1 /*
2 * Copyright (c) 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
30 /*!
31 * MAXWINDOWSIZE_DEFAULT :
32 * maximum window size accepted by DStream __by default__.
33 * Frames requiring more memory will be rejected.
34 * It's possible to set a different limit using ZSTD_DCtx_setMaxWindowSize().
35 */
36 #ifndef ZSTD_MAXWINDOWSIZE_DEFAULT
37 # define ZSTD_MAXWINDOWSIZE_DEFAULT (((U32)1 << ZSTD_WINDOWLOG_LIMIT_DEFAULT) + 1)
38 #endif
39
40 /*!
41 * NO_FORWARD_PROGRESS_MAX :
42 * maximum allowed nb of calls to ZSTD_decompressStream()
43 * without any forward progress
44 * (defined as: no byte read from input, and no byte flushed to output)
45 * before triggering an error.
46 */
47 #ifndef ZSTD_NO_FORWARD_PROGRESS_MAX
48 # define ZSTD_NO_FORWARD_PROGRESS_MAX 16
49 #endif
50
51
52 /*-*******************************************************
53 * Dependencies
54 *********************************************************/
55 #include "../common/zstd_deps.h" /* ZSTD_memcpy, ZSTD_memmove, ZSTD_memset */
56 #include "../common/mem.h" /* low level memory routines */
57 #define FSE_STATIC_LINKING_ONLY
58 #include "../common/fse.h"
59 #define HUF_STATIC_LINKING_ONLY
60 #include "../common/huf.h"
61 #include <linux/xxhash.h> /* xxh64_reset, xxh64_update, xxh64_digest, XXH64 */
62 #include "../common/zstd_internal.h" /* blockProperties_t */
63 #include "zstd_decompress_internal.h" /* ZSTD_DCtx */
64 #include "zstd_ddict.h" /* ZSTD_DDictDictContent */
65 #include "zstd_decompress_block.h" /* ZSTD_decompressBlock_internal */
66
67
68
69
70 /* ***********************************
71 * Multiple DDicts Hashset internals *
72 *************************************/
73
74 #define DDICT_HASHSET_MAX_LOAD_FACTOR_COUNT_MULT 4
75 #define DDICT_HASHSET_MAX_LOAD_FACTOR_SIZE_MULT 3 /* These two constants represent SIZE_MULT/COUNT_MULT load factor without using a float.
76 * Currently, that means a 0.75 load factor.
77 * So, if count * COUNT_MULT / size * SIZE_MULT != 0, then we've exceeded
78 * the load factor of the ddict hash set.
79 */
80
81 #define DDICT_HASHSET_TABLE_BASE_SIZE 64
82 #define DDICT_HASHSET_RESIZE_FACTOR 2
83
84 /* Hash function to determine starting position of dict insertion within the table
85 * Returns an index between [0, hashSet->ddictPtrTableSize]
86 */
ZSTD_DDictHashSet_getIndex(const ZSTD_DDictHashSet * hashSet,U32 dictID)87 static size_t ZSTD_DDictHashSet_getIndex(const ZSTD_DDictHashSet* hashSet, U32 dictID) {
88 const U64 hash = xxh64(&dictID, sizeof(U32), 0);
89 /* DDict ptr table size is a multiple of 2, use size - 1 as mask to get index within [0, hashSet->ddictPtrTableSize) */
90 return hash & (hashSet->ddictPtrTableSize - 1);
91 }
92
93 /* Adds DDict to a hashset without resizing it.
94 * If inserting a DDict with a dictID that already exists in the set, replaces the one in the set.
95 * Returns 0 if successful, or a zstd error code if something went wrong.
96 */
ZSTD_DDictHashSet_emplaceDDict(ZSTD_DDictHashSet * hashSet,const ZSTD_DDict * ddict)97 static size_t ZSTD_DDictHashSet_emplaceDDict(ZSTD_DDictHashSet* hashSet, const ZSTD_DDict* ddict) {
98 const U32 dictID = ZSTD_getDictID_fromDDict(ddict);
99 size_t idx = ZSTD_DDictHashSet_getIndex(hashSet, dictID);
100 const size_t idxRangeMask = hashSet->ddictPtrTableSize - 1;
101 RETURN_ERROR_IF(hashSet->ddictPtrCount == hashSet->ddictPtrTableSize, GENERIC, "Hash set is full!");
102 DEBUGLOG(4, "Hashed index: for dictID: %u is %zu", dictID, idx);
103 while (hashSet->ddictPtrTable[idx] != NULL) {
104 /* Replace existing ddict if inserting ddict with same dictID */
105 if (ZSTD_getDictID_fromDDict(hashSet->ddictPtrTable[idx]) == dictID) {
106 DEBUGLOG(4, "DictID already exists, replacing rather than adding");
107 hashSet->ddictPtrTable[idx] = ddict;
108 return 0;
109 }
110 idx &= idxRangeMask;
111 idx++;
112 }
113 DEBUGLOG(4, "Final idx after probing for dictID %u is: %zu", dictID, idx);
114 hashSet->ddictPtrTable[idx] = ddict;
115 hashSet->ddictPtrCount++;
116 return 0;
117 }
118
119 /* Expands hash table by factor of DDICT_HASHSET_RESIZE_FACTOR and
120 * rehashes all values, allocates new table, frees old table.
121 * Returns 0 on success, otherwise a zstd error code.
122 */
ZSTD_DDictHashSet_expand(ZSTD_DDictHashSet * hashSet,ZSTD_customMem customMem)123 static size_t ZSTD_DDictHashSet_expand(ZSTD_DDictHashSet* hashSet, ZSTD_customMem customMem) {
124 size_t newTableSize = hashSet->ddictPtrTableSize * DDICT_HASHSET_RESIZE_FACTOR;
125 const ZSTD_DDict** newTable = (const ZSTD_DDict**)ZSTD_customCalloc(sizeof(ZSTD_DDict*) * newTableSize, customMem);
126 const ZSTD_DDict** oldTable = hashSet->ddictPtrTable;
127 size_t oldTableSize = hashSet->ddictPtrTableSize;
128 size_t i;
129
130 DEBUGLOG(4, "Expanding DDict hash table! Old size: %zu new size: %zu", oldTableSize, newTableSize);
131 RETURN_ERROR_IF(!newTable, memory_allocation, "Expanded hashset allocation failed!");
132 hashSet->ddictPtrTable = newTable;
133 hashSet->ddictPtrTableSize = newTableSize;
134 hashSet->ddictPtrCount = 0;
135 for (i = 0; i < oldTableSize; ++i) {
136 if (oldTable[i] != NULL) {
137 FORWARD_IF_ERROR(ZSTD_DDictHashSet_emplaceDDict(hashSet, oldTable[i]), "");
138 }
139 }
140 ZSTD_customFree((void*)oldTable, customMem);
141 DEBUGLOG(4, "Finished re-hash");
142 return 0;
143 }
144
145 /* Fetches a DDict with the given dictID
146 * Returns the ZSTD_DDict* with the requested dictID. If it doesn't exist, then returns NULL.
147 */
ZSTD_DDictHashSet_getDDict(ZSTD_DDictHashSet * hashSet,U32 dictID)148 static const ZSTD_DDict* ZSTD_DDictHashSet_getDDict(ZSTD_DDictHashSet* hashSet, U32 dictID) {
149 size_t idx = ZSTD_DDictHashSet_getIndex(hashSet, dictID);
150 const size_t idxRangeMask = hashSet->ddictPtrTableSize - 1;
151 DEBUGLOG(4, "Hashed index: for dictID: %u is %zu", dictID, idx);
152 for (;;) {
153 size_t currDictID = ZSTD_getDictID_fromDDict(hashSet->ddictPtrTable[idx]);
154 if (currDictID == dictID || currDictID == 0) {
155 /* currDictID == 0 implies a NULL ddict entry */
156 break;
157 } else {
158 idx &= idxRangeMask; /* Goes to start of table when we reach the end */
159 idx++;
160 }
161 }
162 DEBUGLOG(4, "Final idx after probing for dictID %u is: %zu", dictID, idx);
163 return hashSet->ddictPtrTable[idx];
164 }
165
166 /* Allocates space for and returns a ddict hash set
167 * The hash set's ZSTD_DDict* table has all values automatically set to NULL to begin with.
168 * Returns NULL if allocation failed.
169 */
ZSTD_createDDictHashSet(ZSTD_customMem customMem)170 static ZSTD_DDictHashSet* ZSTD_createDDictHashSet(ZSTD_customMem customMem) {
171 ZSTD_DDictHashSet* ret = (ZSTD_DDictHashSet*)ZSTD_customMalloc(sizeof(ZSTD_DDictHashSet), customMem);
172 DEBUGLOG(4, "Allocating new hash set");
173 if (!ret)
174 return NULL;
175 ret->ddictPtrTable = (const ZSTD_DDict**)ZSTD_customCalloc(DDICT_HASHSET_TABLE_BASE_SIZE * sizeof(ZSTD_DDict*), customMem);
176 if (!ret->ddictPtrTable) {
177 ZSTD_customFree(ret, customMem);
178 return NULL;
179 }
180 ret->ddictPtrTableSize = DDICT_HASHSET_TABLE_BASE_SIZE;
181 ret->ddictPtrCount = 0;
182 return ret;
183 }
184
185 /* Frees the table of ZSTD_DDict* within a hashset, then frees the hashset itself.
186 * Note: The ZSTD_DDict* within the table are NOT freed.
187 */
ZSTD_freeDDictHashSet(ZSTD_DDictHashSet * hashSet,ZSTD_customMem customMem)188 static void ZSTD_freeDDictHashSet(ZSTD_DDictHashSet* hashSet, ZSTD_customMem customMem) {
189 DEBUGLOG(4, "Freeing ddict hash set");
190 if (hashSet && hashSet->ddictPtrTable) {
191 ZSTD_customFree((void*)hashSet->ddictPtrTable, customMem);
192 }
193 if (hashSet) {
194 ZSTD_customFree(hashSet, customMem);
195 }
196 }
197
198 /* Public function: Adds a DDict into the ZSTD_DDictHashSet, possibly triggering a resize of the hash set.
199 * Returns 0 on success, or a ZSTD error.
200 */
ZSTD_DDictHashSet_addDDict(ZSTD_DDictHashSet * hashSet,const ZSTD_DDict * ddict,ZSTD_customMem customMem)201 static size_t ZSTD_DDictHashSet_addDDict(ZSTD_DDictHashSet* hashSet, const ZSTD_DDict* ddict, ZSTD_customMem customMem) {
202 DEBUGLOG(4, "Adding dict ID: %u to hashset with - Count: %zu Tablesize: %zu", ZSTD_getDictID_fromDDict(ddict), hashSet->ddictPtrCount, hashSet->ddictPtrTableSize);
203 if (hashSet->ddictPtrCount * DDICT_HASHSET_MAX_LOAD_FACTOR_COUNT_MULT / hashSet->ddictPtrTableSize * DDICT_HASHSET_MAX_LOAD_FACTOR_SIZE_MULT != 0) {
204 FORWARD_IF_ERROR(ZSTD_DDictHashSet_expand(hashSet, customMem), "");
205 }
206 FORWARD_IF_ERROR(ZSTD_DDictHashSet_emplaceDDict(hashSet, ddict), "");
207 return 0;
208 }
209
210 /*-*************************************************************
211 * Context management
212 ***************************************************************/
ZSTD_sizeof_DCtx(const ZSTD_DCtx * dctx)213 size_t ZSTD_sizeof_DCtx (const ZSTD_DCtx* dctx)
214 {
215 if (dctx==NULL) return 0; /* support sizeof NULL */
216 return sizeof(*dctx)
217 + ZSTD_sizeof_DDict(dctx->ddictLocal)
218 + dctx->inBuffSize + dctx->outBuffSize;
219 }
220
ZSTD_estimateDCtxSize(void)221 size_t ZSTD_estimateDCtxSize(void) { return sizeof(ZSTD_DCtx); }
222
223
ZSTD_startingInputLength(ZSTD_format_e format)224 static size_t ZSTD_startingInputLength(ZSTD_format_e format)
225 {
226 size_t const startingInputLength = ZSTD_FRAMEHEADERSIZE_PREFIX(format);
227 /* only supports formats ZSTD_f_zstd1 and ZSTD_f_zstd1_magicless */
228 assert( (format == ZSTD_f_zstd1) || (format == ZSTD_f_zstd1_magicless) );
229 return startingInputLength;
230 }
231
ZSTD_DCtx_resetParameters(ZSTD_DCtx * dctx)232 static void ZSTD_DCtx_resetParameters(ZSTD_DCtx* dctx)
233 {
234 assert(dctx->streamStage == zdss_init);
235 dctx->format = ZSTD_f_zstd1;
236 dctx->maxWindowSize = ZSTD_MAXWINDOWSIZE_DEFAULT;
237 dctx->outBufferMode = ZSTD_bm_buffered;
238 dctx->forceIgnoreChecksum = ZSTD_d_validateChecksum;
239 dctx->refMultipleDDicts = ZSTD_rmd_refSingleDDict;
240 }
241
ZSTD_initDCtx_internal(ZSTD_DCtx * dctx)242 static void ZSTD_initDCtx_internal(ZSTD_DCtx* dctx)
243 {
244 dctx->staticSize = 0;
245 dctx->ddict = NULL;
246 dctx->ddictLocal = NULL;
247 dctx->dictEnd = NULL;
248 dctx->ddictIsCold = 0;
249 dctx->dictUses = ZSTD_dont_use;
250 dctx->inBuff = NULL;
251 dctx->inBuffSize = 0;
252 dctx->outBuffSize = 0;
253 dctx->streamStage = zdss_init;
254 dctx->noForwardProgress = 0;
255 dctx->oversizedDuration = 0;
256 #if DYNAMIC_BMI2
257 dctx->bmi2 = ZSTD_cpuSupportsBmi2();
258 #endif
259 dctx->ddictSet = NULL;
260 ZSTD_DCtx_resetParameters(dctx);
261 #ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
262 dctx->dictContentEndForFuzzing = NULL;
263 #endif
264 }
265
ZSTD_initStaticDCtx(void * workspace,size_t workspaceSize)266 ZSTD_DCtx* ZSTD_initStaticDCtx(void *workspace, size_t workspaceSize)
267 {
268 ZSTD_DCtx* const dctx = (ZSTD_DCtx*) workspace;
269
270 if ((size_t)workspace & 7) return NULL; /* 8-aligned */
271 if (workspaceSize < sizeof(ZSTD_DCtx)) return NULL; /* minimum size */
272
273 ZSTD_initDCtx_internal(dctx);
274 dctx->staticSize = workspaceSize;
275 dctx->inBuff = (char*)(dctx+1);
276 return dctx;
277 }
278
ZSTD_createDCtx_internal(ZSTD_customMem customMem)279 static ZSTD_DCtx* ZSTD_createDCtx_internal(ZSTD_customMem customMem) {
280 if ((!customMem.customAlloc) ^ (!customMem.customFree)) return NULL;
281
282 { ZSTD_DCtx* const dctx = (ZSTD_DCtx*)ZSTD_customMalloc(sizeof(*dctx), customMem);
283 if (!dctx) return NULL;
284 dctx->customMem = customMem;
285 ZSTD_initDCtx_internal(dctx);
286 return dctx;
287 }
288 }
289
ZSTD_createDCtx_advanced(ZSTD_customMem customMem)290 ZSTD_DCtx* ZSTD_createDCtx_advanced(ZSTD_customMem customMem)
291 {
292 return ZSTD_createDCtx_internal(customMem);
293 }
294
ZSTD_createDCtx(void)295 ZSTD_DCtx* ZSTD_createDCtx(void)
296 {
297 DEBUGLOG(3, "ZSTD_createDCtx");
298 return ZSTD_createDCtx_internal(ZSTD_defaultCMem);
299 }
300
ZSTD_clearDict(ZSTD_DCtx * dctx)301 static void ZSTD_clearDict(ZSTD_DCtx* dctx)
302 {
303 ZSTD_freeDDict(dctx->ddictLocal);
304 dctx->ddictLocal = NULL;
305 dctx->ddict = NULL;
306 dctx->dictUses = ZSTD_dont_use;
307 }
308
ZSTD_freeDCtx(ZSTD_DCtx * dctx)309 size_t ZSTD_freeDCtx(ZSTD_DCtx* dctx)
310 {
311 if (dctx==NULL) return 0; /* support free on NULL */
312 RETURN_ERROR_IF(dctx->staticSize, memory_allocation, "not compatible with static DCtx");
313 { ZSTD_customMem const cMem = dctx->customMem;
314 ZSTD_clearDict(dctx);
315 ZSTD_customFree(dctx->inBuff, cMem);
316 dctx->inBuff = NULL;
317 if (dctx->ddictSet) {
318 ZSTD_freeDDictHashSet(dctx->ddictSet, cMem);
319 dctx->ddictSet = NULL;
320 }
321 ZSTD_customFree(dctx, cMem);
322 return 0;
323 }
324 }
325
326 /* no longer useful */
ZSTD_copyDCtx(ZSTD_DCtx * dstDCtx,const ZSTD_DCtx * srcDCtx)327 void ZSTD_copyDCtx(ZSTD_DCtx* dstDCtx, const ZSTD_DCtx* srcDCtx)
328 {
329 size_t const toCopy = (size_t)((char*)(&dstDCtx->inBuff) - (char*)dstDCtx);
330 ZSTD_memcpy(dstDCtx, srcDCtx, toCopy); /* no need to copy workspace */
331 }
332
333 /* Given a dctx with a digested frame params, re-selects the correct ZSTD_DDict based on
334 * the requested dict ID from the frame. If there exists a reference to the correct ZSTD_DDict, then
335 * accordingly sets the ddict to be used to decompress the frame.
