]>
Commit | Line | Data |
---|---|---|
11fdf7f2 | 1 | /********************************************************************** |
f67539c2 | 2 | Copyright(c) 2017-2019, Intel Corporation All rights reserved. |
11fdf7f2 TL |
3 | |
4 | Redistribution and use in source and binary forms, with or without | |
5 | modification, are permitted provided that the following conditions | |
6 | are met: | |
7 | * Redistributions of source code must retain the above copyright | |
8 | notice, this list of conditions and the following disclaimer. | |
9 | * Redistributions in binary form must reproduce the above copyright | |
10 | notice, this list of conditions and the following disclaimer in | |
11 | the documentation and/or other materials provided with the | |
12 | distribution. | |
13 | * Neither the name of Intel Corporation nor the names of its | |
14 | contributors may be used to endorse or promote products derived | |
15 | from this software without specific prior written permission. | |
16 | ||
17 | THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS | |
18 | "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT | |
19 | LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR | |
20 | A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT | |
21 | OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, | |
22 | SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT | |
23 | LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, | |
24 | DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY | |
25 | THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT | |
26 | (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE | |
27 | OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. | |
28 | **********************************************************************/ | |
29 | ||
30 | #include <stdio.h> | |
31 | #include <stdlib.h> | |
f67539c2 | 32 | #include <inttypes.h> |
11fdf7f2 TL |
33 | #include <string.h> |
34 | #include <errno.h> | |
9f95a23c | 35 | #include <malloc.h> /* memalign() or _aligned_malloc()/aligned_free() */ |
11fdf7f2 TL |
36 | |
37 | #ifdef _WIN32 | |
38 | #include <windows.h> | |
39 | #include <process.h> | |
40 | #include <intrin.h> | |
f67539c2 | 41 | #define strdup _strdup |
11fdf7f2 TL |
42 | #define __forceinline static __forceinline |
43 | #else | |
44 | #include <x86intrin.h> | |
45 | #define __forceinline static inline __attribute__((always_inline)) | |
46 | #include <unistd.h> | |
47 | #include <pthread.h> | |
48 | #include <sched.h> | |
49 | #endif | |
50 | ||
51 | #include <intel-ipsec-mb.h> | |
52 | ||
53 | #include "msr.h" | |
54 | ||
f67539c2 | 55 | /* memory size for test buffers */ |
11fdf7f2 | 56 | #define BUFSIZE (512 * 1024 * 1024) |
f67539c2 TL |
57 | /* maximum size of a test buffer */ |
58 | #define JOB_SIZE_TOP (16 * 1024) | |
59 | /* min size of a buffer when testing range of buffers */ | |
60 | #define DEFAULT_JOB_SIZE_MIN 16 | |
61 | /* max size of a buffer when testing range of buffers */ | |
62 | #define DEFAULT_JOB_SIZE_MAX (2 * 1024) | |
63 | /* number of bytes to increase buffer size when testing range of buffers */ | |
64 | #define DEFAULT_JOB_SIZE_STEP 16 | |
65 | /* max offset applied to a buffer - this is to avoid collisions in L1 */ | |
66 | #define MAX_BUFFER_OFFSET 4096 | |
67 | /* max value of sha_size_incr */ | |
68 | #define MAX_SHA_SIZE_INCR 128 | |
69 | /* region size for one buffer rounded up to 4K page size */ | |
70 | #define REGION_SIZE (((JOB_SIZE_TOP + (MAX_BUFFER_OFFSET + \ | |
71 | MAX_SHA_SIZE_INCR)) + 4095) & (~4095)) | |
72 | /* number of test buffers */ | |
11fdf7f2 TL |
73 | #define NUM_OFFSETS (BUFSIZE / REGION_SIZE) |
74 | #define NUM_RUNS 16 | |
f67539c2 | 75 | /* maximum number of 128-bit expanded keys */ |
11fdf7f2 | 76 | #define KEYS_PER_JOB 15 |
9f95a23c | 77 | |
f67539c2 TL |
78 | #define AAD_SIZE_MAX JOB_SIZE_TOP |
79 | #define CCM_AAD_SIZE_MAX 46 | |
80 | #define DEFAULT_GCM_AAD_SIZE 12 | |
81 | #define DEFAULT_CCM_AAD_SIZE 8 | |
82 | ||
9f95a23c TL |
83 | #define ITER_SCALE_SMOKE 2048 |
84 | #define ITER_SCALE_SHORT 200000 | |
85 | #define ITER_SCALE_LONG 2000000 | |
86 | ||
11fdf7f2 TL |
87 | #define BITS(x) (sizeof(x) * 8) |
88 | #define DIM(x) (sizeof(x)/sizeof(x[0])) | |
89 | ||
11fdf7f2 TL |
90 | #define MAX_NUM_THREADS 16 /* Maximum number of threads that can be created */ |
91 | ||
f67539c2 TL |
92 | #define CIPHER_MODES_AES 7 /* CBC, CNTR, CNTR+8, CNTR_BITLEN, |
93 | CNTR_BITLEN-4, ECB, NULL_CIPHER */ | |
11fdf7f2 TL |
94 | #define CIPHER_MODES_DOCSIS 4 /* AES DOCSIS, AES DOCSIS+8, DES DOCSIS, |
95 | DES DOCSIS+8 */ | |
96 | #define CIPHER_MODES_DES 1 /* DES */ | |
97 | #define CIPHER_MODES_GCM 1 /* GCM */ | |
98 | #define CIPHER_MODES_CCM 1 /* CCM */ | |
99 | #define CIPHER_MODES_3DES 1 /* 3DES */ | |
f67539c2 | 100 | #define CIPHER_MODES_PON 2 /* PON, NO_CTR PON */ |
11fdf7f2 | 101 | #define DIRECTIONS 2 /* ENC, DEC */ |
f67539c2 TL |
102 | #define HASH_ALGS_AES 10 /* SHA1, SHA256, SHA224, SHA384, SHA512, XCBC, |
103 | MD5, NULL_HASH, CMAC, CMAC_BITLEN */ | |
11fdf7f2 TL |
104 | #define HASH_ALGS_DOCSIS 1 /* NULL_HASH */ |
105 | #define HASH_ALGS_GCM 1 /* GCM */ | |
106 | #define HASH_ALGS_CCM 1 /* CCM */ | |
107 | #define HASH_ALGS_DES 1 /* NULL_HASH for DES */ | |
108 | #define HASH_ALGS_3DES 1 /* NULL_HASH for 3DES */ | |
f67539c2 | 109 | #define HASH_ALGS_PON 1 /* CRC32/BIP for PON */ |
11fdf7f2 TL |
110 | #define KEY_SIZES_AES 3 /* 16, 24, 32 */ |
111 | #define KEY_SIZES_DOCSIS 1 /* 16 or 8 */ | |
112 | #define KEY_SIZES_GCM 3 /* 16, 24, 32 */ | |
113 | #define KEY_SIZES_CCM 1 /* 16 */ | |
114 | #define KEY_SIZES_DES 1 /* 8 */ | |
115 | #define KEY_SIZES_3DES 1 /* 8 x 3 */ | |
f67539c2 | 116 | #define KEY_SIZES_PON 1 /* 16 */ |
11fdf7f2 TL |
117 | |
118 | #define IA32_MSR_FIXED_CTR_CTRL 0x38D | |
119 | #define IA32_MSR_PERF_GLOBAL_CTR 0x38F | |
120 | #define IA32_MSR_CPU_UNHALTED_THREAD 0x30A | |
121 | ||
122 | /* Those defines tell how many different test cases are to be performed. | |
123 | * Have to be multiplied by number of chosen architectures. | |
124 | */ | |
125 | #define VARIANTS_PER_ARCH_AES (CIPHER_MODES_AES * DIRECTIONS * \ | |
126 | HASH_ALGS_AES * KEY_SIZES_AES) | |
127 | #define VARIANTS_PER_ARCH_DOCSIS (CIPHER_MODES_DOCSIS * DIRECTIONS * \ | |
128 | HASH_ALGS_DOCSIS * KEY_SIZES_DOCSIS) | |
129 | #define VARIANTS_PER_ARCH_GCM (CIPHER_MODES_GCM * DIRECTIONS * \ | |
130 | HASH_ALGS_GCM * KEY_SIZES_GCM) | |
131 | #define VARIANTS_PER_ARCH_CCM (CIPHER_MODES_CCM * DIRECTIONS * \ | |
132 | HASH_ALGS_CCM * KEY_SIZES_CCM) | |
133 | #define VARIANTS_PER_ARCH_DES (CIPHER_MODES_DES * DIRECTIONS * \ | |
134 | HASH_ALGS_DES * KEY_SIZES_DES) | |
135 | #define VARIANTS_PER_ARCH_3DES (CIPHER_MODES_3DES * DIRECTIONS * \ | |
136 | HASH_ALGS_3DES * KEY_SIZES_3DES) | |
f67539c2 TL |
137 | #define VARIANTS_PER_ARCH_PON (CIPHER_MODES_PON * DIRECTIONS * \ |
138 | HASH_ALGS_PON * KEY_SIZES_PON) | |
11fdf7f2 | 139 | |
11fdf7f2 TL |
140 | enum arch_type_e { |
141 | ARCH_SSE = 0, | |
142 | ARCH_AVX, | |
143 | ARCH_AVX2, | |
f67539c2 TL |
144 | ARCH_AVX512, |
145 | NUM_ARCHS | |
11fdf7f2 TL |
146 | }; |
147 | ||
148 | enum test_type_e { | |
149 | TTYPE_AES_HMAC, | |
150 | TTYPE_AES_DOCSIS, | |
151 | TTYPE_AES_GCM, | |
152 | TTYPE_AES_CCM, | |
153 | TTYPE_AES_DES, | |
f67539c2 TL |
154 | TTYPE_AES_3DES, |
155 | TTYPE_PON, | |
156 | TTYPE_CUSTOM, | |
157 | NUM_TTYPES | |
11fdf7f2 TL |
158 | }; |
159 | ||
f67539c2 TL |
160 | /* This enum will be mostly translated to JOB_CIPHER_MODE |
161 | * (make sure to update c_mode_names list in print_times function) */ | |
11fdf7f2 TL |
162 | enum test_cipher_mode_e { |
163 | TEST_CBC = 1, | |
164 | TEST_CNTR, | |
165 | TEST_CNTR8, /* CNTR with increased buffer by 8 */ | |
f67539c2 TL |
166 | TEST_CNTR_BITLEN, /* CNTR-BITLEN */ |
167 | TEST_CNTR_BITLEN4, /* CNTR-BITLEN with 4 less bits in the last byte */ | |
168 | TEST_ECB, | |
11fdf7f2 TL |
169 | TEST_NULL_CIPHER, |
170 | TEST_AESDOCSIS, | |
171 | TEST_AESDOCSIS8, /* AES DOCSIS with increased buffer size by 8 */ | |
172 | TEST_DESDOCSIS, | |
173 | TEST_DESDOCSIS4, /* DES DOCSIS with increased buffer size by 4 */ | |
174 | TEST_GCM, /* Additional field used by GCM, not translated */ | |
175 | TEST_CCM, | |
176 | TEST_DES, | |
177 | TEST_3DES, | |
f67539c2 TL |
178 | TEST_PON_CNTR, |
179 | TEST_PON_NO_CNTR, | |
180 | TEST_NUM_CIPHER_TESTS | |
11fdf7f2 TL |
181 | }; |
182 | ||
f67539c2 TL |
183 | /* This enum will be mostly translated to JOB_HASH_ALG |
184 | * (make sure to update h_alg_names list in print_times function) */ | |
11fdf7f2 TL |
185 | enum test_hash_alg_e { |
186 | TEST_SHA1 = 1, | |
187 | TEST_SHA_224, | |
188 | TEST_SHA_256, | |
189 | TEST_SHA_384, | |
190 | TEST_SHA_512, | |
191 | TEST_XCBC, | |
192 | TEST_MD5, | |
193 | TEST_HASH_CMAC, /* added here to be included in AES tests */ | |
f67539c2 | 194 | TEST_HASH_CMAC_BITLEN, |
11fdf7f2 TL |
195 | TEST_NULL_HASH, |
196 | TEST_HASH_GCM, /* Additional field used by GCM, not translated */ | |
197 | TEST_CUSTOM_HASH, /* unused */ | |
f67539c2 TL |
198 | TEST_HASH_CCM, |
199 | TEST_PON_CRC_BIP, | |
200 | TEST_NUM_HASH_TESTS | |
11fdf7f2 TL |
201 | }; |
202 | ||
203 | /* Struct storing cipher parameters */ | |
204 | struct params_s { | |
205 | JOB_CIPHER_DIRECTION cipher_dir; | |
206 | enum test_type_e test_type; /* AES, DOCSIS, GCM */ | |
207 | enum test_cipher_mode_e cipher_mode; | |
208 | enum test_hash_alg_e hash_alg; | |
209 | uint32_t aes_key_size; | |
210 | uint32_t size_aes; | |
f67539c2 | 211 | uint64_t aad_size; |
11fdf7f2 TL |
212 | uint32_t num_sizes; |
213 | uint32_t num_variants; | |
214 | uint32_t core; | |
215 | }; | |
216 | ||
f67539c2 TL |
217 | struct custom_job_params { |
218 | enum test_cipher_mode_e cipher_mode; | |
219 | enum test_hash_alg_e hash_alg; | |
220 | uint32_t aes_key_size; | |
221 | JOB_CIPHER_DIRECTION cipher_dir; | |
222 | }; | |
223 | ||
224 | union params { | |
225 | enum arch_type_e arch_type; | |
226 | struct custom_job_params job_params; | |
227 | }; | |
228 | ||
229 | struct str_value_mapping { | |
230 | const char *name; | |
231 | union params values; | |
232 | }; | |
233 | ||
234 | struct str_value_mapping arch_str_map[] = { | |
235 | {.name = "SSE", .values.arch_type = ARCH_SSE }, | |
236 | {.name = "AVX", .values.arch_type = ARCH_AVX }, | |
237 | {.name = "AVX2", .values.arch_type = ARCH_AVX2 }, | |
238 | {.name = "AVX512", .values.arch_type = ARCH_AVX512 } | |
239 | }; | |
240 | ||
241 | struct str_value_mapping cipher_algo_str_map[] = { | |
242 | { | |
243 | .name = "aes-cbc-128", | |
244 | .values.job_params = { | |
245 | .cipher_mode = TEST_CBC, | |
246 | .aes_key_size = AES_128_BYTES | |
247 | } | |
248 | }, | |
249 | { | |
250 | .name = "aes-cbc-192", | |
251 | .values.job_params = { | |
252 | .cipher_mode = TEST_CBC, | |
253 | .aes_key_size = AES_192_BYTES | |
254 | } | |
255 | }, | |
256 | { | |
257 | .name = "aes-cbc-256", | |
258 | .values.job_params = { | |
259 | .cipher_mode = TEST_CBC, | |
260 | .aes_key_size = AES_256_BYTES | |
261 | } | |
262 | }, | |
263 | { | |
264 | .name = "aes-ctr-128", | |
265 | .values.job_params = { | |
266 | .cipher_mode = TEST_CNTR, | |
267 | .aes_key_size = AES_128_BYTES | |
268 | } | |
269 | }, | |
270 | { | |
271 | .name = "aes-ctr-192", | |
272 | .values.job_params = { | |
273 | .cipher_mode = TEST_CNTR, | |
274 | .aes_key_size = AES_192_BYTES | |
275 | } | |
276 | }, | |
277 | { | |
278 | .name = "aes-ctr-256", | |
279 | .values.job_params = { | |
280 | .cipher_mode = TEST_CNTR, | |
281 | .aes_key_size = AES_256_BYTES | |
282 | } | |
