xref: /btstack/3rd-party/micro-ecc/uECC.c (revision 0561b2d8d5dba972c7daa57d5e677f7a1327edfd)
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 static const uECC_word_t curve_b[uECC_WORDS] = uECC_CONCAT(Curve_B_, uECC_CURVE);
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 static bitcount_t vli_numBits(const uECC_word_t *vli, wordcount_t max_words);
379 static void vli_set(uECC_word_t *dest, const uECC_word_t *src);
380 static cmpresult_t vli_cmp(const uECC_word_t *left, const uECC_word_t *right);
381 static cmpresult_t vli_equal(const uECC_word_t *left, const uECC_word_t *right);
382 static void vli_rshift1(uECC_word_t *vli);
383 static uECC_word_t vli_add(uECC_word_t *result,
384                            const uECC_word_t *left,
385                            const uECC_word_t *right);
386 static uECC_word_t vli_sub(uECC_word_t *result,
387                            const uECC_word_t *left,
388                            const uECC_word_t *right);
389 static void vli_mult(uECC_word_t *result, const uECC_word_t *left, const uECC_word_t *right);
390 static void vli_modAdd(uECC_word_t *result,
391                        const uECC_word_t *left,
392                        const uECC_word_t *right,
393                        const uECC_word_t *mod);
394 static void vli_modSub(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_mmod_fast(uECC_word_t *RESTRICT result, uECC_word_t *RESTRICT product);
399 static void vli_modMult_fast(uECC_word_t *result,
400                              const uECC_word_t *left,
401                              const uECC_word_t *right);
402 static void vli_modInv(uECC_word_t *result, const uECC_word_t *input, const uECC_word_t *mod);
403 #if uECC_SQUARE_FUNC
404 static void vli_square(uECC_word_t *result, const uECC_word_t *left);
405 static void vli_modSquare_fast(uECC_word_t *result, const uECC_word_t *left);
406 #endif
407 
408 #if (defined(_WIN32) || defined(_WIN64))
409 /* Windows */
410 
411 #define WIN32_LEAN_AND_MEAN
412 #include <windows.h>
413 #include <wincrypt.h>
414 
415 static int default_RNG(uint8_t *dest, unsigned size) {
416     HCRYPTPROV prov;
417     if (!CryptAcquireContext(&prov, NULL, NULL, PROV_RSA_FULL, CRYPT_VERIFYCONTEXT)) {
418         return 0;
419     }
420 
421     CryptGenRandom(prov, size, (BYTE *)dest);
422     CryptReleaseContext(prov, 0);
423     return 1;
424 }
425 
426 #elif defined(unix) || defined(__linux__) || defined(__unix__) || defined(__unix) || \
427     (defined(__APPLE__) && defined(__MACH__)) || defined(uECC_POSIX)
428 
429 /* Some POSIX-like system with /dev/urandom or /dev/random. */
430 #include <sys/types.h>
431 #include <fcntl.h>
432 #include <unistd.h>
433 
434 #ifndef O_CLOEXEC
435     #define O_CLOEXEC 0
436 #endif
437 
438 static int default_RNG(uint8_t *dest, unsigned size) {
439     int fd = open("/dev/urandom", O_RDONLY | O_CLOEXEC);
440     if (fd == -1) {
441         fd = open("/dev/random", O_RDONLY | O_CLOEXEC);
442         if (fd == -1) {
443             return 0;
444         }
445     }
446 
447     char *ptr = (char *)dest;
448     size_t left = size;
449     while (left > 0) {
450         ssize_t bytes_read = read(fd, ptr, left);
451         if (bytes_read <= 0) { // read failed
452             close(fd);
453             return 0;
454         }
455         left -= bytes_read;
456         ptr += bytes_read;
457     }
458 
459     close(fd);
460     return 1;
461 }
462 
463 #else /* Some other platform */
464 
465 static int default_RNG(uint8_t *dest, unsigned size) {
466     return 0;
467 }
468 
469 #endif
470 
471 static uECC_RNG_Function g_rng_function = &default_RNG;
472 
473 void uECC_set_rng(uECC_RNG_Function rng_function) {
474     g_rng_function = rng_function;
475 }
476 
477 #ifdef __GNUC__ /* Only support GCC inline asm for now */
478     #if (uECC_ASM && (uECC_PLATFORM == uECC_avr))
479         #include "asm_avr.inc"
480     #endif
481 
482     #if (uECC_ASM && (uECC_PLATFORM == uECC_arm || uECC_PLATFORM == uECC_arm_thumb || \
483                       uECC_PLATFORM == uECC_arm_thumb2))
484         #include "asm_arm.inc"
485     #endif
486 #endif
487 
488 #if !asm_clear
489 static void vli_clear(uECC_word_t *vli) {
490     wordcount_t i;
491     for (i = 0; i < uECC_WORDS; ++i) {
492         vli[i] = 0;
493     }
494 }
495 #endif
496 
497 /* Returns 1 if vli == 0, 0 otherwise. */
498 #if !asm_isZero
499 static uECC_word_t vli_isZero(const uECC_word_t *vli) {
500     wordcount_t i;
501     for (i = 0; i < uECC_WORDS; ++i) {
502         if (vli[i]) {
503             return 0;
504         }
505     }
506     return 1;
507 }
508 #endif
509 
510 /* Returns nonzero if bit 'bit' of vli is set. */
511 #if !asm_testBit
512 static uECC_word_t vli_testBit(const uECC_word_t *vli, bitcount_t bit) {
513     return (vli[bit >> uECC_WORD_BITS_SHIFT] & ((uECC_word_t)1 << (bit & uECC_WORD_BITS_MASK)));
514 }
515 #endif
516 
517 /* Counts the number of words in vli. */
518 #if !asm_numBits
519 static wordcount_t vli_numDigits(const uECC_word_t *vli, wordcount_t max_words) {
520     swordcount_t i;
521     /* Search from the end until we find a non-zero digit.
522        We do it in reverse because we expect that most digits will be nonzero. */
523     for (i = max_words - 1; i >= 0 && vli[i] == 0; --i) {
524     }
525 
526     return (i + 1);
527 }
528 
529 /* Counts the number of bits required to represent vli. */
530 static bitcount_t vli_numBits(const uECC_word_t *vli, wordcount_t max_words) {
531     uECC_word_t i;
532     uECC_word_t digit;
533 
534     wordcount_t num_digits = vli_numDigits(vli, max_words);
535     if (num_digits == 0) {
536         return 0;
537     }
538 
539     digit = vli[num_digits - 1];
540     for (i = 0; digit; ++i) {
541         digit >>= 1;
542     }
543 
544     return (((bitcount_t)(num_digits - 1) << uECC_WORD_BITS_SHIFT) + i);
545 }
546 #endif /* !asm_numBits */
547 
548 /* Sets dest = src. */
549 #if !asm_set
550 static void vli_set(uECC_word_t *dest, const uECC_word_t *src) {
551     wordcount_t i;
552     for (i = 0; i < uECC_WORDS; ++i) {
553         dest[i] = src[i];
554     }
555 }
556 #endif
557 
558 /* Returns sign of left - right. */
559 #if !asm_cmp
560 static cmpresult_t vli_cmp(const uECC_word_t *left, const uECC_word_t *right) {
561     swordcount_t i;
562     for (i = uECC_WORDS - 1; i >= 0; --i) {
563         if (left[i] > right[i]) {
564             return 1;
565         } else if (left[i] < right[i]) {
566             return -1;
567         }
568     }
569     return 0;
570 }
571 #endif
572 
573 static cmpresult_t vli_equal(const uECC_word_t *left, const uECC_word_t *right) {
574     uECC_word_t result = 0;
575     swordcount_t i;
576     for (i = uECC_WORDS - 1; i >= 0; --i) {
577         result |= (left[i] ^ right[i]);
578     }
579     return (result == 0);
580 }
581 
582 /* Computes vli = vli >> 1. */
583 #if !asm_rshift1
584 static void vli_rshift1(uECC_word_t *vli) {
585     uECC_word_t *end = vli;
586     uECC_word_t carry = 0;
587 
588     vli += uECC_WORDS;
589     while (vli-- > end) {
590         uECC_word_t temp = *vli;
591         *vli = (temp >> 1) | carry;
592         carry = temp << (uECC_WORD_BITS - 1);
593     }
594 }
595 #endif
596 
597 /* Computes result = left + right, returning carry. Can modify in place. */
598 #if !asm_add
599 static uECC_word_t vli_add(uECC_word_t *result, const uECC_word_t *left, const uECC_word_t *right) {
600     uECC_word_t carry = 0;
601     wordcount_t i;
602     for (i = 0; i < uECC_WORDS; ++i) {
603         uECC_word_t sum = left[i] + right[i] + carry;
604         if (sum != left[i]) {
605             carry = (sum < left[i]);
606         }
607         result[i] = sum;
608     }
609     return carry;
610 }
611 #endif
612 
613 /* Computes result = left - right, returning borrow. Can modify in place. */
614 #if !asm_sub
615 static uECC_word_t vli_sub(uECC_word_t *result, const uECC_word_t *left, const uECC_word_t *right) {
616     uECC_word_t borrow = 0;
617     wordcount_t i;
618     for (i = 0; i < uECC_WORDS; ++i) {
619         uECC_word_t diff = left[i] - right[i] - borrow;
620         if (diff != left[i]) {
621             borrow = (diff > left[i]);
622         }
623         result[i] = diff;
624     }
625     return borrow;
626 }
627 #endif
628 
629 #if (!asm_mult || (uECC_SQUARE_FUNC && !asm_square) || uECC_CURVE == uECC_secp256k1)
630 static void muladd(uECC_word_t a,
631                    uECC_word_t b,
632                    uECC_word_t *r0,
633                    uECC_word_t *r1,
634                    uECC_word_t *r2) {
635 #if uECC_WORD_SIZE == 8 && !SUPPORTS_INT128
636     uint64_t a0 = a & 0xffffffffull;
637     uint64_t a1 = a >> 32;
638     uint64_t b0 = b & 0xffffffffull;
639     uint64_t b1 = b >> 32;
640 
641     uint64_t i0 = a0 * b0;
642     uint64_t i1 = a0 * b1;
643     uint64_t i2 = a1 * b0;
644     uint64_t i3 = a1 * b1;
645 
646     uint64_t p0, p1;
647 
648     i2 += (i0 >> 32);
649     i2 += i1;
650     if (i2 < i1) { // overflow
651         i3 += 0x100000000ull;
652     }
653 
654     p0 = (i0 & 0xffffffffull) | (i2 << 32);
655     p1 = i3 + (i2 >> 32);
656 
657     *r0 += p0;
658     *r1 += (p1 + (*r0 < p0));
659     *r2 += ((*r1 < p1) || (*r1 == p1 && *r0 < p0));
660 #else
661     uECC_dword_t p = (uECC_dword_t)a * b;
662     uECC_dword_t r01 = ((uECC_dword_t)(*r1) << uECC_WORD_BITS) | *r0;
663     r01 += p;
664     *r2 += (r01 < p);
665     *r1 = r01 >> uECC_WORD_BITS;
666     *r0 = (uECC_word_t)r01;
667 #endif
668 }
669 #define muladd_exists 1
670 #endif
671 
672 #if !asm_mult
673 static void vli_mult(uECC_word_t *result, const uECC_word_t *left, const uECC_word_t *right) {
674     uECC_word_t r0 = 0;
675     uECC_word_t r1 = 0;
676     uECC_word_t r2 = 0;
677     wordcount_t i, k;
678 
679     /* Compute each digit of result in sequence, maintaining the carries. */
680     for (k = 0; k < uECC_WORDS; ++k) {
681         for (i = 0; i <= k; ++i) {
682             muladd(left[i], right[k - i], &r0, &r1, &r2);
683         }
684         result[k] = r0;
685         r0 = r1;
686         r1 = r2;
687         r2 = 0;
688     }
689     for (k = uECC_WORDS; k < uECC_WORDS * 2 - 1; ++k) {
690         for (i = (k + 1) - uECC_WORDS; i < uECC_WORDS; ++i) {
691             muladd(left[i], right[k - i], &r0, &r1, &r2);
692         }
693         result[k] = r0;
694         r0 = r1;
695         r1 = r2;
696         r2 = 0;
697     }
698     result[uECC_WORDS * 2 - 1] = r0;
699 }
700 #endif
701 
702 #if uECC_SQUARE_FUNC
703 
704 #if !asm_square
705 static void mul2add(uECC_word_t a,
706                     uECC_word_t b,
707                     uECC_word_t *r0,
708                     uECC_word_t *r1,
709                     uECC_word_t *r2) {
710 #if uECC_WORD_SIZE == 8 && !SUPPORTS_INT128
711     uint64_t a0 = a & 0xffffffffull;
712     uint64_t a1 = a >> 32;
713     uint64_t b0 = b & 0xffffffffull;
714     uint64_t b1 = b >> 32;
715 
716     uint64_t i0 = a0 * b0;
717     uint64_t i1 = a0 * b1;
718     uint64_t i2 = a1 * b0;
719     uint64_t i3 = a1 * b1;
720 
721     uint64_t p0, p1;
722 
723     i2 += (i0 >> 32);
724     i2 += i1;
725     if (i2 < i1)
726     { // overflow
727         i3 += 0x100000000ull;
728     }
729 
730     p0 = (i0 & 0xffffffffull) | (i2 << 32);
731     p1 = i3 + (i2 >> 32);
732 
733     *r2 += (p1 >> 63);
734     p1 = (p1 << 1) | (p0 >> 63);
735     p0 <<= 1;
736 
737     *r0 += p0;
738     *r1 += (p1 + (*r0 < p0));
739     *r2 += ((*r1 < p1) || (*r1 == p1 && *r0 < p0));
740 #else
741     uECC_dword_t p = (uECC_dword_t)a * b;
742     uECC_dword_t r01 = ((uECC_dword_t)(*r1) << uECC_WORD_BITS) | *r0;
743     *r2 += (p >> (uECC_WORD_BITS * 2 - 1));
744     p *= 2;
745     r01 += p;
746     *r2 += (r01 < p);
747     *r1 = r01 >> uECC_WORD_BITS;
748     *r0 = (uECC_word_t)r01;
749 #endif
750 }
751 
752 static void vli_square(uECC_word_t *result, const uECC_word_t *left) {
753     uECC_word_t r0 = 0;
754     uECC_word_t r1 = 0;
755     uECC_word_t r2 = 0;
756 
757     wordcount_t i, k;
758 
759     for (k = 0; k < uECC_WORDS * 2 - 1; ++k) {
760         uECC_word_t min = (k < uECC_WORDS ? 0 : (k + 1) - uECC_WORDS);
761         for (i = min; i <= k && i <= k - i; ++i) {
762             if (i < k-i) {
763                 mul2add(left[i], left[k - i], &r0, &r1, &r2);
764             } else {
765                 muladd(left[i], left[k - i], &r0, &r1, &r2);
766             }
767         }
768         result[k] = r0;
769         r0 = r1;
770         r1 = r2;
771         r2 = 0;
772     }
773 
774     result[uECC_WORDS * 2 - 1] = r0;
775 }
776 #endif
777 
778 #else /* uECC_SQUARE_FUNC */
779 
780 #define vli_square(result, left, size) vli_mult((result), (left), (left), (size))
781 
782 #endif /* uECC_SQUARE_FUNC */
783 
784 
785 /* Computes result = (left + right) % mod.
