v1.4.3: Merge pull request #1730 from facebook/dev

facebook · Aug 19, 2019 · a3d655d · a3d655d
2 parents ff304e9 + c9072ee
commit a3d655d
Show file tree

Hide file tree

Showing 24 changed files with 300 additions and 177 deletions.
diff --git a/.travis.yml b/.travis.yml
@@ -123,6 +123,10 @@ matrix:
         - make ppcinstall
         - make ppcfuzz
 
+    - name: Trusty (Versions Compatibility Test)
+      script:
+        - make -C tests versionsTest
+
     # check release number
     - name: Tag-Specific Test
       if: tag =~ ^v[0-9]\.[0-9]

diff --git a/CHANGELOG b/CHANGELOG
@@ -1,3 +1,9 @@
+v1.4.3
+bug: Fix Dictionary Compression Ratio Regression by @cyan4973 (#1709)
+bug: Fix Buffer Overflow in v0.3 Decompression by @felixhandte (#1722)
+build: Add support for IAR C/C++ Compiler for Arm by @joseph0918 (#1705)
+misc: Add NULL pointer check in util.c by @leeyoung624 (#1706)
+
 v1.4.2
 bug: Fix bug in zstd-0.5 decoder by @terrelln (#1696)
 bug: Fix seekable decompression in-memory API by @iburinoc (#1695)

diff --git a/doc/zstd_manual.html b/doc/zstd_manual.html
@@ -1,10 +1,10 @@
 <html>
 <head>
 <meta http-equiv="Content-Type" content="text/html; charset=ISO-8859-1">
-<title>zstd 1.4.2 Manual</title>
+<title>zstd 1.4.3 Manual</title>
 </head>
 <body>
-<h1>zstd 1.4.2 Manual</h1>
+<h1>zstd 1.4.3 Manual</h1>
 <hr>
 <a name="Contents"></a><h2>Contents</h2>
 <ol>
@@ -1557,7 +1557,7 @@ <h3>Buffer-less streaming decompression functions</h3><pre></pre><b><pre>typedef
 <a name="Chapter23"></a><h2>Block level API</h2><pre></pre>
 
 <pre><b></b><p>    Frame metadata cost is typically ~18 bytes, which can be non-negligible for very small blocks (< 100 bytes).
-    User will have to take in charge required information to regenerate data, such as compressed and content sizes.
+    But users will have to take in charge needed metadata to regenerate data, such as compressed and content sizes.
 
     A few rules to respect :
     - Compressing and decompressing require a context structure
@@ -1568,12 +1568,14 @@ <h3>Buffer-less streaming decompression functions</h3><pre></pre><b><pre>typedef
       + copyCCtx() and copyDCtx() can be used too
     - Block size is limited, it must be <= ZSTD_getBlockSize() <= ZSTD_BLOCKSIZE_MAX == 128 KB
       + If input is larger than a block size, it's necessary to split input data into multiple blocks
-      + For inputs larger than a single block, really consider using regular ZSTD_compress() instead.
-        Frame metadata is not that costly, and quickly becomes negligible as source size grows larger.
-    - When a block is considered not compressible enough, ZSTD_compressBlock() result will be zero.
-      In which case, nothing is produced into `dst` !
-      + User must test for such outcome and deal directly with uncompressed data
-      + ZSTD_decompressBlock() doesn't accept uncompressed data as input !!!
+      + For inputs larger than a single block, consider using regular ZSTD_compress() instead.
+        Frame metadata is not that costly, and quickly becomes negligible as source size grows larger than a block.
+    - When a block is considered not compressible enough, ZSTD_compressBlock() result will be 0 (zero) !
+      ===> In which case, nothing is produced into `dst` !
+      + User __must__ test for such outcome and deal directly with uncompressed data
+      + A block cannot be declared incompressible if ZSTD_compressBlock() return value was != 0.
+        Doing so would mess up with statistics history, leading to potential data corruption.
+      + ZSTD_decompressBlock() _doesn't accept uncompressed data as input_ !!
       + In case of multiple successive blocks, should some of them be uncompressed,
         decoder must be informed of their existence in order to follow proper history.
         Use ZSTD_insertBlock() for such a case.

