xref: /freebsd/bin/pax/pax.c (revision 38f0b757fd84d17d0fc24739a7cda160c4516d81)
1 /*-
2  * Copyright (c) 1992 Keith Muller.
3  * Copyright (c) 1992, 1993
4  *	The Regents of the University of California.  All rights reserved.
5  *
6  * This code is derived from software contributed to Berkeley by
7  * Keith Muller of the University of California, San Diego.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 #if 0
35 #ifndef lint
36 static char const copyright[] =
37 "@(#) Copyright (c) 1992, 1993\n\
38 	The Regents of the University of California.  All rights reserved.\n";
39 #endif /* not lint */
40 
41 #ifndef lint
42 static char sccsid[] = "@(#)pax.c	8.2 (Berkeley) 4/18/94";
43 #endif /* not lint */
44 #endif
45 #include <sys/cdefs.h>
46 __FBSDID("$FreeBSD$");
47 
48 #include <sys/types.h>
49 #include <sys/stat.h>
50 #include <sys/time.h>
51 #include <sys/resource.h>
52 #include <err.h>
53 #include <errno.h>
54 #include <fcntl.h>
55 #include <locale.h>
56 #include <paths.h>
57 #include <signal.h>
58 #include <stdio.h>
59 #include <stdlib.h>
60 #include <string.h>
61 #include <unistd.h>
62 #include "pax.h"
63 #include "extern.h"
64 static int gen_init(void);
65 
66 /*
67  * PAX main routines, general globals and some simple start up routines
68  */
69 
70 /*
71  * Variables that can be accessed by any routine within pax
72  */
73 int	act = DEFOP;		/* read/write/append/copy */
74 FSUB	*frmt = NULL;		/* archive format type */
75 int	cflag;			/* match all EXCEPT pattern/file */
76 int	cwdfd;			/* starting cwd */
77 int	dflag;			/* directory member match only  */
78 int	iflag;			/* interactive file/archive rename */
79 int	kflag;			/* do not overwrite existing files */
80 int	lflag;			/* use hard links when possible */
81 int	nflag;			/* select first archive member match */
82 int	tflag;			/* restore access time after read */
83 int	uflag;			/* ignore older modification time files */
84 int	vflag;			/* produce verbose output */
85 int	Dflag;			/* same as uflag except inode change time */
86 int	Hflag;			/* follow command line symlinks (write only) */
87 int	Lflag;			/* follow symlinks when writing */
88 int	Xflag;			/* archive files with same device id only */
89 int	Yflag;			/* same as Dflg except after name mode */
90 int	Zflag;			/* same as uflg except after name mode */
91 int	vfpart;			/* is partial verbose output in progress */
92 int	patime = 1;		/* preserve file access time */
93 int	pmtime = 1;		/* preserve file modification times */
94 int	nodirs;			/* do not create directories as needed */
95 int	pmode;			/* preserve file mode bits */
96 int	pids;			/* preserve file uid/gid */
97 int	rmleadslash = 0;	/* remove leading '/' from pathnames */
98 int	exit_val;		/* exit value */
99 int	docrc;			/* check/create file crc */
100 char	*dirptr;		/* destination dir in a copy */
101 const	char *argv0;		/* root of argv[0] */
102 sigset_t s_mask;		/* signal mask for cleanup critical sect */
103 FILE	*listf;			/* file pointer to print file list to */
104 char	*tempfile;		/* tempfile to use for mkstemp(3) */
105 char	*tempbase;		/* basename of tempfile to use for mkstemp(3) */
106 
107 /*
108  *	PAX - Portable Archive Interchange
109  *
110  * 	A utility to read, write, and write lists of the members of archive
111  *	files and copy directory hierarchies. A variety of archive formats
112  *	are supported (some are described in POSIX 1003.1 10.1):
113  *
114  *		ustar - 10.1.1 extended tar interchange format
115  *		cpio  - 10.1.2 extended cpio interchange format
116  *		tar - old BSD 4.3 tar format
117  *		binary cpio - old cpio with binary header format
118  *		sysVR4 cpio -  with and without CRC
119  *
120  * This version is a superset of IEEE Std 1003.2b-d3
121  *
122  * Summary of Extensions to the IEEE Standard:
123  *
124  * 1	READ ENHANCEMENTS
125  * 1.1	Operations which read archives will continue to operate even when
126  *	processing archives which may be damaged, truncated, or fail to meet
127  *	format specs in several different ways. Damaged sections of archives
128  *	are detected and avoided if possible. Attempts will be made to resync
129  *	archive read operations even with badly damaged media.
