xref: /btstack/3rd-party/micro-ecc/uECC.c (revision cd5f23a3250874824c01a2b3326a9522fea3f99f)
1 /* Copyright 2014, Kenneth MacKay. Licensed under the BSD 2-clause license. */
2 
3 #include "uECC.h"
4 
5 // NULL
6 #include "stddef.h"
7 
8 #ifndef uECC_PLATFORM
9     #if __AVR__
10         #define uECC_PLATFORM uECC_avr
11     #elif defined(__thumb2__) || defined(_M_ARMT) /* I think MSVC only supports Thumb-2 targets */
12         #define uECC_PLATFORM uECC_arm_thumb2
13     #elif defined(__thumb__)
14         #define uECC_PLATFORM uECC_arm_thumb
15     #elif defined(__arm__) || defined(_M_ARM)
16         #define uECC_PLATFORM uECC_arm
17     #elif defined(__i386__) || defined(_M_IX86) || defined(_X86_) || defined(__I86__)
18         #define uECC_PLATFORM uECC_x86
19     #elif defined(__amd64__) || defined(_M_X64)
20         #define uECC_PLATFORM uECC_x86_64
21     #else
22         #define uECC_PLATFORM uECC_arch_other
23     #endif
24 #endif
25 
26 #ifndef uECC_WORD_SIZE
27     #if uECC_PLATFORM == uECC_avr
28         #define uECC_WORD_SIZE 1
29     #elif (uECC_PLATFORM == uECC_x86_64)
30         #define uECC_WORD_SIZE 8
31     #else
32         #define uECC_WORD_SIZE 4
33     #endif
34 #endif
35 
36 #if (uECC_CURVE == uECC_secp160r1 || uECC_CURVE == uECC_secp224r1) && (uECC_WORD_SIZE == 8)
37     #undef uECC_WORD_SIZE
38     #define uECC_WORD_SIZE 4
39     #if (uECC_PLATFORM == uECC_x86_64)
40         #undef uECC_PLATFORM
41         #define uECC_PLATFORM uECC_x86
42     #endif
43 #endif
44 
45 #if (uECC_WORD_SIZE != 1) && (uECC_WORD_SIZE != 4) && (uECC_WORD_SIZE != 8)
46     #error "Unsupported value for uECC_WORD_SIZE"
47 #endif
48 
49 #if (uECC_ASM && (uECC_PLATFORM == uECC_avr) && (uECC_WORD_SIZE != 1))
50     #pragma message ("uECC_WORD_SIZE must be 1 when using AVR asm")
51     #undef uECC_WORD_SIZE
52     #define uECC_WORD_SIZE 1
53 #endif
54 
55 #if (uECC_ASM && \
56      (uECC_PLATFORM == uECC_arm || uECC_PLATFORM == uECC_arm_thumb) && \
57      (uECC_WORD_SIZE != 4))
58     #pragma message ("uECC_WORD_SIZE must be 4 when using ARM asm")
59     #undef uECC_WORD_SIZE
60     #define uECC_WORD_SIZE 4
61 #endif
62 
63 #if __STDC_VERSION__ >= 199901L
64     #define RESTRICT restrict
65 #else
66     #define RESTRICT
67 #endif
68 
69 #if defined(__SIZEOF_INT128__) || ((__clang_major__ * 100 + __clang_minor__) >= 302)
70     #define SUPPORTS_INT128 1
71 #else
72     #define SUPPORTS_INT128 0
73 #endif
74 
75 #define MAX_TRIES 64
76 
77 #if (uECC_WORD_SIZE == 1)
78 
79 typedef uint8_t uECC_word_t;
80 typedef uint16_t uECC_dword_t;
81 typedef uint8_t wordcount_t;
82 typedef int8_t swordcount_t;
83 typedef int16_t bitcount_t;
84 typedef int8_t cmpresult_t;
85 
86 #define HIGH_BIT_SET 0x80
87 #define uECC_WORD_BITS 8
88 #define uECC_WORD_BITS_SHIFT 3
89 #define uECC_WORD_BITS_MASK 0x07
90 
91 #define uECC_WORDS_1 20
92 #define uECC_WORDS_2 24
93 #define uECC_WORDS_3 32
94 #define uECC_WORDS_4 32
95 #define uECC_WORDS_5 28
96 
97 #define uECC_N_WORDS_1 21
98 #define uECC_N_WORDS_2 24
99 #define uECC_N_WORDS_3 32
100 #define uECC_N_WORDS_4 32
101 #define uECC_N_WORDS_5 28
102 
103 #define Curve_P_1 {0xFF, 0xFF, 0xFF, 0x7F, 0xFF, 0xFF, 0xFF, 0xFF, \
104                    0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, \
105                    0xFF, 0xFF, 0xFF, 0xFF}
106 #define Curve_P_2 {0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, \
107                    0xFE, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, \
108                    0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF}
109 #define Curve_P_3 {0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, \
110                    0xFF, 0xFF, 0xFF, 0xFF, 0x00, 0x00, 0x00, 0x00, \
111                    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, \
112                    0x01, 0x00, 0x00, 0x00, 0xFF, 0xFF, 0xFF, 0xFF}
113 #define Curve_P_4 {0x2F, 0xFC, 0xFF, 0xFF, 0xFE, 0xFF, 0xFF, 0xFF, \
114                    0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, \
115                    0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, \
116                    0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF}
117 #define Curve_P_5 {0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, \
118                    0x00, 0x00, 0x00, 0x00, 0xFF, 0xFF, 0xFF, 0xFF, \
119                    0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, \
120                    0xFF, 0xFF, 0xFF, 0xFF}
121 
122 #define Curve_B_1 {0x45, 0xFA, 0x65, 0xC5, 0xAD, 0xD4, 0xD4, 0x81, \
123                    0x9F, 0xF8, 0xAC, 0x65, 0x8B, 0x7A, 0xBD, 0x54, \
124                    0xFC, 0xBE, 0x97, 0x1C}
125 #define Curve_B_2 {0xB1, 0xB9, 0x46, 0xC1, 0xEC, 0xDE, 0xB8, 0xFE, \
126                    0x49, 0x30, 0x24, 0x72, 0xAB, 0xE9, 0xA7, 0x0F, \
127                    0xE7, 0x80, 0x9C, 0xE5, 0x19, 0x05, 0x21, 0x64}
128 #define Curve_B_3 {0x4B, 0x60, 0xD2, 0x27, 0x3E, 0x3C, 0xCE, 0x3B, \
129                    0xF6, 0xB0, 0x53, 0xCC, 0xB0, 0x06, 0x1D, 0x65, \
130                    0xBC, 0x86, 0x98, 0x76, 0x55, 0xBD, 0xEB, 0xB3, \
131                    0xE7, 0x93, 0x3A, 0xAA, 0xD8, 0x35, 0xC6, 0x5A}
132 #define Curve_B_4 {0x07, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, \
133                    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, \
134                    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, \
135                    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00}
136 #define Curve_B_5 {0xB4, 0xFF, 0x55, 0x23, 0x43, 0x39, 0x0B, 0x27, \
137                    0xBA, 0xD8, 0xBF, 0xD7, 0xB7, 0xB0, 0x44, 0x50, \
138                    0x56, 0x32, 0x41, 0xF5, 0xAB, 0xB3, 0x04, 0x0C, \
139                    0x85, 0x0A, 0x05, 0xB4}
140 
141 #define Curve_G_1 { \
142     {0x82, 0xFC, 0xCB, 0x13, 0xB9, 0x8B, 0xC3, 0x68, \
143         0x89, 0x69, 0x64, 0x46, 0x28, 0x73, 0xF5, 0x8E, \
144         0x68, 0xB5, 0x96, 0x4A}, \
145     {0x32, 0xFB, 0xC5, 0x7A, 0x37, 0x51, 0x23, 0x04, \
146         0x12, 0xC9, 0xDC, 0x59, 0x7D, 0x94, 0x68, 0x31, \
147         0x55, 0x28, 0xA6, 0x23}}
148 
149 #define Curve_G_2 { \
150     {0x12, 0x10, 0xFF, 0x82, 0xFD, 0x0A, 0xFF, 0xF4, \
151         0x00, 0x88, 0xA1, 0x43, 0xEB, 0x20, 0xBF, 0x7C, \
152         0xF6, 0x90, 0x30, 0xB0, 0x0E, 0xA8, 0x8D, 0x18}, \
153     {0x11, 0x48, 0x79, 0x1E, 0xA1, 0x77, 0xF9, 0x73, \
154         0xD5, 0xCD, 0x24, 0x6B, 0xED, 0x11, 0x10, 0x63, \
155         0x78, 0xDA, 0xC8, 0xFF, 0x95, 0x2B, 0x19, 0x07}}
156 
157 #define Curve_G_3 { \
158     {0x96, 0xC2, 0x98, 0xD8, 0x45, 0x39, 0xA1, 0xF4, \
159         0xA0, 0x33, 0xEB, 0x2D, 0x81, 0x7D, 0x03, 0x77, \
160         0xF2, 0x40, 0xA4, 0x63, 0xE5, 0xE6, 0xBC, 0xF8, \
161         0x47, 0x42, 0x2C, 0xE1, 0xF2, 0xD1, 0x17, 0x6B}, \
162     {0xF5, 0x51, 0xBF, 0x37, 0x68, 0x40, 0xB6, 0xCB, \
163         0xCE, 0x5E, 0x31, 0x6B, 0x57, 0x33, 0xCE, 0x2B, \
164         0x16, 0x9E, 0x0F, 0x7C, 0x4A, 0xEB, 0xE7, 0x8E, \
165         0x9B, 0x7F, 0x1A, 0xFE, 0xE2, 0x42, 0xE3, 0x4F}}
166 
167 #define Curve_G_4 { \
168     {0x98, 0x17, 0xF8, 0x16, 0x5B, 0x81, 0xF2, 0x59, \
169         0xD9, 0x28, 0xCE, 0x2D, 0xDB, 0xFC, 0x9B, 0x02, \
170         0x07, 0x0B, 0x87, 0xCE, 0x95, 0x62, 0xA0, 0x55, \
171         0xAC, 0xBB, 0xDC, 0xF9, 0x7E, 0x66, 0xBE, 0x79}, \
172     {0xB8, 0xD4, 0x10, 0xFB, 0x8F, 0xD0, 0x47, 0x9C, \
173         0x19, 0x54, 0x85, 0xA6, 0x48, 0xB4, 0x17, 0xFD, \
174         0xA8, 0x08, 0x11, 0x0E, 0xFC, 0xFB, 0xA4, 0x5D, \
175         0x65, 0xC4, 0xA3, 0x26, 0x77, 0xDA, 0x3A, 0x48}}
176 
177 #define Curve_G_5 { \
178     {0x21, 0x1D, 0x5C, 0x11, 0xD6, 0x80, 0x32, 0x34, \
179         0x22, 0x11, 0xC2, 0x56, 0xD3, 0xC1, 0x03, 0x4A, \
180         0xB9, 0x90, 0x13, 0x32, 0x7F, 0xBF, 0xB4, 0x6B, \
181         0xBD, 0x0C, 0x0E, 0xB7}, \
182     {0x34, 0x7E, 0x00, 0x85, 0x99, 0x81, 0xD5, 0x44, \
183         0x64, 0x47, 0x07, 0x5A, 0xA0, 0x75, 0x43, 0xCD, \
184         0xE6, 0xDF, 0x22, 0x4C, 0xFB, 0x23, 0xF7, 0xB5, \
185         0x88, 0x63, 0x37, 0xBD}}
186 
187 #define Curve_N_1 {0x57, 0x22, 0x75, 0xCA, 0xD3, 0xAE, 0x27, 0xF9, \
188                    0xC8, 0xF4, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, \
189                    0x00, 0x00, 0x00, 0x00, 0x01}
190 #define Curve_N_2 {0x31, 0x28, 0xD2, 0xB4, 0xB1, 0xC9, 0x6B, 0x14, \
191                    0x36, 0xF8, 0xDE, 0x99, 0xFF, 0xFF, 0xFF, 0xFF, \
192                    0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF}
193 #define Curve_N_3 {0x51, 0x25, 0x63, 0xFC, 0xC2, 0xCA, 0xB9, 0xF3, \
194                    0x84, 0x9E, 0x17, 0xA7, 0xAD, 0xFA, 0xE6, 0xBC, \
195                    0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, \
196                    0x00, 0x00, 0x00, 0x00, 0xFF, 0xFF, 0xFF, 0xFF}
197 #define Curve_N_4 {0x41, 0x41, 0x36, 0xD0, 0x8C, 0x5E, 0xD2, 0xBF, \
198                    0x3B, 0xA0, 0x48, 0xAF, 0xE6, 0xDC, 0xAE, 0xBA, \
199                    0xFE, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, \
200                    0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF}
201 #define Curve_N_5 {0x3D, 0x2A, 0x5C, 0x5C, 0x45, 0x29, 0xDD, 0x13, \
202                    0x3E, 0xF0, 0xB8, 0xE0, 0xA2, 0x16, 0xFF, 0xFF, \
203                    0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, \
204                    0xFF, 0xFF, 0xFF, 0xFF}
205 
206 #elif (uECC_WORD_SIZE == 4)
207 
208 typedef uint32_t uECC_word_t;
209 typedef uint64_t uECC_dword_t;
210 typedef unsigned wordcount_t;
211 typedef int swordcount_t;
212 typedef int bitcount_t;
213 typedef int cmpresult_t;
214 
215 #define HIGH_BIT_SET 0x80000000
216 #define uECC_WORD_BITS 32
217 #define uECC_WORD_BITS_SHIFT 5
218 #define uECC_WORD_BITS_MASK 0x01F
219 
220 #define uECC_WORDS_1 5
221 #define uECC_WORDS_2 6
222 #define uECC_WORDS_3 8
223 #define uECC_WORDS_4 8
224 #define uECC_WORDS_5 7
225 
226 #define uECC_N_WORDS_1 6
227 #define uECC_N_WORDS_2 6
228 #define uECC_N_WORDS_3 8
229 #define uECC_N_WORDS_4 8
230 #define uECC_N_WORDS_5 7
231 
232 #define Curve_P_1 {0x7FFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF}
233 #define Curve_P_2 {0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFE, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF}
234 #define Curve_P_3 {0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0x00000000, \
235                    0x00000000, 0x00000000, 0x00000001, 0xFFFFFFFF}
236 #define Curve_P_4 {0xFFFFFC2F, 0xFFFFFFFE, 0xFFFFFFFF, 0xFFFFFFFF, \
237                    0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF}
238 #define Curve_P_5 {0x00000001, 0x00000000, 0x00000000, 0xFFFFFFFF, \
239                    0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF}
240 
241 #define Curve_B_1 {0xC565FA45, 0x81D4D4AD, 0x65ACF89F, 0x54BD7A8B, 0x1C97BEFC}
242 #define Curve_B_2 {0xC146B9B1, 0xFEB8DEEC, 0x72243049, 0x0FA7E9AB, 0xE59C80E7, 0x64210519}
243 #define Curve_B_3 {0x27D2604B, 0x3BCE3C3E, 0xCC53B0F6, 0x651D06B0, \
244                    0x769886BC, 0xB3EBBD55, 0xAA3A93E7, 0x5AC635D8}
245 #define Curve_B_4 {0x00000007, 0x00000000, 0x00000000, 0x00000000, \
246                    0x00000000, 0x00000000, 0x00000000, 0x00000000}
247 #define Curve_B_5 {0x2355FFB4, 0x270B3943, 0xD7BFD8BA, 0x5044B0B7, \
248                    0xF5413256, 0x0C04B3AB, 0xB4050A85}
249 
250 #define Curve_G_1 { \
251     {0x13CBFC82, 0x68C38BB9, 0x46646989, 0x8EF57328, 0x4A96B568}, \
