2 * Copyright (c) 2003-2007 Tim Kientzle
3 * Copyright (c) 2011-2012 Michihiro NAKAJIMA
4 * Copyright (c) 2016 Martin Matuska
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR
17 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
18 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
19 * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT,
20 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
21 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
22 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
23 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
24 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
25 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
28 #include "archive_platform.h"
29 __FBSDID("$FreeBSD$");
43 #include "archive_acl_private.h" /* For ACL parsing routines. */
44 #include "archive_entry.h"
45 #include "archive_entry_locale.h"
46 #include "archive_private.h"
47 #include "archive_read_private.h"
49 #define tar_min(a,b) ((a) < (b) ? (a) : (b))
52 * Layout of POSIX 'ustar' tar header.
54 struct archive_entry_header_ustar {
63 char linkname[100]; /* "old format" header ends here */
64 char magic[6]; /* For POSIX: "ustar\0" */
65 char version[2]; /* For POSIX: "00" */
74 * Structure of GNU tar header
81 struct archive_entry_header_gnutar {
91 char magic[8]; /* "ustar \0" (note blank/blank/null at end) */
101 struct gnu_sparse sparse[4];
105 * Old GNU format doesn't use POSIX 'prefix' field; they use
106 * the 'L' (longname) entry instead.
111 * Data specific to this format.
113 struct sparse_block {
114 struct sparse_block *next;
121 struct archive_string acl_text;
122 struct archive_string entry_pathname;
123 /* For "GNU.sparse.name" and other similar path extensions. */
124 struct archive_string entry_pathname_override;
125 struct archive_string entry_linkpath;
126 struct archive_string entry_uname;
127 struct archive_string entry_gname;
128 struct archive_string longlink;
129 struct archive_string longname;
130 struct archive_string pax_header;
131 struct archive_string pax_global;
132 struct archive_string line;
133 int pax_hdrcharset_binary;
134 int header_recursion_depth;
135 int64_t entry_bytes_remaining;
136 int64_t entry_offset;
137 int64_t entry_padding;
138 int64_t entry_bytes_unconsumed;
141 struct sparse_block *sparse_list;
142 struct sparse_block *sparse_last;
143 int64_t sparse_offset;
144 int64_t sparse_numbytes;
145 int sparse_gnu_major;
146 int sparse_gnu_minor;
147 char sparse_gnu_pending;
149 struct archive_string localname;
150 struct archive_string_conv *opt_sconv;
151 struct archive_string_conv *sconv;
152 struct archive_string_conv *sconv_acl;
153 struct archive_string_conv *sconv_default;
154 int init_default_conversion;
156 int process_mac_extensions;
157 int read_concatenated_archives;
160 static int archive_block_is_null(const char *p);
161 static char *base64_decode(const char *, size_t, size_t *);
162 static int gnu_add_sparse_entry(struct archive_read *, struct tar *,
163 int64_t offset, int64_t remaining);
165 static void gnu_clear_sparse_list(struct tar *);
166 static int gnu_sparse_old_read(struct archive_read *, struct tar *,
167 const struct archive_entry_header_gnutar *header, size_t *);
168 static int gnu_sparse_old_parse(struct archive_read *, struct tar *,
169 const struct gnu_sparse *sparse, int length);
170 static int gnu_sparse_01_parse(struct archive_read *, struct tar *,
172 static ssize_t gnu_sparse_10_read(struct archive_read *, struct tar *,
174 static int header_Solaris_ACL(struct archive_read *, struct tar *,
175 struct archive_entry *, const void *, size_t *);
176 static int header_common(struct archive_read *, struct tar *,
177 struct archive_entry *, const void *);
178 static int header_old_tar(struct archive_read *, struct tar *,
179 struct archive_entry *, const void *);
180 static int header_pax_extensions(struct archive_read *, struct tar *,
181 struct archive_entry *, const void *, size_t *);
182 static int header_pax_global(struct archive_read *, struct tar *,
183 struct archive_entry *, const void *h, size_t *);
184 static int header_longlink(struct archive_read *, struct tar *,
185 struct archive_entry *, const void *h, size_t *);
186 static int header_longname(struct archive_read *, struct tar *,
187 struct archive_entry *, const void *h, size_t *);
188 static int read_mac_metadata_blob(struct archive_read *, struct tar *,
189 struct archive_entry *, const void *h, size_t *);
190 static int header_volume(struct archive_read *, struct tar *,
191 struct archive_entry *, const void *h, size_t *);
192 static int header_ustar(struct archive_read *, struct tar *,
193 struct archive_entry *, const void *h);
194 static int header_gnutar(struct archive_read *, struct tar *,
195 struct archive_entry *, const void *h, size_t *);
196 static int archive_read_format_tar_bid(struct archive_read *, int);
197 static int archive_read_format_tar_options(struct archive_read *,
198 const char *, const char *);
199 static int archive_read_format_tar_cleanup(struct archive_read *);
200 static int archive_read_format_tar_read_data(struct archive_read *a,
201 const void **buff, size_t *size, int64_t *offset);
202 static int archive_read_format_tar_skip(struct archive_read *a);
203 static int archive_read_format_tar_read_header(struct archive_read *,
204 struct archive_entry *);
205 static int checksum(struct archive_read *, const void *);
206 static int pax_attribute(struct archive_read *, struct tar *,
207 struct archive_entry *, const char *key, const char *value,
208 size_t value_length);
209 static int pax_attribute_acl(struct archive_read *, struct tar *,
210 struct archive_entry *, const char *, int);
211 static int pax_attribute_xattr(struct archive_entry *, const char *,
213 static int pax_header(struct archive_read *, struct tar *,
214 struct archive_entry *, struct archive_string *);
215 static void pax_time(const char *, int64_t *sec, long *nanos);
216 static ssize_t readline(struct archive_read *, struct tar *, const char **,
217 ssize_t limit, size_t *);
218 static int read_body_to_string(struct archive_read *, struct tar *,
219 struct archive_string *, const void *h, size_t *);
220 static int solaris_sparse_parse(struct archive_read *, struct tar *,
221 struct archive_entry *, const char *);
222 static int64_t tar_atol(const char *, size_t);
223 static int64_t tar_atol10(const char *, size_t);
224 static int64_t tar_atol256(const char *, size_t);
225 static int64_t tar_atol8(const char *, size_t);
226 static int tar_read_header(struct archive_read *, struct tar *,
227 struct archive_entry *, size_t *);
228 static int tohex(int c);
229 static char *url_decode(const char *);
230 static void tar_flush_unconsumed(struct archive_read *, size_t *);
234 archive_read_support_format_gnutar(struct archive *a)
236 archive_check_magic(a, ARCHIVE_READ_MAGIC,
237 ARCHIVE_STATE_NEW, "archive_read_support_format_gnutar");
238 return (archive_read_support_format_tar(a));
243 archive_read_support_format_tar(struct archive *_a)
245 struct archive_read *a = (struct archive_read *)_a;
249 archive_check_magic(_a, ARCHIVE_READ_MAGIC,
250 ARCHIVE_STATE_NEW, "archive_read_support_format_tar");
252 tar = (struct tar *)calloc(1, sizeof(*tar));
253 #ifdef HAVE_COPYFILE_H
254 /* Set this by default on Mac OS. */
255 tar->process_mac_extensions = 1;
258 archive_set_error(&a->archive, ENOMEM,
259 "Can't allocate tar data");
260 return (ARCHIVE_FATAL);
263 r = __archive_read_register_format(a, tar, "tar",
264 archive_read_format_tar_bid,
265 archive_read_format_tar_options,
266 archive_read_format_tar_read_header,
267 archive_read_format_tar_read_data,
268 archive_read_format_tar_skip,
270 archive_read_format_tar_cleanup,
280 archive_read_format_tar_cleanup(struct archive_read *a)
284 tar = (struct tar *)(a->format->data);
285 gnu_clear_sparse_list(tar);
286 archive_string_free(&tar->acl_text);
287 archive_string_free(&tar->entry_pathname);
288 archive_string_free(&tar->entry_pathname_override);
289 archive_string_free(&tar->entry_linkpath);
290 archive_string_free(&tar->entry_uname);
291 archive_string_free(&tar->entry_gname);
292 archive_string_free(&tar->line);
293 archive_string_free(&tar->pax_global);
294 archive_string_free(&tar->pax_header);
295 archive_string_free(&tar->longname);
296 archive_string_free(&tar->longlink);
297 archive_string_free(&tar->localname);
299 (a->format->data) = NULL;
304 * Validate number field
306 * This has to be pretty lenient in order to accommodate the enormous
307 * variety of tar writers in the world:
308 * = POSIX (IEEE Std 1003.1-1988) ustar requires octal values with leading
309 * zeros and allows fields to be terminated with space or null characters
310 * = Many writers use different termination (in particular, libarchive
311 * omits terminator bytes to squeeze one or two more digits)
312 * = Many writers pad with space and omit leading zeros
313 * = GNU tar and star write base-256 values if numbers are too
314 * big to be represented in octal
316 * Examples of specific tar headers that we should support:
317 * = Perl Archive::Tar terminates uid, gid, devminor and devmajor with two
318 * null bytes, pads size with spaces and other numeric fields with zeroes
319 * = plexus-archiver prior to 2.6.3 (before switching to commons-compress)
320 * may have uid and gid fields filled with spaces without any octal digits
321 * at all and pads all numeric fields with spaces
323 * This should tolerate all variants in use. It will reject a field
324 * where the writer just left garbage after a trailing NUL.
327 validate_number_field(const char* p_field, size_t i_size)
329 unsigned char marker = (unsigned char)p_field[0];
330 if (marker == 128 || marker == 255 || marker == 0) {
331 /* Base-256 marker, there's nothing we can check. */
336 /* Skip any leading spaces */
337 while (i < i_size && p_field[i] == ' ') {
340 /* Skip octal digits. */
341 while (i < i_size && p_field[i] >= '0' && p_field[i] <= '7') {
344 /* Any remaining characters must be space or NUL padding. */
346 if (p_field[i] != ' ' && p_field[i] != 0) {
356 archive_read_format_tar_bid(struct archive_read *a, int best_bid)
360 const struct archive_entry_header_ustar *header;
362 (void)best_bid; /* UNUSED */
366 /* Now let's look at the actual header and see if it matches. */
367 h = __archive_read_ahead(a, 512, NULL);
371 /* If it's an end-of-archive mark, we can handle it. */
372 if (h[0] == 0 && archive_block_is_null(h)) {
374 * Usually, I bid the number of bits verified, but
375 * in this case, 4096 seems excessive so I picked 10 as
376 * an arbitrary but reasonable-seeming value.