786    Assumes that left < mod and right < mod, and that result does not overlap mod. */
787 #if !asm_modAdd
788 static void vli_modAdd(uECC_word_t *result,
789                        const uECC_word_t *left,
790                        const uECC_word_t *right,
791                        const uECC_word_t *mod) {
792     uECC_word_t carry = vli_add(result, left, right);
793     if (carry || vli_cmp(result, mod) >= 0) {
794         /* result > mod (result = mod + remainder), so subtract mod to get remainder. */
795         vli_sub(result, result, mod);
796     }
797 }
798 #endif
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_modSub
803 static void vli_modSub(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 l_borrow = vli_sub(result, left, right);
808     if (l_borrow) {
809         /* In this case, result == -diff == (max int) - diff. Since -x % d == d - x,
810            we can get the correct result from result + mod (with overflow). */
811         vli_add(result, result, mod);
812     }
813 }
814 #endif
815 
816 #if !asm_modSub_fast
817     #define vli_modSub_fast(result, left, right) vli_modSub((result), (left), (right), curve_p)
818 #endif
819 
820 #if !asm_mmod_fast
821 
822 #if (uECC_CURVE == uECC_secp160r1 || uECC_CURVE == uECC_secp256k1)
823 /* omega_mult() is defined farther below for the different curves / word sizes */
824 static void omega_mult(uECC_word_t * RESTRICT result, const uECC_word_t * RESTRICT right);
825 
826 /* Computes result = product % curve_p
827     see http://www.isys.uni-klu.ac.at/PDF/2001-0126-MT.pdf page 354
828 
829     Note that this only works if log2(omega) < log2(p) / 2 */
830 static void vli_mmod_fast(uECC_word_t *RESTRICT result, uECC_word_t *RESTRICT product) {
831     uECC_word_t tmp[2 * uECC_WORDS];
832     uECC_word_t carry;
833 
834     vli_clear(tmp);
835     vli_clear(tmp + uECC_WORDS);
836 
837     omega_mult(tmp, product + uECC_WORDS); /* (Rq, q) = q * c */
838 
839     carry = vli_add(result, product, tmp); /* (C, r) = r + q       */
840     vli_clear(product);
841     omega_mult(product, tmp + uECC_WORDS); /* Rq*c */
842     carry += vli_add(result, result, product); /* (C1, r) = r + Rq*c */
843 
844     while (carry > 0) {
845         --carry;
846         vli_sub(result, result, curve_p);
847     }
848     if (vli_cmp(result, curve_p) > 0) {
849         vli_sub(result, result, curve_p);
850     }
851 }
852 
853 #endif
854 
855 #if uECC_CURVE == uECC_secp160r1
856 
857 #if uECC_WORD_SIZE == 1
858 static void omega_mult(uint8_t * RESTRICT result, const uint8_t * RESTRICT right) {
859     uint8_t carry;
860     uint8_t i;
861 
862     /* Multiply by (2^31 + 1). */
863     vli_set(result + 4, right); /* 2^32 */
864     vli_rshift1(result + 4); /* 2^31 */
865     result[3] = right[0] << 7; /* get last bit from shift */
866 
867     carry = vli_add(result, result, right); /* 2^31 + 1 */
868     for (i = uECC_WORDS; carry; ++i) {
869         uint16_t sum = (uint16_t)result[i] + carry;
870         result[i] = (uint8_t)sum;
871         carry = sum >> 8;
872     }
873 }
874 #elif uECC_WORD_SIZE == 4
875 static void omega_mult(uint32_t * RESTRICT result, const uint32_t * RESTRICT right) {
876     uint32_t carry;
877     unsigned i;
878 
879     /* Multiply by (2^31 + 1). */
880     vli_set(result + 1, right); /* 2^32 */
881     vli_rshift1(result + 1); /* 2^31 */
882     result[0] = right[0] << 31; /* get last bit from shift */
883 
884     carry = vli_add(result, result, right); /* 2^31 + 1 */
885     for (i = uECC_WORDS; carry; ++i) {
886         uint64_t sum = (uint64_t)result[i] + carry;
887         result[i] = (uint32_t)sum;
888         carry = sum >> 32;
889     }
890 }
891 #endif /* uECC_WORD_SIZE */
892 
893 #elif uECC_CURVE == uECC_secp192r1
894 
895 /* Computes result = product % curve_p.
896    See algorithm 5 and 6 from http://www.isys.uni-klu.ac.at/PDF/2001-0126-MT.pdf */
897 #if uECC_WORD_SIZE == 1
898 static void vli_mmod_fast(uint8_t *RESTRICT result, uint8_t *RESTRICT product) {
899     uint8_t tmp[uECC_WORDS];
900     uint8_t carry;
901 
902     vli_set(result, product);
903 
904     vli_set(tmp, &product[24]);
905     carry = vli_add(result, result, tmp);
906 
907     tmp[0] = tmp[1] = tmp[2] = tmp[3] = tmp[4] = tmp[5] = tmp[6] = tmp[7] = 0;
908     tmp[8] = product[24]; tmp[9] = product[25]; tmp[10] = product[26]; tmp[11] = product[27];
909     tmp[12] = product[28]; tmp[13] = product[29]; tmp[14] = product[30]; tmp[15] = product[31];
910     tmp[16] = product[32]; tmp[17] = product[33]; tmp[18] = product[34]; tmp[19] = product[35];
911     tmp[20] = product[36]; tmp[21] = product[37]; tmp[22] = product[38]; tmp[23] = product[39];
912     carry += vli_add(result, result, tmp);
913 
914     tmp[0] = tmp[8] = product[40];
915     tmp[1] = tmp[9] = product[41];
916     tmp[2] = tmp[10] = product[42];
917     tmp[3] = tmp[11] = product[43];
918     tmp[4] = tmp[12] = product[44];
919     tmp[5] = tmp[13] = product[45];
920     tmp[6] = tmp[14] = product[46];
921     tmp[7] = tmp[15] = product[47];
922     tmp[16] = tmp[17] = tmp[18] = tmp[19] = tmp[20] = tmp[21] = tmp[22] = tmp[23] = 0;
923     carry += vli_add(result, result, tmp);
924 
925     while (carry || vli_cmp(curve_p, result) != 1) {
926         carry -= vli_sub(result, result, curve_p);
927     }
928 }
929 #elif uECC_WORD_SIZE == 4
930 static void vli_mmod_fast(uint32_t *RESTRICT result, uint32_t *RESTRICT product) {
931     uint32_t tmp[uECC_WORDS];
932     int carry;
933 
934     vli_set(result, product);
935 
936     vli_set(tmp, &product[6]);
937     carry = vli_add(result, result, tmp);
938 
939     tmp[0] = tmp[1] = 0;
940     tmp[2] = product[6];
941     tmp[3] = product[7];
942     tmp[4] = product[8];
943     tmp[5] = product[9];
944     carry += vli_add(result, result, tmp);
945 
946     tmp[0] = tmp[2] = product[10];
947     tmp[1] = tmp[3] = product[11];
948     tmp[4] = tmp[5] = 0;
949     carry += vli_add(result, result, tmp);
950 
951     while (carry || vli_cmp(curve_p, result) != 1) {
952         carry -= vli_sub(result, result, curve_p);
953     }
954 }
955 #else
956 static void vli_mmod_fast(uint64_t *RESTRICT result, uint64_t *RESTRICT product) {
957     uint64_t tmp[uECC_WORDS];
958     int carry;
959 
960     vli_set(result, product);
961 
962     vli_set(tmp, &product[3]);
963     carry = vli_add(result, result, tmp);
964 
965     tmp[0] = 0;
966     tmp[1] = product[3];
967     tmp[2] = product[4];
968     carry += vli_add(result, result, tmp);
969 
970     tmp[0] = tmp[1] = product[5];
971     tmp[2] = 0;
972     carry += vli_add(result, result, tmp);
973 
974     while (carry || vli_cmp(curve_p, result) != 1) {
975         carry -= vli_sub(result, result, curve_p);
976     }
977 }
978 #endif /* uECC_WORD_SIZE */
979 
980 #elif uECC_CURVE == uECC_secp256r1
981 
982 /* Computes result = product % curve_p
983    from http://www.nsa.gov/ia/_files/nist-routines.pdf */
984 #if uECC_WORD_SIZE == 1
985 static void vli_mmod_fast(uint8_t *RESTRICT result, uint8_t *RESTRICT product) {
986     uint8_t tmp[uECC_BYTES];
987     int8_t carry;
988 
989     /* t */
990     vli_set(result, product);
991 
992     /* s1 */
993     tmp[0] = tmp[1] = tmp[2] = tmp[3] = 0;
994     tmp[4] = tmp[5] = tmp[6] = tmp[7] = 0;
995     tmp[8] = tmp[9] = tmp[10] = tmp[11] = 0;
996     tmp[12] = product[44]; tmp[13] = product[45]; tmp[14] = product[46]; tmp[15] = product[47];
997     tmp[16] = product[48]; tmp[17] = product[49]; tmp[18] = product[50]; tmp[19] = product[51];
998     tmp[20] = product[52]; tmp[21] = product[53]; tmp[22] = product[54]; tmp[23] = product[55];
999     tmp[24] = product[56]; tmp[25] = product[57]; tmp[26] = product[58]; tmp[27] = product[59];
1000     tmp[28] = product[60]; tmp[29] = product[61]; tmp[30] = product[62]; tmp[31] = product[63];
1001     carry = vli_add(tmp, tmp, tmp);
1002     carry += vli_add(result, result, tmp);
1003 
1004     /* s2 */
1005     tmp[12] = product[48]; tmp[13] = product[49]; tmp[14] = product[50]; tmp[15] = product[51];
1006     tmp[16] = product[52]; tmp[17] = product[53]; tmp[18] = product[54]; tmp[19] = product[55];
1007     tmp[20] = product[56]; tmp[21] = product[57]; tmp[22] = product[58]; tmp[23] = product[59];
1008     tmp[24] = product[60]; tmp[25] = product[61]; tmp[26] = product[62]; tmp[27] = product[63];
1009     tmp[28] = tmp[29] = tmp[30] = tmp[31] = 0;
1010     carry += vli_add(tmp, tmp, tmp);
1011     carry += vli_add(result, result, tmp);
1012 
1013     /* s3 */
1014     tmp[0] = product[32]; tmp[1] = product[33]; tmp[2] = product[34]; tmp[3] = product[35];
1015     tmp[4] = product[36]; tmp[5] = product[37]; tmp[6] = product[38]; tmp[7] = product[39];
1016     tmp[8] = product[40]; tmp[9] = product[41]; tmp[10] = product[42]; tmp[11] = product[43];
1017     tmp[12] = tmp[13] = tmp[14] = tmp[15] = 0;
1018     tmp[16] = tmp[17] = tmp[18] = tmp[19] = 0;
1019     tmp[20] = tmp[21] = tmp[22] = tmp[23] = 0;
1020     tmp[24] = product[56]; tmp[25] = product[57]; tmp[26] = product[58]; tmp[27] = product[59];
1021     tmp[28] = product[60]; tmp[29] = product[61]; tmp[30] = product[62]; tmp[31] = product[63];
1022     carry += vli_add(result, result, tmp);
1023 
1024     /* s4 */
1025     tmp[0] = product[36]; tmp[1] = product[37]; tmp[2] = product[38]; tmp[3] = product[39];
1026     tmp[4] = product[40]; tmp[5] = product[41]; tmp[6] = product[42]; tmp[7] = product[43];
1027     tmp[8] = product[44]; tmp[9] = product[45]; tmp[10] = product[46]; tmp[11] = product[47];
1028     tmp[12] = product[52]; tmp[13] = product[53]; tmp[14] = product[54]; tmp[15] = product[55];
1029     tmp[16] = product[56]; tmp[17] = product[57]; tmp[18] = product[58]; tmp[19] = product[59];
1030     tmp[20] = product[60]; tmp[21] = product[61]; tmp[22] = product[62]; tmp[23] = product[63];
1031     tmp[24] = product[52]; tmp[25] = product[53]; tmp[26] = product[54]; tmp[27] = product[55];
1032     tmp[28] = product[32]; tmp[29] = product[33]; tmp[30] = product[34]; tmp[31] = product[35];
1033     carry += vli_add(result, result, tmp);
1034 
1035     /* d1 */
1036     tmp[0] = product[44]; tmp[1] = product[45]; tmp[2] = product[46]; tmp[3] = product[47];
1037     tmp[4] = product[48]; tmp[5] = product[49]; tmp[6] = product[50]; tmp[7] = product[51];
1038     tmp[8] = product[52]; tmp[9] = product[53]; tmp[10] = product[54]; tmp[11] = product[55];
1039     tmp[12] = tmp[13] = tmp[14] = tmp[15] = 0;
1040     tmp[16] = tmp[17] = tmp[18] = tmp[19] = 0;
1041     tmp[20] = tmp[21] = tmp[22] = tmp[23] = 0;
1042     tmp[24] = product[32]; tmp[25] = product[33]; tmp[26] = product[34]; tmp[27] = product[35];
