xref: /freebsd/bin/dd/dd.c (revision 1f474190)
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 uint64_t	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 	if (close(out.fd) == -1 && errno != EINTR)
128 		err(1, "close");
129 	exit(0);
130 }
131 
132 static int
133 parity(u_char c)
134 {
135 	int i;
136 
137 	i = c ^ (c >> 1) ^ (c >> 2) ^ (c >> 3) ^
138 	    (c >> 4) ^ (c >> 5) ^ (c >> 6) ^ (c >> 7);
139 	return (i & 1);
140 }
141 
142 static void
143 setup(void)
144 {
145 	u_int cnt;
146 	int iflags, oflags;
147 	cap_rights_t rights;
148 	unsigned long cmds[] = { FIODTYPE, MTIOCTOP };
149 
150 	if (in.name == NULL) {
151 		in.name = "stdin";
152 		in.fd = STDIN_FILENO;
153 	} else {
154 		iflags = 0;
155 		if (ddflags & C_IDIRECT)
156 			iflags |= O_DIRECT;
157 		in.fd = open(in.name, O_RDONLY | iflags, 0);
158 		if (in.fd == -1)
159 			err(1, "%s", in.name);
160 	}
161 
162 	getfdtype(&in);
163 
164 	cap_rights_init(&rights, CAP_READ, CAP_SEEK);
165 	if (caph_rights_limit(in.fd, &rights) == -1)
166 		err(1, "unable to limit capability rights");
167 
168 	if (files_cnt > 1 && !(in.flags & ISTAPE))
169 		errx(1, "files is not supported for non-tape devices");
170 
171 	cap_rights_set(&rights, CAP_FTRUNCATE, CAP_IOCTL, CAP_WRITE);
172 	if (ddflags & (C_FDATASYNC | C_FSYNC))
173 		cap_rights_set(&rights, CAP_FSYNC);
174 	if (out.name == NULL) {
175 		/* No way to check for read access here. */
176 		out.fd = STDOUT_FILENO;
177 		out.name = "stdout";
178 		if (ddflags & C_OFSYNC) {
179 			oflags = fcntl(out.fd, F_GETFL);
180 			if (oflags == -1)
181 				err(1, "unable to get fd flags for stdout");
182 			oflags |= O_FSYNC;
183 			if (fcntl(out.fd, F_SETFL, oflags) == -1)
184 				err(1, "unable to set fd flags for stdout");
185 		}
186 	} else {
187 		oflags = O_CREAT;
188 		if (!(ddflags & (C_SEEK | C_NOTRUNC)))
189 			oflags |= O_TRUNC;
190 		if (ddflags & C_OFSYNC)
191 			oflags |= O_FSYNC;
192 		if (ddflags & C_ODIRECT)
193 			oflags |= O_DIRECT;
194 		out.fd = open(out.name, O_RDWR | oflags, DEFFILEMODE);
195 		/*
196 		 * May not have read access, so try again with write only.
197 		 * Without read we may have a problem if output also does
198 		 * not support seeks.
199 		 */
200 		if (out.fd == -1) {
201 			out.fd = open(out.name, O_WRONLY | oflags, DEFFILEMODE);
202 			out.flags |= NOREAD;
203 			cap_rights_clear(&rights, CAP_READ);
204 		}
205 		if (out.fd == -1)
206 			err(1, "%s", out.name);
207 	}
208 
209 	getfdtype(&out);
210 
211 	if (caph_rights_limit(out.fd, &rights) == -1)
212 		err(1, "unable to limit capability rights");
213 	if (caph_ioctls_limit(out.fd, cmds, nitems(cmds)) == -1)
214 		err(1, "unable to limit capability rights");
215 
216 	if (in.fd != STDIN_FILENO && out.fd != STDIN_FILENO) {
217 		if (caph_limit_stdin() == -1)
218 			err(1, "unable to limit capability rights");
219 	}
220 
221 	if (in.fd != STDOUT_FILENO && out.fd != STDOUT_FILENO) {
222 		if (caph_limit_stdout() == -1)
223 			err(1, "unable to limit capability rights");
224 	}
225 
226 	if (in.fd != STDERR_FILENO && out.fd != STDERR_FILENO) {
227 		if (caph_limit_stderr() == -1)
228 			err(1, "unable to limit capability rights");
229 	}
230 
231 	/*
232 	 * Allocate space for the input and output buffers.  If not doing
233 	 * record oriented I/O, only need a single buffer.
