Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(74)

Side by Side Diff: crypto/p224.cc

Issue 1539353003: Switch to standard integer types in crypto/. (Closed) Base URL: https://chromium.googlesource.com/chromium/src.git@master
Patch Set: fix Created 5 years ago
Use n/p to move between diff chunks; N/P to move between comments. Draft comments are only viewable by you.
Jump to:
View unified diff | Download patch
« no previous file with comments | « crypto/p224.h ('k') | crypto/p224_spake.h » ('j') | no next file with comments »
Toggle Intra-line Diffs ('i') | Expand Comments ('e') | Collapse Comments ('c') | Show Comments Hide Comments ('s')
OLDNEW
1 // Copyright (c) 2012 The Chromium Authors. All rights reserved. 1 // Copyright (c) 2012 The Chromium Authors. All rights reserved.
2 // Use of this source code is governed by a BSD-style license that can be 2 // Use of this source code is governed by a BSD-style license that can be
3 // found in the LICENSE file. 3 // found in the LICENSE file.
4 4
5 // This is an implementation of the P224 elliptic curve group. It's written to 5 // This is an implementation of the P224 elliptic curve group. It's written to
6 // be short and simple rather than fast, although it's still constant-time. 6 // be short and simple rather than fast, although it's still constant-time.
7 // 7 //
8 // See http://www.imperialviolet.org/2010/12/04/ecc.html ([1]) for background. 8 // See http://www.imperialviolet.org/2010/12/04/ecc.html ([1]) for background.
9 9
10 #include "crypto/p224.h" 10 #include "crypto/p224.h"
11 11
12 #include <stddef.h>
13 #include <stdint.h>
12 #include <string.h> 14 #include <string.h>
13 15
14 #include "base/sys_byteorder.h" 16 #include "base/sys_byteorder.h"
15 17
16 namespace { 18 namespace {
17 19
18 using base::HostToNet32; 20 using base::HostToNet32;
19 using base::NetToHost32; 21 using base::NetToHost32;
20 22
21 // Field element functions. 23 // Field element functions.
22 // 24 //
23 // The field that we're dealing with is ℤ/pℤ where p = 2**224 - 2**96 + 1. 25 // The field that we're dealing with is ℤ/pℤ where p = 2**224 - 2**96 + 1.
24 // 26 //
25 // Field elements are represented by a FieldElement, which is a typedef to an 27 // Field elements are represented by a FieldElement, which is a typedef to an
26 // array of 8 uint32's. The value of a FieldElement, a, is: 28 // array of 8 uint32_t's. The value of a FieldElement, a, is:
27 // a[0] + 2**28·a[1] + 2**56·a[1] + ... + 2**196·a[7] 29 // a[0] + 2**28·a[1] + 2**56·a[1] + ... + 2**196·a[7]
28 // 30 //
29 // Using 28-bit limbs means that there's only 4 bits of headroom, which is less 31 // Using 28-bit limbs means that there's only 4 bits of headroom, which is less
30 // than we would really like. But it has the useful feature that we hit 2**224 32 // than we would really like. But it has the useful feature that we hit 2**224
31 // exactly, making the reflections during a reduce much nicer. 33 // exactly, making the reflections during a reduce much nicer.
32 34
33 using crypto::p224::FieldElement; 35 using crypto::p224::FieldElement;
34 36
35 // kP is the P224 prime. 37 // kP is the P224 prime.
36 const FieldElement kP = { 38 const FieldElement kP = {
37 1, 0, 0, 268431360, 39 1, 0, 0, 268431360,
38 268435455, 268435455, 268435455, 268435455, 40 268435455, 268435455, 268435455, 268435455,
39 }; 41 };
40 42
41 void Contract(FieldElement* inout); 43 void Contract(FieldElement* inout);
42 44
43 // IsZero returns 0xffffffff if a == 0 mod p and 0 otherwise. 45 // IsZero returns 0xffffffff if a == 0 mod p and 0 otherwise.
44 uint32 IsZero(const FieldElement& a) { 46 uint32_t IsZero(const FieldElement& a) {
45 FieldElement minimal; 47 FieldElement minimal;
46 memcpy(&minimal, &a, sizeof(minimal)); 48 memcpy(&minimal, &a, sizeof(minimal));
47 Contract(&minimal); 49 Contract(&minimal);
48 50
49 uint32 is_zero = 0, is_p = 0; 51 uint32_t is_zero = 0, is_p = 0;
50 for (unsigned i = 0; i < 8; i++) { 52 for (unsigned i = 0; i < 8; i++) {
51 is_zero |= minimal[i]; 53 is_zero |= minimal[i];
52 is_p |= minimal[i] - kP[i]; 54 is_p |= minimal[i] - kP[i];
53 } 55 }
54 56
55 // If either is_zero or is_p is 0, then we should return 1. 57 // If either is_zero or is_p is 0, then we should return 1.
56 is_zero |= is_zero >> 16; 58 is_zero |= is_zero >> 16;
57 is_zero |= is_zero >> 8; 59 is_zero |= is_zero >> 8;
58 is_zero |= is_zero >> 4; 60 is_zero |= is_zero >> 4;
59 is_zero |= is_zero >> 2; 61 is_zero |= is_zero >> 2;
60 is_zero |= is_zero >> 1; 62 is_zero |= is_zero >> 1;
61 63
62 is_p |= is_p >> 16; 64 is_p |= is_p >> 16;
63 is_p |= is_p >> 8; 65 is_p |= is_p >> 8;
64 is_p |= is_p >> 4; 66 is_p |= is_p >> 4;
65 is_p |= is_p >> 2; 67 is_p |= is_p >> 2;
66 is_p |= is_p >> 1; 68 is_p |= is_p >> 1;
67 69
68 // For is_zero and is_p, the LSB is 0 iff all the bits are zero. 70 // For is_zero and is_p, the LSB is 0 iff all the bits are zero.
