diff --git a/.project b/.project index ef47c5d..bc695d4 100644 --- a/.project +++ b/.project @@ -5,6 +5,11 @@ + + org.python.pydev.PyDevBuilder + + + org.eclipse.cdt.managedbuilder.core.genmakebuilder clean,full,incremental, @@ -23,5 +28,6 @@ org.eclipse.cdt.managedbuilder.core.managedBuildNature org.eclipse.cdt.managedbuilder.core.ScannerConfigNature fr.ac6.mcu.ide.core.MCUProjectNature + org.python.pydev.pythonNature diff --git a/.pydevproject b/.pydevproject new file mode 100644 index 0000000..d001f0a --- /dev/null +++ b/.pydevproject @@ -0,0 +1,5 @@ + + +Default +python interpreter + diff --git a/Inc/main.h b/Inc/main.h index a4f8f96..684830b 100644 --- a/Inc/main.h +++ b/Inc/main.h @@ -7,7 +7,7 @@ #define FW_MAGIC 0x46574853 // The constants below define the memory layout documented in the README.md file -#define BOOTLOADER_SIZE 0x1000 +#define BOOTLOADER_SIZE 0x2000 #define FIRMWARE_SIZE ((FLASH_BANK_SIZE - BOOTLOADER_SIZE) / 2) #define FIRMWARE_HEADER_SIZE 0x200 #define FIRMWARE_FIRST_PAGE (BOOTLOADER_SIZE / FLASH_PAGE_SIZE) diff --git a/Inc/uECC.h b/Inc/uECC.h new file mode 100644 index 0000000..c5cbc72 --- /dev/null +++ b/Inc/uECC.h @@ -0,0 +1,123 @@ +/* Copyright 2014, Kenneth MacKay. Licensed under the BSD 2-clause license. */ + +#ifndef _UECC_H_ +#define _UECC_H_ + +#include + +#define uECC_ENABLE_VLI_API 0 +#define uECC_OPTIMIZATION_LEVEL 2 +#define uECC_SUPPORT_COMPRESSED_POINT 0 + +/* Platform selection options. +If uECC_PLATFORM is not defined, the code will try to guess it based on compiler macros. +Possible values for uECC_PLATFORM are defined below: */ +#define uECC_arch_other 0 +#define uECC_x86 1 +#define uECC_x86_64 2 +#define uECC_arm 3 +#define uECC_arm_thumb 4 +#define uECC_arm_thumb2 5 +#define uECC_arm64 6 +#define uECC_avr 7 + +/* If desired, you can define uECC_WORD_SIZE as appropriate for your platform (1, 4, or 8 bytes). +If uECC_WORD_SIZE is not explicitly defined then it will be automatically set based on your +platform. */ + +/* Optimization level; trade speed for code size. + Larger values produce code that is faster but larger. + Currently supported values are 0 - 4; 0 is unusably slow for most applications. + Optimization level 4 currently only has an effect ARM platforms where more than one + curve is enabled. */ +#ifndef uECC_OPTIMIZATION_LEVEL + #define uECC_OPTIMIZATION_LEVEL 2 +#endif + +/* uECC_SQUARE_FUNC - If enabled (defined as nonzero), this will cause a specific function to be +used for (scalar) squaring instead of the generic multiplication function. This can make things +faster somewhat faster, but increases the code size. */ +#ifndef uECC_SQUARE_FUNC + #define uECC_SQUARE_FUNC 0 +#endif + +/* uECC_VLI_NATIVE_LITTLE_ENDIAN - If enabled (defined as nonzero), this will switch to native +little-endian format for *all* arrays passed in and out of the public API. This includes public +and private keys, shared secrets, signatures and message hashes. +Using this switch reduces the amount of call stack memory used by uECC, since less intermediate +translations are required. +Note that this will *only* work on native little-endian processors and it will treat the uint8_t +arrays passed into the public API as word arrays, therefore requiring the provided byte arrays +to be word aligned on architectures that do not support unaligned accesses. +IMPORTANT: Keys and signatures generated with uECC_VLI_NATIVE_LITTLE_ENDIAN=1 are incompatible +with keys and signatures generated with uECC_VLI_NATIVE_LITTLE_ENDIAN=0; all parties must use +the same endianness. */ +#ifndef uECC_VLI_NATIVE_LITTLE_ENDIAN + #define uECC_VLI_NATIVE_LITTLE_ENDIAN 0 +#endif + +/* Curve support selection. Set to 0 to remove that curve. */ +#ifndef uECC_SUPPORTS_secp160r1 + #define uECC_SUPPORTS_secp160r1 0 +#endif +#ifndef uECC_SUPPORTS_secp192r1 + #define uECC_SUPPORTS_secp192r1 0 +#endif +#ifndef uECC_SUPPORTS_secp224r1 + #define uECC_SUPPORTS_secp224r1 0 +#endif +#ifndef uECC_SUPPORTS_secp256r1 + #define uECC_SUPPORTS_secp256r1 0 +#endif +#ifndef uECC_SUPPORTS_secp256k1 + #define uECC_SUPPORTS_secp256k1 1 +#endif + +/* Specifies whether compressed point format is supported. + Set to 0 to disable point compression/decompression functions. */ +#ifndef uECC_SUPPORT_COMPRESSED_POINT + #define uECC_SUPPORT_COMPRESSED_POINT 1 +#endif + +struct uECC_Curve_t; +typedef const struct uECC_Curve_t * uECC_Curve; + +#ifdef __cplusplus +extern "C" +{ +#endif + + +uECC_Curve uECC_secp256k1(void); + +/* uECC_curve_public_key_size() function. + +Returns the size of a public key for the curve in bytes. +*/ +int uECC_curve_public_key_size(uECC_Curve curve); + +/* uECC_verify() function. +Verify an ECDSA signature. + +Usage: Compute the hash of the signed data using the same hash as the signer and +pass it to this function along with the signer's public key and the signature values (r and s). + +Inputs: + public_key - The signer's public key. + message_hash - The hash of the signed data. + hash_size - The size of message_hash in bytes. + signature - The signature value. + +Returns 1 if the signature is valid, 0 if it is invalid. +*/ +int uECC_verify(const uint8_t *public_key, + const uint8_t *message_hash, + unsigned hash_size, + const uint8_t *signature, + uECC_Curve curve); + +#ifdef __cplusplus +} /* end of extern "C" */ +#endif + +#endif /* _UECC_H_ */ diff --git a/Inc/uECC_asm_arm.inc b/Inc/uECC_asm_arm.inc new file mode 100644 index 0000000..30f9a63 --- /dev/null +++ b/Inc/uECC_asm_arm.inc @@ -0,0 +1,225 @@ +/* Copyright 2015, Kenneth MacKay. Licensed under the BSD 2-clause license. */ + +#ifndef _UECC_ASM_ARM_H_ +#define _UECC_ASM_ARM_H_ + +#if (uECC_SUPPORTS_secp256r1 || uECC_SUPPORTS_secp256k1) + #define uECC_MIN_WORDS 8 +#endif +#if uECC_SUPPORTS_secp224r1 + #undef uECC_MIN_WORDS + #define uECC_MIN_WORDS 7 +#endif +#if uECC_SUPPORTS_secp192r1 + #undef uECC_MIN_WORDS + #define uECC_MIN_WORDS 6 +#endif +#if uECC_SUPPORTS_secp160r1 + #undef uECC_MIN_WORDS + #define uECC_MIN_WORDS 5 +#endif + +#if (uECC_PLATFORM == uECC_arm_thumb) + #define REG_RW "+l" + #define REG_WRITE "=l" +#else + #define REG_RW "+r" + #define REG_WRITE "=r" +#endif + +#if (uECC_PLATFORM == uECC_arm_thumb || uECC_PLATFORM == uECC_arm_thumb2) + #define REG_RW_LO "+l" + #define REG_WRITE_LO "=l" +#else + #define REG_RW_LO "+r" + #define REG_WRITE_LO "=r" +#endif + +#if (uECC_PLATFORM == uECC_arm_thumb2) + #define RESUME_SYNTAX +#else + #define RESUME_SYNTAX ".syntax divided \n\t" +#endif + +#if (uECC_OPTIMIZATION_LEVEL >= 2) + +uECC_VLI_API uECC_word_t uECC_vli_add(uECC_word_t *result, + const uECC_word_t *left, + const uECC_word_t *right, + wordcount_t num_words) { +#if (uECC_MAX_WORDS != uECC_MIN_WORDS) + #if (uECC_PLATFORM == uECC_arm_thumb) || (uECC_PLATFORM == uECC_arm_thumb2) + uint32_t jump = (uECC_MAX_WORDS - num_words) * 4 * 2 + 1; + #else /* ARM */ + uint32_t jump = (uECC_MAX_WORDS - num_words) * 4 * 4; + #endif +#endif + uint32_t carry; + uint32_t left_word; + uint32_t right_word; + + __asm__ volatile ( + ".syntax unified \n\t" + "movs %[carry], #0 \n\t" + #if (uECC_MAX_WORDS != uECC_MIN_WORDS) + "adr %[left], 1f \n\t" + ".align 4 \n\t" + "adds %[jump], %[left] \n\t" + #endif + + "ldmia %[lptr]!, {%[left]} \n\t" + "ldmia %[rptr]!, {%[right]} \n\t" + "adds %[left], %[right] \n\t" + "stmia %[dptr]!, {%[left]} \n\t" + + #if (uECC_MAX_WORDS != uECC_MIN_WORDS) + "bx %[jump] \n\t" + #endif + "1: \n\t" + REPEAT(DEC(uECC_MAX_WORDS), + "ldmia %[lptr]!, {%[left]} \n\t" + "ldmia %[rptr]!, {%[right]} \n\t" + "adcs %[left], %[right] \n\t" + "stmia %[dptr]!, {%[left]} \n\t") + + "adcs %[carry], %[carry] \n\t" + RESUME_SYNTAX + : [dptr] REG_RW_LO (result), [lptr] REG_RW_LO (left), [rptr] REG_RW_LO (right), + #if (uECC_MAX_WORDS != uECC_MIN_WORDS) + [jump] REG_RW_LO (jump), + #endif + [carry] REG_WRITE_LO (carry), [left] REG_WRITE_LO (left_word), + [right] REG_WRITE_LO (right_word) + : + : "cc", "memory" + ); + return carry; +} +#define asm_add 1 + +uECC_VLI_API uECC_word_t uECC_vli_sub(uECC_word_t *result, + const uECC_word_t *left, + const uECC_word_t *right, + wordcount_t num_words) { +#if (uECC_MAX_WORDS != uECC_MIN_WORDS) + #if (uECC_PLATFORM == uECC_arm_thumb) || (uECC_PLATFORM == uECC_arm_thumb2) + uint32_t jump = (uECC_MAX_WORDS - num_words) * 4 * 2 + 1; + #else /* ARM */ + uint32_t jump = (uECC_MAX_WORDS - num_words) * 4 * 4; + #endif +#endif + uint32_t carry; + uint32_t left_word; + uint32_t right_word; + + __asm__ volatile ( + ".syntax unified \n\t" + "movs %[carry], #0 \n\t" + #if (uECC_MAX_WORDS != uECC_MIN_WORDS) + "adr %[left], 1f \n\t" + ".align 4 \n\t" + "adds %[jump], %[left] \n\t" + #endif + + "ldmia %[lptr]!, {%[left]} \n\t" + "ldmia %[rptr]!, {%[right]} \n\t" + "subs %[left], %[right] \n\t" + "stmia %[dptr]!, {%[left]} \n\t" + + #if (uECC_MAX_WORDS != uECC_MIN_WORDS) + "bx %[jump] \n\t" + #endif + "1: \n\t" + REPEAT(DEC(uECC_MAX_WORDS), + "ldmia %[lptr]!, {%[left]} \n\t" + "ldmia %[rptr]!, {%[right]} \n\t" + "sbcs %[left], %[right] \n\t" + "stmia %[dptr]!, {%[left]} \n\t") + + "adcs %[carry], %[carry] \n\t" + RESUME_SYNTAX + : [dptr] REG_RW_LO (result), [lptr] REG_RW_LO (left), [rptr] REG_RW_LO (right), + #if (uECC_MAX_WORDS != uECC_MIN_WORDS) + [jump] REG_RW_LO (jump), + #endif + [carry] REG_WRITE_LO (carry), [left] REG_WRITE_LO (left_word), + [right] REG_WRITE_LO (right_word) + : + : "cc", "memory" + ); + return !carry; /* Note that on ARM, carry flag set means "no borrow" when subtracting + (for some reason...) */ +} +#define asm_sub 1 + +#endif /* (uECC_OPTIMIZATION_LEVEL >= 2) */ + +/* ---- "Small" implementations ---- */ + +#if !asm_mult +uECC_VLI_API void uECC_vli_mult(uECC_word_t *result, + const uECC_word_t *left, + const uECC_word_t *right, + wordcount_t num_words) { + uint32_t c0 = 0; + uint32_t c1 = 0; + uint32_t c2 = 0; + uint32_t k = 0; + uint32_t i; + uint32_t t0, t1; + + __asm__ volatile ( + ".syntax unified \n\t" + + "1: \n\t" /* outer loop (k < num_words) */ + "movs %[i], #0 \n\t" /* i = 0 */ + "b 3f \n\t" + + "2: \n\t" /* outer loop (k >= num_words) */ + "movs %[i], %[k] \n\t" /* i = k */ + "subs %[i], %[last_word] \n\t" /* i = k - (num_words - 1) (times 4) */ + + "3: \n\t" /* inner loop */ + "subs %[t0], %[k], %[i] \n\t" /* t0 = k-i */ + + "ldr %[t1], [%[right], %[t0]] \n\t" /* t1 = right[k - i] */ + "ldr %[t0], [%[left], %[i]] \n\t" /* t0 = left[i] */ + + "umull %[t0], %[t1], %[t0], %[t1] \n\t" /* (t0, t1) = left[i] * right[k - i] */ + + "adds %[c0], %[c0], %[t0] \n\t" /* add low word to c0 */ + "adcs %[c1], %[c1], %[t1] \n\t" /* add high word to c1, including carry */ + "adcs %[c2], %[c2], #0 \n\t" /* add carry to c2 */ + + "adds %[i], #4 \n\t" /* i += 4 */ + "cmp %[i], %[last_word] \n\t" /* i > (num_words - 1) (times 4)? */ + "bgt 4f \n\t" /* if so, exit the loop */ + "cmp %[i], %[k] \n\t" /* i <= k? */ + "ble 3b \n\t" /* if so, continue looping */ + + "4: \n\t" /* end inner loop */ + + "str %[c0], [%[result], %[k]] \n\t" /* result[k] = c0 */ + "mov %[c0], %[c1] \n\t" /* c0 = c1 */ + "mov %[c1], %[c2] \n\t" /* c1 = c2 */ + "movs %[c2], #0 \n\t" /* c2 = 0 */ + "adds %[k], #4 \n\t" /* k += 4 */ + "cmp %[k], %[last_word] \n\t" /* k <= (num_words - 1) (times 4) ? */ + "ble 1b \n\t" /* if so, loop back, start with i = 0 */ + "cmp %[k], %[last_word], lsl #1 \n\t" /* k <= (num_words * 2 - 2) (times 4) ? */ + "ble 2b \n\t" /* if so, loop back, start with i = (k + 1) - num_words */ + /* end outer loop */ + + "str %[c0], [%[result], %[k]] \n\t" /* result[num_words * 2 - 1] = c0 */ + RESUME_SYNTAX + : [c0] "+r" (c0), [c1] "+r" (c1), [c2] "+r" (c2), + [k] "+r" (k), [i] "=&r" (i), [t0] "=&r" (t0), [t1] "=&r" (t1) + : [result] "r" (result), [left] "r" (left), [right] "r" (right), + [last_word] "r" ((num_words - 1) * 4) + : "cc", "memory" + ); +} +#define asm_mult 1 +#endif + +#endif /* _UECC_ASM_ARM_H_ */ diff --git a/Inc/uECC_curve-specific.inc b/Inc/uECC_curve-specific.inc new file mode 100644 index 0000000..b450a8e --- /dev/null +++ b/Inc/uECC_curve-specific.inc @@ -0,0 +1,138 @@ +/* Copyright 2015, Kenneth MacKay. Licensed under the BSD 2-clause license. */ + +#ifndef _UECC_CURVE_SPECIFIC_H_ +#define _UECC_CURVE_SPECIFIC_H_ + +#define num_bytes_secp256k1 32 +#define num_words_secp256k1 8 + +#define BYTES_TO_WORDS_8(a, b, c, d, e, f, g, h) 0x##d##c##b##a, 0x##h##g##f##e +#define BYTES_TO_WORDS_4(a, b, c, d) 0x##d##c##b##a + +static void double_jacobian_secp256k1(uECC_word_t * X1, + uECC_word_t * Y1, + uECC_word_t * Z1, + uECC_Curve curve); +static void x_side_secp256k1(uECC_word_t *result, const uECC_word_t *x, uECC_Curve curve); +static void vli_mmod_fast_secp256k1(uECC_word_t *result, uECC_word_t *product); + +static const struct uECC_Curve_t curve_secp256k1 = { + num_words_secp256k1, + num_bytes_secp256k1, + 256, /* num_n_bits */ + { BYTES_TO_WORDS_8(2F, FC, FF, FF, FE, FF, FF, FF), + BYTES_TO_WORDS_8(FF, FF, FF, FF, FF, FF, FF, FF), + BYTES_TO_WORDS_8(FF, FF, FF, FF, FF, FF, FF, FF), + BYTES_TO_WORDS_8(FF, FF, FF, FF, FF, FF, FF, FF) }, + { BYTES_TO_WORDS_8(41, 41, 36, D0, 8C, 5E, D2, BF), + BYTES_TO_WORDS_8(3B, A0, 48, AF, E6, DC, AE, BA), + BYTES_TO_WORDS_8(FE, FF, FF, FF, FF, FF, FF, FF), + BYTES_TO_WORDS_8(FF, FF, FF, FF, FF, FF, FF, FF) }, + { BYTES_TO_WORDS_8(98, 17, F8, 16, 5B, 81, F2, 59), + BYTES_TO_WORDS_8(D9, 28, CE, 2D, DB, FC, 9B, 02), + BYTES_TO_WORDS_8(07, 0B, 87, CE, 95, 62, A0, 55), + BYTES_TO_WORDS_8(AC, BB, DC, F9, 7E, 66, BE, 79), + + BYTES_TO_WORDS_8(B8, D4, 10, FB, 8F, D0, 47, 9C), + BYTES_TO_WORDS_8(19, 54, 85, A6, 48, B4, 17, FD), + BYTES_TO_WORDS_8(A8, 08, 11, 0E, FC, FB, A4, 5D), + BYTES_TO_WORDS_8(65, C4, A3, 26, 77, DA, 3A, 48) }, + { BYTES_TO_WORDS_8(07, 00, 00, 00, 00, 00, 00, 00), + BYTES_TO_WORDS_8(00, 00, 00, 00, 00, 00, 00, 00), + BYTES_TO_WORDS_8(00, 00, 00, 00, 00, 00, 00, 00), + BYTES_TO_WORDS_8(00, 00, 00, 00, 00, 00, 00, 00) }, + &double_jacobian_secp256k1, + &x_side_secp256k1, + &vli_mmod_fast_secp256k1 +}; + +uECC_Curve uECC_secp256k1(void) { return &curve_secp256k1; } + + +/* Double in place */ +static void double_jacobian_secp256k1(uECC_word_t * X1, + uECC_word_t * Y1, + uECC_word_t * Z1, + uECC_Curve curve) { + /* t1 = X, t2 = Y, t3 = Z */ + uECC_word_t t4[num_words_secp256k1]; + uECC_word_t t5[num_words_secp256k1]; + + if (uECC_vli_isZero(Z1, num_words_secp256k1)) { + return; + } + + uECC_vli_modSquare_fast(t5, Y1, curve); /* t5 = y1^2 */ + uECC_vli_modMult_fast(t4, X1, t5, curve); /* t4 = x1*y1^2 = A */ + uECC_vli_modSquare_fast(X1, X1, curve); /* t1 = x1^2 */ + uECC_vli_modSquare_fast(t5, t5, curve); /* t5 = y1^4 */ + uECC_vli_modMult_fast(Z1, Y1, Z1, curve); /* t3 = y1*z1 = z3 */ + + uECC_vli_modAdd(Y1, X1, X1, curve->p, num_words_secp256k1); /* t2 = 2*x1^2 */ + uECC_vli_modAdd(Y1, Y1, X1, curve->p, num_words_secp256k1); /* t2 = 3*x1^2 */ + if (uECC_vli_testBit(Y1, 0)) { + uECC_word_t carry = uECC_vli_add(Y1, Y1, curve->p, num_words_secp256k1); + uECC_vli_rshift1(Y1, num_words_secp256k1); + Y1[num_words_secp256k1 - 1] |= carry << (uECC_WORD_BITS - 1); + } else { + uECC_vli_rshift1(Y1, num_words_secp256k1); + } + /* t2 = 3/2*(x1^2) = B */ + + uECC_vli_modSquare_fast(X1, Y1, curve); /* t1 = B^2 */ + uECC_vli_modSub(X1, X1, t4, curve->p, num_words_secp256k1); /* t1 = B^2 - A */ + uECC_vli_modSub(X1, X1, t4, curve->p, num_words_secp256k1); /* t1 = B^2 - 2A = x3 */ + + uECC_vli_modSub(t4, t4, X1, curve->p, num_words_secp256k1); /* t4 = A - x3 */ + uECC_vli_modMult_fast(Y1, Y1, t4, curve); /* t2 = B * (A - x3) */ + uECC_vli_modSub(Y1, Y1, t5, curve->p, num_words_secp256k1); /* t2 = B * (A - x3) - y1^4 = y3 */ +} + +/* Computes result = x^3 + b. result must not overlap x. */ +static void x_side_secp256k1(uECC_word_t *result, const uECC_word_t *x, uECC_Curve curve) { + uECC_vli_modSquare_fast(result, x, curve); /* r = x^2 */ + uECC_vli_modMult_fast(result, result, x, curve); /* r = x^3 */ + uECC_vli_modAdd(result, result, curve->b, curve->p, num_words_secp256k1); /* r = x^3 + b */ +} + +static void omega_mult_secp256k1(uECC_word_t *result, const uECC_word_t *right); +static void vli_mmod_fast_secp256k1(uECC_word_t *result, uECC_word_t *product) { + uECC_word_t tmp[2 * num_words_secp256k1]; + uECC_word_t carry; + + uECC_vli_clear(tmp, num_words_secp256k1); + uECC_vli_clear(tmp + num_words_secp256k1, num_words_secp256k1); + + omega_mult_secp256k1(tmp, product + num_words_secp256k1); /* (Rq, q) = q * c */ + + carry = uECC_vli_add(result, product, tmp, num_words_secp256k1); /* (C, r) = r + q */ + uECC_vli_clear(product, num_words_secp256k1); + omega_mult_secp256k1(product, tmp + num_words_secp256k1); /* Rq*c */ + carry += uECC_vli_add(result, result, product, num_words_secp256k1); /* (C1, r) = r + Rq*c */ + + while (carry > 0) { + --carry; + uECC_vli_sub(result, result, curve_secp256k1.p, num_words_secp256k1); + } + if (uECC_vli_cmp_unsafe(result, curve_secp256k1.p, num_words_secp256k1) > 0) { + uECC_vli_sub(result, result, curve_secp256k1.p, num_words_secp256k1); + } +} + +static void omega_mult_secp256k1(uint32_t * result, const uint32_t * right) { + /* Multiply by (2^9 + 2^8 + 2^7 + 2^6 + 2^4 + 1). */ + uint32_t carry = 0; + wordcount_t k; + + for (k = 0; k < num_words_secp256k1; ++k) { + uint64_t p = (uint64_t)0x3D1 * right[k] + carry; + result[k] = (uint32_t) p; + carry = p >> 32; + } + result[num_words_secp256k1] = carry; + /* add the 2^32 multiple */ + result[1 + num_words_secp256k1] = + uECC_vli_add(result + 1, result + 1, right, num_words_secp256k1); +} + +#endif /* _UECC_CURVE_SPECIFIC_H_ */ diff --git a/Inc/uECC_types.h b/Inc/uECC_types.h new file mode 100644 index 0000000..ace9bbe --- /dev/null +++ b/Inc/uECC_types.h @@ -0,0 +1,78 @@ +/* Copyright 2015, Kenneth MacKay. Licensed under the BSD 2-clause license. */ + +#ifndef _UECC_TYPES_H_ +#define _UECC_TYPES_H_ + +#define uECC_PLATFORM uECC_arm_thumb2 + +#ifndef uECC_PLATFORM + #if __AVR__ + #define uECC_PLATFORM uECC_avr + #elif defined(__thumb2__) || defined(_M_ARMT) /* I think MSVC only supports Thumb-2 targets */ + #define uECC_PLATFORM uECC_arm_thumb2 + #elif defined(__thumb__) + #define uECC_PLATFORM uECC_arm_thumb + #elif defined(__arm__) || defined(_M_ARM) + #define uECC_PLATFORM uECC_arm + #elif defined(__aarch64__) + #define uECC_PLATFORM uECC_arm64 + #elif defined(__i386__) || defined(_M_IX86) || defined(_X86_) || defined(__I86__) + #define uECC_PLATFORM uECC_x86 + #elif defined(__amd64__) || defined(_M_X64) + #define uECC_PLATFORM uECC_x86_64 + #else + #define uECC_PLATFORM uECC_arch_other + #endif +#endif + +#ifndef uECC_ARM_USE_UMAAL + #if (uECC_PLATFORM == uECC_arm) && (__ARM_ARCH >= 6) + #define uECC_ARM_USE_UMAAL 1 + #elif (uECC_PLATFORM == uECC_arm_thumb2) && (__ARM_ARCH >= 6) && !__ARM_ARCH_7M__ + #define uECC_ARM_USE_UMAAL 1 + #else + #define uECC_ARM_USE_UMAAL 0 + #endif +#endif + +#ifndef uECC_WORD_SIZE + #if uECC_PLATFORM == uECC_avr + #define uECC_WORD_SIZE 1 + #elif (uECC_PLATFORM == uECC_x86_64 || uECC_PLATFORM == uECC_arm64) + #define uECC_WORD_SIZE 8 + #else + #define uECC_WORD_SIZE 4 + #endif +#endif + +#if (uECC_WORD_SIZE != 1) && (uECC_WORD_SIZE != 4) && (uECC_WORD_SIZE != 8) + #error "Unsupported value for uECC_WORD_SIZE" +#endif + +#if ((uECC_PLATFORM == uECC_avr) && (uECC_WORD_SIZE != 1)) + #pragma message ("uECC_WORD_SIZE must be 1 for AVR") + #undef uECC_WORD_SIZE + #define uECC_WORD_SIZE 1 +#endif + +#if ((uECC_PLATFORM == uECC_arm || uECC_PLATFORM == uECC_arm_thumb || \ + uECC_PLATFORM == uECC_arm_thumb2) && \ + (uECC_WORD_SIZE != 4)) + #pragma message ("uECC_WORD_SIZE must be 4 for ARM") + #undef uECC_WORD_SIZE + #define uECC_WORD_SIZE 4 +#endif + +typedef int8_t wordcount_t; +typedef int16_t bitcount_t; +typedef int8_t cmpresult_t; + +typedef uint32_t uECC_word_t; +typedef uint64_t uECC_dword_t; + +#define HIGH_BIT_SET 0x80000000 +#define uECC_WORD_BITS 32 +#define uECC_WORD_BITS_SHIFT 5 +#define uECC_WORD_BITS_MASK 0x01F + +#endif /* _UECC_TYPES_H_ */ diff --git a/Src/main.c b/Src/main.c index a530400..fec9e5a 100644 --- a/Src/main.c +++ b/Src/main.c @@ -1,6 +1,9 @@ #include "main.h" #include "flash.h" #include "sha256.h" +#include "uECC.h" + +uint8_t fw_public_key[] = { 0xfe, 0xcb, 0x28, 0xb9, 0x50, 0xdd, 0x8b, 0x2f, 0xc7, 0x34, 0xd3, 0x60, 0x5b, 0x1a, 0xc6, 0xed, 0x02, 0x50, 0xf2, 0x4a, 0xc4, 0x75, 0xd1, 0x28, 0x7f, 0x7c, 0xb5, 0xce, 0x61, 0xd6, 0x95, 0xb9, 0xb5, 0x27, 0x0b, 0x52, 0x77, 0x42, 0x4b, 0xf3, 0xb4, 0x3c, 0xef, 0xcb, 0x56, 0xd1, 0x98, 0x22, 0x11, 0xc2, 0xe5, 0xd3, 0xf0, 0x22, 0x87, 0xb9, 0xe8, 0x20, 0xdc, 0xee, 0x9f, 0xc2, 0xad, 0x22, }; int main(void) { if (!check_firmware(UPGRADE_FW_START)) { @@ -17,14 +20,21 @@ int check_firmware(uintptr_t addr) { return 