283 | }, | |
284 | { | |
285 | .name = "aes-ctr8-128", | |
286 | .values.job_params = { | |
287 | .cipher_mode = TEST_CNTR8, | |
288 | .aes_key_size = AES_128_BYTES | |
289 | } | |
290 | }, | |
291 | { | |
292 | .name = "aes-ctr8-192", | |
293 | .values.job_params = { | |
294 | .cipher_mode = TEST_CNTR8, | |
295 | .aes_key_size = AES_192_BYTES | |
296 | } | |
297 | }, | |
298 | { | |
299 | .name = "aes-ctr8-256", | |
300 | .values.job_params = { | |
301 | .cipher_mode = TEST_CNTR8, | |
302 | .aes_key_size = AES_256_BYTES | |
303 | } | |
304 | }, | |
305 | { | |
306 | .name = "aes-ctr-bit-128", | |
307 | .values.job_params = { | |
308 | .cipher_mode = TEST_CNTR_BITLEN, | |
309 | .aes_key_size = AES_128_BYTES | |
310 | } | |
311 | }, | |
312 | { | |
313 | .name = "aes-ctr-bit-192", | |
314 | .values.job_params = { | |
315 | .cipher_mode = TEST_CNTR_BITLEN, | |
316 | .aes_key_size = AES_192_BYTES | |
317 | } | |
318 | }, | |
319 | { | |
320 | .name = "aes-ctr-bit-256", | |
321 | .values.job_params = { | |
322 | .cipher_mode = TEST_CNTR_BITLEN, | |
323 | .aes_key_size = AES_256_BYTES | |
324 | } | |
325 | }, | |
326 | { | |
327 | .name = "aes-ctr-bit4-128", | |
328 | .values.job_params = { | |
329 | .cipher_mode = TEST_CNTR_BITLEN4, | |
330 | .aes_key_size = AES_128_BYTES | |
331 | } | |
332 | }, | |
333 | { | |
334 | .name = "aes-ctr-bit4-192", | |
335 | .values.job_params = { | |
336 | .cipher_mode = TEST_CNTR_BITLEN4, | |
337 | .aes_key_size = AES_192_BYTES | |
338 | } | |
339 | }, | |
340 | { | |
341 | .name = "aes-ctr-bit4-256", | |
342 | .values.job_params = { | |
343 | .cipher_mode = TEST_CNTR_BITLEN4, | |
344 | .aes_key_size = AES_256_BYTES | |
345 | } | |
346 | }, | |
347 | { | |
348 | .name = "aes-ecb-128", | |
349 | .values.job_params = { | |
350 | .cipher_mode = TEST_ECB, | |
351 | .aes_key_size = AES_128_BYTES | |
352 | } | |
353 | }, | |
354 | { | |
355 | .name = "aes-ecb-192", | |
356 | .values.job_params = { | |
357 | .cipher_mode = TEST_ECB, | |
358 | .aes_key_size = AES_192_BYTES | |
359 | } | |
360 | }, | |
361 | { | |
362 | .name = "aes-ecb-256", | |
363 | .values.job_params = { | |
364 | .cipher_mode = TEST_ECB, | |
365 | .aes_key_size = AES_256_BYTES | |
366 | } | |
367 | }, | |
368 | { | |
369 | .name = "aes-docsis", | |
370 | .values.job_params = { | |
371 | .cipher_mode = TEST_AESDOCSIS, | |
372 | .aes_key_size = AES_128_BYTES | |
373 | } | |
374 | }, | |
375 | { | |
376 | .name = "aes-docsis8", | |
377 | .values.job_params = { | |
378 | .cipher_mode = TEST_AESDOCSIS8, | |
379 | .aes_key_size = AES_128_BYTES | |
380 | } | |
381 | }, | |
382 | { | |
383 | .name = "des-docsis", | |
384 | .values.job_params = { | |
385 | .cipher_mode = TEST_DESDOCSIS, | |
386 | .aes_key_size = 8 | |
387 | } | |
388 | }, | |
389 | { | |
390 | .name = "des-docsis4", | |
391 | .values.job_params = { | |
392 | .cipher_mode = TEST_DESDOCSIS4, | |
393 | .aes_key_size = 8 | |
394 | } | |
395 | }, | |
396 | { | |
397 | .name = "des-cbc", | |
398 | .values.job_params = { | |
399 | .cipher_mode = TEST_DES, | |
400 | .aes_key_size = 8 | |
401 | } | |
402 | }, | |
403 | { | |
404 | .name = "3des-cbc", | |
405 | .values.job_params = { | |
406 | .cipher_mode = TEST_3DES, | |
407 | .aes_key_size = 8 | |
408 | } | |
409 | }, | |
410 | { | |
411 | .name = "null", | |
412 | .values.job_params = { | |
413 | .cipher_mode = TEST_NULL_CIPHER, | |
414 | .aes_key_size = 0 | |
415 | } | |
416 | } | |
417 | }; | |
418 | ||
419 | struct str_value_mapping hash_algo_str_map[] = { | |
420 | { | |
421 | .name = "sha1-hmac", | |
422 | .values.job_params = { | |
423 | .hash_alg = TEST_SHA1 | |
424 | } | |
425 | }, | |
426 | { | |
427 | .name = "sha224-hmac", | |
428 | .values.job_params = { | |
429 | .hash_alg = TEST_SHA_224 | |
430 | } | |
431 | }, | |
432 | { | |
433 | .name = "sha256-hmac", | |
434 | .values.job_params = { | |
435 | .hash_alg = TEST_SHA_256 | |
436 | } | |
437 | }, | |
438 | { | |
439 | .name = "sha384-hmac", | |
440 | .values.job_params = { | |
441 | .hash_alg = TEST_SHA_384 | |
442 | } | |
443 | }, | |
444 | { | |
445 | .name = "sha512-hmac", | |
446 | .values.job_params = { | |
447 | .hash_alg = TEST_SHA_512 | |
448 | } | |
449 | }, | |
450 | { | |
451 | .name = "aes-xcbc", | |
452 | .values.job_params = { | |
453 | .hash_alg = TEST_XCBC | |
454 | } | |
455 | }, | |
456 | { | |
457 | .name = "md5-hmac", | |
458 | .values.job_params = { | |
459 | .hash_alg = TEST_MD5 | |
460 | } | |
461 | }, | |
462 | { | |
463 | .name = "aes-cmac", | |
464 | .values.job_params = { | |
465 | .hash_alg = TEST_HASH_CMAC | |
466 | } | |
467 | }, | |
468 | { | |
469 | .name = "null", | |
470 | .values.job_params = { | |
471 | .hash_alg = TEST_NULL_HASH | |
472 | } | |
473 | }, | |
474 | { | |
475 | .name = "aes-cmac-bitlen", | |
476 | .values.job_params = { | |
477 | .hash_alg = TEST_HASH_CMAC_BITLEN | |
478 | } | |
479 | }, | |
480 | }; | |
481 | ||
482 | struct str_value_mapping aead_algo_str_map[] = { | |
483 | { | |
484 | .name = "aes-gcm-128", | |
485 | .values.job_params = { | |
486 | .cipher_mode = TEST_GCM, | |
487 | .hash_alg = TEST_HASH_GCM, | |
488 | .aes_key_size = AES_128_BYTES | |
489 | } | |
490 | }, | |
491 | { | |
492 | .name = "aes-gcm-192", | |
493 | .values.job_params = { | |
494 | .cipher_mode = TEST_GCM, | |
495 | .hash_alg = TEST_HASH_GCM, | |
496 | .aes_key_size = AES_192_BYTES | |
497 | } | |
498 | }, | |
499 | { | |
500 | .name = "aes-gcm-256", | |
501 | .values.job_params = { | |
502 | .cipher_mode = TEST_GCM, | |
503 | .hash_alg = TEST_HASH_GCM, | |
504 | .aes_key_size = AES_256_BYTES | |
505 | } | |
506 | }, | |
507 | { | |
508 | .name = "aes-ccm-128", | |
509 | .values.job_params = { | |
510 | .cipher_mode = TEST_CCM, | |
511 | .hash_alg = TEST_HASH_CCM, | |
512 | .aes_key_size = AES_128_BYTES | |
513 | } | |
514 | }, | |
515 | { | |
516 | .name = "pon-128", | |
517 | .values.job_params = { | |
518 | .cipher_mode = TEST_PON_CNTR, | |
519 | .hash_alg = TEST_PON_CRC_BIP, | |
520 | .aes_key_size = AES_128_BYTES | |
521 | } | |
522 | }, | |
523 | { | |
524 | .name = "pon-128-no-ctr", | |
525 | .values.job_params = { | |
526 | .cipher_mode = TEST_PON_NO_CNTR, | |
527 | .hash_alg = TEST_PON_CRC_BIP, | |
528 | .aes_key_size = 0 | |
529 | } | |
530 | }, | |
531 | }; | |
532 | ||
533 | struct str_value_mapping cipher_dir_str_map[] = { | |
534 | {.name = "encrypt", .values.job_params.cipher_dir = ENCRYPT}, | |
535 | {.name = "decrypt", .values.job_params.cipher_dir = DECRYPT} | |
536 | }; | |
537 | ||
11fdf7f2 TL |
538 | /* This struct stores all information about performed test case */ |
539 | struct variant_s { | |
540 | uint32_t arch; | |
541 | struct params_s params; | |
542 | uint64_t *avg_times; | |
543 | }; | |
544 | ||
545 | /* Struct storing information to be passed to threads */ | |
546 | struct thread_info { | |
547 | int print_info; | |
548 | int core; | |
9f95a23c | 549 | MB_MGR *p_mgr; |
11fdf7f2 TL |
550 | } t_info[MAX_NUM_THREADS]; |
551 | ||
552 | enum cache_type_e { | |
553 | WARM = 0, | |
554 | COLD = 1 | |
555 | }; | |
556 | ||
11fdf7f2 | 557 | enum cache_type_e cache_type = WARM; |
f67539c2 TL |
558 | |
559 | const uint32_t auth_tag_length_bytes[19] = { | |
560 | 12, /* SHA1 */ | |
561 | 14, /* SHA_224 */ | |
562 | 16, /* SHA_256 */ | |
563 | 24, /* SHA_384 */ | |
564 | 32, /* SHA_512 */ | |
565 | 12, /* AES_XCBC */ | |
566 | 12, /* MD5 */ | |
567 | 0, /* NULL_HASH */ | |
568 | #ifndef NO_GCM | |
569 | 16, /* AES_GMAC */ | |
570 | #endif | |
571 | 0, /* CUSTOM HASH */ | |
572 | 0, /* AES_CCM */ | |
573 | 16, /* AES_CMAC */ | |
574 | 20, /* PLAIN_SHA1 */ | |
575 | 28, /* PLAIN_SHA_224 */ | |
576 | 32, /* PLAIN_SHA_256 */ | |
577 | 48, /* PLAIN_SHA_384 */ | |
578 | 64, /* PLAIN_SHA_512 */ | |
579 | 4, /* AES_CMAC_BITLEN (3GPP) */ | |
580 | 8, /* PON */ | |
11fdf7f2 | 581 | }; |
11fdf7f2 | 582 | uint32_t index_limit; |
11fdf7f2 TL |
583 | uint32_t key_idxs[NUM_OFFSETS]; |
584 | uint32_t offsets[NUM_OFFSETS]; | |
f67539c2 TL |
585 | uint32_t sha_size_incr = 24; |
586 | ||
587 | enum range { | |
588 | RANGE_MIN = 0, | |
589 | RANGE_STEP, | |
590 | RANGE_MAX, | |
591 | NUM_RANGE | |
592 | }; | |
593 | ||
594 | uint32_t job_sizes[NUM_RANGE] = {DEFAULT_JOB_SIZE_MIN, | |
595 | DEFAULT_JOB_SIZE_STEP, | |
596 | DEFAULT_JOB_SIZE_MAX}; | |
597 | uint32_t job_iter = 0; | |
598 | uint64_t gcm_aad_size = DEFAULT_GCM_AAD_SIZE; | |
599 | uint64_t ccm_aad_size = DEFAULT_CCM_AAD_SIZE; | |
600 | ||
601 | struct custom_job_params custom_job_params = { | |
602 | .cipher_mode = TEST_NULL_CIPHER, | |
603 | .hash_alg = TEST_NULL_HASH, | |
604 | .aes_key_size = 0, | |
605 | .cipher_dir = ENCRYPT | |
606 | }; | |
11fdf7f2 TL |
607 | |
608 | uint8_t archs[NUM_ARCHS] = {1, 1, 1, 1}; /* uses all function sets */ | |
f67539c2 TL |
609 | /* AES, DOCSIS, GCM, CCM, DES, 3DES, PON, CUSTOM */ |
610 | uint8_t test_types[NUM_TTYPES] = {1, 1, 1, 1, 1, 1, 1, 0}; | |
11fdf7f2 TL |
611 | |
612 | int use_gcm_job_api = 0; | |
613 | int use_unhalted_cycles = 0; /* read unhalted cycles instead of tsc */ | |
614 | uint64_t rd_cycles_cost = 0; /* cost of reading unhalted cycles */ | |
615 | uint64_t core_mask = 0; /* bitmap of selected cores */ | |
616 | ||
617 | uint64_t flags = 0; /* flags passed to alloc_mb_mgr() */ | |
618 | ||
9f95a23c | 619 | uint32_t iter_scale = ITER_SCALE_LONG; |
11fdf7f2 | 620 | |
f67539c2 TL |
621 | #define PB_INIT_SIZE 50 |
622 | #define PB_INIT_IDX 2 /* after \r and [ */ | |
623 | static uint32_t PB_SIZE = PB_INIT_SIZE; | |
624 | static uint32_t PB_FINAL_IDX = (PB_INIT_SIZE + (PB_INIT_IDX - 1)); | |
625 | static char prog_bar[PB_INIT_SIZE + 4]; /* 50 + 4 for \r, [, ], \0 */ | |
626 | static uint32_t pb_idx = PB_INIT_IDX; | |
627 | static uint32_t pb_mod = 0; | |
628 | ||
629 | static int silent_progress_bar = 0; | |
630 | ||
631 | static void prog_bar_init(const uint32_t total_num) | |
632 | { | |
633 | if (silent_progress_bar) | |
634 | return; | |
635 | ||
636 | if (total_num < PB_SIZE) { | |
637 | PB_SIZE = total_num; | |
638 | PB_FINAL_IDX = (PB_SIZE + (PB_INIT_IDX - 1)); | |
639 | } | |
640 | pb_idx = PB_INIT_IDX; | |
641 | pb_mod = total_num / PB_SIZE; | |
642 | ||
643 | /* 32 dec == ascii ' ' char */ | |
644 | memset(prog_bar, 32, sizeof(prog_bar)); | |
645 | prog_bar[0] = '\r'; | |
646 | prog_bar[1] = '['; | |
647 | prog_bar[PB_FINAL_IDX + 1] = ']'; | |
648 | prog_bar[PB_FINAL_IDX + 2] = '\0'; | |
649 | ||
650 | fputs(prog_bar, stderr); | |
651 | } | |
652 | ||
653 | static void prog_bar_fini(void) | |
654 | { | |
655 | if (silent_progress_bar) | |
656 | return; | |
657 | ||
658 | prog_bar[PB_FINAL_IDX] = 'X'; /* set final X */ | |
659 | fputs(prog_bar, stderr); | |
660 | } | |
661 | ||
662 | static void prog_bar_update(const uint32_t num) | |
663 | { | |
664 | if (silent_progress_bar) | |
665 | return; | |
666 | ||
667 | if ((pb_mod == 0) || num % pb_mod == 0) { | |
668 | /* print X at every ~50th variant */ | |
669 | prog_bar[pb_idx] = 'X'; | |
670 | fputs(prog_bar, stderr); | |
671 | ||
672 | /* don't overrun final idx */ | |
673 | if (pb_idx < (PB_SIZE + 1)) | |
674 | pb_idx++; | |
675 | } else { | |
676 | const char pb_inter_chars[] = {'|', '/', '-', '\\'}; | |
677 | /* print intermediate chars */ | |
678 | prog_bar[pb_idx] = pb_inter_chars[num % DIM(pb_inter_chars)]; | |
679 | fputs(prog_bar, stderr); | |
680 | } | |
681 | } | |
682 | ||
11fdf7f2 | 683 | /* Read unhalted cycles */ |
f67539c2 | 684 | __forceinline uint64_t read_cycles(const uint32_t core) |
11fdf7f2 TL |
685 | { |
686 | uint64_t val = 0; | |
687 | ||
688 | if (msr_read(core, IA32_MSR_CPU_UNHALTED_THREAD, | |
