rsc | 031ddeb | 2005-07-22 11:47:57 +0000 | [diff] [blame^] | 1 | /* |
| 2 | * tar - `tape archiver', actually usable on any medium. |
| 3 | * POSIX "ustar" compliant when extracting, and by default when creating. |
| 4 | * this tar attempts to read and write multiple Tblock-byte blocks |
| 5 | * at once to and from the filesystem, and does not copy blocks |
| 6 | * around internally. |
| 7 | */ |
| 8 | |
| 9 | #include <u.h> |
| 10 | #include <libc.h> |
| 11 | #include <fcall.h> /* for %M */ |
| 12 | #include <String.h> |
| 13 | |
| 14 | /* |
| 15 | * modified versions of those in libc.h; scans only the first arg for |
| 16 | * keyletters and options. |
| 17 | */ |
| 18 | #define TARGBEGIN {\ |
| 19 | (argv0 || (argv0 = *argv)), argv++, argc--;\ |
| 20 | if (argv[0]) {\ |
| 21 | char *_args, *_argt;\ |
| 22 | Rune _argc;\ |
| 23 | _args = &argv[0][0];\ |
| 24 | _argc = 0;\ |
| 25 | while(*_args && (_args += chartorune(&_argc, _args)))\ |
| 26 | switch(_argc) |
| 27 | #define TARGEND SET(_argt); USED(_argt);USED(_argc);USED(_args); \ |
| 28 | argc--, argv++; } \ |
| 29 | USED(argv); USED(argc); } |
| 30 | #define TARGC() (_argc) |
| 31 | |
| 32 | #define ROUNDUP(a, b) (((a) + (b) - 1)/(b)) |
| 33 | #define BYTES2TBLKS(bytes) ROUNDUP(bytes, Tblock) |
| 34 | |
| 35 | typedef vlong Off; |
| 36 | typedef char *(*Refill)(int ar, char *bufs, int justhdr); |
| 37 | |
| 38 | enum { Stdin, Stdout, Stderr }; |
| 39 | enum { Rd, Wr }; /* pipe fd-array indices */ |
| 40 | enum { Output, Input }; |
| 41 | enum { None, Toc, Xtract, Replace }; |
| 42 | enum { Alldata, Justnxthdr }; |
| 43 | enum { |
| 44 | Tblock = 512, |
| 45 | Nblock = 40, /* maximum blocksize */ |
| 46 | Dblock = 20, /* default blocksize */ |
| 47 | Namsiz = 100, |
| 48 | Maxpfx = 155, /* from POSIX */ |
| 49 | Maxname = Namsiz + 1 + Maxpfx, |
| 50 | DEBUG = 0, |
| 51 | }; |
| 52 | |
| 53 | /* POSIX link flags */ |
| 54 | enum { |
| 55 | LF_PLAIN1 = '\0', |
| 56 | LF_PLAIN2 = '0', |
| 57 | LF_LINK = '1', |
| 58 | LF_SYMLINK1 = '2', |
| 59 | LF_SYMLINK2 = 's', |
| 60 | LF_CHR = '3', |
| 61 | LF_BLK = '4', |
| 62 | LF_DIR = '5', |
| 63 | LF_FIFO = '6', |
| 64 | LF_CONTIG = '7', |
| 65 | /* 'A' - 'Z' are reserved for custom implementations */ |
| 66 | }; |
| 67 | |
| 68 | #define islink(lf) (isreallink(lf) || issymlink(lf)) |
| 69 | #define isreallink(lf) ((lf) == LF_LINK) |
| 70 | #define issymlink(lf) ((lf) == LF_SYMLINK1 || (lf) == LF_SYMLINK2) |
| 71 | |
| 72 | typedef union { |
| 73 | uchar data[Tblock]; |
| 74 | struct { |
| 75 | char name[Namsiz]; |
| 76 | char mode[8]; |
| 77 | char uid[8]; |
| 78 | char gid[8]; |
| 79 | char size[12]; |
| 80 | char mtime[12]; |
| 81 | char chksum[8]; |
| 82 | char linkflag; |
| 83 | char linkname[Namsiz]; |
| 84 | |
| 85 | /* rest are defined by POSIX's ustar format; see p1003.2b */ |
| 86 | char magic[6]; /* "ustar" */ |
| 87 | char version[2]; |
| 88 | char uname[32]; |
| 89 | char gname[32]; |
| 90 | char devmajor[8]; |
| 91 | char devminor[8]; |
| 92 | char prefix[Maxpfx]; /* if non-null, path= prefix "/" name */ |
| 93 | }; |
| 94 | } Hdr; |
| 95 | |
| 96 | typedef struct { |
| 97 | char *comp; |
| 98 | char *decomp; |
| 99 | char *sfx[4]; |
| 100 | } Compress; |
| 101 | |
| 102 | static Compress comps[] = { |
| 103 | "gzip", "gunzip", { ".tar.gz", ".tgz" }, /* default */ |
| 104 | "compress", "uncompress", { ".tar.Z", ".tz" }, |
| 105 | "bzip2", "bunzip2", { ".tar.bz", ".tbz", |
| 106 | ".tar.bz2",".tbz2" }, |
| 107 | }; |
| 108 | |
| 109 | typedef struct { |
| 110 | int kid; |
| 111 | int fd; /* original fd */ |
| 112 | int rfd; /* replacement fd */ |
| 113 | int input; |
| 114 | int open; |
| 115 | } Pushstate; |
| 116 | |
| 117 | #define OTHER(rdwr) (rdwr == Rd? Wr: Rd) |
| 118 | |
