Scott Baker | eee8dd8 | 2019-09-24 12:52:34 -0700 | [diff] [blame^] | 1 | /* |
| 2 | * Copyright (c) 2016-present, Yann Collet, Facebook, Inc. |
| 3 | * All rights reserved. |
| 4 | * |
| 5 | * This source code is licensed under both the BSD-style license (found in the |
| 6 | * LICENSE file in the root directory of this source tree) and the GPLv2 (found |
| 7 | * in the COPYING file in the root directory of this source tree). |
| 8 | * You may select, at your option, one of the above-listed licenses. |
| 9 | */ |
| 10 | |
| 11 | |
| 12 | /* *************************************************************** |
| 13 | * Tuning parameters |
| 14 | *****************************************************************/ |
| 15 | /*! |
| 16 | * HEAPMODE : |
| 17 | * Select how default decompression function ZSTD_decompress() allocates its context, |
| 18 | * on stack (0), or into heap (1, default; requires malloc()). |
| 19 | * Note that functions with explicit context such as ZSTD_decompressDCtx() are unaffected. |
| 20 | */ |
| 21 | #ifndef ZSTD_HEAPMODE |
| 22 | # define ZSTD_HEAPMODE 1 |
| 23 | #endif |
| 24 | |
| 25 | /*! |
| 26 | * LEGACY_SUPPORT : |
| 27 | * if set to 1+, ZSTD_decompress() can decode older formats (v0.1+) |
| 28 | */ |
| 29 | #ifndef ZSTD_LEGACY_SUPPORT |
| 30 | # define ZSTD_LEGACY_SUPPORT 0 |
| 31 | #endif |
| 32 | |
| 33 | /*! |
| 34 | * MAXWINDOWSIZE_DEFAULT : |
| 35 | * maximum window size accepted by DStream __by default__. |
| 36 | * Frames requiring more memory will be rejected. |
| 37 | * It's possible to set a different limit using ZSTD_DCtx_setMaxWindowSize(). |
| 38 | */ |
| 39 | #ifndef ZSTD_MAXWINDOWSIZE_DEFAULT |
| 40 | # define ZSTD_MAXWINDOWSIZE_DEFAULT (((U32)1 << ZSTD_WINDOWLOG_DEFAULTMAX) + 1) |
| 41 | #endif |
| 42 | |
| 43 | |
| 44 | /*-******************************************************* |
| 45 | * Dependencies |
| 46 | *********************************************************/ |
| 47 | #include <string.h> /* memcpy, memmove, memset */ |
| 48 | #include "cpu.h" |
| 49 | #include "mem.h" /* low level memory routines */ |
| 50 | #define FSE_STATIC_LINKING_ONLY |
| 51 | #include "fse.h" |
| 52 | #define HUF_STATIC_LINKING_ONLY |
| 53 | #include "huf.h" |
| 54 | #include "zstd_internal.h" |
| 55 | |
| 56 | #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT>=1) |
| 57 | # include "zstd_legacy.h" |
| 58 | #endif |
| 59 | |
| 60 | |
| 61 | /*-************************************* |
| 62 | * Errors |
| 63 | ***************************************/ |
| 64 | #define ZSTD_isError ERR_isError /* for inlining */ |
| 65 | #define FSE_isError ERR_isError |
| 66 | #define HUF_isError ERR_isError |
| 67 | |
| 68 | |
| 69 | /*_******************************************************* |
| 70 | * Memory operations |
| 71 | **********************************************************/ |
| 72 | static void ZSTD_copy4(void* dst, const void* src) { memcpy(dst, src, 4); } |
| 73 | |
| 74 | |
| 75 | /*-************************************************************* |
| 76 | * Context management |
| 77 | ***************************************************************/ |
| 78 | typedef enum { ZSTDds_getFrameHeaderSize, ZSTDds_decodeFrameHeader, |
| 79 | ZSTDds_decodeBlockHeader, ZSTDds_decompressBlock, |
| 80 | ZSTDds_decompressLastBlock, ZSTDds_checkChecksum, |
| 81 | ZSTDds_decodeSkippableHeader, ZSTDds_skipFrame } ZSTD_dStage; |
| 82 | |
| 83 | typedef enum { zdss_init=0, zdss_loadHeader, |
| 84 | zdss_read, zdss_load, zdss_flush } ZSTD_dStreamStage; |
| 85 | |
| 86 | |
| 87 | typedef struct { |
| 88 | U32 fastMode; |
| 89 | U32 tableLog; |
| 90 | } ZSTD_seqSymbol_header; |
| 91 | |
| 92 | typedef struct { |
| 93 | U16 nextState; |
| 94 | BYTE nbAdditionalBits; |
| 95 | BYTE nbBits; |
| 96 | U32 baseValue; |
| 97 | } ZSTD_seqSymbol; |
| 98 | |
| 99 | #define SEQSYMBOL_TABLE_SIZE(log) (1 + (1 << (log))) |
| 100 | |
| 101 | typedef struct { |
| 102 | ZSTD_seqSymbol LLTable[SEQSYMBOL_TABLE_SIZE(LLFSELog)]; |
| 103 | ZSTD_seqSymbol OFTable[SEQSYMBOL_TABLE_SIZE(OffFSELog)]; |
| 104 | ZSTD_seqSymbol MLTable[SEQSYMBOL_TABLE_SIZE(MLFSELog)]; |
| 105 | HUF_DTable hufTable[HUF_DTABLE_SIZE(HufLog)]; /* can accommodate HUF_decompress4X */ |
| 106 | U32 workspace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32]; |
| 107 | U32 rep[ZSTD_REP_NUM]; |
| 108 | } ZSTD_entropyDTables_t; |
| 109 | |
| 110 | struct ZSTD_DCtx_s |
| 111 | { |
| 112 | const ZSTD_seqSymbol* LLTptr; |
| 113 | const ZSTD_seqSymbol* MLTptr; |
| 114 | const ZSTD_seqSymbol* OFTptr; |
| 115 | const HUF_DTable* HUFptr; |
| 116 | ZSTD_entropyDTables_t entropy; |
| 117 | const void* previousDstEnd; /* detect continuity */ |
| 118 | const void* base; /* start of current segment */ |
| 119 | const void* vBase; /* virtual start of previous segment if it was just before current one */ |
| 120 | const void* dictEnd; /* end of previous segment */ |
| 121 | size_t expected; |
| 122 | ZSTD_frameHeader fParams; |
| 123 | U64 decodedSize; |
| 124 | blockType_e bType; /* used in ZSTD_decompressContinue(), store blockType between block header decoding and block decompression stages */ |
| 125 | ZSTD_dStage stage; |
| 126 | U32 litEntropy; |
| 127 | U32 fseEntropy; |
| 128 | XXH64_state_t xxhState; |
| 129 | size_t headerSize; |
| 130 | U32 dictID; |
| 131 | ZSTD_format_e format; |
| 132 | const BYTE* litPtr; |
| 133 | ZSTD_customMem customMem; |
| 134 | size_t litSize; |
| 135 | size_t rleSize; |
| 136 | size_t staticSize; |
| 137 | int bmi2; /* == 1 if the CPU supports BMI2 and 0 otherwise. CPU support is determined dynamically once per context lifetime. */ |
| 138 | |
| 139 | /* streaming */ |
| 140 | ZSTD_DDict* ddictLocal; |
| 141 | const ZSTD_DDict* ddict; |
| 142 | ZSTD_dStreamStage streamStage; |
| 143 | char* inBuff; |
| 144 | size_t inBuffSize; |
| 145 | size_t inPos; |
| 146 | size_t maxWindowSize; |
| 147 | char* outBuff; |
| 148 | size_t outBuffSize; |
| 149 | size_t outStart; |
| 150 | size_t outEnd; |
| 151 | size_t lhSize; |
| 152 | void* legacyContext; |
| 153 | U32 previousLegacyVersion; |
| 154 | U32 legacyVersion; |
| 155 | U32 hostageByte; |
| 156 | |
| 157 | /* workspace */ |
| 158 | BYTE litBuffer[ZSTD_BLOCKSIZE_MAX + WILDCOPY_OVERLENGTH]; |
| 159 | BYTE headerBuffer[ZSTD_FRAMEHEADERSIZE_MAX]; |
| 160 | }; /* typedef'd to ZSTD_DCtx within "zstd.h" */ |
| 161 | |
| 162 | size_t ZSTD_sizeof_DCtx (const ZSTD_DCtx* dctx) |
| 163 | { |
| 164 | if (dctx==NULL) return 0; /* support sizeof NULL */ |
| 165 | return sizeof(*dctx) |
| 166 | + ZSTD_sizeof_DDict(dctx->ddictLocal) |
| 167 | + dctx->inBuffSize + dctx->outBuffSize; |
| 168 | } |
| 169 | |
| 170 | size_t ZSTD_estimateDCtxSize(void) { return sizeof(ZSTD_DCtx); } |
| 171 | |
| 172 | |
| 173 | static size_t ZSTD_startingInputLength(ZSTD_format_e format) |
| 174 | { |
| 175 | size_t const startingInputLength = (format==ZSTD_f_zstd1_magicless) ? |
| 176 | ZSTD_frameHeaderSize_prefix - ZSTD_frameIdSize : |
| 177 | ZSTD_frameHeaderSize_prefix; |
| 178 | ZSTD_STATIC_ASSERT(ZSTD_FRAMEHEADERSIZE_PREFIX >= ZSTD_FRAMEIDSIZE); |
| 179 | /* only supports formats ZSTD_f_zstd1 and ZSTD_f_zstd1_magicless */ |
| 180 | assert( (format == ZSTD_f_zstd1) || (format == ZSTD_f_zstd1_magicless) ); |
| 181 | return startingInputLength; |
| 182 | } |
| 183 | |
| 184 | static void ZSTD_initDCtx_internal(ZSTD_DCtx* dctx) |
| 185 | { |
| 186 | dctx->format = ZSTD_f_zstd1; /* ZSTD_decompressBegin() invokes ZSTD_startingInputLength() with argument dctx->format */ |
| 187 | dctx->staticSize = 0; |
| 188 | dctx->maxWindowSize = ZSTD_MAXWINDOWSIZE_DEFAULT; |
| 189 | dctx->ddict = NULL; |
| 190 | dctx->ddictLocal = NULL; |
| 191 | dctx->inBuff = NULL; |
| 192 | dctx->inBuffSize = 0; |
| 193 | dctx->outBuffSize = 0; |
| 194 | dctx->streamStage = zdss_init; |
| 195 | dctx->bmi2 = ZSTD_cpuid_bmi2(ZSTD_cpuid()); |
| 196 | } |
| 197 | |
| 198 | ZSTD_DCtx* ZSTD_initStaticDCtx(void *workspace, size_t workspaceSize) |
| 199 | { |
| 200 | ZSTD_DCtx* const dctx = (ZSTD_DCtx*) workspace; |
| 201 | |
| 202 | if ((size_t)workspace & 7) return NULL; /* 8-aligned */ |
| 203 | if (workspaceSize < sizeof(ZSTD_DCtx)) return NULL; /* minimum size */ |
| 204 | |
| 205 | ZSTD_initDCtx_internal(dctx); |
| 206 | dctx->staticSize = workspaceSize; |
| 207 | dctx->inBuff = (char*)(dctx+1); |
| 208 | return dctx; |
| 209 | } |
| 210 | |
| 211 | ZSTD_DCtx* ZSTD_createDCtx_advanced(ZSTD_customMem customMem) |
| 212 | { |
| 213 | if (!customMem.customAlloc ^ !customMem.customFree) return NULL; |
| 214 | |
| 215 | { ZSTD_DCtx* const dctx = (ZSTD_DCtx*)ZSTD_malloc(sizeof(*dctx), customMem); |
| 216 | if (!dctx) return NULL; |
| 217 | dctx->customMem = customMem; |
| 218 | dctx->legacyContext = NULL; |
| 219 | dctx->previousLegacyVersion = 0; |
| 220 | ZSTD_initDCtx_internal(dctx); |
| 221 | return dctx; |
| 222 | } |
| 223 | } |
| 224 | |
| 225 | ZSTD_DCtx* ZSTD_createDCtx(void) |
| 226 | { |
| 227 | DEBUGLOG(3, "ZSTD_createDCtx"); |
| 228 | return ZSTD_createDCtx_advanced(ZSTD_defaultCMem); |
| 229 | } |
| 230 | |
| 231 | size_t ZSTD_freeDCtx(ZSTD_DCtx* dctx) |
| 232 | { |
| 233 | if (dctx==NULL) return 0; /* support free on NULL */ |
| 234 | if (dctx->staticSize) return ERROR(memory_allocation); /* not compatible with static DCtx */ |
| 235 | { ZSTD_customMem const cMem = dctx->customMem; |
| 236 | ZSTD_freeDDict(dctx->ddictLocal); |
| 237 | dctx->ddictLocal = NULL; |
| 238 | ZSTD_free(dctx->inBuff, cMem); |
| 239 | dctx->inBuff = NULL; |
| 240 | #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT >= 1) |
| 241 | if (dctx->legacyContext) |
| 242 | ZSTD_freeLegacyStreamContext(dctx->legacyContext, dctx->previousLegacyVersion); |
| 243 | #endif |
| 244 | ZSTD_free(dctx, cMem); |
| 245 | return 0; |
| 246 | } |
| 247 | } |
| 248 | |
| 249 | /* no longer useful */ |
| 250 | void ZSTD_copyDCtx(ZSTD_DCtx* dstDCtx, const ZSTD_DCtx* srcDCtx) |
| 251 | { |
| 252 | size_t const toCopy = (size_t)((char*)(&dstDCtx->inBuff) - (char*)dstDCtx); |
| 253 | memcpy(dstDCtx, srcDCtx, toCopy); /* no need to copy workspace */ |
| 254 | } |
| 255 | |
| 256 | |
| 257 | /*-************************************************************* |
| 258 | * Frame header decoding |
| 259 | ***************************************************************/ |
| 260 | |
| 261 | /*! ZSTD_isFrame() : |
| 262 | * Tells if the content of `buffer` starts with a valid Frame Identifier. |
| 263 | * Note : Frame Identifier is 4 bytes. If `size < 4`, @return will always be 0. |
| 264 | * Note 2 : Legacy Frame Identifiers are considered valid only if Legacy Support is enabled. |
| 265 | * Note 3 : Skippable Frame Identifiers are considered valid. */ |
| 266 | unsigned ZSTD_isFrame(const void* buffer, size_t size) |
| 267 | { |
| 268 | if (size < ZSTD_frameIdSize) return 0; |
| 269 | { U32 const magic = MEM_readLE32(buffer); |
| 270 | if (magic == ZSTD_MAGICNUMBER) return 1; |
| 271 | if ((magic & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START) return 1; |
| 272 | } |
| 273 | #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT >= 1) |
| 274 | if (ZSTD_isLegacy(buffer, size)) return 1; |
| 275 | #endif |
| 276 | return 0; |
| 277 | } |
| 278 | |
| 279 | /** ZSTD_frameHeaderSize_internal() : |
| 280 | * srcSize must be large enough to reach header size fields. |
| 281 | * note : only works for formats ZSTD_f_zstd1 and ZSTD_f_zstd1_magicless. |
| 282 | * @return : size of the Frame Header |
| 283 | * or an error code, which can be tested with ZSTD_isError() */ |
| 284 | static size_t ZSTD_frameHeaderSize_internal(const void* src, size_t srcSize, ZSTD_format_e format) |
| 285 | { |
| 286 | size_t const minInputSize = ZSTD_startingInputLength(format); |
| 287 | if (srcSize < minInputSize) return ERROR(srcSize_wrong); |
| 288 | |
| 289 | { BYTE const fhd = ((const BYTE*)src)[minInputSize-1]; |
| 290 | U32 const dictID= fhd & 3; |
| 291 | U32 const singleSegment = (fhd >> 5) & 1; |
| 292 | U32 const fcsId = fhd >> 6; |
| 293 | return minInputSize + !singleSegment |
| 294 | + ZSTD_did_fieldSize[dictID] + ZSTD_fcs_fieldSize[fcsId] |
| 295 | + (singleSegment && !fcsId); |
| 296 | } |
| 297 | } |
| 298 | |
| 299 | /** ZSTD_frameHeaderSize() : |
| 300 | * srcSize must be >= ZSTD_frameHeaderSize_prefix. |
| 301 | * @return : size of the Frame Header */ |
| 302 | size_t ZSTD_frameHeaderSize(const void* src, size_t srcSize) |
| 303 | { |
| 304 | return ZSTD_frameHeaderSize_internal(src, srcSize, ZSTD_f_zstd1); |
| 305 | } |
| 306 | |
| 307 | |
| 308 | /** ZSTD_getFrameHeader_internal() : |
| 309 | * decode Frame Header, or require larger `srcSize`. |
| 310 | * note : only works for formats ZSTD_f_zstd1 and ZSTD_f_zstd1_magicless |
| 311 | * @return : 0, `zfhPtr` is correctly filled, |
| 312 | * >0, `srcSize` is too small, value is wanted `srcSize` amount, |
| 313 | * or an error code, which can be tested using ZSTD_isError() */ |
| 314 | static size_t ZSTD_getFrameHeader_internal(ZSTD_frameHeader* zfhPtr, const void* src, size_t srcSize, ZSTD_format_e format) |
| 315 | { |
| 316 | const BYTE* ip = (const BYTE*)src; |
| 317 | size_t const minInputSize = ZSTD_startingInputLength(format); |
| 318 | |
| 319 | if (srcSize < minInputSize) return minInputSize; |
| 320 | |
| 321 | if ( (format != ZSTD_f_zstd1_magicless) |
| 322 | && (MEM_readLE32(src) != ZSTD_MAGICNUMBER) ) { |
| 323 | if ((MEM_readLE32(src) & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START) { |
| 324 | /* skippable frame */ |
| 325 | if (srcSize < ZSTD_skippableHeaderSize) |
| 326 | return ZSTD_skippableHeaderSize; /* magic number + frame length */ |
| 327 | memset(zfhPtr, 0, sizeof(*zfhPtr)); |
| 328 | zfhPtr->frameContentSize = MEM_readLE32((const char *)src + ZSTD_frameIdSize); |
| 329 | zfhPtr->frameType = ZSTD_skippableFrame; |
| 330 | return 0; |
| 331 | } |
| 332 | return ERROR(prefix_unknown); |
| 333 | } |
| 334 | |
| 335 | /* ensure there is enough `srcSize` to fully read/decode frame header */ |
| 336 | { size_t const fhsize = ZSTD_frameHeaderSize_internal(src, srcSize, format); |
| 337 | if (srcSize < fhsize) return fhsize; |
| 338 | zfhPtr->headerSize = (U32)fhsize; |
| 339 | } |
| 340 | |
| 341 | { BYTE const fhdByte = ip[minInputSize-1]; |
| 342 | size_t pos = minInputSize; |
| 343 | U32 const dictIDSizeCode = fhdByte&3; |
| 344 | U32 const checksumFlag = (fhdByte>>2)&1; |
| 345 | U32 const singleSegment = (fhdByte>>5)&1; |
| 346 | U32 const fcsID = fhdByte>>6; |
| 347 | U64 windowSize = 0; |
| 348 | U32 dictID = 0; |
| 349 | U64 frameContentSize = ZSTD_CONTENTSIZE_UNKNOWN; |
| 350 | if ((fhdByte & 0x08) != 0) |
| 351 | return ERROR(frameParameter_unsupported); /* reserved bits, must be zero */ |
| 352 | |
| 353 | if (!singleSegment) { |
| 354 | BYTE const wlByte = ip[pos++]; |
| 355 | U32 const windowLog = (wlByte >> 3) + ZSTD_WINDOWLOG_ABSOLUTEMIN; |
| 356 | if (windowLog > ZSTD_WINDOWLOG_MAX) |
| 357 | return ERROR(frameParameter_windowTooLarge); |
| 358 | windowSize = (1ULL << windowLog); |
| 359 | windowSize += (windowSize >> 3) * (wlByte&7); |
| 360 | } |
| 361 | switch(dictIDSizeCode) |
| 362 | { |
| 363 | default: assert(0); /* impossible */ |
| 364 | case 0 : break; |
| 365 | case 1 : dictID = ip[pos]; pos++; break; |
| 366 | case 2 : dictID = MEM_readLE16(ip+pos); pos+=2; break; |
| 367 | case 3 : dictID = MEM_readLE32(ip+pos); pos+=4; break; |
| 368 | } |
| 369 | switch(fcsID) |
| 370 | { |
| 371 | default: assert(0); /* impossible */ |
| 372 | case 0 : if (singleSegment) frameContentSize = ip[pos]; break; |
| 373 | case 1 : frameContentSize = MEM_readLE16(ip+pos)+256; break; |
| 374 | case 2 : frameContentSize = MEM_readLE32(ip+pos); break; |
| 375 | case 3 : frameContentSize = MEM_readLE64(ip+pos); break; |
| 376 | } |
| 377 | if (singleSegment) windowSize = frameContentSize; |
| 378 | |
| 379 | zfhPtr->frameType = ZSTD_frame; |
| 380 | zfhPtr->frameContentSize = frameContentSize; |
| 381 | zfhPtr->windowSize = windowSize; |
| 382 | zfhPtr->blockSizeMax = (unsigned) MIN(windowSize, ZSTD_BLOCKSIZE_MAX); |
| 383 | zfhPtr->dictID = dictID; |
| 384 | zfhPtr->checksumFlag = checksumFlag; |
| 385 | } |
| 386 | return 0; |
| 387 | } |
| 388 | |
| 389 | /** ZSTD_getFrameHeader() : |
| 390 | * decode Frame Header, or require larger `srcSize`. |
| 391 | * note : this function does not consume input, it only reads it. |
| 392 | * @return : 0, `zfhPtr` is correctly filled, |
| 393 | * >0, `srcSize` is too small, value is wanted `srcSize` amount, |
| 394 | * or an error code, which can be tested using ZSTD_isError() */ |
| 395 | size_t ZSTD_getFrameHeader(ZSTD_frameHeader* zfhPtr, const void* src, size_t srcSize) |
| 396 | { |
