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("$FreeBSD$");
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"
75 #define tar_min(a,b) ((a) < (b) ? (a) : (b))
78 * Layout of POSIX 'ustar' tar header.
80 struct archive_entry_header_ustar {
89 char linkname[100]; /* "old format" header ends here */
90 char magic[6]; /* For POSIX: "ustar\0" */
91 char version[2]; /* For POSIX: "00" */
100 * Structure of GNU tar header
107 struct archive_entry_header_gnutar {
117 char magic[8]; /* "ustar \0" (note blank/blank/null at end) */
127 struct gnu_sparse sparse[4];
131 * Old GNU format doesn't use POSIX 'prefix' field; they use
132 * the 'L' (longname) entry instead.
137 * Data specific to this format.
139 struct sparse_block {
140 struct sparse_block *next;
146 struct archive_string acl_text;
147 struct archive_string entry_pathname;
148 /* For "GNU.sparse.name" and other similar path extensions. */
149 struct archive_string entry_pathname_override;
150 struct archive_string entry_linkpath;
151 struct archive_string entry_uname;
152 struct archive_string entry_gname;
153 struct archive_string longlink;
154 struct archive_string longname;
155 struct archive_string pax_header;
156 struct archive_string pax_global;
157 struct archive_string line;
158 int pax_hdrcharset_binary;
160 size_t pax_entry_length;
161 int header_recursion_depth;
162 off_t entry_bytes_remaining;
166 struct sparse_block *sparse_list;
167 struct sparse_block *sparse_last;
168 int64_t sparse_offset;
169 int64_t sparse_numbytes;
170 int sparse_gnu_major;
171 int sparse_gnu_minor;
172 char sparse_gnu_pending;
175 static ssize_t UTF8_mbrtowc(wchar_t *pwc, const char *s, size_t n);
176 static int archive_block_is_null(const unsigned char *p);
177 static char *base64_decode(const char *, size_t, size_t *);
178 static void gnu_add_sparse_entry(struct tar *,
179 off_t offset, off_t remaining);
180 static void gnu_clear_sparse_list(struct tar *);
181 static int gnu_sparse_old_read(struct archive_read *, struct tar *,
182 const struct archive_entry_header_gnutar *header);
183 static void gnu_sparse_old_parse(struct tar *,
184 const struct gnu_sparse *sparse, int length);
185 static int gnu_sparse_01_parse(struct tar *, const char *);
186 static ssize_t gnu_sparse_10_read(struct archive_read *, struct tar *);
187 static int header_Solaris_ACL(struct archive_read *, struct tar *,
188 struct archive_entry *, const void *);
189 static int header_common(struct archive_read *, struct tar *,
190 struct archive_entry *, const void *);
191 static int header_old_tar(struct archive_read *, struct tar *,
192 struct archive_entry *, const void *);
193 static int header_pax_extensions(struct archive_read *, struct tar *,
194 struct archive_entry *, const void *);
195 static int header_pax_global(struct archive_read *, struct tar *,
196 struct archive_entry *, const void *h);
197 static int header_longlink(struct archive_read *, struct tar *,
198 struct archive_entry *, const void *h);
199 static int header_longname(struct archive_read *, struct tar *,
200 struct archive_entry *, const void *h);
201 static int header_volume(struct archive_read *, struct tar *,
202 struct archive_entry *, const void *h);
203 static int header_ustar(struct archive_read *, struct tar *,
204 struct archive_entry *, const void *h);
205 static int header_gnutar(struct archive_read *, struct tar *,
206 struct archive_entry *, const void *h);
207 static int archive_read_format_tar_bid(struct archive_read *);
208 static int archive_read_format_tar_cleanup(struct archive_read *);
209 static int archive_read_format_tar_read_data(struct archive_read *a,
210 const void **buff, size_t *size, off_t *offset);
211 static int archive_read_format_tar_skip(struct archive_read *a);
212 static int archive_read_format_tar_read_header(struct archive_read *,
213 struct archive_entry *);
214 static int checksum(struct archive_read *, const void *);
215 static int pax_attribute(struct tar *, struct archive_entry *,
216 char *key, char *value);
217 static int pax_header(struct archive_read *, struct tar *,
218 struct archive_entry *, char *attr);
219 static void pax_time(const char *, int64_t *sec, long *nanos);
220 static ssize_t readline(struct archive_read *, struct tar *, const char **,
222 static int read_body_to_string(struct archive_read *, struct tar *,
223 struct archive_string *, const void *h);
224 static int64_t tar_atol(const char *, unsigned);
225 static int64_t tar_atol10(const char *, unsigned);
226 static int64_t tar_atol256(const char *, unsigned);
227 static int64_t tar_atol8(const char *, unsigned);
228 static int tar_read_header(struct archive_read *, struct tar *,
229 struct archive_entry *);
230 static int tohex(int c);
231 static char *url_decode(const char *);
232 static wchar_t *utf8_decode(struct tar *, const char *, size_t length);
235 archive_read_support_format_gnutar(struct archive *a)
237 return (archive_read_support_format_tar(a));
242 archive_read_support_format_tar(struct archive *_a)
244 struct archive_read *a = (struct archive_read *)_a;
248 tar = (struct tar *)malloc(sizeof(*tar));
250 archive_set_error(&a->archive, ENOMEM,
251 "Can't allocate tar data");
252 return (ARCHIVE_FATAL);
254 memset(tar, 0, sizeof(*tar));
256 r = __archive_read_register_format(a, tar, "tar",
257 archive_read_format_tar_bid,
259 archive_read_format_tar_read_header,
260 archive_read_format_tar_read_data,
261 archive_read_format_tar_skip,
262 archive_read_format_tar_cleanup);
270 archive_read_format_tar_cleanup(struct archive_read *a)
274 tar = (struct tar *)(a->format->data);
275 gnu_clear_sparse_list(tar);
276 archive_string_free(&tar->acl_text);
277 archive_string_free(&tar->entry_pathname);
278 archive_string_free(&tar->entry_pathname_override);
279 archive_string_free(&tar->entry_linkpath);
280 archive_string_free(&tar->entry_uname);
281 archive_string_free(&tar->entry_gname);
282 archive_string_free(&tar->line);
283 archive_string_free(&tar->pax_global);
284 archive_string_free(&tar->pax_header);
285 archive_string_free(&tar->longname);
286 archive_string_free(&tar->longlink);
287 free(tar->pax_entry);
289 (a->format->data) = NULL;
295 archive_read_format_tar_bid(struct archive_read *a)
299 const struct archive_entry_header_ustar *header;
303 /* Now let's look at the actual header and see if it matches. */
304 h = __archive_read_ahead(a, 512, NULL);
308 /* If it's an end-of-archive mark, we can handle it. */
309 if ((*(const char *)h) == 0
310 && archive_block_is_null((const unsigned char *)h)) {
312 * Usually, I bid the number of bits verified, but
313 * in this case, 4096 seems excessive so I picked 10 as
314 * an arbitrary but reasonable-seeming value.
319 /* If it's not an end-of-archive mark, it must have a valid checksum.*/
322 bid += 48; /* Checksum is usually 6 octal digits. */
324 header = (const struct archive_entry_header_ustar *)h;
326 /* Recognize POSIX formats. */
327 if ((memcmp(header->magic, "ustar\0", 6) == 0)
328 &&(memcmp(header->version, "00", 2)==0))
331 /* Recognize GNU tar format. */
332 if ((memcmp(header->magic, "ustar ", 6) == 0)
333 &&(memcmp(header->version, " \0", 2)==0))
336 /* Type flag must be null, digit or A-Z, a-z. */
337 if (header->typeflag[0] != 0 &&
338 !( header->typeflag[0] >= '0' && header->typeflag[0] <= '9') &&
339 !( header->typeflag[0] >= 'A' && header->typeflag[0] <= 'Z') &&
340 !( header->typeflag[0] >= 'a' && header->typeflag[0] <= 'z') )
342 bid += 2; /* 6 bits of variation in an 8-bit field leaves 2 bits. */
344 /* Sanity check: Look at first byte of mode field. */
345 switch (255 & (unsigned)header->mode[0]) {
347 /* Base-256 value: No further verification possible! */
349 case ' ': /* Not recommended, but not illegal, either. */
351 case '0': case '1': case '2': case '3':
352 case '4': case '5': case '6': case '7':
354 /* TODO: Check format of remainder of this field. */
357 /* Not a valid mode; bail out here. */
360 /* TODO: Sanity test uid/gid/size/mtime/rdevmajor/rdevminor fields. */
366 * The function invoked by archive_read_header(). This
367 * just sets up a few things and then calls the internal
368 * tar_read_header() function below.