689 | &val) != MACHINE_RETVAL_OK) { | |
690 | fprintf(stderr, "Error reading cycles " | |
691 | "counter on core %u!\n", core); | |
692 | exit(EXIT_FAILURE); | |
693 | } | |
694 | ||
695 | return val; | |
696 | } | |
697 | ||
f67539c2 TL |
698 | /* Method used by qsort to compare 2 values */ |
699 | static int compare_uint64_t(const void *a, const void *b) | |
11fdf7f2 | 700 | { |
f67539c2 | 701 | return (int)(int64_t)(*(const uint64_t *)a - *(const uint64_t *)b); |
11fdf7f2 TL |
702 | } |
703 | ||
704 | /* Get number of bits set in value */ | |
f67539c2 | 705 | static unsigned bitcount(const uint64_t val) |
11fdf7f2 | 706 | { |
f67539c2 | 707 | unsigned i, bits = 0; |
11fdf7f2 TL |
708 | |
709 | for (i = 0; i < BITS(val); i++) | |
710 | if (val & (1ULL << i)) | |
711 | bits++; | |
712 | ||
713 | return bits; | |
714 | } | |
715 | ||
716 | /* Get the next core in core mask | |
717 | Set last_core to negative to start from beginnig of core_mask */ | |
718 | static int next_core(const uint64_t core_mask, | |
719 | const int last_core) | |
720 | { | |
721 | int core = 0; | |
722 | ||
723 | if (last_core >= 0) | |
724 | core = last_core; | |
725 | ||
726 | while (((core_mask >> core) & 1) == 0) { | |
727 | core++; | |
728 | ||
729 | if (core >= (int)BITS(core_mask)) | |
730 | return -1; | |
731 | } | |
732 | ||
733 | return core; | |
734 | } | |
735 | ||
736 | /* Set CPU affinity for current thread */ | |
737 | static int set_affinity(const int cpu) | |
738 | { | |
739 | int ret = 0; | |
11fdf7f2 TL |
740 | int num_cpus = 0; |
741 | ||
742 | /* Get number of cpus in the system */ | |
9f95a23c TL |
743 | #ifdef _WIN32 |
744 | GROUP_AFFINITY NewGroupAffinity; | |
745 | ||
746 | memset(&NewGroupAffinity, 0, sizeof(GROUP_AFFINITY)); | |
747 | num_cpus = GetActiveProcessorCount(ALL_PROCESSOR_GROUPS); | |
748 | #else | |
11fdf7f2 | 749 | num_cpus = sysconf(_SC_NPROCESSORS_CONF); |
9f95a23c | 750 | #endif |
11fdf7f2 TL |
751 | if (num_cpus == 0) { |
752 | fprintf(stderr, "Zero processors in the system!"); | |
753 | return 1; | |
754 | } | |
755 | ||
756 | /* Check if selected core is valid */ | |
757 | if (cpu < 0 || cpu >= num_cpus) { | |
758 | fprintf(stderr, "Invalid CPU selected! " | |
759 | "Max valid CPU is %u\n", num_cpus - 1); | |
760 | return 1; | |
761 | } | |
762 | ||
9f95a23c TL |
763 | #ifdef _WIN32 |
764 | NewGroupAffinity.Mask = 1ULL << cpu; | |
765 | ret = !SetThreadGroupAffinity(GetCurrentThread(), | |
766 | &NewGroupAffinity, NULL); | |
767 | #else | |
768 | cpu_set_t cpuset; | |
769 | ||
11fdf7f2 TL |
770 | CPU_ZERO(&cpuset); |
771 | CPU_SET(cpu, &cpuset); | |
772 | ||
773 | /* Set affinity of current process to cpu */ | |
774 | ret = sched_setaffinity(0, sizeof(cpuset), &cpuset); | |
775 | #endif /* _WIN32 */ | |
9f95a23c | 776 | |
11fdf7f2 TL |
777 | return ret; |
778 | } | |
779 | ||
780 | /* Start counting unhalted cycles */ | |
f67539c2 | 781 | static int start_cycles_ctr(const uint32_t core) |
11fdf7f2 TL |
782 | { |
783 | int ret; | |
784 | ||
785 | if (core >= BITS(core_mask)) | |
786 | return 1; | |
787 | ||
788 | /* Disable cycles counter */ | |
789 | ret = msr_write(core, IA32_MSR_PERF_GLOBAL_CTR, 0); | |
790 | if (ret != MACHINE_RETVAL_OK) | |
791 | return ret; | |
792 | ||
793 | /* Zero cycles counter */ | |
794 | ret = msr_write(core, IA32_MSR_CPU_UNHALTED_THREAD, 0); | |
795 | if (ret != MACHINE_RETVAL_OK) | |
796 | return ret; | |
797 | ||
798 | /* Enable OS and user tracking in FixedCtr1 */ | |
799 | ret = msr_write(core, IA32_MSR_FIXED_CTR_CTRL, 0x30); | |
800 | if (ret != MACHINE_RETVAL_OK) | |
801 | return ret; | |
802 | ||
803 | /* Enable cycles counter */ | |
804 | return msr_write(core, IA32_MSR_PERF_GLOBAL_CTR, (1ULL << 33)); | |
805 | } | |
806 | ||
807 | /* Init MSR module */ | |
808 | static int init_msr_mod(void) | |
809 | { | |
810 | unsigned max_core_count = 0; | |
9f95a23c TL |
811 | #ifdef _WIN32 |
812 | max_core_count = GetActiveProcessorCount(ALL_PROCESSOR_GROUPS); | |
813 | #else | |
11fdf7f2 | 814 | max_core_count = sysconf(_SC_NPROCESSORS_CONF); |
9f95a23c | 815 | #endif |
11fdf7f2 TL |
816 | if (max_core_count == 0) { |
817 | fprintf(stderr, "Zero processors in the system!"); | |
818 | return MACHINE_RETVAL_ERROR; | |
819 | } | |
9f95a23c | 820 | |
11fdf7f2 TL |
821 | return machine_init(max_core_count); |
822 | } | |
823 | ||
824 | /* Set the cost of reading unhalted cycles using RDMSR */ | |
825 | static int set_unhalted_cycle_cost(const int core, uint64_t *value) | |
826 | { | |
827 | uint64_t time1, time2; | |
828 | ||
829 | if (value == NULL || core < 0) | |
830 | return 1; | |
831 | ||
832 | time1 = read_cycles(core); | |
833 | time2 = read_cycles(core); | |
834 | ||
835 | /* Calculate delta */ | |
836 | *value = (time2 - time1); | |
837 | ||
838 | return 0; | |
839 | } | |
840 | ||
841 | /* Calculate the general cost of reading unhalted cycles (median) */ | |
842 | static int set_avg_unhalted_cycle_cost(const int core, uint64_t *value) | |
843 | { | |
844 | unsigned i; | |
845 | uint64_t cycles[10]; | |
846 | ||
847 | if (value == NULL || core_mask == 0 || core < 0) | |
848 | return 1; | |
849 | ||
850 | /* Fill cycles table with read cost values */ | |
851 | for (i = 0; i < DIM(cycles); i++) | |
852 | if (set_unhalted_cycle_cost(core, &cycles[i]) != 0) | |
853 | return 1; | |
854 | ||
855 | /* sort array */ | |
f67539c2 | 856 | qsort(cycles, DIM(cycles), sizeof(uint64_t), compare_uint64_t); |
11fdf7f2 TL |
857 | |
858 | /* set median cost */ | |
859 | *value = cycles[DIM(cycles)/2]; | |
860 | ||
861 | return 0; | |
862 | } | |
863 | ||
864 | /* Freeing allocated memory */ | |
f67539c2 | 865 | static void free_mem(uint8_t **p_buffer, uint128_t **p_keys) |
11fdf7f2 | 866 | { |
f67539c2 TL |
867 | uint128_t *keys = NULL; |
868 | uint8_t *buf = NULL; | |
869 | ||
870 | if (p_keys != NULL) { | |
871 | keys = *p_keys; | |
872 | *p_keys = NULL; | |
873 | } | |
874 | ||
875 | if (p_buffer != NULL) { | |
876 | buf = *p_buffer; | |
877 | *p_buffer = NULL; | |
878 | } | |
879 | ||
9f95a23c | 880 | #ifdef LINUX |
f67539c2 TL |
881 | if (keys != NULL) |
882 | free(keys); | |
9f95a23c | 883 | |
11fdf7f2 TL |
884 | if (buf != NULL) |
885 | free(buf); | |
9f95a23c | 886 | #else |
f67539c2 TL |
887 | if (keys != NULL) |
888 | _aligned_free(keys); | |
9f95a23c TL |
889 | |
890 | if (buf != NULL) | |
891 | _aligned_free(buf); | |
892 | #endif | |
11fdf7f2 TL |
893 | } |
894 | ||
f67539c2 TL |
895 | static const void * |
896 | get_key_pointer(const uint32_t index, const uint128_t *p_keys) | |
897 | { | |
898 | return (const void *) &p_keys[key_idxs[index]]; | |
899 | } | |
900 | ||
901 | static uint8_t *get_src_buffer(const uint32_t index, uint8_t *p_buffer) | |
902 | { | |
903 | return &p_buffer[offsets[index]]; | |
904 | } | |
905 | ||
906 | static uint8_t *get_dst_buffer(const uint32_t index, uint8_t *p_buffer) | |
11fdf7f2 | 907 | { |
f67539c2 TL |
908 | return &p_buffer[offsets[index] + sha_size_incr]; |
909 | } | |
910 | ||
911 | static uint32_t get_next_index(uint32_t index) | |
912 | { | |
913 | if (++index >= index_limit) | |
914 | index = 0; | |
915 | return index; | |
916 | } | |
917 | ||
918 | static void init_buf(void *pb, const size_t length) | |
919 | { | |
920 | const size_t n = length / sizeof(uint64_t); | |
921 | size_t i = 0; | |
922 | ||
923 | if (pb == NULL) | |
924 | return; | |
925 | ||
926 | for (i = 0; i < n; i++) | |
927 | ((uint64_t *)pb)[i] = (uint64_t) rand(); | |
928 | } | |
929 | ||
930 | /* | |
931 | * Packet and key memory allocation and initialization. | |
932 | * init_offsets() needs to be called prior to that so that | |
933 | * index_limit is set up accordingly to hot/cold selection. | |
934 | */ | |
935 | static void init_mem(uint8_t **p_buffer, uint128_t **p_keys) | |
936 | { | |
937 | const size_t bufs_size = index_limit * REGION_SIZE; | |
938 | const size_t keys_size = index_limit * KEYS_PER_JOB * sizeof(uint128_t); | |
9f95a23c | 939 | const size_t alignment = 64; |
f67539c2 TL |
940 | uint8_t *buf = NULL; |
941 | uint128_t *keys = NULL; | |
942 | ||
943 | if (p_keys == NULL || p_buffer == NULL) { | |
944 | fprintf(stderr, "Internal buffer allocation error!\n"); | |
945 | exit(EXIT_FAILURE); | |
946 | } | |
11fdf7f2 | 947 | |
9f95a23c TL |
948 | #ifdef LINUX |
949 | buf = (uint8_t *) memalign(alignment, bufs_size); | |
950 | #else | |
951 | buf = (uint8_t *) _aligned_malloc(bufs_size, alignment); | |
952 | #endif | |
11fdf7f2 TL |
953 | if (!buf) { |
954 | fprintf(stderr, "Could not malloc buf\n"); | |
955 | exit(EXIT_FAILURE); | |
956 | } | |
957 | ||
9f95a23c | 958 | #ifdef LINUX |
f67539c2 | 959 | keys = (uint128_t *) memalign(alignment, keys_size); |
9f95a23c | 960 | #else |
f67539c2 | 961 | keys = (uint128_t *) _aligned_malloc(keys_size, alignment); |
9f95a23c | 962 | #endif |
f67539c2 TL |
963 | if (!keys) { |
964 | fprintf(stderr, "Could not allocate memory for keys!\n"); | |
965 | free_mem(&buf, &keys); | |
11fdf7f2 TL |
966 | exit(EXIT_FAILURE); |
967 | } | |
968 | ||
f67539c2 TL |
969 | *p_keys = keys; |
970 | *p_buffer = buf; | |
11fdf7f2 | 971 | |
f67539c2 TL |
972 | init_buf(buf, bufs_size); |
973 | init_buf(keys, keys_size); | |
974 | } | |
975 | ||
976 | /* | |
977 | * Initialize packet buffer and keys offsets from | |
978 | * the start of the respective buffers | |
979 | */ | |
980 | static void init_offsets(const enum cache_type_e ctype) | |
981 | { | |
11fdf7f2 | 982 | if (ctype == COLD) { |
f67539c2 TL |
983 | uint32_t i; |
984 | ||
11fdf7f2 | 985 | for (i = 0; i < NUM_OFFSETS; i++) { |
f67539c2 | 986 | offsets[i] = (i * REGION_SIZE) + (rand() & 0x3C0); |
11fdf7f2 TL |
987 | key_idxs[i] = i * KEYS_PER_JOB; |
988 | } | |
9f95a23c | 989 | |
f67539c2 TL |
990 | /* swap the entries at random */ |
991 | for (i = 0; i < NUM_OFFSETS; i++) { | |
992 | const uint32_t swap_idx = (rand() % NUM_OFFSETS); | |
993 | const uint32_t tmp_offset = offsets[swap_idx]; | |
994 | const uint32_t tmp_keyidx = key_idxs[swap_idx]; | |
995 | ||
996 | offsets[swap_idx] = offsets[i]; | |
997 | key_idxs[swap_idx] = key_idxs[i]; | |
998 | offsets[i] = tmp_offset; | |
999 | key_idxs[i] = tmp_keyidx; | |
11fdf7f2 | 1000 | } |
f67539c2 | 1001 | |
11fdf7f2 | 1002 | index_limit = NUM_OFFSETS; |
f67539c2 TL |
1003 | } else { /* WARM */ |
1004 | uint32_t i; | |
1005 | ||
1006 | index_limit = 16; | |
1007 | ||
1008 | for (i = 0; i < index_limit; i++) { | |
1009 | /* | |
1010 | * Each buffer starts at different offset from | |
1011 | * start of the page. | |
1012 | * The most optimum determined difference between | |
1013 | * offsets is 4 cache lines. | |
1014 | */ | |
1015 | const uint32_t offset_step = (4 * 64); | |
1016 | const uint32_t L1_way_size = 4096; | |
1017 | ||
1018 | key_idxs[i] = i * KEYS_PER_JOB; | |
1019 | offsets[i] = i * REGION_SIZE + | |
1020 | ((i * offset_step) & (L1_way_size - 1)); | |
11fdf7f2 | 1021 | } |
11fdf7f2 TL |
1022 | } |
1023 | } | |
1024 | ||
f67539c2 TL |
1025 | /* |
1026 | * This function translates enum test_ciper_mode_e to be used by ipsec_mb | |
11fdf7f2 TL |
1027 | * library |
1028 | */ | |
f67539c2 TL |
1029 | static JOB_CIPHER_MODE |
1030 | translate_cipher_mode(const enum test_cipher_mode_e test_mode) | |
11fdf7f2 TL |
1031 | { |
1032 | JOB_CIPHER_MODE c_mode = NULL_CIPHER; | |
1033 | ||
1034 | switch (test_mode) { | |
1035 | case TEST_CBC: | |
1036 | c_mode = CBC; | |
1037 | break; | |
1038 | case TEST_CNTR: | |
1039 | case TEST_CNTR8: | |
1040 | c_mode = CNTR; | |
1041 | break; | |
f67539c2 TL |
1042 | case TEST_CNTR_BITLEN: |
1043 | case TEST_CNTR_BITLEN4: | |