| 397 | return ZSTD_getFrameHeader_internal(zfhPtr, src, srcSize, ZSTD_f_zstd1); |
| 398 | } |
| 399 | |
| 400 | |
| 401 | /** ZSTD_getFrameContentSize() : |
| 402 | * compatible with legacy mode |
| 403 | * @return : decompressed size of the single frame pointed to be `src` if known, otherwise |
| 404 | * - ZSTD_CONTENTSIZE_UNKNOWN if the size cannot be determined |
| 405 | * - ZSTD_CONTENTSIZE_ERROR if an error occurred (e.g. invalid magic number, srcSize too small) */ |
| 406 | unsigned long long ZSTD_getFrameContentSize(const void *src, size_t srcSize) |
| 407 | { |
| 408 | #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT >= 1) |
| 409 | if (ZSTD_isLegacy(src, srcSize)) { |
| 410 | unsigned long long const ret = ZSTD_getDecompressedSize_legacy(src, srcSize); |
| 411 | return ret == 0 ? ZSTD_CONTENTSIZE_UNKNOWN : ret; |
| 412 | } |
| 413 | #endif |
| 414 | { ZSTD_frameHeader zfh; |
| 415 | if (ZSTD_getFrameHeader(&zfh, src, srcSize) != 0) |
| 416 | return ZSTD_CONTENTSIZE_ERROR; |
| 417 | if (zfh.frameType == ZSTD_skippableFrame) { |
| 418 | return 0; |
| 419 | } else { |
| 420 | return zfh.frameContentSize; |
| 421 | } } |
| 422 | } |
| 423 | |
| 424 | /** ZSTD_findDecompressedSize() : |
| 425 | * compatible with legacy mode |
| 426 | * `srcSize` must be the exact length of some number of ZSTD compressed and/or |
| 427 | * skippable frames |
| 428 | * @return : decompressed size of the frames contained */ |
| 429 | unsigned long long ZSTD_findDecompressedSize(const void* src, size_t srcSize) |
| 430 | { |
| 431 | unsigned long long totalDstSize = 0; |
| 432 | |
| 433 | while (srcSize >= ZSTD_frameHeaderSize_prefix) { |
| 434 | U32 const magicNumber = MEM_readLE32(src); |
| 435 | |
| 436 | if ((magicNumber & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START) { |
| 437 | size_t skippableSize; |
| 438 | if (srcSize < ZSTD_skippableHeaderSize) |
| 439 | return ERROR(srcSize_wrong); |
| 440 | skippableSize = MEM_readLE32((const BYTE *)src + ZSTD_frameIdSize) |
| 441 | + ZSTD_skippableHeaderSize; |
| 442 | if (srcSize < skippableSize) { |
| 443 | return ZSTD_CONTENTSIZE_ERROR; |
| 444 | } |
| 445 | |
| 446 | src = (const BYTE *)src + skippableSize; |
| 447 | srcSize -= skippableSize; |
| 448 | continue; |
| 449 | } |
| 450 | |
| 451 | { unsigned long long const ret = ZSTD_getFrameContentSize(src, srcSize); |
| 452 | if (ret >= ZSTD_CONTENTSIZE_ERROR) return ret; |
| 453 | |
| 454 | /* check for overflow */ |
| 455 | if (totalDstSize + ret < totalDstSize) return ZSTD_CONTENTSIZE_ERROR; |
| 456 | totalDstSize += ret; |
| 457 | } |
| 458 | { size_t const frameSrcSize = ZSTD_findFrameCompressedSize(src, srcSize); |
| 459 | if (ZSTD_isError(frameSrcSize)) { |
| 460 | return ZSTD_CONTENTSIZE_ERROR; |
| 461 | } |
| 462 | |
| 463 | src = (const BYTE *)src + frameSrcSize; |
| 464 | srcSize -= frameSrcSize; |
| 465 | } |
| 466 | } /* while (srcSize >= ZSTD_frameHeaderSize_prefix) */ |
| 467 | |
| 468 | if (srcSize) return ZSTD_CONTENTSIZE_ERROR; |
| 469 | |
| 470 | return totalDstSize; |
| 471 | } |
| 472 | |
| 473 | /** ZSTD_getDecompressedSize() : |
| 474 | * compatible with legacy mode |
| 475 | * @return : decompressed size if known, 0 otherwise |
| 476 | note : 0 can mean any of the following : |
| 477 | - frame content is empty |
| 478 | - decompressed size field is not present in frame header |
| 479 | - frame header unknown / not supported |
| 480 | - frame header not complete (`srcSize` too small) */ |
| 481 | unsigned long long ZSTD_getDecompressedSize(const void* src, size_t srcSize) |
| 482 | { |
| 483 | unsigned long long const ret = ZSTD_getFrameContentSize(src, srcSize); |
| 484 | ZSTD_STATIC_ASSERT(ZSTD_CONTENTSIZE_ERROR < ZSTD_CONTENTSIZE_UNKNOWN); |
| 485 | return (ret >= ZSTD_CONTENTSIZE_ERROR) ? 0 : ret; |
| 486 | } |
| 487 | |
| 488 | |
| 489 | /** ZSTD_decodeFrameHeader() : |
| 490 | * `headerSize` must be the size provided by ZSTD_frameHeaderSize(). |
| 491 | * @return : 0 if success, or an error code, which can be tested using ZSTD_isError() */ |
| 492 | static size_t ZSTD_decodeFrameHeader(ZSTD_DCtx* dctx, const void* src, size_t headerSize) |
| 493 | { |
| 494 | size_t const result = ZSTD_getFrameHeader_internal(&(dctx->fParams), src, headerSize, dctx->format); |
| 495 | if (ZSTD_isError(result)) return result; /* invalid header */ |
| 496 | if (result>0) return ERROR(srcSize_wrong); /* headerSize too small */ |
| 497 | if (dctx->fParams.dictID && (dctx->dictID != dctx->fParams.dictID)) |
| 498 | return ERROR(dictionary_wrong); |
| 499 | if (dctx->fParams.checksumFlag) XXH64_reset(&dctx->xxhState, 0); |
| 500 | return 0; |
| 501 | } |
| 502 | |
| 503 | |
| 504 | /*-************************************************************* |
| 505 | * Block decoding |
| 506 | ***************************************************************/ |
| 507 | |
| 508 | /*! ZSTD_getcBlockSize() : |
| 509 | * Provides the size of compressed block from block header `src` */ |
| 510 | size_t ZSTD_getcBlockSize(const void* src, size_t srcSize, |
| 511 | blockProperties_t* bpPtr) |
| 512 | { |
| 513 | if (srcSize < ZSTD_blockHeaderSize) return ERROR(srcSize_wrong); |
| 514 | { U32 const cBlockHeader = MEM_readLE24(src); |
| 515 | U32 const cSize = cBlockHeader >> 3; |
| 516 | bpPtr->lastBlock = cBlockHeader & 1; |
| 517 | bpPtr->blockType = (blockType_e)((cBlockHeader >> 1) & 3); |
| 518 | bpPtr->origSize = cSize; /* only useful for RLE */ |
| 519 | if (bpPtr->blockType == bt_rle) return 1; |
| 520 | if (bpPtr->blockType == bt_reserved) return ERROR(corruption_detected); |
| 521 | return cSize; |
| 522 | } |
| 523 | } |
| 524 | |
| 525 | |
| 526 | static size_t ZSTD_copyRawBlock(void* dst, size_t dstCapacity, |
| 527 | const void* src, size_t srcSize) |
| 528 | { |
| 529 | if (srcSize > dstCapacity) return ERROR(dstSize_tooSmall); |
| 530 | memcpy(dst, src, srcSize); |
| 531 | return srcSize; |
| 532 | } |
| 533 | |
| 534 | |
| 535 | static size_t ZSTD_setRleBlock(void* dst, size_t dstCapacity, |
| 536 | const void* src, size_t srcSize, |
| 537 | size_t regenSize) |
| 538 | { |
| 539 | if (srcSize != 1) return ERROR(srcSize_wrong); |
| 540 | if (regenSize > dstCapacity) return ERROR(dstSize_tooSmall); |
| 541 | memset(dst, *(const BYTE*)src, regenSize); |
| 542 | return regenSize; |
| 543 | } |
| 544 | |
| 545 | /*! ZSTD_decodeLiteralsBlock() : |
| 546 | * @return : nb of bytes read from src (< srcSize ) |
| 547 | * note : symbol not declared but exposed for fullbench */ |
| 548 | size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx, |
| 549 | const void* src, size_t srcSize) /* note : srcSize < BLOCKSIZE */ |
| 550 | { |
| 551 | if (srcSize < MIN_CBLOCK_SIZE) return ERROR(corruption_detected); |
| 552 | |
| 553 | { const BYTE* const istart = (const BYTE*) src; |
| 554 | symbolEncodingType_e const litEncType = (symbolEncodingType_e)(istart[0] & 3); |
| 555 | |
| 556 | switch(litEncType) |
| 557 | { |
| 558 | case set_repeat: |
| 559 | if (dctx->litEntropy==0) return ERROR(dictionary_corrupted); |
| 560 | /* fall-through */ |
| 561 | case set_compressed: |
| 562 | if (srcSize < 5) return ERROR(corruption_detected); /* srcSize >= MIN_CBLOCK_SIZE == 3; here we need up to 5 for case 3 */ |
| 563 | { size_t lhSize, litSize, litCSize; |
| 564 | U32 singleStream=0; |
| 565 | U32 const lhlCode = (istart[0] >> 2) & 3; |
| 566 | U32 const lhc = MEM_readLE32(istart); |
| 567 | switch(lhlCode) |
| 568 | { |
| 569 | case 0: case 1: default: /* note : default is impossible, since lhlCode into [0..3] */ |
| 570 | /* 2 - 2 - 10 - 10 */ |
| 571 | singleStream = !lhlCode; |
| 572 | lhSize = 3; |
| 573 | litSize = (lhc >> 4) & 0x3FF; |
| 574 | litCSize = (lhc >> 14) & 0x3FF; |
| 575 | break; |
| 576 | case 2: |
| 577 | /* 2 - 2 - 14 - 14 */ |
| 578 | lhSize = 4; |
| 579 | litSize = (lhc >> 4) & 0x3FFF; |
| 580 | litCSize = lhc >> 18; |
| 581 | break; |
| 582 | case 3: |
| 583 | /* 2 - 2 - 18 - 18 */ |
| 584 | lhSize = 5; |
| 585 | litSize = (lhc >> 4) & 0x3FFFF; |
| 586 | litCSize = (lhc >> 22) + (istart[4] << 10); |
| 587 | break; |
| 588 | } |
| 589 | if (litSize > ZSTD_BLOCKSIZE_MAX) return ERROR(corruption_detected); |
| 590 | if (litCSize + lhSize > srcSize) return ERROR(corruption_detected); |
| 591 | |
| 592 | if (HUF_isError((litEncType==set_repeat) ? |
| 593 | ( singleStream ? |
| 594 | HUF_decompress1X_usingDTable_bmi2(dctx->litBuffer, litSize, istart+lhSize, litCSize, dctx->HUFptr, dctx->bmi2) : |
| 595 | HUF_decompress4X_usingDTable_bmi2(dctx->litBuffer, litSize, istart+lhSize, litCSize, dctx->HUFptr, dctx->bmi2) ) : |
| 596 | ( singleStream ? |
| 597 | HUF_decompress1X2_DCtx_wksp_bmi2(dctx->entropy.hufTable, dctx->litBuffer, litSize, istart+lhSize, litCSize, |
| 598 | dctx->entropy.workspace, sizeof(dctx->entropy.workspace), dctx->bmi2) : |
| 599 | HUF_decompress4X_hufOnly_wksp_bmi2(dctx->entropy.hufTable, dctx->litBuffer, litSize, istart+lhSize, litCSize, |
| 600 | dctx->entropy.workspace, sizeof(dctx->entropy.workspace), dctx->bmi2)))) |
| 601 | return ERROR(corruption_detected); |
| 602 | |
| 603 | dctx->litPtr = dctx->litBuffer; |
| 604 | dctx->litSize = litSize; |
| 605 | dctx->litEntropy = 1; |
| 606 | if (litEncType==set_compressed) dctx->HUFptr = dctx->entropy.hufTable; |
| 607 | memset(dctx->litBuffer + dctx->litSize, 0, WILDCOPY_OVERLENGTH); |
| 608 | return litCSize + lhSize; |
| 609 | } |
| 610 | |
| 611 | case set_basic: |
| 612 | { size_t litSize, lhSize; |
| 613 | U32 const lhlCode = ((istart[0]) >> 2) & 3; |
| 614 | switch(lhlCode) |
| 615 | { |
| 616 | case 0: case 2: default: /* note : default is impossible, since lhlCode into [0..3] */ |
| 617 | lhSize = 1; |
| 618 | litSize = istart[0] >> 3; |
| 619 | break; |
| 620 | case 1: |
| 621 | lhSize = 2; |
| 622 | litSize = MEM_readLE16(istart) >> 4; |
| 623 | break; |
| 624 | case 3: |
| 625 | lhSize = 3; |
| 626 | litSize = MEM_readLE24(istart) >> 4; |
| 627 | break; |
| 628 | } |
| 629 | |
| 630 | if (lhSize+litSize+WILDCOPY_OVERLENGTH > srcSize) { /* risk reading beyond src buffer with wildcopy */ |
| 631 | if (litSize+lhSize > srcSize) return ERROR(corruption_detected); |
| 632 | memcpy(dctx->litBuffer, istart+lhSize, litSize); |
| 633 | dctx->litPtr = dctx->litBuffer; |
| 634 | dctx->litSize = litSize; |
| 635 | memset(dctx->litBuffer + dctx->litSize, 0, WILDCOPY_OVERLENGTH); |
| 636 | return lhSize+litSize; |
| 637 | } |
| 638 | /* direct reference into compressed stream */ |
| 639 | dctx->litPtr = istart+lhSize; |
| 640 | dctx->litSize = litSize; |
| 641 | return lhSize+litSize; |
| 642 | } |
| 643 | |
| 644 | case set_rle: |
| 645 | { U32 const lhlCode = ((istart[0]) >> 2) & 3; |
| 646 | size_t litSize, lhSize; |
| 647 | switch(lhlCode) |
| 648 | { |
| 649 | case 0: case 2: default: /* note : default is impossible, since lhlCode into [0..3] */ |
| 650 | lhSize = 1; |
| 651 | litSize = istart[0] >> 3; |
| 652 | break; |
| 653 | case 1: |
| 654 | lhSize = 2; |
| 655 | litSize = MEM_readLE16(istart) >> 4; |
| 656 | break; |
| 657 | case 3: |
| 658 | lhSize = 3; |
| 659 | litSize = MEM_readLE24(istart) >> 4; |
| 660 | if (srcSize<4) return ERROR(corruption_detected); /* srcSize >= MIN_CBLOCK_SIZE == 3; here we need lhSize+1 = 4 */ |
| 661 | break; |
| 662 | } |
| 663 | if (litSize > ZSTD_BLOCKSIZE_MAX) return ERROR(corruption_detected); |
| 664 | memset(dctx->litBuffer, istart[lhSize], litSize + WILDCOPY_OVERLENGTH); |
| 665 | dctx->litPtr = dctx->litBuffer; |
| 666 | dctx->litSize = litSize; |
| 667 | return lhSize+1; |
| 668 | } |
| 669 | default: |
| 670 | return ERROR(corruption_detected); /* impossible */ |
| 671 | } |
| 672 | } |
| 673 | } |
| 674 | |
| 675 | /* Default FSE distribution tables. |
| 676 | * These are pre-calculated FSE decoding tables using default distributions as defined in specification : |
| 677 | * https://github.com/facebook/zstd/blob/master/doc/zstd_compression_format.md#default-distributions |
| 678 | * They were generated programmatically with following method : |
| 679 | * - start from default distributions, present in /lib/common/zstd_internal.h |
| 680 | * - generate tables normally, using ZSTD_buildFSETable() |
| 681 | * - printout the content of tables |
| 682 | * - pretify output, report below, test with fuzzer to ensure it's correct */ |
| 683 | |
| 684 | /* Default FSE distribution table for Literal Lengths */ |
| 685 | static const ZSTD_seqSymbol LL_defaultDTable[(1<<LL_DEFAULTNORMLOG)+1] = { |
| 686 | { 1, 1, 1, LL_DEFAULTNORMLOG}, /* header : fastMode, tableLog */ |
| 687 | /* nextState, nbAddBits, nbBits, baseVal */ |
| 688 | { 0, 0, 4, 0}, { 16, 0, 4, 0}, |
| 689 | { 32, 0, 5, 1}, { 0, 0, 5, 3}, |
| 690 | { 0, 0, 5, 4}, { 0, 0, 5, 6}, |
| 691 | { 0, 0, 5, 7}, { 0, 0, 5, 9}, |
| 692 | { 0, 0, 5, 10}, { 0, 0, 5, 12}, |
| 693 | { 0, 0, 6, 14}, { 0, 1, 5, 16}, |
| 694 | { 0, 1, 5, 20}, { 0, 1, 5, 22}, |
| 695 | { 0, 2, 5, 28}, { 0, 3, 5, 32}, |
| 696 | { 0, 4, 5, 48}, { 32, 6, 5, 64}, |
| 697 | { 0, 7, 5, 128}, { 0, 8, 6, 256}, |
| 698 | { 0, 10, 6, 1024}, { 0, 12, 6, 4096}, |
| 699 | { 32, 0, 4, 0}, { 0, 0, 4, 1}, |
| 700 | { 0, 0, 5, 2}, { 32, 0, 5, 4}, |
| 701 | { 0, 0, 5, 5}, { 32, 0, 5, 7}, |
| 702 | { 0, 0, 5, 8}, { 32, 0, 5, 10}, |
| 703 | { 0, 0, 5, 11}, { 0, 0, 6, 13}, |
| 704 | { 32, 1, 5, 16}, { 0, 1, 5, 18}, |
| 705 | { 32, 1, 5, 22}, { 0, 2, 5, 24}, |
| 706 | { 32, 3, 5, 32}, { 0, 3, 5, 40}, |
| 707 | { 0, 6, 4, 64}, { 16, 6, 4, 64}, |
| 708 | { 32, 7, 5, 128}, { 0, 9, 6, 512}, |
| 709 | { 0, 11, 6, 2048}, { 48, 0, 4, 0}, |
| 710 | { 16, 0, 4, 1}, { 32, 0, 5, 2}, |
| 711 | { 32, 0, 5, 3}, { 32, 0, 5, 5}, |
| 712 | { 32, 0, 5, 6}, { 32, 0, 5, 8}, |
| 713 | { 32, 0, 5, 9}, { 32, 0, 5, 11}, |
| 714 | { 32, 0, 5, 12}, { 0, 0, 6, 15}, |
| 715 | { 32, 1, 5, 18}, { 32, 1, 5, 20}, |
| 716 | { 32, 2, 5, 24}, { 32, 2, 5, 28}, |
| 717 | { 32, 3, 5, 40}, { 32, 4, 5, 48}, |
| 718 | { 0, 16, 6,65536}, { 0, 15, 6,32768}, |
| 719 | { 0, 14, 6,16384}, { 0, 13, 6, 8192}, |
| 720 | }; /* LL_defaultDTable */ |
| 721 | |
| 722 | /* Default FSE distribution table for Offset Codes */ |
| 723 | static const ZSTD_seqSymbol OF_defaultDTable[(1<<OF_DEFAULTNORMLOG)+1] = { |
| 724 | { 1, 1, 1, OF_DEFAULTNORMLOG}, /* header : fastMode, tableLog */ |
| 725 | /* nextState, nbAddBits, nbBits, baseVal */ |
| 726 | { 0, 0, 5, 0}, { 0, 6, 4, 61}, |
| 727 | { 0, 9, 5, 509}, { 0, 15, 5,32765}, |
| 728 | { 0, 21, 5,2097149}, { 0, 3, 5, 5}, |
| 729 | { 0, 7, 4, 125}, { 0, 12, 5, 4093}, |
| 730 | { 0, 18, 5,262141}, { 0, 23, 5,8388605}, |
| 731 | { 0, 5, 5, 29}, { 0, 8, 4, 253}, |
| 732 | { 0, 14, 5,16381}, { 0, 20, 5,1048573}, |
| 733 | { 0, 2, 5, 1}, { 16, 7, 4, 125}, |
| 734 | { 0, 11, 5, 2045}, { 0, 17, 5,131069}, |
| 735 | { 0, 22, 5,4194301}, { 0, 4, 5, 13}, |
| 736 | { 16, 8, 4, 253}, { 0, 13, 5, 8189}, |
| 737 | { 0, 19, 5,524285}, { 0, 1, 5, 1}, |
| 738 | { 16, 6, 4, 61}, { 0, 10, 5, 1021}, |
| 739 | { 0, 16, 5,65533}, { 0, 28, 5,268435453}, |
| 740 | { 0, 27, 5,134217725}, { 0, 26, 5,67108861}, |
| 741 | { 0, 25, 5,33554429}, { 0, 24, 5,16777213}, |
| 742 | }; /* OF_defaultDTable */ |
| 743 | |
| 744 | |
| 745 | /* Default FSE distribution table for Match Lengths */ |
| 746 | static const ZSTD_seqSymbol ML_defaultDTable[(1<<ML_DEFAULTNORMLOG)+1] = { |
| 747 | { 1, 1, 1, ML_DEFAULTNORMLOG}, /* header : fastMode, tableLog */ |
| 748 | /* nextState, nbAddBits, nbBits, baseVal */ |
| 749 | { 0, 0, 6, 3}, { 0, 0, 4, 4}, |
| 750 | { 32, 0, 5, 5}, { 0, 0, 5, 6}, |
| 751 | { 0, 0, 5, 8}, { 0, 0, 5, 9}, |
| 752 | { 0, 0, 5, 11}, { 0, 0, 6, 13}, |
| 753 | { 0, 0, 6, 16}, { 0, 0, 6, 19}, |
| 754 | { 0, 0, 6, 22}, { 0, 0, 6, 25}, |
| 755 | { 0, 0, 6, 28}, { 0, 0, 6, 31}, |
| 756 | { 0, 0, 6, 34}, { 0, 1, 6, 37}, |
| 757 | { 0, 1, 6, 41}, { 0, 2, 6, 47}, |
| 758 | { 0, 3, 6, 59}, { 0, 4, 6, 83}, |
| 759 | { 0, 7, 6, 131}, { 0, 9, 6, 515}, |
| 760 | { 16, 0, 4, 4}, { 0, 0, 4, 5}, |
| 761 | { 32, 0, 5, 6}, { 0, 0, 5, 7}, |
| 762 | { 32, 0, 5, 9}, { 0, 0, 5, 10}, |
| 763 | { 0, 0, 6, 12}, { 0, 0, 6, 15}, |
| 764 | { 0, 0, 6, 18}, { 0, 0, 6, 21}, |
| 765 | { 0, 0, 6, 24}, { 0, 0, 6, 27}, |
| 766 | { 0, 0, 6, 30}, { 0, 0, 6, 33}, |
| 767 | { 0, 1, 6, 35}, { 0, 1, 6, 39}, |
| 768 | { 0, 2, 6, 43}, { 0, 3, 6, 51}, |
| 769 | { 0, 4, 6, 67}, { 0, 5, 6, 99}, |
| 770 | { 0, 8, 6, 259}, { 32, 0, 4, 4}, |
| 771 | { 48, 0, 4, 4}, { 16, 0, 4, 5}, |
| 772 | { 32, 0, 5, 7}, { 32, 0, 5, 8}, |
| 773 | { 32, 0, 5, 10}, { 32, 0, 5, 11}, |
| 774 | { 0, 0, 6, 14}, { 0, 0, 6, 17}, |
| 775 | { 0, 0, 6, 20}, { 0, 0, 6, 23}, |
| 776 | { 0, 0, 6, 26}, { 0, 0, 6, 29}, |
| 777 | { 0, 0, 6, 32}, { 0, 16, 6,65539}, |
| 778 | { 0, 15, 6,32771}, { 0, 14, 6,16387}, |
| 779 | { 0, 13, 6, 8195}, { 0, 12, 6, 4099}, |