| 119 | static int debug; |
| 120 | static int verb; |
| 121 | static int posix = 1; |
| 122 | static int docreate; |
| 123 | static int aruid; |
| 124 | static int argid; |
| 125 | static int relative = 1; |
| 126 | static int settime; |
| 127 | static int verbose; |
| 128 | static int docompress; |
| 129 | static int keepexisting; |
| 130 | static Off blkoff; /* offset of the current archive block (not Tblock) */ |
| 131 | static Off nexthdr; |
| 132 | |
| 133 | static int nblock = Dblock; |
| 134 | static char *usefile; |
| 135 | static char origdir[Maxname*2]; |
| 136 | static Hdr *tpblk, *endblk; |
| 137 | static Hdr *curblk; |
| 138 | |
| 139 | static void |
| 140 | usage(void) |
| 141 | { |
| 142 | fprint(2, "usage: %s {crtx}[PRTfgkmpuvz] [archive] file1 file2...\n", |
| 143 | argv0); |
| 144 | exits("usage"); |
| 145 | } |
| 146 | |
| 147 | /* compression */ |
| 148 | |
| 149 | static Compress * |
| 150 | compmethod(char *name) |
| 151 | { |
| 152 | int i, nmlen = strlen(name), sfxlen; |
| 153 | Compress *cp; |
| 154 | |
| 155 | for (cp = comps; cp < comps + nelem(comps); cp++) |
| 156 | for (i = 0; i < nelem(cp->sfx) && cp->sfx[i]; i++) { |
| 157 | sfxlen = strlen(cp->sfx[i]); |
| 158 | if (nmlen > sfxlen && |
| 159 | strcmp(cp->sfx[i], name + nmlen - sfxlen) == 0) |
| 160 | return cp; |
| 161 | } |
| 162 | return docompress? comps: nil; |
| 163 | } |
| 164 | |
| 165 | /* |
| 166 | * push a filter, cmd, onto fd. if input, it's an input descriptor. |
| 167 | * returns a descriptor to replace fd, or -1 on error. |
| 168 | */ |
| 169 | static int |
| 170 | push(int fd, char *cmd, int input, Pushstate *ps) |
| 171 | { |
| 172 | int nfd, pifds[2]; |
| 173 | String *s; |
| 174 | |
| 175 | ps->open = 0; |
| 176 | ps->fd = fd; |
| 177 | ps->input = input; |
| 178 | if (fd < 0 || pipe(pifds) < 0) |
| 179 | return -1; |
| 180 | ps->kid = fork(); |
| 181 | switch (ps->kid) { |
| 182 | case -1: |
| 183 | return -1; |
| 184 | case 0: |
| 185 | if (input) |
| 186 | dup(pifds[Wr], Stdout); |
| 187 | else |
| 188 | dup(pifds[Rd], Stdin); |
| 189 | close(pifds[input? Rd: Wr]); |
| 190 | dup(fd, (input? Stdin: Stdout)); |
| 191 | s = s_new(); |
| 192 | if (cmd[0] != '/') |
| 193 | s_append(s, "/bin/"); |
| 194 | s_append(s, cmd); |
| 195 | execl(s_to_c(s), cmd, nil); |
| 196 | sysfatal("can't exec %s: %r", cmd); |
| 197 | default: |
| 198 | nfd = pifds[input? Rd: Wr]; |
| 199 | close(pifds[input? Wr: Rd]); |
| 200 | break; |
| 201 | } |
| 202 | ps->rfd = nfd; |
| 203 | ps->open = 1; |
| 204 | return nfd; |
| 205 | } |
| 206 | |
| 207 | static char * |
| 208 | pushclose(Pushstate *ps) |
| 209 | { |
| 210 | Waitmsg *wm; |
| 211 | |
| 212 | if (ps->fd < 0 || ps->rfd < 0 || !ps->open) |
| 213 | return "not open"; |
| 214 | close(ps->rfd); |
| 215 | ps->rfd = -1; |
| 216 | ps->open = 0; |
| 217 | while ((wm = wait()) != nil && wm->pid != ps->kid) |
| 218 | continue; |
| 219 | return wm? wm->msg: nil; |
| 220 | } |
| 221 | |
| 222 | /* |
| 223 | * block-buffer management |
| 224 | */ |
| 225 | |
| 226 | static void |
| 227 | initblks(void) |
| 228 | { |
| 229 | free(tpblk); |
| 230 | tpblk = malloc(Tblock * nblock); |
| 231 | assert(tpblk != nil); |
| 232 | endblk = tpblk + nblock; |
| 233 | } |
| 234 | |
| 235 | /* |
| 236 | * (re)fill block buffers from archive. `justhdr' means we don't care |
| 237 | * about the data before the next header block. |
| 238 | */ |
| 239 | static char * |
| 240 | refill(int ar, char *bufs, int justhdr) |
| 241 | { |
| 242 | int i, n; |
| 243 | unsigned bytes = Tblock * nblock; |
| 244 | static int done, first = 1, seekable; |
| 245 | |
| 246 | if (done) |
| 247 | return nil; |
| 248 | |
| 249 | if (first) |
| 250 | seekable = seek(ar, 0, 1) >= 0; |
| 251 | /* try to size non-pipe input at first read */ |