371 archive_read_format_tar_read_header(struct archive_read *a,
372 struct archive_entry *entry)
375 * When converting tar archives to cpio archives, it is
376 * essential that each distinct file have a distinct inode
377 * number. To simplify this, we keep a static count here to
378 * assign fake dev/inode numbers to each tar entry. Note that
379 * pax format archives may overwrite this with something more
382 * Ideally, we would track every file read from the archive so
383 * that we could assign the same dev/ino pair to hardlinks,
384 * but the memory required to store a complete lookup table is
385 * probably not worthwhile just to support the relatively
386 * obscure tar->cpio conversion case.
388 static int default_inode;
389 static int default_dev;
391 struct sparse_block *sp;
396 /* Assign default device/inode values. */
397 archive_entry_set_dev(entry, 1 + default_dev); /* Don't use zero. */
398 archive_entry_set_ino(entry, ++default_inode); /* Don't use zero. */
399 /* Limit generated st_ino number to 16 bits. */
400 if (default_inode >= 0xffff) {
405 tar = (struct tar *)(a->format->data);
406 tar->entry_offset = 0;
407 while (tar->sparse_list != NULL) {
408 sp = tar->sparse_list;
409 tar->sparse_list = sp->next;
412 tar->sparse_last = NULL;
413 tar->realsize = -1; /* Mark this as "unset" */
415 r = tar_read_header(a, tar, entry);
418 * "non-sparse" files are really just sparse files with
421 if (tar->sparse_list == NULL)
422 gnu_add_sparse_entry(tar, 0, tar->entry_bytes_remaining);
424 if (r == ARCHIVE_OK) {
426 * "Regular" entry with trailing '/' is really
427 * directory: This is needed for certain old tar
428 * variants and even for some broken newer ones.
430 p = archive_entry_pathname(entry);
432 if (archive_entry_filetype(entry) == AE_IFREG
434 archive_entry_set_filetype(entry, AE_IFDIR);
440 archive_read_format_tar_read_data(struct archive_read *a,
441 const void **buff, size_t *size, off_t *offset)
445 struct sparse_block *p;
447 tar = (struct tar *)(a->format->data);
449 if (tar->sparse_gnu_pending) {
450 if (tar->sparse_gnu_major == 1 && tar->sparse_gnu_minor == 0) {
451 tar->sparse_gnu_pending = 0;
452 /* Read initial sparse map. */
453 bytes_read = gnu_sparse_10_read(a, tar);
454 tar->entry_bytes_remaining -= bytes_read;
460 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
461 "Unrecognized GNU sparse file format");
462 return (ARCHIVE_WARN);
464 tar->sparse_gnu_pending = 0;
467 /* Remove exhausted entries from sparse list. */
468 while (tar->sparse_list != NULL &&
469 tar->sparse_list->remaining == 0) {
470 p = tar->sparse_list;
471 tar->sparse_list = p->next;
475 /* If we're at end of file, return EOF. */
476 if (tar->sparse_list == NULL || tar->entry_bytes_remaining == 0) {
477 if (__archive_read_skip(a, tar->entry_padding) < 0)
478 return (ARCHIVE_FATAL);
479 tar->entry_padding = 0;
482 *offset = tar->realsize;
483 return (ARCHIVE_EOF);
486 *buff = __archive_read_ahead(a, 1, &bytes_read);
488 return (ARCHIVE_FATAL);
490 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
491 "Truncated tar archive");
492 return (ARCHIVE_FATAL);
494 if (bytes_read > tar->entry_bytes_remaining)
495 bytes_read = tar->entry_bytes_remaining;
496 /* Don't read more than is available in the
497 * current sparse block. */
498 if (tar->sparse_list->remaining < bytes_read)
499 bytes_read = tar->sparse_list->remaining;
501 *offset = tar->sparse_list->offset;
502 tar->sparse_list->remaining -= bytes_read;
503 tar->sparse_list->offset += bytes_read;
504 tar->entry_bytes_remaining -= bytes_read;
505 __archive_read_consume(a, bytes_read);
510 archive_read_format_tar_skip(struct archive_read *a)
515 tar = (struct tar *)(a->format->data);
518 * Compression layer skip functions are required to either skip the
519 * length requested or fail, so we can rely upon the entire entry
520 * plus padding being skipped.
522 bytes_skipped = __archive_read_skip(a,
523 tar->entry_bytes_remaining + tar->entry_padding);
524 if (bytes_skipped < 0)
525 return (ARCHIVE_FATAL);
527 tar->entry_bytes_remaining = 0;
528 tar->entry_padding = 0;
530 /* Free the sparse list. */
531 gnu_clear_sparse_list(tar);
537 * This function recursively interprets all of the headers associated
538 * with a single entry.
541 tar_read_header(struct archive_read *a, struct tar *tar,
542 struct archive_entry *entry)
547 const struct archive_entry_header_ustar *header;
549 /* Read 512-byte header record */
550 h = __archive_read_ahead(a, 512, &bytes);
553 if (bytes < 512) { /* Short read or EOF. */
554 /* Try requesting just one byte and see what happens. */
555 h = __archive_read_ahead(a, 1, &bytes);
558 * The archive ends at a 512-byte boundary but
559 * without a proper end-of-archive marker.
560 * Yes, there are tar writers that do this;
561 * hold our nose and accept it.
563 return (ARCHIVE_EOF);
565 /* Archive ends with a partial block; this is bad. */
566 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
567 "Truncated tar archive");
568 return (ARCHIVE_FATAL);
570 __archive_read_consume(a, 512);
573 /* Check for end-of-archive mark. */
574 if (((*(const char *)h)==0) && archive_block_is_null((const unsigned char *)h)) {
575 /* Try to consume a second all-null record, as well. */
576 h = __archive_read_ahead(a, 512, NULL);
578 __archive_read_consume(a, 512);
579 archive_set_error(&a->archive, 0, NULL);
580 if (a->archive.archive_format_name == NULL) {
581 a->archive.archive_format = ARCHIVE_FORMAT_TAR;
582 a->archive.archive_format_name = "tar";
584 return (ARCHIVE_EOF);
588 * Note: If the checksum fails and we return ARCHIVE_RETRY,
589 * then the client is likely to just retry. This is a very
590 * crude way to search for the next valid header!
592 * TODO: Improve this by implementing a real header scan.
594 if (!checksum(a, h)) {
595 archive_set_error(&a->archive, EINVAL, "Damaged tar archive");
596 return (ARCHIVE_RETRY); /* Retryable: Invalid header */
599 if (++tar->header_recursion_depth > 32) {
600 archive_set_error(&a->archive, EINVAL, "Too many special headers");
601 return (ARCHIVE_WARN);
604 /* Determine the format variant. */
605 header = (const struct archive_entry_header_ustar *)h;
606 switch(header->typeflag[0]) {
607 case 'A': /* Solaris tar ACL */
608 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE;
609 a->archive.archive_format_name = "Solaris tar";
610 err = header_Solaris_ACL(a, tar, entry, h);
612 case 'g': /* POSIX-standard 'g' header. */
613 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE;
614 a->archive.archive_format_name = "POSIX pax interchange format";
615 err = header_pax_global(a, tar, entry, h);
617 case 'K': /* Long link name (GNU tar, others) */
618 err = header_longlink(a, tar, entry, h);
620 case 'L': /* Long filename (GNU tar, others) */
621 err = header_longname(a, tar, entry, h);
623 case 'V': /* GNU volume header */
624 err = header_volume(a, tar, entry, h);
626 case 'X': /* Used by SUN tar; same as 'x'. */
627 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE;
628 a->archive.archive_format_name =
629 "POSIX pax interchange format (Sun variant)";
630 err = header_pax_extensions(a, tar, entry, h);
632 case 'x': /* POSIX-standard 'x' header. */
633 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE;
634 a->archive.archive_format_name = "POSIX pax interchange format";
635 err = header_pax_extensions(a, tar, entry, h);
638 if (memcmp(header->magic, "ustar \0", 8) == 0) {
639 a->archive.archive_format = ARCHIVE_FORMAT_TAR_GNUTAR;
640 a->archive.archive_format_name = "GNU tar format";
641 err = header_gnutar(a, tar, entry, h);
642 } else if (memcmp(header->magic, "ustar", 5) == 0) {
643 if (a->archive.archive_format != ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE) {
644 a->archive.archive_format = ARCHIVE_FORMAT_TAR_USTAR;
645 a->archive.archive_format_name = "POSIX ustar format";
647 err = header_ustar(a, tar, entry, h);
649 a->archive.archive_format = ARCHIVE_FORMAT_TAR;
650 a->archive.archive_format_name = "tar (non-POSIX)";
651 err = header_old_tar(a, tar, entry, h);
654 --tar->header_recursion_depth;
655 /* We return warnings or success as-is. Anything else is fatal. */
656 if (err == ARCHIVE_WARN || err == ARCHIVE_OK)
658 if (err == ARCHIVE_EOF)
659 /* EOF when recursively reading a header is bad. */
660 archive_set_error(&a->archive, EINVAL, "Damaged tar archive");
661 return (ARCHIVE_FATAL);
665 * Return true if block checksum is correct.
