lib/libarchive/archive_read_support_format_iso9660.c

   1 /*-
   2  * Copyright (c) 2003-2007 Tim Kientzle
   3  * Copyright (c) 2009 Andreas Henriksson <andreas@fatal.se>
   4  * Copyright (c) 2009 Michihiro NAKAJIMA
   5  * All rights reserved.
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions
   9  * are met:
  10  * 1. Redistributions of source code must retain the above copyright
  11  *    notice, this list of conditions and the following disclaimer.
  12  * 2. Redistributions in binary form must reproduce the above copyright
  13  *    notice, this list of conditions and the following disclaimer in the
  14  *    documentation and/or other materials provided with the distribution.
  15  *
  16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR
  17  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
  18  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
  19  * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT,
  20  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
  21  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
  22  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
  23  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  24  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
  25  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  26  */
  27
  28 #include "archive_platform.h"
  29 __FBSDID("$FreeBSD$");
  30
  31 #ifdef HAVE_ERRNO_H
  32 #include <errno.h>
  33 #endif
  34 /* #include <stdint.h> */ /* See archive_platform.h */
  35 #include <stdio.h>
  36 #ifdef HAVE_STDLIB_H
  37 #include <stdlib.h>
  38 #endif
  39 #ifdef HAVE_STRING_H
  40 #include <string.h>
  41 #endif
  42 #include <time.h>
  43 #ifdef HAVE_ZLIB_H
  44 #include <zlib.h>
  45 #endif
  46
  47 #include "archive.h"
  48 #include "archive_endian.h"
  49 #include "archive_entry.h"
  50 #include "archive_private.h"
  51 #include "archive_read_private.h"
  52 #include "archive_string.h"
  53
  54 /*
  55  * An overview of ISO 9660 format:
  56  *
  57  * Each disk is laid out as follows:
  58  *   * 32k reserved for private use
  59  *   * Volume descriptor table.  Each volume descriptor
  60  *     is 2k and specifies basic format information.
  61  *     The "Primary Volume Descriptor" (PVD) is defined by the
  62  *     standard and should always be present; other volume
  63  *     descriptors include various vendor-specific extensions.
  64  *   * Files and directories.  Each file/dir is specified by
  65  *     an "extent" (starting sector and length in bytes).
  66  *     Dirs are just files with directory records packed one
  67  *     after another.  The PVD contains a single dir entry
  68  *     specifying the location of the root directory.  Everything
  69  *     else follows from there.
  70  *
  71  * This module works by first reading the volume descriptors, then
  72  * building a list of directory entries, sorted by starting
  73  * sector.  At each step, I look for the earliest dir entry that
  74  * hasn't yet been read, seek forward to that location and read
  75  * that entry.  If it's a dir, I slurp in the new dir entries and
  76  * add them to the heap; if it's a regular file, I return the
  77  * corresponding archive_entry and wait for the client to request
  78  * the file body.  This strategy allows us to read most compliant
  79  * CDs with a single pass through the data, as required by libarchive.
  80  */
  81 #define LOGICAL_BLOCK_SIZE      2048
  82 #define SYSTEM_AREA_BLOCK       16
  83
  84 /* Structure of on-disk primary volume descriptor. */
  85 #define PVD_type_offset 0
  86 #define PVD_type_size 1
  87 #define PVD_id_offset (PVD_type_offset + PVD_type_size)
  88 #define PVD_id_size 5
  89 #define PVD_version_offset (PVD_id_offset + PVD_id_size)
  90 #define PVD_version_size 1
  91 #define PVD_reserved1_offset (PVD_version_offset + PVD_version_size)
  92 #define PVD_reserved1_size 1
  93 #define PVD_system_id_offset (PVD_reserved1_offset + PVD_reserved1_size)
  94 #define PVD_system_id_size 32
  95 #define PVD_volume_id_offset (PVD_system_id_offset + PVD_system_id_size)
  96 #define PVD_volume_id_size 32
  97 #define PVD_reserved2_offset (PVD_volume_id_offset + PVD_volume_id_size)
  98 #define PVD_reserved2_size 8
  99 #define PVD_volume_space_size_offset (PVD_reserved2_offset + PVD_reserved2_size)
 100 #define PVD_volume_space_size_size 8
 101 #define PVD_reserved3_offset (PVD_volume_space_size_offset + PVD_volume_space_size_size)
 102 #define PVD_reserved3_size 32
 103 #define PVD_volume_set_size_offset (PVD_reserved3_offset + PVD_reserved3_size)
 104 #define PVD_volume_set_size_size 4
 105 #define PVD_volume_sequence_number_offset (PVD_volume_set_size_offset + PVD_volume_set_size_size)
 106 #define PVD_volume_sequence_number_size 4
 107 #define PVD_logical_block_size_offset (PVD_volume_sequence_number_offset + PVD_volume_sequence_number_size)
 108 #define PVD_logical_block_size_size 4
 109 #define PVD_path_table_size_offset (PVD_logical_block_size_offset + PVD_logical_block_size_size)
 110 #define PVD_path_table_size_size 8
 111 #define PVD_type_1_path_table_offset (PVD_path_table_size_offset + PVD_path_table_size_size)
 112 #define PVD_type_1_path_table_size 4
 113 #define PVD_opt_type_1_path_table_offset (PVD_type_1_path_table_offset + PVD_type_1_path_table_size)
 114 #define PVD_opt_type_1_path_table_size 4
 115 #define PVD_type_m_path_table_offset (PVD_opt_type_1_path_table_offset + PVD_opt_type_1_path_table_size)
 116 #define PVD_type_m_path_table_size 4
 117 #define PVD_opt_type_m_path_table_offset (PVD_type_m_path_table_offset + PVD_type_m_path_table_size)
 118 #define PVD_opt_type_m_path_table_size 4
 119 #define PVD_root_directory_record_offset (PVD_opt_type_m_path_table_offset + PVD_opt_type_m_path_table_size)
 120 #define PVD_root_directory_record_size 34
 121 #define PVD_volume_set_id_offset (PVD_root_directory_record_offset + PVD_root_directory_record_size)
 122 #define PVD_volume_set_id_size 128
 123 #define PVD_publisher_id_offset (PVD_volume_set_id_offset + PVD_volume_set_id_size)
 124 #define PVD_publisher_id_size 128
 125 #define PVD_preparer_id_offset (PVD_publisher_id_offset + PVD_publisher_id_size)
 126 #define PVD_preparer_id_size 128
 127 #define PVD_application_id_offset (PVD_preparer_id_offset + PVD_preparer_id_size)
 128 #define PVD_application_id_size 128
 129 #define PVD_copyright_file_id_offset (PVD_application_id_offset + PVD_application_id_size)
 130 #define PVD_copyright_file_id_size 37
 131 #define PVD_abstract_file_id_offset (PVD_copyright_file_id_offset + PVD_copyright_file_id_size)
 132 #define PVD_abstract_file_id_size 37
 133 #define PVD_bibliographic_file_id_offset (PVD_abstract_file_id_offset + PVD_abstract_file_id_size)
 134 #define PVD_bibliographic_file_id_size 37
 135 #define PVD_creation_date_offset (PVD_bibliographic_file_id_offset + PVD_bibliographic_file_id_size)
 136 #define PVD_creation_date_size 17
 137 #define PVD_modification_date_offset (PVD_creation_date_offset + PVD_creation_date_size)
 138 #define PVD_modification_date_size 17
 139 #define PVD_expiration_date_offset (PVD_modification_date_offset + PVD_modification_date_size)
 140 #define PVD_expiration_date_size 17
 141 #define PVD_effective_date_offset (PVD_expiration_date_offset + PVD_expiration_date_size)
 142 #define PVD_effective_date_size 17
 143 #define PVD_file_structure_version_offset (PVD_effective_date_offset + PVD_effective_date_size)
 144 #define PVD_file_structure_version_size 1
 145 #define PVD_reserved4_offset (PVD_file_structure_version_offset + PVD_file_structure_version_size)
 146 #define PVD_reserved4_size 1
 147 #define PVD_application_data_offset (PVD_reserved4_offset + PVD_reserved4_size)
 148 #define PVD_application_data_size 512
 149 #define PVD_reserved5_offset (PVD_application_data_offset + PVD_application_data_size)
 150 #define PVD_reserved5_size (2048 - PVD_reserved5_offset)
 151
 152 /* TODO: It would make future maintenance easier to just hardcode the
 153  * above values.  In particular, ECMA119 states the offsets as part of
 154  * the standard.  That would eliminate the need for the following check.*/
 155 #if PVD_reserved5_offset != 1395
 156 #error PVD offset and size definitions are wrong.
 157 #endif
 158
 159
 160 /* Structure of optional on-disk supplementary volume descriptor. */
 161 #define SVD_type_offset 0
 162 #define SVD_type_size 1
 163 #define SVD_id_offset (SVD_type_offset + SVD_type_size)
 164 #define SVD_id_size 5
 165 #define SVD_version_offset (SVD_id_offset + SVD_id_size)
 166 #define SVD_version_size 1
 167 /* ... */
 168 #define SVD_reserved1_offset    72
 169 #define SVD_reserved1_size      8
 170 #define SVD_volume_space_size_offset 80
 171 #define SVD_volume_space_size_size 8
 172 #define SVD_escape_sequences_offset (SVD_volume_space_size_offset + SVD_volume_space_size_size)
 173 #define SVD_escape_sequences_size 32
 174 /* ... */
 175 #define SVD_logical_block_size_offset 128
 176 #define SVD_logical_block_size_size 4
 177 #define SVD_type_L_path_table_offset 140
 178 #define SVD_type_M_path_table_offset 148
 179 /* ... */
 180 #define SVD_root_directory_record_offset 156
 181 #define SVD_root_directory_record_size 34
 182 #define SVD_file_structure_version_offset 881
 183 #define SVD_reserved2_offset    882
 184 #define SVD_reserved2_size      1
 185 #define SVD_reserved3_offset    1395
 186 #define SVD_reserved3_size      653
 187 /* ... */
 188 /* FIXME: validate correctness of last SVD entry offset. */
 189
 190 /* Structure of an on-disk directory record. */
 191 /* Note:  ISO9660 stores each multi-byte integer twice, once in
 192  * each byte order.  The sizes here are the size of just one
 193  * of the two integers.  (This is why the offset of a field isn't
 194  * the same as the offset+size of the previous field.) */
 195 #define DR_length_offset 0
 196 #define DR_length_size 1
 197 #define DR_ext_attr_length_offset 1
 198 #define DR_ext_attr_length_size 1
 199 #define DR_extent_offset 2
 200 #define DR_extent_size 4
 201 #define DR_size_offset 10
 202 #define DR_size_size 4
 203 #define DR_date_offset 18
 204 #define DR_date_size 7
 205 #define DR_flags_offset 25
 206 #define DR_flags_size 1
 207 #define DR_file_unit_size_offset 26
 208 #define DR_file_unit_size_size 1
 209 #define DR_interleave_offset 27
 210 #define DR_interleave_size 1
 211 #define DR_volume_sequence_number_offset 28
 212 #define DR_volume_sequence_number_size 2
 213 #define DR_name_len_offset 32
 214 #define DR_name_len_size 1
 215 #define DR_name_offset 33
 216
 217 #ifdef HAVE_ZLIB_H
 218 static const unsigned char zisofs_magic[8] = {
 219         0x37, 0xE4, 0x53, 0x96, 0xC9, 0xDB, 0xD6, 0x07
 220 };
 221
 222 struct zisofs {
 223         /* Set 1 if this file compressed by paged zlib */
 224         int              pz;
 225         int              pz_log2_bs; /* Log2 of block size */
 226         uint64_t         pz_uncompressed_size;
 227
 228         int              initialized;
 229         unsigned char   *uncompressed_buffer;
 230         size_t           uncompressed_buffer_size;
 231
 232         uint32_t         pz_offset;
 233         unsigned char    header[16];
 234         size_t           header_avail;
 235         int              header_passed;
 236         unsigned char   *block_pointers;
 237         size_t           block_pointers_alloc;
 238         size_t           block_pointers_size;
 239         size_t           block_pointers_avail;
 240         size_t           block_off;
 241         uint32_t         block_avail;
 242
 243         z_stream         stream;
 244         int              stream_valid;
 245 };
 246 #else
 247 struct zisofs {
 248         /* Set 1 if this file compressed by paged zlib */
 249         int              pz;
 250 };
 251 #endif
 252
 253 struct content {
 254         uint64_t         offset;/* Offset on disk.              */
 255         uint64_t         size;  /* File size in bytes.          */
 256         struct content  *next;
 257 };
 258
 259 /* In-memory storage for a directory record. */
 260 struct file_info {
 261         struct file_info        *use_next;
 262         struct file_info        *parent;
 263         struct file_info        *next;
 264         int              subdirs;
 265         uint64_t         key;           /* Heap Key.                    */
 266         uint64_t         offset;        /* Offset on disk.              */
 267         uint64_t         size;          /* File size in bytes.          */
 268         uint32_t         ce_offset;     /* Offset of CE.                */
 269         uint32_t         ce_size;       /* Size of CE.                  */
 270         char             re;            /* Having RRIP "RE" extension.  */
 271         uint64_t         cl_offset;     /* Having RRIP "CL" extension.  */
 272         int              birthtime_is_set;
 273         time_t           birthtime;     /* File created time.           */
 274         time_t           mtime;         /* File last modified time.     */
 275         time_t           atime;         /* File last accessed time.     */
 276         time_t           ctime;         /* File attribute change time.  */
 277         uint64_t         rdev;          /* Device number.               */
 278         mode_t           mode;
 279         uid_t            uid;
 280         gid_t            gid;
 281         int64_t          number;
 282         int              nlinks;
 283         struct archive_string name; /* Pathname */
 284         char             name_continues; /* Non-zero if name continues */
 285         struct archive_string symlink;
 286         char             symlink_continues; /* Non-zero if link continues */
 287         /* Set 1 if this file compressed by paged zlib(zisofs) */
 288         int              pz;
 289         int              pz_log2_bs; /* Log2 of block size */
 290         uint64_t         pz_uncompressed_size;
 291         /* Set 1 if this file is multi extent. */
 292         int              multi_extent;
 293         struct {
 294                 struct content  *first;
 295                 struct content  **last;
 296         } contents;
 297         char             exposed;
 298 };
 299
 300 struct heap_queue {
 301         struct file_info **files;
 302         int              allocated;
 303         int              used;
 304 };
 305
 306 struct iso9660 {
 307         int     magic;
 308 #define ISO9660_MAGIC   0x96609660
 309
 310         int opt_support_joliet;
 311         int opt_support_rockridge;
 312
 313         struct archive_string pathname;
 314         char    seenRockridge;  /* Set true if RR extensions are used. */
 315         char    seenSUSP;       /* Set true if SUSP is beging used. */
 316         char    seenJoliet;
 317
 318         unsigned char   suspOffset;
 319         struct file_info *rr_moved;
 320         struct heap_queue                re_dirs;