69 is_zero &= is_p & 1; 71 is_zero &= is_p & 1;
70 is_zero = (~is_zero) << 31; 72 is_zero = (~is_zero) << 31;
71 is_zero = static_cast<int32>(is_zero) >> 31; 73 is_zero = static_cast<int32_t>(is_zero) >> 31;
72 return is_zero; 74 return is_zero;
73 } 75 }
74 76
75 // Add computes *out = a+b 77 // Add computes *out = a+b
76 // 78 //
77 // a[i] + b[i] < 2**32 79 // a[i] + b[i] < 2**32
78 void Add(FieldElement* out, const FieldElement& a, const FieldElement& b) { 80 void Add(FieldElement* out, const FieldElement& a, const FieldElement& b) {
79 for (int i = 0; i < 8; i++) { 81 for (int i = 0; i < 8; i++) {
80 (*out)[i] = a[i] + b[i]; 82 (*out)[i] = a[i] + b[i];
81 } 83 }
82 } 84 }
83 85
84 static const uint32 kTwo31p3 = (1u<<31) + (1u<<3); 86 static const uint32_t kTwo31p3 = (1u << 31) + (1u << 3);
85 static const uint32 kTwo31m3 = (1u<<31) - (1u<<3); 87 static const uint32_t kTwo31m3 = (1u << 31) - (1u << 3);
86 static const uint32 kTwo31m15m3 = (1u<<31) - (1u<<15) - (1u<<3); 88 static const uint32_t kTwo31m15m3 = (1u << 31) - (1u << 15) - (1u << 3);
87 // kZero31ModP is 0 mod p where bit 31 is set in all limbs so that we can 89 // kZero31ModP is 0 mod p where bit 31 is set in all limbs so that we can
88 // subtract smaller amounts without underflow. See the section "Subtraction" in 90 // subtract smaller amounts without underflow. See the section "Subtraction" in
89 // [1] for why. 91 // [1] for why.
90 static const FieldElement kZero31ModP = { 92 static const FieldElement kZero31ModP = {
91 kTwo31p3, kTwo31m3, kTwo31m3, kTwo31m15m3, 93 kTwo31p3, kTwo31m3, kTwo31m3, kTwo31m15m3,
92 kTwo31m3, kTwo31m3, kTwo31m3, kTwo31m3 94 kTwo31m3, kTwo31m3, kTwo31m3, kTwo31m3
93 }; 95 };
94 96
95 // Subtract computes *out = a-b 97 // Subtract computes *out = a-b
96 // 98 //
97 // a[i], b[i] < 2**30 99 // a[i], b[i] < 2**30
98 // out[i] < 2**32 100 // out[i] < 2**32
99 void Subtract(FieldElement* out, const FieldElement& a, const FieldElement& b) { 101 void Subtract(FieldElement* out, const FieldElement& a, const FieldElement& b) {
100 for (int i = 0; i < 8; i++) { 102 for (int i = 0; i < 8; i++) {
101 // See the section on "Subtraction" in [1] for details. 103 // See the section on "Subtraction" in [1] for details.
102 (*out)[i] = a[i] + kZero31ModP[i] - b[i]; 104 (*out)[i] = a[i] + kZero31ModP[i] - b[i];
103 } 105 }
104 } 106 }
105 107
106 static const uint64 kTwo63p35 = (1ull<<63) + (1ull<<35); 108 static const uint64_t kTwo63p35 = (1ull << 63) + (1ull << 35);
107 static const uint64 kTwo63m35 = (1ull<<63) - (1ull<<35); 109 static const uint64_t kTwo63m35 = (1ull << 63) - (1ull << 35);
108 static const uint64 kTwo63m35m19 = (1ull<<63) - (1ull<<35) - (1ull<<19); 110 static const uint64_t kTwo63m35m19 = (1ull << 63) - (1ull << 35) - (1ull << 19);
109 // kZero63ModP is 0 mod p where bit 63 is set in all limbs. See the section 111 // kZero63ModP is 0 mod p where bit 63 is set in all limbs. See the section
110 // "Subtraction" in [1] for why. 112 // "Subtraction" in [1] for why.
111 static const uint64 kZero63ModP[8] = { 113 static const uint64_t kZero63ModP[8] = {
112 kTwo63p35, kTwo63m35, kTwo63m35, kTwo63m35, 114 kTwo63p35, kTwo63m35, kTwo63m35, kTwo63m35,
113 kTwo63m35m19, kTwo63m35, kTwo63m35, kTwo63m35, 115 kTwo63m35m19, kTwo63m35, kTwo63m35, kTwo63m35,
114 }; 116 };
115 117
116 static const uint32 kBottom28Bits = 0xfffffff; 118 static const uint32_t kBottom28Bits = 0xfffffff;
117 119
118 // LargeFieldElement also represents an element of the field. The limbs are 120 // LargeFieldElement also represents an element of the field. The limbs are
119 // still spaced 28-bits apart and in little-endian order. So the limbs are at 121 // still spaced 28-bits apart and in little-endian order. So the limbs are at
120 // 0, 28, 56, ..., 392 bits, each 64-bits wide. 122 // 0, 28, 56, ..., 392 bits, each 64-bits wide.
121 typedef uint64 LargeFieldElement[15]; 123 typedef uint64_t LargeFieldElement[15];
122 124
123 // ReduceLarge converts a LargeFieldElement to a FieldElement. 125 // ReduceLarge converts a LargeFieldElement to a FieldElement.
124 // 126 //
125 // in[i] < 2**62 127 // in[i] < 2**62
126 void ReduceLarge(FieldElement* out, LargeFieldElement* inptr) { 128 void ReduceLarge(FieldElement* out, LargeFieldElement* inptr) {
127 LargeFieldElement& in(*inptr); 129 LargeFieldElement& in(*inptr);
128 130
129 for (int i = 0; i < 8; i++) { 131 for (int i = 0; i < 8; i++) {
130 in[i] += kZero63ModP[i]; 132 in[i] += kZero63ModP[i];
131 } 133 }
132 134
133 // Eliminate the coefficients at 2**224 and greater while maintaining the 135 // Eliminate the coefficients at 2**224 and greater while maintaining the
134 // same value mod p. 136 // same value mod p.
135 for (int i = 14; i >= 8; i--) { 137 for (int i = 14; i >= 8; i--) {
136 in[i-8] -= in[i]; // reflection off the "+1" term of p. 138 in[i-8] -= in[i]; // reflection off the "+1" term of p.
137 in[i-5] += (in[i] & 0xffff) << 12; // part of the "-2**96" reflection. 139 in[i-5] += (in[i] & 0xffff) << 12; // part of the "-2**96" reflection.
