Malt^2 (1) */
/* These two lines use the observation that either M == Malt or M == 0,
* so M^3 * Malt is either Malt^4 (which is computed by squaring), or
* zero (which is "computed" by cmov). So the cost is one squaring
* versus two multiplications. */
secp256k1_fe_sqr(&n, &n); /* n = Malt^4 (1) */
secp256k1_fe_cmov(&n, &m, degenerate); /* n = M^3 * Malt (GEJ_Y_M+1) */
secp256k1_fe_sqr(&t, &rr_alt); /* t = Ralt^2 (1) */
secp256k1_fe_mul(&r->z, &a->z, &m_alt); /* r->z = Z3 = Malt*Z (1) */
secp256k1_fe_add(&t, &q); /* t = Ralt^2 + Q (2) */
r->x = t; /* r->x = X3 = Ralt^2 + Q (2) */
secp256k1_fe_mul_int(&t, 2); /* t = 2*X3 (4) */
secp256k1_fe_add(&t, &q); /* t = 2*X3 + Q (5) */
secp256k1_fe_mul(&t, &t, &rr_alt); /* t = Ralt*(2*X3 + Q) (1) */
secp256k1_fe_add(&t, &n); /* t = Ralt*(2*X3 + Q) + M^3*Malt (GEJ_Y_M+2) */
secp256k1_fe_negate(&r->y, &t,
SECP256K1_GEJ_Y_MAGNITUDE_MAX + 2); /* r->y = -(Ralt*(2*X3 + Q) + M^3*Malt) (GEJ_Y_M+3) */
secp256k1_fe_half(&r->y); /* r->y = Y3 = -(Ralt*(2*X3 + Q) + M^3*Malt)/2 ((GEJ_Y_M+3)/2 + 1) */
/* In case a->infinity == 1, replace r with (b->x, b->y, 1). */
secp256k1_fe_cmov(&r->x, &b->x, a->infinity);
secp256k1_fe_cmov(&r->y, &b->y, a->infinity);
secp256k1_fe_cmov(&r->z, &secp256k1_fe_one, a->infinity);
/* Set r->infinity if r->z is 0.
*
* If a->infinity is set, then r->infinity = (r->z == 0) = (1 == 0) = false,
* which is correct because the function assumes that b is not infinity.
*
* Now assume !a->infinity. This implies Z = Z1 != 0.
*
* Case y1 = -y2:
* In this case we could have a = -b, namely if x1 = x2.
* We have degenerate = true, r->z = (x1 - x2) * Z.
* Then r->infinity = ((x1 - x2)Z == 0) = (x1 == x2) = (a == -b).
*
* Case y1 != -y2:
* In this case, we can't have a = -b.
* We have degenerate = false, r->z = (y1 + y2) * Z.
* Then r->infinity = ((y1 + y2)Z == 0) = (y1 == -y2) = false. */
r->infinity = secp256k1_fe_normalizes_to_zero(&r->z);
SECP256K1_GEJ_VERIFY(r);
}
static void secp256k1_gej_rescale(secp256k1_gej *r, const secp256k1_fe *s) {
/* Operations: 4 mul, 1 sqr */
secp256k1_fe zz;
SECP256K1_GEJ_VERIFY(r);
SECP256K1_FE_VERIFY(s);
VERIFY_CHECK(!secp256k1_fe_normalizes_to_zero_var(s));
secp256k1_fe_sqr(&zz, s);
secp256k1_fe_mul(&r->x, &r->x, &zz); /* r->x *= s^2 */
secp256k1_fe_mul(&r->y, &r->y, &zz);
secp256k1_fe_mul(&r->y, &r->y, s); /* r->y *= s^3 */
secp256k1_fe_mul(&r->z, &r->z, s); /* r->z *= s */
SECP256K1_GEJ_VERIFY(r);
}
static void secp256k1_ge_to_storage(secp256k1_ge_storage *r, const secp256k1_ge *a) {
secp256k1_fe x, y;
SECP256K1_GE_VERIFY(a);
VERIFY_CHECK(!a->infinity);
x = a->x;
secp256k1_fe_normalize(&x);
y = a->y;
secp256k1_fe_normalize(&y);
secp256k1_fe_to_storage(&r->x, &x);
secp256k1_fe_to_storage(&r->y, &y);
}
static void secp256k1_ge_from_storage(secp256k1_ge *r, const secp256k1_ge_storage *a) {
secp256k1_fe_from_storage(&r->x, &a->x);
secp256k1_fe_from_storage(&r->y, &a->y);
r->infinity = 0;
SECP256K1_GE_VERIFY(r);
}
static SECP256K1_INLINE void secp256k1_gej_cmov(secp256k1_gej *r, const secp256k1_gej *a, int flag) {
SECP256K1_GEJ_VERIFY(r);
SECP256K1_GEJ_VERIFY(a);
secp256k1_fe_cmov(&r->x, &a->x, flag);
secp256k1_fe_cmov(&r->y, &a->y, flag);
secp256k1_fe_cmov(&r->z, &a->z, flag);
r->infinity ^= (r->infinity ^ a->infinity) & flag;
SECP256K1_GEJ_VERIFY(r);
}
static SECP256K1_INLINE void secp256k1_ge_storage_cmov(secp256k1_ge_storage *r, const secp256k1_ge_storage *a, int flag) {
secp256k1_fe_storage_cmov(&r->x, &a->x, flag);
secp256k1_fe_storage_cmov(&r->y, &a->y, flag);
}
static void secp256k1_ge_mul_lambda(secp256k1_ge *r, const secp256k1_ge *a) {
SECP256K1_GE_VERIFY(a);
*r = *a;
secp256k1_fe_mul(&r->x, &r->x, &secp256k1_const_beta);
SECP256K1_GE_VERIFY(r);
}
static int secp256k1_ge_is_in_correct_subgroup(const secp256k1_ge* ge) {
#ifdef EXHAUSTIVE_TEST_ORDER
secp256k1_gej out;
int i;
SECP256K1_GE_VERIFY(ge);
/* A very simple EC multiplication ladder that avoids a dependency on ecmult. */
secp256k1_gej_set_infinity(&out);
for (i = 0; i < 32; ++i) {
secp256k1_gej_double_var(&out, &out, NULL);
if ((((uint32_t)EXHAUSTIVE_TEST_ORDER) >> (31 - i)) & 1) {
secp256k1_gej_add_ge_var(&out, &out, ge, NULL);
}
}
return secp256k1_gej_is_infinity(&out);
#else
SECP256K1_GE_VERIFY(ge);
(void)ge;
/* The real secp256k1 group has cofactor 1, so the subgroup is the entire curve. */
return 1;
#endif
}
static int secp256k1_ge_x_on_curve_var(const secp256k1_fe *x) {
secp256k1_fe c;
secp256k1_fe_sqr(&c, x);
secp256k1_fe_mul(&c, &c, x);
secp256k1_fe_add_int(&c, SECP256K1_B);
return secp256k1_fe_is_square_var(&c);
}
static int secp256k1_ge_x_frac_on_curve_var(const secp256k1_fe *xn, const secp256k1_fe *xd) {
/* We want to determine whether (xn/xd) is on the curve.
*
* (xn/xd)^3 + 7 is square <=> xd*xn^3 + 7*xd^4 is square (multiplying by xd^4, a square).
*/
secp256k1_fe r, t;
VERIFY_CHECK(!secp256k1_fe_normalizes_to_zero_var(xd));
secp256k1_fe_mul(&r, xd, xn); /* r = xd*xn */
secp256k1_fe_sqr(&t, xn); /* t = xn^2 */
secp256k1_fe_mul(&r, &r, &t); /* r = xd*xn^3 */
secp256k1_fe_sqr(&t, xd); /* t = xd^2 */
secp256k1_fe_sqr(&t, &t); /* t = xd^4 */
VERIFY_CHECK(SECP256K1_B <= 31);
secp256k1_fe_mul_int(&t, SECP256K1_B); /* t = 7*xd^4 */
secp256k1_fe_add(&r, &t); /* r = xd*xn^3 + 7*xd^4 */
return secp256k1_fe_is_square_var(&r);
}
static void secp256k1_ge_to_bytes(unsigned char *buf, const secp256k1_ge *a) {
secp256k1_ge_storage s;
/* We require that the secp256k1_ge_storage type is exactly 64 bytes.
* This is formally not guaranteed by the C standard, but should hold on any
* sane compiler in the real world. */
STATIC_ASSERT(sizeof(secp256k1_ge_storage) == 64);
VERIFY_CHECK(!secp256k1_ge_is_infinity(a));
secp256k1_ge_to_storage(&s, a);
memcpy(buf, &s, 64);
}
static void secp256k1_ge_from_bytes(secp256k1_ge *r, const unsigned char *buf) {
secp256k1_ge_storage s;
STATIC_ASSERT(sizeof(secp256k1_ge_storage) == 64);
memcpy(&s, buf, 64);
secp256k1_ge_from_storage(r, &s);
}
static void secp256k1_ge_to_bytes_ext(unsigned char *data, const secp256k1_ge *ge) {
if (secp256k1_ge_is_infinity(ge)) {
memset(data, 0, 64);
} else {
secp256k1_ge_to_bytes(data, ge);
}
}
static void secp256k1_ge_from_bytes_ext(secp256k1_ge *ge, const unsigned char *data) {
static const unsigned char zeros[64] = { 0 };
if (secp256k1_memcmp_var(data, zeros, sizeof(zeros)) == 0) {
secp256k1_ge_set_infinity(ge);
} else {
secp256k1_ge_from_bytes(ge, data);
}
}
#endif /* SECP256K1_GROUP_IMPL_H */

44
libsecp256k1/src/hash.h Normal file
View File

@@ -0,0 +1,44 @@
/***********************************************************************
* Copyright (c) 2014 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_HASH_H
#define SECP256K1_HASH_H
#include <stdlib.h>
#include <stdint.h>
typedef struct {
uint32_t s[8];
unsigned char buf[64];
uint64_t bytes;
} secp256k1_sha256;
static void secp256k1_sha256_initialize(secp256k1_sha256 *hash);
static void secp256k1_sha256_write(secp256k1_sha256 *hash, const unsigned char *data, size_t size);
static void secp256k1_sha256_finalize(secp256k1_sha256 *hash, unsigned char *out32);
static void secp256k1_sha256_clear(secp256k1_sha256 *hash);
typedef struct {
secp256k1_sha256 inner, outer;
} secp256k1_hmac_sha256;
static void secp256k1_hmac_sha256_initialize(secp256k1_hmac_sha256 *hash, const unsigned char *key, size_t size);
static void secp256k1_hmac_sha256_write(secp256k1_hmac_sha256 *hash, const unsigned char *data, size_t size);
static void secp256k1_hmac_sha256_finalize(secp256k1_hmac_sha256 *hash, unsigned char *out32);
static void secp256k1_hmac_sha256_clear(secp256k1_hmac_sha256 *hash);
typedef struct {
unsigned char v[32];
unsigned char k[32];
int retry;
} secp256k1_rfc6979_hmac_sha256;
static void secp256k1_rfc6979_hmac_sha256_initialize(secp256k1_rfc6979_hmac_sha256 *rng, const unsigned char *key, size_t keylen);
static void secp256k1_rfc6979_hmac_sha256_generate(secp256k1_rfc6979_hmac_sha256 *rng, unsigned char *out, size_t outlen);
static void secp256k1_rfc6979_hmac_sha256_finalize(secp256k1_rfc6979_hmac_sha256 *rng);
static void secp256k1_rfc6979_hmac_sha256_clear(secp256k1_rfc6979_hmac_sha256 *rng);
#endif /* SECP256K1_HASH_H */

View File

@@ -0,0 +1,299 @@
/***********************************************************************
* Copyright (c) 2014 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_HASH_IMPL_H
#define SECP256K1_HASH_IMPL_H
#include "hash.h"
#include "util.h"
#include <stdlib.h>
#include <stdint.h>
#include <string.h>
#define Ch(x,y,z) ((z) ^ ((x) & ((y) ^ (z))))
#define Maj(x,y,z) (((x) & (y)) | ((z) & ((x) | (y))))
#define Sigma0(x) (((x) >> 2 | (x) << 30) ^ ((x) >> 13 | (x) << 19) ^ ((x) >> 22 | (x) << 10))
#define Sigma1(x) (((x) >> 6 | (x) << 26) ^ ((x) >> 11 | (x) << 21) ^ ((x) >> 25 | (x) << 7))
#define sigma0(x) (((x) >> 7 | (x) << 25) ^ ((x) >> 18 | (x) << 14) ^ ((x) >> 3))
#define sigma1(x) (((x) >> 17 | (x) << 15) ^ ((x) >> 19 | (x) << 13) ^ ((x) >> 10))
#define Round(a,b,c,d,e,f,g,h,k,w) do { \
uint32_t t1 = (h) + Sigma1(e) + Ch((e), (f), (g)) + (k) + (w); \
uint32_t t2 = Sigma0(a) + Maj((a), (b), (c)); \
(d) += t1; \
(h) = t1 + t2; \
} while(0)
static void secp256k1_sha256_initialize(secp256k1_sha256 *hash) {
hash->s[0] = 0x6a09e667ul;
hash->s[1] = 0xbb67ae85ul;
hash->s[2] = 0x3c6ef372ul;
hash->s[3] = 0xa54ff53aul;
hash->s[4] = 0x510e527ful;
hash->s[5] = 0x9b05688cul;
hash->s[6] = 0x1f83d9abul;
hash->s[7] = 0x5be0cd19ul;
hash->bytes = 0;
}
/** Perform one SHA-256 transformation, processing 16 big endian 32-bit words. */
static void secp256k1_sha256_transform(uint32_t* s, const unsigned char* buf) {
uint32_t a = s[0], b = s[1], c = s[2], d = s[3], e = s[4], f = s[5], g = s[6], h = s[7];
uint32_t w0, w1, w2, w3, w4, w5, w6, w7, w8, w9, w10, w11, w12, w13, w14, w15;
Round(a, b, c, d, e, f, g, h, 0x428a2f98, w0 = secp256k1_read_be32(&buf[0]));
Round(h, a, b, c, d, e, f, g, 0x71374491, w1 = secp256k1_read_be32(&buf[4]));
Round(g, h, a, b, c, d, e, f, 0xb5c0fbcf, w2 = secp256k1_read_be32(&buf[8]));
Round(f, g, h, a, b, c, d, e, 0xe9b5dba5, w3 = secp256k1_read_be32(&buf[12]));
Round(e, f, g, h, a, b, c, d, 0x3956c25b, w4 = secp256k1_read_be32(&buf[16]));
Round(d, e, f, g, h, a, b, c, 0x59f111f1, w5 = secp256k1_read_be32(&buf[20]));
Round(c, d, e, f, g, h, a, b, 0x923f82a4, w6 = secp256k1_read_be32(&buf[24]));
Round(b, c, d, e, f, g, h, a, 0xab1c5ed5, w7 = secp256k1_read_be32(&buf[28]));
Round(a, b, c, d, e, f, g, h, 0xd807aa98, w8 = secp256k1_read_be32(&buf[32]));
Round(h, a, b, c, d, e, f, g, 0x12835b01, w9 = secp256k1_read_be32(&buf[36]));
Round(g, h, a, b, c, d, e, f, 0x243185be, w10 = secp256k1_read_be32(&buf[40]));
Round(f, g, h, a, b, c, d, e, 0x550c7dc3, w11 = secp256k1_read_be32(&buf[44]));
Round(e, f, g, h, a, b, c, d, 0x72be5d74, w12 = secp256k1_read_be32(&buf[48]));
Round(d, e, f, g, h, a, b, c, 0x80deb1fe, w13 = secp256k1_read_be32(&buf[52]));
Round(c, d, e, f, g, h, a, b, 0x9bdc06a7, w14 = secp256k1_read_be32(&buf[56]));
Round(b, c, d, e, f, g, h, a, 0xc19bf174, w15 = secp256k1_read_be32(&buf[60]));
Round(a, b, c, d, e, f, g, h, 0xe49b69c1, w0 += sigma1(w14) + w9 + sigma0(w1));
Round(h, a, b, c, d, e, f, g, 0xefbe4786, w1 += sigma1(w15) + w10 + sigma0(w2));
Round(g, h, a, b, c, d, e, f, 0x0fc19dc6, w2 += sigma1(w0) + w11 + sigma0(w3));
Round(f, g, h, a, b, c, d, e, 0x240ca1cc, w3 += sigma1(w1) + w12 + sigma0(w4));
Round(e, f, g, h, a, b, c, d, 0x2de92c6f, w4 += sigma1(w2) + w13 + sigma0(w5));
Round(d, e, f, g, h, a, b, c, 0x4a7484aa, w5 += sigma1(w3) + w14 + sigma0(w6));
Round(c, d, e, f, g, h, a, b, 0x5cb0a9dc, w6 += sigma1(w4) + w15 + sigma0(w7));
Round(b, c, d, e, f, g, h, a, 0x76f988da, w7 += sigma1(w5) + w0 + sigma0(w8));
Round(a, b, c, d, e, f, g, h, 0x983e5152, w8 += sigma1(w6) + w1 + sigma0(w9));
Round(h, a, b, c, d, e, f, g, 0xa831c66d, w9 += sigma1(w7) + w2 + sigma0(w10));
Round(g, h, a, b, c, d, e, f, 0xb00327c8, w10 += sigma1(w8) + w3 + sigma0(w11));
Round(f, g, h, a, b, c, d, e, 0xbf597fc7, w11 += sigma1(w9) + w4 + sigma0(w12));
Round(e, f, g, h, a, b, c, d, 0xc6e00bf3, w12 += sigma1(w10) + w5 + sigma0(w13));
Round(d, e, f, g, h, a, b, c, 0xd5a79147, w13 += sigma1(w11) + w6 + sigma0(w14));
Round(c, d, e, f, g, h, a, b, 0x06ca6351, w14 += sigma1(w12) + w7 + sigma0(w15));
Round(b, c, d, e, f, g, h, a, 0x14292967, w15 += sigma1(w13) + w8 + sigma0(w0));
Round(a, b, c, d, e, f, g, h, 0x27b70a85, w0 += sigma1(w14) + w9 + sigma0(w1));
Round(h, a, b, c, d, e, f, g, 0x2e1b2138, w1 += sigma1(w15) + w10 + sigma0(w2));
Round(g, h, a, b, c, d, e, f, 0x4d2c6dfc, w2 += sigma1(w0) + w11 + sigma0(w3));
Round(f, g, h, a, b, c, d, e, 0x53380d13, w3 += sigma1(w1) + w12 + sigma0(w4));
Round(e, f, g, h, a, b, c, d, 0x650a7354, w4 += sigma1(w2) + w13 + sigma0(w5));
Round(d, e, f, g, h, a, b, c, 0x766a0abb, w5 += sigma1(w3) + w14 + sigma0(w6));
Round(c, d, e, f, g, h, a, b, 0x81c2c92e, w6 += sigma1(w4) + w15 + sigma0(w7));
Round(b, c, d, e, f, g, h, a, 0x92722c85, w7 += sigma1(w5) + w0 + sigma0(w8));
Round(a, b, c, d, e, f, g, h, 0xa2bfe8a1, w8 += sigma1(w6) + w1 + sigma0(w9));
Round(h, a, b, c, d, e, f, g, 0xa81a664b, w9 += sigma1(w7) + w2 + sigma0(w10));
Round(g, h, a, b, c, d, e, f, 0xc24b8b70, w10 += sigma1(w8) + w3 + sigma0(w11));
Round(f, g, h, a, b, c, d, e, 0xc76c51a3, w11 += sigma1(w9) + w4 + sigma0(w12));
Round(e, f, g, h, a, b, c, d, 0xd192e819, w12 += sigma1(w10) + w5 + sigma0(w13));
Round(d, e, f, g, h, a, b, c, 0xd6990624, w13 += sigma1(w11) + w6 + sigma0(w14));
Round(c, d, e, f, g, h, a, b, 0xf40e3585, w14 += sigma1(w12) + w7 + sigma0(w15));
Round(b, c, d, e, f, g, h, a, 0x106aa070, w15 += sigma1(w13) + w8 + sigma0(w0));
Round(a, b, c, d, e, f, g, h, 0x19a4c116, w0 += sigma1(w14) + w9 + sigma0(w1));
Round(h, a, b, c, d, e, f, g, 0x1e376c08, w1 += sigma1(w15) + w10 + sigma0(w2));
Round(g, h, a, b, c, d, e, f, 0x2748774c, w2 += sigma1(w0) + w11 + sigma0(w3));
Round(f, g, h, a, b, c, d, e, 0x34b0bcb5, w3 += sigma1(w1) + w12 + sigma0(w4));
Round(e, f, g, h, a, b, c, d, 0x391c0cb3, w4 += sigma1(w2) + w13 + sigma0(w5));
Round(d, e, f, g, h, a, b, c, 0x4ed8aa4a, w5 += sigma1(w3) + w14 + sigma0(w6));
Round(c, d, e, f, g, h, a, b, 0x5b9cca4f, w6 += sigma1(w4) + w15 + sigma0(w7));
Round(b, c, d, e, f, g, h, a, 0x682e6ff3, w7 += sigma1(w5) + w0 + sigma0(w8));
Round(a, b, c, d, e, f, g, h, 0x748f82ee, w8 += sigma1(w6) + w1 + sigma0(w9));
Round(h, a, b, c, d, e, f, g, 0x78a5636f, w9 += sigma1(w7) + w2 + sigma0(w10));
Round(g, h, a, b, c, d, e, f, 0x84c87814, w10 += sigma1(w8) + w3 + sigma0(w11));
Round(f, g, h, a, b, c, d, e, 0x8cc70208, w11 += sigma1(w9) + w4 + sigma0(w12));
Round(e, f, g, h, a, b, c, d, 0x90befffa, w12 += sigma1(w10) + w5 + sigma0(w13));
Round(d, e, f, g, h, a, b, c, 0xa4506ceb, w13 += sigma1(w11) + w6 + sigma0(w14));
Round(c, d, e, f, g, h, a, b, 0xbef9a3f7, w14 + sigma1(w12) + w7 + sigma0(w15));
Round(b, c, d, e, f, g, h, a, 0xc67178f2, w15 + sigma1(w13) + w8 + sigma0(w0));
s[0] += a;
s[1] += b;
s[2] += c;
s[3] += d;
s[4] += e;
s[5] += f;
s[6] += g;
s[7] += h;
}
static void secp256k1_sha256_write(secp256k1_sha256 *hash, const unsigned char *data, size_t len) {
size_t bufsize = hash->bytes & 0x3F;
hash->bytes += len;
VERIFY_CHECK(hash->bytes >= len);
while (len >= 64 - bufsize) {
/* Fill the buffer, and process it. */
size_t chunk_len = 64 - bufsize;
memcpy(hash->buf + bufsize, data, chunk_len);
data += chunk_len;
len -= chunk_len;
secp256k1_sha256_transform(hash->s, hash->buf);
bufsize = 0;
}
if (len) {
/* Fill the buffer with what remains. */
memcpy(hash->buf + bufsize, data, len);
}
}
static void secp256k1_sha256_finalize(secp256k1_sha256 *hash, unsigned char *out32) {
static const unsigned char pad[64] = {0x80};
unsigned char sizedesc[8];
int i;
/* The maximum message size of SHA256 is 2^64-1 bits. */
VERIFY_CHECK(hash->bytes < ((uint64_t)1 << 61));
secp256k1_write_be32(&sizedesc[0], hash->bytes >> 29);
secp256k1_write_be32(&sizedesc[4], hash->bytes << 3);
secp256k1_sha256_write(hash, pad, 1 + ((119 - (hash->bytes % 64)) % 64));
secp256k1_sha256_write(hash, sizedesc, 8);
for (i = 0; i < 8; i++) {
secp256k1_write_be32(&out32[4*i], hash->s[i]);
hash->s[i] = 0;
}
}
/* Initializes a sha256 struct and writes the 64 byte string
* SHA256(tag)||SHA256(tag) into it. */
static void secp256k1_sha256_initialize_tagged(secp256k1_sha256 *hash, const unsigned char *tag, size_t taglen) {
unsigned char buf[32];
secp256k1_sha256_initialize(hash);
secp256k1_sha256_write(hash, tag, taglen);
secp256k1_sha256_finalize(hash, buf);
secp256k1_sha256_initialize(hash);
secp256k1_sha256_write(hash, buf, 32);
secp256k1_sha256_write(hash, buf, 32);
}
static void secp256k1_sha256_clear(secp256k1_sha256 *hash) {
secp256k1_memclear(hash, sizeof(*hash));
}
static void secp256k1_hmac_sha256_initialize(secp256k1_hmac_sha256 *hash, const unsigned char *key, size_t keylen) {
size_t n;
unsigned char rkey[64];
if (keylen <= sizeof(rkey)) {
memcpy(rkey, key, keylen);
memset(rkey + keylen, 0, sizeof(rkey) - keylen);
} else {
secp256k1_sha256 sha256;
secp256k1_sha256_initialize(&sha256);
secp256k1_sha256_write(&sha256, key, keylen);
secp256k1_sha256_finalize(&sha256, rkey);
memset(rkey + 32, 0, 32);
}
secp256k1_sha256_initialize(&hash->outer);
for (n = 0; n < sizeof(rkey); n++) {
rkey[n] ^= 0x5c;
}
secp256k1_sha256_write(&hash->outer, rkey, sizeof(rkey));
secp256k1_sha256_initialize(&hash->inner);
for (n = 0; n < sizeof(rkey); n++) {
rkey[n] ^= 0x5c ^ 0x36;
}
secp256k1_sha256_write(&hash->inner, rkey, sizeof(rkey));
secp256k1_memclear(rkey, sizeof(rkey));
}
static void secp256k1_hmac_sha256_write(secp256k1_hmac_sha256 *hash, const unsigned char *data, size_t size) {
secp256k1_sha256_write(&hash->inner, data, size);
}
static void secp256k1_hmac_sha256_finalize(secp256k1_hmac_sha256 *hash, unsigned char *out32) {
unsigned char temp[32];
secp256k1_sha256_finalize(&hash->inner, temp);
secp256k1_sha256_write(&hash->outer, temp, 32);
secp256k1_memclear(temp, sizeof(temp));
secp256k1_sha256_finalize(&hash->outer, out32);
}
static void secp256k1_hmac_sha256_clear(secp256k1_hmac_sha256 *hash) {
secp256k1_memclear(hash, sizeof(*hash));
}
static void secp256k1_rfc6979_hmac_sha256_initialize(secp256k1_rfc6979_hmac_sha256 *rng, const unsigned char *key, size_t keylen) {
secp256k1_hmac_sha256 hmac;
static const unsigned char zero[1] = {0x00};
static const unsigned char one[1] = {0x01};
memset(rng->v, 0x01, 32); /* RFC6979 3.2.b. */
memset(rng->k, 0x00, 32); /* RFC6979 3.2.c. */
/* RFC6979 3.2.d. */
secp256k1_hmac_sha256_initialize(&hmac, rng->k, 32);
secp256k1_hmac_sha256_write(&hmac, rng->v, 32);
secp256k1_hmac_sha256_write(&hmac, zero, 1);
secp256k1_hmac_sha256_write(&hmac, key, keylen);
secp256k1_hmac_sha256_finalize(&hmac, rng->k);
secp256k1_hmac_sha256_initialize(&hmac, rng->k, 32);
secp256k1_hmac_sha256_write(&hmac, rng->v, 32);
secp256k1_hmac_sha256_finalize(&hmac, rng->v);
/* RFC6979 3.2.f. */
secp256k1_hmac_sha256_initialize(&hmac, rng->k, 32);
secp256k1_hmac_sha256_write(&hmac, rng->v, 32);
secp256k1_hmac_sha256_write(&hmac, one, 1);
secp256k1_hmac_sha256_write(&hmac, key, keylen);
secp256k1_hmac_sha256_finalize(&hmac, rng->k);
secp256k1_hmac_sha256_initialize(&hmac, rng->k, 32);
secp256k1_hmac_sha256_write(&hmac, rng->v, 32);
secp256k1_hmac_sha256_finalize(&hmac, rng->v);
rng->retry = 0;
}
static void secp256k1_rfc6979_hmac_sha256_generate(secp256k1_rfc6979_hmac_sha256 *rng, unsigned char *out, size_t outlen) {
/* RFC6979 3.2.h. */
static const unsigned char zero[1] = {0x00};
if (rng->retry) {
secp256k1_hmac_sha256 hmac;
secp256k1_hmac_sha256_initialize(&hmac, rng->k, 32);
secp256k1_hmac_sha256_write(&hmac, rng->v, 32);
secp256k1_hmac_sha256_write(&hmac, zero, 1);
secp256k1_hmac_sha256_finalize(&hmac, rng->k);
secp256k1_hmac_sha256_initialize(&hmac, rng->k, 32);
secp256k1_hmac_sha256_write(&hmac, rng->v, 32);
secp256k1_hmac_sha256_finalize(&hmac, rng->v);
}
while (outlen > 0) {
secp256k1_hmac_sha256 hmac;
int now = outlen;
secp256k1_hmac_sha256_initialize(&hmac, rng->k, 32);
secp256k1_hmac_sha256_write(&hmac, rng->v, 32);
secp256k1_hmac_sha256_finalize(&hmac, rng->v);
if (now > 32) {
now = 32;
}
memcpy(out, rng->v, now);
out += now;
outlen -= now;
}
rng->retry = 1;
}
static void secp256k1_rfc6979_hmac_sha256_finalize(secp256k1_rfc6979_hmac_sha256 *rng) {
(void) rng;
}
static void secp256k1_rfc6979_hmac_sha256_clear(secp256k1_rfc6979_hmac_sha256 *rng) {
secp256k1_memclear(rng, sizeof(*rng));
}
#undef Round
#undef sigma1
#undef sigma0
#undef Sigma1
#undef Sigma0
#undef Maj
#undef Ch
#endif /* SECP256K1_HASH_IMPL_H */

33
libsecp256k1/src/hsort.h Normal file
View File

@@ -0,0 +1,33 @@
/***********************************************************************
* Copyright (c) 2021 Russell O'Connor, Jonas Nick *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_HSORT_H
#define SECP256K1_HSORT_H
#include <stddef.h>
#include <string.h>
/* In-place, iterative heapsort with an interface matching glibc's qsort_r. This
* is preferred over standard library implementations because they generally
* make no guarantee about being fast for malicious inputs.
* Remember that heapsort is unstable.
*
* In/Out: ptr: pointer to the array to sort. The contents of the array are
* sorted in ascending order according to the comparison function.
* In: count: number of elements in the array.
* size: size in bytes of each element.
* cmp: pointer to a comparison function that is called with two
* arguments that point to the objects being compared. The cmp_data
* argument of secp256k1_hsort is passed as third argument. The
* function must return an integer less than, equal to, or greater
* than zero if the first argument is considered to be respectively
* less than, equal to, or greater than the second.
* cmp_data: pointer passed as third argument to cmp.
*/
static void secp256k1_hsort(void *ptr, size_t count, size_t size,
int (*cmp)(const void *, const void *, void *),
void *cmp_data);
#endif

View File

@@ -0,0 +1,125 @@
/***********************************************************************
* Copyright (c) 2021 Russell O'Connor, Jonas Nick *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_HSORT_IMPL_H
#define SECP256K1_HSORT_IMPL_H
#include "hsort.h"
/* An array is a heap when, for all non-zero indexes i, the element at index i
* compares as less than or equal to the element at index parent(i) = (i-1)/2.
*/
static SECP256K1_INLINE size_t secp256k1_heap_child1(size_t i) {
VERIFY_CHECK(i <= (SIZE_MAX - 1)/2);
return 2*i + 1;
}
static SECP256K1_INLINE size_t secp256k1_heap_child2(size_t i) {
VERIFY_CHECK(i <= SIZE_MAX/2 - 1);
return secp256k1_heap_child1(i)+1;
}
static SECP256K1_INLINE void secp256k1_heap_swap64(unsigned char *a, unsigned char *b, size_t len) {
unsigned char tmp[64];
VERIFY_CHECK(len <= 64);
memcpy(tmp, a, len);
memmove(a, b, len);
memcpy(b, tmp, len);
}
static SECP256K1_INLINE void secp256k1_heap_swap(unsigned char *arr, size_t i, size_t j, size_t stride) {
unsigned char *a = arr + i*stride;
unsigned char *b = arr + j*stride;
size_t len = stride;
while (64 < len) {
secp256k1_heap_swap64(a + (len - 64), b + (len - 64), 64);
len -= 64;
}
secp256k1_heap_swap64(a, b, len);
}
/* This function accepts an array arr containing heap_size elements, each of
* size stride. The elements in the array at indices >i satisfy the max-heap
* property, i.e., for any element at index j (where j > i), all of its children
* are smaller than the element itself. The purpose of the function is to update
* the array so that all elements at indices >=i satisfy the max-heap
* property. */
static SECP256K1_INLINE void secp256k1_heap_down(unsigned char *arr, size_t i, size_t heap_size, size_t stride,
int (*cmp)(const void *, const void *, void *), void *cmp_data) {
while (i < heap_size/2) {
VERIFY_CHECK(i <= SIZE_MAX/2 - 1);
/* Proof:
* i < heap_size/2
* i + 1 <= heap_size/2
* 2*i + 2 <= heap_size <= SIZE_MAX
* 2*i <= SIZE_MAX - 2
*/
VERIFY_CHECK(secp256k1_heap_child1(i) < heap_size);
/* Proof:
* i < heap_size/2
* i + 1 <= heap_size/2
* 2*i + 2 <= heap_size
* 2*i + 1 < heap_size
* child1(i) < heap_size
*/
/* Let [x] be notation for the contents at arr[x*stride].
*
* If [child1(i)] > [i] and [child2(i)] > [i],
* swap [i] with the larger child to ensure the new parent is larger
* than both children. When [child1(i)] == [child2(i)], swap [i] with
* [child2(i)].
* Else if [child1(i)] > [i], swap [i] with [child1(i)].
* Else if [child2(i)] > [i], swap [i] with [child2(i)].
*/
if (secp256k1_heap_child2(i) < heap_size
&& 0 <= cmp(arr + secp256k1_heap_child2(i)*stride, arr + secp256k1_heap_child1(i)*stride, cmp_data)) {
if (0 < cmp(arr + secp256k1_heap_child2(i)*stride, arr + i*stride, cmp_data)) {
secp256k1_heap_swap(arr, i, secp256k1_heap_child2(i), stride);
i = secp256k1_heap_child2(i);
} else {
/* At this point we have [child2(i)] >= [child1(i)] and we have
* [child2(i)] <= [i], and thus [child1(i)] <= [i] which means
* that the next comparison can be skipped. */
return;
}
} else if (0 < cmp(arr + secp256k1_heap_child1(i)*stride, arr + i*stride, cmp_data)) {
secp256k1_heap_swap(arr, i, secp256k1_heap_child1(i), stride);
i = secp256k1_heap_child1(i);
} else {
return;
}
}
/* heap_size/2 <= i
* heap_size/2 < i + 1
* heap_size < 2*i + 2
* heap_size <= 2*i + 1
* heap_size <= child1(i)
* Thus child1(i) and child2(i) are now out of bounds and we are at a leaf.
*/
}
/* In-place heap sort. */
static void secp256k1_hsort(void *ptr, size_t count, size_t size,
int (*cmp)(const void *, const void *, void *),
void *cmp_data) {
size_t i;
for (i = count/2; 0 < i; --i) {
secp256k1_heap_down(ptr, i-1, count, size, cmp, cmp_data);
}
for (i = count; 1 < i; --i) {
/* Extract the largest value from the heap */
secp256k1_heap_swap(ptr, 0, i-1, size);
/* Repair the heap condition */
secp256k1_heap_down(ptr, 0, i-1, size, cmp, cmp_data);
}
}
#endif

