contrib/libarchive-2/libarchive/archive_read_support_format_tar.c

   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: src/lib/libarchive/archive_read_support_format_tar.c,v 1.65 2008/01/31 07:41:45 kientzle Exp $");
  28
  29 #ifdef HAVE_ERRNO_H
  30 #include <errno.h>
  31 #endif
  32 #include <stddef.h>
  33 /* #include <stdint.h> */ /* See archive_platform.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 /* Obtain suitable wide-character manipulation functions. */
  42 #ifdef HAVE_WCHAR_H
  43 #include <wchar.h>
  44 #else
  45 /* Good enough for equality testing, which is all we need. */
  46 static int wcscmp(const wchar_t *s1, const wchar_t *s2)
  47 {
  48         int diff = *s1 - *s2;
  49         while (*s1 && diff == 0)
  50                 diff = (int)*++s1 - (int)*++s2;
  51         return diff;
  52 }
  53 /* Good enough for equality testing, which is all we need. */
  54 static int wcsncmp(const wchar_t *s1, const wchar_t *s2, size_t n)
  55 {
  56         int diff = *s1 - *s2;
  57         while (*s1 && diff == 0 && n-- > 0)
  58                 diff = (int)*++s1 - (int)*++s2;
  59         return diff;
  60 }
  61 static size_t wcslen(const wchar_t *s)
  62 {
  63         const wchar_t *p = s;
  64         while (*p)
  65                 p++;
  66         return p - s;
  67 }
  68 #endif
  69
  70 #include "archive.h"
  71 #include "archive_entry.h"
  72 #include "archive_private.h"
  73 #include "archive_read_private.h"
  74
  75 #define tar_min(a,b) ((a) < (b) ? (a) : (b))
  76
  77 /*
  78  * Layout of POSIX 'ustar' tar header.
  79  */
  80 struct archive_entry_header_ustar {
  81         char    name[100];
  82         char    mode[8];
  83         char    uid[8];
  84         char    gid[8];
  85         char    size[12];
  86         char    mtime[12];
  87         char    checksum[8];
  88         char    typeflag[1];
  89         char    linkname[100];  /* "old format" header ends here */
  90         char    magic[6];       /* For POSIX: "ustar\0" */
  91         char    version[2];     /* For POSIX: "00" */
  92         char    uname[32];
  93         char    gname[32];
  94         char    rdevmajor[8];
  95         char    rdevminor[8];
  96         char    prefix[155];
  97 };
  98
  99 /*
 100  * Structure of GNU tar header
 101  */
 102 struct gnu_sparse {
 103         char    offset[12];
 104         char    numbytes[12];
 105 };
 106
 107 struct archive_entry_header_gnutar {
 108         char    name[100];
 109         char    mode[8];
 110         char    uid[8];
 111         char    gid[8];
 112         char    size[12];
 113         char    mtime[12];
 114         char    checksum[8];
 115         char    typeflag[1];
 116         char    linkname[100];
 117         char    magic[8];  /* "ustar  \0" (note blank/blank/null at end) */
 118         char    uname[32];
 119         char    gname[32];
 120         char    rdevmajor[8];
 121         char    rdevminor[8];
 122         char    atime[12];
 123         char    ctime[12];
 124         char    offset[12];
 125         char    longnames[4];
 126         char    unused[1];
 127         struct gnu_sparse sparse[4];
 128         char    isextended[1];
 129         char    realsize[12];
 130         /*
 131          * Old GNU format doesn't use POSIX 'prefix' field; they use
 132          * the 'L' (longname) entry instead.
 133          */
 134 };
 135
 136 /*
 137  * Data specific to this format.
 138  */
 139 struct sparse_block {
 140         struct sparse_block     *next;
 141         off_t   offset;
 142         off_t   remaining;
 143 };
 144
 145 struct tar {
 146         struct archive_string    acl_text;
 147         struct archive_string    entry_pathname;
 148         struct archive_string    entry_linkpath;
 149         struct archive_string    entry_uname;
 150         struct archive_string    entry_gname;
 151         struct archive_string    longlink;
 152         struct archive_string    longname;
 153         struct archive_string    pax_header;
 154         struct archive_string    pax_global;
 155         struct archive_string    line;
 156         int                      pax_hdrcharset_binary;
 157         wchar_t                 *pax_entry;
 158         size_t                   pax_entry_length;
 159         int                      header_recursion_depth;
 160         off_t                    entry_bytes_remaining;
 161         off_t                    entry_offset;
 162         off_t                    entry_padding;
 163         off_t                    realsize;
 164         struct sparse_block     *sparse_list;
 165         struct sparse_block     *sparse_last;
 166         int64_t                  sparse_offset;
 167         int64_t                  sparse_numbytes;
 168         int                      sparse_gnu_major;
 169         int                      sparse_gnu_minor;
 170         char                     sparse_gnu_pending;
 171 };
 172
 173 static ssize_t  UTF8_mbrtowc(wchar_t *pwc, const char *s, size_t n);
 174 static int      archive_block_is_null(const unsigned char *p);
 175 static char     *base64_decode(const char *, size_t, size_t *);
 176 static void      gnu_add_sparse_entry(struct tar *,
 177                     off_t offset, off_t remaining);
 178 static void     gnu_clear_sparse_list(struct tar *);
 179 static int      gnu_sparse_old_read(struct archive_read *, struct tar *,
 180                     const struct archive_entry_header_gnutar *header);
 181 static void     gnu_sparse_old_parse(struct tar *,
 182                     const struct gnu_sparse *sparse, int length);
 183 static int      gnu_sparse_01_parse(struct tar *, const char *);
 184 static ssize_t  gnu_sparse_10_read(struct archive_read *, struct tar *);
 185 static int      header_Solaris_ACL(struct archive_read *,  struct tar *,
 186                     struct archive_entry *, const void *);
 187 static int      header_common(struct archive_read *,  struct tar *,
 188                     struct archive_entry *, const void *);
 189 static int      header_old_tar(struct archive_read *, struct tar *,
 190                     struct archive_entry *, const void *);
 191 static int      header_pax_extensions(struct archive_read *, struct tar *,
 192                     struct archive_entry *, const void *);
 193 static int      header_pax_global(struct archive_read *, struct tar *,
 194                     struct archive_entry *, const void *h);
 195 static int      header_longlink(struct archive_read *, struct tar *,
 196                     struct archive_entry *, const void *h);
 197 static int      header_longname(struct archive_read *, struct tar *,
 198                     struct archive_entry *, const void *h);
 199 static int      header_volume(struct archive_read *, struct tar *,
 200                     struct archive_entry *, const void *h);
 201 static int      header_ustar(struct archive_read *, struct tar *,
 202                     struct archive_entry *, const void *h);
 203 static int      header_gnutar(struct archive_read *, struct tar *,
 204                     struct archive_entry *, const void *h);
 205 static int      archive_read_format_tar_bid(struct archive_read *);
 206 static int      archive_read_format_tar_cleanup(struct archive_read *);
 207 static int      archive_read_format_tar_read_data(struct archive_read *a,
 208                     const void **buff, size_t *size, off_t *offset);
 209 static int      archive_read_format_tar_skip(struct archive_read *a);
 210 static int      archive_read_format_tar_read_header(struct archive_read *,
 211                     struct archive_entry *);
 212 static int      checksum(struct archive_read *, const void *);
 213 static int      pax_attribute(struct tar *, struct archive_entry *,
 214                     char *key, char *value);
 215 static int      pax_header(struct archive_read *, struct tar *,
 216                     struct archive_entry *, char *attr);
 217 static void     pax_time(const char *, int64_t *sec, long *nanos);
 218 static ssize_t  readline(struct archive_read *, struct tar *, const char **,
 219                     ssize_t limit);
 220 static int      read_body_to_string(struct archive_read *, struct tar *,
 221                     struct archive_string *, const void *h);
 222 static int64_t  tar_atol(const char *, unsigned);
 223 static int64_t  tar_atol10(const char *, unsigned);
 224 static int64_t  tar_atol256(const char *, unsigned);
 225 static int64_t  tar_atol8(const char *, unsigned);
 226 static int      tar_read_header(struct archive_read *, struct tar *,
 227                     struct archive_entry *);
 228 static int      tohex(int c);
 229 static char     *url_decode(const char *);
 230 static wchar_t  *utf8_decode(struct tar *, const char *, size_t length);
 231
 232 int
 233 archive_read_support_format_gnutar(struct archive *a)
 234 {
 235         return (archive_read_support_format_tar(a));
 236 }
 237
 238
 239 int
 240 archive_read_support_format_tar(struct archive *_a)
 241 {
 242         struct archive_read *a = (struct archive_read *)_a;
 243         struct tar *tar;
 244         int r;
 245
 246         tar = (struct tar *)malloc(sizeof(*tar));
 247         if (tar == NULL) {
 248                 archive_set_error(&a->archive, ENOMEM,
 249                     "Can't allocate tar data");
 250                 return (ARCHIVE_FATAL);
 251         }
 252         memset(tar, 0, sizeof(*tar));
 253
 254         r = __archive_read_register_format(a, tar,
 255             archive_read_format_tar_bid,
 256             archive_read_format_tar_read_header,
 257             archive_read_format_tar_read_data,
 258             archive_read_format_tar_skip,
 259             archive_read_format_tar_cleanup);
 260
 261         if (r != ARCHIVE_OK)
 262                 free(tar);
 263         return (ARCHIVE_OK);
 264 }
 265
 266 static int
 267 archive_read_format_tar_cleanup(struct archive_read *a)
 268 {
 269         struct tar *tar;
 270
 271         tar = (struct tar *)(a->format->data);
 272         gnu_clear_sparse_list(tar);
 273         archive_string_free(&tar->acl_text);
 274         archive_string_free(&tar->entry_pathname);
 275         archive_string_free(&tar->entry_linkpath);
 276         archive_string_free(&tar->entry_uname);
 277         archive_string_free(&tar->entry_gname);
 278         archive_string_free(&tar->line);
 279         archive_string_free(&tar->pax_global);
 280         archive_string_free(&tar->pax_header);
 281         archive_string_free(&tar->longname);
 282         archive_string_free(&tar->longlink);
 283         free(tar->pax_entry);
 284         free(tar);
 285         (a->format->data) = NULL;
 286         return (ARCHIVE_OK);
 287 }
 288
 289
 290 static int
 291 archive_read_format_tar_bid(struct archive_read *a)
 292 {
 293         int bid;
 294         ssize_t bytes_read;
 295         const void *h;
 296         const struct archive_entry_header_ustar *header;
 297
 298         bid = 0;
 299
 300         /* Now let's look at the actual header and see if it matches. */
 301         if (a->decompressor->read_ahead != NULL)
 302                 bytes_read = (a->decompressor->read_ahead)(a, &h, 512);
 303         else
 304                 bytes_read = 0; /* Empty file. */
 305         if (bytes_read < 0)
 306                 return (ARCHIVE_FATAL);
 307         if (bytes_read < 512)
 308                 return (0);
 309
 310         /* If it's an end-of-archive mark, we can handle it. */
 311         if ((*(const char *)h) == 0
 312             && archive_block_is_null((const unsigned char *)h)) {
 313                 /*
 314                  * Usually, I bid the number of bits verified, but
 315                  * in this case, 4096 seems excessive so I picked 10 as
 316                  * an arbitrary but reasonable-seeming value.
