1 /*- 2 * Copyright (c) 2003-2007 Tim Kientzle 3 * Copyright (c) 2011-2012 Michihiro NAKAJIMA 4 * All rights reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR 16 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 17 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 18 * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, 19 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 20 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 21 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 22 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 23 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 24 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 25 */ 26 27 #include "archive_platform.h" 28 __FBSDID("$FreeBSD: head/lib/libarchive/archive_write_set_format_cpio.c 201170 2009-12-29 06:34:23Z kientzle $"); 29 30 #ifdef HAVE_ERRNO_H 31 #include <errno.h> 32 #endif 33 #include <stdio.h> 34 #ifdef HAVE_STDLIB_H 35 #include <stdlib.h> 36 #endif 37 #ifdef HAVE_STRING_H 38 #include <string.h> 39 #endif 40 41 #include "archive.h" 42 #include "archive_entry.h" 43 #include "archive_entry_locale.h" 44 #include "archive_private.h" 45 #include "archive_write_private.h" 46 #include "archive_write_set_format_private.h" 47 48 static ssize_t archive_write_binary_data(struct archive_write *, 49 const void *buff, size_t s); 50 static int archive_write_binary_close(struct archive_write *); 51 static int archive_write_binary_free(struct archive_write *); 52 static int archive_write_binary_finish_entry(struct archive_write *); 53 static int archive_write_binary_header(struct archive_write *, 54 struct archive_entry *); 55 static int archive_write_binary_options(struct archive_write *, 56 const char *, const char *); 57 static int write_header(struct archive_write *, struct archive_entry *); 58 59 struct cpio { 60 uint64_t entry_bytes_remaining; 61 62 int64_t ino_next; 63 64 struct { int64_t old; int new;} *ino_list; 65 size_t ino_list_size; 66 size_t ino_list_next; 67 68 struct archive_string_conv *opt_sconv; 69 struct archive_string_conv *sconv_default; 70 int init_default_conversion; 71 }; 72 73 /* This struct needs to be packed to get the header right */ 74 75 #if defined(__GNUC__) 76 #define PACKED(x) x __attribute__((packed)) 77 #elif defined(_MSC_VER) 78 #define PACKED(x) __pragma(pack(push, 1)) x __pragma(pack(pop)) 79 #else 80 #define PACKED(x) x 81 #endif 82 83 #define HSIZE 26 84 85 PACKED(struct cpio_binary_header { 86 uint16_t h_magic; 87 uint16_t h_dev; 88 uint16_t h_ino; 89 uint16_t h_mode; 90 uint16_t h_uid; 91 uint16_t h_gid; 92 uint16_t h_nlink; 93 uint16_t h_majmin; 94 uint32_t h_mtime; 95 uint16_t h_namesize; 96 uint32_t h_filesize; 97 }); 98 99 /* Back in the day, the 7th Edition cpio.c had this, to 100 * adapt to, as the comment said, "VAX, Interdata, ...": 101 * 102 * union { long l; short s[2]; char c[4]; } U; 103 * #define MKSHORT(v,lv) {U.l=1L;if(U.c[0]) U.l=lv,v[0]=U.s[1],v[1]=U.s[0]; else U.l=lv,v[0]=U.s[0],v[1]=U.s[1];} 104 * long mklong(v) 105 * short v[]; 106 * { 107 * U.l = 1; 108 * if(U.c[0]) 109 * U.s[0] = v[1], U.s[1] = v[0]; 110 * else 111 * U.s[0] = v[0], U.s[1] = v[1]; 112 * return U.l; 113 * } 114 * 115 * Of course, that assumes that all machines have little-endian shorts, 116 * and just adapts the others to the special endianness of the PDP-11. 