zstd-ruby 1.5.5.0 → 1.5.6.0
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +4 -4
- data/.gitignore +2 -0
- data/README.md +2 -2
- data/ext/zstdruby/extconf.rb +2 -0
- data/ext/zstdruby/libzstd/common/allocations.h +1 -1
- data/ext/zstdruby/libzstd/common/bitstream.h +49 -29
- data/ext/zstdruby/libzstd/common/compiler.h +114 -22
- data/ext/zstdruby/libzstd/common/cpu.h +36 -0
- data/ext/zstdruby/libzstd/common/debug.c +6 -0
- data/ext/zstdruby/libzstd/common/debug.h +20 -11
- data/ext/zstdruby/libzstd/common/error_private.h +45 -36
- data/ext/zstdruby/libzstd/common/fse.h +3 -2
- data/ext/zstdruby/libzstd/common/fse_decompress.c +19 -17
- data/ext/zstdruby/libzstd/common/huf.h +14 -1
- data/ext/zstdruby/libzstd/common/mem.h +0 -9
- data/ext/zstdruby/libzstd/common/pool.c +1 -1
- data/ext/zstdruby/libzstd/common/pool.h +1 -1
- data/ext/zstdruby/libzstd/common/portability_macros.h +2 -0
- data/ext/zstdruby/libzstd/common/threading.c +8 -2
- data/ext/zstdruby/libzstd/common/xxhash.c +5 -11
- data/ext/zstdruby/libzstd/common/xxhash.h +2341 -1007
- data/ext/zstdruby/libzstd/common/zstd_internal.h +5 -5
- data/ext/zstdruby/libzstd/compress/fse_compress.c +8 -7
- data/ext/zstdruby/libzstd/compress/huf_compress.c +54 -25
- data/ext/zstdruby/libzstd/compress/zstd_compress.c +282 -161
- data/ext/zstdruby/libzstd/compress/zstd_compress_internal.h +29 -27
- data/ext/zstdruby/libzstd/compress/zstd_compress_superblock.c +224 -113
- data/ext/zstdruby/libzstd/compress/zstd_cwksp.h +19 -13
- data/ext/zstdruby/libzstd/compress/zstd_double_fast.c +17 -5
- data/ext/zstdruby/libzstd/compress/zstd_double_fast.h +11 -0
- data/ext/zstdruby/libzstd/compress/zstd_fast.c +14 -6
- data/ext/zstdruby/libzstd/compress/zstd_lazy.c +129 -87
- data/ext/zstdruby/libzstd/compress/zstd_lazy.h +103 -28
- data/ext/zstdruby/libzstd/compress/zstd_ldm.c +8 -2
- data/ext/zstdruby/libzstd/compress/zstd_opt.c +216 -112
- data/ext/zstdruby/libzstd/compress/zstd_opt.h +31 -7
- data/ext/zstdruby/libzstd/compress/zstdmt_compress.c +94 -79
- data/ext/zstdruby/libzstd/decompress/huf_decompress.c +188 -126
- data/ext/zstdruby/libzstd/decompress/huf_decompress_amd64.S +38 -19
- data/ext/zstdruby/libzstd/decompress/zstd_decompress.c +84 -32
- data/ext/zstdruby/libzstd/decompress/zstd_decompress_block.c +231 -208
- data/ext/zstdruby/libzstd/decompress/zstd_decompress_block.h +1 -1
- data/ext/zstdruby/libzstd/decompress/zstd_decompress_internal.h +2 -0
- data/ext/zstdruby/libzstd/dictBuilder/cover.c +16 -12
- data/ext/zstdruby/libzstd/dictBuilder/cover.h +2 -8
- data/ext/zstdruby/libzstd/dictBuilder/fastcover.c +2 -2
- data/ext/zstdruby/libzstd/dictBuilder/zdict.c +12 -6
- data/ext/zstdruby/libzstd/zstd.h +129 -60
- data/ext/zstdruby/streaming_compress.c +23 -3
- data/ext/zstdruby/streaming_decompress.c +23 -3
- data/lib/zstd-ruby/version.rb +1 -1
- metadata +2 -2
@@ -51,6 +51,13 @@ static void ZSTD_copy4(void* dst, const void* src) { ZSTD_memcpy(dst, src, 4); }
|
|
51
51
|
* Block decoding
|
52
52
|
***************************************************************/
|
53
53
|
|
54
|
+
static size_t ZSTD_blockSizeMax(ZSTD_DCtx const* dctx)
|
55
|
+
{
|
56
|
+
size_t const blockSizeMax = dctx->isFrameDecompression ? dctx->fParams.blockSizeMax : ZSTD_BLOCKSIZE_MAX;
|
57
|
+
assert(blockSizeMax <= ZSTD_BLOCKSIZE_MAX);
|
58
|
+
return blockSizeMax;
|
59
|
+
}
|
60
|
+
|
54
61
|
/*! ZSTD_getcBlockSize() :
|
55
62
|
* Provides the size of compressed block from block header `src` */
|
56
63
|
size_t ZSTD_getcBlockSize(const void* src, size_t srcSize,
|
@@ -73,41 +80,49 @@ size_t ZSTD_getcBlockSize(const void* src, size_t srcSize,
|
|
73
80
|
static void ZSTD_allocateLiteralsBuffer(ZSTD_DCtx* dctx, void* const dst, const size_t dstCapacity, const size_t litSize,
|
74
81
|
const streaming_operation streaming, const size_t expectedWriteSize, const unsigned splitImmediately)
|
75
82
|
{
|
76
|
-
|
77
|
-
|
78
|
-
|
79
|
-
|
83
|
+
size_t const blockSizeMax = ZSTD_blockSizeMax(dctx);
|
84
|
+
assert(litSize <= blockSizeMax);
|
85
|
+
assert(dctx->isFrameDecompression || streaming == not_streaming);
|
86
|
+
assert(expectedWriteSize <= blockSizeMax);
|
87
|
+
if (streaming == not_streaming && dstCapacity > blockSizeMax + WILDCOPY_OVERLENGTH + litSize + WILDCOPY_OVERLENGTH) {
|
88
|
+
/* If we aren't streaming, we can just put the literals after the output
|
89
|
+
* of the current block. We don't need to worry about overwriting the
|
90
|
+
* extDict of our window, because it doesn't exist.
|
91
|
+
* So if we have space after the end of the block, just put it there.
|
92
|
+
*/
|
93
|
+
dctx->litBuffer = (BYTE*)dst + blockSizeMax + WILDCOPY_OVERLENGTH;
|
80
94
|
dctx->litBufferEnd = dctx->litBuffer + litSize;
|
81
95
|
dctx->litBufferLocation = ZSTD_in_dst;
|
82
|
-
}
|
83
|
-
|
84
|
-
|
85
|
-
|
96
|
+
} else if (litSize <= ZSTD_LITBUFFEREXTRASIZE) {
|
97
|
+
/* Literals fit entirely within the extra buffer, put them there to avoid
|
98
|
+
* having to split the literals.
|
99
|
+
*/
|
100
|
+
dctx->litBuffer = dctx->litExtraBuffer;
|
101
|
+
dctx->litBufferEnd = dctx->litBuffer + litSize;
|
102
|
+
dctx->litBufferLocation = ZSTD_not_in_dst;
|
103
|
+
} else {
|
104
|
+
assert(blockSizeMax > ZSTD_LITBUFFEREXTRASIZE);
|
105
|
+
/* Literals must be split between the output block and the extra lit
|
106
|
+
* buffer. We fill the extra lit buffer with the tail of the literals,
|
107
|
+
* and put the rest of the literals at the end of the block, with
|
108
|
+
* WILDCOPY_OVERLENGTH of buffer room to allow for overreads.
|
109
|
+
* This MUST not write more than our maxBlockSize beyond dst, because in
|
110
|
+
* streaming mode, that could overwrite part of our extDict window.
