1 /*-
2 * Copyright (c) 2003-2007 Tim Kientzle
3 * Copyright (c) 2011-2012 Michihiro NAKAJIMA
4 * Copyright (c) 2016 Martin Matuska
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 *
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR
17 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
18 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
19 * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT,
20 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
21 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
22 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
23 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
24 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
25 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
26 */
27
28 #include "archive_platform.h"
29
30 #ifdef HAVE_ERRNO_H
31 #include <errno.h>
32 #endif
33 #include <stddef.h>
34 #ifdef HAVE_STDLIB_H
35 #include <stdlib.h>
36 #endif
37 #ifdef HAVE_STRING_H
38 #include <string.h>
39 #endif
40
41 #include "archive.h"
42 #include "archive_acl_private.h" /* For ACL parsing routines. */
43 #include "archive_entry.h"
44 #include "archive_entry_locale.h"
45 #include "archive_private.h"
46 #include "archive_read_private.h"
47
48 #define tar_min(a,b) ((a) < (b) ? (a) : (b))
49
50 /*
51 * Layout of POSIX 'ustar' tar header.
52 */
53 struct archive_entry_header_ustar {
54 char name[100];
55 char mode[8];
56 char uid[8];
57 char gid[8];
58 char size[12];
59 char mtime[12];
60 char checksum[8];
61 char typeflag[1];
62 char linkname[100]; /* "old format" header ends here */
63 char magic[6]; /* For POSIX: "ustar\0" */
64 char version[2]; /* For POSIX: "00" */
65 char uname[32];
66 char gname[32];
67 char rdevmajor[8];
68 char rdevminor[8];
69 char prefix[155];
70 };
71
72 /*
73 * Structure of GNU tar header
74 */
75 struct gnu_sparse {
76 char offset[12];
77 char numbytes[12];
78 };
79
80 struct archive_entry_header_gnutar {
81 char name[100];
82 char mode[8];
83 char uid[8];
84 char gid[8];
85 char size[12];
86 char mtime[12];
87 char checksum[8];
88 char typeflag[1];
89 char linkname[100];
90 char magic[8]; /* "ustar \0" (note blank/blank/null at end) */
91 char uname[32];
92 char gname[32];
93 char rdevmajor[8];
94 char rdevminor[8];
95 char atime[12];
96 char ctime[12];
97 char offset[12];
98 char longnames[4];
99 char unused[1];
100 struct gnu_sparse sparse[4];
101 char isextended[1];
102 char realsize[12];
103 /*
104 * Old GNU format doesn't use POSIX 'prefix' field; they use
105 * the 'L' (longname) entry instead.
106 */
107 };
108
109 /*
110 * Data specific to this format.
111 */
112 struct sparse_block {
113 struct sparse_block *next;
114 int64_t offset;
115 int64_t remaining;
116 int hole;
117 };
118
119 struct tar {
120 struct archive_string acl_text;
121 struct archive_string entry_pathname;
122 /* For "GNU.sparse.name" and other similar path extensions. */
123 struct archive_string entry_pathname_override;
124 struct archive_string entry_linkpath;
125 struct archive_string entry_uname;
126 struct archive_string entry_gname;
127 struct archive_string longlink;
128 struct archive_string longname;
129 struct archive_string pax_header;
130 struct archive_string pax_global;
131 struct archive_string line;
132 int pax_hdrcharset_binary;
133 int header_recursion_depth;
134 int64_t entry_bytes_remaining;
135 int64_t entry_offset;
136 int64_t entry_padding;
137 int64_t entry_bytes_unconsumed;
138 int64_t realsize;
139 int sparse_allowed;
140 struct sparse_block *sparse_list;
141 struct sparse_block *sparse_last;
142 int64_t sparse_offset;
143 int64_t sparse_numbytes;
144 int sparse_gnu_major;
145 int sparse_gnu_minor;
146 char sparse_gnu_pending;
147
148 struct archive_string localname;
149 struct archive_string_conv *opt_sconv;
150 struct archive_string_conv *sconv;
151 struct archive_string_conv *sconv_acl;
152 struct archive_string_conv *sconv_default;
153 int init_default_conversion;
154 int compat_2x;
155 int process_mac_extensions;
156 int read_concatenated_archives;
157 int realsize_override;
158 };
159
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);
164
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 *,
171 const char *);
172 static ssize_t gnu_sparse_10_read(struct archive_read *, struct tar *,
173 size_t *);
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 *,
212 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 *);
231
232
233 int
archive_read_support_format_gnutar(struct archive * a)234 archive_read_support_format_gnutar(struct archive *a)
235 {
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));
239 }
240
241
242 int
archive_read_support_format_tar(struct archive * _a)243 archive_read_support_format_tar(struct archive *_a)
244 {
245 struct archive_read *a = (struct archive_read *)_a;
246 struct tar *tar;
247 int r;
248
249 archive_check_magic(_a, ARCHIVE_READ_MAGIC,
250 ARCHIVE_STATE_NEW, "archive_read_support_format_tar");
251
252 tar = (struct tar *)calloc(1, sizeof(*tar));
253 if (tar == NULL) {
254 archive_set_error(&a->archive, ENOMEM,
255 "Can't allocate tar data");
256 return (ARCHIVE_FATAL);
257 }
258 #ifdef HAVE_COPYFILE_H
259 /* Set this by default on Mac OS. */
260 tar->process_mac_extensions = 1;
261 #endif
262
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,
269 NULL,
270 archive_read_format_tar_cleanup,
271 NULL,
272 NULL);
273
274 if (r != ARCHIVE_OK)
275 free(tar);
276 return (ARCHIVE_OK);
277 }
278
279 static int
archive_read_format_tar_cleanup(struct archive_read * a)280 archive_read_format_tar_cleanup(struct archive_read *a)
281 {
282 struct tar *tar;
283
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);
298 free(tar);
299 (a->format->data) = NULL;
300 return (ARCHIVE_OK);
301 }
302
303 /*
304 * Validate number field
305 *
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
315 *
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
322 *
323 * This should tolerate all variants in use. It will reject a field
324 * where the writer just left garbage after a trailing NUL.
325 */
326 static int
validate_number_field(const char * p_field,size_t i_size)327 validate_number_field(const char* p_field, size_t i_size)
328 {
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. */
332 return 1;
333 } else {
334 /* Must be octal */
335 size_t i = 0;
336 /* Skip any leading spaces */
337 while (i < i_size && p_field[i] == ' ') {
338 ++i;
339 }
340 /* Skip octal digits. */
341 while (i < i_size && p_field[i] >= '0' && p_field[i] <= '7') {
342 ++i;
343 }
344 /* Any remaining characters must be space or NUL padding. */
345 while (i < i_size) {
346 if (p_field[i] != ' ' && p_field[i] != 0) {
347 return 0;
348 }
349 ++i;
350 }
351 return 1;
352 }
353 }
354
355 static int
archive_read_format_tar_bid(struct archive_read * a,int best_bid)356 archive_read_format_tar_bid(struct archive_read *a, int best_bid)
357 {
358 int bid;
359 const char *h;
360 const struct archive_entry_header_ustar *header;
361
362 (void)best_bid; /* UNUSED */
363
364 bid = 0;
365
366 /* Now let's look at the actual header and see if it matches. */
367 h = __archive_read_ahead(a, 512, NULL);
368 if (h == NULL)
369 return (-1);
370
371 /* If it's an end-of-archive mark, we can handle it. */
372 if (h[0] == 0 && archive_block_is_null(h)) {
373 /*
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.
377 */
378 return (10);
379 }
380
381 /* If it's not an end-of-archive mark, it must have a valid checksum.*/
382 if (!checksum(a, h))
383 return (0);
384 bid += 48; /* Checksum is usually 6 octal digits. */
385
386 header = (const struct archive_entry_header_ustar *)h;
387
388 /* Recognize POSIX formats. */
389 if ((memcmp(header->magic, "ustar\0", 6) == 0)
390 && (memcmp(header->version, "00", 2) == 0))
391 bid += 56;
392
393 /* Recognize GNU tar format. */
394 if ((memcmp(header->magic, "ustar ", 6) == 0)
395 && (memcmp(header->version, " \0", 2) == 0))
396 bid += 56;
397
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') )
403 return (0);
404 bid += 2; /* 6 bits of variation in an 8-bit field leaves 2 bits. */
405
406 /*
407 * Check format of mode/uid/gid/mtime/size/rdevmajor/rdevminor fields.
408 */
409 if (validate_number_field(header->mode, sizeof(header->mode)) == 0
410 || validate_number_field(header->uid, sizeof(header->uid)) == 0
411 || validate_number_field(header->gid, sizeof(header->gid)) == 0
412 || validate_number_field(header->mtime, sizeof(header->mtime)) == 0
413 || validate_number_field(header->size, sizeof(header->size)) == 0
414 || validate_number_field(header->rdevmajor, sizeof(header->rdevmajor)) == 0
415 || validate_number_field(header->rdevminor, sizeof(header->rdevminor)) == 0) {
416 bid = 0;
417 }
418
419 return (bid);
420 }
421
422 static int
archive_read_format_tar_options(struct archive_read * a,const char * key,const char * val)423 archive_read_format_tar_options(struct archive_read *a,
424 const char *key, const char *val)
425 {
426 struct tar *tar;
427 int ret = ARCHIVE_FAILED;
428
429 tar = (struct tar *)(a->format->data);
430 if (strcmp(key, "compat-2x") == 0) {
431 /* Handle UTF-8 filenames as libarchive 2.x */
432 tar->compat_2x = (val != NULL && val[0] != 0);
433 tar->init_default_conversion = tar->compat_2x;
434 return (ARCHIVE_OK);
435 } else if (strcmp(key, "hdrcharset") == 0) {
436 if (val == NULL || val[0] == 0)
437 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
438 "tar: hdrcharset option needs a character-set name");
439 else {
440 tar->opt_sconv =
441 archive_string_conversion_from_charset(
442 &a->archive, val, 0);
443 if (tar->opt_sconv != NULL)
444 ret = ARCHIVE_OK;
445 else
446 ret = ARCHIVE_FATAL;
447 }
448 return (ret);
449 } else if (strcmp(key, "mac-ext") == 0) {
450 tar->process_mac_extensions = (val != NULL && val[0] != 0);
451 return (ARCHIVE_OK);
452 } else if (strcmp(key, "read_concatenated_archives") == 0) {
453 tar->read_concatenated_archives = (val != NULL && val[0] != 0);
454 return (ARCHIVE_OK);
455 }
456
457 /* Note: The "warn" return is just to inform the options
458 * supervisor that we didn't handle it. It will generate
459 * a suitable error if no one used this option. */
460 return (ARCHIVE_WARN);
461 }
462
463 /* utility function- this exists to centralize the logic of tracking
464 * how much unconsumed data we have floating around, and to consume
465 * anything outstanding since we're going to do read_aheads
466 */
467 static void
tar_flush_unconsumed(struct archive_read * a,size_t * unconsumed)468 tar_flush_unconsumed(struct archive_read *a, size_t *unconsumed)
469 {
470 if (*unconsumed) {
471 /*
472 void *data = (void *)__archive_read_ahead(a, *unconsumed, NULL);
473 * this block of code is to poison claimed unconsumed space, ensuring
474 * things break if it is in use still.
475 * currently it WILL break things, so enable it only for debugging this issue
476 if (data) {
477 memset(data, 0xff, *unconsumed);
478 }
479 */
480 __archive_read_consume(a, *unconsumed);
481 *unconsumed = 0;
482 }
483 }
484
485 /*
486 * The function invoked by archive_read_next_header(). This
487 * just sets up a few things and then calls the internal
488 * tar_read_header() function below.
489 */
490 static int
archive_read_format_tar_read_header(struct archive_read * a,struct archive_entry * entry)491 archive_read_format_tar_read_header(struct archive_read *a,
492 struct archive_entry *entry)
493 {
494 /*
495 * When converting tar archives to cpio archives, it is
496 * essential that each distinct file have a distinct inode
497 * number. To simplify this, we keep a static count here to
498 * assign fake dev/inode numbers to each tar entry. Note that
499 * pax format archives may overwrite this with something more
500 * useful.
501 *
502 * Ideally, we would track every file read from the archive so
503 * that we could assign the same dev/ino pair to hardlinks,
504 * but the memory required to store a complete lookup table is
505 * probably not worthwhile just to support the relatively
506 * obscure tar->cpio conversion case.
507 */
508 static int default_inode;
509 static int default_dev;
510 struct tar *tar;
511 const char *p;
512 const wchar_t *wp;
513 int r;
514 size_t l, unconsumed = 0;
515
516 /* Assign default device/inode values. */
517 archive_entry_set_dev(entry, 1 + default_dev); /* Don't use zero. */
518 archive_entry_set_ino(entry, ++default_inode); /* Don't use zero. */
519 /* Limit generated st_ino number to 16 bits. */
520 if (default_inode >= 0xffff) {
521 ++default_dev;
522 default_inode = 0;
523 }
524
525 tar = (struct tar *)(a->format->data);
526 tar->entry_offset = 0;
527 gnu_clear_sparse_list(tar);
528 tar->realsize = -1; /* Mark this as "unset" */
529 tar->realsize_override = 0;
530
531 /* Setup default string conversion. */
532 tar->sconv = tar->opt_sconv;
533 if (tar->sconv == NULL) {
534 if (!tar->init_default_conversion) {
535 tar->sconv_default =
536 archive_string_default_conversion_for_read(&(a->archive));
537 tar->init_default_conversion = 1;
538 }
539 tar->sconv = tar->sconv_default;
540 }
541
542 r = tar_read_header(a, tar, entry, &unconsumed);
543
544 tar_flush_unconsumed(a, &unconsumed);
545
546 /*
547 * "non-sparse" files are really just sparse files with
548 * a single block.