 317                  */
 318                 return (10);
 319         }
 320
 321         /* If it's not an end-of-archive mark, it must have a valid checksum.*/
 322         if (!checksum(a, h))
 323                 return (0);
 324         bid += 48;  /* Checksum is usually 6 octal digits. */
 325
 326         header = (const struct archive_entry_header_ustar *)h;
 327
 328         /* Recognize POSIX formats. */
 329         if ((memcmp(header->magic, "ustar\0", 6) == 0)
 330             &&(memcmp(header->version, "00", 2)==0))
 331                 bid += 56;
 332
 333         /* Recognize GNU tar format. */
 334         if ((memcmp(header->magic, "ustar ", 6) == 0)
 335             &&(memcmp(header->version, " \0", 2)==0))
 336                 bid += 56;
 337
 338         /* Type flag must be null, digit or A-Z, a-z. */
 339         if (header->typeflag[0] != 0 &&
 340             !( header->typeflag[0] >= '0' && header->typeflag[0] <= '9') &&
 341             !( header->typeflag[0] >= 'A' && header->typeflag[0] <= 'Z') &&
 342             !( header->typeflag[0] >= 'a' && header->typeflag[0] <= 'z') )
 343                 return (0);
 344         bid += 2;  /* 6 bits of variation in an 8-bit field leaves 2 bits. */
 345
 346         /* Sanity check: Look at first byte of mode field. */
 347         switch (255 & (unsigned)header->mode[0]) {
 348         case 0: case 255:
 349                 /* Base-256 value: No further verification possible! */
 350                 break;
 351         case ' ': /* Not recommended, but not illegal, either. */
 352                 break;
 353         case '0': case '1': case '2': case '3':
 354         case '4': case '5': case '6': case '7':
 355                 /* Octal Value. */
 356                 /* TODO: Check format of remainder of this field. */
 357                 break;
 358         default:
 359                 /* Not a valid mode; bail out here. */
 360                 return (0);
 361         }
 362         /* TODO: Sanity test uid/gid/size/mtime/rdevmajor/rdevminor fields. */
 363
 364         return (bid);
 365 }
 366
 367 /*
 368  * The function invoked by archive_read_header().  This
 369  * just sets up a few things and then calls the internal
 370  * tar_read_header() function below.
 371  */
 372 static int
 373 archive_read_format_tar_read_header(struct archive_read *a,
 374     struct archive_entry *entry)
 375 {
 376         /*
 377          * When converting tar archives to cpio archives, it is
 378          * essential that each distinct file have a distinct inode
 379          * number.  To simplify this, we keep a static count here to
 380          * assign fake dev/inode numbers to each tar entry.  Note that
 381          * pax format archives may overwrite this with something more
 382          * useful.
 383          *
 384          * Ideally, we would track every file read from the archive so
 385          * that we could assign the same dev/ino pair to hardlinks,
 386          * but the memory required to store a complete lookup table is
 387          * probably not worthwhile just to support the relatively
 388          * obscure tar->cpio conversion case.
 389          */
 390         static int default_inode;
 391         static int default_dev;
 392         struct tar *tar;
 393         struct sparse_block *sp;
 394         const char *p;
 395         int r;
 396         size_t l;
 397
 398         /* Assign default device/inode values. */
 399         archive_entry_set_dev(entry, 1 + default_dev); /* Don't use zero. */
 400         archive_entry_set_ino(entry, ++default_inode); /* Don't use zero. */
 401         /* Limit generated st_ino number to 16 bits. */
 402         if (default_inode >= 0xffff) {
 403                 ++default_dev;
 404                 default_inode = 0;
 405         }
 406
 407         tar = (struct tar *)(a->format->data);
 408         tar->entry_offset = 0;
 409         while (tar->sparse_list != NULL) {
 410                 sp = tar->sparse_list;
 411                 tar->sparse_list = sp->next;
 412                 free(sp);
 413         }
 414         tar->sparse_last = NULL;
 415         tar->realsize = -1; /* Mark this as "unset" */
 416
 417         r = tar_read_header(a, tar, entry);
 418
 419         /*
 420          * "non-sparse" files are really just sparse files with
 421          * a single block.
 422          */
 423         if (tar->sparse_list == NULL)
 424                 gnu_add_sparse_entry(tar, 0, tar->entry_bytes_remaining);
 425
 426         if (r == ARCHIVE_OK) {
 427                 /*
 428                  * "Regular" entry with trailing '/' is really
 429                  * directory: This is needed for certain old tar
 430                  * variants and even for some broken newer ones.
 431                  */
 432                 p = archive_entry_pathname(entry);
 433                 l = strlen(p);
 434                 if (archive_entry_filetype(entry) == AE_IFREG
 435                     && p[l-1] == '/')
 436                         archive_entry_set_filetype(entry, AE_IFDIR);
 437         }
 438         return (r);
 439 }
 440
 441 static int
 442 archive_read_format_tar_read_data(struct archive_read *a,
 443     const void **buff, size_t *size, off_t *offset)
 444 {
 445         ssize_t bytes_read;
 446         struct tar *tar;
 447         struct sparse_block *p;
 448
 449         tar = (struct tar *)(a->format->data);
 450
 451         if (tar->sparse_gnu_pending) {
 452                 if (tar->sparse_gnu_major == 1 && tar->sparse_gnu_minor == 0) {
 453                         tar->sparse_gnu_pending = 0;
 454                         /* Read initial sparse map. */
 455                         bytes_read = gnu_sparse_10_read(a, tar);
 456                         tar->entry_bytes_remaining -= bytes_read;
 457                         if (bytes_read < 0)
 458                                 return (bytes_read);
 459                 } else {
 460                         *size = 0;
 461                         *offset = 0;
 462                         archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
 463                             "Unrecognized GNU sparse file format");
 464                         return (ARCHIVE_WARN);
 465                 }
 466                 tar->sparse_gnu_pending = 0;
 467         }
 468
 469         /* Remove exhausted entries from sparse list. */
 470         while (tar->sparse_list != NULL &&
 471             tar->sparse_list->remaining == 0) {
 472                 p = tar->sparse_list;
 473                 tar->sparse_list = p->next;
 474                 free(p);
 475         }
 476
 477         /* If we're at end of file, return EOF. */
 478         if (tar->sparse_list == NULL || tar->entry_bytes_remaining == 0) {
 479                 if ((a->decompressor->skip)(a, tar->entry_padding) < 0)
 480                         return (ARCHIVE_FATAL);
 481                 tar->entry_padding = 0;
 482                 *buff = NULL;
 483                 *size = 0;
 484                 *offset = tar->realsize;
 485                 return (ARCHIVE_EOF);
 486         }
 487
 488         bytes_read = (a->decompressor->read_ahead)(a, buff, 1);
 489         if (bytes_read == 0) {
 490                 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
 491                     "Truncated tar archive");
 492                 return (ARCHIVE_FATAL);
 493         }
 494         if (bytes_read < 0)
 495                 return (ARCHIVE_FATAL);
 496         if (bytes_read > tar->entry_bytes_remaining)
 497                 bytes_read = tar->entry_bytes_remaining;
 498         /* Don't read more than is available in the
 499          * current sparse block. */
 500         if (tar->sparse_list->remaining < bytes_read)
 501                 bytes_read = tar->sparse_list->remaining;
 502         *size = bytes_read;
 503         *offset = tar->sparse_list->offset;
 504         tar->sparse_list->remaining -= bytes_read;
 505         tar->sparse_list->offset += bytes_read;
 506         tar->entry_bytes_remaining -= bytes_read;
 507         (a->decompressor->consume)(a, bytes_read);
 508         return (ARCHIVE_OK);
 509 }
 510
 511 static int
 512 archive_read_format_tar_skip(struct archive_read *a)
 513 {
 514         off_t bytes_skipped;
 515         struct tar* tar;
 516
 517         tar = (struct tar *)(a->format->data);
 518
 519         /*
 520          * Compression layer skip functions are required to either skip the
 521          * length requested or fail, so we can rely upon the entire entry
 522          * plus padding being skipped.
 523          */
 524         bytes_skipped = (a->decompressor->skip)(a, tar->entry_bytes_remaining +
 525             tar->entry_padding);
 526         if (bytes_skipped < 0)
 527                 return (ARCHIVE_FATAL);
 528
 529         tar->entry_bytes_remaining = 0;
 530         tar->entry_padding = 0;
 531
 532         /* Free the sparse list. */
 533         gnu_clear_sparse_list(tar);
 534
 535         return (ARCHIVE_OK);
 536 }
 537
 538 /*
 539  * This function recursively interprets all of the headers associated
 540  * with a single entry.
 541  */
 542 static int
 543 tar_read_header(struct archive_read *a, struct tar *tar,
 544     struct archive_entry *entry)
 545 {
 546         ssize_t bytes;
 547         int err;
 548         const void *h;
 549         const struct archive_entry_header_ustar *header;
 550
 551         /* Read 512-byte header record */
 552         bytes = (a->decompressor->read_ahead)(a, &h, 512);
 553         if (bytes < 0)
 554                 return (bytes);
 555         if (bytes == 0) {
 556                 /*
 557                  * An archive that just ends without a proper
 558                  * end-of-archive marker.  Yes, there are tar programs
 559                  * that do this; hold our nose and accept it.
 560                  */
 561                 return (ARCHIVE_EOF);
 562         }
 563         if (bytes < 512) {
 564                 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
 565                     "Truncated tar archive");
 566                 return (ARCHIVE_FATAL);
 567         }
 568         (a->decompressor->consume)(a, 512);
 569
 570
 571         /* Check for end-of-archive mark. */
 572         if (((*(const char *)h)==0) && archive_block_is_null((const unsigned char *)h)) {
 573                 /* Try to consume a second all-null record, as well. */
 574                 bytes = (a->decompressor->read_ahead)(a, &h, 512);
 575                 if (bytes > 0)
 576                         (a->decompressor->consume)(a, bytes);
 577                 archive_set_error(&a->archive, 0, NULL);
 578                 if (a->archive.archive_format_name == NULL) {
 579                         a->archive.archive_format = ARCHIVE_FORMAT_TAR;
 580                         a->archive.archive_format_name = "tar";
 581                 }
 582                 return (ARCHIVE_EOF);
 583         }
 584
 585         /*
 586          * Note: If the checksum fails and we return ARCHIVE_RETRY,
 587          * then the client is likely to just retry.  This is a very
 588          * crude way to search for the next valid header!
 589          *
 590          * TODO: Improve this by implementing a real header scan.