252     {0x7AC5FB32, 0x04235137, 0x59DCC912, 0x3168947D, 0x23A62855}}
253 
254 #define Curve_G_2 { \
255     {0x82FF1012, 0xF4FF0AFD, 0x43A18800, 0x7CBF20EB, 0xB03090F6, 0x188DA80E}, \
256     {0x1E794811, 0x73F977A1, 0x6B24CDD5, 0x631011ED, 0xFFC8DA78, 0x07192B95}}
257 
258 #define Curve_G_3 { \
259     {0xD898C296, 0xF4A13945, 0x2DEB33A0, 0x77037D81,  \
260      0x63A440F2, 0xF8BCE6E5, 0xE12C4247, 0x6B17D1F2}, \
261     {0x37BF51F5, 0xCBB64068, 0x6B315ECE, 0x2BCE3357,  \
262      0x7C0F9E16, 0x8EE7EB4A, 0xFE1A7F9B, 0x4FE342E2}}
263 
264 #define Curve_G_4 { \
265     {0x16F81798, 0x59F2815B, 0x2DCE28D9, 0x029BFCDB,  \
266      0xCE870B07, 0x55A06295, 0xF9DCBBAC, 0x79BE667E}, \
267     {0xFB10D4B8, 0x9C47D08F, 0xA6855419, 0xFD17B448,  \
268      0x0E1108A8, 0x5DA4FBFC, 0x26A3C465, 0x483ADA77}}
269 
270 #define Curve_G_5 { \
271     {0x115C1D21, 0x343280D6, 0x56C21122, 0x4A03C1D3, \
272      0x321390B9, 0x6BB4BF7F, 0xB70E0CBD}, \
273     {0x85007E34, 0x44D58199, 0x5A074764, 0xCD4375A0, \
274      0x4C22DFE6, 0xB5F723FB, 0xBD376388}}
275 
276 #define Curve_N_1 {0xCA752257, 0xF927AED3, 0x0001F4C8, 0x00000000, 0x00000000, 0x00000001}
277 #define Curve_N_2 {0xB4D22831, 0x146BC9B1, 0x99DEF836, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF}
278 #define Curve_N_3 {0xFC632551, 0xF3B9CAC2, 0xA7179E84, 0xBCE6FAAD, \
279                    0xFFFFFFFF, 0xFFFFFFFF, 0x00000000, 0xFFFFFFFF}
280 #define Curve_N_4 {0xD0364141, 0xBFD25E8C, 0xAF48A03B, 0xBAAEDCE6, \
281                    0xFFFFFFFE, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF}
282 #define Curve_N_5 {0x5C5C2A3D, 0x13DD2945, 0xE0B8F03E, 0xFFFF16A2, \
283                    0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF}
284 
285 #elif (uECC_WORD_SIZE == 8)
286 
287 typedef uint64_t uECC_word_t;
288 #if SUPPORTS_INT128
289 typedef unsigned __int128 uECC_dword_t;
290 #endif
291 typedef unsigned wordcount_t;
292 typedef int swordcount_t;
293 typedef int bitcount_t;
294 typedef int cmpresult_t;
295 
296 #define HIGH_BIT_SET 0x8000000000000000ull
297 #define uECC_WORD_BITS 64
298 #define uECC_WORD_BITS_SHIFT 6
299 #define uECC_WORD_BITS_MASK 0x03F
300 
301 #define uECC_WORDS_1 3
302 #define uECC_WORDS_2 3
303 #define uECC_WORDS_3 4
304 #define uECC_WORDS_4 4
305 #define uECC_WORDS_5 4
306 
307 #define uECC_N_WORDS_1 3
308 #define uECC_N_WORDS_2 3
309 #define uECC_N_WORDS_3 4
310 #define uECC_N_WORDS_4 4
311 #define uECC_N_WORDS_5 4
312 
313 #define Curve_P_1 {0xFFFFFFFF7FFFFFFFull, 0xFFFFFFFFFFFFFFFFull, 0x00000000FFFFFFFFull}
314 #define Curve_P_2 {0xFFFFFFFFFFFFFFFFull, 0xFFFFFFFFFFFFFFFEull, 0xFFFFFFFFFFFFFFFFull}
315 #define Curve_P_3 {0xFFFFFFFFFFFFFFFFull, 0x00000000FFFFFFFFull, \
316                    0x0000000000000000ull, 0xFFFFFFFF00000001ull}
317 #define Curve_P_4 {0xFFFFFFFEFFFFFC2Full, 0xFFFFFFFFFFFFFFFFull, \
318                    0xFFFFFFFFFFFFFFFFull, 0xFFFFFFFFFFFFFFFFull}
319 #define Curve_P_5 {0x0000000000000001ull, 0xFFFFFFFF00000000ull, \
320                    0xFFFFFFFFFFFFFFFFull, 0x00000000FFFFFFFFull}
321 
322 #define Curve_B_1 {0x81D4D4ADC565FA45ull, 0x54BD7A8B65ACF89Full, 0x000000001C97BEFCull}
323 #define Curve_B_2 {0xFEB8DEECC146B9B1ull, 0x0FA7E9AB72243049ull, 0x64210519E59C80E7ull}
324 #define Curve_B_3 {0x3BCE3C3E27D2604Bull, 0x651D06B0CC53B0F6ull, \
325                    0xB3EBBD55769886BCull, 0x5AC635D8AA3A93E7ull}
326 #define Curve_B_4 {0x0000000000000007ull, 0x0000000000000000ull, \
327                    0x0000000000000000ull, 0x0000000000000000ull}
328 #define Curve_B_5 {0x270B39432355FFB4ull, 0x5044B0B7D7BFD8BAull, \
329                    0x0C04B3ABF5413256ull, 0x00000000B4050A85ull}
330 
331 #define Curve_G_1 { \
332     {0x68C38BB913CBFC82ull, 0x8EF5732846646989ull, 0x000000004A96B568ull}, \
333     {0x042351377AC5FB32ull, 0x3168947D59DCC912ull, 0x0000000023A62855ull}}
334 
335 #define Curve_G_2 { \
336     {0xF4FF0AFD82FF1012ull, 0x7CBF20EB43A18800ull, 0x188DA80EB03090F6ull}, \
337     {0x73F977A11E794811ull, 0x631011ED6B24CDD5ull, 0x07192B95FFC8DA78ull}}
338 
339 #define Curve_G_3 { \
340     {0xF4A13945D898C296ull, 0x77037D812DEB33A0ull, 0xF8BCE6E563A440F2ull, 0x6B17D1F2E12C4247ull}, \
341     {0xCBB6406837BF51F5ull, 0x2BCE33576B315ECEull, 0x8EE7EB4A7C0F9E16ull, 0x4FE342E2FE1A7F9Bull}}
342 
343 #define Curve_G_4 { \
344     {0x59F2815B16F81798ull, 0x029BFCDB2DCE28D9ull, 0x55A06295CE870B07ull, 0x79BE667EF9DCBBACull}, \
345     {0x9C47D08FFB10D4B8ull, 0xFD17B448A6855419ull, 0x5DA4FBFC0E1108A8ull, 0x483ADA7726A3C465ull}}
346 
347 #define Curve_G_5 { \
348     {0x343280D6115C1D21ull, 0x4A03C1D356C21122ull, 0x6BB4BF7F321390B9ull, 0x00000000B70E0CBDull}, \
349     {0x44D5819985007E34ull, 0xCD4375A05A074764ull, 0xB5F723FB4C22DFE6ull, 0x00000000BD376388ull}}
350 
351 #define Curve_N_1 {0xF927AED3CA752257ull, 0x000000000001F4C8ull, 0x0000000100000000ull}
352 #define Curve_N_2 {0x146BC9B1B4D22831ull, 0xFFFFFFFF99DEF836ull, 0xFFFFFFFFFFFFFFFFull}
353 #define Curve_N_3 {0xF3B9CAC2FC632551ull, 0xBCE6FAADA7179E84ull, \
354                    0xFFFFFFFFFFFFFFFFull, 0xFFFFFFFF00000000ull}
355 #define Curve_N_4 {0xBFD25E8CD0364141ull, 0xBAAEDCE6AF48A03Bull, \
356                    0xFFFFFFFFFFFFFFFEull, 0xFFFFFFFFFFFFFFFFull}
357 #define Curve_N_5 {0x13DD29455C5C2A3Dull, 0xFFFF16A2E0B8F03Eull, \
358                    0xFFFFFFFFFFFFFFFFull, 0x00000000FFFFFFFFull}
359 
360 #endif /* (uECC_WORD_SIZE == 8) */
361 
362 #define uECC_WORDS uECC_CONCAT(uECC_WORDS_, uECC_CURVE)
363 #define uECC_N_WORDS uECC_CONCAT(uECC_N_WORDS_, uECC_CURVE)
364 
365 typedef struct EccPoint {
366     uECC_word_t x[uECC_WORDS];
367     uECC_word_t y[uECC_WORDS];
368 } EccPoint;
369 
370 static const uECC_word_t curve_p[uECC_WORDS] = uECC_CONCAT(Curve_P_, uECC_CURVE);
371 // Global object `curve_b' is only referenced from function `curve_x_side', it should be defined within that functions block scope
372 static const EccPoint curve_G = uECC_CONCAT(Curve_G_, uECC_CURVE);
373 static const uECC_word_t curve_n[uECC_N_WORDS] = uECC_CONCAT(Curve_N_, uECC_CURVE);
374 
375 static void vli_clear(uECC_word_t *vli);
376 static uECC_word_t vli_isZero(const uECC_word_t *vli);
377 static uECC_word_t vli_testBit(const uECC_word_t *vli, bitcount_t bit);
378 #ifdef ENABLE_MICRO_ECC_ECDSA
379 static bitcount_t vli_numBits(const uECC_word_t *vli, wordcount_t max_words);
380 #endif
381 static void vli_set(uECC_word_t *dest, const uECC_word_t *src);
382 static cmpresult_t vli_cmp(const uECC_word_t *left, const uECC_word_t *right);
383 #ifdef ENABLE_MICRO_ECC_ECDSA
384 static cmpresult_t vli_equal(const uECC_word_t *left, const uECC_word_t *right);
385 #endif
386 static void vli_rshift1(uECC_word_t *vli);
387 static uECC_word_t vli_add(uECC_word_t *result,
388                            const uECC_word_t *left,
389                            const uECC_word_t *right);
390 static uECC_word_t vli_sub(uECC_word_t *result,
391                            const uECC_word_t *left,
392                            const uECC_word_t *right);
393 static void vli_mult(uECC_word_t *result, const uECC_word_t *left, const uECC_word_t *right);
394 static void vli_modAdd(uECC_word_t *result,
395                        const uECC_word_t *left,
396                        const uECC_word_t *right,
397                        const uECC_word_t *mod);
398 static void vli_modSub(uECC_word_t *result,
399                        const uECC_word_t *left,
400                        const uECC_word_t *right,
401                        const uECC_word_t *mod);
402 static void vli_mmod_fast(uECC_word_t *RESTRICT result, uECC_word_t *RESTRICT product);
403 static void vli_modMult_fast(uECC_word_t *result,
404                              const uECC_word_t *left,
405                              const uECC_word_t *right);
406 static void vli_modInv(uECC_word_t *result, const uECC_word_t *input, const uECC_word_t *mod);
407 #if uECC_SQUARE_FUNC
408 static void vli_square(uECC_word_t *result, const uECC_word_t *left);
409 static void vli_modSquare_fast(uECC_word_t *result, const uECC_word_t *left);
410 #endif
411 
412 #if ((defined(_WIN32) || defined(_WIN64)) && !defined(uECC_NO_DEFAULT_RNG))
413 /* Windows */
414 
415 #define WIN32_LEAN_AND_MEAN
416 #include <windows.h>
417 #include <wincrypt.h>
418 
419 static int default_RNG(uint8_t *dest, unsigned size) {
420     HCRYPTPROV prov;
421     if (!CryptAcquireContext(&prov, NULL, NULL, PROV_RSA_FULL, CRYPT_VERIFYCONTEXT)) {
422         return 0;
423     }
424 
425     CryptGenRandom(prov, size, (BYTE *)dest);
426     CryptReleaseContext(prov, 0);
427     return 1;
428 }
429 
430 #elif (defined(unix) || defined(__linux__) || defined(__unix__) || defined(__unix) || \
431     (defined(__APPLE__) && defined(__MACH__)) || defined(uECC_POSIX)) && !defined(uECC_NO_DEFAULT_RNG)
432 
433 /* Some POSIX-like system with /dev/urandom or /dev/random. */
434 #include <sys/types.h>
435 #include <fcntl.h>
436 #include <unistd.h>
437 
438 #ifndef O_CLOEXEC
439     #define O_CLOEXEC 0
440 #endif
441 
442 static int default_RNG(uint8_t *dest, unsigned size) {
443     int fd = open("/dev/urandom", O_RDONLY | O_CLOEXEC);
444     if (fd == -1) {
445         fd = open("/dev/random", O_RDONLY | O_CLOEXEC);
446         if (fd == -1) {
447             return 0;
448         }
449     }
450 
451     char *ptr = (char *)dest;
452     size_t left = size;
453     while (left > 0) {
454         ssize_t bytes_read = read(fd, ptr, left);
455         if (bytes_read <= 0) { // read failed
456             close(fd);
457             return 0;
458         }
459         left -= bytes_read;
460         ptr += bytes_read;
461     }
462 
463     close(fd);
464     return 1;
465 }
466 
467 #else /* Some other platform */
468 
469 static int default_RNG(uint8_t *dest, unsigned size) {
470     (void) dest;
471     (void) size;
472     return 0;
473 }
474 
475 #endif
476 
477 static uECC_RNG_Function g_rng_function = &default_RNG;
478 
479 void uECC_set_rng(uECC_RNG_Function rng_function) {
480     g_rng_function = rng_function;
481 }
482 
483 #ifdef __GNUC__ /* Only support GCC inline asm for now */
484     #if (uECC_ASM && (uECC_PLATFORM == uECC_avr))
485         #include "asm_avr.inc"
486     #endif
487 
488     #if (uECC_ASM && (uECC_PLATFORM == uECC_arm || uECC_PLATFORM == uECC_arm_thumb || \
489                       uECC_PLATFORM == uECC_arm_thumb2))
490         #include "asm_arm.inc"
491     #endif
492 #endif
493 
494 #if !asm_clear
495 static void vli_clear(uECC_word_t *vli) {
496     wordcount_t i;
497     for (i = 0; i < uECC_WORDS; ++i) {
498         vli[i] = 0;
499     }
500 }
501 #endif
502 
503 /* Returns 1 if vli == 0, 0 otherwise. */
504 #if !asm_isZero
505 static uECC_word_t vli_isZero(const uECC_word_t *vli) {
506     wordcount_t i;
507     for (i = 0; i < uECC_WORDS; ++i) {
508         if (vli[i]) {
509             return 0;
510         }
511     }
512     return 1;
513 }
514 #endif
515 
516 /* Returns nonzero if bit 'bit' of vli is set. */
517 #if !asm_testBit
518 static uECC_word_t vli_testBit(const uECC_word_t *vli, bitcount_t bit) {
519     return (vli[bit >> uECC_WORD_BITS_SHIFT] & ((uECC_word_t)1 << (bit & uECC_WORD_BITS_MASK)));
520 }
521 #endif
522 
523 #ifdef ENABLE_MICO_ECC_ECDSA
524 
525 /* Counts the number of words in vli. */
526 #if !asm_numBits
527 static wordcount_t vli_numDigits(const uECC_word_t *vli, wordcount_t max_words) {
528     swordcount_t i;
529     /* Search from the end until we find a non-zero digit.