 321         struct heap_queue                cl_files;
 322         struct read_ce_queue {
 323                 struct read_ce_req {
 324                         uint64_t         offset;/* Offset of CE on disk. */
 325                         struct file_info *file;
 326                 }               *reqs;
 327                 int              cnt;
 328                 int              allocated;
 329         }       read_ce_req;
 330
 331         int64_t         previous_number;
 332         struct archive_string previous_pathname;
 333
 334         struct file_info                *use_files;
 335         struct heap_queue                pending_files;
 336         struct {
 337                 struct file_info        *first;
 338                 struct file_info        **last;
 339         }       cache_files;
 340
 341         uint64_t current_position;
 342         ssize_t logical_block_size;
 343         uint64_t volume_size; /* Total size of volume in bytes. */
 344         int32_t  volume_block;/* Total size of volume in logical blocks. */
 345
 346         struct vd {
 347                 int             location;       /* Location of Extent.  */
 348                 uint32_t        size;
 349         } primary, joliet;
 350
 351         off_t   entry_sparse_offset;
 352         int64_t entry_bytes_remaining;
 353         struct zisofs    entry_zisofs;
 354         struct content  *entry_content;
 355 };
 356
 357 static int      archive_read_format_iso9660_bid(struct archive_read *);
 358 static int      archive_read_format_iso9660_options(struct archive_read *,
 359                     const char *, const char *);
 360 static int      archive_read_format_iso9660_cleanup(struct archive_read *);
 361 static int      archive_read_format_iso9660_read_data(struct archive_read *,
 362                     const void **, size_t *, off_t *);
 363 static int      archive_read_format_iso9660_read_data_skip(struct archive_read *);
 364 static int      archive_read_format_iso9660_read_header(struct archive_read *,
 365                     struct archive_entry *);
 366 static const char *build_pathname(struct archive_string *, struct file_info *);
 367 #if DEBUG
 368 static void     dump_isodirrec(FILE *, const unsigned char *isodirrec);
 369 #endif
 370 static time_t   time_from_tm(struct tm *);
 371 static time_t   isodate17(const unsigned char *);
 372 static time_t   isodate7(const unsigned char *);
 373 static int      isBootRecord(struct iso9660 *, const unsigned char *);
 374 static int      isVolumePartition(struct iso9660 *, const unsigned char *);
 375 static int      isVDSetTerminator(struct iso9660 *, const unsigned char *);
 376 static int      isJolietSVD(struct iso9660 *, const unsigned char *);
 377 static int      isSVD(struct iso9660 *, const unsigned char *);
 378 static int      isEVD(struct iso9660 *, const unsigned char *);
 379 static int      isPVD(struct iso9660 *, const unsigned char *);
 380 static struct file_info *next_cache_entry(struct iso9660 *iso9660);
 381 static int      next_entry_seek(struct archive_read *a, struct iso9660 *iso9660,
 382                     struct file_info **pfile);
 383 static struct file_info *
 384                 parse_file_info(struct archive_read *a,
 385                     struct file_info *parent, const unsigned char *isodirrec);
 386 static int      parse_rockridge(struct archive_read *a,
 387                     struct file_info *file, const unsigned char *start,
 388                     const unsigned char *end);
 389 static int      register_CE(struct archive_read *a, int32_t location,
 390                     struct file_info *file);
 391 static int      read_CE(struct archive_read *a, struct iso9660 *iso9660);
 392 static void     parse_rockridge_NM1(struct file_info *,
 393                     const unsigned char *, int);
 394 static void     parse_rockridge_SL1(struct file_info *,
 395                     const unsigned char *, int);
 396 static void     parse_rockridge_TF1(struct file_info *,
 397                     const unsigned char *, int);
 398 static void     parse_rockridge_ZF1(struct file_info *,
 399                     const unsigned char *, int);
 400 static void     register_file(struct iso9660 *, struct file_info *);
 401 static void     release_files(struct iso9660 *);
 402 static unsigned toi(const void *p, int n);
 403 static inline void cache_add_entry(struct iso9660 *iso9660,
 404                     struct file_info *file);
 405 static inline void cache_add_to_next_of_parent(struct iso9660 *iso9660,
 406                     struct file_info *file);
 407 static inline struct file_info *cache_get_entry(struct iso9660 *iso9660);
 408 static void     heap_add_entry(struct heap_queue *heap,
 409                     struct file_info *file, uint64_t key);
 410 static struct file_info *heap_get_entry(struct heap_queue *heap);
 411
 412 #define add_entry(iso9660, file)        \
 413         heap_add_entry(&((iso9660)->pending_files), file, file->offset)
 414 #define next_entry(iso9660)             \
 415         heap_get_entry(&((iso9660)->pending_files))
 416
 417 int
 418 archive_read_support_format_iso9660(struct archive *_a)
 419 {
 420         struct archive_read *a = (struct archive_read *)_a;
 421         struct iso9660 *iso9660;
 422         int r;
 423
 424         iso9660 = (struct iso9660 *)malloc(sizeof(*iso9660));
 425         if (iso9660 == NULL) {
 426                 archive_set_error(&a->archive, ENOMEM, "Can't allocate iso9660 data");
 427                 return (ARCHIVE_FATAL);
 428         }
 429         memset(iso9660, 0, sizeof(*iso9660));
 430         iso9660->magic = ISO9660_MAGIC;
 431         iso9660->cache_files.first = NULL;
 432         iso9660->cache_files.last = &(iso9660->cache_files.first);
 433         /* Enable to support Joliet extensions by default.      */
 434         iso9660->opt_support_joliet = 1;
 435         /* Enable to support Rock Ridge extensions by default.  */
 436         iso9660->opt_support_rockridge = 1;
 437
 438         r = __archive_read_register_format(a,
 439             iso9660,
 440             "iso9660",
 441             archive_read_format_iso9660_bid,
 442             archive_read_format_iso9660_options,
 443             archive_read_format_iso9660_read_header,
 444             archive_read_format_iso9660_read_data,
 445             archive_read_format_iso9660_read_data_skip,
 446             archive_read_format_iso9660_cleanup);
 447
 448         if (r != ARCHIVE_OK) {
 449                 free(iso9660);
 450                 return (r);
 451         }
 452         return (ARCHIVE_OK);
 453 }
 454
 455
 456 static int
 457 archive_read_format_iso9660_bid(struct archive_read *a)
 458 {
 459         struct iso9660 *iso9660;
 460         ssize_t bytes_read;
 461         const void *h;
 462         const unsigned char *p;
 463         int seenTerminator;
 464
 465         iso9660 = (struct iso9660 *)(a->format->data);
 466
 467         /*
 468          * Skip the first 32k (reserved area) and get the first
 469          * 8 sectors of the volume descriptor table.  Of course,
 470          * if the I/O layer gives us more, we'll take it.
 471          */
 472 #define RESERVED_AREA   (SYSTEM_AREA_BLOCK * LOGICAL_BLOCK_SIZE)
 473         h = __archive_read_ahead(a,
 474             RESERVED_AREA + 8 * LOGICAL_BLOCK_SIZE,
 475             &bytes_read);
 476         if (h == NULL)
 477             return (-1);
 478         p = (const unsigned char *)h;
 479
 480         /* Skip the reserved area. */
 481         bytes_read -= RESERVED_AREA;
 482         p += RESERVED_AREA;
 483
 484         /* Check each volume descriptor. */
 485         seenTerminator = 0;
 486         for (; bytes_read > LOGICAL_BLOCK_SIZE;
 487             bytes_read -= LOGICAL_BLOCK_SIZE, p += LOGICAL_BLOCK_SIZE) {
 488                 /* Do not handle undefined Volume Descriptor Type. */
 489                 if (p[0] >= 4 && p[0] <= 254)
 490                         return (0);
 491                 /* Standard Identifier must be "CD001" */
 492                 if (memcmp(p + 1, "CD001", 5) != 0)
 493                         return (0);
 494                 if (!iso9660->primary.location) {
 495                         if (isPVD(iso9660, p))
 496                                 continue;
 497                 }
 498                 if (!iso9660->joliet.location) {
 499                         if (isJolietSVD(iso9660, p))
 500                                 continue;
 501                 }
 502                 if (isBootRecord(iso9660, p))
 503                         continue;
 504                 if (isEVD(iso9660, p))
 505                         continue;
 506                 if (isSVD(iso9660, p))
 507                         continue;
 508                 if (isVolumePartition(iso9660, p))
 509                         continue;
 510                 if (isVDSetTerminator(iso9660, p)) {
 511                         seenTerminator = 1;
 512                         break;
 513                 }
 514                 return (0);
 515         }
 516         /*
 517          * ISO 9660 format must have Primary Volume Descriptor and
 518          * Volume Descriptor Set Terminator.
 519          */
 520         if (seenTerminator && iso9660->primary.location > 16)
 521                 return (48);
 522
 523         /* We didn't find a valid PVD; return a bid of zero. */
 524         return (0);
 525 }
 526
 527 static int
 528 archive_read_format_iso9660_options(struct archive_read *a,
 529                 const char *key, const char *val)
 530 {
 531         struct iso9660 *iso9660;
 532
 533         iso9660 = (struct iso9660 *)(a->format->data);
 534
 535         if (strcmp(key, "joliet") == 0) {
 536                 if (val == NULL || strcmp(val, "off") == 0 ||
 537                                 strcmp(val, "ignore") == 0 ||
 538                                 strcmp(val, "disable") == 0 ||
 539                                 strcmp(val, "0") == 0)
 540                         iso9660->opt_support_joliet = 0;
 541                 else
 542                         iso9660->opt_support_joliet = 1;
 543                 return (ARCHIVE_OK);
 544         }
 545         if (strcmp(key, "rockridge") == 0 ||
 546             strcmp(key, "Rockridge") == 0) {
 547                 iso9660->opt_support_rockridge = val != NULL;
 548                 return (ARCHIVE_OK);
 549         }
 550
 551         /* Note: The "warn" return is just to inform the options
 552          * supervisor that we didn't handle it.  It will generate
 553          * a suitable error if noone used this option. */
 554         return (ARCHIVE_WARN);
 555 }
 556
 557 static int
 558 isBootRecord(struct iso9660 *iso9660, const unsigned char *h)
 559 {
 560         (void)iso9660; /* UNUSED */
 561
 562         /* Type of the Volume Descriptor Boot Record must be 0. */
 563         if (h[0] != 0)
 564                 return (0);
 565
 566         /* Volume Descriptor Version must be 1. */
 567         if (h[6] != 1)
 568                 return (0);
 569
 570         return (1);
 571 }
 572
 573 static int
 574 isVolumePartition(struct iso9660 *iso9660, const unsigned char *h)
 575 {
 576         int32_t location;
 577
 578         /* Type of the Volume Partition Descriptor must be 3. */
 579         if (h[0] != 3)
 580                 return (0);
 581
 582         /* Volume Descriptor Version must be 1. */
 583         if (h[6] != 1)
 584                 return (0);
 585         /* Unused Field */
 586         if (h[7] != 0)
 587                 return (0);
 588
 589         location = archive_le32dec(h + 72);
 590         if (location <= SYSTEM_AREA_BLOCK ||
 591             location >= iso9660->volume_block)
 592                 return (0);
 593         if ((uint32_t)location != archive_be32dec(h + 76))
 594                 return (0);
 595
 596         return (1);
 597 }
 598
 599 static int
 600 isVDSetTerminator(struct iso9660 *iso9660, const unsigned char *h)
 601 {
 602         int i;
 603
 604         (void)iso9660; /* UNUSED */
 605
 606         /* Type of the Volume Descriptor Set Terminator must be 255. */
 607         if (h[0] != 255)
 608                 return (0);
 609
 610         /* Volume Descriptor Version must be 1. */
 611         if (h[6] != 1)
 612                 return (0);
 613
 614         /* Reserved field must be 0. */
 615         for (i = 7; i < 2048; ++i)
 616                 if (h[i] != 0)
 617                         return (0);
 618
 619         return (1);
 620 }
 621
 622 static int
 623 isJolietSVD(struct iso9660 *iso9660, const unsigned char *h)
 624 {
 625         const unsigned char *p;
 626         ssize_t logical_block_size;
 627         int32_t volume_block;
 628
 629         /* Check if current sector is a kind of Supplementary Volume
 630          * Descriptor. */
 631         if (!isSVD(iso9660, h))
 632                 return (0);
 633
 634         /* FIXME: do more validations according to joliet spec. */
 635
 636         /* check if this SVD contains joliet extension! */
 637         p = h + SVD_escape_sequences_offset;
 638         /* N.B. Joliet spec says p[1] == '\\', but.... */
 639         if (p[0] == '%' && p[1] == '/') {
 640                 int level = 0;
 641
 642                 if (p[2] == '@')
 643                         level = 1;
 644                 else if (p[2] == 'C')
 645                         level = 2;
 646                 else if (p[2] == 'E')
 647                         level = 3;
 648                 else /* not joliet */
 649                         return (0);
 650
 651                 iso9660->seenJoliet = level;
 652
 653         } else /* not joliet */
 654                 return (0);
 655
 656         logical_block_size =
 657             archive_le16dec(h + SVD_logical_block_size_offset);
 658         volume_block = archive_le32dec(h + SVD_volume_space_size_offset);
 659
 660         iso9660->logical_block_size = logical_block_size;
 661         iso9660->volume_block = volume_block;
 662         iso9660->volume_size = logical_block_size * (uint64_t)volume_block;
 663         /* Read Root Directory Record in Volume Descriptor. */
 664         p = h + SVD_root_directory_record_offset;
 665         iso9660->joliet.location = archive_le32dec(p + DR_extent_offset);
 666         iso9660->joliet.size = archive_le32dec(p + DR_size_offset);
 667
 668         return (48);
 669 }
 670
 671 static int
 672 isSVD(struct iso9660 *iso9660, const unsigned char *h)
 673 {
 674         const unsigned char *p;
 675         ssize_t logical_block_size;
 676         int32_t volume_block;
 677         int32_t location;
 678         int i;
 679
 680         (void)iso9660; /* UNUSED */
 681
 682         /* Type 2 means it's a SVD. */
 683         if (h[SVD_type_offset] != 2)
 684                 return (0);
 685
