xref: /netbsd/bin/dd/dd.c (revision 38a2b7e9)
1 /*	$NetBSD: dd.c,v 1.53 2019/10/04 08:57:37 mrg Exp $	*/
2 
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 #include <sys/cdefs.h>
37 #ifndef lint
38 __COPYRIGHT("@(#) Copyright (c) 1991, 1993, 1994\
39  The Regents of the University of California.  All rights reserved.");
40 #endif /* not lint */
41 
42 #ifndef lint
43 #if 0
44 static char sccsid[] = "@(#)dd.c	8.5 (Berkeley) 4/2/94";
45 #else
46 __RCSID("$NetBSD: dd.c,v 1.53 2019/10/04 08:57:37 mrg Exp $");
47 #endif
48 #endif /* not lint */
49 
50 #include <sys/param.h>
51 #include <sys/stat.h>
52 #include <sys/ioctl.h>
53 #include <sys/mtio.h>
54 #include <sys/time.h>
55 
56 #include <ctype.h>
57 #include <err.h>
58 #include <errno.h>
59 #include <fcntl.h>
60 #include <locale.h>
61 #include <signal.h>
62 #include <stdio.h>
63 #include <stdlib.h>
64 #include <string.h>
65 #include <unistd.h>
66 
67 #include "dd.h"
68 #include "extern.h"
69 
70 static void dd_close(void);
71 static void dd_in(void);
72 static void getfdtype(IO *);
73 static void redup_clean_fd(IO *);
74 static void setup(void);
75 static void *buffer_alloc(size_t);
76 
77 IO		in, out;		/* input/output state */
78 STAT		st;			/* statistics */
79 void		(*cfunc)(void);		/* conversion function */
80 uint64_t	cpy_cnt;		/* # of blocks to copy */
81 static off_t	pending = 0;		/* pending seek if sparse */
82 u_int		ddflags;		/* conversion options */
83 #ifdef NO_IOFLAG
84 #define		iflag	O_RDONLY
85 #define		oflag	O_CREAT
86 #else
87 u_int		iflag = O_RDONLY;	/* open(2) flags for input file */
88 u_int		oflag = O_CREAT;	/* open(2) flags for output file */
89 #endif /* NO_IOFLAG */
90 uint64_t	cbsz;			/* conversion block size */
91 u_int		files_cnt = 1;		/* # of files to copy */
92 uint64_t	progress = 0;		/* display sign of life */
93 const u_char	*ctab;			/* conversion table */
94 sigset_t	infoset;		/* a set blocking SIGINFO */
95 const char	*msgfmt = "posix";	/* default summary() message format */
96 
97 /*
98  * Ops for stdin/stdout and crunch'd dd.  These are always host ops.
99  */
100 static const struct ddfops ddfops_stdfd = {
101 	.op_open = open,
102 	.op_close = close,
103 	.op_fcntl = fcntl,
104 	.op_ioctl = ioctl,
105 	.op_fstat = fstat,
106 	.op_fsync = fsync,
107 	.op_ftruncate = ftruncate,
108 	.op_lseek = lseek,
109 	.op_read = read,
110 	.op_write = write,
111 };
112 extern const struct ddfops ddfops_prog;
113 
114 int
main(int argc,char * argv[])115 main(int argc, char *argv[])
116 {
117 	int ch;
118 
119 	setprogname(argv[0]);
120 	(void)setlocale(LC_ALL, "");
121 
122 	while ((ch = getopt(argc, argv, "")) != -1) {
123 		switch (ch) {
124 		default:
125 			errx(EXIT_FAILURE, "usage: dd [operand ...]");
126 			/* NOTREACHED */
127 		}
128 	}
129 	argc -= (optind - 1);
130 	argv += (optind - 1);
131 
132 	jcl(argv);
133 #ifndef CRUNCHOPS
134 	if (ddfops_prog.op_init && ddfops_prog.op_init() == -1)
135 		err(1, "prog init");
136 #endif
137 	setup();
138 
139 	(void)signal(SIGINFO, summaryx);
140 	(void)signal(SIGINT, terminate);
141 	(void)sigemptyset(&infoset);
142 	(void)sigaddset(&infoset, SIGINFO);
143 
144 	(void)atexit(summary);
145 
146 	while (files_cnt--)
147 		dd_in();
148 
149 	dd_close();
150 	exit(0);
151 	/* NOTREACHED */
152 }
153 
154 static void *
