1 /* $OpenBSD: ar_subs.c,v 1.33 2009/10/27 23:59:22 deraadt Exp $ */
2 /* $NetBSD: ar_subs.c,v 1.5 1995/03/21 09:07:06 cgd Exp $ */
3
4 /*-
5 * Copyright (c) 2008, 2011, 2012
6 * Thorsten Glaser <tg@mirbsd.org>
7 * Copyright (c) 1992 Keith Muller.
8 * Copyright (c) 1992, 1993
9 * The Regents of the University of California. All rights reserved.
10 *
11 * This code is derived from software contributed to Berkeley by
12 * Keith Muller of the University of California, San Diego.
13 *
14 * Redistribution and use in source and binary forms, with or without
15 * modification, are permitted provided that the following conditions
16 * are met:
17 * 1. Redistributions of source code must retain the above copyright
18 * notice, this list of conditions and the following disclaimer.
19 * 2. Redistributions in binary form must reproduce the above copyright
20 * notice, this list of conditions and the following disclaimer in the
21 * documentation and/or other materials provided with the distribution.
22 * 3. Neither the name of the University nor the names of its contributors
23 * may be used to endorse or promote products derived from this software
24 * without specific prior written permission.
25 *
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * SUCH DAMAGE.
37 */
38
39 #include <sys/param.h>
40 #include <sys/time.h>
41 #include <sys/stat.h>
42 #include <signal.h>
43 #include <string.h>
44 #include <stdio.h>
45 #include <fcntl.h>
46 #include <errno.h>
47 #include <unistd.h>
48 #include <stdlib.h>
49 #include <time.h>
50 #include "pax.h"
51 #include "extern.h"
52 #include "options.h"
53
54 __RCSID("$MirOS: src/bin/pax/ar_subs.c,v 1.14 2012/06/05 18:22:55 tg Exp $");
55
56 static void wr_archive(ARCHD *, int is_app);
57 static int get_arc(void);
58 static int next_head(ARCHD *);
59 extern sigset_t s_mask;
60
61 /*
62 * Routines which control the overall operation modes of pax as specified by
63 * the user: list, append, read ...
64 */
65
66 static char hdbuf[BLKMULT]; /* space for archive header on read */
67 u_long flcnt; /* number of files processed */
68 int ar_do_keepopen = 0; /* see append() below */
69
70 /*
71 * list()
72 * list the contents of an archive which match user supplied pattern(s)
73 * (no pattern matches all).
74 */
75
76 void
list(void)77 list(void)
78 {
79 ARCHD *arcn;
80 int res;
81 ARCHD archd;
82 time_t now;
83
84 arcn = &archd;
85 /*
86 * figure out archive type; pass any format specific options to the
87 * archive option processing routine; call the format init routine. We
88 * also save current time for ls_list() so we do not make a system
89 * call for each file we need to print. If verbose (vflag) start up
90 * the name and group caches.
91 */
92 if ((get_arc() < 0) || ((*frmt->options)() < 0) ||
93 ((*frmt->st_rd)() < 0))
94 return;
95
96 if (vflag && ((uidtb_start() < 0) || (gidtb_start() < 0)))
97 return;
98
99 now = time(NULL);
100
101 /*
102 * step through the archive until the format says it is done
103 */
104 while (next_head(arcn) == 0) {
105 if (arcn->type == PAX_GLL || arcn->type == PAX_GLF) {
106 /*
107 * we need to read, to get the real filename
108 */
109 off_t cnt;
110 if (!(*frmt->rd_data)(arcn, arcn->type == PAX_GLF
111 ? -1 : -2, &cnt))
112 (void)rd_skip(cnt + arcn->pad);
113 continue;
114 }
115
116 /*
117 * check for pattern, and user specified options match.
118 * When all patterns are matched we are done.
119 */
120 if ((res = pat_match(arcn)) < 0)
121 break;
122
123 if ((res == 0) && (sel_chk(arcn) == 0)) {
124 /*
125 * pattern resulted in a selected file
126 */
127 if (pat_sel(arcn) < 0)
128 break;
129
130 /*
131 * modify the name as requested by the user if name
132 * survives modification, do a listing of the file
133 */
134 if ((res = mod_name(arcn)) < 0)
135 break;
136 if (res == 0)
137 ls_list(arcn, now, stdout);
138 }
139
140 /*
141 * skip to next archive format header using values calculated
142 * by the format header read routine
143 */
144 if (rd_skip(arcn->skip + arcn->pad) == 1)
145 break;
146 }
147
148 /*
149 * all done, let format have a chance to cleanup, and make sure that
150 * the patterns supplied by the user were all matched
151 */
152 (void)(*frmt->end_rd)();
153 (void)sigprocmask(SIG_BLOCK, &s_mask, NULL);
154 ar_close();
155 pat_chk();
156 }
157
158 /*
159 * extract()
160 * extract the member(s) of an archive as specified by user supplied
161 * pattern(s) (no patterns extracts all members)
162 */
163
164 void
extract(void)165 extract(void)
166 {
167 ARCHD *arcn;
168 int res;
169 off_t cnt;
170 ARCHD archd;
171 struct stat sb;
172 int fd;
173 time_t now;
174
175 arcn = &archd;
176 /*
177 * figure out archive type; pass any format specific options to the
178 * archive option processing routine; call the format init routine;
179 * start up the directory modification time and access mode database
180 */
181 if ((get_arc() < 0) || ((*frmt->options)() < 0) ||
182 ((*frmt->st_rd)() < 0) || (dir_start() < 0))
183 return;
184
185 /*
186 * When we are doing interactive rename, we store the mapping of names
187 * so we can fix up hard links files later in the archive.
