xref: /freebsd/bin/dd/dd.c (revision 59e50df3cd1493537cfa916daf3c51a01b7ff06e)
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 	struct itimerval itv = { { 1, 0 }, { 1, 0 } }; /* SIGALARM every second, if needed */
99 
100 	(void)setlocale(LC_CTYPE, "");
101 	jcl(argv);
102 	setup();
103 
104 	caph_cache_catpages();
105 	if (caph_enter() < 0)
106 		err(1, "unable to enter capability mode");
107 
108 	(void)signal(SIGINFO, siginfo_handler);
109 	if (ddflags & C_PROGRESS) {
110 		(void)signal(SIGALRM, sigalarm_handler);
111 		setitimer(ITIMER_REAL, &itv, NULL);
112 	}
113 	(void)signal(SIGINT, terminate);
114 
115 	atexit(summary);
116 
117 	while (files_cnt--)
118 		dd_in();
119 
120 	dd_close();
121 	/*
122 	 * Some devices such as cfi(4) may perform significant amounts
123 	 * of work when a write descriptor is closed.  Close the out
124 	 * descriptor explicitly so that the summary handler (called
125 	 * from an atexit() hook) includes this work.
126 	 */
127 	close(out.fd);
128 	exit(0);
129 }
130 
131 static int
132 parity(u_char c)
133 {
134 	int i;
135 
136 	i = c ^ (c >> 1) ^ (c >> 2) ^ (c >> 3) ^
137 	    (c >> 4) ^ (c >> 5) ^ (c >> 6) ^ (c >> 7);
138 	return (i & 1);
139 }
140 
141 static void
142 setup(void)
143 {
144 	u_int cnt;
145 	cap_rights_t rights;
146 	unsigned long cmds[] = { FIODTYPE, MTIOCTOP };
147 
148 	if (in.name == NULL) {
149 		in.name = "stdin";
150 		in.fd = STDIN_FILENO;
151 	} else {
152 		in.fd = open(in.name, O_RDONLY, 0);
153 		if (in.fd == -1)
154 			err(1, "%s", in.name);
155 	}
156 
157 	getfdtype(&in);
158 
159 	cap_rights_init(&rights, CAP_READ, CAP_SEEK);
160 	if (caph_rights_limit(in.fd, &rights) == -1)
161 		err(1, "unable to limit capability rights");
162 
163 	if (files_cnt > 1 && !(in.flags & ISTAPE))
164 		errx(1, "files is not supported for non-tape devices");
165 
166 	cap_rights_set(&rights, CAP_FTRUNCATE, CAP_IOCTL, CAP_WRITE);
167 	if (ddflags & C_FSYNC)
168 		cap_rights_set(&rights, CAP_FSYNC);
169 	if (out.name == NULL) {
170 		/* No way to check for read access here. */
171 		out.fd = STDOUT_FILENO;
172 		out.name = "stdout";
173 	} else {
174 #define	OFLAGS \
175     (O_CREAT | (ddflags & (C_SEEK | C_NOTRUNC) ? 0 : O_TRUNC))
176 		out.fd = open(out.name, O_RDWR | OFLAGS, DEFFILEMODE);
177 		/*
178 		 * May not have read access, so try again with write only.
179 		 * Without read we may have a problem if output also does
180 		 * not support seeks.
181 		 */
182 		if (out.fd == -1) {
183 			out.fd = open(out.name, O_WRONLY | OFLAGS, DEFFILEMODE);
184 			out.flags |= NOREAD;
185 			cap_rights_clear(&rights, CAP_READ);
186 		}
187 		if (out.fd == -1)
188 			err(1, "%s", out.name);
189 	}
190 
191 	getfdtype(&out);
192 
193 	if (caph_rights_limit(out.fd, &rights) == -1)
194 		err(1, "unable to limit capability rights");
195 	if (caph_ioctls_limit(out.fd, cmds, nitems(cmds)) == -1)
196 		err(1, "unable to limit capability rights");
197 
198 	if (in.fd != STDIN_FILENO && out.fd != STDIN_FILENO) {
199 		if (caph_limit_stdin() == -1)
200 			err(1, "unable to limit capability rights");
201 	}
202 
203 	if (in.fd != STDOUT_FILENO && out.fd != STDOUT_FILENO) {
204 		if (caph_limit_stdout() == -1)
205 			err(1, "unable to limit capability rights");
206 	}
207 
208 	if (in.fd != STDERR_FILENO && out.fd != STDERR_FILENO) {
209 		if (caph_limit_stderr() == -1)
210 			err(1, "unable to limit capability rights");
211 	}
212 
213 	/*
214 	 * Allocate space for the input and output buffers.  If not doing
215 	 * record oriented I/O, only need a single buffer.