668 checksum(struct archive_read *a, const void *h)
670 const unsigned char *bytes;
671 const struct archive_entry_header_ustar *header;
674 (void)a; /* UNUSED */
675 bytes = (const unsigned char *)h;
676 header = (const struct archive_entry_header_ustar *)h;
679 * Test the checksum. Note that POSIX specifies _unsigned_
680 * bytes for this calculation.
682 sum = tar_atol(header->checksum, sizeof(header->checksum));
684 for (i = 0; i < 148; i++)
685 check += (unsigned char)bytes[i];
689 check += (unsigned char)bytes[i];
694 * Repeat test with _signed_ bytes, just in case this archive
695 * was created by an old BSD, Solaris, or HP-UX tar with a
696 * broken checksum calculation.
699 for (i = 0; i < 148; i++)
700 check += (signed char)bytes[i];
704 check += (signed char)bytes[i];
712 * Return true if this block contains only nulls.
715 archive_block_is_null(const unsigned char *p)
719 for (i = 0; i < 512; i++)
726 * Interpret 'A' Solaris ACL header
729 header_Solaris_ACL(struct archive_read *a, struct tar *tar,
730 struct archive_entry *entry, const void *h)
732 const struct archive_entry_header_ustar *header;
739 * read_body_to_string adds a NUL terminator, but we need a little
740 * more to make sure that we don't overrun acl_text later.
742 header = (const struct archive_entry_header_ustar *)h;
743 size = tar_atol(header->size, sizeof(header->size));
744 err = read_body_to_string(a, tar, &(tar->acl_text), h);
745 if (err != ARCHIVE_OK)
747 err = tar_read_header(a, tar, entry);
748 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN))
751 /* Skip leading octal number. */
752 /* XXX TODO: Parse the octal number and sanity-check it. */
753 p = acl = tar->acl_text.s;
754 while (*p != '\0' && p < acl + size)
758 if (p >= acl + size) {
759 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
760 "Malformed Solaris ACL attribute");
761 return(ARCHIVE_WARN);
764 /* Skip leading octal number. */
768 while (*p != '\0' && p < acl + size)
771 wp = utf8_decode(tar, acl, p - acl);
772 err = __archive_entry_acl_parse_w(entry, wp,
773 ARCHIVE_ENTRY_ACL_TYPE_ACCESS);
778 * Interpret 'K' long linkname header.
781 header_longlink(struct archive_read *a, struct tar *tar,
782 struct archive_entry *entry, const void *h)
786 err = read_body_to_string(a, tar, &(tar->longlink), h);
787 if (err != ARCHIVE_OK)
789 err = tar_read_header(a, tar, entry);
790 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN))
792 /* Set symlink if symlink already set, else hardlink. */
793 archive_entry_copy_link(entry, tar->longlink.s);
798 * Interpret 'L' long filename header.
801 header_longname(struct archive_read *a, struct tar *tar,
802 struct archive_entry *entry, const void *h)
806 err = read_body_to_string(a, tar, &(tar->longname), h);
807 if (err != ARCHIVE_OK)
809 /* Read and parse "real" header, then override name. */
810 err = tar_read_header(a, tar, entry);
811 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN))
813 archive_entry_copy_pathname(entry, tar->longname.s);
819 * Interpret 'V' GNU tar volume header.
822 header_volume(struct archive_read *a, struct tar *tar,
823 struct archive_entry *entry, const void *h)
827 /* Just skip this and read the next header. */
828 return (tar_read_header(a, tar, entry));
832 * Read body of an archive entry into an archive_string object.
835 read_body_to_string(struct archive_read *a, struct tar *tar,
836 struct archive_string *as, const void *h)
838 off_t size, padded_size;
839 const struct archive_entry_header_ustar *header;
842 (void)tar; /* UNUSED */
843 header = (const struct archive_entry_header_ustar *)h;
844 size = tar_atol(header->size, sizeof(header->size));
845 if ((size > 1048576) || (size < 0)) {
846 archive_set_error(&a->archive, EINVAL,
847 "Special header too large");
848 return (ARCHIVE_FATAL);
851 /* Fail if we can't make our buffer big enough. */
852 if (archive_string_ensure(as, size+1) == NULL) {
853 archive_set_error(&a->archive, ENOMEM,
855 return (ARCHIVE_FATAL);
858 /* Read the body into the string. */
859 padded_size = (size + 511) & ~ 511;
860 src = __archive_read_ahead(a, padded_size, NULL);
862 return (ARCHIVE_FATAL);
863 memcpy(as->s, src, size);
864 __archive_read_consume(a, padded_size);
870 * Parse out common header elements.
872 * This would be the same as header_old_tar, except that the
873 * filename is handled slightly differently for old and POSIX
874 * entries (POSIX entries support a 'prefix'). This factoring
875 * allows header_old_tar and header_ustar
876 * to handle filenames differently, while still putting most of the
877 * common parsing into one place.
880 header_common(struct archive_read *a, struct tar *tar,
881 struct archive_entry *entry, const void *h)
883 const struct archive_entry_header_ustar *header;
886 (void)a; /* UNUSED */
888 header = (const struct archive_entry_header_ustar *)h;
889 if (header->linkname[0])
890 archive_strncpy(&(tar->entry_linkpath), header->linkname,
891 sizeof(header->linkname));
893 archive_string_empty(&(tar->entry_linkpath));
895 /* Parse out the numeric fields (all are octal) */
896 archive_entry_set_mode(entry, tar_atol(header->mode, sizeof(header->mode)));
897 archive_entry_set_uid(entry, tar_atol(header->uid, sizeof(header->uid)));
898 archive_entry_set_gid(entry, tar_atol(header->gid, sizeof(header->gid)));
899 tar->entry_bytes_remaining = tar_atol(header->size, sizeof(header->size));
900 tar->realsize = tar->entry_bytes_remaining;
901 archive_entry_set_size(entry, tar->entry_bytes_remaining);
902 archive_entry_set_mtime(entry, tar_atol(header->mtime, sizeof(header->mtime)), 0);
904 /* Handle the tar type flag appropriately. */
905 tartype = header->typeflag[0];
908 case '1': /* Hard link */
909 archive_entry_copy_hardlink(entry, tar->entry_linkpath.s);
911 * The following may seem odd, but: Technically, tar
912 * does not store the file type for a "hard link"
913 * entry, only the fact that it is a hard link. So, I
914 * leave the type zero normally. But, pax interchange
915 * format allows hard links to have data, which
916 * implies that the underlying entry is a regular
919 if (archive_entry_size(entry) > 0)
920 archive_entry_set_filetype(entry, AE_IFREG);
923 * A tricky point: Traditionally, tar readers have
924 * ignored the size field when reading hardlink
925 * entries, and some writers put non-zero sizes even
926 * though the body is empty. POSIX blessed this
927 * convention in the 1988 standard, but broke with
928 * this tradition in 2001 by permitting hardlink
929 * entries to store valid bodies in pax interchange
930 * format, but not in ustar format. Since there is no
931 * hard and fast way to distinguish pax interchange
932 * from earlier archives (the 'x' and 'g' entries are
933 * optional, after all), we need a heuristic.
