chm_lib.c

又一开源的解压缩chm格式文件的代码

#define _CHM_PMGL_LEN (0x14)
struct chmPmglHeader
{
    char        signature[4];           /*  0 (PMGL) */
    UInt32      free_space;             /*  4 */
    UInt32      unknown_0008;           /*  8 */
    Int32       block_prev;             /*  c */
    Int32       block_next;             /* 10 */
}; /* __attribute__ ((aligned (1))); */

static int _unmarshal_pmgl_header(unsigned char **pData,
                                  unsigned long *pDataLen,
                                  struct chmPmglHeader *dest)
{
    /* we only know how to deal with a 0x14 byte structures */
    if (*pDataLen != _CHM_PMGL_LEN)
        return 0;

    /* unmarshal fields */
    _unmarshal_char_array(pData, pDataLen,  dest->signature, 4);
    _unmarshal_uint32    (pData, pDataLen, &dest->free_space);
    _unmarshal_uint32    (pData, pDataLen, &dest->unknown_0008);
    _unmarshal_int32     (pData, pDataLen, &dest->block_prev);
    _unmarshal_int32     (pData, pDataLen, &dest->block_next);

    /* check structure */
    if (memcmp(dest->signature, _chm_pmgl_marker, 4) != 0)
        return 0;

    return 1;
}

/* structure of PMGI headers */
static const char _chm_pmgi_marker[4] = "PMGI";
#define _CHM_PMGI_LEN (0x08)
struct chmPmgiHeader
{
    char        signature[4];           /*  0 (PMGI) */
    UInt32      free_space;             /*  4 */
}; /* __attribute__ ((aligned (1))); */

static int _unmarshal_pmgi_header(unsigned char **pData,
                                  unsigned long *pDataLen,
                                  struct chmPmgiHeader *dest)
{
    /* we only know how to deal with a 0x8 byte structures */
    if (*pDataLen != _CHM_PMGI_LEN)
        return 0;

    /* unmarshal fields */
    _unmarshal_char_array(pData, pDataLen,  dest->signature, 4);
    _unmarshal_uint32    (pData, pDataLen, &dest->free_space);

    /* check structure */
    if (memcmp(dest->signature, _chm_pmgi_marker, 4) != 0)
        return 0;

    return 1;
}

/* structure of LZXC reset table */
#define _CHM_LZXC_RESETTABLE_V1_LEN (0x28)
struct chmLzxcResetTable
{
    UInt32      version;
    UInt32      block_count;
    UInt32      unknown;
    UInt32      table_offset;
    UInt64      uncompressed_len;
    UInt64      compressed_len;
    UInt64      block_len;     
}; /* __attribute__ ((aligned (1))); */

static int _unmarshal_lzxc_reset_table(unsigned char **pData,
                                       unsigned long *pDataLen,
                                       struct chmLzxcResetTable *dest)
{
    /* we only know how to deal with a 0x28 byte structures */
    if (*pDataLen != _CHM_LZXC_RESETTABLE_V1_LEN)
        return 0;

    /* unmarshal fields */
    _unmarshal_uint32    (pData, pDataLen, &dest->version);
    _unmarshal_uint32    (pData, pDataLen, &dest->block_count);
    _unmarshal_uint32    (pData, pDataLen, &dest->unknown);
    _unmarshal_uint32    (pData, pDataLen, &dest->table_offset);
    _unmarshal_uint64    (pData, pDataLen, &dest->uncompressed_len);
    _unmarshal_uint64    (pData, pDataLen, &dest->compressed_len);
    _unmarshal_uint64    (pData, pDataLen, &dest->block_len);

    /* check structure */
    if (dest->version != 2)
        return 0;

    return 1;
}

/* structure of LZXC control data block */
#define _CHM_LZXC_MIN_LEN (0x18)
#define _CHM_LZXC_V2_LEN (0x1c)
struct chmLzxcControlData
{
    UInt32      size;                   /*  0        */
    char        signature[4];           /*  4 (LZXC) */
    UInt32      version;                /*  8        */
    UInt32      resetInterval;          /*  c        */
    UInt32      windowSize;             /* 10        */
    UInt32      windowsPerReset;        /* 14        */
    UInt32      unknown_18;             /* 18        */
};

static int _unmarshal_lzxc_control_data(unsigned char **pData,
                                        unsigned long *pDataLen,
                                        struct chmLzxcControlData *dest)
{
    /* we want at least 0x18 bytes */
    if (*pDataLen < _CHM_LZXC_MIN_LEN)
        return 0;