1044 | c_mode = CNTR_BITLEN; | |
1045 | break; | |
1046 | case TEST_ECB: | |
1047 | c_mode = ECB; | |
1048 | break; | |
11fdf7f2 TL |
1049 | case TEST_NULL_CIPHER: |
1050 | c_mode = NULL_CIPHER; | |
1051 | break; | |
1052 | case TEST_AESDOCSIS: | |
1053 | case TEST_AESDOCSIS8: | |
1054 | c_mode = DOCSIS_SEC_BPI; | |
1055 | break; | |
1056 | case TEST_DESDOCSIS: | |
1057 | case TEST_DESDOCSIS4: | |
1058 | c_mode = DOCSIS_DES; | |
1059 | break; | |
1060 | case TEST_GCM: | |
1061 | c_mode = GCM; | |
1062 | break; | |
1063 | case TEST_CCM: | |
1064 | c_mode = CCM; | |
1065 | break; | |
1066 | case TEST_DES: | |
1067 | c_mode = DES; | |
1068 | break; | |
1069 | case TEST_3DES: | |
1070 | c_mode = DES3; | |
1071 | break; | |
f67539c2 TL |
1072 | case TEST_PON_CNTR: |
1073 | case TEST_PON_NO_CNTR: | |
1074 | c_mode = PON_AES_CNTR; | |
1075 | break; | |
11fdf7f2 TL |
1076 | default: |
1077 | break; | |
1078 | } | |
1079 | return c_mode; | |
1080 | } | |
1081 | ||
1082 | /* Performs test using AES_HMAC or DOCSIS */ | |
1083 | static uint64_t | |
9f95a23c | 1084 | do_test(MB_MGR *mb_mgr, struct params_s *params, |
f67539c2 | 1085 | const uint32_t num_iter, uint8_t *p_buffer, uint128_t *p_keys) |
11fdf7f2 TL |
1086 | { |
1087 | JOB_AES_HMAC *job; | |
1088 | JOB_AES_HMAC job_template; | |
1089 | uint32_t i; | |
1090 | static uint32_t index = 0; | |
1091 | static DECLARE_ALIGNED(uint128_t iv, 16); | |
1092 | static uint32_t ipad[5], opad[5], digest[3]; | |
1093 | static DECLARE_ALIGNED(uint32_t k1_expanded[11 * 4], 16); | |
1094 | static DECLARE_ALIGNED(uint8_t k2[16], 16); | |
1095 | static DECLARE_ALIGNED(uint8_t k3[16], 16); | |
9f95a23c | 1096 | static DECLARE_ALIGNED(struct gcm_key_data gdata_key, 512); |
f67539c2 | 1097 | uint64_t xgem_hdr = 0; |
11fdf7f2 TL |
1098 | uint32_t size_aes; |
1099 | uint64_t time = 0; | |
1100 | uint32_t aux; | |
1101 | ||
1102 | if ((params->cipher_mode == TEST_AESDOCSIS8) || | |
1103 | (params->cipher_mode == TEST_CNTR8)) | |
1104 | size_aes = params->size_aes + 8; | |
1105 | else if (params->cipher_mode == TEST_DESDOCSIS4) | |
1106 | size_aes = params->size_aes + 4; | |
1107 | else | |
1108 | size_aes = params->size_aes; | |
1109 | ||
f67539c2 TL |
1110 | if (params->cipher_mode == TEST_CNTR_BITLEN) |
1111 | job_template.msg_len_to_cipher_in_bits = size_aes * 8; | |
1112 | else if (params->cipher_mode == TEST_CNTR_BITLEN4) | |
1113 | job_template.msg_len_to_cipher_in_bits = size_aes * 8 - 4; | |
1114 | else | |
1115 | job_template.msg_len_to_cipher_in_bytes = size_aes; | |
1116 | ||
11fdf7f2 TL |
1117 | job_template.msg_len_to_hash_in_bytes = size_aes + sha_size_incr; |
1118 | job_template.hash_start_src_offset_in_bytes = 0; | |
1119 | job_template.cipher_start_src_offset_in_bytes = sha_size_incr; | |
1120 | job_template.iv = (uint8_t *) &iv; | |
1121 | job_template.iv_len_in_bytes = 16; | |
1122 | ||
1123 | job_template.auth_tag_output = (uint8_t *) digest; | |
1124 | ||
1125 | switch (params->hash_alg) { | |
1126 | case TEST_XCBC: | |
1127 | job_template.u.XCBC._k1_expanded = k1_expanded; | |
1128 | job_template.u.XCBC._k2 = k2; | |
1129 | job_template.u.XCBC._k3 = k3; | |
1130 | job_template.hash_alg = AES_XCBC; | |
1131 | break; | |
1132 | case TEST_HASH_CCM: | |
1133 | job_template.hash_alg = AES_CCM; | |
1134 | break; | |
1135 | case TEST_HASH_GCM: | |
1136 | job_template.hash_alg = AES_GMAC; | |
1137 | break; | |
1138 | case TEST_NULL_HASH: | |
1139 | job_template.hash_alg = NULL_HASH; | |
1140 | break; | |
1141 | case TEST_HASH_CMAC: | |
1142 | job_template.u.CMAC._key_expanded = k1_expanded; | |
1143 | job_template.u.CMAC._skey1 = k2; | |
1144 | job_template.u.CMAC._skey2 = k3; | |
1145 | job_template.hash_alg = AES_CMAC; | |
1146 | break; | |
f67539c2 TL |
1147 | case TEST_HASH_CMAC_BITLEN: |
1148 | job_template.u.CMAC._key_expanded = k1_expanded; | |
1149 | job_template.u.CMAC._skey1 = k2; | |
1150 | job_template.u.CMAC._skey2 = k3; | |
1151 | /* | |
1152 | * CMAC bit level version is done in bits (length is | |
1153 | * converted to bits and it is decreased by 4 bits, | |
1154 | * to force the CMAC bitlen path) | |
1155 | */ | |
1156 | job_template.msg_len_to_hash_in_bits = | |
1157 | (job_template.msg_len_to_hash_in_bytes * 8) - 4; | |
1158 | job_template.hash_alg = AES_CMAC_BITLEN; | |
1159 | break; | |
1160 | case TEST_PON_CRC_BIP: | |
1161 | job_template.hash_alg = PON_CRC_BIP; | |
1162 | job_template.msg_len_to_hash_in_bytes = size_aes + 8; | |
1163 | job_template.cipher_start_src_offset_in_bytes = 8; | |
1164 | if (params->cipher_mode == TEST_PON_NO_CNTR) | |
1165 | job_template.msg_len_to_cipher_in_bytes = 0; | |
1166 | break; | |
11fdf7f2 TL |
1167 | default: |
1168 | /* HMAC hash alg is SHA1 or MD5 */ | |
1169 | job_template.u.HMAC._hashed_auth_key_xor_ipad = | |
1170 | (uint8_t *) ipad; | |
1171 | job_template.u.HMAC._hashed_auth_key_xor_opad = | |
1172 | (uint8_t *) opad; | |
1173 | job_template.hash_alg = (JOB_HASH_ALG) params->hash_alg; | |
1174 | break; | |
1175 | } | |
1176 | job_template.auth_tag_output_len_in_bytes = | |
1177 | (uint64_t) auth_tag_length_bytes[job_template.hash_alg - 1]; | |
1178 | ||
1179 | job_template.cipher_direction = params->cipher_dir; | |
1180 | ||
1181 | if (params->cipher_mode == TEST_NULL_CIPHER) { | |
1182 | job_template.chain_order = HASH_CIPHER; | |
f67539c2 TL |
1183 | } else if (params->cipher_mode == TEST_CCM) { |
1184 | if (job_template.cipher_direction == ENCRYPT) | |
1185 | job_template.chain_order = HASH_CIPHER; | |
1186 | else | |
1187 | job_template.chain_order = CIPHER_HASH; | |
11fdf7f2 TL |
1188 | } else { |
1189 | if (job_template.cipher_direction == ENCRYPT) | |
1190 | job_template.chain_order = CIPHER_HASH; | |
1191 | else | |
1192 | job_template.chain_order = HASH_CIPHER; | |
1193 | } | |
1194 | ||
1195 | /* Translating enum to the API's one */ | |
1196 | job_template.cipher_mode = translate_cipher_mode(params->cipher_mode); | |
1197 | job_template.aes_key_len_in_bytes = params->aes_key_size; | |
1198 | if (job_template.cipher_mode == GCM) { | |
1199 | uint8_t key[32]; | |
1200 | ||
9f95a23c TL |
1201 | switch (params->aes_key_size) { |
1202 | case AES_128_BYTES: | |
1203 | IMB_AES128_GCM_PRE(mb_mgr, key, &gdata_key); | |
1204 | break; | |
1205 | case AES_192_BYTES: | |
1206 | IMB_AES192_GCM_PRE(mb_mgr, key, &gdata_key); | |
1207 | break; | |
1208 | case AES_256_BYTES: | |
1209 | default: | |
1210 | IMB_AES256_GCM_PRE(mb_mgr, key, &gdata_key); | |
1211 | break; | |
1212 | } | |
11fdf7f2 TL |
1213 | job_template.aes_enc_key_expanded = &gdata_key; |
1214 | job_template.aes_dec_key_expanded = &gdata_key; | |
f67539c2 | 1215 | job_template.u.GCM.aad_len_in_bytes = params->aad_size; |
11fdf7f2 TL |
1216 | job_template.iv_len_in_bytes = 12; |
1217 | } else if (job_template.cipher_mode == CCM) { | |
1218 | job_template.msg_len_to_cipher_in_bytes = size_aes; | |
1219 | job_template.msg_len_to_hash_in_bytes = size_aes; | |
1220 | job_template.hash_start_src_offset_in_bytes = 0; | |
1221 | job_template.cipher_start_src_offset_in_bytes = 0; | |
f67539c2 | 1222 | job_template.u.CCM.aad_len_in_bytes = params->aad_size; |
11fdf7f2 TL |
1223 | job_template.iv_len_in_bytes = 13; |
1224 | } else if (job_template.cipher_mode == DES || | |
11fdf7f2 TL |
1225 | job_template.cipher_mode == DOCSIS_DES) { |
1226 | job_template.aes_key_len_in_bytes = 8; | |
1227 | job_template.iv_len_in_bytes = 8; | |
9f95a23c TL |
1228 | } else if (job_template.cipher_mode == DES3) { |
1229 | job_template.aes_key_len_in_bytes = 24; | |
1230 | job_template.iv_len_in_bytes = 8; | |
11fdf7f2 TL |
1231 | } |
1232 | ||
f67539c2 TL |
1233 | |
1234 | if (job_template.hash_alg == PON_CRC_BIP) { | |
1235 | /* create XGEM header template */ | |
1236 | const uint64_t pli = | |
1237 | (job_template.msg_len_to_cipher_in_bytes << 2) & 0xffff; | |
1238 | ||
1239 | xgem_hdr = ((pli >> 8) & 0xff) | ((pli & 0xff) << 8); | |
1240 | } | |
1241 | ||
11fdf7f2 TL |
1242 | #ifndef _WIN32 |
1243 | if (use_unhalted_cycles) | |
1244 | time = read_cycles(params->core); | |
1245 | else | |
1246 | #endif | |
1247 | time = __rdtscp(&aux); | |
1248 | ||
1249 | for (i = 0; i < num_iter; i++) { | |
9f95a23c | 1250 | job = IMB_GET_NEXT_JOB(mb_mgr); |
11fdf7f2 TL |
1251 | *job = job_template; |
1252 | ||
f67539c2 TL |
1253 | if (job->hash_alg == PON_CRC_BIP) { |
1254 | uint64_t *p_src = | |
1255 | (uint64_t *) get_src_buffer(index, p_buffer); | |
1256 | ||
1257 | job->src = (const uint8_t *)p_src; | |
1258 | p_src[0] = xgem_hdr; | |
1259 | } else { | |
1260 | job->src = get_src_buffer(index, p_buffer); | |
1261 | } | |
1262 | job->dst = get_dst_buffer(index, p_buffer); | |
11fdf7f2 TL |
1263 | if (job->cipher_mode == GCM) { |
1264 | job->u.GCM.aad = job->src; | |
1265 | } else if (job->cipher_mode == CCM) { | |
1266 | job->u.CCM.aad = job->src; | |
1267 | job->aes_enc_key_expanded = job->aes_dec_key_expanded = | |
f67539c2 TL |
1268 | (const uint32_t *) get_key_pointer(index, |
1269 | p_keys); | |
11fdf7f2 TL |
1270 | } else if (job->cipher_mode == DES3) { |
1271 | static const void *ks_ptr[3]; | |
1272 | ||
1273 | ks_ptr[0] = ks_ptr[1] = ks_ptr[2] = | |
f67539c2 | 1274 | get_key_pointer(index, p_keys); |
11fdf7f2 TL |
1275 | job->aes_enc_key_expanded = |
1276 | job->aes_dec_key_expanded = ks_ptr; | |
1277 | } else { | |
1278 | job->aes_enc_key_expanded = job->aes_dec_key_expanded = | |
f67539c2 TL |
1279 | (const uint32_t *) get_key_pointer(index, |
1280 | p_keys); | |
11fdf7f2 TL |
1281 | } |
1282 | ||
f67539c2 | 1283 | index = get_next_index(index); |
9f95a23c TL |
1284 | #ifdef DEBUG |
1285 | job = IMB_SUBMIT_JOB(mb_mgr); | |
1286 | #else | |
1287 | job = IMB_SUBMIT_JOB_NOCHECK(mb_mgr); | |
1288 | #endif | |
11fdf7f2 TL |
1289 | while (job) { |
1290 | #ifdef DEBUG | |
1291 | if (job->status != STS_COMPLETED) | |
1292 | fprintf(stderr, "failed job, status:%d\n", | |
1293 | job->status); | |
1294 | #endif | |
9f95a23c | 1295 | job = IMB_GET_COMPLETED_JOB(mb_mgr); |
11fdf7f2 TL |
1296 | } |
1297 | } | |
1298 | ||
9f95a23c | 1299 | while ((job = IMB_FLUSH_JOB(mb_mgr))) { |
11fdf7f2 TL |
1300 | #ifdef DEBUG |
1301 | if (job->status != STS_COMPLETED) | |
1302 | fprintf(stderr, "failed job, status:%d\n", job->status); | |
1303 | #endif | |
1304 | } | |
1305 | ||
1306 | #ifndef _WIN32 | |
1307 | if (use_unhalted_cycles) | |
1308 | time = (read_cycles(params->core) - rd_cycles_cost) - time; | |
1309 | else | |
1310 | #endif | |
1311 | time = __rdtscp(&aux) - time; | |
1312 | ||
1313 | return time / num_iter; | |
1314 | } | |
1315 | ||
1316 | /* Performs test using GCM */ | |
1317 | static uint64_t | |
9f95a23c | 1318 | do_test_gcm(struct params_s *params, |
f67539c2 TL |
1319 | const uint32_t num_iter, MB_MGR *mb_mgr, |
1320 | uint8_t *p_buffer, uint128_t *p_keys) | |
11fdf7f2 | 1321 | { |
9f95a23c TL |
1322 | static DECLARE_ALIGNED(struct gcm_key_data gdata_key, 512); |
1323 | static DECLARE_ALIGNED(struct gcm_context_data gdata_ctx, 64); | |
11fdf7f2 TL |
1324 | uint8_t *key; |
1325 | static uint32_t index = 0; | |
11fdf7f2 TL |
1326 | uint32_t size_aes = params->size_aes; |
1327 | uint32_t i; | |
f67539c2 | 1328 | uint8_t *aad = NULL; |
11fdf7f2 TL |
1329 | uint8_t auth_tag[12]; |
1330 | DECLARE_ALIGNED(uint8_t iv[16], 16); | |
1331 | uint64_t time = 0; | |
1332 | uint32_t aux; | |
1333 | ||
1334 | key = (uint8_t *) malloc(sizeof(uint8_t) * params->aes_key_size); | |
1335 | if (!key) { | |
1336 | fprintf(stderr, "Could not malloc key\n"); | |
f67539c2 TL |
1337 | free_mem(&p_buffer, &p_keys); |
1338 | exit(EXIT_FAILURE); | |
