summaryrefslogtreecommitdiffstats
path: root/third_party/zstd/lib/decompress
diff options
context:
space:
mode:
Diffstat (limited to '')
-rw-r--r--third_party/zstd/lib/decompress/huf_decompress.c1944
-rw-r--r--third_party/zstd/lib/decompress/huf_decompress_amd64.S595
-rw-r--r--third_party/zstd/lib/decompress/zstd_ddict.c244
-rw-r--r--third_party/zstd/lib/decompress/zstd_ddict.h44
-rw-r--r--third_party/zstd/lib/decompress/zstd_decompress.c2407
-rw-r--r--third_party/zstd/lib/decompress/zstd_decompress_block.c2215
-rw-r--r--third_party/zstd/lib/decompress/zstd_decompress_block.h73
-rw-r--r--third_party/zstd/lib/decompress/zstd_decompress_internal.h240
8 files changed, 7762 insertions, 0 deletions
diff --git a/third_party/zstd/lib/decompress/huf_decompress.c b/third_party/zstd/lib/decompress/huf_decompress.c
new file mode 100644
index 0000000000..f85dd0beea
--- /dev/null
+++ b/third_party/zstd/lib/decompress/huf_decompress.c
@@ -0,0 +1,1944 @@
+/* ******************************************************************
+ * huff0 huffman decoder,
+ * part of Finite State Entropy library
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
+ *
+ * You can contact the author at :
+ * - FSE+HUF source repository : https://github.com/Cyan4973/FiniteStateEntropy
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+****************************************************************** */
+
+/* **************************************************************
+* Dependencies
+****************************************************************/
+#include "../common/zstd_deps.h" /* ZSTD_memcpy, ZSTD_memset */
+#include "../common/compiler.h"
+#include "../common/bitstream.h" /* BIT_* */
+#include "../common/fse.h" /* to compress headers */
+#include "../common/huf.h"
+#include "../common/error_private.h"
+#include "../common/zstd_internal.h"
+#include "../common/bits.h" /* ZSTD_highbit32, ZSTD_countTrailingZeros64 */
+
+/* **************************************************************
+* Constants
+****************************************************************/
+
+#define HUF_DECODER_FAST_TABLELOG 11
+
+/* **************************************************************
+* Macros
+****************************************************************/
+
+#ifdef HUF_DISABLE_FAST_DECODE
+# define HUF_ENABLE_FAST_DECODE 0
+#else
+# define HUF_ENABLE_FAST_DECODE 1
+#endif
+
+/* These two optional macros force the use one way or another of the two
+ * Huffman decompression implementations. You can't force in both directions
+ * at the same time.
+ */
+#if defined(HUF_FORCE_DECOMPRESS_X1) && \
+ defined(HUF_FORCE_DECOMPRESS_X2)
+#error "Cannot force the use of the X1 and X2 decoders at the same time!"
+#endif
+
+/* When DYNAMIC_BMI2 is enabled, fast decoders are only called when bmi2 is
+ * supported at runtime, so we can add the BMI2 target attribute.
+ * When it is disabled, we will still get BMI2 if it is enabled statically.
+ */
+#if DYNAMIC_BMI2
+# define HUF_FAST_BMI2_ATTRS BMI2_TARGET_ATTRIBUTE
+#else
+# define HUF_FAST_BMI2_ATTRS
+#endif
+
+#ifdef __cplusplus
+# define HUF_EXTERN_C extern "C"
+#else
+# define HUF_EXTERN_C
+#endif
+#define HUF_ASM_DECL HUF_EXTERN_C
+
+#if DYNAMIC_BMI2
+# define HUF_NEED_BMI2_FUNCTION 1
+#else
+# define HUF_NEED_BMI2_FUNCTION 0
+#endif
+
+/* **************************************************************
+* Error Management
+****************************************************************/
+#define HUF_isError ERR_isError
+
+
+/* **************************************************************
+* Byte alignment for workSpace management
+****************************************************************/
+#define HUF_ALIGN(x, a) HUF_ALIGN_MASK((x), (a) - 1)
+#define HUF_ALIGN_MASK(x, mask) (((x) + (mask)) & ~(mask))
+
+
+/* **************************************************************
+* BMI2 Variant Wrappers
+****************************************************************/
+typedef size_t (*HUF_DecompressUsingDTableFn)(void *dst, size_t dstSize,
+ const void *cSrc,
+ size_t cSrcSize,
+ const HUF_DTable *DTable);
+
+#if DYNAMIC_BMI2
+
+#define HUF_DGEN(fn) \
+ \
+ static size_t fn##_default( \
+ void* dst, size_t dstSize, \
+ const void* cSrc, size_t cSrcSize, \
+ const HUF_DTable* DTable) \
+ { \
+ return fn##_body(dst, dstSize, cSrc, cSrcSize, DTable); \
+ } \
+ \
+ static BMI2_TARGET_ATTRIBUTE size_t fn##_bmi2( \
+ void* dst, size_t dstSize, \
+ const void* cSrc, size_t cSrcSize, \
+ const HUF_DTable* DTable) \
+ { \
+ return fn##_body(dst, dstSize, cSrc, cSrcSize, DTable); \
+ } \
+ \
+ static size_t fn(void* dst, size_t dstSize, void const* cSrc, \
+ size_t cSrcSize, HUF_DTable const* DTable, int flags) \
+ { \
+ if (flags & HUF_flags_bmi2) { \
+ return fn##_bmi2(dst, dstSize, cSrc, cSrcSize, DTable); \
+ } \
+ return fn##_default(dst, dstSize, cSrc, cSrcSize, DTable); \
+ }
+
+#else
+
+#define HUF_DGEN(fn) \
+ static size_t fn(void* dst, size_t dstSize, void const* cSrc, \
+ size_t cSrcSize, HUF_DTable const* DTable, int flags) \
+ { \
+ (void)flags; \
+ return fn##_body(dst, dstSize, cSrc, cSrcSize, DTable); \
+ }
+
+#endif
+
+
+/*-***************************/
+/* generic DTableDesc */
+/*-***************************/
+typedef struct { BYTE maxTableLog; BYTE tableType; BYTE tableLog; BYTE reserved; } DTableDesc;
+
+static DTableDesc HUF_getDTableDesc(const HUF_DTable* table)
+{
+ DTableDesc dtd;
+ ZSTD_memcpy(&dtd, table, sizeof(dtd));
+ return dtd;
+}
+
+static size_t HUF_initFastDStream(BYTE const* ip) {
+ BYTE const lastByte = ip[7];
+ size_t const bitsConsumed = lastByte ? 8 - ZSTD_highbit32(lastByte) : 0;
+ size_t const value = MEM_readLEST(ip) | 1;
+ assert(bitsConsumed <= 8);
+ assert(sizeof(size_t) == 8);
+ return value << bitsConsumed;
+}
+
+
+/**
+ * The input/output arguments to the Huffman fast decoding loop:
+ *
+ * ip [in/out] - The input pointers, must be updated to reflect what is consumed.
+ * op [in/out] - The output pointers, must be updated to reflect what is written.
+ * bits [in/out] - The bitstream containers, must be updated to reflect the current state.
+ * dt [in] - The decoding table.
+ * ilowest [in] - The beginning of the valid range of the input. Decoders may read
+ * down to this pointer. It may be below iend[0].
+ * oend [in] - The end of the output stream. op[3] must not cross oend.
+ * iend [in] - The end of each input stream. ip[i] may cross iend[i],
+ * as long as it is above ilowest, but that indicates corruption.
+ */
+typedef struct {
+ BYTE const* ip[4];
+ BYTE* op[4];
+ U64 bits[4];
+ void const* dt;
+ BYTE const* ilowest;
+ BYTE* oend;
+ BYTE const* iend[4];
+} HUF_DecompressFastArgs;
+
+typedef void (*HUF_DecompressFastLoopFn)(HUF_DecompressFastArgs*);
+
+/**
+ * Initializes args for the fast decoding loop.
+ * @returns 1 on success
+ * 0 if the fallback implementation should be used.
+ * Or an error code on failure.
+ */
+static size_t HUF_DecompressFastArgs_init(HUF_DecompressFastArgs* args, void* dst, size_t dstSize, void const* src, size_t srcSize, const HUF_DTable* DTable)
+{
+ void const* dt = DTable + 1;
+ U32 const dtLog = HUF_getDTableDesc(DTable).tableLog;
+
+ const BYTE* const istart = (const BYTE*)src;
+
+ BYTE* const oend = ZSTD_maybeNullPtrAdd((BYTE*)dst, dstSize);
+
+ /* The fast decoding loop assumes 64-bit little-endian.
+ * This condition is false on x32.
+ */
+ if (!MEM_isLittleEndian() || MEM_32bits())
+ return 0;
+
+ /* Avoid nullptr addition */
+ if (dstSize == 0)
+ return 0;
+ assert(dst != NULL);
+
+ /* strict minimum : jump table + 1 byte per stream */
+ if (srcSize < 10)
+ return ERROR(corruption_detected);
+
+ /* Must have at least 8 bytes per stream because we don't handle initializing smaller bit containers.
+ * If table log is not correct at this point, fallback to the old decoder.
+ * On small inputs we don't have enough data to trigger the fast loop, so use the old decoder.
+ */
+ if (dtLog != HUF_DECODER_FAST_TABLELOG)
+ return 0;
+
+ /* Read the jump table. */
+ {
+ size_t const length1 = MEM_readLE16(istart);
+ size_t const length2 = MEM_readLE16(istart+2);
+ size_t const length3 = MEM_readLE16(istart+4);
+ size_t const length4 = srcSize - (length1 + length2 + length3 + 6);
+ args->iend[0] = istart + 6; /* jumpTable */
+ args->iend[1] = args->iend[0] + length1;
+ args->iend[2] = args->iend[1] + length2;
+ args->iend[3] = args->iend[2] + length3;
+
+ /* HUF_initFastDStream() requires this, and this small of an input
+ * won't benefit from the ASM loop anyways.
+ */
+ if (length1 < 8 || length2 < 8 || length3 < 8 || length4 < 8)
+ return 0;
+ if (length4 > srcSize) return ERROR(corruption_detected); /* overflow */
+ }
+ /* ip[] contains the position that is currently loaded into bits[]. */
+ args->ip[0] = args->iend[1] - sizeof(U64);
+ args->ip[1] = args->iend[2] - sizeof(U64);
+ args->ip[2] = args->iend[3] - sizeof(U64);
+ args->ip[3] = (BYTE const*)src + srcSize - sizeof(U64);
+
+ /* op[] contains the output pointers. */
+ args->op[0] = (BYTE*)dst;
+ args->op[1] = args->op[0] + (dstSize+3)/4;
+ args->op[2] = args->op[1] + (dstSize+3)/4;
+ args->op[3] = args->op[2] + (dstSize+3)/4;
+
+ /* No point to call the ASM loop for tiny outputs. */
+ if (args->op[3] >= oend)
+ return 0;
+
+ /* bits[] is the bit container.
+ * It is read from the MSB down to the LSB.
+ * It is shifted left as it is read, and zeros are
+ * shifted in. After the lowest valid bit a 1 is
+ * set, so that CountTrailingZeros(bits[]) can be used
+ * to count how many bits we've consumed.
+ */
+ args->bits[0] = HUF_initFastDStream(args->ip[0]);
+ args->bits[1] = HUF_initFastDStream(args->ip[1]);
+ args->bits[2] = HUF_initFastDStream(args->ip[2]);
+ args->bits[3] = HUF_initFastDStream(args->ip[3]);
+
+ /* The decoders must be sure to never read beyond ilowest.
+ * This is lower than iend[0], but allowing decoders to read
+ * down to ilowest can allow an extra iteration or two in the
+ * fast loop.
+ */
+ args->ilowest = istart;
+
+ args->oend = oend;
+ args->dt = dt;
+
+ return 1;
+}
+
+static size_t HUF_initRemainingDStream(BIT_DStream_t* bit, HUF_DecompressFastArgs const* args, int stream, BYTE* segmentEnd)
+{
+ /* Validate that we haven't overwritten. */
+ if (args->op[stream] > segmentEnd)
+ return ERROR(corruption_detected);
+ /* Validate that we haven't read beyond iend[].
+ * Note that ip[] may be < iend[] because the MSB is
+ * the next bit to read, and we may have consumed 100%
+ * of the stream, so down to iend[i] - 8 is valid.
+ */
+ if (args->ip[stream] < args->iend[stream] - 8)
+ return ERROR(corruption_detected);
+
+ /* Construct the BIT_DStream_t. */
+ assert(sizeof(size_t) == 8);
+ bit->bitContainer = MEM_readLEST(args->ip[stream]);
+ bit->bitsConsumed = ZSTD_countTrailingZeros64(args->bits[stream]);
+ bit->start = (const char*)args->ilowest;
+ bit->limitPtr = bit->start + sizeof(size_t);
+ bit->ptr = (const char*)args->ip[stream];
+
+ return 0;
+}
+
+/* Calls X(N) for each stream 0, 1, 2, 3. */
+#define HUF_4X_FOR_EACH_STREAM(X) \
+ do { \
+ X(0); \
+ X(1); \
+ X(2); \
+ X(3); \
+ } while (0)
+
+/* Calls X(N, var) for each stream 0, 1, 2, 3. */
+#define HUF_4X_FOR_EACH_STREAM_WITH_VAR(X, var) \
+ do { \
+ X(0, (var)); \
+ X(1, (var)); \
+ X(2, (var)); \
+ X(3, (var)); \
+ } while (0)
+
+
+#ifndef HUF_FORCE_DECOMPRESS_X2
+
+/*-***************************/
+/* single-symbol decoding */
+/*-***************************/
+typedef struct { BYTE nbBits; BYTE byte; } HUF_DEltX1; /* single-symbol decoding */
+
+/**
+ * Packs 4 HUF_DEltX1 structs into a U64. This is used to lay down 4 entries at
+ * a time.
+ */
+static U64 HUF_DEltX1_set4(BYTE symbol, BYTE nbBits) {
+ U64 D4;
+ if (MEM_isLittleEndian()) {
+ D4 = (U64)((symbol << 8) + nbBits);
+ } else {
+ D4 = (U64)(symbol + (nbBits << 8));
+ }
+ assert(D4 < (1U << 16));
+ D4 *= 0x0001000100010001ULL;
+ return D4;
+}
+
+/**
+ * Increase the tableLog to targetTableLog and rescales the stats.
+ * If tableLog > targetTableLog this is a no-op.
+ * @returns New tableLog
+ */
+static U32 HUF_rescaleStats(BYTE* huffWeight, U32* rankVal, U32 nbSymbols, U32 tableLog, U32 targetTableLog)
+{
+ if (tableLog > targetTableLog)
+ return tableLog;
+ if (tableLog < targetTableLog) {
+ U32 const scale = targetTableLog - tableLog;
+ U32 s;
+ /* Increase the weight for all non-zero probability symbols by scale. */
+ for (s = 0; s < nbSymbols; ++s) {
+ huffWeight[s] += (BYTE)((huffWeight[s] == 0) ? 0 : scale);
+ }
+ /* Update rankVal to reflect the new weights.
+ * All weights except 0 get moved to weight + scale.
+ * Weights [1, scale] are empty.
+ */
+ for (s = targetTableLog; s > scale; --s) {
+ rankVal[s] = rankVal[s - scale];
+ }
+ for (s = scale; s > 0; --s) {
+ rankVal[s] = 0;
+ }
+ }
+ return targetTableLog;
+}
+
+typedef struct {
+ U32 rankVal[HUF_TABLELOG_ABSOLUTEMAX + 1];
+ U32 rankStart[HUF_TABLELOG_ABSOLUTEMAX + 1];
+ U32 statsWksp[HUF_READ_STATS_WORKSPACE_SIZE_U32];
+ BYTE symbols[HUF_SYMBOLVALUE_MAX + 1];
+ BYTE huffWeight[HUF_SYMBOLVALUE_MAX + 1];
+} HUF_ReadDTableX1_Workspace;
+
+size_t HUF_readDTableX1_wksp(HUF_DTable* DTable, const void* src, size_t srcSize, void* workSpace, size_t wkspSize, int flags)
+{
+ U32 tableLog = 0;
+ U32 nbSymbols = 0;
+ size_t iSize;
+ void* const dtPtr = DTable + 1;
+ HUF_DEltX1* const dt = (HUF_DEltX1*)dtPtr;
+ HUF_ReadDTableX1_Workspace* wksp = (HUF_ReadDTableX1_Workspace*)workSpace;
+
+ DEBUG_STATIC_ASSERT(HUF_DECOMPRESS_WORKSPACE_SIZE >= sizeof(*wksp));
+ if (sizeof(*wksp) > wkspSize) return ERROR(tableLog_tooLarge);
+
+ DEBUG_STATIC_ASSERT(sizeof(DTableDesc) == sizeof(HUF_DTable));
+ /* ZSTD_memset(huffWeight, 0, sizeof(huffWeight)); */ /* is not necessary, even though some analyzer complain ... */
+
+ iSize = HUF_readStats_wksp(wksp->huffWeight, HUF_SYMBOLVALUE_MAX + 1, wksp->rankVal, &nbSymbols, &tableLog, src, srcSize, wksp->statsWksp, sizeof(wksp->statsWksp), flags);
+ if (HUF_isError(iSize)) return iSize;
+
+
+ /* Table header */
+ { DTableDesc dtd = HUF_getDTableDesc(DTable);
+ U32 const maxTableLog = dtd.maxTableLog + 1;
+ U32 const targetTableLog = MIN(maxTableLog, HUF_DECODER_FAST_TABLELOG);
+ tableLog = HUF_rescaleStats(wksp->huffWeight, wksp->rankVal, nbSymbols, tableLog, targetTableLog);
+ if (tableLog > (U32)(dtd.maxTableLog+1)) return ERROR(tableLog_tooLarge); /* DTable too small, Huffman tree cannot fit in */
+ dtd.tableType = 0;
+ dtd.tableLog = (BYTE)tableLog;
+ ZSTD_memcpy(DTable, &dtd, sizeof(dtd));
+ }
+
+ /* Compute symbols and rankStart given rankVal:
+ *
+ * rankVal already contains the number of values of each weight.
+ *
+ * symbols contains the symbols ordered by weight. First are the rankVal[0]
+ * weight 0 symbols, followed by the rankVal[1] weight 1 symbols, and so on.
+ * symbols[0] is filled (but unused) to avoid a branch.
+ *
+ * rankStart contains the offset where each rank belongs in the DTable.
+ * rankStart[0] is not filled because there are no entries in the table for
+ * weight 0.
+ */
+ { int n;
+ U32 nextRankStart = 0;
+ int const unroll = 4;
+ int const nLimit = (int)nbSymbols - unroll + 1;
+ for (n=0; n<(int)tableLog+1; n++) {
+ U32 const curr = nextRankStart;
+ nextRankStart += wksp->rankVal[n];
+ wksp->rankStart[n] = curr;
+ }
+ for (n=0; n < nLimit; n += unroll) {
+ int u;
+ for (u=0; u < unroll; ++u) {
+ size_t const w = wksp->huffWeight[n+u];
+ wksp->symbols[wksp->rankStart[w]++] = (BYTE)(n+u);
+ }
+ }
+ for (; n < (int)nbSymbols; ++n) {
+ size_t const w = wksp->huffWeight[n];
+ wksp->symbols[wksp->rankStart[w]++] = (BYTE)n;
+ }
+ }
+
+ /* fill DTable
+ * We fill all entries of each weight in order.
+ * That way length is a constant for each iteration of the outer loop.
+ * We can switch based on the length to a different inner loop which is
+ * optimized for that particular case.
+ */
+ { U32 w;
+ int symbol = wksp->rankVal[0];
+ int rankStart = 0;
+ for (w=1; w<tableLog+1; ++w) {
+ int const symbolCount = wksp->rankVal[w];
+ int const length = (1 << w) >> 1;
+ int uStart = rankStart;
+ BYTE const nbBits = (BYTE)(tableLog + 1 - w);
+ int s;
+ int u;
+ switch (length) {
+ case 1:
+ for (s=0; s<symbolCount; ++s) {
+ HUF_DEltX1 D;
+ D.byte = wksp->symbols[symbol + s];
+ D.nbBits = nbBits;
+ dt[uStart] = D;
+ uStart += 1;
+ }
+ break;
+ case 2:
+ for (s=0; s<symbolCount; ++s) {
+ HUF_DEltX1 D;
+ D.byte = wksp->symbols[symbol + s];
+ D.nbBits = nbBits;
+ dt[uStart+0] = D;
+ dt[uStart+1] = D;
+ uStart += 2;
+ }
+ break;
+ case 4:
+ for (s=0; s<symbolCount; ++s) {
+ U64 const D4 = HUF_DEltX1_set4(wksp->symbols[symbol + s], nbBits);
+ MEM_write64(dt + uStart, D4);
+ uStart += 4;
+ }
+ break;
+ case 8:
+ for (s=0; s<symbolCount; ++s) {
+ U64 const D4 = HUF_DEltX1_set4(wksp->symbols[symbol + s], nbBits);
+ MEM_write64(dt + uStart, D4);
+ MEM_write64(dt + uStart + 4, D4);
+ uStart += 8;
+ }
+ break;
+ default:
+ for (s=0; s<symbolCount; ++s) {
+ U64 const D4 = HUF_DEltX1_set4(wksp->symbols[symbol + s], nbBits);
+ for (u=0; u < length; u += 16) {
+ MEM_write64(dt + uStart + u + 0, D4);
+ MEM_write64(dt + uStart + u + 4, D4);
+ MEM_write64(dt + uStart + u + 8, D4);
+ MEM_write64(dt + uStart + u + 12, D4);
+ }
+ assert(u == length);
+ uStart += length;
+ }
+ break;
+ }
+ symbol += symbolCount;
+ rankStart += symbolCount * length;
+ }
+ }
+ return iSize;
+}
+
+FORCE_INLINE_TEMPLATE BYTE
+HUF_decodeSymbolX1(BIT_DStream_t* Dstream, const HUF_DEltX1* dt, const U32 dtLog)
+{
+ size_t const val = BIT_lookBitsFast(Dstream, dtLog); /* note : dtLog >= 1 */
+ BYTE const c = dt[val].byte;
+ BIT_skipBits(Dstream, dt[val].nbBits);
+ return c;
+}
+
+#define HUF_DECODE_SYMBOLX1_0(ptr, DStreamPtr) \
+ do { *ptr++ = HUF_decodeSymbolX1(DStreamPtr, dt, dtLog); } while (0)
+
+#define HUF_DECODE_SYMBOLX1_1(ptr, DStreamPtr) \
+ do { \
+ if (MEM_64bits() || (HUF_TABLELOG_MAX<=12)) \
+ HUF_DECODE_SYMBOLX1_0(ptr, DStreamPtr); \
+ } while (0)
+
+#define HUF_DECODE_SYMBOLX1_2(ptr, DStreamPtr) \
+ do { \
+ if (MEM_64bits()) \
+ HUF_DECODE_SYMBOLX1_0(ptr, DStreamPtr); \
+ } while (0)
+
+HINT_INLINE size_t
+HUF_decodeStreamX1(BYTE* p, BIT_DStream_t* const bitDPtr, BYTE* const pEnd, const HUF_DEltX1* const dt, const U32 dtLog)
+{
+ BYTE* const pStart = p;
+
+ /* up to 4 symbols at a time */
+ if ((pEnd - p) > 3) {
+ while ((BIT_reloadDStream(bitDPtr) == BIT_DStream_unfinished) & (p < pEnd-3)) {
+ HUF_DECODE_SYMBOLX1_2(p, bitDPtr);
+ HUF_DECODE_SYMBOLX1_1(p, bitDPtr);
+ HUF_DECODE_SYMBOLX1_2(p, bitDPtr);
+ HUF_DECODE_SYMBOLX1_0(p, bitDPtr);
+ }
+ } else {
+ BIT_reloadDStream(bitDPtr);
+ }
+
+ /* [0-3] symbols remaining */
+ if (MEM_32bits())
+ while ((BIT_reloadDStream(bitDPtr) == BIT_DStream_unfinished) & (p < pEnd))
+ HUF_DECODE_SYMBOLX1_0(p, bitDPtr);
+
+ /* no more data to retrieve from bitstream, no need to reload */
+ while (p < pEnd)
+ HUF_DECODE_SYMBOLX1_0(p, bitDPtr);
+
+ return (size_t)(pEnd-pStart);
+}
+
+FORCE_INLINE_TEMPLATE size_t
+HUF_decompress1X1_usingDTable_internal_body(
+ void* dst, size_t dstSize,
+ const void* cSrc, size_t cSrcSize,
+ const HUF_DTable* DTable)
+{
+ BYTE* op = (BYTE*)dst;
+ BYTE* const oend = ZSTD_maybeNullPtrAdd(op, dstSize);
+ const void* dtPtr = DTable + 1;
+ const HUF_DEltX1* const dt = (const HUF_DEltX1*)dtPtr;
+ BIT_DStream_t bitD;
+ DTableDesc const dtd = HUF_getDTableDesc(DTable);
+ U32 const dtLog = dtd.tableLog;
+
+ CHECK_F( BIT_initDStream(&bitD, cSrc, cSrcSize) );
+
+ HUF_decodeStreamX1(op, &bitD, oend, dt, dtLog);
+
+ if (!BIT_endOfDStream(&bitD)) return ERROR(corruption_detected);
+
+ return dstSize;
+}
+
+/* HUF_decompress4X1_usingDTable_internal_body():
+ * Conditions :
+ * @dstSize >= 6
+ */
+FORCE_INLINE_TEMPLATE size_t
+HUF_decompress4X1_usingDTable_internal_body(
+ void* dst, size_t dstSize,
+ const void* cSrc, size_t cSrcSize,
+ const HUF_DTable* DTable)
+{
+ /* Check */
+ if (cSrcSize < 10) return ERROR(corruption_detected); /* strict minimum : jump table + 1 byte per stream */
+ if (dstSize < 6) return ERROR(corruption_detected); /* stream 4-split doesn't work */
+
+ { const BYTE* const istart = (const BYTE*) cSrc;
+ BYTE* const ostart = (BYTE*) dst;
+ BYTE* const oend = ostart + dstSize;
+ BYTE* const olimit = oend - 3;
+ const void* const dtPtr = DTable + 1;
+ const HUF_DEltX1* const dt = (const HUF_DEltX1*)dtPtr;
+
+ /* Init */
+ BIT_DStream_t bitD1;
+ BIT_DStream_t bitD2;
+ BIT_DStream_t bitD3;
+ BIT_DStream_t bitD4;
+ size_t const length1 = MEM_readLE16(istart);
+ size_t const length2 = MEM_readLE16(istart+2);
+ size_t const length3 = MEM_readLE16(istart+4);
+ size_t const length4 = cSrcSize - (length1 + length2 + length3 + 6);
+ const BYTE* const istart1 = istart + 6; /* jumpTable */
+ const BYTE* const istart2 = istart1 + length1;
+ const BYTE* const istart3 = istart2 + length2;
+ const BYTE* const istart4 = istart3 + length3;
+ const size_t segmentSize = (dstSize+3) / 4;
+ BYTE* const opStart2 = ostart + segmentSize;
+ BYTE* const opStart3 = opStart2 + segmentSize;
+ BYTE* const opStart4 = opStart3 + segmentSize;
+ BYTE* op1 = ostart;
+ BYTE* op2 = opStart2;
+ BYTE* op3 = opStart3;
+ BYTE* op4 = opStart4;
+ DTableDesc const dtd = HUF_getDTableDesc(DTable);
+ U32 const dtLog = dtd.tableLog;
+ U32 endSignal = 1;
+
+ if (length4 > cSrcSize) return ERROR(corruption_detected); /* overflow */
+ if (opStart4 > oend) return ERROR(corruption_detected); /* overflow */
+ assert(dstSize >= 6); /* validated above */
+ CHECK_F( BIT_initDStream(&bitD1, istart1, length1) );
+ CHECK_F( BIT_initDStream(&bitD2, istart2, length2) );
+ CHECK_F( BIT_initDStream(&bitD3, istart3, length3) );
+ CHECK_F( BIT_initDStream(&bitD4, istart4, length4) );
+
+ /* up to 16 symbols per loop (4 symbols per stream) in 64-bit mode */
+ if ((size_t)(oend - op4) >= sizeof(size_t)) {
+ for ( ; (endSignal) & (op4 < olimit) ; ) {
+ HUF_DECODE_SYMBOLX1_2(op1, &bitD1);
+ HUF_DECODE_SYMBOLX1_2(op2, &bitD2);
+ HUF_DECODE_SYMBOLX1_2(op3, &bitD3);
+ HUF_DECODE_SYMBOLX1_2(op4, &bitD4);
+ HUF_DECODE_SYMBOLX1_1(op1, &bitD1);
+ HUF_DECODE_SYMBOLX1_1(op2, &bitD2);
+ HUF_DECODE_SYMBOLX1_1(op3, &bitD3);
+ HUF_DECODE_SYMBOLX1_1(op4, &bitD4);
+ HUF_DECODE_SYMBOLX1_2(op1, &bitD1);
+ HUF_DECODE_SYMBOLX1_2(op2, &bitD2);
+ HUF_DECODE_SYMBOLX1_2(op3, &bitD3);
+ HUF_DECODE_SYMBOLX1_2(op4, &bitD4);
+ HUF_DECODE_SYMBOLX1_0(op1, &bitD1);
+ HUF_DECODE_SYMBOLX1_0(op2, &bitD2);
+ HUF_DECODE_SYMBOLX1_0(op3, &bitD3);
+ HUF_DECODE_SYMBOLX1_0(op4, &bitD4);
+ endSignal &= BIT_reloadDStreamFast(&bitD1) == BIT_DStream_unfinished;
+ endSignal &= BIT_reloadDStreamFast(&bitD2) == BIT_DStream_unfinished;
+ endSignal &= BIT_reloadDStreamFast(&bitD3) == BIT_DStream_unfinished;
+ endSignal &= BIT_reloadDStreamFast(&bitD4) == BIT_DStream_unfinished;
+ }
+ }
+
+ /* check corruption */
+ /* note : should not be necessary : op# advance in lock step, and we control op4.
+ * but curiously, binary generated by gcc 7.2 & 7.3 with -mbmi2 runs faster when >=1 test is present */
+ if (op1 > opStart2) return ERROR(corruption_detected);
+ if (op2 > opStart3) return ERROR(corruption_detected);
+ if (op3 > opStart4) return ERROR(corruption_detected);
+ /* note : op4 supposed already verified within main loop */
+
+ /* finish bitStreams one by one */
+ HUF_decodeStreamX1(op1, &bitD1, opStart2, dt, dtLog);
+ HUF_decodeStreamX1(op2, &bitD2, opStart3, dt, dtLog);
+ HUF_decodeStreamX1(op3, &bitD3, opStart4, dt, dtLog);
+ HUF_decodeStreamX1(op4, &bitD4, oend, dt, dtLog);
+
+ /* check */
+ { U32 const endCheck = BIT_endOfDStream(&bitD1) & BIT_endOfDStream(&bitD2) & BIT_endOfDStream(&bitD3) & BIT_endOfDStream(&bitD4);
+ if (!endCheck) return ERROR(corruption_detected); }
+
+ /* decoded size */
+ return dstSize;
+ }
+}
+
+#if HUF_NEED_BMI2_FUNCTION
+static BMI2_TARGET_ATTRIBUTE
+size_t HUF_decompress4X1_usingDTable_internal_bmi2(void* dst, size_t dstSize, void const* cSrc,
+ size_t cSrcSize, HUF_DTable const* DTable) {
+ return HUF_decompress4X1_usingDTable_internal_body(dst, dstSize, cSrc, cSrcSize, DTable);
+}
+#endif
+
+static
+size_t HUF_decompress4X1_usingDTable_internal_default(void* dst, size_t dstSize, void const* cSrc,
+ size_t cSrcSize, HUF_DTable const* DTable) {
+ return HUF_decompress4X1_usingDTable_internal_body(dst, dstSize, cSrc, cSrcSize, DTable);
+}
+
+#if ZSTD_ENABLE_ASM_X86_64_BMI2
+
+HUF_ASM_DECL void HUF_decompress4X1_usingDTable_internal_fast_asm_loop(HUF_DecompressFastArgs* args) ZSTDLIB_HIDDEN;
+
+#endif
+
+static HUF_FAST_BMI2_ATTRS
+void HUF_decompress4X1_usingDTable_internal_fast_c_loop(HUF_DecompressFastArgs* args)
+{
+ U64 bits[4];
+ BYTE const* ip[4];
+ BYTE* op[4];
+ U16 const* const dtable = (U16 const*)args->dt;
+ BYTE* const oend = args->oend;
+ BYTE const* const ilowest = args->ilowest;
+
+ /* Copy the arguments to local variables */
+ ZSTD_memcpy(&bits, &args->bits, sizeof(bits));
+ ZSTD_memcpy((void*)(&ip), &args->ip, sizeof(ip));
+ ZSTD_memcpy(&op, &args->op, sizeof(op));
+
+ assert(MEM_isLittleEndian());
+ assert(!MEM_32bits());
+
+ for (;;) {
+ BYTE* olimit;
+ int stream;
+
+ /* Assert loop preconditions */
+#ifndef NDEBUG
+ for (stream = 0; stream < 4; ++stream) {
+ assert(op[stream] <= (stream == 3 ? oend : op[stream + 1]));
+ assert(ip[stream] >= ilowest);
+ }
+#endif
+ /* Compute olimit */
+ {
+ /* Each iteration produces 5 output symbols per stream */
+ size_t const oiters = (size_t)(oend - op[3]) / 5;
+ /* Each iteration consumes up to 11 bits * 5 = 55 bits < 7 bytes
+ * per stream.
+ */
+ size_t const iiters = (size_t)(ip[0] - ilowest) / 7;
+ /* We can safely run iters iterations before running bounds checks */
+ size_t const iters = MIN(oiters, iiters);
+ size_t const symbols = iters * 5;
+
+ /* We can simply check that op[3] < olimit, instead of checking all
+ * of our bounds, since we can't hit the other bounds until we've run
+ * iters iterations, which only happens when op[3] == olimit.
+ */
+ olimit = op[3] + symbols;
+
+ /* Exit fast decoding loop once we reach the end. */
+ if (op[3] == olimit)
+ break;
+
+ /* Exit the decoding loop if any input pointer has crossed the
+ * previous one. This indicates corruption, and a precondition
+ * to our loop is that ip[i] >= ip[0].
+ */
+ for (stream = 1; stream < 4; ++stream) {
+ if (ip[stream] < ip[stream - 1])
+ goto _out;
+ }
+ }
+
+#ifndef NDEBUG
+ for (stream = 1; stream < 4; ++stream) {
+ assert(ip[stream] >= ip[stream - 1]);
+ }
+#endif
+
+#define HUF_4X1_DECODE_SYMBOL(_stream, _symbol) \
+ do { \
+ int const index = (int)(bits[(_stream)] >> 53); \
+ int const entry = (int)dtable[index]; \
+ bits[(_stream)] <<= (entry & 0x3F); \
+ op[(_stream)][(_symbol)] = (BYTE)((entry >> 8) & 0xFF); \
+ } while (0)
+
+#define HUF_4X1_RELOAD_STREAM(_stream) \
+ do { \
+ int const ctz = ZSTD_countTrailingZeros64(bits[(_stream)]); \
+ int const nbBits = ctz & 7; \
+ int const nbBytes = ctz >> 3; \
+ op[(_stream)] += 5; \
+ ip[(_stream)] -= nbBytes; \
+ bits[(_stream)] = MEM_read64(ip[(_stream)]) | 1; \
+ bits[(_stream)] <<= nbBits; \
+ } while (0)
+
+ /* Manually unroll the loop because compilers don't consistently
+ * unroll the inner loops, which destroys performance.
+ */
+ do {
+ /* Decode 5 symbols in each of the 4 streams */
+ HUF_4X_FOR_EACH_STREAM_WITH_VAR(HUF_4X1_DECODE_SYMBOL, 0);
+ HUF_4X_FOR_EACH_STREAM_WITH_VAR(HUF_4X1_DECODE_SYMBOL, 1);
+ HUF_4X_FOR_EACH_STREAM_WITH_VAR(HUF_4X1_DECODE_SYMBOL, 2);
+ HUF_4X_FOR_EACH_STREAM_WITH_VAR(HUF_4X1_DECODE_SYMBOL, 3);
+ HUF_4X_FOR_EACH_STREAM_WITH_VAR(HUF_4X1_DECODE_SYMBOL, 4);
+
+ /* Reload each of the 4 the bitstreams */
+ HUF_4X_FOR_EACH_STREAM(HUF_4X1_RELOAD_STREAM);
+ } while (op[3] < olimit);
+
+#undef HUF_4X1_DECODE_SYMBOL
+#undef HUF_4X1_RELOAD_STREAM
+ }
+
+_out:
+
+ /* Save the final values of each of the state variables back to args. */
+ ZSTD_memcpy(&args->bits, &bits, sizeof(bits));
+ ZSTD_memcpy((void*)(&args->ip), &ip, sizeof(ip));
+ ZSTD_memcpy(&args->op, &op, sizeof(op));
+}
+
+/**
+ * @returns @p dstSize on success (>= 6)
+ * 0 if the fallback implementation should be used
+ * An error if an error occurred
+ */
+static HUF_FAST_BMI2_ATTRS
+size_t
+HUF_decompress4X1_usingDTable_internal_fast(
+ void* dst, size_t dstSize,
+ const void* cSrc, size_t cSrcSize,
+ const HUF_DTable* DTable,
+ HUF_DecompressFastLoopFn loopFn)
+{
+ void const* dt = DTable + 1;
+ BYTE const* const ilowest = (BYTE const*)cSrc;
+ BYTE* const oend = ZSTD_maybeNullPtrAdd((BYTE*)dst, dstSize);
+ HUF_DecompressFastArgs args;
+ { size_t const ret = HUF_DecompressFastArgs_init(&args, dst, dstSize, cSrc, cSrcSize, DTable);
+ FORWARD_IF_ERROR(ret, "Failed to init fast loop args");
+ if (ret == 0)
+ return 0;
+ }
+
+ assert(args.ip[0] >= args.ilowest);
+ loopFn(&args);
+
+ /* Our loop guarantees that ip[] >= ilowest and that we haven't
+ * overwritten any op[].
+ */
+ assert(args.ip[0] >= ilowest);
+ assert(args.ip[0] >= ilowest);
+ assert(args.ip[1] >= ilowest);
+ assert(args.ip[2] >= ilowest);
+ assert(args.ip[3] >= ilowest);
+ assert(args.op[3] <= oend);
+
+ assert(ilowest == args.ilowest);
+ assert(ilowest + 6 == args.iend[0]);
+ (void)ilowest;
+
+ /* finish bit streams one by one. */
+ { size_t const segmentSize = (dstSize+3) / 4;
+ BYTE* segmentEnd = (BYTE*)dst;
+ int i;
+ for (i = 0; i < 4; ++i) {
+ BIT_DStream_t bit;
+ if (segmentSize <= (size_t)(oend - segmentEnd))
+ segmentEnd += segmentSize;
+ else
+ segmentEnd = oend;
+ FORWARD_IF_ERROR(HUF_initRemainingDStream(&bit, &args, i, segmentEnd), "corruption");
+ /* Decompress and validate that we've produced exactly the expected length. */
+ args.op[i] += HUF_decodeStreamX1(args.op[i], &bit, segmentEnd, (HUF_DEltX1 const*)dt, HUF_DECODER_FAST_TABLELOG);
+ if (args.op[i] != segmentEnd) return ERROR(corruption_detected);
+ }
+ }
+
+ /* decoded size */
+ assert(dstSize != 0);
+ return dstSize;
+}
+
+HUF_DGEN(HUF_decompress1X1_usingDTable_internal)
+
+static size_t HUF_decompress4X1_usingDTable_internal(void* dst, size_t dstSize, void const* cSrc,
+ size_t cSrcSize, HUF_DTable const* DTable, int flags)
+{
+ HUF_DecompressUsingDTableFn fallbackFn = HUF_decompress4X1_usingDTable_internal_default;
+ HUF_DecompressFastLoopFn loopFn = HUF_decompress4X1_usingDTable_internal_fast_c_loop;
+
+#if DYNAMIC_BMI2
+ if (flags & HUF_flags_bmi2) {
+ fallbackFn = HUF_decompress4X1_usingDTable_internal_bmi2;
+# if ZSTD_ENABLE_ASM_X86_64_BMI2
+ if (!(flags & HUF_flags_disableAsm)) {
+ loopFn = HUF_decompress4X1_usingDTable_internal_fast_asm_loop;
+ }
+# endif
+ } else {
+ return fallbackFn(dst, dstSize, cSrc, cSrcSize, DTable);
+ }
+#endif
+
+#if ZSTD_ENABLE_ASM_X86_64_BMI2 && defined(__BMI2__)
+ if (!(flags & HUF_flags_disableAsm)) {
+ loopFn = HUF_decompress4X1_usingDTable_internal_fast_asm_loop;
+ }
+#endif
+
+ if (HUF_ENABLE_FAST_DECODE && !(flags & HUF_flags_disableFast)) {
+ size_t const ret = HUF_decompress4X1_usingDTable_internal_fast(dst, dstSize, cSrc, cSrcSize, DTable, loopFn);
+ if (ret != 0)
+ return ret;
+ }
+ return fallbackFn(dst, dstSize, cSrc, cSrcSize, DTable);
+}
+
+static size_t HUF_decompress4X1_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize,
+ const void* cSrc, size_t cSrcSize,
+ void* workSpace, size_t wkspSize, int flags)
+{
+ const BYTE* ip = (const BYTE*) cSrc;
+
+ size_t const hSize = HUF_readDTableX1_wksp(dctx, cSrc, cSrcSize, workSpace, wkspSize, flags);
+ if (HUF_isError(hSize)) return hSize;
+ if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
+ ip += hSize; cSrcSize -= hSize;
+
+ return HUF_decompress4X1_usingDTable_internal(dst, dstSize, ip, cSrcSize, dctx, flags);
+}
+
+#endif /* HUF_FORCE_DECOMPRESS_X2 */
+
+
+#ifndef HUF_FORCE_DECOMPRESS_X1
+
+/* *************************/
+/* double-symbols decoding */
+/* *************************/
+
+typedef struct { U16 sequence; BYTE nbBits; BYTE length; } HUF_DEltX2; /* double-symbols decoding */
+typedef struct { BYTE symbol; } sortedSymbol_t;
+typedef U32 rankValCol_t[HUF_TABLELOG_MAX + 1];
+typedef rankValCol_t rankVal_t[HUF_TABLELOG_MAX];
+
+/**
+ * Constructs a HUF_DEltX2 in a U32.
+ */
+static U32 HUF_buildDEltX2U32(U32 symbol, U32 nbBits, U32 baseSeq, int level)
+{
+ U32 seq;
+ DEBUG_STATIC_ASSERT(offsetof(HUF_DEltX2, sequence) == 0);
+ DEBUG_STATIC_ASSERT(offsetof(HUF_DEltX2, nbBits) == 2);
+ DEBUG_STATIC_ASSERT(offsetof(HUF_DEltX2, length) == 3);
+ DEBUG_STATIC_ASSERT(sizeof(HUF_DEltX2) == sizeof(U32));
+ if (MEM_isLittleEndian()) {
+ seq = level == 1 ? symbol : (baseSeq + (symbol << 8));
+ return seq + (nbBits << 16) + ((U32)level << 24);
+ } else {
+ seq = level == 1 ? (symbol << 8) : ((baseSeq << 8) + symbol);
+ return (seq << 16) + (nbBits << 8) + (U32)level;
+ }
+}
+
+/**
+ * Constructs a HUF_DEltX2.
+ */
+static HUF_DEltX2 HUF_buildDEltX2(U32 symbol, U32 nbBits, U32 baseSeq, int level)
+{
+ HUF_DEltX2 DElt;
+ U32 const val = HUF_buildDEltX2U32(symbol, nbBits, baseSeq, level);
+ DEBUG_STATIC_ASSERT(sizeof(DElt) == sizeof(val));
+ ZSTD_memcpy(&DElt, &val, sizeof(val));
+ return DElt;
+}
+
+/**
+ * Constructs 2 HUF_DEltX2s and packs them into a U64.
+ */
+static U64 HUF_buildDEltX2U64(U32 symbol, U32 nbBits, U16 baseSeq, int level)
+{
+ U32 DElt = HUF_buildDEltX2U32(symbol, nbBits, baseSeq, level);
+ return (U64)DElt + ((U64)DElt << 32);
+}
+
+/**
+ * Fills the DTable rank with all the symbols from [begin, end) that are each
+ * nbBits long.
+ *
+ * @param DTableRank The start of the rank in the DTable.
+ * @param begin The first symbol to fill (inclusive).
+ * @param end The last symbol to fill (exclusive).
+ * @param nbBits Each symbol is nbBits long.
+ * @param tableLog The table log.
+ * @param baseSeq If level == 1 { 0 } else { the first level symbol }
+ * @param level The level in the table. Must be 1 or 2.