1; } - cf_sha256_context ctx; - cf_sha256_init(&ctx); - cf_sha256_update(&ctx, UINT8_PTR(addr), UINT32_PTR(addr)[1]); - uint8_t hash[CF_SHA256_HASHSZ]; - cf_sha256_digest(&ctx, hash); - //TODO: verify signature!!! + uint32_t fw_size = UINT32_PTR(addr)[1]; - return 0; + if (fw_size > FIRMWARE_SIZE) { + return 1; + } + + uint8_t hash[CF_SHA256_HASHSZ]; + cf_sha256_context ctx; + uECC_Curve ec_curve = uECC_secp256k1(); + + cf_sha256_init(&ctx); + cf_sha256_update(&ctx, UINT8_PTR(addr + FIRMWARE_HEADER_SIZE), fw_size); + cf_sha256_digest(&ctx, hash); + + return uECC_verify(fw_public_key, hash, CF_SHA256_HASHSZ, UINT8_PTR(addr + 8), ec_curve) != 1; } diff --git a/Src/uECC.c b/Src/uECC.c new file mode 100644 index 0000000..f5afc17 --- /dev/null +++ b/Src/uECC.c @@ -0,0 +1,621 @@ +/* Copyright 2014, Kenneth MacKay. Licensed under the BSD 2-clause license. */ + +#include "uECC.h" +#include "uECC_types.h" + +#ifndef uECC_RNG_MAX_TRIES + #define uECC_RNG_MAX_TRIES 64 +#endif + +#if uECC_ENABLE_VLI_API + #define uECC_VLI_API +#else + #define uECC_VLI_API static +#endif + +#define CONCATX(a, ...) a ## __VA_ARGS__ +#define CONCAT(a, ...) CONCATX(a, __VA_ARGS__) + +#define STRX(a) #a +#define STR(a) STRX(a) + +#define EVAL(...) EVAL1(EVAL1(EVAL1(EVAL1(__VA_ARGS__)))) +#define EVAL1(...) EVAL2(EVAL2(EVAL2(EVAL2(__VA_ARGS__)))) +#define EVAL2(...) EVAL3(EVAL3(EVAL3(EVAL3(__VA_ARGS__)))) +#define EVAL3(...) EVAL4(EVAL4(EVAL4(EVAL4(__VA_ARGS__)))) +#define EVAL4(...) __VA_ARGS__ + +#define DEC_1 0 +#define DEC_2 1 +#define DEC_3 2 +#define DEC_4 3 +#define DEC_5 4 +#define DEC_6 5 +#define DEC_7 6 +#define DEC_8 7 +#define DEC_9 8 +#define DEC_10 9 +#define DEC_11 10 +#define DEC_12 11 +#define DEC_13 12 +#define DEC_14 13 +#define DEC_15 14 +#define DEC_16 15 +#define DEC_17 16 +#define DEC_18 17 +#define DEC_19 18 +#define DEC_20 19 +#define DEC_21 20 +#define DEC_22 21 +#define DEC_23 22 +#define DEC_24 23 +#define DEC_25 24 +#define DEC_26 25 +#define DEC_27 26 +#define DEC_28 27 +#define DEC_29 28 +#define DEC_30 29 +#define DEC_31 30 +#define DEC_32 31 + +#define DEC(N) CONCAT(DEC_, N) + +#define SECOND_ARG(_, val, ...) val +#define SOME_CHECK_0 ~, 0 +#define GET_SECOND_ARG(...) SECOND_ARG(__VA_ARGS__, SOME,) +#define SOME_OR_0(N) GET_SECOND_ARG(CONCAT(SOME_CHECK_, N)) + +#define EMPTY(...) +#define DEFER(...) __VA_ARGS__ EMPTY() + +#define REPEAT_NAME_0() REPEAT_0 +#define REPEAT_NAME_SOME() REPEAT_SOME +#define REPEAT_0(...) +#define REPEAT_SOME(N, stuff) DEFER(CONCAT(REPEAT_NAME_, SOME_OR_0(DEC(N))))()(DEC(N), stuff) stuff +#define REPEAT(N, stuff) EVAL(REPEAT_SOME(N, stuff)) + +#define REPEATM_NAME_0() REPEATM_0 +#define REPEATM_NAME_SOME() REPEATM_SOME +#define REPEATM_0(...) +#define REPEATM_SOME(N, macro) macro(N) \ + DEFER(CONCAT(REPEATM_NAME_, SOME_OR_0(DEC(N))))()(DEC(N), macro) +#define REPEATM(N, macro) EVAL(REPEATM_SOME(N, macro)) + + +#define uECC_MAX_WORDS 8 + +#define BITS_TO_WORDS(num_bits) ((num_bits + ((uECC_WORD_SIZE * 8) - 1)) / (uECC_WORD_SIZE * 8)) +#define BITS_TO_BYTES(num_bits) ((num_bits + 7) / 8) + +struct uECC_Curve_t { + wordcount_t num_words; + wordcount_t num_bytes; + bitcount_t num_n_bits; + uECC_word_t p[uECC_MAX_WORDS]; + uECC_word_t n[uECC_MAX_WORDS]; + uECC_word_t G[uECC_MAX_WORDS * 2]; + uECC_word_t b[uECC_MAX_WORDS]; + void (*double_jacobian)(uECC_word_t * X1, + uECC_word_t * Y1, + uECC_word_t * Z1, + uECC_Curve curve); + void (*x_side)(uECC_word_t *result, const uECC_word_t *x, uECC_Curve curve); + void (*mmod_fast)(uECC_word_t *result, uECC_word_t *product); +}; + +static cmpresult_t uECC_vli_cmp_unsafe(const uECC_word_t *left, + const uECC_word_t *right, + wordcount_t num_words); + +#if (uECC_PLATFORM == uECC_arm || uECC_PLATFORM == uECC_arm_thumb || \ + uECC_PLATFORM == uECC_arm_thumb2) + #include "uECC_asm_arm.inc" +#endif + +int uECC_curve_private_key_size(uECC_Curve curve) { + return BITS_TO_BYTES(curve->num_n_bits); +} + +int uECC_curve_public_key_size(uECC_Curve curve) { + return 2 * curve->num_bytes; +} + +#if !asm_clear +uECC_VLI_API void uECC_vli_clear(uECC_word_t *vli, wordcount_t num_words) { + wordcount_t i; + for (i = 0; i < num_words; ++i) { + vli[i] = 0; + } +} +#endif /* !asm_clear */ + +/* Constant-time comparison to zero - secure way to compare long integers */ +/* Returns 1 if vli == 0, 0 otherwise. */ +uECC_VLI_API uECC_word_t uECC_vli_isZero(const uECC_word_t *vli, wordcount_t num_words) { + uECC_word_t bits = 0; + wordcount_t i; + for (i = 0; i < num_words; ++i) { + bits |= vli[i]; + } + return (bits == 0); +} + +/* Returns nonzero if bit 'bit' of vli is set. */ +uECC_VLI_API uECC_word_t uECC_vli_testBit(const uECC_word_t *vli, bitcount_t bit) { + return (vli[bit >> uECC_WORD_BITS_SHIFT] & ((uECC_word_t)1 << (bit & uECC_WORD_BITS_MASK))); +} + +/* Counts the number of words in vli. */ +static wordcount_t vli_numDigits(const uECC_word_t *vli, const wordcount_t max_words) { + wordcount_t i; + /* Search from the end until we find a non-zero digit. + We do it in reverse because we expect that most digits will be nonzero. */ + for (i = max_words - 1; i >= 0 && vli[i] == 0; --i) { + } + + return (i + 