138 in[i-4] += in[i] >> 16; // the rest of the "-2**96" reflection. 140 in[i-4] += in[i] >> 16; // the rest of the "-2**96" reflection.
139 } 141 }
140 in[8] = 0; 142 in[8] = 0;
141 // in[0..8] < 2**64 143 // in[0..8] < 2**64
142 144
143 // As the values become small enough, we start to store them in |out| and use 145 // As the values become small enough, we start to store them in |out| and use
144 // 32-bit operations. 146 // 32-bit operations.
145 for (int i = 1; i < 8; i++) { 147 for (int i = 1; i < 8; i++) {
146 in[i+1] += in[i] >> 28; 148 in[i+1] += in[i] >> 28;
147 (*out)[i] = static_cast<uint32>(in[i] & kBottom28Bits); 149 (*out)[i] = static_cast<uint32_t>(in[i] & kBottom28Bits);
148 } 150 }
149 // Eliminate the term at 2*224 that we introduced while keeping the same 151 // Eliminate the term at 2*224 that we introduced while keeping the same
150 // value mod p. 152 // value mod p.
151 in[0] -= in[8]; // reflection off the "+1" term of p. 153 in[0] -= in[8]; // reflection off the "+1" term of p.
152 (*out)[3] += static_cast<uint32>(in[8] & 0xffff) << 12; // "-2**96" term 154 (*out)[3] += static_cast<uint32_t>(in[8] & 0xffff) << 12; // "-2**96" term
153 (*out)[4] += static_cast<uint32>(in[8] >> 16); // rest of "-2**96" term 155 (*out)[4] += static_cast<uint32_t>(in[8] >> 16); // rest of "-2**96" term
154 // in[0] < 2**64 156 // in[0] < 2**64
155 // out[3] < 2**29 157 // out[3] < 2**29
156 // out[4] < 2**29 158 // out[4] < 2**29
157 // out[1,2,5..7] < 2**28 159 // out[1,2,5..7] < 2**28
158 160
159 (*out)[0] = static_cast<uint32>(in[0] & kBottom28Bits); 161 (*out)[0] = static_cast<uint32_t>(in[0] & kBottom28Bits);
160 (*out)[1] += static_cast<uint32>((in[0] >> 28) & kBottom28Bits); 162 (*out)[1] += static_cast<uint32_t>((in[0] >> 28) & kBottom28Bits);
161 (*out)[2] += static_cast<uint32>(in[0] >> 56); 163 (*out)[2] += static_cast<uint32_t>(in[0] >> 56);
162 // out[0] < 2**28 164 // out[0] < 2**28
163 // out[1..4] < 2**29 165 // out[1..4] < 2**29
164 // out[5..7] < 2**28 166 // out[5..7] < 2**28
165 } 167 }
166 168
167 // Mul computes *out = a*b 169 // Mul computes *out = a*b
168 // 170 //
169 // a[i] < 2**29, b[i] < 2**30 (or vice versa) 171 // a[i] < 2**29, b[i] < 2**30 (or vice versa)
170 // out[i] < 2**29 172 // out[i] < 2**29
171 void Mul(FieldElement* out, const FieldElement& a, const FieldElement& b) { 173 void Mul(FieldElement* out, const FieldElement& a, const FieldElement& b) {
172 LargeFieldElement tmp; 174 LargeFieldElement tmp;
173 memset(&tmp, 0, sizeof(tmp)); 175 memset(&tmp, 0, sizeof(tmp));
174 176
175 for (int i = 0; i < 8; i++) { 177 for (int i = 0; i < 8; i++) {
176 for (int j = 0; j < 8; j++) { 178 for (int j = 0; j < 8; j++) {
177 tmp[i+j] += static_cast<uint64>(a[i]) * static_cast<uint64>(b[j]); 179 tmp[i + j] += static_cast<uint64_t>(a[i]) * static_cast<uint64_t>(b[j]);
178 } 180 }
179 } 181 }
180 182
181 ReduceLarge(out, &tmp); 183 ReduceLarge(out, &tmp);
182 } 184 }
183 185
184 // Square computes *out = a*a 186 // Square computes *out = a*a
185 // 187 //
186 // a[i] < 2**29 188 // a[i] < 2**29
187 // out[i] < 2**29 189 // out[i] < 2**29
188 void Square(FieldElement* out, const FieldElement& a) { 190 void Square(FieldElement* out, const FieldElement& a) {
189 LargeFieldElement tmp; 191 LargeFieldElement tmp;
190 memset(&tmp, 0, sizeof(tmp)); 192 memset(&tmp, 0, sizeof(tmp));
191 193
192 for (int i = 0; i < 8; i++) { 194 for (int i = 0; i < 8; i++) {
193 for (int j = 0; j <= i; j++) { 195 for (int j = 0; j <= i; j++) {
194 uint64 r = static_cast<uint64>(a[i]) * static_cast<uint64>(a[j]); 196 uint64_t r = static_cast<uint64_t>(a[i]) * static_cast<uint64_t>(a[j]);
195 if (i == j) { 197 if (i == j) {
196 tmp[i+j] += r; 198 tmp[i+j] += r;
197 } else { 199 } else {
198 tmp[i+j] += r << 1; 200 tmp[i+j] += r << 1;
199 } 201 }
200 } 202 }
201 } 203 }
202 204
203 ReduceLarge(out, &tmp); 205 ReduceLarge(out, &tmp);
204 } 206 }
205 207
206 // Reduce reduces the coefficients of in_out to smaller bounds. 208 // Reduce reduces the coefficients of in_out to smaller bounds.
