set zfs_arc_shrinker_limit to 0 by default
[zfs.git] / module / zstd / lib / decompress / zstd_decompress.c
blobbe5c7cfc3347da0291d839379b8ca15aab809e7c
1 /*
2 * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
3 * All rights reserved.
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 */
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.
21 #ifndef ZSTD_HEAPMODE
22 # define ZSTD_HEAPMODE 1
23 #endif
25 /*!
26 * LEGACY_SUPPORT :
27 * if set to 1+, ZSTD_decompress() can decode older formats (v0.1+)
29 #ifndef ZSTD_LEGACY_SUPPORT
30 # define ZSTD_LEGACY_SUPPORT 0
31 #endif
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().
39 #ifndef ZSTD_MAXWINDOWSIZE_DEFAULT
40 # define ZSTD_MAXWINDOWSIZE_DEFAULT (((U32)1 << ZSTD_WINDOWLOG_LIMIT_DEFAULT) + 1)
41 #endif
43 /*!
44 * NO_FORWARD_PROGRESS_MAX :
45 * maximum allowed nb of calls to ZSTD_decompressStream()
46 * without any forward progress
47 * (defined as: no byte read from input, and no byte flushed to output)
48 * before triggering an error.
50 #ifndef ZSTD_NO_FORWARD_PROGRESS_MAX
51 # define ZSTD_NO_FORWARD_PROGRESS_MAX 16
52 #endif
55 /*-*******************************************************
56 * Dependencies
57 *********************************************************/
58 #include <string.h> /* memcpy, memmove, memset */
59 #include "../common/cpu.h" /* bmi2 */
60 #include "../common/mem.h" /* low level memory routines */
61 #define FSE_STATIC_LINKING_ONLY
62 #include "../common/fse.h"
63 #define HUF_STATIC_LINKING_ONLY
64 #include "../common/huf.h"
65 #include "../common/zstd_internal.h" /* blockProperties_t */
66 #include "zstd_decompress_internal.h" /* ZSTD_DCtx */
67 #include "zstd_ddict.h" /* ZSTD_DDictDictContent */
68 #include "zstd_decompress_block.h" /* ZSTD_decompressBlock_internal */
70 #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT>=1)
71 # include "../legacy/zstd_legacy.h"
72 #endif
75 /*-*************************************************************
76 * Context management
77 ***************************************************************/
78 size_t ZSTD_sizeof_DCtx (const ZSTD_DCtx* dctx)
80 if (dctx==NULL) return 0; /* support sizeof NULL */
81 return sizeof(*dctx)
82 + ZSTD_sizeof_DDict(dctx->ddictLocal)
83 + dctx->inBuffSize + dctx->outBuffSize;
86 size_t ZSTD_estimateDCtxSize(void) { return sizeof(ZSTD_DCtx); }
89 static size_t ZSTD_startingInputLength(ZSTD_format_e format)
91 size_t const startingInputLength = ZSTD_FRAMEHEADERSIZE_PREFIX(format);
92 /* only supports formats ZSTD_f_zstd1 and ZSTD_f_zstd1_magicless */
93 assert( (format == ZSTD_f_zstd1) || (format == ZSTD_f_zstd1_magicless) );
94 return startingInputLength;
97 static void ZSTD_initDCtx_internal(ZSTD_DCtx* dctx)
99 dctx->format = ZSTD_f_zstd1; /* ZSTD_decompressBegin() invokes ZSTD_startingInputLength() with argument dctx->format */
100 dctx->staticSize = 0;
101 dctx->maxWindowSize = ZSTD_MAXWINDOWSIZE_DEFAULT;
102 dctx->ddict = NULL;
103 dctx->ddictLocal = NULL;
104 dctx->dictEnd = NULL;
105 dctx->ddictIsCold = 0;
106 dctx->dictUses = ZSTD_dont_use;
107 dctx->inBuff = NULL;
108 dctx->inBuffSize = 0;
109 dctx->outBuffSize = 0;
110 dctx->streamStage = zdss_init;
111 dctx->legacyContext = NULL;
112 dctx->previousLegacyVersion = 0;
113 dctx->noForwardProgress = 0;
114 dctx->oversizedDuration = 0;
115 dctx->bmi2 = ZSTD_cpuid_bmi2(ZSTD_cpuid());
116 dctx->outBufferMode = ZSTD_obm_buffered;
117 #ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
118 dctx->dictContentEndForFuzzing = NULL;
119 #endif
122 ZSTD_DCtx* ZSTD_initStaticDCtx(void *workspace, size_t workspaceSize)
124 ZSTD_DCtx* const dctx = (ZSTD_DCtx*) workspace;
126 if ((size_t)workspace & 7) return NULL; /* 8-aligned */
127 if (workspaceSize < sizeof(ZSTD_DCtx)) return NULL; /* minimum size */
129 ZSTD_initDCtx_internal(dctx);
130 dctx->staticSize = workspaceSize;
131 dctx->inBuff = (char*)(dctx+1);
132 return dctx;
135 ZSTD_DCtx* ZSTD_createDCtx_advanced(ZSTD_customMem customMem)
137 if (!customMem.customAlloc ^ !customMem.customFree) return NULL;
139 { ZSTD_DCtx* const dctx = (ZSTD_DCtx*)ZSTD_malloc(sizeof(*dctx), customMem);
140 if (!dctx) return NULL;
141 dctx->customMem = customMem;
142 ZSTD_initDCtx_internal(dctx);
143 return dctx;
147 ZSTD_DCtx* ZSTD_createDCtx(void)
149 DEBUGLOG(3, "ZSTD_createDCtx");
150 return ZSTD_createDCtx_advanced(ZSTD_defaultCMem);
153 static void ZSTD_clearDict(ZSTD_DCtx* dctx)
155 ZSTD_freeDDict(dctx->ddictLocal);
156 dctx->ddictLocal = NULL;
157 dctx->ddict = NULL;
158 dctx->dictUses = ZSTD_dont_use;
161 size_t ZSTD_freeDCtx(ZSTD_DCtx* dctx)
163 if (dctx==NULL) return 0; /* support free on NULL */
164 RETURN_ERROR_IF(dctx->staticSize, memory_allocation, "not compatible with static DCtx");
165 { ZSTD_customMem const cMem = dctx->customMem;
166 ZSTD_clearDict(dctx);
167 ZSTD_free(dctx->inBuff, cMem);
168 dctx->inBuff = NULL;
169 #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT >= 1)
170 if (dctx->legacyContext)
171 ZSTD_freeLegacyStreamContext(dctx->legacyContext, dctx->previousLegacyVersion);
172 #endif
173 ZSTD_free(dctx, cMem);
174 return 0;
178 /* no longer useful */
179 void ZSTD_copyDCtx(ZSTD_DCtx* dstDCtx, const ZSTD_DCtx* srcDCtx)
181 size_t const toCopy = (size_t)((char*)(&dstDCtx->inBuff) - (char*)dstDCtx);
182 memcpy(dstDCtx, srcDCtx, toCopy); /* no need to copy workspace */
186 /*-*************************************************************
187 * Frame header decoding
188 ***************************************************************/
190 /*! ZSTD_isFrame() :
191 * Tells if the content of `buffer` starts with a valid Frame Identifier.
192 * Note : Frame Identifier is 4 bytes. If `size < 4`, @return will always be 0.
193 * Note 2 : Legacy Frame Identifiers are considered valid only if Legacy Support is enabled.
194 * Note 3 : Skippable Frame Identifiers are considered valid. */
195 unsigned ZSTD_isFrame(const void* buffer, size_t size)
197 if (size < ZSTD_FRAMEIDSIZE) return 0;
198 { U32 const magic = MEM_readLE32(buffer);
199 if (magic == ZSTD_MAGICNUMBER) return 1;
200 if ((magic & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) return 1;
202 #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT >= 1)
203 if (ZSTD_isLegacy(buffer, size)) return 1;
204 #endif
205 return 0;
208 /** ZSTD_frameHeaderSize_internal() :
209 * srcSize must be large enough to reach header size fields.
210 * note : only works for formats ZSTD_f_zstd1 and ZSTD_f_zstd1_magicless.
211 * @return : size of the Frame Header
212 * or an error code, which can be tested with ZSTD_isError() */
213 static size_t ZSTD_frameHeaderSize_internal(const void* src, size_t srcSize, ZSTD_format_e format)
215 size_t const minInputSize = ZSTD_startingInputLength(format);
216 RETURN_ERROR_IF(srcSize < minInputSize, srcSize_wrong, "");
218 { BYTE const fhd = ((const BYTE*)src)[minInputSize-1];
219 U32 const dictID= fhd & 3;
220 U32 const singleSegment = (fhd >> 5) & 1;
221 U32 const fcsId = fhd >> 6;
222 return minInputSize + !singleSegment
223 + ZSTD_did_fieldSize[dictID] + ZSTD_fcs_fieldSize[fcsId]
224 + (singleSegment && !fcsId);
228 /** ZSTD_frameHeaderSize() :
229 * srcSize must be >= ZSTD_frameHeaderSize_prefix.
230 * @return : size of the Frame Header,
231 * or an error code (if srcSize is too small) */
232 size_t ZSTD_frameHeaderSize(const void* src, size_t srcSize)
234 return ZSTD_frameHeaderSize_internal(src, srcSize, ZSTD_f_zstd1);
238 /** ZSTD_getFrameHeader_advanced() :
239 * decode Frame Header, or require larger `srcSize`.
