1/*- 2 * Copyright (c) 2003-2007 Tim Kientzle 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR 15 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 16 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 17 * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, 18 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 19 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 20 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 21 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 22 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 23 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 24 */ 25 26#include "archive_platform.h" 27__FBSDID("$FreeBSD$"); 28 29#ifdef HAVE_ERRNO_H 30#include <errno.h> 31#endif 32#ifdef HAVE_STDLIB_H 33#include <stdlib.h> 34#endif 35#ifdef HAVE_STRING_H 36#include <string.h> 37#endif 38 39#include "archive.h" 40#include "archive_entry.h" 41#include "archive_private.h" 42#include "archive_write_private.h" 43 44struct pax { 45 uint64_t entry_bytes_remaining; 46 uint64_t entry_padding; 47 struct archive_string pax_header; 48}; 49 50static void add_pax_attr(struct archive_string *, const char *key, 51 const char *value); 52static void add_pax_attr_int(struct archive_string *, 53 const char *key, int64_t value); 54static void add_pax_attr_time(struct archive_string *, 55 const char *key, int64_t sec, 56 unsigned long nanos); 57static void add_pax_attr_w(struct archive_string *, 58 const char *key, const wchar_t *wvalue); 59static ssize_t archive_write_pax_data(struct archive_write *, 60 const void *, size_t); 61static int archive_write_pax_finish(struct archive_write *); 62static int archive_write_pax_destroy(struct archive_write *); 63static int archive_write_pax_finish_entry(struct archive_write *); 64static int archive_write_pax_header(struct archive_write *, 65 struct archive_entry *); 66static char *base64_encode(const char *src, size_t len); 67static char *build_pax_attribute_name(char *dest, const char *src); 68static char *build_ustar_entry_name(char *dest, const char *src, 69 size_t src_length, const char *insert); 70static char *format_int(char *dest, int64_t); 71static int has_non_ASCII(const wchar_t *); 72static char *url_encode(const char *in); 73static int write_nulls(struct archive_write *, size_t); 74 75/* 76 * Set output format to 'restricted pax' format. 77 * 78 * This is the same as normal 'pax', but tries to suppress 79 * the pax header whenever possible. This is the default for 80 * bsdtar, for instance. 81 */ 82int 83archive_write_set_format_pax_restricted(struct archive *_a) 84{ 85 struct archive_write *a = (struct archive_write *)_a; 86 int r; 87 r = archive_write_set_format_pax(&a->archive); 88 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_RESTRICTED; 89 a->archive.archive_format_name = "restricted POSIX pax interchange"; 90 return (r); 91} 92 93/* 94 * Set output format to 'pax' format. 95 */ 96int 97archive_write_set_format_pax(struct archive *_a) 98{ 99 struct archive_write *a = (struct archive_write *)_a; 100 struct pax *pax; 101 102 if (a->format_destroy != NULL) 103 (a->format_destroy)(a); 104 105 pax = (struct pax *)malloc(sizeof(*pax)); 106 if (pax == NULL) { 107 archive_set_error(&a->archive, ENOMEM, "Can't allocate pax data"); 108 return (ARCHIVE_FATAL); 109 } 110 memset(pax, 0, sizeof(*pax)); 111 a->format_data = pax; 112 113 a->pad_uncompressed = 1; 114 a->format_name = "pax"; 115 a->format_write_header = archive_write_pax_header; 116 a->format_write_data = archive_write_pax_data; 117 a->format_finish = archive_write_pax_finish; 118 a->format_destroy = archive_write_pax_destroy; 119 a->format_finish_entry = archive_write_pax_finish_entry; 120 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE; 121 a->archive.archive_format_name = "POSIX pax interchange"; 122 return (ARCHIVE_OK); 123} 124 125/* 126 * Note: This code assumes that 'nanos' has the same sign as 'sec', 127 * which implies that sec=-1, nanos=200000000 represents -1.2 seconds 128 * and not -0.8 seconds. This is a pretty pedantic point, as we're 129 * unlikely to encounter many real files created before Jan 1, 1970, 130 * much less ones with timestamps recorded to sub-second resolution. 131 */ 132static void 133add_pax_attr_time(struct archive_string *as, const char *key, 134 int64_t sec, unsigned long nanos) 135{ 136 int digit, i; 137 char *t; 138 /* 139 * Note that each byte contributes fewer than 3 base-10 140 * digits, so this will always be big enough. 141 */ 142 char tmp[1 + 3*sizeof(sec) + 1 + 3*sizeof(nanos)]; 143 144 tmp[sizeof(tmp) - 1] = 0; 145 t = tmp + sizeof(tmp) - 1; 146 147 /* Skip trailing zeros in the fractional part. */ 148 for (digit = 0, i = 10; i > 0 && digit == 0; i--) { 149 digit = nanos % 10; 150 nanos /= 10; 151 } 152 153 /* Only format the fraction if it's non-zero. */ 154 if (i > 0) { 155 while (i > 0) { 156 *--t = "0123456789"[digit]; 157 digit = nanos % 10; 158 nanos /= 10; 159 i--; 160 } 161 *--t = '.'; 162 } 163 t = format_int(t, sec); 164 165 add_pax_attr(as, key, t); 166} 167 168static char * 169format_int(char *t, int64_t i) 170{ 171 int sign; 172 173 if (i < 0) { 174 sign = -1; 175 i = -i; 176 } else 177 sign = 1; 178 179 do { 180 *--t = "0123456789"[i % 10]; 181 } while (i /= 10); 182 if (sign < 0) 183 *--t = '-'; 184 return (t); 185} 186 187static void 188add_pax_attr_int(struct archive_string *as, const char *key, int64_t value) 189{ 190 char tmp[1 + 3 * sizeof(value)]; 191 192 tmp[sizeof(tmp) - 1] = 0; 193 add_pax_attr(as, key, format_int(tmp + sizeof(tmp) - 