2 * Copyright (c) 2003-2007 Tim Kientzle
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR
15 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
16 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
17 * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT,
18 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
19 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
20 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
21 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
22 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
23 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
26 #include "archive_platform.h"
27 __FBSDID("$FreeBSD: src/lib/libarchive/archive_read_support_format_tar.c,v 1.60 2007/07/15 19:13:59 kientzle Exp $");
33 /* #include <stdint.h> */ /* See archive_platform.h */
41 /* Obtain suitable wide-character manipulation functions. */
45 /* Good enough for equality testing, which is all we need. */
46 static int wcscmp(const wchar_t *s1, const wchar_t *s2)
49 while (*s1 && diff == 0)
50 diff = (int)*++s1 - (int)*++s2;
53 /* Good enough for equality testing, which is all we need. */
54 static int wcsncmp(const wchar_t *s1, const wchar_t *s2, size_t n)
57 while (*s1 && diff == 0 && n-- > 0)
58 diff = (int)*++s1 - (int)*++s2;
61 static size_t wcslen(const wchar_t *s)
71 #include "archive_entry.h"
72 #include "archive_private.h"
73 #include "archive_read_private.h"
76 * Layout of POSIX 'ustar' tar header.
78 struct archive_entry_header_ustar {
87 char linkname[100]; /* "old format" header ends here */
88 char magic[6]; /* For POSIX: "ustar\0" */
89 char version[2]; /* For POSIX: "00" */
98 * Structure of GNU tar header
105 struct archive_entry_header_gnutar {
115 char magic[8]; /* "ustar \0" (note blank/blank/null at end) */
125 struct gnu_sparse sparse[4];
129 * GNU doesn't use POSIX 'prefix' field; they use the 'L' (longname)
135 * Data specific to this format.
137 struct sparse_block {
138 struct sparse_block *next;
144 struct archive_string acl_text;
145 struct archive_string entry_name;
146 struct archive_string entry_linkname;
147 struct archive_string entry_uname;
148 struct archive_string entry_gname;
149 struct archive_string longlink;
150 struct archive_string longname;
151 struct archive_string pax_header;
152 struct archive_string pax_global;
153 struct archive_string line;
155 size_t pax_entry_length;
156 int header_recursion_depth;
157 off_t entry_bytes_remaining;
161 struct sparse_block *sparse_list;
162 struct sparse_block *sparse_last;
163 int64_t sparse_offset;
164 int64_t sparse_numbytes;
165 int sparse_gnu_major;
166 int sparse_gnu_minor;
167 char sparse_gnu_pending;
170 static size_t UTF8_mbrtowc(wchar_t *pwc, const char *s, size_t n);
171 static int archive_block_is_null(const unsigned char *p);
172 static char *base64_decode(const wchar_t *, size_t, size_t *);
173 static void gnu_add_sparse_entry(struct tar *,
174 off_t offset, off_t remaining);
175 static int gnu_sparse_old_read(struct archive_read *, struct tar *,
176 const struct archive_entry_header_gnutar *header);
177 static void gnu_sparse_old_parse(struct tar *,
178 const struct gnu_sparse *sparse, int length);
179 static int gnu_sparse_01_parse(struct tar *, const wchar_t *);
180 static ssize_t gnu_sparse_10_read(struct archive_read *, struct tar *);
181 static int header_Solaris_ACL(struct archive_read *, struct tar *,
182 struct archive_entry *, const void *);
183 static int header_common(struct archive_read *, struct tar *,
184 struct archive_entry *, const void *);
185 static int header_old_tar(struct archive_read *, struct tar *,
186 struct archive_entry *, const void *);
187 static int header_pax_extensions(struct archive_read *, struct tar *,
188 struct archive_entry *, const void *);
189 static int header_pax_global(struct archive_read *, struct tar *,
190 struct archive_entry *, const void *h);
191 static int header_longlink(struct archive_read *, struct tar *,
192 struct archive_entry *, const void *h);
193 static int header_longname(struct archive_read *, struct tar *,
194 struct archive_entry *, const void *h);
195 static int header_volume(struct archive_read *, struct tar *,
196 struct archive_entry *, const void *h);
197 static int header_ustar(struct archive_read *, struct tar *,
198 struct archive_entry *, const void *h);
199 static int header_gnutar(struct archive_read *, struct tar *,
200 struct archive_entry *, const void *h);
201 static int archive_read_format_tar_bid(struct archive_read *);
202 static int archive_read_format_tar_cleanup(struct archive_read *);
203 static int archive_read_format_tar_read_data(struct archive_read *a,
204 const void **buff, size_t *size, off_t *offset);
205 static int archive_read_format_tar_skip(struct archive_read *a);
206 static int archive_read_format_tar_read_header(struct archive_read *,
207 struct archive_entry *);
208 static int checksum(struct archive_read *, const void *);
209 static int pax_attribute(struct tar *, struct archive_entry *,
210 wchar_t *key, wchar_t *value);
211 static int pax_header(struct archive_read *, struct tar *,
212 struct archive_entry *, char *attr);
213 static void pax_time(const wchar_t *, int64_t *sec, long *nanos);
214 static ssize_t readline(struct archive_read *, struct tar *, const char **);
215 static int read_body_to_string(struct archive_read *, struct tar *,
216 struct archive_string *, const void *h);
217 static int64_t tar_atol(const char *, unsigned);
218 static int64_t tar_atol10(const wchar_t *, unsigned);
219 static int64_t tar_atol256(const char *, unsigned);
220 static int64_t tar_atol8(const char *, unsigned);
221 static int tar_read_header(struct archive_read *, struct tar *,
222 struct archive_entry *);
223 static int tohex(int c);
224 static char *url_decode(const char *);
225 static int utf8_decode(wchar_t *, const char *, size_t length);
226 static char *wide_to_narrow(const wchar_t *wval);
229 archive_read_support_format_gnutar(struct archive *a)
231 return (archive_read_support_format_tar(a));
236 archive_read_support_format_tar(struct archive *_a)
238 struct archive_read *a = (struct archive_read *)_a;
242 tar = (struct tar *)malloc(sizeof(*tar));
244 archive_set_error(&a->archive, ENOMEM,
245 "Can't allocate tar data");
246 return (ARCHIVE_FATAL);
248 memset(tar, 0, sizeof(*tar));
250 r = __archive_read_register_format(a, tar,
251 archive_read_format_tar_bid,
252 archive_read_format_tar_read_header,
253 archive_read_format_tar_read_data,
254 archive_read_format_tar_skip,
255 archive_read_format_tar_cleanup);
263 archive_read_format_tar_cleanup(struct archive_read *a)
266 struct sparse_block *p;
268 tar = (struct tar *)(a->format->data);
269 while (tar->sparse_list != NULL) {
270 p = tar->sparse_list;
271 tar->sparse_list = p->next;
274 archive_string_free(&tar->acl_text);
275 archive_string_free(&tar->entry_name);
276 archive_string_free(&tar->entry_linkname);
277 archive_string_free(&tar->entry_uname);
278 archive_string_free(&tar->entry_gname);
279 archive_string_free(&tar->line);
280 archive_string_free(&tar->pax_global);
281 archive_string_free(&tar->pax_header);
282 free(tar->pax_entry);
284 (a->format->data) = NULL;
290 archive_read_format_tar_bid(struct archive_read *a)
295 const struct archive_entry_header_ustar *header;
298 * If we're already reading a non-tar file, don't
301 if (a->archive.archive_format != 0 &&
302 (a->archive.archive_format & ARCHIVE_FORMAT_BASE_MASK) !=
308 * If we're already reading a tar format, start the bid at 1 as
311 if ((a->archive.archive_format & ARCHIVE_FORMAT_BASE_MASK) ==
315 /* Now let's look at the actual header and see if it matches. */
316 if (a->decompressor->read_ahead != NULL)
317 bytes_read = (a->decompressor->read_ahead)(a, &h, 512);
319 bytes_read = 0; /* Empty file. */
321 return (ARCHIVE_FATAL);
322 if (bytes_read == 0 && bid > 0) {
323 /* An archive without a proper end-of-archive marker. */
324 /* Hold our nose and bid 1 anyway. */
327 if (bytes_read < 512) {
328 /* If it's a new archive, then just return a zero bid. */
332 * If we already know this is a tar archive,
333 * then we have a problem.