188 */
189 if (iflag && (name_start() < 0))
190 return;
191
192 now = time(NULL);
193
194 /*
195 * step through each entry on the archive until the format read routine
196 * says it is done
197 */
198 while (next_head(arcn) == 0) {
199 if (arcn->type == PAX_GLL || arcn->type == PAX_GLF) {
200 /*
201 * we need to read, to get the real filename
202 */
203 if (!(*frmt->rd_data)(arcn, arcn->type == PAX_GLF
204 ? -1 : -2, &cnt))
205 (void)rd_skip(cnt + arcn->pad);
206 continue;
207 }
208
209 /*
210 * check for pattern, and user specified options match. When
211 * all the patterns are matched we are done
212 */
213 if ((res = pat_match(arcn)) < 0)
214 break;
215
216 if ((res > 0) || (sel_chk(arcn) != 0)) {
217 /*
218 * file is not selected. skip past any file data and
219 * padding and go back for the next archive member
220 */
221 (void)rd_skip(arcn->skip + arcn->pad);
222 continue;
223 }
224
225 /*
226 * with -u or -D only extract when the archive member is newer
227 * than the file with the same name in the file system (no
228 * test of being the same type is required).
229 * NOTE: this test is done BEFORE name modifications as
230 * specified by pax. this operation can be confusing to the
231 * user who might expect the test to be done on an existing
232 * file AFTER the name mod. In honesty the pax spec is probably
233 * flawed in this respect.
234 */
235 if ((uflag || Dflag) && ((lstat(arcn->name, &sb) == 0))) {
236 if (uflag && Dflag) {
237 if ((arcn->sb.st_mtime <= sb.st_mtime) &&
238 (arcn->sb.st_ctime <= sb.st_ctime)) {
239 (void)rd_skip(arcn->skip + arcn->pad);
240 continue;
241 }
242 } else if (Dflag) {
243 if (arcn->sb.st_ctime <= sb.st_ctime) {
244 (void)rd_skip(arcn->skip + arcn->pad);
245 continue;
246 }
247 } else if (arcn->sb.st_mtime <= sb.st_mtime) {
248 (void)rd_skip(arcn->skip + arcn->pad);
249 continue;
250 }
251 }
252
253 /*
254 * this archive member is now been selected. modify the name.
255 */
256 if ((pat_sel(arcn) < 0) || ((res = mod_name(arcn)) < 0))
257 break;
258 if (res > 0) {
259 /*
260 * a bad name mod, skip and purge name from link table
261 */
262 purg_lnk(arcn);
263 (void)rd_skip(arcn->skip + arcn->pad);
264 continue;
265 }
266
267 /*
268 * Non standard -Y and -Z flag. When the existing file is
269 * same age or newer skip
270 */
271 if ((Yflag || Zflag) && ((lstat(arcn->name, &sb) == 0))) {
272 if (Yflag && Zflag) {
273 if ((arcn->sb.st_mtime <= sb.st_mtime) &&
274 (arcn->sb.st_ctime <= sb.st_ctime)) {
275 (void)rd_skip(arcn->skip + arcn->pad);
276 continue;
277 }
278 } else if (Yflag) {
279 if (arcn->sb.st_ctime <= sb.st_ctime) {
280 (void)rd_skip(arcn->skip + arcn->pad);
281 continue;
282 }
283 } else if (arcn->sb.st_mtime <= sb.st_mtime) {
284 (void)rd_skip(arcn->skip + arcn->pad);
285 continue;
286 }
287 }
288
289 if (vflag) {
290 if (vflag > 1)
291 ls_list(arcn, now, listf);
292 else {
293 (void)safe_print(arcn->name, listf);
294 vfpart = 1;
295 }
296 } else if (Vflag) {
297 (void)putc('.', listf);
298 (void)fflush(listf);
299 vfpart = 1;
300 }
301
302 /*
303 * if required, chdir around.
304 */
305 if ((arcn->pat != NULL) && (arcn->pat->chdname != NULL))
306 if (!to_stdout && chdir(arcn->pat->chdname) != 0)
307 syswarn(1, errno, "Cannot chdir to %s",
308 arcn->pat->chdname);
309
310 /*
311 * all ok, extract this member based on type
312 */
313 if ((arcn->type != PAX_REG) && (arcn->type != PAX_CTG)) {
314 /*
315 * process archive members that are not regular files.
316 * throw out padding and any data that might follow the
317 * header (as determined by the format).