| 780 | { 0, 11, 6, 2051}, { 0, 10, 6, 1027}, |
| 781 | }; /* ML_defaultDTable */ |
| 782 | |
| 783 | |
| 784 | static void ZSTD_buildSeqTable_rle(ZSTD_seqSymbol* dt, U32 baseValue, U32 nbAddBits) |
| 785 | { |
| 786 | void* ptr = dt; |
| 787 | ZSTD_seqSymbol_header* const DTableH = (ZSTD_seqSymbol_header*)ptr; |
| 788 | ZSTD_seqSymbol* const cell = dt + 1; |
| 789 | |
| 790 | DTableH->tableLog = 0; |
| 791 | DTableH->fastMode = 0; |
| 792 | |
| 793 | cell->nbBits = 0; |
| 794 | cell->nextState = 0; |
| 795 | assert(nbAddBits < 255); |
| 796 | cell->nbAdditionalBits = (BYTE)nbAddBits; |
| 797 | cell->baseValue = baseValue; |
| 798 | } |
| 799 | |
| 800 | |
| 801 | /* ZSTD_buildFSETable() : |
| 802 | * generate FSE decoding table for one symbol (ll, ml or off) */ |
| 803 | static void |
| 804 | ZSTD_buildFSETable(ZSTD_seqSymbol* dt, |
| 805 | const short* normalizedCounter, unsigned maxSymbolValue, |
| 806 | const U32* baseValue, const U32* nbAdditionalBits, |
| 807 | unsigned tableLog) |
| 808 | { |
| 809 | ZSTD_seqSymbol* const tableDecode = dt+1; |
| 810 | U16 symbolNext[MaxSeq+1]; |
| 811 | |
| 812 | U32 const maxSV1 = maxSymbolValue + 1; |
| 813 | U32 const tableSize = 1 << tableLog; |
| 814 | U32 highThreshold = tableSize-1; |
| 815 | |
| 816 | /* Sanity Checks */ |
| 817 | assert(maxSymbolValue <= MaxSeq); |
| 818 | assert(tableLog <= MaxFSELog); |
| 819 | |
| 820 | /* Init, lay down lowprob symbols */ |
| 821 | { ZSTD_seqSymbol_header DTableH; |
| 822 | DTableH.tableLog = tableLog; |
| 823 | DTableH.fastMode = 1; |
| 824 | { S16 const largeLimit= (S16)(1 << (tableLog-1)); |
| 825 | U32 s; |
| 826 | for (s=0; s<maxSV1; s++) { |
| 827 | if (normalizedCounter[s]==-1) { |
| 828 | tableDecode[highThreshold--].baseValue = s; |
| 829 | symbolNext[s] = 1; |
| 830 | } else { |
| 831 | if (normalizedCounter[s] >= largeLimit) DTableH.fastMode=0; |
| 832 | symbolNext[s] = normalizedCounter[s]; |
| 833 | } } } |
| 834 | memcpy(dt, &DTableH, sizeof(DTableH)); |
| 835 | } |
| 836 | |
| 837 | /* Spread symbols */ |
| 838 | { U32 const tableMask = tableSize-1; |
| 839 | U32 const step = FSE_TABLESTEP(tableSize); |
| 840 | U32 s, position = 0; |
| 841 | for (s=0; s<maxSV1; s++) { |
| 842 | int i; |
| 843 | for (i=0; i<normalizedCounter[s]; i++) { |
| 844 | tableDecode[position].baseValue = s; |
| 845 | position = (position + step) & tableMask; |
| 846 | while (position > highThreshold) position = (position + step) & tableMask; /* lowprob area */ |
| 847 | } } |
| 848 | assert(position == 0); /* position must reach all cells once, otherwise normalizedCounter is incorrect */ |
| 849 | } |
| 850 | |
| 851 | /* Build Decoding table */ |
| 852 | { U32 u; |
| 853 | for (u=0; u<tableSize; u++) { |
| 854 | U32 const symbol = tableDecode[u].baseValue; |
| 855 | U32 const nextState = symbolNext[symbol]++; |
| 856 | tableDecode[u].nbBits = (BYTE) (tableLog - BIT_highbit32(nextState) ); |
| 857 | tableDecode[u].nextState = (U16) ( (nextState << tableDecode[u].nbBits) - tableSize); |
| 858 | assert(nbAdditionalBits[symbol] < 255); |
| 859 | tableDecode[u].nbAdditionalBits = (BYTE)nbAdditionalBits[symbol]; |
| 860 | tableDecode[u].baseValue = baseValue[symbol]; |
| 861 | } } |
| 862 | } |
| 863 | |
| 864 | |
| 865 | /*! ZSTD_buildSeqTable() : |
| 866 | * @return : nb bytes read from src, |
| 867 | * or an error code if it fails */ |
| 868 | static size_t ZSTD_buildSeqTable(ZSTD_seqSymbol* DTableSpace, const ZSTD_seqSymbol** DTablePtr, |
| 869 | symbolEncodingType_e type, U32 max, U32 maxLog, |
| 870 | const void* src, size_t srcSize, |
| 871 | const U32* baseValue, const U32* nbAdditionalBits, |
| 872 | const ZSTD_seqSymbol* defaultTable, U32 flagRepeatTable) |
| 873 | { |
| 874 | switch(type) |
| 875 | { |
| 876 | case set_rle : |
| 877 | if (!srcSize) return ERROR(srcSize_wrong); |
| 878 | if ( (*(const BYTE*)src) > max) return ERROR(corruption_detected); |
| 879 | { U32 const symbol = *(const BYTE*)src; |
| 880 | U32 const baseline = baseValue[symbol]; |
| 881 | U32 const nbBits = nbAdditionalBits[symbol]; |
| 882 | ZSTD_buildSeqTable_rle(DTableSpace, baseline, nbBits); |
| 883 | } |
| 884 | *DTablePtr = DTableSpace; |
| 885 | return 1; |
| 886 | case set_basic : |
| 887 | *DTablePtr = defaultTable; |
| 888 | return 0; |
| 889 | case set_repeat: |
| 890 | if (!flagRepeatTable) return ERROR(corruption_detected); |
| 891 | return 0; |
| 892 | case set_compressed : |
| 893 | { U32 tableLog; |
| 894 | S16 norm[MaxSeq+1]; |
| 895 | size_t const headerSize = FSE_readNCount(norm, &max, &tableLog, src, srcSize); |
| 896 | if (FSE_isError(headerSize)) return ERROR(corruption_detected); |
| 897 | if (tableLog > maxLog) return ERROR(corruption_detected); |
| 898 | ZSTD_buildFSETable(DTableSpace, norm, max, baseValue, nbAdditionalBits, tableLog); |
| 899 | *DTablePtr = DTableSpace; |
| 900 | return headerSize; |
| 901 | } |
| 902 | default : /* impossible */ |
| 903 | assert(0); |
| 904 | return ERROR(GENERIC); |
| 905 | } |
| 906 | } |
| 907 | |
| 908 | static const U32 LL_base[MaxLL+1] = { |
| 909 | 0, 1, 2, 3, 4, 5, 6, 7, |
| 910 | 8, 9, 10, 11, 12, 13, 14, 15, |
| 911 | 16, 18, 20, 22, 24, 28, 32, 40, |
| 912 | 48, 64, 0x80, 0x100, 0x200, 0x400, 0x800, 0x1000, |
| 913 | 0x2000, 0x4000, 0x8000, 0x10000 }; |
| 914 | |
| 915 | static const U32 OF_base[MaxOff+1] = { |
| 916 | 0, 1, 1, 5, 0xD, 0x1D, 0x3D, 0x7D, |
| 917 | 0xFD, 0x1FD, 0x3FD, 0x7FD, 0xFFD, 0x1FFD, 0x3FFD, 0x7FFD, |
| 918 | 0xFFFD, 0x1FFFD, 0x3FFFD, 0x7FFFD, 0xFFFFD, 0x1FFFFD, 0x3FFFFD, 0x7FFFFD, |
| 919 | 0xFFFFFD, 0x1FFFFFD, 0x3FFFFFD, 0x7FFFFFD, 0xFFFFFFD, 0x1FFFFFFD, 0x3FFFFFFD, 0x7FFFFFFD }; |
| 920 | |
| 921 | static const U32 OF_bits[MaxOff+1] = { |
| 922 | 0, 1, 2, 3, 4, 5, 6, 7, |
| 923 | 8, 9, 10, 11, 12, 13, 14, 15, |
| 924 | 16, 17, 18, 19, 20, 21, 22, 23, |
| 925 | 24, 25, 26, 27, 28, 29, 30, 31 }; |
| 926 | |
| 927 | static const U32 ML_base[MaxML+1] = { |
| 928 | 3, 4, 5, 6, 7, 8, 9, 10, |
| 929 | 11, 12, 13, 14, 15, 16, 17, 18, |
| 930 | 19, 20, 21, 22, 23, 24, 25, 26, |
| 931 | 27, 28, 29, 30, 31, 32, 33, 34, |
| 932 | 35, 37, 39, 41, 43, 47, 51, 59, |
| 933 | 67, 83, 99, 0x83, 0x103, 0x203, 0x403, 0x803, |
| 934 | 0x1003, 0x2003, 0x4003, 0x8003, 0x10003 }; |
| 935 | |
| 936 | |
| 937 | size_t ZSTD_decodeSeqHeaders(ZSTD_DCtx* dctx, int* nbSeqPtr, |
| 938 | const void* src, size_t srcSize) |
| 939 | { |
| 940 | const BYTE* const istart = (const BYTE* const)src; |
| 941 | const BYTE* const iend = istart + srcSize; |
| 942 | const BYTE* ip = istart; |
| 943 | DEBUGLOG(5, "ZSTD_decodeSeqHeaders"); |
| 944 | |
| 945 | /* check */ |
| 946 | if (srcSize < MIN_SEQUENCES_SIZE) return ERROR(srcSize_wrong); |
| 947 | |
| 948 | /* SeqHead */ |
| 949 | { int nbSeq = *ip++; |
| 950 | if (!nbSeq) { *nbSeqPtr=0; return 1; } |
| 951 | if (nbSeq > 0x7F) { |
| 952 | if (nbSeq == 0xFF) { |
| 953 | if (ip+2 > iend) return ERROR(srcSize_wrong); |
| 954 | nbSeq = MEM_readLE16(ip) + LONGNBSEQ, ip+=2; |
| 955 | } else { |
| 956 | if (ip >= iend) return ERROR(srcSize_wrong); |
| 957 | nbSeq = ((nbSeq-0x80)<<8) + *ip++; |
| 958 | } |
| 959 | } |
| 960 | *nbSeqPtr = nbSeq; |
| 961 | } |
| 962 | |
| 963 | /* FSE table descriptors */ |
| 964 | if (ip+4 > iend) return ERROR(srcSize_wrong); /* minimum possible size */ |
| 965 | { symbolEncodingType_e const LLtype = (symbolEncodingType_e)(*ip >> 6); |
| 966 | symbolEncodingType_e const OFtype = (symbolEncodingType_e)((*ip >> 4) & 3); |
| 967 | symbolEncodingType_e const MLtype = (symbolEncodingType_e)((*ip >> 2) & 3); |
| 968 | ip++; |
| 969 | |
| 970 | /* Build DTables */ |
| 971 | { size_t const llhSize = ZSTD_buildSeqTable(dctx->entropy.LLTable, &dctx->LLTptr, |
| 972 | LLtype, MaxLL, LLFSELog, |
| 973 | ip, iend-ip, |
| 974 | LL_base, LL_bits, |
| 975 | LL_defaultDTable, dctx->fseEntropy); |
| 976 | if (ZSTD_isError(llhSize)) return ERROR(corruption_detected); |
| 977 | ip += llhSize; |
| 978 | } |
| 979 | |
| 980 | { size_t const ofhSize = ZSTD_buildSeqTable(dctx->entropy.OFTable, &dctx->OFTptr, |
| 981 | OFtype, MaxOff, OffFSELog, |
| 982 | ip, iend-ip, |
| 983 | OF_base, OF_bits, |
| 984 | OF_defaultDTable, dctx->fseEntropy); |
| 985 | if (ZSTD_isError(ofhSize)) return ERROR(corruption_detected); |
| 986 | ip += ofhSize; |
| 987 | } |
| 988 | |
| 989 | { size_t const mlhSize = ZSTD_buildSeqTable(dctx->entropy.MLTable, &dctx->MLTptr, |
| 990 | MLtype, MaxML, MLFSELog, |
| 991 | ip, iend-ip, |
| 992 | ML_base, ML_bits, |
| 993 | ML_defaultDTable, dctx->fseEntropy); |
| 994 | if (ZSTD_isError(mlhSize)) return ERROR(corruption_detected); |
| 995 | ip += mlhSize; |
| 996 | } |
| 997 | } |
| 998 | |
| 999 | return ip-istart; |
| 1000 | } |
| 1001 | |
| 1002 | |
| 1003 | typedef struct { |
| 1004 | size_t litLength; |
| 1005 | size_t matchLength; |
| 1006 | size_t offset; |
| 1007 | const BYTE* match; |
| 1008 | } seq_t; |
| 1009 | |
| 1010 | typedef struct { |
| 1011 | size_t state; |
| 1012 | const ZSTD_seqSymbol* table; |
| 1013 | } ZSTD_fseState; |
| 1014 | |
| 1015 | typedef struct { |
| 1016 | BIT_DStream_t DStream; |
| 1017 | ZSTD_fseState stateLL; |
| 1018 | ZSTD_fseState stateOffb; |
| 1019 | ZSTD_fseState stateML; |
| 1020 | size_t prevOffset[ZSTD_REP_NUM]; |
| 1021 | const BYTE* prefixStart; |
| 1022 | const BYTE* dictEnd; |
| 1023 | size_t pos; |
| 1024 | } seqState_t; |
| 1025 | |
| 1026 | |
| 1027 | FORCE_NOINLINE |
| 1028 | size_t ZSTD_execSequenceLast7(BYTE* op, |
| 1029 | BYTE* const oend, seq_t sequence, |
| 1030 | const BYTE** litPtr, const BYTE* const litLimit, |
| 1031 | const BYTE* const base, const BYTE* const vBase, const BYTE* const dictEnd) |
| 1032 | { |
| 1033 | BYTE* const oLitEnd = op + sequence.litLength; |
| 1034 | size_t const sequenceLength = sequence.litLength + sequence.matchLength; |
| 1035 | BYTE* const oMatchEnd = op + sequenceLength; /* risk : address space overflow (32-bits) */ |
| 1036 | BYTE* const oend_w = oend - WILDCOPY_OVERLENGTH; |
| 1037 | const BYTE* const iLitEnd = *litPtr + sequence.litLength; |
| 1038 | const BYTE* match = oLitEnd - sequence.offset; |
| 1039 | |
| 1040 | /* check */ |
| 1041 | if (oMatchEnd>oend) return ERROR(dstSize_tooSmall); /* last match must start at a minimum distance of WILDCOPY_OVERLENGTH from oend */ |
| 1042 | if (iLitEnd > litLimit) return ERROR(corruption_detected); /* over-read beyond lit buffer */ |
| 1043 | if (oLitEnd <= oend_w) return ERROR(GENERIC); /* Precondition */ |
| 1044 | |
| 1045 | /* copy literals */ |
| 1046 | if (op < oend_w) { |
| 1047 | ZSTD_wildcopy(op, *litPtr, oend_w - op); |
| 1048 | *litPtr += oend_w - op; |
| 1049 | op = oend_w; |
| 1050 | } |
| 1051 | while (op < oLitEnd) *op++ = *(*litPtr)++; |
| 1052 | |
| 1053 | /* copy Match */ |
| 1054 | if (sequence.offset > (size_t)(oLitEnd - base)) { |
| 1055 | /* offset beyond prefix */ |
| 1056 | if (sequence.offset > (size_t)(oLitEnd - vBase)) return ERROR(corruption_detected); |
| 1057 | match = dictEnd - (base-match); |
| 1058 | if (match + sequence.matchLength <= dictEnd) { |
| 1059 | memmove(oLitEnd, match, sequence.matchLength); |
| 1060 | return sequenceLength; |
| 1061 | } |
| 1062 | /* span extDict & currentPrefixSegment */ |
| 1063 | { size_t const length1 = dictEnd - match; |
| 1064 | memmove(oLitEnd, match, length1); |
| 1065 | op = oLitEnd + length1; |
| 1066 | sequence.matchLength -= length1; |
| 1067 | match = base; |
| 1068 | } } |
| 1069 | while (op < oMatchEnd) *op++ = *match++; |
| 1070 | return sequenceLength; |
| 1071 | } |
| 1072 | |
| 1073 | |
| 1074 | HINT_INLINE |
| 1075 | size_t ZSTD_execSequence(BYTE* op, |
| 1076 | BYTE* const oend, seq_t sequence, |
| 1077 | const BYTE** litPtr, const BYTE* const litLimit, |
| 1078 | const BYTE* const base, const BYTE* const vBase, const BYTE* const dictEnd) |
| 1079 | { |
| 1080 | BYTE* const oLitEnd = op + sequence.litLength; |
| 1081 | size_t const sequenceLength = sequence.litLength + sequence.matchLength; |
| 1082 | BYTE* const oMatchEnd = op + sequenceLength; /* risk : address space overflow (32-bits) */ |
| 1083 | BYTE* const oend_w = oend - WILDCOPY_OVERLENGTH; |
| 1084 | const BYTE* const iLitEnd = *litPtr + sequence.litLength; |
| 1085 | const BYTE* match = oLitEnd - sequence.offset; |
| 1086 | |
| 1087 | /* check */ |
| 1088 | if (oMatchEnd>oend) return ERROR(dstSize_tooSmall); /* last match must start at a minimum distance of WILDCOPY_OVERLENGTH from oend */ |
| 1089 | if (iLitEnd > litLimit) return ERROR(corruption_detected); /* over-read beyond lit buffer */ |
| 1090 | if (oLitEnd>oend_w) return ZSTD_execSequenceLast7(op, oend, sequence, litPtr, litLimit, base, vBase, dictEnd); |
| 1091 | |
| 1092 | /* copy Literals */ |
| 1093 | ZSTD_copy8(op, *litPtr); |
| 1094 | if (sequence.litLength > 8) |
| 1095 | ZSTD_wildcopy(op+8, (*litPtr)+8, sequence.litLength - 8); /* note : since oLitEnd <= oend-WILDCOPY_OVERLENGTH, no risk of overwrite beyond oend */ |
| 1096 | op = oLitEnd; |
| 1097 | *litPtr = iLitEnd; /* update for next sequence */ |
| 1098 | |
| 1099 | /* copy Match */ |
| 1100 | if (sequence.offset > (size_t)(oLitEnd - base)) { |
| 1101 | /* offset beyond prefix -> go into extDict */ |
| 1102 | if (sequence.offset > (size_t)(oLitEnd - vBase)) |
| 1103 | return ERROR(corruption_detected); |
| 1104 | match = dictEnd + (match - base); |
| 1105 | if (match + sequence.matchLength <= dictEnd) { |
| 1106 | memmove(oLitEnd, match, sequence.matchLength); |
| 1107 | return sequenceLength; |
| 1108 | } |
| 1109 | /* span extDict & currentPrefixSegment */ |
| 1110 | { size_t const length1 = dictEnd - match; |
| 1111 | memmove(oLitEnd, match, length1); |
| 1112 | op = oLitEnd + length1; |
| 1113 | sequence.matchLength -= length1; |
| 1114 | match = base; |
| 1115 | if (op > oend_w || sequence.matchLength < MINMATCH) { |
| 1116 | U32 i; |
| 1117 | for (i = 0; i < sequence.matchLength; ++i) op[i] = match[i]; |
| 1118 | return sequenceLength; |
| 1119 | } |
| 1120 | } } |
| 1121 | /* Requirement: op <= oend_w && sequence.matchLength >= MINMATCH */ |
| 1122 | |
| 1123 | /* match within prefix */ |
| 1124 | if (sequence.offset < 8) { |
| 1125 | /* close range match, overlap */ |
| 1126 | static const U32 dec32table[] = { 0, 1, 2, 1, 4, 4, 4, 4 }; /* added */ |
| 1127 | static const int dec64table[] = { 8, 8, 8, 7, 8, 9,10,11 }; /* subtracted */ |
| 1128 | int const sub2 = dec64table[sequence.offset]; |
| 1129 | op[0] = match[0]; |
| 1130 | op[1] = match[1]; |
| 1131 | op[2] = match[2]; |
| 1132 | op[3] = match[3]; |
| 1133 | match += dec32table[sequence.offset]; |
| 1134 | ZSTD_copy4(op+4, match); |
| 1135 | match -= sub2; |
| 1136 | } else { |
| 1137 | ZSTD_copy8(op, match); |
| 1138 | } |
| 1139 | op += 8; match += 8; |
| 1140 | |
| 1141 | if (oMatchEnd > oend-(16-MINMATCH)) { |
| 1142 | if (op < oend_w) { |
| 1143 | ZSTD_wildcopy(op, match, oend_w - op); |
| 1144 | match += oend_w - op; |
| 1145 | op = oend_w; |
| 1146 | } |
| 1147 | while (op < oMatchEnd) *op++ = *match++; |
| 1148 | } else { |
| 1149 | ZSTD_wildcopy(op, match, (ptrdiff_t)sequence.matchLength-8); /* works even if matchLength < 8 */ |
| 1150 | } |
| 1151 | return sequenceLength; |
| 1152 | } |
| 1153 | |
| 1154 | |
| 1155 | HINT_INLINE |
| 1156 | size_t ZSTD_execSequenceLong(BYTE* op, |
| 1157 | BYTE* const oend, seq_t sequence, |
| 1158 | const BYTE** litPtr, const BYTE* const litLimit, |
| 1159 | const BYTE* const prefixStart, const BYTE* const dictStart, const BYTE* const dictEnd) |
| 1160 | { |
| 1161 | BYTE* const oLitEnd = op + sequence.litLength; |
| 1162 | size_t const sequenceLength = sequence.litLength + sequence.matchLength; |
| 1163 | BYTE* const oMatchEnd = op + sequenceLength; /* risk : address space overflow (32-bits) */ |
| 1164 | BYTE* const oend_w = oend - WILDCOPY_OVERLENGTH; |
| 1165 | const BYTE* const iLitEnd = *litPtr + sequence.litLength; |
| 1166 | const BYTE* match = sequence.match; |
| 1167 | |
| 1168 | /* check */ |
| 1169 | if (oMatchEnd > oend) return ERROR(dstSize_tooSmall); /* last match must start at a minimum distance of WILDCOPY_OVERLENGTH from oend */ |
| 1170 | if (iLitEnd > litLimit) return ERROR(corruption_detected); /* over-read beyond lit buffer */ |
| 1171 | if (oLitEnd > oend_w) return ZSTD_execSequenceLast7(op, oend, sequence, litPtr, litLimit, prefixStart, dictStart, dictEnd); |
| 1172 | |
| 1173 | /* copy Literals */ |
| 1174 | ZSTD_copy8(op, *litPtr); /* note : op <= oLitEnd <= oend_w == oend - 8 */ |
| 1175 | if (sequence.litLength > 8) |
| 1176 | ZSTD_wildcopy(op+8, (*litPtr)+8, sequence.litLength - 8); /* note : since oLitEnd <= oend-WILDCOPY_OVERLENGTH, no risk of overwrite beyond oend */ |
| 1177 | op = oLitEnd; |