90
libsecp256k1/src/int128.h Normal file
View File

@@ -0,0 +1,90 @@
#ifndef SECP256K1_INT128_H
#define SECP256K1_INT128_H
#include "util.h"
#if defined(SECP256K1_WIDEMUL_INT128)
# if defined(SECP256K1_INT128_NATIVE)
# include "int128_native.h"
# elif defined(SECP256K1_INT128_STRUCT)
# include "int128_struct.h"
# else
# error "Please select int128 implementation"
# endif
/* Construct an unsigned 128-bit value from a high and a low 64-bit value. */
static SECP256K1_INLINE void secp256k1_u128_load(secp256k1_uint128 *r, uint64_t hi, uint64_t lo);
/* Multiply two unsigned 64-bit values a and b and write the result to r. */
static SECP256K1_INLINE void secp256k1_u128_mul(secp256k1_uint128 *r, uint64_t a, uint64_t b);
/* Multiply two unsigned 64-bit values a and b and add the result to r.
* The final result is taken modulo 2^128.
*/
static SECP256K1_INLINE void secp256k1_u128_accum_mul(secp256k1_uint128 *r, uint64_t a, uint64_t b);
/* Add an unsigned 64-bit value a to r.
* The final result is taken modulo 2^128.
*/
static SECP256K1_INLINE void secp256k1_u128_accum_u64(secp256k1_uint128 *r, uint64_t a);
/* Unsigned (logical) right shift.
* Non-constant time in n.
*/
static SECP256K1_INLINE void secp256k1_u128_rshift(secp256k1_uint128 *r, unsigned int n);
/* Return the low 64-bits of a 128-bit value as an unsigned 64-bit value. */
static SECP256K1_INLINE uint64_t secp256k1_u128_to_u64(const secp256k1_uint128 *a);
/* Return the high 64-bits of a 128-bit value as an unsigned 64-bit value. */
static SECP256K1_INLINE uint64_t secp256k1_u128_hi_u64(const secp256k1_uint128 *a);
/* Write an unsigned 64-bit value to r. */
static SECP256K1_INLINE void secp256k1_u128_from_u64(secp256k1_uint128 *r, uint64_t a);
/* Tests if r is strictly less than to 2^n.
* n must be strictly less than 128.
*/
static SECP256K1_INLINE int secp256k1_u128_check_bits(const secp256k1_uint128 *r, unsigned int n);
/* Construct an signed 128-bit value from a high and a low 64-bit value. */
static SECP256K1_INLINE void secp256k1_i128_load(secp256k1_int128 *r, int64_t hi, uint64_t lo);
/* Multiply two signed 64-bit values a and b and write the result to r. */
static SECP256K1_INLINE void secp256k1_i128_mul(secp256k1_int128 *r, int64_t a, int64_t b);
/* Multiply two signed 64-bit values a and b and add the result to r.
* Overflow or underflow from the addition is undefined behaviour.
*/
static SECP256K1_INLINE void secp256k1_i128_accum_mul(secp256k1_int128 *r, int64_t a, int64_t b);
/* Compute a*d - b*c from signed 64-bit values and write the result to r. */
static SECP256K1_INLINE void secp256k1_i128_det(secp256k1_int128 *r, int64_t a, int64_t b, int64_t c, int64_t d);
/* Signed (arithmetic) right shift.
* Non-constant time in b.
*/
static SECP256K1_INLINE void secp256k1_i128_rshift(secp256k1_int128 *r, unsigned int b);
/* Return the input value modulo 2^64. */
static SECP256K1_INLINE uint64_t secp256k1_i128_to_u64(const secp256k1_int128 *a);
/* Return the value as a signed 64-bit value.
* Requires the input to be between INT64_MIN and INT64_MAX.
*/
static SECP256K1_INLINE int64_t secp256k1_i128_to_i64(const secp256k1_int128 *a);
/* Write a signed 64-bit value to r. */
static SECP256K1_INLINE void secp256k1_i128_from_i64(secp256k1_int128 *r, int64_t a);
/* Compare two 128-bit values for equality. */
static SECP256K1_INLINE int secp256k1_i128_eq_var(const secp256k1_int128 *a, const secp256k1_int128 *b);
/* Tests if r is equal to sign*2^n (sign must be 1 or -1).
* n must be strictly less than 127.
*/
static SECP256K1_INLINE int secp256k1_i128_check_pow2(const secp256k1_int128 *r, unsigned int n, int sign);
#endif
#endif

View File

@@ -0,0 +1,18 @@
#ifndef SECP256K1_INT128_IMPL_H
#define SECP256K1_INT128_IMPL_H
#include "util.h"
#include "int128.h"
#if defined(SECP256K1_WIDEMUL_INT128)
# if defined(SECP256K1_INT128_NATIVE)
# include "int128_native_impl.h"
# elif defined(SECP256K1_INT128_STRUCT)
# include "int128_struct_impl.h"
# else
# error "Please select int128 implementation"
# endif
#endif
#endif

View File

@@ -0,0 +1,19 @@
#ifndef SECP256K1_INT128_NATIVE_H
#define SECP256K1_INT128_NATIVE_H
#include <stdint.h>
#include "util.h"
#if !defined(UINT128_MAX) && defined(__SIZEOF_INT128__)
SECP256K1_GNUC_EXT typedef unsigned __int128 uint128_t;
SECP256K1_GNUC_EXT typedef __int128 int128_t;
# define UINT128_MAX ((uint128_t)(-1))
# define INT128_MAX ((int128_t)(UINT128_MAX >> 1))
# define INT128_MIN (-INT128_MAX - 1)
/* No (U)INT128_C macros because compilers providing __int128 do not support 128-bit literals. */
#endif
typedef uint128_t secp256k1_uint128;
typedef int128_t secp256k1_int128;
#endif

View File

@@ -0,0 +1,94 @@
#ifndef SECP256K1_INT128_NATIVE_IMPL_H
#define SECP256K1_INT128_NATIVE_IMPL_H
#include "int128.h"
#include "util.h"
static SECP256K1_INLINE void secp256k1_u128_load(secp256k1_uint128 *r, uint64_t hi, uint64_t lo) {
*r = (((uint128_t)hi) << 64) + lo;
}
static SECP256K1_INLINE void secp256k1_u128_mul(secp256k1_uint128 *r, uint64_t a, uint64_t b) {
*r = (uint128_t)a * b;
}
static SECP256K1_INLINE void secp256k1_u128_accum_mul(secp256k1_uint128 *r, uint64_t a, uint64_t b) {
*r += (uint128_t)a * b;
}
static SECP256K1_INLINE void secp256k1_u128_accum_u64(secp256k1_uint128 *r, uint64_t a) {
*r += a;
}
static SECP256K1_INLINE void secp256k1_u128_rshift(secp256k1_uint128 *r, unsigned int n) {
VERIFY_CHECK(n < 128);
*r >>= n;
}
static SECP256K1_INLINE uint64_t secp256k1_u128_to_u64(const secp256k1_uint128 *a) {
return (uint64_t)(*a);
}
static SECP256K1_INLINE uint64_t secp256k1_u128_hi_u64(const secp256k1_uint128 *a) {
return (uint64_t)(*a >> 64);
}
static SECP256K1_INLINE void secp256k1_u128_from_u64(secp256k1_uint128 *r, uint64_t a) {
*r = a;
}
static SECP256K1_INLINE int secp256k1_u128_check_bits(const secp256k1_uint128 *r, unsigned int n) {
VERIFY_CHECK(n < 128);
return (*r >> n == 0);
}
static SECP256K1_INLINE void secp256k1_i128_load(secp256k1_int128 *r, int64_t hi, uint64_t lo) {
*r = (((uint128_t)(uint64_t)hi) << 64) + lo;
}
static SECP256K1_INLINE void secp256k1_i128_mul(secp256k1_int128 *r, int64_t a, int64_t b) {
*r = (int128_t)a * b;
}
static SECP256K1_INLINE void secp256k1_i128_accum_mul(secp256k1_int128 *r, int64_t a, int64_t b) {
int128_t ab = (int128_t)a * b;
VERIFY_CHECK(0 <= ab ? *r <= INT128_MAX - ab : INT128_MIN - ab <= *r);
*r += ab;
}
static SECP256K1_INLINE void secp256k1_i128_det(secp256k1_int128 *r, int64_t a, int64_t b, int64_t c, int64_t d) {
int128_t ad = (int128_t)a * d;
int128_t bc = (int128_t)b * c;
VERIFY_CHECK(0 <= bc ? INT128_MIN + bc <= ad : ad <= INT128_MAX + bc);
*r = ad - bc;
}
static SECP256K1_INLINE void secp256k1_i128_rshift(secp256k1_int128 *r, unsigned int n) {
VERIFY_CHECK(n < 128);
*r >>= n;
}
static SECP256K1_INLINE uint64_t secp256k1_i128_to_u64(const secp256k1_int128 *a) {
return (uint64_t)*a;
}
static SECP256K1_INLINE int64_t secp256k1_i128_to_i64(const secp256k1_int128 *a) {
VERIFY_CHECK(INT64_MIN <= *a && *a <= INT64_MAX);
return *a;
}
static SECP256K1_INLINE void secp256k1_i128_from_i64(secp256k1_int128 *r, int64_t a) {
*r = a;
}
static SECP256K1_INLINE int secp256k1_i128_eq_var(const secp256k1_int128 *a, const secp256k1_int128 *b) {
return *a == *b;
}
static SECP256K1_INLINE int secp256k1_i128_check_pow2(const secp256k1_int128 *r, unsigned int n, int sign) {
VERIFY_CHECK(n < 127);
VERIFY_CHECK(sign == 1 || sign == -1);
return (*r == (int128_t)((uint128_t)sign << n));
}
#endif

View File

@@ -0,0 +1,14 @@
#ifndef SECP256K1_INT128_STRUCT_H
#define SECP256K1_INT128_STRUCT_H
#include <stdint.h>
#include "util.h"
typedef struct {
uint64_t lo;
uint64_t hi;
} secp256k1_uint128;
typedef secp256k1_uint128 secp256k1_int128;
#endif

View File

@@ -0,0 +1,205 @@
#ifndef SECP256K1_INT128_STRUCT_IMPL_H
#define SECP256K1_INT128_STRUCT_IMPL_H
#include "int128.h"
#include "util.h"
#if defined(_MSC_VER) && (defined(_M_X64) || defined(_M_ARM64)) /* MSVC */
# include <intrin.h>
# if defined(_M_ARM64) || defined(SECP256K1_MSVC_MULH_TEST_OVERRIDE)
/* On ARM64 MSVC, use __(u)mulh for the upper half of 64x64 multiplications.
(Define SECP256K1_MSVC_MULH_TEST_OVERRIDE to test this code path on X64,
which supports both __(u)mulh and _umul128.) */
# if defined(SECP256K1_MSVC_MULH_TEST_OVERRIDE)
# pragma message(__FILE__ ": SECP256K1_MSVC_MULH_TEST_OVERRIDE is defined, forcing use of __(u)mulh.")
# endif
static SECP256K1_INLINE uint64_t secp256k1_umul128(uint64_t a, uint64_t b, uint64_t* hi) {
*hi = __umulh(a, b);
return a * b;
}
static SECP256K1_INLINE int64_t secp256k1_mul128(int64_t a, int64_t b, int64_t* hi) {
*hi = __mulh(a, b);
return (uint64_t)a * (uint64_t)b;
}
# else
/* On x84_64 MSVC, use native _(u)mul128 for 64x64->128 multiplications. */
# define secp256k1_umul128 _umul128
# define secp256k1_mul128 _mul128
# endif
#else
/* On other systems, emulate 64x64->128 multiplications using 32x32->64 multiplications. */
static SECP256K1_INLINE uint64_t secp256k1_umul128(uint64_t a, uint64_t b, uint64_t* hi) {
uint64_t ll = (uint64_t)(uint32_t)a * (uint32_t)b;
uint64_t lh = (uint32_t)a * (b >> 32);
uint64_t hl = (a >> 32) * (uint32_t)b;
uint64_t hh = (a >> 32) * (b >> 32);
uint64_t mid34 = (ll >> 32) + (uint32_t)lh + (uint32_t)hl;
*hi = hh + (lh >> 32) + (hl >> 32) + (mid34 >> 32);
return (mid34 << 32) + (uint32_t)ll;
}
static SECP256K1_INLINE int64_t secp256k1_mul128(int64_t a, int64_t b, int64_t* hi) {
uint64_t ll = (uint64_t)(uint32_t)a * (uint32_t)b;
int64_t lh = (uint32_t)a * (b >> 32);
int64_t hl = (a >> 32) * (uint32_t)b;
int64_t hh = (a >> 32) * (b >> 32);
uint64_t mid34 = (ll >> 32) + (uint32_t)lh + (uint32_t)hl;
*hi = hh + (lh >> 32) + (hl >> 32) + (mid34 >> 32);
return (mid34 << 32) + (uint32_t)ll;
}
#endif
static SECP256K1_INLINE void secp256k1_u128_load(secp256k1_uint128 *r, uint64_t hi, uint64_t lo) {
r->hi = hi;
r->lo = lo;
}
static SECP256K1_INLINE void secp256k1_u128_mul(secp256k1_uint128 *r, uint64_t a, uint64_t b) {
r->lo = secp256k1_umul128(a, b, &r->hi);
}
static SECP256K1_INLINE void secp256k1_u128_accum_mul(secp256k1_uint128 *r, uint64_t a, uint64_t b) {
uint64_t lo, hi;
lo = secp256k1_umul128(a, b, &hi);
r->lo += lo;
r->hi += hi + (r->lo < lo);
}
static SECP256K1_INLINE void secp256k1_u128_accum_u64(secp256k1_uint128 *r, uint64_t a) {
r->lo += a;
r->hi += r->lo < a;
}
/* Unsigned (logical) right shift.
* Non-constant time in n.
*/
static SECP256K1_INLINE void secp256k1_u128_rshift(secp256k1_uint128 *r, unsigned int n) {
VERIFY_CHECK(n < 128);
if (n >= 64) {
r->lo = r->hi >> (n-64);
r->hi = 0;
} else if (n > 0) {
#if defined(_MSC_VER) && defined(_M_X64)
VERIFY_CHECK(n < 64);
r->lo = __shiftright128(r->lo, r->hi, n);
#else
r->lo = ((1U * r->hi) << (64-n)) | r->lo >> n;
#endif
r->hi >>= n;
}
}
static SECP256K1_INLINE uint64_t secp256k1_u128_to_u64(const secp256k1_uint128 *a) {
return a->lo;
}
static SECP256K1_INLINE uint64_t secp256k1_u128_hi_u64(const secp256k1_uint128 *a) {
return a->hi;
}
static SECP256K1_INLINE void secp256k1_u128_from_u64(secp256k1_uint128 *r, uint64_t a) {
r->hi = 0;
r->lo = a;
}
static SECP256K1_INLINE int secp256k1_u128_check_bits(const secp256k1_uint128 *r, unsigned int n) {
VERIFY_CHECK(n < 128);
return n >= 64 ? r->hi >> (n - 64) == 0
: r->hi == 0 && r->lo >> n == 0;
}
static SECP256K1_INLINE void secp256k1_i128_load(secp256k1_int128 *r, int64_t hi, uint64_t lo) {
r->hi = hi;
r->lo = lo;
}
static SECP256K1_INLINE void secp256k1_i128_mul(secp256k1_int128 *r, int64_t a, int64_t b) {
int64_t hi;
r->lo = (uint64_t)secp256k1_mul128(a, b, &hi);
r->hi = (uint64_t)hi;
}
static SECP256K1_INLINE void secp256k1_i128_accum_mul(secp256k1_int128 *r, int64_t a, int64_t b) {
int64_t hi;
uint64_t lo = (uint64_t)secp256k1_mul128(a, b, &hi);
r->lo += lo;
hi += r->lo < lo;
/* Verify no overflow.
* If r represents a positive value (the sign bit is not set) and the value we are adding is a positive value (the sign bit is not set),
* then we require that the resulting value also be positive (the sign bit is not set).
* Note that (X <= Y) means (X implies Y) when X and Y are boolean values (i.e. 0 or 1).
*/
VERIFY_CHECK((r->hi <= 0x7fffffffffffffffu && (uint64_t)hi <= 0x7fffffffffffffffu) <= (r->hi + (uint64_t)hi <= 0x7fffffffffffffffu));
/* Verify no underflow.
* If r represents a negative value (the sign bit is set) and the value we are adding is a negative value (the sign bit is set),
* then we require that the resulting value also be negative (the sign bit is set).
*/
VERIFY_CHECK((r->hi > 0x7fffffffffffffffu && (uint64_t)hi > 0x7fffffffffffffffu) <= (r->hi + (uint64_t)hi > 0x7fffffffffffffffu));
r->hi += hi;
}
static SECP256K1_INLINE void secp256k1_i128_dissip_mul(secp256k1_int128 *r, int64_t a, int64_t b) {
int64_t hi;
uint64_t lo = (uint64_t)secp256k1_mul128(a, b, &hi);
hi += r->lo < lo;
/* Verify no overflow.
* If r represents a positive value (the sign bit is not set) and the value we are subtracting is a negative value (the sign bit is set),
* then we require that the resulting value also be positive (the sign bit is not set).
*/
VERIFY_CHECK((r->hi <= 0x7fffffffffffffffu && (uint64_t)hi > 0x7fffffffffffffffu) <= (r->hi - (uint64_t)hi <= 0x7fffffffffffffffu));
/* Verify no underflow.
* If r represents a negative value (the sign bit is set) and the value we are subtracting is a positive value (the sign sign bit is not set),
* then we require that the resulting value also be negative (the sign bit is set).
*/
VERIFY_CHECK((r->hi > 0x7fffffffffffffffu && (uint64_t)hi <= 0x7fffffffffffffffu) <= (r->hi - (uint64_t)hi > 0x7fffffffffffffffu));
r->hi -= hi;
r->lo -= lo;
}
static SECP256K1_INLINE void secp256k1_i128_det(secp256k1_int128 *r, int64_t a, int64_t b, int64_t c, int64_t d) {
secp256k1_i128_mul(r, a, d);
secp256k1_i128_dissip_mul(r, b, c);
}
/* Signed (arithmetic) right shift.
* Non-constant time in n.
*/
static SECP256K1_INLINE void secp256k1_i128_rshift(secp256k1_int128 *r, unsigned int n) {
VERIFY_CHECK(n < 128);
if (n >= 64) {
r->lo = (uint64_t)((int64_t)(r->hi) >> (n-64));
r->hi = (uint64_t)((int64_t)(r->hi) >> 63);
} else if (n > 0) {
r->lo = ((1U * r->hi) << (64-n)) | r->lo >> n;
r->hi = (uint64_t)((int64_t)(r->hi) >> n);
}
}
static SECP256K1_INLINE uint64_t secp256k1_i128_to_u64(const secp256k1_int128 *a) {
return a->lo;
}
static SECP256K1_INLINE int64_t secp256k1_i128_to_i64(const secp256k1_int128 *a) {
/* Verify that a represents a 64 bit signed value by checking that the high bits are a sign extension of the low bits. */
VERIFY_CHECK(a->hi == -(a->lo >> 63));
return (int64_t)secp256k1_i128_to_u64(a);
}
static SECP256K1_INLINE void secp256k1_i128_from_i64(secp256k1_int128 *r, int64_t a) {
r->hi = (uint64_t)(a >> 63);
r->lo = (uint64_t)a;
}
static SECP256K1_INLINE int secp256k1_i128_eq_var(const secp256k1_int128 *a, const secp256k1_int128 *b) {
return a->hi == b->hi && a->lo == b->lo;
}
static SECP256K1_INLINE int secp256k1_i128_check_pow2(const secp256k1_int128 *r, unsigned int n, int sign) {
VERIFY_CHECK(n < 127);
VERIFY_CHECK(sign == 1 || sign == -1);
return n >= 64 ? r->hi == (uint64_t)sign << (n - 64) && r->lo == 0
: r->hi == (uint64_t)(sign >> 1) && r->lo == (uint64_t)sign << n;
}
#endif

View File

@@ -0,0 +1,43 @@
/***********************************************************************
* Copyright (c) 2020 Peter Dettman *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
**********************************************************************/
#ifndef SECP256K1_MODINV32_H
#define SECP256K1_MODINV32_H
#include "util.h"
/* A signed 30-bit limb representation of integers.
*
* Its value is sum(v[i] * 2^(30*i), i=0..8). */
typedef struct {
int32_t v[9];
} secp256k1_modinv32_signed30;
typedef struct {
/* The modulus in signed30 notation, must be odd and in [3, 2^256]. */
secp256k1_modinv32_signed30 modulus;
/* modulus^{-1} mod 2^30 */
uint32_t modulus_inv30;
} secp256k1_modinv32_modinfo;
/* Replace x with its modular inverse mod modinfo->modulus. x must be in range [0, modulus).
* If x is zero, the result will be zero as well. If not, the inverse must exist (i.e., the gcd of
* x and modulus must be 1). These rules are automatically satisfied if the modulus is prime.
*
* On output, all of x's limbs will be in [0, 2^30).
*/
static void secp256k1_modinv32_var(secp256k1_modinv32_signed30 *x, const secp256k1_modinv32_modinfo *modinfo);
/* Same as secp256k1_modinv32_var, but constant time in x (not in the modulus). */
static void secp256k1_modinv32(secp256k1_modinv32_signed30 *x, const secp256k1_modinv32_modinfo *modinfo);
/* Compute the Jacobi symbol for (x | modinfo->modulus). x must be coprime with modulus (and thus
* cannot be 0, as modulus >= 3). All limbs of x must be non-negative. Returns 0 if the result
* cannot be computed. */
static int secp256k1_jacobi32_maybe_var(const secp256k1_modinv32_signed30 *x, const secp256k1_modinv32_modinfo *modinfo);
#endif /* SECP256K1_MODINV32_H */