240 * note : only works for formats ZSTD_f_zstd1 and ZSTD_f_zstd1_magicless
241 * @return : 0, `zfhPtr` is correctly filled,
242 * >0, `srcSize` is too small, value is wanted `srcSize` amount,
243 * or an error code, which can be tested using ZSTD_isError() */
244 size_t ZSTD_getFrameHeader_advanced(ZSTD_frameHeader* zfhPtr, const void* src, size_t srcSize, ZSTD_format_e format)
246 const BYTE* ip = (const BYTE*)src;
247 size_t const minInputSize = ZSTD_startingInputLength(format);
249 memset(zfhPtr, 0, sizeof(*zfhPtr)); /* not strictly necessary, but static analyzer do not understand that zfhPtr is only going to be read only if return value is zero, since they are 2 different signals */
250 if (srcSize < minInputSize) return minInputSize;
251 RETURN_ERROR_IF(src==NULL, GENERIC, "invalid parameter");
253 if ( (format != ZSTD_f_zstd1_magicless)
254 && (MEM_readLE32(src) != ZSTD_MAGICNUMBER) ) {
255 if ((MEM_readLE32(src) & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) {
256 /* skippable frame */
257 if (srcSize < ZSTD_SKIPPABLEHEADERSIZE)
258 return ZSTD_SKIPPABLEHEADERSIZE; /* magic number + frame length */
259 memset(zfhPtr, 0, sizeof(*zfhPtr));
260 zfhPtr->frameContentSize = MEM_readLE32((const char *)src + ZSTD_FRAMEIDSIZE);
261 zfhPtr->frameType = ZSTD_skippableFrame;
262 return 0;
264 RETURN_ERROR(prefix_unknown, "");
267 /* ensure there is enough `srcSize` to fully read/decode frame header */
268 { size_t const fhsize = ZSTD_frameHeaderSize_internal(src, srcSize, format);
269 if (srcSize < fhsize) return fhsize;
270 zfhPtr->headerSize = (U32)fhsize;
273 { BYTE const fhdByte = ip[minInputSize-1];
274 size_t pos = minInputSize;
275 U32 const dictIDSizeCode = fhdByte&3;
276 U32 const checksumFlag = (fhdByte>>2)&1;
277 U32 const singleSegment = (fhdByte>>5)&1;
278 U32 const fcsID = fhdByte>>6;
279 U64 windowSize = 0;
280 U32 dictID = 0;
281 U64 frameContentSize = ZSTD_CONTENTSIZE_UNKNOWN;
282 RETURN_ERROR_IF((fhdByte & 0x08) != 0, frameParameter_unsupported,
283 "reserved bits, must be zero");
285 if (!singleSegment) {
286 BYTE const wlByte = ip[pos++];
287 U32 const windowLog = (wlByte >> 3) + ZSTD_WINDOWLOG_ABSOLUTEMIN;
288 RETURN_ERROR_IF(windowLog > ZSTD_WINDOWLOG_MAX, frameParameter_windowTooLarge, "");
289 windowSize = (1ULL << windowLog);
290 windowSize += (windowSize >> 3) * (wlByte&7);
292 switch(dictIDSizeCode)
294 default: assert(0); /* impossible */
295 case 0 : break;
296 case 1 : dictID = ip[pos]; pos++; break;
297 case 2 : dictID = MEM_readLE16(ip+pos); pos+=2; break;
298 case 3 : dictID = MEM_readLE32(ip+pos); pos+=4; break;
300 switch(fcsID)
302 default: assert(0); /* impossible */
303 case 0 : if (singleSegment) frameContentSize = ip[pos]; break;
304 case 1 : frameContentSize = MEM_readLE16(ip+pos)+256; break;
305 case 2 : frameContentSize = MEM_readLE32(ip+pos); break;
306 case 3 : frameContentSize = MEM_readLE64(ip+pos); break;
308 if (singleSegment) windowSize = frameContentSize;
310 zfhPtr->frameType = ZSTD_frame;
311 zfhPtr->frameContentSize = frameContentSize;
312 zfhPtr->windowSize = windowSize;
313 zfhPtr->blockSizeMax = (unsigned) MIN(windowSize, ZSTD_BLOCKSIZE_MAX);
314 zfhPtr->dictID = dictID;
315 zfhPtr->checksumFlag = checksumFlag;
317 return 0;
320 /** ZSTD_getFrameHeader() :
321 * decode Frame Header, or require larger `srcSize`.
322 * note : this function does not consume input, it only reads it.
323 * @return : 0, `zfhPtr` is correctly filled,
324 * >0, `srcSize` is too small, value is wanted `srcSize` amount,
325 * or an error code, which can be tested using ZSTD_isError() */
326 size_t ZSTD_getFrameHeader(ZSTD_frameHeader* zfhPtr, const void* src, size_t srcSize)
328 return ZSTD_getFrameHeader_advanced(zfhPtr, src, srcSize, ZSTD_f_zstd1);
332 /** ZSTD_getFrameContentSize() :
333 * compatible with legacy mode
334 * @return : decompressed size of the single frame pointed to be `src` if known, otherwise
335 * - ZSTD_CONTENTSIZE_UNKNOWN if the size cannot be determined
336 * - ZSTD_CONTENTSIZE_ERROR if an error occurred (e.g. invalid magic number, srcSize too small) */
337 unsigned long long ZSTD_getFrameContentSize(const void *src, size_t srcSize)
339 #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT >= 1)
340 if (ZSTD_isLegacy(src, srcSize)) {
341 unsigned long long const ret = ZSTD_getDecompressedSize_legacy(src, srcSize);
342 return ret == 0 ? ZSTD_CONTENTSIZE_UNKNOWN : ret;
344 #endif
345 { ZSTD_frameHeader zfh;
346 if (ZSTD_getFrameHeader(&zfh, src, srcSize) != 0)
347 return ZSTD_CONTENTSIZE_ERROR;
348 if (zfh.frameType == ZSTD_skippableFrame) {
349 return 0;
350 } else {
351 return zfh.frameContentSize;
355 static size_t readSkippableFrameSize(void const* src, size_t srcSize)
357 size_t const skippableHeaderSize = ZSTD_SKIPPABLEHEADERSIZE;
358 U32 sizeU32;
360 RETURN_ERROR_IF(srcSize < ZSTD_SKIPPABLEHEADERSIZE, srcSize_wrong, "");
362 sizeU32 = MEM_readLE32((BYTE const*)src + ZSTD_FRAMEIDSIZE);
363 RETURN_ERROR_IF((U32)(sizeU32 + ZSTD_SKIPPABLEHEADERSIZE) < sizeU32,
364 frameParameter_unsupported, "");
366 size_t const skippableSize = skippableHeaderSize + sizeU32;
367 RETURN_ERROR_IF(skippableSize > srcSize, srcSize_wrong, "");
368 return skippableSize;
372 /** ZSTD_findDecompressedSize() :
373 * compatible with legacy mode
374 * `srcSize` must be the exact length of some number of ZSTD compressed and/or
375 * skippable frames
376 * @return : decompressed size of the frames contained */
377 unsigned long long ZSTD_findDecompressedSize(const void* src, size_t srcSize)
379 unsigned long long totalDstSize = 0;
381 while (srcSize >= ZSTD_startingInputLength(ZSTD_f_zstd1)) {
382 U32 const magicNumber = MEM_readLE32(src);
384 if ((magicNumber & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) {
385 size_t const skippableSize = readSkippableFrameSize(src, srcSize);
386 if (ZSTD_isError(skippableSize)) {
387 return ZSTD_CONTENTSIZE_ERROR;
389 assert(skippableSize <= srcSize);
391 src = (const BYTE *)src + skippableSize;
392 srcSize -= skippableSize;
393 continue;
396 { unsigned long long const ret = ZSTD_getFrameContentSize(src, srcSize);
397 if (ret >= ZSTD_CONTENTSIZE_ERROR) return ret;
399 /* check for overflow */
400 if (totalDstSize + ret < totalDstSize) return ZSTD_CONTENTSIZE_ERROR;
401 totalDstSize += ret;
403 { size_t const frameSrcSize = ZSTD_findFrameCompressedSize(src, srcSize);
404 if (ZSTD_isError(frameSrcSize)) {
405 return ZSTD_CONTENTSIZE_ERROR;
408 src = (const BYTE *)src + frameSrcSize;
409 srcSize -= frameSrcSize;
411 } /* while (srcSize >= ZSTD_frameHeaderSize_prefix) */
413 if (srcSize) return ZSTD_CONTENTSIZE_ERROR;
415 return totalDstSize;
418 /** ZSTD_getDecompressedSize() :
419 * compatible with legacy mode
420 * @return : decompressed size if known, 0 otherwise
421 note : 0 can mean any of the following :
422 - frame content is empty
423 - decompressed size field is not present in frame header
424 - frame header unknown / not supported
425 - frame header not complete (`srcSize` too small) */
426 unsigned long long ZSTD_getDecompressedSize(const void* src, size_t srcSize)
428 unsigned long long const ret = ZSTD_getFrameContentSize(src, srcSize);
429 ZSTD_STATIC_ASSERT(ZSTD_CONTENTSIZE_ERROR < ZSTD_CONTENTSIZE_UNKNOWN);
430 return (ret >= ZSTD_CONTENTSIZE_ERROR) ? 0 : ret;
434 /** ZSTD_decodeFrameHeader() :
435 * `headerSize` must be the size provided by ZSTD_frameHeaderSize().
436 * @return : 0 if success, or an error code, which can be tested using ZSTD_isError() */
437 static size_t ZSTD_decodeFrameHeader(ZSTD_DCtx* dctx, const void* src, size_t headerSize)
439 size_t const result = ZSTD_getFrameHeader_advanced(&(dctx->fParams), src, headerSize, dctx->format);
440 if (ZSTD_isError(result)) return result; /* invalid header */
441 RETURN_ERROR_IF(result>0, srcSize_wrong, "headerSize too small");
442 #ifndef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
443 /* Skip the dictID check in fuzzing mode, because it makes the search
444 * harder.
