xref: /freebsd/bin/dd/dd.c (revision 07d397d74651435ff6f4d474876589e85967350e)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1991, 1993, 1994
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software contributed to Berkeley by
8  * Keith Muller of the University of California, San Diego and Lance
9  * Visser of Convex Computer Corporation.
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) 1991, 1993, 1994\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[] = "@(#)dd.c	8.5 (Berkeley) 4/2/94";
45 #endif /* not lint */
46 #endif
47 #include <sys/cdefs.h>
48 __FBSDID("$FreeBSD$");
49 
50 #include <sys/param.h>
51 #include <sys/stat.h>
52 #include <sys/capsicum.h>
53 #include <sys/conf.h>
54 #include <sys/disklabel.h>
55 #include <sys/filio.h>
56 #include <sys/mtio.h>
57 #include <sys/time.h>
58 
59 #include <assert.h>
60 #include <capsicum_helpers.h>
61 #include <ctype.h>
62 #include <err.h>
63 #include <errno.h>
64 #include <fcntl.h>
65 #include <inttypes.h>
66 #include <locale.h>
67 #include <stdio.h>
68 #include <stdlib.h>
69 #include <string.h>
70 #include <time.h>
71 #include <unistd.h>
72 
73 #include "dd.h"
74 #include "extern.h"
75 
76 static void dd_close(void);
77 static void dd_in(void);
78 static void getfdtype(IO *);
79 static void setup(void);
80 
81 IO	in, out;		/* input/output state */
82 STAT	st;			/* statistics */
83 void	(*cfunc)(void);		/* conversion function */
84 uintmax_t cpy_cnt;		/* # of blocks to copy */
85 static off_t	pending = 0;	/* pending seek if sparse */
86 u_int	ddflags = 0;		/* conversion options */
87 size_t	cbsz;			/* conversion block size */
88 uintmax_t files_cnt = 1;	/* # of files to copy */
89 const	u_char *ctab;		/* conversion table */
90 char	fill_char;		/* Character to fill with if defined */
91 size_t	speed = 0;		/* maximum speed, in bytes per second */
92 volatile sig_atomic_t need_summary;
93 volatile sig_atomic_t need_progress;
94 
95 int
96 main(int argc __unused, char *argv[])
97 {
98 	(void)setlocale(LC_CTYPE, "");
99 	jcl(argv);
100 	setup();
101 
102 	caph_cache_catpages();
103 	if (caph_enter() < 0)
104 		err(1, "unable to enter capability mode");
105 
106 	(void)signal(SIGINFO, siginfo_handler);
107 	(void)signal(SIGALRM, sigalrm_handler);
108 	(void)signal(SIGINT, terminate);
109 
110 	atexit(summary);
111 
112 	while (files_cnt--)
113 		dd_in();
114 
115 	dd_close();
116 	/*
117 	 * Some devices such as cfi(4) may perform significant amounts
118 	 * of work when a write descriptor is closed.  Close the out
119 	 * descriptor explicitly so that the summary handler (called
120 	 * from an atexit() hook) includes this work.
121 	 */
122 	close(out.fd);
123 	exit(0);
124 }
125 
126 static int
127 parity(u_char c)
128 {
129 	int i;
130 
131 	i = c ^ (c >> 1) ^ (c >> 2) ^ (c >> 3) ^
132 	    (c >> 4) ^ (c >> 5) ^ (c >> 6) ^ (c >> 7);
133 	return (i & 1);
134 }
135 
136 static void
137 setup(void)
138 {
139 	u_int cnt;
140 	cap_rights_t rights;
141 	unsigned long cmds[] = { FIODTYPE, MTIOCTOP };
142 
143 	if (in.name == NULL) {
144 		in.name = "stdin";
145 		in.fd = STDIN_FILENO;
146 	} else {
147 		in.fd = open(in.name, O_RDONLY, 0);
148 		if (in.fd == -1)
149 			err(1, "%s", in.name);
150 	}
151 
152 	getfdtype(&in);
153 
154 	cap_rights_init(&rights, CAP_READ, CAP_SEEK);
155 	if (cap_rights_limit(in.fd, &rights) == -1 && errno != ENOSYS)
156 		err(1, "unable to limit capability rights");
157 
158 	if (files_cnt > 1 && !(in.flags & ISTAPE))
159 		errx(1, "files is not supported for non-tape devices");
160 
161 	cap_rights_set(&rights, CAP_FTRUNCATE, CAP_IOCTL, CAP_WRITE);
162 	if (out.name == NULL) {
163 		/* No way to check for read access here. */
164 		out.fd = STDOUT_FILENO;
165 		out.name = "stdout";
166 	} else {
167 #define	OFLAGS \
168     (O_CREAT | (ddflags & (C_SEEK | C_NOTRUNC) ? 0 : O_TRUNC))
169 		out.fd = open(out.name, O_RDWR | OFLAGS, DEFFILEMODE);
170 		/*
171 		 * May not have read access, so try again with write only.
