Annotation of mandoc/compat_fts.c, Revision 1.9
1.1 schwarze 1: #include "config.h"
2:
1.3 schwarze 3: #if HAVE_FTS
1.1 schwarze 4:
5: int dummy;
6:
7: #else
8:
1.9 ! schwarze 9: /* $Id: compat_fts.c,v 1.8 2015/02/07 07:53:01 schwarze Exp $ */
1.7 schwarze 10: /* $OpenBSD: fts.c,v 1.50 2015/01/16 16:48:51 deraadt Exp $ */
1.1 schwarze 11:
12: /*-
13: * Copyright (c) 1990, 1993, 1994
14: * The Regents of the University of California. All rights reserved.
15: *
16: * Redistribution and use in source and binary forms, with or without
17: * modification, are permitted provided that the following conditions
18: * are met:
19: * 1. Redistributions of source code must retain the above copyright
20: * notice, this list of conditions and the following disclaimer.
21: * 2. Redistributions in binary form must reproduce the above copyright
22: * notice, this list of conditions and the following disclaimer in the
23: * documentation and/or other materials provided with the distribution.
24: * 3. Neither the name of the University nor the names of its contributors
25: * may be used to endorse or promote products derived from this software
26: * without specific prior written permission.
27: *
28: * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
29: * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
30: * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
31: * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
32: * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
33: * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
34: * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
35: * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
36: * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
37: * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38: * SUCH DAMAGE.
39: */
40:
41: #include <sys/stat.h>
42: #include <sys/types.h>
43:
44: #include <dirent.h>
45: #include <errno.h>
46: #include <fcntl.h>
47: #include <limits.h>
48: #include <stdlib.h>
49: #include <string.h>
50: #include <unistd.h>
51: #include "compat_fts.h"
52:
1.7 schwarze 53: #define MAXIMUM(a, b) (((a) > (b)) ? (a) : (b))
54:
1.1 schwarze 55: static FTSENT *fts_alloc(FTS *, const char *, size_t);
56: static FTSENT *fts_build(FTS *);
57: static void fts_lfree(FTSENT *);
58: static void fts_load(FTS *, FTSENT *);
59: static size_t fts_maxarglen(char * const *);
60: static void fts_padjust(FTS *, FTSENT *);
61: static int fts_palloc(FTS *, size_t);
62: static unsigned short fts_stat(FTS *, FTSENT *);
63:
64: #define ISDOT(a) (a[0] == '.' && (!a[1] || (a[1] == '.' && !a[2])))
1.6 schwarze 65: #ifndef O_DIRECTORY
66: #define O_DIRECTORY 0
1.8 schwarze 67: #endif
68: #ifndef O_CLOEXEC
69: #define O_CLOEXEC 0
1.6 schwarze 70: #endif
1.1 schwarze 71:
72: #define CLR(opt) (sp->fts_options &= ~(opt))
73: #define ISSET(opt) (sp->fts_options & (opt))
74: #define SET(opt) (sp->fts_options |= (opt))
75:
76: FTS *
77: fts_open(char * const *argv, int options, void *dummy)
78: {
79: FTS *sp;
80: FTSENT *p, *root;
81: int nitems;
82: FTSENT *parent, *tmp;
83: size_t len;
84:
85: /* Options check. */
86: if (options & ~FTS_OPTIONMASK) {
87: errno = EINVAL;
88: return (NULL);
89: }
90:
91: /* Allocate/initialize the stream */
92: if ((sp = calloc(1, sizeof(FTS))) == NULL)
93: return (NULL);
94: sp->fts_options = options;
95:
96: /*
97: * Start out with 1K of path space, and enough, in any case,
98: * to hold the user's paths.
99: */
1.7 schwarze 100: if (fts_palloc(sp, MAXIMUM(fts_maxarglen(argv), PATH_MAX)))
1.1 schwarze 101: goto mem1;
102:
103: /* Allocate/initialize root's parent. */
104: if ((parent = fts_alloc(sp, "", 0)) == NULL)
105: goto mem2;
106: parent->fts_level = FTS_ROOTPARENTLEVEL;
107:
108: /* Allocate/initialize root(s). */
109: for (root = NULL, nitems = 0; *argv; ++argv, ++nitems) {
110: /* Don't allow zero-length paths. */
111: if ((len = strlen(*argv)) == 0) {
112: errno = ENOENT;
113: goto mem3;
114: }
115:
116: if ((p = fts_alloc(sp, *argv, len)) == NULL)
117: goto mem3;
118: p->fts_level = FTS_ROOTLEVEL;
119: p->fts_parent = parent;
120: p->fts_accpath = p->fts_name;
121: p->fts_info = fts_stat(sp, p);
122:
123: /* Command-line "." and ".." are real directories. */
124: if (p->fts_info == FTS_DOT)
125: p->fts_info = FTS_D;
126:
127: p->fts_link = NULL;
128: if (root == NULL)
129: tmp = root = p;
130: else {
131: tmp->fts_link = p;
132: tmp = p;
133: }
134: }
135:
136: /*
137: * Allocate a dummy pointer and make fts_read think that we've just
138: * finished the node before the root(s); set p->fts_info to FTS_INIT
139: * so that everything about the "current" node is ignored.
