1 /*-
2 * Copyright (c) 2003-2007 Tim Kientzle
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 *
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR
15 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
16 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
17 * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT,
18 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
19 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
20 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
21 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
22 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
23 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
24 */
25
26 #include "archive_platform.h"
27 __FBSDID("$FreeBSD: head/lib/libarchive/archive_read_support_format_tar.c 201161 2009-12-29 05:44:39Z kientzle $");
28
29 #ifdef HAVE_ERRNO_H
30 #include <errno.h>
31 #endif
32 #include <stddef.h>
33 /* #include <stdint.h> */ /* See archive_platform.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 /* Obtain suitable wide-character manipulation functions. */
42 #ifdef HAVE_WCHAR_H
43 #include <wchar.h>
44 #else
45 /* Good enough for equality testing, which is all we need. */
wcscmp(const wchar_t * s1,const wchar_t * s2)46 static int wcscmp(const wchar_t *s1, const wchar_t *s2)
47 {
48 int diff = *s1 - *s2;
49 while (*s1 && diff == 0)
50 diff = (int)*++s1 - (int)*++s2;
51 return diff;
52 }
53 /* Good enough for equality testing, which is all we need. */
wcsncmp(const wchar_t * s1,const wchar_t * s2,size_t n)54 static int wcsncmp(const wchar_t *s1, const wchar_t *s2, size_t n)
55 {
56 int diff = *s1 - *s2;
57 while (*s1 && diff == 0 && n-- > 0)
58 diff = (int)*++s1 - (int)*++s2;
59 return diff;
60 }
wcslen(const wchar_t * s)61 static size_t wcslen(const wchar_t *s)
62 {
63 const wchar_t *p = s;
64 while (*p)
65 p++;
66 return p - s;
67 }
68 #endif
69
70 #include "archive.h"
71 #include "archive_entry.h"
72 #include "archive_private.h"
73 #include "archive_read_private.h"
74
75 #define tar_min(a,b) ((a) < (b) ? (a) : (b))
76
77 /*
78 * Layout of POSIX 'ustar' tar header.
79 */
80 struct archive_entry_header_ustar {
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]; /* "old format" header ends here */
90 char magic[6]; /* For POSIX: "ustar\0" */
91 char version[2]; /* For POSIX: "00" */
92 char uname[32];
93 char gname[32];
94 char rdevmajor[8];
95 char rdevminor[8];
96 char prefix[155];
97 };
98
99 /*
100 * Structure of GNU tar header
101 */
102 struct gnu_sparse {
103 char offset[12];
104 char numbytes[12];
105 };
106
107 struct archive_entry_header_gnutar {
108 char name[100];
109 char mode[8];
110 char uid[8];
111 char gid[8];
112 char size[12];
113 char mtime[12];
114 char checksum[8];
115 char typeflag[1];
116 char linkname[100];
117 char magic[8]; /* "ustar \0" (note blank/blank/null at end) */
118 char uname[32];
119 char gname[32];
120 char rdevmajor[8];
121 char rdevminor[8];
122 char atime[12];
123 char ctime[12];
124 char offset[12];
125 char longnames[4];
126 char unused[1];
127 struct gnu_sparse sparse[4];
128 char isextended[1];
129 char realsize[12];
130 /*
131 * Old GNU format doesn't use POSIX 'prefix' field; they use
132 * the 'L' (longname) entry instead.
133 */
134 };
135
136 /*
137 * Data specific to this format.
138 */
139 struct sparse_block {
140 struct sparse_block *next;
141 off_t offset;
142 off_t remaining;
143 };
144
145 struct tar {
146 struct archive_string acl_text;
147 struct archive_string entry_pathname;
148 /* For "GNU.sparse.name" and other similar path extensions. */
149 struct archive_string entry_pathname_override;
150 struct archive_string entry_linkpath;
151 struct archive_string entry_uname;
152 struct archive_string entry_gname;
153 struct archive_string longlink;
154 struct archive_string longname;
155 struct archive_string pax_header;
156 struct archive_string pax_global;
157 struct archive_string line;
158 int pax_hdrcharset_binary;
159 wchar_t *pax_entry;
160 size_t pax_entry_length;
161 int header_recursion_depth;
162 int64_t entry_bytes_remaining;
163 int64_t entry_offset;
164 int64_t entry_padding;
165 int64_t realsize;
166 struct sparse_block *sparse_list;
167 struct sparse_block *sparse_last;
168 int64_t sparse_offset;
169 int64_t sparse_numbytes;
170 int sparse_gnu_major;
171 int sparse_gnu_minor;
172 char sparse_gnu_pending;
173 };
174
175 static ssize_t UTF8_mbrtowc(wchar_t *pwc, const char *s, size_t n);
176 static int archive_block_is_null(const unsigned char *p);
177 static char *base64_decode(const char *, size_t, size_t *);
178 static void gnu_add_sparse_entry(struct tar *,
179 off_t offset, off_t remaining);
180 static void gnu_clear_sparse_list(struct tar *);
181 static int gnu_sparse_old_read(struct archive_read *, struct tar *,
182 const struct archive_entry_header_gnutar *header);
183 static void gnu_sparse_old_parse(struct tar *,
184 const struct gnu_sparse *sparse, int length);
185 static int gnu_sparse_01_parse(struct tar *, const char *);
186 static ssize_t gnu_sparse_10_read(struct archive_read *, struct tar *);
187 static int header_Solaris_ACL(struct archive_read *, struct tar *,
188 struct archive_entry *, const void *);
189 static int header_common(struct archive_read *, struct tar *,
190 struct archive_entry *, const void *);
191 static int header_old_tar(struct archive_read *, struct tar *,
192 struct archive_entry *, const void *);
193 static int header_pax_extensions(struct archive_read *, struct tar *,
194 struct archive_entry *, const void *);
195 static int header_pax_global(struct archive_read *, struct tar *,
196 struct archive_entry *, const void *h);
197 static int header_longlink(struct archive_read *, struct tar *,
198 struct archive_entry *, const void *h);
199 static int header_longname(struct archive_read *, struct tar *,
200 struct archive_entry *, const void *h);
201 static int header_volume(struct archive_read *, struct tar *,
202 struct archive_entry *, const void *h);
203 static int header_ustar(struct archive_read *, struct tar *,
204 struct archive_entry *, const void *h);
205 static int header_gnutar(struct archive_read *, struct tar *,
206 struct archive_entry *, const void *h);
207 static int archive_read_format_tar_bid(struct archive_read *);
208 static int archive_read_format_tar_cleanup(struct archive_read *);
209 static int archive_read_format_tar_read_data(struct archive_read *a,
210 const void **buff, size_t *size, off_t *offset);
211 static int archive_read_format_tar_skip(struct archive_read *a);
212 static int archive_read_format_tar_read_header(struct archive_read *,
213 struct archive_entry *);
214 static int checksum(struct archive_read *, const void *);
215 static int pax_attribute(struct tar *, struct archive_entry *,
216 char *key, char *value);
217 static int pax_header(struct archive_read *, struct tar *,
218 struct archive_entry *, char *attr);
219 static void pax_time(const char *, int64_t *sec, long *nanos);
220 static ssize_t readline(struct archive_read *, struct tar *, const char **,
221 ssize_t limit);
222 static int read_body_to_string(struct archive_read *, struct tar *,
223 struct archive_string *, const void *h);
224 static int64_t tar_atol(const char *, unsigned);
225 static int64_t tar_atol10(const char *, unsigned);
226 static int64_t tar_atol256(const char *, unsigned);
227 static int64_t tar_atol8(const char *, unsigned);
228 static int tar_read_header(struct archive_read *, struct tar *,
229 struct archive_entry *);
230 static int tohex(int c);
231 static char *url_decode(const char *);
232 static wchar_t *utf8_decode(struct tar *, const char *, size_t length);
233
234 int
archive_read_support_format_gnutar(struct archive * a)235 archive_read_support_format_gnutar(struct archive *a)
236 {
237 return (archive_read_support_format_tar(a));
238 }
239
240
241 int
archive_read_support_format_tar(struct archive * _a)242 archive_read_support_format_tar(struct archive *_a)
243 {
244 struct archive_read *a = (struct archive_read *)_a;
245 struct tar *tar;
246 int r;
247
248 tar = (struct tar *)malloc(sizeof(*tar));
249 if (tar == NULL) {
250 archive_set_error(&a->archive, ENOMEM,
251 "Can't allocate tar data");
252 return (ARCHIVE_FATAL);
253 }
254 memset(tar, 0, sizeof(*tar));
255
256 r = __archive_read_register_format(a, tar, "tar",
257 archive_read_format_tar_bid,
258 NULL,
259 archive_read_format_tar_read_header,
260 archive_read_format_tar_read_data,
261 archive_read_format_tar_skip,
262 archive_read_format_tar_cleanup);
263
264 if (r != ARCHIVE_OK)
265 free(tar);
266 return (ARCHIVE_OK);
267 }
268
269 static int
archive_read_format_tar_cleanup(struct archive_read * a)270 archive_read_format_tar_cleanup(struct archive_read *a)
271 {
272 struct tar *tar;
273
274 tar = (struct tar *)(a->format->data);
275 gnu_clear_sparse_list(tar);
276 archive_string_free(&tar->acl_text);
277 archive_string_free(&tar->entry_pathname);
278 archive_string_free(&tar->entry_pathname_override);
279 archive_string_free(&tar->entry_linkpath);
280 archive_string_free(&tar->entry_uname);
281 archive_string_free(&tar->entry_gname);
282 archive_string_free(&tar->line);
283 archive_string_free(&tar->pax_global);
284 archive_string_free(&tar->pax_header);
285 archive_string_free(&tar->longname);
286 archive_string_free(&tar->longlink);
287 free(tar->pax_entry);
288 free(tar);
289 (a->format->data) = NULL;
290 return (ARCHIVE_OK);
291 }
292
293
294 static int
archive_read_format_tar_bid(struct archive_read * a)295 archive_read_format_tar_bid(struct archive_read *a)
296 {
297 int bid;
298 const void *h;
299 const struct archive_entry_header_ustar *header;
300
301 bid = 0;
302
303 /* Now let's look at the actual header and see if it matches. */
304 h = __archive_read_ahead(a, 512, NULL);
305 if (h == NULL)
306 return (-1);
307
308 /* If it's an end-of-archive mark, we can handle it. */
309 if ((*(const char *)h) == 0
310 && archive_block_is_null((const unsigned char *)h)) {
311 /*
312 * Usually, I bid the number of bits verified, but
313 * in this case, 4096 seems excessive so I picked 10 as
314 * an arbitrary but reasonable-seeming value.