381 /* If it's not an end-of-archive mark, it must have a valid checksum.*/
384 bid += 48; /* Checksum is usually 6 octal digits. */
386 header = (const struct archive_entry_header_ustar *)h;
388 /* Recognize POSIX formats. */
389 if ((memcmp(header->magic, "ustar\0", 6) == 0)
390 && (memcmp(header->version, "00", 2) == 0))
393 /* Recognize GNU tar format. */
394 if ((memcmp(header->magic, "ustar ", 6) == 0)
395 && (memcmp(header->version, " \0", 2) == 0))
398 /* Type flag must be null, digit or A-Z, a-z. */
399 if (header->typeflag[0] != 0 &&
400 !( header->typeflag[0] >= '0' && header->typeflag[0] <= '9') &&
401 !( header->typeflag[0] >= 'A' && header->typeflag[0] <= 'Z') &&
402 !( header->typeflag[0] >= 'a' && header->typeflag[0] <= 'z') )
404 bid += 2; /* 6 bits of variation in an 8-bit field leaves 2 bits. */
407 * Check format of mode/uid/gid/mtime/size/rdevmajor/rdevminor fields.
410 validate_number_field(header->mode, sizeof(header->mode)) == 0
411 || validate_number_field(header->uid, sizeof(header->uid)) == 0
412 || validate_number_field(header->gid, sizeof(header->gid)) == 0
413 || validate_number_field(header->mtime, sizeof(header->mtime)) == 0
414 || validate_number_field(header->size, sizeof(header->size)) == 0
415 || validate_number_field(header->rdevmajor, sizeof(header->rdevmajor)) == 0
416 || validate_number_field(header->rdevminor, sizeof(header->rdevminor)) == 0)) {
424 archive_read_format_tar_options(struct archive_read *a,
425 const char *key, const char *val)
428 int ret = ARCHIVE_FAILED;
430 tar = (struct tar *)(a->format->data);
431 if (strcmp(key, "compat-2x") == 0) {
432 /* Handle UTF-8 filenames as libarchive 2.x */
433 tar->compat_2x = (val != NULL && val[0] != 0);
434 tar->init_default_conversion = tar->compat_2x;
436 } else if (strcmp(key, "hdrcharset") == 0) {
437 if (val == NULL || val[0] == 0)
438 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
439 "tar: hdrcharset option needs a character-set name");
442 archive_string_conversion_from_charset(
443 &a->archive, val, 0);
444 if (tar->opt_sconv != NULL)
450 } else if (strcmp(key, "mac-ext") == 0) {
451 tar->process_mac_extensions = (val != NULL && val[0] != 0);
453 } else if (strcmp(key, "read_concatenated_archives") == 0) {
454 tar->read_concatenated_archives = (val != NULL && val[0] != 0);
458 /* Note: The "warn" return is just to inform the options
459 * supervisor that we didn't handle it. It will generate
460 * a suitable error if no one used this option. */
461 return (ARCHIVE_WARN);
464 /* utility function- this exists to centralize the logic of tracking
465 * how much unconsumed data we have floating around, and to consume
466 * anything outstanding since we're going to do read_aheads
469 tar_flush_unconsumed(struct archive_read *a, size_t *unconsumed)
473 void *data = (void *)__archive_read_ahead(a, *unconsumed, NULL);
474 * this block of code is to poison claimed unconsumed space, ensuring
475 * things break if it is in use still.
476 * currently it WILL break things, so enable it only for debugging this issue
478 memset(data, 0xff, *unconsumed);
481 __archive_read_consume(a, *unconsumed);
487 * The function invoked by archive_read_next_header(). This
488 * just sets up a few things and then calls the internal
489 * tar_read_header() function below.
492 archive_read_format_tar_read_header(struct archive_read *a,
493 struct archive_entry *entry)
496 * When converting tar archives to cpio archives, it is
497 * essential that each distinct file have a distinct inode
498 * number. To simplify this, we keep a static count here to
499 * assign fake dev/inode numbers to each tar entry. Note that
500 * pax format archives may overwrite this with something more
503 * Ideally, we would track every file read from the archive so
504 * that we could assign the same dev/ino pair to hardlinks,
505 * but the memory required to store a complete lookup table is
506 * probably not worthwhile just to support the relatively
507 * obscure tar->cpio conversion case.
509 static int default_inode;
510 static int default_dev;
515 size_t l, unconsumed = 0;
517 /* Assign default device/inode values. */
518 archive_entry_set_dev(entry, 1 + default_dev); /* Don't use zero. */
519 archive_entry_set_ino(entry, ++default_inode); /* Don't use zero. */
520 /* Limit generated st_ino number to 16 bits. */
521 if (default_inode >= 0xffff) {
526 tar = (struct tar *)(a->format->data);
527 tar->entry_offset = 0;
528 gnu_clear_sparse_list(tar);
529 tar->realsize = -1; /* Mark this as "unset" */
531 /* Setup default string conversion. */
532 tar->sconv = tar->opt_sconv;
533 if (tar->sconv == NULL) {
534 if (!tar->init_default_conversion) {
536 archive_string_default_conversion_for_read(&(a->archive));
537 tar->init_default_conversion = 1;
539 tar->sconv = tar->sconv_default;
542 r = tar_read_header(a, tar, entry, &unconsumed);
544 tar_flush_unconsumed(a, &unconsumed);
547 * "non-sparse" files are really just sparse files with
550 if (tar->sparse_list == NULL) {
551 if (gnu_add_sparse_entry(a, tar, 0, tar->entry_bytes_remaining)
553 return (ARCHIVE_FATAL);
555 struct sparse_block *sb;
557 for (sb = tar->sparse_list; sb != NULL; sb = sb->next) {
559 archive_entry_sparse_add_entry(entry,
560 sb->offset, sb->remaining);
564 if (r == ARCHIVE_OK && archive_entry_filetype(entry) == AE_IFREG) {
566 * "Regular" entry with trailing '/' is really
567 * directory: This is needed for certain old tar
568 * variants and even for some broken newer ones.
570 if ((wp = archive_entry_pathname_w(entry)) != NULL) {
572 if (l > 0 && wp[l - 1] == L'/') {
573 archive_entry_set_filetype(entry, AE_IFDIR);
575 } else if ((p = archive_entry_pathname(entry)) != NULL) {
577 if (l > 0 && p[l - 1] == '/') {
578 archive_entry_set_filetype(entry, AE_IFDIR);
586 archive_read_format_tar_read_data(struct archive_read *a,
587 const void **buff, size_t *size, int64_t *offset)
591 struct sparse_block *p;
593 tar = (struct tar *)(a->format->data);
596 /* Remove exhausted entries from sparse list. */
597 while (tar->sparse_list != NULL &&
598 tar->sparse_list->remaining == 0) {
599 p = tar->sparse_list;
600 tar->sparse_list = p->next;
604 if (tar->entry_bytes_unconsumed) {
605 __archive_read_consume(a, tar->entry_bytes_unconsumed);
606 tar->entry_bytes_unconsumed = 0;
609 /* If we're at end of file, return EOF. */
610 if (tar->sparse_list == NULL ||
611 tar->entry_bytes_remaining == 0) {
612 if (__archive_read_consume(a, tar->entry_padding) < 0)
613 return (ARCHIVE_FATAL);
614 tar->entry_padding = 0;
617 *offset = tar->realsize;
618 return (ARCHIVE_EOF);
621 *buff = __archive_read_ahead(a, 1, &bytes_read);
623 return (ARCHIVE_FATAL);
625 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
626 "Truncated tar archive");
627 return (ARCHIVE_FATAL);
629 if (bytes_read > tar->entry_bytes_remaining)
630 bytes_read = (ssize_t)tar->entry_bytes_remaining;
631 /* Don't read more than is available in the
632 * current sparse block. */
633 if (tar->sparse_list->remaining < bytes_read)
634 bytes_read = (ssize_t)tar->sparse_list->remaining;
636 *offset = tar->sparse_list->offset;
637 tar->sparse_list->remaining -= bytes_read;
638 tar->sparse_list->offset += bytes_read;
639 tar->entry_bytes_remaining -= bytes_read;
640 tar->entry_bytes_unconsumed = bytes_read;
642 if (!tar->sparse_list->hole)
644 /* Current is hole data and skip this. */
649 archive_read_format_tar_skip(struct archive_read *a)
651 int64_t bytes_skipped;
653 struct sparse_block *p;
656 tar = (struct tar *)(a->format->data);
658 /* Do not consume the hole of a sparse file. */
660 for (p = tar->sparse_list; p != NULL; p = p->next) {
662 if (p->remaining >= INT64_MAX - request) {
663 return ARCHIVE_FATAL;
665 request += p->remaining;
668 if (request > tar->entry_bytes_remaining)
669 request = tar->entry_bytes_remaining;
670 request += tar->entry_padding + tar->entry_bytes_unconsumed;
672 bytes_skipped = __archive_read_consume(a, request);
673 if (bytes_skipped < 0)
674 return (ARCHIVE_FATAL);
676 tar->entry_bytes_remaining = 0;
677 tar->entry_bytes_unconsumed = 0;
678 tar->entry_padding = 0;
680 /* Free the sparse list. */
681 gnu_clear_sparse_list(tar);
687 * This function recursively interprets all of the headers associated
688 * with a single entry.
691 tar_read_header(struct archive_read *a, struct tar *tar,
692 struct archive_entry *entry, size_t *unconsumed)
697 const struct archive_entry_header_ustar *header;
698 const struct archive_entry_header_gnutar *gnuheader;
700 /* Loop until we find a workable header record. */
702 tar_flush_unconsumed(a, unconsumed);
704 /* Read 512-byte header record */
705 h = __archive_read_ahead(a, 512, &bytes);
708 if (bytes == 0) { /* EOF at a block boundary. */
709 /* Some writers do omit the block of nulls. <sigh> */
710 return (ARCHIVE_EOF);
712 if (bytes < 512) { /* Short block at EOF; this is bad. */
713 archive_set_error(&a->archive,
714 ARCHIVE_ERRNO_FILE_FORMAT,
715 "Truncated tar archive");
716 return (ARCHIVE_FATAL);
720 /* Header is workable if it's not an end-of-archive mark. */
721 if (h[0] != 0 || !archive_block_is_null(h))
724 /* Ensure format is set for archives with only null blocks. */
725 if (a->archive.archive_format_name == NULL) {
726 a->archive.archive_format = ARCHIVE_FORMAT_TAR;
727 a->archive.archive_format_name = "tar";
730 if (!tar->read_concatenated_archives) {
731 /* Try to consume a second all-null record, as well. */
732 tar_flush_unconsumed(a, unconsumed);
733 h = __archive_read_ahead(a, 512, NULL);
734 if (h != NULL && h[0] == 0 && archive_block_is_null(h))
735 __archive_read_consume(a, 512);
736 archive_clear_error(&a->archive);
737 return (ARCHIVE_EOF);
741 * We're reading concatenated archives, ignore this block and
742 * loop to get the next.