    /* unmarshal fields */
    _unmarshal_uint32    (pData, pDataLen, &dest->size);
    _unmarshal_char_array(pData, pDataLen,  dest->signature, 4);
    _unmarshal_uint32    (pData, pDataLen, &dest->version);
    _unmarshal_uint32    (pData, pDataLen, &dest->resetInterval);
    _unmarshal_uint32    (pData, pDataLen, &dest->windowSize);
    _unmarshal_uint32    (pData, pDataLen, &dest->windowsPerReset);

    if (*pDataLen >= _CHM_LZXC_V2_LEN)
        _unmarshal_uint32    (pData, pDataLen, &dest->unknown_18);
    else
        dest->unknown_18 = 0;

    if (dest->version == 2)
    {
        dest->resetInterval *= 0x8000;
        dest->windowSize *= 0x8000;
    }
    if (dest->windowSize == 0  ||  dest->resetInterval == 0)
        return 0;

    /* for now, only support resetInterval a multiple of windowSize/2 */
    if (dest->windowSize == 1)
        return 0;
    if ((dest->resetInterval % (dest->windowSize/2)) != 0)
        return 0;

    /* check structure */
    if (memcmp(dest->signature, "LZXC", 4) != 0)
        return 0;

    return 1;
}

/* the structure used for chm file handles */
struct chmFile
{
#ifdef WIN32
    HANDLE              fd;
#else
    int                 fd;
#endif

#ifdef CHM_MT
#ifdef WIN32
    CRITICAL_SECTION    mutex;
    CRITICAL_SECTION    lzx_mutex;
    CRITICAL_SECTION    cache_mutex;
#else
    pthread_mutex_t     mutex;
    pthread_mutex_t     lzx_mutex;
    pthread_mutex_t     cache_mutex;
#endif
#endif

    UInt64              dir_offset;
    UInt64              dir_len;    
    UInt64              data_offset;
    Int32               index_root;
    Int32               index_head;
    UInt32              block_len;     

    UInt64              span;
    struct chmUnitInfo  rt_unit;
    struct chmUnitInfo  cn_unit;
    struct chmLzxcResetTable reset_table;

    /* LZX control data */
    int                 compression_enabled;
    UInt32              window_size;
    UInt32              reset_interval;
    UInt32              reset_blkcount;

    /* decompressor state */
    struct LZXstate    *lzx_state;
    int                 lzx_last_block;

    /* cache for decompressed blocks */
    UChar             **cache_blocks;
    Int64              *cache_block_indices;
    Int32               cache_num_blocks;
};

/*
 * utility functions local to this module
 */

/* utility function to handle differences between {pread,read}(64)? */
static Int64 _chm_fetch_bytes(struct chmFile *h,
                              UChar *buf,
                              UInt64 os,
                              Int64 len)
{
    Int64 readLen=0, oldOs=0;
    if (h->fd  ==  CHM_NULL_FD)
        return readLen;

    CHM_ACQUIRE_LOCK(h->mutex);
#ifdef CHM_USE_WIN32IO
    /* NOTE: this might be better done with CreateFileMapping, et cetera... */
    {
        DWORD origOffsetLo=0, origOffsetHi=0;
        DWORD offsetLo, offsetHi;
        DWORD actualLen=0;

        /* awkward Win32 Seek/Tell */
        offsetLo = (unsigned long)(os & 0xffffffffL);
        offsetHi = (unsigned long)((os >> 32) & 0xffffffffL);
        origOffsetLo = SetFilePointer(h->fd, 0, &origOffsetHi, FILE_CURRENT);
        offsetLo = SetFilePointer(h->fd, offsetLo, &offsetHi, FILE_BEGIN);

        /* read the data */
        if (ReadFile(h->fd,
                     buf,
                     (DWORD)len,
                     &actualLen,
                     NULL) == TRUE)
            readLen = actualLen;
        else
            readLen = 0;

