LXR busybox/archival/libarchive/decompress

   1/* vi: set sw=4 ts=4: */
   2/*
   3 * uncompress for busybox -- (c) 2002 Robert Griebl
   4 *
   5 * based on the original compress42.c source
   6 * (see disclaimer below)
   7 */
   8/* (N)compress42.c - File compression ala IEEE Computer, Mar 1992.
   9 *
  10 * Authors:
  11 *   Spencer W. Thomas   (decvax!harpo!utah-cs!utah-gr!thomas)
  12 *   Jim McKie           (decvax!mcvax!jim)
  13 *   Steve Davies        (decvax!vax135!petsd!peora!srd)
  14 *   Ken Turkowski       (decvax!decwrl!turtlevax!ken)
  15 *   James A. Woods      (decvax!ihnp4!ames!jaw)
  16 *   Joe Orost           (decvax!vax135!petsd!joe)
  17 *   Dave Mack           (csu@alembic.acs.com)
  18 *   Peter Jannesen, Network Communication Systems
  19 *                       (peter@ncs.nl)
  20 *
  21 * marc@suse.de : a small security fix for a buffer overflow
  22 *
  23 * [... History snipped ...]
  24 */
  25#include "libbb.h"
  26#include "bb_archive.h"
  27
  28
  29/* Default input buffer size */
  30#define IBUFSIZ 2048
  31
  32/* Default output buffer size */
  33#define OBUFSIZ 2048
  34
  35/* Defines for third byte of header */
  36#define BIT_MASK        0x1f    /* Mask for 'number of compresssion bits'       */
  37                                /* Masks 0x20 and 0x40 are free.                */
  38                                /* I think 0x20 should mean that there is       */
  39                                /* a fourth header byte (for expansion).        */
  40#define BLOCK_MODE      0x80    /* Block compression if table is full and       */
  41                                /* compression rate is dropping flush tables    */
  42                                /* the next two codes should not be changed lightly, as they must not   */
  43                                /* lie within the contiguous general code space.                        */
  44#define FIRST   257     /* first free entry */
  45#define CLEAR   256     /* table clear output code */
  46
  47#define INIT_BITS 9     /* initial number of bits/code */
  48
  49
  50/* machine variants which require cc -Dmachine:  pdp11, z8000, DOS */
  51#define HBITS      17   /* 50% occupancy */
  52#define HSIZE      (1<<HBITS)
  53#define HMASK      (HSIZE-1)    /* unused */
  54#define HPRIME     9941         /* unused */
  55#define BITS       16
  56#define BITS_STR   "16"
  57#undef  MAXSEG_64K              /* unused */
  58#define MAXCODE(n) (1L << (n))
  59
  60#define htabof(i)               htab[i]
  61#define codetabof(i)            codetab[i]
  62#define tab_prefixof(i)         codetabof(i)
  63#define tab_suffixof(i)         ((unsigned char *)(htab))[i]
  64#define de_stack                ((unsigned char *)&(htab[HSIZE-1]))
  65#define clear_tab_prefixof()    memset(codetab, 0, 256)
  66
  67/*
  68 * Decompress stdin to stdout.  This routine adapts to the codes in the
  69 * file building the "string" table on-the-fly; requiring no table to
  70 * be stored in the compressed file.
