ddt: split internal DDT API into separate header

[mirror_zfs.git] / module / zfs / lz4.c

/*
   LZ4 - Fast LZ compression algorithm
   Copyright (C) 2011-present, Yann Collet.

   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)

   Redistribution and use in source and binary forms, with or without
   modification, are permitted provided that the following conditions are
   met:

       * Redistributions of source code must retain the above copyright
   notice, this list of conditions and the following disclaimer.
       * Redistributions in binary form must reproduce the above
   copyright notice, this list of conditions and the following disclaimer
   in the documentation and/or other materials provided with the
   distribution.

   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

   You can contact the author at :
    - LZ4 homepage : http://www.lz4.org
    - LZ4 source repository : https://github.com/lz4/lz4
*/

/*
 * This file contains unmodified code from lz4 1.9.3's decompressor, plus
 * associated macros and constants.
 *
 * It also contains a couple of defines from the old lz4.c to make things
 * fit together smoothly.
 *
 */

#include <sys/zfs_context.h>

int LZ4_uncompress_unknownOutputSize(const char *source, char *dest,
    int isize, int maxOutputSize);

/*
 * Tuning parameters
 */

/*
 * COMPRESSIONLEVEL: Increasing this value improves compression ratio
 *       Lowering this value reduces memory usage. Reduced memory usage
 *      typically improves speed, due to cache effect (ex: L1 32KB for Intel,
 *      L1 64KB for AMD). Memory usage formula : N->2^(N+2) Bytes
 *      (examples : 12 -> 16KB ; 17 -> 512KB)
 */
#define COMPRESSIONLEVEL 12

/*
 * NOTCOMPRESSIBLE_CONFIRMATION: Decreasing this value will make the
 *      algorithm skip faster data segments considered "incompressible".
 *      This may decrease compression ratio dramatically, but will be
 *      faster on incompressible data. Increasing this value will make
 *      the algorithm search more before declaring a segment "incompressible".
 *      This could improve compression a bit, but will be slower on
 *      incompressible data. The default value (6) is recommended.
 */
#define NOTCOMPRESSIBLE_CONFIRMATION 6

/*
 * Little Endian or Big Endian?
 * Note: overwrite the below #define if you know your architecture endianness.
 */
#if defined(_ZFS_BIG_ENDIAN)
#define LZ4_BIG_ENDIAN 1
#else
/*
 * Little Endian assumed. PDP Endian and other very rare endian format
 * are unsupported.
 */
#undef LZ4_BIG_ENDIAN
#endif

/*-************************************
*  CPU Feature Detection
**************************************/
/* LZ4_FORCE_MEMORY_ACCESS
 * By default, access to unaligned memory is controlled by `memcpy()`, which is safe and portable.
 * Unfortunately, on some target/compiler combinations, the generated assembly is sub-optimal.
 * The below switch allow to select different access method for improved performance.
 * Method 0 (default) : use `memcpy()`. Safe and portable.
 * Method 1 : `__packed` statement. It depends on compiler extension (ie, not portable).
 *            This method is safe if your compiler supports it, and *generally* as fast or faster than `memcpy`.
 * Method 2 : direct access. This method is portable but violate C standard.
 *            It can generate buggy code on targets which assembly generation depends on alignment.
 *            But in some circumstances, it's the only known way to get the most performance (ie GCC + ARMv6)
 * See https://fastcompression.blogspot.fr/2015/08/accessing-unaligned-memory.html for details.
 * Prefer these methods in priority order (0 > 1 > 2)
 */
#ifndef LZ4_FORCE_MEMORY_ACCESS   /* can be defined externally */
#  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 LZ4_FORCE_MEMORY_ACCESS 2
#  elif (defined(__INTEL_COMPILER) && !defined(_WIN32)) || defined(__GNUC__)
#    define LZ4_FORCE_MEMORY_ACCESS 1
#  endif
#endif

/*
 * LZ4_FORCE_SW_BITCOUNT
 * Define this parameter if your target system or compiler does not support hardware bit count
 */
/*
 * Illumos : we can't use GCC's __builtin_ctz family of builtins in the
 * kernel
 * Linux : we can use GCC's __builtin_ctz family of builtins in the
 * kernel
 */
#undef  LZ4_FORCE_SW_BITCOUNT
#if defined(__sunos__)
#define LZ4_FORCE_SW_BITCOUNT
#endif

/*
 * Compiler Options
 */
/* Disable restrict */
#define restrict

/*
 * Linux : GCC_VERSION is defined as of 3.9-rc1, so undefine it.
 * torvalds/linux@3f3f8d2f48acfd8ed3b8e6b7377935da57b27b16
 */
#ifdef GCC_VERSION
#undef GCC_VERSION
#endif