207 // 209 //
208 // On entry: a[i] < 2**31 + 2**30 210 // On entry: a[i] < 2**31 + 2**30
209 // On exit: a[i] < 2**29 211 // On exit: a[i] < 2**29
210 void Reduce(FieldElement* in_out) { 212 void Reduce(FieldElement* in_out) {
211 FieldElement& a = *in_out; 213 FieldElement& a = *in_out;
212 214
213 for (int i = 0; i < 7; i++) { 215 for (int i = 0; i < 7; i++) {
214 a[i+1] += a[i] >> 28; 216 a[i+1] += a[i] >> 28;
215 a[i] &= kBottom28Bits; 217 a[i] &= kBottom28Bits;
216 } 218 }
217 uint32 top = a[7] >> 28; 219 uint32_t top = a[7] >> 28;
218 a[7] &= kBottom28Bits; 220 a[7] &= kBottom28Bits;
219 221
220 // top < 2**4 222 // top < 2**4
221 // Constant-time: mask = (top != 0) ? 0xffffffff : 0 223 // Constant-time: mask = (top != 0) ? 0xffffffff : 0
222 uint32 mask = top; 224 uint32_t mask = top;
223 mask |= mask >> 2; 225 mask |= mask >> 2;
224 mask |= mask >> 1; 226 mask |= mask >> 1;
225 mask <<= 31; 227 mask <<= 31;
226 mask = static_cast<uint32>(static_cast<int32>(mask) >> 31); 228 mask = static_cast<uint32_t>(static_cast<int32_t>(mask) >> 31);
227 229
228 // Eliminate top while maintaining the same value mod p. 230 // Eliminate top while maintaining the same value mod p.
229 a[0] -= top; 231 a[0] -= top;
230 a[3] += top << 12; 232 a[3] += top << 12;
231 233
232 // We may have just made a[0] negative but, if we did, then we must 234 // We may have just made a[0] negative but, if we did, then we must
233 // have added something to a[3], thus it's > 2**12. Therefore we can 235 // have added something to a[3], thus it's > 2**12. Therefore we can
234 // carry down to a[0]. 236 // carry down to a[0].
235 a[3] -= 1 & mask; 237 a[3] -= 1 & mask;
236 a[2] += mask & ((1<<28) - 1); 238 a[2] += mask & ((1<<28) - 1);
(...skipping 56 matching lines...) Expand 10 before | Expand all | Expand 10 after
293 // On entry, in[i] < 2**29 295 // On entry, in[i] < 2**29
294 // On exit, in[i] < 2**28 296 // On exit, in[i] < 2**28
295 void Contract(FieldElement* inout) { 297 void Contract(FieldElement* inout) {
296 FieldElement& out = *inout; 298 FieldElement& out = *inout;
297 299
298 // Reduce the coefficients to < 2**28. 300 // Reduce the coefficients to < 2**28.
299 for (int i = 0; i < 7; i++) { 301 for (int i = 0; i < 7; i++) {
300 out[i+1] += out[i] >> 28; 302 out[i+1] += out[i] >> 28;
301 out[i] &= kBottom28Bits; 303 out[i] &= kBottom28Bits;
302 } 304 }
303 uint32 top = out[7] >> 28; 305 uint32_t top = out[7] >> 28;
304 out[7] &= kBottom28Bits; 306 out[7] &= kBottom28Bits;
305 307
306 // Eliminate top while maintaining the same value mod p. 308 // Eliminate top while maintaining the same value mod p.
307 out[0] -= top; 309 out[0] -= top;
308 out[3] += top << 12; 310 out[3] += top << 12;
309 311
310 // We may just have made out[0] negative. So we carry down. If we made 312 // We may just have made out[0] negative. So we carry down. If we made
311 // out[0] negative then we know that out[3] is sufficiently positive 313 // out[0] negative then we know that out[3] is sufficiently positive
312 // because we just added to it. 314 // because we just added to it.
313 for (int i = 0; i < 3; i++) { 315 for (int i = 0; i < 3; i++) {
314 uint32 mask = static_cast<uint32>(static_cast<int32>(out[i]) >> 31); 316 uint32_t mask = static_cast<uint32_t>(static_cast<int32_t>(out[i]) >> 31);
315 out[i] += (1 << 28) & mask; 317 out[i] += (1 << 28) & mask;
316 out[i+1] -= 1 & mask; 318 out[i+1] -= 1 & mask;
317 } 319 }
318 320
319 // We might have pushed out[3] over 2**28 so we perform another, partial 321 // We might have pushed out[3] over 2**28 so we perform another, partial
320 // carry chain. 322 // carry chain.
321 for (int i = 3; i < 7; i++) { 323 for (int i = 3; i < 7; i++) {
322 out[i+1] += out[i] >> 28; 324 out[i+1] += out[i] >> 28;
323 out[i] &= kBottom28Bits; 325 out[i] &= kBottom28Bits;
324 } 326 }
(...skipping 12 matching lines...) Expand all
337 // The first value of top was in [0..16), therefore, prior to eliminating 339 // The first value of top was in [0..16), therefore, prior to eliminating
338 // the first top, 0xfff1000 <= out[3] <= 0xfffffff. Therefore, after 340 // the first top, 0xfff1000 <= out[3] <= 0xfffffff. Therefore, after
339 // overflowing and being reduced by the second carry chain, out[3] <= 341 // overflowing and being reduced by the second carry chain, out[3] <=
340 // 0xf000. Thus it cannot have overflowed when we eliminated top for the 342 // 0xf000. Thus it cannot have overflowed when we eliminated top for the
341 // second time. 343 // second time.
342 344
343 // Again, we may just have made out[0] negative, so do the same carry down. 345 // Again, we may just have made out[0] negative, so do the same carry down.
344 // As before, if we made out[0] negative then we know that out[3] is 346 // As before, if we made out[0] negative then we know that out[3] is
345 // sufficiently positive. 347 // sufficiently positive.
346 for (int i = 0; i < 3; i++) { 348 for (int i = 0; i < 3; i++) {
347 uint32 mask = static_cast<uint32>(static_cast<int32>(out[i]) >> 31); 349 uint32_t mask = static_cast<uint32_t>(static_cast<int32_t>(out[i]) >> 31);
348 out[i] += (1 << 28) & mask; 350 out[i] += (1 << 28) & mask;
349 out[i+1] -= 1 & mask; 351 out[i+1] -= 1 & mask;
350 } 352 }
351 353
352 // The value is < 2**224, but maybe greater than p. In order to reduce to a 354 // The value is < 2**224, but maybe greater than p. In order to reduce to a
353 // unique, minimal value we see if the value is >= p and, if so, subtract p. 355 // unique, minimal value we see if the value is >= p and, if so, subtract p.
