1 /* $OpenBSD: pax.c,v 1.33 2012/04/19 04:26:46 deraadt Exp $ */
2 /* $NetBSD: pax.c,v 1.5 1996/03/26 23:54:20 mrg Exp $ */
3
4 /*-
5 * Copyright (c) 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/stat.h>
41 #include <sys/time.h>
42 #include <sys/resource.h>
43 #include <signal.h>
44 #include <unistd.h>
45 #include <stdio.h>
46 #include <stdlib.h>
47 #include <string.h>
48 #include <errno.h>
49 #include <err.h>
50 #include <fcntl.h>
51 #include <paths.h>
52 #include <time.h>
53 #include "pax.h"
54 #include "extern.h"
55
56 __RCSID("$MirOS: src/bin/pax/pax.c,v 1.19 2012/06/05 18:22:57 tg Exp $");
57
58 static int gen_init(void);
59 static void sig_cleanup(int) __attribute__((__noreturn__));
60
61 /*
62 * PAX main routines, general globals and some simple start up routines
63 */
64
65 /*
66 * Variables that can be accessed by any routine within pax
67 */
68 int act = ERROR; /* read/write/append/copy */
69 FSUB *frmt = NULL; /* archive format type */
70 int cflag; /* match all EXCEPT pattern/file */
71 int cwdfd; /* starting cwd */
72 int dflag; /* directory member match only */
73 int iflag; /* interactive file/archive rename */
74 int kflag; /* do not overwrite existing files */
75 int lflag; /* use hard links when possible */
76 int nflag; /* select first archive member match */
77 int tflag; /* restore access time after read */
78 int uflag; /* ignore older modification time files */
79 int Vflag = 0; /* print a dot for each file processed */
80 int vflag; /* produce verbose output */
81 int Dflag; /* same as uflag except inode change time */
82 int Hflag; /* follow command line symlinks (write only) */
83 int Lflag; /* follow symlinks when writing */
84 int Xflag; /* archive files with same device id only */
85 int Yflag; /* same as Dflag except after name mode */
86 int Zflag; /* same as uflag except after name mode */
87 int zeroflag; /* use \0 as pathname terminator */
88 int vfpart = 0; /* is partial verbose output in progress */
89 int patime = 1; /* preserve file access time */
90 int pmtime = 1; /* preserve file modification times */
91 int nodirs; /* do not create directories as needed */
92 int pmode; /* preserve file mode bits */
93 int pids; /* preserve file uid/gid */
94 int rmleadslash = 0; /* remove leading '/' from pathnames */
95 int exit_val; /* exit value */
96 int docrc; /* check/create file crc */
97 char *dirptr; /* destination dir in a copy */
98 char *ltmfrmt; /* -v locale time format (if any) */
99 const char *argv0; /* root of argv[0] */
100 sigset_t s_mask; /* signal mask for cleanup critical sect */
101 FILE *listf; /* fp to print file list to (default stderr) */
102 char *tempfile; /* tempfile to use for mkstemp(3) */
103 char *tempbase; /* basename of tempfile to use for mkstemp(3) */
104
105 /*
106 * PAX - Portable Archive Interchange
107 *
108 * A utility to read, write, and write lists of the members of archive
109 * files and copy directory hierarchies. A variety of archive formats
110 * are supported (some are described in POSIX 1003.1 10.1):
111 *
112 * ustar - 10.1.1 extended tar interchange format
113 * cpio - 10.1.2 extended cpio interchange format
114 * tar - old BSD 4.3 tar format
115 * binary cpio - old cpio with binary header format
116 * sysVR4 cpio - with and without CRC
117 *
118 * This version is a superset of IEEE Std 1003.2b-d3
119 *
120 * Summary of Extensions to the IEEE Standard:
121 *
122 * 1 READ ENHANCEMENTS
123 * 1.1 Operations which read archives will continue to operate even when
124 * processing archives which may be damaged, truncated, or fail to meet
125 * format specs in several different ways. Damaged sections of archives
126 * are detected and avoided if possible. Attempts will be made to resync
127 * archive read operations even with badly damaged media.
128 * 1.2 Blocksize requirements are not strictly enforced on archive read.
129 * Tapes which have variable sized records can be read without errors.
