Annotation of embedaddon/rsync/util.c, revision 1.1
1.1 ! misho 1: /*
! 2: * Utility routines used in rsync.
! 3: *
! 4: * Copyright (C) 1996-2000 Andrew Tridgell
! 5: * Copyright (C) 1996 Paul Mackerras
! 6: * Copyright (C) 2001, 2002 Martin Pool <mbp@samba.org>
! 7: * Copyright (C) 2003-2009 Wayne Davison
! 8: *
! 9: * This program is free software; you can redistribute it and/or modify
! 10: * it under the terms of the GNU General Public License as published by
! 11: * the Free Software Foundation; either version 3 of the License, or
! 12: * (at your option) any later version.
! 13: *
! 14: * This program is distributed in the hope that it will be useful,
! 15: * but WITHOUT ANY WARRANTY; without even the implied warranty of
! 16: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
! 17: * GNU General Public License for more details.
! 18: *
! 19: * You should have received a copy of the GNU General Public License along
! 20: * with this program; if not, visit the http://fsf.org website.
! 21: */
! 22:
! 23: #include "rsync.h"
! 24: #include "ifuncs.h"
! 25:
! 26: extern int verbose;
! 27: extern int module_id;
! 28: extern int modify_window;
! 29: extern int relative_paths;
! 30: extern int preserve_times;
! 31: extern int human_readable;
! 32: extern int preserve_xattrs;
! 33: extern char *module_dir;
! 34: extern unsigned int module_dirlen;
! 35: extern mode_t orig_umask;
! 36: extern char *partial_dir;
! 37: extern struct filter_list_struct daemon_filter_list;
! 38:
! 39: int sanitize_paths = 0;
! 40:
! 41: char curr_dir[MAXPATHLEN];
! 42: unsigned int curr_dir_len;
! 43: int curr_dir_depth; /* This is only set for a sanitizing daemon. */
! 44:
! 45: /* Set a fd into nonblocking mode. */
! 46: void set_nonblocking(int fd)
! 47: {
! 48: int val;
! 49:
! 50: if ((val = fcntl(fd, F_GETFL)) == -1)
! 51: return;
! 52: if (!(val & NONBLOCK_FLAG)) {
! 53: val |= NONBLOCK_FLAG;
! 54: fcntl(fd, F_SETFL, val);
! 55: }
! 56: }
! 57:
! 58: /* Set a fd into blocking mode. */
! 59: void set_blocking(int fd)
! 60: {
! 61: int val;
! 62:
! 63: if ((val = fcntl(fd, F_GETFL)) == -1)
! 64: return;
! 65: if (val & NONBLOCK_FLAG) {
! 66: val &= ~NONBLOCK_FLAG;
! 67: fcntl(fd, F_SETFL, val);
! 68: }
! 69: }
! 70:
! 71: /**
! 72: * Create a file descriptor pair - like pipe() but use socketpair if
! 73: * possible (because of blocking issues on pipes).
! 74: *
! 75: * Always set non-blocking.
! 76: */
! 77: int fd_pair(int fd[2])
! 78: {
! 79: int ret;
! 80:
! 81: #ifdef HAVE_SOCKETPAIR
! 82: ret = socketpair(AF_UNIX, SOCK_STREAM, 0, fd);
! 83: #else
! 84: ret = pipe(fd);
! 85: #endif
! 86:
! 87: if (ret == 0) {
! 88: set_nonblocking(fd[0]);
! 89: set_nonblocking(fd[1]);
! 90: }
! 91:
! 92: return ret;
! 93: }
! 94:
! 95: void print_child_argv(const char *prefix, char **cmd)
! 96: {
! 97: rprintf(FCLIENT, "%s ", prefix);
! 98: for (; *cmd; cmd++) {
! 99: /* Look for characters that ought to be quoted. This
! 100: * is not a great quoting algorithm, but it's
! 101: * sufficient for a log message. */
! 102: if (strspn(*cmd, "abcdefghijklmnopqrstuvwxyz"
! 103: "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
! 104: "0123456789"
! 105: ",.-_=+@/") != strlen(*cmd)) {
! 106: rprintf(FCLIENT, "\"%s\" ", *cmd);
! 107: } else {
! 108: rprintf(FCLIENT, "%s ", *cmd);
! 109: }
! 110: }
! 111: rprintf(FCLIENT, "\n");
! 112: }
! 113:
! 114: NORETURN void out_of_memory(const char *str)
! 115: {
! 116: rprintf(FERROR, "ERROR: out of memory in %s [%s]\n", str, who_am_i());
! 117: exit_cleanup(RERR_MALLOC);
! 118: }
! 119:
! 120: NORETURN void overflow_exit(const char *str)
! 121: {
! 122: rprintf(FERROR, "ERROR: buffer overflow in %s [%s]\n", str, who_am_i());
! 123: exit_cleanup(RERR_MALLOC);
! 124: }
! 125:
! 126: /* This returns 0 for success, 1 for a symlink if symlink time-setting
! 127: * is not possible, or -1 for any other error. */
! 128: int set_modtime(const char *fname, time_t modtime, mode_t mode)
! 129: {
! 130: static int switch_step = 0;
! 131:
! 132: if (verbose > 2) {
! 133: rprintf(FINFO, "set modtime of %s to (%ld) %s",
! 134: fname, (long)modtime,
! 135: asctime(localtime(&modtime)));
! 136: }
! 137:
! 138: switch (switch_step) {
! 139: #ifdef HAVE_UTIMENSAT
! 140: #include "case_N.h"
! 141: if (do_utimensat(fname, modtime, 0) == 0)
! 142: break;
! 143: if (errno != ENOSYS)
! 144: return -1;
! 145: switch_step++;
! 146: /* FALLTHROUGH */
! 147: #endif
! 148:
! 149: #ifdef HAVE_LUTIMES
! 150: #include "case_N.h"
! 151: if (do_lutimes(fname, modtime, 0) == 0)
! 152: break;
! 153: if (errno != ENOSYS)
! 154: return -1;
! 155: switch_step++;
! 156: /* FALLTHROUGH */
! 157: #endif
! 158:
! 159: #include "case_N.h"
! 160: switch_step++;
! 161: if (preserve_times & PRESERVE_LINK_TIMES) {
! 162: preserve_times &= ~PRESERVE_LINK_TIMES;
! 163: if (S_ISLNK(mode))
! 164: return 1;
! 165: }
! 166: /* FALLTHROUGH */
! 167:
! 168: #include "case_N.h"
! 169: #ifdef HAVE_UTIMES
! 170: if (do_utimes(fname, modtime, 0) == 0)
! 171: break;
! 172: #else
! 173: if (do_utime(fname, modtime, 0) == 0)
! 174: break;
! 175: #endif
! 176:
! 177: return -1;
! 178: }
! 179:
! 180: return 0;
! 181: }
! 182:
! 183: /* This creates a new directory with default permissions. Since there
! 184: * might be some directory-default permissions affecting this, we can't
! 185: * force the permissions directly using the original umask and mkdir(). */
! 186: int mkdir_defmode(char *fname)
! 187: {
! 188: int ret;
! 189:
! 190: umask(orig_umask);
! 191: ret = do_mkdir(fname, ACCESSPERMS);
! 192: umask(0);
! 193:
! 194: return ret;
! 195: }
! 196:
! 197: /* Create any necessary directories in fname. Any missing directories are
! 198: * created with default permissions. */
! 199: int create_directory_path(char *fname)
! 200: {
! 201: char *p;
! 202: int ret = 0;
! 203:
! 204: while (*fname == '/')
! 205: fname++;
! 206: while (strncmp(fname, "./", 2) == 0)
! 207: fname += 2;
! 208:
! 209: umask(orig_umask);
! 210: p = fname;
! 211: while ((p = strchr(p,'/')) != NULL) {
! 212: *p = '\0';
! 213: if (do_mkdir(fname, ACCESSPERMS) < 0 && errno != EEXIST)
! 214: ret = -1;
! 215: *p++ = '/';
! 216: }
! 217: umask(0);
! 218:
! 219: return ret;
! 220: }
! 221:
! 222: /**
! 223: * Write @p len bytes at @p ptr to descriptor @p desc, retrying if
! 224: * interrupted.
! 225: *
! 226: * @retval len upon success
! 227: *
! 228: * @retval <0 write's (negative) error code
! 229: *
! 230: * Derived from GNU C's cccp.c.