diff --git a/lib/common/bitstream.h b/lib/common/bitstream.h
@@ -57,6 +57,8 @@ extern "C" {
 =========================================*/
 #if defined(__BMI__) && defined(__GNUC__)
 #  include <immintrin.h>   /* support for bextr (experimental) */
+#elif defined(__ICCARM__)
+#  include <intrinsics.h>
 #endif
 
 #define STREAM_ACCUMULATOR_MIN_32  25
@@ -163,6 +165,8 @@ MEM_STATIC unsigned BIT_highbit32 (U32 val)
         return (unsigned) r;
 #   elif defined(__GNUC__) && (__GNUC__ >= 3)   /* Use GCC Intrinsic */
         return 31 - __builtin_clz (val);
+#   elif defined(__ICCARM__)    /* IAR Intrinsic */
+        return 31 - __CLZ(val);
 #   else   /* Software version */
         static const unsigned DeBruijnClz[32] = { 0,  9,  1, 10, 13, 21,  2, 29,
                                                  11, 14, 16, 18, 22, 25,  3, 30,

diff --git a/lib/common/compiler.h b/lib/common/compiler.h
@@ -23,7 +23,7 @@
 #  define INLINE_KEYWORD
 #endif
 
-#if defined(__GNUC__)
+#if defined(__GNUC__) || defined(__ICCARM__)
 #  define FORCE_INLINE_ATTR __attribute__((always_inline))
 #elif defined(_MSC_VER)
 #  define FORCE_INLINE_ATTR __forceinline
@@ -65,7 +65,7 @@
 #ifdef _MSC_VER
 #  define FORCE_NOINLINE static __declspec(noinline)
 #else
-#  ifdef __GNUC__
+#  if defined(__GNUC__) || defined(__ICCARM__)
 #    define FORCE_NOINLINE static __attribute__((__noinline__))
 #  else
 #    define FORCE_NOINLINE static
@@ -76,7 +76,7 @@
 #ifndef __has_attribute
   #define __has_attribute(x) 0  /* Compatibility with non-clang compilers. */
 #endif
-#if defined(__GNUC__)
+#if defined(__GNUC__) || defined(__ICCARM__)
 #  define TARGET_ATTRIBUTE(target) __attribute__((__target__(target)))
 #else
 #  define TARGET_ATTRIBUTE(target)

diff --git a/lib/common/mem.h b/lib/common/mem.h
@@ -102,7 +102,7 @@ MEM_STATIC void MEM_check(void) { MEM_STATIC_ASSERT((sizeof(size_t)==4) || (size
 #ifndef MEM_FORCE_MEMORY_ACCESS   /* can be defined externally, on command line for example */
 #  if defined(__GNUC__) && ( defined(__ARM_ARCH_6__) || defined(__ARM_ARCH_6J__) || defined(__ARM_ARCH_6K__) || defined(__ARM_ARCH_6Z__) || defined(__ARM_ARCH_6ZK__) || defined(__ARM_ARCH_6T2__) )
 #    define MEM_FORCE_MEMORY_ACCESS 2
-#  elif defined(__INTEL_COMPILER) || defined(__GNUC__)
+#  elif defined(__INTEL_COMPILER) || defined(__GNUC__) || defined(__ICCARM__)
 #    define MEM_FORCE_MEMORY_ACCESS 1
 #  endif
 #endif

