1 /*- 2 * Copyright (c) 1991, 1993, 1994 3 * The Regents of the University of California. All rights reserved. 4 * 5 * This code is derived from software contributed to Berkeley by 6 * Keith Muller of the University of California, San Diego and Lance 7 * Visser of Convex Computer Corporation. 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions 11 * are met: 12 * 1. Redistributions of source code must retain the above copyright 13 * notice, this list of conditions and the following disclaimer. 14 * 2. Redistributions in binary form must reproduce the above copyright 15 * notice, this list of conditions and the following disclaimer in the 16 * documentation and/or other materials provided with the distribution. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 */ 33 34 #if 0 35 #ifndef lint 36 static char const copyright[] = 37 "@(#) Copyright (c) 1991, 1993, 1994\n\ 38 The Regents of the University of California. All rights reserved.\n"; 39 #endif /* not lint */ 40 41 #ifndef lint 42 static char sccsid[] = "@(#)dd.c 8.5 (Berkeley) 4/2/94"; 43 #endif /* not lint */ 44 #endif 45 #include <sys/cdefs.h> 46 __FBSDID("$FreeBSD$"); 47 48 #include <sys/param.h> 49 #include <sys/stat.h> 50 #include <sys/conf.h> 51 #include <sys/disklabel.h> 52 #include <sys/filio.h> 53 54 #include <assert.h> 55 #include <ctype.h> 56 #include <err.h> 57 #include <errno.h> 58 #include <fcntl.h> 59 #include <inttypes.h> 60 #include <locale.h> 61 #include <stdio.h> 62 #include <stdlib.h> 63 #include <string.h> 64 #include <time.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 setup(void); 74 75 IO in, out; /* input/output state */ 76 STAT st; /* statistics */ 77 void (*cfunc)(void); /* conversion function */ 78 uintmax_t cpy_cnt; /* # of blocks to copy */ 79 static off_t pending = 0; /* pending seek if sparse */ 80 u_int ddflags = 0; /* conversion options */ 81 size_t cbsz; /* conversion block size */ 82 uintmax_t files_cnt = 1; /* # of files to copy */ 83 const u_char *ctab; /* conversion table */ 84 char fill_char; /* Character to fill with if defined */ 85 volatile sig_atomic_t need_summary; 86 87 int 88 main(int argc __unused, char *argv[]) 89 { 90 (void)setlocale(LC_CTYPE, ""); 91 jcl(argv); 92 setup(); 93 94 (void)signal(SIGINFO, siginfo_handler); 95 (void)signal(SIGINT, terminate); 96 97 atexit(summary); 98 99 while (files_cnt--) 100 dd_in(); 101 102 dd_close(); 103 /* 104 * Some devices such as cfi(4) may perform significant amounts 105 * of work when a write descriptor is closed. Close the out 106 * descriptor explicitly so that the summary handler (called 107 * from an atexit() hook) includes this work. 108 */ 109 close(out.fd); 110 exit(0); 111 } 112 113 static int 114 parity(u_char c) 115 { 116 int i; 117 118 i = c ^ (c >> 1) ^ (c >> 2) ^ (c >> 3) ^ 119 (c >> 4) ^ (c >> 5) ^ (c >> 6) ^ (c >> 7); 120 return (i & 1); 121 } 122 123 static void 124 setup(void) 125 { 126 u_int cnt; 127 128 if (in.name == NULL) { 129 in.name = "stdin"; 130 in.fd = STDIN_FILENO; 131 } else { 132 in.fd = open(in.name, O_RDONLY, 0); 133 if (in.fd == -1) 134 err(1, "%s", in.name); 135 } 136 137 getfdtype(&in); 138 139 if (files_cnt > 1 && !(in.flags & ISTAPE)) 140 errx(1, "files is not supported for non-tape devices"); 141 142 if (out.name == NULL) { 143 /* No way to check for read access here. */ 144 out.fd = STDOUT_FILENO; 145 out.name = "stdout"; 146 } else { 147 #define OFLAGS \ 148 (O_CREAT | (ddflags & (C_SEEK | C_NOTRUNC) ? 