xref: /netbsd-src/sys/kern/vfs_cache.c (revision bdc22b2e01993381dcefeff2bc9b56ca75a4235c)
1 /*	$NetBSD: vfs_cache.c,v 1.120 2017/03/18 22:36:56 riastradh Exp $	*/
2 
3 /*-
4  * Copyright (c) 2008 The NetBSD Foundation, Inc.
5  * 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  *
16  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
17  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
18  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
19  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
20  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
21  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
22  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
23  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
24  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
25  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
26  * POSSIBILITY OF SUCH DAMAGE.
27  */
28 
29 /*
30  * Copyright (c) 1989, 1993
31  *	The Regents of the University of California.  All rights reserved.
32  *
33  * Redistribution and use in source and binary forms, with or without
34  * modification, are permitted provided that the following conditions
35  * are met:
36  * 1. Redistributions of source code must retain the above copyright
37  *    notice, this list of conditions and the following disclaimer.
38  * 2. Redistributions in binary form must reproduce the above copyright
39  *    notice, this list of conditions and the following disclaimer in the
40  *    documentation and/or other materials provided with the distribution.
41  * 3. Neither the name of the University nor the names of its contributors
42  *    may be used to endorse or promote products derived from this software
43  *    without specific prior written permission.
44  *
45  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
46  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
47  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
48  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
49  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
50  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
51  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
52  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
53  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
54  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
55  * SUCH DAMAGE.
56  *
57  *	@(#)vfs_cache.c	8.3 (Berkeley) 8/22/94
58  */
59 
60 #include <sys/cdefs.h>
61 __KERNEL_RCSID(0, "$NetBSD: vfs_cache.c,v 1.120 2017/03/18 22:36:56 riastradh Exp $");
62 
63 #ifdef _KERNEL_OPT
64 #include "opt_ddb.h"
65 #include "opt_dtrace.h"
66 #include "opt_revcache.h"
67 #endif
68 
69 #include <sys/param.h>
70 #include <sys/atomic.h>
71 #include <sys/cpu.h>
72 #include <sys/errno.h>
73 #include <sys/evcnt.h>
74 #include <sys/kernel.h>
75 #include <sys/kthread.h>
76 #include <sys/mount.h>
77 #include <sys/mutex.h>
78 #include <sys/namei.h>
79 #include <sys/pool.h>
80 #include <sys/sdt.h>
81 #include <sys/sysctl.h>
82 #include <sys/systm.h>
83 #include <sys/time.h>
84 #include <sys/vnode_impl.h>
85 
86 #define NAMECACHE_ENTER_REVERSE
87 /*
88  * Name caching works as follows:
89  *
90  * Names found by directory scans are retained in a cache
91  * for future reference.  It is managed LRU, so frequently
92  * used names will hang around.  Cache is indexed by hash value
93  * obtained from (dvp, name) where dvp refers to the directory
94  * containing name.
95  *
96  * For simplicity (and economy of storage), names longer than
97  * a maximum length of NCHNAMLEN are not cached; they occur
98  * infrequently in any case, and are almost never of interest.
99  *
100  * Upon reaching the last segment of a path, if the reference
101  * is for DELETE, or NOCACHE is set (rewrite), and the
102  * name is located in the cache, it will be dropped.
103  */
104 
105 /*
106  * Cache entry lifetime:
107  *
108  *	nonexistent
109  *	---create---> active
110  *	---invalidate---> queued
111  *	---reclaim---> nonexistent.
112  *
113  * States:
114  * - Nonexistent.  Cache entry does not exist.
115  *
116  * - Active.  cache_lookup, cache_lookup_raw, cache_revlookup can look
117  *   up, acquire references, and hand off references to vnodes,
118  *   e.g. via v_interlock.  Marked by nonnull ncp->nc_dvp.
119  *
120  * - Queued.  Pending desstruction by cache_reclaim.  Cannot be used by
121  *   cache_lookup, cache_lookup_raw, or cache_revlookup.  May still be
122  *   on lists.  Marked by null ncp->nc_dvp.
123  *
124  * Transitions:
125  *
126  * - Create: nonexistent--->active
127  *
128  *   Done by cache_enter(dvp, vp, name, namelen, cnflags), called by
129  *   VOP_LOOKUP after the answer is found.  Allocates a struct
130  *   namecache object, initializes it with the above fields, and
131  *   activates it by inserting it into the forward and reverse tables.
132  *
133  * - Invalidate: active--->queued
134  *
135  *   Done by cache_invalidate.  If not already invalidated, nullify
136  *   ncp->nc_dvp and ncp->nc_vp, and add to cache_gcqueue.  Called,
137  *   among various other places, in cache_lookup(dvp, name, namelen,
138  *   nameiop, cnflags, &iswht, &vp) when MAKEENTRY is missing from
139  *   cnflags.
140  *
141  * - Reclaim: queued--->nonexistent
142  *
143  *   Done by cache_reclaim.  Disassociate ncp from any lists it is on
144  *   and free memory.
145  */
146 
147 /*
148  * Locking.
149  *
150  * L namecache_lock		Global lock for namecache table and queues.
