| 1 | /* |
| 2 | * Copyright (c) 2016, Intel Corporation |
| 3 | * |
| 4 | * Redistribution and use in source and binary forms, with or without |
| 5 | * modification, are permitted provided that the following conditions are met: |
| 6 | * |
| 7 | * * Redistributions of source code must retain the above copyright notice, |
| 8 | * 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 the |
| 11 | * documentation and/or other materials provided with the distribution. |
| 12 | * * Neither the name of Intel Corporation nor the names of its contributors |
| 13 | * may be used to endorse or promote products derived from this software |
| 14 | * without specific prior written permission. |
| 15 | * |
| 16 | * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" |
| 17 | * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| 18 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
| 19 | * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE |
| 20 | * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR |
| 21 | * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF |
| 22 | * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS |
| 23 | * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN |
| 24 | * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) |
| 25 | * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE |
| 26 | * POSSIBILITY OF SUCH DAMAGE. |
| 27 | */ |
| 28 | |
| 29 | /** \file |
| 30 | \brief Tamarama: container engine for exclusive engines, runtime code. |
| 31 | */ |
| 32 | #include "config.h" |
| 33 | |
| 34 | #include "tamarama.h" |
| 35 | |
| 36 | #include "tamarama_internal.h" |
| 37 | #include "nfa_api.h" |
| 38 | #include "nfa_api_queue.h" |
| 39 | #include "nfa_api_util.h" |
| 40 | #include "nfa_internal.h" |
| 41 | #include "scratch.h" |
| 42 | #include "util/partial_store.h" |
| 43 | |
| 44 | static really_inline |
| 45 | u32 getSubOffset(const struct Tamarama *t, u32 num) { |
| 46 | DEBUG_PRINTF("subengine:%u\n" , num); |
| 47 | assert(num < t->numSubEngines); |
| 48 | const u32 *sub = |
| 49 | (const u32 *)((const char *)t + sizeof(struct Tamarama) + |
| 50 | t->numSubEngines * sizeof(u32)); |
| 51 | assert(ISALIGNED(sub)); |
| 52 | return sub[num]; |
| 53 | } |
| 54 | |
| 55 | static |
| 56 | const struct NFA *getSubEngine(const struct Tamarama *t, |
| 57 | const u32 activeIdx) { |
| 58 | const u32 offset = getSubOffset(t, activeIdx); |
| 59 | DEBUG_PRINTF("activeIdx:%u offsets:%u\n" , activeIdx, offset); |
| 60 | const char *base = (const char *)t; |
| 61 | return (const struct NFA *)(base + offset); |
| 62 | } |
| 63 | |
| 64 | static |
| 65 | void storeActiveIdx(const struct Tamarama *t, char *state, |
| 66 | const u32 idx) { |
| 67 | assert(idx <= t->numSubEngines); |
| 68 | partial_store_u32(state, idx, t->activeIdxSize); |
| 69 | } |
| 70 | |
| 71 | static |
| 72 | u32 loadActiveIdx(const char *state, |
| 73 | const u32 activeIdxSize) { |
| 74 | return partial_load_u32(state, activeIdxSize); |
| 75 | } |
| 76 | |
| 77 | static really_inline |
| 78 | void copyQueueProperties(const struct mq *q1, struct mq *q2, |
| 79 | const u32 activeIdxSize) { |
| 80 | q2->state = q1->state; |
| 81 | q2->streamState = q1->streamState + activeIdxSize; |
| 82 | q2->offset = q1->offset; |
| 83 | q2->buffer = q1->buffer; |
