lib/lz4: update LZ4 decompressor module
authorGao Xiang <gaoxiang25@huawei.com>
Tue, 30 Oct 2018 22:07:28 +0000 (15:07 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Wed, 31 Oct 2018 15:54:14 +0000 (08:54 -0700)
Update the LZ4 compression module based on LZ4 v1.8.3 in order for the
erofs file system to use the newest LZ4_decompress_safe_partial() which
can now decode exactly the nb of bytes requested [1] to take place of the
open hacked code in the erofs file system itself.

Currently, apart from the erofs file system, no other users use
LZ4_decompress_safe_partial, so no worry about the interface.

In addition, LZ4 v1.8.x boosts up decompression speed compared to the
current code which is based on LZ4 v1.7.3, mainly due to shortcut
optimization for the specific common LZ4-sequences [2].

lzbench testdata (tested in kirin710, 8 cores, 4 big cores
at 2189Mhz, 2GB DDR RAM at 1622Mhz, with enwik8 testdata [3]):

Compressor name         Compress. Decompress. Compr. size  Ratio Filename
memcpy                   5004 MB/s  4924 MB/s   100000000 100.00 enwik8
lz4hc 1.7.3 -9             12 MB/s   653 MB/s    42203253  42.20 enwik8
lz4hc 1.8.0 -9             12 MB/s   908 MB/s    42203096  42.20 enwik8
lz4hc 1.8.3 -9             11 MB/s   965 MB/s    42203094  42.20 enwik8

[1] https://github.com/lz4/lz4/issues/566
    https://github.com/lz4/lz4/commit/08d347b5b217b011ff7487130b79480d8cfdaeb8

[2] v1.8.1 perf: slightly faster compression and decompression speed
    https://github.com/lz4/lz4/commit/a31b7058cb97e4393da55e78a77a1c6f0c9ae038
    v1.8.2 perf: slightly faster HC compression and decompression speed
    https://github.com/lz4/lz4/commit/45f8603aae389d34c689d3ff7427b314071ccd2c
    https://github.com/lz4/lz4/commit/1a191b3f8d26b50a7c1d41590b529ec308d768cd

[3] http://mattmahoney.net/dc/textdata.html
    http://mattmahoney.net/dc/enwik8.zip

Link: http://lkml.kernel.org/r/1537181207-21932-1-git-send-email-gaoxiang25@huawei.com
Signed-off-by: Gao Xiang <gaoxiang25@huawei.com>
Tested-by: Guo Xuenan <guoxuenan@huawei.com>
Cc: Colin Ian King <colin.king@canonical.com>
Cc: Yann Collet <yann.collet.73@gmail.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Fang Wei <fangwei1@huawei.com>
Cc: Chao Yu <yuchao0@huawei.com>
Cc: Miao Xie <miaoxie@huawei.com>
Cc: Sven Schmidt <4sschmid@informatik.uni-hamburg.de>
Cc: Kyungsik Lee <kyungsik.lee@lge.com>
Cc: <weidu.du@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
lib/lz4/lz4_decompress.c
lib/lz4/lz4defs.h

index 141734d255e4b941f888e32ee53885c26cfc8fbd..0c9d3ad17e0fc7b25b75b92d1ebec2c94ce8ecee 100644 (file)
 /*-*****************************
  *     Decompression functions
  *******************************/
-/* LZ4_decompress_generic() :
- * This generic decompression function cover all use cases.
- * It shall be instantiated several times, using different sets of directives
- * Note that it is important this generic function is really inlined,
+
+#define DEBUGLOG(l, ...) {}    /* disabled */
+
+#ifndef assert
+#define assert(condition) ((void)0)
+#endif
+
+/*
+ * LZ4_decompress_generic() :
+ * This generic decompression function covers all use cases.
+ * It shall be instantiated several times, using different sets of directives.
+ * Note that it is important for performance that this function really get inlined,
  * in order to remove useless branches during compilation optimization.
  */
 static FORCE_INLINE int LZ4_decompress_generic(
-        const char * const source,
-        char * const dest,
-        int inputSize,
+        const char * const src,
+        char * const dst,
+        int srcSize,
                /*
                 * If endOnInput == endOnInputSize,
-                * this value is the max size of Output Buffer.
+                * this value is `dstCapacity`
                 */
         int outputSize,
         /* endOnOutputSize, endOnInputSize */
-        int endOnInput,
+        endCondition_directive endOnInput,
         /* full, partial */
-        int partialDecoding,
-        /* only used if partialDecoding == partial */
-        int targetOutputSize,
+        earlyEnd_directive partialDecoding,
         /* noDict, withPrefix64k, usingExtDict */
-        int dict,
-        /* == dest when no prefix */
+        dict_directive dict,
+        /* always <= dst, == dst when no prefix */
         const BYTE * const lowPrefix,
         /* only if dict == usingExtDict */
         const BYTE * const dictStart,
@@ -74,35 +80,43 @@ static FORCE_INLINE int LZ4_decompress_generic(
         const size_t dictSize
         )
 {
-       /* Local Variables */
-       const BYTE *ip = (const BYTE *) source;
-       const BYTE * const iend = ip + inputSize;
+       const BYTE *ip = (const BYTE *) src;
+       const BYTE * const iend = ip + srcSize;
 