234 	 */
235 	if (!(ddflags & (C_BLOCK | C_UNBLOCK))) {
236 		if ((in.db = malloc((size_t)out.dbsz + in.dbsz - 1)) == NULL)
237 			err(1, "input buffer");
238 		out.db = in.db;
239 	} else if ((in.db = malloc(MAX((size_t)in.dbsz, cbsz) + cbsz)) == NULL ||
240 	    (out.db = malloc(out.dbsz + cbsz)) == NULL)
241 		err(1, "output buffer");
242 
243 	/* dbp is the first free position in each buffer. */
244 	in.dbp = in.db;
245 	out.dbp = out.db;
246 
247 	/* Position the input/output streams. */
248 	if (in.offset)
249 		pos_in();
250 	if (out.offset)
251 		pos_out();
252 
253 	/*
254 	 * Truncate the output file.  If it fails on a type of output file
255 	 * that it should _not_ fail on, error out.
256 	 */
257 	if ((ddflags & (C_OF | C_SEEK | C_NOTRUNC)) == (C_OF | C_SEEK) &&
258 	    out.flags & ISTRUNC)
259 		if (ftruncate(out.fd, out.offset * out.dbsz) == -1)
260 			err(1, "truncating %s", out.name);
261 
262 	if (ddflags & (C_LCASE  | C_UCASE | C_ASCII | C_EBCDIC | C_PARITY)) {
263 		if (ctab != NULL) {
264 			for (cnt = 0; cnt <= 0377; ++cnt)
265 				casetab[cnt] = ctab[cnt];
266 		} else {
267 			for (cnt = 0; cnt <= 0377; ++cnt)
268 				casetab[cnt] = cnt;
269 		}
270 		if ((ddflags & C_PARITY) && !(ddflags & C_ASCII)) {
271 			/*
272 			 * If the input is not EBCDIC, and we do parity
273 			 * processing, strip input parity.
274 			 */
275 			for (cnt = 200; cnt <= 0377; ++cnt)
276 				casetab[cnt] = casetab[cnt & 0x7f];
277 		}
278 		if (ddflags & C_LCASE) {
279 			for (cnt = 0; cnt <= 0377; ++cnt)
280 				casetab[cnt] = tolower(casetab[cnt]);
281 		} else if (ddflags & C_UCASE) {
282 			for (cnt = 0; cnt <= 0377; ++cnt)
283 				casetab[cnt] = toupper(casetab[cnt]);
284 		}
285 		if ((ddflags & C_PARITY)) {
286 			/*
287 			 * This should strictly speaking be a no-op, but I
288 			 * wonder what funny LANG settings could get us.
289 			 */
290 			for (cnt = 0; cnt <= 0377; ++cnt)
291 				casetab[cnt] = casetab[cnt] & 0x7f;
292 		}
293 		if ((ddflags & C_PARSET)) {
294 			for (cnt = 0; cnt <= 0377; ++cnt)
295 				casetab[cnt] = casetab[cnt] | 0x80;
296 		}
297 		if ((ddflags & C_PAREVEN)) {
298 			for (cnt = 0; cnt <= 0377; ++cnt)
299 				if (parity(casetab[cnt]))
300 					casetab[cnt] = casetab[cnt] | 0x80;
301 		}
302 		if ((ddflags & C_PARODD)) {
303 			for (cnt = 0; cnt <= 0377; ++cnt)
304 				if (!parity(casetab[cnt]))
305 					casetab[cnt] = casetab[cnt] | 0x80;
306 		}
307 
308 		ctab = casetab;
309 	}
310 
311 	if (clock_gettime(CLOCK_MONOTONIC, &st.start))
312 		err(1, "clock_gettime");
313 }
314 
315 static void
316 getfdtype(IO *io)
317 {
318 	struct stat sb;
319 	int type;
320 
321 	if (fstat(io->fd, &sb) == -1)
322 		err(1, "%s", io->name);
323 	if (S_ISREG(sb.st_mode))
324 		io->flags |= ISTRUNC;
325 	if (S_ISCHR(sb.st_mode) || S_ISBLK(sb.st_mode)) {
326 		if (ioctl(io->fd, FIODTYPE, &type) == -1) {
327 			err(1, "%s", io->name);
328 		} else {
329 			if (type & D_TAPE)
330 				io->flags |= ISTAPE;
331 			else if (type & (D_DISK | D_MEM))
332 				io->flags |= ISSEEK;
333 			if (S_ISCHR(sb.st_mode) && (type & D_TAPE) == 0)
334 				io->flags |= ISCHR;
335 		}
336 		return;
337 	}
338 	errno = 0;
339 	if (lseek(io->fd, (off_t)0, SEEK_CUR) == -1 && errno == ESPIPE)
340 		io->flags |= ISPIPE;
341 	else
342 		io->flags |= ISSEEK;
343 }
344 
345 /*
346  * Limit the speed by adding a delay before every block read.
347  * The delay (t_usleep) is equal to the time computed from block
348  * size and the specified speed limit (t_target) minus the time
349  * spent on actual read and write operations (t_io).