| 84 | q2->length = q1->length; |
| 85 | q2->history = q1->history; |
| 86 | q2->hlength = q1->hlength; |
| 87 | q2->cb = q1->cb; |
| 88 | q2->context = q1->context; |
| 89 | q2->scratch = q1->scratch; |
| 90 | q2->report_current = q1->report_current; |
| 91 | } |
| 92 | |
| 93 | static |
| 94 | void copyQueueItems(const struct Tamarama *t, const struct NFA *sub, |
| 95 | struct mq *q1, struct mq *q2, const u32 activeIdx) { |
| 96 | const u32 *baseTop = (const u32 *)((const char *)t + |
| 97 | sizeof(struct Tamarama)); |
| 98 | |
| 99 | u32 lower = baseTop[activeIdx]; |
| 100 | u32 upper = activeIdx == t->numSubEngines - 1 ? |
| 101 | ~0U : baseTop[activeIdx + 1]; |
| 102 | u32 event_base = isMultiTopType(sub->type) ? MQE_TOP_FIRST : MQE_TOP; |
| 103 | while (q1->cur < q1->end) { |
| 104 | u32 type = q1->items[q1->cur].type; |
| 105 | s64a loc = q1->items[q1->cur].location; |
| 106 | DEBUG_PRINTF("type:%u lower:%u upper:%u\n" , type, lower, upper); |
| 107 | if (type >= lower && type < upper) { |
| 108 | u32 event = event_base; |
| 109 | if (event == MQE_TOP_FIRST) { |
| 110 | event += type - lower; |
| 111 | } |
| 112 | pushQueue(q2, event, loc); |
| 113 | } else { |
| 114 | pushQueueNoMerge(q2, MQE_END, loc); |
| 115 | break; |
| 116 | } |
| 117 | q1->cur++; |
| 118 | } |
| 119 | } |
| 120 | |
| 121 | static |
| 122 | void copyQueue(const struct Tamarama *t, const struct NFA *sub, |
| 123 | struct mq *q1, struct mq *q2, const u32 activeIdx) { |
| 124 | copyQueueProperties(q1, q2, t->activeIdxSize); |
| 125 | |
| 126 | // copy MQE_START item |
| 127 | u32 cur = q1->cur++; |
| 128 | q2->cur = cur; |
| 129 | q2->items[cur] = q1->items[cur]; |
| 130 | q2->end = cur + 1; |
| 131 | |
| 132 | copyQueueItems(t, sub, q1, q2, activeIdx); |
| 133 | // restore cur index of the main queue |
| 134 | q1->cur = cur; |
| 135 | } |
| 136 | |
| 137 | static |
| 138 | u32 findEngineForTop(const u32 *baseTop, const u32 cur, |
| 139 | const u32 numSubEngines) { |
| 140 | u32 i; |
| 141 | for (i = 0; i < numSubEngines; ++i) { |
| 142 | DEBUG_PRINTF("cur:%u base:%u\n" , cur, baseTop[i]); |
| 143 | if (cur >= baseTop[i] && |
| 144 | (i == numSubEngines - 1 || cur < baseTop[i + 1])) { |
| 145 | break; |
| 146 | } |
| 147 | } |
| 148 | return i; |
| 149 | } |
| 150 | |
| 151 | static |
| 152 | void initSubQueue(const struct Tamarama *t, struct mq *q1, |
| 153 | struct mq *q2, const u32 lastActiveIdx, |
| 154 | const u32 activeIdx) { |
| 155 | // Push events to the new queue |
| 156 | const struct NFA *sub = getSubEngine(t, activeIdx); |
| 157 | assert(!isContainerType(sub->type)); |
| 158 | q2->nfa = sub; |
| 159 | |
| 160 | // Reinitialize state if the last active subengine is different |
| 161 | // from current one |
| 162 | if (lastActiveIdx == t->numSubEngines || |
| 163 | lastActiveIdx != activeIdx) { |
| 164 | nfaQueueInitState(q2->nfa, q2); |
| 165 | } |
| 166 | |
| 167 | copyQueueItems(t, sub, q1, q2, activeIdx); |
| 168 | if (q1->items[q1->cur].type == MQE_END) { |
| 169 | q1->cur++; |
| 170 | } |
| 171 | DEBUG_PRINTF("update lastIdx:%u\n" , activeIdx); |
| 172 | storeActiveIdx(t, q1->streamState, activeIdx); |
| 173 | } |
| 174 | |
| 175 | static |
| 176 | void updateQueues(const struct Tamarama *t, struct mq *q1, struct mq *q2) { |
| 177 | q2->cur = q2->end = 0; |
