Merge pull request #859 from terrelln/31

Prepare for ZSTD_WINDOWLOG_MAX == 31

lib/common/bitstream.h

@@ -194,11 +194,14 @@ MEM_STATIC unsigned BIT_highbit32 (register U32 val)
 }
 
 /*=====    Local Constants   =====*/
-static const unsigned BIT_mask[] = { 0, 1, 3, 7, 0xF, 0x1F, 0x3F, 0x7F,
-                                    0xFF, 0x1FF, 0x3FF, 0x7FF, 0xFFF, 0x1FFF, 0x3FFF, 0x7FFF,
-                                    0xFFFF, 0x1FFFF, 0x3FFFF, 0x7FFFF, 0xFFFFF, 0x1FFFFF, 0x3FFFFF, 0x7FFFFF,
-                                    0xFFFFFF, 0x1FFFFFF, 0x3FFFFFF };   /* up to 26 bits */
-
+static const unsigned BIT_mask[] = {
+    0,          1,         3,         7,         0xF,       0x1F,
+    0x3F,       0x7F,      0xFF,      0x1FF,     0x3FF,     0x7FF,
+    0xFFF,      0x1FFF,    0x3FFF,    0x7FFF,    0xFFFF,    0x1FFFF,
+    0x3FFFF,    0x7FFFF,   0xFFFFF,   0x1FFFFF,  0x3FFFFF,  0x7FFFFF,
+    0xFFFFFF,   0x1FFFFFF, 0x3FFFFFF, 0x7FFFFFF, 0xFFFFFFF, 0x1FFFFFFF,
+    0x3FFFFFFF, 0x7FFFFFFF}; /* up to 31 bits */
+#define BIT_MASK_SIZE (sizeof(BIT_mask) / sizeof(BIT_mask[0]))
 
 /*-**************************************************************
 *  bitStream encoding
@@ -220,11 +223,14 @@ MEM_STATIC size_t BIT_initCStream(BIT_CStream_t* bitC,
 }
 
 /*! BIT_addBits() :
- *  can add up to 26 bits into `bitC`.
+ *  can add up to 31 bits into `bitC`.
  *  Note : does not check for register overflow ! */
 MEM_STATIC void BIT_addBits(BIT_CStream_t* bitC,
                             size_t value, unsigned nbBits)
 {
+    MEM_STATIC_ASSERT(BIT_MASK_SIZE == 32);
+    assert(nbBits < BIT_MASK_SIZE);
+    assert(nbBits + bitC->bitPos < sizeof(bitC->bitContainer) * 8);
     bitC->bitContainer |= (value & BIT_mask[nbBits]) << bitC->bitPos;
     bitC->bitPos += nbBits;
 }
@@ -235,6 +241,7 @@ MEM_STATIC void BIT_addBitsFast(BIT_CStream_t* bitC,
                                 size_t value, unsigned nbBits)
 {
     assert((value>>nbBits) == 0);
+    assert(nbBits + bitC->bitPos < sizeof(bitC->bitContainer) * 8);
     bitC->bitContainer |= value << bitC->bitPos;
     bitC->bitPos += nbBits;
 }
@@ -245,7 +252,7 @@ MEM_STATIC void BIT_addBitsFast(BIT_CStream_t* bitC,
 MEM_STATIC void BIT_flushBitsFast(BIT_CStream_t* bitC)
 {
     size_t const nbBytes = bitC->bitPos >> 3;
-    assert( bitC->bitPos <= (sizeof(bitC->bitContainer)*8) );
+    assert(bitC->bitPos < sizeof(bitC->bitContainer) * 8);
     MEM_writeLEST(bitC->ptr, bitC->bitContainer);
     bitC->ptr += nbBytes;
     assert(bitC->ptr <= bitC->endPtr);
@@ -261,7 +268,7 @@ MEM_STATIC void BIT_flushBitsFast(BIT_CStream_t* bitC)
 MEM_STATIC void BIT_flushBits(BIT_CStream_t* bitC)
 {
     size_t const nbBytes = bitC->bitPos >> 3;
-    assert( bitC->bitPos <= (sizeof(bitC->bitContainer)*8) );
+    assert(bitC->bitPos < sizeof(bitC->bitContainer) * 8);
     MEM_writeLEST(bitC->ptr, bitC->bitContainer);
     bitC->ptr += nbBytes;
     if (bitC->ptr > bitC->endPtr) bitC->ptr = bitC->endPtr;
@@ -353,12 +360,14 @@ MEM_STATIC size_t BIT_getMiddleBits(size_t bitContainer, U32 const start, U32 co
 #  endif
         return _bextr_u32(bitContainer, start, nbBits);
 #else
+    assert(nbBits < BIT_MASK_SIZE);
     return (bitContainer >> start) & BIT_mask[nbBits];
 #endif
 }
 
