xref: /freebsd/bin/pax/pax.c (revision 5e53a4f90f82c4345f277dd87cc9292f26e04a29)
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 #if 0
37 #ifndef lint
38 static char const copyright[] =
39 "@(#) Copyright (c) 1992, 1993\n\
40 	The Regents of the University of California.  All rights reserved.\n";
41 #endif /* not lint */
42 
43 #ifndef lint
44 static char sccsid[] = "@(#)pax.c	8.2 (Berkeley) 4/18/94";
45 #endif /* not lint */
46 #endif
47 #include <sys/cdefs.h>
48 __FBSDID("$FreeBSD$");
49 
50 #include <sys/types.h>
51 #include <sys/stat.h>
52 #include <sys/time.h>
53 #include <sys/resource.h>
54 #include <err.h>
55 #include <errno.h>
56 #include <fcntl.h>
57 #include <locale.h>
58 #include <paths.h>
59 #include <signal.h>
60 #include <stdio.h>
61 #include <stdlib.h>
62 #include <string.h>
63 #include "pax.h"
64 #include "extern.h"
65 static int gen_init(void);
66 
67 /*
68  * PAX main routines, general globals and some simple start up routines
69  */
70 
71 /*
72  * Variables that can be accessed by any routine within pax
73  */
74 int	act = DEFOP;		/* read/write/append/copy */
75 FSUB	*frmt = NULL;		/* archive format type */
76 int	cflag;			/* match all EXCEPT pattern/file */
77 int	cwdfd;			/* starting cwd */
78 int	dflag;			/* directory member match only  */
79 int	iflag;			/* interactive file/archive rename */
80 int	kflag;			/* do not overwrite existing files */
81 int	lflag;			/* use hard links when possible */
82 int	nflag;			/* select first archive member match */
83 int	tflag;			/* restore access time after read */
84 int	uflag;			/* ignore older modification time files */
85 int	vflag;			/* produce verbose output */
86 int	Dflag;			/* same as uflag except inode change time */
87 int	Hflag;			/* follow command line symlinks (write only) */
88 int	Lflag;			/* follow symlinks when writing */
89 int	Oflag;			/* limit to single volume */
90 int	Xflag;			/* archive files with same device id only */
91 int	Yflag;			/* same as Dflg except after name mode */
92 int	Zflag;			/* same as uflg except after name mode */
93 int	vfpart;			/* is partial verbose output in progress */
94 int	patime = 1;		/* preserve file access time */
95 int	pmtime = 1;		/* preserve file modification times */
96 int	nodirs;			/* do not create directories as needed */
97 int	pmode;			/* preserve file mode bits */
98 int	pids;			/* preserve file uid/gid */
99 int	rmleadslash = 0;	/* remove leading '/' from pathnames */
100 int	exit_val;		/* exit value */
101 int	docrc;			/* check/create file crc */
102 char	*dirptr;		/* destination dir in a copy */
103 const	char *argv0;		/* root of argv[0] */
104 sigset_t s_mask;		/* signal mask for cleanup critical sect */
105 FILE	*listf;			/* file pointer to print file list to */
106 char	*tempfile;		/* tempfile to use for mkstemp(3) */
107 char	*tempbase;		/* basename of tempfile to use for mkstemp(3) */
108 
109 /*
110  *	PAX - Portable Archive Interchange
111  *
112  * 	A utility to read, write, and write lists of the members of archive
113  *	files and copy directory hierarchies. A variety of archive formats
114  *	are supported (some are described in POSIX 1003.1 10.1):
115  *
116  *		ustar - 10.1.1 extended tar interchange format
117  *		cpio  - 10.1.2 extended cpio interchange format
118  *		tar - old BSD 4.3 tar format
119  *		binary cpio - old cpio with binary header format
120  *		sysVR4 cpio -  with and without CRC
121  *
122  * This version is a superset of IEEE Std 1003.2b-d3
123  *
124  * Summary of Extensions to the IEEE Standard:
125  *
126  * 1	READ ENHANCEMENTS
127  * 1.1	Operations which read archives will continue to operate even when
128  *	processing archives which may be damaged, truncated, or fail to meet
129  *	format specs in several different ways. Damaged sections of archives
130  *	are detected and avoided if possible. Attempts will be made to resync
131  *	archive read operations even with badly damaged media.