140: */
141: if ((sp->fts_cur = fts_alloc(sp, "", 0)) == NULL)
142: goto mem3;
143: sp->fts_cur->fts_link = root;
144: sp->fts_cur->fts_info = FTS_INIT;
145:
146: if (nitems == 0)
147: free(parent);
148:
149: return (sp);
150:
151: mem3: fts_lfree(root);
152: free(parent);
153: mem2: free(sp->fts_path);
154: mem1: free(sp);
155: return (NULL);
156: }
157:
158: static void
159: fts_load(FTS *sp, FTSENT *p)
160: {
161: size_t len;
162: char *cp;
163:
164: /*
165: * Load the stream structure for the next traversal. Since we don't
166: * actually enter the directory until after the preorder visit, set
167: * the fts_accpath field specially so the chdir gets done to the right
168: * place and the user can access the first node. From fts_open it's
169: * known that the path will fit.
170: */
171: len = p->fts_pathlen = p->fts_namelen;
172: memmove(sp->fts_path, p->fts_name, len + 1);
173: if ((cp = strrchr(p->fts_name, '/')) && (cp != p->fts_name || cp[1])) {
174: len = strlen(++cp);
175: memmove(p->fts_name, cp, len + 1);
176: p->fts_namelen = len;
177: }
178: p->fts_accpath = p->fts_path = sp->fts_path;
179: sp->fts_dev = p->fts_dev;
180: }
181:
182: int
183: fts_close(FTS *sp)
184: {
185: FTSENT *freep, *p;
186:
187: /*
188: * This still works if we haven't read anything -- the dummy structure
189: * points to the root list, so we step through to the end of the root
190: * list which has a valid parent pointer.
191: */
192: if (sp->fts_cur) {
193: for (p = sp->fts_cur; p->fts_level >= FTS_ROOTLEVEL;) {
194: freep = p;
195: p = p->fts_link ? p->fts_link : p->fts_parent;
196: free(freep);
197: }
198: free(p);
199: }
200:
201: /* Free up child linked list, sort array, path buffer, stream ptr.*/
202: if (sp->fts_child)
203: fts_lfree(sp->fts_child);
204: free(sp->fts_path);
205: free(sp);
206:
1.9 ! schwarze 207: return (0);
1.1 schwarze 208: }
209:
210: /*
211: * Special case of "/" at the end of the path so that slashes aren't
212: * appended which would cause paths to be written as "....//foo".
213: */
214: #define NAPPEND(p) \
215: (p->fts_path[p->fts_pathlen - 1] == '/' \
216: ? p->fts_pathlen - 1 : p->fts_pathlen)
217:
218: FTSENT *
219: fts_read(FTS *sp)
220: {
221: FTSENT *p, *tmp;
222: int instr;
223: char *t;
224:
225: /* If finished or unrecoverable error, return NULL. */
226: if (sp->fts_cur == NULL || ISSET(FTS_STOP))
227: return (NULL);
228:
229: /* Set current node pointer. */
230: p = sp->fts_cur;
231:
232: /* Save and zero out user instructions. */
233: instr = p->fts_instr;
234: p->fts_instr = FTS_NOINSTR;
235:
236: /* Directory in pre-order. */
237: if (p->fts_info == FTS_D) {
238: /* If skipped or crossed mount point, do post-order visit. */
239: if (instr == FTS_SKIP ||
240: (ISSET(FTS_XDEV) && p->fts_dev != sp->fts_dev)) {
241: if (sp->fts_child) {
242: fts_lfree(sp->fts_child);
243: sp->fts_child = NULL;
244: }
245: p->fts_info = FTS_DP;
246: return (p);
247: }
248:
249: /*
250: * If haven't read do so. If the read fails, fts_build sets
251: * FTS_STOP or the fts_info field of the node.
