2 * Copyright (c) 1989, 1993, 1995
3 * The Regents of the University of California. All rights reserved.
5 * This code is derived from software contributed to Berkeley by
6 * Poul-Henning Kamp of the FreeBSD Project.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 4. Neither the name of the University nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * @(#)vfs_cache.c 8.5 (Berkeley) 3/22/95
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD$");
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/kernel.h>
42 #include <sys/mutex.h>
43 #include <sys/sysctl.h>
44 #include <sys/mount.h>
45 #include <sys/vnode.h>
46 #include <sys/namei.h>
47 #include <sys/malloc.h>
48 #include <sys/syscallsubr.h>
49 #include <sys/sysproto.h>
51 #include <sys/filedesc.h>
52 #include <sys/fnv_hash.h>
57 * This structure describes the elements in the cache of recent
58 * names looked up by namei.
62 LIST_ENTRY(namecache) nc_hash; /* hash chain */
63 LIST_ENTRY(namecache) nc_src; /* source vnode list */
64 TAILQ_ENTRY(namecache) nc_dst; /* destination vnode list */
65 struct vnode *nc_dvp; /* vnode of parent of name */
66 struct vnode *nc_vp; /* vnode the name refers to */
67 u_char nc_flag; /* flag bits */
68 u_char nc_nlen; /* length of name */
69 char nc_name[0]; /* segment name */
73 * Name caching works as follows:
75 * Names found by directory scans are retained in a cache
76 * for future reference. It is managed LRU, so frequently
77 * used names will hang around. Cache is indexed by hash value
78 * obtained from (vp, name) where vp refers to the directory
81 * If it is a "negative" entry, (i.e. for a name that is known NOT to
82 * exist) the vnode pointer will be NULL.
84 * Upon reaching the last segment of a path, if the reference
85 * is for DELETE, or NOCACHE is set (rewrite), and the
86 * name is located in the cache, it will be dropped.
90 * Structures associated with name cacheing.
92 #define NCHHASH(hash) \
93 (&nchashtbl[(hash) & nchash])
94 static LIST_HEAD(nchashhead, namecache) *nchashtbl; /* Hash Table */
95 static TAILQ_HEAD(, namecache) ncneg; /* Hash Table */
96 static u_long nchash; /* size of hash table */
97 SYSCTL_ULONG(_debug, OID_AUTO, nchash, CTLFLAG_RD, &nchash, 0, "");
98 static u_long ncnegfactor = 16; /* ratio of negative entries */
99 SYSCTL_ULONG(_debug, OID_AUTO, ncnegfactor, CTLFLAG_RW, &ncnegfactor, 0, "");
100 static u_long numneg; /* number of cache entries allocated */
101 SYSCTL_ULONG(_debug, OID_AUTO, numneg, CTLFLAG_RD, &numneg, 0, "");
102 static u_long numcache; /* number of cache entries allocated */
103 SYSCTL_ULONG(_debug, OID_AUTO, numcache, CTLFLAG_RD, &numcache, 0, "");
104 static u_long numcachehv; /* number of cache entries with vnodes held */
105 SYSCTL_ULONG(_debug, OID_AUTO, numcachehv, CTLFLAG_RD, &numcachehv, 0, "");
107 static u_long numcachepl; /* number of cache purge for leaf entries */
108 SYSCTL_ULONG(_debug, OID_AUTO, numcachepl, CTLFLAG_RD, &numcachepl, 0, "");
110 struct nchstats nchstats; /* cache effectiveness statistics */
112 static struct mtx cache_lock;
113 MTX_SYSINIT(vfscache, &cache_lock, "Name Cache", MTX_DEF);
115 #define CACHE_LOCK() mtx_lock(&cache_lock)
116 #define CACHE_UNLOCK() mtx_unlock(&cache_lock)
119 * UMA zones for the VFS cache.
121 * The small cache is used for entries with short names, which are the
122 * most common. The large cache is used for entries which are too big to
123 * fit in the small cache.