 686         /* Reserved field must be 0. */
 687         for (i = 0; i < SVD_reserved1_size; ++i)
 688                 if (h[SVD_reserved1_offset + i] != 0)
 689                         return (0);
 690         for (i = 0; i < SVD_reserved2_size; ++i)
 691                 if (h[SVD_reserved2_offset + i] != 0)
 692                         return (0);
 693         for (i = 0; i < SVD_reserved3_size; ++i)
 694                 if (h[SVD_reserved3_offset + i] != 0)
 695                         return (0);
 696
 697         /* File structure version must be 1 for ISO9660/ECMA119. */
 698         if (h[SVD_file_structure_version_offset] != 1)
 699                 return (0);
 700
 701         logical_block_size =
 702             archive_le16dec(h + SVD_logical_block_size_offset);
 703         if (logical_block_size <= 0)
 704                 return (0);
 705
 706         volume_block = archive_le32dec(h + SVD_volume_space_size_offset);
 707         if (volume_block <= SYSTEM_AREA_BLOCK+4)
 708                 return (0);
 709
 710         /* Location of Occurrence of Type L Path Table must be
 711          * available location,
 712          * > SYSTEM_AREA_BLOCK(16) + 2 and < Volume Space Size. */
 713         location = archive_le32dec(h+SVD_type_L_path_table_offset);
 714         if (location <= SYSTEM_AREA_BLOCK+2 || location >= volume_block)
 715                 return (0);
 716
 717         /* Location of Occurrence of Type M Path Table must be
 718          * available location,
 719          * > SYSTEM_AREA_BLOCK(16) + 2 and < Volume Space Size. */
 720         location = archive_be32dec(h+SVD_type_M_path_table_offset);
 721         if (location <= SYSTEM_AREA_BLOCK+2 || location >= volume_block)
 722                 return (0);
 723
 724         /* Read Root Directory Record in Volume Descriptor. */
 725         p = h + SVD_root_directory_record_offset;
 726         if (p[DR_length_offset] != 34)
 727                 return (0);
 728
 729         return (48);
 730 }
 731
 732 static int
 733 isEVD(struct iso9660 *iso9660, const unsigned char *h)
 734 {
 735         const unsigned char *p;
 736         ssize_t logical_block_size;
 737         int32_t volume_block;
 738         int32_t location;
 739         int i;
 740
 741         (void)iso9660; /* UNUSED */
 742
 743         /* Type of the Enhanced Volume Descriptor must be 2. */
 744         if (h[PVD_type_offset] != 2)
 745                 return (0);
 746
 747         /* EVD version must be 2. */
 748         if (h[PVD_version_offset] != 2)
 749                 return (0);
 750
 751         /* Reserved field must be 0. */
 752         if (h[PVD_reserved1_offset] != 0)
 753                 return (0);
 754
 755         /* Reserved field must be 0. */
 756         for (i = 0; i < PVD_reserved2_size; ++i)
 757                 if (h[PVD_reserved2_offset + i] != 0)
 758                         return (0);
 759
 760         /* Reserved field must be 0. */
 761         for (i = 0; i < PVD_reserved3_size; ++i)
 762                 if (h[PVD_reserved3_offset + i] != 0)
 763                         return (0);
 764
 765         /* Logical block size must be > 0. */
 766         /* I've looked at Ecma 119 and can't find any stronger
 767          * restriction on this field. */
 768         logical_block_size =
 769             archive_le16dec(h + PVD_logical_block_size_offset);
 770         if (logical_block_size <= 0)
 771                 return (0);
 772
 773         volume_block =
 774             archive_le32dec(h + PVD_volume_space_size_offset);
 775         if (volume_block <= SYSTEM_AREA_BLOCK+4)
 776                 return (0);
 777
 778         /* File structure version must be 2 for ISO9660:1999. */
 779         if (h[PVD_file_structure_version_offset] != 2)
 780                 return (0);
 781
 782         /* Location of Occurrence of Type L Path Table must be
 783          * available location,
 784          * > SYSTEM_AREA_BLOCK(16) + 2 and < Volume Space Size. */
 785         location = archive_le32dec(h+PVD_type_1_path_table_offset);
 786         if (location <= SYSTEM_AREA_BLOCK+2 || location >= volume_block)
 787                 return (0);
 788
 789         /* Location of Occurrence of Type M Path Table must be
 790          * available location,
 791          * > SYSTEM_AREA_BLOCK(16) + 2 and < Volume Space Size. */
 792         location = archive_be32dec(h+PVD_type_m_path_table_offset);
 793         if (location <= SYSTEM_AREA_BLOCK+2 || location >= volume_block)
 794                 return (0);
 795
 796         /* Reserved field must be 0. */
 797         for (i = 0; i < PVD_reserved4_size; ++i)
 798                 if (h[PVD_reserved4_offset + i] != 0)
 799                         return (0);
 800
 801         /* Reserved field must be 0. */
 802         for (i = 0; i < PVD_reserved5_size; ++i)
 803                 if (h[PVD_reserved5_offset + i] != 0)
 804                         return (0);
 805
 806         /* Read Root Directory Record in Volume Descriptor. */
 807         p = h + PVD_root_directory_record_offset;
 808         if (p[DR_length_offset] != 34)
 809                 return (0);
 810
 811         return (48);
 812 }
 813
 814 static int
 815 isPVD(struct iso9660 *iso9660, const unsigned char *h)
 816 {
 817         const unsigned char *p;
 818         ssize_t logical_block_size;
 819         int32_t volume_block;
 820         int32_t location;
 821         int i;
 822
 823         /* Type of the Primary Volume Descriptor must be 1. */
 824         if (h[PVD_type_offset] != 1)
 825                 return (0);
 826
 827         /* PVD version must be 1. */
 828         if (h[PVD_version_offset] != 1)
 829                 return (0);
 830
 831         /* Reserved field must be 0. */
 832         if (h[PVD_reserved1_offset] != 0)
 833                 return (0);
 834
 835         /* Reserved field must be 0. */
 836         for (i = 0; i < PVD_reserved2_size; ++i)
 837                 if (h[PVD_reserved2_offset + i] != 0)
 838                         return (0);
 839
 840         /* Reserved field must be 0. */
 841         for (i = 0; i < PVD_reserved3_size; ++i)
 842                 if (h[PVD_reserved3_offset + i] != 0)
 843                         return (0);
 844
 845         /* Logical block size must be > 0. */
 846         /* I've looked at Ecma 119 and can't find any stronger
 847          * restriction on this field. */
 848         logical_block_size =
 849             archive_le16dec(h + PVD_logical_block_size_offset);
 850         if (logical_block_size <= 0)
 851                 return (0);
 852
 853         volume_block = archive_le32dec(h + PVD_volume_space_size_offset);
 854         if (volume_block <= SYSTEM_AREA_BLOCK+4)
 855                 return (0);
 856
 857         /* File structure version must be 1 for ISO9660/ECMA119. */
 858         if (h[PVD_file_structure_version_offset] != 1)
 859                 return (0);
 860
 861         /* Location of Occurrence of Type L Path Table must be
 862          * available location,
 863          * > SYSTEM_AREA_BLOCK(16) + 2 and < Volume Space Size. */
 864         location = archive_le32dec(h+PVD_type_1_path_table_offset);
 865         if (location <= SYSTEM_AREA_BLOCK+2 || location >= volume_block)
 866                 return (0);
 867
 868         /* Location of Occurrence of Type M Path Table must be
 869          * available location,
 870          * > SYSTEM_AREA_BLOCK(16) + 2 and < Volume Space Size. */
 871         location = archive_be32dec(h+PVD_type_m_path_table_offset);
 872         if (location <= SYSTEM_AREA_BLOCK+2 || location >= volume_block)
 873                 return (0);
 874
 875         /* Reserved field must be 0. */
 876         for (i = 0; i < PVD_reserved4_size; ++i)
 877                 if (h[PVD_reserved4_offset + i] != 0)
 878                         return (0);
 879
 880         /* Reserved field must be 0. */
 881         for (i = 0; i < PVD_reserved5_size; ++i)
 882                 if (h[PVD_reserved5_offset + i] != 0)
 883                         return (0);
 884
 885         /* XXX TODO: Check other values for sanity; reject more
 886          * malformed PVDs. XXX */
 887
 888         /* Read Root Directory Record in Volume Descriptor. */
 889         p = h + PVD_root_directory_record_offset;
 890         if (p[DR_length_offset] != 34)
 891                 return (0);
 892
 893         iso9660->logical_block_size = logical_block_size;
 894         iso9660->volume_block = volume_block;
 895         iso9660->volume_size = logical_block_size * (uint64_t)volume_block;
 896         iso9660->primary.location = archive_le32dec(p + DR_extent_offset);
 897         iso9660->primary.size = archive_le32dec(p + DR_size_offset);
 898
 899         return (48);
 900 }
 901
 902 static int
 903 read_children(struct archive_read *a, struct file_info *parent)
 904 {
 905         struct iso9660 *iso9660;
 906         const unsigned char *b, *p;
 907         struct file_info *multi;
 908         size_t step;
 909
 910         iso9660 = (struct iso9660 *)(a->format->data);
 911         if (iso9660->current_position > parent->offset) {
 912                 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
 913                     "Ignoring out-of-order directory (%s) %jd > %jd",
 914                     parent->name.s,
 915                     iso9660->current_position,
 916                     parent->offset);
 917                 return (ARCHIVE_WARN);
 918         }
 919         if (parent->offset + parent->size > iso9660->volume_size) {
 920                 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
 921                     "Directory is beyond end-of-media: %s",
 922                     parent->name);
 923                 return (ARCHIVE_WARN);
 924         }
 925         if (iso9660->current_position < parent->offset) {
 926                 int64_t skipsize;
 927
 928                 skipsize = parent->offset - iso9660->current_position;
 929                 skipsize = __archive_read_skip(a, skipsize);
 930                 if (skipsize < 0)
 931                         return ((int)skipsize);
 932                 iso9660->current_position = parent->offset;
 933         }
 934
 935         step = ((parent->size + iso9660->logical_block_size -1) /
 936             iso9660->logical_block_size) * iso9660->logical_block_size;
 937         b = __archive_read_ahead(a, step, NULL);
 938         if (b == NULL) {
 939                 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
 940                     "Failed to read full block when scanning "
 941                     "ISO9660 directory list");
 942                 return (ARCHIVE_FATAL);
 943         }
 944         __archive_read_consume(a, step);
 945         iso9660->current_position += step;
 946         multi = NULL;
 947         while (step) {
 948                 p = b;
 949                 b += iso9660->logical_block_size;
 950                 step -= iso9660->logical_block_size;
 951                 for (; *p != 0 && p < b && p + *p <= b; p += *p) {
 952                         struct file_info *child;
 953
 954                         /* N.B.: these special directory identifiers
 955                          * are 8 bit "values" even on a
 956                          * Joliet CD with UCS-2 (16bit) encoding.
 957                          */
 958
 959                         /* Skip '.' entry. */
 960                         if (*(p + DR_name_len_offset) == 1
 961                             && *(p + DR_name_offset) == '\0')
 962                                 continue;
 963                         /* Skip '..' entry. */
 964                         if (*(p + DR_name_len_offset) == 1
 965                             && *(p + DR_name_offset) == '\001')
 966                                 continue;
 967                         child = parse_file_info(a, parent, p);
 968                         if (child == NULL)
 969                                 return (ARCHIVE_FATAL);
 970                         if (child->cl_offset)
 971                                 heap_add_entry(&(iso9660->cl_files),
 972                                     child, child->cl_offset);
 973                         else {
 974                                 if (child->multi_extent || multi != NULL) {
 975                                         struct content *con;
 976
 977                                         if (multi == NULL) {
 978                                                 multi = child;
 979                                                 multi->contents.first = NULL;
 980                                                 multi->contents.last =
 981                                                     &(multi->contents.first);
 982                                         }
 983                                         con = malloc(sizeof(struct content));
 984                                         if (con == NULL) {
 985                                                 archive_set_error(
 986                                                     &a->archive, ENOMEM,
 987                                                     "No memory for "
 988                                                     "multi extent");
 989                                                 return (ARCHIVE_FATAL);
 990                                         }
 991                                         con->offset = child->offset;
 992                                         con->size = child->size;
 993                                         con->next = NULL;
 994                                         *multi->contents.last = con;
 995                                         multi->contents.last = &(con->next);
 996                                         if (multi == child)
 997                                                 add_entry(iso9660, child);
 998                                         else {
 999                                                 multi->size += child->size;
1000                                                 if (!child->multi_extent)
1001                                                         multi = NULL;
1002                                         }
1003                                 } else
1004                                         add_entry(iso9660, child);
1005                         }
1006                 }
1007         }
1008
1009         /* Read data which recorded by RRIP "CE" extension. */
1010         if (read_CE(a, iso9660) != ARCHIVE_OK)
1011                 return (ARCHIVE_FATAL);
1012
1013         return (ARCHIVE_OK);
1014 }
1015
1016 static int
1017 relocate_dir(struct iso9660 *iso9660, struct file_info *file)
1018 {
1019         struct file_info *re;
1020
1021         re = heap_get_entry(&(iso9660->re_dirs));
1022         while (re != NULL && re->offset < file->cl_offset) {
1023                 /* This case is wrong pattern.