1, value)); 194} 195 196static char * 197utf8_encode(const wchar_t *wval) 198{ 199 int utf8len; 200 const wchar_t *wp; 201 unsigned long wc; 202 char *utf8_value, *p; 203 204 utf8len = 0; 205 for (wp = wval; *wp != L'\0'; ) { 206 wc = *wp++; 207 208 if (wc >= 0xd800 && wc <= 0xdbff 209 && *wp >= 0xdc00 && *wp <= 0xdfff) { 210 /* This is a surrogate pair. Combine into a 211 * full Unicode value before encoding into 212 * UTF-8. */ 213 wc = (wc - 0xd800) << 10; /* High 10 bits */ 214 wc += (*wp++ - 0xdc00); /* Low 10 bits */ 215 wc += 0x10000; /* Skip BMP */ 216 } 217 if (wc <= 0x7f) 218 utf8len++; 219 else if (wc <= 0x7ff) 220 utf8len += 2; 221 else if (wc <= 0xffff) 222 utf8len += 3; 223 else if (wc <= 0x1fffff) 224 utf8len += 4; 225 else if (wc <= 0x3ffffff) 226 utf8len += 5; 227 else if (wc <= 0x7fffffff) 228 utf8len += 6; 229 /* Ignore larger values; UTF-8 can't encode them. */ 230 } 231 232 utf8_value = (char *)malloc(utf8len + 1); 233 if (utf8_value == NULL) { 234 __archive_errx(1, "Not enough memory for attributes"); 235 return (NULL); 236 } 237 238 for (wp = wval, p = utf8_value; *wp != L'\0'; ) { 239 wc = *wp++; 240 if (wc >= 0xd800 && wc <= 0xdbff 241 && *wp >= 0xdc00 && *wp <= 0xdfff) { 242 /* Combine surrogate pair. */ 243 wc = (wc - 0xd800) << 10; 244 wc += *wp++ - 0xdc00 + 0x10000; 245 } 246 if (wc <= 0x7f) { 247 *p++ = (char)wc; 248 } else if (wc <= 0x7ff) { 249 p[0] = 0xc0 | ((wc >> 6) & 0x1f); 250 p[1] = 0x80 | (wc & 0x3f); 251 p += 2; 252 } else if (wc <= 0xffff) { 253 p[0] = 0xe0 | ((wc >> 12) & 0x0f); 254 p[1] = 0x80 | ((wc >> 6) & 0x3f); 255 p[2] = 0x80 | (wc & 0x3f); 256 p += 3; 257 } else if (wc <= 0x1fffff) { 258 p[0] = 0xf0 | ((wc >> 18) & 0x07); 259 p[1] = 0x80 | ((wc >> 12) & 0x3f); 260 p[2] = 0x80 | ((wc >> 6) & 0x3f); 261 p[3] = 0x80 | (wc & 0x3f); 262 p += 4; 263 } else if (wc <= 0x3ffffff) { 264 p[0] = 0xf8 | ((wc >> 24) & 0x03); 265 p[1] = 0x80 | ((wc >> 18) & 0x3f); 266 p[2] = 0x80 | ((wc >> 12) & 0x3f); 267 p[3] = 0x80 | ((wc >> 6) & 0x3f); 268 p[4] = 0x80 | (wc & 0x3f); 269 p += 5; 270 } else if (wc <= 0x7fffffff) { 271 p[0] = 0xfc | ((wc >> 30) & 0x01); 272 p[1] = 0x80 | ((wc >> 24) & 0x3f); 273 p[1] = 0x80 | ((wc >> 18) & 0x3f); 274 p[2] = 0x80 | ((wc >> 12) & 0x3f); 275 p[3] = 0x80 | ((wc >> 6) & 0x3f); 276 p[4] = 0x80 | (wc & 0x3f); 277 p += 6; 278 } 279 /* Ignore larger values; UTF-8 can't encode them. */ 280 } 281 *p = '\0'; 282 283 return (utf8_value); 284} 285 286static void 287add_pax_attr_w(struct archive_string *as, const char *key, const wchar_t *wval) 288{ 289 char *utf8_value = utf8_encode(wval); 290 if (utf8_value == NULL) 291 return; 292 add_pax_attr(as, key, utf8_value); 293 free(utf8_value); 294} 295 296/* 297 * Add a key/value attribute to the pax header. This function handles 298 * the length field and various other syntactic requirements. 299 */ 300static void 301add_pax_attr(struct archive_string *as, const char *key, const char *value) 302{ 303 int digits, i, len, next_ten; 304 char tmp[1 + 3 * sizeof(int)]; /* < 3 base-10 digits per byte */ 305 306 /*- 307 * PAX attributes have the following layout: 308 * <len> <space> <key> <=> <value> <nl> 309 */ 310 len = 1 + (int)strlen(key) + 1 + (int)strlen(value) + 1; 311 312 /* 313 * The <len> field includes the length of the <len> field, so 314 * computing the correct length is tricky. I start by 315 * counting the number of base-10 digits in 'len' and 316 * computing the next higher power of 10. 317 */ 318 next_ten = 1; 319 digits = 0; 320 i = len; 321 while (i > 0) { 322 i = i / 10; 323 digits++; 324 next_ten = next_ten * 10; 325 } 326 /* 327 * For example, if string without the length field is 99 328 * chars, then adding the 2 digit length "99" will force the 329 * total length past 100, requiring an extra digit. The next 330 * statement adjusts for this effect. 331 */ 332 if (len + digits >= next_ten) 333 digits++; 334 335 /* Now, we have the right length so we can build the line. */ 336 tmp[sizeof(tmp) - 1] = 0; /* Null-terminate the work area. */ 337 archive_strcat(as, format_int(tmp + sizeof(tmp) - 1, len + digits)); 338 archive_strappend_char(as, ' '); 339 archive_strcat(as, key); 340 archive_strappend_char(as, '='); 341 archive_strcat(as, value); 342 archive_strappend_char(as, '\n'); 343} 344 345static void 346archive_write_pax_header_xattrs(struct pax *pax, struct archive_entry *entry) 347{ 348 struct archive_string s; 349 int i = archive_entry_xattr_reset(entry); 350 351 while (i--) { 352 const char *name; 353 const void *value; 354 char *encoded_value; 355 char *url_encoded_name = NULL, *encoded_name = NULL; 356 wchar_t *wcs_name = NULL; 357 size_t size; 358 359 archive_entry_xattr_next(entry, &name, &value, &size); 360 /* Name is URL-encoded, then converted to wchar_t, 361 * then UTF-8 encoded. */ 362 url_encoded_name = url_encode(name); 363 if (url_encoded_name != NULL) { 364 /* Convert narrow-character to wide-character. */ 365 size_t wcs_length = strlen(url_encoded_name); 366 wcs_name = (wchar_t *)malloc((wcs_length + 1) * sizeof(wchar_t)); 367 if (wcs_name == NULL) 368 __archive_errx(1, "No memory for xattr conversion"); 369 mbstowcs(wcs_name, url_encoded_name, wcs_length); 370 wcs_name[wcs_length] = 0; 371 free(url_encoded_name); /* Done with this. */ 372 } 373 if (wcs_name != NULL) { 374 encoded_name = utf8_encode(wcs_name); 375 free(wcs_name); /* Done with wchar_t name. */ 376 } 377 378 encoded_value = base64_encode((const char *)value, size); 379 380 if (encoded_name != NULL && encoded_value != NULL) { 381 