 591          */
 592         if (!checksum(a, h)) {
 593                 archive_set_error(&a->archive, EINVAL, "Damaged tar archive");
 594                 return (ARCHIVE_RETRY); /* Retryable: Invalid header */
 595         }
 596
 597         if (++tar->header_recursion_depth > 32) {
 598                 archive_set_error(&a->archive, EINVAL, "Too many special headers");
 599                 return (ARCHIVE_WARN);
 600         }
 601
 602         /* Determine the format variant. */
 603         header = (const struct archive_entry_header_ustar *)h;
 604         switch(header->typeflag[0]) {
 605         case 'A': /* Solaris tar ACL */
 606                 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE;
 607                 a->archive.archive_format_name = "Solaris tar";
 608                 err = header_Solaris_ACL(a, tar, entry, h);
 609                 break;
 610         case 'g': /* POSIX-standard 'g' header. */
 611                 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE;
 612                 a->archive.archive_format_name = "POSIX pax interchange format";
 613                 err = header_pax_global(a, tar, entry, h);
 614                 break;
 615         case 'K': /* Long link name (GNU tar, others) */
 616                 err = header_longlink(a, tar, entry, h);
 617                 break;
 618         case 'L': /* Long filename (GNU tar, others) */
 619                 err = header_longname(a, tar, entry, h);
 620                 break;
 621         case 'V': /* GNU volume header */
 622                 err = header_volume(a, tar, entry, h);
 623                 break;
 624         case 'X': /* Used by SUN tar; same as 'x'. */
 625                 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE;
 626                 a->archive.archive_format_name =
 627                     "POSIX pax interchange format (Sun variant)";
 628                 err = header_pax_extensions(a, tar, entry, h);
 629                 break;
 630         case 'x': /* POSIX-standard 'x' header. */
 631                 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE;
 632                 a->archive.archive_format_name = "POSIX pax interchange format";
 633                 err = header_pax_extensions(a, tar, entry, h);
 634                 break;
 635         default:
 636                 if (memcmp(header->magic, "ustar  \0", 8) == 0) {
 637                         a->archive.archive_format = ARCHIVE_FORMAT_TAR_GNUTAR;
 638                         a->archive.archive_format_name = "GNU tar format";
 639                         err = header_gnutar(a, tar, entry, h);
 640                 } else if (memcmp(header->magic, "ustar", 5) == 0) {
 641                         if (a->archive.archive_format != ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE) {
 642                                 a->archive.archive_format = ARCHIVE_FORMAT_TAR_USTAR;
 643                                 a->archive.archive_format_name = "POSIX ustar format";
 644                         }
 645                         err = header_ustar(a, tar, entry, h);
 646                 } else {
 647                         a->archive.archive_format = ARCHIVE_FORMAT_TAR;
 648                         a->archive.archive_format_name = "tar (non-POSIX)";
 649                         err = header_old_tar(a, tar, entry, h);
 650                 }
 651         }
 652         --tar->header_recursion_depth;
 653         /* We return warnings or success as-is.  Anything else is fatal. */
 654         if (err == ARCHIVE_WARN || err == ARCHIVE_OK)
 655                 return (err);
 656         if (err == ARCHIVE_EOF)
 657                 /* EOF when recursively reading a header is bad. */
 658                 archive_set_error(&a->archive, EINVAL, "Damaged tar archive");
 659         return (ARCHIVE_FATAL);
 660 }
 661
 662 /*
 663  * Return true if block checksum is correct.
 664  */
 665 static int
 666 checksum(struct archive_read *a, const void *h)
 667 {
 668         const unsigned char *bytes;
 669         const struct archive_entry_header_ustar *header;
 670         int check, i, sum;
 671
 672         (void)a; /* UNUSED */
 673         bytes = (const unsigned char *)h;
 674         header = (const struct archive_entry_header_ustar *)h;
 675
 676         /*
 677          * Test the checksum.  Note that POSIX specifies _unsigned_
 678          * bytes for this calculation.
 679          */
 680         sum = tar_atol(header->checksum, sizeof(header->checksum));
 681         check = 0;
 682         for (i = 0; i < 148; i++)
 683                 check += (unsigned char)bytes[i];
 684         for (; i < 156; i++)
 685                 check += 32;
 686         for (; i < 512; i++)
 687                 check += (unsigned char)bytes[i];
 688         if (sum == check)
 689                 return (1);
 690
 691         /*
 692          * Repeat test with _signed_ bytes, just in case this archive
 693          * was created by an old BSD, Solaris, or HP-UX tar with a
 694          * broken checksum calculation.
 695          */
 696         check = 0;
 697         for (i = 0; i < 148; i++)
 698                 check += (signed char)bytes[i];
 699         for (; i < 156; i++)
 700                 check += 32;
 701         for (; i < 512; i++)
 702                 check += (signed char)bytes[i];
 703         if (sum == check)
 704                 return (1);
 705
 706         return (0);
 707 }
 708
 709 /*
 710  * Return true if this block contains only nulls.
 711  */
 712 static int
 713 archive_block_is_null(const unsigned char *p)
 714 {
 715         unsigned i;
 716
 717         for (i = 0; i < ARCHIVE_BYTES_PER_RECORD / sizeof(*p); i++)
 718                 if (*p++)
 719                         return (0);
 720         return (1);
 721 }
 722
 723 /*
 724  * Interpret 'A' Solaris ACL header
 725  */
 726 static int
 727 header_Solaris_ACL(struct archive_read *a, struct tar *tar,
 728     struct archive_entry *entry, const void *h)
 729 {
 730         const struct archive_entry_header_ustar *header;
 731         size_t size;
 732         int err;
 733         char *acl, *p;
 734         wchar_t *wp;
 735
 736         /*
 737          * read_body_to_string adds a NUL terminator, but we need a little
 738          * more to make sure that we don't overrun acl_text later.
 739          */
 740         header = (const struct archive_entry_header_ustar *)h;
 741         size = tar_atol(header->size, sizeof(header->size));
 742         err = read_body_to_string(a, tar, &(tar->acl_text), h);
 743         if (err != ARCHIVE_OK)
 744                 return (err);
 745         err = tar_read_header(a, tar, entry);
 746         if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN))
 747                 return (err);
 748
 749         /* Skip leading octal number. */
 750         /* XXX TODO: Parse the octal number and sanity-check it. */
 751         p = acl = tar->acl_text.s;
 752         while (*p != '\0' && p < acl + size)
 753                 p++;
 754         p++;
 755
 756         if (p >= acl + size) {
 757                 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
 758                     "Malformed Solaris ACL attribute");
 759                 return(ARCHIVE_WARN);
 760         }
 761
 762         /* Skip leading octal number. */
 763         size -= (p - acl);
 764         acl = p;
 765
 766         while (*p != '\0' && p < acl + size)
 767                 p++;
 768
 769         wp = utf8_decode(tar, acl, p - acl);
 770         err = __archive_entry_acl_parse_w(entry, wp,
 771             ARCHIVE_ENTRY_ACL_TYPE_ACCESS);
 772         return (err);
 773 }
 774
 775 /*
 776  * Interpret 'K' long linkname header.
 777  */
 778 static int
 779 header_longlink(struct archive_read *a, struct tar *tar,
 780     struct archive_entry *entry, const void *h)
 781 {
 782         int err;
 783
 784         err = read_body_to_string(a, tar, &(tar->longlink), h);
 785         if (err != ARCHIVE_OK)
 786                 return (err);
 787         err = tar_read_header(a, tar, entry);
 788         if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN))
 789                 return (err);
 790         /* Set symlink if symlink already set, else hardlink. */
 791         archive_entry_copy_link(entry, tar->longlink.s);
 792         return (ARCHIVE_OK);
 793 }
 794
 795 /*
 796  * Interpret 'L' long filename header.
 797  */
 798 static int
 799 header_longname(struct archive_read *a, struct tar *tar,
 800     struct archive_entry *entry, const void *h)
 801 {
 802         int err;
 803
 804         err = read_body_to_string(a, tar, &(tar->longname), h);
 805         if (err != ARCHIVE_OK)
 806                 return (err);
 807         /* Read and parse "real" header, then override name. */
 808         err = tar_read_header(a, tar, entry);
 809         if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN))
 810                 return (err);
 811         archive_entry_copy_pathname(entry, tar->longname.s);
 812         return (ARCHIVE_OK);
 813 }
 814
 815
 816 /*
 817  * Interpret 'V' GNU tar volume header.
 818  */
 819 static int
 820 header_volume(struct archive_read *a, struct tar *tar,
 821     struct archive_entry *entry, const void *h)
 822 {
 823         (void)h;
 824
 825         /* Just skip this and read the next header. */
 826         return (tar_read_header(a, tar, entry));
 827 }
 828
 829 /*
 830  * Read body of an archive entry into an archive_string object.
 831  */
 832 static int
 833 read_body_to_string(struct archive_read *a, struct tar *tar,
 834     struct archive_string *as, const void *h)
 835 {
 836         off_t size, padded_size;
 837         ssize_t bytes_read, bytes_to_copy;
 838         const struct archive_entry_header_ustar *header;
 839         const void *src;
 840         char *dest;
 841
 842         (void)tar; /* UNUSED */
 843         header = (const struct archive_entry_header_ustar *)h;
 844         size  = tar_atol(header->size, sizeof(header->size));
 845         if ((size > 1048576) || (size < 0)) {
 846                 archive_set_error(&a->archive, EINVAL,
 847                     "Special header too large");
 848                 return (ARCHIVE_FATAL);
 849         }
 850
 851         /* Fail if we can't make our buffer big enough. */
 852         if (archive_string_ensure(as, size+1) == NULL) {
 853                 archive_set_error(&a->archive, ENOMEM,
 854                     "No memory");
 855                 return (ARCHIVE_FATAL);
 856         }
 857
 858         /* Read the body into the string. */
 859         padded_size = (size + 511) & ~ 511;
 860         dest = as->s;
 861         while (padded_size > 0) {
 862                 bytes_read = (a->decompressor->read_ahead)(a, &src, padded_size);
 863                 if (bytes_read == 0)
 864                         return (ARCHIVE_EOF);
 865                 if (bytes_read < 0)
 866                         return (ARCHIVE_FATAL);
 867                 if (bytes_read > padded_size)
 868                         bytes_read = padded_size;
 869                 (a->decompressor->consume)(a, bytes_read);
 870                 bytes_to_copy = bytes_read;
 871                 if ((off_t)bytes_to_copy > size)
 872                         bytes_to_copy = (ssize_t)size;
 873                 memcpy(dest, src, bytes_to_copy);
 874                 dest += bytes_to_copy;
 875                 size -= bytes_to_copy;
 876                 padded_size -= bytes_read;
 877         }
 878         *dest = '\0';
 879         return (ARCHIVE_OK);
 880 }
 881
 882 /*
 883  * Parse out common header elements.
 884  *
 885  * This would be the same as header_old_tar, except that the
 886  * filename is handled slightly differently for old and POSIX
 887  * entries  (POSIX entries support a 'prefix').  This factoring
 888  * allows header_old_tar and header_ustar
 889  * to handle filenames differently, while still putting most of the
 890  * common parsing into one place.