1024                  * But dont't reject this directory entry to be robust. */
1025                 cache_add_entry(iso9660, re);
1026                 re = heap_get_entry(&(iso9660->re_dirs));
1027         }
1028         if (re == NULL)
1029                 /* This case is wrong pattern. */
1030                 return (0);
1031         if (re->offset == file->cl_offset) {
1032                 re->parent->subdirs--;
1033                 re->parent = file->parent;
1034                 re->parent->subdirs++;
1035                 cache_add_to_next_of_parent(iso9660, re);
1036                 return (1);
1037         } else
1038                 /* This case is wrong pattern. */
1039                 heap_add_entry(&(iso9660->re_dirs), re, re->offset);
1040         return (0);
1041 }
1042
1043 static int
1044 read_entries(struct archive_read *a)
1045 {
1046         struct iso9660 *iso9660;
1047         struct file_info *file;
1048         int r;
1049
1050         iso9660 = (struct iso9660 *)(a->format->data);
1051
1052         while ((file = next_entry(iso9660)) != NULL &&
1053             (file->mode & AE_IFMT) == AE_IFDIR) {
1054                 r = read_children(a, file);
1055                 if (r != ARCHIVE_OK)
1056                         return (r);
1057
1058                 if (iso9660->seenRockridge &&
1059                     file->parent != NULL &&
1060                     file->parent->parent == NULL &&
1061                     iso9660->rr_moved == NULL &&
1062                     (strcmp(file->name.s, "rr_moved") == 0 ||
1063                      strcmp(file->name.s, ".rr_moved") == 0)) {
1064                         iso9660->rr_moved = file;
1065                 } else if (file->re)
1066                         heap_add_entry(&(iso9660->re_dirs), file,
1067                             file->offset);
1068                 else
1069                         cache_add_entry(iso9660, file);
1070         }
1071         if (file != NULL)
1072                 add_entry(iso9660, file);
1073
1074         if (iso9660->rr_moved != NULL) {
1075                 /*
1076                  * Relocate directory which rr_moved has.
1077                  */
1078                 while ((file = heap_get_entry(&(iso9660->cl_files))) != NULL)
1079                         relocate_dir(iso9660, file);
1080
1081                 /* If rr_moved directory still has children,
1082                  * Add rr_moved into pending_files to show
1083                  */
1084                 if (iso9660->rr_moved->subdirs) {
1085                         cache_add_entry(iso9660, iso9660->rr_moved);
1086                         /* If entries which have "RE" extension are still
1087                          * remaining(this case is unlikely except ISO image
1088                          * is broken), the entries won't be exposed. */
1089                         while ((file = heap_get_entry(&(iso9660->re_dirs))) != NULL)
1090                                 cache_add_entry(iso9660, file);
1091                 } else
1092                         iso9660->rr_moved->parent->subdirs--;
1093         } else {
1094                 /*
1095                  * In case ISO image is broken. If the name of rr_moved
1096                  * directory has been changed by damage, subdirectories
1097                  * of rr_moved entry won't be exposed.
1098                  */
1099                 while ((file = heap_get_entry(&(iso9660->re_dirs))) != NULL)
1100                         cache_add_entry(iso9660, file);
1101         }
1102
1103         return (ARCHIVE_OK);
1104 }
1105
1106 static int
1107 archive_read_format_iso9660_read_header(struct archive_read *a,
1108     struct archive_entry *entry)
1109 {
1110         struct iso9660 *iso9660;
1111         struct file_info *file;
1112         int r, rd_r;
1113
1114         iso9660 = (struct iso9660 *)(a->format->data);
1115
1116         if (!a->archive.archive_format) {
1117                 a->archive.archive_format = ARCHIVE_FORMAT_ISO9660;
1118                 a->archive.archive_format_name = "ISO9660";
1119         }
1120
1121         if (iso9660->current_position == 0) {
1122                 int64_t skipsize;
1123                 struct vd *vd;
1124                 const void *block;
1125                 char seenJoliet;
1126
1127                 vd = &(iso9660->primary);
1128                 if (!iso9660->opt_support_joliet)
1129                         iso9660->seenJoliet = 0;
1130                 if (iso9660->seenJoliet &&
1131                         vd->location > iso9660->joliet.location)
1132                         /* This condition is unlikely; by way of caution. */
1133                         vd = &(iso9660->joliet);
1134
1135                 skipsize = LOGICAL_BLOCK_SIZE * vd->location;
1136                 skipsize = __archive_read_skip(a, skipsize);
1137                 if (skipsize < 0)
1138                         return ((int)skipsize);
1139                 iso9660->current_position = skipsize;
1140
1141                 block = __archive_read_ahead(a, vd->size, NULL);
1142                 if (block == NULL) {
1143                         archive_set_error(&a->archive,
1144                             ARCHIVE_ERRNO_MISC,
1145                             "Failed to read full block when scanning "
1146                             "ISO9660 directory list");
1147                         return (ARCHIVE_FATAL);
1148                 }
1149
1150                 /*
1151                  * While reading Root Directory, flag seenJoliet
1152                  * must be zero to avoid converting special name
1153                  * 0x00(Current Directory) and next byte to UCS2.
1154                  */
1155                 seenJoliet = iso9660->seenJoliet;/* Save flag. */
1156                 iso9660->seenJoliet = 0;
1157                 file = parse_file_info(a, NULL, block);
1158                 if (file == NULL)
1159                         return (ARCHIVE_FATAL);
1160                 iso9660->seenJoliet = seenJoliet;
1161                 if (vd == &(iso9660->primary) && iso9660->seenRockridge
1162                     && iso9660->seenJoliet)
1163                         /*
1164                          * If iso image has RockRidge and Joliet,
1165                          * we use RockRidge Extensions.
1166                          */
1167                         iso9660->seenJoliet = 0;
1168                 if (vd == &(iso9660->primary) && !iso9660->seenRockridge
1169                     && iso9660->seenJoliet) {
1170                         /* Switch reading data from primary to joliet. */
1171                         vd = &(iso9660->joliet);
1172                         skipsize = LOGICAL_BLOCK_SIZE * vd->location;
1173                         skipsize -= iso9660->current_position;
1174                         skipsize = __archive_read_skip(a, skipsize);
1175                         if (skipsize < 0)
1176                                 return ((int)skipsize);
1177                         iso9660->current_position += skipsize;
1178
1179                         block = __archive_read_ahead(a, vd->size, NULL);
1180                         if (block == NULL) {
1181                                 archive_set_error(&a->archive,
1182                                     ARCHIVE_ERRNO_MISC,
1183                                     "Failed to read full block when scanning "
1184                                     "ISO9660 directory list");
1185                                 return (ARCHIVE_FATAL);
1186                         }
1187                         seenJoliet = iso9660->seenJoliet;/* Save flag. */
1188                         iso9660->seenJoliet = 0;
1189                         file = parse_file_info(a, NULL, block);
1190                         if (file == NULL)
1191                                 return (ARCHIVE_FATAL);
1192                         iso9660->seenJoliet = seenJoliet;
1193                 }
1194                 /* Store the root directory in the pending list. */
1195                 add_entry(iso9660, file);
1196                 if (iso9660->seenRockridge) {
1197                         a->archive.archive_format =
1198                             ARCHIVE_FORMAT_ISO9660_ROCKRIDGE;
1199                         a->archive.archive_format_name =
1200                             "ISO9660 with Rockridge extensions";
1201                 }
1202                 rd_r = read_entries(a);
1203                 if (rd_r == ARCHIVE_FATAL)
1204                         return (ARCHIVE_FATAL);
1205         } else
1206                 rd_r = ARCHIVE_OK;
1207
1208         /* Get the next entry that appears after the current offset. */
1209         r = next_entry_seek(a, iso9660, &file);
1210         if (r != ARCHIVE_OK)
1211                 return (r);
1212
1213         iso9660->entry_bytes_remaining = file->size;
1214         iso9660->entry_sparse_offset = 0; /* Offset for sparse-file-aware clients. */
1215
1216         if (file->offset + file->size > iso9660->volume_size) {
1217                 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1218                     "File is beyond end-of-media: %s", file->name);
1219                 iso9660->entry_bytes_remaining = 0;
1220                 iso9660->entry_sparse_offset = 0;
1221                 return (ARCHIVE_WARN);
1222         }
1223
1224         /* Set up the entry structure with information about this entry. */
1225         archive_entry_set_mode(entry, file->mode);
1226         archive_entry_set_uid(entry, file->uid);
1227         archive_entry_set_gid(entry, file->gid);
1228         archive_entry_set_nlink(entry, file->nlinks);
1229         if (file->birthtime_is_set)
1230                 archive_entry_set_birthtime(entry, file->birthtime, 0);
1231         else
1232                 archive_entry_unset_birthtime(entry);
1233         archive_entry_set_mtime(entry, file->mtime, 0);
1234         archive_entry_set_ctime(entry, file->ctime, 0);
1235         archive_entry_set_atime(entry, file->atime, 0);
1236         /* N.B.: Rock Ridge supports 64-bit device numbers. */
1237         archive_entry_set_rdev(entry, (dev_t)file->rdev);
1238         archive_entry_set_size(entry, iso9660->entry_bytes_remaining);
1239         archive_string_empty(&iso9660->pathname);
1240         archive_entry_set_pathname(entry,
1241             build_pathname(&iso9660->pathname, file));
1242         if (file->symlink.s != NULL)
1243                 archive_entry_copy_symlink(entry, file->symlink.s);
1244
1245         /* Note: If the input isn't seekable, we can't rewind to
1246          * return the same body again, so if the next entry refers to
1247          * the same data, we have to return it as a hardlink to the
1248          * original entry. */
1249         if (file->number != -1 &&
1250             file->number == iso9660->previous_number) {
1251                 archive_entry_set_hardlink(entry,
1252                     iso9660->previous_pathname.s);
1253                 archive_entry_unset_size(entry);
1254                 iso9660->entry_bytes_remaining = 0;
1255                 iso9660->entry_sparse_offset = 0;
1256                 return (ARCHIVE_OK);
1257         }
1258
1259         /* Except for the hardlink case above, if the offset of the
1260          * next entry is before our current position, we can't seek
1261          * backwards to extract it, so issue a warning.  Note that
1262          * this can only happen if this entry was added to the heap
1263          * after we passed this offset, that is, only if the directory
1264          * mentioning this entry is later than the body of the entry.
1265          * Such layouts are very unusual; most ISO9660 writers lay out
1266          * and record all directory information first, then store
1267          * all file bodies. */
1268         /* TODO: Someday, libarchive's I/O core will support optional
1269          * seeking.  When that day comes, this code should attempt to
1270          * seek and only return the error if the seek fails.  That
1271          * will give us support for whacky ISO images that require
1272          * seeking while retaining the ability to read almost all ISO
1273          * images in a streaming fashion. */
1274         if ((file->mode & AE_IFMT) != AE_IFDIR &&
1275             file->offset < iso9660->current_position) {
1276                 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1277                     "Ignoring out-of-order file @%x (%s) %jd < %jd",
1278                     file,
1279                     iso9660->pathname.s,
1280                     file->offset, iso9660->current_position);
1281                 iso9660->entry_bytes_remaining = 0;
1282                 iso9660->entry_sparse_offset = 0;
1283                 return (ARCHIVE_WARN);
1284         }
1285
1286         /* Initialize zisofs variables. */
1287         iso9660->entry_zisofs.pz = file->pz;
1288         if (file->pz) {
1289 #ifdef HAVE_ZLIB_H
1290                 struct zisofs  *zisofs;
1291
1292                 zisofs = &iso9660->entry_zisofs;
1293                 zisofs->initialized = 0;
1294                 zisofs->pz_log2_bs = file->pz_log2_bs;
1295                 zisofs->pz_uncompressed_size = file->pz_uncompressed_size;
1296                 zisofs->pz_offset = 0;
1297                 zisofs->header_avail = 0;
1298                 zisofs->header_passed = 0;
1299                 zisofs->block_pointers_avail = 0;
1300 #endif
1301                 archive_entry_set_size(entry, file->pz_uncompressed_size);
1302         }
1303
1304         iso9660->previous_number = file->number;
1305         archive_strcpy(&iso9660->previous_pathname, iso9660->pathname.s);
1306
1307         /* Reset entry_bytes_remaining if the file is multi extent. */
1308         iso9660->entry_content = file->contents.first;
1309         if (iso9660->entry_content != NULL)
1310                 iso9660->entry_bytes_remaining = iso9660->entry_content->size;
1311
1312         if (archive_entry_filetype(entry) == AE_IFDIR) {
1313                 /* Overwrite nlinks by proper link number which is
1314                  * calculated from number of sub directories. */
1315                 archive_entry_set_nlink(entry, 2 + file->subdirs);
1316                 /* Directory data has been read completely. */
1317                 iso9660->entry_bytes_remaining = 0;
1318                 iso9660->entry_sparse_offset = 0;
1319                 file->exposed = 1;
1320         }
1321
1322         if (rd_r != ARCHIVE_OK)
1323                 return (rd_r);
1324         return (ARCHIVE_OK);
1325 }
1326
1327 static int
1328 archive_read_format_iso9660_read_data_skip(struct archive_read *a)
1329 {
1330         /* Because read_next_header always does an explicit skip
1331          * to the next entry, we don't need to do anything here. */
1332         (void)a; /* UNUSED */
1333         return (ARCHIVE_OK);
1334 }
1335
1336 #ifdef HAVE_ZLIB_H
1337
1338 static int
1339 zisofs_read_data(struct archive_read *a,
1340     const void **buff, size_t *size, off_t *offset)
1341 {
1342         struct iso9660 *iso9660;
1343         struct zisofs  *zisofs;
1344         const unsigned char *p;
1345         size_t avail;
1346         ssize_t bytes_read;
1347         size_t uncompressed_size;
1348         int r;
1349
1350         iso9660 = (struct iso9660 *)(a->format->data);
1351         zisofs = &iso9660->entry_zisofs;
1352
1353         p = __archive_read_ahead(a, 1, &bytes_read);
1354         if (bytes_read <= 0) {
1355                 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1356                     "Truncated zisofs file body");
1357                 return (ARCHIVE_FATAL);
1358         }
1359         if (bytes_read > iso9660->entry_bytes_remaining)
1360                 bytes_read = iso9660->entry_bytes_remaining;
1361         avail = bytes_read;
1362         uncompressed_size = 0;
1363
1364         if (!zisofs->initialized) {
1365                 size_t ceil, xsize;
1366
1367                 /* Allocate block pointers buffer. */
1368                 ceil = (zisofs->pz_uncompressed_size +
1369                         (1LL << zisofs->pz_log2_bs) - 1)
1370                         >> zisofs->pz_log2_bs;
1371                 xsize = (ceil + 1) * 4;
1372                 if (zisofs->block_pointers_alloc < xsize) {
1373                         size_t alloc;
1374
1375                         if (zisofs->block_pointers != NULL)
1376                                 free(zisofs->block_pointers);
1377                         alloc = ((xsize >> 10) + 1) << 10;
1378                         zisofs->block_pointers = malloc(alloc);
1379                         if (zisofs->block_pointers == NULL) {
1380                                 archive_set_error(&a->archive, ENOMEM,
1381                                     "No memory for zisofs decompression");
1382                                 return (ARCHIVE_FATAL);
1383                         }
1384                         zisofs->block_pointers_alloc = alloc;
1385                 }
1386                 zisofs->block_pointers_size = xsize;
1387
1388                 /* Allocate uncompressed data buffer. */
1389                 xsize = 1UL << zisofs->pz_log2_bs;
1390                 if (zisofs->uncompressed_buffer_size < xsize) {
1391                         if (zisofs->uncompressed_buffer != NULL)
1392                                 free(zisofs->uncompressed_buffer);
1393                         zisofs->uncompressed_buffer = malloc(xsize);
1394                         if (zisofs->uncompressed_buffer == NULL) {
1395                                 archive_set_error(&a->archive, ENOMEM,
1396                                     "No memory for zisofs decompression");
1397                                 return (ARCHIVE_FATAL);
1398                         }
1399                 }
1400                 zisofs->uncompressed_buffer_size = xsize;
1401
1402                 /*
1403                  * Read the file header, and check the magic code of zisofs.