318 */
319 if (!to_stdout) {
320 if ((arcn->type == PAX_HLK) || (arcn->type == PAX_HRG)) {
321 res = lnk_creat(arcn, &fd);
322 if (fd != -1)
323 goto extdata;
324 } else
325 res = node_creat(arcn);
326 }
327
328 (void)rd_skip(arcn->skip + arcn->pad);
329 if (res < 0)
330 purg_lnk(arcn);
331
332 if (vflag && vfpart) {
333 (void)putc('\n', listf);
334 vfpart = 0;
335 }
336 goto popd;
337 }
338 /*
339 * we have a file with data here. If we cannot create it, skip
340 * over the data and purge the name from hard link table
341 */
342 if (to_stdout)
343 fd = STDOUT_FILENO;
344 else if ((fd = file_creat(arcn)) < 0) {
345 (void)rd_skip(arcn->skip + arcn->pad);
346 purg_lnk(arcn);
347 goto popd;
348 }
349 /*
350 * extract the file from the archive and skip over padding and
351 * any unprocessed data
352 */
353 extdata:
354 res = (*frmt->rd_data)(arcn, fd, &cnt);
355 if (fd != STDOUT_FILENO)
356 file_close(arcn, fd);
357 if (vflag && vfpart) {
358 (void)putc('\n', listf);
359 vfpart = 0;
360 }
361 if (!res)
362 (void)rd_skip(cnt + arcn->pad);
363
364 popd:
365 /*
366 * if required, chdir around.
367 */
368 if ((arcn->pat != NULL) && (arcn->pat->chdname != NULL))
369 if (!to_stdout && fchdir(cwdfd) != 0)
370 syswarn(1, errno,
371 "Can't fchdir to starting directory");
372 }
373
374 /*
375 * all done, restore directory modes and times as required; make sure
376 * all patterns supplied by the user were matched; block off signals
377 * to avoid chance for multiple entry into the cleanup code.
378 */
379 (void)(*frmt->end_rd)();
380 (void)sigprocmask(SIG_BLOCK, &s_mask, NULL);
381 ar_close();
382 proc_dir();
383 pat_chk();
384 }
385
386 /*
387 * wr_archive()
388 * Write an archive. used in both creating a new archive and appends on
389 * previously written archive.
390 */
391
392 static void
wr_archive(ARCHD * arcn,int is_app)393 wr_archive(ARCHD *arcn, int is_app)
394 {
395 int res;
396 int hlk;
397 int wr_one;
398 off_t cnt;
399 int (*wrf)(ARCHD *);
400 int fd = -1;
401 time_t now;
402
403 /*
404 * if this format supports hard link storage, start up the database
405 * that detects them.
406 */
407 if (((hlk = frmt->hlk) == 1) && (lnk_start() < 0))
408 return;
409
410 /*
411 * if this is not append, and there are no files, we do not write a
412 * trailer
413 */
414 wr_one = is_app;
415
416 /*
417 * start up the file traversal code and format specific write
418 */
419 if (ftree_start() < 0) {
420 if (is_app)
421 goto trailer;
422 return;
423 } else if (((*frmt->st_wr)(is_app) < 0))
424 return;
425
426 wrf = frmt->wr;
427
428 /*
429 * When we are doing interactive rename, we store the mapping of names
430 * so we can fix up hard links files later in the archive.
431 */
432 if (iflag && (name_start() < 0))
433 return;
434
435 now = time(NULL);
436
437 /*
438 * while there are files to archive, process them one at at time
439 */
440 while (next_file(arcn) == 0) {
441 /*
442 * check if this file meets user specified options match.
443 */
444 if (sel_chk(arcn) != 0)
445 continue;
446 fd = -1;
447 if (uflag) {
448 /*
449 * only archive if this file is newer than a file with
450 * the same name that is already stored on the archive
451 */
452 if ((res = chk_ftime(arcn)) < 0)
453 break;
454 if (res > 0) {
455 ftree_skipped_newer();
456 continue;
457 }
458 }
459
460 /*
461 * this file is considered selected now. see if this is a hard
462 * link to a file already stored
463 */
464 ftree_sel(arcn);
465 if (hlk && (chk_lnk(arcn) < 0))
466 break;
467
468 if ((arcn->type == PAX_REG) || (arcn->type == PAX_HRG) ||
469 (arcn->type == PAX_CTG)) {
470 /*
471 * we will have to read this file. by opening it now we
472 * can avoid writing a header to the archive for a file
473 * we were later unable to read (we also purge it from
474 * the link table).