172 		 * Without read we may have a problem if output also does
173 		 * not support seeks.
174 		 */
175 		if (out.fd == -1) {
176 			out.fd = open(out.name, O_WRONLY | OFLAGS, DEFFILEMODE);
177 			out.flags |= NOREAD;
178 			cap_rights_clear(&rights, CAP_READ);
179 		}
180 		if (out.fd == -1)
181 			err(1, "%s", out.name);
182 	}
183 
184 	getfdtype(&out);
185 
186 	if (cap_rights_limit(out.fd, &rights) == -1 && errno != ENOSYS)
187 		err(1, "unable to limit capability rights");
188 	if (cap_ioctls_limit(out.fd, cmds, nitems(cmds)) == -1 &&
189 	    errno != ENOSYS)
190 		err(1, "unable to limit capability rights");
191 
192 	if (in.fd != STDIN_FILENO && out.fd != STDIN_FILENO) {
193 		if (caph_limit_stdin() == -1)
194 			err(1, "unable to limit capability rights");
195 	}
196 
197 	if (in.fd != STDOUT_FILENO && out.fd != STDOUT_FILENO) {
198 		if (caph_limit_stdout() == -1)
199 			err(1, "unable to limit capability rights");
200 	}
201 
202 	if (in.fd != STDERR_FILENO && out.fd != STDERR_FILENO) {
203 		if (caph_limit_stderr() == -1)
204 			err(1, "unable to limit capability rights");
205 	}
206 
207 	/*
208 	 * Allocate space for the input and output buffers.  If not doing
209 	 * record oriented I/O, only need a single buffer.
210 	 */
211 	if (!(ddflags & (C_BLOCK | C_UNBLOCK))) {
212 		if ((in.db = malloc((size_t)out.dbsz + in.dbsz - 1)) == NULL)
213 			err(1, "input buffer");
214 		out.db = in.db;
215 	} else if ((in.db = malloc(MAX((size_t)in.dbsz, cbsz) + cbsz)) == NULL ||
216 	    (out.db = malloc(out.dbsz + cbsz)) == NULL)
217 		err(1, "output buffer");
218 
219 	/* dbp is the first free position in each buffer. */
220 	in.dbp = in.db;
221 	out.dbp = out.db;
222 
223 	/* Position the input/output streams. */
224 	if (in.offset)
225 		pos_in();
226 	if (out.offset)
227 		pos_out();
228 
229 	/*
230 	 * Truncate the output file.  If it fails on a type of output file
231 	 * that it should _not_ fail on, error out.
232 	 */
233 	if ((ddflags & (C_OF | C_SEEK | C_NOTRUNC)) == (C_OF | C_SEEK) &&
234 	    out.flags & ISTRUNC)
235 		if (ftruncate(out.fd, out.offset * out.dbsz) == -1)
236 			err(1, "truncating %s", out.name);
237 
238 	if (ddflags & (C_LCASE  | C_UCASE | C_ASCII | C_EBCDIC | C_PARITY)) {
239 		if (ctab != NULL) {
240 			for (cnt = 0; cnt <= 0377; ++cnt)
241 				casetab[cnt] = ctab[cnt];
242 		} else {
243 			for (cnt = 0; cnt <= 0377; ++cnt)
244 				casetab[cnt] = cnt;
245 		}
246 		if ((ddflags & C_PARITY) && !(ddflags & C_ASCII)) {
247 			/*
248 			 * If the input is not EBCDIC, and we do parity
249 			 * processing, strip input parity.