+ */
+static void HUF_fillDTableX2ForWeight(
+ HUF_DEltX2* DTableRank,
+ sortedSymbol_t const* begin, sortedSymbol_t const* end,
+ U32 nbBits, U32 tableLog,
+ U16 baseSeq, int const level)
+{
+ U32 const length = 1U << ((tableLog - nbBits) & 0x1F /* quiet static-analyzer */);
+ const sortedSymbol_t* ptr;
+ assert(level >= 1 && level <= 2);
+ switch (length) {
+ case 1:
+ for (ptr = begin; ptr != end; ++ptr) {
+ HUF_DEltX2 const DElt = HUF_buildDEltX2(ptr->symbol, nbBits, baseSeq, level);
+ *DTableRank++ = DElt;
+ }
+ break;
+ case 2:
+ for (ptr = begin; ptr != end; ++ptr) {
+ HUF_DEltX2 const DElt = HUF_buildDEltX2(ptr->symbol, nbBits, baseSeq, level);
+ DTableRank[0] = DElt;
+ DTableRank[1] = DElt;
+ DTableRank += 2;
+ }
+ break;
+ case 4:
+ for (ptr = begin; ptr != end; ++ptr) {
+ U64 const DEltX2 = HUF_buildDEltX2U64(ptr->symbol, nbBits, baseSeq, level);
+ ZSTD_memcpy(DTableRank + 0, &DEltX2, sizeof(DEltX2));
+ ZSTD_memcpy(DTableRank + 2, &DEltX2, sizeof(DEltX2));
+ DTableRank += 4;
+ }
+ break;
+ case 8:
+ for (ptr = begin; ptr != end; ++ptr) {
+ U64 const DEltX2 = HUF_buildDEltX2U64(ptr->symbol, nbBits, baseSeq, level);
+ ZSTD_memcpy(DTableRank + 0, &DEltX2, sizeof(DEltX2));
+ ZSTD_memcpy(DTableRank + 2, &DEltX2, sizeof(DEltX2));
+ ZSTD_memcpy(DTableRank + 4, &DEltX2, sizeof(DEltX2));
+ ZSTD_memcpy(DTableRank + 6, &DEltX2, sizeof(DEltX2));
+ DTableRank += 8;
+ }
+ break;
+ default:
+ for (ptr = begin; ptr != end; ++ptr) {
+ U64 const DEltX2 = HUF_buildDEltX2U64(ptr->symbol, nbBits, baseSeq, level);
+ HUF_DEltX2* const DTableRankEnd = DTableRank + length;
+ for (; DTableRank != DTableRankEnd; DTableRank += 8) {
+ ZSTD_memcpy(DTableRank + 0, &DEltX2, sizeof(DEltX2));
+ ZSTD_memcpy(DTableRank + 2, &DEltX2, sizeof(DEltX2));
+ ZSTD_memcpy(DTableRank + 4, &DEltX2, sizeof(DEltX2));
+ ZSTD_memcpy(DTableRank + 6, &DEltX2, sizeof(DEltX2));
+ }
+ }
+ break;
+ }
+}
+
+/* HUF_fillDTableX2Level2() :
+ * `rankValOrigin` must be a table of at least (HUF_TABLELOG_MAX + 1) U32 */
+static void HUF_fillDTableX2Level2(HUF_DEltX2* DTable, U32 targetLog, const U32 consumedBits,
+ const U32* rankVal, const int minWeight, const int maxWeight1,
+ const sortedSymbol_t* sortedSymbols, U32 const* rankStart,
+ U32 nbBitsBaseline, U16 baseSeq)
+{
+ /* Fill skipped values (all positions up to rankVal[minWeight]).
+ * These are positions only get a single symbol because the combined weight
+ * is too large.
+ */
+ if (minWeight>1) {
+ U32 const length = 1U << ((targetLog - consumedBits) & 0x1F /* quiet static-analyzer */);
+ U64 const DEltX2 = HUF_buildDEltX2U64(baseSeq, consumedBits, /* baseSeq */ 0, /* level */ 1);
+ int const skipSize = rankVal[minWeight];
+ assert(length > 1);
+ assert((U32)skipSize < length);
+ switch (length) {
+ case 2:
+ assert(skipSize == 1);
+ ZSTD_memcpy(DTable, &DEltX2, sizeof(DEltX2));
+ break;
+ case 4:
+ assert(skipSize <= 4);
+ ZSTD_memcpy(DTable + 0, &DEltX2, sizeof(DEltX2));
+ ZSTD_memcpy(DTable + 2, &DEltX2, sizeof(DEltX2));
+ break;
+ default:
+ {
+ int i;
+ for (i = 0; i < skipSize; i += 8) {
+ ZSTD_memcpy(DTable + i + 0, &DEltX2, sizeof(DEltX2));
+ ZSTD_memcpy(DTable + i + 2, &DEltX2, sizeof(DEltX2));
+ ZSTD_memcpy(DTable + i + 4, &DEltX2, sizeof(DEltX2));
+ ZSTD_memcpy(DTable + i + 6, &DEltX2, sizeof(DEltX2));
+ }
+ }
+ }
+ }
+
+ /* Fill each of the second level symbols by weight. */
+ {
+ int w;
+ for (w = minWeight; w < maxWeight1; ++w) {
+ int const begin = rankStart[w];
+ int const end = rankStart[w+1];
+ U32 const nbBits = nbBitsBaseline - w;
+ U32 const totalBits = nbBits + consumedBits;
+ HUF_fillDTableX2ForWeight(
+ DTable + rankVal[w],
+ sortedSymbols + begin, sortedSymbols + end,
+ totalBits, targetLog,
+ baseSeq, /* level */ 2);
+ }
+ }
+}
+
+static void HUF_fillDTableX2(HUF_DEltX2* DTable, const U32 targetLog,
+ const sortedSymbol_t* sortedList,
+ const U32* rankStart, rankValCol_t* rankValOrigin, const U32 maxWeight,
+ const U32 nbBitsBaseline)
+{
+ U32* const rankVal = rankValOrigin[0];
+ const int scaleLog = nbBitsBaseline - targetLog; /* note : targetLog >= srcLog, hence scaleLog <= 1 */
+ const U32 minBits = nbBitsBaseline - maxWeight;
+ int w;
+ int const wEnd = (int)maxWeight + 1;
+
+ /* Fill DTable in order of weight. */
+ for (w = 1; w < wEnd; ++w) {
+ int const begin = (int)rankStart[w];
+ int const end = (int)rankStart[w+1];
+ U32 const nbBits = nbBitsBaseline - w;
+
+ if (targetLog-nbBits >= minBits) {
+ /* Enough room for a second symbol. */
+ int start = rankVal[w];
+ U32 const length = 1U << ((targetLog - nbBits) & 0x1F /* quiet static-analyzer */);
+ int minWeight = nbBits + scaleLog;
+ int s;
+ if (minWeight < 1) minWeight = 1;
+ /* Fill the DTable for every symbol of weight w.
+ * These symbols get at least 1 second symbol.
+ */
+ for (s = begin; s != end; ++s) {
+ HUF_fillDTableX2Level2(
+ DTable + start, targetLog, nbBits,
+ rankValOrigin[nbBits], minWeight, wEnd,
+ sortedList, rankStart,
+ nbBitsBaseline, sortedList[s].symbol);
+ start += length;
+ }
+ } else {
+ /* Only a single symbol. */
+ HUF_fillDTableX2ForWeight(
+ DTable + rankVal[w],
+ sortedList + begin, sortedList + end,
+ nbBits, targetLog,
+ /* baseSeq */ 0, /* level */ 1);
+ }
+ }
+}
+
+typedef struct {
+ rankValCol_t rankVal[HUF_TABLELOG_MAX];
+ U32 rankStats[HUF_TABLELOG_MAX + 1];
+ U32 rankStart0[HUF_TABLELOG_MAX + 3];
+ sortedSymbol_t sortedSymbol[HUF_SYMBOLVALUE_MAX + 1];
+ BYTE weightList[HUF_SYMBOLVALUE_MAX + 1];
+ U32 calleeWksp[HUF_READ_STATS_WORKSPACE_SIZE_U32];
+} HUF_ReadDTableX2_Workspace;
+
+size_t HUF_readDTableX2_wksp(HUF_DTable* DTable,
+ const void* src, size_t srcSize,
+ void* workSpace, size_t wkspSize, int flags)
+{
+ U32 tableLog, maxW, nbSymbols;
+ DTableDesc dtd = HUF_getDTableDesc(DTable);
+ U32 maxTableLog = dtd.maxTableLog;
+ size_t iSize;
+ void* dtPtr = DTable+1; /* force compiler to avoid strict-aliasing */
+ HUF_DEltX2* const dt = (HUF_DEltX2*)dtPtr;
+ U32 *rankStart;
+
+ HUF_ReadDTableX2_Workspace* const wksp = (HUF_ReadDTableX2_Workspace*)workSpace;
+
+ if (sizeof(*wksp) > wkspSize) return ERROR(GENERIC);
+
+ rankStart = wksp->rankStart0 + 1;
+ ZSTD_memset(wksp->rankStats, 0, sizeof(wksp->rankStats));
+ ZSTD_memset(wksp->rankStart0, 0, sizeof(wksp->rankStart0));
+
+ DEBUG_STATIC_ASSERT(sizeof(HUF_DEltX2) == sizeof(HUF_DTable)); /* if compiler fails here, assertion is wrong */
+ if (maxTableLog > HUF_TABLELOG_MAX) return ERROR(tableLog_tooLarge);
+ /* ZSTD_memset(weightList, 0, sizeof(weightList)); */ /* is not necessary, even though some analyzer complain ... */
+
+ iSize = HUF_readStats_wksp(wksp->weightList, HUF_SYMBOLVALUE_MAX + 1, wksp->rankStats, &nbSymbols, &tableLog, src, srcSize, wksp->calleeWksp, sizeof(wksp->calleeWksp), flags);
+ if (HUF_isError(iSize)) return iSize;
+
+ /* check result */
+ if (tableLog > maxTableLog) return ERROR(tableLog_tooLarge); /* DTable can't fit code depth */
+ if (tableLog <= HUF_DECODER_FAST_TABLELOG && maxTableLog > HUF_DECODER_FAST_TABLELOG) maxTableLog = HUF_DECODER_FAST_TABLELOG;
+
+ /* find maxWeight */
+ for (maxW = tableLog; wksp->rankStats[maxW]==0; maxW--) {} /* necessarily finds a solution before 0 */
+
+ /* Get start index of each weight */
+ { U32 w, nextRankStart = 0;
+ for (w=1; w<maxW+1; w++) {
+ U32 curr = nextRankStart;
+ nextRankStart += wksp->rankStats[w];
+ rankStart[w] = curr;
+ }
+ rankStart[0] = nextRankStart; /* put all 0w symbols at the end of sorted list*/
+ rankStart[maxW+1] = nextRankStart;
+ }
+
+ /* sort symbols by weight */
+ { U32 s;
+ for (s=0; s<nbSymbols; s++) {
+ U32 const w = wksp->weightList[s];
+ U32 const r = rankStart[w]++;
+ wksp->sortedSymbol[r].symbol = (BYTE)s;
+ }
+ rankStart[0] = 0; /* forget 0w symbols; this is beginning of weight(1) */
+ }
+
+ /* Build rankVal */
+ { U32* const rankVal0 = wksp->rankVal[0];
+ { int const rescale = (maxTableLog-tableLog) - 1; /* tableLog <= maxTableLog */
+ U32 nextRankVal = 0;
+ U32 w;
+ for (w=1; w<maxW+1; w++) {
+ U32 curr = nextRankVal;
+ nextRankVal += wksp->rankStats[w] << (w+rescale);
+ rankVal0[w] = curr;
+ } }
+ { U32 const minBits = tableLog+1 - maxW;
+ U32 consumed;
+ for (consumed = minBits; consumed < maxTableLog - minBits + 1; consumed++) {
+ U32* const rankValPtr = wksp->rankVal[consumed];
+ U32 w;
+ for (w = 1; w < maxW+1; w++) {
+ rankValPtr[w] = rankVal0[w] >> consumed;
+ } } } }
+
+ HUF_fillDTableX2(dt, maxTableLog,
+ wksp->sortedSymbol,
+ wksp->rankStart0, wksp->rankVal, maxW,
+ tableLog+1);
+
+ dtd.tableLog = (BYTE)maxTableLog;
+ dtd.tableType = 1;
+ ZSTD_memcpy(DTable, &dtd, sizeof(dtd));
+ return iSize;
+}
+
+
+FORCE_INLINE_TEMPLATE U32
+HUF_decodeSymbolX2(void* op, BIT_DStream_t* DStream, const HUF_DEltX2* dt, const U32 dtLog)
+{
+ size_t const val = BIT_lookBitsFast(DStream, dtLog); /* note : dtLog >= 1 */
+ ZSTD_memcpy(op, &dt[val].sequence, 2);
+ BIT_skipBits(DStream, dt[val].nbBits);
+ return dt[val].length;
+}
+
+FORCE_INLINE_TEMPLATE U32
+HUF_decodeLastSymbolX2(void* op, BIT_DStream_t* DStream, const HUF_DEltX2* dt, const U32 dtLog)
+{
+ size_t const val = BIT_lookBitsFast(DStream, dtLog); /* note : dtLog >= 1 */
+ ZSTD_memcpy(op, &dt[val].sequence, 1);
+ if (dt[val].length==1) {
+ BIT_skipBits(DStream, dt[val].nbBits);
+ } else {
+ if (DStream->bitsConsumed < (sizeof(DStream->bitContainer)*8)) {
+ BIT_skipBits(DStream, dt[val].nbBits);
+ if (DStream->bitsConsumed > (sizeof(DStream->bitContainer)*8))
+ /* ugly hack; works only because it's the last symbol. Note : can't easily extract nbBits from just this symbol */
+ DStream->bitsConsumed = (sizeof(DStream->bitContainer)*8);
+ }
+ }
+ return 1;
+}
+
+#define HUF_DECODE_SYMBOLX2_0(ptr, DStreamPtr) \
+ do { ptr += HUF_decodeSymbolX2(ptr, DStreamPtr, dt, dtLog); } while (0)
+
+#define HUF_DECODE_SYMBOLX2_1(ptr, DStreamPtr) \
+ do { \
+ if (MEM_64bits() || (HUF_TABLELOG_MAX<=12)) \
+ ptr += HUF_decodeSymbolX2(ptr, DStreamPtr, dt, dtLog); \
+ } while (0)
+
+#define HUF_DECODE_SYMBOLX2_2(ptr, DStreamPtr) \
+ do { \
+ if (MEM_64bits()) \
+ ptr += HUF_decodeSymbolX2(ptr, DStreamPtr, dt, dtLog); \
+ } while (0)
+
+HINT_INLINE size_t
+HUF_decodeStreamX2(BYTE* p, BIT_DStream_t* bitDPtr, BYTE* const pEnd,
+ const HUF_DEltX2* const dt, const U32 dtLog)
+{
+ BYTE* const pStart = p;
+
+ /* up to 8 symbols at a time */
+ if ((size_t)(pEnd - p) >= sizeof(bitDPtr->bitContainer)) {
+ if (dtLog <= 11 && MEM_64bits()) {
+ /* up to 10 symbols at a time */
+ while ((BIT_reloadDStream(bitDPtr) == BIT_DStream_unfinished) & (p < pEnd-9)) {
+ HUF_DECODE_SYMBOLX2_0(p, bitDPtr);
+ HUF_DECODE_SYMBOLX2_0(p, bitDPtr);
+ HUF_DECODE_SYMBOLX2_0(p, bitDPtr);
+ HUF_DECODE_SYMBOLX2_0(p, bitDPtr);
+ HUF_DECODE_SYMBOLX2_0(p, bitDPtr);
+ }
+ } else {
+ /* up to 8 symbols at a time */
+ while ((BIT_reloadDStream(bitDPtr) == BIT_DStream_unfinished) & (p < pEnd-(sizeof(bitDPtr->bitContainer)-1))) {
+ HUF_DECODE_SYMBOLX2_2(p, bitDPtr);
+ HUF_DECODE_SYMBOLX2_1(p, bitDPtr);
+ HUF_DECODE_SYMBOLX2_2(p, bitDPtr);
+ HUF_DECODE_SYMBOLX2_0(p, bitDPtr);
+ }
+ }
+ } else {
+ BIT_reloadDStream(bitDPtr);
+ }
+
+ /* closer to end : up to 2 symbols at a time */
+ if ((size_t)(pEnd - p) >= 2) {
+ while ((BIT_reloadDStream(bitDPtr) == BIT_DStream_unfinished) & (p <= pEnd-2))
+ HUF_DECODE_SYMBOLX2_0(p, bitDPtr);
+
+ while (p <= pEnd-2)
+ HUF_DECODE_SYMBOLX2_0(p, bitDPtr); /* no need to reload : reached the end of DStream */
+ }
+
+ if (p < pEnd)
+ p += HUF_decodeLastSymbolX2(p, bitDPtr, dt, dtLog);
+
+ return p-pStart;
+}
+
+FORCE_INLINE_TEMPLATE size_t
+HUF_decompress1X2_usingDTable_internal_body(
+ void* dst, size_t dstSize,
+ const void* cSrc, size_t cSrcSize,
+ const HUF_DTable* DTable)
+{
+ BIT_DStream_t bitD;
+
+ /* Init */
+ CHECK_F( BIT_initDStream(&bitD, cSrc, cSrcSize) );
+
+ /* decode */
+ { BYTE* const ostart = (BYTE*) dst;
+ BYTE* const oend = ZSTD_maybeNullPtrAdd(ostart, dstSize);
+ const void* const dtPtr = DTable+1; /* force compiler to not use strict-aliasing */
+ const HUF_DEltX2* const dt = (const HUF_DEltX2*)dtPtr;
+ DTableDesc const dtd = HUF_getDTableDesc(DTable);
+ HUF_decodeStreamX2(ostart, &bitD, oend, dt, dtd.tableLog);
+ }
+
+ /* check */
+ if (!BIT_endOfDStream(&bitD)) return ERROR(corruption_detected);
+
+ /* decoded size */
+ return dstSize;
+}
+
+/* HUF_decompress4X2_usingDTable_internal_body():
+ * Conditions:
+ * @dstSize >= 6
+ */
+FORCE_INLINE_TEMPLATE size_t
+HUF_decompress4X2_usingDTable_internal_body(
+ void* dst, size_t dstSize,
+ const void* cSrc, size_t cSrcSize,
+ const HUF_DTable* DTable)
+{
+ if (cSrcSize < 10) return ERROR(corruption_detected); /* strict minimum : jump table + 1 byte per stream */
+ if (dstSize < 6) return ERROR(corruption_detected); /* stream 4-split doesn't work */
+
+ { const BYTE* const istart = (const BYTE*) cSrc;
+ BYTE* const ostart = (BYTE*) dst;
+ BYTE* const oend = ostart + dstSize;
+ BYTE* const olimit = oend - (sizeof(size_t)-1);
+ const void* const dtPtr = DTable+1;
+ const HUF_DEltX2* const dt = (const HUF_DEltX2*)dtPtr;
+
+ /* Init */
+ BIT_DStream_t bitD1;
+ BIT_DStream_t bitD2;
+ BIT_DStream_t bitD3;
+ BIT_DStream_t bitD4;
+ size_t const length1 = MEM_readLE16(istart);
+ size_t const length2 = MEM_readLE16(istart+2);
+ size_t const length3 = MEM_readLE16(istart+4);
+ size_t const length4 = cSrcSize - (length1 + length2 + length3 + 6);
+ const BYTE* const istart1 = istart + 6; /* jumpTable */
+ const BYTE* const istart2 = istart1 + length1;
+ const BYTE* const istart3 = istart2 + length2;
+ const BYTE* const istart4 = istart3 + length3;
+ size_t const segmentSize = (dstSize+3) / 4;
+ BYTE* const opStart2 = ostart + segmentSize;
+ BYTE* const opStart3 = opStart2 + segmentSize;
+ BYTE* const opStart4 = opStart3 + segmentSize;
+ BYTE* op1 = ostart;
+ BYTE* op2 = opStart2;
+ BYTE* op3 = opStart3;
+ BYTE* op4 = opStart4;
+ U32 endSignal = 1;
+ DTableDesc const dtd = HUF_getDTableDesc(DTable);
+ U32 const dtLog = dtd.tableLog;
+
+ if (length4 > cSrcSize) return ERROR(corruption_detected); /* overflow */
+ if (opStart4 > oend) return ERROR(corruption_detected); /* overflow */
+ assert(dstSize >= 6 /* validated above */);
+ CHECK_F( BIT_initDStream(&bitD1, istart1, length1) );
+ CHECK_F( BIT_initDStream(&bitD2, istart2, length2) );
+ CHECK_F( BIT_initDStream(&bitD3, istart3, length3) );
+ CHECK_F( BIT_initDStream(&bitD4, istart4, length4) );
+
+ /* 16-32 symbols per loop (4-8 symbols per stream) */
+ if ((size_t)(oend - op4) >= sizeof(size_t)) {
+ for ( ; (endSignal) & (op4 < olimit); ) {
+#if defined(__clang__) && (defined(__x86_64__) || defined(__i386__))
+ HUF_DECODE_SYMBOLX2_2(op1, &bitD1);
+ HUF_DECODE_SYMBOLX2_1(op1, &bitD1);
+ HUF_DECODE_SYMBOLX2_2(op1, &bitD1);
+ HUF_DECODE_SYMBOLX2_0(op1, &bitD1);
+ HUF_DECODE_SYMBOLX2_2(op2, &bitD2);
+ HUF_DECODE_SYMBOLX2_1(op2, &bitD2);
+ HUF_DECODE_SYMBOLX2_2(op2, &bitD2);
+ HUF_DECODE_SYMBOLX2_0(op2, &bitD2);
+ endSignal &= BIT_reloadDStreamFast(&bitD1) == BIT_DStream_unfinished;
+ endSignal &= BIT_reloadDStreamFast(&bitD2) == BIT_DStream_unfinished;
+ HUF_DECODE_SYMBOLX2_2(op3, &bitD3);
+ HUF_DECODE_SYMBOLX2_1(op3, &bitD3);
+ HUF_DECODE_SYMBOLX2_2(op3, &bitD3);
+ HUF_DECODE_SYMBOLX2_0(op3, &bitD3);
+ HUF_DECODE_SYMBOLX2_2(op4, &bitD4);
+ HUF_DECODE_SYMBOLX2_1(op4, &bitD4);
+ HUF_DECODE_SYMBOLX2_2(op4, &bitD4);
+ HUF_DECODE_SYMBOLX2_0(op4, &bitD4);
+ endSignal &= BIT_reloadDStreamFast(&bitD3) == BIT_DStream_unfinished;
+ endSignal &= BIT_reloadDStreamFast(&bitD4) == BIT_DStream_unfinished;
+#else
+ HUF_DECODE_SYMBOLX2_2(op1, &bitD1);
+ HUF_DECODE_SYMBOLX2_2(op2, &bitD2);
+ HUF_DECODE_SYMBOLX2_2(op3, &bitD3);
+ HUF_DECODE_SYMBOLX2_2(op4, &bitD4);
+ HUF_DECODE_SYMBOLX2_1(op1, &bitD1);
+ HUF_DECODE_SYMBOLX2_1(op2, &bitD2);
+ HUF_DECODE_SYMBOLX2_1(op3, &bitD3);
+ HUF_DECODE_SYMBOLX2_1(op4, &bitD4);
+ HUF_DECODE_SYMBOLX2_2(op1, &bitD1);
+ HUF_DECODE_SYMBOLX2_2(op2, &bitD2);
+ HUF_DECODE_SYMBOLX2_2(op3, &bitD3);
+ HUF_DECODE_SYMBOLX2_2(op4, &bitD4);
+ HUF_DECODE_SYMBOLX2_0(op1, &bitD1);
+ HUF_DECODE_SYMBOLX2_0(op2, &bitD2);
+ HUF_DECODE_SYMBOLX2_0(op3, &bitD3);
+ HUF_DECODE_SYMBOLX2_0(op4, &bitD4);
+ endSignal = (U32)LIKELY((U32)
+ (BIT_reloadDStreamFast(&bitD1) == BIT_DStream_unfinished)
+ & (BIT_reloadDStreamFast(&bitD2) == BIT_DStream_unfinished)
+ & (BIT_reloadDStreamFast(&bitD3) == BIT_DStream_unfinished)
+ & (BIT_reloadDStreamFast(&bitD4) == BIT_DStream_unfinished));
+#endif
+ }
+ }
+
+ /* check corruption */
+ if (op1 > opStart2) return ERROR(corruption_detected);
+ if (op2 > opStart3) return ERROR(corruption_detected);
+ if (op3 > opStart4) return ERROR(corruption_detected);
+ /* note : op4 already verified within main loop */
+
+ /* finish bitStreams one by one */
+ HUF_decodeStreamX2(op1, &bitD1, opStart2, dt, dtLog);
+ HUF_decodeStreamX2(op2, &bitD2, opStart3, dt, dtLog);
+ HUF_decodeStreamX2(op3, &bitD3, opStart4, dt, dtLog);
+ HUF_decodeStreamX2(op4, &bitD4, oend, dt, dtLog);
+
+ /* check */
+ { U32 const endCheck = BIT_endOfDStream(&bitD1) & BIT_endOfDStream(&bitD2) & BIT_endOfDStream(&bitD3) & BIT_endOfDStream(&bitD4);
+ if (!endCheck) return ERROR(corruption_detected); }
+
+ /* decoded size */
+ return dstSize;
+ }
+}
+
+#if HUF_NEED_BMI2_FUNCTION
+static BMI2_TARGET_ATTRIBUTE
+size_t HUF_decompress4X2_usingDTable_internal_bmi2(void* dst, size_t dstSize, void const* cSrc,
+ size_t cSrcSize, HUF_DTable const* DTable) {
+ return HUF_decompress4X2_usingDTable_internal_body(dst, dstSize, cSrc, cSrcSize, DTable);
+}
+#endif
+
+static
+size_t HUF_decompress4X2_usingDTable_internal_default(void* dst, size_t dstSize, void const* cSrc,
+ size_t cSrcSize, HUF_DTable const* DTable) {
+ return HUF_decompress4X2_usingDTable_internal_body(dst, dstSize, cSrc, cSrcSize, DTable);
+}
+
+#if ZSTD_ENABLE_ASM_X86_64_BMI2
+
+HUF_ASM_DECL void HUF_decompress4X2_usingDTable_internal_fast_asm_loop(HUF_DecompressFastArgs* args) ZSTDLIB_HIDDEN;
+
+#endif
+
+static HUF_FAST_BMI2_ATTRS
+void HUF_decompress4X2_usingDTable_internal_fast_c_loop(HUF_DecompressFastArgs* args)
+{
+ U64 bits[4];
+ BYTE const* ip[4];
+ BYTE* op[4];
+ BYTE* oend[4];
+ HUF_DEltX2 const* const dtable = (HUF_DEltX2 const*)args->dt;
+ BYTE const* const ilowest = args->ilowest;
+
+ /* Copy the arguments to local registers. */
+ ZSTD_memcpy(&bits, &args->bits, sizeof(bits));
+ ZSTD_memcpy((void*)(&ip), &args->ip, sizeof(ip));
+ ZSTD_memcpy(&op, &args->op, sizeof(op));
+
+ oend[0] = op[1];
+ oend[1] = op[2];
+ oend[2] = op[3];
+ oend[3] = args->oend;
+
+ assert(MEM_isLittleEndian());
+ assert(!MEM_32bits());
+
+ for (;;) {
+ BYTE* olimit;
+ int stream;
+
+ /* Assert loop preconditions */
+#ifndef NDEBUG
+ for (stream = 0; stream < 4; ++stream) {
+ assert(op[stream] <= oend[stream]);
+ assert(ip[stream] >= ilowest);
+ }
+#endif
+ /* Compute olimit */
+ {
+ /* Each loop does 5 table lookups for each of the 4 streams.
+ * Each table lookup consumes up to 11 bits of input, and produces
+ * up to 2 bytes of output.
+ */
+ /* We can consume up to 7 bytes of input per iteration per stream.
+ * We also know that each input pointer is >= ip[0]. So we can run
+ * iters loops before running out of input.
+ */
+ size_t iters = (size_t)(ip[0] - ilowest) / 7;
+ /* Each iteration can produce up to 10 bytes of output per stream.
+ * Each output stream my advance at different rates. So take the
+ * minimum number of safe iterations among all the output streams.
+ */
+ for (stream = 0; stream < 4; ++stream) {
+ size_t const oiters = (size_t)(oend[stream] - op[stream]) / 10;
+ iters = MIN(iters, oiters);
+ }
+
+ /* Each iteration produces at least 5 output symbols. So until
+ * op[3] crosses olimit, we know we haven't executed iters
+ * iterations yet. This saves us maintaining an iters counter,
+ * at the expense of computing the remaining # of iterations
+ * more frequently.
+ */
+ olimit = op[3] + (iters * 5);
+
+ /* Exit the fast decoding loop once we reach the end. */
+ if (op[3] == olimit)
+ break;
+
+ /* Exit the decoding loop if any input pointer has crossed the
+ * previous one. This indicates corruption, and a precondition
+ * to our loop is that ip[i] >= ip[0].
+ */
+ for (stream = 1; stream < 4; ++stream) {
+ if (ip[stream] < ip[stream - 1])
+ goto _out;
+ }
+ }
+
+#ifndef NDEBUG
+ for (stream = 1; stream < 4; ++stream) {
+ assert(ip[stream] >= ip[stream - 1]);
+ }
+#endif
+
+#define HUF_4X2_DECODE_SYMBOL(_stream, _decode3) \
+ do { \
+ if ((_decode3) || (_stream) != 3) { \
+ int const index = (int)(bits[(_stream)] >> 53); \
+ HUF_DEltX2 const entry = dtable[index]; \
+ MEM_write16(op[(_stream)], entry.sequence); \
+ bits[(_stream)] <<= (entry.nbBits) & 0x3F; \
+ op[(_stream)] += (entry.length); \
+ } \
+ } while (0)
+
+#define HUF_4X2_RELOAD_STREAM(_stream) \
+ do { \
+ HUF_4X2_DECODE_SYMBOL(3, 1); \
+ { \
+ int const ctz = ZSTD_countTrailingZeros64(bits[(_stream)]); \
+ int const nbBits = ctz & 7; \
+ int const nbBytes = ctz >> 3; \
+ ip[(_stream)] -= nbBytes; \
+ bits[(_stream)] = MEM_read64(ip[(_stream)]) | 1; \
+ bits[(_stream)] <<= nbBits; \
+ } \
+ } while (0)
+
+ /* Manually unroll the loop because compilers don't consistently
+ * unroll the inner loops, which destroys performance.
+ */
+ do {
+ /* Decode 5 symbols from each of the first 3 streams.
+ * The final stream will be decoded during the reload phase
+ * to reduce register pressure.
+ */
+ HUF_4X_FOR_EACH_STREAM_WITH_VAR(HUF_4X2_DECODE_SYMBOL, 0);
+ HUF_4X_FOR_EACH_STREAM_WITH_VAR(HUF_4X2_DECODE_SYMBOL, 0);
+ HUF_4X_FOR_EACH_STREAM_WITH_VAR(HUF_4X2_DECODE_SYMBOL, 0);
+ HUF_4X_FOR_EACH_STREAM_WITH_VAR(HUF_4X2_DECODE_SYMBOL, 0);
+ HUF_4X_FOR_EACH_STREAM_WITH_VAR(HUF_4X2_DECODE_SYMBOL, 0);
+
+ /* Decode one symbol from the final stream */
+ HUF_4X2_DECODE_SYMBOL(3, 1);
+
+ /* Decode 4 symbols from the final stream & reload bitstreams.
+ * The final stream is reloaded last, meaning that all 5 symbols
+ * are decoded from the final stream before it is reloaded.
+ */
+ HUF_4X_FOR_EACH_STREAM(HUF_4X2_RELOAD_STREAM);
+ } while (op[3] < olimit);
+ }
+
+#undef HUF_4X2_DECODE_SYMBOL
+#undef HUF_4X2_RELOAD_STREAM
+
+_out:
+
+ /* Save the final values of each of the state variables back to args. */
+ ZSTD_memcpy(&args->bits, &bits, sizeof(bits));
+ ZSTD_memcpy((void*)(&args->ip), &ip, sizeof(ip));
+ ZSTD_memcpy(&args->op, &op, sizeof(op));
+}
+
+
+static HUF_FAST_BMI2_ATTRS size_t
+HUF_decompress4X2_usingDTable_internal_fast(
+ void* dst, size_t dstSize,
+ const void* cSrc, size_t cSrcSize,
+ const HUF_DTable* DTable,
+ HUF_DecompressFastLoopFn loopFn) {
+ void const* dt = DTable + 1;
+ const BYTE* const ilowest = (const BYTE*)cSrc;
+ BYTE* const oend = ZSTD_maybeNullPtrAdd((BYTE*)dst, dstSize);
+ HUF_DecompressFastArgs args;
+ {
+ size_t const ret = HUF_DecompressFastArgs_init(&args, dst, dstSize, cSrc, cSrcSize, DTable);
+ FORWARD_IF_ERROR(ret, "Failed to init asm args");
+ if (ret == 0)
+ return 0;
+ }
+
+ assert(args.ip[0] >= args.ilowest);
+ loopFn(&args);
+
+ /* note : op4 already verified within main loop */
+ assert(args.ip[0] >= ilowest);
+ assert(args.ip[1] >= ilowest);
+ assert(args.ip[2] >= ilowest);
+ assert(args.ip[3] >= ilowest);
+ assert(args.op[3] <= oend);
+
+ assert(ilowest == args.ilowest);
+ assert(ilowest + 6 == args.iend[0]);
+ (void)ilowest;
+
+ /* finish bitStreams one by one */
+ {
+ size_t const segmentSize = (dstSize+3) / 4;
+ BYTE* segmentEnd = (BYTE*)dst;
+ int i;
+ for (i = 0; i < 4; ++i) {
+ BIT_DStream_t bit;
+ if (segmentSize <= (size_t)(oend - segmentEnd))
+ segmentEnd += segmentSize;
+ else
+ segmentEnd = oend;
+ FORWARD_IF_ERROR(HUF_initRemainingDStream(&bit, &args, i, segmentEnd), "corruption");
+ args.op[i] += HUF_decodeStreamX2(args.op[i], &bit, segmentEnd, (HUF_DEltX2 const*)dt, HUF_DECODER_FAST_TABLELOG);
+ if (args.op[i] != segmentEnd)
+ return ERROR(corruption_detected);
+ }
+ }
+
+ /* decoded size */
+ return dstSize;
+}
+
+static size_t HUF_decompress4X2_usingDTable_internal(void* dst, size_t dstSize, void const* cSrc,
+ size_t cSrcSize, HUF_DTable const* DTable, int flags)
+{
+ HUF_DecompressUsingDTableFn fallbackFn = HUF_decompress4X2_usingDTable_internal_default;
+ HUF_DecompressFastLoopFn loopFn = HUF_decompress4X2_usingDTable_internal_fast_c_loop;
+
+#if DYNAMIC_BMI2
+ if (flags & HUF_flags_bmi2) {
+ fallbackFn = HUF_decompress4X2_usingDTable_internal_bmi2;
+# if ZSTD_ENABLE_ASM_X86_64_BMI2
+ if (!(flags & HUF_flags_disableAsm)) {
+ loopFn = HUF_decompress4X2_usingDTable_internal_fast_asm_loop;
+ }
+# endif
+ } else {
+ return fallbackFn(dst, dstSize, cSrc, cSrcSize, DTable);
+ }
+#endif
+
+#if ZSTD_ENABLE_ASM_X86_64_BMI2 && defined(__BMI2__)
+ if (!(flags & HUF_flags_disableAsm)) {
+ loopFn = HUF_decompress4X2_usingDTable_internal_fast_asm_loop;
+ }
+#endif
+
+ if (HUF_ENABLE_FAST_DECODE && !(flags & HUF_flags_disableFast)) {
+ size_t const ret = HUF_decompress4X2_usingDTable_internal_fast(dst, dstSize, cSrc, cSrcSize, DTable, loopFn);
+ if (ret != 0)
+ return ret;
+ }
+ return fallbackFn(dst, dstSize, cSrc, cSrcSize, DTable);
+}
+
+HUF_DGEN(HUF_decompress1X2_usingDTable_internal)
+
+size_t HUF_decompress1X2_DCtx_wksp(HUF_DTable* DCtx, void* dst, size_t dstSize,
+ const void* cSrc, size_t cSrcSize,
+ void* workSpace, size_t wkspSize, int flags)
+{
+ const BYTE* ip = (const BYTE*) cSrc;
+
+ size_t const hSize = HUF_readDTableX2_wksp(DCtx, cSrc, cSrcSize,
+ workSpace, wkspSize, flags);
+ if (HUF_isError(hSize)) return hSize;
+ if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
+ ip += hSize; cSrcSize -= hSize;
+
+ return HUF_decompress1X2_usingDTable_internal(dst, dstSize, ip, cSrcSize, DCtx, flags);
+}
+
+static size_t HUF_decompress4X2_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize,
+ const void* cSrc, size_t cSrcSize,
+ void* workSpace, size_t wkspSize, int flags)
+{
+ const BYTE* ip = (const BYTE*) cSrc;
+
+ size_t hSize = HUF_readDTableX2_wksp(dctx, cSrc, cSrcSize,
+ workSpace, wkspSize, flags);
+ if (HUF_isError(hSize)) return hSize;
+ if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
+ ip += hSize; cSrcSize -= hSize;
+
+ return HUF_decompress4X2_usingDTable_internal(dst, dstSize, ip, cSrcSize, dctx, flags);
+}
+
+#endif /* HUF_FORCE_DECOMPRESS_X1 */
+
+
+/* ***********************************/
+/* Universal decompression selectors */
+/* ***********************************/
+
+
+#if !defined(HUF_FORCE_DECOMPRESS_X1) && !defined(HUF_FORCE_DECOMPRESS_X2)
+typedef struct { U32 tableTime; U32 decode256Time; } algo_time_t;
+static const algo_time_t algoTime[16 /* Quantization */][2 /* single, double */] =
+{
+ /* single, double, quad */
+ {{0,0}, {1,1}}, /* Q==0 : impossible */
+ {{0,0}, {1,1}}, /* Q==1 : impossible */
+ {{ 150,216}, { 381,119}}, /* Q == 2 : 12-18% */
+ {{ 170,205}, { 514,112}}, /* Q == 3 : 18-25% */
+ {{ 177,199}, { 539,110}}, /* Q == 4 : 25-32% */
+ {{ 197,194}, { 644,107}}, /* Q == 5 : 32-38% */
+ {{ 221,192}, { 735,107}}, /* Q == 6 : 38-44% */
+ {{ 256,189}, { 881,106}}, /* Q == 7 : 44-50% */
+ {{ 359,188}, {1167,109}}, /* Q == 8 : 50-56% */
+ {{ 582,187}, {1570,114}}, /* Q == 9 : 56-62% */
+ {{ 688,187}, {1712,122}}, /* Q ==10 : 62-69% */
+ {{ 825,186}, {1965,136}}, /* Q ==11 : 69-75% */
+ {{ 976,185}, {2131,150}}, /* Q ==12 : 75-81% */
+ {{1180,186}, {2070,175}}, /* Q ==13 : 81-87% */
+ {{1377,185}, {1731,202}}, /* Q ==14 : 87-93% */
+ {{1412,185}, {1695,202}}, /* Q ==15 : 93-99% */
+};
+#endif
+
+/** HUF_selectDecoder() :
+ * Tells which decoder is likely to decode faster,
+ * based on a set of pre-computed metrics.
+ * @return : 0==HUF_decompress4X1, 1==HUF_decompress4X2 .
+ * Assumption : 0 < dstSize <= 128 KB */
+U32 HUF_selectDecoder (size_t dstSize, size_t cSrcSize)
+{
+ assert(dstSize > 0);
+ assert(dstSize <= 128*1024);
+#if defined(HUF_FORCE_DECOMPRESS_X1)
+ (void)dstSize;
+ (void)cSrcSize;
+ return 0;
+#elif defined(HUF_FORCE_DECOMPRESS_X2)
+ (void)dstSize;
+ (void)cSrcSize;
+ return 1;
+#else
+ /* decoder timing evaluation */
+ { U32 const Q = (cSrcSize >= dstSize) ? 15 : (U32)(cSrcSize * 16 / dstSize); /* Q < 16 */
+ U32 const D256 = (U32)(dstSize >> 8);
+ U32 const DTime0 = algoTime[Q][0].tableTime + (algoTime[Q][0].decode256Time * D256);
+ U32 DTime1 = algoTime[Q][1].tableTime + (algoTime[Q][1].decode256Time * D256);
+ DTime1 += DTime1 >> 5; /* small advantage to algorithm using less memory, to reduce cache eviction */
+ return DTime1 < DTime0;
+ }
+#endif
+}
+
+size_t HUF_decompress1X_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize,
+ const void* cSrc, size_t cSrcSize,
+ void* workSpace, size_t wkspSize, int flags)
+{
+ /* validation checks */
+ if (dstSize == 0) return ERROR(dstSize_tooSmall);
+ if (cSrcSize > dstSize) return ERROR(corruption_detected); /* invalid */
+ if (cSrcSize == dstSize) { ZSTD_memcpy(dst, cSrc, dstSize); return dstSize; } /* not compressed */
+ if (cSrcSize == 1) { ZSTD_memset(dst, *(const BYTE*)cSrc, dstSize); return dstSize; } /* RLE */
+
+ { U32 const algoNb = HUF_selectDecoder(dstSize, cSrcSize);
+#if defined(HUF_FORCE_DECOMPRESS_X1)
+ (void)algoNb;
+ assert(algoNb == 0);
+ return HUF_decompress1X1_DCtx_wksp(dctx, dst, dstSize, cSrc,
+ cSrcSize, workSpace, wkspSize, flags);
+#elif defined(HUF_FORCE_DECOMPRESS_X2)
+ (void)algoNb;
+ assert(algoNb == 1);
+ return HUF_decompress1X2_DCtx_wksp(dctx, dst, dstSize, cSrc,
+ cSrcSize, workSpace, wkspSize, flags);
+#else
+ return algoNb ? HUF_decompress1X2_DCtx_wksp(dctx, dst, dstSize, cSrc,
+ cSrcSize, workSpace, wkspSize, flags):
+ HUF_decompress1X1_DCtx_wksp(dctx, dst, dstSize, cSrc,
+ cSrcSize, workSpace, wkspSize, flags);
+#endif
+ }
+}
+
+
+size_t HUF_decompress1X_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable, int flags)
+{
+ DTableDesc const dtd = HUF_getDTableDesc(DTable);
+#if defined(HUF_FORCE_DECOMPRESS_X1)
+ (void)dtd;
+ assert(dtd.tableType == 0);
+ return HUF_decompress1X1_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, flags);
+#elif defined(HUF_FORCE_DECOMPRESS_X2)
+ (void)dtd;
+ assert(dtd.tableType == 1);
+ return HUF_decompress1X2_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, flags);
+#else
+ return dtd.tableType ? HUF_decompress1X2_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, flags) :
+ HUF_decompress1X1_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, flags);
+#endif
+}
+
+#ifndef HUF_FORCE_DECOMPRESS_X2
+size_t HUF_decompress1X1_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize, int flags)
+{
+ const BYTE* ip = (const BYTE*) cSrc;
+
+ size_t const hSize = HUF_readDTableX1_wksp(dctx, cSrc, cSrcSize, workSpace, wkspSize, flags);
+ if (HUF_isError(hSize)) return hSize;
+ if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
+ ip += hSize; cSrcSize -= hSize;
+
+ return HUF_decompress1X1_usingDTable_internal(dst, dstSize, ip, cSrcSize, dctx, flags);
+}
+#endif
+
+size_t HUF_decompress4X_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable, int flags)
+{
+ DTableDesc const dtd = HUF_getDTableDesc(DTable);
+#if defined(HUF_FORCE_DECOMPRESS_X1)
+ (void)dtd;
+ assert(dtd.tableType == 0);
+ return HUF_decompress4X1_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, flags);
+#elif defined(HUF_FORCE_DECOMPRESS_X2)
+ (void)dtd;
+ assert(dtd.tableType == 1);
+ return HUF_decompress4X2_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, flags);
+#else
+ return dtd.tableType ? HUF_decompress4X2_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, flags) :
+ HUF_decompress4X1_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, flags);
+#endif
+}
+
+size_t HUF_decompress4X_hufOnly_wksp(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize, int flags)
+{
+ /* validation checks */
+ if (dstSize == 0) return ERROR(dstSize_tooSmall);
+ if (cSrcSize == 0) return ERROR(corruption_detected);
+
+ { U32 const algoNb = HUF_selectDecoder(dstSize, cSrcSize);
+#if defined(HUF_FORCE_DECOMPRESS_X1)
+ (void)algoNb;
+ assert(algoNb == 0);
+ return HUF_decompress4X1_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize, workSpace, wkspSize, flags);
+#elif defined(HUF_FORCE_DECOMPRESS_X2)
+ (void)algoNb;
+ assert(algoNb == 1);
+ return HUF_decompress4X2_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize, workSpace, wkspSize, flags);
+#else
+ return algoNb ? HUF_decompress4X2_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize, workSpace, wkspSize, flags) :
+ HUF_decompress4X1_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize, workSpace, wkspSize, flags);
+#endif
+ }
+}
diff --git a/third_party/zstd/lib/decompress/huf_decompress_amd64.S b/third_party/zstd/lib/decompress/huf_decompress_amd64.S
new file mode 100644
index 0000000000..78da291ee3
--- /dev/null
+++ b/third_party/zstd/lib/decompress/huf_decompress_amd64.S
@@ -0,0 +1,595 @@
+/*
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#include "../common/portability_macros.h"
+
+#if defined(__ELF__) && defined(__GNUC__)
+/* Stack marking
+ * ref: https://wiki.gentoo.org/wiki/Hardened/GNU_stack_quickstart
+ */
+.section .note.GNU-stack,"",%progbits
+
+#if defined(__aarch64__)
+/* Mark that this assembly supports BTI & PAC, because it is empty for aarch64.
+ * See: https://github.com/facebook/zstd/issues/3841
+ * See: https://gcc.godbolt.org/z/sqr5T4ffK
+ * See: https://lore.kernel.org/linux-arm-kernel/20200429211641.9279-8-broonie@kernel.org/
+ * See: https://reviews.llvm.org/D62609
+ */
+.pushsection .note.gnu.property, "a"
+.p2align 3
+.long 4 /* size of the name - "GNU\0" */
+.long 0x10 /* size of descriptor */
+.long 0x5 /* NT_GNU_PROPERTY_TYPE_0 */
+.asciz "GNU"
+.long 0xc0000000 /* pr_type - GNU_PROPERTY_AARCH64_FEATURE_1_AND */
+.long 4 /* pr_datasz - 4 bytes */
+.long 3 /* pr_data - GNU_PROPERTY_AARCH64_FEATURE_1_BTI | GNU_PROPERTY_AARCH64_FEATURE_1_PAC */
+.p2align 3 /* pr_padding - bring everything to 8 byte alignment */
+.popsection
+#endif
+
+#endif
+
+#if ZSTD_ENABLE_ASM_X86_64_BMI2
+
+/* Calling convention:
+ *
+ * %rdi contains the first argument: HUF_DecompressAsmArgs*.