! 231: */
! 232: int full_write(int desc, const char *ptr, size_t len)
! 233: {
! 234: int total_written;
! 235:
! 236: total_written = 0;
! 237: while (len > 0) {
! 238: int written = write(desc, ptr, len);
! 239: if (written < 0) {
! 240: if (errno == EINTR)
! 241: continue;
! 242: return written;
! 243: }
! 244: total_written += written;
! 245: ptr += written;
! 246: len -= written;
! 247: }
! 248: return total_written;
! 249: }
! 250:
! 251: /**
! 252: * Read @p len bytes at @p ptr from descriptor @p desc, retrying if
! 253: * interrupted.
! 254: *
! 255: * @retval >0 the actual number of bytes read
! 256: *
! 257: * @retval 0 for EOF
! 258: *
! 259: * @retval <0 for an error.
! 260: *
! 261: * Derived from GNU C's cccp.c. */
! 262: static int safe_read(int desc, char *ptr, size_t len)
! 263: {
! 264: int n_chars;
! 265:
! 266: if (len == 0)
! 267: return len;
! 268:
! 269: do {
! 270: n_chars = read(desc, ptr, len);
! 271: } while (n_chars < 0 && errno == EINTR);
! 272:
! 273: return n_chars;
! 274: }
! 275:
! 276: /* Copy a file. If ofd < 0, copy_file unlinks and opens the "dest" file.
! 277: * Otherwise, it just writes to and closes the provided file descriptor.
! 278: * In either case, if --xattrs are being preserved, the dest file will
! 279: * have its xattrs set from the source file.
! 280: *
! 281: * This is used in conjunction with the --temp-dir, --backup, and
! 282: * --copy-dest options. */
! 283: int copy_file(const char *source, const char *dest, int ofd,
! 284: mode_t mode, int create_bak_dir)
! 285: {
! 286: int ifd;
! 287: char buf[1024 * 8];
! 288: int len; /* Number of bytes read into `buf'. */
! 289:
! 290: if ((ifd = do_open(source, O_RDONLY, 0)) < 0) {
! 291: int save_errno = errno;
! 292: rsyserr(FERROR_XFER, errno, "open %s", full_fname(source));
! 293: errno = save_errno;
! 294: return -1;
! 295: }
! 296:
! 297: if (ofd < 0) {
! 298: if (robust_unlink(dest) && errno != ENOENT) {
! 299: int save_errno = errno;
! 300: rsyserr(FERROR_XFER, errno, "unlink %s", full_fname(dest));
! 301: errno = save_errno;
! 302: return -1;
! 303: }
! 304:
! 305: if ((ofd = do_open(dest, O_WRONLY | O_CREAT | O_TRUNC | O_EXCL, mode)) < 0) {
! 306: int save_errno = errno ? errno : EINVAL; /* 0 paranoia */
! 307: if (create_bak_dir && errno == ENOENT && make_bak_dir(dest) == 0) {
! 308: if ((ofd = do_open(dest, O_WRONLY | O_CREAT | O_TRUNC | O_EXCL, mode)) < 0)
! 309: save_errno = errno ? errno : save_errno;
! 310: else
! 311: save_errno = 0;
! 312: }
! 313: if (save_errno) {
! 314: rsyserr(FERROR_XFER, save_errno, "open %s", full_fname(dest));
! 315: close(ifd);
! 316: errno = save_errno;
! 317: return -1;
! 318: }
! 319: }
! 320: }
! 321:
! 322: while ((len = safe_read(ifd, buf, sizeof buf)) > 0) {
! 323: if (full_write(ofd, buf, len) < 0) {
! 324: int save_errno = errno;
! 325: rsyserr(FERROR_XFER, errno, "write %s", full_fname(dest));
! 326: close(ifd);
! 327: close(ofd);
! 328: errno = save_errno;
! 329: return -1;
! 330: }
! 331: }
! 332:
! 333: if (len < 0) {
! 334: int save_errno = errno;
! 335: rsyserr(FERROR_XFER, errno, "read %s", full_fname(source));
! 336: close(ifd);
! 337: close(ofd);
! 338: errno = save_errno;
! 339: return -1;
! 340: }
! 341:
! 342: if (close(ifd) < 0) {
! 343: rsyserr(FWARNING, errno, "close failed on %s",
! 344: full_fname(source));
! 345: }
! 346:
! 347: if (close(ofd) < 0) {
! 348: int save_errno = errno;
! 349: rsyserr(FERROR_XFER, errno, "close failed on %s",
! 350: full_fname(dest));
! 351: errno = save_errno;
! 352: return -1;
! 353: }
! 354:
! 355: #ifdef SUPPORT_XATTRS
! 356: if (preserve_xattrs)
! 357: copy_xattrs(source, dest);
! 358: #endif
! 359:
! 360: return 0;
! 361: }
! 362:
! 363: /* MAX_RENAMES should be 10**MAX_RENAMES_DIGITS */
! 364: #define MAX_RENAMES_DIGITS 3
! 365: #define MAX_RENAMES 1000
! 366:
! 367: /**
! 368: * Robust unlink: some OS'es (HPUX) refuse to unlink busy files, so
! 369: * rename to <path>/.rsyncNNN instead.
! 370: *
! 371: * Note that successive rsync runs will shuffle the filenames around a
! 372: * bit as long as the file is still busy; this is because this function
! 373: * does not know if the unlink call is due to a new file coming in, or
! 374: * --delete trying to remove old .rsyncNNN files, hence it renames it
! 375: * each time.
! 376: **/
! 377: int robust_unlink(const char *fname)
! 378: {
! 379: #ifndef ETXTBSY
! 380: return do_unlink(fname);
! 381: #else
! 382: static int counter = 1;
! 383: int rc, pos, start;
! 384: char path[MAXPATHLEN];
! 385:
! 386: rc = do_unlink(fname);
! 387: if (rc == 0 || errno != ETXTBSY)
! 388: return rc;
! 389:
! 390: if ((pos = strlcpy(path, fname, MAXPATHLEN)) >= MAXPATHLEN)
! 391: pos = MAXPATHLEN - 1;
! 392:
! 393: while (pos > 0 && path[pos-1] != '/')
! 394: pos--;
! 395: pos += strlcpy(path+pos, ".rsync", MAXPATHLEN-pos);
! 396:
! 397: if (pos > (MAXPATHLEN-MAX_RENAMES_DIGITS-1)) {
! 398: errno = ETXTBSY;
! 399: return -1;
! 400: }
! 401:
! 402: /* start where the last one left off to reduce chance of clashes */
! 403: start = counter;
! 404: do {
! 405: snprintf(&path[pos], MAX_RENAMES_DIGITS+1, "%03d", counter);
! 406: if (++counter >= MAX_RENAMES)
! 407: counter = 1;
! 408: } while ((rc = access(path, 0)) == 0 && counter != start);
! 409:
! 410: if (verbose > 0) {
! 411: rprintf(FWARNING, "renaming %s to %s because of text busy\n",
! 412: fname, path);
! 413: }
! 414:
! 415: /* maybe we should return rename()'s exit status? Nah. */
! 416: if (do_rename(fname, path) != 0) {
! 417: errno = ETXTBSY;
! 418: return -1;
! 419: }
! 420: return 0;
! 421: #endif
! 422: }
! 423:
! 424: /* Returns 0 on successful rename, 1 if we successfully copied the file
! 425: * across filesystems, -2 if copy_file() failed, and -1 on other errors.
! 426: * If partialptr is not NULL and we need to do a copy, copy the file into
! 427: * the active partial-dir instead of over the destination file. */
! 428: int robust_rename(const char *from, const char *to, const char *partialptr,
! 429: int mode)
! 430: {
! 431: int tries = 4;
! 432:
! 433: while (tries--) {
! 434: if (do_rename(from, to) == 0)
! 435: return 0;
! 436:
! 437: switch (errno) {
! 438: #ifdef ETXTBSY
! 439: case ETXTBSY:
! 440: if (robust_unlink(to) != 0) {
! 441: errno = ETXTBSY;
! 442: return -1;
! 443: }
! 444: errno = ETXTBSY;
! 445: break;
! 446: #endif
! 447: case EXDEV:
! 448: if (partialptr) {
! 449: if (!handle_partial_dir(partialptr,PDIR_CREATE))
! 450: return -2;
! 451: to = partialptr;
! 452: }
! 453: if (copy_file(from, to, -1, mode, 0) != 0)
! 454: return -2;
! 455: do_unlink(from);
! 456: return 1;
! 457: default:
! 458: return -1;
! 459: }
! 460: }
! 461: return -1;
! 462: }
! 463:
! 464: static pid_t all_pids[10];
! 465: static int num_pids;
! 466:
! 467: /** Fork and record the pid of the child. **/
! 468: pid_t do_fork(void)
! 469: {
! 470: pid_t newpid = fork();
! 471:
! 472: if (newpid != 0 && newpid != -1) {
! 473: all_pids[num_pids++] = newpid;
! 474: }
! 475: return newpid;
! 476: }
! 477:
! 478: /**
! 479: * Kill all children.