buffer_alloc(size_t sz)155 buffer_alloc(size_t sz)
156 {
157 	size_t align = getpagesize();
158 	void *res;
159 
160 	if (sz < align)
161 		res = malloc(sz);
162 	else if (posix_memalign(&res, align, sz) != 0)
163 		res = NULL;
164 
165 	return res;
166 }
167 
168 static void
setup(void)169 setup(void)
170 {
171 #ifdef CRUNCHOPS
172 	const struct ddfops *prog_ops = &ddfops_stdfd;
173 #else
174 	const struct ddfops *prog_ops = &ddfops_prog;
175 #endif
176 
177 	if (in.name == NULL) {
178 		in.name = "stdin";
179 		in.fd = STDIN_FILENO;
180 		in.ops = &ddfops_stdfd;
181 	} else {
182 		in.ops = prog_ops;
183 		in.fd = ddop_open(in, in.name, iflag, 0);
184 		if (in.fd < 0)
185 			err(EXIT_FAILURE, "%s", in.name);
186 			/* NOTREACHED */
187 
188 		/* Ensure in.fd is outside the stdio descriptor range */
189 		redup_clean_fd(&in);
190 	}
191 
192 	getfdtype(&in);
193 
194 	if (files_cnt > 1 && !(in.flags & ISTAPE)) {
195 		errx(EXIT_FAILURE, "files is not supported for non-tape devices");
196 		/* NOTREACHED */
197 	}
198 
199 	if (out.name == NULL) {
200 		/* No way to check for read access here. */
201 		out.fd = STDOUT_FILENO;
202 		out.name = "stdout";
203 		out.ops = &ddfops_stdfd;
204 	} else {
205 		out.ops = prog_ops;
206 
207 #ifndef NO_IOFLAG
208 		if ((oflag & O_TRUNC) && (ddflags & C_SEEK)) {
209 			errx(EXIT_FAILURE, "oflag=trunc is incompatible "
210 			     "with seek or oseek operands, giving up.");
211 			/* NOTREACHED */
212 		}
213 		if ((oflag & O_TRUNC) && (ddflags & C_NOTRUNC)) {
214 			errx(EXIT_FAILURE, "oflag=trunc is incompatible "
215 			     "with conv=notrunc operand, giving up.");
216 			/* NOTREACHED */
217 		}
218 #endif /* NO_IOFLAG */
219 #define	OFLAGS \
220     (oflag | (ddflags & (C_SEEK | C_NOTRUNC) ? 0 : O_TRUNC))
221 		out.fd = ddop_open(out, out.name, O_RDWR | OFLAGS, DEFFILEMODE);
222 		/*
223 		 * May not have read access, so try again with write only.
224 		 * Without read we may have a problem if output also does
225 		 * not support seeks.
226 		 */
227 		if (out.fd < 0) {
228 			out.fd = ddop_open(out, out.name, O_WRONLY | OFLAGS,
229 			    DEFFILEMODE);
230 			out.flags |= NOREAD;
231 		}
232 		if (out.fd < 0) {
233 			err(EXIT_FAILURE, "%s", out.name);
234 			/* NOTREACHED */
235 		}
236 
237 		/* Ensure out.fd is outside the stdio descriptor range */
238 		redup_clean_fd(&out);
239 	}
240 
241 	getfdtype(&out);
242 
243 	/*
244 	 * Allocate space for the input and output buffers.  If not doing
245 	 * record oriented I/O, only need a single buffer.
246 	 */
247 	if (!(ddflags & (C_BLOCK|C_UNBLOCK))) {
248 		size_t dbsz = out.dbsz;
249 		if (!(ddflags & C_BS))
250 			dbsz += in.dbsz - 1;
251 		if ((in.db = buffer_alloc(dbsz)) == NULL) {
252 			err(EXIT_FAILURE, NULL);
253 			/* NOTREACHED */
254 		}
255 		out.db = in.db;
256 	} else if ((in.db =
257 	    buffer_alloc((u_int)(MAX(in.dbsz, cbsz) + cbsz))) == NULL ||
258 	    (out.db = buffer_alloc((u_int)(out.dbsz + cbsz))) == NULL) {
259 		err(EXIT_FAILURE, NULL);
260 		/* NOTREACHED */
261 	}
262 	in.dbp = in.db;
263 	out.dbp = out.db;
264 
265 	/* Position the input/output streams. */
266 	if (in.offset)
267 		pos_in();
268 	if (out.offset)
269 		pos_out();
270 
271 	/*
272 	 * Truncate the output file; ignore errors because it fails on some
273 	 * kinds of output files, tapes, for example.