335 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
336 "Truncated tar archive");
337 return (ARCHIVE_FATAL);
340 /* If it's an end-of-archive mark, we can handle it. */
341 if ((*(const char *)h) == 0 && archive_block_is_null((const unsigned char *)h)) {
342 /* If it's a known tar file, end-of-archive is definite. */
343 if ((a->archive.archive_format & ARCHIVE_FORMAT_BASE_MASK) ==
350 /* If it's not an end-of-archive mark, it must have a valid checksum.*/
353 bid += 48; /* Checksum is usually 6 octal digits. */
355 header = (const struct archive_entry_header_ustar *)h;
357 /* Recognize POSIX formats. */
358 if ((memcmp(header->magic, "ustar\0", 6) == 0)
359 &&(memcmp(header->version, "00", 2)==0))
362 /* Recognize GNU tar format. */
363 if ((memcmp(header->magic, "ustar ", 6) == 0)
364 &&(memcmp(header->version, " \0", 2)==0))
367 /* Type flag must be null, digit or A-Z, a-z. */
368 if (header->typeflag[0] != 0 &&
369 !( header->typeflag[0] >= '0' && header->typeflag[0] <= '9') &&
370 !( header->typeflag[0] >= 'A' && header->typeflag[0] <= 'Z') &&
371 !( header->typeflag[0] >= 'a' && header->typeflag[0] <= 'z') )
373 bid += 2; /* 6 bits of variation in an 8-bit field leaves 2 bits. */
375 /* Sanity check: Look at first byte of mode field. */
376 switch (255 & (unsigned)header->mode[0]) {
378 /* Base-256 value: No further verification possible! */
380 case ' ': /* Not recommended, but not illegal, either. */
382 case '0': case '1': case '2': case '3':
383 case '4': case '5': case '6': case '7':
385 /* TODO: Check format of remainder of this field. */
388 /* Not a valid mode; bail out here. */
391 /* TODO: Sanity test uid/gid/size/mtime/rdevmajor/rdevminor fields. */
397 * The function invoked by archive_read_header(). This
398 * just sets up a few things and then calls the internal
399 * tar_read_header() function below.
402 archive_read_format_tar_read_header(struct archive_read *a,
403 struct archive_entry *entry)
406 * When converting tar archives to cpio archives, it is
407 * essential that each distinct file have a distinct inode
408 * number. To simplify this, we keep a static count here to
409 * assign fake dev/inode numbers to each tar entry. Note that
410 * pax format archives may overwrite this with something more
413 * Ideally, we would track every file read from the archive so
414 * that we could assign the same dev/ino pair to hardlinks,
415 * but the memory required to store a complete lookup table is
416 * probably not worthwhile just to support the relatively
417 * obscure tar->cpio conversion case.
419 static int default_inode;
420 static int default_dev;
422 struct sparse_block *sp;
428 /* Assign default device/inode values. */
429 archive_entry_set_dev(entry, 1 + default_dev); /* Don't use zero. */
430 archive_entry_set_ino(entry, ++default_inode); /* Don't use zero. */
431 /* Limit generated st_ino number to 16 bits. */
432 if (default_inode >= 0xffff) {
437 tar = (struct tar *)(a->format->data);
438 tar->entry_offset = 0;
439 while (tar->sparse_list != NULL) {
440 sp = tar->sparse_list;
441 tar->sparse_list = sp->next;
444 tar->sparse_last = NULL;
446 r = tar_read_header(a, tar, entry);
449 * Yuck. See comments for gnu_sparse_10_read for why this
450 * is here and not in _read_data where it "should" go.
452 if (tar->sparse_gnu_pending
453 && tar->sparse_gnu_major == 1
454 && tar->sparse_gnu_minor == 0) {
455 tar->sparse_gnu_pending = 0;
456 /* Read initial sparse map. */
457 size = gnu_sparse_10_read(a, tar);
460 tar->entry_bytes_remaining -= size;
461 tar->entry_padding += size;
465 * "non-sparse" files are really just sparse files with
468 if (tar->sparse_list == NULL)
469 gnu_add_sparse_entry(tar, 0, tar->entry_bytes_remaining);
471 tar->realsize = archive_entry_size(entry);
473 if (r == ARCHIVE_OK) {
475 * "Regular" entry with trailing '/' is really
476 * directory: This is needed for certain old tar
477 * variants and even for some broken newer ones.
479 p = archive_entry_pathname(entry);
481 if (archive_entry_filetype(entry) == AE_IFREG
483 archive_entry_set_filetype(entry, AE_IFDIR);
489 archive_read_format_tar_read_data(struct archive_read *a,
490 const void **buff, size_t *size, off_t *offset)
494 struct sparse_block *p;
496 tar = (struct tar *)(a->format->data);
498 if (tar->sparse_gnu_pending) {
499 if (tar->sparse_gnu_major == 1 && tar->sparse_gnu_minor == 0) {
501 * <sigh> We should parse the sparse data
502 * here, but have to parse it as part of the
503 * header because of a bug in GNU tar 1.16.1.
508 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
509 "Unrecognized GNU sparse file format");
510 return (ARCHIVE_WARN);
512 tar->sparse_gnu_pending = 0;
515 /* Remove exhausted entries from sparse list. */
516 while (tar->sparse_list != NULL &&
517 tar->sparse_list->remaining == 0) {
518 p = tar->sparse_list;
519 tar->sparse_list = p->next;
523 /* If we're at end of file, return EOF. */
524 if (tar->sparse_list == NULL || tar->entry_bytes_remaining == 0) {
525 if ((a->decompressor->skip)(a, tar->entry_padding) < 0)
526 return (ARCHIVE_FATAL);
527 tar->entry_padding = 0;
530 *offset = tar->realsize;
531 return (ARCHIVE_EOF);
534 bytes_read = (a->decompressor->read_ahead)(a, buff, 1);
535 if (bytes_read == 0) {
536 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
537 "Truncated tar archive");
538 return (ARCHIVE_FATAL);
541 return (ARCHIVE_FATAL);
542 if (bytes_read > tar->entry_bytes_remaining)
543 bytes_read = tar->entry_bytes_remaining;
544 /* Don't read more than is available in the
545 * current sparse block. */
546 if (tar->sparse_list->remaining < bytes_read)
547 bytes_read = tar->sparse_list->remaining;
549 *offset = tar->sparse_list->offset;
550 tar->sparse_list->remaining -= bytes_read;
551 tar->sparse_list->offset += bytes_read;
552 tar->entry_bytes_remaining -= bytes_read;
553 (a->decompressor->consume)(a, bytes_read);
558 archive_read_format_tar_skip(struct archive_read *a)
562 struct sparse_block *p;
564 tar = (struct tar *)(a->format->data);
567 * Compression layer skip functions are required to either skip the
568 * length requested or fail, so we can rely upon the entire entry
569 * plus padding being skipped.
571 bytes_skipped = (a->decompressor->skip)(a, tar->entry_bytes_remaining +
573 if (bytes_skipped < 0)
574 return (ARCHIVE_FATAL);
576 tar->entry_bytes_remaining = 0;
577 tar->entry_padding = 0;
579 /* Free the sparse list. */
580 while (tar->sparse_list != NULL) {
581 p = tar->sparse_list;
582 tar->sparse_list = p->next;
585 tar->sparse_last = NULL;
591 * This function recursively interprets all of the headers associated
592 * with a single entry.
595 tar_read_header(struct archive_read *a, struct tar *tar,
596 struct archive_entry *entry)
601 const struct archive_entry_header_ustar *header;
603 /* Read 512-byte header record */
604 bytes = (a->decompressor->read_ahead)(a, &h, 512);
607 * If we're here, it's becase the _bid function accepted
608 * this file. So just call a short read end-of-archive
609 * and be done with it.