0 : O_TRUNC)) 149 out.fd = open(out.name, O_RDWR | OFLAGS, DEFFILEMODE); 150 /* 151 * May not have read access, so try again with write only. 152 * Without read we may have a problem if output also does 153 * not support seeks. 154 */ 155 if (out.fd == -1) { 156 out.fd = open(out.name, O_WRONLY | OFLAGS, DEFFILEMODE); 157 out.flags |= NOREAD; 158 } 159 if (out.fd == -1) 160 err(1, "%s", out.name); 161 } 162 163 getfdtype(&out); 164 165 /* 166 * Allocate space for the input and output buffers. If not doing 167 * record oriented I/O, only need a single buffer. 168 */ 169 if (!(ddflags & (C_BLOCK | C_UNBLOCK))) { 170 if ((in.db = malloc(out.dbsz + in.dbsz - 1)) == NULL) 171 err(1, "input buffer"); 172 out.db = in.db; 173 } else if ((in.db = malloc(MAX(in.dbsz, cbsz) + cbsz)) == NULL || 174 (out.db = malloc(out.dbsz + cbsz)) == NULL) 175 err(1, "output buffer"); 176 177 /* dbp is the first free position in each buffer. */ 178 in.dbp = in.db; 179 out.dbp = out.db; 180 181 /* Position the input/output streams. */ 182 if (in.offset) 183 pos_in(); 184 if (out.offset) 185 pos_out(); 186 187 /* 188 * Truncate the output file. If it fails on a type of output file 189 * that it should _not_ fail on, error out. 190 */ 191 if ((ddflags & (C_OF | C_SEEK | C_NOTRUNC)) == (C_OF | C_SEEK) && 192 out.flags & ISTRUNC) 193 if (ftruncate(out.fd, out.offset * out.dbsz) == -1) 194 err(1, "truncating %s", out.name); 195 196 if (ddflags & (C_LCASE | C_UCASE | C_ASCII | C_EBCDIC | C_PARITY)) { 197 if (ctab != NULL) { 198 for (cnt = 0; cnt <= 0377; ++cnt) 199 casetab[cnt] = ctab[cnt]; 200 } else { 201 for (cnt = 0; cnt <= 0377; ++cnt) 202 casetab[cnt] = cnt; 203 } 204 if ((ddflags & C_PARITY) && !(ddflags & C_ASCII)) { 205 /* 206 * If the input is not EBCDIC, and we do parity 207 * processing, strip input parity. 208 */ 209 for (cnt = 200; cnt <= 0377; ++cnt) 210 casetab[cnt] = casetab[cnt & 0x7f]; 211 } 212 if (ddflags & C_LCASE) { 213 for (cnt = 0; cnt <= 0377; ++cnt) 214 casetab[cnt] = tolower(casetab[cnt]); 215 } else if (ddflags & C_UCASE) { 216 for (cnt = 0; cnt <= 0377; ++cnt) 217 casetab[cnt] = toupper(casetab[cnt]); 218 } 219 if ((ddflags & C_PARITY)) { 220 /* 221 * This should strictly speaking be a no-op, but I 222 * wonder what funny LANG settings could get us. 223 */ 224 for (cnt = 0; cnt <= 0377; ++cnt) 225 casetab[cnt] = casetab[cnt] & 0x7f; 226 } 227 if ((ddflags & C_PARSET)) { 228 for (cnt = 0; cnt <= 0377; ++cnt) 229 casetab[cnt] = casetab[cnt] | 0x80; 230 } 231 if ((ddflags & C_PAREVEN)) { 232 for (cnt = 0; cnt <= 0377; ++cnt) 233 if (parity(casetab[cnt])) 234 casetab[cnt] = casetab[cnt] | 0x80; 235 } 236 if ((ddflags & C_PARODD)) { 237 for (cnt = 0; cnt <= 0377; ++cnt) 238 if (!parity(casetab[cnt])) 239 casetab[cnt] = casetab[cnt] | 0x80; 240 } 241 242 ctab = casetab; 243 } 244 245 if (clock_gettime(CLOCK_MONOTONIC, &st.start)) 246 err(1, "clock_gettime"); 247 } 248 249 static void 250 getfdtype(IO *io) 251 { 252 struct stat sb; 253 int type; 254 255 if (fstat(io->fd, &sb) == -1) 256 err(1, "%s", io->name); 257 if (S_ISREG(sb.st_mode)) 258 io->flags |= ISTRUNC; 259 if (S_ISCHR(sb.st_mode) || S_ISBLK(sb.st_mode)) { 260 if (ioctl(io->fd, FIODTYPE, &type) == -1) { 