src/xzdec/xzdec.c

   1 ///////////////////////////////////////////////////////////////////////////////
   2 //
   3 /// \file       xzdec.c
   4 /// \brief      Simple single-threaded tool to uncompress .xz or .lzma files
   5 //
   6 //  Copyright (C) 2007 Lasse Collin
   7 //
   8 //  This program is free software; you can redistribute it and/or
   9 //  modify it under the terms of the GNU Lesser General Public
  10 //  License as published by the Free Software Foundation; either
  11 //  version 2.1 of the License, or (at your option) any later version.
  12 //
  13 //  This program is distributed in the hope that it will be useful,
  14 //  but WITHOUT ANY WARRANTY; without even the implied warranty of
  15 //  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  16 //  Lesser General Public License for more details.
  17 //
  18 ///////////////////////////////////////////////////////////////////////////////
  19
  20 #include "sysdefs.h"
  21 #include "lzma.h"
  22
  23 #include <errno.h>
  24 #include <stdio.h>
  25 #include <unistd.h>
  26
  27 #ifdef _WIN32
  28 #       include <fcntl.h>
  29 #endif
  30
  31 #include "getopt.h"
  32 #include "physmem.h"
  33
  34
  35 #ifdef LZMADEC
  36 #       define TOOL_FORMAT "lzma"
  37 #else
  38 #       define TOOL_FORMAT "xz"
  39 #endif
  40
  41
  42 /// Number of bytes to use memory at maximum
  43 static uint64_t memlimit;
  44
  45 /// Program name to be shown in error messages
  46 static const char *argv0;
  47
  48
  49 static void lzma_attribute((noreturn))
  50 my_exit(void)
  51 {
  52         int status = EXIT_SUCCESS;
  53
  54         // Close stdout. We don't care about stderr, because we write to it
  55         // only when an error has already occurred.
  56         const int ferror_err = ferror(stdout);
  57         const int fclose_err = fclose(stdout);
  58
  59         if (ferror_err || fclose_err) {
  60                 // If it was fclose() that failed, we have the reason
  61                 // in errno. If only ferror() indicated an error,
  62                 // we have no idea what the reason was.
  63                 fprintf(stderr, "%s: Cannot write to standard output: %s\n",
  64                                 argv0, fclose_err
  65                                         ? strerror(errno) : "Unknown error");
  66                 status = EXIT_FAILURE;
  67         }
  68
  69         exit(status);
  70 }
  71
  72
  73 static void lzma_attribute((noreturn))
  74 help(void)
  75 {
  76         printf(
  77 "Usage: %s [OPTION]... [FILE]...\n"
  78 "Uncompress files in the ." TOOL_FORMAT " format to the standard output.\n"
  79 "\n"
  80 "  -c, --stdout       (ignored)\n"
  81 "  -d, --decompress   (ignored)\n"
  82 "  -k, --keep         (ignored)\n"
  83 "  -f, --force        (ignored)\n"
  84 "  -M, --memory=NUM   use NUM bytes of memory at maximum (0 means default);\n"
  85 "                     the suffixes k, M, G, Ki, Mi, and Gi are supported.\n"
  86 "  -h, --help         display this help and exit\n"
  87 "  -V, --version      display version and license information and exit\n"
  88 "\n"
  89 "With no FILE, or when FILE is -, read standard input.\n"
  90 "\n"
  91 "On this configuration, the tool will use about %" PRIu64
  92                 " MiB of memory at maximum.\n"
  93 "\n"
  94 "Report bugs to <" PACKAGE_BUGREPORT "> (in English or Finnish).\n",
  95                 argv0, (memlimit + 512 * 1024) / (1024 * 1024));
  96         my_exit();
  97 }
  98
  99
 100 static void lzma_attribute((noreturn))
 101 version(void)
 102 {
 103         printf(TOOL_FORMAT "dec " PACKAGE_VERSION "\n"
 104                         "liblzma %s\n", lzma_version_string());
 105
 106         my_exit();
 107 }
 108
 109
 110 /// Finds out the amount of physical memory in the system, and sets
 111 /// a default memory usage limit.