| 252 | if (first && usefile) { |
| 253 | blkoff = seek(ar, 0, 1); /* note position */ |
| 254 | n = read(ar, bufs, bytes); |
| 255 | if (n <= 0) |
| 256 | sysfatal("error reading archive: %r"); |
| 257 | i = n; |
| 258 | if (i % Tblock != 0) { |
| 259 | fprint(2, "%s: archive block size (%d) error\n", |
| 260 | argv0, i); |
| 261 | exits("blocksize"); |
| 262 | } |
| 263 | i /= Tblock; |
| 264 | if (i != nblock) { |
| 265 | nblock = i; |
| 266 | fprint(2, "%s: blocking = %d\n", argv0, nblock); |
| 267 | endblk = (Hdr *)bufs + nblock; |
| 268 | bytes = n; |
| 269 | } |
| 270 | } else if (justhdr && seekable && nexthdr - seek(ar, 0, 1) >= bytes) { |
| 271 | /* optimisation for huge archive members on seekable media */ |
| 272 | if (seek(ar, bytes, 1) < 0) |
| 273 | sysfatal("can't seek on archive: %r"); |
| 274 | n = bytes; |
| 275 | } else |
| 276 | n = readn(ar, bufs, bytes); |
| 277 | first = 0; |
| 278 | |
| 279 | if (n == 0) |
| 280 | sysfatal("unexpected EOF reading archive"); |
| 281 | else if (n < 0) |
| 282 | sysfatal("error reading archive: %r"); |
| 283 | else if (n%Tblock != 0) |
| 284 | sysfatal("partial block read from archive"); |
| 285 | if (n != bytes) { |
| 286 | done = 1; |
| 287 | memset(bufs + n, 0, bytes - n); |
| 288 | } |
| 289 | return bufs; |
| 290 | } |
| 291 | |
| 292 | static Hdr * |
| 293 | getblk(int ar, Refill rfp, int justhdr) |
| 294 | { |
| 295 | if (curblk == nil || curblk >= endblk) { /* input block exhausted? */ |
| 296 | if (rfp != nil && (*rfp)(ar, (char *)tpblk, justhdr) == nil) |
| 297 | return nil; |
| 298 | curblk = tpblk; |
| 299 | } |
| 300 | return curblk++; |
| 301 | } |
| 302 | |
| 303 | static Hdr * |
| 304 | getblkrd(int ar, int justhdr) |
| 305 | { |
| 306 | return getblk(ar, refill, justhdr); |
| 307 | } |
| 308 | |
| 309 | static Hdr * |
| 310 | getblke(int ar) |
| 311 | { |
| 312 | return getblk(ar, nil, Alldata); |
| 313 | } |
| 314 | |
| 315 | static Hdr * |
| 316 | getblkz(int ar) |
| 317 | { |
| 318 | Hdr *hp = getblke(ar); |
| 319 | |
| 320 | if (hp != nil) |
| 321 | memset(hp->data, 0, Tblock); |
| 322 | return hp; |
| 323 | } |
| 324 | |
| 325 | /* |
| 326 | * how many block buffers are available, starting at the address |
| 327 | * just returned by getblk*? |
| 328 | */ |
| 329 | static int |
| 330 | gothowmany(int max) |
| 331 | { |
| 332 | int n = endblk - (curblk - 1); |
| 333 | |
| 334 | return n > max? max: n; |
| 335 | } |
| 336 | |
| 337 | /* |
| 338 | * indicate that one is done with the last block obtained from getblke |
| 339 | * and it is now available to be written into the archive. |
| 340 | */ |
| 341 | static void |
| 342 | putlastblk(int ar) |
| 343 | { |
| 344 | unsigned bytes = Tblock * nblock; |
| 345 | |
| 346 | /* if writing end-of-archive, aid compression (good hygiene too) */ |
| 347 | if (curblk < endblk) |
| 348 | memset(curblk, 0, (char *)endblk - (char *)curblk); |
| 349 | if (write(ar, tpblk, bytes) != bytes) |
| 350 | sysfatal("error writing archive: %r"); |
| 351 | } |
| 352 | |
| 353 | static void |
| 354 | putblk(int ar) |
| 355 | { |
| 356 | if (curblk >= endblk) |
| 357 | putlastblk(ar); |
| 358 | } |
| 359 | |
| 360 | static void |
| 361 | putbackblk(int ar) |
| 362 | { |
| 363 | curblk--; |
| 364 | USED(ar); |
| 365 | } |
| 366 | |
| 367 | static void |
| 368 | putreadblks(int ar, int blks) |
| 369 | { |
| 370 | curblk += blks - 1; |
| 371 | USED(ar); |
| 372 | } |
| 373 | |
| 374 | static void |
| 375 | putblkmany(int ar, int blks) |
| 376 | { |
| 377 | curblk += blks - 1; |
| 378 | putblk(ar); |
| 379 | } |
| 380 | |
| 381 | /* |
| 382 | * common routines |
| 383 | */ |
| 384 | |
| 385 | /* |
| 386 | * modifies hp->chksum but restores it; important for the last block of the |