diff --git a/lib/common/xxhash.c b/lib/common/xxhash.c
@@ -53,7 +53,8 @@
 #  if defined(__GNUC__) && ( defined(__ARM_ARCH_6__) || defined(__ARM_ARCH_6J__) || defined(__ARM_ARCH_6K__) || defined(__ARM_ARCH_6Z__) || defined(__ARM_ARCH_6ZK__) || defined(__ARM_ARCH_6T2__) )
 #    define XXH_FORCE_MEMORY_ACCESS 2
 #  elif (defined(__INTEL_COMPILER) && !defined(WIN32)) || \
-  (defined(__GNUC__) && ( defined(__ARM_ARCH_7__) || defined(__ARM_ARCH_7A__) || defined(__ARM_ARCH_7R__) || defined(__ARM_ARCH_7M__) || defined(__ARM_ARCH_7S__) ))
+  (defined(__GNUC__) && ( defined(__ARM_ARCH_7__) || defined(__ARM_ARCH_7A__) || defined(__ARM_ARCH_7R__) || defined(__ARM_ARCH_7M__) || defined(__ARM_ARCH_7S__) )) || \
+  defined(__ICCARM__)
 #    define XXH_FORCE_MEMORY_ACCESS 1
 #  endif
 #endif
@@ -120,7 +121,7 @@ static void* XXH_memcpy(void* dest, const void* src, size_t size) { return memcp
 #  define INLINE_KEYWORD
 #endif
 
-#if defined(__GNUC__)
+#if defined(__GNUC__) || defined(__ICCARM__)
 #  define FORCE_INLINE_ATTR __attribute__((always_inline))
 #elif defined(_MSC_VER)
 #  define FORCE_INLINE_ATTR __forceinline
@@ -206,7 +207,12 @@ static U64 XXH_read64(const void* memPtr)
 #  define XXH_rotl32(x,r) _rotl(x,r)
 #  define XXH_rotl64(x,r) _rotl64(x,r)
 #else
+#if defined(__ICCARM__)
+#  include <intrinsics.h>
+#  define XXH_rotl32(x,r) __ROR(x,(32 - r))
+#else
 #  define XXH_rotl32(x,r) ((x << r) | (x >> (32 - r)))
+#endif
 #  define XXH_rotl64(x,r) ((x << r) | (x >> (64 - r)))
 #endif
 

diff --git a/lib/common/zstd_internal.h b/lib/common/zstd_internal.h
@@ -56,9 +56,9 @@ extern "C" {
 /**
  * Return the specified error if the condition evaluates to true.
  *
- * In debug modes, prints additional information. In order to do that
- * (particularly, printing the conditional that failed), this can't just wrap
- * RETURN_ERROR().
+ * In debug modes, prints additional information.
+ * In order to do that (particularly, printing the conditional that failed),
+ * this can't just wrap RETURN_ERROR().
  */
 #define RETURN_ERROR_IF(cond, err, ...) \
   if (cond) { \
@@ -324,6 +324,8 @@ MEM_STATIC U32 ZSTD_highbit32(U32 val)   /* compress, dictBuilder, decodeCorpus
         return (unsigned)r;
 #   elif defined(__GNUC__) && (__GNUC__ >= 3)   /* GCC Intrinsic */
         return 31 - __builtin_clz(val);
+#   elif defined(__ICCARM__)    /* IAR Intrinsic */
+        return 31 - __CLZ(val);
 #   else   /* Software version */
         static const U32 DeBruijnClz[32] = { 0, 9, 1, 10, 13, 21, 2, 29, 11, 14, 16, 18, 22, 25, 3, 30, 8, 12, 20, 28, 15, 17, 24, 7, 19, 27, 23, 6, 26, 5, 4, 31 };
         U32 v = val;

diff --git a/lib/compress/zstd_compress.c b/lib/compress/zstd_compress.c
@@ -1955,7 +1955,7 @@ ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
     BYTE* const ostart = (BYTE*)dst;
     BYTE* const oend = ostart + dstCapacity;
     BYTE* op = ostart;
-    size_t const nbSeq = seqStorePtr->sequences - seqStorePtr->sequencesStart;
+    size_t const nbSeq = (size_t)(seqStorePtr->sequences - seqStorePtr->sequencesStart);
     BYTE* seqHead;
     BYTE* lastNCount = NULL;
 