+ * %rbp isn't maintained (no frame pointer).
+ * %rsp contains the stack pointer that grows down.
+ * No red-zone is assumed, only addresses >= %rsp are used.
+ * All register contents are preserved.
+ *
+ * TODO: Support Windows calling convention.
+ */
+
+ZSTD_HIDE_ASM_FUNCTION(HUF_decompress4X1_usingDTable_internal_fast_asm_loop)
+ZSTD_HIDE_ASM_FUNCTION(HUF_decompress4X2_usingDTable_internal_fast_asm_loop)
+ZSTD_HIDE_ASM_FUNCTION(_HUF_decompress4X2_usingDTable_internal_fast_asm_loop)
+ZSTD_HIDE_ASM_FUNCTION(_HUF_decompress4X1_usingDTable_internal_fast_asm_loop)
+.global HUF_decompress4X1_usingDTable_internal_fast_asm_loop
+.global HUF_decompress4X2_usingDTable_internal_fast_asm_loop
+.global _HUF_decompress4X1_usingDTable_internal_fast_asm_loop
+.global _HUF_decompress4X2_usingDTable_internal_fast_asm_loop
+.text
+
+/* Sets up register mappings for clarity.
+ * op[], bits[], dtable & ip[0] each get their own register.
+ * ip[1,2,3] & olimit alias var[].
+ * %rax is a scratch register.
+ */
+
+#define op0 rsi
+#define op1 rbx
+#define op2 rcx
+#define op3 rdi
+
+#define ip0 r8
+#define ip1 r9
+#define ip2 r10
+#define ip3 r11
+
+#define bits0 rbp
+#define bits1 rdx
+#define bits2 r12
+#define bits3 r13
+#define dtable r14
+#define olimit r15
+
+/* var[] aliases ip[1,2,3] & olimit
+ * ip[1,2,3] are saved every iteration.
+ * olimit is only used in compute_olimit.
+ */
+#define var0 r15
+#define var1 r9
+#define var2 r10
+#define var3 r11
+
+/* 32-bit var registers */
+#define vard0 r15d
+#define vard1 r9d
+#define vard2 r10d
+#define vard3 r11d
+
+/* Calls X(N) for each stream 0, 1, 2, 3. */
+#define FOR_EACH_STREAM(X) \
+ X(0); \
+ X(1); \
+ X(2); \
+ X(3)
+
+/* Calls X(N, idx) for each stream 0, 1, 2, 3. */
+#define FOR_EACH_STREAM_WITH_INDEX(X, idx) \
+ X(0, idx); \
+ X(1, idx); \
+ X(2, idx); \
+ X(3, idx)
+
+/* Define both _HUF_* & HUF_* symbols because MacOS
+ * C symbols are prefixed with '_' & Linux symbols aren't.
+ */
+_HUF_decompress4X1_usingDTable_internal_fast_asm_loop:
+HUF_decompress4X1_usingDTable_internal_fast_asm_loop:
+ ZSTD_CET_ENDBRANCH
+ /* Save all registers - even if they are callee saved for simplicity. */
+ push %rax
+ push %rbx
+ push %rcx
+ push %rdx
+ push %rbp
+ push %rsi
+ push %rdi
+ push %r8
+ push %r9
+ push %r10
+ push %r11
+ push %r12
+ push %r13
+ push %r14
+ push %r15
+
+ /* Read HUF_DecompressAsmArgs* args from %rax */
+ movq %rdi, %rax
+ movq 0(%rax), %ip0
+ movq 8(%rax), %ip1
+ movq 16(%rax), %ip2
+ movq 24(%rax), %ip3
+ movq 32(%rax), %op0
+ movq 40(%rax), %op1
+ movq 48(%rax), %op2
+ movq 56(%rax), %op3
+ movq 64(%rax), %bits0
+ movq 72(%rax), %bits1
+ movq 80(%rax), %bits2
+ movq 88(%rax), %bits3
+ movq 96(%rax), %dtable
+ push %rax /* argument */
+ push 104(%rax) /* ilowest */
+ push 112(%rax) /* oend */
+ push %olimit /* olimit space */
+
+ subq $24, %rsp
+
+.L_4X1_compute_olimit:
+ /* Computes how many iterations we can do safely
+ * %r15, %rax may be clobbered
+ * rbx, rdx must be saved
+ * op3 & ip0 mustn't be clobbered
+ */
+ movq %rbx, 0(%rsp)
+ movq %rdx, 8(%rsp)
+
+ movq 32(%rsp), %rax /* rax = oend */
+ subq %op3, %rax /* rax = oend - op3 */
+
+ /* r15 = (oend - op3) / 5 */
+ movabsq $-3689348814741910323, %rdx
+ mulq %rdx
+ movq %rdx, %r15
+ shrq $2, %r15
+
+ movq %ip0, %rax /* rax = ip0 */
+ movq 40(%rsp), %rdx /* rdx = ilowest */
+ subq %rdx, %rax /* rax = ip0 - ilowest */
+ movq %rax, %rbx /* rbx = ip0 - ilowest */
+
+ /* rdx = (ip0 - ilowest) / 7 */
+ movabsq $2635249153387078803, %rdx
+ mulq %rdx
+ subq %rdx, %rbx
+ shrq %rbx
+ addq %rbx, %rdx
+ shrq $2, %rdx
+
+ /* r15 = min(%rdx, %r15) */
+ cmpq %rdx, %r15
+ cmova %rdx, %r15
+
+ /* r15 = r15 * 5 */
+ leaq (%r15, %r15, 4), %r15
+
+ /* olimit = op3 + r15 */
+ addq %op3, %olimit
+
+ movq 8(%rsp), %rdx
+ movq 0(%rsp), %rbx
+
+ /* If (op3 + 20 > olimit) */
+ movq %op3, %rax /* rax = op3 */
+ cmpq %rax, %olimit /* op3 == olimit */
+ je .L_4X1_exit
+
+ /* If (ip1 < ip0) go to exit */
+ cmpq %ip0, %ip1
+ jb .L_4X1_exit
+
+ /* If (ip2 < ip1) go to exit */
+ cmpq %ip1, %ip2
+ jb .L_4X1_exit
+
+ /* If (ip3 < ip2) go to exit */
+ cmpq %ip2, %ip3
+ jb .L_4X1_exit
+
+/* Reads top 11 bits from bits[n]
+ * Loads dt[bits[n]] into var[n]
+ */
+#define GET_NEXT_DELT(n) \
+ movq $53, %var##n; \
+ shrxq %var##n, %bits##n, %var##n; \
+ movzwl (%dtable,%var##n,2),%vard##n
+
+/* var[n] must contain the DTable entry computed with GET_NEXT_DELT
+ * Moves var[n] to %rax
+ * bits[n] <<= var[n] & 63
+ * op[n][idx] = %rax >> 8
+ * %ah is a way to access bits [8, 16) of %rax
+ */
+#define DECODE_FROM_DELT(n, idx) \
+ movq %var##n, %rax; \
+ shlxq %var##n, %bits##n, %bits##n; \
+ movb %ah, idx(%op##n)
+
+/* Assumes GET_NEXT_DELT has been called.
+ * Calls DECODE_FROM_DELT then GET_NEXT_DELT
+ */
+#define DECODE_AND_GET_NEXT(n, idx) \
+ DECODE_FROM_DELT(n, idx); \
+ GET_NEXT_DELT(n) \
+
+/* // ctz & nbBytes is stored in bits[n]
+ * // nbBits is stored in %rax
+ * ctz = CTZ[bits[n]]
+ * nbBits = ctz & 7
+ * nbBytes = ctz >> 3
+ * op[n] += 5
+ * ip[n] -= nbBytes
+ * // Note: x86-64 is little-endian ==> no bswap
+ * bits[n] = MEM_readST(ip[n]) | 1
+ * bits[n] <<= nbBits
+ */
+#define RELOAD_BITS(n) \
+ bsfq %bits##n, %bits##n; \
+ movq %bits##n, %rax; \
+ andq $7, %rax; \
+ shrq $3, %bits##n; \
+ leaq 5(%op##n), %op##n; \
+ subq %bits##n, %ip##n; \
+ movq (%ip##n), %bits##n; \
+ orq $1, %bits##n; \
+ shlx %rax, %bits##n, %bits##n
+
+ /* Store clobbered variables on the stack */
+ movq %olimit, 24(%rsp)
+ movq %ip1, 0(%rsp)
+ movq %ip2, 8(%rsp)
+ movq %ip3, 16(%rsp)
+
+ /* Call GET_NEXT_DELT for each stream */
+ FOR_EACH_STREAM(GET_NEXT_DELT)
+
+ .p2align 6
+
+.L_4X1_loop_body:
+ /* Decode 5 symbols in each of the 4 streams (20 total)
+ * Must have called GET_NEXT_DELT for each stream
+ */
+ FOR_EACH_STREAM_WITH_INDEX(DECODE_AND_GET_NEXT, 0)
+ FOR_EACH_STREAM_WITH_INDEX(DECODE_AND_GET_NEXT, 1)
+ FOR_EACH_STREAM_WITH_INDEX(DECODE_AND_GET_NEXT, 2)
+ FOR_EACH_STREAM_WITH_INDEX(DECODE_AND_GET_NEXT, 3)
+ FOR_EACH_STREAM_WITH_INDEX(DECODE_FROM_DELT, 4)
+
+ /* Load ip[1,2,3] from stack (var[] aliases them)
+ * ip[] is needed for RELOAD_BITS
+ * Each will be stored back to the stack after RELOAD
+ */
+ movq 0(%rsp), %ip1
+ movq 8(%rsp), %ip2
+ movq 16(%rsp), %ip3
+
+ /* Reload each stream & fetch the next table entry
+ * to prepare for the next iteration
+ */
+ RELOAD_BITS(0)
+ GET_NEXT_DELT(0)
+
+ RELOAD_BITS(1)
+ movq %ip1, 0(%rsp)
+ GET_NEXT_DELT(1)
+
+ RELOAD_BITS(2)
+ movq %ip2, 8(%rsp)
+ GET_NEXT_DELT(2)
+
+ RELOAD_BITS(3)
+ movq %ip3, 16(%rsp)
+ GET_NEXT_DELT(3)
+
+ /* If op3 < olimit: continue the loop */
+ cmp %op3, 24(%rsp)
+ ja .L_4X1_loop_body
+
+ /* Reload ip[1,2,3] from stack */
+ movq 0(%rsp), %ip1
+ movq 8(%rsp), %ip2
+ movq 16(%rsp), %ip3
+
+ /* Re-compute olimit */
+ jmp .L_4X1_compute_olimit
+
+#undef GET_NEXT_DELT
+#undef DECODE_FROM_DELT
+#undef DECODE
+#undef RELOAD_BITS
+.L_4X1_exit:
+ addq $24, %rsp
+
+ /* Restore stack (oend & olimit) */
+ pop %rax /* olimit */
+ pop %rax /* oend */
+ pop %rax /* ilowest */
+ pop %rax /* arg */
+
+ /* Save ip / op / bits */
+ movq %ip0, 0(%rax)
+ movq %ip1, 8(%rax)
+ movq %ip2, 16(%rax)
+ movq %ip3, 24(%rax)
+ movq %op0, 32(%rax)
+ movq %op1, 40(%rax)
+ movq %op2, 48(%rax)
+ movq %op3, 56(%rax)
+ movq %bits0, 64(%rax)
+ movq %bits1, 72(%rax)
+ movq %bits2, 80(%rax)
+ movq %bits3, 88(%rax)
+
+ /* Restore registers */
+ pop %r15
+ pop %r14
+ pop %r13
+ pop %r12
+ pop %r11
+ pop %r10
+ pop %r9
+ pop %r8
+ pop %rdi
+ pop %rsi
+ pop %rbp
+ pop %rdx
+ pop %rcx
+ pop %rbx
+ pop %rax
+ ret
+
+_HUF_decompress4X2_usingDTable_internal_fast_asm_loop:
+HUF_decompress4X2_usingDTable_internal_fast_asm_loop:
+ ZSTD_CET_ENDBRANCH
+ /* Save all registers - even if they are callee saved for simplicity. */
+ push %rax
+ push %rbx
+ push %rcx
+ push %rdx
+ push %rbp
+ push %rsi
+ push %rdi
+ push %r8
+ push %r9
+ push %r10
+ push %r11
+ push %r12
+ push %r13
+ push %r14
+ push %r15
+
+ movq %rdi, %rax
+ movq 0(%rax), %ip0
+ movq 8(%rax), %ip1
+ movq 16(%rax), %ip2
+ movq 24(%rax), %ip3
+ movq 32(%rax), %op0
+ movq 40(%rax), %op1
+ movq 48(%rax), %op2
+ movq 56(%rax), %op3
+ movq 64(%rax), %bits0
+ movq 72(%rax), %bits1
+ movq 80(%rax), %bits2
+ movq 88(%rax), %bits3
+ movq 96(%rax), %dtable
+ push %rax /* argument */
+ push %rax /* olimit */
+ push 104(%rax) /* ilowest */
+
+ movq 112(%rax), %rax
+ push %rax /* oend3 */
+
+ movq %op3, %rax
+ push %rax /* oend2 */
+
+ movq %op2, %rax
+ push %rax /* oend1 */
+
+ movq %op1, %rax
+ push %rax /* oend0 */
+
+ /* Scratch space */
+ subq $8, %rsp
+
+.L_4X2_compute_olimit:
+ /* Computes how many iterations we can do safely
+ * %r15, %rax may be clobbered
+ * rdx must be saved
+ * op[1,2,3,4] & ip0 mustn't be clobbered
+ */
+ movq %rdx, 0(%rsp)
+
+ /* We can consume up to 7 input bytes each iteration. */
+ movq %ip0, %rax /* rax = ip0 */
+ movq 40(%rsp), %rdx /* rdx = ilowest */
+ subq %rdx, %rax /* rax = ip0 - ilowest */
+ movq %rax, %r15 /* r15 = ip0 - ilowest */
+
+ /* rdx = rax / 7 */
+ movabsq $2635249153387078803, %rdx
+ mulq %rdx
+ subq %rdx, %r15
+ shrq %r15
+ addq %r15, %rdx
+ shrq $2, %rdx
+
+ /* r15 = (ip0 - ilowest) / 7 */
+ movq %rdx, %r15
+
+ /* r15 = min(r15, min(oend0 - op0, oend1 - op1, oend2 - op2, oend3 - op3) / 10) */
+ movq 8(%rsp), %rax /* rax = oend0 */
+ subq %op0, %rax /* rax = oend0 - op0 */
+ movq 16(%rsp), %rdx /* rdx = oend1 */
+ subq %op1, %rdx /* rdx = oend1 - op1 */
+
+ cmpq %rax, %rdx
+ cmova %rax, %rdx /* rdx = min(%rdx, %rax) */
+
+ movq 24(%rsp), %rax /* rax = oend2 */
+ subq %op2, %rax /* rax = oend2 - op2 */
+
+ cmpq %rax, %rdx
+ cmova %rax, %rdx /* rdx = min(%rdx, %rax) */
+
+ movq 32(%rsp), %rax /* rax = oend3 */
+ subq %op3, %rax /* rax = oend3 - op3 */
+
+ cmpq %rax, %rdx
+ cmova %rax, %rdx /* rdx = min(%rdx, %rax) */
+
+ movabsq $-3689348814741910323, %rax
+ mulq %rdx
+ shrq $3, %rdx /* rdx = rdx / 10 */
+
+ /* r15 = min(%rdx, %r15) */
+ cmpq %rdx, %r15
+ cmova %rdx, %r15
+
+ /* olimit = op3 + 5 * r15 */
+ movq %r15, %rax
+ leaq (%op3, %rax, 4), %olimit
+ addq %rax, %olimit
+
+ movq 0(%rsp), %rdx
+
+ /* If (op3 + 10 > olimit) */
+ movq %op3, %rax /* rax = op3 */
+ cmpq %rax, %olimit /* op3 == olimit */
+ je .L_4X2_exit
+
+ /* If (ip1 < ip0) go to exit */
+ cmpq %ip0, %ip1
+ jb .L_4X2_exit
+
+ /* If (ip2 < ip1) go to exit */
+ cmpq %ip1, %ip2
+ jb .L_4X2_exit
+
+ /* If (ip3 < ip2) go to exit */
+ cmpq %ip2, %ip3
+ jb .L_4X2_exit
+
+#define DECODE(n, idx) \
+ movq %bits##n, %rax; \
+ shrq $53, %rax; \
+ movzwl 0(%dtable,%rax,4),%r8d; \
+ movzbl 2(%dtable,%rax,4),%r15d; \
+ movzbl 3(%dtable,%rax,4),%eax; \
+ movw %r8w, (%op##n); \
+ shlxq %r15, %bits##n, %bits##n; \
+ addq %rax, %op##n
+
+#define RELOAD_BITS(n) \
+ bsfq %bits##n, %bits##n; \
+ movq %bits##n, %rax; \
+ shrq $3, %bits##n; \
+ andq $7, %rax; \
+ subq %bits##n, %ip##n; \
+ movq (%ip##n), %bits##n; \
+ orq $1, %bits##n; \
+ shlxq %rax, %bits##n, %bits##n
+
+
+ movq %olimit, 48(%rsp)
+
+ .p2align 6
+
+.L_4X2_loop_body:
+ /* We clobber r8, so store it on the stack */
+ movq %r8, 0(%rsp)
+
+ /* Decode 5 symbols from each of the 4 streams (20 symbols total). */
+ FOR_EACH_STREAM_WITH_INDEX(DECODE, 0)
+ FOR_EACH_STREAM_WITH_INDEX(DECODE, 1)
+ FOR_EACH_STREAM_WITH_INDEX(DECODE, 2)
+ FOR_EACH_STREAM_WITH_INDEX(DECODE, 3)
+ FOR_EACH_STREAM_WITH_INDEX(DECODE, 4)
+
+ /* Reload r8 */
+ movq 0(%rsp), %r8
+
+ FOR_EACH_STREAM(RELOAD_BITS)
+
+ cmp %op3, 48(%rsp)
+ ja .L_4X2_loop_body
+ jmp .L_4X2_compute_olimit
+
+#undef DECODE
+#undef RELOAD_BITS
+.L_4X2_exit:
+ addq $8, %rsp
+ /* Restore stack (oend & olimit) */
+ pop %rax /* oend0 */
+ pop %rax /* oend1 */
+ pop %rax /* oend2 */
+ pop %rax /* oend3 */
+ pop %rax /* ilowest */
+ pop %rax /* olimit */
+ pop %rax /* arg */
+
+ /* Save ip / op / bits */
+ movq %ip0, 0(%rax)
+ movq %ip1, 8(%rax)
+ movq %ip2, 16(%rax)
+ movq %ip3, 24(%rax)
+ movq %op0, 32(%rax)
+ movq %op1, 40(%rax)
+ movq %op2, 48(%rax)
+ movq %op3, 56(%rax)
+ movq %bits0, 64(%rax)
+ movq %bits1, 72(%rax)
+ movq %bits2, 80(%rax)
+ movq %bits3, 88(%rax)
+
+ /* Restore registers */
+ pop %r15
+ pop %r14
+ pop %r13
+ pop %r12
+ pop %r11
+ pop %r10
+ pop %r9
+ pop %r8
+ pop %rdi
+ pop %rsi
+ pop %rbp
+ pop %rdx
+ pop %rcx
+ pop %rbx
+ pop %rax
+ ret
+
+#endif
diff --git a/third_party/zstd/lib/decompress/zstd_ddict.c b/third_party/zstd/lib/decompress/zstd_ddict.c
new file mode 100644
index 0000000000..309ec0d036
--- /dev/null
+++ b/third_party/zstd/lib/decompress/zstd_ddict.c
@@ -0,0 +1,244 @@
+/*
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+/* zstd_ddict.c :
+ * concentrates all logic that needs to know the internals of ZSTD_DDict object */
+
+/*-*******************************************************
+* Dependencies
+*********************************************************/
+#include "../common/allocations.h" /* ZSTD_customMalloc, ZSTD_customFree */
+#include "../common/zstd_deps.h" /* ZSTD_memcpy, ZSTD_memmove, ZSTD_memset */
+#include "../common/cpu.h" /* bmi2 */
+#include "../common/mem.h" /* low level memory routines */
+#define FSE_STATIC_LINKING_ONLY
+#include "../common/fse.h"
+#include "../common/huf.h"
+#include "zstd_decompress_internal.h"
+#include "zstd_ddict.h"
+
+#if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT>=1)
+# include "../legacy/zstd_legacy.h"
+#endif
+
+
+
+/*-*******************************************************
+* Types
+*********************************************************/
+struct ZSTD_DDict_s {
+ void* dictBuffer;
+ const void* dictContent;
+ size_t dictSize;
+ ZSTD_entropyDTables_t entropy;
+ U32 dictID;
+ U32 entropyPresent;
+ ZSTD_customMem cMem;
+}; /* typedef'd to ZSTD_DDict within "zstd.h" */
+
+const void* ZSTD_DDict_dictContent(const ZSTD_DDict* ddict)
+{
+ assert(ddict != NULL);
+ return ddict->dictContent;
+}
+
+size_t ZSTD_DDict_dictSize(const ZSTD_DDict* ddict)
+{
+ assert(ddict != NULL);
+ return ddict->dictSize;
+}
+
+void ZSTD_copyDDictParameters(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict)
+{
+ DEBUGLOG(4, "ZSTD_copyDDictParameters");
+ assert(dctx != NULL);
+ assert(ddict != NULL);
+ dctx->dictID = ddict->dictID;
+ dctx->prefixStart = ddict->dictContent;
+ dctx->virtualStart = ddict->dictContent;
+ dctx->dictEnd = (const BYTE*)ddict->dictContent + ddict->dictSize;
+ dctx->previousDstEnd = dctx->dictEnd;
+#ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
+ dctx->dictContentBeginForFuzzing = dctx->prefixStart;
+ dctx->dictContentEndForFuzzing = dctx->previousDstEnd;
+#endif
+ if (ddict->entropyPresent) {
+ dctx->litEntropy = 1;
+ dctx->fseEntropy = 1;
+ dctx->LLTptr = ddict->entropy.LLTable;
+ dctx->MLTptr = ddict->entropy.MLTable;
+ dctx->OFTptr = ddict->entropy.OFTable;
+ dctx->HUFptr = ddict->entropy.hufTable;
+ dctx->entropy.rep[0] = ddict->entropy.rep[0];
+ dctx->entropy.rep[1] = ddict->entropy.rep[1];
+ dctx->entropy.rep[2] = ddict->entropy.rep[2];
+ } else {
+ dctx->litEntropy = 0;
+ dctx->fseEntropy = 0;
+ }
+}
+
+
+static size_t
+ZSTD_loadEntropy_intoDDict(ZSTD_DDict* ddict,
+ ZSTD_dictContentType_e dictContentType)
+{
+ ddict->dictID = 0;
+ ddict->entropyPresent = 0;
+ if (dictContentType == ZSTD_dct_rawContent) return 0;
+
+ if (ddict->dictSize < 8) {
+ if (dictContentType == ZSTD_dct_fullDict)
+ return ERROR(dictionary_corrupted); /* only accept specified dictionaries */
+ return 0; /* pure content mode */
+ }
+ { U32 const magic = MEM_readLE32(ddict->dictContent);
+ if (magic != ZSTD_MAGIC_DICTIONARY) {
+ if (dictContentType == ZSTD_dct_fullDict)
+ return ERROR(dictionary_corrupted); /* only accept specified dictionaries */
+ return 0; /* pure content mode */
+ }
+ }
+ ddict->dictID = MEM_readLE32((const char*)ddict->dictContent + ZSTD_FRAMEIDSIZE);
+
+ /* load entropy tables */
+ RETURN_ERROR_IF(ZSTD_isError(ZSTD_loadDEntropy(
+ &ddict->entropy, ddict->dictContent, ddict->dictSize)),
+ dictionary_corrupted, "");
+ ddict->entropyPresent = 1;
+ return 0;
+}
+
+
+static size_t ZSTD_initDDict_internal(ZSTD_DDict* ddict,
+ const void* dict, size_t dictSize,
+ ZSTD_dictLoadMethod_e dictLoadMethod,
+ ZSTD_dictContentType_e dictContentType)
+{
+ if ((dictLoadMethod == ZSTD_dlm_byRef) || (!dict) || (!dictSize)) {
+ ddict->dictBuffer = NULL;
+ ddict->dictContent = dict;
+ if (!dict) dictSize = 0;
+ } else {
+ void* const internalBuffer = ZSTD_customMalloc(dictSize, ddict->cMem);
+ ddict->dictBuffer = internalBuffer;
+ ddict->dictContent = internalBuffer;
+ if (!internalBuffer) return ERROR(memory_allocation);
+ ZSTD_memcpy(internalBuffer, dict, dictSize);
+ }
+ ddict->dictSize = dictSize;
+ ddict->entropy.hufTable[0] = (HUF_DTable)((ZSTD_HUFFDTABLE_CAPACITY_LOG)*0x1000001); /* cover both little and big endian */
+
+ /* parse dictionary content */
+ FORWARD_IF_ERROR( ZSTD_loadEntropy_intoDDict(ddict, dictContentType) , "");
+
+ return 0;
+}
+
+ZSTD_DDict* ZSTD_createDDict_advanced(const void* dict, size_t dictSize,
+ ZSTD_dictLoadMethod_e dictLoadMethod,
+ ZSTD_dictContentType_e dictContentType,
+ ZSTD_customMem customMem)
+{
+ if ((!customMem.customAlloc) ^ (!customMem.customFree)) return NULL;
+
+ { ZSTD_DDict* const ddict = (ZSTD_DDict*) ZSTD_customMalloc(sizeof(ZSTD_DDict), customMem);
+ if (ddict == NULL) return NULL;
+ ddict->cMem = customMem;
+ { size_t const initResult = ZSTD_initDDict_internal(ddict,
+ dict, dictSize,
+ dictLoadMethod, dictContentType);
+ if (ZSTD_isError(initResult)) {
+ ZSTD_freeDDict(ddict);
+ return NULL;
+ } }
+ return ddict;
+ }
+}
+
+/*! ZSTD_createDDict() :
+* Create a digested dictionary, to start decompression without startup delay.
+* `dict` content is copied inside DDict.
+* Consequently, `dict` can be released after `ZSTD_DDict` creation */
+ZSTD_DDict* ZSTD_createDDict(const void* dict, size_t dictSize)
+{
+ ZSTD_customMem const allocator = { NULL, NULL, NULL };
+ return ZSTD_createDDict_advanced(dict, dictSize, ZSTD_dlm_byCopy, ZSTD_dct_auto, allocator);
+}
+
+/*! ZSTD_createDDict_byReference() :
+ * Create a digested dictionary, to start decompression without startup delay.
+ * Dictionary content is simply referenced, it will be accessed during decompression.
+ * Warning : dictBuffer must outlive DDict (DDict must be freed before dictBuffer) */
+ZSTD_DDict* ZSTD_createDDict_byReference(const void* dictBuffer, size_t dictSize)
+{
+ ZSTD_customMem const allocator = { NULL, NULL, NULL };
+ return ZSTD_createDDict_advanced(dictBuffer, dictSize, ZSTD_dlm_byRef, ZSTD_dct_auto, allocator);
+}
+
+
+const ZSTD_DDict* ZSTD_initStaticDDict(
+ void* sBuffer, size_t sBufferSize,
+ const void* dict, size_t dictSize,
+ ZSTD_dictLoadMethod_e dictLoadMethod,
+ ZSTD_dictContentType_e dictContentType)
+{
+ size_t const neededSpace = sizeof(ZSTD_DDict)
+ + (dictLoadMethod == ZSTD_dlm_byRef ? 0 : dictSize);
+ ZSTD_DDict* const ddict = (ZSTD_DDict*)sBuffer;
+ assert(sBuffer != NULL);
+ assert(dict != NULL);
+ if ((size_t)sBuffer & 7) return NULL; /* 8-aligned */
+ if (sBufferSize < neededSpace) return NULL;
+ if (dictLoadMethod == ZSTD_dlm_byCopy) {
+ ZSTD_memcpy(ddict+1, dict, dictSize); /* local copy */
+ dict = ddict+1;
+ }
+ if (ZSTD_isError( ZSTD_initDDict_internal(ddict,
+ dict, dictSize,
+ ZSTD_dlm_byRef, dictContentType) ))
+ return NULL;
+ return ddict;
+}
+
+
+size_t ZSTD_freeDDict(ZSTD_DDict* ddict)
+{
+ if (ddict==NULL) return 0; /* support free on NULL */
+ { ZSTD_customMem const cMem = ddict->cMem;
+ ZSTD_customFree(ddict->dictBuffer, cMem);
+ ZSTD_customFree(ddict, cMem);
+ return 0;
+ }
+}
+
+/*! ZSTD_estimateDDictSize() :
+ * Estimate amount of memory that will be needed to create a dictionary for decompression.
+ * Note : dictionary created by reference using ZSTD_dlm_byRef are smaller */
+size_t ZSTD_estimateDDictSize(size_t dictSize, ZSTD_dictLoadMethod_e dictLoadMethod)
+{
+ return sizeof(ZSTD_DDict) + (dictLoadMethod == ZSTD_dlm_byRef ? 0 : dictSize);
+}
+
+size_t ZSTD_sizeof_DDict(const ZSTD_DDict* ddict)
+{
+ if (ddict==NULL) return 0; /* support sizeof on NULL */
+ return sizeof(*ddict) + (ddict->dictBuffer ? ddict->dictSize : 0) ;
+}
+
+/*! ZSTD_getDictID_fromDDict() :
+ * Provides the dictID of the dictionary loaded into `ddict`.
+ * If @return == 0, the dictionary is not conformant to Zstandard specification, or empty.
+ * Non-conformant dictionaries can still be loaded, but as content-only dictionaries. */
+unsigned ZSTD_getDictID_fromDDict(const ZSTD_DDict* ddict)
+{
+ if (ddict==NULL) return 0;
+ return ddict->dictID;
+}
diff --git a/third_party/zstd/lib/decompress/zstd_ddict.h b/third_party/zstd/lib/decompress/zstd_ddict.h
new file mode 100644
index 0000000000..c4ca8877a0
--- /dev/null
+++ b/third_party/zstd/lib/decompress/zstd_ddict.h
@@ -0,0 +1,44 @@
+/*
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+
+#ifndef ZSTD_DDICT_H
+#define ZSTD_DDICT_H
+
+/*-*******************************************************
+ * Dependencies
+ *********************************************************/
+#include "../common/zstd_deps.h" /* size_t */
+#include "../zstd.h" /* ZSTD_DDict, and several public functions */
+
+
+/*-*******************************************************
+ * Interface
+ *********************************************************/
+
+/* note: several prototypes are already published in `zstd.h` :
+ * ZSTD_createDDict()
+ * ZSTD_createDDict_byReference()
+ * ZSTD_createDDict_advanced()
+ * ZSTD_freeDDict()
+ * ZSTD_initStaticDDict()
+ * ZSTD_sizeof_DDict()
+ * ZSTD_estimateDDictSize()
+ * ZSTD_getDictID_fromDict()
+ */
+
+const void* ZSTD_DDict_dictContent(const ZSTD_DDict* ddict);
+size_t ZSTD_DDict_dictSize(const ZSTD_DDict* ddict);
+
+void ZSTD_copyDDictParameters(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict);
+
+
+
+#endif /* ZSTD_DDICT_H */
diff --git a/third_party/zstd/lib/decompress/zstd_decompress.c b/third_party/zstd/lib/decompress/zstd_decompress.c
new file mode 100644
index 0000000000..2f03cf7b0c
--- /dev/null
+++ b/third_party/zstd/lib/decompress/zstd_decompress.c
@@ -0,0 +1,2407 @@
+/*
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+
+/* ***************************************************************
+* Tuning parameters
+*****************************************************************/
+/*!
+ * HEAPMODE :
+ * Select how default decompression function ZSTD_decompress() allocates its context,
+ * on stack (0), or into heap (1, default; requires malloc()).
+ * Note that functions with explicit context such as ZSTD_decompressDCtx() are unaffected.
+ */
+#ifndef ZSTD_HEAPMODE
+# define ZSTD_HEAPMODE 1
+#endif
+
+/*!
+* LEGACY_SUPPORT :
+* if set to 1+, ZSTD_decompress() can decode older formats (v0.1+)
+*/
+#ifndef ZSTD_LEGACY_SUPPORT
+# define ZSTD_LEGACY_SUPPORT 0
+#endif
+
+/*!
+ * MAXWINDOWSIZE_DEFAULT :
+ * maximum window size accepted by DStream __by default__.
+ * Frames requiring more memory will be rejected.
+ * It's possible to set a different limit using ZSTD_DCtx_setMaxWindowSize().
+ */
+#ifndef ZSTD_MAXWINDOWSIZE_DEFAULT
+# define ZSTD_MAXWINDOWSIZE_DEFAULT (((U32)1 << ZSTD_WINDOWLOG_LIMIT_DEFAULT) + 1)
+#endif
+
+/*!
+ * NO_FORWARD_PROGRESS_MAX :
+ * maximum allowed nb of calls to ZSTD_decompressStream()
+ * without any forward progress
+ * (defined as: no byte read from input, and no byte flushed to output)
+ * before triggering an error.
+ */
+#ifndef ZSTD_NO_FORWARD_PROGRESS_MAX
+# define ZSTD_NO_FORWARD_PROGRESS_MAX 16
+#endif
+
+
+/*-*******************************************************
+* Dependencies
+*********************************************************/
+#include "../common/zstd_deps.h" /* ZSTD_memcpy, ZSTD_memmove, ZSTD_memset */
+#include "../common/allocations.h" /* ZSTD_customMalloc, ZSTD_customCalloc, ZSTD_customFree */
+#include "../common/error_private.h"
+#include "../common/zstd_internal.h" /* blockProperties_t */
+#include "../common/mem.h" /* low level memory routines */
+#include "../common/bits.h" /* ZSTD_highbit32 */
+#define FSE_STATIC_LINKING_ONLY
+#include "../common/fse.h"
+#include "../common/huf.h"
+#include "../common/xxhash.h" /* XXH64_reset, XXH64_update, XXH64_digest, XXH64 */
+#include "zstd_decompress_internal.h" /* ZSTD_DCtx */
+#include "zstd_ddict.h" /* ZSTD_DDictDictContent */
+#include "zstd_decompress_block.h" /* ZSTD_decompressBlock_internal */
+
+#if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT>=1)
+# include "../legacy/zstd_legacy.h"
+#endif
+
+
+
+/*************************************
+ * Multiple DDicts Hashset internals *
+ *************************************/
+
+#define DDICT_HASHSET_MAX_LOAD_FACTOR_COUNT_MULT 4
+#define DDICT_HASHSET_MAX_LOAD_FACTOR_SIZE_MULT 3 /* These two constants represent SIZE_MULT/COUNT_MULT load factor without using a float.
+ * Currently, that means a 0.75 load factor.
+ * So, if count * COUNT_MULT / size * SIZE_MULT != 0, then we've exceeded
+ * the load factor of the ddict hash set.
+ */
+
+#define DDICT_HASHSET_TABLE_BASE_SIZE 64
+#define DDICT_HASHSET_RESIZE_FACTOR 2
+
+/* Hash function to determine starting position of dict insertion within the table
+ * Returns an index between [0, hashSet->ddictPtrTableSize]
+ */
+static size_t ZSTD_DDictHashSet_getIndex(const ZSTD_DDictHashSet* hashSet, U32 dictID) {
+ const U64 hash = XXH64(&dictID, sizeof(U32), 0);
+ /* DDict ptr table size is a multiple of 2, use size - 1 as mask to get index within [0, hashSet->ddictPtrTableSize) */
+ return hash & (hashSet->ddictPtrTableSize - 1);
+}
+
+/* Adds DDict to a hashset without resizing it.
+ * If inserting a DDict with a dictID that already exists in the set, replaces the one in the set.
+ * Returns 0 if successful, or a zstd error code if something went wrong.
+ */
+static size_t ZSTD_DDictHashSet_emplaceDDict(ZSTD_DDictHashSet* hashSet, const ZSTD_DDict* ddict) {
+ const U32 dictID = ZSTD_getDictID_fromDDict(ddict);
+ size_t idx = ZSTD_DDictHashSet_getIndex(hashSet, dictID);
+ const size_t idxRangeMask = hashSet->ddictPtrTableSize - 1;
+ RETURN_ERROR_IF(hashSet->ddictPtrCount == hashSet->ddictPtrTableSize, GENERIC, "Hash set is full!");
+ DEBUGLOG(4, "Hashed index: for dictID: %u is %zu", dictID, idx);
+ while (hashSet->ddictPtrTable[idx] != NULL) {
+ /* Replace existing ddict if inserting ddict with same dictID */
+ if (ZSTD_getDictID_fromDDict(hashSet->ddictPtrTable[idx]) == dictID) {
+ DEBUGLOG(4, "DictID already exists, replacing rather than adding");
+ hashSet->ddictPtrTable[idx] = ddict;
+ return 0;
+ }
+ idx &= idxRangeMask;
+ idx++;
+ }
+ DEBUGLOG(4, "Final idx after probing for dictID %u is: %zu", dictID, idx);
+ hashSet->ddictPtrTable[idx] = ddict;
+ hashSet->ddictPtrCount++;
+ return 0;
+}
+
+/* Expands hash table by factor of DDICT_HASHSET_RESIZE_FACTOR and
+ * rehashes all values, allocates new table, frees old table.
+ * Returns 0 on success, otherwise a zstd error code.
+ */
+static size_t ZSTD_DDictHashSet_expand(ZSTD_DDictHashSet* hashSet, ZSTD_customMem customMem) {
+ size_t newTableSize = hashSet->ddictPtrTableSize * DDICT_HASHSET_RESIZE_FACTOR;
+ const ZSTD_DDict** newTable = (const ZSTD_DDict**)ZSTD_customCalloc(sizeof(ZSTD_DDict*) * newTableSize, customMem);
+ const ZSTD_DDict** oldTable = hashSet->ddictPtrTable;
+ size_t oldTableSize = hashSet->ddictPtrTableSize;
+ size_t i;
+
+ DEBUGLOG(4, "Expanding DDict hash table! Old size: %zu new size: %zu", oldTableSize, newTableSize);
+ RETURN_ERROR_IF(!newTable, memory_allocation, "Expanded hashset allocation failed!");
+ hashSet->ddictPtrTable = newTable;
+ hashSet->ddictPtrTableSize = newTableSize;
+ hashSet->ddictPtrCount = 0;
+ for (i = 0; i < oldTableSize; ++i) {
+ if (oldTable[i] != NULL) {
+ FORWARD_IF_ERROR(ZSTD_DDictHashSet_emplaceDDict(hashSet, oldTable[i]), "");
+ }
+ }
+ ZSTD_customFree((void*)oldTable, customMem);
+ DEBUGLOG(4, "Finished re-hash");
+ return 0;
+}
+
+/* Fetches a DDict with the given dictID
+ * Returns the ZSTD_DDict* with the requested dictID. If it doesn't exist, then returns NULL.
+ */
+static const ZSTD_DDict* ZSTD_DDictHashSet_getDDict(ZSTD_DDictHashSet* hashSet, U32 dictID) {
+ size_t idx = ZSTD_DDictHashSet_getIndex(hashSet, dictID);
+ const size_t idxRangeMask = hashSet->ddictPtrTableSize - 1;
+ DEBUGLOG(4, "Hashed index: for dictID: %u is %zu", dictID, idx);
+ for (;;) {
+ size_t currDictID = ZSTD_getDictID_fromDDict(hashSet->ddictPtrTable[idx]);
+ if (currDictID == dictID || currDictID == 0) {
+ /* currDictID == 0 implies a NULL ddict entry */
+ break;
+ } else {
+ idx &= idxRangeMask; /* Goes to start of table when we reach the end */
+ idx++;
+ }
+ }
+ DEBUGLOG(4, "Final idx after probing for dictID %u is: %zu", dictID, idx);
+ return hashSet->ddictPtrTable[idx];
+}
+
+/* Allocates space for and returns a ddict hash set
+ * The hash set's ZSTD_DDict* table has all values automatically set to NULL to begin with.
+ * Returns NULL if allocation failed.
+ */
+static ZSTD_DDictHashSet* ZSTD_createDDictHashSet(ZSTD_customMem customMem) {
+ ZSTD_DDictHashSet* ret = (ZSTD_DDictHashSet*)ZSTD_customMalloc(sizeof(ZSTD_DDictHashSet), customMem);
+ DEBUGLOG(4, "Allocating new hash set");
+ if (!ret)
+ return NULL;
+ ret->ddictPtrTable = (const ZSTD_DDict**)ZSTD_customCalloc(DDICT_HASHSET_TABLE_BASE_SIZE * sizeof(ZSTD_DDict*), customMem);
+ if (!ret->ddictPtrTable) {
+ ZSTD_customFree(ret, customMem);
+ return NULL;
+ }
+ ret->ddictPtrTableSize = DDICT_HASHSET_TABLE_BASE_SIZE;
+ ret->ddictPtrCount = 0;
+ return ret;
+}
+
+/* Frees the table of ZSTD_DDict* within a hashset, then frees the hashset itself.
+ * Note: The ZSTD_DDict* within the table are NOT freed.
+ */
+static void ZSTD_freeDDictHashSet(ZSTD_DDictHashSet* hashSet, ZSTD_customMem customMem) {
+ DEBUGLOG(4, "Freeing ddict hash set");
+ if (hashSet && hashSet->ddictPtrTable) {
+ ZSTD_customFree((void*)hashSet->ddictPtrTable, customMem);
+ }
+ if (hashSet) {
+ ZSTD_customFree(hashSet, customMem);
+ }
+}
+
+/* Public function: Adds a DDict into the ZSTD_DDictHashSet, possibly triggering a resize of the hash set.
+ * Returns 0 on success, or a ZSTD error.
+ */
+static size_t ZSTD_DDictHashSet_addDDict(ZSTD_DDictHashSet* hashSet, const ZSTD_DDict* ddict, ZSTD_customMem customMem) {
+ DEBUGLOG(4, "Adding dict ID: %u to hashset with - Count: %zu Tablesize: %zu", ZSTD_getDictID_fromDDict(ddict), hashSet->ddictPtrCount, hashSet->ddictPtrTableSize);
+ if (hashSet->ddictPtrCount * DDICT_HASHSET_MAX_LOAD_FACTOR_COUNT_MULT / hashSet->ddictPtrTableSize * DDICT_HASHSET_MAX_LOAD_FACTOR_SIZE_MULT != 0) {
+ FORWARD_IF_ERROR(ZSTD_DDictHashSet_expand(hashSet, customMem), "");
+ }
+ FORWARD_IF_ERROR(ZSTD_DDictHashSet_emplaceDDict(hashSet, ddict), "");
+ return 0;
+}
+
+/*-*************************************************************
+* Context management
+***************************************************************/
+size_t ZSTD_sizeof_DCtx (const ZSTD_DCtx* dctx)
+{
+ if (dctx==NULL) return 0; /* support sizeof NULL */
+ return sizeof(*dctx)
+ + ZSTD_sizeof_DDict(dctx->ddictLocal)
+ + dctx->inBuffSize + dctx->outBuffSize;
+}
+
+size_t ZSTD_estimateDCtxSize(void) { return sizeof(ZSTD_DCtx); }
+
+
+static size_t ZSTD_startingInputLength(ZSTD_format_e format)
+{
+ size_t const startingInputLength = ZSTD_FRAMEHEADERSIZE_PREFIX(format);
+ /* only supports formats ZSTD_f_zstd1 and ZSTD_f_zstd1_magicless */
+ assert( (format == ZSTD_f_zstd1) || (format == ZSTD_f_zstd1_magicless) );
+ return startingInputLength;
+}
+
+static void ZSTD_DCtx_resetParameters(ZSTD_DCtx* dctx)
+{
+ assert(dctx->streamStage == zdss_init);
+ dctx->format = ZSTD_f_zstd1;
+ dctx->maxWindowSize = ZSTD_MAXWINDOWSIZE_DEFAULT;
+ dctx->outBufferMode = ZSTD_bm_buffered;
+ dctx->forceIgnoreChecksum = ZSTD_d_validateChecksum;
+ dctx->refMultipleDDicts = ZSTD_rmd_refSingleDDict;
+ dctx->disableHufAsm = 0;
+ dctx->maxBlockSizeParam = 0;
+}
+
+static void ZSTD_initDCtx_internal(ZSTD_DCtx* dctx)
+{
+ dctx->staticSize = 0;
+ dctx->ddict = NULL;
+ dctx->ddictLocal = NULL;
+ dctx->dictEnd = NULL;
+ dctx->ddictIsCold = 0;
+ dctx->dictUses = ZSTD_dont_use;
+ dctx->inBuff = NULL;
+ dctx->inBuffSize = 0;
+ dctx->outBuffSize = 0;
+ dctx->streamStage = zdss_init;
+#if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT>=1)
+ dctx->legacyContext = NULL;
+ dctx->previousLegacyVersion = 0;
+#endif
+ dctx->noForwardProgress = 0;
+ dctx->oversizedDuration = 0;
+ dctx->isFrameDecompression = 1;
+#if DYNAMIC_BMI2
+ dctx->bmi2 = ZSTD_cpuSupportsBmi2();
+#endif
+ dctx->ddictSet = NULL;
+ ZSTD_DCtx_resetParameters(dctx);
+#ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
+ dctx->dictContentEndForFuzzing = NULL;
+#endif
+}
+
+ZSTD_DCtx* ZSTD_initStaticDCtx(void *workspace, size_t workspaceSize)
+{
+ ZSTD_DCtx* const dctx = (ZSTD_DCtx*) workspace;
+
+ if ((size_t)workspace & 7) return NULL; /* 8-aligned */
+ if (workspaceSize < sizeof(ZSTD_DCtx)) return NULL; /* minimum size */
+
+ ZSTD_initDCtx_internal(dctx);
+ dctx->staticSize = workspaceSize;
+ dctx->inBuff = (char*)(dctx+1);
+ return dctx;
+}
+
+static ZSTD_DCtx* ZSTD_createDCtx_internal(ZSTD_customMem customMem) {
+ if ((!customMem.customAlloc) ^ (!customMem.customFree)) return NULL;
+
+ { ZSTD_DCtx* const dctx = (ZSTD_DCtx*)ZSTD_customMalloc(sizeof(*dctx), customMem);
+ if (!dctx) return NULL;
+ dctx->customMem = customMem;
+ ZSTD_initDCtx_internal(dctx);
+ return dctx;
+ }
+}
+
+ZSTD_DCtx* ZSTD_createDCtx_advanced(ZSTD_customMem customMem)
+{
+ return ZSTD_createDCtx_internal(customMem);
+}
+
+ZSTD_DCtx* ZSTD_createDCtx(void)
+{
+ DEBUGLOG(3, "ZSTD_createDCtx");
+ return ZSTD_createDCtx_internal(ZSTD_defaultCMem);
+}
+
+static void ZSTD_clearDict(ZSTD_DCtx* dctx)
+{
+ ZSTD_freeDDict(dctx->ddictLocal);
+ dctx->ddictLocal = NULL;
+ dctx->ddict = NULL;
+ dctx->dictUses = ZSTD_dont_use;
+}
+
+size_t ZSTD_freeDCtx(ZSTD_DCtx* dctx)
+{
+ if (dctx==NULL) return 0; /* support free on NULL */
+ RETURN_ERROR_IF(dctx->staticSize, memory_allocation, "not compatible with static DCtx");
+ { ZSTD_customMem const cMem = dctx->customMem;
+ ZSTD_clearDict(dctx);
+ ZSTD_customFree(dctx->inBuff, cMem);
+ dctx->inBuff = NULL;
+#if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT >= 1)
+ if (dctx->legacyContext)
+ ZSTD_freeLegacyStreamContext(dctx->legacyContext, dctx->previousLegacyVersion);
+#endif
+ if (dctx->ddictSet) {
+ ZSTD_freeDDictHashSet(dctx->ddictSet, cMem);
+ dctx->ddictSet = NULL;
+ }
+ ZSTD_customFree(dctx, cMem);
+ return 0;
+ }
+}
+
+/* no longer useful */
+void ZSTD_copyDCtx(ZSTD_DCtx* dstDCtx, const ZSTD_DCtx* srcDCtx)
+{
+ size_t const toCopy = (size_t)((char*)(&dstDCtx->inBuff) - (char*)dstDCtx);
+ ZSTD_memcpy(dstDCtx, srcDCtx, toCopy); /* no need to copy workspace */
+}
+
+/* Given a dctx with a digested frame params, re-selects the correct ZSTD_DDict based on
+ * the requested dict ID from the frame. If there exists a reference to the correct ZSTD_DDict, then
+ * accordingly sets the ddict to be used to decompress the frame.