 MEM_STATIC size_t BIT_getLowerBits(size_t bitContainer, U32 const nbBits)
 {
+    assert(nbBits < BIT_MASK_SIZE);
     return bitContainer & BIT_mask[nbBits];
 }
 

lib/compress/fse_compress.c

@@ -461,6 +461,7 @@ static unsigned FSE_minTableLog(size_t srcSize, unsigned maxSymbolValue)
     U32 minBitsSrc = BIT_highbit32((U32)(srcSize - 1)) + 1;
     U32 minBitsSymbols = BIT_highbit32(maxSymbolValue) + 2;
     U32 minBits = minBitsSrc < minBitsSymbols ? minBitsSrc : minBitsSymbols;
+    assert(srcSize > 1); /* Not supported, RLE should be used instead */
     return minBits;
 }
 
@@ -469,6 +470,7 @@ unsigned FSE_optimalTableLog_internal(unsigned maxTableLog, size_t srcSize, unsi
     U32 maxBitsSrc = BIT_highbit32((U32)(srcSize - 1)) - minus;
     U32 tableLog = maxTableLog;
     U32 minBits = FSE_minTableLog(srcSize, maxSymbolValue);
+    assert(srcSize > 1); /* Not supported, RLE should be used instead */
     if (tableLog==0) tableLog = FSE_DEFAULT_TABLELOG;
     if (maxBitsSrc < tableLog) tableLog = maxBitsSrc;   /* Accuracy can be reduced */
     if (minBits > tableLog) tableLog = minBits;   /* Need a minimum to safely represent all symbol values */

lib/compress/zstd_compress.c

@@ -675,7 +675,7 @@ size_t ZSTD_estimateCCtxSize_advanced_usingCCtxParams(const ZSTD_CCtx_params* pa
         size_t const maxNbSeq = blockSize / divider;
         size_t const tokenSpace = blockSize + 11*maxNbSeq;
         size_t const chainSize =
-                (cParams.strategy == ZSTD_fast) ? 0 : (1 << cParams.chainLog);
+                (cParams.strategy == ZSTD_fast) ? 0 : ((size_t)1 << cParams.chainLog);
         size_t const hSize = ((size_t)1) << cParams.hashLog;
         U32    const hashLog3 = (cParams.searchLength>3) ?
                                 0 : MIN(ZSTD_HASHLOG3_MAX, cParams.windowLog);
@@ -833,7 +833,7 @@ static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc,
         size_t const maxNbSeq = blockSize / divider;
         size_t const tokenSpace = blockSize + 11*maxNbSeq;
         size_t const chainSize = (params.cParams.strategy == ZSTD_fast) ?
-                                0 : (1 << params.cParams.chainLog);
+                                0 : ((size_t)1 << params.cParams.chainLog);
         size_t const hSize = ((size_t)1) << params.cParams.hashLog;
         U32    const hashLog3 = (params.cParams.searchLength>3) ?
                                 0 : MIN(ZSTD_HASHLOG3_MAX, params.cParams.windowLog);
@@ -1005,7 +1005,7 @@ static size_t ZSTD_copyCCtx_internal(ZSTD_CCtx* dstCCtx,
     }
 
     /* copy tables */
-    {   size_t const chainSize = (srcCCtx->appliedParams.cParams.strategy == ZSTD_fast) ? 0 : (1 << srcCCtx->appliedParams.cParams.chainLog);
+    {   size_t const chainSize = (srcCCtx->appliedParams.cParams.strategy == ZSTD_fast) ? 0 : ((size_t)1 << srcCCtx->appliedParams.cParams.chainLog);
         size_t const hSize =  (size_t)1 << srcCCtx->appliedParams.cParams.hashLog;
         size_t const h3Size = (size_t)1 << srcCCtx->hashLog3;
         size_t const tableSpace = (chainSize + hSize + h3Size) * sizeof(U32);
@@ -1599,13 +1599,33 @@ static size_t ZSTD_compress_frameChunk (ZSTD_CCtx* cctx,
             return ERROR(dstSize_tooSmall);   /* not enough space to store compressed block */
         if (remaining < blockSize) blockSize = remaining;
 