@@ -1964,7 +1964,7 @@ ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
 
     /* Compress literals */
     {   const BYTE* const literals = seqStorePtr->litStart;
-        size_t const litSize = seqStorePtr->lit - literals;
+        size_t const litSize = (size_t)(seqStorePtr->lit - literals);
         size_t const cSize = ZSTD_compressLiterals(
                                     &prevEntropy->huf, &nextEntropy->huf,
                                     cctxParams->cParams.strategy,
@@ -1991,7 +1991,7 @@ ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
     if (nbSeq==0) {
         /* Copy the old tables over as if we repeated them */
         memcpy(&nextEntropy->fse, &prevEntropy->fse, sizeof(prevEntropy->fse));
-        return op - ostart;
+        return (size_t)(op - ostart);
     }
 
     /* seqHead : flags for FSE encoding type */
@@ -2012,7 +2012,7 @@ ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
                                         ZSTD_defaultAllowed, strategy);
         assert(set_basic < set_compressed && set_rle < set_compressed);
         assert(!(LLtype < set_compressed && nextEntropy->fse.litlength_repeatMode != FSE_repeat_none)); /* We don't copy tables */
-        {   size_t const countSize = ZSTD_buildCTable(op, oend - op, CTable_LitLength, LLFSELog, (symbolEncodingType_e)LLtype,
+        {   size_t const countSize = ZSTD_buildCTable(op, (size_t)(oend - op), CTable_LitLength, LLFSELog, (symbolEncodingType_e)LLtype,
                                                     count, max, llCodeTable, nbSeq, LL_defaultNorm, LL_defaultNormLog, MaxLL,
                                                     prevEntropy->fse.litlengthCTable, sizeof(prevEntropy->fse.litlengthCTable),
                                                     workspace, wkspSize);
@@ -2035,7 +2035,7 @@ ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
                                         OF_defaultNorm, OF_defaultNormLog,
                                         defaultPolicy, strategy);
         assert(!(Offtype < set_compressed && nextEntropy->fse.offcode_repeatMode != FSE_repeat_none)); /* We don't copy tables */
-        {   size_t const countSize = ZSTD_buildCTable(op, oend - op, CTable_OffsetBits, OffFSELog, (symbolEncodingType_e)Offtype,
+        {   size_t const countSize = ZSTD_buildCTable(op, (size_t)(oend - op), CTable_OffsetBits, OffFSELog, (symbolEncodingType_e)Offtype,
                                                     count, max, ofCodeTable, nbSeq, OF_defaultNorm, OF_defaultNormLog, DefaultMaxOff,
                                                     prevEntropy->fse.offcodeCTable, sizeof(prevEntropy->fse.offcodeCTable),
                                                     workspace, wkspSize);
@@ -2056,7 +2056,7 @@ ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
                                         ML_defaultNorm, ML_defaultNormLog,
                                         ZSTD_defaultAllowed, strategy);
         assert(!(MLtype < set_compressed && nextEntropy->fse.matchlength_repeatMode != FSE_repeat_none)); /* We don't copy tables */
-        {   size_t const countSize = ZSTD_buildCTable(op, oend - op, CTable_MatchLength, MLFSELog, (symbolEncodingType_e)MLtype,
+        {   size_t const countSize = ZSTD_buildCTable(op, (size_t)(oend - op), CTable_MatchLength, MLFSELog, (symbolEncodingType_e)MLtype,
                                                     count, max, mlCodeTable, nbSeq, ML_defaultNorm, ML_defaultNormLog, MaxML,
                                                     prevEntropy->fse.matchlengthCTable, sizeof(prevEntropy->fse.matchlengthCTable),
                                                     workspace, wkspSize);
@@ -2070,7 +2070,7 @@ ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
     *seqHead = (BYTE)((LLtype<<6) + (Offtype<<4) + (MLtype<<2));
 
     {   size_t const bitstreamSize = ZSTD_encodeSequences(
-                                        op, oend - op,
+                                        op, (size_t)(oend - op),
                                         CTable_MatchLength, mlCodeTable,
                                         CTable_OffsetBits, ofCodeTable,
                                         CTable_LitLength, llCodeTable,
@@ -2097,7 +2097,7 @@ ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
     }
 