|
111
|
+
*/
|
86
112
|
if (splitImmediately) {
|
87
113
|
/* won't fit in litExtraBuffer, so it will be split between end of dst and extra buffer */
|
88
114
|
dctx->litBuffer = (BYTE*)dst + expectedWriteSize - litSize + ZSTD_LITBUFFEREXTRASIZE - WILDCOPY_OVERLENGTH;
|
89
115
|
dctx->litBufferEnd = dctx->litBuffer + litSize - ZSTD_LITBUFFEREXTRASIZE;
|
90
|
-
}
|
91
|
-
else {
|
116
|
+
} else {
|
92
117
|
/* initially this will be stored entirely in dst during huffman decoding, it will partially be shifted to litExtraBuffer after */
|
93
118
|
dctx->litBuffer = (BYTE*)dst + expectedWriteSize - litSize;
|
94
119
|
dctx->litBufferEnd = (BYTE*)dst + expectedWriteSize;
|
95
120
|
}
|
96
121
|
dctx->litBufferLocation = ZSTD_split;
|
97
|
-
|
98
|
-
else
|
99
|
-
{
|
100
|
-
/* fits entirely within litExtraBuffer, so no split is necessary */
|
101
|
-
dctx->litBuffer = dctx->litExtraBuffer;
|
102
|
-
dctx->litBufferEnd = dctx->litBuffer + litSize;
|
103
|
-
dctx->litBufferLocation = ZSTD_not_in_dst;
|
122
|
+
assert(dctx->litBufferEnd <= (BYTE*)dst + expectedWriteSize);
|
104
123
|
}
|
105
124
|
}
|
106
125
|
|
107
|
-
/* Hidden declaration for fullbench */
|
108
|
-
size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
|
109
|
-
const void* src, size_t srcSize,
|
110
|
-
void* dst, size_t dstCapacity, const streaming_operation streaming);
|
111
126
|
/*! ZSTD_decodeLiteralsBlock() :
|
112
127
|
* Where it is possible to do so without being stomped by the output during decompression, the literals block will be stored
|
113
128
|
* in the dstBuffer. If there is room to do so, it will be stored in full in the excess dst space after where the current
|
@@ -116,7 +131,7 @@ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
|
|
116
131
|
*
|
117
132
|
* @return : nb of bytes read from src (< srcSize )
|
118
133
|
* note : symbol not declared but exposed for fullbench */
|
119
|
-
size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
|
134
|
+
static size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
|
120
135
|
const void* src, size_t srcSize, /* note : srcSize < BLOCKSIZE */
|
121
136
|
void* dst, size_t dstCapacity, const streaming_operation streaming)
|
122
137
|
{
|
@@ -125,6 +140,7 @@ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
|
|
125
140
|
|
126
141
|
{ const BYTE* const istart = (const BYTE*) src;
|
127
142
|
symbolEncodingType_e const litEncType = (symbolEncodingType_e)(istart[0] & 3);
|
143
|
+
size_t const blockSizeMax = ZSTD_blockSizeMax(dctx);
|
128
144
|
|
129
145
|
switch(litEncType)
|
130
146
|
{
|
@@ -140,7 +156,7 @@ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
|
|
140
156
|
U32 const lhlCode = (istart[0] >> 2) & 3;
|
141
157
|
U32 const lhc = MEM_readLE32(istart);
|
142
158
|
size_t hufSuccess;
|
143
|
-
size_t expectedWriteSize = MIN(
|
159
|
+
size_t expectedWriteSize = MIN(blockSizeMax, dstCapacity);
|
144
160
|
int const flags = 0
|
145
161
|
| (ZSTD_DCtx_get_bmi2(dctx) ? HUF_flags_bmi2 : 0)
|
146
162
|
| (dctx->disableHufAsm ? HUF_flags_disableAsm : 0);
|
@@ -167,7 +183,7 @@ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
|
|
167
183
|
break;
|
168
184
|
}
|
169
185
|
RETURN_ERROR_IF(litSize > 0 && dst == NULL, dstSize_tooSmall, "NULL not handled");
|
170
|
-
RETURN_ERROR_IF(litSize >
|
186
|
+
RETURN_ERROR_IF(litSize > blockSizeMax, corruption_detected, "");
|
171
187
|
if (!singleStream)
|
172
188
|
RETURN_ERROR_IF(litSize < MIN_LITERALS_FOR_4_STREAMS, literals_headerWrong,
|
173
189
|
"Not enough literals (%zu) for the 4-streams mode (min %u)",
|
@@ -214,10 +230,12 @@ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
|
|
214
230
|
}
|
215
231
|
if (dctx->litBufferLocation == ZSTD_split)
|
216
232
|
{
|
233
|
+
assert(litSize > ZSTD_LITBUFFEREXTRASIZE);
|
217
234
|
ZSTD_memcpy(dctx->litExtraBuffer, dctx->litBufferEnd - ZSTD_LITBUFFEREXTRASIZE, ZSTD_LITBUFFEREXTRASIZE);
|
218
235
|
ZSTD_memmove(dctx->litBuffer + ZSTD_LITBUFFEREXTRASIZE - WILDCOPY_OVERLENGTH, dctx->litBuffer, litSize - ZSTD_LITBUFFEREXTRASIZE);
|
219
236
|
dctx->litBuffer += ZSTD_LITBUFFEREXTRASIZE - WILDCOPY_OVERLENGTH;
|
220
237
|
dctx->litBufferEnd -= WILDCOPY_OVERLENGTH;
|
238
|
+
assert(dctx->litBufferEnd <= (BYTE*)dst + blockSizeMax);
|
221
239
|
}
|
222
240
|
|
223
241
|
RETURN_ERROR_IF(HUF_isError(hufSuccess), corruption_detected, "");
|
@@ -232,7 +250,7 @@ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
|
|
232
250
|
case set_basic:
|
233
251
|
{ size_t litSize, lhSize;
|
234
252
|
U32 const lhlCode = ((istart[0]) >> 2) & 3;
|
235
|
-
size_t expectedWriteSize = MIN(
|
253
|
+
size_t expectedWriteSize = MIN(blockSizeMax, dstCapacity);
|
236
254
|
switch(lhlCode)
|
237
255
|
{
|
238
256
|
case 0: case 2: default: /* note : default is impossible, since lhlCode into [0..3] */
|
@@ -251,6 +269,7 @@ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
|
|
251
269
|
}
|
252
270
|
|
253
271
|
RETURN_ERROR_IF(litSize > 0 && dst == NULL, dstSize_tooSmall, "NULL not handled");
|
272
|
+
RETURN_ERROR_IF(litSize > blockSizeMax, corruption_detected, "");
|
254
273
|
RETURN_ERROR_IF(expectedWriteSize < litSize, dstSize_tooSmall, "");
|
255
274
|
ZSTD_allocateLiteralsBuffer(dctx, dst, dstCapacity, litSize, streaming, expectedWriteSize, 1);
|
256
275
|
if (lhSize+litSize+WILDCOPY_OVERLENGTH > srcSize) { /* risk reading beyond src buffer with wildcopy */
|
@@ -279,7 +298,7 @@ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
|
|
279
298
|
case set_rle:
|
280
299
|
{ U32 const lhlCode = ((istart[0]) >> 2) & 3;
|
281
300
|
size_t litSize, lhSize;
|
282
|
-
size_t expectedWriteSize = MIN(
|
301
|
+
size_t expectedWriteSize = MIN(blockSizeMax, dstCapacity);
|
283
302
|
switch(lhlCode)
|
284
303
|
{
|
285
304
|
case 0: case 2: default: /* note : default is impossible, since lhlCode into [0..3] */
|
@@ -298,7 +317,7 @@ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
|
|
298
317
|
break;
|
299
318
|
}
|
300
319
|
RETURN_ERROR_IF(litSize > 0 && dst == NULL, dstSize_tooSmall, "NULL not handled");
|
301
|
-
RETURN_ERROR_IF(litSize >
|
320
|
+
RETURN_ERROR_IF(litSize > blockSizeMax, corruption_detected, "");
|
302
321
|
RETURN_ERROR_IF(expectedWriteSize < litSize, dstSize_tooSmall, "");
|
303
322
|
ZSTD_allocateLiteralsBuffer(dctx, dst, dstCapacity, litSize, streaming, expectedWriteSize, 1);
|
304
323
|
if (dctx->litBufferLocation == ZSTD_split)
|
@@ -320,6 +339,18 @@ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
|
|
320
339
|
}
|
321
340
|
}
|
322
341
|
|
342
|
+
/* Hidden declaration for fullbench */
|
343
|
+
size_t ZSTD_decodeLiteralsBlock_wrapper(ZSTD_DCtx* dctx,
|
344
|
+
const void* src, size_t srcSize,
|
345
|
+
void* dst, size_t dstCapacity);
|
346
|
+
size_t ZSTD_decodeLiteralsBlock_wrapper(ZSTD_DCtx* dctx,
|
347
|
+
const void* src, size_t srcSize,
|
348
|
+
void* dst, size_t dstCapacity)
|
349
|
+
{
|
350
|
+
dctx->isFrameDecompression = 0;
|
351
|
+
return ZSTD_decodeLiteralsBlock(dctx, src, srcSize, dst, dstCapacity, not_streaming);
|
352
|
+
}
|
353
|
+
|
323
354
|
/* Default FSE distribution tables.
|
324
355
|
* These are pre-calculated FSE decoding tables using default distributions as defined in specification :
|
325
356
|
* https://github.com/facebook/zstd/blob/release/doc/zstd_compression_format.md#default-distributions
|
@@ -675,11 +706,6 @@ size_t ZSTD_decodeSeqHeaders(ZSTD_DCtx* dctx, int* nbSeqPtr,
|
|
675
706
|
|
676
707
|
/* SeqHead */
|
677
708
|
nbSeq = *ip++;
|
678
|
-
if (!nbSeq) {
|
679
|
-
*nbSeqPtr=0;
|
680
|
-
RETURN_ERROR_IF(srcSize != 1, srcSize_wrong, "");
|
681
|
-
return 1;
|
682
|
-
}
|
683
709
|
if (nbSeq > 0x7F) {
|
684
710
|
if (nbSeq == 0xFF) {
|
685
711
|
RETURN_ERROR_IF(ip+2 > iend, srcSize_wrong, "");
|
@@ -692,8 +718,16 @@ size_t ZSTD_decodeSeqHeaders(ZSTD_DCtx* dctx, int* nbSeqPtr,
|
|
692
718
|
}
|
693
719
|
*nbSeqPtr = nbSeq;
|
694
720
|
|
721
|
+
if (nbSeq == 0) {
|
722
|
+
/* No sequence : section ends immediately */
|
723
|
+
RETURN_ERROR_IF(ip != iend, corruption_detected,
|
724
|
+
"extraneous data present in the Sequences section");
|
725
|
+
return (size_t)(ip - istart);
|
726
|
+
}
|
727
|
+
|
695
728
|
/* FSE table descriptors */
|
696
729
|
RETURN_ERROR_IF(ip+1 > iend, srcSize_wrong, ""); /* minimum possible size: 1 byte for symbol encoding types */
|
730
|
+
RETURN_ERROR_IF(*ip & 3, corruption_detected, ""); /* The last field, Reserved, must be all-zeroes. */
|
697
731
|
{ symbolEncodingType_e const LLtype = (symbolEncodingType_e)(*ip >> 6);
|
698
732
|
symbolEncodingType_e const OFtype = (symbolEncodingType_e)((*ip >> 4) & 3);
|
699
733
|
symbolEncodingType_e const MLtype = (symbolEncodingType_e)((*ip >> 2) & 3);
|
@@ -840,7 +874,7 @@ static void ZSTD_safecopy(BYTE* op, const BYTE* const oend_w, BYTE const* ip, pt
|
|
840
874
|
/* ZSTD_safecopyDstBeforeSrc():
|
841
875
|
* This version allows overlap with dst before src, or handles the non-overlap case with dst after src
|
842
876
|
* Kept separate from more common ZSTD_safecopy case to avoid performance impact to the safecopy common case */
|
843
|
-
static void ZSTD_safecopyDstBeforeSrc(BYTE* op, BYTE
|
877
|
+
static void ZSTD_safecopyDstBeforeSrc(BYTE* op, const BYTE* ip, ptrdiff_t length) {
|
844
878
|
ptrdiff_t const diff = op - ip;
|
845
879
|
BYTE* const oend = op + length;
|
846
880
|
|
@@ -869,6 +903,7 @@ static void ZSTD_safecopyDstBeforeSrc(BYTE* op, BYTE const* ip, ptrdiff_t length
|
|
869
903
|
* to be optimized for many small sequences, since those fall into ZSTD_execSequence().