 891  */
 892 static int
 893 header_common(struct archive_read *a, struct tar *tar,
 894     struct archive_entry *entry, const void *h)
 895 {
 896         const struct archive_entry_header_ustar *header;
 897         char    tartype;
 898
 899         (void)a; /* UNUSED */
 900
 901         header = (const struct archive_entry_header_ustar *)h;
 902         if (header->linkname[0])
 903                 archive_strncpy(&(tar->entry_linkpath), header->linkname,
 904                     sizeof(header->linkname));
 905         else
 906                 archive_string_empty(&(tar->entry_linkpath));
 907
 908         /* Parse out the numeric fields (all are octal) */
 909         archive_entry_set_mode(entry, tar_atol(header->mode, sizeof(header->mode)));
 910         archive_entry_set_uid(entry, tar_atol(header->uid, sizeof(header->uid)));
 911         archive_entry_set_gid(entry, tar_atol(header->gid, sizeof(header->gid)));
 912         tar->entry_bytes_remaining = tar_atol(header->size, sizeof(header->size));
 913         tar->realsize = tar->entry_bytes_remaining;
 914         archive_entry_set_size(entry, tar->entry_bytes_remaining);
 915         archive_entry_set_mtime(entry, tar_atol(header->mtime, sizeof(header->mtime)), 0);
 916
 917         /* Handle the tar type flag appropriately. */
 918         tartype = header->typeflag[0];
 919
 920         switch (tartype) {
 921         case '1': /* Hard link */
 922                 archive_entry_copy_hardlink(entry, tar->entry_linkpath.s);
 923                 /*
 924                  * The following may seem odd, but: Technically, tar
 925                  * does not store the file type for a "hard link"
 926                  * entry, only the fact that it is a hard link.  So, I
 927                  * leave the type zero normally.  But, pax interchange
 928                  * format allows hard links to have data, which
 929                  * implies that the underlying entry is a regular
 930                  * file.
 931                  */
 932                 if (archive_entry_size(entry) > 0)
 933                         archive_entry_set_filetype(entry, AE_IFREG);
 934
 935                 /*
 936                  * A tricky point: Traditionally, tar readers have
 937                  * ignored the size field when reading hardlink
 938                  * entries, and some writers put non-zero sizes even
 939                  * though the body is empty.  POSIX blessed this
 940                  * convention in the 1988 standard, but broke with
 941                  * this tradition in 2001 by permitting hardlink
 942                  * entries to store valid bodies in pax interchange
 943                  * format, but not in ustar format.  Since there is no
 944                  * hard and fast way to distinguish pax interchange
 945                  * from earlier archives (the 'x' and 'g' entries are
 946                  * optional, after all), we need a heuristic.
 947                  */
 948                 if (archive_entry_size(entry) == 0) {
 949                         /* If the size is already zero, we're done. */
 950                 }  else if (a->archive.archive_format
 951                     == ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE) {
 952                         /* Definitely pax extended; must obey hardlink size. */
 953                 } else if (a->archive.archive_format == ARCHIVE_FORMAT_TAR
 954                     || a->archive.archive_format == ARCHIVE_FORMAT_TAR_GNUTAR)
 955                 {
 956                         /* Old-style or GNU tar: we must ignore the size. */
 957                         archive_entry_set_size(entry, 0);
 958                         tar->entry_bytes_remaining = 0;
 959                 } else if (archive_read_format_tar_bid(a) > 50) {
 960                         /*
 961                          * We don't know if it's pax: If the bid
 962                          * function sees a valid ustar header
 963                          * immediately following, then let's ignore
 964                          * the hardlink size.
 965                          */
 966                         archive_entry_set_size(entry, 0);
 967                         tar->entry_bytes_remaining = 0;
 968                 }
 969                 /*
 970                  * TODO: There are still two cases I'd like to handle:
 971                  *   = a ustar non-pax archive with a hardlink entry at
 972                  *     end-of-archive.  (Look for block of nulls following?)
 973                  *   = a pax archive that has not seen any pax headers
 974                  *     and has an entry which is a hardlink entry storing
 975                  *     a body containing an uncompressed tar archive.
 976                  * The first is worth addressing; I don't see any reliable
 977                  * way to deal with the second possibility.
 978                  */
 979                 break;
 980         case '2': /* Symlink */
 981                 archive_entry_set_filetype(entry, AE_IFLNK);
 982                 archive_entry_set_size(entry, 0);
 983                 tar->entry_bytes_remaining = 0;
 984                 archive_entry_copy_symlink(entry, tar->entry_linkpath.s);
 985                 break;
 986         case '3': /* Character device */
 987                 archive_entry_set_filetype(entry, AE_IFCHR);
 988                 archive_entry_set_size(entry, 0);
 989                 tar->entry_bytes_remaining = 0;
 990                 break;
 991         case '4': /* Block device */
 992                 archive_entry_set_filetype(entry, AE_IFBLK);
 993                 archive_entry_set_size(entry, 0);
 994                 tar->entry_bytes_remaining = 0;
 995                 break;
 996         case '5': /* Dir */
 997                 archive_entry_set_filetype(entry, AE_IFDIR);
 998                 archive_entry_set_size(entry, 0);
 999                 tar->entry_bytes_remaining = 0;
1000                 break;
1001         case '6': /* FIFO device */
1002                 archive_entry_set_filetype(entry, AE_IFIFO);
1003                 archive_entry_set_size(entry, 0);
1004                 tar->entry_bytes_remaining = 0;
1005                 break;
1006         case 'D': /* GNU incremental directory type */
1007                 /*
1008                  * No special handling is actually required here.
1009                  * It might be nice someday to preprocess the file list and
1010                  * provide it to the client, though.
1011                  */
1012                 archive_entry_set_filetype(entry, AE_IFDIR);
1013                 break;
1014         case 'M': /* GNU "Multi-volume" (remainder of file from last archive)*/
1015                 /*
1016                  * As far as I can tell, this is just like a regular file
1017                  * entry, except that the contents should be _appended_ to
1018                  * the indicated file at the indicated offset.  This may
1019                  * require some API work to fully support.
1020                  */
1021                 break;
1022         case 'N': /* Old GNU "long filename" entry. */
1023                 /* The body of this entry is a script for renaming
1024                  * previously-extracted entries.  Ugh.  It will never
1025                  * be supported by libarchive. */
1026                 archive_entry_set_filetype(entry, AE_IFREG);
1027                 break;
1028         case 'S': /* GNU sparse files */
1029                 /*
1030                  * Sparse files are really just regular files with
1031                  * sparse information in the extended area.
1032                  */
1033                 /* FALLTHROUGH */
1034         default: /* Regular file  and non-standard types */
1035                 /*
1036                  * Per POSIX: non-recognized types should always be
1037                  * treated as regular files.
1038                  */
1039                 archive_entry_set_filetype(entry, AE_IFREG);
1040                 break;
1041         }
1042         return (0);
1043 }
1044
1045 /*
1046  * Parse out header elements for "old-style" tar archives.
1047  */
1048 static int
1049 header_old_tar(struct archive_read *a, struct tar *tar,
1050     struct archive_entry *entry, const void *h)
1051 {
1052         const struct archive_entry_header_ustar *header;
1053
1054         /* Copy filename over (to ensure null termination). */
1055         header = (const struct archive_entry_header_ustar *)h;
1056         archive_strncpy(&(tar->entry_pathname), header->name, sizeof(header->name));
1057         archive_entry_copy_pathname(entry, tar->entry_pathname.s);
1058
1059         /* Grab rest of common fields */
1060         header_common(a, tar, entry, h);
1061
1062         tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining);
1063         return (0);
1064 }
1065
1066 /*
1067  * Parse a file header for a pax extended archive entry.
1068  */
1069 static int
1070 header_pax_global(struct archive_read *a, struct tar *tar,
1071     struct archive_entry *entry, const void *h)
1072 {
1073         int err;
1074
1075         err = read_body_to_string(a, tar, &(tar->pax_global), h);
1076         if (err != ARCHIVE_OK)
1077                 return (err);
1078         err = tar_read_header(a, tar, entry);
1079         return (err);
1080 }
1081
1082 static int
1083 header_pax_extensions(struct archive_read *a, struct tar *tar,
1084     struct archive_entry *entry, const void *h)
1085 {
1086         int err, err2;
1087
1088         err = read_body_to_string(a, tar, &(tar->pax_header), h);
1089         if (err != ARCHIVE_OK)
1090                 return (err);
1091
1092         /* Parse the next header. */
1093         err = tar_read_header(a, tar, entry);
1094         if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN))
1095                 return (err);
1096
1097         /*
1098          * TODO: Parse global/default options into 'entry' struct here
1099          * before handling file-specific options.
1100          *
1101          * This design (parse standard header, then overwrite with pax
1102          * extended attribute data) usually works well, but isn't ideal;
1103          * it would be better to parse the pax extended attributes first
1104          * and then skip any fields in the standard header that were
1105          * defined in the pax header.
1106          */
1107         err2 = pax_header(a, tar, entry, tar->pax_header.s);
1108         err =  err_combine(err, err2);
1109         tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining);
1110         return (err);
1111 }
1112
1113
1114 /*
1115  * Parse a file header for a Posix "ustar" archive entry.  This also
1116  * handles "pax" or "extended ustar" entries.
1117  */
1118 static int
1119 header_ustar(struct archive_read *a, struct tar *tar,
1120     struct archive_entry *entry, const void *h)
1121 {
1122         const struct archive_entry_header_ustar *header;
1123         struct archive_string *as;
1124
1125         header = (const struct archive_entry_header_ustar *)h;
1126
1127         /* Copy name into an internal buffer to ensure null-termination. */
1128         as = &(tar->entry_pathname);
1129         if (header->prefix[0]) {
1130                 archive_strncpy(as, header->prefix, sizeof(header->prefix));
1131                 if (as->s[archive_strlen(as) - 1] != '/')
1132                         archive_strappend_char(as, '/');
1133                 archive_strncat(as, header->name, sizeof(header->name));
1134         } else
1135                 archive_strncpy(as, header->name, sizeof(header->name));
1136
1137         archive_entry_copy_pathname(entry, as->s);
1138
1139         /* Handle rest of common fields. */
1140         header_common(a, tar, entry, h);
1141
1142         /* Handle POSIX ustar fields. */
1143         archive_strncpy(&(tar->entry_uname), header->uname,
1144             sizeof(header->uname));
1145         archive_entry_copy_uname(entry, tar->entry_uname.s);
1146
1147         archive_strncpy(&(tar->entry_gname), header->gname,
1148             sizeof(header->gname));
1149         archive_entry_copy_gname(entry, tar->entry_gname.s);
1150
1151         /* Parse out device numbers only for char and block specials. */
1152         if (header->typeflag[0] == '3' || header->typeflag[0] == '4') {
1153                 archive_entry_set_rdevmajor(entry,
1154                     tar_atol(header->rdevmajor, sizeof(header->rdevmajor)));
1155                 archive_entry_set_rdevminor(entry,
1156                     tar_atol(header->rdevminor, sizeof(header->rdevminor)));
1157         }
1158
1159         tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining);
1160
1161         return (0);
1162 }
1163
1164
1165 /*
1166  * Parse the pax extended attributes record.
1167  *
1168  * Returns non-zero if there's an error in the data.
