/src/openvswitch/lib/hash.c
Line | Count | Source (jump to first uncovered line) |
1 | | /* |
2 | | * Copyright (c) 2008, 2009, 2010, 2012, 2013, 2014 Nicira, Inc. |
3 | | * |
4 | | * Licensed under the Apache License, Version 2.0 (the "License"); |
5 | | * you may not use this file except in compliance with the License. |
6 | | * You may obtain a copy of the License at: |
7 | | * |
8 | | * http://www.apache.org/licenses/LICENSE-2.0 |
9 | | * |
10 | | * Unless required by applicable law or agreed to in writing, software |
11 | | * distributed under the License is distributed on an "AS IS" BASIS, |
12 | | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
13 | | * See the License for the specific language governing permissions and |
14 | | * limitations under the License. |
15 | | */ |
16 | | #include <config.h> |
17 | | #include "hash.h" |
18 | | #include <string.h> |
19 | | #include "unaligned.h" |
20 | | |
21 | | /* Returns the hash of 'a', 'b', and 'c'. */ |
22 | | uint32_t |
23 | | hash_3words(uint32_t a, uint32_t b, uint32_t c) |
24 | 0 | { |
25 | 0 | return hash_finish(hash_add(hash_add(hash_add(a, 0), b), c), 12); |
26 | 0 | } |
27 | | |
28 | | /* Returns the hash of the 'n' bytes at 'p', starting from 'basis'. */ |
29 | | uint32_t |
30 | | hash_bytes(const void *p_, size_t n, uint32_t basis) |
31 | 0 | { |
32 | 0 | const uint8_t *p = p_; |
33 | 0 | size_t orig_n = n; |
34 | 0 | uint32_t hash; |
35 | |
|
36 | 0 | hash = basis; |
37 | 0 | while (n >= 4) { |
38 | 0 | hash = hash_add(hash, |
39 | 0 | get_unaligned_u32(ALIGNED_CAST(const uint32_t *, p))); |
40 | 0 | n -= 4; |
41 | 0 | p += 4; |
42 | 0 | } |
43 | |
|
44 | 0 | if (n) { |
45 | 0 | uint32_t tmp = 0; |
46 | |
|
47 | 0 | memcpy(&tmp, p, n); |
48 | 0 | hash = hash_add(hash, tmp); |
49 | 0 | } |
50 | |
|
51 | 0 | return hash_finish(hash, orig_n); |
52 | 0 | } |
53 | | |
54 | | uint32_t |
55 | | hash_double(double x, uint32_t basis) |
56 | 0 | { |
57 | 0 | uint32_t value[2]; |
58 | 0 | BUILD_ASSERT_DECL(sizeof x == sizeof value); |
59 | |
|
60 | 0 | memcpy(value, &x, sizeof value); |
61 | 0 | return hash_3words(value[0], value[1], basis); |
62 | 0 | } |
63 | | |
64 | | uint32_t |
65 | | hash_words__(const uint32_t *p, size_t n_words, uint32_t basis) |
66 | 0 | { |
67 | 0 | return hash_words_inline(p, n_words, basis); |
68 | 0 | } |
69 | | |
70 | | uint32_t |
71 | | hash_words64__(const uint64_t *p, size_t n_words, uint32_t basis) |
72 | 0 | { |
73 | 0 | return hash_words64_inline(p, n_words, basis); |
74 | 0 | } |
75 | | |
76 | | #if !(defined(__x86_64__)) && !(defined(__aarch64__)) |
77 | | void |