-       BYTE *op = (BYTE *) dest;
+       BYTE *op = (BYTE *) dst;
        BYTE * const oend = op + outputSize;
        BYTE *cpy;
-       BYTE *oexit = op + targetOutputSize;
-       const BYTE * const lowLimit = lowPrefix - dictSize;
 
        const BYTE * const dictEnd = (const BYTE *)dictStart + dictSize;
-       static const unsigned int dec32table[] = { 0, 1, 2, 1, 4, 4, 4, 4 };
-       static const int dec64table[] = { 0, 0, 0, -1, 0, 1, 2, 3 };
+       static const unsigned int inc32table[8] = {0, 1, 2, 1, 0, 4, 4, 4};
+       static const int dec64table[8] = {0, 0, 0, -1, -4, 1, 2, 3};
 
        const int safeDecode = (endOnInput == endOnInputSize);
        const int checkOffset = ((safeDecode) && (dictSize < (int)(64 * KB)));
 
+       /* Set up the "end" pointers for the shortcut. */
+       const BYTE *const shortiend = iend -
+               (endOnInput ? 14 : 8) /*maxLL*/ - 2 /*offset*/;
+       const BYTE *const shortoend = oend -
+               (endOnInput ? 14 : 8) /*maxLL*/ - 18 /*maxML*/;
+
+       DEBUGLOG(5, "%s (srcSize:%i, dstSize:%i)", __func__,
+                srcSize, outputSize);
+
        /* Special cases */
-       /* targetOutputSize too high => decode everything */
-       if ((partialDecoding) && (oexit > oend - MFLIMIT))
-               oexit = oend - MFLIMIT;
+       assert(lowPrefix <= op);
+       assert(src != NULL);
 
        /* Empty output buffer */
        if ((endOnInput) && (unlikely(outputSize == 0)))
-               return ((inputSize == 1) && (*ip == 0)) ? 0 : -1;
+               return ((srcSize == 1) && (*ip == 0)) ? 0 : -1;
 
        if ((!endOnInput) && (unlikely(outputSize == 0)))
                return (*ip == 0 ? 1 : -1);
 
+       if ((endOnInput) && unlikely(srcSize == 0))
+               return -1;
+
        /* Main Loop : decode sequences */
        while (1) {
                size_t length;
@@ -111,12 +125,74 @@ static FORCE_INLINE int LZ4_decompress_generic(
 
                /* get literal length */
                unsigned int const token = *ip++;
-
                length = token>>ML_BITS;
 