336 *
337 * If no DDict is found, then no action is taken, and the ZSTD_DCtx::ddict remains as-is.
338 *
339 * ZSTD_d_refMultipleDDicts must be enabled for this function to be called.
340 */
ZSTD_DCtx_selectFrameDDict(ZSTD_DCtx * dctx)341 static void ZSTD_DCtx_selectFrameDDict(ZSTD_DCtx* dctx) {
342 assert(dctx->refMultipleDDicts && dctx->ddictSet);
343 DEBUGLOG(4, "Adjusting DDict based on requested dict ID from frame");
344 if (dctx->ddict) {
345 const ZSTD_DDict* frameDDict = ZSTD_DDictHashSet_getDDict(dctx->ddictSet, dctx->fParams.dictID);
346 if (frameDDict) {
347 DEBUGLOG(4, "DDict found!");
348 ZSTD_clearDict(dctx);
349 dctx->dictID = dctx->fParams.dictID;
350 dctx->ddict = frameDDict;
351 dctx->dictUses = ZSTD_use_indefinitely;
352 }
353 }
354 }
355
356
357 /*-*************************************************************
358 * Frame header decoding
359 ***************************************************************/
360
361 /*! ZSTD_isFrame() :
362 * Tells if the content of `buffer` starts with a valid Frame Identifier.
363 * Note : Frame Identifier is 4 bytes. If `size < 4`, @return will always be 0.
364 * Note 2 : Legacy Frame Identifiers are considered valid only if Legacy Support is enabled.
365 * Note 3 : Skippable Frame Identifiers are considered valid. */
ZSTD_isFrame(const void * buffer,size_t size)366 unsigned ZSTD_isFrame(const void* buffer, size_t size)
367 {
368 if (size < ZSTD_FRAMEIDSIZE) return 0;
369 { U32 const magic = MEM_readLE32(buffer);
370 if (magic == ZSTD_MAGICNUMBER) return 1;
371 if ((magic & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) return 1;
372 }
373 return 0;
374 }
375
376 /*! ZSTD_isSkippableFrame() :
377 * Tells if the content of `buffer` starts with a valid Frame Identifier for a skippable frame.
378 * Note : Frame Identifier is 4 bytes. If `size < 4`, @return will always be 0.
379 */
ZSTD_isSkippableFrame(const void * buffer,size_t size)380 unsigned ZSTD_isSkippableFrame(const void* buffer, size_t size)
381 {
382 if (size < ZSTD_FRAMEIDSIZE) return 0;
383 { U32 const magic = MEM_readLE32(buffer);
384 if ((magic & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) return 1;
385 }
386 return 0;
387 }
388
389 /* ZSTD_frameHeaderSize_internal() :
390 * srcSize must be large enough to reach header size fields.
391 * note : only works for formats ZSTD_f_zstd1 and ZSTD_f_zstd1_magicless.
392 * @return : size of the Frame Header
393 * or an error code, which can be tested with ZSTD_isError() */
ZSTD_frameHeaderSize_internal(const void * src,size_t srcSize,ZSTD_format_e format)394 static size_t ZSTD_frameHeaderSize_internal(const void* src, size_t srcSize, ZSTD_format_e format)
395 {
396 size_t const minInputSize = ZSTD_startingInputLength(format);
397 RETURN_ERROR_IF(srcSize < minInputSize, srcSize_wrong, "");
398
399 { BYTE const fhd = ((const BYTE*)src)[minInputSize-1];
400 U32 const dictID= fhd & 3;
401 U32 const singleSegment = (fhd >> 5) & 1;
402 U32 const fcsId = fhd >> 6;
403 return minInputSize + !singleSegment
404 + ZSTD_did_fieldSize[dictID] + ZSTD_fcs_fieldSize[fcsId]
405 + (singleSegment && !fcsId);
406 }
407 }
408
409 /* ZSTD_frameHeaderSize() :
410 * srcSize must be >= ZSTD_frameHeaderSize_prefix.
411 * @return : size of the Frame Header,
412 * or an error code (if srcSize is too small) */
ZSTD_frameHeaderSize(const void * src,size_t srcSize)413 size_t ZSTD_frameHeaderSize(const void* src, size_t srcSize)
414 {
415 return ZSTD_frameHeaderSize_internal(src, srcSize, ZSTD_f_zstd1);
416 }
417
418
419 /* ZSTD_getFrameHeader_advanced() :
420 * decode Frame Header, or require larger `srcSize`.
421 * note : only works for formats ZSTD_f_zstd1 and ZSTD_f_zstd1_magicless
422 * @return : 0, `zfhPtr` is correctly filled,
423 * >0, `srcSize` is too small, value is wanted `srcSize` amount,
424 * or an error code, which can be tested using ZSTD_isError() */
ZSTD_getFrameHeader_advanced(ZSTD_frameHeader * zfhPtr,const void * src,size_t srcSize,ZSTD_format_e format)425 size_t ZSTD_getFrameHeader_advanced(ZSTD_frameHeader* zfhPtr, const void* src, size_t srcSize, ZSTD_format_e format)
426 {
427 const BYTE* ip = (const BYTE*)src;
428 size_t const minInputSize = ZSTD_startingInputLength(format);
429
430 ZSTD_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 */
431 if (srcSize < minInputSize) return minInputSize;
432 RETURN_ERROR_IF(src==NULL, GENERIC, "invalid parameter");
433
434 if ( (format != ZSTD_f_zstd1_magicless)
435 && (MEM_readLE32(src) != ZSTD_MAGICNUMBER) ) {
436 if ((MEM_readLE32(src) & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) {
437 /* skippable frame */
438 if (srcSize < ZSTD_SKIPPABLEHEADERSIZE)
439 return ZSTD_SKIPPABLEHEADERSIZE; /* magic number + frame length */
440 ZSTD_memset(zfhPtr, 0, sizeof(*zfhPtr));
441 zfhPtr->frameContentSize = MEM_readLE32((const char *)src + ZSTD_FRAMEIDSIZE);
442 zfhPtr->frameType = ZSTD_skippableFrame;
443 return 0;
444 }
445 RETURN_ERROR(prefix_unknown, "");
446 }
447
448 /* ensure there is enough `srcSize` to fully read/decode frame header */
449 { size_t const fhsize = ZSTD_frameHeaderSize_internal(src, srcSize, format);
450 if (srcSize < fhsize) return fhsize;
451 zfhPtr->headerSize = (U32)fhsize;
452 }
453
454 { BYTE const fhdByte = ip[minInputSize-1];
455 size_t pos = minInputSize;
456 U32 const dictIDSizeCode = fhdByte&3;
457 U32 const checksumFlag = (fhdByte>>2)&1;
458 U32 const singleSegment = (fhdByte>>5)&1;
459 U32 const fcsID = fhdByte>>6;
460 U64 windowSize = 0;
461 U32 dictID = 0;
462 U64 frameContentSize = ZSTD_CONTENTSIZE_UNKNOWN;
463 RETURN_ERROR_IF((fhdByte & 0x08) != 0, frameParameter_unsupported,
464 "reserved bits, must be zero");
465
466 if (!singleSegment) {
467 BYTE const wlByte = ip[pos++];
468 U32 const windowLog = (wlByte >> 3) + ZSTD_WINDOWLOG_ABSOLUTEMIN;
469 RETURN_ERROR_IF(windowLog > ZSTD_WINDOWLOG_MAX, frameParameter_windowTooLarge, "");
470 windowSize = (1ULL << windowLog);
471 windowSize += (windowSize >> 3) * (wlByte&7);
472 }
473 switch(dictIDSizeCode)
474 {
475 default:
476 assert(0); /* impossible */
477 ZSTD_FALLTHROUGH;
478 case 0 : break;
479 case 1 : dictID = ip[pos]; pos++; break;
480 case 2 : dictID = MEM_readLE16(ip+pos); pos+=2; break;
481 case 3 : dictID = MEM_readLE32(ip+pos); pos+=4; break;
482 }
483 switch(fcsID)
484 {
485 default:
486 assert(0); /* impossible */
487 ZSTD_FALLTHROUGH;
488 case 0 : if (singleSegment) frameContentSize = ip[pos]; break;
489 case 1 : frameContentSize = MEM_readLE16(ip+pos)+256; break;
490 case 2 : frameContentSize = MEM_readLE32(ip+pos); break;
491 case 3 : frameContentSize = MEM_readLE64(ip+pos); break;
492 }
493 if (singleSegment) windowSize = frameContentSize;
494
495 zfhPtr->frameType = ZSTD_frame;
496 zfhPtr->frameContentSize = frameContentSize;
497 zfhPtr->windowSize = windowSize;
498 zfhPtr->blockSizeMax = (unsigned) MIN(windowSize, ZSTD_BLOCKSIZE_MAX);
499 zfhPtr->dictID = dictID;
500 zfhPtr->checksumFlag = checksumFlag;
501 }
502 return 0;
503 }
504
505 /* ZSTD_getFrameHeader() :
506 * decode Frame Header, or require larger `srcSize`.
507 * note : this function does not consume input, it only reads it.
508 * @return : 0, `zfhPtr` is correctly filled,
509 * >0, `srcSize` is too small, value is wanted `srcSize` amount,
510 * or an error code, which can be tested using ZSTD_isError() */
ZSTD_getFrameHeader(ZSTD_frameHeader * zfhPtr,const void * src,size_t srcSize)511 size_t ZSTD_getFrameHeader(ZSTD_frameHeader* zfhPtr, const void* src, size_t srcSize)
512 {
513 return ZSTD_getFrameHeader_advanced(zfhPtr, src, srcSize, ZSTD_f_zstd1);
514 }
515
516 /* ZSTD_getFrameContentSize() :
517 * compatible with legacy mode
518 * @return : decompressed size of the single frame pointed to be `src` if known, otherwise
519 * - ZSTD_CONTENTSIZE_UNKNOWN if the size cannot be determined
520 * - ZSTD_CONTENTSIZE_ERROR if an error occurred (e.g. invalid magic number, srcSize too small) */
ZSTD_getFrameContentSize(const void * src,size_t srcSize)521 unsigned long long ZSTD_getFrameContentSize(const void *src, size_t srcSize)
522 {
523 { ZSTD_frameHeader zfh;
524 if (ZSTD_getFrameHeader(&zfh, src, srcSize) != 0)
525 return ZSTD_CONTENTSIZE_ERROR;
526 if (zfh.frameType == ZSTD_skippableFrame) {
527 return 0;
528 } else {
529 return zfh.frameContentSize;
530 } }
531 }
532
readSkippableFrameSize(void const * src,size_t srcSize)533 static size_t readSkippableFrameSize(void const* src, size_t srcSize)
534 {
535 size_t const skippableHeaderSize = ZSTD_SKIPPABLEHEADERSIZE;
536 U32 sizeU32;
537
538 RETURN_ERROR_IF(srcSize < ZSTD_SKIPPABLEHEADERSIZE, srcSize_wrong, "");
539
540 sizeU32 = MEM_readLE32((BYTE const*)src + ZSTD_FRAMEIDSIZE);
541 RETURN_ERROR_IF((U32)(sizeU32 + ZSTD_SKIPPABLEHEADERSIZE) < sizeU32,
542 frameParameter_unsupported, "");
543 {
544 size_t const skippableSize = skippableHeaderSize + sizeU32;
545 RETURN_ERROR_IF(skippableSize > srcSize, srcSize_wrong, "");
546 return skippableSize;
547 }
548 }
549
550 /*! ZSTD_readSkippableFrame() :
551 * Retrieves a zstd skippable frame containing data given by src, and writes it to dst buffer.
552 *
553 * The parameter magicVariant will receive the magicVariant that was supplied when the frame was written,
554 * i.e. magicNumber - ZSTD_MAGIC_SKIPPABLE_START. This can be NULL if the caller is not interested
555 * in the magicVariant.