| 387 | * old archive when updating with `tar rf archive' |
| 388 | */ |
| 389 | long |
| 390 | chksum(Hdr *hp) |
| 391 | { |
| 392 | int n = Tblock; |
| 393 | long i = 0; |
| 394 | uchar *cp = hp->data; |
| 395 | char oldsum[sizeof hp->chksum]; |
| 396 | |
| 397 | memmove(oldsum, hp->chksum, sizeof oldsum); |
| 398 | memset(hp->chksum, ' ', sizeof hp->chksum); |
| 399 | while (n-- > 0) |
| 400 | i += *cp++; |
| 401 | memmove(hp->chksum, oldsum, sizeof oldsum); |
| 402 | return i; |
| 403 | } |
| 404 | |
| 405 | static int |
| 406 | isustar(Hdr *hp) |
| 407 | { |
| 408 | return strcmp(hp->magic, "ustar") == 0; |
| 409 | } |
| 410 | |
| 411 | /* |
| 412 | * s is at most n bytes long, but need not be NUL-terminated. |
| 413 | * if shorter than n bytes, all bytes after the first NUL must also |
| 414 | * be NUL. |
| 415 | */ |
| 416 | static int |
| 417 | strnlen(char *s, int n) |
| 418 | { |
| 419 | return s[n - 1] != '\0'? n: strlen(s); |
| 420 | } |
| 421 | |
| 422 | /* set fullname from header */ |
| 423 | static char * |
| 424 | name(Hdr *hp) |
| 425 | { |
| 426 | int pfxlen, namlen; |
| 427 | static char fullnamebuf[2 + Maxname + 1]; /* 2 at beginning for ./ on relative names */ |
| 428 | char *fullname; |
| 429 | |
| 430 | fullname = fullnamebuf+2; |
| 431 | namlen = strnlen(hp->name, sizeof hp->name); |
| 432 | if (hp->prefix[0] == '\0' || !isustar(hp)) { /* old-style name? */ |
| 433 | memmove(fullname, hp->name, namlen); |
| 434 | fullname[namlen] = '\0'; |
| 435 | return fullname; |
| 436 | } |
| 437 | |
| 438 | /* name is in two pieces */ |
| 439 | pfxlen = strnlen(hp->prefix, sizeof hp->prefix); |
| 440 | memmove(fullname, hp->prefix, pfxlen); |
| 441 | fullname[pfxlen] = '/'; |
| 442 | memmove(fullname + pfxlen + 1, hp->name, namlen); |
| 443 | fullname[pfxlen + 1 + namlen] = '\0'; |
| 444 | return fullname; |
| 445 | } |
| 446 | |
| 447 | static int |
| 448 | isdir(Hdr *hp) |
| 449 | { |
| 450 | /* the mode test is ugly but sometimes necessary */ |
| 451 | return hp->linkflag == LF_DIR || |
| 452 | strrchr(name(hp), '\0')[-1] == '/' || |
| 453 | (strtoul(hp->mode, nil, 8)&0170000) == 040000; |
| 454 | } |
| 455 | |
| 456 | static int |
| 457 | eotar(Hdr *hp) |
| 458 | { |
| 459 | return name(hp)[0] == '\0'; |
| 460 | } |
| 461 | |
| 462 | Off |
| 463 | hdrsize(Hdr *hp) |
| 464 | { |
| 465 | Off bytes = strtoull(hp->size, nil, 8); |
| 466 | |
| 467 | if(isdir(hp)) |
| 468 | bytes = 0; |
| 469 | return bytes; |
| 470 | } |
| 471 | |
| 472 | static Hdr * |
| 473 | readhdr(int ar) |
| 474 | { |
| 475 | long hdrcksum; |
| 476 | Hdr *hp; |
| 477 | |
| 478 | hp = getblkrd(ar, Alldata); |
| 479 | if (hp == nil) |
| 480 | sysfatal("unexpected EOF instead of archive header"); |
| 481 | if (eotar(hp)) /* end-of-archive block? */ |
| 482 | return nil; |
| 483 | hdrcksum = strtoul(hp->chksum, nil, 8); |
| 484 | if (chksum(hp) != hdrcksum) |
| 485 | sysfatal("bad archive header checksum: name %.64s...", |
| 486 | hp->name); |
| 487 | nexthdr += Tblock*(1 + BYTES2TBLKS(hdrsize(hp))); |
| 488 | return hp; |
| 489 | } |
| 490 | |
| 491 | /* |
| 492 | * tar r[c] |
| 493 | */ |
| 494 | |
| 495 | /* |
| 496 | * if name is longer than Namsiz bytes, try to split it at a slash and fit the |
| 497 | * pieces into hp->prefix and hp->name. |
| 498 | */ |
| 499 | static int |
| 500 | putfullname(Hdr *hp, char *name) |
| 501 | { |
| 502 | int namlen, pfxlen; |
| 503 | char *sl, *osl; |
| 504 | String *slname = nil; |
| 505 | |
| 506 | if (isdir(hp)) { |
| 507 | slname = s_new(); |
| 508 | s_append(slname, name); |
| 509 | s_append(slname, "/"); /* posix requires this */ |
| 510 | name = s_to_c(slname); |
| 511 | } |
| 512 | |
| 513 | namlen = strlen(name); |
| 514 | if (namlen <= Namsiz) { |
| 515 | strncpy(hp->name, name, Namsiz); |