315 */
316 return (10);
317 }
318
319 /* If it's not an end-of-archive mark, it must have a valid checksum.*/
320 if (!checksum(a, h))
321 return (0);
322 bid += 48; /* Checksum is usually 6 octal digits. */
323
324 header = (const struct archive_entry_header_ustar *)h;
325
326 /* Recognize POSIX formats. */
327 if ((memcmp(header->magic, "ustar\0", 6) == 0)
328 &&(memcmp(header->version, "00", 2)==0))
329 bid += 56;
330
331 /* Recognize GNU tar format. */
332 if ((memcmp(header->magic, "ustar ", 6) == 0)
333 &&(memcmp(header->version, " \0", 2)==0))
334 bid += 56;
335
336 /* Type flag must be null, digit or A-Z, a-z. */
337 if (header->typeflag[0] != 0 &&
338 !( header->typeflag[0] >= '0' && header->typeflag[0] <= '9') &&
339 !( header->typeflag[0] >= 'A' && header->typeflag[0] <= 'Z') &&
340 !( header->typeflag[0] >= 'a' && header->typeflag[0] <= 'z') )
341 return (0);
342 bid += 2; /* 6 bits of variation in an 8-bit field leaves 2 bits. */
343
344 /* Sanity check: Look at first byte of mode field. */
345 switch (255 & (unsigned)header->mode[0]) {
346 case 0: case 255:
347 /* Base-256 value: No further verification possible! */
348 break;
349 case ' ': /* Not recommended, but not illegal, either. */
350 break;
351 case '0': case '1': case '2': case '3':
352 case '4': case '5': case '6': case '7':
353 /* Octal Value. */
354 /* TODO: Check format of remainder of this field. */
355 break;
356 default:
357 /* Not a valid mode; bail out here. */
358 return (0);
359 }
360 /* TODO: Sanity test uid/gid/size/mtime/rdevmajor/rdevminor fields. */
361
362 return (bid);
363 }
364
365 /*
366 * The function invoked by archive_read_header(). This
367 * just sets up a few things and then calls the internal
368 * tar_read_header() function below.
369 */
370 static int
archive_read_format_tar_read_header(struct archive_read * a,struct archive_entry * entry)371 archive_read_format_tar_read_header(struct archive_read *a,
372 struct archive_entry *entry)
373 {
374 /*
375 * When converting tar archives to cpio archives, it is
376 * essential that each distinct file have a distinct inode
377 * number. To simplify this, we keep a static count here to
378 * assign fake dev/inode numbers to each tar entry. Note that
379 * pax format archives may overwrite this with something more
380 * useful.
381 *
382 * Ideally, we would track every file read from the archive so
383 * that we could assign the same dev/ino pair to hardlinks,
384 * but the memory required to store a complete lookup table is
385 * probably not worthwhile just to support the relatively
386 * obscure tar->cpio conversion case.
387 */
388 static int default_inode;
389 static int default_dev;
390 struct tar *tar;
391 struct sparse_block *sp;
392 const char *p;
393 int r;
394 size_t l;
395
396 /* Assign default device/inode values. */
397 archive_entry_set_dev(entry, 1 + default_dev); /* Don't use zero. */
398 archive_entry_set_ino(entry, ++default_inode); /* Don't use zero. */
399 /* Limit generated st_ino number to 16 bits. */
400 if (default_inode >= 0xffff) {
401 ++default_dev;
402 default_inode = 0;
403 }
404
405 tar = (struct tar *)(a->format->data);
406 tar->entry_offset = 0;
407 while (tar->sparse_list != NULL) {
408 sp = tar->sparse_list;
409 tar->sparse_list = sp->next;
410 free(sp);
411 }
412 tar->sparse_last = NULL;
413 tar->realsize = -1; /* Mark this as "unset" */
414
415 r = tar_read_header(a, tar, entry);
416
417 /*
418 * "non-sparse" files are really just sparse files with
419 * a single block.
420 */
421 if (tar->sparse_list == NULL)
422 gnu_add_sparse_entry(tar, 0, tar->entry_bytes_remaining);
423
424 if (r == ARCHIVE_OK) {
425 /*
426 * "Regular" entry with trailing '/' is really
427 * directory: This is needed for certain old tar
428 * variants and even for some broken newer ones.
429 */
430 p = archive_entry_pathname(entry);
431 l = strlen(p);
432 if (archive_entry_filetype(entry) == AE_IFREG
433 && p[l-1] == '/')
434 archive_entry_set_filetype(entry, AE_IFDIR);
435 }
436 return (r);
437 }
438
439 static int
archive_read_format_tar_read_data(struct archive_read * a,const void ** buff,size_t * size,off_t * offset)440 archive_read_format_tar_read_data(struct archive_read *a,
441 const void **buff, size_t *size, off_t *offset)
442 {
443 ssize_t bytes_read;
444 struct tar *tar;
445 struct sparse_block *p;
446
447 tar = (struct tar *)(a->format->data);
448
449 if (tar->sparse_gnu_pending) {
450 if (tar->sparse_gnu_major == 1 && tar->sparse_gnu_minor == 0) {
451 tar->sparse_gnu_pending = 0;
452 /* Read initial sparse map. */
453 bytes_read = gnu_sparse_10_read(a, tar);
454 tar->entry_bytes_remaining -= bytes_read;
455 if (bytes_read < 0)
456 return (bytes_read);
457 } else {
458 *size = 0;
459 *offset = 0;
460 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
461 "Unrecognized GNU sparse file format");
462 return (ARCHIVE_WARN);
463 }
464 tar->sparse_gnu_pending = 0;
465 }
466
467 /* Remove exhausted entries from sparse list. */
468 while (tar->sparse_list != NULL &&
469 tar->sparse_list->remaining == 0) {
470 p = tar->sparse_list;
471 tar->sparse_list = p->next;
472 free(p);
473 }
474
475 /* If we're at end of file, return EOF. */
476 if (tar->sparse_list == NULL || tar->entry_bytes_remaining == 0) {
477 if (__archive_read_skip(a, tar->entry_padding) < 0)
478 return (ARCHIVE_FATAL);
479 tar->entry_padding = 0;
480 *buff = NULL;
481 *size = 0;
482 *offset = tar->realsize;
483 return (ARCHIVE_EOF);
484 }
485
486 *buff = __archive_read_ahead(a, 1, &bytes_read);
487 if (bytes_read < 0)
488 return (ARCHIVE_FATAL);
489 if (*buff == NULL) {
490 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
491 "Truncated tar archive");
492 return (ARCHIVE_FATAL);
493 }
494 if (bytes_read > tar->entry_bytes_remaining)
495 bytes_read = tar->entry_bytes_remaining;
496 /* Don't read more than is available in the
497 * current sparse block. */
498 if (tar->sparse_list->remaining < bytes_read)
499 bytes_read = tar->sparse_list->remaining;
500 *size = bytes_read;
501 *offset = tar->sparse_list->offset;
502 tar->sparse_list->remaining -= bytes_read;
503 tar->sparse_list->offset += bytes_read;
504 tar->entry_bytes_remaining -= bytes_read;
505 __archive_read_consume(a, bytes_read);
506 return (ARCHIVE_OK);
507 }
508
509 static int
archive_read_format_tar_skip(struct archive_read * a)510 archive_read_format_tar_skip(struct archive_read *a)
511 {
512 int64_t bytes_skipped;
513 struct tar* tar;
514
515 tar = (struct tar *)(a->format->data);
516
517 /*
518 * Compression layer skip functions are required to either skip the
519 * length requested or fail, so we can rely upon the entire entry
520 * plus padding being skipped.
521 */
522 bytes_skipped = __archive_read_skip(a,
523 tar->entry_bytes_remaining + tar->entry_padding);
524 if (bytes_skipped < 0)
525 return (ARCHIVE_FATAL);
526
527 tar->entry_bytes_remaining = 0;
528 tar->entry_padding = 0;
529
530 /* Free the sparse list. */
531 gnu_clear_sparse_list(tar);
532
533 return (ARCHIVE_OK);
534 }
535
536 /*
537 * This function recursively interprets all of the headers associated
538 * with a single entry.
539 */
540 static int
tar_read_header(struct archive_read * a,struct tar * tar,struct archive_entry * entry)541 tar_read_header(struct archive_read *a, struct tar *tar,
542 struct archive_entry *entry)
543 {
544 ssize_t bytes;
545 int err;
546 const void *h;
547 const struct archive_entry_header_ustar *header;
548
549 /* Read 512-byte header record */
550 h = __archive_read_ahead(a, 512, &bytes);
551 if (bytes < 0)
552 return (bytes);
553 if (bytes < 512) { /* Short read or EOF. */
554 /* Try requesting just one byte and see what happens. */
555 (void)__archive_read_ahead(a, 1, &bytes);
556 if (bytes == 0) {
557 /*
558 * The archive ends at a 512-byte boundary but
559 * without a proper end-of-archive marker.
560 * Yes, there are tar writers that do this;
561 * hold our nose and accept it.
562 */
563 return (ARCHIVE_EOF);
564 }
565 /* Archive ends with a partial block; this is bad. */
566 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
567 "Truncated tar archive");
568 return (ARCHIVE_FATAL);
569 }
570 __archive_read_consume(a, 512);
571
572
573 /* Check for end-of-archive mark. */
574 if (((*(const char *)h)==0) && archive_block_is_null((const unsigned char *)h)) {
575 /* Try to consume a second all-null record, as well. */
576 h = __archive_read_ahead(a, 512, NULL);
577 if (h != NULL)
578 __archive_read_consume(a, 512);
579 archive_set_error(&a->archive, 0, NULL);
580 if (a->archive.archive_format_name == NULL) {
581 a->archive.archive_format = ARCHIVE_FORMAT_TAR;
582 a->archive.archive_format_name = "tar";
583 }
584 return (ARCHIVE_EOF);
585 }
586
587 /*
588 * Note: If the checksum fails and we return ARCHIVE_RETRY,
589 * then the client is likely to just retry. This is a very
590 * crude way to search for the next valid header!
591 *
592 * TODO: Improve this by implementing a real header scan.
593 */
594 if (!checksum(a, h)) {
595 archive_set_error(&a->archive, EINVAL, "Damaged tar archive");
596 return (ARCHIVE_RETRY); /* Retryable: Invalid header */
597 }
598
599 if (++tar->header_recursion_depth > 32) {
600 archive_set_error(&a->archive, EINVAL, "Too many special headers");
601 return (ARCHIVE_WARN);
602 }
603
604 /* Determine the format variant. */
605 header = (const struct archive_entry_header_ustar *)h;
606 switch(header->typeflag[0]) {
607 case 'A': /* Solaris tar ACL */
608 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE;
609 a->archive.archive_format_name = "Solaris tar";
610 err = header_Solaris_ACL(a, tar, entry, h);
611 break;
612 case 'g': /* POSIX-standard 'g' header. */
613 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE;
614 a->archive.archive_format_name = "POSIX pax interchange format";
615 err = header_pax_global(a, tar, entry, h);
616 break;
617 case 'K': /* Long link name (GNU tar, others) */
618 err = header_longlink(a, tar, entry, h);
619 break;
620 case 'L': /* Long filename (GNU tar, others) */
621 err = header_longname(a, tar, entry, h);
622 break;
623 case 'V': /* GNU volume header */
624 err = header_volume(a, tar, entry, h);
625 break;
626 case 'X': /* Used by SUN tar; same as 'x'. */
627 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE;
628 a->archive.archive_format_name =
629 "POSIX pax interchange format (Sun variant)";
630 err = header_pax_extensions(a, tar, entry, h);
631 break;
632 case 'x': /* POSIX-standard 'x' header. */
633 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE;
634 a->archive.archive_format_name = "POSIX pax interchange format";
635 err = header_pax_extensions(a, tar, entry, h);
636 break;
637 default:
638 if (memcmp(header->magic, "ustar \0", 8) == 0) {
639 a->archive.archive_format = ARCHIVE_FORMAT_TAR_GNUTAR;
640 a->archive.archive_format_name = "GNU tar format";
641 err = header_gnutar(a, tar, entry, h);
642 } else if (memcmp(header->magic, "ustar", 5) == 0) {
643 if (a->archive.archive_format != ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE) {
644 a->archive.archive_format = ARCHIVE_FORMAT_TAR_USTAR;
645 a->archive.archive_format_name = "POSIX ustar format";
646 }
647 err = header_ustar(a, tar, entry, h);
648 } else {
649 a->archive.archive_format = ARCHIVE_FORMAT_TAR;
650 a->archive.archive_format_name = "tar (non-POSIX)";
651 err = header_old_tar(a, tar, entry, h);
652 }
653 }
654 --tar->header_recursion_depth;
655 /* We return warnings or success as-is. Anything else is fatal. */
656 if (err == ARCHIVE_WARN || err == ARCHIVE_OK)
657 return (err);
658 if (err == ARCHIVE_EOF)
659 /* EOF when recursively reading a header is bad. */
660 archive_set_error(&a->archive, EINVAL, "Damaged tar archive");
661 return (ARCHIVE_FATAL);
662 }
663
664 /*
665 * Return true if block checksum is correct.