1169  */
1170 static int
1171 pax_header(struct archive_read *a, struct tar *tar,
1172     struct archive_entry *entry, char *attr)
1173 {
1174         size_t attr_length, l, line_length;
1175         char *line, *p;
1176         char *key, *value;
1177         wchar_t *wp;
1178         int err, err2;
1179
1180         attr_length = strlen(attr);
1181         tar->pax_hdrcharset_binary = 0;
1182         archive_string_empty(&(tar->entry_gname));
1183         archive_string_empty(&(tar->entry_linkpath));
1184         archive_string_empty(&(tar->entry_pathname));
1185         archive_string_empty(&(tar->entry_uname));
1186         err = ARCHIVE_OK;
1187         while (attr_length > 0) {
1188                 /* Parse decimal length field at start of line. */
1189                 line_length = 0;
1190                 l = attr_length;
1191                 line = p = attr; /* Record start of line. */
1192                 while (l>0) {
1193                         if (*p == ' ') {
1194                                 p++;
1195                                 l--;
1196                                 break;
1197                         }
1198                         if (*p < '0' || *p > '9') {
1199                                 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1200                                     "Ignoring malformed pax extended attributes");
1201                                 return (ARCHIVE_WARN);
1202                         }
1203                         line_length *= 10;
1204                         line_length += *p - '0';
1205                         if (line_length > 999999) {
1206                                 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1207                                     "Rejecting pax extended attribute > 1MB");
1208                                 return (ARCHIVE_WARN);
1209                         }
1210                         p++;
1211                         l--;
1212                 }
1213
1214                 /*
1215                  * Parsed length must be no bigger than available data,
1216                  * at least 1, and the last character of the line must
1217                  * be '\n'.
1218                  */
1219                 if (line_length > attr_length
1220                     || line_length < 1
1221                     || attr[line_length - 1] != '\n')
1222                 {
1223                         archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1224                             "Ignoring malformed pax extended attribute");
1225                         return (ARCHIVE_WARN);
1226                 }
1227
1228                 /* Null-terminate the line. */
1229                 attr[line_length - 1] = '\0';
1230
1231                 /* Find end of key and null terminate it. */
1232                 key = p;
1233                 if (key[0] == '=')
1234                         return (-1);
1235                 while (*p && *p != '=')
1236                         ++p;
1237                 if (*p == '\0') {
1238                         archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1239                             "Invalid pax extended attributes");
1240                         return (ARCHIVE_WARN);
1241                 }
1242                 *p = '\0';
1243
1244                 /* Identify null-terminated 'value' portion. */
1245                 value = p + 1;
1246
1247                 /* Identify this attribute and set it in the entry. */
1248                 err2 = pax_attribute(tar, entry, key, value);
1249                 err = err_combine(err, err2);
1250
1251                 /* Skip to next line */
1252                 attr += line_length;
1253                 attr_length -= line_length;
1254         }
1255         if (archive_strlen(&(tar->entry_gname)) > 0) {
1256                 value = tar->entry_gname.s;
1257                 if (tar->pax_hdrcharset_binary)
1258                         archive_entry_copy_gname(entry, value);
1259                 else {
1260                         wp = utf8_decode(tar, value, strlen(value));
1261                         if (wp == NULL)
1262                                 archive_entry_copy_gname(entry, value);
1263                         else
1264                                 archive_entry_copy_gname_w(entry, wp);
1265                 }
1266         }
1267         if (archive_strlen(&(tar->entry_linkpath)) > 0) {
1268                 value = tar->entry_linkpath.s;
1269                 if (tar->pax_hdrcharset_binary)
1270                         archive_entry_copy_link(entry, value);
1271                 else {
1272                         wp = utf8_decode(tar, value, strlen(value));
1273                         if (wp == NULL)
1274                                 archive_entry_copy_link(entry, value);
1275                         else
1276                                 archive_entry_copy_link_w(entry, wp);
1277                 }
1278         }
1279         if (archive_strlen(&(tar->entry_pathname)) > 0) {
1280                 value = tar->entry_pathname.s;
1281                 if (tar->pax_hdrcharset_binary)
1282                         archive_entry_copy_pathname(entry, value);
1283                 else {
1284                         wp = utf8_decode(tar, value, strlen(value));
1285                         if (wp == NULL)
1286                                 archive_entry_copy_pathname(entry, value);
1287                         else
1288                                 archive_entry_copy_pathname_w(entry, wp);
1289                 }
1290         }
1291         if (archive_strlen(&(tar->entry_uname)) > 0) {
1292                 value = tar->entry_uname.s;
1293                 if (tar->pax_hdrcharset_binary)
1294                         archive_entry_copy_uname(entry, value);
1295                 else {
1296                         wp = utf8_decode(tar, value, strlen(value));
1297                         if (wp == NULL)
1298                                 archive_entry_copy_uname(entry, value);
1299                         else
1300                                 archive_entry_copy_uname_w(entry, wp);
1301                 }
1302         }
1303         return (err);
1304 }
1305
1306 static int
1307 pax_attribute_xattr(struct archive_entry *entry,
1308         char *name, char *value)
1309 {
1310         char *name_decoded;
1311         void *value_decoded;
1312         size_t value_len;
1313
1314         if (strlen(name) < 18 || (strncmp(name, "LIBARCHIVE.xattr.", 17)) != 0)
1315                 return 3;
1316
1317         name += 17;
1318
1319         /* URL-decode name */
1320         name_decoded = url_decode(name);
1321         if (name_decoded == NULL)
1322                 return 2;
1323
1324         /* Base-64 decode value */
1325         value_decoded = base64_decode(value, strlen(value), &value_len);
1326         if (value_decoded == NULL) {
1327                 free(name_decoded);
1328                 return 1;
1329         }
1330
1331         archive_entry_xattr_add_entry(entry, name_decoded,
1332                 value_decoded, value_len);
1333
1334         free(name_decoded);
1335         free(value_decoded);
1336         return 0;
1337 }
1338
1339 /*
1340  * Parse a single key=value attribute.  key/value pointers are
1341  * assumed to point into reasonably long-lived storage.
1342  *
1343  * Note that POSIX reserves all-lowercase keywords.  Vendor-specific
1344  * extensions should always have keywords of the form "VENDOR.attribute"
1345  * In particular, it's quite feasible to support many different
1346  * vendor extensions here.  I'm using "LIBARCHIVE" for extensions
1347  * unique to this library.
1348  *
1349  * Investigate other vendor-specific extensions and see if
1350  * any of them look useful.
1351  */
1352 static int
1353 pax_attribute(struct tar *tar, struct archive_entry *entry,
1354     char *key, char *value)
1355 {
1356         int64_t s;
1357         long n;
1358         wchar_t *wp;
1359
1360         switch (key[0]) {
1361         case 'G':
1362                 /* GNU "0.0" sparse pax format. */
1363                 if (strcmp(key, "GNU.sparse.numblocks") == 0) {
1364                         tar->sparse_offset = -1;
1365                         tar->sparse_numbytes = -1;
1366                         tar->sparse_gnu_major = 0;
1367                         tar->sparse_gnu_minor = 0;
1368                 }
1369                 if (strcmp(key, "GNU.sparse.offset") == 0) {
1370                         tar->sparse_offset = tar_atol10(value, strlen(value));
1371                         if (tar->sparse_numbytes != -1) {
1372                                 gnu_add_sparse_entry(tar,
1373                                     tar->sparse_offset, tar->sparse_numbytes);
1374                                 tar->sparse_offset = -1;
1375                                 tar->sparse_numbytes = -1;
1376                         }
1377                 }
1378                 if (strcmp(key, "GNU.sparse.numbytes") == 0) {
1379                         tar->sparse_numbytes = tar_atol10(value, strlen(value));
1380                         if (tar->sparse_numbytes != -1) {
1381                                 gnu_add_sparse_entry(tar,
1382                                     tar->sparse_offset, tar->sparse_numbytes);
1383                                 tar->sparse_offset = -1;
1384                                 tar->sparse_numbytes = -1;
1385                         }
1386                 }
1387                 if (strcmp(key, "GNU.sparse.size") == 0) {
1388                         tar->realsize = tar_atol10(value, strlen(value));
1389                         archive_entry_set_size(entry, tar->realsize);
1390                 }
1391
1392                 /* GNU "0.1" sparse pax format. */
1393                 if (strcmp(key, "GNU.sparse.map") == 0) {
1394                         tar->sparse_gnu_major = 0;
1395                         tar->sparse_gnu_minor = 1;
1396                         if (gnu_sparse_01_parse(tar, value) != ARCHIVE_OK)
1397                                 return (ARCHIVE_WARN);
1398                 }
1399
1400                 /* GNU "1.0" sparse pax format */
1401                 if (strcmp(key, "GNU.sparse.major") == 0) {
1402                         tar->sparse_gnu_major = tar_atol10(value, strlen(value));
1403                         tar->sparse_gnu_pending = 1;
1404                 }
1405                 if (strcmp(key, "GNU.sparse.minor") == 0) {
1406                         tar->sparse_gnu_minor = tar_atol10(value, strlen(value));
1407                         tar->sparse_gnu_pending = 1;
1408                 }
1409                 if (strcmp(key, "GNU.sparse.name") == 0) {
1410                         wp = utf8_decode(tar, value, strlen(value));
1411                         if (wp != NULL)
1412                                 archive_entry_copy_pathname_w(entry, wp);
1413                         else
1414                                 archive_entry_copy_pathname(entry, value);
1415                 }
1416                 if (strcmp(key, "GNU.sparse.realsize") == 0) {
1417                         tar->realsize = tar_atol10(value, strlen(value));
1418                         archive_entry_set_size(entry, tar->realsize);
1419                 }
1420                 break;
1421         case 'L':
1422                 /* Our extensions */
1423 /* TODO: Handle arbitrary extended attributes... */
1424 /*
1425                 if (strcmp(key, "LIBARCHIVE.xxxxxxx")==0)
1426                         archive_entry_set_xxxxxx(entry, value);
1427 */
1428                 if (strncmp(key, "LIBARCHIVE.xattr.", 17)==0)
1429                         pax_attribute_xattr(entry, key, value);
1430                 break;
1431         case 'S':
1432                 /* We support some keys used by the "star" archiver */
1433                 if (strcmp(key, "SCHILY.acl.access")==0) {
1434                         wp = utf8_decode(tar, value, strlen(value));
1435                         /* TODO: if (wp == NULL) */
1436                         __archive_entry_acl_parse_w(entry, wp,
1437                             ARCHIVE_ENTRY_ACL_TYPE_ACCESS);
1438                 } else if (strcmp(key, "SCHILY.acl.default")==0) {
1439                         wp = utf8_decode(tar, value, strlen(value));
1440                         /* TODO: if (wp == NULL) */
1441                         __archive_entry_acl_parse_w(entry, wp,
1442                             ARCHIVE_ENTRY_ACL_TYPE_DEFAULT);
1443                 } else if (strcmp(key, "SCHILY.devmajor")==0) {
1444                         archive_entry_set_rdevmajor(entry,
1445                             tar_atol10(value, strlen(value)));
1446                 } else if (strcmp(key, "SCHILY.devminor")==0) {
1447                         archive_entry_set_rdevminor(entry,
1448                             tar_atol10(value, strlen(value)));
1449                 } else if (strcmp(key, "SCHILY.fflags")==0) {
1450                         wp = utf8_decode(tar, value, strlen(value));
1451                         /* TODO: if (wp == NULL) */
1452                         archive_entry_copy_fflags_text_w(entry, wp);