| 1178 | *litPtr = iLitEnd; /* update for next sequence */ |
| 1179 | |
| 1180 | /* copy Match */ |
| 1181 | if (sequence.offset > (size_t)(oLitEnd - prefixStart)) { |
| 1182 | /* offset beyond prefix */ |
| 1183 | if (sequence.offset > (size_t)(oLitEnd - dictStart)) return ERROR(corruption_detected); |
| 1184 | if (match + sequence.matchLength <= dictEnd) { |
| 1185 | memmove(oLitEnd, match, sequence.matchLength); |
| 1186 | return sequenceLength; |
| 1187 | } |
| 1188 | /* span extDict & currentPrefixSegment */ |
| 1189 | { size_t const length1 = dictEnd - match; |
| 1190 | memmove(oLitEnd, match, length1); |
| 1191 | op = oLitEnd + length1; |
| 1192 | sequence.matchLength -= length1; |
| 1193 | match = prefixStart; |
| 1194 | if (op > oend_w || sequence.matchLength < MINMATCH) { |
| 1195 | U32 i; |
| 1196 | for (i = 0; i < sequence.matchLength; ++i) op[i] = match[i]; |
| 1197 | return sequenceLength; |
| 1198 | } |
| 1199 | } } |
| 1200 | assert(op <= oend_w); |
| 1201 | assert(sequence.matchLength >= MINMATCH); |
| 1202 | |
| 1203 | /* match within prefix */ |
| 1204 | if (sequence.offset < 8) { |
| 1205 | /* close range match, overlap */ |
| 1206 | static const U32 dec32table[] = { 0, 1, 2, 1, 4, 4, 4, 4 }; /* added */ |
| 1207 | static const int dec64table[] = { 8, 8, 8, 7, 8, 9,10,11 }; /* subtracted */ |
| 1208 | int const sub2 = dec64table[sequence.offset]; |
| 1209 | op[0] = match[0]; |
| 1210 | op[1] = match[1]; |
| 1211 | op[2] = match[2]; |
| 1212 | op[3] = match[3]; |
| 1213 | match += dec32table[sequence.offset]; |
| 1214 | ZSTD_copy4(op+4, match); |
| 1215 | match -= sub2; |
| 1216 | } else { |
| 1217 | ZSTD_copy8(op, match); |
| 1218 | } |
| 1219 | op += 8; match += 8; |
| 1220 | |
| 1221 | if (oMatchEnd > oend-(16-MINMATCH)) { |
| 1222 | if (op < oend_w) { |
| 1223 | ZSTD_wildcopy(op, match, oend_w - op); |
| 1224 | match += oend_w - op; |
| 1225 | op = oend_w; |
| 1226 | } |
| 1227 | while (op < oMatchEnd) *op++ = *match++; |
| 1228 | } else { |
| 1229 | ZSTD_wildcopy(op, match, (ptrdiff_t)sequence.matchLength-8); /* works even if matchLength < 8 */ |
| 1230 | } |
| 1231 | return sequenceLength; |
| 1232 | } |
| 1233 | |
| 1234 | static void |
| 1235 | ZSTD_initFseState(ZSTD_fseState* DStatePtr, BIT_DStream_t* bitD, const ZSTD_seqSymbol* dt) |
| 1236 | { |
| 1237 | const void* ptr = dt; |
| 1238 | const ZSTD_seqSymbol_header* const DTableH = (const ZSTD_seqSymbol_header*)ptr; |
| 1239 | DStatePtr->state = BIT_readBits(bitD, DTableH->tableLog); |
| 1240 | DEBUGLOG(6, "ZSTD_initFseState : val=%u using %u bits", |
| 1241 | (U32)DStatePtr->state, DTableH->tableLog); |
| 1242 | BIT_reloadDStream(bitD); |
| 1243 | DStatePtr->table = dt + 1; |
| 1244 | } |
| 1245 | |
| 1246 | FORCE_INLINE_TEMPLATE void |
| 1247 | ZSTD_updateFseState(ZSTD_fseState* DStatePtr, BIT_DStream_t* bitD) |
| 1248 | { |
| 1249 | ZSTD_seqSymbol const DInfo = DStatePtr->table[DStatePtr->state]; |
| 1250 | U32 const nbBits = DInfo.nbBits; |
| 1251 | size_t const lowBits = BIT_readBits(bitD, nbBits); |
| 1252 | DStatePtr->state = DInfo.nextState + lowBits; |
| 1253 | } |
| 1254 | |
| 1255 | /* We need to add at most (ZSTD_WINDOWLOG_MAX_32 - 1) bits to read the maximum |
| 1256 | * offset bits. But we can only read at most (STREAM_ACCUMULATOR_MIN_32 - 1) |
| 1257 | * bits before reloading. This value is the maximum number of bytes we read |
| 1258 | * after reloading when we are decoding long offets. |
| 1259 | */ |
| 1260 | #define LONG_OFFSETS_MAX_EXTRA_BITS_32 \ |
| 1261 | (ZSTD_WINDOWLOG_MAX_32 > STREAM_ACCUMULATOR_MIN_32 \ |
| 1262 | ? ZSTD_WINDOWLOG_MAX_32 - STREAM_ACCUMULATOR_MIN_32 \ |
| 1263 | : 0) |
| 1264 | |
| 1265 | typedef enum { ZSTD_lo_isRegularOffset, ZSTD_lo_isLongOffset=1 } ZSTD_longOffset_e; |
| 1266 | |
| 1267 | FORCE_INLINE_TEMPLATE seq_t |
| 1268 | ZSTD_decodeSequence(seqState_t* seqState, const ZSTD_longOffset_e longOffsets) |
| 1269 | { |
| 1270 | seq_t seq; |
| 1271 | U32 const llBits = seqState->stateLL.table[seqState->stateLL.state].nbAdditionalBits; |
| 1272 | U32 const mlBits = seqState->stateML.table[seqState->stateML.state].nbAdditionalBits; |
| 1273 | U32 const ofBits = seqState->stateOffb.table[seqState->stateOffb.state].nbAdditionalBits; |
| 1274 | U32 const totalBits = llBits+mlBits+ofBits; |
| 1275 | U32 const llBase = seqState->stateLL.table[seqState->stateLL.state].baseValue; |
| 1276 | U32 const mlBase = seqState->stateML.table[seqState->stateML.state].baseValue; |
| 1277 | U32 const ofBase = seqState->stateOffb.table[seqState->stateOffb.state].baseValue; |
| 1278 | |
| 1279 | /* sequence */ |
| 1280 | { size_t offset; |
| 1281 | if (!ofBits) |
| 1282 | offset = 0; |
| 1283 | else { |
| 1284 | ZSTD_STATIC_ASSERT(ZSTD_lo_isLongOffset == 1); |
| 1285 | ZSTD_STATIC_ASSERT(LONG_OFFSETS_MAX_EXTRA_BITS_32 == 5); |
| 1286 | assert(ofBits <= MaxOff); |
| 1287 | if (MEM_32bits() && longOffsets && (ofBits >= STREAM_ACCUMULATOR_MIN_32)) { |
| 1288 | U32 const extraBits = ofBits - MIN(ofBits, 32 - seqState->DStream.bitsConsumed); |
| 1289 | offset = ofBase + (BIT_readBitsFast(&seqState->DStream, ofBits - extraBits) << extraBits); |
| 1290 | BIT_reloadDStream(&seqState->DStream); |
| 1291 | if (extraBits) offset += BIT_readBitsFast(&seqState->DStream, extraBits); |
| 1292 | assert(extraBits <= LONG_OFFSETS_MAX_EXTRA_BITS_32); /* to avoid another reload */ |
| 1293 | } else { |
| 1294 | offset = ofBase + BIT_readBitsFast(&seqState->DStream, ofBits/*>0*/); /* <= (ZSTD_WINDOWLOG_MAX-1) bits */ |
| 1295 | if (MEM_32bits()) BIT_reloadDStream(&seqState->DStream); |
| 1296 | } |
| 1297 | } |
| 1298 | |
| 1299 | if (ofBits <= 1) { |
| 1300 | offset += (llBase==0); |
| 1301 | if (offset) { |
| 1302 | size_t temp = (offset==3) ? seqState->prevOffset[0] - 1 : seqState->prevOffset[offset]; |
| 1303 | temp += !temp; /* 0 is not valid; input is corrupted; force offset to 1 */ |
| 1304 | if (offset != 1) seqState->prevOffset[2] = seqState->prevOffset[1]; |
| 1305 | seqState->prevOffset[1] = seqState->prevOffset[0]; |
| 1306 | seqState->prevOffset[0] = offset = temp; |
| 1307 | } else { /* offset == 0 */ |
| 1308 | offset = seqState->prevOffset[0]; |
| 1309 | } |
| 1310 | } else { |
| 1311 | seqState->prevOffset[2] = seqState->prevOffset[1]; |
| 1312 | seqState->prevOffset[1] = seqState->prevOffset[0]; |
| 1313 | seqState->prevOffset[0] = offset; |
| 1314 | } |
| 1315 | seq.offset = offset; |
| 1316 | } |
| 1317 | |
| 1318 | seq.matchLength = mlBase |
| 1319 | + ((mlBits>0) ? BIT_readBitsFast(&seqState->DStream, mlBits/*>0*/) : 0); /* <= 16 bits */ |
| 1320 | if (MEM_32bits() && (mlBits+llBits >= STREAM_ACCUMULATOR_MIN_32-LONG_OFFSETS_MAX_EXTRA_BITS_32)) |
| 1321 | BIT_reloadDStream(&seqState->DStream); |
| 1322 | if (MEM_64bits() && (totalBits >= STREAM_ACCUMULATOR_MIN_64-(LLFSELog+MLFSELog+OffFSELog))) |
| 1323 | BIT_reloadDStream(&seqState->DStream); |
| 1324 | /* Ensure there are enough bits to read the rest of data in 64-bit mode. */ |
| 1325 | ZSTD_STATIC_ASSERT(16+LLFSELog+MLFSELog+OffFSELog < STREAM_ACCUMULATOR_MIN_64); |
| 1326 | |
| 1327 | seq.litLength = llBase |
| 1328 | + ((llBits>0) ? BIT_readBitsFast(&seqState->DStream, llBits/*>0*/) : 0); /* <= 16 bits */ |
| 1329 | if (MEM_32bits()) |
| 1330 | BIT_reloadDStream(&seqState->DStream); |
| 1331 | |
| 1332 | DEBUGLOG(6, "seq: litL=%u, matchL=%u, offset=%u", |
| 1333 | (U32)seq.litLength, (U32)seq.matchLength, (U32)seq.offset); |
| 1334 | |
| 1335 | /* ANS state update */ |
| 1336 | ZSTD_updateFseState(&seqState->stateLL, &seqState->DStream); /* <= 9 bits */ |
| 1337 | ZSTD_updateFseState(&seqState->stateML, &seqState->DStream); /* <= 9 bits */ |
| 1338 | if (MEM_32bits()) BIT_reloadDStream(&seqState->DStream); /* <= 18 bits */ |
| 1339 | ZSTD_updateFseState(&seqState->stateOffb, &seqState->DStream); /* <= 8 bits */ |
| 1340 | |
| 1341 | return seq; |
| 1342 | } |
| 1343 | |
| 1344 | FORCE_INLINE_TEMPLATE size_t |
| 1345 | ZSTD_decompressSequences_body( ZSTD_DCtx* dctx, |
| 1346 | void* dst, size_t maxDstSize, |
| 1347 | const void* seqStart, size_t seqSize, int nbSeq, |
| 1348 | const ZSTD_longOffset_e isLongOffset) |
| 1349 | { |
| 1350 | const BYTE* ip = (const BYTE*)seqStart; |
| 1351 | const BYTE* const iend = ip + seqSize; |
| 1352 | BYTE* const ostart = (BYTE* const)dst; |
| 1353 | BYTE* const oend = ostart + maxDstSize; |
| 1354 | BYTE* op = ostart; |
| 1355 | const BYTE* litPtr = dctx->litPtr; |
| 1356 | const BYTE* const litEnd = litPtr + dctx->litSize; |
| 1357 | const BYTE* const base = (const BYTE*) (dctx->base); |
| 1358 | const BYTE* const vBase = (const BYTE*) (dctx->vBase); |
| 1359 | const BYTE* const dictEnd = (const BYTE*) (dctx->dictEnd); |
| 1360 | DEBUGLOG(5, "ZSTD_decompressSequences"); |
| 1361 | |
| 1362 | /* Regen sequences */ |
| 1363 | if (nbSeq) { |
| 1364 | seqState_t seqState; |
| 1365 | dctx->fseEntropy = 1; |
| 1366 | { U32 i; for (i=0; i<ZSTD_REP_NUM; i++) seqState.prevOffset[i] = dctx->entropy.rep[i]; } |
| 1367 | CHECK_E(BIT_initDStream(&seqState.DStream, ip, iend-ip), corruption_detected); |
| 1368 | ZSTD_initFseState(&seqState.stateLL, &seqState.DStream, dctx->LLTptr); |
| 1369 | ZSTD_initFseState(&seqState.stateOffb, &seqState.DStream, dctx->OFTptr); |
| 1370 | ZSTD_initFseState(&seqState.stateML, &seqState.DStream, dctx->MLTptr); |
| 1371 | |
| 1372 | for ( ; (BIT_reloadDStream(&(seqState.DStream)) <= BIT_DStream_completed) && nbSeq ; ) { |
| 1373 | nbSeq--; |
| 1374 | { seq_t const sequence = ZSTD_decodeSequence(&seqState, isLongOffset); |
| 1375 | size_t const oneSeqSize = ZSTD_execSequence(op, oend, sequence, &litPtr, litEnd, base, vBase, dictEnd); |
| 1376 | DEBUGLOG(6, "regenerated sequence size : %u", (U32)oneSeqSize); |
| 1377 | if (ZSTD_isError(oneSeqSize)) return oneSeqSize; |
| 1378 | op += oneSeqSize; |
| 1379 | } } |
| 1380 | |
| 1381 | /* check if reached exact end */ |
| 1382 | DEBUGLOG(5, "ZSTD_decompressSequences: after decode loop, remaining nbSeq : %i", nbSeq); |
| 1383 | if (nbSeq) return ERROR(corruption_detected); |
| 1384 | /* save reps for next block */ |
| 1385 | { U32 i; for (i=0; i<ZSTD_REP_NUM; i++) dctx->entropy.rep[i] = (U32)(seqState.prevOffset[i]); } |
| 1386 | } |
| 1387 | |
| 1388 | /* last literal segment */ |
| 1389 | { size_t const lastLLSize = litEnd - litPtr; |
| 1390 | if (lastLLSize > (size_t)(oend-op)) return ERROR(dstSize_tooSmall); |
| 1391 | memcpy(op, litPtr, lastLLSize); |
| 1392 | op += lastLLSize; |
| 1393 | } |
| 1394 | |
| 1395 | return op-ostart; |
| 1396 | } |
| 1397 | |
| 1398 | static size_t |
| 1399 | ZSTD_decompressSequences_default(ZSTD_DCtx* dctx, |
| 1400 | void* dst, size_t maxDstSize, |
| 1401 | const void* seqStart, size_t seqSize, int nbSeq, |
| 1402 | const ZSTD_longOffset_e isLongOffset) |
| 1403 | { |
| 1404 | return ZSTD_decompressSequences_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset); |
| 1405 | } |
| 1406 | |
| 1407 | |
| 1408 | |
| 1409 | FORCE_INLINE_TEMPLATE seq_t |
| 1410 | ZSTD_decodeSequenceLong(seqState_t* seqState, ZSTD_longOffset_e const longOffsets) |
| 1411 | { |
| 1412 | seq_t seq; |
| 1413 | U32 const llBits = seqState->stateLL.table[seqState->stateLL.state].nbAdditionalBits; |
| 1414 | U32 const mlBits = seqState->stateML.table[seqState->stateML.state].nbAdditionalBits; |
| 1415 | U32 const ofBits = seqState->stateOffb.table[seqState->stateOffb.state].nbAdditionalBits; |
| 1416 | U32 const totalBits = llBits+mlBits+ofBits; |
| 1417 | U32 const llBase = seqState->stateLL.table[seqState->stateLL.state].baseValue; |
| 1418 | U32 const mlBase = seqState->stateML.table[seqState->stateML.state].baseValue; |
| 1419 | U32 const ofBase = seqState->stateOffb.table[seqState->stateOffb.state].baseValue; |
| 1420 | |
| 1421 | /* sequence */ |
| 1422 | { size_t offset; |
| 1423 | if (!ofBits) |
| 1424 | offset = 0; |
| 1425 | else { |
| 1426 | ZSTD_STATIC_ASSERT(ZSTD_lo_isLongOffset == 1); |
| 1427 | ZSTD_STATIC_ASSERT(LONG_OFFSETS_MAX_EXTRA_BITS_32 == 5); |
| 1428 | assert(ofBits <= MaxOff); |
| 1429 | if (MEM_32bits() && longOffsets) { |
| 1430 | U32 const extraBits = ofBits - MIN(ofBits, STREAM_ACCUMULATOR_MIN_32-1); |
| 1431 | offset = ofBase + (BIT_readBitsFast(&seqState->DStream, ofBits - extraBits) << extraBits); |
| 1432 | if (MEM_32bits() || extraBits) BIT_reloadDStream(&seqState->DStream); |
| 1433 | if (extraBits) offset += BIT_readBitsFast(&seqState->DStream, extraBits); |
| 1434 | } else { |
| 1435 | offset = ofBase + BIT_readBitsFast(&seqState->DStream, ofBits); /* <= (ZSTD_WINDOWLOG_MAX-1) bits */ |
| 1436 | if (MEM_32bits()) BIT_reloadDStream(&seqState->DStream); |
| 1437 | } |
| 1438 | } |
| 1439 | |
| 1440 | if (ofBits <= 1) { |
| 1441 | offset += (llBase==0); |
| 1442 | if (offset) { |
| 1443 | size_t temp = (offset==3) ? seqState->prevOffset[0] - 1 : seqState->prevOffset[offset]; |
| 1444 | temp += !temp; /* 0 is not valid; input is corrupted; force offset to 1 */ |
| 1445 | if (offset != 1) seqState->prevOffset[2] = seqState->prevOffset[1]; |
| 1446 | seqState->prevOffset[1] = seqState->prevOffset[0]; |
| 1447 | seqState->prevOffset[0] = offset = temp; |
| 1448 | } else { |
| 1449 | offset = seqState->prevOffset[0]; |
| 1450 | } |
| 1451 | } else { |
| 1452 | seqState->prevOffset[2] = seqState->prevOffset[1]; |
| 1453 | seqState->prevOffset[1] = seqState->prevOffset[0]; |
| 1454 | seqState->prevOffset[0] = offset; |
| 1455 | } |
| 1456 | seq.offset = offset; |
| 1457 | } |
| 1458 | |
| 1459 | seq.matchLength = mlBase + ((mlBits>0) ? BIT_readBitsFast(&seqState->DStream, mlBits) : 0); /* <= 16 bits */ |
| 1460 | if (MEM_32bits() && (mlBits+llBits >= STREAM_ACCUMULATOR_MIN_32-LONG_OFFSETS_MAX_EXTRA_BITS_32)) |
| 1461 | BIT_reloadDStream(&seqState->DStream); |
| 1462 | if (MEM_64bits() && (totalBits >= STREAM_ACCUMULATOR_MIN_64-(LLFSELog+MLFSELog+OffFSELog))) |
| 1463 | BIT_reloadDStream(&seqState->DStream); |
| 1464 | /* Verify that there is enough bits to read the rest of the data in 64-bit mode. */ |
| 1465 | ZSTD_STATIC_ASSERT(16+LLFSELog+MLFSELog+OffFSELog < STREAM_ACCUMULATOR_MIN_64); |
| 1466 | |
| 1467 | seq.litLength = llBase + ((llBits>0) ? BIT_readBitsFast(&seqState->DStream, llBits) : 0); /* <= 16 bits */ |
| 1468 | if (MEM_32bits()) |
| 1469 | BIT_reloadDStream(&seqState->DStream); |
| 1470 | |
| 1471 | { size_t const pos = seqState->pos + seq.litLength; |
| 1472 | const BYTE* const matchBase = (seq.offset > pos) ? seqState->dictEnd : seqState->prefixStart; |
| 1473 | seq.match = matchBase + pos - seq.offset; /* note : this operation can overflow when seq.offset is really too large, which can only happen when input is corrupted. |
| 1474 | * No consequence though : no memory access will occur, overly large offset will be detected in ZSTD_execSequenceLong() */ |
| 1475 | seqState->pos = pos + seq.matchLength; |
| 1476 | } |
| 1477 | |
| 1478 | /* ANS state update */ |
| 1479 | ZSTD_updateFseState(&seqState->stateLL, &seqState->DStream); /* <= 9 bits */ |
| 1480 | ZSTD_updateFseState(&seqState->stateML, &seqState->DStream); /* <= 9 bits */ |
| 1481 | if (MEM_32bits()) BIT_reloadDStream(&seqState->DStream); /* <= 18 bits */ |
| 1482 | ZSTD_updateFseState(&seqState->stateOffb, &seqState->DStream); /* <= 8 bits */ |
| 1483 | |
| 1484 | return seq; |
| 1485 | } |
| 1486 | |
| 1487 | FORCE_INLINE_TEMPLATE size_t |
| 1488 | ZSTD_decompressSequencesLong_body( |
| 1489 | ZSTD_DCtx* dctx, |
| 1490 | void* dst, size_t maxDstSize, |
| 1491 | const void* seqStart, size_t seqSize, int nbSeq, |
| 1492 | const ZSTD_longOffset_e isLongOffset) |
| 1493 | { |
| 1494 | const BYTE* ip = (const BYTE*)seqStart; |
| 1495 | const BYTE* const iend = ip + seqSize; |
| 1496 | BYTE* const ostart = (BYTE* const)dst; |
| 1497 | BYTE* const oend = ostart + maxDstSize; |
| 1498 | BYTE* op = ostart; |
| 1499 | const BYTE* litPtr = dctx->litPtr; |
| 1500 | const BYTE* const litEnd = litPtr + dctx->litSize; |
| 1501 | const BYTE* const prefixStart = (const BYTE*) (dctx->base); |
| 1502 | const BYTE* const dictStart = (const BYTE*) (dctx->vBase); |
| 1503 | const BYTE* const dictEnd = (const BYTE*) (dctx->dictEnd); |
| 1504 | |
| 1505 | /* Regen sequences */ |
| 1506 | if (nbSeq) { |
| 1507 | #define STORED_SEQS 4 |
| 1508 | #define STOSEQ_MASK (STORED_SEQS-1) |
| 1509 | #define ADVANCED_SEQS 4 |
| 1510 | seq_t sequences[STORED_SEQS]; |
| 1511 | int const seqAdvance = MIN(nbSeq, ADVANCED_SEQS); |
| 1512 | seqState_t seqState; |
| 1513 | int seqNb; |
| 1514 | dctx->fseEntropy = 1; |
| 1515 | { U32 i; for (i=0; i<ZSTD_REP_NUM; i++) seqState.prevOffset[i] = dctx->entropy.rep[i]; } |
| 1516 | seqState.prefixStart = prefixStart; |
| 1517 | seqState.pos = (size_t)(op-prefixStart); |