549 */
550 if (tar->sparse_list == NULL) {
551 if (gnu_add_sparse_entry(a, tar, 0, tar->entry_bytes_remaining)
552 != ARCHIVE_OK)
553 return (ARCHIVE_FATAL);
554 } else {
555 struct sparse_block *sb;
556
557 for (sb = tar->sparse_list; sb != NULL; sb = sb->next) {
558 if (!sb->hole)
559 archive_entry_sparse_add_entry(entry,
560 sb->offset, sb->remaining);
561 }
562 }
563
564 if (r == ARCHIVE_OK && archive_entry_filetype(entry) == AE_IFREG) {
565 /*
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.
569 */
570 if ((wp = archive_entry_pathname_w(entry)) != NULL) {
571 l = wcslen(wp);
572 if (l > 0 && wp[l - 1] == L'/') {
573 archive_entry_set_filetype(entry, AE_IFDIR);
574 tar->entry_bytes_remaining = 0;
575 tar->entry_padding = 0;
576 }
577 } else if ((p = archive_entry_pathname(entry)) != NULL) {
578 l = strlen(p);
579 if (l > 0 && p[l - 1] == '/') {
580 archive_entry_set_filetype(entry, AE_IFDIR);
581 tar->entry_bytes_remaining = 0;
582 tar->entry_padding = 0;
583 }
584 }
585 }
586 return (r);
587 }
588
589 static int
archive_read_format_tar_read_data(struct archive_read * a,const void ** buff,size_t * size,int64_t * offset)590 archive_read_format_tar_read_data(struct archive_read *a,
591 const void **buff, size_t *size, int64_t *offset)
592 {
593 ssize_t bytes_read;
594 struct tar *tar;
595 struct sparse_block *p;
596
597 tar = (struct tar *)(a->format->data);
598
599 for (;;) {
600 /* Remove exhausted entries from sparse list. */
601 while (tar->sparse_list != NULL &&
602 tar->sparse_list->remaining == 0) {
603 p = tar->sparse_list;
604 tar->sparse_list = p->next;
605 free(p);
606 }
607
608 if (tar->entry_bytes_unconsumed) {
609 __archive_read_consume(a, tar->entry_bytes_unconsumed);
610 tar->entry_bytes_unconsumed = 0;
611 }
612
613 /* If we're at end of file, return EOF. */
614 if (tar->sparse_list == NULL ||
615 tar->entry_bytes_remaining == 0) {
616 if (__archive_read_consume(a, tar->entry_padding) < 0)
617 return (ARCHIVE_FATAL);
618 tar->entry_padding = 0;
619 *buff = NULL;
620 *size = 0;
621 *offset = tar->realsize;
622 return (ARCHIVE_EOF);
623 }
624
625 *buff = __archive_read_ahead(a, 1, &bytes_read);
626 if (bytes_read < 0)
627 return (ARCHIVE_FATAL);
628 if (*buff == NULL) {
629 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
630 "Truncated tar archive");
631 return (ARCHIVE_FATAL);
632 }
633 if (bytes_read > tar->entry_bytes_remaining)
634 bytes_read = (ssize_t)tar->entry_bytes_remaining;
635 /* Don't read more than is available in the
636 * current sparse block. */
637 if (tar->sparse_list->remaining < bytes_read)
638 bytes_read = (ssize_t)tar->sparse_list->remaining;
639 *size = bytes_read;
640 *offset = tar->sparse_list->offset;
641 tar->sparse_list->remaining -= bytes_read;
642 tar->sparse_list->offset += bytes_read;
643 tar->entry_bytes_remaining -= bytes_read;
644 tar->entry_bytes_unconsumed = bytes_read;
645
646 if (!tar->sparse_list->hole)
647 return (ARCHIVE_OK);
648 /* Current is hole data and skip this. */
649 }
650 }
651
652 static int
archive_read_format_tar_skip(struct archive_read * a)653 archive_read_format_tar_skip(struct archive_read *a)
654 {
655 int64_t bytes_skipped;
656 int64_t request;
657 struct sparse_block *p;
658 struct tar* tar;
659
660 tar = (struct tar *)(a->format->data);
661
662 /* Do not consume the hole of a sparse file. */
663 request = 0;
664 for (p = tar->sparse_list; p != NULL; p = p->next) {
665 if (!p->hole) {
666 if (p->remaining >= INT64_MAX - request) {
667 return ARCHIVE_FATAL;
668 }
669 request += p->remaining;
670 }
671 }
672 if (request > tar->entry_bytes_remaining)
673 request = tar->entry_bytes_remaining;
674 request += tar->entry_padding + tar->entry_bytes_unconsumed;
675
676 bytes_skipped = __archive_read_consume(a, request);
677 if (bytes_skipped < 0)
678 return (ARCHIVE_FATAL);
679
680 tar->entry_bytes_remaining = 0;
681 tar->entry_bytes_unconsumed = 0;
682 tar->entry_padding = 0;
683
684 /* Free the sparse list. */
685 gnu_clear_sparse_list(tar);
686
687 return (ARCHIVE_OK);
688 }
689
690 /*
691 * This function recursively interprets all of the headers associated
692 * with a single entry.
693 */
694 static int
tar_read_header(struct archive_read * a,struct tar * tar,struct archive_entry * entry,size_t * unconsumed)695 tar_read_header(struct archive_read *a, struct tar *tar,
696 struct archive_entry *entry, size_t *unconsumed)
697 {
698 ssize_t bytes;
699 int err, eof_vol_header;
700 const char *h;
701 const struct archive_entry_header_ustar *header;
702 const struct archive_entry_header_gnutar *gnuheader;
703
704 eof_vol_header = 0;
705
706 /* Loop until we find a workable header record. */
707 for (;;) {
708 tar_flush_unconsumed(a, unconsumed);
709
710 /* Read 512-byte header record */
711 h = __archive_read_ahead(a, 512, &bytes);
712 if (bytes < 0)
713 return ((int)bytes);
714 if (bytes == 0) { /* EOF at a block boundary. */
715 /* Some writers do omit the block of nulls. <sigh> */
716 return (ARCHIVE_EOF);
717 }
718 if (bytes < 512) { /* Short block at EOF; this is bad. */
719 archive_set_error(&a->archive,
720 ARCHIVE_ERRNO_FILE_FORMAT,
721 "Truncated tar archive");
722 return (ARCHIVE_FATAL);
723 }
724 *unconsumed = 512;
725
726 /* Header is workable if it's not an end-of-archive mark. */
727 if (h[0] != 0 || !archive_block_is_null(h))
728 break;
729
730 /* Ensure format is set for archives with only null blocks. */
731 if (a->archive.archive_format_name == NULL) {
732 a->archive.archive_format = ARCHIVE_FORMAT_TAR;
733 a->archive.archive_format_name = "tar";
734 }
735
736 if (!tar->read_concatenated_archives) {
737 /* Try to consume a second all-null record, as well. */
738 tar_flush_unconsumed(a, unconsumed);
739 h = __archive_read_ahead(a, 512, NULL);
740 if (h != NULL && h[0] == 0 && archive_block_is_null(h))
741 __archive_read_consume(a, 512);
742 archive_clear_error(&a->archive);
743 return (ARCHIVE_EOF);
744 }
745
746 /*
747 * We're reading concatenated archives, ignore this block and
748 * loop to get the next.
749 */
750 }
751
752 /*
753 * Note: If the checksum fails and we return ARCHIVE_RETRY,
754 * then the client is likely to just retry. This is a very
755 * crude way to search for the next valid header!
756 *
757 * TODO: Improve this by implementing a real header scan.
758 */
759 if (!checksum(a, h)) {
760 tar_flush_unconsumed(a, unconsumed);
761 archive_set_error(&a->archive, EINVAL, "Damaged tar archive");
762 return (ARCHIVE_RETRY); /* Retryable: Invalid header */
763 }
764
765 if (++tar->header_recursion_depth > 32) {
766 tar_flush_unconsumed(a, unconsumed);
767 archive_set_error(&a->archive, EINVAL, "Too many special headers");
768 return (ARCHIVE_WARN);
769 }
770
771 /* Determine the format variant. */
772 header = (const struct archive_entry_header_ustar *)h;
773
774 switch(header->typeflag[0]) {
775 case 'A': /* Solaris tar ACL */
776 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE;
777 a->archive.archive_format_name = "Solaris tar";
778 err = header_Solaris_ACL(a, tar, entry, h, unconsumed);
779 break;
780 case 'g': /* POSIX-standard 'g' header. */
781 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE;
782 a->archive.archive_format_name = "POSIX pax interchange format";
783 err = header_pax_global(a, tar, entry, h, unconsumed);
784 if (err == ARCHIVE_EOF)
785 return (err);
786 break;
787 case 'K': /* Long link name (GNU tar, others) */
788 err = header_longlink(a, tar, entry, h, unconsumed);
789 break;
790 case 'L': /* Long filename (GNU tar, others) */
791 err = header_longname(a, tar, entry, h, unconsumed);
792 break;
793 case 'V': /* GNU volume header */
794 err = header_volume(a, tar, entry, h, unconsumed);
795 if (err == ARCHIVE_EOF)
796 eof_vol_header = 1;
797 break;
798 case 'X': /* Used by SUN tar; same as 'x'. */
799 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE;
800 a->archive.archive_format_name =
801 "POSIX pax interchange format (Sun variant)";
802 err = header_pax_extensions(a, tar, entry, h, unconsumed);
803 break;
804 case 'x': /* POSIX-standard 'x' header. */
805 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE;
806 a->archive.archive_format_name = "POSIX pax interchange format";
807 err = header_pax_extensions(a, tar, entry, h, unconsumed);
808 break;
809 default:
810 gnuheader = (const struct archive_entry_header_gnutar *)h;
811 if (memcmp(gnuheader->magic, "ustar \0", 8) == 0) {
812 a->archive.archive_format = ARCHIVE_FORMAT_TAR_GNUTAR;
813 a->archive.archive_format_name = "GNU tar format";
814 err = header_gnutar(a, tar, entry, h, unconsumed);
815 } else if (memcmp(header->magic, "ustar", 5) == 0) {
816 if (a->archive.archive_format != ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE) {
817 a->archive.archive_format = ARCHIVE_FORMAT_TAR_USTAR;
818 a->archive.archive_format_name = "POSIX ustar format";
819 }
820 err = header_ustar(a, tar, entry, h);
821 } else {
822 a->archive.archive_format = ARCHIVE_FORMAT_TAR;
823 a->archive.archive_format_name = "tar (non-POSIX)";
824 err = header_old_tar(a, tar, entry, h);
825 }
826 }
827 if (err == ARCHIVE_FATAL)
828 return (err);
829
830 tar_flush_unconsumed(a, unconsumed);
831
832 h = NULL;
833 header = NULL;
834
835 --tar->header_recursion_depth;
836 /* Yuck. Apple's design here ends up storing long pathname
837 * extensions for both the AppleDouble extension entry and the
838 * regular entry.
839 */
840 if ((err == ARCHIVE_WARN || err == ARCHIVE_OK) &&
841 tar->header_recursion_depth == 0 &&
842 tar->process_mac_extensions) {
843 int err2 = read_mac_metadata_blob(a, tar, entry, h, unconsumed);
844 if (err2 < err)
845 err = err2;
846 }
847
848 /* We return warnings or success as-is. Anything else is fatal. */
849 if (err == ARCHIVE_WARN || err == ARCHIVE_OK) {
850 if (tar->sparse_gnu_pending) {
851 if (tar->sparse_gnu_major == 1 &&
852 tar->sparse_gnu_minor == 0) {
853 ssize_t bytes_read;
854
855 tar->sparse_gnu_pending = 0;
856 /* Read initial sparse map. */
857 bytes_read = gnu_sparse_10_read(a, tar, unconsumed);
858 if (bytes_read < 0)
859 return ((int)bytes_read);
860 tar->entry_bytes_remaining -= bytes_read;
861 } else {
862 archive_set_error(&a->archive,
863 ARCHIVE_ERRNO_MISC,
864 "Unrecognized GNU sparse file format");
865 return (ARCHIVE_WARN);
866 }
867 tar->sparse_gnu_pending = 0;
868 }
869 return (err);
870 }
871 if (err == ARCHIVE_EOF) {
872 if (!eof_vol_header) {
873 /* EOF when recursively reading a header is bad. */
874 archive_set_error(&a->archive, EINVAL,
875 "Damaged tar archive");
876 } else {
877 /* If we encounter just a GNU volume header treat
878 * this situation as an empty archive */
879 return (ARCHIVE_EOF);
880 }
881 }
882 return (ARCHIVE_FATAL);
883 }
884
885 /*
886 * Return true if block checksum is correct.