556 *
557 * Returns an error if destination buffer is not large enough, or if the frame is not skippable.
558 *
559 * @return : number of bytes written or a ZSTD error.
560 */
ZSTD_readSkippableFrame(void * dst,size_t dstCapacity,unsigned * magicVariant,const void * src,size_t srcSize)561 ZSTDLIB_API size_t ZSTD_readSkippableFrame(void* dst, size_t dstCapacity, unsigned* magicVariant,
562 const void* src, size_t srcSize)
563 {
564 U32 const magicNumber = MEM_readLE32(src);
565 size_t skippableFrameSize = readSkippableFrameSize(src, srcSize);
566 size_t skippableContentSize = skippableFrameSize - ZSTD_SKIPPABLEHEADERSIZE;
567
568 /* check input validity */
569 RETURN_ERROR_IF(!ZSTD_isSkippableFrame(src, srcSize), frameParameter_unsupported, "");
570 RETURN_ERROR_IF(skippableFrameSize < ZSTD_SKIPPABLEHEADERSIZE || skippableFrameSize > srcSize, srcSize_wrong, "");
571 RETURN_ERROR_IF(skippableContentSize > dstCapacity, dstSize_tooSmall, "");
572
573 /* deliver payload */
574 if (skippableContentSize > 0 && dst != NULL)
575 ZSTD_memcpy(dst, (const BYTE *)src + ZSTD_SKIPPABLEHEADERSIZE, skippableContentSize);
576 if (magicVariant != NULL)
577 *magicVariant = magicNumber - ZSTD_MAGIC_SKIPPABLE_START;
578 return skippableContentSize;
579 }
580
581 /* ZSTD_findDecompressedSize() :
582 * compatible with legacy mode
583 * `srcSize` must be the exact length of some number of ZSTD compressed and/or
584 * skippable frames
585 * @return : decompressed size of the frames contained */
ZSTD_findDecompressedSize(const void * src,size_t srcSize)586 unsigned long long ZSTD_findDecompressedSize(const void* src, size_t srcSize)
587 {
588 unsigned long long totalDstSize = 0;
589
590 while (srcSize >= ZSTD_startingInputLength(ZSTD_f_zstd1)) {
591 U32 const magicNumber = MEM_readLE32(src);
592
593 if ((magicNumber & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) {
594 size_t const skippableSize = readSkippableFrameSize(src, srcSize);
595 if (ZSTD_isError(skippableSize)) {
596 return ZSTD_CONTENTSIZE_ERROR;
597 }
598 assert(skippableSize <= srcSize);
599
600 src = (const BYTE *)src + skippableSize;
601 srcSize -= skippableSize;
602 continue;
603 }
604
605 { unsigned long long const ret = ZSTD_getFrameContentSize(src, srcSize);
606 if (ret >= ZSTD_CONTENTSIZE_ERROR) return ret;
607
608 /* check for overflow */
609 if (totalDstSize + ret < totalDstSize) return ZSTD_CONTENTSIZE_ERROR;
610 totalDstSize += ret;
611 }
612 { size_t const frameSrcSize = ZSTD_findFrameCompressedSize(src, srcSize);
613 if (ZSTD_isError(frameSrcSize)) {
614 return ZSTD_CONTENTSIZE_ERROR;
615 }
616
617 src = (const BYTE *)src + frameSrcSize;
618 srcSize -= frameSrcSize;
619 }
620 } /* while (srcSize >= ZSTD_frameHeaderSize_prefix) */
621
622 if (srcSize) return ZSTD_CONTENTSIZE_ERROR;
623
624 return totalDstSize;
625 }
626
627 /* ZSTD_getDecompressedSize() :
628 * compatible with legacy mode
629 * @return : decompressed size if known, 0 otherwise
630 note : 0 can mean any of the following :
631 - frame content is empty
632 - decompressed size field is not present in frame header
633 - frame header unknown / not supported
634 - frame header not complete (`srcSize` too small) */
ZSTD_getDecompressedSize(const void * src,size_t srcSize)635 unsigned long long ZSTD_getDecompressedSize(const void* src, size_t srcSize)
636 {
637 unsigned long long const ret = ZSTD_getFrameContentSize(src, srcSize);
638 ZSTD_STATIC_ASSERT(ZSTD_CONTENTSIZE_ERROR < ZSTD_CONTENTSIZE_UNKNOWN);
639 return (ret >= ZSTD_CONTENTSIZE_ERROR) ? 0 : ret;
640 }
641
642
643 /* ZSTD_decodeFrameHeader() :
644 * `headerSize` must be the size provided by ZSTD_frameHeaderSize().
645 * If multiple DDict references are enabled, also will choose the correct DDict to use.
646 * @return : 0 if success, or an error code, which can be tested using ZSTD_isError() */
ZSTD_decodeFrameHeader(ZSTD_DCtx * dctx,const void * src,size_t headerSize)647 static size_t ZSTD_decodeFrameHeader(ZSTD_DCtx* dctx, const void* src, size_t headerSize)
648 {
649 size_t const result = ZSTD_getFrameHeader_advanced(&(dctx->fParams), src, headerSize, dctx->format);
650 if (ZSTD_isError(result)) return result; /* invalid header */
651 RETURN_ERROR_IF(result>0, srcSize_wrong, "headerSize too small");
652
653 /* Reference DDict requested by frame if dctx references multiple ddicts */
654 if (dctx->refMultipleDDicts == ZSTD_rmd_refMultipleDDicts && dctx->ddictSet) {
655 ZSTD_DCtx_selectFrameDDict(dctx);
656 }
657
658 #ifndef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
659 /* Skip the dictID check in fuzzing mode, because it makes the search
660 * harder.
661 */
662 RETURN_ERROR_IF(dctx->fParams.dictID && (dctx->dictID != dctx->fParams.dictID),
663 dictionary_wrong, "");
664 #endif
665 dctx->validateChecksum = (dctx->fParams.checksumFlag && !dctx->forceIgnoreChecksum) ? 1 : 0;
666 if (dctx->validateChecksum) xxh64_reset(&dctx->xxhState, 0);
667 dctx->processedCSize += headerSize;
668 return 0;
669 }
670
ZSTD_errorFrameSizeInfo(size_t ret)671 static ZSTD_frameSizeInfo ZSTD_errorFrameSizeInfo(size_t ret)
672 {
673 ZSTD_frameSizeInfo frameSizeInfo;
674 frameSizeInfo.compressedSize = ret;
675 frameSizeInfo.decompressedBound = ZSTD_CONTENTSIZE_ERROR;
676 return frameSizeInfo;
677 }
678
ZSTD_findFrameSizeInfo(const void * src,size_t srcSize)679 static ZSTD_frameSizeInfo ZSTD_findFrameSizeInfo(const void* src, size_t srcSize)
680 {
681 ZSTD_frameSizeInfo frameSizeInfo;
682 ZSTD_memset(&frameSizeInfo, 0, sizeof(ZSTD_frameSizeInfo));
683
684
685 if ((srcSize >= ZSTD_SKIPPABLEHEADERSIZE)
686 && (MEM_readLE32(src) & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) {
687 frameSizeInfo.compressedSize = readSkippableFrameSize(src, srcSize);
688 assert(ZSTD_isError(frameSizeInfo.compressedSize) ||
689 frameSizeInfo.compressedSize <= srcSize);
690 return frameSizeInfo;
691 } else {
692 const BYTE* ip = (const BYTE*)src;
693 const BYTE* const ipstart = ip;
694 size_t remainingSize = srcSize;
695 size_t nbBlocks = 0;
696 ZSTD_frameHeader zfh;
697
698 /* Extract Frame Header */
699 { size_t const ret = ZSTD_getFrameHeader(&zfh, src, srcSize);
700 if (ZSTD_isError(ret))
701 return ZSTD_errorFrameSizeInfo(ret);
702 if (ret > 0)
703 return ZSTD_errorFrameSizeInfo(ERROR(srcSize_wrong));
704 }
705
706 ip += zfh.headerSize;
707 remainingSize -= zfh.headerSize;
708
709 /* Iterate over each block */
710 while (1) {
711 blockProperties_t blockProperties;
712 size_t const cBlockSize = ZSTD_getcBlockSize(ip, remainingSize, &blockProperties);
713 if (ZSTD_isError(cBlockSize))
714 return ZSTD_errorFrameSizeInfo(cBlockSize);
715
716 if (ZSTD_blockHeaderSize + cBlockSize > remainingSize)
717 return ZSTD_errorFrameSizeInfo(ERROR(srcSize_wrong));
718
719 ip += ZSTD_blockHeaderSize + cBlockSize;
720 remainingSize -= ZSTD_blockHeaderSize + cBlockSize;
721 nbBlocks++;
722
723 if (blockProperties.lastBlock) break;
724 }
725
726 /* Final frame content checksum */
727 if (zfh.checksumFlag) {
728 if (remainingSize < 4)
729 return ZSTD_errorFrameSizeInfo(ERROR(srcSize_wrong));
730 ip += 4;
731 }
732
733 frameSizeInfo.compressedSize = (size_t)(ip - ipstart);
734 frameSizeInfo.decompressedBound = (zfh.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN)
735 ? zfh.frameContentSize
736 : nbBlocks * zfh.blockSizeMax;
737 return frameSizeInfo;
738 }
739 }
740
741 /* ZSTD_findFrameCompressedSize() :
742 * compatible with legacy mode
743 * `src` must point to the start of a ZSTD frame, ZSTD legacy frame, or skippable frame
744 * `srcSize` must be at least as large as the frame contained
745 * @return : the compressed size of the frame starting at `src` */
ZSTD_findFrameCompressedSize(const void * src,size_t srcSize)746 size_t ZSTD_findFrameCompressedSize(const void *src, size_t srcSize)
747 {
748 ZSTD_frameSizeInfo const frameSizeInfo = ZSTD_findFrameSizeInfo(src, srcSize);
749 return frameSizeInfo.compressedSize;
750 }
751
752 /* ZSTD_decompressBound() :
753 * compatible with legacy mode
754 * `src` must point to the start of a ZSTD frame or a skippeable frame
755 * `srcSize` must be at least as large as the frame contained
756 * @return : the maximum decompressed size of the compressed source
757 */
ZSTD_decompressBound(const void * src,size_t srcSize)758 unsigned long long ZSTD_decompressBound(const void* src, size_t srcSize)
759 {
760 unsigned long long bound = 0;
761 /* Iterate over each frame */
762 while (srcSize > 0) {
763 ZSTD_frameSizeInfo const frameSizeInfo = ZSTD_findFrameSizeInfo(src, srcSize);
764 size_t const compressedSize = frameSizeInfo.compressedSize;
765 unsigned long long const decompressedBound = frameSizeInfo.decompressedBound;
766 if (ZSTD_isError(compressedSize) || decompressedBound == ZSTD_CONTENTSIZE_ERROR)
767 return ZSTD_CONTENTSIZE_ERROR;
768 assert(srcSize >= compressedSize);
769 src = (const BYTE*)src + compressedSize;
770 srcSize -= compressedSize;
771 bound += decompressedBound;
772 }
773 return bound;
774 }
775
776
777 /*-*************************************************************
778 * Frame decoding
779 ***************************************************************/
780
781 /* ZSTD_insertBlock() :
782 * insert `src` block into `dctx` history. Useful to track uncompressed blocks. */
ZSTD_insertBlock(ZSTD_DCtx * dctx,const void * blockStart,size_t blockSize)783 size_t ZSTD_insertBlock(ZSTD_DCtx* dctx, const void* blockStart, size_t blockSize)
784 {
785 DEBUGLOG(5, "ZSTD_insertBlock: %u bytes", (unsigned)blockSize);
786 ZSTD_checkContinuity(dctx, blockStart, blockSize);
787 dctx->previousDstEnd = (const char*)blockStart + blockSize;
788 return blockSize;
789 }
790
791
ZSTD_copyRawBlock(void * dst,size_t dstCapacity,const void * src,size_t srcSize)792 static size_t ZSTD_copyRawBlock(void* dst, size_t dstCapacity,
793 const void* src, size_t srcSize)
794 {
795 DEBUGLOG(5, "ZSTD_copyRawBlock");
796 RETURN_ERROR_IF(srcSize > dstCapacity, dstSize_tooSmall, "");
797 if (dst == NULL) {
798 if (srcSize == 0) return 0;
799 RETURN_ERROR(dstBuffer_null, "");
800 }
801 ZSTD_memmove(dst, src, srcSize);
802 return srcSize;
803 }
804
ZSTD_setRleBlock(void * dst,size_t dstCapacity,BYTE b,size_t regenSize)805 static size_t ZSTD_setRleBlock(void* dst, size_t dstCapacity,
806 BYTE b,
807 size_t regenSize)
808 {
809 RETURN_ERROR_IF(regenSize > dstCapacity, dstSize_tooSmall, "");
810 if (dst == NULL) {
811 if (regenSize == 0) return 0;
812 RETURN_ERROR(dstBuffer_null, "");
813 }
814 ZSTD_memset(dst, b, regenSize);
815 return regenSize;
816 }
817
ZSTD_DCtx_trace_end(ZSTD_DCtx const * dctx,U64 uncompressedSize,U64 compressedSize,unsigned streaming)818 static void ZSTD_DCtx_trace_end(ZSTD_DCtx const* dctx, U64 uncompressedSize, U64 compressedSize, unsigned streaming)
819 {
820 (void)dctx;
821 (void)uncompressedSize;
822 (void)compressedSize;
823 (void)streaming;
824 }
825
826
827 /*! ZSTD_decompressFrame() :
828 * @dctx must be properly initialized
829 * will update *srcPtr and *srcSizePtr,
830 * to make *srcPtr progress by one frame. */
ZSTD_decompressFrame(ZSTD_DCtx * dctx,void * dst,size_t dstCapacity,const void ** srcPtr,size_t * srcSizePtr)831 static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx,
832 void* dst, size_t dstCapacity,
833 const void** srcPtr, size_t *srcSizePtr)
834 {
835 const BYTE* const istart = (const BYTE*)(*srcPtr);
836 const BYTE* ip = istart;
837 BYTE* const ostart = (BYTE*)dst;
838 BYTE* const oend = dstCapacity != 0 ? ostart + dstCapacity : ostart;
839 BYTE* op = ostart;
840 size_t remainingSrcSize = *srcSizePtr;
841
842 DEBUGLOG(4, "ZSTD_decompressFrame (srcSize:%i)", (int)*srcSizePtr);
843
844 /* check */
845 RETURN_ERROR_IF(
846 remainingSrcSize < ZSTD_FRAMEHEADERSIZE_MIN(dctx->format)+ZSTD_blockHeaderSize,
847 srcSize_wrong, "");
848
849 /* Frame Header */
850 { size_t const frameHeaderSize = ZSTD_frameHeaderSize_internal(
851 ip, ZSTD_FRAMEHEADERSIZE_PREFIX(dctx->format), dctx->format);
852 if (ZSTD_isError(frameHeaderSize)) return frameHeaderSize;
853 RETURN_ERROR_IF(remainingSrcSize < frameHeaderSize+ZSTD_blockHeaderSize,
854 srcSize_wrong, "");
855 FORWARD_IF_ERROR( ZSTD_decodeFrameHeader(dctx, ip, frameHeaderSize) , "");
856 ip += frameHeaderSize; remainingSrcSize -= frameHeaderSize;
857 }
858
859 /* Loop on each block */
860 while (1) {
861 BYTE* oBlockEnd = oend;
862 size_t decodedSize;
863 blockProperties_t blockProperties;
864 size_t const cBlockSize = ZSTD_getcBlockSize(ip, remainingSrcSize, &blockProperties);
865 if (ZSTD_isError(cBlockSize)) return cBlockSize;
866
867 ip += ZSTD_blockHeaderSize;
868 remainingSrcSize -= ZSTD_blockHeaderSize;
869 RETURN_ERROR_IF(cBlockSize > remainingSrcSize, srcSize_wrong, "");
870
871 if (ip >= op && ip < oBlockEnd) {
872 /* We are decompressing in-place. Limit the output pointer so that we
873 * don't overwrite the block that we are currently reading. This will
874 * fail decompression if the input & output pointers aren't spaced
875 * far enough apart.