1339 | } | |
1340 | ||
1341 | aad = (uint8_t *) malloc(sizeof(uint8_t) * params->aad_size); | |
1342 | if (!aad) { | |
1343 | free(key); | |
1344 | fprintf(stderr, "Could not malloc AAD\n"); | |
1345 | free_mem(&p_buffer, &p_keys); | |
11fdf7f2 TL |
1346 | exit(EXIT_FAILURE); |
1347 | } | |
1348 | ||
9f95a23c TL |
1349 | switch (params->aes_key_size) { |
1350 | case AES_128_BYTES: | |
1351 | IMB_AES128_GCM_PRE(mb_mgr, key, &gdata_key); | |
1352 | break; | |
1353 | case AES_192_BYTES: | |
1354 | IMB_AES192_GCM_PRE(mb_mgr, key, &gdata_key); | |
1355 | break; | |
1356 | case AES_256_BYTES: | |
1357 | default: | |
1358 | IMB_AES256_GCM_PRE(mb_mgr, key, &gdata_key); | |
1359 | break; | |
1360 | } | |
1361 | ||
11fdf7f2 TL |
1362 | if (params->cipher_dir == ENCRYPT) { |
1363 | #ifndef _WIN32 | |
1364 | if (use_unhalted_cycles) | |
1365 | time = read_cycles(params->core); | |
1366 | else | |
1367 | #endif | |
1368 | time = __rdtscp(&aux); | |
1369 | ||
9f95a23c TL |
1370 | if (params->aes_key_size == AES_128_BYTES) { |
1371 | for (i = 0; i < num_iter; i++) { | |
f67539c2 TL |
1372 | uint8_t *pb = get_dst_buffer(index, p_buffer); |
1373 | ||
9f95a23c TL |
1374 | IMB_AES128_GCM_ENC(mb_mgr, &gdata_key, |
1375 | &gdata_ctx, | |
f67539c2 TL |
1376 | pb, |
1377 | pb, | |
9f95a23c | 1378 | size_aes, iv, |
f67539c2 | 1379 | aad, params->aad_size, |
9f95a23c | 1380 | auth_tag, sizeof(auth_tag)); |
f67539c2 | 1381 | index = get_next_index(index); |
9f95a23c TL |
1382 | } |
1383 | } else if (params->aes_key_size == AES_192_BYTES) { | |
1384 | for (i = 0; i < num_iter; i++) { | |
f67539c2 TL |
1385 | uint8_t *pb = get_dst_buffer(index, p_buffer); |
1386 | ||
9f95a23c TL |
1387 | IMB_AES192_GCM_ENC(mb_mgr, &gdata_key, |
1388 | &gdata_ctx, | |
f67539c2 TL |
1389 | pb, |
1390 | pb, | |
9f95a23c | 1391 | size_aes, iv, |
f67539c2 | 1392 | aad, params->aad_size, |
9f95a23c | 1393 | auth_tag, sizeof(auth_tag)); |
f67539c2 | 1394 | index = get_next_index(index); |
9f95a23c TL |
1395 | } |
1396 | } else { /* 256 */ | |
1397 | for (i = 0; i < num_iter; i++) { | |
f67539c2 TL |
1398 | uint8_t *pb = get_dst_buffer(index, p_buffer); |
1399 | ||
9f95a23c TL |
1400 | IMB_AES256_GCM_ENC(mb_mgr, &gdata_key, |
1401 | &gdata_ctx, | |
f67539c2 TL |
1402 | pb, |
1403 | pb, | |
9f95a23c | 1404 | size_aes, iv, |
f67539c2 | 1405 | aad, params->aad_size, |
9f95a23c | 1406 | auth_tag, sizeof(auth_tag)); |
f67539c2 | 1407 | index = get_next_index(index); |
9f95a23c | 1408 | } |
11fdf7f2 TL |
1409 | } |
1410 | #ifndef _WIN32 | |
1411 | if (use_unhalted_cycles) | |
1412 | time = (read_cycles(params->core) - | |
1413 | rd_cycles_cost) - time; | |
1414 | else | |
1415 | #endif | |
1416 | time = __rdtscp(&aux) - time; | |
1417 | } else { /*DECRYPT*/ | |
1418 | #ifndef _WIN32 | |
1419 | if (use_unhalted_cycles) | |
1420 | time = read_cycles(params->core); | |
1421 | else | |
1422 | #endif | |
1423 | time = __rdtscp(&aux); | |
1424 | ||
9f95a23c TL |
1425 | if (params->aes_key_size == AES_128_BYTES) { |
1426 | for (i = 0; i < num_iter; i++) { | |
f67539c2 TL |
1427 | uint8_t *pb = get_dst_buffer(index, p_buffer); |
1428 | ||
9f95a23c TL |
1429 | IMB_AES128_GCM_DEC(mb_mgr, &gdata_key, |
1430 | &gdata_ctx, | |
f67539c2 TL |
1431 | pb, |
1432 | pb, | |
9f95a23c | 1433 | size_aes, iv, |
f67539c2 | 1434 | aad, params->aad_size, |
9f95a23c | 1435 | auth_tag, sizeof(auth_tag)); |
f67539c2 | 1436 | index = get_next_index(index); |
9f95a23c TL |
1437 | } |
1438 | } else if (params->aes_key_size == AES_192_BYTES) { | |
1439 | for (i = 0; i < num_iter; i++) { | |
f67539c2 TL |
1440 | uint8_t *pb = get_dst_buffer(index, p_buffer); |
1441 | ||
9f95a23c TL |
1442 | IMB_AES192_GCM_DEC(mb_mgr, &gdata_key, |
1443 | &gdata_ctx, | |
f67539c2 TL |
1444 | pb, |
1445 | pb, | |
9f95a23c | 1446 | size_aes, iv, |
f67539c2 | 1447 | aad, params->aad_size, |
9f95a23c | 1448 | auth_tag, sizeof(auth_tag)); |
f67539c2 | 1449 | index = get_next_index(index); |
9f95a23c TL |
1450 | } |
1451 | } else { /* 256 */ | |
1452 | for (i = 0; i < num_iter; i++) { | |
f67539c2 TL |
1453 | uint8_t *pb = get_dst_buffer(index, p_buffer); |
1454 | ||
9f95a23c TL |
1455 | IMB_AES256_GCM_DEC(mb_mgr, &gdata_key, |
1456 | &gdata_ctx, | |
f67539c2 TL |
1457 | pb, |
1458 | pb, | |
9f95a23c | 1459 | size_aes, iv, |
f67539c2 | 1460 | aad, params->aad_size, |
9f95a23c | 1461 | auth_tag, sizeof(auth_tag)); |
f67539c2 | 1462 | index = get_next_index(index); |
9f95a23c | 1463 | } |
11fdf7f2 TL |
1464 | } |
1465 | #ifndef _WIN32 | |
1466 | if (use_unhalted_cycles) | |
1467 | time = (read_cycles(params->core) - | |
1468 | rd_cycles_cost) - time; | |
1469 | else | |
1470 | #endif | |
1471 | time = __rdtscp(&aux) - time; | |
1472 | } | |
1473 | ||
1474 | free(key); | |
f67539c2 | 1475 | free(aad); |
11fdf7f2 TL |
1476 | return time / num_iter; |
1477 | } | |
1478 | ||
11fdf7f2 | 1479 | /* Computes mean of set of times after dropping bottom and top quarters */ |
f67539c2 TL |
1480 | static uint64_t |
1481 | mean_median(uint64_t *array, uint32_t size, | |
1482 | uint8_t *p_buffer, uint128_t *p_keys) | |
11fdf7f2 | 1483 | { |
f67539c2 | 1484 | const uint32_t quarter = size / 4; |
11fdf7f2 TL |
1485 | uint32_t i; |
1486 | uint64_t sum; | |
1487 | ||
1488 | /* these are single threaded runs, so we skip | |
1489 | * the hardware thread related skew clipping | |
1490 | * thus skipping "ignore first and last eighth" | |
1491 | */ | |
1492 | ||
1493 | /* ignore lowest and highest quarter */ | |
1494 | qsort(array, size, sizeof(uint64_t), compare_uint64_t); | |
1495 | ||
1496 | /* dropping the bottom and top quarters | |
1497 | * after sorting to remove noise/variations | |
1498 | */ | |
1499 | array += quarter; | |
1500 | size -= quarter * 2; | |
1501 | ||
1502 | ||
1503 | if ((size == 0) || (size & 0x80000000)) { | |
f67539c2 TL |
1504 | fprintf(stderr, "Not enough data points!\n"); |
1505 | free_mem(&p_buffer, &p_keys); | |
11fdf7f2 TL |
1506 | exit(EXIT_FAILURE); |
1507 | } | |
1508 | sum = 0; | |
1509 | for (i = 0; i < size; i++) | |
1510 | sum += array[i]; | |
1511 | ||
1512 | sum = (sum + size / 2) / size; | |
1513 | return sum; | |
1514 | } | |
1515 | ||
1516 | /* Runs test for each buffer size and stores averaged execution time */ | |
1517 | static void | |
1518 | process_variant(MB_MGR *mgr, const uint32_t arch, struct params_s *params, | |
f67539c2 TL |
1519 | struct variant_s *variant_ptr, const uint32_t run, |
1520 | uint8_t *p_buffer, uint128_t *p_keys) | |
11fdf7f2 TL |
1521 | { |
1522 | const uint32_t sizes = params->num_sizes; | |
1523 | uint64_t *times = &variant_ptr->avg_times[run]; | |
1524 | uint32_t sz; | |
1525 | ||
1526 | for (sz = 0; sz < sizes; sz++) { | |
f67539c2 TL |
1527 | const uint32_t size_aes = job_sizes[RANGE_MIN] + |
1528 | (sz * job_sizes[RANGE_STEP]); | |
1529 | uint32_t num_iter; | |
1530 | ||
1531 | params->aad_size = 0; | |
1532 | if (params->cipher_mode == TEST_GCM) | |
1533 | params->aad_size = gcm_aad_size; | |
1534 | ||
1535 | if (params->cipher_mode == TEST_CCM) | |
1536 | params->aad_size = ccm_aad_size; | |
1537 | ||
1538 | /* | |
1539 | * If job size == 0, check AAD size | |
1540 | * (only allowed for GCM/CCM) | |
1541 | */ | |
1542 | if (size_aes == 0 && params->aad_size != 0) | |
1543 | num_iter = (iter_scale >= (uint32_t)params->aad_size) ? | |
1544 | (iter_scale / (uint32_t)params->aad_size) : | |
1545 | 1; | |
1546 | else if (size_aes != 0) | |
1547 | num_iter = (iter_scale >= size_aes) ? | |
1548 | (iter_scale / size_aes) : 1; | |
1549 | else | |
1550 | num_iter = iter_scale; | |
11fdf7f2 TL |
1551 | |
1552 | params->size_aes = size_aes; | |
f67539c2 TL |
1553 | if (params->cipher_mode == TEST_GCM && (!use_gcm_job_api)) { |
1554 | if (job_iter == 0) | |
1555 | *times = do_test_gcm(params, 2 * num_iter, mgr, | |
1556 | p_buffer, p_keys); | |
1557 | else | |
1558 | *times = do_test_gcm(params, job_iter, mgr, | |
1559 | p_buffer, p_keys); | |
1560 | } else { | |
1561 | if (job_iter == 0) | |
1562 | *times = do_test(mgr, params, num_iter, | |
1563 | p_buffer, p_keys); | |
1564 | else | |
1565 | *times = do_test(mgr, params, job_iter, | |
1566 | p_buffer, p_keys); | |
1567 | } | |
11fdf7f2 TL |
1568 | times += NUM_RUNS; |
1569 | } | |
1570 | ||
1571 | variant_ptr->params = *params; | |
1572 | variant_ptr->arch = arch; | |
1573 | } | |
1574 | ||
1575 | /* Sets cipher mode, hash algorithm */ | |
1576 | static void | |
1577 | do_variants(MB_MGR *mgr, const uint32_t arch, struct params_s *params, | |
1578 | const uint32_t run, struct variant_s **variant_ptr, | |
f67539c2 TL |
1579 | uint32_t *variant, uint8_t *p_buffer, uint128_t *p_keys, |
1580 | const int print_info) | |
11fdf7f2 TL |
1581 | { |
1582 | uint32_t hash_alg; | |
1583 | uint32_t h_start = TEST_SHA1; | |
1584 | uint32_t h_end = TEST_NULL_HASH; | |
1585 | uint32_t c_mode; | |
1586 | uint32_t c_start = TEST_CBC; | |
1587 | uint32_t c_end = TEST_NULL_CIPHER; | |
1588 | ||
1589 | switch (params->test_type) { | |
1590 | case TTYPE_AES_DOCSIS: | |
1591 | h_start = TEST_NULL_HASH; | |
1592 | c_start = TEST_AESDOCSIS; | |
1593 | c_end = TEST_DESDOCSIS4; | |
1594 | break; | |
1595 | case TTYPE_AES_GCM: | |
1596 | h_start = TEST_HASH_GCM; | |
1597 | h_end = TEST_HASH_GCM; | |
1598 | c_start = TEST_GCM; | |
1599 | c_end = TEST_GCM; | |
1600 | break; | |
1601 | case TTYPE_AES_CCM: | |
1602 | h_start = TEST_HASH_CCM; | |
1603 | h_end = TEST_HASH_CCM; | |
1604 | c_start = TEST_CCM; | |
1605 | c_end = TEST_CCM; | |
1606 | break; | |
1607 | case TTYPE_AES_DES: | |
1608 | h_start = TEST_NULL_HASH; | |
1609 | h_end = TEST_NULL_HASH; | |
1610 | c_start = TEST_DES; | |
1611 | c_end = TEST_DES; | |
1612 | break; | |
1613 | case TTYPE_AES_3DES: | |
1614 | h_start = TEST_NULL_HASH; | |
1615 | h_end = TEST_NULL_HASH; | |
1616 | c_start = TEST_3DES; | |
1617 | c_end = TEST_3DES; | |
1618 | break; | |
f67539c2 TL |
1619 | case TTYPE_PON: |
1620 | h_start = TEST_PON_CRC_BIP; | |
1621 | h_end = TEST_PON_CRC_BIP; | |
1622 | c_start = TEST_PON_CNTR; | |
1623 | c_end = TEST_PON_NO_CNTR; | |
1624 | break; | |
1625 | case TTYPE_CUSTOM: | |
1626 | h_start = params->hash_alg; | |
1627 | h_end = params->hash_alg; | |
1628 | c_start = params->cipher_mode; | |
1629 | c_end = params->cipher_mode; | |
1630 | break; | |
11fdf7f2 TL |
1631 | default: |
1632 | break; | |
1633 | } | |
1634 | ||
1635 | for (c_mode = c_start; c_mode <= c_end; c_mode++) { | |
1636 | params->cipher_mode = (enum test_cipher_mode_e) c_mode; | |
1637 | for (hash_alg = h_start; hash_alg <= h_end; hash_alg++) { | |
1638 | params->hash_alg = (enum test_hash_alg_e) hash_alg; | |
f67539c2 TL |
1639 | process_variant(mgr, arch, params, *variant_ptr, run, |
1640 | p_buffer, p_keys); | |
1641 | /* update and print progress bar */ | |
1642 | if (print_info) | |
1643 | prog_bar_update(*variant); | |
11fdf7f2 TL |
1644 | (*variant)++; |
1645 | (*variant_ptr)++; | |
1646 | } | |
1647 | } | |
1648 | } | |
1649 | ||
1650 | /* Sets cipher direction and key size */ | |
1651 | static void | |
1652 | run_dir_test(MB_MGR *mgr, const uint32_t arch, struct params_s *params, | |
1653 | const uint32_t run, struct variant_s **variant_ptr, | |
f67539c2 TL |
1654 | uint32_t *variant, uint8_t *p_buffer, uint128_t *p_keys, |
1655 | const int print_info) | |
11fdf7f2 TL |
1656 | { |
1657 | uint32_t dir; | |
1658 | uint32_t k; /* Key size */ | |
1659 | uint32_t limit = AES_256_BYTES; /* Key size value limit */ | |
1660 | ||
1661 | if (params->test_type == TTYPE_AES_DOCSIS || | |
1662 | params->test_type == TTYPE_AES_DES || | |
1663 | params->test_type == TTYPE_AES_3DES || | |