1453                 } else if (strcmp(key, "SCHILY.dev")==0) {
1454                         archive_entry_set_dev(entry,
1455                             tar_atol10(value, strlen(value)));
1456                 } else if (strcmp(key, "SCHILY.ino")==0) {
1457                         archive_entry_set_ino(entry,
1458                             tar_atol10(value, strlen(value)));
1459                 } else if (strcmp(key, "SCHILY.nlink")==0) {
1460                         archive_entry_set_nlink(entry,
1461                             tar_atol10(value, strlen(value)));
1462                 } else if (strcmp(key, "SCHILY.realsize")==0) {
1463                         tar->realsize = tar_atol10(value, strlen(value));
1464                         archive_entry_set_size(entry, tar->realsize);
1465                 }
1466                 break;
1467         case 'a':
1468                 if (strcmp(key, "atime")==0) {
1469                         pax_time(value, &s, &n);
1470                         archive_entry_set_atime(entry, s, n);
1471                 }
1472                 break;
1473         case 'c':
1474                 if (strcmp(key, "ctime")==0) {
1475                         pax_time(value, &s, &n);
1476                         archive_entry_set_ctime(entry, s, n);
1477                 } else if (strcmp(key, "charset")==0) {
1478                         /* TODO: Publish charset information in entry. */
1479                 } else if (strcmp(key, "comment")==0) {
1480                         /* TODO: Publish comment in entry. */
1481                 }
1482                 break;
1483         case 'g':
1484                 if (strcmp(key, "gid")==0) {
1485                         archive_entry_set_gid(entry,
1486                             tar_atol10(value, strlen(value)));
1487                 } else if (strcmp(key, "gname")==0) {
1488                         archive_strcpy(&(tar->entry_gname), value);
1489                 }
1490                 break;
1491         case 'h':
1492                 if (strcmp(key, "hdrcharset") == 0) {
1493                         if (strcmp(value, "BINARY") == 0)
1494                                 tar->pax_hdrcharset_binary = 1;
1495                         else if (strcmp(value, "ISO-IR 10646 2000 UTF-8") == 0)
1496                                 tar->pax_hdrcharset_binary = 0;
1497                         else {
1498                                 /* TODO: Warn about unsupported hdrcharset */
1499                         }
1500                 }
1501                 break;
1502         case 'l':
1503                 /* pax interchange doesn't distinguish hardlink vs. symlink. */
1504                 if (strcmp(key, "linkpath")==0) {
1505                         archive_strcpy(&(tar->entry_linkpath), value);
1506                 }
1507                 break;
1508         case 'm':
1509                 if (strcmp(key, "mtime")==0) {
1510                         pax_time(value, &s, &n);
1511                         archive_entry_set_mtime(entry, s, n);
1512                 }
1513                 break;
1514         case 'p':
1515                 if (strcmp(key, "path")==0) {
1516                         archive_strcpy(&(tar->entry_pathname), value);
1517                 }
1518                 break;
1519         case 'r':
1520                 /* POSIX has reserved 'realtime.*' */
1521                 break;
1522         case 's':
1523                 /* POSIX has reserved 'security.*' */
1524                 /* Someday: if (strcmp(key, "security.acl")==0) { ... } */
1525                 if (strcmp(key, "size")==0) {
1526                         /* "size" is the size of the data in the entry. */
1527                         tar->entry_bytes_remaining
1528                             = tar_atol10(value, strlen(value));
1529                         /*
1530                          * But, "size" is not necessarily the size of
1531                          * the file on disk; if this is a sparse file,
1532                          * the disk size may have already been set from
1533                          * GNU.sparse.realsize or GNU.sparse.size or
1534                          * an old GNU header field or SCHILY.realsize
1535                          * or ....
1536                          */
1537                         if (tar->realsize < 0) {
1538                                 archive_entry_set_size(entry,
1539                                     tar->entry_bytes_remaining);
1540                                 tar->realsize
1541                                     = tar->entry_bytes_remaining;
1542                         }
1543                 }
1544                 break;
1545         case 'u':
1546                 if (strcmp(key, "uid")==0) {
1547                         archive_entry_set_uid(entry,
1548                             tar_atol10(value, strlen(value)));
1549                 } else if (strcmp(key, "uname")==0) {
1550                         archive_strcpy(&(tar->entry_uname), value);
1551                 }
1552                 break;
1553         }
1554         return (0);
1555 }
1556
1557
1558
1559 /*
1560  * parse a decimal time value, which may include a fractional portion
1561  */
1562 static void
1563 pax_time(const char *p, int64_t *ps, long *pn)
1564 {
1565         char digit;
1566         int64_t s;
1567         unsigned long l;
1568         int sign;
1569         int64_t limit, last_digit_limit;
1570
1571         limit = INT64_MAX / 10;
1572         last_digit_limit = INT64_MAX % 10;
1573
1574         s = 0;
1575         sign = 1;
1576         if (*p == '-') {
1577                 sign = -1;
1578                 p++;
1579         }
1580         while (*p >= '0' && *p <= '9') {
1581                 digit = *p - '0';
1582                 if (s > limit ||
1583                     (s == limit && digit > last_digit_limit)) {
1584                         s = UINT64_MAX;
1585                         break;
1586                 }
1587                 s = (s * 10) + digit;
1588                 ++p;
1589         }
1590
1591         *ps = s * sign;
1592
1593         /* Calculate nanoseconds. */
1594         *pn = 0;
1595
1596         if (*p != '.')
1597                 return;
1598
1599         l = 100000000UL;
1600         do {
1601                 ++p;
1602                 if (*p >= '0' && *p <= '9')
1603                         *pn += (*p - '0') * l;
1604                 else
1605                         break;
1606         } while (l /= 10);
1607 }
1608
1609 /*
1610  * Parse GNU tar header
1611  */
1612 static int
1613 header_gnutar(struct archive_read *a, struct tar *tar,
1614     struct archive_entry *entry, const void *h)
1615 {
1616         const struct archive_entry_header_gnutar *header;
1617
1618         (void)a;
1619
1620         /*
1621          * GNU header is like POSIX ustar, except 'prefix' is
1622          * replaced with some other fields. This also means the
1623          * filename is stored as in old-style archives.
1624          */
1625
1626         /* Grab fields common to all tar variants. */
1627         header_common(a, tar, entry, h);
1628
1629         /* Copy filename over (to ensure null termination). */
1630         header = (const struct archive_entry_header_gnutar *)h;
1631         archive_strncpy(&(tar->entry_pathname), header->name,
1632             sizeof(header->name));
1633         archive_entry_copy_pathname(entry, tar->entry_pathname.s);
1634
1635         /* Fields common to ustar and GNU */
1636         /* XXX Can the following be factored out since it's common
1637          * to ustar and gnu tar?  Is it okay to move it down into
1638          * header_common, perhaps?  */
1639         archive_strncpy(&(tar->entry_uname),
1640             header->uname, sizeof(header->uname));
1641         archive_entry_copy_uname(entry, tar->entry_uname.s);
1642
1643         archive_strncpy(&(tar->entry_gname),
1644             header->gname, sizeof(header->gname));
1645         archive_entry_copy_gname(entry, tar->entry_gname.s);
1646
1647         /* Parse out device numbers only for char and block specials */
1648         if (header->typeflag[0] == '3' || header->typeflag[0] == '4') {
1649                 archive_entry_set_rdevmajor(entry,
1650                     tar_atol(header->rdevmajor, sizeof(header->rdevmajor)));
1651                 archive_entry_set_rdevminor(entry,
1652                     tar_atol(header->rdevminor, sizeof(header->rdevminor)));
1653         } else
1654                 archive_entry_set_rdev(entry, 0);
1655
1656         tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining);
1657
1658         /* Grab GNU-specific fields. */
1659         archive_entry_set_atime(entry,
1660             tar_atol(header->atime, sizeof(header->atime)), 0);
1661         archive_entry_set_ctime(entry,
1662             tar_atol(header->ctime, sizeof(header->ctime)), 0);
1663         if (header->realsize[0] != 0) {
1664                 tar->realsize
1665                     = tar_atol(header->realsize, sizeof(header->realsize));
1666                 archive_entry_set_size(entry, tar->realsize);
1667         }
1668
1669         if (header->sparse[0].offset[0] != 0) {
1670                 gnu_sparse_old_read(a, tar, header);
1671         } else {
1672                 if (header->isextended[0] != 0) {
1673                         /* XXX WTF? XXX */
1674                 }
1675         }
1676
1677         return (0);
1678 }
1679
1680 static void
1681 gnu_add_sparse_entry(struct tar *tar, off_t offset, off_t remaining)
1682 {
1683         struct sparse_block *p;
1684
1685         p = (struct sparse_block *)malloc(sizeof(*p));
1686         if (p == NULL)
1687                 __archive_errx(1, "Out of memory");
1688         memset(p, 0, sizeof(*p));
1689         if (tar->sparse_last != NULL)
1690                 tar->sparse_last->next = p;
1691         else
1692                 tar->sparse_list = p;
1693         tar->sparse_last = p;
1694         p->offset = offset;
1695         p->remaining = remaining;
1696 }
1697
1698 static void
1699 gnu_clear_sparse_list(struct tar *tar)
1700 {
1701         struct sparse_block *p;
1702
1703         while (tar->sparse_list != NULL) {
1704                 p = tar->sparse_list;
1705                 tar->sparse_list = p->next;
1706                 free(p);
1707         }
1708         tar->sparse_last = NULL;
1709 }
1710
1711 /*
1712  * GNU tar old-format sparse data.
1713  *
1714  * GNU old-format sparse data is stored in a fixed-field
1715  * format.  Offset/size values are 11-byte octal fields (same
1716  * format as 'size' field in ustart header).  These are
1717  * stored in the header, allocating subsequent header blocks
1718  * as needed.  Extending the header in this way is a pretty
1719  * severe POSIX violation; this design has earned GNU tar a
1720  * lot of criticism.
1721  */
1722
1723 static int
1724 gnu_sparse_old_read(struct archive_read *a, struct tar *tar,
1725     const struct archive_entry_header_gnutar *header)
1726 {
1727         ssize_t bytes_read;
1728         const void *data;
1729         struct extended {
1730                 struct gnu_sparse sparse[21];
1731                 char    isextended[1];
1732                 char    padding[7];
1733         };
1734         const struct extended *ext;
1735
1736         gnu_sparse_old_parse(tar, header->sparse, 4);
1737         if (header->isextended[0] == 0)
1738                 return (ARCHIVE_OK);
1739
1740         do {
1741                 bytes_read = (a->decompressor->read_ahead)(a, &data, 512);
1742                 if (bytes_read < 0)
1743                         return (ARCHIVE_FATAL);
1744                 if (bytes_read < 512) {
1745                         archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1746                             "Truncated tar archive "
1747                             "detected while reading sparse file data");
1748                         return (ARCHIVE_FATAL);
1749                 }
1750                 (a->decompressor->consume)(a, 512);
1751                 ext = (const struct extended *)data;
1752                 gnu_sparse_old_parse(tar, ext->sparse, 21);
1753         } while (ext->isextended[0] != 0);
1754         if (tar->sparse_list != NULL)
1755                 tar->entry_offset = tar->sparse_list->offset;
1756         return (ARCHIVE_OK);
1757 }
1758
1759 static void
1760 gnu_sparse_old_parse(struct tar *tar,
1761     const struct gnu_sparse *sparse, int length)
1762 {
1763         while (length > 0 && sparse->offset[0] != 0) {
1764                 gnu_add_sparse_entry(tar,
1765                     tar_atol(sparse->offset, sizeof(sparse->offset)),
1766                     tar_atol(sparse->numbytes, sizeof(sparse->numbytes)));
1767                 sparse++;
1768                 length--;
1769         }
1770 }
1771
1772 /*
1773  * GNU tar sparse format 0.0
1774  *
1775  * Beginning with GNU tar 1.15, sparse files are stored using
1776  * information in the pax extended header.  The GNU tar maintainers
1777  * have gone through a number of variations in the process of working
1778  * out this scheme; furtunately, they're all numbered.