View File

@@ -0,0 +1,725 @@
/***********************************************************************
* Copyright (c) 2020 Peter Dettman *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
**********************************************************************/
#ifndef SECP256K1_MODINV32_IMPL_H
#define SECP256K1_MODINV32_IMPL_H
#include "modinv32.h"
#include "util.h"
#include <stdlib.h>
/* This file implements modular inversion based on the paper "Fast constant-time gcd computation and
* modular inversion" by Daniel J. Bernstein and Bo-Yin Yang.
*
* For an explanation of the algorithm, see doc/safegcd_implementation.md. This file contains an
* implementation for N=30, using 30-bit signed limbs represented as int32_t.
*/
#ifdef VERIFY
static const secp256k1_modinv32_signed30 SECP256K1_SIGNED30_ONE = {{1}};
/* Compute a*factor and put it in r. All but the top limb in r will be in range [0,2^30). */
static void secp256k1_modinv32_mul_30(secp256k1_modinv32_signed30 *r, const secp256k1_modinv32_signed30 *a, int alen, int32_t factor) {
const int32_t M30 = (int32_t)(UINT32_MAX >> 2);
int64_t c = 0;
int i;
for (i = 0; i < 8; ++i) {
if (i < alen) c += (int64_t)a->v[i] * factor;
r->v[i] = (int32_t)c & M30; c >>= 30;
}
if (8 < alen) c += (int64_t)a->v[8] * factor;
VERIFY_CHECK(c == (int32_t)c);
r->v[8] = (int32_t)c;
}
/* Return -1 for a<b*factor, 0 for a==b*factor, 1 for a>b*factor. A consists of alen limbs; b has 9. */
static int secp256k1_modinv32_mul_cmp_30(const secp256k1_modinv32_signed30 *a, int alen, const secp256k1_modinv32_signed30 *b, int32_t factor) {
int i;
secp256k1_modinv32_signed30 am, bm;
secp256k1_modinv32_mul_30(&am, a, alen, 1); /* Normalize all but the top limb of a. */
secp256k1_modinv32_mul_30(&bm, b, 9, factor);
for (i = 0; i < 8; ++i) {
/* Verify that all but the top limb of a and b are normalized. */
VERIFY_CHECK(am.v[i] >> 30 == 0);
VERIFY_CHECK(bm.v[i] >> 30 == 0);
}
for (i = 8; i >= 0; --i) {
if (am.v[i] < bm.v[i]) return -1;
if (am.v[i] > bm.v[i]) return 1;
}
return 0;
}
#endif
/* Take as input a signed30 number in range (-2*modulus,modulus), and add a multiple of the modulus
* to it to bring it to range [0,modulus). If sign < 0, the input will also be negated in the
* process. The input must have limbs in range (-2^30,2^30). The output will have limbs in range
* [0,2^30). */
static void secp256k1_modinv32_normalize_30(secp256k1_modinv32_signed30 *r, int32_t sign, const secp256k1_modinv32_modinfo *modinfo) {
const int32_t M30 = (int32_t)(UINT32_MAX >> 2);
int32_t r0 = r->v[0], r1 = r->v[1], r2 = r->v[2], r3 = r->v[3], r4 = r->v[4],
r5 = r->v[5], r6 = r->v[6], r7 = r->v[7], r8 = r->v[8];
volatile int32_t cond_add, cond_negate;
#ifdef VERIFY
/* Verify that all limbs are in range (-2^30,2^30). */
int i;
for (i = 0; i < 9; ++i) {
VERIFY_CHECK(r->v[i] >= -M30);
VERIFY_CHECK(r->v[i] <= M30);
}
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(r, 9, &modinfo->modulus, -2) > 0); /* r > -2*modulus */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(r, 9, &modinfo->modulus, 1) < 0); /* r < modulus */
#endif
/* In a first step, add the modulus if the input is negative, and then negate if requested.
* This brings r from range (-2*modulus,modulus) to range (-modulus,modulus). As all input
* limbs are in range (-2^30,2^30), this cannot overflow an int32_t. Note that the right
* shifts below are signed sign-extending shifts (see assumptions.h for tests that that is
* indeed the behavior of the right shift operator). */
cond_add = r8 >> 31;
r0 += modinfo->modulus.v[0] & cond_add;
r1 += modinfo->modulus.v[1] & cond_add;
r2 += modinfo->modulus.v[2] & cond_add;
r3 += modinfo->modulus.v[3] & cond_add;
r4 += modinfo->modulus.v[4] & cond_add;
r5 += modinfo->modulus.v[5] & cond_add;
r6 += modinfo->modulus.v[6] & cond_add;
r7 += modinfo->modulus.v[7] & cond_add;
r8 += modinfo->modulus.v[8] & cond_add;
cond_negate = sign >> 31;
r0 = (r0 ^ cond_negate) - cond_negate;
r1 = (r1 ^ cond_negate) - cond_negate;
r2 = (r2 ^ cond_negate) - cond_negate;
r3 = (r3 ^ cond_negate) - cond_negate;
r4 = (r4 ^ cond_negate) - cond_negate;
r5 = (r5 ^ cond_negate) - cond_negate;
r6 = (r6 ^ cond_negate) - cond_negate;
r7 = (r7 ^ cond_negate) - cond_negate;
r8 = (r8 ^ cond_negate) - cond_negate;
/* Propagate the top bits, to bring limbs back to range (-2^30,2^30). */
r1 += r0 >> 30; r0 &= M30;
r2 += r1 >> 30; r1 &= M30;
r3 += r2 >> 30; r2 &= M30;
r4 += r3 >> 30; r3 &= M30;
r5 += r4 >> 30; r4 &= M30;
r6 += r5 >> 30; r5 &= M30;
r7 += r6 >> 30; r6 &= M30;
r8 += r7 >> 30; r7 &= M30;
/* In a second step add the modulus again if the result is still negative, bringing r to range
* [0,modulus). */
cond_add = r8 >> 31;
r0 += modinfo->modulus.v[0] & cond_add;
r1 += modinfo->modulus.v[1] & cond_add;
r2 += modinfo->modulus.v[2] & cond_add;
r3 += modinfo->modulus.v[3] & cond_add;
r4 += modinfo->modulus.v[4] & cond_add;
r5 += modinfo->modulus.v[5] & cond_add;
r6 += modinfo->modulus.v[6] & cond_add;
r7 += modinfo->modulus.v[7] & cond_add;
r8 += modinfo->modulus.v[8] & cond_add;
/* And propagate again. */
r1 += r0 >> 30; r0 &= M30;
r2 += r1 >> 30; r1 &= M30;
r3 += r2 >> 30; r2 &= M30;
r4 += r3 >> 30; r3 &= M30;
r5 += r4 >> 30; r4 &= M30;
r6 += r5 >> 30; r5 &= M30;
r7 += r6 >> 30; r6 &= M30;
r8 += r7 >> 30; r7 &= M30;
r->v[0] = r0;
r->v[1] = r1;
r->v[2] = r2;
r->v[3] = r3;
r->v[4] = r4;
r->v[5] = r5;
r->v[6] = r6;
r->v[7] = r7;
r->v[8] = r8;
VERIFY_CHECK(r0 >> 30 == 0);
VERIFY_CHECK(r1 >> 30 == 0);
VERIFY_CHECK(r2 >> 30 == 0);
VERIFY_CHECK(r3 >> 30 == 0);
VERIFY_CHECK(r4 >> 30 == 0);
VERIFY_CHECK(r5 >> 30 == 0);
VERIFY_CHECK(r6 >> 30 == 0);
VERIFY_CHECK(r7 >> 30 == 0);
VERIFY_CHECK(r8 >> 30 == 0);
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(r, 9, &modinfo->modulus, 0) >= 0); /* r >= 0 */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(r, 9, &modinfo->modulus, 1) < 0); /* r < modulus */
}
/* Data type for transition matrices (see section 3 of explanation).
*
* t = [ u v ]
* [ q r ]
*/
typedef struct {
int32_t u, v, q, r;
} secp256k1_modinv32_trans2x2;
/* Compute the transition matrix and zeta for 30 divsteps.
*
* Input: zeta: initial zeta
* f0: bottom limb of initial f
* g0: bottom limb of initial g
* Output: t: transition matrix
* Return: final zeta
*
* Implements the divsteps_n_matrix function from the explanation.
*/
static int32_t secp256k1_modinv32_divsteps_30(int32_t zeta, uint32_t f0, uint32_t g0, secp256k1_modinv32_trans2x2 *t) {
/* u,v,q,r are the elements of the transformation matrix being built up,
* starting with the identity matrix. Semantically they are signed integers
* in range [-2^30,2^30], but here represented as unsigned mod 2^32. This
* permits left shifting (which is UB for negative numbers). The range
* being inside [-2^31,2^31) means that casting to signed works correctly.
*/
uint32_t u = 1, v = 0, q = 0, r = 1;
volatile uint32_t c1, c2;
uint32_t mask1, mask2, f = f0, g = g0, x, y, z;
int i;
for (i = 0; i < 30; ++i) {
VERIFY_CHECK((f & 1) == 1); /* f must always be odd */
VERIFY_CHECK((u * f0 + v * g0) == f << i);
VERIFY_CHECK((q * f0 + r * g0) == g << i);
/* Compute conditional masks for (zeta < 0) and for (g & 1). */
c1 = zeta >> 31;
mask1 = c1;
c2 = g & 1;
mask2 = -c2;
/* Compute x,y,z, conditionally negated versions of f,u,v. */
x = (f ^ mask1) - mask1;
y = (u ^ mask1) - mask1;
z = (v ^ mask1) - mask1;
/* Conditionally add x,y,z to g,q,r. */
g += x & mask2;
q += y & mask2;
r += z & mask2;
/* In what follows, mask1 is a condition mask for (zeta < 0) and (g & 1). */
mask1 &= mask2;
/* Conditionally change zeta into -zeta-2 or zeta-1. */
zeta = (zeta ^ mask1) - 1;
/* Conditionally add g,q,r to f,u,v. */
f += g & mask1;
u += q & mask1;
v += r & mask1;
/* Shifts */
g >>= 1;
u <<= 1;
v <<= 1;
/* Bounds on zeta that follow from the bounds on iteration count (max 20*30 divsteps). */
VERIFY_CHECK(zeta >= -601 && zeta <= 601);
}
/* Return data in t and return value. */
t->u = (int32_t)u;
t->v = (int32_t)v;
t->q = (int32_t)q;
t->r = (int32_t)r;
/* The determinant of t must be a power of two. This guarantees that multiplication with t
* does not change the gcd of f and g, apart from adding a power-of-2 factor to it (which
* will be divided out again). As each divstep's individual matrix has determinant 2, the
* aggregate of 30 of them will have determinant 2^30. */
VERIFY_CHECK((int64_t)t->u * t->r - (int64_t)t->v * t->q == ((int64_t)1) << 30);
return zeta;
}
/* secp256k1_modinv32_inv256[i] = -(2*i+1)^-1 (mod 256) */
static const uint8_t secp256k1_modinv32_inv256[128] = {
0xFF, 0x55, 0x33, 0x49, 0xC7, 0x5D, 0x3B, 0x11, 0x0F, 0xE5, 0xC3, 0x59,
0xD7, 0xED, 0xCB, 0x21, 0x1F, 0x75, 0x53, 0x69, 0xE7, 0x7D, 0x5B, 0x31,
0x2F, 0x05, 0xE3, 0x79, 0xF7, 0x0D, 0xEB, 0x41, 0x3F, 0x95, 0x73, 0x89,
0x07, 0x9D, 0x7B, 0x51, 0x4F, 0x25, 0x03, 0x99, 0x17, 0x2D, 0x0B, 0x61,
0x5F, 0xB5, 0x93, 0xA9, 0x27, 0xBD, 0x9B, 0x71, 0x6F, 0x45, 0x23, 0xB9,
0x37, 0x4D, 0x2B, 0x81, 0x7F, 0xD5, 0xB3, 0xC9, 0x47, 0xDD, 0xBB, 0x91,
0x8F, 0x65, 0x43, 0xD9, 0x57, 0x6D, 0x4B, 0xA1, 0x9F, 0xF5, 0xD3, 0xE9,
0x67, 0xFD, 0xDB, 0xB1, 0xAF, 0x85, 0x63, 0xF9, 0x77, 0x8D, 0x6B, 0xC1,
0xBF, 0x15, 0xF3, 0x09, 0x87, 0x1D, 0xFB, 0xD1, 0xCF, 0xA5, 0x83, 0x19,
0x97, 0xAD, 0x8B, 0xE1, 0xDF, 0x35, 0x13, 0x29, 0xA7, 0x3D, 0x1B, 0xF1,
0xEF, 0xC5, 0xA3, 0x39, 0xB7, 0xCD, 0xAB, 0x01
};
/* Compute the transition matrix and eta for 30 divsteps (variable time).
*
* Input: eta: initial eta
* f0: bottom limb of initial f
* g0: bottom limb of initial g
* Output: t: transition matrix
* Return: final eta
*
* Implements the divsteps_n_matrix_var function from the explanation.
*/
static int32_t secp256k1_modinv32_divsteps_30_var(int32_t eta, uint32_t f0, uint32_t g0, secp256k1_modinv32_trans2x2 *t) {
/* Transformation matrix; see comments in secp256k1_modinv32_divsteps_30. */
uint32_t u = 1, v = 0, q = 0, r = 1;
uint32_t f = f0, g = g0, m;
uint16_t w;
int i = 30, limit, zeros;
for (;;) {
/* Use a sentinel bit to count zeros only up to i. */
zeros = secp256k1_ctz32_var(g | (UINT32_MAX << i));
/* Perform zeros divsteps at once; they all just divide g by two. */
g >>= zeros;
u <<= zeros;
v <<= zeros;
eta -= zeros;
i -= zeros;
/* We're done once we've done 30 divsteps. */
if (i == 0) break;
VERIFY_CHECK((f & 1) == 1);
VERIFY_CHECK((g & 1) == 1);
VERIFY_CHECK((u * f0 + v * g0) == f << (30 - i));
VERIFY_CHECK((q * f0 + r * g0) == g << (30 - i));
/* Bounds on eta that follow from the bounds on iteration count (max 25*30 divsteps). */
VERIFY_CHECK(eta >= -751 && eta <= 751);
/* If eta is negative, negate it and replace f,g with g,-f. */
if (eta < 0) {
uint32_t tmp;
eta = -eta;
tmp = f; f = g; g = -tmp;
tmp = u; u = q; q = -tmp;
tmp = v; v = r; r = -tmp;
}
/* eta is now >= 0. In what follows we're going to cancel out the bottom bits of g. No more
* than i can be cancelled out (as we'd be done before that point), and no more than eta+1
* can be done as its sign will flip once that happens. */
limit = ((int)eta + 1) > i ? i : ((int)eta + 1);
/* m is a mask for the bottom min(limit, 8) bits (our table only supports 8 bits). */
VERIFY_CHECK(limit > 0 && limit <= 30);
m = (UINT32_MAX >> (32 - limit)) & 255U;
/* Find what multiple of f must be added to g to cancel its bottom min(limit, 8) bits. */
w = (g * secp256k1_modinv32_inv256[(f >> 1) & 127]) & m;
/* Do so. */
g += f * w;
q += u * w;
r += v * w;
VERIFY_CHECK((g & m) == 0);
}
/* Return data in t and return value. */
t->u = (int32_t)u;
t->v = (int32_t)v;
t->q = (int32_t)q;
t->r = (int32_t)r;
/* The determinant of t must be a power of two. This guarantees that multiplication with t
* does not change the gcd of f and g, apart from adding a power-of-2 factor to it (which
* will be divided out again). As each divstep's individual matrix has determinant 2, the
* aggregate of 30 of them will have determinant 2^30. */
VERIFY_CHECK((int64_t)t->u * t->r - (int64_t)t->v * t->q == ((int64_t)1) << 30);
return eta;
}
/* Compute the transition matrix and eta for 30 posdivsteps (variable time, eta=-delta), and keeps track
* of the Jacobi symbol along the way. f0 and g0 must be f and g mod 2^32 rather than 2^30, because
* Jacobi tracking requires knowing (f mod 8) rather than just (f mod 2).
*
* Input: eta: initial eta
* f0: bottom limb of initial f
* g0: bottom limb of initial g
* Output: t: transition matrix
* Input/Output: (*jacp & 1) is bitflipped if and only if the Jacobi symbol of (f | g) changes sign
* by applying the returned transformation matrix to it. The other bits of *jacp may
* change, but are meaningless.
* Return: final eta
*/
static int32_t secp256k1_modinv32_posdivsteps_30_var(int32_t eta, uint32_t f0, uint32_t g0, secp256k1_modinv32_trans2x2 *t, int *jacp) {
/* Transformation matrix. */
uint32_t u = 1, v = 0, q = 0, r = 1;
uint32_t f = f0, g = g0, m;
uint16_t w;
int i = 30, limit, zeros;
int jac = *jacp;
for (;;) {
/* Use a sentinel bit to count zeros only up to i. */
zeros = secp256k1_ctz32_var(g | (UINT32_MAX << i));
/* Perform zeros divsteps at once; they all just divide g by two. */
g >>= zeros;
u <<= zeros;
v <<= zeros;
eta -= zeros;
i -= zeros;
/* Update the bottom bit of jac: when dividing g by an odd power of 2,
* if (f mod 8) is 3 or 5, the Jacobi symbol changes sign. */
jac ^= (zeros & ((f >> 1) ^ (f >> 2)));
/* We're done once we've done 30 posdivsteps. */
if (i == 0) break;
VERIFY_CHECK((f & 1) == 1);
VERIFY_CHECK((g & 1) == 1);
VERIFY_CHECK((u * f0 + v * g0) == f << (30 - i));
VERIFY_CHECK((q * f0 + r * g0) == g << (30 - i));
/* If eta is negative, negate it and replace f,g with g,f. */
if (eta < 0) {
uint32_t tmp;
eta = -eta;
/* Update bottom bit of jac: when swapping f and g, the Jacobi symbol changes sign
* if both f and g are 3 mod 4. */
jac ^= ((f & g) >> 1);
tmp = f; f = g; g = tmp;
tmp = u; u = q; q = tmp;
tmp = v; v = r; r = tmp;
}
/* eta is now >= 0. In what follows we're going to cancel out the bottom bits of g. No more
* than i can be cancelled out (as we'd be done before that point), and no more than eta+1
* can be done as its sign will flip once that happens. */
limit = ((int)eta + 1) > i ? i : ((int)eta + 1);
/* m is a mask for the bottom min(limit, 8) bits (our table only supports 8 bits). */
VERIFY_CHECK(limit > 0 && limit <= 30);
m = (UINT32_MAX >> (32 - limit)) & 255U;
/* Find what multiple of f must be added to g to cancel its bottom min(limit, 8) bits. */
w = (g * secp256k1_modinv32_inv256[(f >> 1) & 127]) & m;
/* Do so. */
g += f * w;
q += u * w;
r += v * w;
VERIFY_CHECK((g & m) == 0);
}
/* Return data in t and return value. */
t->u = (int32_t)u;
t->v = (int32_t)v;
t->q = (int32_t)q;
t->r = (int32_t)r;
/* The determinant of t must be a power of two. This guarantees that multiplication with t
* does not change the gcd of f and g, apart from adding a power-of-2 factor to it (which
* will be divided out again). As each divstep's individual matrix has determinant 2 or -2,
* the aggregate of 30 of them will have determinant 2^30 or -2^30. */
VERIFY_CHECK((int64_t)t->u * t->r - (int64_t)t->v * t->q == ((int64_t)1) << 30 ||
(int64_t)t->u * t->r - (int64_t)t->v * t->q == -(((int64_t)1) << 30));
*jacp = jac;
return eta;
}
/* Compute (t/2^30) * [d, e] mod modulus, where t is a transition matrix for 30 divsteps.
*
* On input and output, d and e are in range (-2*modulus,modulus). All output limbs will be in range
* (-2^30,2^30).
*
* This implements the update_de function from the explanation.
*/
static void secp256k1_modinv32_update_de_30(secp256k1_modinv32_signed30 *d, secp256k1_modinv32_signed30 *e, const secp256k1_modinv32_trans2x2 *t, const secp256k1_modinv32_modinfo* modinfo) {
const int32_t M30 = (int32_t)(UINT32_MAX >> 2);
const int32_t u = t->u, v = t->v, q = t->q, r = t->r;
int32_t di, ei, md, me, sd, se;
int64_t cd, ce;
int i;
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(d, 9, &modinfo->modulus, -2) > 0); /* d > -2*modulus */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(d, 9, &modinfo->modulus, 1) < 0); /* d < modulus */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(e, 9, &modinfo->modulus, -2) > 0); /* e > -2*modulus */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(e, 9, &modinfo->modulus, 1) < 0); /* e < modulus */
VERIFY_CHECK(labs(u) <= (M30 + 1 - labs(v))); /* |u|+|v| <= 2^30 */
VERIFY_CHECK(labs(q) <= (M30 + 1 - labs(r))); /* |q|+|r| <= 2^30 */
/* [md,me] start as zero; plus [u,q] if d is negative; plus [v,r] if e is negative. */
sd = d->v[8] >> 31;
se = e->v[8] >> 31;
md = (u & sd) + (v & se);
me = (q & sd) + (r & se);
/* Begin computing t*[d,e]. */
di = d->v[0];
ei = e->v[0];
cd = (int64_t)u * di + (int64_t)v * ei;
ce = (int64_t)q * di + (int64_t)r * ei;
/* Correct md,me so that t*[d,e]+modulus*[md,me] has 30 zero bottom bits. */
md -= (modinfo->modulus_inv30 * (uint32_t)cd + md) & M30;
me -= (modinfo->modulus_inv30 * (uint32_t)ce + me) & M30;
/* Update the beginning of computation for t*[d,e]+modulus*[md,me] now md,me are known. */
cd += (int64_t)modinfo->modulus.v[0] * md;
ce += (int64_t)modinfo->modulus.v[0] * me;
/* Verify that the low 30 bits of the computation are indeed zero, and then throw them away. */
VERIFY_CHECK(((int32_t)cd & M30) == 0); cd >>= 30;
VERIFY_CHECK(((int32_t)ce & M30) == 0); ce >>= 30;
/* Now iteratively compute limb i=1..8 of t*[d,e]+modulus*[md,me], and store them in output
* limb i-1 (shifting down by 30 bits). */
for (i = 1; i < 9; ++i) {
di = d->v[i];
ei = e->v[i];
cd += (int64_t)u * di + (int64_t)v * ei;
ce += (int64_t)q * di + (int64_t)r * ei;
cd += (int64_t)modinfo->modulus.v[i] * md;
ce += (int64_t)modinfo->modulus.v[i] * me;
d->v[i - 1] = (int32_t)cd & M30; cd >>= 30;
e->v[i - 1] = (int32_t)ce & M30; ce >>= 30;
}
/* What remains is limb 9 of t*[d,e]+modulus*[md,me]; store it as output limb 8. */
d->v[8] = (int32_t)cd;
e->v[8] = (int32_t)ce;
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(d, 9, &modinfo->modulus, -2) > 0); /* d > -2*modulus */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(d, 9, &modinfo->modulus, 1) < 0); /* d < modulus */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(e, 9, &modinfo->modulus, -2) > 0); /* e > -2*modulus */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(e, 9, &modinfo->modulus, 1) < 0); /* e < modulus */
}
/* Compute (t/2^30) * [f, g], where t is a transition matrix for 30 divsteps.
*
* This implements the update_fg function from the explanation.
*/
static void secp256k1_modinv32_update_fg_30(secp256k1_modinv32_signed30 *f, secp256k1_modinv32_signed30 *g, const secp256k1_modinv32_trans2x2 *t) {
const int32_t M30 = (int32_t)(UINT32_MAX >> 2);
const int32_t u = t->u, v = t->v, q = t->q, r = t->r;
int32_t fi, gi;
int64_t cf, cg;
int i;
/* Start computing t*[f,g]. */
fi = f->v[0];
gi = g->v[0];
cf = (int64_t)u * fi + (int64_t)v * gi;
cg = (int64_t)q * fi + (int64_t)r * gi;
/* Verify that the bottom 30 bits of the result are zero, and then throw them away. */
VERIFY_CHECK(((int32_t)cf & M30) == 0); cf >>= 30;
VERIFY_CHECK(((int32_t)cg & M30) == 0); cg >>= 30;
/* Now iteratively compute limb i=1..8 of t*[f,g], and store them in output limb i-1 (shifting
* down by 30 bits). */
for (i = 1; i < 9; ++i) {
fi = f->v[i];
gi = g->v[i];
cf += (int64_t)u * fi + (int64_t)v * gi;
cg += (int64_t)q * fi + (int64_t)r * gi;
f->v[i - 1] = (int32_t)cf & M30; cf >>= 30;
g->v[i - 1] = (int32_t)cg & M30; cg >>= 30;
}
/* What remains is limb 9 of t*[f,g]; store it as output limb 8. */
f->v[8] = (int32_t)cf;
g->v[8] = (int32_t)cg;
}
/* Compute (t/2^30) * [f, g], where t is a transition matrix for 30 divsteps.
*
* Version that operates on a variable number of limbs in f and g.
*
* This implements the update_fg function from the explanation in modinv64_impl.h.
*/
static void secp256k1_modinv32_update_fg_30_var(int len, secp256k1_modinv32_signed30 *f, secp256k1_modinv32_signed30 *g, const secp256k1_modinv32_trans2x2 *t) {
const int32_t M30 = (int32_t)(UINT32_MAX >> 2);
const int32_t u = t->u, v = t->v, q = t->q, r = t->r;
int32_t fi, gi;
int64_t cf, cg;
int i;
VERIFY_CHECK(len > 0);
/* Start computing t*[f,g]. */
fi = f->v[0];
gi = g->v[0];
cf = (int64_t)u * fi + (int64_t)v * gi;
cg = (int64_t)q * fi + (int64_t)r * gi;
/* Verify that the bottom 62 bits of the result are zero, and then throw them away. */
VERIFY_CHECK(((int32_t)cf & M30) == 0); cf >>= 30;
VERIFY_CHECK(((int32_t)cg & M30) == 0); cg >>= 30;
/* Now iteratively compute limb i=1..len of t*[f,g], and store them in output limb i-1 (shifting
* down by 30 bits). */
for (i = 1; i < len; ++i) {
fi = f->v[i];
gi = g->v[i];
cf += (int64_t)u * fi + (int64_t)v * gi;
cg += (int64_t)q * fi + (int64_t)r * gi;
f->v[i - 1] = (int32_t)cf & M30; cf >>= 30;
g->v[i - 1] = (int32_t)cg & M30; cg >>= 30;
}
/* What remains is limb (len) of t*[f,g]; store it as output limb (len-1). */
f->v[len - 1] = (int32_t)cf;
g->v[len - 1] = (int32_t)cg;
}
/* Compute the inverse of x modulo modinfo->modulus, and replace x with it (constant time in x). */
static void secp256k1_modinv32(secp256k1_modinv32_signed30 *x, const secp256k1_modinv32_modinfo *modinfo) {
/* Start with d=0, e=1, f=modulus, g=x, zeta=-1. */
secp256k1_modinv32_signed30 d = {{0}};
secp256k1_modinv32_signed30 e = {{1}};
secp256k1_modinv32_signed30 f = modinfo->modulus;
secp256k1_modinv32_signed30 g = *x;
int i;
int32_t zeta = -1; /* zeta = -(delta+1/2); delta is initially 1/2. */
/* Do 20 iterations of 30 divsteps each = 600 divsteps. 590 suffices for 256-bit inputs. */
for (i = 0; i < 20; ++i) {
/* Compute transition matrix and new zeta after 30 divsteps. */
secp256k1_modinv32_trans2x2 t;
zeta = secp256k1_modinv32_divsteps_30(zeta, f.v[0], g.v[0], &t);
/* Update d,e using that transition matrix. */
secp256k1_modinv32_update_de_30(&d, &e, &t, modinfo);
/* Update f,g using that transition matrix. */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&f, 9, &modinfo->modulus, -1) > 0); /* f > -modulus */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&f, 9, &modinfo->modulus, 1) <= 0); /* f <= modulus */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&g, 9, &modinfo->modulus, -1) > 0); /* g > -modulus */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&g, 9, &modinfo->modulus, 1) < 0); /* g < modulus */
secp256k1_modinv32_update_fg_30(&f, &g, &t);
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&f, 9, &modinfo->modulus, -1) > 0); /* f > -modulus */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&f, 9, &modinfo->modulus, 1) <= 0); /* f <= modulus */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&g, 9, &modinfo->modulus, -1) > 0); /* g > -modulus */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&g, 9, &modinfo->modulus, 1) < 0); /* g < modulus */
}
/* At this point sufficient iterations have been performed that g must have reached 0
* and (if g was not originally 0) f must now equal +/- GCD of the initial f, g
* values i.e. +/- 1, and d now contains +/- the modular inverse. */
/* g == 0 */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&g, 9, &SECP256K1_SIGNED30_ONE, 0) == 0);
/* |f| == 1, or (x == 0 and d == 0 and f == modulus) */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&f, 9, &SECP256K1_SIGNED30_ONE, -1) == 0 ||
secp256k1_modinv32_mul_cmp_30(&f, 9, &SECP256K1_SIGNED30_ONE, 1) == 0 ||
(secp256k1_modinv32_mul_cmp_30(x, 9, &SECP256K1_SIGNED30_ONE, 0) == 0 &&
secp256k1_modinv32_mul_cmp_30(&d, 9, &SECP256K1_SIGNED30_ONE, 0) == 0 &&
secp256k1_modinv32_mul_cmp_30(&f, 9, &modinfo->modulus, 1) == 0));
/* Optionally negate d, normalize to [0,modulus), and return it. */
secp256k1_modinv32_normalize_30(&d, f.v[8], modinfo);
*x = d;
}
/* Compute the inverse of x modulo modinfo->modulus, and replace x with it (variable time). */
static void secp256k1_modinv32_var(secp256k1_modinv32_signed30 *x, const secp256k1_modinv32_modinfo *modinfo) {
/* Start with d=0, e=1, f=modulus, g=x, eta=-1. */
secp256k1_modinv32_signed30 d = {{0, 0, 0, 0, 0, 0, 0, 0, 0}};
secp256k1_modinv32_signed30 e = {{1, 0, 0, 0, 0, 0, 0, 0, 0}};
secp256k1_modinv32_signed30 f = modinfo->modulus;
secp256k1_modinv32_signed30 g = *x;
#ifdef VERIFY
int i = 0;
#endif
int j, len = 9;
int32_t eta = -1; /* eta = -delta; delta is initially 1 (faster for the variable-time code) */
int32_t cond, fn, gn;
/* Do iterations of 30 divsteps each until g=0. */
while (1) {
/* Compute transition matrix and new eta after 30 divsteps. */
secp256k1_modinv32_trans2x2 t;
eta = secp256k1_modinv32_divsteps_30_var(eta, f.v[0], g.v[0], &t);
/* Update d,e using that transition matrix. */
secp256k1_modinv32_update_de_30(&d, &e, &t, modinfo);
/* Update f,g using that transition matrix. */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&f, len, &modinfo->modulus, -1) > 0); /* f > -modulus */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&f, len, &modinfo->modulus, 1) <= 0); /* f <= modulus */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&g, len, &modinfo->modulus, -1) > 0); /* g > -modulus */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&g, len, &modinfo->modulus, 1) < 0); /* g < modulus */
secp256k1_modinv32_update_fg_30_var(len, &f, &g, &t);
/* If the bottom limb of g is 0, there is a chance g=0. */
if (g.v[0] == 0) {
cond = 0;
/* Check if all other limbs are also 0. */
for (j = 1; j < len; ++j) {
cond |= g.v[j];
}
/* If so, we're done. */
if (cond == 0) break;
}
/* Determine if len>1 and limb (len-1) of both f and g is 0 or -1. */
fn = f.v[len - 1];
gn = g.v[len - 1];
cond = ((int32_t)len - 2) >> 31;
cond |= fn ^ (fn >> 31);
cond |= gn ^ (gn >> 31);
/* If so, reduce length, propagating the sign of f and g's top limb into the one below. */
if (cond == 0) {
f.v[len - 2] |= (uint32_t)fn << 30;
g.v[len - 2] |= (uint32_t)gn << 30;
--len;
}
VERIFY_CHECK(++i < 25); /* We should never need more than 25*30 = 750 divsteps */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&f, len, &modinfo->modulus, -1) > 0); /* f > -modulus */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&f, len, &modinfo->modulus, 1) <= 0); /* f <= modulus */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&g, len, &modinfo->modulus, -1) > 0); /* g > -modulus */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&g, len, &modinfo->modulus, 1) < 0); /* g < modulus */
}
/* At this point g is 0 and (if g was not originally 0) f must now equal +/- GCD of
* the initial f, g values i.e. +/- 1, and d now contains +/- the modular inverse. */
/* g == 0 */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&g, len, &SECP256K1_SIGNED30_ONE, 0) == 0);
/* |f| == 1, or (x == 0 and d == 0 and f == modulus) */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&f, len, &SECP256K1_SIGNED30_ONE, -1) == 0 ||
secp256k1_modinv32_mul_cmp_30(&f, len, &SECP256K1_SIGNED30_ONE, 1) == 0 ||
(secp256k1_modinv32_mul_cmp_30(x, 9, &SECP256K1_SIGNED30_ONE, 0) == 0 &&
secp256k1_modinv32_mul_cmp_30(&d, 9, &SECP256K1_SIGNED30_ONE, 0) == 0 &&
secp256k1_modinv32_mul_cmp_30(&f, len, &modinfo->modulus, 1) == 0));
/* Optionally negate d, normalize to [0,modulus), and return it. */
secp256k1_modinv32_normalize_30(&d, f.v[len - 1], modinfo);
*x = d;
}
/* Do up to 50 iterations of 30 posdivsteps (up to 1500 steps; more is extremely rare) each until f=1.
* In VERIFY mode use a lower number of iterations (750, close to the median 756), so failure actually occurs. */
#ifdef VERIFY
#define JACOBI32_ITERATIONS 25
#else
#define JACOBI32_ITERATIONS 50
#endif
/* Compute the Jacobi symbol of x modulo modinfo->modulus (variable time). gcd(x,modulus) must be 1. */
static int secp256k1_jacobi32_maybe_var(const secp256k1_modinv32_signed30 *x, const secp256k1_modinv32_modinfo *modinfo) {
/* Start with f=modulus, g=x, eta=-1. */
secp256k1_modinv32_signed30 f = modinfo->modulus;
secp256k1_modinv32_signed30 g = *x;
int j, len = 9;
int32_t eta = -1; /* eta = -delta; delta is initially 1 */
int32_t cond, fn, gn;
int jac = 0;
int count;
/* The input limbs must all be non-negative. */
VERIFY_CHECK(g.v[0] >= 0 && g.v[1] >= 0 && g.v[2] >= 0 && g.v[3] >= 0 && g.v[4] >= 0 && g.v[5] >= 0 && g.v[6] >= 0 && g.v[7] >= 0 && g.v[8] >= 0);
/* If x > 0, then if the loop below converges, it converges to f=g=gcd(x,modulus). Since we
* require that gcd(x,modulus)=1 and modulus>=3, x cannot be 0. Thus, we must reach f=1 (or
* time out). */
VERIFY_CHECK((g.v[0] | g.v[1] | g.v[2] | g.v[3] | g.v[4] | g.v[5] | g.v[6] | g.v[7] | g.v[8]) != 0);
for (count = 0; count < JACOBI32_ITERATIONS; ++count) {
/* Compute transition matrix and new eta after 30 posdivsteps. */
secp256k1_modinv32_trans2x2 t;
eta = secp256k1_modinv32_posdivsteps_30_var(eta, f.v[0] | ((uint32_t)f.v[1] << 30), g.v[0] | ((uint32_t)g.v[1] << 30), &t, &jac);
/* Update f,g using that transition matrix. */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&f, len, &modinfo->modulus, 0) > 0); /* f > 0 */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&f, len, &modinfo->modulus, 1) <= 0); /* f <= modulus */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&g, len, &modinfo->modulus, 0) > 0); /* g > 0 */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&g, len, &modinfo->modulus, 1) < 0); /* g < modulus */
secp256k1_modinv32_update_fg_30_var(len, &f, &g, &t);
/* If the bottom limb of f is 1, there is a chance that f=1. */
if (f.v[0] == 1) {
cond = 0;
/* Check if the other limbs are also 0. */
for (j = 1; j < len; ++j) {
cond |= f.v[j];
}
/* If so, we're done. If f=1, the Jacobi symbol (g | f)=1. */
if (cond == 0) return 1 - 2*(jac & 1);
}
/* Determine if len>1 and limb (len-1) of both f and g is 0. */
fn = f.v[len - 1];
gn = g.v[len - 1];
cond = ((int32_t)len - 2) >> 31;
cond |= fn;
cond |= gn;
/* If so, reduce length. */
if (cond == 0) --len;
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&f, len, &modinfo->modulus, 0) > 0); /* f > 0 */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&f, len, &modinfo->modulus, 1) <= 0); /* f <= modulus */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&g, len, &modinfo->modulus, 0) > 0); /* g > 0 */
VERIFY_CHECK(secp256k1_modinv32_mul_cmp_30(&g, len, &modinfo->modulus, 1) < 0); /* g < modulus */
}
/* The loop failed to converge to f=g after 1500 iterations. Return 0, indicating unknown result. */
return 0;
}
#endif /* SECP256K1_MODINV32_IMPL_H */

View File

@@ -0,0 +1,47 @@
/***********************************************************************
* Copyright (c) 2020 Peter Dettman *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
**********************************************************************/
#ifndef SECP256K1_MODINV64_H
#define SECP256K1_MODINV64_H
#include "util.h"
#ifndef SECP256K1_WIDEMUL_INT128
#error "modinv64 requires 128-bit wide multiplication support"
#endif
/* A signed 62-bit limb representation of integers.
*
* Its value is sum(v[i] * 2^(62*i), i=0..4). */
typedef struct {
int64_t v[5];
} secp256k1_modinv64_signed62;
typedef struct {
/* The modulus in signed62 notation, must be odd and in [3, 2^256]. */
secp256k1_modinv64_signed62 modulus;
/* modulus^{-1} mod 2^62 */
uint64_t modulus_inv62;
} secp256k1_modinv64_modinfo;
/* Replace x with its modular inverse mod modinfo->modulus. x must be in range [0, modulus).
* If x is zero, the result will be zero as well. If not, the inverse must exist (i.e., the gcd of
* x and modulus must be 1). These rules are automatically satisfied if the modulus is prime.
*
* On output, all of x's limbs will be in [0, 2^62).
*/
static void secp256k1_modinv64_var(secp256k1_modinv64_signed62 *x, const secp256k1_modinv64_modinfo *modinfo);
/* Same as secp256k1_modinv64_var, but constant time in x (not in the modulus). */
static void secp256k1_modinv64(secp256k1_modinv64_signed62 *x, const secp256k1_modinv64_modinfo *modinfo);
/* Compute the Jacobi symbol for (x | modinfo->modulus). x must be coprime with modulus (and thus
* cannot be 0, as modulus >= 3). All limbs of x must be non-negative. Returns 0 if the result
* cannot be computed. */
static int secp256k1_jacobi64_maybe_var(const secp256k1_modinv64_signed62 *x, const secp256k1_modinv64_modinfo *modinfo);
#endif /* SECP256K1_MODINV64_H */