350  */
351 static void
352 speed_limit(void)
353 {
354 	static double t_prev, t_usleep;
355 	double t_now, t_io, t_target;
356 
357 	t_now = secs_elapsed();
358 	t_io = t_now - t_prev - t_usleep;
359 	t_target = (double)in.dbsz / (double)speed;
360 	t_usleep = t_target - t_io;
361 	if (t_usleep > 0)
362 		usleep(t_usleep * 1000000);
363 	else
364 		t_usleep = 0;
365 	t_prev = t_now;
366 }
367 
368 static void
369 swapbytes(void *v, size_t len)
370 {
371 	unsigned char *p = v;
372 	unsigned char t;
373 
374 	while (len > 1) {
375 		t = p[0];
376 		p[0] = p[1];
377 		p[1] = t;
378 		p += 2;
379 		len -= 2;
380 	}
381 }
382 
383 static void
384 dd_in(void)
385 {
386 	ssize_t n;
387 
388 	for (;;) {
389 		switch (cpy_cnt) {
390 		case -1:			/* count=0 was specified */
391 			return;
392 		case 0:
393 			break;
394 		default:
395 			if (st.in_full + st.in_part >= (uintmax_t)cpy_cnt)
396 				return;
397 			break;
398 		}
399 
400 		if (speed > 0)
401 			speed_limit();
402 
403 		/*
404 		 * Zero the buffer first if sync; if doing block operations,
405 		 * use spaces.
406 		 */
407 		if (ddflags & C_SYNC) {
408 			if (ddflags & C_FILL)
409 				memset(in.dbp, fill_char, in.dbsz);
410 			else if (ddflags & (C_BLOCK | C_UNBLOCK))
411 				memset(in.dbp, ' ', in.dbsz);
412 			else
413 				memset(in.dbp, 0, in.dbsz);
414 		}
415 
416 		in.dbrcnt = 0;
417 fill:
418 		n = read(in.fd, in.dbp + in.dbrcnt, in.dbsz - in.dbrcnt);
419 
420 		/* EOF */
421 		if (n == 0 && in.dbrcnt == 0)
422 			return;
423 
424 		/* Read error */
425 		if (n == -1) {
426 			/*
427 			 * If noerror not specified, die.  POSIX requires that
428 			 * the warning message be followed by an I/O display.
429 			 */
430 			if (!(ddflags & C_NOERROR))
431 				err(1, "%s", in.name);
432 			warn("%s", in.name);
433 			summary();
434 
435 			/*
436 			 * If it's a seekable file descriptor, seek past the
437 			 * error.  If your OS doesn't do the right thing for
438 			 * raw disks this section should be modified to re-read
439 			 * in sector size chunks.
440 			 */
441 			if (in.flags & ISSEEK &&
442 			    lseek(in.fd, (off_t)in.dbsz, SEEK_CUR))
443 				warn("%s", in.name);
444 
445 			/* If sync not specified, omit block and continue. */
446 			if (!(ddflags & C_SYNC))
447 				continue;
448 		}
449 
450 		/* If conv=sync, use the entire block. */
451 		if (ddflags & C_SYNC)
452 			n = in.dbsz;
453 
454 		/* Count the bytes read for this block. */
455 		in.dbrcnt += n;
456 
457 		/* Count the number of full and partial blocks. */
458 		if (in.dbrcnt == in.dbsz)
459 			++st.in_full;
460 		else if (ddflags & C_IFULLBLOCK && n != 0)
461 			goto fill; /* these don't count */
462 		else
463 			++st.in_part;
464 
465 		/* Count the total bytes read for this file. */
466 		in.dbcnt += in.dbrcnt;
467 
468 		/*
469 		 * POSIX states that if bs is set and no other conversions
470 		 * than noerror, notrunc or sync are specified, the block
471 		 * is output without buffering as it is read.
472 		 */
473 		if ((ddflags & ~(C_NOERROR | C_NOTRUNC | C_SYNC)) == C_BS) {
474 			out.dbcnt = in.dbcnt;
475 			dd_out(1);
476 			in.dbcnt = 0;
477 			continue;
478 		}
479 
480 		if (ddflags & C_SWAB) {
481 			if ((n = in.dbrcnt) & 1) {
482 				++st.swab;
483 				--n;
484 			}
485 			swapbytes(in.dbp, (size_t)n);
486 		}
487 
488 		/* Advance to the next block. */
489 		in.dbp += in.dbrcnt;
490 		(*cfunc)();
491 		if (need_summary)
492 			summary();
493 		if (need_progress)
494 			progress();
495 	}
496 }
497 
498 /*
499  * Clean up any remaining I/O and flush output.  If necessary, the output file
500  * is truncated.