475 */
476 if ((fd = open(arcn->org_name, O_RDONLY, 0)) < 0) {
477 syswarn(1,errno, "Unable to open %s to read",
478 arcn->org_name);
479 purg_lnk(arcn);
480 continue;
481 }
482 }
483
484 /*
485 * Now modify the name as requested by the user
486 */
487 if ((res = mod_name(arcn)) < 0) {
488 /*
489 * name modification says to skip this file, close the
490 * file and purge link table entry
491 */
492 rdfile_close(arcn, &fd);
493 purg_lnk(arcn);
494 break;
495 }
496
497 if ((res > 0) || (docrc && (set_crc(arcn, fd) < 0))) {
498 /*
499 * unable to obtain the crc we need, close the file,
500 * purge link table entry
501 */
502 rdfile_close(arcn, &fd);
503 purg_lnk(arcn);
504 continue;
505 }
506
507 if (vflag) {
508 if (vflag > 1)
509 ls_list(arcn, now, listf);
510 else {
511 (void)safe_print(arcn->name, listf);
512 vfpart = 1;
513 }
514 } else if (Vflag) {
515 (void)putc('.', listf);
516 (void)fflush(listf);
517 vfpart = 1;
518 }
519 ++flcnt;
520
521 /*
522 * looks safe to store the file, have the format specific
523 * routine write routine store the file header on the archive
524 */
525 if ((res = (*wrf)(arcn)) < 0) {
526 rdfile_close(arcn, &fd);
527 break;
528 }
529 wr_one = 1;
530 if (res > 0) {
531 /*
532 * format write says no file data needs to be stored
533 * so we are done messing with this file
534 */
535 if (vflag && vfpart) {
536 (void)putc('\n', listf);
537 vfpart = 0;
538 }
539 rdfile_close(arcn, &fd);
540 continue;
541 }
542
543 /*
544 * Add file data to the archive, quit on write error. if we
545 * cannot write the entire file contents to the archive we
546 * must pad the archive to replace the missing file data
547 * (otherwise during an extract the file header for the file
548 * which FOLLOWS this one will not be where we expect it to
549 * be).
550 */
551 res = (*frmt->wr_data)(arcn, fd, &cnt);
552 rdfile_close(arcn, &fd);
553 if (vflag && vfpart) {
554 (void)putc('\n', listf);
555 vfpart = 0;
556 }
557 if (res < 0)
558 break;
559
560 /*
561 * pad as required, cnt is number of bytes not written
562 */
563 if (((cnt > 0) && (wr_skip(cnt) < 0)) ||
564 ((arcn->pad > 0) && (wr_skip(arcn->pad) < 0)))
565 break;
566 }
567
568 trailer:
569 /*
570 * tell format to write trailer; pad to block boundary; reset directory
571 * mode/access times, and check if all patterns supplied by the user
572 * were matched. block off signals to avoid chance for multiple entry
573 * into the cleanup code
574 */
575 if (wr_one) {
576 (*frmt->end_wr)();
577 wr_fin();
578 }
579 (void)sigprocmask(SIG_BLOCK, &s_mask, NULL);
580 ar_close();
581 if (tflag)
582 proc_dir();
583 ftree_chk();
584 }
585
586 /*
587 * append()
588 * Add file to previously written archive. Archive format specified by the
589 * user must agree with archive. The archive is read first to collect
590 * modification times (if -u) and locate the archive trailer. The archive
591 * is positioned in front of the record with the trailer and wr_archive()
592 * is called to add the new members.
593 * PAX IMPLEMENTATION DETAIL NOTE:
594 * -u is implemented by adding the new members to the end of the archive.
595 * Care is taken so that these do not end up as links to the older
596 * version of the same file already stored in the archive. It is expected
597 * when extraction occurs these newer versions will over-write the older
598 * ones stored "earlier" in the archive (this may be a bad assumption as
599 * it depends on the implementation of the program doing the extraction).
600 * It is really difficult to splice in members without either re-writing
601 * the entire archive (from the point were the old version was), or having
602 * assistance of the format specification in terms of a special update
603 * header that invalidates a previous archive record. The posix spec left
604 * the method used to implement -u unspecified. This pax is able to
605 * over write existing files that it creates.
606 */
607
608 void
append(void)609 append(void)
610 {
611 ARCHD *arcn;
612 int res;
613 ARCHD archd;
614 FSUB *orgfrmt;
615 int udev;
616 off_t tlen;
617
618 arcn = &archd;
619 orgfrmt = frmt;
620
621 /*
622 * Do not allow an append operation if the actual archive is of a
623 * different format than the user specified format.
624 */
625 if (get_arc() < 0)
626 return;
627 if ((orgfrmt != NULL) && (orgfrmt != frmt)) {
628 paxwarn(1, "Cannot mix current archive format %s with %s",
629 frmt->name, orgfrmt->name);
630 return;
631 }
632
633 /*
634 * pass the format any options and start up format
635 */
636 if (((*frmt->options)() < 0) || ((*frmt->st_rd)() < 0))
637 return;
638
639 /* hack to allow appending to Unix Archiver libraries */
640 if (frmt->is_uar)
641 ar_do_keepopen = 1;
642
643 /*
644 * if we only are adding members that are newer, we need to save the
645 * mod times for all files we see.
646 */
647 if (uflag && (ftime_start() < 0))
648 return;
649
650 /*
651 * some archive formats encode hard links by recording the device and
652 * file serial number (inode) but copy the file anyway (multiple times)
653 * to the archive. When we append, we run the risk that newly added
654 * files may have the same device and inode numbers as those recorded
655 * on the archive but during a previous run. If this happens, when the
656 * archive is extracted we get INCORRECT hard links. We avoid this by
657 * remapping the device numbers so that newly added files will never
658 * use the same device number as one found on the archive. remapping
659 * allows new members to safely have links among themselves. remapping
660 * also avoids problems with file inode (serial number) truncations
661 * when the inode number is larger than storage space in the archive
662 * header. See the remap routines for more details.