261 err(1, "%s", io->name); 262 } else { 263 if (type & D_TAPE) 264 io->flags |= ISTAPE; 265 else if (type & (D_DISK | D_MEM)) 266 io->flags |= ISSEEK; 267 if (S_ISCHR(sb.st_mode) && (type & D_TAPE) == 0) 268 io->flags |= ISCHR; 269 } 270 return; 271 } 272 errno = 0; 273 if (lseek(io->fd, (off_t)0, SEEK_CUR) == -1 && errno == ESPIPE) 274 io->flags |= ISPIPE; 275 else 276 io->flags |= ISSEEK; 277 } 278 279 static void 280 dd_in(void) 281 { 282 ssize_t n; 283 284 for (;;) { 285 switch (cpy_cnt) { 286 case -1: /* count=0 was specified */ 287 return; 288 case 0: 289 break; 290 default: 291 if (st.in_full + st.in_part >= (uintmax_t)cpy_cnt) 292 return; 293 break; 294 } 295 296 /* 297 * Zero the buffer first if sync; if doing block operations, 298 * use spaces. 299 */ 300 if (ddflags & C_SYNC) { 301 if (ddflags & C_FILL) 302 memset(in.dbp, fill_char, in.dbsz); 303 else if (ddflags & (C_BLOCK | C_UNBLOCK)) 304 memset(in.dbp, ' ', in.dbsz); 305 else 306 memset(in.dbp, 0, in.dbsz); 307 } 308 309 n = read(in.fd, in.dbp, in.dbsz); 310 if (n == 0) { 311 in.dbrcnt = 0; 312 return; 313 } 314 315 /* Read error. */ 316 if (n == -1) { 317 /* 318 * If noerror not specified, die. POSIX requires that 319 * the warning message be followed by an I/O display. 320 */ 321 if (!(ddflags & C_NOERROR)) 322 err(1, "%s", in.name); 323 warn("%s", in.name); 324 summary(); 325 326 /* 327 * If it's a seekable file descriptor, seek past the 328 * error. If your OS doesn't do the right thing for 329 * raw disks this section should be modified to re-read 330 * in sector size chunks. 331 */ 332 if (in.flags & ISSEEK && 333 lseek(in.fd, (off_t)in.dbsz, SEEK_CUR)) 334 warn("%s", in.name); 335 336 /* If sync not specified, omit block and continue. */ 337 if (!(ddflags & C_SYNC)) 338 continue; 339 340 /* Read errors count as full blocks. */ 341 in.dbcnt += in.dbrcnt = in.dbsz; 342 ++st.in_full; 343 344 /* Handle full input blocks. */ 345 } else if ((size_t)n == in.dbsz) { 346 in.dbcnt += in.dbrcnt = n; 347 ++st.in_full; 348 349 /* Handle partial input blocks. */ 350 } else { 351 /* If sync, use the entire block. */ 352 if (ddflags & C_SYNC) 353 in.dbcnt += in.dbrcnt = in.dbsz; 354 else 355 in.dbcnt += in.dbrcnt = n; 356 ++st.in_part; 357 } 358 359 /* 360 * POSIX states that if bs is set and no other conversions 361 * than noerror, notrunc or sync are specified, the block 362 * is output without buffering as it is read. 363 */ 364 if ((ddflags & ~(C_NOERROR | C_NOTRUNC | C_SYNC)) == C_BS) { 365 out.dbcnt = in.dbcnt; 366 dd_out(1); 367 in.dbcnt = 0; 368 continue; 369 } 370 371 if (ddflags & C_SWAB) { 372 if ((n = in.dbrcnt) & 1) { 373 ++st.swab; 374 --n; 375 } 376 swab(in.dbp, in.dbp, (size_t)n); 377 } 378 379 in.dbp += in.dbrcnt; 380 (*cfunc)(); 381 if (need_summary) { 382 summary(); 383 } 384 } 385 } 386 387 /* 388 * Clean up any remaining I/O and flush output. If necessary, the output file 389 * is truncated. 390 */ 391 static void 392 dd_close(void) 393 { 394 if (cfunc == def) 395 def_close(); 396 else if (cfunc == block) 397 block_close(); 398 else if (cfunc == unblock) 399 unblock_close(); 400 if (ddflags & C_OSYNC && out.dbcnt && out.dbcnt < out.dbsz) { 401 if (ddflags & C_FILL) 402 memset(out.dbp, fill_char, out.dbsz - out.dbcnt); 403 else if (ddflags & (C_BLOCK | C_UNBLOCK)) 404 memset(out.dbp, ' ', out.dbsz - out.dbcnt); 405 else 406 memset(out.dbp, 0, out.dbsz - out.dbcnt); 407 out.dbcnt = out.dbsz; 408 } 409 if (out.dbcnt || pending) 410 dd_out(1); 411 412 /* 413 * If the file ends with a hole, ftruncate it to extend its size 414 * up to the end of the hole (without having to write any data). 