1779  *
1780  * Sparse format 0.0 uses attribute GNU.sparse.numblocks to store the
1781  * number of blocks, and GNU.sparse.offset/GNU.sparse.numbytes to
1782  * store offset/size for each block.  The repeated instances of these
1783  * latter fields violate the pax specification (which frowns on
1784  * duplicate keys), so this format was quickly replaced.
1785  */
1786
1787 /*
1788  * GNU tar sparse format 0.1
1789  *
1790  * This version replaced the offset/numbytes attributes with
1791  * a single "map" attribute that stored a list of integers.  This
1792  * format had two problems: First, the "map" attribute could be very
1793  * long, which caused problems for some implementations.  More
1794  * importantly, the sparse data was lost when extracted by archivers
1795  * that didn't recognize this extension.
1796  */
1797
1798 static int
1799 gnu_sparse_01_parse(struct tar *tar, const char *p)
1800 {
1801         const char *e;
1802         off_t offset = -1, size = -1;
1803
1804         for (;;) {
1805                 e = p;
1806                 while (*e != '\0' && *e != ',') {
1807                         if (*e < '0' || *e > '9')
1808                                 return (ARCHIVE_WARN);
1809                         e++;
1810                 }
1811                 if (offset < 0) {
1812                         offset = tar_atol10(p, e - p);
1813                         if (offset < 0)
1814                                 return (ARCHIVE_WARN);
1815                 } else {
1816                         size = tar_atol10(p, e - p);
1817                         if (size < 0)
1818                                 return (ARCHIVE_WARN);
1819                         gnu_add_sparse_entry(tar, offset, size);
1820                         offset = -1;
1821                 }
1822                 if (*e == '\0')
1823                         return (ARCHIVE_OK);
1824                 p = e + 1;
1825         }
1826 }
1827
1828 /*
1829  * GNU tar sparse format 1.0
1830  *
1831  * The idea: The offset/size data is stored as a series of base-10
1832  * ASCII numbers prepended to the file data, so that dearchivers that
1833  * don't support this format will extract the block map along with the
1834  * data and a separate post-process can restore the sparseness.
1835  *
1836  * Unfortunately, GNU tar 1.16 had a bug that added unnecessary
1837  * padding to the body of the file when using this format.  GNU tar
1838  * 1.17 corrected this bug without bumping the version number, so
1839  * it's not possible to support both variants.  This code supports
1840  * the later variant at the expense of not supporting the former.
1841  *
1842  * This variant also replaced GNU.sparse.size with GNU.sparse.realsize
1843  * and introduced the GNU.sparse.major/GNU.sparse.minor attributes.
1844  */
1845
1846 /*
1847  * Read the next line from the input, and parse it as a decimal
1848  * integer followed by '\n'.  Returns positive integer value or
1849  * negative on error.
1850  */
1851 static int64_t
1852 gnu_sparse_10_atol(struct archive_read *a, struct tar *tar,
1853     ssize_t *remaining)
1854 {
1855         int64_t l, limit, last_digit_limit;
1856         const char *p;
1857         ssize_t bytes_read;
1858         int base, digit;
1859
1860         base = 10;
1861         limit = INT64_MAX / base;
1862         last_digit_limit = INT64_MAX % base;
1863
1864         /*
1865          * Skip any lines starting with '#'; GNU tar specs
1866          * don't require this, but they should.
1867          */
1868         do {
1869                 bytes_read = readline(a, tar, &p, tar_min(*remaining, 100));
1870                 if (bytes_read <= 0)
1871                         return (ARCHIVE_FATAL);
1872                 *remaining -= bytes_read;
1873         } while (p[0] == '#');
1874
1875         l = 0;
1876         while (bytes_read > 0) {
1877                 if (*p == '\n')
1878                         return (l);
1879                 if (*p < '0' || *p >= '0' + base)
1880                         return (ARCHIVE_WARN);
1881                 digit = *p - '0';
1882                 if (l > limit || (l == limit && digit > last_digit_limit))
1883                         l = UINT64_MAX; /* Truncate on overflow. */
1884                 else
1885                         l = (l * base) + digit;
1886                 p++;
1887                 bytes_read--;
1888         }
1889         /* TODO: Error message. */
1890         return (ARCHIVE_WARN);
1891 }
1892
1893 /*
1894  * Returns length (in bytes) of the sparse data description
1895  * that was read.
1896  */
1897 static ssize_t
1898 gnu_sparse_10_read(struct archive_read *a, struct tar *tar)
1899 {
1900         ssize_t remaining, bytes_read;
1901         int entries;
1902         off_t offset, size, to_skip;
1903
1904         /* Clear out the existing sparse list. */
1905         gnu_clear_sparse_list(tar);
1906
1907         remaining = tar->entry_bytes_remaining;
1908
1909         /* Parse entries. */
1910         entries = gnu_sparse_10_atol(a, tar, &remaining);
1911         if (entries < 0)
1912                 return (ARCHIVE_FATAL);
1913         /* Parse the individual entries. */
1914         while (entries-- > 0) {
1915                 /* Parse offset/size */
1916                 offset = gnu_sparse_10_atol(a, tar, &remaining);
1917                 if (offset < 0)
1918                         return (ARCHIVE_FATAL);
1919                 size = gnu_sparse_10_atol(a, tar, &remaining);
1920                 if (size < 0)
1921                         return (ARCHIVE_FATAL);
1922                 /* Add a new sparse entry. */
1923                 gnu_add_sparse_entry(tar, offset, size);
1924         }
1925         /* Skip rest of block... */
1926         bytes_read = tar->entry_bytes_remaining - remaining;
1927         to_skip = 0x1ff & -bytes_read;
1928         if (to_skip != (a->decompressor->skip)(a, to_skip))
1929                 return (ARCHIVE_FATAL);
1930         return (bytes_read + to_skip);
1931 }
1932
1933 /*-
1934  * Convert text->integer.
1935  *
1936  * Traditional tar formats (including POSIX) specify base-8 for
1937  * all of the standard numeric fields.  This is a significant limitation
1938  * in practice:
1939  *   = file size is limited to 8GB
1940  *   = rdevmajor and rdevminor are limited to 21 bits
1941  *   = uid/gid are limited to 21 bits
1942  *
1943  * There are two workarounds for this:
1944  *   = pax extended headers, which use variable-length string fields
1945  *   = GNU tar and STAR both allow either base-8 or base-256 in
1946  *      most fields.  The high bit is set to indicate base-256.
1947  *
1948  * On read, this implementation supports both extensions.
1949  */
1950 static int64_t
1951 tar_atol(const char *p, unsigned char_cnt)
1952 {
1953         /*
1954          * Technically, GNU tar considers a field to be in base-256
1955          * only if the first byte is 0xff or 0x80.
1956          */
1957         if (*p & 0x80)
1958                 return (tar_atol256(p, char_cnt));
1959         return (tar_atol8(p, char_cnt));
1960 }
1961
1962 /*
1963  * Note that this implementation does not (and should not!) obey
1964  * locale settings; you cannot simply substitute strtol here, since
1965  * it does obey locale.
1966  */
1967 static int64_t
1968 tar_atol8(const char *p, unsigned char_cnt)
1969 {
1970         int64_t l, limit, last_digit_limit;
1971         int digit, sign, base;
1972
1973         base = 8;
1974         limit = INT64_MAX / base;
1975         last_digit_limit = INT64_MAX % base;
1976
1977         while (*p == ' ' || *p == '\t')
1978                 p++;
1979         if (*p == '-') {
1980                 sign = -1;
1981                 p++;
1982         } else
1983                 sign = 1;
1984
1985         l = 0;
1986         digit = *p - '0';
1987         while (digit >= 0 && digit < base  && char_cnt-- > 0) {
1988                 if (l>limit || (l == limit && digit > last_digit_limit)) {
1989                         l = UINT64_MAX; /* Truncate on overflow. */
1990                         break;
1991                 }
1992                 l = (l * base) + digit;
1993                 digit = *++p - '0';
1994         }
1995         return (sign < 0) ? -l : l;
1996 }
1997
1998 /*
1999  * Note that this implementation does not (and should not!) obey
2000  * locale settings; you cannot simply substitute strtol here, since
2001  * it does obey locale.
2002  */
2003 static int64_t
2004 tar_atol10(const char *p, unsigned char_cnt)
2005 {
2006         int64_t l, limit, last_digit_limit;
2007         int base, digit, sign;
2008
2009         base = 10;
2010         limit = INT64_MAX / base;
2011         last_digit_limit = INT64_MAX % base;
2012
2013         while (*p == ' ' || *p == '\t')
2014                 p++;
2015         if (*p == '-') {
2016                 sign = -1;
2017                 p++;
2018         } else
2019                 sign = 1;
2020
2021         l = 0;
2022         digit = *p - '0';
2023         while (digit >= 0 && digit < base  && char_cnt-- > 0) {
2024                 if (l > limit || (l == limit && digit > last_digit_limit)) {
2025                         l = UINT64_MAX; /* Truncate on overflow. */
2026                         break;
2027                 }
2028                 l = (l * base) + digit;
2029                 digit = *++p - '0';
2030         }
2031         return (sign < 0) ? -l : l;
2032 }
2033
2034 /*
2035  * Parse a base-256 integer.  This is just a straight signed binary
2036  * value in big-endian order, except that the high-order bit is
2037  * ignored.
2038  */
2039 static int64_t
2040 tar_atol256(const char *_p, unsigned char_cnt)
2041 {
2042         int64_t l, upper_limit, lower_limit;
2043         const unsigned char *p = (const unsigned char *)_p;
2044
2045         upper_limit = INT64_MAX / 256;
2046         lower_limit = INT64_MIN / 256;
2047
2048         /* Pad with 1 or 0 bits, depending on sign. */
2049         if ((0x40 & *p) == 0x40)
2050                 l = (int64_t)-1;
2051         else
2052                 l = 0;
2053         l = (l << 6) | (0x3f & *p++);
2054         while (--char_cnt > 0) {
2055                 if (l > upper_limit) {
2056                         l = INT64_MAX; /* Truncate on overflow */
2057                         break;
2058                 } else if (l < lower_limit) {
2059                         l = INT64_MIN;
2060                         break;
2061                 }
2062                 l = (l << 8) | (0xff & (int64_t)*p++);
2063         }
2064         return (l);
2065 }
2066
2067 /*
2068  * Returns length of line (including trailing newline)
2069  * or negative on error.  'start' argument is updated to
2070  * point to first character of line.  This avoids copying
2071  * when possible.
