xref: /freebsd/bin/dd/dd.c (revision 0dbdecfd4f23486da21345fe446a28a1c27db3f2)
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 
58 #include <assert.h>
59 #include <capsicum_helpers.h>
60 #include <ctype.h>
61 #include <err.h>
62 #include <errno.h>
63 #include <fcntl.h>
64 #include <inttypes.h>
65 #include <locale.h>
66 #include <stdio.h>
67 #include <stdlib.h>
68 #include <string.h>
69 #include <time.h>
70 #include <unistd.h>
71 
72 #include "dd.h"
73 #include "extern.h"
74 
75 static void dd_close(void);
76 static void dd_in(void);
77 static void getfdtype(IO *);
78 static void setup(void);
79 
80 IO	in, out;		/* input/output state */
81 STAT	st;			/* statistics */
82 void	(*cfunc)(void);		/* conversion function */
83 uintmax_t cpy_cnt;		/* # of blocks to copy */
84 static off_t	pending = 0;	/* pending seek if sparse */
85 u_int	ddflags = 0;		/* conversion options */
86 size_t	cbsz;			/* conversion block size */
87 uintmax_t files_cnt = 1;	/* # of files to copy */
88 const	u_char *ctab;		/* conversion table */
89 char	fill_char;		/* Character to fill with if defined */
90 size_t	speed = 0;		/* maximum speed, in bytes per second */
91 volatile sig_atomic_t need_summary;
92 
93 int
94 main(int argc __unused, char *argv[])
95 {
96 	(void)setlocale(LC_CTYPE, "");
97 	jcl(argv);
98 	setup();
99 
100 	caph_cache_catpages();
101 	if (caph_enter() < 0)
102 		err(1, "unable to enter capability mode");
103 
104 	(void)signal(SIGINFO, siginfo_handler);
105 	(void)signal(SIGINT, terminate);
106 
107 	atexit(summary);
108 
109 	while (files_cnt--)
110 		dd_in();
111 
112 	dd_close();
113 	/*
114 	 * Some devices such as cfi(4) may perform significant amounts
115 	 * of work when a write descriptor is closed.  Close the out
116 	 * descriptor explicitly so that the summary handler (called
117 	 * from an atexit() hook) includes this work.
118 	 */
119 	close(out.fd);
120 	exit(0);
121 }
122 
123 static int
124 parity(u_char c)
125 {
126 	int i;
127 
128 	i = c ^ (c >> 1) ^ (c >> 2) ^ (c >> 3) ^
129 	    (c >> 4) ^ (c >> 5) ^ (c >> 6) ^ (c >> 7);
130 	return (i & 1);
131 }
132 
133 static void
134 setup(void)
135 {
136 	u_int cnt;
137 	cap_rights_t rights;
138 	unsigned long cmds[] = { FIODTYPE, MTIOCTOP };
139 
140 	if (in.name == NULL) {
141 		in.name = "stdin";
142 		in.fd = STDIN_FILENO;
143 	} else {
144 		in.fd = open(in.name, O_RDONLY, 0);
145 		if (in.fd == -1)
146 			err(1, "%s", in.name);
147 	}
148 
149 	getfdtype(&in);
150 
151 	cap_rights_init(&rights, CAP_READ, CAP_SEEK);
152 	if (cap_rights_limit(in.fd, &rights) == -1 && errno != ENOSYS)
153 		err(1, "unable to limit capability rights");
154 
155 	if (files_cnt > 1 && !(in.flags & ISTAPE))
156 		errx(1, "files is not supported for non-tape devices");
157 
158 	cap_rights_set(&rights, CAP_FTRUNCATE, CAP_IOCTL, CAP_WRITE);
159 	if (out.name == NULL) {
160 		/* No way to check for read access here. */
161 		out.fd = STDOUT_FILENO;
162 		out.name = "stdout";
163 	} else {
164 #define	OFLAGS \
165     (O_CREAT | (ddflags & (C_SEEK | C_NOTRUNC) ? 0 : O_TRUNC))
166 		out.fd = open(out.name, O_RDWR | OFLAGS, DEFFILEMODE);
167 		/*
168 		 * May not have read access, so try again with write only.