250 			 */
251 			for (cnt = 200; cnt <= 0377; ++cnt)
252 				casetab[cnt] = casetab[cnt & 0x7f];
253 		}
254 		if (ddflags & C_LCASE) {
255 			for (cnt = 0; cnt <= 0377; ++cnt)
256 				casetab[cnt] = tolower(casetab[cnt]);
257 		} else if (ddflags & C_UCASE) {
258 			for (cnt = 0; cnt <= 0377; ++cnt)
259 				casetab[cnt] = toupper(casetab[cnt]);
260 		}
261 		if ((ddflags & C_PARITY)) {
262 			/*
263 			 * This should strictly speaking be a no-op, but I
264 			 * wonder what funny LANG settings could get us.
265 			 */
266 			for (cnt = 0; cnt <= 0377; ++cnt)
267 				casetab[cnt] = casetab[cnt] & 0x7f;
268 		}
269 		if ((ddflags & C_PARSET)) {
270 			for (cnt = 0; cnt <= 0377; ++cnt)
271 				casetab[cnt] = casetab[cnt] | 0x80;
272 		}
273 		if ((ddflags & C_PAREVEN)) {
274 			for (cnt = 0; cnt <= 0377; ++cnt)
275 				if (parity(casetab[cnt]))
276 					casetab[cnt] = casetab[cnt] | 0x80;
277 		}
278 		if ((ddflags & C_PARODD)) {
279 			for (cnt = 0; cnt <= 0377; ++cnt)
280 				if (!parity(casetab[cnt]))
281 					casetab[cnt] = casetab[cnt] | 0x80;
282 		}
283 
284 		ctab = casetab;
285 	}
286 
287 	if ((ddflags & C_PROGRESS)) {
288 		struct itimerval timer = {
289 			.it_interval = { .tv_sec = 1, .tv_usec = 0 },
290 			.it_value = { .tv_sec = 1, .tv_usec = 0 },
291 		};
292 		setitimer(ITIMER_REAL, &timer, NULL);
293 	}
294 
295 	if (clock_gettime(CLOCK_MONOTONIC, &st.start))
296 		err(1, "clock_gettime");
297 }
298 
299 static void
300 getfdtype(IO *io)
301 {
302 	struct stat sb;
303 	int type;
304 
305 	if (fstat(io->fd, &sb) == -1)
306 		err(1, "%s", io->name);
307 	if (S_ISREG(sb.st_mode))
308 		io->flags |= ISTRUNC;
309 	if (S_ISCHR(sb.st_mode) || S_ISBLK(sb.st_mode)) {
310 		if (ioctl(io->fd, FIODTYPE, &type) == -1) {
311 			err(1, "%s", io->name);
312 		} else {
313 			if (type & D_TAPE)
314 				io->flags |= ISTAPE;
315 			else if (type & (D_DISK | D_MEM))
316 				io->flags |= ISSEEK;
317 			if (S_ISCHR(sb.st_mode) && (type & D_TAPE) == 0)
318 				io->flags |= ISCHR;
319 		}
320 		return;
321 	}
322 	errno = 0;
323 	if (lseek(io->fd, (off_t)0, SEEK_CUR) == -1 && errno == ESPIPE)
324 		io->flags |= ISPIPE;
325 	else
326 		io->flags |= ISSEEK;
327 }
328 
329 /*
330  * Limit the speed by adding a delay before every block read.
331  * The delay (t_usleep) is equal to the time computed from block
332  * size and the specified speed limit (t_target) minus the time
333  * spent on actual read and write operations (t_io).
334  */
335 static void
336 speed_limit(void)
337 {
338 	static double t_prev, t_usleep;
339 	double t_now, t_io, t_target;
340 
341 	t_now = secs_elapsed();
342 	t_io = t_now - t_prev - t_usleep;
343 	t_target = (double)in.dbsz / (double)speed;
344 	t_usleep = t_target - t_io;
345 	if (t_usleep > 0)
346 		usleep(t_usleep * 1000000);
347 	else
348 		t_usleep = 0;
349 	t_prev = t_now;
350 }
351 
352 static void
353 swapbytes(void *v, size_t len)
354 {
355 	unsigned char *p = v;
356 	unsigned char t;
357 
358 	while (len > 1) {
359 		t = p[0];
360 		p[0] = p[1];
361 		p[1] = t;
362 		p += 2;
363 		len -= 2;
364 	}
365 }
366 
367 static void
368 dd_in(void)
369 {
370 	ssize_t n;
371 
372 	for (;;) {
373 		switch (cpy_cnt) {
374 		case -1:			/* count=0 was specified */
375 			return;
376 		case 0:
377 			break;
378 		default:
379 			if (st.in_full + st.in_part >= (uintmax_t)cpy_cnt)
380 				return;
381 			break;
382 		}
383 
384 		if (speed > 0)
385 			speed_limit();
386 
387 		/*
388 		 * Zero the buffer first if sync; if doing block operations,
389 		 * use spaces.