-        /* preemptive overflow correction */
+        /* preemptive overflow correction:
+         * 1. correction is large enough:
+         *    lowLimit > (3<<29) ==> current > 3<<29 + 1<<windowLog - blockSize
+         *    1<<windowLog <= newCurrent < 1<<chainLog + 1<<windowLog
+         *
+         *    current - newCurrent
+         *    > (3<<29 + 1<<windowLog - blockSize) - (1<<windowLog + 1<<chainLog)
+         *    > (3<<29 - blockSize) - (1<<chainLog)
+         *    > (3<<29 - blockSize) - (1<<30)             (NOTE: chainLog <= 30)
+         *    > 1<<29 - 1<<17
+         *
+         * 2. (ip+blockSize - cctx->base) doesn't overflow:
+         *    In 32 bit mode we limit windowLog to 30 so we don't get
+         *    differences larger than 1<<31-1.
+         * 3. cctx->lowLimit < 1<<32:
+         *    windowLog <= 31 ==> 3<<29 + 1<<windowLog < 7<<29 < 1<<32.
+         */
         if (cctx->lowLimit > (3U<<29)) {
-            U32 const cycleMask = (1 << ZSTD_cycleLog(cctx->appliedParams.cParams.hashLog, cctx->appliedParams.cParams.strategy)) - 1;
+            U32 const cycleMask = (1 << ZSTD_cycleLog(cctx->appliedParams.cParams.chainLog, cctx->appliedParams.cParams.strategy)) - 1;
             U32 const current = (U32)(ip - cctx->base);
             U32 const newCurrent = (current & cycleMask) + (1 << cctx->appliedParams.cParams.windowLog);
             U32 const correction = current - newCurrent;
-            ZSTD_STATIC_ASSERT(ZSTD_WINDOWLOG_MAX_64 <= 30);
+            ZSTD_STATIC_ASSERT(ZSTD_CHAINLOG_MAX <= 30);
+            ZSTD_STATIC_ASSERT(ZSTD_WINDOWLOG_MAX_32 <= 30);
+            ZSTD_STATIC_ASSERT(ZSTD_WINDOWLOG_MAX <= 31);
+            assert(current > newCurrent);
+            assert(correction > 1<<28); /* Loose bound, should be about 1<<29 */
             ZSTD_reduceIndex(cctx, correction);
             cctx->base += correction;
             cctx->dictBase += correction;
@@ -1613,6 +1633,7 @@ static size_t ZSTD_compress_frameChunk (ZSTD_CCtx* cctx,
             cctx->dictLimit -= correction;
             if (cctx->nextToUpdate < correction) cctx->nextToUpdate = 0;
             else cctx->nextToUpdate -= correction;
+            DEBUGLOG(4, "Correction of 0x%x bytes to lowLimit=0x%x\n", correction, cctx->lowLimit);
         }
 
         if ((U32)(ip+blockSize - cctx->base) > cctx->loadedDictEnd + maxDist) {

lib/compress/zstd_ldm.c

@@ -295,7 +295,7 @@ size_t ZSTD_compressBlock_ldm_generic(ZSTD_CCtx* cctx,
     const U32   lowestIndex = cctx->dictLimit;
     const BYTE* const lowest = base + lowestIndex;
     const BYTE* const iend = istart + srcSize;
-    const BYTE* const ilimit = iend - ldmParams.minMatchLength;
+    const BYTE* const ilimit = iend - MAX(ldmParams.minMatchLength, HASH_READ_SIZE);
 
     const ZSTD_blockCompressor blockCompressor =
         ZSTD_selectBlockCompressor(cctx->appliedParams.cParams.strategy, 0);
@@ -499,7 +499,7 @@ static size_t ZSTD_compressBlock_ldm_extDict_generic(
     const BYTE* const lowPrefixPtr = base + dictLimit;
     const BYTE* const dictEnd = dictBase + dictLimit;
     const BYTE* const iend = istart + srcSize;
-    const BYTE* const ilimit = iend - ldmParams.minMatchLength;
+    const BYTE* const ilimit = iend - MAX(ldmParams.minMatchLength, HASH_READ_SIZE);
 
     const ZSTD_blockCompressor blockCompressor =
         ZSTD_selectBlockCompressor(ctx->appliedParams.cParams.strategy, 1);

lib/decompress/zstd_decompress.c

@@ -35,7 +35,7 @@
 *  Frames requiring more memory will be rejected.
 */
 #ifndef ZSTD_MAXWINDOWSIZE_DEFAULT
-#  define ZSTD_MAXWINDOWSIZE_DEFAULT ((1 << ZSTD_WINDOWLOG_MAX) + 1)   /* defined within zstd.h */
+#  define ZSTD_MAXWINDOWSIZE_DEFAULT (((U64)1 << ZSTD_WINDOWLOG_MAX) + 1)   /* defined within zstd.h */
 #endif
 