876 *
877 * This is important to set, even when the pointers are far enough
878 * apart, because ZSTD_decompressBlock_internal() can decide to store
879 * literals in the output buffer, after the block it is decompressing.
880 * Since we don't want anything to overwrite our input, we have to tell
881 * ZSTD_decompressBlock_internal to never write past ip.
882 *
883 * See ZSTD_allocateLiteralsBuffer() for reference.
884 */
885 oBlockEnd = op + (ip - op);
886 }
887
888 switch(blockProperties.blockType)
889 {
890 case bt_compressed:
891 decodedSize = ZSTD_decompressBlock_internal(dctx, op, (size_t)(oBlockEnd-op), ip, cBlockSize, /* frame */ 1, not_streaming);
892 break;
893 case bt_raw :
894 /* Use oend instead of oBlockEnd because this function is safe to overlap. It uses memmove. */
895 decodedSize = ZSTD_copyRawBlock(op, (size_t)(oend-op), ip, cBlockSize);
896 break;
897 case bt_rle :
898 decodedSize = ZSTD_setRleBlock(op, (size_t)(oBlockEnd-op), *ip, blockProperties.origSize);
899 break;
900 case bt_reserved :
901 default:
902 RETURN_ERROR(corruption_detected, "invalid block type");
903 }
904
905 if (ZSTD_isError(decodedSize)) return decodedSize;
906 if (dctx->validateChecksum)
907 xxh64_update(&dctx->xxhState, op, decodedSize);
908 if (decodedSize != 0)
909 op += decodedSize;
910 assert(ip != NULL);
911 ip += cBlockSize;
912 remainingSrcSize -= cBlockSize;
913 if (blockProperties.lastBlock) break;
914 }
915
916 if (dctx->fParams.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN) {
917 RETURN_ERROR_IF((U64)(op-ostart) != dctx->fParams.frameContentSize,
918 corruption_detected, "");
919 }
920 if (dctx->fParams.checksumFlag) { /* Frame content checksum verification */
921 RETURN_ERROR_IF(remainingSrcSize<4, checksum_wrong, "");
922 if (!dctx->forceIgnoreChecksum) {
923 U32 const checkCalc = (U32)xxh64_digest(&dctx->xxhState);
924 U32 checkRead;
925 checkRead = MEM_readLE32(ip);
926 RETURN_ERROR_IF(checkRead != checkCalc, checksum_wrong, "");
927 }
928 ip += 4;
929 remainingSrcSize -= 4;
930 }
931 ZSTD_DCtx_trace_end(dctx, (U64)(op-ostart), (U64)(ip-istart), /* streaming */ 0);
932 /* Allow caller to get size read */
933 *srcPtr = ip;
934 *srcSizePtr = remainingSrcSize;
935 return (size_t)(op-ostart);
936 }
937
ZSTD_decompressMultiFrame(ZSTD_DCtx * dctx,void * dst,size_t dstCapacity,const void * src,size_t srcSize,const void * dict,size_t dictSize,const ZSTD_DDict * ddict)938 static size_t ZSTD_decompressMultiFrame(ZSTD_DCtx* dctx,
939 void* dst, size_t dstCapacity,
940 const void* src, size_t srcSize,
941 const void* dict, size_t dictSize,
942 const ZSTD_DDict* ddict)
943 {
944 void* const dststart = dst;
945 int moreThan1Frame = 0;
946
947 DEBUGLOG(5, "ZSTD_decompressMultiFrame");
948 assert(dict==NULL || ddict==NULL); /* either dict or ddict set, not both */
949
950 if (ddict) {
951 dict = ZSTD_DDict_dictContent(ddict);
952 dictSize = ZSTD_DDict_dictSize(ddict);
953 }
954
955 while (srcSize >= ZSTD_startingInputLength(dctx->format)) {
956
957
958 { U32 const magicNumber = MEM_readLE32(src);
959 DEBUGLOG(4, "reading magic number %08X (expecting %08X)",
960 (unsigned)magicNumber, ZSTD_MAGICNUMBER);
961 if ((magicNumber & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) {
962 size_t const skippableSize = readSkippableFrameSize(src, srcSize);
963 FORWARD_IF_ERROR(skippableSize, "readSkippableFrameSize failed");
964 assert(skippableSize <= srcSize);
965
966 src = (const BYTE *)src + skippableSize;
967 srcSize -= skippableSize;
968 continue;
969 } }
970
971 if (ddict) {
972 /* we were called from ZSTD_decompress_usingDDict */
973 FORWARD_IF_ERROR(ZSTD_decompressBegin_usingDDict(dctx, ddict), "");
974 } else {
975 /* this will initialize correctly with no dict if dict == NULL, so
976 * use this in all cases but ddict */
977 FORWARD_IF_ERROR(ZSTD_decompressBegin_usingDict(dctx, dict, dictSize), "");
978 }
979 ZSTD_checkContinuity(dctx, dst, dstCapacity);
980
981 { const size_t res = ZSTD_decompressFrame(dctx, dst, dstCapacity,
982 &src, &srcSize);
983 RETURN_ERROR_IF(
984 (ZSTD_getErrorCode(res) == ZSTD_error_prefix_unknown)
985 && (moreThan1Frame==1),
986 srcSize_wrong,
987 "At least one frame successfully completed, "
988 "but following bytes are garbage: "
989 "it's more likely to be a srcSize error, "
990 "specifying more input bytes than size of frame(s). "
991 "Note: one could be unlucky, it might be a corruption error instead, "
992 "happening right at the place where we expect zstd magic bytes. "
993 "But this is _much_ less likely than a srcSize field error.");
994 if (ZSTD_isError(res)) return res;
995 assert(res <= dstCapacity);
996 if (res != 0)
997 dst = (BYTE*)dst + res;
998 dstCapacity -= res;
999 }
1000 moreThan1Frame = 1;
1001 } /* while (srcSize >= ZSTD_frameHeaderSize_prefix) */
1002
1003 RETURN_ERROR_IF(srcSize, srcSize_wrong, "input not entirely consumed");
1004
1005 return (size_t)((BYTE*)dst - (BYTE*)dststart);
1006 }
1007
ZSTD_decompress_usingDict(ZSTD_DCtx * dctx,void * dst,size_t dstCapacity,const void * src,size_t srcSize,const void * dict,size_t dictSize)1008 size_t ZSTD_decompress_usingDict(ZSTD_DCtx* dctx,
1009 void* dst, size_t dstCapacity,
1010 const void* src, size_t srcSize,
1011 const void* dict, size_t dictSize)
1012 {
1013 return ZSTD_decompressMultiFrame(dctx, dst, dstCapacity, src, srcSize, dict, dictSize, NULL);
1014 }
1015
1016
ZSTD_getDDict(ZSTD_DCtx * dctx)1017 static ZSTD_DDict const* ZSTD_getDDict(ZSTD_DCtx* dctx)
1018 {
1019 switch (dctx->dictUses) {
1020 default:
1021 assert(0 /* Impossible */);
1022 ZSTD_FALLTHROUGH;
1023 case ZSTD_dont_use:
1024 ZSTD_clearDict(dctx);
1025 return NULL;
1026 case ZSTD_use_indefinitely:
1027 return dctx->ddict;
1028 case ZSTD_use_once:
1029 dctx->dictUses = ZSTD_dont_use;
1030 return dctx->ddict;
1031 }
1032 }
1033
ZSTD_decompressDCtx(ZSTD_DCtx * dctx,void * dst,size_t dstCapacity,const void * src,size_t srcSize)1034 size_t ZSTD_decompressDCtx(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize)
1035 {
1036 return ZSTD_decompress_usingDDict(dctx, dst, dstCapacity, src, srcSize, ZSTD_getDDict(dctx));
1037 }
1038
1039
ZSTD_decompress(void * dst,size_t dstCapacity,const void * src,size_t srcSize)1040 size_t ZSTD_decompress(void* dst, size_t dstCapacity, const void* src, size_t srcSize)
1041 {
1042 #if defined(ZSTD_HEAPMODE) && (ZSTD_HEAPMODE>=1)
1043 size_t regenSize;
1044 ZSTD_DCtx* const dctx = ZSTD_createDCtx_internal(ZSTD_defaultCMem);
1045 RETURN_ERROR_IF(dctx==NULL, memory_allocation, "NULL pointer!");
1046 regenSize = ZSTD_decompressDCtx(dctx, dst, dstCapacity, src, srcSize);
1047 ZSTD_freeDCtx(dctx);
1048 return regenSize;
1049 #else /* stack mode */
1050 ZSTD_DCtx dctx;
1051 ZSTD_initDCtx_internal(&dctx);
1052 return ZSTD_decompressDCtx(&dctx, dst, dstCapacity, src, srcSize);
1053 #endif
1054 }
1055
1056
1057 /*-**************************************
1058 * Advanced Streaming Decompression API
1059 * Bufferless and synchronous
1060 ****************************************/
ZSTD_nextSrcSizeToDecompress(ZSTD_DCtx * dctx)1061 size_t ZSTD_nextSrcSizeToDecompress(ZSTD_DCtx* dctx) { return dctx->expected; }
1062
1063 /*
1064 * Similar to ZSTD_nextSrcSizeToDecompress(), but when a block input can be streamed,
1065 * we allow taking a partial block as the input. Currently only raw uncompressed blocks can
1066 * be streamed.
1067 *
1068 * For blocks that can be streamed, this allows us to reduce the latency until we produce
1069 * output, and avoid copying the input.
1070 *
1071 * @param inputSize - The total amount of input that the caller currently has.
