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