1043     tmp[28] = product[40]; tmp[29] = product[41]; tmp[30] = product[42]; tmp[31] = product[43];
1044     carry -= vli_sub(result, result, tmp);
1045 
1046     /* d2 */
1047     tmp[0] = product[48]; tmp[1] = product[49]; tmp[2] = product[50]; tmp[3] = product[51];
1048     tmp[4] = product[52]; tmp[5] = product[53]; tmp[6] = product[54]; tmp[7] = product[55];
1049     tmp[8] = product[56]; tmp[9] = product[57]; tmp[10] = product[58]; tmp[11] = product[59];
1050     tmp[12] = product[60]; tmp[13] = product[61]; tmp[14] = product[62]; tmp[15] = product[63];
1051     tmp[16] = tmp[17] = tmp[18] = tmp[19] = 0;
1052     tmp[20] = tmp[21] = tmp[22] = tmp[23] = 0;
1053     tmp[24] = product[36]; tmp[25] = product[37]; tmp[26] = product[38]; tmp[27] = product[39];
1054     tmp[28] = product[44]; tmp[29] = product[45]; tmp[30] = product[46]; tmp[31] = product[47];
1055     carry -= vli_sub(result, result, tmp);
1056 
1057     /* d3 */
1058     tmp[0] = product[52]; tmp[1] = product[53]; tmp[2] = product[54]; tmp[3] = product[55];
1059     tmp[4] = product[56]; tmp[5] = product[57]; tmp[6] = product[58]; tmp[7] = product[59];
1060     tmp[8] = product[60]; tmp[9] = product[61]; tmp[10] = product[62]; tmp[11] = product[63];
1061     tmp[12] = product[32]; tmp[13] = product[33]; tmp[14] = product[34]; tmp[15] = product[35];
1062     tmp[16] = product[36]; tmp[17] = product[37]; tmp[18] = product[38]; tmp[19] = product[39];
1063     tmp[20] = product[40]; tmp[21] = product[41]; tmp[22] = product[42]; tmp[23] = product[43];
1064     tmp[24] = tmp[25] = tmp[26] = tmp[27] = 0;
1065     tmp[28] = product[48]; tmp[29] = product[49]; tmp[30] = product[50]; tmp[31] = product[51];
1066     carry -= vli_sub(result, result, tmp);
1067 
1068     /* d4 */
1069     tmp[0] = product[56]; tmp[1] = product[57]; tmp[2] = product[58]; tmp[3] = product[59];
1070     tmp[4] = product[60]; tmp[5] = product[61]; tmp[6] = product[62]; tmp[7] = product[63];
1071     tmp[8] = tmp[9] = tmp[10] = tmp[11] = 0;
1072     tmp[12] = product[36]; tmp[13] = product[37]; tmp[14] = product[38]; tmp[15] = product[39];
1073     tmp[16] = product[40]; tmp[17] = product[41]; tmp[18] = product[42]; tmp[19] = product[43];
1074     tmp[20] = product[44]; tmp[21] = product[45]; tmp[22] = product[46]; tmp[23] = product[47];
1075     tmp[24] = tmp[25] = tmp[26] = tmp[27] = 0;
1076     tmp[28] = product[52]; tmp[29] = product[53]; tmp[30] = product[54]; tmp[31] = product[55];
1077     carry -= vli_sub(result, result, tmp);
1078 
1079     if (carry < 0) {
1080         do {
1081             carry += vli_add(result, result, curve_p);
1082         } while (carry < 0);
1083     } else {
1084         while (carry || vli_cmp(curve_p, result) != 1) {
1085             carry -= vli_sub(result, result, curve_p);
1086         }
1087     }
1088 }
1089 #elif uECC_WORD_SIZE == 4
1090 static void vli_mmod_fast(uint32_t *RESTRICT result, uint32_t *RESTRICT product) {
1091     uint32_t tmp[uECC_WORDS];
1092     int carry;
1093 
1094     /* t */
1095     vli_set(result, product);
1096 
1097     /* s1 */
1098     tmp[0] = tmp[1] = tmp[2] = 0;
1099     tmp[3] = product[11];
1100     tmp[4] = product[12];
1101     tmp[5] = product[13];
1102     tmp[6] = product[14];
1103     tmp[7] = product[15];
1104     carry = vli_add(tmp, tmp, tmp);
1105     carry += vli_add(result, result, tmp);
1106 
1107     /* s2 */
1108     tmp[3] = product[12];
1109     tmp[4] = product[13];
1110     tmp[5] = product[14];
1111     tmp[6] = product[15];
1112     tmp[7] = 0;
1113     carry += vli_add(tmp, tmp, tmp);
1114     carry += vli_add(result, result, tmp);
1115 
1116     /* s3 */
1117     tmp[0] = product[8];
1118     tmp[1] = product[9];
1119     tmp[2] = product[10];
1120     tmp[3] = tmp[4] = tmp[5] = 0;
1121     tmp[6] = product[14];
1122     tmp[7] = product[15];
1123     carry += vli_add(result, result, tmp);
1124 
1125     /* s4 */
1126     tmp[0] = product[9];
1127     tmp[1] = product[10];
1128     tmp[2] = product[11];
1129     tmp[3] = product[13];
1130     tmp[4] = product[14];
1131     tmp[5] = product[15];
1132     tmp[6] = product[13];
1133     tmp[7] = product[8];
1134     carry += vli_add(result, result, tmp);
1135 
1136     /* d1 */
1137     tmp[0] = product[11];
1138     tmp[1] = product[12];
1139     tmp[2] = product[13];
1140     tmp[3] = tmp[4] = tmp[5] = 0;
1141     tmp[6] = product[8];
1142     tmp[7] = product[10];
1143     carry -= vli_sub(result, result, tmp);
1144 
1145     /* d2 */
1146     tmp[0] = product[12];
1147     tmp[1] = product[13];
1148     tmp[2] = product[14];
1149     tmp[3] = product[15];
1150     tmp[4] = tmp[5] = 0;
1151     tmp[6] = product[9];
1152     tmp[7] = product[11];
1153     carry -= vli_sub(result, result, tmp);
1154 
1155     /* d3 */
1156     tmp[0] = product[13];
1157     tmp[1] = product[14];
1158     tmp[2] = product[15];
1159     tmp[3] = product[8];
1160     tmp[4] = product[9];
1161     tmp[5] = product[10];
1162     tmp[6] = 0;
1163     tmp[7] = product[12];
1164     carry -= vli_sub(result, result, tmp);
1165 
1166     /* d4 */
1167     tmp[0] = product[14];
1168     tmp[1] = product[15];
1169     tmp[2] = 0;
1170     tmp[3] = product[9];
1171     tmp[4] = product[10];
1172     tmp[5] = product[11];
1173     tmp[6] = 0;
1174     tmp[7] = product[13];
1175     carry -= vli_sub(result, result, tmp);
1176 
1177     if (carry < 0) {
1178         do {
1179             carry += vli_add(result, result, curve_p);
1180         } while (carry < 0);
1181     } else {
1182         while (carry || vli_cmp(curve_p, result) != 1) {
1183             carry -= vli_sub(result, result, curve_p);
1184         }
1185     }
1186 }
1187 #else
1188 static void vli_mmod_fast(uint64_t *RESTRICT result, uint64_t *RESTRICT product) {
1189     uint64_t tmp[uECC_WORDS];
1190     int carry;
1191 
1192     /* t */
1193     vli_set(result, product);
1194 
1195     /* s1 */
1196     tmp[0] = 0;
1197     tmp[1] = product[5] & 0xffffffff00000000ull;
1198     tmp[2] = product[6];
1199     tmp[3] = product[7];
1200     carry = vli_add(tmp, tmp, tmp);
1201     carry += vli_add(result, result, tmp);
1202 
1203     /* s2 */
1204     tmp[1] = product[6] << 32;
1205     tmp[2] = (product[6] >> 32) | (product[7] << 32);
1206     tmp[3] = product[7] >> 32;
1207     carry += vli_add(tmp, tmp, tmp);
1208     carry += vli_add(result, result, tmp);
1209 
1210     /* s3 */
1211     tmp[0] = product[4];
1212     tmp[1] = product[5] & 0xffffffff;
1213     tmp[2] = 0;
1214     tmp[3] = product[7];
1215     carry += vli_add(result, result, tmp);
1216 
1217     /* s4 */
1218     tmp[0] = (product[4] >> 32) | (product[5] << 32);
1219     tmp[1] = (product[5] >> 32) | (product[6] & 0xffffffff00000000ull);
1220     tmp[2] = product[7];
1221     tmp[3] = (product[6] >> 32) | (product[4] << 32);
1222     carry += vli_add(result, result, tmp);
1223 
1224     /* d1 */
1225     tmp[0] = (product[5] >> 32) | (product[6] << 32);
1226     tmp[1] = (product[6] >> 32);
1227     tmp[2] = 0;
1228     tmp[3] = (product[4] & 0xffffffff) | (product[5] << 32);
1229     carry -= vli_sub(result, result, tmp);
1230 
1231     /* d2 */
1232     tmp[0] = product[6];
1233     tmp[1] = product[7];
1234     tmp[2] = 0;
1235     tmp[3] = (product[4] >> 32) | (product[5] & 0xffffffff00000000ull);
1236     carry -= vli_sub(result, result, tmp);
1237 
1238     /* d3 */
1239     tmp[0] = (product[6] >> 32) | (product[7] << 32);
1240     tmp[1] = (product[7] >> 32) | (product[4] << 32);
1241     tmp[2] = (product[4] >> 32) | (product[5] << 32);
1242     tmp[3] = (product[6] << 32);
1243     carry -= vli_sub(result, result, tmp);
1244 
1245     /* d4 */
1246     tmp[0] = product[7];
1247     tmp[1] = product[4] & 0xffffffff00000000ull;
1248     tmp[2] = product[5];
1249     tmp[3] = product[6] & 0xffffffff00000000ull;
1250     carry -= vli_sub(result, result, tmp);
1251 
1252     if (carry < 0) {
1253         do {
1254             carry += vli_add(result, result, curve_p);
1255         } while (carry < 0);
1256     } else {
1257         while (carry || vli_cmp(curve_p, result) != 1) {
1258             carry -= vli_sub(result, result, curve_p);
1259         }
1260     }
1261 }
1262 #endif /* uECC_WORD_SIZE */
1263 
1264 #elif uECC_CURVE == uECC_secp256k1
1265 
1266 #if uECC_WORD_SIZE == 1
1267 static void omega_mult(uint8_t * RESTRICT result, const uint8_t * RESTRICT right) {
1268     /* Multiply by (2^32 + 2^9 + 2^8 + 2^7 + 2^6 + 2^4 + 1). */
1269     uECC_word_t r0 = 0;
1270     uECC_word_t r1 = 0;
1271     uECC_word_t r2 = 0;
1272     wordcount_t k;
1273 
1274     /* Multiply by (2^9 + 2^8 + 2^7 + 2^6 + 2^4 + 1). */
1275     muladd(0xD1, right[0], &r0, &r1, &r2);
1276     result[0] = r0;
1277     r0 = r1;
1278     r1 = r2;
1279     /* r2 is still 0 */
1280 
1281     for (k = 1; k < uECC_WORDS; ++k) {
1282         muladd(0x03, right[k - 1], &r0, &r1, &r2);
1283         muladd(0xD1, right[k], &r0, &r1, &r2);
1284         result[k] = r0;
1285         r0 = r1;
1286         r1 = r2;
1287         r2 = 0;
1288     }
1289     muladd(0x03, right[uECC_WORDS - 1], &r0, &r1, &r2);
1290     result[uECC_WORDS] = r0;
1291     result[uECC_WORDS + 1] = r1;
1292 
1293     result[4 + uECC_WORDS] = vli_add(result + 4, result + 4, right); /* add the 2^32 multiple */
1294 }
1295 #elif uECC_WORD_SIZE == 4
1296 static void omega_mult(uint32_t * RESTRICT result, const uint32_t * RESTRICT right) {
1297     /* Multiply by (2^9 + 2^8 + 2^7 + 2^6 + 2^4 + 1). */
1298     uint32_t carry = 0;
1299     wordcount_t k;
1300 
1301     for (k = 0; k < uECC_WORDS; ++k) {
1302         uint64_t p = (uint64_t)0x3D1 * right[k] + carry;
1303         result[k] = (p & 0xffffffff);
1304         carry = p >> 32;
1305     }
1306     result[uECC_WORDS] = carry;
1307 
1308     result[1 + uECC_WORDS] = vli_add(result + 1, result + 1, right); /* add the 2^32 multiple */
1309 }
1310 #else
1311 static void omega_mult(uint64_t * RESTRICT result, const uint64_t * RESTRICT right) {
1312     uECC_word_t r0 = 0;
1313     uECC_word_t r1 = 0;
1314     uECC_word_t r2 = 0;
1315     wordcount_t k;
1316 
1317     /* Multiply by (2^32 + 2^9 + 2^8 + 2^7 + 2^6 + 2^4 + 1). */
1318     for (k = 0; k < uECC_WORDS; ++k) {
1319         muladd(0x1000003D1ull, right[k], &r0, &r1, &r2);
1320         result[k] = r0;
1321         r0 = r1;
1322         r1 = r2;
1323         r2 = 0;
1324     }
1325     result[uECC_WORDS] = r0;
1326 }
1327 #endif /* uECC_WORD_SIZE */
1328 
1329 #elif uECC_CURVE == uECC_secp224r1
1330 
1331 /* Computes result = product % curve_p
1332    from http://www.nsa.gov/ia/_files/nist-routines.pdf */
1333 #if uECC_WORD_SIZE == 1
1334 // TODO it may be faster to use the omega_mult method when fully asm optimized.