747 * Note: If the checksum fails and we return ARCHIVE_RETRY,
748 * then the client is likely to just retry. This is a very
749 * crude way to search for the next valid header!
751 * TODO: Improve this by implementing a real header scan.
753 if (!checksum(a, h)) {
754 tar_flush_unconsumed(a, unconsumed);
755 archive_set_error(&a->archive, EINVAL, "Damaged tar archive");
756 return (ARCHIVE_RETRY); /* Retryable: Invalid header */
759 if (++tar->header_recursion_depth > 32) {
760 tar_flush_unconsumed(a, unconsumed);
761 archive_set_error(&a->archive, EINVAL, "Too many special headers");
762 return (ARCHIVE_WARN);
765 /* Determine the format variant. */
766 header = (const struct archive_entry_header_ustar *)h;
768 switch(header->typeflag[0]) {
769 case 'A': /* Solaris tar ACL */
770 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE;
771 a->archive.archive_format_name = "Solaris tar";
772 err = header_Solaris_ACL(a, tar, entry, h, unconsumed);
774 case 'g': /* POSIX-standard 'g' header. */
775 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE;
776 a->archive.archive_format_name = "POSIX pax interchange format";
777 err = header_pax_global(a, tar, entry, h, unconsumed);
778 if (err == ARCHIVE_EOF)
781 case 'K': /* Long link name (GNU tar, others) */
782 err = header_longlink(a, tar, entry, h, unconsumed);
784 case 'L': /* Long filename (GNU tar, others) */
785 err = header_longname(a, tar, entry, h, unconsumed);
787 case 'V': /* GNU volume header */
788 err = header_volume(a, tar, entry, h, unconsumed);
790 case 'X': /* Used by SUN tar; same as 'x'. */
791 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE;
792 a->archive.archive_format_name =
793 "POSIX pax interchange format (Sun variant)";
794 err = header_pax_extensions(a, tar, entry, h, unconsumed);
796 case 'x': /* POSIX-standard 'x' header. */
797 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE;
798 a->archive.archive_format_name = "POSIX pax interchange format";
799 err = header_pax_extensions(a, tar, entry, h, unconsumed);
802 gnuheader = (const struct archive_entry_header_gnutar *)h;
803 if (memcmp(gnuheader->magic, "ustar \0", 8) == 0) {
804 a->archive.archive_format = ARCHIVE_FORMAT_TAR_GNUTAR;
805 a->archive.archive_format_name = "GNU tar format";
806 err = header_gnutar(a, tar, entry, h, unconsumed);
807 } else if (memcmp(header->magic, "ustar", 5) == 0) {
808 if (a->archive.archive_format != ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE) {
809 a->archive.archive_format = ARCHIVE_FORMAT_TAR_USTAR;
810 a->archive.archive_format_name = "POSIX ustar format";
812 err = header_ustar(a, tar, entry, h);
814 a->archive.archive_format = ARCHIVE_FORMAT_TAR;
815 a->archive.archive_format_name = "tar (non-POSIX)";
816 err = header_old_tar(a, tar, entry, h);
819 if (err == ARCHIVE_FATAL)
822 tar_flush_unconsumed(a, unconsumed);
827 --tar->header_recursion_depth;
828 /* Yuck. Apple's design here ends up storing long pathname
829 * extensions for both the AppleDouble extension entry and the
832 if ((err == ARCHIVE_WARN || err == ARCHIVE_OK) &&
833 tar->header_recursion_depth == 0 &&
834 tar->process_mac_extensions) {
835 int err2 = read_mac_metadata_blob(a, tar, entry, h, unconsumed);
840 /* We return warnings or success as-is. Anything else is fatal. */
841 if (err == ARCHIVE_WARN || err == ARCHIVE_OK) {
842 if (tar->sparse_gnu_pending) {
843 if (tar->sparse_gnu_major == 1 &&
844 tar->sparse_gnu_minor == 0) {
847 tar->sparse_gnu_pending = 0;
848 /* Read initial sparse map. */
849 bytes_read = gnu_sparse_10_read(a, tar, unconsumed);
850 tar->entry_bytes_remaining -= bytes_read;
852 return ((int)bytes_read);
854 archive_set_error(&a->archive,
856 "Unrecognized GNU sparse file format");
857 return (ARCHIVE_WARN);
859 tar->sparse_gnu_pending = 0;
863 if (err == ARCHIVE_EOF)
864 /* EOF when recursively reading a header is bad. */
865 archive_set_error(&a->archive, EINVAL, "Damaged tar archive");
866 return (ARCHIVE_FATAL);
870 * Return true if block checksum is correct.
873 checksum(struct archive_read *a, const void *h)
875 const unsigned char *bytes;
876 const struct archive_entry_header_ustar *header;
880 (void)a; /* UNUSED */
881 bytes = (const unsigned char *)h;
882 header = (const struct archive_entry_header_ustar *)h;
884 /* Checksum field must hold an octal number */
885 for (i = 0; i < sizeof(header->checksum); ++i) {
886 char c = header->checksum[i];
887 if (c != ' ' && c != '\0' && (c < '0' || c > '7'))
892 * Test the checksum. Note that POSIX specifies _unsigned_
893 * bytes for this calculation.
895 sum = (int)tar_atol(header->checksum, sizeof(header->checksum));
897 for (i = 0; i < 148; i++)
898 check += (unsigned char)bytes[i];
902 check += (unsigned char)bytes[i];
907 * Repeat test with _signed_ bytes, just in case this archive
908 * was created by an old BSD, Solaris, or HP-UX tar with a
909 * broken checksum calculation.
912 for (i = 0; i < 148; i++)
913 check += (signed char)bytes[i];
917 check += (signed char)bytes[i];
925 * Return true if this block contains only nulls.
928 archive_block_is_null(const char *p)
932 for (i = 0; i < 512; i++)
939 * Interpret 'A' Solaris ACL header
942 header_Solaris_ACL(struct archive_read *a, struct tar *tar,
943 struct archive_entry *entry, const void *h, size_t *unconsumed)
945 const struct archive_entry_header_ustar *header;
952 * read_body_to_string adds a NUL terminator, but we need a little
953 * more to make sure that we don't overrun acl_text later.
955 header = (const struct archive_entry_header_ustar *)h;
956 size = (size_t)tar_atol(header->size, sizeof(header->size));
957 err = read_body_to_string(a, tar, &(tar->acl_text), h, unconsumed);
958 if (err != ARCHIVE_OK)
961 /* Recursively read next header */
962 err = tar_read_header(a, tar, entry, unconsumed);
963 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN))
966 /* TODO: Examine the first characters to see if this
967 * is an AIX ACL descriptor. We'll likely never support
968 * them, but it would be polite to recognize and warn when
971 /* Leading octal number indicates ACL type and number of entries. */
972 p = acl = tar->acl_text.s;
974 while (*p != '\0' && p < acl + size) {
975 if (*p < '0' || *p > '7') {
976 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
977 "Malformed Solaris ACL attribute (invalid digit)");
978 return(ARCHIVE_WARN);
982 if (type > 077777777) {
983 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
984 "Malformed Solaris ACL attribute (count too large)");
985 return (ARCHIVE_WARN);
989 switch ((int)type & ~0777777) {
994 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
995 "Solaris NFSv4 ACLs not supported");
996 return (ARCHIVE_WARN);
998 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
999 "Malformed Solaris ACL attribute (unsupported type %o)",
1001 return (ARCHIVE_WARN);
1005 if (p >= acl + size) {
1006 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1007 "Malformed Solaris ACL attribute (body overflow)");
1008 return(ARCHIVE_WARN);
1011 /* ACL text is null-terminated; find the end. */
1015 while (*p != '\0' && p < acl + size)
1018 if (tar->sconv_acl == NULL) {
1019 tar->sconv_acl = archive_string_conversion_from_charset(
1020 &(a->archive), "UTF-8", 1);
1021 if (tar->sconv_acl == NULL)
1022 return (ARCHIVE_FATAL);
1024 archive_strncpy(&(tar->localname), acl, p - acl);
1025 err = archive_acl_from_text_l(archive_entry_acl(entry),
1026 tar->localname.s, ARCHIVE_ENTRY_ACL_TYPE_ACCESS, tar->sconv_acl);
1027 if (err != ARCHIVE_OK) {
1028 if (errno == ENOMEM) {
1029 archive_set_error(&a->archive, ENOMEM,
1030 "Can't allocate memory for ACL");
1032 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1033 "Malformed Solaris ACL attribute (unparsable)");
1039 * Interpret 'K' long linkname header.
1042 header_longlink(struct archive_read *a, struct tar *tar,
1043 struct archive_entry *entry, const void *h, size_t *unconsumed)
1047 err = read_body_to_string(a, tar, &(tar->longlink), h, unconsumed);
1048 if (err != ARCHIVE_OK)
1050 err = tar_read_header(a, tar, entry, unconsumed);
1051 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN))
1053 /* Set symlink if symlink already set, else hardlink. */
1054 archive_entry_copy_link(entry, tar->longlink.s);
1055 return (ARCHIVE_OK);
1059 set_conversion_failed_error(struct archive_read *a,
1060 struct archive_string_conv *sconv, const char *name)
1062 if (errno == ENOMEM) {
1063 archive_set_error(&a->archive, ENOMEM,
1064 "Can't allocate memory for %s", name);
1065 return (ARCHIVE_FATAL);
1067 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1068 "%s can't be converted from %s to current locale.",
1069 name, archive_string_conversion_charset_name(sconv));
1070 return (ARCHIVE_WARN);
1074 * Interpret 'L' long filename header.
1077 header_longname(struct archive_read *a, struct tar *tar,
1078 struct archive_entry *entry, const void *h, size_t *unconsumed)
1082 err = read_body_to_string(a, tar, &(tar->longname), h, unconsumed);
1083 if (err != ARCHIVE_OK)
1085 /* Read and parse "real" header, then override name. */
1086 err = tar_read_header(a, tar, entry, unconsumed);
1087 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN))
1089 if (archive_entry_copy_pathname_l(entry, tar->longname.s,
1090 archive_strlen(&(tar->longname)), tar->sconv) != 0)
1091 err = set_conversion_failed_error(a, tar->sconv, "Pathname");
1097 * Interpret 'V' GNU tar volume header.
1100 header_volume(struct archive_read *a, struct tar *tar,
1101 struct archive_entry *entry, const void *h, size_t *unconsumed)
1105 /* Just skip this and read the next header. */
1106 return (tar_read_header(a, tar, entry, unconsumed));
1110 * Read body of an archive entry into an archive_string object.