530        We do it in reverse because we expect that most digits will be nonzero. */
531     for (i = max_words - 1; i >= 0 && vli[i] == 0; --i) {
532     }
533 
534     return (i + 1);
535 }
536 
537 /* Counts the number of bits required to represent vli. */
538 static bitcount_t vli_numBits(const uECC_word_t *vli, wordcount_t max_words) {
539     uECC_word_t i;
540     uECC_word_t digit;
541 
542     wordcount_t num_digits = vli_numDigits(vli, max_words);
543     if (num_digits == 0) {
544         return 0;
545     }
546 
547     digit = vli[num_digits - 1];
548     for (i = 0; digit; ++i) {
549         digit >>= 1;
550     }
551 
552     return (((bitcount_t)(num_digits - 1) << uECC_WORD_BITS_SHIFT) + i);
553 }
554 
555 #endif /* ENABLE_MICO_ECC_ECDSA */
556 
557 #endif /* !asm_numBits */
558 
559 /* Sets dest = src. */
560 #if !asm_set
561 static void vli_set(uECC_word_t *dest, const uECC_word_t *src) {
562     wordcount_t i;
563     for (i = 0; i < uECC_WORDS; ++i) {
564         dest[i] = src[i];
565     }
566 }
567 #endif
568 
569 /* Returns sign of left - right. */
570 #if !asm_cmp
571 static cmpresult_t vli_cmp(const uECC_word_t *left, const uECC_word_t *right) {
572     swordcount_t i;
573     for (i = uECC_WORDS - 1; i >= 0; --i) {
574         if (left[i] > right[i]) {
575             return 1;
576         } else if (left[i] < right[i]) {
577             return -1;
578         }
579     }
580     return 0;
581 }
582 #endif
583 
584 #ifdef ENABLE_MICRO_ECC_ECDSA
585 
586 static cmpresult_t vli_equal(const uECC_word_t *left, const uECC_word_t *right) {
587     uECC_word_t result = 0;
588     swordcount_t i;
589     for (i = uECC_WORDS - 1; i >= 0; --i) {
590         result |= (left[i] ^ right[i]);
591     }
592     return (result == 0);
593 }
594 
595 #endif
596 
597 /* Computes vli = vli >> 1. */
598 #if !asm_rshift1
599 static void vli_rshift1(uECC_word_t *vli) {
600     uECC_word_t *end = vli;
601     uECC_word_t carry = 0;
602 
603     vli += uECC_WORDS;
604     while (vli-- > end) {
605         uECC_word_t temp = *vli;
606         *vli = (temp >> 1) | carry;
607         carry = temp << (uECC_WORD_BITS - 1);
608     }
609 }
610 #endif
611 
612 /* Computes result = left + right, returning carry. Can modify in place. */
613 #if !asm_add
614 static uECC_word_t vli_add(uECC_word_t *result, const uECC_word_t *left, const uECC_word_t *right) {
615     uECC_word_t carry = 0;
616     wordcount_t i;
617     for (i = 0; i < uECC_WORDS; ++i) {
618         uECC_word_t sum = left[i] + right[i] + carry;
619         if (sum != left[i]) {
620             carry = (sum < left[i]);
621         }
622         result[i] = sum;
623     }
624     return carry;
625 }
626 #endif
627 
628 /* Computes result = left - right, returning borrow. Can modify in place. */
629 #if !asm_sub
630 static uECC_word_t vli_sub(uECC_word_t *result, const uECC_word_t *left, const uECC_word_t *right) {
631     uECC_word_t borrow = 0;
632     wordcount_t i;
633     for (i = 0; i < uECC_WORDS; ++i) {
634         uECC_word_t diff = left[i] - right[i] - borrow;
635         if (diff != left[i]) {
636             borrow = (diff > left[i]);
637         }
638         result[i] = diff;
639     }
640     return borrow;
641 }
642 #endif
643 
644 #if (!asm_mult || (uECC_SQUARE_FUNC && !asm_square) || uECC_CURVE == uECC_secp256k1)
645 static void muladd(uECC_word_t a,
646                    uECC_word_t b,
647                    uECC_word_t *r0,
648                    uECC_word_t *r1,
649                    uECC_word_t *r2) {
650 #if uECC_WORD_SIZE == 8 && !SUPPORTS_INT128
651     uint64_t a0 = a & 0xffffffffull;
652     uint64_t a1 = a >> 32;
653     uint64_t b0 = b & 0xffffffffull;
654     uint64_t b1 = b >> 32;
655 
656     uint64_t i0 = a0 * b0;
657     uint64_t i1 = a0 * b1;
658     uint64_t i2 = a1 * b0;
659     uint64_t i3 = a1 * b1;
660 
661     uint64_t p0, p1;
662 
663     i2 += (i0 >> 32);
664     i2 += i1;
665     if (i2 < i1) { // overflow
666         i3 += 0x100000000ull;
667     }
668 
669     p0 = (i0 & 0xffffffffull) | (i2 << 32);
670     p1 = i3 + (i2 >> 32);
671 
672     *r0 += p0;
673     *r1 += (p1 + (*r0 < p0));
674     *r2 += ((*r1 < p1) || (*r1 == p1 && *r0 < p0));
675 #else
676     uECC_dword_t p = (uECC_dword_t)a * b;
677     uECC_dword_t r01 = ((uECC_dword_t)(*r1) << uECC_WORD_BITS) | *r0;
678     r01 += p;
679     *r2 += (r01 < p);
680     *r1 = r01 >> uECC_WORD_BITS;
681     *r0 = (uECC_word_t)r01;
682 #endif
683 }
684 #define muladd_exists 1
685 #endif
686 
687 #if !asm_mult
688 static void vli_mult(uECC_word_t *result, const uECC_word_t *left, const uECC_word_t *right) {
689     uECC_word_t r0 = 0;
690     uECC_word_t r1 = 0;
691     uECC_word_t r2 = 0;
692     wordcount_t i, k;
693 
694     /* Compute each digit of result in sequence, maintaining the carries. */
695     for (k = 0; k < uECC_WORDS; ++k) {
696         for (i = 0; i <= k; ++i) {
697             muladd(left[i], right[k - i], &r0, &r1, &r2);
698         }
699         result[k] = r0;
700         r0 = r1;
701         r1 = r2;
702         r2 = 0;
703     }
704     for (k = uECC_WORDS; k < uECC_WORDS * 2 - 1; ++k) {
705         for (i = (k + 1) - uECC_WORDS; i < uECC_WORDS; ++i) {
706             muladd(left[i], right[k - i], &r0, &r1, &r2);
707         }
708         result[k] = r0;
709         r0 = r1;
710         r1 = r2;
711         r2 = 0;
712     }
713     result[uECC_WORDS * 2 - 1] = r0;
714 }
715 #endif
716 
717 #if uECC_SQUARE_FUNC
718 
719 #if !asm_square
720 static void mul2add(uECC_word_t a,
721                     uECC_word_t b,
722                     uECC_word_t *r0,
723                     uECC_word_t *r1,
724                     uECC_word_t *r2) {
725 #if uECC_WORD_SIZE == 8 && !SUPPORTS_INT128
726     uint64_t a0 = a & 0xffffffffull;
727     uint64_t a1 = a >> 32;
728     uint64_t b0 = b & 0xffffffffull;
729     uint64_t b1 = b >> 32;
730 
731     uint64_t i0 = a0 * b0;
732     uint64_t i1 = a0 * b1;
733     uint64_t i2 = a1 * b0;
734     uint64_t i3 = a1 * b1;
735 
736     uint64_t p0, p1;
737 
738     i2 += (i0 >> 32);
739     i2 += i1;
740     if (i2 < i1)
741     { // overflow
742         i3 += 0x100000000ull;
743     }
744 
745     p0 = (i0 & 0xffffffffull) | (i2 << 32);
746     p1 = i3 + (i2 >> 32);
747 
748     *r2 += (p1 >> 63);
749     p1 = (p1 << 1) | (p0 >> 63);
750     p0 <<= 1;
751 
752     *r0 += p0;
753     *r1 += (p1 + (*r0 < p0));
754     *r2 += ((*r1 < p1) || (*r1 == p1 && *r0 < p0));
755 #else
756     uECC_dword_t p = (uECC_dword_t)a * b;
757     uECC_dword_t r01 = ((uECC_dword_t)(*r1) << uECC_WORD_BITS) | *r0;
758     *r2 += (p >> (uECC_WORD_BITS * 2 - 1));
759     p *= 2;
760     r01 += p;
761     *r2 += (r01 < p);
762     *r1 = r01 >> uECC_WORD_BITS;
763     *r0 = (uECC_word_t)r01;
764 #endif
765 }
766 
767 static void vli_square(uECC_word_t *result, const uECC_word_t *left) {
768     uECC_word_t r0 = 0;
769     uECC_word_t r1 = 0;
770     uECC_word_t r2 = 0;
771 
772     wordcount_t i, k;
773 
774     for (k = 0; k < uECC_WORDS * 2 - 1; ++k) {
775         uECC_word_t min = (k < uECC_WORDS ? 0 : (k + 1) - uECC_WORDS);
776         for (i = min; i <= k && i <= k - i; ++i) {
777             if (i < k-i) {
778                 mul2add(left[i], left[k - i], &r0, &r1, &r2);
779             } else {
780                 muladd(left[i], left[k - i], &r0, &r1, &r2);
781             }
782         }
783         result[k] = r0;
784         r0 = r1;
785         r1 = r2;
786         r2 = 0;
787     }
788 
789     result[uECC_WORDS * 2 - 1] = r0;
790 }
791 #endif
792 
793 #else /* uECC_SQUARE_FUNC */
794 
795 #define vli_square(result, left, size) vli_mult((result), (left), (left), (size))
796 
797 #endif /* uECC_SQUARE_FUNC */
798 
799 
800 /* Computes result = (left + right) % mod.
801    Assumes that left < mod and right < mod, and that result does not overlap mod. */
802 #if !asm_modAdd
803 static void vli_modAdd(uECC_word_t *result,
804                        const uECC_word_t *left,
805                        const uECC_word_t *right,
806                        const uECC_word_t *mod) {
807     uECC_word_t carry = vli_add(result, left, right);
808     if (carry || vli_cmp(result, mod) >= 0) {
809         /* result > mod (result = mod + remainder), so subtract mod to get remainder. */
810         vli_sub(result, result, mod);
811     }
812 }
813 #endif
814 
815 /* Computes result = (left - right) % mod.
816    Assumes that left < mod and right < mod, and that result does not overlap mod. */
817 #if !asm_modSub
818 static void vli_modSub(uECC_word_t *result,
819                        const uECC_word_t *left,
820                        const uECC_word_t *right,
821                        const uECC_word_t *mod) {
822     uECC_word_t l_borrow = vli_sub(result, left, right);
823     if (l_borrow) {
824         /* In this case, result == -diff == (max int) - diff. Since -x % d == d - x,
825            we can get the correct result from result + mod (with overflow). */
826         vli_add(result, result, mod);
827     }
828 }
829 #endif
830 
831 #if !asm_modSub_fast
832     #define vli_modSub_fast(result, left, right) vli_modSub((result), (left), (right), curve_p)
833 #endif
834 
835 #if !asm_mmod_fast
836 
837 #if (uECC_CURVE == uECC_secp160r1 || uECC_CURVE == uECC_secp256k1)
838 /* omega_mult() is defined farther below for the different curves / word sizes */
839 static void omega_mult(uECC_word_t * RESTRICT result, const uECC_word_t * RESTRICT right);
840 
841 /* Computes result = product % curve_p
842     see http://www.isys.uni-klu.ac.at/PDF/2001-0126-MT.pdf page 354
843 
844     Note that this only works if log2(omega) < log2(p) / 2 */
845 static void vli_mmod_fast(uECC_word_t *RESTRICT result, uECC_word_t *RESTRICT product) {
846     uECC_word_t tmp[2 * uECC_WORDS];
847     uECC_word_t carry;
848 
849     vli_clear(tmp);
850     vli_clear(tmp + uECC_WORDS);
851 
852     omega_mult(tmp, product + uECC_WORDS); /* (Rq, q) = q * c */
853 
854     carry = vli_add(result, product, tmp); /* (C, r) = r + q       */
855     vli_clear(product);
856     omega_mult(product, tmp + uECC_WORDS); /* Rq*c */
857     carry += vli_add(result, result, product); /* (C1, r) = r + Rq*c */
858 
859     while (carry > 0) {
860         --carry;
861         vli_sub(result, result, curve_p);
862     }
863     if (vli_cmp(result, curve_p) > 0) {
864         vli_sub(result, result, curve_p);
865     }
866 }
867 
868 #endif
869 
870 #if uECC_CURVE == uECC_secp160r1
871 
872 #if uECC_WORD_SIZE == 1
873 static void omega_mult(uint8_t * RESTRICT result, const uint8_t * RESTRICT right) {
874     uint8_t carry;
875     uint8_t i;
876 
877     /* Multiply by (2^31 + 1). */
878     vli_set(result + 4, right); /* 2^32 */
879     vli_rshift1(result + 4); /* 2^31 */
880     result[3] = right[0] << 7; /* get last bit from shift */
881 
882     carry = vli_add(result, result, right); /* 2^31 + 1 */
883     for (i = uECC_WORDS; carry; ++i) {
884         uint16_t sum = (uint16_t)result[i] + carry;
885         result[i] = (uint8_t)sum;
886         carry = sum >> 8;
887     }
888 }
889 #elif uECC_WORD_SIZE == 4
890 static void omega_mult(uint32_t * RESTRICT result, const uint32_t * RESTRICT right) {
891     uint32_t carry;
892     unsigned i;
893 
894     /* Multiply by (2^31 + 1). */
895     vli_set(result + 1, right); /* 2^32 */
896     vli_rshift1(result + 1); /* 2^31 */
897     result[0] = right[0] << 31; /* get last bit from shift */
898 
899     carry = vli_add(result, result, right); /* 2^31 + 1 */
900     for (i = uECC_WORDS; carry; ++i) {
901         uint64_t sum = (uint64_t)result[i] + carry;
902         result[i] = (uint32_t)sum;
903         carry = sum >> 32;
904     }
905 }
906 #endif /* uECC_WORD_SIZE */
907 
908 #elif uECC_CURVE == uECC_secp192r1
909 
910 /* Computes result = product % curve_p.
911    See algorithm 5 and 6 from http://www.isys.uni-klu.ac.at/PDF/2001-0126-MT.pdf */
912 #if uECC_WORD_SIZE == 1
913 static void vli_mmod_fast(uint8_t *RESTRICT result, uint8_t *RESTRICT product) {
914     uint8_t tmp[uECC_WORDS];
915     uint8_t carry;
916 
917     vli_set(result, product);
918 
919     vli_set(tmp, &product[24]);
920     carry = vli_add(result, result, tmp);
921 
922     tmp[0] = tmp[1] = tmp[2] = tmp[3] = tmp[4] = tmp[5] = tmp[6] = tmp[7] = 0;
923     tmp[8] = product[24]; tmp[9] = product[25]; tmp[10] = product[26]; tmp[11] = product[27];
924     tmp[12] = product[28]; tmp[13] = product[29]; tmp[14] = product[30]; tmp[15] = product[31];
925     tmp[16] = product[32]; tmp[17] = product[33]; tmp[18] = product[34]; tmp[19] = product[35];
926     tmp[20] = product[36]; tmp[21] = product[37]; tmp[22] = product[38]; tmp[23] = product[39];
927     carry += vli_add(result, result, tmp);
928 
929     tmp[0] = tmp[8] = product[40];
930     tmp[1] = tmp[9] = product[41];
931     tmp[2] = tmp[10] = product[42];
932     tmp[3] = tmp[11] = product[43];
933     tmp[4] = tmp[12] = product[44];
934     tmp[5] = tmp[13] = product[45];
935     tmp[6] = tmp[14] = product[46];
936     tmp[7] = tmp[15] = product[47];
937     tmp[16] = tmp[17] = tmp[18] = tmp[19] = tmp[20] = tmp[21] = tmp[22] = tmp[23] = 0;
938     carry += vli_add(result, result, tmp);
939 
940     while (carry || vli_cmp(curve_p, result) != 1) {
941         carry -= vli_sub(result, result, curve_p);
942     }
943 }
944 #elif uECC_WORD_SIZE == 4
945 static void vli_mmod_fast(uint32_t *RESTRICT result, uint32_t *RESTRICT product) {
946     uint32_t tmp[uECC_WORDS];
947     int carry;
948 
949     vli_set(result, product);
950 
951     vli_set(tmp, &product[6]);
952     carry = vli_add(result, result, tmp);
953 
954     tmp[0] = tmp[1] = 0;
955     tmp[2] = product[6];
956     tmp[3] = product[7];
957     tmp[4] = product[8];
958     tmp[5] = product[9];
959     carry += vli_add(result, result, tmp);
960 
961     tmp[0] = tmp[2] = product[10];
962     tmp[1] = tmp[3] = product[11];
963     tmp[4] = tmp[5] = 0;
964     carry += vli_add(result, result, tmp);
965 
966     while (carry || vli_cmp(curve_p, result) != 1) {
967         carry -= vli_sub(result, result, curve_p);
968     }
969 }
970 #else
971 static void vli_mmod_fast(uint64_t *RESTRICT result, uint64_t *RESTRICT product) {
972     uint64_t tmp[uECC_WORDS];
973     int carry;
974 
975     vli_set(result, product);
976 
977     vli_set(tmp, &product[3]);
978     carry = vli_add(result, result, tmp);
979 
980     tmp[0] = 0;
981     tmp[1] = product[3];
982     tmp[2] = product[4];
983     carry += vli_add(result, result, tmp);
984 
985     tmp[0] = tmp[1] = product[5];
986     tmp[2] = 0;
987     carry += vli_add(result, result, tmp);
988 
989     while (carry || vli_cmp(curve_p, result) != 1) {