View File

@@ -0,0 +1,780 @@
/***********************************************************************
* Copyright (c) 2020 Peter Dettman *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
**********************************************************************/
#ifndef SECP256K1_MODINV64_IMPL_H
#define SECP256K1_MODINV64_IMPL_H
#include "int128.h"
#include "modinv64.h"
/* This file implements modular inversion based on the paper "Fast constant-time gcd computation and
* modular inversion" by Daniel J. Bernstein and Bo-Yin Yang.
*
* For an explanation of the algorithm, see doc/safegcd_implementation.md. This file contains an
* implementation for N=62, using 62-bit signed limbs represented as int64_t.
*/
/* Data type for transition matrices (see section 3 of explanation).
*
* t = [ u v ]
* [ q r ]
*/
typedef struct {
int64_t u, v, q, r;
} secp256k1_modinv64_trans2x2;
#ifdef VERIFY
/* Helper function to compute the absolute value of an int64_t.
* (we don't use abs/labs/llabs as it depends on the int sizes). */
static int64_t secp256k1_modinv64_abs(int64_t v) {
VERIFY_CHECK(v > INT64_MIN);
if (v < 0) return -v;
return v;
}
static const secp256k1_modinv64_signed62 SECP256K1_SIGNED62_ONE = {{1}};
/* Compute a*factor and put it in r. All but the top limb in r will be in range [0,2^62). */
static void secp256k1_modinv64_mul_62(secp256k1_modinv64_signed62 *r, const secp256k1_modinv64_signed62 *a, int alen, int64_t factor) {
const uint64_t M62 = UINT64_MAX >> 2;
secp256k1_int128 c, d;
int i;
secp256k1_i128_from_i64(&c, 0);
for (i = 0; i < 4; ++i) {
if (i < alen) secp256k1_i128_accum_mul(&c, a->v[i], factor);
r->v[i] = secp256k1_i128_to_u64(&c) & M62; secp256k1_i128_rshift(&c, 62);
}
if (4 < alen) secp256k1_i128_accum_mul(&c, a->v[4], factor);
secp256k1_i128_from_i64(&d, secp256k1_i128_to_i64(&c));
VERIFY_CHECK(secp256k1_i128_eq_var(&c, &d));
r->v[4] = secp256k1_i128_to_i64(&c);
}
/* Return -1 for a<b*factor, 0 for a==b*factor, 1 for a>b*factor. A has alen limbs; b has 5. */
static int secp256k1_modinv64_mul_cmp_62(const secp256k1_modinv64_signed62 *a, int alen, const secp256k1_modinv64_signed62 *b, int64_t factor) {
int i;
secp256k1_modinv64_signed62 am, bm;
secp256k1_modinv64_mul_62(&am, a, alen, 1); /* Normalize all but the top limb of a. */
secp256k1_modinv64_mul_62(&bm, b, 5, factor);
for (i = 0; i < 4; ++i) {
/* Verify that all but the top limb of a and b are normalized. */
VERIFY_CHECK(am.v[i] >> 62 == 0);
VERIFY_CHECK(bm.v[i] >> 62 == 0);
}
for (i = 4; i >= 0; --i) {
if (am.v[i] < bm.v[i]) return -1;
if (am.v[i] > bm.v[i]) return 1;
}
return 0;
}
/* Check if the determinant of t is equal to 1 << n. If abs, check if |det t| == 1 << n. */
static int secp256k1_modinv64_det_check_pow2(const secp256k1_modinv64_trans2x2 *t, unsigned int n, int abs) {
secp256k1_int128 a;
secp256k1_i128_det(&a, t->u, t->v, t->q, t->r);
if (secp256k1_i128_check_pow2(&a, n, 1)) return 1;
if (abs && secp256k1_i128_check_pow2(&a, n, -1)) return 1;
return 0;
}
#endif
/* Take as input a signed62 number in range (-2*modulus,modulus), and add a multiple of the modulus
* to it to bring it to range [0,modulus). If sign < 0, the input will also be negated in the
* process. The input must have limbs in range (-2^62,2^62). The output will have limbs in range
* [0,2^62). */
static void secp256k1_modinv64_normalize_62(secp256k1_modinv64_signed62 *r, int64_t sign, const secp256k1_modinv64_modinfo *modinfo) {
const int64_t M62 = (int64_t)(UINT64_MAX >> 2);
int64_t r0 = r->v[0], r1 = r->v[1], r2 = r->v[2], r3 = r->v[3], r4 = r->v[4];
volatile int64_t cond_add, cond_negate;
#ifdef VERIFY
/* Verify that all limbs are in range (-2^62,2^62). */
int i;
for (i = 0; i < 5; ++i) {
VERIFY_CHECK(r->v[i] >= -M62);
VERIFY_CHECK(r->v[i] <= M62);
}
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(r, 5, &modinfo->modulus, -2) > 0); /* r > -2*modulus */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(r, 5, &modinfo->modulus, 1) < 0); /* r < modulus */
#endif
/* In a first step, add the modulus if the input is negative, and then negate if requested.
* This brings r from range (-2*modulus,modulus) to range (-modulus,modulus). As all input
* limbs are in range (-2^62,2^62), this cannot overflow an int64_t. Note that the right
* shifts below are signed sign-extending shifts (see assumptions.h for tests that that is
* indeed the behavior of the right shift operator). */
cond_add = r4 >> 63;
r0 += modinfo->modulus.v[0] & cond_add;
r1 += modinfo->modulus.v[1] & cond_add;
r2 += modinfo->modulus.v[2] & cond_add;
r3 += modinfo->modulus.v[3] & cond_add;
r4 += modinfo->modulus.v[4] & cond_add;
cond_negate = sign >> 63;
r0 = (r0 ^ cond_negate) - cond_negate;
r1 = (r1 ^ cond_negate) - cond_negate;
r2 = (r2 ^ cond_negate) - cond_negate;
r3 = (r3 ^ cond_negate) - cond_negate;
r4 = (r4 ^ cond_negate) - cond_negate;
/* Propagate the top bits, to bring limbs back to range (-2^62,2^62). */
r1 += r0 >> 62; r0 &= M62;
r2 += r1 >> 62; r1 &= M62;
r3 += r2 >> 62; r2 &= M62;
r4 += r3 >> 62; r3 &= M62;
/* In a second step add the modulus again if the result is still negative, bringing
* r to range [0,modulus). */
cond_add = r4 >> 63;
r0 += modinfo->modulus.v[0] & cond_add;
r1 += modinfo->modulus.v[1] & cond_add;
r2 += modinfo->modulus.v[2] & cond_add;
r3 += modinfo->modulus.v[3] & cond_add;
r4 += modinfo->modulus.v[4] & cond_add;
/* And propagate again. */
r1 += r0 >> 62; r0 &= M62;
r2 += r1 >> 62; r1 &= M62;
r3 += r2 >> 62; r2 &= M62;
r4 += r3 >> 62; r3 &= M62;
r->v[0] = r0;
r->v[1] = r1;
r->v[2] = r2;
r->v[3] = r3;
r->v[4] = r4;
VERIFY_CHECK(r0 >> 62 == 0);
VERIFY_CHECK(r1 >> 62 == 0);
VERIFY_CHECK(r2 >> 62 == 0);
VERIFY_CHECK(r3 >> 62 == 0);
VERIFY_CHECK(r4 >> 62 == 0);
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(r, 5, &modinfo->modulus, 0) >= 0); /* r >= 0 */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(r, 5, &modinfo->modulus, 1) < 0); /* r < modulus */
}
/* Compute the transition matrix and eta for 59 divsteps (where zeta=-(delta+1/2)).
* Note that the transformation matrix is scaled by 2^62 and not 2^59.
*
* Input: zeta: initial zeta
* f0: bottom limb of initial f
* g0: bottom limb of initial g
* Output: t: transition matrix
* Return: final zeta
*
* Implements the divsteps_n_matrix function from the explanation.
*/
static int64_t secp256k1_modinv64_divsteps_59(int64_t zeta, uint64_t f0, uint64_t g0, secp256k1_modinv64_trans2x2 *t) {
/* u,v,q,r are the elements of the transformation matrix being built up,
* starting with the identity matrix times 8 (because the caller expects
* a result scaled by 2^62). Semantically they are signed integers
* in range [-2^62,2^62], but here represented as unsigned mod 2^64. This
* permits left shifting (which is UB for negative numbers). The range
* being inside [-2^63,2^63) means that casting to signed works correctly.
*/
uint64_t u = 8, v = 0, q = 0, r = 8;
volatile uint64_t c1, c2;
uint64_t mask1, mask2, f = f0, g = g0, x, y, z;
int i;
for (i = 3; i < 62; ++i) {
VERIFY_CHECK((f & 1) == 1); /* f must always be odd */
VERIFY_CHECK((u * f0 + v * g0) == f << i);
VERIFY_CHECK((q * f0 + r * g0) == g << i);
/* Compute conditional masks for (zeta < 0) and for (g & 1). */
c1 = zeta >> 63;
mask1 = c1;
c2 = g & 1;
mask2 = -c2;
/* Compute x,y,z, conditionally negated versions of f,u,v. */
x = (f ^ mask1) - mask1;
y = (u ^ mask1) - mask1;
z = (v ^ mask1) - mask1;
/* Conditionally add x,y,z to g,q,r. */
g += x & mask2;
q += y & mask2;
r += z & mask2;
/* In what follows, c1 is a condition mask for (zeta < 0) and (g & 1). */
mask1 &= mask2;
/* Conditionally change zeta into -zeta-2 or zeta-1. */
zeta = (zeta ^ mask1) - 1;
/* Conditionally add g,q,r to f,u,v. */
f += g & mask1;
u += q & mask1;
v += r & mask1;
/* Shifts */
g >>= 1;
u <<= 1;
v <<= 1;
/* Bounds on zeta that follow from the bounds on iteration count (max 10*59 divsteps). */
VERIFY_CHECK(zeta >= -591 && zeta <= 591);
}
/* Return data in t and return value. */
t->u = (int64_t)u;
t->v = (int64_t)v;
t->q = (int64_t)q;
t->r = (int64_t)r;
/* The determinant of t must be a power of two. This guarantees that multiplication with t
* does not change the gcd of f and g, apart from adding a power-of-2 factor to it (which
* will be divided out again). As each divstep's individual matrix has determinant 2, the
* aggregate of 59 of them will have determinant 2^59. Multiplying with the initial
* 8*identity (which has determinant 2^6) means the overall outputs has determinant
* 2^65. */
VERIFY_CHECK(secp256k1_modinv64_det_check_pow2(t, 65, 0));
return zeta;
}
/* Compute the transition matrix and eta for 62 divsteps (variable time, eta=-delta).
*
* Input: eta: initial eta
* f0: bottom limb of initial f
* g0: bottom limb of initial g
* Output: t: transition matrix
* Return: final eta
*
* Implements the divsteps_n_matrix_var function from the explanation.
*/
static int64_t secp256k1_modinv64_divsteps_62_var(int64_t eta, uint64_t f0, uint64_t g0, secp256k1_modinv64_trans2x2 *t) {
/* Transformation matrix; see comments in secp256k1_modinv64_divsteps_62. */
uint64_t u = 1, v = 0, q = 0, r = 1;
uint64_t f = f0, g = g0, m;
uint32_t w;
int i = 62, limit, zeros;
for (;;) {
/* Use a sentinel bit to count zeros only up to i. */
zeros = secp256k1_ctz64_var(g | (UINT64_MAX << i));
/* Perform zeros divsteps at once; they all just divide g by two. */
g >>= zeros;
u <<= zeros;
v <<= zeros;
eta -= zeros;
i -= zeros;
/* We're done once we've done 62 divsteps. */
if (i == 0) break;
VERIFY_CHECK((f & 1) == 1);
VERIFY_CHECK((g & 1) == 1);
VERIFY_CHECK((u * f0 + v * g0) == f << (62 - i));
VERIFY_CHECK((q * f0 + r * g0) == g << (62 - i));
/* Bounds on eta that follow from the bounds on iteration count (max 12*62 divsteps). */
VERIFY_CHECK(eta >= -745 && eta <= 745);
/* If eta is negative, negate it and replace f,g with g,-f. */
if (eta < 0) {
uint64_t tmp;
eta = -eta;
tmp = f; f = g; g = -tmp;
tmp = u; u = q; q = -tmp;
tmp = v; v = r; r = -tmp;
/* Use a formula to cancel out up to 6 bits of g. Also, no more than i can be cancelled
* out (as we'd be done before that point), and no more than eta+1 can be done as its
* sign will flip again once that happens. */
limit = ((int)eta + 1) > i ? i : ((int)eta + 1);
VERIFY_CHECK(limit > 0 && limit <= 62);
/* m is a mask for the bottom min(limit, 6) bits. */
m = (UINT64_MAX >> (64 - limit)) & 63U;
/* Find what multiple of f must be added to g to cancel its bottom min(limit, 6)
* bits. */
w = (f * g * (f * f - 2)) & m;
} else {
/* In this branch, use a simpler formula that only lets us cancel up to 4 bits of g, as
* eta tends to be smaller here. */
limit = ((int)eta + 1) > i ? i : ((int)eta + 1);
VERIFY_CHECK(limit > 0 && limit <= 62);
/* m is a mask for the bottom min(limit, 4) bits. */
m = (UINT64_MAX >> (64 - limit)) & 15U;
/* Find what multiple of f must be added to g to cancel its bottom min(limit, 4)
* bits. */
w = f + (((f + 1) & 4) << 1);
w = (-w * g) & m;
}
g += f * w;
q += u * w;
r += v * w;
VERIFY_CHECK((g & m) == 0);
}
/* Return data in t and return value. */
t->u = (int64_t)u;
t->v = (int64_t)v;
t->q = (int64_t)q;
t->r = (int64_t)r;
/* The determinant of t must be a power of two. This guarantees that multiplication with t
* does not change the gcd of f and g, apart from adding a power-of-2 factor to it (which
* will be divided out again). As each divstep's individual matrix has determinant 2, the
* aggregate of 62 of them will have determinant 2^62. */
VERIFY_CHECK(secp256k1_modinv64_det_check_pow2(t, 62, 0));
return eta;
}
/* Compute the transition matrix and eta for 62 posdivsteps (variable time, eta=-delta), and keeps track
* of the Jacobi symbol along the way. f0 and g0 must be f and g mod 2^64 rather than 2^62, because
* Jacobi tracking requires knowing (f mod 8) rather than just (f mod 2).
*
* Input: eta: initial eta
* f0: bottom limb of initial f
* g0: bottom limb of initial g
* Output: t: transition matrix
* Input/Output: (*jacp & 1) is bitflipped if and only if the Jacobi symbol of (f | g) changes sign
* by applying the returned transformation matrix to it. The other bits of *jacp may
* change, but are meaningless.
* Return: final eta
*/
static int64_t secp256k1_modinv64_posdivsteps_62_var(int64_t eta, uint64_t f0, uint64_t g0, secp256k1_modinv64_trans2x2 *t, int *jacp) {
/* Transformation matrix; see comments in secp256k1_modinv64_divsteps_62. */
uint64_t u = 1, v = 0, q = 0, r = 1;
uint64_t f = f0, g = g0, m;
uint32_t w;
int i = 62, limit, zeros;
int jac = *jacp;
for (;;) {
/* Use a sentinel bit to count zeros only up to i. */
zeros = secp256k1_ctz64_var(g | (UINT64_MAX << i));
/* Perform zeros divsteps at once; they all just divide g by two. */
g >>= zeros;
u <<= zeros;
v <<= zeros;
eta -= zeros;
i -= zeros;
/* Update the bottom bit of jac: when dividing g by an odd power of 2,
* if (f mod 8) is 3 or 5, the Jacobi symbol changes sign. */
jac ^= (zeros & ((f >> 1) ^ (f >> 2)));
/* We're done once we've done 62 posdivsteps. */
if (i == 0) break;
VERIFY_CHECK((f & 1) == 1);
VERIFY_CHECK((g & 1) == 1);
VERIFY_CHECK((u * f0 + v * g0) == f << (62 - i));
VERIFY_CHECK((q * f0 + r * g0) == g << (62 - i));
/* If eta is negative, negate it and replace f,g with g,f. */
if (eta < 0) {
uint64_t tmp;
eta = -eta;
tmp = f; f = g; g = tmp;
tmp = u; u = q; q = tmp;
tmp = v; v = r; r = tmp;
/* Update bottom bit of jac: when swapping f and g, the Jacobi symbol changes sign
* if both f and g are 3 mod 4. */
jac ^= ((f & g) >> 1);
/* Use a formula to cancel out up to 6 bits of g. Also, no more than i can be cancelled
* out (as we'd be done before that point), and no more than eta+1 can be done as its
* sign will flip again once that happens. */
limit = ((int)eta + 1) > i ? i : ((int)eta + 1);
VERIFY_CHECK(limit > 0 && limit <= 62);
/* m is a mask for the bottom min(limit, 6) bits. */
m = (UINT64_MAX >> (64 - limit)) & 63U;
/* Find what multiple of f must be added to g to cancel its bottom min(limit, 6)
* bits. */
w = (f * g * (f * f - 2)) & m;
} else {
/* In this branch, use a simpler formula that only lets us cancel up to 4 bits of g, as
* eta tends to be smaller here. */
limit = ((int)eta + 1) > i ? i : ((int)eta + 1);
VERIFY_CHECK(limit > 0 && limit <= 62);
/* m is a mask for the bottom min(limit, 4) bits. */
m = (UINT64_MAX >> (64 - limit)) & 15U;
/* Find what multiple of f must be added to g to cancel its bottom min(limit, 4)
* bits. */
w = f + (((f + 1) & 4) << 1);
w = (-w * g) & m;
}
g += f * w;
q += u * w;
r += v * w;
VERIFY_CHECK((g & m) == 0);
}
/* Return data in t and return value. */
t->u = (int64_t)u;
t->v = (int64_t)v;
t->q = (int64_t)q;
t->r = (int64_t)r;
/* The determinant of t must be a power of two. This guarantees that multiplication with t
* does not change the gcd of f and g, apart from adding a power-of-2 factor to it (which
* will be divided out again). As each divstep's individual matrix has determinant 2 or -2,
* the aggregate of 62 of them will have determinant 2^62 or -2^62. */
VERIFY_CHECK(secp256k1_modinv64_det_check_pow2(t, 62, 1));
*jacp = jac;
return eta;
}
/* Compute (t/2^62) * [d, e] mod modulus, where t is a transition matrix scaled by 2^62.
*
* On input and output, d and e are in range (-2*modulus,modulus). All output limbs will be in range
* (-2^62,2^62).
*
* This implements the update_de function from the explanation.
*/
static void secp256k1_modinv64_update_de_62(secp256k1_modinv64_signed62 *d, secp256k1_modinv64_signed62 *e, const secp256k1_modinv64_trans2x2 *t, const secp256k1_modinv64_modinfo* modinfo) {
const uint64_t M62 = UINT64_MAX >> 2;
const int64_t d0 = d->v[0], d1 = d->v[1], d2 = d->v[2], d3 = d->v[3], d4 = d->v[4];
const int64_t e0 = e->v[0], e1 = e->v[1], e2 = e->v[2], e3 = e->v[3], e4 = e->v[4];
const int64_t u = t->u, v = t->v, q = t->q, r = t->r;
int64_t md, me, sd, se;
secp256k1_int128 cd, ce;
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(d, 5, &modinfo->modulus, -2) > 0); /* d > -2*modulus */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(d, 5, &modinfo->modulus, 1) < 0); /* d < modulus */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(e, 5, &modinfo->modulus, -2) > 0); /* e > -2*modulus */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(e, 5, &modinfo->modulus, 1) < 0); /* e < modulus */
VERIFY_CHECK(secp256k1_modinv64_abs(u) <= (((int64_t)1 << 62) - secp256k1_modinv64_abs(v))); /* |u|+|v| <= 2^62 */
VERIFY_CHECK(secp256k1_modinv64_abs(q) <= (((int64_t)1 << 62) - secp256k1_modinv64_abs(r))); /* |q|+|r| <= 2^62 */
/* [md,me] start as zero; plus [u,q] if d is negative; plus [v,r] if e is negative. */
sd = d4 >> 63;
se = e4 >> 63;
md = (u & sd) + (v & se);
me = (q & sd) + (r & se);
/* Begin computing t*[d,e]. */
secp256k1_i128_mul(&cd, u, d0);
secp256k1_i128_accum_mul(&cd, v, e0);
secp256k1_i128_mul(&ce, q, d0);
secp256k1_i128_accum_mul(&ce, r, e0);
/* Correct md,me so that t*[d,e]+modulus*[md,me] has 62 zero bottom bits. */
md -= (modinfo->modulus_inv62 * secp256k1_i128_to_u64(&cd) + md) & M62;
me -= (modinfo->modulus_inv62 * secp256k1_i128_to_u64(&ce) + me) & M62;
/* Update the beginning of computation for t*[d,e]+modulus*[md,me] now md,me are known. */
secp256k1_i128_accum_mul(&cd, modinfo->modulus.v[0], md);
secp256k1_i128_accum_mul(&ce, modinfo->modulus.v[0], me);
/* Verify that the low 62 bits of the computation are indeed zero, and then throw them away. */
VERIFY_CHECK((secp256k1_i128_to_u64(&cd) & M62) == 0); secp256k1_i128_rshift(&cd, 62);
VERIFY_CHECK((secp256k1_i128_to_u64(&ce) & M62) == 0); secp256k1_i128_rshift(&ce, 62);
/* Compute limb 1 of t*[d,e]+modulus*[md,me], and store it as output limb 0 (= down shift). */
secp256k1_i128_accum_mul(&cd, u, d1);
secp256k1_i128_accum_mul(&cd, v, e1);
secp256k1_i128_accum_mul(&ce, q, d1);
secp256k1_i128_accum_mul(&ce, r, e1);
if (modinfo->modulus.v[1]) { /* Optimize for the case where limb of modulus is zero. */
secp256k1_i128_accum_mul(&cd, modinfo->modulus.v[1], md);
secp256k1_i128_accum_mul(&ce, modinfo->modulus.v[1], me);
}
d->v[0] = secp256k1_i128_to_u64(&cd) & M62; secp256k1_i128_rshift(&cd, 62);
e->v[0] = secp256k1_i128_to_u64(&ce) & M62; secp256k1_i128_rshift(&ce, 62);
/* Compute limb 2 of t*[d,e]+modulus*[md,me], and store it as output limb 1. */
secp256k1_i128_accum_mul(&cd, u, d2);
secp256k1_i128_accum_mul(&cd, v, e2);
secp256k1_i128_accum_mul(&ce, q, d2);
secp256k1_i128_accum_mul(&ce, r, e2);
if (modinfo->modulus.v[2]) { /* Optimize for the case where limb of modulus is zero. */
secp256k1_i128_accum_mul(&cd, modinfo->modulus.v[2], md);
secp256k1_i128_accum_mul(&ce, modinfo->modulus.v[2], me);
}
d->v[1] = secp256k1_i128_to_u64(&cd) & M62; secp256k1_i128_rshift(&cd, 62);
e->v[1] = secp256k1_i128_to_u64(&ce) & M62; secp256k1_i128_rshift(&ce, 62);
/* Compute limb 3 of t*[d,e]+modulus*[md,me], and store it as output limb 2. */
secp256k1_i128_accum_mul(&cd, u, d3);
secp256k1_i128_accum_mul(&cd, v, e3);
secp256k1_i128_accum_mul(&ce, q, d3);
secp256k1_i128_accum_mul(&ce, r, e3);
if (modinfo->modulus.v[3]) { /* Optimize for the case where limb of modulus is zero. */
secp256k1_i128_accum_mul(&cd, modinfo->modulus.v[3], md);
secp256k1_i128_accum_mul(&ce, modinfo->modulus.v[3], me);
}
d->v[2] = secp256k1_i128_to_u64(&cd) & M62; secp256k1_i128_rshift(&cd, 62);
e->v[2] = secp256k1_i128_to_u64(&ce) & M62; secp256k1_i128_rshift(&ce, 62);
/* Compute limb 4 of t*[d,e]+modulus*[md,me], and store it as output limb 3. */
secp256k1_i128_accum_mul(&cd, u, d4);
secp256k1_i128_accum_mul(&cd, v, e4);
secp256k1_i128_accum_mul(&ce, q, d4);
secp256k1_i128_accum_mul(&ce, r, e4);
secp256k1_i128_accum_mul(&cd, modinfo->modulus.v[4], md);
secp256k1_i128_accum_mul(&ce, modinfo->modulus.v[4], me);
d->v[3] = secp256k1_i128_to_u64(&cd) & M62; secp256k1_i128_rshift(&cd, 62);
e->v[3] = secp256k1_i128_to_u64(&ce) & M62; secp256k1_i128_rshift(&ce, 62);
/* What remains is limb 5 of t*[d,e]+modulus*[md,me]; store it as output limb 4. */
d->v[4] = secp256k1_i128_to_i64(&cd);
e->v[4] = secp256k1_i128_to_i64(&ce);
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(d, 5, &modinfo->modulus, -2) > 0); /* d > -2*modulus */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(d, 5, &modinfo->modulus, 1) < 0); /* d < modulus */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(e, 5, &modinfo->modulus, -2) > 0); /* e > -2*modulus */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(e, 5, &modinfo->modulus, 1) < 0); /* e < modulus */
}
/* Compute (t/2^62) * [f, g], where t is a transition matrix scaled by 2^62.
*
* This implements the update_fg function from the explanation.
*/
static void secp256k1_modinv64_update_fg_62(secp256k1_modinv64_signed62 *f, secp256k1_modinv64_signed62 *g, const secp256k1_modinv64_trans2x2 *t) {
const uint64_t M62 = UINT64_MAX >> 2;
const int64_t f0 = f->v[0], f1 = f->v[1], f2 = f->v[2], f3 = f->v[3], f4 = f->v[4];
const int64_t g0 = g->v[0], g1 = g->v[1], g2 = g->v[2], g3 = g->v[3], g4 = g->v[4];
const int64_t u = t->u, v = t->v, q = t->q, r = t->r;
secp256k1_int128 cf, cg;
/* Start computing t*[f,g]. */
secp256k1_i128_mul(&cf, u, f0);
secp256k1_i128_accum_mul(&cf, v, g0);
secp256k1_i128_mul(&cg, q, f0);
secp256k1_i128_accum_mul(&cg, r, g0);
/* Verify that the bottom 62 bits of the result are zero, and then throw them away. */
VERIFY_CHECK((secp256k1_i128_to_u64(&cf) & M62) == 0); secp256k1_i128_rshift(&cf, 62);
VERIFY_CHECK((secp256k1_i128_to_u64(&cg) & M62) == 0); secp256k1_i128_rshift(&cg, 62);
/* Compute limb 1 of t*[f,g], and store it as output limb 0 (= down shift). */
secp256k1_i128_accum_mul(&cf, u, f1);
secp256k1_i128_accum_mul(&cf, v, g1);
secp256k1_i128_accum_mul(&cg, q, f1);
secp256k1_i128_accum_mul(&cg, r, g1);
f->v[0] = secp256k1_i128_to_u64(&cf) & M62; secp256k1_i128_rshift(&cf, 62);
g->v[0] = secp256k1_i128_to_u64(&cg) & M62; secp256k1_i128_rshift(&cg, 62);
/* Compute limb 2 of t*[f,g], and store it as output limb 1. */
secp256k1_i128_accum_mul(&cf, u, f2);
secp256k1_i128_accum_mul(&cf, v, g2);
secp256k1_i128_accum_mul(&cg, q, f2);
secp256k1_i128_accum_mul(&cg, r, g2);
f->v[1] = secp256k1_i128_to_u64(&cf) & M62; secp256k1_i128_rshift(&cf, 62);
g->v[1] = secp256k1_i128_to_u64(&cg) & M62; secp256k1_i128_rshift(&cg, 62);
/* Compute limb 3 of t*[f,g], and store it as output limb 2. */
secp256k1_i128_accum_mul(&cf, u, f3);
secp256k1_i128_accum_mul(&cf, v, g3);
secp256k1_i128_accum_mul(&cg, q, f3);
secp256k1_i128_accum_mul(&cg, r, g3);
f->v[2] = secp256k1_i128_to_u64(&cf) & M62; secp256k1_i128_rshift(&cf, 62);
g->v[2] = secp256k1_i128_to_u64(&cg) & M62; secp256k1_i128_rshift(&cg, 62);
/* Compute limb 4 of t*[f,g], and store it as output limb 3. */
secp256k1_i128_accum_mul(&cf, u, f4);
secp256k1_i128_accum_mul(&cf, v, g4);
secp256k1_i128_accum_mul(&cg, q, f4);
secp256k1_i128_accum_mul(&cg, r, g4);
f->v[3] = secp256k1_i128_to_u64(&cf) & M62; secp256k1_i128_rshift(&cf, 62);
g->v[3] = secp256k1_i128_to_u64(&cg) & M62; secp256k1_i128_rshift(&cg, 62);
/* What remains is limb 5 of t*[f,g]; store it as output limb 4. */
f->v[4] = secp256k1_i128_to_i64(&cf);
g->v[4] = secp256k1_i128_to_i64(&cg);
}
/* Compute (t/2^62) * [f, g], where t is a transition matrix for 62 divsteps.
*
* Version that operates on a variable number of limbs in f and g.
*
* This implements the update_fg function from the explanation.
*/
static void secp256k1_modinv64_update_fg_62_var(int len, secp256k1_modinv64_signed62 *f, secp256k1_modinv64_signed62 *g, const secp256k1_modinv64_trans2x2 *t) {
const uint64_t M62 = UINT64_MAX >> 2;
const int64_t u = t->u, v = t->v, q = t->q, r = t->r;
int64_t fi, gi;
secp256k1_int128 cf, cg;
int i;
VERIFY_CHECK(len > 0);
/* Start computing t*[f,g]. */
fi = f->v[0];
gi = g->v[0];
secp256k1_i128_mul(&cf, u, fi);
secp256k1_i128_accum_mul(&cf, v, gi);
secp256k1_i128_mul(&cg, q, fi);
secp256k1_i128_accum_mul(&cg, r, gi);
/* Verify that the bottom 62 bits of the result are zero, and then throw them away. */
VERIFY_CHECK((secp256k1_i128_to_u64(&cf) & M62) == 0); secp256k1_i128_rshift(&cf, 62);
VERIFY_CHECK((secp256k1_i128_to_u64(&cg) & M62) == 0); secp256k1_i128_rshift(&cg, 62);
/* Now iteratively compute limb i=1..len of t*[f,g], and store them in output limb i-1 (shifting
* down by 62 bits). */
for (i = 1; i < len; ++i) {
fi = f->v[i];
gi = g->v[i];
secp256k1_i128_accum_mul(&cf, u, fi);
secp256k1_i128_accum_mul(&cf, v, gi);
secp256k1_i128_accum_mul(&cg, q, fi);
secp256k1_i128_accum_mul(&cg, r, gi);
f->v[i - 1] = secp256k1_i128_to_u64(&cf) & M62; secp256k1_i128_rshift(&cf, 62);
g->v[i - 1] = secp256k1_i128_to_u64(&cg) & M62; secp256k1_i128_rshift(&cg, 62);
}
/* What remains is limb (len) of t*[f,g]; store it as output limb (len-1). */
f->v[len - 1] = secp256k1_i128_to_i64(&cf);
g->v[len - 1] = secp256k1_i128_to_i64(&cg);
}
/* Compute the inverse of x modulo modinfo->modulus, and replace x with it (constant time in x). */
static void secp256k1_modinv64(secp256k1_modinv64_signed62 *x, const secp256k1_modinv64_modinfo *modinfo) {
/* Start with d=0, e=1, f=modulus, g=x, zeta=-1. */
secp256k1_modinv64_signed62 d = {{0, 0, 0, 0, 0}};
secp256k1_modinv64_signed62 e = {{1, 0, 0, 0, 0}};
secp256k1_modinv64_signed62 f = modinfo->modulus;
secp256k1_modinv64_signed62 g = *x;
int i;
int64_t zeta = -1; /* zeta = -(delta+1/2); delta starts at 1/2. */
/* Do 10 iterations of 59 divsteps each = 590 divsteps. This suffices for 256-bit inputs. */
for (i = 0; i < 10; ++i) {
/* Compute transition matrix and new zeta after 59 divsteps. */
secp256k1_modinv64_trans2x2 t;
zeta = secp256k1_modinv64_divsteps_59(zeta, f.v[0], g.v[0], &t);
/* Update d,e using that transition matrix. */
secp256k1_modinv64_update_de_62(&d, &e, &t, modinfo);
/* Update f,g using that transition matrix. */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&f, 5, &modinfo->modulus, -1) > 0); /* f > -modulus */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&f, 5, &modinfo->modulus, 1) <= 0); /* f <= modulus */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&g, 5, &modinfo->modulus, -1) > 0); /* g > -modulus */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&g, 5, &modinfo->modulus, 1) < 0); /* g < modulus */
secp256k1_modinv64_update_fg_62(&f, &g, &t);
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&f, 5, &modinfo->modulus, -1) > 0); /* f > -modulus */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&f, 5, &modinfo->modulus, 1) <= 0); /* f <= modulus */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&g, 5, &modinfo->modulus, -1) > 0); /* g > -modulus */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&g, 5, &modinfo->modulus, 1) < 0); /* g < modulus */
}
/* At this point sufficient iterations have been performed that g must have reached 0
* and (if g was not originally 0) f must now equal +/- GCD of the initial f, g
* values i.e. +/- 1, and d now contains +/- the modular inverse. */
/* g == 0 */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&g, 5, &SECP256K1_SIGNED62_ONE, 0) == 0);
/* |f| == 1, or (x == 0 and d == 0 and f == modulus) */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&f, 5, &SECP256K1_SIGNED62_ONE, -1) == 0 ||
secp256k1_modinv64_mul_cmp_62(&f, 5, &SECP256K1_SIGNED62_ONE, 1) == 0 ||
(secp256k1_modinv64_mul_cmp_62(x, 5, &SECP256K1_SIGNED62_ONE, 0) == 0 &&
secp256k1_modinv64_mul_cmp_62(&d, 5, &SECP256K1_SIGNED62_ONE, 0) == 0 &&
secp256k1_modinv64_mul_cmp_62(&f, 5, &modinfo->modulus, 1) == 0));
/* Optionally negate d, normalize to [0,modulus), and return it. */
secp256k1_modinv64_normalize_62(&d, f.v[4], modinfo);
*x = d;
}
/* Compute the inverse of x modulo modinfo->modulus, and replace x with it (variable time). */
static void secp256k1_modinv64_var(secp256k1_modinv64_signed62 *x, const secp256k1_modinv64_modinfo *modinfo) {
/* Start with d=0, e=1, f=modulus, g=x, eta=-1. */
secp256k1_modinv64_signed62 d = {{0, 0, 0, 0, 0}};
secp256k1_modinv64_signed62 e = {{1, 0, 0, 0, 0}};
secp256k1_modinv64_signed62 f = modinfo->modulus;
secp256k1_modinv64_signed62 g = *x;
#ifdef VERIFY
int i = 0;
#endif
int j, len = 5;
int64_t eta = -1; /* eta = -delta; delta is initially 1 */
int64_t cond, fn, gn;
/* Do iterations of 62 divsteps each until g=0. */
while (1) {
/* Compute transition matrix and new eta after 62 divsteps. */
secp256k1_modinv64_trans2x2 t;
eta = secp256k1_modinv64_divsteps_62_var(eta, f.v[0], g.v[0], &t);
/* Update d,e using that transition matrix. */
secp256k1_modinv64_update_de_62(&d, &e, &t, modinfo);
/* Update f,g using that transition matrix. */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&f, len, &modinfo->modulus, -1) > 0); /* f > -modulus */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&f, len, &modinfo->modulus, 1) <= 0); /* f <= modulus */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&g, len, &modinfo->modulus, -1) > 0); /* g > -modulus */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&g, len, &modinfo->modulus, 1) < 0); /* g < modulus */
secp256k1_modinv64_update_fg_62_var(len, &f, &g, &t);
/* If the bottom limb of g is zero, there is a chance that g=0. */
if (g.v[0] == 0) {
cond = 0;
/* Check if the other limbs are also 0. */
for (j = 1; j < len; ++j) {
cond |= g.v[j];
}
/* If so, we're done. */
if (cond == 0) break;
}
/* Determine if len>1 and limb (len-1) of both f and g is 0 or -1. */
fn = f.v[len - 1];
gn = g.v[len - 1];
cond = ((int64_t)len - 2) >> 63;
cond |= fn ^ (fn >> 63);
cond |= gn ^ (gn >> 63);
/* If so, reduce length, propagating the sign of f and g's top limb into the one below. */
if (cond == 0) {
f.v[len - 2] |= (uint64_t)fn << 62;
g.v[len - 2] |= (uint64_t)gn << 62;
--len;
}
VERIFY_CHECK(++i < 12); /* We should never need more than 12*62 = 744 divsteps */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&f, len, &modinfo->modulus, -1) > 0); /* f > -modulus */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&f, len, &modinfo->modulus, 1) <= 0); /* f <= modulus */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&g, len, &modinfo->modulus, -1) > 0); /* g > -modulus */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&g, len, &modinfo->modulus, 1) < 0); /* g < modulus */
}
/* At this point g is 0 and (if g was not originally 0) f must now equal +/- GCD of
* the initial f, g values i.e. +/- 1, and d now contains +/- the modular inverse. */
/* g == 0 */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&g, len, &SECP256K1_SIGNED62_ONE, 0) == 0);
/* |f| == 1, or (x == 0 and d == 0 and f == modulus) */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&f, len, &SECP256K1_SIGNED62_ONE, -1) == 0 ||
secp256k1_modinv64_mul_cmp_62(&f, len, &SECP256K1_SIGNED62_ONE, 1) == 0 ||
(secp256k1_modinv64_mul_cmp_62(x, 5, &SECP256K1_SIGNED62_ONE, 0) == 0 &&
secp256k1_modinv64_mul_cmp_62(&d, 5, &SECP256K1_SIGNED62_ONE, 0) == 0 &&
secp256k1_modinv64_mul_cmp_62(&f, len, &modinfo->modulus, 1) == 0));
/* Optionally negate d, normalize to [0,modulus), and return it. */
secp256k1_modinv64_normalize_62(&d, f.v[len - 1], modinfo);
*x = d;
}
/* Do up to 25 iterations of 62 posdivsteps (up to 1550 steps; more is extremely rare) each until f=1.
* In VERIFY mode use a lower number of iterations (744, close to the median 756), so failure actually occurs. */
#ifdef VERIFY
#define JACOBI64_ITERATIONS 12
#else
#define JACOBI64_ITERATIONS 25
#endif
/* Compute the Jacobi symbol of x modulo modinfo->modulus (variable time). gcd(x,modulus) must be 1. */
static int secp256k1_jacobi64_maybe_var(const secp256k1_modinv64_signed62 *x, const secp256k1_modinv64_modinfo *modinfo) {
/* Start with f=modulus, g=x, eta=-1. */
secp256k1_modinv64_signed62 f = modinfo->modulus;
secp256k1_modinv64_signed62 g = *x;
int j, len = 5;
int64_t eta = -1; /* eta = -delta; delta is initially 1 */
int64_t cond, fn, gn;
int jac = 0;
int count;
/* The input limbs must all be non-negative. */
VERIFY_CHECK(g.v[0] >= 0 && g.v[1] >= 0 && g.v[2] >= 0 && g.v[3] >= 0 && g.v[4] >= 0);
/* If x > 0, then if the loop below converges, it converges to f=g=gcd(x,modulus). Since we
* require that gcd(x,modulus)=1 and modulus>=3, x cannot be 0. Thus, we must reach f=1 (or
* time out). */
VERIFY_CHECK((g.v[0] | g.v[1] | g.v[2] | g.v[3] | g.v[4]) != 0);
for (count = 0; count < JACOBI64_ITERATIONS; ++count) {
/* Compute transition matrix and new eta after 62 posdivsteps. */
secp256k1_modinv64_trans2x2 t;
eta = secp256k1_modinv64_posdivsteps_62_var(eta, f.v[0] | ((uint64_t)f.v[1] << 62), g.v[0] | ((uint64_t)g.v[1] << 62), &t, &jac);
/* Update f,g using that transition matrix. */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&f, len, &modinfo->modulus, 0) > 0); /* f > 0 */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&f, len, &modinfo->modulus, 1) <= 0); /* f <= modulus */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&g, len, &modinfo->modulus, 0) > 0); /* g > 0 */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&g, len, &modinfo->modulus, 1) < 0); /* g < modulus */
secp256k1_modinv64_update_fg_62_var(len, &f, &g, &t);
/* If the bottom limb of f is 1, there is a chance that f=1. */
if (f.v[0] == 1) {
cond = 0;
/* Check if the other limbs are also 0. */
for (j = 1; j < len; ++j) {
cond |= f.v[j];
}
/* If so, we're done. When f=1, the Jacobi symbol (g | f)=1. */
if (cond == 0) return 1 - 2*(jac & 1);
}
/* Determine if len>1 and limb (len-1) of both f and g is 0. */
fn = f.v[len - 1];
gn = g.v[len - 1];
cond = ((int64_t)len - 2) >> 63;
cond |= fn;
cond |= gn;
/* If so, reduce length. */
if (cond == 0) --len;
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&f, len, &modinfo->modulus, 0) > 0); /* f > 0 */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&f, len, &modinfo->modulus, 1) <= 0); /* f <= modulus */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&g, len, &modinfo->modulus, 0) > 0); /* g > 0 */
VERIFY_CHECK(secp256k1_modinv64_mul_cmp_62(&g, len, &modinfo->modulus, 1) < 0); /* g < modulus */
}
/* The loop failed to converge to f=g after 1550 iterations. Return 0, indicating unknown result. */
return 0;
}
#endif /* SECP256K1_MODINV64_IMPL_H */

View File

@@ -0,0 +1,285 @@
/***********************************************************************
* Copyright (c) 2020 Jonas Nick *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_MODULE_EXTRAKEYS_MAIN_H
#define SECP256K1_MODULE_EXTRAKEYS_MAIN_H
#include "../../../include/secp256k1.h"
#include "../../../include/secp256k1_extrakeys.h"
#include "../../util.h"
static SECP256K1_INLINE int secp256k1_xonly_pubkey_load(const secp256k1_context* ctx, secp256k1_ge *ge, const secp256k1_xonly_pubkey *pubkey) {
return secp256k1_pubkey_load(ctx, ge, (const secp256k1_pubkey *) pubkey);
}
static SECP256K1_INLINE void secp256k1_xonly_pubkey_save(secp256k1_xonly_pubkey *pubkey, secp256k1_ge *ge) {
secp256k1_pubkey_save((secp256k1_pubkey *) pubkey, ge);
}
int secp256k1_xonly_pubkey_parse(const secp256k1_context* ctx, secp256k1_xonly_pubkey *pubkey, const unsigned char *input32) {
secp256k1_ge pk;
secp256k1_fe x;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(pubkey != NULL);
memset(pubkey, 0, sizeof(*pubkey));
ARG_CHECK(input32 != NULL);
if (!secp256k1_fe_set_b32_limit(&x, input32)) {
return 0;
}
if (!secp256k1_ge_set_xo_var(&pk, &x, 0)) {
return 0;
}
if (!secp256k1_ge_is_in_correct_subgroup(&pk)) {
return 0;
}
secp256k1_xonly_pubkey_save(pubkey, &pk);
return 1;
}
int secp256k1_xonly_pubkey_serialize(const secp256k1_context* ctx, unsigned char *output32, const secp256k1_xonly_pubkey *pubkey) {
secp256k1_ge pk;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(output32 != NULL);
memset(output32, 0, 32);
ARG_CHECK(pubkey != NULL);
if (!secp256k1_xonly_pubkey_load(ctx, &pk, pubkey)) {
return 0;
}
secp256k1_fe_get_b32(output32, &pk.x);
return 1;
}
int secp256k1_xonly_pubkey_cmp(const secp256k1_context* ctx, const secp256k1_xonly_pubkey* pk0, const secp256k1_xonly_pubkey* pk1) {
unsigned char out[2][32];
const secp256k1_xonly_pubkey* pk[2];
int i;
VERIFY_CHECK(ctx != NULL);
pk[0] = pk0; pk[1] = pk1;
for (i = 0; i < 2; i++) {
/* If the public key is NULL or invalid, xonly_pubkey_serialize will
* call the illegal_callback and return 0. In that case we will
* serialize the key as all zeros which is less than any valid public
* key. This results in consistent comparisons even if NULL or invalid
* pubkeys are involved and prevents edge cases such as sorting
* algorithms that use this function and do not terminate as a
* result. */
if (!secp256k1_xonly_pubkey_serialize(ctx, out[i], pk[i])) {
/* Note that xonly_pubkey_serialize should already set the output to
* zero in that case, but it's not guaranteed by the API, we can't
* test it and writing a VERIFY_CHECK is more complex than
* explicitly memsetting (again). */
memset(out[i], 0, sizeof(out[i]));
}
}
return secp256k1_memcmp_var(out[0], out[1], sizeof(out[1]));
}
/** Keeps a group element as is if it has an even Y and otherwise negates it.
* y_parity is set to 0 in the former case and to 1 in the latter case.
* Requires that the coordinates of r are normalized. */
static int secp256k1_extrakeys_ge_even_y(secp256k1_ge *r) {
int y_parity = 0;
VERIFY_CHECK(!secp256k1_ge_is_infinity(r));
if (secp256k1_fe_is_odd(&r->y)) {
secp256k1_fe_negate(&r->y, &r->y, 1);
y_parity = 1;
}
return y_parity;
}
int secp256k1_xonly_pubkey_from_pubkey(const secp256k1_context* ctx, secp256k1_xonly_pubkey *xonly_pubkey, int *pk_parity, const secp256k1_pubkey *pubkey) {
secp256k1_ge pk;
int tmp;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(xonly_pubkey != NULL);
ARG_CHECK(pubkey != NULL);
if (!secp256k1_pubkey_load(ctx, &pk, pubkey)) {
return 0;
}
tmp = secp256k1_extrakeys_ge_even_y(&pk);
if (pk_parity != NULL) {
*pk_parity = tmp;
}
secp256k1_xonly_pubkey_save(xonly_pubkey, &pk);
return 1;
}
int secp256k1_xonly_pubkey_tweak_add(const secp256k1_context* ctx, secp256k1_pubkey *output_pubkey, const secp256k1_xonly_pubkey *internal_pubkey, const unsigned char *tweak32) {
secp256k1_ge pk;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(output_pubkey != NULL);
memset(output_pubkey, 0, sizeof(*output_pubkey));
ARG_CHECK(internal_pubkey != NULL);
ARG_CHECK(tweak32 != NULL);
if (!secp256k1_xonly_pubkey_load(ctx, &pk, internal_pubkey)
|| !secp256k1_ec_pubkey_tweak_add_helper(&pk, tweak32)) {
return 0;
}
secp256k1_pubkey_save(output_pubkey, &pk);
return 1;
}
int secp256k1_xonly_pubkey_tweak_add_check(const secp256k1_context* ctx, const unsigned char *tweaked_pubkey32, int tweaked_pk_parity, const secp256k1_xonly_pubkey *internal_pubkey, const unsigned char *tweak32) {
secp256k1_ge pk;
unsigned char pk_expected32[32];
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(internal_pubkey != NULL);
ARG_CHECK(tweaked_pubkey32 != NULL);
ARG_CHECK(tweak32 != NULL);
if (!secp256k1_xonly_pubkey_load(ctx, &pk, internal_pubkey)
|| !secp256k1_ec_pubkey_tweak_add_helper(&pk, tweak32)) {
return 0;
}
secp256k1_fe_normalize_var(&pk.x);
secp256k1_fe_normalize_var(&pk.y);
secp256k1_fe_get_b32(pk_expected32, &pk.x);
return secp256k1_memcmp_var(&pk_expected32, tweaked_pubkey32, 32) == 0
&& secp256k1_fe_is_odd(&pk.y) == tweaked_pk_parity;
}
static void secp256k1_keypair_save(secp256k1_keypair *keypair, const secp256k1_scalar *sk, secp256k1_ge *pk) {
secp256k1_scalar_get_b32(&keypair->data[0], sk);
secp256k1_pubkey_save((secp256k1_pubkey *)&keypair->data[32], pk);
}
static int secp256k1_keypair_seckey_load(const secp256k1_context* ctx, secp256k1_scalar *sk, const secp256k1_keypair *keypair) {
int ret;
ret = secp256k1_scalar_set_b32_seckey(sk, &keypair->data[0]);
/* We can declassify ret here because sk is only zero if a keypair function
* failed (which zeroes the keypair) and its return value is ignored. */
secp256k1_declassify(ctx, &ret, sizeof(ret));
ARG_CHECK(ret);
return ret;
}
/* Load a keypair into pk and sk (if non-NULL). This function declassifies pk
* and ARG_CHECKs that the keypair is not invalid. It always initializes sk and
* pk with dummy values. */
static int secp256k1_keypair_load(const secp256k1_context* ctx, secp256k1_scalar *sk, secp256k1_ge *pk, const secp256k1_keypair *keypair) {
int ret;
const secp256k1_pubkey *pubkey = (const secp256k1_pubkey *)&keypair->data[32];
/* Need to declassify the pubkey because pubkey_load ARG_CHECKs if it's
* invalid. */
secp256k1_declassify(ctx, pubkey, sizeof(*pubkey));
ret = secp256k1_pubkey_load(ctx, pk, pubkey);
if (sk != NULL) {
ret = ret && secp256k1_keypair_seckey_load(ctx, sk, keypair);
}
if (!ret) {
*pk = secp256k1_ge_const_g;
if (sk != NULL) {
*sk = secp256k1_scalar_one;
}
}
return ret;
}
int secp256k1_keypair_create(const secp256k1_context* ctx, secp256k1_keypair *keypair, const unsigned char *seckey32) {
secp256k1_scalar sk;
secp256k1_ge pk;
int ret = 0;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(keypair != NULL);
memset(keypair, 0, sizeof(*keypair));
ARG_CHECK(secp256k1_ecmult_gen_context_is_built(&ctx->ecmult_gen_ctx));
ARG_CHECK(seckey32 != NULL);
ret = secp256k1_ec_pubkey_create_helper(&ctx->ecmult_gen_ctx, &sk, &pk, seckey32);
secp256k1_keypair_save(keypair, &sk, &pk);
secp256k1_memczero(keypair, sizeof(*keypair), !ret);
secp256k1_scalar_clear(&sk);
return ret;
}
int secp256k1_keypair_sec(const secp256k1_context* ctx, unsigned char *seckey, const secp256k1_keypair *keypair) {
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(seckey != NULL);
memset(seckey, 0, 32);
ARG_CHECK(keypair != NULL);
memcpy(seckey, &keypair->data[0], 32);
return 1;
}
int secp256k1_keypair_pub(const secp256k1_context* ctx, secp256k1_pubkey *pubkey, const secp256k1_keypair *keypair) {
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(pubkey != NULL);
memset(pubkey, 0, sizeof(*pubkey));
ARG_CHECK(keypair != NULL);
memcpy(pubkey->data, &keypair->data[32], sizeof(*pubkey));
return 1;
}
int secp256k1_keypair_xonly_pub(const secp256k1_context* ctx, secp256k1_xonly_pubkey *pubkey, int *pk_parity, const secp256k1_keypair *keypair) {
secp256k1_ge pk;
int tmp;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(pubkey != NULL);
memset(pubkey, 0, sizeof(*pubkey));
ARG_CHECK(keypair != NULL);
if (!secp256k1_keypair_load(ctx, NULL, &pk, keypair)) {
return 0;
}
tmp = secp256k1_extrakeys_ge_even_y(&pk);
if (pk_parity != NULL) {
*pk_parity = tmp;
}
secp256k1_xonly_pubkey_save(pubkey, &pk);
return 1;
}
int secp256k1_keypair_xonly_tweak_add(const secp256k1_context* ctx, secp256k1_keypair *keypair, const unsigned char *tweak32) {
secp256k1_ge pk;
secp256k1_scalar sk;
int y_parity;
int ret;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(keypair != NULL);
ARG_CHECK(tweak32 != NULL);
ret = secp256k1_keypair_load(ctx, &sk, &pk, keypair);
memset(keypair, 0, sizeof(*keypair));
y_parity = secp256k1_extrakeys_ge_even_y(&pk);
if (y_parity == 1) {
secp256k1_scalar_negate(&sk, &sk);
}
ret &= secp256k1_ec_seckey_tweak_add_helper(&sk, tweak32);
ret &= secp256k1_ec_pubkey_tweak_add_helper(&pk, tweak32);
secp256k1_declassify(ctx, &ret, sizeof(ret));
if (ret) {
secp256k1_keypair_save(keypair, &sk, &pk);
}
secp256k1_scalar_clear(&sk);
return ret;
}
#endif