130 * 1.3 The user can specify via the non-standard option flag -E if error
131 * resync operation should stop on a media error, try a specified number
132 * of times to correct, or try to correct forever.
133 * 1.4 Sparse files (lseek holes) stored on the archive (but stored with blocks
134 * of all zeros will be restored with holes appropriate for the target
135 * filesystem
136 * 1.5 The user is notified whenever something is found during archive
137 * read operations which violates spec (but the read will continue).
138 * 1.6 Multiple archive volumes can be read and may span over different
139 * archive devices
140 * 1.7 Rigidly restores all file attributes exactly as they are stored on the
141 * archive.
142 * 1.8 Modification change time ranges can be specified via multiple -T
143 * options. These allow a user to select files whose modification time
144 * lies within a specific time range.
145 * 1.9 Files can be selected based on owner (user name or uid) via one or more
146 * -U options.
147 * 1.10 Files can be selected based on group (group name or gid) via one o
148 * more -G options.
149 * 1.11 File modification time can be checked against existing file after
150 * name modification (-Z)
151 *
152 * 2 WRITE ENHANCEMENTS
153 * 2.1 Write operation will stop instead of allowing a user to create a flawed
154 * flawed archive (due to any problem).
155 * 2.2 Archives written by pax are forced to strictly conform to both the
156 * archive and pax the specific format specifications.
157 * 2.3 Blocking size and format is rigidly enforced on writes.
158 * 2.4 Formats which may exhibit header overflow problems (they have fields
159 * too small for large file systems, such as inode number storage), use
160 * routines designed to repair this problem. These techniques still
161 * conform to both pax and format specifications, but no longer truncate
162 * these fields. This removes any restrictions on using these archive
163 * formats on large file systems.
164 * 2.5 Multiple archive volumes can be written and may span over different
165 * archive devices
166 * 2.6 A archive volume record limit allows the user to specify the number
167 * of bytes stored on an archive volume. When reached the user is
168 * prompted for the next archive volume. This is specified with the
169 * non-standard -B flag. The limit is rounded up to the next blocksize.
170 * 2.7 All archive padding during write use zero filled sections. This makes
171 * it much easier to pull data out of flawed archive during read
172 * operations.
173 * 2.8 Access time reset with the -t applies to all file nodes (including
174 * directories).
175 * 2.9 Symbolic links can be followed with -L (optional in the spec).
176 * 2.10 Modification or inode change time ranges can be specified via
177 * multiple -T options. These allow a user to select files whose
178 * modification or inode change time lies within a specific time range.
179 * 2.11 Files can be selected based on owner (user name or uid) via one or more
180 * -U options.
181 * 2.12 Files can be selected based on group (group name or gid) via one o
182 * more -G options.
183 * 2.13 Symlinks which appear on the command line can be followed (without
184 * following other symlinks; -H flag)
185 *
186 * 3 COPY ENHANCEMENTS
187 * 3.1 Sparse files (lseek holes) can be copied without expanding the holes
188 * into zero filled blocks. The file copy is created with holes which are
189 * appropriate for the target filesystem
190 * 3.2 Access time as well as modification time on copied file trees can be
191 * preserved with the appropriate -p options.
192 * 3.3 Access time reset with the -t applies to all file nodes (including
193 * directories).
194 * 3.4 Symbolic links can be followed with -L (optional in the spec).
195 * 3.5 Modification or inode change time ranges can be specified via
196 * multiple -T options. These allow a user to select files whose
197 * modification or inode change time lies within a specific time range.
198 * 3.6 Files can be selected based on owner (user name or uid) via one or more
199 * -U options.
200 * 3.7 Files can be selected based on group (group name or gid) via one o
201 * more -G options.
202 * 3.8 Symlinks which appear on the command line can be followed (without
203 * following other symlinks; -H flag)
204 * 3.9 File inode change time can be checked against existing file before
205 * name modification (-D)
206 * 3.10 File inode change time can be checked against existing file after
207 * name modification (-Y)
208 * 3.11 File modification time can be checked against existing file after
209 * name modification (-Z)
210 *
211 * 4 GENERAL ENHANCEMENTS
212 * 4.1 Internal structure is designed to isolate format dependent and
213 * independent functions. Formats are selected via a format driver table.
214 * This encourages the addition of new archive formats by only having to
215 * write those routines which id, read and write the archive header.