887 */
888 static int
checksum(struct archive_read * a,const void * h)889 checksum(struct archive_read *a, const void *h)
890 {
891 const unsigned char *bytes;
892 const struct archive_entry_header_ustar *header;
893 int check, sum;
894 size_t i;
895
896 (void)a; /* UNUSED */
897 bytes = (const unsigned char *)h;
898 header = (const struct archive_entry_header_ustar *)h;
899
900 /* Checksum field must hold an octal number */
901 for (i = 0; i < sizeof(header->checksum); ++i) {
902 char c = header->checksum[i];
903 if (c != ' ' && c != '\0' && (c < '0' || c > '7'))
904 return 0;
905 }
906
907 /*
908 * Test the checksum. Note that POSIX specifies _unsigned_
909 * bytes for this calculation.
910 */
911 sum = (int)tar_atol(header->checksum, sizeof(header->checksum));
912 check = 0;
913 for (i = 0; i < 148; i++)
914 check += (unsigned char)bytes[i];
915 for (; i < 156; i++)
916 check += 32;
917 for (; i < 512; i++)
918 check += (unsigned char)bytes[i];
919 if (sum == check)
920 return (1);
921
922 /*
923 * Repeat test with _signed_ bytes, just in case this archive
924 * was created by an old BSD, Solaris, or HP-UX tar with a
925 * broken checksum calculation.
926 */
927 check = 0;
928 for (i = 0; i < 148; i++)
929 check += (signed char)bytes[i];
930 for (; i < 156; i++)
931 check += 32;
932 for (; i < 512; i++)
933 check += (signed char)bytes[i];
934 if (sum == check)
935 return (1);
936
937 return (0);
938 }
939
940 /*
941 * Return true if this block contains only nulls.
942 */
943 static int
archive_block_is_null(const char * p)944 archive_block_is_null(const char *p)
945 {
946 unsigned i;
947
948 for (i = 0; i < 512; i++)
949 if (*p++)
950 return (0);
951 return (1);
952 }
953
954 /*
955 * Interpret 'A' Solaris ACL header
956 */
957 static int
header_Solaris_ACL(struct archive_read * a,struct tar * tar,struct archive_entry * entry,const void * h,size_t * unconsumed)958 header_Solaris_ACL(struct archive_read *a, struct tar *tar,
959 struct archive_entry *entry, const void *h, size_t *unconsumed)
960 {
961 const struct archive_entry_header_ustar *header;
962 size_t size;
963 int err, acl_type;
964 int64_t type;
965 char *acl, *p;
966
967 /*
968 * read_body_to_string adds a NUL terminator, but we need a little
969 * more to make sure that we don't overrun acl_text later.
970 */
971 header = (const struct archive_entry_header_ustar *)h;
972 size = (size_t)tar_atol(header->size, sizeof(header->size));
973 err = read_body_to_string(a, tar, &(tar->acl_text), h, unconsumed);
974 if (err != ARCHIVE_OK)
975 return (err);
976
977 /* Recursively read next header */
978 err = tar_read_header(a, tar, entry, unconsumed);
979 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN))
980 return (err);
981
982 /* TODO: Examine the first characters to see if this
983 * is an AIX ACL descriptor. We'll likely never support
984 * them, but it would be polite to recognize and warn when
985 * we do see them. */
986
987 /* Leading octal number indicates ACL type and number of entries. */
988 p = acl = tar->acl_text.s;
989 type = 0;
990 while (*p != '\0' && p < acl + size) {
991 if (*p < '0' || *p > '7') {
992 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
993 "Malformed Solaris ACL attribute (invalid digit)");
994 return(ARCHIVE_WARN);
995 }
996 type <<= 3;
997 type += *p - '0';
998 if (type > 077777777) {
999 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1000 "Malformed Solaris ACL attribute (count too large)");
1001 return (ARCHIVE_WARN);
1002 }
1003 p++;
1004 }
1005 switch ((int)type & ~0777777) {
1006 case 01000000:
1007 /* POSIX.1e ACL */
1008 acl_type = ARCHIVE_ENTRY_ACL_TYPE_ACCESS;
1009 break;
1010 case 03000000:
1011 /* NFSv4 ACL */
1012 acl_type = ARCHIVE_ENTRY_ACL_TYPE_NFS4;
1013 break;
1014 default:
1015 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1016 "Malformed Solaris ACL attribute (unsupported type %o)",
1017 (int)type);
1018 return (ARCHIVE_WARN);
1019 }
1020 p++;
1021
1022 if (p >= acl + size) {
1023 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1024 "Malformed Solaris ACL attribute (body overflow)");
1025 return(ARCHIVE_WARN);
1026 }
1027
1028 /* ACL text is null-terminated; find the end. */
1029 size -= (p - acl);
1030 acl = p;
1031
1032 while (*p != '\0' && p < acl + size)
1033 p++;
1034
1035 if (tar->sconv_acl == NULL) {
1036 tar->sconv_acl = archive_string_conversion_from_charset(
1037 &(a->archive), "UTF-8", 1);
1038 if (tar->sconv_acl == NULL)
1039 return (ARCHIVE_FATAL);
1040 }
1041 archive_strncpy(&(tar->localname), acl, p - acl);
1042 err = archive_acl_from_text_l(archive_entry_acl(entry),
1043 tar->localname.s, acl_type, tar->sconv_acl);
1044 if (err != ARCHIVE_OK) {
1045 if (errno == ENOMEM) {
1046 archive_set_error(&a->archive, ENOMEM,
1047 "Can't allocate memory for ACL");
1048 } else
1049 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1050 "Malformed Solaris ACL attribute (unparsable)");
1051 }
1052 return (err);
1053 }
1054
1055 /*
1056 * Interpret 'K' long linkname header.
1057 */
1058 static int
header_longlink(struct archive_read * a,struct tar * tar,struct archive_entry * entry,const void * h,size_t * unconsumed)1059 header_longlink(struct archive_read *a, struct tar *tar,
1060 struct archive_entry *entry, const void *h, size_t *unconsumed)
1061 {
1062 int err;
1063
1064 err = read_body_to_string(a, tar, &(tar->longlink), h, unconsumed);
1065 if (err != ARCHIVE_OK)
1066 return (err);
1067 err = tar_read_header(a, tar, entry, unconsumed);
1068 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN))
1069 return (err);
1070 /* Set symlink if symlink already set, else hardlink. */
1071 archive_entry_copy_link(entry, tar->longlink.s);
1072 return (ARCHIVE_OK);
1073 }
1074
1075 static int
set_conversion_failed_error(struct archive_read * a,struct archive_string_conv * sconv,const char * name)1076 set_conversion_failed_error(struct archive_read *a,
1077 struct archive_string_conv *sconv, const char *name)
1078 {
1079 if (errno == ENOMEM) {
1080 archive_set_error(&a->archive, ENOMEM,
1081 "Can't allocate memory for %s", name);
1082 return (ARCHIVE_FATAL);
1083 }
1084 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1085 "%s can't be converted from %s to current locale.",
1086 name, archive_string_conversion_charset_name(sconv));
1087 return (ARCHIVE_WARN);
1088 }
1089
1090 /*
1091 * Interpret 'L' long filename header.
1092 */
1093 static int
header_longname(struct archive_read * a,struct tar * tar,struct archive_entry * entry,const void * h,size_t * unconsumed)1094 header_longname(struct archive_read *a, struct tar *tar,
1095 struct archive_entry *entry, const void *h, size_t *unconsumed)
1096 {
1097 int err;
1098
1099 err = read_body_to_string(a, tar, &(tar->longname), h, unconsumed);
1100 if (err != ARCHIVE_OK)
1101 return (err);
1102 /* Read and parse "real" header, then override name. */
1103 err = tar_read_header(a, tar, entry, unconsumed);
1104 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN))
1105 return (err);
1106 if (archive_entry_copy_pathname_l(entry, tar->longname.s,
1107 archive_strlen(&(tar->longname)), tar->sconv) != 0)
1108 err = set_conversion_failed_error(a, tar->sconv, "Pathname");
1109 return (err);
1110 }
1111
1112
1113 /*
1114 * Interpret 'V' GNU tar volume header.
1115 */
1116 static int
header_volume(struct archive_read * a,struct tar * tar,struct archive_entry * entry,const void * h,size_t * unconsumed)1117 header_volume(struct archive_read *a, struct tar *tar,
1118 struct archive_entry *entry, const void *h, size_t *unconsumed)
1119 {
1120 (void)h;
1121
1122 /* Just skip this and read the next header. */
1123 return (tar_read_header(a, tar, entry, unconsumed));
1124 }
1125
1126 /*
1127 * Read body of an archive entry into an archive_string object.
1128 */
1129 static int
read_body_to_string(struct archive_read * a,struct tar * tar,struct archive_string * as,const void * h,size_t * unconsumed)1130 read_body_to_string(struct archive_read *a, struct tar *tar,
1131 struct archive_string *as, const void *h, size_t *unconsumed)
1132 {
1133 int64_t size;
1134 const struct archive_entry_header_ustar *header;
1135 const void *src;
1136
1137 (void)tar; /* UNUSED */
1138 header = (const struct archive_entry_header_ustar *)h;
1139 size = tar_atol(header->size, sizeof(header->size));
1140 if ((size > 1048576) || (size < 0)) {
1141 archive_set_error(&a->archive, EINVAL,
1142 "Special header too large");
1143 return (ARCHIVE_FATAL);
1144 }
1145
1146 /* Fail if we can't make our buffer big enough. */
1147 if (archive_string_ensure(as, (size_t)size+1) == NULL) {
1148 archive_set_error(&a->archive, ENOMEM,
1149 "No memory");
1150 return (ARCHIVE_FATAL);
1151 }
1152
1153 tar_flush_unconsumed(a, unconsumed);
1154
1155 /* Read the body into the string. */
1156 *unconsumed = (size_t)((size + 511) & ~ 511);
1157 src = __archive_read_ahead(a, *unconsumed, NULL);
1158 if (src == NULL) {
1159 *unconsumed = 0;
1160 return (ARCHIVE_FATAL);
1161 }
1162 memcpy(as->s, src, (size_t)size);
1163 as->s[size] = '\0';
1164 as->length = (size_t)size;
1165 return (ARCHIVE_OK);
1166 }
1167
1168 /*
1169 * Parse out common header elements.
1170 *
1171 * This would be the same as header_old_tar, except that the
1172 * filename is handled slightly differently for old and POSIX
1173 * entries (POSIX entries support a 'prefix'). This factoring
1174 * allows header_old_tar and header_ustar
1175 * to handle filenames differently, while still putting most of the
1176 * common parsing into one place.
1177 */
1178 static int
header_common(struct archive_read * a,struct tar * tar,struct archive_entry * entry,const void * h)1179 header_common(struct archive_read *a, struct tar *tar,
1180 struct archive_entry *entry, const void *h)
1181 {
1182 const struct archive_entry_header_ustar *header;
1183 char tartype;
1184 int err = ARCHIVE_OK;
1185
1186 header = (const struct archive_entry_header_ustar *)h;
1187 if (header->linkname[0])
1188 archive_strncpy(&(tar->entry_linkpath),
1189 header->linkname, sizeof(header->linkname));
1190 else
1191 archive_string_empty(&(tar->entry_linkpath));
1192
1193 /* Parse out the numeric fields (all are octal) */
1194 archive_entry_set_mode(entry,
1195 (mode_t)tar_atol(header->mode, sizeof(header->mode)));
1196 archive_entry_set_uid(entry, tar_atol(header->uid, sizeof(header->uid)));
1197 archive_entry_set_gid(entry, tar_atol(header->gid, sizeof(header->gid)));
1198 tar->entry_bytes_remaining = tar_atol(header->size, sizeof(header->size));
1199 if (tar->entry_bytes_remaining < 0) {
1200 tar->entry_bytes_remaining = 0;
1201 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1202 "Tar entry has negative size");
1203 return (ARCHIVE_FATAL);
1204 }
1205 if (tar->entry_bytes_remaining == INT64_MAX) {
1206 /* Note: tar_atol returns INT64_MAX on overflow */
1207 tar->entry_bytes_remaining = 0;
1208 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1209 "Tar entry size overflow");
1210 return (ARCHIVE_FATAL);
1211 }
1212 tar->realsize = tar->entry_bytes_remaining;
1213 archive_entry_set_size(entry, tar->entry_bytes_remaining);
1214 archive_entry_set_mtime(entry, tar_atol(header->mtime, sizeof(header->mtime)), 0);
1215
1216 /* Handle the tar type flag appropriately. */
1217 tartype = header->typeflag[0];
1218
1219 switch (tartype) {
1220 case '1': /* Hard link */
1221 if (archive_entry_copy_hardlink_l(entry, tar->entry_linkpath.s,
1222 archive_strlen(&(tar->entry_linkpath)), tar->sconv) != 0) {
1223 err = set_conversion_failed_error(a, tar->sconv,
1224 "Linkname");
1225 if (err == ARCHIVE_FATAL)
1226 return (err);
1227 }
1228 /*
1229 * The following may seem odd, but: Technically, tar
1230 * does not store the file type for a "hard link"
1231 * entry, only the fact that it is a hard link. So, I
1232 * leave the type zero normally. But, pax interchange
1233 * format allows hard links to have data, which
1234 * implies that the underlying entry is a regular
1235 * file.
1236 */
1237 if (archive_entry_size(entry) > 0)
1238 archive_entry_set_filetype(entry, AE_IFREG);
1239
1240 /*
1241 * A tricky point: Traditionally, tar readers have
1242 * ignored the size field when reading hardlink
1243 * entries, and some writers put non-zero sizes even
1244 * though the body is empty. POSIX blessed this
1245 * convention in the 1988 standard, but broke with
1246 * this tradition in 2001 by permitting hardlink
1247 * entries to store valid bodies in pax interchange
1248 * format, but not in ustar format. Since there is no
1249 * hard and fast way to distinguish pax interchange
1250 * from earlier archives (the 'x' and 'g' entries are
1251 * optional, after all), we need a heuristic.