78 | | hash_bytes128(const void *p_, size_t len, uint32_t basis, ovs_u128 *out) |
79 | | { |
80 | | const uint32_t c1 = 0x239b961b; |
81 | | const uint32_t c2 = 0xab0e9789; |
82 | | const uint32_t c3 = 0x38b34ae5; |
83 | | const uint32_t c4 = 0xa1e38b93; |
84 | | const uint8_t *tail, *data = (const uint8_t *)p_; |
85 | | const uint32_t *blocks = (const uint32_t *)p_; |
86 | | const int nblocks = len / 16; |
87 | | uint32_t h1 = basis; |
88 | | uint32_t h2 = basis; |
89 | | uint32_t h3 = basis; |
90 | | uint32_t h4 = basis; |
91 | | |
92 | | /* Body */ |
93 | | for (int i = 0; i < nblocks; i++) { |
94 | | uint32_t k1 = get_unaligned_u32(&blocks[i * 4 + 0]); |
95 | | uint32_t k2 = get_unaligned_u32(&blocks[i * 4 + 1]); |
96 | | uint32_t k3 = get_unaligned_u32(&blocks[i * 4 + 2]); |
97 | | uint32_t k4 = get_unaligned_u32(&blocks[i * 4 + 3]); |
98 | | |
99 | | k1 *= c1; |
100 | | k1 = hash_rot(k1, 15); |
101 | | k1 *= c2; |
102 | | h1 ^= k1; |
103 | | |
104 | | h1 = hash_rot(h1, 19); |
105 | | h1 += h2; |
106 | | h1 = h1 * 5 + 0x561ccd1b; |
107 | | |
108 | | k2 *= c2; |
109 | | k2 = hash_rot(k2, 16); |
110 | | k2 *= c3; |
111 | | h2 ^= k2; |
112 | | |
113 | | h2 = hash_rot(h2, 17); |
114 | | h2 += h3; |
115 | | h2 = h2 * 5 + 0x0bcaa747; |
116 | | |
117 | | k3 *= c3; |
118 | | k3 = hash_rot(k3, 17); |
119 | | k3 *= c4; |
120 | | h3 ^= k3; |
121 | | |
122 | | h3 = hash_rot(h3, 15); |
123 | | h3 += h4; |
124 | | h3 = h3 * 5 + 0x96cd1c35; |
125 | | |
126 | | k4 *= c4; |
127 | | k4 = hash_rot(k4, 18); |
128 | | k4 *= c1; |
129 | | h4 ^= k4; |
130 | | |
131 | | h4 = hash_rot(h4, 13); |
132 | | h4 += h1; |
133 | | h4 = h4 * 5 + 0x32ac3b17; |
134 | | } |
135 | | |
136 | | /* Tail */ |
137 | | uint32_t k1, k2, k3, k4; |
138 | | k1 = k2 = k3 = k4 = 0; |
139 | | tail = data + nblocks * 16; |
140 | | switch (len & 15) { |
141 | | case 15: |
142 | | k4 ^= tail[14] << 16; |
143 | | /* fall through */ |
144 | | case 14: |
145 | | k4 ^= tail[13] << 8; |
146 | | /* fall through */ |
147 | | case 13: |
148 | | k4 ^= tail[12] << 0; |
149 | | k4 *= c4; |
150 | | k4 = hash_rot(k4, 18); |
151 | | k4 *= c1; |
152 | | h4 ^= k4; |
153 | | /* fall through */ |
154 | | |
155 | | case 12: |
156 | | k3 ^= tail[11] << 24; |
157 | | /* fall through */ |
158 | | case 11: |
159 | | k3 ^= tail[10] << 16; |
160 | | /* fall through */ |
161 | | case 10: |
162 | | k3 ^= tail[9] << 8; |
163 | | /* fall through */ |
164 | | case 9: |
165 | | k3 ^= tail[8] << 0; |
166 | | k3 *= c3; |
167 | | k3 = hash_rot(k3, 17); |
168 | | k3 *= c4; |
169 | | h3 ^= k3; |
170 | | /* fall through */ |
171 | | |
172 | | case 8: |