446 RETURN_ERROR_IF(dctx->fParams.dictID && (dctx->dictID != dctx->fParams.dictID),
447 dictionary_wrong, "");
448 #endif
449 if (dctx->fParams.checksumFlag) XXH64_reset(&dctx->xxhState, 0);
450 return 0;
453 static ZSTD_frameSizeInfo ZSTD_errorFrameSizeInfo(size_t ret)
455 ZSTD_frameSizeInfo frameSizeInfo;
456 frameSizeInfo.compressedSize = ret;
457 frameSizeInfo.decompressedBound = ZSTD_CONTENTSIZE_ERROR;
458 return frameSizeInfo;
461 static ZSTD_frameSizeInfo ZSTD_findFrameSizeInfo(const void* src, size_t srcSize)
463 ZSTD_frameSizeInfo frameSizeInfo;
464 memset(&frameSizeInfo, 0, sizeof(ZSTD_frameSizeInfo));
466 #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT >= 1)
467 if (ZSTD_isLegacy(src, srcSize))
468 return ZSTD_findFrameSizeInfoLegacy(src, srcSize);
469 #endif
471 if ((srcSize >= ZSTD_SKIPPABLEHEADERSIZE)
472 && (MEM_readLE32(src) & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) {
473 frameSizeInfo.compressedSize = readSkippableFrameSize(src, srcSize);
474 assert(ZSTD_isError(frameSizeInfo.compressedSize) ||
475 frameSizeInfo.compressedSize <= srcSize);
476 return frameSizeInfo;
477 } else {
478 const BYTE* ip = (const BYTE*)src;
479 const BYTE* const ipstart = ip;
480 size_t remainingSize = srcSize;
481 size_t nbBlocks = 0;
482 ZSTD_frameHeader zfh;
484 /* Extract Frame Header */
485 { size_t const ret = ZSTD_getFrameHeader(&zfh, src, srcSize);
486 if (ZSTD_isError(ret))
487 return ZSTD_errorFrameSizeInfo(ret);
488 if (ret > 0)
489 return ZSTD_errorFrameSizeInfo(ERROR(srcSize_wrong));
492 ip += zfh.headerSize;
493 remainingSize -= zfh.headerSize;
495 /* Iterate over each block */
496 while (1) {
497 blockProperties_t blockProperties;
498 size_t const cBlockSize = ZSTD_getcBlockSize(ip, remainingSize, &blockProperties);
499 if (ZSTD_isError(cBlockSize))
500 return ZSTD_errorFrameSizeInfo(cBlockSize);
502 if (ZSTD_blockHeaderSize + cBlockSize > remainingSize)
503 return ZSTD_errorFrameSizeInfo(ERROR(srcSize_wrong));
505 ip += ZSTD_blockHeaderSize + cBlockSize;
506 remainingSize -= ZSTD_blockHeaderSize + cBlockSize;
507 nbBlocks++;
509 if (blockProperties.lastBlock) break;
512 /* Final frame content checksum */
513 if (zfh.checksumFlag) {
514 if (remainingSize < 4)
515 return ZSTD_errorFrameSizeInfo(ERROR(srcSize_wrong));
516 ip += 4;
519 frameSizeInfo.compressedSize = ip - ipstart;
520 frameSizeInfo.decompressedBound = (zfh.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN)
521 ? zfh.frameContentSize
522 : nbBlocks * zfh.blockSizeMax;
523 return frameSizeInfo;
527 /** ZSTD_findFrameCompressedSize() :
528 * compatible with legacy mode
529 * `src` must point to the start of a ZSTD frame, ZSTD legacy frame, or skippable frame
530 * `srcSize` must be at least as large as the frame contained
531 * @return : the compressed size of the frame starting at `src` */
532 size_t ZSTD_findFrameCompressedSize(const void *src, size_t srcSize)
534 ZSTD_frameSizeInfo const frameSizeInfo = ZSTD_findFrameSizeInfo(src, srcSize);
535 return frameSizeInfo.compressedSize;
538 /** ZSTD_decompressBound() :
539 * compatible with legacy mode
540 * `src` must point to the start of a ZSTD frame or a skippeable frame
541 * `srcSize` must be at least as large as the frame contained
542 * @return : the maximum decompressed size of the compressed source
544 unsigned long long ZSTD_decompressBound(const void* src, size_t srcSize)
546 unsigned long long bound = 0;
547 /* Iterate over each frame */
548 while (srcSize > 0) {
549 ZSTD_frameSizeInfo const frameSizeInfo = ZSTD_findFrameSizeInfo(src, srcSize);
550 size_t const compressedSize = frameSizeInfo.compressedSize;
551 unsigned long long const decompressedBound = frameSizeInfo.decompressedBound;
552 if (ZSTD_isError(compressedSize) || decompressedBound == ZSTD_CONTENTSIZE_ERROR)
553 return ZSTD_CONTENTSIZE_ERROR;
554 assert(srcSize >= compressedSize);
555 src = (const BYTE*)src + compressedSize;
556 srcSize -= compressedSize;
557 bound += decompressedBound;
559 return bound;
563 /*-*************************************************************
564 * Frame decoding
565 ***************************************************************/
567 /** ZSTD_insertBlock() :
568 * insert `src` block into `dctx` history. Useful to track uncompressed blocks. */
569 size_t ZSTD_insertBlock(ZSTD_DCtx* dctx, const void* blockStart, size_t blockSize)
571 DEBUGLOG(5, "ZSTD_insertBlock: %u bytes", (unsigned)blockSize);
572 ZSTD_checkContinuity(dctx, blockStart);
573 dctx->previousDstEnd = (const char*)blockStart + blockSize;
574 return blockSize;
578 static size_t ZSTD_copyRawBlock(void* dst, size_t dstCapacity,
579 const void* src, size_t srcSize)
581 DEBUGLOG(5, "ZSTD_copyRawBlock");
582 if (dst == NULL) {
583 if (srcSize == 0) return 0;
584 RETURN_ERROR(dstBuffer_null, "");
586 RETURN_ERROR_IF(srcSize > dstCapacity, dstSize_tooSmall, "");
587 memcpy(dst, src, srcSize);
588 return srcSize;
591 static size_t ZSTD_setRleBlock(void* dst, size_t dstCapacity,
592 BYTE b,
593 size_t regenSize)
595 if (dst == NULL) {
596 if (regenSize == 0) return 0;
597 RETURN_ERROR(dstBuffer_null, "");
599 RETURN_ERROR_IF(regenSize > dstCapacity, dstSize_tooSmall, "");
600 memset(dst, b, regenSize);
601 return regenSize;
605 /*! ZSTD_decompressFrame() :
606 * @dctx must be properly initialized
607 * will update *srcPtr and *srcSizePtr,
608 * to make *srcPtr progress by one frame. */
609 static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx,
610 void* dst, size_t dstCapacity,
611 const void** srcPtr, size_t *srcSizePtr)
613 const BYTE* ip = (const BYTE*)(*srcPtr);
614 BYTE* const ostart = (BYTE* const)dst;
615 BYTE* const oend = dstCapacity != 0 ? ostart + dstCapacity : ostart;
616 BYTE* op = ostart;
617 size_t remainingSrcSize = *srcSizePtr;
619 DEBUGLOG(4, "ZSTD_decompressFrame (srcSize:%i)", (int)*srcSizePtr);
621 /* check */
622 RETURN_ERROR_IF(
623 remainingSrcSize < ZSTD_FRAMEHEADERSIZE_MIN(dctx->format)+ZSTD_blockHeaderSize,
624 srcSize_wrong, "");
626 /* Frame Header */
627 { size_t const frameHeaderSize = ZSTD_frameHeaderSize_internal(
628 ip, ZSTD_FRAMEHEADERSIZE_PREFIX(dctx->format), dctx->format);
629 if (ZSTD_isError(frameHeaderSize)) return frameHeaderSize;
630 RETURN_ERROR_IF(remainingSrcSize < frameHeaderSize+ZSTD_blockHeaderSize,
631 srcSize_wrong, "");
632 FORWARD_IF_ERROR( ZSTD_decodeFrameHeader(dctx, ip, frameHeaderSize) , "");
633 ip += frameHeaderSize; remainingSrcSize -= frameHeaderSize;
636 /* Loop on each block */
637 while (1) {
638 size_t decodedSize;
639 blockProperties_t blockProperties;
640 size_t const cBlockSize = ZSTD_getcBlockSize(ip, remainingSrcSize, &blockProperties);
641 if (ZSTD_isError(cBlockSize)) return cBlockSize;
643 ip += ZSTD_blockHeaderSize;
644 remainingSrcSize -= ZSTD_blockHeaderSize;
645 RETURN_ERROR_IF(cBlockSize > remainingSrcSize, srcSize_wrong, "");
647 switch(blockProperties.blockType)
649 case bt_compressed:
650 decodedSize = ZSTD_decompressBlock_internal(dctx, op, oend-op, ip, cBlockSize, /* frame */ 1);
651 break;
652 case bt_raw :
653 decodedSize = ZSTD_copyRawBlock(op, oend-op, ip, cBlockSize);
654 break;
655 case bt_rle :
656 decodedSize = ZSTD_setRleBlock(op, oend-op, *ip, blockProperties.origSize);
657 break;
658 case bt_reserved :
659 default:
660 RETURN_ERROR(corruption_detected, "invalid block type");
663 if (ZSTD_isError(decodedSize)) return decodedSize;
664 if (dctx->fParams.checksumFlag)
665 XXH64_update(&dctx->xxhState, op, decodedSize);
666 if (decodedSize != 0)
667 op += decodedSize;
668 assert(ip != NULL);
669 ip += cBlockSize;
670 remainingSrcSize -= cBlockSize;
671 if (blockProperties.lastBlock) break;
674 if (dctx->fParams.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN) {
675 RETURN_ERROR_IF((U64)(op-ostart) != dctx->fParams.frameContentSize,
676 corruption_detected, "");
678 if (dctx->fParams.checksumFlag) { /* Frame content checksum verification */
679 U32 const checkCalc = (U32)XXH64_digest(&dctx->xxhState);
680 U32 checkRead;
681 RETURN_ERROR_IF(remainingSrcSize<4, checksum_wrong, "");
682 checkRead = MEM_readLE32(ip);
683 RETURN_ERROR_IF(checkRead != checkCalc, checksum_wrong, "");
684 ip += 4;
685 remainingSrcSize -= 4;
688 /* Allow caller to get size read */
689 *srcPtr = ip;
690 *srcSizePtr = remainingSrcSize;
691 return op-ostart;
694 static size_t ZSTD_decompressMultiFrame(ZSTD_DCtx* dctx,
695 void* dst, size_t dstCapacity,
696 const void* src, size_t srcSize,
697 const void* dict, size_t dictSize,
698 const ZSTD_DDict* ddict)
700 void* const dststart = dst;
701 int moreThan1Frame = 0;
703 DEBUGLOG(5, "ZSTD_decompressMultiFrame");
704 assert(dict==NULL || ddict==NULL); /* either dict or ddict set, not both */
706 if (ddict) {
707 dict = ZSTD_DDict_dictContent(ddict);
708 dictSize = ZSTD_DDict_dictSize(ddict);
711 while (srcSize >= ZSTD_startingInputLength(dctx->format)) {
713 #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT >= 1)
714 if (ZSTD_isLegacy(src, srcSize)) {
715 size_t decodedSize;
716 size_t const frameSize = ZSTD_findFrameCompressedSizeLegacy(src, srcSize);
717 if (ZSTD_isError(frameSize)) return frameSize;
718 RETURN_ERROR_IF(dctx->staticSize, memory_allocation,
719 "legacy support is not compatible with static dctx");
721 decodedSize = ZSTD_decompressLegacy(dst, dstCapacity, src, frameSize, dict, dictSize);
722 if (ZSTD_isError(decodedSize)) return decodedSize;
724 assert(decodedSize <=- dstCapacity);
725 dst = (BYTE*)dst + decodedSize;
726 dstCapacity -= decodedSize;
728 src = (const BYTE*)src + frameSize;
729 srcSize -= frameSize;
731 continue;
733 #endif
735 { U32 const magicNumber = MEM_readLE32(src);
736 DEBUGLOG(4, "reading magic number %08X (expecting %08X)",
737 (unsigned)magicNumber, ZSTD_MAGICNUMBER);
738 if ((magicNumber & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) {
739 size_t const skippableSize = readSkippableFrameSize(src, srcSize);
740 FORWARD_IF_ERROR(skippableSize, "readSkippableFrameSize failed");
741 assert(skippableSize <= srcSize);
743 src = (const BYTE *)src + skippableSize;
744 srcSize -= skippableSize;
745 continue;
748 if (ddict) {
749 /* we were called from ZSTD_decompress_usingDDict */
750 FORWARD_IF_ERROR(ZSTD_decompressBegin_usingDDict(dctx, ddict), "");
751 } else {
752 /* this will initialize correctly with no dict if dict == NULL, so
753 * use this in all cases but ddict */
754 FORWARD_IF_ERROR(ZSTD_decompressBegin_usingDict(dctx, dict, dictSize), "");
756 ZSTD_checkContinuity(dctx, dst);
758 { const size_t res = ZSTD_decompressFrame(dctx, dst, dstCapacity,
759 &src, &srcSize);
760 RETURN_ERROR_IF(
761 (ZSTD_getErrorCode(res) == ZSTD_error_prefix_unknown)
762 && (moreThan1Frame==1),
763 srcSize_wrong,
764 "at least one frame successfully completed, but following "
765 "bytes are garbage: it's more likely to be a srcSize error, "
766 "specifying more bytes than compressed size of frame(s). This "
767 "error message replaces ERROR(prefix_unknown), which would be "
768 "confusing, as the first header is actually correct. Note that "
769 "one could be unlucky, it might be a corruption error instead, "
770 "happening right at the place where we expect zstd magic "
771 "bytes. But this is _much_ less likely than a srcSize field "
772 "error.");
773 if (ZSTD_isError(res)) return res;
774 assert(res <= dstCapacity);
775 if (res != 0)
776 dst = (BYTE*)dst + res;
777 dstCapacity -= res;
779 moreThan1Frame = 1;
780 } /* while (srcSize >= ZSTD_frameHeaderSize_prefix) */
782 RETURN_ERROR_IF(srcSize, srcSize_wrong, "input not entirely consumed");
784 return (BYTE*)dst - (BYTE*)dststart;
787 size_t ZSTD_decompress_usingDict(ZSTD_DCtx* dctx,
788 void* dst, size_t dstCapacity,
789 const void* src, size_t srcSize,
790 const void* dict, size_t dictSize)
792 return ZSTD_decompressMultiFrame(dctx, dst, dstCapacity, src, srcSize, dict, dictSize, NULL);
796 static ZSTD_DDict const* ZSTD_getDDict(ZSTD_DCtx* dctx)
798 switch (dctx->dictUses) {
799 default:
800 assert(0 /* Impossible */);
801 /* fall-through */
802 case ZSTD_dont_use:
803 ZSTD_clearDict(dctx);
804 return NULL;
805 case ZSTD_use_indefinitely:
806 return dctx->ddict;
807 case ZSTD_use_once:
808 dctx->dictUses = ZSTD_dont_use;
809 return dctx->ddict;
813 size_t ZSTD_decompressDCtx(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize)
815 return ZSTD_decompress_usingDDict(dctx, dst, dstCapacity, src, srcSize, ZSTD_getDDict(dctx));
819 size_t ZSTD_decompress(void* dst, size_t dstCapacity, const void* src, size_t srcSize)
821 #if defined(ZSTD_HEAPMODE) && (ZSTD_HEAPMODE>=1)
822 size_t regenSize;
823 ZSTD_DCtx* const dctx = ZSTD_createDCtx();
824 RETURN_ERROR_IF(dctx==NULL, memory_allocation, "NULL pointer!");
825 regenSize = ZSTD_decompressDCtx(dctx, dst, dstCapacity, src, srcSize);
826 ZSTD_freeDCtx(dctx);
827 return regenSize;
828 #else /* stack mode */
829 ZSTD_DCtx dctx;
830 ZSTD_initDCtx_internal(&dctx);
831 return ZSTD_decompressDCtx(&dctx, dst, dstCapacity, src, srcSize);
832 #endif
836 /*-**************************************
837 * Advanced Streaming Decompression API
838 * Bufferless and synchronous
839 ****************************************/
840 size_t ZSTD_nextSrcSizeToDecompress(ZSTD_DCtx* dctx) { return dctx->expected; }
843 * Similar to ZSTD_nextSrcSizeToDecompress(), but when when a block input can be streamed,
844 * we allow taking a partial block as the input. Currently only raw uncompressed blocks can
845 * be streamed.