f67539c2 | 1664 | params->test_type == TTYPE_PON || |
11fdf7f2 TL |
1665 | params->test_type == TTYPE_AES_CCM) |
1666 | limit = AES_128_BYTES; | |
1667 | ||
9f95a23c TL |
1668 | switch (arch) { |
1669 | case 0: | |
1670 | init_mb_mgr_sse(mgr); | |
1671 | break; | |
1672 | case 1: | |
1673 | init_mb_mgr_avx(mgr); | |
1674 | break; | |
1675 | case 2: | |
1676 | init_mb_mgr_avx2(mgr); | |
1677 | break; | |
1678 | default: | |
1679 | case 3: | |
1680 | init_mb_mgr_avx512(mgr); | |
1681 | break; | |
1682 | } | |
11fdf7f2 | 1683 | |
f67539c2 TL |
1684 | if (params->test_type == TTYPE_CUSTOM) { |
1685 | params->cipher_dir = custom_job_params.cipher_dir; | |
1686 | params->aes_key_size = custom_job_params.aes_key_size; | |
1687 | params->cipher_mode = custom_job_params.cipher_mode; | |
1688 | params->hash_alg = custom_job_params.hash_alg; | |
1689 | do_variants(mgr, arch, params, run, variant_ptr, | |
1690 | variant, p_buffer, p_keys, print_info); | |
1691 | return; | |
1692 | } | |
1693 | ||
11fdf7f2 TL |
1694 | for (dir = ENCRYPT; dir <= DECRYPT; dir++) { |
1695 | params->cipher_dir = (JOB_CIPHER_DIRECTION) dir; | |
1696 | for (k = AES_128_BYTES; k <= limit; k += 8) { | |
1697 | params->aes_key_size = k; | |
1698 | do_variants(mgr, arch, params, run, variant_ptr, | |
f67539c2 | 1699 | variant, p_buffer, p_keys, print_info); |
11fdf7f2 TL |
1700 | } |
1701 | } | |
1702 | } | |
1703 | ||
1704 | /* Generates output containing averaged times for each test variant */ | |
f67539c2 TL |
1705 | static void |
1706 | print_times(struct variant_s *variant_list, struct params_s *params, | |
1707 | const uint32_t total_variants, uint8_t *p_buffer, uint128_t *p_keys) | |
11fdf7f2 TL |
1708 | { |
1709 | const uint32_t sizes = params->num_sizes; | |
1710 | uint32_t col; | |
1711 | uint32_t sz; | |
1712 | ||
1713 | /* Temporary variables */ | |
1714 | struct params_s par; | |
1715 | uint8_t c_mode; | |
1716 | uint8_t c_dir; | |
1717 | uint8_t h_alg; | |
1718 | const char *func_names[4] = { | |
1719 | "SSE", "AVX", "AVX2", "AVX512" | |
1720 | }; | |
f67539c2 TL |
1721 | const char *c_mode_names[TEST_NUM_CIPHER_TESTS - 1] = { |
1722 | "CBC", "CNTR", "CNTR+8", "CNTR_BITLEN", "CNTR_BITLEN4", "ECB", | |
1723 | "NULL_CIPHER", "DOCAES", "DOCAES+8", "DOCDES", "DOCDES+4", | |
1724 | "GCM", "CCM", "DES", "3DES", "PON", "PON_NO_CTR" | |
11fdf7f2 TL |
1725 | }; |
1726 | const char *c_dir_names[2] = { | |
1727 | "ENCRYPT", "DECRYPT" | |
1728 | }; | |
f67539c2 | 1729 | const char *h_alg_names[TEST_NUM_HASH_TESTS - 1] = { |
11fdf7f2 | 1730 | "SHA1", "SHA_224", "SHA_256", "SHA_384", "SHA_512", "XCBC", |
f67539c2 TL |
1731 | "MD5", "CMAC", "CMAC_BITLEN", "NULL_HASH", "GCM", "CUSTOM", |
1732 | "CCM", "BIP-CRC32" | |
11fdf7f2 TL |
1733 | }; |
1734 | printf("ARCH"); | |
1735 | for (col = 0; col < total_variants; col++) | |
1736 | printf("\t%s", func_names[variant_list[col].arch]); | |
1737 | printf("\n"); | |
1738 | printf("CIPHER"); | |
1739 | for (col = 0; col < total_variants; col++) { | |
1740 | par = variant_list[col].params; | |
f67539c2 | 1741 | c_mode = par.cipher_mode - TEST_CBC; |
11fdf7f2 TL |
1742 | printf("\t%s", c_mode_names[c_mode]); |
1743 | } | |
1744 | printf("\n"); | |
1745 | printf("DIR"); | |
1746 | for (col = 0; col < total_variants; col++) { | |
1747 | par = variant_list[col].params; | |
1748 | c_dir = par.cipher_dir - ENCRYPT; | |
1749 | printf("\t%s", c_dir_names[c_dir]); | |
1750 | } | |
1751 | printf("\n"); | |
1752 | printf("HASH_ALG"); | |
1753 | for (col = 0; col < total_variants; col++) { | |
1754 | par = variant_list[col].params; | |
f67539c2 | 1755 | h_alg = par.hash_alg - TEST_SHA1; |
11fdf7f2 TL |
1756 | printf("\t%s", h_alg_names[h_alg]); |
1757 | } | |
1758 | printf("\n"); | |
1759 | printf("KEY_SIZE"); | |
1760 | for (col = 0; col < total_variants; col++) { | |
1761 | par = variant_list[col].params; | |
1762 | printf("\tAES-%u", par.aes_key_size * 8); | |
1763 | } | |
1764 | printf("\n"); | |
1765 | for (sz = 0; sz < sizes; sz++) { | |
f67539c2 TL |
1766 | printf("%d", job_sizes[RANGE_MIN] + |
1767 | (sz * job_sizes[RANGE_STEP])); | |
11fdf7f2 TL |
1768 | for (col = 0; col < total_variants; col++) { |
1769 | uint64_t *time_ptr = | |
1770 | &variant_list[col].avg_times[sz * NUM_RUNS]; | |
1771 | const unsigned long long val = | |
f67539c2 TL |
1772 | mean_median(time_ptr, NUM_RUNS, |
1773 | p_buffer, p_keys); | |
11fdf7f2 TL |
1774 | |
1775 | printf("\t%llu", val); | |
1776 | } | |
1777 | printf("\n"); | |
1778 | } | |
1779 | } | |
1780 | ||
1781 | /* Prepares data structure for test variants storage, sets test configuration */ | |
1782 | #ifdef _WIN32 | |
1783 | static void | |
1784 | #else | |
1785 | static void * | |
1786 | #endif | |
1787 | run_tests(void *arg) | |
1788 | { | |
1789 | uint32_t i; | |
1790 | struct thread_info *info = (struct thread_info *)arg; | |
1791 | MB_MGR *p_mgr = NULL; | |
1792 | struct params_s params; | |
f67539c2 | 1793 | uint32_t num_variants[NUM_TTYPES] = {0}; |
11fdf7f2 TL |
1794 | uint32_t type, at_size, run, arch; |
1795 | uint32_t variants_per_arch, max_arch; | |
1796 | uint32_t variant; | |
1797 | uint32_t total_variants = 0; | |
1798 | struct variant_s *variant_ptr = NULL; | |
1799 | struct variant_s *variant_list = NULL; | |
f67539c2 TL |
1800 | const uint32_t min_size = job_sizes[RANGE_MIN]; |
1801 | const uint32_t max_size = job_sizes[RANGE_MAX]; | |
1802 | const uint32_t step_size = job_sizes[RANGE_STEP]; | |
1803 | uint8_t *buf = NULL; | |
1804 | uint128_t *keys = NULL; | |
11fdf7f2 | 1805 | |
9f95a23c | 1806 | p_mgr = info->p_mgr; |
f67539c2 TL |
1807 | |
1808 | params.num_sizes = ((max_size - min_size) / step_size) + 1; | |
1809 | ||
11fdf7f2 TL |
1810 | params.core = (uint32_t)info->core; |
1811 | ||
1812 | /* if cores selected then set affinity */ | |
1813 | if (core_mask) | |
1814 | if (set_affinity(info->core) != 0) { | |
1815 | fprintf(stderr, "Failed to set cpu " | |
1816 | "affinity on core %d\n", info->core); | |
1817 | goto exit_failure; | |
1818 | } | |
1819 | ||
1820 | /* If unhalted cycles selected and this is | |
1821 | the primary thread then start counter */ | |
1822 | if (use_unhalted_cycles && info->print_info) { | |
1823 | int ret; | |
1824 | ||
1825 | ret = start_cycles_ctr(params.core); | |
1826 | if (ret != 0) { | |
1827 | fprintf(stderr, "Failed to start cycles " | |
1828 | "counter on core %u\n", params.core); | |
1829 | goto exit_failure; | |
1830 | } | |
1831 | /* Get average cost of reading counter */ | |
1832 | ret = set_avg_unhalted_cycle_cost(params.core, &rd_cycles_cost); | |
1833 | if (ret != 0 || rd_cycles_cost == 0) { | |
1834 | fprintf(stderr, "Error calculating unhalted " | |
1835 | "cycles read overhead!\n"); | |
1836 | goto exit_failure; | |
1837 | } else | |
1838 | fprintf(stderr, "Started counting unhalted cycles on " | |
1839 | "core %d\nUnhalted cycles read cost = %lu " | |
1840 | "cycles\n", params.core, | |
1841 | (unsigned long)rd_cycles_cost); | |
1842 | } | |
1843 | ||
f67539c2 TL |
1844 | init_mem(&buf, &keys); |
1845 | ||
1846 | for (type = TTYPE_AES_HMAC; type < NUM_TTYPES; type++) { | |
11fdf7f2 TL |
1847 | if (test_types[type] == 0) |
1848 | continue; | |
1849 | ||
1850 | switch (type) { | |
1851 | default: | |
1852 | case TTYPE_AES_HMAC: | |
1853 | variants_per_arch = VARIANTS_PER_ARCH_AES; | |
1854 | max_arch = NUM_ARCHS; | |
1855 | break; | |
1856 | case TTYPE_AES_DOCSIS: | |
1857 | variants_per_arch = VARIANTS_PER_ARCH_DOCSIS; | |
1858 | max_arch = NUM_ARCHS; | |
1859 | break; | |
1860 | case TTYPE_AES_GCM: | |
1861 | variants_per_arch = VARIANTS_PER_ARCH_GCM; | |
9f95a23c | 1862 | max_arch = NUM_ARCHS; |
11fdf7f2 TL |
1863 | break; |
1864 | case TTYPE_AES_CCM: | |
1865 | variants_per_arch = VARIANTS_PER_ARCH_CCM; | |
1866 | max_arch = NUM_ARCHS; | |
1867 | break; | |
1868 | case TTYPE_AES_DES: | |
1869 | variants_per_arch = VARIANTS_PER_ARCH_DES; | |
1870 | max_arch = NUM_ARCHS; | |
1871 | break; | |
1872 | case TTYPE_AES_3DES: | |
1873 | variants_per_arch = VARIANTS_PER_ARCH_3DES; | |
1874 | max_arch = NUM_ARCHS; | |
1875 | break; | |
f67539c2 TL |
1876 | case TTYPE_PON: |
1877 | variants_per_arch = VARIANTS_PER_ARCH_PON; | |
1878 | max_arch = NUM_ARCHS; | |
1879 | break; | |
1880 | case TTYPE_CUSTOM: | |
1881 | variants_per_arch = 1; | |
1882 | max_arch = NUM_ARCHS; | |
1883 | break; | |
11fdf7f2 TL |
1884 | } |
1885 | ||
1886 | /* Calculating number of all variants */ | |
1887 | for (arch = 0; arch < max_arch; arch++) { | |
1888 | if (archs[arch] == 0) | |
1889 | continue; | |
1890 | num_variants[type] += variants_per_arch; | |
1891 | } | |
1892 | total_variants += num_variants[type]; | |
1893 | } | |
1894 | ||
f67539c2 TL |
1895 | if (total_variants == 0) { |
1896 | fprintf(stderr, "No tests to be run\n"); | |
1897 | goto exit; | |
1898 | } | |
1899 | ||
1900 | if (info->print_info && !silent_progress_bar) | |
1901 | fprintf(stderr, "Total number of combinations (algos, " | |
1902 | "key sizes, cipher directions) to test = %u\n", | |
1903 | total_variants); | |
1904 | ||
11fdf7f2 TL |
1905 | variant_list = (struct variant_s *) |
1906 | malloc(total_variants * sizeof(struct variant_s)); | |
1907 | if (variant_list == NULL) { | |
1908 | fprintf(stderr, "Cannot allocate memory\n"); | |
1909 | goto exit_failure; | |
1910 | } | |
1911 | ||
1912 | at_size = NUM_RUNS * params.num_sizes * sizeof(uint64_t); | |
1913 | for (variant = 0, variant_ptr = variant_list; | |
1914 | variant < total_variants; | |
1915 | variant++, variant_ptr++) { | |
1916 | variant_ptr->avg_times = (uint64_t *) malloc(at_size); | |
1917 | if (!variant_ptr->avg_times) { | |
1918 | fprintf(stderr, "Cannot allocate memory\n"); | |
1919 | goto exit_failure; | |
1920 | } | |
1921 | } | |
f67539c2 | 1922 | |
11fdf7f2 | 1923 | for (run = 0; run < NUM_RUNS; run++) { |
f67539c2 TL |
1924 | if (info->print_info) |
1925 | fprintf(stderr, "\nStarting run %d of %d%c", | |
1926 | run+1, NUM_RUNS, | |
1927 | silent_progress_bar ? '\r' : '\n' ); | |
11fdf7f2 TL |
1928 | |
1929 | variant = 0; | |
1930 | variant_ptr = variant_list; | |
1931 | ||
9f95a23c TL |
1932 | if (iter_scale == ITER_SCALE_SMOKE && run != 0) |
1933 | continue; | |
1934 | ||
f67539c2 TL |
1935 | if (info->print_info) |
1936 | prog_bar_init(total_variants); | |
1937 | ||
1938 | for (type = TTYPE_AES_HMAC; type < NUM_TTYPES; type++) { | |
11fdf7f2 TL |
1939 | if (test_types[type] == 0) |
1940 | continue; | |
1941 | ||
9f95a23c | 1942 | max_arch = NUM_ARCHS; |
11fdf7f2 TL |
1943 | |
1944 | params.num_variants = num_variants[type]; | |
1945 | params.test_type = type; | |
1946 | /* Performing tests for each selected architecture */ | |
1947 | for (arch = 0; arch < max_arch; arch++) { | |
1948 | if (archs[arch] == 0) | |
1949 | continue; | |
1950 | run_dir_test(p_mgr, arch, ¶ms, run, | |
f67539c2 TL |
1951 | &variant_ptr, &variant, buf, |
1952 | keys, info->print_info); | |
11fdf7f2 TL |
1953 | } |
1954 | } /* end for type */ | |
f67539c2 TL |
1955 | if (info->print_info) |
1956 | prog_bar_fini(); | |
1957 | ||
11fdf7f2 | 1958 | } /* end for run */ |
f67539c2 TL |
1959 | if (info->print_info == 1 && iter_scale != ITER_SCALE_SMOKE) { |
1960 | fprintf(stderr, "\n"); | |
1961 | print_times(variant_list, ¶ms, total_variants, buf, keys); | |
1962 | } | |
11fdf7f2 | 1963 | |
f67539c2 | 1964 | exit: |
11fdf7f2 TL |
1965 | if (variant_list != NULL) { |
1966 | /* Freeing variants list */ | |
1967 | for (i = 0; i < total_variants; i++) | |
1968 | free(variant_list[i].avg_times); | |
1969 | free(variant_list); | |
1970 | } | |
f67539c2 | 1971 | free_mem(&buf, &keys); |