169 		 * Without read we may have a problem if output also does
170 		 * not support seeks.
171 		 */
172 		if (out.fd == -1) {
173 			out.fd = open(out.name, O_WRONLY | OFLAGS, DEFFILEMODE);
174 			out.flags |= NOREAD;
175 			cap_rights_clear(&rights, CAP_READ);
176 		}
177 		if (out.fd == -1)
178 			err(1, "%s", out.name);
179 	}
180 
181 	getfdtype(&out);
182 
183 	if (cap_rights_limit(out.fd, &rights) == -1 && errno != ENOSYS)
184 		err(1, "unable to limit capability rights");
185 	if (cap_ioctls_limit(out.fd, cmds, nitems(cmds)) == -1 &&
186 	    errno != ENOSYS)
187 		err(1, "unable to limit capability rights");
188 
189 	if (in.fd != STDIN_FILENO && out.fd != STDIN_FILENO) {
190 		if (caph_limit_stdin() == -1)
191 			err(1, "unable to limit capability rights");
192 	}
193 
194 	if (in.fd != STDOUT_FILENO && out.fd != STDOUT_FILENO) {
195 		if (caph_limit_stdout() == -1)
196 			err(1, "unable to limit capability rights");
197 	}
198 
199 	if (in.fd != STDERR_FILENO && out.fd != STDERR_FILENO) {
200 		if (caph_limit_stderr() == -1)
201 			err(1, "unable to limit capability rights");
202 	}
203 
204 	/*
205 	 * Allocate space for the input and output buffers.  If not doing
206 	 * record oriented I/O, only need a single buffer.
207 	 */
208 	if (!(ddflags & (C_BLOCK | C_UNBLOCK))) {
209 		if ((in.db = malloc((size_t)out.dbsz + in.dbsz - 1)) == NULL)
210 			err(1, "input buffer");
211 		out.db = in.db;
212 	} else if ((in.db = malloc(MAX((size_t)in.dbsz, cbsz) + cbsz)) == NULL ||
213 	    (out.db = malloc(out.dbsz + cbsz)) == NULL)
214 		err(1, "output buffer");
215 
216 	/* dbp is the first free position in each buffer. */
217 	in.dbp = in.db;
218 	out.dbp = out.db;
219 
220 	/* Position the input/output streams. */
221 	if (in.offset)
222 		pos_in();
223 	if (out.offset)
224 		pos_out();
225 
226 	/*
227 	 * Truncate the output file.  If it fails on a type of output file
228 	 * that it should _not_ fail on, error out.
229 	 */
230 	if ((ddflags & (C_OF | C_SEEK | C_NOTRUNC)) == (C_OF | C_SEEK) &&
231 	    out.flags & ISTRUNC)
232 		if (ftruncate(out.fd, out.offset * out.dbsz) == -1)
233 			err(1, "truncating %s", out.name);
234 
235 	if (ddflags & (C_LCASE  | C_UCASE | C_ASCII | C_EBCDIC | C_PARITY)) {
236 		if (ctab != NULL) {
237 			for (cnt = 0; cnt <= 0377; ++cnt)
238 				casetab[cnt] = ctab[cnt];
239 		} else {
240 			for (cnt = 0; cnt <= 0377; ++cnt)
241 				casetab[cnt] = cnt;
242 		}
243 		if ((ddflags & C_PARITY) && !(ddflags & C_ASCII)) {
244 			/*
245 			 * If the input is not EBCDIC, and we do parity
246 			 * processing, strip input parity.
247 			 */
248 			for (cnt = 200; cnt <= 0377; ++cnt)
249 				casetab[cnt] = casetab[cnt & 0x7f];
250 		}
251 		if (ddflags & C_LCASE) {
252 			for (cnt = 0; cnt <= 0377; ++cnt)
253 				casetab[cnt] = tolower(casetab[cnt]);
254 		} else if (ddflags & C_UCASE) {
255 			for (cnt = 0; cnt <= 0377; ++cnt)
256 				casetab[cnt] = toupper(casetab[cnt]);
257 		}
258 		if ((ddflags & C_PARITY)) {
259 			/*
260 			 * This should strictly speaking be a no-op, but I
261 			 * wonder what funny LANG settings could get us.