1252 */
1253 if (archive_entry_size(entry) == 0) {
1254 /* If the size is already zero, we're done. */
1255 } else if (a->archive.archive_format
1256 == ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE) {
1257 /* Definitely pax extended; must obey hardlink size. */
1258 } else if (a->archive.archive_format == ARCHIVE_FORMAT_TAR
1259 || a->archive.archive_format == ARCHIVE_FORMAT_TAR_GNUTAR)
1260 {
1261 /* Old-style or GNU tar: we must ignore the size. */
1262 archive_entry_set_size(entry, 0);
1263 tar->entry_bytes_remaining = 0;
1264 } else if (archive_read_format_tar_bid(a, 50) > 50) {
1265 /*
1266 * We don't know if it's pax: If the bid
1267 * function sees a valid ustar header
1268 * immediately following, then let's ignore
1269 * the hardlink size.
1270 */
1271 archive_entry_set_size(entry, 0);
1272 tar->entry_bytes_remaining = 0;
1273 }
1274 /*
1275 * TODO: There are still two cases I'd like to handle:
1276 * = a ustar non-pax archive with a hardlink entry at
1277 * end-of-archive. (Look for block of nulls following?)
1278 * = a pax archive that has not seen any pax headers
1279 * and has an entry which is a hardlink entry storing
1280 * a body containing an uncompressed tar archive.
1281 * The first is worth addressing; I don't see any reliable
1282 * way to deal with the second possibility.
1283 */
1284 break;
1285 case '2': /* Symlink */
1286 archive_entry_set_filetype(entry, AE_IFLNK);
1287 archive_entry_set_size(entry, 0);
1288 tar->entry_bytes_remaining = 0;
1289 if (archive_entry_copy_symlink_l(entry, tar->entry_linkpath.s,
1290 archive_strlen(&(tar->entry_linkpath)), tar->sconv) != 0) {
1291 err = set_conversion_failed_error(a, tar->sconv,
1292 "Linkname");
1293 if (err == ARCHIVE_FATAL)
1294 return (err);
1295 }
1296 break;
1297 case '3': /* Character device */
1298 archive_entry_set_filetype(entry, AE_IFCHR);
1299 archive_entry_set_size(entry, 0);
1300 tar->entry_bytes_remaining = 0;
1301 break;
1302 case '4': /* Block device */
1303 archive_entry_set_filetype(entry, AE_IFBLK);
1304 archive_entry_set_size(entry, 0);
1305 tar->entry_bytes_remaining = 0;
1306 break;
1307 case '5': /* Dir */
1308 archive_entry_set_filetype(entry, AE_IFDIR);
1309 archive_entry_set_size(entry, 0);
1310 tar->entry_bytes_remaining = 0;
1311 break;
1312 case '6': /* FIFO device */
1313 archive_entry_set_filetype(entry, AE_IFIFO);
1314 archive_entry_set_size(entry, 0);
1315 tar->entry_bytes_remaining = 0;
1316 break;
1317 case 'D': /* GNU incremental directory type */
1318 /*
1319 * No special handling is actually required here.
1320 * It might be nice someday to preprocess the file list and
1321 * provide it to the client, though.
1322 */
1323 archive_entry_set_filetype(entry, AE_IFDIR);
1324 break;
1325 case 'M': /* GNU "Multi-volume" (remainder of file from last archive)*/
1326 /*
1327 * As far as I can tell, this is just like a regular file
1328 * entry, except that the contents should be _appended_ to
1329 * the indicated file at the indicated offset. This may
1330 * require some API work to fully support.
1331 */
1332 break;
1333 case 'N': /* Old GNU "long filename" entry. */
1334 /* The body of this entry is a script for renaming
1335 * previously-extracted entries. Ugh. It will never
1336 * be supported by libarchive. */
1337 archive_entry_set_filetype(entry, AE_IFREG);
1338 break;
1339 case 'S': /* GNU sparse files */
1340 /*
1341 * Sparse files are really just regular files with
1342 * sparse information in the extended area.
1343 */
1344 /* FALLTHROUGH */
1345 case '0':
1346 /*
1347 * Enable sparse file "read" support only for regular
1348 * files and explicit GNU sparse files. However, we
1349 * don't allow non-standard file types to be sparse.
1350 */
1351 tar->sparse_allowed = 1;
1352 /* FALLTHROUGH */
1353 default: /* Regular file and non-standard types */
1354 /*
1355 * Per POSIX: non-recognized types should always be
1356 * treated as regular files.
1357 */
1358 archive_entry_set_filetype(entry, AE_IFREG);
1359 break;
1360 }
1361 return (err);
1362 }
1363
1364 /*
1365 * Parse out header elements for "old-style" tar archives.
1366 */
1367 static int
header_old_tar(struct archive_read * a,struct tar * tar,struct archive_entry * entry,const void * h)1368 header_old_tar(struct archive_read *a, struct tar *tar,
1369 struct archive_entry *entry, const void *h)
1370 {
1371 const struct archive_entry_header_ustar *header;
1372 int err = ARCHIVE_OK, err2;
1373
1374 /* Copy filename over (to ensure null termination). */
1375 header = (const struct archive_entry_header_ustar *)h;
1376 if (archive_entry_copy_pathname_l(entry,
1377 header->name, sizeof(header->name), tar->sconv) != 0) {
1378 err = set_conversion_failed_error(a, tar->sconv, "Pathname");
1379 if (err == ARCHIVE_FATAL)
1380 return (err);
1381 }
1382
1383 /* Grab rest of common fields */
1384 err2 = header_common(a, tar, entry, h);
1385 if (err > err2)
1386 err = err2;
1387
1388 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining);
1389 return (err);
1390 }
1391
1392 /*
1393 * Read a Mac AppleDouble-encoded blob of file metadata,
1394 * if there is one.
1395 */
1396 static int
read_mac_metadata_blob(struct archive_read * a,struct tar * tar,struct archive_entry * entry,const void * h,size_t * unconsumed)1397 read_mac_metadata_blob(struct archive_read *a, struct tar *tar,
1398 struct archive_entry *entry, const void *h, size_t *unconsumed)
1399 {
1400 int64_t size;
1401 size_t msize;
1402 const void *data;
1403 const char *p, *name;
1404 const wchar_t *wp, *wname;
1405
1406 (void)h; /* UNUSED */
1407
1408 wname = wp = archive_entry_pathname_w(entry);
1409 if (wp != NULL) {
1410 /* Find the last path element. */
1411 for (; *wp != L'\0'; ++wp) {
1412 if (wp[0] == '/' && wp[1] != L'\0')
1413 wname = wp + 1;
1414 }
1415 /*
1416 * If last path element starts with "._", then
1417 * this is a Mac extension.
1418 */
1419 if (wname[0] != L'.' || wname[1] != L'_' || wname[2] == L'\0')
1420 return ARCHIVE_OK;
1421 } else {
1422 /* Find the last path element. */
1423 name = p = archive_entry_pathname(entry);
1424 if (p == NULL)
1425 return (ARCHIVE_FAILED);
1426 for (; *p != '\0'; ++p) {
1427 if (p[0] == '/' && p[1] != '\0')
1428 name = p + 1;
1429 }
1430 /*
1431 * If last path element starts with "._", then
1432 * this is a Mac extension.
1433 */
1434 if (name[0] != '.' || name[1] != '_' || name[2] == '\0')
1435 return ARCHIVE_OK;
1436 }
1437
1438 /* Read the body as a Mac OS metadata blob. */
1439 size = archive_entry_size(entry);
1440 msize = (size_t)size;
1441 if (size < 0 || (uintmax_t)msize != (uintmax_t)size) {
1442 *unconsumed = 0;
1443 return (ARCHIVE_FATAL);
1444 }
1445
1446 /*
1447 * TODO: Look beyond the body here to peek at the next header.
1448 * If it's a regular header (not an extension header)
1449 * that has the wrong name, just return the current
1450 * entry as-is, without consuming the body here.
1451 * That would reduce the risk of us mis-identifying
1452 * an ordinary file that just happened to have
1453 * a name starting with "._".
1454 *
1455 * Q: Is the above idea really possible? Even
1456 * when there are GNU or pax extension entries?
1457 */
1458 data = __archive_read_ahead(a, msize, NULL);
1459 if (data == NULL) {
1460 *unconsumed = 0;
1461 return (ARCHIVE_FATAL);
1462 }
1463 archive_entry_copy_mac_metadata(entry, data, msize);
1464 *unconsumed = (msize + 511) & ~ 511;
1465 tar_flush_unconsumed(a, unconsumed);
1466 return (tar_read_header(a, tar, entry, unconsumed));
1467 }
1468
1469 /*
1470 * Parse a file header for a pax extended archive entry.
1471 */
1472 static int
header_pax_global(struct archive_read * a,struct tar * tar,struct archive_entry * entry,const void * h,size_t * unconsumed)1473 header_pax_global(struct archive_read *a, struct tar *tar,
1474 struct archive_entry *entry, const void *h, size_t *unconsumed)
1475 {
1476 int err;
1477
1478 err = read_body_to_string(a, tar, &(tar->pax_global), h, unconsumed);
1479 if (err != ARCHIVE_OK)
1480 return (err);
1481 err = tar_read_header(a, tar, entry, unconsumed);
1482 return (err);
1483 }
1484
1485 static int
header_pax_extensions(struct archive_read * a,struct tar * tar,struct archive_entry * entry,const void * h,size_t * unconsumed)1486 header_pax_extensions(struct archive_read *a, struct tar *tar,
1487 struct archive_entry *entry, const void *h, size_t *unconsumed)
1488 {
1489 int err, err2;
1490
1491 err = read_body_to_string(a, tar, &(tar->pax_header), h, unconsumed);
1492 if (err != ARCHIVE_OK)
1493 return (err);
1494
1495 /* Parse the next header. */
1496 err = tar_read_header(a, tar, entry, unconsumed);
1497 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN))
1498 return (err);
1499
1500 /*
1501 * TODO: Parse global/default options into 'entry' struct here
1502 * before handling file-specific options.
1503 *
1504 * This design (parse standard header, then overwrite with pax
1505 * extended attribute data) usually works well, but isn't ideal;
1506 * it would be better to parse the pax extended attributes first
1507 * and then skip any fields in the standard header that were
1508 * defined in the pax header.
1509 */
1510 err2 = pax_header(a, tar, entry, &tar->pax_header);
1511 err = err_combine(err, err2);
1512 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining);
1513 return (err);
1514 }
1515
1516
1517 /*
1518 * Parse a file header for a Posix "ustar" archive entry. This also
1519 * handles "pax" or "extended ustar" entries.
1520 */
1521 static int
header_ustar(struct archive_read * a,struct tar * tar,struct archive_entry * entry,const void * h)1522 header_ustar(struct archive_read *a, struct tar *tar,
1523 struct archive_entry *entry, const void *h)
1524 {
1525 const struct archive_entry_header_ustar *header;
1526 struct archive_string *as;
1527 int err = ARCHIVE_OK, r;
1528
1529 header = (const struct archive_entry_header_ustar *)h;
1530
1531 /* Copy name into an internal buffer to ensure null-termination. */
1532 as = &(tar->entry_pathname);
1533 if (header->prefix[0]) {
1534 archive_strncpy(as, header->prefix, sizeof(header->prefix));
1535 if (as->s[archive_strlen(as) - 1] != '/')
1536 archive_strappend_char(as, '/');
1537 archive_strncat(as, header->name, sizeof(header->name));
1538 } else {
1539 archive_strncpy(as, header->name, sizeof(header->name));
1540 }
1541 if (archive_entry_copy_pathname_l(entry, as->s, archive_strlen(as),
1542 tar->sconv) != 0) {
1543 err = set_conversion_failed_error(a, tar->sconv, "Pathname");
1544 if (err == ARCHIVE_FATAL)
1545 return (err);
1546 }
1547
1548 /* Handle rest of common fields. */
1549 r = header_common(a, tar, entry, h);
1550 if (r == ARCHIVE_FATAL)
1551 return (r);
1552 if (r < err)
1553 err = r;
1554
1555 /* Handle POSIX ustar fields. */
1556 if (archive_entry_copy_uname_l(entry,
1557 header->uname, sizeof(header->uname), tar->sconv) != 0) {
1558 err = set_conversion_failed_error(a, tar->sconv, "Uname");
1559 if (err == ARCHIVE_FATAL)
1560 return (err);
1561 }
1562
1563 if (archive_entry_copy_gname_l(entry,
1564 header->gname, sizeof(header->gname), tar->sconv) != 0) {
1565 err = set_conversion_failed_error(a, tar->sconv, "Gname");
1566 if (err == ARCHIVE_FATAL)
1567 return (err);
1568 }
1569
1570 /* Parse out device numbers only for char and block specials. */
1571 if (header->typeflag[0] == '3' || header->typeflag[0] == '4') {
1572 archive_entry_set_rdevmajor(entry, (dev_t)
1573 tar_atol(header->rdevmajor, sizeof(header->rdevmajor)));
1574 archive_entry_set_rdevminor(entry, (dev_t)
1575 tar_atol(header->rdevminor, sizeof(header->rdevminor)));
1576 }
1577
1578 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining);
1579
1580 return (err);
1581 }
1582
1583
1584 /*
1585 * Parse the pax extended attributes record.
