Home | History | Annotate | Download | only in dd
      1 /*	$NetBSD: dd.c,v 1.49 2012/02/21 01:49:01 matt 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 1.49 2012/02/21 01:49:01 matt 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 
     76 int main(int, char *[]);
     77 
     78 IO		in, out;		/* input/output state */
     79 STAT		st;			/* statistics */
     80 void		(*cfunc)(void);		/* conversion function */
     81 uint64_t	cpy_cnt;		/* # of blocks to copy */
     82 static off_t	pending = 0;		/* pending seek if sparse */
     83 u_int		ddflags;		/* conversion options */
     84 uint64_t	cbsz;			/* conversion block size */
     85 u_int		files_cnt = 1;		/* # of files to copy */
     86 uint64_t	progress = 0;		/* display sign of life */
     87 const u_char	*ctab;			/* conversion table */
     88 sigset_t	infoset;		/* a set blocking SIGINFO */
     89 const char	*msgfmt = "posix";	/* default summary() message format */
     90 
     91 /*
     92  * Ops for stdin/stdout and crunch'd dd.  These are always host ops.
     93  */
     94 static const struct ddfops ddfops_stdfd = {
     95 	.op_open = open,
     96 	.op_close = close,
     97 	.op_fcntl = fcntl,
     98 	.op_ioctl = ioctl,
     99 	.op_fstat = fstat,
    100 	.op_fsync = fsync,
    101 	.op_ftruncate = ftruncate,
    102 	.op_lseek = lseek,
    103 	.op_read = read,
    104 	.op_write = write,
    105 };
    106 extern const struct ddfops ddfops_prog;
    107 
    108 int
    109 main(int argc, char *argv[])
    110 {
    111 	int ch;
    112 
    113 	setprogname(argv[0]);
    114 	(void)setlocale(LC_ALL, "");
    115 
    116 	while ((ch = getopt(argc, argv, "")) != -1) {
    117 		switch (ch) {
    118 		default:
    119 			errx(EXIT_FAILURE, "usage: dd [operand ...]");
    120 			/* NOTREACHED */
    121 		}
    122 	}
    123 	argc -= (optind - 1);
    124 	argv += (optind - 1);
    125 
    126 	jcl(argv);
    127 #ifndef CRUNCHOPS
    128 	if (ddfops_prog.op_init && ddfops_prog.op_init() == -1)
    129 		err(1, "prog init");
    130 #endif
    131 	setup();
    132 
    133 	(void)signal(SIGINFO, summaryx);
    134 	(void)signal(SIGINT, terminate);
    135 	(void)sigemptyset(&infoset);
    136 	(void)sigaddset(&infoset, SIGINFO);
    137 
    138 	(void)atexit(summary);
    139 
    140 	while (files_cnt--)
    141 		dd_in();
    142 
    143 	dd_close();
    144 	exit(0);
    145 	/* NOTREACHED */
    146 }
    147 
    148 static void
    149 setup(void)
    150 {
    151 #ifdef CRUNCHOPS
    152 	const struct ddfops *prog_ops = &ddfops_stdfd;
    153 #else
    154 	const struct ddfops *prog_ops = &ddfops_prog;
    155 #endif
    156 
    157 	if (in.name == NULL) {
    158 		in.name = "stdin";
    159 		in.fd = STDIN_FILENO;
    160 		in.ops = &ddfops_stdfd;
    161 	} else {
    162 		in.ops = prog_ops;
    163 		in.fd = ddop_open(in, in.name, O_RDONLY, 0);
    164 		if (in.fd < 0)
    165 			err(EXIT_FAILURE, "%s", in.name);
    166 			/* NOTREACHED */
    167 
    168 		/* Ensure in.fd is outside the stdio descriptor range */
    169 		redup_clean_fd(&in);
    170 	}
    171 
    172 	getfdtype(&in);
    173 
    174 	if (files_cnt > 1 && !(in.flags & ISTAPE)) {
    175 		errx(EXIT_FAILURE, "files is not supported for non-tape devices");
    176 		/* NOTREACHED */
    177 	}
    178 
    179 	if (out.name == NULL) {
    180 		/* No way to check for read access here. */
    181 		out.fd = STDOUT_FILENO;
    182 		out.name = "stdout";
    183 		out.ops = &ddfops_stdfd;
    184 	} else {
    185 		out.ops = prog_ops;
    186 #define	OFLAGS \
    187     (O_CREAT | (ddflags & (C_SEEK | C_NOTRUNC) ? 0 : O_TRUNC))
    188 		out.fd = ddop_open(out, out.name, O_RDWR | OFLAGS, DEFFILEMODE);
    189 		/*
    190 		 * May not have read access, so try again with write only.