#define GCC_VERSION (__GNUC__ * 100 + __GNUC_MINOR__)

#ifndef LZ4_FORCE_INLINE
#  ifdef _MSC_VER    /* Visual Studio */
#    define LZ4_FORCE_INLINE static __forceinline
#  else
#    if defined (__cplusplus) || defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   /* C99 */
#      ifdef __GNUC__
#        define LZ4_FORCE_INLINE static inline __attribute__((always_inline))
#      else
#        define LZ4_FORCE_INLINE static inline
#      endif
#    else
#      define LZ4_FORCE_INLINE static
#    endif /* __STDC_VERSION__ */
#  endif  /* _MSC_VER */
#endif /* LZ4_FORCE_INLINE */

/* LZ4_FORCE_O2 and LZ4_FORCE_INLINE
 * gcc on ppc64le generates an unrolled SIMDized loop for LZ4_wildCopy8,
 * together with a simple 8-byte copy loop as a fall-back path.
 * However, this optimization hurts the decompression speed by >30%,
 * because the execution does not go to the optimized loop
 * for typical compressible data, and all of the preamble checks
 * before going to the fall-back path become useless overhead.
 * This optimization happens only with the -O3 flag, and -O2 generates
 * a simple 8-byte copy loop.
 * With gcc on ppc64le, all of the LZ4_decompress_* and LZ4_wildCopy8
 * functions are annotated with __attribute__((optimize("O2"))),
 * and also LZ4_wildCopy8 is forcibly inlined, so that the O2 attribute
 * of LZ4_wildCopy8 does not affect the compression speed.
 */
#if defined(__PPC64__) && defined(__LITTLE_ENDIAN__) && defined(__GNUC__) && !defined(__clang__)
#  define LZ4_FORCE_O2  __attribute__((optimize("O2")))
#  undef LZ4_FORCE_INLINE
#  define LZ4_FORCE_INLINE  static __inline __attribute__((optimize("O2"),always_inline))
#else
#  define LZ4_FORCE_O2
#endif

#ifndef expect
#if (defined(__GNUC__) && (__GNUC__ >= 3)) || (defined(__INTEL_COMPILER) && (__INTEL_COMPILER >= 800)) || defined(__clang__)
#  define expect(expr,value)    (__builtin_expect ((expr),(value)) )
#else
#  define expect(expr,value)    (expr)
#endif
#endif

#ifndef likely
#define likely(expr)    expect((expr) != 0, 1)
#endif

#ifndef unlikely
#define unlikely(expr)  expect((expr) != 0, 0)
#endif

#ifndef _KERNEL
#include <stdlib.h>   /* malloc, calloc, free */
#include <string.h>   /* memset, memcpy */
#endif
#define ALLOC(s)          malloc(s)
#define ALLOC_AND_ZERO(s) calloc(1,s)
#define FREEMEM(p)        free(p)

#define MEM_INIT(p,v,s)   memset((p),(v),(s))


/*-************************************
*  Common Constants
**************************************/
#define MINMATCH 4

#define WILDCOPYLENGTH 8
#define LASTLITERALS   5   /* see ../doc/lz4_Block_format.md#parsing-restrictions */
#define MFLIMIT       12   /* see ../doc/lz4_Block_format.md#parsing-restrictions */
#define MATCH_SAFEGUARD_DISTANCE  ((2*WILDCOPYLENGTH) - MINMATCH)   /* ensure it's possible to write 2 x wildcopyLength without overflowing output buffer */
#define FASTLOOP_SAFE_DISTANCE 64

#define KB *(1 <<10)
#define MB *(1 <<20)
#define GB *(1U<<30)

#ifndef LZ4_DISTANCE_MAX   /* history window size; can be user-defined at compile time */
#  define LZ4_DISTANCE_MAX 65535   /* set to maximum value by default */
#endif

#define LZ4_DISTANCE_ABSOLUTE_MAX 65535
#if (LZ4_DISTANCE_MAX > LZ4_DISTANCE_ABSOLUTE_MAX)   /* max supported by LZ4 format */
#  error "LZ4_DISTANCE_MAX is too big : must be <= 65535"
#endif

#define ML_BITS  4
#define ML_MASK  ((1U<<ML_BITS)-1)
#define RUN_BITS (8-ML_BITS)
#define RUN_MASK ((1U<<RUN_BITS)-1)