390 		 */
391 		if (ddflags & C_SYNC) {
392 			if (ddflags & C_FILL)
393 				memset(in.dbp, fill_char, in.dbsz);
394 			else if (ddflags & (C_BLOCK | C_UNBLOCK))
395 				memset(in.dbp, ' ', in.dbsz);
396 			else
397 				memset(in.dbp, 0, in.dbsz);
398 		}
399 
400 		n = read(in.fd, in.dbp, in.dbsz);
401 		if (n == 0) {
402 			in.dbrcnt = 0;
403 			return;
404 		}
405 
406 		/* Read error. */
407 		if (n == -1) {
408 			/*
409 			 * If noerror not specified, die.  POSIX requires that
410 			 * the warning message be followed by an I/O display.
411 			 */
412 			if (!(ddflags & C_NOERROR))
413 				err(1, "%s", in.name);
414 			warn("%s", in.name);
415 			summary();
416 
417 			/*
418 			 * If it's a seekable file descriptor, seek past the
419 			 * error.  If your OS doesn't do the right thing for
420 			 * raw disks this section should be modified to re-read
421 			 * in sector size chunks.
422 			 */
423 			if (in.flags & ISSEEK &&
424 			    lseek(in.fd, (off_t)in.dbsz, SEEK_CUR))
425 				warn("%s", in.name);
426 
427 			/* If sync not specified, omit block and continue. */
428 			if (!(ddflags & C_SYNC))
429 				continue;
430 
431 			/* Read errors count as full blocks. */
432 			in.dbcnt += in.dbrcnt = in.dbsz;
433 			++st.in_full;
434 
435 		/* Handle full input blocks. */
436 		} else if ((size_t)n == (size_t)in.dbsz) {
437 			in.dbcnt += in.dbrcnt = n;
438 			++st.in_full;
439 
440 		/* Handle partial input blocks. */
441 		} else {
442 			/* If sync, use the entire block. */
443 			if (ddflags & C_SYNC)
444 				in.dbcnt += in.dbrcnt = in.dbsz;
445 			else
446 				in.dbcnt += in.dbrcnt = n;
447 			++st.in_part;
448 		}
449 
450 		/*
451 		 * POSIX states that if bs is set and no other conversions
452 		 * than noerror, notrunc or sync are specified, the block
453 		 * is output without buffering as it is read.
454 		 */
455 		if ((ddflags & ~(C_NOERROR | C_NOTRUNC | C_SYNC)) == C_BS) {
456 			out.dbcnt = in.dbcnt;
457 			dd_out(1);
458 			in.dbcnt = 0;
459 			continue;
460 		}
461 
462 		if (ddflags & C_SWAB) {
463 			if ((n = in.dbrcnt) & 1) {
464 				++st.swab;
465 				--n;
466 			}
467 			swapbytes(in.dbp, (size_t)n);
468 		}
469 
470 		in.dbp += in.dbrcnt;
471 		(*cfunc)();
472 		if (need_summary) {
473 			summary();
474 		}
475 		if (need_progress) {
476 			progress();
477 		}
478 	}
479 }
480 
481 /*
482  * Clean up any remaining I/O and flush output.  If necessary, the output file
483  * is truncated.
484  */
485 static void
486 dd_close(void)
487 {
488 	if (cfunc == def)
489 		def_close();
490 	else if (cfunc == block)
491 		block_close();
492 	else if (cfunc == unblock)
493 		unblock_close();
494 	if (ddflags & C_OSYNC && out.dbcnt && out.dbcnt < out.dbsz) {
495 		if (ddflags & C_FILL)
496 			memset(out.dbp, fill_char, out.dbsz - out.dbcnt);
497 		else if (ddflags & (C_BLOCK | C_UNBLOCK))
498 			memset(out.dbp, ' ', out.dbsz - out.dbcnt);
499 		else
500 			memset(out.dbp, 0, out.dbsz - out.dbcnt);
501 		out.dbcnt = out.dbsz;
502 	}
503 	if (out.dbcnt || pending)
504 		dd_out(1);
505 
506 	/*
507 	 * If the file ends with a hole, ftruncate it to extend its size
508 	 * up to the end of the hole (without having to write any data).