| 1518 | seqState.dictEnd = dictEnd; |
| 1519 | CHECK_E(BIT_initDStream(&seqState.DStream, ip, iend-ip), corruption_detected); |
| 1520 | ZSTD_initFseState(&seqState.stateLL, &seqState.DStream, dctx->LLTptr); |
| 1521 | ZSTD_initFseState(&seqState.stateOffb, &seqState.DStream, dctx->OFTptr); |
| 1522 | ZSTD_initFseState(&seqState.stateML, &seqState.DStream, dctx->MLTptr); |
| 1523 | |
| 1524 | /* prepare in advance */ |
| 1525 | for (seqNb=0; (BIT_reloadDStream(&seqState.DStream) <= BIT_DStream_completed) && (seqNb<seqAdvance); seqNb++) { |
| 1526 | sequences[seqNb] = ZSTD_decodeSequenceLong(&seqState, isLongOffset); |
| 1527 | } |
| 1528 | if (seqNb<seqAdvance) return ERROR(corruption_detected); |
| 1529 | |
| 1530 | /* decode and decompress */ |
| 1531 | for ( ; (BIT_reloadDStream(&(seqState.DStream)) <= BIT_DStream_completed) && (seqNb<nbSeq) ; seqNb++) { |
| 1532 | seq_t const sequence = ZSTD_decodeSequenceLong(&seqState, isLongOffset); |
| 1533 | size_t const oneSeqSize = ZSTD_execSequenceLong(op, oend, sequences[(seqNb-ADVANCED_SEQS) & STOSEQ_MASK], &litPtr, litEnd, prefixStart, dictStart, dictEnd); |
| 1534 | if (ZSTD_isError(oneSeqSize)) return oneSeqSize; |
| 1535 | PREFETCH(sequence.match); /* note : it's safe to invoke PREFETCH() on any memory address, including invalid ones */ |
| 1536 | sequences[seqNb&STOSEQ_MASK] = sequence; |
| 1537 | op += oneSeqSize; |
| 1538 | } |
| 1539 | if (seqNb<nbSeq) return ERROR(corruption_detected); |
| 1540 | |
| 1541 | /* finish queue */ |
| 1542 | seqNb -= seqAdvance; |
| 1543 | for ( ; seqNb<nbSeq ; seqNb++) { |
| 1544 | size_t const oneSeqSize = ZSTD_execSequenceLong(op, oend, sequences[seqNb&STOSEQ_MASK], &litPtr, litEnd, prefixStart, dictStart, dictEnd); |
| 1545 | if (ZSTD_isError(oneSeqSize)) return oneSeqSize; |
| 1546 | op += oneSeqSize; |
| 1547 | } |
| 1548 | |
| 1549 | /* save reps for next block */ |
| 1550 | { U32 i; for (i=0; i<ZSTD_REP_NUM; i++) dctx->entropy.rep[i] = (U32)(seqState.prevOffset[i]); } |
| 1551 | #undef STORED_SEQS |
| 1552 | #undef STOSEQ_MASK |
| 1553 | #undef ADVANCED_SEQS |
| 1554 | } |
| 1555 | |
| 1556 | /* last literal segment */ |
| 1557 | { size_t const lastLLSize = litEnd - litPtr; |
| 1558 | if (lastLLSize > (size_t)(oend-op)) return ERROR(dstSize_tooSmall); |
| 1559 | memcpy(op, litPtr, lastLLSize); |
| 1560 | op += lastLLSize; |
| 1561 | } |
| 1562 | |
| 1563 | return op-ostart; |
| 1564 | } |
| 1565 | |
| 1566 | static size_t |
| 1567 | ZSTD_decompressSequencesLong_default(ZSTD_DCtx* dctx, |
| 1568 | void* dst, size_t maxDstSize, |
| 1569 | const void* seqStart, size_t seqSize, int nbSeq, |
| 1570 | const ZSTD_longOffset_e isLongOffset) |
| 1571 | { |
| 1572 | return ZSTD_decompressSequencesLong_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset); |
| 1573 | } |
| 1574 | |
| 1575 | |
| 1576 | |
| 1577 | #if DYNAMIC_BMI2 |
| 1578 | |
| 1579 | static TARGET_ATTRIBUTE("bmi2") size_t |
| 1580 | ZSTD_decompressSequences_bmi2(ZSTD_DCtx* dctx, |
| 1581 | void* dst, size_t maxDstSize, |
| 1582 | const void* seqStart, size_t seqSize, int nbSeq, |
| 1583 | const ZSTD_longOffset_e isLongOffset) |
| 1584 | { |
| 1585 | return ZSTD_decompressSequences_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset); |
| 1586 | } |
| 1587 | |
| 1588 | static TARGET_ATTRIBUTE("bmi2") size_t |
| 1589 | ZSTD_decompressSequencesLong_bmi2(ZSTD_DCtx* dctx, |
| 1590 | void* dst, size_t maxDstSize, |
| 1591 | const void* seqStart, size_t seqSize, int nbSeq, |
| 1592 | const ZSTD_longOffset_e isLongOffset) |
| 1593 | { |
| 1594 | return ZSTD_decompressSequencesLong_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset); |
| 1595 | } |
| 1596 | |
| 1597 | #endif |
| 1598 | |
| 1599 | typedef size_t (*ZSTD_decompressSequences_t)( |
| 1600 | ZSTD_DCtx *dctx, void *dst, size_t maxDstSize, |
| 1601 | const void *seqStart, size_t seqSize, int nbSeq, |
| 1602 | const ZSTD_longOffset_e isLongOffset); |
| 1603 | |
| 1604 | static size_t ZSTD_decompressSequences(ZSTD_DCtx* dctx, void* dst, size_t maxDstSize, |
| 1605 | const void* seqStart, size_t seqSize, int nbSeq, |
| 1606 | const ZSTD_longOffset_e isLongOffset) |
| 1607 | { |
| 1608 | DEBUGLOG(5, "ZSTD_decompressSequences"); |
| 1609 | #if DYNAMIC_BMI2 |
| 1610 | if (dctx->bmi2) { |
| 1611 | return ZSTD_decompressSequences_bmi2(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset); |
| 1612 | } |
| 1613 | #endif |
| 1614 | return ZSTD_decompressSequences_default(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset); |
| 1615 | } |
| 1616 | |
| 1617 | static size_t ZSTD_decompressSequencesLong(ZSTD_DCtx* dctx, |
| 1618 | void* dst, size_t maxDstSize, |
| 1619 | const void* seqStart, size_t seqSize, int nbSeq, |
| 1620 | const ZSTD_longOffset_e isLongOffset) |
| 1621 | { |
| 1622 | DEBUGLOG(5, "ZSTD_decompressSequencesLong"); |
| 1623 | #if DYNAMIC_BMI2 |
| 1624 | if (dctx->bmi2) { |
| 1625 | return ZSTD_decompressSequencesLong_bmi2(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset); |
| 1626 | } |
| 1627 | #endif |
| 1628 | return ZSTD_decompressSequencesLong_default(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset); |
| 1629 | } |
| 1630 | |
| 1631 | /* ZSTD_getLongOffsetsShare() : |
| 1632 | * condition : offTable must be valid |
| 1633 | * @return : "share" of long offsets (arbitrarily defined as > (1<<23)) |
| 1634 | * compared to maximum possible of (1<<OffFSELog) */ |
| 1635 | static unsigned |
| 1636 | ZSTD_getLongOffsetsShare(const ZSTD_seqSymbol* offTable) |
| 1637 | { |
| 1638 | const void* ptr = offTable; |
| 1639 | U32 const tableLog = ((const ZSTD_seqSymbol_header*)ptr)[0].tableLog; |
| 1640 | const ZSTD_seqSymbol* table = offTable + 1; |
| 1641 | U32 const max = 1 << tableLog; |
| 1642 | U32 u, total = 0; |
| 1643 | DEBUGLOG(5, "ZSTD_getLongOffsetsShare: (tableLog=%u)", tableLog); |
| 1644 | |
| 1645 | assert(max <= (1 << OffFSELog)); /* max not too large */ |
| 1646 | for (u=0; u<max; u++) { |
| 1647 | if (table[u].nbAdditionalBits > 22) total += 1; |
| 1648 | } |
| 1649 | |
| 1650 | assert(tableLog <= OffFSELog); |
| 1651 | total <<= (OffFSELog - tableLog); /* scale to OffFSELog */ |
| 1652 | |
| 1653 | return total; |
| 1654 | } |
| 1655 | |
| 1656 | |
| 1657 | static size_t ZSTD_decompressBlock_internal(ZSTD_DCtx* dctx, |
| 1658 | void* dst, size_t dstCapacity, |
| 1659 | const void* src, size_t srcSize, const int frame) |
| 1660 | { /* blockType == blockCompressed */ |
| 1661 | const BYTE* ip = (const BYTE*)src; |
| 1662 | /* isLongOffset must be true if there are long offsets. |
| 1663 | * Offsets are long if they are larger than 2^STREAM_ACCUMULATOR_MIN. |
| 1664 | * We don't expect that to be the case in 64-bit mode. |
| 1665 | * In block mode, window size is not known, so we have to be conservative. (note: but it could be evaluated from current-lowLimit) |
| 1666 | */ |
| 1667 | ZSTD_longOffset_e const isLongOffset = (ZSTD_longOffset_e)(MEM_32bits() && (!frame || dctx->fParams.windowSize > (1ULL << STREAM_ACCUMULATOR_MIN))); |
| 1668 | DEBUGLOG(5, "ZSTD_decompressBlock_internal (size : %u)", (U32)srcSize); |
| 1669 | |
| 1670 | if (srcSize >= ZSTD_BLOCKSIZE_MAX) return ERROR(srcSize_wrong); |
| 1671 | |
| 1672 | /* Decode literals section */ |
| 1673 | { size_t const litCSize = ZSTD_decodeLiteralsBlock(dctx, src, srcSize); |
| 1674 | DEBUGLOG(5, "ZSTD_decodeLiteralsBlock : %u", (U32)litCSize); |
| 1675 | if (ZSTD_isError(litCSize)) return litCSize; |
| 1676 | ip += litCSize; |
| 1677 | srcSize -= litCSize; |
| 1678 | } |
| 1679 | |
| 1680 | /* Build Decoding Tables */ |
| 1681 | { int nbSeq; |
| 1682 | size_t const seqHSize = ZSTD_decodeSeqHeaders(dctx, &nbSeq, ip, srcSize); |
| 1683 | if (ZSTD_isError(seqHSize)) return seqHSize; |
| 1684 | ip += seqHSize; |
| 1685 | srcSize -= seqHSize; |
| 1686 | |
| 1687 | if ( (!frame || dctx->fParams.windowSize > (1<<24)) |
| 1688 | && (nbSeq>0) ) { /* could probably use a larger nbSeq limit */ |
| 1689 | U32 const shareLongOffsets = ZSTD_getLongOffsetsShare(dctx->OFTptr); |
| 1690 | U32 const minShare = MEM_64bits() ? 7 : 20; /* heuristic values, correspond to 2.73% and 7.81% */ |
| 1691 | if (shareLongOffsets >= minShare) |
| 1692 | return ZSTD_decompressSequencesLong(dctx, dst, dstCapacity, ip, srcSize, nbSeq, isLongOffset); |
| 1693 | } |
| 1694 | |
| 1695 | return ZSTD_decompressSequences(dctx, dst, dstCapacity, ip, srcSize, nbSeq, isLongOffset); |
| 1696 | } |
| 1697 | } |
| 1698 | |
| 1699 | |
| 1700 | static void ZSTD_checkContinuity(ZSTD_DCtx* dctx, const void* dst) |
| 1701 | { |
| 1702 | if (dst != dctx->previousDstEnd) { /* not contiguous */ |
| 1703 | dctx->dictEnd = dctx->previousDstEnd; |
| 1704 | dctx->vBase = (const char*)dst - ((const char*)(dctx->previousDstEnd) - (const char*)(dctx->base)); |
| 1705 | dctx->base = dst; |
| 1706 | dctx->previousDstEnd = dst; |
| 1707 | } |
| 1708 | } |
| 1709 | |
| 1710 | size_t ZSTD_decompressBlock(ZSTD_DCtx* dctx, |
| 1711 | void* dst, size_t dstCapacity, |
| 1712 | const void* src, size_t srcSize) |
| 1713 | { |
| 1714 | size_t dSize; |
| 1715 | ZSTD_checkContinuity(dctx, dst); |
| 1716 | dSize = ZSTD_decompressBlock_internal(dctx, dst, dstCapacity, src, srcSize, /* frame */ 0); |
| 1717 | dctx->previousDstEnd = (char*)dst + dSize; |
| 1718 | return dSize; |
| 1719 | } |
| 1720 | |
| 1721 | |
| 1722 | /** ZSTD_insertBlock() : |
| 1723 | insert `src` block into `dctx` history. Useful to track uncompressed blocks. */ |
| 1724 | ZSTDLIB_API size_t ZSTD_insertBlock(ZSTD_DCtx* dctx, const void* blockStart, size_t blockSize) |
| 1725 | { |
| 1726 | ZSTD_checkContinuity(dctx, blockStart); |
| 1727 | dctx->previousDstEnd = (const char*)blockStart + blockSize; |
| 1728 | return blockSize; |
| 1729 | } |
| 1730 | |
| 1731 | |
| 1732 | static size_t ZSTD_generateNxBytes(void* dst, size_t dstCapacity, BYTE byte, size_t length) |
| 1733 | { |
| 1734 | if (length > dstCapacity) return ERROR(dstSize_tooSmall); |
| 1735 | memset(dst, byte, length); |
| 1736 | return length; |
| 1737 | } |
| 1738 | |
| 1739 | /** ZSTD_findFrameCompressedSize() : |
| 1740 | * compatible with legacy mode |
| 1741 | * `src` must point to the start of a ZSTD frame, ZSTD legacy frame, or skippable frame |
| 1742 | * `srcSize` must be at least as large as the frame contained |
| 1743 | * @return : the compressed size of the frame starting at `src` */ |
| 1744 | size_t ZSTD_findFrameCompressedSize(const void *src, size_t srcSize) |
| 1745 | { |
| 1746 | #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT >= 1) |
| 1747 | if (ZSTD_isLegacy(src, srcSize)) |
| 1748 | return ZSTD_findFrameCompressedSizeLegacy(src, srcSize); |
| 1749 | #endif |
| 1750 | if ( (srcSize >= ZSTD_skippableHeaderSize) |
| 1751 | && (MEM_readLE32(src) & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START ) { |
| 1752 | return ZSTD_skippableHeaderSize + MEM_readLE32((const BYTE*)src + ZSTD_frameIdSize); |
| 1753 | } else { |
| 1754 | const BYTE* ip = (const BYTE*)src; |
| 1755 | const BYTE* const ipstart = ip; |
| 1756 | size_t remainingSize = srcSize; |
| 1757 | ZSTD_frameHeader zfh; |
| 1758 | |
| 1759 | /* Extract Frame Header */ |
| 1760 | { size_t const ret = ZSTD_getFrameHeader(&zfh, src, srcSize); |
| 1761 | if (ZSTD_isError(ret)) return ret; |
| 1762 | if (ret > 0) return ERROR(srcSize_wrong); |
| 1763 | } |
| 1764 | |
| 1765 | ip += zfh.headerSize; |
| 1766 | remainingSize -= zfh.headerSize; |
| 1767 | |
| 1768 | /* Loop on each block */ |
| 1769 | while (1) { |
| 1770 | blockProperties_t blockProperties; |
| 1771 | size_t const cBlockSize = ZSTD_getcBlockSize(ip, remainingSize, &blockProperties); |
| 1772 | if (ZSTD_isError(cBlockSize)) return cBlockSize; |
| 1773 | |
| 1774 | if (ZSTD_blockHeaderSize + cBlockSize > remainingSize) |
| 1775 | return ERROR(srcSize_wrong); |
| 1776 | |
| 1777 | ip += ZSTD_blockHeaderSize + cBlockSize; |
| 1778 | remainingSize -= ZSTD_blockHeaderSize + cBlockSize; |
| 1779 | |
| 1780 | if (blockProperties.lastBlock) break; |
| 1781 | } |
| 1782 | |
| 1783 | if (zfh.checksumFlag) { /* Final frame content checksum */ |
| 1784 | if (remainingSize < 4) return ERROR(srcSize_wrong); |
| 1785 | ip += 4; |
| 1786 | remainingSize -= 4; |
| 1787 | } |
| 1788 | |
| 1789 | return ip - ipstart; |
| 1790 | } |
| 1791 | } |
| 1792 | |
| 1793 | /*! ZSTD_decompressFrame() : |
| 1794 | * @dctx must be properly initialized */ |
| 1795 | static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx, |
| 1796 | void* dst, size_t dstCapacity, |
| 1797 | const void** srcPtr, size_t *srcSizePtr) |
| 1798 | { |
| 1799 | const BYTE* ip = (const BYTE*)(*srcPtr); |
| 1800 | BYTE* const ostart = (BYTE* const)dst; |
| 1801 | BYTE* const oend = ostart + dstCapacity; |
| 1802 | BYTE* op = ostart; |
| 1803 | size_t remainingSize = *srcSizePtr; |
| 1804 | |
| 1805 | /* check */ |
| 1806 | if (remainingSize < ZSTD_frameHeaderSize_min+ZSTD_blockHeaderSize) |
| 1807 | return ERROR(srcSize_wrong); |
| 1808 | |
| 1809 | /* Frame Header */ |
| 1810 | { size_t const frameHeaderSize = ZSTD_frameHeaderSize(ip, ZSTD_frameHeaderSize_prefix); |
| 1811 | if (ZSTD_isError(frameHeaderSize)) return frameHeaderSize; |
| 1812 | if (remainingSize < frameHeaderSize+ZSTD_blockHeaderSize) |
| 1813 | return ERROR(srcSize_wrong); |
| 1814 | CHECK_F( ZSTD_decodeFrameHeader(dctx, ip, frameHeaderSize) ); |
| 1815 | ip += frameHeaderSize; remainingSize -= frameHeaderSize; |
| 1816 | } |
| 1817 | |
| 1818 | /* Loop on each block */ |
| 1819 | while (1) { |
| 1820 | size_t decodedSize; |
| 1821 | blockProperties_t blockProperties; |
| 1822 | size_t const cBlockSize = ZSTD_getcBlockSize(ip, remainingSize, &blockProperties); |
| 1823 | if (ZSTD_isError(cBlockSize)) return cBlockSize; |
| 1824 | |
| 1825 | ip += ZSTD_blockHeaderSize; |
| 1826 | remainingSize -= ZSTD_blockHeaderSize; |
| 1827 | if (cBlockSize > remainingSize) return ERROR(srcSize_wrong); |
| 1828 | |
| 1829 | switch(blockProperties.blockType) |
| 1830 | { |
| 1831 | case bt_compressed: |
| 1832 | decodedSize = ZSTD_decompressBlock_internal(dctx, op, oend-op, ip, cBlockSize, /* frame */ 1); |
| 1833 | break; |
| 1834 | case bt_raw : |
| 1835 | decodedSize = ZSTD_copyRawBlock(op, oend-op, ip, cBlockSize); |
| 1836 | break; |
| 1837 | case bt_rle : |
| 1838 | decodedSize = ZSTD_generateNxBytes(op, oend-op, *ip, blockProperties.origSize); |
| 1839 | break; |
| 1840 | case bt_reserved : |
| 1841 | default: |
| 1842 | return ERROR(corruption_detected); |
| 1843 | } |
| 1844 | |
| 1845 | if (ZSTD_isError(decodedSize)) return decodedSize; |
| 1846 | if (dctx->fParams.checksumFlag) |
| 1847 | XXH64_update(&dctx->xxhState, op, decodedSize); |
| 1848 | op += decodedSize; |
| 1849 | ip += cBlockSize; |
| 1850 | remainingSize -= cBlockSize; |
| 1851 | if (blockProperties.lastBlock) break; |
| 1852 | } |
| 1853 | |
| 1854 | if (dctx->fParams.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN) { |
| 1855 | if ((U64)(op-ostart) != dctx->fParams.frameContentSize) { |
| 1856 | return ERROR(corruption_detected); |
| 1857 | } } |
| 1858 | if (dctx->fParams.checksumFlag) { /* Frame content checksum verification */ |
| 1859 | U32 const checkCalc = (U32)XXH64_digest(&dctx->xxhState); |
| 1860 | U32 checkRead; |
| 1861 | if (remainingSize<4) return ERROR(checksum_wrong); |
| 1862 | checkRead = MEM_readLE32(ip); |
| 1863 | if (checkRead != checkCalc) return ERROR(checksum_wrong); |
| 1864 | ip += 4; |
| 1865 | remainingSize -= 4; |
| 1866 | } |
| 1867 | |
| 1868 | /* Allow caller to get size read */ |
| 1869 | *srcPtr = ip; |
| 1870 | *srcSizePtr = remainingSize; |
| 1871 | return op-ostart; |
| 1872 | } |
| 1873 | |
| 1874 | static const void* ZSTD_DDictDictContent(const ZSTD_DDict* ddict); |
| 1875 | static size_t ZSTD_DDictDictSize(const ZSTD_DDict* ddict); |
| 1876 | |
| 1877 | static size_t ZSTD_decompressMultiFrame(ZSTD_DCtx* dctx, |
| 1878 | void* dst, size_t dstCapacity, |
| 1879 | const void* src, size_t srcSize, |
| 1880 | const void* dict, size_t dictSize, |
| 1881 | const ZSTD_DDict* ddict) |
| 1882 | { |
| 1883 | void* const dststart = dst; |
| 1884 | assert(dict==NULL || ddict==NULL); /* either dict or ddict set, not both */ |
| 1885 | |
| 1886 | if (ddict) { |
| 1887 | dict = ZSTD_DDictDictContent(ddict); |
| 1888 | dictSize = ZSTD_DDictDictSize(ddict); |
| 1889 | } |
| 1890 | |
| 1891 | while (srcSize >= ZSTD_frameHeaderSize_prefix) { |
| 1892 | U32 magicNumber; |
| 1893 | |
| 1894 | #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT >= 1) |
| 1895 | if (ZSTD_isLegacy(src, srcSize)) { |
| 1896 | size_t decodedSize; |
| 1897 | size_t const frameSize = ZSTD_findFrameCompressedSizeLegacy(src, srcSize); |