 
@@ -1372,8 +1372,9 @@ static size_t ZSTD_decompressBlock_internal(ZSTD_DCtx* dctx,
      * conservative.
      */
     ZSTD_longOffset_e const isLongOffset = (ZSTD_longOffset_e)(MEM_32bits() && (!frame || dctx->fParams.windowSize > (1ULL << STREAM_ACCUMULATOR_MIN)));
-    /* We don't expect window sizes this big. */
-    assert(!frame || dctx->fParams.windowSize <= (1ULL << STREAM_ACCUMULATOR_MIN_64));
+    /* windowSize could be any value at this point, since it is only validated
+     * in the streaming API.
+     */
     DEBUGLOG(5, "ZSTD_decompressBlock_internal");
 
     if (srcSize >= ZSTD_BLOCKSIZE_MAX) return ERROR(srcSize_wrong);
@@ -1545,6 +1546,10 @@ static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx,
         if (blockProperties.lastBlock) break;
     }
 
+    if (dctx->fParams.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN) {
+        if ((U64)(op-ostart) != dctx->fParams.frameContentSize) {
+            return ERROR(corruption_detected);
+    }   }
     if (dctx->fParams.checksumFlag) { /* Frame content checksum verification */
         U32 const checkCalc = (U32)XXH64_digest(&dctx->xxhState);
         U32 checkRead;

lib/zstd.h

@@ -380,9 +380,9 @@ ZSTDLIB_API size_t ZSTD_DStreamOutSize(void);   /*!< recommended size for output
 #define ZSTD_WINDOWLOG_MAX_64  27
 #define ZSTD_WINDOWLOG_MAX    ((unsigned)(sizeof(size_t) == 4 ? ZSTD_WINDOWLOG_MAX_32 : ZSTD_WINDOWLOG_MAX_64))
 #define ZSTD_WINDOWLOG_MIN     10
-#define ZSTD_HASHLOG_MAX       ZSTD_WINDOWLOG_MAX
+#define ZSTD_HASHLOG_MAX       MIN(ZSTD_WINDOWLOG_MAX, 30)
 #define ZSTD_HASHLOG_MIN        6
-#define ZSTD_CHAINLOG_MAX     (ZSTD_WINDOWLOG_MAX+1)
+#define ZSTD_CHAINLOG_MAX      MIN(ZSTD_WINDOWLOG_MAX+1, 30)
 #define ZSTD_CHAINLOG_MIN      ZSTD_HASHLOG_MIN
 #define ZSTD_HASHLOG3_MAX      17
 #define ZSTD_SEARCHLOG_MAX    (ZSTD_WINDOWLOG_MAX-1)

tests/fuzzer.c

@@ -811,7 +811,7 @@ static int basicUnitTests(U32 seed, double compressibility)
         }
         DISPLAYLEVEL(4, "OK \n");
 
-        DISPLAYLEVEL(4, "test%3i : Loading rawContent starting with dict header w/ ZSTD_dm_auto should fail", testNb++);
+        DISPLAYLEVEL(4, "test%3i : Loading rawContent starting with dict header w/ ZSTD_dm_auto should fail : ", testNb++);
         {
             size_t ret;
             MEM_writeLE32((char*)dictBuffer+2, ZSTD_MAGIC_DICTIONARY);
@@ -821,7 +821,7 @@ static int basicUnitTests(U32 seed, double compressibility)
         }
         DISPLAYLEVEL(4, "OK \n");
 
-        DISPLAYLEVEL(4, "test%3i : Loading rawContent starting with dict header w/ ZSTD_dm_rawContent should pass", testNb++);
+        DISPLAYLEVEL(4, "test%3i : Loading rawContent starting with dict header w/ ZSTD_dm_rawContent should pass : ", testNb++);
         {
             size_t ret;
             MEM_writeLE32((char*)dictBuffer+2, ZSTD_MAGIC_DICTIONARY);