| 516 | hp->prefix[0] = '\0'; /* ustar paranoia */ |
| 517 | return 0; |
| 518 | } |
| 519 | |
| 520 | if (!posix || namlen > Maxname) { |
| 521 | fprint(2, "%s: name too long for tar header: %s\n", |
| 522 | argv0, name); |
| 523 | return -1; |
| 524 | } |
| 525 | /* |
| 526 | * try various splits until one results in pieces that fit into the |
| 527 | * appropriate fields of the header. look for slashes from right |
| 528 | * to left, in the hopes of putting the largest part of the name into |
| 529 | * hp->prefix, which is larger than hp->name. |
| 530 | */ |
| 531 | sl = strrchr(name, '/'); |
| 532 | while (sl != nil) { |
| 533 | pfxlen = sl - name; |
| 534 | if (pfxlen <= sizeof hp->prefix && namlen-1 - pfxlen <= Namsiz) |
| 535 | break; |
| 536 | osl = sl; |
| 537 | *osl = '\0'; |
| 538 | sl = strrchr(name, '/'); |
| 539 | *osl = '/'; |
| 540 | } |
| 541 | if (sl == nil) { |
| 542 | fprint(2, "%s: name can't be split to fit tar header: %s\n", |
| 543 | argv0, name); |
| 544 | return -1; |
| 545 | } |
| 546 | *sl = '\0'; |
| 547 | strncpy(hp->prefix, name, sizeof hp->prefix); |
| 548 | *sl++ = '/'; |
| 549 | strncpy(hp->name, sl, sizeof hp->name); |
| 550 | if (slname) |
| 551 | s_free(slname); |
| 552 | return 0; |
| 553 | } |
| 554 | |
| 555 | static int |
| 556 | mkhdr(Hdr *hp, Dir *dir, char *file) |
| 557 | { |
| 558 | /* |
| 559 | * these fields run together, so we format them in order and don't use |
| 560 | * snprint. |
| 561 | */ |
| 562 | sprint(hp->mode, "%6lo ", dir->mode & 0777); |
| 563 | sprint(hp->uid, "%6o ", aruid); |
| 564 | sprint(hp->gid, "%6o ", argid); |
| 565 | /* |
| 566 | * files > 2 bytes can't be described |
| 567 | * (unless we resort to xustar or exustar formats). |
| 568 | */ |
| 569 | if (dir->length >= (Off)1<<33) { |
| 570 | fprint(2, "%s: %s: too large for tar header format\n", |
| 571 | argv0, file); |
| 572 | return -1; |
| 573 | } |
| 574 | sprint(hp->size, "%11lluo ", dir->length); |
| 575 | sprint(hp->mtime, "%11luo ", dir->mtime); |
| 576 | hp->linkflag = (dir->mode&DMDIR? LF_DIR: LF_PLAIN1); |
| 577 | putfullname(hp, file); |
| 578 | if (posix) { |
| 579 | strncpy(hp->magic, "ustar", sizeof hp->magic); |
| 580 | strncpy(hp->version, "00", sizeof hp->version); |
| 581 | strncpy(hp->uname, dir->uid, sizeof hp->uname); |
| 582 | strncpy(hp->gname, dir->gid, sizeof hp->gname); |
| 583 | } |
| 584 | sprint(hp->chksum, "%6luo", chksum(hp)); |
| 585 | return 0; |
| 586 | } |
| 587 | |
| 588 | static void addtoar(int ar, char *file, char *shortf); |
| 589 | |
| 590 | static void |
| 591 | addtreetoar(int ar, char *file, char *shortf, int fd) |
| 592 | { |
| 593 | int n; |
| 594 | Dir *dent, *dirents; |
| 595 | String *name = s_new(); |
| 596 | |
| 597 | n = dirreadall(fd, &dirents); |
| 598 | close(fd); |
| 599 | if (n == 0) |
| 600 | return; |
| 601 | |
| 602 | if (chdir(shortf) < 0) |
| 603 | sysfatal("chdir %s: %r", file); |
| 604 | if (DEBUG) |
| 605 | fprint(2, "chdir %s\t# %s\n", shortf, file); |
| 606 | |
| 607 | for (dent = dirents; dent < dirents + n; dent++) { |
| 608 | s_reset(name); |
| 609 | s_append(name, file); |
| 610 | s_append(name, "/"); |
| 611 | s_append(name, dent->name); |
| 612 | addtoar(ar, s_to_c(name), dent->name); |
| 613 | } |
| 614 | s_free(name); |
| 615 | free(dirents); |
| 616 | |
| 617 | /* |
| 618 | * this assumes that shortf is just one component, which is true |
| 619 | * during directory descent, but not necessarily true of command-line |
| 620 | * arguments. Our caller (or addtoar's) must reset the working |
| 621 | * directory if necessary. |
| 622 | */ |
| 623 | if (chdir("..") < 0) |
| 624 | sysfatal("chdir %s/..: %r", file); |
| 625 | if (DEBUG) |
| 626 | fprint(2, "chdir ..\n"); |
| 627 | } |