935 if (archive_entry_size(entry) == 0) {
936 /* If the size is already zero, we're done. */
937 } else if (a->archive.archive_format
938 == ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE) {
939 /* Definitely pax extended; must obey hardlink size. */
940 } else if (a->archive.archive_format == ARCHIVE_FORMAT_TAR
941 || a->archive.archive_format == ARCHIVE_FORMAT_TAR_GNUTAR)
943 /* Old-style or GNU tar: we must ignore the size. */
944 archive_entry_set_size(entry, 0);
945 tar->entry_bytes_remaining = 0;
946 } else if (archive_read_format_tar_bid(a) > 50) {
948 * We don't know if it's pax: If the bid
949 * function sees a valid ustar header
950 * immediately following, then let's ignore
953 archive_entry_set_size(entry, 0);
954 tar->entry_bytes_remaining = 0;
957 * TODO: There are still two cases I'd like to handle:
958 * = a ustar non-pax archive with a hardlink entry at
959 * end-of-archive. (Look for block of nulls following?)
960 * = a pax archive that has not seen any pax headers
961 * and has an entry which is a hardlink entry storing
962 * a body containing an uncompressed tar archive.
963 * The first is worth addressing; I don't see any reliable
964 * way to deal with the second possibility.
967 case '2': /* Symlink */
968 archive_entry_set_filetype(entry, AE_IFLNK);
969 archive_entry_set_size(entry, 0);
970 tar->entry_bytes_remaining = 0;
971 archive_entry_copy_symlink(entry, tar->entry_linkpath.s);
973 case '3': /* Character device */
974 archive_entry_set_filetype(entry, AE_IFCHR);
975 archive_entry_set_size(entry, 0);
976 tar->entry_bytes_remaining = 0;
978 case '4': /* Block device */
979 archive_entry_set_filetype(entry, AE_IFBLK);
980 archive_entry_set_size(entry, 0);
981 tar->entry_bytes_remaining = 0;
984 archive_entry_set_filetype(entry, AE_IFDIR);
985 archive_entry_set_size(entry, 0);
986 tar->entry_bytes_remaining = 0;
988 case '6': /* FIFO device */
989 archive_entry_set_filetype(entry, AE_IFIFO);
990 archive_entry_set_size(entry, 0);
991 tar->entry_bytes_remaining = 0;
993 case 'D': /* GNU incremental directory type */
995 * No special handling is actually required here.
996 * It might be nice someday to preprocess the file list and
997 * provide it to the client, though.
999 archive_entry_set_filetype(entry, AE_IFDIR);
1001 case 'M': /* GNU "Multi-volume" (remainder of file from last archive)*/
1003 * As far as I can tell, this is just like a regular file
1004 * entry, except that the contents should be _appended_ to
1005 * the indicated file at the indicated offset. This may
1006 * require some API work to fully support.
1009 case 'N': /* Old GNU "long filename" entry. */
1010 /* The body of this entry is a script for renaming
1011 * previously-extracted entries. Ugh. It will never
1012 * be supported by libarchive. */
1013 archive_entry_set_filetype(entry, AE_IFREG);
1015 case 'S': /* GNU sparse files */
1017 * Sparse files are really just regular files with
1018 * sparse information in the extended area.
1021 default: /* Regular file and non-standard types */
1023 * Per POSIX: non-recognized types should always be
1024 * treated as regular files.
1026 archive_entry_set_filetype(entry, AE_IFREG);
1033 * Parse out header elements for "old-style" tar archives.
1036 header_old_tar(struct archive_read *a, struct tar *tar,
1037 struct archive_entry *entry, const void *h)
1039 const struct archive_entry_header_ustar *header;
1041 /* Copy filename over (to ensure null termination). */
1042 header = (const struct archive_entry_header_ustar *)h;
1043 archive_strncpy(&(tar->entry_pathname), header->name, sizeof(header->name));
1044 archive_entry_copy_pathname(entry, tar->entry_pathname.s);
1046 /* Grab rest of common fields */
1047 header_common(a, tar, entry, h);
1049 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining);
1054 * Parse a file header for a pax extended archive entry.
1057 header_pax_global(struct archive_read *a, struct tar *tar,
1058 struct archive_entry *entry, const void *h)
1062 err = read_body_to_string(a, tar, &(tar->pax_global), h);
1063 if (err != ARCHIVE_OK)
1065 err = tar_read_header(a, tar, entry);
1070 header_pax_extensions(struct archive_read *a, struct tar *tar,
1071 struct archive_entry *entry, const void *h)
1075 err = read_body_to_string(a, tar, &(tar->pax_header), h);
1076 if (err != ARCHIVE_OK)
1079 /* Parse the next header. */
1080 err = tar_read_header(a, tar, entry);
1081 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN))
1085 * TODO: Parse global/default options into 'entry' struct here
1086 * before handling file-specific options.
1088 * This design (parse standard header, then overwrite with pax
1089 * extended attribute data) usually works well, but isn't ideal;
1090 * it would be better to parse the pax extended attributes first
1091 * and then skip any fields in the standard header that were
1092 * defined in the pax header.
1094 err2 = pax_header(a, tar, entry, tar->pax_header.s);
1095 err = err_combine(err, err2);
1096 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining);
1102 * Parse a file header for a Posix "ustar" archive entry. This also
1103 * handles "pax" or "extended ustar" entries.
1106 header_ustar(struct archive_read *a, struct tar *tar,
1107 struct archive_entry *entry, const void *h)
1109 const struct archive_entry_header_ustar *header;
1110 struct archive_string *as;
1112 header = (const struct archive_entry_header_ustar *)h;
1114 /* Copy name into an internal buffer to ensure null-termination. */
1115 as = &(tar->entry_pathname);
1116 if (header->prefix[0]) {
1117 archive_strncpy(as, header->prefix, sizeof(header->prefix));
1118 if (as->s[archive_strlen(as) - 1] != '/')
1119 archive_strappend_char(as, '/');
1120 archive_strncat(as, header->name, sizeof(header->name));
1122 archive_strncpy(as, header->name, sizeof(header->name));
1124 archive_entry_copy_pathname(entry, as->s);
1126 /* Handle rest of common fields. */
1127 header_common(a, tar, entry, h);
1129 /* Handle POSIX ustar fields. */
1130 archive_strncpy(&(tar->entry_uname), header->uname,
1131 sizeof(header->uname));
1132 archive_entry_copy_uname(entry, tar->entry_uname.s);
1134 archive_strncpy(&(tar->entry_gname), header->gname,
1135 sizeof(header->gname));
1136 archive_entry_copy_gname(entry, tar->entry_gname.s);
1138 /* Parse out device numbers only for char and block specials. */
1139 if (header->typeflag[0] == '3' || header->typeflag[0] == '4') {
1140 archive_entry_set_rdevmajor(entry,
1141 tar_atol(header->rdevmajor, sizeof(header->rdevmajor)));
1142 archive_entry_set_rdevminor(entry,
1143 tar_atol(header->rdevminor, sizeof(header->rdevminor)));
1146 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining);
1153 * Parse the pax extended attributes record.
1155 * Returns non-zero if there's an error in the data.