666 */
667 static int
checksum(struct archive_read * a,const void * h)668 checksum(struct archive_read *a, const void *h)
669 {
670 const unsigned char *bytes;
671 const struct archive_entry_header_ustar *header;
672 int check, i, sum;
673
674 (void)a; /* UNUSED */
675 bytes = (const unsigned char *)h;
676 header = (const struct archive_entry_header_ustar *)h;
677
678 /*
679 * Test the checksum. Note that POSIX specifies _unsigned_
680 * bytes for this calculation.
681 */
682 sum = tar_atol(header->checksum, sizeof(header->checksum));
683 check = 0;
684 for (i = 0; i < 148; i++)
685 check += (unsigned char)bytes[i];
686 for (; i < 156; i++)
687 check += 32;
688 for (; i < 512; i++)
689 check += (unsigned char)bytes[i];
690 if (sum == check)
691 return (1);
692
693 /*
694 * Repeat test with _signed_ bytes, just in case this archive
695 * was created by an old BSD, Solaris, or HP-UX tar with a
696 * broken checksum calculation.
697 */
698 check = 0;
699 for (i = 0; i < 148; i++)
700 check += (signed char)bytes[i];
701 for (; i < 156; i++)
702 check += 32;
703 for (; i < 512; i++)
704 check += (signed char)bytes[i];
705 if (sum == check)
706 return (1);
707
708 return (0);
709 }
710
711 /*
712 * Return true if this block contains only nulls.
713 */
714 static int
archive_block_is_null(const unsigned char * p)715 archive_block_is_null(const unsigned char *p)
716 {
717 unsigned i;
718
719 for (i = 0; i < 512; i++)
720 if (*p++)
721 return (0);
722 return (1);
723 }
724
725 /*
726 * Interpret 'A' Solaris ACL header
727 */
728 static int
header_Solaris_ACL(struct archive_read * a,struct tar * tar,struct archive_entry * entry,const void * h)729 header_Solaris_ACL(struct archive_read *a, struct tar *tar,
730 struct archive_entry *entry, const void *h)
731 {
732 const struct archive_entry_header_ustar *header;
733 size_t size;
734 int err;
735 int64_t type;
736 char *acl, *p;
737 wchar_t *wp;
738
739 /*
740 * read_body_to_string adds a NUL terminator, but we need a little
741 * more to make sure that we don't overrun acl_text later.
742 */
743 header = (const struct archive_entry_header_ustar *)h;
744 size = tar_atol(header->size, sizeof(header->size));
745 err = read_body_to_string(a, tar, &(tar->acl_text), h);
746 if (err != ARCHIVE_OK)
747 return (err);
748 /* Recursively read next header */
749 err = tar_read_header(a, tar, entry);
750 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN))
751 return (err);
752
753 /* TODO: Examine the first characters to see if this
754 * is an AIX ACL descriptor. We'll likely never support
755 * them, but it would be polite to recognize and warn when
756 * we do see them. */
757
758 /* Leading octal number indicates ACL type and number of entries. */
759 p = acl = tar->acl_text.s;
760 type = 0;
761 while (*p != '\0' && p < acl + size) {
762 if (*p < '0' || *p > '7') {
763 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
764 "Malformed Solaris ACL attribute (invalid digit)");
765 return(ARCHIVE_WARN);
766 }
767 type <<= 3;
768 type += *p - '0';
769 if (type > 077777777) {
770 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
771 "Malformed Solaris ACL attribute (count too large)");
772 return (ARCHIVE_WARN);
773 }
774 p++;
775 }
776 switch ((int)type & ~0777777) {
777 case 01000000:
778 /* POSIX.1e ACL */
779 break;
780 case 03000000:
781 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
782 "Solaris NFSv4 ACLs not supported");
783 return (ARCHIVE_WARN);
784 default:
785 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
786 "Malformed Solaris ACL attribute (unsupported type %o)",
787 (int)type);
788 return (ARCHIVE_WARN);
789 }
790 p++;
791
792 if (p >= acl + size) {
793 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
794 "Malformed Solaris ACL attribute (body overflow)");
795 return(ARCHIVE_WARN);
796 }
797
798 /* ACL text is null-terminated; find the end. */
799 size -= (p - acl);
800 acl = p;
801
802 while (*p != '\0' && p < acl + size)
803 p++;
804
805 wp = utf8_decode(tar, acl, p - acl);
806 err = __archive_entry_acl_parse_w(entry, wp,
807 ARCHIVE_ENTRY_ACL_TYPE_ACCESS);
808 if (err != ARCHIVE_OK)
809 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
810 "Malformed Solaris ACL attribute (unparsable)");
811 return (err);
812 }
813
814 /*
815 * Interpret 'K' long linkname header.
816 */
817 static int
header_longlink(struct archive_read * a,struct tar * tar,struct archive_entry * entry,const void * h)818 header_longlink(struct archive_read *a, struct tar *tar,
819 struct archive_entry *entry, const void *h)
820 {
821 int err;
822
823 err = read_body_to_string(a, tar, &(tar->longlink), h);
824 if (err != ARCHIVE_OK)
825 return (err);
826 err = tar_read_header(a, tar, entry);
827 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN))
828 return (err);
829 /* Set symlink if symlink already set, else hardlink. */
830 archive_entry_copy_link(entry, tar->longlink.s);
831 return (ARCHIVE_OK);
832 }
833
834 /*
835 * Interpret 'L' long filename header.
836 */
837 static int
header_longname(struct archive_read * a,struct tar * tar,struct archive_entry * entry,const void * h)838 header_longname(struct archive_read *a, struct tar *tar,
839 struct archive_entry *entry, const void *h)
840 {
841 int err;
842
843 err = read_body_to_string(a, tar, &(tar->longname), h);
844 if (err != ARCHIVE_OK)
845 return (err);
846 /* Read and parse "real" header, then override name. */
847 err = tar_read_header(a, tar, entry);
848 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN))
849 return (err);
850 archive_entry_copy_pathname(entry, tar->longname.s);
851 return (ARCHIVE_OK);
852 }
853
854
855 /*
856 * Interpret 'V' GNU tar volume header.
857 */
858 static int
header_volume(struct archive_read * a,struct tar * tar,struct archive_entry * entry,const void * h)859 header_volume(struct archive_read *a, struct tar *tar,
860 struct archive_entry *entry, const void *h)
861 {
862 (void)h;
863
864 /* Just skip this and read the next header. */
865 return (tar_read_header(a, tar, entry));
866 }
867
868 /*
869 * Read body of an archive entry into an archive_string object.
870 */
871 static int
read_body_to_string(struct archive_read * a,struct tar * tar,struct archive_string * as,const void * h)872 read_body_to_string(struct archive_read *a, struct tar *tar,
873 struct archive_string *as, const void *h)
874 {
875 off_t size, padded_size;
876 const struct archive_entry_header_ustar *header;
877 const void *src;
878
879 (void)tar; /* UNUSED */
880 header = (const struct archive_entry_header_ustar *)h;
881 size = tar_atol(header->size, sizeof(header->size));
882 if ((size > 1048576) || (size < 0)) {
883 archive_set_error(&a->archive, EINVAL,
884 "Special header too large");
885 return (ARCHIVE_FATAL);
886 }
887
888 /* Fail if we can't make our buffer big enough. */
889 if (archive_string_ensure(as, size+1) == NULL) {
890 archive_set_error(&a->archive, ENOMEM,
891 "No memory");
892 return (ARCHIVE_FATAL);
893 }
894
895 /* Read the body into the string. */
896 padded_size = (size + 511) & ~ 511;
897 src = __archive_read_ahead(a, padded_size, NULL);
898 if (src == NULL)
899 return (ARCHIVE_FATAL);
900 memcpy(as->s, src, size);
901 __archive_read_consume(a, padded_size);
902 as->s[size] = '\0';
903 return (ARCHIVE_OK);
904 }
905
906 /*
907 * Parse out common header elements.
908 *
909 * This would be the same as header_old_tar, except that the
910 * filename is handled slightly differently for old and POSIX
911 * entries (POSIX entries support a 'prefix'). This factoring
912 * allows header_old_tar and header_ustar
913 * to handle filenames differently, while still putting most of the
914 * common parsing into one place.
915 */
916 static int
header_common(struct archive_read * a,struct tar * tar,struct archive_entry * entry,const void * h)917 header_common(struct archive_read *a, struct tar *tar,
918 struct archive_entry *entry, const void *h)
919 {
920 const struct archive_entry_header_ustar *header;
921 char tartype;
922
923 (void)a; /* UNUSED */
924
925 header = (const struct archive_entry_header_ustar *)h;
926 if (header->linkname[0])
927 archive_strncpy(&(tar->entry_linkpath), header->linkname,
928 sizeof(header->linkname));
929 else
930 archive_string_empty(&(tar->entry_linkpath));
931
932 /* Parse out the numeric fields (all are octal) */
933 archive_entry_set_mode(entry, tar_atol(header->mode, sizeof(header->mode)));
934 archive_entry_set_uid(entry, tar_atol(header->uid, sizeof(header->uid)));
935 archive_entry_set_gid(entry, tar_atol(header->gid, sizeof(header->gid)));
936 tar->entry_bytes_remaining = tar_atol(header->size, sizeof(header->size));
937 tar->realsize = tar->entry_bytes_remaining;
938 archive_entry_set_size(entry, tar->entry_bytes_remaining);
939 archive_entry_set_mtime(entry, tar_atol(header->mtime, sizeof(header->mtime)), 0);
940
941 /* Handle the tar type flag appropriately. */
942 tartype = header->typeflag[0];
943
944 switch (tartype) {
945 case '1': /* Hard link */
946 archive_entry_copy_hardlink(entry, tar->entry_linkpath.s);
947 /*
948 * The following may seem odd, but: Technically, tar
949 * does not store the file type for a "hard link"
950 * entry, only the fact that it is a hard link. So, I
951 * leave the type zero normally. But, pax interchange
952 * format allows hard links to have data, which
953 * implies that the underlying entry is a regular
954 * file.