252: */
253: if (sp->fts_child) {
1.9 ! schwarze 254: /* nothing */
1.1 schwarze 255: } else if ((sp->fts_child = fts_build(sp)) == NULL) {
256: if (ISSET(FTS_STOP))
257: return (NULL);
258: return (p);
259: }
260: p = sp->fts_child;
261: sp->fts_child = NULL;
262: goto name;
263: }
264:
265: /* Move to the next node on this level. */
266: next: tmp = p;
267: if ((p = p->fts_link)) {
268: free(tmp);
269:
270: /*
271: * If reached the top, return to the original directory (or
272: * the root of the tree), and load the paths for the next root.
273: */
274: if (p->fts_level == FTS_ROOTLEVEL) {
275: fts_load(sp, p);
276: return (sp->fts_cur = p);
277: }
278:
279: /*
280: * User may have called fts_set on the node. If skipped,
281: * ignore. If followed, get a file descriptor so we can
282: * get back if necessary.
283: */
284: if (p->fts_instr == FTS_SKIP)
285: goto next;
286:
287: name: t = sp->fts_path + NAPPEND(p->fts_parent);
288: *t++ = '/';
289: memmove(t, p->fts_name, p->fts_namelen + 1);
290: return (sp->fts_cur = p);
291: }
292:
293: /* Move up to the parent node. */
294: p = tmp->fts_parent;
295: free(tmp);
296:
297: if (p->fts_level == FTS_ROOTPARENTLEVEL) {
298: /*
299: * Done; free everything up and set errno to 0 so the user
300: * can distinguish between error and EOF.
301: */
302: free(p);
303: errno = 0;
304: return (sp->fts_cur = NULL);
305: }
306:
307: /* NUL terminate the pathname. */
308: sp->fts_path[p->fts_pathlen] = '\0';
309:
310: p->fts_info = p->fts_errno ? FTS_ERR : FTS_DP;
311: return (sp->fts_cur = p);
312: }
313:
314: /*
315: * Fts_set takes the stream as an argument although it's not used in this
316: * implementation; it would be necessary if anyone wanted to add global
317: * semantics to fts using fts_set. An error return is allowed for similar
318: * reasons.
319: */
320: /* ARGSUSED */
321: int
322: fts_set(FTS *sp, FTSENT *p, int instr)
323: {
324: if (instr && instr != FTS_NOINSTR && instr != FTS_SKIP) {
325: errno = EINVAL;
326: return (1);
327: }
328: p->fts_instr = instr;
329: return (0);
330: }
331:
332: /*
333: * This is the tricky part -- do not casually change *anything* in here. The
334: * idea is to build the linked list of entries that are used by fts_children
335: * and fts_read. There are lots of special cases.
336: *
337: * The real slowdown in walking the tree is the stat calls. If FTS_NOSTAT is
338: * set and it's a physical walk (so that symbolic links can't be directories),
339: * we can do things quickly. First, if it's a 4.4BSD file system, the type
340: * of the file is in the directory entry. Otherwise, we assume that the number
341: * of subdirectories in a node is equal to the number of links to the parent.
342: * The former skips all stat calls. The latter skips stat calls in any leaf
343: * directories and for any files after the subdirectories in the directory have
344: * been found, cutting the stat calls by about 2/3.
345: */
346: static FTSENT *
347: fts_build(FTS *sp)
348: {
349: struct dirent *dp;
350: FTSENT *p, *head;
351: FTSENT *cur, *tail;
352: DIR *dirp;
353: void *oldaddr;
1.2 schwarze 354: size_t dlen, len, maxlen;
1.9 ! schwarze 355: int nitems, level, doadjust;
1.1 schwarze 356: int saved_errno;
357: char *cp;
358:
359: /* Set current node pointer. */
360: cur = sp->fts_cur;
361:
362: /*
363: * Open the directory for reading. If this fails, we're done.
364: * If being called from fts_read, set the fts_info field.
365: */
366: if ((dirp = opendir(cur->fts_accpath)) == NULL) {
367: cur->fts_info = FTS_DNR;
368: cur->fts_errno = errno;
369: return (NULL);
370: }
371:
372: /*
373: * Figure out the max file name length that can be stored in the
374: * current path -- the inner loop allocates more path as necessary.
375: * We really wouldn't have to do the maxlen calculations here, we
376: * could do them in fts_read before returning the path, but it's a
377: * lot easier here since the length is part of the dirent structure.
378: *
379: * If not changing directories set a pointer so that can just append
380: * each new name into the path.