+               /* ip < iend before the increment */
+               assert(!endOnInput || ip <= iend);
+
+               /*
+                * A two-stage shortcut for the most common case:
+                * 1) If the literal length is 0..14, and there is enough
+                * space, enter the shortcut and copy 16 bytes on behalf
+                * of the literals (in the fast mode, only 8 bytes can be
+                * safely copied this way).
+                * 2) Further if the match length is 4..18, copy 18 bytes
+                * in a similar manner; but we ensure that there's enough
+                * space in the output for those 18 bytes earlier, upon
+                * entering the shortcut (in other words, there is a
+                * combined check for both stages).
+                */
+               if ((endOnInput ? length != RUN_MASK : length <= 8)
+                  /*
+                   * strictly "less than" on input, to re-enter
+                   * the loop with at least one byte
+                   */
+                  && likely((endOnInput ? ip < shortiend : 1) &
+                            (op <= shortoend))) {
+                       /* Copy the literals */
+                       memcpy(op, ip, endOnInput ? 16 : 8);
+                       op += length; ip += length;
+
+                       /*
+                        * The second stage:
+                        * prepare for match copying, decode full info.
+                        * If it doesn't work out, the info won't be wasted.
+                        */
+                       length = token & ML_MASK; /* match length */
+                       offset = LZ4_readLE16(ip);
+                       ip += 2;
+                       match = op - offset;
+                       assert(match <= op); /* check overflow */
+
+                       /* Do not deal with overlapping matches. */
+                       if ((length != ML_MASK) &&
+                           (offset >= 8) &&
+                           (dict == withPrefix64k || match >= lowPrefix)) {
+                               /* Copy the match. */
+                               memcpy(op + 0, match + 0, 8);
+                               memcpy(op + 8, match + 8, 8);
+                               memcpy(op + 16, match + 16, 2);
+                               op += length + MINMATCH;
+                               /* Both stages worked, load the next token. */
+                               continue;
+                       }
+
+                       /*
+                        * The second stage didn't work out, but the info
+                        * is ready. Propel it right to the point of match
+                        * copying.
+                        */
+                       goto _copy_match;
+               }
+
+               /* decode literal length */
                if (length == RUN_MASK) {
                        unsigned int s;
 
+                       if (unlikely(endOnInput ? ip >= iend - RUN_MASK : 0)) {
+                               /* overflow detection */
+                               goto _output_error;
+                       }
                        do {
                                s = *ip++;
                                length += s;
@@ -125,14 +201,14 @@ static FORCE_INLINE int LZ4_decompress_generic(
                                : 1) & (s == 255));
 
                        if ((safeDecode)
-                               && unlikely(
-                                       (size_t)(op + length) < (size_t)(op))) {
+                           && unlikely((uptrval)(op) +
+                                       length < (uptrval)(op))) {
                                /* overflow detection */
                                goto _output_error;
                        }
                        if ((safeDecode)
-                               && unlikely(
-                                       (size_t)(ip + length) < (size_t)(ip))) {
+                           && unlikely((uptrval)(ip) +
+                                       length < (uptrval)(ip))) {
                                /* overflow detection */
                                goto _output_error;
                        }
@@ -140,16 +216,19 @@ static FORCE_INLINE int LZ4_decompress_generic(
 
                /* copy literals */
                cpy = op + length;
-               if (((endOnInput) && ((cpy > (partialDecoding ? oexit : oend - MFLIMIT))
+               LZ4_STATIC_ASSERT(MFLIMIT >= WILDCOPYLENGTH);
+
+               if (((endOnInput) && ((cpy > oend - MFLIMIT)
                        || (ip + length > iend - (2 + 1 + LASTLITERALS))))
                        || ((!endOnInput) && (cpy > oend - WILDCOPYLENGTH))) {
                        if (partialDecoding) {
                                if (cpy > oend) {
                                        /*
-                                        * Error :
-                                        * write attempt beyond end of output buffer
+                                        * Partial decoding :
+                                        * stop in the middle of literal segment
                                         */
-                                       goto _output_error;
+                                       cpy = oend;
+                                       length = oend - op;
                                }
                                if ((endOnInput)
                                        && (ip + length > iend)) {
@@ -184,29 +263,43 @@ static FORCE_INLINE int LZ4_decompress_generic(
                        memcpy(op, ip, length);
                        ip += length;
                        op += length;
+
                        /* Necessarily EOF, due to parsing restrictions */
-                       break;
+                       if (!partialDecoding || (cpy == oend))
+                               break;
+               } else {
+                       /* may overwrite up to WILDCOPYLENGTH beyond cpy */
+                       LZ4_wildCopy(op, ip, cpy);
+                       ip += length;
+                       op = cpy;
                }
 
-               LZ4_wildCopy(op, ip, cpy);
-               ip += length;
-               op = cpy;
-
                /* get offset */
                offset = LZ4_readLE16(ip);
                ip += 2;
                match = op - offset;
 