125 static uma_zone_t cache_zone_small;
126 static uma_zone_t cache_zone_large;
128 #define CACHE_PATH_CUTOFF 32
129 #define CACHE_ZONE_SMALL (sizeof(struct namecache) + CACHE_PATH_CUTOFF)
130 #define CACHE_ZONE_LARGE (sizeof(struct namecache) + NAME_MAX)
132 #define cache_alloc(len) uma_zalloc(((len) <= CACHE_PATH_CUTOFF) ? \
133 cache_zone_small : cache_zone_large, M_WAITOK)
134 #define cache_free(ncp) do { \
136 uma_zfree(((ncp)->nc_nlen <= CACHE_PATH_CUTOFF) ? \
137 cache_zone_small : cache_zone_large, (ncp)); \
140 static int doingcache = 1; /* 1 => enable the cache */
141 SYSCTL_INT(_debug, OID_AUTO, vfscache, CTLFLAG_RW, &doingcache, 0, "");
143 /* Export size information to userland */
144 SYSCTL_INT(_debug, OID_AUTO, vnsize, CTLFLAG_RD, 0, sizeof(struct vnode), "");
145 SYSCTL_INT(_debug, OID_AUTO, ncsize, CTLFLAG_RD, 0, sizeof(struct namecache), "");
148 * The new name cache statistics
150 static SYSCTL_NODE(_vfs, OID_AUTO, cache, CTLFLAG_RW, 0, "Name cache statistics");
151 #define STATNODE(mode, name, var) \
152 SYSCTL_ULONG(_vfs_cache, OID_AUTO, name, mode, var, 0, "");
153 STATNODE(CTLFLAG_RD, numneg, &numneg);
154 STATNODE(CTLFLAG_RD, numcache, &numcache);
155 static u_long numcalls; STATNODE(CTLFLAG_RD, numcalls, &numcalls);
156 static u_long dothits; STATNODE(CTLFLAG_RD, dothits, &dothits);
157 static u_long dotdothits; STATNODE(CTLFLAG_RD, dotdothits, &dotdothits);
158 static u_long numchecks; STATNODE(CTLFLAG_RD, numchecks, &numchecks);
159 static u_long nummiss; STATNODE(CTLFLAG_RD, nummiss, &nummiss);
160 static u_long nummisszap; STATNODE(CTLFLAG_RD, nummisszap, &nummisszap);
161 static u_long numposzaps; STATNODE(CTLFLAG_RD, numposzaps, &numposzaps);
162 static u_long numposhits; STATNODE(CTLFLAG_RD, numposhits, &numposhits);
163 static u_long numnegzaps; STATNODE(CTLFLAG_RD, numnegzaps, &numnegzaps);
164 static u_long numneghits; STATNODE(CTLFLAG_RD, numneghits, &numneghits);
166 SYSCTL_OPAQUE(_vfs_cache, OID_AUTO, nchstats, CTLFLAG_RD, &nchstats,
167 sizeof(nchstats), "LU", "VFS cache effectiveness statistics");
171 static void cache_zap(struct namecache *ncp);
173 static MALLOC_DEFINE(M_VFSCACHE, "vfscache", "VFS name cache entries");
176 * Flags in namecache.nc_flag
181 * Grab an atomic snapshot of the name cache hash chain lengths
183 SYSCTL_NODE(_debug, OID_AUTO, hashstat, CTLFLAG_RW, NULL, "hash table stats");
186 sysctl_debug_hashstat_rawnchash(SYSCTL_HANDLER_ARGS)
189 struct nchashhead *ncpp;
190 struct namecache *ncp;
194 n_nchash = nchash + 1; /* nchash is max index, not count */
196 return SYSCTL_OUT(req, 0, n_nchash * sizeof(int));