130  * 1.2	Blocksize requirements are not strictly enforced on archive read.
131  *	Tapes which have variable sized records can be read without errors.
132  * 1.3	The user can specify via the non-standard option flag -E if error
133  *	resync operation should stop on a media error, try a specified number
134  *	of times to correct, or try to correct forever.
135  * 1.4	Sparse files (lseek holes) stored on the archive (but stored with blocks
136  *	of all zeros will be restored with holes appropriate for the target
137  *	file system
138  * 1.5	The user is notified whenever something is found during archive
139  *	read operations which violates spec (but the read will continue).
140  * 1.6	Multiple archive volumes can be read and may span over different
141  *	archive devices
142  * 1.7	Rigidly restores all file attributes exactly as they are stored on the
143  *	archive.
144  * 1.8	Modification change time ranges can be specified via multiple -T
145  *	options. These allow a user to select files whose modification time
146  *	lies within a specific time range.
147  * 1.9	Files can be selected based on owner (user name or uid) via one or more
148  *	-U options.
149  * 1.10	Files can be selected based on group (group name or gid) via one o
150  *	more -G options.
151  * 1.11	File modification time can be checked against existing file after
152  *	name modification (-Z)
153  *
154  * 2	WRITE ENHANCEMENTS
155  * 2.1	Write operation will stop instead of allowing a user to create a flawed
156  *	flawed archive (due to any problem).
157  * 2.2	Archives written by pax are forced to strictly conform to both the
158  *	archive and pax the specific format specifications.
159  * 2.3	Blocking size and format is rigidly enforced on writes.
160  * 2.4	Formats which may exhibit header overflow problems (they have fields
161  *	too small for large file systems, such as inode number storage), use
162  *	routines designed to repair this problem. These techniques still
163  *	conform to both pax and format specifications, but no longer truncate
164  *	these fields. This removes any restrictions on using these archive
165  *	formats on large file systems.
166  * 2.5	Multiple archive volumes can be written and may span over different
167  *	archive devices
168  * 2.6	A archive volume record limit allows the user to specify the number
169  *	of bytes stored on an archive volume. When reached the user is
170  *	prompted for the next archive volume. This is specified with the
171  *	non-standard -B flag. The limit is rounded up to the next blocksize.
172  * 2.7	All archive padding during write use zero filled sections. This makes
173  *	it much easier to pull data out of flawed archive during read
174  *	operations.
175  * 2.8	Access time reset with the -t applies to all file nodes (including
176  *	directories).
177  * 2.9	Symbolic links can be followed with -L (optional in the spec).
178  * 2.10	Modification or inode change time ranges can be specified via
179  *	multiple -T options. These allow a user to select files whose
180  *	modification or inode change time lies within a specific time range.
181  * 2.11	Files can be selected based on owner (user name or uid) via one or more
182  *	-U options.
183  * 2.12	Files can be selected based on group (group name or gid) via one o
184  *	more -G options.
185  * 2.13	Symlinks which appear on the command line can be followed (without
186  *	following other symlinks; -H flag)
187  *
188  * 3	COPY ENHANCEMENTS
189  * 3.1	Sparse files (lseek holes) can be copied without expanding the holes
190  *	into zero filled blocks. The file copy is created with holes which are
191  *	appropriate for the target file system
192  * 3.2	Access time as well as modification time on copied file trees can be
193  *	preserved with the appropriate -p options.
194  * 3.3	Access time reset with the -t applies to all file nodes (including
195  *	directories).
196  * 3.4	Symbolic links can be followed with -L (optional in the spec).
197  * 3.5	Modification or inode change time ranges can be specified via
198  *	multiple -T options. These allow a user to select files whose
199  *	modification or inode change time lies within a specific time range.
200  * 3.6	Files can be selected based on owner (user name or uid) via one or more
201  *	-U options.
202  * 3.7	Files can be selected based on group (group name or gid) via one o
203  *	more -G options.