-               if ((checkOffset) && (unlikely(match < lowLimit))) {
+               /* get matchlength */
+               length = token & ML_MASK;
+
+_copy_match:
+               if ((checkOffset) && (unlikely(match + dictSize < lowPrefix))) {
                        /* Error : offset outside buffers */
                        goto _output_error;
                }
 
                /* costs ~1%; silence an msan warning when offset == 0 */
-               LZ4_write32(op, (U32)offset);
+               /*
+                * note : when partialDecoding, there is no guarantee that
+                * at least 4 bytes remain available in output buffer
+                */
+               if (!partialDecoding) {
+                       assert(oend > op);
+                       assert(oend - op >= 4);
+
+                       LZ4_write32(op, (U32)offset);
+               }
 
-               /* get matchlength */
-               length = token & ML_MASK;
                if (length == ML_MASK) {
                        unsigned int s;
 
@@ -221,7 +314,7 @@ static FORCE_INLINE int LZ4_decompress_generic(
 
                        if ((safeDecode)
                                && unlikely(
-                                       (size_t)(op + length) < (size_t)op)) {
+                                       (uptrval)(op) + length < (uptrval)op)) {
                                /* overflow detection */
                                goto _output_error;
                        }
@@ -229,24 +322,26 @@ static FORCE_INLINE int LZ4_decompress_generic(
 
                length += MINMATCH;
 
-               /* check external dictionary */
+               /* match starting within external dictionary */
                if ((dict == usingExtDict) && (match < lowPrefix)) {
                        if (unlikely(op + length > oend - LASTLITERALS)) {
                                /* doesn't respect parsing restriction */
-                               goto _output_error;
+                               if (!partialDecoding)
+                                       goto _output_error;
+                               length = min(length, (size_t)(oend - op));
                        }
 
                        if (length <= (size_t)(lowPrefix - match)) {
                                /*
-                                * match can be copied as a single segment
-                                * from external dictionary
+                                * match fits entirely within external
+                                * dictionary : just copy
                                 */
                                memmove(op, dictEnd - (lowPrefix - match),
                                        length);
                                op += length;
                        } else {
                                /*
-                                * match encompass external
+                                * match stretches into both external
                                 * dictionary and current block
                                 */
                                size_t const copySize = (size_t)(lowPrefix - match);
@@ -254,7 +349,6 @@ static FORCE_INLINE int LZ4_decompress_generic(
 
                                memcpy(op, dictEnd - copySize, copySize);
                                op += copySize;
-
                                if (restSize > (size_t)(op - lowPrefix)) {
                                        /* overlap copy */
                                        BYTE * const endOfMatch = op + restSize;
@@ -267,23 +361,44 @@ static FORCE_INLINE int LZ4_decompress_generic(
                                        op += restSize;
                                }
                        }
-
                        continue;
                }
 
                /* copy match within block */
                cpy = op + length;
 
-               if (unlikely(offset < 8)) {
-                       const int dec64 = dec64table[offset];
+               /*
+                * partialDecoding :
+                * may not respect endBlock parsing restrictions
+                */
+               assert(op <= oend);
+               if (partialDecoding &&
+                   (cpy > oend - MATCH_SAFEGUARD_DISTANCE)) {
+                       size_t const mlen = min(length, (size_t)(oend - op));
+                       const BYTE * const matchEnd = match + mlen;
+                       BYTE * const copyEnd = op + mlen;
+
+                       if (matchEnd > op) {
+                               /* overlap copy */
+                               while (op < copyEnd)
+                                       *op++ = *match++;
+                       } else {
+                               memcpy(op, match, mlen);
+                       }
+                       op = copyEnd;
+                       if (op == oend)
+                               break;
+                       continue;
+               }
 
+               if (unlikely(offset < 8)) {
                        op[0] = match[0];
                        op[1] = match[1];
                        op[2] = match[2];
                        op[3] = match[3];
-                       match += dec32table[offset];
+                       match += inc32table[offset];
                        memcpy(op + 4, match, 4);
-                       match -= dec64;
+                       match -= dec64table[offset];
                } else {
                        LZ4_copy8(op, match);
                        match += 8;
@@ -291,7 +406,7 @@ static FORCE_INLINE int LZ4_decompress_generic(
 
                op += 8;
 