1335 void vli_mmod_fast(uint8_t *RESTRICT result, uint8_t *RESTRICT product) {
1336     uint8_t tmp[uECC_WORDS];
1337     int8_t carry;
1338 
1339     /* t */
1340     vli_set(result, product);
1341 
1342     /* s1 */
1343     tmp[0] = tmp[1] = tmp[2] = tmp[3] = 0;
1344     tmp[4] = tmp[5] = tmp[6] = tmp[7] = 0;
1345     tmp[8] = tmp[9] = tmp[10] = tmp[11] = 0;
1346     tmp[12] = product[28]; tmp[13] = product[29]; tmp[14] = product[30]; tmp[15] = product[31];
1347     tmp[16] = product[32]; tmp[17] = product[33]; tmp[18] = product[34]; tmp[19] = product[35];
1348     tmp[20] = product[36]; tmp[21] = product[37]; tmp[22] = product[38]; tmp[23] = product[39];
1349     tmp[24] = product[40]; tmp[25] = product[41]; tmp[26] = product[42]; tmp[27] = product[43];
1350     carry = vli_add(result, result, tmp);
1351 
1352     /* s2 */
1353     tmp[12] = product[44]; tmp[13] = product[45]; tmp[14] = product[46]; tmp[15] = product[47];
1354     tmp[16] = product[48]; tmp[17] = product[49]; tmp[18] = product[50]; tmp[19] = product[51];
1355     tmp[20] = product[52]; tmp[21] = product[53]; tmp[22] = product[54]; tmp[23] = product[55];
1356     tmp[24] = tmp[25] = tmp[26] = tmp[27] = 0;
1357     carry += vli_add(result, result, tmp);
1358 
1359     /* d1 */
1360     tmp[0]  = product[28]; tmp[1]  = product[29]; tmp[2]  = product[30]; tmp[3]  = product[31];
1361     tmp[4]  = product[32]; tmp[5]  = product[33]; tmp[6]  = product[34]; tmp[7]  = product[35];
1362     tmp[8]  = product[36]; tmp[9]  = product[37]; tmp[10] = product[38]; tmp[11] = product[39];
1363     tmp[12] = product[40]; tmp[13] = product[41]; tmp[14] = product[42]; tmp[15] = product[43];
1364     tmp[16] = product[44]; tmp[17] = product[45]; tmp[18] = product[46]; tmp[19] = product[47];
1365     tmp[20] = product[48]; tmp[21] = product[49]; tmp[22] = product[50]; tmp[23] = product[51];
1366     tmp[24] = product[52]; tmp[25] = product[53]; tmp[26] = product[54]; tmp[27] = product[55];
1367     carry -= vli_sub(result, result, tmp);
1368 
1369     /* d2 */
1370     tmp[0]  = product[44]; tmp[1]  = product[45]; tmp[2]  = product[46]; tmp[3]  = product[47];
1371     tmp[4]  = product[48]; tmp[5]  = product[49]; tmp[6]  = product[50]; tmp[7]  = product[51];
1372     tmp[8]  = product[52]; tmp[9]  = product[53]; tmp[10] = product[54]; tmp[11] = product[55];
1373     tmp[12] = tmp[13] = tmp[14] = tmp[15] = 0;
1374     tmp[16] = tmp[17] = tmp[18] = tmp[19] = 0;
1375     tmp[20] = tmp[21] = tmp[22] = tmp[23] = 0;
1376     tmp[24] = tmp[25] = tmp[26] = tmp[27] = 0;
1377     carry -= vli_sub(result, result, tmp);
1378 
1379     if (carry < 0) {
1380         do {
1381             carry += vli_add(result, result, curve_p);
1382         } while (carry < 0);
1383     } else {
1384         while (carry || vli_cmp(curve_p, result) != 1) {
1385             carry -= vli_sub(result, result, curve_p);
1386         }
1387     }
1388 }
1389 #elif uECC_WORD_SIZE == 4
1390 void vli_mmod_fast(uint32_t *RESTRICT result, uint32_t *RESTRICT product)
1391 {
1392     uint32_t tmp[uECC_WORDS];
1393     int carry;
1394 
1395     /* t */
1396     vli_set(result, product);
1397 
1398     /* s1 */
1399     tmp[0] = tmp[1] = tmp[2] = 0;
1400     tmp[3] = product[7];
1401     tmp[4] = product[8];
1402     tmp[5] = product[9];
1403     tmp[6] = product[10];
1404     carry = vli_add(result, result, tmp);
1405 
1406     /* s2 */
1407     tmp[3] = product[11];
1408     tmp[4] = product[12];
1409     tmp[5] = product[13];
1410     tmp[6] = 0;
1411     carry += vli_add(result, result, tmp);
1412 
1413     /* d1 */
1414     tmp[0] = product[7];
1415     tmp[1] = product[8];
1416     tmp[2] = product[9];
1417     tmp[3] = product[10];
1418     tmp[4] = product[11];
1419     tmp[5] = product[12];
1420     tmp[6] = product[13];
1421     carry -= vli_sub(result, result, tmp);
1422 
1423     /* d2 */
1424     tmp[0] = product[11];
1425     tmp[1] = product[12];
1426     tmp[2] = product[13];
1427     tmp[3] = tmp[4] = tmp[5] = tmp[6] = 0;
1428     carry -= vli_sub(result, result, tmp);
1429 
1430     if (carry < 0) {
1431         do {
1432             carry += vli_add(result, result, curve_p);
1433         } while (carry < 0);
1434     } else {
1435         while (carry || vli_cmp(curve_p, result) != 1) {
1436             carry -= vli_sub(result, result, curve_p);
1437         }
1438     }
1439 }
1440 #endif /* uECC_WORD_SIZE */
1441 
1442 #endif /* uECC_CURVE */
1443 #endif /* !asm_mmod_fast */
1444 
1445 /* Computes result = (left * right) % curve_p. */
1446 static void vli_modMult_fast(uECC_word_t *result,
1447                              const uECC_word_t *left,
1448                              const uECC_word_t *right) {
1449     uECC_word_t product[2 * uECC_WORDS];
1450     vli_mult(product, left, right);
1451     vli_mmod_fast(result, product);
1452 }
1453 
1454 #if uECC_SQUARE_FUNC
1455 
1456 /* Computes result = left^2 % curve_p. */
1457 static void vli_modSquare_fast(uECC_word_t *result, const uECC_word_t *left) {
1458     uECC_word_t product[2 * uECC_WORDS];
1459     vli_square(product, left);
1460     vli_mmod_fast(result, product);
1461 }
1462 
1463 #else /* uECC_SQUARE_FUNC */
1464 
1465 #define vli_modSquare_fast(result, left) vli_modMult_fast((result), (left), (left))
1466 
1467 #endif /* uECC_SQUARE_FUNC */
1468 
1469 
1470 #define EVEN(vli) (!(vli[0] & 1))
1471 /* Computes result = (1 / input) % mod. All VLIs are the same size.
1472    See "From Euclid's GCD to Montgomery Multiplication to the Great Divide"
1473    https://labs.oracle.com/techrep/2001/smli_tr-2001-95.pdf */
1474 #if !asm_modInv
1475 static void vli_modInv(uECC_word_t *result, const uECC_word_t *input, const uECC_word_t *mod) {
1476     uECC_word_t a[uECC_WORDS], b[uECC_WORDS], u[uECC_WORDS], v[uECC_WORDS];
1477     uECC_word_t carry;
1478     cmpresult_t cmpResult;
1479 
1480     if (vli_isZero(input)) {
1481         vli_clear(result);
1482         return;
1483     }
1484 
1485     vli_set(a, input);
1486     vli_set(b, mod);
1487     vli_clear(u);
1488     u[0] = 1;
1489     vli_clear(v);
1490     while ((cmpResult = vli_cmp(a, b)) != 0) {
1491         carry = 0;
1492         if (EVEN(a)) {
1493             vli_rshift1(a);
1494             if (!EVEN(u)) {
1495                 carry = vli_add(u, u, mod);
1496             }
1497             vli_rshift1(u);
1498             if (carry) {
1499                 u[uECC_WORDS - 1] |= HIGH_BIT_SET;
1500             }
1501         } else if (EVEN(b)) {
1502             vli_rshift1(b);
1503             if (!EVEN(v)) {
1504                 carry = vli_add(v, v, mod);
1505             }
1506             vli_rshift1(v);
1507             if (carry) {
1508                 v[uECC_WORDS - 1] |= HIGH_BIT_SET;
1509             }
1510         } else if (cmpResult > 0) {
1511             vli_sub(a, a, b);
1512             vli_rshift1(a);
1513             if (vli_cmp(u, v) < 0) {
1514                 vli_add(u, u, mod);
1515             }
1516             vli_sub(u, u, v);
1517             if (!EVEN(u)) {
1518                 carry = vli_add(u, u, mod);
1519             }
1520             vli_rshift1(u);
1521             if (carry) {
1522                 u[uECC_WORDS - 1] |= HIGH_BIT_SET;
1523             }
1524         } else {
1525             vli_sub(b, b, a);
1526             vli_rshift1(b);
1527             if (vli_cmp(v, u) < 0) {
1528                 vli_add(v, v, mod);
1529             }
1530             vli_sub(v, v, u);
1531             if (!EVEN(v)) {
1532                 carry = vli_add(v, v, mod);
1533             }
1534             vli_rshift1(v);
1535             if (carry) {
1536                 v[uECC_WORDS - 1] |= HIGH_BIT_SET;
1537             }
1538         }
1539     }
1540     vli_set(result, u);
1541 }
1542 #endif /* !asm_modInv */
1543 
1544 /* ------ Point operations ------ */
1545 
1546 /* Returns 1 if 'point' is the point at infinity, 0 otherwise. */
1547 static cmpresult_t EccPoint_isZero(const EccPoint *point) {
1548     return (vli_isZero(point->x) && vli_isZero(point->y));
1549 }
1550 
1551 /* Point multiplication algorithm using Montgomery's ladder with co-Z coordinates.