847 * For blocks that can be streamed, this allows us to reduce the latency until we produce
848 * output, and avoid copying the input.
850 * @param inputSize - The total amount of input that the caller currently has.
852 static size_t ZSTD_nextSrcSizeToDecompressWithInputSize(ZSTD_DCtx* dctx, size_t inputSize) {
853 if (!(dctx->stage == ZSTDds_decompressBlock || dctx->stage == ZSTDds_decompressLastBlock))
854 return dctx->expected;
855 if (dctx->bType != bt_raw)
856 return dctx->expected;
857 return MIN(MAX(inputSize, 1), dctx->expected);
860 ZSTD_nextInputType_e ZSTD_nextInputType(ZSTD_DCtx* dctx) {
861 switch(dctx->stage)
863 default: /* should not happen */
864 assert(0);
865 case ZSTDds_getFrameHeaderSize:
866 case ZSTDds_decodeFrameHeader:
867 return ZSTDnit_frameHeader;
868 case ZSTDds_decodeBlockHeader:
869 return ZSTDnit_blockHeader;
870 case ZSTDds_decompressBlock:
871 return ZSTDnit_block;
872 case ZSTDds_decompressLastBlock:
873 return ZSTDnit_lastBlock;
874 case ZSTDds_checkChecksum:
875 return ZSTDnit_checksum;
876 case ZSTDds_decodeSkippableHeader:
877 case ZSTDds_skipFrame:
878 return ZSTDnit_skippableFrame;
882 static int ZSTD_isSkipFrame(ZSTD_DCtx* dctx) { return dctx->stage == ZSTDds_skipFrame; }
884 /** ZSTD_decompressContinue() :
885 * srcSize : must be the exact nb of bytes expected (see ZSTD_nextSrcSizeToDecompress())
886 * @return : nb of bytes generated into `dst` (necessarily <= `dstCapacity)
887 * or an error code, which can be tested using ZSTD_isError() */
888 size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize)
890 DEBUGLOG(5, "ZSTD_decompressContinue (srcSize:%u)", (unsigned)srcSize);
891 /* Sanity check */
892 RETURN_ERROR_IF(srcSize != ZSTD_nextSrcSizeToDecompressWithInputSize(dctx, srcSize), srcSize_wrong, "not allowed");
893 if (dstCapacity) ZSTD_checkContinuity(dctx, dst);
895 switch (dctx->stage)
897 case ZSTDds_getFrameHeaderSize :
898 assert(src != NULL);
899 if (dctx->format == ZSTD_f_zstd1) { /* allows header */
900 assert(srcSize >= ZSTD_FRAMEIDSIZE); /* to read skippable magic number */
901 if ((MEM_readLE32(src) & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) { /* skippable frame */
902 memcpy(dctx->headerBuffer, src, srcSize);
903 dctx->expected = ZSTD_SKIPPABLEHEADERSIZE - srcSize; /* remaining to load to get full skippable frame header */
904 dctx->stage = ZSTDds_decodeSkippableHeader;
905 return 0;
907 dctx->headerSize = ZSTD_frameHeaderSize_internal(src, srcSize, dctx->format);
908 if (ZSTD_isError(dctx->headerSize)) return dctx->headerSize;
909 memcpy(dctx->headerBuffer, src, srcSize);
910 dctx->expected = dctx->headerSize - srcSize;
911 dctx->stage = ZSTDds_decodeFrameHeader;
912 return 0;
914 case ZSTDds_decodeFrameHeader:
915 assert(src != NULL);
916 memcpy(dctx->headerBuffer + (dctx->headerSize - srcSize), src, srcSize);
917 FORWARD_IF_ERROR(ZSTD_decodeFrameHeader(dctx, dctx->headerBuffer, dctx->headerSize), "");
918 dctx->expected = ZSTD_blockHeaderSize;
919 dctx->stage = ZSTDds_decodeBlockHeader;
920 return 0;
922 case ZSTDds_decodeBlockHeader:
923 { blockProperties_t bp;
924 size_t const cBlockSize = ZSTD_getcBlockSize(src, ZSTD_blockHeaderSize, &bp);
925 if (ZSTD_isError(cBlockSize)) return cBlockSize;
926 RETURN_ERROR_IF(cBlockSize > dctx->fParams.blockSizeMax, corruption_detected, "Block Size Exceeds Maximum");
927 dctx->expected = cBlockSize;
928 dctx->bType = bp.blockType;
929 dctx->rleSize = bp.origSize;
930 if (cBlockSize) {
931 dctx->stage = bp.lastBlock ? ZSTDds_decompressLastBlock : ZSTDds_decompressBlock;
932 return 0;
934 /* empty block */
935 if (bp.lastBlock) {
936 if (dctx->fParams.checksumFlag) {
937 dctx->expected = 4;
938 dctx->stage = ZSTDds_checkChecksum;
939 } else {
940 dctx->expected = 0; /* end of frame */
941 dctx->stage = ZSTDds_getFrameHeaderSize;
943 } else {
944 dctx->expected = ZSTD_blockHeaderSize; /* jump to next header */
945 dctx->stage = ZSTDds_decodeBlockHeader;
947 return 0;
950 case ZSTDds_decompressLastBlock:
951 case ZSTDds_decompressBlock:
952 DEBUGLOG(5, "ZSTD_decompressContinue: case ZSTDds_decompressBlock");
953 { size_t rSize;
954 switch(dctx->bType)
956 case bt_compressed:
957 DEBUGLOG(5, "ZSTD_decompressContinue: case bt_compressed");
958 rSize = ZSTD_decompressBlock_internal(dctx, dst, dstCapacity, src, srcSize, /* frame */ 1);
959 dctx->expected = 0; /* Streaming not supported */
960 break;
961 case bt_raw :
962 assert(srcSize <= dctx->expected);
963 rSize = ZSTD_copyRawBlock(dst, dstCapacity, src, srcSize);
964 FORWARD_IF_ERROR(rSize, "ZSTD_copyRawBlock failed");
965 assert(rSize == srcSize);
966 dctx->expected -= rSize;
967 break;
968 case bt_rle :
969 rSize = ZSTD_setRleBlock(dst, dstCapacity, *(const BYTE*)src, dctx->rleSize);
970 dctx->expected = 0; /* Streaming not supported */
971 break;
972 case bt_reserved : /* should never happen */
973 default:
974 RETURN_ERROR(corruption_detected, "invalid block type");
976 FORWARD_IF_ERROR(rSize, "");
977 RETURN_ERROR_IF(rSize > dctx->fParams.blockSizeMax, corruption_detected, "Decompressed Block Size Exceeds Maximum");
978 DEBUGLOG(5, "ZSTD_decompressContinue: decoded size from block : %u", (unsigned)rSize);
979 dctx->decodedSize += rSize;
980 if (dctx->fParams.checksumFlag) XXH64_update(&dctx->xxhState, dst, rSize);
981 dctx->previousDstEnd = (char*)dst + rSize;
983 /* Stay on the same stage until we are finished streaming the block. */
984 if (dctx->expected > 0) {
985 return rSize;
988 if (dctx->stage == ZSTDds_decompressLastBlock) { /* end of frame */
989 DEBUGLOG(4, "ZSTD_decompressContinue: decoded size from frame : %u", (unsigned)dctx->decodedSize);
990 RETURN_ERROR_IF(
991 dctx->fParams.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN
992 && dctx->decodedSize != dctx->fParams.frameContentSize,
993 corruption_detected, "");
994 if (dctx->fParams.checksumFlag) { /* another round for frame checksum */
995 dctx->expected = 4;
996 dctx->stage = ZSTDds_checkChecksum;
997 } else {
998 dctx->expected = 0; /* ends here */
999 dctx->stage = ZSTDds_getFrameHeaderSize;
1001 } else {
1002 dctx->stage = ZSTDds_decodeBlockHeader;
1003 dctx->expected = ZSTD_blockHeaderSize;
1005 return rSize;
1008 case ZSTDds_checkChecksum:
1009 assert(srcSize == 4); /* guaranteed by dctx->expected */
1010 { U32 const h32 = (U32)XXH64_digest(&dctx->xxhState);
1011 U32 const check32 = MEM_readLE32(src);
1012 DEBUGLOG(4, "ZSTD_decompressContinue: checksum : calculated %08X :: %08X read", (unsigned)h32, (unsigned)check32);
1013 RETURN_ERROR_IF(check32 != h32, checksum_wrong, "");
1014 dctx->expected = 0;
1015 dctx->stage = ZSTDds_getFrameHeaderSize;
1016 return 0;
1019 case ZSTDds_decodeSkippableHeader:
1020 assert(src != NULL);
1021 assert(srcSize <= ZSTD_SKIPPABLEHEADERSIZE);
1022 memcpy(dctx->headerBuffer + (ZSTD_SKIPPABLEHEADERSIZE - srcSize), src, srcSize); /* complete skippable header */
1023 dctx->expected = MEM_readLE32(dctx->headerBuffer + ZSTD_FRAMEIDSIZE); /* note : dctx->expected can grow seriously large, beyond local buffer size */
1024 dctx->stage = ZSTDds_skipFrame;
1025 return 0;
1027 case ZSTDds_skipFrame:
1028 dctx->expected = 0;
1029 dctx->stage = ZSTDds_getFrameHeaderSize;
1030 return 0;
1032 default:
1033 assert(0); /* impossible */
1034 RETURN_ERROR(GENERIC, "impossible to reach"); /* some compiler require default to do something */
1039 static size_t ZSTD_refDictContent(ZSTD_DCtx* dctx, const void* dict, size_t dictSize)
1041 dctx->dictEnd = dctx->previousDstEnd;
1042 dctx->virtualStart = (const char*)dict - ((const char*)(dctx->previousDstEnd) - (const char*)(dctx->prefixStart));
1043 dctx->prefixStart = dict;
1044 dctx->previousDstEnd = (const char*)dict + dictSize;
1045 #ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
1046 dctx->dictContentBeginForFuzzing = dctx->prefixStart;
1047 dctx->dictContentEndForFuzzing = dctx->previousDstEnd;
1048 #endif
1049 return 0;
1052 /*! ZSTD_loadDEntropy() :
1053 * dict : must point at beginning of a valid zstd dictionary.