132  * 1.2	Blocksize requirements are not strictly enforced on archive read.
133  *	Tapes which have variable sized records can be read without errors.
134  * 1.3	The user can specify via the non-standard option flag -E if error
135  *	resync operation should stop on a media error, try a specified number
136  *	of times to correct, or try to correct forever.
137  * 1.4	Sparse files (lseek holes) stored on the archive (but stored with blocks
138  *	of all zeros will be restored with holes appropriate for the target
139  *	file system
140  * 1.5	The user is notified whenever something is found during archive
141  *	read operations which violates spec (but the read will continue).
142  * 1.6	Multiple archive volumes can be read and may span over different
143  *	archive devices
144  * 1.7	Rigidly restores all file attributes exactly as they are stored on the
145  *	archive.
146  * 1.8	Modification change time ranges can be specified via multiple -T
147  *	options. These allow a user to select files whose modification time
148  *	lies within a specific time range.
149  * 1.9	Files can be selected based on owner (user name or uid) via one or more
150  *	-U options.
151  * 1.10	Files can be selected based on group (group name or gid) via one o
152  *	more -G options.
153  * 1.11	File modification time can be checked against existing file after
154  *	name modification (-Z)
155  *
156  * 2	WRITE ENHANCEMENTS
157  * 2.1	Write operation will stop instead of allowing a user to create a flawed
158  *	flawed archive (due to any problem).
159  * 2.2	Archives written by pax are forced to strictly conform to both the
160  *	archive and pax the specific format specifications.
161  * 2.3	Blocking size and format is rigidly enforced on writes.
162  * 2.4	Formats which may exhibit header overflow problems (they have fields
163  *	too small for large file systems, such as inode number storage), use
164  *	routines designed to repair this problem. These techniques still
165  *	conform to both pax and format specifications, but no longer truncate
166  *	these fields. This removes any restrictions on using these archive
167  *	formats on large file systems.
168  * 2.5	Multiple archive volumes can be written and may span over different
169  *	archive devices
170  * 2.6	A archive volume record limit allows the user to specify the number
171  *	of bytes stored on an archive volume. When reached the user is
172  *	prompted for the next archive volume. This is specified with the
173  *	non-standard -B flag. The limit is rounded up to the next blocksize.
174  * 2.7	All archive padding during write use zero filled sections. This makes
175  *	it much easier to pull data out of flawed archive during read
176  *	operations.
177  * 2.8	Access time reset with the -t applies to all file nodes (including
178  *	directories).
179  * 2.9	Symbolic links can be followed with -L (optional in the spec).
180  * 2.10	Modification or inode change time ranges can be specified via
181  *	multiple -T options. These allow a user to select files whose
182  *	modification or inode change time lies within a specific time range.
183  * 2.11	Files can be selected based on owner (user name or uid) via one or more
184  *	-U options.
185  * 2.12	Files can be selected based on group (group name or gid) via one o
186  *	more -G options.
187  * 2.13	Symlinks which appear on the command line can be followed (without
188  *	following other symlinks; -H flag)
189  *
190  * 3	COPY ENHANCEMENTS
191  * 3.1	Sparse files (lseek holes) can be copied without expanding the holes
192  *	into zero filled blocks. The file copy is created with holes which are
193  *	appropriate for the target file system
194  * 3.2	Access time as well as modification time on copied file trees can be
195  *	preserved with the appropriate -p options.
196  * 3.3	Access time reset with the -t applies to all file nodes (including
197  *	directories).
198  * 3.4	Symbolic links can be followed with -L (optional in the spec).
199  * 3.5	Modification or inode change time ranges can be specified via
200  *	multiple -T options. These allow a user to select files whose
201  *	modification or inode change time lies within a specific time range.
202  * 3.6	Files can be selected based on owner (user name or uid) via one or more
203  *	-U options.
204  * 3.7	Files can be selected based on group (group name or gid) via one o
205  *	more -G options.