198 /* Scan hash tables for applicable entries */
199 for (ncpp = nchashtbl; n_nchash > 0; n_nchash--, ncpp++) {
201 LIST_FOREACH(ncp, ncpp, nc_hash) {
204 error = SYSCTL_OUT(req, &count, sizeof(count));
210 SYSCTL_PROC(_debug_hashstat, OID_AUTO, rawnchash, CTLTYPE_INT|CTLFLAG_RD,
211 0, 0, sysctl_debug_hashstat_rawnchash, "S,int", "nchash chain lengths");
214 sysctl_debug_hashstat_nchash(SYSCTL_HANDLER_ARGS)
217 struct nchashhead *ncpp;
218 struct namecache *ncp;
220 int count, maxlength, used, pct;
223 return SYSCTL_OUT(req, 0, 4 * sizeof(int));
225 n_nchash = nchash + 1; /* nchash is max index, not count */
229 /* Scan hash tables for applicable entries */
230 for (ncpp = nchashtbl; n_nchash > 0; n_nchash--, ncpp++) {
232 LIST_FOREACH(ncp, ncpp, nc_hash) {
237 if (maxlength < count)
240 n_nchash = nchash + 1;
241 pct = (used * 100 * 100) / n_nchash;
242 error = SYSCTL_OUT(req, &n_nchash, sizeof(n_nchash));
245 error = SYSCTL_OUT(req, &used, sizeof(used));
248 error = SYSCTL_OUT(req, &maxlength, sizeof(maxlength));
251 error = SYSCTL_OUT(req, &pct, sizeof(pct));
256 SYSCTL_PROC(_debug_hashstat, OID_AUTO, nchash, CTLTYPE_INT|CTLFLAG_RD,
257 0, 0, sysctl_debug_hashstat_nchash, "I", "nchash chain lengths");
262 * Removes a namecache entry from cache, whether it contains an actual
263 * pointer to a vnode or if it is just a negative cache entry.
267 struct namecache *ncp;
271 mtx_assert(&cache_lock, MA_OWNED);
273 LIST_REMOVE(ncp, nc_hash);
274 LIST_REMOVE(ncp, nc_src);
275 if (LIST_EMPTY(&ncp->nc_dvp->v_cache_src)) {
280 TAILQ_REMOVE(&ncp->nc_vp->v_cache_dst, ncp, nc_dst);
282 TAILQ_REMOVE(&ncneg, ncp, nc_dst);
294 * Test whether this (directory) vnode's namei cache entry contains
295 * subdirectories or not. Used to determine whether the directory is
296 * a leaf in the namei cache or not. Note: the directory may still
297 * contain files in the namei cache.
299 * Returns 0 if the directory is a leaf, -1 if it isn't.
302 cache_leaf_test(struct vnode *vp)
304 struct namecache *ncpc;
309 for (ncpc = LIST_FIRST(&vp->v_cache_src);
311 ncpc = LIST_NEXT(ncpc, nc_src)
313 if (ncpc->nc_vp != NULL && ncpc->nc_vp->v_type == VDIR) {
323 * Lookup an entry in the cache
325 * Lookup is called with dvp pointing to the directory to search,
326 * cnp pointing to the name of the entry being sought. If the lookup
327 * succeeds, the vnode is returned in *vpp, and a status of -1 is
328 * returned. If the lookup determines that the name does not exist
329 * (negative cacheing), a status of ENOENT is returned. If the lookup
330 * fails, a status of zero is returned.