archive_string_init(&s); 382 archive_strcpy(&s, "LIBARCHIVE.xattr."); 383 archive_strcat(&s, encoded_name); 384 add_pax_attr(&(pax->pax_header), s.s, encoded_value); 385 archive_string_free(&s); 386 } 387 free(encoded_name); 388 free(encoded_value); 389 } 390} 391 392/* 393 * TODO: Consider adding 'comment' and 'charset' fields to 394 * archive_entry so that clients can specify them. Also, consider 395 * adding generic key/value tags so clients can add arbitrary 396 * key/value data. 397 */ 398static int 399archive_write_pax_header(struct archive_write *a, 400 struct archive_entry *entry_original) 401{ 402 struct archive_entry *entry_main; 403 const char *p; 404 char *t; 405 const wchar_t *wp; 406 const char *suffix; 407 int need_extension, r, ret; 408 struct pax *pax; 409 const char *hdrcharset = NULL; 410 const char *hardlink; 411 const char *path = NULL, *linkpath = NULL; 412 const char *uname = NULL, *gname = NULL; 413 const wchar_t *path_w = NULL, *linkpath_w = NULL; 414 const wchar_t *uname_w = NULL, *gname_w = NULL; 415 416 char paxbuff[512]; 417 char ustarbuff[512]; 418 char ustar_entry_name[256]; 419 char pax_entry_name[256]; 420 421 ret = ARCHIVE_OK; 422 need_extension = 0; 423 pax = (struct pax *)a->format_data; 424 425 hardlink = archive_entry_hardlink(entry_original); 426 427 /* Make sure this is a type of entry that we can handle here */ 428 if (hardlink == NULL) { 429 switch (archive_entry_filetype(entry_original)) { 430 case AE_IFBLK: 431 case AE_IFCHR: 432 case AE_IFIFO: 433 case AE_IFLNK: 434 case AE_IFREG: 435 break; 436 case AE_IFDIR: 437 /* 438 * Ensure a trailing '/'. Modify the original 439 * entry so the client sees the change. 440 */ 441 p = archive_entry_pathname(entry_original); 442 if (p[strlen(p) - 1] != '/') { 443 t = (char *)malloc(strlen(p) + 2); 444 if (t == NULL) { 445 archive_set_error(&a->archive, ENOMEM, 446 "Can't allocate pax data"); 447 return(ARCHIVE_FATAL); 448 } 449 strcpy(t, p); 450 strcat(t, "/"); 451 archive_entry_copy_pathname(entry_original, t); 452 free(t); 453 } 454 break; 455 case AE_IFSOCK: 456 archive_set_error(&a->archive, 457 ARCHIVE_ERRNO_FILE_FORMAT, 458 "tar format cannot archive socket"); 459 return (ARCHIVE_WARN); 460 default: 461 archive_set_error(&a->archive, 462 ARCHIVE_ERRNO_FILE_FORMAT, 463 "tar format cannot archive this (type=0%lo)", 464 (unsigned long)archive_entry_filetype(entry_original)); 465 return (ARCHIVE_WARN); 466 } 467 } 468 469 /* Copy entry so we can modify it as needed. */ 470 entry_main = archive_entry_clone(entry_original); 471 archive_string_empty(&(pax->pax_header)); /* Blank our work area. */ 472 473 /* 474 * First, check the name fields and see if any of them 475 * require binary coding. If any of them does, then all of 476 * them do. 477 */ 478 hdrcharset = NULL; 479 path = archive_entry_pathname(entry_main); 480 path_w = archive_entry_pathname_w(entry_main); 481 if (path != NULL && path_w == NULL) { 482 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, 483 "Can't translate pathname '%s' to UTF-8", path); 484 ret = ARCHIVE_WARN; 485 hdrcharset = "BINARY"; 486 } 487 uname = archive_entry_uname(entry_main); 488 uname_w = archive_entry_uname_w(entry_main); 489 if (uname != NULL && uname_w == NULL) { 490 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, 491 "Can't translate uname '%s' to UTF-8", uname); 492 ret = ARCHIVE_WARN; 493 hdrcharset = "BINARY"; 494 } 495 gname = archive_entry_gname(entry_main); 496 gname_w = archive_entry_gname_w(entry_main); 497 if (gname != NULL && gname_w == NULL) { 498 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, 499 "Can't translate gname '%s' to UTF-8", gname); 500 ret = ARCHIVE_WARN; 501 hdrcharset = "BINARY"; 502 } 503 linkpath = hardlink; 504 if (linkpath != NULL) { 505 linkpath_w = archive_entry_hardlink_w(entry_main); 506 } else { 507 linkpath = archive_entry_symlink(entry_main); 508 if (linkpath != NULL) 509 linkpath_w = archive_entry_symlink_w(entry_main); 510 } 511 if (linkpath != NULL && linkpath_w == NULL) { 512 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, 513 "Can't translate linkpath '%s' to UTF-8", linkpath); 514 ret = ARCHIVE_WARN; 515 hdrcharset = "BINARY"; 516 } 517 518 /* Store the header encoding first, to be nice to readers. */ 519 if (hdrcharset != NULL) 520 add_pax_attr(&(pax->pax_header), "hdrcharset", hdrcharset); 521 522 523 /* 524 * If name is too long, or has non-ASCII characters, add 525 * 'path' to pax extended attrs. (Note that an unconvertible 526 * name must have non-ASCII characters.) 527 */ 528 if (path == NULL) { 529 /* We don't have a narrow version, so we have to store 530 * the wide version. */ 531 add_pax_attr_w(&(pax->pax_header), "path", path_w); 532 archive_entry_set_pathname(entry_main, "@WidePath"); 533 need_extension = 1; 534 } else if (has_non_ASCII(path_w)) { 535 /* We have non-ASCII characters. */ 536 if (path_w == NULL || hdrcharset != NULL) { 537 /* Can't do UTF-8, so store it raw. */ 538 add_pax_attr(&(pax->pax_header), "path", path); 539 } else { 540 /* Store UTF-8 */ 541 add_pax_attr_w(&(pax->pax_header), 542 "path", path_w); 543 } 544 archive_entry_set_pathname(entry_main, 545 build_ustar_entry_name(ustar_entry_name, 546 path, strlen(path), NULL)); 547 need_extension = 1; 548 } else { 549 /* We have an all-ASCII path; we'd like to just store 550 * it in the ustar header if it will fit. Yes, this 551 * duplicates some of the logic in 552 * write_set_format_ustar.c 553 */ 554 if (strlen(path) <= 100) { 555 /* Fits in the old 100-char tar name field. */ 556 } else { 557 /* Find largest suffix that will fit. */ 558 /* Note: strlen() > 100, so strlen() - 100 - 1 >= 0 */ 559 suffix = strchr(path + strlen(path) - 100 - 1, '/'); 560 /* Don't attempt an empty prefix. */ 561 if (suffix == path) 562 suffix = strchr(suffix + 1, '/'); 563 /* We can put it in the ustar header if it's 564 * all ASCII and it's either <= 100 characters 565 * or can be split at a '/' into a prefix <= 566 * 155 chars and a suffix <= 100 chars. (Note 567 * the strchr() above will return NULL exactly 568 * when the path can't be split.) 