xref: /netbsd-src/sys/kern/vfs_cache.c (revision 7cc2f76925f078d01ddc9e640a98f4ccfc9f8c3b)
1 /*	$NetBSD: vfs_cache.c,v 1.28 2000/11/24 07:25:51 chs Exp $	*/
2 
3 /*
4  * Copyright (c) 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. All advertising materials mentioning features or use of this software
16  *    must display the following acknowledgement:
17  *	This product includes software developed by the University of
18  *	California, Berkeley and its contributors.
19  * 4. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  *	@(#)vfs_cache.c	8.3 (Berkeley) 8/22/94
36  */
37 
38 #include "opt_ddb.h"
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/time.h>
43 #include <sys/mount.h>
44 #include <sys/vnode.h>
45 #include <sys/namei.h>
46 #include <sys/errno.h>
47 #include <sys/malloc.h>
48 #include <sys/pool.h>
49 
50 /*
51  * Name caching works as follows:
52  *
53  * Names found by directory scans are retained in a cache
54  * for future reference.  It is managed LRU, so frequently
55  * used names will hang around.  Cache is indexed by hash value
56  * obtained from (dvp, name) where dvp refers to the directory
57  * containing name.
58  *
59  * For simplicity (and economy of storage), names longer than
60  * a maximum length of NCHNAMLEN are not cached; they occur
61  * infrequently in any case, and are almost never of interest.
62  *
63  * Upon reaching the last segment of a path, if the reference
64  * is for DELETE, or NOCACHE is set (rewrite), and the
65  * name is located in the cache, it will be dropped.
66  * The entry is dropped also when it was not possible to lock
67  * the cached vnode, either because vget() failed or the generation
68  * number has changed while waiting for the lock.
69  */
70 
71 /*
72  * Structures associated with name cacheing.
73  */
74 LIST_HEAD(nchashhead, namecache) *nchashtbl;
75 u_long	nchash;				/* size of hash table - 1 */
76 long	numcache;			/* number of cache entries allocated */
77 
78 LIST_HEAD(ncvhashhead, namecache) *ncvhashtbl;
79 u_long	ncvhash;			/* size of hash table - 1 */
80 
81 TAILQ_HEAD(, namecache) nclruhead;		/* LRU chain */
82 struct	nchstats nchstats;		/* cache effectiveness statistics */
83 
84 struct pool namecache_pool;
85 
86 int doingcache = 1;			/* 1 => enable the cache */
87 
88 /*
89  * Look for a the name in the cache. We don't do this
90  * if the segment name is long, simply so the cache can avoid
91  * holding long names (which would either waste space, or
92  * add greatly to the complexity).
93  *
94  * Lookup is called with ni_dvp pointing to the directory to search,
95  * ni_ptr pointing to the name of the entry being sought, ni_namelen
96  * tells the length of the name, and ni_hash contains a hash of
97  * the name. If the lookup succeeds, the vnode is locked, stored in ni_vp
98  * and a status of zero is returned. If the locking fails for whatever
99  * reason, the vnode is unlocked and the error is returned to caller.
100  * If the lookup determines that the name does not exist (negative cacheing),
101  * a status of ENOENT is returned. If the lookup fails, a status of -1
102  * is returned.
103  */
104 int
105 cache_lookup(dvp, vpp, cnp)
106 	struct vnode *dvp;
107 	struct vnode **vpp;
108 	struct componentname *cnp;
109 {
110 	struct namecache *ncp;
111 	struct nchashhead *ncpp;
112 	struct vnode *vp;
113 	int vpid, error;
114 
115 	if (!doingcache) {
116 		cnp->cn_flags &= ~MAKEENTRY;
117 		*vpp = 0;
118 		return (-1);
119 	}
120 	if (cnp->cn_namelen > NCHNAMLEN) {
121 		nchstats.ncs_long++;
122 		cnp->cn_flags &= ~MAKEENTRY;
123 		*vpp = 0;
124 		return (-1);
125 	}
126 	ncpp = &nchashtbl[(cnp->cn_hash ^ dvp->v_id) & nchash];
127 	LIST_FOREACH(ncp, ncpp, nc_hash) {
128 		if (ncp->nc_dvp == dvp &&
129 		    ncp->nc_dvpid == dvp->v_id &&
130 		    ncp->nc_nlen == cnp->cn_namelen &&
131 		    !memcmp(ncp->nc_name, cnp->cn_nameptr, (u_int)ncp->nc_nlen))
132 			break;
133 	}
134 	if (ncp == 0) {
135 		nchstats.ncs_miss++;
136 		*vpp = 0;
137 		return (-1);
138 	}
139 	if ((cnp->cn_flags & MAKEENTRY) == 0) {
140 		nchstats.ncs_badhits++;
141 		goto remove;
142 	} else if (ncp->nc_vp == NULL) {
143 		/*
144 		 * Restore the ISWHITEOUT flag saved earlier.