663 */
664 if ((udev = frmt->udev) && (dev_start() < 0))
665 return;
666
667 /*
668 * reading the archive may take a long time. If verbose tell the user
669 */
670 if (vflag) {
671 (void)fprintf(listf,
672 "%s: Reading archive to position at the end...", argv0);
673 vfpart = 1;
674 }
675
676 /*
677 * step through the archive until the format says it is done
678 */
679 while (next_head(arcn) == 0) {
680 /*
681 * check if this file meets user specified options.
682 */
683 if (sel_chk(arcn) != 0) {
684 if (rd_skip(arcn->skip + arcn->pad) == 1)
685 break;
686 continue;
687 }
688
689 if (uflag) {
690 /*
691 * see if this is the newest version of this file has
692 * already been seen, if so skip.
693 */
694 if ((res = chk_ftime(arcn)) < 0)
695 break;
696 if (res > 0) {
697 if (rd_skip(arcn->skip + arcn->pad) == 1)
698 break;
699 continue;
700 }
701 }
702
703 /*
704 * Store this device number. Device numbers seen during the
705 * read phase of append will cause newly appended files with a
706 * device number seen in the old part of the archive to be
707 * remapped to an unused device number.
708 */
709 if ((udev && (add_dev(arcn) < 0)) ||
710 (rd_skip(arcn->skip + arcn->pad) == 1))
711 break;
712 }
713
714 /*
715 * done, finish up read and get the number of bytes to back up so we
716 * can add new members. The format might have used the hard link table,
717 * purge it.
718 */
719 tlen = (*frmt->end_rd)();
720 lnk_end();
721
722 /*
723 * try to position for write, if this fails quit. if any error occurs,
724 * we will refuse to write
725 */
726 if (appnd_start(tlen) < 0)
727 return;
728
729 /*
730 * tell the user we are done reading.
731 */
732 if (vflag && vfpart) {
733 (void)fputs("done.\n", listf);
734 vfpart = 0;
735 }
736
737 /*
738 * go to the writing phase to add the new members
739 */
740 wr_archive(arcn, 1);
741 }
742
743 /*
744 * archive()
745 * write a new archive
746 */
747
748 void
archive(void)749 archive(void)
750 {
751 ARCHD archd;
752
753 /*
754 * if we only are adding members that are newer, we need to save the
755 * mod times for all files; set up for writing; pass the format any
756 * options write the archive
757 */
758 if ((uflag && (ftime_start() < 0)) || (wr_start() < 0))
759 return;
760 if ((*frmt->options)() < 0)
761 return;
762
763 wr_archive(&archd, 0);
764 }
765
766 /*
767 * copy()
768 * copy files from one part of the file system to another. this does not
769 * use any archive storage. The EFFECT OF THE COPY IS THE SAME as if an
770 * archive was written and then extracted in the destination directory
771 * (except the files are forced to be under the destination directory).
772 */
773
774 void
copy(void)775 copy(void)
776 {
777 ARCHD *arcn;
778 int res;
779 int fddest;
780 char *dest_pt;
781 size_t dlen;
782 size_t drem;
783 int fdsrc = -1;
784 struct stat sb;
785 ARCHD archd;
786 char dirbuf[PAXPATHLEN+1];
787
788 arcn = &archd;
789 /*
790 * set up the destination dir path and make sure it is a directory. We
791 * make sure we have a trailing / on the destination
792 */
793 dlen = strlcpy(dirbuf, dirptr, sizeof(dirbuf));
794 if (dlen >= sizeof(dirbuf) ||
795 (dlen == sizeof(dirbuf) - 1 && dirbuf[dlen - 1] != '/')) {
796 paxwarn(1, "directory name is too long %s", dirptr);
797 return;
798 }
799 dest_pt = dirbuf + dlen;
800 if (*(dest_pt-1) != '/') {
801 *dest_pt++ = '/';
802 *dest_pt = '\0';
803 ++dlen;
804 }
805 drem = PAXPATHLEN - dlen;
806
807 if (stat(dirptr, &sb) < 0) {
808 syswarn(1, errno, "Cannot access destination directory %s",
809 dirptr);
810 return;
811 }
812 if (!S_ISDIR(sb.st_mode)) {
813 paxwarn(1, "Destination is not a directory %s", dirptr);
814 return;
815 }
816
817 /*
818 * start up the hard link table; file traversal routines and the
819 * modification time and access mode database
820 */
821 if ((lnk_start() < 0) || (ftree_start() < 0) || (dir_start() < 0))
822 return;
823
824 /*
825 * When we are doing interactive rename, we store the mapping of names
826 * so we can fix up hard links files later in the archive.