-               if (unlikely(cpy > oend - 12)) {
+               if (unlikely(cpy > oend - MATCH_SAFEGUARD_DISTANCE)) {
                        BYTE * const oCopyLimit = oend - (WILDCOPYLENGTH - 1);
 
                        if (cpy > oend - LASTLITERALS) {
@@ -307,60 +422,139 @@ static FORCE_INLINE int LZ4_decompress_generic(
                                match += oCopyLimit - op;
                                op = oCopyLimit;
                        }
-
                        while (op < cpy)
                                *op++ = *match++;
                } else {
                        LZ4_copy8(op, match);
-
                        if (length > 16)
                                LZ4_wildCopy(op + 8, match + 8, cpy);
                }
-
-               op = cpy; /* correction */
+               op = cpy; /* wildcopy correction */
        }
 
        /* end of decoding */
        if (endOnInput) {
                /* Nb of output bytes decoded */
-               return (int) (((char *)op) - dest);
+               return (int) (((char *)op) - dst);
        } else {
                /* Nb of input bytes read */
-               return (int) (((const char *)ip) - source);
+               return (int) (((const char *)ip) - src);
        }
 
        /* Overflow error detected */
 _output_error:
-       return -1;
+       return (int) (-(((const char *)ip) - src)) - 1;
 }
 
 int LZ4_decompress_safe(const char *source, char *dest,
        int compressedSize, int maxDecompressedSize)
 {
-       return LZ4_decompress_generic(source, dest, compressedSize,
-               maxDecompressedSize, endOnInputSize, full, 0,
-               noDict, (BYTE *)dest, NULL, 0);
+       return LZ4_decompress_generic(source, dest,
+                                     compressedSize, maxDecompressedSize,
+                                     endOnInputSize, decode_full_block,
+                                     noDict, (BYTE *)dest, NULL, 0);
 }
 
-int LZ4_decompress_safe_partial(const char *source, char *dest,
-       int compressedSize, int targetOutputSize, int maxDecompressedSize)
+int LZ4_decompress_safe_partial(const char *src, char *dst,
+       int compressedSize, int targetOutputSize, int dstCapacity)
 {
-       return LZ4_decompress_generic(source, dest, compressedSize,
-               maxDecompressedSize, endOnInputSize, partial,
-               targetOutputSize, noDict, (BYTE *)dest, NULL, 0);
+       dstCapacity = min(targetOutputSize, dstCapacity);
+       return LZ4_decompress_generic(src, dst, compressedSize, dstCapacity,
+                                     endOnInputSize, partial_decode,
+                                     noDict, (BYTE *)dst, NULL, 0);
 }
 
 int LZ4_decompress_fast(const char *source, char *dest, int originalSize)
 {
        return LZ4_decompress_generic(source, dest, 0, originalSize,
-               endOnOutputSize, full, 0, withPrefix64k,
-               (BYTE *)(dest - 64 * KB), NULL, 64 * KB);
+                                     endOnOutputSize, decode_full_block,
+                                     withPrefix64k,
+                                     (BYTE *)dest - 64 * KB, NULL, 0);
+}
+
+/* ===== Instantiate a few more decoding cases, used more than once. ===== */
+
+int LZ4_decompress_safe_withPrefix64k(const char *source, char *dest,
+                                     int compressedSize, int maxOutputSize)
+{
+       return LZ4_decompress_generic(source, dest,
+                                     compressedSize, maxOutputSize,
+                                     endOnInputSize, decode_full_block,
+                                     withPrefix64k,
+                                     (BYTE *)dest - 64 * KB, NULL, 0);
+}
+
+static int LZ4_decompress_safe_withSmallPrefix(const char *source, char *dest,
+                                              int compressedSize,
+                                              int maxOutputSize,
+                                              size_t prefixSize)
+{
+       return LZ4_decompress_generic(source, dest,
+                                     compressedSize, maxOutputSize,
+                                     endOnInputSize, decode_full_block,
+                                     noDict,
+                                     (BYTE *)dest - prefixSize, NULL, 0);
+}
+
+int LZ4_decompress_safe_forceExtDict(const char *source, char *dest,
+                                    int compressedSize, int maxOutputSize,
+                                    const void *dictStart, size_t dictSize)
+{
+       return LZ4_decompress_generic(source, dest,
+                                     compressedSize, maxOutputSize,
+                                     endOnInputSize, decode_full_block,
+                                     usingExtDict, (BYTE *)dest,
+                                     (const BYTE *)dictStart, dictSize);
 }
 