1072 */
ZSTD_nextSrcSizeToDecompressWithInputSize(ZSTD_DCtx * dctx,size_t inputSize)1073 static size_t ZSTD_nextSrcSizeToDecompressWithInputSize(ZSTD_DCtx* dctx, size_t inputSize) {
1074 if (!(dctx->stage == ZSTDds_decompressBlock || dctx->stage == ZSTDds_decompressLastBlock))
1075 return dctx->expected;
1076 if (dctx->bType != bt_raw)
1077 return dctx->expected;
1078 return BOUNDED(1, inputSize, dctx->expected);
1079 }
1080
ZSTD_nextInputType(ZSTD_DCtx * dctx)1081 ZSTD_nextInputType_e ZSTD_nextInputType(ZSTD_DCtx* dctx) {
1082 switch(dctx->stage)
1083 {
1084 default: /* should not happen */
1085 assert(0);
1086 ZSTD_FALLTHROUGH;
1087 case ZSTDds_getFrameHeaderSize:
1088 ZSTD_FALLTHROUGH;
1089 case ZSTDds_decodeFrameHeader:
1090 return ZSTDnit_frameHeader;
1091 case ZSTDds_decodeBlockHeader:
1092 return ZSTDnit_blockHeader;
1093 case ZSTDds_decompressBlock:
1094 return ZSTDnit_block;
1095 case ZSTDds_decompressLastBlock:
1096 return ZSTDnit_lastBlock;
1097 case ZSTDds_checkChecksum:
1098 return ZSTDnit_checksum;
1099 case ZSTDds_decodeSkippableHeader:
1100 ZSTD_FALLTHROUGH;
1101 case ZSTDds_skipFrame:
1102 return ZSTDnit_skippableFrame;
1103 }
1104 }
1105
ZSTD_isSkipFrame(ZSTD_DCtx * dctx)1106 static int ZSTD_isSkipFrame(ZSTD_DCtx* dctx) { return dctx->stage == ZSTDds_skipFrame; }
1107
1108 /* ZSTD_decompressContinue() :
1109 * srcSize : must be the exact nb of bytes expected (see ZSTD_nextSrcSizeToDecompress())
1110 * @return : nb of bytes generated into `dst` (necessarily <= `dstCapacity)
1111 * or an error code, which can be tested using ZSTD_isError() */
ZSTD_decompressContinue(ZSTD_DCtx * dctx,void * dst,size_t dstCapacity,const void * src,size_t srcSize)1112 size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize)
1113 {
1114 DEBUGLOG(5, "ZSTD_decompressContinue (srcSize:%u)", (unsigned)srcSize);
1115 /* Sanity check */
1116 RETURN_ERROR_IF(srcSize != ZSTD_nextSrcSizeToDecompressWithInputSize(dctx, srcSize), srcSize_wrong, "not allowed");
1117 ZSTD_checkContinuity(dctx, dst, dstCapacity);
1118
1119 dctx->processedCSize += srcSize;
1120
1121 switch (dctx->stage)
1122 {
1123 case ZSTDds_getFrameHeaderSize :
1124 assert(src != NULL);
1125 if (dctx->format == ZSTD_f_zstd1) { /* allows header */
1126 assert(srcSize >= ZSTD_FRAMEIDSIZE); /* to read skippable magic number */
1127 if ((MEM_readLE32(src) & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) { /* skippable frame */
1128 ZSTD_memcpy(dctx->headerBuffer, src, srcSize);
1129 dctx->expected = ZSTD_SKIPPABLEHEADERSIZE - srcSize; /* remaining to load to get full skippable frame header */
1130 dctx->stage = ZSTDds_decodeSkippableHeader;
1131 return 0;
1132 } }
1133 dctx->headerSize = ZSTD_frameHeaderSize_internal(src, srcSize, dctx->format);
1134 if (ZSTD_isError(dctx->headerSize)) return dctx->headerSize;
1135 ZSTD_memcpy(dctx->headerBuffer, src, srcSize);
1136 dctx->expected = dctx->headerSize - srcSize;
1137 dctx->stage = ZSTDds_decodeFrameHeader;
1138 return 0;
1139
1140 case ZSTDds_decodeFrameHeader:
1141 assert(src != NULL);
1142 ZSTD_memcpy(dctx->headerBuffer + (dctx->headerSize - srcSize), src, srcSize);
1143 FORWARD_IF_ERROR(ZSTD_decodeFrameHeader(dctx, dctx->headerBuffer, dctx->headerSize), "");
1144 dctx->expected = ZSTD_blockHeaderSize;
1145 dctx->stage = ZSTDds_decodeBlockHeader;
1146 return 0;
1147
1148 case ZSTDds_decodeBlockHeader:
1149 { blockProperties_t bp;
1150 size_t const cBlockSize = ZSTD_getcBlockSize(src, ZSTD_blockHeaderSize, &bp);
1151 if (ZSTD_isError(cBlockSize)) return cBlockSize;
1152 RETURN_ERROR_IF(cBlockSize > dctx->fParams.blockSizeMax, corruption_detected, "Block Size Exceeds Maximum");
1153 dctx->expected = cBlockSize;
1154 dctx->bType = bp.blockType;
1155 dctx->rleSize = bp.origSize;
1156 if (cBlockSize) {
1157 dctx->stage = bp.lastBlock ? ZSTDds_decompressLastBlock : ZSTDds_decompressBlock;
1158 return 0;
1159 }
1160 /* empty block */
1161 if (bp.lastBlock) {
1162 if (dctx->fParams.checksumFlag) {
1163 dctx->expected = 4;
1164 dctx->stage = ZSTDds_checkChecksum;
1165 } else {
1166 dctx->expected = 0; /* end of frame */
1167 dctx->stage = ZSTDds_getFrameHeaderSize;
1168 }
1169 } else {
1170 dctx->expected = ZSTD_blockHeaderSize; /* jump to next header */
1171 dctx->stage = ZSTDds_decodeBlockHeader;
1172 }
1173 return 0;
1174 }
1175
1176 case ZSTDds_decompressLastBlock:
1177 case ZSTDds_decompressBlock:
1178 DEBUGLOG(5, "ZSTD_decompressContinue: case ZSTDds_decompressBlock");
1179 { size_t rSize;
1180 switch(dctx->bType)
1181 {
1182 case bt_compressed:
1183 DEBUGLOG(5, "ZSTD_decompressContinue: case bt_compressed");
1184 rSize = ZSTD_decompressBlock_internal(dctx, dst, dstCapacity, src, srcSize, /* frame */ 1, is_streaming);
1185 dctx->expected = 0; /* Streaming not supported */
1186 break;
1187 case bt_raw :
1188 assert(srcSize <= dctx->expected);
1189 rSize = ZSTD_copyRawBlock(dst, dstCapacity, src, srcSize);
1190 FORWARD_IF_ERROR(rSize, "ZSTD_copyRawBlock failed");
1191 assert(rSize == srcSize);
1192 dctx->expected -= rSize;
1193 break;
1194 case bt_rle :
1195 rSize = ZSTD_setRleBlock(dst, dstCapacity, *(const BYTE*)src, dctx->rleSize);
1196 dctx->expected = 0; /* Streaming not supported */
1197 break;
1198 case bt_reserved : /* should never happen */
1199 default:
1200 RETURN_ERROR(corruption_detected, "invalid block type");
1201 }
1202 FORWARD_IF_ERROR(rSize, "");
1203 RETURN_ERROR_IF(rSize > dctx->fParams.blockSizeMax, corruption_detected, "Decompressed Block Size Exceeds Maximum");
1204 DEBUGLOG(5, "ZSTD_decompressContinue: decoded size from block : %u", (unsigned)rSize);
1205 dctx->decodedSize += rSize;
1206 if (dctx->validateChecksum) xxh64_update(&dctx->xxhState, dst, rSize);
1207 dctx->previousDstEnd = (char*)dst + rSize;
1208
1209 /* Stay on the same stage until we are finished streaming the block. */
1210 if (dctx->expected > 0) {
1211 return rSize;
1212 }
1213
1214 if (dctx->stage == ZSTDds_decompressLastBlock) { /* end of frame */
1215 DEBUGLOG(4, "ZSTD_decompressContinue: decoded size from frame : %u", (unsigned)dctx->decodedSize);
1216 RETURN_ERROR_IF(
1217 dctx->fParams.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN
1218 && dctx->decodedSize != dctx->fParams.frameContentSize,
1219 corruption_detected, "");
1220 if (dctx->fParams.checksumFlag) { /* another round for frame checksum */
1221 dctx->expected = 4;
1222 dctx->stage = ZSTDds_checkChecksum;
1223 } else {
1224 ZSTD_DCtx_trace_end(dctx, dctx->decodedSize, dctx->processedCSize, /* streaming */ 1);
1225 dctx->expected = 0; /* ends here */
1226 dctx->stage = ZSTDds_getFrameHeaderSize;
1227 }
1228 } else {
1229 dctx->stage = ZSTDds_decodeBlockHeader;
1230 dctx->expected = ZSTD_blockHeaderSize;
1231 }
1232 return rSize;
1233 }
1234
1235 case ZSTDds_checkChecksum:
1236 assert(srcSize == 4); /* guaranteed by dctx->expected */
1237 {
1238 if (dctx->validateChecksum) {
1239 U32 const h32 = (U32)xxh64_digest(&dctx->xxhState);
1240 U32 const check32 = MEM_readLE32(src);
1241 DEBUGLOG(4, "ZSTD_decompressContinue: checksum : calculated %08X :: %08X read", (unsigned)h32, (unsigned)check32);
1242 RETURN_ERROR_IF(check32 != h32, checksum_wrong, "");
1243 }
1244 ZSTD_DCtx_trace_end(dctx, dctx->decodedSize, dctx->processedCSize, /* streaming */ 1);
1245 dctx->expected = 0;
1246 dctx->stage = ZSTDds_getFrameHeaderSize;
1247 return 0;
1248 }
1249
1250 case ZSTDds_decodeSkippableHeader:
1251 assert(src != NULL);
1252 assert(srcSize <= ZSTD_SKIPPABLEHEADERSIZE);
1253 ZSTD_memcpy(dctx->headerBuffer + (ZSTD_SKIPPABLEHEADERSIZE - srcSize), src, srcSize); /* complete skippable header */
1254 dctx->expected = MEM_readLE32(dctx->headerBuffer + ZSTD_FRAMEIDSIZE); /* note : dctx->expected can grow seriously large, beyond local buffer size */
1255 dctx->stage = ZSTDds_skipFrame;
1256 return 0;
1257
1258 case ZSTDds_skipFrame:
1259 dctx->expected = 0;
1260 dctx->stage = ZSTDds_getFrameHeaderSize;
1261 return 0;
1262
1263 default:
1264 assert(0); /* impossible */
1265 RETURN_ERROR(GENERIC, "impossible to reach"); /* some compiler require default to do something */
1266 }
1267 }
1268
1269
ZSTD_refDictContent(ZSTD_DCtx * dctx,const void * dict,size_t dictSize)1270 static size_t ZSTD_refDictContent(ZSTD_DCtx* dctx, const void* dict, size_t dictSize)
1271 {
1272 dctx->dictEnd = dctx->previousDstEnd;
1273 dctx->virtualStart = (const char*)dict - ((const char*)(dctx->previousDstEnd) - (const char*)(dctx->prefixStart));
1274 dctx->prefixStart = dict;
1275 dctx->previousDstEnd = (const char*)dict + dictSize;
1276 #ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
1277 dctx->dictContentBeginForFuzzing = dctx->prefixStart;
1278 dctx->dictContentEndForFuzzing = dctx->previousDstEnd;
1279 #endif
1280 return 0;
1281 }
1282
1283 /*! ZSTD_loadDEntropy() :
1284 * dict : must point at beginning of a valid zstd dictionary.
1285 * @return : size of entropy tables read */
1286 size_t
ZSTD_loadDEntropy(ZSTD_entropyDTables_t * entropy,const void * const dict,size_t const dictSize)1287 ZSTD_loadDEntropy(ZSTD_entropyDTables_t* entropy,
1288 const void* const dict, size_t const dictSize)
1289 {
1290 const BYTE* dictPtr = (const BYTE*)dict;
1291 const BYTE* const dictEnd = dictPtr + dictSize;
1292
1293 RETURN_ERROR_IF(dictSize <= 8, dictionary_corrupted, "dict is too small");
1294 assert(MEM_readLE32(dict) == ZSTD_MAGIC_DICTIONARY); /* dict must be valid */
1295 dictPtr += 8; /* skip header = magic + dictID */
1296
1297 ZSTD_STATIC_ASSERT(offsetof(ZSTD_entropyDTables_t, OFTable) == offsetof(ZSTD_entropyDTables_t, LLTable) + sizeof(entropy->LLTable));
1298 ZSTD_STATIC_ASSERT(offsetof(ZSTD_entropyDTables_t, MLTable) == offsetof(ZSTD_entropyDTables_t, OFTable) + sizeof(entropy->OFTable));
1299 ZSTD_STATIC_ASSERT(sizeof(entropy->LLTable) + sizeof(entropy->OFTable) + sizeof(entropy->MLTable) >= HUF_DECOMPRESS_WORKSPACE_SIZE);
1300 { void* const workspace = &entropy->LLTable; /* use fse tables as temporary workspace; implies fse tables are grouped together */
1301 size_t const workspaceSize = sizeof(entropy->LLTable) + sizeof(entropy->OFTable) + sizeof(entropy->MLTable);
1302 #ifdef HUF_FORCE_DECOMPRESS_X1
1303 /* in minimal huffman, we always use X1 variants */
1304 size_t const hSize = HUF_readDTableX1_wksp(entropy->hufTable,
1305 dictPtr, dictEnd - dictPtr,
1306 workspace, workspaceSize);
1307 #else
1308 size_t const hSize = HUF_readDTableX2_wksp(entropy->hufTable,
1309 dictPtr, (size_t)(dictEnd - dictPtr),
1310 workspace, workspaceSize);
1311 #endif
1312 RETURN_ERROR_IF(HUF_isError(hSize), dictionary_corrupted, "");
1313 dictPtr += hSize;
1314 }
1315
1316 { short offcodeNCount[MaxOff+1];
1317 unsigned offcodeMaxValue = MaxOff, offcodeLog;
1318 size_t const offcodeHeaderSize = FSE_readNCount(offcodeNCount, &offcodeMaxValue, &offcodeLog, dictPtr, (size_t)(dictEnd-dictPtr));
1319 RETURN_ERROR_IF(FSE_isError(offcodeHeaderSize), dictionary_corrupted, "");
1320 RETURN_ERROR_IF(offcodeMaxValue > MaxOff, dictionary_corrupted, "");
1321 RETURN_ERROR_IF(offcodeLog > OffFSELog, dictionary_corrupted, "");
1322 ZSTD_buildFSETable( entropy->OFTable,
1323 offcodeNCount, offcodeMaxValue,
1324 OF_base, OF_bits,
1325 offcodeLog,
1326 entropy->workspace, sizeof(entropy->workspace),
1327 /* bmi2 */0);
1328 dictPtr += offcodeHeaderSize;
1329 }
1330
1331 { short matchlengthNCount[MaxML+1];
1332 unsigned matchlengthMaxValue = MaxML, matchlengthLog;
1333 size_t const matchlengthHeaderSize = FSE_readNCount(matchlengthNCount, &matchlengthMaxValue, &matchlengthLog, dictPtr, (size_t)(dictEnd-dictPtr));
1334 RETURN_ERROR_IF(FSE_isError(matchlengthHeaderSize), dictionary_corrupted, "");
1335 RETURN_ERROR_IF(matchlengthMaxValue > MaxML, dictionary_corrupted, "");
1336 RETURN_ERROR_IF(matchlengthLog > MLFSELog, dictionary_corrupted, "");
1337 ZSTD_buildFSETable( entropy->MLTable,
1338 matchlengthNCount, matchlengthMaxValue,
1339 ML_base, ML_bits,
1340 matchlengthLog,
1341 entropy->workspace, sizeof(entropy->workspace),
1342 /* bmi2 */ 0);
1343 dictPtr += matchlengthHeaderSize;
1344 }
1345
1346 { short litlengthNCount[MaxLL+1];
1347 unsigned litlengthMaxValue = MaxLL, litlengthLog;
1348 size_t const litlengthHeaderSize = FSE_readNCount(litlengthNCount, &litlengthMaxValue, &litlengthLog, dictPtr, (size_t)(dictEnd-dictPtr));
1349 RETURN_ERROR_IF(FSE_isError(litlengthHeaderSize), dictionary_corrupted, "");
1350 RETURN_ERROR_IF(litlengthMaxValue > MaxLL, dictionary_corrupted, "");
1351 RETURN_ERROR_IF(litlengthLog > LLFSELog, dictionary_corrupted, "");
1352 ZSTD_buildFSETable( entropy->LLTable,
1353 litlengthNCount, litlengthMaxValue,
1354 LL_base, LL_bits,
1355 litlengthLog,
1356 entropy->workspace, sizeof(entropy->workspace),
1357 /* bmi2 */ 0);
1358 dictPtr += litlengthHeaderSize;
1359 }
1360
1361 RETURN_ERROR_IF(dictPtr+12 > dictEnd, dictionary_corrupted, "");
1362 { int i;
1363 size_t const dictContentSize = (size_t)(dictEnd - (dictPtr+12));
1364 for (i=0; i<3; i++) {
1365 U32 const rep = MEM_readLE32(dictPtr); dictPtr += 4;
1366 RETURN_ERROR_IF(rep==0 || rep > dictContentSize,
1367 dictionary_corrupted, "");
1368 entropy->rep[i] = rep;
1369 } }
1370
1371 return (size_t)(dictPtr - (const BYTE*)dict);
1372 }
1373
ZSTD_decompress_insertDictionary(ZSTD_DCtx * dctx,const void * dict,size_t dictSize)1374 static size_t ZSTD_decompress_insertDictionary(ZSTD_DCtx* dctx, const void* dict, size_t dictSize)
1375 {
1376 if (dictSize < 8) return ZSTD_refDictContent(dctx, dict, dictSize);
1377 { U32 const magic = MEM_readLE32(dict);
1378 if (magic != ZSTD_MAGIC_DICTIONARY) {
1379 return ZSTD_refDictContent(dctx, dict, dictSize); /* pure content mode */
1380 } }
1381 dctx->dictID = MEM_readLE32((const char*)dict + ZSTD_FRAMEIDSIZE);
1382
1383 /* load entropy tables */
1384 { size_t const eSize = ZSTD_loadDEntropy(&dctx->entropy, dict, dictSize);
1385 RETURN_ERROR_IF(ZSTD_isError(eSize), dictionary_corrupted, "");
1386 dict = (const char*)dict + eSize;
1387 dictSize -= eSize;
1388 }
1389 dctx->litEntropy = dctx->fseEntropy = 1;
1390
1391 /* reference dictionary content */
1392 return ZSTD_refDictContent(dctx, dict, dictSize);
1393 }
1394
ZSTD_decompressBegin(ZSTD_DCtx * dctx)1395 size_t ZSTD_decompressBegin(ZSTD_DCtx* dctx)
1396 {
1397 assert(dctx != NULL);
1398 dctx->expected = ZSTD_startingInputLength(dctx->format); /* dctx->format must be properly set */
1399 dctx->stage = ZSTDds_getFrameHeaderSize;
1400 dctx->processedCSize = 0;
1401 dctx->decodedSize = 0;
1402 dctx->previousDstEnd = NULL;
1403 dctx->prefixStart = NULL;
1404 dctx->virtualStart = NULL;
1405 dctx->dictEnd = NULL;
1406 dctx->entropy.hufTable[0] = (HUF_DTable)((HufLog)*0x1000001); /* cover both little and big endian */
1407 dctx->litEntropy = dctx->fseEntropy = 0;
1408 dctx->dictID = 0;
1409 dctx->bType = bt_reserved;
1410 ZSTD_STATIC_ASSERT(sizeof(dctx->entropy.rep) == sizeof(repStartValue));
1411 ZSTD_memcpy(dctx->entropy.rep, repStartValue, sizeof(repStartValue)); /* initial repcodes */
1412 dctx->LLTptr = dctx->entropy.LLTable;
1413 dctx->MLTptr = dctx->entropy.MLTable;
1414 dctx->OFTptr = dctx->entropy.OFTable;
1415 dctx->HUFptr = dctx->entropy.hufTable;
1416 return 0;
1417 }
1418
ZSTD_decompressBegin_usingDict(ZSTD_DCtx * dctx,const void * dict,size_t dictSize)1419 size_t ZSTD_decompressBegin_usingDict(ZSTD_DCtx* dctx, const void* dict, size_t dictSize)
1420 {
1421 FORWARD_IF_ERROR( ZSTD_decompressBegin(dctx) , "");
1422 if (dict && dictSize)
1423 RETURN_ERROR_IF(
1424 ZSTD_isError(ZSTD_decompress_insertDictionary(dctx, dict, dictSize)),
1425 dictionary_corrupted, "");
1426 return 0;
1427 }
1428
1429
1430 /* ====== ZSTD_DDict ====== */
1431
ZSTD_decompressBegin_usingDDict(ZSTD_DCtx * dctx,const ZSTD_DDict * ddict)1432 size_t ZSTD_decompressBegin_usingDDict(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict)
1433 {
1434 DEBUGLOG(4, "ZSTD_decompressBegin_usingDDict");
1435 assert(dctx != NULL);
1436 if (ddict) {
1437 const char* const dictStart = (const char*)ZSTD_DDict_dictContent(ddict);
1438 size_t const dictSize = ZSTD_DDict_dictSize(ddict);
1439 const void* const dictEnd = dictStart + dictSize;
1440 dctx->ddictIsCold = (dctx->dictEnd != dictEnd);
1441 DEBUGLOG(4, "DDict is %s",
1442 dctx->ddictIsCold ? "~cold~" : "hot!");
1443 }
1444 FORWARD_IF_ERROR( ZSTD_decompressBegin(dctx) , "");
1445 if (ddict) { /* NULL ddict is equivalent to no dictionary */
1446 ZSTD_copyDDictParameters(dctx, ddict);
1447 }
1448 return 0;
1449 }
1450
1451 /*! ZSTD_getDictID_fromDict() :
1452 * Provides the dictID stored within dictionary.