216 	 */
217 	if (!(ddflags & (C_BLOCK | C_UNBLOCK))) {
218 		if ((in.db = malloc((size_t)out.dbsz + in.dbsz - 1)) == NULL)
219 			err(1, "input buffer");
220 		out.db = in.db;
221 	} else if ((in.db = malloc(MAX((size_t)in.dbsz, cbsz) + cbsz)) == NULL ||
222 	    (out.db = malloc(out.dbsz + cbsz)) == NULL)
223 		err(1, "output buffer");
224 
225 	/* dbp is the first free position in each buffer. */
226 	in.dbp = in.db;
227 	out.dbp = out.db;
228 
229 	/* Position the input/output streams. */
230 	if (in.offset)
231 		pos_in();
232 	if (out.offset)
233 		pos_out();
234 
235 	/*
236 	 * Truncate the output file.  If it fails on a type of output file
237 	 * that it should _not_ fail on, error out.
238 	 */
239 	if ((ddflags & (C_OF | C_SEEK | C_NOTRUNC)) == (C_OF | C_SEEK) &&
240 	    out.flags & ISTRUNC)
241 		if (ftruncate(out.fd, out.offset * out.dbsz) == -1)
242 			err(1, "truncating %s", out.name);
243 
244 	if (ddflags & (C_LCASE  | C_UCASE | C_ASCII | C_EBCDIC | C_PARITY)) {
245 		if (ctab != NULL) {
246 			for (cnt = 0; cnt <= 0377; ++cnt)
247 				casetab[cnt] = ctab[cnt];
248 		} else {
249 			for (cnt = 0; cnt <= 0377; ++cnt)
250 				casetab[cnt] = cnt;
251 		}
252 		if ((ddflags & C_PARITY) && !(ddflags & C_ASCII)) {
253 			/*
254 			 * If the input is not EBCDIC, and we do parity
255 			 * processing, strip input parity.
256 			 */
257 			for (cnt = 200; cnt <= 0377; ++cnt)
258 				casetab[cnt] = casetab[cnt & 0x7f];
259 		}
260 		if (ddflags & C_LCASE) {
261 			for (cnt = 0; cnt <= 0377; ++cnt)
262 				casetab[cnt] = tolower(casetab[cnt]);
263 		} else if (ddflags & C_UCASE) {
264 			for (cnt = 0; cnt <= 0377; ++cnt)
265 				casetab[cnt] = toupper(casetab[cnt]);
266 		}
267 		if ((ddflags & C_PARITY)) {
268 			/*
269 			 * This should strictly speaking be a no-op, but I
270 			 * wonder what funny LANG settings could get us.
271 			 */
272 			for (cnt = 0; cnt <= 0377; ++cnt)
273 				casetab[cnt] = casetab[cnt] & 0x7f;
274 		}
275 		if ((ddflags & C_PARSET)) {
276 			for (cnt = 0; cnt <= 0377; ++cnt)
277 				casetab[cnt] = casetab[cnt] | 0x80;
278 		}
279 		if ((ddflags & C_PAREVEN)) {
280 			for (cnt = 0; cnt <= 0377; ++cnt)
281 				if (parity(casetab[cnt]))
282 					casetab[cnt] = casetab[cnt] | 0x80;
283 		}
284 		if ((ddflags & C_PARODD)) {
285 			for (cnt = 0; cnt <= 0377; ++cnt)
286 				if (!parity(casetab[cnt]))
287 					casetab[cnt] = casetab[cnt] | 0x80;
288 		}
289 
290 		ctab = casetab;
291 	}
292 
293 	if (clock_gettime(CLOCK_MONOTONIC, &st.start))
294 		err(1, "clock_gettime");
295 }
296 
297 static void
298 getfdtype(IO *io)
299 {
300 	struct stat sb;
301 	int type;
302 
303 	if (fstat(io->fd, &sb) == -1)
304 		err(1, "%s", io->name);
305 	if (S_ISREG(sb.st_mode))
306 		io->flags |= ISTRUNC;
307 	if (S_ISCHR(sb.st_mode) || S_ISBLK(sb.st_mode)) {
308 		if (ioctl(io->fd, FIODTYPE, &type) == -1) {
309 			err(1, "%s", io->name);
310 		} else {
311 			if (type & D_TAPE)
312 				io->flags |= ISTAPE;
313 			else if (type & (D_DISK | D_MEM))
314 				io->flags |= ISSEEK;
315 			if (S_ISCHR(sb.st_mode) && (type & D_TAPE) == 0)
316 				io->flags |= ISCHR;
317 		}
318 		return;
319 	}
320 	errno = 0;
321 	if (lseek(io->fd, (off_t)0, SEEK_CUR) == -1 && errno == ESPIPE)
322 		io->flags |= ISPIPE;
323 	else
324 		io->flags |= ISSEEK;
325 }
326 
327 /*
328  * Limit the speed by adding a delay before every block read.