990         carry -= vli_sub(result, result, curve_p);
991     }
992 }
993 #endif /* uECC_WORD_SIZE */
994 
995 #elif uECC_CURVE == uECC_secp256r1
996 
997 /* Computes result = product % curve_p
998    from http://www.nsa.gov/ia/_files/nist-routines.pdf */
999 #if uECC_WORD_SIZE == 1
1000 static void vli_mmod_fast(uint8_t *RESTRICT result, uint8_t *RESTRICT product) {
1001     uint8_t tmp[uECC_BYTES];
1002     int8_t carry;
1003 
1004     /* t */
1005     vli_set(result, product);
1006 
1007     /* s1 */
1008     tmp[0] = tmp[1] = tmp[2] = tmp[3] = 0;
1009     tmp[4] = tmp[5] = tmp[6] = tmp[7] = 0;
1010     tmp[8] = tmp[9] = tmp[10] = tmp[11] = 0;
1011     tmp[12] = product[44]; tmp[13] = product[45]; tmp[14] = product[46]; tmp[15] = product[47];
1012     tmp[16] = product[48]; tmp[17] = product[49]; tmp[18] = product[50]; tmp[19] = product[51];
1013     tmp[20] = product[52]; tmp[21] = product[53]; tmp[22] = product[54]; tmp[23] = product[55];
1014     tmp[24] = product[56]; tmp[25] = product[57]; tmp[26] = product[58]; tmp[27] = product[59];
1015     tmp[28] = product[60]; tmp[29] = product[61]; tmp[30] = product[62]; tmp[31] = product[63];
1016     carry = vli_add(tmp, tmp, tmp);
1017     carry += vli_add(result, result, tmp);
1018 
1019     /* s2 */
1020     tmp[12] = product[48]; tmp[13] = product[49]; tmp[14] = product[50]; tmp[15] = product[51];
1021     tmp[16] = product[52]; tmp[17] = product[53]; tmp[18] = product[54]; tmp[19] = product[55];
1022     tmp[20] = product[56]; tmp[21] = product[57]; tmp[22] = product[58]; tmp[23] = product[59];
1023     tmp[24] = product[60]; tmp[25] = product[61]; tmp[26] = product[62]; tmp[27] = product[63];
1024     tmp[28] = tmp[29] = tmp[30] = tmp[31] = 0;
1025     carry += vli_add(tmp, tmp, tmp);
1026     carry += vli_add(result, result, tmp);
1027 
1028     /* s3 */
1029     tmp[0] = product[32]; tmp[1] = product[33]; tmp[2] = product[34]; tmp[3] = product[35];
1030     tmp[4] = product[36]; tmp[5] = product[37]; tmp[6] = product[38]; tmp[7] = product[39];
1031     tmp[8] = product[40]; tmp[9] = product[41]; tmp[10] = product[42]; tmp[11] = product[43];
1032     tmp[12] = tmp[13] = tmp[14] = tmp[15] = 0;
1033     tmp[16] = tmp[17] = tmp[18] = tmp[19] = 0;
1034     tmp[20] = tmp[21] = tmp[22] = tmp[23] = 0;
1035     tmp[24] = product[56]; tmp[25] = product[57]; tmp[26] = product[58]; tmp[27] = product[59];
1036     tmp[28] = product[60]; tmp[29] = product[61]; tmp[30] = product[62]; tmp[31] = product[63];
1037     carry += vli_add(result, result, tmp);
1038 
1039     /* s4 */
1040     tmp[0] = product[36]; tmp[1] = product[37]; tmp[2] = product[38]; tmp[3] = product[39];
1041     tmp[4] = product[40]; tmp[5] = product[41]; tmp[6] = product[42]; tmp[7] = product[43];
1042     tmp[8] = product[44]; tmp[9] = product[45]; tmp[10] = product[46]; tmp[11] = product[47];
1043     tmp[12] = product[52]; tmp[13] = product[53]; tmp[14] = product[54]; tmp[15] = product[55];
1044     tmp[16] = product[56]; tmp[17] = product[57]; tmp[18] = product[58]; tmp[19] = product[59];
1045     tmp[20] = product[60]; tmp[21] = product[61]; tmp[22] = product[62]; tmp[23] = product[63];
1046     tmp[24] = product[52]; tmp[25] = product[53]; tmp[26] = product[54]; tmp[27] = product[55];
1047     tmp[28] = product[32]; tmp[29] = product[33]; tmp[30] = product[34]; tmp[31] = product[35];
1048     carry += vli_add(result, result, tmp);
1049 
1050     /* d1 */
1051     tmp[0] = product[44]; tmp[1] = product[45]; tmp[2] = product[46]; tmp[3] = product[47];
1052     tmp[4] = product[48]; tmp[5] = product[49]; tmp[6] = product[50]; tmp[7] = product[51];
1053     tmp[8] = product[52]; tmp[9] = product[53]; tmp[10] = product[54]; tmp[11] = product[55];
1054     tmp[12] = tmp[13] = tmp[14] = tmp[15] = 0;
1055     tmp[16] = tmp[17] = tmp[18] = tmp[19] = 0;
1056     tmp[20] = tmp[21] = tmp[22] = tmp[23] = 0;
1057     tmp[24] = product[32]; tmp[25] = product[33]; tmp[26] = product[34]; tmp[27] = product[35];
1058     tmp[28] = product[40]; tmp[29] = product[41]; tmp[30] = product[42]; tmp[31] = product[43];
1059     carry -= vli_sub(result, result, tmp);
1060 
1061     /* d2 */
1062     tmp[0] = product[48]; tmp[1] = product[49]; tmp[2] = product[50]; tmp[3] = product[51];
1063     tmp[4] = product[52]; tmp[5] = product[53]; tmp[6] = product[54]; tmp[7] = product[55];
1064     tmp[8] = product[56]; tmp[9] = product[57]; tmp[10] = product[58]; tmp[11] = product[59];
1065     tmp[12] = product[60]; tmp[13] = product[61]; tmp[14] = product[62]; tmp[15] = product[63];
1066     tmp[16] = tmp[17] = tmp[18] = tmp[19] = 0;
1067     tmp[20] = tmp[21] = tmp[22] = tmp[23] = 0;
1068     tmp[24] = product[36]; tmp[25] = product[37]; tmp[26] = product[38]; tmp[27] = product[39];
1069     tmp[28] = product[44]; tmp[29] = product[45]; tmp[30] = product[46]; tmp[31] = product[47];
1070     carry -= vli_sub(result, result, tmp);
1071 
1072     /* d3 */
1073     tmp[0] = product[52]; tmp[1] = product[53]; tmp[2] = product[54]; tmp[3] = product[55];
1074     tmp[4] = product[56]; tmp[5] = product[57]; tmp[6] = product[58]; tmp[7] = product[59];
1075     tmp[8] = product[60]; tmp[9] = product[61]; tmp[10] = product[62]; tmp[11] = product[63];
1076     tmp[12] = product[32]; tmp[13] = product[33]; tmp[14] = product[34]; tmp[15] = product[35];
1077     tmp[16] = product[36]; tmp[17] = product[37]; tmp[18] = product[38]; tmp[19] = product[39];
1078     tmp[20] = product[40]; tmp[21] = product[41]; tmp[22] = product[42]; tmp[23] = product[43];
1079     tmp[24] = tmp[25] = tmp[26] = tmp[27] = 0;
1080     tmp[28] = product[48]; tmp[29] = product[49]; tmp[30] = product[50]; tmp[31] = product[51];
1081     carry -= vli_sub(result, result, tmp);
1082 
1083     /* d4 */
1084     tmp[0] = product[56]; tmp[1] = product[57]; tmp[2] = product[58]; tmp[3] = product[59];
1085     tmp[4] = product[60]; tmp[5] = product[61]; tmp[6] = product[62]; tmp[7] = product[63];
1086     tmp[8] = tmp[9] = tmp[10] = tmp[11] = 0;
1087     tmp[12] = product[36]; tmp[13] = product[37]; tmp[14] = product[38]; tmp[15] = product[39];
1088     tmp[16] = product[40]; tmp[17] = product[41]; tmp[18] = product[42]; tmp[19] = product[43];
1089     tmp[20] = product[44]; tmp[21] = product[45]; tmp[22] = product[46]; tmp[23] = product[47];
1090     tmp[24] = tmp[25] = tmp[26] = tmp[27] = 0;
1091     tmp[28] = product[52]; tmp[29] = product[53]; tmp[30] = product[54]; tmp[31] = product[55];
1092     carry -= vli_sub(result, result, tmp);
1093 
1094     if (carry < 0) {
1095         do {
1096             carry += vli_add(result, result, curve_p);
1097         } while (carry < 0);
1098     } else {
1099         while (carry || vli_cmp(curve_p, result) != 1) {
1100             carry -= vli_sub(result, result, curve_p);
1101         }
1102     }
1103 }
1104 #elif uECC_WORD_SIZE == 4
1105 static void vli_mmod_fast(uint32_t *RESTRICT result, uint32_t *RESTRICT product) {
1106     uint32_t tmp[uECC_WORDS];
1107     int carry;
1108 
1109     /* t */
1110     vli_set(result, product);
1111 
1112     /* s1 */
1113     tmp[0] = tmp[1] = tmp[2] = 0;
1114     tmp[3] = product[11];
1115     tmp[4] = product[12];
1116     tmp[5] = product[13];
1117     tmp[6] = product[14];
1118     tmp[7] = product[15];
1119     carry = vli_add(tmp, tmp, tmp);
1120     carry += vli_add(result, result, tmp);
1121 
1122     /* s2 */
1123     tmp[3] = product[12];
1124     tmp[4] = product[13];
1125     tmp[5] = product[14];
1126     tmp[6] = product[15];
1127     tmp[7] = 0;
1128     carry += vli_add(tmp, tmp, tmp);
1129     carry += vli_add(result, result, tmp);
1130 
1131     /* s3 */
1132     tmp[0] = product[8];
1133     tmp[1] = product[9];
1134     tmp[2] = product[10];
1135     tmp[3] = tmp[4] = tmp[5] = 0;
1136     tmp[6] = product[14];
1137     tmp[7] = product[15];
1138     carry += vli_add(result, result, tmp);
1139 
1140     /* s4 */
1141     tmp[0] = product[9];
1142     tmp[1] = product[10];
1143     tmp[2] = product[11];
1144     tmp[3] = product[13];
1145     tmp[4] = product[14];
1146     tmp[5] = product[15];
1147     tmp[6] = product[13];
1148     tmp[7] = product[8];
1149     carry += vli_add(result, result, tmp);
1150 
1151     /* d1 */
1152     tmp[0] = product[11];
1153     tmp[1] = product[12];
1154     tmp[2] = product[13];
1155     tmp[3] = tmp[4] = tmp[5] = 0;
1156     tmp[6] = product[8];
1157     tmp[7] = product[10];
1158     carry -= vli_sub(result, result, tmp);
1159 
1160     /* d2 */
1161     tmp[0] = product[12];
1162     tmp[1] = product[13];
1163     tmp[2] = product[14];
1164     tmp[3] = product[15];
1165     tmp[4] = tmp[5] = 0;
1166     tmp[6] = product[9];
1167     tmp[7] = product[11];
1168     carry -= vli_sub(result, result, tmp);
1169 
1170     /* d3 */
1171     tmp[0] = product[13];
1172     tmp[1] = product[14];
1173     tmp[2] = product[15];
1174     tmp[3] = product[8];
1175     tmp[4] = product[9];
1176     tmp[5] = product[10];
1177     tmp[6] = 0;
1178     tmp[7] = product[12];
1179     carry -= vli_sub(result, result, tmp);
1180 
1181     /* d4 */
1182     tmp[0] = product[14];
1183     tmp[1] = product[15];
1184     tmp[2] = 0;
1185     tmp[3] = product[9];
1186     tmp[4] = product[10];
1187     tmp[5] = product[11];
1188     tmp[6] = 0;
1189     tmp[7] = product[13];
1190     carry -= vli_sub(result, result, tmp);
1191 
1192     if (carry < 0) {
1193         do {
1194             carry += vli_add(result, result, curve_p);
1195         } while (carry < 0);
1196     } else {
1197         while (carry || vli_cmp(curve_p, result) != 1) {
1198             carry -= vli_sub(result, result, curve_p);
1199         }
1200     }
1201 }
1202 #else
1203 static void vli_mmod_fast(uint64_t *RESTRICT result, uint64_t *RESTRICT product) {
1204     uint64_t tmp[uECC_WORDS];
1205     int carry;
1206 
1207     /* t */
1208     vli_set(result, product);
1209 
1210     /* s1 */
1211     tmp[0] = 0;
1212     tmp[1] = product[5] & 0xffffffff00000000ull;
1213     tmp[2] = product[6];
1214     tmp[3] = product[7];
1215     carry = vli_add(tmp, tmp, tmp);
1216     carry += vli_add(result, result, tmp);
1217 
1218     /* s2 */
1219     tmp[1] = product[6] << 32;
1220     tmp[2] = (product[6] >> 32) | (product[7] << 32);
1221     tmp[3] = product[7] >> 32;
1222     carry += vli_add(tmp, tmp, tmp);
1223     carry += vli_add(result, result, tmp);
1224 
1225     /* s3 */
1226     tmp[0] = product[4];
1227     tmp[1] = product[5] & 0xffffffff;
1228     tmp[2] = 0;
1229     tmp[3] = product[7];
1230     carry += vli_add(result, result, tmp);
1231 
1232     /* s4 */
1233     tmp[0] = (product[4] >> 32) | (product[5] << 32);
1234     tmp[1] = (product[5] >> 32) | (product[6] & 0xffffffff00000000ull);
1235     tmp[2] = product[7];
1236     tmp[3] = (product[6] >> 32) | (product[4] << 32);
1237     carry += vli_add(result, result, tmp);
1238 
1239     /* d1 */
1240     tmp[0] = (product[5] >> 32) | (product[6] << 32);
1241     tmp[1] = (product[6] >> 32);
1242     tmp[2] = 0;
1243     tmp[3] = (product[4] & 0xffffffff) | (product[5] << 32);
1244     carry -= vli_sub(result, result, tmp);
1245 
1246     /* d2 */
1247     tmp[0] = product[6];
1248     tmp[1] = product[7];
1249     tmp[2] = 0;
1250     tmp[3] = (product[4] >> 32) | (product[5] & 0xffffffff00000000ull);
1251     carry -= vli_sub(result, result, tmp);
1252 
1253     /* d3 */
1254     tmp[0] = (product[6] >> 32) | (product[7] << 32);
1255     tmp[1] = (product[7] >> 32) | (product[4] << 32);
1256     tmp[2] = (product[4] >> 32) | (product[5] << 32);
1257     tmp[3] = (product[6] << 32);
1258     carry -= vli_sub(result, result, tmp);
1259 
1260     /* d4 */
1261     tmp[0] = product[7];
1262     tmp[1] = product[4] & 0xffffffff00000000ull;
1263     tmp[2] = product[5];
1264     tmp[3] = product[6] & 0xffffffff00000000ull;
1265     carry -= vli_sub(result, result, tmp);
1266 
1267     if (carry < 0) {
1268         do {
1269             carry += vli_add(result, result, curve_p);
1270         } while (carry < 0);
1271     } else {
1272         while (carry || vli_cmp(curve_p, result) != 1) {
1273             carry -= vli_sub(result, result, curve_p);
1274         }
1275     }
1276 }
1277 #endif /* uECC_WORD_SIZE */
1278 
1279 #elif uECC_CURVE == uECC_secp256k1
1280 
1281 #if uECC_WORD_SIZE == 1
1282 static void omega_mult(uint8_t * RESTRICT result, const uint8_t * RESTRICT right) {
1283     /* Multiply by (2^32 + 2^9 + 2^8 + 2^7 + 2^6 + 2^4 + 1). */
1284     uECC_word_t r0 = 0;
1285     uECC_word_t r1 = 0;
1286     uECC_word_t r2 = 0;
1287     wordcount_t k;
1288 
1289     /* Multiply by (2^9 + 2^8 + 2^7 + 2^6 + 2^4 + 1). */
1290     muladd(0xD1, right[0], &r0, &r1, &r2);
1291     result[0] = r0;
1292     r0 = r1;
1293     r1 = r2;
1294     /* r2 is still 0 */
1295 
1296     for (k = 1; k < uECC_WORDS; ++k) {
1297         muladd(0x03, right[k - 1], &r0, &r1, &r2);
1298         muladd(0xD1, right[k], &r0, &r1, &r2);
1299         result[k] = r0;
1300         r0 = r1;
1301         r1 = r2;
1302         r2 = 0;
1303     }
1304     muladd(0x03, right[uECC_WORDS - 1], &r0, &r1, &r2);
1305     result[uECC_WORDS] = r0;
1306     result[uECC_WORDS + 1] = r1;
1307 
1308     result[4 + uECC_WORDS] = vli_add(result + 4, result + 4, right); /* add the 2^32 multiple */
1309 }
1310 #elif uECC_WORD_SIZE == 4
1311 static void omega_mult(uint32_t * RESTRICT result, const uint32_t * RESTRICT right) {
1312     /* Multiply by (2^9 + 2^8 + 2^7 + 2^6 + 2^4 + 1). */
1313     uint32_t carry = 0;
1314     wordcount_t k;
1315 
1316     for (k = 0; k < uECC_WORDS; ++k) {
1317         uint64_t p = (uint64_t)0x3D1 * right[k] + carry;
1318         result[k] = (p & 0xffffffff);
1319         carry = p >> 32;
1320     }
1321     result[uECC_WORDS] = carry;
1322 
1323     result[1 + uECC_WORDS] = vli_add(result + 1, result + 1, right); /* add the 2^32 multiple */
1324 }
1325 #else
1326 static void omega_mult(uint64_t * RESTRICT result, const uint64_t * RESTRICT right) {
1327     uECC_word_t r0 = 0;
1328     uECC_word_t r1 = 0;
1329     uECC_word_t r2 = 0;
1330     wordcount_t k;
1331 
1332     /* Multiply by (2^32 + 2^9 + 2^8 + 2^7 + 2^6 + 2^4 + 1). */
1333     for (k = 0; k < uECC_WORDS; ++k) {
1334         muladd(0x1000003D1ull, right[k], &r0, &r1, &r2);
1335         result[k] = r0;
1336         r0 = r1;
1337         r1 = r2;
1338         r2 = 0;
1339     }
1340     result[uECC_WORDS] = r0;
1341 }
1342 #endif /* uECC_WORD_SIZE */
1343 
1344 #elif uECC_CURVE == uECC_secp224r1
1345 
1346 /* Computes result = product % curve_p
1347    from http://www.nsa.gov/ia/_files/nist-routines.pdf */
1348 #if uECC_WORD_SIZE == 1
1349 // TODO it may be faster to use the omega_mult method when fully asm optimized.