334 cache_lookup(dvp, vpp, cnp)
337 struct componentname *cnp;
339 struct namecache *ncp;
343 cnp->cn_flags &= ~MAKEENTRY;
350 if (cnp->cn_nameptr[0] == '.') {
351 if (cnp->cn_namelen == 1) {
357 if (cnp->cn_namelen == 2 && cnp->cn_nameptr[1] == '.') {
359 if (dvp->v_dd->v_id != dvp->v_ddid ||
360 (cnp->cn_flags & MAKEENTRY) == 0) {
371 hash = fnv_32_buf(cnp->cn_nameptr, cnp->cn_namelen, FNV1_32_INIT);
372 hash = fnv_32_buf(&dvp->v_id, sizeof(dvp->v_id), hash);
373 LIST_FOREACH(ncp, (NCHHASH(hash)), nc_hash) {
375 if (ncp->nc_dvp == dvp && ncp->nc_nlen == cnp->cn_namelen &&
376 !bcmp(ncp->nc_name, cnp->cn_nameptr, ncp->nc_nlen))
380 /* We failed to find an entry */
382 if ((cnp->cn_flags & MAKEENTRY) == 0) {
392 /* We don't want to have an entry, so dump it */
393 if ((cnp->cn_flags & MAKEENTRY) == 0) {
395 nchstats.ncs_badhits++;
401 /* We found a "positive" match, return the vnode */
404 nchstats.ncs_goodhits++;
410 /* We found a negative match, and want to create it, so purge */
411 if (cnp->cn_nameiop == CREATE) {
413 nchstats.ncs_badhits++;
421 * We found a "negative" match, so we shift it to the end of
422 * the "negative" cache entries queue to satisfy LRU. Also,
423 * check to see if the entry is a whiteout; indicate this to
424 * the componentname, if so.
426 TAILQ_REMOVE(&ncneg, ncp, nc_dst);
427 TAILQ_INSERT_TAIL(&ncneg, ncp, nc_dst);
428 nchstats.ncs_neghits++;
429 if (ncp->nc_flag & NCF_WHITE)
430 cnp->cn_flags |= ISWHITEOUT;
436 * Add an entry to the cache.
439 cache_enter(dvp, vp, cnp)
442 struct componentname *cnp;
444 struct namecache *ncp;
445 struct nchashhead *ncpp;
454 if (cnp->cn_nameptr[0] == '.') {
455 if (cnp->cn_namelen == 1) {
458 if (cnp->cn_namelen == 2 && cnp->cn_nameptr[1] == '.') {
461 dvp->v_ddid = vp->v_id;
472 ncp = cache_alloc(cnp->cn_namelen);
477 ncp->nc_flag = cnp->cn_flags & ISWHITEOUT ? NCF_WHITE : 0;
478 } else if (vp->v_type == VDIR) {
480 vp->v_ddid = dvp->v_id;
484 * Set the rest of the namecache entry elements, calculate it's
485 * hash key and insert it into the appropriate chain within
486 * the cache entries table.
490 len = ncp->nc_nlen = cnp->cn_namelen;
491 hash = fnv_32_buf(cnp->cn_nameptr, len, FNV1_32_INIT);
492 bcopy(cnp->cn_nameptr, ncp->nc_name, len);
493 hash = fnv_32_buf(&dvp->v_id, sizeof(dvp->v_id), hash);
494 ncpp = NCHHASH(hash);
495 LIST_INSERT_HEAD(ncpp, ncp, nc_hash);
496 if (LIST_EMPTY(&dvp->v_cache_src)) {
500 LIST_INSERT_HEAD(&dvp->v_cache_src, ncp, nc_src);
502 * If the entry is "negative", we place it into the
503 * "negative" cache queue, otherwise, we place it into the
504 * destination vnode's cache entries queue.
507 TAILQ_INSERT_HEAD(&vp->v_cache_dst, ncp, nc_dst);
509 TAILQ_INSERT_TAIL(&ncneg, ncp, nc_dst);
511 if (numneg * ncnegfactor > numcache) {
512 ncp = TAILQ_FIRST(&ncneg);
523 * Name cache initialization, from vfs_init() when we are booting
526 nchinit(void *dummy __unused)
531 cache_zone_small = uma_zcreate("S VFS Cache", CACHE_ZONE_SMALL, NULL,
532 NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_ZINIT);
533 cache_zone_large = uma_zcreate("L VFS Cache", CACHE_ZONE_LARGE, NULL,
534 NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_ZINIT);
536 nchashtbl = hashinit(desiredvnodes * 2, M_VFSCACHE, &nchash);
538 SYSINIT(vfs, SI_SUB_VFS, SI_ORDER_SECOND, nchinit, NULL)
542 * Invalidate all entries to a particular vnode.