| 178 | copyQueueProperties(q1, q2, t->activeIdxSize); |
| 179 | |
| 180 | const u32 numSubEngines = t->numSubEngines; |
| 181 | u32 lastActiveIdx = loadActiveIdx(q1->streamState, |
| 182 | t->activeIdxSize); |
| 183 | #ifdef DEBUG |
| 184 | DEBUG_PRINTF("external queue\n" ); |
| 185 | debugQueue(q1); |
| 186 | #endif |
| 187 | |
| 188 | // Push MQE_START event to the subqueue |
| 189 | s64a loc = q1->items[q1->cur].location; |
| 190 | pushQueueAt(q2, 0, MQE_START, loc); |
| 191 | char hasStart = 0; |
| 192 | if (q1->items[q1->cur].type == MQE_START) { |
| 193 | hasStart = 1; |
| 194 | q1->cur++; |
| 195 | } |
| 196 | |
| 197 | u32 activeIdx = lastActiveIdx; |
| 198 | // If we have top events in the main queue, update current active id |
| 199 | if (q1->cur < q1->end - 1) { |
| 200 | const u32 *baseTop = (const u32 *)((const char *)t + |
| 201 | sizeof(struct Tamarama)); |
| 202 | u32 curTop = q1->items[q1->cur].type; |
| 203 | activeIdx = findEngineForTop(baseTop, curTop, numSubEngines); |
| 204 | } |
| 205 | |
| 206 | assert(activeIdx < numSubEngines); |
| 207 | DEBUG_PRINTF("last id:%u, current id:%u, num of subengines:%u\n" , |
| 208 | lastActiveIdx, activeIdx, numSubEngines); |
| 209 | // Handle unfinished last alive subengine |
| 210 | if (lastActiveIdx != activeIdx && |
| 211 | lastActiveIdx != numSubEngines && hasStart) { |
| 212 | loc = q1->items[q1->cur].location; |
| 213 | pushQueueNoMerge(q2, MQE_END, loc); |
| 214 | q2->nfa = getSubEngine(t, lastActiveIdx); |
| 215 | return; |
| 216 | } |
| 217 | |
| 218 | initSubQueue(t, q1, q2, lastActiveIdx, activeIdx); |
| 219 | DEBUG_PRINTF("finish queues\n" ); |
| 220 | } |
| 221 | |
| 222 | // After processing subqueue items for subengines, we need to copy back |
| 223 | // remaining items in subqueue if there are any to Tamarama main queue |
| 224 | static |
| 225 | void copyBack(const struct Tamarama *t, struct mq *q, struct mq *q1) { |
| 226 | DEBUG_PRINTF("copy back %u, %u\n" , q1->cur, q1->end); |
| 227 | q->report_current = q1->report_current; |
| 228 | if (q->cur >= q->end && q1->cur >= q1->end) { |
| 229 | return; |
| 230 | } |
| 231 | |
| 232 | const u32 *baseTop = (const u32 *)((const char *)t + |
| 233 | sizeof(struct Tamarama)); |
| 234 | const u32 lastIdx = loadActiveIdx(q->streamState, |
| 235 | t->activeIdxSize); |
| 236 | u32 base = 0, event_base = 0; |
| 237 | if (lastIdx != t->numSubEngines) { |
| 238 | base = baseTop[lastIdx]; |
| 239 | const struct NFA *sub = getSubEngine(t, lastIdx); |
| 240 | event_base = isMultiTopType(sub->type) ? MQE_TOP_FIRST : MQE_TOP; |
| 241 | } |
| 242 | |
| 243 | u32 numItems = q1->end > q1->cur + 1 ? q1->end - q1->cur - 1 : 1; |
| 244 | // Also need to copy MQE_END if the main queue is empty |
| 245 | if (q->cur == q->end) { |
| 246 | assert(q->cur > 1 && q1->items[q1->end - 1].type == MQE_END); |
| 247 | q->items[--q->cur] = q1->items[q1->end - 1]; |
| 248 | } |
| 249 | u32 cur = q->cur - numItems; |
| 250 | q->items[cur] = q1->items[q1->cur++]; |
| 251 | q->items[cur].type = MQE_START; |
| 252 | q->cur = cur++; |
| 253 | for (u32 i = 0; i < numItems - 1; ++i) { |
| 254 | assert(q1->cur < q1->end); |
| 255 | u32 type = q1->items[q1->cur].type; |
| 256 | if (type > MQE_END) { |
| 257 | q1->items[q1->cur].type = type - event_base + base; |
