xref: /openbsd-src/bin/pax/pax.c (revision 9c38e9f3b7a3ed3332c151ebb7f9f8547b5313dc)
1 /*	$OpenBSD: pax.c,v 1.35 2014/01/09 03:12:25 guenther Exp $	*/
2 /*	$NetBSD: pax.c,v 1.5 1996/03/26 23:54:20 mrg Exp $	*/
3 
4 /*-
5  * Copyright (c) 1992 Keith Muller.
6  * Copyright (c) 1992, 1993
7  *	The Regents of the University of California.  All rights reserved.
8  *
9  * This code is derived from software contributed to Berkeley by
10  * Keith Muller of the University of California, San Diego.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  */
36 
37 #include <sys/types.h>
38 #include <sys/stat.h>
39 #include <sys/time.h>
40 #include <sys/resource.h>
41 #include <signal.h>
42 #include <unistd.h>
43 #include <stdlib.h>
44 #include <string.h>
45 #include <errno.h>
46 #include <err.h>
47 #include <fcntl.h>
48 #include <paths.h>
49 #include <stdio.h>
50 
51 #include "pax.h"
52 #include "extern.h"
53 static int gen_init(void);
54 
55 /*
56  * PAX main routines, general globals and some simple start up routines
57  */
58 
59 /*
60  * Variables that can be accessed by any routine within pax
61  */
62 int	act = DEFOP;		/* read/write/append/copy */
63 FSUB	*frmt = NULL;		/* archive format type */
64 int	cflag;			/* match all EXCEPT pattern/file */
65 int	cwdfd;			/* starting cwd */
66 int	dflag;			/* directory member match only  */
67 int	iflag;			/* interactive file/archive rename */
68 int	kflag;			/* do not overwrite existing files */
69 int	lflag;			/* use hard links when possible */
70 int	nflag;			/* select first archive member match */
71 int	tflag;			/* restore access time after read */
72 int	uflag;			/* ignore older modification time files */
73 int	vflag;			/* produce verbose output */
74 int	Dflag;			/* same as uflag except inode change time */
75 int	Hflag;			/* follow command line symlinks (write only) */
76 int	Lflag;			/* follow symlinks when writing */
77 int	Nflag;			/* only use numeric uid and gid */
78 int	Xflag;			/* archive files with same device id only */
79 int	Yflag;			/* same as Dflag except after name mode */
80 int	Zflag;			/* same as uflag except after name mode */
81 int	zeroflag;		/* use \0 as pathname terminator */
82 int	vfpart;			/* is partial verbose output in progress */
83 int	patime = 1;		/* preserve file access time */
84 int	pmtime = 1;		/* preserve file modification times */
85 int	nodirs;			/* do not create directories as needed */
86 int	pmode;			/* preserve file mode bits */
87 int	pids;			/* preserve file uid/gid */
88 int	rmleadslash = 0;	/* remove leading '/' from pathnames */
89 int	exit_val;		/* exit value */
90 int	docrc;			/* check/create file crc */
91 char	*dirptr;		/* destination dir in a copy */
92 char	*argv0;			/* root of argv[0] */
93 sigset_t s_mask;		/* signal mask for cleanup critical sect */
94 FILE	*listf = stderr;	/* file pointer to print file list to */
95 char	*tempfile;		/* tempfile to use for mkstemp(3) */
96 char	*tempbase;		/* basename of tempfile to use for mkstemp(3) */
97 
98 /*
99  *	PAX - Portable Archive Interchange
100  *
101  *	A utility to read, write, and write lists of the members of archive
102  *	files and copy directory hierarchies. A variety of archive formats
103  *	are supported (some are described in POSIX 1003.1 10.1):
104  *
105  *		ustar - 10.1.1 extended tar interchange format
106  *		cpio  - 10.1.2 extended cpio interchange format
107  *		tar - old BSD 4.3 tar format
108  *		binary cpio - old cpio with binary header format
109  *		sysVR4 cpio -  with and without CRC
110  *
111  * This version is a superset of IEEE Std 1003.2b-d3
112  *
113  * Summary of Extensions to the IEEE Standard:
114  *
115  * 1	READ ENHANCEMENTS
116  * 1.1	Operations which read archives will continue to operate even when
117  *	processing archives which may be damaged, truncated, or fail to meet
118  *	format specs in several different ways. Damaged sections of archives
119  *	are detected and avoided if possible. Attempts will be made to resync
120  *	archive read operations even with badly damaged media.