381: */
382: len = NAPPEND(cur);
1.9 ! schwarze 383: cp = sp->fts_path + len;
! 384: *cp++ = '/';
1.1 schwarze 385: len++;
386: maxlen = sp->fts_pathlen - len;
387:
388: /*
389: * fts_level is signed so we must prevent it from wrapping
390: * around to FTS_ROOTLEVEL and FTS_ROOTPARENTLEVEL.
391: */
392: level = cur->fts_level;
393: if (level < FTS_MAXLEVEL)
394: level++;
395:
396: /* Read the directory, attaching each entry to the `link' pointer. */
397: doadjust = 0;
398: for (head = tail = NULL, nitems = 0; dirp && (dp = readdir(dirp));) {
399: if (ISDOT(dp->d_name))
400: continue;
401:
1.4 schwarze 402: #if HAVE_DIRENT_NAMLEN
1.2 schwarze 403: dlen = dp->d_namlen;
404: #else
405: dlen = strlen(dp->d_name);
406: #endif
407:
408: if (!(p = fts_alloc(sp, dp->d_name, dlen)))
1.1 schwarze 409: goto mem1;
1.2 schwarze 410: if (dlen >= maxlen) { /* include space for NUL */
1.1 schwarze 411: oldaddr = sp->fts_path;
1.2 schwarze 412: if (fts_palloc(sp, dlen + len + 1)) {
1.1 schwarze 413: /*
414: * No more memory for path or structures. Save
415: * errno, free up the current structure and the
416: * structures already allocated.
417: */
418: mem1: saved_errno = errno;
419: if (p)
420: free(p);
421: fts_lfree(head);
422: (void)closedir(dirp);
423: cur->fts_info = FTS_ERR;
424: SET(FTS_STOP);
425: errno = saved_errno;
426: return (NULL);
427: }
428: /* Did realloc() change the pointer? */
429: if (oldaddr != sp->fts_path) {
430: doadjust = 1;
1.9 ! schwarze 431: cp = sp->fts_path + len;
1.1 schwarze 432: }
433: maxlen = sp->fts_pathlen - len;
434: }
435:
436: p->fts_level = level;
437: p->fts_parent = sp->fts_cur;
1.2 schwarze 438: p->fts_pathlen = len + dlen;
1.1 schwarze 439: if (p->fts_pathlen < len) {
440: /*
441: * If we wrap, free up the current structure and
442: * the structures already allocated, then error
443: * out with ENAMETOOLONG.
444: */
445: free(p);
446: fts_lfree(head);
447: (void)closedir(dirp);
448: cur->fts_info = FTS_ERR;
449: SET(FTS_STOP);
450: errno = ENAMETOOLONG;
451: return (NULL);
452: }
453:
1.9 ! schwarze 454: /* Build a file name for fts_stat to stat. */
! 455: p->fts_accpath = p->fts_path;
! 456: memmove(cp, p->fts_name, p->fts_namelen + 1);
! 457: /* Stat it. */
! 458: p->fts_info = fts_stat(sp, p);
1.1 schwarze 459:
460: /* We walk in directory order so "ls -f" doesn't get upset. */
461: p->fts_link = NULL;
462: if (head == NULL)
463: head = tail = p;
464: else {
465: tail->fts_link = p;
466: tail = p;
467: }
468: ++nitems;
469: }
470: if (dirp)
471: (void)closedir(dirp);
472:
473: /*
474: * If realloc() changed the address of the path, adjust the
475: * addresses for the rest of the tree and the dir list.
476: */
477: if (doadjust)
478: fts_padjust(sp, head);
479:
480: /*
481: * If not changing directories, reset the path back to original
482: * state.
483: */
1.9 ! schwarze 484: if (len == sp->fts_pathlen || nitems == 0)
! 485: --cp;
! 486: *cp = '\0';
1.1 schwarze 487:
488: /* If didn't find anything, return NULL. */
489: if (!nitems) {
490: cur->fts_info = FTS_DP;
491: return (NULL);
492: }
493: return (head);
494: }
495:
496: static unsigned short
497: fts_stat(FTS *sp, FTSENT *p)
498: {
499: FTSENT *t;
500: dev_t dev;
501: ino_t ino;
502: struct stat *sbp;
503:
504: /* If user needs stat info, stat buffer already allocated. */
505: sbp = p->fts_statp;
506:
507: if (lstat(p->fts_accpath, sbp)) {
508: p->fts_errno = errno;
509: memset(sbp, 0, sizeof(struct stat));
510: return (FTS_NS);
511: }
512:
513: if (S_ISDIR(sbp->st_mode)) {
514: /*
515: * Set the device/inode. Used to find cycles and check for
516: * crossing mount points. Also remember the link count, used
517: * in fts_build to limit the number of stat calls. It is
518: * understood that these fields are only referenced if fts_info
519: * is set to FTS_D.