145 		 */
146 		cnp->cn_flags |= ncp->nc_vpid;
147 		if (cnp->cn_nameiop != CREATE ||
148 		    (cnp->cn_flags & ISLASTCN) == 0) {
149 			nchstats.ncs_neghits++;
150 			/*
151 			 * Move this slot to end of LRU chain,
152 			 * if not already there.
153 			 */
154 			if (ncp->nc_lru.tqe_next != 0) {
155 				TAILQ_REMOVE(&nclruhead, ncp, nc_lru);
156 				TAILQ_INSERT_TAIL(&nclruhead, ncp, nc_lru);
157 			}
158 			return (ENOENT);
159 		} else {
160 			nchstats.ncs_badhits++;
161 			goto remove;
162 		}
163 	} else if (ncp->nc_vpid != ncp->nc_vp->v_id) {
164 		nchstats.ncs_falsehits++;
165 		goto remove;
166 	}
167 
168 	vp = ncp->nc_vp;
169 	vpid = vp->v_id;
170 	if (vp == dvp) {	/* lookup on "." */
171 		VREF(dvp);
172 		error = 0;
173 	} else if (cnp->cn_flags & ISDOTDOT) {
174 		VOP_UNLOCK(dvp, 0);
175 		cnp->cn_flags |= PDIRUNLOCK;
176 		error = vget(vp, LK_EXCLUSIVE);
177 		/* if the above vget() succeeded and both LOCKPARENT and
178 		 * ISLASTCN is set, lock the directory vnode as well */
179 		if (!error && (~cnp->cn_flags & (LOCKPARENT|ISLASTCN)) == 0) {
180 			if ((error = vn_lock(dvp, LK_EXCLUSIVE)) != 0) {
181 				vput(vp);
182 				return (error);
183 			}
184 			cnp->cn_flags &= ~PDIRUNLOCK;
185 		}
186 	} else {
187 		error = vget(vp, LK_EXCLUSIVE);
188 		/* if the above vget() failed or either of LOCKPARENT or
189 		 * ISLASTCN is set, unlock the directory vnode */
190 		if (error || (~cnp->cn_flags & (LOCKPARENT|ISLASTCN)) != 0) {
191 			VOP_UNLOCK(dvp, 0);
192 			cnp->cn_flags |= PDIRUNLOCK;
193 		}
194 	}
195 
196 	/*
197 	 * Check that the lock succeeded, and that the capability number did
198 	 * not change while we were waiting for the lock.
199 	 */
200 	if (error || vpid != vp->v_id) {
201 		if (!error) {
202 			vput(vp);
203 			nchstats.ncs_falsehits++;
204 		} else
205 			nchstats.ncs_badhits++;
206 		/*
207 		 * The parent needs to be locked when we return to VOP_LOOKUP().
208 		 * The `.' case here should be extremely rare (if it can happen
209 		 * at all), so we don't bother optimizing out the unlock/relock.
210 		 */
211 		if (vp == dvp ||
212 		    error || (~cnp->cn_flags & (LOCKPARENT|ISLASTCN)) != 0) {
213 			if ((error = vn_lock(dvp, LK_EXCLUSIVE)) != 0)
214 				return (error);
215 			cnp->cn_flags &= ~PDIRUNLOCK;
216 		}
217 		goto remove;
218 	}
219 
220 	nchstats.ncs_goodhits++;
221 	/*
222 	 * move this slot to end of LRU chain, if not already there
223 	 */
224 	if (ncp->nc_lru.tqe_next != 0) {
225 		TAILQ_REMOVE(&nclruhead, ncp, nc_lru);
226 		TAILQ_INSERT_TAIL(&nclruhead, ncp, nc_lru);
227 	}
228 	*vpp = vp;
229 	return (0);
230 
231 remove:
232 	/*
233 	 * Last component and we are renaming or deleting,
234 	 * the cache entry is invalid, or otherwise don't
235 	 * want cache entry to exist.
236 	 */
237 	TAILQ_REMOVE(&nclruhead, ncp, nc_lru);
238 	LIST_REMOVE(ncp, nc_hash);
239 	ncp->nc_hash.le_prev = 0;
240 	if (ncp->nc_vhash.le_prev != NULL) {
241 		LIST_REMOVE(ncp, nc_vhash);
242 		ncp->nc_vhash.le_prev = 0;
243 	}
244 	TAILQ_INSERT_HEAD(&nclruhead, ncp, nc_lru);
245 	*vpp = 0;
246 	return (-1);
247 }
248 
249 /*
250  * Scan cache looking for name of directory entry pointing at vp.
251  *
252  * Fill in dvpp.