1552 From http://eprint.iacr.org/2011/338.pdf
1553 */
1554 
1555 /* Double in place */
1556 #if (uECC_CURVE == uECC_secp256k1)
1557 static void EccPoint_double_jacobian(uECC_word_t * RESTRICT X1,
1558                                      uECC_word_t * RESTRICT Y1,
1559                                      uECC_word_t * RESTRICT Z1) {
1560     /* t1 = X, t2 = Y, t3 = Z */
1561     uECC_word_t t4[uECC_WORDS];
1562     uECC_word_t t5[uECC_WORDS];
1563 
1564     if (vli_isZero(Z1)) {
1565         return;
1566     }
1567 
1568     vli_modSquare_fast(t5, Y1);   /* t5 = y1^2 */
1569     vli_modMult_fast(t4, X1, t5); /* t4 = x1*y1^2 = A */
1570     vli_modSquare_fast(X1, X1);   /* t1 = x1^2 */
1571     vli_modSquare_fast(t5, t5);   /* t5 = y1^4 */
1572     vli_modMult_fast(Z1, Y1, Z1); /* t3 = y1*z1 = z3 */
1573 
1574     vli_modAdd(Y1, X1, X1, curve_p); /* t2 = 2*x1^2 */
1575     vli_modAdd(Y1, Y1, X1, curve_p); /* t2 = 3*x1^2 */
1576     if (vli_testBit(Y1, 0)) {
1577         uECC_word_t carry = vli_add(Y1, Y1, curve_p);
1578         vli_rshift1(Y1);
1579         Y1[uECC_WORDS - 1] |= carry << (uECC_WORD_BITS - 1);
1580     } else {
1581         vli_rshift1(Y1);
1582     }
1583     /* t2 = 3/2*(x1^2) = B */
1584 
1585     vli_modSquare_fast(X1, Y1);      /* t1 = B^2 */
1586     vli_modSub(X1, X1, t4, curve_p); /* t1 = B^2 - A */
1587     vli_modSub(X1, X1, t4, curve_p); /* t1 = B^2 - 2A = x3 */
1588 
1589     vli_modSub(t4, t4, X1, curve_p); /* t4 = A - x3 */
1590     vli_modMult_fast(Y1, Y1, t4);    /* t2 = B * (A - x3) */
1591     vli_modSub(Y1, Y1, t5, curve_p); /* t2 = B * (A - x3) - y1^4 = y3 */
1592 }
1593 #else
1594 static void EccPoint_double_jacobian(uECC_word_t * RESTRICT X1,
1595                                      uECC_word_t * RESTRICT Y1,
1596                                      uECC_word_t * RESTRICT Z1) {
1597     /* t1 = X, t2 = Y, t3 = Z */
1598     uECC_word_t t4[uECC_WORDS];
1599     uECC_word_t t5[uECC_WORDS];
1600 
1601     if (vli_isZero(Z1)) {
1602         return;
1603     }
1604 
1605     vli_modSquare_fast(t4, Y1);   /* t4 = y1^2 */
1606     vli_modMult_fast(t5, X1, t4); /* t5 = x1*y1^2 = A */
1607     vli_modSquare_fast(t4, t4);   /* t4 = y1^4 */
1608     vli_modMult_fast(Y1, Y1, Z1); /* t2 = y1*z1 = z3 */
1609     vli_modSquare_fast(Z1, Z1);   /* t3 = z1^2 */
1610 
1611     vli_modAdd(X1, X1, Z1, curve_p); /* t1 = x1 + z1^2 */
1612     vli_modAdd(Z1, Z1, Z1, curve_p); /* t3 = 2*z1^2 */
1613     vli_modSub_fast(Z1, X1, Z1);     /* t3 = x1 - z1^2 */
1614     vli_modMult_fast(X1, X1, Z1);    /* t1 = x1^2 - z1^4 */
1615 
1616     vli_modAdd(Z1, X1, X1, curve_p); /* t3 = 2*(x1^2 - z1^4) */
1617     vli_modAdd(X1, X1, Z1, curve_p); /* t1 = 3*(x1^2 - z1^4) */
1618     if (vli_testBit(X1, 0)) {
1619         uECC_word_t l_carry = vli_add(X1, X1, curve_p);
1620         vli_rshift1(X1);
1621         X1[uECC_WORDS - 1] |= l_carry << (uECC_WORD_BITS - 1);
1622     } else {
1623         vli_rshift1(X1);
1624     }
1625     /* t1 = 3/2*(x1^2 - z1^4) = B */
1626 
1627     vli_modSquare_fast(Z1, X1);   /* t3 = B^2 */
1628     vli_modSub_fast(Z1, Z1, t5);  /* t3 = B^2 - A */
1629     vli_modSub_fast(Z1, Z1, t5);  /* t3 = B^2 - 2A = x3 */
1630     vli_modSub_fast(t5, t5, Z1);  /* t5 = A - x3 */
1631     vli_modMult_fast(X1, X1, t5); /* t1 = B * (A - x3) */
1632     vli_modSub_fast(t4, X1, t4);  /* t4 = B * (A - x3) - y1^4 = y3 */
1633 
1634     vli_set(X1, Z1);
1635     vli_set(Z1, Y1);
1636     vli_set(Y1, t4);
1637 }
1638 #endif
1639 
1640 /* Modify (x1, y1) => (x1 * z^2, y1 * z^3) */
1641 static void apply_z(uECC_word_t * RESTRICT X1,
1642                     uECC_word_t * RESTRICT Y1,
1643                     const uECC_word_t * RESTRICT Z) {
1644     uECC_word_t t1[uECC_WORDS];
1645 
1646     vli_modSquare_fast(t1, Z);    /* z^2 */
1647     vli_modMult_fast(X1, X1, t1); /* x1 * z^2 */
1648     vli_modMult_fast(t1, t1, Z);  /* z^3 */
1649     vli_modMult_fast(Y1, Y1, t1); /* y1 * z^3 */
1650 }
1651 
1652 /* P = (x1, y1) => 2P, (x2, y2) => P' */
1653 static void XYcZ_initial_double(uECC_word_t * RESTRICT X1,
1654                                 uECC_word_t * RESTRICT Y1,
1655                                 uECC_word_t * RESTRICT X2,
1656                                 uECC_word_t * RESTRICT Y2,
1657                                 const uECC_word_t * RESTRICT initial_Z) {
1658     uECC_word_t z[uECC_WORDS];
1659     if (initial_Z) {
1660         vli_set(z, initial_Z);
1661     } else {
1662         vli_clear(z);
1663         z[0] = 1;
1664     }
1665 
1666     vli_set(X2, X1);
1667     vli_set(Y2, Y1);
1668 
1669     apply_z(X1, Y1, z);
1670     EccPoint_double_jacobian(X1, Y1, z);
1671     apply_z(X2, Y2, z);
1672 }
1673 
1674 /* Input P = (x1, y1, Z), Q = (x2, y2, Z)
1675    Output P' = (x1', y1', Z3), P + Q = (x3, y3, Z3)
1676    or P => P', Q => P + Q
1677 */
1678 static void XYcZ_add(uECC_word_t * RESTRICT X1,
1679                      uECC_word_t * RESTRICT Y1,
1680                      uECC_word_t * RESTRICT X2,
1681                      uECC_word_t * RESTRICT Y2) {
1682     /* t1 = X1, t2 = Y1, t3 = X2, t4 = Y2 */
1683     uECC_word_t t5[uECC_WORDS];
1684 
1685     vli_modSub_fast(t5, X2, X1);  /* t5 = x2 - x1 */
1686     vli_modSquare_fast(t5, t5);   /* t5 = (x2 - x1)^2 = A */
1687     vli_modMult_fast(X1, X1, t5); /* t1 = x1*A = B */
1688     vli_modMult_fast(X2, X2, t5); /* t3 = x2*A = C */
1689     vli_modSub_fast(Y2, Y2, Y1);  /* t4 = y2 - y1 */
1690     vli_modSquare_fast(t5, Y2);   /* t5 = (y2 - y1)^2 = D */
1691 
1692     vli_modSub_fast(t5, t5, X1);  /* t5 = D - B */
1693     vli_modSub_fast(t5, t5, X2);  /* t5 = D - B - C = x3 */
1694     vli_modSub_fast(X2, X2, X1);  /* t3 = C - B */
1695     vli_modMult_fast(Y1, Y1, X2); /* t2 = y1*(C - B) */
1696     vli_modSub_fast(X2, X1, t5);  /* t3 = B - x3 */
1697     vli_modMult_fast(Y2, Y2, X2); /* t4 = (y2 - y1)*(B - x3) */
1698     vli_modSub_fast(Y2, Y2, Y1);  /* t4 = y3 */
1699 
1700     vli_set(X2, t5);
1701 }
1702 
1703 /* Input P = (x1, y1, Z), Q = (x2, y2, Z)
1704    Output P + Q = (x3, y3, Z3), P - Q = (x3', y3', Z3)
1705    or P => P - Q, Q => P + Q
1706 */
1707 static void XYcZ_addC(uECC_word_t * RESTRICT X1,
1708                       uECC_word_t * RESTRICT Y1,
1709                       uECC_word_t * RESTRICT X2,
1710                       uECC_word_t * RESTRICT Y2) {
1711     /* t1 = X1, t2 = Y1, t3 = X2, t4 = Y2 */
1712     uECC_word_t t5[uECC_WORDS];
1713     uECC_word_t t6[uECC_WORDS];
1714     uECC_word_t t7[uECC_WORDS];
1715 
1716     vli_modSub_fast(t5, X2, X1);     /* t5 = x2 - x1 */
1717     vli_modSquare_fast(t5, t5);      /* t5 = (x2 - x1)^2 = A */
1718     vli_modMult_fast(X1, X1, t5);    /* t1 = x1*A = B */
1719     vli_modMult_fast(X2, X2, t5);    /* t3 = x2*A = C */
1720     vli_modAdd(t5, Y2, Y1, curve_p); /* t5 = y2 + y1 */
1721     vli_modSub_fast(Y2, Y2, Y1);     /* t4 = y2 - y1 */
1722 
1723     vli_modSub_fast(t6, X2, X1);     /* t6 = C - B */
1724     vli_modMult_fast(Y1, Y1, t6);    /* t2 = y1 * (C - B) = E */
1725     vli_modAdd(t6, X1, X2, curve_p); /* t6 = B + C */
1726     vli_modSquare_fast(X2, Y2);      /* t3 = (y2 - y1)^2 = D */
1727     vli_modSub_fast(X2, X2, t6);     /* t3 = D - (B + C) = x3 */
1728 
1729     vli_modSub_fast(t7, X1, X2);  /* t7 = B - x3 */
1730     vli_modMult_fast(Y2, Y2, t7); /* t4 = (y2 - y1)*(B - x3) */
1731     vli_modSub_fast(Y2, Y2, Y1);  /* t4 = (y2 - y1)*(B - x3) - E = y3 */
1732 
1733     vli_modSquare_fast(t7, t5);   /* t7 = (y2 + y1)^2 = F */
1734     vli_modSub_fast(t7, t7, t6);  /* t7 = F - (B + C) = x3' */
1735     vli_modSub_fast(t6, t7, X1);  /* t6 = x3' - B */
1736     vli_modMult_fast(t6, t6, t5); /* t6 = (y2 + y1)*(x3' - B) */
1737     vli_modSub_fast(Y1, t6, Y1);  /* t2 = (y2 + y1)*(x3' - B) - E = y3' */
1738 
1739     vli_set(X1, t7);
1740 }
1741 
1742 static void EccPoint_mult(EccPoint * RESTRICT result,
1743                           const EccPoint * RESTRICT point,
1744                           const uECC_word_t * RESTRICT scalar,
1745                           const uECC_word_t * RESTRICT initialZ,
1746                           bitcount_t numBits) {
1747     /* R0 and R1 */
1748     uECC_word_t Rx[2][uECC_WORDS];
1749     uECC_word_t Ry[2][uECC_WORDS];
1750     uECC_word_t z[uECC_WORDS];
1751     bitcount_t i;
1752     uECC_word_t nb;
1753 
1754     vli_set(Rx[1], point->x);
1755     vli_set(Ry[1], point->y);
1756 
1757     XYcZ_initial_double(Rx[1], Ry[1], Rx[0], Ry[0], initialZ);
1758 
1759     for (i = numBits - 2; i > 0; --i) {
1760         nb = !vli_testBit(scalar, i);
1761         XYcZ_addC(Rx[1 - nb], Ry[1 - nb], Rx[nb], Ry[nb]);
1762         XYcZ_add(Rx[nb], Ry[nb], Rx[1 - nb], Ry[1 - nb]);
1763     }
1764 
1765     nb = !vli_testBit(scalar, 0);
1766     XYcZ_addC(Rx[1 - nb], Ry[1 - nb], Rx[nb], Ry[nb]);
1767 
1768     /* Find final 1/Z value. */
1769     vli_modSub_fast(z, Rx[1], Rx[0]);   /* X1 - X0 */
1770     vli_modMult_fast(z, z, Ry[1 - nb]); /* Yb * (X1 - X0) */
1771     vli_modMult_fast(z, z, point->x); /* xP * Yb * (X1 - X0) */
1772     vli_modInv(z, z, curve_p);          /* 1 / (xP * Yb * (X1 - X0)) */
1773     vli_modMult_fast(z, z, point->y); /* yP / (xP * Yb * (X1 - X0)) */
1774     vli_modMult_fast(z, z, Rx[1 - nb]); /* Xb * yP / (xP * Yb * (X1 - X0)) */
1775     /* End 1/Z calculation */
1776 
1777     XYcZ_add(Rx[nb], Ry[nb], Rx[1 - nb], Ry[1 - nb]);
1778     apply_z(Rx[0], Ry[0], z);
1779 
1780     vli_set(result->x, Rx[0]);
1781     vli_set(result->y, Ry[0]);
1782 }
1783 
1784 static int EccPoint_compute_public_key(EccPoint *result, uECC_word_t *private) {
1785     uECC_word_t tmp1[uECC_WORDS];
1786     uECC_word_t tmp2[uECC_WORDS];
1787     uECC_word_t *p2[2] = {tmp1, tmp2};
1788     uECC_word_t carry;
1789 
1790     /* Make sure the private key is in the range [1, n-1]. */
1791     if (vli_isZero(private)) {
1792         return 0;
1793     }
1794 
1795 #if (uECC_CURVE == uECC_secp160r1)
1796     // Don't regularize the bitcount for secp160r1, since it would have a larger performance
1797     // impact (about 2% slower on average) and requires the vli_xxx_n functions, leading to
1798     // a significant increase in code size.
1799 
1800     EccPoint_mult(result, &curve_G, private, NULL, vli_numBits(private, uECC_WORDS));
1801 #else
1802     if (vli_cmp(curve_n, private) != 1) {
1803         return 0;
1804     }
1805 
1806     // Regularize the bitcount for the private key so that attackers cannot use a side channel
1807     // attack to learn the number of leading zeros.
