2 * Copyright (c) 1992 Keith Muller.
3 * Copyright (c) 1992, 1993
4 * The Regents of the University of California. All rights reserved.
6 * This code is derived from software contributed to Berkeley by
7 * Keith Muller of the University of California, San Diego.
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
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 * 3. 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.
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
33 * @(#) Copyright (c) 1992, 1993 The Regents of the University of California. All rights reserved.
34 * @(#)pax.c 8.2 (Berkeley) 4/18/94
35 * $FreeBSD: src/bin/pax/pax.c,v 1.13.2.4 2002/11/07 15:29:53 imp Exp $
36 * $DragonFly: src/bin/pax/pax.c,v 1.6 2006/09/27 21:58:08 pavalos Exp $
39 #include <sys/types.h>
42 #include <sys/resource.h>
55 static int gen_init (void);
58 * PAX main routines, general globals and some simple start up routines
62 * Variables that can be accessed by any routine within pax
64 int act
= DEFOP
; /* read/write/append/copy */
65 FSUB
*frmt
= NULL
; /* archive format type */
66 int cflag
; /* match all EXCEPT pattern/file */
67 int cwdfd
; /* starting cwd */
68 int dflag
; /* directory member match only */
69 int iflag
; /* interactive file/archive rename */
70 int kflag
; /* do not overwrite existing files */
71 int lflag
; /* use hard links when possible */
72 int nflag
; /* select first archive member match */
73 int tflag
; /* restore access time after read */
74 int uflag
; /* ignore older modification time files */
75 int vflag
; /* produce verbose output */
76 int Dflag
; /* same as uflag except inode change time */
77 int Hflag
; /* follow command line symlinks (write only) */
78 int Lflag
; /* follow symlinks when writing */
79 int Xflag
; /* archive files with same device id only */
80 int Yflag
; /* same as Dflag except after name mode */
81 int Zflag
; /* same as uflag except after name mode */
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
; /* 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) */
99 * PAX - Portable Archive Interchange
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):
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
111 * This version is a superset of IEEE Std 1003.2b-d3
113 * Summary of Extensions to the IEEE Standard:
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
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
133 * 1.7 Rigidly restores all file attributes exactly as they are stored on the
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
140 * 1.10 Files can be selected based on group (group name or gid) via one o
142 * 1.11 File modification time can be checked against existing file after
143 * name modification (-Z)
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
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
166 * 2.8 Access time reset with the -t applies to all file nodes (including
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
174 * 2.12 Files can be selected based on group (group name or gid) via one o
176 * 2.13 Symlinks which appear on the command line can be followed (without
177 * following other symlinks; -H flag)
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
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
193 * 3.7 Files can be selected based on group (group name or gid) via one o
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)
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.
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
219 main(int argc
, char **argv
)
224 setlocale(LC_ALL
, "");
227 * Keep a reference to cwd, so we can always come back home.
229 cwdfd
= open(".", O_RDONLY
);
231 syswarn(0, errno
, "Can't open current working directory.");
236 * Where should we put temporary files?
238 if ((tmpdir
= getenv("TMPDIR")) == NULL
|| *tmpdir
== '\0')
240 tdlen
= strlen(tmpdir
);
241 while(tdlen
> 0 && tmpdir
[tdlen
- 1] == '/')
243 tempfile
= malloc(tdlen
+ 1 + sizeof(_TFILE_BASE
));
244 if (tempfile
== NULL
) {
245 paxwarn(1, "Cannot allocate memory for temp file name.");
249 memcpy(tempfile
, tmpdir
, tdlen
);
250 tempbase
= tempfile
+ tdlen
;
254 * parse options, determine operational mode, general init
257 if ((gen_init() < 0) || (tty_init() < 0))
261 * select a primary operation mode
271 if (gzip_program
!= NULL
)
272 err(1, "can not gzip while appending");
288 * when interrupted we try to do whatever delayed processing we can.
289 * This is not critical, but we really ought to limit our damage when we
290 * are aborted by the user.
296 sig_cleanup(int which_sig
)
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.