509 	 */
510 	if (out.seek_offset > 0 && (out.flags & ISTRUNC)) {
511 		if (ftruncate(out.fd, out.seek_offset) == -1)
512 			err(1, "truncating %s", out.name);
513 	}
514 }
515 
516 void
517 dd_out(int force)
518 {
519 	u_char *outp;
520 	size_t cnt, i, n;
521 	ssize_t nw;
522 	static int warned;
523 	int sparse;
524 
525 	/*
526 	 * Write one or more blocks out.  The common case is writing a full
527 	 * output block in a single write; increment the full block stats.
528 	 * Otherwise, we're into partial block writes.  If a partial write,
529 	 * and it's a character device, just warn.  If a tape device, quit.
530 	 *
531 	 * The partial writes represent two cases.  1: Where the input block
532 	 * was less than expected so the output block was less than expected.
533 	 * 2: Where the input block was the right size but we were forced to
534 	 * write the block in multiple chunks.  The original versions of dd(1)
535 	 * never wrote a block in more than a single write, so the latter case
536 	 * never happened.
537 	 *
538 	 * One special case is if we're forced to do the write -- in that case
539 	 * we play games with the buffer size, and it's usually a partial write.
540 	 */
541 	outp = out.db;
542 
543 	/*
544 	 * If force, first try to write all pending data, else try to write
545 	 * just one block. Subsequently always write data one full block at
546 	 * a time at most.
547 	 */
548 	for (n = force ? out.dbcnt : out.dbsz;; n = out.dbsz) {
549 		cnt = n;
550 		do {
551 			sparse = 0;
552 			if (ddflags & C_SPARSE) {
553 				sparse = 1;	/* Is buffer sparse? */
554 				for (i = 0; i < cnt; i++)
555 					if (outp[i] != 0) {
556 						sparse = 0;
557 						break;
558 					}
559 			}
560 			if (sparse && !force) {
561 				pending += cnt;
562 				nw = cnt;
563 			} else {
564 				if (pending != 0) {
565 					/*
566 					 * Seek past hole.  Note that we need to record the
567 					 * reached offset, because we might have no more data
568 					 * to write, in which case we'll need to call
569 					 * ftruncate to extend the file size.
570 					 */
571 					out.seek_offset = lseek(out.fd, pending, SEEK_CUR);
572 					if (out.seek_offset == -1)
573 						err(2, "%s: seek error creating sparse file",
574 						    out.name);
575 					pending = 0;
576 				}
577 				if (cnt) {
578 					nw = write(out.fd, outp, cnt);
579 					out.seek_offset = 0;
580 				} else {
581 					return;
582 				}
583 			}
584 
585 			if (nw <= 0) {
586 				if (nw == 0)
587 					errx(1, "%s: end of device", out.name);
588 				if (errno != EINTR)
589 					err(1, "%s", out.name);
590 				nw = 0;
591 			}
592 
593 			outp += nw;
594 			st.bytes += nw;
595 
596 			if ((size_t)nw == n && n == (size_t)out.dbsz)
597 				++st.out_full;
598 			else
599 				++st.out_part;
600 
601 			if ((size_t) nw != cnt) {
602 				if (out.flags & ISTAPE)
603 					errx(1, "%s: short write on tape device",
604 				    	out.name);
605 				if (out.flags & ISCHR && !warned) {
606 					warned = 1;
607 					warnx("%s: short write on character device",
608 				    	out.name);
609 				}
610 			}
611 
612 			cnt -= nw;
613 		} while (cnt != 0);
614 
615 		if ((out.dbcnt -= n) < out.dbsz)
616 			break;
617 	}
618 
619 	/* Reassemble the output block. */
620 	if (out.dbcnt)
621 		(void)memmove(out.db, out.dbp - out.dbcnt, out.dbcnt);
622 	out.dbp = out.db + out.dbcnt;
623 }
624