274 	 */
275 	if ((ddflags & (C_OF | C_SEEK | C_NOTRUNC)) == (C_OF | C_SEEK))
276 		(void)ddop_ftruncate(out, out.fd, (off_t)out.offset * out.dbsz);
277 
278 	/*
279 	 * If converting case at the same time as another conversion, build a
280 	 * table that does both at once.  If just converting case, use the
281 	 * built-in tables.
282 	 */
283 	if (ddflags & (C_LCASE|C_UCASE)) {
284 #ifdef	NO_CONV
285 		/* Should not get here, but just in case... */
286 		errx(EXIT_FAILURE, "case conv and -DNO_CONV");
287 		/* NOTREACHED */
288 #else	/* NO_CONV */
289 		u_int cnt;
290 
291 		if (ddflags & C_ASCII || ddflags & C_EBCDIC) {
292 			if (ddflags & C_LCASE) {
293 				for (cnt = 0; cnt < 256; ++cnt)
294 					casetab[cnt] = tolower(ctab[cnt]);
295 			} else {
296 				for (cnt = 0; cnt < 256; ++cnt)
297 					casetab[cnt] = toupper(ctab[cnt]);
298 			}
299 		} else {
300 			if (ddflags & C_LCASE) {
301 				for (cnt = 0; cnt < 256; ++cnt)
302 					casetab[cnt] = tolower(cnt);
303 			} else {
304 				for (cnt = 0; cnt < 256; ++cnt)
305 					casetab[cnt] = toupper(cnt);
306 			}
307 		}
308 
309 		ctab = casetab;
310 #endif	/* NO_CONV */
311 	}
312 
313 	(void)gettimeofday(&st.start, NULL);	/* Statistics timestamp. */
314 }
315 
316 static void
getfdtype(IO * io)317 getfdtype(IO *io)
318 {
319 	struct mtget mt;
320 	struct stat sb;
321 
322 	if (io->ops->op_fstat(io->fd, &sb)) {
323 		err(EXIT_FAILURE, "%s", io->name);
324 		/* NOTREACHED */
325 	}
326 	if (S_ISCHR(sb.st_mode))
327 		io->flags |= io->ops->op_ioctl(io->fd, MTIOCGET, &mt)
328 		    ? ISCHR : ISTAPE;
329 	else if (io->ops->op_lseek(io->fd, (off_t)0, SEEK_CUR) == -1
330 	    && errno == ESPIPE)
331 		io->flags |= ISPIPE;		/* XXX fixed in 4.4BSD */
332 }
333 
334 /*
335  * Move the parameter file descriptor to a descriptor that is outside the
336  * stdio descriptor range, if necessary.  This is required to avoid
337  * accidentally outputting completion or error messages into the
338  * output file that were intended for the tty.
339  */
340 static void
redup_clean_fd(IO * io)341 redup_clean_fd(IO *io)
342 {
343 	int fd = io->fd;
344 	int newfd;
345 
346 	if (fd != STDIN_FILENO && fd != STDOUT_FILENO &&
347 	    fd != STDERR_FILENO)
348 		/* File descriptor is ok, return immediately. */
349 		return;
350 
351 	/*
352 	 * 3 is the first descriptor greater than STD*_FILENO.  Any
353 	 * free descriptor valued 3 or above is acceptable...
354 	 */
355 	newfd = io->ops->op_fcntl(fd, F_DUPFD, 3);
356 	if (newfd < 0) {
357 		err(EXIT_FAILURE, "dupfd IO");
358 		/* NOTREACHED */
359 	}
360 
361 	io->ops->op_close(fd);
362 	io->fd = newfd;
363 }
364 
365 static void
dd_in(void)366 dd_in(void)
367 {
368 	int flags;
369 	int64_t n;
370 
371 	for (flags = ddflags;;) {
372 		if (cpy_cnt && (st.in_full + st.in_part) >= cpy_cnt)
373 			return;
374 
375 		/*
376 		 * Clear the buffer first if doing "sync" on input.
377 		 * If doing block operations use spaces.  This will
378 		 * affect not only the C_NOERROR case, but also the
379 		 * last partial input block which should be padded
380 		 * with zero and not garbage.