|
870
904
|
*/
|
871
905
|
FORCE_NOINLINE
|
906
|
+
ZSTD_ALLOW_POINTER_OVERFLOW_ATTR
|
872
907
|
size_t ZSTD_execSequenceEnd(BYTE* op,
|
873
908
|
BYTE* const oend, seq_t sequence,
|
874
909
|
const BYTE** litPtr, const BYTE* const litLimit,
|
@@ -916,6 +951,7 @@ size_t ZSTD_execSequenceEnd(BYTE* op,
|
|
916
951
|
* This version is intended to be used during instances where the litBuffer is still split. It is kept separate to avoid performance impact for the good case.
|
917
952
|
*/
|
918
953
|
FORCE_NOINLINE
|
954
|
+
ZSTD_ALLOW_POINTER_OVERFLOW_ATTR
|
919
955
|
size_t ZSTD_execSequenceEndSplitLitBuffer(BYTE* op,
|
920
956
|
BYTE* const oend, const BYTE* const oend_w, seq_t sequence,
|
921
957
|
const BYTE** litPtr, const BYTE* const litLimit,
|
@@ -961,6 +997,7 @@ size_t ZSTD_execSequenceEndSplitLitBuffer(BYTE* op,
|
|
961
997
|
}
|
962
998
|
|
963
999
|
HINT_INLINE
|
1000
|
+
ZSTD_ALLOW_POINTER_OVERFLOW_ATTR
|
964
1001
|
size_t ZSTD_execSequence(BYTE* op,
|
965
1002
|
BYTE* const oend, seq_t sequence,
|
966
1003
|
const BYTE** litPtr, const BYTE* const litLimit,
|
@@ -1059,6 +1096,7 @@ size_t ZSTD_execSequence(BYTE* op,
|
|
1059
1096
|
}
|
1060
1097
|
|
1061
1098
|
HINT_INLINE
|
1099
|
+
ZSTD_ALLOW_POINTER_OVERFLOW_ATTR
|
1062
1100
|
size_t ZSTD_execSequenceSplitLitBuffer(BYTE* op,
|
1063
1101
|
BYTE* const oend, const BYTE* const oend_w, seq_t sequence,
|
1064
1102
|
const BYTE** litPtr, const BYTE* const litLimit,
|
@@ -1181,14 +1219,20 @@ ZSTD_updateFseStateWithDInfo(ZSTD_fseState* DStatePtr, BIT_DStream_t* bitD, U16
|
|
1181
1219
|
|
1182
1220
|
typedef enum { ZSTD_lo_isRegularOffset, ZSTD_lo_isLongOffset=1 } ZSTD_longOffset_e;
|
1183
1221
|
|
1222
|
+
/**
|
1223
|
+
* ZSTD_decodeSequence():
|
1224
|
+
* @p longOffsets : tells the decoder to reload more bit while decoding large offsets
|
1225
|
+
* only used in 32-bit mode
|
1226
|
+
* @return : Sequence (litL + matchL + offset)
|
1227
|
+
*/
|
1184
1228
|
FORCE_INLINE_TEMPLATE seq_t
|
1185
|
-
ZSTD_decodeSequence(seqState_t* seqState, const ZSTD_longOffset_e longOffsets)
|
1229
|
+
ZSTD_decodeSequence(seqState_t* seqState, const ZSTD_longOffset_e longOffsets, const int isLastSeq)
|
1186
1230
|
{
|
1187
1231
|
seq_t seq;
|
1188
1232
|
/*
|
1189
|
-
* ZSTD_seqSymbol is a
|
1190
|
-
* loaded in one operation
|
1191
|
-
* bit-extracting on aarch64.
|
1233
|
+
* ZSTD_seqSymbol is a 64 bits wide structure.
|
1234
|
+
* It can be loaded in one operation
|
1235
|
+
* and its fields extracted by simply shifting or bit-extracting on aarch64.
|
1192
1236
|
* GCC doesn't recognize this and generates more unnecessary ldr/ldrb/ldrh
|
1193
1237
|
* operations that cause performance drop. This can be avoided by using this
|
1194
1238
|
* ZSTD_memcpy hack.
|
@@ -1261,7 +1305,7 @@ ZSTD_decodeSequence(seqState_t* seqState, const ZSTD_longOffset_e longOffsets)
|
|
1261
1305
|
} else {
|
1262
1306
|
offset = ofBase + ll0 + BIT_readBitsFast(&seqState->DStream, 1);
|
1263
1307
|
{ size_t temp = (offset==3) ? seqState->prevOffset[0] - 1 : seqState->prevOffset[offset];
|
1264
|
-
temp
|
1308
|
+
temp -= !temp; /* 0 is not valid: input corrupted => force offset to -1 => corruption detected at execSequence */
|
1265
1309
|
if (offset != 1) seqState->prevOffset[2] = seqState->prevOffset[1];
|
1266
1310
|
seqState->prevOffset[1] = seqState->prevOffset[0];
|
1267
1311
|
seqState->prevOffset[0] = offset = temp;
|
@@ -1288,17 +1332,22 @@ ZSTD_decodeSequence(seqState_t* seqState, const ZSTD_longOffset_e longOffsets)
|
|
1288
1332
|
DEBUGLOG(6, "seq: litL=%u, matchL=%u, offset=%u",
|
1289
1333
|
(U32)seq.litLength, (U32)seq.matchLength, (U32)seq.offset);
|
1290
1334
|
|
1291
|
-
|
1292
|
-
|
1293
|
-
|
1294
|
-
|
1335
|
+
if (!isLastSeq) {
|
1336
|
+
/* don't update FSE state for last Sequence */
|
1337
|
+
ZSTD_updateFseStateWithDInfo(&seqState->stateLL, &seqState->DStream, llNext, llnbBits); /* <= 9 bits */
|
1338
|
+
ZSTD_updateFseStateWithDInfo(&seqState->stateML, &seqState->DStream, mlNext, mlnbBits); /* <= 9 bits */
|
1339
|
+
if (MEM_32bits()) BIT_reloadDStream(&seqState->DStream); /* <= 18 bits */
|
1340
|
+
ZSTD_updateFseStateWithDInfo(&seqState->stateOffb, &seqState->DStream, ofNext, ofnbBits); /* <= 8 bits */
|
1341
|
+
BIT_reloadDStream(&seqState->DStream);
|
1342
|
+
}
|
1295
1343
|
}
|
1296
1344
|
|
1297
1345
|
return seq;
|
1298
1346
|
}
|
1299
1347
|
|
1300
|
-
#
|
1301
|
-
|
1348
|
+
#if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE)
|
1349
|
+
#if DEBUGLEVEL >= 1
|
1350
|
+
static int ZSTD_dictionaryIsActive(ZSTD_DCtx const* dctx, BYTE const* prefixStart, BYTE const* oLitEnd)
|
1302
1351
|
{
|
1303
1352
|
size_t const windowSize = dctx->fParams.windowSize;
|
1304
1353
|
/* No dictionary used. */
|
@@ -1312,30 +1361,33 @@ MEM_STATIC int ZSTD_dictionaryIsActive(ZSTD_DCtx const* dctx, BYTE const* prefix
|
|
1312
1361
|
/* Dictionary is active. */
|
1313
1362
|
return 1;
|
1314
1363
|
}
|
1364
|
+
#endif
|
1315
1365
|
|
1316
|
-
|
1366
|
+
static void ZSTD_assertValidSequence(
|
1317
1367
|
ZSTD_DCtx const* dctx,
|
1318
1368
|
BYTE const* op, BYTE const* oend,
|
1319
1369
|
seq_t const seq,
|
1320
1370
|
BYTE const* prefixStart, BYTE const* virtualStart)
|
1321
1371
|
{
|
1322
1372
|
#if DEBUGLEVEL >= 1
|
1323
|
-
|
1324
|
-
|
1325
|
-
|
1326
|
-
|
1327
|
-
|
1328
|
-
|
1329
|
-
|
1330
|
-
|
1331
|
-
|
1332
|
-
|
1333
|
-
|
1334
|
-
|
1335
|
-
|
1336
|
-
|
1337
|
-
|
1338
|
-
|
1373
|
+
if (dctx->isFrameDecompression) {
|
1374
|
+
size_t const windowSize = dctx->fParams.windowSize;
|
1375
|
+
size_t const sequenceSize = seq.litLength + seq.matchLength;
|
1376
|
+
BYTE const* const oLitEnd = op + seq.litLength;
|
1377
|
+
DEBUGLOG(6, "Checking sequence: litL=%u matchL=%u offset=%u",
|
1378
|
+
(U32)seq.litLength, (U32)seq.matchLength, (U32)seq.offset);
|
1379
|
+
assert(op <= oend);
|
1380
|
+
assert((size_t)(oend - op) >= sequenceSize);
|
1381
|
+
assert(sequenceSize <= ZSTD_blockSizeMax(dctx));
|
1382
|
+
if (ZSTD_dictionaryIsActive(dctx, prefixStart, oLitEnd)) {
|
1383
|
+
size_t const dictSize = (size_t)((char const*)dctx->dictContentEndForFuzzing - (char const*)dctx->dictContentBeginForFuzzing);
|
1384
|
+
/* Offset must be within the dictionary. */
|
1385
|
+
assert(seq.offset <= (size_t)(oLitEnd - virtualStart));
|
1386
|
+
assert(seq.offset <= windowSize + dictSize);
|
1387
|
+
} else {
|
1388
|
+
/* Offset must be within our window. */
|
1389
|
+
assert(seq.offset <= windowSize);
|
1390
|
+
}
|
1339
1391
|
}
|
1340