+ *
+ * If no DDict is found, then no action is taken, and the ZSTD_DCtx::ddict remains as-is.
+ *
+ * ZSTD_d_refMultipleDDicts must be enabled for this function to be called.
+ */
+static void ZSTD_DCtx_selectFrameDDict(ZSTD_DCtx* dctx) {
+ assert(dctx->refMultipleDDicts && dctx->ddictSet);
+ DEBUGLOG(4, "Adjusting DDict based on requested dict ID from frame");
+ if (dctx->ddict) {
+ const ZSTD_DDict* frameDDict = ZSTD_DDictHashSet_getDDict(dctx->ddictSet, dctx->fParams.dictID);
+ if (frameDDict) {
+ DEBUGLOG(4, "DDict found!");
+ ZSTD_clearDict(dctx);
+ dctx->dictID = dctx->fParams.dictID;
+ dctx->ddict = frameDDict;
+ dctx->dictUses = ZSTD_use_indefinitely;
+ }
+ }
+}
+
+
+/*-*************************************************************
+ * Frame header decoding
+ ***************************************************************/
+
+/*! ZSTD_isFrame() :
+ * Tells if the content of `buffer` starts with a valid Frame Identifier.
+ * Note : Frame Identifier is 4 bytes. If `size < 4`, @return will always be 0.
+ * Note 2 : Legacy Frame Identifiers are considered valid only if Legacy Support is enabled.
+ * Note 3 : Skippable Frame Identifiers are considered valid. */
+unsigned ZSTD_isFrame(const void* buffer, size_t size)
+{
+ if (size < ZSTD_FRAMEIDSIZE) return 0;
+ { U32 const magic = MEM_readLE32(buffer);
+ if (magic == ZSTD_MAGICNUMBER) return 1;
+ if ((magic & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) return 1;
+ }
+#if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT >= 1)
+ if (ZSTD_isLegacy(buffer, size)) return 1;
+#endif
+ return 0;
+}
+
+/*! ZSTD_isSkippableFrame() :
+ * Tells if the content of `buffer` starts with a valid Frame Identifier for a skippable frame.
+ * Note : Frame Identifier is 4 bytes. If `size < 4`, @return will always be 0.
+ */
+unsigned ZSTD_isSkippableFrame(const void* buffer, size_t size)
+{
+ if (size < ZSTD_FRAMEIDSIZE) return 0;
+ { U32 const magic = MEM_readLE32(buffer);
+ if ((magic & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) return 1;
+ }
+ return 0;
+}
+
+/** ZSTD_frameHeaderSize_internal() :
+ * srcSize must be large enough to reach header size fields.
+ * note : only works for formats ZSTD_f_zstd1 and ZSTD_f_zstd1_magicless.
+ * @return : size of the Frame Header
+ * or an error code, which can be tested with ZSTD_isError() */
+static size_t ZSTD_frameHeaderSize_internal(const void* src, size_t srcSize, ZSTD_format_e format)
+{
+ size_t const minInputSize = ZSTD_startingInputLength(format);
+ RETURN_ERROR_IF(srcSize < minInputSize, srcSize_wrong, "");
+
+ { BYTE const fhd = ((const BYTE*)src)[minInputSize-1];
+ U32 const dictID= fhd & 3;
+ U32 const singleSegment = (fhd >> 5) & 1;
+ U32 const fcsId = fhd >> 6;
+ return minInputSize + !singleSegment
+ + ZSTD_did_fieldSize[dictID] + ZSTD_fcs_fieldSize[fcsId]
+ + (singleSegment && !fcsId);
+ }
+}
+
+/** ZSTD_frameHeaderSize() :
+ * srcSize must be >= ZSTD_frameHeaderSize_prefix.
+ * @return : size of the Frame Header,
+ * or an error code (if srcSize is too small) */
+size_t ZSTD_frameHeaderSize(const void* src, size_t srcSize)
+{
+ return ZSTD_frameHeaderSize_internal(src, srcSize, ZSTD_f_zstd1);
+}
+
+
+/** ZSTD_getFrameHeader_advanced() :
+ * decode Frame Header, or require larger `srcSize`.
+ * note : only works for formats ZSTD_f_zstd1 and ZSTD_f_zstd1_magicless
+ * @return : 0, `zfhPtr` is correctly filled,
+ * >0, `srcSize` is too small, value is wanted `srcSize` amount,
+** or an error code, which can be tested using ZSTD_isError() */
+size_t ZSTD_getFrameHeader_advanced(ZSTD_frameHeader* zfhPtr, const void* src, size_t srcSize, ZSTD_format_e format)
+{
+ const BYTE* ip = (const BYTE*)src;
+ size_t const minInputSize = ZSTD_startingInputLength(format);
+
+ DEBUGLOG(5, "ZSTD_getFrameHeader_advanced: minInputSize = %zu, srcSize = %zu", minInputSize, srcSize);
+
+ if (srcSize > 0) {
+ /* note : technically could be considered an assert(), since it's an invalid entry */
+ RETURN_ERROR_IF(src==NULL, GENERIC, "invalid parameter : src==NULL, but srcSize>0");
+ }
+ if (srcSize < minInputSize) {
+ if (srcSize > 0 && format != ZSTD_f_zstd1_magicless) {
+ /* when receiving less than @minInputSize bytes,
+ * control these bytes at least correspond to a supported magic number
+ * in order to error out early if they don't.
+ **/
+ size_t const toCopy = MIN(4, srcSize);
+ unsigned char hbuf[4]; MEM_writeLE32(hbuf, ZSTD_MAGICNUMBER);
+ assert(src != NULL);
+ ZSTD_memcpy(hbuf, src, toCopy);
+ if ( MEM_readLE32(hbuf) != ZSTD_MAGICNUMBER ) {
+ /* not a zstd frame : let's check if it's a skippable frame */
+ MEM_writeLE32(hbuf, ZSTD_MAGIC_SKIPPABLE_START);
+ ZSTD_memcpy(hbuf, src, toCopy);
+ if ((MEM_readLE32(hbuf) & ZSTD_MAGIC_SKIPPABLE_MASK) != ZSTD_MAGIC_SKIPPABLE_START) {
+ RETURN_ERROR(prefix_unknown,
+ "first bytes don't correspond to any supported magic number");
+ } } }
+ return minInputSize;
+ }
+
+ ZSTD_memset(zfhPtr, 0, sizeof(*zfhPtr)); /* not strictly necessary, but static analyzers may not understand that zfhPtr will be read only if return value is zero, since they are 2 different signals */
+ if ( (format != ZSTD_f_zstd1_magicless)
+ && (MEM_readLE32(src) != ZSTD_MAGICNUMBER) ) {
+ if ((MEM_readLE32(src) & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) {
+ /* skippable frame */
+ if (srcSize < ZSTD_SKIPPABLEHEADERSIZE)
+ return ZSTD_SKIPPABLEHEADERSIZE; /* magic number + frame length */
+ ZSTD_memset(zfhPtr, 0, sizeof(*zfhPtr));
+ zfhPtr->frameContentSize = MEM_readLE32((const char *)src + ZSTD_FRAMEIDSIZE);
+ zfhPtr->frameType = ZSTD_skippableFrame;
+ return 0;
+ }
+ RETURN_ERROR(prefix_unknown, "");
+ }
+
+ /* ensure there is enough `srcSize` to fully read/decode frame header */
+ { size_t const fhsize = ZSTD_frameHeaderSize_internal(src, srcSize, format);
+ if (srcSize < fhsize) return fhsize;
+ zfhPtr->headerSize = (U32)fhsize;
+ }
+
+ { BYTE const fhdByte = ip[minInputSize-1];
+ size_t pos = minInputSize;
+ U32 const dictIDSizeCode = fhdByte&3;
+ U32 const checksumFlag = (fhdByte>>2)&1;
+ U32 const singleSegment = (fhdByte>>5)&1;
+ U32 const fcsID = fhdByte>>6;
+ U64 windowSize = 0;
+ U32 dictID = 0;
+ U64 frameContentSize = ZSTD_CONTENTSIZE_UNKNOWN;
+ RETURN_ERROR_IF((fhdByte & 0x08) != 0, frameParameter_unsupported,
+ "reserved bits, must be zero");
+
+ if (!singleSegment) {
+ BYTE const wlByte = ip[pos++];
+ U32 const windowLog = (wlByte >> 3) + ZSTD_WINDOWLOG_ABSOLUTEMIN;
+ RETURN_ERROR_IF(windowLog > ZSTD_WINDOWLOG_MAX, frameParameter_windowTooLarge, "");
+ windowSize = (1ULL << windowLog);
+ windowSize += (windowSize >> 3) * (wlByte&7);
+ }
+ switch(dictIDSizeCode)
+ {
+ default:
+ assert(0); /* impossible */
+ ZSTD_FALLTHROUGH;
+ case 0 : break;
+ case 1 : dictID = ip[pos]; pos++; break;
+ case 2 : dictID = MEM_readLE16(ip+pos); pos+=2; break;
+ case 3 : dictID = MEM_readLE32(ip+pos); pos+=4; break;
+ }
+ switch(fcsID)
+ {
+ default:
+ assert(0); /* impossible */
+ ZSTD_FALLTHROUGH;
+ case 0 : if (singleSegment) frameContentSize = ip[pos]; break;
+ case 1 : frameContentSize = MEM_readLE16(ip+pos)+256; break;
+ case 2 : frameContentSize = MEM_readLE32(ip+pos); break;
+ case 3 : frameContentSize = MEM_readLE64(ip+pos); break;
+ }
+ if (singleSegment) windowSize = frameContentSize;
+
+ zfhPtr->frameType = ZSTD_frame;
+ zfhPtr->frameContentSize = frameContentSize;
+ zfhPtr->windowSize = windowSize;
+ zfhPtr->blockSizeMax = (unsigned) MIN(windowSize, ZSTD_BLOCKSIZE_MAX);
+ zfhPtr->dictID = dictID;
+ zfhPtr->checksumFlag = checksumFlag;
+ }
+ return 0;
+}
+
+/** ZSTD_getFrameHeader() :
+ * decode Frame Header, or require larger `srcSize`.
+ * note : this function does not consume input, it only reads it.
+ * @return : 0, `zfhPtr` is correctly filled,
+ * >0, `srcSize` is too small, value is wanted `srcSize` amount,
+ * or an error code, which can be tested using ZSTD_isError() */
+size_t ZSTD_getFrameHeader(ZSTD_frameHeader* zfhPtr, const void* src, size_t srcSize)
+{
+ return ZSTD_getFrameHeader_advanced(zfhPtr, src, srcSize, ZSTD_f_zstd1);
+}
+
+/** ZSTD_getFrameContentSize() :
+ * compatible with legacy mode
+ * @return : decompressed size of the single frame pointed to be `src` if known, otherwise
+ * - ZSTD_CONTENTSIZE_UNKNOWN if the size cannot be determined
+ * - ZSTD_CONTENTSIZE_ERROR if an error occurred (e.g. invalid magic number, srcSize too small) */
+unsigned long long ZSTD_getFrameContentSize(const void *src, size_t srcSize)
+{
+#if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT >= 1)
+ if (ZSTD_isLegacy(src, srcSize)) {
+ unsigned long long const ret = ZSTD_getDecompressedSize_legacy(src, srcSize);
+ return ret == 0 ? ZSTD_CONTENTSIZE_UNKNOWN : ret;
+ }
+#endif
+ { ZSTD_frameHeader zfh;
+ if (ZSTD_getFrameHeader(&zfh, src, srcSize) != 0)
+ return ZSTD_CONTENTSIZE_ERROR;
+ if (zfh.frameType == ZSTD_skippableFrame) {
+ return 0;
+ } else {
+ return zfh.frameContentSize;
+ } }
+}
+
+static size_t readSkippableFrameSize(void const* src, size_t srcSize)
+{
+ size_t const skippableHeaderSize = ZSTD_SKIPPABLEHEADERSIZE;
+ U32 sizeU32;
+
+ RETURN_ERROR_IF(srcSize < ZSTD_SKIPPABLEHEADERSIZE, srcSize_wrong, "");
+
+ sizeU32 = MEM_readLE32((BYTE const*)src + ZSTD_FRAMEIDSIZE);
+ RETURN_ERROR_IF((U32)(sizeU32 + ZSTD_SKIPPABLEHEADERSIZE) < sizeU32,
+ frameParameter_unsupported, "");
+ { size_t const skippableSize = skippableHeaderSize + sizeU32;
+ RETURN_ERROR_IF(skippableSize > srcSize, srcSize_wrong, "");
+ return skippableSize;
+ }
+}
+
+/*! ZSTD_readSkippableFrame() :
+ * Retrieves content of a skippable frame, and writes it to dst buffer.
+ *
+ * The parameter magicVariant will receive the magicVariant that was supplied when the frame was written,
+ * i.e. magicNumber - ZSTD_MAGIC_SKIPPABLE_START. This can be NULL if the caller is not interested
+ * in the magicVariant.
+ *
+ * Returns an error if destination buffer is not large enough, or if this is not a valid skippable frame.
+ *
+ * @return : number of bytes written or a ZSTD error.
+ */
+size_t ZSTD_readSkippableFrame(void* dst, size_t dstCapacity,
+ unsigned* magicVariant, /* optional, can be NULL */
+ const void* src, size_t srcSize)
+{
+ RETURN_ERROR_IF(srcSize < ZSTD_SKIPPABLEHEADERSIZE, srcSize_wrong, "");
+
+ { U32 const magicNumber = MEM_readLE32(src);
+ size_t skippableFrameSize = readSkippableFrameSize(src, srcSize);
+ size_t skippableContentSize = skippableFrameSize - ZSTD_SKIPPABLEHEADERSIZE;
+
+ /* check input validity */
+ RETURN_ERROR_IF(!ZSTD_isSkippableFrame(src, srcSize), frameParameter_unsupported, "");
+ RETURN_ERROR_IF(skippableFrameSize < ZSTD_SKIPPABLEHEADERSIZE || skippableFrameSize > srcSize, srcSize_wrong, "");
+ RETURN_ERROR_IF(skippableContentSize > dstCapacity, dstSize_tooSmall, "");
+
+ /* deliver payload */
+ if (skippableContentSize > 0 && dst != NULL)
+ ZSTD_memcpy(dst, (const BYTE *)src + ZSTD_SKIPPABLEHEADERSIZE, skippableContentSize);
+ if (magicVariant != NULL)
+ *magicVariant = magicNumber - ZSTD_MAGIC_SKIPPABLE_START;
+ return skippableContentSize;
+ }
+}
+
+/** ZSTD_findDecompressedSize() :
+ * `srcSize` must be the exact length of some number of ZSTD compressed and/or
+ * skippable frames
+ * note: compatible with legacy mode
+ * @return : decompressed size of the frames contained */
+unsigned long long ZSTD_findDecompressedSize(const void* src, size_t srcSize)
+{
+ unsigned long long totalDstSize = 0;
+
+ while (srcSize >= ZSTD_startingInputLength(ZSTD_f_zstd1)) {
+ U32 const magicNumber = MEM_readLE32(src);
+
+ if ((magicNumber & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) {
+ size_t const skippableSize = readSkippableFrameSize(src, srcSize);
+ if (ZSTD_isError(skippableSize)) return ZSTD_CONTENTSIZE_ERROR;
+ assert(skippableSize <= srcSize);
+
+ src = (const BYTE *)src + skippableSize;
+ srcSize -= skippableSize;
+ continue;
+ }
+
+ { unsigned long long const fcs = ZSTD_getFrameContentSize(src, srcSize);
+ if (fcs >= ZSTD_CONTENTSIZE_ERROR) return fcs;
+
+ if (totalDstSize + fcs < totalDstSize)
+ return ZSTD_CONTENTSIZE_ERROR; /* check for overflow */
+ totalDstSize += fcs;
+ }
+ /* skip to next frame */
+ { size_t const frameSrcSize = ZSTD_findFrameCompressedSize(src, srcSize);
+ if (ZSTD_isError(frameSrcSize)) return ZSTD_CONTENTSIZE_ERROR;
+ assert(frameSrcSize <= srcSize);
+
+ src = (const BYTE *)src + frameSrcSize;
+ srcSize -= frameSrcSize;
+ }
+ } /* while (srcSize >= ZSTD_frameHeaderSize_prefix) */
+
+ if (srcSize) return ZSTD_CONTENTSIZE_ERROR;
+
+ return totalDstSize;
+}
+
+/** ZSTD_getDecompressedSize() :
+ * compatible with legacy mode
+ * @return : decompressed size if known, 0 otherwise
+ note : 0 can mean any of the following :
+ - frame content is empty
+ - decompressed size field is not present in frame header
+ - frame header unknown / not supported
+ - frame header not complete (`srcSize` too small) */
+unsigned long long ZSTD_getDecompressedSize(const void* src, size_t srcSize)
+{
+ unsigned long long const ret = ZSTD_getFrameContentSize(src, srcSize);
+ ZSTD_STATIC_ASSERT(ZSTD_CONTENTSIZE_ERROR < ZSTD_CONTENTSIZE_UNKNOWN);
+ return (ret >= ZSTD_CONTENTSIZE_ERROR) ? 0 : ret;
+}
+
+
+/** ZSTD_decodeFrameHeader() :
+ * `headerSize` must be the size provided by ZSTD_frameHeaderSize().
+ * If multiple DDict references are enabled, also will choose the correct DDict to use.
+ * @return : 0 if success, or an error code, which can be tested using ZSTD_isError() */
+static size_t ZSTD_decodeFrameHeader(ZSTD_DCtx* dctx, const void* src, size_t headerSize)
+{
+ size_t const result = ZSTD_getFrameHeader_advanced(&(dctx->fParams), src, headerSize, dctx->format);
+ if (ZSTD_isError(result)) return result; /* invalid header */
+ RETURN_ERROR_IF(result>0, srcSize_wrong, "headerSize too small");
+
+ /* Reference DDict requested by frame if dctx references multiple ddicts */
+ if (dctx->refMultipleDDicts == ZSTD_rmd_refMultipleDDicts && dctx->ddictSet) {
+ ZSTD_DCtx_selectFrameDDict(dctx);
+ }
+
+#ifndef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
+ /* Skip the dictID check in fuzzing mode, because it makes the search
+ * harder.
+ */
+ RETURN_ERROR_IF(dctx->fParams.dictID && (dctx->dictID != dctx->fParams.dictID),
+ dictionary_wrong, "");
+#endif
+ dctx->validateChecksum = (dctx->fParams.checksumFlag && !dctx->forceIgnoreChecksum) ? 1 : 0;
+ if (dctx->validateChecksum) XXH64_reset(&dctx->xxhState, 0);
+ dctx->processedCSize += headerSize;
+ return 0;
+}
+
+static ZSTD_frameSizeInfo ZSTD_errorFrameSizeInfo(size_t ret)
+{
+ ZSTD_frameSizeInfo frameSizeInfo;
+ frameSizeInfo.compressedSize = ret;
+ frameSizeInfo.decompressedBound = ZSTD_CONTENTSIZE_ERROR;
+ return frameSizeInfo;
+}
+
+static ZSTD_frameSizeInfo ZSTD_findFrameSizeInfo(const void* src, size_t srcSize, ZSTD_format_e format)
+{
+ ZSTD_frameSizeInfo frameSizeInfo;
+ ZSTD_memset(&frameSizeInfo, 0, sizeof(ZSTD_frameSizeInfo));
+
+#if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT >= 1)
+ if (format == ZSTD_f_zstd1 && ZSTD_isLegacy(src, srcSize))
+ return ZSTD_findFrameSizeInfoLegacy(src, srcSize);
+#endif
+
+ if (format == ZSTD_f_zstd1 && (srcSize >= ZSTD_SKIPPABLEHEADERSIZE)
+ && (MEM_readLE32(src) & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) {
+ frameSizeInfo.compressedSize = readSkippableFrameSize(src, srcSize);
+ assert(ZSTD_isError(frameSizeInfo.compressedSize) ||
+ frameSizeInfo.compressedSize <= srcSize);
+ return frameSizeInfo;
+ } else {
+ const BYTE* ip = (const BYTE*)src;
+ const BYTE* const ipstart = ip;
+ size_t remainingSize = srcSize;
+ size_t nbBlocks = 0;
+ ZSTD_frameHeader zfh;
+
+ /* Extract Frame Header */
+ { size_t const ret = ZSTD_getFrameHeader_advanced(&zfh, src, srcSize, format);
+ if (ZSTD_isError(ret))
+ return ZSTD_errorFrameSizeInfo(ret);
+ if (ret > 0)
+ return ZSTD_errorFrameSizeInfo(ERROR(srcSize_wrong));
+ }
+
+ ip += zfh.headerSize;
+ remainingSize -= zfh.headerSize;
+
+ /* Iterate over each block */
+ while (1) {
+ blockProperties_t blockProperties;
+ size_t const cBlockSize = ZSTD_getcBlockSize(ip, remainingSize, &blockProperties);
+ if (ZSTD_isError(cBlockSize))
+ return ZSTD_errorFrameSizeInfo(cBlockSize);
+
+ if (ZSTD_blockHeaderSize + cBlockSize > remainingSize)
+ return ZSTD_errorFrameSizeInfo(ERROR(srcSize_wrong));
+
+ ip += ZSTD_blockHeaderSize + cBlockSize;
+ remainingSize -= ZSTD_blockHeaderSize + cBlockSize;
+ nbBlocks++;
+
+ if (blockProperties.lastBlock) break;
+ }
+
+ /* Final frame content checksum */
+ if (zfh.checksumFlag) {
+ if (remainingSize < 4)
+ return ZSTD_errorFrameSizeInfo(ERROR(srcSize_wrong));
+ ip += 4;
+ }
+
+ frameSizeInfo.nbBlocks = nbBlocks;
+ frameSizeInfo.compressedSize = (size_t)(ip - ipstart);
+ frameSizeInfo.decompressedBound = (zfh.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN)
+ ? zfh.frameContentSize
+ : (unsigned long long)nbBlocks * zfh.blockSizeMax;
+ return frameSizeInfo;
+ }
+}
+
+static size_t ZSTD_findFrameCompressedSize_advanced(const void *src, size_t srcSize, ZSTD_format_e format) {
+ ZSTD_frameSizeInfo const frameSizeInfo = ZSTD_findFrameSizeInfo(src, srcSize, format);
+ return frameSizeInfo.compressedSize;
+}
+
+/** ZSTD_findFrameCompressedSize() :
+ * See docs in zstd.h
+ * Note: compatible with legacy mode */
+size_t ZSTD_findFrameCompressedSize(const void *src, size_t srcSize)
+{
+ return ZSTD_findFrameCompressedSize_advanced(src, srcSize, ZSTD_f_zstd1);
+}
+
+/** ZSTD_decompressBound() :
+ * compatible with legacy mode
+ * `src` must point to the start of a ZSTD frame or a skippeable frame
+ * `srcSize` must be at least as large as the frame contained
+ * @return : the maximum decompressed size of the compressed source
+ */
+unsigned long long ZSTD_decompressBound(const void* src, size_t srcSize)
+{
+ unsigned long long bound = 0;
+ /* Iterate over each frame */
+ while (srcSize > 0) {
+ ZSTD_frameSizeInfo const frameSizeInfo = ZSTD_findFrameSizeInfo(src, srcSize, ZSTD_f_zstd1);
+ size_t const compressedSize = frameSizeInfo.compressedSize;
+ unsigned long long const decompressedBound = frameSizeInfo.decompressedBound;
+ if (ZSTD_isError(compressedSize) || decompressedBound == ZSTD_CONTENTSIZE_ERROR)
+ return ZSTD_CONTENTSIZE_ERROR;
+ assert(srcSize >= compressedSize);
+ src = (const BYTE*)src + compressedSize;
+ srcSize -= compressedSize;
+ bound += decompressedBound;
+ }
+ return bound;
+}
+
+size_t ZSTD_decompressionMargin(void const* src, size_t srcSize)
+{
+ size_t margin = 0;
+ unsigned maxBlockSize = 0;
+
+ /* Iterate over each frame */
+ while (srcSize > 0) {
+ ZSTD_frameSizeInfo const frameSizeInfo = ZSTD_findFrameSizeInfo(src, srcSize, ZSTD_f_zstd1);
+ size_t const compressedSize = frameSizeInfo.compressedSize;
+ unsigned long long const decompressedBound = frameSizeInfo.decompressedBound;
+ ZSTD_frameHeader zfh;
+
+ FORWARD_IF_ERROR(ZSTD_getFrameHeader(&zfh, src, srcSize), "");
+ if (ZSTD_isError(compressedSize) || decompressedBound == ZSTD_CONTENTSIZE_ERROR)
+ return ERROR(corruption_detected);
+
+ if (zfh.frameType == ZSTD_frame) {
+ /* Add the frame header to our margin */
+ margin += zfh.headerSize;
+ /* Add the checksum to our margin */
+ margin += zfh.checksumFlag ? 4 : 0;
+ /* Add 3 bytes per block */
+ margin += 3 * frameSizeInfo.nbBlocks;
+
+ /* Compute the max block size */
+ maxBlockSize = MAX(maxBlockSize, zfh.blockSizeMax);
+ } else {
+ assert(zfh.frameType == ZSTD_skippableFrame);
+ /* Add the entire skippable frame size to our margin. */
+ margin += compressedSize;
+ }
+
+ assert(srcSize >= compressedSize);
+ src = (const BYTE*)src + compressedSize;
+ srcSize -= compressedSize;
+ }
+
+ /* Add the max block size back to the margin. */
+ margin += maxBlockSize;
+
+ return margin;
+}
+
+/*-*************************************************************
+ * Frame decoding
+ ***************************************************************/
+
+/** ZSTD_insertBlock() :
+ * insert `src` block into `dctx` history. Useful to track uncompressed blocks. */
+size_t ZSTD_insertBlock(ZSTD_DCtx* dctx, const void* blockStart, size_t blockSize)
+{
+ DEBUGLOG(5, "ZSTD_insertBlock: %u bytes", (unsigned)blockSize);
+ ZSTD_checkContinuity(dctx, blockStart, blockSize);
+ dctx->previousDstEnd = (const char*)blockStart + blockSize;
+ return blockSize;
+}
+
+
+static size_t ZSTD_copyRawBlock(void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize)
+{
+ DEBUGLOG(5, "ZSTD_copyRawBlock");
+ RETURN_ERROR_IF(srcSize > dstCapacity, dstSize_tooSmall, "");
+ if (dst == NULL) {
+ if (srcSize == 0) return 0;
+ RETURN_ERROR(dstBuffer_null, "");
+ }
+ ZSTD_memmove(dst, src, srcSize);
+ return srcSize;
+}
+
+static size_t ZSTD_setRleBlock(void* dst, size_t dstCapacity,
+ BYTE b,
+ size_t regenSize)
+{
+ RETURN_ERROR_IF(regenSize > dstCapacity, dstSize_tooSmall, "");
+ if (dst == NULL) {
+ if (regenSize == 0) return 0;
+ RETURN_ERROR(dstBuffer_null, "");
+ }
+ ZSTD_memset(dst, b, regenSize);
+ return regenSize;
+}
+
+static void ZSTD_DCtx_trace_end(ZSTD_DCtx const* dctx, U64 uncompressedSize, U64 compressedSize, unsigned streaming)
+{
+#if ZSTD_TRACE
+ if (dctx->traceCtx && ZSTD_trace_decompress_end != NULL) {
+ ZSTD_Trace trace;
+ ZSTD_memset(&trace, 0, sizeof(trace));
+ trace.version = ZSTD_VERSION_NUMBER;
+ trace.streaming = streaming;
+ if (dctx->ddict) {
+ trace.dictionaryID = ZSTD_getDictID_fromDDict(dctx->ddict);
+ trace.dictionarySize = ZSTD_DDict_dictSize(dctx->ddict);
+ trace.dictionaryIsCold = dctx->ddictIsCold;
+ }
+ trace.uncompressedSize = (size_t)uncompressedSize;
+ trace.compressedSize = (size_t)compressedSize;
+ trace.dctx = dctx;
+ ZSTD_trace_decompress_end(dctx->traceCtx, &trace);
+ }
+#else
+ (void)dctx;
+ (void)uncompressedSize;
+ (void)compressedSize;
+ (void)streaming;
+#endif
+}
+
+
+/*! ZSTD_decompressFrame() :
+ * @dctx must be properly initialized
+ * will update *srcPtr and *srcSizePtr,
+ * to make *srcPtr progress by one frame. */
+static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx,
+ void* dst, size_t dstCapacity,
+ const void** srcPtr, size_t *srcSizePtr)
+{
+ const BYTE* const istart = (const BYTE*)(*srcPtr);
+ const BYTE* ip = istart;
+ BYTE* const ostart = (BYTE*)dst;
+ BYTE* const oend = dstCapacity != 0 ? ostart + dstCapacity : ostart;
+ BYTE* op = ostart;
+ size_t remainingSrcSize = *srcSizePtr;
+
+ DEBUGLOG(4, "ZSTD_decompressFrame (srcSize:%i)", (int)*srcSizePtr);
+
+ /* check */
+ RETURN_ERROR_IF(
+ remainingSrcSize < ZSTD_FRAMEHEADERSIZE_MIN(dctx->format)+ZSTD_blockHeaderSize,
+ srcSize_wrong, "");
+
+ /* Frame Header */
+ { size_t const frameHeaderSize = ZSTD_frameHeaderSize_internal(
+ ip, ZSTD_FRAMEHEADERSIZE_PREFIX(dctx->format), dctx->format);
+ if (ZSTD_isError(frameHeaderSize)) return frameHeaderSize;
+ RETURN_ERROR_IF(remainingSrcSize < frameHeaderSize+ZSTD_blockHeaderSize,
+ srcSize_wrong, "");
+ FORWARD_IF_ERROR( ZSTD_decodeFrameHeader(dctx, ip, frameHeaderSize) , "");
+ ip += frameHeaderSize; remainingSrcSize -= frameHeaderSize;
+ }
+
+ /* Shrink the blockSizeMax if enabled */
+ if (dctx->maxBlockSizeParam != 0)
+ dctx->fParams.blockSizeMax = MIN(dctx->fParams.blockSizeMax, (unsigned)dctx->maxBlockSizeParam);
+
+ /* Loop on each block */
+ while (1) {
+ BYTE* oBlockEnd = oend;
+ size_t decodedSize;
+ blockProperties_t blockProperties;
+ size_t const cBlockSize = ZSTD_getcBlockSize(ip, remainingSrcSize, &blockProperties);
+ if (ZSTD_isError(cBlockSize)) return cBlockSize;
+
+ ip += ZSTD_blockHeaderSize;
+ remainingSrcSize -= ZSTD_blockHeaderSize;
+ RETURN_ERROR_IF(cBlockSize > remainingSrcSize, srcSize_wrong, "");
+
+ if (ip >= op && ip < oBlockEnd) {
+ /* We are decompressing in-place. Limit the output pointer so that we
+ * don't overwrite the block that we are currently reading. This will
+ * fail decompression if the input & output pointers aren't spaced
+ * far enough apart.
+ *
+ * This is important to set, even when the pointers are far enough
+ * apart, because ZSTD_decompressBlock_internal() can decide to store
+ * literals in the output buffer, after the block it is decompressing.
+ * Since we don't want anything to overwrite our input, we have to tell
+ * ZSTD_decompressBlock_internal to never write past ip.
+ *
+ * See ZSTD_allocateLiteralsBuffer() for reference.
+ */
+ oBlockEnd = op + (ip - op);
+ }
+
+ switch(blockProperties.blockType)
+ {
+ case bt_compressed:
+ assert(dctx->isFrameDecompression == 1);
+ decodedSize = ZSTD_decompressBlock_internal(dctx, op, (size_t)(oBlockEnd-op), ip, cBlockSize, not_streaming);
+ break;
+ case bt_raw :
+ /* Use oend instead of oBlockEnd because this function is safe to overlap. It uses memmove. */
+ decodedSize = ZSTD_copyRawBlock(op, (size_t)(oend-op), ip, cBlockSize);
+ break;
+ case bt_rle :
+ decodedSize = ZSTD_setRleBlock(op, (size_t)(oBlockEnd-op), *ip, blockProperties.origSize);
+ break;
+ case bt_reserved :
+ default:
+ RETURN_ERROR(corruption_detected, "invalid block type");
+ }
+ FORWARD_IF_ERROR(decodedSize, "Block decompression failure");
+ DEBUGLOG(5, "Decompressed block of dSize = %u", (unsigned)decodedSize);
+ if (dctx->validateChecksum) {
+ XXH64_update(&dctx->xxhState, op, decodedSize);
+ }
+ if (decodedSize) /* support dst = NULL,0 */ {
+ op += decodedSize;
+ }
+ assert(ip != NULL);
+ ip += cBlockSize;
+ remainingSrcSize -= cBlockSize;
+ if (blockProperties.lastBlock) break;
+ }
+
+ if (dctx->fParams.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN) {
+ RETURN_ERROR_IF((U64)(op-ostart) != dctx->fParams.frameContentSize,
+ corruption_detected, "");
+ }
+ if (dctx->fParams.checksumFlag) { /* Frame content checksum verification */
+ RETURN_ERROR_IF(remainingSrcSize<4, checksum_wrong, "");
+ if (!dctx->forceIgnoreChecksum) {
+ U32 const checkCalc = (U32)XXH64_digest(&dctx->xxhState);
+ U32 checkRead;
+ checkRead = MEM_readLE32(ip);
+ RETURN_ERROR_IF(checkRead != checkCalc, checksum_wrong, "");
+ }
+ ip += 4;
+ remainingSrcSize -= 4;
+ }
+ ZSTD_DCtx_trace_end(dctx, (U64)(op-ostart), (U64)(ip-istart), /* streaming */ 0);
+ /* Allow caller to get size read */
+ DEBUGLOG(4, "ZSTD_decompressFrame: decompressed frame of size %zi, consuming %zi bytes of input", op-ostart, ip - (const BYTE*)*srcPtr);
+ *srcPtr = ip;
+ *srcSizePtr = remainingSrcSize;
+ return (size_t)(op-ostart);
+}
+
+static
+ZSTD_ALLOW_POINTER_OVERFLOW_ATTR
+size_t 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)
+{
+ void* const dststart = dst;
+ int moreThan1Frame = 0;
+
+ DEBUGLOG(5, "ZSTD_decompressMultiFrame");
+ assert(dict==NULL || ddict==NULL); /* either dict or ddict set, not both */
+
+ if (ddict) {
+ dict = ZSTD_DDict_dictContent(ddict);
+ dictSize = ZSTD_DDict_dictSize(ddict);
+ }
+
+ while (srcSize >= ZSTD_startingInputLength(dctx->format)) {
+
+#if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT >= 1)
+ if (dctx->format == ZSTD_f_zstd1 && ZSTD_isLegacy(src, srcSize)) {
+ size_t decodedSize;
+ size_t const frameSize = ZSTD_findFrameCompressedSizeLegacy(src, srcSize);
+ if (ZSTD_isError(frameSize)) return frameSize;
+ RETURN_ERROR_IF(dctx->staticSize, memory_allocation,
+ "legacy support is not compatible with static dctx");
+
+ decodedSize = ZSTD_decompressLegacy(dst, dstCapacity, src, frameSize, dict, dictSize);
+ if (ZSTD_isError(decodedSize)) return decodedSize;
+
+ {
+ unsigned long long const expectedSize = ZSTD_getFrameContentSize(src, srcSize);
+ RETURN_ERROR_IF(expectedSize == ZSTD_CONTENTSIZE_ERROR, corruption_detected, "Corrupted frame header!");
+ if (expectedSize != ZSTD_CONTENTSIZE_UNKNOWN) {
+ RETURN_ERROR_IF(expectedSize != decodedSize, corruption_detected,
+ "Frame header size does not match decoded size!");
+ }
+ }
+
+ assert(decodedSize <= dstCapacity);
+ dst = (BYTE*)dst + decodedSize;
+ dstCapacity -= decodedSize;
+
+ src = (const BYTE*)src + frameSize;
+ srcSize -= frameSize;
+
+ continue;
+ }
+#endif
+
+ if (dctx->format == ZSTD_f_zstd1 && srcSize >= 4) {
+ U32 const magicNumber = MEM_readLE32(src);
+ DEBUGLOG(5, "reading magic number %08X", (unsigned)magicNumber);
+ if ((magicNumber & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) {
+ /* skippable frame detected : skip it */
+ size_t const skippableSize = readSkippableFrameSize(src, srcSize);
+ FORWARD_IF_ERROR(skippableSize, "invalid skippable frame");
+ assert(skippableSize <= srcSize);
+
+ src = (const BYTE *)src + skippableSize;
+ srcSize -= skippableSize;
+ continue; /* check next frame */
+ } }
+
+ if (ddict) {
+ /* we were called from ZSTD_decompress_usingDDict */
+ FORWARD_IF_ERROR(ZSTD_decompressBegin_usingDDict(dctx, ddict), "");
+ } else {
+ /* this will initialize correctly with no dict if dict == NULL, so
+ * use this in all cases but ddict */
+ FORWARD_IF_ERROR(ZSTD_decompressBegin_usingDict(dctx, dict, dictSize), "");
+ }
+ ZSTD_checkContinuity(dctx, dst, dstCapacity);
+
+ { const size_t res = ZSTD_decompressFrame(dctx, dst, dstCapacity,
+ &src, &srcSize);
+ RETURN_ERROR_IF(
+ (ZSTD_getErrorCode(res) == ZSTD_error_prefix_unknown)
+ && (moreThan1Frame==1),
+ srcSize_wrong,
+ "At least one frame successfully completed, "
+ "but following bytes are garbage: "
+ "it's more likely to be a srcSize error, "
+ "specifying more input bytes than size of frame(s). "
+ "Note: one could be unlucky, it might be a corruption error instead, "
+ "happening right at the place where we expect zstd magic bytes. "
+ "But this is _much_ less likely than a srcSize field error.");
+ if (ZSTD_isError(res)) return res;
+ assert(res <= dstCapacity);
+ if (res != 0)
+ dst = (BYTE*)dst + res;
+ dstCapacity -= res;
+ }
+ moreThan1Frame = 1;
+ } /* while (srcSize >= ZSTD_frameHeaderSize_prefix) */
+
+ RETURN_ERROR_IF(srcSize, srcSize_wrong, "input not entirely consumed");
+
+ return (size_t)((BYTE*)dst - (BYTE*)dststart);
+}
+
+size_t ZSTD_decompress_usingDict(ZSTD_DCtx* dctx,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize,
+ const void* dict, size_t dictSize)
+{
+ return ZSTD_decompressMultiFrame(dctx, dst, dstCapacity, src, srcSize, dict, dictSize, NULL);
+}
+
+
+static ZSTD_DDict const* ZSTD_getDDict(ZSTD_DCtx* dctx)
+{
+ switch (dctx->dictUses) {
+ default:
+ assert(0 /* Impossible */);
+ ZSTD_FALLTHROUGH;
+ case ZSTD_dont_use:
+ ZSTD_clearDict(dctx);
+ return NULL;
+ case ZSTD_use_indefinitely:
+ return dctx->ddict;
+ case ZSTD_use_once:
+ dctx->dictUses = ZSTD_dont_use;
+ return dctx->ddict;
+ }
+}
+
+size_t ZSTD_decompressDCtx(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize)
+{
+ return ZSTD_decompress_usingDDict(dctx, dst, dstCapacity, src, srcSize, ZSTD_getDDict(dctx));
+}
+
+
+size_t ZSTD_decompress(void* dst, size_t dstCapacity, const void* src, size_t srcSize)
+{
+#if defined(ZSTD_HEAPMODE) && (ZSTD_HEAPMODE>=1)
+ size_t regenSize;
+ ZSTD_DCtx* const dctx = ZSTD_createDCtx_internal(ZSTD_defaultCMem);
+ RETURN_ERROR_IF(dctx==NULL, memory_allocation, "NULL pointer!");
+ regenSize = ZSTD_decompressDCtx(dctx, dst, dstCapacity, src, srcSize);
+ ZSTD_freeDCtx(dctx);
+ return regenSize;
+#else /* stack mode */
+ ZSTD_DCtx dctx;
+ ZSTD_initDCtx_internal(&dctx);
+ return ZSTD_decompressDCtx(&dctx, dst, dstCapacity, src, srcSize);
+#endif
+}
+
+
+/*-**************************************
+* Advanced Streaming Decompression API
+* Bufferless and synchronous
+****************************************/
+size_t ZSTD_nextSrcSizeToDecompress(ZSTD_DCtx* dctx) { return dctx->expected; }
+
+/**
+ * Similar to ZSTD_nextSrcSizeToDecompress(), but when a block input can be streamed, we
+ * allow taking a partial block as the input. Currently only raw uncompressed blocks can
+ * be streamed.
+ *
+ * For blocks that can be streamed, this allows us to reduce the latency until we produce
+ * output, and avoid copying the input.
+ *
+ * @param inputSize - The total amount of input that the caller currently has.