| 628 | |
| 629 | static void |
| 630 | addtoar(int ar, char *file, char *shortf) |
| 631 | { |
| 632 | int n, fd, isdir; |
| 633 | long bytes; |
| 634 | ulong blksleft, blksread; |
| 635 | Hdr *hbp; |
| 636 | Dir *dir; |
| 637 | |
| 638 | fd = open(shortf, OREAD); |
| 639 | if (fd < 0) { |
| 640 | fprint(2, "%s: can't open %s: %r\n", argv0, file); |
| 641 | return; |
| 642 | } |
| 643 | dir = dirfstat(fd); |
| 644 | if (dir == nil) |
| 645 | sysfatal("can't fstat %s: %r", file); |
| 646 | |
| 647 | hbp = getblkz(ar); |
| 648 | isdir = !!(dir->qid.type&QTDIR); |
| 649 | if (mkhdr(hbp, dir, file) < 0) { |
| 650 | putbackblk(ar); |
| 651 | free(dir); |
| 652 | close(fd); |
| 653 | return; |
| 654 | } |
| 655 | putblk(ar); |
| 656 | |
| 657 | blksleft = BYTES2TBLKS(dir->length); |
| 658 | free(dir); |
| 659 | |
| 660 | if (isdir) |
| 661 | addtreetoar(ar, file, shortf, fd); |
| 662 | else { |
| 663 | for (; blksleft > 0; blksleft -= blksread) { |
| 664 | hbp = getblke(ar); |
| 665 | blksread = gothowmany(blksleft); |
| 666 | bytes = blksread * Tblock; |
| 667 | n = readn(fd, hbp->data, bytes); |
| 668 | if (n < 0) |
| 669 | sysfatal("error reading %s: %r", file); |
| 670 | /* |
| 671 | * ignore EOF. zero any partial block to aid |
| 672 | * compression and emergency recovery of data. |
| 673 | */ |
| 674 | if (n < Tblock) |
| 675 | memset(hbp->data + n, 0, bytes - n); |
| 676 | putblkmany(ar, blksread); |
| 677 | } |
| 678 | close(fd); |
| 679 | if (verbose) |
| 680 | fprint(2, "%s\n", file); |
| 681 | } |
| 682 | } |
| 683 | |
| 684 | static char * |
| 685 | replace(char **argv) |
| 686 | { |
| 687 | int i, ar; |
| 688 | ulong blksleft, blksread; |
| 689 | Off bytes; |
| 690 | Hdr *hp; |
| 691 | Compress *comp = nil; |
| 692 | Pushstate ps; |
| 693 | |
| 694 | if (usefile && docreate) { |
| 695 | ar = create(usefile, OWRITE, 0666); |
| 696 | if (docompress) |
| 697 | comp = compmethod(usefile); |
| 698 | } else if (usefile) |
| 699 | ar = open(usefile, ORDWR); |
| 700 | else |
| 701 | ar = Stdout; |
| 702 | if (comp) |
| 703 | ar = push(ar, comp->comp, Output, &ps); |
| 704 | if (ar < 0) |
| 705 | sysfatal("can't open archive %s: %r", usefile); |
| 706 | |
| 707 | if (usefile && !docreate) { |
| 708 | /* skip quickly to the end */ |
| 709 | while ((hp = readhdr(ar)) != nil) { |
| 710 | bytes = hdrsize(hp); |
| 711 | for (blksleft = BYTES2TBLKS(bytes); |
| 712 | blksleft > 0 && getblkrd(ar, Justnxthdr) != nil; |
| 713 | blksleft -= blksread) { |
| 714 | blksread = gothowmany(blksleft); |
| 715 | putreadblks(ar, blksread); |
| 716 | } |
| 717 | } |
| 718 | /* |
| 719 | * we have just read the end-of-archive Tblock. |
| 720 | * now seek back over the (big) archive block containing it, |
| 721 | * and back up curblk ptr over end-of-archive Tblock in memory. |
| 722 | */ |
| 723 | if (seek(ar, blkoff, 0) < 0) |
| 724 | sysfatal("can't seek back over end-of-archive: %r"); |
| 725 | curblk--; |
| 726 | } |
| 727 | |
| 728 | for (i = 0; argv[i] != nil; i++) { |
| 729 | addtoar(ar, argv[i], argv[i]); |
| 730 | chdir(origdir); /* for correctness & profiling */ |
| 731 | } |
| 732 | |
| 733 | /* write end-of-archive marker */ |
| 734 | getblkz(ar); |
| 735 | putblk(ar); |
| 736 | getblkz(ar); |
| 737 | putlastblk(ar); |
| 738 | |
| 739 | if (comp) |
| 740 | return pushclose(&ps); |
| 741 | if (ar > Stderr) |
| 742 | close(ar); |
| 743 | return nil; |
| 744 | } |
| 745 | |
| 746 | /* |
| 747 | * tar [xt] |
| 748 | */ |
| 749 | |
| 750 | /* is pfx a file-name prefix of name? */ |
| 751 | static int |
| 752 | prefix(char *name, char *pfx) |
| 753 | { |
| 754 | int pfxlen = strlen(pfx); |
| 755 | char clpfx[Maxname+1]; |
| 756 | |
| 757 | if (pfxlen > Maxname) |
| 758 | return 0; |
| 759 | strcpy(clpfx, pfx); |