354 356
355 // First we build a mask from the top four limbs, which must all be 357 // First we build a mask from the top four limbs, which must all be
356 // equal to bottom28Bits if the whole value is >= p. If top_4_all_ones 358 // equal to bottom28Bits if the whole value is >= p. If top_4_all_ones
357 // ends up with any zero bits in the bottom 28 bits, then this wasn't 359 // ends up with any zero bits in the bottom 28 bits, then this wasn't
358 // true. 360 // true.
359 uint32 top_4_all_ones = 0xffffffffu; 361 uint32_t top_4_all_ones = 0xffffffffu;
360 for (int i = 4; i < 8; i++) { 362 for (int i = 4; i < 8; i++) {
361 top_4_all_ones &= out[i]; 363 top_4_all_ones &= out[i];
362 } 364 }
363 top_4_all_ones |= 0xf0000000; 365 top_4_all_ones |= 0xf0000000;
364 // Now we replicate any zero bits to all the bits in top_4_all_ones. 366 // Now we replicate any zero bits to all the bits in top_4_all_ones.
365 top_4_all_ones &= top_4_all_ones >> 16; 367 top_4_all_ones &= top_4_all_ones >> 16;
366 top_4_all_ones &= top_4_all_ones >> 8; 368 top_4_all_ones &= top_4_all_ones >> 8;
367 top_4_all_ones &= top_4_all_ones >> 4; 369 top_4_all_ones &= top_4_all_ones >> 4;
368 top_4_all_ones &= top_4_all_ones >> 2; 370 top_4_all_ones &= top_4_all_ones >> 2;
369 top_4_all_ones &= top_4_all_ones >> 1; 371 top_4_all_ones &= top_4_all_ones >> 1;
370 top_4_all_ones = 372 top_4_all_ones =
371 static_cast<uint32>(static_cast<int32>(top_4_all_ones << 31) >> 31); 373 static_cast<uint32_t>(static_cast<int32_t>(top_4_all_ones << 31) >> 31);
372 374
373 // Now we test whether the bottom three limbs are non-zero. 375 // Now we test whether the bottom three limbs are non-zero.
374 uint32 bottom_3_non_zero = out[0] | out[1] | out[2]; 376 uint32_t bottom_3_non_zero = out[0] | out[1] | out[2];
375 bottom_3_non_zero |= bottom_3_non_zero >> 16; 377 bottom_3_non_zero |= bottom_3_non_zero >> 16;
376 bottom_3_non_zero |= bottom_3_non_zero >> 8; 378 bottom_3_non_zero |= bottom_3_non_zero >> 8;
377 bottom_3_non_zero |= bottom_3_non_zero >> 4; 379 bottom_3_non_zero |= bottom_3_non_zero >> 4;
378 bottom_3_non_zero |= bottom_3_non_zero >> 2; 380 bottom_3_non_zero |= bottom_3_non_zero >> 2;
379 bottom_3_non_zero |= bottom_3_non_zero >> 1; 381 bottom_3_non_zero |= bottom_3_non_zero >> 1;
380 bottom_3_non_zero = 382 bottom_3_non_zero =
381 static_cast<uint32>(static_cast<int32>(bottom_3_non_zero) >> 31); 383 static_cast<uint32_t>(static_cast<int32_t>(bottom_3_non_zero) >> 31);
382 384
383 // Everything depends on the value of out[3]. 385 // Everything depends on the value of out[3].
384 // If it's > 0xffff000 and top_4_all_ones != 0 then the whole value is >= p 386 // If it's > 0xffff000 and top_4_all_ones != 0 then the whole value is >= p
385 // If it's = 0xffff000 and top_4_all_ones != 0 and bottom_3_non_zero != 0, 387 // If it's = 0xffff000 and top_4_all_ones != 0 and bottom_3_non_zero != 0,
386 // then the whole value is >= p 388 // then the whole value is >= p
387 // If it's < 0xffff000, then the whole value is < p 389 // If it's < 0xffff000, then the whole value is < p
388 uint32 n = out[3] - 0xffff000; 390 uint32_t n = out[3] - 0xffff000;
389 uint32 out_3_equal = n; 391 uint32_t out_3_equal = n;
390 out_3_equal |= out_3_equal >> 16; 392 out_3_equal |= out_3_equal >> 16;
391 out_3_equal |= out_3_equal >> 8; 393 out_3_equal |= out_3_equal >> 8;
392 out_3_equal |= out_3_equal >> 4; 394 out_3_equal |= out_3_equal >> 4;
393 out_3_equal |= out_3_equal >> 2; 395 out_3_equal |= out_3_equal >> 2;
394 out_3_equal |= out_3_equal >> 1; 396 out_3_equal |= out_3_equal >> 1;
395 out_3_equal = 397 out_3_equal =
396 ~static_cast<uint32>(static_cast<int32>(out_3_equal << 31) >> 31); 398 ~static_cast<uint32_t>(static_cast<int32_t>(out_3_equal << 31) >> 31);
397 399
398 // If out[3] > 0xffff000 then n's MSB will be zero. 400 // If out[3] > 0xffff000 then n's MSB will be zero.
399 uint32 out_3_gt = ~static_cast<uint32>(static_cast<int32>(n << 31) >> 31); 401 uint32_t out_3_gt =
402 ~static_cast<uint32_t>(static_cast<int32_t>(n << 31) >> 31);
400 403
401 uint32 mask = top_4_all_ones & ((out_3_equal & bottom_3_non_zero) | out_3_gt); 404 uint32_t mask =
405 top_4_all_ones & ((out_3_equal & bottom_3_non_zero) | out_3_gt);
402 out[0] -= 1 & mask; 406 out[0] -= 1 & mask;
403 out[3] -= 0xffff000 & mask; 407 out[3] -= 0xffff000 & mask;
404 out[4] -= 0xfffffff & mask; 408 out[4] -= 0xfffffff & mask;
405 out[5] -= 0xfffffff & mask; 409 out[5] -= 0xfffffff & mask;
406 out[6] -= 0xfffffff & mask; 410 out[6] -= 0xfffffff & mask;
407 out[7] -= 0xfffffff & mask; 411 out[7] -= 0xfffffff & mask;
408 } 412 }
409 413
410 414
411 // Group element functions. 415 // Group element functions.