1392
|
#else
|
1341
1393
|
(void)dctx, (void)op, (void)oend, (void)seq, (void)prefixStart, (void)virtualStart;
|
@@ -1351,23 +1403,21 @@ DONT_VECTORIZE
|
|
1351
1403
|
ZSTD_decompressSequences_bodySplitLitBuffer( ZSTD_DCtx* dctx,
|
1352
1404
|
void* dst, size_t maxDstSize,
|
1353
1405
|
const void* seqStart, size_t seqSize, int nbSeq,
|
1354
|
-
const ZSTD_longOffset_e isLongOffset
|
1355
|
-
const int frame)
|
1406
|
+
const ZSTD_longOffset_e isLongOffset)
|
1356
1407
|
{
|
1357
1408
|
const BYTE* ip = (const BYTE*)seqStart;
|
1358
1409
|
const BYTE* const iend = ip + seqSize;
|
1359
1410
|
BYTE* const ostart = (BYTE*)dst;
|
1360
|
-
BYTE* const oend = ostart
|
1411
|
+
BYTE* const oend = ZSTD_maybeNullPtrAdd(ostart, maxDstSize);
|
1361
1412
|
BYTE* op = ostart;
|
1362
1413
|
const BYTE* litPtr = dctx->litPtr;
|
1363
1414
|
const BYTE* litBufferEnd = dctx->litBufferEnd;
|
1364
1415
|
const BYTE* const prefixStart = (const BYTE*) (dctx->prefixStart);
|
1365
1416
|
const BYTE* const vBase = (const BYTE*) (dctx->virtualStart);
|
1366
1417
|
const BYTE* const dictEnd = (const BYTE*) (dctx->dictEnd);
|
1367
|
-
DEBUGLOG(5, "ZSTD_decompressSequences_bodySplitLitBuffer");
|
1368
|
-
(void)frame;
|
1418
|
+
DEBUGLOG(5, "ZSTD_decompressSequences_bodySplitLitBuffer (%i seqs)", nbSeq);
|
1369
1419
|
|
1370
|
-
/*
|
1420
|
+
/* Literals are split between internal buffer & output buffer */
|
1371
1421
|
if (nbSeq) {
|
1372
1422
|
seqState_t seqState;
|
1373
1423
|
dctx->fseEntropy = 1;
|
@@ -1386,8 +1436,7 @@ ZSTD_decompressSequences_bodySplitLitBuffer( ZSTD_DCtx* dctx,
|
|
1386
1436
|
BIT_DStream_completed < BIT_DStream_overflow);
|
1387
1437
|
|
1388
1438
|
/* decompress without overrunning litPtr begins */
|
1389
|
-
{
|
1390
|
-
seq_t sequence = ZSTD_decodeSequence(&seqState, isLongOffset);
|
1439
|
+
{ seq_t sequence = {0,0,0}; /* some static analyzer believe that @sequence is not initialized (it necessarily is, since for(;;) loop as at least one iteration) */
|
1391
1440
|
/* Align the decompression loop to 32 + 16 bytes.
|
1392
1441
|
*
|
1393
1442
|
* zstd compiled with gcc-9 on an Intel i9-9900k shows 10% decompression
|
@@ -1449,27 +1498,26 @@ ZSTD_decompressSequences_bodySplitLitBuffer( ZSTD_DCtx* dctx,
|
|
1449
1498
|
#endif
|
1450
1499
|
|
1451
1500
|
/* Handle the initial state where litBuffer is currently split between dst and litExtraBuffer */
|
1452
|
-
for (;
|
1453
|
-
|
1501
|
+
for ( ; nbSeq; nbSeq--) {
|
1502
|
+
sequence = ZSTD_decodeSequence(&seqState, isLongOffset, nbSeq==1);
|
1503
|
+
if (litPtr + sequence.litLength > dctx->litBufferEnd) break;
|
1504
|
+
{ size_t const oneSeqSize = ZSTD_execSequenceSplitLitBuffer(op, oend, litPtr + sequence.litLength - WILDCOPY_OVERLENGTH, sequence, &litPtr, litBufferEnd, prefixStart, vBase, dictEnd);
|
1454
1505
|
#if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE)
|
1455
|
-
|
1456
|
-
|
1506
|
+
assert(!ZSTD_isError(oneSeqSize));
|
1507
|
+
ZSTD_assertValidSequence(dctx, op, oend, sequence, prefixStart, vBase);
|
1457
1508
|
#endif
|
1458
|
-
|
1459
|
-
|
1460
|
-
|
1461
|
-
|
1462
|
-
|
1463
|
-
|
1464
|
-
BIT_reloadDStream(&(seqState.DStream));
|
1465
|
-
sequence = ZSTD_decodeSequence(&seqState, isLongOffset);
|
1466
|
-
}
|
1509
|
+
if (UNLIKELY(ZSTD_isError(oneSeqSize)))
|
1510
|
+
return oneSeqSize;
|
1511
|
+
DEBUGLOG(6, "regenerated sequence size : %u", (U32)oneSeqSize);
|
1512
|
+
op += oneSeqSize;
|
1513
|
+
} }
|
1514
|
+
DEBUGLOG(6, "reached: (litPtr + sequence.litLength > dctx->litBufferEnd)");
|
1467
1515
|
|
1468
1516
|
/* If there are more sequences, they will need to read literals from litExtraBuffer; copy over the remainder from dst and update litPtr and litEnd */
|
1469
1517
|
if (nbSeq > 0) {
|
1470
1518
|
const size_t leftoverLit = dctx->litBufferEnd - litPtr;
|
1471
|
-
|
1472
|
-
{
|
1519
|
+
DEBUGLOG(6, "There are %i sequences left, and %zu/%zu literals left in buffer", nbSeq, leftoverLit, sequence.litLength);
|
1520
|
+
if (leftoverLit) {
|
1473
1521
|
RETURN_ERROR_IF(leftoverLit > (size_t)(oend - op), dstSize_tooSmall, "remaining lit must fit within dstBuffer");
|
1474
1522
|
ZSTD_safecopyDstBeforeSrc(op, litPtr, leftoverLit);
|
1475
1523
|
sequence.litLength -= leftoverLit;
|
@@ -1478,24 +1526,22 @@ ZSTD_decompressSequences_bodySplitLitBuffer( ZSTD_DCtx* dctx,
|
|
1478
1526
|
litPtr = dctx->litExtraBuffer;
|
1479
1527
|
litBufferEnd = dctx->litExtraBuffer + ZSTD_LITBUFFEREXTRASIZE;
|
1480
1528
|
dctx->litBufferLocation = ZSTD_not_in_dst;
|
1481
|
-
{
|
1482
|
-
size_t const oneSeqSize = ZSTD_execSequence(op, oend, sequence, &litPtr, litBufferEnd, prefixStart, vBase, dictEnd);
|
1529
|
+
{ size_t const oneSeqSize = ZSTD_execSequence(op, oend, sequence, &litPtr, litBufferEnd, prefixStart, vBase, dictEnd);
|
1483
1530
|
#if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE)
|
1484
1531
|
assert(!ZSTD_isError(oneSeqSize));
|
1485
|
-
|
1532
|
+
ZSTD_assertValidSequence(dctx, op, oend, sequence, prefixStart, vBase);
|
1486
1533
|
#endif
|
1487
1534
|
if (UNLIKELY(ZSTD_isError(oneSeqSize)))
|
1488
1535
|
return oneSeqSize;
|
1489
1536
|
DEBUGLOG(6, "regenerated sequence size : %u", (U32)oneSeqSize);
|
1490
1537
|
op += oneSeqSize;
|
1491
|
-
if (--nbSeq)
|
1492
|
-
BIT_reloadDStream(&(seqState.DStream));
|
1493
1538
|
}
|
1539
|
+
nbSeq--;
|
1494
1540
|
}
|
1495
1541
|
}
|
1496
1542
|
|
1497
|
-
if (nbSeq > 0)
|
1498
|
-
|
1543
|
+
if (nbSeq > 0) {
|
1544
|
+
/* there is remaining lit from extra buffer */
|
1499
1545
|
|
1500
1546
|
#if defined(__GNUC__) && defined(__x86_64__)
|
1501
1547
|
__asm__(".p2align 6");
|
@@ -1514,35 +1560,34 @@ ZSTD_decompressSequences_bodySplitLitBuffer( ZSTD_DCtx* dctx,
|
|
1514
1560
|
# endif
|
1515
1561
|
#endif
|
1516
1562
|
|
1517
|
-
for (; ; ) {
|
1518
|
-
seq_t const sequence = ZSTD_decodeSequence(&seqState, isLongOffset);
|
1563
|
+
for ( ; nbSeq ; nbSeq--) {
|
1564
|
+
seq_t const sequence = ZSTD_decodeSequence(&seqState, isLongOffset, nbSeq==1);
|
1519
1565
|
size_t const oneSeqSize = ZSTD_execSequence(op, oend, sequence, &litPtr, litBufferEnd, prefixStart, vBase, dictEnd);
|
1520
1566
|
#if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE)
|
1521
1567
|
assert(!ZSTD_isError(oneSeqSize));
|
1522
|
-
|
1568
|
+
ZSTD_assertValidSequence(dctx, op, oend, sequence, prefixStart, vBase);
|
1523
1569
|
#endif
|
1524
1570
|
if (UNLIKELY(ZSTD_isError(oneSeqSize)))
|
1525
1571
|
return oneSeqSize;
|
1526
1572
|
DEBUGLOG(6, "regenerated sequence size : %u", (U32)oneSeqSize);