2072  */
2073 static ssize_t
2074 readline(struct archive_read *a, struct tar *tar, const char **start,
2075     ssize_t limit)
2076 {
2077         ssize_t bytes_read;
2078         ssize_t total_size = 0;
2079         const void *t;
2080         const char *s;
2081         void *p;
2082
2083         bytes_read = (a->decompressor->read_ahead)(a, &t, 1);
2084         if (bytes_read <= 0)
2085                 return (ARCHIVE_FATAL);
2086         s = t;  /* Start of line? */
2087         p = memchr(t, '\n', bytes_read);
2088         /* If we found '\n' in the read buffer, return pointer to that. */
2089         if (p != NULL) {
2090                 bytes_read = 1 + ((const char *)p) - s;
2091                 if (bytes_read > limit) {
2092                         archive_set_error(&a->archive,
2093                             ARCHIVE_ERRNO_FILE_FORMAT,
2094                             "Line too long");
2095                         return (ARCHIVE_FATAL);
2096                 }
2097                 (a->decompressor->consume)(a, bytes_read);
2098                 *start = s;
2099                 return (bytes_read);
2100         }
2101         /* Otherwise, we need to accumulate in a line buffer. */
2102         for (;;) {
2103                 if (total_size + bytes_read > limit) {
2104                         archive_set_error(&a->archive,
2105                             ARCHIVE_ERRNO_FILE_FORMAT,
2106                             "Line too long");
2107                         return (ARCHIVE_FATAL);
2108                 }
2109                 if (archive_string_ensure(&tar->line, total_size + bytes_read) == NULL) {
2110                         archive_set_error(&a->archive, ENOMEM,
2111                             "Can't allocate working buffer");
2112                         return (ARCHIVE_FATAL);
2113                 }
2114                 memcpy(tar->line.s + total_size, t, bytes_read);
2115                 (a->decompressor->consume)(a, bytes_read);
2116                 total_size += bytes_read;
2117                 /* If we found '\n', clean up and return. */
2118                 if (p != NULL) {
2119                         *start = tar->line.s;
2120                         return (total_size);
2121                 }
2122                 /* Read some more. */
2123                 bytes_read = (a->decompressor->read_ahead)(a, &t, 1);
2124                 if (bytes_read <= 0)
2125                         return (ARCHIVE_FATAL);
2126                 s = t;  /* Start of line? */
2127                 p = memchr(t, '\n', bytes_read);
2128                 /* If we found '\n', trim the read. */
2129                 if (p != NULL) {
2130                         bytes_read = 1 + ((const char *)p) - s;
2131                 }
2132         }
2133 }
2134
2135 static wchar_t *
2136 utf8_decode(struct tar *tar, const char *src, size_t length)
2137 {
2138         wchar_t *dest;
2139         ssize_t n;
2140         int err;
2141
2142         /* Ensure pax_entry buffer is big enough. */
2143         if (tar->pax_entry_length <= length) {
2144                 wchar_t *old_entry = tar->pax_entry;
2145
2146                 if (tar->pax_entry_length <= 0)
2147                         tar->pax_entry_length = 1024;
2148                 while (tar->pax_entry_length <= length + 1)
2149                         tar->pax_entry_length *= 2;
2150
2151                 old_entry = tar->pax_entry;
2152                 tar->pax_entry = (wchar_t *)realloc(tar->pax_entry,
2153                     tar->pax_entry_length * sizeof(wchar_t));
2154                 if (tar->pax_entry == NULL) {
2155                         free(old_entry);
2156                         /* TODO: Handle this error. */
2157                         return (NULL);
2158                 }
2159         }
2160
2161         dest = tar->pax_entry;
2162         err = 0;
2163         while (length > 0) {
2164                 n = UTF8_mbrtowc(dest, src, length);
2165                 if (n < 0)
2166                         return (NULL);
2167                 if (n == 0)
2168                         break;
2169                 dest++;
2170                 src += n;
2171                 length -= n;
2172         }
2173         *dest++ = L'\0';
2174         return (tar->pax_entry);
2175 }
2176
2177 /*
2178  * Copied and simplified from FreeBSD libc/locale.
2179  */
2180 static ssize_t
2181 UTF8_mbrtowc(wchar_t *pwc, const char *s, size_t n)
2182 {
2183         int ch, i, len, mask;
2184         unsigned long wch;
2185
2186         if (s == NULL || n == 0 || pwc == NULL)
2187                 return (0);
2188
2189         /*
2190          * Determine the number of octets that make up this character from
2191          * the first octet, and a mask that extracts the interesting bits of
2192          * the first octet.
2193          */
2194         ch = (unsigned char)*s;
2195         if ((ch & 0x80) == 0) {
2196                 mask = 0x7f;
2197                 len = 1;
2198         } else if ((ch & 0xe0) == 0xc0) {
2199                 mask = 0x1f;
2200                 len = 2;
2201         } else if ((ch & 0xf0) == 0xe0) {
2202                 mask = 0x0f;
2203                 len = 3;
2204         } else if ((ch & 0xf8) == 0xf0) {
2205                 mask = 0x07;
2206                 len = 4;
2207         } else {
2208                 /* Invalid first byte. */
2209                 return (-1);
2210         }
2211
2212         if (n < (size_t)len) {
2213                 /* Valid first byte but truncated. */
2214                 return (-2);
2215         }
2216
2217         /*
2218          * Decode the octet sequence representing the character in chunks
2219          * of 6 bits, most significant first.
2220          */
2221         wch = (unsigned char)*s++ & mask;
2222         i = len;
2223         while (--i != 0) {
2224                 if ((*s & 0xc0) != 0x80) {
2225                         /* Invalid intermediate byte; consume one byte and
2226                          * emit '?' */
2227                         *pwc = '?';
2228                         return (1);
2229                 }
2230                 wch <<= 6;
2231                 wch |= *s++ & 0x3f;
2232         }
2233
2234         /* Assign the value to the output; out-of-range values
2235          * just get truncated. */
2236         *pwc = (wchar_t)wch;
2237 #ifdef WCHAR_MAX
2238         /*
2239          * If platform has WCHAR_MAX, we can do something
2240          * more sensible with out-of-range values.
2241          */
2242         if (wch >= WCHAR_MAX)
2243                 *pwc = '?';
2244 #endif
2245         /* Return number of bytes input consumed: 0 for end-of-string. */
2246         return (wch == L'\0' ? 0 : len);
2247 }
2248
2249
2250 /*
2251  * base64_decode - Base64 decode
2252  *
2253  * This accepts most variations of base-64 encoding, including:
2254  *    * with or without line breaks
2255  *    * with or without the final group padded with '=' or '_' characters
2256  * (The most economical Base-64 variant does not pad the last group and
2257  * omits line breaks; RFC1341 used for MIME requires both.)
2258  */
2259 static char *
2260 base64_decode(const char *s, size_t len, size_t *out_len)
2261 {
2262         static const unsigned char digits[64] = {
2263                 'A','B','C','D','E','F','G','H','I','J','K','L','M','N',
2264                 'O','P','Q','R','S','T','U','V','W','X','Y','Z','a','b',
2265                 'c','d','e','f','g','h','i','j','k','l','m','n','o','p',
2266                 'q','r','s','t','u','v','w','x','y','z','0','1','2','3',
2267                 '4','5','6','7','8','9','+','/' };
2268         static unsigned char decode_table[128];
2269         char *out, *d;
2270         const unsigned char *src = (const unsigned char *)s;
2271
2272         /* If the decode table is not yet initialized, prepare it. */
2273         if (decode_table[digits[1]] != 1) {
2274                 size_t i;
2275                 memset(decode_table, 0xff, sizeof(decode_table));
2276                 for (i = 0; i < sizeof(digits); i++)
2277                         decode_table[digits[i]] = i;
2278         }
2279
2280         /* Allocate enough space to hold the entire output. */
2281         /* Note that we may not use all of this... */
2282         out = (char *)malloc((len * 3 + 3) / 4);
2283         if (out == NULL) {
2284                 *out_len = 0;
2285                 return (NULL);
2286         }
2287         d = out;
2288
2289         while (len > 0) {
2290                 /* Collect the next group of (up to) four characters. */
2291                 int v = 0;
2292                 int group_size = 0;
2293                 while (group_size < 4 && len > 0) {
2294                         /* '=' or '_' padding indicates final group. */
2295                         if (*src == '=' || *src == '_') {
2296                                 len = 0;
2297                                 break;
2298                         }
2299                         /* Skip illegal characters (including line breaks) */
2300                         if (*src > 127 || *src < 32
2301                             || decode_table[*src] == 0xff) {
2302                                 len--;
2303                                 src++;
2304                                 continue;
2305                         }
2306                         v <<= 6;
2307                         v |= decode_table[*src++];
2308                         len --;
2309                         group_size++;
2310                 }
2311                 /* Align a short group properly. */
2312                 v <<= 6 * (4 - group_size);
2313                 /* Unpack the group we just collected. */
2314                 switch (group_size) {
2315                 case 4: d[2] = v & 0xff;
2316                         /* FALLTHROUGH */
2317                 case 3: d[1] = (v >> 8) & 0xff;
2318                         /* FALLTHROUGH */
2319                 case 2: d[0] = (v >> 16) & 0xff;
2320                         break;
2321                 case 1: /* this is invalid! */
2322                         break;
2323                 }
2324                 d += group_size * 3 / 4;
2325         }
2326
2327         *out_len = d - out;
2328         return (out);
2329 }
2330
2331 static char *
2332 url_decode(const char *in)
2333 {
2334         char *out, *d;
2335         const char *s;
2336
2337         out = (char *)malloc(strlen(in) + 1);
2338         if (out == NULL)
2339                 return (NULL);
2340         for (s = in, d = out; *s != '\0'; ) {
2341                 if (*s == '%') {
2342                         /* Try to convert % escape */
2343                         int digit1 = tohex(s[1]);
2344                         int digit2 = tohex(s[2]);
2345                         if (digit1 >= 0 && digit2 >= 0) {
2346                                 /* Looks good, consume three chars */
2347                                 s += 3;
2348                                 /* Convert output */
2349                                 *d++ = ((digit1 << 4) | digit2);
2350                                 continue;
2351                         }
2352                         /* Else fall through and treat '%' as normal char */
2353                 }
2354                 *d++ = *s++;
2355         }
2356         *d = '\0';
2357         return (out);
2358 }
2359
2360 static int
2361 tohex(int c)
2362 {
2363         if (c >= '0' && c <= '9')
2364                 return (c - '0');
2365         else if (c >= 'A' && c <= 'F')
2366                 return (c - 'A' + 10);
2367         else if (c >= 'a' && c <= 'f')
2368                 return (c - 'a' + 10);
2369         else
2370                 return (-1);
2371 }