diff --git a/depend/secp256k1/.travis.yml b/depend/secp256k1/.travis.yml index c4154e9..74f658f 100644 --- a/depend/secp256k1/.travis.yml +++ b/depend/secp256k1/.travis.yml @@ -1,5 +1,5 @@ language: c -sudo: false +os: linux addons: apt: packages: libgmp-dev @@ -66,4 +66,3 @@ script: - if [ -n "$HOST" ]; then export USE_HOST="--host=$HOST"; fi - if [ "x$HOST" = "xi686-linux-gnu" ]; then export CC="$CC -m32"; fi - ./configure --enable-experimental=$EXPERIMENTAL --enable-endomorphism=$ENDOMORPHISM --with-field=$FIELD --with-bignum=$BIGNUM --with-scalar=$SCALAR --enable-ecmult-static-precomputation=$STATICPRECOMPUTATION --enable-module-ecdh=$ECDH --enable-module-recovery=$RECOVERY --enable-jni=$JNI $EXTRAFLAGS $USE_HOST && make -j2 $BUILD -os: linux diff --git a/depend/secp256k1/Makefile.am b/depend/secp256k1/Makefile.am index 01fd0cd..21df09f 100644 --- a/depend/secp256k1/Makefile.am +++ b/depend/secp256k1/Makefile.am @@ -8,6 +8,7 @@ else JNI_LIB = endif include_HEADERS = include/secp256k1.h +include_HEADERS += include/secp256k1_preallocated.h noinst_HEADERS = noinst_HEADERS += src/scalar.h noinst_HEADERS += src/scalar_4x64.h @@ -114,7 +115,7 @@ exhaustive_tests_CPPFLAGS = -DSECP256K1_BUILD -I$(top_srcdir)/src $(SECP_INCLUDE if !ENABLE_COVERAGE exhaustive_tests_CPPFLAGS += -DVERIFY endif -exhaustive_tests_LDADD = $(SECP_LIBS) +exhaustive_tests_LDADD = $(SECP_LIBS) $(COMMON_LIB) exhaustive_tests_LDFLAGS = -static TESTS += exhaustive_tests endif @@ -151,7 +152,6 @@ endif if USE_ECMULT_STATIC_PRECOMPUTATION CPPFLAGS_FOR_BUILD +=-I$(top_srcdir) -CFLAGS_FOR_BUILD += -Wall -Wextra -Wno-unused-function gen_context_OBJECTS = gen_context.o gen_context_BIN = gen_context$(BUILD_EXEEXT) @@ -159,7 +159,7 @@ gen_%.o: src/gen_%.c $(CC_FOR_BUILD) $(CPPFLAGS_FOR_BUILD) $(CFLAGS_FOR_BUILD) -c $< -o $@ $(gen_context_BIN): $(gen_context_OBJECTS) - $(CC_FOR_BUILD) $^ -o $@ + $(CC_FOR_BUILD) $(CFLAGS_FOR_BUILD) $(LDFLAGS_FOR_BUILD) $^ -o $@ $(libsecp256k1_la_OBJECTS): src/ecmult_static_context.h $(tests_OBJECTS): src/ecmult_static_context.h diff --git a/depend/secp256k1/configure.ac b/depend/secp256k1/configure.ac index 68c45a5..b8340b7 100644 --- a/depend/secp256k1/configure.ac +++ b/depend/secp256k1/configure.ac @@ -85,42 +85,42 @@ AC_COMPILE_IFELSE([AC_LANG_SOURCE([[char foo;]])], ]) AC_ARG_ENABLE(benchmark, - AS_HELP_STRING([--enable-benchmark],[compile benchmark (default is yes)]), + AS_HELP_STRING([--enable-benchmark],[compile benchmark [default=yes]]), [use_benchmark=$enableval], [use_benchmark=yes]) AC_ARG_ENABLE(coverage, - AS_HELP_STRING([--enable-coverage],[enable compiler flags to support kcov coverage analysis]), + AS_HELP_STRING([--enable-coverage],[enable compiler flags to support kcov coverage analysis [default=no]]), [enable_coverage=$enableval], [enable_coverage=no]) AC_ARG_ENABLE(tests, - AS_HELP_STRING([--enable-tests],[compile tests (default is yes)]), + AS_HELP_STRING([--enable-tests],[compile tests [default=yes]]), [use_tests=$enableval], [use_tests=yes]) AC_ARG_ENABLE(openssl_tests, - AS_HELP_STRING([--enable-openssl-tests],[enable OpenSSL tests, if OpenSSL is available (default is auto)]), + AS_HELP_STRING([--enable-openssl-tests],[enable OpenSSL tests [default=auto]]), [enable_openssl_tests=$enableval], [enable_openssl_tests=auto]) AC_ARG_ENABLE(experimental, - AS_HELP_STRING([--enable-experimental],[allow experimental configure options (default is no)]), + AS_HELP_STRING([--enable-experimental],[allow experimental configure options [default=no]]), [use_experimental=$enableval], [use_experimental=no]) AC_ARG_ENABLE(exhaustive_tests, - AS_HELP_STRING([--enable-exhaustive-tests],[compile exhaustive tests (default is yes)]), + AS_HELP_STRING([--enable-exhaustive-tests],[compile exhaustive tests [default=yes]]), [use_exhaustive_tests=$enableval], [use_exhaustive_tests=yes]) AC_ARG_ENABLE(endomorphism, - AS_HELP_STRING([--enable-endomorphism],[enable endomorphism (default is no)]), + AS_HELP_STRING([--enable-endomorphism],[enable endomorphism [default=no]]), [use_endomorphism=$enableval], [use_endomorphism=no]) AC_ARG_ENABLE(ecmult_static_precomputation, - AS_HELP_STRING([--enable-ecmult-static-precomputation],[enable precomputed ecmult table for signing (default is yes)]), + AS_HELP_STRING([--enable-ecmult-static-precomputation],[enable precomputed ecmult table for signing [default=auto]]), [use_ecmult_static_precomputation=$enableval], [use_ecmult_static_precomputation=auto]) @@ -130,65 +130,100 @@ AC_ARG_ENABLE(module_ecdh, [enable_module_ecdh=no]) AC_ARG_ENABLE(module_recovery, - AS_HELP_STRING([--enable-module-recovery],[enable ECDSA pubkey recovery module (default is no)]), + AS_HELP_STRING([--enable-module-recovery],[enable ECDSA pubkey recovery module [default=no]]), [enable_module_recovery=$enableval], [enable_module_recovery=no]) +AC_ARG_ENABLE(external_default_callbacks, + AS_HELP_STRING([--enable-external-default-callbacks],[enable external default callback functions (default is no)]), + [use_external_default_callbacks=$enableval], + [use_external_default_callbacks=no]) + AC_ARG_ENABLE(jni, - AS_HELP_STRING([--enable-jni],[enable libsecp256k1_jni (default is no)]), + AS_HELP_STRING([--enable-jni],[enable libsecp256k1_jni [default=no]]), [use_jni=$enableval], [use_jni=no]) AC_ARG_WITH([field], [AS_HELP_STRING([--with-field=64bit|32bit|auto], -[Specify Field Implementation. Default is auto])],[req_field=$withval], [req_field=auto]) +[finite field implementation to use [default=auto]])],[req_field=$withval], [req_field=auto]) AC_ARG_WITH([bignum], [AS_HELP_STRING([--with-bignum=gmp|no|auto], -[Specify Bignum Implementation. Default is auto])],[req_bignum=$withval], [req_bignum=auto]) +[bignum implementation to use [default=auto]])],[req_bignum=$withval], [req_bignum=auto]) AC_ARG_WITH([scalar], [AS_HELP_STRING([--with-scalar=64bit|32bit|auto], -[Specify scalar implementation. Default is auto])],[req_scalar=$withval], [req_scalar=auto]) +[scalar implementation to use [default=auto]])],[req_scalar=$withval], [req_scalar=auto]) -AC_ARG_WITH([asm], [AS_HELP_STRING([--with-asm=x86_64|arm|no|auto] -[Specify assembly optimizations to use. Default is auto (experimental: arm)])],[req_asm=$withval], [req_asm=auto]) +AC_ARG_WITH([asm], [AS_HELP_STRING([--with-asm=x86_64|arm|no|auto], +[assembly optimizations to use (experimental: arm) [default=auto]])],[req_asm=$withval], [req_asm=auto]) + +AC_ARG_WITH([ecmult-window], [AS_HELP_STRING([--with-ecmult-window=SIZE|auto], +[window size for ecmult precomputation for verification, specified as integer in range [2..24].] +[Larger values result in possibly better performance at the cost of an exponentially larger precomputed table.] +[The table will store 2^(SIZE-2) * 64 bytes of data but can be larger in memory due to platform-specific padding and alignment.] +[If the endomorphism optimization is enabled, two tables of this size are used instead of only one.] +["auto" is a reasonable setting for desktop machines (currently 15). [default=auto]] +)], +[req_ecmult_window=$withval], [req_ecmult_window=auto]) AC_CHECK_TYPES([__int128]) -AC_MSG_CHECKING([for __builtin_expect]) -AC_COMPILE_IFELSE([AC_LANG_SOURCE([[void myfunc() {__builtin_expect(0,0);}]])], - [ AC_MSG_RESULT([yes]);AC_DEFINE(HAVE_BUILTIN_EXPECT,1,[Define this symbol if __builtin_expect is available]) ], - [ AC_MSG_RESULT([no]) - ]) - if test x"$enable_coverage" = x"yes"; then AC_DEFINE(COVERAGE, 1, [Define this symbol to compile out all VERIFY code]) CFLAGS="$CFLAGS -O0 --coverage" - LDFLAGS="--coverage" + LDFLAGS="$LDFLAGS --coverage" else CFLAGS="$CFLAGS -O3" fi if test x"$use_ecmult_static_precomputation" != x"no"; then + # Temporarily switch to an environment for the native compiler save_cross_compiling=$cross_compiling cross_compiling=no - TEMP_CC="$CC" + SAVE_CC="$CC" CC="$CC_FOR_BUILD" - AC_MSG_CHECKING([native compiler: ${CC_FOR_BUILD}]) + SAVE_CFLAGS="$CFLAGS" + CFLAGS="$CFLAGS_FOR_BUILD" + SAVE_CPPFLAGS="$CPPFLAGS" + CPPFLAGS="$CPPFLAGS_FOR_BUILD" + SAVE_LDFLAGS="$LDFLAGS" + LDFLAGS="$LDFLAGS_FOR_BUILD" + + warn_CFLAGS_FOR_BUILD="-Wall -Wextra -Wno-unused-function" + saved_CFLAGS="$CFLAGS" + CFLAGS="$CFLAGS $warn_CFLAGS_FOR_BUILD" + AC_MSG_CHECKING([if native ${CC_FOR_BUILD} supports ${warn_CFLAGS_FOR_BUILD}]) + AC_COMPILE_IFELSE([AC_LANG_SOURCE([[char foo;]])], + [ AC_MSG_RESULT([yes]) ], + [ AC_MSG_RESULT([no]) + CFLAGS="$saved_CFLAGS" + ]) + + AC_MSG_CHECKING([for working native compiler: ${CC_FOR_BUILD}]) AC_RUN_IFELSE( - [AC_LANG_PROGRAM([], [return 0])], + [AC_LANG_PROGRAM([], [])], [working_native_cc=yes], [working_native_cc=no],[dnl]) - CC="$TEMP_CC" + + CFLAGS_FOR_BUILD="$CFLAGS" + + # Restore the environment cross_compiling=$save_cross_compiling + CC="$SAVE_CC" + CFLAGS="$SAVE_CFLAGS" + CPPFLAGS="$SAVE_CPPFLAGS" + LDFLAGS="$SAVE_LDFLAGS" if test x"$working_native_cc" = x"no"; then + AC_MSG_RESULT([no]) set_precomp=no + m4_define([please_set_for_build], [Please set CC_FOR_BUILD, CFLAGS_FOR_BUILD, CPPFLAGS_FOR_BUILD, and/or LDFLAGS_FOR_BUILD.]) if test x"$use_ecmult_static_precomputation" = x"yes"; then - AC_MSG_ERROR([${CC_FOR_BUILD} does not produce working binaries. Please set CC_FOR_BUILD]) + AC_MSG_ERROR([native compiler ${CC_FOR_BUILD} does not produce working binaries. please_set_for_build]) else - AC_MSG_RESULT([${CC_FOR_BUILD} does not produce working binaries. Please set CC_FOR_BUILD]) + AC_MSG_WARN([Disabling statically generated ecmult table because the native compiler ${CC_FOR_BUILD} does not produce working binaries. please_set_for_build]) fi else - AC_MSG_RESULT([ok]) + AC_MSG_RESULT([yes]) set_precomp=yes fi else @@ -366,6 +401,28 @@ case $set_scalar in ;; esac +#set ecmult window size +if test x"$req_ecmult_window" = x"auto"; then + set_ecmult_window=15 +else + set_ecmult_window=$req_ecmult_window +fi + +error_window_size=['window size for ecmult precomputation not an integer in range [2..24] or "auto"'] +case $set_ecmult_window in +''|*[[!0-9]]*) + # no valid integer + AC_MSG_ERROR($error_window_size) + ;; +*) + if test "$set_ecmult_window" -lt 2 -o "$set_ecmult_window" -gt 24 ; then + # not in range + AC_MSG_ERROR($error_window_size) + fi + AC_DEFINE_UNQUOTED(ECMULT_WINDOW_SIZE, $set_ecmult_window, [Set window size for ecmult precomputation]) + ;; +esac + if test x"$use_tests" = x"yes"; then SECP_OPENSSL_CHECK if test x"$has_openssl_ec" = x"yes"; then @@ -441,17 +498,9 @@ if test x"$use_external_asm" = x"yes"; then AC_DEFINE(USE_EXTERNAL_ASM, 1, [Define this symbol if an external (non-inline) assembly implementation is used]) fi -AC_MSG_NOTICE([Using static precomputation: $set_precomp]) -AC_MSG_NOTICE([Using assembly optimizations: $set_asm]) -AC_MSG_NOTICE([Using field implementation: $set_field]) -AC_MSG_NOTICE([Using bignum implementation: $set_bignum]) -AC_MSG_NOTICE([Using scalar implementation: $set_scalar]) -AC_MSG_NOTICE([Using endomorphism optimizations: $use_endomorphism]) -AC_MSG_NOTICE([Building benchmarks: $use_benchmark]) -AC_MSG_NOTICE([Building for coverage analysis: $enable_coverage]) -AC_MSG_NOTICE([Building ECDH module: $enable_module_ecdh]) -AC_MSG_NOTICE([Building ECDSA pubkey recovery module: $enable_module_recovery]) -AC_MSG_NOTICE([Using jni: $use_jni]) +if test x"$use_external_default_callbacks" = x"yes"; then + AC_DEFINE(USE_EXTERNAL_DEFAULT_CALLBACKS, 1, [Define this symbol if an external implementation of the default callbacks is used]) +fi if test x"$enable_experimental" = x"yes"; then AC_MSG_NOTICE([******]) @@ -482,7 +531,7 @@ AM_CONDITIONAL([USE_BENCHMARK], [test x"$use_benchmark" = x"yes"]) AM_CONDITIONAL([USE_ECMULT_STATIC_PRECOMPUTATION], [test x"$set_precomp" = x"yes"]) AM_CONDITIONAL([ENABLE_MODULE_ECDH], [test x"$enable_module_ecdh" = x"yes"]) AM_CONDITIONAL([ENABLE_MODULE_RECOVERY], [test x"$enable_module_recovery" = x"yes"]) -AM_CONDITIONAL([USE_JNI], [test x"$use_jni" == x"yes"]) +AM_CONDITIONAL([USE_JNI], [test x"$use_jni" = x"yes"]) AM_CONDITIONAL([USE_EXTERNAL_ASM], [test x"$use_external_asm" = x"yes"]) AM_CONDITIONAL([USE_ASM_ARM], [test x"$set_asm" = x"arm"]) @@ -492,3 +541,26 @@ unset PKG_CONFIG_PATH PKG_CONFIG_PATH="$PKGCONFIG_PATH_TEMP" AC_OUTPUT + +echo +echo "Build Options:" +echo " with endomorphism = $use_endomorphism" +echo " with ecmult precomp = $set_precomp" +echo " with external callbacks = $use_external_default_callbacks" +echo " with jni = $use_jni" +echo " with benchmarks = $use_benchmark" +echo " with coverage = $enable_coverage" +echo " module ecdh = $enable_module_ecdh" +echo " module recovery = $enable_module_recovery" +echo +echo " asm = $set_asm" +echo " bignum = $set_bignum" +echo " field = $set_field" +echo " scalar = $set_scalar" +echo " ecmult window size = $set_ecmult_window" +echo +echo " CC = $CC" +echo " CFLAGS = $CFLAGS" +echo " CPPFLAGS = $CPPFLAGS" +echo " LDFLAGS = $LDFLAGS" +echo diff --git a/depend/secp256k1/include/secp256k1.h b/depend/secp256k1/include/secp256k1.h index f1f78ab..3e90b1b 100644 --- a/depend/secp256k1/include/secp256k1.h +++ b/depend/secp256k1/include/secp256k1.h @@ -33,9 +33,10 @@ extern "C" { * verification). * * A constructed context can safely be used from multiple threads - * simultaneously, but API call that take a non-const pointer to a context + * 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 and secp256k1_context_randomize. + * 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. @@ -163,7 +164,8 @@ typedef int (*secp256k1_nonce_function)( #define SECP256K1_FLAGS_BIT_CONTEXT_SIGN (1 << 9) #define SECP256K1_FLAGS_BIT_COMPRESSION (1 << 8) -/** Flags to pass to secp256k1_context_create. */ +/** Flags to pass to secp256k1_context_create, secp256k1_context_preallocated_size, and + * secp256k1_context_preallocated_create. */ #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) #define SECP256K1_CONTEXT_NONE (SECP256K1_FLAGS_TYPE_CONTEXT) @@ -186,7 +188,11 @@ typedef int (*secp256k1_nonce_function)( */ SECP256K1_API extern const secp256k1_context *secp256k1_context_no_precomp; -/** Create a secp256k1 context object. +/** 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 the functions in secp256k1_preallocated.h. * * Returns: a newly created context object. * In: flags: which parts of the context to initialize. @@ -197,7 +203,11 @@ SECP256K1_API secp256k1_context* secp256k1_context_create( unsigned int flags ) SECP256K1_WARN_UNUSED_RESULT; -/** Copies a secp256k1 