121  * 1.2	Blocksize requirements are not strictly enforced on archive read.
122  *	Tapes which have variable sized records can be read without errors.
123  * 1.3	The user can specify via the non-standard option flag -E if error
124  *	resync operation should stop on a media error, try a specified number
125  *	of times to correct, or try to correct forever.
126  * 1.4	Sparse files (lseek holes) stored on the archive (but stored with blocks
127  *	of all zeros will be restored with holes appropriate for the target
128  *	filesystem
129  * 1.5	The user is notified whenever something is found during archive
130  *	read operations which violates spec (but the read will continue).
131  * 1.6	Multiple archive volumes can be read and may span over different
132  *	archive devices
133  * 1.7	Rigidly restores all file attributes exactly as they are stored on the
134  *	archive.
135  * 1.8	Modification change time ranges can be specified via multiple -T
136  *	options. These allow a user to select files whose modification time
137  *	lies within a specific time range.
138  * 1.9	Files can be selected based on owner (user name or uid) via one or more
139  *	-U options.
140  * 1.10	Files can be selected based on group (group name or gid) via one o
141  *	more -G options.
142  * 1.11	File modification time can be checked against existing file after
143  *	name modification (-Z)
144  *
145  * 2	WRITE ENHANCEMENTS
146  * 2.1	Write operation will stop instead of allowing a user to create a flawed
147  *	flawed archive (due to any problem).
148  * 2.2	Archives written by pax are forced to strictly conform to both the
149  *	archive and pax the specific format specifications.
150  * 2.3	Blocking size and format is rigidly enforced on writes.
151  * 2.4	Formats which may exhibit header overflow problems (they have fields
152  *	too small for large file systems, such as inode number storage), use
153  *	routines designed to repair this problem. These techniques still
154  *	conform to both pax and format specifications, but no longer truncate
155  *	these fields. This removes any restrictions on using these archive
156  *	formats on large file systems.
157  * 2.5	Multiple archive volumes can be written and may span over different
158  *	archive devices
159  * 2.6	A archive volume record limit allows the user to specify the number
160  *	of bytes stored on an archive volume. When reached the user is
161  *	prompted for the next archive volume. This is specified with the
162  *	non-standard -B flag. The limit is rounded up to the next blocksize.
163  * 2.7	All archive padding during write use zero filled sections. This makes
164  *	it much easier to pull data out of flawed archive during read
165  *	operations.
166  * 2.8	Access time reset with the -t applies to all file nodes (including
167  *	directories).
168  * 2.9	Symbolic links can be followed with -L (optional in the spec).
169  * 2.10	Modification or inode change time ranges can be specified via
170  *	multiple -T options. These allow a user to select files whose
171  *	modification or inode change time lies within a specific time range.
172  * 2.11	Files can be selected based on owner (user name or uid) via one or more
173  *	-U options.
174  * 2.12	Files can be selected based on group (group name or gid) via one o
175  *	more -G options.
176  * 2.13	Symlinks which appear on the command line can be followed (without
177  *	following other symlinks; -H flag)
178  *
179  * 3	COPY ENHANCEMENTS
180  * 3.1	Sparse files (lseek holes) can be copied without expanding the holes
181  *	into zero filled blocks. The file copy is created with holes which are
182  *	appropriate for the target filesystem
183  * 3.2	Access time as well as modification time on copied file trees can be
184  *	preserved with the appropriate -p options.
185  * 3.3	Access time reset with the -t applies to all file nodes (including
186  *	directories).
187  * 3.4	Symbolic links can be followed with -L (optional in the spec).
188  * 3.5	Modification or inode change time ranges can be specified via
189  *	multiple -T options. These allow a user to select files whose
190  *	modification or inode change time lies within a specific time range.
191  * 3.6	Files can be selected based on owner (user name or uid) via one or more
192  *	-U options.
193  * 3.7	Files can be selected based on group (group name or gid) via one o
194  *	more -G options.