1113 read_body_to_string(struct archive_read *a, struct tar *tar,
1114 struct archive_string *as, const void *h, size_t *unconsumed)
1117 const struct archive_entry_header_ustar *header;
1120 (void)tar; /* UNUSED */
1121 header = (const struct archive_entry_header_ustar *)h;
1122 size = tar_atol(header->size, sizeof(header->size));
1123 if ((size > 1048576) || (size < 0)) {
1124 archive_set_error(&a->archive, EINVAL,
1125 "Special header too large");
1126 return (ARCHIVE_FATAL);
1129 /* Fail if we can't make our buffer big enough. */
1130 if (archive_string_ensure(as, (size_t)size+1) == NULL) {
1131 archive_set_error(&a->archive, ENOMEM,
1133 return (ARCHIVE_FATAL);
1136 tar_flush_unconsumed(a, unconsumed);
1138 /* Read the body into the string. */
1139 *unconsumed = (size_t)((size + 511) & ~ 511);
1140 src = __archive_read_ahead(a, *unconsumed, NULL);
1143 return (ARCHIVE_FATAL);
1145 memcpy(as->s, src, (size_t)size);
1147 as->length = (size_t)size;
1148 return (ARCHIVE_OK);
1152 * Parse out common header elements.
1154 * This would be the same as header_old_tar, except that the
1155 * filename is handled slightly differently for old and POSIX
1156 * entries (POSIX entries support a 'prefix'). This factoring
1157 * allows header_old_tar and header_ustar
1158 * to handle filenames differently, while still putting most of the
1159 * common parsing into one place.
1162 header_common(struct archive_read *a, struct tar *tar,
1163 struct archive_entry *entry, const void *h)
1165 const struct archive_entry_header_ustar *header;
1167 int err = ARCHIVE_OK;
1169 header = (const struct archive_entry_header_ustar *)h;
1170 if (header->linkname[0])
1171 archive_strncpy(&(tar->entry_linkpath),
1172 header->linkname, sizeof(header->linkname));
1174 archive_string_empty(&(tar->entry_linkpath));
1176 /* Parse out the numeric fields (all are octal) */
1177 archive_entry_set_mode(entry,
1178 (mode_t)tar_atol(header->mode, sizeof(header->mode)));
1179 archive_entry_set_uid(entry, tar_atol(header->uid, sizeof(header->uid)));
1180 archive_entry_set_gid(entry, tar_atol(header->gid, sizeof(header->gid)));
1181 tar->entry_bytes_remaining = tar_atol(header->size, sizeof(header->size));
1182 if (tar->entry_bytes_remaining < 0) {
1183 tar->entry_bytes_remaining = 0;
1184 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1185 "Tar entry has negative size");
1186 return (ARCHIVE_FATAL);
1188 if (tar->entry_bytes_remaining == INT64_MAX) {
1189 /* Note: tar_atol returns INT64_MAX on overflow */
1190 tar->entry_bytes_remaining = 0;
1191 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1192 "Tar entry size overflow");
1193 return (ARCHIVE_FATAL);
1195 tar->realsize = tar->entry_bytes_remaining;
1196 archive_entry_set_size(entry, tar->entry_bytes_remaining);
1197 archive_entry_set_mtime(entry, tar_atol(header->mtime, sizeof(header->mtime)), 0);
1199 /* Handle the tar type flag appropriately. */
1200 tartype = header->typeflag[0];
1203 case '1': /* Hard link */
1204 if (archive_entry_copy_hardlink_l(entry, tar->entry_linkpath.s,
1205 archive_strlen(&(tar->entry_linkpath)), tar->sconv) != 0) {
1206 err = set_conversion_failed_error(a, tar->sconv,
1208 if (err == ARCHIVE_FATAL)
1212 * The following may seem odd, but: Technically, tar
1213 * does not store the file type for a "hard link"
1214 * entry, only the fact that it is a hard link. So, I
1215 * leave the type zero normally. But, pax interchange
1216 * format allows hard links to have data, which
1217 * implies that the underlying entry is a regular
1220 if (archive_entry_size(entry) > 0)
1221 archive_entry_set_filetype(entry, AE_IFREG);
1224 * A tricky point: Traditionally, tar readers have
1225 * ignored the size field when reading hardlink
1226 * entries, and some writers put non-zero sizes even
1227 * though the body is empty. POSIX blessed this
1228 * convention in the 1988 standard, but broke with
1229 * this tradition in 2001 by permitting hardlink
1230 * entries to store valid bodies in pax interchange
1231 * format, but not in ustar format. Since there is no
1232 * hard and fast way to distinguish pax interchange
1233 * from earlier archives (the 'x' and 'g' entries are
1234 * optional, after all), we need a heuristic.
1236 if (archive_entry_size(entry) == 0) {
1237 /* If the size is already zero, we're done. */
1238 } else if (a->archive.archive_format
1239 == ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE) {
1240 /* Definitely pax extended; must obey hardlink size. */
1241 } else if (a->archive.archive_format == ARCHIVE_FORMAT_TAR
1242 || a->archive.archive_format == ARCHIVE_FORMAT_TAR_GNUTAR)
1244 /* Old-style or GNU tar: we must ignore the size. */
1245 archive_entry_set_size(entry, 0);
1246 tar->entry_bytes_remaining = 0;
1247 } else if (archive_read_format_tar_bid(a, 50) > 50) {
1249 * We don't know if it's pax: If the bid
1250 * function sees a valid ustar header
1251 * immediately following, then let's ignore
1252 * the hardlink size.
1254 archive_entry_set_size(entry, 0);
1255 tar->entry_bytes_remaining = 0;
1258 * TODO: There are still two cases I'd like to handle:
1259 * = a ustar non-pax archive with a hardlink entry at
1260 * end-of-archive. (Look for block of nulls following?)
1261 * = a pax archive that has not seen any pax headers
1262 * and has an entry which is a hardlink entry storing
1263 * a body containing an uncompressed tar archive.
1264 * The first is worth addressing; I don't see any reliable
1265 * way to deal with the second possibility.
1268 case '2': /* Symlink */
1269 archive_entry_set_filetype(entry, AE_IFLNK);
1270 archive_entry_set_size(entry, 0);
1271 tar->entry_bytes_remaining = 0;
1272 if (archive_entry_copy_symlink_l(entry, tar->entry_linkpath.s,
1273 archive_strlen(&(tar->entry_linkpath)), tar->sconv) != 0) {
1274 err = set_conversion_failed_error(a, tar->sconv,
1276 if (err == ARCHIVE_FATAL)
1280 case '3': /* Character device */
1281 archive_entry_set_filetype(entry, AE_IFCHR);
1282 archive_entry_set_size(entry, 0);
1283 tar->entry_bytes_remaining = 0;
1285 case '4': /* Block device */
1286 archive_entry_set_filetype(entry, AE_IFBLK);
1287 archive_entry_set_size(entry, 0);
1288 tar->entry_bytes_remaining = 0;
1291 archive_entry_set_filetype(entry, AE_IFDIR);
1292 archive_entry_set_size(entry, 0);
1293 tar->entry_bytes_remaining = 0;
1295 case '6': /* FIFO device */
1296 archive_entry_set_filetype(entry, AE_IFIFO);
1297 archive_entry_set_size(entry, 0);
1298 tar->entry_bytes_remaining = 0;
1300 case 'D': /* GNU incremental directory type */
1302 * No special handling is actually required here.
1303 * It might be nice someday to preprocess the file list and
1304 * provide it to the client, though.
1306 archive_entry_set_filetype(entry, AE_IFDIR);
1308 case 'M': /* GNU "Multi-volume" (remainder of file from last archive)*/
1310 * As far as I can tell, this is just like a regular file
1311 * entry, except that the contents should be _appended_ to
1312 * the indicated file at the indicated offset. This may
1313 * require some API work to fully support.
1316 case 'N': /* Old GNU "long filename" entry. */
1317 /* The body of this entry is a script for renaming
1318 * previously-extracted entries. Ugh. It will never
1319 * be supported by libarchive. */
1320 archive_entry_set_filetype(entry, AE_IFREG);
1322 case 'S': /* GNU sparse files */
1324 * Sparse files are really just regular files with
1325 * sparse information in the extended area.
1330 * Enable sparse file "read" support only for regular
1331 * files and explicit GNU sparse files. However, we
1332 * don't allow non-standard file types to be sparse.
1334 tar->sparse_allowed = 1;
1336 default: /* Regular file and non-standard types */
1338 * Per POSIX: non-recognized types should always be
1339 * treated as regular files.
1341 archive_entry_set_filetype(entry, AE_IFREG);
1348 * Parse out header elements for "old-style" tar archives.
1351 header_old_tar(struct archive_read *a, struct tar *tar,
1352 struct archive_entry *entry, const void *h)
1354 const struct archive_entry_header_ustar *header;
1355 int err = ARCHIVE_OK, err2;
1357 /* Copy filename over (to ensure null termination). */
1358 header = (const struct archive_entry_header_ustar *)h;
1359 if (archive_entry_copy_pathname_l(entry,
1360 header->name, sizeof(header->name), tar->sconv) != 0) {
1361 err = set_conversion_failed_error(a, tar->sconv, "Pathname");
1362 if (err == ARCHIVE_FATAL)
1366 /* Grab rest of common fields */
1367 err2 = header_common(a, tar, entry, h);
1371 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining);
1376 * Read a Mac AppleDouble-encoded blob of file metadata,
1380 read_mac_metadata_blob(struct archive_read *a, struct tar *tar,
1381 struct archive_entry *entry, const void *h, size_t *unconsumed)
1385 const char *p, *name;
1386 const wchar_t *wp, *wname;
1388 (void)h; /* UNUSED */
1390 wname = wp = archive_entry_pathname_w(entry);
1392 /* Find the last path element. */
1393 for (; *wp != L'\0'; ++wp) {
1394 if (wp[0] == '/' && wp[1] != L'\0')
1398 * If last path element starts with "._", then
1399 * this is a Mac extension.
1401 if (wname[0] != L'.' || wname[1] != L'_' || wname[2] == L'\0')
1404 /* Find the last path element. */
1405 name = p = archive_entry_pathname(entry);
1407 return (ARCHIVE_FAILED);
1408 for (; *p != '\0'; ++p) {
1409 if (p[0] == '/' && p[1] != '\0')
1413 * If last path element starts with "._", then
1414 * this is a Mac extension.
1416 if (name[0] != '.' || name[1] != '_' || name[2] == '\0')
1420 /* Read the body as a Mac OS metadata blob. */
1421 size = archive_entry_size(entry);
1424 * TODO: Look beyond the body here to peek at the next header.
1425 * If it's a regular header (not an extension header)
1426 * that has the wrong name, just return the current
1427 * entry as-is, without consuming the body here.