955 */
956 if (archive_entry_size(entry) > 0)
957 archive_entry_set_filetype(entry, AE_IFREG);
958
959 /*
960 * A tricky point: Traditionally, tar readers have
961 * ignored the size field when reading hardlink
962 * entries, and some writers put non-zero sizes even
963 * though the body is empty. POSIX blessed this
964 * convention in the 1988 standard, but broke with
965 * this tradition in 2001 by permitting hardlink
966 * entries to store valid bodies in pax interchange
967 * format, but not in ustar format. Since there is no
968 * hard and fast way to distinguish pax interchange
969 * from earlier archives (the 'x' and 'g' entries are
970 * optional, after all), we need a heuristic.
971 */
972 if (archive_entry_size(entry) == 0) {
973 /* If the size is already zero, we're done. */
974 } else if (a->archive.archive_format
975 == ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE) {
976 /* Definitely pax extended; must obey hardlink size. */
977 } else if (a->archive.archive_format == ARCHIVE_FORMAT_TAR
978 || a->archive.archive_format == ARCHIVE_FORMAT_TAR_GNUTAR)
979 {
980 /* Old-style or GNU tar: we must ignore the size. */
981 archive_entry_set_size(entry, 0);
982 tar->entry_bytes_remaining = 0;
983 } else if (archive_read_format_tar_bid(a) > 50) {
984 /*
985 * We don't know if it's pax: If the bid
986 * function sees a valid ustar header
987 * immediately following, then let's ignore
988 * the hardlink size.
989 */
990 archive_entry_set_size(entry, 0);
991 tar->entry_bytes_remaining = 0;
992 }
993 /*
994 * TODO: There are still two cases I'd like to handle:
995 * = a ustar non-pax archive with a hardlink entry at
996 * end-of-archive. (Look for block of nulls following?)
997 * = a pax archive that has not seen any pax headers
998 * and has an entry which is a hardlink entry storing
999 * a body containing an uncompressed tar archive.
1000 * The first is worth addressing; I don't see any reliable
1001 * way to deal with the second possibility.
1002 */
1003 break;
1004 case '2': /* Symlink */
1005 archive_entry_set_filetype(entry, AE_IFLNK);
1006 archive_entry_set_size(entry, 0);
1007 tar->entry_bytes_remaining = 0;
1008 archive_entry_copy_symlink(entry, tar->entry_linkpath.s);
1009 break;
1010 case '3': /* Character device */
1011 archive_entry_set_filetype(entry, AE_IFCHR);
1012 archive_entry_set_size(entry, 0);
1013 tar->entry_bytes_remaining = 0;
1014 break;
1015 case '4': /* Block device */
1016 archive_entry_set_filetype(entry, AE_IFBLK);
1017 archive_entry_set_size(entry, 0);
1018 tar->entry_bytes_remaining = 0;
1019 break;
1020 case '5': /* Dir */
1021 archive_entry_set_filetype(entry, AE_IFDIR);
1022 archive_entry_set_size(entry, 0);
1023 tar->entry_bytes_remaining = 0;
1024 break;
1025 case '6': /* FIFO device */
1026 archive_entry_set_filetype(entry, AE_IFIFO);
1027 archive_entry_set_size(entry, 0);
1028 tar->entry_bytes_remaining = 0;
1029 break;
1030 case 'D': /* GNU incremental directory type */
1031 /*
1032 * No special handling is actually required here.
1033 * It might be nice someday to preprocess the file list and
1034 * provide it to the client, though.
1035 */
1036 archive_entry_set_filetype(entry, AE_IFDIR);
1037 break;
1038 case 'M': /* GNU "Multi-volume" (remainder of file from last archive)*/
1039 /*
1040 * As far as I can tell, this is just like a regular file
1041 * entry, except that the contents should be _appended_ to
1042 * the indicated file at the indicated offset. This may
1043 * require some API work to fully support.
1044 */
1045 break;
1046 case 'N': /* Old GNU "long filename" entry. */
1047 /* The body of this entry is a script for renaming
1048 * previously-extracted entries. Ugh. It will never
1049 * be supported by libarchive. */
1050 archive_entry_set_filetype(entry, AE_IFREG);
1051 break;
1052 case 'S': /* GNU sparse files */
1053 /*
1054 * Sparse files are really just regular files with
1055 * sparse information in the extended area.
1056 */
1057 /* FALLTHROUGH */
1058 default: /* Regular file and non-standard types */
1059 /*
1060 * Per POSIX: non-recognized types should always be
1061 * treated as regular files.
1062 */
1063 archive_entry_set_filetype(entry, AE_IFREG);
1064 break;
1065 }
1066 return (0);
1067 }
1068
1069 /*
1070 * Parse out header elements for "old-style" tar archives.
1071 */
1072 static int
header_old_tar(struct archive_read * a,struct tar * tar,struct archive_entry * entry,const void * h)1073 header_old_tar(struct archive_read *a, struct tar *tar,
1074 struct archive_entry *entry, const void *h)
1075 {
1076 const struct archive_entry_header_ustar *header;
1077
1078 /* Copy filename over (to ensure null termination). */
1079 header = (const struct archive_entry_header_ustar *)h;
1080 archive_strncpy(&(tar->entry_pathname), header->name, sizeof(header->name));
1081 archive_entry_copy_pathname(entry, tar->entry_pathname.s);
1082
1083 /* Grab rest of common fields */
1084 header_common(a, tar, entry, h);
1085
1086 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining);
1087 return (0);
1088 }
1089
1090 /*
1091 * Parse a file header for a pax extended archive entry.
1092 */
1093 static int
header_pax_global(struct archive_read * a,struct tar * tar,struct archive_entry * entry,const void * h)1094 header_pax_global(struct archive_read *a, struct tar *tar,
1095 struct archive_entry *entry, const void *h)
1096 {
1097 int err;
1098
1099 err = read_body_to_string(a, tar, &(tar->pax_global), h);
1100 if (err != ARCHIVE_OK)
1101 return (err);
1102 err = tar_read_header(a, tar, entry);
1103 return (err);
1104 }
1105
1106 static int
header_pax_extensions(struct archive_read * a,struct tar * tar,struct archive_entry * entry,const void * h)1107 header_pax_extensions(struct archive_read *a, struct tar *tar,
1108 struct archive_entry *entry, const void *h)
1109 {
1110 int err, err2;
1111
1112 err = read_body_to_string(a, tar, &(tar->pax_header), h);
1113 if (err != ARCHIVE_OK)
1114 return (err);
1115
1116 /* Parse the next header. */
1117 err = tar_read_header(a, tar, entry);
1118 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN))
1119 return (err);
1120
1121 /*
1122 * TODO: Parse global/default options into 'entry' struct here
1123 * before handling file-specific options.
1124 *
1125 * This design (parse standard header, then overwrite with pax
1126 * extended attribute data) usually works well, but isn't ideal;
1127 * it would be better to parse the pax extended attributes first
1128 * and then skip any fields in the standard header that were
1129 * defined in the pax header.
1130 */
1131 err2 = pax_header(a, tar, entry, tar->pax_header.s);
1132 err = err_combine(err, err2);
1133 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining);
1134 return (err);
1135 }
1136
1137
1138 /*
1139 * Parse a file header for a Posix "ustar" archive entry. This also
1140 * handles "pax" or "extended ustar" entries.
1141 */
1142 static int
header_ustar(struct archive_read * a,struct tar * tar,struct archive_entry * entry,const void * h)1143 header_ustar(struct archive_read *a, struct tar *tar,
1144 struct archive_entry *entry, const void *h)
1145 {
1146 const struct archive_entry_header_ustar *header;
1147 struct archive_string *as;
1148
1149 header = (const struct archive_entry_header_ustar *)h;
1150
1151 /* Copy name into an internal buffer to ensure null-termination. */
1152 as = &(tar->entry_pathname);
1153 if (header->prefix[0]) {
1154 archive_strncpy(as, header->prefix, sizeof(header->prefix));
1155 if (as->s[archive_strlen(as) - 1] != '/')
1156 archive_strappend_char(as, '/');
1157 archive_strncat(as, header->name, sizeof(header->name));
1158 } else
1159 archive_strncpy(as, header->name, sizeof(header->name));
1160
1161 archive_entry_copy_pathname(entry, as->s);
1162
1163 /* Handle rest of common fields. */
1164 header_common(a, tar, entry, h);
1165
1166 /* Handle POSIX ustar fields. */
1167 archive_strncpy(&(tar->entry_uname), header->uname,
1168 sizeof(header->uname));
1169 archive_entry_copy_uname(entry, tar->entry_uname.s);
1170
1171 archive_strncpy(&(tar->entry_gname), header->gname,
1172 sizeof(header->gname));
1173 archive_entry_copy_gname(entry, tar->entry_gname.s);
1174
1175 /* Parse out device numbers only for char and block specials. */
1176 if (header->typeflag[0] == '3' || header->typeflag[0] == '4') {
1177 archive_entry_set_rdevmajor(entry,
1178 tar_atol(header->rdevmajor, sizeof(header->rdevmajor)));
1179 archive_entry_set_rdevminor(entry,
1180 tar_atol(header->rdevminor, sizeof(header->rdevminor)));
1181 }
1182
1183 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining);
1184
1185 return (0);
1186 }
1187
1188
1189 /*
1190 * Parse the pax extended attributes record.
1191 *
1192 * Returns non-zero if there's an error in the data.
1193 */
1194 static int
pax_header(struct archive_read * a,struct tar * tar,struct archive_entry * entry,char * attr)1195 pax_header(struct archive_read *a, struct tar *tar,
1196 struct archive_entry *entry, char *attr)
1197 {
1198 size_t attr_length, l, line_length;
1199 char *p;
1200 char *key, *value;
1201 int err, err2;
1202
1203 attr_length = strlen(attr);
1204 tar->pax_hdrcharset_binary = 0;
1205 archive_string_empty(&(tar->entry_gname));
1206 archive_string_empty(&(tar->entry_linkpath));
1207 archive_string_empty(&(tar->entry_pathname));
1208 archive_string_empty(&(tar->entry_pathname_override));
1209 archive_string_empty(&(tar->entry_uname));
1210 err = ARCHIVE_OK;
1211 while (attr_length > 0) {
1212 /* Parse decimal length field at start of line. */
1213 line_length = 0;
1214 l = attr_length;
1215 p = attr; /* Record start of line. */
1216 while (l>0) {
1217 if (*p == ' ') {
1218 p++;
1219 l--;
1220 break;
1221 }
1222 if (*p < '0' || *p > '9') {
1223 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1224 "Ignoring malformed pax extended attributes");
1225 return (ARCHIVE_WARN);
1226 }
1227 line_length *= 10;
1228 line_length += *p - '0';
1229 if (line_length > 999999) {
1230 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1231 "Rejecting pax extended attribute > 1MB");
1232 return (ARCHIVE_WARN);
1233 }
1234 p++;
1235 l--;
1236 }
1237
1238 /*
1239 * Parsed length must be no bigger than available data,
1240 * at least 1, and the last character of the line must
1241 * be '\n'.