+ */
+static size_t ZSTD_nextSrcSizeToDecompressWithInputSize(ZSTD_DCtx* dctx, size_t inputSize) {
+ if (!(dctx->stage == ZSTDds_decompressBlock || dctx->stage == ZSTDds_decompressLastBlock))
+ return dctx->expected;
+ if (dctx->bType != bt_raw)
+ return dctx->expected;
+ return BOUNDED(1, inputSize, dctx->expected);
+}
+
+ZSTD_nextInputType_e ZSTD_nextInputType(ZSTD_DCtx* dctx) {
+ switch(dctx->stage)
+ {
+ default: /* should not happen */
+ assert(0);
+ ZSTD_FALLTHROUGH;
+ case ZSTDds_getFrameHeaderSize:
+ ZSTD_FALLTHROUGH;
+ case ZSTDds_decodeFrameHeader:
+ return ZSTDnit_frameHeader;
+ case ZSTDds_decodeBlockHeader:
+ return ZSTDnit_blockHeader;
+ case ZSTDds_decompressBlock:
+ return ZSTDnit_block;
+ case ZSTDds_decompressLastBlock:
+ return ZSTDnit_lastBlock;
+ case ZSTDds_checkChecksum:
+ return ZSTDnit_checksum;
+ case ZSTDds_decodeSkippableHeader:
+ ZSTD_FALLTHROUGH;
+ case ZSTDds_skipFrame:
+ return ZSTDnit_skippableFrame;
+ }
+}
+
+static int ZSTD_isSkipFrame(ZSTD_DCtx* dctx) { return dctx->stage == ZSTDds_skipFrame; }
+
+/** ZSTD_decompressContinue() :
+ * srcSize : must be the exact nb of bytes expected (see ZSTD_nextSrcSizeToDecompress())
+ * @return : nb of bytes generated into `dst` (necessarily <= `dstCapacity)
+ * or an error code, which can be tested using ZSTD_isError() */
+size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize)
+{
+ DEBUGLOG(5, "ZSTD_decompressContinue (srcSize:%u)", (unsigned)srcSize);
+ /* Sanity check */
+ RETURN_ERROR_IF(srcSize != ZSTD_nextSrcSizeToDecompressWithInputSize(dctx, srcSize), srcSize_wrong, "not allowed");
+ ZSTD_checkContinuity(dctx, dst, dstCapacity);
+
+ dctx->processedCSize += srcSize;
+
+ switch (dctx->stage)
+ {
+ case ZSTDds_getFrameHeaderSize :
+ assert(src != NULL);
+ if (dctx->format == ZSTD_f_zstd1) { /* allows header */
+ assert(srcSize >= ZSTD_FRAMEIDSIZE); /* to read skippable magic number */
+ if ((MEM_readLE32(src) & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) { /* skippable frame */
+ ZSTD_memcpy(dctx->headerBuffer, src, srcSize);
+ dctx->expected = ZSTD_SKIPPABLEHEADERSIZE - srcSize; /* remaining to load to get full skippable frame header */
+ dctx->stage = ZSTDds_decodeSkippableHeader;
+ return 0;
+ } }
+ dctx->headerSize = ZSTD_frameHeaderSize_internal(src, srcSize, dctx->format);
+ if (ZSTD_isError(dctx->headerSize)) return dctx->headerSize;
+ ZSTD_memcpy(dctx->headerBuffer, src, srcSize);
+ dctx->expected = dctx->headerSize - srcSize;
+ dctx->stage = ZSTDds_decodeFrameHeader;
+ return 0;
+
+ case ZSTDds_decodeFrameHeader:
+ assert(src != NULL);
+ ZSTD_memcpy(dctx->headerBuffer + (dctx->headerSize - srcSize), src, srcSize);
+ FORWARD_IF_ERROR(ZSTD_decodeFrameHeader(dctx, dctx->headerBuffer, dctx->headerSize), "");
+ dctx->expected = ZSTD_blockHeaderSize;
+ dctx->stage = ZSTDds_decodeBlockHeader;
+ return 0;
+
+ case ZSTDds_decodeBlockHeader:
+ { blockProperties_t bp;
+ size_t const cBlockSize = ZSTD_getcBlockSize(src, ZSTD_blockHeaderSize, &bp);
+ if (ZSTD_isError(cBlockSize)) return cBlockSize;
+ RETURN_ERROR_IF(cBlockSize > dctx->fParams.blockSizeMax, corruption_detected, "Block Size Exceeds Maximum");
+ dctx->expected = cBlockSize;
+ dctx->bType = bp.blockType;
+ dctx->rleSize = bp.origSize;
+ if (cBlockSize) {
+ dctx->stage = bp.lastBlock ? ZSTDds_decompressLastBlock : ZSTDds_decompressBlock;
+ return 0;
+ }
+ /* empty block */
+ if (bp.lastBlock) {
+ if (dctx->fParams.checksumFlag) {
+ dctx->expected = 4;
+ dctx->stage = ZSTDds_checkChecksum;
+ } else {
+ dctx->expected = 0; /* end of frame */
+ dctx->stage = ZSTDds_getFrameHeaderSize;
+ }
+ } else {
+ dctx->expected = ZSTD_blockHeaderSize; /* jump to next header */
+ dctx->stage = ZSTDds_decodeBlockHeader;
+ }
+ return 0;
+ }
+
+ case ZSTDds_decompressLastBlock:
+ case ZSTDds_decompressBlock:
+ DEBUGLOG(5, "ZSTD_decompressContinue: case ZSTDds_decompressBlock");
+ { size_t rSize;
+ switch(dctx->bType)
+ {
+ case bt_compressed:
+ DEBUGLOG(5, "ZSTD_decompressContinue: case bt_compressed");
+ assert(dctx->isFrameDecompression == 1);
+ rSize = ZSTD_decompressBlock_internal(dctx, dst, dstCapacity, src, srcSize, is_streaming);
+ dctx->expected = 0; /* Streaming not supported */
+ break;
+ case bt_raw :
+ assert(srcSize <= dctx->expected);
+ rSize = ZSTD_copyRawBlock(dst, dstCapacity, src, srcSize);
+ FORWARD_IF_ERROR(rSize, "ZSTD_copyRawBlock failed");
+ assert(rSize == srcSize);
+ dctx->expected -= rSize;
+ break;
+ case bt_rle :
+ rSize = ZSTD_setRleBlock(dst, dstCapacity, *(const BYTE*)src, dctx->rleSize);
+ dctx->expected = 0; /* Streaming not supported */
+ break;
+ case bt_reserved : /* should never happen */
+ default:
+ RETURN_ERROR(corruption_detected, "invalid block type");
+ }
+ FORWARD_IF_ERROR(rSize, "");
+ RETURN_ERROR_IF(rSize > dctx->fParams.blockSizeMax, corruption_detected, "Decompressed Block Size Exceeds Maximum");
+ DEBUGLOG(5, "ZSTD_decompressContinue: decoded size from block : %u", (unsigned)rSize);
+ dctx->decodedSize += rSize;
+ if (dctx->validateChecksum) XXH64_update(&dctx->xxhState, dst, rSize);
+ dctx->previousDstEnd = (char*)dst + rSize;
+
+ /* Stay on the same stage until we are finished streaming the block. */
+ if (dctx->expected > 0) {
+ return rSize;
+ }
+
+ if (dctx->stage == ZSTDds_decompressLastBlock) { /* end of frame */
+ DEBUGLOG(4, "ZSTD_decompressContinue: decoded size from frame : %u", (unsigned)dctx->decodedSize);
+ RETURN_ERROR_IF(
+ dctx->fParams.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN
+ && dctx->decodedSize != dctx->fParams.frameContentSize,
+ corruption_detected, "");
+ if (dctx->fParams.checksumFlag) { /* another round for frame checksum */
+ dctx->expected = 4;
+ dctx->stage = ZSTDds_checkChecksum;
+ } else {
+ ZSTD_DCtx_trace_end(dctx, dctx->decodedSize, dctx->processedCSize, /* streaming */ 1);
+ dctx->expected = 0; /* ends here */
+ dctx->stage = ZSTDds_getFrameHeaderSize;
+ }
+ } else {
+ dctx->stage = ZSTDds_decodeBlockHeader;
+ dctx->expected = ZSTD_blockHeaderSize;
+ }
+ return rSize;
+ }
+
+ case ZSTDds_checkChecksum:
+ assert(srcSize == 4); /* guaranteed by dctx->expected */
+ {
+ if (dctx->validateChecksum) {
+ U32 const h32 = (U32)XXH64_digest(&dctx->xxhState);
+ U32 const check32 = MEM_readLE32(src);
+ DEBUGLOG(4, "ZSTD_decompressContinue: checksum : calculated %08X :: %08X read", (unsigned)h32, (unsigned)check32);
+ RETURN_ERROR_IF(check32 != h32, checksum_wrong, "");
+ }
+ ZSTD_DCtx_trace_end(dctx, dctx->decodedSize, dctx->processedCSize, /* streaming */ 1);
+ dctx->expected = 0;
+ dctx->stage = ZSTDds_getFrameHeaderSize;
+ return 0;
+ }
+
+ case ZSTDds_decodeSkippableHeader:
+ assert(src != NULL);
+ assert(srcSize <= ZSTD_SKIPPABLEHEADERSIZE);
+ assert(dctx->format != ZSTD_f_zstd1_magicless);
+ ZSTD_memcpy(dctx->headerBuffer + (ZSTD_SKIPPABLEHEADERSIZE - srcSize), src, srcSize); /* complete skippable header */
+ dctx->expected = MEM_readLE32(dctx->headerBuffer + ZSTD_FRAMEIDSIZE); /* note : dctx->expected can grow seriously large, beyond local buffer size */
+ dctx->stage = ZSTDds_skipFrame;
+ return 0;
+
+ case ZSTDds_skipFrame:
+ dctx->expected = 0;
+ dctx->stage = ZSTDds_getFrameHeaderSize;
+ return 0;
+
+ default:
+ assert(0); /* impossible */
+ RETURN_ERROR(GENERIC, "impossible to reach"); /* some compilers require default to do something */
+ }
+}
+
+
+static size_t ZSTD_refDictContent(ZSTD_DCtx* dctx, const void* dict, size_t dictSize)
+{
+ dctx->dictEnd = dctx->previousDstEnd;
+ dctx->virtualStart = (const char*)dict - ((const char*)(dctx->previousDstEnd) - (const char*)(dctx->prefixStart));
+ dctx->prefixStart = dict;
+ dctx->previousDstEnd = (const char*)dict + dictSize;
+#ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
+ dctx->dictContentBeginForFuzzing = dctx->prefixStart;
+ dctx->dictContentEndForFuzzing = dctx->previousDstEnd;
+#endif
+ return 0;
+}
+
+/*! ZSTD_loadDEntropy() :
+ * dict : must point at beginning of a valid zstd dictionary.
+ * @return : size of entropy tables read */
+size_t
+ZSTD_loadDEntropy(ZSTD_entropyDTables_t* entropy,
+ const void* const dict, size_t const dictSize)
+{
+ const BYTE* dictPtr = (const BYTE*)dict;
+ const BYTE* const dictEnd = dictPtr + dictSize;
+
+ RETURN_ERROR_IF(dictSize <= 8, dictionary_corrupted, "dict is too small");
+ assert(MEM_readLE32(dict) == ZSTD_MAGIC_DICTIONARY); /* dict must be valid */
+ dictPtr += 8; /* skip header = magic + dictID */
+
+ ZSTD_STATIC_ASSERT(offsetof(ZSTD_entropyDTables_t, OFTable) == offsetof(ZSTD_entropyDTables_t, LLTable) + sizeof(entropy->LLTable));
+ ZSTD_STATIC_ASSERT(offsetof(ZSTD_entropyDTables_t, MLTable) == offsetof(ZSTD_entropyDTables_t, OFTable) + sizeof(entropy->OFTable));
+ ZSTD_STATIC_ASSERT(sizeof(entropy->LLTable) + sizeof(entropy->OFTable) + sizeof(entropy->MLTable) >= HUF_DECOMPRESS_WORKSPACE_SIZE);
+ { void* const workspace = &entropy->LLTable; /* use fse tables as temporary workspace; implies fse tables are grouped together */
+ size_t const workspaceSize = sizeof(entropy->LLTable) + sizeof(entropy->OFTable) + sizeof(entropy->MLTable);
+#ifdef HUF_FORCE_DECOMPRESS_X1
+ /* in minimal huffman, we always use X1 variants */
+ size_t const hSize = HUF_readDTableX1_wksp(entropy->hufTable,
+ dictPtr, dictEnd - dictPtr,
+ workspace, workspaceSize, /* flags */ 0);
+#else
+ size_t const hSize = HUF_readDTableX2_wksp(entropy->hufTable,
+ dictPtr, (size_t)(dictEnd - dictPtr),
+ workspace, workspaceSize, /* flags */ 0);
+#endif
+ RETURN_ERROR_IF(HUF_isError(hSize), dictionary_corrupted, "");
+ dictPtr += hSize;
+ }
+
+ { short offcodeNCount[MaxOff+1];
+ unsigned offcodeMaxValue = MaxOff, offcodeLog;
+ size_t const offcodeHeaderSize = FSE_readNCount(offcodeNCount, &offcodeMaxValue, &offcodeLog, dictPtr, (size_t)(dictEnd-dictPtr));
+ RETURN_ERROR_IF(FSE_isError(offcodeHeaderSize), dictionary_corrupted, "");
+ RETURN_ERROR_IF(offcodeMaxValue > MaxOff, dictionary_corrupted, "");
+ RETURN_ERROR_IF(offcodeLog > OffFSELog, dictionary_corrupted, "");
+ ZSTD_buildFSETable( entropy->OFTable,
+ offcodeNCount, offcodeMaxValue,
+ OF_base, OF_bits,
+ offcodeLog,
+ entropy->workspace, sizeof(entropy->workspace),
+ /* bmi2 */0);
+ dictPtr += offcodeHeaderSize;
+ }
+
+ { short matchlengthNCount[MaxML+1];
+ unsigned matchlengthMaxValue = MaxML, matchlengthLog;
+ size_t const matchlengthHeaderSize = FSE_readNCount(matchlengthNCount, &matchlengthMaxValue, &matchlengthLog, dictPtr, (size_t)(dictEnd-dictPtr));
+ RETURN_ERROR_IF(FSE_isError(matchlengthHeaderSize), dictionary_corrupted, "");
+ RETURN_ERROR_IF(matchlengthMaxValue > MaxML, dictionary_corrupted, "");
+ RETURN_ERROR_IF(matchlengthLog > MLFSELog, dictionary_corrupted, "");
+ ZSTD_buildFSETable( entropy->MLTable,
+ matchlengthNCount, matchlengthMaxValue,
+ ML_base, ML_bits,
+ matchlengthLog,
+ entropy->workspace, sizeof(entropy->workspace),
+ /* bmi2 */ 0);
+ dictPtr += matchlengthHeaderSize;
+ }
+
+ { short litlengthNCount[MaxLL+1];
+ unsigned litlengthMaxValue = MaxLL, litlengthLog;
+ size_t const litlengthHeaderSize = FSE_readNCount(litlengthNCount, &litlengthMaxValue, &litlengthLog, dictPtr, (size_t)(dictEnd-dictPtr));
+ RETURN_ERROR_IF(FSE_isError(litlengthHeaderSize), dictionary_corrupted, "");
+ RETURN_ERROR_IF(litlengthMaxValue > MaxLL, dictionary_corrupted, "");
+ RETURN_ERROR_IF(litlengthLog > LLFSELog, dictionary_corrupted, "");
+ ZSTD_buildFSETable( entropy->LLTable,
+ litlengthNCount, litlengthMaxValue,
+ LL_base, LL_bits,
+ litlengthLog,
+ entropy->workspace, sizeof(entropy->workspace),
+ /* bmi2 */ 0);
+ dictPtr += litlengthHeaderSize;
+ }
+
+ RETURN_ERROR_IF(dictPtr+12 > dictEnd, dictionary_corrupted, "");
+ { int i;
+ size_t const dictContentSize = (size_t)(dictEnd - (dictPtr+12));
+ for (i=0; i<3; i++) {
+ U32 const rep = MEM_readLE32(dictPtr); dictPtr += 4;
+ RETURN_ERROR_IF(rep==0 || rep > dictContentSize,
+ dictionary_corrupted, "");
+ entropy->rep[i] = rep;
+ } }
+
+ return (size_t)(dictPtr - (const BYTE*)dict);
+}
+
+static size_t ZSTD_decompress_insertDictionary(ZSTD_DCtx* dctx, const void* dict, size_t dictSize)
+{
+ if (dictSize < 8) return ZSTD_refDictContent(dctx, dict, dictSize);
+ { U32 const magic = MEM_readLE32(dict);
+ if (magic != ZSTD_MAGIC_DICTIONARY) {
+ return ZSTD_refDictContent(dctx, dict, dictSize); /* pure content mode */
+ } }
+ dctx->dictID = MEM_readLE32((const char*)dict + ZSTD_FRAMEIDSIZE);
+
+ /* load entropy tables */
+ { size_t const eSize = ZSTD_loadDEntropy(&dctx->entropy, dict, dictSize);
+ RETURN_ERROR_IF(ZSTD_isError(eSize), dictionary_corrupted, "");
+ dict = (const char*)dict + eSize;
+ dictSize -= eSize;
+ }
+ dctx->litEntropy = dctx->fseEntropy = 1;
+
+ /* reference dictionary content */
+ return ZSTD_refDictContent(dctx, dict, dictSize);
+}
+
+size_t ZSTD_decompressBegin(ZSTD_DCtx* dctx)
+{
+ assert(dctx != NULL);
+#if ZSTD_TRACE
+ dctx->traceCtx = (ZSTD_trace_decompress_begin != NULL) ? ZSTD_trace_decompress_begin(dctx) : 0;
+#endif
+ dctx->expected = ZSTD_startingInputLength(dctx->format); /* dctx->format must be properly set */
+ dctx->stage = ZSTDds_getFrameHeaderSize;
+ dctx->processedCSize = 0;
+ dctx->decodedSize = 0;
+ dctx->previousDstEnd = NULL;
+ dctx->prefixStart = NULL;
+ dctx->virtualStart = NULL;
+ dctx->dictEnd = NULL;
+ dctx->entropy.hufTable[0] = (HUF_DTable)((ZSTD_HUFFDTABLE_CAPACITY_LOG)*0x1000001); /* cover both little and big endian */
+ dctx->litEntropy = dctx->fseEntropy = 0;
+ dctx->dictID = 0;
+ dctx->bType = bt_reserved;
+ dctx->isFrameDecompression = 1;
+ ZSTD_STATIC_ASSERT(sizeof(dctx->entropy.rep) == sizeof(repStartValue));
+ ZSTD_memcpy(dctx->entropy.rep, repStartValue, sizeof(repStartValue)); /* initial repcodes */
+ dctx->LLTptr = dctx->entropy.LLTable;
+ dctx->MLTptr = dctx->entropy.MLTable;
+ dctx->OFTptr = dctx->entropy.OFTable;
+ dctx->HUFptr = dctx->entropy.hufTable;
+ return 0;
+}
+
+size_t ZSTD_decompressBegin_usingDict(ZSTD_DCtx* dctx, const void* dict, size_t dictSize)
+{
+ FORWARD_IF_ERROR( ZSTD_decompressBegin(dctx) , "");
+ if (dict && dictSize)
+ RETURN_ERROR_IF(
+ ZSTD_isError(ZSTD_decompress_insertDictionary(dctx, dict, dictSize)),
+ dictionary_corrupted, "");
+ return 0;
+}
+
+
+/* ====== ZSTD_DDict ====== */
+
+size_t ZSTD_decompressBegin_usingDDict(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict)
+{
+ DEBUGLOG(4, "ZSTD_decompressBegin_usingDDict");
+ assert(dctx != NULL);
+ if (ddict) {
+ const char* const dictStart = (const char*)ZSTD_DDict_dictContent(ddict);
+ size_t const dictSize = ZSTD_DDict_dictSize(ddict);
+ const void* const dictEnd = dictStart + dictSize;
+ dctx->ddictIsCold = (dctx->dictEnd != dictEnd);
+ DEBUGLOG(4, "DDict is %s",
+ dctx->ddictIsCold ? "~cold~" : "hot!");
+ }
+ FORWARD_IF_ERROR( ZSTD_decompressBegin(dctx) , "");
+ if (ddict) { /* NULL ddict is equivalent to no dictionary */
+ ZSTD_copyDDictParameters(dctx, ddict);
+ }
+ return 0;
+}
+
+/*! ZSTD_getDictID_fromDict() :
+ * Provides the dictID stored within dictionary.
+ * if @return == 0, the dictionary is not conformant with Zstandard specification.
+ * It can still be loaded, but as a content-only dictionary. */
+unsigned ZSTD_getDictID_fromDict(const void* dict, size_t dictSize)
+{
+ if (dictSize < 8) return 0;
+ if (MEM_readLE32(dict) != ZSTD_MAGIC_DICTIONARY) return 0;
+ return MEM_readLE32((const char*)dict + ZSTD_FRAMEIDSIZE);
+}
+
+/*! ZSTD_getDictID_fromFrame() :
+ * Provides the dictID required to decompress frame stored within `src`.
+ * If @return == 0, the dictID could not be decoded.
+ * This could for one of the following reasons :
+ * - The frame does not require a dictionary (most common case).
+ * - The frame was built with dictID intentionally removed.
+ * Needed dictionary is a hidden piece of information.
+ * Note : this use case also happens when using a non-conformant dictionary.
+ * - `srcSize` is too small, and as a result, frame header could not be decoded.
+ * Note : possible if `srcSize < ZSTD_FRAMEHEADERSIZE_MAX`.
+ * - This is not a Zstandard frame.
+ * When identifying the exact failure cause, it's possible to use
+ * ZSTD_getFrameHeader(), which will provide a more precise error code. */
+unsigned ZSTD_getDictID_fromFrame(const void* src, size_t srcSize)
+{
+ ZSTD_frameHeader zfp = { 0, 0, 0, ZSTD_frame, 0, 0, 0, 0, 0 };
+ size_t const hError = ZSTD_getFrameHeader(&zfp, src, srcSize);
+ if (ZSTD_isError(hError)) return 0;
+ return zfp.dictID;
+}
+
+
+/*! ZSTD_decompress_usingDDict() :
+* Decompression using a pre-digested Dictionary
+* Use dictionary without significant overhead. */
+size_t ZSTD_decompress_usingDDict(ZSTD_DCtx* dctx,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize,
+ const ZSTD_DDict* ddict)
+{
+ /* pass content and size in case legacy frames are encountered */
+ return ZSTD_decompressMultiFrame(dctx, dst, dstCapacity, src, srcSize,
+ NULL, 0,
+ ddict);
+}
+
+
+/*=====================================
+* Streaming decompression
+*====================================*/
+
+ZSTD_DStream* ZSTD_createDStream(void)
+{
+ DEBUGLOG(3, "ZSTD_createDStream");
+ return ZSTD_createDCtx_internal(ZSTD_defaultCMem);
+}
+
+ZSTD_DStream* ZSTD_initStaticDStream(void *workspace, size_t workspaceSize)
+{
+ return ZSTD_initStaticDCtx(workspace, workspaceSize);
+}
+
+ZSTD_DStream* ZSTD_createDStream_advanced(ZSTD_customMem customMem)
+{
+ return ZSTD_createDCtx_internal(customMem);
+}
+
+size_t ZSTD_freeDStream(ZSTD_DStream* zds)
+{
+ return ZSTD_freeDCtx(zds);
+}
+
+
+/* *** Initialization *** */
+
+size_t ZSTD_DStreamInSize(void) { return ZSTD_BLOCKSIZE_MAX + ZSTD_blockHeaderSize; }
+size_t ZSTD_DStreamOutSize(void) { return ZSTD_BLOCKSIZE_MAX; }
+
+size_t ZSTD_DCtx_loadDictionary_advanced(ZSTD_DCtx* dctx,
+ const void* dict, size_t dictSize,
+ ZSTD_dictLoadMethod_e dictLoadMethod,
+ ZSTD_dictContentType_e dictContentType)
+{
+ RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong, "");
+ ZSTD_clearDict(dctx);
+ if (dict && dictSize != 0) {
+ dctx->ddictLocal = ZSTD_createDDict_advanced(dict, dictSize, dictLoadMethod, dictContentType, dctx->customMem);
+ RETURN_ERROR_IF(dctx->ddictLocal == NULL, memory_allocation, "NULL pointer!");
+ dctx->ddict = dctx->ddictLocal;
+ dctx->dictUses = ZSTD_use_indefinitely;
+ }
+ return 0;
+}
+
+size_t ZSTD_DCtx_loadDictionary_byReference(ZSTD_DCtx* dctx, const void* dict, size_t dictSize)
+{
+ return ZSTD_DCtx_loadDictionary_advanced(dctx, dict, dictSize, ZSTD_dlm_byRef, ZSTD_dct_auto);
+}
+
+size_t ZSTD_DCtx_loadDictionary(ZSTD_DCtx* dctx, const void* dict, size_t dictSize)
+{
+ return ZSTD_DCtx_loadDictionary_advanced(dctx, dict, dictSize, ZSTD_dlm_byCopy, ZSTD_dct_auto);
+}
+
+size_t ZSTD_DCtx_refPrefix_advanced(ZSTD_DCtx* dctx, const void* prefix, size_t prefixSize, ZSTD_dictContentType_e dictContentType)
+{
+ FORWARD_IF_ERROR(ZSTD_DCtx_loadDictionary_advanced(dctx, prefix, prefixSize, ZSTD_dlm_byRef, dictContentType), "");
+ dctx->dictUses = ZSTD_use_once;
+ return 0;
+}
+
+size_t ZSTD_DCtx_refPrefix(ZSTD_DCtx* dctx, const void* prefix, size_t prefixSize)
+{
+ return ZSTD_DCtx_refPrefix_advanced(dctx, prefix, prefixSize, ZSTD_dct_rawContent);
+}
+
+
+/* ZSTD_initDStream_usingDict() :
+ * return : expected size, aka ZSTD_startingInputLength().
+ * this function cannot fail */
+size_t ZSTD_initDStream_usingDict(ZSTD_DStream* zds, const void* dict, size_t dictSize)
+{
+ DEBUGLOG(4, "ZSTD_initDStream_usingDict");
+ FORWARD_IF_ERROR( ZSTD_DCtx_reset(zds, ZSTD_reset_session_only) , "");
+ FORWARD_IF_ERROR( ZSTD_DCtx_loadDictionary(zds, dict, dictSize) , "");
+ return ZSTD_startingInputLength(zds->format);
+}
+
+/* note : this variant can't fail */
+size_t ZSTD_initDStream(ZSTD_DStream* zds)
+{
+ DEBUGLOG(4, "ZSTD_initDStream");
+ FORWARD_IF_ERROR(ZSTD_DCtx_reset(zds, ZSTD_reset_session_only), "");
+ FORWARD_IF_ERROR(ZSTD_DCtx_refDDict(zds, NULL), "");
+ return ZSTD_startingInputLength(zds->format);
+}
+
+/* ZSTD_initDStream_usingDDict() :
+ * ddict will just be referenced, and must outlive decompression session
+ * this function cannot fail */
+size_t ZSTD_initDStream_usingDDict(ZSTD_DStream* dctx, const ZSTD_DDict* ddict)
+{
+ DEBUGLOG(4, "ZSTD_initDStream_usingDDict");
+ FORWARD_IF_ERROR( ZSTD_DCtx_reset(dctx, ZSTD_reset_session_only) , "");
+ FORWARD_IF_ERROR( ZSTD_DCtx_refDDict(dctx, ddict) , "");
+ return ZSTD_startingInputLength(dctx->format);
+}
+
+/* ZSTD_resetDStream() :
+ * return : expected size, aka ZSTD_startingInputLength().
+ * this function cannot fail */
+size_t ZSTD_resetDStream(ZSTD_DStream* dctx)
+{
+ DEBUGLOG(4, "ZSTD_resetDStream");
+ FORWARD_IF_ERROR(ZSTD_DCtx_reset(dctx, ZSTD_reset_session_only), "");
+ return ZSTD_startingInputLength(dctx->format);
+}
+
+
+size_t ZSTD_DCtx_refDDict(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict)
+{
+ RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong, "");
+ ZSTD_clearDict(dctx);
+ if (ddict) {
+ dctx->ddict = ddict;
+ dctx->dictUses = ZSTD_use_indefinitely;
+ if (dctx->refMultipleDDicts == ZSTD_rmd_refMultipleDDicts) {
+ if (dctx->ddictSet == NULL) {
+ dctx->ddictSet = ZSTD_createDDictHashSet(dctx->customMem);
+ if (!dctx->ddictSet) {
+ RETURN_ERROR(memory_allocation, "Failed to allocate memory for hash set!");
+ }
+ }
+ assert(!dctx->staticSize); /* Impossible: ddictSet cannot have been allocated if static dctx */
+ FORWARD_IF_ERROR(ZSTD_DDictHashSet_addDDict(dctx->ddictSet, ddict, dctx->customMem), "");
+ }
+ }
+ return 0;
+}
+
+/* ZSTD_DCtx_setMaxWindowSize() :
+ * note : no direct equivalence in ZSTD_DCtx_setParameter,
+ * since this version sets windowSize, and the other sets windowLog */
+size_t ZSTD_DCtx_setMaxWindowSize(ZSTD_DCtx* dctx, size_t maxWindowSize)
+{
+ ZSTD_bounds const bounds = ZSTD_dParam_getBounds(ZSTD_d_windowLogMax);
+ size_t const min = (size_t)1 << bounds.lowerBound;
+ size_t const max = (size_t)1 << bounds.upperBound;
+ RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong, "");
+ RETURN_ERROR_IF(maxWindowSize < min, parameter_outOfBound, "");
+ RETURN_ERROR_IF(maxWindowSize > max, parameter_outOfBound, "");
+ dctx->maxWindowSize = maxWindowSize;
+ return 0;
+}
+
+size_t ZSTD_DCtx_setFormat(ZSTD_DCtx* dctx, ZSTD_format_e format)
+{
+ return ZSTD_DCtx_setParameter(dctx, ZSTD_d_format, (int)format);
+}
+
+ZSTD_bounds ZSTD_dParam_getBounds(ZSTD_dParameter dParam)
+{
+ ZSTD_bounds bounds = { 0, 0, 0 };
+ switch(dParam) {
+ case ZSTD_d_windowLogMax:
+ bounds.lowerBound = ZSTD_WINDOWLOG_ABSOLUTEMIN;
+ bounds.upperBound = ZSTD_WINDOWLOG_MAX;
+ return bounds;
+ case ZSTD_d_format:
+ bounds.lowerBound = (int)ZSTD_f_zstd1;
+ bounds.upperBound = (int)ZSTD_f_zstd1_magicless;
+ ZSTD_STATIC_ASSERT(ZSTD_f_zstd1 < ZSTD_f_zstd1_magicless);
+ return bounds;
+ case ZSTD_d_stableOutBuffer:
+ bounds.lowerBound = (int)ZSTD_bm_buffered;
+ bounds.upperBound = (int)ZSTD_bm_stable;
+ return bounds;
+ case ZSTD_d_forceIgnoreChecksum:
+ bounds.lowerBound = (int)ZSTD_d_validateChecksum;
+ bounds.upperBound = (int)ZSTD_d_ignoreChecksum;
+ return bounds;
+ case ZSTD_d_refMultipleDDicts:
+ bounds.lowerBound = (int)ZSTD_rmd_refSingleDDict;
+ bounds.upperBound = (int)ZSTD_rmd_refMultipleDDicts;
+ return bounds;
+ case ZSTD_d_disableHuffmanAssembly:
+ bounds.lowerBound = 0;
+ bounds.upperBound = 1;
+ return bounds;
+ case ZSTD_d_maxBlockSize:
+ bounds.lowerBound = ZSTD_BLOCKSIZE_MAX_MIN;
+ bounds.upperBound = ZSTD_BLOCKSIZE_MAX;
+ return bounds;
+
+ default:;
+ }
+ bounds.error = ERROR(parameter_unsupported);
+ return bounds;
+}
+
+/* ZSTD_dParam_withinBounds:
+ * @return 1 if value is within dParam bounds,
+ * 0 otherwise */
+static int ZSTD_dParam_withinBounds(ZSTD_dParameter dParam, int value)
+{
+ ZSTD_bounds const bounds = ZSTD_dParam_getBounds(dParam);
+ if (ZSTD_isError(bounds.error)) return 0;
+ if (value < bounds.lowerBound) return 0;
+ if (value > bounds.upperBound) return 0;
+ return 1;
+}
+
+#define CHECK_DBOUNDS(p,v) { \
+ RETURN_ERROR_IF(!ZSTD_dParam_withinBounds(p, v), parameter_outOfBound, ""); \
+}
+
+size_t ZSTD_DCtx_getParameter(ZSTD_DCtx* dctx, ZSTD_dParameter param, int* value)
+{
+ switch (param) {
+ case ZSTD_d_windowLogMax:
+ *value = (int)ZSTD_highbit32((U32)dctx->maxWindowSize);
+ return 0;
+ case ZSTD_d_format:
+ *value = (int)dctx->format;
+ return 0;
+ case ZSTD_d_stableOutBuffer:
+ *value = (int)dctx->outBufferMode;
+ return 0;
+ case ZSTD_d_forceIgnoreChecksum:
+ *value = (int)dctx->forceIgnoreChecksum;
+ return 0;
+ case ZSTD_d_refMultipleDDicts:
+ *value = (int)dctx->refMultipleDDicts;
+ return 0;
+ case ZSTD_d_disableHuffmanAssembly:
+ *value = (int)dctx->disableHufAsm;
+ return 0;
+ case ZSTD_d_maxBlockSize:
+ *value = dctx->maxBlockSizeParam;
+ return 0;
+ default:;
+ }
+ RETURN_ERROR(parameter_unsupported, "");
+}
+
+size_t ZSTD_DCtx_setParameter(ZSTD_DCtx* dctx, ZSTD_dParameter dParam, int value)
+{
+ RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong, "");
+ switch(dParam) {
+ case ZSTD_d_windowLogMax:
+ if (value == 0) value = ZSTD_WINDOWLOG_LIMIT_DEFAULT;
+ CHECK_DBOUNDS(ZSTD_d_windowLogMax, value);
+ dctx->maxWindowSize = ((size_t)1) << value;
+ return 0;
+ case ZSTD_d_format:
+ CHECK_DBOUNDS(ZSTD_d_format, value);
+ dctx->format = (ZSTD_format_e)value;
+ return 0;
+ case ZSTD_d_stableOutBuffer:
+ CHECK_DBOUNDS(ZSTD_d_stableOutBuffer, value);
+ dctx->outBufferMode = (ZSTD_bufferMode_e)value;
+ return 0;
+ case ZSTD_d_forceIgnoreChecksum:
+ CHECK_DBOUNDS(ZSTD_d_forceIgnoreChecksum, value);
+ dctx->forceIgnoreChecksum = (ZSTD_forceIgnoreChecksum_e)value;
+ return 0;
+ case ZSTD_d_refMultipleDDicts:
+ CHECK_DBOUNDS(ZSTD_d_refMultipleDDicts, value);
+ if (dctx->staticSize != 0) {
+ RETURN_ERROR(parameter_unsupported, "Static dctx does not support multiple DDicts!");
+ }
+ dctx->refMultipleDDicts = (ZSTD_refMultipleDDicts_e)value;
+ return 0;
+ case ZSTD_d_disableHuffmanAssembly:
+ CHECK_DBOUNDS(ZSTD_d_disableHuffmanAssembly, value);
+ dctx->disableHufAsm = value != 0;
+ return 0;
+ case ZSTD_d_maxBlockSize:
+ if (value != 0) CHECK_DBOUNDS(ZSTD_d_maxBlockSize, value);
+ dctx->maxBlockSizeParam = value;
+ return 0;
+ default:;
+ }
+ RETURN_ERROR(parameter_unsupported, "");
+}
+
+size_t ZSTD_DCtx_reset(ZSTD_DCtx* dctx, ZSTD_ResetDirective reset)
+{
+ if ( (reset == ZSTD_reset_session_only)
+ || (reset == ZSTD_reset_session_and_parameters) ) {
+ dctx->streamStage = zdss_init;
+ dctx->noForwardProgress = 0;
+ dctx->isFrameDecompression = 1;
+ }
+ if ( (reset == ZSTD_reset_parameters)
+ || (reset == ZSTD_reset_session_and_parameters) ) {
+ RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong, "");
+ ZSTD_clearDict(dctx);
+ ZSTD_DCtx_resetParameters(dctx);
+ }
+ return 0;
+}
+
+
+size_t ZSTD_sizeof_DStream(const ZSTD_DStream* dctx)
+{
+ return ZSTD_sizeof_DCtx(dctx);
+}
+
+static size_t ZSTD_decodingBufferSize_internal(unsigned long long windowSize, unsigned long long frameContentSize, size_t blockSizeMax)
+{
+ size_t const blockSize = MIN((size_t)MIN(windowSize, ZSTD_BLOCKSIZE_MAX), blockSizeMax);
+ /* We need blockSize + WILDCOPY_OVERLENGTH worth of buffer so that if a block
+ * ends at windowSize + WILDCOPY_OVERLENGTH + 1 bytes, we can start writing
+ * the block at the beginning of the output buffer, and maintain a full window.
+ *
+ * We need another blockSize worth of buffer so that we can store split
+ * literals at the end of the block without overwriting the extDict window.
+ */
+ unsigned long long const neededRBSize = windowSize + (blockSize * 2) + (WILDCOPY_OVERLENGTH * 2);
+ unsigned long long const neededSize = MIN(frameContentSize, neededRBSize);
+ size_t const minRBSize = (size_t) neededSize;
+ RETURN_ERROR_IF((unsigned long long)minRBSize != neededSize,
+ frameParameter_windowTooLarge, "");
+ return minRBSize;
+}
+
+size_t ZSTD_decodingBufferSize_min(unsigned long long windowSize, unsigned long long frameContentSize)
+{
+ return ZSTD_decodingBufferSize_internal(windowSize, frameContentSize, ZSTD_BLOCKSIZE_MAX);
+}
+
+size_t ZSTD_estimateDStreamSize(size_t windowSize)
+{
+ size_t const blockSize = MIN(windowSize, ZSTD_BLOCKSIZE_MAX);
+ size_t const inBuffSize = blockSize; /* no block can be larger */
+ size_t const outBuffSize = ZSTD_decodingBufferSize_min(windowSize, ZSTD_CONTENTSIZE_UNKNOWN);
+ return ZSTD_estimateDCtxSize() + inBuffSize + outBuffSize;
+}
+
+size_t ZSTD_estimateDStreamSize_fromFrame(const void* src, size_t srcSize)
+{
+ U32 const windowSizeMax = 1U << ZSTD_WINDOWLOG_MAX; /* note : should be user-selectable, but requires an additional parameter (or a dctx) */
+ ZSTD_frameHeader zfh;
+ size_t const err = ZSTD_getFrameHeader(&zfh, src, srcSize);
+ if (ZSTD_isError(err)) return err;
+ RETURN_ERROR_IF(err>0, srcSize_wrong, "");
+ RETURN_ERROR_IF(zfh.windowSize > windowSizeMax,
+ frameParameter_windowTooLarge, "");
+ return ZSTD_estimateDStreamSize((size_t)zfh.windowSize);
+}
+
+
+/* ***** Decompression ***** */
+
+static int ZSTD_DCtx_isOverflow(ZSTD_DStream* zds, size_t const neededInBuffSize, size_t const neededOutBuffSize)
+{
+ return (zds->inBuffSize + zds->outBuffSize) >= (neededInBuffSize + neededOutBuffSize) * ZSTD_WORKSPACETOOLARGE_FACTOR;
+}
+
+static void ZSTD_DCtx_updateOversizedDuration(ZSTD_DStream* zds, size_t const neededInBuffSize, size_t const neededOutBuffSize)
+{
+ if (ZSTD_DCtx_isOverflow(zds, neededInBuffSize, neededOutBuffSize))
+ zds->oversizedDuration++;
+ else
+ zds->oversizedDuration = 0;
+}
+
+static int ZSTD_DCtx_isOversizedTooLong(ZSTD_DStream* zds)
+{
+ return zds->oversizedDuration >= ZSTD_WORKSPACETOOLARGE_MAXDURATION;
+}
+
+/* Checks that the output buffer hasn't changed if ZSTD_obm_stable is used. */
+static size_t ZSTD_checkOutBuffer(ZSTD_DStream const* zds, ZSTD_outBuffer const* output)
+{
+ ZSTD_outBuffer const expect = zds->expectedOutBuffer;
+ /* No requirement when ZSTD_obm_stable is not enabled. */
+ if (zds->outBufferMode != ZSTD_bm_stable)
+ return 0;
+ /* Any buffer is allowed in zdss_init, this must be the same for every other call until
+ * the context is reset.
+ */
+ if (zds->streamStage == zdss_init)
+ return 0;
+ /* The buffer must match our expectation exactly. */
+ if (expect.dst == output->dst && expect.pos == output->pos && expect.size == output->size)
+ return 0;
+ RETURN_ERROR(dstBuffer_wrong, "ZSTD_d_stableOutBuffer enabled but output differs!");
+}
+
+/* Calls ZSTD_decompressContinue() with the right parameters for ZSTD_decompressStream()
+ * and updates the stage and the output buffer state. This call is extracted so it can be
+ * used both when reading directly from the ZSTD_inBuffer, and in buffered input mode.
+ * NOTE: You must break after calling this function since the streamStage is modified.