    191 		 * Without read we may have a problem if output also does
    192 		 * not support seeks.
    193 		 */
    194 		if (out.fd < 0) {
    195 			out.fd = ddop_open(out, out.name, O_WRONLY | OFLAGS,
    196 			    DEFFILEMODE);
    197 			out.flags |= NOREAD;
    198 		}
    199 		if (out.fd < 0) {
    200 			err(EXIT_FAILURE, "%s", out.name);
    201 			/* NOTREACHED */
    202 		}
    203 
    204 		/* Ensure out.fd is outside the stdio descriptor range */
    205 		redup_clean_fd(&out);
    206 	}
    207 
    208 	getfdtype(&out);
    209 
    210 	/*
    211 	 * Allocate space for the input and output buffers.  If not doing
    212 	 * record oriented I/O, only need a single buffer.
    213 	 */
    214 	if (!(ddflags & (C_BLOCK|C_UNBLOCK))) {
    215 		size_t dbsz = out.dbsz;
    216 		if (!(ddflags & C_BS))
    217 			dbsz += in.dbsz - 1;
    218 		if ((in.db = malloc(dbsz)) == NULL) {
    219 			err(EXIT_FAILURE, NULL);
    220 			/* NOTREACHED */
    221 		}
    222 		out.db = in.db;
    223 	} else if ((in.db =
    224 	    malloc((u_int)(MAX(in.dbsz, cbsz) + cbsz))) == NULL ||
    225 	    (out.db = malloc((u_int)(out.dbsz + cbsz))) == NULL) {
    226 		err(EXIT_FAILURE, NULL);
    227 		/* NOTREACHED */
    228 	}
    229 	in.dbp = in.db;
    230 	out.dbp = out.db;
    231 
    232 	/* Position the input/output streams. */
    233 	if (in.offset)
    234 		pos_in();
    235 	if (out.offset)
    236 		pos_out();
    237 
    238 	/*
    239 	 * Truncate the output file; ignore errors because it fails on some
    240 	 * kinds of output files, tapes, for example.
    241 	 */
    242 	if ((ddflags & (C_OF | C_SEEK | C_NOTRUNC)) == (C_OF | C_SEEK))
    243 		(void)ddop_ftruncate(out, out.fd, (off_t)out.offset * out.dbsz);
    244 
    245 	/*
    246 	 * If converting case at the same time as another conversion, build a
    247 	 * table that does both at once.  If just converting case, use the
    248 	 * built-in tables.