173 | | k2 ^= tail[7] << 24; |
174 | | /* fall through */ |
175 | | case 7: |
176 | | k2 ^= tail[6] << 16; |
177 | | /* fall through */ |
178 | | case 6: |
179 | | k2 ^= tail[5] << 8; |
180 | | /* fall through */ |
181 | | case 5: |
182 | | k2 ^= tail[4] << 0; |
183 | | k2 *= c2; |
184 | | k2 = hash_rot(k2, 16); |
185 | | k2 *= c3; |
186 | | h2 ^= k2; |
187 | | /* fall through */ |
188 | | |
189 | | case 4: |
190 | | k1 ^= tail[3] << 24; |
191 | | /* fall through */ |
192 | | case 3: |
193 | | k1 ^= tail[2] << 16; |
194 | | /* fall through */ |
195 | | case 2: |
196 | | k1 ^= tail[1] << 8; |
197 | | /* fall through */ |
198 | | case 1: |
199 | | k1 ^= tail[0] << 0; |
200 | | k1 *= c1; |
201 | | k1 = hash_rot(k1, 15); |
202 | | k1 *= c2; |
203 | | h1 ^= k1; |
204 | | }; |
205 | | |
206 | | /* Finalization */ |
207 | | h1 ^= len; |
208 | | h2 ^= len; |
209 | | h3 ^= len; |
210 | | h4 ^= len; |
211 | | |
212 | | h1 += h2; |
213 | | h1 += h3; |
214 | | h1 += h4; |
215 | | h2 += h1; |
216 | | h3 += h1; |
217 | | h4 += h1; |
218 | | |
219 | | h1 = mhash_finish(h1); |
220 | | h2 = mhash_finish(h2); |
221 | | h3 = mhash_finish(h3); |
222 | | h4 = mhash_finish(h4); |
223 | | |
224 | | h1 += h2; |
225 | | h1 += h3; |
226 | | h1 += h4; |
227 | | h2 += h1; |
228 | | h3 += h1; |
229 | | h4 += h1; |
230 | | |
231 | | out->u32[0] = h1; |
232 | | out->u32[1] = h2; |
233 | | out->u32[2] = h3; |
234 | | out->u32[3] = h4; |
235 | | } |
236 | | |
237 | | #else /* __x86_64__ or __aarch64__*/ |
238 | | |
239 | | static inline uint64_t |
240 | | hash_rot64(uint64_t x, int8_t r) |
241 | 0 | { |
242 | 0 | return (x << r) | (x >> (64 - r)); |
243 | 0 | } |
244 | | |
245 | | static inline uint64_t |
246 | | fmix64(uint64_t k) |
247 | 0 | { |
248 | 0 | k ^= k >> 33; |
249 | 0 | k *= 0xff51afd7ed558ccdULL; |
250 | 0 | k ^= k >> 33; |
251 | 0 | k *= 0xc4ceb9fe1a85ec53ULL; |
252 | 0 | k ^= k >> 33; |
253 | |
|
254 | 0 | return k; |
255 | 0 | } |
256 | | |
257 | | void |
258 | | hash_bytes128(const void *p_, size_t len, uint32_t basis, ovs_u128 *out) |
259 | 0 | { |
260 | 0 | const uint64_t c1 = 0x87c37b91114253d5ULL; |
261 | 0 | const uint64_t c2 = 0x4cf5ad432745937fULL; |
262 | 0 | const uint8_t *tail, *data = (const uint8_t *)p_; |
263 | 0 | const uint64_t *blocks = (const uint64_t *)p_; |
264 | 0 | const int nblocks = len / 16; |
265 | 0 | uint64_t h1 = basis; |
266 | 0 | uint64_t h2 = basis; |
267 | 0 | uint64_t k1, k2; |
268 | | |
269 | | /* Body */ |
270 | 0 | for (int i = 0; i < nblocks; i++) { |
271 | 0 | k1 = get_unaligned_u64(&blocks[i * 2 + 0]); |
272 | 0 | k2 = get_unaligned_u64(&blocks[i * 2 + 1]); |