1054 * @return : size of entropy tables read */
1055 size_t
1056 ZSTD_loadDEntropy(ZSTD_entropyDTables_t* entropy,
1057 const void* const dict, size_t const dictSize)
1059 const BYTE* dictPtr = (const BYTE*)dict;
1060 const BYTE* const dictEnd = dictPtr + dictSize;
1062 RETURN_ERROR_IF(dictSize <= 8, dictionary_corrupted, "dict is too small");
1063 assert(MEM_readLE32(dict) == ZSTD_MAGIC_DICTIONARY); /* dict must be valid */
1064 dictPtr += 8; /* skip header = magic + dictID */
1066 ZSTD_STATIC_ASSERT(offsetof(ZSTD_entropyDTables_t, OFTable) == offsetof(ZSTD_entropyDTables_t, LLTable) + sizeof(entropy->LLTable));
1067 ZSTD_STATIC_ASSERT(offsetof(ZSTD_entropyDTables_t, MLTable) == offsetof(ZSTD_entropyDTables_t, OFTable) + sizeof(entropy->OFTable));
1068 ZSTD_STATIC_ASSERT(sizeof(entropy->LLTable) + sizeof(entropy->OFTable) + sizeof(entropy->MLTable) >= HUF_DECOMPRESS_WORKSPACE_SIZE);
1069 { void* const workspace = &entropy->LLTable; /* use fse tables as temporary workspace; implies fse tables are grouped together */
1070 size_t const workspaceSize = sizeof(entropy->LLTable) + sizeof(entropy->OFTable) + sizeof(entropy->MLTable);
1071 #ifdef HUF_FORCE_DECOMPRESS_X1
1072 /* in minimal huffman, we always use X1 variants */
1073 size_t const hSize = HUF_readDTableX1_wksp(entropy->hufTable,
1074 dictPtr, dictEnd - dictPtr,
1075 workspace, workspaceSize);
1076 #else
1077 size_t const hSize = HUF_readDTableX2_wksp(entropy->hufTable,
1078 dictPtr, dictEnd - dictPtr,
1079 workspace, workspaceSize);
1080 #endif
1081 RETURN_ERROR_IF(HUF_isError(hSize), dictionary_corrupted, "");
1082 dictPtr += hSize;
1085 { short offcodeNCount[MaxOff+1];
1086 unsigned offcodeMaxValue = MaxOff, offcodeLog;
1087 size_t const offcodeHeaderSize = FSE_readNCount(offcodeNCount, &offcodeMaxValue, &offcodeLog, dictPtr, dictEnd-dictPtr);
1088 RETURN_ERROR_IF(FSE_isError(offcodeHeaderSize), dictionary_corrupted, "");
1089 RETURN_ERROR_IF(offcodeMaxValue > MaxOff, dictionary_corrupted, "");
1090 RETURN_ERROR_IF(offcodeLog > OffFSELog, dictionary_corrupted, "");
1091 ZSTD_buildFSETable( entropy->OFTable,
1092 offcodeNCount, offcodeMaxValue,
1093 OF_base, OF_bits,
1094 offcodeLog);
1095 dictPtr += offcodeHeaderSize;
1098 { short matchlengthNCount[MaxML+1];
1099 unsigned matchlengthMaxValue = MaxML, matchlengthLog;
1100 size_t const matchlengthHeaderSize = FSE_readNCount(matchlengthNCount, &matchlengthMaxValue, &matchlengthLog, dictPtr, dictEnd-dictPtr);
1101 RETURN_ERROR_IF(FSE_isError(matchlengthHeaderSize), dictionary_corrupted, "");
1102 RETURN_ERROR_IF(matchlengthMaxValue > MaxML, dictionary_corrupted, "");
1103 RETURN_ERROR_IF(matchlengthLog > MLFSELog, dictionary_corrupted, "");
1104 ZSTD_buildFSETable( entropy->MLTable,
1105 matchlengthNCount, matchlengthMaxValue,
1106 ML_base, ML_bits,
1107 matchlengthLog);
1108 dictPtr += matchlengthHeaderSize;
1111 { short litlengthNCount[MaxLL+1];
1112 unsigned litlengthMaxValue = MaxLL, litlengthLog;
1113 size_t const litlengthHeaderSize = FSE_readNCount(litlengthNCount, &litlengthMaxValue, &litlengthLog, dictPtr, dictEnd-dictPtr);
1114 RETURN_ERROR_IF(FSE_isError(litlengthHeaderSize), dictionary_corrupted, "");
1115 RETURN_ERROR_IF(litlengthMaxValue > MaxLL, dictionary_corrupted, "");
1116 RETURN_ERROR_IF(litlengthLog > LLFSELog, dictionary_corrupted, "");
1117 ZSTD_buildFSETable( entropy->LLTable,
1118 litlengthNCount, litlengthMaxValue,
1119 LL_base, LL_bits,
1120 litlengthLog);
1121 dictPtr += litlengthHeaderSize;
1124 RETURN_ERROR_IF(dictPtr+12 > dictEnd, dictionary_corrupted, "");
1125 { int i;
1126 size_t const dictContentSize = (size_t)(dictEnd - (dictPtr+12));
1127 for (i=0; i<3; i++) {
1128 U32 const rep = MEM_readLE32(dictPtr); dictPtr += 4;
1129 RETURN_ERROR_IF(rep==0 || rep > dictContentSize,
1130 dictionary_corrupted, "");
1131 entropy->rep[i] = rep;
1134 return dictPtr - (const BYTE*)dict;
1137 static size_t ZSTD_decompress_insertDictionary(ZSTD_DCtx* dctx, const void* dict, size_t dictSize)
1139 if (dictSize < 8) return ZSTD_refDictContent(dctx, dict, dictSize);
1140 { U32 const magic = MEM_readLE32(dict);
1141 if (magic != ZSTD_MAGIC_DICTIONARY) {
1142 return ZSTD_refDictContent(dctx, dict, dictSize); /* pure content mode */
1144 dctx->dictID = MEM_readLE32((const char*)dict + ZSTD_FRAMEIDSIZE);
1146 /* load entropy tables */
1147 { size_t const eSize = ZSTD_loadDEntropy(&dctx->entropy, dict, dictSize);
1148 RETURN_ERROR_IF(ZSTD_isError(eSize), dictionary_corrupted, "");
1149 dict = (const char*)dict + eSize;
1150 dictSize -= eSize;
1152 dctx->litEntropy = dctx->fseEntropy = 1;
1154 /* reference dictionary content */
1155 return ZSTD_refDictContent(dctx, dict, dictSize);
1158 size_t ZSTD_decompressBegin(ZSTD_DCtx* dctx)
1160 assert(dctx != NULL);
1161 dctx->expected = ZSTD_startingInputLength(dctx->format); /* dctx->format must be properly set */
1162 dctx->stage = ZSTDds_getFrameHeaderSize;
1163 dctx->decodedSize = 0;
1164 dctx->previousDstEnd = NULL;
1165 dctx->prefixStart = NULL;
1166 dctx->virtualStart = NULL;
1167 dctx->dictEnd = NULL;
1168 dctx->entropy.hufTable[0] = (HUF_DTable)((HufLog)*0x1000001); /* cover both little and big endian */
1169 dctx->litEntropy = dctx->fseEntropy = 0;
1170 dctx->dictID = 0;
1171 dctx->bType = bt_reserved;
1172 ZSTD_STATIC_ASSERT(sizeof(dctx->entropy.rep) == sizeof(repStartValue));
1173 memcpy(dctx->entropy.rep, repStartValue, sizeof(repStartValue)); /* initial repcodes */
1174 dctx->LLTptr = dctx->entropy.LLTable;
1175 dctx->MLTptr = dctx->entropy.MLTable;
1176 dctx->OFTptr = dctx->entropy.OFTable;
1177 dctx->HUFptr = dctx->entropy.hufTable;
1178 return 0;
1181 size_t ZSTD_decompressBegin_usingDict(ZSTD_DCtx* dctx, const void* dict, size_t dictSize)
1183 FORWARD_IF_ERROR( ZSTD_decompressBegin(dctx) , "");
1184 if (dict && dictSize)
1185 RETURN_ERROR_IF(
1186 ZSTD_isError(ZSTD_decompress_insertDictionary(dctx, dict, dictSize)),
1187 dictionary_corrupted, "");
1188 return 0;
1192 /* ====== ZSTD_DDict ====== */
1194 size_t ZSTD_decompressBegin_usingDDict(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict)
1196 DEBUGLOG(4, "ZSTD_decompressBegin_usingDDict");
1197 assert(dctx != NULL);
1198 if (ddict) {
1199 const char* const dictStart = (const char*)ZSTD_DDict_dictContent(ddict);
1200 size_t const dictSize = ZSTD_DDict_dictSize(ddict);
1201 const void* const dictEnd = dictStart + dictSize;
1202 dctx->ddictIsCold = (dctx->dictEnd != dictEnd);
1203 DEBUGLOG(4, "DDict is %s",
1204 dctx->ddictIsCold ? "~cold~" : "hot!");
1206 FORWARD_IF_ERROR( ZSTD_decompressBegin(dctx) , "");
1207 if (ddict) { /* NULL ddict is equivalent to no dictionary */
1208 ZSTD_copyDDictParameters(dctx, ddict);
1210 return 0;
1213 /*! ZSTD_getDictID_fromDict() :
1214 * Provides the dictID stored within dictionary.
1215 * if @return == 0, the dictionary is not conformant with Zstandard specification.