569 */ 570 if (suffix == NULL /* Suffix > 100 chars. */ 571 || suffix[1] == '\0' /* empty suffix */ 572 || suffix - path > 155) /* Prefix > 155 chars */ 573 { 574 if (path_w == NULL || hdrcharset != NULL) { 575 /* Can't do UTF-8, so store it raw. */ 576 add_pax_attr(&(pax->pax_header), 577 "path", path); 578 } else { 579 /* Store UTF-8 */ 580 add_pax_attr_w(&(pax->pax_header), 581 "path", path_w); 582 } 583 archive_entry_set_pathname(entry_main, 584 build_ustar_entry_name(ustar_entry_name, 585 path, strlen(path), NULL)); 586 need_extension = 1; 587 } 588 } 589 } 590 591 if (linkpath != NULL) { 592 /* If link name is too long or has non-ASCII characters, add 593 * 'linkpath' to pax extended attrs. */ 594 if (strlen(linkpath) > 100 || linkpath_w == NULL 595 || linkpath_w == NULL || has_non_ASCII(linkpath_w)) { 596 if (linkpath_w == NULL || hdrcharset != NULL) 597 /* If the linkpath is not convertible 598 * to wide, or we're encoding in 599 * binary anyway, store it raw. */ 600 add_pax_attr(&(pax->pax_header), 601 "linkpath", linkpath); 602 else 603 /* If the link is long or has a 604 * non-ASCII character, store it as a 605 * pax extended attribute. */ 606 add_pax_attr_w(&(pax->pax_header), 607 "linkpath", linkpath_w); 608 if (strlen(linkpath) > 100) { 609 if (hardlink != NULL) 610 archive_entry_set_hardlink(entry_main, 611 "././@LongHardLink"); 612 else 613 archive_entry_set_symlink(entry_main, 614 "././@LongSymLink"); 615 } 616 need_extension = 1; 617 } 618 } 619 620 /* If file size is too large, add 'size' to pax extended attrs. */ 621 if (archive_entry_size(entry_main) >= (((int64_t)1) << 33)) { 622 add_pax_attr_int(&(pax->pax_header), "size", 623 archive_entry_size(entry_main)); 624 need_extension = 1; 625 } 626 627 /* If numeric GID is too large, add 'gid' to pax extended attrs. */ 628 if ((unsigned int)archive_entry_gid(entry_main) >= (1 << 18)) { 629 add_pax_attr_int(&(pax->pax_header), "gid", 630 archive_entry_gid(entry_main)); 631 need_extension = 1; 632 } 633 634 /* If group name is too large or has non-ASCII characters, add 635 * 'gname' to pax extended attrs. */ 636 if (gname != NULL) { 637 if (strlen(gname) > 31 638 || gname_w == NULL 639 || has_non_ASCII(gname_w)) 640 { 641 if (gname_w == NULL || hdrcharset != NULL) { 642 add_pax_attr(&(pax->pax_header), 643 "gname", gname); 644 } else { 645 add_pax_attr_w(&(pax->pax_header), 646 "gname", gname_w); 647 } 648 need_extension = 1; 649 } 650 } 651 652 /* If numeric UID is too large, add 'uid' to pax extended attrs. */ 653 if ((unsigned int)archive_entry_uid(entry_main) >= (1 << 18)) { 654 add_pax_attr_int(&(pax->pax_header), "uid", 655 archive_entry_uid(entry_main)); 656 need_extension = 1; 657 } 658 659 /* Add 'uname' to pax extended attrs if necessary. */ 660 if (uname != NULL) { 661 if (strlen(uname) > 31 662 || uname_w == NULL 663 || has_non_ASCII(uname_w)) 664 { 665 if (uname_w == NULL || hdrcharset != NULL) { 666 add_pax_attr(&(pax->pax_header), 667 "uname", uname); 668 } else { 669 add_pax_attr_w(&(pax->pax_header), 670 "uname", uname_w); 671 } 672 need_extension = 1; 673 } 674 } 675 676 /* 677 * POSIX/SUSv3 doesn't provide a standard key for large device 678 * numbers. I use the same keys here that Joerg Schilling 679 * used for 'star.' (Which, somewhat confusingly, are called 680 * "devXXX" even though they code "rdev" values.) No doubt, 681 * other implementations use other keys. Note that there's no 682 * reason we can't write the same information into a number of 683 * different keys. 684 * 685 * Of course, this is only needed for block or char device entries. 686 */ 687 if (archive_entry_filetype(entry_main) == AE_IFBLK 688 || archive_entry_filetype(entry_main) == AE_IFCHR) { 689 /* 690 * If rdevmajor is too large, add 'SCHILY.devmajor' to 691 * extended attributes. 692 */ 693 dev_t rdevmajor, rdevminor; 694 rdevmajor = archive_entry_rdevmajor(entry_main); 695 rdevminor = archive_entry_rdevminor(entry_main); 696 if (rdevmajor >= (1 << 18)) { 697 add_pax_attr_int(&(pax->pax_header), "SCHILY.devmajor", 698 rdevmajor); 699 /* 700 * Non-strict formatting below means we don't 701 * have to truncate here. Not truncating improves 702 * the chance that some more modern tar archivers 703 * (such as GNU tar 1.13) can restore the full 704 * value even if they don't understand the pax 705 * extended attributes. See my rant below about 706 * file size fields for additional details. 707 */ 708 /* archive_entry_set_rdevmajor(entry_main, 709 rdevmajor & ((1 << 18) - 1)); */ 710 need_extension = 1; 711 } 712 713 /* 714 * If devminor is too large, add 'SCHILY.devminor' to 715 * extended attributes. 716 */ 717 if (rdevminor >= (1 << 18)) { 718 add_pax_attr_int(&(pax->pax_header), "SCHILY.devminor", 719 rdevminor); 720 /* Truncation is not necessary here, either. */ 721 /* archive_entry_set_rdevminor(entry_main, 722 rdevminor & ((1 << 18) - 1)); */ 723 need_extension = 1; 724 } 725 } 726 727 /* 728 * Technically, the mtime field in the ustar header can 729 * support 33 bits, but many platforms use signed 32-bit time 730 * values. The cutoff of 0x7fffffff here is a compromise. 