329  * The delay (t_usleep) is equal to the time computed from block
330  * size and the specified speed limit (t_target) minus the time
331  * spent on actual read and write operations (t_io).
332  */
333 static void
334 speed_limit(void)
335 {
336 	static double t_prev, t_usleep;
337 	double t_now, t_io, t_target;
338 
339 	t_now = secs_elapsed();
340 	t_io = t_now - t_prev - t_usleep;
341 	t_target = (double)in.dbsz / (double)speed;
342 	t_usleep = t_target - t_io;
343 	if (t_usleep > 0)
344 		usleep(t_usleep * 1000000);
345 	else
346 		t_usleep = 0;
347 	t_prev = t_now;
348 }
349 
350 static void
351 swapbytes(void *v, size_t len)
352 {
353 	unsigned char *p = v;
354 	unsigned char t;
355 
356 	while (len > 1) {
357 		t = p[0];
358 		p[0] = p[1];
359 		p[1] = t;
360 		p += 2;
361 		len -= 2;
362 	}
363 }
364 
365 static void
366 dd_in(void)
367 {
368 	ssize_t n;
369 
370 	for (;;) {
371 		switch (cpy_cnt) {
372 		case -1:			/* count=0 was specified */
373 			return;
374 		case 0:
375 			break;
376 		default:
377 			if (st.in_full + st.in_part >= (uintmax_t)cpy_cnt)
378 				return;
379 			break;
380 		}
381 
382 		if (speed > 0)
383 			speed_limit();
384 
385 		/*
386 		 * Zero the buffer first if sync; if doing block operations,
387 		 * use spaces.
388 		 */
389 		if (ddflags & C_SYNC) {
390 			if (ddflags & C_FILL)
391 				memset(in.dbp, fill_char, in.dbsz);
392 			else if (ddflags & (C_BLOCK | C_UNBLOCK))
393 				memset(in.dbp, ' ', in.dbsz);
394 			else
395 				memset(in.dbp, 0, in.dbsz);
396 		}
397 
398 		n = read(in.fd, in.dbp, in.dbsz);
399 		if (n == 0) {
400 			in.dbrcnt = 0;
401 			return;
402 		}
403 
404 		/* Read error. */
405 		if (n == -1) {
406 			/*
407 			 * If noerror not specified, die.  POSIX requires that
408 			 * the warning message be followed by an I/O display.
409 			 */
410 			if (!(ddflags & C_NOERROR))
411 				err(1, "%s", in.name);
412 			warn("%s", in.name);
413 			summary();
414 
415 			/*
416 			 * If it's a seekable file descriptor, seek past the
417 			 * error.  If your OS doesn't do the right thing for
418 			 * raw disks this section should be modified to re-read
419 			 * in sector size chunks.
420 			 */
421 			if (in.flags & ISSEEK &&
422 			    lseek(in.fd, (off_t)in.dbsz, SEEK_CUR))
423 				warn("%s", in.name);
424 
425 			/* If sync not specified, omit block and continue. */
426 			if (!(ddflags & C_SYNC))
427 				continue;
428 
429 			/* Read errors count as full blocks. */
430 			in.dbcnt += in.dbrcnt = in.dbsz;
431 			++st.in_full;
432 
433 		/* Handle full input blocks. */
434 		} else if ((size_t)n == (size_t)in.dbsz) {
435 			in.dbcnt += in.dbrcnt = n;
436 			++st.in_full;
437 
438 		/* Handle partial input blocks. */
439 		} else {
440 			/* If sync, use the entire block. */
441 			if (ddflags & C_SYNC)
442 				in.dbcnt += in.dbrcnt = in.dbsz;
443 			else
444 				in.dbcnt += in.dbrcnt = n;
445 			++st.in_part;
446 		}
447 
448 		/*
449 		 * POSIX states that if bs is set and no other conversions
450 		 * than noerror, notrunc or sync are specified, the block
451 		 * is output without buffering as it is read.
452 		 */
453 		if ((ddflags & ~(C_NOERROR | C_NOTRUNC | C_SYNC)) == C_BS) {
454 			out.dbcnt = in.dbcnt;
455 			dd_out(1);
456 			in.dbcnt = 0;
457 			continue;
458 		}
459 
460 		if (ddflags & C_SWAB) {
461 			if ((n = in.dbrcnt) & 1) {
462 				++st.swab;
463 				--n;
464 			}
465 			swapbytes(in.dbp, (size_t)n);
466 		}
467 
468 		in.dbp += in.dbrcnt;
469 		(*cfunc)();
470 		if (need_summary)
471 			summary();
472 		if (need_progress)
473 			progress();
474 	}
475 }
476 
477 /*
478  * Clean up any remaining I/O and flush output.  If necessary, the output file
479  * is truncated.