1158 pax_header(struct archive_read *a, struct tar *tar,
1159 struct archive_entry *entry, char *attr)
1161 size_t attr_length, l, line_length;
1166 attr_length = strlen(attr);
1167 tar->pax_hdrcharset_binary = 0;
1168 archive_string_empty(&(tar->entry_gname));
1169 archive_string_empty(&(tar->entry_linkpath));
1170 archive_string_empty(&(tar->entry_pathname));
1171 archive_string_empty(&(tar->entry_pathname_override));
1172 archive_string_empty(&(tar->entry_uname));
1174 while (attr_length > 0) {
1175 /* Parse decimal length field at start of line. */
1178 line = p = attr; /* Record start of line. */
1185 if (*p < '0' || *p > '9') {
1186 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1187 "Ignoring malformed pax extended attributes");
1188 return (ARCHIVE_WARN);
1191 line_length += *p - '0';
1192 if (line_length > 999999) {
1193 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1194 "Rejecting pax extended attribute > 1MB");
1195 return (ARCHIVE_WARN);
1202 * Parsed length must be no bigger than available data,
1203 * at least 1, and the last character of the line must
1206 if (line_length > attr_length
1208 || attr[line_length - 1] != '\n')
1210 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1211 "Ignoring malformed pax extended attribute");
1212 return (ARCHIVE_WARN);
1215 /* Null-terminate the line. */
1216 attr[line_length - 1] = '\0';
1218 /* Find end of key and null terminate it. */
1222 while (*p && *p != '=')
1225 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1226 "Invalid pax extended attributes");
1227 return (ARCHIVE_WARN);
1231 /* Identify null-terminated 'value' portion. */
1234 /* Identify this attribute and set it in the entry. */
1235 err2 = pax_attribute(tar, entry, key, value);
1236 err = err_combine(err, err2);
1238 /* Skip to next line */
1239 attr += line_length;
1240 attr_length -= line_length;
1242 if (archive_strlen(&(tar->entry_gname)) > 0) {
1243 value = tar->entry_gname.s;
1244 if (tar->pax_hdrcharset_binary)
1245 archive_entry_copy_gname(entry, value);
1247 if (!archive_entry_update_gname_utf8(entry, value)) {
1249 archive_set_error(&a->archive,
1250 ARCHIVE_ERRNO_FILE_FORMAT,
1251 "Gname in pax header can't "
1252 "be converted to current locale.");
1256 if (archive_strlen(&(tar->entry_linkpath)) > 0) {
1257 value = tar->entry_linkpath.s;
1258 if (tar->pax_hdrcharset_binary)
1259 archive_entry_copy_link(entry, value);
1261 if (!archive_entry_update_link_utf8(entry, value)) {
1263 archive_set_error(&a->archive,
1264 ARCHIVE_ERRNO_FILE_FORMAT,
1265 "Linkname in pax header can't "
1266 "be converted to current locale.");
1271 * Some extensions (such as the GNU sparse file extensions)
1272 * deliberately store a synthetic name under the regular 'path'
1273 * attribute and the real file name under a different attribute.
1274 * Since we're supposed to not care about the order, we
1275 * have no choice but to store all of the various filenames
1276 * we find and figure it all out afterwards. This is the
1277 * figuring out part.
1280 if (archive_strlen(&(tar->entry_pathname_override)) > 0)
1281 value = tar->entry_pathname_override.s;
1282 else if (archive_strlen(&(tar->entry_pathname)) > 0)
1283 value = tar->entry_pathname.s;
1284 if (value != NULL) {
1285 if (tar->pax_hdrcharset_binary)
1286 archive_entry_copy_pathname(entry, value);
1288 if (!archive_entry_update_pathname_utf8(entry, value)) {
1290 archive_set_error(&a->archive,
1291 ARCHIVE_ERRNO_FILE_FORMAT,
1292 "Pathname in pax header can't be "
1293 "converted to current locale.");
1297 if (archive_strlen(&(tar->entry_uname)) > 0) {
1298 value = tar->entry_uname.s;
1299 if (tar->pax_hdrcharset_binary)
1300 archive_entry_copy_uname(entry, value);
1302 if (!archive_entry_update_uname_utf8(entry, value)) {
1304 archive_set_error(&a->archive,
1305 ARCHIVE_ERRNO_FILE_FORMAT,
1306 "Uname in pax header can't "
1307 "be converted to current locale.");
1315 pax_attribute_xattr(struct archive_entry *entry,
1316 char *name, char *value)
1319 void *value_decoded;
1322 if (strlen(name) < 18 || (strncmp(name, "LIBARCHIVE.xattr.", 17)) != 0)
1327 /* URL-decode name */
1328 name_decoded = url_decode(name);
1329 if (name_decoded == NULL)
1332 /* Base-64 decode value */
1333 value_decoded = base64_decode(value, strlen(value), &value_len);
1334 if (value_decoded == NULL) {
1339 archive_entry_xattr_add_entry(entry, name_decoded,
1340 value_decoded, value_len);
1343 free(value_decoded);
1348 * Parse a single key=value attribute. key/value pointers are
1349 * assumed to point into reasonably long-lived storage.
1351 * Note that POSIX reserves all-lowercase keywords. Vendor-specific
1352 * extensions should always have keywords of the form "VENDOR.attribute"
1353 * In particular, it's quite feasible to support many different
1354 * vendor extensions here. I'm using "LIBARCHIVE" for extensions
1355 * unique to this library.
1357 * Investigate other vendor-specific extensions and see if
1358 * any of them look useful.
1361 pax_attribute(struct tar *tar, struct archive_entry *entry,
1362 char *key, char *value)
1370 /* GNU "0.0" sparse pax format. */
1371 if (strcmp(key, "GNU.sparse.numblocks") == 0) {
1372 tar->sparse_offset = -1;
1373 tar->sparse_numbytes = -1;
1374 tar->sparse_gnu_major = 0;
1375 tar->sparse_gnu_minor = 0;
1377 if (strcmp(key, "GNU.sparse.offset") == 0) {
1378 tar->sparse_offset = tar_atol10(value, strlen(value));
1379 if (tar->sparse_numbytes != -1) {
1380 gnu_add_sparse_entry(tar,
1381 tar->sparse_offset, tar->sparse_numbytes);
1382 tar->sparse_offset = -1;
1383 tar->sparse_numbytes = -1;
1386 if (strcmp(key, "GNU.sparse.numbytes") == 0) {
1387 tar->sparse_numbytes = tar_atol10(value, strlen(value));
1388 if (tar->sparse_numbytes != -1) {
1389 gnu_add_sparse_entry(tar,
1390 tar->sparse_offset, tar->sparse_numbytes);
1391 tar->sparse_offset = -1;
1392 tar->sparse_numbytes = -1;
1395 if (strcmp(key, "GNU.sparse.size") == 0) {
1396 tar->realsize = tar_atol10(value, strlen(value));
1397 archive_entry_set_size(entry, tar->realsize);
1400 /* GNU "0.1" sparse pax format. */
1401 if (strcmp(key, "GNU.sparse.map") == 0) {
1402 tar->sparse_gnu_major = 0;
1403 tar->sparse_gnu_minor = 1;
1404 if (gnu_sparse_01_parse(tar, value) != ARCHIVE_OK)
1405 return (ARCHIVE_WARN);
1408 /* GNU "1.0" sparse pax format */
1409 if (strcmp(key, "GNU.sparse.major") == 0) {
1410 tar->sparse_gnu_major = tar_atol10(value, strlen(value));
1411 tar->sparse_gnu_pending = 1;
1413 if (strcmp(key, "GNU.sparse.minor") == 0) {
1414 tar->sparse_gnu_minor = tar_atol10(value, strlen(value));
1415 tar->sparse_gnu_pending = 1;
1417 if (strcmp(key, "GNU.sparse.name") == 0) {
1419 * The real filename; when storing sparse
1420 * files, GNU tar puts a synthesized name into
1421 * the regular 'path' attribute in an attempt
1422 * to limit confusion. ;-)
1424 archive_strcpy(&(tar->entry_pathname_override), value);
1426 if (strcmp(key, "GNU.sparse.realsize") == 0) {
1427 tar->realsize = tar_atol10(value, strlen(value));
1428 archive_entry_set_size(entry, tar->realsize);
1432 /* Our extensions */