| 760 | cleanname(clpfx); |
| 761 | return strncmp(pfx, name, pfxlen) == 0 && |
| 762 | (name[pfxlen] == '\0' || name[pfxlen] == '/'); |
| 763 | } |
| 764 | |
| 765 | static int |
| 766 | match(char *name, char **argv) |
| 767 | { |
| 768 | int i; |
| 769 | char clname[Maxname+1]; |
| 770 | |
| 771 | if (argv[0] == nil) |
| 772 | return 1; |
| 773 | strcpy(clname, name); |
| 774 | cleanname(clname); |
| 775 | for (i = 0; argv[i] != nil; i++) |
| 776 | if (prefix(clname, argv[i])) |
| 777 | return 1; |
| 778 | return 0; |
| 779 | } |
| 780 | |
| 781 | static int |
| 782 | makedir(char *s) |
| 783 | { |
| 784 | int f; |
| 785 | |
| 786 | if (access(s, AEXIST) == 0) |
| 787 | return -1; |
| 788 | f = create(s, OREAD, DMDIR | 0777); |
| 789 | if (f >= 0) |
| 790 | close(f); |
| 791 | return f; |
| 792 | } |
| 793 | |
| 794 | static void |
| 795 | mkpdirs(char *s) |
| 796 | { |
| 797 | int done = 0; |
| 798 | char *p = s; |
| 799 | |
| 800 | while (!done && (p = strchr(p + 1, '/')) != nil) { |
| 801 | *p = '\0'; |
| 802 | done = (access(s, AEXIST) < 0 && makedir(s) < 0); |
| 803 | *p = '/'; |
| 804 | } |
| 805 | } |
| 806 | |
| 807 | /* copy a file from the archive into the filesystem */ |
| 808 | /* fname is result of name(), so has two extra bytes at beginning */ |
| 809 | static void |
| 810 | extract1(int ar, Hdr *hp, char *fname) |
| 811 | { |
| 812 | int wrbytes, fd = -1, dir = 0; |
| 813 | long mtime = strtol(hp->mtime, nil, 8); |
| 814 | ulong mode = strtoul(hp->mode, nil, 8) & 0777; |
| 815 | Off bytes = strtoll(hp->size, nil, 8); /* for printing */ |
| 816 | ulong blksread, blksleft = BYTES2TBLKS(hdrsize(hp)); |
| 817 | Hdr *hbp; |
| 818 | |
| 819 | if (isdir(hp)) { |
| 820 | mode |= DMDIR|0700; |
| 821 | dir = 1; |
| 822 | } |
| 823 | switch (hp->linkflag) { |
| 824 | case LF_LINK: |
| 825 | case LF_SYMLINK1: |
| 826 | case LF_SYMLINK2: |
| 827 | case LF_FIFO: |
| 828 | blksleft = 0; |
| 829 | break; |
| 830 | } |
| 831 | if (relative) { |
| 832 | if(fname[0] == '/') |
| 833 | *--fname = '.'; |
| 834 | else if(fname[0] == '#'){ |
| 835 | *--fname = '/'; |
| 836 | *--fname = '.'; |
| 837 | } |
| 838 | } |
| 839 | if (verb == Xtract) { |
| 840 | cleanname(fname); |
| 841 | switch (hp->linkflag) { |
| 842 | case LF_LINK: |
| 843 | case LF_SYMLINK1: |
| 844 | case LF_SYMLINK2: |
| 845 | fprint(2, "%s: can't make (sym)link %s\n", |
| 846 | argv0, fname); |
| 847 | break; |
| 848 | case LF_FIFO: |
| 849 | fprint(2, "%s: can't make fifo %s\n", argv0, fname); |
| 850 | break; |
| 851 | default: |
| 852 | if (!keepexisting || access(fname, AEXIST) < 0) { |
| 853 | int rw = (dir? OREAD: OWRITE); |
| 854 | |
| 855 | fd = create(fname, rw, mode); |
| 856 | if (fd < 0) { |
| 857 | mkpdirs(fname); |
| 858 | fd = create(fname, rw, mode); |
| 859 | } |
| 860 | if (fd < 0 && |
| 861 | (!dir || access(fname, AEXIST) < 0)) |
| 862 | fprint(2, "%s: can't create %s: %r\n", |
| 863 | argv0, fname); |
| 864 | } |
| 865 | if (fd >= 0 && verbose) |
| 866 | fprint(2, "%s\n", fname); |
| 867 | break; |
| 868 | } |
| 869 | } else if (verbose) { |
| 870 | char *cp = ctime(mtime); |
| 871 | |
| 872 | print("%M %8lld %-12.12s %-4.4s %s\n", |
| 873 | mode, bytes, cp+4, cp+24, fname); |
| 874 | } else |
| 875 | print("%s\n", fname); |
| 876 | |
| 877 | for (; blksleft > 0; blksleft -= blksread) { |
| 878 | hbp = getblkrd(ar, (fd >= 0? Alldata: Justnxthdr)); |
| 879 | if (hbp == nil) |
| 880 | sysfatal("unexpected EOF on archive extracting %s", |
| 881 | fname); |
| 882 | blksread = gothowmany(blksleft); |
| 883 | wrbytes = Tblock*blksread; |
| 884 | if(wrbytes > bytes) |
| 885 | wrbytes = bytes; |
| 886 | if (fd >= 0 && write(fd, hbp->data, wrbytes) != wrbytes) |
| 887 | sysfatal("write error on %s: %r", fname); |