480  */
481 static void
482 dd_close(void)
483 {
484 	if (cfunc == def)
485 		def_close();
486 	else if (cfunc == block)
487 		block_close();
488 	else if (cfunc == unblock)
489 		unblock_close();
490 	if (ddflags & C_OSYNC && out.dbcnt && out.dbcnt < out.dbsz) {
491 		if (ddflags & C_FILL)
492 			memset(out.dbp, fill_char, out.dbsz - out.dbcnt);
493 		else if (ddflags & (C_BLOCK | C_UNBLOCK))
494 			memset(out.dbp, ' ', out.dbsz - out.dbcnt);
495 		else
496 			memset(out.dbp, 0, out.dbsz - out.dbcnt);
497 		out.dbcnt = out.dbsz;
498 	}
499 	if (out.dbcnt || pending)
500 		dd_out(1);
501 
502 	/*
503 	 * If the file ends with a hole, ftruncate it to extend its size
504 	 * up to the end of the hole (without having to write any data).
505 	 */
506 	if (out.seek_offset > 0 && (out.flags & ISTRUNC)) {
507 		if (ftruncate(out.fd, out.seek_offset) == -1)
508 			err(1, "truncating %s", out.name);
509 	}
510 
511 	if (ddflags & C_FSYNC) {
512 		if (fsync(out.fd) == -1)
513 			err(1, "fsyncing %s", out.name);
514 	}
515 }
516 
517 void
518 dd_out(int force)
519 {
520 	u_char *outp;
521 	size_t cnt, n;
522 	ssize_t nw;
523 	static int warned;
524 	int sparse;
525 
526 	/*
527 	 * Write one or more blocks out.  The common case is writing a full
528 	 * output block in a single write; increment the full block stats.
529 	 * Otherwise, we're into partial block writes.  If a partial write,
530 	 * and it's a character device, just warn.  If a tape device, quit.
531 	 *
532 	 * The partial writes represent two cases.  1: Where the input block
533 	 * was less than expected so the output block was less than expected.
534 	 * 2: Where the input block was the right size but we were forced to
535 	 * write the block in multiple chunks.  The original versions of dd(1)
536 	 * never wrote a block in more than a single write, so the latter case
537 	 * never happened.
538 	 *
539 	 * One special case is if we're forced to do the write -- in that case
540 	 * we play games with the buffer size, and it's usually a partial write.
541 	 */
542 	outp = out.db;
543 
544 	/*
545 	 * If force, first try to write all pending data, else try to write
546 	 * just one block. Subsequently always write data one full block at
547 	 * a time at most.
548 	 */
549 	for (n = force ? out.dbcnt : out.dbsz;; n = out.dbsz) {
550 		cnt = n;
551 		do {
552 			sparse = 0;
553 			if (ddflags & C_SPARSE) {
554 				/* Is buffer sparse? */
555 				sparse = BISZERO(outp, cnt);
556 			}
557 			if (sparse && !force) {
558 				pending += cnt;
559 				nw = cnt;
560 			} else {
561 				if (pending != 0) {
562 					/*
563 					 * Seek past hole.  Note that we need to record the
564 					 * reached offset, because we might have no more data
565 					 * to write, in which case we'll need to call
566 					 * ftruncate to extend the file size.
567 					 */
568 					out.seek_offset = lseek(out.fd, pending, SEEK_CUR);
569 					if (out.seek_offset == -1)
570 						err(2, "%s: seek error creating sparse file",
571 						    out.name);
572 					pending = 0;
573 				}
574 				if (cnt) {
575 					nw = write(out.fd, outp, cnt);
576 					out.seek_offset = 0;
577 				} else {
578 					return;
579 				}
580 			}
581 
582 			if (nw <= 0) {
583 				if (nw == 0)
584 					errx(1, "%s: end of device", out.name);
585 				if (errno != EINTR)
586 					err(1, "%s", out.name);
587 				nw = 0;
588 			}
589 
590 			outp += nw;
591 			st.bytes += nw;
592 
593 			if ((size_t)nw == n && n == (size_t)out.dbsz)
594 				++st.out_full;
595 			else
596 				++st.out_part;
597 
598 			if ((size_t) nw != cnt) {
599 				if (out.flags & ISTAPE)
600 					errx(1, "%s: short write on tape device",
601 				    	out.name);
602 				if (out.flags & ISCHR && !warned) {
603 					warned = 1;
604 					warnx("%s: short write on character device",
605 				    	out.name);
606 				}
607 			}
608 
609 			cnt -= nw;
610 		} while (cnt != 0);
611 
612 		if ((out.dbcnt -= n) < out.dbsz)
613 			break;
614 	}
615 
616 	/* Reassemble the output block. */
617 	if (out.dbcnt)
618 		(void)memmove(out.db, out.dbp - out.dbcnt, out.dbcnt);
619 	out.dbp = out.db + out.dbcnt;
620 }
621