#define DEBUGLOG(l, ...) {}    /* disabled */

#ifndef assert
#define assert ASSERT
#endif

/*-************************************
*  Types
**************************************/
#ifndef _KERNEL
#include <limits.h>
#endif
#if defined(__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */)
#ifndef _KERNEL
#include <stdint.h>
#endif
  typedef  uint8_t BYTE;
  typedef uint16_t U16;
  typedef uint32_t U32;
  typedef  int32_t S32;
  typedef uint64_t U64;
  typedef uintptr_t uptrval;
#else
# if UINT_MAX != 4294967295UL
#   error "LZ4 code (when not C++ or C99) assumes that sizeof(int) == 4"
# endif
  typedef unsigned char       BYTE;
  typedef unsigned short      U16;
  typedef unsigned int        U32;
  typedef   signed int        S32;
  typedef unsigned long long  U64;
  typedef size_t              uptrval;   /* generally true, except OpenVMS-64 */
#endif

#if defined(__x86_64__)
  typedef U64    reg_t;   /* 64-bits in x32 mode */
#else
  typedef size_t reg_t;   /* 32-bits in x32 mode */
#endif

typedef enum {
    notLimited = 0,
    limitedOutput = 1,
    fillOutput = 2
} limitedOutput_directive;


/*-************************************
*  Reading and writing into memory
**************************************/

/**
 * LZ4 relies on memcpy with a constant size being inlined. In freestanding
 * environments, the compiler can't assume the implementation of memcpy() is
 * standard compliant, so it can't apply its specialized memcpy() inlining
 * logic. When possible, use __builtin_memcpy() to tell the compiler to analyze
 * memcpy() as if it were standard compliant, so it can inline it in freestanding
 * environments. This is needed when decompressing the Linux Kernel, for example.
 */
#if defined(__GNUC__) && (__GNUC__ >= 4)
#define LZ4_memcpy(dst, src, size) __builtin_memcpy(dst, src, size)
#else
#define LZ4_memcpy(dst, src, size) memcpy(dst, src, size)
#endif

static unsigned LZ4_isLittleEndian(void)
{
    const union { U32 u; BYTE c[4]; } one = { 1 };   /* don't use static : performance detrimental */
    return one.c[0];
}


#if defined(LZ4_FORCE_MEMORY_ACCESS) && (LZ4_FORCE_MEMORY_ACCESS==2)
/* lie to the compiler about data alignment; use with caution */

static U16 LZ4_read16(const void* memPtr) { return *(const U16*) memPtr; }

static void LZ4_write16(void* memPtr, U16 value) { *(U16*)memPtr = value; }
static void LZ4_write32(void* memPtr, U32 value) { *(U32*)memPtr = value; }

#elif defined(LZ4_FORCE_MEMORY_ACCESS) && (LZ4_FORCE_MEMORY_ACCESS==1)

/* __pack instructions are safer, but compiler specific, hence potentially problematic for some compilers */
/* currently only defined for gcc and icc */
typedef union { U16 u16; U32 u32; reg_t uArch; } __attribute__((packed)) unalign;

static U16 LZ4_read16(const void* ptr) { return ((const unalign*)ptr)->u16; }

static void LZ4_write32(void* memPtr, U32 value) { ((unalign*)memPtr)->u32 = value; }

#else  /* safe and portable access using memcpy() */

static U16 LZ4_read16(const void* memPtr)
{
    U16 val; LZ4_memcpy(&val, memPtr, sizeof(val)); return val;
}

static void LZ4_write32(void* memPtr, U32 value)
{
    LZ4_memcpy(memPtr, &value, sizeof(value));
}

#endif /* LZ4_FORCE_MEMORY_ACCESS */

static U16 LZ4_readLE16(const void* memPtr)
{
    if (LZ4_isLittleEndian()) {
        return LZ4_read16(memPtr);
    } else {
        const BYTE* p = (const BYTE*)memPtr;
        return (U16)((U16)p[0] + (p[1]<<8));
    }
}

/* customized variant of memcpy, which can overwrite up to 8 bytes beyond dstEnd */
LZ4_FORCE_INLINE
void LZ4_wildCopy8(void* dstPtr, const void* srcPtr, void* dstEnd)
{
    BYTE* d = (BYTE*)dstPtr;
    const BYTE* s = (const BYTE*)srcPtr;
    BYTE* const e = (BYTE*)dstEnd;

    do { LZ4_memcpy(d,s,8); d+=8; s+=8; } while (d<e);
}

static const unsigned inc32table[8] = {0, 1, 2,  1,  0,  4, 4, 4};
static const int      dec64table[8] = {0, 0, 0, -1, -4,  1, 2, 3};


#ifndef LZ4_FAST_DEC_LOOP
#  if defined __i386__ || defined _M_IX86 || defined __x86_64__ || defined _M_X64
#    define LZ4_FAST_DEC_LOOP 1
#  elif defined(__aarch64__) && !defined(__clang__)
     /* On aarch64, we disable this optimization for clang because on certain
      * mobile chipsets, performance is reduced with clang. For information
      * refer to https://github.com/lz4/lz4/pull/707 */
#    define LZ4_FAST_DEC_LOOP 1
#  else
#    define LZ4_FAST_DEC_LOOP 0
#  endif
#endif