544 * Remove all entries in the namecache relating to this vnode and
545 * change the v_id. We take the v_id from a global counter, since
546 * it becomes a handy sequence number in crash-dumps that way.
547 * No valid vnode will ever have (v_id == 0).
549 * XXX: Only time and the size of v_id prevents this from failing:
550 * XXX: In theory we should hunt down all (struct vnode*, v_id)
551 * XXX: soft references and nuke them, at least on the global
552 * XXX: v_id wraparound. The period of resistance can be extended
553 * XXX: by incrementing each vnodes v_id individually instead of
554 * XXX: using the global v_id.
558 * XXX This is sometimes called when a vnode may still be re-used, in which
559 * case v_dd may be invalid. Need to look this up.
565 static u_long nextid;
568 while (!LIST_EMPTY(&vp->v_cache_src))
569 cache_zap(LIST_FIRST(&vp->v_cache_src));
570 while (!TAILQ_EMPTY(&vp->v_cache_dst))
571 cache_zap(TAILQ_FIRST(&vp->v_cache_dst));
575 while (nextid == vp->v_id || !nextid);
583 * Flush all entries referencing a particular filesystem.
585 * Since we need to check it anyway, we will flush all the invalid
586 * entries at the same time.
592 struct nchashhead *ncpp;
593 struct namecache *ncp, *nnp;
594 struct nchashhead mplist;
599 /* Scan hash tables for applicable entries */
601 for (ncpp = &nchashtbl[nchash]; ncpp >= nchashtbl; ncpp--) {
602 for (ncp = LIST_FIRST(ncpp); ncp != 0; ncp = nnp) {
603 nnp = LIST_NEXT(ncp, nc_hash);
604 if (ncp->nc_dvp->v_mount == mp) {
605 LIST_REMOVE(ncp, nc_hash);
606 LIST_INSERT_HEAD(&mplist, ncp, nc_hash);
610 while (!LIST_EMPTY(&mplist))
611 cache_zap(LIST_FIRST(&mplist));
616 * Perform canonical checks and cache lookup and pass on to filesystem
617 * through the vop_cachedlookup only if needed.
622 struct vop_lookup_args /* {
624 struct vnode **a_vpp;
625 struct componentname *a_cnp;
628 struct vnode *dvp, *vp;
631 struct vnode **vpp = ap->a_vpp;
632 struct componentname *cnp = ap->a_cnp;
633 struct ucred *cred = cnp->cn_cred;
634 int flags = cnp->cn_flags;
635 struct thread *td = cnp->cn_thread;
636 u_long vpid; /* capability number of vnode */
640 lockparent = flags & LOCKPARENT;
642 if (dvp->v_type != VDIR)
645 if ((flags & ISLASTCN) && (dvp->v_mount->mnt_flag & MNT_RDONLY) &&
646 (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME))