    249 	 */
    250 	if (ddflags & (C_LCASE|C_UCASE)) {
    251 #ifdef	NO_CONV
    252 		/* Should not get here, but just in case... */
    253 		errx(EXIT_FAILURE, "case conv and -DNO_CONV");
    254 		/* NOTREACHED */
    255 #else	/* NO_CONV */
    256 		u_int cnt;
    257 
    258 		if (ddflags & C_ASCII || ddflags & C_EBCDIC) {
    259 			if (ddflags & C_LCASE) {
    260 				for (cnt = 0; cnt < 256; ++cnt)
    261 					casetab[cnt] = tolower(ctab[cnt]);
    262 			} else {
    263 				for (cnt = 0; cnt < 256; ++cnt)
    264 					casetab[cnt] = toupper(ctab[cnt]);
    265 			}
    266 		} else {
    267 			if (ddflags & C_LCASE) {
    268 				for (cnt = 0; cnt < 256; ++cnt)
    269 					casetab[cnt] = tolower(cnt);
    270 			} else {
    271 				for (cnt = 0; cnt < 256; ++cnt)
    272 					casetab[cnt] = toupper(cnt);
    273 			}
    274 		}
    275 
    276 		ctab = casetab;
    277 #endif	/* NO_CONV */
    278 	}
    279 
    280 	(void)gettimeofday(&st.start, NULL);	/* Statistics timestamp. */
    281 }
    282 
    283 static void
    284 getfdtype(IO *io)
    285 {
    286 	struct mtget mt;
    287 	struct stat sb;
    288 
    289 	if (io->ops->op_fstat(io->fd, &sb)) {
    290 		err(EXIT_FAILURE, "%s", io->name);
    291 		/* NOTREACHED */
    292 	}
    293 	if (S_ISCHR(sb.st_mode))
    294 		io->flags |= io->ops->op_ioctl(io->fd, MTIOCGET, &mt)
    295 		    ? ISCHR : ISTAPE;
    296 	else if (io->ops->op_lseek(io->fd, (off_t)0, SEEK_CUR) == -1
    297 	    && errno == ESPIPE)
    298 		io->flags |= ISPIPE;		/* XXX fixed in 4.4BSD */
    299 }
    300 
    301 /*
    302  * Move the parameter file descriptor to a descriptor that is outside the
    303  * stdio descriptor range, if necessary.  This is required to avoid
    304  * accidentally outputting completion or error messages into the
    305  * output file that were intended for the tty.
    306  */
    307 static void
    308 redup_clean_fd(IO *io)
    309 {
    310 	int fd = io->fd;
    311 	int newfd;
    312 
    313 	if (fd != STDIN_FILENO && fd != STDOUT_FILENO &&
    314 	    fd != STDERR_FILENO)
    315 		/* File descriptor is ok, return immediately. */
    316 		return;
    317 
    318 	/*
    319 	 * 3 is the first descriptor greater than STD*_FILENO.  Any
    320 	 * free descriptor valued 3 or above is acceptable...
    321 	 */
    322 	newfd = io->ops->op_fcntl(fd, F_DUPFD, 3);
    323 	if (newfd < 0) {
    324 		err(EXIT_FAILURE, "dupfd IO");
    325 		/* NOTREACHED */
    326 	}
    327 
    328 	io->ops->op_close(fd);
    329 	io->fd = newfd;
    330 }
    331 
    332 static void
    333 dd_in(void)
    334 {
    335 	int flags;
    336 	int64_t n;
    337 
    338 	for (flags = ddflags;;) {
    339 		if (cpy_cnt && (st.in_full + st.in_part) >= cpy_cnt)
    340 			return;
    341 
    342 		/*
    343 		 * Clear the buffer first if doing "sync" on input.
    344 		 * If doing block operations use spaces.  This will
    345 		 * affect not only the C_NOERROR case, but also the
    346 		 * last partial input block which should be padded
    347 		 * with zero and not garbage.
    348 		 */
    349 		if (flags & C_SYNC) {
    350 			if (flags & (C_BLOCK|C_UNBLOCK))
    351 				(void)memset(in.dbp, ' ', in.dbsz);
    352 			else
    353 				(void)memset(in.dbp, 0, in.dbsz);
    354 		}
    355 
    356 		n = ddop_read(in, in.fd, in.dbp, in.dbsz);
    357 		if (n == 0) {
    358 			in.dbrcnt = 0;
    359 			return;
    360 		}
    361 
    362 		/* Read error. */
    363 		if (n < 0) {
    364 
    365 			/*
    366 			 * If noerror not specified, die.  POSIX requires that
    367 			 * the warning message be followed by an I/O display.
    368 			 */
    369 			if (!(flags & C_NOERROR)) {
    370 				err(EXIT_FAILURE, "%s", in.name);
    371 				/* NOTREACHED */
    372 			}
    373 			warn("%s", in.name);
    374 			summary();
    375 
    376 			/*
    377 			 * If it's not a tape drive or a pipe, seek past the
    378 			 * error.  If your OS doesn't do the right thing for
    379 			 * raw disks this section should be modified to re-read
    380 			 * in sector size chunks.