206  * 3.8	Symlinks which appear on the command line can be followed (without
207  *	following other symlinks; -H flag)
208  * 3.9  File inode change time can be checked against existing file before
209  *	name modification (-D)
210  * 3.10 File inode change time can be checked against existing file after
211  *	name modification (-Y)
212  * 3.11	File modification time can be checked against existing file after
213  *	name modification (-Z)
214  *
215  * 4	GENERAL ENHANCEMENTS
216  * 4.1	Internal structure is designed to isolate format dependent and
217  *	independent functions. Formats are selected via a format driver table.
218  *	This encourages the addition of new archive formats by only having to
219  *	write those routines which id, read and write the archive header.
220  */
221 
222 /*
223  * main()
224  *	parse options, set up and operate as specified by the user.
225  *	any operational flaw will set exit_val to non-zero
226  * Return: 0 if ok, 1 otherwise
227  */
228 
229 int
230 main(int argc, char *argv[])
231 {
232 	const char *tmpdir;
233 	size_t tdlen;
234 
235 	(void) setlocale(LC_ALL, "");
236 	listf = stderr;
237 	/*
238 	 * Keep a reference to cwd, so we can always come back home.
239 	 */
240 	cwdfd = open(".", O_RDONLY);
241 	if (cwdfd < 0) {
242 		syswarn(0, errno, "Can't open current working directory.");
243 		return(exit_val);
244 	}
245 
246 	/*
247 	 * Where should we put temporary files?
248 	 */
249 	if ((tmpdir = getenv("TMPDIR")) == NULL || *tmpdir == '\0')
250 		tmpdir = _PATH_TMP;
251 	tdlen = strlen(tmpdir);
252 	while(tdlen > 0 && tmpdir[tdlen - 1] == '/')
253 		tdlen--;
254 	tempfile = malloc(tdlen + 1 + sizeof(_TFILE_BASE));
255 	if (tempfile == NULL) {
256 		paxwarn(1, "Cannot allocate memory for temp file name.");
257 		return(exit_val);
258 	}
259 	if (tdlen)
260 		memcpy(tempfile, tmpdir, tdlen);
261 	tempbase = tempfile + tdlen;
262 	*tempbase++ = '/';
263 
264 	/*
265 	 * parse options, determine operational mode, general init
266 	 */
267 	options(argc, argv);
268 	if ((gen_init() < 0) || (tty_init() < 0))
269 		return(exit_val);
270 
271 	/*
272 	 * select a primary operation mode
273 	 */
274 	switch(act) {
275 	case EXTRACT:
276 		extract();
277 		break;
278 	case ARCHIVE:
279 		archive();
280 		break;
281 	case APPND:
282 		if (gzip_program != NULL)
283 			err(1, "can not gzip while appending");
284 		append();
285 		break;
286 	case COPY:
287 		copy();
288 		break;
289 	default:
290 	case LIST:
291 		list();
292 		break;
293 	}
294 	return(exit_val);
295 }
296 
297 /*
298  * sig_cleanup()
299  *	when interrupted we try to do whatever delayed processing we can.
300  *	This is not critical, but we really ought to limit our damage when we
301  *	are aborted by the user.
302  * Return:
303  *	never....
304  */
305 
306 void
307 sig_cleanup(int which_sig)
308 {
309 	/*
310 	 * restore modes and times for any dirs we may have created
311 	 * or any dirs we may have read. Set vflag and vfpart so the user
312 	 * will clearly see the message on a line by itself.
313 	 */
314 	vflag = vfpart = 1;
315 	if (which_sig == SIGXCPU)
316 		paxwarn(0, "Cpu time limit reached, cleaning up.");
317 	else
318 		paxwarn(0, "Signal caught, cleaning up.");
319 
320 	ar_close();
321 	proc_dir();
322 	if (tflag)
323 		atdir_end();
324 	exit(1);
325 }
326 
327 /*
328  * gen_init()
329  *	general setup routines. Not all are required, but they really help
330  *	when dealing with a medium to large sized archives.
331  */
332 
333 static int
334 gen_init(void)
335 {
336 	struct rlimit reslimit;
337 	struct sigaction n_hand;
338 	struct sigaction o_hand;
339 
340 	/*
341 	 * Really needed to handle large archives. We can run out of memory for
342 	 * internal tables really fast when we have a whole lot of files...