1404                  */
1405                 if (zisofs->header_avail < sizeof(zisofs->header)) {
1406                         xsize = sizeof(zisofs->header) - zisofs->header_avail;
1407                         if (avail < xsize)
1408                                 xsize = avail;
1409                         memcpy(zisofs->header + zisofs->header_avail, p, xsize);
1410                         zisofs->header_avail += xsize;
1411                         avail -= xsize;
1412                         p += xsize;
1413                 }
1414                 if (!zisofs->header_passed &&
1415                     zisofs->header_avail == sizeof(zisofs->header)) {
1416                         int err = 0;
1417
1418                         if (memcmp(zisofs->header, zisofs_magic,
1419                             sizeof(zisofs_magic)) != 0)
1420                                 err = 1;
1421                         if (archive_le32dec(zisofs->header + 8)
1422                             != zisofs->pz_uncompressed_size)
1423                                 err = 1;
1424                         if (zisofs->header[12] != 4)
1425                                 err = 1;
1426                         if (zisofs->header[13] != zisofs->pz_log2_bs)
1427                                 err = 1;
1428                         if (err) {
1429                                 archive_set_error(&a->archive,
1430                                     ARCHIVE_ERRNO_FILE_FORMAT,
1431                                     "Illegal zisofs file body");
1432                                 return (ARCHIVE_FATAL);
1433                         }
1434                         zisofs->header_passed = 1;
1435                 }
1436                 /*
1437                  * Read block pointers.
1438                  */
1439                 if (zisofs->header_passed &&
1440                     zisofs->block_pointers_avail < zisofs->block_pointers_size) {
1441                         xsize = zisofs->block_pointers_size
1442                             - zisofs->block_pointers_avail;
1443                         if (avail < xsize)
1444                                 xsize = avail;
1445                         memcpy(zisofs->block_pointers
1446                             + zisofs->block_pointers_avail, p, xsize);
1447                         zisofs->block_pointers_avail += xsize;
1448                         avail -= xsize;
1449                         p += xsize;
1450                         if (zisofs->block_pointers_avail
1451                             == zisofs->block_pointers_size) {
1452                                 /* We've got all block pointers and initialize
1453                                  * related variables.   */
1454                                 zisofs->block_off = 0;
1455                                 zisofs->block_avail = 0;
1456                                 /* Complete a initialization */
1457                                 zisofs->initialized = 1;
1458                         }
1459                 }
1460
1461                 if (!zisofs->initialized)
1462                         goto next_data; /* We need more datas. */
1463         }
1464
1465         /*
1466          * Get block offsets from block pointers.
1467          */
1468         if (zisofs->block_avail == 0) {
1469                 uint32_t bst, bed;
1470
1471                 if (zisofs->block_off + 4 >= zisofs->block_pointers_size) {
1472                         /* There isn't a pair of offsets. */
1473                         archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1474                             "Illegal zisofs block pointers");
1475                         return (ARCHIVE_FATAL);
1476                 }
1477                 bst = archive_le32dec(zisofs->block_pointers + zisofs->block_off);
1478                 if (bst != zisofs->pz_offset + (bytes_read - avail)) {
1479                         /* TODO: Should we seek offset of current file by bst ? */
1480                         archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1481                             "Illegal zisofs block pointers(cannot seek)");
1482                         return (ARCHIVE_FATAL);
1483                 }
1484                 bed = archive_le32dec(
1485                     zisofs->block_pointers + zisofs->block_off + 4);
1486                 if (bed < bst) {
1487                         archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1488                             "Illegal zisofs block pointers");
1489                         return (ARCHIVE_FATAL);
1490                 }
1491                 zisofs->block_avail = bed - bst;
1492                 zisofs->block_off += 4;
1493
1494                 /* Initialize compression library for new block. */
1495                 if (zisofs->stream_valid)
1496                         r = inflateReset(&zisofs->stream);
1497                 else
1498                         r = inflateInit(&zisofs->stream);
1499                 if (r != Z_OK) {
1500                         archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1501                             "Can't initialize zisofs decompression.");
1502                         return (ARCHIVE_FATAL);
1503                 }
1504                 zisofs->stream_valid = 1;
1505                 zisofs->stream.total_in = 0;
1506                 zisofs->stream.total_out = 0;
1507         }
1508
1509         /*
1510          * Make uncompressed datas.
1511          */
1512         if (zisofs->block_avail == 0) {
1513                 memset(zisofs->uncompressed_buffer, 0,
1514                     zisofs->uncompressed_buffer_size);
1515                 uncompressed_size = zisofs->uncompressed_buffer_size;
1516         } else {
1517                 zisofs->stream.next_in = (Bytef *)(uintptr_t)(const void *)p;
1518                 if (avail > zisofs->block_avail)
1519                         zisofs->stream.avail_in = zisofs->block_avail;
1520                 else
1521                         zisofs->stream.avail_in = avail;
1522                 zisofs->stream.next_out = zisofs->uncompressed_buffer;
1523                 zisofs->stream.avail_out = zisofs->uncompressed_buffer_size;
1524
1525                 r = inflate(&zisofs->stream, 0);
1526                 switch (r) {
1527                 case Z_OK: /* Decompressor made some progress.*/
1528                 case Z_STREAM_END: /* Found end of stream. */
1529                         break;
1530                 default:
1531                         archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1532                             "zisofs decompression failed (%d)", r);
1533                         return (ARCHIVE_FATAL);
1534                 }
1535                 uncompressed_size =
1536                     zisofs->uncompressed_buffer_size - zisofs->stream.avail_out;
1537                 avail -= zisofs->stream.next_in - p;
1538                 zisofs->block_avail -= zisofs->stream.next_in - p;
1539         }
1540 next_data:
1541         bytes_read -= avail;
1542         *buff = zisofs->uncompressed_buffer;
1543         *size = uncompressed_size;
1544         *offset = iso9660->entry_sparse_offset;
1545         iso9660->entry_sparse_offset += uncompressed_size;
1546         iso9660->entry_bytes_remaining -= bytes_read;
1547         iso9660->current_position += bytes_read;
1548         zisofs->pz_offset += bytes_read;
1549         __archive_read_consume(a, bytes_read);
1550
1551         return (ARCHIVE_OK);
1552 }
1553
1554 #else /* HAVE_ZLIB_H */
1555
1556 static int
1557 zisofs_read_data(struct archive_read *a,
1558     const void **buff, size_t *size, off_t *offset)
1559 {
1560
1561         (void)buff;/* UNUSED */
1562         (void)size;/* UNUSED */
1563         (void)offset;/* UNUSED */
1564         archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1565             "zisofs is not supported on this platform.");
1566         return (ARCHIVE_FAILED);
1567 }
1568
1569 #endif /* HAVE_ZLIB_H */
1570
1571 static int
1572 archive_read_format_iso9660_read_data(struct archive_read *a,
1573     const void **buff, size_t *size, off_t *offset)
1574 {
1575         ssize_t bytes_read;
1576         struct iso9660 *iso9660;
1577
1578         iso9660 = (struct iso9660 *)(a->format->data);
1579         if (iso9660->entry_bytes_remaining <= 0) {
1580                 if (iso9660->entry_content != NULL)
1581                         iso9660->entry_content = iso9660->entry_content->next;
1582                 if (iso9660->entry_content == NULL) {
1583                         *buff = NULL;
1584                         *size = 0;
1585                         *offset = iso9660->entry_sparse_offset;
1586                         return (ARCHIVE_EOF);
1587                 }
1588                 /* Seek forward to the start of the entry. */
1589                 if (iso9660->current_position < iso9660->entry_content->offset) {
1590                         int64_t step;
1591
1592                         step = iso9660->entry_content->offset -
1593                             iso9660->current_position;
1594                         step = __archive_read_skip(a, step);
1595                         if (step < 0)
1596                                 return ((int)step);
1597                         iso9660->current_position =
1598                             iso9660->entry_content->offset;
1599                 }
1600                 if (iso9660->entry_content->offset < iso9660->current_position) {
1601                         archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1602                             "Ignoring out-of-order file (%s) %jd < %jd",
1603                             iso9660->pathname.s,
1604                             iso9660->entry_content->offset,
1605                             iso9660->current_position);
1606                         *buff = NULL;
1607                         *size = 0;
1608                         *offset = iso9660->entry_sparse_offset;
1609                         return (ARCHIVE_WARN);
1610                 }
1611                 iso9660->entry_bytes_remaining = iso9660->entry_content->size;
1612         }
1613         if (iso9660->entry_zisofs.pz)
1614                 return (zisofs_read_data(a, buff, size, offset));
1615
1616         *buff = __archive_read_ahead(a, 1, &bytes_read);
1617         if (bytes_read == 0)
1618                 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1619                     "Truncated input file");
1620         if (*buff == NULL)
1621                 return (ARCHIVE_FATAL);
1622         if (bytes_read > iso9660->entry_bytes_remaining)
1623                 bytes_read = iso9660->entry_bytes_remaining;
1624         *size = bytes_read;
1625         *offset = iso9660->entry_sparse_offset;
1626         iso9660->entry_sparse_offset += bytes_read;
1627         iso9660->entry_bytes_remaining -= bytes_read;
1628         iso9660->current_position += bytes_read;
1629         __archive_read_consume(a, bytes_read);
1630         return (ARCHIVE_OK);
1631 }
1632
1633 static int
1634 archive_read_format_iso9660_cleanup(struct archive_read *a)
1635 {
1636         struct iso9660 *iso9660;
1637         int r = ARCHIVE_OK;
1638
1639         iso9660 = (struct iso9660 *)(a->format->data);
1640         release_files(iso9660);
1641         free(iso9660->read_ce_req.reqs);
1642         archive_string_free(&iso9660->pathname);
1643         archive_string_free(&iso9660->previous_pathname);
1644         if (iso9660->pending_files.files)
1645                 free(iso9660->pending_files.files);
1646         if (iso9660->re_dirs.files)
1647                 free(iso9660->re_dirs.files);
1648         if (iso9660->cl_files.files)
1649                 free(iso9660->cl_files.files);
1650 #ifdef HAVE_ZLIB_H
1651         free(iso9660->entry_zisofs.uncompressed_buffer);
1652         free(iso9660->entry_zisofs.block_pointers);
1653         if (iso9660->entry_zisofs.stream_valid) {
1654                 if (inflateEnd(&iso9660->entry_zisofs.stream) != Z_OK) {
1655                         archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1656                             "Failed to clean up zlib decompressor");
1657                         r = ARCHIVE_FATAL;
1658                 }
1659         }
1660 #endif
1661         free(iso9660);
1662         (a->format->data) = NULL;
1663         return (r);
1664 }
1665
1666 /*
1667  * This routine parses a single ISO directory record, makes sense
1668  * of any extensions, and stores the result in memory.