    381 			 */
    382 			if (!(in.flags & (ISPIPE|ISTAPE)) &&
    383 			    ddop_lseek(in, in.fd, (off_t)in.dbsz, SEEK_CUR))
    384 				warn("%s", in.name);
    385 
    386 			/* If sync not specified, omit block and continue. */
    387 			if (!(ddflags & C_SYNC))
    388 				continue;
    389 
    390 			/* Read errors count as full blocks. */
    391 			in.dbcnt += in.dbrcnt = in.dbsz;
    392 			++st.in_full;
    393 
    394 		/* Handle full input blocks. */
    395 		} else if ((uint64_t)n == in.dbsz) {
    396 			in.dbcnt += in.dbrcnt = n;
    397 			++st.in_full;
    398 
    399 		/* Handle partial input blocks. */
    400 		} else {
    401 			/* If sync, use the entire block. */
    402 			if (ddflags & C_SYNC)
    403 				in.dbcnt += in.dbrcnt = in.dbsz;
    404 			else
    405 				in.dbcnt += in.dbrcnt = n;
    406 			++st.in_part;
    407 		}
    408 
    409 		/*
    410 		 * POSIX states that if bs is set and no other conversions
    411 		 * than noerror, notrunc or sync are specified, the block
    412 		 * is output without buffering as it is read.
    413 		 */
    414 		if (ddflags & C_BS) {
    415 			out.dbcnt = in.dbcnt;
    416 			dd_out(1);
    417 			in.dbcnt = 0;
    418 			continue;
    419 		}
    420 
    421 		if (ddflags & C_SWAB) {
    422 			if ((n = in.dbrcnt) & 1) {
    423 				++st.swab;
    424 				--n;
    425 			}
    426 			swab(in.dbp, in.dbp, n);
    427 		}
    428 
    429 		in.dbp += in.dbrcnt;
    430 		(*cfunc)();
    431 	}
    432 }
    433 
    434 /*
    435  * Cleanup any remaining I/O and flush output.  If necessary, output file
    436  * is truncated.
    437  */
    438 static void
    439 dd_close(void)
    440 {
    441 
    442 	if (cfunc == def)
    443 		def_close();
    444 	else if (cfunc == block)
    445 		block_close();
    446 	else if (cfunc == unblock)
    447 		unblock_close();
    448 	if (ddflags & C_OSYNC && out.dbcnt < out.dbsz) {
    449 		(void)memset(out.dbp, 0, out.dbsz - out.dbcnt);
    450 		out.dbcnt = out.dbsz;
    451 	}
    452 	/* If there are pending sparse blocks, make sure
    453 	 * to write out the final block un-sparse
    454 	 */
    455 	if ((out.dbcnt == 0) && pending) {
    456 		memset(out.db, 0, out.dbsz);
    457 		out.dbcnt = out.dbsz;
    458 		out.dbp = out.db + out.dbcnt;
    459 		pending -= out.dbsz;
    460 	}
    461 	if (out.dbcnt)
    462 		dd_out(1);
    463 
    464 	/*
    465 	 * Reporting nfs write error may be deferred until next
    466 	 * write(2) or close(2) system call.  So, we need to do an
    467 	 * extra check.  If an output is stdout, the file structure
    468 	 * may be shared with other processes and close(2) just
    469 	 * decreases the reference count.
    470 	 */
    471 	if (out.fd == STDOUT_FILENO && ddop_fsync(out, out.fd) == -1
    472 	    && errno != EINVAL) {
    473 		err(EXIT_FAILURE, "fsync stdout");
    474 		/* NOTREACHED */
    475 	}
    476 	if (ddop_close(out, out.fd) == -1) {
    477 		err(EXIT_FAILURE, "close");
    478 		/* NOTREACHED */
    479 	}
    480 }
    481 
    482 void
    483 dd_out(int force)
    484 {
    485 	static int warned;
    486 	int64_t cnt, n, nw;
    487 	u_char *outp;
    488 
    489 	/*
    490 	 * Write one or more blocks out.  The common case is writing a full
    491 	 * output block in a single write; increment the full block stats.