! 480: *
! 481: * @todo It would be kind of nice to make sure that they are actually
! 482: * all our children before we kill them, because their pids may have
! 483: * been recycled by some other process. Perhaps when we wait for a
! 484: * child, we should remove it from this array. Alternatively we could
! 485: * perhaps use process groups, but I think that would not work on
! 486: * ancient Unix versions that don't support them.
! 487: **/
! 488: void kill_all(int sig)
! 489: {
! 490: int i;
! 491:
! 492: for (i = 0; i < num_pids; i++) {
! 493: /* Let's just be a little careful where we
! 494: * point that gun, hey? See kill(2) for the
! 495: * magic caused by negative values. */
! 496: pid_t p = all_pids[i];
! 497:
! 498: if (p == getpid())
! 499: continue;
! 500: if (p <= 0)
! 501: continue;
! 502:
! 503: kill(p, sig);
! 504: }
! 505: }
! 506:
! 507: /** Turn a user name into a uid */
! 508: int name_to_uid(const char *name, uid_t *uid_p)
! 509: {
! 510: struct passwd *pass;
! 511: if (!name || !*name)
! 512: return 0;
! 513: if (!(pass = getpwnam(name)))
! 514: return 0;
! 515: *uid_p = pass->pw_uid;
! 516: return 1;
! 517: }
! 518:
! 519: /** Turn a group name into a gid */
! 520: int name_to_gid(const char *name, gid_t *gid_p)
! 521: {
! 522: struct group *grp;
! 523: if (!name || !*name)
! 524: return 0;
! 525: if (!(grp = getgrnam(name)))
! 526: return 0;
! 527: *gid_p = grp->gr_gid;
! 528: return 1;
! 529: }
! 530:
! 531: /** Lock a byte range in a open file */
! 532: int lock_range(int fd, int offset, int len)
! 533: {
! 534: struct flock lock;
! 535:
! 536: lock.l_type = F_WRLCK;
! 537: lock.l_whence = SEEK_SET;
! 538: lock.l_start = offset;
! 539: lock.l_len = len;
! 540: lock.l_pid = 0;
! 541:
! 542: return fcntl(fd,F_SETLK,&lock) == 0;
! 543: }
! 544:
! 545: #define ENSURE_MEMSPACE(buf, type, sz, req) \
! 546: if ((req) > sz && !(buf = realloc_array(buf, type, sz = MAX(sz * 2, req)))) \
! 547: out_of_memory("glob_expand")
! 548:
! 549: static inline void call_glob_match(const char *name, int len, int from_glob,
! 550: char *arg, int abpos, int fbpos);
! 551:
! 552: static struct glob_data {
! 553: char *arg_buf, *filt_buf, **argv;
! 554: int absize, fbsize, maxargs, argc;
! 555: } glob;
! 556:
! 557: static void glob_match(char *arg, int abpos, int fbpos)
! 558: {
! 559: int len;
! 560: char *slash;
! 561:
! 562: while (*arg == '.' && arg[1] == '/') {
! 563: if (fbpos < 0) {
! 564: ENSURE_MEMSPACE(glob.filt_buf, char, glob.fbsize, glob.absize);
! 565: memcpy(glob.filt_buf, glob.arg_buf, abpos + 1);
! 566: fbpos = abpos;
! 567: }
! 568: ENSURE_MEMSPACE(glob.arg_buf, char, glob.absize, abpos + 3);
! 569: glob.arg_buf[abpos++] = *arg++;
! 570: glob.arg_buf[abpos++] = *arg++;
! 571: glob.arg_buf[abpos] = '\0';
! 572: }
! 573: if ((slash = strchr(arg, '/')) != NULL) {
! 574: *slash = '\0';
! 575: len = slash - arg;
! 576: } else
! 577: len = strlen(arg);
! 578: if (strpbrk(arg, "*?[")) {
! 579: struct dirent *di;
! 580: DIR *d;
! 581:
! 582: if (!(d = opendir(abpos ? glob.arg_buf : ".")))
! 583: return;
! 584: while ((di = readdir(d)) != NULL) {
! 585: char *dname = d_name(di);
! 586: if (dname[0] == '.' && (dname[1] == '\0'
! 587: || (dname[1] == '.' && dname[2] == '\0')))
! 588: continue;
! 589: if (!wildmatch(arg, dname))
! 590: continue;
! 591: call_glob_match(dname, strlen(dname), 1,
! 592: slash ? arg + len + 1 : NULL,
! 593: abpos, fbpos);
! 594: }
! 595: closedir(d);
! 596: } else {
! 597: call_glob_match(arg, len, 0,
! 598: slash ? arg + len + 1 : NULL,
! 599: abpos, fbpos);
! 600: }
! 601: if (slash)
! 602: *slash = '/';
! 603: }
! 604:
! 605: static inline void call_glob_match(const char *name, int len, int from_glob,
! 606: char *arg, int abpos, int fbpos)
! 607: {
! 608: char *use_buf;
! 609:
! 610: ENSURE_MEMSPACE(glob.arg_buf, char, glob.absize, abpos + len + 2);
! 611: memcpy(glob.arg_buf + abpos, name, len);
! 612: abpos += len;
! 613: glob.arg_buf[abpos] = '\0';
! 614:
! 615: if (fbpos >= 0) {
! 616: ENSURE_MEMSPACE(glob.filt_buf, char, glob.fbsize, fbpos + len + 2);
! 617: memcpy(glob.filt_buf + fbpos, name, len);
! 618: fbpos += len;
! 619: glob.filt_buf[fbpos] = '\0';
! 620: use_buf = glob.filt_buf;
! 621: } else
! 622: use_buf = glob.arg_buf;
! 623:
! 624: if (from_glob || (arg && len)) {
! 625: STRUCT_STAT st;
! 626: int is_dir;
! 627:
! 628: if (do_stat(glob.arg_buf, &st) != 0)
! 629: return;
! 630: is_dir = S_ISDIR(st.st_mode) != 0;
! 631: if (arg && !is_dir)
! 632: return;
! 633:
! 634: if (daemon_filter_list.head
! 635: && check_filter(&daemon_filter_list, FLOG, use_buf, is_dir) < 0)
! 636: return;
! 637: }
! 638:
! 639: if (arg) {
! 640: glob.arg_buf[abpos++] = '/';
! 641: glob.arg_buf[abpos] = '\0';
! 642: if (fbpos >= 0) {
! 643: glob.filt_buf[fbpos++] = '/';
! 644: glob.filt_buf[fbpos] = '\0';
! 645: }
! 646: glob_match(arg, abpos, fbpos);
! 647: } else {
! 648: ENSURE_MEMSPACE(glob.argv, char *, glob.maxargs, glob.argc + 1);
! 649: if (!(glob.argv[glob.argc++] = strdup(glob.arg_buf)))
! 650: out_of_memory("glob_match");
! 651: }
! 652: }
! 653:
! 654: /* This routine performs wild-card expansion of the pathname in "arg". Any
! 655: * daemon-excluded files/dirs will not be matched by the wildcards. Returns 0
! 656: * if a wild-card string is the only returned item (due to matching nothing). */
! 657: int glob_expand(const char *arg, char ***argv_p, int *argc_p, int *maxargs_p)
! 658: {
! 659: int ret, save_argc;
! 660: char *s;
! 661:
! 662: if (!arg) {
! 663: if (glob.filt_buf)
! 664: free(glob.filt_buf);
! 665: free(glob.arg_buf);
! 666: memset(&glob, 0, sizeof glob);
! 667: return -1;
! 668: }
! 669:
! 670: if (sanitize_paths)
! 671: s = sanitize_path(NULL, arg, "", 0, SP_KEEP_DOT_DIRS);
! 672: else {
! 673: s = strdup(arg);
! 674: if (!s)
! 675: out_of_memory("glob_expand");
! 676: clean_fname(s, CFN_KEEP_DOT_DIRS
! 677: | CFN_KEEP_TRAILING_SLASH
! 678: | CFN_COLLAPSE_DOT_DOT_DIRS);
! 679: }
! 680:
! 681: ENSURE_MEMSPACE(glob.arg_buf, char, glob.absize, MAXPATHLEN);
! 682: *glob.arg_buf = '\0';
! 683:
! 684: glob.argc = save_argc = *argc_p;
! 685: glob.argv = *argv_p;
! 686: glob.maxargs = *maxargs_p;
! 687:
! 688: ENSURE_MEMSPACE(glob.argv, char *, glob.maxargs, 100);
! 689:
! 690: glob_match(s, 0, -1);
! 691:
! 692: /* The arg didn't match anything, so add the failed arg to the list. */
! 693: if (glob.argc == save_argc) {
! 694: ENSURE_MEMSPACE(glob.argv, char *, glob.maxargs, glob.argc + 1);
! 695: glob.argv[glob.argc++] = s;
! 696: ret = 0;
! 697: } else {
! 698: free(s);
! 699: ret = 1;
! 700: }
! 701:
! 702: *maxargs_p = glob.maxargs;
! 703: *argv_p = glob.argv;
! 704: *argc_p = glob.argc;
! 705:
! 706: return ret;
! 707: }
! 708:
! 709: /* This routine is only used in daemon mode. */
! 710: void glob_expand_module(char *base1, char *arg, char ***argv_p, int *argc_p, int *maxargs_p)
! 711: {
! 712: char *p, *s;
! 713: char *base = base1;
! 714: int base_len = strlen(base);
! 715:
! 716: if (!arg || !*arg)
! 717: return;
! 718:
! 719: if (strncmp(arg, base, base_len) == 0)
! 720: arg += base_len;
! 721:
! 722: if (!(arg = strdup(arg)))
! 723: out_of_memory("glob_expand_module");
! 724:
! 725: if (asprintf(&base," %s/", base1) <= 0)
! 726: out_of_memory("glob_expand_module");
! 727: base_len++;
! 728:
! 729: for (s = arg; *s; s = p + base_len) {
! 730: if ((p = strstr(s, base)) != NULL)
! 731: *p = '\0'; /* split it at this point */
! 732: glob_expand(s, argv_p, argc_p, maxargs_p);
! 733: if (!p)
! 734: break;
! 735: }
! 736:
! 737: free(arg);
! 738: free(base);
! 739: }
! 740:
! 741: /**
! 742: * Convert a string to lower case
! 743: **/
! 744: void strlower(char *s)
! 745: {
! 746: while (*s) {
! 747: if (isUpper(s))
! 748: *s = toLower(s);
! 749: s++;
! 750: }
! 751: }
! 752:
! 753: /* Join strings p1 & p2 into "dest" with a guaranteed '/' between them. (If
! 754: * p1 ends with a '/', no extra '/' is inserted.) Returns the length of both
! 755: * strings + 1 (if '/' was inserted), regardless of whether the null-terminated
! 756: * string fits into destsize. */
! 757: size_t pathjoin(char *dest, size_t destsize, const char *p1, const char *p2)
! 758: {
! 759: size_t len = strlcpy(dest, p1, destsize);
! 760: if (len < destsize - 1) {
! 761: if (!len || dest[len-1] != '/')
! 762: dest[len++] = '/';
! 763: if (len < destsize - 1)
! 764: len += strlcpy(dest + len, p2, destsize - len);
! 765: else {
! 766: dest[len] = '\0';
! 767: len += strlen(p2);
! 768: }
! 769: }
! 770: else
! 771: len += strlen(p2) + 1; /* Assume we'd insert a '/'. */
! 772: return len;
! 773: }
! 774:
! 775: /* Join any number of strings together, putting them in "dest". The return
! 776: * value is the length of all the strings, regardless of whether the null-
! 777: * terminated whole fits in destsize. Your list of string pointers must end
! 778: * with a NULL to indicate the end of the list. */
! 779: size_t stringjoin(char *dest, size_t destsize, ...)
! 780: {
! 781: va_list ap;
! 782: size_t len, ret = 0;
! 783: const char *src;
! 784:
! 785: va_start(ap, destsize);
! 786: while (1) {
! 787: if (!(src = va_arg(ap, const char *)))
! 788: break;
! 789: len = strlen(src);
! 790: ret += len;
! 791: if (destsize > 1) {
! 792: if (len >= destsize)
! 793: len = destsize - 1;
! 794: memcpy(dest, src, len);
! 795: destsize -= len;
! 796: dest += len;
! 797: }
! 798: }
! 799: *dest = '\0';
! 800: va_end(ap);
! 801:
! 802: return ret;
! 803: }
! 804:
! 805: int count_dir_elements(const char *p)
! 806: {
! 807: int cnt = 0, new_component = 1;
! 808: while (*p) {
! 809: if (*p++ == '/')
! 810: new_component = (*p != '.' || (p[1] != '/' && p[1] != '\0'));
! 811: else if (new_component) {
! 812: new_component = 0;
! 813: cnt++;
! 814: }
! 815: }
! 816: return cnt;
! 817: }
! 818:
! 819: /* Turns multiple adjacent slashes into a single slash (possible exception:
! 820: * the preserving of two leading slashes at the start), drops all leading or
! 821: * interior "." elements unless CFN_KEEP_DOT_DIRS is flagged. Will also drop
! 822: * a trailing '.' after a '/' if CFN_DROP_TRAILING_DOT_DIR is flagged, removes
! 823: * a trailing slash (perhaps after removing the aforementioned dot) unless
! 824: * CFN_KEEP_TRAILING_SLASH is flagged, and will also collapse ".." elements
! 825: * (except at the start) if CFN_COLLAPSE_DOT_DOT_DIRS is flagged. If the
! 826: * resulting name would be empty, returns ".". */
! 827: unsigned int clean_fname(char *name, int flags)
! 828: {
! 829: char *limit = name - 1, *t = name, *f = name;
! 830: int anchored;
! 831:
! 832: if (!name)
! 833: return 0;
! 834:
! 835: if ((anchored = *f == '/') != 0) {
! 836: *t++ = *f++;
! 837: #ifdef __CYGWIN__
! 838: /* If there are exactly 2 slashes at the start, preserve
! 839: * them. Would break daemon excludes unless the paths are
! 840: * really treated differently, so used this sparingly. */
! 841: if (*f == '/' && f[1] != '/')
! 842: *t++ = *f++;
! 843: #endif
! 844: } else if (flags & CFN_KEEP_DOT_DIRS && *f == '.' && f[1] == '/') {
! 845: *t++ = *f++;
! 846: *t++ = *f++;
! 847: }
! 848: while (*f) {
! 849: /* discard extra slashes */
! 850: if (*f == '/') {
! 851: f++;
! 852: continue;
! 853: }
! 854: if (*f == '.') {
! 855: /* discard interior "." dirs */
! 856: if (f[1] == '/' && !(flags & CFN_KEEP_DOT_DIRS)) {
! 857: f += 2;
! 858: continue;
! 859: }
! 860: if (f[1] == '\0' && flags & CFN_DROP_TRAILING_DOT_DIR)
! 861: break;
! 862: /* collapse ".." dirs */
! 863: if (flags & CFN_COLLAPSE_DOT_DOT_DIRS
! 864: && f[1] == '.' && (f[2] == '/' || !f[2])) {
! 865: char *s = t - 1;
! 866: if (s == name && anchored) {
! 867: f += 2;
! 868: continue;
! 869: }
! 870: while (s > limit && *--s != '/') {}
! 871: if (s != t - 1 && (s < name || *s == '/')) {
! 872: t = s + 1;
! 873: f += 2;
! 874: continue;
! 875: }
! 876: limit = t + 2;
! 877: }
! 878: }
! 879: while (*f && (*t++ = *f++) != '/') {}
! 880: }
! 881:
! 882: if (t > name+anchored && t[-1] == '/' && !(flags & CFN_KEEP_TRAILING_SLASH))
! 883: t--;
! 884: if (t == name)
! 885: *t++ = '.';
! 886: *t = '\0';
! 887:
! 888: return t - name;
! 889: }
! 890:
! 891: /* Make path appear as if a chroot had occurred. This handles a leading
! 892: * "/" (either removing it or expanding it) and any leading or embedded
! 893: * ".." components that attempt to escape past the module's top dir.