| 258 | } |
| 259 | q->items[cur++] = q1->items[q1->cur++]; |
| 260 | } |
| 261 | |
| 262 | #ifdef DEBUG |
| 263 | DEBUG_PRINTF("external queue\n" ); |
| 264 | debugQueue(q); |
| 265 | #endif |
| 266 | } |
| 267 | |
| 268 | char nfaExecTamarama_testEOD(const struct NFA *n, const char *state, |
| 269 | const char *streamState, u64a offset, |
| 270 | NfaCallback callback, void *context) { |
| 271 | const struct Tamarama *t = getImplNfa(n); |
| 272 | u32 activeIdx = loadActiveIdx(streamState, t->activeIdxSize); |
| 273 | if (activeIdx == t->numSubEngines) { |
| 274 | return MO_CONTINUE_MATCHING; |
| 275 | } |
| 276 | |
| 277 | const struct NFA *sub = getSubEngine(t, activeIdx); |
| 278 | if (nfaAcceptsEod(sub)) { |
| 279 | assert(!isContainerType(sub->type)); |
| 280 | const char *subStreamState = streamState + t->activeIdxSize; |
| 281 | return nfaCheckFinalState(sub, state, subStreamState, offset, callback, |
| 282 | context); |
| 283 | } |
| 284 | |
| 285 | return MO_CONTINUE_MATCHING; |
| 286 | } |
| 287 | |
| 288 | char nfaExecTamarama_QR(const struct NFA *n, struct mq *q, ReportID report) { |
| 289 | DEBUG_PRINTF("exec rose\n" ); |
| 290 | struct mq q1; |
| 291 | q1.cur = q1.end = 0; |
| 292 | char rv = 0; |
| 293 | const struct Tamarama *t = getImplNfa(n); |
| 294 | while (q->cur < q->end) { |
| 295 | updateQueues(t, q, &q1); |
| 296 | } |
| 297 | |
| 298 | if (q1.cur < q1.end) { |
| 299 | rv = nfaQueueExecRose(q1.nfa, &q1, report); |
| 300 | } |
| 301 | |
| 302 | DEBUG_PRINTF("exec rose rv:%u\n" , rv); |
| 303 | return rv; |
| 304 | } |
| 305 | |
| 306 | char nfaExecTamarama_reportCurrent(const struct NFA *n, struct mq *q) { |
| 307 | const struct Tamarama *t = getImplNfa(n); |
| 308 | u32 activeIdx = loadActiveIdx(q->streamState, t->activeIdxSize); |
| 309 | if (activeIdx == t->numSubEngines) { |
| 310 | return 1; |
| 311 | } |
| 312 | |
| 313 | const struct NFA *sub = getSubEngine(t, activeIdx); |
| 314 | struct mq q1; |
| 315 | copyQueue(t, sub, q, &q1, activeIdx); |
| 316 | return nfaReportCurrentMatches(sub, &q1); |
| 317 | } |
| 318 | |
| 319 | char nfaExecTamarama_inAccept(const struct NFA *n, ReportID report, |
| 320 | struct mq *q) { |
| 321 | const struct Tamarama *t = getImplNfa(n); |
| 322 | u32 activeIdx = loadActiveIdx(q->streamState, t->activeIdxSize); |
| 323 | if (activeIdx == t->numSubEngines) { |
| 324 | return 0; |
| 325 | } |
| 326 | const struct NFA *sub = getSubEngine(t, activeIdx); |
| 327 | |
| 328 | struct mq q1; |
| 329 | copyQueue(t, sub, q, &q1, activeIdx); |
| 330 | return nfaInAcceptState(sub, report, &q1); |
| 331 | } |
| 332 | |
| 333 | char nfaExecTamarama_inAnyAccept(const struct NFA *n, struct mq *q) { |
| 334 | const struct Tamarama *t = getImplNfa(n); |
| 335 | u32 activeIdx = loadActiveIdx(q->streamState, t->activeIdxSize); |
| 336 | if (activeIdx == t->numSubEngines) { |
| 337 | return 0; |
| 338 | } |
| 339 | const struct NFA *sub = getSubEngine(t, activeIdx); |
| 340 | |
| 341 | struct mq q1; |
| 342 | copyQueue(t, sub, q, &q1, activeIdx); |
| 343 | return nfaInAnyAcceptState(sub, &q1); |
| 344 | } |
| 345 | |
| 346 | char nfaExecTamarama_queueInitState(const struct NFA *n, struct mq *q) { |
| 347 | DEBUG_PRINTF("init state\n" ); |
| 348 | const struct Tamarama *t = getImplNfa(n); |
| 349 | char *ptr = q->streamState; |