1216 * It can still be loaded, but as a content-only dictionary. */
1217 unsigned ZSTD_getDictID_fromDict(const void* dict, size_t dictSize)
1219 if (dictSize < 8) return 0;
1220 if (MEM_readLE32(dict) != ZSTD_MAGIC_DICTIONARY) return 0;
1221 return MEM_readLE32((const char*)dict + ZSTD_FRAMEIDSIZE);
1224 /*! ZSTD_getDictID_fromFrame() :
1225 * Provides the dictID required to decompress frame stored within `src`.
1226 * If @return == 0, the dictID could not be decoded.
1227 * This could for one of the following reasons :
1228 * - The frame does not require a dictionary (most common case).
1229 * - The frame was built with dictID intentionally removed.
1230 * Needed dictionary is a hidden information.
1231 * Note : this use case also happens when using a non-conformant dictionary.
1232 * - `srcSize` is too small, and as a result, frame header could not be decoded.
1233 * Note : possible if `srcSize < ZSTD_FRAMEHEADERSIZE_MAX`.
1234 * - This is not a Zstandard frame.
1235 * When identifying the exact failure cause, it's possible to use
1236 * ZSTD_getFrameHeader(), which will provide a more precise error code. */
1237 unsigned ZSTD_getDictID_fromFrame(const void* src, size_t srcSize)
1239 ZSTD_frameHeader zfp = { 0, 0, 0, ZSTD_frame, 0, 0, 0 };
1240 size_t const hError = ZSTD_getFrameHeader(&zfp, src, srcSize);
1241 if (ZSTD_isError(hError)) return 0;
1242 return zfp.dictID;
1246 /*! ZSTD_decompress_usingDDict() :
1247 * Decompression using a pre-digested Dictionary
1248 * Use dictionary without significant overhead. */
1249 size_t ZSTD_decompress_usingDDict(ZSTD_DCtx* dctx,
1250 void* dst, size_t dstCapacity,
1251 const void* src, size_t srcSize,
1252 const ZSTD_DDict* ddict)
1254 /* pass content and size in case legacy frames are encountered */
1255 return ZSTD_decompressMultiFrame(dctx, dst, dstCapacity, src, srcSize,
1256 NULL, 0,
1257 ddict);
1261 /*=====================================
1262 * Streaming decompression
1263 *====================================*/
1265 ZSTD_DStream* ZSTD_createDStream(void)
1267 DEBUGLOG(3, "ZSTD_createDStream");
1268 return ZSTD_createDStream_advanced(ZSTD_defaultCMem);
1271 ZSTD_DStream* ZSTD_initStaticDStream(void *workspace, size_t workspaceSize)
1273 return ZSTD_initStaticDCtx(workspace, workspaceSize);
1276 ZSTD_DStream* ZSTD_createDStream_advanced(ZSTD_customMem customMem)
1278 return ZSTD_createDCtx_advanced(customMem);
1281 size_t ZSTD_freeDStream(ZSTD_DStream* zds)
1283 return ZSTD_freeDCtx(zds);
1287 /* *** Initialization *** */
1289 size_t ZSTD_DStreamInSize(void) { return ZSTD_BLOCKSIZE_MAX + ZSTD_blockHeaderSize; }
1290 size_t ZSTD_DStreamOutSize(void) { return ZSTD_BLOCKSIZE_MAX; }
1292 size_t ZSTD_DCtx_loadDictionary_advanced(ZSTD_DCtx* dctx,
1293 const void* dict, size_t dictSize,
1294 ZSTD_dictLoadMethod_e dictLoadMethod,
1295 ZSTD_dictContentType_e dictContentType)
1297 RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong, "");
1298 ZSTD_clearDict(dctx);
1299 if (dict && dictSize != 0) {
1300 dctx->ddictLocal = ZSTD_createDDict_advanced(dict, dictSize, dictLoadMethod, dictContentType, dctx->customMem);
1301 RETURN_ERROR_IF(dctx->ddictLocal == NULL, memory_allocation, "NULL pointer!");
1302 dctx->ddict = dctx->ddictLocal;
1303 dctx->dictUses = ZSTD_use_indefinitely;
1305 return 0;
1308 size_t ZSTD_DCtx_loadDictionary_byReference(ZSTD_DCtx* dctx, const void* dict, size_t dictSize)
1310 return ZSTD_DCtx_loadDictionary_advanced(dctx, dict, dictSize, ZSTD_dlm_byRef, ZSTD_dct_auto);
1313 size_t ZSTD_DCtx_loadDictionary(ZSTD_DCtx* dctx, const void* dict, size_t dictSize)
1315 return ZSTD_DCtx_loadDictionary_advanced(dctx, dict, dictSize, ZSTD_dlm_byCopy, ZSTD_dct_auto);
1318 size_t ZSTD_DCtx_refPrefix_advanced(ZSTD_DCtx* dctx, const void* prefix, size_t prefixSize, ZSTD_dictContentType_e dictContentType)
1320 FORWARD_IF_ERROR(ZSTD_DCtx_loadDictionary_advanced(dctx, prefix, prefixSize, ZSTD_dlm_byRef, dictContentType), "");
1321 dctx->dictUses = ZSTD_use_once;
1322 return 0;
1325 size_t ZSTD_DCtx_refPrefix(ZSTD_DCtx* dctx, const void* prefix, size_t prefixSize)
1327 return ZSTD_DCtx_refPrefix_advanced(dctx, prefix, prefixSize, ZSTD_dct_rawContent);
1331 /* ZSTD_initDStream_usingDict() :
1332 * return : expected size, aka ZSTD_startingInputLength().
1333 * this function cannot fail */
1334 size_t ZSTD_initDStream_usingDict(ZSTD_DStream* zds, const void* dict, size_t dictSize)
1336 DEBUGLOG(4, "ZSTD_initDStream_usingDict");
1337 FORWARD_IF_ERROR( ZSTD_DCtx_reset(zds, ZSTD_reset_session_only) , "");
1338 FORWARD_IF_ERROR( ZSTD_DCtx_loadDictionary(zds, dict, dictSize) , "");
1339 return ZSTD_startingInputLength(zds->format);
1342 /* note : this variant can't fail */
1343 size_t ZSTD_initDStream(ZSTD_DStream* zds)
1345 DEBUGLOG(4, "ZSTD_initDStream");
1346 return ZSTD_initDStream_usingDDict(zds, NULL);
1349 /* ZSTD_initDStream_usingDDict() :
1350 * ddict will just be referenced, and must outlive decompression session
1351 * this function cannot fail */
1352 size_t ZSTD_initDStream_usingDDict(ZSTD_DStream* dctx, const ZSTD_DDict* ddict)
1354 FORWARD_IF_ERROR( ZSTD_DCtx_reset(dctx, ZSTD_reset_session_only) , "");
1355 FORWARD_IF_ERROR( ZSTD_DCtx_refDDict(dctx, ddict) , "");
1356 return ZSTD_startingInputLength(dctx->format);
1359 /* ZSTD_resetDStream() :
1360 * return : expected size, aka ZSTD_startingInputLength().
1361 * this function cannot fail */
1362 size_t ZSTD_resetDStream(ZSTD_DStream* dctx)
1364 FORWARD_IF_ERROR(ZSTD_DCtx_reset(dctx, ZSTD_reset_session_only), "");
1365 return ZSTD_startingInputLength(dctx->format);
1369 size_t ZSTD_DCtx_refDDict(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict)
1371 RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong, "");
1372 ZSTD_clearDict(dctx);
1373 if (ddict) {
1374 dctx->ddict = ddict;
1375 dctx->dictUses = ZSTD_use_indefinitely;
1377 return 0;
1380 /* ZSTD_DCtx_setMaxWindowSize() :
1381 * note : no direct equivalence in ZSTD_DCtx_setParameter,
1382 * since this version sets windowSize, and the other sets windowLog */
1383 size_t ZSTD_DCtx_setMaxWindowSize(ZSTD_DCtx* dctx, size_t maxWindowSize)
1385 ZSTD_bounds const bounds = ZSTD_dParam_getBounds(ZSTD_d_windowLogMax);
1386 size_t const min = (size_t)1 << bounds.lowerBound;
1387 size_t const max = (size_t)1 << bounds.upperBound;
1388 RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong, "");
1389 RETURN_ERROR_IF(maxWindowSize < min, parameter_outOfBound, "");
1390 RETURN_ERROR_IF(maxWindowSize > max, parameter_outOfBound, "");
1391 dctx->maxWindowSize = maxWindowSize;
1392 return 0;
1395 size_t ZSTD_DCtx_setFormat(ZSTD_DCtx* dctx, ZSTD_format_e format)
1397 return ZSTD_DCtx_setParameter(dctx, ZSTD_d_format, format);
1400 ZSTD_bounds ZSTD_dParam_getBounds(ZSTD_dParameter dParam)
1402 ZSTD_bounds bounds = { 0, 0, 0 };
1403 switch(dParam) {
1404 case ZSTD_d_windowLogMax:
1405 bounds.lowerBound = ZSTD_WINDOWLOG_ABSOLUTEMIN;
1406 bounds.upperBound = ZSTD_WINDOWLOG_MAX;
1407 return bounds;
1408 case ZSTD_d_format:
1409 bounds.lowerBound = (int)ZSTD_f_zstd1;
1410 bounds.upperBound = (int)ZSTD_f_zstd1_magicless;
1411 ZSTD_STATIC_ASSERT(ZSTD_f_zstd1 < ZSTD_f_zstd1_magicless);
1412 return bounds;
1413 case ZSTD_d_stableOutBuffer:
1414 bounds.lowerBound = (int)ZSTD_obm_buffered;
1415 bounds.upperBound = (int)ZSTD_obm_stable;
1416 return bounds;
1417 default:;
1419 bounds.error = ERROR(parameter_unsupported);
1420 return bounds;
1423 /* ZSTD_dParam_withinBounds:
1424 * @return 1 if value is within dParam bounds,
1425 * 0 otherwise */
1426 static int ZSTD_dParam_withinBounds(ZSTD_dParameter dParam, int value)
1428 ZSTD_bounds const bounds = ZSTD_dParam_getBounds(dParam);
1429 if (ZSTD_isError(bounds.error)) return 0;
1430 if (value < bounds.lowerBound) return 0;
1431 if (value > bounds.upperBound) return 0;
1432 return 1;
1435 #define CHECK_DBOUNDS(p,v) { \
1436 RETURN_ERROR_IF(!ZSTD_dParam_withinBounds(p, v), parameter_outOfBound, ""); \
1439 size_t ZSTD_DCtx_setParameter(ZSTD_DCtx* dctx, ZSTD_dParameter dParam, int value)
1441 RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong, "");
1442 switch(dParam) {
1443 case ZSTD_d_windowLogMax:
1444 if (value == 0) value = ZSTD_WINDOWLOG_LIMIT_DEFAULT;
1445 CHECK_DBOUNDS(ZSTD_d_windowLogMax, value);
1446 dctx->maxWindowSize = ((size_t)1) << value;
1447 return 0;
1448 case ZSTD_d_format:
1449 CHECK_DBOUNDS(ZSTD_d_format, value);
1450 dctx->format = (ZSTD_format_e)value;
1451 return 0;
1452 case ZSTD_d_stableOutBuffer:
1453 CHECK_DBOUNDS(ZSTD_d_stableOutBuffer, value);
1454 dctx->outBufferMode = (ZSTD_outBufferMode_e)value;
1455 return 0;
1456 default:;
1458 RETURN_ERROR(parameter_unsupported, "");
1461 size_t ZSTD_DCtx_reset(ZSTD_DCtx* dctx, ZSTD_ResetDirective reset)