! 894: *
! 895: * If dest is NULL, a buffer is allocated to hold the result. It is legal
! 896: * to call with the dest and the path (p) pointing to the same buffer, but
! 897: * rootdir will be ignored to avoid expansion of the string.
! 898: *
! 899: * The rootdir string contains a value to use in place of a leading slash.
! 900: * Specify NULL to get the default of "module_dir".
! 901: *
! 902: * The depth var is a count of how many '..'s to allow at the start of the
! 903: * path.
! 904: *
! 905: * We also clean the path in a manner similar to clean_fname() but with a
! 906: * few differences:
! 907: *
! 908: * Turns multiple adjacent slashes into a single slash, gets rid of "." dir
! 909: * elements (INCLUDING a trailing dot dir), PRESERVES a trailing slash, and
! 910: * ALWAYS collapses ".." elements (except for those at the start of the
! 911: * string up to "depth" deep). If the resulting name would be empty,
! 912: * change it into a ".". */
! 913: char *sanitize_path(char *dest, const char *p, const char *rootdir, int depth,
! 914: int flags)
! 915: {
! 916: char *start, *sanp;
! 917: int rlen = 0, drop_dot_dirs = !relative_paths || !(flags & SP_KEEP_DOT_DIRS);
! 918:
! 919: if (dest != p) {
! 920: int plen = strlen(p);
! 921: if (*p == '/') {
! 922: if (!rootdir)
! 923: rootdir = module_dir;
! 924: rlen = strlen(rootdir);
! 925: depth = 0;
! 926: p++;
! 927: }
! 928: if (dest) {
! 929: if (rlen + plen + 1 >= MAXPATHLEN)
! 930: return NULL;
! 931: } else if (!(dest = new_array(char, rlen + plen + 1)))
! 932: out_of_memory("sanitize_path");
! 933: if (rlen) {
! 934: memcpy(dest, rootdir, rlen);
! 935: if (rlen > 1)
! 936: dest[rlen++] = '/';
! 937: }
! 938: }
! 939:
! 940: if (drop_dot_dirs) {
! 941: while (*p == '.' && p[1] == '/')
! 942: p += 2;
! 943: }
! 944:
! 945: start = sanp = dest + rlen;
! 946: /* This loop iterates once per filename component in p, pointing at
! 947: * the start of the name (past any prior slash) for each iteration. */
! 948: while (*p) {
! 949: /* discard leading or extra slashes */
! 950: if (*p == '/') {
! 951: p++;
! 952: continue;
! 953: }
! 954: if (drop_dot_dirs) {
! 955: if (*p == '.' && (p[1] == '/' || p[1] == '\0')) {
! 956: /* skip "." component */
! 957: p++;
! 958: continue;
! 959: }
! 960: }
! 961: if (*p == '.' && p[1] == '.' && (p[2] == '/' || p[2] == '\0')) {
! 962: /* ".." component followed by slash or end */
! 963: if (depth <= 0 || sanp != start) {
! 964: p += 2;
! 965: if (sanp != start) {
! 966: /* back up sanp one level */
! 967: --sanp; /* now pointing at slash */
! 968: while (sanp > start && sanp[-1] != '/')
! 969: sanp--;
! 970: }
! 971: continue;
! 972: }
! 973: /* allow depth levels of .. at the beginning */
! 974: depth--;
! 975: /* move the virtual beginning to leave the .. alone */
! 976: start = sanp + 3;
! 977: }
! 978: /* copy one component through next slash */
! 979: while (*p && (*sanp++ = *p++) != '/') {}
! 980: }
! 981: if (sanp == dest) {
! 982: /* ended up with nothing, so put in "." component */
! 983: *sanp++ = '.';
! 984: }
! 985: *sanp = '\0';
! 986:
! 987: return dest;
! 988: }
! 989:
! 990: /* Like chdir(), but it keeps track of the current directory (in the
! 991: * global "curr_dir"), and ensures that the path size doesn't overflow.
! 992: * Also cleans the path using the clean_fname() function. */
! 993: int change_dir(const char *dir, int set_path_only)
! 994: {
! 995: static int initialised;
! 996: unsigned int len;
! 997:
! 998: if (!initialised) {
! 999: initialised = 1;
! 1000: if (getcwd(curr_dir, sizeof curr_dir - 1) == NULL) {
! 1001: rsyserr(FERROR, errno, "getcwd()");
! 1002: exit_cleanup(RERR_FILESELECT);
! 1003: }
! 1004: curr_dir_len = strlen(curr_dir);
! 1005: }
! 1006:
! 1007: if (!dir) /* this call was probably just to initialize */
! 1008: return 0;
! 1009:
! 1010: len = strlen(dir);
! 1011: if (len == 1 && *dir == '.')
! 1012: return 1;
! 1013:
! 1014: if (*dir == '/') {
! 1015: if (len >= sizeof curr_dir) {
! 1016: errno = ENAMETOOLONG;
! 1017: return 0;
! 1018: }
! 1019: if (!set_path_only && chdir(dir))
! 1020: return 0;
! 1021: memcpy(curr_dir, dir, len + 1);
! 1022: } else {
! 1023: if (curr_dir_len + 1 + len >= sizeof curr_dir) {
! 1024: errno = ENAMETOOLONG;
! 1025: return 0;
! 1026: }
! 1027: if (!(curr_dir_len && curr_dir[curr_dir_len-1] == '/'))
! 1028: curr_dir[curr_dir_len++] = '/';
! 1029: memcpy(curr_dir + curr_dir_len, dir, len + 1);
! 1030:
! 1031: if (!set_path_only && chdir(curr_dir)) {
! 1032: curr_dir[curr_dir_len] = '\0';
! 1033: return 0;
! 1034: }
! 1035: }
! 1036:
! 1037: curr_dir_len = clean_fname(curr_dir, CFN_COLLAPSE_DOT_DOT_DIRS);
! 1038: if (sanitize_paths) {
! 1039: if (module_dirlen > curr_dir_len)
! 1040: module_dirlen = curr_dir_len;
! 1041: curr_dir_depth = count_dir_elements(curr_dir + module_dirlen);
! 1042: }
! 1043:
! 1044: if (verbose >= 5 && !set_path_only)
! 1045: rprintf(FINFO, "[%s] change_dir(%s)\n", who_am_i(), curr_dir);
! 1046:
! 1047: return 1;
! 1048: }
! 1049:
! 1050: /* This will make a relative path absolute and clean it up via clean_fname().
! 1051: * Returns the string, which might be newly allocated, or NULL on error. */
! 1052: char *normalize_path(char *path, BOOL force_newbuf, unsigned int *len_ptr)
! 1053: {
! 1054: unsigned int len;
! 1055:
! 1056: if (*path != '/') { /* Make path absolute. */
! 1057: int len = strlen(path);
! 1058: if (curr_dir_len + 1 + len >= sizeof curr_dir)
! 1059: return NULL;
! 1060: curr_dir[curr_dir_len] = '/';
! 1061: memcpy(curr_dir + curr_dir_len + 1, path, len + 1);
! 1062: if (!(path = strdup(curr_dir)))
! 1063: out_of_memory("normalize_path");
! 1064: curr_dir[curr_dir_len] = '\0';
! 1065: } else if (force_newbuf) {
! 1066: if (!(path = strdup(path)))
! 1067: out_of_memory("normalize_path");
! 1068: }
! 1069:
! 1070: len = clean_fname(path, CFN_COLLAPSE_DOT_DOT_DIRS | CFN_DROP_TRAILING_DOT_DIR);
! 1071:
! 1072: if (len_ptr)
! 1073: *len_ptr = len;
! 1074:
! 1075: return path;
! 1076: }
! 1077:
! 1078: /**
! 1079: * Return a quoted string with the full pathname of the indicated filename.
! 1080: * The string " (in MODNAME)" may also be appended. The returned pointer
! 1081: * remains valid until the next time full_fname() is called.