262 			 */
263 			for (cnt = 0; cnt <= 0377; ++cnt)
264 				casetab[cnt] = casetab[cnt] & 0x7f;
265 		}
266 		if ((ddflags & C_PARSET)) {
267 			for (cnt = 0; cnt <= 0377; ++cnt)
268 				casetab[cnt] = casetab[cnt] | 0x80;
269 		}
270 		if ((ddflags & C_PAREVEN)) {
271 			for (cnt = 0; cnt <= 0377; ++cnt)
272 				if (parity(casetab[cnt]))
273 					casetab[cnt] = casetab[cnt] | 0x80;
274 		}
275 		if ((ddflags & C_PARODD)) {
276 			for (cnt = 0; cnt <= 0377; ++cnt)
277 				if (!parity(casetab[cnt]))
278 					casetab[cnt] = casetab[cnt] | 0x80;
279 		}
280 
281 		ctab = casetab;
282 	}
283 
284 	if (clock_gettime(CLOCK_MONOTONIC, &st.start))
285 		err(1, "clock_gettime");
286 }
287 
288 static void
289 getfdtype(IO *io)
290 {
291 	struct stat sb;
292 	int type;
293 
294 	if (fstat(io->fd, &sb) == -1)
295 		err(1, "%s", io->name);
296 	if (S_ISREG(sb.st_mode))
297 		io->flags |= ISTRUNC;
298 	if (S_ISCHR(sb.st_mode) || S_ISBLK(sb.st_mode)) {
299 		if (ioctl(io->fd, FIODTYPE, &type) == -1) {
300 			err(1, "%s", io->name);
301 		} else {
302 			if (type & D_TAPE)
303 				io->flags |= ISTAPE;
304 			else if (type & (D_DISK | D_MEM))
305 				io->flags |= ISSEEK;
306 			if (S_ISCHR(sb.st_mode) && (type & D_TAPE) == 0)
307 				io->flags |= ISCHR;
308 		}
309 		return;
310 	}
311 	errno = 0;
312 	if (lseek(io->fd, (off_t)0, SEEK_CUR) == -1 && errno == ESPIPE)
313 		io->flags |= ISPIPE;
314 	else
315 		io->flags |= ISSEEK;
316 }
317 
318 /*
319  * Limit the speed by adding a delay before every block read.
320  * The delay (t_usleep) is equal to the time computed from block
321  * size and the specified speed limit (t_target) minus the time
322  * spent on actual read and write operations (t_io).
323  */
324 static void
325 speed_limit(void)
326 {
327 	static double t_prev, t_usleep;
328 	double t_now, t_io, t_target;
329 
330 	t_now = secs_elapsed();
331 	t_io = t_now - t_prev - t_usleep;
332 	t_target = (double)in.dbsz / (double)speed;
333 	t_usleep = t_target - t_io;
334 	if (t_usleep > 0)
335 		usleep(t_usleep * 1000000);
336 	else
337 		t_usleep = 0;
338 	t_prev = t_now;
339 }
340 
341 static void
342 swapbytes(void *v, size_t len)
343 {
344 	unsigned char *p = v;
345 	unsigned char t;
346 
347 	while (len > 1) {
348 		t = p[0];
349 		p[0] = p[1];
350 		p[1] = t;
351 		p += 2;
352 		len -= 2;
353 	}
354 }
355 
356 static void
357 dd_in(void)
358 {
359 	ssize_t n;
360 
361 	for (;;) {
362 		switch (cpy_cnt) {
363 		case -1:			/* count=0 was specified */
364 			return;
365 		case 0:
366 			break;
367 		default:
368 			if (st.in_full + st.in_part >= (uintmax_t)cpy_cnt)
369 				return;
370 			break;
371 		}
372 
373 		if (speed > 0)
374 			speed_limit();
375 
376 		/*
377 		 * Zero the buffer first if sync; if doing block operations,
378 		 * use spaces.