| 1898 | if (ZSTD_isError(frameSize)) return frameSize; |
| 1899 | /* legacy support is not compatible with static dctx */ |
| 1900 | if (dctx->staticSize) return ERROR(memory_allocation); |
| 1901 | |
| 1902 | decodedSize = ZSTD_decompressLegacy(dst, dstCapacity, src, frameSize, dict, dictSize); |
| 1903 | |
| 1904 | dst = (BYTE*)dst + decodedSize; |
| 1905 | dstCapacity -= decodedSize; |
| 1906 | |
| 1907 | src = (const BYTE*)src + frameSize; |
| 1908 | srcSize -= frameSize; |
| 1909 | |
| 1910 | continue; |
| 1911 | } |
| 1912 | #endif |
| 1913 | |
| 1914 | magicNumber = MEM_readLE32(src); |
| 1915 | DEBUGLOG(4, "reading magic number %08X (expecting %08X)", |
| 1916 | (U32)magicNumber, (U32)ZSTD_MAGICNUMBER); |
| 1917 | if (magicNumber != ZSTD_MAGICNUMBER) { |
| 1918 | if ((magicNumber & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START) { |
| 1919 | size_t skippableSize; |
| 1920 | if (srcSize < ZSTD_skippableHeaderSize) |
| 1921 | return ERROR(srcSize_wrong); |
| 1922 | skippableSize = MEM_readLE32((const BYTE*)src + ZSTD_frameIdSize) |
| 1923 | + ZSTD_skippableHeaderSize; |
| 1924 | if (srcSize < skippableSize) return ERROR(srcSize_wrong); |
| 1925 | |
| 1926 | src = (const BYTE *)src + skippableSize; |
| 1927 | srcSize -= skippableSize; |
| 1928 | continue; |
| 1929 | } |
| 1930 | return ERROR(prefix_unknown); |
| 1931 | } |
| 1932 | |
| 1933 | if (ddict) { |
| 1934 | /* we were called from ZSTD_decompress_usingDDict */ |
| 1935 | CHECK_F(ZSTD_decompressBegin_usingDDict(dctx, ddict)); |
| 1936 | } else { |
| 1937 | /* this will initialize correctly with no dict if dict == NULL, so |
| 1938 | * use this in all cases but ddict */ |
| 1939 | CHECK_F(ZSTD_decompressBegin_usingDict(dctx, dict, dictSize)); |
| 1940 | } |
| 1941 | ZSTD_checkContinuity(dctx, dst); |
| 1942 | |
| 1943 | { const size_t res = ZSTD_decompressFrame(dctx, dst, dstCapacity, |
| 1944 | &src, &srcSize); |
| 1945 | if (ZSTD_isError(res)) return res; |
| 1946 | /* no need to bound check, ZSTD_decompressFrame already has */ |
| 1947 | dst = (BYTE*)dst + res; |
| 1948 | dstCapacity -= res; |
| 1949 | } |
| 1950 | } /* while (srcSize >= ZSTD_frameHeaderSize_prefix) */ |
| 1951 | |
| 1952 | if (srcSize) return ERROR(srcSize_wrong); /* input not entirely consumed */ |
| 1953 | |
| 1954 | return (BYTE*)dst - (BYTE*)dststart; |
| 1955 | } |
| 1956 | |
| 1957 | size_t ZSTD_decompress_usingDict(ZSTD_DCtx* dctx, |
| 1958 | void* dst, size_t dstCapacity, |
| 1959 | const void* src, size_t srcSize, |
| 1960 | const void* dict, size_t dictSize) |
| 1961 | { |
| 1962 | return ZSTD_decompressMultiFrame(dctx, dst, dstCapacity, src, srcSize, dict, dictSize, NULL); |
| 1963 | } |
| 1964 | |
| 1965 | |
| 1966 | size_t ZSTD_decompressDCtx(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize) |
| 1967 | { |
| 1968 | return ZSTD_decompress_usingDict(dctx, dst, dstCapacity, src, srcSize, NULL, 0); |
| 1969 | } |
| 1970 | |
| 1971 | |
| 1972 | size_t ZSTD_decompress(void* dst, size_t dstCapacity, const void* src, size_t srcSize) |
| 1973 | { |
| 1974 | #if defined(ZSTD_HEAPMODE) && (ZSTD_HEAPMODE>=1) |
| 1975 | size_t regenSize; |
| 1976 | ZSTD_DCtx* const dctx = ZSTD_createDCtx(); |
| 1977 | if (dctx==NULL) return ERROR(memory_allocation); |
| 1978 | regenSize = ZSTD_decompressDCtx(dctx, dst, dstCapacity, src, srcSize); |
| 1979 | ZSTD_freeDCtx(dctx); |
| 1980 | return regenSize; |
| 1981 | #else /* stack mode */ |
| 1982 | ZSTD_DCtx dctx; |
| 1983 | return ZSTD_decompressDCtx(&dctx, dst, dstCapacity, src, srcSize); |
| 1984 | #endif |
| 1985 | } |
| 1986 | |
| 1987 | |
| 1988 | /*-************************************** |
| 1989 | * Advanced Streaming Decompression API |
| 1990 | * Bufferless and synchronous |
| 1991 | ****************************************/ |
| 1992 | size_t ZSTD_nextSrcSizeToDecompress(ZSTD_DCtx* dctx) { return dctx->expected; } |
| 1993 | |
| 1994 | ZSTD_nextInputType_e ZSTD_nextInputType(ZSTD_DCtx* dctx) { |
| 1995 | switch(dctx->stage) |
| 1996 | { |
| 1997 | default: /* should not happen */ |
| 1998 | assert(0); |
| 1999 | case ZSTDds_getFrameHeaderSize: |
| 2000 | case ZSTDds_decodeFrameHeader: |
| 2001 | return ZSTDnit_frameHeader; |
| 2002 | case ZSTDds_decodeBlockHeader: |
| 2003 | return ZSTDnit_blockHeader; |
| 2004 | case ZSTDds_decompressBlock: |
| 2005 | return ZSTDnit_block; |
| 2006 | case ZSTDds_decompressLastBlock: |
| 2007 | return ZSTDnit_lastBlock; |
| 2008 | case ZSTDds_checkChecksum: |
| 2009 | return ZSTDnit_checksum; |
| 2010 | case ZSTDds_decodeSkippableHeader: |
| 2011 | case ZSTDds_skipFrame: |
| 2012 | return ZSTDnit_skippableFrame; |
| 2013 | } |
| 2014 | } |
| 2015 | |
| 2016 | static int ZSTD_isSkipFrame(ZSTD_DCtx* dctx) { return dctx->stage == ZSTDds_skipFrame; } |
| 2017 | |
| 2018 | /** ZSTD_decompressContinue() : |
| 2019 | * srcSize : must be the exact nb of bytes expected (see ZSTD_nextSrcSizeToDecompress()) |
| 2020 | * @return : nb of bytes generated into `dst` (necessarily <= `dstCapacity) |
| 2021 | * or an error code, which can be tested using ZSTD_isError() */ |
| 2022 | size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize) |
| 2023 | { |
| 2024 | DEBUGLOG(5, "ZSTD_decompressContinue (srcSize:%u)", (U32)srcSize); |
| 2025 | /* Sanity check */ |
| 2026 | if (srcSize != dctx->expected) return ERROR(srcSize_wrong); /* not allowed */ |
| 2027 | if (dstCapacity) ZSTD_checkContinuity(dctx, dst); |
| 2028 | |
| 2029 | switch (dctx->stage) |
| 2030 | { |
| 2031 | case ZSTDds_getFrameHeaderSize : |
| 2032 | assert(src != NULL); |
| 2033 | if (dctx->format == ZSTD_f_zstd1) { /* allows header */ |
| 2034 | assert(srcSize >= ZSTD_frameIdSize); /* to read skippable magic number */ |
| 2035 | if ((MEM_readLE32(src) & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START) { /* skippable frame */ |
| 2036 | memcpy(dctx->headerBuffer, src, srcSize); |
| 2037 | dctx->expected = ZSTD_skippableHeaderSize - srcSize; /* remaining to load to get full skippable frame header */ |
| 2038 | dctx->stage = ZSTDds_decodeSkippableHeader; |
| 2039 | return 0; |
| 2040 | } } |
| 2041 | dctx->headerSize = ZSTD_frameHeaderSize_internal(src, srcSize, dctx->format); |
| 2042 | if (ZSTD_isError(dctx->headerSize)) return dctx->headerSize; |
| 2043 | memcpy(dctx->headerBuffer, src, srcSize); |
| 2044 | dctx->expected = dctx->headerSize - srcSize; |
| 2045 | dctx->stage = ZSTDds_decodeFrameHeader; |
| 2046 | return 0; |
| 2047 | |
| 2048 | case ZSTDds_decodeFrameHeader: |
| 2049 | assert(src != NULL); |
| 2050 | memcpy(dctx->headerBuffer + (dctx->headerSize - srcSize), src, srcSize); |
| 2051 | CHECK_F(ZSTD_decodeFrameHeader(dctx, dctx->headerBuffer, dctx->headerSize)); |
| 2052 | dctx->expected = ZSTD_blockHeaderSize; |
| 2053 | dctx->stage = ZSTDds_decodeBlockHeader; |
| 2054 | return 0; |
| 2055 | |
| 2056 | case ZSTDds_decodeBlockHeader: |
| 2057 | { blockProperties_t bp; |
| 2058 | size_t const cBlockSize = ZSTD_getcBlockSize(src, ZSTD_blockHeaderSize, &bp); |
| 2059 | if (ZSTD_isError(cBlockSize)) return cBlockSize; |
| 2060 | dctx->expected = cBlockSize; |
| 2061 | dctx->bType = bp.blockType; |
| 2062 | dctx->rleSize = bp.origSize; |
| 2063 | if (cBlockSize) { |
| 2064 | dctx->stage = bp.lastBlock ? ZSTDds_decompressLastBlock : ZSTDds_decompressBlock; |
| 2065 | return 0; |
| 2066 | } |
| 2067 | /* empty block */ |
| 2068 | if (bp.lastBlock) { |
| 2069 | if (dctx->fParams.checksumFlag) { |
| 2070 | dctx->expected = 4; |
| 2071 | dctx->stage = ZSTDds_checkChecksum; |
| 2072 | } else { |
| 2073 | dctx->expected = 0; /* end of frame */ |
| 2074 | dctx->stage = ZSTDds_getFrameHeaderSize; |
| 2075 | } |
| 2076 | } else { |
| 2077 | dctx->expected = ZSTD_blockHeaderSize; /* jump to next header */ |
| 2078 | dctx->stage = ZSTDds_decodeBlockHeader; |
| 2079 | } |
| 2080 | return 0; |
| 2081 | } |
| 2082 | |
| 2083 | case ZSTDds_decompressLastBlock: |
| 2084 | case ZSTDds_decompressBlock: |
| 2085 | DEBUGLOG(5, "ZSTD_decompressContinue: case ZSTDds_decompressBlock"); |
| 2086 | { size_t rSize; |
| 2087 | switch(dctx->bType) |
| 2088 | { |
| 2089 | case bt_compressed: |
| 2090 | DEBUGLOG(5, "ZSTD_decompressContinue: case bt_compressed"); |
| 2091 | rSize = ZSTD_decompressBlock_internal(dctx, dst, dstCapacity, src, srcSize, /* frame */ 1); |
| 2092 | break; |
| 2093 | case bt_raw : |
| 2094 | rSize = ZSTD_copyRawBlock(dst, dstCapacity, src, srcSize); |
| 2095 | break; |
| 2096 | case bt_rle : |
| 2097 | rSize = ZSTD_setRleBlock(dst, dstCapacity, src, srcSize, dctx->rleSize); |
| 2098 | break; |
| 2099 | case bt_reserved : /* should never happen */ |
| 2100 | default: |
| 2101 | return ERROR(corruption_detected); |
| 2102 | } |
| 2103 | if (ZSTD_isError(rSize)) return rSize; |
| 2104 | DEBUGLOG(5, "ZSTD_decompressContinue: decoded size from block : %u", (U32)rSize); |
| 2105 | dctx->decodedSize += rSize; |
| 2106 | if (dctx->fParams.checksumFlag) XXH64_update(&dctx->xxhState, dst, rSize); |
| 2107 | |
| 2108 | if (dctx->stage == ZSTDds_decompressLastBlock) { /* end of frame */ |
| 2109 | DEBUGLOG(4, "ZSTD_decompressContinue: decoded size from frame : %u", (U32)dctx->decodedSize); |
| 2110 | if (dctx->fParams.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN) { |
| 2111 | if (dctx->decodedSize != dctx->fParams.frameContentSize) { |
| 2112 | return ERROR(corruption_detected); |
| 2113 | } } |
| 2114 | if (dctx->fParams.checksumFlag) { /* another round for frame checksum */ |
| 2115 | dctx->expected = 4; |
| 2116 | dctx->stage = ZSTDds_checkChecksum; |
| 2117 | } else { |
| 2118 | dctx->expected = 0; /* ends here */ |
| 2119 | dctx->stage = ZSTDds_getFrameHeaderSize; |
| 2120 | } |
| 2121 | } else { |
| 2122 | dctx->stage = ZSTDds_decodeBlockHeader; |
| 2123 | dctx->expected = ZSTD_blockHeaderSize; |
| 2124 | dctx->previousDstEnd = (char*)dst + rSize; |
| 2125 | } |
| 2126 | return rSize; |
| 2127 | } |
| 2128 | |
| 2129 | case ZSTDds_checkChecksum: |
| 2130 | assert(srcSize == 4); /* guaranteed by dctx->expected */ |
| 2131 | { U32 const h32 = (U32)XXH64_digest(&dctx->xxhState); |
| 2132 | U32 const check32 = MEM_readLE32(src); |
| 2133 | DEBUGLOG(4, "ZSTD_decompressContinue: checksum : calculated %08X :: %08X read", h32, check32); |
| 2134 | if (check32 != h32) return ERROR(checksum_wrong); |
| 2135 | dctx->expected = 0; |
| 2136 | dctx->stage = ZSTDds_getFrameHeaderSize; |
| 2137 | return 0; |
| 2138 | } |
| 2139 | |
| 2140 | case ZSTDds_decodeSkippableHeader: |
| 2141 | assert(src != NULL); |
| 2142 | assert(srcSize <= ZSTD_skippableHeaderSize); |
| 2143 | memcpy(dctx->headerBuffer + (ZSTD_skippableHeaderSize - srcSize), src, srcSize); /* complete skippable header */ |
| 2144 | dctx->expected = MEM_readLE32(dctx->headerBuffer + ZSTD_frameIdSize); /* note : dctx->expected can grow seriously large, beyond local buffer size */ |
| 2145 | dctx->stage = ZSTDds_skipFrame; |
| 2146 | return 0; |
| 2147 | |
| 2148 | case ZSTDds_skipFrame: |
| 2149 | dctx->expected = 0; |
| 2150 | dctx->stage = ZSTDds_getFrameHeaderSize; |
| 2151 | return 0; |
| 2152 | |
| 2153 | default: |
| 2154 | return ERROR(GENERIC); /* impossible */ |
| 2155 | } |
| 2156 | } |
| 2157 | |
| 2158 | |
| 2159 | static size_t ZSTD_refDictContent(ZSTD_DCtx* dctx, const void* dict, size_t dictSize) |
| 2160 | { |
| 2161 | dctx->dictEnd = dctx->previousDstEnd; |
| 2162 | dctx->vBase = (const char*)dict - ((const char*)(dctx->previousDstEnd) - (const char*)(dctx->base)); |
| 2163 | dctx->base = dict; |
| 2164 | dctx->previousDstEnd = (const char*)dict + dictSize; |
| 2165 | return 0; |
| 2166 | } |
| 2167 | |
| 2168 | /* ZSTD_loadEntropy() : |
| 2169 | * dict : must point at beginning of a valid zstd dictionary |
| 2170 | * @return : size of entropy tables read */ |
| 2171 | static size_t ZSTD_loadEntropy(ZSTD_entropyDTables_t* entropy, const void* const dict, size_t const dictSize) |
| 2172 | { |
| 2173 | const BYTE* dictPtr = (const BYTE*)dict; |
| 2174 | const BYTE* const dictEnd = dictPtr + dictSize; |
| 2175 | |
| 2176 | if (dictSize <= 8) return ERROR(dictionary_corrupted); |
| 2177 | dictPtr += 8; /* skip header = magic + dictID */ |
| 2178 | |
| 2179 | |
| 2180 | { size_t const hSize = HUF_readDTableX4_wksp( |
| 2181 | entropy->hufTable, dictPtr, dictEnd - dictPtr, |
| 2182 | entropy->workspace, sizeof(entropy->workspace)); |
| 2183 | if (HUF_isError(hSize)) return ERROR(dictionary_corrupted); |
| 2184 | dictPtr += hSize; |
| 2185 | } |
| 2186 | |
| 2187 | { short offcodeNCount[MaxOff+1]; |
| 2188 | U32 offcodeMaxValue = MaxOff, offcodeLog; |
| 2189 | size_t const offcodeHeaderSize = FSE_readNCount(offcodeNCount, &offcodeMaxValue, &offcodeLog, dictPtr, dictEnd-dictPtr); |
| 2190 | if (FSE_isError(offcodeHeaderSize)) return ERROR(dictionary_corrupted); |
| 2191 | if (offcodeMaxValue > MaxOff) return ERROR(dictionary_corrupted); |
| 2192 | if (offcodeLog > OffFSELog) return ERROR(dictionary_corrupted); |
| 2193 | ZSTD_buildFSETable(entropy->OFTable, |
| 2194 | offcodeNCount, offcodeMaxValue, |
| 2195 | OF_base, OF_bits, |
| 2196 | offcodeLog); |
| 2197 | dictPtr += offcodeHeaderSize; |
| 2198 | } |
| 2199 | |
| 2200 | { short matchlengthNCount[MaxML+1]; |
| 2201 | unsigned matchlengthMaxValue = MaxML, matchlengthLog; |
| 2202 | size_t const matchlengthHeaderSize = FSE_readNCount(matchlengthNCount, &matchlengthMaxValue, &matchlengthLog, dictPtr, dictEnd-dictPtr); |
| 2203 | if (FSE_isError(matchlengthHeaderSize)) return ERROR(dictionary_corrupted); |
| 2204 | if (matchlengthMaxValue > MaxML) return ERROR(dictionary_corrupted); |
| 2205 | if (matchlengthLog > MLFSELog) return ERROR(dictionary_corrupted); |
| 2206 | ZSTD_buildFSETable(entropy->MLTable, |
| 2207 | matchlengthNCount, matchlengthMaxValue, |
| 2208 | ML_base, ML_bits, |
| 2209 | matchlengthLog); |
| 2210 | dictPtr += matchlengthHeaderSize; |
| 2211 | } |
| 2212 | |
| 2213 | { short litlengthNCount[MaxLL+1]; |
| 2214 | unsigned litlengthMaxValue = MaxLL, litlengthLog; |
| 2215 | size_t const litlengthHeaderSize = FSE_readNCount(litlengthNCount, &litlengthMaxValue, &litlengthLog, dictPtr, dictEnd-dictPtr); |
| 2216 | if (FSE_isError(litlengthHeaderSize)) return ERROR(dictionary_corrupted); |
| 2217 | if (litlengthMaxValue > MaxLL) return ERROR(dictionary_corrupted); |
| 2218 | if (litlengthLog > LLFSELog) return ERROR(dictionary_corrupted); |
| 2219 | ZSTD_buildFSETable(entropy->LLTable, |
| 2220 | litlengthNCount, litlengthMaxValue, |
| 2221 | LL_base, LL_bits, |
| 2222 | litlengthLog); |
| 2223 | dictPtr += litlengthHeaderSize; |
| 2224 | } |
| 2225 | |
| 2226 | if (dictPtr+12 > dictEnd) return ERROR(dictionary_corrupted); |
| 2227 | { int i; |
| 2228 | size_t const dictContentSize = (size_t)(dictEnd - (dictPtr+12)); |
| 2229 | for (i=0; i<3; i++) { |
| 2230 | U32 const rep = MEM_readLE32(dictPtr); dictPtr += 4; |
| 2231 | if (rep==0 || rep >= dictContentSize) return ERROR(dictionary_corrupted); |
| 2232 | entropy->rep[i] = rep; |
| 2233 | } } |
| 2234 | |
| 2235 | return dictPtr - (const BYTE*)dict; |
| 2236 | } |
| 2237 | |
| 2238 | static size_t ZSTD_decompress_insertDictionary(ZSTD_DCtx* dctx, const void* dict, size_t dictSize) |
| 2239 | { |
| 2240 | if (dictSize < 8) return ZSTD_refDictContent(dctx, dict, dictSize); |
| 2241 | { U32 const magic = MEM_readLE32(dict); |
| 2242 | if (magic != ZSTD_MAGIC_DICTIONARY) { |
| 2243 | return ZSTD_refDictContent(dctx, dict, dictSize); /* pure content mode */ |
| 2244 | } } |
| 2245 | dctx->dictID = MEM_readLE32((const char*)dict + ZSTD_frameIdSize); |
| 2246 | |
| 2247 | /* load entropy tables */ |
| 2248 | { size_t const eSize = ZSTD_loadEntropy(&dctx->entropy, dict, dictSize); |
| 2249 | if (ZSTD_isError(eSize)) return ERROR(dictionary_corrupted); |
| 2250 | dict = (const char*)dict + eSize; |
| 2251 | dictSize -= eSize; |
| 2252 | } |
| 2253 | dctx->litEntropy = dctx->fseEntropy = 1; |
| 2254 | |
| 2255 | /* reference dictionary content */ |
| 2256 | return ZSTD_refDictContent(dctx, dict, dictSize); |
| 2257 | } |
| 2258 | |
| 2259 | /* Note : this function cannot fail */ |
| 2260 | size_t ZSTD_decompressBegin(ZSTD_DCtx* dctx) |
| 2261 | { |
| 2262 | assert(dctx != NULL); |
| 2263 | dctx->expected = ZSTD_startingInputLength(dctx->format); /* dctx->format must be properly set */ |
| 2264 | dctx->stage = ZSTDds_getFrameHeaderSize; |
| 2265 | dctx->decodedSize = 0; |
| 2266 | dctx->previousDstEnd = NULL; |
| 2267 | dctx->base = NULL; |
| 2268 | dctx->vBase = NULL; |
| 2269 | dctx->dictEnd = NULL; |
| 2270 | dctx->entropy.hufTable[0] = (HUF_DTable)((HufLog)*0x1000001); /* cover both little and big endian */ |
| 2271 | dctx->litEntropy = dctx->fseEntropy = 0; |