204  * 3.8	Symlinks which appear on the command line can be followed (without
205  *	following other symlinks; -H flag)
206  * 3.9  File inode change time can be checked against existing file before
207  *	name modification (-D)
208  * 3.10 File inode change time can be checked against existing file after
209  *	name modification (-Y)
210  * 3.11	File modification time can be checked against existing file after
211  *	name modification (-Z)
212  *
213  * 4	GENERAL ENHANCEMENTS
214  * 4.1	Internal structure is designed to isolate format dependent and
215  *	independent functions. Formats are selected via a format driver table.
216  *	This encourages the addition of new archive formats by only having to
217  *	write those routines which id, read and write the archive header.
218  */
219 
220 /*
221  * main()
222  *	parse options, set up and operate as specified by the user.
223  *	any operational flaw will set exit_val to non-zero
224  * Return: 0 if ok, 1 otherwise
225  */
226 
227 int
228 main(int argc, char *argv[])
229 {
230 	const char *tmpdir;
231 	size_t tdlen;
232 
233 	(void) setlocale(LC_ALL, "");
234 	listf = stderr;
235 	/*
236 	 * Keep a reference to cwd, so we can always come back home.
237 	 */
238 	cwdfd = open(".", O_RDONLY);
239 	if (cwdfd < 0) {
240 		syswarn(0, errno, "Can't open current working directory.");
241 		return(exit_val);
242 	}
243 
244 	/*
245 	 * Where should we put temporary files?
246 	 */
247 	if ((tmpdir = getenv("TMPDIR")) == NULL || *tmpdir == '\0')
248 		tmpdir = _PATH_TMP;
249 	tdlen = strlen(tmpdir);
250 	while(tdlen > 0 && tmpdir[tdlen - 1] == '/')
251 		tdlen--;
252 	tempfile = malloc(tdlen + 1 + sizeof(_TFILE_BASE));
253 	if (tempfile == NULL) {
254 		paxwarn(1, "Cannot allocate memory for temp file name.");
255 		return(exit_val);
256 	}
257 	if (tdlen)
258 		memcpy(tempfile, tmpdir, tdlen);
259 	tempbase = tempfile + tdlen;
260 	*tempbase++ = '/';
261 
262 	/*
263 	 * parse options, determine operational mode, general init
264 	 */
265 	options(argc, argv);
266 	if ((gen_init() < 0) || (tty_init() < 0))
267 		return(exit_val);
268 
269 	/*
270 	 * select a primary operation mode
271 	 */
272 	switch(act) {
273 	case EXTRACT:
274 		extract();
275 		break;
276 	case ARCHIVE:
277 		archive();
278 		break;
279 	case APPND:
280 		if (gzip_program != NULL)
281 			err(1, "can not gzip while appending");
282 		append();
283 		break;
284 	case COPY:
285 		copy();
286 		break;
287 	default:
288 	case LIST:
289 		list();
290 		break;
291 	}
292 	return(exit_val);
293 }
294 
295 /*
296  * sig_cleanup()
297  *	when interrupted we try to do whatever delayed processing we can.
298  *	This is not critical, but we really ought to limit our damage when we
299  *	are aborted by the user.
300  * Return:
301  *	never....
302  */
303 
304 void
305 sig_cleanup(int which_sig)
306 {
307 	/*
308 	 * restore modes and times for any dirs we may have created
309 	 * or any dirs we may have read. Set vflag and vfpart so the user
310 	 * will clearly see the message on a line by itself.
311 	 */
312 	vflag = vfpart = 1;
313 	if (which_sig == SIGXCPU)
314 		paxwarn(0, "Cpu time limit reached, cleaning up.");
315 	else
316 		paxwarn(0, "Signal caught, cleaning up.");
317 
318 	ar_close();
319 	proc_dir();
320 	if (tflag)
321 		atdir_end();
322 	exit(1);
323 }
324 
325 /*
326  * gen_init()
327  *	general setup routines. Not all are required, but they really help
328  *	when dealing with a medium to large sized archives.
329  */
330 
331 static int
332 gen_init(void)
333 {
334 	struct rlimit reslimit;
335 	struct sigaction n_hand;
336 	struct sigaction o_hand;
337 
338 	/*
339 	 * Really needed to handle large archives. We can run out of memory for
340 	 * internal tables really fast when we have a whole lot of files...