View File

@@ -0,0 +1,269 @@
/***********************************************************************
* Copyright (c) 2018-2020 Andrew Poelstra, Jonas Nick *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_MODULE_SCHNORRSIG_MAIN_H
#define SECP256K1_MODULE_SCHNORRSIG_MAIN_H
#include "../../../include/secp256k1.h"
#include "../../../include/secp256k1_schnorrsig.h"
#include "../../hash.h"
/* Initializes SHA256 with fixed midstate. This midstate was computed by applying
* SHA256 to SHA256("BIP0340/nonce")||SHA256("BIP0340/nonce"). */
static void secp256k1_nonce_function_bip340_sha256_tagged(secp256k1_sha256 *sha) {
secp256k1_sha256_initialize(sha);
sha->s[0] = 0x46615b35ul;
sha->s[1] = 0xf4bfbff7ul;
sha->s[2] = 0x9f8dc671ul;
sha->s[3] = 0x83627ab3ul;
sha->s[4] = 0x60217180ul;
sha->s[5] = 0x57358661ul;
sha->s[6] = 0x21a29e54ul;
sha->s[7] = 0x68b07b4cul;
sha->bytes = 64;
}
/* Initializes SHA256 with fixed midstate. This midstate was computed by applying
* SHA256 to SHA256("BIP0340/aux")||SHA256("BIP0340/aux"). */
static void secp256k1_nonce_function_bip340_sha256_tagged_aux(secp256k1_sha256 *sha) {
secp256k1_sha256_initialize(sha);
sha->s[0] = 0x24dd3219ul;
sha->s[1] = 0x4eba7e70ul;
sha->s[2] = 0xca0fabb9ul;
sha->s[3] = 0x0fa3166dul;
sha->s[4] = 0x3afbe4b1ul;
sha->s[5] = 0x4c44df97ul;
sha->s[6] = 0x4aac2739ul;
sha->s[7] = 0x249e850aul;
sha->bytes = 64;
}
/* algo argument for nonce_function_bip340 to derive the nonce exactly as stated in BIP-340
* by using the correct tagged hash function. */
static const unsigned char bip340_algo[] = {'B', 'I', 'P', '0', '3', '4', '0', '/', 'n', 'o', 'n', 'c', 'e'};
static const unsigned char schnorrsig_extraparams_magic[4] = SECP256K1_SCHNORRSIG_EXTRAPARAMS_MAGIC;
static int nonce_function_bip340(unsigned char *nonce32, const unsigned char *msg, size_t msglen, const unsigned char *key32, const unsigned char *xonly_pk32, const unsigned char *algo, size_t algolen, void *data) {
secp256k1_sha256 sha;
unsigned char masked_key[32];
int i;
if (algo == NULL) {
return 0;
}
if (data != NULL) {
secp256k1_nonce_function_bip340_sha256_tagged_aux(&sha);
secp256k1_sha256_write(&sha, data, 32);
secp256k1_sha256_finalize(&sha, masked_key);
for (i = 0; i < 32; i++) {
masked_key[i] ^= key32[i];
}
} else {
/* Precomputed TaggedHash("BIP0340/aux", 0x0000...00); */
static const unsigned char ZERO_MASK[32] = {
84, 241, 105, 207, 201, 226, 229, 114,
116, 128, 68, 31, 144, 186, 37, 196,
136, 244, 97, 199, 11, 94, 165, 220,
170, 247, 175, 105, 39, 10, 165, 20
};
for (i = 0; i < 32; i++) {
masked_key[i] = key32[i] ^ ZERO_MASK[i];
}
}
/* Tag the hash with algo which is important to avoid nonce reuse across
* algorithms. If this nonce function is used in BIP-340 signing as defined
* in the spec, an optimized tagging implementation is used. */
if (algolen == sizeof(bip340_algo)
&& secp256k1_memcmp_var(algo, bip340_algo, algolen) == 0) {
secp256k1_nonce_function_bip340_sha256_tagged(&sha);
} else {
secp256k1_sha256_initialize_tagged(&sha, algo, algolen);
}
/* Hash masked-key||pk||msg using the tagged hash as per the spec */
secp256k1_sha256_write(&sha, masked_key, 32);
secp256k1_sha256_write(&sha, xonly_pk32, 32);
secp256k1_sha256_write(&sha, msg, msglen);
secp256k1_sha256_finalize(&sha, nonce32);
secp256k1_sha256_clear(&sha);
return 1;
}
const secp256k1_nonce_function_hardened secp256k1_nonce_function_bip340 = nonce_function_bip340;
/* Initializes SHA256 with fixed midstate. This midstate was computed by applying
* SHA256 to SHA256("BIP0340/challenge")||SHA256("BIP0340/challenge"). */
static void secp256k1_schnorrsig_sha256_tagged(secp256k1_sha256 *sha) {
secp256k1_sha256_initialize(sha);
sha->s[0] = 0x9cecba11ul;
sha->s[1] = 0x23925381ul;
sha->s[2] = 0x11679112ul;
sha->s[3] = 0xd1627e0ful;
sha->s[4] = 0x97c87550ul;
sha->s[5] = 0x003cc765ul;
sha->s[6] = 0x90f61164ul;
sha->s[7] = 0x33e9b66aul;
sha->bytes = 64;
}
static void secp256k1_schnorrsig_challenge(secp256k1_scalar* e, const unsigned char *r32, const unsigned char *msg, size_t msglen, const unsigned char *pubkey32)
{
unsigned char buf[32];
secp256k1_sha256 sha;
/* tagged hash(r.x, pk.x, msg) */
secp256k1_schnorrsig_sha256_tagged(&sha);
secp256k1_sha256_write(&sha, r32, 32);
secp256k1_sha256_write(&sha, pubkey32, 32);
secp256k1_sha256_write(&sha, msg, msglen);
secp256k1_sha256_finalize(&sha, buf);
/* Set scalar e to the challenge hash modulo the curve order as per
* BIP340. */
secp256k1_scalar_set_b32(e, buf, NULL);
}
static int secp256k1_schnorrsig_sign_internal(const secp256k1_context* ctx, unsigned char *sig64, const unsigned char *msg, size_t msglen, const secp256k1_keypair *keypair, secp256k1_nonce_function_hardened noncefp, void *ndata) {
secp256k1_scalar sk;
secp256k1_scalar e;
secp256k1_scalar k;
secp256k1_gej rj;
secp256k1_ge pk;
secp256k1_ge r;
unsigned char buf[32] = { 0 };
unsigned char pk_buf[32];
unsigned char seckey[32];
int ret = 1;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(secp256k1_ecmult_gen_context_is_built(&ctx->ecmult_gen_ctx));
ARG_CHECK(sig64 != NULL);
ARG_CHECK(msg != NULL || msglen == 0);
ARG_CHECK(keypair != NULL);
if (noncefp == NULL) {
noncefp = secp256k1_nonce_function_bip340;
}
ret &= secp256k1_keypair_load(ctx, &sk, &pk, keypair);
/* Because we are signing for a x-only pubkey, the secret key is negated
* before signing if the point corresponding to the secret key does not
* have an even Y. */
if (secp256k1_fe_is_odd(&pk.y)) {
secp256k1_scalar_negate(&sk, &sk);
}
secp256k1_scalar_get_b32(seckey, &sk);
secp256k1_fe_get_b32(pk_buf, &pk.x);
ret &= !!noncefp(buf, msg, msglen, seckey, pk_buf, bip340_algo, sizeof(bip340_algo), ndata);
secp256k1_scalar_set_b32(&k, buf, NULL);
ret &= !secp256k1_scalar_is_zero(&k);
secp256k1_scalar_cmov(&k, &secp256k1_scalar_one, !ret);
secp256k1_ecmult_gen(&ctx->ecmult_gen_ctx, &rj, &k);
secp256k1_ge_set_gej(&r, &rj);
/* We declassify r to allow using it as a branch point. This is fine
* because r is not a secret. */
secp256k1_declassify(ctx, &r, sizeof(r));
secp256k1_fe_normalize_var(&r.y);
if (secp256k1_fe_is_odd(&r.y)) {
secp256k1_scalar_negate(&k, &k);
}
secp256k1_fe_normalize_var(&r.x);
secp256k1_fe_get_b32(&sig64[0], &r.x);
secp256k1_schnorrsig_challenge(&e, &sig64[0], msg, msglen, pk_buf);
secp256k1_scalar_mul(&e, &e, &sk);
secp256k1_scalar_add(&e, &e, &k);
secp256k1_scalar_get_b32(&sig64[32], &e);
secp256k1_memczero(sig64, 64, !ret);
secp256k1_scalar_clear(&k);
secp256k1_scalar_clear(&sk);
secp256k1_memclear(seckey, sizeof(seckey));
secp256k1_gej_clear(&rj);
return ret;
}
int secp256k1_schnorrsig_sign32(const secp256k1_context* ctx, unsigned char *sig64, const unsigned char *msg32, const secp256k1_keypair *keypair, const unsigned char *aux_rand32) {
/* We cast away const from the passed aux_rand32 argument since we know the default nonce function does not modify it. */
return secp256k1_schnorrsig_sign_internal(ctx, sig64, msg32, 32, keypair, secp256k1_nonce_function_bip340, (unsigned char*)aux_rand32);
}
int secp256k1_schnorrsig_sign(const secp256k1_context* ctx, unsigned char *sig64, const unsigned char *msg32, const secp256k1_keypair *keypair, const unsigned char *aux_rand32) {
return secp256k1_schnorrsig_sign32(ctx, sig64, msg32, keypair, aux_rand32);
}
int secp256k1_schnorrsig_sign_custom(const secp256k1_context* ctx, unsigned char *sig64, const unsigned char *msg, size_t msglen, const secp256k1_keypair *keypair, secp256k1_schnorrsig_extraparams *extraparams) {
secp256k1_nonce_function_hardened noncefp = NULL;
void *ndata = NULL;
VERIFY_CHECK(ctx != NULL);
if (extraparams != NULL) {
ARG_CHECK(secp256k1_memcmp_var(extraparams->magic,
schnorrsig_extraparams_magic,
sizeof(extraparams->magic)) == 0);
noncefp = extraparams->noncefp;
ndata = extraparams->ndata;
}
return secp256k1_schnorrsig_sign_internal(ctx, sig64, msg, msglen, keypair, noncefp, ndata);
}
int secp256k1_schnorrsig_verify(const secp256k1_context* ctx, const unsigned char *sig64, const unsigned char *msg, size_t msglen, const secp256k1_xonly_pubkey *pubkey) {
secp256k1_scalar s;
secp256k1_scalar e;
secp256k1_gej rj;
secp256k1_ge pk;
secp256k1_gej pkj;
secp256k1_fe rx;
secp256k1_ge r;
unsigned char buf[32];
int overflow;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(sig64 != NULL);
ARG_CHECK(msg != NULL || msglen == 0);
ARG_CHECK(pubkey != NULL);
if (!secp256k1_fe_set_b32_limit(&rx, &sig64[0])) {
return 0;
}
secp256k1_scalar_set_b32(&s, &sig64[32], &overflow);
if (overflow) {
return 0;
}
if (!secp256k1_xonly_pubkey_load(ctx, &pk, pubkey)) {
return 0;
}
/* Compute e. */
secp256k1_fe_get_b32(buf, &pk.x);
secp256k1_schnorrsig_challenge(&e, &sig64[0], msg, msglen, buf);
/* Compute rj = s*G + (-e)*pkj */
secp256k1_scalar_negate(&e, &e);
secp256k1_gej_set_ge(&pkj, &pk);
secp256k1_ecmult(&rj, &pkj, &e, &s);
secp256k1_ge_set_gej_var(&r, &rj);
if (secp256k1_ge_is_infinity(&r)) {
return 0;
}
secp256k1_fe_normalize_var(&r.y);
return !secp256k1_fe_is_odd(&r.y) &&
secp256k1_fe_equal(&rx, &r.x);
}
#endif

View File

@@ -0,0 +1,90 @@
/*****************************************************************************************************
* Copyright (c) 2013, 2014, 2017, 2021 Pieter Wuille, Andrew Poelstra, Jonas Nick, Russell O'Connor *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php. *
*****************************************************************************************************/
#include <inttypes.h>
#include <stdio.h>
#include "../include/secp256k1.h"
#include "assumptions.h"
#include "util.h"
#include "field_impl.h"
#include "group_impl.h"
#include "int128_impl.h"
#include "ecmult.h"
#include "ecmult_compute_table_impl.h"
static void print_table(FILE *fp, const char *name, int window_g, const secp256k1_ge_storage* table) {
int j;
int i;
fprintf(fp, "const secp256k1_ge_storage %s[ECMULT_TABLE_SIZE(WINDOW_G)] = {\n", name);
fprintf(fp, " S(%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32
",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32")\n",
SECP256K1_GE_STORAGE_CONST_GET(table[0]));
j = 1;
for(i = 3; i <= window_g; ++i) {
fprintf(fp, "#if WINDOW_G > %d\n", i-1);
for(;j < ECMULT_TABLE_SIZE(i); ++j) {
fprintf(fp, ",S(%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32
",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32")\n",
SECP256K1_GE_STORAGE_CONST_GET(table[j]));
}
fprintf(fp, "#endif\n");
}
fprintf(fp, "};\n");
}
static void print_two_tables(FILE *fp, int window_g) {
secp256k1_ge_storage* table = malloc(ECMULT_TABLE_SIZE(window_g) * sizeof(secp256k1_ge_storage));
secp256k1_ge_storage* table_128 = malloc(ECMULT_TABLE_SIZE(window_g) * sizeof(secp256k1_ge_storage));
secp256k1_ecmult_compute_two_tables(table, table_128, window_g, &secp256k1_ge_const_g);
print_table(fp, "secp256k1_pre_g", window_g, table);
print_table(fp, "secp256k1_pre_g_128", window_g, table_128);
free(table);
free(table_128);
}
int main(void) {
/* Always compute all tables for window sizes up to 15. */
int window_g = (ECMULT_WINDOW_SIZE < 15) ? 15 : ECMULT_WINDOW_SIZE;
const char outfile[] = "src/precomputed_ecmult.c";
FILE* fp;
fp = fopen(outfile, "w");
if (fp == NULL) {
fprintf(stderr, "Could not open %s for writing!\n", outfile);
return -1;
}
fprintf(fp, "/* This file was automatically generated by precompute_ecmult. */\n");
fprintf(fp, "/* This file contains an array secp256k1_pre_g with odd multiples of the base point G and\n");
fprintf(fp, " * an array secp256k1_pre_g_128 with odd multiples of 2^128*G for accelerating the computation of a*P + b*G.\n");
fprintf(fp, " */\n");
fprintf(fp, "#include \"group.h\"\n");
fprintf(fp, "#include \"ecmult.h\"\n");
fprintf(fp, "#include \"precomputed_ecmult.h\"\n");
fprintf(fp, "#define S(a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p) SECP256K1_GE_STORAGE_CONST(0x##a##u,0x##b##u,0x##c##u,0x##d##u,0x##e##u,0x##f##u,0x##g##u,0x##h##u,0x##i##u,0x##j##u,0x##k##u,0x##l##u,0x##m##u,0x##n##u,0x##o##u,0x##p##u)\n");
fprintf(fp, "#if ECMULT_WINDOW_SIZE > %d\n", window_g);
fprintf(fp, " #error configuration mismatch, invalid ECMULT_WINDOW_SIZE. Try deleting precomputed_ecmult.c before the build.\n");
fprintf(fp, "#endif\n");
fprintf(fp, "#ifdef EXHAUSTIVE_TEST_ORDER\n");
fprintf(fp, "# error Cannot compile precomputed_ecmult.c in exhaustive test mode\n");
fprintf(fp, "#endif /* EXHAUSTIVE_TEST_ORDER */\n");
fprintf(fp, "#define WINDOW_G ECMULT_WINDOW_SIZE\n");
print_two_tables(fp, window_g);
fprintf(fp, "#undef S\n");
fclose(fp);
return 0;
}

View File

@@ -0,0 +1,100 @@
/*********************************************************************************
* Copyright (c) 2013, 2014, 2015, 2021 Thomas Daede, Cory Fields, Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php. *
*********************************************************************************/
#include <inttypes.h>
#include <stdio.h>
#include "../include/secp256k1.h"
#include "assumptions.h"
#include "util.h"
#include "group.h"
#include "int128_impl.h"
#include "ecmult_gen.h"
#include "ecmult_gen_compute_table_impl.h"
static const int CONFIGS[][2] = {
{2, 5},
{11, 6},
{43, 6}
};
static void print_table(FILE* fp, int blocks, int teeth) {
int spacing = CEIL_DIV(256, blocks * teeth);
size_t points = ((size_t)1) << (teeth - 1);
int outer;
size_t inner;
secp256k1_ge_storage* table = checked_malloc(&default_error_callback, blocks * points * sizeof(secp256k1_ge_storage));
secp256k1_ecmult_gen_compute_table(table, &secp256k1_ge_const_g, blocks, teeth, spacing);
fprintf(fp, "#elif (COMB_BLOCKS == %d) && (COMB_TEETH == %d) && (COMB_SPACING == %d)\n", blocks, teeth, spacing);
for (outer = 0; outer != blocks; outer++) {
fprintf(fp,"{");
for (inner = 0; inner != points; inner++) {
fprintf(fp, "S(%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32
",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32",%"PRIx32")",
SECP256K1_GE_STORAGE_CONST_GET(table[outer * points + inner]));
if (inner != points - 1) {
fprintf(fp,",\n");
}
}
if (outer != blocks - 1) {
fprintf(fp,"},\n");
} else {
fprintf(fp,"}\n");
}
}
free(table);
}
int main(int argc, char **argv) {
const char outfile[] = "src/precomputed_ecmult_gen.c";
FILE* fp;
size_t config;
int did_current_config = 0;
(void)argc;
(void)argv;
fp = fopen(outfile, "w");
if (fp == NULL) {
fprintf(stderr, "Could not open %s for writing!\n", outfile);
return -1;
}
fprintf(fp, "/* This file was automatically generated by precompute_ecmult_gen. */\n");
fprintf(fp, "/* See ecmult_gen_impl.h for details about the contents of this file. */\n");
fprintf(fp, "#include \"group.h\"\n");
fprintf(fp, "#include \"ecmult_gen.h\"\n");
fprintf(fp, "#include \"precomputed_ecmult_gen.h\"\n");
fprintf(fp, "#ifdef EXHAUSTIVE_TEST_ORDER\n");
fprintf(fp, "# error Cannot compile precomputed_ecmult_gen.c in exhaustive test mode\n");
fprintf(fp, "#endif /* EXHAUSTIVE_TEST_ORDER */\n");
fprintf(fp, "#define S(a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p) SECP256K1_GE_STORAGE_CONST(0x##a##u,0x##b##u,0x##c##u,0x##d##u,0x##e##u,0x##f##u,0x##g##u,0x##h##u,0x##i##u,0x##j##u,0x##k##u,0x##l##u,0x##m##u,0x##n##u,0x##o##u,0x##p##u)\n");
fprintf(fp, "const secp256k1_ge_storage secp256k1_ecmult_gen_prec_table[COMB_BLOCKS][COMB_POINTS] = {\n");
fprintf(fp, "#if 0\n");
for (config = 0; config < sizeof(CONFIGS) / sizeof(*CONFIGS); ++config) {
print_table(fp, CONFIGS[config][0], CONFIGS[config][1]);
if (CONFIGS[config][0] == COMB_BLOCKS && CONFIGS[config][1] == COMB_TEETH) {
did_current_config = 1;
}
}
if (!did_current_config) {
print_table(fp, COMB_BLOCKS, COMB_TEETH);
}
fprintf(fp, "#else\n");
fprintf(fp, "# error Configuration mismatch, invalid COMB_* parameters. Try deleting precomputed_ecmult_gen.c before the build.\n");
fprintf(fp, "#endif\n");
fprintf(fp, "};\n");
fprintf(fp, "#undef S\n");
fclose(fp);
return 0;
}

File diff suppressed because it is too large Load Diff

View File

@@ -0,0 +1,38 @@
/*****************************************************************************************************
* Copyright (c) 2013, 2014, 2017, 2021 Pieter Wuille, Andrew Poelstra, Jonas Nick, Russell O'Connor *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php. *
*****************************************************************************************************/
#ifndef SECP256K1_PRECOMPUTED_ECMULT_H
#define SECP256K1_PRECOMPUTED_ECMULT_H
#ifdef __cplusplus
extern "C" {
#endif
#include "ecmult.h"
#include "group.h"
#if defined(EXHAUSTIVE_TEST_ORDER)
# if EXHAUSTIVE_TEST_ORDER == 7
# define WINDOW_G 3
# elif EXHAUSTIVE_TEST_ORDER == 13
# define WINDOW_G 4
# elif EXHAUSTIVE_TEST_ORDER == 199
# define WINDOW_G 8
# else
# error No known generator for the specified exhaustive test group order.
# endif
static secp256k1_ge_storage secp256k1_pre_g[ECMULT_TABLE_SIZE(WINDOW_G)];
static secp256k1_ge_storage secp256k1_pre_g_128[ECMULT_TABLE_SIZE(WINDOW_G)];
#else /* !defined(EXHAUSTIVE_TEST_ORDER) */
# define WINDOW_G ECMULT_WINDOW_SIZE
extern const secp256k1_ge_storage secp256k1_pre_g[ECMULT_TABLE_SIZE(WINDOW_G)];
extern const secp256k1_ge_storage secp256k1_pre_g_128[ECMULT_TABLE_SIZE(WINDOW_G)];
#endif /* defined(EXHAUSTIVE_TEST_ORDER) */
#ifdef __cplusplus
}
#endif
#endif /* SECP256K1_PRECOMPUTED_ECMULT_H */

File diff suppressed because it is too large Load Diff

View File

@@ -0,0 +1,26 @@
/*********************************************************************************
* Copyright (c) 2013, 2014, 2015, 2021 Thomas Daede, Cory Fields, Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php. *
*********************************************************************************/
#ifndef SECP256K1_PRECOMPUTED_ECMULT_GEN_H
#define SECP256K1_PRECOMPUTED_ECMULT_GEN_H
#ifdef __cplusplus
extern "C" {
#endif
#include "group.h"
#include "ecmult_gen.h"
#ifdef EXHAUSTIVE_TEST_ORDER
static secp256k1_ge_storage secp256k1_ecmult_gen_prec_table[COMB_BLOCKS][COMB_POINTS];
#else
extern const secp256k1_ge_storage secp256k1_ecmult_gen_prec_table[COMB_BLOCKS][COMB_POINTS];
#endif /* defined(EXHAUSTIVE_TEST_ORDER) */
#ifdef __cplusplus
}
#endif
#endif /* SECP256K1_PRECOMPUTED_ECMULT_GEN_H */

105
libsecp256k1/src/scalar.h Normal file
View File

@@ -0,0 +1,105 @@
/***********************************************************************
* Copyright (c) 2014 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_SCALAR_H
#define SECP256K1_SCALAR_H
#include "util.h"
#if defined(EXHAUSTIVE_TEST_ORDER)
#include "scalar_low.h"
#elif defined(SECP256K1_WIDEMUL_INT128)
#include "scalar_4x64.h"
#elif defined(SECP256K1_WIDEMUL_INT64)
#include "scalar_8x32.h"
#else
#error "Please select wide multiplication implementation"
#endif
/** Clear a scalar to prevent the leak of sensitive data. */
static void secp256k1_scalar_clear(secp256k1_scalar *r);
/** Access bits (1 < count <= 32) from a scalar. All requested bits must belong to the same 32-bit limb. */
static uint32_t secp256k1_scalar_get_bits_limb32(const secp256k1_scalar *a, unsigned int offset, unsigned int count);
/** Access bits (1 < count <= 32) from a scalar. offset + count must be < 256. Not constant time in offset and count. */
static uint32_t secp256k1_scalar_get_bits_var(const secp256k1_scalar *a, unsigned int offset, unsigned int count);
/** Set a scalar from a big endian byte array. The scalar will be reduced modulo group order `n`.
* In: bin: pointer to a 32-byte array.
* Out: r: scalar to be set.
* overflow: non-zero if the scalar was bigger or equal to `n` before reduction, zero otherwise (can be NULL).
*/
static void secp256k1_scalar_set_b32(secp256k1_scalar *r, const unsigned char *bin, int *overflow);
/** Set a scalar from a big endian byte array and returns 1 if it is a valid
* seckey and 0 otherwise. */
static int secp256k1_scalar_set_b32_seckey(secp256k1_scalar *r, const unsigned char *bin);
/** Set a scalar to an unsigned integer. */
static void secp256k1_scalar_set_int(secp256k1_scalar *r, unsigned int v);
/** Convert a scalar to a byte array. */
static void secp256k1_scalar_get_b32(unsigned char *bin, const secp256k1_scalar* a);
/** Add two scalars together (modulo the group order). Returns whether it overflowed. */
static int secp256k1_scalar_add(secp256k1_scalar *r, const secp256k1_scalar *a, const secp256k1_scalar *b);
/** Conditionally add a power of two to a scalar. The result is not allowed to overflow. */
static void secp256k1_scalar_cadd_bit(secp256k1_scalar *r, unsigned int bit, int flag);
/** Multiply two scalars (modulo the group order). */
static void secp256k1_scalar_mul(secp256k1_scalar *r, const secp256k1_scalar *a, const secp256k1_scalar *b);
/** Compute the inverse of a scalar (modulo the group order). */
static void secp256k1_scalar_inverse(secp256k1_scalar *r, const secp256k1_scalar *a);
/** Compute the inverse of a scalar (modulo the group order), without constant-time guarantee. */
static void secp256k1_scalar_inverse_var(secp256k1_scalar *r, const secp256k1_scalar *a);
/** Compute the complement of a scalar (modulo the group order). */
static void secp256k1_scalar_negate(secp256k1_scalar *r, const secp256k1_scalar *a);
/** Multiply a scalar with the multiplicative inverse of 2. */
static void secp256k1_scalar_half(secp256k1_scalar *r, const secp256k1_scalar *a);
/** Check whether a scalar equals zero. */
static int secp256k1_scalar_is_zero(const secp256k1_scalar *a);
/** Check whether a scalar equals one. */
static int secp256k1_scalar_is_one(const secp256k1_scalar *a);
/** Check whether a scalar, considered as an nonnegative integer, is even. */
static int secp256k1_scalar_is_even(const secp256k1_scalar *a);
/** Check whether a scalar is higher than the group order divided by 2. */
static int secp256k1_scalar_is_high(const secp256k1_scalar *a);
/** Conditionally negate a number, in constant time.
* Returns -1 if the number was negated, 1 otherwise */
static int secp256k1_scalar_cond_negate(secp256k1_scalar *a, int flag);
/** Compare two scalars. */
static int secp256k1_scalar_eq(const secp256k1_scalar *a, const secp256k1_scalar *b);
/** Find r1 and r2 such that r1+r2*2^128 = k. */
static void secp256k1_scalar_split_128(secp256k1_scalar *r1, secp256k1_scalar *r2, const secp256k1_scalar *k);
/** Find r1 and r2 such that r1+r2*lambda = k, where r1 and r2 or their
* negations are maximum 128 bits long (see secp256k1_ge_mul_lambda). It is
* required that r1, r2, and k all point to different objects. */
static void secp256k1_scalar_split_lambda(secp256k1_scalar * SECP256K1_RESTRICT r1, secp256k1_scalar * SECP256K1_RESTRICT r2, const secp256k1_scalar * SECP256K1_RESTRICT k);
/** Multiply a and b (without taking the modulus!), divide by 2**shift, and round to the nearest integer. Shift must be at least 256. */
static void secp256k1_scalar_mul_shift_var(secp256k1_scalar *r, const secp256k1_scalar *a, const secp256k1_scalar *b, unsigned int shift);
/** If flag is true, set *r equal to *a; otherwise leave it. Constant-time. Both *r and *a must be initialized.*/
static void secp256k1_scalar_cmov(secp256k1_scalar *r, const secp256k1_scalar *a, int flag);
/** Check invariants on a scalar (no-op unless VERIFY is enabled). */
static void secp256k1_scalar_verify(const secp256k1_scalar *r);
#define SECP256K1_SCALAR_VERIFY(r) secp256k1_scalar_verify(r)
#endif /* SECP256K1_SCALAR_H */

View File

@@ -0,0 +1,19 @@
/***********************************************************************
* Copyright (c) 2014 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_SCALAR_REPR_H
#define SECP256K1_SCALAR_REPR_H
#include <stdint.h>
/** A scalar modulo the group order of the secp256k1 curve. */
typedef struct {
uint64_t d[4];
} secp256k1_scalar;
#define SECP256K1_SCALAR_CONST(d7, d6, d5, d4, d3, d2, d1, d0) {{((uint64_t)(d1)) << 32 | (d0), ((uint64_t)(d3)) << 32 | (d2), ((uint64_t)(d5)) << 32 | (d4), ((uint64_t)(d7)) << 32 | (d6)}}
#endif /* SECP256K1_SCALAR_REPR_H */

File diff suppressed because it is too large Load Diff

View File

@@ -0,0 +1,19 @@
/***********************************************************************
* Copyright (c) 2014 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_SCALAR_REPR_H
#define SECP256K1_SCALAR_REPR_H
#include <stdint.h>
/** A scalar modulo the group order of the secp256k1 curve. */
typedef struct {
uint32_t d[8];
} secp256k1_scalar;
#define SECP256K1_SCALAR_CONST(d7, d6, d5, d4, d3, d2, d1, d0) {{(d0), (d1), (d2), (d3), (d4), (d5), (d6), (d7)}}
#endif /* SECP256K1_SCALAR_REPR_H */