379 		 */
380 		if (ddflags & C_SYNC) {
381 			if (ddflags & C_FILL)
382 				memset(in.dbp, fill_char, in.dbsz);
383 			else if (ddflags & (C_BLOCK | C_UNBLOCK))
384 				memset(in.dbp, ' ', in.dbsz);
385 			else
386 				memset(in.dbp, 0, in.dbsz);
387 		}
388 
389 		n = read(in.fd, in.dbp, in.dbsz);
390 		if (n == 0) {
391 			in.dbrcnt = 0;
392 			return;
393 		}
394 
395 		/* Read error. */
396 		if (n == -1) {
397 			/*
398 			 * If noerror not specified, die.  POSIX requires that
399 			 * the warning message be followed by an I/O display.
400 			 */
401 			if (!(ddflags & C_NOERROR))
402 				err(1, "%s", in.name);
403 			warn("%s", in.name);
404 			summary();
405 
406 			/*
407 			 * If it's a seekable file descriptor, seek past the
408 			 * error.  If your OS doesn't do the right thing for
409 			 * raw disks this section should be modified to re-read
410 			 * in sector size chunks.
411 			 */
412 			if (in.flags & ISSEEK &&
413 			    lseek(in.fd, (off_t)in.dbsz, SEEK_CUR))
414 				warn("%s", in.name);
415 
416 			/* If sync not specified, omit block and continue. */
417 			if (!(ddflags & C_SYNC))
418 				continue;
419 
420 			/* Read errors count as full blocks. */
421 			in.dbcnt += in.dbrcnt = in.dbsz;
422 			++st.in_full;
423 
424 		/* Handle full input blocks. */
425 		} else if ((size_t)n == (size_t)in.dbsz) {
426 			in.dbcnt += in.dbrcnt = n;
427 			++st.in_full;
428 
429 		/* Handle partial input blocks. */
430 		} else {
431 			/* If sync, use the entire block. */
432 			if (ddflags & C_SYNC)
433 				in.dbcnt += in.dbrcnt = in.dbsz;
434 			else
435 				in.dbcnt += in.dbrcnt = n;
436 			++st.in_part;
437 		}
438 
439 		/*
440 		 * POSIX states that if bs is set and no other conversions
441 		 * than noerror, notrunc or sync are specified, the block
442 		 * is output without buffering as it is read.
443 		 */
444 		if ((ddflags & ~(C_NOERROR | C_NOTRUNC | C_SYNC)) == C_BS) {
445 			out.dbcnt = in.dbcnt;
446 			dd_out(1);
447 			in.dbcnt = 0;
448 			continue;
449 		}
450 
451 		if (ddflags & C_SWAB) {
452 			if ((n = in.dbrcnt) & 1) {
453 				++st.swab;
454 				--n;
455 			}
456 			swapbytes(in.dbp, (size_t)n);
457 		}
458 
459 		in.dbp += in.dbrcnt;
460 		(*cfunc)();
461 		if (need_summary) {
462 			summary();
463 		}
464 	}
465 }
466 
467 /*
468  * Clean up any remaining I/O and flush output.  If necessary, the output file
469  * is truncated.
470  */
471 static void
472 dd_close(void)
473 {
474 	if (cfunc == def)
475 		def_close();
476 	else if (cfunc == block)
477 		block_close();
478 	else if (cfunc == unblock)
479 		unblock_close();
480 	if (ddflags & C_OSYNC && out.dbcnt && out.dbcnt < out.dbsz) {
481 		if (ddflags & C_FILL)
482 			memset(out.dbp, fill_char, out.dbsz - out.dbcnt);
483 		else if (ddflags & (C_BLOCK | C_UNBLOCK))
484 			memset(out.dbp, ' ', out.dbsz - out.dbcnt);
485 		else
486 			memset(out.dbp, 0, out.dbsz - out.dbcnt);
487 		out.dbcnt = out.dbsz;
488 	}
489 	if (out.dbcnt || pending)
490 		dd_out(1);
491 
492 	/*
493 	 * If the file ends with a hole, ftruncate it to extend its size
494 	 * up to the end of the hole (without having to write any data).