1428 * That would reduce the risk of us mis-identifying
1429 * an ordinary file that just happened to have
1430 * a name starting with "._".
1432 * Q: Is the above idea really possible? Even
1433 * when there are GNU or pax extension entries?
1435 data = __archive_read_ahead(a, (size_t)size, NULL);
1438 return (ARCHIVE_FATAL);
1440 archive_entry_copy_mac_metadata(entry, data, (size_t)size);
1441 *unconsumed = (size_t)((size + 511) & ~ 511);
1442 tar_flush_unconsumed(a, unconsumed);
1443 return (tar_read_header(a, tar, entry, unconsumed));
1447 * Parse a file header for a pax extended archive entry.
1450 header_pax_global(struct archive_read *a, struct tar *tar,
1451 struct archive_entry *entry, const void *h, size_t *unconsumed)
1455 err = read_body_to_string(a, tar, &(tar->pax_global), h, unconsumed);
1456 if (err != ARCHIVE_OK)
1458 err = tar_read_header(a, tar, entry, unconsumed);
1463 header_pax_extensions(struct archive_read *a, struct tar *tar,
1464 struct archive_entry *entry, const void *h, size_t *unconsumed)
1468 err = read_body_to_string(a, tar, &(tar->pax_header), h, unconsumed);
1469 if (err != ARCHIVE_OK)
1472 /* Parse the next header. */
1473 err = tar_read_header(a, tar, entry, unconsumed);
1474 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN))
1478 * TODO: Parse global/default options into 'entry' struct here
1479 * before handling file-specific options.
1481 * This design (parse standard header, then overwrite with pax
1482 * extended attribute data) usually works well, but isn't ideal;
1483 * it would be better to parse the pax extended attributes first
1484 * and then skip any fields in the standard header that were
1485 * defined in the pax header.
1487 err2 = pax_header(a, tar, entry, &tar->pax_header);
1488 err = err_combine(err, err2);
1489 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining);
1495 * Parse a file header for a Posix "ustar" archive entry. This also
1496 * handles "pax" or "extended ustar" entries.
1499 header_ustar(struct archive_read *a, struct tar *tar,
1500 struct archive_entry *entry, const void *h)
1502 const struct archive_entry_header_ustar *header;
1503 struct archive_string *as;
1504 int err = ARCHIVE_OK, r;
1506 header = (const struct archive_entry_header_ustar *)h;
1508 /* Copy name into an internal buffer to ensure null-termination. */
1509 as = &(tar->entry_pathname);
1510 if (header->prefix[0]) {
1511 archive_strncpy(as, header->prefix, sizeof(header->prefix));
1512 if (as->s[archive_strlen(as) - 1] != '/')
1513 archive_strappend_char(as, '/');
1514 archive_strncat(as, header->name, sizeof(header->name));
1516 archive_strncpy(as, header->name, sizeof(header->name));
1518 if (archive_entry_copy_pathname_l(entry, as->s, archive_strlen(as),
1520 err = set_conversion_failed_error(a, tar->sconv, "Pathname");
1521 if (err == ARCHIVE_FATAL)
1525 /* Handle rest of common fields. */
1526 r = header_common(a, tar, entry, h);
1527 if (r == ARCHIVE_FATAL)
1532 /* Handle POSIX ustar fields. */
1533 if (archive_entry_copy_uname_l(entry,
1534 header->uname, sizeof(header->uname), tar->sconv) != 0) {
1535 err = set_conversion_failed_error(a, tar->sconv, "Uname");
1536 if (err == ARCHIVE_FATAL)
1540 if (archive_entry_copy_gname_l(entry,
1541 header->gname, sizeof(header->gname), tar->sconv) != 0) {
1542 err = set_conversion_failed_error(a, tar->sconv, "Gname");
1543 if (err == ARCHIVE_FATAL)
1547 /* Parse out device numbers only for char and block specials. */
1548 if (header->typeflag[0] == '3' || header->typeflag[0] == '4') {
1549 archive_entry_set_rdevmajor(entry, (dev_t)
1550 tar_atol(header->rdevmajor, sizeof(header->rdevmajor)));
1551 archive_entry_set_rdevminor(entry, (dev_t)
1552 tar_atol(header->rdevminor, sizeof(header->rdevminor)));
1555 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining);
1562 * Parse the pax extended attributes record.
1564 * Returns non-zero if there's an error in the data.
1567 pax_header(struct archive_read *a, struct tar *tar,
1568 struct archive_entry *entry, struct archive_string *in_as)
1570 size_t attr_length, l, line_length, value_length;
1573 struct archive_string *as;
1574 struct archive_string_conv *sconv;
1576 char *attr = in_as->s;
1578 attr_length = in_as->length;
1579 tar->pax_hdrcharset_binary = 0;
1580 archive_string_empty(&(tar->entry_gname));
1581 archive_string_empty(&(tar->entry_linkpath));
1582 archive_string_empty(&(tar->entry_pathname));
1583 archive_string_empty(&(tar->entry_pathname_override));
1584 archive_string_empty(&(tar->entry_uname));
1586 while (attr_length > 0) {
1587 /* Parse decimal length field at start of line. */
1590 p = attr; /* Record start of line. */
1597 if (*p < '0' || *p > '9') {
1598 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1599 "Ignoring malformed pax extended attributes");
1600 return (ARCHIVE_WARN);
1603 line_length += *p - '0';
1604 if (line_length > 999999) {
1605 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1606 "Rejecting pax extended attribute > 1MB");
1607 return (ARCHIVE_WARN);
1614 * Parsed length must be no bigger than available data,
1615 * at least 1, and the last character of the line must
1618 if (line_length > attr_length
1620 || attr[line_length - 1] != '\n')
1622 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1623 "Ignoring malformed pax extended attribute");
1624 return (ARCHIVE_WARN);
1627 /* Null-terminate the line. */
1628 attr[line_length - 1] = '\0';
1630 /* Find end of key and null terminate it. */
1634 while (*p && *p != '=')
1637 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1638 "Invalid pax extended attributes");
1639 return (ARCHIVE_WARN);
1645 /* Some values may be binary data */
1646 value_length = attr + line_length - 1 - value;
1648 /* Identify this attribute and set it in the entry. */
1649 err2 = pax_attribute(a, tar, entry, key, value, value_length);
1650 if (err2 == ARCHIVE_FATAL)
1652 err = err_combine(err, err2);
1654 /* Skip to next line */
1655 attr += line_length;
1656 attr_length -= line_length;
1660 * PAX format uses UTF-8 as default charset for its metadata
1661 * unless hdrcharset=BINARY is present in its header.
1662 * We apply the charset specified by the hdrcharset option only
1663 * when the hdrcharset attribute(in PAX header) is BINARY because
1664 * we respect the charset described in PAX header and BINARY also
1665 * means that metadata(filename,uname and gname) character-set
1668 if (tar->pax_hdrcharset_binary)
1669 sconv = tar->opt_sconv;
1671 sconv = archive_string_conversion_from_charset(
1672 &(a->archive), "UTF-8", 1);
1674 return (ARCHIVE_FATAL);
1676 archive_string_conversion_set_opt(sconv,
1677 SCONV_SET_OPT_UTF8_LIBARCHIVE2X);
1680 if (archive_strlen(&(tar->entry_gname)) > 0) {
1681 if (archive_entry_copy_gname_l(entry, tar->entry_gname.s,
1682 archive_strlen(&(tar->entry_gname)), sconv) != 0) {
1683 err = set_conversion_failed_error(a, sconv, "Gname");
1684 if (err == ARCHIVE_FATAL)
1686 /* Use a converted an original name. */
1687 archive_entry_copy_gname(entry, tar->entry_gname.s);
1690 if (archive_strlen(&(tar->entry_linkpath)) > 0) {
1691 if (archive_entry_copy_link_l(entry, tar->entry_linkpath.s,
1692 archive_strlen(&(tar->entry_linkpath)), sconv) != 0) {
1693 err = set_conversion_failed_error(a, sconv, "Linkname");
1694 if (err == ARCHIVE_FATAL)
1696 /* Use a converted an original name. */
1697 archive_entry_copy_link(entry, tar->entry_linkpath.s);
1701 * Some extensions (such as the GNU sparse file extensions)
1702 * deliberately store a synthetic name under the regular 'path'
1703 * attribute and the real file name under a different attribute.
1704 * Since we're supposed to not care about the order, we
1705 * have no choice but to store all of the various filenames
1706 * we find and figure it all out afterwards. This is the
1707 * figuring out part.
1710 if (archive_strlen(&(tar->entry_pathname_override)) > 0)
1711 as = &(tar->entry_pathname_override);
1712 else if (archive_strlen(&(tar->entry_pathname)) > 0)
1713 as = &(tar->entry_pathname);
1715 if (archive_entry_copy_pathname_l(entry, as->s,
1716 archive_strlen(as), sconv) != 0) {
1717 err = set_conversion_failed_error(a, sconv, "Pathname");
1718 if (err == ARCHIVE_FATAL)
1720 /* Use a converted an original name. */
1721 archive_entry_copy_pathname(entry, as->s);
1724 if (archive_strlen(&(tar->entry_uname)) > 0) {
1725 if (archive_entry_copy_uname_l(entry, tar->entry_uname.s,
1726 archive_strlen(&(tar->entry_uname)), sconv) != 0) {
1727 err = set_conversion_failed_error(a, sconv, "Uname");
1728 if (err == ARCHIVE_FATAL)
1730 /* Use a converted an original name. */
1731 archive_entry_copy_uname(entry, tar->entry_uname.s);
1738 pax_attribute_xattr(struct archive_entry *entry,
1739 const char *name, const char *value)
1742 void *value_decoded;
1745 if (strlen(name) < 18 || (memcmp(name, "LIBARCHIVE.xattr.", 17)) != 0)
1750 /* URL-decode name */
1751 name_decoded = url_decode(name);
1752 if (name_decoded == NULL)
1755 /* Base-64 decode value */
1756 value_decoded = base64_decode(value, strlen(value), &value_len);
1757 if (value_decoded == NULL) {
1762 archive_entry_xattr_add_entry(entry, name_decoded,
1763 value_decoded, value_len);
1766 free(value_decoded);
1771 pax_attribute_schily_xattr(struct archive_entry *entry,
1772 const char *name, const char *value, size_t value_length)
1774 if (strlen(name) < 14 || (memcmp(name, "SCHILY.xattr.", 13)) != 0)
1779 archive_entry_xattr_add_entry(entry, name, value, value_length);
1785 pax_attribute_acl(struct archive_read *a, struct tar *tar,
1786 struct archive_entry *entry, const char *value, int type)
1792 case ARCHIVE_ENTRY_ACL_TYPE_ACCESS:
1793 errstr = "SCHILY.acl.access";
1795 case ARCHIVE_ENTRY_ACL_TYPE_DEFAULT:
1796 errstr = "SCHILY.acl.default";
1798 case ARCHIVE_ENTRY_ACL_TYPE_NFS4:
1799 errstr = "SCHILY.acl.ace";
1802 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1803 "Unknown ACL type: %d", type);
1804 return(ARCHIVE_FATAL);
1807 if (tar->sconv_acl == NULL) {
1809 archive_string_conversion_from_charset(
1810 &(a->archive), "UTF-8", 1);
1811 if (tar->sconv_acl == NULL)
1812 return (ARCHIVE_FATAL);
1815 r = archive_acl_from_text_l(archive_entry_acl(entry), value, type,
1817 if (r != ARCHIVE_OK) {
1818 if (r == ARCHIVE_FATAL) {
1819 archive_set_error(&a->archive, ENOMEM,
1820 "%s %s", "Can't allocate memory for ",
1824 archive_set_error(&a->archive,
1825 ARCHIVE_ERRNO_MISC, "%s %s", "Parse error: ", errstr);
1831 * Parse a single key=value attribute. key/value pointers are
1832 * assumed to point into reasonably long-lived storage.