1453 * if @return == 0, the dictionary is not conformant with Zstandard specification.
1454 * It can still be loaded, but as a content-only dictionary. */
ZSTD_getDictID_fromDict(const void * dict,size_t dictSize)1455 unsigned ZSTD_getDictID_fromDict(const void* dict, size_t dictSize)
1456 {
1457 if (dictSize < 8) return 0;
1458 if (MEM_readLE32(dict) != ZSTD_MAGIC_DICTIONARY) return 0;
1459 return MEM_readLE32((const char*)dict + ZSTD_FRAMEIDSIZE);
1460 }
1461
1462 /*! ZSTD_getDictID_fromFrame() :
1463 * Provides the dictID required to decompress frame stored within `src`.
1464 * If @return == 0, the dictID could not be decoded.
1465 * This could for one of the following reasons :
1466 * - The frame does not require a dictionary (most common case).
1467 * - The frame was built with dictID intentionally removed.
1468 * Needed dictionary is a hidden information.
1469 * Note : this use case also happens when using a non-conformant dictionary.
1470 * - `srcSize` is too small, and as a result, frame header could not be decoded.
1471 * Note : possible if `srcSize < ZSTD_FRAMEHEADERSIZE_MAX`.
1472 * - This is not a Zstandard frame.
1473 * When identifying the exact failure cause, it's possible to use
1474 * ZSTD_getFrameHeader(), which will provide a more precise error code. */
ZSTD_getDictID_fromFrame(const void * src,size_t srcSize)1475 unsigned ZSTD_getDictID_fromFrame(const void* src, size_t srcSize)
1476 {
1477 ZSTD_frameHeader zfp = { 0, 0, 0, ZSTD_frame, 0, 0, 0 };
1478 size_t const hError = ZSTD_getFrameHeader(&zfp, src, srcSize);
1479 if (ZSTD_isError(hError)) return 0;
1480 return zfp.dictID;
1481 }
1482
1483
1484 /*! ZSTD_decompress_usingDDict() :
1485 * Decompression using a pre-digested Dictionary
1486 * Use dictionary without significant overhead. */
ZSTD_decompress_usingDDict(ZSTD_DCtx * dctx,void * dst,size_t dstCapacity,const void * src,size_t srcSize,const ZSTD_DDict * ddict)1487 size_t ZSTD_decompress_usingDDict(ZSTD_DCtx* dctx,
1488 void* dst, size_t dstCapacity,
1489 const void* src, size_t srcSize,
1490 const ZSTD_DDict* ddict)
1491 {
1492 /* pass content and size in case legacy frames are encountered */
1493 return ZSTD_decompressMultiFrame(dctx, dst, dstCapacity, src, srcSize,
1494 NULL, 0,
1495 ddict);
1496 }
1497
1498
1499 /*=====================================
1500 * Streaming decompression
1501 *====================================*/
1502
ZSTD_createDStream(void)1503 ZSTD_DStream* ZSTD_createDStream(void)
1504 {
1505 DEBUGLOG(3, "ZSTD_createDStream");
1506 return ZSTD_createDCtx_internal(ZSTD_defaultCMem);
1507 }
1508
ZSTD_initStaticDStream(void * workspace,size_t workspaceSize)1509 ZSTD_DStream* ZSTD_initStaticDStream(void *workspace, size_t workspaceSize)
1510 {
1511 return ZSTD_initStaticDCtx(workspace, workspaceSize);
1512 }
1513
ZSTD_createDStream_advanced(ZSTD_customMem customMem)1514 ZSTD_DStream* ZSTD_createDStream_advanced(ZSTD_customMem customMem)
1515 {
1516 return ZSTD_createDCtx_internal(customMem);
1517 }
1518
ZSTD_freeDStream(ZSTD_DStream * zds)1519 size_t ZSTD_freeDStream(ZSTD_DStream* zds)
1520 {
1521 return ZSTD_freeDCtx(zds);
1522 }
1523
1524
1525 /* *** Initialization *** */
1526
ZSTD_DStreamInSize(void)1527 size_t ZSTD_DStreamInSize(void) { return ZSTD_BLOCKSIZE_MAX + ZSTD_blockHeaderSize; }
ZSTD_DStreamOutSize(void)1528 size_t ZSTD_DStreamOutSize(void) { return ZSTD_BLOCKSIZE_MAX; }
1529
ZSTD_DCtx_loadDictionary_advanced(ZSTD_DCtx * dctx,const void * dict,size_t dictSize,ZSTD_dictLoadMethod_e dictLoadMethod,ZSTD_dictContentType_e dictContentType)1530 size_t ZSTD_DCtx_loadDictionary_advanced(ZSTD_DCtx* dctx,
1531 const void* dict, size_t dictSize,
1532 ZSTD_dictLoadMethod_e dictLoadMethod,
1533 ZSTD_dictContentType_e dictContentType)
1534 {
1535 RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong, "");
1536 ZSTD_clearDict(dctx);
1537 if (dict && dictSize != 0) {
1538 dctx->ddictLocal = ZSTD_createDDict_advanced(dict, dictSize, dictLoadMethod, dictContentType, dctx->customMem);
1539 RETURN_ERROR_IF(dctx->ddictLocal == NULL, memory_allocation, "NULL pointer!");
1540 dctx->ddict = dctx->ddictLocal;
1541 dctx->dictUses = ZSTD_use_indefinitely;
1542 }
1543 return 0;
1544 }
1545
ZSTD_DCtx_loadDictionary_byReference(ZSTD_DCtx * dctx,const void * dict,size_t dictSize)1546 size_t ZSTD_DCtx_loadDictionary_byReference(ZSTD_DCtx* dctx, const void* dict, size_t dictSize)
1547 {
1548 return ZSTD_DCtx_loadDictionary_advanced(dctx, dict, dictSize, ZSTD_dlm_byRef, ZSTD_dct_auto);
1549 }
1550
ZSTD_DCtx_loadDictionary(ZSTD_DCtx * dctx,const void * dict,size_t dictSize)1551 size_t ZSTD_DCtx_loadDictionary(ZSTD_DCtx* dctx, const void* dict, size_t dictSize)
1552 {
1553 return ZSTD_DCtx_loadDictionary_advanced(dctx, dict, dictSize, ZSTD_dlm_byCopy, ZSTD_dct_auto);
1554 }
1555
ZSTD_DCtx_refPrefix_advanced(ZSTD_DCtx * dctx,const void * prefix,size_t prefixSize,ZSTD_dictContentType_e dictContentType)1556 size_t ZSTD_DCtx_refPrefix_advanced(ZSTD_DCtx* dctx, const void* prefix, size_t prefixSize, ZSTD_dictContentType_e dictContentType)
1557 {
1558 FORWARD_IF_ERROR(ZSTD_DCtx_loadDictionary_advanced(dctx, prefix, prefixSize, ZSTD_dlm_byRef, dictContentType), "");
1559 dctx->dictUses = ZSTD_use_once;
1560 return 0;
1561 }
1562
ZSTD_DCtx_refPrefix(ZSTD_DCtx * dctx,const void * prefix,size_t prefixSize)1563 size_t ZSTD_DCtx_refPrefix(ZSTD_DCtx* dctx, const void* prefix, size_t prefixSize)
1564 {
1565 return ZSTD_DCtx_refPrefix_advanced(dctx, prefix, prefixSize, ZSTD_dct_rawContent);
1566 }
1567
1568
1569 /* ZSTD_initDStream_usingDict() :
1570 * return : expected size, aka ZSTD_startingInputLength().
1571 * this function cannot fail */
ZSTD_initDStream_usingDict(ZSTD_DStream * zds,const void * dict,size_t dictSize)1572 size_t ZSTD_initDStream_usingDict(ZSTD_DStream* zds, const void* dict, size_t dictSize)
1573 {
1574 DEBUGLOG(4, "ZSTD_initDStream_usingDict");
1575 FORWARD_IF_ERROR( ZSTD_DCtx_reset(zds, ZSTD_reset_session_only) , "");
1576 FORWARD_IF_ERROR( ZSTD_DCtx_loadDictionary(zds, dict, dictSize) , "");
1577 return ZSTD_startingInputLength(zds->format);
1578 }
1579
1580 /* note : this variant can't fail */
ZSTD_initDStream(ZSTD_DStream * zds)1581 size_t ZSTD_initDStream(ZSTD_DStream* zds)
1582 {
1583 DEBUGLOG(4, "ZSTD_initDStream");
1584 return ZSTD_initDStream_usingDDict(zds, NULL);
1585 }
1586
1587 /* ZSTD_initDStream_usingDDict() :
1588 * ddict will just be referenced, and must outlive decompression session
1589 * this function cannot fail */
ZSTD_initDStream_usingDDict(ZSTD_DStream * dctx,const ZSTD_DDict * ddict)1590 size_t ZSTD_initDStream_usingDDict(ZSTD_DStream* dctx, const ZSTD_DDict* ddict)
1591 {
1592 FORWARD_IF_ERROR( ZSTD_DCtx_reset(dctx, ZSTD_reset_session_only) , "");
1593 FORWARD_IF_ERROR( ZSTD_DCtx_refDDict(dctx, ddict) , "");
1594 return ZSTD_startingInputLength(dctx->format);
1595 }
1596
1597 /* ZSTD_resetDStream() :
1598 * return : expected size, aka ZSTD_startingInputLength().