1586 *
1587 * Returns non-zero if there's an error in the data.
1588 */
1589 static int
pax_header(struct archive_read * a,struct tar * tar,struct archive_entry * entry,struct archive_string * in_as)1590 pax_header(struct archive_read *a, struct tar *tar,
1591 struct archive_entry *entry, struct archive_string *in_as)
1592 {
1593 size_t attr_length, l, line_length, value_length;
1594 char *p;
1595 char *key, *value;
1596 struct archive_string *as;
1597 struct archive_string_conv *sconv;
1598 int err, err2;
1599 char *attr = in_as->s;
1600
1601 attr_length = in_as->length;
1602 tar->pax_hdrcharset_binary = 0;
1603 archive_string_empty(&(tar->entry_gname));
1604 archive_string_empty(&(tar->entry_linkpath));
1605 archive_string_empty(&(tar->entry_pathname));
1606 archive_string_empty(&(tar->entry_pathname_override));
1607 archive_string_empty(&(tar->entry_uname));
1608 err = ARCHIVE_OK;
1609 while (attr_length > 0) {
1610 /* Parse decimal length field at start of line. */
1611 line_length = 0;
1612 l = attr_length;
1613 p = attr; /* Record start of line. */
1614 while (l>0) {
1615 if (*p == ' ') {
1616 p++;
1617 l--;
1618 break;
1619 }
1620 if (*p < '0' || *p > '9') {
1621 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1622 "Ignoring malformed pax extended attributes");
1623 return (ARCHIVE_WARN);
1624 }
1625 line_length *= 10;
1626 line_length += *p - '0';
1627 if (line_length > 999999) {
1628 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1629 "Rejecting pax extended attribute > 1MB");
1630 return (ARCHIVE_WARN);
1631 }
1632 p++;
1633 l--;
1634 }
1635
1636 /*
1637 * Parsed length must be no bigger than available data,
1638 * at least 1, and the last character of the line must
1639 * be '\n'.
1640 */
1641 if (line_length > attr_length
1642 || line_length < 1
1643 || attr[line_length - 1] != '\n')
1644 {
1645 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1646 "Ignoring malformed pax extended attribute");
1647 return (ARCHIVE_WARN);
1648 }
1649
1650 /* Null-terminate the line. */
1651 attr[line_length - 1] = '\0';
1652
1653 /* Find end of key and null terminate it. */
1654 key = p;
1655 if (key[0] == '=')
1656 return (-1);
1657 while (*p && *p != '=')
1658 ++p;
1659 if (*p == '\0') {
1660 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1661 "Invalid pax extended attributes");
1662 return (ARCHIVE_WARN);
1663 }
1664 *p = '\0';
1665
1666 value = p + 1;
1667
1668 /* Some values may be binary data */
1669 value_length = attr + line_length - 1 - value;
1670
1671 /* Identify this attribute and set it in the entry. */
1672 err2 = pax_attribute(a, tar, entry, key, value, value_length);
1673 if (err2 == ARCHIVE_FATAL)
1674 return (err2);
1675 err = err_combine(err, err2);
1676
1677 /* Skip to next line */
1678 attr += line_length;
1679 attr_length -= line_length;
1680 }
1681
1682 /*
1683 * PAX format uses UTF-8 as default charset for its metadata
1684 * unless hdrcharset=BINARY is present in its header.
1685 * We apply the charset specified by the hdrcharset option only
1686 * when the hdrcharset attribute(in PAX header) is BINARY because
1687 * we respect the charset described in PAX header and BINARY also
1688 * means that metadata(filename,uname and gname) character-set
1689 * is unknown.
1690 */
1691 if (tar->pax_hdrcharset_binary)
1692 sconv = tar->opt_sconv;
1693 else {
1694 sconv = archive_string_conversion_from_charset(
1695 &(a->archive), "UTF-8", 1);
1696 if (sconv == NULL)
1697 return (ARCHIVE_FATAL);
1698 if (tar->compat_2x)
1699 archive_string_conversion_set_opt(sconv,
1700 SCONV_SET_OPT_UTF8_LIBARCHIVE2X);
1701 }
1702
1703 if (archive_strlen(&(tar->entry_gname)) > 0) {
1704 if (archive_entry_copy_gname_l(entry, tar->entry_gname.s,
1705 archive_strlen(&(tar->entry_gname)), sconv) != 0) {
1706 err = set_conversion_failed_error(a, sconv, "Gname");
1707 if (err == ARCHIVE_FATAL)
1708 return (err);
1709 /* Use a converted an original name. */
1710 archive_entry_copy_gname(entry, tar->entry_gname.s);
1711 }
1712 }
1713 if (archive_strlen(&(tar->entry_linkpath)) > 0) {
1714 if (archive_entry_copy_link_l(entry, tar->entry_linkpath.s,
1715 archive_strlen(&(tar->entry_linkpath)), sconv) != 0) {
1716 err = set_conversion_failed_error(a, sconv, "Linkname");
1717 if (err == ARCHIVE_FATAL)
1718 return (err);
1719 /* Use a converted an original name. */
1720 archive_entry_copy_link(entry, tar->entry_linkpath.s);
1721 }
1722 }
1723 /*
1724 * Some extensions (such as the GNU sparse file extensions)
1725 * deliberately store a synthetic name under the regular 'path'
1726 * attribute and the real file name under a different attribute.
1727 * Since we're supposed to not care about the order, we
1728 * have no choice but to store all of the various filenames
1729 * we find and figure it all out afterwards. This is the
1730 * figuring out part.
1731 */
1732 as = NULL;
1733 if (archive_strlen(&(tar->entry_pathname_override)) > 0)
1734 as = &(tar->entry_pathname_override);
1735 else if (archive_strlen(&(tar->entry_pathname)) > 0)
1736 as = &(tar->entry_pathname);
1737 if (as != NULL) {
1738 if (archive_entry_copy_pathname_l(entry, as->s,
1739 archive_strlen(as), sconv) != 0) {
1740 err = set_conversion_failed_error(a, sconv, "Pathname");
1741 if (err == ARCHIVE_FATAL)
1742 return (err);
1743 /* Use a converted an original name. */
1744 archive_entry_copy_pathname(entry, as->s);
1745 }
1746 }
1747 if (archive_strlen(&(tar->entry_uname)) > 0) {
1748 if (archive_entry_copy_uname_l(entry, tar->entry_uname.s,
1749 archive_strlen(&(tar->entry_uname)), sconv) != 0) {
1750 err = set_conversion_failed_error(a, sconv, "Uname");
1751 if (err == ARCHIVE_FATAL)
1752 return (err);
1753 /* Use a converted an original name. */
1754 archive_entry_copy_uname(entry, tar->entry_uname.s);
1755 }
1756 }
1757 return (err);
1758 }
1759
1760 static int
pax_attribute_xattr(struct archive_entry * entry,const char * name,const char * value)1761 pax_attribute_xattr(struct archive_entry *entry,
1762 const char *name, const char *value)
1763 {
1764 char *name_decoded;
1765 void *value_decoded;
1766 size_t value_len;
1767
1768 if (strlen(name) < 18 || (memcmp(name, "LIBARCHIVE.xattr.", 17)) != 0)
1769 return 3;
1770
1771 name += 17;
1772
1773 /* URL-decode name */
1774 name_decoded = url_decode(name);
1775 if (name_decoded == NULL)
1776 return 2;
1777
1778 /* Base-64 decode value */
1779 value_decoded = base64_decode(value, strlen(value), &value_len);
1780 if (value_decoded == NULL) {
1781 free(name_decoded);
1782 return 1;
1783 }
1784
1785 archive_entry_xattr_add_entry(entry, name_decoded,
1786 value_decoded, value_len);
1787
1788 free(name_decoded);
1789 free(value_decoded);
1790 return 0;
1791 }
1792
1793 static int
pax_attribute_schily_xattr(struct archive_entry * entry,const char * name,const char * value,size_t value_length)1794 pax_attribute_schily_xattr(struct archive_entry *entry,
1795 const char *name, const char *value, size_t value_length)
1796 {
1797 if (strlen(name) < 14 || (memcmp(name, "SCHILY.xattr.", 13)) != 0)
1798 return 1;
1799
1800 name += 13;
1801
1802 archive_entry_xattr_add_entry(entry, name, value, value_length);
1803
1804 return 0;
1805 }
1806
1807 static int
pax_attribute_rht_security_selinux(struct archive_entry * entry,const char * value,size_t value_length)1808 pax_attribute_rht_security_selinux(struct archive_entry *entry,
1809 const char *value, size_t value_length)
1810 {
1811 archive_entry_xattr_add_entry(entry, "security.selinux",
1812 value, value_length);
1813
1814 return 0;
1815 }
1816
1817 static int
pax_attribute_acl(struct archive_read * a,struct tar * tar,struct archive_entry * entry,const char * value,int type)1818 pax_attribute_acl(struct archive_read *a, struct tar *tar,
1819 struct archive_entry *entry, const char *value, int type)
1820 {
1821 int r;
1822 const char* errstr;
1823
1824 switch (type) {
1825 case ARCHIVE_ENTRY_ACL_TYPE_ACCESS:
1826 errstr = "SCHILY.acl.access";
1827 break;
1828 case ARCHIVE_ENTRY_ACL_TYPE_DEFAULT:
1829 errstr = "SCHILY.acl.default";
1830 break;
1831 case ARCHIVE_ENTRY_ACL_TYPE_NFS4:
1832 errstr = "SCHILY.acl.ace";
1833 break;
1834 default:
1835 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1836 "Unknown ACL type: %d", type);
1837 return(ARCHIVE_FATAL);
1838 }
1839
1840 if (tar->sconv_acl == NULL) {
1841 tar->sconv_acl =
1842 archive_string_conversion_from_charset(
1843 &(a->archive), "UTF-8", 1);
1844 if (tar->sconv_acl == NULL)
1845 return (ARCHIVE_FATAL);
1846 }
1847
1848 r = archive_acl_from_text_l(archive_entry_acl(entry), value, type,
1849 tar->sconv_acl);
1850 if (r != ARCHIVE_OK) {
1851 if (r == ARCHIVE_FATAL) {
1852 archive_set_error(&a->archive, ENOMEM,
1853 "%s %s", "Can't allocate memory for ",
1854 errstr);
1855 return (r);
1856 }
1857 archive_set_error(&a->archive,
1858 ARCHIVE_ERRNO_MISC, "%s %s", "Parse error: ", errstr);
1859 }
1860 return (r);
1861 }
1862
1863 /*
1864 * Parse a single key=value attribute. key/value pointers are
1865 * assumed to point into reasonably long-lived storage.
1866 *
1867 * Note that POSIX reserves all-lowercase keywords. Vendor-specific
1868 * extensions should always have keywords of the form "VENDOR.attribute"
1869 * In particular, it's quite feasible to support many different
1870 * vendor extensions here. I'm using "LIBARCHIVE" for extensions
1871 * unique to this library.
1872 *
1873 * Investigate other vendor-specific extensions and see if
1874 * any of them look useful.