1834 * Note that POSIX reserves all-lowercase keywords. Vendor-specific
1835 * extensions should always have keywords of the form "VENDOR.attribute"
1836 * In particular, it's quite feasible to support many different
1837 * vendor extensions here. I'm using "LIBARCHIVE" for extensions
1838 * unique to this library.
1840 * Investigate other vendor-specific extensions and see if
1841 * any of them look useful.
1844 pax_attribute(struct archive_read *a, struct tar *tar,
1845 struct archive_entry *entry, const char *key, const char *value, size_t value_length)
1849 int err = ARCHIVE_OK, r;
1853 value = ""; /* Disable compiler warning; do not pass
1854 * NULL pointer to strlen(). */
1858 /* Reject GNU.sparse.* headers on non-regular files. */
1859 if (strncmp(key, "GNU.sparse", 10) == 0 &&
1860 !tar->sparse_allowed) {
1861 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1862 "Non-regular file cannot be sparse");
1863 return (ARCHIVE_FATAL);
1866 /* GNU "0.0" sparse pax format. */
1867 if (strcmp(key, "GNU.sparse.numblocks") == 0) {
1868 tar->sparse_offset = -1;
1869 tar->sparse_numbytes = -1;
1870 tar->sparse_gnu_major = 0;
1871 tar->sparse_gnu_minor = 0;
1873 if (strcmp(key, "GNU.sparse.offset") == 0) {
1874 tar->sparse_offset = tar_atol10(value, strlen(value));
1875 if (tar->sparse_numbytes != -1) {
1876 if (gnu_add_sparse_entry(a, tar,
1877 tar->sparse_offset, tar->sparse_numbytes)
1879 return (ARCHIVE_FATAL);
1880 tar->sparse_offset = -1;
1881 tar->sparse_numbytes = -1;
1884 if (strcmp(key, "GNU.sparse.numbytes") == 0) {
1885 tar->sparse_numbytes = tar_atol10(value, strlen(value));
1886 if (tar->sparse_numbytes != -1) {
1887 if (gnu_add_sparse_entry(a, tar,
1888 tar->sparse_offset, tar->sparse_numbytes)
1890 return (ARCHIVE_FATAL);
1891 tar->sparse_offset = -1;
1892 tar->sparse_numbytes = -1;
1895 if (strcmp(key, "GNU.sparse.size") == 0) {
1896 tar->realsize = tar_atol10(value, strlen(value));
1897 archive_entry_set_size(entry, tar->realsize);
1900 /* GNU "0.1" sparse pax format. */
1901 if (strcmp(key, "GNU.sparse.map") == 0) {
1902 tar->sparse_gnu_major = 0;
1903 tar->sparse_gnu_minor = 1;
1904 if (gnu_sparse_01_parse(a, tar, value) != ARCHIVE_OK)
1905 return (ARCHIVE_WARN);
1908 /* GNU "1.0" sparse pax format */
1909 if (strcmp(key, "GNU.sparse.major") == 0) {
1910 tar->sparse_gnu_major = (int)tar_atol10(value, strlen(value));
1911 tar->sparse_gnu_pending = 1;
1913 if (strcmp(key, "GNU.sparse.minor") == 0) {
1914 tar->sparse_gnu_minor = (int)tar_atol10(value, strlen(value));
1915 tar->sparse_gnu_pending = 1;
1917 if (strcmp(key, "GNU.sparse.name") == 0) {
1919 * The real filename; when storing sparse
1920 * files, GNU tar puts a synthesized name into
1921 * the regular 'path' attribute in an attempt
1922 * to limit confusion. ;-)
1924 archive_strcpy(&(tar->entry_pathname_override), value);
1926 if (strcmp(key, "GNU.sparse.realsize") == 0) {
1927 tar->realsize = tar_atol10(value, strlen(value));
1928 archive_entry_set_size(entry, tar->realsize);
1932 /* Our extensions */
1933 /* TODO: Handle arbitrary extended attributes... */
1935 if (strcmp(key, "LIBARCHIVE.xxxxxxx") == 0)
1936 archive_entry_set_xxxxxx(entry, value);
1938 if (strcmp(key, "LIBARCHIVE.creationtime") == 0) {
1939 pax_time(value, &s, &n);
1940 archive_entry_set_birthtime(entry, s, n);
1942 if (memcmp(key, "LIBARCHIVE.xattr.", 17) == 0)
1943 pax_attribute_xattr(entry, key, value);
1946 /* We support some keys used by the "star" archiver */
1947 if (strcmp(key, "SCHILY.acl.access") == 0) {
1948 r = pax_attribute_acl(a, tar, entry, value,
1949 ARCHIVE_ENTRY_ACL_TYPE_ACCESS);
1950 if (r == ARCHIVE_FATAL)
1952 } else if (strcmp(key, "SCHILY.acl.default") == 0) {
1953 r = pax_attribute_acl(a, tar, entry, value,
1954 ARCHIVE_ENTRY_ACL_TYPE_DEFAULT);
1955 if (r == ARCHIVE_FATAL)
1957 } else if (strcmp(key, "SCHILY.acl.ace") == 0) {
1958 r = pax_attribute_acl(a, tar, entry, value,
1959 ARCHIVE_ENTRY_ACL_TYPE_NFS4);
1960 if (r == ARCHIVE_FATAL)
1962 } else if (strcmp(key, "SCHILY.devmajor") == 0) {
1963 archive_entry_set_rdevmajor(entry,
1964 (dev_t)tar_atol10(value, strlen(value)));
1965 } else if (strcmp(key, "SCHILY.devminor") == 0) {
1966 archive_entry_set_rdevminor(entry,
1967 (dev_t)tar_atol10(value, strlen(value)));
1968 } else if (strcmp(key, "SCHILY.fflags") == 0) {
1969 archive_entry_copy_fflags_text(entry, value);
1970 } else if (strcmp(key, "SCHILY.dev") == 0) {
1971 archive_entry_set_dev(entry,
1972 (dev_t)tar_atol10(value, strlen(value)));
1973 } else if (strcmp(key, "SCHILY.ino") == 0) {
1974 archive_entry_set_ino(entry,
1975 tar_atol10(value, strlen(value)));
1976 } else if (strcmp(key, "SCHILY.nlink") == 0) {
1977 archive_entry_set_nlink(entry, (unsigned)
1978 tar_atol10(value, strlen(value)));
1979 } else if (strcmp(key, "SCHILY.realsize") == 0) {
1980 tar->realsize = tar_atol10(value, strlen(value));
1981 archive_entry_set_size(entry, tar->realsize);
1982 } else if (strncmp(key, "SCHILY.xattr.", 13) == 0) {
1983 pax_attribute_schily_xattr(entry, key, value,
1985 } else if (strcmp(key, "SUN.holesdata") == 0) {
1986 /* A Solaris extension for sparse. */
1987 r = solaris_sparse_parse(a, tar, entry, value);
1989 if (r == ARCHIVE_FATAL)
1992 archive_set_error(&a->archive,
1994 "Parse error: SUN.holesdata");
1999 if (strcmp(key, "atime") == 0) {
2000 pax_time(value, &s, &n);
2001 archive_entry_set_atime(entry, s, n);
2005 if (strcmp(key, "ctime") == 0) {
2006 pax_time(value, &s, &n);
2007 archive_entry_set_ctime(entry, s, n);
2008 } else if (strcmp(key, "charset") == 0) {
2009 /* TODO: Publish charset information in entry. */
2010 } else if (strcmp(key, "comment") == 0) {
2011 /* TODO: Publish comment in entry. */
2015 if (strcmp(key, "gid") == 0) {
2016 archive_entry_set_gid(entry,
2017 tar_atol10(value, strlen(value)));
2018 } else if (strcmp(key, "gname") == 0) {
2019 archive_strcpy(&(tar->entry_gname), value);
2023 if (strcmp(key, "hdrcharset") == 0) {
2024 if (strcmp(value, "BINARY") == 0)
2026 tar->pax_hdrcharset_binary = 1;
2027 else if (strcmp(value, "ISO-IR 10646 2000 UTF-8") == 0)
2028 tar->pax_hdrcharset_binary = 0;
2032 /* pax interchange doesn't distinguish hardlink vs. symlink. */
2033 if (strcmp(key, "linkpath") == 0) {
2034 archive_strcpy(&(tar->entry_linkpath), value);
2038 if (strcmp(key, "mtime") == 0) {
2039 pax_time(value, &s, &n);
2040 archive_entry_set_mtime(entry, s, n);
2044 if (strcmp(key, "path") == 0) {
2045 archive_strcpy(&(tar->entry_pathname), value);
2049 /* POSIX has reserved 'realtime.*' */
2052 /* POSIX has reserved 'security.*' */
2053 /* Someday: if (strcmp(key, "security.acl") == 0) { ... } */
2054 if (strcmp(key, "size") == 0) {
2055 /* "size" is the size of the data in the entry. */
2056 tar->entry_bytes_remaining
2057 = tar_atol10(value, strlen(value));
2059 * But, "size" is not necessarily the size of
2060 * the file on disk; if this is a sparse file,
2061 * the disk size may have already been set from
2062 * GNU.sparse.realsize or GNU.sparse.size or
2063 * an old GNU header field or SCHILY.realsize
2066 if (tar->realsize < 0) {
2067 archive_entry_set_size(entry,
2068 tar->entry_bytes_remaining);
2070 = tar->entry_bytes_remaining;
2075 if (strcmp(key, "uid") == 0) {
2076 archive_entry_set_uid(entry,
2077 tar_atol10(value, strlen(value)));
2078 } else if (strcmp(key, "uname") == 0) {
2079 archive_strcpy(&(tar->entry_uname), value);
2089 * parse a decimal time value, which may include a fractional portion
2092 pax_time(const char *p, int64_t *ps, long *pn)
2098 int64_t limit, last_digit_limit;
2100 limit = INT64_MAX / 10;
2101 last_digit_limit = INT64_MAX % 10;
2109 while (*p >= '0' && *p <= '9') {
2112 (s == limit && digit > last_digit_limit)) {
2116 s = (s * 10) + digit;
2122 /* Calculate nanoseconds. */
2131 if (*p >= '0' && *p <= '9')
2132 *pn += (*p - '0') * l;
2139 * Parse GNU tar header
2142 header_gnutar(struct archive_read *a, struct tar *tar,
2143 struct archive_entry *entry, const void *h, size_t *unconsumed)
2145 const struct archive_entry_header_gnutar *header;
2147 int err = ARCHIVE_OK;
2150 * GNU header is like POSIX ustar, except 'prefix' is
2151 * replaced with some other fields. This also means the
2152 * filename is stored as in old-style archives.