1599 * this function cannot fail */
ZSTD_resetDStream(ZSTD_DStream * dctx)1600 size_t ZSTD_resetDStream(ZSTD_DStream* dctx)
1601 {
1602 FORWARD_IF_ERROR(ZSTD_DCtx_reset(dctx, ZSTD_reset_session_only), "");
1603 return ZSTD_startingInputLength(dctx->format);
1604 }
1605
1606
ZSTD_DCtx_refDDict(ZSTD_DCtx * dctx,const ZSTD_DDict * ddict)1607 size_t ZSTD_DCtx_refDDict(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict)
1608 {
1609 RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong, "");
1610 ZSTD_clearDict(dctx);
1611 if (ddict) {
1612 dctx->ddict = ddict;
1613 dctx->dictUses = ZSTD_use_indefinitely;
1614 if (dctx->refMultipleDDicts == ZSTD_rmd_refMultipleDDicts) {
1615 if (dctx->ddictSet == NULL) {
1616 dctx->ddictSet = ZSTD_createDDictHashSet(dctx->customMem);
1617 if (!dctx->ddictSet) {
1618 RETURN_ERROR(memory_allocation, "Failed to allocate memory for hash set!");
1619 }
1620 }
1621 assert(!dctx->staticSize); /* Impossible: ddictSet cannot have been allocated if static dctx */
1622 FORWARD_IF_ERROR(ZSTD_DDictHashSet_addDDict(dctx->ddictSet, ddict, dctx->customMem), "");
1623 }
1624 }
1625 return 0;
1626 }
1627
1628 /* ZSTD_DCtx_setMaxWindowSize() :
1629 * note : no direct equivalence in ZSTD_DCtx_setParameter,
1630 * since this version sets windowSize, and the other sets windowLog */
ZSTD_DCtx_setMaxWindowSize(ZSTD_DCtx * dctx,size_t maxWindowSize)1631 size_t ZSTD_DCtx_setMaxWindowSize(ZSTD_DCtx* dctx, size_t maxWindowSize)
1632 {
1633 ZSTD_bounds const bounds = ZSTD_dParam_getBounds(ZSTD_d_windowLogMax);
1634 size_t const min = (size_t)1 << bounds.lowerBound;
1635 size_t const max = (size_t)1 << bounds.upperBound;
1636 RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong, "");
1637 RETURN_ERROR_IF(maxWindowSize < min, parameter_outOfBound, "");
1638 RETURN_ERROR_IF(maxWindowSize > max, parameter_outOfBound, "");
1639 dctx->maxWindowSize = maxWindowSize;
1640 return 0;
1641 }
1642
ZSTD_DCtx_setFormat(ZSTD_DCtx * dctx,ZSTD_format_e format)1643 size_t ZSTD_DCtx_setFormat(ZSTD_DCtx* dctx, ZSTD_format_e format)
1644 {
1645 return ZSTD_DCtx_setParameter(dctx, ZSTD_d_format, (int)format);
1646 }
1647
ZSTD_dParam_getBounds(ZSTD_dParameter dParam)1648 ZSTD_bounds ZSTD_dParam_getBounds(ZSTD_dParameter dParam)
1649 {
1650 ZSTD_bounds bounds = { 0, 0, 0 };
1651 switch(dParam) {
1652 case ZSTD_d_windowLogMax:
1653 bounds.lowerBound = ZSTD_WINDOWLOG_ABSOLUTEMIN;
1654 bounds.upperBound = ZSTD_WINDOWLOG_MAX;
1655 return bounds;
1656 case ZSTD_d_format:
1657 bounds.lowerBound = (int)ZSTD_f_zstd1;
1658 bounds.upperBound = (int)ZSTD_f_zstd1_magicless;
1659 ZSTD_STATIC_ASSERT(ZSTD_f_zstd1 < ZSTD_f_zstd1_magicless);
1660 return bounds;
1661 case ZSTD_d_stableOutBuffer:
1662 bounds.lowerBound = (int)ZSTD_bm_buffered;
1663 bounds.upperBound = (int)ZSTD_bm_stable;
1664 return bounds;
1665 case ZSTD_d_forceIgnoreChecksum:
1666 bounds.lowerBound = (int)ZSTD_d_validateChecksum;
1667 bounds.upperBound = (int)ZSTD_d_ignoreChecksum;
1668 return bounds;
1669 case ZSTD_d_refMultipleDDicts:
1670 bounds.lowerBound = (int)ZSTD_rmd_refSingleDDict;
1671 bounds.upperBound = (int)ZSTD_rmd_refMultipleDDicts;
1672 return bounds;
1673 default:;
1674 }
1675 bounds.error = ERROR(parameter_unsupported);
1676 return bounds;
1677 }
1678
1679 /* ZSTD_dParam_withinBounds:
1680 * @return 1 if value is within dParam bounds,
1681 * 0 otherwise */
ZSTD_dParam_withinBounds(ZSTD_dParameter dParam,int value)1682 static int ZSTD_dParam_withinBounds(ZSTD_dParameter dParam, int value)
1683 {
1684 ZSTD_bounds const bounds = ZSTD_dParam_getBounds(dParam);
1685 if (ZSTD_isError(bounds.error)) return 0;
1686 if (value < bounds.lowerBound) return 0;
1687 if (value > bounds.upperBound) return 0;
1688 return 1;
1689 }
1690
1691 #define CHECK_DBOUNDS(p,v) { \
1692 RETURN_ERROR_IF(!ZSTD_dParam_withinBounds(p, v), parameter_outOfBound, ""); \
1693 }
1694
ZSTD_DCtx_getParameter(ZSTD_DCtx * dctx,ZSTD_dParameter param,int * value)1695 size_t ZSTD_DCtx_getParameter(ZSTD_DCtx* dctx, ZSTD_dParameter param, int* value)
1696 {
1697 switch (param) {
1698 case ZSTD_d_windowLogMax:
1699 *value = (int)ZSTD_highbit32((U32)dctx->maxWindowSize);
1700 return 0;
1701 case ZSTD_d_format:
1702 *value = (int)dctx->format;
1703 return 0;
1704 case ZSTD_d_stableOutBuffer:
1705 *value = (int)dctx->outBufferMode;
1706 return 0;
1707 case ZSTD_d_forceIgnoreChecksum:
1708 *value = (int)dctx->forceIgnoreChecksum;
1709 return 0;
1710 case ZSTD_d_refMultipleDDicts:
1711 *value = (int)dctx->refMultipleDDicts;
1712 return 0;
1713 default:;
1714 }
1715 RETURN_ERROR(parameter_unsupported, "");
1716 }
1717
ZSTD_DCtx_setParameter(ZSTD_DCtx * dctx,ZSTD_dParameter dParam,int value)1718 size_t ZSTD_DCtx_setParameter(ZSTD_DCtx* dctx, ZSTD_dParameter dParam, int value)
1719 {
1720 RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong, "");
1721 switch(dParam) {
1722 case ZSTD_d_windowLogMax:
1723 if (value == 0) value = ZSTD_WINDOWLOG_LIMIT_DEFAULT;
1724 CHECK_DBOUNDS(ZSTD_d_windowLogMax, value);
1725 dctx->maxWindowSize = ((size_t)1) << value;
1726 return 0;
1727 case ZSTD_d_format:
1728 CHECK_DBOUNDS(ZSTD_d_format, value);
1729 dctx->format = (ZSTD_format_e)value;
1730 return 0;
1731 case ZSTD_d_stableOutBuffer:
1732 CHECK_DBOUNDS(ZSTD_d_stableOutBuffer, value);
1733 dctx->outBufferMode = (ZSTD_bufferMode_e)value;
1734 return 0;
1735 case ZSTD_d_forceIgnoreChecksum:
1736 CHECK_DBOUNDS(ZSTD_d_forceIgnoreChecksum, value);
1737 dctx->forceIgnoreChecksum = (ZSTD_forceIgnoreChecksum_e)value;
1738 return 0;
1739 case ZSTD_d_refMultipleDDicts:
1740 CHECK_DBOUNDS(ZSTD_d_refMultipleDDicts, value);
1741 if (dctx->staticSize != 0) {
1742 RETURN_ERROR(parameter_unsupported, "Static dctx does not support multiple DDicts!");
1743 }
1744 dctx->refMultipleDDicts = (ZSTD_refMultipleDDicts_e)value;
1745 return 0;
1746 default:;
1747 }
1748 RETURN_ERROR(parameter_unsupported, "");
1749 }
1750
ZSTD_DCtx_reset(ZSTD_DCtx * dctx,ZSTD_ResetDirective reset)1751 size_t ZSTD_DCtx_reset(ZSTD_DCtx* dctx, ZSTD_ResetDirective reset)
1752 {
1753 if ( (reset == ZSTD_reset_session_only)
1754 || (reset == ZSTD_reset_session_and_parameters) ) {
1755 dctx->streamStage = zdss_init;
1756 dctx->noForwardProgress = 0;
1757 }
1758 if ( (reset == ZSTD_reset_parameters)
1759 || (reset == ZSTD_reset_session_and_parameters) ) {
1760 RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong, "");
1761 ZSTD_clearDict(dctx);
1762 ZSTD_DCtx_resetParameters(dctx);
1763 }
1764 return 0;
1765 }
1766
1767
ZSTD_sizeof_DStream(const ZSTD_DStream * dctx)1768 size_t ZSTD_sizeof_DStream(const ZSTD_DStream* dctx)
1769 {
1770 return ZSTD_sizeof_DCtx(dctx);
1771 }
1772
ZSTD_decodingBufferSize_min(unsigned long long windowSize,unsigned long long frameContentSize)1773 size_t ZSTD_decodingBufferSize_min(unsigned long long windowSize, unsigned long long frameContentSize)
1774 {
1775 size_t const blockSize = (size_t) MIN(windowSize, ZSTD_BLOCKSIZE_MAX);
1776 /* space is needed to store the litbuffer after the output of a given block without stomping the extDict of a previous run, as well as to cover both windows against wildcopy*/
1777 unsigned long long const neededRBSize = windowSize + blockSize + ZSTD_BLOCKSIZE_MAX + (WILDCOPY_OVERLENGTH * 2);
1778 unsigned long long const neededSize = MIN(frameContentSize, neededRBSize);
1779 size_t const minRBSize = (size_t) neededSize;
1780 RETURN_ERROR_IF((unsigned long long)minRBSize != neededSize,
1781 frameParameter_windowTooLarge, "");
1782 return minRBSize;
1783 }
1784
ZSTD_estimateDStreamSize(size_t windowSize)1785 size_t ZSTD_estimateDStreamSize(size_t windowSize)
1786 {
1787 size_t const blockSize = MIN(windowSize, ZSTD_BLOCKSIZE_MAX);
1788 size_t const inBuffSize = blockSize; /* no block can be larger */
1789 size_t const outBuffSize = ZSTD_decodingBufferSize_min(windowSize, ZSTD_CONTENTSIZE_UNKNOWN);
1790 return ZSTD_estimateDCtxSize() + inBuffSize + outBuffSize;
1791 }
1792
ZSTD_estimateDStreamSize_fromFrame(const void * src,size_t srcSize)1793 size_t ZSTD_estimateDStreamSize_fromFrame(const void* src, size_t srcSize)
1794 {
1795 U32 const windowSizeMax = 1U << ZSTD_WINDOWLOG_MAX; /* note : should be user-selectable, but requires an additional parameter (or a dctx) */
1796 ZSTD_frameHeader zfh;
1797 size_t const err = ZSTD_getFrameHeader(&zfh, src, srcSize);
1798 if (ZSTD_isError(err)) return err;
1799 RETURN_ERROR_IF(err>0, srcSize_wrong, "");
1800 RETURN_ERROR_IF(zfh.windowSize > windowSizeMax,
1801 frameParameter_windowTooLarge, "");
1802 return ZSTD_estimateDStreamSize((size_t)zfh.windowSize);
1803 }
1804
1805
1806 /* ***** Decompression ***** */
1807
ZSTD_DCtx_isOverflow(ZSTD_DStream * zds,size_t const neededInBuffSize,size_t const neededOutBuffSize)1808 static int ZSTD_DCtx_isOverflow(ZSTD_DStream* zds, size_t const neededInBuffSize, size_t const neededOutBuffSize)
1809 {
1810 return (zds->inBuffSize + zds->outBuffSize) >= (neededInBuffSize + neededOutBuffSize) * ZSTD_WORKSPACETOOLARGE_FACTOR;
1811 }
1812
ZSTD_DCtx_updateOversizedDuration(ZSTD_DStream * zds,size_t const neededInBuffSize,size_t const neededOutBuffSize)1813 static void ZSTD_DCtx_updateOversizedDuration(ZSTD_DStream* zds, size_t const neededInBuffSize, size_t const neededOutBuffSize)
1814 {
1815 if (ZSTD_DCtx_isOverflow(zds, neededInBuffSize, neededOutBuffSize))
1816 zds->oversizedDuration++;
1817 else
1818 zds->oversizedDuration = 0;
1819 }
1820
ZSTD_DCtx_isOversizedTooLong(ZSTD_DStream * zds)1821 static int ZSTD_DCtx_isOversizedTooLong(ZSTD_DStream* zds)
1822 {
1823 return zds->oversizedDuration >= ZSTD_WORKSPACETOOLARGE_MAXDURATION;
1824 }
1825
1826 /* Checks that the output buffer hasn't changed if ZSTD_obm_stable is used. */
ZSTD_checkOutBuffer(ZSTD_DStream const * zds,ZSTD_outBuffer const * output)1827 static size_t ZSTD_checkOutBuffer(ZSTD_DStream const* zds, ZSTD_outBuffer const* output)
1828 {
1829 ZSTD_outBuffer const expect = zds->expectedOutBuffer;
1830 /* No requirement when ZSTD_obm_stable is not enabled. */
1831 if (zds->outBufferMode != ZSTD_bm_stable)
1832 return 0;
1833 /* Any buffer is allowed in zdss_init, this must be the same for every other call until
1834 * the context is reset.
1835 */
1836 if (zds->streamStage == zdss_init)
1837 return 0;
1838 /* The buffer must match our expectation exactly. */
1839 if (expect.dst == output->dst && expect.pos == output->pos && expect.size == output->size)
1840 return 0;
1841 RETURN_ERROR(dstBuffer_wrong, "ZSTD_d_stableOutBuffer enabled but output differs!");
1842 }
1843
1844 /* Calls ZSTD_decompressContinue() with the right parameters for ZSTD_decompressStream()
1845 * and updates the stage and the output buffer state. This call is extracted so it can be
1846 * used both when reading directly from the ZSTD_inBuffer, and in buffered input mode.
1847 * NOTE: You must break after calling this function since the streamStage is modified.