216 */
217
218 /*
219 * main()
220 * parse options, set up and operate as specified by the user.
221 * any operational flaw will set exit_val to non-zero
222 * Return: 0 if ok, 1 otherwise
223 */
224
225 int
main(int argc,char ** argv)226 main(int argc, char **argv)
227 {
228 const char *tmpdir;
229 size_t tdlen;
230
231 /* may not be a constant, thus initialising early */
232 listf = stderr;
233
234 /*
235 * Keep a reference to cwd, so we can always come back home.
236 */
237 cwdfd = open(".", O_RDONLY);
238 if (cwdfd < 0) {
239 syswarn(1, errno, "Can't open current working directory.");
240 return(exit_val);
241 }
242
243 /*
244 * Where should we put temporary files?
245 */
246 if ((tmpdir = getenv("TMPDIR")) == NULL || *tmpdir == '\0')
247 tmpdir = _PATH_TMP;
248 tdlen = strlen(tmpdir);
249 while (tdlen > 0 && tmpdir[tdlen - 1] == '/')
250 tdlen--;
251 tempfile = malloc(tdlen + 1 + sizeof(_TFILE_BASE));
252 if (tempfile == NULL) {
253 paxwarn(1, "Cannot allocate memory for temp file name.");
254 return(exit_val);
255 }
256 if (tdlen)
257 memcpy(tempfile, tmpdir, tdlen);
258 tempbase = tempfile + tdlen;
259 *tempbase++ = '/';
260
261 /*
262 * parse options, determine operational mode, general init
263 */
264 options(argc, argv);
265 if ((gen_init() < 0) || (tty_init() < 0))
266 return(exit_val);
267
268 /*
269 * select a primary operation mode
270 */
271 switch (act) {
272 case EXTRACT:
273 extract();
274 break;
275 case ARCHIVE:
276 archive();
277 break;
278 case APPND:
279 if (compress_program != NULL)
280 errx(1, "cannot compress while appending");
281 append();
282 break;
283 case COPY:
284 copy();
285 break;
286 default:
287 /* for ar_io.c etc. */
288 act = LIST;
289 case LIST:
290 list();
291 break;
292 }
293 return(exit_val);
294 }
295
296 /*
297 * sig_cleanup()
298 * when interrupted we try to do whatever delayed processing we can.
299 * This is not critical, but we really ought to limit our damage when we
300 * are aborted by the user.
301 * Return:
302 * never....
303 */
304
305 static void
sig_cleanup(int which_sig)306 sig_cleanup(int which_sig)
307 {
308 /*
309 * The definition of this array doubles as compile-time assert
310 * on the size of long, off_t, and whether LONG_OFF_T is used,
311 * or not, correctly; target size is 80, error size -1.
312 */
313 char errbuf[((sizeof(long) >= 4) &&
314 (sizeof(ot_type) >= 4) &&
315 (sizeof(ot_type) == sizeof(off_t))) ? 80 : -1];
316
317 /*
318 * restore modes and times for any dirs we may have created
319 * or any dirs we may have read. Set vflag and vfpart so the user
320 * will clearly see the message on a line by itself.
321 */
322 vflag = vfpart = 1;
323
324 /* paxwarn() uses stdio; fake it as well as we can */
325 if (which_sig == SIGXCPU)
326 strlcpy(errbuf, "CPU time limit reached, cleaning up.\n",
327 sizeof errbuf);
328 else
329 strlcpy(errbuf, "Signal caught, cleaning up.\n",
330 sizeof errbuf);
331 if (!write(STDERR_FILENO, errbuf, strlen(errbuf))) {
332 /* dummy, to keep fortified gcc quiet */
333 errbuf[0] = '\0';
334 }
335
336 ar_close(); /* XXX signal race */
337 proc_dir(); /* XXX signal race */
338 if (tflag)
339 atdir_end(); /* XXX signal race */
340 _exit(1);
341 }
342
343 /*
344 * gen_init()
345 * general setup routines. Not all are required, but they really help
346 * when dealing with a medium to large sized archives.
347 */
348
349 static int
gen_init(void)350 gen_init(void)
351 {
352 struct rlimit reslimit;
353 struct sigaction n_hand;
354 struct sigaction o_hand;
355
356 /*
357 * Really needed to handle large archives. We can run out of memory for
358 * internal tables really fast when we have a whole lot of files...