1350 void vli_mmod_fast(uint8_t *RESTRICT result, uint8_t *RESTRICT product) {
1351     uint8_t tmp[uECC_WORDS];
1352     int8_t carry;
1353 
1354     /* t */
1355     vli_set(result, product);
1356 
1357     /* s1 */
1358     tmp[0] = tmp[1] = tmp[2] = tmp[3] = 0;
1359     tmp[4] = tmp[5] = tmp[6] = tmp[7] = 0;
1360     tmp[8] = tmp[9] = tmp[10] = tmp[11] = 0;
1361     tmp[12] = product[28]; tmp[13] = product[29]; tmp[14] = product[30]; tmp[15] = product[31];
1362     tmp[16] = product[32]; tmp[17] = product[33]; tmp[18] = product[34]; tmp[19] = product[35];
1363     tmp[20] = product[36]; tmp[21] = product[37]; tmp[22] = product[38]; tmp[23] = product[39];
1364     tmp[24] = product[40]; tmp[25] = product[41]; tmp[26] = product[42]; tmp[27] = product[43];
1365     carry = vli_add(result, result, tmp);
1366 
1367     /* s2 */
1368     tmp[12] = product[44]; tmp[13] = product[45]; tmp[14] = product[46]; tmp[15] = product[47];
1369     tmp[16] = product[48]; tmp[17] = product[49]; tmp[18] = product[50]; tmp[19] = product[51];
1370     tmp[20] = product[52]; tmp[21] = product[53]; tmp[22] = product[54]; tmp[23] = product[55];
1371     tmp[24] = tmp[25] = tmp[26] = tmp[27] = 0;
1372     carry += vli_add(result, result, tmp);
1373 
1374     /* d1 */
1375     tmp[0]  = product[28]; tmp[1]  = product[29]; tmp[2]  = product[30]; tmp[3]  = product[31];
1376     tmp[4]  = product[32]; tmp[5]  = product[33]; tmp[6]  = product[34]; tmp[7]  = product[35];
1377     tmp[8]  = product[36]; tmp[9]  = product[37]; tmp[10] = product[38]; tmp[11] = product[39];
1378     tmp[12] = product[40]; tmp[13] = product[41]; tmp[14] = product[42]; tmp[15] = product[43];
1379     tmp[16] = product[44]; tmp[17] = product[45]; tmp[18] = product[46]; tmp[19] = product[47];
1380     tmp[20] = product[48]; tmp[21] = product[49]; tmp[22] = product[50]; tmp[23] = product[51];
1381     tmp[24] = product[52]; tmp[25] = product[53]; tmp[26] = product[54]; tmp[27] = product[55];
1382     carry -= vli_sub(result, result, tmp);
1383 
1384     /* d2 */
1385     tmp[0]  = product[44]; tmp[1]  = product[45]; tmp[2]  = product[46]; tmp[3]  = product[47];
1386     tmp[4]  = product[48]; tmp[5]  = product[49]; tmp[6]  = product[50]; tmp[7]  = product[51];
1387     tmp[8]  = product[52]; tmp[9]  = product[53]; tmp[10] = product[54]; tmp[11] = product[55];
1388     tmp[12] = tmp[13] = tmp[14] = tmp[15] = 0;
1389     tmp[16] = tmp[17] = tmp[18] = tmp[19] = 0;
1390     tmp[20] = tmp[21] = tmp[22] = tmp[23] = 0;
1391     tmp[24] = tmp[25] = tmp[26] = tmp[27] = 0;
1392     carry -= vli_sub(result, result, tmp);
1393 
1394     if (carry < 0) {
1395         do {
1396             carry += vli_add(result, result, curve_p);
1397         } while (carry < 0);
1398     } else {
1399         while (carry || vli_cmp(curve_p, result) != 1) {
1400             carry -= vli_sub(result, result, curve_p);
1401         }
1402     }
1403 }
1404 #elif uECC_WORD_SIZE == 4
1405 void vli_mmod_fast(uint32_t *RESTRICT result, uint32_t *RESTRICT product)
1406 {
1407     uint32_t tmp[uECC_WORDS];
1408     int carry;
1409 
1410     /* t */
1411     vli_set(result, product);
1412 
1413     /* s1 */
1414     tmp[0] = tmp[1] = tmp[2] = 0;
1415     tmp[3] = product[7];
1416     tmp[4] = product[8];
1417     tmp[5] = product[9];
1418     tmp[6] = product[10];
1419     carry = vli_add(result, result, tmp);
1420 
1421     /* s2 */
1422     tmp[3] = product[11];
1423     tmp[4] = product[12];
1424     tmp[5] = product[13];
1425     tmp[6] = 0;
1426     carry += vli_add(result, result, tmp);
1427 
1428     /* d1 */
1429     tmp[0] = product[7];
1430     tmp[1] = product[8];
1431     tmp[2] = product[9];
1432     tmp[3] = product[10];
1433     tmp[4] = product[11];
1434     tmp[5] = product[12];
1435     tmp[6] = product[13];
1436     carry -= vli_sub(result, result, tmp);
1437 
1438     /* d2 */
1439     tmp[0] = product[11];
1440     tmp[1] = product[12];
1441     tmp[2] = product[13];
1442     tmp[3] = tmp[4] = tmp[5] = tmp[6] = 0;
1443     carry -= vli_sub(result, result, tmp);
1444 
1445     if (carry < 0) {
1446         do {
1447             carry += vli_add(result, result, curve_p);
1448         } while (carry < 0);
1449     } else {
1450         while (carry || vli_cmp(curve_p, result) != 1) {
1451             carry -= vli_sub(result, result, curve_p);
1452         }
1453     }
1454 }
1455 #endif /* uECC_WORD_SIZE */
1456 
1457 #endif /* uECC_CURVE */
1458 #endif /* !asm_mmod_fast */
1459 
1460 /* Computes result = (left * right) % curve_p. */
1461 static void vli_modMult_fast(uECC_word_t *result,
1462                              const uECC_word_t *left,
1463                              const uECC_word_t *right) {
1464     uECC_word_t product[2 * uECC_WORDS];
1465     vli_mult(product, left, right);
1466     vli_mmod_fast(result, product);
1467 }
1468 
1469 #if uECC_SQUARE_FUNC
1470 
1471 /* Computes result = left^2 % curve_p. */
1472 static void vli_modSquare_fast(uECC_word_t *result, const uECC_word_t *left) {
1473     uECC_word_t product[2 * uECC_WORDS];
1474     vli_square(product, left);
1475     vli_mmod_fast(result, product);
1476 }
1477 
1478 #else /* uECC_SQUARE_FUNC */
1479 
1480 #define vli_modSquare_fast(result, left) vli_modMult_fast((result), (left), (left))
1481 
1482 #endif /* uECC_SQUARE_FUNC */
1483 
1484 
1485 #define EVEN(vli) (!(vli[0] & 1))
1486 /* Computes result = (1 / input) % mod. All VLIs are the same size.
1487    See "From Euclid's GCD to Montgomery Multiplication to the Great Divide"
1488    https://labs.oracle.com/techrep/2001/smli_tr-2001-95.pdf */
1489 #if !asm_modInv
1490 static void vli_modInv(uECC_word_t *result, const uECC_word_t *input, const uECC_word_t *mod) {
1491     uECC_word_t a[uECC_WORDS], b[uECC_WORDS], u[uECC_WORDS], v[uECC_WORDS];
1492     uECC_word_t carry;
1493     cmpresult_t cmpResult;
1494 
1495     if (vli_isZero(input)) {
1496         vli_clear(result);
1497         return;
1498     }
1499 
1500     vli_set(a, input);
1501     vli_set(b, mod);
1502     vli_clear(u);
1503     u[0] = 1;
1504     vli_clear(v);
1505     while ((cmpResult = vli_cmp(a, b)) != 0) {
1506         carry = 0;
1507         if (EVEN(a)) {
1508             vli_rshift1(a);
1509             if (!EVEN(u)) {
1510                 carry = vli_add(u, u, mod);
1511             }
1512             vli_rshift1(u);
1513             if (carry) {
1514                 u[uECC_WORDS - 1] |= HIGH_BIT_SET;
1515             }
1516         } else if (EVEN(b)) {
1517             vli_rshift1(b);
1518             if (!EVEN(v)) {
1519                 carry = vli_add(v, v, mod);
1520             }
1521             vli_rshift1(v);
1522             if (carry) {
1523                 v[uECC_WORDS - 1] |= HIGH_BIT_SET;
1524             }
1525         } else if (cmpResult > 0) {
1526             vli_sub(a, a, b);
1527             vli_rshift1(a);
1528             if (vli_cmp(u, v) < 0) {
1529                 vli_add(u, u, mod);
1530             }
1531             vli_sub(u, u, v);
1532             if (!EVEN(u)) {
1533                 carry = vli_add(u, u, mod);
1534             }
1535             vli_rshift1(u);
1536             if (carry) {
1537                 u[uECC_WORDS - 1] |= HIGH_BIT_SET;
1538             }
1539         } else {
1540             vli_sub(b, b, a);
1541             vli_rshift1(b);
1542             if (vli_cmp(v, u) < 0) {
1543                 vli_add(v, v, mod);
1544             }
1545             vli_sub(v, v, u);
1546             if (!EVEN(v)) {
1547                 carry = vli_add(v, v, mod);
1548             }
1549             vli_rshift1(v);
1550             if (carry) {
1551                 v[uECC_WORDS - 1] |= HIGH_BIT_SET;
1552             }
1553         }
1554     }
1555     vli_set(result, u);
1556 }
1557 #endif /* !asm_modInv */
1558 
1559 /* ------ Point operations ------ */
1560 
1561 /* Returns 1 if 'point' is the point at infinity, 0 otherwise. */
1562 static cmpresult_t EccPoint_isZero(const EccPoint *point) {
1563     return (vli_isZero(point->x) && vli_isZero(point->y));
1564 }
1565 
1566 /* Point multiplication algorithm using Montgomery's ladder with co-Z coordinates.
1567 From http://eprint.iacr.org/2011/338.pdf
1568 */
1569 
1570 /* Double in place */
1571 #if (uECC_CURVE == uECC_secp256k1)
1572 static void EccPoint_double_jacobian(uECC_word_t * RESTRICT X1,
1573                                      uECC_word_t * RESTRICT Y1,
1574                                      uECC_word_t * RESTRICT Z1) {
1575     /* t1 = X, t2 = Y, t3 = Z */
1576     uECC_word_t t4[uECC_WORDS];
1577     uECC_word_t t5[uECC_WORDS];
1578 
1579     if (vli_isZero(Z1)) {
1580         return;
1581     }
1582 
1583     vli_modSquare_fast(t5, Y1);   /* t5 = y1^2 */
1584     vli_modMult_fast(t4, X1, t5); /* t4 = x1*y1^2 = A */
1585     vli_modSquare_fast(X1, X1);   /* t1 = x1^2 */
1586     vli_modSquare_fast(t5, t5);   /* t5 = y1^4 */
1587     vli_modMult_fast(Z1, Y1, Z1); /* t3 = y1*z1 = z3 */
1588 
1589     vli_modAdd(Y1, X1, X1, curve_p); /* t2 = 2*x1^2 */
1590     vli_modAdd(Y1, Y1, X1, curve_p); /* t2 = 3*x1^2 */
1591     if (vli_testBit(Y1, 0)) {
1592         uECC_word_t carry = vli_add(Y1, Y1, curve_p);
1593         vli_rshift1(Y1);
1594         Y1[uECC_WORDS - 1] |= carry << (uECC_WORD_BITS - 1);
1595     } else {
1596         vli_rshift1(Y1);
1597     }
1598     /* t2 = 3/2*(x1^2) = B */
1599 
1600     vli_modSquare_fast(X1, Y1);      /* t1 = B^2 */
1601     vli_modSub(X1, X1, t4, curve_p); /* t1 = B^2 - A */
1602     vli_modSub(X1, X1, t4, curve_p); /* t1 = B^2 - 2A = x3 */
1603 
1604     vli_modSub(t4, t4, X1, curve_p); /* t4 = A - x3 */
1605     vli_modMult_fast(Y1, Y1, t4);    /* t2 = B * (A - x3) */
1606     vli_modSub(Y1, Y1, t5, curve_p); /* t2 = B * (A - x3) - y1^4 = y3 */
1607 }
1608 #else
1609 static void EccPoint_double_jacobian(uECC_word_t * RESTRICT X1,
1610                                      uECC_word_t * RESTRICT Y1,
1611                                      uECC_word_t * RESTRICT Z1) {
1612     /* t1 = X, t2 = Y, t3 = Z */
1613     uECC_word_t t4[uECC_WORDS];
1614     uECC_word_t t5[uECC_WORDS];
1615 
1616     if (vli_isZero(Z1)) {
1617         return;
1618     }
1619 
1620     vli_modSquare_fast(t4, Y1);   /* t4 = y1^2 */
1621     vli_modMult_fast(t5, X1, t4); /* t5 = x1*y1^2 = A */
1622     vli_modSquare_fast(t4, t4);   /* t4 = y1^4 */
1623     vli_modMult_fast(Y1, Y1, Z1); /* t2 = y1*z1 = z3 */
1624     vli_modSquare_fast(Z1, Z1);   /* t3 = z1^2 */
1625 
1626     vli_modAdd(X1, X1, Z1, curve_p); /* t1 = x1 + z1^2 */
1627     vli_modAdd(Z1, Z1, Z1, curve_p); /* t3 = 2*z1^2 */
1628     vli_modSub_fast(Z1, X1, Z1);     /* t3 = x1 - z1^2 */
1629     vli_modMult_fast(X1, X1, Z1);    /* t1 = x1^2 - z1^4 */
1630 
1631     vli_modAdd(Z1, X1, X1, curve_p); /* t3 = 2*(x1^2 - z1^4) */
1632     vli_modAdd(X1, X1, Z1, curve_p); /* t1 = 3*(x1^2 - z1^4) */
1633     if (vli_testBit(X1, 0)) {
1634         uECC_word_t l_carry = vli_add(X1, X1, curve_p);
1635         vli_rshift1(X1);
1636         X1[uECC_WORDS - 1] |= l_carry << (uECC_WORD_BITS - 1);
1637     } else {
1638         vli_rshift1(X1);
1639     }
1640     /* t1 = 3/2*(x1^2 - z1^4) = B */
1641 
1642     vli_modSquare_fast(Z1, X1);   /* t3 = B^2 */
1643     vli_modSub_fast(Z1, Z1, t5);  /* t3 = B^2 - A */
1644     vli_modSub_fast(Z1, Z1, t5);  /* t3 = B^2 - 2A = x3 */
1645     vli_modSub_fast(t5, t5, Z1);  /* t5 = A - x3 */
1646     vli_modMult_fast(X1, X1, t5); /* t1 = B * (A - x3) */
1647     vli_modSub_fast(t4, X1, t4);  /* t4 = B * (A - x3) - y1^4 = y3 */
1648 
1649     vli_set(X1, Z1);
1650     vli_set(Z1, Y1);
1651     vli_set(Y1, t4);
1652 }
1653 #endif
1654 
1655 /* Modify (x1, y1) => (x1 * z^2, y1 * z^3) */
1656 static void apply_z(uECC_word_t * RESTRICT X1,
1657                     uECC_word_t * RESTRICT Y1,
1658                     const uECC_word_t * RESTRICT Z) {
1659     uECC_word_t t1[uECC_WORDS];
1660 
1661     vli_modSquare_fast(t1, Z);    /* z^2 */
1662     vli_modMult_fast(X1, X1, t1); /* x1 * z^2 */
1663     vli_modMult_fast(t1, t1, Z);  /* z^3 */
1664     vli_modMult_fast(Y1, Y1, t1); /* y1 * z^3 */
1665 }
1666 
1667 /* P = (x1, y1) => 2P, (x2, y2) => P' */
1668 static void XYcZ_initial_double(uECC_word_t * RESTRICT X1,
1669                                 uECC_word_t * RESTRICT Y1,
1670                                 uECC_word_t * RESTRICT X2,
1671                                 uECC_word_t * RESTRICT Y2,
1672                                 const uECC_word_t * RESTRICT initial_Z) {
1673     uECC_word_t z[uECC_WORDS];
1674     if (initial_Z) {
1675         vli_set(z, initial_Z);
1676     } else {
1677         vli_clear(z);
1678         z[0] = 1;
1679     }
1680 
1681     vli_set(X2, X1);
1682     vli_set(Y2, Y1);
1683 
1684     apply_z(X1, Y1, z);
1685     EccPoint_double_jacobian(X1, Y1, z);
1686     apply_z(X2, Y2, z);
1687 }
1688 
1689 /* Input P = (x1, y1, Z), Q = (x2, y2, Z)
1690    Output P' = (x1', y1', Z3), P + Q = (x3, y3, Z3)
1691    or P => P', Q => P + Q
1692 */
1693 static void XYcZ_add(uECC_word_t * RESTRICT X1,
1694                      uECC_word_t * RESTRICT Y1,
1695                      uECC_word_t * RESTRICT X2,
1696                      uECC_word_t * RESTRICT Y2) {
1697     /* t1 = X1, t2 = Y1, t3 = X2, t4 = Y2 */
1698     uECC_word_t t5[uECC_WORDS];
1699 
1700     vli_modSub_fast(t5, X2, X1);  /* t5 = x2 - x1 */
1701     vli_modSquare_fast(t5, t5);   /* t5 = (x2 - x1)^2 = A */
1702     vli_modMult_fast(X1, X1, t5); /* t1 = x1*A = B */
1703     vli_modMult_fast(X2, X2, t5); /* t3 = x2*A = C */
1704     vli_modSub_fast(Y2, Y2, Y1);  /* t4 = y2 - y1 */
1705     vli_modSquare_fast(t5, Y2);   /* t5 = (y2 - y1)^2 = D */
1706 
1707     vli_modSub_fast(t5, t5, X1);  /* t5 = D - B */
1708     vli_modSub_fast(t5, t5, X2);  /* t5 = D - B - C = x3 */
1709     vli_modSub_fast(X2, X2, X1);  /* t3 = C - B */
1710     vli_modMult_fast(Y1, Y1, X2); /* t2 = y1*(C - B) */
1711     vli_modSub_fast(X2, X1, t5);  /* t3 = B - x3 */
1712     vli_modMult_fast(Y2, Y2, X2); /* t4 = (y2 - y1)*(B - x3) */
1713     vli_modSub_fast(Y2, Y2, Y1);  /* t4 = y3 */
1714 
1715     vli_set(X2, t5);
1716 }
1717 
1718 /* Input P = (x1, y1, Z), Q = (x2, y2, Z)
1719    Output P + Q = (x3, y3, Z3), P - Q = (x3', y3', Z3)
1720    or P => P - Q, Q => P + Q
1721 */
1722 static void XYcZ_addC(uECC_word_t * RESTRICT X1,
1723                       uECC_word_t * RESTRICT Y1,
1724                       uECC_word_t * RESTRICT X2,
1725                       uECC_word_t * RESTRICT Y2) {
1726     /* t1 = X1, t2 = Y1, t3 = X2, t4 = Y2 */
1727     uECC_word_t t5[uECC_WORDS];
1728     uECC_word_t t6[uECC_WORDS];
1729     uECC_word_t t7[uECC_WORDS];
1730 
1731     vli_modSub_fast(t5, X2, X1);     /* t5 = x2 - x1 */
1732     vli_modSquare_fast(t5, t5);      /* t5 = (x2 - x1)^2 = A */