1669  */
1670 static struct file_info *
1671 parse_file_info(struct archive_read *a, struct file_info *parent,
1672     const unsigned char *isodirrec)
1673 {
1674         struct iso9660 *iso9660;
1675         struct file_info *file;
1676         size_t name_len;
1677         const unsigned char *rr_start, *rr_end;
1678         const unsigned char *p;
1679         size_t dr_len;
1680         int32_t location;
1681         int flags;
1682
1683         iso9660 = (struct iso9660 *)(a->format->data);
1684
1685         dr_len = (size_t)isodirrec[DR_length_offset];
1686         name_len = (size_t)isodirrec[DR_name_len_offset];
1687         location = archive_le32dec(isodirrec + DR_extent_offset);
1688         /* Sanity check that dr_len needs at least 34. */
1689         if (dr_len < 34) {
1690                 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1691                     "Invalid length of directory record");
1692                 return (NULL);
1693         }
1694         /* Sanity check that name_len doesn't exceed dr_len. */
1695         if (dr_len - 33 < name_len || name_len == 0) {
1696                 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1697                     "Invalid length of file identifier");
1698                 return (NULL);
1699         }
1700         /* Sanity check that location doesn't exceed volume block.
1701          * Don't check lower limit of location; it's possibility
1702          * the location has negative value when file type is symbolic
1703          * link or file size is zero. As far as I know latest mkisofs
1704          * do that.
1705          */
1706         if (location >= iso9660->volume_block) {
1707                 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1708                     "Invalid location of extent of file");
1709                 return (NULL);
1710         }
1711
1712         /* Create a new file entry and copy data from the ISO dir record. */
1713         file = (struct file_info *)malloc(sizeof(*file));
1714         if (file == NULL) {
1715                 archive_set_error(&a->archive, ENOMEM,
1716                     "No memory for file entry");
1717                 return (NULL);
1718         }
1719         memset(file, 0, sizeof(*file));
1720         file->parent = parent;
1721         file->offset = iso9660->logical_block_size * (uint64_t)location;
1722         file->size = toi(isodirrec + DR_size_offset, DR_size_size);
1723         file->mtime = isodate7(isodirrec + DR_date_offset);
1724         file->ctime = file->atime = file->mtime;
1725
1726         p = isodirrec + DR_name_offset;
1727         /* Rockridge extensions (if any) follow name.  Compute this
1728          * before fidgeting the name_len below. */
1729         rr_start = p + name_len + (name_len & 1 ? 0 : 1);
1730         rr_end = isodirrec + dr_len;
1731
1732         if (iso9660->seenJoliet) {
1733                 /* Joliet names are max 64 chars (128 bytes) according to spec,
1734                  * but genisoimage/mkisofs allows recording longer Joliet
1735                  * names which are 103 UCS2 characters(206 bytes) by their
1736                  * option '-joliet-long'.
1737                  */
1738                 wchar_t wbuff[103+1], *wp;
1739                 const unsigned char *c;
1740
1741                 if (name_len > 206)
1742                         name_len = 206;
1743                 /* convert BE UTF-16 to wchar_t */
1744                 for (c = p, wp = wbuff;
1745                                 c < (p + name_len) &&
1746                                 wp < (wbuff + sizeof(wbuff)/sizeof(*wbuff) - 1);
1747                                 c += 2) {
1748                         *wp++ = (((255 & (int)c[0]) << 8) | (255 & (int)c[1]));
1749                 }
1750                 *wp = L'\0';
1751
1752 #if 0 /* untested code, is it at all useful on Joliet? */
1753                 /* trim trailing first version and dot from filename.
1754                  *
1755                  * Remember we where in UTF-16BE land!
1756                  * SEPARATOR 1 (.) and SEPARATOR 2 (;) are both
1757                  * 16 bits big endian characters on Joliet.
1758                  *
1759                  * TODO: sanitize filename?
1760                  *       Joliet allows any UCS-2 char except:
1761                  *       *, /, :, ;, ? and \.
1762                  */
1763                 /* Chop off trailing ';1' from files. */
1764                 if (*(wp-2) == ';' && *(wp-1) == '1') {
1765                         wp-=2;
1766                         *wp = L'\0';
1767                 }
1768
1769                 /* Chop off trailing '.' from filenames. */
1770                 if (*(wp-1) == '.')
1771                         *(--wp) = L'\0';
1772 #endif
1773
1774                 /* store the result in the file name field. */
1775                 archive_strappend_w_utf8(&file->name, wbuff);
1776         } else {
1777                 /* Chop off trailing ';1' from files. */
1778                 if (name_len > 2 && p[name_len - 2] == ';' &&
1779                                 p[name_len - 1] == '1')
1780                         name_len -= 2;
1781                 /* Chop off trailing '.' from filenames. */
1782                 if (name_len > 1 && p[name_len - 1] == '.')
1783                         --name_len;
1784
1785                 archive_strncpy(&file->name, (const char *)p, name_len);
1786         }
1787
1788         flags = isodirrec[DR_flags_offset];
1789         if (flags & 0x02)
1790                 file->mode = AE_IFDIR | 0700;
1791         else
1792                 file->mode = AE_IFREG | 0400;
1793         if (flags & 0x80)
1794                 file->multi_extent = 1;
1795         else
1796                 file->multi_extent = 0;
1797         /*
1798          * Use location for file number.
1799          * File number is treated as inode number to find out harlink
1800          * target. If Rockridge extensions is being used, file number
1801          * will be overwritten by FILE SERIAL NUMBER of RRIP "PX"
1802          * extension.
1803          * NOTE: Old mkisofs did not record that FILE SERIAL NUMBER
1804          * in ISO images.
1805          */
1806         if (file->size == 0 && location >= 0)
1807                 /* If file->size is zero, its location points wrong place.
1808                  * Dot not use it for file number.
1809                  * When location has negative value, it can be used
1810                  * for file number.
1811                  */
1812                 file->number = -1;
1813         else
1814                 file->number = (int64_t)(uint32_t)location;
1815
1816         /* Rockridge extensions overwrite information from above. */
1817         if (iso9660->opt_support_rockridge) {
1818                 if (parent == NULL && rr_end - rr_start >= 7) {
1819                         p = rr_start;
1820                         if (p[0] == 'S' && p[1] == 'P'
1821                             && p[2] == 7 && p[3] == 1
1822                             && p[4] == 0xBE && p[5] == 0xEF) {
1823                                 /*
1824                                  * SP extension stores the suspOffset
1825                                  * (Number of bytes to skip between
1826                                  * filename and SUSP records.)
1827                                  * It is mandatory by the SUSP standard
1828                                  * (IEEE 1281).
1829                                  *
1830                                  * It allows SUSP to coexist with
1831                                  * non-SUSP uses of the System
1832                                  * Use Area by placing non-SUSP data
1833                                  * before SUSP data.
1834                                  *
1835                                  * SP extension must be in the root
1836                                  * directory entry, disable all SUSP
1837                                  * processing if not found.
1838                                  */
1839                                 iso9660->suspOffset = p[6];
1840                                 iso9660->seenSUSP = 1;
1841                                 rr_start += 7;
1842                         }
1843                 }
1844                 if (iso9660->seenSUSP) {
1845                         int r;
1846
1847                         file->name_continues = 0;
1848                         file->symlink_continues = 0;
1849                         rr_start += iso9660->suspOffset;
1850                         r = parse_rockridge(a, file, rr_start, rr_end);
1851                         if (r != ARCHIVE_OK) {
1852                                 free(file);
1853                                 return (NULL);
1854                         }
1855                 } else
1856                         /* If there isn't SUSP, disable parsing
1857                          * rock ridge extensions. */
1858                         iso9660->opt_support_rockridge = 0;
1859         }
1860
1861         file->nlinks = 1;/* Reset nlink. we'll calculate it later. */
1862         /* Tell file's parent how many children that parent has. */
1863         if (parent != NULL && (flags & 0x02) && file->cl_offset == 0)
1864                 parent->subdirs++;
1865
1866 #if DEBUG
1867         /* DEBUGGING: Warn about attributes I don't yet fully support. */
1868         if ((flags & ~0x02) != 0) {
1869                 fprintf(stderr, "\n ** Unrecognized flag: ");
1870                 dump_isodirrec(stderr, isodirrec);
1871                 fprintf(stderr, "\n");
1872         } else if (toi(isodirrec + DR_volume_sequence_number_offset, 2) != 1) {
1873                 fprintf(stderr, "\n ** Unrecognized sequence number: ");
1874                 dump_isodirrec(stderr, isodirrec);
1875                 fprintf(stderr, "\n");
1876         } else if (*(isodirrec + DR_file_unit_size_offset) != 0) {
1877                 fprintf(stderr, "\n ** Unexpected file unit size: ");
1878                 dump_isodirrec(stderr, isodirrec);
1879                 fprintf(stderr, "\n");
1880         } else if (*(isodirrec + DR_interleave_offset) != 0) {
1881                 fprintf(stderr, "\n ** Unexpected interleave: ");
1882                 dump_isodirrec(stderr, isodirrec);
1883                 fprintf(stderr, "\n");
1884         } else if (*(isodirrec + DR_ext_attr_length_offset) != 0) {
1885                 fprintf(stderr, "\n ** Unexpected extended attribute length: ");
1886                 dump_isodirrec(stderr, isodirrec);
1887                 fprintf(stderr, "\n");
1888         }
1889 #endif
1890         register_file(iso9660, file);
1891         return (file);
1892 }
1893
1894 static int
1895 parse_rockridge(struct archive_read *a, struct file_info *file,
1896     const unsigned char *p, const unsigned char *end)
1897 {
1898         struct iso9660 *iso9660;
1899
1900         iso9660 = (struct iso9660 *)(a->format->data);
1901
1902         while (p + 4 <= end  /* Enough space for another entry. */
1903             && p[0] >= 'A' && p[0] <= 'Z' /* Sanity-check 1st char of name. */
1904             && p[1] >= 'A' && p[1] <= 'Z' /* Sanity-check 2nd char of name. */
1905             && p[2] >= 4 /* Sanity-check length. */
1906             && p + p[2] <= end) { /* Sanity-check length. */
1907                 const unsigned char *data = p + 4;
1908                 int data_length = p[2] - 4;
1909                 int version = p[3];
1910
1911                 /*
1912                  * Yes, each 'if' here does test p[0] again.
1913                  * Otherwise, the fall-through handling to catch
1914                  * unsupported extensions doesn't work.
1915                  */
1916                 switch(p[0]) {
1917                 case 'C':
1918                         if (p[0] == 'C' && p[1] == 'E') {
1919                                 if (version == 1 && data_length == 24) {
1920                                         /*
1921                                          * CE extension comprises:
1922                                          *   8 byte sector containing extension
1923                                          *   8 byte offset w/in above sector
1924                                          *   8 byte length of continuation
1925                                          */
1926                                         int32_t location =
1927                                             archive_le32dec(data);
1928                                         file->ce_offset =
1929                                             archive_le32dec(data+8);
1930                                         file->ce_size =
1931                                             archive_le32dec(data+16);
1932                                         if (register_CE(a, location, file)
1933                                             != ARCHIVE_OK)
1934                                                 return (ARCHIVE_FATAL);
1935                                 }
1936                                 break;
1937                         }
1938                         if (p[0] == 'C' && p[1] == 'L') {
1939                                 if (version == 1 && data_length == 8) {
1940                                         file->cl_offset = (uint64_t)
1941                                             iso9660->logical_block_size *
1942                                             (uint64_t)archive_le32dec(data);
1943                                         iso9660->seenRockridge = 1;
1944                                 }
1945                                 break;
1946                         }
1947                         /* FALLTHROUGH */
1948                 case 'N':
1949                         if (p[0] == 'N' && p[1] == 'M') {
1950                                 if (version == 1) {
1951                                         parse_rockridge_NM1(file,
1952                                             data, data_length);
1953                                         iso9660->seenRockridge = 1;
1954                                 }
1955                                 break;
1956                         }
1957                         /* FALLTHROUGH */
1958                 case 'P':
1959                         if (p[0] == 'P' && p[1] == 'D') {
1960                                 /*
1961                                  * PD extension is padding;
1962                                  * contents are always ignored.
1963                                  */
1964                                 break;
1965                         }
1966                         if (p[0] == 'P' && p[1] == 'N') {
1967                                 if (version == 1 && data_length == 16) {
1968                                         file->rdev = toi(data,4);
1969                                         file->rdev <<= 32;
1970                                         file->rdev |= toi(data + 8, 4);
1971                                         iso9660->seenRockridge = 1;
1972                                 }
1973                                 break;
1974                         }
1975                         if (p[0] == 'P' && p[1] == 'X') {
1976                                 /*
1977                                  * PX extension comprises:
1978                                  *   8 bytes for mode,
1979                                  *   8 bytes for nlinks,
1980                                  *   8 bytes for uid,
1981                                  *   8 bytes for gid,
1982                                  *   8 bytes for inode.
1983                                  */
1984                                 if (version == 1) {
1985                                         if (data_length >= 8)
1986                                                 file->mode
1987                                                     = toi(data, 4);
1988                                         if (data_length >= 16)
1989                                                 file->nlinks
1990                                                     = toi(data + 8, 4);
1991                                         if (data_length >= 24)
1992                                                 file->uid
1993                                                     = toi(data + 16, 4);
1994                                         if (data_length >= 32)
1995                                                 file->gid
1996                                                     = toi(data + 24, 4);
1997                                         if (data_length >= 40)
1998                                                 file->number
1999                                                     = toi(data + 32, 4);
2000                                         iso9660->seenRockridge = 1;
2001                                 }
2002                                 break;
2003                         }
2004                         /* FALLTHROUGH */
2005                 case 'R':
2006                         if (p[0] == 'R' && p[1] == 'E' && version == 1) {
2007                                 file->re = 1;
2008                                 iso9660->seenRockridge = 1;
2009                                 break;
2010                         }
2011                         if (p[0] == 'R' && p[1] == 'R' && version == 1) {
2012                                 /*
2013                                  * RR extension comprises:
2014                                  *    one byte flag value
2015                                  * This extension is obsolete,
2016                                  * so contents are always ignored.
2017                                  */
2018                                 break;
2019                         }
2020                         /* FALLTHROUGH */
2021                 case 'S':
2022                         if (p[0] == 'S' && p[1] == 'L') {
2023                                 if (version == 1) {
2024                                         parse_rockridge_SL1(file,
2025                                             data, data_length);
2026                                         iso9660->seenRockridge = 1;
2027                                 }
2028                                 break;
2029                         }
2030                         if (p[0] == 'S' && p[1] == 'T'
2031                             && data_length == 0 && version == 1) {
2032                                 /*
2033                                  * ST extension marks end of this
2034                                  * block of SUSP entries.