| 2272 | dctx->dictID = 0; |
| 2273 | ZSTD_STATIC_ASSERT(sizeof(dctx->entropy.rep) == sizeof(repStartValue)); |
| 2274 | memcpy(dctx->entropy.rep, repStartValue, sizeof(repStartValue)); /* initial repcodes */ |
| 2275 | dctx->LLTptr = dctx->entropy.LLTable; |
| 2276 | dctx->MLTptr = dctx->entropy.MLTable; |
| 2277 | dctx->OFTptr = dctx->entropy.OFTable; |
| 2278 | dctx->HUFptr = dctx->entropy.hufTable; |
| 2279 | return 0; |
| 2280 | } |
| 2281 | |
| 2282 | size_t ZSTD_decompressBegin_usingDict(ZSTD_DCtx* dctx, const void* dict, size_t dictSize) |
| 2283 | { |
| 2284 | CHECK_F( ZSTD_decompressBegin(dctx) ); |
| 2285 | if (dict && dictSize) |
| 2286 | CHECK_E(ZSTD_decompress_insertDictionary(dctx, dict, dictSize), dictionary_corrupted); |
| 2287 | return 0; |
| 2288 | } |
| 2289 | |
| 2290 | |
| 2291 | /* ====== ZSTD_DDict ====== */ |
| 2292 | |
| 2293 | struct ZSTD_DDict_s { |
| 2294 | void* dictBuffer; |
| 2295 | const void* dictContent; |
| 2296 | size_t dictSize; |
| 2297 | ZSTD_entropyDTables_t entropy; |
| 2298 | U32 dictID; |
| 2299 | U32 entropyPresent; |
| 2300 | ZSTD_customMem cMem; |
| 2301 | }; /* typedef'd to ZSTD_DDict within "zstd.h" */ |
| 2302 | |
| 2303 | static const void* ZSTD_DDictDictContent(const ZSTD_DDict* ddict) |
| 2304 | { |
| 2305 | return ddict->dictContent; |
| 2306 | } |
| 2307 | |
| 2308 | static size_t ZSTD_DDictDictSize(const ZSTD_DDict* ddict) |
| 2309 | { |
| 2310 | return ddict->dictSize; |
| 2311 | } |
| 2312 | |
| 2313 | size_t ZSTD_decompressBegin_usingDDict(ZSTD_DCtx* dstDCtx, const ZSTD_DDict* ddict) |
| 2314 | { |
| 2315 | CHECK_F( ZSTD_decompressBegin(dstDCtx) ); |
| 2316 | if (ddict) { /* support begin on NULL */ |
| 2317 | dstDCtx->dictID = ddict->dictID; |
| 2318 | dstDCtx->base = ddict->dictContent; |
| 2319 | dstDCtx->vBase = ddict->dictContent; |
| 2320 | dstDCtx->dictEnd = (const BYTE*)ddict->dictContent + ddict->dictSize; |
| 2321 | dstDCtx->previousDstEnd = dstDCtx->dictEnd; |
| 2322 | if (ddict->entropyPresent) { |
| 2323 | dstDCtx->litEntropy = 1; |
| 2324 | dstDCtx->fseEntropy = 1; |
| 2325 | dstDCtx->LLTptr = ddict->entropy.LLTable; |
| 2326 | dstDCtx->MLTptr = ddict->entropy.MLTable; |
| 2327 | dstDCtx->OFTptr = ddict->entropy.OFTable; |
| 2328 | dstDCtx->HUFptr = ddict->entropy.hufTable; |
| 2329 | dstDCtx->entropy.rep[0] = ddict->entropy.rep[0]; |
| 2330 | dstDCtx->entropy.rep[1] = ddict->entropy.rep[1]; |
| 2331 | dstDCtx->entropy.rep[2] = ddict->entropy.rep[2]; |
| 2332 | } else { |
| 2333 | dstDCtx->litEntropy = 0; |
| 2334 | dstDCtx->fseEntropy = 0; |
| 2335 | } |
| 2336 | } |
| 2337 | return 0; |
| 2338 | } |
| 2339 | |
| 2340 | static size_t ZSTD_loadEntropy_inDDict(ZSTD_DDict* ddict, ZSTD_dictContentType_e dictContentType) |
| 2341 | { |
| 2342 | ddict->dictID = 0; |
| 2343 | ddict->entropyPresent = 0; |
| 2344 | if (dictContentType == ZSTD_dct_rawContent) return 0; |
| 2345 | |
| 2346 | if (ddict->dictSize < 8) { |
| 2347 | if (dictContentType == ZSTD_dct_fullDict) |
| 2348 | return ERROR(dictionary_corrupted); /* only accept specified dictionaries */ |
| 2349 | return 0; /* pure content mode */ |
| 2350 | } |
| 2351 | { U32 const magic = MEM_readLE32(ddict->dictContent); |
| 2352 | if (magic != ZSTD_MAGIC_DICTIONARY) { |
| 2353 | if (dictContentType == ZSTD_dct_fullDict) |
| 2354 | return ERROR(dictionary_corrupted); /* only accept specified dictionaries */ |
| 2355 | return 0; /* pure content mode */ |
| 2356 | } |
| 2357 | } |
| 2358 | ddict->dictID = MEM_readLE32((const char*)ddict->dictContent + ZSTD_frameIdSize); |
| 2359 | |
| 2360 | /* load entropy tables */ |
| 2361 | CHECK_E( ZSTD_loadEntropy(&ddict->entropy, ddict->dictContent, ddict->dictSize), dictionary_corrupted ); |
| 2362 | ddict->entropyPresent = 1; |
| 2363 | return 0; |
| 2364 | } |
| 2365 | |
| 2366 | |
| 2367 | static size_t ZSTD_initDDict_internal(ZSTD_DDict* ddict, |
| 2368 | const void* dict, size_t dictSize, |
| 2369 | ZSTD_dictLoadMethod_e dictLoadMethod, |
| 2370 | ZSTD_dictContentType_e dictContentType) |
| 2371 | { |
| 2372 | if ((dictLoadMethod == ZSTD_dlm_byRef) || (!dict) || (!dictSize)) { |
| 2373 | ddict->dictBuffer = NULL; |
| 2374 | ddict->dictContent = dict; |
| 2375 | } else { |
| 2376 | void* const internalBuffer = ZSTD_malloc(dictSize, ddict->cMem); |
| 2377 | ddict->dictBuffer = internalBuffer; |
| 2378 | ddict->dictContent = internalBuffer; |
| 2379 | if (!internalBuffer) return ERROR(memory_allocation); |
| 2380 | memcpy(internalBuffer, dict, dictSize); |
| 2381 | } |
| 2382 | ddict->dictSize = dictSize; |
| 2383 | ddict->entropy.hufTable[0] = (HUF_DTable)((HufLog)*0x1000001); /* cover both little and big endian */ |
| 2384 | |
| 2385 | /* parse dictionary content */ |
| 2386 | CHECK_F( ZSTD_loadEntropy_inDDict(ddict, dictContentType) ); |
| 2387 | |
| 2388 | return 0; |
| 2389 | } |
| 2390 | |
| 2391 | ZSTD_DDict* ZSTD_createDDict_advanced(const void* dict, size_t dictSize, |
| 2392 | ZSTD_dictLoadMethod_e dictLoadMethod, |
| 2393 | ZSTD_dictContentType_e dictContentType, |
| 2394 | ZSTD_customMem customMem) |
| 2395 | { |
| 2396 | if (!customMem.customAlloc ^ !customMem.customFree) return NULL; |
| 2397 | |
| 2398 | { ZSTD_DDict* const ddict = (ZSTD_DDict*) ZSTD_malloc(sizeof(ZSTD_DDict), customMem); |
| 2399 | if (!ddict) return NULL; |
| 2400 | ddict->cMem = customMem; |
| 2401 | |
| 2402 | if (ZSTD_isError( ZSTD_initDDict_internal(ddict, dict, dictSize, dictLoadMethod, dictContentType) )) { |
| 2403 | ZSTD_freeDDict(ddict); |
| 2404 | return NULL; |
| 2405 | } |
| 2406 | |
| 2407 | return ddict; |
| 2408 | } |
| 2409 | } |
| 2410 | |
| 2411 | /*! ZSTD_createDDict() : |
| 2412 | * Create a digested dictionary, to start decompression without startup delay. |
| 2413 | * `dict` content is copied inside DDict. |
| 2414 | * Consequently, `dict` can be released after `ZSTD_DDict` creation */ |
| 2415 | ZSTD_DDict* ZSTD_createDDict(const void* dict, size_t dictSize) |
| 2416 | { |
| 2417 | ZSTD_customMem const allocator = { NULL, NULL, NULL }; |
| 2418 | return ZSTD_createDDict_advanced(dict, dictSize, ZSTD_dlm_byCopy, ZSTD_dct_auto, allocator); |
| 2419 | } |
| 2420 | |
| 2421 | /*! ZSTD_createDDict_byReference() : |
| 2422 | * Create a digested dictionary, to start decompression without startup delay. |
| 2423 | * Dictionary content is simply referenced, it will be accessed during decompression. |
| 2424 | * Warning : dictBuffer must outlive DDict (DDict must be freed before dictBuffer) */ |
| 2425 | ZSTD_DDict* ZSTD_createDDict_byReference(const void* dictBuffer, size_t dictSize) |
| 2426 | { |
| 2427 | ZSTD_customMem const allocator = { NULL, NULL, NULL }; |
| 2428 | return ZSTD_createDDict_advanced(dictBuffer, dictSize, ZSTD_dlm_byRef, ZSTD_dct_auto, allocator); |
| 2429 | } |
| 2430 | |
| 2431 | |
| 2432 | const ZSTD_DDict* ZSTD_initStaticDDict( |
| 2433 | void* workspace, size_t workspaceSize, |
| 2434 | const void* dict, size_t dictSize, |
| 2435 | ZSTD_dictLoadMethod_e dictLoadMethod, |
| 2436 | ZSTD_dictContentType_e dictContentType) |
| 2437 | { |
| 2438 | size_t const neededSpace = |
| 2439 | sizeof(ZSTD_DDict) + (dictLoadMethod == ZSTD_dlm_byRef ? 0 : dictSize); |
| 2440 | ZSTD_DDict* const ddict = (ZSTD_DDict*)workspace; |
| 2441 | assert(workspace != NULL); |
| 2442 | assert(dict != NULL); |
| 2443 | if ((size_t)workspace & 7) return NULL; /* 8-aligned */ |
| 2444 | if (workspaceSize < neededSpace) return NULL; |
| 2445 | if (dictLoadMethod == ZSTD_dlm_byCopy) { |
| 2446 | memcpy(ddict+1, dict, dictSize); /* local copy */ |
| 2447 | dict = ddict+1; |
| 2448 | } |
| 2449 | if (ZSTD_isError( ZSTD_initDDict_internal(ddict, dict, dictSize, ZSTD_dlm_byRef, dictContentType) )) |
| 2450 | return NULL; |
| 2451 | return ddict; |
| 2452 | } |
| 2453 | |
| 2454 | |
| 2455 | size_t ZSTD_freeDDict(ZSTD_DDict* ddict) |
| 2456 | { |
| 2457 | if (ddict==NULL) return 0; /* support free on NULL */ |
| 2458 | { ZSTD_customMem const cMem = ddict->cMem; |
| 2459 | ZSTD_free(ddict->dictBuffer, cMem); |
| 2460 | ZSTD_free(ddict, cMem); |
| 2461 | return 0; |
| 2462 | } |
| 2463 | } |
| 2464 | |
| 2465 | /*! ZSTD_estimateDDictSize() : |
| 2466 | * Estimate amount of memory that will be needed to create a dictionary for decompression. |
| 2467 | * Note : dictionary created by reference using ZSTD_dlm_byRef are smaller */ |
| 2468 | size_t ZSTD_estimateDDictSize(size_t dictSize, ZSTD_dictLoadMethod_e dictLoadMethod) |
| 2469 | { |
| 2470 | return sizeof(ZSTD_DDict) + (dictLoadMethod == ZSTD_dlm_byRef ? 0 : dictSize); |
| 2471 | } |
| 2472 | |
| 2473 | size_t ZSTD_sizeof_DDict(const ZSTD_DDict* ddict) |
| 2474 | { |
| 2475 | if (ddict==NULL) return 0; /* support sizeof on NULL */ |
| 2476 | return sizeof(*ddict) + (ddict->dictBuffer ? ddict->dictSize : 0) ; |
| 2477 | } |
| 2478 | |
| 2479 | /*! ZSTD_getDictID_fromDict() : |
| 2480 | * Provides the dictID stored within dictionary. |
| 2481 | * if @return == 0, the dictionary is not conformant with Zstandard specification. |
| 2482 | * It can still be loaded, but as a content-only dictionary. */ |
| 2483 | unsigned ZSTD_getDictID_fromDict(const void* dict, size_t dictSize) |
| 2484 | { |
| 2485 | if (dictSize < 8) return 0; |
| 2486 | if (MEM_readLE32(dict) != ZSTD_MAGIC_DICTIONARY) return 0; |
| 2487 | return MEM_readLE32((const char*)dict + ZSTD_frameIdSize); |
| 2488 | } |
| 2489 | |
| 2490 | /*! ZSTD_getDictID_fromDDict() : |
| 2491 | * Provides the dictID of the dictionary loaded into `ddict`. |
| 2492 | * If @return == 0, the dictionary is not conformant to Zstandard specification, or empty. |
| 2493 | * Non-conformant dictionaries can still be loaded, but as content-only dictionaries. */ |
| 2494 | unsigned ZSTD_getDictID_fromDDict(const ZSTD_DDict* ddict) |
| 2495 | { |
| 2496 | if (ddict==NULL) return 0; |
| 2497 | return ZSTD_getDictID_fromDict(ddict->dictContent, ddict->dictSize); |
| 2498 | } |
| 2499 | |
| 2500 | /*! ZSTD_getDictID_fromFrame() : |
| 2501 | * Provides the dictID required to decompresse frame stored within `src`. |
| 2502 | * If @return == 0, the dictID could not be decoded. |
| 2503 | * This could for one of the following reasons : |
| 2504 | * - The frame does not require a dictionary (most common case). |
| 2505 | * - The frame was built with dictID intentionally removed. |
| 2506 | * Needed dictionary is a hidden information. |
| 2507 | * Note : this use case also happens when using a non-conformant dictionary. |
| 2508 | * - `srcSize` is too small, and as a result, frame header could not be decoded. |
| 2509 | * Note : possible if `srcSize < ZSTD_FRAMEHEADERSIZE_MAX`. |
| 2510 | * - This is not a Zstandard frame. |
| 2511 | * When identifying the exact failure cause, it's possible to use |
| 2512 | * ZSTD_getFrameHeader(), which will provide a more precise error code. */ |
| 2513 | unsigned ZSTD_getDictID_fromFrame(const void* src, size_t srcSize) |
| 2514 | { |
| 2515 | ZSTD_frameHeader zfp = { 0, 0, 0, ZSTD_frame, 0, 0, 0 }; |
| 2516 | size_t const hError = ZSTD_getFrameHeader(&zfp, src, srcSize); |
| 2517 | if (ZSTD_isError(hError)) return 0; |
| 2518 | return zfp.dictID; |
| 2519 | } |
| 2520 | |
| 2521 | |
| 2522 | /*! ZSTD_decompress_usingDDict() : |
| 2523 | * Decompression using a pre-digested Dictionary |
| 2524 | * Use dictionary without significant overhead. */ |
| 2525 | size_t ZSTD_decompress_usingDDict(ZSTD_DCtx* dctx, |
| 2526 | void* dst, size_t dstCapacity, |
| 2527 | const void* src, size_t srcSize, |
| 2528 | const ZSTD_DDict* ddict) |
| 2529 | { |
| 2530 | /* pass content and size in case legacy frames are encountered */ |
| 2531 | return ZSTD_decompressMultiFrame(dctx, dst, dstCapacity, src, srcSize, |
| 2532 | NULL, 0, |
| 2533 | ddict); |
| 2534 | } |
| 2535 | |
| 2536 | |
| 2537 | /*===================================== |
| 2538 | * Streaming decompression |
| 2539 | *====================================*/ |
| 2540 | |
| 2541 | ZSTD_DStream* ZSTD_createDStream(void) |
| 2542 | { |
| 2543 | DEBUGLOG(3, "ZSTD_createDStream"); |
| 2544 | return ZSTD_createDStream_advanced(ZSTD_defaultCMem); |
| 2545 | } |
| 2546 | |
| 2547 | ZSTD_DStream* ZSTD_initStaticDStream(void *workspace, size_t workspaceSize) |
| 2548 | { |
| 2549 | return ZSTD_initStaticDCtx(workspace, workspaceSize); |
| 2550 | } |
| 2551 | |
| 2552 | ZSTD_DStream* ZSTD_createDStream_advanced(ZSTD_customMem customMem) |
| 2553 | { |
| 2554 | return ZSTD_createDCtx_advanced(customMem); |
| 2555 | } |
| 2556 | |
| 2557 | size_t ZSTD_freeDStream(ZSTD_DStream* zds) |
| 2558 | { |
| 2559 | return ZSTD_freeDCtx(zds); |
| 2560 | } |
| 2561 | |
| 2562 | |
| 2563 | /* *** Initialization *** */ |
| 2564 | |
| 2565 | size_t ZSTD_DStreamInSize(void) { return ZSTD_BLOCKSIZE_MAX + ZSTD_blockHeaderSize; } |
| 2566 | size_t ZSTD_DStreamOutSize(void) { return ZSTD_BLOCKSIZE_MAX; } |
| 2567 | |
| 2568 | size_t ZSTD_DCtx_loadDictionary_advanced(ZSTD_DCtx* dctx, const void* dict, size_t dictSize, ZSTD_dictLoadMethod_e dictLoadMethod, ZSTD_dictContentType_e dictContentType) |
| 2569 | { |
| 2570 | if (dctx->streamStage != zdss_init) return ERROR(stage_wrong); |
| 2571 | ZSTD_freeDDict(dctx->ddictLocal); |
| 2572 | if (dict && dictSize >= 8) { |
| 2573 | dctx->ddictLocal = ZSTD_createDDict_advanced(dict, dictSize, dictLoadMethod, dictContentType, dctx->customMem); |
| 2574 | if (dctx->ddictLocal == NULL) return ERROR(memory_allocation); |
| 2575 | } else { |
| 2576 | dctx->ddictLocal = NULL; |
| 2577 | } |
| 2578 | dctx->ddict = dctx->ddictLocal; |
| 2579 | return 0; |
| 2580 | } |
| 2581 | |
| 2582 | size_t ZSTD_DCtx_loadDictionary_byReference(ZSTD_DCtx* dctx, const void* dict, size_t dictSize) |
| 2583 | { |
| 2584 | return ZSTD_DCtx_loadDictionary_advanced(dctx, dict, dictSize, ZSTD_dlm_byRef, ZSTD_dct_auto); |
| 2585 | } |
| 2586 | |
| 2587 | size_t ZSTD_DCtx_loadDictionary(ZSTD_DCtx* dctx, const void* dict, size_t dictSize) |
| 2588 | { |
| 2589 | return ZSTD_DCtx_loadDictionary_advanced(dctx, dict, dictSize, ZSTD_dlm_byCopy, ZSTD_dct_auto); |
| 2590 | } |
| 2591 | |
| 2592 | size_t ZSTD_DCtx_refPrefix_advanced(ZSTD_DCtx* dctx, const void* prefix, size_t prefixSize, ZSTD_dictContentType_e dictContentType) |
| 2593 | { |
| 2594 | return ZSTD_DCtx_loadDictionary_advanced(dctx, prefix, prefixSize, ZSTD_dlm_byRef, dictContentType); |
| 2595 | } |
| 2596 | |
| 2597 | size_t ZSTD_DCtx_refPrefix(ZSTD_DCtx* dctx, const void* prefix, size_t prefixSize) |
| 2598 | { |
| 2599 | return ZSTD_DCtx_refPrefix_advanced(dctx, prefix, prefixSize, ZSTD_dct_rawContent); |
| 2600 | } |
| 2601 | |
| 2602 | |
| 2603 | /* ZSTD_initDStream_usingDict() : |
| 2604 | * return : expected size, aka ZSTD_frameHeaderSize_prefix. |
| 2605 | * this function cannot fail */ |
| 2606 | size_t ZSTD_initDStream_usingDict(ZSTD_DStream* zds, const void* dict, size_t dictSize) |
| 2607 | { |
| 2608 | DEBUGLOG(4, "ZSTD_initDStream_usingDict"); |
| 2609 | zds->streamStage = zdss_init; |
| 2610 | CHECK_F( ZSTD_DCtx_loadDictionary(zds, dict, dictSize) ); |
| 2611 | return ZSTD_frameHeaderSize_prefix; |
| 2612 | } |
| 2613 | |
| 2614 | /* note : this variant can't fail */ |
| 2615 | size_t ZSTD_initDStream(ZSTD_DStream* zds) |
| 2616 | { |
| 2617 | DEBUGLOG(4, "ZSTD_initDStream"); |
| 2618 | return ZSTD_initDStream_usingDict(zds, NULL, 0); |
| 2619 | } |
| 2620 | |
| 2621 | size_t ZSTD_DCtx_refDDict(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict) |
| 2622 | { |
| 2623 | if (dctx->streamStage != zdss_init) return ERROR(stage_wrong); |
| 2624 | dctx->ddict = ddict; |
| 2625 | return 0; |
| 2626 | } |
| 2627 | |
| 2628 | /* ZSTD_initDStream_usingDDict() : |
| 2629 | * ddict will just be referenced, and must outlive decompression session |
| 2630 | * this function cannot fail */ |
| 2631 | size_t ZSTD_initDStream_usingDDict(ZSTD_DStream* dctx, const ZSTD_DDict* ddict) |
| 2632 | { |
| 2633 | size_t const initResult = ZSTD_initDStream(dctx); |
| 2634 | dctx->ddict = ddict; |
| 2635 | return initResult; |
| 2636 | } |
| 2637 | |
| 2638 | /* ZSTD_resetDStream() : |
| 2639 | * return : expected size, aka ZSTD_frameHeaderSize_prefix. |
| 2640 | * this function cannot fail */ |
| 2641 | size_t ZSTD_resetDStream(ZSTD_DStream* dctx) |
| 2642 | { |
| 2643 | DEBUGLOG(4, "ZSTD_resetDStream"); |
| 2644 | dctx->streamStage = zdss_loadHeader; |
| 2645 | dctx->lhSize = dctx->inPos = dctx->outStart = dctx->outEnd = 0; |
| 2646 | dctx->legacyVersion = 0; |
| 2647 | dctx->hostageByte = 0; |
| 2648 | return ZSTD_frameHeaderSize_prefix; |
| 2649 | } |
| 2650 | |
| 2651 | size_t ZSTD_setDStreamParameter(ZSTD_DStream* dctx, |
| 2652 | ZSTD_DStreamParameter_e paramType, unsigned paramValue) |
| 2653 | { |