611 return (ARCHIVE_EOF);
613 (a->decompressor->consume)(a, 512);
615 /* Check for end-of-archive mark. */
616 if (((*(const char *)h)==0) && archive_block_is_null((const unsigned char *)h)) {
617 /* Try to consume a second all-null record, as well. */
618 bytes = (a->decompressor->read_ahead)(a, &h, 512);
620 (a->decompressor->consume)(a, bytes);
621 archive_set_error(&a->archive, 0, NULL);
622 return (ARCHIVE_EOF);
626 * Note: If the checksum fails and we return ARCHIVE_RETRY,
627 * then the client is likely to just retry. This is a very
628 * crude way to search for the next valid header!
630 * TODO: Improve this by implementing a real header scan.
632 if (!checksum(a, h)) {
633 archive_set_error(&a->archive, EINVAL, "Damaged tar archive");
634 return (ARCHIVE_RETRY); /* Retryable: Invalid header */
637 if (++tar->header_recursion_depth > 32) {
638 archive_set_error(&a->archive, EINVAL, "Too many special headers");
639 return (ARCHIVE_WARN);
642 /* Determine the format variant. */
643 header = (const struct archive_entry_header_ustar *)h;
644 switch(header->typeflag[0]) {
645 case 'A': /* Solaris tar ACL */
646 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE;
647 a->archive.archive_format_name = "Solaris tar";
648 err = header_Solaris_ACL(a, tar, entry, h);
650 case 'g': /* POSIX-standard 'g' header. */
651 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE;
652 a->archive.archive_format_name = "POSIX pax interchange format";
653 err = header_pax_global(a, tar, entry, h);
655 case 'K': /* Long link name (GNU tar, others) */
656 err = header_longlink(a, tar, entry, h);
658 case 'L': /* Long filename (GNU tar, others) */
659 err = header_longname(a, tar, entry, h);
661 case 'V': /* GNU volume header */
662 err = header_volume(a, tar, entry, h);
664 case 'X': /* Used by SUN tar; same as 'x'. */
665 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE;
666 a->archive.archive_format_name =
667 "POSIX pax interchange format (Sun variant)";
668 err = header_pax_extensions(a, tar, entry, h);
670 case 'x': /* POSIX-standard 'x' header. */
671 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE;
672 a->archive.archive_format_name = "POSIX pax interchange format";
673 err = header_pax_extensions(a, tar, entry, h);
676 if (memcmp(header->magic, "ustar \0", 8) == 0) {
677 a->archive.archive_format = ARCHIVE_FORMAT_TAR_GNUTAR;
678 a->archive.archive_format_name = "GNU tar format";
679 err = header_gnutar(a, tar, entry, h);
680 } else if (memcmp(header->magic, "ustar", 5) == 0) {
681 if (a->archive.archive_format != ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE) {
682 a->archive.archive_format = ARCHIVE_FORMAT_TAR_USTAR;
683 a->archive.archive_format_name = "POSIX ustar format";
685 err = header_ustar(a, tar, entry, h);
687 a->archive.archive_format = ARCHIVE_FORMAT_TAR;
688 a->archive.archive_format_name = "tar (non-POSIX)";
689 err = header_old_tar(a, tar, entry, h);
692 --tar->header_recursion_depth;
693 /* We return warnings or success as-is. Anything else is fatal. */
694 if (err == ARCHIVE_WARN || err == ARCHIVE_OK)
696 if (err == ARCHIVE_EOF)
697 /* EOF when recursively reading a header is bad. */
698 archive_set_error(&a->archive, EINVAL, "Damaged tar archive");
699 return (ARCHIVE_FATAL);
703 * Return true if block checksum is correct.
706 checksum(struct archive_read *a, const void *h)
708 const unsigned char *bytes;
709 const struct archive_entry_header_ustar *header;
712 (void)a; /* UNUSED */
713 bytes = (const unsigned char *)h;
714 header = (const struct archive_entry_header_ustar *)h;
717 * Test the checksum. Note that POSIX specifies _unsigned_
718 * bytes for this calculation.
720 sum = tar_atol(header->checksum, sizeof(header->checksum));
722 for (i = 0; i < 148; i++)
723 check += (unsigned char)bytes[i];
727 check += (unsigned char)bytes[i];
732 * Repeat test with _signed_ bytes, just in case this archive
733 * was created by an old BSD, Solaris, or HP-UX tar with a
734 * broken checksum calculation.
737 for (i = 0; i < 148; i++)
738 check += (signed char)bytes[i];
742 check += (signed char)bytes[i];
750 * Return true if this block contains only nulls.
753 archive_block_is_null(const unsigned char *p)
757 for (i = 0; i < ARCHIVE_BYTES_PER_RECORD / sizeof(*p); i++)
764 * Interpret 'A' Solaris ACL header
767 header_Solaris_ACL(struct archive_read *a, struct tar *tar,
768 struct archive_entry *entry, const void *h)
770 const struct archive_entry_header_ustar *header;
777 * read_body_to_string adds a NUL terminator, but we need a little
778 * more to make sure that we don't overrun acl_text later.
780 header = (const struct archive_entry_header_ustar *)h;
781 size = tar_atol(header->size, sizeof(header->size));
782 err = read_body_to_string(a, tar, &(tar->acl_text), h);
783 if (err != ARCHIVE_OK)
785 err = tar_read_header(a, tar, entry);
786 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN))
789 /* Skip leading octal number. */
790 /* XXX TODO: Parse the octal number and sanity-check it. */
791 p = acl = tar->acl_text.s;
792 while (*p != '\0' && p < acl + size)
796 if (p >= acl + size) {
797 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
798 "Malformed Solaris ACL attribute");
799 return(ARCHIVE_WARN);
802 /* Skip leading octal number. */
806 while (*p != '\0' && p < acl + size)
809 wp = (wchar_t *)malloc((p - acl + 1) * sizeof(wchar_t));
811 archive_set_error(&a->archive, ENOMEM,
812 "Can't allocate work buffer for ACL parsing");
813 return (ARCHIVE_FATAL);
815 utf8_decode(wp, acl, p - acl);
816 err = __archive_entry_acl_parse_w(entry, wp,
817 ARCHIVE_ENTRY_ACL_TYPE_ACCESS);
823 * Interpret 'K' long linkname header.
826 header_longlink(struct archive_read *a, struct tar *tar,
827 struct archive_entry *entry, const void *h)
831 err = read_body_to_string(a, tar, &(tar->longlink), h);
832 if (err != ARCHIVE_OK)
834 err = tar_read_header(a, tar, entry);
835 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN))
837 /* Set symlink if symlink already set, else hardlink. */
838 archive_entry_set_link(entry, tar->longlink.s);
843 * Interpret 'L' long filename header.
846 header_longname(struct archive_read *a, struct tar *tar,
847 struct archive_entry *entry, const void *h)
851 err = read_body_to_string(a, tar, &(tar->longname), h);
852 if (err != ARCHIVE_OK)
854 /* Read and parse "real" header, then override name. */
855 err = tar_read_header(a, tar, entry);
856 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN))
858 archive_entry_set_pathname(entry, tar->longname.s);
864 * Interpret 'V' GNU tar volume header.
867 header_volume(struct archive_read *a, struct tar *tar,
868 struct archive_entry *entry, const void *h)
872 /* Just skip this and read the next header. */
873 return (tar_read_header(a, tar, entry));
877 * Read body of an archive entry into an archive_string object.
880 read_body_to_string(struct archive_read *a, struct tar *tar,
881 struct archive_string *as, const void *h)
883 off_t size, padded_size;
884 ssize_t bytes_read, bytes_to_copy;
885 const struct archive_entry_header_ustar *header;
889 (void)tar; /* UNUSED */
890 header = (const struct archive_entry_header_ustar *)h;
891 size = tar_atol(header->size, sizeof(header->size));
892 if ((size > 1048576) || (size < 0)) {
893 archive_set_error(&a->archive, EINVAL,
894 "Special header too large");
895 return (ARCHIVE_FATAL);
898 /* Fail if we can't make our buffer big enough. */
899 if (archive_string_ensure(as, size+1) == NULL) {
900 archive_set_error(&a->archive, ENOMEM,
902 return (ARCHIVE_FATAL);
905 /* Read the body into the string. */
906 padded_size = (size + 511) & ~ 511;
908 while (padded_size > 0) {
909 bytes_read = (a->decompressor->read_ahead)(a, &src, padded_size);
911 return (ARCHIVE_EOF);
913 return (ARCHIVE_FATAL);
914 if (bytes_read > padded_size)
915 bytes_read = padded_size;
916 (a->decompressor->consume)(a, bytes_read);
917 bytes_to_copy = bytes_read;
918 if ((off_t)bytes_to_copy > size)
919 bytes_to_copy = (ssize_t)size;
920 memcpy(dest, src, bytes_to_copy);
921 dest += bytes_to_copy;
922 size -= bytes_to_copy;
923 padded_size -= bytes_read;
930 * Parse out common header elements.