520: */
521: dev = p->fts_dev = sbp->st_dev;
522: ino = p->fts_ino = sbp->st_ino;
523: p->fts_nlink = sbp->st_nlink;
524:
525: if (ISDOT(p->fts_name))
526: return (FTS_DOT);
527:
528: /*
529: * Cycle detection is done by brute force when the directory
530: * is first encountered. If the tree gets deep enough or the
531: * number of symbolic links to directories is high enough,
532: * something faster might be worthwhile.
533: */
534: for (t = p->fts_parent;
535: t->fts_level >= FTS_ROOTLEVEL; t = t->fts_parent)
536: if (ino == t->fts_ino && dev == t->fts_dev) {
537: p->fts_cycle = t;
538: return (FTS_DC);
539: }
540: return (FTS_D);
541: }
542: if (S_ISLNK(sbp->st_mode))
543: return (FTS_SL);
544: if (S_ISREG(sbp->st_mode))
545: return (FTS_F);
546: return (FTS_DEFAULT);
547: }
548:
549: static FTSENT *
550: fts_alloc(FTS *sp, const char *name, size_t namelen)
551: {
552: FTSENT *p;
553: size_t len;
554:
555: len = sizeof(FTSENT) + namelen;
556: if ((p = calloc(1, len)) == NULL)
557: return (NULL);
558:
559: p->fts_path = sp->fts_path;
560: p->fts_namelen = namelen;
561: p->fts_instr = FTS_NOINSTR;
1.2 schwarze 562: p->fts_statp = malloc(sizeof(struct stat));
563: if (p->fts_statp == NULL) {
564: free(p);
565: return (NULL);
566: }
1.1 schwarze 567: memcpy(p->fts_name, name, namelen);
568:
569: return (p);
570: }
571:
572: static void
573: fts_lfree(FTSENT *head)
574: {
575: FTSENT *p;
576:
577: /* Free a linked list of structures. */
578: while ((p = head)) {
579: head = head->fts_link;
580: free(p);
581: }
582: }
583:
584: /*
585: * Allow essentially unlimited paths; find, rm, ls should all work on any tree.
586: * Most systems will allow creation of paths much longer than PATH_MAX, even
587: * though the kernel won't resolve them. Add the size (not just what's needed)
588: * plus 256 bytes so don't realloc the path 2 bytes at a time.
589: */
590: static int
591: fts_palloc(FTS *sp, size_t more)
592: {
593: char *p;
594:
595: /*
596: * Check for possible wraparound.
597: */
598: more += 256;
599: if (sp->fts_pathlen + more < sp->fts_pathlen) {
600: if (sp->fts_path)
601: free(sp->fts_path);
602: sp->fts_path = NULL;
603: errno = ENAMETOOLONG;
604: return (1);
605: }
606: sp->fts_pathlen += more;
607: p = realloc(sp->fts_path, sp->fts_pathlen);
608: if (p == NULL) {
609: if (sp->fts_path)
610: free(sp->fts_path);
611: sp->fts_path = NULL;
612: return (1);
613: }
614: sp->fts_path = p;
615: return (0);
616: }
617:
618: /*
619: * When the path is realloc'd, have to fix all of the pointers in structures
620: * already returned.
621: */
622: static void
623: fts_padjust(FTS *sp, FTSENT *head)
624: {
625: FTSENT *p;
626: char *addr = sp->fts_path;
627:
628: #define ADJUST(p) { \
629: if ((p)->fts_accpath != (p)->fts_name) { \
630: (p)->fts_accpath = \
631: (char *)addr + ((p)->fts_accpath - (p)->fts_path); \
632: } \
633: (p)->fts_path = addr; \
634: }
635: /* Adjust the current set of children. */
636: for (p = sp->fts_child; p; p = p->fts_link)
637: ADJUST(p);
638:
639: /* Adjust the rest of the tree, including the current level. */
640: for (p = head; p->fts_level >= FTS_ROOTLEVEL;) {
641: ADJUST(p);
642: p = p->fts_link ? p->fts_link : p->fts_parent;
643: }
644: }
645:
646: static size_t
647: fts_maxarglen(char * const *argv)
648: {
649: size_t len, max;
650:
651: for (max = 0; *argv; ++argv)
652: if ((len = strlen(*argv)) > max)
653: max = len;
654: return (max + 1);
655: }
656:
657: #endif
CVSweb