| 350 | // Use activeIdxSize as a sentinel value and initialize the state to |
| 351 | // an invalid engine as nothing has been triggered yet |
| 352 | storeActiveIdx(t, ptr, t->numSubEngines); |
| 353 | return 0; |
| 354 | } |
| 355 | |
| 356 | char nfaExecTamarama_queueCompressState(const struct NFA *n, const struct mq *q, |
| 357 | s64a loc) { |
| 358 | const struct Tamarama *t = getImplNfa(n); |
| 359 | u32 activeIdx = loadActiveIdx(q->streamState, t->activeIdxSize); |
| 360 | if (activeIdx == t->numSubEngines) { |
| 361 | return 0; |
| 362 | } |
| 363 | |
| 364 | const struct NFA *sub = getSubEngine(t, activeIdx); |
| 365 | |
| 366 | struct mq q1; |
| 367 | copyQueueProperties(q, &q1, t->activeIdxSize); |
| 368 | return nfaQueueCompressState(sub, &q1, loc); |
| 369 | } |
| 370 | |
| 371 | char nfaExecTamarama_expandState(const struct NFA *n, void *dest, |
| 372 | const void *src, u64a offset, u8 key) { |
| 373 | const struct Tamarama *t = getImplNfa(n); |
| 374 | u32 activeIdx = loadActiveIdx(src, t->activeIdxSize); |
| 375 | if (activeIdx == t->numSubEngines) { |
| 376 | return 0; |
| 377 | } |
| 378 | |
| 379 | const struct NFA *sub = getSubEngine(t, activeIdx); |
| 380 | |
| 381 | const char *subStreamState = (const char *)src + t->activeIdxSize; |
| 382 | return nfaExpandState(sub, dest, subStreamState, offset, key); |
| 383 | } |
| 384 | |
| 385 | enum nfa_zombie_status nfaExecTamarama_zombie_status(const struct NFA *n, |
| 386 | struct mq *q, s64a loc) { |
| 387 | const struct Tamarama *t = getImplNfa(n); |
| 388 | u32 activeIdx = loadActiveIdx(q->streamState, t->activeIdxSize); |
| 389 | if (activeIdx == t->numSubEngines) { |
| 390 | return NFA_ZOMBIE_NO; |
| 391 | } |
| 392 | const struct NFA *sub = getSubEngine(t, activeIdx); |
| 393 | |
| 394 | struct mq q1; |
| 395 | copyQueue(t, sub, q, &q1, activeIdx); |
| 396 | return nfaGetZombieStatus(sub, &q1, loc); |
| 397 | } |
| 398 | |
| 399 | char nfaExecTamarama_Q(const struct NFA *n, struct mq *q, s64a end) { |
| 400 | DEBUG_PRINTF("exec\n" ); |
| 401 | struct mq q1; |
| 402 | char rv = MO_ALIVE; |
| 403 | char copy = 0; |
| 404 | const struct Tamarama *t = getImplNfa(n); |
| 405 | while (q->cur < q->end && q_cur_loc(q) <= end) { |
| 406 | updateQueues(t, q, &q1); |
| 407 | rv = nfaQueueExec_raw(q1.nfa, &q1, end); |
| 408 | q->report_current = q1.report_current; |
| 409 | copy = 1; |
| 410 | if (can_stop_matching(q->scratch)) { |
| 411 | break; |
| 412 | } |
| 413 | } |
| 414 | if (copy) { |
| 415 | copyBack(t, q, &q1); |
| 416 | } |
| 417 | return rv; |
| 418 | } |
| 419 | |
| 420 | char nfaExecTamarama_Q2(const struct NFA *n, struct mq *q, s64a end) { |
| 421 | DEBUG_PRINTF("exec to match\n" ); |
| 422 | struct mq q1; |
| 423 | char rv = 0; |
| 424 | char copy = 0; |
| 425 | const struct Tamarama *t = getImplNfa(n); |
| 426 | while (q->cur < q->end && q_cur_loc(q) <= end && |
| 427 | rv != MO_MATCHES_PENDING) { |
| 428 | updateQueues(t, q, &q1); |
| 429 | rv = nfaQueueExec2_raw(q1.nfa, &q1, end); |
| 430 | q->report_current = q1.report_current; |
| 431 | copy = 1; |
| 432 | if (can_stop_matching(q->scratch)) { |
| 433 | break; |
| 434 | } |
| 435 | } |
| 436 | if (copy) { |
| 437 | copyBack(t, q, &q1); |
| 438 | } |
| 439 | return rv; |
| 440 | } |
| 441 | |
| 442 | |