|
1527
1573
|
op += oneSeqSize;
|
1528
|
-
if (UNLIKELY(!--nbSeq))
|
1529
|
-
break;
|
1530
|
-
BIT_reloadDStream(&(seqState.DStream));
|
1531
1574
|
}
|
1532
1575
|
}
|
1533
1576
|
|
1534
1577
|
/* check if reached exact end */
|
1535
1578
|
DEBUGLOG(5, "ZSTD_decompressSequences_bodySplitLitBuffer: after decode loop, remaining nbSeq : %i", nbSeq);
|
1536
1579
|
RETURN_ERROR_IF(nbSeq, corruption_detected, "");
|
1537
|
-
|
1580
|
+
DEBUGLOG(5, "bitStream : start=%p, ptr=%p, bitsConsumed=%u", seqState.DStream.start, seqState.DStream.ptr, seqState.DStream.bitsConsumed);
|
1581
|
+
RETURN_ERROR_IF(!BIT_endOfDStream(&seqState.DStream), corruption_detected, "");
|
1538
1582
|
/* save reps for next block */
|
1539
1583
|
{ U32 i; for (i=0; i<ZSTD_REP_NUM; i++) dctx->entropy.rep[i] = (U32)(seqState.prevOffset[i]); }
|
1540
1584
|
}
|
1541
1585
|
|
1542
1586
|
/* last literal segment */
|
1543
|
-
if (dctx->litBufferLocation == ZSTD_split)
|
1544
|
-
|
1545
|
-
size_t const lastLLSize = litBufferEnd - litPtr;
|
1587
|
+
if (dctx->litBufferLocation == ZSTD_split) {
|
1588
|
+
/* split hasn't been reached yet, first get dst then copy litExtraBuffer */
|
1589
|
+
size_t const lastLLSize = (size_t)(litBufferEnd - litPtr);
|
1590
|
+
DEBUGLOG(6, "copy last literals from segment : %u", (U32)lastLLSize);
|
1546
1591
|
RETURN_ERROR_IF(lastLLSize > (size_t)(oend - op), dstSize_tooSmall, "");
|
1547
1592
|
if (op != NULL) {
|
1548
1593
|
ZSTD_memmove(op, litPtr, lastLLSize);
|
@@ -1552,15 +1597,17 @@ ZSTD_decompressSequences_bodySplitLitBuffer( ZSTD_DCtx* dctx,
|
|
1552
1597
|
litBufferEnd = dctx->litExtraBuffer + ZSTD_LITBUFFEREXTRASIZE;
|
1553
1598
|
dctx->litBufferLocation = ZSTD_not_in_dst;
|
1554
1599
|
}
|
1555
|
-
|
1600
|
+
/* copy last literals from internal buffer */
|
1601
|
+
{ size_t const lastLLSize = (size_t)(litBufferEnd - litPtr);
|
1602
|
+
DEBUGLOG(6, "copy last literals from internal buffer : %u", (U32)lastLLSize);
|
1556
1603
|
RETURN_ERROR_IF(lastLLSize > (size_t)(oend-op), dstSize_tooSmall, "");
|
1557
1604
|
if (op != NULL) {
|
1558
1605
|
ZSTD_memcpy(op, litPtr, lastLLSize);
|
1559
1606
|
op += lastLLSize;
|
1560
|
-
|
1561
|
-
}
|
1607
|
+
} }
|
1562
1608
|
|
1563
|
-
|
1609
|
+
DEBUGLOG(6, "decoded block of size %u bytes", (U32)(op - ostart));
|
1610
|
+
return (size_t)(op - ostart);
|
1564
1611
|
}
|
1565
1612
|
|
1566
1613
|
FORCE_INLINE_TEMPLATE size_t
|
@@ -1568,13 +1615,12 @@ DONT_VECTORIZE
|
|
1568
1615
|
ZSTD_decompressSequences_body(ZSTD_DCtx* dctx,
|
1569
1616
|
void* dst, size_t maxDstSize,
|
1570
1617
|
const void* seqStart, size_t seqSize, int nbSeq,
|
1571
|
-
const ZSTD_longOffset_e isLongOffset
|
1572
|
-
const int frame)
|
1618
|
+
const ZSTD_longOffset_e isLongOffset)
|
1573
1619
|
{
|
1574
1620
|
const BYTE* ip = (const BYTE*)seqStart;
|
1575
1621
|
const BYTE* const iend = ip + seqSize;
|
1576
1622
|
BYTE* const ostart = (BYTE*)dst;
|
1577
|
-
BYTE* const oend = dctx->litBufferLocation == ZSTD_not_in_dst ? ostart
|
1623
|
+
BYTE* const oend = dctx->litBufferLocation == ZSTD_not_in_dst ? ZSTD_maybeNullPtrAdd(ostart, maxDstSize) : dctx->litBuffer;
|
1578
1624
|
BYTE* op = ostart;
|
1579
1625
|
const BYTE* litPtr = dctx->litPtr;
|
1580
1626
|
const BYTE* const litEnd = litPtr + dctx->litSize;
|
@@ -1582,7 +1628,6 @@ ZSTD_decompressSequences_body(ZSTD_DCtx* dctx,
|
|
1582
1628
|
const BYTE* const vBase = (const BYTE*)(dctx->virtualStart);
|
1583
1629
|
const BYTE* const dictEnd = (const BYTE*)(dctx->dictEnd);
|
1584
1630
|
DEBUGLOG(5, "ZSTD_decompressSequences_body: nbSeq = %d", nbSeq);
|
1585
|
-
(void)frame;
|
1586
1631
|
|
1587
1632
|
/* Regen sequences */
|
1588
1633
|
if (nbSeq) {
|
@@ -1597,11 +1642,6 @@ ZSTD_decompressSequences_body(ZSTD_DCtx* dctx,
|
|
1597
1642
|
ZSTD_initFseState(&seqState.stateML, &seqState.DStream, dctx->MLTptr);
|
1598
1643
|
assert(dst != NULL);
|
1599
1644
|
|
1600
|
-
ZSTD_STATIC_ASSERT(
|
1601
|
-
BIT_DStream_unfinished < BIT_DStream_completed &&
|
1602
|
-
BIT_DStream_endOfBuffer < BIT_DStream_completed &&
|
1603
|
-
BIT_DStream_completed < BIT_DStream_overflow);
|
1604
|
-
|
1605
1645
|
#if defined(__GNUC__) && defined(__x86_64__)
|
1606
1646
|
__asm__(".p2align 6");
|
1607
1647
|
__asm__("nop");
|
@@ -1616,73 +1656,70 @@ ZSTD_decompressSequences_body(ZSTD_DCtx* dctx,
|
|
1616
1656
|
# endif
|
1617
1657
|
#endif
|
1618
1658
|
|
1619
|
-
for ( ; ; ) {
|
1620
|
-
seq_t const sequence = ZSTD_decodeSequence(&seqState, isLongOffset);
|
1659
|
+
for ( ; nbSeq ; nbSeq--) {
|
1660
|
+
seq_t const sequence = ZSTD_decodeSequence(&seqState, isLongOffset, nbSeq==1);
|
1621
1661
|
size_t const oneSeqSize = ZSTD_execSequence(op, oend, sequence, &litPtr, litEnd, prefixStart, vBase, dictEnd);
|
1622
1662
|
#if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE)
|
1623
1663
|
assert(!ZSTD_isError(oneSeqSize));
|
1624
|
-
|
1664
|
+
ZSTD_assertValidSequence(dctx, op, oend, sequence, prefixStart, vBase);
|
1625
1665
|
#endif
|
1626
1666
|
if (UNLIKELY(ZSTD_isError(oneSeqSize)))
|
1627
1667
|
return oneSeqSize;
|
1628
1668
|
DEBUGLOG(6, "regenerated sequence size : %u", (U32)oneSeqSize);
|
1629
1669
|
op += oneSeqSize;
|
1630
|
-
if (UNLIKELY(!--nbSeq))
|
1631
|
-
break;
|
1632
|
-
BIT_reloadDStream(&(seqState.DStream));
|
1633
1670
|
}
|
1634
1671
|
|
1635
1672
|
/* check if reached exact end */
|
1636
|
-
|
1637
|
-
RETURN_ERROR_IF(
|
1638
|
-
RETURN_ERROR_IF(BIT_reloadDStream(&seqState.DStream) < BIT_DStream_completed, corruption_detected, "");
|
1673
|
+
assert(nbSeq == 0);
|
1674
|
+
RETURN_ERROR_IF(!BIT_endOfDStream(&seqState.DStream), corruption_detected, "");
|
1639
1675
|
/* save reps for next block */
|
1640
1676
|
{ U32 i; for (i=0; i<ZSTD_REP_NUM; i++) dctx->entropy.rep[i] = (U32)(seqState.prevOffset[i]); }
|
1641
1677
|
}
|
1642
1678
|
|
1643
1679
|
/* last literal segment */
|
1644
|
-
{ size_t const lastLLSize = litEnd - litPtr;
|
1680
|
+
{ size_t const lastLLSize = (size_t)(litEnd - litPtr);
|
1681
|
+
DEBUGLOG(6, "copy last literals : %u", (U32)lastLLSize);
|
1645
1682
|
RETURN_ERROR_IF(lastLLSize > (size_t)(oend-op), dstSize_tooSmall, "");
|
1646
1683
|
if (op != NULL) {
|
1647
1684
|
ZSTD_memcpy(op, litPtr, lastLLSize);
|
1648
1685
|
op += lastLLSize;
|
1649
|
-
|
1650
|
-
}
|
1686
|
+
} }
|
1651
1687
|
|
1652
|
-
|
1688
|
+
DEBUGLOG(6, "decoded block of size %u bytes", (U32)(op - ostart));
|
1689
|
+
return (size_t)(op - ostart);
|
1653
1690
|
}
|
1654
1691
|
|
1655
1692
|
static size_t
|
1656