827 */
828 if (iflag && (name_start() < 0))
829 return;
830
831 /*
832 * set up to cp file trees
833 */
834 cp_start();
835
836 /*
837 * while there are files to archive, process them
838 */
839 while (next_file(arcn) == 0) {
840 fdsrc = -1;
841
842 /*
843 * check if this file meets user specified options
844 */
845 if (sel_chk(arcn) != 0)
846 continue;
847
848 /*
849 * if there is already a file in the destination directory with
850 * the same name and it is newer, skip the one stored on the
851 * archive.
852 * NOTE: this test is done BEFORE name modifications as
853 * specified by pax. this can be confusing to the user who
854 * might expect the test to be done on an existing file AFTER
855 * the name mod. In honesty the pax spec is probably flawed in
856 * this respect
857 */
858 if (uflag || Dflag) {
859 /*
860 * create the destination name
861 */
862 if (strlcpy(dest_pt, arcn->name + (*arcn->name == '/'),
863 drem + 1) > drem) {
864 paxwarn(1, "Destination pathname too long %s",
865 arcn->name);
866 continue;
867 }
868
869 /*
870 * if existing file is same age or newer skip
871 */
872 res = lstat(dirbuf, &sb);
873 *dest_pt = '\0';
874
875 if (res == 0) {
876 ftree_skipped_newer();
877 if (uflag && Dflag) {
878 if ((arcn->sb.st_mtime<=sb.st_mtime) &&
879 (arcn->sb.st_ctime<=sb.st_ctime))
880 continue;
881 } else if (Dflag) {
882 if (arcn->sb.st_ctime <= sb.st_ctime)
883 continue;
884 } else if (arcn->sb.st_mtime <= sb.st_mtime)
885 continue;
886 }
887 }
888
889 /*
890 * this file is considered selected. See if this is a hard link
891 * to a previous file; modify the name as requested by the
892 * user; set the final destination.
893 */
894 ftree_sel(arcn);
895 if ((chk_lnk(arcn) < 0) || ((res = mod_name(arcn)) < 0))
896 break;
897 if ((res > 0) || (set_dest(arcn, dirbuf, dlen) < 0)) {
898 /*
899 * skip file, purge from link table
900 */
901 purg_lnk(arcn);
902 continue;
903 }
904
905 /*
906 * Non standard -Y and -Z flag. When the existing file is
907 * same age or newer skip
908 */
909 if ((Yflag || Zflag) && ((lstat(arcn->name, &sb) == 0))) {
910 if (Yflag && Zflag) {
911 if ((arcn->sb.st_mtime <= sb.st_mtime) &&
912 (arcn->sb.st_ctime <= sb.st_ctime))
913 continue;
914 } else if (Yflag) {
915 if (arcn->sb.st_ctime <= sb.st_ctime)
916 continue;
917 } else if (arcn->sb.st_mtime <= sb.st_mtime)
918 continue;
919 }
920
921 if (vflag) {
922 (void)safe_print(arcn->name, listf);
923 vfpart = 1;
924 } else if (Vflag) {
925 (void)putc('.', listf);
926 (void)fflush(listf);
927 vfpart = 1;
928 }
929 ++flcnt;
930
931 /*
932 * try to create a hard link to the src file if requested
933 * but make sure we are not trying to overwrite ourselves.
934 */
935 if (lflag)
936 res = cross_lnk(arcn);
937 else
938 res = chk_same(arcn);
939 if (res <= 0) {
940 if (vflag && vfpart) {
941 (void)putc('\n', listf);
942 vfpart = 0;
943 }
944 continue;
945 }
946
947 /*
948 * have to create a new file
949 */
950 if ((arcn->type != PAX_REG) && (arcn->type != PAX_CTG)) {
951 /*
952 * create a link or special file
953 */
954 if ((arcn->type == PAX_HLK) || (arcn->type == PAX_HRG))
955 res = lnk_creat(arcn, NULL);
956 else
957 res = node_creat(arcn);
958 if (res < 0)
959 purg_lnk(arcn);
960 if (vflag && vfpart) {
961 (void)putc('\n', listf);
962 vfpart = 0;
963 }
964 continue;
965 }
966
967 /*
968 * have to copy a regular file to the destination directory.
969 * first open source file and then create the destination file
970 */
971 if ((fdsrc = open(arcn->org_name, O_RDONLY, 0)) < 0) {
972 syswarn(1, errno, "Unable to open %s to read",
973 arcn->org_name);
974 purg_lnk(arcn);
975 continue;
976 }
977 if ((fddest = file_creat(arcn)) < 0) {
978 rdfile_close(arcn, &fdsrc);
979 purg_lnk(arcn);
980 continue;
981 }
982
983 /*
984 * copy source file data to the destination file
985 */
986 cp_file(arcn, fdsrc, fddest);
987 file_close(arcn, fddest);
988 rdfile_close(arcn, &fdsrc);
989
990 if (vflag && vfpart) {
991 (void)putc('\n', listf);
992 vfpart = 0;
993 }
994 }
995
996 /*
997 * restore directory modes and times as required; make sure all
998 * patterns were selected block off signals to avoid chance for
999 * multiple entry into the cleanup code.