1733     vli_modMult_fast(X1, X1, t5);    /* t1 = x1*A = B */
1734     vli_modMult_fast(X2, X2, t5);    /* t3 = x2*A = C */
1735     vli_modAdd(t5, Y2, Y1, curve_p); /* t5 = y2 + y1 */
1736     vli_modSub_fast(Y2, Y2, Y1);     /* t4 = y2 - y1 */
1737 
1738     vli_modSub_fast(t6, X2, X1);     /* t6 = C - B */
1739     vli_modMult_fast(Y1, Y1, t6);    /* t2 = y1 * (C - B) = E */
1740     vli_modAdd(t6, X1, X2, curve_p); /* t6 = B + C */
1741     vli_modSquare_fast(X2, Y2);      /* t3 = (y2 - y1)^2 = D */
1742     vli_modSub_fast(X2, X2, t6);     /* t3 = D - (B + C) = x3 */
1743 
1744     vli_modSub_fast(t7, X1, X2);  /* t7 = B - x3 */
1745     vli_modMult_fast(Y2, Y2, t7); /* t4 = (y2 - y1)*(B - x3) */
1746     vli_modSub_fast(Y2, Y2, Y1);  /* t4 = (y2 - y1)*(B - x3) - E = y3 */
1747 
1748     vli_modSquare_fast(t7, t5);   /* t7 = (y2 + y1)^2 = F */
1749     vli_modSub_fast(t7, t7, t6);  /* t7 = F - (B + C) = x3' */
1750     vli_modSub_fast(t6, t7, X1);  /* t6 = x3' - B */
1751     vli_modMult_fast(t6, t6, t5); /* t6 = (y2 + y1)*(x3' - B) */
1752     vli_modSub_fast(Y1, t6, Y1);  /* t2 = (y2 + y1)*(x3' - B) - E = y3' */
1753 
1754     vli_set(X1, t7);
1755 }
1756 
1757 static void EccPoint_mult(EccPoint * RESTRICT result,
1758                           const EccPoint * RESTRICT point,
1759                           const uECC_word_t * RESTRICT scalar,
1760                           const uECC_word_t * RESTRICT initialZ,
1761                           bitcount_t numBits) {
1762     /* R0 and R1 */
1763     uECC_word_t Rx[2][uECC_WORDS];
1764     uECC_word_t Ry[2][uECC_WORDS];
1765     uECC_word_t z[uECC_WORDS];
1766     bitcount_t i;
1767     uECC_word_t nb;
1768 
1769     vli_set(Rx[1], point->x);
1770     vli_set(Ry[1], point->y);
1771 
1772     XYcZ_initial_double(Rx[1], Ry[1], Rx[0], Ry[0], initialZ);
1773 
1774     for (i = numBits - 2; i > 0; --i) {
1775         nb = !vli_testBit(scalar, i);
1776         XYcZ_addC(Rx[1 - nb], Ry[1 - nb], Rx[nb], Ry[nb]);
1777         XYcZ_add(Rx[nb], Ry[nb], Rx[1 - nb], Ry[1 - nb]);
1778     }
1779 
1780     nb = !vli_testBit(scalar, 0);
1781     XYcZ_addC(Rx[1 - nb], Ry[1 - nb], Rx[nb], Ry[nb]);
1782 
1783     /* Find final 1/Z value. */
1784     vli_modSub_fast(z, Rx[1], Rx[0]);   /* X1 - X0 */
1785     vli_modMult_fast(z, z, Ry[1 - nb]); /* Yb * (X1 - X0) */
1786     vli_modMult_fast(z, z, point->x); /* xP * Yb * (X1 - X0) */
1787     vli_modInv(z, z, curve_p);          /* 1 / (xP * Yb * (X1 - X0)) */
1788     vli_modMult_fast(z, z, point->y); /* yP / (xP * Yb * (X1 - X0)) */
1789     vli_modMult_fast(z, z, Rx[1 - nb]); /* Xb * yP / (xP * Yb * (X1 - X0)) */
1790     /* End 1/Z calculation */
1791 
1792     XYcZ_add(Rx[nb], Ry[nb], Rx[1 - nb], Ry[1 - nb]);
1793     apply_z(Rx[0], Ry[0], z);
1794 
1795     vli_set(result->x, Rx[0]);
1796     vli_set(result->y, Ry[0]);
1797 }
1798 
1799 static int EccPoint_compute_public_key(EccPoint *result, uECC_word_t *private) {
1800     uECC_word_t tmp1[uECC_WORDS];
1801     uECC_word_t tmp2[uECC_WORDS];
1802     uECC_word_t *p2[2] = {tmp1, tmp2};
1803     uECC_word_t carry;
1804 
1805     /* Make sure the private key is in the range [1, n-1]. */
1806     if (vli_isZero(private)) {
1807         return 0;
1808     }
1809 
1810 #if (uECC_CURVE == uECC_secp160r1)
1811     // Don't regularize the bitcount for secp160r1, since it would have a larger performance
1812     // impact (about 2% slower on average) and requires the vli_xxx_n functions, leading to
1813     // a significant increase in code size.
1814 
1815     EccPoint_mult(result, &curve_G, private, NULL, vli_numBits(private, uECC_WORDS));
1816 #else
1817     if (vli_cmp(curve_n, private) != 1) {
1818         return 0;
1819     }
1820 
1821     // Regularize the bitcount for the private key so that attackers cannot use a side channel
1822     // attack to learn the number of leading zeros.
1823     carry = vli_add(tmp1, private, curve_n);
1824     vli_add(tmp2, tmp1, curve_n);
1825     EccPoint_mult(result, &curve_G, p2[!carry], NULL, (uECC_BYTES * 8) + 1);
1826 #endif
1827 
1828     if (EccPoint_isZero(result)) {
1829         return 0;
1830     }
1831     return 1;
1832 }
1833 
1834 #ifdef ENABLE_MICRO_ECC_COMPRESSION
1835 
1836 #if uECC_CURVE == uECC_secp224r1
1837 
1838 /* Routine 3.2.4 RS;  from http://www.nsa.gov/ia/_files/nist-routines.pdf */
1839 static void mod_sqrt_secp224r1_rs(uECC_word_t *d1,
1840                                   uECC_word_t *e1,
1841                                   uECC_word_t *f1,
1842                                   const uECC_word_t *d0,
1843                                   const uECC_word_t *e0,
1844                                   const uECC_word_t *f0) {
1845     uECC_word_t t[uECC_WORDS];
1846 
1847     vli_modSquare_fast(t, d0);                 /* t <-- d0 ^ 2 */
1848     vli_modMult_fast(e1, d0, e0);              /* e1 <-- d0 * e0 */
1849     vli_modAdd(d1, t, f0, curve_p);            /* d1 <-- t  + f0 */
1850     vli_modAdd(e1, e1, e1, curve_p);           /* e1 <-- e1 + e1 */
1851     vli_modMult_fast(f1, t, f0);               /* f1 <-- t  * f0 */
1852     vli_modAdd(f1, f1, f1, curve_p);           /* f1 <-- f1 + f1 */
1853     vli_modAdd(f1, f1, f1, curve_p);           /* f1 <-- f1 + f1 */
1854 }
1855 
1856 /* Routine 3.2.5 RSS;  from http://www.nsa.gov/ia/_files/nist-routines.pdf */
1857 static void mod_sqrt_secp224r1_rss(uECC_word_t *d1,
1858                                    uECC_word_t *e1,
1859                                    uECC_word_t *f1,
1860                                    const uECC_word_t *d0,
1861                                    const uECC_word_t *e0,
1862                                    const uECC_word_t *f0,
1863                                    const bitcount_t j) {
1864     bitcount_t i;
1865 
1866     vli_set(d1, d0);                           /* d1 <-- d0 */
1867     vli_set(e1, e0);                           /* e1 <-- e0 */
1868     vli_set(f1, f0);                           /* f1 <-- f0 */
1869     for (i = 1; i <= j; i++) {
1870         mod_sqrt_secp224r1_rs(d1, e1, f1, d1, e1, f1); /* RS (d1,e1,f1,d1,e1,f1) */
1871     }
1872 }
1873 
1874 /* Routine 3.2.6 RM;  from http://www.nsa.gov/ia/_files/nist-routines.pdf */
1875 static void mod_sqrt_secp224r1_rm(uECC_word_t *d2,
1876                                   uECC_word_t *e2,
1877                                   uECC_word_t *f2,
1878                                   const uECC_word_t *c,
1879                                   const uECC_word_t *d0,
1880                                   const uECC_word_t *e0,
1881                                   const uECC_word_t *d1,
1882                                   const uECC_word_t *e1) {
1883     uECC_word_t t1[uECC_WORDS];
1884     uECC_word_t t2[uECC_WORDS];
1885 
1886     vli_modMult_fast(t1, e0, e1);              /* t1 <-- e0 * e1 */
1887     vli_modMult_fast(t1, t1, c);               /* t1 <-- t1 * c */
1888     vli_modSub_fast(t1, curve_p, t1);          /* t1 <-- p  - t1 */
1889     vli_modMult_fast(t2, d0, d1);              /* t2 <-- d0 * d1 */
1890     vli_modAdd(t2, t2, t1, curve_p);           /* t2 <-- t2 + t1 */
1891     vli_modMult_fast(t1, d0, e1);              /* t1 <-- d0 * e1 */
1892     vli_modMult_fast(e2, d1, e0);              /* e2 <-- d1 * e0 */
1893     vli_modAdd(e2, e2, t1, curve_p);           /* e2 <-- e2 + t1 */
1894     vli_modSquare_fast(f2, e2);                /* f2 <-- e2^2 */
1895     vli_modMult_fast(f2, f2, c);               /* f2 <-- f2 * c */
1896     vli_modSub_fast(f2, curve_p, f2);          /* f2 <-- p  - f2 */
1897     vli_set(d2, t2);                           /* d2 <-- t2 */
1898 }
1899 
1900 /* Routine 3.2.7 RP;  from http://www.nsa.gov/ia/_files/nist-routines.pdf */
1901 static void mod_sqrt_secp224r1_rp(uECC_word_t *d1,
1902                                   uECC_word_t *e1,
1903                                   uECC_word_t *f1,
1904                                   const uECC_word_t *c,
1905                                   const uECC_word_t *r) {
1906     wordcount_t i;
1907     wordcount_t pow2i = 1;
1908     uECC_word_t d0[uECC_WORDS];
1909     uECC_word_t e0[uECC_WORDS] = {1};          /* e0 <-- 1 */
1910     uECC_word_t f0[uECC_WORDS];
1911 
1912     vli_set(d0, r);                            /* d0 <-- r */
1913     vli_modSub_fast(f0, curve_p, c);           /* f0 <-- p  - c */
1914     for (i = 0; i <= 6; i++) {
1915         mod_sqrt_secp224r1_rss(d1, e1, f1, d0, e0, f0, pow2i); /* RSS (d1,e1,f1,d0,e0,f0,2^i) */
1916         mod_sqrt_secp224r1_rm(d1, e1, f1, c, d1, e1, d0, e0);  /* RM (d1,e1,f1,c,d1,e1,d0,e0) */
1917         vli_set(d0, d1);                       /* d0 <-- d1 */
1918         vli_set(e0, e1);                       /* e0 <-- e1 */
1919         vli_set(f0, f1);                       /* f0 <-- f1 */
1920         pow2i *= 2;
1921     }
1922 }
1923 
1924 /* Compute a = sqrt(a) (mod curve_p). */
1925 /* Routine 3.2.8 mp_mod_sqrt_224; from http://www.nsa.gov/ia/_files/nist-routines.pdf */
1926 static void mod_sqrt(uECC_word_t *a) {
1927     bitcount_t i;
1928     uECC_word_t e1[uECC_WORDS];
1929     uECC_word_t f1[uECC_WORDS];
1930     uECC_word_t d0[uECC_WORDS];
1931     uECC_word_t e0[uECC_WORDS];
1932     uECC_word_t f0[uECC_WORDS];
1933     uECC_word_t d1[uECC_WORDS];
1934 
1935     // s = a; using constant instead of random value
1936     mod_sqrt_secp224r1_rp(d0, e0, f0, a, a);           /* RP (d0, e0, f0, c, s) */
1937     mod_sqrt_secp224r1_rs(d1, e1, f1, d0, e0, f0);     /* RS (d1, e1, f1, d0, e0, f0) */
1938     for (i = 1; i <= 95; i++) {
1939         vli_set(d0, d1);                               /* d0 <-- d1 */
1940         vli_set(e0, e1);                               /* e0 <-- e1 */
1941         vli_set(f0, f1);                               /* f0 <-- f1 */
1942         mod_sqrt_secp224r1_rs(d1, e1, f1, d0, e0, f0); /* RS (d1, e1, f1, d0, e0, f0) */
1943         if (vli_isZero(d1)) {                          /* if d1 == 0 */
1944 	        break;
1945         }
1946     }
1947     vli_modInv(f1, e0, curve_p);                       /* f1 <-- 1 / e0 */
1948     vli_modMult_fast(a, d0, f1);                       /* a  <-- d0 / e0 */
1949 }
1950 
1951 #else /* uECC_CURVE */
1952 
1953 /* Compute a = sqrt(a) (mod curve_p). */
1954 static void mod_sqrt(uECC_word_t *a) {
1955     bitcount_t i;
1956     uECC_word_t p1[uECC_WORDS] = {1};
1957     uECC_word_t l_result[uECC_WORDS] = {1};
1958 
1959     /* Since curve_p == 3 (mod 4) for all supported curves, we can
1960        compute sqrt(a) = a^((curve_p + 1) / 4) (mod curve_p). */
1961     vli_add(p1, curve_p, p1); /* p1 = curve_p + 1 */
1962     for (i = vli_numBits(p1, uECC_WORDS) - 1; i > 1; --i) {
1963         vli_modSquare_fast(l_result, l_result);
1964         if (vli_testBit(p1, i)) {
1965             vli_modMult_fast(l_result, l_result, a);
1966         }
1967     }
1968     vli_set(a, l_result);
1969 }
1970 #endif /* uECC_CURVE */
1971 
1972 #endif /* ENABLE_MICRO_ECC_COMPRESSION */
1973 
1974 
1975 #if uECC_WORD_SIZE == 1
1976 
1977 static void vli_nativeToBytes(uint8_t * RESTRICT dest, const uint8_t * RESTRICT src) {
1978     uint8_t i;
1979     for (i = 0; i < uECC_BYTES; ++i) {
1980         dest[i] = src[(uECC_BYTES - 1) - i];
1981     }
1982 }
1983 
1984 #define vli_bytesToNative(dest, src) vli_nativeToBytes((dest), (src))
1985 
1986 #elif uECC_WORD_SIZE == 4
1987 
1988 static void vli_nativeToBytes(uint8_t *bytes, const uint32_t *native) {
1989     unsigned i;
1990     for (i = 0; i < uECC_WORDS; ++i) {
1991         uint8_t *digit = bytes + 4 * (uECC_WORDS - 1 - i);
1992         digit[0] = native[i] >> 24;
1993         digit[1] = native[i] >> 16;
1994         digit[2] = native[i] >> 8;
1995         digit[3] = native[i];
1996     }
1997 }
1998 
1999 static void vli_bytesToNative(uint32_t *native, const uint8_t *bytes) {
2000     unsigned i;
2001     for (i = 0; i < uECC_WORDS; ++i) {
2002         const uint8_t *digit = bytes + 4 * (uECC_WORDS - 1 - i);
2003         native[i] = ((uint32_t)digit[0] << 24) | ((uint32_t)digit[1] << 16) |
2004                     ((uint32_t)digit[2] << 8) | (uint32_t)digit[3];
2005     }
2006 }
2007 
2008 #else
2009 
2010 static void vli_nativeToBytes(uint8_t *bytes, const uint64_t *native) {
2011     unsigned i;
2012     for (i = 0; i < uECC_WORDS; ++i) {
2013         uint8_t *digit = bytes + 8 * (uECC_WORDS - 1 - i);
2014         digit[0] = native[i] >> 56;
2015         digit[1] = native[i] >> 48;
2016         digit[2] = native[i] >> 40;
2017         digit[3] = native[i] >> 32;
2018         digit[4] = native[i] >> 24;
2019         digit[5] = native[i] >> 16;
2020         digit[6] = native[i] >> 8;
2021         digit[7] = native[i];
2022     }
2023 }
2024 
2025 static void vli_bytesToNative(uint64_t *native, const uint8_t *bytes) {
2026     unsigned i;
2027     for (i = 0; i < uECC_WORDS; ++i) {
2028         const uint8_t *digit = bytes + 8 * (uECC_WORDS - 1 - i);
2029         native[i] = ((uint64_t)digit[0] << 56) | ((uint64_t)digit[1] << 48) |
2030                     ((uint64_t)digit[2] << 40) | ((uint64_t)digit[3] << 32) |
2031                     ((uint64_t)digit[4] << 24) | ((uint64_t)digit[5] << 16) |
2032                     ((uint64_t)digit[6] << 8) | (uint64_t)digit[7];
2033     }
2034 }
2035 
2036 #endif /* uECC_WORD_SIZE */
2037 
2038 int uECC_make_key(uint8_t public_key[uECC_BYTES*2], uint8_t private_key[uECC_BYTES]) {
2039     uECC_word_t private[uECC_WORDS];
2040     EccPoint public;
2041     uECC_word_t tries;
2042     for (tries = 0; tries < MAX_TRIES; ++tries) {
2043         if (g_rng_function((uint8_t *)private, sizeof(private)) &&
2044                 EccPoint_compute_public_key(&public, private)) {
2045             vli_nativeToBytes(private_key, private);
2046             vli_nativeToBytes(public_key, public.x);
2047             vli_nativeToBytes(public_key + uECC_BYTES, public.y);
2048             return 1;
2049         }
2050     }
2051     return 0;
2052 }
2053 
2054 int uECC_shared_secret(const uint8_t public_key[uECC_BYTES*2],
2055                        const uint8_t private_key[uECC_BYTES],
2056                        uint8_t secret[uECC_BYTES]) {
2057     EccPoint public;
2058     EccPoint product;
2059     uECC_word_t private[uECC_WORDS];
2060     uECC_word_t tmp[uECC_WORDS];
2061     uECC_word_t *p2[2] = {private, tmp};
2062     uECC_word_t random[uECC_WORDS];
2063     uECC_word_t *initial_Z = NULL;
2064     uECC_word_t tries;
2065     uECC_word_t carry;
2066 
2067     // Try to get a random initial Z value to improve protection against side-channel
2068     // attacks. If the RNG fails every time (eg it was not defined), we continue so that
2069     // uECC_shared_secret() can still work without an RNG defined.