412 // 416 //
413 // These functions deal with group elements. The group is an elliptic curve 417 // These functions deal with group elements. The group is an elliptic curve
414 // group with a = -3 defined in FIPS 186-3, section D.2.2. 418 // group with a = -3 defined in FIPS 186-3, section D.2.2.
415 419
416 using crypto::p224::Point; 420 using crypto::p224::Point;
417 421
418 // kB is parameter of the elliptic curve. 422 // kB is parameter of the elliptic curve.
419 const FieldElement kB = { 423 const FieldElement kB = {
420 55967668, 11768882, 265861671, 185302395, 424 55967668, 11768882, 265861671, 185302395,
421 39211076, 180311059, 84673715, 188764328, 425 39211076, 180311059, 84673715, 188764328,
422 }; 426 };
423 427
424 void CopyConditional(Point* out, const Point& a, uint32 mask); 428 void CopyConditional(Point* out, const Point& a, uint32_t mask);
425 void DoubleJacobian(Point* out, const Point& a); 429 void DoubleJacobian(Point* out, const Point& a);
426 430
427 // AddJacobian computes *out = a+b where a != b. 431 // AddJacobian computes *out = a+b where a != b.
428 void AddJacobian(Point *out, 432 void AddJacobian(Point *out,
429 const Point& a, 433 const Point& a,
430 const Point& b) { 434 const Point& b) {
431 // See http://hyperelliptic.org/EFD/g1p/auto-shortw-jacobian-3.html#addition-a dd-2007-bl 435 // See http://hyperelliptic.org/EFD/g1p/auto-shortw-jacobian-3.html#addition-a dd-2007-bl
432 FieldElement z1z1, z2z2, u1, u2, s1, s2, h, i, j, r, v; 436 FieldElement z1z1, z2z2, u1, u2, s1, s2, h, i, j, r, v;
433 437
434 uint32 z1_is_zero = IsZero(a.z); 438 uint32_t z1_is_zero = IsZero(a.z);
435 uint32 z2_is_zero = IsZero(b.z); 439 uint32_t z2_is_zero = IsZero(b.z);
436 440
437 // Z1Z1 = Z1² 441 // Z1Z1 = Z1²
438 Square(&z1z1, a.z); 442 Square(&z1z1, a.z);
439 443
440 // Z2Z2 = Z2² 444 // Z2Z2 = Z2²
441 Square(&z2z2, b.z); 445 Square(&z2z2, b.z);
442 446
443 // U1 = X1*Z2Z2 447 // U1 = X1*Z2Z2
444 Mul(&u1, a.x, z2z2); 448 Mul(&u1, a.x, z2z2);
445 449
446 // U2 = X2*Z1Z1 450 // U2 = X2*Z1Z1
447 Mul(&u2, b.x, z1z1); 451 Mul(&u2, b.x, z1z1);
448 452
449 // S1 = Y1*Z2*Z2Z2 453 // S1 = Y1*Z2*Z2Z2
450 Mul(&s1, b.z, z2z2); 454 Mul(&s1, b.z, z2z2);
451 Mul(&s1, a.y, s1); 455 Mul(&s1, a.y, s1);
452 456
453 // S2 = Y2*Z1*Z1Z1 457 // S2 = Y2*Z1*Z1Z1
454 Mul(&s2, a.z, z1z1); 458 Mul(&s2, a.z, z1z1);
455 Mul(&s2, b.y, s2); 459 Mul(&s2, b.y, s2);
456 460
457 // H = U2-U1 461 // H = U2-U1
458 Subtract(&h, u2, u1); 462 Subtract(&h, u2, u1);
459 Reduce(&h); 463 Reduce(&h);
460 uint32 x_equal = IsZero(h); 464 uint32_t x_equal = IsZero(h);
461 465
462 // I = (2*H)² 466 // I = (2*H)²
463 for (int k = 0; k < 8; k++) { 467 for (int k = 0; k < 8; k++) {
464 i[k] = h[k] << 1; 468 i[k] = h[k] << 1;
465 } 469 }
466 Reduce(&i); 470 Reduce(&i);
467 Square(&i, i); 471 Square(&i, i);
468 472
469 // J = H*I 473 // J = H*I
470 Mul(&j, h, i); 474 Mul(&j, h, i);
471 // r = 2*(S2-S1) 475 // r = 2*(S2-S1)
472 Subtract(&r, s2, s1); 476 Subtract(&r, s2, s1);
473 Reduce(&r); 477 Reduce(&r);
474 uint32 y_equal = IsZero(r); 478 uint32_t y_equal = IsZero(r);
475 479
476 if (x_equal && y_equal && !z1_is_zero && !z2_is_zero) { 480 if (x_equal && y_equal && !z1_is_zero && !z2_is_zero) {
477 // The two input points are the same therefore we must use the dedicated 481 // The two input points are the same therefore we must use the dedicated
478 // doubling function as the slope of the line is undefined. 482 // doubling function as the slope of the line is undefined.
479 DoubleJacobian(out, a); 483 DoubleJacobian(out, a);
480 return; 484 return;
481 } 485 }
482 486
483 for (int k = 0; k < 8; k++) { 487 for (int k = 0; k < 8; k++) {
484 r[k] <<= 1; 488 r[k] <<= 1;
(...skipping 86 matching lines...) Expand 10 before | Expand all | Expand 10 after
571 gamma[i] <<= 3; 575 gamma[i] <<= 3;
572 } 576 }
573 Reduce(&gamma); 577 Reduce(&gamma);
574 Mul(&out->y, alpha, beta); 578 Mul(&out->y, alpha, beta);
575 Subtract(&out->y, out->y, gamma); 579 Subtract(&out->y, out->y, gamma);
576 Reduce(&out->y); 580 Reduce(&out->y);
577 } 581 }
578 582
579 // CopyConditional sets *out=a if mask is 0xffffffff. mask must be either 0 of 583 // CopyConditional sets *out=a if mask is 0xffffffff. mask must be either 0 of
580 // 0xffffffff. 584 // 0xffffffff.