649 error = VOP_ACCESS(dvp, VEXEC, cred, td);
654 error = cache_lookup(dvp, vpp, cnp);
658 /* We do this because the rest of the system now expects to get
659 * a shared lock, which is later upgraded if LOCKSHARED is not
660 * set. We have so many cases here because of bugs that yield
661 * inconsistant lock states. This all badly needs to be fixed
663 error = VOP_CACHEDLOOKUP(dvp, vpp, cnp);
667 flock = VOP_ISLOCKED(*vpp, td);
668 if (flock != LK_EXCLUSIVE) {
670 if ((flags & ISLASTCN) &&
671 (flags & LOCKSHARED))
672 VOP_LOCK(*vpp, LK_SHARED, td);
674 VOP_LOCK(*vpp, LK_EXCLUSIVE, td);
676 } else if ((flags & ISLASTCN) && (flags & LOCKSHARED))
677 VOP_LOCK(*vpp, LK_DOWNGRADE, td);
683 return (VOP_CACHEDLOOKUP(dvp, vpp, cnp));
691 cnp->cn_flags &= ~PDIRUNLOCK;
692 if (dvp == vp) { /* lookup on "." */
695 } else if (flags & ISDOTDOT) {
696 VOP_UNLOCK(dvp, 0, td);
697 cnp->cn_flags |= PDIRUNLOCK;
699 if ((flags & ISLASTCN) && (flags & LOCKSHARED))
700 error = vget(vp, LK_SHARED, td);
702 error = vget(vp, LK_EXCLUSIVE, td);
704 error = vget(vp, LK_EXCLUSIVE, td);
707 if (!error && lockparent && (flags & ISLASTCN)) {
708 if ((error = vn_lock(dvp, LK_EXCLUSIVE, td)) == 0)
709 cnp->cn_flags &= ~PDIRUNLOCK;
713 if ((flags & ISLASTCN) && (flags & LOCKSHARED))
714 error = vget(vp, LK_SHARED, td);
716 error = vget(vp, LK_EXCLUSIVE, td);
718 error = vget(vp, LK_EXCLUSIVE, td);
720 if (!lockparent || error || !(flags & ISLASTCN)) {
721 VOP_UNLOCK(dvp, 0, td);
722 cnp->cn_flags |= PDIRUNLOCK;
726 * Check that the capability number did not change
727 * while we were waiting for the lock.
730 if (vpid == vp->v_id)
733 if (lockparent && dvp != vp && (flags & ISLASTCN)) {
734 VOP_UNLOCK(dvp, 0, td);
735 cnp->cn_flags |= PDIRUNLOCK;
738 if (cnp->cn_flags & PDIRUNLOCK) {
739 error = vn_lock(dvp, LK_EXCLUSIVE, td);
742 cnp->cn_flags &= ~PDIRUNLOCK;
745 error = VOP_CACHEDLOOKUP(dvp, vpp, cnp);
750 flock = VOP_ISLOCKED(*vpp, td);
751 if (flock != LK_EXCLUSIVE) {
753 if ((flags & ISLASTCN) && (flags & LOCKSHARED))
754 VOP_LOCK(*vpp, LK_SHARED, td);
756 VOP_LOCK(*vpp, LK_EXCLUSIVE, td);
758 } else if ((flags & ISLASTCN) && (flags & LOCKSHARED))
759 VOP_LOCK(*vpp, LK_DOWNGRADE, td);
764 return (VOP_CACHEDLOOKUP(dvp, vpp, cnp));
769 #ifndef _SYS_SYSPROTO_H_
770 struct __getcwd_args {
777 * XXX All of these sysctls would probably be more productive dead.