    492 	 * Otherwise, we're into partial block writes.  If a partial write,
    493 	 * and it's a character device, just warn.  If a tape device, quit.
    494 	 *
    495 	 * The partial writes represent two cases.  1: Where the input block
    496 	 * was less than expected so the output block was less than expected.
    497 	 * 2: Where the input block was the right size but we were forced to
    498 	 * write the block in multiple chunks.  The original versions of dd(1)
    499 	 * never wrote a block in more than a single write, so the latter case
    500 	 * never happened.
    501 	 *
    502 	 * One special case is if we're forced to do the write -- in that case
    503 	 * we play games with the buffer size, and it's usually a partial write.
    504 	 */
    505 	outp = out.db;
    506 	for (n = force ? out.dbcnt : out.dbsz;; n = out.dbsz) {
    507 		for (cnt = n;; cnt -= nw) {
    508 
    509 			if (!force && ddflags & C_SPARSE) {
    510 				int sparse, i;
    511 				sparse = 1;	/* Is buffer sparse? */
    512 				for (i = 0; i < cnt; i++)
    513 					if (outp[i] != 0) {
    514 						sparse = 0;
    515 						break;
    516 					}
    517 				if (sparse) {
    518 					pending += cnt;
    519 					outp += cnt;
    520 					nw = 0;
    521 					break;
    522 				}
    523 			}
    524 			if (pending != 0) {
    525 				if (ddop_lseek(out,
    526 				    out.fd, pending, SEEK_CUR) == -1)
    527 					err(EXIT_FAILURE, "%s: seek error creating sparse file",
    528 					    out.name);
    529 			}
    530 			nw = bwrite(&out, outp, cnt);
    531 			if (nw <= 0) {
    532 				if (nw == 0)
    533 					errx(EXIT_FAILURE,
    534 						"%s: end of device", out.name);
    535 					/* NOTREACHED */
    536 				if (errno != EINTR)
    537 					err(EXIT_FAILURE, "%s", out.name);
    538 					/* NOTREACHED */
    539 				nw = 0;
    540 			}
    541 			if (pending) {
    542 				st.bytes += pending;
    543 				st.sparse += pending/out.dbsz;
    544 				st.out_full += pending/out.dbsz;
    545 				pending = 0;
    546 			}
    547 			outp += nw;
    548 			st.bytes += nw;
    549 			if (nw == n) {
    550 				if ((uint64_t)n != out.dbsz)
    551 					++st.out_part;
    552 				else
    553 					++st.out_full;
    554 				break;
    555 			}
    556 			++st.out_part;
    557 			if (nw == cnt)
    558 				break;
    559 			if (out.flags & ISCHR && !warned) {
    560 				warned = 1;
    561 				warnx("%s: short write on character device", out.name);
    562 			}
    563 			if (out.flags & ISTAPE)
    564 				errx(EXIT_FAILURE,
    565 					"%s: short write on tape device", out.name);
    566 				/* NOTREACHED */
    567 
    568 		}
    569 		if ((out.dbcnt -= n) < out.dbsz)
    570 			break;
    571 	}
    572 
    573 	/* Reassemble the output block. */
    574 	if (out.dbcnt)
    575 		(void)memmove(out.db, out.dbp - out.dbcnt, out.dbcnt);
    576 	out.dbp = out.db + out.dbcnt;
    577 
    578 	if (progress && (st.out_full + st.out_part) % progress == 0)
    579 		(void)write(STDERR_FILENO, ".", 1);
    580 }
    581 
    582 /*
    583  * A protected against SIGINFO write
    584  */
    585 ssize_t
    586 bwrite(IO *io, const void *buf, size_t len)
    587 {
    588 	sigset_t oset;
    589 	ssize_t rv;
    590 	int oerrno;
    591 
    592 	(void)sigprocmask(SIG_BLOCK, &infoset, &oset);
    593 	rv = io->ops->op_write(io->fd, buf, len);
    594 	oerrno = errno;
    595 	(void)sigprocmask(SIG_SETMASK, &oset, NULL);
    596 	errno = oerrno;
    597 	return (rv);
    598 }
    599