117 * 118 * Now, we could do this: 119 * 120 * union { uint32_t l; uint16_t s[2]; uint8_t c[4]; } U; 121 * #define PUTI16(v,sv) {U.s[0]=1;if(U.c[0]) v=sv; else U.s[0]=sv,U.c[2]=U.c[1],U.c[3]=U.c[0],v=U.s[1];} 122 * #define PUTI32(v,lv) {char_t Ut;U.l=1;if(U.c[0]) U.l=lv,v[0]=U.s[1],v[1]=U.s[0]; else U.l=lv,Ut=U.c[0],U.c[0]=U.c[1],U.c[1]=Ut,Ut=U.c[2],U.c[2]=U.c[3],U.c[3]=Ut,v[0]=U.s[0],v[1]=U.s[1];} 123 * 124 * ...but it feels a little better to do it like this: 125 */ 126 127 static uint16_t la_swap16(uint16_t in) { 128 union { 129 uint16_t s[2]; 130 uint8_t c[4]; 131 } U; 132 U.s[0] = 1; 133 if (U.c[0]) 134 return in; 135 else { 136 U.s[0] = in; 137 U.c[2] = U.c[1]; 138 U.c[3] = U.c[0]; 139 return U.s[1]; 140 } 141 /* NOTREACHED */ 142 } 143 144 static uint32_t la_swap32(uint32_t in) { 145 union { 146 uint32_t l; 147 uint16_t s[2]; 148 uint8_t c[4]; 149 } U; 150 U.l = 1; 151 if (U.c[0]) { /* Little-endian */ 152 uint16_t t; 153 U.l = in; 154 t = U.s[0]; 155 U.s[0] = U.s[1]; 156 U.s[1] = t; 157 } else if (U.c[3]) { /* Big-endian */ 158 U.l = in; 159 U.s[0] = la_swap16(U.s[0]); 160 U.s[1] = la_swap16(U.s[1]); 161 } else { /* PDP-endian */ 162 U.l = in; 163 } 164 return U.l; 165 } 166 167 /* 168 * Set output format to the selected binary variant 169 */ 170 static int 171 archive_write_set_format_cpio_binary(struct archive *_a, int format) 172 { 173 struct archive_write *a = (struct archive_write *)_a; 174 struct cpio *cpio; 175 176 if (sizeof(struct cpio_binary_header) != HSIZE) { 177 archive_set_error(&a->archive, EINVAL, 178 "Binary cpio format not supported on this platform"); 179 return (ARCHIVE_FATAL); 180 } 181 182 archive_check_magic(_a, ARCHIVE_WRITE_MAGIC, 183 ARCHIVE_STATE_NEW, "archive_write_set_format_cpio_binary"); 184 185 /* If someone else was already registered, unregister them. */ 186 if (a->format_free != NULL) 187 (a->format_free)(a); 188 189 cpio = (struct cpio *)calloc(1, sizeof(*cpio)); 190 if (cpio == NULL) { 191 archive_set_error(&a->archive, ENOMEM, "Can't allocate cpio data"); 192 return (ARCHIVE_FATAL); 193 } 194 a->format_data = cpio; 195 a->format_name = "cpio"; 196 a->format_options = archive_write_binary_options; 197 a->format_write_header = archive_write_binary_header; 198 a->format_write_data = archive_write_binary_data; 199 a->format_finish_entry = archive_write_binary_finish_entry; 200 a->format_close = archive_write_binary_close; 201 a->format_free = archive_write_binary_free; 202 a->archive.archive_format = format; 203 switch (format) { 204 case ARCHIVE_FORMAT_CPIO_PWB: 205 a->archive.archive_format_name = "PWB cpio"; 206 break; 207 case ARCHIVE_FORMAT_CPIO_BIN_LE: 208 a->archive.archive_format_name = "7th Edition cpio"; 209 break; 210 default: 211 archive_set_error(&a->archive, EINVAL, "binary format must be 'pwb' or 'bin'"); 212 return (ARCHIVE_FATAL); 213 } 214 return (ARCHIVE_OK); 215 } 216 217 /* 218 * Set output format to PWB (6th Edition) binary format 219 */ 220 int 221 archive_write_set_format_cpio_pwb(struct archive *_a) 222 { 223 return archive_write_set_format_cpio_binary(_a, ARCHIVE_FORMAT_CPIO_PWB); 224 } 225 226 /* 227 * Set output format to 7th Edition binary format 228 */ 229 int 230 archive_write_set_format_cpio_bin(struct archive *_a) 231 { 232 return archive_write_set_format_cpio_binary(_a, ARCHIVE_FORMAT_CPIO_BIN_LE); 233 } 234 235 static int 236 archive_write_binary_options(struct archive_write *a, const char *key, 237 const char *val) 238 { 239 struct cpio *cpio = (struct cpio *)a->format_data; 240 int ret = ARCHIVE_FAILED; 241 242 if (strcmp(key, "hdrcharset") == 0) { 243 if (val == NULL || val[0] == 0) 244 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 245 "%s: hdrcharset option needs a character-set name", 246 a->format_name); 247 else { 248 cpio->opt_sconv = archive_string_conversion_to_charset( 249 &a->archive, val, 0); 250 if (cpio->opt_sconv != NULL) 251 ret = ARCHIVE_OK; 252 else 253 ret = ARCHIVE_FATAL; 254 } 255 return (ret); 256 } 257 258 /* Note: The "warn" return is just to