581 void CopyConditional(Point* out, 585 void CopyConditional(Point* out, const Point& a, uint32_t mask) {
582 const Point& a,
583 uint32 mask) {
584 for (int i = 0; i < 8; i++) { 586 for (int i = 0; i < 8; i++) {
585 out->x[i] ^= mask & (a.x[i] ^ out->x[i]); 587 out->x[i] ^= mask & (a.x[i] ^ out->x[i]);
586 out->y[i] ^= mask & (a.y[i] ^ out->y[i]); 588 out->y[i] ^= mask & (a.y[i] ^ out->y[i]);
587 out->z[i] ^= mask & (a.z[i] ^ out->z[i]); 589 out->z[i] ^= mask & (a.z[i] ^ out->z[i]);
588 } 590 }
589 } 591 }
590 592
591 // ScalarMult calculates *out = a*scalar where scalar is a big-endian number of 593 // ScalarMult calculates *out = a*scalar where scalar is a big-endian number of
592 // length scalar_len and != 0. 594 // length scalar_len and != 0.
593 void ScalarMult(Point* out, const Point& a, 595 void ScalarMult(Point* out,
594 const uint8* scalar, size_t scalar_len) { 596 const Point& a,
597 const uint8_t* scalar,
598 size_t scalar_len) {
595 memset(out, 0, sizeof(*out)); 599 memset(out, 0, sizeof(*out));
596 Point tmp; 600 Point tmp;
597 601
598 for (size_t i = 0; i < scalar_len; i++) { 602 for (size_t i = 0; i < scalar_len; i++) {
599 for (unsigned int bit_num = 0; bit_num < 8; bit_num++) { 603 for (unsigned int bit_num = 0; bit_num < 8; bit_num++) {
600 DoubleJacobian(out, *out); 604 DoubleJacobian(out, *out);
601 uint32 bit = static_cast<uint32>(static_cast<int32>( 605 uint32_t bit = static_cast<uint32_t>(static_cast<int32_t>(
602 (((scalar[i] >> (7 - bit_num)) & 1) << 31) >> 31)); 606 (((scalar[i] >> (7 - bit_num)) & 1) << 31) >> 31));
603 AddJacobian(&tmp, a, *out); 607 AddJacobian(&tmp, a, *out);
604 CopyConditional(out, tmp, bit); 608 CopyConditional(out, tmp, bit);
605 } 609 }
606 } 610 }
607 } 611 }
608 612
609 // Get224Bits reads 7 words from in and scatters their contents in 613 // Get224Bits reads 7 words from in and scatters their contents in
610 // little-endian form into 8 words at out, 28 bits per output word. 614 // little-endian form into 8 words at out, 28 bits per output word.
611 void Get224Bits(uint32* out, const uint32* in) { 615 void Get224Bits(uint32_t* out, const uint32_t* in) {
612 out[0] = NetToHost32(in[6]) & kBottom28Bits; 616 out[0] = NetToHost32(in[6]) & kBottom28Bits;
613 out[1] = ((NetToHost32(in[5]) << 4) | 617 out[1] = ((NetToHost32(in[5]) << 4) |
614 (NetToHost32(in[6]) >> 28)) & kBottom28Bits; 618 (NetToHost32(in[6]) >> 28)) & kBottom28Bits;
615 out[2] = ((NetToHost32(in[4]) << 8) | 619 out[2] = ((NetToHost32(in[4]) << 8) |
616 (NetToHost32(in[5]) >> 24)) & kBottom28Bits; 620 (NetToHost32(in[5]) >> 24)) & kBottom28Bits;
617 out[3] = ((NetToHost32(in[3]) << 12) | 621 out[3] = ((NetToHost32(in[3]) << 12) |
618 (NetToHost32(in[4]) >> 20)) & kBottom28Bits; 622 (NetToHost32(in[4]) >> 20)) & kBottom28Bits;
619 out[4] = ((NetToHost32(in[2]) << 16) | 623 out[4] = ((NetToHost32(in[2]) << 16) |
620 (NetToHost32(in[3]) >> 16)) & kBottom28Bits; 624 (NetToHost32(in[3]) >> 16)) & kBottom28Bits;
621 out[5] = ((NetToHost32(in[1]) << 20) | 625 out[5] = ((NetToHost32(in[1]) << 20) |
622 (NetToHost32(in[2]) >> 12)) & kBottom28Bits; 626 (NetToHost32(in[2]) >> 12)) & kBottom28Bits;
623 out[6] = ((NetToHost32(in[0]) << 24) | 627 out[6] = ((NetToHost32(in[0]) << 24) |
624 (NetToHost32(in[1]) >> 8)) & kBottom28Bits; 628 (NetToHost32(in[1]) >> 8)) & kBottom28Bits;
625 out[7] = (NetToHost32(in[0]) >> 4) & kBottom28Bits; 629 out[7] = (NetToHost32(in[0]) >> 4) & kBottom28Bits;
626 } 630 }
627 631
628 // Put224Bits performs the inverse operation to Get224Bits: taking 28 bits from 632 // Put224Bits performs the inverse operation to Get224Bits: taking 28 bits from
629 // each of 8 input words and writing them in big-endian order to 7 words at 633 // each of 8 input words and writing them in big-endian order to 7 words at
630 // out. 634 // out.