731 * Yes, this check is duplicated just below; this helps to 732 * avoid writing an mtime attribute just to handle a 733 * high-resolution timestamp in "restricted pax" mode. 734 */ 735 if (!need_extension && 736 ((archive_entry_mtime(entry_main) < 0) 737 || (archive_entry_mtime(entry_main) >= 0x7fffffff))) 738 need_extension = 1; 739 740 /* I use a star-compatible file flag attribute. */ 741 p = archive_entry_fflags_text(entry_main); 742 if (!need_extension && p != NULL && *p != '\0') 743 need_extension = 1; 744 745 /* If there are non-trivial ACL entries, we need an extension. */ 746 if (!need_extension && archive_entry_acl_count(entry_original, 747 ARCHIVE_ENTRY_ACL_TYPE_ACCESS) > 0) 748 need_extension = 1; 749 750 /* If there are non-trivial ACL entries, we need an extension. */ 751 if (!need_extension && archive_entry_acl_count(entry_original, 752 ARCHIVE_ENTRY_ACL_TYPE_DEFAULT) > 0) 753 need_extension = 1; 754 755 /* If there are extended attributes, we need an extension */ 756 if (!need_extension && archive_entry_xattr_count(entry_original) > 0) 757 need_extension = 1; 758 759 /* 760 * The following items are handled differently in "pax 761 * restricted" format. In particular, in "pax restricted" 762 * format they won't be added unless need_extension is 763 * already set (we're already generating an extended header, so 764 * may as well include these). 765 */ 766 if (a->archive.archive_format != ARCHIVE_FORMAT_TAR_PAX_RESTRICTED || 767 need_extension) { 768 769 if (archive_entry_mtime(entry_main) < 0 || 770 archive_entry_mtime(entry_main) >= 0x7fffffff || 771 archive_entry_mtime_nsec(entry_main) != 0) 772 add_pax_attr_time(&(pax->pax_header), "mtime", 773 archive_entry_mtime(entry_main), 774 archive_entry_mtime_nsec(entry_main)); 775 776 if (archive_entry_ctime(entry_main) != 0 || 777 archive_entry_ctime_nsec(entry_main) != 0) 778 add_pax_attr_time(&(pax->pax_header), "ctime", 779 archive_entry_ctime(entry_main), 780 archive_entry_ctime_nsec(entry_main)); 781 782 if (archive_entry_atime(entry_main) != 0 || 783 archive_entry_atime_nsec(entry_main) != 0) 784 add_pax_attr_time(&(pax->pax_header), "atime", 785 archive_entry_atime(entry_main), 786 archive_entry_atime_nsec(entry_main)); 787 788 /* Store birth/creationtime only if it's earlier than mtime */ 789 if (archive_entry_birthtime_is_set(entry_main) && 790 archive_entry_birthtime(entry_main) 791 < archive_entry_mtime(entry_main)) 792 add_pax_attr_time(&(pax->pax_header), 793 "LIBARCHIVE.creationtime", 794 archive_entry_birthtime(entry_main), 795 archive_entry_birthtime_nsec(entry_main)); 796 797 /* I use a star-compatible file flag attribute. */ 798 p = archive_entry_fflags_text(entry_main); 799 if (p != NULL && *p != '\0') 800 add_pax_attr(&(pax->pax_header), "SCHILY.fflags", p); 801 802 /* I use star-compatible ACL attributes. */ 803 wp = archive_entry_acl_text_w(entry_original, 804 ARCHIVE_ENTRY_ACL_TYPE_ACCESS | 805 ARCHIVE_ENTRY_ACL_STYLE_EXTRA_ID); 806 if (wp != NULL && *wp != L'\0') 807 add_pax_attr_w(&(pax->pax_header), 808 "SCHILY.acl.access", wp); 809 wp = archive_entry_acl_text_w(entry_original, 810 ARCHIVE_ENTRY_ACL_TYPE_DEFAULT | 811 ARCHIVE_ENTRY_ACL_STYLE_EXTRA_ID); 812 if (wp != NULL && *wp != L'\0') 813 add_pax_attr_w(&(pax->pax_header), 814 "SCHILY.acl.default", wp); 815 816 /* Include star-compatible metadata info. */ 817 /* Note: "SCHILY.dev{major,minor}" are NOT the 818 * major/minor portions of "SCHILY.dev". */ 819 add_pax_attr_int(&(pax->pax_header), "SCHILY.dev", 820 archive_entry_dev(entry_main)); 821 add_pax_attr_int(&(pax->pax_header), "SCHILY.ino", 822 archive_entry_ino64(entry_main)); 823 add_pax_attr_int(&(pax->pax_header), "SCHILY.nlink", 824 archive_entry_nlink(entry_main)); 825 826 /* Store extended attributes */ 827 archive_write_pax_header_xattrs(pax, entry_original); 828 } 829 830 /* Only regular files have data. */ 831 if (archive_entry_filetype(entry_main) != AE_IFREG) 832 archive_entry_set_size(entry_main, 0); 833 834 /* 835 * Pax-restricted does not store data for hardlinks, in order 836 * to improve compatibility with ustar. 837 */ 838 if (a->archive.archive_format != ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE && 839 hardlink != NULL) 840 archive_entry_set_size(entry_main, 0); 841 842 /* 843 * XXX Full pax interchange format does permit a hardlink 844 * entry to have data associated with it. I'm not supporting 845 * that here because the client expects me to tell them whether 846 * or not this format expects data for hardlinks. If I 847 * don't check here, then every pax archive will end up with 848 * duplicated data for hardlinks. Someday, there may be 849 * need to select this behavior, in which case the following 850 * will need to be revisited. XXX 851 */ 852 if (hardlink != NULL) 853 archive_entry_set_size(entry_main, 0); 854 855 /* Format 'ustar' header for main entry. 856 * 857 * The trouble with file size: If the reader can't understand 858 * the file size, they may not be able to locate the next 859 * entry and the rest of the archive is toast. Pax-compliant 860 * readers are supposed to ignore the file size in the main 861 * header, so the question becomes how to maximize portability 862 * for readers that don't support pax attribute extensions. 