1875 */
1876 static int
pax_attribute(struct archive_read * a,struct tar * tar,struct archive_entry * entry,const char * key,const char * value,size_t value_length)1877 pax_attribute(struct archive_read *a, struct tar *tar,
1878 struct archive_entry *entry, const char *key, const char *value, size_t value_length)
1879 {
1880 int64_t s;
1881 long n;
1882 int err = ARCHIVE_OK, r;
1883
1884 #ifndef __FreeBSD__
1885 if (value == NULL)
1886 value = ""; /* Disable compiler warning; do not pass
1887 * NULL pointer to strlen(). */
1888 #endif
1889 switch (key[0]) {
1890 case 'G':
1891 /* Reject GNU.sparse.* headers on non-regular files. */
1892 if (strncmp(key, "GNU.sparse", 10) == 0 &&
1893 !tar->sparse_allowed) {
1894 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1895 "Non-regular file cannot be sparse");
1896 return (ARCHIVE_FATAL);
1897 }
1898
1899 /* GNU "0.0" sparse pax format. */
1900 if (strcmp(key, "GNU.sparse.numblocks") == 0) {
1901 tar->sparse_offset = -1;
1902 tar->sparse_numbytes = -1;
1903 tar->sparse_gnu_major = 0;
1904 tar->sparse_gnu_minor = 0;
1905 }
1906 if (strcmp(key, "GNU.sparse.offset") == 0) {
1907 tar->sparse_offset = tar_atol10(value, strlen(value));
1908 if (tar->sparse_numbytes != -1) {
1909 if (gnu_add_sparse_entry(a, tar,
1910 tar->sparse_offset, tar->sparse_numbytes)
1911 != ARCHIVE_OK)
1912 return (ARCHIVE_FATAL);
1913 tar->sparse_offset = -1;
1914 tar->sparse_numbytes = -1;
1915 }
1916 }
1917 if (strcmp(key, "GNU.sparse.numbytes") == 0) {
1918 tar->sparse_numbytes = tar_atol10(value, strlen(value));
1919 if (tar->sparse_offset != -1) {
1920 if (gnu_add_sparse_entry(a, tar,
1921 tar->sparse_offset, tar->sparse_numbytes)
1922 != ARCHIVE_OK)
1923 return (ARCHIVE_FATAL);
1924 tar->sparse_offset = -1;
1925 tar->sparse_numbytes = -1;
1926 }
1927 }
1928 if (strcmp(key, "GNU.sparse.size") == 0) {
1929 tar->realsize = tar_atol10(value, strlen(value));
1930 archive_entry_set_size(entry, tar->realsize);
1931 tar->realsize_override = 1;
1932 }
1933
1934 /* GNU "0.1" sparse pax format. */
1935 if (strcmp(key, "GNU.sparse.map") == 0) {
1936 tar->sparse_gnu_major = 0;
1937 tar->sparse_gnu_minor = 1;
1938 if (gnu_sparse_01_parse(a, tar, value) != ARCHIVE_OK)
1939 return (ARCHIVE_WARN);
1940 }
1941
1942 /* GNU "1.0" sparse pax format */
1943 if (strcmp(key, "GNU.sparse.major") == 0) {
1944 tar->sparse_gnu_major = (int)tar_atol10(value, strlen(value));
1945 tar->sparse_gnu_pending = 1;
1946 }
1947 if (strcmp(key, "GNU.sparse.minor") == 0) {
1948 tar->sparse_gnu_minor = (int)tar_atol10(value, strlen(value));
1949 tar->sparse_gnu_pending = 1;
1950 }
1951 if (strcmp(key, "GNU.sparse.name") == 0) {
1952 /*
1953 * The real filename; when storing sparse
1954 * files, GNU tar puts a synthesized name into
1955 * the regular 'path' attribute in an attempt
1956 * to limit confusion. ;-)
1957 */
1958 archive_strcpy(&(tar->entry_pathname_override), value);
1959 }
1960 if (strcmp(key, "GNU.sparse.realsize") == 0) {
1961 tar->realsize = tar_atol10(value, strlen(value));
1962 archive_entry_set_size(entry, tar->realsize);
1963 tar->realsize_override = 1;
1964 }
1965 break;
1966 case 'L':
1967 /* Our extensions */
1968 /* TODO: Handle arbitrary extended attributes... */
1969 /*
1970 if (strcmp(key, "LIBARCHIVE.xxxxxxx") == 0)
1971 archive_entry_set_xxxxxx(entry, value);
1972 */
1973 if (strcmp(key, "LIBARCHIVE.creationtime") == 0) {
1974 pax_time(value, &s, &n);
1975 archive_entry_set_birthtime(entry, s, n);
1976 }
1977 if (strcmp(key, "LIBARCHIVE.symlinktype") == 0) {
1978 if (strcmp(value, "file") == 0) {
1979 archive_entry_set_symlink_type(entry,
1980 AE_SYMLINK_TYPE_FILE);
1981 } else if (strcmp(value, "dir") == 0) {
1982 archive_entry_set_symlink_type(entry,
1983 AE_SYMLINK_TYPE_DIRECTORY);
1984 }
1985 }
1986 if (memcmp(key, "LIBARCHIVE.xattr.", 17) == 0)
1987 pax_attribute_xattr(entry, key, value);
1988 break;
1989 case 'R':
1990 /* GNU tar uses RHT.security header to store SELinux xattrs
1991 * SCHILY.xattr.security.selinux == RHT.security.selinux */
1992 if (strcmp(key, "RHT.security.selinux") == 0) {
1993 pax_attribute_rht_security_selinux(entry, value,
1994 value_length);
1995 }
1996 break;
1997 case 'S':
1998 /* We support some keys used by the "star" archiver */
1999 if (strcmp(key, "SCHILY.acl.access") == 0) {
2000 r = pax_attribute_acl(a, tar, entry, value,
2001 ARCHIVE_ENTRY_ACL_TYPE_ACCESS);
2002 if (r == ARCHIVE_FATAL)
2003 return (r);
2004 } else if (strcmp(key, "SCHILY.acl.default") == 0) {
2005 r = pax_attribute_acl(a, tar, entry, value,
2006 ARCHIVE_ENTRY_ACL_TYPE_DEFAULT);
2007 if (r == ARCHIVE_FATAL)
2008 return (r);
2009 } else if (strcmp(key, "SCHILY.acl.ace") == 0) {
2010 r = pax_attribute_acl(a, tar, entry, value,
2011 ARCHIVE_ENTRY_ACL_TYPE_NFS4);
2012 if (r == ARCHIVE_FATAL)
2013 return (r);
2014 } else if (strcmp(key, "SCHILY.devmajor") == 0) {
2015 archive_entry_set_rdevmajor(entry,
2016 (dev_t)tar_atol10(value, strlen(value)));
2017 } else if (strcmp(key, "SCHILY.devminor") == 0) {
2018 archive_entry_set_rdevminor(entry,
2019 (dev_t)tar_atol10(value, strlen(value)));
2020 } else if (strcmp(key, "SCHILY.fflags") == 0) {
2021 archive_entry_copy_fflags_text(entry, value);
2022 } else if (strcmp(key, "SCHILY.dev") == 0) {
2023 archive_entry_set_dev(entry,
2024 (dev_t)tar_atol10(value, strlen(value)));
2025 } else if (strcmp(key, "SCHILY.ino") == 0) {
2026 archive_entry_set_ino(entry,
2027 tar_atol10(value, strlen(value)));
2028 } else if (strcmp(key, "SCHILY.nlink") == 0) {
2029 archive_entry_set_nlink(entry, (unsigned)
2030 tar_atol10(value, strlen(value)));
2031 } else if (strcmp(key, "SCHILY.realsize") == 0) {
2032 tar->realsize = tar_atol10(value, strlen(value));
2033 tar->realsize_override = 1;
2034 archive_entry_set_size(entry, tar->realsize);
2035 } else if (strncmp(key, "SCHILY.xattr.", 13) == 0) {
2036 pax_attribute_schily_xattr(entry, key, value,
2037 value_length);
2038 } else if (strcmp(key, "SUN.holesdata") == 0) {
2039 /* A Solaris extension for sparse. */
2040 r = solaris_sparse_parse(a, tar, entry, value);
2041 if (r < err) {
2042 if (r == ARCHIVE_FATAL)
2043 return (r);
2044 err = r;
2045 archive_set_error(&a->archive,
2046 ARCHIVE_ERRNO_MISC,
2047 "Parse error: SUN.holesdata");
2048 }
2049 }
2050 break;
2051 case 'a':
2052 if (strcmp(key, "atime") == 0) {
2053 pax_time(value, &s, &n);
2054 archive_entry_set_atime(entry, s, n);
2055 }
2056 break;
2057 case 'c':
2058 if (strcmp(key, "ctime") == 0) {
2059 pax_time(value, &s, &n);
2060 archive_entry_set_ctime(entry, s, n);
2061 } else if (strcmp(key, "charset") == 0) {
2062 /* TODO: Publish charset information in entry. */
2063 } else if (strcmp(key, "comment") == 0) {
2064 /* TODO: Publish comment in entry. */
2065 }
2066 break;
2067 case 'g':
2068 if (strcmp(key, "gid") == 0) {
2069 archive_entry_set_gid(entry,
2070 tar_atol10(value, strlen(value)));
2071 } else if (strcmp(key, "gname") == 0) {
2072 archive_strcpy(&(tar->entry_gname), value);
2073 }
2074 break;
2075 case 'h':
2076 if (strcmp(key, "hdrcharset") == 0) {
2077 if (strcmp(value, "BINARY") == 0)
2078 /* Binary mode. */
2079 tar->pax_hdrcharset_binary = 1;
2080 else if (strcmp(value, "ISO-IR 10646 2000 UTF-8") == 0)
2081 tar->pax_hdrcharset_binary = 0;
2082 }
2083 break;
2084 case 'l':
2085 /* pax interchange doesn't distinguish hardlink vs. symlink. */
2086 if (strcmp(key, "linkpath") == 0) {
2087 archive_strcpy(&(tar->entry_linkpath), value);
2088 }
2089 break;
2090 case 'm':
2091 if (strcmp(key, "mtime") == 0) {
2092 pax_time(value, &s, &n);
2093 archive_entry_set_mtime(entry, s, n);
2094 }
2095 break;
2096 case 'p':
2097 if (strcmp(key, "path") == 0) {
2098 archive_strcpy(&(tar->entry_pathname), value);
2099 }
2100 break;
2101 case 'r':
2102 /* POSIX has reserved 'realtime.*' */
2103 break;
2104 case 's':
2105 /* POSIX has reserved 'security.*' */
2106 /* Someday: if (strcmp(key, "security.acl") == 0) { ... } */
2107 if (strcmp(key, "size") == 0) {
2108 /* "size" is the size of the data in the entry. */
2109 tar->entry_bytes_remaining
2110 = tar_atol10(value, strlen(value));
2111 if (tar->entry_bytes_remaining < 0) {
2112 tar->entry_bytes_remaining = 0;
2113 archive_set_error(&a->archive,
2114 ARCHIVE_ERRNO_MISC,
2115 "Tar size attribute is negative");
2116 return (ARCHIVE_FATAL);
2117 }
2118 if (tar->entry_bytes_remaining == INT64_MAX) {
2119 /* Note: tar_atol returns INT64_MAX on overflow */
2120 tar->entry_bytes_remaining = 0;
2121 archive_set_error(&a->archive,
2122 ARCHIVE_ERRNO_MISC,
2123 "Tar size attribute overflow");
2124 return (ARCHIVE_FATAL);
2125 }
2126 /*
2127 * The "size" pax header keyword always overrides the
2128 * "size" field in the tar header.
2129 * GNU.sparse.realsize, GNU.sparse.size and
2130 * SCHILY.realsize override this value.
2131 */
2132 if (!tar->realsize_override) {
2133 archive_entry_set_size(entry,
2134 tar->entry_bytes_remaining);
2135 tar->realsize
2136 = tar->entry_bytes_remaining;
2137 }
2138 }
2139 break;
2140 case 'u':
2141 if (strcmp(key, "uid") == 0) {
2142 archive_entry_set_uid(entry,
2143 tar_atol10(value, strlen(value)));
2144 } else if (strcmp(key, "uname") == 0) {
2145 archive_strcpy(&(tar->entry_uname), value);
2146 }
2147 break;
2148 }
2149 return (err);
2150 }
2151
2152
2153
2154 /*
2155 * parse a decimal time value, which may include a fractional portion
2156 */
2157 static void
pax_time(const char * p,int64_t * ps,long * pn)2158 pax_time(const char *p, int64_t *ps, long *pn)
2159 {
2160 char digit;
2161 int64_t s;
2162 unsigned long l;
2163 int sign;
2164 int64_t limit, last_digit_limit;
2165
2166 limit = INT64_MAX / 10;
2167 last_digit_limit = INT64_MAX % 10;
2168
2169 s = 0;
2170 sign = 1;
2171 if (*p == '-') {
2172 sign = -1;
2173 p++;
2174 }
2175 while (*p >= '0' && *p <= '9') {
2176 digit = *p - '0';
2177 if (s > limit ||
2178 (s == limit && digit > last_digit_limit)) {
2179 s = INT64_MAX;
2180 break;
2181 }
2182 s = (s * 10) + digit;
2183 ++p;
2184 }
2185
2186 *ps = s * sign;
2187
2188 /* Calculate nanoseconds. */
2189 *pn = 0;
2190
2191 if (*p != '.')
2192 return;
2193
2194 l = 100000000UL;
2195 do {
2196 ++p;
2197 if (*p >= '0' && *p <= '9')
2198 *pn += (*p - '0') * l;
2199 else
2200 break;
2201 } while (l /= 10);
2202 }
2203
2204 /*
2205 * Parse GNU tar header
2206 */
2207 static int
header_gnutar(struct archive_read * a,struct tar * tar,struct archive_entry * entry,const void * h,size_t * unconsumed)2208 header_gnutar(struct archive_read *a, struct tar *tar,
2209 struct archive_entry *entry, const void *h, size_t *unconsumed)
2210 {
2211 const struct archive_entry_header_gnutar *header;
2212 int64_t t;
2213 int err = ARCHIVE_OK;
2214
2215 /*
2216 * GNU header is like POSIX ustar, except 'prefix' is
2217 * replaced with some other fields. This also means the
2218 * filename is stored as in old-style archives.