501  */
502 static void
503 dd_close(void)
504 {
505 	if (cfunc == def)
506 		def_close();
507 	else if (cfunc == block)
508 		block_close();
509 	else if (cfunc == unblock)
510 		unblock_close();
511 	if (ddflags & C_OSYNC && out.dbcnt && out.dbcnt < out.dbsz) {
512 		if (ddflags & C_FILL)
513 			memset(out.dbp, fill_char, out.dbsz - out.dbcnt);
514 		else if (ddflags & (C_BLOCK | C_UNBLOCK))
515 			memset(out.dbp, ' ', out.dbsz - out.dbcnt);
516 		else
517 			memset(out.dbp, 0, out.dbsz - out.dbcnt);
518 		out.dbcnt = out.dbsz;
519 	}
520 	if (out.dbcnt || pending)
521 		dd_out(1);
522 
523 	/*
524 	 * If the file ends with a hole, ftruncate it to extend its size
525 	 * up to the end of the hole (without having to write any data).
526 	 */
527 	if (out.seek_offset > 0 && (out.flags & ISTRUNC)) {
528 		if (ftruncate(out.fd, out.seek_offset) == -1)
529 			err(1, "truncating %s", out.name);
530 	}
531 
532 	if (ddflags & C_FSYNC) {
533 		if (fsync(out.fd) == -1)
534 			err(1, "fsyncing %s", out.name);
535 	} else if (ddflags & C_FDATASYNC) {
536 		if (fdatasync(out.fd) == -1)
537 			err(1, "fdatasyncing %s", out.name);
538 	}
539 }
540 
541 void
542 dd_out(int force)
543 {
544 	u_char *outp;
545 	size_t cnt, n;
546 	ssize_t nw;
547 	static int warned;
548 	int sparse;
549 
550 	/*
551 	 * Write one or more blocks out.  The common case is writing a full
552 	 * output block in a single write; increment the full block stats.
553 	 * Otherwise, we're into partial block writes.  If a partial write,
554 	 * and it's a character device, just warn.  If a tape device, quit.
555 	 *
556 	 * The partial writes represent two cases.  1: Where the input block
557 	 * was less than expected so the output block was less than expected.
558 	 * 2: Where the input block was the right size but we were forced to
559 	 * write the block in multiple chunks.  The original versions of dd(1)
560 	 * never wrote a block in more than a single write, so the latter case
561 	 * never happened.
562 	 *
563 	 * One special case is if we're forced to do the write -- in that case
564 	 * we play games with the buffer size, and it's usually a partial write.
565 	 */
566 	outp = out.db;
567 
568 	/*
569 	 * If force, first try to write all pending data, else try to write
570 	 * just one block. Subsequently always write data one full block at
571 	 * a time at most.
572 	 */
573 	for (n = force ? out.dbcnt : out.dbsz;; n = out.dbsz) {
574 		cnt = n;
575 		do {
576 			sparse = 0;
577 			if (ddflags & C_SPARSE) {
578 				/* Is buffer sparse? */
579 				sparse = BISZERO(outp, cnt);
580 			}
581 			if (sparse && !force) {
582 				pending += cnt;
583 				nw = cnt;
584 			} else {
585 				if (pending != 0) {
586 					/*
587 					 * Seek past hole.  Note that we need to record the
588 					 * reached offset, because we might have no more data
589 					 * to write, in which case we'll need to call
590 					 * ftruncate to extend the file size.
591 					 */
592 					out.seek_offset = lseek(out.fd, pending, SEEK_CUR);
593 					if (out.seek_offset == -1)
594 						err(2, "%s: seek error creating sparse file",
595 						    out.name);
596 					pending = 0;
597 				}
598 				if (cnt) {
599 					nw = write(out.fd, outp, cnt);
600 					out.seek_offset = 0;
601 				} else {
602 					return;
603 				}
604 			}
605 
606 			if (nw <= 0) {
607 				if (nw == 0)
608 					errx(1, "%s: end of device", out.name);
609 				if (errno != EINTR)
610 					err(1, "%s", out.name);
611 				nw = 0;
612 			}
613 
614 			outp += nw;
615 			st.bytes += nw;
616 
617 			if ((size_t)nw == n && n == (size_t)out.dbsz)
618 				++st.out_full;
619 			else
620 				++st.out_part;
621 
622 			if ((size_t) nw != cnt) {
623 				if (out.flags & ISTAPE)
624 					errx(1, "%s: short write on tape device",
625 				    	out.name);
626 				if (out.flags & ISCHR && !warned) {
627 					warned = 1;
628 					warnx("%s: short write on character device",
629 				    	out.name);
630 				}
631 			}
632 
633 			cnt -= nw;
634 		} while (cnt != 0);
635 
636 		if ((out.dbcnt -= n) < out.dbsz)
637 			break;
638 	}
639 
640 	/* Reassemble the output block. */
641 	if (out.dbcnt)
642 		(void)memmove(out.db, out.dbp - out.dbcnt, out.dbcnt);
643 	out.dbp = out.db + out.dbcnt;
644 }
645