context object. +/** 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. * * Returns: a newly created context object. * Args: ctx: an existing context to copy (cannot be NULL) @@ -206,10 +216,18 @@ SECP256K1_API secp256k1_context* secp256k1_context_clone( const secp256k1_context* ctx ) SECP256K1_ARG_NONNULL(1) SECP256K1_WARN_UNUSED_RESULT; -/** Destroy a secp256k1 context object. +/** Destroy a secp256k1 context object (created in dynamically allocated memory). * * The context pointer may not be used afterwards. - * Args: ctx: an existing context to destroy (cannot be NULL) + * + * 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: an existing context to destroy, constructed using + * secp256k1_context_create or secp256k1_context_clone */ SECP256K1_API void secp256k1_context_destroy( secp256k1_context* ctx @@ -229,11 +247,28 @@ SECP256K1_API void secp256k1_context_destroy( * 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_illegal_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: an existing context object (cannot be NULL) * In: fun: a pointer to a function to call when an illegal argument is - * passed to the API, taking a message and an opaque pointer - * (NULL restores a default handler that calls abort). + * 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. + * + * See also secp256k1_context_set_error_callback. */ SECP256K1_API void secp256k1_context_set_illegal_callback( secp256k1_context* ctx, @@ -253,9 +288,12 @@ SECP256K1_API void secp256k1_context_set_illegal_callback( * * Args: ctx: an existing context object (cannot be NULL) * In: fun: a pointer to a function to call when an internal error occurs, - * taking a message and an opaque pointer (NULL restores a default - * handler that calls abort). + * 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. + * + * See also secp256k1_context_set_illegal_callback. */ SECP256K1_API void secp256k1_context_set_error_callback( secp256k1_context* ctx, @@ -267,21 +305,24 @@ SECP256K1_API void secp256k1_context_set_error_callback( * * Returns: a newly created scratch space. * Args: ctx: an existing context object (cannot be NULL) - * In: max_size: maximum amount of memory to allocate + * In: size: amount of memory to be available as scratch space. Some extra + * (<100 bytes) will be allocated for extra accounting. */ SECP256K1_API SECP256K1_WARN_UNUSED_RESULT secp256k1_scratch_space* secp256k1_scratch_space_create( const secp256k1_context* ctx, - size_t max_size + size_t size ) SECP256K1_ARG_NONNULL(1); /** Destroy a secp256k1 scratch space. * * The pointer may not be used afterwards. - * Args: scratch: space to destroy + * Args: ctx: a secp256k1 context object. + * scratch: space to destroy */ SECP256K1_API void secp256k1_scratch_space_destroy( + const secp256k1_context* ctx, secp256k1_scratch_space* scratch -); +) SECP256K1_ARG_NONNULL(1); /** Parse a variable-length public key into the pubkey object. * @@ -615,7 +656,7 @@ SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_ec_pubkey_tweak_mul( ) SECP256K1_ARG_NONNULL(1) SECP256K1_ARG_NONNULL(2) SECP256K1_ARG_NONNULL(3); /** Updates the context randomization to protect against side-channel leakage. - * Returns: 1: randomization successfully updated + * Returns: 1: randomization successfully updated or nothing to randomize * 0: error * Args: ctx: pointer to a context object (cannot be NULL) * In: seed32: pointer to a 32-byte random seed (NULL resets to initial state) @@ -630,8 +671,14 @@ SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_ec_pubkey_tweak_mul( * that it does not affect function results, but shields against attacks which * rely on any input-dependent behaviour. * + * This function has currently an effect only on contexts initialized for signing + * because randomization is currently used only for signing. However, this is not + * guaranteed and may change in the future. It is safe to call this function on + * contexts not initialized for signing; then it will have no effect and return 1. + * * You should call this after secp256k1_context_create or - * secp256k1_context_clone, and may call this repeatedly afterwards. + * secp256k1_context_clone (and secp256k1_context_preallocated_create or + * secp256k1_context_clone, resp.), and you may call this repeatedly afterwards. */ SECP256K1_API SECP256K1_WARN_UNUSED_RESULT int secp256k1_context_randomize( secp256k1_context* ctx, diff --git a/depend/secp256k1/include/secp256k1_preallocated.h b/depend/secp256k1/include/secp256k1_preallocated.h new file mode 100644 index 0000000..0fb64a5 --- /dev/null +++ b/depend/secp256k1/include/secp256k1_preallocated.h @@ -0,0 +1,128 @@ +#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 excepts 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 by 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: a newly created context object. + * In: prealloc: a pointer to a rewritable contiguous block of memory of + * size at least secp256k1_context_preallocated_size(flags) + * bytes, as detailed above (cannot be NULL) + * flags: which parts of the context to initialize. + * + * 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: an existing context to copy (cannot be NULL) + */ +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. + * + * Returns: a newly created context object. + * Args: ctx: an existing context to copy (cannot be NULL) + * In: prealloc: a pointer to a rewritable contiguous block of memory of + * size at least secp256k1_context_preallocated_size(flags) + * bytes, as detailed above (cannot be NULL) + */ +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: an existing context to destroy, constructed using + * secp256k1_context_preallocated_create or + * secp256k1_context_preallocated_clone (cannot be NULL) + */ +SECP256K1_API void secp256k1_context_preallocated_destroy( + secp256k1_context* ctx +); + +#ifdef __cplusplus +} +#endif + +#endif /* SECP256K1_PREALLOCATED_H */ diff --git a/depend/secp256k1/libsecp256k1.pc.in b/depend/secp256k1/libsecp256k1.pc.in index a0d006f..694e98e 100644 --- a/depend/secp256k1/libsecp256k1.pc.in +++ b/depend/secp256k1/libsecp256k1.pc.in @@ -8,6 +8,6 @@ Description: Optimized C library for EC operations on curve secp256k1 URL: https://github.com/bitcoin-core/secp256k1 Version: @PACKAGE_VERSION@ Cflags: -I${includedir} -Libs.private: @SECP_LIBS@ Libs: -L${libdir} -lsecp256k1 +Libs.private: @SECP_LIBS@ diff --git a/depend/secp256k1/src/asm/field_10x26_arm.s b/depend/secp256k1/src/asm/field_10x26_arm.s index 5a9cc3f..9a5bd06 100644 --- a/depend/secp256k1/src/asm/field_10x26_arm.s +++ b/depend/secp256k1/src/asm/field_10x26_arm.s @@ -16,15 +16,9 @@ Note: */ .syntax unified - .arch armv7-a @ eabi attributes - see readelf -A - .eabi_attribute 8, 1 @ Tag_ARM_ISA_use = yes - .eabi_attribute 9, 0 @ Tag_Thumb_ISA_use = no - .eabi_attribute 10, 0 @ Tag_FP_arch = none .eabi_attribute 24, 1 @ Tag_ABI_align_needed = 8-byte .eabi_attribute 25, 1 @ Tag_ABI_align_preserved = 8-byte, except leaf SP - .eabi_attribute 30, 2 @ Tag_ABI_optimization_goals = Aggressive Speed - .eabi_attribute 34, 1 @ Tag_CPU_unaligned_access = v6 .text @ Field constants diff --git a/depend/secp256k1/src/basic-config.h b/depend/secp256k1/src/basic-config.h index fc58806..3a3969d 100644 --- a/depend/secp256k1/src/basic-config.h +++ b/depend/secp256k1/src/basic-config.h @@ -10,7 +10,10 @@ #ifdef USE_BASIC_CONFIG #undef USE_ASM_X86_64 +#undef USE_ECMULT_STATIC_PRECOMPUTATION #undef USE_ENDOMORPHISM +#undef USE_EXTERNAL_ASM +#undef USE_EXTERNAL_DEFAULT_CALLBACKS #undef USE_FIELD_10X26 #undef USE_FIELD_5X52 #undef USE_FIELD_INV_BUILTIN @@ -27,6 +30,7 @@ #define USE_SCALAR_INV_BUILTIN 1 #define USE_FIELD_10X26 1 #define USE_SCALAR_8X32 1 +#define ECMULT_WINDOW_SIZE 15 #endif /* USE_BASIC_CONFIG */ diff --git a/depend/secp256k1/src/bench_ecmult.c b/depend/secp256k1/src/bench_ecmult.c index 52d0476..7b5d185 100644 --- a/depend/secp256k1/src/bench_ecmult.c +++ b/depend/secp256k1/src/bench_ecmult.c @@ -64,7 +64,7 @@ static void bench_ecmult(void* arg) { size_t iter; for (iter = 0; iter < iters; ++iter) { - data->ecmult_multi(&data->ctx->ecmult_ctx, data->scratch, &data->output[iter], data->includes_g ? &data->scalars[data->offset1] : NULL, bench_callback, arg, count - includes_g); + data->ecmult_multi(&data->ctx->error_callback, &data->ctx->ecmult_ctx, data->scratch, &data->output[iter], data->includes_g ? &data->scalars[data->offset1] : NULL, bench_callback, arg, count - includes_g); data->offset1 = (data->offset1 + count) % POINTS; data->offset2 = (data->offset2 + count - 1) % POINTS; } @@ -139,6 +139,11 @@ int main(int argc, char **argv) { secp256k1_gej* pubkeys_gej; size_t scratch_size; + data.ctx = secp256k1_context_create(SECP256K1_CONTEXT_SIGN | SECP256K1_CONTEXT_VERIFY); + scratch_size = secp256k1_strauss_scratch_size(POINTS) + STRAUSS_SCRATCH_OBJECTS*16; + data.scratch = secp256k1_scratch_space_create(data.ctx, scratch_size); + data.ecmult_multi = secp256k1_ecmult_multi_var; + if (argc > 1) { if(have_flag(argc, argv, "pippenger_wnaf")) { printf("Using pippenger_wnaf:\n"); @@ -146,15 +151,19 @@ int main(int argc, char **argv) { } 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"); + data.ecmult_multi = secp256k1_ecmult_multi_var; + secp256k1_scratch_space_destroy(data.ctx, data.scratch); + data.scratch = NULL; + } else { + fprintf(stderr, "%s: unrecognized argument '%s'.\n", argv[0], argv[1]); + fprintf(stderr, "Use 'pippenger_wnaf', 'strauss_wnaf', 'simple' or no argument to benchmark a combined algorithm.\n"); + return 1; } - } else { - data.ecmult_multi = secp256k1_ecmult_multi_var; } /* Allocate stuff */ - data.ctx = secp256k1_context_create(SECP256K1_CONTEXT_SIGN | SECP256K1_CONTEXT_VERIFY); - scratch_size = secp256k1_strauss_scratch_size(POINTS) + STRAUSS_SCRATCH_OBJECTS*16; - data.scratch = secp256k1_scratch_space_create(data.ctx, scratch_size); data.scalars = malloc(sizeof(secp256k1_scalar) * POINTS); data.seckeys = malloc(sizeof(secp256k1_scalar) * POINTS); data.pubkeys = malloc(sizeof(secp256k1_ge) * POINTS); @@ -172,7 +181,7 @@ int main(int argc, char **argv) { secp256k1_scalar_add(&data.seckeys[i], &data.seckeys[i - 1], &data.seckeys[i - 1]); } } - secp256k1_ge_set_all_gej_var(data.pubkeys, pubkeys_gej, POINTS, &data.ctx->error_callback); + secp256k1_ge_set_all_gej_var(data.pubkeys, pubkeys_gej, POINTS); free(pubkeys_gej); for (i = 1; i <= 8; ++i) { @@ -184,8 +193,10 @@ int main(int argc, char **argv) { run_test(&data, i << p, 1); } } + if (data.scratch != NULL) { + secp256k1_scratch_space_destroy(data.ctx, data.scratch); + } secp256k1_context_destroy(data.ctx); - secp256k1_scratch_space_destroy(data.scratch); free(data.scalars); free(data.pubkeys); free(data.seckeys); diff --git a/depend/secp256k1/src/bench_internal.c b/depend/secp256k1/src/bench_internal.c index 9c0a07f..a8f4e9e 100644 --- a/depend/secp256k1/src/bench_internal.c +++ b/depend/secp256k1/src/bench_internal.c @@ -184,9 +184,11 @@ void bench_field_inverse_var(void* arg) { void bench_field_sqrt(void* arg) { int i; bench_inv *data = (bench_inv*)arg; + secp256k1_fe t; for (i = 0; i < 20000; i++) { - secp256k1_fe_sqrt(&data->fe_x, &data->fe_x); + t = data->fe_x; + secp256k1_fe_sqrt(&data->fe_x, &t); secp256k1_fe_add(&data->fe_x, &data->fe_y); } } @@ -251,7 +253,7 @@ void bench_wnaf_const(void* arg) { bench_inv *data = (bench_inv*)arg; for (i = 0; i < 20000; i++) { - secp256k1_wnaf_const(data->wnaf, data->scalar_x, WINDOW_A, 256); + secp256k1_wnaf_const(data->wnaf, &data->scalar_x, WINDOW_A, 256); secp256k1_scalar_add(&data->scalar_x, &data->scalar_x, &data->scalar_y); } } diff --git a/depend/secp256k1/src/eckey_impl.h b/depend/secp256k1/src/eckey_impl.h index 1ab9a68..7c5b789 100644 --- a/depend/secp256k1/src/eckey_impl.h +++ b/depend/secp256k1/src/eckey_impl.h @@ -18,7 +18,7 @@ static int secp256k1_eckey_pubkey_parse(secp256k1_ge *elem, const unsigned char if (size == 33 && (pub[0] == SECP256K1_TAG_PUBKEY_EVEN || pub[0] == SECP256K1_TAG_PUBKEY_ODD)) { secp256k1_fe x; return secp256k1_fe_set_b32(&x, pub+1) && secp256k1_ge_set_xo_var(elem, &x, pub[0] == SECP256K1_TAG_PUBKEY_ODD); - } else if (size == 65 && (pub[0] == 0x04 || pub[0] == 0x06 || pub[0] == 0x07)) { + } 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(&x, pub+1) || !secp256k1_fe_set_b32(&y, pub+33)) { return 0; diff --git a/depend/secp256k1/src/ecmult.h b/depend/secp256k1/src/ecmult.h index ea1cd8a..c9b1982 100644 --- a/depend/secp256k1/src/ecmult.h +++ b/depend/secp256k1/src/ecmult.h @@ -20,10 +20,10 @@ typedef struct { #endif } secp256k1_ecmult_context; +static const size_t SECP256K1_ECMULT_CONTEXT_PREALLOCATED_SIZE; static void secp256k1_ecmult_context_init(secp256k1_ecmult_context *ctx); -static void secp256k1_ecmult_context_build(secp256k1_ecmult_context *ctx, const secp256k1_callback *cb); -static void secp256k1_ecmult_context_clone(secp256k1_ecmult_context *dst, - const secp256k1_ecmult_context *src, const secp256k1_callback *cb); +static void secp256k1_ecmult_context_build(secp256k1_ecmult_context *ctx, void **prealloc); +static void secp256k1_ecmult_context_finalize_memcpy(secp256k1_ecmult_context *dst, const secp256k1_ecmult_context *src); static void secp256k1_ecmult_context_clear(secp256k1_ecmult_context *ctx); static int secp256k1_ecmult_context_is_built(const secp256k1_ecmult_context *ctx); @@ -37,11 +37,12 @@ typedef int (secp256k1_ecmult_multi_callback)(secp256k1_scalar *sc, secp256k1_ge * 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. + * 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_ecmult_context *ctx, secp256k1_scratch *scratch, secp256k1_gej *r, const secp256k1_scalar *inp_g_sc, secp256k1_ecmult_multi_callback cb, void *cbdata, size_t n); +static int secp256k1_ecmult_multi_var(const secp256k1_callback* error_callback, const secp256k1_ecmult_context *ctx, 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 */ diff --git a/depend/secp256k1/src/ecmult_const_impl.h b/depend/secp256k1/src/ecmult_const_impl.h index 8411752..aaa576a 100644 --- a/depend/secp256k1/src/ecmult_const_impl.h +++ b/depend/secp256k1/src/ecmult_const_impl.h @@ -48,7 +48,7 @@ * * Numbers reference steps of `Algorithm SPA-resistant Width-w NAF with Odd Scalar` on pp. 335 */ -static int secp256k1_wnaf_const(int *wnaf, secp256k1_scalar s, int w, int size) { +static int secp256k1_wnaf_const(int *wnaf, const secp256k1_scalar *scalar, int