779 static int disablecwd;
780 SYSCTL_INT(_debug, OID_AUTO, disablecwd, CTLFLAG_RW, &disablecwd, 0,
781 "Disable the getcwd syscall");
783 /* Various statistics for the getcwd syscall */
784 static u_long numcwdcalls; STATNODE(CTLFLAG_RD, numcwdcalls, &numcwdcalls);
785 static u_long numcwdfail1; STATNODE(CTLFLAG_RD, numcwdfail1, &numcwdfail1);
786 static u_long numcwdfail2; STATNODE(CTLFLAG_RD, numcwdfail2, &numcwdfail2);
787 static u_long numcwdfail3; STATNODE(CTLFLAG_RD, numcwdfail3, &numcwdfail3);
788 static u_long numcwdfail4; STATNODE(CTLFLAG_RD, numcwdfail4, &numcwdfail4);
789 static u_long numcwdfound; STATNODE(CTLFLAG_RD, numcwdfound, &numcwdfound);
791 /* Implementation of the getcwd syscall */
795 struct __getcwd_args *uap;
798 return (kern___getcwd(td, uap->buf, UIO_USERSPACE, uap->buflen));
802 kern___getcwd(struct thread *td, u_char *buf, enum uio_seg bufseg, u_int buflen)
805 int error, i, slash_prefixed;
806 struct filedesc *fdp;
807 struct namecache *ncp;
815 if (buflen > MAXPATHLEN)
819 tmpbuf = bp = malloc(buflen, M_TEMP, M_WAITOK);
822 fdp = td->td_proc->p_fd;
825 for (vp = fdp->fd_cdir; vp != fdp->fd_rdir && vp != rootvnode;) {
826 if (vp->v_vflag & VV_ROOT) {
827 if (vp->v_mount == NULL) { /* forced unmount */
831 vp = vp->v_mount->mnt_vnodecovered;
834 if (vp->v_dd->v_id != vp->v_ddid) {
840 ncp = TAILQ_FIRST(&vp->v_cache_dst);
847 if (ncp->nc_dvp != vp->v_dd) {
853 for (i = ncp->nc_nlen - 1; i >= 0; i--) {
860 *--bp = ncp->nc_name[i];
873 if (!slash_prefixed) {
881 FILEDESC_UNLOCK(fdp);
884 if (bufseg == UIO_SYSSPACE)
885 bcopy(bp, buf, strlen(bp) + 1);
887 error = copyout(bp, buf, strlen(bp) + 1);
888 free(tmpbuf, M_TEMP);
891 FILEDESC_UNLOCK(fdp);
893 free(tmpbuf, M_TEMP);
898 * Thus begins the fullpath magic.
902 #define STATNODE(name) \
904 SYSCTL_UINT(_vfs_cache, OID_AUTO, name, CTLFLAG_RD, &name, 0, "")
906 static int disablefullpath;
907 SYSCTL_INT(_debug, OID_AUTO, disablefullpath, CTLFLAG_RW, &disablefullpath, 0,
908 "Disable the vn_fullpath function");
910 STATNODE(numfullpathcalls);
911 STATNODE(numfullpathfail1);
912 STATNODE(numfullpathfail2);
913 STATNODE(numfullpathfail3);
914 STATNODE(numfullpathfail4);
915 STATNODE(numfullpathfound);
918 * Retrieve the full filesystem path that correspond to a vnode from the name
919 * cache (if available)
922 vn_fullpath(struct thread *td, struct vnode *vn, char **retbuf, char **freebuf)
925 int i, slash_prefixed;
926 struct filedesc *fdp;
927 struct namecache *ncp;
935 buf = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
936 bp = buf + MAXPATHLEN - 1;
938 fdp = td->td_proc->p_fd;
940 ASSERT_VOP_LOCKED(vn, "vn_fullpath");
942 for (vp = vn; vp != fdp->fd_rdir && vp != rootvnode;) {
943 if (vp->v_vflag & VV_ROOT) {
944 if (vp->v_mount == NULL) { /* forced unmount */
945 FILEDESC_UNLOCK(fdp);
949 vp = vp->v_mount->mnt_vnodecovered;
952 if (vp != vn && vp->v_dd->v_id != vp->v_ddid) {
953 FILEDESC_UNLOCK(fdp);
959 ncp = TAILQ_FIRST(&vp->v_cache_dst);
963 FILEDESC_UNLOCK(fdp);
967 if (vp != vn && ncp->nc_dvp != vp->v_dd) {
970 FILEDESC_UNLOCK(fdp);
974 for (i = ncp->nc_nlen - 1; i >= 0; i--) {
978 FILEDESC_UNLOCK(fdp);
982 *--bp = ncp->nc_name[i];
987 FILEDESC_UNLOCK(fdp);
996 if (!slash_prefixed) {
999 FILEDESC_UNLOCK(fdp);
1005 FILEDESC_UNLOCK(fdp);