1808     carry = vli_add(tmp1, private, curve_n);
1809     vli_add(tmp2, tmp1, curve_n);
1810     EccPoint_mult(result, &curve_G, p2[!carry], NULL, (uECC_BYTES * 8) + 1);
1811 #endif
1812 
1813     if (EccPoint_isZero(result)) {
1814         return 0;
1815     }
1816     return 1;
1817 }
1818 
1819 #if uECC_CURVE == uECC_secp224r1
1820 
1821 /* Routine 3.2.4 RS;  from http://www.nsa.gov/ia/_files/nist-routines.pdf */
1822 static void mod_sqrt_secp224r1_rs(uECC_word_t *d1,
1823                                   uECC_word_t *e1,
1824                                   uECC_word_t *f1,
1825                                   const uECC_word_t *d0,
1826                                   const uECC_word_t *e0,
1827                                   const uECC_word_t *f0) {
1828     uECC_word_t t[uECC_WORDS];
1829 
1830     vli_modSquare_fast(t, d0);                 /* t <-- d0 ^ 2 */
1831     vli_modMult_fast(e1, d0, e0);              /* e1 <-- d0 * e0 */
1832     vli_modAdd(d1, t, f0, curve_p);            /* d1 <-- t  + f0 */
1833     vli_modAdd(e1, e1, e1, curve_p);           /* e1 <-- e1 + e1 */
1834     vli_modMult_fast(f1, t, f0);               /* f1 <-- t  * f0 */
1835     vli_modAdd(f1, f1, f1, curve_p);           /* f1 <-- f1 + f1 */
1836     vli_modAdd(f1, f1, f1, curve_p);           /* f1 <-- f1 + f1 */
1837 }
1838 
1839 /* Routine 3.2.5 RSS;  from http://www.nsa.gov/ia/_files/nist-routines.pdf */
1840 static void mod_sqrt_secp224r1_rss(uECC_word_t *d1,
1841                                    uECC_word_t *e1,
1842                                    uECC_word_t *f1,
1843                                    const uECC_word_t *d0,
1844                                    const uECC_word_t *e0,
1845                                    const uECC_word_t *f0,
1846                                    const bitcount_t j) {
1847     bitcount_t i;
1848 
1849     vli_set(d1, d0);                           /* d1 <-- d0 */
1850     vli_set(e1, e0);                           /* e1 <-- e0 */
1851     vli_set(f1, f0);                           /* f1 <-- f0 */
1852     for (i = 1; i <= j; i++) {
1853         mod_sqrt_secp224r1_rs(d1, e1, f1, d1, e1, f1); /* RS (d1,e1,f1,d1,e1,f1) */
1854     }
1855 }
1856 
1857 /* Routine 3.2.6 RM;  from http://www.nsa.gov/ia/_files/nist-routines.pdf */
1858 static void mod_sqrt_secp224r1_rm(uECC_word_t *d2,
1859                                   uECC_word_t *e2,
1860                                   uECC_word_t *f2,
1861                                   const uECC_word_t *c,
1862                                   const uECC_word_t *d0,
1863                                   const uECC_word_t *e0,
1864                                   const uECC_word_t *d1,
1865                                   const uECC_word_t *e1) {
1866     uECC_word_t t1[uECC_WORDS];
1867     uECC_word_t t2[uECC_WORDS];
1868 
1869     vli_modMult_fast(t1, e0, e1);              /* t1 <-- e0 * e1 */
1870     vli_modMult_fast(t1, t1, c);               /* t1 <-- t1 * c */
1871     vli_modSub_fast(t1, curve_p, t1);          /* t1 <-- p  - t1 */
1872     vli_modMult_fast(t2, d0, d1);              /* t2 <-- d0 * d1 */
1873     vli_modAdd(t2, t2, t1, curve_p);           /* t2 <-- t2 + t1 */
1874     vli_modMult_fast(t1, d0, e1);              /* t1 <-- d0 * e1 */
1875     vli_modMult_fast(e2, d1, e0);              /* e2 <-- d1 * e0 */
1876     vli_modAdd(e2, e2, t1, curve_p);           /* e2 <-- e2 + t1 */
1877     vli_modSquare_fast(f2, e2);                /* f2 <-- e2^2 */
1878     vli_modMult_fast(f2, f2, c);               /* f2 <-- f2 * c */
1879     vli_modSub_fast(f2, curve_p, f2);          /* f2 <-- p  - f2 */
1880     vli_set(d2, t2);                           /* d2 <-- t2 */
1881 }
1882 
1883 /* Routine 3.2.7 RP;  from http://www.nsa.gov/ia/_files/nist-routines.pdf */
1884 static void mod_sqrt_secp224r1_rp(uECC_word_t *d1,
1885                                   uECC_word_t *e1,
1886                                   uECC_word_t *f1,
1887                                   const uECC_word_t *c,
1888                                   const uECC_word_t *r) {
1889     wordcount_t i;
1890     wordcount_t pow2i = 1;
1891     uECC_word_t d0[uECC_WORDS];
1892     uECC_word_t e0[uECC_WORDS] = {1};          /* e0 <-- 1 */
1893     uECC_word_t f0[uECC_WORDS];
1894 
1895     vli_set(d0, r);                            /* d0 <-- r */
1896     vli_modSub_fast(f0, curve_p, c);           /* f0 <-- p  - c */
1897     for (i = 0; i <= 6; i++) {
1898         mod_sqrt_secp224r1_rss(d1, e1, f1, d0, e0, f0, pow2i); /* RSS (d1,e1,f1,d0,e0,f0,2^i) */
1899         mod_sqrt_secp224r1_rm(d1, e1, f1, c, d1, e1, d0, e0);  /* RM (d1,e1,f1,c,d1,e1,d0,e0) */
1900         vli_set(d0, d1);                       /* d0 <-- d1 */
1901         vli_set(e0, e1);                       /* e0 <-- e1 */
1902         vli_set(f0, f1);                       /* f0 <-- f1 */
1903         pow2i *= 2;
1904     }
1905 }
1906 
1907 /* Compute a = sqrt(a) (mod curve_p). */
1908 /* Routine 3.2.8 mp_mod_sqrt_224; from http://www.nsa.gov/ia/_files/nist-routines.pdf */
1909 static void mod_sqrt(uECC_word_t *a) {
1910     bitcount_t i;
1911     uECC_word_t e1[uECC_WORDS];
1912     uECC_word_t f1[uECC_WORDS];
1913     uECC_word_t d0[uECC_WORDS];
1914     uECC_word_t e0[uECC_WORDS];
1915     uECC_word_t f0[uECC_WORDS];
1916     uECC_word_t d1[uECC_WORDS];
1917 
1918     // s = a; using constant instead of random value
1919     mod_sqrt_secp224r1_rp(d0, e0, f0, a, a);           /* RP (d0, e0, f0, c, s) */
1920     mod_sqrt_secp224r1_rs(d1, e1, f1, d0, e0, f0);     /* RS (d1, e1, f1, d0, e0, f0) */
1921     for (i = 1; i <= 95; i++) {
1922         vli_set(d0, d1);                               /* d0 <-- d1 */
1923         vli_set(e0, e1);                               /* e0 <-- e1 */
1924         vli_set(f0, f1);                               /* f0 <-- f1 */
1925         mod_sqrt_secp224r1_rs(d1, e1, f1, d0, e0, f0); /* RS (d1, e1, f1, d0, e0, f0) */
1926         if (vli_isZero(d1)) {                          /* if d1 == 0 */
1927 	        break;
1928         }
1929     }
1930     vli_modInv(f1, e0, curve_p);                       /* f1 <-- 1 / e0 */
1931     vli_modMult_fast(a, d0, f1);                       /* a  <-- d0 / e0 */
1932 }
1933 
1934 #else /* uECC_CURVE */
1935 
1936 /* Compute a = sqrt(a) (mod curve_p). */
1937 static void mod_sqrt(uECC_word_t *a) {
1938     bitcount_t i;
1939     uECC_word_t p1[uECC_WORDS] = {1};
1940     uECC_word_t l_result[uECC_WORDS] = {1};
1941 
1942     /* Since curve_p == 3 (mod 4) for all supported curves, we can
1943        compute sqrt(a) = a^((curve_p + 1) / 4) (mod curve_p). */
1944     vli_add(p1, curve_p, p1); /* p1 = curve_p + 1 */
1945     for (i = vli_numBits(p1, uECC_WORDS) - 1; i > 1; --i) {
1946         vli_modSquare_fast(l_result, l_result);
1947         if (vli_testBit(p1, i)) {
1948             vli_modMult_fast(l_result, l_result, a);
1949         }
1950     }
1951     vli_set(a, l_result);
1952 }
1953 #endif /* uECC_CURVE */
1954 
1955 #if uECC_WORD_SIZE == 1
1956 
1957 static void vli_nativeToBytes(uint8_t * RESTRICT dest, const uint8_t * RESTRICT src) {
1958     uint8_t i;
1959     for (i = 0; i < uECC_BYTES; ++i) {
1960         dest[i] = src[(uECC_BYTES - 1) - i];
1961     }
1962 }
1963 
1964 #define vli_bytesToNative(dest, src) vli_nativeToBytes((dest), (src))
1965 
1966 #elif uECC_WORD_SIZE == 4
1967 
1968 static void vli_nativeToBytes(uint8_t *bytes, const uint32_t *native) {
1969     unsigned i;
1970     for (i = 0; i < uECC_WORDS; ++i) {
1971         uint8_t *digit = bytes + 4 * (uECC_WORDS - 1 - i);
1972         digit[0] = native[i] >> 24;
1973         digit[1] = native[i] >> 16;
1974         digit[2] = native[i] >> 8;
1975         digit[3] = native[i];
1976     }
1977 }
1978 
1979 static void vli_bytesToNative(uint32_t *native, const uint8_t *bytes) {
1980     unsigned i;
1981     for (i = 0; i < uECC_WORDS; ++i) {
1982         const uint8_t *digit = bytes + 4 * (uECC_WORDS - 1 - i);
1983         native[i] = ((uint32_t)digit[0] << 24) | ((uint32_t)digit[1] << 16) |
1984                     ((uint32_t)digit[2] << 8) | (uint32_t)digit[3];
1985     }
1986 }
1987 
1988 #else
1989 
1990 static void vli_nativeToBytes(uint8_t *bytes, const uint64_t *native) {
1991     unsigned i;
1992     for (i = 0; i < uECC_WORDS; ++i) {
1993         uint8_t *digit = bytes + 8 * (uECC_WORDS - 1 - i);
1994         digit[0] = native[i] >> 56;
1995         digit[1] = native[i] >> 48;
1996         digit[2] = native[i] >> 40;
1997         digit[3] = native[i] >> 32;
1998         digit[4] = native[i] >> 24;
1999         digit[5] = native[i] >> 16;
2000         digit[6] = native[i] >> 8;
2001         digit[7] = native[i];
2002     }
2003 }
2004 
2005 static void vli_bytesToNative(uint64_t *native, const uint8_t *bytes) {
2006     unsigned i;
2007     for (i = 0; i < uECC_WORDS; ++i) {
2008         const uint8_t *digit = bytes + 8 * (uECC_WORDS - 1 - i);
2009         native[i] = ((uint64_t)digit[0] << 56) | ((uint64_t)digit[1] << 48) |
2010                     ((uint64_t)digit[2] << 40) | ((uint64_t)digit[3] << 32) |
2011                     ((uint64_t)digit[4] << 24) | ((uint64_t)digit[5] << 16) |
2012                     ((uint64_t)digit[6] << 8) | (uint64_t)digit[7];
2013     }
2014 }
2015 
2016 #endif /* uECC_WORD_SIZE */
2017 
2018 int uECC_make_key(uint8_t public_key[uECC_BYTES*2], uint8_t private_key[uECC_BYTES]) {
2019     uECC_word_t private[uECC_WORDS];
2020     EccPoint public;
2021     uECC_word_t tries;
2022     for (tries = 0; tries < MAX_TRIES; ++tries) {
2023         if (g_rng_function((uint8_t *)private, sizeof(private)) &&
2024                 EccPoint_compute_public_key(&public, private)) {
2025             vli_nativeToBytes(private_key, private);
2026             vli_nativeToBytes(public_key, public.x);
2027             vli_nativeToBytes(public_key + uECC_BYTES, public.y);
2028             return 1;
2029         }
2030     }
2031     return 0;
2032 }
2033 
2034 int uECC_shared_secret(const uint8_t public_key[uECC_BYTES*2],
2035                        const uint8_t private_key[uECC_BYTES],
2036                        uint8_t secret[uECC_BYTES]) {
2037     EccPoint public;
2038     EccPoint product;
2039     uECC_word_t private[uECC_WORDS];
2040     uECC_word_t tmp[uECC_WORDS];
2041     uECC_word_t *p2[2] = {private, tmp};
2042     uECC_word_t random[uECC_WORDS];
2043     uECC_word_t *initial_Z = NULL;
2044     uECC_word_t tries;
2045     uECC_word_t carry;
2046 
2047     // Try to get a random initial Z value to improve protection against side-channel
2048     // attacks. If the RNG fails every time (eg it was not defined), we continue so that
2049     // uECC_shared_secret() can still work without an RNG defined.
2050     for (tries = 0; tries < MAX_TRIES; ++tries) {
2051         if (g_rng_function((uint8_t *)random, sizeof(random)) && !vli_isZero(random)) {
2052             initial_Z = random;
2053             break;
2054         }
2055     }
2056 
2057     vli_bytesToNative(private, private_key);
2058     vli_bytesToNative(public.x, public_key);
2059     vli_bytesToNative(public.y, public_key + uECC_BYTES);
2060 
2061 #if (uECC_CURVE == uECC_secp160r1)
2062     // Don't regularize the bitcount for secp160r1.
2063     EccPoint_mult(&product, &public, private, initial_Z, vli_numBits(private, uECC_WORDS));
2064 #else
2065     // Regularize the bitcount for the private key so that attackers cannot use a side channel
2066     // attack to learn the number of leading zeros.
2067     carry = vli_add(private, private, curve_n);
2068     vli_add(tmp, private, curve_n);
2069     EccPoint_mult(&product, &public, p2[!carry], initial_Z, (uECC_BYTES * 8) + 1);
2070 #endif
2071 
2072     vli_nativeToBytes(secret, product.x);
2073     return !EccPoint_isZero(&product);
2074 }
2075 
2076 void uECC_compress(const uint8_t public_key[uECC_BYTES*2], uint8_t compressed[uECC_BYTES+1]) {
2077     wordcount_t i;
2078     for (i = 0; i < uECC_BYTES; ++i) {
2079         compressed[i+1] = public_key[i];
2080     }
2081     compressed[0] = 2 + (public_key[uECC_BYTES * 2 - 1] & 0x01);
2082 }
2083 
2084 /* Computes result = x^3 + ax + b. result must not overlap x. */
2085 static void curve_x_side(uECC_word_t * RESTRICT result, const uECC_word_t * RESTRICT x) {
2086 #if (uECC_CURVE == uECC_secp256k1)
2087     vli_modSquare_fast(result, x); /* r = x^2 */
2088     vli_modMult_fast(result, result, x); /* r = x^3 */
2089     vli_modAdd(result, result, curve_b, curve_p); /* r = x^3 + b */
2090 #else
2091     uECC_word_t _3[uECC_WORDS] = {3}; /* -a = 3 */
2092 
2093     vli_modSquare_fast(result, x); /* r = x^2 */
2094     vli_modSub_fast(result, result, _3); /* r = x^2 - 3 */
2095     vli_modMult_fast(result, result, x); /* r = x^3 - 3x */
2096     vli_modAdd(result, result, curve_b, curve_p); /* r = x^3 - 3x + b */
2097 #endif
2098 }
2099 
2100 void uECC_decompress(const uint8_t compressed[uECC_BYTES+1], uint8_t public_key[uECC_BYTES*2]) {
2101     EccPoint point;
2102     vli_bytesToNative(point.x, compressed + 1);
2103     curve_x_side(point.y, point.x);
2104     mod_sqrt(point.y);
2105 
2106     if ((point.y[0] & 0x01) != (compressed[0] & 0x01)) {
2107         vli_sub(point.y, curve_p, point.y);
2108     }
2109 
2110     vli_nativeToBytes(public_key, point.x);
2111     vli_nativeToBytes(public_key + uECC_BYTES, point.y);
2112 }
2113 
2114 int uECC_valid_public_key(const uint8_t public_key[uECC_BYTES*2]) {
2115     uECC_word_t tmp1[uECC_WORDS];
2116     uECC_word_t tmp2[uECC_WORDS];
2117     EccPoint public;
2118 
2119     vli_bytesToNative(public.x, public_key);
2120     vli_bytesToNative(public.y, public_key + uECC_BYTES);
2121 
2122     // The point at infinity is invalid.