! 1082: **/
! 1083: char *full_fname(const char *fn)
! 1084: {
! 1085: static char *result = NULL;
! 1086: char *m1, *m2, *m3;
! 1087: char *p1, *p2;
! 1088:
! 1089: if (result)
! 1090: free(result);
! 1091:
! 1092: if (*fn == '/')
! 1093: p1 = p2 = "";
! 1094: else {
! 1095: p1 = curr_dir + module_dirlen;
! 1096: for (p2 = p1; *p2 == '/'; p2++) {}
! 1097: if (*p2)
! 1098: p2 = "/";
! 1099: }
! 1100: if (module_id >= 0) {
! 1101: m1 = " (in ";
! 1102: m2 = lp_name(module_id);
! 1103: m3 = ")";
! 1104: } else
! 1105: m1 = m2 = m3 = "";
! 1106:
! 1107: if (asprintf(&result, "\"%s%s%s\"%s%s%s", p1, p2, fn, m1, m2, m3) <= 0)
! 1108: out_of_memory("full_fname");
! 1109:
! 1110: return result;
! 1111: }
! 1112:
! 1113: static char partial_fname[MAXPATHLEN];
! 1114:
! 1115: char *partial_dir_fname(const char *fname)
! 1116: {
! 1117: char *t = partial_fname;
! 1118: int sz = sizeof partial_fname;
! 1119: const char *fn;
! 1120:
! 1121: if ((fn = strrchr(fname, '/')) != NULL) {
! 1122: fn++;
! 1123: if (*partial_dir != '/') {
! 1124: int len = fn - fname;
! 1125: strncpy(t, fname, len); /* safe */
! 1126: t += len;
! 1127: sz -= len;
! 1128: }
! 1129: } else
! 1130: fn = fname;
! 1131: if ((int)pathjoin(t, sz, partial_dir, fn) >= sz)
! 1132: return NULL;
! 1133: if (daemon_filter_list.head) {
! 1134: t = strrchr(partial_fname, '/');
! 1135: *t = '\0';
! 1136: if (check_filter(&daemon_filter_list, FLOG, partial_fname, 1) < 0)
! 1137: return NULL;
! 1138: *t = '/';
! 1139: if (check_filter(&daemon_filter_list, FLOG, partial_fname, 0) < 0)
! 1140: return NULL;
! 1141: }
! 1142:
! 1143: return partial_fname;
! 1144: }
! 1145:
! 1146: /* If no --partial-dir option was specified, we don't need to do anything
! 1147: * (the partial-dir is essentially '.'), so just return success. */
! 1148: int handle_partial_dir(const char *fname, int create)
! 1149: {
! 1150: char *fn, *dir;
! 1151:
! 1152: if (fname != partial_fname)
! 1153: return 1;
! 1154: if (!create && *partial_dir == '/')
! 1155: return 1;
! 1156: if (!(fn = strrchr(partial_fname, '/')))
! 1157: return 1;
! 1158:
! 1159: *fn = '\0';
! 1160: dir = partial_fname;
! 1161: if (create) {
! 1162: STRUCT_STAT st;
! 1163: int statret = do_lstat(dir, &st);
! 1164: if (statret == 0 && !S_ISDIR(st.st_mode)) {
! 1165: if (do_unlink(dir) < 0) {
! 1166: *fn = '/';
! 1167: return 0;
! 1168: }
! 1169: statret = -1;
! 1170: }
! 1171: if (statret < 0 && do_mkdir(dir, 0700) < 0) {
! 1172: *fn = '/';
! 1173: return 0;
! 1174: }
! 1175: } else
! 1176: do_rmdir(dir);
! 1177: *fn = '/';
! 1178:
! 1179: return 1;
! 1180: }
! 1181:
! 1182: /* Determine if a symlink points outside the current directory tree.
! 1183: * This is considered "unsafe" because e.g. when mirroring somebody
! 1184: * else's machine it might allow them to establish a symlink to
! 1185: * /etc/passwd, and then read it through a web server.
! 1186: *
! 1187: * Returns 1 if unsafe, 0 if safe.
! 1188: *
! 1189: * Null symlinks and absolute symlinks are always unsafe.
! 1190: *
! 1191: * Basically here we are concerned with symlinks whose target contains
! 1192: * "..", because this might cause us to walk back up out of the
! 1193: * transferred directory. We are not allowed to go back up and
! 1194: * reenter.
! 1195: *
! 1196: * "dest" is the target of the symlink in question.
! 1197: *
! 1198: * "src" is the top source directory currently applicable at the level
! 1199: * of the referenced symlink. This is usually the symlink's full path
! 1200: * (including its name), as referenced from the root of the transfer. */
! 1201: int unsafe_symlink(const char *dest, const char *src)
! 1202: {
! 1203: const char *name, *slash;
! 1204: int depth = 0;
! 1205:
! 1206: /* all absolute and null symlinks are unsafe */
! 1207: if (!dest || !*dest || *dest == '/')
! 1208: return 1;
! 1209:
! 1210: /* find out what our safety margin is */
! 1211: for (name = src; (slash = strchr(name, '/')) != 0; name = slash+1) {
! 1212: /* ".." segment starts the count over. "." segment is ignored. */
! 1213: if (*name == '.' && (name[1] == '/' || (name[1] == '.' && name[2] == '/'))) {
! 1214: if (name[1] == '.')
! 1215: depth = 0;
! 1216: } else
! 1217: depth++;
! 1218: while (slash[1] == '/') slash++; /* just in case src isn't clean */
! 1219: }
! 1220: if (*name == '.' && name[1] == '.' && name[2] == '\0')
! 1221: depth = 0;
! 1222:
! 1223: for (name = dest; (slash = strchr(name, '/')) != 0; name = slash+1) {
! 1224: if (*name == '.' && (name[1] == '/' || (name[1] == '.' && name[2] == '/'))) {
! 1225: if (name[1] == '.') {
! 1226: /* if at any point we go outside the current directory
! 1227: then stop - it is unsafe */
! 1228: if (--depth < 0)
! 1229: return 1;
! 1230: }
! 1231: } else
! 1232: depth++;
! 1233: while (slash[1] == '/') slash++;
! 1234: }
! 1235: if (*name == '.' && name[1] == '.' && name[2] == '\0')
! 1236: depth--;
! 1237:
! 1238: return depth < 0;
! 1239: }
! 1240:
! 1241: #define HUMANIFY(mult) \
! 1242: do { \
! 1243: if (num >= mult || num <= -mult) { \
! 1244: double dnum = (double)num / mult; \
! 1245: char units; \
! 1246: if (num < 0) \
! 1247: dnum = -dnum; \
! 1248: if (dnum < mult) \
! 1249: units = 'K'; \
! 1250: else if ((dnum /= mult) < mult) \
! 1251: units = 'M'; \
! 1252: else { \
! 1253: dnum /= mult; \
! 1254: units = 'G'; \
! 1255: } \
! 1256: if (num < 0) \
! 1257: dnum = -dnum; \
! 1258: snprintf(bufs[n], sizeof bufs[0], "%.2f%c", dnum, units); \
! 1259: return bufs[n]; \
! 1260: } \
! 1261: } while (0)
! 1262:
! 1263: /* Return the int64 number as a string. If the --human-readable option was
! 1264: * specified, we may output the number in K, M, or G units. We can return
! 1265: * up to 4 buffers at a time. */
! 1266: char *human_num(int64 num)
! 1267: {
! 1268: static char bufs[4][128]; /* more than enough room */
! 1269: static unsigned int n;
! 1270: char *s;
! 1271: int negated;
! 1272:
! 1273: n = (n + 1) % (sizeof bufs / sizeof bufs[0]);
! 1274:
! 1275: if (human_readable) {
! 1276: if (human_readable == 1)
! 1277: HUMANIFY(1000);
! 1278: else
! 1279: HUMANIFY(1024);
! 1280: }
! 1281:
! 1282: s = bufs[n] + sizeof bufs[0] - 1;
! 1283: *s = '\0';
! 1284:
! 1285: if (!num)
! 1286: *--s = '0';
! 1287: if (num < 0) {
! 1288: /* A maximum-size negated number can't fit as a positive,
! 1289: * so do one digit in negated form to start us off. */
! 1290: *--s = (char)(-(num % 10)) + '0';
! 1291: num = -(num / 10);
! 1292: negated = 1;
! 1293: } else
! 1294: negated = 0;
! 1295:
! 1296: while (num) {
! 1297: *--s = (char)(num % 10) + '0';
! 1298: num /= 10;
! 1299: }
! 1300:
! 1301: if (negated)
! 1302: *--s = '-';
! 1303:
! 1304: return s;
! 1305: }
! 1306:
! 1307: /* Return the double number as a string. If the --human-readable option was
! 1308: * specified, we may output the number in K, M, or G units. We use a buffer
! 1309: * from human_num() to return our result. */
! 1310: char *human_dnum(double dnum, int decimal_digits)
! 1311: {
! 1312: char *buf = human_num(dnum);
! 1313: int len = strlen(buf);
! 1314: if (isDigit(buf + len - 1)) {
! 1315: /* There's extra room in buf prior to the start of the num. */
! 1316: buf -= decimal_digits + 2;
! 1317: snprintf(buf, len + decimal_digits + 3, "%.*f", decimal_digits, dnum);
! 1318: }
! 1319: return buf;
! 1320: }
! 1321:
! 1322: /* Return the date and time as a string. Some callers tweak returned buf. */
! 1323: char *timestring(time_t t)
! 1324: {
! 1325: static char TimeBuf[200];
! 1326: struct tm *tm = localtime(&t);
! 1327: char *p;
! 1328:
! 1329: #ifdef HAVE_STRFTIME
! 1330: strftime(TimeBuf, sizeof TimeBuf - 1, "%Y/%m/%d %H:%M:%S", tm);
! 1331: #else
! 1332: strlcpy(TimeBuf, asctime(tm), sizeof TimeBuf);
! 1333: #endif
! 1334:
! 1335: if ((p = strchr(TimeBuf, '\n')) != NULL)
! 1336: *p = '\0';
! 1337:
! 1338: return TimeBuf;
! 1339: }
! 1340:
! 1341: /**
! 1342: * Sleep for a specified number of milliseconds.