932 * This would be the same as header_old_tar, except that the
933 * filename is handled slightly differently for old and POSIX
934 * entries (POSIX entries support a 'prefix'). This factoring
935 * allows header_old_tar and header_ustar
936 * to handle filenames differently, while still putting most of the
937 * common parsing into one place.
940 header_common(struct archive_read *a, struct tar *tar,
941 struct archive_entry *entry, const void *h)
943 const struct archive_entry_header_ustar *header;
946 (void)a; /* UNUSED */
948 header = (const struct archive_entry_header_ustar *)h;
949 if (header->linkname[0])
950 archive_strncpy(&(tar->entry_linkname), header->linkname,
951 sizeof(header->linkname));
953 archive_string_empty(&(tar->entry_linkname));
955 /* Parse out the numeric fields (all are octal) */
956 archive_entry_set_mode(entry, tar_atol(header->mode, sizeof(header->mode)));
957 archive_entry_set_uid(entry, tar_atol(header->uid, sizeof(header->uid)));
958 archive_entry_set_gid(entry, tar_atol(header->gid, sizeof(header->gid)));
959 tar->entry_bytes_remaining = tar_atol(header->size, sizeof(header->size));
960 archive_entry_set_size(entry, tar->entry_bytes_remaining);
961 archive_entry_set_mtime(entry, tar_atol(header->mtime, sizeof(header->mtime)), 0);
963 /* Handle the tar type flag appropriately. */
964 tartype = header->typeflag[0];
967 case '1': /* Hard link */
968 archive_entry_set_hardlink(entry, tar->entry_linkname.s);
970 * The following may seem odd, but: Technically, tar
971 * does not store the file type for a "hard link"
972 * entry, only the fact that it is a hard link. So, I
973 * leave the type zero normally. But, pax interchange
974 * format allows hard links to have data, which
975 * implies that the underlying entry is a regular
978 if (archive_entry_size(entry) > 0)
979 archive_entry_set_filetype(entry, AE_IFREG);
982 * A tricky point: Traditionally, tar readers have
983 * ignored the size field when reading hardlink
984 * entries, and some writers put non-zero sizes even
985 * though the body is empty. POSIX.1-2001 broke with
986 * this tradition by permitting hardlink entries to
987 * store valid bodies in pax interchange format, but
988 * not in ustar format. Since there is no hard and
989 * fast way to distinguish pax interchange from
990 * earlier archives (the 'x' and 'g' entries are
991 * optional, after all), we need a heuristic. Here, I
992 * use the bid function to test whether or not there's
993 * a valid header following. Of course, if we know
994 * this is pax interchange format, then we must obey
997 * This heuristic will only fail for a pax interchange
998 * archive that is storing hardlink bodies, no pax
999 * extended attribute entries have yet occurred, and
1000 * we encounter a hardlink entry for a file that is
1001 * itself an uncompressed tar archive.
1003 if (archive_entry_size(entry) > 0 &&
1004 a->archive.archive_format != ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE &&
1005 archive_read_format_tar_bid(a) > 50) {
1006 archive_entry_set_size(entry, 0);
1007 tar->entry_bytes_remaining = 0;
1010 case '2': /* Symlink */
1011 archive_entry_set_filetype(entry, AE_IFLNK);
1012 archive_entry_set_size(entry, 0);
1013 tar->entry_bytes_remaining = 0;
1014 archive_entry_set_symlink(entry, tar->entry_linkname.s);
1016 case '3': /* Character device */
1017 archive_entry_set_filetype(entry, AE_IFCHR);
1018 archive_entry_set_size(entry, 0);
1019 tar->entry_bytes_remaining = 0;
1021 case '4': /* Block device */
1022 archive_entry_set_filetype(entry, AE_IFBLK);
1023 archive_entry_set_size(entry, 0);
1024 tar->entry_bytes_remaining = 0;
1027 archive_entry_set_filetype(entry, AE_IFDIR);
1028 archive_entry_set_size(entry, 0);
1029 tar->entry_bytes_remaining = 0;
1031 case '6': /* FIFO device */
1032 archive_entry_set_filetype(entry, AE_IFIFO);
1033 archive_entry_set_size(entry, 0);
1034 tar->entry_bytes_remaining = 0;
1036 case 'D': /* GNU incremental directory type */
1038 * No special handling is actually required here.
1039 * It might be nice someday to preprocess the file list and
1040 * provide it to the client, though.
1042 archive_entry_set_filetype(entry, AE_IFDIR);
1044 case 'M': /* GNU "Multi-volume" (remainder of file from last archive)*/
1046 * As far as I can tell, this is just like a regular file
1047 * entry, except that the contents should be _appended_ to
1048 * the indicated file at the indicated offset. This may
1049 * require some API work to fully support.
1052 case 'N': /* Old GNU "long filename" entry. */
1053 /* The body of this entry is a script for renaming
1054 * previously-extracted entries. Ugh. It will never
1055 * be supported by libarchive. */
1056 archive_entry_set_filetype(entry, AE_IFREG);
1058 case 'S': /* GNU sparse files */
1060 * Sparse files are really just regular files with
1061 * sparse information in the extended area.
1064 default: /* Regular file and non-standard types */
1066 * Per POSIX: non-recognized types should always be
1067 * treated as regular files.
1069 archive_entry_set_filetype(entry, AE_IFREG);
1076 * Parse out header elements for "old-style" tar archives.
1079 header_old_tar(struct archive_read *a, struct tar *tar,
1080 struct archive_entry *entry, const void *h)
1082 const struct archive_entry_header_ustar *header;
1084 /* Copy filename over (to ensure null termination). */
1085 header = (const struct archive_entry_header_ustar *)h;
1086 archive_strncpy(&(tar->entry_name), header->name, sizeof(header->name));
1087 archive_entry_set_pathname(entry, tar->entry_name.s);
1089 /* Grab rest of common fields */
1090 header_common(a, tar, entry, h);
1092 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining);
1097 * Parse a file header for a pax extended archive entry.
1100 header_pax_global(struct archive_read *a, struct tar *tar,
1101 struct archive_entry *entry, const void *h)
1105 err = read_body_to_string(a, tar, &(tar->pax_global), h);
1106 if (err != ARCHIVE_OK)
1108 err = tar_read_header(a, tar, entry);
1113 header_pax_extensions(struct archive_read *a, struct tar *tar,
1114 struct archive_entry *entry, const void *h)
1118 err = read_body_to_string(a, tar, &(tar->pax_header), h);
1119 if (err != ARCHIVE_OK)
1122 /* Parse the next header. */
1123 err = tar_read_header(a, tar, entry);
1124 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN))
1128 * TODO: Parse global/default options into 'entry' struct here
1129 * before handling file-specific options.
1131 * This design (parse standard header, then overwrite with pax
1132 * extended attribute data) usually works well, but isn't ideal;
1133 * it would be better to parse the pax extended attributes first
1134 * and then skip any fields in the standard header that were
1135 * defined in the pax header.
1137 err2 = pax_header(a, tar, entry, tar->pax_header.s);
1138 err = err_combine(err, err2);
1139 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining);
1145 * Parse a file header for a Posix "ustar" archive entry. This also
1146 * handles "pax" or "extended ustar" entries.