1242 */
1243 if (line_length > attr_length
1244 || line_length < 1
1245 || attr[line_length - 1] != '\n')
1246 {
1247 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1248 "Ignoring malformed pax extended attribute");
1249 return (ARCHIVE_WARN);
1250 }
1251
1252 /* Null-terminate the line. */
1253 attr[line_length - 1] = '\0';
1254
1255 /* Find end of key and null terminate it. */
1256 key = p;
1257 if (key[0] == '=')
1258 return (-1);
1259 while (*p && *p != '=')
1260 ++p;
1261 if (*p == '\0') {
1262 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1263 "Invalid pax extended attributes");
1264 return (ARCHIVE_WARN);
1265 }
1266 *p = '\0';
1267
1268 /* Identify null-terminated 'value' portion. */
1269 value = p + 1;
1270
1271 /* Identify this attribute and set it in the entry. */
1272 err2 = pax_attribute(tar, entry, key, value);
1273 err = err_combine(err, err2);
1274
1275 /* Skip to next line */
1276 attr += line_length;
1277 attr_length -= line_length;
1278 }
1279 if (archive_strlen(&(tar->entry_gname)) > 0) {
1280 value = tar->entry_gname.s;
1281 if (tar->pax_hdrcharset_binary)
1282 archive_entry_copy_gname(entry, value);
1283 else {
1284 if (!archive_entry_update_gname_utf8(entry, value)) {
1285 err = ARCHIVE_WARN;
1286 archive_set_error(&a->archive,
1287 ARCHIVE_ERRNO_FILE_FORMAT,
1288 "Gname in pax header can't "
1289 "be converted to current locale.");
1290 }
1291 }
1292 }
1293 if (archive_strlen(&(tar->entry_linkpath)) > 0) {
1294 value = tar->entry_linkpath.s;
1295 if (tar->pax_hdrcharset_binary)
1296 archive_entry_copy_link(entry, value);
1297 else {
1298 if (!archive_entry_update_link_utf8(entry, value)) {
1299 err = ARCHIVE_WARN;
1300 archive_set_error(&a->archive,
1301 ARCHIVE_ERRNO_FILE_FORMAT,
1302 "Linkname in pax header can't "
1303 "be converted to current locale.");
1304 }
1305 }
1306 }
1307 /*
1308 * Some extensions (such as the GNU sparse file extensions)
1309 * deliberately store a synthetic name under the regular 'path'
1310 * attribute and the real file name under a different attribute.
1311 * Since we're supposed to not care about the order, we
1312 * have no choice but to store all of the various filenames
1313 * we find and figure it all out afterwards. This is the
1314 * figuring out part.
1315 */
1316 value = NULL;
1317 if (archive_strlen(&(tar->entry_pathname_override)) > 0)
1318 value = tar->entry_pathname_override.s;
1319 else if (archive_strlen(&(tar->entry_pathname)) > 0)
1320 value = tar->entry_pathname.s;
1321 if (value != NULL) {
1322 if (tar->pax_hdrcharset_binary)
1323 archive_entry_copy_pathname(entry, value);
1324 else {
1325 if (!archive_entry_update_pathname_utf8(entry, value)) {
1326 err = ARCHIVE_WARN;
1327 archive_set_error(&a->archive,
1328 ARCHIVE_ERRNO_FILE_FORMAT,
1329 "Pathname in pax header can't be "
1330 "converted to current locale.");
1331 }
1332 }
1333 }
1334 if (archive_strlen(&(tar->entry_uname)) > 0) {
1335 value = tar->entry_uname.s;
1336 if (tar->pax_hdrcharset_binary)
1337 archive_entry_copy_uname(entry, value);
1338 else {
1339 if (!archive_entry_update_uname_utf8(entry, value)) {
1340 err = ARCHIVE_WARN;
1341 archive_set_error(&a->archive,
1342 ARCHIVE_ERRNO_FILE_FORMAT,
1343 "Uname in pax header can't "
1344 "be converted to current locale.");
1345 }
1346 }
1347 }
1348 return (err);
1349 }
1350
1351 static int
pax_attribute_xattr(struct archive_entry * entry,char * name,char * value)1352 pax_attribute_xattr(struct archive_entry *entry,
1353 char *name, char *value)
1354 {
1355 char *name_decoded;
1356 void *value_decoded;
1357 size_t value_len;
1358
1359 if (strlen(name) < 18 || (strncmp(name, "LIBARCHIVE.xattr.", 17)) != 0)
1360 return 3;
1361
1362 name += 17;
1363
1364 /* URL-decode name */
1365 name_decoded = url_decode(name);
1366 if (name_decoded == NULL)
1367 return 2;
1368
1369 /* Base-64 decode value */
1370 value_decoded = base64_decode(value, strlen(value), &value_len);
1371 if (value_decoded == NULL) {
1372 free(name_decoded);
1373 return 1;
1374 }
1375
1376 archive_entry_xattr_add_entry(entry, name_decoded,
1377 value_decoded, value_len);
1378
1379 free(name_decoded);
1380 free(value_decoded);
1381 return 0;
1382 }
1383
1384 /*
1385 * Parse a single key=value attribute. key/value pointers are
1386 * assumed to point into reasonably long-lived storage.
1387 *
1388 * Note that POSIX reserves all-lowercase keywords. Vendor-specific
1389 * extensions should always have keywords of the form "VENDOR.attribute"
1390 * In particular, it's quite feasible to support many different
1391 * vendor extensions here. I'm using "LIBARCHIVE" for extensions
1392 * unique to this library.
1393 *
1394 * Investigate other vendor-specific extensions and see if
1395 * any of them look useful.
1396 */
1397 static int
pax_attribute(struct tar * tar,struct archive_entry * entry,char * key,char * value)1398 pax_attribute(struct tar *tar, struct archive_entry *entry,
1399 char *key, char *value)
1400 {
1401 int64_t s;
1402 long n;
1403 wchar_t *wp;
1404
1405 switch (key[0]) {
1406 case 'G':
1407 /* GNU "0.0" sparse pax format. */
1408 if (strcmp(key, "GNU.sparse.numblocks") == 0) {
1409 tar->sparse_offset = -1;
1410 tar->sparse_numbytes = -1;
1411 tar->sparse_gnu_major = 0;
1412 tar->sparse_gnu_minor = 0;
1413 }
1414 if (strcmp(key, "GNU.sparse.offset") == 0) {
1415 tar->sparse_offset = tar_atol10(value, strlen(value));
1416 if (tar->sparse_numbytes != -1) {
1417 gnu_add_sparse_entry(tar,
1418 tar->sparse_offset, tar->sparse_numbytes);
1419 tar->sparse_offset = -1;
1420 tar->sparse_numbytes = -1;
1421 }
1422 }
1423 if (strcmp(key, "GNU.sparse.numbytes") == 0) {
1424 tar->sparse_numbytes = tar_atol10(value, strlen(value));
1425 if (tar->sparse_numbytes != -1) {
1426 gnu_add_sparse_entry(tar,
1427 tar->sparse_offset, tar->sparse_numbytes);
1428 tar->sparse_offset = -1;
1429 tar->sparse_numbytes = -1;
1430 }
1431 }
1432 if (strcmp(key, "GNU.sparse.size") == 0) {
1433 tar->realsize = tar_atol10(value, strlen(value));
1434 archive_entry_set_size(entry, tar->realsize);
1435 }
1436
1437 /* GNU "0.1" sparse pax format. */
1438 if (strcmp(key, "GNU.sparse.map") == 0) {
1439 tar->sparse_gnu_major = 0;
1440 tar->sparse_gnu_minor = 1;
1441 if (gnu_sparse_01_parse(tar, value) != ARCHIVE_OK)
1442 return (ARCHIVE_WARN);
1443 }
1444
1445 /* GNU "1.0" sparse pax format */
1446 if (strcmp(key, "GNU.sparse.major") == 0) {
1447 tar->sparse_gnu_major = tar_atol10(value, strlen(value));
1448 tar->sparse_gnu_pending = 1;
1449 }
1450 if (strcmp(key, "GNU.sparse.minor") == 0) {
1451 tar->sparse_gnu_minor = tar_atol10(value, strlen(value));
1452 tar->sparse_gnu_pending = 1;
1453 }
1454 if (strcmp(key, "GNU.sparse.name") == 0) {
1455 /*
1456 * The real filename; when storing sparse
1457 * files, GNU tar puts a synthesized name into
1458 * the regular 'path' attribute in an attempt
1459 * to limit confusion. ;-)
1460 */
1461 archive_strcpy(&(tar->entry_pathname_override), value);
1462 }
1463 if (strcmp(key, "GNU.sparse.realsize") == 0) {
1464 tar->realsize = tar_atol10(value, strlen(value));
1465 archive_entry_set_size(entry, tar->realsize);
1466 }
1467 break;
1468 case 'L':
1469 /* Our extensions */
1470 /* TODO: Handle arbitrary extended attributes... */
1471 /*
1472 if (strcmp(key, "LIBARCHIVE.xxxxxxx")==0)
1473 archive_entry_set_xxxxxx(entry, value);
1474 */
1475 if (strcmp(key, "LIBARCHIVE.creationtime")==0) {
1476 pax_time(value, &s, &n);
1477 archive_entry_set_birthtime(entry, s, n);
1478 }
1479 if (strncmp(key, "LIBARCHIVE.xattr.", 17)==0)
1480 pax_attribute_xattr(entry, key, value);
1481 break;
1482 case 'S':
1483 /* We support some keys used by the "star" archiver */
1484 if (strcmp(key, "SCHILY.acl.access")==0) {
1485 wp = utf8_decode(tar, value, strlen(value));
1486 /* TODO: if (wp == NULL) */
1487 __archive_entry_acl_parse_w(entry, wp,
1488 ARCHIVE_ENTRY_ACL_TYPE_ACCESS);
1489 } else if (strcmp(key, "SCHILY.acl.default")==0) {
1490 wp = utf8_decode(tar, value, strlen(value));
1491 /* TODO: if (wp == NULL) */
1492 __archive_entry_acl_parse_w(entry, wp,
1493 ARCHIVE_ENTRY_ACL_TYPE_DEFAULT);
1494 } else if (strcmp(key, "SCHILY.devmajor")==0) {
1495 archive_entry_set_rdevmajor(entry,
1496 tar_atol10(value, strlen(value)));
1497 } else if (strcmp(key, "SCHILY.devminor")==0) {
1498 archive_entry_set_rdevminor(entry,
1499 tar_atol10(value, strlen(value)));
1500 } else if (strcmp(key, "SCHILY.fflags")==0) {
1501 archive_entry_copy_fflags_text(entry, value);
1502 } else if (strcmp(key, "SCHILY.dev")==0) {
1503 archive_entry_set_dev(entry,
1504 tar_atol10(value, strlen(value)));
1505 } else if (strcmp(key, "SCHILY.ino")==0) {
1506 archive_entry_set_ino(entry,
1507 tar_atol10(value, strlen(value)));
1508 } else if (strcmp(key, "SCHILY.nlink")==0) {
1509 archive_entry_set_nlink(entry,
1510 tar_atol10(value, strlen(value)));
1511 } else if (strcmp(key, "SCHILY.realsize")==0) {
1512 tar->realsize = tar_atol10(value, strlen(value));
1513 archive_entry_set_size(entry, tar->realsize);
1514 }
1515 break;
1516 case 'a':
1517 if (strcmp(key, "atime")==0) {
1518 pax_time(value, &s, &n);
1519 archive_entry_set_atime(entry, s, n);
1520 }
1521 break;
1522 case 'c':