151  * C struct nchcpu::cpu_lock	Per-CPU lock to reduce read contention.
152  * N struct namecache::nc_lock	Per-entry lock.
153  * V struct vnode::v_interlock	Vnode interlock.
154  *
155  * Lock order: L -> C -> N -> V
156  *
157  *	Examples:
158  *	. L->C: cache_reclaim
159  *	. C->N->V: cache_lookup
160  *	. L->N->V: cache_purge1, cache_revlookup
161  *
162  * All use serialized by namecache_lock:
163  *
164  *	nclruhead / struct namecache::nc_lru
165  *	ncvhashtbl / struct namecache::nc_vhash
166  *	struct vnode_impl::vi_dnclist / struct namecache::nc_dvlist
167  *	struct vnode_impl::vi_nclist / struct namecache::nc_vlist
168  *	nchstats
169  *
170  * - Insertion serialized by namecache_lock,
171  * - read protected by per-CPU lock,
172  * - insert/read ordering guaranteed by memory barriers, and
173  * - deletion allowed only under namecache_lock and *all* per-CPU locks
174  *   in CPU_INFO_FOREACH order:
175  *
176  *	nchashtbl / struct namecache::nc_hash
177  *
178  *   The per-CPU locks exist only to reduce the probability of
179  *   contention between readers.  We do not bind to a CPU, so
180  *   contention is still possible.
181  *
182  * All use serialized by struct namecache::nc_lock:
183  *
184  *	struct namecache::nc_dvp
185  *	struct namecache::nc_vp
186  *	struct namecache::nc_gcqueue (*)
187  *	struct namecache::nc_hittime (**)
188  *
189  * (*) Once on the queue, only cache_thread uses this nc_gcqueue, unlocked.
190  * (**) cache_prune reads nc_hittime unlocked, since approximate is OK.
191  *
192  * Unlocked because stable after initialization:
193  *
194  *	struct namecache::nc_dvp
195  *	struct namecache::nc_vp
196  *	struct namecache::nc_flags
197  *	struct namecache::nc_nlen
198  *	struct namecache::nc_name
199  *
200  * Unlocked because approximation is OK:
201  *
202  *	struct nchcpu::cpu_stats
203  *	struct nchcpu::cpu_stats_last
204  *
205  * Updates under namecache_lock or any per-CPU lock are marked with
206  * COUNT, while updates outside those locks are marked with COUNT_UNL.
207  *
208  * - The theory seems to have been that you could replace COUNT_UNL by
209  *   atomic operations -- except that doesn't help unless you also
210  *   replace COUNT by atomic operations, because mixing atomics and
211  *   nonatomics is a recipe for failure.
212  * - We use 32-bit per-CPU counters and 64-bit global counters under
213  *   the theory that 32-bit counters are less likely to be hosed by
214  *   nonatomic increment.
215  */
216 
217 /*
218  * The comment below is preserved for posterity in case it is
219  * important, but it is clear that everywhere the namecache_count_*()
220  * functions are called, other cache_*() functions that take the same
221  * locks are also called, so I can't imagine how this could be a
222  * problem:
223  *
224  * N.B.: Attempting to protect COUNT_UNL() increments by taking
225  * a per-cpu lock in the namecache_count_*() functions causes
226  * a deadlock.  Don't do that, use atomic increments instead if
227  * the imperfections here bug you.
228  */
229 
230 /*
231  * struct nchstats_percpu:
232  *
233  *	Per-CPU counters.
234  */
235 struct nchstats_percpu _NAMEI_CACHE_STATS(uint32_t);
236 
237 /*
238  * struct nchcpu:
239  *
240  *	Per-CPU namecache state: lock and per-CPU counters.
241  */
242 struct nchcpu {
243 	kmutex_t		cpu_lock;
244 	struct nchstats_percpu	cpu_stats;
245 	/* XXX maybe __cacheline_aligned would improve this? */
246 	struct nchstats_percpu	cpu_stats_last;	/* from last sample */
247 };
248 
249 /*
250  * The type for the hash code. While the hash function generates a
251  * u32, the hash code has historically been passed around as a u_long,
252  * and the value is modified by xor'ing a uintptr_t, so it's not
253  * entirely clear what the best type is. For now I'll leave it
254  * unchanged as u_long.
255  */
256 
257 typedef u_long nchash_t;
258 
259 /*
260  * Structures associated with name cacheing.
261  */
262 
263 static kmutex_t *namecache_lock __read_mostly;
264 static pool_cache_t namecache_cache __read_mostly;
265 static TAILQ_HEAD(, namecache) nclruhead __cacheline_aligned;
266 
267 static LIST_HEAD(nchashhead, namecache) *nchashtbl __read_mostly;
268 static u_long	nchash __read_mostly;
269 
270 #define	NCHASH2(hash, dvp)	\
271 	(((hash) ^ ((uintptr_t)(dvp) >> 3)) & nchash)
272 
273 static LIST_HEAD(ncvhashhead, namecache) *ncvhashtbl __read_mostly;
274 static u_long	ncvhash __read_mostly;
275 
276 #define	NCVHASH(vp)		(((uintptr_t)(vp) >> 3) & ncvhash)
277 
278 /* Number of cache entries allocated. */
279 static long	numcache __cacheline_aligned;
280 
281 /* Garbage collection queue and number of entries pending in it. */
282 static void	*cache_gcqueue;
283 static u_int	cache_gcpend;
284 
285 /* Cache effectiveness statistics.  This holds total from per-cpu stats */
286 struct nchstats	nchstats __cacheline_aligned;
287 
288 /*
289  * Macros to count an event, update the central stats with per-cpu
290  * values and add current per-cpu increments to the subsystem total
291  * last collected by cache_reclaim().