2219 */
2220
2221 /* Grab fields common to all tar variants. */
2222 err = header_common(a, tar, entry, h);
2223 if (err == ARCHIVE_FATAL)
2224 return (err);
2225
2226 /* Copy filename over (to ensure null termination). */
2227 header = (const struct archive_entry_header_gnutar *)h;
2228 if (archive_entry_copy_pathname_l(entry,
2229 header->name, sizeof(header->name), tar->sconv) != 0) {
2230 err = set_conversion_failed_error(a, tar->sconv, "Pathname");
2231 if (err == ARCHIVE_FATAL)
2232 return (err);
2233 }
2234
2235 /* Fields common to ustar and GNU */
2236 /* XXX Can the following be factored out since it's common
2237 * to ustar and gnu tar? Is it okay to move it down into
2238 * header_common, perhaps? */
2239 if (archive_entry_copy_uname_l(entry,
2240 header->uname, sizeof(header->uname), tar->sconv) != 0) {
2241 err = set_conversion_failed_error(a, tar->sconv, "Uname");
2242 if (err == ARCHIVE_FATAL)
2243 return (err);
2244 }
2245
2246 if (archive_entry_copy_gname_l(entry,
2247 header->gname, sizeof(header->gname), tar->sconv) != 0) {
2248 err = set_conversion_failed_error(a, tar->sconv, "Gname");
2249 if (err == ARCHIVE_FATAL)
2250 return (err);
2251 }
2252
2253 /* Parse out device numbers only for char and block specials */
2254 if (header->typeflag[0] == '3' || header->typeflag[0] == '4') {
2255 archive_entry_set_rdevmajor(entry, (dev_t)
2256 tar_atol(header->rdevmajor, sizeof(header->rdevmajor)));
2257 archive_entry_set_rdevminor(entry, (dev_t)
2258 tar_atol(header->rdevminor, sizeof(header->rdevminor)));
2259 } else
2260 archive_entry_set_rdev(entry, 0);
2261
2262 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining);
2263
2264 /* Grab GNU-specific fields. */
2265 t = tar_atol(header->atime, sizeof(header->atime));
2266 if (t > 0)
2267 archive_entry_set_atime(entry, t, 0);
2268 t = tar_atol(header->ctime, sizeof(header->ctime));
2269 if (t > 0)
2270 archive_entry_set_ctime(entry, t, 0);
2271
2272 if (header->realsize[0] != 0) {
2273 tar->realsize
2274 = tar_atol(header->realsize, sizeof(header->realsize));
2275 archive_entry_set_size(entry, tar->realsize);
2276 tar->realsize_override = 1;
2277 }
2278
2279 if (header->sparse[0].offset[0] != 0) {
2280 if (gnu_sparse_old_read(a, tar, header, unconsumed)
2281 != ARCHIVE_OK)
2282 return (ARCHIVE_FATAL);
2283 } else {
2284 if (header->isextended[0] != 0) {
2285 /* XXX WTF? XXX */
2286 }
2287 }
2288
2289 return (err);
2290 }
2291
2292 static int
gnu_add_sparse_entry(struct archive_read * a,struct tar * tar,int64_t offset,int64_t remaining)2293 gnu_add_sparse_entry(struct archive_read *a, struct tar *tar,
2294 int64_t offset, int64_t remaining)
2295 {
2296 struct sparse_block *p;
2297
2298 p = (struct sparse_block *)calloc(1, sizeof(*p));
2299 if (p == NULL) {
2300 archive_set_error(&a->archive, ENOMEM, "Out of memory");
2301 return (ARCHIVE_FATAL);
2302 }
2303 if (tar->sparse_last != NULL)
2304 tar->sparse_last->next = p;
2305 else
2306 tar->sparse_list = p;
2307 tar->sparse_last = p;
2308 if (remaining < 0 || offset < 0 || offset > INT64_MAX - remaining) {
2309 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Malformed sparse map data");
2310 return (ARCHIVE_FATAL);
2311 }
2312 p->offset = offset;
2313 p->remaining = remaining;
2314 return (ARCHIVE_OK);
2315 }
2316
2317 static void
gnu_clear_sparse_list(struct tar * tar)2318 gnu_clear_sparse_list(struct tar *tar)
2319 {
2320 struct sparse_block *p;
2321
2322 while (tar->sparse_list != NULL) {
2323 p = tar->sparse_list;
2324 tar->sparse_list = p->next;
2325 free(p);
2326 }
2327 tar->sparse_last = NULL;
2328 }
2329
2330 /*
2331 * GNU tar old-format sparse data.
2332 *
2333 * GNU old-format sparse data is stored in a fixed-field
2334 * format. Offset/size values are 11-byte octal fields (same
2335 * format as 'size' field in ustart header). These are
2336 * stored in the header, allocating subsequent header blocks
2337 * as needed. Extending the header in this way is a pretty
2338 * severe POSIX violation; this design has earned GNU tar a
2339 * lot of criticism.
2340 */
2341
2342 static int
gnu_sparse_old_read(struct archive_read * a,struct tar * tar,const struct archive_entry_header_gnutar * header,size_t * unconsumed)2343 gnu_sparse_old_read(struct archive_read *a, struct tar *tar,
2344 const struct archive_entry_header_gnutar *header, size_t *unconsumed)
2345 {
2346 ssize_t bytes_read;
2347 const void *data;
2348 struct extended {
2349 struct gnu_sparse sparse[21];
2350 char isextended[1];
2351 char padding[7];
2352 };
2353 const struct extended *ext;
2354
2355 if (gnu_sparse_old_parse(a, tar, header->sparse, 4) != ARCHIVE_OK)
2356 return (ARCHIVE_FATAL);
2357 if (header->isextended[0] == 0)
2358 return (ARCHIVE_OK);
2359
2360 do {
2361 tar_flush_unconsumed(a, unconsumed);
2362 data = __archive_read_ahead(a, 512, &bytes_read);
2363 if (bytes_read < 0)
2364 return (ARCHIVE_FATAL);
2365 if (bytes_read < 512) {
2366 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
2367 "Truncated tar archive "
2368 "detected while reading sparse file data");
2369 return (ARCHIVE_FATAL);
2370 }
2371 *unconsumed = 512;
2372 ext = (const struct extended *)data;
2373 if (gnu_sparse_old_parse(a, tar, ext->sparse, 21) != ARCHIVE_OK)
2374 return (ARCHIVE_FATAL);
2375 } while (ext->isextended[0] != 0);
2376 if (tar->sparse_list != NULL)
2377 tar->entry_offset = tar->sparse_list->offset;
2378 return (ARCHIVE_OK);
2379 }
2380
2381 static int
gnu_sparse_old_parse(struct archive_read * a,struct tar * tar,const struct gnu_sparse * sparse,int length)2382 gnu_sparse_old_parse(struct archive_read *a, struct tar *tar,
2383 const struct gnu_sparse *sparse, int length)
2384 {
2385 while (length > 0 && sparse->offset[0] != 0) {
2386 if (gnu_add_sparse_entry(a, tar,
2387 tar_atol(sparse->offset, sizeof(sparse->offset)),
2388 tar_atol(sparse->numbytes, sizeof(sparse->numbytes)))
2389 != ARCHIVE_OK)
2390 return (ARCHIVE_FATAL);
2391 sparse++;
2392 length--;
2393 }
2394 return (ARCHIVE_OK);
2395 }
2396
2397 /*
2398 * GNU tar sparse format 0.0
2399 *
2400 * Beginning with GNU tar 1.15, sparse files are stored using
2401 * information in the pax extended header. The GNU tar maintainers
2402 * have gone through a number of variations in the process of working
2403 * out this scheme; fortunately, they're all numbered.
2404 *
2405 * Sparse format 0.0 uses attribute GNU.sparse.numblocks to store the
2406 * number of blocks, and GNU.sparse.offset/GNU.sparse.numbytes to
2407 * store offset/size for each block. The repeated instances of these
2408 * latter fields violate the pax specification (which frowns on
2409 * duplicate keys), so this format was quickly replaced.
2410 */
2411
2412 /*
2413 * GNU tar sparse format 0.1
2414 *
2415 * This version replaced the offset/numbytes attributes with
2416 * a single "map" attribute that stored a list of integers. This
2417 * format had two problems: First, the "map" attribute could be very
2418 * long, which caused problems for some implementations. More
2419 * importantly, the sparse data was lost when extracted by archivers
2420 * that didn't recognize this extension.
2421 */
2422
2423 static int
gnu_sparse_01_parse(struct archive_read * a,struct tar * tar,const char * p)2424 gnu_sparse_01_parse(struct archive_read *a, struct tar *tar, const char *p)
2425 {
2426 const char *e;
2427 int64_t offset = -1, size = -1;
2428
2429 for (;;) {
2430 e = p;
2431 while (*e != '\0' && *e != ',') {
2432 if (*e < '0' || *e > '9')
2433 return (ARCHIVE_WARN);
2434 e++;
2435 }
2436 if (offset < 0) {
2437 offset = tar_atol10(p, e - p);
2438 if (offset < 0)
2439 return (ARCHIVE_WARN);
2440 } else {
2441 size = tar_atol10(p, e - p);
2442 if (size < 0)
2443 return (ARCHIVE_WARN);
2444 if (gnu_add_sparse_entry(a, tar, offset, size)
2445 != ARCHIVE_OK)
2446 return (ARCHIVE_FATAL);
2447 offset = -1;
2448 }
2449 if (*e == '\0')
2450 return (ARCHIVE_OK);
2451 p = e + 1;
2452 }
2453 }
2454
2455 /*
2456 * GNU tar sparse format 1.0
2457 *
2458 * The idea: The offset/size data is stored as a series of base-10
2459 * ASCII numbers prepended to the file data, so that dearchivers that
2460 * don't support this format will extract the block map along with the
2461 * data and a separate post-process can restore the sparseness.
2462 *
2463 * Unfortunately, GNU tar 1.16 had a bug that added unnecessary
2464 * padding to the body of the file when using this format. GNU tar
2465 * 1.17 corrected this bug without bumping the version number, so
2466 * it's not possible to support both variants. This code supports
2467 * the later variant at the expense of not supporting the former.
2468 *
2469 * This variant also replaced GNU.sparse.size with GNU.sparse.realsize
2470 * and introduced the GNU.sparse.major/GNU.sparse.minor attributes.
2471 */
2472
2473 /*
2474 * Read the next line from the input, and parse it as a decimal
2475 * integer followed by '\n'. Returns positive integer value or
2476 * negative on error.
2477 */
2478 static int64_t
gnu_sparse_10_atol(struct archive_read * a,struct tar * tar,int64_t * remaining,size_t * unconsumed)2479 gnu_sparse_10_atol(struct archive_read *a, struct tar *tar,
2480 int64_t *remaining, size_t *unconsumed)
2481 {
2482 int64_t l, limit, last_digit_limit;
2483 const char *p;
2484 ssize_t bytes_read;
2485 int base, digit;
2486
2487 base = 10;
2488 limit = INT64_MAX / base;
2489 last_digit_limit = INT64_MAX % base;
2490
2491 /*
2492 * Skip any lines starting with '#'; GNU tar specs
2493 * don't require this, but they should.
2494 */
2495 do {
2496 bytes_read = readline(a, tar, &p,
2497 (ssize_t)tar_min(*remaining, 100), unconsumed);
2498 if (bytes_read <= 0)
2499 return (ARCHIVE_FATAL);
2500 *remaining -= bytes_read;
2501 } while (p[0] == '#');
2502
2503 l = 0;
2504 while (bytes_read > 0) {
2505 if (*p == '\n')
2506 return (l);
2507 if (*p < '0' || *p >= '0' + base)
2508 return (ARCHIVE_WARN);
2509 digit = *p - '0';
2510 if (l > limit || (l == limit && digit > last_digit_limit))
2511 l = INT64_MAX; /* Truncate on overflow. */
2512 else
2513 l = (l * base) + digit;
2514 p++;
2515 bytes_read--;
2516 }
2517 /* TODO: Error message. */
2518 return (ARCHIVE_WARN);
2519 }
2520
2521 /*
2522 * Returns length (in bytes) of the sparse data description
2523 * that was read.
2524 */
2525 static ssize_t
gnu_sparse_10_read(struct archive_read * a,struct tar * tar,size_t * unconsumed)2526 gnu_sparse_10_read(struct archive_read *a, struct tar *tar, size_t *unconsumed)
2527 {
2528 ssize_t bytes_read;
2529 int entries;
2530 int64_t offset, size, to_skip, remaining;
2531
2532 /* Clear out the existing sparse list. */
2533 gnu_clear_sparse_list(tar);
2534
2535 remaining = tar->entry_bytes_remaining;
2536
2537 /* Parse entries. */
2538 entries = (int)gnu_sparse_10_atol(a, tar, &remaining, unconsumed);
2539 if (entries < 0)
2540 return (ARCHIVE_FATAL);
2541 /* Parse the individual entries. */
2542 while (entries-- > 0) {
2543 /* Parse offset/size */
2544 offset = gnu_sparse_10_atol(a, tar, &remaining, unconsumed);
2545 if (offset < 0)
2546 return (ARCHIVE_FATAL);
2547 size = gnu_sparse_10_atol(a, tar, &remaining, unconsumed);
2548 if (size < 0)
2549 return (ARCHIVE_FATAL);
2550 /* Add a new sparse entry. */
2551 if (gnu_add_sparse_entry(a, tar, offset, size) != ARCHIVE_OK)
2552 return (ARCHIVE_FATAL);
2553 }
2554 /* Skip rest of block... */
2555 tar_flush_unconsumed(a, unconsumed);
2556 bytes_read = (ssize_t)(tar->entry_bytes_remaining - remaining);
2557 to_skip = 0x1ff & -bytes_read;
2558 /* Fail if tar->entry_bytes_remaing would get negative */
2559 if (to_skip > remaining)
2560 return (ARCHIVE_FATAL);
2561 if (to_skip != __archive_read_consume(a, to_skip))
2562 return (ARCHIVE_FATAL);
2563 return ((ssize_t)(bytes_read + to_skip));
2564 }
2565
2566 /*
2567 * Solaris pax extension for a sparse file. This is recorded with the
2568 * data and hole pairs. The way recording sparse information by Solaris'
2569 * pax simply indicates where data and sparse are, so the stored contents
2570 * consist of both data and hole.