w, int size) { int global_sign; int skew = 0; int word = 0; @@ -59,8 +59,12 @@ static int secp256k1_wnaf_const(int *wnaf, secp256k1_scalar s, int w, int size) int flip; int bit; - secp256k1_scalar neg_s; + secp256k1_scalar s; int not_neg_one; + + VERIFY_CHECK(w > 0); + VERIFY_CHECK(size > 0); + /* Note that we cannot handle even numbers by negating them to be odd, as is * done in other implementations, since if our scalars were specified to have * width < 256 for performance reasons, their negations would have width 256 @@ -75,12 +79,13 @@ static int secp256k1_wnaf_const(int *wnaf, secp256k1_scalar s, int w, int size) * {1, 2} we want to add to the scalar when ensuring that it's odd. Further * complicating things, -1 interacts badly with `secp256k1_scalar_cadd_bit` and * we need to special-case it in this logic. */ - flip = secp256k1_scalar_is_high(&s); + flip = secp256k1_scalar_is_high(scalar); /* We add 1 to even numbers, 2 to odd ones, noting that negation flips parity */ - bit = flip ^ !secp256k1_scalar_is_even(&s); + bit = flip ^ !secp256k1_scalar_is_even(scalar); /* We check for negative one, since adding 2 to it will cause an overflow */ - secp256k1_scalar_negate(&neg_s, &s); - not_neg_one = !secp256k1_scalar_is_one(&neg_s); + secp256k1_scalar_negate(&s, scalar); + not_neg_one = !secp256k1_scalar_is_one(&s); + s = *scalar; secp256k1_scalar_cadd_bit(&s, bit, not_neg_one); /* If we had negative one, flip == 1, s.d[0] == 0, bit == 1, so caller expects * that we added two to it and flipped it. In fact for -1 these operations are @@ -93,7 +98,7 @@ static int secp256k1_wnaf_const(int *wnaf, secp256k1_scalar s, int w, int size) /* 4 */ u_last = secp256k1_scalar_shr_int(&s, w); - while (word * w < size) { + do { int sign; int even; @@ -109,7 +114,7 @@ static int secp256k1_wnaf_const(int *wnaf, secp256k1_scalar s, int w, int size) wnaf[word++] = u_last * global_sign; u_last = u; - } + } while (word * w < size); wnaf[word] = u * global_sign; VERIFY_CHECK(secp256k1_scalar_is_zero(&s)); @@ -132,7 +137,6 @@ static void secp256k1_ecmult_const(secp256k1_gej *r, const secp256k1_ge *a, cons int wnaf_1[1 + WNAF_SIZE(WINDOW_A - 1)]; int i; - secp256k1_scalar sc = *scalar; /* build wnaf representation for q. */ int rsize = size; @@ -140,13 +144,13 @@ static void secp256k1_ecmult_const(secp256k1_gej *r, const secp256k1_ge *a, cons if (size > 128) { rsize = 128; /* split q into q_1 and q_lam (where q = q_1 + q_lam*lambda, and q_1 and q_lam are ~128 bit) */ - secp256k1_scalar_split_lambda(&q_1, &q_lam, &sc); - skew_1 = secp256k1_wnaf_const(wnaf_1, q_1, WINDOW_A - 1, 128); - skew_lam = secp256k1_wnaf_const(wnaf_lam, q_lam, WINDOW_A - 1, 128); + secp256k1_scalar_split_lambda(&q_1, &q_lam, scalar); + skew_1 = secp256k1_wnaf_const(wnaf_1, &q_1, WINDOW_A - 1, 128); + skew_lam = secp256k1_wnaf_const(wnaf_lam, &q_lam, WINDOW_A - 1, 128); } else #endif { - skew_1 = secp256k1_wnaf_const(wnaf_1, sc, WINDOW_A - 1, size); + skew_1 = secp256k1_wnaf_const(wnaf_1, scalar, WINDOW_A - 1, size); #ifdef USE_ENDOMORPHISM skew_lam = 0; #endif diff --git a/depend/secp256k1/src/ecmult_gen.h b/depend/secp256k1/src/ecmult_gen.h index 7564b70..b136e94 100644 --- a/depend/secp256k1/src/ecmult_gen.h +++ b/depend/secp256k1/src/ecmult_gen.h @@ -28,10 +28,10 @@ typedef struct { secp256k1_gej initial; } secp256k1_ecmult_gen_context; +static const size_t SECP256K1_ECMULT_GEN_CONTEXT_PREALLOCATED_SIZE; static void secp256k1_ecmult_gen_context_init(secp256k1_ecmult_gen_context* ctx); -static void secp256k1_ecmult_gen_context_build(secp256k1_ecmult_gen_context* ctx, const secp256k1_callback* cb); -static void secp256k1_ecmult_gen_context_clone(secp256k1_ecmult_gen_context *dst, - const secp256k1_ecmult_gen_context* src, const secp256k1_callback* cb); +static void secp256k1_ecmult_gen_context_build(secp256k1_ecmult_gen_context* ctx, void **prealloc); +static void secp256k1_ecmult_gen_context_finalize_memcpy(secp256k1_ecmult_gen_context *dst, const secp256k1_ecmult_gen_context* src); static void secp256k1_ecmult_gen_context_clear(secp256k1_ecmult_gen_context* ctx); static int secp256k1_ecmult_gen_context_is_built(const secp256k1_ecmult_gen_context* ctx); diff --git a/depend/secp256k1/src/ecmult_gen_impl.h b/depend/secp256k1/src/ecmult_gen_impl.h index 714f02e..f818d45 100644 --- a/depend/secp256k1/src/ecmult_gen_impl.h +++ b/depend/secp256k1/src/ecmult_gen_impl.h @@ -7,6 +7,7 @@ #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" @@ -14,23 +15,32 @@ #ifdef USE_ECMULT_STATIC_PRECOMPUTATION #include "ecmult_static_context.h" #endif + +#ifndef USE_ECMULT_STATIC_PRECOMPUTATION + static const size_t SECP256K1_ECMULT_GEN_CONTEXT_PREALLOCATED_SIZE = ROUND_TO_ALIGN(sizeof(*((secp256k1_ecmult_gen_context*) NULL)->prec)); +#else + static const size_t SECP256K1_ECMULT_GEN_CONTEXT_PREALLOCATED_SIZE = 0; +#endif + static void secp256k1_ecmult_gen_context_init(secp256k1_ecmult_gen_context *ctx) { ctx->prec = NULL; } -static void secp256k1_ecmult_gen_context_build(secp256k1_ecmult_gen_context *ctx, const secp256k1_callback* cb) { +static void secp256k1_ecmult_gen_context_build(secp256k1_ecmult_gen_context *ctx, void **prealloc) { #ifndef USE_ECMULT_STATIC_PRECOMPUTATION secp256k1_ge prec[1024]; secp256k1_gej gj; secp256k1_gej nums_gej; int i, j; + size_t const prealloc_size = SECP256K1_ECMULT_GEN_CONTEXT_PREALLOCATED_SIZE; + void* const base = *prealloc; #endif if (ctx->prec != NULL) { return; } #ifndef USE_ECMULT_STATIC_PRECOMPUTATION - ctx->prec = (secp256k1_ge_storage (*)[64][16])checked_malloc(cb, sizeof(*ctx->prec)); + ctx->prec = (secp256k1_ge_storage (*)[64][16])manual_alloc(prealloc, prealloc_size, base, prealloc_size); /* get the generator */ secp256k1_gej_set_ge(&gj, &secp256k1_ge_const_g); @@ -77,7 +87,7 @@ static void secp256k1_ecmult_gen_context_build(secp256k1_ecmult_gen_context *ctx secp256k1_gej_add_var(&numsbase, &numsbase, &nums_gej, NULL); } } - secp256k1_ge_set_all_gej_var(prec, precj, 1024, cb); + secp256k1_ge_set_all_gej_var(prec, precj, 1024); } for (j = 0; j < 64; j++) { for (i = 0; i < 16; i++) { @@ -85,7 +95,7 @@ static void secp256k1_ecmult_gen_context_build(secp256k1_ecmult_gen_context *ctx } } #else - (void)cb; + (void)prealloc; ctx->prec = (secp256k1_ge_storage (*)[64][16])secp256k1_ecmult_static_context; #endif secp256k1_ecmult_gen_blind(ctx, NULL); @@ -95,27 +105,18 @@ static int secp256k1_ecmult_gen_context_is_built(const secp256k1_ecmult_gen_cont return ctx->prec != NULL; } -static void secp256k1_ecmult_gen_context_clone(secp256k1_ecmult_gen_context *dst, - const secp256k1_ecmult_gen_context *src, const secp256k1_callback* cb) { - if (src->prec == NULL) { - dst->prec = NULL; - } else { +static void secp256k1_ecmult_gen_context_finalize_memcpy(secp256k1_ecmult_gen_context *dst, const secp256k1_ecmult_gen_context *src) { #ifndef USE_ECMULT_STATIC_PRECOMPUTATION - dst->prec = (secp256k1_ge_storage (*)[64][16])checked_malloc(cb, sizeof(*dst->prec)); - memcpy(dst->prec, src->prec, sizeof(*dst->prec)); -#else - (void)cb; - dst->prec = src->prec; -#endif - dst->initial = src->initial; - dst->blind = src->blind; + if (src->prec != NULL) { + /* We cast to void* first to suppress a -Wcast-align warning. */ + dst->prec = (secp256k1_ge_storage (*)[64][16])(void*)((unsigned char*)dst + ((unsigned char*)src->prec - (unsigned char*)src)); } +#else + (void)dst, (void)src; +#endif } static void secp256k1_ecmult_gen_context_clear(secp256k1_ecmult_gen_context *ctx) { -#ifndef USE_ECMULT_STATIC_PRECOMPUTATION - free(ctx->prec); -#endif secp256k1_scalar_clear(&ctx->blind); secp256k1_gej_clear(&ctx->initial); ctx->prec = NULL; diff --git a/depend/secp256k1/src/ecmult_impl.h b/depend/secp256k1/src/ecmult_impl.h index d5fb6c5..bb7aded 100644 --- a/depend/secp256k1/src/ecmult_impl.h +++ b/depend/secp256k1/src/ecmult_impl.h @@ -10,6 +10,7 @@ #include #include +#include "util.h" #include "group.h" #include "scalar.h" #include "ecmult.h" @@ -30,16 +31,32 @@ # endif #else /* optimal for 128-bit and 256-bit exponents. */ -#define WINDOW_A 5 -/** larger numbers may result in slightly better performance, at the cost of - exponentially larger precomputed tables. */ -#ifdef USE_ENDOMORPHISM -/** Two tables for window size 15: 1.375 MiB. */ -#define WINDOW_G 15 -#else -/** One table for window size 16: 1.375 MiB. */ -#define WINDOW_G 16 +# 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. + * If the endomorphism optimization is enabled (USE_ENDOMORMPHSIM) + * two tables of this size are used instead of only one. + */ +# define WINDOW_G ECMULT_WINDOW_SIZE #endif + +/* Noone 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 + * 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 #ifdef USE_ENDOMORPHISM @@ -137,24 +154,135 @@ static void secp256k1_ecmult_odd_multiples_table_globalz_windowa(secp256k1_ge *p secp256k1_ge_globalz_set_table_gej(ECMULT_TABLE_SIZE(WINDOW_A), pre, globalz, prej, zr); } -static void secp256k1_ecmult_odd_multiples_table_storage_var(int n, secp256k1_ge_storage *pre, const secp256k1_gej *a, const secp256k1_callback *cb) { - secp256k1_gej *prej = (secp256k1_gej*)checked_malloc(cb, sizeof(secp256k1_gej) * n); - secp256k1_ge *prea = (secp256k1_ge*)checked_malloc(cb, sizeof(secp256k1_ge) * n); - secp256k1_fe *zr = (secp256k1_fe*)checked_malloc(cb, sizeof(secp256k1_fe) * n); +static void secp256k1_ecmult_odd_multiples_table_storage_var(const int n, secp256k1_ge_storage *pre, const secp256k1_gej *a) { + secp256k1_gej d; + secp256k1_ge d_ge, p_ge; + secp256k1_gej pj; + secp256k1_fe zi; + secp256k1_fe zr; + secp256k1_fe dx_over_dz_squared; int i; - /* Compute the odd multiples in Jacobian form. */ - secp256k1_ecmult_odd_multiples_table(n, prej, zr, a); - /* Convert them in batch to affine coordinates. */ - secp256k1_ge_set_table_gej_var(prea, prej, zr, n); - /* Convert them to compact storage form. */ - for (i = 0; i < n; i++) { - secp256k1_ge_to_storage(&pre[i], &prea[i]); + VERIFY_CHECK(!a->infinity); + + secp256k1_gej_double_var(&d, a, NULL); + + /* First, we perform all the additions in an isomorphic curve obtained by multiplying + * all `z` coordinates by 1/`d.z`. In these coordinates `d` is affine so we can use + * `secp256k1_gej_add_ge_var` to perform the additions. For each addition, we store + * the resulting y-coordinate and the z-ratio, since we only have enough memory to + * store two field elements. These are sufficient to efficiently undo the isomorphism + * and recompute all the `x`s. + */ + d_ge.x = d.x; + d_ge.y = d.y; + d_ge.infinity = 0; + + secp256k1_ge_set_gej_zinv(&p_ge, a, &d.z); + pj.x = p_ge.x; + pj.y = p_ge.y; + pj.z = a->z; + pj.infinity = 0; + + for (i = 0; i < (n - 1); i++) { + secp256k1_fe_normalize_var(&pj.y); + secp256k1_fe_to_storage(&pre[i].y, &pj.y); + secp256k1_gej_add_ge_var(&pj, &pj, &d_ge, &zr); + secp256k1_fe_normalize_var(&zr); + secp256k1_fe_to_storage(&pre[i].x, &zr); } - free(prea); - free(prej); - free(zr); + /* Invert d.z in the same batch, preserving pj.z so we can extract 1/d.z */ + secp256k1_fe_mul(&zi, &pj.z, &d.z); + secp256k1_fe_inv_var(&zi, &zi); + + /* Directly set `pre[n - 1]` to `pj`, saving the inverted z-coordinate so + * that we can combine it with the saved z-ratios to compute the other zs + * without any more inversions. */ + secp256k1_ge_set_gej_zinv(&p_ge, &pj, &zi); + secp256k1_ge_to_storage(&pre[n - 1], &p_ge); + + /* Compute the actual x-coordinate of D, which will be needed below. */ + secp256k1_fe_mul(&d.z, &zi, &pj.z); /* d.z = 1/d.z */ + secp256k1_fe_sqr(&dx_over_dz_squared, &d.z); + secp256k1_fe_mul(&dx_over_dz_squared, &dx_over_dz_squared, &d.x); + + /* Going into the second loop, we have set `pre[n-1]` to its final affine + * form, but still need to set `pre[i]` for `i` in 0 through `n-2`. We + * have `zi = (p.z * d.z)^-1`, where + * + * `p.z` is the z-coordinate of the point on the isomorphic curve + * which was ultimately assigned to `pre[n-1]`. + * `d.z` is the multiplier that must be applied to all z-coordinates + * to move from our isomorphic curve back to secp256k1; so the + * product `p.z * d.z` is the z-coordinate of the secp256k1 + * point assigned to `pre[n-1]`. + * + * All subsequent inverse-z-coordinates can be obtained by multiplying this + * factor by successive z-ratios, which is much more efficient than directly + * computing each one. + * + * Importantly, these inverse-zs will be coordinates of points on secp256k1, + * while our other stored values come from computations on the isomorphic + * curve. So in the below loop, we will take care not to actually use `zi` + * or any derived values until we're back on secp256k1. + */ + i = n - 1; + while (i > 0) { + secp256k1_fe zi2, zi3; + const secp256k1_fe *rzr; + i--; + + secp256k1_ge_from_storage(&p_ge, &pre[i]); + + /* For each remaining point, we extract the z-ratio from the stored + * x-coordinate, compute its z^-1 from that, and compute the full + * point from that. */ + rzr = &p_ge.x; + secp256k1_fe_mul(&zi, &zi, rzr); + secp256k1_fe_sqr(&zi2, &zi); + secp256k1_fe_mul(&zi3, &zi2, &zi); + /* To compute the actual x-coordinate, we use the stored z ratio and + * y-coordinate, which we obtained from `secp256k1_gej_add_ge_var` + * in the loop above, as well as the inverse of the square of its + * z-coordinate. We store the latter in the `zi2` variable, which is + * computed iteratively starting from the overall Z inverse then + * multiplying by each z-ratio in turn. + * + * Denoting the z-ratio as `rzr`, we observe that it is equal to `h` + * from the inside of the above `gej_add_ge_var` call. This satisfies + * + * rzr = d_x * z^2 - x * d_z^2 + * + * where (`d_x`, `d_z`) are Jacobian coordinates of `D` and `(x, z)` + * are Jacobian coordinates of our desired point -- except both are on + * the isomorphic curve that we were using when we called `gej_add_ge_var`. + * To get back to secp256k1, we must multiply both `z`s by `d_z`, or + * equivalently divide both `x`s by `d_z^2`. Our equation then becomes + * + * rzr = d_x * z^2 / d_z^2 - x + * + * (The left-hand-side, being a ratio of z-coordinates, is unaffected + * by the isomorphism.) + * + * Rearranging to solve for `x`, we have + * + * x = d_x * z^2 / d_z^2 - rzr + * + * But what we actually want is the affine coordinate `X = x/z^2`, + * which will satisfy + * + * X = d_x / d_z^2 - rzr / z^2 + * = dx_over_dz_squared - rzr * zi2 + */ + secp256k1_fe_mul(&p_ge.x, rzr, &zi2); + secp256k1_fe_negate(&p_ge.x, &p_ge.x, 1); + secp256k1_fe_add(&p_ge.x, &dx_over_dz_squared); + /* y is stored_y/z^3, as we expect */ + secp256k1_fe_mul(&p_ge.y, &p_ge.y, &zi3); + /* Store */ + secp256k1_ge_to_storage(&pre[i], &p_ge); + } } /** The following two macro retrieves a particular odd multiple from a table @@ -166,7 +294,8 @@ static void secp256k1_ecmult_odd_multiples_table_storage_var(int n, secp256k1_ge if ((n) > 0) { \ *(r) = (pre)[((n)-1)/2]; \ } else { \ - secp256k1_ge_neg((r), &(pre)[(-(n)-1)/2]); \ + *(r) = (pre)[(-(n)-1)/2]; \ + secp256k1_fe_negate(&((r)->y), &((r)->y), 1); \ } \ } while(0) @@ -178,10 +307,17 @@ static void secp256k1_ecmult_odd_multiples_table_storage_var(int n, secp256k1_ge secp256k1_ge_from_storage((r), &(pre)[((n)-1)/2]); \ } else { \ secp256k1_ge_from_storage((r), &(pre)[(-(n)-1)/2]); \ - secp256k1_ge_neg((r), (r)); \ + secp256k1_fe_negate(&((r)->y), &((r)->y), 1); \ } \ } while(0) +static const size_t SECP256K1_ECMULT_CONTEXT_PREALLOCATED_SIZE = + ROUND_TO_ALIGN(sizeof((*((secp256k1_ecmult_context*) NULL)->pre_g)[0]) * ECMULT_TABLE_SIZE(WINDOW_G)) +#ifdef USE_ENDOMORPHISM + + ROUND_TO_ALIGN(sizeof((*((secp256k1_ecmult_context*) NULL)->pre_g_128)[0]) * ECMULT_TABLE_SIZE(WINDOW_G)) +#endif + ; + static void secp256k1_ecmult_context_init(secp256k1_ecmult_context *ctx) { ctx->pre_g = NULL; #ifdef USE_ENDOMORPHISM @@ -189,8 +325,10 @@ static void secp256k1_ecmult_context_init(secp256k1_ecmult_context *ctx) { #endif } -static void secp256k1_ecmult_context_build(secp256k1_ecmult_context *ctx, const secp256k1_callback *cb) { +static void secp256k1_ecmult_context_build(secp256k1_ecmult_context *ctx, void **prealloc) { secp256k1_gej gj; + void* const base = *prealloc; + size_t const prealloc_size = SECP256K1_ECMULT_CONTEXT_PREALLOCATED_SIZE; if (ctx->pre_g != NULL) { return; @@ -199,44 +337,44 @@ static void secp256k1_ecmult_context_build(secp256k1_ecmult_context *ctx, const /* get the generator */ secp256k1_gej_set_ge(&gj, &secp256k1_ge_const_g); - ctx->pre_g = (secp256k1_ge_storage (*)[])checked_malloc(cb, sizeof((*ctx->pre_g)[0]) * ECMULT_TABLE_SIZE(WINDOW_G)); + { + size_t size = sizeof((*ctx->pre_g)[0]) * ((size_t)ECMULT_TABLE_SIZE(WINDOW_G)); + /* check for overflow */ + VERIFY_CHECK(size / sizeof((*ctx->pre_g)[0]) == ((size_t)ECMULT_TABLE_SIZE(WINDOW_G))); + ctx->pre_g = (secp256k1_ge_storage (*)[])manual_alloc(prealloc, sizeof((*ctx->pre_g)[0]) * ECMULT_TABLE_SIZE(WINDOW_G), base, prealloc_size); + } /* precompute the tables with odd multiples */ - secp256k1_ecmult_odd_multiples_table_storage_var(ECMULT_TABLE_SIZE(WINDOW_G), *ctx->pre_g, &gj, cb); + secp256k1_ecmult_odd_multiples_table_storage_var(ECMULT_TABLE_SIZE(WINDOW_G), *ctx->pre_g, &gj); #ifdef USE_ENDOMORPHISM { secp256k1_gej g_128j; int i; - ctx->pre_g_128 = (secp256k1_ge_storage (*)[])checked_malloc(cb, sizeof((*ctx->pre_g_128)[0]) * ECMULT_TABLE_SIZE(WINDOW_G)); + size_t size = sizeof((*ctx->pre_g_128)[0]) * ((size_t) ECMULT_TABLE_SIZE(WINDOW_G)); + /* check for overflow */ + VERIFY_CHECK(size / sizeof((*ctx->pre_g_128)[0]) == ((size_t)ECMULT_TABLE_SIZE(WINDOW_G))); + ctx->pre_g_128 = (secp256k1_ge_storage (*)[])manual_alloc(prealloc, sizeof((*ctx->pre_g_128)[0]) * ECMULT_TABLE_SIZE(WINDOW_G), base, prealloc_size); /* calculate 2^128*generator */ g_128j = gj; for (i = 0; i < 128; i++) { secp256k1_gej_double_var(&g_128j, &g_128j, NULL); } - secp256k1_ecmult_odd_multiples_table_storage_var(ECMULT_TABLE_SIZE(WINDOW_G), *ctx->pre_g_128, &g_128j, cb); + secp256k1_ecmult_odd_multiples_table_storage_var(ECMULT_TABLE_SIZE(WINDOW_G), *ctx->pre_g_128, &g_128j); } #endif } -static void secp256k1_ecmult_context_clone(secp256k1_ecmult_context *dst, - const secp256k1_ecmult_context *src, const secp256k1_callback *cb) { - if (src->pre_g == NULL) { - dst->pre_g = NULL; - } else { - size_t size = sizeof((*dst->pre_g)[0]) * ECMULT_TABLE_SIZE(WINDOW_G); - dst->pre_g = (secp256k1_ge_storage (*)[])checked_malloc(cb, size); - memcpy(dst->pre_g, src->pre_g, size); +static void secp256k1_ecmult_context_finalize_memcpy(secp256k1_ecmult_context *dst, const secp256k1_ecmult_context *src) { + if (src->pre_g != NULL) { + /* We cast to void* first to suppress a -Wcast-align warning. */ + dst->pre_g = (secp256k1_ge_storage (*)[])(void*)((unsigned char*)dst + ((unsigned char*)(src->pre_g) - (unsigned char*)src)); } #ifdef USE_ENDOMORPHISM - if (src->pre_g_128 == NULL) { - dst->pre_g_128 = NULL; - } else { - size_t size = sizeof((*dst->pre_g_128)[0]) * ECMULT_TABLE_SIZE(WINDOW_G); - dst->pre_g_128 = (secp256k1_ge_storage (*)[])checked_malloc(cb, size); - memcpy(dst->pre_g_128, src->pre_g_128, size); + if (src->pre_g_128 != NULL) { + dst->pre_g_128 = (secp256k1_ge_storage (*)[])(void*)((unsigned char*)dst + ((unsigned char*)(src->pre_g_128) - (unsigned char*)src)); } #endif } @@ -246,10 +384,6 @@ static int secp256k1_ecmult_context_is_built(const secp256k1_ecmult_context *ctx } static void secp256k1_ecmult_context_clear(secp256k1_ecmult_context *ctx) { - free(ctx->pre_g); -#ifdef USE_ENDOMORPHISM - free(ctx->pre_g_128); -#endif secp256k1_ecmult_context_init(ctx); } @@ -306,7 +440,7 @@ static int secp256k1_ecmult_wnaf(int *wnaf, int len, const secp256k1_scalar *a, CHECK(carry == 0); while (bit < 256) { CHECK(secp256k1_scalar_get_bits(&s, bit++, 1) == 0); - } + } #endif return last_set_bit + 1; } @@ -514,52 +648,55 @@ static size_t secp256k1_strauss_scratch_size(size_t n_points) { return n_points*point_size; } -static int secp256k1_ecmult_strauss_batch(const secp256k1_ecmult_context *ctx, 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) { +static int secp256k1_ecmult_strauss_batch(const secp256k1_callback* error_callback, const secp256k1_ecmult_context *ctx, 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; } - if (!secp256k1_scratch_allocate_frame(scratch, secp256k1_strauss_scratch_size(n_points), STRAUSS_SCRATCH_OBJECTS)) { - return 0; - } - points = (secp256k1_gej*)secp256k1_scratch_alloc(scratch, n_points * sizeof(secp256k1_gej)); - scalars = (secp256k1_scalar*)secp256k1_scratch_alloc(scratch, n_points * sizeof(secp256k1_scalar)); - state.prej = (secp256k1_gej*)secp256k1_scratch_alloc(scratch, n_points * ECMULT_TABLE_SIZE(WINDOW_A) * sizeof(secp256k1_gej)); - state.zr = (secp256k1_fe*)secp256k1_scratch_alloc(scratch, n_points * ECMULT_TABLE_SIZE(WINDOW_A) * sizeof(secp256k1_fe)); + 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.prej = (secp256k1_gej*)secp256k1_scratch_alloc(error_callback, scratch, n_points * ECMULT_TABLE_SIZE(WINDOW_A) * sizeof(secp256k1_gej)); + state.zr = (secp256k1_fe*)secp256k1_scratch_alloc(error_callback, scratch, n_points * ECMULT_TABLE_SIZE(WINDOW_A) * sizeof(secp256k1_fe)); #ifdef USE_ENDOMORPHISM - state.pre_a = (secp256k1_ge*)secp256k1_scratch_alloc(scratch, n_points * 2 * ECMULT_TABLE_SIZE(WINDOW_A) * sizeof(secp256k1_ge)); + state.pre_a = (secp256k1_ge*)secp256k1_scratch_alloc(error_callback, scratch, n_points * 2 * ECMULT_TABLE_SIZE(WINDOW_A) * sizeof(secp256k1_ge)); state.pre_a_lam = state.pre_a + n_points * ECMULT_TABLE_SIZE(WINDOW_A); #else - state.pre_a = (secp256k1_ge*)secp256k1_scratch_alloc(scratch, n_points * ECMULT_TABLE_SIZE(WINDOW_A) * sizeof(secp256k1_ge)); + state.pre_a = (secp256k1_ge*)secp256k1_scratch_alloc(error_callback, scratch, n_points * ECMULT_TABLE_SIZE(WINDOW_A) * sizeof(secp256k1_ge)); #endif - state.ps = (struct secp256k1_strauss_point_state*)secp256k1_scratch_alloc(scratch, n_points * sizeof(struct secp256k1_strauss_point_state)); + 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.prej == NULL || state.zr == NULL || state.pre_a == 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_deallocate_frame(scratch); + secp256k1_scratch_apply_checkpoint(error_callback, scratch, scratch_checkpoint); return 0; } secp256k1_gej_set_ge(&points[i], &point); } secp256k1_ecmult_strauss_wnaf(ctx, &state, r, n_points, points, scalars, inp_g_sc); - secp256k1_scratch_deallocate_frame(scratch); + 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_ecmult_context *actx, 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(actx, scratch, r, inp_g_sc, cb, cbdata, n, 0); +static int secp256k1_ecmult_strauss_batch_single(const secp256k1_callback* error_callback, const secp256k1_ecmult_context *actx, 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, actx, scratch, r, inp_g_sc, cb, cbdata, n, 0); } -static size_t secp256k1_strauss_max_points(secp256k1_scratch *scratch) { - return secp256k1_scratch_max_allocation(scratch, STRAUSS_SCRATCH_OBJECTS) / secp256k1_strauss_scratch_size(1); +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. @@ -848,10 +985,11 @@ static size_t secp256k1_pippenger_scratch_size(size_t n_points, int bucket_windo size_t entries = n_points + 1; #endif 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 ((1<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, (1<ps == NULL || state_space->wnaf_na == NULL || buckets == NULL) { + secp256k1_scratch_apply_checkpoint(error_callback, scratch, scratch_checkpoint); return 0; } - points = (secp256k1_ge *) secp256k1_scratch_alloc(scratch, entries * sizeof(*points)); - scalars = (secp256k1_scalar *) secp256k1_scratch_alloc(scratch, entries * sizeof(*scalars)); - state_space = (struct secp256k1_pippenger_state *) secp256k1_scratch_alloc(scratch, sizeof(*state_space)); - state_space->ps = (struct secp256k1_pippenger_point_state *) secp256k1_scratch_alloc(scratch, entries * sizeof(*state_space->ps)); - state_space->wnaf_na = (int *) secp256k1_scratch_alloc(scratch, entries*(WNAF_SIZE(bucket_window+1)) * sizeof(int)); - buckets = (secp256k1_gej *) secp256k1_scratch_alloc(scratch, (1< max_alloc) { break; } @@ -972,12 +1116,58 @@ static size_t secp256k1_pippenger_max_points(secp256k1_scratch *scratch) { return res; } -typedef int (*secp256k1_ecmult_multi_func)(const secp256k1_ecmult_context*, secp256k1_scratch*, secp256k1_gej*, const secp256k1_scalar*, secp256k1_ecmult_multi_callback cb, void*, size_t); -static int secp256k1_ecmult_multi_var(const secp256k1_ecmult_context *ctx, secp256k1_scratch *scratch, secp256k1_gej *r, const secp256k1_scalar *inp_g_sc, secp256k1_ecmult_multi_callback cb, void *cbdata, size_t n) { +/* Computes ecmult_multi by simply multiplying and adding each point. Does not + * require a scratch space */ +static int secp256k1_ecmult_multi_simple_var(const secp256k1_ecmult_context *ctx, 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_scalar szero; + secp256k1_gej tmpj; + + secp256k1_scalar_set_int(&szero, 0); + secp256k1_gej_set_infinity(r); + secp256k1_gej_set_infinity(&tmpj); + /* r = inp_g_sc*G */ + secp256k1_ecmult(ctx, r, &tmpj, &szero, 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(ctx, &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 = 1 + (n - 1) / max_n_batch_points; + *n_batch_points = 1 + (n - 1) / *n_batches; + return 1; +} + +typedef int (*secp256k1_ecmult_multi_func)(const secp256k1_callback* error_callback, const secp256k1_ecmult_context*, 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, const secp256k1_ecmult_context *ctx, 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_ecmult_context*, secp256k1_scratch*, secp256k1_gej*, const secp256k1_scalar*, secp256k1_ecmult_multi_callback cb, void*, size_t, size_t); - size_t max_points; + int (*f)(const secp256k1_callback* error_callback, const secp256k1_ecmult_context*, 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; @@ -990,32 +1180,30 @@ static int secp256k1_ecmult_multi_var(const secp256k1_ecmult_context *ctx, secp2 secp256k1_ecmult(ctx, r, r, &szero, inp_g_sc); return 1; } - - max_points = secp256k1_pippenger_max_points(scratch); - if (max_points == 0) { - return 0; - } else if (max_points > ECMULT_MAX_POINTS_PER_BATCH) { - max_points = ECMULT_MAX_POINTS_PER_BATCH; + if (scratch == NULL) { + return secp256k1_ecmult_multi_simple_var(ctx, r, inp_g_sc, cb, cbdata, n); } - n_batches = (n+max_points-1)/max_points; - n_batch_points = (n+n_batches-1)/n_batches; + /* 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(ctx, r, inp_g_sc, cb, cbdata, n); + } if (n_batch_points >= ECMULT_PIPPENGER_THRESHOLD) { f = secp256k1_ecmult_pippenger_batch; } else { - max_points = secp256k1_strauss_max_points(scratch); - if (max_points == 0) { - return 0; + 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(ctx, r, inp_g_sc, cb, cbdata, n); } - n_batches = (n+max_points-1)/max_points; - n_batch_points = (n+n_batches-1)/n_batches; 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(ctx, scratch, &tmp, i == 0 ? inp_g_sc : NULL, cb, cbdata, nbp, offset)) { + if (!f(error_callback, ctx, scratch, &tmp, i == 0 ? inp_g_sc : NULL, cb, cbdata, nbp, offset)) { return 0; } secp256k1_gej_add_var(r, r, &tmp, NULL); diff --git a/depend/secp256k1/src/field_10x26.h b/depend/secp256k1/src/field_10x26.h index 727c526..5ff03c8 100644 --- a/depend/secp256k1/src/field_10x26.h +++ b/depend/secp256k1/src/field_10x26.h @@ -10,7 +10,9 @@ #include typedef struct { - /* X = sum(i=0..9, elem[i]*2^26) mod n */ + /* X = sum(i=0..9, n[i]*2^(i*26)) mod p + * where p = 2^256 - 0x1000003D1 + */ uint32_t n[10]; #ifdef VERIFY int magnitude; diff --git a/depend/secp256k1/src/field_10x26_impl.h b/depend/secp256k1/src/field_10x26_impl.h index 94f8132..4ae4fdc 100644 --- a/depend/secp256k1/src/field_10x26_impl.h +++ b/depend/secp256k1/src/field_10x26_impl.h @@ -8,7 +8,6 @@ #define SECP256K1_FIELD_REPR_IMPL_H #include "util.h" -#include "num.h" #include "field.h" #ifdef VERIFY @@ -486,7 +485,8 @@ SECP256K1_INLINE static void secp256k1_fe_mul_inner(uint32_t *r, const uint32_t VERIFY_BITS(b[9], 26); /** [... a b c] is a shorthand for ... + a<<52 + b<<26 + c<<0 mod n. - * px is a shorthand for sum(a[i]*b[x-i], i=0..x). + * for 0 <= x <= 9, px is a shorthand for sum(a[i]*b[x-i], i=0..x). + * for 9 <= x <= 18, px is a shorthand for sum(a[i]*b[x-i], i=(x-9)..9) * Note that [x 0 0 0 0 0 0 0 0 0 0] = [x*R1 x*R0]. */ @@ -1069,6 +1069,7 @@ static void secp256k1_fe_mul(secp256k1_fe *r, const secp256k1_fe *a, const secp2 secp256k1_fe_verify(a); secp256k1_fe_verify(b); VERIFY_CHECK(r != b); + VERIFY_CHECK(a != b); #endif secp256k1_fe_mul_inner(r->n, a->n, b->n); #ifdef VERIFY diff --git a/depend/secp256k1/src/field_5x52.h b/depend/secp256k1/src/field_5x52.h index bccd8fe..fc5bfe3 100644 --- a/depend/secp256k1/src/field_5x52.h +++ b/depend/secp256k1/src/field_5x52.h @@ -10,7 +10,9 @@ #include typedef struct { - /* X = sum(i=0..4, elem[i]*2^52) mod n */ + /* X = sum(i=0..4, n[i]*2^(i*52)) mod p + * where p = 2^256 - 0x1000003D1 + */ uint64_t n[5]; #ifdef VERIFY int magnitude; diff --git a/depend/secp256k1/src/field_5x52_impl.h b/depend/secp256k1/src/field_5x52_impl.h index 957c61b..f426332 100644 --- a/depend/secp256k1/src/field_5x52_impl.h +++ b/depend/secp256k1/src/field_5x52_impl.h @@ -12,7 +12,6 @@ #endif #include "util.h" -#include "num.h" #include "field.h" #if defined(USE_ASM_X86_64) @@ -422,6 +421,7 @@ static void secp256k1_fe_mul(secp256k1_fe *r, const secp256k1_fe *a, const secp2 secp256k1_fe_verify(a); secp256k1_fe_verify(b); VERIFY_CHECK(r != b); + VERIFY_CHECK(a != b); #endif secp256k1_fe_mul_inner(r->n, a->n, b->n); #ifdef VERIFY diff --git a/depend/secp256k1/src/field_5x52_int128_impl.h b/depend/secp256k1/src/field_5x52_int128_impl.h index 95a0d17..bcbfb92 100644 --- a/depend/secp256k1/src/field_5x52_int128_impl.h +++ b/depend/secp256k1/src/field_5x52_int128_impl.h @@ -32,9 +32,11 @@ SECP256K1_INLINE static void secp256k1_fe_mul_inner(uint64_t *r, const uint64_t 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. - * px is a shorthand for sum(a[i]*b[x-i], i=0..x). + * 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]. */ diff --git a/depend/secp256k1/src/field_impl.h b/depend/secp256k1/src/field_impl.h index 2042864..6070cac 100644 --- a/depend/secp256k1/src/field_impl.h +++ b/depend/secp256k1/src/field_impl.h @@ -12,6 +12,7 @@ #endif #include "util.h" +#include "num.h" #if defined(USE_FIELD_10X26) #include "field_10x26_impl.h" @@ -48,6 +49,8 @@ static int secp256k1_fe_sqrt(secp256k1_fe *r, const secp256k1_fe *a) { secp256k1_fe x2, x3, x6, x9, x11, x22, x44, x88, x176, x220, x223, t1; int j; + VERIFY_CHECK(r != a); + /** 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] diff --git a/depend/secp256k1/src/gen_context.c b/depend/secp256k1/src/gen_context.c index 87d296e..82c605c 100644 --- a/depend/secp256k1/src/gen_context.c +++ b/depend/secp256k1/src/gen_context.c @@ -8,6 +8,7 @@ #include "basic-config.h" #include "include/secp256k1.h" +#include "util.h" #include "field_impl.h" #include "scalar_impl.h" #include "group_impl.h" @@ -26,6 +27,7 @@ static const secp256k1_callback default_error_callback = { int main(int argc, char **argv) { secp256k1_ecmult_gen_context ctx; + void *prealloc, *base; int inner; int outer; FILE* fp; @@ -38,15 +40,17 @@ int main(int argc, char **argv) { fprintf(stderr, "Could not open src/ecmult_static_context.h for writing!\n"); return -1; } - + fprintf(fp, "#ifndef _SECP256K1_ECMULT_STATIC_CONTEXT_\n"); fprintf(fp, "#define _SECP256K1_ECMULT_STATIC_CONTEXT_\n"); fprintf(fp, "#include \"src/group.h\"\n"); fprintf(fp, "#define SC SECP256K1_GE_STORAGE_CONST\n"); fprintf(fp, "static const secp256k1_ge_storage secp256k1_ecmult_static_context[64][16] = {\n"); + base = checked_malloc(&default_error_callback, SECP256K1_ECMULT_GEN_CONTEXT_PREALLOCATED_SIZE); + prealloc = base; secp256k1_ecmult_gen_context_init(&ctx); - secp256k1_ecmult_gen_context_build(&ctx, &default_error_callback); + secp256k1_ecmult_gen_context_build(&ctx, &prealloc); for(outer = 0; outer != 64; outer++) { fprintf(fp,"{\n"); for(inner = 0; inner != 16; inner++) { @@ -65,10 +69,11 @@ int main(int argc, char **argv) { } fprintf(fp,"};\n"); secp256k1_ecmult_gen_context_clear(&ctx); - + free(base); + fprintf(fp, "#undef SC\n"); fprintf(fp, "#endif\n"); fclose(fp); - + return 0; } diff --git a/depend/secp256k1/src/group.h b/depend/secp256k1/src/group.h index 3947ea2..8e122ab 100644 --- a/depend/secp256k1/src/group.h +++ b/depend/secp256k1/src/group.h @@ -65,12 +65,7 @@ static void secp256k1_ge_neg(secp256k1_ge *r, const secp256k1_ge *a); static void secp256k1_ge_set_gej(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, const secp256k1_callback *cb); - -/** Set a batch of group elements equal to the inputs given in jacobian - * coordinates (with known z-ratios). zr must contain the known z-ratios such - * that mul(a[i].z, zr[i+1]) == a[i+1].z. zr[0] is ignored. */ -static void secp256k1_ge_set_table_gej_var(secp256k1_ge *r, const secp256k1_gej *a, const secp256k1_fe *zr, size_t len); +static void secp256k1_ge_set_all_gej_var(secp256k1_ge *r, const secp256k1_gej *a, size_t len); /** Bring a batch inputs given in jacobian coordinates (with known z-ratios) to * the same global z "denominator". zr must contain the known z-ratios such diff --git a/depend/secp256k1/src/group_impl.h b/depend/secp256k1/src/group_impl.h index b1ace87..9b93c39 100644 --- a/depend/secp256k1/src/group_impl.h +++ b/depend/secp256k1/src/group_impl.h @@ -38,22 +38,22 @@ */ #if defined(EXHAUSTIVE_TEST_ORDER) # if EXHAUSTIVE_TEST_ORDER == 199 -const secp256k1_ge secp256k1_ge_const_g = SECP256K1_GE_CONST( +static const secp256k1_ge secp256k1_ge_const_g = SECP256K1_GE_CONST( 0xFA7CC9A7, 0x0737F2DB, 0xA749DD39, 0x2B4FB069, 0x3B017A7D, 0xA808C2F1, 0xFB12940C, 0x9EA66C18, 0x78AC123A, 0x5ED8AEF3, 0x8732BC91, 0x1F3A2868, 0x48DF246C, 0x808DAE72, 0xCFE52572, 0x7F0501ED ); -const int CURVE_B = 4; +static const int CURVE_B = 4; # elif EXHAUSTIVE_TEST_ORDER == 13 -const secp256k1_ge secp256k1_ge_const_g = SECP256K1_GE_CONST( +static const secp256k1_ge secp256k1_ge_const_g = SECP256K1_GE_CONST( 0xedc60018, 0xa51a786b, 0x2ea91f4d, 0x4c9416c0, 0x9de54c3b, 0xa1316554, 0x6cf4345c, 0x7277ef15, 0x54cb1b6b, 0xdc8c1273, 0x087844ea, 0x43f4603e, 0x0eaf9a43, 0xf6effe55, 0x939f806d, 0x37adf8ac ); -const int CURVE_B = 2; +static const int CURVE_B = 2; # else # error No known generator for the specified exhaustive test group order. # endif @@ -68,7 +68,7 @@ static const secp256k1_ge secp256k1_ge_const_g = SECP256K1_GE_CONST( 0xFD17B448UL, 0xA6855419UL, 0x9C47D08FUL, 0xFB10D4B8UL ); -const int CURVE_B = 7; +static const int CURVE_B = 7; #endif static void secp256k1_ge_set_gej_zinv(secp256k1_ge *r, const secp256k1_gej *a, const secp256k1_fe *zi) { @@ -126,46 +126,43 @@ static void secp256k1_ge_set_gej_var(secp256k1_ge *r, secp256k1_gej *a) { r->y = a->y; } -static void secp256k1_ge_set_all_gej_var(secp256k1_ge *r, const secp256k1_gej *a, size_t len, const secp256k1_callback *cb) { - secp256k1_fe *az; - secp256k1_fe *azi; +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 count = 0; - az = (secp256k1_fe *)checked_malloc(cb, sizeof(secp256k1_fe) * len); + size_t last_i = SIZE_MAX; + for (i = 0; i < len; i++) { if (!a[i].infinity) { - az[count++] = a[i].z; + /* 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); - azi = (secp256k1_fe *)checked_malloc(cb, sizeof(secp256k1_fe) * count); - secp256k1_fe_inv_all_var(azi, az, count); - free(az); + 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; - count = 0; for (i = 0; i < len; i++) { r[i].infinity = a[i].infinity; if (!a[i].infinity) { - secp256k1_ge_set_gej_zinv(&r[i], &a[i], &azi[count++]); - } - } - free(azi); -} - -static void secp256k1_ge_set_table_gej_var(secp256k1_ge *r, const secp256k1_gej *a, const secp256k1_fe *zr, size_t len) { - size_t i = len - 1; - secp256k1_fe zi; - - if (len > 0) { - /* Compute the inverse of the last z coordinate, and use it to compute the last affine output. */ - secp256k1_fe_inv(&zi, &a[i].z); - secp256k1_ge_set_gej_zinv(&r[i], &a[i], &zi); - - /* Work out way backwards, using the z-ratios to scale the x/y values. */ - while (i > 0) { - secp256k1_fe_mul(&zi, &zi, &zr[i]); - i--; - secp256k1_ge_set_gej_zinv(&r[i], &a[i], &zi); + secp256k1_ge_set_gej_zinv(&r[i], &a[i], &r[i].x); } } } @@ -178,6 +175,8 @@ static void secp256k1_ge_globalz_set_table_gej(size_t len, secp256k1_ge *r, secp /* The z of the final point gives us the "global Z" for the table. */ r[i].x = a[i].x; r[i].y = a[i].y; + /* Ensure all y values are in weak normal form for fast negation of points */ + secp256k1_fe_normalize_weak(&r[i].y); *globalz = a[i].z; r[i].infinity = 0; zs = zr[i]; diff --git a/depend/secp256k1/src/scalar_4x64_impl.h b/depend/secp256k1/src/scalar_4x64_impl.h index db1ebf9..d378335 100644 --- a/depend/secp256k1/src/scalar_4x64_impl.h +++ b/depend/secp256k1/src/scalar_4x64_impl.h @@ -376,7 +376,7 @@ static void secp256k1_scalar_reduce_512(secp256k1_scalar *r, const uint64_t *l) /* extract m6 */ "movq %%r8, %q6\n" : "=g"(m0), "=g"(m1), "=g"(m2), "=g"(m3), "=g"(m4), "=g"(m5), "=g"(m6) - : "S"(l), "n"(SECP256K1_N_C_0), "n"(SECP256K1_N_C_1) + : "S"(l), "i"(SECP256K1_N_C_0), "i"(SECP256K1_N_C_1) : "rax", "rdx", "r8", "r9", "r10", "r11", "r12", "r13", "r14", "cc"); /* Reduce 385 bits into 258. */ @@ -455,7 +455,7 @@ static void secp256k1_scalar_reduce_512(secp256k1_scalar *r, const uint64_t *l) /* extract p4 */ "movq %%r9, %q4\n" : "=&g"(p0), "=&g"(p1), "=&g"(p2), "=g"(p3), "=g"(p4) - : "g"(m0), "g"(m1), "g"(m2), "g"(m3), "g"(m4), "g"(m5), "g"(m6), "n"(SECP256K1_N_C_0), "n"(SECP256K1_N_C_1) + : "g"(m0), "g"(m1), "g"(m2), "g"(m3), "g"(m4), "g"(m5), "g"(m6), "i"(SECP256K1_N_C_0), "i"(SECP256K1_N_C_1) : "rax", "rdx", "r8", "r9", "r10", "r11", "r12", "r13", "cc"); /* Reduce 258 bits into 256. */ @@ -501,7 +501,7 @@ static void secp256k1_scalar_reduce_512(secp256k1_scalar *r, const uint64_t *l) /* Extract c */ "movq %%r9, %q0\n" : "=g"(c) - : "g"(p0), "g"(p1), "g"(p2), "g"(p3), "g"(p4), "D"(r), "n"(SECP256K1_N_C_0), "n"(SECP256K1_N_C_1) + : "g"(p0), "g"(p1), "g"(p2), "g"(p3), "g"(p4), "D"(r), "i"(SECP256K1_N_C_0), "i"(SECP256K1_N_C_1) : "rax", "rdx", "r8", "r9", "r10", "cc", "memory"); #else uint128_t c; diff --git a/depend/secp256k1/src/scratch.h b/depend/secp256k1/src/scratch.h index fef377a..77b35d1 100644 --- a/depend/secp256k1/src/scratch.h +++ b/depend/secp256k1/src/scratch.h @@ -7,33 +7,36 @@ #ifndef _SECP256K1_SCRATCH_ #define _SECP256K1_SCRATCH_ -#define SECP256K1_SCRATCH_MAX_FRAMES 5 - /* 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 { - void *data[SECP256K1_SCRATCH_MAX_FRAMES]; - size_t offset[SECP256K1_SCRATCH_MAX_FRAMES]; - size_t frame_size[SECP256K1_SCRATCH_MAX_FRAMES]; - size_t frame; + /** 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; - const secp256k1_callback* error_callback; } secp256k1_scratch; static secp256k1_scratch* secp256k1_scratch_create(const secp256k1_callback* error_callback, size_t max_size); -static void secp256k1_scratch_destroy(secp256k1_scratch* scratch); +static void secp256k1_scratch_destroy(const secp256k1_callback* error_callback, secp256k1_scratch* scratch); -/** Attempts to allocate a new stack frame with `n` available bytes. Returns 1 on success, 0 on failure */ -static int secp256k1_scratch_allocate_frame(secp256k1_scratch* scratch, size_t n, size_t objects); +/** 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); -/** Deallocates a stack frame */ -static void secp256k1_scratch_deallocate_frame(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_scratch* scratch, size_t n_objects); +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(secp256k1_scratch* scratch, size_t n); +static void *secp256k1_scratch_alloc(const secp256k1_callback* error_callback, secp256k1_scratch* scratch, size_t n); #endif diff --git a/depend/secp256k1/src/scratch_impl.h b/depend/secp256k1/src/scratch_impl.h index abed713..4cee700 100644 --- a/depend/secp256k1/src/scratch_impl.h +++ b/depend/secp256k1/src/scratch_impl.h @@ -7,78 +7,80 @@ #ifndef _SECP256K1_SCRATCH_IMPL_H_ #define _SECP256K1_SCRATCH_IMPL_H_ +#include "util.h" #include "scratch.h" -/* 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. - * TODO: Determine this at configure time. */ -#define ALIGNMENT 16 - -static secp256k1_scratch* secp256k1_scratch_create(const secp256k1_callback* error_callback, size_t max_size) { - secp256k1_scratch* ret = (secp256k1_scratch*)checked_malloc(error_callback, sizeof(*ret)); +static secp256k1_scratch* secp256k1_scratch_create(const secp256k1_callback* error_callback, size_t size) { + const size_t base_alloc = ((sizeof(secp256k1_scratch) + ALIGNMENT - 1) / ALIGNMENT) * ALIGNMENT; + void *alloc = checked_malloc(error_callback, base_alloc + size); + secp256k1_scratch* ret = (secp256k1_scratch *)alloc; if (ret != NULL) { memset(ret, 0, sizeof(*ret)); - ret->max_size = max_size; - ret->error_callback = error_callback; + memcpy(ret->magic, "scratch", 8); + ret->data = (void *) ((char *) alloc + base_alloc); + ret->max_size = size; } return ret; } -static void secp256k1_scratch_destroy(secp256k1_scratch* scratch) { +static void secp256k1_scratch_destroy(const secp256k1_callback* error_callback, secp256k1_scratch* scratch) { if (scratch != NULL) { - VERIFY_CHECK(scratch->frame == 0); + VERIFY_CHECK(scratch->alloc_size == 0); /* all checkpoints should be applied */ + if (memcmp(scratch->magic, "scratch", 8) != 0) { + secp256k1_callback_call(error_callback, "invalid scratch space"); + return; + } + memset(scratch->magic, 0, sizeof(scratch->magic)); free(scratch); } } -static size_t secp256k1_scratch_max_allocation(const secp256k1_scratch* scratch, size_t objects) { - size_t i = 0; - size_t allocated = 0; - for (i = 0; i < scratch->frame; i++) { - allocated += scratch->frame_size[i]; - } - if (scratch->max_size - allocated <= objects * ALIGNMENT) { +static size_t secp256k1_scratch_checkpoint(const secp256k1_callback* error_callback, const secp256k1_scratch* scratch) { + if (memcmp(scratch->magic, "scratch", 8) != 0) { + secp256k1_callback_call(error_callback, "invalid scratch space"); return 0; } - return scratch->max_size - allocated - objects * ALIGNMENT; + return scratch->alloc_size; } -static int secp256k1_scratch_allocate_frame(secp256k1_scratch* scratch, size_t n, size_t objects) { - VERIFY_CHECK(scratch->frame < SECP256K1_SCRATCH_MAX_FRAMES); +static void secp256k1_scratch_apply_checkpoint(const secp256k1_callback* error_callback, secp256k1_scratch* scratch, size_t checkpoint) { + if (memcmp(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; +} - if (n <= secp256k1_scratch_max_allocation(scratch, objects)) { - n += objects * ALIGNMENT; - scratch->data[scratch->frame] = checked_malloc(scratch->error_callback, n); - if (scratch->data[scratch->frame] == NULL) { - return 0; - } - scratch->frame_size[scratch->frame] = n; - scratch->offset[scratch->frame] = 0; - scratch->frame++; - return 1; - } else { +static size_t secp256k1_scratch_max_allocation(const secp256k1_callback* error_callback, const secp256k1_scratch* scratch, size_t objects) { + if (memcmp(scratch->magic, "scratch", 8) != 0) { + secp256k1_callback_call(error_callback, "invalid scratch space"); 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_deallocate_frame(secp256k1_scratch* scratch) { - VERIFY_CHECK(scratch->frame > 0); - scratch->frame -= 1; - free(scratch->data[scratch->frame]); -} - -static void *secp256k1_scratch_alloc(secp256k1_scratch* scratch, size_t size) { +static void *secp256k1_scratch_alloc(const secp256k1_callback* error_callback, secp256k1_scratch* scratch, size_t size) { void *ret; - size_t frame = scratch->frame - 1; - size = ((size + ALIGNMENT - 1) / ALIGNMENT) * ALIGNMENT; + size = ROUND_TO_ALIGN(size); - if (scratch->frame == 0 || size + scratch->offset[frame] > scratch->frame_size[frame]) { + if (memcmp(scratch->magic, "scratch", 8) != 0) { + secp256k1_callback_call(error_callback, "invalid scratch space"); return NULL; } - ret = (void *) ((unsigned char *) scratch->data[frame] + scratch->offset[frame]); + + if (size > scratch->max_size - scratch->alloc_size) { + return NULL; + } + ret = (void *) ((char *) scratch->data + scratch->alloc_size); memset(ret, 0, size); - scratch->offset[frame] += size; + scratch->alloc_size += size; return ret; } diff --git a/depend/secp256k1/src/secp256k1.c b/depend/secp256k1/src/secp256k1.c index a1e3908..6954e1b 100644 --- a/depend/secp256k1/src/secp256k1.c +++ b/depend/secp256k1/src/secp256k1.c @@ -5,6 +5,7 @@ **********************************************************************/ #include "include/secp256k1.h" +#include "include/secp256k1_preallocated.h" #include "util.h" #include "num_impl.h" @@ -26,28 +27,39 @@ } \ } while(0) -static void default_illegal_callback_fn(const char* str, void* data) { +#define ARG_CHECK_NO_RETURN(cond) do { \ + if (EXPECT(!