+ */
+static size_t ZSTD_decompressContinueStream(
+ ZSTD_DStream* zds, char** op, char* oend,
+ void const* src, size_t srcSize) {
+ int const isSkipFrame = ZSTD_isSkipFrame(zds);
+ if (zds->outBufferMode == ZSTD_bm_buffered) {
+ size_t const dstSize = isSkipFrame ? 0 : zds->outBuffSize - zds->outStart;
+ size_t const decodedSize = ZSTD_decompressContinue(zds,
+ zds->outBuff + zds->outStart, dstSize, src, srcSize);
+ FORWARD_IF_ERROR(decodedSize, "");
+ if (!decodedSize && !isSkipFrame) {
+ zds->streamStage = zdss_read;
+ } else {
+ zds->outEnd = zds->outStart + decodedSize;
+ zds->streamStage = zdss_flush;
+ }
+ } else {
+ /* Write directly into the output buffer */
+ size_t const dstSize = isSkipFrame ? 0 : (size_t)(oend - *op);
+ size_t const decodedSize = ZSTD_decompressContinue(zds, *op, dstSize, src, srcSize);
+ FORWARD_IF_ERROR(decodedSize, "");
+ *op += decodedSize;
+ /* Flushing is not needed. */
+ zds->streamStage = zdss_read;
+ assert(*op <= oend);
+ assert(zds->outBufferMode == ZSTD_bm_stable);
+ }
+ return 0;
+}
+
+size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inBuffer* input)
+{
+ const char* const src = (const char*)input->src;
+ const char* const istart = input->pos != 0 ? src + input->pos : src;
+ const char* const iend = input->size != 0 ? src + input->size : src;
+ const char* ip = istart;
+ char* const dst = (char*)output->dst;
+ char* const ostart = output->pos != 0 ? dst + output->pos : dst;
+ char* const oend = output->size != 0 ? dst + output->size : dst;
+ char* op = ostart;
+ U32 someMoreWork = 1;
+
+ DEBUGLOG(5, "ZSTD_decompressStream");
+ RETURN_ERROR_IF(
+ input->pos > input->size,
+ srcSize_wrong,
+ "forbidden. in: pos: %u vs size: %u",
+ (U32)input->pos, (U32)input->size);
+ RETURN_ERROR_IF(
+ output->pos > output->size,
+ dstSize_tooSmall,
+ "forbidden. out: pos: %u vs size: %u",
+ (U32)output->pos, (U32)output->size);
+ DEBUGLOG(5, "input size : %u", (U32)(input->size - input->pos));
+ FORWARD_IF_ERROR(ZSTD_checkOutBuffer(zds, output), "");
+
+ while (someMoreWork) {
+ switch(zds->streamStage)
+ {
+ case zdss_init :
+ DEBUGLOG(5, "stage zdss_init => transparent reset ");
+ zds->streamStage = zdss_loadHeader;
+ zds->lhSize = zds->inPos = zds->outStart = zds->outEnd = 0;
+#if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT>=1)
+ zds->legacyVersion = 0;
+#endif
+ zds->hostageByte = 0;
+ zds->expectedOutBuffer = *output;
+ ZSTD_FALLTHROUGH;
+
+ case zdss_loadHeader :
+ DEBUGLOG(5, "stage zdss_loadHeader (srcSize : %u)", (U32)(iend - ip));
+#if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT>=1)
+ if (zds->legacyVersion) {
+ RETURN_ERROR_IF(zds->staticSize, memory_allocation,
+ "legacy support is incompatible with static dctx");
+ { size_t const hint = ZSTD_decompressLegacyStream(zds->legacyContext, zds->legacyVersion, output, input);
+ if (hint==0) zds->streamStage = zdss_init;
+ return hint;
+ } }
+#endif
+ { size_t const hSize = ZSTD_getFrameHeader_advanced(&zds->fParams, zds->headerBuffer, zds->lhSize, zds->format);
+ if (zds->refMultipleDDicts && zds->ddictSet) {
+ ZSTD_DCtx_selectFrameDDict(zds);
+ }
+ if (ZSTD_isError(hSize)) {
+#if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT>=1)
+ U32 const legacyVersion = ZSTD_isLegacy(istart, iend-istart);
+ if (legacyVersion) {
+ ZSTD_DDict const* const ddict = ZSTD_getDDict(zds);
+ const void* const dict = ddict ? ZSTD_DDict_dictContent(ddict) : NULL;
+ size_t const dictSize = ddict ? ZSTD_DDict_dictSize(ddict) : 0;
+ DEBUGLOG(5, "ZSTD_decompressStream: detected legacy version v0.%u", legacyVersion);
+ RETURN_ERROR_IF(zds->staticSize, memory_allocation,
+ "legacy support is incompatible with static dctx");
+ FORWARD_IF_ERROR(ZSTD_initLegacyStream(&zds->legacyContext,
+ zds->previousLegacyVersion, legacyVersion,
+ dict, dictSize), "");
+ zds->legacyVersion = zds->previousLegacyVersion = legacyVersion;
+ { size_t const hint = ZSTD_decompressLegacyStream(zds->legacyContext, legacyVersion, output, input);
+ if (hint==0) zds->streamStage = zdss_init; /* or stay in stage zdss_loadHeader */
+ return hint;
+ } }
+#endif
+ return hSize; /* error */
+ }
+ if (hSize != 0) { /* need more input */
+ size_t const toLoad = hSize - zds->lhSize; /* if hSize!=0, hSize > zds->lhSize */
+ size_t const remainingInput = (size_t)(iend-ip);
+ assert(iend >= ip);
+ if (toLoad > remainingInput) { /* not enough input to load full header */
+ if (remainingInput > 0) {
+ ZSTD_memcpy(zds->headerBuffer + zds->lhSize, ip, remainingInput);
+ zds->lhSize += remainingInput;
+ }
+ input->pos = input->size;
+ /* check first few bytes */
+ FORWARD_IF_ERROR(
+ ZSTD_getFrameHeader_advanced(&zds->fParams, zds->headerBuffer, zds->lhSize, zds->format),
+ "First few bytes detected incorrect" );
+ /* return hint input size */
+ return (MAX((size_t)ZSTD_FRAMEHEADERSIZE_MIN(zds->format), hSize) - zds->lhSize) + ZSTD_blockHeaderSize; /* remaining header bytes + next block header */
+ }
+ assert(ip != NULL);
+ ZSTD_memcpy(zds->headerBuffer + zds->lhSize, ip, toLoad); zds->lhSize = hSize; ip += toLoad;
+ break;
+ } }
+
+ /* check for single-pass mode opportunity */
+ if (zds->fParams.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN
+ && zds->fParams.frameType != ZSTD_skippableFrame
+ && (U64)(size_t)(oend-op) >= zds->fParams.frameContentSize) {
+ size_t const cSize = ZSTD_findFrameCompressedSize_advanced(istart, (size_t)(iend-istart), zds->format);
+ if (cSize <= (size_t)(iend-istart)) {
+ /* shortcut : using single-pass mode */
+ size_t const decompressedSize = ZSTD_decompress_usingDDict(zds, op, (size_t)(oend-op), istart, cSize, ZSTD_getDDict(zds));
+ if (ZSTD_isError(decompressedSize)) return decompressedSize;
+ DEBUGLOG(4, "shortcut to single-pass ZSTD_decompress_usingDDict()");
+ assert(istart != NULL);
+ ip = istart + cSize;
+ op = op ? op + decompressedSize : op; /* can occur if frameContentSize = 0 (empty frame) */
+ zds->expected = 0;
+ zds->streamStage = zdss_init;
+ someMoreWork = 0;
+ break;
+ } }
+
+ /* Check output buffer is large enough for ZSTD_odm_stable. */
+ if (zds->outBufferMode == ZSTD_bm_stable
+ && zds->fParams.frameType != ZSTD_skippableFrame
+ && zds->fParams.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN
+ && (U64)(size_t)(oend-op) < zds->fParams.frameContentSize) {
+ RETURN_ERROR(dstSize_tooSmall, "ZSTD_obm_stable passed but ZSTD_outBuffer is too small");
+ }
+
+ /* Consume header (see ZSTDds_decodeFrameHeader) */
+ DEBUGLOG(4, "Consume header");
+ FORWARD_IF_ERROR(ZSTD_decompressBegin_usingDDict(zds, ZSTD_getDDict(zds)), "");
+
+ if (zds->format == ZSTD_f_zstd1
+ && (MEM_readLE32(zds->headerBuffer) & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) { /* skippable frame */
+ zds->expected = MEM_readLE32(zds->headerBuffer + ZSTD_FRAMEIDSIZE);
+ zds->stage = ZSTDds_skipFrame;
+ } else {
+ FORWARD_IF_ERROR(ZSTD_decodeFrameHeader(zds, zds->headerBuffer, zds->lhSize), "");
+ zds->expected = ZSTD_blockHeaderSize;
+ zds->stage = ZSTDds_decodeBlockHeader;
+ }
+
+ /* control buffer memory usage */
+ DEBUGLOG(4, "Control max memory usage (%u KB <= max %u KB)",
+ (U32)(zds->fParams.windowSize >>10),
+ (U32)(zds->maxWindowSize >> 10) );
+ zds->fParams.windowSize = MAX(zds->fParams.windowSize, 1U << ZSTD_WINDOWLOG_ABSOLUTEMIN);
+ RETURN_ERROR_IF(zds->fParams.windowSize > zds->maxWindowSize,
+ frameParameter_windowTooLarge, "");
+ if (zds->maxBlockSizeParam != 0)
+ zds->fParams.blockSizeMax = MIN(zds->fParams.blockSizeMax, (unsigned)zds->maxBlockSizeParam);
+
+ /* Adapt buffer sizes to frame header instructions */
+ { size_t const neededInBuffSize = MAX(zds->fParams.blockSizeMax, 4 /* frame checksum */);
+ size_t const neededOutBuffSize = zds->outBufferMode == ZSTD_bm_buffered
+ ? ZSTD_decodingBufferSize_internal(zds->fParams.windowSize, zds->fParams.frameContentSize, zds->fParams.blockSizeMax)
+ : 0;
+
+ ZSTD_DCtx_updateOversizedDuration(zds, neededInBuffSize, neededOutBuffSize);
+
+ { int const tooSmall = (zds->inBuffSize < neededInBuffSize) || (zds->outBuffSize < neededOutBuffSize);
+ int const tooLarge = ZSTD_DCtx_isOversizedTooLong(zds);
+
+ if (tooSmall || tooLarge) {
+ size_t const bufferSize = neededInBuffSize + neededOutBuffSize;
+ DEBUGLOG(4, "inBuff : from %u to %u",
+ (U32)zds->inBuffSize, (U32)neededInBuffSize);
+ DEBUGLOG(4, "outBuff : from %u to %u",
+ (U32)zds->outBuffSize, (U32)neededOutBuffSize);
+ if (zds->staticSize) { /* static DCtx */
+ DEBUGLOG(4, "staticSize : %u", (U32)zds->staticSize);
+ assert(zds->staticSize >= sizeof(ZSTD_DCtx)); /* controlled at init */
+ RETURN_ERROR_IF(
+ bufferSize > zds->staticSize - sizeof(ZSTD_DCtx),
+ memory_allocation, "");
+ } else {
+ ZSTD_customFree(zds->inBuff, zds->customMem);
+ zds->inBuffSize = 0;
+ zds->outBuffSize = 0;
+ zds->inBuff = (char*)ZSTD_customMalloc(bufferSize, zds->customMem);
+ RETURN_ERROR_IF(zds->inBuff == NULL, memory_allocation, "");
+ }
+ zds->inBuffSize = neededInBuffSize;
+ zds->outBuff = zds->inBuff + zds->inBuffSize;
+ zds->outBuffSize = neededOutBuffSize;
+ } } }
+ zds->streamStage = zdss_read;
+ ZSTD_FALLTHROUGH;
+
+ case zdss_read:
+ DEBUGLOG(5, "stage zdss_read");
+ { size_t const neededInSize = ZSTD_nextSrcSizeToDecompressWithInputSize(zds, (size_t)(iend - ip));
+ DEBUGLOG(5, "neededInSize = %u", (U32)neededInSize);
+ if (neededInSize==0) { /* end of frame */
+ zds->streamStage = zdss_init;
+ someMoreWork = 0;
+ break;
+ }
+ if ((size_t)(iend-ip) >= neededInSize) { /* decode directly from src */
+ FORWARD_IF_ERROR(ZSTD_decompressContinueStream(zds, &op, oend, ip, neededInSize), "");
+ assert(ip != NULL);
+ ip += neededInSize;
+ /* Function modifies the stage so we must break */
+ break;
+ } }
+ if (ip==iend) { someMoreWork = 0; break; } /* no more input */
+ zds->streamStage = zdss_load;
+ ZSTD_FALLTHROUGH;
+
+ case zdss_load:
+ { size_t const neededInSize = ZSTD_nextSrcSizeToDecompress(zds);
+ size_t const toLoad = neededInSize - zds->inPos;
+ int const isSkipFrame = ZSTD_isSkipFrame(zds);
+ size_t loadedSize;
+ /* At this point we shouldn't be decompressing a block that we can stream. */
+ assert(neededInSize == ZSTD_nextSrcSizeToDecompressWithInputSize(zds, (size_t)(iend - ip)));
+ if (isSkipFrame) {
+ loadedSize = MIN(toLoad, (size_t)(iend-ip));
+ } else {
+ RETURN_ERROR_IF(toLoad > zds->inBuffSize - zds->inPos,
+ corruption_detected,
+ "should never happen");
+ loadedSize = ZSTD_limitCopy(zds->inBuff + zds->inPos, toLoad, ip, (size_t)(iend-ip));
+ }
+ if (loadedSize != 0) {
+ /* ip may be NULL */
+ ip += loadedSize;
+ zds->inPos += loadedSize;
+ }
+ if (loadedSize < toLoad) { someMoreWork = 0; break; } /* not enough input, wait for more */
+
+ /* decode loaded input */
+ zds->inPos = 0; /* input is consumed */
+ FORWARD_IF_ERROR(ZSTD_decompressContinueStream(zds, &op, oend, zds->inBuff, neededInSize), "");
+ /* Function modifies the stage so we must break */
+ break;
+ }
+ case zdss_flush:
+ {
+ size_t const toFlushSize = zds->outEnd - zds->outStart;
+ size_t const flushedSize = ZSTD_limitCopy(op, (size_t)(oend-op), zds->outBuff + zds->outStart, toFlushSize);
+
+ op = op ? op + flushedSize : op;
+
+ zds->outStart += flushedSize;
+ if (flushedSize == toFlushSize) { /* flush completed */
+ zds->streamStage = zdss_read;
+ if ( (zds->outBuffSize < zds->fParams.frameContentSize)
+ && (zds->outStart + zds->fParams.blockSizeMax > zds->outBuffSize) ) {
+ DEBUGLOG(5, "restart filling outBuff from beginning (left:%i, needed:%u)",
+ (int)(zds->outBuffSize - zds->outStart),
+ (U32)zds->fParams.blockSizeMax);
+ zds->outStart = zds->outEnd = 0;
+ }
+ break;
+ } }
+ /* cannot complete flush */
+ someMoreWork = 0;
+ break;
+
+ default:
+ assert(0); /* impossible */
+ RETURN_ERROR(GENERIC, "impossible to reach"); /* some compilers require default to do something */
+ } }
+
+ /* result */
+ input->pos = (size_t)(ip - (const char*)(input->src));
+ output->pos = (size_t)(op - (char*)(output->dst));
+
+ /* Update the expected output buffer for ZSTD_obm_stable. */
+ zds->expectedOutBuffer = *output;
+
+ if ((ip==istart) && (op==ostart)) { /* no forward progress */
+ zds->noForwardProgress ++;
+ if (zds->noForwardProgress >= ZSTD_NO_FORWARD_PROGRESS_MAX) {
+ RETURN_ERROR_IF(op==oend, noForwardProgress_destFull, "");
+ RETURN_ERROR_IF(ip==iend, noForwardProgress_inputEmpty, "");
+ assert(0);
+ }
+ } else {
+ zds->noForwardProgress = 0;
+ }
+ { size_t nextSrcSizeHint = ZSTD_nextSrcSizeToDecompress(zds);
+ if (!nextSrcSizeHint) { /* frame fully decoded */
+ if (zds->outEnd == zds->outStart) { /* output fully flushed */
+ if (zds->hostageByte) {
+ if (input->pos >= input->size) {
+ /* can't release hostage (not present) */
+ zds->streamStage = zdss_read;
+ return 1;
+ }
+ input->pos++; /* release hostage */
+ } /* zds->hostageByte */
+ return 0;
+ } /* zds->outEnd == zds->outStart */
+ if (!zds->hostageByte) { /* output not fully flushed; keep last byte as hostage; will be released when all output is flushed */
+ input->pos--; /* note : pos > 0, otherwise, impossible to finish reading last block */
+ zds->hostageByte=1;
+ }
+ return 1;
+ } /* nextSrcSizeHint==0 */
+ nextSrcSizeHint += ZSTD_blockHeaderSize * (ZSTD_nextInputType(zds) == ZSTDnit_block); /* preload header of next block */
+ assert(zds->inPos <= nextSrcSizeHint);
+ nextSrcSizeHint -= zds->inPos; /* part already loaded*/
+ return nextSrcSizeHint;
+ }
+}
+
+size_t ZSTD_decompressStream_simpleArgs (
+ ZSTD_DCtx* dctx,
+ void* dst, size_t dstCapacity, size_t* dstPos,
+ const void* src, size_t srcSize, size_t* srcPos)
+{
+ ZSTD_outBuffer output;
+ ZSTD_inBuffer input;
+ output.dst = dst;
+ output.size = dstCapacity;
+ output.pos = *dstPos;
+ input.src = src;
+ input.size = srcSize;
+ input.pos = *srcPos;
+ { size_t const cErr = ZSTD_decompressStream(dctx, &output, &input);
+ *dstPos = output.pos;
+ *srcPos = input.pos;
+ return cErr;
+ }
+}
diff --git a/third_party/zstd/lib/decompress/zstd_decompress_block.c b/third_party/zstd/lib/decompress/zstd_decompress_block.c
new file mode 100644
index 0000000000..76d7332e88
--- /dev/null
+++ b/third_party/zstd/lib/decompress/zstd_decompress_block.c
@@ -0,0 +1,2215 @@
+/*
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+/* zstd_decompress_block :
+ * this module takes care of decompressing _compressed_ block */
+
+/*-*******************************************************
+* Dependencies
+*********************************************************/
+#include "../common/zstd_deps.h" /* ZSTD_memcpy, ZSTD_memmove, ZSTD_memset */
+#include "../common/compiler.h" /* prefetch */
+#include "../common/cpu.h" /* bmi2 */
+#include "../common/mem.h" /* low level memory routines */
+#define FSE_STATIC_LINKING_ONLY
+#include "../common/fse.h"
+#include "../common/huf.h"
+#include "../common/zstd_internal.h"
+#include "zstd_decompress_internal.h" /* ZSTD_DCtx */
+#include "zstd_ddict.h" /* ZSTD_DDictDictContent */
+#include "zstd_decompress_block.h"
+#include "../common/bits.h" /* ZSTD_highbit32 */
+
+/*_*******************************************************
+* Macros
+**********************************************************/
+
+/* These two optional macros force the use one way or another of the two
+ * ZSTD_decompressSequences implementations. You can't force in both directions
+ * at the same time.
+ */
+#if defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT) && \
+ defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG)
+#error "Cannot force the use of the short and the long ZSTD_decompressSequences variants!"
+#endif
+
+
+/*_*******************************************************
+* Memory operations
+**********************************************************/
+static void ZSTD_copy4(void* dst, const void* src) { ZSTD_memcpy(dst, src, 4); }
+
+
+/*-*************************************************************
+ * Block decoding
+ ***************************************************************/
+
+static size_t ZSTD_blockSizeMax(ZSTD_DCtx const* dctx)
+{
+ size_t const blockSizeMax = dctx->isFrameDecompression ? dctx->fParams.blockSizeMax : ZSTD_BLOCKSIZE_MAX;
+ assert(blockSizeMax <= ZSTD_BLOCKSIZE_MAX);
+ return blockSizeMax;
+}
+
+/*! ZSTD_getcBlockSize() :
+ * Provides the size of compressed block from block header `src` */
+size_t ZSTD_getcBlockSize(const void* src, size_t srcSize,
+ blockProperties_t* bpPtr)
+{
+ RETURN_ERROR_IF(srcSize < ZSTD_blockHeaderSize, srcSize_wrong, "");
+
+ { U32 const cBlockHeader = MEM_readLE24(src);
+ U32 const cSize = cBlockHeader >> 3;
+ bpPtr->lastBlock = cBlockHeader & 1;
+ bpPtr->blockType = (blockType_e)((cBlockHeader >> 1) & 3);
+ bpPtr->origSize = cSize; /* only useful for RLE */
+ if (bpPtr->blockType == bt_rle) return 1;
+ RETURN_ERROR_IF(bpPtr->blockType == bt_reserved, corruption_detected, "");
+ return cSize;
+ }
+}
+
+/* Allocate buffer for literals, either overlapping current dst, or split between dst and litExtraBuffer, or stored entirely within litExtraBuffer */
+static void ZSTD_allocateLiteralsBuffer(ZSTD_DCtx* dctx, void* const dst, const size_t dstCapacity, const size_t litSize,
+ const streaming_operation streaming, const size_t expectedWriteSize, const unsigned splitImmediately)
+{
+ size_t const blockSizeMax = ZSTD_blockSizeMax(dctx);
+ assert(litSize <= blockSizeMax);
+ assert(dctx->isFrameDecompression || streaming == not_streaming);
+ assert(expectedWriteSize <= blockSizeMax);
+ if (streaming == not_streaming && dstCapacity > blockSizeMax + WILDCOPY_OVERLENGTH + litSize + WILDCOPY_OVERLENGTH) {
+ /* If we aren't streaming, we can just put the literals after the output
+ * of the current block. We don't need to worry about overwriting the
+ * extDict of our window, because it doesn't exist.
+ * So if we have space after the end of the block, just put it there.
+ */
+ dctx->litBuffer = (BYTE*)dst + blockSizeMax + WILDCOPY_OVERLENGTH;
+ dctx->litBufferEnd = dctx->litBuffer + litSize;
+ dctx->litBufferLocation = ZSTD_in_dst;
+ } else if (litSize <= ZSTD_LITBUFFEREXTRASIZE) {
+ /* Literals fit entirely within the extra buffer, put them there to avoid
+ * having to split the literals.
+ */
+ dctx->litBuffer = dctx->litExtraBuffer;
+ dctx->litBufferEnd = dctx->litBuffer + litSize;
+ dctx->litBufferLocation = ZSTD_not_in_dst;
+ } else {
+ assert(blockSizeMax > ZSTD_LITBUFFEREXTRASIZE);
+ /* Literals must be split between the output block and the extra lit
+ * buffer. We fill the extra lit buffer with the tail of the literals,
+ * and put the rest of the literals at the end of the block, with
+ * WILDCOPY_OVERLENGTH of buffer room to allow for overreads.
+ * This MUST not write more than our maxBlockSize beyond dst, because in
+ * streaming mode, that could overwrite part of our extDict window.
+ */
+ if (splitImmediately) {
+ /* won't fit in litExtraBuffer, so it will be split between end of dst and extra buffer */
+ dctx->litBuffer = (BYTE*)dst + expectedWriteSize - litSize + ZSTD_LITBUFFEREXTRASIZE - WILDCOPY_OVERLENGTH;
+ dctx->litBufferEnd = dctx->litBuffer + litSize - ZSTD_LITBUFFEREXTRASIZE;
+ } else {
+ /* initially this will be stored entirely in dst during huffman decoding, it will partially be shifted to litExtraBuffer after */
+ dctx->litBuffer = (BYTE*)dst + expectedWriteSize - litSize;
+ dctx->litBufferEnd = (BYTE*)dst + expectedWriteSize;
+ }
+ dctx->litBufferLocation = ZSTD_split;
+ assert(dctx->litBufferEnd <= (BYTE*)dst + expectedWriteSize);
+ }
+}
+
+/*! ZSTD_decodeLiteralsBlock() :
+ * Where it is possible to do so without being stomped by the output during decompression, the literals block will be stored
+ * 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
+ * block will be output. Otherwise it will be stored at the end of the current dst blockspace, with a small portion being
+ * stored in dctx->litExtraBuffer to help keep it "ahead" of the current output write.
+ *
+ * @return : nb of bytes read from src (< srcSize )
+ * note : symbol not declared but exposed for fullbench */
+static size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
+ const void* src, size_t srcSize, /* note : srcSize < BLOCKSIZE */
+ void* dst, size_t dstCapacity, const streaming_operation streaming)
+{
+ DEBUGLOG(5, "ZSTD_decodeLiteralsBlock");
+ RETURN_ERROR_IF(srcSize < MIN_CBLOCK_SIZE, corruption_detected, "");
+
+ { const BYTE* const istart = (const BYTE*) src;
+ symbolEncodingType_e const litEncType = (symbolEncodingType_e)(istart[0] & 3);
+ size_t const blockSizeMax = ZSTD_blockSizeMax(dctx);
+
+ switch(litEncType)
+ {
+ case set_repeat:
+ DEBUGLOG(5, "set_repeat flag : re-using stats from previous compressed literals block");
+ RETURN_ERROR_IF(dctx->litEntropy==0, dictionary_corrupted, "");
+ ZSTD_FALLTHROUGH;
+
+ case set_compressed:
+ RETURN_ERROR_IF(srcSize < 5, corruption_detected, "srcSize >= MIN_CBLOCK_SIZE == 2; here we need up to 5 for case 3");
+ { size_t lhSize, litSize, litCSize;
+ U32 singleStream=0;
+ U32 const lhlCode = (istart[0] >> 2) & 3;
+ U32 const lhc = MEM_readLE32(istart);
+ size_t hufSuccess;
+ size_t expectedWriteSize = MIN(blockSizeMax, dstCapacity);
+ int const flags = 0
+ | (ZSTD_DCtx_get_bmi2(dctx) ? HUF_flags_bmi2 : 0)
+ | (dctx->disableHufAsm ? HUF_flags_disableAsm : 0);
+ switch(lhlCode)
+ {
+ case 0: case 1: default: /* note : default is impossible, since lhlCode into [0..3] */
+ /* 2 - 2 - 10 - 10 */
+ singleStream = !lhlCode;
+ lhSize = 3;
+ litSize = (lhc >> 4) & 0x3FF;
+ litCSize = (lhc >> 14) & 0x3FF;
+ break;
+ case 2:
+ /* 2 - 2 - 14 - 14 */
+ lhSize = 4;
+ litSize = (lhc >> 4) & 0x3FFF;
+ litCSize = lhc >> 18;
+ break;
+ case 3:
+ /* 2 - 2 - 18 - 18 */
+ lhSize = 5;
+ litSize = (lhc >> 4) & 0x3FFFF;
+ litCSize = (lhc >> 22) + ((size_t)istart[4] << 10);
+ break;
+ }
+ RETURN_ERROR_IF(litSize > 0 && dst == NULL, dstSize_tooSmall, "NULL not handled");
+ RETURN_ERROR_IF(litSize > blockSizeMax, corruption_detected, "");
+ if (!singleStream)
+ RETURN_ERROR_IF(litSize < MIN_LITERALS_FOR_4_STREAMS, literals_headerWrong,
+ "Not enough literals (%zu) for the 4-streams mode (min %u)",
+ litSize, MIN_LITERALS_FOR_4_STREAMS);
+ RETURN_ERROR_IF(litCSize + lhSize > srcSize, corruption_detected, "");
+ RETURN_ERROR_IF(expectedWriteSize < litSize , dstSize_tooSmall, "");
+ ZSTD_allocateLiteralsBuffer(dctx, dst, dstCapacity, litSize, streaming, expectedWriteSize, 0);
+
+ /* prefetch huffman table if cold */
+ if (dctx->ddictIsCold && (litSize > 768 /* heuristic */)) {
+ PREFETCH_AREA(dctx->HUFptr, sizeof(dctx->entropy.hufTable));
+ }
+
+ if (litEncType==set_repeat) {
+ if (singleStream) {
+ hufSuccess = HUF_decompress1X_usingDTable(
+ dctx->litBuffer, litSize, istart+lhSize, litCSize,
+ dctx->HUFptr, flags);
+ } else {
+ assert(litSize >= MIN_LITERALS_FOR_4_STREAMS);
+ hufSuccess = HUF_decompress4X_usingDTable(
+ dctx->litBuffer, litSize, istart+lhSize, litCSize,
+ dctx->HUFptr, flags);
+ }
+ } else {
+ if (singleStream) {
+#if defined(HUF_FORCE_DECOMPRESS_X2)
+ hufSuccess = HUF_decompress1X_DCtx_wksp(
+ dctx->entropy.hufTable, dctx->litBuffer, litSize,
+ istart+lhSize, litCSize, dctx->workspace,
+ sizeof(dctx->workspace), flags);
+#else
+ hufSuccess = HUF_decompress1X1_DCtx_wksp(
+ dctx->entropy.hufTable, dctx->litBuffer, litSize,
+ istart+lhSize, litCSize, dctx->workspace,
+ sizeof(dctx->workspace), flags);
+#endif
+ } else {
+ hufSuccess = HUF_decompress4X_hufOnly_wksp(
+ dctx->entropy.hufTable, dctx->litBuffer, litSize,
+ istart+lhSize, litCSize, dctx->workspace,
+ sizeof(dctx->workspace), flags);
+ }
+ }
+ if (dctx->litBufferLocation == ZSTD_split)
+ {
+ assert(litSize > ZSTD_LITBUFFEREXTRASIZE);
+ ZSTD_memcpy(dctx->litExtraBuffer, dctx->litBufferEnd - ZSTD_LITBUFFEREXTRASIZE, ZSTD_LITBUFFEREXTRASIZE);
+ ZSTD_memmove(dctx->litBuffer + ZSTD_LITBUFFEREXTRASIZE - WILDCOPY_OVERLENGTH, dctx->litBuffer, litSize - ZSTD_LITBUFFEREXTRASIZE);
+ dctx->litBuffer += ZSTD_LITBUFFEREXTRASIZE - WILDCOPY_OVERLENGTH;
+ dctx->litBufferEnd -= WILDCOPY_OVERLENGTH;
+ assert(dctx->litBufferEnd <= (BYTE*)dst + blockSizeMax);
+ }
+
+ RETURN_ERROR_IF(HUF_isError(hufSuccess), corruption_detected, "");
+
+ dctx->litPtr = dctx->litBuffer;
+ dctx->litSize = litSize;
+ dctx->litEntropy = 1;
+ if (litEncType==set_compressed) dctx->HUFptr = dctx->entropy.hufTable;
+ return litCSize + lhSize;
+ }
+
+ case set_basic:
+ { size_t litSize, lhSize;
+ U32 const lhlCode = ((istart[0]) >> 2) & 3;
+ size_t expectedWriteSize = MIN(blockSizeMax, dstCapacity);
+ switch(lhlCode)
+ {
+ case 0: case 2: default: /* note : default is impossible, since lhlCode into [0..3] */
+ lhSize = 1;
+ litSize = istart[0] >> 3;
+ break;
+ case 1:
+ lhSize = 2;
+ litSize = MEM_readLE16(istart) >> 4;
+ break;
+ case 3:
+ lhSize = 3;
+ RETURN_ERROR_IF(srcSize<3, corruption_detected, "srcSize >= MIN_CBLOCK_SIZE == 2; here we need lhSize = 3");
+ litSize = MEM_readLE24(istart) >> 4;
+ break;
+ }
+
+ RETURN_ERROR_IF(litSize > 0 && dst == NULL, dstSize_tooSmall, "NULL not handled");
+ RETURN_ERROR_IF(litSize > blockSizeMax, corruption_detected, "");
+ RETURN_ERROR_IF(expectedWriteSize < litSize, dstSize_tooSmall, "");
+ ZSTD_allocateLiteralsBuffer(dctx, dst, dstCapacity, litSize, streaming, expectedWriteSize, 1);
+ if (lhSize+litSize+WILDCOPY_OVERLENGTH > srcSize) { /* risk reading beyond src buffer with wildcopy */
+ RETURN_ERROR_IF(litSize+lhSize > srcSize, corruption_detected, "");
+ if (dctx->litBufferLocation == ZSTD_split)
+ {
+ ZSTD_memcpy(dctx->litBuffer, istart + lhSize, litSize - ZSTD_LITBUFFEREXTRASIZE);
+ ZSTD_memcpy(dctx->litExtraBuffer, istart + lhSize + litSize - ZSTD_LITBUFFEREXTRASIZE, ZSTD_LITBUFFEREXTRASIZE);
+ }
+ else
+ {
+ ZSTD_memcpy(dctx->litBuffer, istart + lhSize, litSize);
+ }
+ dctx->litPtr = dctx->litBuffer;
+ dctx->litSize = litSize;
+ return lhSize+litSize;
+ }
+ /* direct reference into compressed stream */
+ dctx->litPtr = istart+lhSize;
+ dctx->litSize = litSize;
+ dctx->litBufferEnd = dctx->litPtr + litSize;
+ dctx->litBufferLocation = ZSTD_not_in_dst;
+ return lhSize+litSize;
+ }
+
+ case set_rle:
+ { U32 const lhlCode = ((istart[0]) >> 2) & 3;
+ size_t litSize, lhSize;
+ size_t expectedWriteSize = MIN(blockSizeMax, dstCapacity);
+ switch(lhlCode)
+ {
+ case 0: case 2: default: /* note : default is impossible, since lhlCode into [0..3] */
+ lhSize = 1;
+ litSize = istart[0] >> 3;
+ break;
+ case 1:
+ lhSize = 2;
+ RETURN_ERROR_IF(srcSize<3, corruption_detected, "srcSize >= MIN_CBLOCK_SIZE == 2; here we need lhSize+1 = 3");
+ litSize = MEM_readLE16(istart) >> 4;
+ break;
+ case 3:
+ lhSize = 3;
+ RETURN_ERROR_IF(srcSize<4, corruption_detected, "srcSize >= MIN_CBLOCK_SIZE == 2; here we need lhSize+1 = 4");
+ litSize = MEM_readLE24(istart) >> 4;
+ break;
+ }
+ RETURN_ERROR_IF(litSize > 0 && dst == NULL, dstSize_tooSmall, "NULL not handled");
+ RETURN_ERROR_IF(litSize > blockSizeMax, corruption_detected, "");
+ RETURN_ERROR_IF(expectedWriteSize < litSize, dstSize_tooSmall, "");
+ ZSTD_allocateLiteralsBuffer(dctx, dst, dstCapacity, litSize, streaming, expectedWriteSize, 1);
+ if (dctx->litBufferLocation == ZSTD_split)
+ {
+ ZSTD_memset(dctx->litBuffer, istart[lhSize], litSize - ZSTD_LITBUFFEREXTRASIZE);
+ ZSTD_memset(dctx->litExtraBuffer, istart[lhSize], ZSTD_LITBUFFEREXTRASIZE);
+ }
+ else
+ {
+ ZSTD_memset(dctx->litBuffer, istart[lhSize], litSize);
+ }
+ dctx->litPtr = dctx->litBuffer;
+ dctx->litSize = litSize;
+ return lhSize+1;
+ }
+ default:
+ RETURN_ERROR(corruption_detected, "impossible");
+ }
+ }
+}
+
+/* Hidden declaration for fullbench */
+size_t ZSTD_decodeLiteralsBlock_wrapper(ZSTD_DCtx* dctx,
+ const void* src, size_t srcSize,
+ void* dst, size_t dstCapacity);
+size_t ZSTD_decodeLiteralsBlock_wrapper(ZSTD_DCtx* dctx,
+ const void* src, size_t srcSize,
+ void* dst, size_t dstCapacity)
+{
+ dctx->isFrameDecompression = 0;
+ return ZSTD_decodeLiteralsBlock(dctx, src, srcSize, dst, dstCapacity, not_streaming);
+}
+
+/* Default FSE distribution tables.
+ * These are pre-calculated FSE decoding tables using default distributions as defined in specification :
+ * https://github.com/facebook/zstd/blob/release/doc/zstd_compression_format.md#default-distributions
+ * They were generated programmatically with following method :
+ * - start from default distributions, present in /lib/common/zstd_internal.h
+ * - generate tables normally, using ZSTD_buildFSETable()
+ * - printout the content of tables
+ * - pretify output, report below, test with fuzzer to ensure it's correct */
+
+/* Default FSE distribution table for Literal Lengths */
+static const ZSTD_seqSymbol LL_defaultDTable[(1<<LL_DEFAULTNORMLOG)+1] = {
+ { 1, 1, 1, LL_DEFAULTNORMLOG}, /* header : fastMode, tableLog */
+ /* nextState, nbAddBits, nbBits, baseVal */
+ { 0, 0, 4, 0}, { 16, 0, 4, 0},
+ { 32, 0, 5, 1}, { 0, 0, 5, 3},
+ { 0, 0, 5, 4}, { 0, 0, 5, 6},
+ { 0, 0, 5, 7}, { 0, 0, 5, 9},
+ { 0, 0, 5, 10}, { 0, 0, 5, 12},
+ { 0, 0, 6, 14}, { 0, 1, 5, 16},
+ { 0, 1, 5, 20}, { 0, 1, 5, 22},
+ { 0, 2, 5, 28}, { 0, 3, 5, 32},
+ { 0, 4, 5, 48}, { 32, 6, 5, 64},
+ { 0, 7, 5, 128}, { 0, 8, 6, 256},
+ { 0, 10, 6, 1024}, { 0, 12, 6, 4096},
+ { 32, 0, 4, 0}, { 0, 0, 4, 1},
+ { 0, 0, 5, 2}, { 32, 0, 5, 4},
+ { 0, 0, 5, 5}, { 32, 0, 5, 7},
+ { 0, 0, 5, 8}, { 32, 0, 5, 10},
+ { 0, 0, 5, 11}, { 0, 0, 6, 13},
+ { 32, 1, 5, 16}, { 0, 1, 5, 18},
+ { 32, 1, 5, 22}, { 0, 2, 5, 24},
+ { 32, 3, 5, 32}, { 0, 3, 5, 40},
+ { 0, 6, 4, 64}, { 16, 6, 4, 64},
+ { 32, 7, 5, 128}, { 0, 9, 6, 512},
+ { 0, 11, 6, 2048}, { 48, 0, 4, 0},
+ { 16, 0, 4, 1}, { 32, 0, 5, 2},
+ { 32, 0, 5, 3}, { 32, 0, 5, 5},
+ { 32, 0, 5, 6}, { 32, 0, 5, 8},
+ { 32, 0, 5, 9}, { 32, 0, 5, 11},
+ { 32, 0, 5, 12}, { 0, 0, 6, 15},
+ { 32, 1, 5, 18}, { 32, 1, 5, 20},
+ { 32, 2, 5, 24}, { 32, 2, 5, 28},
+ { 32, 3, 5, 40}, { 32, 4, 5, 48},
+ { 0, 16, 6,65536}, { 0, 15, 6,32768},
+ { 0, 14, 6,16384}, { 0, 13, 6, 8192},
+}; /* LL_defaultDTable */
+
+/* Default FSE distribution table for Offset Codes */
+static const ZSTD_seqSymbol OF_defaultDTable[(1<<OF_DEFAULTNORMLOG)+1] = {
+ { 1, 1, 1, OF_DEFAULTNORMLOG}, /* header : fastMode, tableLog */
+ /* nextState, nbAddBits, nbBits, baseVal */
+ { 0, 0, 5, 0}, { 0, 6, 4, 61},
+ { 0, 9, 5, 509}, { 0, 15, 5,32765},
+ { 0, 21, 5,2097149}, { 0, 3, 5, 5},
+ { 0, 7, 4, 125}, { 0, 12, 5, 4093},
+ { 0, 18, 5,262141}, { 0, 23, 5,8388605},
+ { 0, 5, 5, 29}, { 0, 8, 4, 253},
+ { 0, 14, 5,16381}, { 0, 20, 5,1048573},
+ { 0, 2, 5, 1}, { 16, 7, 4, 125},
+ { 0, 11, 5, 2045}, { 0, 17, 5,131069},
+ { 0, 22, 5,4194301}, { 0, 4, 5, 13},
+ { 16, 8, 4, 253}, { 0, 13, 5, 8189},
+ { 0, 19, 5,524285}, { 0, 1, 5, 1},
+ { 16, 6, 4, 61}, { 0, 10, 5, 1021},
+ { 0, 16, 5,65533}, { 0, 28, 5,268435453},
+ { 0, 27, 5,134217725}, { 0, 26, 5,67108861},
+ { 0, 25, 5,33554429}, { 0, 24, 5,16777213},
+}; /* OF_defaultDTable */
+
+
+/* Default FSE distribution table for Match Lengths */
+static const ZSTD_seqSymbol ML_defaultDTable[(1<<ML_DEFAULTNORMLOG)+1] = {
+ { 1, 1, 1, ML_DEFAULTNORMLOG}, /* header : fastMode, tableLog */
+ /* nextState, nbAddBits, nbBits, baseVal */
+ { 0, 0, 6, 3}, { 0, 0, 4, 4},
+ { 32, 0, 5, 5}, { 0, 0, 5, 6},
+ { 0, 0, 5, 8}, { 0, 0, 5, 9},
+ { 0, 0, 5, 11}, { 0, 0, 6, 13},
+ { 0, 0, 6, 16}, { 0, 0, 6, 19},
+ { 0, 0, 6, 22}, { 0, 0, 6, 25},
+ { 0, 0, 6, 28}, { 0, 0, 6, 31},
+ { 0, 0, 6, 34}, { 0, 1, 6, 37},
+ { 0, 1, 6, 41}, { 0, 2, 6, 47},
+ { 0, 3, 6, 59}, { 0, 4, 6, 83},
+ { 0, 7, 6, 131}, { 0, 9, 6, 515},
+ { 16, 0, 4, 4}, { 0, 0, 4, 5},
+ { 32, 0, 5, 6}, { 0, 0, 5, 7},
+ { 32, 0, 5, 9}, { 0, 0, 5, 10},
+ { 0, 0, 6, 12}, { 0, 0, 6, 15},
+ { 0, 0, 6, 18}, { 0, 0, 6, 21},
+ { 0, 0, 6, 24}, { 0, 0, 6, 27},
+ { 0, 0, 6, 30}, { 0, 0, 6, 33},
+ { 0, 1, 6, 35}, { 0, 1, 6, 39},
+ { 0, 2, 6, 43}, { 0, 3, 6, 51},
+ { 0, 4, 6, 67}, { 0, 5, 6, 99},
+ { 0, 8, 6, 259}, { 32, 0, 4, 4},
+ { 48, 0, 4, 4}, { 16, 0, 4, 5},
+ { 32, 0, 5, 7}, { 32, 0, 5, 8},
+ { 32, 0, 5, 10}, { 32, 0, 5, 11},
+ { 0, 0, 6, 14}, { 0, 0, 6, 17},
+ { 0, 0, 6, 20}, { 0, 0, 6, 23},
+ { 0, 0, 6, 26}, { 0, 0, 6, 29},
+ { 0, 0, 6, 32}, { 0, 16, 6,65539},
+ { 0, 15, 6,32771}, { 0, 14, 6,16387},
+ { 0, 13, 6, 8195}, { 0, 12, 6, 4099},
+ { 0, 11, 6, 2051}, { 0, 10, 6, 1027},
+}; /* ML_defaultDTable */
+
+
+static void ZSTD_buildSeqTable_rle(ZSTD_seqSymbol* dt, U32 baseValue, U8 nbAddBits)
+{
+ void* ptr = dt;
+ ZSTD_seqSymbol_header* const DTableH = (ZSTD_seqSymbol_header*)ptr;
+ ZSTD_seqSymbol* const cell = dt + 1;
+
+ DTableH->tableLog = 0;
+ DTableH->fastMode = 0;
+
+ cell->nbBits = 0;
+ cell->nextState = 0;
+ assert(nbAddBits < 255);
+ cell->nbAdditionalBits = nbAddBits;
+ cell->baseValue = baseValue;
+}
+
+
+/* ZSTD_buildFSETable() :
+ * generate FSE decoding table for one symbol (ll, ml or off)
+ * cannot fail if input is valid =>
+ * all inputs are presumed validated at this stage */
+FORCE_INLINE_TEMPLATE
+void ZSTD_buildFSETable_body(ZSTD_seqSymbol* dt,
+ const short* normalizedCounter, unsigned maxSymbolValue,
+ const U32* baseValue, const U8* nbAdditionalBits,
+ unsigned tableLog, void* wksp, size_t wkspSize)
+{
+ ZSTD_seqSymbol* const tableDecode = dt+1;
+ U32 const maxSV1 = maxSymbolValue + 1;
+ U32 const tableSize = 1 << tableLog;
+
+ U16* symbolNext = (U16*)wksp;
+ BYTE* spread = (BYTE*)(symbolNext + MaxSeq + 1);
+ U32 highThreshold = tableSize - 1;
+
+
+ /* Sanity Checks */
+ assert(maxSymbolValue <= MaxSeq);
+ assert(tableLog <= MaxFSELog);
+ assert(wkspSize >= ZSTD_BUILD_FSE_TABLE_WKSP_SIZE);
+ (void)wkspSize;
+ /* Init, lay down lowprob symbols */
+ { ZSTD_seqSymbol_header DTableH;
+ DTableH.tableLog = tableLog;
+ DTableH.fastMode = 1;
+ { S16 const largeLimit= (S16)(1 << (tableLog-1));
+ U32 s;
+ for (s=0; s<maxSV1; s++) {
+ if (normalizedCounter[s]==-1) {
+ tableDecode[highThreshold--].baseValue = s;
+ symbolNext[s] = 1;
+ } else {
+ if (normalizedCounter[s] >= largeLimit) DTableH.fastMode=0;
+ assert(normalizedCounter[s]>=0);
+ symbolNext[s] = (U16)normalizedCounter[s];
+ } } }
+ ZSTD_memcpy(dt, &DTableH, sizeof(DTableH));
+ }
+
+ /* Spread symbols */
+ assert(tableSize <= 512);
+ /* Specialized symbol spreading for the case when there are
+ * no low probability (-1 count) symbols. When compressing
+ * small blocks we avoid low probability symbols to hit this
+ * case, since header decoding speed matters more.
+ */
+ if (highThreshold == tableSize - 1) {
+ size_t const tableMask = tableSize-1;
+ size_t const step = FSE_TABLESTEP(tableSize);
+ /* First lay down the symbols in order.
+ * We use a uint64_t to lay down 8 bytes at a time. This reduces branch
+ * misses since small blocks generally have small table logs, so nearly
+ * all symbols have counts <= 8. We ensure we have 8 bytes at the end of
+ * our buffer to handle the over-write.
+ */
+ {
+ U64 const add = 0x0101010101010101ull;
+ size_t pos = 0;
+ U64 sv = 0;
+ U32 s;
+ for (s=0; s<maxSV1; ++s, sv += add) {
+ int i;
+ int const n = normalizedCounter[s];
+ MEM_write64(spread + pos, sv);
+ for (i = 8; i < n; i += 8) {
+ MEM_write64(spread + pos + i, sv);
+ }
+ assert(n>=0);
+ pos += (size_t)n;
+ }
+ }
+ /* Now we spread those positions across the table.
+ * The benefit of doing it in two stages is that we avoid the
+ * variable size inner loop, which caused lots of branch misses.
+ * Now we can run through all the positions without any branch misses.
+ * We unroll the loop twice, since that is what empirically worked best.
+ */
+ {
+ size_t position = 0;
+ size_t s;
+ size_t const unroll = 2;
+ assert(tableSize % unroll == 0); /* FSE_MIN_TABLELOG is 5 */
+ for (s = 0; s < (size_t)tableSize; s += unroll) {
+ size_t u;
+ for (u = 0; u < unroll; ++u) {
+ size_t const uPosition = (position + (u * step)) & tableMask;
+ tableDecode[uPosition].baseValue = spread[s + u];
+ }
+ position = (position + (unroll * step)) & tableMask;
+ }
+ assert(position == 0);
+ }
+ } else {
+ U32 const tableMask = tableSize-1;
+ U32 const step = FSE_TABLESTEP(tableSize);
+ U32 s, position = 0;
+ for (s=0; s<maxSV1; s++) {
+ int i;
+ int const n = normalizedCounter[s];
+ for (i=0; i<n; i++) {
+ tableDecode[position].baseValue = s;
+ position = (position + step) & tableMask;
+ while (UNLIKELY(position > highThreshold)) position = (position + step) & tableMask; /* lowprob area */
+ } }
+ assert(position == 0); /* position must reach all cells once, otherwise normalizedCounter is incorrect */
+ }
+
+ /* Build Decoding table */
+ {
+ U32 u;
+ for (u=0; u<tableSize; u++) {
+ U32 const symbol = tableDecode[u].baseValue;
+ U32 const nextState = symbolNext[symbol]++;
+ tableDecode[u].nbBits = (BYTE) (tableLog - ZSTD_highbit32(nextState) );
+ tableDecode[u].nextState = (U16) ( (nextState << tableDecode[u].nbBits) - tableSize);
+ assert(nbAdditionalBits[symbol] < 255);
+ tableDecode[u].nbAdditionalBits = nbAdditionalBits[symbol];
+ tableDecode[u].baseValue = baseValue[symbol];
+ }
+ }
+}
+
+/* Avoids the FORCE_INLINE of the _body() function. */
+static void ZSTD_buildFSETable_body_default(ZSTD_seqSymbol* dt,
+ const short* normalizedCounter, unsigned maxSymbolValue,
+ const U32* baseValue, const U8* nbAdditionalBits,
+ unsigned tableLog, void* wksp, size_t wkspSize)
+{
+ ZSTD_buildFSETable_body(dt, normalizedCounter, maxSymbolValue,
+ baseValue, nbAdditionalBits, tableLog, wksp, wkspSize);
+}
+
+#if DYNAMIC_BMI2
+BMI2_TARGET_ATTRIBUTE static void ZSTD_buildFSETable_body_bmi2(ZSTD_seqSymbol* dt,
+ const short* normalizedCounter, unsigned maxSymbolValue,
+ const U32* baseValue, const U8* nbAdditionalBits,
+ unsigned tableLog, void* wksp, size_t wkspSize)
+{
+ ZSTD_buildFSETable_body(dt, normalizedCounter, maxSymbolValue,
+ baseValue, nbAdditionalBits, tableLog, wksp, wkspSize);
+}
+#endif
+
+void ZSTD_buildFSETable(ZSTD_seqSymbol* dt,
+ const short* normalizedCounter, unsigned maxSymbolValue,
+ const U32* baseValue, const U8* nbAdditionalBits,
+ unsigned tableLog, void* wksp, size_t wkspSize, int bmi2)
+{
+#if DYNAMIC_BMI2
+ if (bmi2) {
+ ZSTD_buildFSETable_body_bmi2(dt, normalizedCounter, maxSymbolValue,
+ baseValue, nbAdditionalBits, tableLog, wksp, wkspSize);
+ return;
+ }
+#endif
+ (void)bmi2;
+ ZSTD_buildFSETable_body_default(dt, normalizedCounter, maxSymbolValue,
+ baseValue, nbAdditionalBits, tableLog, wksp, wkspSize);
+}
+
+
+/*! ZSTD_buildSeqTable() :
+ * @return : nb bytes read from src,
+ * or an error code if it fails */
+static size_t ZSTD_buildSeqTable(ZSTD_seqSymbol* DTableSpace, const ZSTD_seqSymbol** DTablePtr,
+ symbolEncodingType_e type, unsigned max, U32 maxLog,
+ const void* src, size_t srcSize,
+ const U32* baseValue, const U8* nbAdditionalBits,
+ const ZSTD_seqSymbol* defaultTable, U32 flagRepeatTable,
+ int ddictIsCold, int nbSeq, U32* wksp, size_t wkspSize,
+ int bmi2)
+{
+ switch(type)
+ {
+ case set_rle :
+ RETURN_ERROR_IF(!srcSize, srcSize_wrong, "");
+ RETURN_ERROR_IF((*(const BYTE*)src) > max, corruption_detected, "");
+ { U32 const symbol = *(const BYTE*)src;
+ U32 const baseline = baseValue[symbol];
+ U8 const nbBits = nbAdditionalBits[symbol];
+ ZSTD_buildSeqTable_rle(DTableSpace, baseline, nbBits);
+ }
+ *DTablePtr = DTableSpace;
+ return 1;
+ case set_basic :
+ *DTablePtr = defaultTable;
+ return 0;
+ case set_repeat:
+ RETURN_ERROR_IF(!flagRepeatTable, corruption_detected, "");
+ /* prefetch FSE table if used */
+ if (ddictIsCold && (nbSeq > 24 /* heuristic */)) {
+ const void* const pStart = *DTablePtr;
+ size_t const pSize = sizeof(ZSTD_seqSymbol) * (SEQSYMBOL_TABLE_SIZE(maxLog));
+ PREFETCH_AREA(pStart, pSize);
+ }
+ return 0;
+ case set_compressed :
+ { unsigned tableLog;
+ S16 norm[MaxSeq+1];
+ size_t const headerSize = FSE_readNCount(norm, &max, &tableLog, src, srcSize);
+ RETURN_ERROR_IF(FSE_isError(headerSize), corruption_detected, "");
+ RETURN_ERROR_IF(tableLog > maxLog, corruption_detected, "");
+ ZSTD_buildFSETable(DTableSpace, norm, max, baseValue, nbAdditionalBits, tableLog, wksp, wkspSize, bmi2);
+ *DTablePtr = DTableSpace;
+ return headerSize;
+ }
+ default :
+ assert(0);
+ RETURN_ERROR(GENERIC, "impossible");
+ }
+}
+
+size_t ZSTD_decodeSeqHeaders(ZSTD_DCtx* dctx, int* nbSeqPtr,
+ const void* src, size_t srcSize)
+{
+ const BYTE* const istart = (const BYTE*)src;
+ const BYTE* const iend = istart + srcSize;
+ const BYTE* ip = istart;
+ int nbSeq;
+ DEBUGLOG(5, "ZSTD_decodeSeqHeaders");
+
+ /* check */
+ RETURN_ERROR_IF(srcSize < MIN_SEQUENCES_SIZE, srcSize_wrong, "");
+
+ /* SeqHead */
+ nbSeq = *ip++;
+ if (nbSeq > 0x7F) {
+ if (nbSeq == 0xFF) {
+ RETURN_ERROR_IF(ip+2 > iend, srcSize_wrong, "");
+ nbSeq = MEM_readLE16(ip) + LONGNBSEQ;
+ ip+=2;
+ } else {
+ RETURN_ERROR_IF(ip >= iend, srcSize_wrong, "");
+ nbSeq = ((nbSeq-0x80)<<8) + *ip++;
+ }
+ }
+ *nbSeqPtr = nbSeq;
+
+ if (nbSeq == 0) {
+ /* No sequence : section ends immediately */
+ RETURN_ERROR_IF(ip != iend, corruption_detected,
+ "extraneous data present in the Sequences section");
+ return (size_t)(ip - istart);
+ }
+
+ /* FSE table descriptors */
+ RETURN_ERROR_IF(ip+1 > iend, srcSize_wrong, ""); /* minimum possible size: 1 byte for symbol encoding types */
+ RETURN_ERROR_IF(*ip & 3, corruption_detected, ""); /* The last field, Reserved, must be all-zeroes. */
+ { symbolEncodingType_e const LLtype = (symbolEncodingType_e)(*ip >> 6);
+ symbolEncodingType_e const OFtype = (symbolEncodingType_e)((*ip >> 4) & 3);
+ symbolEncodingType_e const MLtype = (symbolEncodingType_e)((*ip >> 2) & 3);
+ ip++;
+
+ /* Build DTables */
+ { size_t const llhSize = ZSTD_buildSeqTable(dctx->entropy.LLTable, &dctx->LLTptr,
+ LLtype, MaxLL, LLFSELog,
+ ip, iend-ip,
+ LL_base, LL_bits,
+ LL_defaultDTable, dctx->fseEntropy,
+ dctx->ddictIsCold, nbSeq,
+ dctx->workspace, sizeof(dctx->workspace),
+ ZSTD_DCtx_get_bmi2(dctx));
+ RETURN_ERROR_IF(ZSTD_isError(llhSize), corruption_detected, "ZSTD_buildSeqTable failed");
+ ip += llhSize;
+ }
+
+ { size_t const ofhSize = ZSTD_buildSeqTable(dctx->entropy.OFTable, &dctx->OFTptr,
+ OFtype, MaxOff, OffFSELog,
+ ip, iend-ip,
+ OF_base, OF_bits,
+ OF_defaultDTable, dctx->fseEntropy,
+ dctx->ddictIsCold, nbSeq,
+ dctx->workspace, sizeof(dctx->workspace),
+ ZSTD_DCtx_get_bmi2(dctx));
+ RETURN_ERROR_IF(ZSTD_isError(ofhSize), corruption_detected, "ZSTD_buildSeqTable failed");
+ ip += ofhSize;
+ }
+
+ { size_t const mlhSize = ZSTD_buildSeqTable(dctx->entropy.MLTable, &dctx->MLTptr,
+ MLtype, MaxML, MLFSELog,
+ ip, iend-ip,
+ ML_base, ML_bits,
+ ML_defaultDTable, dctx->fseEntropy,
+ dctx->ddictIsCold, nbSeq,
+ dctx->workspace, sizeof(dctx->workspace),
+ ZSTD_DCtx_get_bmi2(dctx));
+ RETURN_ERROR_IF(ZSTD_isError(mlhSize), corruption_detected, "ZSTD_buildSeqTable failed");
+ ip += mlhSize;
+ }
+ }
+
+ return ip-istart;
+}
+
+
+typedef struct {
+ size_t litLength;
+ size_t matchLength;
+ size_t offset;
+} seq_t;
+
+typedef struct {
+ size_t state;
+ const ZSTD_seqSymbol* table;
+} ZSTD_fseState;
+
+typedef struct {
+ BIT_DStream_t DStream;
+ ZSTD_fseState stateLL;
+ ZSTD_fseState stateOffb;
+ ZSTD_fseState stateML;
+ size_t prevOffset[ZSTD_REP_NUM];
+} seqState_t;
+
+/*! ZSTD_overlapCopy8() :
+ * Copies 8 bytes from ip to op and updates op and ip where ip <= op.