        /* restore original position */
        SetFilePointer(h->fd, origOffsetLo, &origOffsetHi, FILE_BEGIN);
    }
#else
#ifdef CHM_USE_PREAD
#ifdef CHM_USE_IO64
    readLen = pread64(h->fd, buf, (long)len, os);
#else
    readLen = pread(h->fd, buf, (long)len, (unsigned long)os);
#endif
#else
#ifdef CHM_USE_IO64
    oldOs = lseek64(h->fd, 0, SEEK_CUR);
    lseek64(h->fd, os, SEEK_SET);
    readLen = read(h->fd, buf, len);
    lseek64(h->fd, oldOs, SEEK_SET);
#else
    oldOs = lseek(h->fd, 0, SEEK_CUR);
    lseek(h->fd, (long)os, SEEK_SET);
    readLen = read(h->fd, buf, len);
    lseek(h->fd, (long)oldOs, SEEK_SET);
#endif
#endif
#endif
    CHM_RELEASE_LOCK(h->mutex);
    return readLen;
}

/* open an ITS archive */
#ifdef PPC_BSTR
/* RWE 6/12/2003 */
struct chmFile *chm_open(BSTR filename)
#else
struct chmFile *chm_open(const char *filename)
#endif
{
    unsigned char               sbuffer[256];
    unsigned long               sremain;
    unsigned char              *sbufpos;
    struct chmFile             *newHandle=NULL;
    struct chmItsfHeader        itsfHeader;
    struct chmItspHeader        itspHeader;
    struct chmUnitInfo          uiSpan;
    struct chmUnitInfo          uiLzxc;
    struct chmLzxcControlData   ctlData;

    /* allocate handle */
    newHandle = (struct chmFile *)malloc(sizeof(struct chmFile));
    newHandle->fd = CHM_NULL_FD;
    newHandle->lzx_state = NULL;
    newHandle->cache_blocks = NULL;
    newHandle->cache_block_indices = NULL;
    newHandle->cache_num_blocks = 0;

    /* open file */
#ifdef WIN32
#ifdef PPC_BSTR
    if ((newHandle->fd=CreateFile(filename,
				  GENERIC_READ,
                                  FILE_SHARE_READ,
				  NULL,
				  OPEN_EXISTING,
				  FILE_ATTRIBUTE_NORMAL,
				  NULL)) == CHM_NULL_FD)
    {
        free(newHandle);
        return NULL;
    }
#else
    if ((newHandle->fd=CreateFileA(filename,
                                   GENERIC_READ,
                                   0,
                                   NULL,
                                   OPEN_EXISTING,
                                   FILE_ATTRIBUTE_NORMAL,
                                   NULL)) == CHM_NULL_FD)
    {
        free(newHandle);
        return NULL;
    }
#endif
#else
    if ((newHandle->fd=open(filename, O_RDONLY)) == CHM_NULL_FD)
    {
        free(newHandle);
        return NULL;
    }
#endif

    /* initialize mutexes, if needed */
#ifdef CHM_MT
#ifdef WIN32
    InitializeCriticalSection(&newHandle->mutex);
    InitializeCriticalSection(&newHandle->lzx_mutex);
    InitializeCriticalSection(&newHandle->cache_mutex);
#else
    pthread_mutex_init(&newHandle->mutex, NULL);
    pthread_mutex_init(&newHandle->lzx_mutex, NULL);
    pthread_mutex_init(&newHandle->cache_mutex, NULL);
#endif
#endif

    /* read and verify header */
    sremain = _CHM_ITSF_V3_LEN;
    sbufpos = sbuffer;
    if (_chm_fetch_bytes(newHandle, sbuffer, (UInt64)0, sremain) != sremain    ||
        !_unmarshal_itsf_header(&sbufpos, &sremain, &itsfHeader))
    {
        chm_close(newHandle);
        return NULL;
    }

    /* stash important values from header */
    newHandle->dir_offset  = itsfHeader.dir_offset;
    newHandle->dir_len     = itsfHeader.dir_len;
    newHandle->data_offset = itsfHeader.data_offset;

    /* now, read and verify the directory header chunk */
    sremain = _CHM_ITSP_V1_LEN;
    sbufpos = sbuffer;
    if (_chm_fetch_bytes(newHandle, sbuffer,
                         (UInt64)itsfHeader.dir_offset, sremain) != sremain       ||
        !_unmarshal_itsp_header(&sbufpos, &sremain, &itspHeader))
    {
        chm_close(newHandle);
        return NULL;
    }