1523 if (strcmp(key, "ctime")==0) {
1524 pax_time(value, &s, &n);
1525 archive_entry_set_ctime(entry, s, n);
1526 } else if (strcmp(key, "charset")==0) {
1527 /* TODO: Publish charset information in entry. */
1528 } else if (strcmp(key, "comment")==0) {
1529 /* TODO: Publish comment in entry. */
1530 }
1531 break;
1532 case 'g':
1533 if (strcmp(key, "gid")==0) {
1534 archive_entry_set_gid(entry,
1535 tar_atol10(value, strlen(value)));
1536 } else if (strcmp(key, "gname")==0) {
1537 archive_strcpy(&(tar->entry_gname), value);
1538 }
1539 break;
1540 case 'h':
1541 if (strcmp(key, "hdrcharset") == 0) {
1542 if (strcmp(value, "BINARY") == 0)
1543 tar->pax_hdrcharset_binary = 1;
1544 else if (strcmp(value, "ISO-IR 10646 2000 UTF-8") == 0)
1545 tar->pax_hdrcharset_binary = 0;
1546 else {
1547 /* TODO: Warn about unsupported hdrcharset */
1548 }
1549 }
1550 break;
1551 case 'l':
1552 /* pax interchange doesn't distinguish hardlink vs. symlink. */
1553 if (strcmp(key, "linkpath")==0) {
1554 archive_strcpy(&(tar->entry_linkpath), value);
1555 }
1556 break;
1557 case 'm':
1558 if (strcmp(key, "mtime")==0) {
1559 pax_time(value, &s, &n);
1560 archive_entry_set_mtime(entry, s, n);
1561 }
1562 break;
1563 case 'p':
1564 if (strcmp(key, "path")==0) {
1565 archive_strcpy(&(tar->entry_pathname), value);
1566 }
1567 break;
1568 case 'r':
1569 /* POSIX has reserved 'realtime.*' */
1570 break;
1571 case 's':
1572 /* POSIX has reserved 'security.*' */
1573 /* Someday: if (strcmp(key, "security.acl")==0) { ... } */
1574 if (strcmp(key, "size")==0) {
1575 /* "size" is the size of the data in the entry. */
1576 tar->entry_bytes_remaining
1577 = tar_atol10(value, strlen(value));
1578 /*
1579 * But, "size" is not necessarily the size of
1580 * the file on disk; if this is a sparse file,
1581 * the disk size may have already been set from
1582 * GNU.sparse.realsize or GNU.sparse.size or
1583 * an old GNU header field or SCHILY.realsize
1584 * or ....
1585 */
1586 if (tar->realsize < 0) {
1587 archive_entry_set_size(entry,
1588 tar->entry_bytes_remaining);
1589 tar->realsize
1590 = tar->entry_bytes_remaining;
1591 }
1592 }
1593 break;
1594 case 'u':
1595 if (strcmp(key, "uid")==0) {
1596 archive_entry_set_uid(entry,
1597 tar_atol10(value, strlen(value)));
1598 } else if (strcmp(key, "uname")==0) {
1599 archive_strcpy(&(tar->entry_uname), value);
1600 }
1601 break;
1602 }
1603 return (0);
1604 }
1605
1606
1607
1608 /*
1609 * parse a decimal time value, which may include a fractional portion
1610 */
1611 static void
pax_time(const char * p,int64_t * ps,long * pn)1612 pax_time(const char *p, int64_t *ps, long *pn)
1613 {
1614 char digit;
1615 int64_t s;
1616 unsigned long l;
1617 int sign;
1618 int64_t limit, last_digit_limit;
1619
1620 limit = INT64_MAX / 10;
1621 last_digit_limit = INT64_MAX % 10;
1622
1623 s = 0;
1624 sign = 1;
1625 if (*p == '-') {
1626 sign = -1;
1627 p++;
1628 }
1629 while (*p >= '0' && *p <= '9') {
1630 digit = *p - '0';
1631 if (s > limit ||
1632 (s == limit && digit > last_digit_limit)) {
1633 s = INT64_MAX;
1634 break;
1635 }
1636 s = (s * 10) + digit;
1637 ++p;
1638 }
1639
1640 *ps = s * sign;
1641
1642 /* Calculate nanoseconds. */
1643 *pn = 0;
1644
1645 if (*p != '.')
1646 return;
1647
1648 l = 100000000UL;
1649 do {
1650 ++p;
1651 if (*p >= '0' && *p <= '9')
1652 *pn += (*p - '0') * l;
1653 else
1654 break;
1655 } while (l /= 10);
1656 }
1657
1658 /*
1659 * Parse GNU tar header
1660 */
1661 static int
header_gnutar(struct archive_read * a,struct tar * tar,struct archive_entry * entry,const void * h)1662 header_gnutar(struct archive_read *a, struct tar *tar,
1663 struct archive_entry *entry, const void *h)
1664 {
1665 const struct archive_entry_header_gnutar *header;
1666
1667 (void)a;
1668
1669 /*
1670 * GNU header is like POSIX ustar, except 'prefix' is
1671 * replaced with some other fields. This also means the
1672 * filename is stored as in old-style archives.
1673 */
1674
1675 /* Grab fields common to all tar variants. */
1676 header_common(a, tar, entry, h);
1677
1678 /* Copy filename over (to ensure null termination). */
1679 header = (const struct archive_entry_header_gnutar *)h;
1680 archive_strncpy(&(tar->entry_pathname), header->name,
1681 sizeof(header->name));
1682 archive_entry_copy_pathname(entry, tar->entry_pathname.s);
1683
1684 /* Fields common to ustar and GNU */
1685 /* XXX Can the following be factored out since it's common
1686 * to ustar and gnu tar? Is it okay to move it down into
1687 * header_common, perhaps? */
1688 archive_strncpy(&(tar->entry_uname),
1689 header->uname, sizeof(header->uname));
1690 archive_entry_copy_uname(entry, tar->entry_uname.s);
1691
1692 archive_strncpy(&(tar->entry_gname),
1693 header->gname, sizeof(header->gname));
1694 archive_entry_copy_gname(entry, tar->entry_gname.s);
1695
1696 /* Parse out device numbers only for char and block specials */
1697 if (header->typeflag[0] == '3' || header->typeflag[0] == '4') {
1698 archive_entry_set_rdevmajor(entry,
1699 tar_atol(header->rdevmajor, sizeof(header->rdevmajor)));
1700 archive_entry_set_rdevminor(entry,
1701 tar_atol(header->rdevminor, sizeof(header->rdevminor)));
1702 } else
1703 archive_entry_set_rdev(entry, 0);
1704
1705 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining);
1706
1707 /* Grab GNU-specific fields. */
1708 archive_entry_set_atime(entry,
1709 tar_atol(header->atime, sizeof(header->atime)), 0);
1710 archive_entry_set_ctime(entry,
1711 tar_atol(header->ctime, sizeof(header->ctime)), 0);
1712 if (header->realsize[0] != 0) {
1713 tar->realsize
1714 = tar_atol(header->realsize, sizeof(header->realsize));
1715 archive_entry_set_size(entry, tar->realsize);
1716 }
1717
1718 if (header->sparse[0].offset[0] != 0) {
1719 gnu_sparse_old_read(a, tar, header);
1720 } else {
1721 if (header->isextended[0] != 0) {
1722 /* XXX WTF? XXX */
1723 }
1724 }
1725
1726 return (0);
1727 }
1728
1729 static void
gnu_add_sparse_entry(struct tar * tar,off_t offset,off_t remaining)1730 gnu_add_sparse_entry(struct tar *tar, off_t offset, off_t remaining)
1731 {
1732 struct sparse_block *p;
1733
1734 p = (struct sparse_block *)malloc(sizeof(*p));
1735 if (p == NULL)
1736 __archive_errx(1, "Out of memory");
1737 memset(p, 0, sizeof(*p));
1738 if (tar->sparse_last != NULL)
1739 tar->sparse_last->next = p;
1740 else
1741 tar->sparse_list = p;
1742 tar->sparse_last = p;
1743 p->offset = offset;
1744 p->remaining = remaining;
1745 }
1746
1747 static void
gnu_clear_sparse_list(struct tar * tar)1748 gnu_clear_sparse_list(struct tar *tar)
1749 {
1750 struct sparse_block *p;
1751
1752 while (tar->sparse_list != NULL) {
1753 p = tar->sparse_list;
1754 tar->sparse_list = p->next;
1755 free(p);
1756 }
1757 tar->sparse_last = NULL;
1758 }
1759
1760 /*
1761 * GNU tar old-format sparse data.
1762 *
1763 * GNU old-format sparse data is stored in a fixed-field
1764 * format. Offset/size values are 11-byte octal fields (same
1765 * format as 'size' field in ustart header). These are
1766 * stored in the header, allocating subsequent header blocks
1767 * as needed. Extending the header in this way is a pretty
1768 * severe POSIX violation; this design has earned GNU tar a
1769 * lot of criticism.
1770 */
1771
1772 static int
gnu_sparse_old_read(struct archive_read * a,struct tar * tar,const struct archive_entry_header_gnutar * header)1773 gnu_sparse_old_read(struct archive_read *a, struct tar *tar,
1774 const struct archive_entry_header_gnutar *header)
1775 {
1776 ssize_t bytes_read;
1777 const void *data;
1778 struct extended {
1779 struct gnu_sparse sparse[21];
1780 char isextended[1];
1781 char padding[7];
1782 };
1783 const struct extended *ext;
1784
1785 gnu_sparse_old_parse(tar, header->sparse, 4);
1786 if (header->isextended[0] == 0)
1787 return (ARCHIVE_OK);
1788
1789 do {
1790 data = __archive_read_ahead(a, 512, &bytes_read);
1791 if (bytes_read < 0)
1792 return (ARCHIVE_FATAL);
1793 if (bytes_read < 512) {
1794 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1795 "Truncated tar archive "
1796 "detected while reading sparse file data");
1797 return (ARCHIVE_FATAL);
1798 }
1799 __archive_read_consume(a, 512);
1800 ext = (const struct extended *)data;
1801 gnu_sparse_old_parse(tar, ext->sparse, 21);
1802 } while (ext->isextended[0] != 0);
1803 if (tar->sparse_list != NULL)
1804 tar->entry_offset = tar->sparse_list->offset;
1805 return (ARCHIVE_OK);
1806 }
1807
1808 static void
gnu_sparse_old_parse(struct tar * tar,const struct gnu_sparse * sparse,int length)1809 gnu_sparse_old_parse(struct tar *tar,
1810 const struct gnu_sparse *sparse, int length)
1811 {
1812 while (length > 0 && sparse->offset[0] != 0) {
1813 gnu_add_sparse_entry(tar,
1814 tar_atol(sparse->offset, sizeof(sparse->offset)),
1815 tar_atol(sparse->numbytes, sizeof(sparse->numbytes)));
1816 sparse++;
1817 length--;
1818 }
1819 }
1820
1821 /*
1822 * GNU tar sparse format 0.0
1823 *
1824 * Beginning with GNU tar 1.15, sparse files are stored using
1825 * information in the pax extended header. The GNU tar maintainers
1826 * have gone through a number of variations in the process of working
1827 * out this scheme; furtunately, they're all numbered.