1433 /* TODO: Handle arbitrary extended attributes... */
1435 if (strcmp(key, "LIBARCHIVE.xxxxxxx")==0)
1436 archive_entry_set_xxxxxx(entry, value);
1438 if (strcmp(key, "LIBARCHIVE.creationtime")==0) {
1439 pax_time(value, &s, &n);
1440 archive_entry_set_birthtime(entry, s, n);
1442 if (strncmp(key, "LIBARCHIVE.xattr.", 17)==0)
1443 pax_attribute_xattr(entry, key, value);
1446 /* We support some keys used by the "star" archiver */
1447 if (strcmp(key, "SCHILY.acl.access")==0) {
1448 wp = utf8_decode(tar, value, strlen(value));
1449 /* TODO: if (wp == NULL) */
1450 __archive_entry_acl_parse_w(entry, wp,
1451 ARCHIVE_ENTRY_ACL_TYPE_ACCESS);
1452 } else if (strcmp(key, "SCHILY.acl.default")==0) {
1453 wp = utf8_decode(tar, value, strlen(value));
1454 /* TODO: if (wp == NULL) */
1455 __archive_entry_acl_parse_w(entry, wp,
1456 ARCHIVE_ENTRY_ACL_TYPE_DEFAULT);
1457 } else if (strcmp(key, "SCHILY.devmajor")==0) {
1458 archive_entry_set_rdevmajor(entry,
1459 tar_atol10(value, strlen(value)));
1460 } else if (strcmp(key, "SCHILY.devminor")==0) {
1461 archive_entry_set_rdevminor(entry,
1462 tar_atol10(value, strlen(value)));
1463 } else if (strcmp(key, "SCHILY.fflags")==0) {
1464 archive_entry_copy_fflags_text(entry, value);
1465 } else if (strcmp(key, "SCHILY.dev")==0) {
1466 archive_entry_set_dev(entry,
1467 tar_atol10(value, strlen(value)));
1468 } else if (strcmp(key, "SCHILY.ino")==0) {
1469 archive_entry_set_ino(entry,
1470 tar_atol10(value, strlen(value)));
1471 } else if (strcmp(key, "SCHILY.nlink")==0) {
1472 archive_entry_set_nlink(entry,
1473 tar_atol10(value, strlen(value)));
1474 } else if (strcmp(key, "SCHILY.realsize")==0) {
1475 tar->realsize = tar_atol10(value, strlen(value));
1476 archive_entry_set_size(entry, tar->realsize);
1480 if (strcmp(key, "atime")==0) {
1481 pax_time(value, &s, &n);
1482 archive_entry_set_atime(entry, s, n);
1486 if (strcmp(key, "ctime")==0) {
1487 pax_time(value, &s, &n);
1488 archive_entry_set_ctime(entry, s, n);
1489 } else if (strcmp(key, "charset")==0) {
1490 /* TODO: Publish charset information in entry. */
1491 } else if (strcmp(key, "comment")==0) {
1492 /* TODO: Publish comment in entry. */
1496 if (strcmp(key, "gid")==0) {
1497 archive_entry_set_gid(entry,
1498 tar_atol10(value, strlen(value)));
1499 } else if (strcmp(key, "gname")==0) {
1500 archive_strcpy(&(tar->entry_gname), value);
1504 if (strcmp(key, "hdrcharset") == 0) {
1505 if (strcmp(value, "BINARY") == 0)
1506 tar->pax_hdrcharset_binary = 1;
1507 else if (strcmp(value, "ISO-IR 10646 2000 UTF-8") == 0)
1508 tar->pax_hdrcharset_binary = 0;
1510 /* TODO: Warn about unsupported hdrcharset */
1515 /* pax interchange doesn't distinguish hardlink vs. symlink. */
1516 if (strcmp(key, "linkpath")==0) {
1517 archive_strcpy(&(tar->entry_linkpath), value);
1521 if (strcmp(key, "mtime")==0) {
1522 pax_time(value, &s, &n);
1523 archive_entry_set_mtime(entry, s, n);
1527 if (strcmp(key, "path")==0) {
1528 archive_strcpy(&(tar->entry_pathname), value);
1532 /* POSIX has reserved 'realtime.*' */
1535 /* POSIX has reserved 'security.*' */
1536 /* Someday: if (strcmp(key, "security.acl")==0) { ... } */
1537 if (strcmp(key, "size")==0) {
1538 /* "size" is the size of the data in the entry. */
1539 tar->entry_bytes_remaining
1540 = tar_atol10(value, strlen(value));
1542 * But, "size" is not necessarily the size of
1543 * the file on disk; if this is a sparse file,
1544 * the disk size may have already been set from
1545 * GNU.sparse.realsize or GNU.sparse.size or
1546 * an old GNU header field or SCHILY.realsize
1549 if (tar->realsize < 0) {
1550 archive_entry_set_size(entry,
1551 tar->entry_bytes_remaining);
1553 = tar->entry_bytes_remaining;
1558 if (strcmp(key, "uid")==0) {
1559 archive_entry_set_uid(entry,
1560 tar_atol10(value, strlen(value)));
1561 } else if (strcmp(key, "uname")==0) {
1562 archive_strcpy(&(tar->entry_uname), value);
1572 * parse a decimal time value, which may include a fractional portion
1575 pax_time(const char *p, int64_t *ps, long *pn)
1581 int64_t limit, last_digit_limit;
1583 limit = INT64_MAX / 10;
1584 last_digit_limit = INT64_MAX % 10;
1592 while (*p >= '0' && *p <= '9') {
1595 (s == limit && digit > last_digit_limit)) {
1599 s = (s * 10) + digit;
1605 /* Calculate nanoseconds. */
1614 if (*p >= '0' && *p <= '9')
1615 *pn += (*p - '0') * l;
1622 * Parse GNU tar header
1625 header_gnutar(struct archive_read *a, struct tar *tar,
1626 struct archive_entry *entry, const void *h)
1628 const struct archive_entry_header_gnutar *header;
1633 * GNU header is like POSIX ustar, except 'prefix' is
1634 * replaced with some other fields. This also means the
1635 * filename is stored as in old-style archives.
1638 /* Grab fields common to all tar variants. */
1639 header_common(a, tar, entry, h);
1641 /* Copy filename over (to ensure null termination). */
1642 header = (const struct archive_entry_header_gnutar *)h;
1643 archive_strncpy(&(tar->entry_pathname), header->name,
1644 sizeof(header->name));
1645 archive_entry_copy_pathname(entry, tar->entry_pathname.s);
1647 /* Fields common to ustar and GNU */
1648 /* XXX Can the following be factored out since it's common
1649 * to ustar and gnu tar? Is it okay to move it down into
1650 * header_common, perhaps? */
1651 archive_strncpy(&(tar->entry_uname),
1652 header->uname, sizeof(header->uname));
1653 archive_entry_copy_uname(entry, tar->entry_uname.s);
1655 archive_strncpy(&(tar->entry_gname),
1656 header->gname, sizeof(header->gname));
1657 archive_entry_copy_gname(entry, tar->entry_gname.s);
1659 /* Parse out device numbers only for char and block specials */
1660 if (header->typeflag[0] == '3' || header->typeflag[0] == '4') {
1661 archive_entry_set_rdevmajor(entry,
1662 tar_atol(header->rdevmajor, sizeof(header->rdevmajor)));
1663 archive_entry_set_rdevminor(entry,
1664 tar_atol(header->rdevminor, sizeof(header->rdevminor)));
1666 archive_entry_set_rdev(entry, 0);
1668 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining);
1670 /* Grab GNU-specific fields. */
1671 archive_entry_set_atime(entry,
1672 tar_atol(header->atime, sizeof(header->atime)), 0);
1673 archive_entry_set_ctime(entry,
1674 tar_atol(header->ctime, sizeof(header->ctime)), 0);
1675 if (header->realsize[0] != 0) {
1677 = tar_atol(header->realsize, sizeof(header->realsize));
1678 archive_entry_set_size(entry, tar->realsize);
1681 if (header->sparse[0].offset[0] != 0) {
1682 gnu_sparse_old_read(a, tar, header);
1684 if (header->isextended[0] != 0) {
1693 gnu_add_sparse_entry(struct tar *tar, off_t offset, off_t remaining)
1695 struct sparse_block *p;
1697 p = (struct sparse_block *)malloc(sizeof(*p));
1699 __archive_errx(1, "Out of memory");
1700 memset(p, 0, sizeof(*p));
1701 if (tar->sparse_last != NULL)
1702 tar->sparse_last->next = p;
1704 tar->sparse_list = p;
1705 tar->sparse_last = p;
1707 p->remaining = remaining;
1711 gnu_clear_sparse_list(struct tar *tar)
1713 struct sparse_block *p;
1715 while (tar->sparse_list != NULL) {
1716 p = tar->sparse_list;
1717 tar->sparse_list = p->next;
1720 tar->sparse_last = NULL;
1724 * GNU tar old-format sparse data.