2123     if (EccPoint_isZero(&public)) {
2124         return 0;
2125     }
2126 
2127     // x and y must be smaller than p.
2128     if (vli_cmp(curve_p, public.x) != 1 || vli_cmp(curve_p, public.y) != 1) {
2129         return 0;
2130     }
2131 
2132     vli_modSquare_fast(tmp1, public.y); /* tmp1 = y^2 */
2133     curve_x_side(tmp2, public.x); /* tmp2 = x^3 + ax + b */
2134 
2135     /* Make sure that y^2 == x^3 + ax + b */
2136     return (vli_cmp(tmp1, tmp2) == 0);
2137 }
2138 
2139 int uECC_compute_public_key(const uint8_t private_key[uECC_BYTES],
2140                             uint8_t public_key[uECC_BYTES * 2]) {
2141     uECC_word_t private[uECC_WORDS];
2142     EccPoint public;
2143 
2144     vli_bytesToNative(private, private_key);
2145 
2146     if (!EccPoint_compute_public_key(&public, private)) {
2147         return 0;
2148     }
2149 
2150     vli_nativeToBytes(public_key, public.x);
2151     vli_nativeToBytes(public_key + uECC_BYTES, public.y);
2152     return 1;
2153 }
2154 
2155 int uECC_bytes(void) {
2156     return uECC_BYTES;
2157 }
2158 
2159 int uECC_curve(void) {
2160     return uECC_CURVE;
2161 }
2162 
2163 /* -------- ECDSA code -------- */
2164 
2165 #if (uECC_CURVE == uECC_secp160r1)
2166 static void vli_clear_n(uECC_word_t *vli) {
2167     vli_clear(vli);
2168     vli[uECC_N_WORDS - 1] = 0;
2169 }
2170 
2171 static uECC_word_t vli_isZero_n(const uECC_word_t *vli) {
2172     if (vli[uECC_N_WORDS - 1]) {
2173         return 0;
2174     }
2175     return vli_isZero(vli);
2176 }
2177 
2178 static void vli_set_n(uECC_word_t *dest, const uECC_word_t *src) {
2179     vli_set(dest, src);
2180     dest[uECC_N_WORDS - 1] = src[uECC_N_WORDS - 1];
2181 }
2182 
2183 static cmpresult_t vli_cmp_n(const uECC_word_t *left, const uECC_word_t *right) {
2184     if (left[uECC_N_WORDS - 1] > right[uECC_N_WORDS - 1]) {
2185         return 1;
2186     } else if (left[uECC_N_WORDS - 1] < right[uECC_N_WORDS - 1]) {
2187         return -1;
2188     }
2189     return vli_cmp(left, right);
2190 }
2191 
2192 static void vli_rshift1_n(uECC_word_t *vli) {
2193     vli_rshift1(vli);
2194     vli[uECC_N_WORDS - 2] |= vli[uECC_N_WORDS - 1] << (uECC_WORD_BITS - 1);
2195     vli[uECC_N_WORDS - 1] = vli[uECC_N_WORDS - 1] >> 1;
2196 }
2197 
2198 static uECC_word_t vli_add_n(uECC_word_t *result,
2199                              const uECC_word_t *left,
2200                              const uECC_word_t *right) {
2201     uECC_word_t carry = vli_add(result, left, right);
2202     uECC_word_t sum = left[uECC_N_WORDS - 1] + right[uECC_N_WORDS - 1] + carry;
2203     if (sum != left[uECC_N_WORDS - 1]) {
2204         carry = (sum < left[uECC_N_WORDS - 1]);
2205     }
2206     result[uECC_N_WORDS - 1] = sum;
2207     return carry;
2208 }
2209 
2210 static uECC_word_t vli_sub_n(uECC_word_t *result,
2211                              const uECC_word_t *left,
2212                              const uECC_word_t *right) {
2213     uECC_word_t borrow = vli_sub(result, left, right);
2214     uECC_word_t diff = left[uECC_N_WORDS - 1] - right[uECC_N_WORDS - 1] - borrow;
2215     if (diff != left[uECC_N_WORDS - 1]) {
2216         borrow = (diff > left[uECC_N_WORDS - 1]);
2217     }
2218     result[uECC_N_WORDS - 1] = diff;
2219     return borrow;
2220 }
2221 
2222 #if !muladd_exists
2223 static void muladd(uECC_word_t a,
2224                    uECC_word_t b,
2225                    uECC_word_t *r0,
2226                    uECC_word_t *r1,
2227                    uECC_word_t *r2) {
2228     uECC_dword_t p = (uECC_dword_t)a * b;
2229     uECC_dword_t r01 = ((uECC_dword_t)(*r1) << uECC_WORD_BITS) | *r0;
2230     r01 += p;
2231     *r2 += (r01 < p);
2232     *r1 = r01 >> uECC_WORD_BITS;
2233     *r0 = (uECC_word_t)r01;
2234 }
2235 #define muladd_exists 1
2236 #endif
2237 
2238 static void vli_mult_n(uECC_word_t *result, const uECC_word_t *left, const uECC_word_t *right) {
2239     uECC_word_t r0 = 0;
2240     uECC_word_t r1 = 0;
2241     uECC_word_t r2 = 0;
2242     wordcount_t i, k;
2243 
2244     for (k = 0; k < uECC_N_WORDS * 2 - 1; ++k) {
2245         wordcount_t min = (k < uECC_N_WORDS ? 0 : (k + 1) - uECC_N_WORDS);
2246         wordcount_t max = (k < uECC_N_WORDS ? k : uECC_N_WORDS - 1);
2247         for (i = min; i <= max; ++i) {
2248             muladd(left[i], right[k - i], &r0, &r1, &r2);
2249         }
2250         result[k] = r0;
2251         r0 = r1;
2252         r1 = r2;
2253         r2 = 0;
2254     }
2255     result[uECC_N_WORDS * 2 - 1] = r0;
2256 }
2257 
2258 static void vli_modAdd_n(uECC_word_t *result,
2259                          const uECC_word_t *left,
2260                          const uECC_word_t *right,
2261                          const uECC_word_t *mod) {
2262     uECC_word_t carry = vli_add_n(result, left, right);
2263     if (carry || vli_cmp_n(result, mod) >= 0) {
2264         vli_sub_n(result, result, mod);
2265     }
2266 }
2267 
2268 static void vli_modInv_n(uECC_word_t *result, const uECC_word_t *input, const uECC_word_t *mod) {
2269     uECC_word_t a[uECC_N_WORDS], b[uECC_N_WORDS], u[uECC_N_WORDS], v[uECC_N_WORDS];
2270     uECC_word_t carry;
2271     cmpresult_t cmpResult;
2272 
2273     if (vli_isZero_n(input)) {
2274         vli_clear_n(result);
2275         return;
2276     }
2277 
2278     vli_set_n(a, input);
2279     vli_set_n(b, mod);
2280     vli_clear_n(u);
2281     u[0] = 1;
2282     vli_clear_n(v);
2283     while ((cmpResult = vli_cmp_n(a, b)) != 0) {
2284         carry = 0;
2285         if (EVEN(a)) {
2286             vli_rshift1_n(a);
2287             if (!EVEN(u)) {
2288                 carry = vli_add_n(u, u, mod);
2289             }
2290             vli_rshift1_n(u);
2291             if (carry) {
2292                 u[uECC_N_WORDS - 1] |= HIGH_BIT_SET;
2293             }
2294         } else if (EVEN(b)) {
2295             vli_rshift1_n(b);
2296             if (!EVEN(v)) {
2297                 carry = vli_add_n(v, v, mod);
2298             }
2299             vli_rshift1_n(v);
2300             if (carry) {
2301                 v[uECC_N_WORDS - 1] |= HIGH_BIT_SET;
2302             }
2303         } else if (cmpResult > 0) {
2304             vli_sub_n(a, a, b);
2305             vli_rshift1_n(a);
2306             if (vli_cmp_n(u, v) < 0) {
2307                 vli_add_n(u, u, mod);
2308             }
2309             vli_sub_n(u, u, v);
2310             if (!EVEN(u)) {
2311                 carry = vli_add_n(u, u, mod);
2312             }
2313             vli_rshift1_n(u);
2314             if (carry) {
2315                 u[uECC_N_WORDS - 1] |= HIGH_BIT_SET;
2316             }
2317         } else {
2318             vli_sub_n(b, b, a);
2319             vli_rshift1_n(b);
2320             if (vli_cmp_n(v, u) < 0) {
2321                 vli_add_n(v, v, mod);
2322             }
2323             vli_sub_n(v, v, u);
2324             if (!EVEN(v)) {
2325                 carry = vli_add_n(v, v, mod);
2326             }
2327             vli_rshift1_n(v);
2328             if (carry) {
2329                 v[uECC_N_WORDS - 1] |= HIGH_BIT_SET;
2330             }
2331         }
2332     }
2333     vli_set_n(result, u);
2334 }
2335 
2336 static void vli2_rshift1_n(uECC_word_t *vli) {
2337     vli_rshift1_n(vli);
2338     vli[uECC_N_WORDS - 1] |= vli[uECC_N_WORDS] << (uECC_WORD_BITS - 1);
2339     vli_rshift1_n(vli + uECC_N_WORDS);
2340 }
2341 
2342 static uECC_word_t vli2_sub_n(uECC_word_t *result,
2343                               const uECC_word_t *left,
2344                               const uECC_word_t *right) {
2345     uECC_word_t borrow = 0;
2346     wordcount_t i;
2347     for (i = 0; i < uECC_N_WORDS * 2; ++i) {
2348         uECC_word_t diff = left[i] - right[i] - borrow;
2349         if (diff != left[i]) {
2350             borrow = (diff > left[i]);
2351         }
2352         result[i] = diff;
2353     }
2354     return borrow;
2355 }
2356 
2357 /* Computes result = (left * right) % curve_n. */
2358 static void vli_modMult_n(uECC_word_t *result, const uECC_word_t *left, const uECC_word_t *right) {
2359     bitcount_t i;
2360     uECC_word_t product[2 * uECC_N_WORDS];
2361     uECC_word_t modMultiple[2 * uECC_N_WORDS];
2362     uECC_word_t tmp[2 * uECC_N_WORDS];
2363     uECC_word_t *v[2] = {tmp, product};
2364     uECC_word_t index = 1;
2365 
2366     vli_mult_n(product, left, right);
2367     vli_clear_n(modMultiple);
2368     vli_set(modMultiple + uECC_N_WORDS + 1, curve_n);
2369     vli_rshift1(modMultiple + uECC_N_WORDS + 1);
2370     modMultiple[2 * uECC_N_WORDS - 1] |= HIGH_BIT_SET;
2371     modMultiple[uECC_N_WORDS] = HIGH_BIT_SET;
2372 
2373     for (i = 0;
2374          i <= ((((bitcount_t)uECC_N_WORDS) << uECC_WORD_BITS_SHIFT) + (uECC_WORD_BITS - 1));
2375          ++i) {
2376         uECC_word_t borrow = vli2_sub_n(v[1 - index], v[index], modMultiple);
2377         index = !(index ^ borrow); /* Swap the index if there was no borrow */
2378         vli2_rshift1_n(modMultiple);
2379     }
2380     vli_set_n(result, v[index]);
2381 }
2382 
2383 #else
2384 
2385 #define vli_cmp_n vli_cmp
2386 #define vli_modInv_n vli_modInv
2387 #define vli_modAdd_n vli_modAdd
2388 
2389 static void vli2_rshift1(uECC_word_t *vli) {
2390     vli_rshift1(vli);
2391     vli[uECC_WORDS - 1] |= vli[uECC_WORDS] << (uECC_WORD_BITS - 1);
2392     vli_rshift1(vli + uECC_WORDS);
2393 }
2394 
2395 static uECC_word_t vli2_sub(uECC_word_t *result,
2396                             const uECC_word_t *left,
2397                             const uECC_word_t *right) {
2398     uECC_word_t borrow = 0;
2399     wordcount_t i;
2400     for (i = 0; i < uECC_WORDS * 2; ++i) {
2401         uECC_word_t diff = left[i] - right[i] - borrow;
2402         if (diff != left[i]) {
2403             borrow = (diff > left[i]);
2404         }
2405         result[i] = diff;
2406     }
2407     return borrow;
2408 }
2409 
2410 /* Computes result = (left * right) % curve_n. */
2411 static void vli_modMult_n(uECC_word_t *result, const uECC_word_t *left, const uECC_word_t *right) {
2412     uECC_word_t product[2 * uECC_WORDS];
2413     uECC_word_t modMultiple[2 * uECC_WORDS];
2414     uECC_word_t tmp[2 * uECC_WORDS];
2415     uECC_word_t *v[2] = {tmp, product};
2416     bitcount_t i;
2417     uECC_word_t index = 1;
2418 
2419     vli_mult(product, left, right);
2420     vli_set(modMultiple + uECC_WORDS, curve_n); /* works if curve_n has its highest bit set */
2421     vli_clear(modMultiple);
2422 
2423     for (i = 0; i <= uECC_BYTES * 8; ++i) {
2424         uECC_word_t borrow = vli2_sub(v[1 - index], v[index], modMultiple);
2425         index = !(index ^ borrow); /* Swap the index if there was no borrow */
2426         vli2_rshift1(modMultiple);
2427     }
2428     vli_set(result, v[index]);
2429 }
2430 #endif /* (uECC_CURVE != uECC_secp160r1) */
2431 
2432 static int uECC_sign_with_k(const uint8_t private_key[uECC_BYTES],
2433                             const uint8_t message_hash[uECC_BYTES],
2434                             uECC_word_t k[uECC_N_WORDS],
2435                             uint8_t signature[uECC_BYTES*2]) {
2436     uECC_word_t tmp[uECC_N_WORDS];
2437     uECC_word_t s[uECC_N_WORDS];
2438     uECC_word_t *k2[2] = {tmp, s};
2439     EccPoint p;
2440     uECC_word_t carry;
2441     uECC_word_t tries;
2442 
2443     /* Make sure 0 < k < curve_n */
2444     if (vli_isZero(k) || vli_cmp_n(curve_n, k) != 1) {
2445         return 0;
2446     }
2447 
2448 #if (uECC_CURVE == uECC_secp160r1)
2449     /* Make sure that we don't leak timing information about k.