1149 header_ustar(struct archive_read *a, struct tar *tar,
1150 struct archive_entry *entry, const void *h)
1152 const struct archive_entry_header_ustar *header;
1153 struct archive_string *as;
1155 header = (const struct archive_entry_header_ustar *)h;
1157 /* Copy name into an internal buffer to ensure null-termination. */
1158 as = &(tar->entry_name);
1159 if (header->prefix[0]) {
1160 archive_strncpy(as, header->prefix, sizeof(header->prefix));
1161 if (as->s[archive_strlen(as) - 1] != '/')
1162 archive_strappend_char(as, '/');
1163 archive_strncat(as, header->name, sizeof(header->name));
1165 archive_strncpy(as, header->name, sizeof(header->name));
1167 archive_entry_set_pathname(entry, as->s);
1169 /* Handle rest of common fields. */
1170 header_common(a, tar, entry, h);
1172 /* Handle POSIX ustar fields. */
1173 archive_strncpy(&(tar->entry_uname), header->uname,
1174 sizeof(header->uname));
1175 archive_entry_set_uname(entry, tar->entry_uname.s);
1177 archive_strncpy(&(tar->entry_gname), header->gname,
1178 sizeof(header->gname));
1179 archive_entry_set_gname(entry, tar->entry_gname.s);
1181 /* Parse out device numbers only for char and block specials. */
1182 if (header->typeflag[0] == '3' || header->typeflag[0] == '4') {
1183 archive_entry_set_rdevmajor(entry,
1184 tar_atol(header->rdevmajor, sizeof(header->rdevmajor)));
1185 archive_entry_set_rdevminor(entry,
1186 tar_atol(header->rdevminor, sizeof(header->rdevminor)));
1189 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining);
1196 * Parse the pax extended attributes record.
1198 * Returns non-zero if there's an error in the data.
1201 pax_header(struct archive_read *a, struct tar *tar,
1202 struct archive_entry *entry, char *attr)
1204 size_t attr_length, l, line_length;
1206 wchar_t *key, *wp, *value;
1209 attr_length = strlen(attr);
1211 while (attr_length > 0) {
1212 /* Parse decimal length field at start of line. */
1215 line = p = attr; /* Record start of line. */
1222 if (*p < '0' || *p > '9') {
1223 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1224 "Ignoring malformed pax extended attributes");
1225 return (ARCHIVE_WARN);
1228 line_length += *p - '0';
1229 if (line_length > 999999) {
1230 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1231 "Rejecting pax extended attribute > 1MB");
1232 return (ARCHIVE_WARN);
1239 * Parsed length must be no bigger than available data,
1240 * at least 1, and the last character of the line must
1243 if (line_length > attr_length
1245 || attr[line_length - 1] != '\n')
1247 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1248 "Ignoring malformed pax extended attribute");
1249 return (ARCHIVE_WARN);
1252 /* Ensure pax_entry buffer is big enough. */
1253 if (tar->pax_entry_length <= line_length) {
1254 wchar_t *old_entry = tar->pax_entry;
1256 if (tar->pax_entry_length <= 0)
1257 tar->pax_entry_length = 1024;
1258 while (tar->pax_entry_length <= line_length + 1)
1259 tar->pax_entry_length *= 2;
1261 old_entry = tar->pax_entry;
1262 tar->pax_entry = (wchar_t *)realloc(tar->pax_entry,
1263 tar->pax_entry_length * sizeof(wchar_t));
1264 if (tar->pax_entry == NULL) {
1266 archive_set_error(&a->archive, ENOMEM,
1268 return (ARCHIVE_FATAL);
1272 /* Decode UTF-8 to wchar_t, null-terminate result. */
1273 if (utf8_decode(tar->pax_entry, p,
1274 line_length - (p - attr) - 1)) {
1275 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1276 "Invalid UTF8 character in pax extended attribute");
1277 err = err_combine(err, ARCHIVE_WARN);
1280 /* Null-terminate 'key' value. */
1281 wp = key = tar->pax_entry;
1284 while (*wp && *wp != L'=')
1287 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1288 "Invalid pax extended attributes");
1289 return (ARCHIVE_WARN);
1293 /* Identify null-terminated 'value' portion. */
1296 /* Identify this attribute and set it in the entry. */
1297 err2 = pax_attribute(tar, entry, key, value);
1298 err = err_combine(err, err2);
1300 /* Skip to next line */
1301 attr += line_length;
1302 attr_length -= line_length;
1308 pax_attribute_xattr(struct archive_entry *entry,
1309 wchar_t *name, wchar_t *value)
1311 char *name_decoded, *name_narrow;
1312 void *value_decoded;
1315 if (wcslen(name) < 18 || (wcsncmp(name, L"LIBARCHIVE.xattr.", 17)) != 0)
1320 /* URL-decode name */
1321 name_narrow = wide_to_narrow(name);
1322 if (name_narrow == NULL)
1324 name_decoded = url_decode(name_narrow);
1326 if (name_decoded == NULL)
1329 /* Base-64 decode value */
1330 value_decoded = base64_decode(value, wcslen(value), &value_len);
1331 if (value_decoded == NULL) {
1336 archive_entry_xattr_add_entry(entry, name_decoded,
1337 value_decoded, value_len);
1340 free(value_decoded);
1345 * Parse a single key=value attribute. key/value pointers are
1346 * assumed to point into reasonably long-lived storage.
1348 * Note that POSIX reserves all-lowercase keywords. Vendor-specific
1349 * extensions should always have keywords of the form "VENDOR.attribute"
1350 * In particular, it's quite feasible to support many different
1351 * vendor extensions here. I'm using "LIBARCHIVE" for extensions
1352 * unique to this library.
1354 * Investigate other vendor-specific extensions and see if
1355 * any of them look useful.
1358 pax_attribute(struct tar *tar, struct archive_entry *entry,
1359 wchar_t *key, wchar_t *value)
1366 /* GNU "0.0" sparse pax format. */
1367 if (wcscmp(key, L"GNU.sparse.numblocks") == 0) {
1368 tar->sparse_offset = -1;
1369 tar->sparse_numbytes = -1;
1370 tar->sparse_gnu_major = 0;
1371 tar->sparse_gnu_minor = 0;
1373 if (wcscmp(key, L"GNU.sparse.offset") == 0) {
1374 tar->sparse_offset = tar_atol10(value, wcslen(value));
1375 if (tar->sparse_numbytes != -1) {
1376 gnu_add_sparse_entry(tar,
1377 tar->sparse_offset, tar->sparse_numbytes);
1378 tar->sparse_offset = -1;
1379 tar->sparse_numbytes = -1;
1382 if (wcscmp(key, L"GNU.sparse.numbytes") == 0) {
1383 tar->sparse_numbytes = tar_atol10(value, wcslen(value));
1384 if (tar->sparse_numbytes != -1) {
1385 gnu_add_sparse_entry(tar,
1386 tar->sparse_offset, tar->sparse_numbytes);
1387 tar->sparse_offset = -1;
1388 tar->sparse_numbytes = -1;
1391 if (wcscmp(key, L"GNU.sparse.size") == 0)
1392 archive_entry_set_size(entry,
1393 tar_atol10(value, wcslen(value)));
1395 /* GNU "0.1" sparse pax format. */
1396 if (wcscmp(key, L"GNU.sparse.map") == 0) {
1397 tar->sparse_gnu_major = 0;
1398 tar->sparse_gnu_minor = 1;
1399 if (gnu_sparse_01_parse(tar, value) != ARCHIVE_OK)