863 * For maximum compatibility, I permit numeric extensions in 864 * the main header so that the file size stored will always be 865 * correct, even if it's in a format that only some 866 * implementations understand. The technique used here is: 867 * 868 * a) If possible, follow the standard exactly. This handles 869 * files up to 8 gigabytes minus 1. 870 * 871 * b) If that fails, try octal but omit the field terminator. 872 * That handles files up to 64 gigabytes minus 1. 873 * 874 * c) Otherwise, use base-256 extensions. That handles files 875 * up to 2^63 in this implementation, with the potential to 876 * go up to 2^94. That should hold us for a while. ;-) 877 * 878 * The non-strict formatter uses similar logic for other 879 * numeric fields, though they're less critical. 880 */ 881 __archive_write_format_header_ustar(a, ustarbuff, entry_main, -1, 0); 882 883 /* If we built any extended attributes, write that entry first. */ 884 if (archive_strlen(&(pax->pax_header)) > 0) { 885 struct archive_entry *pax_attr_entry; 886 time_t s; 887 uid_t uid; 888 gid_t gid; 889 mode_t mode; 890 891 pax_attr_entry = archive_entry_new(); 892 p = archive_entry_pathname(entry_main); 893 archive_entry_set_pathname(pax_attr_entry, 894 build_pax_attribute_name(pax_entry_name, p)); 895 archive_entry_set_size(pax_attr_entry, 896 archive_strlen(&(pax->pax_header))); 897 /* Copy uid/gid (but clip to ustar limits). */ 898 uid = archive_entry_uid(entry_main); 899 if ((unsigned int)uid >= 1 << 18) 900 uid = (uid_t)(1 << 18) - 1; 901 archive_entry_set_uid(pax_attr_entry, uid); 902 gid = archive_entry_gid(entry_main); 903 if ((unsigned int)gid >= 1 << 18) 904 gid = (gid_t)(1 << 18) - 1; 905 archive_entry_set_gid(pax_attr_entry, gid); 906 /* Copy mode over (but not setuid/setgid bits) */ 907 mode = archive_entry_mode(entry_main); 908#ifdef S_ISUID 909 mode &= ~S_ISUID; 910#endif 911#ifdef S_ISGID 912 mode &= ~S_ISGID; 913#endif 914#ifdef S_ISVTX 915 mode &= ~S_ISVTX; 916#endif 917 archive_entry_set_mode(pax_attr_entry, mode); 918 919 /* Copy uname/gname. */ 920 archive_entry_set_uname(pax_attr_entry, 921 archive_entry_uname(entry_main)); 922 archive_entry_set_gname(pax_attr_entry, 923 archive_entry_gname(entry_main)); 924 925 /* Copy mtime, but clip to ustar limits. */ 926 s = archive_entry_mtime(entry_main); 927 if (s < 0) { s = 0; } 928 if (s >= 0x7fffffff) { s = 0x7fffffff; } 929 archive_entry_set_mtime(pax_attr_entry, s, 0); 930 931 /* Standard ustar doesn't support atime. */ 932 archive_entry_set_atime(pax_attr_entry, 0, 0); 933 934 /* Standard ustar doesn't support ctime. */ 935 archive_entry_set_ctime(pax_attr_entry, 0, 0); 936 937 r = __archive_write_format_header_ustar(a, paxbuff, 938 pax_attr_entry, 'x', 1); 939 940 archive_entry_free(pax_attr_entry); 941 942 /* Note that the 'x' header shouldn't ever fail to format */ 943 if (r != 0) { 944 const char *msg = "archive_write_pax_header: " 945 "'x' header failed?! This can't happen.\n"; 946 size_t u = write(2, msg, strlen(msg)); 947 (void)u; /* UNUSED */ 948 exit(1); 949 } 950 r = (a->compressor.write)(a, paxbuff, 512); 951 if (r != ARCHIVE_OK) { 952 pax->entry_bytes_remaining = 0; 953 pax->entry_padding = 0; 954 return (ARCHIVE_FATAL); 955 } 956 957 pax->entry_bytes_remaining = archive_strlen(&(pax->pax_header)); 958 pax->entry_padding = 0x1ff & (-(int64_t)pax->entry_bytes_remaining); 959 960 r = (a->compressor.write)(a, pax->pax_header.s, 961 archive_strlen(&(pax->pax_header))); 962 if (r != ARCHIVE_OK) { 963 /* If a write fails, we're pretty much toast. */ 964 return (ARCHIVE_FATAL); 965 } 966 /* Pad out the end of the entry. */ 967 r = write_nulls(a, pax->entry_padding); 968 if (r != ARCHIVE_OK) { 969 /* If a write fails, we're pretty much toast. */ 970 return (ARCHIVE_FATAL); 971 } 972 pax->entry_bytes_remaining = pax->entry_padding = 0; 973 } 974 975 /* Write the header for main entry. */ 976 r = (a->compressor.write)(a, ustarbuff, 512); 977 if (r != ARCHIVE_OK) 978 return (r); 979 980 /* 981 * Inform the client of the on-disk size we're using, so 982 * they can avoid unnecessarily writing a body for something 983 * that we're just going to ignore. 984 */ 985 archive_entry_set_size(entry_original, archive_entry_size(entry_main)); 986 pax->entry_bytes_remaining = archive_entry_size(entry_main); 987 pax->entry_padding = 0x1ff & (-(int64_t)pax->entry_bytes_remaining); 988 archive_entry_free(entry_main); 989 990 return (ret); 991} 992 993/* 994 * We need a valid name for the regular 'ustar' entry. This routine 995 * tries to hack something more-or-less reasonable. 996 * 997 * The approach here tries to preserve leading dir names. We do so by 998 * working with four sections: 999 * 1) "prefix" directory names, 1000 * 2) "suffix" directory names, 1001 * 3) inserted dir name (optional), 1002 * 4) filename. 1003 * 1004 * These sections must satisfy the following requirements: 1005 * * Parts 1 & 2 together form an initial portion of the dir name. 1006 * * Part 3 is specified by the caller. (It should not contain a leading 1007 * or trailing '/'.) 1008 * * Part 4 forms an initial portion of the base filename. 1009 * * The filename must be <= 99 chars to fit the ustar 'name' field. 1010 * * Parts 2, 3, 4 together must be <= 99 chars to fit the ustar 'name' fld. 1011 * * Part 1 must be <= 155 chars to fit the ustar 'prefix' field. 1012 * * If the original name ends in a '/', the new name must also end in a '/' 1013 * * Trailing '/.' sequences may be stripped. 1014 * 1015 * Note: Recall that the ustar format does not store the '/' separating 1016 * parts 1 & 2, but does store the '/' separating parts 2 & 3. 