495 	 */
496 	if (out.seek_offset > 0 && (out.flags & ISTRUNC)) {
497 		if (ftruncate(out.fd, out.seek_offset) == -1)
498 			err(1, "truncating %s", out.name);
499 	}
500 }
501 
502 void
503 dd_out(int force)
504 {
505 	u_char *outp;
506 	size_t cnt, i, n;
507 	ssize_t nw;
508 	static int warned;
509 	int sparse;
510 
511 	/*
512 	 * Write one or more blocks out.  The common case is writing a full
513 	 * output block in a single write; increment the full block stats.
514 	 * Otherwise, we're into partial block writes.  If a partial write,
515 	 * and it's a character device, just warn.  If a tape device, quit.
516 	 *
517 	 * The partial writes represent two cases.  1: Where the input block
518 	 * was less than expected so the output block was less than expected.
519 	 * 2: Where the input block was the right size but we were forced to
520 	 * write the block in multiple chunks.  The original versions of dd(1)
521 	 * never wrote a block in more than a single write, so the latter case
522 	 * never happened.
523 	 *
524 	 * One special case is if we're forced to do the write -- in that case
525 	 * we play games with the buffer size, and it's usually a partial write.
526 	 */
527 	outp = out.db;
528 
529 	/*
530 	 * If force, first try to write all pending data, else try to write
531 	 * just one block. Subsequently always write data one full block at
532 	 * a time at most.
533 	 */
534 	for (n = force ? out.dbcnt : out.dbsz;; n = out.dbsz) {
535 		cnt = n;
536 		do {
537 			sparse = 0;
538 			if (ddflags & C_SPARSE) {
539 				sparse = 1;	/* Is buffer sparse? */
540 				for (i = 0; i < cnt; i++)
541 					if (outp[i] != 0) {
542 						sparse = 0;
543 						break;
544 					}
545 			}
546 			if (sparse && !force) {
547 				pending += cnt;
548 				nw = cnt;
549 			} else {
550 				if (pending != 0) {
551 					/*
552 					 * Seek past hole.  Note that we need to record the
553 					 * reached offset, because we might have no more data
554 					 * to write, in which case we'll need to call
555 					 * ftruncate to extend the file size.
556 					 */
557 					out.seek_offset = lseek(out.fd, pending, SEEK_CUR);
558 					if (out.seek_offset == -1)
559 						err(2, "%s: seek error creating sparse file",
560 						    out.name);
561 					pending = 0;
562 				}
563 				if (cnt) {
564 					nw = write(out.fd, outp, cnt);
565 					out.seek_offset = 0;
566 				} else {
567 					return;
568 				}
569 			}
570 
571 			if (nw <= 0) {
572 				if (nw == 0)
573 					errx(1, "%s: end of device", out.name);
574 				if (errno != EINTR)
575 					err(1, "%s", out.name);
576 				nw = 0;
577 			}
578 
579 			outp += nw;
580 			st.bytes += nw;
581 
582 			if ((size_t)nw == n && n == (size_t)out.dbsz)
583 				++st.out_full;
584 			else
585 				++st.out_part;
586 
587 			if ((size_t) nw != cnt) {
588 				if (out.flags & ISTAPE)
589 					errx(1, "%s: short write on tape device",
590 				    	out.name);
591 				if (out.flags & ISCHR && !warned) {
592 					warned = 1;
593 					warnx("%s: short write on character device",
594 				    	out.name);
595 				}
596 			}
597 
598 			cnt -= nw;
599 		} while (cnt != 0);
600 
601 		if ((out.dbcnt -= n) < out.dbsz)
602 			break;
603 	}
604 
605 	/* Reassemble the output block. */
606 	if (out.dbcnt)
607 		(void)memmove(out.db, out.dbp - out.dbcnt, out.dbcnt);
608 	out.dbp = out.db + out.dbcnt;
609 }
610