1400 return (ARCHIVE_WARN);
1403 /* GNU "1.0" sparse pax format */
1404 if (wcscmp(key, L"GNU.sparse.major") == 0) {
1405 tar->sparse_gnu_major = tar_atol10(value, wcslen(value));
1406 tar->sparse_gnu_pending = 1;
1408 if (wcscmp(key, L"GNU.sparse.minor") == 0) {
1409 tar->sparse_gnu_minor = tar_atol10(value, wcslen(value));
1410 tar->sparse_gnu_pending = 1;
1412 if (wcscmp(key, L"GNU.sparse.name") == 0)
1413 archive_entry_copy_pathname_w(entry, value);
1414 if (wcscmp(key, L"GNU.sparse.realsize") == 0)
1415 archive_entry_set_size(entry,
1416 tar_atol10(value, wcslen(value)));
1419 /* Our extensions */
1420 /* TODO: Handle arbitrary extended attributes... */
1422 if (strcmp(key, "LIBARCHIVE.xxxxxxx")==0)
1423 archive_entry_set_xxxxxx(entry, value);
1425 if (wcsncmp(key, L"LIBARCHIVE.xattr.", 17)==0)
1426 pax_attribute_xattr(entry, key, value);
1429 /* We support some keys used by the "star" archiver */
1430 if (wcscmp(key, L"SCHILY.acl.access")==0)
1431 __archive_entry_acl_parse_w(entry, value,
1432 ARCHIVE_ENTRY_ACL_TYPE_ACCESS);
1433 else if (wcscmp(key, L"SCHILY.acl.default")==0)
1434 __archive_entry_acl_parse_w(entry, value,
1435 ARCHIVE_ENTRY_ACL_TYPE_DEFAULT);
1436 else if (wcscmp(key, L"SCHILY.devmajor")==0)
1437 archive_entry_set_rdevmajor(entry, tar_atol10(value, wcslen(value)));
1438 else if (wcscmp(key, L"SCHILY.devminor")==0)
1439 archive_entry_set_rdevminor(entry, tar_atol10(value, wcslen(value)));
1440 else if (wcscmp(key, L"SCHILY.fflags")==0)
1441 archive_entry_copy_fflags_text_w(entry, value);
1442 else if (wcscmp(key, L"SCHILY.dev")==0)
1443 archive_entry_set_dev(entry, tar_atol10(value, wcslen(value)));
1444 else if (wcscmp(key, L"SCHILY.ino")==0)
1445 archive_entry_set_ino(entry, tar_atol10(value, wcslen(value)));
1446 else if (wcscmp(key, L"SCHILY.nlink")==0)
1447 archive_entry_set_nlink(entry, tar_atol10(value, wcslen(value)));
1450 if (wcscmp(key, L"atime")==0) {
1451 pax_time(value, &s, &n);
1452 archive_entry_set_atime(entry, s, n);
1456 if (wcscmp(key, L"ctime")==0) {
1457 pax_time(value, &s, &n);
1458 archive_entry_set_ctime(entry, s, n);
1459 } else if (wcscmp(key, L"charset")==0) {
1460 /* TODO: Publish charset information in entry. */
1461 } else if (wcscmp(key, L"comment")==0) {
1462 /* TODO: Publish comment in entry. */
1466 if (wcscmp(key, L"gid")==0)
1467 archive_entry_set_gid(entry, tar_atol10(value, wcslen(value)));
1468 else if (wcscmp(key, L"gname")==0)
1469 archive_entry_copy_gname_w(entry, value);
1472 /* pax interchange doesn't distinguish hardlink vs. symlink. */
1473 if (wcscmp(key, L"linkpath")==0) {
1474 if (archive_entry_hardlink(entry))
1475 archive_entry_copy_hardlink_w(entry, value);
1477 archive_entry_copy_symlink_w(entry, value);
1481 if (wcscmp(key, L"mtime")==0) {
1482 pax_time(value, &s, &n);
1483 archive_entry_set_mtime(entry, s, n);
1487 if (wcscmp(key, L"path")==0)
1488 archive_entry_copy_pathname_w(entry, value);
1491 /* POSIX has reserved 'realtime.*' */
1494 /* POSIX has reserved 'security.*' */
1495 /* Someday: if (wcscmp(key, L"security.acl")==0) { ... } */
1496 if (wcscmp(key, L"size")==0) {
1497 tar->entry_bytes_remaining = tar_atol10(value, wcslen(value));
1498 archive_entry_set_size(entry, tar->entry_bytes_remaining);
1500 tar->entry_bytes_remaining = 0;
1504 if (wcscmp(key, L"uid")==0)
1505 archive_entry_set_uid(entry, tar_atol10(value, wcslen(value)));
1506 else if (wcscmp(key, L"uname")==0)
1507 archive_entry_copy_uname_w(entry, value);
1516 * parse a decimal time value, which may include a fractional portion
1519 pax_time(const wchar_t *p, int64_t *ps, long *pn)
1525 int64_t limit, last_digit_limit;
1527 limit = INT64_MAX / 10;
1528 last_digit_limit = INT64_MAX % 10;
1536 while (*p >= '0' && *p <= '9') {
1539 (s == limit && digit > last_digit_limit)) {
1543 s = (s * 10) + digit;
1549 /* Calculate nanoseconds. */
1558 if (*p >= '0' && *p <= '9')
1559 *pn += (*p - '0') * l;
1566 * Parse GNU tar header
1569 header_gnutar(struct archive_read *a, struct tar *tar,
1570 struct archive_entry *entry, const void *h)
1572 const struct archive_entry_header_gnutar *header;
1577 * GNU header is like POSIX ustar, except 'prefix' is
1578 * replaced with some other fields. This also means the
1579 * filename is stored as in old-style archives.
1582 /* Grab fields common to all tar variants. */
1583 header_common(a, tar, entry, h);
1585 /* Copy filename over (to ensure null termination). */
1586 header = (const struct archive_entry_header_gnutar *)h;
1587 archive_strncpy(&(tar->entry_name), header->name,
1588 sizeof(header->name));
1589 archive_entry_set_pathname(entry, tar->entry_name.s);
1591 /* Fields common to ustar and GNU */
1592 /* XXX Can the following be factored out since it's common
1593 * to ustar and gnu tar? Is it okay to move it down into
1594 * header_common, perhaps? */
1595 archive_strncpy(&(tar->entry_uname),
1596 header->uname, sizeof(header->uname));
1597 archive_entry_set_uname(entry, tar->entry_uname.s);
1599 archive_strncpy(&(tar->entry_gname),
1600 header->gname, sizeof(header->gname));
1601 archive_entry_set_gname(entry, tar->entry_gname.s);
1603 /* Parse out device numbers only for char and block specials */
1604 if (header->typeflag[0] == '3' || header->typeflag[0] == '4') {
1605 archive_entry_set_rdevmajor(entry,
1606 tar_atol(header->rdevmajor, sizeof(header->rdevmajor)));
1607 archive_entry_set_rdevminor(entry,
1608 tar_atol(header->rdevminor, sizeof(header->rdevminor)));
1610 archive_entry_set_rdev(entry, 0);
1612 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining);
1614 /* Grab GNU-specific fields. */
1615 archive_entry_set_atime(entry,
1616 tar_atol(header->atime, sizeof(header->atime)), 0);
1617 archive_entry_set_ctime(entry,
1618 tar_atol(header->ctime, sizeof(header->ctime)), 0);
1619 if (header->realsize[0] != 0) {
1620 archive_entry_set_size(entry,
1621 tar_atol(header->realsize, sizeof(header->realsize)));
1624 if (header->sparse[0].offset[0] != 0) {
1625 gnu_sparse_old_read(a, tar, header);
1627 if (header->isextended[0] != 0) {
1636 gnu_add_sparse_entry(struct tar *tar, off_t offset, off_t remaining)
1638 struct sparse_block *p;
1640 p = (struct sparse_block *)malloc(sizeof(*p));
1642 __archive_errx(1, "Out of memory");
1643 memset(p, 0, sizeof(*p));
1644 if (tar->sparse_last != NULL)
1645 tar->sparse_last->next = p;
1647 tar->sparse_list = p;
1648 tar->sparse_last = p;
1650 p->remaining = remaining;
1654 * GNU tar old-format sparse data.