| 888 | putreadblks(ar, blksread); |
| 889 | bytes -= wrbytes; |
| 890 | } |
| 891 | if (fd >= 0) { |
| 892 | /* |
| 893 | * directories should be wstated after we're done |
| 894 | * creating files in them. |
| 895 | */ |
| 896 | if (settime) { |
| 897 | Dir nd; |
| 898 | |
| 899 | nulldir(&nd); |
| 900 | nd.mtime = mtime; |
| 901 | if (isustar(hp)) |
| 902 | nd.gid = hp->gname; |
| 903 | dirfwstat(fd, &nd); |
| 904 | } |
| 905 | close(fd); |
| 906 | } |
| 907 | } |
| 908 | |
| 909 | static void |
| 910 | skip(int ar, Hdr *hp, char *fname) |
| 911 | { |
| 912 | ulong blksleft, blksread; |
| 913 | Hdr *hbp; |
| 914 | |
| 915 | for (blksleft = BYTES2TBLKS(hdrsize(hp)); blksleft > 0; |
| 916 | blksleft -= blksread) { |
| 917 | hbp = getblkrd(ar, Justnxthdr); |
| 918 | if (hbp == nil) |
| 919 | sysfatal("unexpected EOF on archive extracting %s", |
| 920 | fname); |
| 921 | blksread = gothowmany(blksleft); |
| 922 | putreadblks(ar, blksread); |
| 923 | } |
| 924 | } |
| 925 | |
| 926 | static char * |
| 927 | extract(char **argv) |
| 928 | { |
| 929 | int ar; |
| 930 | char *longname; |
| 931 | Hdr *hp; |
| 932 | Compress *comp = nil; |
| 933 | Pushstate ps; |
| 934 | |
| 935 | if (usefile) { |
| 936 | ar = open(usefile, OREAD); |
| 937 | comp = compmethod(usefile); |
| 938 | } else |
| 939 | ar = Stdin; |
| 940 | if (comp) |
| 941 | ar = push(ar, comp->decomp, Input, &ps); |
| 942 | if (ar < 0) |
| 943 | sysfatal("can't open archive %s: %r", usefile); |
| 944 | |
| 945 | while ((hp = readhdr(ar)) != nil) { |
| 946 | longname = name(hp); |
| 947 | if (match(longname, argv)) |
| 948 | extract1(ar, hp, longname); |
| 949 | else |
| 950 | skip(ar, hp, longname); |
| 951 | } |
| 952 | |
| 953 | if (comp) |
| 954 | return pushclose(&ps); |
| 955 | if (ar > Stderr) |
| 956 | close(ar); |
| 957 | return nil; |
| 958 | } |
| 959 | |
| 960 | void |
| 961 | main(int argc, char *argv[]) |
| 962 | { |
| 963 | int errflg = 0; |
| 964 | char *ret = nil; |
| 965 | |
| 966 | quotefmtinstall(); |
| 967 | fmtinstall('M', dirmodefmt); |
| 968 | |
| 969 | TARGBEGIN { |
| 970 | case 'c': |
| 971 | docreate++; |
| 972 | verb = Replace; |
| 973 | break; |
| 974 | case 'f': |
| 975 | usefile = EARGF(usage()); |
| 976 | break; |
| 977 | case 'g': |
| 978 | argid = strtoul(EARGF(usage()), 0, 0); |
| 979 | break; |
| 980 | case 'k': |
| 981 | keepexisting++; |
| 982 | break; |
| 983 | case 'm': /* compatibility */ |
| 984 | settime = 0; |
| 985 | break; |
| 986 | case 'p': |
| 987 | posix++; |
| 988 | break; |
| 989 | case 'P': |
| 990 | posix = 0; |
| 991 | break; |
| 992 | case 'r': |
| 993 | verb = Replace; |
| 994 | break; |
| 995 | case 'R': |
| 996 | relative = 0; |
| 997 | break; |
| 998 | case 't': |
| 999 | verb = Toc; |
| 1000 | break; |
| 1001 | case 'T': |
| 1002 | settime++; |
| 1003 | break; |
| 1004 | case 'u': |
| 1005 | aruid = strtoul(EARGF(usage()), 0, 0); |
| 1006 | break; |
| 1007 | case 'v': |
| 1008 | verbose++; |
| 1009 | break; |
| 1010 | case 'x': |
| 1011 | verb = Xtract; |
| 1012 | break; |
| 1013 | case 'z': |
| 1014 | docompress++; |
| 1015 | break; |
| 1016 | case '-': |
| 1017 | break; |
| 1018 | default: |
| 1019 | fprint(2, "tar: unknown letter %C\n", TARGC()); |
| 1020 | errflg++; |
| 1021 | break; |
| 1022 | } TARGEND |
| 1023 | |
| 1024 | if (argc < 0 || errflg) |
| 1025 | usage(); |
| 1026 | |
| 1027 | initblks(); |
| 1028 | switch (verb) { |
| 1029 | case Toc: |
| 1030 | case Xtract: |
| 1031 | ret = extract(argv); |
| 1032 | break; |
| 1033 | case Replace: |
| 1034 | if (getwd(origdir, sizeof origdir) == nil) |
| 1035 | strcpy(origdir, "/tmp"); |
| 1036 | ret = replace(argv); |
| 1037 | break; |
| 1038 | default: |
| 1039 | usage(); |
| 1040 | break; |
| 1041 | } |
| 1042 | exits(ret); |
| 1043 | } |