341 	 */
342 	if (getrlimit(RLIMIT_DATA , &reslimit) == 0){
343 		reslimit.rlim_cur = reslimit.rlim_max;
344 		(void)setrlimit(RLIMIT_DATA , &reslimit);
345 	}
346 
347 	/*
348 	 * should file size limits be waived? if the os limits us, this is
349 	 * needed if we want to write a large archive
350 	 */
351 	if (getrlimit(RLIMIT_FSIZE , &reslimit) == 0){
352 		reslimit.rlim_cur = reslimit.rlim_max;
353 		(void)setrlimit(RLIMIT_FSIZE , &reslimit);
354 	}
355 
356 	/*
357 	 * increase the size the stack can grow to
358 	 */
359 	if (getrlimit(RLIMIT_STACK , &reslimit) == 0){
360 		reslimit.rlim_cur = reslimit.rlim_max;
361 		(void)setrlimit(RLIMIT_STACK , &reslimit);
362 	}
363 
364 	/*
365 	 * not really needed, but doesn't hurt
366 	 */
367 	if (getrlimit(RLIMIT_RSS , &reslimit) == 0){
368 		reslimit.rlim_cur = reslimit.rlim_max;
369 		(void)setrlimit(RLIMIT_RSS , &reslimit);
370 	}
371 
372 	/*
373 	 * signal handling to reset stored directory times and modes. Since
374 	 * we deal with broken pipes via failed writes we ignore it. We also
375 	 * deal with any file size limit thorough failed writes. Cpu time
376 	 * limits are caught and a cleanup is forced.
377 	 */
378 	if ((sigemptyset(&s_mask) < 0) || (sigaddset(&s_mask, SIGTERM) < 0) ||
379 	    (sigaddset(&s_mask,SIGINT) < 0)||(sigaddset(&s_mask,SIGHUP) < 0) ||
380 	    (sigaddset(&s_mask,SIGPIPE) < 0)||(sigaddset(&s_mask,SIGQUIT)<0) ||
381 	    (sigaddset(&s_mask,SIGXCPU) < 0)||(sigaddset(&s_mask,SIGXFSZ)<0)) {
382 		paxwarn(1, "Unable to set up signal mask");
383 		return(-1);
384 	}
385 	memset(&n_hand, 0, sizeof n_hand);
386 	n_hand.sa_mask = s_mask;
387 	n_hand.sa_flags = 0;
388 	n_hand.sa_handler = sig_cleanup;
389 
390 	if ((sigaction(SIGHUP, &n_hand, &o_hand) < 0) &&
391 	    (o_hand.sa_handler == SIG_IGN) &&
392 	    (sigaction(SIGHUP, &o_hand, &o_hand) < 0))
393 		goto out;
394 
395 	if ((sigaction(SIGTERM, &n_hand, &o_hand) < 0) &&
396 	    (o_hand.sa_handler == SIG_IGN) &&
397 	    (sigaction(SIGTERM, &o_hand, &o_hand) < 0))
398 		goto out;
399 
400 	if ((sigaction(SIGINT, &n_hand, &o_hand) < 0) &&
401 	    (o_hand.sa_handler == SIG_IGN) &&
402 	    (sigaction(SIGINT, &o_hand, &o_hand) < 0))
403 		goto out;
404 
405 	if ((sigaction(SIGQUIT, &n_hand, &o_hand) < 0) &&
406 	    (o_hand.sa_handler == SIG_IGN) &&
407 	    (sigaction(SIGQUIT, &o_hand, &o_hand) < 0))
408 		goto out;
409 
410 	if ((sigaction(SIGXCPU, &n_hand, &o_hand) < 0) &&
411 	    (o_hand.sa_handler == SIG_IGN) &&
412 	    (sigaction(SIGXCPU, &o_hand, &o_hand) < 0))
413 		goto out;
414 
415 	n_hand.sa_handler = SIG_IGN;
416 	if ((sigaction(SIGPIPE, &n_hand, &o_hand) < 0) ||
417 	    (sigaction(SIGXFSZ, &n_hand, &o_hand) < 0))
418 		goto out;
419 	return(0);
420 
421     out:
422 	syswarn(1, errno, "Unable to set up signal handler");
423 	return(-1);
424 }
425