1); +} + +/* Counts the number of bits required to represent vli. */ +uECC_VLI_API bitcount_t uECC_vli_numBits(const uECC_word_t *vli, const wordcount_t max_words) { + uECC_word_t i; + uECC_word_t digit; + + wordcount_t num_digits = vli_numDigits(vli, max_words); + if (num_digits == 0) { + return 0; + } + + digit = vli[num_digits - 1]; + for (i = 0; digit; ++i) { + digit >>= 1; + } + + return (((bitcount_t)(num_digits - 1) << uECC_WORD_BITS_SHIFT) + i); +} + +/* Sets dest = src. */ +#if !asm_set +uECC_VLI_API void uECC_vli_set(uECC_word_t *dest, const uECC_word_t *src, wordcount_t num_words) { + wordcount_t i; + for (i = 0; i < num_words; ++i) { + dest[i] = src[i]; + } +} +#endif /* !asm_set */ + +/* Returns sign of left - right. */ +static cmpresult_t uECC_vli_cmp_unsafe(const uECC_word_t *left, + const uECC_word_t *right, + wordcount_t num_words) { + wordcount_t i; + for (i = num_words - 1; i >= 0; --i) { + if (left[i] > right[i]) { + return 1; + } else if (left[i] < right[i]) { + return -1; + } + } + return 0; +} + +/* Constant-time comparison function - secure way to compare long integers */ +/* Returns one if left == right, zero otherwise. */ +uECC_VLI_API uECC_word_t uECC_vli_equal(const uECC_word_t *left, + const uECC_word_t *right, + wordcount_t num_words) { + uECC_word_t diff = 0; + wordcount_t i; + for (i = num_words - 1; i >= 0; --i) { + diff |= (left[i] ^ right[i]); + } + return (diff == 0); +} + +/* Computes vli = vli >> 1. */ +#if !asm_rshift1 +uECC_VLI_API void uECC_vli_rshift1(uECC_word_t *vli, wordcount_t num_words) { + uECC_word_t *end = vli; + uECC_word_t carry = 0; + + vli += num_words; + while (vli-- > end) { + uECC_word_t temp = *vli; + *vli = (temp >> 1) | carry; + carry = temp << (uECC_WORD_BITS - 1); + } +} +#endif /* !asm_rshift1 */ + +/* Computes result = (left + right) % mod. + Assumes that left < mod and right < mod, and that result does not overlap mod. */ +uECC_VLI_API void uECC_vli_modAdd(uECC_word_t *result, + const uECC_word_t *left, + const uECC_word_t *right, + const uECC_word_t *mod, + wordcount_t num_words) { + uECC_word_t carry = uECC_vli_add(result, left, right, num_words); + if (carry || uECC_vli_cmp_unsafe(mod, result, num_words) != 1) { + /* result > mod (result = mod + remainder), so subtract mod to get remainder. */ + uECC_vli_sub(result, result, mod, num_words); + } +} + +/* Computes result = (left - right) % mod. + Assumes that left < mod and right < mod, and that result does not overlap mod. */ +uECC_VLI_API void uECC_vli_modSub(uECC_word_t *result, + const uECC_word_t *left, + const uECC_word_t *right, + const uECC_word_t *mod, + wordcount_t num_words) { + uECC_word_t l_borrow = uECC_vli_sub(result, left, right, num_words); + if (l_borrow) { + /* In this case, result == -diff == (max int) - diff. Since -x % d == d - x, + we can get the correct result from result + mod (with overflow). */ + uECC_vli_add(result, result, mod, num_words); + } +} + +/* Computes result = product % mod, where product is 2N words long. */ +/* Currently only designed to work for curve_p or curve_n. */ +uECC_VLI_API void uECC_vli_mmod(uECC_word_t *result, + uECC_word_t *product, + const uECC_word_t *mod, + wordcount_t num_words) { + uECC_word_t mod_multiple[2 * uECC_MAX_WORDS]; + uECC_word_t tmp[2 * uECC_MAX_WORDS]; + uECC_word_t *v[2] = {tmp, product}; + uECC_word_t index; + + /* Shift mod so its highest set bit is at the maximum position. */ + bitcount_t shift = (num_words * 2 * uECC_WORD_BITS) - uECC_vli_numBits(mod, num_words); + wordcount_t word_shift = shift / uECC_WORD_BITS; + wordcount_t bit_shift = shift % uECC_WORD_BITS; + uECC_word_t carry = 0; + uECC_vli_clear(mod_multiple, word_shift); + if (bit_shift > 0) { + for(index = 0; index < (uECC_word_t)num_words; ++index) { + mod_multiple[word_shift + index] = (mod[index] << bit_shift) | carry; + carry = mod[index] >> (uECC_WORD_BITS - bit_shift); + } + } else { + uECC_vli_set(mod_multiple + word_shift, mod, num_words); + } + + for (index = 1; shift >= 0; --shift) { + uECC_word_t borrow = 0; + wordcount_t i; + for (i = 0; i < num_words * 2; ++i) { + uECC_word_t diff = v[index][i] - mod_multiple[i] - borrow; + if (diff != v[index][i]) { + borrow = (diff > v[index][i]); + } + v[1 - index][i] = diff; + } + index = !(index ^ borrow); /* Swap the index if there was no borrow */ + uECC_vli_rshift1(mod_multiple, num_words); + mod_multiple[num_words - 1] |= mod_multiple[num_words] << (uECC_WORD_BITS - 1); + uECC_vli_rshift1(mod_multiple + num_words, num_words); + } + uECC_vli_set(result, v[index], num_words); +} + +/* Computes result = (left * right) % mod. */ +uECC_VLI_API void uECC_vli_modMult(uECC_word_t *result, + const uECC_word_t *left, + const uECC_word_t *right, + const uECC_word_t *mod, + wordcount_t num_words) { + uECC_word_t product[2 * uECC_MAX_WORDS]; + uECC_vli_mult(product, left, right, num_words); + uECC_vli_mmod(result, product, mod, num_words); +} + +uECC_VLI_API void uECC_vli_modMult_fast(uECC_word_t *result, + const uECC_word_t *left, + const uECC_word_t *right, + uECC_Curve curve) { + uECC_word_t product[2 * uECC_MAX_WORDS]; + uECC_vli_mult(product, left, right, curve->num_words); + curve->mmod_fast(result, product); +} + +uECC_VLI_API void uECC_vli_modSquare_fast(uECC_word_t *result, + const uECC_word_t *left, + uECC_Curve curve) { + uECC_vli_modMult_fast(result, left, left, curve); +} + +#define EVEN(vli) (!