1017 */ 1018static char * 1019build_ustar_entry_name(char *dest, const char *src, size_t src_length, 1020 const char *insert) 1021{ 1022 const char *prefix, *prefix_end; 1023 const char *suffix, *suffix_end; 1024 const char *filename, *filename_end; 1025 char *p; 1026 int need_slash = 0; /* Was there a trailing slash? */ 1027 size_t suffix_length = 99; 1028 size_t insert_length; 1029 1030 /* Length of additional dir element to be added. */ 1031 if (insert == NULL) 1032 insert_length = 0; 1033 else 1034 /* +2 here allows for '/' before and after the insert. */ 1035 insert_length = strlen(insert) + 2; 1036 1037 /* Step 0: Quick bailout in a common case. */ 1038 if (src_length < 100 && insert == NULL) { 1039 strncpy(dest, src, src_length); 1040 dest[src_length] = '\0'; 1041 return (dest); 1042 } 1043 1044 /* Step 1: Locate filename and enforce the length restriction. */ 1045 filename_end = src + src_length; 1046 /* Remove trailing '/' chars and '/.' pairs. */ 1047 for (;;) { 1048 if (filename_end > src && filename_end[-1] == '/') { 1049 filename_end --; 1050 need_slash = 1; /* Remember to restore trailing '/'. */ 1051 continue; 1052 } 1053 if (filename_end > src + 1 && filename_end[-1] == '.' 1054 && filename_end[-2] == '/') { 1055 filename_end -= 2; 1056 need_slash = 1; /* "foo/." will become "foo/" */ 1057 continue; 1058 } 1059 break; 1060 } 1061 if (need_slash) 1062 suffix_length--; 1063 /* Find start of filename. */ 1064 filename = filename_end - 1; 1065 while ((filename > src) && (*filename != '/')) 1066 filename --; 1067 if ((*filename == '/') && (filename < filename_end - 1)) 1068 filename ++; 1069 /* Adjust filename_end so that filename + insert fits in 99 chars. */ 1070 suffix_length -= insert_length; 1071 if (filename_end > filename + suffix_length) 1072 filename_end = filename + suffix_length; 1073 /* Calculate max size for "suffix" section (#3 above). */ 1074 suffix_length -= filename_end - filename; 1075 1076 /* Step 2: Locate the "prefix" section of the dirname, including 1077 * trailing '/'. */ 1078 prefix = src; 1079 prefix_end = prefix + 155; 1080 if (prefix_end > filename) 1081 prefix_end = filename; 1082 while (prefix_end > prefix && *prefix_end != '/') 1083 prefix_end--; 1084 if ((prefix_end < filename) && (*prefix_end == '/')) 1085 prefix_end++; 1086 1087 /* Step 3: Locate the "suffix" section of the dirname, 1088 * including trailing '/'. */ 1089 suffix = prefix_end; 1090 suffix_end = suffix + suffix_length; /* Enforce limit. */ 1091 if (suffix_end > filename) 1092 suffix_end = filename; 1093 if (suffix_end < suffix) 1094 suffix_end = suffix; 1095 while (suffix_end > suffix && *suffix_end != '/') 1096 suffix_end--; 1097 if ((suffix_end < filename) && (*suffix_end == '/')) 1098 suffix_end++; 1099 1100 /* Step 4: Build the new name. */ 1101 /* The OpenBSD strlcpy function is safer, but less portable. */ 1102 /* Rather than maintain two versions, just use the strncpy version. */ 1103 p = dest; 1104 if (prefix_end > prefix) { 1105 strncpy(p, prefix, prefix_end - prefix); 1106 p += prefix_end - prefix; 1107 } 1108 if (suffix_end > suffix) { 1109 strncpy(p, suffix, suffix_end - suffix); 1110 p += suffix_end - suffix; 1111 } 1112 if (insert != NULL) { 1113 /* Note: assume insert does not have leading or trailing '/' */ 1114 strcpy(p, insert); 1115 p += strlen(insert); 1116 *p++ = '/'; 1117 } 1118 strncpy(p, filename, filename_end - filename); 1119 p += filename_end - filename; 1120 if (need_slash) 1121 *p++ = '/'; 1122 *p = '\0'; 1123 1124 return (dest); 1125} 1126 1127/* 1128 * The ustar header for the pax extended attributes must have a 1129 * reasonable name: SUSv3 requires 'dirname'/PaxHeader.'pid'/'filename' 1130 * where 'pid' is the PID of the archiving process. Unfortunately, 1131 * that makes testing a pain since the output varies for each run, 1132 * so I'm sticking with the simpler 'dirname'/PaxHeader/'filename' 1133 * for now. (Someday, I'll make this settable. Then I can use the 1134 * SUS recommendation as default and test harnesses can override it 1135 * to get predictable results.) 1136 * 1137 * Joerg Schilling has argued that this is unnecessary because, in 1138 * practice, if the pax extended attributes get extracted as regular 1139 * files, noone is going to bother reading those attributes to 1140 * manually restore them. Based on this, 'star' uses 1141 * /tmp/PaxHeader/'basename' as the ustar header name. This is a 1142 * tempting argument, in part because it's simpler than the SUSv3 1143 * recommendation, but I'm not entirely convinced. I'm also 1144 * uncomfortable with the fact that "/tmp" is a Unix-ism. 1145 * 1146 * The following routine leverages build_ustar_entry_name() above and 1147 * so is simpler than you might think. It just needs to provide the 1148 * additional path element and handle a few pathological cases). 1149 */ 1150static char * 1151build_pax_attribute_name(char *dest, const char *src) 1152{ 1153 char buff[64]; 1154 const char *p; 1155 1156 /* Handle the null filename case. */ 1157 if (src == NULL || *src == '\0') { 1158 strcpy(dest, "PaxHeader/blank"); 1159 return (dest); 1160 } 1161 1162 /* Prune final '/' and other unwanted final elements. */ 1163 p = src + strlen(src); 1164 for (;;) { 1165 /* Ends in "/", remove the '/' */ 1166 if (p > src && p[-1] == '/') { 1167 --p; 1168 continue; 1169 } 1170 /* Ends in "/.", remove the '.' */ 1171 if (p > src + 1 && p[-1] == '.' 1172 && p[-2] == '/') { 1173 --p; 1174 continue; 1175 } 1176 break; 1177 } 1178 1179 /* Pathological case: After above, there was nothing left. 1180 * This includes "/." "/./." "/.