 112 static void
 113 set_default_memlimit(void)
 114 {
 115         const uint64_t mem = physmem();
 116
 117         if (mem == 0)
 118                 // Cannot autodetect, use 10 MiB as the default limit.
 119                 memlimit = (1U << 23) + (1U << 21);
 120         else
 121                 // Limit is 33 % of RAM.
 122                 memlimit = mem / 3;
 123
 124         return;
 125 }
 126
 127
 128 /// \brief      Converts a string to uint64_t
 129 ///
 130 /// This is rudely copied from src/xz/util.c and modified a little. :-(
 131 ///
 132 static uint64_t
 133 str_to_uint64(const char *value)
 134 {
 135         uint64_t result = 0;
 136
 137         if (*value < '0' || *value > '9') {
 138                 fprintf(stderr, "%s: %s: Not a number\n", argv0, value);
 139                 exit(EXIT_FAILURE);
 140         }
 141
 142         do {
 143                 // Don't overflow.
 144                 if (result > (UINT64_MAX - 9) / 10)
 145                         return UINT64_MAX;
 146
 147                 result *= 10;
 148                 result += *value - '0';
 149                 ++value;
 150         } while (*value >= '0' && *value <= '9');
 151
 152         if (*value != '\0') {
 153                 // Look for suffix.
 154                 static const struct {
 155                         const char name[4];
 156                         uint32_t multiplier;
 157                 } suffixes[] = {
 158                         { "k",   1000 },
 159                         { "kB",  1000 },
 160                         { "M",   1000000 },
 161                         { "MB",  1000000 },
 162                         { "G",   1000000000 },
 163                         { "GB",  1000000000 },
 164                         { "Ki",  1024 },
 165                         { "KiB", 1024 },
 166                         { "Mi",  1048576 },
 167                         { "MiB", 1048576 },
 168                         { "Gi",  1073741824 },
 169                         { "GiB", 1073741824 }
 170                 };
 171
 172                 uint32_t multiplier = 0;
 173                 for (size_t i = 0; i < ARRAY_SIZE(suffixes); ++i) {
 174                         if (strcmp(value, suffixes[i].name) == 0) {
 175                                 multiplier = suffixes[i].multiplier;
 176                                 break;
 177                         }
 178                 }
 179
 180                 if (multiplier == 0) {
 181                         fprintf(stderr, "%s: %s: Invalid suffix\n",
 182                                         argv0, value);
 183                         exit(EXIT_FAILURE);
 184                 }
 185
 186                 // Don't overflow here either.
 187                 if (result > UINT64_MAX / multiplier)
 188                         result = UINT64_MAX;
 189                 else
 190                         result *= multiplier;
 191         }
 192
 193         return result;
 194 }
 195
 196
 197 /// Parses command line options.