1848 */
ZSTD_decompressContinueStream(ZSTD_DStream * zds,char ** op,char * oend,void const * src,size_t srcSize)1849 static size_t ZSTD_decompressContinueStream(
1850 ZSTD_DStream* zds, char** op, char* oend,
1851 void const* src, size_t srcSize) {
1852 int const isSkipFrame = ZSTD_isSkipFrame(zds);
1853 if (zds->outBufferMode == ZSTD_bm_buffered) {
1854 size_t const dstSize = isSkipFrame ? 0 : zds->outBuffSize - zds->outStart;
1855 size_t const decodedSize = ZSTD_decompressContinue(zds,
1856 zds->outBuff + zds->outStart, dstSize, src, srcSize);
1857 FORWARD_IF_ERROR(decodedSize, "");
1858 if (!decodedSize && !isSkipFrame) {
1859 zds->streamStage = zdss_read;
1860 } else {
1861 zds->outEnd = zds->outStart + decodedSize;
1862 zds->streamStage = zdss_flush;
1863 }
1864 } else {
1865 /* Write directly into the output buffer */
1866 size_t const dstSize = isSkipFrame ? 0 : (size_t)(oend - *op);
1867 size_t const decodedSize = ZSTD_decompressContinue(zds, *op, dstSize, src, srcSize);
1868 FORWARD_IF_ERROR(decodedSize, "");
1869 *op += decodedSize;
1870 /* Flushing is not needed. */
1871 zds->streamStage = zdss_read;
1872 assert(*op <= oend);
1873 assert(zds->outBufferMode == ZSTD_bm_stable);
1874 }
1875 return 0;
1876 }
1877
ZSTD_decompressStream(ZSTD_DStream * zds,ZSTD_outBuffer * output,ZSTD_inBuffer * input)1878 size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inBuffer* input)
1879 {
1880 const char* const src = (const char*)input->src;
1881 const char* const istart = input->pos != 0 ? src + input->pos : src;
1882 const char* const iend = input->size != 0 ? src + input->size : src;
1883 const char* ip = istart;
1884 char* const dst = (char*)output->dst;
1885 char* const ostart = output->pos != 0 ? dst + output->pos : dst;
1886 char* const oend = output->size != 0 ? dst + output->size : dst;
1887 char* op = ostart;
1888 U32 someMoreWork = 1;
1889
1890 DEBUGLOG(5, "ZSTD_decompressStream");
1891 RETURN_ERROR_IF(
1892 input->pos > input->size,
1893 srcSize_wrong,
1894 "forbidden. in: pos: %u vs size: %u",
1895 (U32)input->pos, (U32)input->size);
1896 RETURN_ERROR_IF(
1897 output->pos > output->size,
1898 dstSize_tooSmall,
1899 "forbidden. out: pos: %u vs size: %u",
1900 (U32)output->pos, (U32)output->size);
1901 DEBUGLOG(5, "input size : %u", (U32)(input->size - input->pos));
1902 FORWARD_IF_ERROR(ZSTD_checkOutBuffer(zds, output), "");
1903
1904 while (someMoreWork) {
1905 switch(zds->streamStage)
1906 {
1907 case zdss_init :
1908 DEBUGLOG(5, "stage zdss_init => transparent reset ");
1909 zds->streamStage = zdss_loadHeader;
1910 zds->lhSize = zds->inPos = zds->outStart = zds->outEnd = 0;
1911 zds->hostageByte = 0;
1912 zds->expectedOutBuffer = *output;
1913 ZSTD_FALLTHROUGH;
1914
1915 case zdss_loadHeader :
1916 DEBUGLOG(5, "stage zdss_loadHeader (srcSize : %u)", (U32)(iend - ip));
1917 { size_t const hSize = ZSTD_getFrameHeader_advanced(&zds->fParams, zds->headerBuffer, zds->lhSize, zds->format);
1918 if (zds->refMultipleDDicts && zds->ddictSet) {
1919 ZSTD_DCtx_selectFrameDDict(zds);
1920 }
1921 DEBUGLOG(5, "header size : %u", (U32)hSize);
1922 if (ZSTD_isError(hSize)) {
1923 return hSize; /* error */
1924 }
1925 if (hSize != 0) { /* need more input */
1926 size_t const toLoad = hSize - zds->lhSize; /* if hSize!=0, hSize > zds->lhSize */
1927 size_t const remainingInput = (size_t)(iend-ip);
1928 assert(iend >= ip);
1929 if (toLoad > remainingInput) { /* not enough input to load full header */
1930 if (remainingInput > 0) {
1931 ZSTD_memcpy(zds->headerBuffer + zds->lhSize, ip, remainingInput);
1932 zds->lhSize += remainingInput;
1933 }
1934 input->pos = input->size;
1935 return (MAX((size_t)ZSTD_FRAMEHEADERSIZE_MIN(zds->format), hSize) - zds->lhSize) + ZSTD_blockHeaderSize; /* remaining header bytes + next block header */
1936 }
1937 assert(ip != NULL);
1938 ZSTD_memcpy(zds->headerBuffer + zds->lhSize, ip, toLoad); zds->lhSize = hSize; ip += toLoad;
1939 break;
1940 } }
1941
1942 /* check for single-pass mode opportunity */
1943 if (zds->fParams.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN
1944 && zds->fParams.frameType != ZSTD_skippableFrame
1945 && (U64)(size_t)(oend-op) >= zds->fParams.frameContentSize) {
1946 size_t const cSize = ZSTD_findFrameCompressedSize(istart, (size_t)(iend-istart));
1947 if (cSize <= (size_t)(iend-istart)) {
1948 /* shortcut : using single-pass mode */
1949 size_t const decompressedSize = ZSTD_decompress_usingDDict(zds, op, (size_t)(oend-op), istart, cSize, ZSTD_getDDict(zds));
1950 if (ZSTD_isError(decompressedSize)) return decompressedSize;
1951 DEBUGLOG(4, "shortcut to single-pass ZSTD_decompress_usingDDict()")
1952 ip = istart + cSize;
1953 op += decompressedSize;
1954 zds->expected = 0;
1955 zds->streamStage = zdss_init;
1956 someMoreWork = 0;
1957 break;
1958 } }
1959
1960 /* Check output buffer is large enough for ZSTD_odm_stable. */
1961 if (zds->outBufferMode == ZSTD_bm_stable
1962 && zds->fParams.frameType != ZSTD_skippableFrame
1963 && zds->fParams.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN
1964 && (U64)(size_t)(oend-op) < zds->fParams.frameContentSize) {
1965 RETURN_ERROR(dstSize_tooSmall, "ZSTD_obm_stable passed but ZSTD_outBuffer is too small");
1966 }
1967
1968 /* Consume header (see ZSTDds_decodeFrameHeader) */
1969 DEBUGLOG(4, "Consume header");
1970 FORWARD_IF_ERROR(ZSTD_decompressBegin_usingDDict(zds, ZSTD_getDDict(zds)), "");
1971
1972 if ((MEM_readLE32(zds->headerBuffer) & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) { /* skippable frame */
1973 zds->expected = MEM_readLE32(zds->headerBuffer + ZSTD_FRAMEIDSIZE);
1974 zds->stage = ZSTDds_skipFrame;
1975 } else {
1976 FORWARD_IF_ERROR(ZSTD_decodeFrameHeader(zds, zds->headerBuffer, zds->lhSize), "");
1977 zds->expected = ZSTD_blockHeaderSize;
1978 zds->stage = ZSTDds_decodeBlockHeader;
1979 }
1980
1981 /* control buffer memory usage */
1982 DEBUGLOG(4, "Control max memory usage (%u KB <= max %u KB)",
1983 (U32)(zds->fParams.windowSize >>10),
1984 (U32)(zds->maxWindowSize >> 10) );
1985 zds->fParams.windowSize = MAX(zds->fParams.windowSize, 1U << ZSTD_WINDOWLOG_ABSOLUTEMIN);
1986 RETURN_ERROR_IF(zds->fParams.windowSize > zds->maxWindowSize,
1987 frameParameter_windowTooLarge, "");
1988
1989 /* Adapt buffer sizes to frame header instructions */
1990 { size_t const neededInBuffSize = MAX(zds->fParams.blockSizeMax, 4 /* frame checksum */);
1991 size_t const neededOutBuffSize = zds->outBufferMode == ZSTD_bm_buffered
1992 ? ZSTD_decodingBufferSize_min(zds->fParams.windowSize, zds->fParams.frameContentSize)
1993 : 0;
1994
1995 ZSTD_DCtx_updateOversizedDuration(zds, neededInBuffSize, neededOutBuffSize);
1996
1997 { int const tooSmall = (zds->inBuffSize < neededInBuffSize) || (zds->outBuffSize < neededOutBuffSize);
1998 int const tooLarge = ZSTD_DCtx_isOversizedTooLong(zds);
1999
2000 if (tooSmall || tooLarge) {
2001 size_t const bufferSize = neededInBuffSize + neededOutBuffSize;
2002 DEBUGLOG(4, "inBuff : from %u to %u",
2003 (U32)zds->inBuffSize, (U32)neededInBuffSize);
2004 DEBUGLOG(4, "outBuff : from %u to %u",
2005 (U32)zds->outBuffSize, (U32)neededOutBuffSize);
2006 if (zds->staticSize) { /* static DCtx */
2007 DEBUGLOG(4, "staticSize : %u", (U32)zds->staticSize);
2008 assert(zds->staticSize >= sizeof(ZSTD_DCtx)); /* controlled at init */
2009 RETURN_ERROR_IF(
2010 bufferSize > zds->staticSize - sizeof(ZSTD_DCtx),
2011 memory_allocation, "");
2012 } else {
2013 ZSTD_customFree(zds->inBuff, zds->customMem);
2014 zds->inBuffSize = 0;
2015 zds->outBuffSize = 0;
2016 zds->inBuff = (char*)ZSTD_customMalloc(bufferSize, zds->customMem);
2017 RETURN_ERROR_IF(zds->inBuff == NULL, memory_allocation, "");
2018 }
2019 zds->inBuffSize = neededInBuffSize;
2020 zds->outBuff = zds->inBuff + zds->inBuffSize;
2021 zds->outBuffSize = neededOutBuffSize;
2022 } } }
2023 zds->streamStage = zdss_read;
2024 ZSTD_FALLTHROUGH;
2025
2026 case zdss_read:
2027 DEBUGLOG(5, "stage zdss_read");
2028 { size_t const neededInSize = ZSTD_nextSrcSizeToDecompressWithInputSize(zds, (size_t)(iend - ip));
2029 DEBUGLOG(5, "neededInSize = %u", (U32)neededInSize);
2030 if (neededInSize==0) { /* end of frame */
2031 zds->streamStage = zdss_init;
2032 someMoreWork = 0;
2033 break;
2034 }
2035 if ((size_t)(iend-ip) >= neededInSize) { /* decode directly from src */
2036 FORWARD_IF_ERROR(ZSTD_decompressContinueStream(zds, &op, oend, ip, neededInSize), "");
2037 ip += neededInSize;
2038 /* Function modifies the stage so we must break */
2039 break;
2040 } }
2041 if (ip==iend) { someMoreWork = 0; break; } /* no more input */
2042 zds->streamStage = zdss_load;
2043 ZSTD_FALLTHROUGH;
2044
2045 case zdss_load:
2046 { size_t const neededInSize = ZSTD_nextSrcSizeToDecompress(zds);
2047 size_t const toLoad = neededInSize - zds->inPos;
2048 int const isSkipFrame = ZSTD_isSkipFrame(zds);
2049 size_t loadedSize;
2050 /* At this point we shouldn't be decompressing a block that we can stream. */
2051 assert(neededInSize == ZSTD_nextSrcSizeToDecompressWithInputSize(zds, iend - ip));
2052 if (isSkipFrame) {
2053 loadedSize = MIN(toLoad, (size_t)(iend-ip));
2054 } else {
2055 RETURN_ERROR_IF(toLoad > zds->inBuffSize - zds->inPos,
2056 corruption_detected,
2057 "should never happen");
2058 loadedSize = ZSTD_limitCopy(zds->inBuff + zds->inPos, toLoad, ip, (size_t)(iend-ip));
2059 }
2060 ip += loadedSize;
2061 zds->inPos += loadedSize;
2062 if (loadedSize < toLoad) { someMoreWork = 0; break; } /* not enough input, wait for more */
2063
2064 /* decode loaded input */
2065 zds->inPos = 0; /* input is consumed */
2066 FORWARD_IF_ERROR(ZSTD_decompressContinueStream(zds, &op, oend, zds->inBuff, neededInSize), "");
2067 /* Function modifies the stage so we must break */
2068 break;
2069 }
2070 case zdss_flush:
2071 { size_t const toFlushSize = zds->outEnd - zds->outStart;
2072 size_t const flushedSize = ZSTD_limitCopy(op, (size_t)(oend-op), zds->outBuff + zds->outStart, toFlushSize);
2073 op += flushedSize;
2074 zds->outStart += flushedSize;
2075 if (flushedSize == toFlushSize) { /* flush completed */
2076 zds->streamStage = zdss_read;
2077 if ( (zds->outBuffSize < zds->fParams.frameContentSize)
2078 && (zds->outStart + zds->fParams.blockSizeMax > zds->outBuffSize) ) {
2079 DEBUGLOG(5, "restart filling outBuff from beginning (left:%i, needed:%u)",
2080 (int)(zds->outBuffSize - zds->outStart),
2081 (U32)zds->fParams.blockSizeMax);
2082 zds->outStart = zds->outEnd = 0;
2083 }
2084 break;
2085 } }
2086 /* cannot complete flush */
2087 someMoreWork = 0;
2088 break;
2089
2090 default:
2091 assert(0); /* impossible */
2092 RETURN_ERROR(GENERIC, "impossible to reach"); /* some compiler require default to do something */
2093 } }
2094
2095 /* result */
2096 input->pos = (size_t)(ip - (const char*)(input->src));
2097 output->pos = (size_t)(op - (char*)(output->dst));
2098
2099 /* Update the expected output buffer for ZSTD_obm_stable. */
2100 zds->expectedOutBuffer = *output;
2101
2102 if ((ip==istart) && (op==ostart)) { /* no forward progress */
2103 zds->noForwardProgress ++;
2104 if (zds->noForwardProgress >= ZSTD_NO_FORWARD_PROGRESS_MAX) {
2105 RETURN_ERROR_IF(op==oend, dstSize_tooSmall, "");
2106 RETURN_ERROR_IF(ip==iend, srcSize_wrong, "");
2107 assert(0);
2108 }
2109 } else {
2110 zds->noForwardProgress = 0;
2111 }
2112 { size_t nextSrcSizeHint = ZSTD_nextSrcSizeToDecompress(zds);
2113 if (!nextSrcSizeHint) { /* frame fully decoded */
2114 if (zds->outEnd == zds->outStart) { /* output fully flushed */
2115 if (zds->hostageByte) {
2116 if (input->pos >= input->size) {
2117 /* can't release hostage (not present) */
2118 zds->streamStage = zdss_read;
2119 return 1;
2120 }
2121 input->pos++; /* release hostage */
2122 } /* zds->hostageByte */
2123 return 0;
2124 } /* zds->outEnd == zds->outStart */
2125 if (!zds->hostageByte) { /* output not fully flushed; keep last byte as hostage; will be released when all output is flushed */
2126 input->pos--; /* note : pos > 0, otherwise, impossible to finish reading last block */
2127 zds->hostageByte=1;
2128 }
2129 return 1;
2130 } /* nextSrcSizeHint==0 */
2131 nextSrcSizeHint += ZSTD_blockHeaderSize * (ZSTD_nextInputType(zds) == ZSTDnit_block); /* preload header of next block */
2132 assert(zds->inPos <= nextSrcSizeHint);
2133 nextSrcSizeHint -= zds->inPos; /* part already loaded*/
2134 return nextSrcSizeHint;
2135 }
2136 }
2137
ZSTD_decompressStream_simpleArgs(ZSTD_DCtx * dctx,void * dst,size_t dstCapacity,size_t * dstPos,const void * src,size_t srcSize,size_t * srcPos)2138 size_t ZSTD_decompressStream_simpleArgs (
2139 ZSTD_DCtx* dctx,
2140 void* dst, size_t dstCapacity, size_t* dstPos,
2141 const void* src, size_t srcSize, size_t* srcPos)
2142 {
2143 ZSTD_outBuffer output = { dst, dstCapacity, *dstPos };
2144 ZSTD_inBuffer input = { src, srcSize, *srcPos };
2145 /* ZSTD_compress_generic() will check validity of dstPos and srcPos */
2146 size_t const cErr = ZSTD_decompressStream(dctx, &output, &input);
2147 *dstPos = output.pos;
2148 *srcPos = input.pos;
2149 return cErr;
2150 }
2151