1000 */
1001 (void)sigprocmask(SIG_BLOCK, &s_mask, NULL);
1002 ar_close();
1003 proc_dir();
1004 ftree_chk();
1005 }
1006
1007 /*
1008 * next_head()
1009 * try to find a valid header in the archive. Uses format specific
1010 * routines to extract the header and id the trailer. Trailers may be
1011 * located within a valid header or in an invalid header (the location
1012 * is format specific. The inhead field from the option table tells us
1013 * where to look for the trailer).
1014 * We keep reading (and resyncing) until we get enough contiguous data
1015 * to check for a header. If we cannot find one, we shift by a byte
1016 * add a new byte from the archive to the end of the buffer and try again.
1017 * If we get a read error, we throw out what we have (as we must have
1018 * contiguous data) and start over again.
1019 * ASSUMED: headers fit within a BLKMULT header.
1020 * Return:
1021 * 0 if we got a header, -1 if we are unable to ever find another one
1022 * (we reached the end of input, or we reached the limit on retries. see
1023 * the specs for rd_wrbuf() for more details)
1024 */
1025
1026 static int
next_head(ARCHD * arcn)1027 next_head(ARCHD *arcn)
1028 {
1029 int ret;
1030 char *hdend;
1031 int res;
1032 int shftsz;
1033 int hsz;
1034 int in_resync = 0; /* set when we are in resync mode */
1035 int cnt = 0; /* counter for trailer function */
1036 int first = 1; /* on 1st read, EOF isn't premature. */
1037
1038 /*
1039 * set up initial conditions, we want a whole frmt->hsz block as we
1040 * have no data yet.
1041 */
1042 res = hsz = frmt->hsz;
1043 hdend = hdbuf;
1044 shftsz = hsz - 1;
1045 for (;;) {
1046 /*
1047 * keep looping until we get a contiguous FULL buffer
1048 * (frmt->hsz is the proper size)
1049 */
1050 for (;;) {
1051 if ((ret = rd_wrbuf(hdend, res)) == res)
1052 break;
1053
1054 /*
1055 * If we read 0 bytes (EOF) from an archive when we
1056 * expect to find a header, we have stepped upon
1057 * an archive without the customary block of zeroes
1058 * end marker. It's just stupid to error out on
1059 * them, so exit gracefully.
1060 */
1061 if (first && ret == 0)
1062 return(-1);
1063 first = 0;
1064
1065 /*
1066 * some kind of archive read problem, try to resync the
1067 * storage device, better give the user the bad news.
1068 */
1069 if ((ret == 0) || (rd_sync() < 0) || frmt->is_uar) {
1070 no_header:
1071 paxwarn(1,"Premature end of file on archive read");
1072 return(-1);
1073 }
1074 if (!in_resync) {
1075 if (act == APPND) {
1076 paxwarn(1,
1077 "Archive I/O error, cannot continue");
1078 return(-1);
1079 }
1080 paxwarn(1,"Archive I/O error. Trying to recover.");
1081 ++in_resync;
1082 }
1083
1084 /*
1085 * oh well, throw it all out and start over
1086 */
1087 res = hsz;
1088 hdend = hdbuf;
1089 }
1090
1091 /*
1092 * ok we have a contiguous buffer of the right size. Call the
1093 * format read routine. If this was not a valid header and this
1094 * format stores trailers outside of the header, call the
1095 * format specific trailer routine to check for a trailer. We
1096 * have to watch out that we do not mis-identify file data or
1097 * block padding as a header or trailer. Format specific
1098 * trailer functions must NOT check for the trailer while we
1099 * are running in resync mode. Some trailer functions may tell
1100 * us that this block cannot contain a valid header either, so
1101 * we then throw out the entire block and start over.
1102 */
1103 if ((*frmt->rd)(arcn, hdbuf) == 0)
1104 break;
1105
1106 if (frmt->is_uar)
1107 goto no_header;
1108
1109 if (!frmt->inhead) {
1110 /*
1111 * this format has trailers outside of valid headers
1112 */
1113 if ((ret = (*frmt->trail)(arcn,hdbuf,in_resync,&cnt)) == 0){
1114 /*
1115 * valid trailer found, drain input as required
1116 */
1117 ar_drain();
1118 return(-1);
1119 }
1120
1121 if (ret == 1) {
1122 /*
1123 * we are in resync and we were told to throw
1124 * the whole block out because none of the
1125 * bytes in this block can be used to form a
1126 * valid header
1127 */
1128 res = hsz;
1129 hdend = hdbuf;
1130 continue;
1131 }
1132 }
1133
1134 /*
1135 * Brute force section.
1136 * not a valid header. We may be able to find a header yet. So
1137 * we shift over by one byte, and set up to read one byte at a
1138 * time from the archive and place it at the end of the buffer.
1139 * We will keep moving byte at a time until we find a header or
1140 * get a read error and have to start over.
1141 */
1142 if (!in_resync) {
1143 if (act == APPND) {
1144 paxwarn(1,"Unable to append, archive header flaw");
1145 return(-1);
1146 }
1147 paxwarn(1,"Invalid header, starting valid header search.");
1148 ++in_resync;
1149 }
1150 memmove(hdbuf, hdbuf+1, shftsz);
1151 res = 1;
1152 hdend = hdbuf + shftsz;
1153 }
1154
1155 /*
1156 * ok got a valid header, check for trailer if format encodes it in the
1157 * the header. NOTE: the parameters are different than trailer routines
1158 * which encode trailers outside of the header!