+static int LZ4_decompress_fast_extDict(const char *source, char *dest,
+                                      int originalSize,
+                                      const void *dictStart, size_t dictSize)
+{
+       return LZ4_decompress_generic(source, dest,
+                                     0, originalSize,
+                                     endOnOutputSize, decode_full_block,
+                                     usingExtDict, (BYTE *)dest,
+                                     (const BYTE *)dictStart, dictSize);
+}
+
+/*
+ * The "double dictionary" mode, for use with e.g. ring buffers: the first part
+ * of the dictionary is passed as prefix, and the second via dictStart + dictSize.
+ * These routines are used only once, in LZ4_decompress_*_continue().
+ */
+static FORCE_INLINE
+int LZ4_decompress_safe_doubleDict(const char *source, char *dest,
+                                  int compressedSize, int maxOutputSize,
+                                  size_t prefixSize,
+                                  const void *dictStart, size_t dictSize)
+{
+       return LZ4_decompress_generic(source, dest,
+                                     compressedSize, maxOutputSize,
+                                     endOnInputSize, decode_full_block,
+                                     usingExtDict, (BYTE *)dest - prefixSize,
+                                     (const BYTE *)dictStart, dictSize);
+}
+
+static FORCE_INLINE
+int LZ4_decompress_fast_doubleDict(const char *source, char *dest,
+                                  int originalSize, size_t prefixSize,
+                                  const void *dictStart, size_t dictSize)
+{
+       return LZ4_decompress_generic(source, dest,
+                                     0, originalSize,
+                                     endOnOutputSize, decode_full_block,
+                                     usingExtDict, (BYTE *)dest - prefixSize,
+                                     (const BYTE *)dictStart, dictSize);
+}
+
+/* ===== streaming decompression functions ===== */
+
 int LZ4_setStreamDecode(LZ4_streamDecode_t *LZ4_streamDecode,
        const char *dictionary, int dictSize)
 {
-       LZ4_streamDecode_t_internal *lz4sd = (LZ4_streamDecode_t_internal *) LZ4_streamDecode;
+       LZ4_streamDecode_t_internal *lz4sd =
+               &LZ4_streamDecode->internal_donotuse;
 
        lz4sd->prefixSize = (size_t) dictSize;
        lz4sd->prefixEnd = (const BYTE *) dictionary + dictSize;
@@ -382,35 +576,51 @@ int LZ4_setStreamDecode(LZ4_streamDecode_t *LZ4_streamDecode,
 int LZ4_decompress_safe_continue(LZ4_streamDecode_t *LZ4_streamDecode,
        const char *source, char *dest, int compressedSize, int maxOutputSize)
 {
-       LZ4_streamDecode_t_internal *lz4sd = &LZ4_streamDecode->internal_donotuse;
+       LZ4_streamDecode_t_internal *lz4sd =
+               &LZ4_streamDecode->internal_donotuse;
        int result;
 