195  * 3.8	Symlinks which appear on the command line can be followed (without
196  *	following other symlinks; -H flag)
197  * 3.9  File inode change time can be checked against existing file before
198  *	name modification (-D)
199  * 3.10 File inode change time can be checked against existing file after
200  *	name modification (-Y)
201  * 3.11	File modification time can be checked against existing file after
202  *	name modification (-Z)
203  *
204  * 4	GENERAL ENHANCEMENTS
205  * 4.1	Internal structure is designed to isolate format dependent and
206  *	independent functions. Formats are selected via a format driver table.
207  *	This encourages the addition of new archive formats by only having to
208  *	write those routines which id, read and write the archive header.
209  */
210 
211 /*
212  * main()
213  *	parse options, set up and operate as specified by the user.
214  *	any operational flaw will set exit_val to non-zero
215  * Return: 0 if ok, 1 otherwise
216  */
217 
218 int
219 main(int argc, char **argv)
220 {
221 	char *tmpdir;
222 	size_t tdlen;
223 
224 	/*
225 	 * Keep a reference to cwd, so we can always come back home.
226 	 */
227 	cwdfd = open(".", O_RDONLY);
228 	if (cwdfd < 0) {
229 		syswarn(1, errno, "Can't open current working directory.");
230 		return(exit_val);
231 	}
232 
233 	/*
234 	 * Where should we put temporary files?
235 	 */
236 	if ((tmpdir = getenv("TMPDIR")) == NULL || *tmpdir == '\0')
237 		tmpdir = _PATH_TMP;
238 	tdlen = strlen(tmpdir);
239 	while (tdlen > 0 && tmpdir[tdlen - 1] == '/')
240 		tdlen--;
241 	tempfile = malloc(tdlen + 1 + sizeof(_TFILE_BASE));
242 	if (tempfile == NULL) {
243 		paxwarn(1, "Cannot allocate memory for temp file name.");
244 		return(exit_val);
245 	}
246 	if (tdlen)
247 		memcpy(tempfile, tmpdir, tdlen);
248 	tempbase = tempfile + tdlen;
249 	*tempbase++ = '/';
250 
251 	/*
252 	 * parse options, determine operational mode, general init
253 	 */
254 	options(argc, argv);
255 	if ((gen_init() < 0) || (tty_init() < 0))
256 		return(exit_val);
257 
258 	/*
259 	 * select a primary operation mode
260 	 */
261 	switch (act) {
262 	case EXTRACT:
263 		extract();
264 		break;
265 	case ARCHIVE:
266 		archive();
267 		break;
268 	case APPND:
269 		if (gzip_program != NULL)
270 			errx(1, "can not gzip while appending");
271 		append();
272 		break;
273 	case COPY:
274 		copy();
275 		break;
276 	default:
277 	case LIST:
278 		list();
279 		break;
280 	}
281 	return(exit_val);
282 }
283 
284 /*
285  * sig_cleanup()
286  *	when interrupted we try to do whatever delayed processing we can.
287  *	This is not critical, but we really ought to limit our damage when we
288  *	are aborted by the user.
289  * Return:
290  *	never....
291  */
292 
293 void
294 sig_cleanup(int which_sig)
295 {
296 	char errbuf[80];
297 
298 	/*
299 	 * restore modes and times for any dirs we may have created
300 	 * or any dirs we may have read. Set vflag and vfpart so the user
301 	 * will clearly see the message on a line by itself.
302 	 */
303 	vflag = vfpart = 1;
304 
305 	/* paxwarn() uses stdio; fake it as well as we can */
306 	if (which_sig == SIGXCPU)
307 		strlcpy(errbuf, "CPU time limit reached, cleaning up.\n",
308 		    sizeof errbuf);
309 	else
310 		strlcpy(errbuf, "Signal caught, cleaning up.\n",
311 		    sizeof errbuf);
312 	(void) write(STDERR_FILENO, errbuf, strlen(errbuf));
313 
314 	ar_close();			/* XXX signal race */
315 	proc_dir();			/* XXX signal race */
316 	if (tflag)
317 		atdir_end();		/* XXX signal race */
318 	_exit(1);
319 }
320 
321 /*
322  * gen_init()
323  *	general setup routines. Not all are required, but they really help
324  *	when dealing with a medium to large sized archives.
325  */
326 
327 static int
328 gen_init(void)
329 {
330 	struct rlimit reslimit;
331 	struct sigaction n_hand;
332 	struct sigaction o_hand;
333 
334 	/*
335 	 * Really needed to handle large archives. We can run out of memory for
336 	 * internal tables really fast when we have a whole lot of files...