2070     for (tries = 0; tries < MAX_TRIES; ++tries) {
2071         if (g_rng_function((uint8_t *)random, sizeof(random)) && !vli_isZero(random)) {
2072             initial_Z = random;
2073             break;
2074         }
2075     }
2076 
2077     vli_bytesToNative(private, private_key);
2078     vli_bytesToNative(public.x, public_key);
2079     vli_bytesToNative(public.y, public_key + uECC_BYTES);
2080 
2081 #if (uECC_CURVE == uECC_secp160r1)
2082     // Don't regularize the bitcount for secp160r1.
2083     EccPoint_mult(&product, &public, private, initial_Z, vli_numBits(private, uECC_WORDS));
2084 #else
2085     // Regularize the bitcount for the private key so that attackers cannot use a side channel
2086     // attack to learn the number of leading zeros.
2087     carry = vli_add(private, private, curve_n);
2088     vli_add(tmp, private, curve_n);
2089     EccPoint_mult(&product, &public, p2[!carry], initial_Z, (uECC_BYTES * 8) + 1);
2090 #endif
2091 
2092     vli_nativeToBytes(secret, product.x);
2093     return !EccPoint_isZero(&product);
2094 }
2095 
2096 #ifdef ENABLE_MICRO_ECC_COMPRESSION
2097 
2098 void uECC_compress(const uint8_t public_key[uECC_BYTES*2], uint8_t compressed[uECC_BYTES+1]) {
2099     wordcount_t i;
2100     for (i = 0; i < uECC_BYTES; ++i) {
2101         compressed[i+1] = public_key[i];
2102     }
2103     compressed[0] = 2 + (public_key[uECC_BYTES * 2 - 1] & 0x01);
2104 }
2105 
2106 #endif
2107 
2108 /* Computes result = x^3 + ax + b. result must not overlap x. */
2109 static void curve_x_side(uECC_word_t * RESTRICT result, const uECC_word_t * RESTRICT x) {
2110     static const uECC_word_t curve_b[uECC_WORDS] = uECC_CONCAT(Curve_B_, uECC_CURVE);
2111 #if (uECC_CURVE == uECC_secp256k1)
2112     vli_modSquare_fast(result, x); /* r = x^2 */
2113     vli_modMult_fast(result, result, x); /* r = x^3 */
2114     vli_modAdd(result, result, curve_b, curve_p); /* r = x^3 + b */
2115 #else
2116     uECC_word_t _3[uECC_WORDS] = {3}; /* -a = 3 */
2117 
2118     vli_modSquare_fast(result, x); /* r = x^2 */
2119     vli_modSub_fast(result, result, _3); /* r = x^2 - 3 */
2120     vli_modMult_fast(result, result, x); /* r = x^3 - 3x */
2121     vli_modAdd(result, result, curve_b, curve_p); /* r = x^3 - 3x + b */
2122 #endif
2123 }
2124 
2125 #ifdef ENABLE_MICRO_ECC_COMPRESSION
2126 
2127 void uECC_decompress(const uint8_t compressed[uECC_BYTES+1], uint8_t public_key[uECC_BYTES*2]) {
2128     EccPoint point;
2129     vli_bytesToNative(point.x, compressed + 1);
2130     curve_x_side(point.y, point.x);
2131     mod_sqrt(point.y);
2132 
2133     if ((point.y[0] & 0x01) != (compressed[0] & 0x01)) {
2134         vli_sub(point.y, curve_p, point.y);
2135     }
2136 
2137     vli_nativeToBytes(public_key, point.x);
2138     vli_nativeToBytes(public_key + uECC_BYTES, point.y);
2139 }
2140 
2141 #endif /* ENABLE_MICRO_ECC_COMPRESSION */
2142 
2143 int uECC_valid_public_key(const uint8_t public_key[uECC_BYTES*2]) {
2144     uECC_word_t tmp1[uECC_WORDS];
2145     uECC_word_t tmp2[uECC_WORDS];
2146     EccPoint public;
2147 
2148     vli_bytesToNative(public.x, public_key);
2149     vli_bytesToNative(public.y, public_key + uECC_BYTES);
2150 
2151     // The point at infinity is invalid.
2152     if (EccPoint_isZero(&public)) {
2153         return 0;
2154     }
2155 
2156     // x and y must be smaller than p.
2157     if (vli_cmp(curve_p, public.x) != 1 || vli_cmp(curve_p, public.y) != 1) {
2158         return 0;
2159     }
2160 
2161     vli_modSquare_fast(tmp1, public.y); /* tmp1 = y^2 */
2162     curve_x_side(tmp2, public.x); /* tmp2 = x^3 + ax + b */
2163 
2164     /* Make sure that y^2 == x^3 + ax + b */
2165     return (vli_cmp(tmp1, tmp2) == 0);
2166 }
2167 
2168 int uECC_compute_public_key(const uint8_t private_key[uECC_BYTES],
2169                             uint8_t public_key[uECC_BYTES * 2]) {
2170     uECC_word_t private[uECC_WORDS];
2171     EccPoint public;
2172 
2173     vli_bytesToNative(private, private_key);
2174 
2175     if (!EccPoint_compute_public_key(&public, private)) {
2176         return 0;
2177     }
2178 
2179     vli_nativeToBytes(public_key, public.x);
2180     vli_nativeToBytes(public_key + uECC_BYTES, public.y);
2181     return 1;
2182 }
2183 
2184 int uECC_bytes(void) {
2185     return uECC_BYTES;
2186 }
2187 
2188 int uECC_curve(void) {
2189     return uECC_CURVE;
2190 }
2191 
2192 /* -------- ECDSA code -------- */
2193 
2194 #ifdef ENABLE_MICRO_ECC_ECDSA
2195 
2196 #if (uECC_CURVE == uECC_secp160r1)
2197 static void vli_clear_n(uECC_word_t *vli) {
2198     vli_clear(vli);
2199     vli[uECC_N_WORDS - 1] = 0;
2200 }
2201 
2202 static uECC_word_t vli_isZero_n(const uECC_word_t *vli) {
2203     if (vli[uECC_N_WORDS - 1]) {
2204         return 0;
2205     }
2206     return vli_isZero(vli);
2207 }
2208 
2209 static void vli_set_n(uECC_word_t *dest, const uECC_word_t *src) {
2210     vli_set(dest, src);
2211     dest[uECC_N_WORDS - 1] = src[uECC_N_WORDS - 1];
2212 }
2213 
2214 static cmpresult_t vli_cmp_n(const uECC_word_t *left, const uECC_word_t *right) {
2215     if (left[uECC_N_WORDS - 1] > right[uECC_N_WORDS - 1]) {
2216         return 1;
2217     } else if (left[uECC_N_WORDS - 1] < right[uECC_N_WORDS - 1]) {
2218         return -1;
2219     }
2220     return vli_cmp(left, right);
2221 }
2222 
2223 static void vli_rshift1_n(uECC_word_t *vli) {
2224     vli_rshift1(vli);
2225     vli[uECC_N_WORDS - 2] |= vli[uECC_N_WORDS - 1] << (uECC_WORD_BITS - 1);
2226     vli[uECC_N_WORDS - 1] = vli[uECC_N_WORDS - 1] >> 1;
2227 }
2228 
2229 static uECC_word_t vli_add_n(uECC_word_t *result,
2230                              const uECC_word_t *left,
2231                              const uECC_word_t *right) {
2232     uECC_word_t carry = vli_add(result, left, right);
2233     uECC_word_t sum = left[uECC_N_WORDS - 1] + right[uECC_N_WORDS - 1] + carry;
2234     if (sum != left[uECC_N_WORDS - 1]) {
2235         carry = (sum < left[uECC_N_WORDS - 1]);
2236     }
2237     result[uECC_N_WORDS - 1] = sum;
2238     return carry;
2239 }
2240 
2241 static uECC_word_t vli_sub_n(uECC_word_t *result,
2242                              const uECC_word_t *left,
2243                              const uECC_word_t *right) {
2244     uECC_word_t borrow = vli_sub(result, left, right);
2245     uECC_word_t diff = left[uECC_N_WORDS - 1] - right[uECC_N_WORDS - 1] - borrow;
2246     if (diff != left[uECC_N_WORDS - 1]) {
2247         borrow = (diff > left[uECC_N_WORDS - 1]);
2248     }
2249     result[uECC_N_WORDS - 1] = diff;
2250     return borrow;
2251 }
2252 
2253 #if !muladd_exists
2254 static void muladd(uECC_word_t a,
2255                    uECC_word_t b,
2256                    uECC_word_t *r0,
2257                    uECC_word_t *r1,
2258                    uECC_word_t *r2) {
2259     uECC_dword_t p = (uECC_dword_t)a * b;
2260     uECC_dword_t r01 = ((uECC_dword_t)(*r1) << uECC_WORD_BITS) | *r0;
2261     r01 += p;
2262     *r2 += (r01 < p);
2263     *r1 = r01 >> uECC_WORD_BITS;
2264     *r0 = (uECC_word_t)r01;
2265 }
2266 #define muladd_exists 1
2267 #endif
2268 
2269 static void vli_mult_n(uECC_word_t *result, const uECC_word_t *left, const uECC_word_t *right) {
2270     uECC_word_t r0 = 0;
2271     uECC_word_t r1 = 0;
2272     uECC_word_t r2 = 0;
2273     wordcount_t i, k;
2274 
2275     for (k = 0; k < uECC_N_WORDS * 2 - 1; ++k) {
2276         wordcount_t min = (k < uECC_N_WORDS ? 0 : (k + 1) - uECC_N_WORDS);
2277         wordcount_t max = (k < uECC_N_WORDS ? k : uECC_N_WORDS - 1);
2278         for (i = min; i <= max; ++i) {
2279             muladd(left[i], right[k - i], &r0, &r1, &r2);
2280         }
2281         result[k] = r0;
2282         r0 = r1;
2283         r1 = r2;
2284         r2 = 0;
2285     }
2286     result[uECC_N_WORDS * 2 - 1] = r0;
2287 }
2288 
2289 static void vli_modAdd_n(uECC_word_t *result,
2290                          const uECC_word_t *left,
2291                          const uECC_word_t *right,
2292                          const uECC_word_t *mod) {
2293     uECC_word_t carry = vli_add_n(result, left, right);
2294     if (carry || vli_cmp_n(result, mod) >= 0) {
2295         vli_sub_n(result, result, mod);
2296     }
2297 }
2298 
2299 static void vli_modInv_n(uECC_word_t *result, const uECC_word_t *input, const uECC_word_t *mod) {
2300     uECC_word_t a[uECC_N_WORDS], b[uECC_N_WORDS], u[uECC_N_WORDS], v[uECC_N_WORDS];
2301     uECC_word_t carry;
2302     cmpresult_t cmpResult;
2303 
2304     if (vli_isZero_n(input)) {
2305         vli_clear_n(result);
2306         return;
2307     }
2308 
2309     vli_set_n(a, input);
2310     vli_set_n(b, mod);
2311     vli_clear_n(u);
2312     u[0] = 1;
2313     vli_clear_n(v);
2314     while ((cmpResult = vli_cmp_n(a, b)) != 0) {
2315         carry = 0;
2316         if (EVEN(a)) {
2317             vli_rshift1_n(a);
2318             if (!EVEN(u)) {
2319                 carry = vli_add_n(u, u, mod);
2320             }
2321             vli_rshift1_n(u);
2322             if (carry) {
2323                 u[uECC_N_WORDS - 1] |= HIGH_BIT_SET;
2324             }
2325         } else if (EVEN(b)) {
2326             vli_rshift1_n(b);
2327             if (!EVEN(v)) {
2328                 carry = vli_add_n(v, v, mod);
2329             }
2330             vli_rshift1_n(v);
2331             if (carry) {
2332                 v[uECC_N_WORDS - 1] |= HIGH_BIT_SET;
2333             }
2334         } else if (cmpResult > 0) {
2335             vli_sub_n(a, a, b);
2336             vli_rshift1_n(a);
2337             if (vli_cmp_n(u, v) < 0) {
2338                 vli_add_n(u, u, mod);
2339             }
2340             vli_sub_n(u, u, v);
2341             if (!EVEN(u)) {
2342                 carry = vli_add_n(u, u, mod);
2343             }
2344             vli_rshift1_n(u);
2345             if (carry) {
2346                 u[uECC_N_WORDS - 1] |= HIGH_BIT_SET;
2347             }
2348         } else {
2349             vli_sub_n(b, b, a);
2350             vli_rshift1_n(b);
2351             if (vli_cmp_n(v, u) < 0) {
2352                 vli_add_n(v, v, mod);
2353             }
2354             vli_sub_n(v, v, u);
2355             if (!EVEN(v)) {
2356                 carry = vli_add_n(v, v, mod);
2357             }
2358             vli_rshift1_n(v);
2359             if (carry) {
2360                 v[uECC_N_WORDS - 1] |= HIGH_BIT_SET;
2361             }
2362         }
2363     }
2364     vli_set_n(result, u);
2365 }
2366 
2367 static void vli2_rshift1_n(uECC_word_t *vli) {
2368     vli_rshift1_n(vli);
2369     vli[uECC_N_WORDS - 1] |= vli[uECC_N_WORDS] << (uECC_WORD_BITS - 1);
2370     vli_rshift1_n(vli + uECC_N_WORDS);
2371 }
2372 
2373 static uECC_word_t vli2_sub_n(uECC_word_t *result,
2374                               const uECC_word_t *left,
2375                               const uECC_word_t *right) {
2376     uECC_word_t borrow = 0;
2377     wordcount_t i;
2378     for (i = 0; i < uECC_N_WORDS * 2; ++i) {
2379         uECC_word_t diff = left[i] - right[i] - borrow;
2380         if (diff != left[i]) {
2381             borrow = (diff > left[i]);
2382         }
2383         result[i] = diff;
2384     }
2385     return borrow;
2386 }
2387 
2388 /* Computes result = (left * right) % curve_n. */
2389 static void vli_modMult_n(uECC_word_t *result, const uECC_word_t *left, const uECC_word_t *right) {
2390     bitcount_t i;
2391     uECC_word_t product[2 * uECC_N_WORDS];
2392     uECC_word_t modMultiple[2 * uECC_N_WORDS];
2393     uECC_word_t tmp[2 * uECC_N_WORDS];
2394     uECC_word_t *v[2] = {tmp, product};
2395     uECC_word_t index = 1;
2396 
2397     vli_mult_n(product, left, right);
2398     vli_clear_n(modMultiple);
2399     vli_set(modMultiple + uECC_N_WORDS + 1, curve_n);
2400     vli_rshift1(modMultiple + uECC_N_WORDS + 1);
2401     modMultiple[2 * uECC_N_WORDS - 1] |= HIGH_BIT_SET;
2402     modMultiple[uECC_N_WORDS] = HIGH_BIT_SET;
2403 
2404     for (i = 0;
2405          i <= ((((bitcount_t)uECC_N_WORDS) << uECC_WORD_BITS_SHIFT) + (uECC_WORD_BITS - 1));
2406          ++i) {
2407         uECC_word_t borrow = vli2_sub_n(v[1 - index], v[index], modMultiple);
2408         index = !(index ^ borrow); /* Swap the index if there was no borrow */
2409         vli2_rshift1_n(modMultiple);
2410     }
2411     vli_set_n(result, v[index]);
2412 }
2413 
2414 #else
2415 
2416 #define vli_cmp_n vli_cmp
2417 #define vli_modInv_n vli_modInv
2418 #define vli_modAdd_n vli_modAdd
2419 
2420 static void vli2_rshift1(uECC_word_t *vli) {
2421     vli_rshift1(vli);
2422     vli[uECC_WORDS - 1] |= vli[uECC_WORDS] << (uECC_WORD_BITS - 1);
2423     vli_rshift1(vli + uECC_WORDS);
2424 }
2425 
2426 static uECC_word_t vli2_sub(uECC_word_t *result,
2427                             const uECC_word_t *left,
2428                             const uECC_word_t *right) {
2429     uECC_word_t borrow = 0;
2430     wordcount_t i;
2431     for (i = 0; i < uECC_WORDS * 2; ++i) {
2432         uECC_word_t diff = left[i] - right[i] - borrow;
2433         if (diff != left[i]) {
2434             borrow = (diff > left[i]);
2435         }
2436         result[i] = diff;
2437     }
2438     return borrow;
2439 }
2440 
2441 /* Computes result = (left * right) % curve_n. */
2442 static void vli_modMult_n(uECC_word_t *result, const uECC_word_t *left, const uECC_word_t *right) {
2443     uECC_word_t product[2 * uECC_WORDS];
2444     uECC_word_t modMultiple[2 * uECC_WORDS];
2445     uECC_word_t tmp[2 * uECC_WORDS];
2446     uECC_word_t *v[2] = {tmp, product};
2447     bitcount_t i;
2448     uECC_word_t index = 1;
2449 
2450     vli_mult(product, left, right);
2451     vli_set(modMultiple + uECC_WORDS, curve_n); /* works if curve_n has its highest bit set */
2452     vli_clear(modMultiple);
2453 
2454     for (i = 0; i <= uECC_BYTES * 8; ++i) {
2455         uECC_word_t borrow = vli2_sub(v[1 - index], v[index], modMultiple);
2456         index = !(index ^ borrow); /* Swap the index if there was no borrow */
2457         vli2_rshift1(modMultiple);
2458     }
2459     vli_set(result, v[index]);
2460 }
2461 #endif /* (uECC_CURVE != uECC_secp160r1) */
2462 
2463 static int uECC_sign_with_k(const uint8_t private_key[uECC_BYTES],
2464                             const uint8_t message_hash[uECC_BYTES],
2465                             uECC_word_t k[uECC_N_WORDS],
2466                             uint8_t signature[uECC_BYTES*2]) {
2467     uECC_word_t tmp[uECC_N_WORDS];
2468     uECC_word_t s[uECC_N_WORDS];
2469     uECC_word_t *k2[2] = {tmp, s};
2470     EccPoint p;
2471     uECC_word_t carry;
2472     uECC_word_t tries;
2473 
2474     /* Make sure 0 < k < curve_n */
2475     if (vli_isZero(k) || vli_cmp_n(curve_n, k) != 1) {
2476         return 0;
2477     }
2478 
2479 #if (uECC_CURVE == uECC_secp160r1)
2480     /* Make sure that we don't leak timing information about k.