View File

@@ -0,0 +1,816 @@
/***********************************************************************
* Copyright (c) 2014 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_SCALAR_REPR_IMPL_H
#define SECP256K1_SCALAR_REPR_IMPL_H
#include "checkmem.h"
#include "modinv32_impl.h"
#include "util.h"
/* Limbs of the secp256k1 order. */
#define SECP256K1_N_0 ((uint32_t)0xD0364141UL)
#define SECP256K1_N_1 ((uint32_t)0xBFD25E8CUL)
#define SECP256K1_N_2 ((uint32_t)0xAF48A03BUL)
#define SECP256K1_N_3 ((uint32_t)0xBAAEDCE6UL)
#define SECP256K1_N_4 ((uint32_t)0xFFFFFFFEUL)
#define SECP256K1_N_5 ((uint32_t)0xFFFFFFFFUL)
#define SECP256K1_N_6 ((uint32_t)0xFFFFFFFFUL)
#define SECP256K1_N_7 ((uint32_t)0xFFFFFFFFUL)
/* Limbs of 2^256 minus the secp256k1 order. */
#define SECP256K1_N_C_0 (~SECP256K1_N_0 + 1)
#define SECP256K1_N_C_1 (~SECP256K1_N_1)
#define SECP256K1_N_C_2 (~SECP256K1_N_2)
#define SECP256K1_N_C_3 (~SECP256K1_N_3)
#define SECP256K1_N_C_4 (1)
/* Limbs of half the secp256k1 order. */
#define SECP256K1_N_H_0 ((uint32_t)0x681B20A0UL)
#define SECP256K1_N_H_1 ((uint32_t)0xDFE92F46UL)
#define SECP256K1_N_H_2 ((uint32_t)0x57A4501DUL)
#define SECP256K1_N_H_3 ((uint32_t)0x5D576E73UL)
#define SECP256K1_N_H_4 ((uint32_t)0xFFFFFFFFUL)
#define SECP256K1_N_H_5 ((uint32_t)0xFFFFFFFFUL)
#define SECP256K1_N_H_6 ((uint32_t)0xFFFFFFFFUL)
#define SECP256K1_N_H_7 ((uint32_t)0x7FFFFFFFUL)
SECP256K1_INLINE static void secp256k1_scalar_set_int(secp256k1_scalar *r, unsigned int v) {
r->d[0] = v;
r->d[1] = 0;
r->d[2] = 0;
r->d[3] = 0;
r->d[4] = 0;
r->d[5] = 0;
r->d[6] = 0;
r->d[7] = 0;
SECP256K1_SCALAR_VERIFY(r);
}
SECP256K1_INLINE static uint32_t secp256k1_scalar_get_bits_limb32(const secp256k1_scalar *a, unsigned int offset, unsigned int count) {
SECP256K1_SCALAR_VERIFY(a);
VERIFY_CHECK(count > 0 && count <= 32);
VERIFY_CHECK((offset + count - 1) >> 5 == offset >> 5);
return (a->d[offset >> 5] >> (offset & 0x1F)) & (0xFFFFFFFF >> (32 - count));
}
SECP256K1_INLINE static uint32_t secp256k1_scalar_get_bits_var(const secp256k1_scalar *a, unsigned int offset, unsigned int count) {
SECP256K1_SCALAR_VERIFY(a);
VERIFY_CHECK(count > 0 && count <= 32);
VERIFY_CHECK(offset + count <= 256);
if ((offset + count - 1) >> 5 == offset >> 5) {
return secp256k1_scalar_get_bits_limb32(a, offset, count);
} else {
VERIFY_CHECK((offset >> 5) + 1 < 8);
return ((a->d[offset >> 5] >> (offset & 0x1F)) | (a->d[(offset >> 5) + 1] << (32 - (offset & 0x1F)))) & (0xFFFFFFFF >> (32 - count));
}
}
SECP256K1_INLINE static int secp256k1_scalar_check_overflow(const secp256k1_scalar *a) {
int yes = 0;
int no = 0;
no |= (a->d[7] < SECP256K1_N_7); /* No need for a > check. */
no |= (a->d[6] < SECP256K1_N_6); /* No need for a > check. */
no |= (a->d[5] < SECP256K1_N_5); /* No need for a > check. */
no |= (a->d[4] < SECP256K1_N_4);
yes |= (a->d[4] > SECP256K1_N_4) & ~no;
no |= (a->d[3] < SECP256K1_N_3) & ~yes;
yes |= (a->d[3] > SECP256K1_N_3) & ~no;
no |= (a->d[2] < SECP256K1_N_2) & ~yes;
yes |= (a->d[2] > SECP256K1_N_2) & ~no;
no |= (a->d[1] < SECP256K1_N_1) & ~yes;
yes |= (a->d[1] > SECP256K1_N_1) & ~no;
yes |= (a->d[0] >= SECP256K1_N_0) & ~no;
return yes;
}
SECP256K1_INLINE static int secp256k1_scalar_reduce(secp256k1_scalar *r, uint32_t overflow) {
uint64_t t;
VERIFY_CHECK(overflow <= 1);
t = (uint64_t)r->d[0] + overflow * SECP256K1_N_C_0;
r->d[0] = t & 0xFFFFFFFFUL; t >>= 32;
t += (uint64_t)r->d[1] + overflow * SECP256K1_N_C_1;
r->d[1] = t & 0xFFFFFFFFUL; t >>= 32;
t += (uint64_t)r->d[2] + overflow * SECP256K1_N_C_2;
r->d[2] = t & 0xFFFFFFFFUL; t >>= 32;
t += (uint64_t)r->d[3] + overflow * SECP256K1_N_C_3;
r->d[3] = t & 0xFFFFFFFFUL; t >>= 32;
t += (uint64_t)r->d[4] + overflow * SECP256K1_N_C_4;
r->d[4] = t & 0xFFFFFFFFUL; t >>= 32;
t += (uint64_t)r->d[5];
r->d[5] = t & 0xFFFFFFFFUL; t >>= 32;
t += (uint64_t)r->d[6];
r->d[6] = t & 0xFFFFFFFFUL; t >>= 32;
t += (uint64_t)r->d[7];
r->d[7] = t & 0xFFFFFFFFUL;
SECP256K1_SCALAR_VERIFY(r);
return overflow;
}
static int secp256k1_scalar_add(secp256k1_scalar *r, const secp256k1_scalar *a, const secp256k1_scalar *b) {
int overflow;
uint64_t t = (uint64_t)a->d[0] + b->d[0];
SECP256K1_SCALAR_VERIFY(a);
SECP256K1_SCALAR_VERIFY(b);
r->d[0] = t & 0xFFFFFFFFULL; t >>= 32;
t += (uint64_t)a->d[1] + b->d[1];
r->d[1] = t & 0xFFFFFFFFULL; t >>= 32;
t += (uint64_t)a->d[2] + b->d[2];
r->d[2] = t & 0xFFFFFFFFULL; t >>= 32;
t += (uint64_t)a->d[3] + b->d[3];
r->d[3] = t & 0xFFFFFFFFULL; t >>= 32;
t += (uint64_t)a->d[4] + b->d[4];
r->d[4] = t & 0xFFFFFFFFULL; t >>= 32;
t += (uint64_t)a->d[5] + b->d[5];
r->d[5] = t & 0xFFFFFFFFULL; t >>= 32;
t += (uint64_t)a->d[6] + b->d[6];
r->d[6] = t & 0xFFFFFFFFULL; t >>= 32;
t += (uint64_t)a->d[7] + b->d[7];
r->d[7] = t & 0xFFFFFFFFULL; t >>= 32;
overflow = t + secp256k1_scalar_check_overflow(r);
VERIFY_CHECK(overflow == 0 || overflow == 1);
secp256k1_scalar_reduce(r, overflow);
SECP256K1_SCALAR_VERIFY(r);
return overflow;
}
static void secp256k1_scalar_cadd_bit(secp256k1_scalar *r, unsigned int bit, int flag) {
uint64_t t;
volatile int vflag = flag;
SECP256K1_SCALAR_VERIFY(r);
VERIFY_CHECK(bit < 256);
bit += ((uint32_t) vflag - 1) & 0x100; /* forcing (bit >> 5) > 7 makes this a noop */
t = (uint64_t)r->d[0] + (((uint32_t)((bit >> 5) == 0)) << (bit & 0x1F));
r->d[0] = t & 0xFFFFFFFFULL; t >>= 32;
t += (uint64_t)r->d[1] + (((uint32_t)((bit >> 5) == 1)) << (bit & 0x1F));
r->d[1] = t & 0xFFFFFFFFULL; t >>= 32;
t += (uint64_t)r->d[2] + (((uint32_t)((bit >> 5) == 2)) << (bit & 0x1F));
r->d[2] = t & 0xFFFFFFFFULL; t >>= 32;
t += (uint64_t)r->d[3] + (((uint32_t)((bit >> 5) == 3)) << (bit & 0x1F));
r->d[3] = t & 0xFFFFFFFFULL; t >>= 32;
t += (uint64_t)r->d[4] + (((uint32_t)((bit >> 5) == 4)) << (bit & 0x1F));
r->d[4] = t & 0xFFFFFFFFULL; t >>= 32;
t += (uint64_t)r->d[5] + (((uint32_t)((bit >> 5) == 5)) << (bit & 0x1F));
r->d[5] = t & 0xFFFFFFFFULL; t >>= 32;
t += (uint64_t)r->d[6] + (((uint32_t)((bit >> 5) == 6)) << (bit & 0x1F));
r->d[6] = t & 0xFFFFFFFFULL; t >>= 32;
t += (uint64_t)r->d[7] + (((uint32_t)((bit >> 5) == 7)) << (bit & 0x1F));
r->d[7] = t & 0xFFFFFFFFULL;
SECP256K1_SCALAR_VERIFY(r);
VERIFY_CHECK((t >> 32) == 0);
}
static void secp256k1_scalar_set_b32(secp256k1_scalar *r, const unsigned char *b32, int *overflow) {
int over;
r->d[0] = secp256k1_read_be32(&b32[28]);
r->d[1] = secp256k1_read_be32(&b32[24]);
r->d[2] = secp256k1_read_be32(&b32[20]);
r->d[3] = secp256k1_read_be32(&b32[16]);
r->d[4] = secp256k1_read_be32(&b32[12]);
r->d[5] = secp256k1_read_be32(&b32[8]);
r->d[6] = secp256k1_read_be32(&b32[4]);
r->d[7] = secp256k1_read_be32(&b32[0]);
over = secp256k1_scalar_reduce(r, secp256k1_scalar_check_overflow(r));
if (overflow) {
*overflow = over;
}
SECP256K1_SCALAR_VERIFY(r);
}
static void secp256k1_scalar_get_b32(unsigned char *bin, const secp256k1_scalar* a) {
SECP256K1_SCALAR_VERIFY(a);
secp256k1_write_be32(&bin[0], a->d[7]);
secp256k1_write_be32(&bin[4], a->d[6]);
secp256k1_write_be32(&bin[8], a->d[5]);
secp256k1_write_be32(&bin[12], a->d[4]);
secp256k1_write_be32(&bin[16], a->d[3]);
secp256k1_write_be32(&bin[20], a->d[2]);
secp256k1_write_be32(&bin[24], a->d[1]);
secp256k1_write_be32(&bin[28], a->d[0]);
}
SECP256K1_INLINE static int secp256k1_scalar_is_zero(const secp256k1_scalar *a) {
SECP256K1_SCALAR_VERIFY(a);
return (a->d[0] | a->d[1] | a->d[2] | a->d[3] | a->d[4] | a->d[5] | a->d[6] | a->d[7]) == 0;
}
static void secp256k1_scalar_negate(secp256k1_scalar *r, const secp256k1_scalar *a) {
uint32_t nonzero = 0xFFFFFFFFUL * (secp256k1_scalar_is_zero(a) == 0);
uint64_t t = (uint64_t)(~a->d[0]) + SECP256K1_N_0 + 1;
SECP256K1_SCALAR_VERIFY(a);
r->d[0] = t & nonzero; t >>= 32;
t += (uint64_t)(~a->d[1]) + SECP256K1_N_1;
r->d[1] = t & nonzero; t >>= 32;
t += (uint64_t)(~a->d[2]) + SECP256K1_N_2;
r->d[2] = t & nonzero; t >>= 32;
t += (uint64_t)(~a->d[3]) + SECP256K1_N_3;
r->d[3] = t & nonzero; t >>= 32;
t += (uint64_t)(~a->d[4]) + SECP256K1_N_4;
r->d[4] = t & nonzero; t >>= 32;
t += (uint64_t)(~a->d[5]) + SECP256K1_N_5;
r->d[5] = t & nonzero; t >>= 32;
t += (uint64_t)(~a->d[6]) + SECP256K1_N_6;
r->d[6] = t & nonzero; t >>= 32;
t += (uint64_t)(~a->d[7]) + SECP256K1_N_7;
r->d[7] = t & nonzero;
SECP256K1_SCALAR_VERIFY(r);
}
static void secp256k1_scalar_half(secp256k1_scalar *r, const secp256k1_scalar *a) {
/* Writing `/` for field division and `//` for integer division, we compute
*
* a/2 = (a - (a&1))/2 + (a&1)/2
* = (a >> 1) + (a&1 ? 1/2 : 0)
* = (a >> 1) + (a&1 ? n//2+1 : 0),
*
* where n is the group order and in the last equality we have used 1/2 = n//2+1 (mod n).
* For n//2, we have the constants SECP256K1_N_H_0, ...
*
* This sum does not overflow. The most extreme case is a = -2, the largest odd scalar. Here:
* - the left summand is: a >> 1 = (a - a&1)/2 = (n-2-1)//2 = (n-3)//2
* - the right summand is: a&1 ? n//2+1 : 0 = n//2+1 = (n-1)//2 + 2//2 = (n+1)//2
* Together they sum to (n-3)//2 + (n+1)//2 = (2n-2)//2 = n - 1, which is less than n.
*/
uint32_t mask = -(uint32_t)(a->d[0] & 1U);
uint64_t t = (uint32_t)((a->d[0] >> 1) | (a->d[1] << 31));
SECP256K1_SCALAR_VERIFY(a);
t += (SECP256K1_N_H_0 + 1U) & mask;
r->d[0] = t; t >>= 32;
t += (uint32_t)((a->d[1] >> 1) | (a->d[2] << 31));
t += SECP256K1_N_H_1 & mask;
r->d[1] = t; t >>= 32;
t += (uint32_t)((a->d[2] >> 1) | (a->d[3] << 31));
t += SECP256K1_N_H_2 & mask;
r->d[2] = t; t >>= 32;
t += (uint32_t)((a->d[3] >> 1) | (a->d[4] << 31));
t += SECP256K1_N_H_3 & mask;
r->d[3] = t; t >>= 32;
t += (uint32_t)((a->d[4] >> 1) | (a->d[5] << 31));
t += SECP256K1_N_H_4 & mask;
r->d[4] = t; t >>= 32;
t += (uint32_t)((a->d[5] >> 1) | (a->d[6] << 31));
t += SECP256K1_N_H_5 & mask;
r->d[5] = t; t >>= 32;
t += (uint32_t)((a->d[6] >> 1) | (a->d[7] << 31));
t += SECP256K1_N_H_6 & mask;
r->d[6] = t; t >>= 32;
r->d[7] = (uint32_t)t + (uint32_t)(a->d[7] >> 1) + (SECP256K1_N_H_7 & mask);
/* The line above only computed the bottom 32 bits of r->d[7]. Redo the computation
* in full 64 bits to make sure the top 32 bits are indeed zero. */
VERIFY_CHECK((t + (a->d[7] >> 1) + (SECP256K1_N_H_7 & mask)) >> 32 == 0);
SECP256K1_SCALAR_VERIFY(r);
}
SECP256K1_INLINE static int secp256k1_scalar_is_one(const secp256k1_scalar *a) {
SECP256K1_SCALAR_VERIFY(a);
return ((a->d[0] ^ 1) | a->d[1] | a->d[2] | a->d[3] | a->d[4] | a->d[5] | a->d[6] | a->d[7]) == 0;
}
static int secp256k1_scalar_is_high(const secp256k1_scalar *a) {
int yes = 0;
int no = 0;
SECP256K1_SCALAR_VERIFY(a);
no |= (a->d[7] < SECP256K1_N_H_7);
yes |= (a->d[7] > SECP256K1_N_H_7) & ~no;
no |= (a->d[6] < SECP256K1_N_H_6) & ~yes; /* No need for a > check. */
no |= (a->d[5] < SECP256K1_N_H_5) & ~yes; /* No need for a > check. */
no |= (a->d[4] < SECP256K1_N_H_4) & ~yes; /* No need for a > check. */
no |= (a->d[3] < SECP256K1_N_H_3) & ~yes;
yes |= (a->d[3] > SECP256K1_N_H_3) & ~no;
no |= (a->d[2] < SECP256K1_N_H_2) & ~yes;
yes |= (a->d[2] > SECP256K1_N_H_2) & ~no;
no |= (a->d[1] < SECP256K1_N_H_1) & ~yes;
yes |= (a->d[1] > SECP256K1_N_H_1) & ~no;
yes |= (a->d[0] > SECP256K1_N_H_0) & ~no;
return yes;
}
static int secp256k1_scalar_cond_negate(secp256k1_scalar *r, int flag) {
/* If we are flag = 0, mask = 00...00 and this is a no-op;
* if we are flag = 1, mask = 11...11 and this is identical to secp256k1_scalar_negate */
volatile int vflag = flag;
uint32_t mask = -vflag;
uint32_t nonzero = 0xFFFFFFFFUL * (secp256k1_scalar_is_zero(r) == 0);
uint64_t t = (uint64_t)(r->d[0] ^ mask) + ((SECP256K1_N_0 + 1) & mask);
SECP256K1_SCALAR_VERIFY(r);
r->d[0] = t & nonzero; t >>= 32;
t += (uint64_t)(r->d[1] ^ mask) + (SECP256K1_N_1 & mask);
r->d[1] = t & nonzero; t >>= 32;
t += (uint64_t)(r->d[2] ^ mask) + (SECP256K1_N_2 & mask);
r->d[2] = t & nonzero; t >>= 32;
t += (uint64_t)(r->d[3] ^ mask) + (SECP256K1_N_3 & mask);
r->d[3] = t & nonzero; t >>= 32;
t += (uint64_t)(r->d[4] ^ mask) + (SECP256K1_N_4 & mask);
r->d[4] = t & nonzero; t >>= 32;
t += (uint64_t)(r->d[5] ^ mask) + (SECP256K1_N_5 & mask);
r->d[5] = t & nonzero; t >>= 32;
t += (uint64_t)(r->d[6] ^ mask) + (SECP256K1_N_6 & mask);
r->d[6] = t & nonzero; t >>= 32;
t += (uint64_t)(r->d[7] ^ mask) + (SECP256K1_N_7 & mask);
r->d[7] = t & nonzero;
SECP256K1_SCALAR_VERIFY(r);
return 2 * (mask == 0) - 1;
}
/* Inspired by the macros in OpenSSL's crypto/bn/asm/x86_64-gcc.c. */
/** Add a*b to the number defined by (c0,c1,c2). c2 must never overflow. */
#define muladd(a,b) { \
uint32_t tl, th; \
{ \
uint64_t t = (uint64_t)a * b; \
th = t >> 32; /* at most 0xFFFFFFFE */ \
tl = t; \
} \
c0 += tl; /* overflow is handled on the next line */ \
th += (c0 < tl); /* at most 0xFFFFFFFF */ \
c1 += th; /* overflow is handled on the next line */ \
c2 += (c1 < th); /* never overflows by contract (verified in the next line) */ \
VERIFY_CHECK((c1 >= th) || (c2 != 0)); \
}
/** Add a*b to the number defined by (c0,c1). c1 must never overflow. */
#define muladd_fast(a,b) { \
uint32_t tl, th; \
{ \
uint64_t t = (uint64_t)a * b; \
th = t >> 32; /* at most 0xFFFFFFFE */ \
tl = t; \
} \
c0 += tl; /* overflow is handled on the next line */ \
th += (c0 < tl); /* at most 0xFFFFFFFF */ \
c1 += th; /* never overflows by contract (verified in the next line) */ \
VERIFY_CHECK(c1 >= th); \
}
/** Add a to the number defined by (c0,c1,c2). c2 must never overflow. */
#define sumadd(a) { \
unsigned int over; \
c0 += (a); /* overflow is handled on the next line */ \
over = (c0 < (a)); \
c1 += over; /* overflow is handled on the next line */ \
c2 += (c1 < over); /* never overflows by contract */ \
}
/** Add a to the number defined by (c0,c1). c1 must never overflow, c2 must be zero. */
#define sumadd_fast(a) { \
c0 += (a); /* overflow is handled on the next line */ \
c1 += (c0 < (a)); /* never overflows by contract (verified the next line) */ \
VERIFY_CHECK((c1 != 0) | (c0 >= (a))); \
VERIFY_CHECK(c2 == 0); \
}
/** Extract the lowest 32 bits of (c0,c1,c2) into n, and left shift the number 32 bits. */
#define extract(n) { \
(n) = c0; \
c0 = c1; \
c1 = c2; \
c2 = 0; \
}
/** Extract the lowest 32 bits of (c0,c1,c2) into n, and left shift the number 32 bits. c2 is required to be zero. */
#define extract_fast(n) { \
(n) = c0; \
c0 = c1; \
c1 = 0; \
VERIFY_CHECK(c2 == 0); \
}
static void secp256k1_scalar_reduce_512(secp256k1_scalar *r, const uint32_t *l) {
uint64_t c;
uint32_t n0 = l[8], n1 = l[9], n2 = l[10], n3 = l[11], n4 = l[12], n5 = l[13], n6 = l[14], n7 = l[15];
uint32_t m0, m1, m2, m3, m4, m5, m6, m7, m8, m9, m10, m11, m12;
uint32_t p0, p1, p2, p3, p4, p5, p6, p7, p8;
/* 96 bit accumulator. */
uint32_t c0, c1, c2;
/* Reduce 512 bits into 385. */
/* m[0..12] = l[0..7] + n[0..7] * SECP256K1_N_C. */
c0 = l[0]; c1 = 0; c2 = 0;
muladd_fast(n0, SECP256K1_N_C_0);
extract_fast(m0);
sumadd_fast(l[1]);
muladd(n1, SECP256K1_N_C_0);
muladd(n0, SECP256K1_N_C_1);
extract(m1);
sumadd(l[2]);
muladd(n2, SECP256K1_N_C_0);
muladd(n1, SECP256K1_N_C_1);
muladd(n0, SECP256K1_N_C_2);
extract(m2);
sumadd(l[3]);
muladd(n3, SECP256K1_N_C_0);
muladd(n2, SECP256K1_N_C_1);
muladd(n1, SECP256K1_N_C_2);
muladd(n0, SECP256K1_N_C_3);
extract(m3);
sumadd(l[4]);
muladd(n4, SECP256K1_N_C_0);
muladd(n3, SECP256K1_N_C_1);
muladd(n2, SECP256K1_N_C_2);
muladd(n1, SECP256K1_N_C_3);
sumadd(n0);
extract(m4);
sumadd(l[5]);
muladd(n5, SECP256K1_N_C_0);
muladd(n4, SECP256K1_N_C_1);
muladd(n3, SECP256K1_N_C_2);
muladd(n2, SECP256K1_N_C_3);
sumadd(n1);
extract(m5);
sumadd(l[6]);
muladd(n6, SECP256K1_N_C_0);
muladd(n5, SECP256K1_N_C_1);
muladd(n4, SECP256K1_N_C_2);
muladd(n3, SECP256K1_N_C_3);
sumadd(n2);
extract(m6);
sumadd(l[7]);
muladd(n7, SECP256K1_N_C_0);
muladd(n6, SECP256K1_N_C_1);
muladd(n5, SECP256K1_N_C_2);
muladd(n4, SECP256K1_N_C_3);
sumadd(n3);
extract(m7);
muladd(n7, SECP256K1_N_C_1);
muladd(n6, SECP256K1_N_C_2);
muladd(n5, SECP256K1_N_C_3);
sumadd(n4);
extract(m8);
muladd(n7, SECP256K1_N_C_2);
muladd(n6, SECP256K1_N_C_3);
sumadd(n5);
extract(m9);
muladd(n7, SECP256K1_N_C_3);
sumadd(n6);
extract(m10);
sumadd_fast(n7);
extract_fast(m11);
VERIFY_CHECK(c0 <= 1);
m12 = c0;
/* Reduce 385 bits into 258. */
/* p[0..8] = m[0..7] + m[8..12] * SECP256K1_N_C. */
c0 = m0; c1 = 0; c2 = 0;
muladd_fast(m8, SECP256K1_N_C_0);
extract_fast(p0);
sumadd_fast(m1);
muladd(m9, SECP256K1_N_C_0);
muladd(m8, SECP256K1_N_C_1);
extract(p1);
sumadd(m2);
muladd(m10, SECP256K1_N_C_0);
muladd(m9, SECP256K1_N_C_1);
muladd(m8, SECP256K1_N_C_2);
extract(p2);
sumadd(m3);
muladd(m11, SECP256K1_N_C_0);
muladd(m10, SECP256K1_N_C_1);
muladd(m9, SECP256K1_N_C_2);
muladd(m8, SECP256K1_N_C_3);
extract(p3);
sumadd(m4);
muladd(m12, SECP256K1_N_C_0);
muladd(m11, SECP256K1_N_C_1);
muladd(m10, SECP256K1_N_C_2);
muladd(m9, SECP256K1_N_C_3);
sumadd(m8);
extract(p4);
sumadd(m5);
muladd(m12, SECP256K1_N_C_1);
muladd(m11, SECP256K1_N_C_2);
muladd(m10, SECP256K1_N_C_3);
sumadd(m9);
extract(p5);
sumadd(m6);
muladd(m12, SECP256K1_N_C_2);
muladd(m11, SECP256K1_N_C_3);
sumadd(m10);
extract(p6);
sumadd_fast(m7);
muladd_fast(m12, SECP256K1_N_C_3);
sumadd_fast(m11);
extract_fast(p7);
p8 = c0 + m12;
VERIFY_CHECK(p8 <= 2);
/* Reduce 258 bits into 256. */
/* r[0..7] = p[0..7] + p[8] * SECP256K1_N_C. */
c = p0 + (uint64_t)SECP256K1_N_C_0 * p8;
r->d[0] = c & 0xFFFFFFFFUL; c >>= 32;
c += p1 + (uint64_t)SECP256K1_N_C_1 * p8;
r->d[1] = c & 0xFFFFFFFFUL; c >>= 32;
c += p2 + (uint64_t)SECP256K1_N_C_2 * p8;
r->d[2] = c & 0xFFFFFFFFUL; c >>= 32;
c += p3 + (uint64_t)SECP256K1_N_C_3 * p8;
r->d[3] = c & 0xFFFFFFFFUL; c >>= 32;
c += p4 + (uint64_t)p8;
r->d[4] = c & 0xFFFFFFFFUL; c >>= 32;
c += p5;
r->d[5] = c & 0xFFFFFFFFUL; c >>= 32;
c += p6;
r->d[6] = c & 0xFFFFFFFFUL; c >>= 32;
c += p7;
r->d[7] = c & 0xFFFFFFFFUL; c >>= 32;
/* Final reduction of r. */
secp256k1_scalar_reduce(r, c + secp256k1_scalar_check_overflow(r));
}
static void secp256k1_scalar_mul_512(uint32_t *l, const secp256k1_scalar *a, const secp256k1_scalar *b) {
/* 96 bit accumulator. */
uint32_t c0 = 0, c1 = 0, c2 = 0;
/* l[0..15] = a[0..7] * b[0..7]. */
muladd_fast(a->d[0], b->d[0]);
extract_fast(l[0]);
muladd(a->d[0], b->d[1]);
muladd(a->d[1], b->d[0]);
extract(l[1]);
muladd(a->d[0], b->d[2]);
muladd(a->d[1], b->d[1]);
muladd(a->d[2], b->d[0]);
extract(l[2]);
muladd(a->d[0], b->d[3]);
muladd(a->d[1], b->d[2]);
muladd(a->d[2], b->d[1]);
muladd(a->d[3], b->d[0]);
extract(l[3]);
muladd(a->d[0], b->d[4]);
muladd(a->d[1], b->d[3]);
muladd(a->d[2], b->d[2]);
muladd(a->d[3], b->d[1]);
muladd(a->d[4], b->d[0]);
extract(l[4]);
muladd(a->d[0], b->d[5]);
muladd(a->d[1], b->d[4]);
muladd(a->d[2], b->d[3]);
muladd(a->d[3], b->d[2]);
muladd(a->d[4], b->d[1]);
muladd(a->d[5], b->d[0]);
extract(l[5]);
muladd(a->d[0], b->d[6]);
muladd(a->d[1], b->d[5]);
muladd(a->d[2], b->d[4]);
muladd(a->d[3], b->d[3]);
muladd(a->d[4], b->d[2]);
muladd(a->d[5], b->d[1]);
muladd(a->d[6], b->d[0]);
extract(l[6]);
muladd(a->d[0], b->d[7]);
muladd(a->d[1], b->d[6]);
muladd(a->d[2], b->d[5]);
muladd(a->d[3], b->d[4]);
muladd(a->d[4], b->d[3]);
muladd(a->d[5], b->d[2]);
muladd(a->d[6], b->d[1]);
muladd(a->d[7], b->d[0]);
extract(l[7]);
muladd(a->d[1], b->d[7]);
muladd(a->d[2], b->d[6]);
muladd(a->d[3], b->d[5]);
muladd(a->d[4], b->d[4]);
muladd(a->d[5], b->d[3]);
muladd(a->d[6], b->d[2]);
muladd(a->d[7], b->d[1]);
extract(l[8]);
muladd(a->d[2], b->d[7]);
muladd(a->d[3], b->d[6]);
muladd(a->d[4], b->d[5]);
muladd(a->d[5], b->d[4]);
muladd(a->d[6], b->d[3]);
muladd(a->d[7], b->d[2]);
extract(l[9]);
muladd(a->d[3], b->d[7]);
muladd(a->d[4], b->d[6]);
muladd(a->d[5], b->d[5]);
muladd(a->d[6], b->d[4]);
muladd(a->d[7], b->d[3]);
extract(l[10]);
muladd(a->d[4], b->d[7]);
muladd(a->d[5], b->d[6]);
muladd(a->d[6], b->d[5]);
muladd(a->d[7], b->d[4]);
extract(l[11]);
muladd(a->d[5], b->d[7]);
muladd(a->d[6], b->d[6]);
muladd(a->d[7], b->d[5]);
extract(l[12]);
muladd(a->d[6], b->d[7]);
muladd(a->d[7], b->d[6]);
extract(l[13]);
muladd_fast(a->d[7], b->d[7]);
extract_fast(l[14]);
VERIFY_CHECK(c1 == 0);
l[15] = c0;
}
#undef sumadd
#undef sumadd_fast
#undef muladd
#undef muladd_fast
#undef extract
#undef extract_fast
static void secp256k1_scalar_mul(secp256k1_scalar *r, const secp256k1_scalar *a, const secp256k1_scalar *b) {
uint32_t l[16];
SECP256K1_SCALAR_VERIFY(a);
SECP256K1_SCALAR_VERIFY(b);
secp256k1_scalar_mul_512(l, a, b);
secp256k1_scalar_reduce_512(r, l);
SECP256K1_SCALAR_VERIFY(r);
}
static void secp256k1_scalar_split_128(secp256k1_scalar *r1, secp256k1_scalar *r2, const secp256k1_scalar *k) {
SECP256K1_SCALAR_VERIFY(k);
r1->d[0] = k->d[0];
r1->d[1] = k->d[1];
r1->d[2] = k->d[2];
r1->d[3] = k->d[3];
r1->d[4] = 0;
r1->d[5] = 0;
r1->d[6] = 0;
r1->d[7] = 0;
r2->d[0] = k->d[4];
r2->d[1] = k->d[5];
r2->d[2] = k->d[6];
r2->d[3] = k->d[7];
r2->d[4] = 0;
r2->d[5] = 0;
r2->d[6] = 0;
r2->d[7] = 0;
SECP256K1_SCALAR_VERIFY(r1);
SECP256K1_SCALAR_VERIFY(r2);
}
SECP256K1_INLINE static int secp256k1_scalar_eq(const secp256k1_scalar *a, const secp256k1_scalar *b) {
SECP256K1_SCALAR_VERIFY(a);
SECP256K1_SCALAR_VERIFY(b);
return ((a->d[0] ^ b->d[0]) | (a->d[1] ^ b->d[1]) | (a->d[2] ^ b->d[2]) | (a->d[3] ^ b->d[3]) | (a->d[4] ^ b->d[4]) | (a->d[5] ^ b->d[5]) | (a->d[6] ^ b->d[6]) | (a->d[7] ^ b->d[7])) == 0;
}
SECP256K1_INLINE static void secp256k1_scalar_mul_shift_var(secp256k1_scalar *r, const secp256k1_scalar *a, const secp256k1_scalar *b, unsigned int shift) {
uint32_t l[16];
unsigned int shiftlimbs;
unsigned int shiftlow;
unsigned int shifthigh;
SECP256K1_SCALAR_VERIFY(a);
SECP256K1_SCALAR_VERIFY(b);
VERIFY_CHECK(shift >= 256);
secp256k1_scalar_mul_512(l, a, b);
shiftlimbs = shift >> 5;
shiftlow = shift & 0x1F;
shifthigh = 32 - shiftlow;
r->d[0] = shift < 512 ? (l[0 + shiftlimbs] >> shiftlow | (shift < 480 && shiftlow ? (l[1 + shiftlimbs] << shifthigh) : 0)) : 0;
r->d[1] = shift < 480 ? (l[1 + shiftlimbs] >> shiftlow | (shift < 448 && shiftlow ? (l[2 + shiftlimbs] << shifthigh) : 0)) : 0;
r->d[2] = shift < 448 ? (l[2 + shiftlimbs] >> shiftlow | (shift < 416 && shiftlow ? (l[3 + shiftlimbs] << shifthigh) : 0)) : 0;
r->d[3] = shift < 416 ? (l[3 + shiftlimbs] >> shiftlow | (shift < 384 && shiftlow ? (l[4 + shiftlimbs] << shifthigh) : 0)) : 0;
r->d[4] = shift < 384 ? (l[4 + shiftlimbs] >> shiftlow | (shift < 352 && shiftlow ? (l[5 + shiftlimbs] << shifthigh) : 0)) : 0;
r->d[5] = shift < 352 ? (l[5 + shiftlimbs] >> shiftlow | (shift < 320 && shiftlow ? (l[6 + shiftlimbs] << shifthigh) : 0)) : 0;
r->d[6] = shift < 320 ? (l[6 + shiftlimbs] >> shiftlow | (shift < 288 && shiftlow ? (l[7 + shiftlimbs] << shifthigh) : 0)) : 0;
r->d[7] = shift < 288 ? (l[7 + shiftlimbs] >> shiftlow) : 0;
secp256k1_scalar_cadd_bit(r, 0, (l[(shift - 1) >> 5] >> ((shift - 1) & 0x1f)) & 1);
SECP256K1_SCALAR_VERIFY(r);
}
static SECP256K1_INLINE void secp256k1_scalar_cmov(secp256k1_scalar *r, const secp256k1_scalar *a, int flag) {
uint32_t mask0, mask1;
volatile int vflag = flag;
SECP256K1_SCALAR_VERIFY(a);
SECP256K1_CHECKMEM_CHECK_VERIFY(r->d, sizeof(r->d));
mask0 = vflag + ~((uint32_t)0);
mask1 = ~mask0;
r->d[0] = (r->d[0] & mask0) | (a->d[0] & mask1);
r->d[1] = (r->d[1] & mask0) | (a->d[1] & mask1);
r->d[2] = (r->d[2] & mask0) | (a->d[2] & mask1);
r->d[3] = (r->d[3] & mask0) | (a->d[3] & mask1);
r->d[4] = (r->d[4] & mask0) | (a->d[4] & mask1);
r->d[5] = (r->d[5] & mask0) | (a->d[5] & mask1);
r->d[6] = (r->d[6] & mask0) | (a->d[6] & mask1);
r->d[7] = (r->d[7] & mask0) | (a->d[7] & mask1);
SECP256K1_SCALAR_VERIFY(r);
}
static void secp256k1_scalar_from_signed30(secp256k1_scalar *r, const secp256k1_modinv32_signed30 *a) {
const uint32_t a0 = a->v[0], a1 = a->v[1], a2 = a->v[2], a3 = a->v[3], a4 = a->v[4],
a5 = a->v[5], a6 = a->v[6], a7 = a->v[7], a8 = a->v[8];
/* The output from secp256k1_modinv32{_var} should be normalized to range [0,modulus), and
* have limbs in [0,2^30). The modulus is < 2^256, so the top limb must be below 2^(256-30*8).
*/
VERIFY_CHECK(a0 >> 30 == 0);
VERIFY_CHECK(a1 >> 30 == 0);
VERIFY_CHECK(a2 >> 30 == 0);
VERIFY_CHECK(a3 >> 30 == 0);
VERIFY_CHECK(a4 >> 30 == 0);
VERIFY_CHECK(a5 >> 30 == 0);
VERIFY_CHECK(a6 >> 30 == 0);
VERIFY_CHECK(a7 >> 30 == 0);
VERIFY_CHECK(a8 >> 16 == 0);
r->d[0] = a0 | a1 << 30;
r->d[1] = a1 >> 2 | a2 << 28;
r->d[2] = a2 >> 4 | a3 << 26;
r->d[3] = a3 >> 6 | a4 << 24;
r->d[4] = a4 >> 8 | a5 << 22;
r->d[5] = a5 >> 10 | a6 << 20;
r->d[6] = a6 >> 12 | a7 << 18;
r->d[7] = a7 >> 14 | a8 << 16;
SECP256K1_SCALAR_VERIFY(r);
}
static void secp256k1_scalar_to_signed30(secp256k1_modinv32_signed30 *r, const secp256k1_scalar *a) {
const uint32_t M30 = UINT32_MAX >> 2;
const uint32_t a0 = a->d[0], a1 = a->d[1], a2 = a->d[2], a3 = a->d[3],
a4 = a->d[4], a5 = a->d[5], a6 = a->d[6], a7 = a->d[7];
SECP256K1_SCALAR_VERIFY(a);
r->v[0] = a0 & M30;
r->v[1] = (a0 >> 30 | a1 << 2) & M30;
r->v[2] = (a1 >> 28 | a2 << 4) & M30;
r->v[3] = (a2 >> 26 | a3 << 6) & M30;
r->v[4] = (a3 >> 24 | a4 << 8) & M30;
r->v[5] = (a4 >> 22 | a5 << 10) & M30;
r->v[6] = (a5 >> 20 | a6 << 12) & M30;
r->v[7] = (a6 >> 18 | a7 << 14) & M30;
r->v[8] = a7 >> 16;
}
static const secp256k1_modinv32_modinfo secp256k1_const_modinfo_scalar = {
{{0x10364141L, 0x3F497A33L, 0x348A03BBL, 0x2BB739ABL, -0x146L, 0, 0, 0, 65536}},
0x2A774EC1L
};
static void secp256k1_scalar_inverse(secp256k1_scalar *r, const secp256k1_scalar *x) {
secp256k1_modinv32_signed30 s;
#ifdef VERIFY
int zero_in = secp256k1_scalar_is_zero(x);
#endif
SECP256K1_SCALAR_VERIFY(x);
secp256k1_scalar_to_signed30(&s, x);
secp256k1_modinv32(&s, &secp256k1_const_modinfo_scalar);
secp256k1_scalar_from_signed30(r, &s);
SECP256K1_SCALAR_VERIFY(r);
VERIFY_CHECK(secp256k1_scalar_is_zero(r) == zero_in);
}
static void secp256k1_scalar_inverse_var(secp256k1_scalar *r, const secp256k1_scalar *x) {
secp256k1_modinv32_signed30 s;
#ifdef VERIFY
int zero_in = secp256k1_scalar_is_zero(x);
#endif
SECP256K1_SCALAR_VERIFY(x);
secp256k1_scalar_to_signed30(&s, x);
secp256k1_modinv32_var(&s, &secp256k1_const_modinfo_scalar);
secp256k1_scalar_from_signed30(r, &s);
SECP256K1_SCALAR_VERIFY(r);
VERIFY_CHECK(secp256k1_scalar_is_zero(r) == zero_in);
}
SECP256K1_INLINE static int secp256k1_scalar_is_even(const secp256k1_scalar *a) {
SECP256K1_SCALAR_VERIFY(a);
return !(a->d[0] & 1);
}
#endif /* SECP256K1_SCALAR_REPR_IMPL_H */