inform the options 259 * supervisor that we didn't handle it. It will generate 260 * a suitable error if no one used this option. */ 261 return (ARCHIVE_WARN); 262 } 263 264 /* 265 * Ino values are as long as 64 bits on some systems; cpio format 266 * only allows 16 bits and relies on the ino values to identify hardlinked 267 * files. So, we can't merely "hash" the ino numbers since collisions 268 * would corrupt the archive. Instead, we generate synthetic ino values 269 * to store in the archive and maintain a map of original ino values to 270 * synthetic ones so we can preserve hardlink information. 271 * 272 * TODO: Make this more efficient. It's not as bad as it looks (most 273 * files don't have any hardlinks and we don't do any work here for those), 274 * but it wouldn't be hard to do better. 275 * 276 * TODO: Work with dev/ino pairs here instead of just ino values. 277 */ 278 static int 279 synthesize_ino_value(struct cpio *cpio, struct archive_entry *entry) 280 { 281 int64_t ino = archive_entry_ino64(entry); 282 int ino_new; 283 size_t i; 284 285 /* 286 * If no index number was given, don't assign one. In 287 * particular, this handles the end-of-archive marker 288 * correctly by giving it a zero index value. (This is also 289 * why we start our synthetic index numbers with one below.) 290 */ 291 if (ino == 0) 292 return (0); 293 294 /* Don't store a mapping if we don't need to. */ 295 if (archive_entry_nlink(entry) < 2) { 296 return (int)(++cpio->ino_next); 297 } 298 299 /* Look up old ino; if we have it, this is a hardlink 300 * and we reuse the same value. */ 301 for (i = 0; i < cpio->ino_list_next; ++i) { 302 if (cpio->ino_list[i].old == ino) 303 return (cpio->ino_list[i].new); 304 } 305 306 /* Assign a new index number. */ 307 ino_new = (int)(++cpio->ino_next); 308 309 /* Ensure space for the new mapping. */ 310 if (cpio->ino_list_size <= cpio->ino_list_next) { 311 size_t newsize = cpio->ino_list_size < 512 312 ? 512 : cpio->ino_list_size * 2; 313 void *newlist = realloc(cpio->ino_list, 314 sizeof(cpio->ino_list[0]) * newsize); 315 if (newlist == NULL) 316 return (-1); 317 318 cpio->ino_list_size = newsize; 319 cpio->ino_list = newlist; 320 } 321 322 /* Record and return the new value. */ 323 cpio->ino_list[cpio->ino_list_next].old = ino; 324 cpio->ino_list[cpio->ino_list_next].new = ino_new; 325 ++cpio->ino_list_next; 326 return (ino_new); 327 } 328 329 330 static struct archive_string_conv * 331 get_sconv(struct archive_write *a) 332 { 333 struct cpio *cpio; 334 struct archive_string_conv *sconv; 335 336 cpio = (struct cpio *)a->format_data; 337 sconv = cpio->opt_sconv; 338 if (sconv == NULL) { 339 if (!cpio->init_default_conversion) { 340 cpio->sconv_default = 341 archive_string_default_conversion_for_write( 342 &(a->archive)); 343 cpio->init_default_conversion = 1; 344 } 345 sconv = cpio->sconv_default; 346 } 347 return (sconv); 348 } 349 350 static int 351 archive_write_binary_header(struct archive_write *a, struct archive_entry *entry) 352 { 353 const char *path; 354 size_t len; 355 356 if (archive_entry_filetype(entry) == 0 && archive_entry_hardlink(entry) == NULL) { 357 archive_set_error(&a->archive, -1, "Filetype required"); 358 return (ARCHIVE_FAILED); 359 } 360 361 if (archive_entry_pathname_l(entry, &path, &len, get_sconv(a)) != 0 362 && errno == ENOMEM) { 363 archive_set_error(&a->archive, ENOMEM, 364 "Can't allocate memory for Pathname"); 365 return (ARCHIVE_FATAL); 366 } 367 if (len == 0 || path == NULL || path[0] == '\0') { 368 archive_set_error(&a->archive, -1, "Pathname required"); 369 return (ARCHIVE_FAILED); 370 } 371 372 if (!archive_entry_size_is_set(entry) || archive_entry_size(entry) < 