 198 static void
 199 parse_options(int argc, char **argv)
 200 {
 201         static const char short_opts[] = "cdkfM:hV";
 202         static const struct option long_opts[] = {
 203                 { "stdout",       no_argument,         NULL, 'c' },
 204                 { "to-stdout",    no_argument,         NULL, 'c' },
 205                 { "decompress",   no_argument,         NULL, 'd' },
 206                 { "uncompress",   no_argument,         NULL, 'd' },
 207                 { "force",        no_argument,         NULL, 'f' },
 208                 { "keep",         no_argument,         NULL, 'k' },
 209                 { "memory",       required_argument,   NULL, 'M' },
 210                 { "help",         no_argument,         NULL, 'h' },
 211                 { "version",      no_argument,         NULL, 'V' },
 212                 { NULL,           0,                   NULL, 0   }
 213         };
 214
 215         int c;
 216
 217         while ((c = getopt_long(argc, argv, short_opts, long_opts, NULL))
 218                         != -1) {
 219                 switch (c) {
 220                 case 'c':
 221                 case 'd':
 222                 case 'f':
 223                 case 'k':
 224                         break;
 225
 226                 case 'M':
 227                         memlimit = str_to_uint64(optarg);
 228                         if (memlimit == 0)
 229                                 set_default_memlimit();
 230
 231                         break;
 232
 233                 case 'h':
 234                         help();
 235
 236                 case 'V':
 237                         version();
 238
 239                 default:
 240                         exit(EXIT_FAILURE);
 241                 }
 242         }
 243
 244         return;
 245 }
 246
 247
 248 static void
 249 uncompress(lzma_stream *strm, FILE *file, const char *filename)
 250 {
 251         lzma_ret ret;
 252
 253         // Initialize the decoder
 254 #ifdef LZMADEC
 255         ret = lzma_alone_decoder(strm, memlimit);
 256 #else
 257         ret = lzma_stream_decoder(strm, memlimit, LZMA_CONCATENATED);
 258 #endif
 259
 260         // The only reasonable error here is LZMA_MEM_ERROR.
 261         // FIXME: Maybe also LZMA_MEMLIMIT_ERROR in future?
 262         if (ret != LZMA_OK) {
 263                 fprintf(stderr, "%s: ", argv0);
 264
 265                 if (ret == LZMA_MEM_ERROR)
 266                         fprintf(stderr, "%s\n", strerror(ENOMEM));
 267                 else
 268                         fprintf(stderr, "Internal program error (bug)\n");
 269
 270                 exit(EXIT_FAILURE);
 271         }
 272
 273         // Input and output buffers
 274         uint8_t in_buf[BUFSIZ];
 275         uint8_t out_buf[BUFSIZ];
 276
 277         strm->avail_in = 0;
 278         strm->next_out = out_buf;
 279         strm->avail_out = BUFSIZ;
 280
 281         lzma_action action = LZMA_RUN;
 282
 283         while (true) {
 284                 if (strm->avail_in == 0) {
 285                         strm->next_in = in_buf;
 286                         strm->avail_in = fread(in_buf, 1, BUFSIZ, file);
 287
 288                         if (ferror(file)) {
 289                                 // POSIX says that fread() sets errno if
 290                                 // an error occurred. ferror() doesn't
 291                                 // touch errno.
 292                                 fprintf(stderr, "%s: %s: Error reading "
 293                                                 "input file: %s\n",
 294                                                 argv0, filename,
 295                                                 strerror(errno));
 296                                 exit(EXIT_FAILURE);
 297                         }
 298
 299 #ifndef LZMADEC
 300                         // When using LZMA_CONCATENATED, we need to tell
 301                         // liblzma when it has got all the input.
 302                         if (feof(file))
 303                                 action = LZMA_FINISH;
 304 #endif
 305                 }
 306
 307                 ret = lzma_code(strm, action);
 308
 309                 // Write and check write error before checking decoder error.
 310                 // This way as much data as possible gets written to output
 311                 // even if decoder detected an error.
 312                 if (strm->avail_out == 0 || ret != LZMA_OK) {
 313                         const size_t write_size = BUFSIZ - strm->avail_out;
 314
 315                         if (fwrite(out_buf, 1, write_size, stdout)
 316                                         != write_size) {
 317                                 // Wouldn't be a surprise if writing to stderr
 318                                 // would fail too but at least try to show an
 319                                 // error message.
 320                                 fprintf(stderr, "%s: Cannot write to "
 321                                                 "standard output: %s\n", argv0,
 322                                                 strerror(errno));
 323                                 exit(EXIT_FAILURE);
 324                         }
 325
 326                         strm->next_out = out_buf;
 327                         strm->avail_out = BUFSIZ;
 328                 }
 329
 330                 if (ret != LZMA_OK) {
 331                         if (ret == LZMA_STREAM_END) {
 332 #ifdef LZMADEC
 333                                 // Check that there's no trailing garbage.