View File

@@ -0,0 +1,321 @@
/***********************************************************************
* Copyright (c) 2014 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_SCALAR_IMPL_H
#define SECP256K1_SCALAR_IMPL_H
#ifdef VERIFY
#include <string.h>
#endif
#include "scalar.h"
#include "util.h"
#if defined(EXHAUSTIVE_TEST_ORDER)
#include "scalar_low_impl.h"
#elif defined(SECP256K1_WIDEMUL_INT128)
#include "scalar_4x64_impl.h"
#elif defined(SECP256K1_WIDEMUL_INT64)
#include "scalar_8x32_impl.h"
#else
#error "Please select wide multiplication implementation"
#endif
static const secp256k1_scalar secp256k1_scalar_one = SECP256K1_SCALAR_CONST(0, 0, 0, 0, 0, 0, 0, 1);
static const secp256k1_scalar secp256k1_scalar_zero = SECP256K1_SCALAR_CONST(0, 0, 0, 0, 0, 0, 0, 0);
SECP256K1_INLINE static void secp256k1_scalar_clear(secp256k1_scalar *r) {
secp256k1_memclear(r, sizeof(secp256k1_scalar));
}
static int secp256k1_scalar_set_b32_seckey(secp256k1_scalar *r, const unsigned char *bin) {
int overflow;
secp256k1_scalar_set_b32(r, bin, &overflow);
SECP256K1_SCALAR_VERIFY(r);
return (!overflow) & (!secp256k1_scalar_is_zero(r));
}
static void secp256k1_scalar_verify(const secp256k1_scalar *r) {
VERIFY_CHECK(secp256k1_scalar_check_overflow(r) == 0);
(void)r;
}
#if defined(EXHAUSTIVE_TEST_ORDER)
/* Begin of section generated by sage/gen_exhaustive_groups.sage. */
# if EXHAUSTIVE_TEST_ORDER == 7
# define EXHAUSTIVE_TEST_LAMBDA 2
# elif EXHAUSTIVE_TEST_ORDER == 13
# define EXHAUSTIVE_TEST_LAMBDA 9
# elif EXHAUSTIVE_TEST_ORDER == 199
# define EXHAUSTIVE_TEST_LAMBDA 92
# else
# error No known lambda for the specified exhaustive test group order.
# endif
/* End of section generated by sage/gen_exhaustive_groups.sage. */
/**
* Find r1 and r2 given k, such that r1 + r2 * lambda == k mod n; unlike in the
* full case we don't bother making r1 and r2 be small, we just want them to be
* nontrivial to get full test coverage for the exhaustive tests. We therefore
* (arbitrarily) set r2 = k + 5 (mod n) and r1 = k - r2 * lambda (mod n).
*/
static void secp256k1_scalar_split_lambda(secp256k1_scalar * SECP256K1_RESTRICT r1, secp256k1_scalar * SECP256K1_RESTRICT r2, const secp256k1_scalar * SECP256K1_RESTRICT k) {
SECP256K1_SCALAR_VERIFY(k);
VERIFY_CHECK(r1 != k);
VERIFY_CHECK(r2 != k);
VERIFY_CHECK(r1 != r2);
*r2 = (*k + 5) % EXHAUSTIVE_TEST_ORDER;
*r1 = (*k + (EXHAUSTIVE_TEST_ORDER - *r2) * EXHAUSTIVE_TEST_LAMBDA) % EXHAUSTIVE_TEST_ORDER;
SECP256K1_SCALAR_VERIFY(r1);
SECP256K1_SCALAR_VERIFY(r2);
}
#else
/**
* The Secp256k1 curve has an endomorphism, where lambda * (x, y) = (beta * x, y), where
* lambda is: */
static const secp256k1_scalar secp256k1_const_lambda = SECP256K1_SCALAR_CONST(
0x5363AD4CUL, 0xC05C30E0UL, 0xA5261C02UL, 0x8812645AUL,
0x122E22EAUL, 0x20816678UL, 0xDF02967CUL, 0x1B23BD72UL
);
#ifdef VERIFY
static void secp256k1_scalar_split_lambda_verify(const secp256k1_scalar *r1, const secp256k1_scalar *r2, const secp256k1_scalar *k);
#endif
/*
* Both lambda and beta are primitive cube roots of unity. That is lamba^3 == 1 mod n and
* beta^3 == 1 mod p, where n is the curve order and p is the field order.
*
* Furthermore, because (X^3 - 1) = (X - 1)(X^2 + X + 1), the primitive cube roots of unity are
* roots of X^2 + X + 1. Therefore lambda^2 + lamba == -1 mod n and beta^2 + beta == -1 mod p.
* (The other primitive cube roots of unity are lambda^2 and beta^2 respectively.)
*
* Let l = -1/2 + i*sqrt(3)/2, the complex root of X^2 + X + 1. We can define a ring
* homomorphism phi : Z[l] -> Z_n where phi(a + b*l) == a + b*lambda mod n. The kernel of phi
* is a lattice over Z[l] (considering Z[l] as a Z-module). This lattice is generated by a
* reduced basis {a1 + b1*l, a2 + b2*l} where
*
* - a1 = {0x30,0x86,0xd2,0x21,0xa7,0xd4,0x6b,0xcd,0xe8,0x6c,0x90,0xe4,0x92,0x84,0xeb,0x15}
* - b1 = -{0xe4,0x43,0x7e,0xd6,0x01,0x0e,0x88,0x28,0x6f,0x54,0x7f,0xa9,0x0a,0xbf,0xe4,0xc3}
* - a2 = {0x01,0x14,0xca,0x50,0xf7,0xa8,0xe2,0xf3,0xf6,0x57,0xc1,0x10,0x8d,0x9d,0x44,0xcf,0xd8}
* - b2 = {0x30,0x86,0xd2,0x21,0xa7,0xd4,0x6b,0xcd,0xe8,0x6c,0x90,0xe4,0x92,0x84,0xeb,0x15}
*
* "Guide to Elliptic Curve Cryptography" (Hankerson, Menezes, Vanstone) gives an algorithm
* (algorithm 3.74) to find k1 and k2 given k, such that k1 + k2 * lambda == k mod n, and k1
* and k2 are small in absolute value.
*
* The algorithm computes c1 = round(b2 * k / n) and c2 = round((-b1) * k / n), and gives
* k1 = k - (c1*a1 + c2*a2) and k2 = -(c1*b1 + c2*b2). Instead, we use modular arithmetic, and
* compute r2 = k2 mod n, and r1 = k1 mod n = (k - r2 * lambda) mod n, avoiding the need for
* the constants a1 and a2.
*
* g1, g2 are precomputed constants used to replace division with a rounded multiplication
* when decomposing the scalar for an endomorphism-based point multiplication.
*
* The possibility of using precomputed estimates is mentioned in "Guide to Elliptic Curve
* Cryptography" (Hankerson, Menezes, Vanstone) in section 3.5.
*
* The derivation is described in the paper "Efficient Software Implementation of Public-Key
* Cryptography on Sensor Networks Using the MSP430X Microcontroller" (Gouvea, Oliveira, Lopez),
* Section 4.3 (here we use a somewhat higher-precision estimate):
* d = a1*b2 - b1*a2
* g1 = round(2^384 * b2/d)
* g2 = round(2^384 * (-b1)/d)
*
* (Note that d is also equal to the curve order, n, here because [a1,b1] and [a2,b2]
* can be found as outputs of the Extended Euclidean Algorithm on inputs n and lambda).
*
* The function below splits k into r1 and r2, such that
* - r1 + lambda * r2 == k (mod n)
* - either r1 < 2^128 or -r1 mod n < 2^128
* - either r2 < 2^128 or -r2 mod n < 2^128
*
* See proof below.
*/
static void secp256k1_scalar_split_lambda(secp256k1_scalar * SECP256K1_RESTRICT r1, secp256k1_scalar * SECP256K1_RESTRICT r2, const secp256k1_scalar * SECP256K1_RESTRICT k) {
secp256k1_scalar c1, c2;
static const secp256k1_scalar minus_b1 = SECP256K1_SCALAR_CONST(
0x00000000UL, 0x00000000UL, 0x00000000UL, 0x00000000UL,
0xE4437ED6UL, 0x010E8828UL, 0x6F547FA9UL, 0x0ABFE4C3UL
);
static const secp256k1_scalar minus_b2 = SECP256K1_SCALAR_CONST(
0xFFFFFFFFUL, 0xFFFFFFFFUL, 0xFFFFFFFFUL, 0xFFFFFFFEUL,
0x8A280AC5UL, 0x0774346DUL, 0xD765CDA8UL, 0x3DB1562CUL
);
static const secp256k1_scalar g1 = SECP256K1_SCALAR_CONST(
0x3086D221UL, 0xA7D46BCDUL, 0xE86C90E4UL, 0x9284EB15UL,
0x3DAA8A14UL, 0x71E8CA7FUL, 0xE893209AUL, 0x45DBB031UL
);
static const secp256k1_scalar g2 = SECP256K1_SCALAR_CONST(
0xE4437ED6UL, 0x010E8828UL, 0x6F547FA9UL, 0x0ABFE4C4UL,
0x221208ACUL, 0x9DF506C6UL, 0x1571B4AEUL, 0x8AC47F71UL
);
SECP256K1_SCALAR_VERIFY(k);
VERIFY_CHECK(r1 != k);
VERIFY_CHECK(r2 != k);
VERIFY_CHECK(r1 != r2);
/* these _var calls are constant time since the shift amount is constant */
secp256k1_scalar_mul_shift_var(&c1, k, &g1, 384);
secp256k1_scalar_mul_shift_var(&c2, k, &g2, 384);
secp256k1_scalar_mul(&c1, &c1, &minus_b1);
secp256k1_scalar_mul(&c2, &c2, &minus_b2);
secp256k1_scalar_add(r2, &c1, &c2);
secp256k1_scalar_mul(r1, r2, &secp256k1_const_lambda);
secp256k1_scalar_negate(r1, r1);
secp256k1_scalar_add(r1, r1, k);
SECP256K1_SCALAR_VERIFY(r1);
SECP256K1_SCALAR_VERIFY(r2);
#ifdef VERIFY
secp256k1_scalar_split_lambda_verify(r1, r2, k);
#endif
}
#ifdef VERIFY
/*
* Proof for secp256k1_scalar_split_lambda's bounds.
*
* Let
* - epsilon1 = 2^256 * |g1/2^384 - b2/d|
* - epsilon2 = 2^256 * |g2/2^384 - (-b1)/d|
* - c1 = round(k*g1/2^384)
* - c2 = round(k*g2/2^384)
*
* Lemma 1: |c1 - k*b2/d| < 2^-1 + epsilon1
*
* |c1 - k*b2/d|
* =
* |c1 - k*g1/2^384 + k*g1/2^384 - k*b2/d|
* <= {triangle inequality}
* |c1 - k*g1/2^384| + |k*g1/2^384 - k*b2/d|
* =
* |c1 - k*g1/2^384| + k*|g1/2^384 - b2/d|
* < {rounding in c1 and 0 <= k < 2^256}
* 2^-1 + 2^256 * |g1/2^384 - b2/d|
* = {definition of epsilon1}
* 2^-1 + epsilon1
*
* Lemma 2: |c2 - k*(-b1)/d| < 2^-1 + epsilon2
*
* |c2 - k*(-b1)/d|
* =
* |c2 - k*g2/2^384 + k*g2/2^384 - k*(-b1)/d|
* <= {triangle inequality}
* |c2 - k*g2/2^384| + |k*g2/2^384 - k*(-b1)/d|
* =
* |c2 - k*g2/2^384| + k*|g2/2^384 - (-b1)/d|
* < {rounding in c2 and 0 <= k < 2^256}
* 2^-1 + 2^256 * |g2/2^384 - (-b1)/d|
* = {definition of epsilon2}
* 2^-1 + epsilon2
*
* Let
* - k1 = k - c1*a1 - c2*a2
* - k2 = - c1*b1 - c2*b2
*
* Lemma 3: |k1| < (a1 + a2 + 1)/2 < 2^128
*
* |k1|
* = {definition of k1}
* |k - c1*a1 - c2*a2|
* = {(a1*b2 - b1*a2)/n = 1}
* |k*(a1*b2 - b1*a2)/n - c1*a1 - c2*a2|
* =
* |a1*(k*b2/n - c1) + a2*(k*(-b1)/n - c2)|
* <= {triangle inequality}
* a1*|k*b2/n - c1| + a2*|k*(-b1)/n - c2|
* < {Lemma 1 and Lemma 2}
* a1*(2^-1 + epsilon1) + a2*(2^-1 + epsilon2)
* < {rounding up to an integer}
* (a1 + a2 + 1)/2
* < {rounding up to a power of 2}
* 2^128
*
* Lemma 4: |k2| < (-b1 + b2)/2 + 1 < 2^128
*
* |k2|
* = {definition of k2}
* |- c1*a1 - c2*a2|
* = {(b1*b2 - b1*b2)/n = 0}
* |k*(b1*b2 - b1*b2)/n - c1*b1 - c2*b2|
* =
* |b1*(k*b2/n - c1) + b2*(k*(-b1)/n - c2)|
* <= {triangle inequality}
* (-b1)*|k*b2/n - c1| + b2*|k*(-b1)/n - c2|
* < {Lemma 1 and Lemma 2}
* (-b1)*(2^-1 + epsilon1) + b2*(2^-1 + epsilon2)
* < {rounding up to an integer}
* (-b1 + b2)/2 + 1
* < {rounding up to a power of 2}
* 2^128
*
* Let
* - r2 = k2 mod n
* - r1 = k - r2*lambda mod n.
*
* Notice that r1 is defined such that r1 + r2 * lambda == k (mod n).
*
* Lemma 5: r1 == k1 mod n.
*
* r1
* == {definition of r1 and r2}
* k - k2*lambda
* == {definition of k2}
* k - (- c1*b1 - c2*b2)*lambda
* ==
* k + c1*b1*lambda + c2*b2*lambda
* == {a1 + b1*lambda == 0 mod n and a2 + b2*lambda == 0 mod n}
* k - c1*a1 - c2*a2
* == {definition of k1}
* k1
*
* From Lemma 3, Lemma 4, Lemma 5 and the definition of r2, we can conclude that
*
* - either r1 < 2^128 or -r1 mod n < 2^128
* - either r2 < 2^128 or -r2 mod n < 2^128.
*
* Q.E.D.
*/
static void secp256k1_scalar_split_lambda_verify(const secp256k1_scalar *r1, const secp256k1_scalar *r2, const secp256k1_scalar *k) {
secp256k1_scalar s;
unsigned char buf1[32];
unsigned char buf2[32];
/* (a1 + a2 + 1)/2 is 0xa2a8918ca85bafe22016d0b917e4dd77 */
static const unsigned char k1_bound[32] = {
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0xa2, 0xa8, 0x91, 0x8c, 0xa8, 0x5b, 0xaf, 0xe2, 0x20, 0x16, 0xd0, 0xb9, 0x17, 0xe4, 0xdd, 0x77
};
/* (-b1 + b2)/2 + 1 is 0x8a65287bd47179fb2be08846cea267ed */
static const unsigned char k2_bound[32] = {
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x8a, 0x65, 0x28, 0x7b, 0xd4, 0x71, 0x79, 0xfb, 0x2b, 0xe0, 0x88, 0x46, 0xce, 0xa2, 0x67, 0xed
};
secp256k1_scalar_mul(&s, &secp256k1_const_lambda, r2);
secp256k1_scalar_add(&s, &s, r1);
VERIFY_CHECK(secp256k1_scalar_eq(&s, k));
secp256k1_scalar_negate(&s, r1);
secp256k1_scalar_get_b32(buf1, r1);
secp256k1_scalar_get_b32(buf2, &s);
VERIFY_CHECK(secp256k1_memcmp_var(buf1, k1_bound, 32) < 0 || secp256k1_memcmp_var(buf2, k1_bound, 32) < 0);
secp256k1_scalar_negate(&s, r2);
secp256k1_scalar_get_b32(buf1, r2);
secp256k1_scalar_get_b32(buf2, &s);
VERIFY_CHECK(secp256k1_memcmp_var(buf1, k2_bound, 32) < 0 || secp256k1_memcmp_var(buf2, k2_bound, 32) < 0);
}
#endif /* VERIFY */
#endif /* !defined(EXHAUSTIVE_TEST_ORDER) */
#endif /* SECP256K1_SCALAR_IMPL_H */

View File

@@ -0,0 +1,24 @@
/***********************************************************************
* Copyright (c) 2015, 2022 Andrew Poelstra, Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_SCALAR_REPR_H
#define SECP256K1_SCALAR_REPR_H
#include <stdint.h>
/** A scalar modulo the group order of the secp256k1 curve. */
typedef uint32_t secp256k1_scalar;
/* A compile-time constant equal to 2^32 (modulo order). */
#define SCALAR_2P32 ((0xffffffffUL % EXHAUSTIVE_TEST_ORDER) + 1U)
/* Compute a*2^32 + b (modulo order). */
#define SCALAR_HORNER(a, b) (((uint64_t)(a) * SCALAR_2P32 + (b)) % EXHAUSTIVE_TEST_ORDER)
/* Evaluates to the provided 256-bit constant reduced modulo order. */
#define SECP256K1_SCALAR_CONST(d7, d6, d5, d4, d3, d2, d1, d0) SCALAR_HORNER(SCALAR_HORNER(SCALAR_HORNER(SCALAR_HORNER(SCALAR_HORNER(SCALAR_HORNER(SCALAR_HORNER((d7), (d6)), (d5)), (d4)), (d3)), (d2)), (d1)), (d0))
#endif /* SECP256K1_SCALAR_REPR_H */

View File

@@ -0,0 +1,206 @@
/***********************************************************************
* Copyright (c) 2015 Andrew Poelstra *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_SCALAR_REPR_IMPL_H
#define SECP256K1_SCALAR_REPR_IMPL_H
#include "checkmem.h"
#include "scalar.h"
#include "util.h"
#include <string.h>
SECP256K1_INLINE static int secp256k1_scalar_is_even(const secp256k1_scalar *a) {
SECP256K1_SCALAR_VERIFY(a);
return !(*a & 1);
}
SECP256K1_INLINE static void secp256k1_scalar_set_int(secp256k1_scalar *r, unsigned int v) {
*r = v % EXHAUSTIVE_TEST_ORDER;
SECP256K1_SCALAR_VERIFY(r);
}
SECP256K1_INLINE static uint32_t secp256k1_scalar_get_bits_limb32(const secp256k1_scalar *a, unsigned int offset, unsigned int count) {
SECP256K1_SCALAR_VERIFY(a);
VERIFY_CHECK(count > 0 && count <= 32);
if (offset < 32) {
return (*a >> offset) & (0xFFFFFFFF >> (32 - count));
} else {
return 0;
}
}
SECP256K1_INLINE static uint32_t secp256k1_scalar_get_bits_var(const secp256k1_scalar *a, unsigned int offset, unsigned int count) {
SECP256K1_SCALAR_VERIFY(a);
return secp256k1_scalar_get_bits_limb32(a, offset, count);
}
SECP256K1_INLINE static int secp256k1_scalar_check_overflow(const secp256k1_scalar *a) { return *a >= EXHAUSTIVE_TEST_ORDER; }
static int secp256k1_scalar_add(secp256k1_scalar *r, const secp256k1_scalar *a, const secp256k1_scalar *b) {
SECP256K1_SCALAR_VERIFY(a);
SECP256K1_SCALAR_VERIFY(b);
*r = (*a + *b) % EXHAUSTIVE_TEST_ORDER;
SECP256K1_SCALAR_VERIFY(r);
return *r < *b;
}
static void secp256k1_scalar_cadd_bit(secp256k1_scalar *r, unsigned int bit, int flag) {
SECP256K1_SCALAR_VERIFY(r);
if (flag && bit < 32)
*r += ((uint32_t)1 << bit);
SECP256K1_SCALAR_VERIFY(r);
VERIFY_CHECK(bit < 32);
/* Verify that adding (1 << bit) will not overflow any in-range scalar *r by overflowing the underlying uint32_t. */
VERIFY_CHECK(((uint32_t)1 << bit) - 1 <= UINT32_MAX - EXHAUSTIVE_TEST_ORDER);
}
static void secp256k1_scalar_set_b32(secp256k1_scalar *r, const unsigned char *b32, int *overflow) {
int i;
int over = 0;
*r = 0;
for (i = 0; i < 32; i++) {
*r = (*r * 0x100) + b32[i];
if (*r >= EXHAUSTIVE_TEST_ORDER) {
over = 1;
*r %= EXHAUSTIVE_TEST_ORDER;
}
}
if (overflow) *overflow = over;
SECP256K1_SCALAR_VERIFY(r);
}
static void secp256k1_scalar_get_b32(unsigned char *bin, const secp256k1_scalar* a) {
SECP256K1_SCALAR_VERIFY(a);
memset(bin, 0, 32);
bin[28] = *a >> 24; bin[29] = *a >> 16; bin[30] = *a >> 8; bin[31] = *a;
}
SECP256K1_INLINE static int secp256k1_scalar_is_zero(const secp256k1_scalar *a) {
SECP256K1_SCALAR_VERIFY(a);
return *a == 0;
}
static void secp256k1_scalar_negate(secp256k1_scalar *r, const secp256k1_scalar *a) {
SECP256K1_SCALAR_VERIFY(a);
if (*a == 0) {
*r = 0;
} else {
*r = EXHAUSTIVE_TEST_ORDER - *a;
}
SECP256K1_SCALAR_VERIFY(r);
}
SECP256K1_INLINE static int secp256k1_scalar_is_one(const secp256k1_scalar *a) {
SECP256K1_SCALAR_VERIFY(a);
return *a == 1;
}
static int secp256k1_scalar_is_high(const secp256k1_scalar *a) {
SECP256K1_SCALAR_VERIFY(a);
return *a > EXHAUSTIVE_TEST_ORDER / 2;
}
static int secp256k1_scalar_cond_negate(secp256k1_scalar *r, int flag) {
SECP256K1_SCALAR_VERIFY(r);
if (flag) secp256k1_scalar_negate(r, r);
SECP256K1_SCALAR_VERIFY(r);
return flag ? -1 : 1;
}
static void secp256k1_scalar_mul(secp256k1_scalar *r, const secp256k1_scalar *a, const secp256k1_scalar *b) {
SECP256K1_SCALAR_VERIFY(a);
SECP256K1_SCALAR_VERIFY(b);
*r = (*a * *b) % EXHAUSTIVE_TEST_ORDER;
SECP256K1_SCALAR_VERIFY(r);
}
static void secp256k1_scalar_split_128(secp256k1_scalar *r1, secp256k1_scalar *r2, const secp256k1_scalar *a) {
SECP256K1_SCALAR_VERIFY(a);
*r1 = *a;
*r2 = 0;
SECP256K1_SCALAR_VERIFY(r1);
SECP256K1_SCALAR_VERIFY(r2);
}
SECP256K1_INLINE static int secp256k1_scalar_eq(const secp256k1_scalar *a, const secp256k1_scalar *b) {
SECP256K1_SCALAR_VERIFY(a);
SECP256K1_SCALAR_VERIFY(b);
return *a == *b;
}
static SECP256K1_INLINE void secp256k1_scalar_cmov(secp256k1_scalar *r, const secp256k1_scalar *a, int flag) {
uint32_t mask0, mask1;
volatile int vflag = flag;
SECP256K1_SCALAR_VERIFY(a);
SECP256K1_CHECKMEM_CHECK_VERIFY(r, sizeof(*r));
mask0 = vflag + ~((uint32_t)0);
mask1 = ~mask0;
*r = (*r & mask0) | (*a & mask1);
SECP256K1_SCALAR_VERIFY(r);
}
static void secp256k1_scalar_inverse(secp256k1_scalar *r, const secp256k1_scalar *x) {
int i;
uint32_t res = 0;
SECP256K1_SCALAR_VERIFY(x);
for (i = 0; i < EXHAUSTIVE_TEST_ORDER; i++) {
if ((i * *x) % EXHAUSTIVE_TEST_ORDER == 1) {
res = i;
break;
}
}
/* If this VERIFY_CHECK triggers we were given a noninvertible scalar (and thus
* have a composite group order; fix it in exhaustive_tests.c). */
VERIFY_CHECK(res != 0);
*r = res;
SECP256K1_SCALAR_VERIFY(r);
}
static void secp256k1_scalar_inverse_var(secp256k1_scalar *r, const secp256k1_scalar *x) {
SECP256K1_SCALAR_VERIFY(x);
secp256k1_scalar_inverse(r, x);
SECP256K1_SCALAR_VERIFY(r);
}
static void secp256k1_scalar_half(secp256k1_scalar *r, const secp256k1_scalar *a) {
SECP256K1_SCALAR_VERIFY(a);
*r = (*a + ((-(uint32_t)(*a & 1)) & EXHAUSTIVE_TEST_ORDER)) >> 1;
SECP256K1_SCALAR_VERIFY(r);
}
#endif /* SECP256K1_SCALAR_REPR_IMPL_H */

View File

@@ -0,0 +1,44 @@
/***********************************************************************
* Copyright (c) 2017 Andrew Poelstra *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_SCRATCH_H
#define SECP256K1_SCRATCH_H
/* The typedef is used internally; the struct name is used in the public API
* (where it is exposed as a different typedef) */
typedef struct secp256k1_scratch_space_struct {
/** guard against interpreting this object as other types */
unsigned char magic[8];
/** actual allocated data */
void *data;
/** amount that has been allocated (i.e. `data + offset` is the next
* available pointer) */
size_t alloc_size;
/** maximum size available to allocate */
size_t max_size;
} secp256k1_scratch;
typedef struct secp256k1_scratch_space_struct secp256k1_scratch_space;
static secp256k1_scratch* secp256k1_scratch_create(const secp256k1_callback* error_callback, size_t max_size);
static void secp256k1_scratch_destroy(const secp256k1_callback* error_callback, secp256k1_scratch* scratch);
/** Returns an opaque object used to "checkpoint" a scratch space. Used
* with `secp256k1_scratch_apply_checkpoint` to undo allocations. */
static size_t secp256k1_scratch_checkpoint(const secp256k1_callback* error_callback, const secp256k1_scratch* scratch);
/** Applies a check point received from `secp256k1_scratch_checkpoint`,
* undoing all allocations since that point. */
static void secp256k1_scratch_apply_checkpoint(const secp256k1_callback* error_callback, secp256k1_scratch* scratch, size_t checkpoint);
/** Returns the maximum allocation the scratch space will allow */
static size_t secp256k1_scratch_max_allocation(const secp256k1_callback* error_callback, const secp256k1_scratch* scratch, size_t n_objects);
/** Returns a pointer into the most recently allocated frame, or NULL if there is insufficient available space */
static void *secp256k1_scratch_alloc(const secp256k1_callback* error_callback, secp256k1_scratch* scratch, size_t n);
#endif

View File

@@ -0,0 +1,99 @@
/***********************************************************************
* Copyright (c) 2017 Andrew Poelstra *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_SCRATCH_IMPL_H
#define SECP256K1_SCRATCH_IMPL_H
#include "util.h"
#include "scratch.h"
static secp256k1_scratch* secp256k1_scratch_create(const secp256k1_callback* error_callback, size_t size) {
const size_t base_alloc = ROUND_TO_ALIGN(sizeof(secp256k1_scratch));
void *alloc = checked_malloc(error_callback, base_alloc + size);
secp256k1_scratch* ret = (secp256k1_scratch *)alloc;
if (ret != NULL) {
memset(ret, 0, sizeof(*ret));
memcpy(ret->magic, "scratch", 8);
ret->data = (void *) ((char *) alloc + base_alloc);
ret->max_size = size;
}
return ret;
}
static void secp256k1_scratch_destroy(const secp256k1_callback* error_callback, secp256k1_scratch* scratch) {
if (scratch != NULL) {
if (secp256k1_memcmp_var(scratch->magic, "scratch", 8) != 0) {
secp256k1_callback_call(error_callback, "invalid scratch space");
return;
}
VERIFY_CHECK(scratch->alloc_size == 0); /* all checkpoints should be applied */
memset(scratch->magic, 0, sizeof(scratch->magic));
free(scratch);
}
}
static size_t secp256k1_scratch_checkpoint(const secp256k1_callback* error_callback, const secp256k1_scratch* scratch) {
if (secp256k1_memcmp_var(scratch->magic, "scratch", 8) != 0) {
secp256k1_callback_call(error_callback, "invalid scratch space");
return 0;
}
return scratch->alloc_size;
}
static void secp256k1_scratch_apply_checkpoint(const secp256k1_callback* error_callback, secp256k1_scratch* scratch, size_t checkpoint) {
if (secp256k1_memcmp_var(scratch->magic, "scratch", 8) != 0) {
secp256k1_callback_call(error_callback, "invalid scratch space");
return;
}
if (checkpoint > scratch->alloc_size) {
secp256k1_callback_call(error_callback, "invalid checkpoint");
return;
}
scratch->alloc_size = checkpoint;
}
static size_t secp256k1_scratch_max_allocation(const secp256k1_callback* error_callback, const secp256k1_scratch* scratch, size_t objects) {
if (secp256k1_memcmp_var(scratch->magic, "scratch", 8) != 0) {
secp256k1_callback_call(error_callback, "invalid scratch space");
return 0;
}
/* Ensure that multiplication will not wrap around */
if (ALIGNMENT > 1 && objects > SIZE_MAX/(ALIGNMENT - 1)) {
return 0;
}
if (scratch->max_size - scratch->alloc_size <= objects * (ALIGNMENT - 1)) {
return 0;
}
return scratch->max_size - scratch->alloc_size - objects * (ALIGNMENT - 1);
}
static void *secp256k1_scratch_alloc(const secp256k1_callback* error_callback, secp256k1_scratch* scratch, size_t size) {
void *ret;
size_t rounded_size;
rounded_size = ROUND_TO_ALIGN(size);
/* Check that rounding did not wrap around */
if (rounded_size < size) {
return NULL;
}
size = rounded_size;
if (secp256k1_memcmp_var(scratch->magic, "scratch", 8) != 0) {
secp256k1_callback_call(error_callback, "invalid scratch space");
return NULL;
}
if (size > scratch->max_size - scratch->alloc_size) {
return NULL;
}
ret = (void *) ((char *) scratch->data + scratch->alloc_size);
memset(ret, 0, size);
scratch->alloc_size += size;
return ret;
}
#endif