     DEBUGLOG(5, "compressed block size : %u", (unsigned)(op - ostart));
-    return op - ostart;
+    return (size_t)(op - ostart);
 }
 
 MEM_STATIC size_t
@@ -2270,7 +2270,8 @@ static size_t ZSTD_compressBlock_internal(ZSTD_CCtx* zc,
 {
     size_t cSize;
     DEBUGLOG(5, "ZSTD_compressBlock_internal (dstCapacity=%u, dictLimit=%u, nextToUpdate=%u)",
-                (unsigned)dstCapacity, (unsigned)zc->blockState.matchState.window.dictLimit, (unsigned)zc->blockState.matchState.nextToUpdate);
+                (unsigned)dstCapacity, (unsigned)zc->blockState.matchState.window.dictLimit,
+                (unsigned)zc->blockState.matchState.nextToUpdate);
 
     {   const size_t bss = ZSTD_buildSeqStore(zc, src, srcSize);
         FORWARD_IF_ERROR(bss);
@@ -2538,8 +2539,9 @@ size_t ZSTD_getBlockSize(const ZSTD_CCtx* cctx)
 
 size_t ZSTD_compressBlock(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize)
 {
-    size_t const blockSizeMax = ZSTD_getBlockSize(cctx);
-    RETURN_ERROR_IF(srcSize > blockSizeMax, srcSize_wrong);
+    DEBUGLOG(5, "ZSTD_compressBlock: srcSize = %u", (unsigned)srcSize);
+    { size_t const blockSizeMax = ZSTD_getBlockSize(cctx);
+      RETURN_ERROR_IF(srcSize > blockSizeMax, srcSize_wrong); }
 
     return ZSTD_compressContinue_internal(cctx, dst, dstCapacity, src, srcSize, 0 /* frame mode */, 0 /* last chunk */);
 }
@@ -2564,7 +2566,7 @@ static size_t ZSTD_loadDictionaryContent(ZSTD_matchState_t* ms,
     if (srcSize <= HASH_READ_SIZE) return 0;
 
     while (iend - ip > HASH_READ_SIZE) {
-        size_t const remaining = iend - ip;
+        size_t const remaining = (size_t)(iend - ip);
         size_t const chunk = MIN(remaining, ZSTD_CHUNKSIZE_MAX);
         const BYTE* const ichunk = ip + chunk;
 

diff --git a/lib/compress/zstd_compress_internal.h b/lib/compress/zstd_compress_internal.h
@@ -134,9 +134,15 @@ typedef struct {
 typedef struct ZSTD_matchState_t ZSTD_matchState_t;
 struct ZSTD_matchState_t {
     ZSTD_window_t window;   /* State for window round buffer management */
-    U32 loadedDictEnd;      /* index of end of dictionary, within context's referential. When dict referential is copied into active context (i.e. not attached), effectively same value as dictSize, since referential starts from zero */
+    U32 loadedDictEnd;      /* index of end of dictionary, within context's referential.
+                             * When loadedDictEnd != 0, a dictionary is in use, and still valid.
+                             * This relies on a mechanism to set loadedDictEnd=0 when dictionary is no longer within distance.
+                             * Such mechanism is provided within ZSTD_window_enforceMaxDist() and ZSTD_checkDictValidity().
+                             * When dict referential is copied into active context (i.e. not attached),
+                             * loadedDictEnd == dictSize, since referential starts from zero.
+                             */
     U32 nextToUpdate;       /* index from which to continue table update */
-    U32 hashLog3;           /* dispatch table : larger == faster, more memory */
+    U32 hashLog3;           /* dispatch table for matches of len==3 : larger == faster, more memory */
     U32* hashTable;
     U32* hashTable3;
     U32* chainTable;
@@ -350,7 +356,7 @@ MEM_STATIC void ZSTD_storeSeq(seqStore_t* seqStorePtr, size_t litLength, const v
     /* copy Literals */
     assert(seqStorePtr->maxNbLit <= 128 KB);
     assert(seqStorePtr->lit + litLength <= seqStorePtr->litStart + seqStorePtr->maxNbLit);
-    ZSTD_wildcopy(seqStorePtr->lit, literals, litLength, ZSTD_no_overlap);
+    ZSTD_wildcopy(seqStorePtr->lit, literals, (ptrdiff_t)litLength, ZSTD_no_overlap);
     seqStorePtr->lit += litLength;
 