#if LZ4_FAST_DEC_LOOP

LZ4_FORCE_INLINE void
LZ4_memcpy_using_offset_base(BYTE* dstPtr, const BYTE* srcPtr, BYTE* dstEnd, const size_t offset)
{
    assert(srcPtr + offset == dstPtr);
    if (offset < 8) {
        LZ4_write32(dstPtr, 0);   /* silence an msan warning when offset==0 */
        dstPtr[0] = srcPtr[0];
        dstPtr[1] = srcPtr[1];
        dstPtr[2] = srcPtr[2];
        dstPtr[3] = srcPtr[3];
        srcPtr += inc32table[offset];
        LZ4_memcpy(dstPtr+4, srcPtr, 4);
        srcPtr -= dec64table[offset];
        dstPtr += 8;
    } else {
        LZ4_memcpy(dstPtr, srcPtr, 8);
        dstPtr += 8;
        srcPtr += 8;
    }

    LZ4_wildCopy8(dstPtr, srcPtr, dstEnd);
}

/* customized variant of memcpy, which can overwrite up to 32 bytes beyond dstEnd
 * this version copies two times 16 bytes (instead of one time 32 bytes)
 * because it must be compatible with offsets >= 16. */
LZ4_FORCE_INLINE void
LZ4_wildCopy32(void* dstPtr, const void* srcPtr, void* dstEnd)
{
    BYTE* d = (BYTE*)dstPtr;
    const BYTE* s = (const BYTE*)srcPtr;
    BYTE* const e = (BYTE*)dstEnd;

    do { LZ4_memcpy(d,s,16); LZ4_memcpy(d+16,s+16,16); d+=32; s+=32; } while (d<e);
}

/* LZ4_memcpy_using_offset()  presumes :
 * - dstEnd >= dstPtr + MINMATCH
 * - there is at least 8 bytes available to write after dstEnd */
LZ4_FORCE_INLINE void
LZ4_memcpy_using_offset(BYTE* dstPtr, const BYTE* srcPtr, BYTE* dstEnd, const size_t offset)
{
    BYTE v[8];

    assert(dstEnd >= dstPtr + MINMATCH);

    switch(offset) {
    case 1:
        MEM_INIT(v, *srcPtr, 8);
        break;
    case 2:
        LZ4_memcpy(v, srcPtr, 2);
        LZ4_memcpy(&v[2], srcPtr, 2);
        LZ4_memcpy(&v[4], v, 4);
        break;
    case 4:
        LZ4_memcpy(v, srcPtr, 4);
        LZ4_memcpy(&v[4], srcPtr, 4);
        break;
    default:
        LZ4_memcpy_using_offset_base(dstPtr, srcPtr, dstEnd, offset);
        return;
    }

    LZ4_memcpy(dstPtr, v, 8);
    dstPtr += 8;
    while (dstPtr < dstEnd) {
        LZ4_memcpy(dstPtr, v, 8);
        dstPtr += 8;
    }
}
#endif


/*-************************************
*  Local Structures and types
**************************************/
typedef enum { clearedTable = 0, byPtr, byU32, byU16 } tableType_t;

/**
 * This enum distinguishes several different modes of accessing previous
 * content in the stream.
 *
 * - noDict        : There is no preceding content.
 * - withPrefix64k : Table entries up to ctx->dictSize before the current blob
 *                   blob being compressed are valid and refer to the preceding
 *                   content (of length ctx->dictSize), which is available
 *                   contiguously preceding in memory the content currently
 *                   being compressed.
 * - usingExtDict  : Like withPrefix64k, but the preceding content is somewhere
 *                   else in memory, starting at ctx->dictionary with length
 *                   ctx->dictSize.
 * - usingDictCtx  : Like usingExtDict, but everything concerning the preceding
 *                   content is in a separate context, pointed to by
 *                   ctx->dictCtx. ctx->dictionary, ctx->dictSize, and table
 *                   entries in the current context that refer to positions
 *                   preceding the beginning of the current compression are
 *                   ignored. Instead, ctx->dictCtx->dictionary and ctx->dictCtx
 *                   ->dictSize describe the location and size of the preceding
 *                   content, and matches are found by looking in the ctx
 *                   ->dictCtx->hashTable.
 */
typedef enum { noDict = 0, withPrefix64k, usingExtDict, usingDictCtx } dict_directive;
typedef enum { noDictIssue = 0, dictSmall } dictIssue_directive;