+ * If the offset is < 8 then the offset is spread to at least 8 bytes.
+ *
+ * Precondition: *ip <= *op
+ * Postcondition: *op - *op >= 8
+ */
+HINT_INLINE void ZSTD_overlapCopy8(BYTE** op, BYTE const** ip, size_t offset) {
+ assert(*ip <= *op);
+ if (offset < 8) {
+ /* close range match, overlap */
+ static const U32 dec32table[] = { 0, 1, 2, 1, 4, 4, 4, 4 }; /* added */
+ static const int dec64table[] = { 8, 8, 8, 7, 8, 9,10,11 }; /* subtracted */
+ int const sub2 = dec64table[offset];
+ (*op)[0] = (*ip)[0];
+ (*op)[1] = (*ip)[1];
+ (*op)[2] = (*ip)[2];
+ (*op)[3] = (*ip)[3];
+ *ip += dec32table[offset];
+ ZSTD_copy4(*op+4, *ip);
+ *ip -= sub2;
+ } else {
+ ZSTD_copy8(*op, *ip);
+ }
+ *ip += 8;
+ *op += 8;
+ assert(*op - *ip >= 8);
+}
+
+/*! ZSTD_safecopy() :
+ * Specialized version of memcpy() that is allowed to READ up to WILDCOPY_OVERLENGTH past the input buffer
+ * and write up to 16 bytes past oend_w (op >= oend_w is allowed).
+ * This function is only called in the uncommon case where the sequence is near the end of the block. It
+ * should be fast for a single long sequence, but can be slow for several short sequences.
+ *
+ * @param ovtype controls the overlap detection
+ * - ZSTD_no_overlap: The source and destination are guaranteed to be at least WILDCOPY_VECLEN bytes apart.
+ * - ZSTD_overlap_src_before_dst: The src and dst may overlap and may be any distance apart.
+ * The src buffer must be before the dst buffer.
+ */
+static void ZSTD_safecopy(BYTE* op, const BYTE* const oend_w, BYTE const* ip, ptrdiff_t length, ZSTD_overlap_e ovtype) {
+ ptrdiff_t const diff = op - ip;
+ BYTE* const oend = op + length;
+
+ assert((ovtype == ZSTD_no_overlap && (diff <= -8 || diff >= 8 || op >= oend_w)) ||
+ (ovtype == ZSTD_overlap_src_before_dst && diff >= 0));
+
+ if (length < 8) {
+ /* Handle short lengths. */
+ while (op < oend) *op++ = *ip++;
+ return;
+ }
+ if (ovtype == ZSTD_overlap_src_before_dst) {
+ /* Copy 8 bytes and ensure the offset >= 8 when there can be overlap. */
+ assert(length >= 8);
+ ZSTD_overlapCopy8(&op, &ip, diff);
+ length -= 8;
+ assert(op - ip >= 8);
+ assert(op <= oend);
+ }
+
+ if (oend <= oend_w) {
+ /* No risk of overwrite. */
+ ZSTD_wildcopy(op, ip, length, ovtype);
+ return;
+ }
+ if (op <= oend_w) {
+ /* Wildcopy until we get close to the end. */
+ assert(oend > oend_w);
+ ZSTD_wildcopy(op, ip, oend_w - op, ovtype);
+ ip += oend_w - op;
+ op += oend_w - op;
+ }
+ /* Handle the leftovers. */
+ while (op < oend) *op++ = *ip++;
+}
+
+/* ZSTD_safecopyDstBeforeSrc():
+ * This version allows overlap with dst before src, or handles the non-overlap case with dst after src
+ * Kept separate from more common ZSTD_safecopy case to avoid performance impact to the safecopy common case */
+static void ZSTD_safecopyDstBeforeSrc(BYTE* op, const BYTE* ip, ptrdiff_t length) {
+ ptrdiff_t const diff = op - ip;
+ BYTE* const oend = op + length;
+
+ if (length < 8 || diff > -8) {
+ /* Handle short lengths, close overlaps, and dst not before src. */
+ while (op < oend) *op++ = *ip++;
+ return;
+ }
+
+ if (op <= oend - WILDCOPY_OVERLENGTH && diff < -WILDCOPY_VECLEN) {
+ ZSTD_wildcopy(op, ip, oend - WILDCOPY_OVERLENGTH - op, ZSTD_no_overlap);
+ ip += oend - WILDCOPY_OVERLENGTH - op;
+ op += oend - WILDCOPY_OVERLENGTH - op;
+ }
+
+ /* Handle the leftovers. */
+ while (op < oend) *op++ = *ip++;
+}
+
+/* ZSTD_execSequenceEnd():
+ * This version handles cases that are near the end of the output buffer. It requires
+ * more careful checks to make sure there is no overflow. By separating out these hard
+ * and unlikely cases, we can speed up the common cases.
+ *
+ * NOTE: This function needs to be fast for a single long sequence, but doesn't need
+ * to be optimized for many small sequences, since those fall into ZSTD_execSequence().
+ */
+FORCE_NOINLINE
+ZSTD_ALLOW_POINTER_OVERFLOW_ATTR
+size_t ZSTD_execSequenceEnd(BYTE* op,
+ BYTE* const oend, seq_t sequence,
+ const BYTE** litPtr, const BYTE* const litLimit,
+ const BYTE* const prefixStart, const BYTE* const virtualStart, const BYTE* const dictEnd)
+{
+ BYTE* const oLitEnd = op + sequence.litLength;
+ size_t const sequenceLength = sequence.litLength + sequence.matchLength;
+ const BYTE* const iLitEnd = *litPtr + sequence.litLength;
+ const BYTE* match = oLitEnd - sequence.offset;
+ BYTE* const oend_w = oend - WILDCOPY_OVERLENGTH;
+
+ /* bounds checks : careful of address space overflow in 32-bit mode */
+ RETURN_ERROR_IF(sequenceLength > (size_t)(oend - op), dstSize_tooSmall, "last match must fit within dstBuffer");
+ RETURN_ERROR_IF(sequence.litLength > (size_t)(litLimit - *litPtr), corruption_detected, "try to read beyond literal buffer");
+ assert(op < op + sequenceLength);
+ assert(oLitEnd < op + sequenceLength);
+
+ /* copy literals */
+ ZSTD_safecopy(op, oend_w, *litPtr, sequence.litLength, ZSTD_no_overlap);
+ op = oLitEnd;
+ *litPtr = iLitEnd;
+
+ /* copy Match */
+ if (sequence.offset > (size_t)(oLitEnd - prefixStart)) {
+ /* offset beyond prefix */
+ RETURN_ERROR_IF(sequence.offset > (size_t)(oLitEnd - virtualStart), corruption_detected, "");
+ match = dictEnd - (prefixStart - match);
+ if (match + sequence.matchLength <= dictEnd) {
+ ZSTD_memmove(oLitEnd, match, sequence.matchLength);
+ return sequenceLength;
+ }
+ /* span extDict & currentPrefixSegment */
+ { size_t const length1 = dictEnd - match;
+ ZSTD_memmove(oLitEnd, match, length1);
+ op = oLitEnd + length1;
+ sequence.matchLength -= length1;
+ match = prefixStart;
+ }
+ }
+ ZSTD_safecopy(op, oend_w, match, sequence.matchLength, ZSTD_overlap_src_before_dst);
+ return sequenceLength;
+}
+
+/* ZSTD_execSequenceEndSplitLitBuffer():
+ * 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.
+ */
+FORCE_NOINLINE
+ZSTD_ALLOW_POINTER_OVERFLOW_ATTR
+size_t ZSTD_execSequenceEndSplitLitBuffer(BYTE* op,
+ BYTE* const oend, const BYTE* const oend_w, seq_t sequence,
+ const BYTE** litPtr, const BYTE* const litLimit,
+ const BYTE* const prefixStart, const BYTE* const virtualStart, const BYTE* const dictEnd)
+{
+ BYTE* const oLitEnd = op + sequence.litLength;
+ size_t const sequenceLength = sequence.litLength + sequence.matchLength;
+ const BYTE* const iLitEnd = *litPtr + sequence.litLength;
+ const BYTE* match = oLitEnd - sequence.offset;
+
+
+ /* bounds checks : careful of address space overflow in 32-bit mode */
+ RETURN_ERROR_IF(sequenceLength > (size_t)(oend - op), dstSize_tooSmall, "last match must fit within dstBuffer");
+ RETURN_ERROR_IF(sequence.litLength > (size_t)(litLimit - *litPtr), corruption_detected, "try to read beyond literal buffer");
+ assert(op < op + sequenceLength);
+ assert(oLitEnd < op + sequenceLength);
+
+ /* copy literals */
+ RETURN_ERROR_IF(op > *litPtr && op < *litPtr + sequence.litLength, dstSize_tooSmall, "output should not catch up to and overwrite literal buffer");
+ ZSTD_safecopyDstBeforeSrc(op, *litPtr, sequence.litLength);
+ op = oLitEnd;
+ *litPtr = iLitEnd;
+
+ /* copy Match */
+ if (sequence.offset > (size_t)(oLitEnd - prefixStart)) {
+ /* offset beyond prefix */
+ RETURN_ERROR_IF(sequence.offset > (size_t)(oLitEnd - virtualStart), corruption_detected, "");
+ match = dictEnd - (prefixStart - match);
+ if (match + sequence.matchLength <= dictEnd) {
+ ZSTD_memmove(oLitEnd, match, sequence.matchLength);
+ return sequenceLength;
+ }
+ /* span extDict & currentPrefixSegment */
+ { size_t const length1 = dictEnd - match;
+ ZSTD_memmove(oLitEnd, match, length1);
+ op = oLitEnd + length1;
+ sequence.matchLength -= length1;
+ match = prefixStart;
+ }
+ }
+ ZSTD_safecopy(op, oend_w, match, sequence.matchLength, ZSTD_overlap_src_before_dst);
+ return sequenceLength;
+}
+
+HINT_INLINE
+ZSTD_ALLOW_POINTER_OVERFLOW_ATTR
+size_t ZSTD_execSequence(BYTE* op,
+ BYTE* const oend, seq_t sequence,
+ const BYTE** litPtr, const BYTE* const litLimit,
+ const BYTE* const prefixStart, const BYTE* const virtualStart, const BYTE* const dictEnd)
+{
+ BYTE* const oLitEnd = op + sequence.litLength;
+ size_t const sequenceLength = sequence.litLength + sequence.matchLength;
+ BYTE* const oMatchEnd = op + sequenceLength; /* risk : address space overflow (32-bits) */
+ BYTE* const oend_w = oend - WILDCOPY_OVERLENGTH; /* risk : address space underflow on oend=NULL */
+ const BYTE* const iLitEnd = *litPtr + sequence.litLength;
+ const BYTE* match = oLitEnd - sequence.offset;
+
+ assert(op != NULL /* Precondition */);
+ assert(oend_w < oend /* No underflow */);
+
+#if defined(__aarch64__)
+ /* prefetch sequence starting from match that will be used for copy later */
+ PREFETCH_L1(match);
+#endif
+ /* Handle edge cases in a slow path:
+ * - Read beyond end of literals
+ * - Match end is within WILDCOPY_OVERLIMIT of oend
+ * - 32-bit mode and the match length overflows
+ */
+ if (UNLIKELY(
+ iLitEnd > litLimit ||
+ oMatchEnd > oend_w ||
+ (MEM_32bits() && (size_t)(oend - op) < sequenceLength + WILDCOPY_OVERLENGTH)))
+ return ZSTD_execSequenceEnd(op, oend, sequence, litPtr, litLimit, prefixStart, virtualStart, dictEnd);
+
+ /* Assumptions (everything else goes into ZSTD_execSequenceEnd()) */
+ assert(op <= oLitEnd /* No overflow */);
+ assert(oLitEnd < oMatchEnd /* Non-zero match & no overflow */);
+ assert(oMatchEnd <= oend /* No underflow */);
+ assert(iLitEnd <= litLimit /* Literal length is in bounds */);
+ assert(oLitEnd <= oend_w /* Can wildcopy literals */);
+ assert(oMatchEnd <= oend_w /* Can wildcopy matches */);
+
+ /* Copy Literals:
+ * Split out litLength <= 16 since it is nearly always true. +1.6% on gcc-9.
+ * We likely don't need the full 32-byte wildcopy.
+ */
+ assert(WILDCOPY_OVERLENGTH >= 16);
+ ZSTD_copy16(op, (*litPtr));
+ if (UNLIKELY(sequence.litLength > 16)) {
+ ZSTD_wildcopy(op + 16, (*litPtr) + 16, sequence.litLength - 16, ZSTD_no_overlap);
+ }
+ op = oLitEnd;
+ *litPtr = iLitEnd; /* update for next sequence */
+
+ /* Copy Match */
+ if (sequence.offset > (size_t)(oLitEnd - prefixStart)) {
+ /* offset beyond prefix -> go into extDict */
+ RETURN_ERROR_IF(UNLIKELY(sequence.offset > (size_t)(oLitEnd - virtualStart)), corruption_detected, "");
+ match = dictEnd + (match - prefixStart);
+ if (match + sequence.matchLength <= dictEnd) {
+ ZSTD_memmove(oLitEnd, match, sequence.matchLength);
+ return sequenceLength;
+ }
+ /* span extDict & currentPrefixSegment */
+ { size_t const length1 = dictEnd - match;
+ ZSTD_memmove(oLitEnd, match, length1);
+ op = oLitEnd + length1;
+ sequence.matchLength -= length1;
+ match = prefixStart;
+ }
+ }
+ /* Match within prefix of 1 or more bytes */
+ assert(op <= oMatchEnd);
+ assert(oMatchEnd <= oend_w);
+ assert(match >= prefixStart);
+ assert(sequence.matchLength >= 1);
+
+ /* Nearly all offsets are >= WILDCOPY_VECLEN bytes, which means we can use wildcopy
+ * without overlap checking.
+ */
+ if (LIKELY(sequence.offset >= WILDCOPY_VECLEN)) {
+ /* We bet on a full wildcopy for matches, since we expect matches to be
+ * longer than literals (in general). In silesia, ~10% of matches are longer
+ * than 16 bytes.
+ */
+ ZSTD_wildcopy(op, match, (ptrdiff_t)sequence.matchLength, ZSTD_no_overlap);
+ return sequenceLength;
+ }
+ assert(sequence.offset < WILDCOPY_VECLEN);
+
+ /* Copy 8 bytes and spread the offset to be >= 8. */
+ ZSTD_overlapCopy8(&op, &match, sequence.offset);
+
+ /* If the match length is > 8 bytes, then continue with the wildcopy. */
+ if (sequence.matchLength > 8) {
+ assert(op < oMatchEnd);
+ ZSTD_wildcopy(op, match, (ptrdiff_t)sequence.matchLength - 8, ZSTD_overlap_src_before_dst);
+ }
+ return sequenceLength;
+}
+
+HINT_INLINE
+ZSTD_ALLOW_POINTER_OVERFLOW_ATTR
+size_t ZSTD_execSequenceSplitLitBuffer(BYTE* op,
+ BYTE* const oend, const BYTE* const oend_w, seq_t sequence,
+ const BYTE** litPtr, const BYTE* const litLimit,
+ const BYTE* const prefixStart, const BYTE* const virtualStart, const BYTE* const dictEnd)
+{
+ BYTE* const oLitEnd = op + sequence.litLength;
+ size_t const sequenceLength = sequence.litLength + sequence.matchLength;
+ BYTE* const oMatchEnd = op + sequenceLength; /* risk : address space overflow (32-bits) */
+ const BYTE* const iLitEnd = *litPtr + sequence.litLength;
+ const BYTE* match = oLitEnd - sequence.offset;
+
+ assert(op != NULL /* Precondition */);
+ assert(oend_w < oend /* No underflow */);
+ /* Handle edge cases in a slow path:
+ * - Read beyond end of literals
+ * - Match end is within WILDCOPY_OVERLIMIT of oend
+ * - 32-bit mode and the match length overflows
+ */
+ if (UNLIKELY(
+ iLitEnd > litLimit ||
+ oMatchEnd > oend_w ||
+ (MEM_32bits() && (size_t)(oend - op) < sequenceLength + WILDCOPY_OVERLENGTH)))
+ return ZSTD_execSequenceEndSplitLitBuffer(op, oend, oend_w, sequence, litPtr, litLimit, prefixStart, virtualStart, dictEnd);
+
+ /* Assumptions (everything else goes into ZSTD_execSequenceEnd()) */
+ assert(op <= oLitEnd /* No overflow */);
+ assert(oLitEnd < oMatchEnd /* Non-zero match & no overflow */);
+ assert(oMatchEnd <= oend /* No underflow */);
+ assert(iLitEnd <= litLimit /* Literal length is in bounds */);
+ assert(oLitEnd <= oend_w /* Can wildcopy literals */);
+ assert(oMatchEnd <= oend_w /* Can wildcopy matches */);
+
+ /* Copy Literals:
+ * Split out litLength <= 16 since it is nearly always true. +1.6% on gcc-9.
+ * We likely don't need the full 32-byte wildcopy.
+ */
+ assert(WILDCOPY_OVERLENGTH >= 16);
+ ZSTD_copy16(op, (*litPtr));
+ if (UNLIKELY(sequence.litLength > 16)) {
+ ZSTD_wildcopy(op+16, (*litPtr)+16, sequence.litLength-16, ZSTD_no_overlap);
+ }
+ op = oLitEnd;
+ *litPtr = iLitEnd; /* update for next sequence */
+
+ /* Copy Match */
+ if (sequence.offset > (size_t)(oLitEnd - prefixStart)) {
+ /* offset beyond prefix -> go into extDict */
+ RETURN_ERROR_IF(UNLIKELY(sequence.offset > (size_t)(oLitEnd - virtualStart)), corruption_detected, "");
+ match = dictEnd + (match - prefixStart);
+ if (match + sequence.matchLength <= dictEnd) {
+ ZSTD_memmove(oLitEnd, match, sequence.matchLength);
+ return sequenceLength;
+ }
+ /* span extDict & currentPrefixSegment */
+ { size_t const length1 = dictEnd - match;
+ ZSTD_memmove(oLitEnd, match, length1);
+ op = oLitEnd + length1;
+ sequence.matchLength -= length1;
+ match = prefixStart;
+ } }
+ /* Match within prefix of 1 or more bytes */
+ assert(op <= oMatchEnd);
+ assert(oMatchEnd <= oend_w);
+ assert(match >= prefixStart);
+ assert(sequence.matchLength >= 1);
+
+ /* Nearly all offsets are >= WILDCOPY_VECLEN bytes, which means we can use wildcopy
+ * without overlap checking.
+ */
+ if (LIKELY(sequence.offset >= WILDCOPY_VECLEN)) {
+ /* We bet on a full wildcopy for matches, since we expect matches to be
+ * longer than literals (in general). In silesia, ~10% of matches are longer
+ * than 16 bytes.
+ */
+ ZSTD_wildcopy(op, match, (ptrdiff_t)sequence.matchLength, ZSTD_no_overlap);
+ return sequenceLength;
+ }
+ assert(sequence.offset < WILDCOPY_VECLEN);
+
+ /* Copy 8 bytes and spread the offset to be >= 8. */
+ ZSTD_overlapCopy8(&op, &match, sequence.offset);
+
+ /* If the match length is > 8 bytes, then continue with the wildcopy. */
+ if (sequence.matchLength > 8) {
+ assert(op < oMatchEnd);
+ ZSTD_wildcopy(op, match, (ptrdiff_t)sequence.matchLength-8, ZSTD_overlap_src_before_dst);
+ }
+ return sequenceLength;
+}
+
+
+static void
+ZSTD_initFseState(ZSTD_fseState* DStatePtr, BIT_DStream_t* bitD, const ZSTD_seqSymbol* dt)
+{
+ const void* ptr = dt;
+ const ZSTD_seqSymbol_header* const DTableH = (const ZSTD_seqSymbol_header*)ptr;
+ DStatePtr->state = BIT_readBits(bitD, DTableH->tableLog);
+ DEBUGLOG(6, "ZSTD_initFseState : val=%u using %u bits",
+ (U32)DStatePtr->state, DTableH->tableLog);
+ BIT_reloadDStream(bitD);
+ DStatePtr->table = dt + 1;
+}
+
+FORCE_INLINE_TEMPLATE void
+ZSTD_updateFseStateWithDInfo(ZSTD_fseState* DStatePtr, BIT_DStream_t* bitD, U16 nextState, U32 nbBits)
+{
+ size_t const lowBits = BIT_readBits(bitD, nbBits);
+ DStatePtr->state = nextState + lowBits;
+}
+
+/* We need to add at most (ZSTD_WINDOWLOG_MAX_32 - 1) bits to read the maximum
+ * offset bits. But we can only read at most STREAM_ACCUMULATOR_MIN_32
+ * bits before reloading. This value is the maximum number of bytes we read
+ * after reloading when we are decoding long offsets.
+ */
+#define LONG_OFFSETS_MAX_EXTRA_BITS_32 \
+ (ZSTD_WINDOWLOG_MAX_32 > STREAM_ACCUMULATOR_MIN_32 \
+ ? ZSTD_WINDOWLOG_MAX_32 - STREAM_ACCUMULATOR_MIN_32 \
+ : 0)
+
+typedef enum { ZSTD_lo_isRegularOffset, ZSTD_lo_isLongOffset=1 } ZSTD_longOffset_e;
+
+/**
+ * ZSTD_decodeSequence():
+ * @p longOffsets : tells the decoder to reload more bit while decoding large offsets
+ * only used in 32-bit mode
+ * @return : Sequence (litL + matchL + offset)
+ */
+FORCE_INLINE_TEMPLATE seq_t
+ZSTD_decodeSequence(seqState_t* seqState, const ZSTD_longOffset_e longOffsets, const int isLastSeq)
+{
+ seq_t seq;
+ /*
+ * ZSTD_seqSymbol is a 64 bits wide structure.
+ * It can be loaded in one operation
+ * and its fields extracted by simply shifting or bit-extracting on aarch64.
+ * GCC doesn't recognize this and generates more unnecessary ldr/ldrb/ldrh
+ * operations that cause performance drop. This can be avoided by using this
+ * ZSTD_memcpy hack.
+ */
+#if defined(__aarch64__) && (defined(__GNUC__) && !defined(__clang__))
+ ZSTD_seqSymbol llDInfoS, mlDInfoS, ofDInfoS;
+ ZSTD_seqSymbol* const llDInfo = &llDInfoS;
+ ZSTD_seqSymbol* const mlDInfo = &mlDInfoS;
+ ZSTD_seqSymbol* const ofDInfo = &ofDInfoS;
+ ZSTD_memcpy(llDInfo, seqState->stateLL.table + seqState->stateLL.state, sizeof(ZSTD_seqSymbol));
+ ZSTD_memcpy(mlDInfo, seqState->stateML.table + seqState->stateML.state, sizeof(ZSTD_seqSymbol));
+ ZSTD_memcpy(ofDInfo, seqState->stateOffb.table + seqState->stateOffb.state, sizeof(ZSTD_seqSymbol));
+#else
+ const ZSTD_seqSymbol* const llDInfo = seqState->stateLL.table + seqState->stateLL.state;
+ const ZSTD_seqSymbol* const mlDInfo = seqState->stateML.table + seqState->stateML.state;
+ const ZSTD_seqSymbol* const ofDInfo = seqState->stateOffb.table + seqState->stateOffb.state;
+#endif
+ seq.matchLength = mlDInfo->baseValue;
+ seq.litLength = llDInfo->baseValue;
+ { U32 const ofBase = ofDInfo->baseValue;
+ BYTE const llBits = llDInfo->nbAdditionalBits;
+ BYTE const mlBits = mlDInfo->nbAdditionalBits;
+ BYTE const ofBits = ofDInfo->nbAdditionalBits;
+ BYTE const totalBits = llBits+mlBits+ofBits;
+
+ U16 const llNext = llDInfo->nextState;
+ U16 const mlNext = mlDInfo->nextState;
+ U16 const ofNext = ofDInfo->nextState;
+ U32 const llnbBits = llDInfo->nbBits;
+ U32 const mlnbBits = mlDInfo->nbBits;
+ U32 const ofnbBits = ofDInfo->nbBits;
+
+ assert(llBits <= MaxLLBits);
+ assert(mlBits <= MaxMLBits);
+ assert(ofBits <= MaxOff);
+ /*
+ * As gcc has better branch and block analyzers, sometimes it is only
+ * valuable to mark likeliness for clang, it gives around 3-4% of
+ * performance.
+ */
+
+ /* sequence */
+ { size_t offset;
+ if (ofBits > 1) {
+ ZSTD_STATIC_ASSERT(ZSTD_lo_isLongOffset == 1);
+ ZSTD_STATIC_ASSERT(LONG_OFFSETS_MAX_EXTRA_BITS_32 == 5);
+ ZSTD_STATIC_ASSERT(STREAM_ACCUMULATOR_MIN_32 > LONG_OFFSETS_MAX_EXTRA_BITS_32);
+ ZSTD_STATIC_ASSERT(STREAM_ACCUMULATOR_MIN_32 - LONG_OFFSETS_MAX_EXTRA_BITS_32 >= MaxMLBits);
+ if (MEM_32bits() && longOffsets && (ofBits >= STREAM_ACCUMULATOR_MIN_32)) {
+ /* Always read extra bits, this keeps the logic simple,
+ * avoids branches, and avoids accidentally reading 0 bits.
+ */
+ U32 const extraBits = LONG_OFFSETS_MAX_EXTRA_BITS_32;
+ offset = ofBase + (BIT_readBitsFast(&seqState->DStream, ofBits - extraBits) << extraBits);
+ BIT_reloadDStream(&seqState->DStream);
+ offset += BIT_readBitsFast(&seqState->DStream, extraBits);
+ } else {
+ offset = ofBase + BIT_readBitsFast(&seqState->DStream, ofBits/*>0*/); /* <= (ZSTD_WINDOWLOG_MAX-1) bits */
+ if (MEM_32bits()) BIT_reloadDStream(&seqState->DStream);
+ }
+ seqState->prevOffset[2] = seqState->prevOffset[1];
+ seqState->prevOffset[1] = seqState->prevOffset[0];
+ seqState->prevOffset[0] = offset;
+ } else {
+ U32 const ll0 = (llDInfo->baseValue == 0);
+ if (LIKELY((ofBits == 0))) {
+ offset = seqState->prevOffset[ll0];
+ seqState->prevOffset[1] = seqState->prevOffset[!ll0];
+ seqState->prevOffset[0] = offset;
+ } else {
+ offset = ofBase + ll0 + BIT_readBitsFast(&seqState->DStream, 1);
+ { size_t temp = (offset==3) ? seqState->prevOffset[0] - 1 : seqState->prevOffset[offset];
+ temp -= !temp; /* 0 is not valid: input corrupted => force offset to -1 => corruption detected at execSequence */
+ if (offset != 1) seqState->prevOffset[2] = seqState->prevOffset[1];
+ seqState->prevOffset[1] = seqState->prevOffset[0];
+ seqState->prevOffset[0] = offset = temp;
+ } } }
+ seq.offset = offset;
+ }
+
+ if (mlBits > 0)
+ seq.matchLength += BIT_readBitsFast(&seqState->DStream, mlBits/*>0*/);
+
+ if (MEM_32bits() && (mlBits+llBits >= STREAM_ACCUMULATOR_MIN_32-LONG_OFFSETS_MAX_EXTRA_BITS_32))
+ BIT_reloadDStream(&seqState->DStream);
+ if (MEM_64bits() && UNLIKELY(totalBits >= STREAM_ACCUMULATOR_MIN_64-(LLFSELog+MLFSELog+OffFSELog)))
+ BIT_reloadDStream(&seqState->DStream);
+ /* Ensure there are enough bits to read the rest of data in 64-bit mode. */
+ ZSTD_STATIC_ASSERT(16+LLFSELog+MLFSELog+OffFSELog < STREAM_ACCUMULATOR_MIN_64);
+
+ if (llBits > 0)
+ seq.litLength += BIT_readBitsFast(&seqState->DStream, llBits/*>0*/);
+
+ if (MEM_32bits())
+ BIT_reloadDStream(&seqState->DStream);
+
+ DEBUGLOG(6, "seq: litL=%u, matchL=%u, offset=%u",
+ (U32)seq.litLength, (U32)seq.matchLength, (U32)seq.offset);
+
+ if (!isLastSeq) {
+ /* don't update FSE state for last Sequence */
+ ZSTD_updateFseStateWithDInfo(&seqState->stateLL, &seqState->DStream, llNext, llnbBits); /* <= 9 bits */
+ ZSTD_updateFseStateWithDInfo(&seqState->stateML, &seqState->DStream, mlNext, mlnbBits); /* <= 9 bits */
+ if (MEM_32bits()) BIT_reloadDStream(&seqState->DStream); /* <= 18 bits */
+ ZSTD_updateFseStateWithDInfo(&seqState->stateOffb, &seqState->DStream, ofNext, ofnbBits); /* <= 8 bits */
+ BIT_reloadDStream(&seqState->DStream);
+ }
+ }
+
+ return seq;
+}
+
+#if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE)
+#if DEBUGLEVEL >= 1
+static int ZSTD_dictionaryIsActive(ZSTD_DCtx const* dctx, BYTE const* prefixStart, BYTE const* oLitEnd)
+{
+ size_t const windowSize = dctx->fParams.windowSize;
+ /* No dictionary used. */
+ if (dctx->dictContentEndForFuzzing == NULL) return 0;
+ /* Dictionary is our prefix. */
+ if (prefixStart == dctx->dictContentBeginForFuzzing) return 1;
+ /* Dictionary is not our ext-dict. */
+ if (dctx->dictEnd != dctx->dictContentEndForFuzzing) return 0;
+ /* Dictionary is not within our window size. */
+ if ((size_t)(oLitEnd - prefixStart) >= windowSize) return 0;
+ /* Dictionary is active. */
+ return 1;
+}
+#endif
+
+static void ZSTD_assertValidSequence(
+ ZSTD_DCtx const* dctx,
+ BYTE const* op, BYTE const* oend,
+ seq_t const seq,
+ BYTE const* prefixStart, BYTE const* virtualStart)
+{
+#if DEBUGLEVEL >= 1
+ if (dctx->isFrameDecompression) {
+ size_t const windowSize = dctx->fParams.windowSize;
+ size_t const sequenceSize = seq.litLength + seq.matchLength;
+ BYTE const* const oLitEnd = op + seq.litLength;
+ DEBUGLOG(6, "Checking sequence: litL=%u matchL=%u offset=%u",
+ (U32)seq.litLength, (U32)seq.matchLength, (U32)seq.offset);
+ assert(op <= oend);
+ assert((size_t)(oend - op) >= sequenceSize);
+ assert(sequenceSize <= ZSTD_blockSizeMax(dctx));
+ if (ZSTD_dictionaryIsActive(dctx, prefixStart, oLitEnd)) {
+ size_t const dictSize = (size_t)((char const*)dctx->dictContentEndForFuzzing - (char const*)dctx->dictContentBeginForFuzzing);
+ /* Offset must be within the dictionary. */
+ assert(seq.offset <= (size_t)(oLitEnd - virtualStart));
+ assert(seq.offset <= windowSize + dictSize);
+ } else {
+ /* Offset must be within our window. */
+ assert(seq.offset <= windowSize);
+ }
+ }
+#else
+ (void)dctx, (void)op, (void)oend, (void)seq, (void)prefixStart, (void)virtualStart;
+#endif
+}
+#endif
+
+#ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG
+
+
+FORCE_INLINE_TEMPLATE size_t
+DONT_VECTORIZE
+ZSTD_decompressSequences_bodySplitLitBuffer( ZSTD_DCtx* dctx,
+ void* dst, size_t maxDstSize,
+ const void* seqStart, size_t seqSize, int nbSeq,
+ const ZSTD_longOffset_e isLongOffset)
+{
+ const BYTE* ip = (const BYTE*)seqStart;
+ const BYTE* const iend = ip + seqSize;
+ BYTE* const ostart = (BYTE*)dst;
+ BYTE* const oend = ZSTD_maybeNullPtrAdd(ostart, maxDstSize);
+ BYTE* op = ostart;
+ const BYTE* litPtr = dctx->litPtr;
+ const BYTE* litBufferEnd = dctx->litBufferEnd;
+ const BYTE* const prefixStart = (const BYTE*) (dctx->prefixStart);
+ const BYTE* const vBase = (const BYTE*) (dctx->virtualStart);
+ const BYTE* const dictEnd = (const BYTE*) (dctx->dictEnd);
+ DEBUGLOG(5, "ZSTD_decompressSequences_bodySplitLitBuffer (%i seqs)", nbSeq);
+
+ /* Literals are split between internal buffer & output buffer */
+ if (nbSeq) {
+ seqState_t seqState;
+ dctx->fseEntropy = 1;
+ { U32 i; for (i=0; i<ZSTD_REP_NUM; i++) seqState.prevOffset[i] = dctx->entropy.rep[i]; }
+ RETURN_ERROR_IF(
+ ERR_isError(BIT_initDStream(&seqState.DStream, ip, iend-ip)),
+ corruption_detected, "");
+ ZSTD_initFseState(&seqState.stateLL, &seqState.DStream, dctx->LLTptr);
+ ZSTD_initFseState(&seqState.stateOffb, &seqState.DStream, dctx->OFTptr);
+ ZSTD_initFseState(&seqState.stateML, &seqState.DStream, dctx->MLTptr);
+ assert(dst != NULL);
+
+ ZSTD_STATIC_ASSERT(
+ BIT_DStream_unfinished < BIT_DStream_completed &&
+ BIT_DStream_endOfBuffer < BIT_DStream_completed &&
+ BIT_DStream_completed < BIT_DStream_overflow);
+
+ /* decompress without overrunning litPtr begins */
+ { 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) */
+ /* Align the decompression loop to 32 + 16 bytes.
+ *
+ * zstd compiled with gcc-9 on an Intel i9-9900k shows 10% decompression
+ * speed swings based on the alignment of the decompression loop. This
+ * performance swing is caused by parts of the decompression loop falling
+ * out of the DSB. The entire decompression loop should fit in the DSB,
+ * when it can't we get much worse performance. You can measure if you've
+ * hit the good case or the bad case with this perf command for some
+ * compressed file test.zst:
+ *
+ * perf stat -e cycles -e instructions -e idq.all_dsb_cycles_any_uops \
+ * -e idq.all_mite_cycles_any_uops -- ./zstd -tq test.zst
+ *
+ * If you see most cycles served out of the MITE you've hit the bad case.
+ * If you see most cycles served out of the DSB you've hit the good case.
+ * If it is pretty even then you may be in an okay case.
+ *
+ * This issue has been reproduced on the following CPUs:
+ * - Kabylake: Macbook Pro (15-inch, 2019) 2.4 GHz Intel Core i9
+ * Use Instruments->Counters to get DSB/MITE cycles.
+ * I never got performance swings, but I was able to
+ * go from the good case of mostly DSB to half of the
+ * cycles served from MITE.
+ * - Coffeelake: Intel i9-9900k
+ * - Coffeelake: Intel i7-9700k
+ *
+ * I haven't been able to reproduce the instability or DSB misses on any
+ * of the following CPUS:
+ * - Haswell
+ * - Broadwell: Intel(R) Xeon(R) CPU E5-2680 v4 @ 2.40GH
+ * - Skylake
+ *
+ * Alignment is done for each of the three major decompression loops:
+ * - ZSTD_decompressSequences_bodySplitLitBuffer - presplit section of the literal buffer
+ * - ZSTD_decompressSequences_bodySplitLitBuffer - postsplit section of the literal buffer
+ * - ZSTD_decompressSequences_body
+ * Alignment choices are made to minimize large swings on bad cases and influence on performance
+ * from changes external to this code, rather than to overoptimize on the current commit.
+ *
+ * If you are seeing performance stability this script can help test.
+ * It tests on 4 commits in zstd where I saw performance change.