(vli[0] & 1)) +static void vli_modInv_update(uECC_word_t *uv, + const uECC_word_t *mod, + wordcount_t num_words) { + uECC_word_t carry = 0; + if (!EVEN(uv)) { + carry = uECC_vli_add(uv, uv, mod, num_words); + } + uECC_vli_rshift1(uv, num_words); + if (carry) { + uv[num_words - 1] |= HIGH_BIT_SET; + } +} + +/* Computes result = (1 / input) % mod. All VLIs are the same size. + See "From Euclid's GCD to Montgomery Multiplication to the Great Divide" */ +uECC_VLI_API void uECC_vli_modInv(uECC_word_t *result, + const uECC_word_t *input, + const uECC_word_t *mod, + wordcount_t num_words) { + uECC_word_t a[uECC_MAX_WORDS], b[uECC_MAX_WORDS], u[uECC_MAX_WORDS], v[uECC_MAX_WORDS]; + cmpresult_t cmpResult; + + if (uECC_vli_isZero(input, num_words)) { + uECC_vli_clear(result, num_words); + return; + } + + uECC_vli_set(a, input, num_words); + uECC_vli_set(b, mod, num_words); + uECC_vli_clear(u, num_words); + u[0] = 1; + uECC_vli_clear(v, num_words); + while ((cmpResult = uECC_vli_cmp_unsafe(a, b, num_words)) != 0) { + if (EVEN(a)) { + uECC_vli_rshift1(a, num_words); + vli_modInv_update(u, mod, num_words); + } else if (EVEN(b)) { + uECC_vli_rshift1(b, num_words); + vli_modInv_update(v, mod, num_words); + } else if (cmpResult > 0) { + uECC_vli_sub(a, a, b, num_words); + uECC_vli_rshift1(a, num_words); + if (uECC_vli_cmp_unsafe(u, v, num_words) < 0) { + uECC_vli_add(u, u, mod, num_words); + } + uECC_vli_sub(u, u, v, num_words); + vli_modInv_update(u, mod, num_words); + } else { + uECC_vli_sub(b, b, a, num_words); + uECC_vli_rshift1(b, num_words); + if (uECC_vli_cmp_unsafe(v, u, num_words) < 0) { + uECC_vli_add(v, v, mod, num_words); + } + uECC_vli_sub(v, v, u, num_words); + vli_modInv_update(v, mod, num_words); + } + } + uECC_vli_set(result, u, num_words); +} + +/* ------ Point operations ------ */ + +#include "uECC_curve-specific.inc" + +/* Returns 1 if 'point' is the point at infinity, 0 otherwise. */ +#define EccPoint_isZero(point, curve) uECC_vli_isZero((point), (curve)->num_words * 2) + +/* Point multiplication algorithm using Montgomery's ladder with co-Z coordinates. +From http://eprint.iacr.org/2011/338.pdf +*/ + +/* Modify (x1, y1) => (x1 * z^2, y1 * z^3) */ +static void apply_z(uECC_word_t * X1, + uECC_word_t * Y1, + const uECC_word_t * const Z, + uECC_Curve curve) { + uECC_word_t t1[uECC_MAX_WORDS]; + + uECC_vli_modSquare_fast(t1, Z, curve); /* z^2 */ + uECC_vli_modMult_fast(X1, X1, t1, curve); /* x1 * z^2 */ + uECC_vli_modMult_fast(t1, t1, Z, curve); /* z^3 */ + uECC_vli_modMult_fast(Y1, Y1, t1, curve); /* y1 * z^3 */ +} + +/* Input P = (x1, y1, Z), Q = (x2, y2, Z) + Output P' = (x1', y1', Z3), P + Q = (x3, y3, Z3) + or P => P', Q => P + Q +*/ +static void XYcZ_add(uECC_word_t * X1, + uECC_word_t * Y1, + uECC_word_t * X2, + uECC_word_t * Y2, + uECC_Curve curve) { + /* t1 = X1, t2 = Y1, t3 = X2, t4 = Y2 */ + uECC_word_t t5[uECC_MAX_WORDS]; + wordcount_t num_words = curve->num_words; + + uECC_vli_modSub(t5, X2, X1, curve->p, num_words); /* t5 = x2 - x1 */ + uECC_vli_modSquare_fast(t5, t5, curve); /* t5 = (x2 - x1)^2 = A */ + uECC_vli_modMult_fast(X1, X1, t5, curve); /* t1 = x1*A = B */ + uECC_vli_modMult_fast(X2, X2, t5, curve); /* t3 = x2*A = C */ + uECC_vli_modSub(Y2, Y2, Y1, curve->p, num_words); /* t4 = y2 - y1 */ + uECC_vli_modSquare_fast(t5, Y2, curve); /* t5 = (y2 - y1)^2 = D */ + + uECC_vli_modSub(t5, t5, X1, curve->p, num_words); /* t5 = D - B */ + uECC_vli_modSub(t5, t5, X2, curve->p, num_words); /* t5 = D - B - C = x3 */ + uECC_vli_modSub(X2, X2, X1, curve->p, num_words); /* t3 = C - B */ + uECC_vli_modMult_fast(Y1, Y1, X2, curve); /* t2 = y1*(C - B) */ + uECC_vli_modSub(X2, X1, t5, curve->p, num_words); /* t3 = B - x3 */ + uECC_vli_modMult_fast(Y2, Y2, X2, curve); /* t4 = (y2 - y1)*(B - x3) */ + uECC_vli_modSub(Y2, Y2, Y1, curve->p, num_words); /* t4 = y3 */ + + uECC_vli_set(X2, t5, num_words); +} + +uECC_VLI_API void uECC_vli_bytesToNative(uECC_word_t *native, + const uint8_t *bytes, + int num_bytes) { + wordcount_t i; + uECC_vli_clear(native, (num_bytes + (uECC_WORD_SIZE - 1)) / uECC_WORD_SIZE); + for (i = 0; i < num_bytes; ++i) { + unsigned b = num_bytes - 1 - i; + native[b / uECC_WORD_SIZE] |= + (uECC_word_t)bytes[i] << (8 * (b % uECC_WORD_SIZE)); + } +} + +int uECC_valid_point(const uECC_word_t *point, uECC_Curve curve) { + uECC_word_t tmp1[uECC_MAX_WORDS]; + uECC_word_t tmp2[uECC_MAX_WORDS]; + wordcount_t num_words = curve->num_words; + + /* The point at infinity is invalid. */ + if (EccPoint_isZero(point, curve)) { + return 0; + } + + /* x and y must be smaller than p. */ + if (uECC_vli_cmp_unsafe(curve->p, point, num_words) != 1 || + uECC_vli_cmp_unsafe(curve->p, point + num_words, num_words) != 1) { + return 0; + } + + uECC_vli_modSquare_fast(tmp1, point + num_words, curve); + curve->x_side(tmp2, point, curve); /* tmp2 = x^3 + ax + b */ + + /* Make sure that y^2 == x^3 + ax + b */ + return (int)(uECC_vli_equal(tmp1, tmp2, num_words)); +} + +/* -------- ECDSA code -------- */ + +static void bits2int(uECC_word_t *native, + const uint8_t *bits, + unsigned bits_size, + uECC_Curve curve) { + unsigned num_n_bytes = BITS_TO_BYTES(curve->num_n_bits); + unsigned num_n_words = BITS_TO_WORDS(curve->num_n_bits); + int shift; + uECC_word_t carry; + uECC_word_t *ptr; + + if (bits_size > num_n_bytes) { + bits_size = num_n_bytes; + } + + uECC_vli_clear(native, num_n_words); + uECC_vli_bytesToNative(native, bits, bits_size); + + if (bits_size * 8 <= (unsigned)curve->num_n_bits) { + return; + } + shift = bits_size * 8 - curve->num_n_bits; + carry = 0; + ptr = native + num_n_words; + while (ptr-- > native) { + uECC_word_t temp = *ptr; + *ptr = (temp >> shift) | carry; + carry = temp << (uECC_WORD_BITS - shift); + } + + /* Reduce mod curve_n */ + if (uECC_vli_cmp_unsafe(curve->n, native, num_n_words) != 1) { + uECC_vli_sub(native, native, curve->n, num_n_words); + } +} + +static bitcount_t smax(bitcount_t a, bitcount_t b) { + return (a > b ? a : b); +} + +int uECC_verify(const uint8_t *public_key, + const uint8_t *message_hash, + unsigned hash_size, + const uint8_t *signature, + uECC_Curve curve) { + uECC_word_t u1[uECC_MAX_WORDS], u2[uECC_MAX_WORDS]; + uECC_word_t z[uECC_MAX_WORDS]; + uECC_word_t sum[uECC_MAX_WORDS * 2]; + uECC_word_t rx[uECC_MAX_WORDS]; + uECC_word_t ry[uECC_MAX_WORDS]; + uECC_word_t tx[uECC_MAX_WORDS]; + uECC_word_t ty[uECC_MAX_WORDS]; + uECC_word_t tz[uECC_MAX_WORDS]; + const uECC_word_t *points[4]; + const uECC_word_t *point; + bitcount_t num_bits; + bitcount_t i; + uECC_word_t _public[uECC_MAX_WORDS * 2]; + uECC_word_t r[uECC_MAX_WORDS], s[uECC_MAX_WORDS]; + wordcount_t num_words = curve->num_words; + wordcount_t num_n_words = BITS_TO_WORDS(curve->num_n_bits); + + rx[num_n_words - 1] = 0; + r[num_n_words - 1] = 0; + s[num_n_words - 1] = 0; + + uECC_vli_bytesToNative(_public, public_key, curve->num_bytes); + uECC_vli_bytesToNative(_public + num_words, public_key + curve->num_bytes, curve->num_bytes); + uECC_vli_bytesToNative(r, signature, curve->num_bytes); + uECC_vli_bytesToNative(s, signature + curve->num_bytes, curve->num_bytes); + + /* r, s must not be 0. */ + if (uECC_vli_isZero(r, num_words) || uECC_vli_isZero(s, num_words)) { + return 0; + } + + /* r, s must be < n. */ + if (uECC_vli_cmp_unsafe(curve->n, r, num_n_words) != 1 || + uECC_vli_cmp_unsafe(curve->n, s, num_n_words) != 1) { + return 0; + } + + /* Calculate u1 and u2. */ + uECC_vli_modInv(z, s, curve->n, num_n_words); /* z = 1/s */ + u1[num_n_words - 1] = 0; + bits2int(u1, message_hash, hash_size, curve); + uECC_vli_modMult(u1, u1, z, curve->n, num_n_words); /* u1 = e/s */ + uECC_vli_modMult(u2, r, z, curve->n, num_n_words); /* u2 = r/s */ + + /* Calculate sum = G + Q. */ + uECC_vli_set(sum, _public, num_words); + uECC_vli_set(sum + num_words, _public + num_words, num_words); + uECC_vli_set(tx, curve->G, num_words); + uECC_vli_set(ty, curve->G + num_words, num_words); + uECC_vli_modSub(z, sum, tx, curve->p, num_words); /* z = x2 - x1 */ + XYcZ_add(tx, ty, sum, sum + num_words, curve); + uECC_vli_modInv(z, z, curve->p, num_words); /* z = 1/z */ + apply_z(sum, sum + num_words, z, curve); + + /* Use Shamir's trick to calculate u1*G + u2*Q */ + points[0] = 0; + points[1] = curve->G; + points[2] = _public; + points[3] = sum; + num_bits = smax(uECC_vli_numBits(u1, num_n_words), + uECC_vli_numBits(u2, num_n_words)); + + point = points[(!!uECC_vli_testBit(u1, num_bits - 1)) | + ((!!uECC_vli_testBit(u2, num_bits - 1)) << 1)]; + uECC_vli_set(rx, point, num_words); + uECC_vli_set(ry, point + num_words, num_words); + uECC_vli_clear(z, num_words); + z[0] = 1; + + for (i = num_bits - 2; i >= 0; --i) { + uECC_word_t index; + curve->double_jacobian(rx, ry, z, curve); + + index = (!!uECC_vli_testBit(u1, i)) | ((!!uECC_vli_testBit(u2, i)) << 1); + point = points[index]; + if (point) { + uECC_vli_set(tx, point, num_words); + uECC_vli_set(ty, point + num_words, num_words); + apply_z(tx, ty, z, curve); + uECC_vli_modSub(tz, rx, tx, curve->p, num_words); /* Z = x2 - x1 */ + XYcZ_add(tx, ty, rx, ry, curve); + uECC_vli_modMult_fast(z, z, tz, curve); + } + } + + uECC_vli_modInv(z, z, curve->p, num_words); /* Z = 1/Z */ + apply_z(rx, ry, z, curve); + + /* v = x1 (mod n) */ + if (uECC_vli_cmp_unsafe(curve->n, rx, num_n_words) != 1) { + uECC_vli_sub(rx, rx, curve->n, num_n_words); + } + + /* Accept only if v == r. */ + return (int)(uECC_vli_equal(rx, r, num_words)); +} diff --git a/genkeys.py b/genkeys.py new file mode 100755 index 0000000..878b7e4 --- /dev/null +++ b/genkeys.py @@ -0,0 +1,18 @@ +#!/usr/bin/env python + +import ecdsa +import hashlib +import binascii +import re + +def main(): + sk = ecdsa.SigningKey.generate(curve=ecdsa.curves.SECP256k1, hashfunc=hashlib.sha256) + sk_pem = sk.to_pem() + vk = sk.get_verifying_key() + + vk_hex = binascii.hexlify(vk.to_string()) + vk_h = "uint8_t fw_public_key[] = { " + re.sub(r"(..)", "0x\\1, ", vk_hex) + "};" + open("secret_key.pem","w").write(sk.to_pem()) + open("public_key.h","w").write(vk_h) +if __name__ == '__main__': + main() \ No newline at end of file