/*-*******************************
 *  Decompression functions
 ********************************/

typedef enum { endOnOutputSize = 0, endOnInputSize = 1 } endCondition_directive;
typedef enum { decode_full_block = 0, partial_decode = 1 } earlyEnd_directive;

typedef enum { loop_error = -2, initial_error = -1, ok = 0 } variable_length_error;

LZ4_FORCE_INLINE unsigned
read_variable_length(const BYTE**ip, const BYTE* lencheck,
                     int loop_check, int initial_check,
                     variable_length_error* error)
{
    U32 length = 0;
    U32 s;
    if (initial_check && unlikely((*ip) >= lencheck)) {    /* overflow detection */
        *error = initial_error;
        return length;
    }
    do {
        s = **ip;
        (*ip)++;
        length += s;
        if (loop_check && unlikely((*ip) >= lencheck)) {    /* overflow detection */
            *error = loop_error;
            return length;
        }
    } while (s==255);

    return length;
}

#define LZ4_STATIC_ASSERT(c)    ASSERT(c)


/*! 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.
 */
LZ4_FORCE_INLINE int
LZ4_decompress_generic(
                 const char* const src,
                 char* const dst,
                 int srcSize,
                 int outputSize,         /* If endOnInput==endOnInputSize, this value is `dstCapacity` */

                 endCondition_directive endOnInput,   /* endOnOutputSize, endOnInputSize */
                 earlyEnd_directive partialDecoding,  /* full, partial */
                 dict_directive dict,                 /* noDict, withPrefix64k, usingExtDict */
                 const BYTE* const lowPrefix,  /* always <= dst, == dst when no prefix */
                 const BYTE* const dictStart,  /* only if dict==usingExtDict */
                 const size_t dictSize         /* note : = 0 if noDict */
                 )
{
    if ((src == NULL) || (outputSize < 0)) { return -1; }

    {   const BYTE* ip = (const BYTE*) src;
        const BYTE* const iend = ip + srcSize;

        BYTE* op = (BYTE*) dst;
        BYTE* const oend = op + outputSize;
        BYTE* cpy;

        const BYTE* const dictEnd = (dictStart == NULL) ? NULL : dictStart + dictSize;

        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*/;

        const BYTE* match;
        size_t offset;
        unsigned token;
        size_t length;


        DEBUGLOG(5, "LZ4_decompress_generic (srcSize:%i, dstSize:%i)", srcSize, outputSize);

        /* Special cases */
        assert(lowPrefix <= op);
        if ((endOnInput) && (unlikely(outputSize==0))) {
            /* Empty output buffer */
            if (partialDecoding) return 0;
            return ((srcSize==1) && (*ip==0)) ? 0 : -1;
        }
        if ((!endOnInput) && (unlikely(outputSize==0))) { return (*ip==0 ? 1 : -1); }
        if ((endOnInput) && unlikely(srcSize==0)) { return -1; }

        /* Currently the fast loop shows a regression on qualcomm arm chips. */
#if LZ4_FAST_DEC_LOOP
        if ((oend - op) < FASTLOOP_SAFE_DISTANCE) {
            DEBUGLOG(6, "skip fast decode loop");
            goto safe_decode;
        }

        /* Fast loop : decode sequences as long as output < iend-FASTLOOP_SAFE_DISTANCE */
        while (1) {
            /* Main fastloop assertion: We can always wildcopy FASTLOOP_SAFE_DISTANCE */
            assert(oend - op >= FASTLOOP_SAFE_DISTANCE);
            if (endOnInput) { assert(ip < iend); }
            token = *ip++;
            length = token >> ML_BITS;  /* literal length */

            assert(!endOnInput || ip <= iend); /* ip < iend before the increment */

            /* decode literal length */
            if (length == RUN_MASK) {
                variable_length_error error = ok;
                length += read_variable_length(&ip, iend-RUN_MASK, (int)endOnInput, (int)endOnInput, &error);
                if (error == initial_error) { goto _output_error; }
                if ((safeDecode) && unlikely((uptrval)(op)+length<(uptrval)(op))) { goto _output_error; } /* overflow detection */
                if ((safeDecode) && unlikely((uptrval)(ip)+length<(uptrval)(ip))) { goto _output_error; } /* overflow detection */