1693
|
ZSTD_decompressSequences_default(ZSTD_DCtx* dctx,
|
1657
1694
|
void* dst, size_t maxDstSize,
|
1658
1695
|
const void* seqStart, size_t seqSize, int nbSeq,
|
1659
|
-
const ZSTD_longOffset_e isLongOffset
|
1660
|
-
const int frame)
|
1696
|
+
const ZSTD_longOffset_e isLongOffset)
|
1661
1697
|
{
|
1662
|
-
return ZSTD_decompressSequences_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset
|
1698
|
+
return ZSTD_decompressSequences_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
|
1663
1699
|
}
|
1664
1700
|
|
1665
1701
|
static size_t
|
1666
1702
|
ZSTD_decompressSequencesSplitLitBuffer_default(ZSTD_DCtx* dctx,
|
1667
1703
|
void* dst, size_t maxDstSize,
|
1668
1704
|
const void* seqStart, size_t seqSize, int nbSeq,
|
1669
|
-
const ZSTD_longOffset_e isLongOffset
|
1670
|
-
const int frame)
|
1705
|
+
const ZSTD_longOffset_e isLongOffset)
|
1671
1706
|
{
|
1672
|
-
return ZSTD_decompressSequences_bodySplitLitBuffer(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset
|
1707
|
+
return ZSTD_decompressSequences_bodySplitLitBuffer(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
|
1673
1708
|
}
|
1674
1709
|
#endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG */
|
1675
1710
|
|
1676
1711
|
#ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT
|
1677
1712
|
|
1678
|
-
FORCE_INLINE_TEMPLATE
|
1679
|
-
|
1713
|
+
FORCE_INLINE_TEMPLATE
|
1714
|
+
|
1715
|
+
size_t ZSTD_prefetchMatch(size_t prefetchPos, seq_t const sequence,
|
1680
1716
|
const BYTE* const prefixStart, const BYTE* const dictEnd)
|
1681
1717
|
{
|
1682
1718
|
prefetchPos += sequence.litLength;
|
1683
1719
|
{ const BYTE* const matchBase = (sequence.offset > prefetchPos) ? dictEnd : prefixStart;
|
1684
|
-
|
1685
|
-
|
1720
|
+
/* note : this operation can overflow when seq.offset is really too large, which can only happen when input is corrupted.
|
1721
|
+
* No consequence though : memory address is only used for prefetching, not for dereferencing */
|
1722
|
+
const BYTE* const match = ZSTD_wrappedPtrSub(ZSTD_wrappedPtrAdd(matchBase, prefetchPos), sequence.offset);
|
1686
1723
|
PREFETCH_L1(match); PREFETCH_L1(match+CACHELINE_SIZE); /* note : it's safe to invoke PREFETCH() on any memory address, including invalid ones */
|
1687
1724
|
}
|
1688
1725
|
return prefetchPos + sequence.matchLength;
|
@@ -1697,20 +1734,18 @@ ZSTD_decompressSequencesLong_body(
|
|
1697
1734
|
ZSTD_DCtx* dctx,
|
1698
1735
|
void* dst, size_t maxDstSize,
|
1699
1736
|
const void* seqStart, size_t seqSize, int nbSeq,
|
1700
|
-
const ZSTD_longOffset_e isLongOffset
|
1701
|
-
const int frame)
|
1737
|
+
const ZSTD_longOffset_e isLongOffset)
|
1702
1738
|
{
|
1703
1739
|
const BYTE* ip = (const BYTE*)seqStart;
|
1704
1740
|
const BYTE* const iend = ip + seqSize;
|
1705
1741
|
BYTE* const ostart = (BYTE*)dst;
|
1706
|
-
BYTE* const oend = dctx->litBufferLocation == ZSTD_in_dst ? dctx->litBuffer : ostart
|
1742
|
+
BYTE* const oend = dctx->litBufferLocation == ZSTD_in_dst ? dctx->litBuffer : ZSTD_maybeNullPtrAdd(ostart, maxDstSize);
|
1707
1743
|
BYTE* op = ostart;
|
1708
1744
|
const BYTE* litPtr = dctx->litPtr;
|
1709
1745
|
const BYTE* litBufferEnd = dctx->litBufferEnd;
|
1710
1746
|
const BYTE* const prefixStart = (const BYTE*) (dctx->prefixStart);
|
1711
1747
|
const BYTE* const dictStart = (const BYTE*) (dctx->virtualStart);
|
1712
1748
|
const BYTE* const dictEnd = (const BYTE*) (dctx->dictEnd);
|
1713
|
-
(void)frame;
|
1714
1749
|
|
1715
1750
|
/* Regen sequences */
|
1716
1751
|
if (nbSeq) {
|
@@ -1735,20 +1770,17 @@ ZSTD_decompressSequencesLong_body(
|
|
1735
1770
|
ZSTD_initFseState(&seqState.stateML, &seqState.DStream, dctx->MLTptr);
|
1736
1771
|
|
1737
1772
|
/* prepare in advance */
|
1738
|
-
for (seqNb=0;
|
1739
|
-
seq_t const sequence = ZSTD_decodeSequence(&seqState, isLongOffset);
|
1773
|
+
for (seqNb=0; seqNb<seqAdvance; seqNb++) {
|
1774
|
+
seq_t const sequence = ZSTD_decodeSequence(&seqState, isLongOffset, seqNb == nbSeq-1);
|
1740
1775
|
prefetchPos = ZSTD_prefetchMatch(prefetchPos, sequence, prefixStart, dictEnd);
|
1741
1776
|
sequences[seqNb] = sequence;
|
1742
1777
|
}
|
1743
|
-
RETURN_ERROR_IF(seqNb<seqAdvance, corruption_detected, "");
|
1744
1778
|
|
1745
1779
|
/* decompress without stomping litBuffer */
|
1746
|
-
for (;
|
1747
|
-
seq_t sequence = ZSTD_decodeSequence(&seqState, isLongOffset);
|
1748
|
-
size_t oneSeqSize;
|
1780
|
+
for (; seqNb < nbSeq; seqNb++) {
|
1781
|
+
seq_t sequence = ZSTD_decodeSequence(&seqState, isLongOffset, seqNb == nbSeq-1);
|
1749
1782
|
|
1750
|
-
if (dctx->litBufferLocation == ZSTD_split && litPtr + sequences[(seqNb - ADVANCED_SEQS) & STORED_SEQS_MASK].litLength > dctx->litBufferEnd)
|
1751
|
-
{
|
1783
|
+
if (dctx->litBufferLocation == ZSTD_split && litPtr + sequences[(seqNb - ADVANCED_SEQS) & STORED_SEQS_MASK].litLength > dctx->litBufferEnd) {
|
1752
1784
|
/* lit buffer is reaching split point, empty out the first buffer and transition to litExtraBuffer */
|
1753
1785
|
const size_t leftoverLit = dctx->litBufferEnd - litPtr;
|
1754
1786
|
if (leftoverLit)
|
@@ -1761,26 +1793,26 @@ ZSTD_decompressSequencesLong_body(
|
|
1761
1793
|
litPtr = dctx->litExtraBuffer;
|
1762
1794
|
litBufferEnd = dctx->litExtraBuffer + ZSTD_LITBUFFEREXTRASIZE;
|
1763
1795
|
dctx->litBufferLocation = ZSTD_not_in_dst;
|
1764
|
-
oneSeqSize = ZSTD_execSequence(op, oend, sequences[(seqNb - ADVANCED_SEQS) & STORED_SEQS_MASK], &litPtr, litBufferEnd, prefixStart, dictStart, dictEnd);
|
1796
|
+
{ size_t const oneSeqSize = ZSTD_execSequence(op, oend, sequences[(seqNb - ADVANCED_SEQS) & STORED_SEQS_MASK], &litPtr, litBufferEnd, prefixStart, dictStart, dictEnd);
|
1765
1797
|
#if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE)
|
1766
|
-
|
1767
|
-
|
1798
|
+
assert(!ZSTD_isError(oneSeqSize));
|
1799
|
+
ZSTD_assertValidSequence(dctx, op, oend, sequences[(seqNb - ADVANCED_SEQS) & STORED_SEQS_MASK], prefixStart, dictStart);
|
1768
1800
|
#endif
|
1769
|
-
|
1801
|
+
if (ZSTD_isError(oneSeqSize)) return oneSeqSize;
|
1770
1802
|
|
1771
|
-
|
1772
|
-
|
1773
|
-
|
1774
|
-
}
|
1803
|
+
prefetchPos = ZSTD_prefetchMatch(prefetchPos, sequence, prefixStart, dictEnd);
|
1804
|
+
sequences[seqNb & STORED_SEQS_MASK] = sequence;
|
1805
|
+
op += oneSeqSize;
|
1806
|
+
} }
|
1775
1807
|
else
|
1776
1808
|
{
|
1777
1809
|
/* lit buffer is either wholly contained in first or second split, or not split at all*/
|
1778
|
-
oneSeqSize = dctx->litBufferLocation == ZSTD_split ?