//./." etc. */ 1181 if (p == src) { 1182 strcpy(dest, "/PaxHeader/rootdir"); 1183 return (dest); 1184 } 1185 1186 /* Convert unadorned "." into a suitable filename. */ 1187 if (*src == '.' && p == src + 1) { 1188 strcpy(dest, "PaxHeader/currentdir"); 1189 return (dest); 1190 } 1191 1192 /* 1193 * TODO: Push this string into the 'pax' structure to avoid 1194 * recomputing it every time. That will also open the door 1195 * to having clients override it. 1196 */ 1197#if HAVE_GETPID && 0 /* Disable this for now; see above comment. */ 1198 sprintf(buff, "PaxHeader.%d", getpid()); 1199#else 1200 /* If the platform can't fetch the pid, don't include it. */ 1201 strcpy(buff, "PaxHeader"); 1202#endif 1203 /* General case: build a ustar-compatible name adding "/PaxHeader/". */ 1204 build_ustar_entry_name(dest, src, p - src, buff); 1205 1206 return (dest); 1207} 1208 1209/* Write two null blocks for the end of archive */ 1210static int 1211archive_write_pax_finish(struct archive_write *a) 1212{ 1213 int r; 1214 1215 if (a->compressor.write == NULL) 1216 return (ARCHIVE_OK); 1217 1218 r = write_nulls(a, 512 * 2); 1219 return (r); 1220} 1221 1222static int 1223archive_write_pax_destroy(struct archive_write *a) 1224{ 1225 struct pax *pax; 1226 1227 pax = (struct pax *)a->format_data; 1228 if (pax == NULL) 1229 return (ARCHIVE_OK); 1230 1231 archive_string_free(&pax->pax_header); 1232 free(pax); 1233 a->format_data = NULL; 1234 return (ARCHIVE_OK); 1235} 1236 1237static int 1238archive_write_pax_finish_entry(struct archive_write *a) 1239{ 1240 struct pax *pax; 1241 int ret; 1242 1243 pax = (struct pax *)a->format_data; 1244 ret = write_nulls(a, pax->entry_bytes_remaining + pax->entry_padding); 1245 pax->entry_bytes_remaining = pax->entry_padding = 0; 1246 return (ret); 1247} 1248 1249static int 1250write_nulls(struct archive_write *a, size_t padding) 1251{ 1252 int ret; 1253 size_t to_write; 1254 1255 while (padding > 0) { 1256 to_write = padding < a->null_length ? padding : a->null_length; 1257 ret = (a->compressor.write)(a, a->nulls, to_write); 1258 if (ret != ARCHIVE_OK) 1259 return (ret); 1260 padding -= to_write; 1261 } 1262 return (ARCHIVE_OK); 1263} 1264 1265static ssize_t 1266archive_write_pax_data(struct archive_write *a, const void *buff, size_t s) 1267{ 1268 struct pax *pax; 1269 int ret; 1270 1271 pax = (struct pax *)a->format_data; 1272 if (s > pax->entry_bytes_remaining) 1273 s = pax->entry_bytes_remaining; 1274 1275 ret = (a->compressor.write)(a, buff, s); 1276 pax->entry_bytes_remaining -= s; 1277 if (ret == ARCHIVE_OK) 1278 return (s); 1279 else 1280 return (ret); 1281} 1282 1283static int 1284has_non_ASCII(const wchar_t *wp) 1285{ 1286 if (wp == NULL) 1287 return (1); 1288 while (*wp != L'\0' && *wp < 128) 1289 wp++; 1290 return (*wp != L'\0'); 1291} 1292 1293/* 1294 * Used by extended attribute support; encodes the name 1295 * so that there will be no '=' characters in the result. 1296 */ 1297static char * 1298url_encode(const char *in) 1299{ 1300 const char *s; 1301 char *d; 1302 int out_len = 0; 1303 char *out; 1304 1305 for (s = in; *s != '\0'; s++) { 1306 if (*s < 33 || *s > 126 || *s == '%' || *s == '=') 1307 out_len += 3; 1308 else 1309 out_len++; 1310 } 1311 1312 out = (char *)malloc(out_len + 1); 1313 if (out == NULL) 1314 return (NULL); 1315 1316 for (s = in, d = out; *s != '\0'; s++) { 1317 /* encode any non-printable ASCII character or '%' or '=' */ 1318 if (*s < 33 || *s > 126 || *s == '%' || *s == '=') { 1319 /* URL encoding is '%' followed by two hex digits */ 1320 *d++ = '%'; 1321 *d++ = "0123456789ABCDEF"[0x0f & (*s >> 4)]; 1322 *d++ = "0123456789ABCDEF"[0x0f & *s]; 1323 } else { 1324 *d++ = *s; 1325 } 1326 } 1327 *d = '\0'; 1328 return (out); 1329} 1330 1331/* 1332 * Encode a sequence of bytes into a C string using base-64 encoding. 1333 * 1334 * Returns a null-terminated C string allocated with malloc(); caller 1335 * is responsible for freeing the result. 1336 */ 1337static char * 1338base64_encode(const char *s, size_t len) 1339{ 1340 static const char digits[64] = 1341 { 'A','B','C','D','E','F','G','H','I','J','K','L','M','N','O', 1342 'P','Q','R','S','T','U','V','W','X','Y','Z','a','b','c','d', 1343 'e','f','g','h','i','j','k','l','m','n','o','p','q','r','s', 1344 't','u','v','w','x','y','z','0','1','2','3','4','5','6','7', 1345 '8','9','+','/' }; 1346 int v; 1347 char *d, *out; 1348 1349 /* 3 bytes becomes 4 chars, but round up and allow for trailing NUL */ 1350 out = (char *)malloc((len * 4 + 2) / 3 + 1); 1351 if (out == NULL) 1352 return (NULL); 1353 d = out; 1354 1355 /* Convert each group of 3 bytes into 4 characters. */ 1356 while (len >= 3) { 1357 v = (((int)s[0] << 16) & 0xff0000) 1358 | (((int)s[1] << 8) & 0xff00) 1359 | (((int)s[2]) & 0x00ff); 1360 s += 3; 1361 len -= 3; 1362 *d++ = digits[(v >> 18) & 0x3f]; 1363 *d++ = digits[(v >> 12) & 0x3f]; 1364 *d++ = digits[(v >> 6) & 0x3f]; 1365 *d++ = digits[(v) & 0x3f]; 1366 } 1367 /* Handle final group of 1 byte (2 chars) or 2 bytes (3 chars). */ 1368 switch (len) { 1369 case 0: break; 1370 case 1: 1371 v = (((int)s[0] << 16) & 0xff0000); 1372 *d++ = digits[(v >> 18) & 0x3f]; 1373 *d++ = digits[(v >> 12) & 0x3f]; 1374 break; 1375 case 2: 1376 v = (((int)s[0] << 16) & 0xff0000) 1377 | (((int)s[1] << 8) & 0xff00); 1378 *d++ = digits[(v >> 18) & 0x3f]; 1379 *d++ = digits[(v >> 12) & 0x3f]; 1380 *d++ = digits[(v >> 6) & 0x3f]; 1381 break; 1382 } 1383 /* Add trailing NUL character so output is a valid C string. */ 1384 *d = '\0'; 1385 return (out); 1386} 1387