  71 */
  72
  73IF_DESKTOP(long long) int FAST_FUNC
  74unpack_Z_stream(transformer_state_t *xstate)
  75{
  76        IF_DESKTOP(long long total_written = 0;)
  77        IF_DESKTOP(long long) int retval = -1;
  78        unsigned char *stackp;
  79        int finchar;
  80        long oldcode;
  81        long incode;
  82        int inbits;
  83        int posbits;
  84        int outpos;
  85        int insize;
  86        int bitmask;
  87        long free_ent;
  88        long maxcode;
  89        long maxmaxcode;
  90        int n_bits;
  91        int rsize = 0;
  92        unsigned char *inbuf; /* were eating insane amounts of stack - */
  93        unsigned char *outbuf; /* bad for some embedded targets */
  94        unsigned char *htab;
  95        unsigned short *codetab;
  96
  97        /* Hmm, these were statics - why?! */
  98        /* user settable max # bits/code */
  99        int maxbits; /* = BITS; */
 100        /* block compress mode -C compatible with 2.0 */
 101        int block_mode; /* = BLOCK_MODE; */
 102
 103        if (check_signature16(xstate, COMPRESS_MAGIC))
 104                return -1;
 105
 106        inbuf = xzalloc(IBUFSIZ + 64);
 107        outbuf = xzalloc(OBUFSIZ + 2048);
 108        htab = xzalloc(HSIZE);  /* wasn't zeroed out before, maybe can xmalloc? */
 109        codetab = xzalloc(HSIZE * sizeof(codetab[0]));
 110
 111        insize = 0;
 112
 113        /* xread isn't good here, we have to return - caller may want
 114         * to do some cleanup (e.g. delete incomplete unpacked file etc) */
 115        if (full_read(xstate->src_fd, inbuf, 1) != 1) {
 116                bb_error_msg("short read");
 117                goto err;
 118        }
 119
 120        maxbits = inbuf[0] & BIT_MASK;
 121        block_mode = inbuf[0] & BLOCK_MODE;
 122        maxmaxcode = MAXCODE(maxbits);
 123
 124        if (maxbits > BITS) {
 125                bb_error_msg("compressed with %d bits, can only handle "
 126                                BITS_STR" bits", maxbits);
 127                goto err;
 128        }
 129
 130        n_bits = INIT_BITS;
 131        maxcode = MAXCODE(INIT_BITS) - 1;
 132        bitmask = (1 << INIT_BITS) - 1;
 133        oldcode = -1;
 134        finchar = 0;
 135        outpos = 0;
 136        posbits = 0 << 3;
 137
 138        free_ent = ((block_mode) ? FIRST : 256);
 139
 140        /* As above, initialize the first 256 entries in the table. */
 141        /*clear_tab_prefixof(); - done by xzalloc */
 142
 143        {
 144                int i;
 145                for (i = 255; i >= 0; --i)
 146                        tab_suffixof(i) = (unsigned char) i;
 147        }
 148
 149        do {
 150 resetbuf:
 151                {
 152                        int i;
 153                        int e;
 154                        int o;
 155
 156                        o = posbits >> 3;
 157                        e = insize - o;
 158
 159                        for (i = 0; i < e; ++i)
 160                                inbuf[i] = inbuf[i + o];
 161
 162                        insize = e;
 163                        posbits = 0;
 164                }
 165
 166                if (insize < (int) (IBUFSIZ + 64) - IBUFSIZ) {
 167                        rsize = safe_read(xstate->src_fd, inbuf + insize, IBUFSIZ);
 168                        if (rsize < 0)
 169                                bb_error_msg_and_die(bb_msg_read_error);
 170                        insize += rsize;
 171                }
 172
 173                inbits = ((rsize > 0) ? (insize - insize % n_bits) << 3 :
 174                                  (insize << 3) - (n_bits - 1));
 175
 176                while (inbits > posbits) {
 177                        long code;
 178
 179                        if (free_ent > maxcode) {
 180                                posbits =
 181                                        ((posbits - 1) +
 182                                         ((n_bits << 3) -
 183                                          (posbits - 1 + (n_bits << 3)) % (n_bits << 3)));
 184                                ++n_bits;
 185                                if (n_bits == maxbits) {
 186                                        maxcode = maxmaxcode;
 187                                } else {
 188                                        maxcode = MAXCODE(n_bits) - 1;
 189                                }
 190                                bitmask = (1 << n_bits) - 1;
 191                                goto resetbuf;
 192                        }
 193                        {
 194                                unsigned char *p = &inbuf[posbits >> 3];
 195                                code = ((p[0]
 196                                        | ((long) (p[1]) << 8)
 197                                        | ((long) (p[2]) << 16)) >> (posbits & 0x7)) & bitmask;
 198                        }
 199                        posbits += n_bits;
 200
 201                        if (oldcode == -1) {
 202                                if (code >= 256)
 203                                        bb_error_msg_and_die("corrupted data"); /* %ld", code); */
 204                                oldcode = code;