|
1810
|
+
size_t const oneSeqSize = dctx->litBufferLocation == ZSTD_split ?
|
1779
1811
|
ZSTD_execSequenceSplitLitBuffer(op, oend, litPtr + sequences[(seqNb - ADVANCED_SEQS) & STORED_SEQS_MASK].litLength - WILDCOPY_OVERLENGTH, sequences[(seqNb - ADVANCED_SEQS) & STORED_SEQS_MASK], &litPtr, litBufferEnd, prefixStart, dictStart, dictEnd) :
|
1780
1812
|
ZSTD_execSequence(op, oend, sequences[(seqNb - ADVANCED_SEQS) & STORED_SEQS_MASK], &litPtr, litBufferEnd, prefixStart, dictStart, dictEnd);
|
1781
1813
|
#if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE)
|
1782
1814
|
assert(!ZSTD_isError(oneSeqSize));
|
1783
|
-
|
1815
|
+
ZSTD_assertValidSequence(dctx, op, oend, sequences[(seqNb - ADVANCED_SEQS) & STORED_SEQS_MASK], prefixStart, dictStart);
|
1784
1816
|
#endif
|
1785
1817
|
if (ZSTD_isError(oneSeqSize)) return oneSeqSize;
|
1786
1818
|
|
@@ -1789,17 +1821,15 @@ ZSTD_decompressSequencesLong_body(
|
|
1789
1821
|
op += oneSeqSize;
|
1790
1822
|
}
|
1791
1823
|
}
|
1792
|
-
RETURN_ERROR_IF(
|
1824
|
+
RETURN_ERROR_IF(!BIT_endOfDStream(&seqState.DStream), corruption_detected, "");
|
1793
1825
|
|
1794
1826
|
/* finish queue */
|
1795
1827
|
seqNb -= seqAdvance;
|
1796
1828
|
for ( ; seqNb<nbSeq ; seqNb++) {
|
1797
1829
|
seq_t *sequence = &(sequences[seqNb&STORED_SEQS_MASK]);
|
1798
|
-
if (dctx->litBufferLocation == ZSTD_split && litPtr + sequence->litLength > dctx->litBufferEnd)
|
1799
|
-
{
|
1830
|
+
if (dctx->litBufferLocation == ZSTD_split && litPtr + sequence->litLength > dctx->litBufferEnd) {
|
1800
1831
|
const size_t leftoverLit = dctx->litBufferEnd - litPtr;
|
1801
|
-
if (leftoverLit)
|
1802
|
-
{
|
1832
|
+
if (leftoverLit) {
|
1803
1833
|
RETURN_ERROR_IF(leftoverLit > (size_t)(oend - op), dstSize_tooSmall, "remaining lit must fit within dstBuffer");
|
1804
1834
|
ZSTD_safecopyDstBeforeSrc(op, litPtr, leftoverLit);
|
1805
1835
|
sequence->litLength -= leftoverLit;
|
@@ -1808,11 +1838,10 @@ ZSTD_decompressSequencesLong_body(
|
|
1808
1838
|
litPtr = dctx->litExtraBuffer;
|
1809
1839
|
litBufferEnd = dctx->litExtraBuffer + ZSTD_LITBUFFEREXTRASIZE;
|
1810
1840
|
dctx->litBufferLocation = ZSTD_not_in_dst;
|
1811
|
-
{
|
1812
|
-
size_t const oneSeqSize = ZSTD_execSequence(op, oend, *sequence, &litPtr, litBufferEnd, prefixStart, dictStart, dictEnd);
|
1841
|
+
{ size_t const oneSeqSize = ZSTD_execSequence(op, oend, *sequence, &litPtr, litBufferEnd, prefixStart, dictStart, dictEnd);
|
1813
1842
|
#if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE)
|
1814
1843
|
assert(!ZSTD_isError(oneSeqSize));
|
1815
|
-
|
1844
|
+
ZSTD_assertValidSequence(dctx, op, oend, sequences[seqNb&STORED_SEQS_MASK], prefixStart, dictStart);
|
1816
1845
|
#endif
|
1817
1846
|
if (ZSTD_isError(oneSeqSize)) return oneSeqSize;
|
1818
1847
|
op += oneSeqSize;
|
@@ -1825,7 +1854,7 @@ ZSTD_decompressSequencesLong_body(
|
|
1825
1854
|
ZSTD_execSequence(op, oend, *sequence, &litPtr, litBufferEnd, prefixStart, dictStart, dictEnd);
|
1826
1855
|
#if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE)
|
1827
1856
|
assert(!ZSTD_isError(oneSeqSize));
|
1828
|
-
|
1857
|
+
ZSTD_assertValidSequence(dctx, op, oend, sequences[seqNb&STORED_SEQS_MASK], prefixStart, dictStart);
|
1829
1858
|
#endif
|
1830
1859
|
if (ZSTD_isError(oneSeqSize)) return oneSeqSize;
|
1831
1860
|
op += oneSeqSize;
|
@@ -1837,8 +1866,7 @@ ZSTD_decompressSequencesLong_body(
|
|
1837
1866
|
}
|
1838
1867
|
|
1839
1868
|
/* last literal segment */
|
1840
|
-
if (dctx->litBufferLocation == ZSTD_split)
|
1841
|
-
{
|
1869
|
+
if (dctx->litBufferLocation == ZSTD_split) { /* first deplete literal buffer in dst, then copy litExtraBuffer */
|
1842
1870
|
size_t const lastLLSize = litBufferEnd - litPtr;
|
1843
1871
|
RETURN_ERROR_IF(lastLLSize > (size_t)(oend - op), dstSize_tooSmall, "");
|
1844
1872
|
if (op != NULL) {
|
@@ -1856,17 +1884,16 @@ ZSTD_decompressSequencesLong_body(
|
|
1856
1884
|
}
|
1857
1885
|
}
|
1858
1886
|
|
1859
|
-
return op-ostart;
|
1887
|
+
return (size_t)(op - ostart);
|
1860
1888
|
}
|
1861
1889
|
|
1862
1890
|
static size_t
|
1863
1891
|
ZSTD_decompressSequencesLong_default(ZSTD_DCtx* dctx,
|
1864
1892
|
void* dst, size_t maxDstSize,
|
1865
1893
|
const void* seqStart, size_t seqSize, int nbSeq,
|
1866
|
-
const ZSTD_longOffset_e isLongOffset
|
1867
|
-
const int frame)
|
1894
|
+
const ZSTD_longOffset_e isLongOffset)
|
1868
1895
|
{
|
1869
|
-
return ZSTD_decompressSequencesLong_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset
|
1896
|
+
return ZSTD_decompressSequencesLong_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
|
1870
1897
|
}
|
1871
1898
|
#endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT */
|
1872
1899
|
|
@@ -1880,20 +1907,18 @@ DONT_VECTORIZE
|
|
1880
1907
|
ZSTD_decompressSequences_bmi2(ZSTD_DCtx* dctx,
|
1881
1908
|
void* dst, size_t maxDstSize,
|
1882
1909
|
const void* seqStart, size_t seqSize, int nbSeq,
|
1883
|
-
const ZSTD_longOffset_e isLongOffset
|
1884
|
-
const int frame)
|
1910
|
+
const ZSTD_longOffset_e isLongOffset)
|
1885
1911
|
{
|
1886
|
-
return ZSTD_decompressSequences_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset
|
1912
|
+
return ZSTD_decompressSequences_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
|
1887
1913
|
}
|
1888
1914
|
static BMI2_TARGET_ATTRIBUTE size_t
|
1889
1915
|
DONT_VECTORIZE
|
1890
1916
|
ZSTD_decompressSequencesSplitLitBuffer_bmi2(ZSTD_DCtx* dctx,
|
1891
1917
|
void* dst, size_t maxDstSize,
|
1892
1918
|
const void* seqStart, size_t seqSize, int nbSeq,
|
1893
|
-
const ZSTD_longOffset_e isLongOffset
|
1894
|
-
const int frame)
|
1919
|
+
const ZSTD_longOffset_e isLongOffset)
|
1895
1920
|
{
|
1896
|
-
return ZSTD_decompressSequences_bodySplitLitBuffer(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset
|
1921
|
+
return ZSTD_decompressSequences_bodySplitLitBuffer(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
|
1897
1922
|
}
|
1898
1923
|
#endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG */
|
1899
1924
|
|
@@ -1902,10 +1927,9 @@ static BMI2_TARGET_ATTRIBUTE size_t
|
|
1902
1927
|
ZSTD_decompressSequencesLong_bmi2(ZSTD_DCtx* dctx,
|
1903
1928
|
void* dst, size_t maxDstSize,
|
1904
1929
|
const void* seqStart, size_t seqSize, int nbSeq,
|
1905
|
-
const ZSTD_longOffset_e isLongOffset
|
1906
|
-
const int frame)
|
1930
|
+
const ZSTD_longOffset_e isLongOffset)
|
1907
1931
|
{
|
1908
|
-
return ZSTD_decompressSequencesLong_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset
|
1932
|
+
return ZSTD_decompressSequencesLong_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
|
1909
1933
|
}
|
1910
1934
|
#endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT */
|
1911
1935
|
|
@@ -1915,37 +1939,34 @@ typedef size_t (*ZSTD_decompressSequences_t)(
|
|
1915
1939
|
ZSTD_DCtx* dctx,
|
1916
1940
|
void* dst, size_t maxDstSize,
|
1917
1941
|
const void* seqStart, size_t seqSize, int nbSeq,
|
1918
|
-
const ZSTD_longOffset_e isLongOffset
|
1919
|
-
const int frame);
|
1942
|
+
const ZSTD_longOffset_e isLongOffset);
|
1920
1943
|
|
1921
1944
|
#ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG
|
1922
1945
|
static size_t
|
1923
1946
|
ZSTD_decompressSequences(ZSTD_DCtx* dctx, void* dst, size_t maxDstSize,
|
1924
1947
|
const void* seqStart, size_t seqSize, int nbSeq,
|
1925
|
-
const ZSTD_longOffset_e isLongOffset
|
1926
|
-