381 		 */
382 		if (flags & C_SYNC) {
383 			if (flags & (C_BLOCK|C_UNBLOCK))
384 				(void)memset(in.dbp, ' ', in.dbsz);
385 			else
386 				(void)memset(in.dbp, 0, in.dbsz);
387 		}
388 
389 		n = ddop_read(in, in.fd, in.dbp, in.dbsz);
390 		if (n == 0) {
391 			in.dbrcnt = 0;
392 			return;
393 		}
394 
395 		/* Read error. */
396 		if (n < 0) {
397 
398 			/*
399 			 * If noerror not specified, die.  POSIX requires that
400 			 * the warning message be followed by an I/O display.
401 			 */
402 			if (!(flags & C_NOERROR)) {
403 				err(EXIT_FAILURE, "%s", in.name);
404 				/* NOTREACHED */
405 			}
406 			warn("%s", in.name);
407 			summary();
408 
409 			/*
410 			 * If it's not a tape drive or a pipe, seek past the
411 			 * error.  If your OS doesn't do the right thing for
412 			 * raw disks this section should be modified to re-read
413 			 * in sector size chunks.
414 			 */
415 			if (!(in.flags & (ISPIPE|ISTAPE)) &&
416 			    ddop_lseek(in, in.fd, (off_t)in.dbsz, SEEK_CUR))
417 				warn("%s", in.name);
418 
419 			/* If sync not specified, omit block and continue. */
420 			if (!(ddflags & C_SYNC))
421 				continue;
422 
423 			/* Read errors count as full blocks. */
424 			in.dbcnt += in.dbrcnt = in.dbsz;
425 			++st.in_full;
426 
427 		/* Handle full input blocks. */
428 		} else if ((uint64_t)n == in.dbsz) {
429 			in.dbcnt += in.dbrcnt = n;
430 			++st.in_full;
431 
432 		/* Handle partial input blocks. */
433 		} else {
434 			/* If sync, use the entire block. */
435 			if (ddflags & C_SYNC)
436 				in.dbcnt += in.dbrcnt = in.dbsz;
437 			else
438 				in.dbcnt += in.dbrcnt = n;
439 			++st.in_part;
440 		}
441 
442 		/*
443 		 * POSIX states that if bs is set and no other conversions
444 		 * than noerror, notrunc or sync are specified, the block
445 		 * is output without buffering as it is read.
446 		 */
447 		if (ddflags & C_BS) {
448 			out.dbcnt = in.dbcnt;
449 			dd_out(1);
450 			in.dbcnt = 0;
451 			continue;
452 		}
453 
454 		if (ddflags & C_SWAB) {
455 			if ((n = in.dbrcnt) & 1) {
456 				++st.swab;
457 				--n;
458 			}
459 			dd_swab(in.dbp, in.dbp, n);
460 		}
461 
462 		in.dbp += in.dbrcnt;
463 		(*cfunc)();
464 	}
465 }
466 
467 /*
468  * Cleanup any remaining I/O and flush output.  If necessary, output file
469  * is truncated.
470  */
471 static void
dd_close(void)472 dd_close(void)
473 {
474 
475 	if (cfunc == def)
476 		def_close();
477 	else if (cfunc == block)
478 		block_close();
479 	else if (cfunc == unblock)
480 		unblock_close();
481 	if (ddflags & C_OSYNC && out.dbcnt < out.dbsz) {
482 		(void)memset(out.dbp, 0, out.dbsz - out.dbcnt);
483 		out.dbcnt = out.dbsz;
484 	}
485 	/* If there are pending sparse blocks, make sure
486 	 * to write out the final block un-sparse
487 	 */
488 	if ((out.dbcnt == 0) && pending) {
489 		memset(out.db, 0, out.dbsz);
490 		out.dbcnt = out.dbsz;
491 		out.dbp = out.db + out.dbcnt;
492 		pending -= out.dbsz;
493 	}
494 	if (out.dbcnt)
495 		dd_out(1);
496 
497 	/*
498 	 * Reporting nfs write error may be deferred until next
499 	 * write(2) or close(2) system call.  So, we need to do an
500 	 * extra check.  If an output is stdout, the file structure
501 	 * may be shared with other processes and close(2) just
502 	 * decreases the reference count.