0) { 373 archive_set_error(&a->archive, -1, "Size required"); 374 return (ARCHIVE_FAILED); 375 } 376 return write_header(a, entry); 377 } 378 379 static int 380 write_header(struct archive_write *a, struct archive_entry *entry) 381 { 382 struct cpio *cpio; 383 const char *p, *path; 384 int pathlength, ret, ret_final; 385 int64_t ino; 386 struct cpio_binary_header h; 387 struct archive_string_conv *sconv; 388 struct archive_entry *entry_main; 389 size_t len; 390 391 cpio = (struct cpio *)a->format_data; 392 ret_final = ARCHIVE_OK; 393 sconv = get_sconv(a); 394 395 #if defined(_WIN32) && !defined(__CYGWIN__) 396 /* Make sure the path separators in pathname, hardlink and symlink 397 * are all slash '/', not the Windows path separator '\'. */ 398 entry_main = __la_win_entry_in_posix_pathseparator(entry); 399 if (entry_main == NULL) { 400 archive_set_error(&a->archive, ENOMEM, 401 "Can't allocate ustar data"); 402 return(ARCHIVE_FATAL); 403 } 404 if (entry != entry_main) 405 entry = entry_main; 406 else 407 entry_main = NULL; 408 #else 409 entry_main = NULL; 410 #endif 411 412 ret = archive_entry_pathname_l(entry, &path, &len, sconv); 413 if (ret != 0) { 414 if (errno == ENOMEM) { 415 archive_set_error(&a->archive, ENOMEM, 416 "Can't allocate memory for Pathname"); 417 ret_final = ARCHIVE_FATAL; 418 goto exit_write_header; 419 } 420 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, 421 "Can't translate pathname '%s' to %s", 422 archive_entry_pathname(entry), 423 archive_string_conversion_charset_name(sconv)); 424 ret_final = ARCHIVE_WARN; 425 } 426 /* Include trailing null */ 427 pathlength = (int)len + 1; 428 429 h.h_magic = la_swap16(070707); 430 h.h_dev = la_swap16(archive_entry_dev(entry)); 431 432 ino = synthesize_ino_value(cpio, entry); 433 if (ino < 0) { 434 archive_set_error(&a->archive, ENOMEM, 435 "No memory for ino translation table"); 436 ret_final = ARCHIVE_FATAL; 437 goto exit_write_header; 438 } else if (ino > 077777) { 439 archive_set_error(&a->archive, ERANGE, 440 "Too many files for this cpio format"); 441 ret_final = ARCHIVE_FATAL; 442 goto exit_write_header; 443 } 444 h.h_ino = la_swap16((uint16_t)ino); 445 446 h.h_mode = archive_entry_mode(entry); 447 if (((h.h_mode & AE_IFMT) == AE_IFSOCK) || ((h.h_mode & AE_IFMT) == AE_IFIFO)) { 448 archive_set_error(&a->archive, EINVAL, 449 "sockets and fifos cannot be represented in the binary cpio formats"); 450 ret_final = ARCHIVE_FATAL; 451 goto exit_write_header; 452 } 453 if (a->archive.archive_format == ARCHIVE_FORMAT_CPIO_PWB) { 454 if ((h.h_mode & AE_IFMT) == AE_IFLNK) { 455 archive_set_error(&a->archive, EINVAL, 456 "symbolic links cannot be represented in the PWB cpio format"); 457 ret_final = ARCHIVE_FATAL; 458 goto exit_write_header; 459 } 460 /* we could turn off AE_IFREG here, but it does no harm, */ 461 /* and allows v7 cpio to read the entry without confusion */ 462 } 463 h.h_mode = la_swap16(h.h_mode); 464 465 h.h_uid = la_swap16((uint16_t)archive_entry_uid(entry)); 466 h.h_gid = la_swap16((uint16_t)archive_entry_gid(entry)); 467 h.h_nlink = la_swap16((uint16_t)archive_entry_nlink(entry)); 468 469 if (archive_entry_filetype(entry) == AE_IFBLK 470 || archive_entry_filetype(entry) == AE_IFCHR) 471 h.h_majmin = la_swap16(archive_entry_rdev(entry)); 472 else 473 h.h_majmin = 0; 474 475 h.h_mtime = la_swap32((uint32_t)archive_entry_mtime(entry)); 476 h.h_namesize = la_swap16(pathlength); 477 478 /* Non-regular files don't store bodies. */ 479 if (archive_entry_filetype(entry) != AE_IFREG) 480 archive_entry_set_size(entry, 0); 481 482 /* Symlinks get the link written as the body of the entry. */ 483 ret = archive_entry_symlink_l(entry, &p, &len, sconv); 