1656 * GNU old-format sparse data is stored in a fixed-field
1657 * format. Offset/size values are 11-byte octal fields (same
1658 * format as 'size' field in ustart header). These are
1659 * stored in the header, allocating subsequent header blocks
1660 * as needed. Extending the header in this way is a pretty
1661 * severe POSIX violation; this design has earned GNU tar a
1666 gnu_sparse_old_read(struct archive_read *a, struct tar *tar,
1667 const struct archive_entry_header_gnutar *header)
1672 struct gnu_sparse sparse[21];
1676 const struct extended *ext;
1678 gnu_sparse_old_parse(tar, header->sparse, 4);
1679 if (header->isextended[0] == 0)
1680 return (ARCHIVE_OK);
1683 bytes_read = (a->decompressor->read_ahead)(a, &data, 512);
1685 return (ARCHIVE_FATAL);
1686 if (bytes_read < 512) {
1687 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1688 "Truncated tar archive "
1689 "detected while reading sparse file data");
1690 return (ARCHIVE_FATAL);
1692 (a->decompressor->consume)(a, 512);
1693 ext = (const struct extended *)data;
1694 gnu_sparse_old_parse(tar, ext->sparse, 21);
1695 } while (ext->isextended[0] != 0);
1696 if (tar->sparse_list != NULL)
1697 tar->entry_offset = tar->sparse_list->offset;
1698 return (ARCHIVE_OK);
1702 gnu_sparse_old_parse(struct tar *tar,
1703 const struct gnu_sparse *sparse, int length)
1705 while (length > 0 && sparse->offset[0] != 0) {
1706 gnu_add_sparse_entry(tar,
1707 tar_atol(sparse->offset, sizeof(sparse->offset)),
1708 tar_atol(sparse->numbytes, sizeof(sparse->numbytes)));
1715 * GNU tar sparse format 0.0
1717 * Beginning with GNU tar 1.15, sparse files are stored using
1718 * information in the pax extended header. The GNU tar maintainers
1719 * have gone through a number of variations in the process of working
1720 * out this scheme; furtunately, they're all numbered.
1722 * Sparse format 0.0 uses attribute GNU.sparse.numblocks to store the
1723 * number of blocks, and GNU.sparse.offset/GNU.sparse.numbytes to
1724 * store offset/size for each block. The repeated instances of these
1725 * latter fields violate the pax specification (which frowns on
1726 * duplicate keys), so this format was quickly replaced.
1730 * GNU tar sparse format 0.1
1732 * This version replaced the offset/numbytes attributes with
1733 * a single "map" attribute that stored a list of integers. This
1734 * format had two problems: First, the "map" attribute could be very
1735 * long, which caused problems for some implementations. More
1736 * importantly, the sparse data was lost when extracted by archivers
1737 * that didn't recognize this extension.
1741 gnu_sparse_01_parse(struct tar *tar, const wchar_t *p)
1744 off_t offset = -1, size = -1;
1748 while (*e != '\0' && *e != ',') {
1749 if (*e < '0' || *e > '9')
1750 return (ARCHIVE_WARN);
1754 offset = tar_atol10(p, e - p);
1756 return (ARCHIVE_WARN);
1758 size = tar_atol10(p, e - p);
1760 return (ARCHIVE_WARN);
1761 gnu_add_sparse_entry(tar, offset, size);
1765 return (ARCHIVE_OK);
1771 * GNU tar sparse format 1.0
1773 * The idea: The offset/size data is stored as a series of base-10
1774 * ASCII numbers prepended to the file data, so that dearchivers that
1775 * don't support this format will extract the block map along with the
1776 * data and a separate post-process can restore the sparseness.
1778 * Unfortunately, GNU tar 1.16 adds bogus padding to the end of the
1779 * entry that depends on the size of the map; this means we have to
1780 * parse the sparse map when we read the header (otherwise, entry_skip
1781 * will fail). This is why sparse_10_read is called from read_header
1782 * above, instead of at the beginning of read_data, where it "should"
1785 * This variant also replaced GNU.sparse.size with GNU.sparse.realsize
1786 * and introduced the GNU.sparse.major/GNU.sparse.minor attributes.
1790 * Read the next line from the input, and parse it as a decimal
1791 * integer followed by '\n'. Returns positive integer value or
1792 * negative on error.
1795 gnu_sparse_10_atol(struct archive_read *a, struct tar *tar,
1796 ssize_t *total_read)
1798 int64_t l, limit, last_digit_limit;
1804 limit = INT64_MAX / base;
1805 last_digit_limit = INT64_MAX % base;
1807 bytes_read = readline(a, tar, &p);
1808 if (bytes_read <= 0)
1809 return (ARCHIVE_FATAL);
1810 *total_read += bytes_read;
1813 while (bytes_read > 0) {
1816 if (*p < '0' || *p >= '0' + base)
1817 return (ARCHIVE_WARN);
1819 if (l > limit || (l == limit && digit > last_digit_limit))
1820 l = UINT64_MAX; /* Truncate on overflow. */
1822 l = (l * base) + digit;
1826 /* TODO: Error message. */
1827 return (ARCHIVE_WARN);
1831 * Returns number of bytes consumed to read the sparse block data.
1834 gnu_sparse_10_read(struct archive_read *a, struct tar *tar)
1836 ssize_t bytes_read = 0;
1838 off_t offset, size, to_skip;
1840 /* Parse entries. */
1841 entries = gnu_sparse_10_atol(a, tar, &bytes_read);
1843 return (ARCHIVE_FATAL);
1844 /* Parse the individual entries. */
1845 while (entries-- > 0) {
1846 /* Parse offset/size */
1847 offset = gnu_sparse_10_atol(a, tar, &bytes_read);
1849 return (ARCHIVE_FATAL);
1850 size = gnu_sparse_10_atol(a, tar, &bytes_read);
1852 return (ARCHIVE_FATAL);
1853 /* Add a new sparse entry. */
1854 gnu_add_sparse_entry(tar, offset, size);
1856 /* Skip rest of block... */
1857 to_skip = 0x1ff & -bytes_read;
1858 if (to_skip != (a->decompressor->skip)(a, to_skip))
1859 return (ARCHIVE_FATAL);
1860 return (bytes_read + to_skip);
1864 * Convert text->integer.
1866 * Traditional tar formats (including POSIX) specify base-8 for
1867 * all of the standard numeric fields. This is a significant limitation
1869 * = file size is limited to 8GB
1870 * = rdevmajor and rdevminor are limited to 21 bits
1871 * = uid/gid are limited to 21 bits
1873 * There are two workarounds for this:
1874 * = pax extended headers, which use variable-length string fields
1875 * = GNU tar and STAR both allow either base-8 or base-256 in
1876 * most fields. The high bit is set to indicate base-256.
1878 * On read, this implementation supports both extensions.
1881 tar_atol(const char *p, unsigned char_cnt)
1884 * Technically, GNU tar considers a field to be in base-256
1885 * only if the first byte is 0xff or 0x80.
1888 return (tar_atol256(p, char_cnt));
1889 return (tar_atol8(p, char_cnt));
1893 * Note that this implementation does not (and should not!) obey
1894 * locale settings; you cannot simply substitute strtol here, since
1895 * it does obey locale.
1898 tar_atol8(const char *p, unsigned char_cnt)
1900 int64_t l, limit, last_digit_limit;
1901 int digit, sign, base;
1904 limit = INT64_MAX / base;
1905 last_digit_limit = INT64_MAX % base;
1907 while (*p == ' ' || *p == '\t')
1917 while (digit >= 0 && digit < base && char_cnt-- > 0) {
1918 if (l>limit || (l == limit && digit > last_digit_limit)) {
1919 l = UINT64_MAX; /* Truncate on overflow. */
1922 l = (l * base) + digit;
1925 return (sign < 0) ? -l : l;
1929 * Note that this implementation does not (and should not!) obey
1930 * locale settings; you cannot simply substitute strtol here, since
1931 * it does obey locale.
1934 tar_atol10(const wchar_t *p, unsigned char_cnt)
1936 int64_t l, limit, last_digit_limit;
1937 int base, digit, sign;
1940 limit = INT64_MAX / base;
1941 last_digit_limit = INT64_MAX % base;
1943 while (*p == ' ' || *p == '\t')
1953 while (digit >= 0 && digit < base && char_cnt-- > 0) {
1954 if (l > limit || (l == limit && digit > last_digit_limit)) {
1955 l = UINT64_MAX; /* Truncate on overflow. */
1958 l = (l * base) + digit;
1961 return (sign < 0) ? -l : l;
1965 * Parse a base-256 integer. This is just a straight signed binary
1966 * value in big-endian order, except that the high-order bit is
1967 * ignored. Remember that "int64_t" may or may not be exactly 64
1968 * bits; the implementation here tries to avoid making any assumptions
1969 * about the actual size of an int64_t. It does assume we're using
1970 * twos-complement arithmetic, though.