2155 /* Grab fields common to all tar variants. */
2156 err = header_common(a, tar, entry, h);
2157 if (err == ARCHIVE_FATAL)
2160 /* Copy filename over (to ensure null termination). */
2161 header = (const struct archive_entry_header_gnutar *)h;
2162 if (archive_entry_copy_pathname_l(entry,
2163 header->name, sizeof(header->name), tar->sconv) != 0) {
2164 err = set_conversion_failed_error(a, tar->sconv, "Pathname");
2165 if (err == ARCHIVE_FATAL)
2169 /* Fields common to ustar and GNU */
2170 /* XXX Can the following be factored out since it's common
2171 * to ustar and gnu tar? Is it okay to move it down into
2172 * header_common, perhaps? */
2173 if (archive_entry_copy_uname_l(entry,
2174 header->uname, sizeof(header->uname), tar->sconv) != 0) {
2175 err = set_conversion_failed_error(a, tar->sconv, "Uname");
2176 if (err == ARCHIVE_FATAL)
2180 if (archive_entry_copy_gname_l(entry,
2181 header->gname, sizeof(header->gname), tar->sconv) != 0) {
2182 err = set_conversion_failed_error(a, tar->sconv, "Gname");
2183 if (err == ARCHIVE_FATAL)
2187 /* Parse out device numbers only for char and block specials */
2188 if (header->typeflag[0] == '3' || header->typeflag[0] == '4') {
2189 archive_entry_set_rdevmajor(entry, (dev_t)
2190 tar_atol(header->rdevmajor, sizeof(header->rdevmajor)));
2191 archive_entry_set_rdevminor(entry, (dev_t)
2192 tar_atol(header->rdevminor, sizeof(header->rdevminor)));
2194 archive_entry_set_rdev(entry, 0);
2196 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining);
2198 /* Grab GNU-specific fields. */
2199 t = tar_atol(header->atime, sizeof(header->atime));
2201 archive_entry_set_atime(entry, t, 0);
2202 t = tar_atol(header->ctime, sizeof(header->ctime));
2204 archive_entry_set_ctime(entry, t, 0);
2206 if (header->realsize[0] != 0) {
2208 = tar_atol(header->realsize, sizeof(header->realsize));
2209 archive_entry_set_size(entry, tar->realsize);
2212 if (header->sparse[0].offset[0] != 0) {
2213 if (gnu_sparse_old_read(a, tar, header, unconsumed)
2215 return (ARCHIVE_FATAL);
2217 if (header->isextended[0] != 0) {
2226 gnu_add_sparse_entry(struct archive_read *a, struct tar *tar,
2227 int64_t offset, int64_t remaining)
2229 struct sparse_block *p;
2231 p = (struct sparse_block *)calloc(1, sizeof(*p));
2233 archive_set_error(&a->archive, ENOMEM, "Out of memory");
2234 return (ARCHIVE_FATAL);
2236 if (tar->sparse_last != NULL)
2237 tar->sparse_last->next = p;
2239 tar->sparse_list = p;
2240 tar->sparse_last = p;
2241 if (remaining < 0 || offset < 0) {
2242 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Malformed sparse map data");
2243 return (ARCHIVE_FATAL);
2246 p->remaining = remaining;
2247 return (ARCHIVE_OK);
2251 gnu_clear_sparse_list(struct tar *tar)
2253 struct sparse_block *p;
2255 while (tar->sparse_list != NULL) {
2256 p = tar->sparse_list;
2257 tar->sparse_list = p->next;
2260 tar->sparse_last = NULL;
2264 * GNU tar old-format sparse data.
2266 * GNU old-format sparse data is stored in a fixed-field
2267 * format. Offset/size values are 11-byte octal fields (same
2268 * format as 'size' field in ustart header). These are
2269 * stored in the header, allocating subsequent header blocks
2270 * as needed. Extending the header in this way is a pretty
2271 * severe POSIX violation; this design has earned GNU tar a
2276 gnu_sparse_old_read(struct archive_read *a, struct tar *tar,
2277 const struct archive_entry_header_gnutar *header, size_t *unconsumed)
2282 struct gnu_sparse sparse[21];
2286 const struct extended *ext;
2288 if (gnu_sparse_old_parse(a, tar, header->sparse, 4) != ARCHIVE_OK)
2289 return (ARCHIVE_FATAL);
2290 if (header->isextended[0] == 0)
2291 return (ARCHIVE_OK);
2294 tar_flush_unconsumed(a, unconsumed);
2295 data = __archive_read_ahead(a, 512, &bytes_read);
2297 return (ARCHIVE_FATAL);
2298 if (bytes_read < 512) {
2299 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
2300 "Truncated tar archive "
2301 "detected while reading sparse file data");
2302 return (ARCHIVE_FATAL);
2305 ext = (const struct extended *)data;
2306 if (gnu_sparse_old_parse(a, tar, ext->sparse, 21) != ARCHIVE_OK)
2307 return (ARCHIVE_FATAL);
2308 } while (ext->isextended[0] != 0);
2309 if (tar->sparse_list != NULL)
2310 tar->entry_offset = tar->sparse_list->offset;
2311 return (ARCHIVE_OK);
2315 gnu_sparse_old_parse(struct archive_read *a, struct tar *tar,
2316 const struct gnu_sparse *sparse, int length)
2318 while (length > 0 && sparse->offset[0] != 0) {
2319 if (gnu_add_sparse_entry(a, tar,
2320 tar_atol(sparse->offset, sizeof(sparse->offset)),
2321 tar_atol(sparse->numbytes, sizeof(sparse->numbytes)))
2323 return (ARCHIVE_FATAL);
2327 return (ARCHIVE_OK);
2331 * GNU tar sparse format 0.0
2333 * Beginning with GNU tar 1.15, sparse files are stored using
2334 * information in the pax extended header. The GNU tar maintainers
2335 * have gone through a number of variations in the process of working
2336 * out this scheme; fortunately, they're all numbered.
2338 * Sparse format 0.0 uses attribute GNU.sparse.numblocks to store the
2339 * number of blocks, and GNU.sparse.offset/GNU.sparse.numbytes to
2340 * store offset/size for each block. The repeated instances of these
2341 * latter fields violate the pax specification (which frowns on
2342 * duplicate keys), so this format was quickly replaced.
2346 * GNU tar sparse format 0.1
2348 * This version replaced the offset/numbytes attributes with
2349 * a single "map" attribute that stored a list of integers. This
2350 * format had two problems: First, the "map" attribute could be very
2351 * long, which caused problems for some implementations. More
2352 * importantly, the sparse data was lost when extracted by archivers
2353 * that didn't recognize this extension.
2357 gnu_sparse_01_parse(struct archive_read *a, struct tar *tar, const char *p)
2360 int64_t offset = -1, size = -1;
2364 while (*e != '\0' && *e != ',') {
2365 if (*e < '0' || *e > '9')
2366 return (ARCHIVE_WARN);
2370 offset = tar_atol10(p, e - p);
2372 return (ARCHIVE_WARN);
2374 size = tar_atol10(p, e - p);
2376 return (ARCHIVE_WARN);
2377 if (gnu_add_sparse_entry(a, tar, offset, size)
2379 return (ARCHIVE_FATAL);
2383 return (ARCHIVE_OK);
2389 * GNU tar sparse format 1.0
2391 * The idea: The offset/size data is stored as a series of base-10
2392 * ASCII numbers prepended to the file data, so that dearchivers that
2393 * don't support this format will extract the block map along with the
2394 * data and a separate post-process can restore the sparseness.
2396 * Unfortunately, GNU tar 1.16 had a bug that added unnecessary
2397 * padding to the body of the file when using this format. GNU tar
2398 * 1.17 corrected this bug without bumping the version number, so
2399 * it's not possible to support both variants. This code supports
2400 * the later variant at the expense of not supporting the former.
2402 * This variant also replaced GNU.sparse.size with GNU.sparse.realsize
2403 * and introduced the GNU.sparse.major/GNU.sparse.minor attributes.
2407 * Read the next line from the input, and parse it as a decimal
2408 * integer followed by '\n'. Returns positive integer value or
2409 * negative on error.
2412 gnu_sparse_10_atol(struct archive_read *a, struct tar *tar,
2413 int64_t *remaining, size_t *unconsumed)
2415 int64_t l, limit, last_digit_limit;
2421 limit = INT64_MAX / base;
2422 last_digit_limit = INT64_MAX % base;
2425 * Skip any lines starting with '#'; GNU tar specs
2426 * don't require this, but they should.