    /* grab essential information from ITSP header */
    newHandle->dir_offset += itspHeader.header_len;
    newHandle->dir_len    -= itspHeader.header_len;
    newHandle->index_root  = itspHeader.index_root;
    newHandle->index_head  = itspHeader.index_head;
    newHandle->block_len   = itspHeader.block_len;

    /* if the index root is -1, this means we don't have any PMGI blocks.
     * as a result, we must use the sole PMGL block as the index root
     */
    if (newHandle->index_root == -1)
        newHandle->index_root = newHandle->index_head;

    /* By default, compression is enabled. */
    newHandle->compression_enabled = 1;

/* Jed, Sun Jun 27: 'span' doesn't seem to be used anywhere?! */
#if 0
    /* fetch span */
    if (CHM_RESOLVE_SUCCESS != chm_resolve_object(newHandle,
                                                  _CHMU_SPANINFO,
                                                  &uiSpan)                ||
        uiSpan.space == CHM_COMPRESSED)
    {
        chm_close(newHandle);
        return NULL;
    }

    /* N.B.: we've already checked that uiSpan is in the uncompressed section,
     *       so this should not require attempting to decompress, which may
     *       rely on having a valid "span"
     */
    sremain = 8;
    sbufpos = sbuffer;
    if (chm_retrieve_object(newHandle, &uiSpan, sbuffer,
                            0, sremain) != sremain                        ||
        !_unmarshal_uint64(&sbufpos, &sremain, &newHandle->span))
    {
        chm_close(newHandle);
        return NULL;
    }
#endif

    /* prefetch most commonly needed unit infos */
    if (CHM_RESOLVE_SUCCESS != chm_resolve_object(newHandle,
                                                  _CHMU_RESET_TABLE,
                                                  &newHandle->rt_unit)    ||
        newHandle->rt_unit.space == CHM_COMPRESSED                        ||
        CHM_RESOLVE_SUCCESS != chm_resolve_object(newHandle,
                                                  _CHMU_CONTENT,
                                                  &newHandle->cn_unit)    ||
        newHandle->cn_unit.space == CHM_COMPRESSED                        ||
        CHM_RESOLVE_SUCCESS != chm_resolve_object(newHandle,
                                                  _CHMU_LZXC_CONTROLDATA,
                                                  &uiLzxc)                ||
        uiLzxc.space == CHM_COMPRESSED)
    {
        newHandle->compression_enabled = 0;
    }

    /* read reset table info */
    if (newHandle->compression_enabled)
    {
        sremain = _CHM_LZXC_RESETTABLE_V1_LEN;
        sbufpos = sbuffer;
        if (chm_retrieve_object(newHandle, &newHandle->rt_unit, sbuffer,
                                0, sremain) != sremain                        ||
            !_unmarshal_lzxc_reset_table(&sbufpos, &sremain,
                                         &newHandle->reset_table))
        {
            newHandle->compression_enabled = 0;
        }
    }

    /* read control data */
    if (newHandle->compression_enabled)
    {
        sremain = (unsigned long)uiLzxc.length;
        sbufpos = sbuffer;
        if (chm_retrieve_object(newHandle, &uiLzxc, sbuffer,
                                0, sremain) != sremain                       ||
            !_unmarshal_lzxc_control_data(&sbufpos, &sremain,
                                          &ctlData))
        {
            newHandle->compression_enabled = 0;
        }

        newHandle->window_size = ctlData.windowSize;
        newHandle->reset_interval = ctlData.resetInterval;

/* Jed, Mon Jun 28: Experimentally, it appears that the reset block count */
/*       must be multiplied by this formerly unknown ctrl data field in   */
/*       order to decompress some files.                                  */
#if 0
        newHandle->reset_blkcount = newHandle->reset_interval /
                    (newHandle->window_size / 2);
#else
        newHandle->reset_blkcount = newHandle->reset_interval    /
                                    (newHandle->window_size / 2) *
                                    ctlData.windowsPerReset;
#endif
    }

    /* initialize cache */
    chm_set_param(newHandle, CHM_PARAM_MAX_BLOCKS_CACHED,