1828 *
1829 * Sparse format 0.0 uses attribute GNU.sparse.numblocks to store the
1830 * number of blocks, and GNU.sparse.offset/GNU.sparse.numbytes to
1831 * store offset/size for each block. The repeated instances of these
1832 * latter fields violate the pax specification (which frowns on
1833 * duplicate keys), so this format was quickly replaced.
1834 */
1835
1836 /*
1837 * GNU tar sparse format 0.1
1838 *
1839 * This version replaced the offset/numbytes attributes with
1840 * a single "map" attribute that stored a list of integers. This
1841 * format had two problems: First, the "map" attribute could be very
1842 * long, which caused problems for some implementations. More
1843 * importantly, the sparse data was lost when extracted by archivers
1844 * that didn't recognize this extension.
1845 */
1846
1847 static int
gnu_sparse_01_parse(struct tar * tar,const char * p)1848 gnu_sparse_01_parse(struct tar *tar, const char *p)
1849 {
1850 const char *e;
1851 off_t offset = -1, size = -1;
1852
1853 for (;;) {
1854 e = p;
1855 while (*e != '\0' && *e != ',') {
1856 if (*e < '0' || *e > '9')
1857 return (ARCHIVE_WARN);
1858 e++;
1859 }
1860 if (offset < 0) {
1861 offset = tar_atol10(p, e - p);
1862 if (offset < 0)
1863 return (ARCHIVE_WARN);
1864 } else {
1865 size = tar_atol10(p, e - p);
1866 if (size < 0)
1867 return (ARCHIVE_WARN);
1868 gnu_add_sparse_entry(tar, offset, size);
1869 offset = -1;
1870 }
1871 if (*e == '\0')
1872 return (ARCHIVE_OK);
1873 p = e + 1;
1874 }
1875 }
1876
1877 /*
1878 * GNU tar sparse format 1.0
1879 *
1880 * The idea: The offset/size data is stored as a series of base-10
1881 * ASCII numbers prepended to the file data, so that dearchivers that
1882 * don't support this format will extract the block map along with the
1883 * data and a separate post-process can restore the sparseness.
1884 *
1885 * Unfortunately, GNU tar 1.16 had a bug that added unnecessary
1886 * padding to the body of the file when using this format. GNU tar
1887 * 1.17 corrected this bug without bumping the version number, so
1888 * it's not possible to support both variants. This code supports
1889 * the later variant at the expense of not supporting the former.
1890 *
1891 * This variant also replaced GNU.sparse.size with GNU.sparse.realsize
1892 * and introduced the GNU.sparse.major/GNU.sparse.minor attributes.
1893 */
1894
1895 /*
1896 * Read the next line from the input, and parse it as a decimal
1897 * integer followed by '\n'. Returns positive integer value or
1898 * negative on error.
1899 */
1900 static int64_t
gnu_sparse_10_atol(struct archive_read * a,struct tar * tar,ssize_t * remaining)1901 gnu_sparse_10_atol(struct archive_read *a, struct tar *tar,
1902 ssize_t *remaining)
1903 {
1904 int64_t l, limit, last_digit_limit;
1905 const char *p;
1906 ssize_t bytes_read;
1907 int base, digit;
1908
1909 base = 10;
1910 limit = INT64_MAX / base;
1911 last_digit_limit = INT64_MAX % base;
1912
1913 /*
1914 * Skip any lines starting with '#'; GNU tar specs
1915 * don't require this, but they should.
1916 */
1917 do {
1918 bytes_read = readline(a, tar, &p, tar_min(*remaining, 100));
1919 if (bytes_read <= 0)
1920 return (ARCHIVE_FATAL);
1921 *remaining -= bytes_read;
1922 } while (p[0] == '#');
1923
1924 l = 0;
1925 while (bytes_read > 0) {
1926 if (*p == '\n')
1927 return (l);
1928 if (*p < '0' || *p >= '0' + base)
1929 return (ARCHIVE_WARN);
1930 digit = *p - '0';
1931 if (l > limit || (l == limit && digit > last_digit_limit))
1932 l = INT64_MAX; /* Truncate on overflow. */
1933 else
1934 l = (l * base) + digit;
1935 p++;
1936 bytes_read--;
1937 }
1938 /* TODO: Error message. */
1939 return (ARCHIVE_WARN);
1940 }
1941
1942 /*
1943 * Returns length (in bytes) of the sparse data description
1944 * that was read.
1945 */
1946 static ssize_t
gnu_sparse_10_read(struct archive_read * a,struct tar * tar)1947 gnu_sparse_10_read(struct archive_read *a, struct tar *tar)
1948 {
1949 ssize_t remaining, bytes_read;
1950 int entries;
1951 off_t offset, size, to_skip;
1952
1953 /* Clear out the existing sparse list. */
1954 gnu_clear_sparse_list(tar);
1955
1956 remaining = tar->entry_bytes_remaining;
1957
1958 /* Parse entries. */
1959 entries = gnu_sparse_10_atol(a, tar, &remaining);
1960 if (entries < 0)
1961 return (ARCHIVE_FATAL);
1962 /* Parse the individual entries. */
1963 while (entries-- > 0) {
1964 /* Parse offset/size */
1965 offset = gnu_sparse_10_atol(a, tar, &remaining);
1966 if (offset < 0)
1967 return (ARCHIVE_FATAL);
1968 size = gnu_sparse_10_atol(a, tar, &remaining);
1969 if (size < 0)
1970 return (ARCHIVE_FATAL);
1971 /* Add a new sparse entry. */
1972 gnu_add_sparse_entry(tar, offset, size);
1973 }
1974 /* Skip rest of block... */
1975 bytes_read = tar->entry_bytes_remaining - remaining;
1976 to_skip = 0x1ff & -bytes_read;
1977 if (to_skip != __archive_read_skip(a, to_skip))
1978 return (ARCHIVE_FATAL);
1979 return (bytes_read + to_skip);
1980 }
1981
1982 /*-
1983 * Convert text->integer.
1984 *
1985 * Traditional tar formats (including POSIX) specify base-8 for
1986 * all of the standard numeric fields. This is a significant limitation
1987 * in practice:
1988 * = file size is limited to 8GB
1989 * = rdevmajor and rdevminor are limited to 21 bits
1990 * = uid/gid are limited to 21 bits
1991 *
1992 * There are two workarounds for this:
1993 * = pax extended headers, which use variable-length string fields
1994 * = GNU tar and STAR both allow either base-8 or base-256 in
1995 * most fields. The high bit is set to indicate base-256.
1996 *
1997 * On read, this implementation supports both extensions.
1998 */
1999 static int64_t
tar_atol(const char * p,unsigned char_cnt)2000 tar_atol(const char *p, unsigned char_cnt)
2001 {
2002 /*
2003 * Technically, GNU tar considers a field to be in base-256
2004 * only if the first byte is 0xff or 0x80.
2005 */
2006 if (*p & 0x80)
2007 return (tar_atol256(p, char_cnt));
2008 return (tar_atol8(p, char_cnt));
2009 }
2010
2011 /*
2012 * Note that this implementation does not (and should not!) obey
2013 * locale settings; you cannot simply substitute strtol here, since
2014 * it does obey locale.
2015 */
2016 static int64_t
tar_atol8(const char * p,unsigned char_cnt)2017 tar_atol8(const char *p, unsigned char_cnt)
2018 {
2019 int64_t l, limit, last_digit_limit;
2020 int digit, sign, base;
2021
2022 base = 8;
2023 limit = INT64_MAX / base;
2024 last_digit_limit = INT64_MAX % base;
2025
2026 while (*p == ' ' || *p == '\t')
2027 p++;
2028 if (*p == '-') {
2029 sign = -1;
2030 p++;
2031 } else
2032 sign = 1;
2033
2034 l = 0;
2035 digit = *p - '0';
2036 while (digit >= 0 && digit < base && char_cnt-- > 0) {
2037 if (l>limit || (l == limit && digit > last_digit_limit)) {
2038 l = INT64_MAX; /* Truncate on overflow. */
2039 break;
2040 }
2041 l = (l * base) + digit;
2042 digit = *++p - '0';
2043 }
2044 return (sign < 0) ? -l : l;
2045 }
2046
2047 /*
2048 * Note that this implementation does not (and should not!) obey
2049 * locale settings; you cannot simply substitute strtol here, since
2050 * it does obey locale.
2051 */
2052 static int64_t
tar_atol10(const char * p,unsigned char_cnt)2053 tar_atol10(const char *p, unsigned char_cnt)
2054 {
2055 int64_t l, limit, last_digit_limit;
2056 int base, digit, sign;
2057
2058 base = 10;
2059 limit = INT64_MAX / base;
2060 last_digit_limit = INT64_MAX % base;
2061
2062 while (*p == ' ' || *p == '\t')
2063 p++;
2064 if (*p == '-') {
2065 sign = -1;
2066 p++;
2067 } else
2068 sign = 1;
2069
2070 l = 0;
2071 digit = *p - '0';
2072 while (digit >= 0 && digit < base && char_cnt-- > 0) {
2073 if (l > limit || (l == limit && digit > last_digit_limit)) {
2074 l = INT64_MAX; /* Truncate on overflow. */
2075 break;
2076 }
2077 l = (l * base) + digit;
2078 digit = *++p - '0';
2079 }
2080 return (sign < 0) ? -l : l;
2081 }
2082
2083 /*
2084 * Parse a base-256 integer. This is just a straight signed binary
2085 * value in big-endian order, except that the high-order bit is
2086 * ignored.
2087 */
2088 static int64_t
tar_atol256(const char * _p,unsigned char_cnt)2089 tar_atol256(const char *_p, unsigned char_cnt)
2090 {
2091 int64_t l, upper_limit, lower_limit;
2092 const unsigned char *p = (const unsigned char *)_p;
2093
2094 upper_limit = INT64_MAX / 256;
2095 lower_limit = INT64_MIN / 256;
2096
2097 /* Pad with 1 or 0 bits, depending on sign. */
2098 if ((0x40 & *p) == 0x40)
2099 l = (int64_t)-1;
2100 else
2101 l = 0;
2102 l = (l << 6) | (0x3f & *p++);
2103 while (--char_cnt > 0) {
2104 if (l > upper_limit) {
2105 l = INT64_MAX; /* Truncate on overflow */
2106 break;
2107 } else if (l < lower_limit) {
2108 l = INT64_MIN;
2109 break;
2110 }
2111 l = (l << 8) | (0xff & (int64_t)*p++);
2112 }
2113 return (l);
2114 }
2115
2116 /*
2117 * Returns length of line (including trailing newline)
2118 * or negative on error. 'start' argument is updated to
2119 * point to first character of line. This avoids copying
2120 * when possible.