-       if (lz4sd->prefixEnd == (BYTE *)dest) {
-               result = LZ4_decompress_generic(source, dest,
-                       compressedSize,
-                       maxOutputSize,
-                       endOnInputSize, full, 0,
-                       usingExtDict, lz4sd->prefixEnd - lz4sd->prefixSize,
-                       lz4sd->externalDict,
-                       lz4sd->extDictSize);
-
+       if (lz4sd->prefixSize == 0) {
+               /* The first call, no dictionary yet. */
+               assert(lz4sd->extDictSize == 0);
+               result = LZ4_decompress_safe(source, dest,
+                       compressedSize, maxOutputSize);
+               if (result <= 0)
+                       return result;
+               lz4sd->prefixSize = result;
+               lz4sd->prefixEnd = (BYTE *)dest + result;
+       } else if (lz4sd->prefixEnd == (BYTE *)dest) {
+               /* They're rolling the current segment. */
+               if (lz4sd->prefixSize >= 64 * KB - 1)
+                       result = LZ4_decompress_safe_withPrefix64k(source, dest,
+                               compressedSize, maxOutputSize);
+               else if (lz4sd->extDictSize == 0)
+                       result = LZ4_decompress_safe_withSmallPrefix(source,
+                               dest, compressedSize, maxOutputSize,
+                               lz4sd->prefixSize);
+               else
+                       result = LZ4_decompress_safe_doubleDict(source, dest,
+                               compressedSize, maxOutputSize,
+                               lz4sd->prefixSize,
+                               lz4sd->externalDict, lz4sd->extDictSize);
                if (result <= 0)
                        return result;
-
                lz4sd->prefixSize += result;
-               lz4sd->prefixEnd        += result;
+               lz4sd->prefixEnd  += result;
        } else {
+               /*
+                * The buffer wraps around, or they're
+                * switching to another buffer.
+                */
                lz4sd->extDictSize = lz4sd->prefixSize;
                lz4sd->externalDict = lz4sd->prefixEnd - lz4sd->extDictSize;
-               result = LZ4_decompress_generic(source, dest,
+               result = LZ4_decompress_safe_forceExtDict(source, dest,
                        compressedSize, maxOutputSize,
-                       endOnInputSize, full, 0,
-                       usingExtDict, (BYTE *)dest,
                        lz4sd->externalDict, lz4sd->extDictSize);
                if (result <= 0)
                        return result;
                lz4sd->prefixSize = result;
-               lz4sd->prefixEnd        = (BYTE *)dest + result;
+               lz4sd->prefixEnd  = (BYTE *)dest + result;
        }
 
        return result;
@@ -422,75 +632,66 @@ int LZ4_decompress_fast_continue(LZ4_streamDecode_t *LZ4_streamDecode,
        LZ4_streamDecode_t_internal *lz4sd = &LZ4_streamDecode->internal_donotuse;
        int result;
 
-       if (lz4sd->prefixEnd == (BYTE *)dest) {
-               result = LZ4_decompress_generic(source, dest, 0, originalSize,
-                       endOnOutputSize, full, 0,
-                       usingExtDict,
-                       lz4sd->prefixEnd - lz4sd->prefixSize,
-                       lz4sd->externalDict, lz4sd->extDictSize);
-
+       if (lz4sd->prefixSize == 0) {
+               assert(lz4sd->extDictSize == 0);
+               result = LZ4_decompress_fast(source, dest, originalSize);
+               if (result <= 0)
+                       return result;
+               lz4sd->prefixSize = originalSize;
+               lz4sd->prefixEnd = (BYTE *)dest + originalSize;
+       } else if (lz4sd->prefixEnd == (BYTE *)dest) {
+               if (lz4sd->prefixSize >= 64 * KB - 1 ||
+                   lz4sd->extDictSize == 0)
+                       result = LZ4_decompress_fast(source, dest,
+                                                    originalSize);
+               else
+                       result = LZ4_decompress_fast_doubleDict(source, dest,
+                               originalSize, lz4sd->prefixSize,
+                               lz4sd->externalDict, lz4sd->extDictSize);
                if (result <= 0)
                        return result;
-
                lz4sd->prefixSize += originalSize;
-               lz4sd->prefixEnd        += originalSize;
+               lz4sd->prefixEnd  += originalSize;
        } else {
                lz4sd->extDictSize = lz4sd->prefixSize;
                lz4sd->externalDict = lz4sd->prefixEnd - lz4sd->extDictSize;
-               result = LZ4_decompress_generic(source, dest, 0, originalSize,
-                       endOnOutputSize, full, 0,
-                       usingExtDict, (BYTE *)dest,
-                       lz4sd->externalDict, lz4sd->extDictSize);
+               result = LZ4_decompress_fast_extDict(source, dest,
+                       originalSize, lz4sd->externalDict, lz4sd->extDictSize);
                if (result <= 0)
                        return result;
                lz4sd->prefixSize = originalSize;
-               lz4sd->prefixEnd        = (BYTE *)dest + originalSize;
+               lz4sd->prefixEnd = (BYTE *)dest + originalSize;
        }
-
        return result;
 }
 