2429 bytes_read = readline(a, tar, &p,
2430 (ssize_t)tar_min(*remaining, 100), unconsumed);
2431 if (bytes_read <= 0)
2432 return (ARCHIVE_FATAL);
2433 *remaining -= bytes_read;
2434 } while (p[0] == '#');
2437 while (bytes_read > 0) {
2440 if (*p < '0' || *p >= '0' + base)
2441 return (ARCHIVE_WARN);
2443 if (l > limit || (l == limit && digit > last_digit_limit))
2444 l = INT64_MAX; /* Truncate on overflow. */
2446 l = (l * base) + digit;
2450 /* TODO: Error message. */
2451 return (ARCHIVE_WARN);
2455 * Returns length (in bytes) of the sparse data description
2459 gnu_sparse_10_read(struct archive_read *a, struct tar *tar, size_t *unconsumed)
2463 int64_t offset, size, to_skip, remaining;
2465 /* Clear out the existing sparse list. */
2466 gnu_clear_sparse_list(tar);
2468 remaining = tar->entry_bytes_remaining;
2470 /* Parse entries. */
2471 entries = (int)gnu_sparse_10_atol(a, tar, &remaining, unconsumed);
2473 return (ARCHIVE_FATAL);
2474 /* Parse the individual entries. */
2475 while (entries-- > 0) {
2476 /* Parse offset/size */
2477 offset = gnu_sparse_10_atol(a, tar, &remaining, unconsumed);
2479 return (ARCHIVE_FATAL);
2480 size = gnu_sparse_10_atol(a, tar, &remaining, unconsumed);
2482 return (ARCHIVE_FATAL);
2483 /* Add a new sparse entry. */
2484 if (gnu_add_sparse_entry(a, tar, offset, size) != ARCHIVE_OK)
2485 return (ARCHIVE_FATAL);
2487 /* Skip rest of block... */
2488 tar_flush_unconsumed(a, unconsumed);
2489 bytes_read = (ssize_t)(tar->entry_bytes_remaining - remaining);
2490 to_skip = 0x1ff & -bytes_read;
2491 if (to_skip != __archive_read_consume(a, to_skip))
2492 return (ARCHIVE_FATAL);
2493 return ((ssize_t)(bytes_read + to_skip));
2497 * Solaris pax extension for a sparse file. This is recorded with the
2498 * data and hole pairs. The way recording sparse information by Solaris'
2499 * pax simply indicates where data and sparse are, so the stored contents
2500 * consist of both data and hole.
2503 solaris_sparse_parse(struct archive_read *a, struct tar *tar,
2504 struct archive_entry *entry, const char *p)
2510 (void)entry; /* UNUSED */
2516 return (ARCHIVE_WARN);
2519 while (*e != '\0' && *e != ' ') {
2520 if (*e < '0' || *e > '9')
2521 return (ARCHIVE_WARN);
2525 end = tar_atol10(p, e - p);
2527 return (ARCHIVE_WARN);
2529 if (gnu_add_sparse_entry(a, tar, start,
2530 end - start) != ARCHIVE_OK)
2531 return (ARCHIVE_FATAL);
2532 tar->sparse_last->hole = hole;
2535 return (ARCHIVE_OK);
2542 * Convert text->integer.
2544 * Traditional tar formats (including POSIX) specify base-8 for
2545 * all of the standard numeric fields. This is a significant limitation
2547 * = file size is limited to 8GB
2548 * = rdevmajor and rdevminor are limited to 21 bits
2549 * = uid/gid are limited to 21 bits
2551 * There are two workarounds for this:
2552 * = pax extended headers, which use variable-length string fields
2553 * = GNU tar and STAR both allow either base-8 or base-256 in
2554 * most fields. The high bit is set to indicate base-256.
2556 * On read, this implementation supports both extensions.
2559 tar_atol(const char *p, size_t char_cnt)
2562 * Technically, GNU tar considers a field to be in base-256
2563 * only if the first byte is 0xff or 0x80.
2566 return (tar_atol256(p, char_cnt));
2567 return (tar_atol8(p, char_cnt));
2571 * Note that this implementation does not (and should not!) obey
2572 * locale settings; you cannot simply substitute strtol here, since
2573 * it does obey locale.
2576 tar_atol_base_n(const char *p, size_t char_cnt, int base)
2578 int64_t l, maxval, limit, last_digit_limit;
2582 limit = INT64_MAX / base;
2583 last_digit_limit = INT64_MAX % base;
2585 /* the pointer will not be dereferenced if char_cnt is zero
2586 * due to the way the && operator is evaluated.
2588 while (char_cnt != 0 && (*p == ' ' || *p == '\t')) {
2594 if (char_cnt != 0 && *p == '-') {
2600 limit = -(INT64_MIN / base);
2601 last_digit_limit = INT64_MIN % base;
2605 if (char_cnt != 0) {
2607 while (digit >= 0 && digit < base && char_cnt != 0) {
2608 if (l>limit || (l == limit && digit > last_digit_limit)) {
2609 return maxval; /* Truncate on overflow. */
2611 l = (l * base) + digit;
2616 return (sign < 0) ? -l : l;
2620 tar_atol8(const char *p, size_t char_cnt)
2622 return tar_atol_base_n(p, char_cnt, 8);
2626 tar_atol10(const char *p, size_t char_cnt)
2628 return tar_atol_base_n(p, char_cnt, 10);
2632 * Parse a base-256 integer. This is just a variable-length
2633 * twos-complement signed binary value in big-endian order, except
2634 * that the high-order bit is ignored. The values here can be up to
2635 * 12 bytes, so we need to be careful about overflowing 64-bit
2636 * (8-byte) integers.
2638 * This code unashamedly assumes that the local machine uses 8-bit
2639 * bytes and twos-complement arithmetic.
2642 tar_atol256(const char *_p, size_t char_cnt)
2645 const unsigned char *p = (const unsigned char *)_p;
2646 unsigned char c, neg;
2648 /* Extend 7-bit 2s-comp to 8-bit 2s-comp, decide sign. */
2653 l = ~ARCHIVE_LITERAL_ULL(0);
2660 /* If more than 8 bytes, check that we can ignore
2661 * high-order bits without overflow. */
2662 while (char_cnt > sizeof(int64_t)) {
2665 return neg ? INT64_MIN : INT64_MAX;
2669 /* c is first byte that fits; if sign mismatch, return overflow */
2670 if ((c ^ neg) & 0x80) {
2671 return neg ? INT64_MIN : INT64_MAX;
2674 /* Accumulate remaining bytes. */
2675 while (--char_cnt > 0) {
2680 /* Return signed twos-complement value. */
2681 return (int64_t)(l);
2685 * Returns length of line (including trailing newline)
2686 * or negative on error. 'start' argument is updated to
2687 * point to first character of line. This avoids copying
2691 readline(struct archive_read *a, struct tar *tar, const char **start,
2692 ssize_t limit, size_t *unconsumed)
2695 ssize_t total_size = 0;
2700 tar_flush_unconsumed(a, unconsumed);
2702 t = __archive_read_ahead(a, 1, &bytes_read);
2703 if (bytes_read <= 0)
2704 return (ARCHIVE_FATAL);
2705 s = t; /* Start of line? */
2706 p = memchr(t, '\n', bytes_read);
2707 /* If we found '\n' in the read buffer, return pointer to that. */
2709 bytes_read = 1 + ((const char *)p) - s;
2710 if (bytes_read > limit) {
2711 archive_set_error(&a->archive,
2712 ARCHIVE_ERRNO_FILE_FORMAT,
2714 return (ARCHIVE_FATAL);
2716 *unconsumed = bytes_read;
2718 return (bytes_read);
2720 *unconsumed = bytes_read;
2721 /* Otherwise, we need to accumulate in a line buffer. */
2723 if (total_size + bytes_read > limit) {
2724 archive_set_error(&a->archive,
2725 ARCHIVE_ERRNO_FILE_FORMAT,
2727 return (ARCHIVE_FATAL);
2729 if (archive_string_ensure(&tar->line, total_size + bytes_read) == NULL) {
2730 archive_set_error(&a->archive, ENOMEM,
2731 "Can't allocate working buffer");
2732 return (ARCHIVE_FATAL);
2734 memcpy(tar->line.s + total_size, t, bytes_read);
2735 tar_flush_unconsumed(a, unconsumed);
2736 total_size += bytes_read;
2737 /* If we found '\n', clean up and return. */
2739 *start = tar->line.s;
2740 return (total_size);
2742 /* Read some more. */
2743 t = __archive_read_ahead(a, 1, &bytes_read);
2744 if (bytes_read <= 0)
2745 return (ARCHIVE_FATAL);
2746 s = t; /* Start of line? */
2747 p = memchr(t, '\n', bytes_read);
2748 /* If we found '\n', trim the read. */
2750 bytes_read = 1 + ((const char *)p) - s;
2752 *unconsumed = bytes_read;
2757 * base64_decode - Base64 decode
2759 * This accepts most variations of base-64 encoding, including:
2760 * * with or without line breaks
2761 * * with or without the final group padded with '=' or '_' characters
2762 * (The most economical Base-64 variant does not pad the last group and
2763 * omits line breaks; RFC1341 used for MIME requires both.)
2766 base64_decode(const char *s, size_t len, size_t *out_len)
2768 static const unsigned char digits[64] = {
2769 'A','B','C','D','E','F','G','H','I','J','K','L','M','N',
2770 'O','P','Q','R','S','T','U','V','W','X','Y','Z','a','b',
2771 'c','d','e','f','g','h','i','j','k','l','m','n','o','p',
2772 'q','r','s','t','u','v','w','x','y','z','0','1','2','3',
2773 '4','5','6','7','8','9','+','/' };
2774 static unsigned char decode_table[128];
2776 const unsigned char *src = (const unsigned char *)s;
2778 /* If the decode table is not yet initialized, prepare it. */
2779 if (decode_table[digits[1]] != 1) {
2781 memset(decode_table, 0xff, sizeof(decode_table));
2782 for (i = 0; i < sizeof(digits); i++)
2783 decode_table[digits[i]] = i;
2786 /* Allocate enough space to hold the entire output. */
2787 /* Note that we may not use all of this... */
2788 out = (char *)malloc(len - len / 4 + 1);
2796 /* Collect the next group of (up to) four characters. */
2799 while (group_size < 4 && len > 0) {
2800 /* '=' or '_' padding indicates final group. */
2801 if (*src == '=' || *src == '_') {
2805 /* Skip illegal characters (including line breaks) */
2806 if (*src > 127 || *src < 32
2807 || decode_table[*src] == 0xff) {
2813 v |= decode_table[*src++];
2817 /* Align a short group properly. */
2818 v <<= 6 * (4 - group_size);
2819 /* Unpack the group we just collected. */
2820 switch (group_size) {
2821 case 4: d[2] = v & 0xff;
2823 case 3: d[1] = (v >> 8) & 0xff;
2825 case 2: d[0] = (v >> 16) & 0xff;
2827 case 1: /* this is invalid! */
2830 d += group_size * 3 / 4;
2838 url_decode(const char *in)
2843 out = (char *)malloc(strlen(in) + 1);
2846 for (s = in, d = out; *s != '\0'; ) {
2847 if (s[0] == '%' && s[1] != '\0' && s[2] != '\0') {
2848 /* Try to convert % escape */
2849 int digit1 = tohex(s[1]);
2850 int digit2 = tohex(s[2]);
2851 if (digit1 >= 0 && digit2 >= 0) {
2852 /* Looks good, consume three chars */
2854 /* Convert output */
2855 *d++ = ((digit1 << 4) | digit2);
2858 /* Else fall through and treat '%' as normal char */
2869 if (c >= '0' && c <= '9')
2871 else if (c >= 'A' && c <= 'F')
2872 return (c - 'A' + 10);
2873 else if (c >= 'a' && c <= 'f')
2874 return (c - 'a' + 10);