11fdf7f2 TL |
1972 | free_mb_mgr(p_mgr); |
1973 | #ifndef _WIN32 | |
1974 | return NULL; | |
1975 | ||
1976 | #else | |
1977 | return; | |
1978 | #endif | |
1979 | exit_failure: | |
1980 | if (variant_list != NULL) | |
1981 | free(variant_list); | |
f67539c2 | 1982 | free_mem(&buf, &keys); |
11fdf7f2 TL |
1983 | free_mb_mgr(p_mgr); |
1984 | exit(EXIT_FAILURE); | |
1985 | } | |
1986 | ||
1987 | static void usage(void) | |
1988 | { | |
1989 | fprintf(stderr, "Usage: ipsec_perf [args], " | |
1990 | "where args are zero or more\n" | |
1991 | "-h: print this message\n" | |
1992 | "-c: Use cold cache, it uses warm as default\n" | |
1993 | "-w: Use warm cache\n" | |
f67539c2 TL |
1994 | "--arch: run only tests on specified architecture (SSE/AVX/AVX2/AVX512)\n" |
1995 | "--cipher-algo: Select cipher algorithm to run on the custom test\n" | |
1996 | "--cipher-dir: Select cipher direction to run on the custom test " | |
1997 | "(encrypt/decrypt) (default = encrypt)\n" | |
1998 | "--hash-algo: Select hash algorithm to run on the custom test\n" | |
1999 | "--aead-algo: Select AEAD algorithm to run on the custom test\n" | |
11fdf7f2 TL |
2000 | "--no-avx512: Don't do AVX512\n" |
2001 | "--no-avx2: Don't do AVX2\n" | |
2002 | "--no-avx: Don't do AVX\n" | |
2003 | "--no-sse: Don't do SSE\n" | |
2004 | "-o val: Use <val> for the SHA size increment, default is 24\n" | |
2005 | "--shani-on: use SHA extensions, default: auto-detect\n" | |
2006 | "--shani-off: don't use SHA extensions\n" | |
2007 | "--no-gcm: do not run GCM perf tests\n" | |
2008 | "--no-aes: do not run standard AES + HMAC perf tests\n" | |
2009 | "--no-docsis: do not run DOCSIS cipher perf tests\n" | |
2010 | "--no-ccm: do not run CCM cipher perf tests\n" | |
2011 | "--no-des: do not run DES cipher perf tests\n" | |
2012 | "--no-3des: do not run 3DES cipher perf tests\n" | |
f67539c2 | 2013 | "--no-pon: do not run PON cipher perf tests\n" |
11fdf7f2 TL |
2014 | "--gcm-job-api: use JOB API for GCM perf tests" |
2015 | " (raw GCM API is default)\n" | |
2016 | "--threads num: <num> for the number of threads to run" | |
2017 | " Max: %d\n" | |
2018 | "--cores mask: <mask> CPU's to run threads\n" | |
2019 | "--unhalted-cycles: measure using unhalted cycles (requires root).\n" | |
9f95a23c | 2020 | " Note: RDTSC is used by default.\n" |
f67539c2 TL |
2021 | "--quick: reduces number of test iterations by x10\n" |
2022 | " (less precise but quicker)\n" | |
2023 | "--smoke: very quick, unprecise and without print out\n" | |
2024 | " (for validation only)\n" | |
2025 | "--job-size: size of the cipher & MAC job in bytes. It can be:\n" | |
2026 | " - single value: test single size\n" | |
2027 | " - range: test multiple sizes with following format" | |
2028 | " min:step:max (e.g. 16:16:256)\n" | |
2029 | " (-o still applies for MAC)\n" | |
2030 | "--aad-size: size of AAD for AEAD algorithms\n" | |
2031 | "--job-iter: number of tests iterations for each job size\n" | |
2032 | "--no-progress-bar: Don't display progress bar\n", | |
11fdf7f2 TL |
2033 | MAX_NUM_THREADS + 1); |
2034 | } | |
2035 | ||
f67539c2 TL |
2036 | static int |
2037 | get_next_num_arg(const char * const *argv, const int index, const int argc, | |
2038 | void *dst, const size_t dst_size) | |
2039 | { | |
2040 | char *endptr = NULL; | |
2041 | uint64_t val; | |
2042 | ||
2043 | if (dst == NULL || argv == NULL || index < 0 || argc < 0) { | |
2044 | fprintf(stderr, "%s() internal error!\n", __func__); | |
2045 | exit(EXIT_FAILURE); | |
2046 | } | |
2047 | ||
2048 | if (index >= (argc - 1)) { | |
2049 | fprintf(stderr, "'%s' requires an argument!\n", argv[index]); | |
2050 | exit(EXIT_FAILURE); | |
2051 | } | |
2052 | ||
2053 | #ifdef _WIN32 | |
2054 | val = _strtoui64(argv[index + 1], &endptr, 0); | |
2055 | #else | |
2056 | val = strtoull(argv[index + 1], &endptr, 0); | |
2057 | #endif | |
2058 | if (endptr == argv[index + 1] || (endptr != NULL && *endptr != '\0')) { | |
2059 | fprintf(stderr, "Error converting '%s' as value for '%s'!\n", | |
2060 | argv[index + 1], argv[index]); | |
2061 | exit(EXIT_FAILURE); | |
2062 | } | |
2063 | ||
2064 | switch (dst_size) { | |
2065 | case (sizeof(uint8_t)): | |
2066 | *((uint8_t *)dst) = (uint8_t) val; | |
2067 | break; | |
2068 | case (sizeof(uint16_t)): | |
2069 | *((uint16_t *)dst) = (uint16_t) val; | |
2070 | break; | |
2071 | case (sizeof(uint32_t)): | |
2072 | *((uint32_t *)dst) = (uint32_t) val; | |
2073 | break; | |
2074 | case (sizeof(uint64_t)): | |
2075 | *((uint64_t *)dst) = val; | |
2076 | break; | |
2077 | default: | |
2078 | fprintf(stderr, "%s() invalid dst_size %u!\n", | |
2079 | __func__, (unsigned) dst_size); | |
2080 | exit(EXIT_FAILURE); | |
2081 | break; | |
2082 | } | |
2083 | ||
2084 | return index + 1; | |
2085 | } | |
2086 | ||
2087 | static int | |
2088 | detect_arch(unsigned int arch_support[NUM_ARCHS]) | |
2089 | { | |
2090 | const uint64_t detect_sse = | |
2091 | IMB_FEATURE_SSE4_2 | IMB_FEATURE_CMOV | IMB_FEATURE_AESNI; | |
2092 | const uint64_t detect_avx = | |
2093 | IMB_FEATURE_AVX | IMB_FEATURE_CMOV | IMB_FEATURE_AESNI; | |
2094 | const uint64_t detect_avx2 = IMB_FEATURE_AVX2 | detect_avx; | |
2095 | const uint64_t detect_avx512 = IMB_FEATURE_AVX512_SKX | detect_avx2; | |
2096 | MB_MGR *p_mgr = NULL; | |
2097 | enum arch_type_e arch_id; | |
2098 | ||
2099 | if (arch_support == NULL) { | |
2100 | fprintf(stderr, "Array not passed correctly\n"); | |
2101 | return -1; | |
2102 | } | |
2103 | ||
2104 | for (arch_id = ARCH_SSE; arch_id < NUM_ARCHS; arch_id++) | |
2105 | arch_support[arch_id] = 1; | |
2106 | ||
2107 | p_mgr = alloc_mb_mgr(0); | |
2108 | if (p_mgr == NULL) { | |
2109 | fprintf(stderr, "Architecture detect error!\n"); | |
2110 | return -1; | |
2111 | } | |
2112 | ||
2113 | if ((p_mgr->features & detect_avx512) != detect_avx512) | |
2114 | arch_support[ARCH_AVX512] = 0; | |
2115 | ||
2116 | if ((p_mgr->features & detect_avx2) != detect_avx2) | |
2117 | arch_support[ARCH_AVX2] = 0; | |
2118 | ||
2119 | if ((p_mgr->features & detect_avx) != detect_avx) | |
2120 | arch_support[ARCH_AVX] = 0; | |
2121 | ||
2122 | if ((p_mgr->features & detect_sse) != detect_sse) | |
2123 | arch_support[ARCH_SSE] = 0; | |
2124 | ||
2125 | free_mb_mgr(p_mgr); | |
2126 | ||
2127 | return 0; | |
2128 | } | |
2129 | ||
2130 | /* | |
2131 | * Check string argument is supported and if it is, return values associated | |
2132 | * with it. | |
2133 | */ | |
2134 | static const union params * | |
2135 | check_string_arg(const char *param, const char *arg, | |
2136 | const struct str_value_mapping *map, | |
2137 | const unsigned int num_avail_opts) | |
2138 | { | |
2139 | unsigned int i; | |
2140 | ||
2141 | if (arg == NULL) { | |
2142 | fprintf(stderr, "%s requires an argument\n", param); | |
2143 | goto exit; | |
2144 | } | |
2145 | ||
2146 | for (i = 0; i < num_avail_opts; i++) | |
2147 | if (strcmp(arg, map[i].name) == 0) | |
2148 | return &(map[i].values); | |
2149 | ||
2150 | /* Argument is not listed in the available options */ | |
2151 | fprintf(stderr, "Invalid argument for %s\n", param); | |
2152 | exit: | |
2153 | fprintf(stderr, "Accepted arguments: "); | |
2154 | for (i = 0; i < num_avail_opts; i++) | |
2155 | fprintf(stderr, "%s ", map[i].name); | |
2156 | fprintf(stderr, "\n"); | |
2157 | ||
2158 | return NULL; | |
2159 | } | |
2160 | ||
2161 | static int | |
2162 | parse_range(const char * const *argv, const int index, const int argc, | |
2163 | uint32_t range_values[NUM_RANGE]) | |
2164 | { | |
2165 | char *token; | |
2166 | uint32_t number; | |
2167 | unsigned int i; | |
2168 | ||
2169 | ||
2170 | if (range_values == NULL || argv == NULL || index < 0 || argc < 0) { | |
2171 | fprintf(stderr, "%s() internal error!\n", __func__); | |
2172 | exit(EXIT_FAILURE); | |
2173 | } | |
2174 | ||
2175 | if (index >= (argc - 1)) { | |
2176 | fprintf(stderr, "'%s' requires an argument!\n", argv[index]); | |
2177 | exit(EXIT_FAILURE); | |
2178 | } | |
2179 | ||
2180 | char *copy_arg = strdup(argv[index + 1]); | |
2181 | ||
2182 | if (copy_arg == NULL) { | |
2183 | fprintf(stderr, "%s() internal error!\n", __func__); | |
2184 | exit(EXIT_FAILURE); | |
2185 | } | |
2186 | ||
2187 | errno = 0; | |
2188 | token = strtok(copy_arg, ":"); | |
2189 | ||
2190 | /* Try parsing range (minimum, step and maximum values) */ | |
2191 | for (i = 0; i < NUM_RANGE; i++) { | |
2192 | if (token == NULL) | |
2193 | goto no_range; | |
2194 | ||
2195 | number = strtoul(token, NULL, 10); | |
2196 | ||
2197 | if (errno != 0) | |
2198 | goto no_range; | |
2199 | ||
2200 | range_values[i] = number; | |
2201 | token = strtok(NULL, ":"); | |
2202 | } | |
2203 | ||
2204 | if (token != NULL) | |
2205 | goto no_range; | |
2206 | ||
2207 | if (range_values[RANGE_MAX] < range_values[RANGE_MIN]) { | |
2208 | fprintf(stderr, "Maximum value of range cannot be lower " | |
2209 | "than minimum value\n"); | |
2210 | exit(EXIT_FAILURE); | |
2211 | } | |
2212 | ||
2213 | if (range_values[RANGE_STEP] == 0) { | |
2214 | fprintf(stderr, "Step value in range cannot be 0\n"); | |
2215 | exit(EXIT_FAILURE); | |
2216 | } | |
2217 | ||
2218 | goto end_range; | |
2219 | no_range: | |
2220 | /* Try parsing as single value */ | |
2221 | get_next_num_arg(argv, index, argc, &job_sizes[RANGE_MIN], | |
2222 | sizeof(job_sizes[RANGE_MIN])); | |
2223 | ||
2224 | job_sizes[RANGE_MAX] = job_sizes[RANGE_MIN]; | |
2225 | ||
2226 | end_range: | |
2227 | free(copy_arg); | |
2228 | return (index + 1); | |
2229 | ||
2230 | } | |
2231 | ||
11fdf7f2 TL |
2232 | int main(int argc, char *argv[]) |
2233 | { | |
f67539c2 TL |
2234 | uint32_t num_t = 0; |
2235 | int i, core = 0; | |
11fdf7f2 | 2236 | struct thread_info *thread_info_p = t_info; |
f67539c2 TL |
2237 | unsigned int arch_id; |
2238 | unsigned int arch_support[NUM_ARCHS]; | |
2239 | const union params *values; | |
2240 | unsigned int cipher_algo_set = 0; | |
2241 | unsigned int hash_algo_set = 0; | |
2242 | unsigned int aead_algo_set = 0; | |
2243 | unsigned int cipher_dir_set = 0; | |
11fdf7f2 TL |
2244 | #ifdef _WIN32 |
2245 | HANDLE threads[MAX_NUM_THREADS]; | |
2246 | #else | |
2247 | pthread_t tids[MAX_NUM_THREADS]; | |
2248 | #endif | |
2249 | ||
2250 | for (i = 1; i < argc; i++) | |
2251 | if (strcmp(argv[i], "-h") == 0) { | |
2252 | usage(); | |
2253 | return EXIT_SUCCESS; | |
2254 | } else if (strcmp(argv[i], "-c") == 0) { | |
2255 | cache_type = COLD; | |
2256 | fprintf(stderr, "Cold cache, "); | |
2257 | } else if (strcmp(argv[i], "-w") == 0) { | |
2258 | cache_type = WARM; | |
2259 | fprintf(stderr, "Warm cache, "); | |
2260 | } else if (strcmp(argv[i], "--no-avx512") == 0) { | |
2261 | archs[ARCH_AVX512] = 0; | |
2262 | } else if (strcmp(argv[i], "--no-avx2") == 0) { | |
2263 | archs[ARCH_AVX2] = 0; | |
2264 | } else if (strcmp(argv[i], "--no-avx") == 0) { | |
2265 | archs[ARCH_AVX] = 0; | |
2266 | } else if (strcmp(argv[i], "--no-sse") == 0) { | |
2267 | archs[ARCH_SSE] = 0; | |
2268 | } else if (strcmp(argv[i], "--shani-on") == 0) { | |
2269 | flags &= (~IMB_FLAG_SHANI_OFF); | |
2270 | } else if (strcmp(argv[i], "--shani-off") == 0) { | |
2271 | flags |= IMB_FLAG_SHANI_OFF; | |
2272 | } else if (strcmp(argv[i], "--no-gcm") == 0) { | |
2273 | test_types[TTYPE_AES_GCM] = 0; | |
2274 | } else if (strcmp(argv[i], "--no-aes") == 0) { | |
2275 | test_types[TTYPE_AES_HMAC] = 0; | |
2276 | } else if (strcmp(argv[i], "--no-docsis") == 0) { | |
2277 | test_types[TTYPE_AES_DOCSIS] = 0; | |
2278 | } else if (strcmp(argv[i], "--no-ccm") == 0) { | |
2279 | test_types[TTYPE_AES_CCM] = 0; | |