! 1343: *
! 1344: * Always returns TRUE. (In the future it might return FALSE if
! 1345: * interrupted.)
! 1346: **/
! 1347: int msleep(int t)
! 1348: {
! 1349: int tdiff = 0;
! 1350: struct timeval tval, t1, t2;
! 1351:
! 1352: gettimeofday(&t1, NULL);
! 1353:
! 1354: while (tdiff < t) {
! 1355: tval.tv_sec = (t-tdiff)/1000;
! 1356: tval.tv_usec = 1000*((t-tdiff)%1000);
! 1357:
! 1358: errno = 0;
! 1359: select(0,NULL,NULL, NULL, &tval);
! 1360:
! 1361: gettimeofday(&t2, NULL);
! 1362: tdiff = (t2.tv_sec - t1.tv_sec)*1000 +
! 1363: (t2.tv_usec - t1.tv_usec)/1000;
! 1364: }
! 1365:
! 1366: return True;
! 1367: }
! 1368:
! 1369: /* Determine if two time_t values are equivalent (either exact, or in
! 1370: * the modification timestamp window established by --modify-window).
! 1371: *
! 1372: * @retval 0 if the times should be treated as the same
! 1373: *
! 1374: * @retval +1 if the first is later
! 1375: *
! 1376: * @retval -1 if the 2nd is later
! 1377: **/
! 1378: int cmp_time(time_t file1, time_t file2)
! 1379: {
! 1380: if (file2 > file1) {
! 1381: if (file2 - file1 <= modify_window)
! 1382: return 0;
! 1383: return -1;
! 1384: }
! 1385: if (file1 - file2 <= modify_window)
! 1386: return 0;
! 1387: return 1;
! 1388: }
! 1389:
! 1390:
! 1391: #ifdef __INSURE__XX
! 1392: #include <dlfcn.h>
! 1393:
! 1394: /**
! 1395: This routine is a trick to immediately catch errors when debugging
! 1396: with insure. A xterm with a gdb is popped up when insure catches
! 1397: a error. It is Linux specific.
! 1398: **/
! 1399: int _Insure_trap_error(int a1, int a2, int a3, int a4, int a5, int a6)
! 1400: {
! 1401: static int (*fn)();
! 1402: int ret;
! 1403: char *cmd;
! 1404:
! 1405: asprintf(&cmd, "/usr/X11R6/bin/xterm -display :0 -T Panic -n Panic -e /bin/sh -c 'cat /tmp/ierrs.*.%d ; gdb /proc/%d/exe %d'",
! 1406: getpid(), getpid(), getpid());
! 1407:
! 1408: if (!fn) {
! 1409: static void *h;
! 1410: h = dlopen("/usr/local/parasoft/insure++lite/lib.linux2/libinsure.so", RTLD_LAZY);
! 1411: fn = dlsym(h, "_Insure_trap_error");
! 1412: }
! 1413:
! 1414: ret = fn(a1, a2, a3, a4, a5, a6);
! 1415:
! 1416: system(cmd);
! 1417:
! 1418: free(cmd);
! 1419:
! 1420: return ret;
! 1421: }
! 1422: #endif
! 1423:
! 1424: #define MALLOC_MAX 0x40000000
! 1425:
! 1426: void *_new_array(unsigned long num, unsigned int size, int use_calloc)
! 1427: {
! 1428: if (num >= MALLOC_MAX/size)
! 1429: return NULL;
! 1430: return use_calloc ? calloc(num, size) : malloc(num * size);
! 1431: }
! 1432:
! 1433: void *_realloc_array(void *ptr, unsigned int size, size_t num)
! 1434: {
! 1435: if (num >= MALLOC_MAX/size)
! 1436: return NULL;
! 1437: if (!ptr)
! 1438: return malloc(size * num);
! 1439: return realloc(ptr, size * num);
! 1440: }
! 1441:
! 1442: /* Take a filename and filename length and return the most significant
! 1443: * filename suffix we can find. This ignores suffixes such as "~",
! 1444: * ".bak", ".orig", ".~1~", etc. */
! 1445: const char *find_filename_suffix(const char *fn, int fn_len, int *len_ptr)
! 1446: {
! 1447: const char *suf, *s;
! 1448: BOOL had_tilde;
! 1449: int s_len;
! 1450:
! 1451: /* One or more dots at the start aren't a suffix. */
! 1452: while (fn_len && *fn == '.') fn++, fn_len--;
! 1453:
! 1454: /* Ignore the ~ in a "foo~" filename. */
! 1455: if (fn_len > 1 && fn[fn_len-1] == '~')
! 1456: fn_len--, had_tilde = True;
! 1457: else
! 1458: had_tilde = False;
! 1459:
! 1460: /* Assume we don't find an suffix. */
! 1461: suf = "";
! 1462: *len_ptr = 0;
! 1463:
! 1464: /* Find the last significant suffix. */
! 1465: for (s = fn + fn_len; fn_len > 1; ) {
! 1466: while (*--s != '.' && s != fn) {}
! 1467: if (s == fn)
! 1468: break;
! 1469: s_len = fn_len - (s - fn);
! 1470: fn_len = s - fn;
! 1471: if (s_len == 4) {
! 1472: if (strcmp(s+1, "bak") == 0
! 1473: || strcmp(s+1, "old") == 0)
! 1474: continue;
! 1475: } else if (s_len == 5) {
! 1476: if (strcmp(s+1, "orig") == 0)
! 1477: continue;
! 1478: } else if (s_len > 2 && had_tilde
! 1479: && s[1] == '~' && isDigit(s + 2))
! 1480: continue;
! 1481: *len_ptr = s_len;
! 1482: suf = s;
! 1483: if (s_len == 1)
! 1484: break;
! 1485: /* Determine if the suffix is all digits. */
! 1486: for (s++, s_len--; s_len > 0; s++, s_len--) {
! 1487: if (!isDigit(s))
! 1488: return suf;
! 1489: }
! 1490: /* An all-digit suffix may not be that signficant. */
! 1491: s = suf;
! 1492: }
! 1493:
! 1494: return suf;
! 1495: }
! 1496:
! 1497: /* This is an implementation of the Levenshtein distance algorithm. It
! 1498: * was implemented to avoid needing a two-dimensional matrix (to save
! 1499: * memory). It was also tweaked to try to factor in the ASCII distance
! 1500: * between changed characters as a minor distance quantity. The normal
! 1501: * Levenshtein units of distance (each signifying a single change between
! 1502: * the two strings) are defined as a "UNIT". */
! 1503:
! 1504: #define UNIT (1 << 16)
! 1505:
! 1506: uint32 fuzzy_distance(const char *s1, int len1, const char *s2, int len2)
! 1507: {
! 1508: uint32 a[MAXPATHLEN], diag, above, left, diag_inc, above_inc, left_inc;
! 1509: int32 cost;
! 1510: int i1, i2;
! 1511:
! 1512: if (!len1 || !len2) {
! 1513: if (!len1) {
! 1514: s1 = s2;
! 1515: len1 = len2;
! 1516: }
! 1517: for (i1 = 0, cost = 0; i1 < len1; i1++)
! 1518: cost += s1[i1];
! 1519: return (int32)len1 * UNIT + cost;
! 1520: }
! 1521:
! 1522: for (i2 = 0; i2 < len2; i2++)
! 1523: a[i2] = (i2+1) * UNIT;
! 1524:
! 1525: for (i1 = 0; i1 < len1; i1++) {
! 1526: diag = i1 * UNIT;