1726 * GNU old-format sparse data is stored in a fixed-field
1727 * format. Offset/size values are 11-byte octal fields (same
1728 * format as 'size' field in ustart header). These are
1729 * stored in the header, allocating subsequent header blocks
1730 * as needed. Extending the header in this way is a pretty
1731 * severe POSIX violation; this design has earned GNU tar a
1736 gnu_sparse_old_read(struct archive_read *a, struct tar *tar,
1737 const struct archive_entry_header_gnutar *header)
1742 struct gnu_sparse sparse[21];
1746 const struct extended *ext;
1748 gnu_sparse_old_parse(tar, header->sparse, 4);
1749 if (header->isextended[0] == 0)
1750 return (ARCHIVE_OK);
1753 data = __archive_read_ahead(a, 512, &bytes_read);
1755 return (ARCHIVE_FATAL);
1756 if (bytes_read < 512) {
1757 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1758 "Truncated tar archive "
1759 "detected while reading sparse file data");
1760 return (ARCHIVE_FATAL);
1762 __archive_read_consume(a, 512);
1763 ext = (const struct extended *)data;
1764 gnu_sparse_old_parse(tar, ext->sparse, 21);
1765 } while (ext->isextended[0] != 0);
1766 if (tar->sparse_list != NULL)
1767 tar->entry_offset = tar->sparse_list->offset;
1768 return (ARCHIVE_OK);
1772 gnu_sparse_old_parse(struct tar *tar,
1773 const struct gnu_sparse *sparse, int length)
1775 while (length > 0 && sparse->offset[0] != 0) {
1776 gnu_add_sparse_entry(tar,
1777 tar_atol(sparse->offset, sizeof(sparse->offset)),
1778 tar_atol(sparse->numbytes, sizeof(sparse->numbytes)));
1785 * GNU tar sparse format 0.0
1787 * Beginning with GNU tar 1.15, sparse files are stored using
1788 * information in the pax extended header. The GNU tar maintainers
1789 * have gone through a number of variations in the process of working
1790 * out this scheme; furtunately, they're all numbered.
1792 * Sparse format 0.0 uses attribute GNU.sparse.numblocks to store the
1793 * number of blocks, and GNU.sparse.offset/GNU.sparse.numbytes to
1794 * store offset/size for each block. The repeated instances of these
1795 * latter fields violate the pax specification (which frowns on
1796 * duplicate keys), so this format was quickly replaced.
1800 * GNU tar sparse format 0.1
1802 * This version replaced the offset/numbytes attributes with
1803 * a single "map" attribute that stored a list of integers. This
1804 * format had two problems: First, the "map" attribute could be very
1805 * long, which caused problems for some implementations. More
1806 * importantly, the sparse data was lost when extracted by archivers
1807 * that didn't recognize this extension.
1811 gnu_sparse_01_parse(struct tar *tar, const char *p)
1814 off_t offset = -1, size = -1;
1818 while (*e != '\0' && *e != ',') {
1819 if (*e < '0' || *e > '9')
1820 return (ARCHIVE_WARN);
1824 offset = tar_atol10(p, e - p);
1826 return (ARCHIVE_WARN);
1828 size = tar_atol10(p, e - p);
1830 return (ARCHIVE_WARN);
1831 gnu_add_sparse_entry(tar, offset, size);
1835 return (ARCHIVE_OK);
1841 * GNU tar sparse format 1.0
1843 * The idea: The offset/size data is stored as a series of base-10
1844 * ASCII numbers prepended to the file data, so that dearchivers that
1845 * don't support this format will extract the block map along with the
1846 * data and a separate post-process can restore the sparseness.
1848 * Unfortunately, GNU tar 1.16 had a bug that added unnecessary
1849 * padding to the body of the file when using this format. GNU tar
1850 * 1.17 corrected this bug without bumping the version number, so
1851 * it's not possible to support both variants. This code supports
1852 * the later variant at the expense of not supporting the former.
1854 * This variant also replaced GNU.sparse.size with GNU.sparse.realsize
1855 * and introduced the GNU.sparse.major/GNU.sparse.minor attributes.
1859 * Read the next line from the input, and parse it as a decimal
1860 * integer followed by '\n'. Returns positive integer value or
1861 * negative on error.
1864 gnu_sparse_10_atol(struct archive_read *a, struct tar *tar,
1867 int64_t l, limit, last_digit_limit;
1873 limit = INT64_MAX / base;
1874 last_digit_limit = INT64_MAX % base;
1877 * Skip any lines starting with '#'; GNU tar specs
1878 * don't require this, but they should.
1881 bytes_read = readline(a, tar, &p, tar_min(*remaining, 100));
1882 if (bytes_read <= 0)
1883 return (ARCHIVE_FATAL);
1884 *remaining -= bytes_read;
1885 } while (p[0] == '#');
1888 while (bytes_read > 0) {
1891 if (*p < '0' || *p >= '0' + base)
1892 return (ARCHIVE_WARN);
1894 if (l > limit || (l == limit && digit > last_digit_limit))
1895 l = UINT64_MAX; /* Truncate on overflow. */
1897 l = (l * base) + digit;
1901 /* TODO: Error message. */
1902 return (ARCHIVE_WARN);
1906 * Returns length (in bytes) of the sparse data description
1910 gnu_sparse_10_read(struct archive_read *a, struct tar *tar)
1912 ssize_t remaining, bytes_read;
1914 off_t offset, size, to_skip;
1916 /* Clear out the existing sparse list. */
1917 gnu_clear_sparse_list(tar);
1919 remaining = tar->entry_bytes_remaining;
1921 /* Parse entries. */
1922 entries = gnu_sparse_10_atol(a, tar, &remaining);
1924 return (ARCHIVE_FATAL);
1925 /* Parse the individual entries. */
1926 while (entries-- > 0) {
1927 /* Parse offset/size */
1928 offset = gnu_sparse_10_atol(a, tar, &remaining);
1930 return (ARCHIVE_FATAL);
1931 size = gnu_sparse_10_atol(a, tar, &remaining);
1933 return (ARCHIVE_FATAL);
1934 /* Add a new sparse entry. */
1935 gnu_add_sparse_entry(tar, offset, size);
1937 /* Skip rest of block... */
1938 bytes_read = tar->entry_bytes_remaining - remaining;
1939 to_skip = 0x1ff & -bytes_read;
1940 if (to_skip != __archive_read_skip(a, to_skip))
1941 return (ARCHIVE_FATAL);
1942 return (bytes_read + to_skip);
1946 * Convert text->integer.
1948 * Traditional tar formats (including POSIX) specify base-8 for
1949 * all of the standard numeric fields. This is a significant limitation
1951 * = file size is limited to 8GB
1952 * = rdevmajor and rdevminor are limited to 21 bits
1953 * = uid/gid are limited to 21 bits
1955 * There are two workarounds for this:
1956 * = pax extended headers, which use variable-length string fields
1957 * = GNU tar and STAR both allow either base-8 or base-256 in
1958 * most fields. The high bit is set to indicate base-256.
1960 * On read, this implementation supports both extensions.
1963 tar_atol(const char *p, unsigned char_cnt)
1966 * Technically, GNU tar considers a field to be in base-256
1967 * only if the first byte is 0xff or 0x80.
1970 return (tar_atol256(p, char_cnt));
1971 return (tar_atol8(p, char_cnt));
1975 * Note that this implementation does not (and should not!) obey
1976 * locale settings; you cannot simply substitute strtol here, since
1977 * it does obey locale.
1980 tar_atol8(const char *p, unsigned char_cnt)
1982 int64_t l, limit, last_digit_limit;
1983 int digit, sign, base;
1986 limit = INT64_MAX / base;
1987 last_digit_limit = INT64_MAX % base;
1989 while (*p == ' ' || *p == '\t')
1999 while (digit >= 0 && digit < base && char_cnt-- > 0) {
2000 if (l>limit || (l == limit && digit > last_digit_limit)) {
2001 l = UINT64_MAX; /* Truncate on overflow. */
2004 l = (l * base) + digit;
2007 return (sign < 0) ? -l : l;
2011 * Note that this implementation does not (and should not!) obey
2012 * locale settings; you cannot simply substitute strtol here, since
2013 * it does obey locale.