2571 */
2572 static int
solaris_sparse_parse(struct archive_read * a,struct tar * tar,struct archive_entry * entry,const char * p)2573 solaris_sparse_parse(struct archive_read *a, struct tar *tar,
2574 struct archive_entry *entry, const char *p)
2575 {
2576 const char *e;
2577 int64_t start, end;
2578 int hole = 1;
2579
2580 (void)entry; /* UNUSED */
2581
2582 end = 0;
2583 if (*p == ' ')
2584 p++;
2585 else
2586 return (ARCHIVE_WARN);
2587 for (;;) {
2588 e = p;
2589 while (*e != '\0' && *e != ' ') {
2590 if (*e < '0' || *e > '9')
2591 return (ARCHIVE_WARN);
2592 e++;
2593 }
2594 start = end;
2595 end = tar_atol10(p, e - p);
2596 if (end < 0)
2597 return (ARCHIVE_WARN);
2598 if (start < end) {
2599 if (gnu_add_sparse_entry(a, tar, start,
2600 end - start) != ARCHIVE_OK)
2601 return (ARCHIVE_FATAL);
2602 tar->sparse_last->hole = hole;
2603 }
2604 if (*e == '\0')
2605 return (ARCHIVE_OK);
2606 p = e + 1;
2607 hole = hole == 0;
2608 }
2609 }
2610
2611 /*-
2612 * Convert text->integer.
2613 *
2614 * Traditional tar formats (including POSIX) specify base-8 for
2615 * all of the standard numeric fields. This is a significant limitation
2616 * in practice:
2617 * = file size is limited to 8GB
2618 * = rdevmajor and rdevminor are limited to 21 bits
2619 * = uid/gid are limited to 21 bits
2620 *
2621 * There are two workarounds for this:
2622 * = pax extended headers, which use variable-length string fields
2623 * = GNU tar and STAR both allow either base-8 or base-256 in
2624 * most fields. The high bit is set to indicate base-256.
2625 *
2626 * On read, this implementation supports both extensions.
2627 */
2628 static int64_t
tar_atol(const char * p,size_t char_cnt)2629 tar_atol(const char *p, size_t char_cnt)
2630 {
2631 /*
2632 * Technically, GNU tar considers a field to be in base-256
2633 * only if the first byte is 0xff or 0x80.
2634 */
2635 if (*p & 0x80)
2636 return (tar_atol256(p, char_cnt));
2637 return (tar_atol8(p, char_cnt));
2638 }
2639
2640 /*
2641 * Note that this implementation does not (and should not!) obey
2642 * locale settings; you cannot simply substitute strtol here, since
2643 * it does obey locale.
2644 */
2645 static int64_t
tar_atol_base_n(const char * p,size_t char_cnt,int base)2646 tar_atol_base_n(const char *p, size_t char_cnt, int base)
2647 {
2648 int64_t l, maxval, limit, last_digit_limit;
2649 int digit, sign;
2650
2651 maxval = INT64_MAX;
2652 limit = INT64_MAX / base;
2653 last_digit_limit = INT64_MAX % base;
2654
2655 /* the pointer will not be dereferenced if char_cnt is zero
2656 * due to the way the && operator is evaluated.
2657 */
2658 while (char_cnt != 0 && (*p == ' ' || *p == '\t')) {
2659 p++;
2660 char_cnt--;
2661 }
2662
2663 sign = 1;
2664 if (char_cnt != 0 && *p == '-') {
2665 sign = -1;
2666 p++;
2667 char_cnt--;
2668
2669 maxval = INT64_MIN;
2670 limit = -(INT64_MIN / base);
2671 last_digit_limit = -(INT64_MIN % base);
2672 }
2673
2674 l = 0;
2675 if (char_cnt != 0) {
2676 digit = *p - '0';
2677 while (digit >= 0 && digit < base && char_cnt != 0) {
2678 if (l>limit || (l == limit && digit >= last_digit_limit)) {
2679 return maxval; /* Truncate on overflow. */
2680 }
2681 l = (l * base) + digit;
2682 digit = *++p - '0';
2683 char_cnt--;
2684 }
2685 }
2686 return (sign < 0) ? -l : l;
2687 }
2688
2689 static int64_t
tar_atol8(const char * p,size_t char_cnt)2690 tar_atol8(const char *p, size_t char_cnt)
2691 {
2692 return tar_atol_base_n(p, char_cnt, 8);
2693 }
2694
2695 static int64_t
tar_atol10(const char * p,size_t char_cnt)2696 tar_atol10(const char *p, size_t char_cnt)
2697 {
2698 return tar_atol_base_n(p, char_cnt, 10);
2699 }
2700
2701 /*
2702 * Parse a base-256 integer. This is just a variable-length
2703 * twos-complement signed binary value in big-endian order, except
2704 * that the high-order bit is ignored. The values here can be up to
2705 * 12 bytes, so we need to be careful about overflowing 64-bit
2706 * (8-byte) integers.
2707 *
2708 * This code unashamedly assumes that the local machine uses 8-bit
2709 * bytes and twos-complement arithmetic.
2710 */
2711 static int64_t
tar_atol256(const char * _p,size_t char_cnt)2712 tar_atol256(const char *_p, size_t char_cnt)
2713 {
2714 uint64_t l;
2715 const unsigned char *p = (const unsigned char *)_p;
2716 unsigned char c, neg;
2717
2718 /* Extend 7-bit 2s-comp to 8-bit 2s-comp, decide sign. */
2719 c = *p;
2720 if (c & 0x40) {
2721 neg = 0xff;
2722 c |= 0x80;
2723 l = ~ARCHIVE_LITERAL_ULL(0);
2724 } else {
2725 neg = 0;
2726 c &= 0x7f;
2727 l = 0;
2728 }
2729
2730 /* If more than 8 bytes, check that we can ignore
2731 * high-order bits without overflow. */
2732 while (char_cnt > sizeof(int64_t)) {
2733 --char_cnt;
2734 if (c != neg)
2735 return neg ? INT64_MIN : INT64_MAX;
2736 c = *++p;
2737 }
2738
2739 /* c is first byte that fits; if sign mismatch, return overflow */
2740 if ((c ^ neg) & 0x80) {
2741 return neg ? INT64_MIN : INT64_MAX;
2742 }
2743
2744 /* Accumulate remaining bytes. */
2745 while (--char_cnt > 0) {
2746 l = (l << 8) | c;
2747 c = *++p;
2748 }
2749 l = (l << 8) | c;
2750 /* Return signed twos-complement value. */
2751 return (int64_t)(l);
2752 }
2753
2754 /*
2755 * Returns length of line (including trailing newline)
2756 * or negative on error. 'start' argument is updated to
2757 * point to first character of line. This avoids copying
2758 * when possible.
2759 */
2760 static ssize_t
readline(struct archive_read * a,struct tar * tar,const char ** start,ssize_t limit,size_t * unconsumed)2761 readline(struct archive_read *a, struct tar *tar, const char **start,
2762 ssize_t limit, size_t *unconsumed)
2763 {
2764 ssize_t bytes_read;
2765 ssize_t total_size = 0;
2766 const void *t;
2767 const char *s;
2768 void *p;
2769
2770 tar_flush_unconsumed(a, unconsumed);
2771
2772 t = __archive_read_ahead(a, 1, &bytes_read);
2773 if (bytes_read <= 0)
2774 return (ARCHIVE_FATAL);
2775 s = t; /* Start of line? */
2776 p = memchr(t, '\n', bytes_read);
2777 /* If we found '\n' in the read buffer, return pointer to that. */
2778 if (p != NULL) {
2779 bytes_read = 1 + ((const char *)p) - s;
2780 if (bytes_read > limit) {
2781 archive_set_error(&a->archive,
2782 ARCHIVE_ERRNO_FILE_FORMAT,
2783 "Line too long");
2784 return (ARCHIVE_FATAL);
2785 }
2786 *unconsumed = bytes_read;
2787 *start = s;
2788 return (bytes_read);
2789 }
2790 *unconsumed = bytes_read;
2791 /* Otherwise, we need to accumulate in a line buffer. */
2792 for (;;) {
2793 if (total_size + bytes_read > limit) {
2794 archive_set_error(&a->archive,
2795 ARCHIVE_ERRNO_FILE_FORMAT,
2796 "Line too long");
2797 return (ARCHIVE_FATAL);
2798 }
2799 if (archive_string_ensure(&tar->line, total_size + bytes_read) == NULL) {
2800 archive_set_error(&a->archive, ENOMEM,
2801 "Can't allocate working buffer");
2802 return (ARCHIVE_FATAL);
2803 }
2804 memcpy(tar->line.s + total_size, t, bytes_read);
2805 tar_flush_unconsumed(a, unconsumed);
2806 total_size += bytes_read;
2807 /* If we found '\n', clean up and return. */
2808 if (p != NULL) {
2809 *start = tar->line.s;
2810 return (total_size);
2811 }
2812 /* Read some more. */
2813 t = __archive_read_ahead(a, 1, &bytes_read);
2814 if (bytes_read <= 0)
2815 return (ARCHIVE_FATAL);
2816 s = t; /* Start of line? */
2817 p = memchr(t, '\n', bytes_read);
2818 /* If we found '\n', trim the read. */
2819 if (p != NULL) {
2820 bytes_read = 1 + ((const char *)p) - s;
2821 }
2822 *unconsumed = bytes_read;
2823 }
2824 }
2825
2826 /*
2827 * base64_decode - Base64 decode
2828 *
2829 * This accepts most variations of base-64 encoding, including:
2830 * * with or without line breaks
2831 * * with or without the final group padded with '=' or '_' characters
2832 * (The most economical Base-64 variant does not pad the last group and
2833 * omits line breaks; RFC1341 used for MIME requires both.)
2834 */
2835 static char *
base64_decode(const char * s,size_t len,size_t * out_len)2836 base64_decode(const char *s, size_t len, size_t *out_len)
2837 {
2838 static const unsigned char digits[64] = {
2839 'A','B','C','D','E','F','G','H','I','J','K','L','M','N',
2840 'O','P','Q','R','S','T','U','V','W','X','Y','Z','a','b',
2841 'c','d','e','f','g','h','i','j','k','l','m','n','o','p',
2842 'q','r','s','t','u','v','w','x','y','z','0','1','2','3',
2843 '4','5','6','7','8','9','+','/' };
2844 static unsigned char decode_table[128];
2845 char *out, *d;
2846 const unsigned char *src = (const unsigned char *)s;
2847
2848 /* If the decode table is not yet initialized, prepare it. */
2849 if (decode_table[digits[1]] != 1) {
2850 unsigned i;
2851 memset(decode_table, 0xff, sizeof(decode_table));
2852 for (i = 0; i < sizeof(digits); i++)
2853 decode_table[digits[i]] = i;
2854 }
2855
2856 /* Allocate enough space to hold the entire output. */
2857 /* Note that we may not use all of this... */
2858 out = (char *)malloc(len - len / 4 + 1);
2859 if (out == NULL) {
2860 *out_len = 0;
2861 return (NULL);
2862 }
2863 d = out;
2864
2865 while (len > 0) {
2866 /* Collect the next group of (up to) four characters. */
2867 int v = 0;
2868 int group_size = 0;
2869 while (group_size < 4 && len > 0) {
2870 /* '=' or '_' padding indicates final group. */
2871 if (*src == '=' || *src == '_') {
2872 len = 0;
2873 break;
2874 }
2875 /* Skip illegal characters (including line breaks) */
2876 if (*src > 127 || *src < 32
2877 || decode_table[*src] == 0xff) {
2878 len--;
2879 src++;
2880 continue;
2881 }
2882 v <<= 6;
2883 v |= decode_table[*src++];
2884 len --;
2885 group_size++;
2886 }
2887 /* Align a short group properly. */
2888 v <<= 6 * (4 - group_size);
2889 /* Unpack the group we just collected. */
2890 switch (group_size) {
2891 case 4: d[2] = v & 0xff;
2892 /* FALLTHROUGH */
2893 case 3: d[1] = (v >> 8) & 0xff;
2894 /* FALLTHROUGH */
2895 case 2: d[0] = (v >> 16) & 0xff;
2896 break;
2897 case 1: /* this is invalid! */
2898 break;
2899 }
2900 d += group_size * 3 / 4;
2901 }
2902
2903 *out_len = d - out;
2904 return (out);
2905 }
2906
2907 static char *
url_decode(const char * in)2908 url_decode(const char *in)
2909 {
2910 char *out, *d;
2911 const char *s;
2912
2913 out = (char *)malloc(strlen(in) + 1);
2914 if (out == NULL)
2915 return (NULL);
2916 for (s = in, d = out; *s != '\0'; ) {
2917 if (s[0] == '%' && s[1] != '\0' && s[2] != '\0') {
2918 /* Try to convert % escape */
2919 int digit1 = tohex(s[1]);
2920 int digit2 = tohex(s[2]);
2921 if (digit1 >= 0 && digit2 >= 0) {
2922 /* Looks good, consume three chars */
2923 s += 3;
2924 /* Convert output */
2925 *d++ = ((digit1 << 4) | digit2);
2926 continue;
2927 }
2928 /* Else fall through and treat '%' as normal char */
2929 }
2930 *d++ = *s++;
2931 }
2932 *d = '\0';
2933 return (out);
2934 }
2935
2936 static int
tohex(int c)2937 tohex(int c)
2938 {
2939 if (c >= '0' && c <= '9')
2940 return (c - '0');
2941 else if (c >= 'A' && c <= 'F')
2942 return (c - 'A' + 10);
2943 else if (c >= 'a' && c <= 'f')
2944 return (c - 'a' + 10);
2945 else
2946 return (-1);
2947 }
2948