2280 | } else if (strcmp(argv[i], "--no-des") == 0) { | |
2281 | test_types[TTYPE_AES_DES] = 0; | |
2282 | } else if (strcmp(argv[i], "--no-3des") == 0) { | |
2283 | test_types[TTYPE_AES_3DES] = 0; | |
f67539c2 TL |
2284 | } else if (strcmp(argv[i], "--no-pon") == 0) { |
2285 | test_types[TTYPE_PON] = 0; | |
11fdf7f2 TL |
2286 | } else if (strcmp(argv[i], "--gcm-job-api") == 0) { |
2287 | use_gcm_job_api = 1; | |
9f95a23c TL |
2288 | } else if (strcmp(argv[i], "--quick") == 0) { |
2289 | iter_scale = ITER_SCALE_SHORT; | |
2290 | } else if (strcmp(argv[i], "--smoke") == 0) { | |
2291 | iter_scale = ITER_SCALE_SMOKE; | |
f67539c2 TL |
2292 | } else if (strcmp(argv[i], "--arch") == 0) { |
2293 | values = check_string_arg(argv[i], argv[i+1], | |
2294 | arch_str_map, | |
2295 | DIM(arch_str_map)); | |
2296 | if (values == NULL) | |
2297 | return EXIT_FAILURE; | |
2298 | ||
2299 | /* | |
2300 | * Disable all the other architectures | |
2301 | * and enable only the specified | |
2302 | */ | |
2303 | memset(archs, 0, sizeof(archs)); | |
2304 | archs[values->arch_type] = 1; | |
2305 | i++; | |
2306 | } else if (strcmp(argv[i], "--cipher-algo") == 0) { | |
2307 | values = check_string_arg(argv[i], argv[i+1], | |
2308 | cipher_algo_str_map, | |
2309 | DIM(cipher_algo_str_map)); | |
2310 | if (values == NULL) | |
2311 | return EXIT_FAILURE; | |
2312 | ||
2313 | custom_job_params.cipher_mode = | |
2314 | values->job_params.cipher_mode; | |
2315 | custom_job_params.aes_key_size = | |
2316 | values->job_params.aes_key_size; | |
2317 | test_types[TTYPE_CUSTOM] = 1; | |
2318 | cipher_algo_set = 1; | |
2319 | i++; | |
2320 | } else if (strcmp(argv[i], "--cipher-dir") == 0) { | |
2321 | values = check_string_arg(argv[i], argv[i+1], | |
2322 | cipher_dir_str_map, | |
2323 | DIM(cipher_dir_str_map)); | |
2324 | if (values == NULL) | |
2325 | return EXIT_FAILURE; | |
2326 | ||
2327 | custom_job_params.cipher_dir = | |
2328 | values->job_params.cipher_dir; | |
2329 | cipher_dir_set = 1; | |
2330 | i++; | |
2331 | } else if (strcmp(argv[i], "--hash-algo") == 0) { | |
2332 | values = check_string_arg(argv[i], argv[i+1], | |
2333 | hash_algo_str_map, | |
2334 | DIM(hash_algo_str_map)); | |
2335 | if (values == NULL) | |
2336 | return EXIT_FAILURE; | |
2337 | ||
2338 | custom_job_params.hash_alg = | |
2339 | values->job_params.hash_alg; | |
2340 | test_types[TTYPE_CUSTOM] = 1; | |
2341 | hash_algo_set = 1; | |
2342 | i++; | |
2343 | } else if (strcmp(argv[i], "--aead-algo") == 0) { | |
2344 | values = check_string_arg(argv[i], argv[i+1], | |
2345 | aead_algo_str_map, | |
2346 | DIM(aead_algo_str_map)); | |
2347 | if (values == NULL) | |
2348 | return EXIT_FAILURE; | |
2349 | ||
2350 | custom_job_params.cipher_mode = | |
2351 | values->job_params.cipher_mode; | |
2352 | custom_job_params.aes_key_size = | |
2353 | values->job_params.aes_key_size; | |
2354 | custom_job_params.hash_alg = | |
2355 | values->job_params.hash_alg; | |
2356 | test_types[TTYPE_CUSTOM] = 1; | |
2357 | aead_algo_set = 1; | |
11fdf7f2 | 2358 | i++; |
f67539c2 TL |
2359 | } else if (strcmp(argv[i], "-o") == 0) { |
2360 | i = get_next_num_arg((const char * const *)argv, i, | |
2361 | argc, &sha_size_incr, | |
2362 | sizeof(sha_size_incr)); | |
2363 | } else if (strcmp(argv[i], "--job-size") == 0) { | |
2364 | /* Try parsing the argument as a range first */ | |
2365 | i = parse_range((const char * const *)argv, i, argc, | |
2366 | job_sizes); | |
2367 | if (job_sizes[RANGE_MAX] > JOB_SIZE_TOP) { | |
2368 | fprintf(stderr, | |
2369 | "Invalid job size %u (max %u)\n", | |
2370 | (unsigned) job_sizes[RANGE_MAX], | |
2371 | JOB_SIZE_TOP); | |
2372 | return EXIT_FAILURE; | |
2373 | } | |
2374 | } else if (strcmp(argv[i], "--aad-size") == 0) { | |
2375 | /* Get AAD size for both GCM and CCM */ | |
2376 | i = get_next_num_arg((const char * const *)argv, i, | |
2377 | argc, &gcm_aad_size, | |
2378 | sizeof(gcm_aad_size)); | |
2379 | if (gcm_aad_size > AAD_SIZE_MAX) { | |
2380 | fprintf(stderr, | |
2381 | "Invalid AAD size %u (max %u)!\n", | |
2382 | (unsigned) gcm_aad_size, | |
2383 | AAD_SIZE_MAX); | |
2384 | return EXIT_FAILURE; | |
2385 | } | |
2386 | ccm_aad_size = gcm_aad_size; | |
2387 | } else if (strcmp(argv[i], "--job-iter") == 0) { | |
2388 | i = get_next_num_arg((const char * const *)argv, i, | |
2389 | argc, &job_iter, sizeof(job_iter)); | |
11fdf7f2 | 2390 | } else if (strcmp(argv[i], "--threads") == 0) { |
f67539c2 TL |
2391 | i = get_next_num_arg((const char * const *)argv, i, |
2392 | argc, &num_t, sizeof(num_t)); | |
11fdf7f2 TL |
2393 | if (num_t > (MAX_NUM_THREADS + 1)) { |
2394 | fprintf(stderr, "Invalid number of threads!\n"); | |
2395 | return EXIT_FAILURE; | |
2396 | } | |
2397 | } else if (strcmp(argv[i], "--cores") == 0) { | |
f67539c2 TL |
2398 | i = get_next_num_arg((const char * const *)argv, i, |
2399 | argc, &core_mask, | |
2400 | sizeof(core_mask)); | |
11fdf7f2 | 2401 | } else if (strcmp(argv[i], "--unhalted-cycles") == 0) { |
11fdf7f2 | 2402 | use_unhalted_cycles = 1; |
f67539c2 TL |
2403 | } else if (strcmp(argv[i], "--no-progress-bar") == 0) { |
2404 | silent_progress_bar = 1; | |
11fdf7f2 TL |
2405 | } else { |
2406 | usage(); | |
2407 | return EXIT_FAILURE; | |
2408 | } | |
2409 | ||
f67539c2 TL |
2410 | if (test_types[TTYPE_CUSTOM]) { |
2411 | /* Disable all other tests when custom test is selected */ | |
2412 | memset(test_types, 0, sizeof(test_types)); | |
2413 | test_types[TTYPE_CUSTOM] = 1; | |
2414 | if (aead_algo_set && (cipher_algo_set || hash_algo_set)) { | |
2415 | fprintf(stderr, "AEAD algorithm cannot be used " | |
2416 | "combined with another cipher/hash " | |
2417 | "algorithm\n"); | |
2418 | return EXIT_FAILURE; | |
2419 | } | |
2420 | } | |
2421 | ||
2422 | if (cipher_algo_set == 0 && aead_algo_set == 0 && cipher_dir_set) { | |
2423 | fprintf(stderr, "--cipher-dir can only be used with " | |
2424 | "--cipher-algo or --aead-algo\n"); | |
2425 | return EXIT_FAILURE; | |
2426 | } | |
2427 | ||
2428 | if (test_types[TTYPE_AES_CCM] || | |
2429 | custom_job_params.cipher_mode == TEST_CCM) { | |
2430 | if (ccm_aad_size > CCM_AAD_SIZE_MAX) { | |
2431 | fprintf(stderr, "AAD cannot be higher than %u in CCM\n", | |
2432 | CCM_AAD_SIZE_MAX); | |
2433 | return EXIT_FAILURE; | |
2434 | } | |
2435 | } | |
2436 | ||
2437 | if (job_sizes[RANGE_MIN] == 0) { | |
2438 | if (test_types[TTYPE_AES_HMAC] || | |
2439 | test_types[TTYPE_AES_DOCSIS] || | |
2440 | test_types[TTYPE_AES_DES] || | |
2441 | test_types[TTYPE_AES_3DES] || | |
2442 | (test_types[TTYPE_CUSTOM] && | |
2443 | aead_algo_set == 0)) { | |
2444 | fprintf(stderr, "Buffer size cannot be 0 unless only " | |
2445 | "an AEAD algorithm is tested\n"); | |
2446 | return EXIT_FAILURE; | |
2447 | } | |
2448 | } | |
2449 | ||
11fdf7f2 TL |
2450 | /* Check num cores >= number of threads */ |
2451 | if ((core_mask != 0 && num_t != 0) && (num_t > bitcount(core_mask))) { | |
2452 | fprintf(stderr, "Insufficient number of cores in " | |
2453 | "core mask (0x%lx) to run %d threads!\n", | |
2454 | (unsigned long) core_mask, num_t); | |
2455 | return EXIT_FAILURE; | |
2456 | } | |
2457 | ||
2458 | /* if cycles selected then init MSR module */ | |
2459 | if (use_unhalted_cycles) { | |
2460 | if (core_mask == 0) { | |
2461 | fprintf(stderr, "Must specify core mask " | |
2462 | "when reading unhalted cycles!\n"); | |
2463 | return EXIT_FAILURE; | |
2464 | } | |
2465 | ||
2466 | if (init_msr_mod() != 0) { | |
2467 | fprintf(stderr, "Error initializing MSR module!\n"); | |
2468 | return EXIT_FAILURE; | |
2469 | } | |
2470 | } | |
2471 | ||
f67539c2 TL |
2472 | if (detect_arch(arch_support) < 0) |
2473 | return EXIT_FAILURE; | |
2474 | ||
2475 | /* disable tests depending on instruction sets supported */ | |
2476 | for (arch_id = 0; arch_id < NUM_ARCHS; arch_id++) { | |
2477 | if (archs[arch_id] == 1 && arch_support[arch_id] == 0) { | |
2478 | archs[arch_id] = 0; | |
2479 | fprintf(stderr, | |
2480 | "%s not supported. Disabling %s tests\n", | |
2481 | arch_str_map[arch_id].name, | |
2482 | arch_str_map[arch_id].name); | |
2483 | } | |
2484 | } | |
2485 | ||
11fdf7f2 | 2486 | fprintf(stderr, "SHA size incr = %d\n", sha_size_incr); |
f67539c2 TL |
2487 | |
2488 | if (test_types[TTYPE_AES_GCM] || | |
2489 | (custom_job_params.cipher_mode == TEST_GCM)) | |
2490 | fprintf(stderr, "GCM AAD = %"PRIu64"\n", gcm_aad_size); | |
2491 | ||
2492 | if (test_types[TTYPE_AES_CCM] || | |
2493 | (custom_job_params.cipher_mode == TEST_CCM)) | |
2494 | fprintf(stderr, "CCM AAD = %"PRIu64"\n", ccm_aad_size); | |
2495 | ||
11fdf7f2 TL |
2496 | if (archs[ARCH_SSE]) { |
2497 | MB_MGR *p_mgr = alloc_mb_mgr(flags); | |
2498 | ||
2499 | if (p_mgr == NULL) { | |
2500 | fprintf(stderr, "Error allocating MB_MGR structure!\n"); | |
2501 | return EXIT_FAILURE; | |
2502 | } | |
2503 | init_mb_mgr_sse(p_mgr); | |
2504 | fprintf(stderr, "%s SHA extensions (shani) for SSE arch\n", | |
2505 | (p_mgr->features & IMB_FEATURE_SHANI) ? | |
2506 | "Using" : "Not using"); | |
2507 | free_mb_mgr(p_mgr); | |
2508 | } | |
2509 | ||
2510 | memset(t_info, 0, sizeof(t_info)); | |
f67539c2 | 2511 | init_offsets(cache_type); |
9f95a23c TL |
2512 | |
2513 | srand(ITER_SCALE_LONG + ITER_SCALE_SHORT + ITER_SCALE_SMOKE); | |
9f95a23c | 2514 | |
f67539c2 TL |
2515 | if (num_t > 1) { |
2516 | uint32_t n; | |
2517 | ||
2518 | for (n = 0; n < (num_t - 1); n++, thread_info_p++) { | |
11fdf7f2 TL |
2519 | /* Set core if selected */ |
2520 | if (core_mask) { | |
2521 | core = next_core(core_mask, core); | |
2522 | thread_info_p->core = core++; | |
2523 | } | |
9f95a23c TL |
2524 | |
2525 | /* Allocate MB manager for each thread */ | |
2526 | thread_info_p->p_mgr = alloc_mb_mgr(flags); | |
2527 | if (thread_info_p->p_mgr == NULL) { | |
2528 | fprintf(stderr, "Failed to allocate MB_MGR " | |
f67539c2 TL |
2529 | "structure for thread %u!\n", |
2530 | (unsigned)(n + 1)); | |
9f95a23c TL |
2531 | exit(EXIT_FAILURE); |
2532 | } | |
11fdf7f2 | 2533 | #ifdef _WIN32 |
f67539c2 | 2534 | threads[n] = (HANDLE) |
11fdf7f2 TL |
2535 | _beginthread(&run_tests, 0, |
2536 | (void *)thread_info_p); | |
2537 | #else | |
2538 | pthread_attr_t attr; | |
2539 | ||
2540 | pthread_attr_init(&attr); | |
f67539c2 | 2541 | pthread_create(&tids[n], &attr, run_tests, |
11fdf7f2 TL |
2542 | (void *)thread_info_p); |
2543 | #endif | |
2544 | } | |
f67539c2 | 2545 | } |
11fdf7f2 TL |
2546 | |
2547 | thread_info_p->print_info = 1; | |
9f95a23c TL |
2548 | thread_info_p->p_mgr = alloc_mb_mgr(flags); |
2549 | if (thread_info_p->p_mgr == NULL) { | |
2550 | fprintf(stderr, "Failed to allocate MB_MGR " | |
2551 | "structure for main thread!\n"); | |
9f95a23c TL |
2552 | exit(EXIT_FAILURE); |
2553 | } | |
11fdf7f2 TL |
2554 | if (core_mask) { |
2555 | core = next_core(core_mask, core); | |
2556 | thread_info_p->core = core; | |
2557 | } | |
2558 | ||
2559 | run_tests((void *)thread_info_p); | |
2560 | if (num_t > 1) { | |
f67539c2 TL |
2561 | uint32_t n; |
2562 | ||
11fdf7f2 TL |
2563 | #ifdef _WIN32 |
2564 | WaitForMultipleObjects(num_t, threads, FALSE, INFINITE); | |
2565 | #endif | |
f67539c2 TL |
2566 | for (n = 0; n < (num_t - 1); n++) { |
2567 | fprintf(stderr, "Waiting on thread %u to finish...\n", | |
2568 | (unsigned)(n + 2)); | |
11fdf7f2 | 2569 | #ifdef _WIN32 |
f67539c2 | 2570 | CloseHandle(threads[n]); |
11fdf7f2 | 2571 | #else |
f67539c2 | 2572 | pthread_join(tids[n], NULL); |
11fdf7f2 TL |
2573 | #endif |
2574 | } | |
2575 | } | |
2576 | ||
2577 | if (use_unhalted_cycles) | |
2578 | machine_fini(); | |
2579 | ||
11fdf7f2 TL |
2580 | return EXIT_SUCCESS; |
2581 | } |