484 if (ret != 0) { 485 if (errno == ENOMEM) { 486 archive_set_error(&a->archive, ENOMEM, 487 "Can't allocate memory for Linkname"); 488 ret_final = ARCHIVE_FATAL; 489 goto exit_write_header; 490 } 491 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, 492 "Can't translate linkname '%s' to %s", 493 archive_entry_symlink(entry), 494 archive_string_conversion_charset_name(sconv)); 495 ret_final = ARCHIVE_WARN; 496 } 497 498 if (len > 0 && p != NULL && *p != '\0') { 499 if (a->archive.archive_format == ARCHIVE_FORMAT_CPIO_PWB) { 500 archive_set_error(&a->archive, EINVAL, 501 "symlinks are not supported by UNIX V6 or by PWB cpio"); 502 ret_final = ARCHIVE_FATAL; 503 goto exit_write_header; 504 } 505 h.h_filesize = la_swap32((uint32_t)strlen(p)); /* symlink */ 506 } else { 507 if ((a->archive.archive_format == ARCHIVE_FORMAT_CPIO_PWB) && 508 (archive_entry_size(entry) > 256*256*256-1)) { 509 archive_set_error(&a->archive, ERANGE, 510 "File is too large for PWB binary cpio format."); 511 ret_final = ARCHIVE_FAILED; 512 goto exit_write_header; 513 } else if (archive_entry_size(entry) > INT32_MAX) { 514 archive_set_error(&a->archive, ERANGE, 515 "File is too large for binary cpio format."); 516 ret_final = ARCHIVE_FAILED; 517 goto exit_write_header; 518 } 519 h.h_filesize = la_swap32((uint32_t)archive_entry_size(entry)); /* file */ 520 } 521 522 ret = __archive_write_output(a, &h, HSIZE); 523 if (ret != ARCHIVE_OK) { 524 ret_final = ARCHIVE_FATAL; 525 goto exit_write_header; 526 } 527 528 ret = __archive_write_output(a, path, pathlength); 529 if ((ret == ARCHIVE_OK) && ((pathlength % 2) != 0)) 530 ret = __archive_write_nulls(a, 1); 531 if (ret != ARCHIVE_OK) { 532 ret_final = ARCHIVE_FATAL; 533 goto exit_write_header; 534 } 535 536 cpio->entry_bytes_remaining = archive_entry_size(entry); 537 if ((cpio->entry_bytes_remaining % 2) != 0) 538 cpio->entry_bytes_remaining++; 539 540 /* Write the symlink now. */ 541 if (p != NULL && *p != '\0') { 542 ret = __archive_write_output(a, p, strlen(p)); 543 if ((ret == ARCHIVE_OK) && ((strlen(p) % 2) != 0)) 544 ret = __archive_write_nulls(a, 1); 545 if (ret != ARCHIVE_OK) { 546 ret_final = ARCHIVE_FATAL; 547 goto exit_write_header; 548 } 549 } 550 551 exit_write_header: 552 archive_entry_free(entry_main); 553 return (ret_final); 554 } 555 556 static ssize_t 557 archive_write_binary_data(struct archive_write *a, const void *buff, size_t s) 558 { 559 struct cpio *cpio; 560 int ret; 561 562 cpio = (struct cpio *)a->format_data; 563 if (s > cpio->entry_bytes_remaining) 564 s = (size_t)cpio->entry_bytes_remaining; 565 566 ret = __archive_write_output(a, buff, s); 567 cpio->entry_bytes_remaining -= s; 568 if (ret >= 0) 569 return (s); 570 else 571 return (ret); 572 } 573 574 static int 575 archive_write_binary_close(struct archive_write *a) 576 { 577 int er; 578 struct archive_entry *trailer; 579 580 trailer = archive_entry_new2(NULL); 581 /* nlink = 1 here for GNU cpio compat. */ 582 archive_entry_set_nlink(trailer, 1); 583 archive_entry_set_size(trailer, 0); 584 archive_entry_set_pathname(trailer, "TRAILER!!!"); 585 er = write_header(a, trailer); 586 archive_entry_free(trailer); 587 return (er); 588 } 589 590 static int 591 archive_write_binary_free(struct archive_write *a) 592 { 593 struct cpio *cpio; 594 595 cpio = (struct cpio *)a->format_data; 596 free(cpio->ino_list); 597 free(cpio); 598 a->format_data = NULL; 599 return (ARCHIVE_OK); 600 } 601 602 static int 603 archive_write_binary_finish_entry(struct archive_write *a) 604 { 605 struct cpio *cpio; 606 607 cpio = (struct cpio *)a->format_data; 608 return (__archive_write_nulls(a, 609 (size_t)cpio->entry_bytes_remaining)); 610 } 611