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