1159 */
1160 if (frmt->inhead && ((*frmt->trail)(arcn,NULL,0,NULL) == 0)) {
1161 /*
1162 * valid trailer found, drain input as required
1163 */
1164 ar_drain();
1165 return(-1);
1166 }
1167
1168 ++flcnt;
1169 return(0);
1170 }
1171
1172 /*
1173 * get_arc()
1174 * Figure out what format an archive is. Handles archive with flaws by
1175 * brute force searches for a legal header in any supported format. The
1176 * format id routines have to be careful to NOT mis-identify a format.
1177 * ASSUMED: headers fit within a BLKMULT header.
1178 * Return:
1179 * 0 if archive found -1 otherwise
1180 */
1181
1182 static int
get_arc(void)1183 get_arc(void)
1184 {
1185 int i;
1186 int hdsz = 0;
1187 int res;
1188 int minhd = BLKMULT;
1189 char *hdend;
1190 int notice = 0;
1191
1192 /*
1193 * find the smallest header size in all archive formats and then set up
1194 * to read the archive.
1195 */
1196 for (i = 0; ford[i] >= 0; ++i) {
1197 if (fsub[ford[i]].hsz < minhd)
1198 minhd = fsub[ford[i]].hsz;
1199 }
1200 if (rd_start() < 0)
1201 return(-1);
1202 res = BLKMULT;
1203 hdsz = 0;
1204 hdend = hdbuf;
1205
1206 /* try to verify against ar first */
1207 if (buf_fill_internal(8) == 8) {
1208 i = rd_wrbuf(hdend, 8);
1209 if (i == 8 && uar_ismagic(hdbuf) == 0) {
1210 extern int F_UAR;
1211
1212 frmt = &(fsub[F_UAR]);
1213 return (0);
1214 }
1215 if (i > 0)
1216 pback(hdend, i);
1217 }
1218
1219 for (;;) {
1220 for (;;) {
1221 /*
1222 * fill the buffer with at least the smallest header
1223 */
1224 i = rd_wrbuf(hdend, res);
1225 if (i > 0)
1226 hdsz += i;
1227 if (hdsz >= minhd)
1228 break;
1229
1230 /*
1231 * if we cannot recover from a read error quit
1232 */
1233 if ((i == 0) || (rd_sync() < 0))
1234 goto out;
1235
1236 /*
1237 * when we get an error none of the data we already
1238 * have can be used to create a legal header (we just
1239 * got an error in the middle), so we throw it all out
1240 * and refill the buffer with fresh data.
1241 */
1242 res = BLKMULT;
1243 hdsz = 0;
1244 hdend = hdbuf;
1245 if (!notice) {
1246 if (act == APPND)
1247 return(-1);
1248 paxwarn(1,"Cannot identify format. Searching...");
1249 ++notice;
1250 }
1251 }
1252
1253 /*
1254 * we have at least the size of the smallest header in any
1255 * archive format. Look to see if we have a match. The array
1256 * ford[] is used to specify the header id order to reduce the
1257 * chance of incorrectly id'ing a valid header (some formats
1258 * may be subsets of each other and the order would then be
1259 * important).
1260 */
1261 for (i = 0; ford[i] >= 0; ++i) {
1262 if ((*fsub[ford[i]].id)(hdbuf, hdsz) < 0)
1263 continue;
1264 frmt = &(fsub[ford[i]]);
1265 /*
1266 * yuck, to avoid slow special case code in the extract
1267 * routines, just push this header back as if it was
1268 * not seen. We have left extra space at start of the
1269 * buffer for this purpose. This is a bit ugly, but
1270 * adding all the special case code is far worse.
1271 */
1272 pback(hdbuf, hdsz);
1273 return(0);
1274 }
1275
1276 /*
1277 * We have a flawed archive, no match. we start searching, but
1278 * we never allow additions to flawed archives
1279 */
1280 if (!notice) {
1281 if (act == APPND)
1282 return(-1);
1283 paxwarn(1, "Cannot identify format. Searching...");
1284 ++notice;
1285 }
1286
1287 /*
1288 * brute force search for a header that we can id.
1289 * we shift through byte at a time. this is slow, but we cannot
1290 * determine the nature of the flaw in the archive in a
1291 * portable manner
1292 */
1293 if (--hdsz > 0) {
1294 memmove(hdbuf, hdbuf+1, hdsz);
1295 res = BLKMULT - hdsz;
1296 hdend = hdbuf + hdsz;
1297 } else {
1298 res = BLKMULT;
1299 hdend = hdbuf;
1300 hdsz = 0;
1301 }
1302 }
1303
1304 out:
1305 /*
1306 * we cannot find a header, bow, apologise and quit
1307 */
1308 paxwarn(1, "Sorry, unable to determine archive format.");
1309 return(-1);
1310 }
1311