                /* copy literals */
                cpy = op+length;
                LZ4_STATIC_ASSERT(MFLIMIT >= WILDCOPYLENGTH);
                if (endOnInput) {  /* LZ4_decompress_safe() */
                    if ((cpy>oend-32) || (ip+length>iend-32)) { goto safe_literal_copy; }
                    LZ4_wildCopy32(op, ip, cpy);
                } else {   /* LZ4_decompress_fast() */
                    if (cpy>oend-8) { goto safe_literal_copy; }
                    LZ4_wildCopy8(op, ip, cpy); /* LZ4_decompress_fast() cannot copy more than 8 bytes at a time :
                                                 * it doesn't know input length, and only relies on end-of-block properties */
                }
                ip += length; op = cpy;
            } else {
                cpy = op+length;
                if (endOnInput) {  /* LZ4_decompress_safe() */
                    DEBUGLOG(7, "copy %u bytes in a 16-bytes stripe", (unsigned)length);
                    /* We don't need to check oend, since we check it once for each loop below */
                    if (ip > iend-(16 + 1/*max lit + offset + nextToken*/)) { goto safe_literal_copy; }
                    /* Literals can only be 14, but hope compilers optimize if we copy by a register size */
                    LZ4_memcpy(op, ip, 16);
                } else {  /* LZ4_decompress_fast() */
                    /* LZ4_decompress_fast() cannot copy more than 8 bytes at a time :
                     * it doesn't know input length, and relies on end-of-block properties */
                    LZ4_memcpy(op, ip, 8);
                    if (length > 8) { LZ4_memcpy(op+8, ip+8, 8); }
                }
                ip += length; op = cpy;
            }

            /* get offset */
            offset = LZ4_readLE16(ip); ip+=2;
            match = op - offset;
            assert(match <= op);

            /* get matchlength */
            length = token & ML_MASK;

            if (length == ML_MASK) {
                variable_length_error error = ok;
                if ((checkOffset) && (unlikely(match + dictSize < lowPrefix))) { goto _output_error; } /* Error : offset outside buffers */
                length += read_variable_length(&ip, iend - LASTLITERALS + 1, (int)endOnInput, 0, &error);
                if (error != ok) { goto _output_error; }
                if ((safeDecode) && unlikely((uptrval)(op)+length<(uptrval)op)) { goto _output_error; } /* overflow detection */
                length += MINMATCH;
                if (op + length >= oend - FASTLOOP_SAFE_DISTANCE) {
                    goto safe_match_copy;
                }
            } else {
                length += MINMATCH;
                if (op + length >= oend - FASTLOOP_SAFE_DISTANCE) {
                    goto safe_match_copy;
                }

                /* Fastpath check: Avoids a branch in LZ4_wildCopy32 if true */
                if ((dict == withPrefix64k) || (match >= lowPrefix)) {
                    if (offset >= 8) {
                        assert(match >= lowPrefix);
                        assert(match <= op);
                        assert(op + 18 <= oend);

                        LZ4_memcpy(op, match, 8);
                        LZ4_memcpy(op+8, match+8, 8);
                        LZ4_memcpy(op+16, match+16, 2);
                        op += length;
                        continue;
            }   }   }

            if (checkOffset && (unlikely(match + dictSize < lowPrefix))) { goto _output_error; } /* Error : offset outside buffers */
            /* match starting within external dictionary */
            if ((dict==usingExtDict) && (match < lowPrefix)) {
                if (unlikely(op+length > oend-LASTLITERALS)) {
                    if (partialDecoding) {
                        DEBUGLOG(7, "partialDecoding: dictionary match, close to dstEnd");
                        length = MIN(length, (size_t)(oend-op));
                    } else {
                        goto _output_error;  /* end-of-block condition violated */
                }   }

                if (length <= (size_t)(lowPrefix-match)) {
                    /* match fits entirely within external dictionary : just copy */
                    memmove(op, dictEnd - (lowPrefix-match), length);
                    op += length;
                } else {
                    /* match stretches into both external dictionary and current block */
                    size_t const copySize = (size_t)(lowPrefix - match);
                    size_t const restSize = length - copySize;
                    LZ4_memcpy(op, dictEnd - copySize, copySize);
                    op += copySize;
                    if (restSize > (size_t)(op - lowPrefix)) {  /* overlap copy */
                        BYTE* const endOfMatch = op + restSize;
                        const BYTE* copyFrom = lowPrefix;
                        while (op < endOfMatch) { *op++ = *copyFrom++; }
                    } else {
                        LZ4_memcpy(op, lowPrefix, restSize);
                        op += restSize;
                }   }
                continue;
            }