1463 if ( (reset == ZSTD_reset_session_only)
1464 || (reset == ZSTD_reset_session_and_parameters) ) {
1465 dctx->streamStage = zdss_init;
1466 dctx->noForwardProgress = 0;
1468 if ( (reset == ZSTD_reset_parameters)
1469 || (reset == ZSTD_reset_session_and_parameters) ) {
1470 RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong, "");
1471 ZSTD_clearDict(dctx);
1472 dctx->format = ZSTD_f_zstd1;
1473 dctx->maxWindowSize = ZSTD_MAXWINDOWSIZE_DEFAULT;
1475 return 0;
1479 size_t ZSTD_sizeof_DStream(const ZSTD_DStream* dctx)
1481 return ZSTD_sizeof_DCtx(dctx);
1484 size_t ZSTD_decodingBufferSize_min(unsigned long long windowSize, unsigned long long frameContentSize)
1486 size_t const blockSize = (size_t) MIN(windowSize, ZSTD_BLOCKSIZE_MAX);
1487 unsigned long long const neededRBSize = windowSize + blockSize + (WILDCOPY_OVERLENGTH * 2);
1488 unsigned long long const neededSize = MIN(frameContentSize, neededRBSize);
1489 size_t const minRBSize = (size_t) neededSize;
1490 RETURN_ERROR_IF((unsigned long long)minRBSize != neededSize,
1491 frameParameter_windowTooLarge, "");
1492 return minRBSize;
1495 size_t ZSTD_estimateDStreamSize(size_t windowSize)
1497 size_t const blockSize = MIN(windowSize, ZSTD_BLOCKSIZE_MAX);
1498 size_t const inBuffSize = blockSize; /* no block can be larger */
1499 size_t const outBuffSize = ZSTD_decodingBufferSize_min(windowSize, ZSTD_CONTENTSIZE_UNKNOWN);
1500 return ZSTD_estimateDCtxSize() + inBuffSize + outBuffSize;
1503 size_t ZSTD_estimateDStreamSize_fromFrame(const void* src, size_t srcSize)
1505 U32 const windowSizeMax = 1U << ZSTD_WINDOWLOG_MAX; /* note : should be user-selectable, but requires an additional parameter (or a dctx) */
1506 ZSTD_frameHeader zfh;
1507 size_t const err = ZSTD_getFrameHeader(&zfh, src, srcSize);
1508 if (ZSTD_isError(err)) return err;
1509 RETURN_ERROR_IF(err>0, srcSize_wrong, "");
1510 RETURN_ERROR_IF(zfh.windowSize > windowSizeMax,
1511 frameParameter_windowTooLarge, "");
1512 return ZSTD_estimateDStreamSize((size_t)zfh.windowSize);
1516 /* ***** Decompression ***** */
1518 static int ZSTD_DCtx_isOverflow(ZSTD_DStream* zds, size_t const neededInBuffSize, size_t const neededOutBuffSize)
1520 return (zds->inBuffSize + zds->outBuffSize) >= (neededInBuffSize + neededOutBuffSize) * ZSTD_WORKSPACETOOLARGE_FACTOR;
1523 static void ZSTD_DCtx_updateOversizedDuration(ZSTD_DStream* zds, size_t const neededInBuffSize, size_t const neededOutBuffSize)
1525 if (ZSTD_DCtx_isOverflow(zds, neededInBuffSize, neededOutBuffSize))
1526 zds->oversizedDuration++;
1527 else
1528 zds->oversizedDuration = 0;
1531 static int ZSTD_DCtx_isOversizedTooLong(ZSTD_DStream* zds)
1533 return zds->oversizedDuration >= ZSTD_WORKSPACETOOLARGE_MAXDURATION;
1536 /* Checks that the output buffer hasn't changed if ZSTD_obm_stable is used. */
1537 static size_t ZSTD_checkOutBuffer(ZSTD_DStream const* zds, ZSTD_outBuffer const* output)
1539 ZSTD_outBuffer const expect = zds->expectedOutBuffer;
1540 /* No requirement when ZSTD_obm_stable is not enabled. */
1541 if (zds->outBufferMode != ZSTD_obm_stable)
1542 return 0;
1543 /* Any buffer is allowed in zdss_init, this must be the same for every other call until
1544 * the context is reset.
1546 if (zds->streamStage == zdss_init)
1547 return 0;
1548 /* The buffer must match our expectation exactly. */
1549 if (expect.dst == output->dst && expect.pos == output->pos && expect.size == output->size)
1550 return 0;
1551 RETURN_ERROR(dstBuffer_wrong, "ZSTD_obm_stable enabled but output differs!");
1554 /* Calls ZSTD_decompressContinue() with the right parameters for ZSTD_decompressStream()
1555 * and updates the stage and the output buffer state. This call is extracted so it can be
1556 * used both when reading directly from the ZSTD_inBuffer, and in buffered input mode.
1557 * NOTE: You must break after calling this function since the streamStage is modified.
1559 static size_t ZSTD_decompressContinueStream(
1560 ZSTD_DStream* zds, char** op, char* oend,
1561 void const* src, size_t srcSize) {
1562 int const isSkipFrame = ZSTD_isSkipFrame(zds);
1563 if (zds->outBufferMode == ZSTD_obm_buffered) {
1564 size_t const dstSize = isSkipFrame ? 0 : zds->outBuffSize - zds->outStart;
1565 size_t const decodedSize = ZSTD_decompressContinue(zds,
1566 zds->outBuff + zds->outStart, dstSize, src, srcSize);
1567 FORWARD_IF_ERROR(decodedSize, "");
1568 if (!decodedSize && !isSkipFrame) {
1569 zds->streamStage = zdss_read;
1570 } else {
1571 zds->outEnd = zds->outStart + decodedSize;
1572 zds->streamStage = zdss_flush;
1574 } else {
1575 /* Write directly into the output buffer */
1576 size_t const dstSize = isSkipFrame ? 0 : oend - *op;
1577 size_t const decodedSize = ZSTD_decompressContinue(zds, *op, dstSize, src, srcSize);
1578 FORWARD_IF_ERROR(decodedSize, "");
1579 *op += decodedSize;
1580 /* Flushing is not needed. */
1581 zds->streamStage = zdss_read;
1582 assert(*op <= oend);
1583 assert(zds->outBufferMode == ZSTD_obm_stable);
1585 return 0;
1588 size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inBuffer* input)
1590 const char* const src = (const char*)input->src;
1591 const char* const istart = input->pos != 0 ? src + input->pos : src;
1592 const char* const iend = input->size != 0 ? src + input->size : src;
1593 const char* ip = istart;
1594 char* const dst = (char*)output->dst;
1595 char* const ostart = output->pos != 0 ? dst + output->pos : dst;
1596 char* const oend = output->size != 0 ? dst + output->size : dst;
1597 char* op = ostart;
1598 U32 someMoreWork = 1;
1600 DEBUGLOG(5, "ZSTD_decompressStream");
1601 RETURN_ERROR_IF(
1602 input->pos > input->size,
1603 srcSize_wrong,
1604 "forbidden. in: pos: %u vs size: %u",
1605 (U32)input->pos, (U32)input->size);
1606 RETURN_ERROR_IF(
1607 output->pos > output->size,
1608 dstSize_tooSmall,
1609 "forbidden. out: pos: %u vs size: %u",
1610 (U32)output->pos, (U32)output->size);
1611 DEBUGLOG(5, "input size : %u", (U32)(input->size - input->pos));
1612 FORWARD_IF_ERROR(ZSTD_checkOutBuffer(zds, output), "");
1614 while (someMoreWork) {
1615 switch(zds->streamStage)
1617 case zdss_init :
1618 DEBUGLOG(5, "stage zdss_init => transparent reset ");
1619 zds->streamStage = zdss_loadHeader;
1620 zds->lhSize = zds->inPos = zds->outStart = zds->outEnd = 0;
1621 zds->legacyVersion = 0;
1622 zds->hostageByte = 0;
1623 zds->expectedOutBuffer = *output;
1624 /* fall-through */
1626 case zdss_loadHeader :
1627 DEBUGLOG(5, "stage zdss_loadHeader (srcSize : %u)", (U32)(iend - ip));
1628 #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT>=1)
1629 if (zds->legacyVersion) {
1630 RETURN_ERROR_IF(zds->staticSize, memory_allocation,
1631 "legacy support is incompatible with static dctx");
1632 { size_t const hint = ZSTD_decompressLegacyStream(zds->legacyContext, zds->legacyVersion, output, input);
1633 if (hint==0) zds->streamStage = zdss_init;
1634 return hint;
1636 #endif
1637 { size_t const hSize = ZSTD_getFrameHeader_advanced(&zds->fParams, zds->headerBuffer, zds->lhSize, zds->format);
1638 DEBUGLOG(5, "header size : %u", (U32)hSize);
1639 if (ZSTD_isError(hSize)) {
1640 #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT>=1)
1641 U32 const legacyVersion = ZSTD_isLegacy(istart, iend-istart);
1642 if (legacyVersion) {
1643 ZSTD_DDict const* const ddict = ZSTD_getDDict(zds);
1644 const void* const dict = ddict ? ZSTD_DDict_dictContent(ddict) : NULL;
1645 size_t const dictSize = ddict ? ZSTD_DDict_dictSize(ddict) : 0;
1646 DEBUGLOG(5, "ZSTD_decompressStream: detected legacy version v0.%u", legacyVersion);
1647 RETURN_ERROR_IF(zds->staticSize, memory_allocation,
1648 "legacy support is incompatible with static dctx");
1649 FORWARD_IF_ERROR(ZSTD_initLegacyStream(&zds->legacyContext,
1650 zds->previousLegacyVersion, legacyVersion,
1651 dict, dictSize), "");
1652 zds->legacyVersion = zds->previousLegacyVersion = legacyVersion;
1653 { size_t const hint = ZSTD_decompressLegacyStream(zds->legacyContext, legacyVersion, output, input);
1654 if (hint==0) zds->streamStage = zdss_init; /* or stay in stage zdss_loadHeader */
1655 return hint;
1657 #endif
1658 return hSize; /* error */
1660 if (hSize != 0) { /* need more input */
1661 size_t const toLoad = hSize - zds->lhSize; /* if hSize!=0, hSize > zds->lhSize */
1662 size_t const remainingInput = (size_t)(iend-ip);
1663 assert(iend >= ip);
1664 if (toLoad > remainingInput) { /* not enough input to load full header */
1665 if (remainingInput > 0) {
1666 memcpy(zds->headerBuffer + zds->lhSize, ip, remainingInput);
1667 zds->lhSize += remainingInput;
1669 input->pos = input->size;
1670 return (MAX((size_t)ZSTD_FRAMEHEADERSIZE_MIN(zds->format), hSize) - zds->lhSize) + ZSTD_blockHeaderSize; /* remaining header bytes + next block header */
1672 assert(ip != NULL);