+ *
+ * https://gist.github.com/terrelln/9889fc06a423fd5ca6e99351564473f4
+ */
+#if defined(__GNUC__) && defined(__x86_64__)
+ __asm__(".p2align 6");
+# if __GNUC__ >= 7
+ /* good for gcc-7, gcc-9, and gcc-11 */
+ __asm__("nop");
+ __asm__(".p2align 5");
+ __asm__("nop");
+ __asm__(".p2align 4");
+# if __GNUC__ == 8 || __GNUC__ == 10
+ /* good for gcc-8 and gcc-10 */
+ __asm__("nop");
+ __asm__(".p2align 3");
+# endif
+# endif
+#endif
+
+ /* Handle the initial state where litBuffer is currently split between dst and litExtraBuffer */
+ for ( ; nbSeq; nbSeq--) {
+ sequence = ZSTD_decodeSequence(&seqState, isLongOffset, nbSeq==1);
+ if (litPtr + sequence.litLength > dctx->litBufferEnd) break;
+ { size_t const oneSeqSize = ZSTD_execSequenceSplitLitBuffer(op, oend, litPtr + sequence.litLength - WILDCOPY_OVERLENGTH, sequence, &litPtr, litBufferEnd, prefixStart, vBase, dictEnd);
+#if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE)
+ assert(!ZSTD_isError(oneSeqSize));
+ ZSTD_assertValidSequence(dctx, op, oend, sequence, prefixStart, vBase);
+#endif
+ if (UNLIKELY(ZSTD_isError(oneSeqSize)))
+ return oneSeqSize;
+ DEBUGLOG(6, "regenerated sequence size : %u", (U32)oneSeqSize);
+ op += oneSeqSize;
+ } }
+ DEBUGLOG(6, "reached: (litPtr + sequence.litLength > dctx->litBufferEnd)");
+
+ /* If there are more sequences, they will need to read literals from litExtraBuffer; copy over the remainder from dst and update litPtr and litEnd */
+ if (nbSeq > 0) {
+ const size_t leftoverLit = dctx->litBufferEnd - litPtr;
+ DEBUGLOG(6, "There are %i sequences left, and %zu/%zu literals left in buffer", nbSeq, leftoverLit, sequence.litLength);
+ if (leftoverLit) {
+ RETURN_ERROR_IF(leftoverLit > (size_t)(oend - op), dstSize_tooSmall, "remaining lit must fit within dstBuffer");
+ ZSTD_safecopyDstBeforeSrc(op, litPtr, leftoverLit);
+ sequence.litLength -= leftoverLit;
+ op += leftoverLit;
+ }
+ litPtr = dctx->litExtraBuffer;
+ litBufferEnd = dctx->litExtraBuffer + ZSTD_LITBUFFEREXTRASIZE;
+ dctx->litBufferLocation = ZSTD_not_in_dst;
+ { size_t const oneSeqSize = ZSTD_execSequence(op, oend, sequence, &litPtr, litBufferEnd, prefixStart, vBase, dictEnd);
+#if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE)
+ assert(!ZSTD_isError(oneSeqSize));
+ ZSTD_assertValidSequence(dctx, op, oend, sequence, prefixStart, vBase);
+#endif
+ if (UNLIKELY(ZSTD_isError(oneSeqSize)))
+ return oneSeqSize;
+ DEBUGLOG(6, "regenerated sequence size : %u", (U32)oneSeqSize);
+ op += oneSeqSize;
+ }
+ nbSeq--;
+ }
+ }
+
+ if (nbSeq > 0) {
+ /* there is remaining lit from extra buffer */
+
+#if defined(__GNUC__) && defined(__x86_64__)
+ __asm__(".p2align 6");
+ __asm__("nop");
+# if __GNUC__ != 7
+ /* worse for gcc-7 better for gcc-8, gcc-9, and gcc-10 and clang */
+ __asm__(".p2align 4");
+ __asm__("nop");
+ __asm__(".p2align 3");
+# elif __GNUC__ >= 11
+ __asm__(".p2align 3");
+# else
+ __asm__(".p2align 5");
+ __asm__("nop");
+ __asm__(".p2align 3");
+# endif
+#endif
+
+ for ( ; nbSeq ; nbSeq--) {
+ seq_t const sequence = ZSTD_decodeSequence(&seqState, isLongOffset, nbSeq==1);
+ size_t const oneSeqSize = ZSTD_execSequence(op, oend, sequence, &litPtr, litBufferEnd, prefixStart, vBase, dictEnd);
+#if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE)
+ assert(!ZSTD_isError(oneSeqSize));
+ ZSTD_assertValidSequence(dctx, op, oend, sequence, prefixStart, vBase);
+#endif
+ if (UNLIKELY(ZSTD_isError(oneSeqSize)))
+ return oneSeqSize;
+ DEBUGLOG(6, "regenerated sequence size : %u", (U32)oneSeqSize);
+ op += oneSeqSize;
+ }
+ }
+
+ /* check if reached exact end */
+ DEBUGLOG(5, "ZSTD_decompressSequences_bodySplitLitBuffer: after decode loop, remaining nbSeq : %i", nbSeq);
+ RETURN_ERROR_IF(nbSeq, corruption_detected, "");
+ DEBUGLOG(5, "bitStream : start=%p, ptr=%p, bitsConsumed=%u", seqState.DStream.start, seqState.DStream.ptr, seqState.DStream.bitsConsumed);
+ RETURN_ERROR_IF(!BIT_endOfDStream(&seqState.DStream), corruption_detected, "");
+ /* save reps for next block */
+ { U32 i; for (i=0; i<ZSTD_REP_NUM; i++) dctx->entropy.rep[i] = (U32)(seqState.prevOffset[i]); }
+ }
+
+ /* last literal segment */
+ if (dctx->litBufferLocation == ZSTD_split) {
+ /* split hasn't been reached yet, first get dst then copy litExtraBuffer */
+ size_t const lastLLSize = (size_t)(litBufferEnd - litPtr);
+ DEBUGLOG(6, "copy last literals from segment : %u", (U32)lastLLSize);
+ RETURN_ERROR_IF(lastLLSize > (size_t)(oend - op), dstSize_tooSmall, "");
+ if (op != NULL) {
+ ZSTD_memmove(op, litPtr, lastLLSize);
+ op += lastLLSize;
+ }
+ litPtr = dctx->litExtraBuffer;
+ litBufferEnd = dctx->litExtraBuffer + ZSTD_LITBUFFEREXTRASIZE;
+ dctx->litBufferLocation = ZSTD_not_in_dst;
+ }
+ /* copy last literals from internal buffer */
+ { size_t const lastLLSize = (size_t)(litBufferEnd - litPtr);
+ DEBUGLOG(6, "copy last literals from internal buffer : %u", (U32)lastLLSize);
+ RETURN_ERROR_IF(lastLLSize > (size_t)(oend-op), dstSize_tooSmall, "");
+ if (op != NULL) {
+ ZSTD_memcpy(op, litPtr, lastLLSize);
+ op += lastLLSize;
+ } }
+
+ DEBUGLOG(6, "decoded block of size %u bytes", (U32)(op - ostart));
+ return (size_t)(op - ostart);
+}
+
+FORCE_INLINE_TEMPLATE size_t
+DONT_VECTORIZE
+ZSTD_decompressSequences_body(ZSTD_DCtx* dctx,
+ void* dst, size_t maxDstSize,
+ const void* seqStart, size_t seqSize, int nbSeq,
+ const ZSTD_longOffset_e isLongOffset)
+{
+ const BYTE* ip = (const BYTE*)seqStart;
+ const BYTE* const iend = ip + seqSize;
+ BYTE* const ostart = (BYTE*)dst;
+ BYTE* const oend = dctx->litBufferLocation == ZSTD_not_in_dst ? ZSTD_maybeNullPtrAdd(ostart, maxDstSize) : dctx->litBuffer;
+ BYTE* op = ostart;
+ const BYTE* litPtr = dctx->litPtr;
+ const BYTE* const litEnd = litPtr + dctx->litSize;
+ const BYTE* const prefixStart = (const BYTE*)(dctx->prefixStart);
+ const BYTE* const vBase = (const BYTE*)(dctx->virtualStart);
+ const BYTE* const dictEnd = (const BYTE*)(dctx->dictEnd);
+ DEBUGLOG(5, "ZSTD_decompressSequences_body: nbSeq = %d", nbSeq);
+
+ /* Regen sequences */
+ if (nbSeq) {
+ seqState_t seqState;
+ dctx->fseEntropy = 1;
+ { U32 i; for (i = 0; i < ZSTD_REP_NUM; i++) seqState.prevOffset[i] = dctx->entropy.rep[i]; }
+ RETURN_ERROR_IF(
+ ERR_isError(BIT_initDStream(&seqState.DStream, ip, iend - ip)),
+ corruption_detected, "");
+ ZSTD_initFseState(&seqState.stateLL, &seqState.DStream, dctx->LLTptr);
+ ZSTD_initFseState(&seqState.stateOffb, &seqState.DStream, dctx->OFTptr);
+ ZSTD_initFseState(&seqState.stateML, &seqState.DStream, dctx->MLTptr);
+ assert(dst != NULL);
+
+#if defined(__GNUC__) && defined(__x86_64__)
+ __asm__(".p2align 6");
+ __asm__("nop");
+# if __GNUC__ >= 7
+ __asm__(".p2align 5");
+ __asm__("nop");
+ __asm__(".p2align 3");
+# else
+ __asm__(".p2align 4");
+ __asm__("nop");
+ __asm__(".p2align 3");
+# endif
+#endif
+
+ for ( ; nbSeq ; nbSeq--) {
+ seq_t const sequence = ZSTD_decodeSequence(&seqState, isLongOffset, nbSeq==1);
+ size_t const oneSeqSize = ZSTD_execSequence(op, oend, sequence, &litPtr, litEnd, prefixStart, vBase, dictEnd);
+#if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE)
+ assert(!ZSTD_isError(oneSeqSize));
+ ZSTD_assertValidSequence(dctx, op, oend, sequence, prefixStart, vBase);
+#endif
+ if (UNLIKELY(ZSTD_isError(oneSeqSize)))
+ return oneSeqSize;
+ DEBUGLOG(6, "regenerated sequence size : %u", (U32)oneSeqSize);
+ op += oneSeqSize;
+ }
+
+ /* check if reached exact end */
+ assert(nbSeq == 0);
+ RETURN_ERROR_IF(!BIT_endOfDStream(&seqState.DStream), corruption_detected, "");
+ /* save reps for next block */
+ { U32 i; for (i=0; i<ZSTD_REP_NUM; i++) dctx->entropy.rep[i] = (U32)(seqState.prevOffset[i]); }
+ }
+
+ /* last literal segment */
+ { size_t const lastLLSize = (size_t)(litEnd - litPtr);
+ DEBUGLOG(6, "copy last literals : %u", (U32)lastLLSize);
+ RETURN_ERROR_IF(lastLLSize > (size_t)(oend-op), dstSize_tooSmall, "");
+ if (op != NULL) {
+ ZSTD_memcpy(op, litPtr, lastLLSize);
+ op += lastLLSize;
+ } }
+
+ DEBUGLOG(6, "decoded block of size %u bytes", (U32)(op - ostart));
+ return (size_t)(op - ostart);
+}
+
+static size_t
+ZSTD_decompressSequences_default(ZSTD_DCtx* dctx,
+ void* dst, size_t maxDstSize,
+ const void* seqStart, size_t seqSize, int nbSeq,
+ const ZSTD_longOffset_e isLongOffset)
+{
+ return ZSTD_decompressSequences_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+}
+
+static size_t
+ZSTD_decompressSequencesSplitLitBuffer_default(ZSTD_DCtx* dctx,
+ void* dst, size_t maxDstSize,
+ const void* seqStart, size_t seqSize, int nbSeq,
+ const ZSTD_longOffset_e isLongOffset)
+{
+ return ZSTD_decompressSequences_bodySplitLitBuffer(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+}
+#endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG */
+
+#ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT
+
+FORCE_INLINE_TEMPLATE
+
+size_t ZSTD_prefetchMatch(size_t prefetchPos, seq_t const sequence,
+ const BYTE* const prefixStart, const BYTE* const dictEnd)
+{
+ prefetchPos += sequence.litLength;
+ { const BYTE* const matchBase = (sequence.offset > prefetchPos) ? dictEnd : prefixStart;
+ /* note : this operation can overflow when seq.offset is really too large, which can only happen when input is corrupted.
+ * No consequence though : memory address is only used for prefetching, not for dereferencing */
+ const BYTE* const match = ZSTD_wrappedPtrSub(ZSTD_wrappedPtrAdd(matchBase, prefetchPos), sequence.offset);
+ PREFETCH_L1(match); PREFETCH_L1(match+CACHELINE_SIZE); /* note : it's safe to invoke PREFETCH() on any memory address, including invalid ones */
+ }
+ return prefetchPos + sequence.matchLength;
+}
+
+/* This decoding function employs prefetching
+ * to reduce latency impact of cache misses.
+ * It's generally employed when block contains a significant portion of long-distance matches
+ * or when coupled with a "cold" dictionary */
+FORCE_INLINE_TEMPLATE size_t
+ZSTD_decompressSequencesLong_body(
+ ZSTD_DCtx* dctx,
+ void* dst, size_t maxDstSize,
+ const void* seqStart, size_t seqSize, int nbSeq,
+ const ZSTD_longOffset_e isLongOffset)
+{
+ const BYTE* ip = (const BYTE*)seqStart;
+ const BYTE* const iend = ip + seqSize;
+ BYTE* const ostart = (BYTE*)dst;
+ BYTE* const oend = dctx->litBufferLocation == ZSTD_in_dst ? dctx->litBuffer : ZSTD_maybeNullPtrAdd(ostart, maxDstSize);
+ BYTE* op = ostart;
+ const BYTE* litPtr = dctx->litPtr;
+ const BYTE* litBufferEnd = dctx->litBufferEnd;
+ const BYTE* const prefixStart = (const BYTE*) (dctx->prefixStart);
+ const BYTE* const dictStart = (const BYTE*) (dctx->virtualStart);
+ const BYTE* const dictEnd = (const BYTE*) (dctx->dictEnd);
+
+ /* Regen sequences */
+ if (nbSeq) {
+#define STORED_SEQS 8
+#define STORED_SEQS_MASK (STORED_SEQS-1)
+#define ADVANCED_SEQS STORED_SEQS
+ seq_t sequences[STORED_SEQS];
+ int const seqAdvance = MIN(nbSeq, ADVANCED_SEQS);
+ seqState_t seqState;
+ int seqNb;
+ size_t prefetchPos = (size_t)(op-prefixStart); /* track position relative to prefixStart */
+
+ dctx->fseEntropy = 1;
+ { int i; for (i=0; i<ZSTD_REP_NUM; i++) seqState.prevOffset[i] = dctx->entropy.rep[i]; }
+ assert(dst != NULL);
+ assert(iend >= ip);
+ RETURN_ERROR_IF(
+ ERR_isError(BIT_initDStream(&seqState.DStream, ip, iend-ip)),
+ corruption_detected, "");
+ ZSTD_initFseState(&seqState.stateLL, &seqState.DStream, dctx->LLTptr);
+ ZSTD_initFseState(&seqState.stateOffb, &seqState.DStream, dctx->OFTptr);
+ ZSTD_initFseState(&seqState.stateML, &seqState.DStream, dctx->MLTptr);
+
+ /* prepare in advance */
+ for (seqNb=0; seqNb<seqAdvance; seqNb++) {
+ seq_t const sequence = ZSTD_decodeSequence(&seqState, isLongOffset, seqNb == nbSeq-1);
+ prefetchPos = ZSTD_prefetchMatch(prefetchPos, sequence, prefixStart, dictEnd);
+ sequences[seqNb] = sequence;
+ }
+
+ /* decompress without stomping litBuffer */
+ for (; seqNb < nbSeq; seqNb++) {
+ seq_t sequence = ZSTD_decodeSequence(&seqState, isLongOffset, seqNb == nbSeq-1);
+
+ if (dctx->litBufferLocation == ZSTD_split && litPtr + sequences[(seqNb - ADVANCED_SEQS) & STORED_SEQS_MASK].litLength > dctx->litBufferEnd) {
+ /* lit buffer is reaching split point, empty out the first buffer and transition to litExtraBuffer */
+ const size_t leftoverLit = dctx->litBufferEnd - litPtr;
+ if (leftoverLit)
+ {
+ RETURN_ERROR_IF(leftoverLit > (size_t)(oend - op), dstSize_tooSmall, "remaining lit must fit within dstBuffer");
+ ZSTD_safecopyDstBeforeSrc(op, litPtr, leftoverLit);
+ sequences[(seqNb - ADVANCED_SEQS) & STORED_SEQS_MASK].litLength -= leftoverLit;
+ op += leftoverLit;
+ }
+ litPtr = dctx->litExtraBuffer;
+ litBufferEnd = dctx->litExtraBuffer + ZSTD_LITBUFFEREXTRASIZE;
+ dctx->litBufferLocation = ZSTD_not_in_dst;
+ { size_t const oneSeqSize = ZSTD_execSequence(op, oend, sequences[(seqNb - ADVANCED_SEQS) & STORED_SEQS_MASK], &litPtr, litBufferEnd, prefixStart, dictStart, dictEnd);
+#if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE)
+ assert(!ZSTD_isError(oneSeqSize));
+ ZSTD_assertValidSequence(dctx, op, oend, sequences[(seqNb - ADVANCED_SEQS) & STORED_SEQS_MASK], prefixStart, dictStart);
+#endif
+ if (ZSTD_isError(oneSeqSize)) return oneSeqSize;
+
+ prefetchPos = ZSTD_prefetchMatch(prefetchPos, sequence, prefixStart, dictEnd);
+ sequences[seqNb & STORED_SEQS_MASK] = sequence;
+ op += oneSeqSize;
+ } }
+ else
+ {
+ /* lit buffer is either wholly contained in first or second split, or not split at all*/
+ size_t const oneSeqSize = dctx->litBufferLocation == ZSTD_split ?
+ 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) :
+ ZSTD_execSequence(op, oend, sequences[(seqNb - ADVANCED_SEQS) & STORED_SEQS_MASK], &litPtr, litBufferEnd, prefixStart, dictStart, dictEnd);
+#if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE)
+ assert(!ZSTD_isError(oneSeqSize));
+ ZSTD_assertValidSequence(dctx, op, oend, sequences[(seqNb - ADVANCED_SEQS) & STORED_SEQS_MASK], prefixStart, dictStart);
+#endif
+ if (ZSTD_isError(oneSeqSize)) return oneSeqSize;
+
+ prefetchPos = ZSTD_prefetchMatch(prefetchPos, sequence, prefixStart, dictEnd);
+ sequences[seqNb & STORED_SEQS_MASK] = sequence;
+ op += oneSeqSize;
+ }
+ }
+ RETURN_ERROR_IF(!BIT_endOfDStream(&seqState.DStream), corruption_detected, "");
+
+ /* finish queue */
+ seqNb -= seqAdvance;
+ for ( ; seqNb<nbSeq ; seqNb++) {
+ seq_t *sequence = &(sequences[seqNb&STORED_SEQS_MASK]);
+ if (dctx->litBufferLocation == ZSTD_split && litPtr + sequence->litLength > dctx->litBufferEnd) {
+ const size_t leftoverLit = dctx->litBufferEnd - litPtr;
+ if (leftoverLit) {
+ RETURN_ERROR_IF(leftoverLit > (size_t)(oend - op), dstSize_tooSmall, "remaining lit must fit within dstBuffer");
+ ZSTD_safecopyDstBeforeSrc(op, litPtr, leftoverLit);
+ sequence->litLength -= leftoverLit;
+ op += leftoverLit;
+ }
+ litPtr = dctx->litExtraBuffer;
+ litBufferEnd = dctx->litExtraBuffer + ZSTD_LITBUFFEREXTRASIZE;
+ dctx->litBufferLocation = ZSTD_not_in_dst;
+ { size_t const oneSeqSize = ZSTD_execSequence(op, oend, *sequence, &litPtr, litBufferEnd, prefixStart, dictStart, dictEnd);
+#if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE)
+ assert(!ZSTD_isError(oneSeqSize));
+ ZSTD_assertValidSequence(dctx, op, oend, sequences[seqNb&STORED_SEQS_MASK], prefixStart, dictStart);
+#endif
+ if (ZSTD_isError(oneSeqSize)) return oneSeqSize;
+ op += oneSeqSize;
+ }
+ }
+ else
+ {
+ size_t const oneSeqSize = dctx->litBufferLocation == ZSTD_split ?
+ ZSTD_execSequenceSplitLitBuffer(op, oend, litPtr + sequence->litLength - WILDCOPY_OVERLENGTH, *sequence, &litPtr, litBufferEnd, prefixStart, dictStart, dictEnd) :
+ ZSTD_execSequence(op, oend, *sequence, &litPtr, litBufferEnd, prefixStart, dictStart, dictEnd);
+#if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE)
+ assert(!ZSTD_isError(oneSeqSize));
+ ZSTD_assertValidSequence(dctx, op, oend, sequences[seqNb&STORED_SEQS_MASK], prefixStart, dictStart);
+#endif
+ if (ZSTD_isError(oneSeqSize)) return oneSeqSize;
+ op += oneSeqSize;
+ }
+ }
+
+ /* save reps for next block */
+ { U32 i; for (i=0; i<ZSTD_REP_NUM; i++) dctx->entropy.rep[i] = (U32)(seqState.prevOffset[i]); }
+ }
+
+ /* last literal segment */
+ if (dctx->litBufferLocation == ZSTD_split) { /* first deplete literal buffer in dst, then copy litExtraBuffer */
+ size_t const lastLLSize = litBufferEnd - litPtr;
+ RETURN_ERROR_IF(lastLLSize > (size_t)(oend - op), dstSize_tooSmall, "");
+ if (op != NULL) {
+ ZSTD_memmove(op, litPtr, lastLLSize);
+ op += lastLLSize;
+ }
+ litPtr = dctx->litExtraBuffer;
+ litBufferEnd = dctx->litExtraBuffer + ZSTD_LITBUFFEREXTRASIZE;
+ }
+ { size_t const lastLLSize = litBufferEnd - litPtr;
+ RETURN_ERROR_IF(lastLLSize > (size_t)(oend-op), dstSize_tooSmall, "");
+ if (op != NULL) {
+ ZSTD_memmove(op, litPtr, lastLLSize);
+ op += lastLLSize;
+ }
+ }
+
+ return (size_t)(op - ostart);
+}
+
+static size_t
+ZSTD_decompressSequencesLong_default(ZSTD_DCtx* dctx,
+ void* dst, size_t maxDstSize,
+ const void* seqStart, size_t seqSize, int nbSeq,
+ const ZSTD_longOffset_e isLongOffset)
+{
+ return ZSTD_decompressSequencesLong_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+}
+#endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT */
+
+
+
+#if DYNAMIC_BMI2
+
+#ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG
+static BMI2_TARGET_ATTRIBUTE size_t
+DONT_VECTORIZE
+ZSTD_decompressSequences_bmi2(ZSTD_DCtx* dctx,
+ void* dst, size_t maxDstSize,
+ const void* seqStart, size_t seqSize, int nbSeq,
+ const ZSTD_longOffset_e isLongOffset)
+{
+ return ZSTD_decompressSequences_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+}
+static BMI2_TARGET_ATTRIBUTE size_t
+DONT_VECTORIZE
+ZSTD_decompressSequencesSplitLitBuffer_bmi2(ZSTD_DCtx* dctx,
+ void* dst, size_t maxDstSize,
+ const void* seqStart, size_t seqSize, int nbSeq,
+ const ZSTD_longOffset_e isLongOffset)
+{
+ return ZSTD_decompressSequences_bodySplitLitBuffer(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+}
+#endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG */
+
+#ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT
+static BMI2_TARGET_ATTRIBUTE size_t
+ZSTD_decompressSequencesLong_bmi2(ZSTD_DCtx* dctx,
+ void* dst, size_t maxDstSize,
+ const void* seqStart, size_t seqSize, int nbSeq,
+ const ZSTD_longOffset_e isLongOffset)
+{
+ return ZSTD_decompressSequencesLong_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+}
+#endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT */
+
+#endif /* DYNAMIC_BMI2 */
+
+typedef size_t (*ZSTD_decompressSequences_t)(
+ ZSTD_DCtx* dctx,
+ void* dst, size_t maxDstSize,
+ const void* seqStart, size_t seqSize, int nbSeq,
+ const ZSTD_longOffset_e isLongOffset);
+
+#ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG
+static size_t
+ZSTD_decompressSequences(ZSTD_DCtx* dctx, void* dst, size_t maxDstSize,
+ const void* seqStart, size_t seqSize, int nbSeq,
+ const ZSTD_longOffset_e isLongOffset)
+{
+ DEBUGLOG(5, "ZSTD_decompressSequences");
+#if DYNAMIC_BMI2
+ if (ZSTD_DCtx_get_bmi2(dctx)) {
+ return ZSTD_decompressSequences_bmi2(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+ }
+#endif
+ return ZSTD_decompressSequences_default(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+}
+static size_t
+ZSTD_decompressSequencesSplitLitBuffer(ZSTD_DCtx* dctx, void* dst, size_t maxDstSize,
+ const void* seqStart, size_t seqSize, int nbSeq,
+ const ZSTD_longOffset_e isLongOffset)
+{
+ DEBUGLOG(5, "ZSTD_decompressSequencesSplitLitBuffer");
+#if DYNAMIC_BMI2
+ if (ZSTD_DCtx_get_bmi2(dctx)) {
+ return ZSTD_decompressSequencesSplitLitBuffer_bmi2(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+ }
+#endif
+ return ZSTD_decompressSequencesSplitLitBuffer_default(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+}
+#endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG */
+
+
+#ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT
+/* ZSTD_decompressSequencesLong() :
+ * decompression function triggered when a minimum share of offsets is considered "long",
+ * aka out of cache.
+ * note : "long" definition seems overloaded here, sometimes meaning "wider than bitstream register", and sometimes meaning "farther than memory cache distance".
+ * This function will try to mitigate main memory latency through the use of prefetching */
+static size_t
+ZSTD_decompressSequencesLong(ZSTD_DCtx* dctx,
+ void* dst, size_t maxDstSize,
+ const void* seqStart, size_t seqSize, int nbSeq,
+ const ZSTD_longOffset_e isLongOffset)
+{
+ DEBUGLOG(5, "ZSTD_decompressSequencesLong");
+#if DYNAMIC_BMI2
+ if (ZSTD_DCtx_get_bmi2(dctx)) {
+ return ZSTD_decompressSequencesLong_bmi2(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+ }
+#endif
+ return ZSTD_decompressSequencesLong_default(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+}
+#endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT */
+
+
+/**
+ * @returns The total size of the history referenceable by zstd, including
+ * both the prefix and the extDict. At @p op any offset larger than this
+ * is invalid.
+ */
+static size_t ZSTD_totalHistorySize(BYTE* op, BYTE const* virtualStart)
+{
+ return (size_t)(op - virtualStart);
+}
+
+typedef struct {
+ unsigned longOffsetShare;
+ unsigned maxNbAdditionalBits;
+} ZSTD_OffsetInfo;
+
+/* ZSTD_getOffsetInfo() :
+ * condition : offTable must be valid
+ * @return : "share" of long offsets (arbitrarily defined as > (1<<23))
+ * compared to maximum possible of (1<<OffFSELog),
+ * as well as the maximum number additional bits required.
+ */
+static ZSTD_OffsetInfo
+ZSTD_getOffsetInfo(const ZSTD_seqSymbol* offTable, int nbSeq)
+{
+ ZSTD_OffsetInfo info = {0, 0};
+ /* If nbSeq == 0, then the offTable is uninitialized, but we have
+ * no sequences, so both values should be 0.
+ */
+ if (nbSeq != 0) {
+ const void* ptr = offTable;
+ U32 const tableLog = ((const ZSTD_seqSymbol_header*)ptr)[0].tableLog;
+ const ZSTD_seqSymbol* table = offTable + 1;
+ U32 const max = 1 << tableLog;
+ U32 u;
+ DEBUGLOG(5, "ZSTD_getLongOffsetsShare: (tableLog=%u)", tableLog);
+
+ assert(max <= (1 << OffFSELog)); /* max not too large */
+ for (u=0; u<max; u++) {
+ info.maxNbAdditionalBits = MAX(info.maxNbAdditionalBits, table[u].nbAdditionalBits);
+ if (table[u].nbAdditionalBits > 22) info.longOffsetShare += 1;
+ }
+
+ assert(tableLog <= OffFSELog);
+ info.longOffsetShare <<= (OffFSELog - tableLog); /* scale to OffFSELog */
+ }
+
+ return info;
+}
+
+/**
+ * @returns The maximum offset we can decode in one read of our bitstream, without
+ * reloading more bits in the middle of the offset bits read. Any offsets larger
+ * than this must use the long offset decoder.
+ */
+static size_t ZSTD_maxShortOffset(void)
+{
+ if (MEM_64bits()) {
+ /* We can decode any offset without reloading bits.
+ * This might change if the max window size grows.
+ */
+ ZSTD_STATIC_ASSERT(ZSTD_WINDOWLOG_MAX <= 31);
+ return (size_t)-1;
+ } else {
+ /* The maximum offBase is (1 << (STREAM_ACCUMULATOR_MIN + 1)) - 1.
+ * This offBase would require STREAM_ACCUMULATOR_MIN extra bits.
+ * Then we have to subtract ZSTD_REP_NUM to get the maximum possible offset.
+ */
+ size_t const maxOffbase = ((size_t)1 << (STREAM_ACCUMULATOR_MIN + 1)) - 1;
+ size_t const maxOffset = maxOffbase - ZSTD_REP_NUM;
+ assert(ZSTD_highbit32((U32)maxOffbase) == STREAM_ACCUMULATOR_MIN);
+ return maxOffset;
+ }
+}
+
+size_t
+ZSTD_decompressBlock_internal(ZSTD_DCtx* dctx,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize, const streaming_operation streaming)
+{ /* blockType == blockCompressed */
+ const BYTE* ip = (const BYTE*)src;
+ DEBUGLOG(5, "ZSTD_decompressBlock_internal (cSize : %u)", (unsigned)srcSize);
+
+ /* Note : the wording of the specification
+ * allows compressed block to be sized exactly ZSTD_blockSizeMax(dctx).
+ * This generally does not happen, as it makes little sense,
+ * since an uncompressed block would feature same size and have no decompression cost.
+ * Also, note that decoder from reference libzstd before < v1.5.4
+ * would consider this edge case as an error.
+ * As a consequence, avoid generating compressed blocks of size ZSTD_blockSizeMax(dctx)
+ * for broader compatibility with the deployed ecosystem of zstd decoders */
+ RETURN_ERROR_IF(srcSize > ZSTD_blockSizeMax(dctx), srcSize_wrong, "");
+
+ /* Decode literals section */
+ { size_t const litCSize = ZSTD_decodeLiteralsBlock(dctx, src, srcSize, dst, dstCapacity, streaming);
+ DEBUGLOG(5, "ZSTD_decodeLiteralsBlock : cSize=%u, nbLiterals=%zu", (U32)litCSize, dctx->litSize);
+ if (ZSTD_isError(litCSize)) return litCSize;
+ ip += litCSize;
+ srcSize -= litCSize;
+ }
+
+ /* Build Decoding Tables */
+ {
+ /* Compute the maximum block size, which must also work when !frame and fParams are unset.
+ * Additionally, take the min with dstCapacity to ensure that the totalHistorySize fits in a size_t.
+ */
+ size_t const blockSizeMax = MIN(dstCapacity, ZSTD_blockSizeMax(dctx));
+ size_t const totalHistorySize = ZSTD_totalHistorySize(ZSTD_maybeNullPtrAdd((BYTE*)dst, blockSizeMax), (BYTE const*)dctx->virtualStart);
+ /* isLongOffset must be true if there are long offsets.
+ * Offsets are long if they are larger than ZSTD_maxShortOffset().
+ * We don't expect that to be the case in 64-bit mode.
+ *
+ * We check here to see if our history is large enough to allow long offsets.
+ * If it isn't, then we can't possible have (valid) long offsets. If the offset
+ * is invalid, then it is okay to read it incorrectly.
+ *
+ * If isLongOffsets is true, then we will later check our decoding table to see
+ * if it is even possible to generate long offsets.
+ */
+ ZSTD_longOffset_e isLongOffset = (ZSTD_longOffset_e)(MEM_32bits() && (totalHistorySize > ZSTD_maxShortOffset()));
+ /* These macros control at build-time which decompressor implementation
+ * we use. If neither is defined, we do some inspection and dispatch at
+ * runtime.
+ */
+#if !defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT) && \
+ !defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG)
+ int usePrefetchDecoder = dctx->ddictIsCold;
+#else
+ /* Set to 1 to avoid computing offset info if we don't need to.
+ * Otherwise this value is ignored.
+ */
+ int usePrefetchDecoder = 1;
+#endif
+ int nbSeq;
+ size_t const seqHSize = ZSTD_decodeSeqHeaders(dctx, &nbSeq, ip, srcSize);
+ if (ZSTD_isError(seqHSize)) return seqHSize;
+ ip += seqHSize;
+ srcSize -= seqHSize;
+
+ RETURN_ERROR_IF((dst == NULL || dstCapacity == 0) && nbSeq > 0, dstSize_tooSmall, "NULL not handled");
+ RETURN_ERROR_IF(MEM_64bits() && sizeof(size_t) == sizeof(void*) && (size_t)(-1) - (size_t)dst < (size_t)(1 << 20), dstSize_tooSmall,
+ "invalid dst");
+
+ /* If we could potentially have long offsets, or we might want to use the prefetch decoder,
+ * compute information about the share of long offsets, and the maximum nbAdditionalBits.
+ * NOTE: could probably use a larger nbSeq limit
+ */
+ if (isLongOffset || (!usePrefetchDecoder && (totalHistorySize > (1u << 24)) && (nbSeq > 8))) {
+ ZSTD_OffsetInfo const info = ZSTD_getOffsetInfo(dctx->OFTptr, nbSeq);
+ if (isLongOffset && info.maxNbAdditionalBits <= STREAM_ACCUMULATOR_MIN) {
+ /* If isLongOffset, but the maximum number of additional bits that we see in our table is small
+ * enough, then we know it is impossible to have too long an offset in this block, so we can
+ * use the regular offset decoder.
+ */
+ isLongOffset = ZSTD_lo_isRegularOffset;
+ }
+ if (!usePrefetchDecoder) {
+ U32 const minShare = MEM_64bits() ? 7 : 20; /* heuristic values, correspond to 2.73% and 7.81% */
+ usePrefetchDecoder = (info.longOffsetShare >= minShare);
+ }
+ }
+
+ dctx->ddictIsCold = 0;
+
+#if !defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT) && \
+ !defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG)
+ if (usePrefetchDecoder) {
+#else
+ (void)usePrefetchDecoder;
+ {
+#endif
+#ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT
+ return ZSTD_decompressSequencesLong(dctx, dst, dstCapacity, ip, srcSize, nbSeq, isLongOffset);
+#endif
+ }
+
+#ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG
+ /* else */
+ if (dctx->litBufferLocation == ZSTD_split)
+ return ZSTD_decompressSequencesSplitLitBuffer(dctx, dst, dstCapacity, ip, srcSize, nbSeq, isLongOffset);
+ else
+ return ZSTD_decompressSequences(dctx, dst, dstCapacity, ip, srcSize, nbSeq, isLongOffset);
+#endif
+ }
+}
+
+
+ZSTD_ALLOW_POINTER_OVERFLOW_ATTR
+void ZSTD_checkContinuity(ZSTD_DCtx* dctx, const void* dst, size_t dstSize)
+{
+ if (dst != dctx->previousDstEnd && dstSize > 0) { /* not contiguous */
+ dctx->dictEnd = dctx->previousDstEnd;
+ dctx->virtualStart = (const char*)dst - ((const char*)(dctx->previousDstEnd) - (const char*)(dctx->prefixStart));
+ dctx->prefixStart = dst;
+ dctx->previousDstEnd = dst;
+ }
+}
+
+
+size_t ZSTD_decompressBlock_deprecated(ZSTD_DCtx* dctx,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize)
+{
+ size_t dSize;
+ dctx->isFrameDecompression = 0;
+ ZSTD_checkContinuity(dctx, dst, dstCapacity);
+ dSize = ZSTD_decompressBlock_internal(dctx, dst, dstCapacity, src, srcSize, not_streaming);
+ FORWARD_IF_ERROR(dSize, "");
+ dctx->previousDstEnd = (char*)dst + dSize;
+ return dSize;
+}
+
+
+/* NOTE: Must just wrap ZSTD_decompressBlock_deprecated() */
+size_t ZSTD_decompressBlock(ZSTD_DCtx* dctx,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize)
+{
+ return ZSTD_decompressBlock_deprecated(dctx, dst, dstCapacity, src, srcSize);
+}
diff --git a/third_party/zstd/lib/decompress/zstd_decompress_block.h b/third_party/zstd/lib/decompress/zstd_decompress_block.h
new file mode 100644
index 0000000000..ab152404ba
--- /dev/null
+++ b/third_party/zstd/lib/decompress/zstd_decompress_block.h
@@ -0,0 +1,73 @@
+/*
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+
+#ifndef ZSTD_DEC_BLOCK_H
+#define ZSTD_DEC_BLOCK_H
+
+/*-*******************************************************
+ * Dependencies
+ *********************************************************/
+#include "../common/zstd_deps.h" /* size_t */
+#include "../zstd.h" /* DCtx, and some public functions */
+#include "../common/zstd_internal.h" /* blockProperties_t, and some public functions */
+#include "zstd_decompress_internal.h" /* ZSTD_seqSymbol */
+
+
+/* === Prototypes === */
+
+/* note: prototypes already published within `zstd.h` :
+ * ZSTD_decompressBlock()
+ */
+
+/* note: prototypes already published within `zstd_internal.h` :
+ * ZSTD_getcBlockSize()
+ * ZSTD_decodeSeqHeaders()
+ */
+
+
+ /* Streaming state is used to inform allocation of the literal buffer */
+typedef enum {
+ not_streaming = 0,
+ is_streaming = 1
+} streaming_operation;
+
+/* ZSTD_decompressBlock_internal() :
+ * decompress block, starting at `src`,
+ * into destination buffer `dst`.
+ * @return : decompressed block size,
+ * or an error code (which can be tested using ZSTD_isError())
+ */
+size_t ZSTD_decompressBlock_internal(ZSTD_DCtx* dctx,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize, const streaming_operation streaming);
+
+/* ZSTD_buildFSETable() :
+ * generate FSE decoding table for one symbol (ll, ml or off)
+ * this function must be called with valid parameters only
+ * (dt is large enough, normalizedCounter distribution total is a power of 2, max is within range, etc.)
+ * in which case it cannot fail.
+ * The workspace must be 4-byte aligned and at least ZSTD_BUILD_FSE_TABLE_WKSP_SIZE bytes, which is
+ * defined in zstd_decompress_internal.h.
+ * Internal use only.
+ */
+void ZSTD_buildFSETable(ZSTD_seqSymbol* dt,
+ const short* normalizedCounter, unsigned maxSymbolValue,
+ const U32* baseValue, const U8* nbAdditionalBits,
+ unsigned tableLog, void* wksp, size_t wkspSize,
+ int bmi2);
+
+/* Internal definition of ZSTD_decompressBlock() to avoid deprecation warnings. */
+size_t ZSTD_decompressBlock_deprecated(ZSTD_DCtx* dctx,
+ void* dst, size_t dstCapacity,
+ const void* src, size_t srcSize);
+
+
+#endif /* ZSTD_DEC_BLOCK_H */
diff --git a/third_party/zstd/lib/decompress/zstd_decompress_internal.h b/third_party/zstd/lib/decompress/zstd_decompress_internal.h
new file mode 100644
index 0000000000..83a7a0115f
--- /dev/null
+++ b/third_party/zstd/lib/decompress/zstd_decompress_internal.h
@@ -0,0 +1,240 @@
+/*
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+
+/* zstd_decompress_internal:
+ * objects and definitions shared within lib/decompress modules */
+
+ #ifndef ZSTD_DECOMPRESS_INTERNAL_H
+ #define ZSTD_DECOMPRESS_INTERNAL_H
+
+
+/*-*******************************************************
+ * Dependencies
+ *********************************************************/
+#include "../common/mem.h" /* BYTE, U16, U32 */
+#include "../common/zstd_internal.h" /* constants : MaxLL, MaxML, MaxOff, LLFSELog, etc. */
+
+
+
+/*-*******************************************************
+ * Constants
+ *********************************************************/
+static UNUSED_ATTR const U32 LL_base[MaxLL+1] = {
+ 0, 1, 2, 3, 4, 5, 6, 7,
+ 8, 9, 10, 11, 12, 13, 14, 15,
+ 16, 18, 20, 22, 24, 28, 32, 40,
+ 48, 64, 0x80, 0x100, 0x200, 0x400, 0x800, 0x1000,
+ 0x2000, 0x4000, 0x8000, 0x10000 };
+
+static UNUSED_ATTR const U32 OF_base[MaxOff+1] = {
+ 0, 1, 1, 5, 0xD, 0x1D, 0x3D, 0x7D,
+ 0xFD, 0x1FD, 0x3FD, 0x7FD, 0xFFD, 0x1FFD, 0x3FFD, 0x7FFD,
+ 0xFFFD, 0x1FFFD, 0x3FFFD, 0x7FFFD, 0xFFFFD, 0x1FFFFD, 0x3FFFFD, 0x7FFFFD,
+ 0xFFFFFD, 0x1FFFFFD, 0x3FFFFFD, 0x7FFFFFD, 0xFFFFFFD, 0x1FFFFFFD, 0x3FFFFFFD, 0x7FFFFFFD };
+
+static UNUSED_ATTR const U8 OF_bits[MaxOff+1] = {
+ 0, 1, 2, 3, 4, 5, 6, 7,
+ 8, 9, 10, 11, 12, 13, 14, 15,
+ 16, 17, 18, 19, 20, 21, 22, 23,
+ 24, 25, 26, 27, 28, 29, 30, 31 };
+
+static UNUSED_ATTR const U32 ML_base[MaxML+1] = {
+ 3, 4, 5, 6, 7, 8, 9, 10,
+ 11, 12, 13, 14, 15, 16, 17, 18,
+ 19, 20, 21, 22, 23, 24, 25, 26,
+ 27, 28, 29, 30, 31, 32, 33, 34,
+ 35, 37, 39, 41, 43, 47, 51, 59,
+ 67, 83, 99, 0x83, 0x103, 0x203, 0x403, 0x803,
+ 0x1003, 0x2003, 0x4003, 0x8003, 0x10003 };
+
+
+/*-*******************************************************
+ * Decompression types
+ *********************************************************/
+ typedef struct {
+ U32 fastMode;
+ U32 tableLog;
+ } ZSTD_seqSymbol_header;
+
+ typedef struct {
+ U16 nextState;
+ BYTE nbAdditionalBits;
+ BYTE nbBits;
+ U32 baseValue;
+ } ZSTD_seqSymbol;
+
+ #define SEQSYMBOL_TABLE_SIZE(log) (1 + (1 << (log)))
+
+#define ZSTD_BUILD_FSE_TABLE_WKSP_SIZE (sizeof(S16) * (MaxSeq + 1) + (1u << MaxFSELog) + sizeof(U64))
+#define ZSTD_BUILD_FSE_TABLE_WKSP_SIZE_U32 ((ZSTD_BUILD_FSE_TABLE_WKSP_SIZE + sizeof(U32) - 1) / sizeof(U32))
+#define ZSTD_HUFFDTABLE_CAPACITY_LOG 12
+
+typedef struct {
+ ZSTD_seqSymbol LLTable[SEQSYMBOL_TABLE_SIZE(LLFSELog)]; /* Note : Space reserved for FSE Tables */
+ ZSTD_seqSymbol OFTable[SEQSYMBOL_TABLE_SIZE(OffFSELog)]; /* is also used as temporary workspace while building hufTable during DDict creation */
+ ZSTD_seqSymbol MLTable[SEQSYMBOL_TABLE_SIZE(MLFSELog)]; /* and therefore must be at least HUF_DECOMPRESS_WORKSPACE_SIZE large */
+ HUF_DTable hufTable[HUF_DTABLE_SIZE(ZSTD_HUFFDTABLE_CAPACITY_LOG)]; /* can accommodate HUF_decompress4X */
+ U32 rep[ZSTD_REP_NUM];
+ U32 workspace[ZSTD_BUILD_FSE_TABLE_WKSP_SIZE_U32];
+} ZSTD_entropyDTables_t;
+
+typedef enum { ZSTDds_getFrameHeaderSize, ZSTDds_decodeFrameHeader,
+ ZSTDds_decodeBlockHeader, ZSTDds_decompressBlock,
+ ZSTDds_decompressLastBlock, ZSTDds_checkChecksum,
+ ZSTDds_decodeSkippableHeader, ZSTDds_skipFrame } ZSTD_dStage;
+
+typedef enum { zdss_init=0, zdss_loadHeader,
+ zdss_read, zdss_load, zdss_flush } ZSTD_dStreamStage;
+
+typedef enum {
+ ZSTD_use_indefinitely = -1, /* Use the dictionary indefinitely */
+ ZSTD_dont_use = 0, /* Do not use the dictionary (if one exists free it) */
+ ZSTD_use_once = 1 /* Use the dictionary once and set to ZSTD_dont_use */
+} ZSTD_dictUses_e;
+
+/* Hashset for storing references to multiple ZSTD_DDict within ZSTD_DCtx */
+typedef struct {
+ const ZSTD_DDict** ddictPtrTable;
+ size_t ddictPtrTableSize;
+ size_t ddictPtrCount;
+} ZSTD_DDictHashSet;
+
+#ifndef ZSTD_DECODER_INTERNAL_BUFFER
+# define ZSTD_DECODER_INTERNAL_BUFFER (1 << 16)
+#endif
+
+#define ZSTD_LBMIN 64
+#define ZSTD_LBMAX (128 << 10)
+
+/* extra buffer, compensates when dst is not large enough to store litBuffer */
+#define ZSTD_LITBUFFEREXTRASIZE BOUNDED(ZSTD_LBMIN, ZSTD_DECODER_INTERNAL_BUFFER, ZSTD_LBMAX)
+
+typedef enum {
+ ZSTD_not_in_dst = 0, /* Stored entirely within litExtraBuffer */
+ ZSTD_in_dst = 1, /* Stored entirely within dst (in memory after current output write) */
+ ZSTD_split = 2 /* Split between litExtraBuffer and dst */
+} ZSTD_litLocation_e;
+
+struct ZSTD_DCtx_s
+{
+ const ZSTD_seqSymbol* LLTptr;
+ const ZSTD_seqSymbol* MLTptr;
+ const ZSTD_seqSymbol* OFTptr;
+ const HUF_DTable* HUFptr;
+ ZSTD_entropyDTables_t entropy;
+ U32 workspace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32]; /* space needed when building huffman tables */
+ const void* previousDstEnd; /* detect continuity */
+ const void* prefixStart; /* start of current segment */
+ const void* virtualStart; /* virtual start of previous segment if it was just before current one */
+ const void* dictEnd; /* end of previous segment */
+ size_t expected;
+ ZSTD_frameHeader fParams;
+ U64 processedCSize;
+ U64 decodedSize;
+ blockType_e bType; /* used in ZSTD_decompressContinue(), store blockType between block header decoding and block decompression stages */
+ ZSTD_dStage stage;
+ U32 litEntropy;
+ U32 fseEntropy;
+ XXH64_state_t xxhState;
+ size_t headerSize;
+ ZSTD_format_e format;
+ ZSTD_forceIgnoreChecksum_e forceIgnoreChecksum; /* User specified: if == 1, will ignore checksums in compressed frame. Default == 0 */
+ U32 validateChecksum; /* if == 1, will validate checksum. Is == 1 if (fParams.checksumFlag == 1) and (forceIgnoreChecksum == 0). */
+ const BYTE* litPtr;
+ ZSTD_customMem customMem;
+ size_t litSize;
+ size_t rleSize;
+ size_t staticSize;
+ int isFrameDecompression;
+#if DYNAMIC_BMI2 != 0
+ int bmi2; /* == 1 if the CPU supports BMI2 and 0 otherwise. CPU support is determined dynamically once per context lifetime. */
+#endif
+
+ /* dictionary */
+ ZSTD_DDict* ddictLocal;
+ const ZSTD_DDict* ddict; /* set by ZSTD_initDStream_usingDDict(), or ZSTD_DCtx_refDDict() */
+ U32 dictID;
+ int ddictIsCold; /* if == 1 : dictionary is "new" for working context, and presumed "cold" (not in cpu cache) */
+ ZSTD_dictUses_e dictUses;
+ ZSTD_DDictHashSet* ddictSet; /* Hash set for multiple ddicts */
+ ZSTD_refMultipleDDicts_e refMultipleDDicts; /* User specified: if == 1, will allow references to multiple DDicts. Default == 0 (disabled) */
+ int disableHufAsm;
+ int maxBlockSizeParam;
+
+ /* streaming */
+ ZSTD_dStreamStage streamStage;
+ char* inBuff;
+ size_t inBuffSize;
+ size_t inPos;
+ size_t maxWindowSize;
+ char* outBuff;
+ size_t outBuffSize;
+ size_t outStart;
+ size_t outEnd;
+ size_t lhSize;
+#if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT>=1)
+ void* legacyContext;
+ U32 previousLegacyVersion;
+ U32 legacyVersion;
+#endif
+ U32 hostageByte;
+ int noForwardProgress;
+ ZSTD_bufferMode_e outBufferMode;
+ ZSTD_outBuffer expectedOutBuffer;
+
+ /* workspace */
+ BYTE* litBuffer;
+ const BYTE* litBufferEnd;
+ ZSTD_litLocation_e litBufferLocation;
+ BYTE litExtraBuffer[ZSTD_LITBUFFEREXTRASIZE + WILDCOPY_OVERLENGTH]; /* literal buffer can be split between storage within dst and within this scratch buffer */
+ BYTE headerBuffer[ZSTD_FRAMEHEADERSIZE_MAX];
+
+ size_t oversizedDuration;
+
+#ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
+ void const* dictContentBeginForFuzzing;
+ void const* dictContentEndForFuzzing;
+#endif
+
+ /* Tracing */
+#if ZSTD_TRACE
+ ZSTD_TraceCtx traceCtx;
+#endif
+}; /* typedef'd to ZSTD_DCtx within "zstd.h" */
+
+MEM_STATIC int ZSTD_DCtx_get_bmi2(const struct ZSTD_DCtx_s *dctx) {
+#if DYNAMIC_BMI2 != 0
+ return dctx->bmi2;
+#else
+ (void)dctx;
+ return 0;
+#endif
+}
+
+/*-*******************************************************
+ * Shared internal functions
+ *********************************************************/
+
+/*! ZSTD_loadDEntropy() :
+ * dict : must point at beginning of a valid zstd dictionary.
+ * @return : size of dictionary header (size of magic number + dict ID + entropy tables) */
+size_t ZSTD_loadDEntropy(ZSTD_entropyDTables_t* entropy,
+ const void* const dict, size_t const dictSize);
+
+/*! ZSTD_checkContinuity() :
+ * check if next `dst` follows previous position, where decompression ended.
+ * If yes, do nothing (continue on current segment).
+ * If not, classify previous segment as "external dictionary", and start a new segment.
+ * This function cannot fail. */
+void ZSTD_checkContinuity(ZSTD_DCtx* dctx, const void* dst, size_t dstSize);
+
+
+#endif /* ZSTD_DECOMPRESS_INTERNAL_H */
generated by cgit v1.2.3 (git 2.39.1) at 2024-07-19 04:32:52 +0000