2121 */
2122 static ssize_t
readline(struct archive_read * a,struct tar * tar,const char ** start,ssize_t limit)2123 readline(struct archive_read *a, struct tar *tar, const char **start,
2124 ssize_t limit)
2125 {
2126 ssize_t bytes_read;
2127 ssize_t total_size = 0;
2128 const void *t;
2129 const char *s;
2130 void *p;
2131
2132 t = __archive_read_ahead(a, 1, &bytes_read);
2133 if (bytes_read <= 0)
2134 return (ARCHIVE_FATAL);
2135 s = t; /* Start of line? */
2136 p = memchr(t, '\n', bytes_read);
2137 /* If we found '\n' in the read buffer, return pointer to that. */
2138 if (p != NULL) {
2139 bytes_read = 1 + ((const char *)p) - s;
2140 if (bytes_read > limit) {
2141 archive_set_error(&a->archive,
2142 ARCHIVE_ERRNO_FILE_FORMAT,
2143 "Line too long");
2144 return (ARCHIVE_FATAL);
2145 }
2146 __archive_read_consume(a, bytes_read);
2147 *start = s;
2148 return (bytes_read);
2149 }
2150 /* Otherwise, we need to accumulate in a line buffer. */
2151 for (;;) {
2152 if (total_size + bytes_read > limit) {
2153 archive_set_error(&a->archive,
2154 ARCHIVE_ERRNO_FILE_FORMAT,
2155 "Line too long");
2156 return (ARCHIVE_FATAL);
2157 }
2158 if (archive_string_ensure(&tar->line, total_size + bytes_read) == NULL) {
2159 archive_set_error(&a->archive, ENOMEM,
2160 "Can't allocate working buffer");
2161 return (ARCHIVE_FATAL);
2162 }
2163 memcpy(tar->line.s + total_size, t, bytes_read);
2164 __archive_read_consume(a, bytes_read);
2165 total_size += bytes_read;
2166 /* If we found '\n', clean up and return. */
2167 if (p != NULL) {
2168 *start = tar->line.s;
2169 return (total_size);
2170 }
2171 /* Read some more. */
2172 t = __archive_read_ahead(a, 1, &bytes_read);
2173 if (bytes_read <= 0)
2174 return (ARCHIVE_FATAL);
2175 s = t; /* Start of line? */
2176 p = memchr(t, '\n', bytes_read);
2177 /* If we found '\n', trim the read. */
2178 if (p != NULL) {
2179 bytes_read = 1 + ((const char *)p) - s;
2180 }
2181 }
2182 }
2183
2184 static wchar_t *
utf8_decode(struct tar * tar,const char * src,size_t length)2185 utf8_decode(struct tar *tar, const char *src, size_t length)
2186 {
2187 wchar_t *dest;
2188 ssize_t n;
2189
2190 /* Ensure pax_entry buffer is big enough. */
2191 if (tar->pax_entry_length <= length) {
2192 wchar_t *old_entry;
2193
2194 if (tar->pax_entry_length <= 0)
2195 tar->pax_entry_length = 1024;
2196 while (tar->pax_entry_length <= length + 1)
2197 tar->pax_entry_length *= 2;
2198
2199 old_entry = tar->pax_entry;
2200 tar->pax_entry = (wchar_t *)realloc(tar->pax_entry,
2201 tar->pax_entry_length * sizeof(wchar_t));
2202 if (tar->pax_entry == NULL) {
2203 free(old_entry);
2204 /* TODO: Handle this error. */
2205 return (NULL);
2206 }
2207 }
2208
2209 dest = tar->pax_entry;
2210 while (length > 0) {
2211 n = UTF8_mbrtowc(dest, src, length);
2212 if (n < 0)
2213 return (NULL);
2214 if (n == 0)
2215 break;
2216 dest++;
2217 src += n;
2218 length -= n;
2219 }
2220 *dest = L'\0';
2221 return (tar->pax_entry);
2222 }
2223
2224 /*
2225 * Copied and simplified from FreeBSD libc/locale.
2226 */
2227 static ssize_t
UTF8_mbrtowc(wchar_t * pwc,const char * s,size_t n)2228 UTF8_mbrtowc(wchar_t *pwc, const char *s, size_t n)
2229 {
2230 int ch, i, len, mask;
2231 unsigned long wch;
2232
2233 if (s == NULL || n == 0 || pwc == NULL)
2234 return (0);
2235
2236 /*
2237 * Determine the number of octets that make up this character from
2238 * the first octet, and a mask that extracts the interesting bits of
2239 * the first octet.
2240 */
2241 ch = (unsigned char)*s;
2242 if ((ch & 0x80) == 0) {
2243 mask = 0x7f;
2244 len = 1;
2245 } else if ((ch & 0xe0) == 0xc0) {
2246 mask = 0x1f;
2247 len = 2;
2248 } else if ((ch & 0xf0) == 0xe0) {
2249 mask = 0x0f;
2250 len = 3;
2251 } else if ((ch & 0xf8) == 0xf0) {
2252 mask = 0x07;
2253 len = 4;
2254 } else {
2255 /* Invalid first byte. */
2256 return (-1);
2257 }
2258
2259 if (n < (size_t)len) {
2260 /* Valid first byte but truncated. */
2261 return (-2);
2262 }
2263
2264 /*
2265 * Decode the octet sequence representing the character in chunks
2266 * of 6 bits, most significant first.
2267 */
2268 wch = (unsigned char)*s++ & mask;
2269 i = len;
2270 while (--i != 0) {
2271 if ((*s & 0xc0) != 0x80) {
2272 /* Invalid intermediate byte; consume one byte and
2273 * emit '?' */
2274 *pwc = '?';
2275 return (1);
2276 }
2277 wch <<= 6;
2278 wch |= *s++ & 0x3f;
2279 }
2280
2281 /* Assign the value to the output; out-of-range values
2282 * just get truncated. */
2283 *pwc = (wchar_t)wch;
2284 #ifdef WCHAR_MAX
2285 /*
2286 * If platform has WCHAR_MAX, we can do something
2287 * more sensible with out-of-range values.
2288 */
2289 if (wch >= WCHAR_MAX)
2290 *pwc = '?';
2291 #endif
2292 /* Return number of bytes input consumed: 0 for end-of-string. */
2293 return (wch == L'\0' ? 0 : len);
2294 }
2295
2296
2297 /*
2298 * base64_decode - Base64 decode
2299 *
2300 * This accepts most variations of base-64 encoding, including:
2301 * * with or without line breaks
2302 * * with or without the final group padded with '=' or '_' characters
2303 * (The most economical Base-64 variant does not pad the last group and
2304 * omits line breaks; RFC1341 used for MIME requires both.)
2305 */
2306 static char *
base64_decode(const char * s,size_t len,size_t * out_len)2307 base64_decode(const char *s, size_t len, size_t *out_len)
2308 {
2309 static const unsigned char digits[64] = {
2310 'A','B','C','D','E','F','G','H','I','J','K','L','M','N',
2311 'O','P','Q','R','S','T','U','V','W','X','Y','Z','a','b',
2312 'c','d','e','f','g','h','i','j','k','l','m','n','o','p',
2313 'q','r','s','t','u','v','w','x','y','z','0','1','2','3',
2314 '4','5','6','7','8','9','+','/' };
2315 static unsigned char decode_table[128];
2316 char *out, *d;
2317 const unsigned char *src = (const unsigned char *)s;
2318
2319 /* If the decode table is not yet initialized, prepare it. */
2320 if (decode_table[digits[1]] != 1) {
2321 unsigned i;
2322 memset(decode_table, 0xff, sizeof(decode_table));
2323 for (i = 0; i < sizeof(digits); i++)
2324 decode_table[digits[i]] = i;
2325 }
2326
2327 /* Allocate enough space to hold the entire output. */
2328 /* Note that we may not use all of this... */
2329 out = (char *)malloc(len - len / 4 + 1);
2330 if (out == NULL) {
2331 *out_len = 0;
2332 return (NULL);
2333 }
2334 d = out;
2335
2336 while (len > 0) {
2337 /* Collect the next group of (up to) four characters. */
2338 int v = 0;
2339 int group_size = 0;
2340 while (group_size < 4 && len > 0) {
2341 /* '=' or '_' padding indicates final group. */
2342 if (*src == '=' || *src == '_') {
2343 len = 0;
2344 break;
2345 }
2346 /* Skip illegal characters (including line breaks) */
2347 if (*src > 127 || *src < 32
2348 || decode_table[*src] == 0xff) {
2349 len--;
2350 src++;
2351 continue;
2352 }
2353 v <<= 6;
2354 v |= decode_table[*src++];
2355 len --;
2356 group_size++;
2357 }
2358 /* Align a short group properly. */
2359 v <<= 6 * (4 - group_size);
2360 /* Unpack the group we just collected. */
2361 switch (group_size) {
2362 case 4: d[2] = v & 0xff;
2363 /* FALLTHROUGH */
2364 case 3: d[1] = (v >> 8) & 0xff;
2365 /* FALLTHROUGH */
2366 case 2: d[0] = (v >> 16) & 0xff;
2367 break;
2368 case 1: /* this is invalid! */
2369 break;
2370 }
2371 d += group_size * 3 / 4;
2372 }
2373
2374 *out_len = d - out;
2375 return (out);
2376 }
2377
2378 static char *
url_decode(const char * in)2379 url_decode(const char *in)
2380 {
2381 char *out, *d;
2382 const char *s;
2383
2384 out = (char *)malloc(strlen(in) + 1);
2385 if (out == NULL)
2386 return (NULL);
2387 for (s = in, d = out; *s != '\0'; ) {
2388 if (s[0] == '%' && s[1] != '\0' && s[2] != '\0') {
2389 /* Try to convert % escape */
2390 int digit1 = tohex(s[1]);
2391 int digit2 = tohex(s[2]);
2392 if (digit1 >= 0 && digit2 >= 0) {
2393 /* Looks good, consume three chars */
2394 s += 3;
2395 /* Convert output */
2396 *d++ = ((digit1 << 4) | digit2);
2397 continue;
2398 }
2399 /* Else fall through and treat '%' as normal char */
2400 }
2401 *d++ = *s++;
2402 }
2403 *d = '\0';
2404 return (out);
2405 }
2406
2407 static int
tohex(int c)2408 tohex(int c)
2409 {
2410 if (c >= '0' && c <= '9')
2411 return (c - '0');
2412 else if (c >= 'A' && c <= 'F')
2413 return (c - 'A' + 10);
2414 else if (c >= 'a' && c <= 'f')
2415 return (c - 'a' + 10);
2416 else
2417 return (-1);
2418 }
2419