304 if (which_sig
== SIGXCPU
)
305 paxwarn(0, "Cpu time limit reached, cleaning up.");
307 paxwarn(0, "Signal caught, cleaning up.");
318 * general setup routines. Not all are required, but they really help
319 * when dealing with a medium to large sized archives.
325 struct rlimit reslimit
;
326 struct sigaction n_hand
;
327 struct sigaction o_hand
;
330 * Really needed to handle large archives. We can run out of memory for
331 * internal tables really fast when we have a whole lot of files...
333 if (getrlimit(RLIMIT_DATA
, &reslimit
) == 0){
334 reslimit
.rlim_cur
= reslimit
.rlim_max
;
335 setrlimit(RLIMIT_DATA
, &reslimit
);
339 * should file size limits be waived? if the os limits us, this is
340 * needed if we want to write a large archive
342 if (getrlimit(RLIMIT_FSIZE
, &reslimit
) == 0){
343 reslimit
.rlim_cur
= reslimit
.rlim_max
;
344 setrlimit(RLIMIT_FSIZE
, &reslimit
);
348 * increase the size the stack can grow to
350 if (getrlimit(RLIMIT_STACK
, &reslimit
) == 0){
351 reslimit
.rlim_cur
= reslimit
.rlim_max
;
352 setrlimit(RLIMIT_STACK
, &reslimit
);
356 * not really needed, but doesn't hurt
358 if (getrlimit(RLIMIT_RSS
, &reslimit
) == 0){
359 reslimit
.rlim_cur
= reslimit
.rlim_max
;
360 setrlimit(RLIMIT_RSS
, &reslimit
);
364 * signal handling to reset stored directory times and modes. Since
365 * we deal with broken pipes via failed writes we ignore it. We also
366 * deal with any file size limit through failed writes. Cpu time
367 * limits are caught and a cleanup is forced.
369 if ((sigemptyset(&s_mask
) < 0) || (sigaddset(&s_mask
, SIGTERM
) < 0) ||
370 (sigaddset(&s_mask
,SIGINT
) < 0)||(sigaddset(&s_mask
,SIGHUP
) < 0) ||
371 (sigaddset(&s_mask
,SIGPIPE
) < 0)||(sigaddset(&s_mask
,SIGQUIT
)<0) ||
372 (sigaddset(&s_mask
,SIGXCPU
) < 0)||(sigaddset(&s_mask
,SIGXFSZ
)<0)) {
373 paxwarn(1, "Unable to set up signal mask");
376 memset(&n_hand
, 0, sizeof n_hand
);
377 n_hand
.sa_mask
= s_mask
;
379 n_hand
.sa_handler
= sig_cleanup
;
381 if ((sigaction(SIGHUP
, &n_hand
, &o_hand
) < 0) &&
382 (o_hand
.sa_handler
== SIG_IGN
) &&
383 (sigaction(SIGHUP
, &o_hand
, &o_hand
) < 0))
386 if ((sigaction(SIGTERM
, &n_hand
, &o_hand
) < 0) &&
387 (o_hand
.sa_handler
== SIG_IGN
) &&
388 (sigaction(SIGTERM
, &o_hand
, &o_hand
) < 0))
391 if ((sigaction(SIGINT
, &n_hand
, &o_hand
) < 0) &&
392 (o_hand
.sa_handler
== SIG_IGN
) &&
393 (sigaction(SIGINT
, &o_hand
, &o_hand
) < 0))
396 if ((sigaction(SIGQUIT
, &n_hand
, &o_hand
) < 0) &&
397 (o_hand
.sa_handler
== SIG_IGN
) &&
398 (sigaction(SIGQUIT
, &o_hand
, &o_hand
) < 0))
401 if ((sigaction(SIGXCPU
, &n_hand
, &o_hand
) < 0) &&
402 (o_hand
.sa_handler
== SIG_IGN
) &&
403 (sigaction(SIGXCPU
, &o_hand
, &o_hand
) < 0))
406 n_hand
.sa_handler
= SIG_IGN
;
407 if ((sigaction(SIGPIPE
, &n_hand
, &o_hand
) < 0) ||
408 (sigaction(SIGXFSZ
, &n_hand
, &o_hand
) < 0))
413 syswarn(1, errno
, "Unable to set up signal handler");