View File

@@ -0,0 +1,831 @@
/***********************************************************************
* Copyright (c) 2013-2015 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
/* This is a C project. It should not be compiled with a C++ compiler,
* and we error out if we detect one.
*
* We still want to be able to test the project with a C++ compiler
* because it is still good to know if this will lead to real trouble, so
* there is a possibility to override the check. But be warned that
* compiling with a C++ compiler is not supported. */
#if defined(__cplusplus) && !defined(SECP256K1_CPLUSPLUS_TEST_OVERRIDE)
#error Trying to compile a C project with a C++ compiler.
#endif
#define SECP256K1_BUILD
#include "../include/secp256k1.h"
#include "../include/secp256k1_preallocated.h"
#include "assumptions.h"
#include "checkmem.h"
#include "util.h"
#include "field_impl.h"
#include "scalar_impl.h"
#include "group_impl.h"
#include "ecmult_impl.h"
#include "ecmult_const_impl.h"
#include "ecmult_gen_impl.h"
#include "ecdsa_impl.h"
#include "eckey_impl.h"
#include "hash_impl.h"
#include "int128_impl.h"
#include "scratch_impl.h"
#include "selftest.h"
#include "hsort_impl.h"
#ifdef SECP256K1_NO_BUILD
# error "secp256k1.h processed without SECP256K1_BUILD defined while building secp256k1.c"
#endif
#define ARG_CHECK(cond) do { \
if (EXPECT(!(cond), 0)) { \
secp256k1_callback_call(&ctx->illegal_callback, #cond); \
return 0; \
} \
} while(0)
#define ARG_CHECK_VOID(cond) do { \
if (EXPECT(!(cond), 0)) { \
secp256k1_callback_call(&ctx->illegal_callback, #cond); \
return; \
} \
} while(0)
/* Note that whenever you change the context struct, you must also change the
* context_eq function. */
struct secp256k1_context_struct {
secp256k1_ecmult_gen_context ecmult_gen_ctx;
secp256k1_callback illegal_callback;
secp256k1_callback error_callback;
int declassify;
};
static const secp256k1_context secp256k1_context_static_ = {
{ 0 },
{ secp256k1_default_illegal_callback_fn, 0 },
{ secp256k1_default_error_callback_fn, 0 },
0
};
const secp256k1_context *secp256k1_context_static = &secp256k1_context_static_;
const secp256k1_context *secp256k1_context_no_precomp = &secp256k1_context_static_;
/* Helper function that determines if a context is proper, i.e., is not the static context or a copy thereof.
*
* This is intended for "context" functions such as secp256k1_context_clone. Functions that need specific
* features of a context should still check for these features directly. For example, a function that needs
* ecmult_gen should directly check for the existence of the ecmult_gen context. */
static int secp256k1_context_is_proper(const secp256k1_context* ctx) {
return secp256k1_ecmult_gen_context_is_built(&ctx->ecmult_gen_ctx);
}
void secp256k1_selftest(void) {
if (!secp256k1_selftest_passes()) {
secp256k1_callback_call(&default_error_callback, "self test failed");
}
}
size_t secp256k1_context_preallocated_size(unsigned int flags) {
size_t ret = sizeof(secp256k1_context);
/* A return value of 0 is reserved as an indicator for errors when we call this function internally. */
VERIFY_CHECK(ret != 0);
if (EXPECT((flags & SECP256K1_FLAGS_TYPE_MASK) != SECP256K1_FLAGS_TYPE_CONTEXT, 0)) {
secp256k1_callback_call(&default_illegal_callback,
"Invalid flags");
return 0;
}
if (EXPECT(!SECP256K1_CHECKMEM_RUNNING() && (flags & SECP256K1_FLAGS_BIT_CONTEXT_DECLASSIFY), 0)) {
secp256k1_callback_call(&default_illegal_callback,
"Declassify flag requires running with memory checking");
return 0;
}
return ret;
}
size_t secp256k1_context_preallocated_clone_size(const secp256k1_context* ctx) {
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(secp256k1_context_is_proper(ctx));
return sizeof(secp256k1_context);
}
secp256k1_context* secp256k1_context_preallocated_create(void* prealloc, unsigned int flags) {
size_t prealloc_size;
secp256k1_context* ret;
secp256k1_selftest();
prealloc_size = secp256k1_context_preallocated_size(flags);
if (prealloc_size == 0) {
return NULL;
}
VERIFY_CHECK(prealloc != NULL);
ret = (secp256k1_context*)prealloc;
ret->illegal_callback = default_illegal_callback;
ret->error_callback = default_error_callback;
/* Flags have been checked by secp256k1_context_preallocated_size. */
VERIFY_CHECK((flags & SECP256K1_FLAGS_TYPE_MASK) == SECP256K1_FLAGS_TYPE_CONTEXT);
secp256k1_ecmult_gen_context_build(&ret->ecmult_gen_ctx);
ret->declassify = !!(flags & SECP256K1_FLAGS_BIT_CONTEXT_DECLASSIFY);
return ret;
}
secp256k1_context* secp256k1_context_create(unsigned int flags) {
size_t const prealloc_size = secp256k1_context_preallocated_size(flags);
secp256k1_context* ctx = (secp256k1_context*)checked_malloc(&default_error_callback, prealloc_size);
if (EXPECT(secp256k1_context_preallocated_create(ctx, flags) == NULL, 0)) {
free(ctx);
return NULL;
}
return ctx;
}
secp256k1_context* secp256k1_context_preallocated_clone(const secp256k1_context* ctx, void* prealloc) {
secp256k1_context* ret;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(prealloc != NULL);
ARG_CHECK(secp256k1_context_is_proper(ctx));
ret = (secp256k1_context*)prealloc;
*ret = *ctx;
return ret;
}
secp256k1_context* secp256k1_context_clone(const secp256k1_context* ctx) {
secp256k1_context* ret;
size_t prealloc_size;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(secp256k1_context_is_proper(ctx));
prealloc_size = secp256k1_context_preallocated_clone_size(ctx);
ret = (secp256k1_context*)checked_malloc(&ctx->error_callback, prealloc_size);
ret = secp256k1_context_preallocated_clone(ctx, ret);
return ret;
}
void secp256k1_context_preallocated_destroy(secp256k1_context* ctx) {
ARG_CHECK_VOID(ctx == NULL || secp256k1_context_is_proper(ctx));
/* Defined as noop */
if (ctx == NULL) {
return;
}
secp256k1_ecmult_gen_context_clear(&ctx->ecmult_gen_ctx);
}
void secp256k1_context_destroy(secp256k1_context* ctx) {
ARG_CHECK_VOID(ctx == NULL || secp256k1_context_is_proper(ctx));
/* Defined as noop */
if (ctx == NULL) {
return;
}
secp256k1_context_preallocated_destroy(ctx);
free(ctx);
}
void secp256k1_context_set_illegal_callback(secp256k1_context* ctx, void (*fun)(const char* message, void* data), const void* data) {
/* We compare pointers instead of checking secp256k1_context_is_proper() here
because setting callbacks is allowed on *copies* of the static context:
it's harmless and makes testing easier. */
ARG_CHECK_VOID(ctx != secp256k1_context_static);
if (fun == NULL) {
fun = secp256k1_default_illegal_callback_fn;
}
ctx->illegal_callback.fn = fun;
ctx->illegal_callback.data = data;
}
void secp256k1_context_set_error_callback(secp256k1_context* ctx, void (*fun)(const char* message, void* data), const void* data) {
/* We compare pointers instead of checking secp256k1_context_is_proper() here
because setting callbacks is allowed on *copies* of the static context:
it's harmless and makes testing easier. */
ARG_CHECK_VOID(ctx != secp256k1_context_static);
if (fun == NULL) {
fun = secp256k1_default_error_callback_fn;
}
ctx->error_callback.fn = fun;
ctx->error_callback.data = data;
}
static secp256k1_scratch_space* secp256k1_scratch_space_create(const secp256k1_context* ctx, size_t max_size) {
VERIFY_CHECK(ctx != NULL);
return secp256k1_scratch_create(&ctx->error_callback, max_size);
}
static void secp256k1_scratch_space_destroy(const secp256k1_context *ctx, secp256k1_scratch_space* scratch) {
VERIFY_CHECK(ctx != NULL);
secp256k1_scratch_destroy(&ctx->error_callback, scratch);
}
/* Mark memory as no-longer-secret for the purpose of analysing constant-time behaviour
* of the software.
*/
static SECP256K1_INLINE void secp256k1_declassify(const secp256k1_context* ctx, const void *p, size_t len) {
if (EXPECT(ctx->declassify, 0)) SECP256K1_CHECKMEM_DEFINE(p, len);
}
static int secp256k1_pubkey_load(const secp256k1_context* ctx, secp256k1_ge* ge, const secp256k1_pubkey* pubkey) {
secp256k1_ge_from_bytes(ge, pubkey->data);
ARG_CHECK(!secp256k1_fe_is_zero(&ge->x));
return 1;
}
static void secp256k1_pubkey_save(secp256k1_pubkey* pubkey, secp256k1_ge* ge) {
secp256k1_ge_to_bytes(pubkey->data, ge);
}
int secp256k1_ec_pubkey_parse(const secp256k1_context* ctx, secp256k1_pubkey* pubkey, const unsigned char *input, size_t inputlen) {
secp256k1_ge Q;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(pubkey != NULL);
memset(pubkey, 0, sizeof(*pubkey));
ARG_CHECK(input != NULL);
if (!secp256k1_eckey_pubkey_parse(&Q, input, inputlen)) {
return 0;
}
if (!secp256k1_ge_is_in_correct_subgroup(&Q)) {
return 0;
}
secp256k1_pubkey_save(pubkey, &Q);
secp256k1_ge_clear(&Q);
return 1;
}
int secp256k1_ec_pubkey_serialize(const secp256k1_context* ctx, unsigned char *output, size_t *outputlen, const secp256k1_pubkey* pubkey, unsigned int flags) {
secp256k1_ge Q;
size_t len;
int ret = 0;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(outputlen != NULL);
ARG_CHECK(*outputlen >= ((flags & SECP256K1_FLAGS_BIT_COMPRESSION) ? 33u : 65u));
len = *outputlen;
*outputlen = 0;
ARG_CHECK(output != NULL);
memset(output, 0, len);
ARG_CHECK(pubkey != NULL);
ARG_CHECK((flags & SECP256K1_FLAGS_TYPE_MASK) == SECP256K1_FLAGS_TYPE_COMPRESSION);
if (secp256k1_pubkey_load(ctx, &Q, pubkey)) {
ret = secp256k1_eckey_pubkey_serialize(&Q, output, &len, flags & SECP256K1_FLAGS_BIT_COMPRESSION);
if (ret) {
*outputlen = len;
}
}
return ret;
}
int secp256k1_ec_pubkey_cmp(const secp256k1_context* ctx, const secp256k1_pubkey* pubkey0, const secp256k1_pubkey* pubkey1) {
unsigned char out[2][33];
const secp256k1_pubkey* pk[2];
int i;
VERIFY_CHECK(ctx != NULL);
pk[0] = pubkey0; pk[1] = pubkey1;
for (i = 0; i < 2; i++) {
size_t out_size = sizeof(out[i]);
/* If the public key is NULL or invalid, ec_pubkey_serialize will call
* the illegal_callback and return 0. In that case we will serialize the
* key as all zeros which is less than any valid public key. This
* results in consistent comparisons even if NULL or invalid pubkeys are
* involved and prevents edge cases such as sorting algorithms that use
* this function and do not terminate as a result. */
if (!secp256k1_ec_pubkey_serialize(ctx, out[i], &out_size, pk[i], SECP256K1_EC_COMPRESSED)) {
/* Note that ec_pubkey_serialize should already set the output to
* zero in that case, but it's not guaranteed by the API, we can't
* test it and writing a VERIFY_CHECK is more complex than
* explicitly memsetting (again). */
memset(out[i], 0, sizeof(out[i]));
}
}
return secp256k1_memcmp_var(out[0], out[1], sizeof(out[0]));
}
static int secp256k1_ec_pubkey_sort_cmp(const void* pk1, const void* pk2, void *ctx) {
return secp256k1_ec_pubkey_cmp((secp256k1_context *)ctx,
*(secp256k1_pubkey **)pk1,
*(secp256k1_pubkey **)pk2);
}
int secp256k1_ec_pubkey_sort(const secp256k1_context* ctx, const secp256k1_pubkey **pubkeys, size_t n_pubkeys) {
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(pubkeys != NULL);
/* Suppress wrong warning (fixed in MSVC 19.33) */
#if defined(_MSC_VER) && (_MSC_VER < 1933)
#pragma warning(push)
#pragma warning(disable: 4090)
#endif
/* Casting away const is fine because neither secp256k1_hsort nor
* secp256k1_ec_pubkey_sort_cmp modify the data pointed to by the cmp_data
* argument. */
secp256k1_hsort(pubkeys, n_pubkeys, sizeof(*pubkeys), secp256k1_ec_pubkey_sort_cmp, (void *)ctx);
#if defined(_MSC_VER) && (_MSC_VER < 1933)
#pragma warning(pop)
#endif
return 1;
}
static void secp256k1_ecdsa_signature_load(const secp256k1_context* ctx, secp256k1_scalar* r, secp256k1_scalar* s, const secp256k1_ecdsa_signature* sig) {
(void)ctx;
if (sizeof(secp256k1_scalar) == 32) {
/* When the secp256k1_scalar type is exactly 32 byte, use its
* representation inside secp256k1_ecdsa_signature, as conversion is very fast.
* Note that secp256k1_ecdsa_signature_save must use the same representation. */
memcpy(r, &sig->data[0], 32);
memcpy(s, &sig->data[32], 32);
} else {
secp256k1_scalar_set_b32(r, &sig->data[0], NULL);
secp256k1_scalar_set_b32(s, &sig->data[32], NULL);
}
}
static void secp256k1_ecdsa_signature_save(secp256k1_ecdsa_signature* sig, const secp256k1_scalar* r, const secp256k1_scalar* s) {
if (sizeof(secp256k1_scalar) == 32) {
memcpy(&sig->data[0], r, 32);
memcpy(&sig->data[32], s, 32);
} else {
secp256k1_scalar_get_b32(&sig->data[0], r);
secp256k1_scalar_get_b32(&sig->data[32], s);
}
}
int secp256k1_ecdsa_signature_parse_der(const secp256k1_context* ctx, secp256k1_ecdsa_signature* sig, const unsigned char *input, size_t inputlen) {
secp256k1_scalar r, s;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(sig != NULL);
ARG_CHECK(input != NULL);
if (secp256k1_ecdsa_sig_parse(&r, &s, input, inputlen)) {
secp256k1_ecdsa_signature_save(sig, &r, &s);
return 1;
} else {
memset(sig, 0, sizeof(*sig));
return 0;
}
}
int secp256k1_ecdsa_signature_parse_compact(const secp256k1_context* ctx, secp256k1_ecdsa_signature* sig, const unsigned char *input64) {
secp256k1_scalar r, s;
int ret = 1;
int overflow = 0;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(sig != NULL);
ARG_CHECK(input64 != NULL);
secp256k1_scalar_set_b32(&r, &input64[0], &overflow);
ret &= !overflow;
secp256k1_scalar_set_b32(&s, &input64[32], &overflow);
ret &= !overflow;
if (ret) {
secp256k1_ecdsa_signature_save(sig, &r, &s);
} else {
memset(sig, 0, sizeof(*sig));
}
return ret;
}
int secp256k1_ecdsa_signature_serialize_der(const secp256k1_context* ctx, unsigned char *output, size_t *outputlen, const secp256k1_ecdsa_signature* sig) {
secp256k1_scalar r, s;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(output != NULL);
ARG_CHECK(outputlen != NULL);
ARG_CHECK(sig != NULL);
secp256k1_ecdsa_signature_load(ctx, &r, &s, sig);
return secp256k1_ecdsa_sig_serialize(output, outputlen, &r, &s);
}
int secp256k1_ecdsa_signature_serialize_compact(const secp256k1_context* ctx, unsigned char *output64, const secp256k1_ecdsa_signature* sig) {
secp256k1_scalar r, s;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(output64 != NULL);
ARG_CHECK(sig != NULL);
secp256k1_ecdsa_signature_load(ctx, &r, &s, sig);
secp256k1_scalar_get_b32(&output64[0], &r);
secp256k1_scalar_get_b32(&output64[32], &s);
return 1;
}
int secp256k1_ecdsa_signature_normalize(const secp256k1_context* ctx, secp256k1_ecdsa_signature *sigout, const secp256k1_ecdsa_signature *sigin) {
secp256k1_scalar r, s;
int ret = 0;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(sigin != NULL);
secp256k1_ecdsa_signature_load(ctx, &r, &s, sigin);
ret = secp256k1_scalar_is_high(&s);
if (sigout != NULL) {
if (ret) {
secp256k1_scalar_negate(&s, &s);
}
secp256k1_ecdsa_signature_save(sigout, &r, &s);
}
return ret;
}
int secp256k1_ecdsa_verify(const secp256k1_context* ctx, const secp256k1_ecdsa_signature *sig, const unsigned char *msghash32, const secp256k1_pubkey *pubkey) {
secp256k1_ge q;
secp256k1_scalar r, s;
secp256k1_scalar m;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(msghash32 != NULL);
ARG_CHECK(sig != NULL);
ARG_CHECK(pubkey != NULL);
secp256k1_scalar_set_b32(&m, msghash32, NULL);
secp256k1_ecdsa_signature_load(ctx, &r, &s, sig);
return (!secp256k1_scalar_is_high(&s) &&
secp256k1_pubkey_load(ctx, &q, pubkey) &&
secp256k1_ecdsa_sig_verify(&r, &s, &q, &m));
}
static SECP256K1_INLINE void buffer_append(unsigned char *buf, unsigned int *offset, const void *data, unsigned int len) {
memcpy(buf + *offset, data, len);
*offset += len;
}
static int nonce_function_rfc6979(unsigned char *nonce32, const unsigned char *msg32, const unsigned char *key32, const unsigned char *algo16, void *data, unsigned int counter) {
unsigned char keydata[112];
unsigned int offset = 0;
secp256k1_rfc6979_hmac_sha256 rng;
unsigned int i;
secp256k1_scalar msg;
unsigned char msgmod32[32];
secp256k1_scalar_set_b32(&msg, msg32, NULL);
secp256k1_scalar_get_b32(msgmod32, &msg);
/* We feed a byte array to the PRNG as input, consisting of:
* - the private key (32 bytes) and reduced message (32 bytes), see RFC 6979 3.2d.
* - optionally 32 extra bytes of data, see RFC 6979 3.6 Additional Data.
* - optionally 16 extra bytes with the algorithm name.
* Because the arguments have distinct fixed lengths it is not possible for
* different argument mixtures to emulate each other and result in the same
* nonces.
*/
buffer_append(keydata, &offset, key32, 32);
buffer_append(keydata, &offset, msgmod32, 32);
if (data != NULL) {
buffer_append(keydata, &offset, data, 32);
}
if (algo16 != NULL) {
buffer_append(keydata, &offset, algo16, 16);
}
secp256k1_rfc6979_hmac_sha256_initialize(&rng, keydata, offset);
for (i = 0; i <= counter; i++) {
secp256k1_rfc6979_hmac_sha256_generate(&rng, nonce32, 32);
}
secp256k1_rfc6979_hmac_sha256_finalize(&rng);
secp256k1_memclear(keydata, sizeof(keydata));
secp256k1_rfc6979_hmac_sha256_clear(&rng);
return 1;
}
const secp256k1_nonce_function secp256k1_nonce_function_rfc6979 = nonce_function_rfc6979;
const secp256k1_nonce_function secp256k1_nonce_function_default = nonce_function_rfc6979;
static int secp256k1_ecdsa_sign_inner(const secp256k1_context* ctx, secp256k1_scalar* r, secp256k1_scalar* s, int* recid, const unsigned char *msg32, const unsigned char *seckey, secp256k1_nonce_function noncefp, const void* noncedata) {
secp256k1_scalar sec, non, msg;
int ret = 0;
int is_sec_valid;
unsigned char nonce32[32];
unsigned int count = 0;
/* Default initialization here is important so we won't pass uninit values to the cmov in the end */
*r = secp256k1_scalar_zero;
*s = secp256k1_scalar_zero;
if (recid) {
*recid = 0;
}
if (noncefp == NULL) {
noncefp = secp256k1_nonce_function_default;
}
/* Fail if the secret key is invalid. */
is_sec_valid = secp256k1_scalar_set_b32_seckey(&sec, seckey);
secp256k1_scalar_cmov(&sec, &secp256k1_scalar_one, !is_sec_valid);
secp256k1_scalar_set_b32(&msg, msg32, NULL);
while (1) {
int is_nonce_valid;
ret = !!noncefp(nonce32, msg32, seckey, NULL, (void*)noncedata, count);
if (!ret) {
break;
}
is_nonce_valid = secp256k1_scalar_set_b32_seckey(&non, nonce32);
/* The nonce is still secret here, but it being invalid is less likely than 1:2^255. */
secp256k1_declassify(ctx, &is_nonce_valid, sizeof(is_nonce_valid));
if (is_nonce_valid) {
ret = secp256k1_ecdsa_sig_sign(&ctx->ecmult_gen_ctx, r, s, &sec, &msg, &non, recid);
/* The final signature is no longer a secret, nor is the fact that we were successful or not. */
secp256k1_declassify(ctx, &ret, sizeof(ret));
if (ret) {
break;
}
}
count++;
}
/* We don't want to declassify is_sec_valid and therefore the range of
* seckey. As a result is_sec_valid is included in ret only after ret was
* used as a branching variable. */
ret &= is_sec_valid;
secp256k1_memclear(nonce32, sizeof(nonce32));
secp256k1_scalar_clear(&msg);
secp256k1_scalar_clear(&non);
secp256k1_scalar_clear(&sec);
secp256k1_scalar_cmov(r, &secp256k1_scalar_zero, !ret);
secp256k1_scalar_cmov(s, &secp256k1_scalar_zero, !ret);
if (recid) {
const int zero = 0;
secp256k1_int_cmov(recid, &zero, !ret);
}
return ret;
}
int secp256k1_ecdsa_sign(const secp256k1_context* ctx, secp256k1_ecdsa_signature *signature, const unsigned char *msghash32, const unsigned char *seckey, secp256k1_nonce_function noncefp, const void* noncedata) {
secp256k1_scalar r, s;
int ret;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(secp256k1_ecmult_gen_context_is_built(&ctx->ecmult_gen_ctx));
ARG_CHECK(msghash32 != NULL);
ARG_CHECK(signature != NULL);
ARG_CHECK(seckey != NULL);
ret = secp256k1_ecdsa_sign_inner(ctx, &r, &s, NULL, msghash32, seckey, noncefp, noncedata);
secp256k1_ecdsa_signature_save(signature, &r, &s);
return ret;
}
int secp256k1_ec_seckey_verify(const secp256k1_context* ctx, const unsigned char *seckey) {
secp256k1_scalar sec;
int ret;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(seckey != NULL);
ret = secp256k1_scalar_set_b32_seckey(&sec, seckey);
secp256k1_scalar_clear(&sec);
return ret;
}
static int secp256k1_ec_pubkey_create_helper(const secp256k1_ecmult_gen_context *ecmult_gen_ctx, secp256k1_scalar *seckey_scalar, secp256k1_ge *p, const unsigned char *seckey) {
secp256k1_gej pj;
int ret;
ret = secp256k1_scalar_set_b32_seckey(seckey_scalar, seckey);
secp256k1_scalar_cmov(seckey_scalar, &secp256k1_scalar_one, !ret);
secp256k1_ecmult_gen(ecmult_gen_ctx, &pj, seckey_scalar);
secp256k1_ge_set_gej(p, &pj);
secp256k1_gej_clear(&pj);
return ret;
}
int secp256k1_ec_pubkey_create(const secp256k1_context* ctx, secp256k1_pubkey *pubkey, const unsigned char *seckey) {
secp256k1_ge p;
secp256k1_scalar seckey_scalar;
int ret = 0;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(pubkey != NULL);
memset(pubkey, 0, sizeof(*pubkey));
ARG_CHECK(secp256k1_ecmult_gen_context_is_built(&ctx->ecmult_gen_ctx));
ARG_CHECK(seckey != NULL);
ret = secp256k1_ec_pubkey_create_helper(&ctx->ecmult_gen_ctx, &seckey_scalar, &p, seckey);
secp256k1_pubkey_save(pubkey, &p);
secp256k1_memczero(pubkey, sizeof(*pubkey), !ret);
secp256k1_scalar_clear(&seckey_scalar);
return ret;
}
int secp256k1_ec_seckey_negate(const secp256k1_context* ctx, unsigned char *seckey) {
secp256k1_scalar sec;
int ret = 0;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(seckey != NULL);
ret = secp256k1_scalar_set_b32_seckey(&sec, seckey);
secp256k1_scalar_cmov(&sec, &secp256k1_scalar_zero, !ret);
secp256k1_scalar_negate(&sec, &sec);
secp256k1_scalar_get_b32(seckey, &sec);
secp256k1_scalar_clear(&sec);
return ret;
}
int secp256k1_ec_privkey_negate(const secp256k1_context* ctx, unsigned char *seckey) {
return secp256k1_ec_seckey_negate(ctx, seckey);
}
int secp256k1_ec_pubkey_negate(const secp256k1_context* ctx, secp256k1_pubkey *pubkey) {
int ret = 0;
secp256k1_ge p;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(pubkey != NULL);
ret = secp256k1_pubkey_load(ctx, &p, pubkey);
memset(pubkey, 0, sizeof(*pubkey));
if (ret) {
secp256k1_ge_neg(&p, &p);
secp256k1_pubkey_save(pubkey, &p);
}
return ret;
}
static int secp256k1_ec_seckey_tweak_add_helper(secp256k1_scalar *sec, const unsigned char *tweak32) {
secp256k1_scalar term;
int overflow = 0;
int ret = 0;
secp256k1_scalar_set_b32(&term, tweak32, &overflow);
ret = (!overflow) & secp256k1_eckey_privkey_tweak_add(sec, &term);
secp256k1_scalar_clear(&term);
return ret;
}
int secp256k1_ec_seckey_tweak_add(const secp256k1_context* ctx, unsigned char *seckey, const unsigned char *tweak32) {
secp256k1_scalar sec;
int ret = 0;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(seckey != NULL);
ARG_CHECK(tweak32 != NULL);
ret = secp256k1_scalar_set_b32_seckey(&sec, seckey);
ret &= secp256k1_ec_seckey_tweak_add_helper(&sec, tweak32);
secp256k1_scalar_cmov(&sec, &secp256k1_scalar_zero, !ret);
secp256k1_scalar_get_b32(seckey, &sec);
secp256k1_scalar_clear(&sec);
return ret;
}
int secp256k1_ec_privkey_tweak_add(const secp256k1_context* ctx, unsigned char *seckey, const unsigned char *tweak32) {
return secp256k1_ec_seckey_tweak_add(ctx, seckey, tweak32);
}
static int secp256k1_ec_pubkey_tweak_add_helper(secp256k1_ge *p, const unsigned char *tweak32) {
secp256k1_scalar term;
int overflow = 0;
secp256k1_scalar_set_b32(&term, tweak32, &overflow);
return !overflow && secp256k1_eckey_pubkey_tweak_add(p, &term);
}
int secp256k1_ec_pubkey_tweak_add(const secp256k1_context* ctx, secp256k1_pubkey *pubkey, const unsigned char *tweak32) {
secp256k1_ge p;
int ret = 0;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(pubkey != NULL);
ARG_CHECK(tweak32 != NULL);
ret = secp256k1_pubkey_load(ctx, &p, pubkey);
memset(pubkey, 0, sizeof(*pubkey));
ret = ret && secp256k1_ec_pubkey_tweak_add_helper(&p, tweak32);
if (ret) {
secp256k1_pubkey_save(pubkey, &p);
}
return ret;
}
int secp256k1_ec_seckey_tweak_mul(const secp256k1_context* ctx, unsigned char *seckey, const unsigned char *tweak32) {
secp256k1_scalar factor;
secp256k1_scalar sec;
int ret = 0;
int overflow = 0;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(seckey != NULL);
ARG_CHECK(tweak32 != NULL);
secp256k1_scalar_set_b32(&factor, tweak32, &overflow);
ret = secp256k1_scalar_set_b32_seckey(&sec, seckey);
ret &= (!overflow) & secp256k1_eckey_privkey_tweak_mul(&sec, &factor);
secp256k1_scalar_cmov(&sec, &secp256k1_scalar_zero, !ret);
secp256k1_scalar_get_b32(seckey, &sec);
secp256k1_scalar_clear(&sec);
secp256k1_scalar_clear(&factor);
return ret;
}
int secp256k1_ec_privkey_tweak_mul(const secp256k1_context* ctx, unsigned char *seckey, const unsigned char *tweak32) {
return secp256k1_ec_seckey_tweak_mul(ctx, seckey, tweak32);
}
int secp256k1_ec_pubkey_tweak_mul(const secp256k1_context* ctx, secp256k1_pubkey *pubkey, const unsigned char *tweak32) {
secp256k1_ge p;
secp256k1_scalar factor;
int ret = 0;
int overflow = 0;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(pubkey != NULL);
ARG_CHECK(tweak32 != NULL);
secp256k1_scalar_set_b32(&factor, tweak32, &overflow);
ret = !overflow && secp256k1_pubkey_load(ctx, &p, pubkey);
memset(pubkey, 0, sizeof(*pubkey));
if (ret) {
if (secp256k1_eckey_pubkey_tweak_mul(&p, &factor)) {
secp256k1_pubkey_save(pubkey, &p);
} else {
ret = 0;
}
}
return ret;
}
int secp256k1_context_randomize(secp256k1_context* ctx, const unsigned char *seed32) {
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(secp256k1_context_is_proper(ctx));
if (secp256k1_ecmult_gen_context_is_built(&ctx->ecmult_gen_ctx)) {
secp256k1_ecmult_gen_blind(&ctx->ecmult_gen_ctx, seed32);
}
return 1;
}
int secp256k1_ec_pubkey_combine(const secp256k1_context* ctx, secp256k1_pubkey *pubnonce, const secp256k1_pubkey * const *pubnonces, size_t n) {
size_t i;
secp256k1_gej Qj;
secp256k1_ge Q;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(pubnonce != NULL);
memset(pubnonce, 0, sizeof(*pubnonce));
ARG_CHECK(n >= 1);
ARG_CHECK(pubnonces != NULL);
secp256k1_gej_set_infinity(&Qj);
for (i = 0; i < n; i++) {
ARG_CHECK(pubnonces[i] != NULL);
secp256k1_pubkey_load(ctx, &Q, pubnonces[i]);
secp256k1_gej_add_ge(&Qj, &Qj, &Q);
}
if (secp256k1_gej_is_infinity(&Qj)) {
return 0;
}
secp256k1_ge_set_gej(&Q, &Qj);
secp256k1_pubkey_save(pubnonce, &Q);
return 1;
}
int secp256k1_tagged_sha256(const secp256k1_context* ctx, unsigned char *hash32, const unsigned char *tag, size_t taglen, const unsigned char *msg, size_t msglen) {
secp256k1_sha256 sha;
VERIFY_CHECK(ctx != NULL);
ARG_CHECK(hash32 != NULL);
ARG_CHECK(tag != NULL);
ARG_CHECK(msg != NULL);
secp256k1_sha256_initialize_tagged(&sha, tag, taglen);
secp256k1_sha256_write(&sha, msg, msglen);
secp256k1_sha256_finalize(&sha, hash32);
secp256k1_sha256_clear(&sha);
return 1;
}
#ifdef ENABLE_MODULE_ECDH
# include "modules/ecdh/main_impl.h"
#endif
#ifdef ENABLE_MODULE_RECOVERY
# include "modules/recovery/main_impl.h"
#endif
#ifdef ENABLE_MODULE_EXTRAKEYS
# include "modules/extrakeys/main_impl.h"
#endif
#ifdef ENABLE_MODULE_SCHNORRSIG
# include "modules/schnorrsig/main_impl.h"
#endif
#ifdef ENABLE_MODULE_MUSIG
# include "modules/musig/main_impl.h"
#endif
#ifdef ENABLE_MODULE_ELLSWIFT
# include "modules/ellswift/main_impl.h"
#endif

View File

@@ -0,0 +1,32 @@
/***********************************************************************
* Copyright (c) 2020 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_SELFTEST_H
#define SECP256K1_SELFTEST_H
#include "hash.h"
#include <string.h>
static int secp256k1_selftest_sha256(void) {
static const char *input63 = "For this sample, this 63-byte string will be used as input data";
static const unsigned char output32[32] = {
0xf0, 0x8a, 0x78, 0xcb, 0xba, 0xee, 0x08, 0x2b, 0x05, 0x2a, 0xe0, 0x70, 0x8f, 0x32, 0xfa, 0x1e,
0x50, 0xc5, 0xc4, 0x21, 0xaa, 0x77, 0x2b, 0xa5, 0xdb, 0xb4, 0x06, 0xa2, 0xea, 0x6b, 0xe3, 0x42,
};
unsigned char out[32];
secp256k1_sha256 hasher;
secp256k1_sha256_initialize(&hasher);
secp256k1_sha256_write(&hasher, (const unsigned char*)input63, 63);
secp256k1_sha256_finalize(&hasher, out);
return secp256k1_memcmp_var(out, output32, 32) == 0;
}
static int secp256k1_selftest_passes(void) {
return secp256k1_selftest_sha256();
}
#endif /* SECP256K1_SELFTEST_H */

451
libsecp256k1/src/util.h Normal file
View File

@@ -0,0 +1,451 @@
/***********************************************************************
* Copyright (c) 2013, 2014 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php.*
***********************************************************************/
#ifndef SECP256K1_UTIL_H
#define SECP256K1_UTIL_H
#include "../include/secp256k1.h"
#include "checkmem.h"
#include <string.h>
#include <stdlib.h>
#include <stdint.h>
#include <stdio.h>
#include <limits.h>
#if defined(_MSC_VER)
/* For SecureZeroMemory */
#include <Windows.h>
#endif
#define STR_(x) #x
#define STR(x) STR_(x)
#define DEBUG_CONFIG_MSG(x) "DEBUG_CONFIG: " x
#define DEBUG_CONFIG_DEF(x) DEBUG_CONFIG_MSG(#x "=" STR(x))
/* Debug helper for printing arrays of unsigned char. */
#define PRINT_BUF(buf, len) do { \
printf("%s[%lu] = ", #buf, (unsigned long)len); \
print_buf_plain(buf, len); \
} while(0)
static void print_buf_plain(const unsigned char *buf, size_t len) {
size_t i;
printf("{");
for (i = 0; i < len; i++) {
if (i % 8 == 0) {
printf("\n ");
} else {
printf(" ");
}
printf("0x%02X,", buf[i]);
}
printf("\n}\n");
}
# if (!defined(__STDC_VERSION__) || (__STDC_VERSION__ < 199901L) )
# if SECP256K1_GNUC_PREREQ(2,7)
# define SECP256K1_INLINE __inline__
# elif (defined(_MSC_VER))
# define SECP256K1_INLINE __inline
# else
# define SECP256K1_INLINE
# endif
# else
# define SECP256K1_INLINE inline
# endif
/** Assert statically that expr is true.
*
* This is a statement-like macro and can only be used inside functions.
*/
#define STATIC_ASSERT(expr) do { \
switch(0) { \
case 0: \
/* If expr evaluates to 0, we have two case labels "0", which is illegal. */ \
case /* ERROR: static assertion failed */ (expr): \
; \
} \
} while(0)
/** Assert statically that expr is an integer constant expression, and run stmt.
*
* Useful for example to enforce that magnitude arguments are constant.
*/
#define ASSERT_INT_CONST_AND_DO(expr, stmt) do { \
switch(42) { \
/* C allows only integer constant expressions as case labels. */ \
case /* ERROR: integer argument is not constant */ (expr): \
break; \
default: ; \
} \
stmt; \
} while(0)
typedef struct {
void (*fn)(const char *text, void* data);
const void* data;
} secp256k1_callback;
static SECP256K1_INLINE void secp256k1_callback_call(const secp256k1_callback * const cb, const char * const text) {
cb->fn(text, (void*)cb->data);
}
#ifndef USE_EXTERNAL_DEFAULT_CALLBACKS
static void secp256k1_default_illegal_callback_fn(const char* str, void* data) {
(void)data;
fprintf(stderr, "[libsecp256k1] illegal argument: %s\n", str);
abort();
}
static void secp256k1_default_error_callback_fn(const char* str, void* data) {
(void)data;
fprintf(stderr, "[libsecp256k1] internal consistency check failed: %s\n", str);
abort();
}
#else
void secp256k1_default_illegal_callback_fn(const char* str, void* data);
void secp256k1_default_error_callback_fn(const char* str, void* data);
#endif
static const secp256k1_callback default_illegal_callback = {
secp256k1_default_illegal_callback_fn,
NULL
};
static const secp256k1_callback default_error_callback = {
secp256k1_default_error_callback_fn,
NULL
};
#ifdef DETERMINISTIC
#define TEST_FAILURE(msg) do { \
fprintf(stderr, "%s\n", msg); \
abort(); \
} while(0);
#else
#define TEST_FAILURE(msg) do { \
fprintf(stderr, "%s:%d: %s\n", __FILE__, __LINE__, msg); \
abort(); \
} while(0)
#endif
#if SECP256K1_GNUC_PREREQ(3, 0)
#define EXPECT(x,c) __builtin_expect((x),(c))
#else
#define EXPECT(x,c) (x)
#endif
#ifdef DETERMINISTIC
#define CHECK(cond) do { \
if (EXPECT(!(cond), 0)) { \
TEST_FAILURE("test condition failed"); \
} \
} while(0)
#else
#define CHECK(cond) do { \
if (EXPECT(!(cond), 0)) { \
TEST_FAILURE("test condition failed: " #cond); \
} \
} while(0)
#endif
/* Like assert(), but when VERIFY is defined. */
#if defined(VERIFY)
#define VERIFY_CHECK CHECK
#else
#define VERIFY_CHECK(cond)
#endif
static SECP256K1_INLINE void *checked_malloc(const secp256k1_callback* cb, size_t size) {
void *ret = malloc(size);
if (ret == NULL) {
secp256k1_callback_call(cb, "Out of memory");
}
return ret;
}
#if defined(__BIGGEST_ALIGNMENT__)
#define ALIGNMENT __BIGGEST_ALIGNMENT__
#else
/* Using 16 bytes alignment because common architectures never have alignment
* requirements above 8 for any of the types we care about. In addition we
* leave some room because currently we don't care about a few bytes. */
#define ALIGNMENT 16
#endif
/* ceil(x/y) for integers x > 0 and y > 0. Here, / denotes rational division. */
#define CEIL_DIV(x, y) (1 + ((x) - 1) / (y))
#define ROUND_TO_ALIGN(size) (CEIL_DIV(size, ALIGNMENT) * ALIGNMENT)
/* Macro for restrict, when available and not in a VERIFY build. */
#if defined(SECP256K1_BUILD) && defined(VERIFY)
# define SECP256K1_RESTRICT
#else
# if (!defined(__STDC_VERSION__) || (__STDC_VERSION__ < 199901L) )
# if SECP256K1_GNUC_PREREQ(3,0)
# define SECP256K1_RESTRICT __restrict__
# elif (defined(_MSC_VER) && _MSC_VER >= 1400)
# define SECP256K1_RESTRICT __restrict
# else
# define SECP256K1_RESTRICT
# endif
# else
# define SECP256K1_RESTRICT restrict
# endif
#endif
#if defined(__GNUC__)
# define SECP256K1_GNUC_EXT __extension__
#else
# define SECP256K1_GNUC_EXT
#endif
/* Zero memory if flag == 1. Flag must be 0 or 1. Constant time. */
static SECP256K1_INLINE void secp256k1_memczero(void *s, size_t len, int flag) {
unsigned char *p = (unsigned char *)s;
/* Access flag with a volatile-qualified lvalue.
This prevents clang from figuring out (after inlining) that flag can
take only be 0 or 1, which leads to variable time code. */
volatile int vflag = flag;
unsigned char mask = -(unsigned char) vflag;
while (len) {
*p &= ~mask;
p++;
len--;
}
}
/* Cleanses memory to prevent leaking sensitive info. Won't be optimized out. */
static SECP256K1_INLINE void secp256k1_memclear(void *ptr, size_t len) {
#if defined(_MSC_VER)
/* SecureZeroMemory is guaranteed not to be optimized out by MSVC. */
SecureZeroMemory(ptr, len);
#elif defined(__GNUC__)
/* We use a memory barrier that scares the compiler away from optimizing out the memset.
*
* Quoting Adam Langley <agl@google.com> in commit ad1907fe73334d6c696c8539646c21b11178f20f
* in BoringSSL (ISC License):
* As best as we can tell, this is sufficient to break any optimisations that
* might try to eliminate "superfluous" memsets.
* This method is used in memzero_explicit() the Linux kernel, too. Its advantage is that it
* is pretty efficient, because the compiler can still implement the memset() efficently,
* just not remove it entirely. See "Dead Store Elimination (Still) Considered Harmful" by
* Yang et al. (USENIX Security 2017) for more background.
*/
memset(ptr, 0, len);
__asm__ __volatile__("" : : "r"(ptr) : "memory");
#else
void *(*volatile const volatile_memset)(void *, int, size_t) = memset;
volatile_memset(ptr, 0, len);
#endif
#ifdef VERIFY
SECP256K1_CHECKMEM_UNDEFINE(ptr, len);
#endif
}
/** Semantics like memcmp. Variable-time.
*
* We use this to avoid possible compiler bugs with memcmp, e.g.
* https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95189
*/
static SECP256K1_INLINE int secp256k1_memcmp_var(const void *s1, const void *s2, size_t n) {
const unsigned char *p1 = s1, *p2 = s2;
size_t i;
for (i = 0; i < n; i++) {
int diff = p1[i] - p2[i];
if (diff != 0) {
return diff;
}
}
return 0;
}
/* Return 1 if all elements of array s are 0 and otherwise return 0.
* Constant-time. */
static SECP256K1_INLINE int secp256k1_is_zero_array(const unsigned char *s, size_t len) {
unsigned char acc = 0;
int ret;
size_t i;
for (i = 0; i < len; i++) {
acc |= s[i];
}
ret = (acc == 0);
/* acc may contain secret values. Try to explicitly clear it. */
secp256k1_memclear(&acc, sizeof(acc));
return ret;
}
/** If flag is true, set *r equal to *a; otherwise leave it. Constant-time. Both *r and *a must be initialized and non-negative.*/
static SECP256K1_INLINE void secp256k1_int_cmov(int *r, const int *a, int flag) {
unsigned int mask0, mask1, r_masked, a_masked;
/* Access flag with a volatile-qualified lvalue.
This prevents clang from figuring out (after inlining) that flag can
take only be 0 or 1, which leads to variable time code. */
volatile int vflag = flag;
/* Casting a negative int to unsigned and back to int is implementation defined behavior */
VERIFY_CHECK(*r >= 0 && *a >= 0);
mask0 = (unsigned int)vflag + ~0u;
mask1 = ~mask0;
r_masked = ((unsigned int)*r & mask0);
a_masked = ((unsigned int)*a & mask1);
*r = (int)(r_masked | a_masked);
}
#if defined(USE_FORCE_WIDEMUL_INT128_STRUCT)
/* If USE_FORCE_WIDEMUL_INT128_STRUCT is set, use int128_struct. */
# define SECP256K1_WIDEMUL_INT128 1
# define SECP256K1_INT128_STRUCT 1
#elif defined(USE_FORCE_WIDEMUL_INT128)
/* If USE_FORCE_WIDEMUL_INT128 is set, use int128. */
# define SECP256K1_WIDEMUL_INT128 1
# define SECP256K1_INT128_NATIVE 1
#elif defined(USE_FORCE_WIDEMUL_INT64)
/* If USE_FORCE_WIDEMUL_INT64 is set, use int64. */
# define SECP256K1_WIDEMUL_INT64 1
#elif defined(UINT128_MAX) || defined(__SIZEOF_INT128__)
/* If a native 128-bit integer type exists, use int128. */
# define SECP256K1_WIDEMUL_INT128 1
# define SECP256K1_INT128_NATIVE 1
#elif defined(_MSC_VER) && (defined(_M_X64) || defined(_M_ARM64))
/* On 64-bit MSVC targets (x86_64 and arm64), use int128_struct
* (which has special logic to implement using intrinsics on those systems). */
# define SECP256K1_WIDEMUL_INT128 1
# define SECP256K1_INT128_STRUCT 1
#elif SIZE_MAX > 0xffffffff
/* Systems with 64-bit pointers (and thus registers) very likely benefit from
* using 64-bit based arithmetic (even if we need to fall back to 32x32->64 based
* multiplication logic). */
# define SECP256K1_WIDEMUL_INT128 1
# define SECP256K1_INT128_STRUCT 1
#else
/* Lastly, fall back to int64 based arithmetic. */
# define SECP256K1_WIDEMUL_INT64 1
#endif
#ifndef __has_builtin
#define __has_builtin(x) 0
#endif
/* Determine the number of trailing zero bits in a (non-zero) 32-bit x.
* This function is only intended to be used as fallback for
* secp256k1_ctz32_var, but permits it to be tested separately. */
static SECP256K1_INLINE int secp256k1_ctz32_var_debruijn(uint32_t x) {
static const uint8_t debruijn[32] = {
0x00, 0x01, 0x02, 0x18, 0x03, 0x13, 0x06, 0x19, 0x16, 0x04, 0x14, 0x0A,
0x10, 0x07, 0x0C, 0x1A, 0x1F, 0x17, 0x12, 0x05, 0x15, 0x09, 0x0F, 0x0B,
0x1E, 0x11, 0x08, 0x0E, 0x1D, 0x0D, 0x1C, 0x1B
};
return debruijn[(uint32_t)((x & -x) * 0x04D7651FU) >> 27];
}
/* Determine the number of trailing zero bits in a (non-zero) 64-bit x.
* This function is only intended to be used as fallback for
* secp256k1_ctz64_var, but permits it to be tested separately. */
static SECP256K1_INLINE int secp256k1_ctz64_var_debruijn(uint64_t x) {
static const uint8_t debruijn[64] = {
0, 1, 2, 53, 3, 7, 54, 27, 4, 38, 41, 8, 34, 55, 48, 28,
62, 5, 39, 46, 44, 42, 22, 9, 24, 35, 59, 56, 49, 18, 29, 11,
63, 52, 6, 26, 37, 40, 33, 47, 61, 45, 43, 21, 23, 58, 17, 10,
51, 25, 36, 32, 60, 20, 57, 16, 50, 31, 19, 15, 30, 14, 13, 12
};
return debruijn[(uint64_t)((x & -x) * 0x022FDD63CC95386DU) >> 58];
}
/* Determine the number of trailing zero bits in a (non-zero) 32-bit x. */
static SECP256K1_INLINE int secp256k1_ctz32_var(uint32_t x) {
VERIFY_CHECK(x != 0);
#if (__has_builtin(__builtin_ctz) || SECP256K1_GNUC_PREREQ(3,4))
/* If the unsigned type is sufficient to represent the largest uint32_t, consider __builtin_ctz. */
if (((unsigned)UINT32_MAX) == UINT32_MAX) {
return __builtin_ctz(x);
}
#endif
#if (__has_builtin(__builtin_ctzl) || SECP256K1_GNUC_PREREQ(3,4))
/* Otherwise consider __builtin_ctzl (the unsigned long type is always at least 32 bits). */
return __builtin_ctzl(x);
#else
/* If no suitable CTZ builtin is available, use a (variable time) software emulation. */
return secp256k1_ctz32_var_debruijn(x);
#endif
}
/* Determine the number of trailing zero bits in a (non-zero) 64-bit x. */
static SECP256K1_INLINE int secp256k1_ctz64_var(uint64_t x) {
VERIFY_CHECK(x != 0);
#if (__has_builtin(__builtin_ctzl) || SECP256K1_GNUC_PREREQ(3,4))
/* If the unsigned long type is sufficient to represent the largest uint64_t, consider __builtin_ctzl. */
if (((unsigned long)UINT64_MAX) == UINT64_MAX) {
return __builtin_ctzl(x);
}
#endif
#if (__has_builtin(__builtin_ctzll) || SECP256K1_GNUC_PREREQ(3,4))
/* Otherwise consider __builtin_ctzll (the unsigned long long type is always at least 64 bits). */
return __builtin_ctzll(x);
#else
/* If no suitable CTZ builtin is available, use a (variable time) software emulation. */
return secp256k1_ctz64_var_debruijn(x);
#endif
}
/* Read a uint32_t in big endian */
SECP256K1_INLINE static uint32_t secp256k1_read_be32(const unsigned char* p) {
return (uint32_t)p[0] << 24 |
(uint32_t)p[1] << 16 |
(uint32_t)p[2] << 8 |
(uint32_t)p[3];
}
/* Write a uint32_t in big endian */
SECP256K1_INLINE static void secp256k1_write_be32(unsigned char* p, uint32_t x) {
p[3] = x;
p[2] = x >> 8;
p[1] = x >> 16;
p[0] = x >> 24;
}
/* Read a uint64_t in big endian */
SECP256K1_INLINE static uint64_t secp256k1_read_be64(const unsigned char* p) {
return (uint64_t)p[0] << 56 |
(uint64_t)p[1] << 48 |
(uint64_t)p[2] << 40 |
(uint64_t)p[3] << 32 |
(uint64_t)p[4] << 24 |
(uint64_t)p[5] << 16 |
(uint64_t)p[6] << 8 |
(uint64_t)p[7];
}
/* Write a uint64_t in big endian */
SECP256K1_INLINE static void secp256k1_write_be64(unsigned char* p, uint64_t x) {
p[7] = x;
p[6] = x >> 8;
p[5] = x >> 16;
p[4] = x >> 24;
p[3] = x >> 32;
p[2] = x >> 40;
p[1] = x >> 48;
p[0] = x >> 56;
}
/* Rotate a uint32_t to the right. */
SECP256K1_INLINE static uint32_t secp256k1_rotr32(const uint32_t x, const unsigned int by) {
#if defined(_MSC_VER)
return _rotr(x, by); /* needs <stdlib.h> */
#else
/* Reduce rotation amount to avoid UB when shifting. */
const unsigned int mask = CHAR_BIT * sizeof(x) - 1;
/* Turned into a rot instruction by GCC and clang. */
return (x >> (by & mask)) | (x << ((-by) & mask));
#endif
}
#endif /* SECP256K1_UTIL_H */

View File

@@ -3,10 +3,23 @@
package nostr
/*
#cgo LDFLAGS: -lsecp256k1
#include <secp256k1.h>
#include <secp256k1_schnorrsig.h>
#include <secp256k1_extrakeys.h>
#cgo CFLAGS: -I${SRCDIR}/libsecp256k1/include -I${SRCDIR}/libsecp256k1/src
#cgo CFLAGS: -DECMULT_GEN_PREC_BITS=4
#cgo CFLAGS: -DECMULT_WINDOW_SIZE=15
#cgo CFLAGS: -DENABLE_MODULE_SCHNORRSIG=1
#cgo CFLAGS: -DENABLE_MODULE_EXTRAKEYS=1
#include "./libsecp256k1/src/secp256k1.c"
#include "./libsecp256k1/src/precomputed_ecmult.c"
#include "./libsecp256k1/src/precomputed_ecmult_gen.c"
#include "./libsecp256k1/src/ecmult_gen.h"
#include "./libsecp256k1/src/ecmult.h"
#include "./libsecp256k1/src/modules/extrakeys/main_impl.h"
#include "./libsecp256k1/src/modules/schnorrsig/main_impl.h"
#include "./libsecp256k1/include/secp256k1.h"
#include "./libsecp256k1/include/secp256k1_extrakeys.h"
#include "./libsecp256k1/include/secp256k1_schnorrsig.h"
*/
import "C"