1973 tar_atol256(const char *_p, unsigned char_cnt)
1975 int64_t l, upper_limit, lower_limit;
1976 const unsigned char *p = (const unsigned char *)_p;
1978 upper_limit = INT64_MAX / 256;
1979 lower_limit = INT64_MIN / 256;
1981 /* Pad with 1 or 0 bits, depending on sign. */
1982 if ((0x40 & *p) == 0x40)
1986 l = (l << 6) | (0x3f & *p++);
1987 while (--char_cnt > 0) {
1988 if (l > upper_limit) {
1989 l = INT64_MAX; /* Truncate on overflow */
1991 } else if (l < lower_limit) {
1995 l = (l << 8) | (0xff & (int64_t)*p++);
2001 * Returns length of line (including trailing newline)
2002 * or negative on error. 'start' argument is updated to
2003 * point to first character of line. This avoids copying
2007 readline(struct archive_read *a, struct tar *tar, const char **start)
2010 ssize_t total_size = 0;
2015 bytes_read = (a->decompressor->read_ahead)(a, &t, 1);
2016 if (bytes_read <= 0)
2017 return (ARCHIVE_FATAL);
2018 s = t; /* Start of line? */
2019 p = memchr(t, '\n', bytes_read);
2020 /* If we found '\n' in the read buffer, return pointer to that. */
2022 bytes_read = 1 + ((const char *)p) - s;
2023 (a->decompressor->consume)(a, bytes_read);
2025 return (bytes_read);
2027 /* Otherwise, we need to accumulate in a line buffer. */
2029 if (archive_string_ensure(&tar->line, total_size + bytes_read) == NULL) {
2030 archive_set_error(&a->archive, ENOMEM,
2031 "Can't allocate working buffer");
2032 return (ARCHIVE_FATAL);
2034 memcpy(tar->line.s + total_size, t, bytes_read);
2035 (a->decompressor->consume)(a, bytes_read);
2036 total_size += bytes_read;
2037 /* If we found '\n', clean up and return. */
2039 *start = tar->line.s;
2040 return (total_size);
2042 /* Read some more. */
2043 bytes_read = (a->decompressor->read_ahead)(a, &t, 1);
2044 if (bytes_read <= 0)
2045 return (ARCHIVE_FATAL);
2046 s = t; /* Start of line? */
2047 p = memchr(t, '\n', bytes_read);
2048 /* If we found '\n', trim the read. */
2050 bytes_read = 1 + ((const char *)p) - s;
2056 utf8_decode(wchar_t *dest, const char *src, size_t length)
2062 while (length > 0) {
2063 n = UTF8_mbrtowc(dest, src, length);
2075 * Copied and simplified from FreeBSD libc/locale.
2078 UTF8_mbrtowc(wchar_t *pwc, const char *s, size_t n)
2080 int ch, i, len, mask;
2083 if (s == NULL || n == 0 || pwc == NULL)
2087 * Determine the number of octets that make up this character from
2088 * the first octet, and a mask that extracts the interesting bits of
2091 ch = (unsigned char)*s;
2092 if ((ch & 0x80) == 0) {
2095 } else if ((ch & 0xe0) == 0xc0) {
2098 } else if ((ch & 0xf0) == 0xe0) {
2101 } else if ((ch & 0xf8) == 0xf0) {
2104 } else if ((ch & 0xfc) == 0xf8) {
2107 } else if ((ch & 0xfe) == 0xfc) {
2111 /* Invalid first byte; convert to '?' */
2116 if (n < (size_t)len) {
2117 /* Invalid first byte; convert to '?' */
2123 * Decode the octet sequence representing the character in chunks
2124 * of 6 bits, most significant first.
2126 wch = (unsigned char)*s++ & mask;
2129 if ((*s & 0xc0) != 0x80) {
2130 /* Invalid intermediate byte; consume one byte and
2139 /* Assign the value to the output; out-of-range values
2140 * just get truncated. */
2141 *pwc = (wchar_t)wch;
2144 * If platform has WCHAR_MAX, we can do something
2145 * more sensible with out-of-range values.
2147 if (wch >= WCHAR_MAX)
2150 /* Return number of bytes input consumed: 0 for end-of-string. */
2151 return (wch == L'\0' ? 0 : len);
2156 * base64_decode - Base64 decode
2158 * This accepts most variations of base-64 encoding, including:
2159 * * with or without line breaks
2160 * * with or without the final group padded with '=' or '_' characters
2161 * (The most economical Base-64 variant does not pad the last group and
2162 * omits line breaks; RFC1341 used for MIME requires both.)
2165 base64_decode(const wchar_t *src, size_t len, size_t *out_len)
2167 static const unsigned char digits[64] = {
2168 'A','B','C','D','E','F','G','H','I','J','K','L','M','N',
2169 'O','P','Q','R','S','T','U','V','W','X','Y','Z','a','b',
2170 'c','d','e','f','g','h','i','j','k','l','m','n','o','p',
2171 'q','r','s','t','u','v','w','x','y','z','0','1','2','3',
2172 '4','5','6','7','8','9','+','/' };
2173 static unsigned char decode_table[128];
2176 /* If the decode table is not yet initialized, prepare it. */
2177 if (decode_table[digits[1]] != 1) {
2179 memset(decode_table, 0xff, sizeof(decode_table));
2180 for (i = 0; i < sizeof(digits); i++)
2181 decode_table[digits[i]] = i;
2184 /* Allocate enough space to hold the entire output. */
2185 /* Note that we may not use all of this... */
2186 out = (char *)malloc((len * 3 + 3) / 4);
2194 /* Collect the next group of (up to) four characters. */
2197 while (group_size < 4 && len > 0) {
2198 /* '=' or '_' padding indicates final group. */
2199 if (*src == '=' || *src == '_') {
2203 /* Skip illegal characters (including line breaks) */
2204 if (*src > 127 || *src < 32
2205 || decode_table[*src] == 0xff) {
2211 v |= decode_table[*src++];
2215 /* Align a short group properly. */
2216 v <<= 6 * (4 - group_size);
2217 /* Unpack the group we just collected. */
2218 switch (group_size) {
2219 case 4: d[2] = v & 0xff;
2221 case 3: d[1] = (v >> 8) & 0xff;
2223 case 2: d[0] = (v >> 16) & 0xff;
2225 case 1: /* this is invalid! */
2228 d += group_size * 3 / 4;
2236 * This is a little tricky because the C99 standard wcstombs()
2237 * function returns the number of bytes that were converted,
2238 * not the number that should be converted. As a result,
2239 * we can never accurately size the output buffer (without
2240 * doing a tedious output size calculation in advance).
2241 * This approach (try a conversion, then try again if it fails)
2242 * will almost always succeed on the first try, and is thus
2243 * much faster, at the cost of sometimes requiring multiple
2244 * passes while we expand the buffer.
2247 wide_to_narrow(const wchar_t *wval)
2249 int converted_length;
2250 /* Guess an output buffer size and try the conversion. */
2251 int alloc_length = wcslen(wval) * 3;
2252 char *mbs_val = (char *)malloc(alloc_length + 1);
2253 if (mbs_val == NULL)
2255 converted_length = wcstombs(mbs_val, wval, alloc_length);
2257 /* If we exhausted the buffer, resize and try again. */
2258 while (converted_length >= alloc_length) {
2261 mbs_val = (char *)malloc(alloc_length + 1);
2262 if (mbs_val == NULL)
2264 converted_length = wcstombs(mbs_val, wval, alloc_length);
2267 /* Ensure a trailing null and return the final string. */
2268 mbs_val[alloc_length] = '\0';
2273 url_decode(const char *in)
2278 out = (char *)malloc(strlen(in) + 1);
2281 for (s = in, d = out; *s != '\0'; ) {
2283 /* Try to convert % escape */
2284 int digit1 = tohex(s[1]);
2285 int digit2 = tohex(s[2]);
2286 if (digit1 >= 0 && digit2 >= 0) {
2287 /* Looks good, consume three chars */
2289 /* Convert output */
2290 *d++ = ((digit1 << 4) | digit2);
2293 /* Else fall through and treat '%' as normal char */
2304 if (c >= '0' && c <= '9')
2306 else if (c >= 'A' && c <= 'F')
2307 return (c - 'A' + 10);
2308 else if (c >= 'a' && c <= 'f')
2309 return (c - 'a' + 10);