2035                                  *
2036                                  * It allows SUSP to coexist with
2037                                  * non-SUSP uses of the System
2038                                  * Use Area by placing non-SUSP data
2039                                  * after SUSP data.
2040                                  */
2041                                 iso9660->seenSUSP = 0;
2042                                 iso9660->seenRockridge = 0;
2043                                 return (ARCHIVE_OK);
2044                         }
2045                 case 'T':
2046                         if (p[0] == 'T' && p[1] == 'F') {
2047                                 if (version == 1) {
2048                                         parse_rockridge_TF1(file,
2049                                             data, data_length);
2050                                         iso9660->seenRockridge = 1;
2051                                 }
2052                                 break;
2053                         }
2054                         /* FALLTHROUGH */
2055                 case 'Z':
2056                         if (p[0] == 'Z' && p[1] == 'F') {
2057                                 if (version == 1)
2058                                         parse_rockridge_ZF1(file,
2059                                             data, data_length);
2060                                 break;
2061                         }
2062                         /* FALLTHROUGH */
2063                 default:
2064                         /* The FALLTHROUGHs above leave us here for
2065                          * any unsupported extension. */
2066                         break;
2067                 }
2068
2069
2070
2071                 p += p[2];
2072         }
2073         return (ARCHIVE_OK);
2074 }
2075
2076 static int
2077 register_CE(struct archive_read *a, int32_t location,
2078     struct file_info *file)
2079 {
2080         struct iso9660 *iso9660;
2081         struct read_ce_queue *heap;
2082         struct read_ce_req *p;
2083         uint64_t offset, parent_offset;
2084         int hole, parent;
2085
2086         iso9660 = (struct iso9660 *)(a->format->data);
2087         offset = ((uint64_t)location) * (uint64_t)iso9660->logical_block_size;
2088         if (((file->mode & AE_IFMT) == AE_IFREG &&
2089             offset >= file->offset) ||
2090             offset < iso9660->current_position) {
2091                 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
2092                     "Invalid location in SUSP \"CE\" extension");
2093                 return (ARCHIVE_FATAL);
2094         }
2095
2096         /* Expand our CE list as necessary. */
2097         heap = &(iso9660->read_ce_req);
2098         if (heap->cnt >= heap->allocated) {
2099                 int new_size;
2100
2101                 if (heap->allocated < 16)
2102                         new_size = 16;
2103                 else
2104                         new_size = heap->allocated * 2;
2105                 /* Overflow might keep us from growing the list. */
2106                 if (new_size <= heap->allocated)
2107                         __archive_errx(1, "Out of memory");
2108                 p = malloc(new_size * sizeof(p[0]));
2109                 if (p == NULL)
2110                         __archive_errx(1, "Out of memory");
2111                 if (heap->reqs != NULL) {
2112                         memcpy(p, heap->reqs, heap->cnt * sizeof(*p));
2113                         free(heap->reqs);
2114                 }
2115                 heap->reqs = p;
2116                 heap->allocated = new_size;
2117         }
2118
2119         /*
2120          * Start with hole at end, walk it up tree to find insertion point.
2121          */
2122         hole = heap->cnt++;
2123         while (hole > 0) {
2124                 parent = (hole - 1)/2;
2125                 parent_offset = heap->reqs[parent].offset;
2126                 if (offset >= parent_offset) {
2127                         heap->reqs[hole].offset = offset;
2128                         heap->reqs[hole].file = file;
2129                         return (ARCHIVE_OK);
2130                 }
2131                 // Move parent into hole <==> move hole up tree.
2132                 heap->reqs[hole] = heap->reqs[parent];
2133                 hole = parent;
2134         }
2135         heap->reqs[0].offset = offset;
2136         heap->reqs[0].file = file;
2137         return (ARCHIVE_OK);
2138 }
2139
2140 static void
2141 next_CE(struct read_ce_queue *heap)
2142 {
2143         uint64_t a_offset, b_offset, c_offset;
2144         int a, b, c;
2145         struct read_ce_req tmp;
2146
2147         if (heap->cnt < 1)
2148                 return;
2149
2150         /*
2151          * Move the last item in the heap to the root of the tree
2152          */
2153         heap->reqs[0] = heap->reqs[--(heap->cnt)];
2154
2155         /*
2156          * Rebalance the heap.
2157          */
2158         a = 0; // Starting element and its offset
2159         a_offset = heap->reqs[a].offset;
2160         for (;;) {
2161                 b = a + a + 1; // First child
2162                 if (b >= heap->cnt)
2163                         return;
2164                 b_offset = heap->reqs[b].offset;
2165                 c = b + 1; // Use second child if it is smaller.
2166                 if (c < heap->cnt) {
2167                         c_offset = heap->reqs[c].offset;
2168                         if (c_offset < b_offset) {
2169                                 b = c;
2170                                 b_offset = c_offset;
2171                         }
2172                 }
2173                 if (a_offset <= b_offset)
2174                         return;
2175                 tmp = heap->reqs[a];
2176                 heap->reqs[a] = heap->reqs[b];
2177                 heap->reqs[b] = tmp;
2178                 a = b;
2179         }
2180 }
2181
2182
2183 static int
2184 read_CE(struct archive_read *a, struct iso9660 *iso9660)
2185 {
2186         struct read_ce_queue *heap;
2187         const unsigned char *b, *p, *end;
2188         struct file_info *file;
2189         size_t step;
2190         int r;
2191
2192         /* Read data which RRIP "CE" extension points. */
2193         heap = &(iso9660->read_ce_req);
2194         step = iso9660->logical_block_size;
2195         while (heap->cnt &&
2196             heap->reqs[0].offset == iso9660->current_position) {
2197                 b = __archive_read_ahead(a, step, NULL);
2198                 if (b == NULL) {
2199                         archive_set_error(&a->archive,
2200                             ARCHIVE_ERRNO_MISC,
2201                             "Failed to read full block when scanning "
2202                             "ISO9660 directory list");
2203                         return (ARCHIVE_FATAL);
2204                 }
2205                 do {
2206                         file = heap->reqs[0].file;
2207                         p = b + file->ce_offset;
2208                         end = p + file->ce_size;
2209                         next_CE(heap);
2210                         r = parse_rockridge(a, file, p, end);
2211                         if (r != ARCHIVE_OK)
2212                                 return (ARCHIVE_FATAL);
2213                 } while (heap->cnt &&
2214                     heap->reqs[0].offset == iso9660->current_position);
2215                 /* NOTE: Do not move this consume's code to fron of
2216                  * do-while loop. Registration of nested CE extension
2217                  * might cause error because of current position. */
2218                 __archive_read_consume(a, step);
2219                 iso9660->current_position += step;
2220         }
2221         return (ARCHIVE_OK);
2222 }
2223
2224 static void
2225 parse_rockridge_NM1(struct file_info *file,
2226                     const unsigned char *data, int data_length)
2227 {
2228         if (!file->name_continues)
2229                 archive_string_empty(&file->name);
2230         file->name_continues = 0;
2231         if (data_length < 1)
2232                 return;
2233         /*
2234          * NM version 1 extension comprises:
2235          *   1 byte flag, value is one of:
2236          *     = 0: remainder is name
2237          *     = 1: remainder is name, next NM entry continues name
2238          *     = 2: "."
2239          *     = 4: ".."
2240          *     = 32: Implementation specific
2241          *     All other values are reserved.
2242          */
2243         switch(data[0]) {
2244         case 0:
2245                 if (data_length < 2)
2246                         return;
2247                 archive_strncat(&file->name, (const char *)data + 1, data_length - 1);
2248                 break;
2249         case 1:
2250                 if (data_length < 2)
2251                         return;
2252                 archive_strncat(&file->name, (const char *)data + 1, data_length - 1);
2253                 file->name_continues = 1;
2254                 break;
2255         case 2:
2256                 archive_strcat(&file->name, ".");
2257                 break;
2258         case 4:
2259                 archive_strcat(&file->name, "..");
2260                 break;
2261         default:
2262                 return;
2263         }
2264
2265 }
2266
2267 static void
2268 parse_rockridge_TF1(struct file_info *file, const unsigned char *data,
2269     int data_length)
2270 {
2271         char flag;
2272         /*
2273          * TF extension comprises:
2274          *   one byte flag
2275          *   create time (optional)
2276          *   modify time (optional)
2277          *   access time (optional)
2278          *   attribute time (optional)
2279          *  Time format and presence of fields
2280          *  is controlled by flag bits.
2281          */
2282         if (data_length < 1)
2283                 return;
2284         flag = data[0];
2285         ++data;
2286         --data_length;
2287         if (flag & 0x80) {
2288                 /* Use 17-byte time format. */
2289                 if ((flag & 1) && data_length >= 17) {
2290                         /* Create time. */
2291                         file->birthtime_is_set = 1;
2292                         file->birthtime = isodate17(data);
2293                         data += 17;
2294                         data_length -= 17;
2295                 }
2296                 if ((flag & 2) && data_length >= 17) {
2297                         /* Modify time. */
2298                         file->mtime = isodate17(data);
2299                         data += 17;
2300                         data_length -= 17;
2301                 }
2302                 if ((flag & 4) && data_length >= 17) {
2303                         /* Access time. */
2304                         file->atime = isodate17(data);
2305                         data += 17;
2306                         data_length -= 17;
2307                 }
2308                 if ((flag & 8) && data_length >= 17) {
2309                         /* Attribute change time. */
2310                         file->ctime = isodate17(data);
2311                 }
2312         } else {
2313                 /* Use 7-byte time format. */
2314                 if ((flag & 1) && data_length >= 7) {
2315                         /* Create time. */
2316                         file->birthtime_is_set = 1;
2317                         file->birthtime = isodate7(data);
2318                         data += 7;
2319                         data_length -= 7;
2320                 }
2321                 if ((flag & 2) && data_length >= 7) {
2322                         /* Modify time. */
2323                         file->mtime = isodate7(data);
2324                         data += 7;
2325                         data_length -= 7;
2326                 }
2327                 if ((flag & 4) && data_length >= 7) {
2328                         /* Access time. */
2329                         file->atime = isodate7(data);
2330                         data += 7;
2331                         data_length -= 7;
2332                 }
2333                 if ((flag & 8) && data_length >= 7) {
2334                         /* Attribute change time. */
2335                         file->ctime = isodate7(data);
2336                 }
2337         }
2338 }
2339
2340 static void
2341 parse_rockridge_SL1(struct file_info *file, const unsigned char *data,
2342     int data_length)
2343 {
2344         const char *separator = "";
2345
2346         if (!file->symlink_continues || file->symlink.length < 1)
2347                 archive_string_empty(&file->symlink);
2348         else if (!file->symlink_continues &&
2349             file->symlink.s[file->symlink.length - 1] != '/')
2350                 separator = "/";
2351         file->symlink_continues = 0;
2352
2353         /*
2354          * Defined flag values:
2355          *  0: This is the last SL record for this symbolic link
2356          *  1: this symbolic link field continues in next SL entry
2357          *  All other values are reserved.
2358          */
2359         if (data_length < 1)
2360                 return;
2361         switch(*data) {
2362         case 0:
2363                 break;
2364         case 1:
2365                 file->symlink_continues = 1;
2366                 break;
2367         default:
2368                 return;
2369         }
2370         ++data;  /* Skip flag byte. */
2371         --data_length;
2372
2373         /*
2374          * SL extension body stores "components".
2375          * Basically, this is a complicated way of storing
2376          * a POSIX path.  It also interferes with using
2377          * symlinks for storing non-path data. <sigh>
2378          *
2379          * Each component is 2 bytes (flag and length)
2380          * possibly followed by name data.