2481        See http://eprint.iacr.org/2011/232.pdf */
2482     vli_add_n(tmp, k, curve_n);
2483     carry = (tmp[uECC_WORDS] & 0x02);
2484     vli_add_n(s, tmp, curve_n);
2485 
2486     /* p = k * G */
2487     EccPoint_mult(&p, &curve_G, k2[!carry], NULL, (uECC_BYTES * 8) + 2);
2488 #else
2489     /* Make sure that we don't leak timing information about k.
2490        See http://eprint.iacr.org/2011/232.pdf */
2491     carry = vli_add(tmp, k, curve_n);
2492     vli_add(s, tmp, curve_n);
2493 
2494     /* p = k * G */
2495     EccPoint_mult(&p, &curve_G, k2[!carry], NULL, (uECC_BYTES * 8) + 1);
2496 
2497     /* r = x1 (mod n) */
2498     if (vli_cmp(curve_n, p.x) != 1) {
2499         vli_sub(p.x, p.x, curve_n);
2500     }
2501 #endif
2502     if (vli_isZero(p.x)) {
2503         return 0;
2504     }
2505 
2506     // Attempt to get a random number to prevent side channel analysis of k.
2507     // If the RNG fails every time (eg it was not defined), we continue so that
2508     // deterministic signing can still work (with reduced security) without
2509     // an RNG defined.
2510     carry = 0; // use to signal that the RNG succeeded at least once.
2511     for (tries = 0; tries < MAX_TRIES; ++tries) {
2512         if (!g_rng_function((uint8_t *)tmp, sizeof(tmp))) {
2513             continue;
2514         }
2515         carry = 1;
2516         if (!vli_isZero(tmp)) {
2517             break;
2518         }
2519     }
2520     if (!carry) {
2521         vli_clear(tmp);
2522         tmp[0] = 1;
2523     }
2524 
2525     /* Prevent side channel analysis of vli_modInv() to determine
2526        bits of k / the private key by premultiplying by a random number */
2527     vli_modMult_n(k, k, tmp); /* k' = rand * k */
2528     vli_modInv_n(k, k, curve_n); /* k = 1 / k' */
2529     vli_modMult_n(k, k, tmp); /* k = 1 / k */
2530 
2531     vli_nativeToBytes(signature, p.x); /* store r */
2532 
2533     tmp[uECC_N_WORDS - 1] = 0;
2534     vli_bytesToNative(tmp, private_key); /* tmp = d */
2535     s[uECC_N_WORDS - 1] = 0;
2536     vli_set(s, p.x);
2537     vli_modMult_n(s, tmp, s); /* s = r*d */
2538 
2539     vli_bytesToNative(tmp, message_hash);
2540     vli_modAdd_n(s, tmp, s, curve_n); /* s = e + r*d */
2541     vli_modMult_n(s, s, k); /* s = (e + r*d) / k */
2542 #if (uECC_CURVE == uECC_secp160r1)
2543     if (s[uECC_N_WORDS - 1]) {
2544         return 0;
2545     }
2546 #endif
2547     vli_nativeToBytes(signature + uECC_BYTES, s);
2548     return 1;
2549 }
2550 
2551 int uECC_sign(const uint8_t private_key[uECC_BYTES],
2552               const uint8_t message_hash[uECC_BYTES],
2553               uint8_t signature[uECC_BYTES*2]) {
2554     uECC_word_t k[uECC_N_WORDS];
2555     uECC_word_t tries;
2556 
2557     for (tries = 0; tries < MAX_TRIES; ++tries) {
2558         if(g_rng_function((uint8_t *)k, sizeof(k))) {
2559         #if (uECC_CURVE == uECC_secp160r1)
2560             k[uECC_WORDS] &= 0x01;
2561         #endif
2562             if (uECC_sign_with_k(private_key, message_hash, k, signature)) {
2563                 return 1;
2564             }
2565         }
2566     }
2567     return 0;
2568 }
2569 
2570 /* Compute an HMAC using K as a key (as in RFC 6979). Note that K is always
2571    the same size as the hash result size. */
2572 static void HMAC_init(uECC_HashContext *hash_context, const uint8_t *K) {
2573     uint8_t *pad = hash_context->tmp + 2 * hash_context->result_size;
2574     unsigned i;
2575     for (i = 0; i < hash_context->result_size; ++i)
2576         pad[i] = K[i] ^ 0x36;
2577     for (; i < hash_context->block_size; ++i)
2578         pad[i] = 0x36;
2579 
2580     hash_context->init_hash(hash_context);
2581     hash_context->update_hash(hash_context, pad, hash_context->block_size);
2582 }
2583 
2584 static void HMAC_update(uECC_HashContext *hash_context,
2585                         const uint8_t *message,
2586                         unsigned message_size) {
2587     hash_context->update_hash(hash_context, message, message_size);
2588 }
2589 
2590 static void HMAC_finish(uECC_HashContext *hash_context, const uint8_t *K, uint8_t *result) {
2591     uint8_t *pad = hash_context->tmp + 2 * hash_context->result_size;
2592     unsigned i;
2593     for (i = 0; i < hash_context->result_size; ++i)
2594         pad[i] = K[i] ^ 0x5c;
2595     for (; i < hash_context->block_size; ++i)
2596         pad[i] = 0x5c;
2597 
2598     hash_context->finish_hash(hash_context, result);
2599 
2600     hash_context->init_hash(hash_context);
2601     hash_context->update_hash(hash_context, pad, hash_context->block_size);
2602     hash_context->update_hash(hash_context, result, hash_context->result_size);
2603     hash_context->finish_hash(hash_context, result);
2604 }
2605 
2606 /* V = HMAC_K(V) */
2607 static void update_V(uECC_HashContext *hash_context, uint8_t *K, uint8_t *V) {
2608     HMAC_init(hash_context, K);
2609     HMAC_update(hash_context, V, hash_context->result_size);
2610     HMAC_finish(hash_context, K, V);
2611 }
2612 
2613 /* Deterministic signing, similar to RFC 6979. Differences are:
2614     * We just use (truncated) H(m) directly rather than bits2octets(H(m))
2615       (it is not reduced modulo curve_n).
2616     * We generate a value for k (aka T) directly rather than converting endianness.
2617 
2618    Layout of hash_context->tmp: <K> | <V> | (1 byte overlapped 0x00 or 0x01) / <HMAC pad> */
2619 int uECC_sign_deterministic(const uint8_t private_key[uECC_BYTES],
2620                             const uint8_t message_hash[uECC_BYTES],
2621                             uECC_HashContext *hash_context,
2622                             uint8_t signature[uECC_BYTES*2]) {
2623     uint8_t *K = hash_context->tmp;
2624     uint8_t *V = K + hash_context->result_size;
2625     uECC_word_t tries;
2626     unsigned i;
2627     for (i = 0; i < hash_context->result_size; ++i) {
2628         V[i] = 0x01;
2629         K[i] = 0;
2630     }
2631 
2632     // K = HMAC_K(V || 0x00 || int2octets(x) || h(m))
2633     HMAC_init(hash_context, K);
2634     V[hash_context->result_size] = 0x00;
2635     HMAC_update(hash_context, V, hash_context->result_size + 1);
2636     HMAC_update(hash_context, private_key, uECC_BYTES);
2637     HMAC_update(hash_context, message_hash, uECC_BYTES);
2638     HMAC_finish(hash_context, K, K);
2639 
2640     update_V(hash_context, K, V);
2641 
2642     // K = HMAC_K(V || 0x01 || int2octets(x) || h(m))
2643     HMAC_init(hash_context, K);
2644     V[hash_context->result_size] = 0x01;
2645     HMAC_update(hash_context, V, hash_context->result_size + 1);
2646     HMAC_update(hash_context, private_key, uECC_BYTES);
2647     HMAC_update(hash_context, message_hash, uECC_BYTES);
2648     HMAC_finish(hash_context, K, K);
2649 
2650     update_V(hash_context, K, V);
2651 
2652     for (tries = 0; tries < MAX_TRIES; ++tries) {
2653         uECC_word_t T[uECC_N_WORDS];
2654         uint8_t *T_ptr = (uint8_t *)T;
2655         unsigned T_bytes = 0;
2656         while (T_bytes < sizeof(T)) {
2657             update_V(hash_context, K, V);
2658             for (i = 0; i < hash_context->result_size && T_bytes < sizeof(T); ++i, ++T_bytes) {
2659                 T_ptr[T_bytes] = V[i];
2660             }
2661         }
2662     #if (uECC_CURVE == uECC_secp160r1)
2663         T[uECC_WORDS] &= 0x01;
2664     #endif
2665 
2666         if (uECC_sign_with_k(private_key, message_hash, T, signature)) {
2667             return 1;
2668         }
2669 
2670         // K = HMAC_K(V || 0x00)
2671         HMAC_init(hash_context, K);
2672         V[hash_context->result_size] = 0x00;
2673         HMAC_update(hash_context, V, hash_context->result_size + 1);
2674         HMAC_finish(hash_context, K, K);
2675 
2676         update_V(hash_context, K, V);
2677     }
2678     return 0;
2679 }
2680 
2681 static bitcount_t smax(bitcount_t a, bitcount_t b) {
2682     return (a > b ? a : b);
2683 }
2684 
2685 int uECC_verify(const uint8_t public_key[uECC_BYTES*2],
2686                 const uint8_t hash[uECC_BYTES],
2687                 const uint8_t signature[uECC_BYTES*2]) {
2688     uECC_word_t u1[uECC_N_WORDS], u2[uECC_N_WORDS];
2689     uECC_word_t z[uECC_N_WORDS];
2690     EccPoint public, sum;
2691     uECC_word_t rx[uECC_WORDS];
2692     uECC_word_t ry[uECC_WORDS];
2693     uECC_word_t tx[uECC_WORDS];
2694     uECC_word_t ty[uECC_WORDS];
2695     uECC_word_t tz[uECC_WORDS];
2696     const EccPoint *points[4];
2697     const EccPoint *point;
2698     bitcount_t numBits;
2699     bitcount_t i;
2700     uECC_word_t r[uECC_N_WORDS], s[uECC_N_WORDS];
2701     r[uECC_N_WORDS - 1] = 0;
2702     s[uECC_N_WORDS - 1] = 0;
2703 
2704     vli_bytesToNative(public.x, public_key);
2705     vli_bytesToNative(public.y, public_key + uECC_BYTES);
2706     vli_bytesToNative(r, signature);
2707     vli_bytesToNative(s, signature + uECC_BYTES);
2708 
2709     if (vli_isZero(r) || vli_isZero(s)) { /* r, s must not be 0. */
2710         return 0;
2711     }
2712 
2713 #if (uECC_CURVE != uECC_secp160r1)
2714     if (vli_cmp(curve_n, r) != 1 || vli_cmp(curve_n, s) != 1) { /* r, s must be < n. */
2715         return 0;
2716     }
2717 #endif
2718 
2719     /* Calculate u1 and u2. */
2720     vli_modInv_n(z, s, curve_n); /* Z = s^-1 */
2721     u1[uECC_N_WORDS - 1] = 0;
2722     vli_bytesToNative(u1, hash);
2723     vli_modMult_n(u1, u1, z); /* u1 = e/s */
2724     vli_modMult_n(u2, r, z); /* u2 = r/s */
2725 
2726     /* Calculate sum = G + Q. */
2727     vli_set(sum.x, public.x);
2728     vli_set(sum.y, public.y);
2729     vli_set(tx, curve_G.x);
2730     vli_set(ty, curve_G.y);
2731     vli_modSub_fast(z, sum.x, tx); /* Z = x2 - x1 */
2732     XYcZ_add(tx, ty, sum.x, sum.y);
2733     vli_modInv(z, z, curve_p); /* Z = 1/Z */
2734     apply_z(sum.x, sum.y, z);
2735 
2736     /* Use Shamir's trick to calculate u1*G + u2*Q */
2737     points[0] = 0;
2738     points[1] = &curve_G;
2739     points[2] = &public;
2740     points[3] = &sum;
2741     numBits = smax(vli_numBits(u1, uECC_N_WORDS), vli_numBits(u2, uECC_N_WORDS));
2742 
2743     point = points[(!!vli_testBit(u1, numBits - 1)) | ((!!vli_testBit(u2, numBits - 1)) << 1)];
2744     vli_set(rx, point->x);
2745     vli_set(ry, point->y);
2746     vli_clear(z);
2747     z[0] = 1;
2748 
2749     for (i = numBits - 2; i >= 0; --i) {
2750         uECC_word_t index;
2751         EccPoint_double_jacobian(rx, ry, z);
2752 
2753         index = (!!vli_testBit(u1, i)) | ((!!vli_testBit(u2, i)) << 1);
2754         point = points[index];
2755         if (point) {
2756             vli_set(tx, point->x);
2757             vli_set(ty, point->y);
2758             apply_z(tx, ty, z);
2759             vli_modSub_fast(tz, rx, tx); /* Z = x2 - x1 */
2760             XYcZ_add(tx, ty, rx, ry);
2761             vli_modMult_fast(z, z, tz);
2762         }
2763     }
2764 
2765     vli_modInv(z, z, curve_p); /* Z = 1/Z */
2766     apply_z(rx, ry, z);
2767 
2768     /* v = x1 (mod n) */
2769 #if (uECC_CURVE != uECC_secp160r1)
2770     if (vli_cmp(curve_n, rx) != 1) {
2771         vli_sub(rx, rx, curve_n);
2772     }
2773 #endif
2774 
2775     /* Accept only if v == r. */
2776     return vli_equal(rx, r);
2777 }
2778 
2779 #endif
2780