343 	 */
344 	if (getrlimit(RLIMIT_DATA , &reslimit) == 0){
345 		reslimit.rlim_cur = reslimit.rlim_max;
346 		(void)setrlimit(RLIMIT_DATA , &reslimit);
347 	}
348 
349 	/*
350 	 * should file size limits be waived? if the os limits us, this is
351 	 * needed if we want to write a large archive
352 	 */
353 	if (getrlimit(RLIMIT_FSIZE , &reslimit) == 0){
354 		reslimit.rlim_cur = reslimit.rlim_max;
355 		(void)setrlimit(RLIMIT_FSIZE , &reslimit);
356 	}
357 
358 	/*
359 	 * increase the size the stack can grow to
360 	 */
361 	if (getrlimit(RLIMIT_STACK , &reslimit) == 0){
362 		reslimit.rlim_cur = reslimit.rlim_max;
363 		(void)setrlimit(RLIMIT_STACK , &reslimit);
364 	}
365 
366 	/*
367 	 * not really needed, but doesn't hurt
368 	 */
369 	if (getrlimit(RLIMIT_RSS , &reslimit) == 0){
370 		reslimit.rlim_cur = reslimit.rlim_max;
371 		(void)setrlimit(RLIMIT_RSS , &reslimit);
372 	}
373 
374 	/*
375 	 * signal handling to reset stored directory times and modes. Since
376 	 * we deal with broken pipes via failed writes we ignore it. We also
377 	 * deal with any file size limit thorough failed writes. Cpu time
378 	 * limits are caught and a cleanup is forced.
379 	 */
380 	if ((sigemptyset(&s_mask) < 0) || (sigaddset(&s_mask, SIGTERM) < 0) ||
381 	    (sigaddset(&s_mask,SIGINT) < 0)||(sigaddset(&s_mask,SIGHUP) < 0) ||
382 	    (sigaddset(&s_mask,SIGPIPE) < 0)||(sigaddset(&s_mask,SIGQUIT)<0) ||
383 	    (sigaddset(&s_mask,SIGXCPU) < 0)||(sigaddset(&s_mask,SIGXFSZ)<0)) {
384 		paxwarn(1, "Unable to set up signal mask");
385 		return(-1);
386 	}
387 	memset(&n_hand, 0, sizeof n_hand);
388 	n_hand.sa_mask = s_mask;
389 	n_hand.sa_flags = 0;
390 	n_hand.sa_handler = sig_cleanup;
391 
392 	if ((sigaction(SIGHUP, &n_hand, &o_hand) < 0) &&
393 	    (o_hand.sa_handler == SIG_IGN) &&
394 	    (sigaction(SIGHUP, &o_hand, &o_hand) < 0))
395 		goto out;
396 
397 	if ((sigaction(SIGTERM, &n_hand, &o_hand) < 0) &&
398 	    (o_hand.sa_handler == SIG_IGN) &&
399 	    (sigaction(SIGTERM, &o_hand, &o_hand) < 0))
400 		goto out;
401 
402 	if ((sigaction(SIGINT, &n_hand, &o_hand) < 0) &&
403 	    (o_hand.sa_handler == SIG_IGN) &&
404 	    (sigaction(SIGINT, &o_hand, &o_hand) < 0))
405 		goto out;
406 
407 	if ((sigaction(SIGQUIT, &n_hand, &o_hand) < 0) &&
408 	    (o_hand.sa_handler == SIG_IGN) &&
409 	    (sigaction(SIGQUIT, &o_hand, &o_hand) < 0))
410 		goto out;
411 
412 	if ((sigaction(SIGXCPU, &n_hand, &o_hand) < 0) &&
413 	    (o_hand.sa_handler == SIG_IGN) &&
414 	    (sigaction(SIGXCPU, &o_hand, &o_hand) < 0))
415 		goto out;
416 
417 	n_hand.sa_handler = SIG_IGN;
418 	if ((sigaction(SIGPIPE, &n_hand, &o_hand) < 0) ||
419 	    (sigaction(SIGXFSZ, &n_hand, &o_hand) < 0))
420 		goto out;
421 	return(0);
422 
423     out:
424 	syswarn(1, errno, "Unable to set up signal handler");
425 	return(-1);
426 }
427