503 	 */
504 	if (out.fd == STDOUT_FILENO && ddop_fsync(out, out.fd) == -1
505 	    && errno != EINVAL) {
506 		err(EXIT_FAILURE, "fsync stdout");
507 		/* NOTREACHED */
508 	}
509 	if (ddop_close(out, out.fd) == -1) {
510 		err(EXIT_FAILURE, "close");
511 		/* NOTREACHED */
512 	}
513 }
514 
515 void
dd_out(int force)516 dd_out(int force)
517 {
518 	static int warned;
519 	int64_t cnt, n, nw;
520 	u_char *outp;
521 
522 	/*
523 	 * Write one or more blocks out.  The common case is writing a full
524 	 * output block in a single write; increment the full block stats.
525 	 * Otherwise, we're into partial block writes.  If a partial write,
526 	 * and it's a character device, just warn.  If a tape device, quit.
527 	 *
528 	 * The partial writes represent two cases.  1: Where the input block
529 	 * was less than expected so the output block was less than expected.
530 	 * 2: Where the input block was the right size but we were forced to
531 	 * write the block in multiple chunks.  The original versions of dd(1)
532 	 * never wrote a block in more than a single write, so the latter case
533 	 * never happened.
534 	 *
535 	 * One special case is if we're forced to do the write -- in that case
536 	 * we play games with the buffer size, and it's usually a partial write.
537 	 */
538 	outp = out.db;
539 	for (n = force ? out.dbcnt : out.dbsz;; n = out.dbsz) {
540 		for (cnt = n;; cnt -= nw) {
541 
542 			if (!force && ddflags & C_SPARSE) {
543 				int sparse, i;
544 				sparse = 1;	/* Is buffer sparse? */
545 				for (i = 0; i < cnt; i++)
546 					if (outp[i] != 0) {
547 						sparse = 0;
548 						break;
549 					}
550 				if (sparse) {
551 					pending += cnt;
552 					outp += cnt;
553 					nw = 0;
554 					break;
555 				}
556 			}
557 			if (pending != 0) {
558 				if (ddop_lseek(out,
559 				    out.fd, pending, SEEK_CUR) == -1)
560 					err(EXIT_FAILURE, "%s: seek error creating sparse file",
561 					    out.name);
562 			}
563 			nw = bwrite(&out, outp, cnt);
564 			if (nw <= 0) {
565 				if (nw == 0)
566 					errx(EXIT_FAILURE,
567 						"%s: end of device", out.name);
568 					/* NOTREACHED */
569 				if (errno != EINTR)
570 					err(EXIT_FAILURE, "%s", out.name);
571 					/* NOTREACHED */
572 				nw = 0;
573 			}
574 			if (pending) {
575 				st.bytes += pending;
576 				st.sparse += pending/out.dbsz;
577 				st.out_full += pending/out.dbsz;
578 				pending = 0;
579 			}
580 			outp += nw;
581 			st.bytes += nw;
582 			if (nw == n) {
583 				if ((uint64_t)n != out.dbsz)
584 					++st.out_part;
585 				else
586 					++st.out_full;
587 				break;
588 			}
589 			++st.out_part;
590 			if (nw == cnt)
591 				break;
592 			if (out.flags & ISCHR && !warned) {
593 				warned = 1;
594 				warnx("%s: short write on character device", out.name);
595 			}
596 			if (out.flags & ISTAPE)
597 				errx(EXIT_FAILURE,
598 					"%s: short write on tape device", out.name);
599 				/* NOTREACHED */
600 
601 		}
602 		if ((out.dbcnt -= n) < out.dbsz)
603 			break;
604 	}
605 
606 	/* Reassemble the output block. */
607 	if (out.dbcnt)
608 		(void)memmove(out.db, out.dbp - out.dbcnt, out.dbcnt);
609 	out.dbp = out.db + out.dbcnt;
610 
611 	if (progress && (st.out_full + st.out_part) % progress == 0)
612 		(void)write(STDERR_FILENO, ".", 1);
613 }
614 
615 /*
616  * A protected against SIGINFO write
617  */
618 ssize_t
bwrite(IO * io,const void * buf,size_t len)619 bwrite(IO *io, const void *buf, size_t len)
620 {
621 	sigset_t oset;
622 	ssize_t rv;
623 	int oerrno;
624 
625 	(void)sigprocmask(SIG_BLOCK, &infoset, &oset);
626 	rv = io->ops->op_write(io->fd, buf, len);
627 	oerrno = errno;
628 	(void)sigprocmask(SIG_SETMASK, &oset, NULL);
629 	errno = oerrno;
630 	return (rv);
631 }
632