1673 memcpy(zds->headerBuffer + zds->lhSize, ip, toLoad); zds->lhSize = hSize; ip += toLoad;
1674 break;
1677 /* check for single-pass mode opportunity */
1678 if (zds->fParams.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN
1679 && zds->fParams.frameType != ZSTD_skippableFrame
1680 && (U64)(size_t)(oend-op) >= zds->fParams.frameContentSize) {
1681 size_t const cSize = ZSTD_findFrameCompressedSize(istart, iend-istart);
1682 if (cSize <= (size_t)(iend-istart)) {
1683 /* shortcut : using single-pass mode */
1684 size_t const decompressedSize = ZSTD_decompress_usingDDict(zds, op, oend-op, istart, cSize, ZSTD_getDDict(zds));
1685 if (ZSTD_isError(decompressedSize)) return decompressedSize;
1686 DEBUGLOG(4, "shortcut to single-pass ZSTD_decompress_usingDDict()")
1687 ip = istart + cSize;
1688 op += decompressedSize;
1689 zds->expected = 0;
1690 zds->streamStage = zdss_init;
1691 someMoreWork = 0;
1692 break;
1695 /* Check output buffer is large enough for ZSTD_odm_stable. */
1696 if (zds->outBufferMode == ZSTD_obm_stable
1697 && zds->fParams.frameType != ZSTD_skippableFrame
1698 && zds->fParams.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN
1699 && (U64)(size_t)(oend-op) < zds->fParams.frameContentSize) {
1700 RETURN_ERROR(dstSize_tooSmall, "ZSTD_obm_stable passed but ZSTD_outBuffer is too small");
1703 /* Consume header (see ZSTDds_decodeFrameHeader) */
1704 DEBUGLOG(4, "Consume header");
1705 FORWARD_IF_ERROR(ZSTD_decompressBegin_usingDDict(zds, ZSTD_getDDict(zds)), "");
1707 if ((MEM_readLE32(zds->headerBuffer) & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) { /* skippable frame */
1708 zds->expected = MEM_readLE32(zds->headerBuffer + ZSTD_FRAMEIDSIZE);
1709 zds->stage = ZSTDds_skipFrame;
1710 } else {
1711 FORWARD_IF_ERROR(ZSTD_decodeFrameHeader(zds, zds->headerBuffer, zds->lhSize), "");
1712 zds->expected = ZSTD_blockHeaderSize;
1713 zds->stage = ZSTDds_decodeBlockHeader;
1716 /* control buffer memory usage */
1717 DEBUGLOG(4, "Control max memory usage (%u KB <= max %u KB)",
1718 (U32)(zds->fParams.windowSize >>10),
1719 (U32)(zds->maxWindowSize >> 10) );
1720 zds->fParams.windowSize = MAX(zds->fParams.windowSize, 1U << ZSTD_WINDOWLOG_ABSOLUTEMIN);
1721 RETURN_ERROR_IF(zds->fParams.windowSize > zds->maxWindowSize,
1722 frameParameter_windowTooLarge, "");
1724 /* Adapt buffer sizes to frame header instructions */
1725 { size_t const neededInBuffSize = MAX(zds->fParams.blockSizeMax, 4 /* frame checksum */);
1726 size_t const neededOutBuffSize = zds->outBufferMode == ZSTD_obm_buffered
1727 ? ZSTD_decodingBufferSize_min(zds->fParams.windowSize, zds->fParams.frameContentSize)
1728 : 0;
1730 ZSTD_DCtx_updateOversizedDuration(zds, neededInBuffSize, neededOutBuffSize);
1732 { int const tooSmall = (zds->inBuffSize < neededInBuffSize) || (zds->outBuffSize < neededOutBuffSize);
1733 int const tooLarge = ZSTD_DCtx_isOversizedTooLong(zds);
1735 if (tooSmall || tooLarge) {
1736 size_t const bufferSize = neededInBuffSize + neededOutBuffSize;
1737 DEBUGLOG(4, "inBuff : from %u to %u",
1738 (U32)zds->inBuffSize, (U32)neededInBuffSize);
1739 DEBUGLOG(4, "outBuff : from %u to %u",
1740 (U32)zds->outBuffSize, (U32)neededOutBuffSize);
1741 if (zds->staticSize) { /* static DCtx */
1742 DEBUGLOG(4, "staticSize : %u", (U32)zds->staticSize);
1743 assert(zds->staticSize >= sizeof(ZSTD_DCtx)); /* controlled at init */
1744 RETURN_ERROR_IF(
1745 bufferSize > zds->staticSize - sizeof(ZSTD_DCtx),
1746 memory_allocation, "");
1747 } else {
1748 ZSTD_free(zds->inBuff, zds->customMem);
1749 zds->inBuffSize = 0;
1750 zds->outBuffSize = 0;
1751 zds->inBuff = (char*)ZSTD_malloc(bufferSize, zds->customMem);
1752 RETURN_ERROR_IF(zds->inBuff == NULL, memory_allocation, "");
1754 zds->inBuffSize = neededInBuffSize;
1755 zds->outBuff = zds->inBuff + zds->inBuffSize;
1756 zds->outBuffSize = neededOutBuffSize;
1757 } } }
1758 zds->streamStage = zdss_read;
1759 /* fall-through */
1761 case zdss_read:
1762 DEBUGLOG(5, "stage zdss_read");
1763 { size_t const neededInSize = ZSTD_nextSrcSizeToDecompressWithInputSize(zds, iend - ip);
1764 DEBUGLOG(5, "neededInSize = %u", (U32)neededInSize);
1765 if (neededInSize==0) { /* end of frame */
1766 zds->streamStage = zdss_init;
1767 someMoreWork = 0;
1768 break;
1770 if ((size_t)(iend-ip) >= neededInSize) { /* decode directly from src */
1771 FORWARD_IF_ERROR(ZSTD_decompressContinueStream(zds, &op, oend, ip, neededInSize), "");
1772 ip += neededInSize;
1773 /* Function modifies the stage so we must break */
1774 break;
1776 if (ip==iend) { someMoreWork = 0; break; } /* no more input */
1777 zds->streamStage = zdss_load;
1778 /* fall-through */
1780 case zdss_load:
1781 { size_t const neededInSize = ZSTD_nextSrcSizeToDecompress(zds);
1782 size_t const toLoad = neededInSize - zds->inPos;
1783 int const isSkipFrame = ZSTD_isSkipFrame(zds);
1784 size_t loadedSize;
1785 /* At this point we shouldn't be decompressing a block that we can stream. */
1786 assert(neededInSize == ZSTD_nextSrcSizeToDecompressWithInputSize(zds, iend - ip));
1787 if (isSkipFrame) {
1788 loadedSize = MIN(toLoad, (size_t)(iend-ip));
1789 } else {
1790 RETURN_ERROR_IF(toLoad > zds->inBuffSize - zds->inPos,
1791 corruption_detected,
1792 "should never happen");
1793 loadedSize = ZSTD_limitCopy(zds->inBuff + zds->inPos, toLoad, ip, iend-ip);
1795 ip += loadedSize;
1796 zds->inPos += loadedSize;
1797 if (loadedSize < toLoad) { someMoreWork = 0; break; } /* not enough input, wait for more */
1799 /* decode loaded input */
1800 zds->inPos = 0; /* input is consumed */
1801 FORWARD_IF_ERROR(ZSTD_decompressContinueStream(zds, &op, oend, zds->inBuff, neededInSize), "");
1802 /* Function modifies the stage so we must break */
1803 break;
1805 case zdss_flush:
1806 { size_t const toFlushSize = zds->outEnd - zds->outStart;
1807 size_t const flushedSize = ZSTD_limitCopy(op, oend-op, zds->outBuff + zds->outStart, toFlushSize);
1808 op += flushedSize;
1809 zds->outStart += flushedSize;
1810 if (flushedSize == toFlushSize) { /* flush completed */
1811 zds->streamStage = zdss_read;
1812 if ( (zds->outBuffSize < zds->fParams.frameContentSize)
1813 && (zds->outStart + zds->fParams.blockSizeMax > zds->outBuffSize) ) {
1814 DEBUGLOG(5, "restart filling outBuff from beginning (left:%i, needed:%u)",
1815 (int)(zds->outBuffSize - zds->outStart),
1816 (U32)zds->fParams.blockSizeMax);
1817 zds->outStart = zds->outEnd = 0;
1819 break;
1821 /* cannot complete flush */
1822 someMoreWork = 0;
1823 break;
1825 default:
1826 assert(0); /* impossible */
1827 RETURN_ERROR(GENERIC, "impossible to reach"); /* some compiler require default to do something */
1830 /* result */
1831 input->pos = (size_t)(ip - (const char*)(input->src));
1832 output->pos = (size_t)(op - (char*)(output->dst));
1834 /* Update the expected output buffer for ZSTD_obm_stable. */
1835 zds->expectedOutBuffer = *output;
1837 if ((ip==istart) && (op==ostart)) { /* no forward progress */
1838 zds->noForwardProgress ++;
1839 if (zds->noForwardProgress >= ZSTD_NO_FORWARD_PROGRESS_MAX) {
1840 RETURN_ERROR_IF(op==oend, dstSize_tooSmall, "");
1841 RETURN_ERROR_IF(ip==iend, srcSize_wrong, "");
1842 assert(0);
1844 } else {
1845 zds->noForwardProgress = 0;
1847 { size_t nextSrcSizeHint = ZSTD_nextSrcSizeToDecompress(zds);
1848 if (!nextSrcSizeHint) { /* frame fully decoded */
1849 if (zds->outEnd == zds->outStart) { /* output fully flushed */
1850 if (zds->hostageByte) {
1851 if (input->pos >= input->size) {
1852 /* can't release hostage (not present) */
1853 zds->streamStage = zdss_read;
1854 return 1;
1856 input->pos++; /* release hostage */
1857 } /* zds->hostageByte */
1858 return 0;
1859 } /* zds->outEnd == zds->outStart */
1860 if (!zds->hostageByte) { /* output not fully flushed; keep last byte as hostage; will be released when all output is flushed */
1861 input->pos--; /* note : pos > 0, otherwise, impossible to finish reading last block */
1862 zds->hostageByte=1;
1864 return 1;
1865 } /* nextSrcSizeHint==0 */
1866 nextSrcSizeHint += ZSTD_blockHeaderSize * (ZSTD_nextInputType(zds) == ZSTDnit_block); /* preload header of next block */
1867 assert(zds->inPos <= nextSrcSizeHint);
1868 nextSrcSizeHint -= zds->inPos; /* part already loaded*/
1869 return nextSrcSizeHint;
1873 size_t ZSTD_decompressStream_simpleArgs (
1874 ZSTD_DCtx* dctx,
1875 void* dst, size_t dstCapacity, size_t* dstPos,
1876 const void* src, size_t srcSize, size_t* srcPos)
1878 ZSTD_outBuffer output = { dst, dstCapacity, *dstPos };
1879 ZSTD_inBuffer input = { src, srcSize, *srcPos };
1880 /* ZSTD_compress_generic() will check validity of dstPos and srcPos */
1881 size_t const cErr = ZSTD_decompressStream(dctx, &output, &input);
1882 *dstPos = output.pos;
1883 *srcPos = input.pos;
1884 return cErr;