 205                                finchar = (int) oldcode;
 206                                outbuf[outpos++] = (unsigned char) finchar;
 207                                continue;
 208                        }
 209
 210                        if (code == CLEAR && block_mode) {
 211                                clear_tab_prefixof();
 212                                free_ent = FIRST - 1;
 213                                posbits =
 214                                        ((posbits - 1) +
 215                                         ((n_bits << 3) -
 216                                          (posbits - 1 + (n_bits << 3)) % (n_bits << 3)));
 217                                n_bits = INIT_BITS;
 218                                maxcode = MAXCODE(INIT_BITS) - 1;
 219                                bitmask = (1 << INIT_BITS) - 1;
 220                                goto resetbuf;
 221                        }
 222
 223                        incode = code;
 224                        stackp = de_stack;
 225
 226                        /* Special case for KwKwK string. */
 227                        if (code >= free_ent) {
 228                                if (code > free_ent) {
 229/*
 230                                        unsigned char *p;
 231
 232                                        posbits -= n_bits;
 233                                        p = &inbuf[posbits >> 3];
 234                                        bb_error_msg
 235                                                ("insize:%d posbits:%d inbuf:%02X %02X %02X %02X %02X (%d)",
 236                                                insize, posbits, p[-1], p[0], p[1], p[2], p[3],
 237                                                (posbits & 07));
 238*/
 239                                        bb_error_msg("corrupted data");
 240                                        goto err;
 241                                }
 242
 243                                *--stackp = (unsigned char) finchar;
 244                                code = oldcode;
 245                        }
 246
 247                        /* Generate output characters in reverse order */
 248                        while (code >= 256) {
 249                                if (stackp <= &htabof(0))
 250                                        bb_error_msg_and_die("corrupted data");
 251                                *--stackp = tab_suffixof(code);
 252                                code = tab_prefixof(code);
 253                        }
 254
 255                        finchar = tab_suffixof(code);
 256                        *--stackp = (unsigned char) finchar;
 257
 258                        /* And put them out in forward order */
 259                        {
 260                                int i;
 261
 262                                i = de_stack - stackp;
 263                                if (outpos + i >= OBUFSIZ) {
 264                                        do {
 265                                                if (i > OBUFSIZ - outpos) {
 266                                                        i = OBUFSIZ - outpos;
 267                                                }
 268
 269                                                if (i > 0) {
 270                                                        memcpy(outbuf + outpos, stackp, i);
 271                                                        outpos += i;
 272                                                }
 273
 274                                                if (outpos >= OBUFSIZ) {
 275                                                        xtransformer_write(xstate, outbuf, outpos);
 276                                                        IF_DESKTOP(total_written += outpos;)
 277                                                        outpos = 0;
 278                                                }
 279                                                stackp += i;
 280                                                i = de_stack - stackp;
 281                                        } while (i > 0);
 282                                } else {
 283                                        memcpy(outbuf + outpos, stackp, i);
 284                                        outpos += i;
 285                                }
 286                        }
 287
 288                        /* Generate the new entry. */
 289                        if (free_ent < maxmaxcode) {
 290                                tab_prefixof(free_ent) = (unsigned short) oldcode;
 291                                tab_suffixof(free_ent) = (unsigned char) finchar;
 292                                free_ent++;
 293                        }
 294
 295                        /* Remember previous code.  */
 296                        oldcode = incode;
 297                }
 298        } while (rsize > 0);
 299
 300        if (outpos > 0) {
 301                xtransformer_write(xstate, outbuf, outpos);
 302                IF_DESKTOP(total_written += outpos;)
 303        }
 304
 305        retval = IF_DESKTOP(total_written) + 0;
 306 err:
 307        free(inbuf);
 308        free(outbuf);
 309        free(htab);
 310        free(codetab);
 311        return retval;
 312}
 313