337 	 */
338 	if (getrlimit(RLIMIT_DATA , &reslimit) == 0){
339 		reslimit.rlim_cur = reslimit.rlim_max;
340 		(void)setrlimit(RLIMIT_DATA , &reslimit);
341 	}
342 
343 	/*
344 	 * should file size limits be waived? if the os limits us, this is
345 	 * needed if we want to write a large archive
346 	 */
347 	if (getrlimit(RLIMIT_FSIZE , &reslimit) == 0){
348 		reslimit.rlim_cur = reslimit.rlim_max;
349 		(void)setrlimit(RLIMIT_FSIZE , &reslimit);
350 	}
351 
352 	/*
353 	 * increase the size the stack can grow to
354 	 */
355 	if (getrlimit(RLIMIT_STACK , &reslimit) == 0){
356 		reslimit.rlim_cur = reslimit.rlim_max;
357 		(void)setrlimit(RLIMIT_STACK , &reslimit);
358 	}
359 
360 	/*
361 	 * not really needed, but doesn't hurt
362 	 */
363 	if (getrlimit(RLIMIT_RSS , &reslimit) == 0){
364 		reslimit.rlim_cur = reslimit.rlim_max;
365 		(void)setrlimit(RLIMIT_RSS , &reslimit);
366 	}
367 
368 	/*
369 	 * signal handling to reset stored directory times and modes. Since
370 	 * we deal with broken pipes via failed writes we ignore it. We also
371 	 * deal with any file size limit through failed writes. Cpu time
372 	 * limits are caught and a cleanup is forced.
373 	 */
374 	if ((sigemptyset(&s_mask) < 0) || (sigaddset(&s_mask, SIGTERM) < 0) ||
375 	    (sigaddset(&s_mask,SIGINT) < 0)||(sigaddset(&s_mask,SIGHUP) < 0) ||
376 	    (sigaddset(&s_mask,SIGPIPE) < 0)||(sigaddset(&s_mask,SIGQUIT)<0) ||
377 	    (sigaddset(&s_mask,SIGXCPU) < 0)||(sigaddset(&s_mask,SIGXFSZ)<0)) {
378 		paxwarn(1, "Unable to set up signal mask");
379 		return(-1);
380 	}
381 	memset(&n_hand, 0, sizeof n_hand);
382 	n_hand.sa_mask = s_mask;
383 	n_hand.sa_flags = 0;
384 	n_hand.sa_handler = sig_cleanup;
385 
386 	if ((sigaction(SIGHUP, &n_hand, &o_hand) < 0) ||
387 	    (o_hand.sa_handler == SIG_IGN) &&
388 	    (sigaction(SIGHUP, &o_hand, &o_hand) < 0))
389 		goto out;
390 
391 	if ((sigaction(SIGTERM, &n_hand, &o_hand) < 0) ||
392 	    (o_hand.sa_handler == SIG_IGN) &&
393 	    (sigaction(SIGTERM, &o_hand, &o_hand) < 0))
394 		goto out;
395 
396 	if ((sigaction(SIGINT, &n_hand, &o_hand) < 0) ||
397 	    (o_hand.sa_handler == SIG_IGN) &&
398 	    (sigaction(SIGINT, &o_hand, &o_hand) < 0))
399 		goto out;
400 
401 	if ((sigaction(SIGQUIT, &n_hand, &o_hand) < 0) ||
402 	    (o_hand.sa_handler == SIG_IGN) &&
403 	    (sigaction(SIGQUIT, &o_hand, &o_hand) < 0))
404 		goto out;
405 
406 	if ((sigaction(SIGXCPU, &n_hand, &o_hand) < 0) ||
407 	    (o_hand.sa_handler == SIG_IGN) &&
408 	    (sigaction(SIGXCPU, &o_hand, &o_hand) < 0))
409 		goto out;
410 
411 	n_hand.sa_handler = SIG_IGN;
412 	if ((sigaction(SIGPIPE, &n_hand, &o_hand) < 0) ||
413 	    (sigaction(SIGXFSZ, &n_hand, &o_hand) < 0))
414 		goto out;
415 	return(0);
416 
417     out:
418 	syswarn(1, errno, "Unable to set up signal handler");
419 	return(-1);
420 }
421