359 */
360 if (getrlimit(RLIMIT_DATA , &reslimit) == 0){
361 reslimit.rlim_cur = reslimit.rlim_max;
362 (void)setrlimit(RLIMIT_DATA , &reslimit);
363 }
364
365 /*
366 * should file size limits be waived? if the os limits us, this is
367 * needed if we want to write a large archive
368 */
369 if (getrlimit(RLIMIT_FSIZE , &reslimit) == 0){
370 reslimit.rlim_cur = reslimit.rlim_max;
371 (void)setrlimit(RLIMIT_FSIZE , &reslimit);
372 }
373
374 /*
375 * increase the size the stack can grow to
376 */
377 if (getrlimit(RLIMIT_STACK , &reslimit) == 0){
378 reslimit.rlim_cur = reslimit.rlim_max;
379 (void)setrlimit(RLIMIT_STACK , &reslimit);
380 }
381
382 /*
383 * not really needed, but doesn't hurt
384 */
385 #ifdef RLIMIT_RSS
386 if (getrlimit(RLIMIT_RSS , &reslimit) == 0){
387 reslimit.rlim_cur = reslimit.rlim_max;
388 (void)setrlimit(RLIMIT_RSS , &reslimit);
389 }
390 #endif
391
392 /*
393 * Handle posix locale
394 *
395 * set user defines time printing format for -v option
396 */
397 ltmfrmt = getenv("LC_TIME");
398
399 /*
400 * signal handling to reset stored directory times and modes. Since
401 * we deal with broken pipes via failed writes we ignore it. We also
402 * deal with any file size limit through failed writes. CPU time
403 * limits are caught and a cleanup is forced.
404 */
405 if ((sigemptyset(&s_mask) < 0) || (sigaddset(&s_mask, SIGTERM) < 0) ||
406 (sigaddset(&s_mask,SIGINT) < 0)||(sigaddset(&s_mask,SIGHUP) < 0) ||
407 (sigaddset(&s_mask,SIGPIPE) < 0)||(sigaddset(&s_mask,SIGQUIT)<0) ||
408 (sigaddset(&s_mask,SIGXCPU) < 0)||(sigaddset(&s_mask,SIGXFSZ)<0)) {
409 paxwarn(1, "Unable to set up signal mask");
410 return(-1);
411 }
412 memset(&n_hand, 0, sizeof n_hand);
413 n_hand.sa_mask = s_mask;
414 n_hand.sa_flags = 0;
415 n_hand.sa_handler = sig_cleanup;
416
417 if ((sigaction(SIGHUP, &n_hand, &o_hand) < 0) || (
418 (o_hand.sa_handler == SIG_IGN) &&
419 (sigaction(SIGHUP, &o_hand, &o_hand) < 0)))
420 goto out;
421
422 if ((sigaction(SIGTERM, &n_hand, &o_hand) < 0) || (
423 (o_hand.sa_handler == SIG_IGN) &&
424 (sigaction(SIGTERM, &o_hand, &o_hand) < 0)))
425 goto out;
426
427 if ((sigaction(SIGINT, &n_hand, &o_hand) < 0) || (
428 (o_hand.sa_handler == SIG_IGN) &&
429 (sigaction(SIGINT, &o_hand, &o_hand) < 0)))
430 goto out;
431
432 if ((sigaction(SIGQUIT, &n_hand, &o_hand) < 0) || (
433 (o_hand.sa_handler == SIG_IGN) &&
434 (sigaction(SIGQUIT, &o_hand, &o_hand) < 0)))
435 goto out;
436
437 if ((sigaction(SIGXCPU, &n_hand, &o_hand) < 0) || (
438 (o_hand.sa_handler == SIG_IGN) &&
439 (sigaction(SIGXCPU, &o_hand, &o_hand) < 0)))
440 goto out;
441
442 n_hand.sa_handler = SIG_IGN;
443 if ((sigaction(SIGPIPE, &n_hand, &o_hand) < 0) ||
444 (sigaction(SIGXFSZ, &n_hand, &o_hand) < 0))
445 goto out;
446 return(0);
447
448 out:
449 syswarn(1, errno, "Unable to set up signal handler");
450 return(-1);
451 }
452