            /* copy match within block */
            cpy = op + length;

            assert((op <= oend) && (oend-op >= 32));
            if (unlikely(offset<16)) {
                LZ4_memcpy_using_offset(op, match, cpy, offset);
            } else {
                LZ4_wildCopy32(op, match, cpy);
            }

            op = cpy;   /* wildcopy correction */
        }
    safe_decode:
#endif

        /* Main Loop : decode remaining sequences where output < FASTLOOP_SAFE_DISTANCE */
        while (1) {
            token = *ip++;
            length = token >> ML_BITS;  /* literal length */

            assert(!endOnInput || ip <= iend); /* ip < iend before the increment */

            /* 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 */
                LZ4_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. */
                    LZ4_memcpy(op + 0, match + 0, 8);
                    LZ4_memcpy(op + 8, match + 8, 8);
                    LZ4_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) {
                variable_length_error error = ok;
                length += read_variable_length(&ip, iend-RUN_MASK, (int)endOnInput, (int)endOnInput, &error);
                if (error == initial_error) { goto _output_error; }
                if ((safeDecode) && unlikely((uptrval)(op)+length<(uptrval)(op))) { goto _output_error; } /* overflow detection */
                if ((safeDecode) && unlikely((uptrval)(ip)+length<(uptrval)(ip))) { goto _output_error; } /* overflow detection */
            }

            /* copy literals */
            cpy = op+length;
#if LZ4_FAST_DEC_LOOP
        safe_literal_copy:
#endif
            LZ4_STATIC_ASSERT(MFLIMIT >= WILDCOPYLENGTH);
            if ( ((endOnInput) && ((cpy>oend-MFLIMIT) || (ip+length>iend-(2+1+LASTLITERALS))) )
              || ((!endOnInput) && (cpy>oend-WILDCOPYLENGTH)) )
            {
                /* We've either hit the input parsing restriction or the output parsing restriction.
                 * In the normal scenario, decoding a full block, it must be the last sequence,
                 * otherwise it's an error (invalid input or dimensions).
                 * In partialDecoding scenario, it's necessary to ensure there is no buffer overflow.
                 */
                if (partialDecoding) {
                    /* Since we are partial decoding we may be in this block because of the output parsing
                     * restriction, which is not valid since the output buffer is allowed to be undersized.
                     */
                    assert(endOnInput);
                    DEBUGLOG(7, "partialDecoding: copying literals, close to input or output end")
                    DEBUGLOG(7, "partialDecoding: literal length = %u", (unsigned)length);
                    DEBUGLOG(7, "partialDecoding: remaining space in dstBuffer : %i", (int)(oend - op));
                    DEBUGLOG(7, "partialDecoding: remaining space in srcBuffer : %i", (int)(iend - ip));
                    /* Finishing in the middle of a literals segment,
                     * due to lack of input.
                     */
                    if (ip+length > iend) {
                        length = (size_t)(iend-ip);
                        cpy = op + length;
                    }
                    /* Finishing in the middle of a literals segment,
                     * due to lack of output space.
                     */
                    if (cpy > oend) {
                        cpy = oend;
                        assert(op<=oend);
                        length = (size_t)(oend-op);
                    }
                } else {
                    /* We must be on the last sequence because of the parsing limitations so check
                     * that we exactly regenerate the original size (must be exact when !endOnInput).
                     */
                    if ((!endOnInput) && (cpy != oend)) { goto _output_error; }
                     /* We must be on the last sequence (or invalid) because of the parsing limitations
                      * so check that we exactly consume the input and don't overrun the output buffer.
                      */
                    if ((endOnInput) && ((ip+length != iend) || (cpy > oend))) {
                        DEBUGLOG(6, "should have been last run of literals")
                        DEBUGLOG(6, "ip(%p) + length(%i) = %p != iend (%p)", ip, (int)length, ip+length, iend);
                        DEBUGLOG(6, "or cpy(%p) > oend(%p)", cpy, oend);
                        goto _output_error;
                    }
                }
                memmove(op, ip, length);  /* supports overlapping memory regions; only matters for in-place decompression scenarios */
                ip += length;
                op += length;
                /* Necessarily EOF when !partialDecoding.
                 * When partialDecoding, it is EOF if we've either
                 * filled the output buffer or
                 * can't proceed with reading an offset for following match.
                 */
                if (!partialDecoding || (cpy == oend) || (ip >= (iend-2))) {
                    break;
                }
            } else {
                LZ4_wildCopy8(op, ip, cpy);   /* may overwrite up to WILDCOPYLENGTH beyond cpy */
                ip += length; op = cpy;
            }

            /* get offset */
            offset = LZ4_readLE16(ip); ip+=2;
            match = op - offset;

            /* get matchlength */
            length = token & ML_MASK;

    _copy_match:
            if (length == ML_MASK) {
              variable_length_error error = ok;
              length += read_variable_length(&ip, iend - LASTLITERALS + 1, (int)endOnInput, 0, &error);
              if (error != ok) goto _output_error;
                if ((safeDecode) && unlikely((uptrval)(op)+length<(uptrval)op)) goto _output_error;   /* overflow detection */
            }
            length += MINMATCH;