-/*
- * Advanced decoding functions :
- * *_usingDict() :
- * These decoding functions work the same as "_continue" ones,
- * the dictionary must be explicitly provided within parameters
- */
-static FORCE_INLINE int LZ4_decompress_usingDict_generic(const char *source,
-       char *dest, int compressedSize, int maxOutputSize, int safe,
-       const char *dictStart, int dictSize)
+int LZ4_decompress_safe_usingDict(const char *source, char *dest,
+                                 int compressedSize, int maxOutputSize,
+                                 const char *dictStart, int dictSize)
 {
        if (dictSize == 0)
-               return LZ4_decompress_generic(source, dest,
-                       compressedSize, maxOutputSize, safe, full, 0,
-                       noDict, (BYTE *)dest, NULL, 0);
-       if (dictStart + dictSize == dest) {
-               if (dictSize >= (int)(64 * KB - 1))
-                       return LZ4_decompress_generic(source, dest,
-                               compressedSize, maxOutputSize, safe, full, 0,
-                               withPrefix64k, (BYTE *)dest - 64 * KB, NULL, 0);
-               return LZ4_decompress_generic(source, dest, compressedSize,
-                       maxOutputSize, safe, full, 0, noDict,
-                       (BYTE *)dest - dictSize, NULL, 0);
+               return LZ4_decompress_safe(source, dest,
+                                          compressedSize, maxOutputSize);
+       if (dictStart+dictSize == dest) {
+               if (dictSize >= 64 * KB - 1)
+                       return LZ4_decompress_safe_withPrefix64k(source, dest,
+                               compressedSize, maxOutputSize);
+               return LZ4_decompress_safe_withSmallPrefix(source, dest,
+                       compressedSize, maxOutputSize, dictSize);
        }
-       return LZ4_decompress_generic(source, dest, compressedSize,
-               maxOutputSize, safe, full, 0, usingExtDict,
-               (BYTE *)dest, (const BYTE *)dictStart, dictSize);
-}
-
-int LZ4_decompress_safe_usingDict(const char *source, char *dest,
-       int compressedSize, int maxOutputSize,
-       const char *dictStart, int dictSize)
-{
-       return LZ4_decompress_usingDict_generic(source, dest,
-               compressedSize, maxOutputSize, 1, dictStart, dictSize);
+       return LZ4_decompress_safe_forceExtDict(source, dest,
+               compressedSize, maxOutputSize, dictStart, dictSize);
 }
 
 int LZ4_decompress_fast_usingDict(const char *source, char *dest,
-       int originalSize, const char *dictStart, int dictSize)
+                                 int originalSize,
+                                 const char *dictStart, int dictSize)
 {
-       return LZ4_decompress_usingDict_generic(source, dest, 0,
-               originalSize, 0, dictStart, dictSize);
+       if (dictSize == 0 || dictStart + dictSize == dest)
+               return LZ4_decompress_fast(source, dest, originalSize);
+
+       return LZ4_decompress_fast_extDict(source, dest, originalSize,
+               dictStart, dictSize);
 }
 
 #ifndef STATIC
index 00a0b58a0871bc3c1f0afca8aa9bbd03aab09edd..1a7fa9d9170fb6007eb66a14b3425570294c92d9 100644 (file)
@@ -75,6 +75,11 @@ typedef uintptr_t uptrval;
 #define WILDCOPYLENGTH 8
 #define LASTLITERALS 5
 #define MFLIMIT (WILDCOPYLENGTH + MINMATCH)
+/*
+ * ensure it's possible to write 2 x wildcopyLength
+ * without overflowing output buffer
+ */
+#define MATCH_SAFEGUARD_DISTANCE  ((2 * WILDCOPYLENGTH) - MINMATCH)
 
 /* Increase this value ==> compression run slower on incompressible data */
 #define LZ4_SKIPTRIGGER 6
@@ -222,6 +227,8 @@ typedef enum { noDict = 0, withPrefix64k, usingExtDict } dict_directive;
 typedef enum { noDictIssue = 0, dictSmall } dictIssue_directive;
 
 typedef enum { endOnOutputSize = 0, endOnInputSize = 1 } endCondition_directive;
-typedef enum { full = 0, partial = 1 } earlyEnd_directive;
+typedef enum { decode_full_block = 0, partial_decode = 1 } earlyEnd_directive;
+
+#define LZ4_STATIC_ASSERT(c)   BUILD_BUG_ON(!(c))
 
 #endif