2016 tar_atol10(const char *p, unsigned char_cnt)
2018 int64_t l, limit, last_digit_limit;
2019 int base, digit, sign;
2022 limit = INT64_MAX / base;
2023 last_digit_limit = INT64_MAX % base;
2025 while (*p == ' ' || *p == '\t')
2035 while (digit >= 0 && digit < base && char_cnt-- > 0) {
2036 if (l > limit || (l == limit && digit > last_digit_limit)) {
2037 l = UINT64_MAX; /* Truncate on overflow. */
2040 l = (l * base) + digit;
2043 return (sign < 0) ? -l : l;
2047 * Parse a base-256 integer. This is just a straight signed binary
2048 * value in big-endian order, except that the high-order bit is
2052 tar_atol256(const char *_p, unsigned char_cnt)
2054 int64_t l, upper_limit, lower_limit;
2055 const unsigned char *p = (const unsigned char *)_p;
2057 upper_limit = INT64_MAX / 256;
2058 lower_limit = INT64_MIN / 256;
2060 /* Pad with 1 or 0 bits, depending on sign. */
2061 if ((0x40 & *p) == 0x40)
2065 l = (l << 6) | (0x3f & *p++);
2066 while (--char_cnt > 0) {
2067 if (l > upper_limit) {
2068 l = INT64_MAX; /* Truncate on overflow */
2070 } else if (l < lower_limit) {
2074 l = (l << 8) | (0xff & (int64_t)*p++);
2080 * Returns length of line (including trailing newline)
2081 * or negative on error. 'start' argument is updated to
2082 * point to first character of line. This avoids copying
2086 readline(struct archive_read *a, struct tar *tar, const char **start,
2090 ssize_t total_size = 0;
2095 t = __archive_read_ahead(a, 1, &bytes_read);
2096 if (bytes_read <= 0)
2097 return (ARCHIVE_FATAL);
2098 s = t; /* Start of line? */
2099 p = memchr(t, '\n', bytes_read);
2100 /* If we found '\n' in the read buffer, return pointer to that. */
2102 bytes_read = 1 + ((const char *)p) - s;
2103 if (bytes_read > limit) {
2104 archive_set_error(&a->archive,
2105 ARCHIVE_ERRNO_FILE_FORMAT,
2107 return (ARCHIVE_FATAL);
2109 __archive_read_consume(a, bytes_read);
2111 return (bytes_read);
2113 /* Otherwise, we need to accumulate in a line buffer. */
2115 if (total_size + bytes_read > limit) {
2116 archive_set_error(&a->archive,
2117 ARCHIVE_ERRNO_FILE_FORMAT,
2119 return (ARCHIVE_FATAL);
2121 if (archive_string_ensure(&tar->line, total_size + bytes_read) == NULL) {
2122 archive_set_error(&a->archive, ENOMEM,
2123 "Can't allocate working buffer");
2124 return (ARCHIVE_FATAL);
2126 memcpy(tar->line.s + total_size, t, bytes_read);
2127 __archive_read_consume(a, bytes_read);
2128 total_size += bytes_read;
2129 /* If we found '\n', clean up and return. */
2131 *start = tar->line.s;
2132 return (total_size);
2134 /* Read some more. */
2135 t = __archive_read_ahead(a, 1, &bytes_read);
2136 if (bytes_read <= 0)
2137 return (ARCHIVE_FATAL);
2138 s = t; /* Start of line? */
2139 p = memchr(t, '\n', bytes_read);
2140 /* If we found '\n', trim the read. */
2142 bytes_read = 1 + ((const char *)p) - s;
2148 utf8_decode(struct tar *tar, const char *src, size_t length)
2153 /* Ensure pax_entry buffer is big enough. */
2154 if (tar->pax_entry_length <= length) {
2155 wchar_t *old_entry = tar->pax_entry;
2157 if (tar->pax_entry_length <= 0)
2158 tar->pax_entry_length = 1024;
2159 while (tar->pax_entry_length <= length + 1)
2160 tar->pax_entry_length *= 2;
2162 old_entry = tar->pax_entry;
2163 tar->pax_entry = (wchar_t *)realloc(tar->pax_entry,
2164 tar->pax_entry_length * sizeof(wchar_t));
2165 if (tar->pax_entry == NULL) {
2167 /* TODO: Handle this error. */
2172 dest = tar->pax_entry;
2173 while (length > 0) {
2174 n = UTF8_mbrtowc(dest, src, length);
2184 return (tar->pax_entry);
2188 * Copied and simplified from FreeBSD libc/locale.
2191 UTF8_mbrtowc(wchar_t *pwc, const char *s, size_t n)
2193 int ch, i, len, mask;
2196 if (s == NULL || n == 0 || pwc == NULL)
2200 * Determine the number of octets that make up this character from
2201 * the first octet, and a mask that extracts the interesting bits of
2204 ch = (unsigned char)*s;
2205 if ((ch & 0x80) == 0) {
2208 } else if ((ch & 0xe0) == 0xc0) {
2211 } else if ((ch & 0xf0) == 0xe0) {
2214 } else if ((ch & 0xf8) == 0xf0) {
2218 /* Invalid first byte. */
2222 if (n < (size_t)len) {
2223 /* Valid first byte but truncated. */
2228 * Decode the octet sequence representing the character in chunks
2229 * of 6 bits, most significant first.
2231 wch = (unsigned char)*s++ & mask;
2234 if ((*s & 0xc0) != 0x80) {
2235 /* Invalid intermediate byte; consume one byte and
2244 /* Assign the value to the output; out-of-range values
2245 * just get truncated. */
2246 *pwc = (wchar_t)wch;
2249 * If platform has WCHAR_MAX, we can do something
2250 * more sensible with out-of-range values.
2252 if (wch >= WCHAR_MAX)
2255 /* Return number of bytes input consumed: 0 for end-of-string. */
2256 return (wch == L'\0' ? 0 : len);
2261 * base64_decode - Base64 decode
2263 * This accepts most variations of base-64 encoding, including:
2264 * * with or without line breaks
2265 * * with or without the final group padded with '=' or '_' characters
2266 * (The most economical Base-64 variant does not pad the last group and
2267 * omits line breaks; RFC1341 used for MIME requires both.)
2270 base64_decode(const char *s, size_t len, size_t *out_len)
2272 static const unsigned char digits[64] = {
2273 'A','B','C','D','E','F','G','H','I','J','K','L','M','N',
2274 'O','P','Q','R','S','T','U','V','W','X','Y','Z','a','b',
2275 'c','d','e','f','g','h','i','j','k','l','m','n','o','p',
2276 'q','r','s','t','u','v','w','x','y','z','0','1','2','3',
2277 '4','5','6','7','8','9','+','/' };
2278 static unsigned char decode_table[128];
2280 const unsigned char *src = (const unsigned char *)s;
2282 /* If the decode table is not yet initialized, prepare it. */
2283 if (decode_table[digits[1]] != 1) {
2285 memset(decode_table, 0xff, sizeof(decode_table));
2286 for (i = 0; i < sizeof(digits); i++)
2287 decode_table[digits[i]] = i;
2290 /* Allocate enough space to hold the entire output. */
2291 /* Note that we may not use all of this... */
2292 out = (char *)malloc(len - len / 4 + 1);
2300 /* Collect the next group of (up to) four characters. */
2303 while (group_size < 4 && len > 0) {
2304 /* '=' or '_' padding indicates final group. */
2305 if (*src == '=' || *src == '_') {
2309 /* Skip illegal characters (including line breaks) */
2310 if (*src > 127 || *src < 32
2311 || decode_table[*src] == 0xff) {
2317 v |= decode_table[*src++];
2321 /* Align a short group properly. */
2322 v <<= 6 * (4 - group_size);
2323 /* Unpack the group we just collected. */
2324 switch (group_size) {
2325 case 4: d[2] = v & 0xff;
2327 case 3: d[1] = (v >> 8) & 0xff;
2329 case 2: d[0] = (v >> 16) & 0xff;
2331 case 1: /* this is invalid! */
2334 d += group_size * 3 / 4;
2342 url_decode(const char *in)
2347 out = (char *)malloc(strlen(in) + 1);
2350 for (s = in, d = out; *s != '\0'; ) {
2351 if (s[0] == '%' && s[1] != '\0' && s[2] != '\0') {
2352 /* Try to convert % escape */
2353 int digit1 = tohex(s[1]);
2354 int digit2 = tohex(s[2]);
2355 if (digit1 >= 0 && digit2 >= 0) {
2356 /* Looks good, consume three chars */
2358 /* Convert output */
2359 *d++ = ((digit1 << 4) | digit2);
2362 /* Else fall through and treat '%' as normal char */
2373 if (c >= '0' && c <= '9')
2375 else if (c >= 'A' && c <= 'F')
2376 return (c - 'A' + 10);
2377 else if (c >= 'a' && c <= 'f')
2378 return (c - 'a' + 10);