253  *
254  * If bufp is non-NULL, also place the name in the buffer which starts
255  * at bufp, immediately before *bpp, and move bpp backwards to point
256  * at the start of it.  (Yes, this is a little baroque, but it's done
257  * this way to cater to the whims of getcwd).
258  *
259  * Returns 0 on success, -1 on cache miss, positive errno on failure.
260  */
261 int
262 cache_revlookup (vp, dvpp, bpp, bufp)
263 	struct vnode *vp, **dvpp;
264 	char **bpp;
265 	char *bufp;
266 {
267 	struct namecache *ncp;
268 	struct vnode *dvp;
269 	struct ncvhashhead *nvcpp;
270 
271 	if (!doingcache)
272 		goto out;
273 
274 	nvcpp = &ncvhashtbl[(vp->v_id & ncvhash)];
275 
276 	LIST_FOREACH(ncp, nvcpp, nc_vhash) {
277 		if ((ncp->nc_vp == vp) &&
278 		    (ncp->nc_vpid == vp->v_id) &&
279 		    ((dvp = ncp->nc_dvp) != 0) &&
280 		    (dvp != vp) && 		/* avoid pesky . entries.. */
281 		    (dvp->v_id == ncp->nc_dvpid))
282 		{
283 			char *bp;
284 
285 #ifdef DIAGNOSTIC
286 			if ((ncp->nc_nlen == 1) &&
287 			    (ncp->nc_name[0] == '.'))
288 				panic("cache_revlookup: found entry for .");
289 
290 			if ((ncp->nc_nlen == 2) &&
291 			    (ncp->nc_name[0] == '.') &&
292 			    (ncp->nc_name[1] == '.'))
293 				panic("cache_revlookup: found entry for ..");
294 #endif
295 			nchstats.ncs_revhits++;
296 
297 			if (bufp) {
298 				bp = *bpp;
299 				bp -= ncp->nc_nlen;
300 				if (bp <= bufp) {
301 					*dvpp = 0;
302 					return ERANGE;
303 				}
304 				memcpy(bp, ncp->nc_name, ncp->nc_nlen);
305 				*bpp = bp;
306 			}
307 
308 			/* XXX MP: how do we know dvp won't evaporate? */
309 			*dvpp = dvp;
310 			return 0;
311 		}
312 	}
313 	nchstats.ncs_revmiss++;
314  out:
315 	*dvpp = 0;
316 	return -1;
317 }
318 
319 /*
320  * Add an entry to the cache
321  */
322 void
323 cache_enter(dvp, vp, cnp)
324 	struct vnode *dvp;
325 	struct vnode *vp;
326 	struct componentname *cnp;
327 {
328 	struct namecache *ncp;
329 	struct nchashhead *ncpp;
330 	struct ncvhashhead *nvcpp;
331 
332 #ifdef DIAGNOSTIC
333 	if (cnp->cn_namelen > NCHNAMLEN)
334 		panic("cache_enter: name too long");
335 #endif
336 	if (!doingcache)
337 		return;
338 	/*
339 	 * Free the cache slot at head of lru chain.
340 	 */
341 	if (numcache < numvnodes) {
342 		ncp = pool_get(&namecache_pool, PR_WAITOK);
343 		memset(ncp, 0, sizeof(*ncp));
344 		numcache++;
345 	} else if ((ncp = TAILQ_FIRST(&nclruhead)) != NULL) {
346 		TAILQ_REMOVE(&nclruhead, ncp, nc_lru);
347 		if (ncp->nc_hash.le_prev != 0) {
348 			LIST_REMOVE(ncp, nc_hash);
349 			ncp->nc_hash.le_prev = 0;
350 		}
351 		if (ncp->nc_vhash.le_prev != 0) {
352 			LIST_REMOVE(ncp, nc_vhash);
353 			ncp->nc_vhash.le_prev = 0;
354 		}
355 	} else
356 		return;
357 	/* grab the vnode we just found */
358 	ncp->nc_vp = vp;
359 	if (vp)
360 		ncp->nc_vpid = vp->v_id;
361 	else {
362 		/*
363 		 * For negative hits, save the ISWHITEOUT flag so we can
364 		 * restore it later when the cache entry is used again.
365 		 */
366 		ncp->nc_vpid = cnp->cn_flags & ISWHITEOUT;
367 	}
368 	/* fill in cache info */
369 	ncp->nc_dvp = dvp;
370 	ncp->nc_dvpid = dvp->v_id;
371 	ncp->nc_nlen = cnp->cn_namelen;
372 	memcpy(ncp->nc_name, cnp->cn_nameptr, (unsigned)ncp->nc_nlen);
373 	TAILQ_INSERT_TAIL(&nclruhead, ncp, nc_lru);
374 	ncpp = &nchashtbl[(cnp->cn_hash ^ dvp->v_id) & nchash];
375 	LIST_INSERT_HEAD(ncpp, ncp, nc_hash);
376 
377 	ncp->nc_vhash.le_prev = 0;
378 	ncp->nc_vhash.le_next = 0;
379 
380 	/*
381 	 * Create reverse-cache entries (used in getcwd) for directories.