(cond), 0)) { \ + secp256k1_callback_call(&ctx->illegal_callback, #cond); \ + } \ +} while(0) + +#ifndef USE_EXTERNAL_DEFAULT_CALLBACKS +#include +#include +static void secp256k1_default_illegal_callback_fn(const char* str, void* data) { (void)data; fprintf(stderr, "[libsecp256k1] illegal argument: %s\n", str); abort(); } - -static const secp256k1_callback default_illegal_callback = { - default_illegal_callback_fn, - NULL -}; - -static void default_error_callback_fn(const char* str, void* data) { +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_error_callback = { - default_error_callback_fn, +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 +}; struct secp256k1_context_struct { secp256k1_ecmult_context ecmult_ctx; @@ -59,20 +71,55 @@ struct secp256k1_context_struct { static const secp256k1_context secp256k1_context_no_precomp_ = { { 0 }, { 0 }, - { default_illegal_callback_fn, 0 }, - { default_error_callback_fn, 0 } + { secp256k1_default_illegal_callback_fn, 0 }, + { secp256k1_default_error_callback_fn, 0 } }; const secp256k1_context *secp256k1_context_no_precomp = &secp256k1_context_no_precomp_; -secp256k1_context* secp256k1_context_create(unsigned int flags) { - secp256k1_context* ret = (secp256k1_context*)checked_malloc(&default_error_callback, sizeof(secp256k1_context)); +size_t secp256k1_context_preallocated_size(unsigned int flags) { + size_t ret = ROUND_TO_ALIGN(sizeof(secp256k1_context)); + + if (EXPECT((flags & SECP256K1_FLAGS_TYPE_MASK) != SECP256K1_FLAGS_TYPE_CONTEXT, 0)) { + secp256k1_callback_call(&default_illegal_callback, + "Invalid flags"); + return 0; + } + + if (flags & SECP256K1_FLAGS_BIT_CONTEXT_SIGN) { + ret += SECP256K1_ECMULT_GEN_CONTEXT_PREALLOCATED_SIZE; + } + if (flags & SECP256K1_FLAGS_BIT_CONTEXT_VERIFY) { + ret += SECP256K1_ECMULT_CONTEXT_PREALLOCATED_SIZE; + } + return ret; +} + +size_t secp256k1_context_preallocated_clone_size(const secp256k1_context* ctx) { + size_t ret = ROUND_TO_ALIGN(sizeof(secp256k1_context)); + VERIFY_CHECK(ctx != NULL); + if (secp256k1_ecmult_gen_context_is_built(&ctx->ecmult_gen_ctx)) { + ret += SECP256K1_ECMULT_GEN_CONTEXT_PREALLOCATED_SIZE; + } + if (secp256k1_ecmult_context_is_built(&ctx->ecmult_ctx)) { + ret += SECP256K1_ECMULT_CONTEXT_PREALLOCATED_SIZE; + } + return ret; +} + +secp256k1_context* secp256k1_context_preallocated_create(void* prealloc, unsigned int flags) { + void* const base = prealloc; + size_t prealloc_size; + secp256k1_context* ret; + + VERIFY_CHECK(prealloc != NULL); + prealloc_size = secp256k1_context_preallocated_size(flags); + ret = (secp256k1_context*)manual_alloc(&prealloc, sizeof(secp256k1_context), base, prealloc_size); ret->illegal_callback = default_illegal_callback; ret->error_callback = default_error_callback; if (EXPECT((flags & SECP256K1_FLAGS_TYPE_MASK) != SECP256K1_FLAGS_TYPE_CONTEXT, 0)) { secp256k1_callback_call(&ret->illegal_callback, "Invalid flags"); - free(ret); return NULL; } @@ -80,47 +127,79 @@ secp256k1_context* secp256k1_context_create(unsigned int flags) { secp256k1_ecmult_gen_context_init(&ret->ecmult_gen_ctx); if (flags & SECP256K1_FLAGS_BIT_CONTEXT_SIGN) { - secp256k1_ecmult_gen_context_build(&ret->ecmult_gen_ctx, &ret->error_callback); + secp256k1_ecmult_gen_context_build(&ret->ecmult_gen_ctx, &prealloc); } if (flags & SECP256K1_FLAGS_BIT_CONTEXT_VERIFY) { - secp256k1_ecmult_context_build(&ret->ecmult_ctx, &ret->error_callback); + secp256k1_ecmult_context_build(&ret->ecmult_ctx, &prealloc); } + return (secp256k1_context*) 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) { + size_t prealloc_size; + secp256k1_context* ret; + VERIFY_CHECK(ctx != NULL); + ARG_CHECK(prealloc != NULL); + + prealloc_size = secp256k1_context_preallocated_clone_size(ctx); + ret = (secp256k1_context*)prealloc; + memcpy(ret, ctx, prealloc_size); + secp256k1_ecmult_gen_context_finalize_memcpy(&ret->ecmult_gen_ctx, &ctx->ecmult_gen_ctx); + secp256k1_ecmult_context_finalize_memcpy(&ret->ecmult_ctx, &ctx->ecmult_ctx); return ret; } secp256k1_context* secp256k1_context_clone(const secp256k1_context* ctx) { - secp256k1_context* ret = (secp256k1_context*)checked_malloc(&ctx->error_callback, sizeof(secp256k1_context)); - ret->illegal_callback = ctx->illegal_callback; - ret->error_callback = ctx->error_callback; - secp256k1_ecmult_context_clone(&ret->ecmult_ctx, &ctx->ecmult_ctx, &ctx->error_callback); - secp256k1_ecmult_gen_context_clone(&ret->ecmult_gen_ctx, &ctx->ecmult_gen_ctx, &ctx->error_callback); + secp256k1_context* ret; + size_t prealloc_size; + + VERIFY_CHECK(ctx != NULL); + 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_destroy(secp256k1_context* ctx) { - CHECK(ctx != secp256k1_context_no_precomp); +void secp256k1_context_preallocated_destroy(secp256k1_context* ctx) { + ARG_CHECK_NO_RETURN(ctx != secp256k1_context_no_precomp); if (ctx != NULL) { secp256k1_ecmult_context_clear(&ctx->ecmult_ctx); secp256k1_ecmult_gen_context_clear(&ctx->ecmult_gen_ctx); + } +} +void secp256k1_context_destroy(secp256k1_context* ctx) { + if (ctx != NULL) { + 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) { - CHECK(ctx != secp256k1_context_no_precomp); + ARG_CHECK_NO_RETURN(ctx != secp256k1_context_no_precomp); if (fun == NULL) { - fun = default_illegal_callback_fn; + 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) { - CHECK(ctx != secp256k1_context_no_precomp); + ARG_CHECK_NO_RETURN(ctx != secp256k1_context_no_precomp); if (fun == NULL) { - fun = default_error_callback_fn; + fun = secp256k1_default_error_callback_fn; } ctx->error_callback.fn = fun; ctx->error_callback.data = data; @@ -131,8 +210,9 @@ secp256k1_scratch_space* secp256k1_scratch_space_create(const secp256k1_context* return secp256k1_scratch_create(&ctx->error_callback, max_size); } -void secp256k1_scratch_space_destroy(secp256k1_scratch_space* scratch) { - secp256k1_scratch_destroy(scratch); +void secp256k1_scratch_space_destroy(const secp256k1_context *ctx, secp256k1_scratch_space* scratch) { + VERIFY_CHECK(ctx != NULL); + secp256k1_scratch_destroy(&ctx->error_callback, scratch); } static int secp256k1_pubkey_load(const secp256k1_context* ctx, secp256k1_ge* ge, const secp256k1_pubkey* pubkey) { @@ -457,6 +537,7 @@ int secp256k1_ec_privkey_negate(const secp256k1_context* ctx, unsigned char *sec secp256k1_scalar_negate(&sec, &sec); secp256k1_scalar_get_b32(seckey, &sec); + secp256k1_scalar_clear(&sec); return 1; } @@ -570,9 +651,9 @@ int secp256k1_ec_pubkey_tweak_mul(const secp256k1_context* ctx, secp256k1_pubkey int secp256k1_context_randomize(secp256k1_context* ctx, const unsigned char *seed32) { VERIFY_CHECK(ctx != NULL); - CHECK(ctx != secp256k1_context_no_precomp); - ARG_CHECK(secp256k1_ecmult_gen_context_is_built(&ctx->ecmult_gen_ctx)); - secp256k1_ecmult_gen_blind(&ctx->ecmult_gen_ctx, seed32); + if (secp256k1_ecmult_gen_context_is_built(&ctx->ecmult_gen_ctx)) { + secp256k1_ecmult_gen_blind(&ctx->ecmult_gen_ctx, seed32); + } return 1; } diff --git a/depend/secp256k1/src/tests.c b/depend/secp256k1/src/tests.c index c72a742..1d813f1 100644 --- a/depend/secp256k1/src/tests.c +++ b/depend/secp256k1/src/tests.c @@ -16,6 +16,7 @@ #include "secp256k1.c" #include "include/secp256k1.h" +#include "include/secp256k1_preallocated.h" #include "testrand_impl.h" #ifdef ENABLE_OPENSSL_TESTS @@ -137,23 +138,47 @@ void random_scalar_order(secp256k1_scalar *num) { } while(1); } -void run_context_tests(void) { +void run_context_tests(int use_prealloc) { secp256k1_pubkey pubkey; secp256k1_pubkey zero_pubkey; secp256k1_ecdsa_signature sig; unsigned char ctmp[32]; int32_t ecount; int32_t ecount2; - secp256k1_context *none = secp256k1_context_create(SECP256K1_CONTEXT_NONE); - secp256k1_context *sign = secp256k1_context_create(SECP256K1_CONTEXT_SIGN); - secp256k1_context *vrfy = secp256k1_context_create(SECP256K1_CONTEXT_VERIFY); - secp256k1_context *both = secp256k1_context_create(SECP256K1_CONTEXT_SIGN | SECP256K1_CONTEXT_VERIFY); + secp256k1_context *none; + secp256k1_context *sign; + secp256k1_context *vrfy; + secp256k1_context *both; + void *none_prealloc = NULL; + void *sign_prealloc = NULL; + void *vrfy_prealloc = NULL; + void *both_prealloc = NULL; secp256k1_gej pubj; secp256k1_ge pub; secp256k1_scalar msg, key, nonce; secp256k1_scalar sigr, sigs; + if (use_prealloc) { + none_prealloc = malloc(secp256k1_context_preallocated_size(SECP256K1_CONTEXT_NONE)); + sign_prealloc = malloc(secp256k1_context_preallocated_size(SECP256K1_CONTEXT_SIGN)); + vrfy_prealloc = malloc(secp256k1_context_preallocated_size(SECP256K1_CONTEXT_VERIFY)); + both_prealloc = malloc(secp256k1_context_preallocated_size(SECP256K1_CONTEXT_SIGN | SECP256K1_CONTEXT_VERIFY)); + CHECK(none_prealloc != NULL); + CHECK(sign_prealloc != NULL); + CHECK(vrfy_prealloc != NULL); + CHECK(both_prealloc != NULL); + none = secp256k1_context_preallocated_create(none_prealloc, SECP256K1_CONTEXT_NONE); + sign = secp256k1_context_preallocated_create(sign_prealloc, SECP256K1_CONTEXT_SIGN); + vrfy = secp256k1_context_preallocated_create(vrfy_prealloc, SECP256K1_CONTEXT_VERIFY); + both = secp256k1_context_preallocated_create(both_prealloc, SECP256K1_CONTEXT_SIGN | SECP256K1_CONTEXT_VERIFY); + } else { + none = secp256k1_context_create(SECP256K1_CONTEXT_NONE); + sign = secp256k1_context_create(SECP256K1_CONTEXT_SIGN); + vrfy = secp256k1_context_create(SECP256K1_CONTEXT_VERIFY); + both = secp256k1_context_create(SECP256K1_CONTEXT_SIGN | SECP256K1_CONTEXT_VERIFY); + } + memset(&zero_pubkey, 0, sizeof(zero_pubkey)); ecount = 0; @@ -163,14 +188,57 @@ void run_context_tests(void) { secp256k1_context_set_error_callback(sign, counting_illegal_callback_fn, NULL); CHECK(vrfy->error_callback.fn != sign->error_callback.fn); + /* check if sizes for cloning are consistent */ + CHECK(secp256k1_context_preallocated_clone_size(none) == secp256k1_context_preallocated_size(SECP256K1_CONTEXT_NONE)); + CHECK(secp256k1_context_preallocated_clone_size(sign) == secp256k1_context_preallocated_size(SECP256K1_CONTEXT_SIGN)); + CHECK(secp256k1_context_preallocated_clone_size(vrfy) == secp256k1_context_preallocated_size(SECP256K1_CONTEXT_VERIFY)); + CHECK(secp256k1_context_preallocated_clone_size(both) == secp256k1_context_preallocated_size(SECP256K1_CONTEXT_SIGN | SECP256K1_CONTEXT_VERIFY)); + /*** clone and destroy all of them to make sure cloning was complete ***/ { secp256k1_context *ctx_tmp; - ctx_tmp = none; none = secp256k1_context_clone(none); secp256k1_context_destroy(ctx_tmp); - ctx_tmp = sign; sign = secp256k1_context_clone(sign); secp256k1_context_destroy(ctx_tmp); - ctx_tmp = vrfy; vrfy = secp256k1_context_clone(vrfy); secp256k1_context_destroy(ctx_tmp); - ctx_tmp = both; both = secp256k1_context_clone(both); secp256k1_context_destroy(ctx_tmp); + if (use_prealloc) { + /* clone into a non-preallocated context and then again into a new preallocated one. */ + ctx_tmp = none; none = secp256k1_context_clone(none); secp256k1_context_preallocated_destroy(ctx_tmp); + free(none_prealloc); none_prealloc = malloc(secp256k1_context_preallocated_size(SECP256K1_CONTEXT_NONE)); CHECK(none_prealloc != NULL); + ctx_tmp = none; none = secp256k1_context_preallocated_clone(none, none_prealloc); secp256k1_context_destroy(ctx_tmp); + + ctx_tmp = sign; sign = secp256k1_context_clone(sign); secp256k1_context_preallocated_destroy(ctx_tmp); + free(sign_prealloc); sign_prealloc = malloc(secp256k1_context_preallocated_size(SECP256K1_CONTEXT_SIGN)); CHECK(sign_prealloc != NULL); + ctx_tmp = sign; sign = secp256k1_context_preallocated_clone(sign, sign_prealloc); secp256k1_context_destroy(ctx_tmp); + + ctx_tmp = vrfy; vrfy = secp256k1_context_clone(vrfy); secp256k1_context_preallocated_destroy(ctx_tmp); + free(vrfy_prealloc); vrfy_prealloc = malloc(secp256k1_context_preallocated_size(SECP256K1_CONTEXT_VERIFY)); CHECK(vrfy_prealloc != NULL); + ctx_tmp = vrfy; vrfy = secp256k1_context_preallocated_clone(vrfy, vrfy_prealloc); secp256k1_context_destroy(ctx_tmp); + + ctx_tmp = both; both = secp256k1_context_clone(both); secp256k1_context_preallocated_destroy(ctx_tmp); + free(both_prealloc); both_prealloc = malloc(secp256k1_context_preallocated_size(SECP256K1_CONTEXT_SIGN | SECP256K1_CONTEXT_VERIFY)); CHECK(both_prealloc != NULL); + ctx_tmp = both; both = secp256k1_context_preallocated_clone(both, both_prealloc); secp256k1_context_destroy(ctx_tmp); + } else { + /* clone into a preallocated context and then again into a new non-preallocated one. */ + void *prealloc_tmp; + + prealloc_tmp = malloc(secp256k1_context_preallocated_size(SECP256K1_CONTEXT_NONE)); CHECK(prealloc_tmp != NULL); + ctx_tmp = none; none = secp256k1_context_preallocated_clone(none, prealloc_tmp); secp256k1_context_destroy(ctx_tmp); + ctx_tmp = none; none = secp256k1_context_clone(none); secp256k1_context_preallocated_destroy(ctx_tmp); + free(prealloc_tmp); + + prealloc_tmp = malloc(secp256k1_context_preallocated_size(SECP256K1_CONTEXT_SIGN)); CHECK(prealloc_tmp != NULL); + ctx_tmp = sign; sign = secp256k1_context_preallocated_clone(sign, prealloc_tmp); secp256k1_context_destroy(ctx_tmp); + ctx_tmp = sign; sign = secp256k1_context_clone(sign); secp256k1_context_preallocated_destroy(ctx_tmp); + free(prealloc_tmp); + + prealloc_tmp = malloc(secp256k1_context_preallocated_size(SECP256K1_CONTEXT_VERIFY)); CHECK(prealloc_tmp != NULL); + ctx_tmp = vrfy; vrfy = secp256k1_context_preallocated_clone(vrfy, prealloc_tmp); secp256k1_context_destroy(ctx_tmp); + ctx_tmp = vrfy; vrfy = secp256k1_context_clone(vrfy); secp256k1_context_preallocated_destroy(ctx_tmp); + free(prealloc_tmp); + + prealloc_tmp = malloc(secp256k1_context_preallocated_size(SECP256K1_CONTEXT_SIGN | SECP256K1_CONTEXT_VERIFY)); CHECK(prealloc_tmp != NULL); + ctx_tmp = both; both = secp256k1_context_preallocated_clone(both, prealloc_tmp); secp256k1_context_destroy(ctx_tmp); + ctx_tmp = both; both = secp256k1_context_clone(both); secp256k1_context_preallocated_destroy(ctx_tmp); + free(prealloc_tmp); + } } /* Verify that the error callback makes it across the clone. */ @@ -218,17 +286,17 @@ void run_context_tests(void) { CHECK(ecount == 3); CHECK(secp256k1_ec_pubkey_tweak_mul(vrfy, &pubkey, ctmp) == 1); CHECK(ecount == 3); - CHECK(secp256k1_context_randomize(vrfy, ctmp) == 0); - CHECK(ecount == 4); + CHECK(secp256k1_context_randomize(vrfy, ctmp) == 1); + CHECK(ecount == 3); + CHECK(secp256k1_context_randomize(vrfy, NULL) == 1); + CHECK(ecount == 3); + CHECK(secp256k1_context_randomize(sign, ctmp) == 1); + CHECK(ecount2 == 14); CHECK(secp256k1_context_randomize(sign, NULL) == 1); CHECK(ecount2 == 14); secp256k1_context_set_illegal_callback(vrfy, NULL, NULL); secp256k1_context_set_illegal_callback(sign, NULL, NULL); - /* This shouldn't leak memory, due to already-set tests. */ - secp256k1_ecmult_gen_context_build(&sign->ecmult_gen_ctx, NULL); - secp256k1_ecmult_context_build(&vrfy->ecmult_ctx, NULL); - /* obtain a working nonce */ do { random_scalar_order_test(&nonce); @@ -243,49 +311,95 @@ void run_context_tests(void) { CHECK(secp256k1_ecdsa_sig_verify(&both->ecmult_ctx, &sigr, &sigs, &pub, &msg)); /* cleanup */ - secp256k1_context_destroy(none); - secp256k1_context_destroy(sign); - secp256k1_context_destroy(vrfy); - secp256k1_context_destroy(both); + if (use_prealloc) { + secp256k1_context_preallocated_destroy(none); + secp256k1_context_preallocated_destroy(sign); + secp256k1_context_preallocated_destroy(vrfy); + secp256k1_context_preallocated_destroy(both); + free(none_prealloc); + free(sign_prealloc); + free(vrfy_prealloc); + free(both_prealloc); + } else { + secp256k1_context_destroy(none); + secp256k1_context_destroy(sign); + secp256k1_context_destroy(vrfy); + secp256k1_context_destroy(both); + } /* Defined as no-op. */ secp256k1_context_destroy(NULL); + secp256k1_context_preallocated_destroy(NULL); + } void run_scratch_tests(void) { + const size_t adj_alloc = ((500 + ALIGNMENT - 1) / ALIGNMENT) * ALIGNMENT; + int32_t ecount = 0; + size_t checkpoint; + size_t checkpoint_2; secp256k1_context *none = secp256k1_context_create(SECP256K1_CONTEXT_NONE); secp256k1_scratch_space *scratch; + secp256k1_scratch_space local_scratch; /* Test public API */ secp256k1_context_set_illegal_callback(none, counting_illegal_callback_fn, &ecount); + secp256k1_context_set_error_callback(none, counting_illegal_callback_fn, &ecount); scratch = secp256k1_scratch_space_create(none, 1000); CHECK(scratch != NULL); CHECK(ecount == 0); /* Test internal API */ - CHECK(secp256k1_scratch_max_allocation(scratch, 0) == 1000); - CHECK(secp256k1_scratch_max_allocation(scratch, 1) < 1000); + CHECK(secp256k1_scratch_max_allocation(&none->error_callback, scratch, 0) == 1000); + CHECK(secp256k1_scratch_max_allocation(&none->error_callback, scratch, 1) == 1000 - (ALIGNMENT - 1)); + CHECK(scratch->alloc_size == 0); + CHECK(scratch->alloc_size % ALIGNMENT == 0); - /* Allocating 500 bytes with no frame fails */ - CHECK(secp256k1_scratch_alloc(scratch, 500) == NULL); - CHECK(secp256k1_scratch_max_allocation(scratch, 0) == 1000); + /* Allocating 500 bytes succeeds */ + checkpoint = secp256k1_scratch_checkpoint(&none->error_callback, scratch); + CHECK(secp256k1_scratch_alloc(&none->error_callback, scratch, 500) != NULL); + CHECK(secp256k1_scratch_max_allocation(&none->error_callback, scratch, 0) == 1000 - adj_alloc); + CHECK(secp256k1_scratch_max_allocation(&none->error_callback, scratch, 1) == 1000 - adj_alloc - (ALIGNMENT - 1)); + CHECK(scratch->alloc_size != 0); + CHECK(scratch->alloc_size % ALIGNMENT == 0); - /* ...but pushing a new stack frame does affect the max allocation */ - CHECK(secp256k1_scratch_allocate_frame(scratch, 500, 1 == 1)); - CHECK(secp256k1_scratch_max_allocation(scratch, 1) < 500); /* 500 - ALIGNMENT */ - CHECK(secp256k1_scratch_alloc(scratch, 500) != NULL); - CHECK(secp256k1_scratch_alloc(scratch, 500) == NULL); + /* Allocating another 500 bytes fails */ + CHECK(secp256k1_scratch_alloc(&none->error_callback, scratch, 500) == NULL); + CHECK(secp256k1_scratch_max_allocation(&none->error_callback, scratch, 0) == 1000 - adj_alloc); + CHECK(secp256k1_scratch_max_allocation(&none->error_callback, scratch, 1) == 1000 - adj_alloc - (ALIGNMENT - 1)); + CHECK(scratch->alloc_size != 0); + CHECK(scratch->alloc_size % ALIGNMENT == 0); - CHECK(secp256k1_scratch_allocate_frame(scratch, 500, 1) == 0); + /* ...but it succeeds once we apply the checkpoint to undo it */ + secp256k1_scratch_apply_checkpoint(&none->error_callback, scratch, checkpoint); + CHECK(scratch->alloc_size == 0); + CHECK(secp256k1_scratch_max_allocation(&none->error_callback, scratch, 0) == 1000); + CHECK(secp256k1_scratch_alloc(&none->error_callback, scratch, 500) != NULL); + CHECK(scratch->alloc_size != 0); - /* ...and this effect is undone by popping the frame */ - secp256k1_scratch_deallocate_frame(scratch); - CHECK(secp256k1_scratch_max_allocation(scratch, 0) == 1000); - CHECK(secp256k1_scratch_alloc(scratch, 500) == NULL); + /* try to apply a bad checkpoint */ + checkpoint_2 = secp256k1_scratch_checkpoint(&none->error_callback, scratch); + secp256k1_scratch_apply_checkpoint(&none->error_callback, scratch, checkpoint); + CHECK(ecount == 0); + secp256k1_scratch_apply_checkpoint(&none->error_callback, scratch, checkpoint_2); /* checkpoint_2 is after checkpoint */ + CHECK(ecount == 1); + secp256k1_scratch_apply_checkpoint(&none->error_callback, scratch, (size_t) -1); /* this is just wildly invalid */ + CHECK(ecount == 2); + + /* try to use badly initialized scratch space */ + secp256k1_scratch_space_destroy(none, scratch); + memset(&local_scratch, 0, sizeof(local_scratch)); + scratch = &local_scratch; + CHECK(!secp256k1_scratch_max_allocation(&none->error_callback, scratch, 0)); + CHECK(ecount == 3); + CHECK(secp256k1_scratch_alloc(&none->error_callback, scratch, 500) == NULL); + CHECK(ecount == 4); + secp256k1_scratch_space_destroy(none, scratch); + CHECK(ecount == 5); /* cleanup */ - secp256k1_scratch_space_destroy(scratch); + secp256k1_scratch_space_destroy(none, NULL); /* no-op */ secp256k1_context_destroy(none); } @@ -2095,7 +2209,6 @@ void test_ge(void) { /* Test batch gej -> ge conversion with and without known z ratios. */ { secp256k1_fe *zr = (secp256k1_fe *)checked_malloc(&ctx->error_callback, (4 * runs + 1) * sizeof(secp256k1_fe)); - secp256k1_ge *ge_set_table = (secp256k1_ge *)checked_malloc(&ctx->error_callback, (4 * runs + 1) * sizeof(secp256k1_ge)); secp256k1_ge *ge_set_all = (secp256k1_ge *)checked_malloc(&ctx->error_callback, (4 * runs + 1) * sizeof(secp256k1_ge)); for (i = 0; i < 4 * runs + 1; i++) { /* Compute gej[i + 1].z / gez[i].z (with gej[n].z taken to be 1). */ @@ -2103,20 +2216,33 @@ void test_ge(void) { secp256k1_fe_mul(&zr[i + 1], &zinv[i], &gej[i + 1].z); } } - secp256k1_ge_set_table_gej_var(ge_set_table, gej, zr, 4 * runs + 1); - secp256k1_ge_set_all_gej_var(ge_set_all, gej, 4 * runs + 1, &ctx->error_callback); + secp256k1_ge_set_all_gej_var(ge_set_all, gej, 4 * runs + 1); for (i = 0; i < 4 * runs + 1; i++) { secp256k1_fe s; random_fe_non_zero(&s); secp256k1_gej_rescale(&gej[i], &s); - ge_equals_gej(&ge_set_table[i], &gej[i]); ge_equals_gej(&ge_set_all[i], &gej[i]); } - free(ge_set_table); free(ge_set_all); free(zr); } + /* Test batch gej -> ge conversion with many infinities. */ + for (i = 0; i < 4 * runs + 1; i++) { + random_group_element_test(&ge[i]); + /* randomly set half the points to infinitiy */ + if(secp256k1_fe_is_odd(&ge[i].x)) { + secp256k1_ge_set_infinity(&ge[i]); + } + secp256k1_gej_set_ge(&gej[i], &ge[i]); + } + /* batch invert */ + secp256k1_ge_set_all_gej_var(ge, gej, 4 * runs + 1); + /* check result */ + for (i = 0; i < 4 * runs + 1; i++) { + ge_equals_gej(&ge[i], &gej[i]); + } + free(ge); free(gej); free(zinv); @@ -2556,14 +2682,13 @@ void test_ecmult_multi(secp256k1_scratch *scratch, secp256k1_ecmult_multi_func e secp256k1_gej r; secp256k1_gej r2; ecmult_multi_data data; - secp256k1_scratch *scratch_empty; data.sc = sc; data.pt = pt; secp256k1_scalar_set_int(&szero, 0); /* No points to multiply */ - CHECK(ecmult_multi(&ctx->ecmult_ctx, scratch, &r, NULL, ecmult_multi_callback, &data, 0)); + CHECK(ecmult_multi(&ctx->error_callback, &ctx->ecmult_ctx, scratch, &r, NULL, ecmult_multi_callback, &data, 0)); /* Check 1- and 2-point multiplies against ecmult */ for (ncount = 0; ncount < count; ncount++) { @@ -2579,36 +2704,31 @@ void test_ecmult_multi(secp256k1_scratch *scratch, secp256k1_ecmult_multi_func e /* only G scalar */ secp256k1_ecmult(&ctx->ecmult_ctx, &r2, &ptgj, &szero, &sc[0]); - CHECK(ecmult_multi(&ctx->ecmult_ctx, scratch, &r, &sc[0], ecmult_multi_callback, &data, 0)); + CHECK(ecmult_multi(&ctx->error_callback, &ctx->ecmult_ctx, scratch, &r, &sc[0], ecmult_multi_callback, &data, 0)); secp256k1_gej_neg(&r2, &r2); secp256k1_gej_add_var(&r, &r, &r2, NULL); CHECK(secp256k1_gej_is_infinity(&r)); /* 1-point */ secp256k1_ecmult(&ctx->ecmult_ctx, &r2, &ptgj, &sc[0], &szero); - CHECK(ecmult_multi(&ctx->ecmult_ctx, scratch, &r, &szero, ecmult_multi_callback, &data, 1)); + CHECK(ecmult_multi(&ctx->error_callback, &ctx->ecmult_ctx, scratch, &r, &szero, ecmult_multi_callback, &data, 1)); secp256k1_gej_neg(&r2, &r2); secp256k1_gej_add_var(&r, &r, &r2, NULL); CHECK(secp256k1_gej_is_infinity(&r)); - /* Try to multiply 1 point, but scratch space is empty */ - scratch_empty = secp256k1_scratch_create(&ctx->error_callback, 0); - CHECK(!ecmult_multi(&ctx->ecmult_ctx, scratch_empty, &r, &szero, ecmult_multi_callback, &data, 1)); - secp256k1_scratch_destroy(scratch_empty); - /* Try to multiply 1 point, but callback returns false */ - CHECK(!ecmult_multi(&ctx->ecmult_ctx, scratch, &r, &szero, ecmult_multi_false_callback, &data, 1)); + CHECK(!ecmult_multi(&ctx->error_callback, &ctx->ecmult_ctx, scratch, &r, &szero, ecmult_multi_false_callback, &data, 1)); /* 2-point */ secp256k1_ecmult(&ctx->ecmult_ctx, &r2, &ptgj, &sc[0], &sc[1]); - CHECK(ecmult_multi(&ctx->ecmult_ctx, scratch, &r, &szero, ecmult_multi_callback, &data, 2)); + CHECK(ecmult_multi(&ctx->error_callback, &ctx->ecmult_ctx, scratch, &r, &szero, ecmult_multi_callback, &data, 2)); secp256k1_gej_neg(&r2, &r2); secp256k1_gej_add_var(&r, &r, &r2, NULL); CHECK(secp256k1_gej_is_infinity(&r)); /* 2-point with G scalar */ secp256k1_ecmult(&ctx->ecmult_ctx, &r2, &ptgj, &sc[0], &sc[1]); - CHECK(ecmult_multi(&ctx->ecmult_ctx, scratch, &r, &sc[1], ecmult_multi_callback, &data, 1)); + CHECK(ecmult_multi(&ctx->error_callback, &ctx->ecmult_ctx, scratch, &r, &sc[1], ecmult_multi_callback, &data, 1)); secp256k1_gej_neg(&r2, &r2); secp256k1_gej_add_var(&r, &r, &r2, NULL); CHECK(secp256k1_gej_is_infinity(&r)); @@ -2625,7 +2745,7 @@ void test_ecmult_multi(secp256k1_scratch *scratch, secp256k1_ecmult_multi_func e random_scalar_order(&sc[i]); secp256k1_ge_set_infinity(&pt[i]); } - CHECK(ecmult_multi(&ctx->ecmult_ctx, scratch, &r, &szero, ecmult_multi_callback, &data, sizes[j])); + CHECK(ecmult_multi(&ctx->error_callback, &ctx->ecmult_ctx, scratch, &r, &szero, ecmult_multi_callback, &data, sizes[j])); CHECK(secp256k1_gej_is_infinity(&r)); } @@ -2635,7 +2755,7 @@ void test_ecmult_multi(secp256k1_scratch *scratch, secp256k1_ecmult_multi_func e pt[i] = ptg; secp256k1_scalar_set_int(&sc[i], 0); } - CHECK(ecmult_multi(&ctx->ecmult_ctx, scratch, &r, &szero, ecmult_multi_callback, &data, sizes[j])); + CHECK(ecmult_multi(&ctx->error_callback, &ctx->ecmult_ctx, scratch, &r, &szero, ecmult_multi_callback, &data, sizes[j])); CHECK(secp256k1_gej_is_infinity(&r)); } @@ -2648,7 +2768,7 @@ void test_ecmult_multi(secp256k1_scratch *scratch, secp256k1_ecmult_multi_func e pt[2 * i + 1] = ptg; } - CHECK(ecmult_multi(&ctx->ecmult_ctx, scratch, &r, &szero, ecmult_multi_callback, &data, sizes[j])); + CHECK(ecmult_multi(&ctx->error_callback, &ctx->ecmult_ctx, scratch, &r, &szero, ecmult_multi_callback, &data, sizes[j])); CHECK(secp256k1_gej_is_infinity(&r)); random_scalar_order(&sc[0]); @@ -2661,7 +2781,7 @@ void test_ecmult_multi(secp256k1_scratch *scratch, secp256k1_ecmult_multi_func e secp256k1_ge_neg(&pt[2*i+1], &pt[2*i]); } - CHECK(ecmult_multi(&ctx->ecmult_ctx, scratch, &r, &szero, ecmult_multi_callback, &data, sizes[j])); + CHECK(ecmult_multi(&ctx->error_callback, &ctx->ecmult_ctx, scratch, &r, &szero, ecmult_multi_callback, &data, sizes[j])); CHECK(secp256k1_gej_is_infinity(&r)); } @@ -2676,7 +2796,7 @@ void test_ecmult_multi(secp256k1_scratch *scratch, secp256k1_ecmult_multi_func e secp256k1_scalar_negate(&sc[i], &sc[i]); } - CHECK(ecmult_multi(&ctx->ecmult_ctx, scratch, &r, &szero, ecmult_multi_callback, &data, 32)); + CHECK(ecmult_multi(&ctx->error_callback, &ctx->ecmult_ctx, scratch, &r, &szero, ecmult_multi_callback, &data, 32)); CHECK(secp256k1_gej_is_infinity(&r)); } @@ -2695,7 +2815,7 @@ void test_ecmult_multi(secp256k1_scratch *scratch, secp256k1_ecmult_multi_func e } secp256k1_ecmult(&ctx->ecmult_ctx, &r2, &r, &sc[0], &szero); - CHECK(ecmult_multi(&ctx->ecmult_ctx, scratch, &r, &szero, ecmult_multi_callback, &data, 20)); + CHECK(ecmult_multi(&ctx->error_callback, &ctx->ecmult_ctx, scratch, &r, &szero, ecmult_multi_callback, &data, 