631 void Put224Bits(uint32* out, const uint32* in) { 635 void Put224Bits(uint32_t* out, const uint32_t* in) {
632 out[6] = HostToNet32((in[0] >> 0) | (in[1] << 28)); 636 out[6] = HostToNet32((in[0] >> 0) | (in[1] << 28));
633 out[5] = HostToNet32((in[1] >> 4) | (in[2] << 24)); 637 out[5] = HostToNet32((in[1] >> 4) | (in[2] << 24));
634 out[4] = HostToNet32((in[2] >> 8) | (in[3] << 20)); 638 out[4] = HostToNet32((in[2] >> 8) | (in[3] << 20));
635 out[3] = HostToNet32((in[3] >> 12) | (in[4] << 16)); 639 out[3] = HostToNet32((in[3] >> 12) | (in[4] << 16));
636 out[2] = HostToNet32((in[4] >> 16) | (in[5] << 12)); 640 out[2] = HostToNet32((in[4] >> 16) | (in[5] << 12));
637 out[1] = HostToNet32((in[5] >> 20) | (in[6] << 8)); 641 out[1] = HostToNet32((in[5] >> 20) | (in[6] << 8));
638 out[0] = HostToNet32((in[6] >> 24) | (in[7] << 4)); 642 out[0] = HostToNet32((in[6] >> 24) | (in[7] << 4));
639 } 643 }
640 644
641 } // anonymous namespace 645 } // anonymous namespace
642 646
643 namespace crypto { 647 namespace crypto {
644 648
645 namespace p224 { 649 namespace p224 {
646 650
647 bool Point::SetFromString(const base::StringPiece& in) { 651 bool Point::SetFromString(const base::StringPiece& in) {
648 if (in.size() != 2*28) 652 if (in.size() != 2*28)
649 return false; 653 return false;
650 const uint32* inwords = reinterpret_cast<const uint32*>(in.data()); 654 const uint32_t* inwords = reinterpret_cast<const uint32_t*>(in.data());
651 Get224Bits(x, inwords); 655 Get224Bits(x, inwords);
652 Get224Bits(y, inwords + 7); 656 Get224Bits(y, inwords + 7);
653 memset(&z, 0, sizeof(z)); 657 memset(&z, 0, sizeof(z));
654 z[0] = 1; 658 z[0] = 1;
655 659
656 // Check that the point is on the curve, i.e. that y² = x³ - 3x + b. 660 // Check that the point is on the curve, i.e. that y² = x³ - 3x + b.
657 FieldElement lhs; 661 FieldElement lhs;
658 Square(&lhs, y); 662 Square(&lhs, y);
659 Contract(&lhs); 663 Contract(&lhs);
660 664
(...skipping 25 matching lines...) Expand all
686 690
687 Invert(&zinv, this->z); 691 Invert(&zinv, this->z);
688 Square(&zinv_sq, zinv); 692 Square(&zinv_sq, zinv);
689 Mul(&xx, x, zinv_sq); 693 Mul(&xx, x, zinv_sq);
690 Mul(&zinv_sq, zinv_sq, zinv); 694 Mul(&zinv_sq, zinv_sq, zinv);
691 Mul(&yy, y, zinv_sq); 695 Mul(&yy, y, zinv_sq);
692 696
693 Contract(&xx); 697 Contract(&xx);
694 Contract(&yy); 698 Contract(&yy);
695 699
696 uint32 outwords[14]; 700 uint32_t outwords[14];
697 Put224Bits(outwords, xx); 701 Put224Bits(outwords, xx);
698 Put224Bits(outwords + 7, yy); 702 Put224Bits(outwords + 7, yy);
699 return std::string(reinterpret_cast<const char*>(outwords), sizeof(outwords)); 703 return std::string(reinterpret_cast<const char*>(outwords), sizeof(outwords));
700 } 704 }
701 705
702 void ScalarMult(const Point& in, const uint8* scalar, Point* out) { 706 void ScalarMult(const Point& in, const uint8_t* scalar, Point* out) {
703 ::ScalarMult(out, in, scalar, 28); 707 ::ScalarMult(out, in, scalar, 28);
704 } 708 }
705 709
706 // kBasePoint is the base point (generator) of the elliptic curve group. 710 // kBasePoint is the base point (generator) of the elliptic curve group.
707 static const Point kBasePoint = { 711 static const Point kBasePoint = {
708 {22813985, 52956513, 34677300, 203240812, 712 {22813985, 52956513, 34677300, 203240812,
709 12143107, 133374265, 225162431, 191946955}, 713 12143107, 133374265, 225162431, 191946955},
710 {83918388, 223877528, 122119236, 123340192, 714 {83918388, 223877528, 122119236, 123340192,
711 266784067, 263504429, 146143011, 198407736}, 715 266784067, 263504429, 146143011, 198407736},
712 {1, 0, 0, 0, 0, 0, 0, 0}, 716 {1, 0, 0, 0, 0, 0, 0, 0},
713 }; 717 };
714 718
715 void ScalarBaseMult(const uint8* scalar, Point* out) { 719 void ScalarBaseMult(const uint8_t* scalar, Point* out) {
716 ::ScalarMult(out, kBasePoint, scalar, 28); 720 ::ScalarMult(out, kBasePoint, scalar, 28);
717 } 721 }
718 722
719 void Add(const Point& a, const Point& b, Point* out) { 723 void Add(const Point& a, const Point& b, Point* out) {
720 AddJacobian(out, a, b); 724 AddJacobian(out, a, b);
721 } 725 }
722 726
723 void Negate(const Point& in, Point* out) { 727 void Negate(const Point& in, Point* out) {
724 // Guide to elliptic curve cryptography, page 89 suggests that (X : X+Y : Z) 728 // Guide to elliptic curve cryptography, page 89 suggests that (X : X+Y : Z)
725 // is the negative in Jacobian coordinates, but it doesn't actually appear to 729 // is the negative in Jacobian coordinates, but it doesn't actually appear to
726 // be true in testing so this performs the negation in affine coordinates. 730 // be true in testing so this performs the negation in affine coordinates.
727 FieldElement zinv, zinv_sq, y; 731 FieldElement zinv, zinv_sq, y;
728 Invert(&zinv, in.z); 732 Invert(&zinv, in.z);
729 Square(&zinv_sq, zinv); 733 Square(&zinv_sq, zinv);
730 Mul(&out->x, in.x, zinv_sq); 734 Mul(&out->x, in.x, zinv_sq);
731 Mul(&zinv_sq, zinv_sq, zinv); 735 Mul(&zinv_sq, zinv_sq, zinv);
732 Mul(&y, in.y, zinv_sq); 736 Mul(&y, in.y, zinv_sq);
733 737
734 Subtract(&out->y, kP, y); 738 Subtract(&out->y, kP, y);
735 Reduce(&out->y); 739 Reduce(&out->y);
736 740
737 memset(&out->z, 0, sizeof(out->z)); 741 memset(&out->z, 0, sizeof(out->z));
738 out->z[0] = 1; 742 out->z[0] = 1;
739 } 743 }
740 744
741 } // namespace p224 745 } // namespace p224
742 746
743 } // namespace crypto 747 } // namespace crypto
OLDNEW
« no previous file with comments | « crypto/p224.h ('k') | crypto/p224_spake.h » ('j') | no next file with comments »

Powered by Google App Engine
This is Rietveld 408576698