382 	 */
383 	if (vp &&
384 	    (vp != dvp) &&
385 	    (vp->v_type == VDIR) &&
386 	    ((ncp->nc_nlen > 2) ||
387 	     ((ncp->nc_nlen == 2) && (ncp->nc_name[0] != '.') && (ncp->nc_name[1] != '.')) ||
388 	     ((ncp->nc_nlen == 1) && (ncp->nc_name[0] != '.'))))
389 	{
390 		nvcpp = &ncvhashtbl[(vp->v_id & ncvhash)];
391 		LIST_INSERT_HEAD(nvcpp, ncp, nc_vhash);
392 	}
393 
394 }
395 
396 /*
397  * Name cache initialization, from vfs_init() when we are booting
398  */
399 void
400 nchinit()
401 {
402 
403 	TAILQ_INIT(&nclruhead);
404 	nchashtbl =
405 	    hashinit(desiredvnodes, HASH_LIST, M_CACHE, M_WAITOK, &nchash);
406 	ncvhashtbl =
407 	    hashinit(desiredvnodes/8, HASH_LIST, M_CACHE, M_WAITOK, &ncvhash);
408 	pool_init(&namecache_pool, sizeof(struct namecache), 0, 0, 0,
409 	    "ncachepl", 0, pool_page_alloc_nointr, pool_page_free_nointr,
410 	    M_CACHE);
411 }
412 
413 /*
414  * Cache flush, a particular vnode; called when a vnode is renamed to
415  * hide entries that would now be invalid
416  */
417 void
418 cache_purge(vp)
419 	struct vnode *vp;
420 {
421 	struct namecache *ncp;
422 	struct nchashhead *ncpp;
423 	static u_long nextvnodeid;
424 
425 	vp->v_id = ++nextvnodeid;
426 	if (nextvnodeid != 0)
427 		return;
428 	for (ncpp = &nchashtbl[nchash]; ncpp >= nchashtbl; ncpp--) {
429 		LIST_FOREACH(ncp, ncpp, nc_hash) {
430 			ncp->nc_vpid = 0;
431 			ncp->nc_dvpid = 0;
432 		}
433 	}
434 	vp->v_id = ++nextvnodeid;
435 }
436 
437 /*
438  * Cache flush, a whole filesystem; called when filesys is umounted to
439  * remove entries that would now be invalid.
440  */
441 void
442 cache_purgevfs(mp)
443 	struct mount *mp;
444 {
445 	struct namecache *ncp, *nxtcp;
446 
447 	for (ncp = TAILQ_FIRST(&nclruhead); ncp != NULL; ncp = nxtcp) {
448 		nxtcp = TAILQ_NEXT(ncp, nc_lru);
449 		if (ncp->nc_dvp == NULL || ncp->nc_dvp->v_mount != mp) {
450 			continue;
451 		}
452 		/* free the resources we had */
453 		ncp->nc_vp = NULL;
454 		ncp->nc_dvp = NULL;
455 		TAILQ_REMOVE(&nclruhead, ncp, nc_lru);
456 		if (ncp->nc_hash.le_prev != 0) {
457 			LIST_REMOVE(ncp, nc_hash);
458 			ncp->nc_hash.le_prev = 0;
459 		}
460 		if (ncp->nc_vhash.le_prev != 0) {
461 			LIST_REMOVE(ncp, nc_vhash);
462 			ncp->nc_vhash.le_prev = 0;
463 		}
464 		TAILQ_INSERT_HEAD(&nclruhead, ncp, nc_lru);
465 	}
466 }
467 
468 #ifdef DDB
469 void
470 namecache_print(struct vnode *vp, void (*pr)(const char *, ...))
471 {
472 	struct vnode *dvp = NULL;
473 	struct namecache *ncp;
474 
475 	TAILQ_FOREACH(ncp, &nclruhead, nc_lru) {
476 		if (ncp->nc_vp == vp && ncp->nc_vpid == vp->v_id) {
477 			(*pr)("name %.*s\n", ncp->nc_nlen, ncp->nc_name);
478 			dvp = ncp->nc_dvp;
479 		}
480 	}
481 	if (dvp == NULL) {
482 		(*pr)("name not found\n");
483 		return;
484 	}
485 	vp = dvp;
486 	TAILQ_FOREACH(ncp, &nclruhead, nc_lru) {
487 		if (ncp->nc_vp == vp && ncp->nc_vpid == vp->v_id) {
488 			(*pr)("parent %.*s\n", ncp->nc_nlen, ncp->nc_name);
489 		}
490 	}
491 }
492 #endif
493