415 */ 416 if (out.seek_offset > 0 && (out.flags & ISTRUNC)) { 417 if (ftruncate(out.fd, out.seek_offset) == -1) 418 err(1, "truncating %s", out.name); 419 } 420 } 421 422 void 423 dd_out(int force) 424 { 425 u_char *outp; 426 size_t cnt, i, n; 427 ssize_t nw; 428 static int warned; 429 int sparse; 430 431 /* 432 * Write one or more blocks out. The common case is writing a full 433 * output block in a single write; increment the full block stats. 434 * Otherwise, we're into partial block writes. If a partial write, 435 * and it's a character device, just warn. If a tape device, quit. 436 * 437 * The partial writes represent two cases. 1: Where the input block 438 * was less than expected so the output block was less than expected. 439 * 2: Where the input block was the right size but we were forced to 440 * write the block in multiple chunks. The original versions of dd(1) 441 * never wrote a block in more than a single write, so the latter case 442 * never happened. 443 * 444 * One special case is if we're forced to do the write -- in that case 445 * we play games with the buffer size, and it's usually a partial write. 446 */ 447 outp = out.db; 448 449 /* 450 * If force, first try to write all pending data, else try to write 451 * just one block. Subsequently always write data one full block at 452 * a time at most. 453 */ 454 for (n = force ? out.dbcnt : out.dbsz;; n = out.dbsz) { 455 cnt = n; 456 do { 457 sparse = 0; 458 if (ddflags & C_SPARSE) { 459 sparse = 1; /* Is buffer sparse? */ 460 for (i = 0; i < cnt; i++) 461 if (outp[i] != 0) { 462 sparse = 0; 463 break; 464 } 465 } 466 if (sparse && !force) { 467 pending += cnt; 468 nw = cnt; 469 } else { 470 if (pending != 0) { 471 /* 472 * Seek past hole. Note that we need to record the 473 * reached offset, because we might have no more data 474 * to write, in which case we'll need to call 475 * ftruncate to extend the file size. 476 */ 477 out.seek_offset = lseek(out.fd, pending, SEEK_CUR); 478 if (out.seek_offset == -1) 479 err(2, "%s: seek error creating sparse file", 480 out.name); 481 pending = 0; 482 } 483 if (cnt) { 484 nw = write(out.fd, outp, cnt); 485 out.seek_offset = 0; 486 } else { 487 return; 488 } 489 } 490 491 if (nw <= 0) { 492 if (nw == 0) 493 errx(1, "%s: end of device", out.name); 494 if (errno != EINTR) 495 err(1, "%s", out.name); 496 nw = 0; 497 } 498 499 outp += nw; 500 st.bytes += nw; 501 502 if ((size_t)nw == n && n == out.dbsz) 503 ++st.out_full; 504 else 505 ++st.out_part; 506 507 if ((size_t) nw != cnt) { 508 if (out.flags & ISTAPE) 509 errx(1, "%s: short write on tape device", 510 out.name); 511 if (out.flags & ISCHR && !warned) { 512 warned = 1; 513 warnx("%s: short write on character device", 514 out.name); 515 } 516 } 517 518 cnt -= nw; 519 } while (cnt != 0); 520 521 if ((out.dbcnt -= n) < out.dbsz) 522 break; 523 } 524 525 /* Reassemble the output block. */ 526 if (out.dbcnt) 527 (void)memmove(out.db, out.dbp - out.dbcnt, out.dbcnt); 528 out.dbp = out.db + out.dbcnt; 529 } 530