 334                                 if (strm->avail_in != 0
 335                                                 || fread(in_buf, 1, 1, file)
 336                                                         != 0
 337                                                 || !feof(file))
 338                                         ret = LZMA_DATA_ERROR;
 339                                 else
 340                                         return;
 341 #else
 342                                 // lzma_stream_decoder() already guarantees
 343                                 // that there's no trailing garbage.
 344                                 assert(strm->avail_in == 0);
 345                                 assert(action == LZMA_FINISH);
 346                                 assert(feof(file));
 347                                 return;
 348 #endif
 349                         }
 350
 351                         const char *msg;
 352                         switch (ret) {
 353                         case LZMA_MEM_ERROR:
 354                                 msg = strerror(ENOMEM);
 355                                 break;
 356
 357                         case LZMA_MEMLIMIT_ERROR:
 358                                 msg = "Memory usage limit reached";
 359                                 break;
 360
 361                         case LZMA_FORMAT_ERROR:
 362                                 msg = "File format not recognized";
 363                                 break;
 364
 365                         case LZMA_OPTIONS_ERROR:
 366                                 // FIXME: Better message?
 367                                 msg = "Unsupported compression options";
 368                                 break;
 369
 370                         case LZMA_DATA_ERROR:
 371                                 msg = "File is corrupt";
 372                                 break;
 373
 374                         case LZMA_BUF_ERROR:
 375                                 msg = "Unexpected end of input";
 376                                 break;
 377
 378                         default:
 379                                 msg = "Internal program error (bug)";
 380                                 break;
 381                         }
 382
 383                         fprintf(stderr, "%s: %s: %s\n", argv0, filename, msg);
 384
 385                         exit(EXIT_FAILURE);
 386                 }
 387         }
 388 }
 389
 390
 391 int
 392 main(int argc, char **argv)
 393 {
 394         // Set the argv0 global so that we can print the command name in
 395         // error and help messages.
 396         argv0 = argv[0];
 397
 398         // Detect amount of installed RAM and set the memory usage limit.
 399         // This is needed before parsing the command line arguments.
 400         set_default_memlimit();
 401
 402         // Parse the command line options.
 403         parse_options(argc, argv);
 404
 405         // The same lzma_stream is used for all files that we decode. This way
 406         // we don't need to reallocate memory for every file if they use same
 407         // compression settings.
 408         lzma_stream strm = LZMA_STREAM_INIT;
 409
 410         // Some systems require setting stdin and stdout to binary mode.
 411 #ifdef _WIN32
 412         setmode(fileno(stdin), O_BINARY);
 413         setmode(fileno(stdout), O_BINARY);
 414 #endif
 415
 416         if (optind == argc) {
 417                 // No filenames given, decode from stdin.
 418                 uncompress(&strm, stdin, "(stdin)");
 419         } else {
 420                 // Loop through the filenames given on the command line.
 421                 do {
 422                         // "-" indicates stdin.
 423                         if (strcmp(argv[optind], "-") == 0) {
 424                                 uncompress(&strm, stdin, "(stdin)");
 425                         } else {
 426                                 FILE *file = fopen(argv[optind], "rb");
 427                                 if (file == NULL) {
 428                                         fprintf(stderr, "%s: %s: %s\n",
 429                                                         argv0, argv[optind],
 430                                                         strerror(errno));
 431                                         exit(EXIT_FAILURE);
 432                                 }
 433
 434                                 uncompress(&strm, file, argv[optind]);
 435                                 fclose(file);
 436                         }
 437                 } while (++optind < argc);
 438         }
 439
 440 #ifndef NDEBUG
 441         // Free the memory only when debugging. Freeing wastes some time,
 442         // but allows detecting possible memory leaks with Valgrind.
 443         lzma_end(&strm);
 444 #endif
 445
 446         my_exit();
 447 }