! 1527: above = (i1+1) * UNIT;
! 1528: for (i2 = 0; i2 < len2; i2++) {
! 1529: left = a[i2];
! 1530: if ((cost = *((uchar*)s1+i1) - *((uchar*)s2+i2)) != 0) {
! 1531: if (cost < 0)
! 1532: cost = UNIT - cost;
! 1533: else
! 1534: cost = UNIT + cost;
! 1535: }
! 1536: diag_inc = diag + cost;
! 1537: left_inc = left + UNIT + *((uchar*)s1+i1);
! 1538: above_inc = above + UNIT + *((uchar*)s2+i2);
! 1539: a[i2] = above = left < above
! 1540: ? (left_inc < diag_inc ? left_inc : diag_inc)
! 1541: : (above_inc < diag_inc ? above_inc : diag_inc);
! 1542: diag = left;
! 1543: }
! 1544: }
! 1545:
! 1546: return a[len2-1];
! 1547: }
! 1548:
! 1549: #define BB_SLOT_SIZE (16*1024) /* Desired size in bytes */
! 1550: #define BB_PER_SLOT_BITS (BB_SLOT_SIZE * 8) /* Number of bits per slot */
! 1551: #define BB_PER_SLOT_INTS (BB_SLOT_SIZE / 4) /* Number of int32s per slot */
! 1552:
! 1553: struct bitbag {
! 1554: uint32 **bits;
! 1555: int slot_cnt;
! 1556: };
! 1557:
! 1558: struct bitbag *bitbag_create(int max_ndx)
! 1559: {
! 1560: struct bitbag *bb = new(struct bitbag);
! 1561: bb->slot_cnt = (max_ndx + BB_PER_SLOT_BITS - 1) / BB_PER_SLOT_BITS;
! 1562:
! 1563: if (!(bb->bits = (uint32**)calloc(bb->slot_cnt, sizeof (uint32*))))
! 1564: out_of_memory("bitbag_create");
! 1565:
! 1566: return bb;
! 1567: }
! 1568:
! 1569: void bitbag_set_bit(struct bitbag *bb, int ndx)
! 1570: {
! 1571: int slot = ndx / BB_PER_SLOT_BITS;
! 1572: ndx %= BB_PER_SLOT_BITS;
! 1573:
! 1574: if (!bb->bits[slot]) {
! 1575: if (!(bb->bits[slot] = (uint32*)calloc(BB_PER_SLOT_INTS, 4)))
! 1576: out_of_memory("bitbag_set_bit");
! 1577: }
! 1578:
! 1579: bb->bits[slot][ndx/32] |= 1u << (ndx % 32);
! 1580: }
! 1581:
! 1582: #if 0 /* not needed yet */
! 1583: void bitbag_clear_bit(struct bitbag *bb, int ndx)
! 1584: {
! 1585: int slot = ndx / BB_PER_SLOT_BITS;
! 1586: ndx %= BB_PER_SLOT_BITS;
! 1587:
! 1588: if (!bb->bits[slot])
! 1589: return;
! 1590:
! 1591: bb->bits[slot][ndx/32] &= ~(1u << (ndx % 32));
! 1592: }
! 1593:
! 1594: int bitbag_check_bit(struct bitbag *bb, int ndx)
! 1595: {
! 1596: int slot = ndx / BB_PER_SLOT_BITS;
! 1597: ndx %= BB_PER_SLOT_BITS;
! 1598:
! 1599: if (!bb->bits[slot])
! 1600: return 0;
! 1601:
! 1602: return bb->bits[slot][ndx/32] & (1u << (ndx % 32)) ? 1 : 0;
! 1603: }
! 1604: #endif
! 1605:
! 1606: /* Call this with -1 to start checking from 0. Returns -1 at the end. */
! 1607: int bitbag_next_bit(struct bitbag *bb, int after)
! 1608: {
! 1609: uint32 bits, mask;
! 1610: int i, ndx = after + 1;
! 1611: int slot = ndx / BB_PER_SLOT_BITS;
! 1612: ndx %= BB_PER_SLOT_BITS;
! 1613:
! 1614: mask = (1u << (ndx % 32)) - 1;
! 1615: for (i = ndx / 32; slot < bb->slot_cnt; slot++, i = mask = 0) {
! 1616: if (!bb->bits[slot])
! 1617: continue;
! 1618: for ( ; i < BB_PER_SLOT_INTS; i++, mask = 0) {
! 1619: if (!(bits = bb->bits[slot][i] & ~mask))
! 1620: continue;
! 1621: /* The xor magic figures out the lowest enabled bit in
! 1622: * bits, and the switch quickly computes log2(bit). */
! 1623: switch (bits ^ (bits & (bits-1))) {
! 1624: #define LOG2(n) case 1u << n: return slot*BB_PER_SLOT_BITS + i*32 + n
! 1625: LOG2(0); LOG2(1); LOG2(2); LOG2(3);
! 1626: LOG2(4); LOG2(5); LOG2(6); LOG2(7);
! 1627: LOG2(8); LOG2(9); LOG2(10); LOG2(11);
! 1628: LOG2(12); LOG2(13); LOG2(14); LOG2(15);
! 1629: LOG2(16); LOG2(17); LOG2(18); LOG2(19);
! 1630: LOG2(20); LOG2(21); LOG2(22); LOG2(23);
! 1631: LOG2(24); LOG2(25); LOG2(26); LOG2(27);
! 1632: LOG2(28); LOG2(29); LOG2(30); LOG2(31);
! 1633: }
! 1634: return -1; /* impossible... */
! 1635: }
! 1636: }
! 1637:
! 1638: return -1;
! 1639: }
! 1640:
! 1641: void flist_ndx_push(flist_ndx_list *lp, int ndx)
! 1642: {
! 1643: struct flist_ndx_item *item;
! 1644:
! 1645: if (!(item = new(struct flist_ndx_item)))
! 1646: out_of_memory("flist_ndx_push");
! 1647: item->next = NULL;
! 1648: item->ndx = ndx;
! 1649: if (lp->tail)
! 1650: lp->tail->next = item;
! 1651: else
! 1652: lp->head = item;
! 1653: lp->tail = item;
! 1654: }
! 1655:
! 1656: int flist_ndx_pop(flist_ndx_list *lp)
! 1657: {
! 1658: struct flist_ndx_item *next;
! 1659: int ndx;
! 1660:
! 1661: if (!lp->head)
! 1662: return -1;
! 1663:
! 1664: ndx = lp->head->ndx;
! 1665: next = lp->head->next;
! 1666: free(lp->head);
! 1667: lp->head = next;
! 1668: if (!next)
! 1669: lp->tail = NULL;
! 1670:
! 1671: return ndx;
! 1672: }
! 1673:
! 1674: void *expand_item_list(item_list *lp, size_t item_size,
! 1675: const char *desc, int incr)
! 1676: {
! 1677: /* First time through, 0 <= 0, so list is expanded. */
! 1678: if (lp->malloced <= lp->count) {
! 1679: void *new_ptr;
! 1680: size_t new_size = lp->malloced;
! 1681: if (incr < 0)
! 1682: new_size += -incr; /* increase slowly */
! 1683: else if (new_size < (size_t)incr)
! 1684: new_size += incr;
! 1685: else
! 1686: new_size *= 2;
! 1687: if (new_size < lp->malloced)
! 1688: overflow_exit("expand_item_list");
! 1689: /* Using _realloc_array() lets us pass the size, not a type. */
! 1690: new_ptr = _realloc_array(lp->items, item_size, new_size);
! 1691: if (verbose >= 4) {
! 1692: rprintf(FINFO, "[%s] expand %s to %.0f bytes, did%s move\n",
! 1693: who_am_i(), desc, (double)new_size * item_size,
! 1694: new_ptr == lp->items ? " not" : "");
! 1695: }
! 1696: if (!new_ptr)
! 1697: out_of_memory("expand_item_list");
! 1698:
! 1699: lp->items = new_ptr;
! 1700: lp->malloced = new_size;
! 1701: }
! 1702: return (char*)lp->items + (lp->count++ * item_size);
! 1703: }
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>