20)); secp256k1_gej_neg(&r2, &r2); secp256k1_gej_add_var(&r, &r, &r2, NULL); CHECK(secp256k1_gej_is_infinity(&r)); @@ -2718,7 +2838,7 @@ void test_ecmult_multi(secp256k1_scratch *scratch, secp256k1_ecmult_multi_func e secp256k1_gej_set_ge(&p0j, &pt[0]); secp256k1_ecmult(&ctx->ecmult_ctx, &r2, &p0j, &rs, &szero); - CHECK(ecmult_multi(&ctx->ecmult_ctx, scratch, &r, &szero, ecmult_multi_callback, &data, 20)); + CHECK(ecmult_multi(&ctx->error_callback, &ctx->ecmult_ctx, scratch, &r, &szero, ecmult_multi_callback, &data, 20)); secp256k1_gej_neg(&r2, &r2); secp256k1_gej_add_var(&r, &r, &r2, NULL); CHECK(secp256k1_gej_is_infinity(&r)); @@ -2731,13 +2851,13 @@ void test_ecmult_multi(secp256k1_scratch *scratch, secp256k1_ecmult_multi_func e } secp256k1_scalar_clear(&sc[0]); - CHECK(ecmult_multi(&ctx->ecmult_ctx, scratch, &r, &szero, ecmult_multi_callback, &data, 20)); + CHECK(ecmult_multi(&ctx->error_callback, &ctx->ecmult_ctx, scratch, &r, &szero, ecmult_multi_callback, &data, 20)); secp256k1_scalar_clear(&sc[1]); secp256k1_scalar_clear(&sc[2]); secp256k1_scalar_clear(&sc[3]); secp256k1_scalar_clear(&sc[4]); - CHECK(ecmult_multi(&ctx->ecmult_ctx, scratch, &r, &szero, ecmult_multi_callback, &data, 6)); - CHECK(ecmult_multi(&ctx->ecmult_ctx, scratch, &r, &szero, ecmult_multi_callback, &data, 5)); + CHECK(ecmult_multi(&ctx->error_callback, &ctx->ecmult_ctx, scratch, &r, &szero, ecmult_multi_callback, &data, 6)); + CHECK(ecmult_multi(&ctx->error_callback, &ctx->ecmult_ctx, scratch, &r, &szero, ecmult_multi_callback, &data, 5)); CHECK(secp256k1_gej_is_infinity(&r)); /* Run through s0*(t0*P) + s1*(t1*P) exhaustively for many small values of s0, s1, t0, t1 */ @@ -2782,7 +2902,7 @@ void test_ecmult_multi(secp256k1_scratch *scratch, secp256k1_ecmult_multi_func e secp256k1_scalar_add(&tmp1, &tmp1, &tmp2); secp256k1_ecmult(&ctx->ecmult_ctx, &expected, &ptgj, &tmp1, &szero); - CHECK(ecmult_multi(&ctx->ecmult_ctx, scratch, &actual, &szero, ecmult_multi_callback, &data, 2)); + CHECK(ecmult_multi(&ctx->error_callback, &ctx->ecmult_ctx, scratch, &actual, &szero, ecmult_multi_callback, &data, 2)); secp256k1_gej_neg(&expected, &expected); secp256k1_gej_add_var(&actual, &actual, &expected, NULL); CHECK(secp256k1_gej_is_infinity(&actual)); @@ -2793,6 +2913,24 @@ void test_ecmult_multi(secp256k1_scratch *scratch, secp256k1_ecmult_multi_func e } } +void test_ecmult_multi_batch_single(secp256k1_ecmult_multi_func ecmult_multi) { + secp256k1_scalar szero; + secp256k1_scalar sc[32]; + secp256k1_ge pt[32]; + secp256k1_gej r; + ecmult_multi_data data; + secp256k1_scratch *scratch_empty; + + data.sc = sc; + data.pt = pt; + secp256k1_scalar_set_int(&szero, 0); + + /* Try to multiply 1 point, but scratch space is empty.*/ + scratch_empty = secp256k1_scratch_create(&ctx->error_callback, 0); + CHECK(!ecmult_multi(&ctx->error_callback, &ctx->ecmult_ctx, scratch_empty, &r, &szero, ecmult_multi_callback, &data, 1)); + secp256k1_scratch_destroy(&ctx->error_callback, scratch_empty); +} + void test_secp256k1_pippenger_bucket_window_inv(void) { int i; @@ -2823,21 +2961,75 @@ void test_ecmult_multi_pippenger_max_points(void) { int bucket_window = 0; for(; scratch_size < max_size; scratch_size+=256) { + size_t i; + size_t total_alloc; + size_t checkpoint; scratch = secp256k1_scratch_create(&ctx->error_callback, scratch_size); CHECK(scratch != NULL); - n_points_supported = secp256k1_pippenger_max_points(scratch); + checkpoint = secp256k1_scratch_checkpoint(&ctx->error_callback, scratch); + n_points_supported = secp256k1_pippenger_max_points(&ctx->error_callback, scratch); if (n_points_supported == 0) { - secp256k1_scratch_destroy(scratch); + secp256k1_scratch_destroy(&ctx->error_callback, scratch); continue; } bucket_window = secp256k1_pippenger_bucket_window(n_points_supported); - CHECK(secp256k1_scratch_allocate_frame(scratch, secp256k1_pippenger_scratch_size(n_points_supported, bucket_window), PIPPENGER_SCRATCH_OBJECTS)); - secp256k1_scratch_deallocate_frame(scratch); - secp256k1_scratch_destroy(scratch); + /* allocate `total_alloc` bytes over `PIPPENGER_SCRATCH_OBJECTS` many allocations */ + total_alloc = secp256k1_pippenger_scratch_size(n_points_supported, bucket_window); + for (i = 0; i < PIPPENGER_SCRATCH_OBJECTS - 1; i++) { + CHECK(secp256k1_scratch_alloc(&ctx->error_callback, scratch, 1)); + total_alloc--; + } + CHECK(secp256k1_scratch_alloc(&ctx->error_callback, scratch, total_alloc)); + secp256k1_scratch_apply_checkpoint(&ctx->error_callback, scratch, checkpoint); + secp256k1_scratch_destroy(&ctx->error_callback, scratch); } CHECK(bucket_window == PIPPENGER_MAX_BUCKET_WINDOW); } +void test_ecmult_multi_batch_size_helper(void) { + size_t n_batches, n_batch_points, max_n_batch_points, n; + + max_n_batch_points = 0; + n = 1; + CHECK(secp256k1_ecmult_multi_batch_size_helper(&n_batches, &n_batch_points, max_n_batch_points, n) == 0); + + max_n_batch_points = 1; + n = 0; + CHECK(secp256k1_ecmult_multi_batch_size_helper(&n_batches, &n_batch_points, max_n_batch_points, n) == 1); + CHECK(n_batches == 0); + CHECK(n_batch_points == 0); + + max_n_batch_points = 2; + n = 5; + CHECK(secp256k1_ecmult_multi_batch_size_helper(&n_batches, &n_batch_points, max_n_batch_points, n) == 1); + CHECK(n_batches == 3); + CHECK(n_batch_points == 2); + + max_n_batch_points = ECMULT_MAX_POINTS_PER_BATCH; + n = ECMULT_MAX_POINTS_PER_BATCH; + CHECK(secp256k1_ecmult_multi_batch_size_helper(&n_batches, &n_batch_points, max_n_batch_points, n) == 1); + CHECK(n_batches == 1); + CHECK(n_batch_points == ECMULT_MAX_POINTS_PER_BATCH); + + max_n_batch_points = ECMULT_MAX_POINTS_PER_BATCH + 1; + n = ECMULT_MAX_POINTS_PER_BATCH + 1; + CHECK(secp256k1_ecmult_multi_batch_size_helper(&n_batches, &n_batch_points, max_n_batch_points, n) == 1); + CHECK(n_batches == 2); + CHECK(n_batch_points == ECMULT_MAX_POINTS_PER_BATCH/2 + 1); + + max_n_batch_points = 1; + n = SIZE_MAX; + CHECK(secp256k1_ecmult_multi_batch_size_helper(&n_batches, &n_batch_points, max_n_batch_points, n) == 1); + CHECK(n_batches == SIZE_MAX); + CHECK(n_batch_points == 1); + + max_n_batch_points = 2; + n = SIZE_MAX; + CHECK(secp256k1_ecmult_multi_batch_size_helper(&n_batches, &n_batch_points, max_n_batch_points, n) == 1); + CHECK(n_batches == SIZE_MAX/2 + 1); + CHECK(n_batch_points == 2); +} + /** * Run secp256k1_ecmult_multi_var with num points and a scratch space restricted to * 1 <= i <= num points. @@ -2872,19 +3064,25 @@ void test_ecmult_multi_batching(void) { } data.sc = sc; data.pt = pt; + secp256k1_gej_neg(&r2, &r2); - /* Test with empty scratch space */ + /* Test with empty scratch space. It should compute the correct result using + * ecmult_mult_simple algorithm which doesn't require a scratch space. */ scratch = secp256k1_scratch_create(&ctx->error_callback, 0); - CHECK(!secp256k1_ecmult_multi_var(&ctx->ecmult_ctx, scratch, &r, &scG, ecmult_multi_callback, &data, 1)); - secp256k1_scratch_destroy(scratch); + CHECK(secp256k1_ecmult_multi_var(&ctx->error_callback, &ctx->ecmult_ctx, scratch, &r, &scG, ecmult_multi_callback, &data, n_points)); + secp256k1_gej_add_var(&r, &r, &r2, NULL); + CHECK(secp256k1_gej_is_infinity(&r)); + secp256k1_scratch_destroy(&ctx->error_callback, scratch); /* Test with space for 1 point in pippenger. That's not enough because - * ecmult_multi selects strauss which requires more memory. */ + * ecmult_multi selects strauss which requires more memory. It should + * therefore select the simple algorithm. */ scratch = secp256k1_scratch_create(&ctx->error_callback, secp256k1_pippenger_scratch_size(1, 1) + PIPPENGER_SCRATCH_OBJECTS*ALIGNMENT); - CHECK(!secp256k1_ecmult_multi_var(&ctx->ecmult_ctx, scratch, &r, &scG, ecmult_multi_callback, &data, 1)); - secp256k1_scratch_destroy(scratch); + CHECK(secp256k1_ecmult_multi_var(&ctx->error_callback, &ctx->ecmult_ctx, scratch, &r, &scG, ecmult_multi_callback, &data, n_points)); + secp256k1_gej_add_var(&r, &r, &r2, NULL); + CHECK(secp256k1_gej_is_infinity(&r)); + secp256k1_scratch_destroy(&ctx->error_callback, scratch); - secp256k1_gej_neg(&r2, &r2); for(i = 1; i <= n_points; i++) { if (i > ECMULT_PIPPENGER_THRESHOLD) { int bucket_window = secp256k1_pippenger_bucket_window(i); @@ -2894,10 +3092,10 @@ void test_ecmult_multi_batching(void) { size_t scratch_size = secp256k1_strauss_scratch_size(i); scratch = secp256k1_scratch_create(&ctx->error_callback, scratch_size + STRAUSS_SCRATCH_OBJECTS*ALIGNMENT); } - CHECK(secp256k1_ecmult_multi_var(&ctx->ecmult_ctx, scratch, &r, &scG, ecmult_multi_callback, &data, n_points)); + CHECK(secp256k1_ecmult_multi_var(&ctx->error_callback, &ctx->ecmult_ctx, scratch, &r, &scG, ecmult_multi_callback, &data, n_points)); secp256k1_gej_add_var(&r, &r, &r2, NULL); CHECK(secp256k1_gej_is_infinity(&r)); - secp256k1_scratch_destroy(scratch); + secp256k1_scratch_destroy(&ctx->error_callback, scratch); } free(sc); free(pt); @@ -2910,15 +3108,19 @@ void run_ecmult_multi_tests(void) { test_ecmult_multi_pippenger_max_points(); scratch = secp256k1_scratch_create(&ctx->error_callback, 819200); test_ecmult_multi(scratch, secp256k1_ecmult_multi_var); + test_ecmult_multi(NULL, secp256k1_ecmult_multi_var); test_ecmult_multi(scratch, secp256k1_ecmult_pippenger_batch_single); + test_ecmult_multi_batch_single(secp256k1_ecmult_pippenger_batch_single); test_ecmult_multi(scratch, secp256k1_ecmult_strauss_batch_single); - secp256k1_scratch_destroy(scratch); + test_ecmult_multi_batch_single(secp256k1_ecmult_strauss_batch_single); + secp256k1_scratch_destroy(&ctx->error_callback, scratch); /* Run test_ecmult_multi with space for exactly one point */ scratch = secp256k1_scratch_create(&ctx->error_callback, secp256k1_strauss_scratch_size(1) + STRAUSS_SCRATCH_OBJECTS*ALIGNMENT); test_ecmult_multi(scratch, secp256k1_ecmult_multi_var); - secp256k1_scratch_destroy(scratch); + secp256k1_scratch_destroy(&ctx->error_callback, scratch); + test_ecmult_multi_batch_size_helper(); test_ecmult_multi_batching(); } @@ -2988,7 +3190,7 @@ void test_constant_wnaf(const secp256k1_scalar *number, int w) { } bits = 128; #endif - skew = secp256k1_wnaf_const(wnaf, num, w, bits); + skew = secp256k1_wnaf_const(wnaf, &num, w, bits); for (i = WNAF_SIZE_BITS(bits, w); i >= 0; --i) { secp256k1_scalar t; @@ -4978,8 +5180,9 @@ int main(int argc, char **argv) { } } else { FILE *frand = fopen("/dev/urandom", "r"); - if ((frand == NULL) || fread(&seed16, sizeof(seed16), 1, frand) != sizeof(seed16)) { + if ((frand == NULL) || fread(&seed16, 1, sizeof(seed16), frand) != sizeof(seed16)) { uint64_t t = time(NULL) * (uint64_t)1337; + fprintf(stderr, "WARNING: could not read 16 bytes from /dev/urandom; falling back to insecure PRNG\n"); seed16[0] ^= t; seed16[1] ^= t >> 8; seed16[2] ^= t >> 16; @@ -4999,7 +5202,8 @@ int main(int argc, char **argv) { printf("random seed = %02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x\n", seed16[0], seed16[1], seed16[2], seed16[3], seed16[4], seed16[5], seed16[6], seed16[7], seed16[8], seed16[9], seed16[10], seed16[11], seed16[12], seed16[13], seed16[14], seed16[15]); /* initialize */ - run_context_tests(); + run_context_tests(0); + run_context_tests(1); run_scratch_tests(); ctx = secp256k1_context_create(SECP256K1_CONTEXT_SIGN | SECP256K1_CONTEXT_VERIFY); if (secp256k1_rand_bits(1)) { diff --git a/depend/secp256k1/src/tests_exhaustive.c b/depend/secp256k1/src/tests_exhaustive.c index ab9779b..b44e357 100644 --- a/depend/secp256k1/src/tests_exhaustive.c +++ b/depend/secp256k1/src/tests_exhaustive.c @@ -212,14 +212,14 @@ void test_exhaustive_ecmult_multi(const secp256k1_context *ctx, const secp256k1_ data.pt[0] = group[x]; data.pt[1] = group[y]; - secp256k1_ecmult_multi_var(&ctx->ecmult_ctx, scratch, &tmp, &g_sc, ecmult_multi_callback, &data, 2); + secp256k1_ecmult_multi_var(&ctx->error_callback, &ctx->ecmult_ctx, scratch, &tmp, &g_sc, ecmult_multi_callback, &data, 2); ge_equals_gej(&group[(i * x + j * y + k) % order], &tmp); } } } } } - secp256k1_scratch_destroy(scratch); + secp256k1_scratch_destroy(&ctx->error_callback, scratch); } void r_from_k(secp256k1_scalar *r, const secp256k1_ge *group, int k) { diff --git a/depend/secp256k1/src/util.h b/depend/secp256k1/src/util.h index e014750..9deb61b 100644 --- a/depend/secp256k1/src/util.h +++ b/depend/secp256k1/src/util.h @@ -36,7 +36,7 @@ static SECP256K1_INLINE void secp256k1_callback_call(const secp256k1_callback * } while(0) #endif -#ifdef HAVE_BUILTIN_EXPECT +#if SECP256K1_GNUC_PREREQ(3, 0) #define EXPECT(x,c) __builtin_expect((x),(c)) #else #define EXPECT(x,c) (x) @@ -84,6 +84,47 @@ static SECP256K1_INLINE void *checked_realloc(const secp256k1_callback* cb, void 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 + +#define ROUND_TO_ALIGN(size) (((size + ALIGNMENT - 1) / ALIGNMENT) * ALIGNMENT) + +/* Assume there is a contiguous memory object with bounds [base, base + max_size) + * of which the memory range [base, *prealloc_ptr) is already allocated for usage, + * where *prealloc_ptr is an aligned pointer. In that setting, this functions + * reserves the subobject [*prealloc_ptr, *prealloc_ptr + alloc_size) of + * alloc_size bytes by increasing *prealloc_ptr accordingly, taking into account + * alignment requirements. + * + * The function returns an aligned pointer to the newly allocated subobject. + * + * This is useful for manual memory management: if we're simply given a block + * [base, base + max_size), the caller can use this function to allocate memory + * in this block and keep track of the current allocation state with *prealloc_ptr. + * + * It is VERIFY_CHECKed that there is enough space left in the memory object and + * *prealloc_ptr is aligned relative to base. + */ +static SECP256K1_INLINE void *manual_alloc(void** prealloc_ptr, size_t alloc_size, void* base, size_t max_size) { + size_t aligned_alloc_size = ROUND_TO_ALIGN(alloc_size); + void* ret; + VERIFY_CHECK(prealloc_ptr != NULL); + VERIFY_CHECK(*prealloc_ptr != NULL); + VERIFY_CHECK(base != NULL); + VERIFY_CHECK((unsigned char*)*prealloc_ptr >= (unsigned char*)base); + VERIFY_CHECK(((unsigned char*)*prealloc_ptr - (unsigned char*)base) % ALIGNMENT == 0); + VERIFY_CHECK((unsigned char*)*prealloc_ptr - (unsigned char*)base + aligned_alloc_size <= max_size); + ret = *prealloc_ptr; + *((unsigned char**)prealloc_ptr) += aligned_alloc_size; + return ret; +} + /* Macro for restrict, when available and not in a VERIFY build. */ #if defined(SECP256K1_BUILD) && defined(VERIFY) # define SECP256K1_RESTRICT