2381          */
2382         while (data_length >= 2) {
2383                 unsigned char flag = *data++;
2384                 unsigned char nlen = *data++;
2385                 data_length -= 2;
2386
2387                 archive_strcat(&file->symlink, separator);
2388                 separator = "/";
2389
2390                 switch(flag) {
2391                 case 0: /* Usual case, this is text. */
2392                         if (data_length < nlen)
2393                                 return;
2394                         archive_strncat(&file->symlink,
2395                             (const char *)data, nlen);
2396                         break;
2397                 case 0x01: /* Text continues in next component. */
2398                         if (data_length < nlen)
2399                                 return;
2400                         archive_strncat(&file->symlink,
2401                             (const char *)data, nlen);
2402                         separator = "";
2403                         break;
2404                 case 0x02: /* Current dir. */
2405                         archive_strcat(&file->symlink, ".");
2406                         break;
2407                 case 0x04: /* Parent dir. */
2408                         archive_strcat(&file->symlink, "..");
2409                         break;
2410                 case 0x08: /* Root of filesystem. */
2411                         archive_strcat(&file->symlink, "/");
2412                         separator = "";
2413                         break;
2414                 case 0x10: /* Undefined (historically "volume root" */
2415                         archive_string_empty(&file->symlink);
2416                         archive_strcat(&file->symlink, "ROOT");
2417                         break;
2418                 case 0x20: /* Undefined (historically "hostname") */
2419                         archive_strcat(&file->symlink, "hostname");
2420                         break;
2421                 default:
2422                         /* TODO: issue a warning ? */
2423                         return;
2424                 }
2425                 data += nlen;
2426                 data_length -= nlen;
2427         }
2428 }
2429
2430 static void
2431 parse_rockridge_ZF1(struct file_info *file, const unsigned char *data,
2432     int data_length)
2433 {
2434
2435         if (data[0] == 0x70 && data[1] == 0x7a && data_length == 12) {
2436                 /* paged zlib */
2437                 file->pz = 1;
2438                 file->pz_log2_bs = data[3];
2439                 file->pz_uncompressed_size = archive_le32dec(&data[4]);
2440         }
2441 }
2442
2443 static void
2444 register_file(struct iso9660 *iso9660, struct file_info *file)
2445 {
2446
2447         file->use_next = iso9660->use_files;
2448         iso9660->use_files = file;
2449 }
2450
2451 static void
2452 release_files(struct iso9660 *iso9660)
2453 {
2454         struct content *con, *connext;
2455         struct file_info *file;
2456
2457         file = iso9660->use_files;
2458         while (file != NULL) {
2459                 struct file_info *next = file->use_next;
2460
2461                 archive_string_free(&file->name);
2462                 archive_string_free(&file->symlink);
2463                 con = file->contents.first;
2464                 while (con != NULL) {
2465                         connext = con->next;
2466                         free(con);
2467                         con = connext;
2468                 }
2469                 free(file);
2470                 file = next;
2471         }
2472 }
2473
2474 static int
2475 next_entry_seek(struct archive_read *a, struct iso9660 *iso9660,
2476     struct file_info **pfile)
2477 {
2478         struct file_info *file;
2479
2480         *pfile = file = next_cache_entry(iso9660);
2481         if (file == NULL)
2482                 return (ARCHIVE_EOF);
2483
2484         /* Don't waste time seeking for zero-length bodies. */
2485         if (file->size == 0)
2486                 file->offset = iso9660->current_position;
2487
2488         /* Seek forward to the start of the entry. */
2489         if (iso9660->current_position < file->offset) {
2490                 int64_t step;
2491
2492                 step = file->offset - iso9660->current_position;
2493                 step = __archive_read_skip(a, step);
2494                 if (step < 0)
2495                         return ((int)step);
2496                 iso9660->current_position = file->offset;
2497         }
2498
2499         /* We found body of file; handle it now. */
2500         return (ARCHIVE_OK);
2501 }
2502
2503 static struct file_info *
2504 next_cache_entry(struct iso9660 *iso9660)
2505 {
2506         struct file_info *file;
2507         struct {
2508                 struct file_info        *first;
2509                 struct file_info        **last;
2510         }       empty_files;
2511         int64_t number;
2512         int count;
2513
2514         file = cache_get_entry(iso9660);
2515         if (file != NULL) {
2516                 while (file->parent != NULL && !file->parent->exposed) {
2517                         /* If file's parent is not exposed, it's moved
2518                          * to next entry of its parent. */
2519                         cache_add_to_next_of_parent(iso9660, file);
2520                         file = cache_get_entry(iso9660);
2521                 }
2522                 return (file);
2523         }
2524
2525         file = next_entry(iso9660);
2526         if (file == NULL)
2527                 return (NULL);
2528
2529         if ((file->mode & AE_IFMT) != AE_IFREG || file->number == -1)
2530                 return (file);
2531
2532         count = 0;
2533         number = file->number;
2534         iso9660->cache_files.first = NULL;
2535         iso9660->cache_files.last = &(iso9660->cache_files.first);
2536         empty_files.first = NULL;
2537         empty_files.last = &empty_files.first;
2538         /* Collect files which has the same file serial number.
2539          * Peek pending_files so that file which number is different
2540          * is not put bak. */
2541         while (iso9660->pending_files.used > 0 &&
2542             (iso9660->pending_files.files[0]->number == -1 ||
2543              iso9660->pending_files.files[0]->number == number)) {
2544                 if (file->number == -1) {
2545                         /* This file has the same offset
2546                          * but it's wrong offset which empty files
2547                          * and symlink files have.
2548                          * NOTE: This wrong offse was recorded by
2549                          * old mkisofs utility. If ISO images is
2550                          * created by latest mkisofs, this does not
2551                          * happen.
2552                          */
2553                         file->next = NULL;
2554                         *empty_files.last = file;
2555                         empty_files.last = &(file->next);
2556                 } else {
2557                         count++;
2558                         cache_add_entry(iso9660, file);
2559                 }
2560                 file = next_entry(iso9660);
2561         }
2562
2563         if (count == 0)
2564                 return (file);
2565         if (file->number == -1) {
2566                 file->next = NULL;
2567                 *empty_files.last = file;
2568                 empty_files.last = &(file->next);
2569         } else {
2570                 count++;
2571                 cache_add_entry(iso9660, file);
2572         }
2573
2574         if (count > 1) {
2575                 /* The count is the same as number of hardlink,
2576                  * so much so that each nlinks of files in cache_file
2577                  * is overwritten by value of the count.
2578                  */
2579                 for (file = iso9660->cache_files.first;
2580                     file != NULL; file = file->next)
2581                         file->nlinks = count;
2582         }
2583         /* If there are empty files, that files are added
2584          * to the tail of the cache_files. */
2585         if (empty_files.first != NULL) {
2586                 *iso9660->cache_files.last = empty_files.first;
2587                 iso9660->cache_files.last = empty_files.last;
2588         }
2589         return (cache_get_entry(iso9660));
2590 }
2591
2592 static inline void
2593 cache_add_entry(struct iso9660 *iso9660, struct file_info *file)
2594 {
2595         file->next = NULL;
2596         *iso9660->cache_files.last = file;
2597         iso9660->cache_files.last = &(file->next);
2598 }
2599
2600 static inline void
2601 cache_add_to_next_of_parent(struct iso9660 *iso9660, struct file_info *file)
2602 {
2603         file->next = file->parent->next;
2604         file->parent->next = file;
2605         if (iso9660->cache_files.last == &(file->parent->next))
2606                 iso9660->cache_files.last = &(file->next);
2607 }
2608
2609 static inline struct file_info *
2610 cache_get_entry(struct iso9660 *iso9660)
2611 {
2612         struct file_info *file;
2613
2614         if ((file = iso9660->cache_files.first) != NULL) {
2615                 iso9660->cache_files.first = file->next;
2616                 if (iso9660->cache_files.first == NULL)
2617                         iso9660->cache_files.last = &(iso9660->cache_files.first);
2618         }
2619         return (file);
2620 }
2621
2622 static void
2623 heap_add_entry(struct heap_queue *heap, struct file_info *file, uint64_t key)
2624 {
2625         uint64_t file_key, parent_key;
2626         int hole, parent;
2627
2628         /* Expand our pending files list as necessary. */
2629         if (heap->used >= heap->allocated) {
2630                 struct file_info **new_pending_files;
2631                 int new_size = heap->allocated * 2;
2632
2633                 if (heap->allocated < 1024)
2634                         new_size = 1024;
2635                 /* Overflow might keep us from growing the list. */
2636                 if (new_size <= heap->allocated)
2637                         __archive_errx(1, "Out of memory");
2638                 new_pending_files = (struct file_info **)
2639                     malloc(new_size * sizeof(new_pending_files[0]));
2640                 if (new_pending_files == NULL)
2641                         __archive_errx(1, "Out of memory");
2642                 memcpy(new_pending_files, heap->files,
2643                     heap->allocated * sizeof(new_pending_files[0]));
2644                 if (heap->files != NULL)
2645                         free(heap->files);
2646                 heap->files = new_pending_files;
2647                 heap->allocated = new_size;
2648         }
2649
2650         file_key = file->key = key;
2651
2652         /*
2653          * Start with hole at end, walk it up tree to find insertion point.
2654          */
2655         hole = heap->used++;
2656         while (hole > 0) {
2657                 parent = (hole - 1)/2;
2658                 parent_key = heap->files[parent]->key;
2659                 if (file_key >= parent_key) {
2660                         heap->files[hole] = file;
2661                         return;
2662                 }
2663                 // Move parent into hole <==> move hole up tree.
2664                 heap->files[hole] = heap->files[parent];
2665                 hole = parent;
2666         }
2667         heap->files[0] = file;
2668 }
2669
2670 static struct file_info *
2671 heap_get_entry(struct heap_queue *heap)
2672 {
2673         uint64_t a_key, b_key, c_key;
2674         int a, b, c;
2675         struct file_info *r, *tmp;
2676
2677         if (heap->used < 1)
2678                 return (NULL);
2679
2680         /*
2681          * The first file in the list is the earliest; we'll return this.
2682          */
2683         r = heap->files[0];
2684
2685         /*
2686          * Move the last item in the heap to the root of the tree
2687          */
2688         heap->files[0] = heap->files[--(heap->used)];
2689
2690         /*
2691          * Rebalance the heap.
2692          */
2693         a = 0; // Starting element and its heap key
2694         a_key = heap->files[a]->key;
2695         for (;;) {
2696                 b = a + a + 1; // First child
2697                 if (b >= heap->used)
2698                         return (r);
2699                 b_key = heap->files[b]->key;
2700                 c = b + 1; // Use second child if it is smaller.
2701                 if (c < heap->used) {
2702                         c_key = heap->files[c]->key;
2703                         if (c_key < b_key) {
2704                                 b = c;
2705                                 b_key = c_key;
2706                         }
2707                 }
2708                 if (a_key <= b_key)
2709                         return (r);
2710                 tmp = heap->files[a];
2711                 heap->files[a] = heap->files[b];
2712                 heap->files[b] = tmp;
2713                 a = b;
2714         }
2715 }
2716
2717 static unsigned int
2718 toi(const void *p, int n)
2719 {
2720         const unsigned char *v = (const unsigned char *)p;
2721         if (n > 1)
2722                 return v[0] + 256 * toi(v + 1, n - 1);
2723         if (n == 1)
2724                 return v[0];
2725         return (0);
2726 }
2727
2728 static time_t
2729 isodate7(const unsigned char *v)
2730 {
2731         struct tm tm;
2732         int offset;
2733         memset(&tm, 0, sizeof(tm));
2734         tm.tm_year = v[0];
2735         tm.tm_mon = v[1] - 1;
2736         tm.tm_mday = v[2];
2737         tm.tm_hour = v[3];
2738         tm.tm_min = v[4];
2739         tm.tm_sec = v[5];
2740         /* v[6] is the signed timezone offset, in 1/4-hour increments. */
2741         offset = ((const signed char *)v)[6];
2742         if (offset > -48 && offset < 52) {
2743                 tm.tm_hour -= offset / 4;
2744                 tm.tm_min -= (offset % 4) * 15;
2745         }
2746         return (time_from_tm(&tm));
2747 }
2748
2749 static time_t
2750 isodate17(const unsigned char *v)
2751 {
2752         struct tm tm;
2753         int offset;
2754         memset(&tm, 0, sizeof(tm));
2755         tm.tm_year = (v[0] - '0') * 1000 + (v[1] - '0') * 100
2756             + (v[2] - '0') * 10 + (v[3] - '0')
2757             - 1900;
2758         tm.tm_mon = (v[4] - '0') * 10 + (v[5] - '0');
2759         tm.tm_mday = (v[6] - '0') * 10 + (v[7] - '0');
2760         tm.tm_hour = (v[8] - '0') * 10 + (v[9] - '0');
2761         tm.tm_min = (v[10] - '0') * 10 + (v[11] - '0');
2762         tm.tm_sec = (v[12] - '0') * 10 + (v[13] - '0');
2763         /* v[16] is the signed timezone offset, in 1/4-hour increments. */
2764         offset = ((const signed char *)v)[16];
2765         if (offset > -48 && offset < 52) {
2766                 tm.tm_hour -= offset / 4;
2767                 tm.tm_min -= (offset % 4) * 15;
2768         }
2769         return (time_from_tm(&tm));
2770 }
2771
2772 static time_t
2773 time_from_tm(struct tm *t)
2774 {
2775 #if HAVE_TIMEGM
2776         /* Use platform timegm() if available. */
2777         return (timegm(t));
2778 #else
2779         /* Else use direct calculation using POSIX assumptions. */
2780         /* First, fix up tm_yday based on the year/month/day. */
2781         mktime(t);
2782         /* Then we can compute timegm() from first principles. */
2783         return (t->tm_sec + t->tm_min * 60 + t->tm_hour * 3600
2784             + t->tm_yday * 86400 + (t->tm_year - 70) * 31536000
2785             + ((t->tm_year - 69) / 4) * 86400 -
2786             ((t->tm_year - 1) / 100) * 86400
2787             + ((t->tm_year + 299) / 400) * 86400);
2788 #endif
2789 }
2790
2791 static const char *
2792 build_pathname(struct archive_string *as, struct file_info *file)
2793 {
2794         if (file->parent != NULL && archive_strlen(&file->parent->name) > 0) {
2795                 build_pathname(as, file->parent);
2796                 archive_strcat(as, "/");
2797         }
2798         if (archive_strlen(&file->name) == 0)
2799                 archive_strcat(as, ".");
2800         else
2801                 archive_string_concat(as, &file->name);
2802         return (as->s);
2803 }
2804
2805 #if DEBUG
2806 static void
2807 dump_isodirrec(FILE *out, const unsigned char *isodirrec)
2808 {
2809         fprintf(out, " l %d,",
2810             toi(isodirrec + DR_length_offset, DR_length_size));
2811         fprintf(out, " a %d,",
2812             toi(isodirrec + DR_ext_attr_length_offset, DR_ext_attr_length_size));
2813         fprintf(out, " ext 0x%x,",
2814             toi(isodirrec + DR_extent_offset, DR_extent_size));
2815         fprintf(out, " s %d,",
2816             toi(isodirrec + DR_size_offset, DR_extent_size));
2817         fprintf(out, " f 0x%02x,",
2818             toi(isodirrec + DR_flags_offset, DR_flags_size));
2819         fprintf(out, " u %d,",
2820             toi(isodirrec + DR_file_unit_size_offset, DR_file_unit_size_size));
2821         fprintf(out, " ilv %d,",
2822             toi(isodirrec + DR_interleave_offset, DR_interleave_size));
2823         fprintf(out, " seq %d,",
2824             toi(isodirrec + DR_volume_sequence_number_offset, DR_volume_sequence_number_size));
2825         fprintf(out, " nl %d:",
2826             toi(isodirrec + DR_name_len_offset, DR_name_len_size));
2827         fprintf(out, " `%.*s'",
2828             toi(isodirrec + DR_name_len_offset, DR_name_len_size), isodirrec + DR_name_offset);
2829 }
2830 #endif