| 2654 | if (dctx->streamStage != zdss_init) return ERROR(stage_wrong); |
| 2655 | switch(paramType) |
| 2656 | { |
| 2657 | default : return ERROR(parameter_unsupported); |
| 2658 | case DStream_p_maxWindowSize : |
| 2659 | DEBUGLOG(4, "setting maxWindowSize = %u KB", paramValue >> 10); |
| 2660 | dctx->maxWindowSize = paramValue ? paramValue : (U32)(-1); |
| 2661 | break; |
| 2662 | } |
| 2663 | return 0; |
| 2664 | } |
| 2665 | |
| 2666 | size_t ZSTD_DCtx_setMaxWindowSize(ZSTD_DCtx* dctx, size_t maxWindowSize) |
| 2667 | { |
| 2668 | if (dctx->streamStage != zdss_init) return ERROR(stage_wrong); |
| 2669 | dctx->maxWindowSize = maxWindowSize; |
| 2670 | return 0; |
| 2671 | } |
| 2672 | |
| 2673 | size_t ZSTD_DCtx_setFormat(ZSTD_DCtx* dctx, ZSTD_format_e format) |
| 2674 | { |
| 2675 | DEBUGLOG(4, "ZSTD_DCtx_setFormat : %u", (unsigned)format); |
| 2676 | if (dctx->streamStage != zdss_init) return ERROR(stage_wrong); |
| 2677 | dctx->format = format; |
| 2678 | return 0; |
| 2679 | } |
| 2680 | |
| 2681 | |
| 2682 | size_t ZSTD_sizeof_DStream(const ZSTD_DStream* dctx) |
| 2683 | { |
| 2684 | return ZSTD_sizeof_DCtx(dctx); |
| 2685 | } |
| 2686 | |
| 2687 | size_t ZSTD_decodingBufferSize_min(unsigned long long windowSize, unsigned long long frameContentSize) |
| 2688 | { |
| 2689 | size_t const blockSize = (size_t) MIN(windowSize, ZSTD_BLOCKSIZE_MAX); |
| 2690 | unsigned long long const neededRBSize = windowSize + blockSize + (WILDCOPY_OVERLENGTH * 2); |
| 2691 | unsigned long long const neededSize = MIN(frameContentSize, neededRBSize); |
| 2692 | size_t const minRBSize = (size_t) neededSize; |
| 2693 | if ((unsigned long long)minRBSize != neededSize) return ERROR(frameParameter_windowTooLarge); |
| 2694 | return minRBSize; |
| 2695 | } |
| 2696 | |
| 2697 | size_t ZSTD_estimateDStreamSize(size_t windowSize) |
| 2698 | { |
| 2699 | size_t const blockSize = MIN(windowSize, ZSTD_BLOCKSIZE_MAX); |
| 2700 | size_t const inBuffSize = blockSize; /* no block can be larger */ |
| 2701 | size_t const outBuffSize = ZSTD_decodingBufferSize_min(windowSize, ZSTD_CONTENTSIZE_UNKNOWN); |
| 2702 | return ZSTD_estimateDCtxSize() + inBuffSize + outBuffSize; |
| 2703 | } |
| 2704 | |
| 2705 | size_t ZSTD_estimateDStreamSize_fromFrame(const void* src, size_t srcSize) |
| 2706 | { |
| 2707 | U32 const windowSizeMax = 1U << ZSTD_WINDOWLOG_MAX; /* note : should be user-selectable */ |
| 2708 | ZSTD_frameHeader zfh; |
| 2709 | size_t const err = ZSTD_getFrameHeader(&zfh, src, srcSize); |
| 2710 | if (ZSTD_isError(err)) return err; |
| 2711 | if (err>0) return ERROR(srcSize_wrong); |
| 2712 | if (zfh.windowSize > windowSizeMax) |
| 2713 | return ERROR(frameParameter_windowTooLarge); |
| 2714 | return ZSTD_estimateDStreamSize((size_t)zfh.windowSize); |
| 2715 | } |
| 2716 | |
| 2717 | |
| 2718 | /* ***** Decompression ***** */ |
| 2719 | |
| 2720 | MEM_STATIC size_t ZSTD_limitCopy(void* dst, size_t dstCapacity, const void* src, size_t srcSize) |
| 2721 | { |
| 2722 | size_t const length = MIN(dstCapacity, srcSize); |
| 2723 | memcpy(dst, src, length); |
| 2724 | return length; |
| 2725 | } |
| 2726 | |
| 2727 | |
| 2728 | size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inBuffer* input) |
| 2729 | { |
| 2730 | const char* const istart = (const char*)(input->src) + input->pos; |
| 2731 | const char* const iend = (const char*)(input->src) + input->size; |
| 2732 | const char* ip = istart; |
| 2733 | char* const ostart = (char*)(output->dst) + output->pos; |
| 2734 | char* const oend = (char*)(output->dst) + output->size; |
| 2735 | char* op = ostart; |
| 2736 | U32 someMoreWork = 1; |
| 2737 | |
| 2738 | DEBUGLOG(5, "ZSTD_decompressStream"); |
| 2739 | if (input->pos > input->size) { /* forbidden */ |
| 2740 | DEBUGLOG(5, "in: pos: %u vs size: %u", |
| 2741 | (U32)input->pos, (U32)input->size); |
| 2742 | return ERROR(srcSize_wrong); |
| 2743 | } |
| 2744 | if (output->pos > output->size) { /* forbidden */ |
| 2745 | DEBUGLOG(5, "out: pos: %u vs size: %u", |
| 2746 | (U32)output->pos, (U32)output->size); |
| 2747 | return ERROR(dstSize_tooSmall); |
| 2748 | } |
| 2749 | DEBUGLOG(5, "input size : %u", (U32)(input->size - input->pos)); |
| 2750 | |
| 2751 | while (someMoreWork) { |
| 2752 | switch(zds->streamStage) |
| 2753 | { |
| 2754 | case zdss_init : |
| 2755 | DEBUGLOG(5, "stage zdss_init => transparent reset "); |
| 2756 | ZSTD_resetDStream(zds); /* transparent reset on starting decoding a new frame */ |
| 2757 | /* fall-through */ |
| 2758 | |
| 2759 | case zdss_loadHeader : |
| 2760 | DEBUGLOG(5, "stage zdss_loadHeader (srcSize : %u)", (U32)(iend - ip)); |
| 2761 | #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT>=1) |
| 2762 | if (zds->legacyVersion) { |
| 2763 | /* legacy support is incompatible with static dctx */ |
| 2764 | if (zds->staticSize) return ERROR(memory_allocation); |
| 2765 | { size_t const hint = ZSTD_decompressLegacyStream(zds->legacyContext, zds->legacyVersion, output, input); |
| 2766 | if (hint==0) zds->streamStage = zdss_init; |
| 2767 | return hint; |
| 2768 | } } |
| 2769 | #endif |
| 2770 | { size_t const hSize = ZSTD_getFrameHeader_internal(&zds->fParams, zds->headerBuffer, zds->lhSize, zds->format); |
| 2771 | DEBUGLOG(5, "header size : %u", (U32)hSize); |
| 2772 | if (ZSTD_isError(hSize)) { |
| 2773 | #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT>=1) |
| 2774 | U32 const legacyVersion = ZSTD_isLegacy(istart, iend-istart); |
| 2775 | if (legacyVersion) { |
| 2776 | const void* const dict = zds->ddict ? zds->ddict->dictContent : NULL; |
| 2777 | size_t const dictSize = zds->ddict ? zds->ddict->dictSize : 0; |
| 2778 | DEBUGLOG(5, "ZSTD_decompressStream: detected legacy version v0.%u", legacyVersion); |
| 2779 | /* legacy support is incompatible with static dctx */ |
| 2780 | if (zds->staticSize) return ERROR(memory_allocation); |
| 2781 | CHECK_F(ZSTD_initLegacyStream(&zds->legacyContext, |
| 2782 | zds->previousLegacyVersion, legacyVersion, |
| 2783 | dict, dictSize)); |
| 2784 | zds->legacyVersion = zds->previousLegacyVersion = legacyVersion; |
| 2785 | { size_t const hint = ZSTD_decompressLegacyStream(zds->legacyContext, legacyVersion, output, input); |
| 2786 | if (hint==0) zds->streamStage = zdss_init; /* or stay in stage zdss_loadHeader */ |
| 2787 | return hint; |
| 2788 | } } |
| 2789 | #endif |
| 2790 | return hSize; /* error */ |
| 2791 | } |
| 2792 | if (hSize != 0) { /* need more input */ |
| 2793 | size_t const toLoad = hSize - zds->lhSize; /* if hSize!=0, hSize > zds->lhSize */ |
| 2794 | size_t const remainingInput = (size_t)(iend-ip); |
| 2795 | assert(iend >= ip); |
| 2796 | if (toLoad > remainingInput) { /* not enough input to load full header */ |
| 2797 | if (remainingInput > 0) { |
| 2798 | memcpy(zds->headerBuffer + zds->lhSize, ip, remainingInput); |
| 2799 | zds->lhSize += remainingInput; |
| 2800 | } |
| 2801 | input->pos = input->size; |
| 2802 | return (MAX(ZSTD_frameHeaderSize_min, hSize) - zds->lhSize) + ZSTD_blockHeaderSize; /* remaining header bytes + next block header */ |
| 2803 | } |
| 2804 | assert(ip != NULL); |
| 2805 | memcpy(zds->headerBuffer + zds->lhSize, ip, toLoad); zds->lhSize = hSize; ip += toLoad; |
| 2806 | break; |
| 2807 | } } |
| 2808 | |
| 2809 | /* check for single-pass mode opportunity */ |
| 2810 | if (zds->fParams.frameContentSize && zds->fParams.windowSize /* skippable frame if == 0 */ |
| 2811 | && (U64)(size_t)(oend-op) >= zds->fParams.frameContentSize) { |
| 2812 | size_t const cSize = ZSTD_findFrameCompressedSize(istart, iend-istart); |
| 2813 | if (cSize <= (size_t)(iend-istart)) { |
| 2814 | /* shortcut : using single-pass mode */ |
| 2815 | size_t const decompressedSize = ZSTD_decompress_usingDDict(zds, op, oend-op, istart, cSize, zds->ddict); |
| 2816 | if (ZSTD_isError(decompressedSize)) return decompressedSize; |
| 2817 | DEBUGLOG(4, "shortcut to single-pass ZSTD_decompress_usingDDict()") |
| 2818 | ip = istart + cSize; |
| 2819 | op += decompressedSize; |
| 2820 | zds->expected = 0; |
| 2821 | zds->streamStage = zdss_init; |
| 2822 | someMoreWork = 0; |
| 2823 | break; |
| 2824 | } } |
| 2825 | |
| 2826 | /* Consume header (see ZSTDds_decodeFrameHeader) */ |
| 2827 | DEBUGLOG(4, "Consume header"); |
| 2828 | CHECK_F(ZSTD_decompressBegin_usingDDict(zds, zds->ddict)); |
| 2829 | |
| 2830 | if ((MEM_readLE32(zds->headerBuffer) & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START) { /* skippable frame */ |
| 2831 | zds->expected = MEM_readLE32(zds->headerBuffer + ZSTD_frameIdSize); |
| 2832 | zds->stage = ZSTDds_skipFrame; |
| 2833 | } else { |
| 2834 | CHECK_F(ZSTD_decodeFrameHeader(zds, zds->headerBuffer, zds->lhSize)); |
| 2835 | zds->expected = ZSTD_blockHeaderSize; |
| 2836 | zds->stage = ZSTDds_decodeBlockHeader; |
| 2837 | } |
| 2838 | |
| 2839 | /* control buffer memory usage */ |
| 2840 | DEBUGLOG(4, "Control max memory usage (%u KB <= max %u KB)", |
| 2841 | (U32)(zds->fParams.windowSize >>10), |
| 2842 | (U32)(zds->maxWindowSize >> 10) ); |
| 2843 | zds->fParams.windowSize = MAX(zds->fParams.windowSize, 1U << ZSTD_WINDOWLOG_ABSOLUTEMIN); |
| 2844 | if (zds->fParams.windowSize > zds->maxWindowSize) return ERROR(frameParameter_windowTooLarge); |
| 2845 | |
| 2846 | /* Adapt buffer sizes to frame header instructions */ |
| 2847 | { size_t const neededInBuffSize = MAX(zds->fParams.blockSizeMax, 4 /* frame checksum */); |
| 2848 | size_t const neededOutBuffSize = ZSTD_decodingBufferSize_min(zds->fParams.windowSize, zds->fParams.frameContentSize); |
| 2849 | if ((zds->inBuffSize < neededInBuffSize) || (zds->outBuffSize < neededOutBuffSize)) { |
| 2850 | size_t const bufferSize = neededInBuffSize + neededOutBuffSize; |
| 2851 | DEBUGLOG(4, "inBuff : from %u to %u", |
| 2852 | (U32)zds->inBuffSize, (U32)neededInBuffSize); |
| 2853 | DEBUGLOG(4, "outBuff : from %u to %u", |
| 2854 | (U32)zds->outBuffSize, (U32)neededOutBuffSize); |
| 2855 | if (zds->staticSize) { /* static DCtx */ |
| 2856 | DEBUGLOG(4, "staticSize : %u", (U32)zds->staticSize); |
| 2857 | assert(zds->staticSize >= sizeof(ZSTD_DCtx)); /* controlled at init */ |
| 2858 | if (bufferSize > zds->staticSize - sizeof(ZSTD_DCtx)) |
| 2859 | return ERROR(memory_allocation); |
| 2860 | } else { |
| 2861 | ZSTD_free(zds->inBuff, zds->customMem); |
| 2862 | zds->inBuffSize = 0; |
| 2863 | zds->outBuffSize = 0; |
| 2864 | zds->inBuff = (char*)ZSTD_malloc(bufferSize, zds->customMem); |
| 2865 | if (zds->inBuff == NULL) return ERROR(memory_allocation); |
| 2866 | } |
| 2867 | zds->inBuffSize = neededInBuffSize; |
| 2868 | zds->outBuff = zds->inBuff + zds->inBuffSize; |
| 2869 | zds->outBuffSize = neededOutBuffSize; |
| 2870 | } } |
| 2871 | zds->streamStage = zdss_read; |
| 2872 | /* fall-through */ |
| 2873 | |
| 2874 | case zdss_read: |
| 2875 | DEBUGLOG(5, "stage zdss_read"); |
| 2876 | { size_t const neededInSize = ZSTD_nextSrcSizeToDecompress(zds); |
| 2877 | DEBUGLOG(5, "neededInSize = %u", (U32)neededInSize); |
| 2878 | if (neededInSize==0) { /* end of frame */ |
| 2879 | zds->streamStage = zdss_init; |
| 2880 | someMoreWork = 0; |
| 2881 | break; |
| 2882 | } |
| 2883 | if ((size_t)(iend-ip) >= neededInSize) { /* decode directly from src */ |
| 2884 | int const isSkipFrame = ZSTD_isSkipFrame(zds); |
| 2885 | size_t const decodedSize = ZSTD_decompressContinue(zds, |
| 2886 | zds->outBuff + zds->outStart, (isSkipFrame ? 0 : zds->outBuffSize - zds->outStart), |
| 2887 | ip, neededInSize); |
| 2888 | if (ZSTD_isError(decodedSize)) return decodedSize; |
| 2889 | ip += neededInSize; |
| 2890 | if (!decodedSize && !isSkipFrame) break; /* this was just a header */ |
| 2891 | zds->outEnd = zds->outStart + decodedSize; |
| 2892 | zds->streamStage = zdss_flush; |
| 2893 | break; |
| 2894 | } } |
| 2895 | if (ip==iend) { someMoreWork = 0; break; } /* no more input */ |
| 2896 | zds->streamStage = zdss_load; |
| 2897 | /* fall-through */ |
| 2898 | |
| 2899 | case zdss_load: |
| 2900 | { size_t const neededInSize = ZSTD_nextSrcSizeToDecompress(zds); |
| 2901 | size_t const toLoad = neededInSize - zds->inPos; |
| 2902 | int const isSkipFrame = ZSTD_isSkipFrame(zds); |
| 2903 | size_t loadedSize; |
| 2904 | if (isSkipFrame) { |
| 2905 | loadedSize = MIN(toLoad, (size_t)(iend-ip)); |
| 2906 | } else { |
| 2907 | if (toLoad > zds->inBuffSize - zds->inPos) return ERROR(corruption_detected); /* should never happen */ |
| 2908 | loadedSize = ZSTD_limitCopy(zds->inBuff + zds->inPos, toLoad, ip, iend-ip); |
| 2909 | } |
| 2910 | ip += loadedSize; |
| 2911 | zds->inPos += loadedSize; |
| 2912 | if (loadedSize < toLoad) { someMoreWork = 0; break; } /* not enough input, wait for more */ |
| 2913 | |
| 2914 | /* decode loaded input */ |
| 2915 | { size_t const decodedSize = ZSTD_decompressContinue(zds, |
| 2916 | zds->outBuff + zds->outStart, zds->outBuffSize - zds->outStart, |
| 2917 | zds->inBuff, neededInSize); |
| 2918 | if (ZSTD_isError(decodedSize)) return decodedSize; |
| 2919 | zds->inPos = 0; /* input is consumed */ |
| 2920 | if (!decodedSize && !isSkipFrame) { zds->streamStage = zdss_read; break; } /* this was just a header */ |
| 2921 | zds->outEnd = zds->outStart + decodedSize; |
| 2922 | } } |
| 2923 | zds->streamStage = zdss_flush; |
| 2924 | /* fall-through */ |
| 2925 | |
| 2926 | case zdss_flush: |
| 2927 | { size_t const toFlushSize = zds->outEnd - zds->outStart; |
| 2928 | size_t const flushedSize = ZSTD_limitCopy(op, oend-op, zds->outBuff + zds->outStart, toFlushSize); |
| 2929 | op += flushedSize; |
| 2930 | zds->outStart += flushedSize; |
| 2931 | if (flushedSize == toFlushSize) { /* flush completed */ |
| 2932 | zds->streamStage = zdss_read; |
| 2933 | if ( (zds->outBuffSize < zds->fParams.frameContentSize) |
| 2934 | && (zds->outStart + zds->fParams.blockSizeMax > zds->outBuffSize) ) { |
| 2935 | DEBUGLOG(5, "restart filling outBuff from beginning (left:%i, needed:%u)", |
| 2936 | (int)(zds->outBuffSize - zds->outStart), |
| 2937 | (U32)zds->fParams.blockSizeMax); |
| 2938 | zds->outStart = zds->outEnd = 0; |
| 2939 | } |
| 2940 | break; |
| 2941 | } } |
| 2942 | /* cannot complete flush */ |
| 2943 | someMoreWork = 0; |
| 2944 | break; |
| 2945 | |
| 2946 | default: return ERROR(GENERIC); /* impossible */ |
| 2947 | } } |
| 2948 | |
| 2949 | /* result */ |
| 2950 | input->pos += (size_t)(ip-istart); |
| 2951 | output->pos += (size_t)(op-ostart); |
| 2952 | { size_t nextSrcSizeHint = ZSTD_nextSrcSizeToDecompress(zds); |
| 2953 | if (!nextSrcSizeHint) { /* frame fully decoded */ |
| 2954 | if (zds->outEnd == zds->outStart) { /* output fully flushed */ |
| 2955 | if (zds->hostageByte) { |
| 2956 | if (input->pos >= input->size) { |
| 2957 | /* can't release hostage (not present) */ |
| 2958 | zds->streamStage = zdss_read; |
| 2959 | return 1; |
| 2960 | } |
| 2961 | input->pos++; /* release hostage */ |
| 2962 | } /* zds->hostageByte */ |
| 2963 | return 0; |
| 2964 | } /* zds->outEnd == zds->outStart */ |
| 2965 | if (!zds->hostageByte) { /* output not fully flushed; keep last byte as hostage; will be released when all output is flushed */ |
| 2966 | input->pos--; /* note : pos > 0, otherwise, impossible to finish reading last block */ |
| 2967 | zds->hostageByte=1; |
| 2968 | } |
| 2969 | return 1; |
| 2970 | } /* nextSrcSizeHint==0 */ |
| 2971 | nextSrcSizeHint += ZSTD_blockHeaderSize * (ZSTD_nextInputType(zds) == ZSTDnit_block); /* preload header of next block */ |
| 2972 | assert(zds->inPos <= nextSrcSizeHint); |
| 2973 | nextSrcSizeHint -= zds->inPos; /* part already loaded*/ |
| 2974 | return nextSrcSizeHint; |
| 2975 | } |
| 2976 | } |
| 2977 | |
| 2978 | |
| 2979 | size_t ZSTD_decompress_generic(ZSTD_DCtx* dctx, ZSTD_outBuffer* output, ZSTD_inBuffer* input) |
| 2980 | { |
| 2981 | return ZSTD_decompressStream(dctx, output, input); |
| 2982 | } |
| 2983 | |
| 2984 | size_t ZSTD_decompress_generic_simpleArgs ( |
| 2985 | ZSTD_DCtx* dctx, |
| 2986 | void* dst, size_t dstCapacity, size_t* dstPos, |
| 2987 | const void* src, size_t srcSize, size_t* srcPos) |
| 2988 | { |
| 2989 | ZSTD_outBuffer output = { dst, dstCapacity, *dstPos }; |
| 2990 | ZSTD_inBuffer input = { src, srcSize, *srcPos }; |
| 2991 | /* ZSTD_compress_generic() will check validity of dstPos and srcPos */ |
| 2992 | size_t const cErr = ZSTD_decompress_generic(dctx, &output, &input); |
| 2993 | *dstPos = output.pos; |
| 2994 | *srcPos = input.pos; |
| 2995 | return cErr; |
| 2996 | } |
| 2997 | |
| 2998 | void ZSTD_DCtx_reset(ZSTD_DCtx* dctx) |
| 2999 | { |
| 3000 | (void)ZSTD_initDStream(dctx); |
| 3001 | dctx->format = ZSTD_f_zstd1; |
| 3002 | dctx->maxWindowSize = ZSTD_MAXWINDOWSIZE_DEFAULT; |
| 3003 | } |