     /* literal Length */
@@ -763,24 +769,37 @@ ZSTD_window_enforceMaxDist(ZSTD_window_t* window,
 
 /* Similar to ZSTD_window_enforceMaxDist(),
  * but only invalidates dictionary
- * when input progresses beyond window size. */
+ * when input progresses beyond window size.
+ * assumption : loadedDictEndPtr and dictMatchStatePtr are valid (non NULL)
+ *              loadedDictEnd uses same referential as window->base
+ *              maxDist is the window size */
 MEM_STATIC void
-ZSTD_checkDictValidity(ZSTD_window_t* window,
+ZSTD_checkDictValidity(const ZSTD_window_t* window,
                        const void* blockEnd,
                              U32   maxDist,
                              U32*  loadedDictEndPtr,
                        const ZSTD_matchState_t** dictMatchStatePtr)
 {
-    U32 const blockEndIdx = (U32)((BYTE const*)blockEnd - window->base);
-    U32 const loadedDictEnd = (loadedDictEndPtr != NULL) ? *loadedDictEndPtr : 0;
-    DEBUGLOG(5, "ZSTD_checkDictValidity: blockEndIdx=%u, maxDist=%u, loadedDictEnd=%u",
-                (unsigned)blockEndIdx, (unsigned)maxDist, (unsigned)loadedDictEnd);
-
-    if (loadedDictEnd && (blockEndIdx > maxDist + loadedDictEnd)) {
-        /* On reaching window size, dictionaries are invalidated */
-        if (loadedDictEndPtr) *loadedDictEndPtr = 0;
-        if (dictMatchStatePtr) *dictMatchStatePtr = NULL;
-    }
+    assert(loadedDictEndPtr != NULL);
+    assert(dictMatchStatePtr != NULL);
+    {   U32 const blockEndIdx = (U32)((BYTE const*)blockEnd - window->base);
+        U32 const loadedDictEnd = *loadedDictEndPtr;
+        DEBUGLOG(5, "ZSTD_checkDictValidity: blockEndIdx=%u, maxDist=%u, loadedDictEnd=%u",
+                    (unsigned)blockEndIdx, (unsigned)maxDist, (unsigned)loadedDictEnd);
+        assert(blockEndIdx >= loadedDictEnd);
+
+        if (blockEndIdx > loadedDictEnd + maxDist) {
+            /* On reaching window size, dictionaries are invalidated.
+             * For simplification, if window size is reached anywhere within next block,
+             * the dictionary is invalidated for the full block.
+             */
+            DEBUGLOG(6, "invalidating dictionary for current block (distance > windowSize)");
+            *loadedDictEndPtr = 0;
+            *dictMatchStatePtr = NULL;
+        } else {
+            if (*loadedDictEndPtr != 0) {
+                DEBUGLOG(6, "dictionary considered valid for current block");
+    }   }   }
 }
 
 /**
@@ -822,6 +841,17 @@ MEM_STATIC U32 ZSTD_window_update(ZSTD_window_t* window,
     return contiguous;
 }
 
+MEM_STATIC U32 ZSTD_getLowestMatchIndex(const ZSTD_matchState_t* ms, U32 current, unsigned windowLog)
+{
+    U32    const maxDistance = 1U << windowLog;
+    U32    const lowestValid = ms->window.lowLimit;
+    U32    const withinWindow = (current - lowestValid > maxDistance) ? current - maxDistance : lowestValid;
+    U32    const isDictionary = (ms->loadedDictEnd != 0);
+    U32    const matchLowest = isDictionary ? lowestValid : withinWindow;
+    return matchLowest;
+}
+
+
 
 /* debug functions */
 #if (DEBUGLEVEL>=2)