2450        See http://eprint.iacr.org/2011/232.pdf */
2451     vli_add_n(tmp, k, curve_n);
2452     carry = (tmp[uECC_WORDS] & 0x02);
2453     vli_add_n(s, tmp, curve_n);
2454 
2455     /* p = k * G */
2456     EccPoint_mult(&p, &curve_G, k2[!carry], NULL, (uECC_BYTES * 8) + 2);
2457 #else
2458     /* Make sure that we don't leak timing information about k.
2459        See http://eprint.iacr.org/2011/232.pdf */
2460     carry = vli_add(tmp, k, curve_n);
2461     vli_add(s, tmp, curve_n);
2462 
2463     /* p = k * G */
2464     EccPoint_mult(&p, &curve_G, k2[!carry], NULL, (uECC_BYTES * 8) + 1);
2465 
2466     /* r = x1 (mod n) */
2467     if (vli_cmp(curve_n, p.x) != 1) {
2468         vli_sub(p.x, p.x, curve_n);
2469     }
2470 #endif
2471     if (vli_isZero(p.x)) {
2472         return 0;
2473     }
2474 
2475     // Attempt to get a random number to prevent side channel analysis of k.
2476     // If the RNG fails every time (eg it was not defined), we continue so that
2477     // deterministic signing can still work (with reduced security) without
2478     // an RNG defined.
2479     carry = 0; // use to signal that the RNG succeeded at least once.
2480     for (tries = 0; tries < MAX_TRIES; ++tries) {
2481         if (!g_rng_function((uint8_t *)tmp, sizeof(tmp))) {
2482             continue;
2483         }
2484         carry = 1;
2485         if (!vli_isZero(tmp)) {
2486             break;
2487         }
2488     }
2489     if (!carry) {
2490         vli_clear(tmp);
2491         tmp[0] = 1;
2492     }
2493 
2494     /* Prevent side channel analysis of vli_modInv() to determine
2495        bits of k / the private key by premultiplying by a random number */
2496     vli_modMult_n(k, k, tmp); /* k' = rand * k */
2497     vli_modInv_n(k, k, curve_n); /* k = 1 / k' */
2498     vli_modMult_n(k, k, tmp); /* k = 1 / k */
2499 
2500     vli_nativeToBytes(signature, p.x); /* store r */
2501 
2502     tmp[uECC_N_WORDS - 1] = 0;
2503     vli_bytesToNative(tmp, private_key); /* tmp = d */
2504     s[uECC_N_WORDS - 1] = 0;
2505     vli_set(s, p.x);
2506     vli_modMult_n(s, tmp, s); /* s = r*d */
2507 
2508     vli_bytesToNative(tmp, message_hash);
2509     vli_modAdd_n(s, tmp, s, curve_n); /* s = e + r*d */
2510     vli_modMult_n(s, s, k); /* s = (e + r*d) / k */
2511 #if (uECC_CURVE == uECC_secp160r1)
2512     if (s[uECC_N_WORDS - 1]) {
2513         return 0;
2514     }
2515 #endif
2516     vli_nativeToBytes(signature + uECC_BYTES, s);
2517     return 1;
2518 }
2519 
2520 int uECC_sign(const uint8_t private_key[uECC_BYTES],
2521               const uint8_t message_hash[uECC_BYTES],
2522               uint8_t signature[uECC_BYTES*2]) {
2523     uECC_word_t k[uECC_N_WORDS];
2524     uECC_word_t tries;
2525 
2526     for (tries = 0; tries < MAX_TRIES; ++tries) {
2527         if(g_rng_function((uint8_t *)k, sizeof(k))) {
2528         #if (uECC_CURVE == uECC_secp160r1)
2529             k[uECC_WORDS] &= 0x01;
2530         #endif
2531             if (uECC_sign_with_k(private_key, message_hash, k, signature)) {
2532                 return 1;
2533             }
2534         }
2535     }
2536     return 0;
2537 }
2538 
2539 /* Compute an HMAC using K as a key (as in RFC 6979). Note that K is always
2540    the same size as the hash result size. */
2541 static void HMAC_init(uECC_HashContext *hash_context, const uint8_t *K) {
2542     uint8_t *pad = hash_context->tmp + 2 * hash_context->result_size;
2543     unsigned i;
2544     for (i = 0; i < hash_context->result_size; ++i)
2545         pad[i] = K[i] ^ 0x36;
2546     for (; i < hash_context->block_size; ++i)
2547         pad[i] = 0x36;
2548 
2549     hash_context->init_hash(hash_context);
2550     hash_context->update_hash(hash_context, pad, hash_context->block_size);
2551 }
2552 
2553 static void HMAC_update(uECC_HashContext *hash_context,
2554                         const uint8_t *message,
2555                         unsigned message_size) {
2556     hash_context->update_hash(hash_context, message, message_size);
2557 }
2558 
2559 static void HMAC_finish(uECC_HashContext *hash_context, const uint8_t *K, uint8_t *result) {
2560     uint8_t *pad = hash_context->tmp + 2 * hash_context->result_size;
2561     unsigned i;
2562     for (i = 0; i < hash_context->result_size; ++i)
2563         pad[i] = K[i] ^ 0x5c;
2564     for (; i < hash_context->block_size; ++i)
2565         pad[i] = 0x5c;
2566 
2567     hash_context->finish_hash(hash_context, result);
2568 
2569     hash_context->init_hash(hash_context);
2570     hash_context->update_hash(hash_context, pad, hash_context->block_size);
2571     hash_context->update_hash(hash_context, result, hash_context->result_size);
2572     hash_context->finish_hash(hash_context, result);
2573 }
2574 
2575 /* V = HMAC_K(V) */
2576 static void update_V(uECC_HashContext *hash_context, uint8_t *K, uint8_t *V) {
2577     HMAC_init(hash_context, K);
2578     HMAC_update(hash_context, V, hash_context->result_size);
2579     HMAC_finish(hash_context, K, V);
2580 }
2581 
2582 /* Deterministic signing, similar to RFC 6979. Differences are:
2583     * We just use (truncated) H(m) directly rather than bits2octets(H(m))
2584       (it is not reduced modulo curve_n).
2585     * We generate a value for k (aka T) directly rather than converting endianness.
2586 
2587    Layout of hash_context->tmp: <K> | <V> | (1 byte overlapped 0x00 or 0x01) / <HMAC pad> */
2588 int uECC_sign_deterministic(const uint8_t private_key[uECC_BYTES],
2589                             const uint8_t message_hash[uECC_BYTES],
2590                             uECC_HashContext *hash_context,
2591                             uint8_t signature[uECC_BYTES*2]) {
2592     uint8_t *K = hash_context->tmp;
2593     uint8_t *V = K + hash_context->result_size;
2594     uECC_word_t tries;
2595     unsigned i;
2596     for (i = 0; i < hash_context->result_size; ++i) {
2597         V[i] = 0x01;
2598         K[i] = 0;
2599     }
2600 
2601     // K = HMAC_K(V || 0x00 || int2octets(x) || h(m))
2602     HMAC_init(hash_context, K);
2603     V[hash_context->result_size] = 0x00;
2604     HMAC_update(hash_context, V, hash_context->result_size + 1);
2605     HMAC_update(hash_context, private_key, uECC_BYTES);
2606     HMAC_update(hash_context, message_hash, uECC_BYTES);
2607     HMAC_finish(hash_context, K, K);
2608 
2609     update_V(hash_context, K, V);
2610 
2611     // K = HMAC_K(V || 0x01 || int2octets(x) || h(m))
2612     HMAC_init(hash_context, K);
2613     V[hash_context->result_size] = 0x01;
2614     HMAC_update(hash_context, V, hash_context->result_size + 1);
2615     HMAC_update(hash_context, private_key, uECC_BYTES);
2616     HMAC_update(hash_context, message_hash, uECC_BYTES);
2617     HMAC_finish(hash_context, K, K);
2618 
2619     update_V(hash_context, K, V);
2620 
2621     for (tries = 0; tries < MAX_TRIES; ++tries) {
2622         uECC_word_t T[uECC_N_WORDS];
2623         uint8_t *T_ptr = (uint8_t *)T;
2624         unsigned T_bytes = 0;
2625         while (T_bytes < sizeof(T)) {
2626             update_V(hash_context, K, V);
2627             for (i = 0; i < hash_context->result_size && T_bytes < sizeof(T); ++i, ++T_bytes) {
2628                 T_ptr[T_bytes] = V[i];
2629             }
2630         }
2631     #if (uECC_CURVE == uECC_secp160r1)
2632         T[uECC_WORDS] &= 0x01;
2633     #endif
2634 
2635         if (uECC_sign_with_k(private_key, message_hash, T, signature)) {
2636             return 1;
2637         }
2638 
2639         // K = HMAC_K(V || 0x00)
2640         HMAC_init(hash_context, K);
2641         V[hash_context->result_size] = 0x00;
2642         HMAC_update(hash_context, V, hash_context->result_size + 1);
2643         HMAC_finish(hash_context, K, K);
2644 
2645         update_V(hash_context, K, V);
2646     }
2647     return 0;
2648 }
2649 
2650 static bitcount_t smax(bitcount_t a, bitcount_t b) {
2651     return (a > b ? a : b);
2652 }
2653 
2654 int uECC_verify(const uint8_t public_key[uECC_BYTES*2],
2655                 const uint8_t hash[uECC_BYTES],
2656                 const uint8_t signature[uECC_BYTES*2]) {
2657     uECC_word_t u1[uECC_N_WORDS], u2[uECC_N_WORDS];
2658     uECC_word_t z[uECC_N_WORDS];
2659     EccPoint public, sum;
2660     uECC_word_t rx[uECC_WORDS];
2661     uECC_word_t ry[uECC_WORDS];
2662     uECC_word_t tx[uECC_WORDS];
2663     uECC_word_t ty[uECC_WORDS];
2664     uECC_word_t tz[uECC_WORDS];
2665     const EccPoint *points[4];
2666     const EccPoint *point;
2667     bitcount_t numBits;
2668     bitcount_t i;
2669     uECC_word_t r[uECC_N_WORDS], s[uECC_N_WORDS];
2670     r[uECC_N_WORDS - 1] = 0;
2671     s[uECC_N_WORDS - 1] = 0;
2672 
2673     vli_bytesToNative(public.x, public_key);
2674     vli_bytesToNative(public.y, public_key + uECC_BYTES);
2675     vli_bytesToNative(r, signature);
2676     vli_bytesToNative(s, signature + uECC_BYTES);
2677 
2678     if (vli_isZero(r) || vli_isZero(s)) { /* r, s must not be 0. */
2679         return 0;
2680     }
2681 
2682 #if (uECC_CURVE != uECC_secp160r1)
2683     if (vli_cmp(curve_n, r) != 1 || vli_cmp(curve_n, s) != 1) { /* r, s must be < n. */
2684         return 0;
2685     }
2686 #endif
2687 
2688     /* Calculate u1 and u2. */
2689     vli_modInv_n(z, s, curve_n); /* Z = s^-1 */
2690     u1[uECC_N_WORDS - 1] = 0;
2691     vli_bytesToNative(u1, hash);
2692     vli_modMult_n(u1, u1, z); /* u1 = e/s */
2693     vli_modMult_n(u2, r, z); /* u2 = r/s */
2694 
2695     /* Calculate sum = G + Q. */
2696     vli_set(sum.x, public.x);
2697     vli_set(sum.y, public.y);
2698     vli_set(tx, curve_G.x);
2699     vli_set(ty, curve_G.y);
2700     vli_modSub_fast(z, sum.x, tx); /* Z = x2 - x1 */
2701     XYcZ_add(tx, ty, sum.x, sum.y);
2702     vli_modInv(z, z, curve_p); /* Z = 1/Z */
2703     apply_z(sum.x, sum.y, z);
2704 
2705     /* Use Shamir's trick to calculate u1*G + u2*Q */
2706     points[0] = 0;
2707     points[1] = &curve_G;
2708     points[2] = &public;
2709     points[3] = &sum;
2710     numBits = smax(vli_numBits(u1, uECC_N_WORDS), vli_numBits(u2, uECC_N_WORDS));
2711 
2712     point = points[(!!vli_testBit(u1, numBits - 1)) | ((!!vli_testBit(u2, numBits - 1)) << 1)];
2713     vli_set(rx, point->x);
2714     vli_set(ry, point->y);
2715     vli_clear(z);
2716     z[0] = 1;
2717 
2718     for (i = numBits - 2; i >= 0; --i) {
2719         uECC_word_t index;
2720         EccPoint_double_jacobian(rx, ry, z);
2721 
2722         index = (!!vli_testBit(u1, i)) | ((!!vli_testBit(u2, i)) << 1);
2723         point = points[index];
2724         if (point) {
2725             vli_set(tx, point->x);
2726             vli_set(ty, point->y);
2727             apply_z(tx, ty, z);
2728             vli_modSub_fast(tz, rx, tx); /* Z = x2 - x1 */
2729             XYcZ_add(tx, ty, rx, ry);
2730             vli_modMult_fast(z, z, tz);
2731         }
2732     }
2733 
2734     vli_modInv(z, z, curve_p); /* Z = 1/Z */
2735     apply_z(rx, ry, z);
2736 
2737     /* v = x1 (mod n) */
2738 #if (uECC_CURVE != uECC_secp160r1)
2739     if (vli_cmp(curve_n, rx) != 1) {
2740         vli_sub(rx, rx, curve_n);
2741     }
2742 #endif
2743 
2744     /* Accept only if v == r. */
2745     return vli_equal(rx, r);
2746 }
2747