273 | |
|
274 | 0 | k1 *= c1; |
275 | 0 | k1 = hash_rot64(k1, 31); |
276 | 0 | k1 *= c2; |
277 | 0 | h1 ^= k1; |
278 | |
|
279 | 0 | h1 = hash_rot64(h1, 27); |
280 | 0 | h1 += h2; |
281 | 0 | h1 = h1 * 5 + 0x52dce729; |
282 | |
|
283 | 0 | k2 *= c2; |
284 | 0 | k2 = hash_rot64(k2, 33); |
285 | 0 | k2 *= c1; |
286 | 0 | h2 ^= k2; |
287 | |
|
288 | 0 | h2 = hash_rot64(h2, 31); |
289 | 0 | h2 += h1; |
290 | 0 | h2 = h2 * 5 + 0x38495ab5; |
291 | 0 | } |
292 | | |
293 | | /* Tail */ |
294 | 0 | k1 = 0; |
295 | 0 | k2 = 0; |
296 | 0 | tail = data + nblocks * 16; |
297 | 0 | switch (len & 15) { |
298 | 0 | case 15: |
299 | 0 | k2 ^= ((uint64_t) tail[14]) << 48; |
300 | | /* fall through */ |
301 | 0 | case 14: |
302 | 0 | k2 ^= ((uint64_t) tail[13]) << 40; |
303 | | /* fall through */ |
304 | 0 | case 13: |
305 | 0 | k2 ^= ((uint64_t) tail[12]) << 32; |
306 | | /* fall through */ |
307 | 0 | case 12: |
308 | 0 | k2 ^= ((uint64_t) tail[11]) << 24; |
309 | | /* fall through */ |
310 | 0 | case 11: |
311 | 0 | k2 ^= ((uint64_t) tail[10]) << 16; |
312 | | /* fall through */ |
313 | 0 | case 10: |
314 | 0 | k2 ^= ((uint64_t) tail[9]) << 8; |
315 | | /* fall through */ |
316 | 0 | case 9: |
317 | 0 | k2 ^= ((uint64_t) tail[8]) << 0; |
318 | 0 | k2 *= c2; |
319 | 0 | k2 = hash_rot64(k2, 33); |
320 | 0 | k2 *= c1; |
321 | 0 | h2 ^= k2; |
322 | | /* fall through */ |
323 | 0 | case 8: |
324 | 0 | k1 ^= ((uint64_t) tail[7]) << 56; |
325 | | /* fall through */ |
326 | 0 | case 7: |
327 | 0 | k1 ^= ((uint64_t) tail[6]) << 48; |
328 | | /* fall through */ |
329 | 0 | case 6: |
330 | 0 | k1 ^= ((uint64_t) tail[5]) << 40; |
331 | | /* fall through */ |
332 | 0 | case 5: |
333 | 0 | k1 ^= ((uint64_t) tail[4]) << 32; |
334 | | /* fall through */ |
335 | 0 | case 4: |
336 | 0 | k1 ^= ((uint64_t) tail[3]) << 24; |
337 | | /* fall through */ |
338 | 0 | case 3: |
339 | 0 | k1 ^= ((uint64_t) tail[2]) << 16; |
340 | | /* fall through */ |
341 | 0 | case 2: |
342 | 0 | k1 ^= ((uint64_t) tail[1]) << 8; |
343 | | /* fall through */ |
344 | 0 | case 1: |
345 | 0 | k1 ^= ((uint64_t) tail[0]) << 0; |
346 | 0 | k1 *= c1; |
347 | 0 | k1 = hash_rot64(k1, 31); |
348 | 0 | k1 *= c2; |
349 | 0 | h1 ^= k1; |
350 | 0 | }; |
351 | | |
352 | | /* Finalization */ |
353 | 0 | h1 ^= len; |
354 | 0 | h2 ^= len; |
355 | 0 | h1 += h2; |
356 | 0 | h2 += h1; |
357 | 0 | h1 = fmix64(h1); |
358 | 0 | h2 = fmix64(h2); |
359 | 0 | h1 += h2; |
360 | 0 | h2 += h1; |
361 | |
|
362 | 0 | out->u64.lo = h1; |
363 | 0 | out->u64.hi = h2; |
364 | 0 | } |
365 | | #endif /* __x86_64__ or __aarch64__*/ |