const int frame)
|
1948
|
+
const ZSTD_longOffset_e isLongOffset)
|
1927
1949
|
{
|
1928
1950
|
DEBUGLOG(5, "ZSTD_decompressSequences");
|
1929
1951
|
#if DYNAMIC_BMI2
|
1930
1952
|
if (ZSTD_DCtx_get_bmi2(dctx)) {
|
1931
|
-
return ZSTD_decompressSequences_bmi2(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset
|
1953
|
+
return ZSTD_decompressSequences_bmi2(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
|
1932
1954
|
}
|
1933
1955
|
#endif
|
1934
|
-
return ZSTD_decompressSequences_default(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset
|
1956
|
+
return ZSTD_decompressSequences_default(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
|
1935
1957
|
}
|
1936
1958
|
static size_t
|
1937
1959
|
ZSTD_decompressSequencesSplitLitBuffer(ZSTD_DCtx* dctx, void* dst, size_t maxDstSize,
|
1938
1960
|
const void* seqStart, size_t seqSize, int nbSeq,
|
1939
|
-
const ZSTD_longOffset_e isLongOffset
|
1940
|
-
const int frame)
|
1961
|
+
const ZSTD_longOffset_e isLongOffset)
|
1941
1962
|
{
|
1942
1963
|
DEBUGLOG(5, "ZSTD_decompressSequencesSplitLitBuffer");
|
1943
1964
|
#if DYNAMIC_BMI2
|
1944
1965
|
if (ZSTD_DCtx_get_bmi2(dctx)) {
|
1945
|
-
return ZSTD_decompressSequencesSplitLitBuffer_bmi2(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset
|
1966
|
+
return ZSTD_decompressSequencesSplitLitBuffer_bmi2(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
|
1946
1967
|
}
|
1947
1968
|
#endif
|
1948
|
-
return ZSTD_decompressSequencesSplitLitBuffer_default(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset
|
1969
|
+
return ZSTD_decompressSequencesSplitLitBuffer_default(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
|
1949
1970
|
}
|
1950
1971
|
#endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG */
|
1951
1972
|
|
@@ -1960,16 +1981,15 @@ static size_t
|
|
1960
1981
|
ZSTD_decompressSequencesLong(ZSTD_DCtx* dctx,
|
1961
1982
|
void* dst, size_t maxDstSize,
|
1962
1983
|
const void* seqStart, size_t seqSize, int nbSeq,
|
1963
|
-
const ZSTD_longOffset_e isLongOffset
|
1964
|
-
const int frame)
|
1984
|
+
const ZSTD_longOffset_e isLongOffset)
|
1965
1985
|
{
|
1966
1986
|
DEBUGLOG(5, "ZSTD_decompressSequencesLong");
|
1967
1987
|
#if DYNAMIC_BMI2
|
1968
1988
|
if (ZSTD_DCtx_get_bmi2(dctx)) {
|
1969
|
-
return ZSTD_decompressSequencesLong_bmi2(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset
|
1989
|
+
return ZSTD_decompressSequencesLong_bmi2(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
|
1970
1990
|
}
|
1971
1991
|
#endif
|
1972
|
-
return ZSTD_decompressSequencesLong_default(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset
|
1992
|
+
return ZSTD_decompressSequencesLong_default(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
|
1973
1993
|
}
|
1974
1994
|
#endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT */
|
1975
1995
|
|
@@ -2051,20 +2071,20 @@ static size_t ZSTD_maxShortOffset(void)
|
|
2051
2071
|
size_t
|
2052
2072
|
ZSTD_decompressBlock_internal(ZSTD_DCtx* dctx,
|
2053
2073
|
void* dst, size_t dstCapacity,
|
2054
|
-
const void* src, size_t srcSize, const
|
2074
|
+
const void* src, size_t srcSize, const streaming_operation streaming)
|
2055
2075
|
{ /* blockType == blockCompressed */
|
2056
2076
|
const BYTE* ip = (const BYTE*)src;
|
2057
|
-
DEBUGLOG(5, "ZSTD_decompressBlock_internal (
|
2077
|
+
DEBUGLOG(5, "ZSTD_decompressBlock_internal (cSize : %u)", (unsigned)srcSize);
|
2058
2078
|
|
2059
2079
|
/* Note : the wording of the specification
|
2060
|
-
* allows compressed block to be sized exactly
|
2080
|
+
* allows compressed block to be sized exactly ZSTD_blockSizeMax(dctx).
|
2061
2081
|
* This generally does not happen, as it makes little sense,
|
2062
2082
|
* since an uncompressed block would feature same size and have no decompression cost.
|
2063
2083
|
* Also, note that decoder from reference libzstd before < v1.5.4
|
2064
2084
|
* would consider this edge case as an error.
|
2065
|
-
* As a consequence, avoid generating compressed blocks of size
|
2085
|
+
* As a consequence, avoid generating compressed blocks of size ZSTD_blockSizeMax(dctx)
|
2066
2086
|
* for broader compatibility with the deployed ecosystem of zstd decoders */
|
2067
|
-
RETURN_ERROR_IF(srcSize >
|
2087
|
+
RETURN_ERROR_IF(srcSize > ZSTD_blockSizeMax(dctx), srcSize_wrong, "");
|
2068
2088
|
|
2069
2089
|
/* Decode literals section */
|
2070
2090
|
{ size_t const litCSize = ZSTD_decodeLiteralsBlock(dctx, src, srcSize, dst, dstCapacity, streaming);
|
@@ -2079,8 +2099,8 @@ ZSTD_decompressBlock_internal(ZSTD_DCtx* dctx,
|
|
2079
2099
|
/* Compute the maximum block size, which must also work when !frame and fParams are unset.
|
2080
2100
|
* Additionally, take the min with dstCapacity to ensure that the totalHistorySize fits in a size_t.
|
2081
2101
|
*/
|
2082
|
-
size_t const blockSizeMax = MIN(dstCapacity, (
|
2083
|
-
size_t const totalHistorySize = ZSTD_totalHistorySize((BYTE*)dst
|
2102
|
+
size_t const blockSizeMax = MIN(dstCapacity, ZSTD_blockSizeMax(dctx));
|
2103
|
+
size_t const totalHistorySize = ZSTD_totalHistorySize(ZSTD_maybeNullPtrAdd((BYTE*)dst, blockSizeMax), (BYTE const*)dctx->virtualStart);
|
2084
2104
|
/* isLongOffset must be true if there are long offsets.
|
2085
2105
|
* Offsets are long if they are larger than ZSTD_maxShortOffset().
|
2086
2106
|
* We don't expect that to be the case in 64-bit mode.
|
@@ -2145,21 +2165,22 @@ ZSTD_decompressBlock_internal(ZSTD_DCtx* dctx,
|
|
2145
2165
|
{
|
2146
2166
|
#endif
|
2147
2167
|
#ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT
|
2148
|
-
return ZSTD_decompressSequencesLong(dctx, dst, dstCapacity, ip, srcSize, nbSeq, isLongOffset
|
2168
|
+
return ZSTD_decompressSequencesLong(dctx, dst, dstCapacity, ip, srcSize, nbSeq, isLongOffset);
|
2149
2169
|
#endif
|
2150
2170
|
}
|
2151
2171
|
|
2152
2172
|
#ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG
|
2153
2173
|
/* else */
|
2154
2174
|
if (dctx->litBufferLocation == ZSTD_split)
|
2155
|
-
return ZSTD_decompressSequencesSplitLitBuffer(dctx, dst, dstCapacity, ip, srcSize, nbSeq, isLongOffset
|
2175
|
+
return ZSTD_decompressSequencesSplitLitBuffer(dctx, dst, dstCapacity, ip, srcSize, nbSeq, isLongOffset);
|
2156
2176
|
else
|
2157
|
-
return ZSTD_decompressSequences(dctx, dst, dstCapacity, ip, srcSize, nbSeq, isLongOffset
|
2177
|
+
return ZSTD_decompressSequences(dctx, dst, dstCapacity, ip, srcSize, nbSeq, isLongOffset);
|
2158
2178
|
#endif
|
2159
2179
|
}
|
2160
2180
|
}
|
2161
2181
|
|
2162
2182
|
|
2183
|
+
ZSTD_ALLOW_POINTER_OVERFLOW_ATTR
|
2163
2184
|
void ZSTD_checkContinuity(ZSTD_DCtx* dctx, const void* dst, size_t dstSize)
|
2164
2185
|
{
|
2165
2186
|
if (dst != dctx->previousDstEnd && dstSize > 0) { /* not contiguous */
|
@@ -2176,8 +2197,10 @@ size_t ZSTD_decompressBlock_deprecated(ZSTD_DCtx* dctx,
|
|
2176
2197
|
const void* src, size_t srcSize)
|
2177
2198
|
{
|
2178
2199
|
size_t dSize;
|
2200
|
+
dctx->isFrameDecompression = 0;
|
2179
2201
|
ZSTD_checkContinuity(dctx, dst, dstCapacity);
|
2180
|
-
dSize = ZSTD_decompressBlock_internal(dctx, dst, dstCapacity, src, srcSize,
|
2202
|
+
dSize = ZSTD_decompressBlock_internal(dctx, dst, dstCapacity, src, srcSize, not_streaming);
|
2203
|
+
FORWARD_IF_ERROR(dSize, "");
|
2181
2204
|
dctx->previousDstEnd = (char*)dst + dSize;
|
2182
2205
|
return dSize;
|
2183
2206
|
}
|