#if LZ4_FAST_DEC_LOOP
        safe_match_copy:
#endif
            if ((checkOffset) && (unlikely(match + dictSize < lowPrefix))) goto _output_error;   /* Error : offset outside buffers */
            /* match starting within external dictionary */
            if ((dict==usingExtDict) && (match < lowPrefix)) {
                if (unlikely(op+length > oend-LASTLITERALS)) {
                    if (partialDecoding) length = MIN(length, (size_t)(oend-op));
                    else goto _output_error;   /* doesn't respect parsing restriction */
                }

                if (length <= (size_t)(lowPrefix-match)) {
                    /* match fits entirely within external dictionary : just copy */
                    memmove(op, dictEnd - (lowPrefix-match), length);
                    op += length;
                } else {
                    /* match stretches into both external dictionary and current block */
                    size_t const copySize = (size_t)(lowPrefix - match);
                    size_t const restSize = length - copySize;
                    LZ4_memcpy(op, dictEnd - copySize, copySize);
                    op += copySize;
                    if (restSize > (size_t)(op - lowPrefix)) {  /* overlap copy */
                        BYTE* const endOfMatch = op + restSize;
                        const BYTE* copyFrom = lowPrefix;
                        while (op < endOfMatch) *op++ = *copyFrom++;
                    } else {
                        LZ4_memcpy(op, lowPrefix, restSize);
                        op += restSize;
                }   }
                continue;
            }
            assert(match >= lowPrefix);

            /* copy match within block */
            cpy = op + length;

            /* partialDecoding : may end anywhere within the block */
            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 {
                    LZ4_memcpy(op, match, mlen);
                }
                op = copyEnd;
                if (op == oend) { break; }
                continue;
            }

            if (unlikely(offset<8)) {
                LZ4_write32(op, 0);   /* silence msan warning when offset==0 */
                op[0] = match[0];
                op[1] = match[1];
                op[2] = match[2];
                op[3] = match[3];
                match += inc32table[offset];
                LZ4_memcpy(op+4, match, 4);
                match -= dec64table[offset];
            } else {
                LZ4_memcpy(op, match, 8);
                match += 8;
            }
            op += 8;

            if (unlikely(cpy > oend-MATCH_SAFEGUARD_DISTANCE)) {
                BYTE* const oCopyLimit = oend - (WILDCOPYLENGTH-1);
                if (cpy > oend-LASTLITERALS) { goto _output_error; } /* Error : last LASTLITERALS bytes must be literals (uncompressed) */
                if (op < oCopyLimit) {
                    LZ4_wildCopy8(op, match, oCopyLimit);
                    match += oCopyLimit - op;
                    op = oCopyLimit;
                }
                while (op < cpy) { *op++ = *match++; }
            } else {
                LZ4_memcpy(op, match, 8);
                if (length > 16)  { LZ4_wildCopy8(op+8, match+8, cpy); }
            }
            op = cpy;   /* wildcopy correction */
        }

        /* end of decoding */
        if (endOnInput) {
            DEBUGLOG(5, "decoded %i bytes", (int) (((char*)op)-dst));
           return (int) (((char*)op)-dst);     /* Nb of output bytes decoded */
       } else {
           return (int) (((const char*)ip)-src);   /* Nb of input bytes read */
       }

        /* Overflow error detected */
    _output_error:
        return (int) (-(((const char*)ip)-src))-1;
    }
}

/*
 * LZ4_uncompress_unknownOutputSize() :
 *      isize  : is the input size, therefore the compressed size
 *      maxOutputSize : is the size of the destination buffer (which must be
 *              already allocated)
 *      return : the number of bytes decoded in the destination buffer
 *              (necessarily <= maxOutputSize). If the source stream is
 *              malformed, the function will stop decoding and return a
 *              negative result, indicating the byte position of the faulty
 *              instruction. This function never writes beyond dest +
 *              maxOutputSize, and is therefore protected against malicious
 *              data packets.
 *      note   : Destination buffer must be already allocated.
 *              This version is slightly slower than real_LZ4_uncompress()
 *
 */

/*
 * Note: In upstream code, LZ4_uncompress_unknownOutputSize is now a legacy
 *       wrapper for LZ4_decompress_safe which is a wrapper for
 *       LZ4_decompress_generic; this wrapper flattens that, rather than
 *       rewriting the callers.
 */
int LZ4_uncompress_unknownOutputSize(const char* source, char* dest, int compressedSize, int maxDecompressedSize)
{
    return LZ4_decompress_generic(source, dest, compressedSize, maxDecompressedSize,
                                  endOnInputSize, decode_full_block, noDict,
                                  (BYTE*)dest, NULL, 0);
}