292  */
293 #define	CACHE_STATS_CURRENT	/* nothing */
294 
295 #define	COUNT(cpup, f)	((cpup)->cpu_stats.f++)
296 
297 #define	UPDATE(cpup, f) do { \
298 	struct nchcpu *Xcpup = (cpup); \
299 	uint32_t Xcnt = (volatile uint32_t) Xcpup->cpu_stats.f; \
300 	nchstats.f += Xcnt - Xcpup->cpu_stats_last.f; \
301 	Xcpup->cpu_stats_last.f = Xcnt; \
302 } while (/* CONSTCOND */ 0)
303 
304 #define	ADD(stats, cpup, f) do { \
305 	struct nchcpu *Xcpup = (cpup); \
306 	stats.f += Xcpup->cpu_stats.f - Xcpup->cpu_stats_last.f; \
307 } while (/* CONSTCOND */ 0)
308 
309 /* Do unlocked stats the same way. Use a different name to allow mind changes */
310 #define	COUNT_UNL(cpup, f)	COUNT((cpup), f)
311 
312 static const int cache_lowat = 95;
313 static const int cache_hiwat = 98;
314 static const int cache_hottime = 5;	/* number of seconds */
315 static int doingcache = 1;		/* 1 => enable the cache */
316 
317 static struct evcnt cache_ev_scan;
318 static struct evcnt cache_ev_gc;
319 static struct evcnt cache_ev_over;
320 static struct evcnt cache_ev_under;
321 static struct evcnt cache_ev_forced;
322 
323 static struct namecache *cache_lookup_entry(
324     const struct vnode *, const char *, size_t);
325 static void cache_thread(void *);
326 static void cache_invalidate(struct namecache *);
327 static void cache_disassociate(struct namecache *);
328 static void cache_reclaim(void);
329 static int cache_ctor(void *, void *, int);
330 static void cache_dtor(void *, void *);
331 
332 static struct sysctllog *sysctllog;
333 static void sysctl_cache_stat_setup(void);
334 
335 SDT_PROVIDER_DEFINE(vfs);
336 
337 SDT_PROBE_DEFINE1(vfs, namecache, invalidate, done, "struct vnode *");
338 SDT_PROBE_DEFINE1(vfs, namecache, purge, parents, "struct vnode *");
339 SDT_PROBE_DEFINE1(vfs, namecache, purge, children, "struct vnode *");
340 SDT_PROBE_DEFINE2(vfs, namecache, purge, name, "char *", "size_t");
341 SDT_PROBE_DEFINE1(vfs, namecache, purge, vfs, "struct mount *");
342 SDT_PROBE_DEFINE3(vfs, namecache, lookup, hit, "struct vnode *",
343     "char *", "size_t");
344 SDT_PROBE_DEFINE3(vfs, namecache, lookup, miss, "struct vnode *",
345     "char *", "size_t");
346 SDT_PROBE_DEFINE3(vfs, namecache, lookup, toolong, "struct vnode *",
347     "char *", "size_t");
348 SDT_PROBE_DEFINE2(vfs, namecache, revlookup, success, "struct vnode *",
349      "struct vnode *");
350 SDT_PROBE_DEFINE2(vfs, namecache, revlookup, fail, "struct vnode *",
351      "int");
352 SDT_PROBE_DEFINE2(vfs, namecache, prune, done, "int", "int");
353 SDT_PROBE_DEFINE3(vfs, namecache, enter, toolong, "struct vnode *",
354     "char *", "size_t");
355 SDT_PROBE_DEFINE3(vfs, namecache, enter, done, "struct vnode *",
356     "char *", "size_t");
357 
358 /*
359  * Compute the hash for an entry.
360  *
361  * (This is for now a wrapper around namei_hash, whose interface is
362  * for the time being slightly inconvenient.)
363  */
364 static nchash_t
365 cache_hash(const char *name, size_t namelen)
366 {
367 	const char *endptr;
368 
369 	endptr = name + namelen;
370 	return namei_hash(name, &endptr);
371 }
372 
373 /*
374  * Invalidate a cache entry and enqueue it for garbage collection.
375  * The caller needs to hold namecache_lock or a per-cpu lock to hold
376  * off cache_reclaim().
377  */
378 static void
379 cache_invalidate(struct namecache *ncp)
380 {
381 	void *head;
382 
383 	KASSERT(mutex_owned(&ncp->nc_lock));
384 
385 	if (ncp->nc_dvp != NULL) {
386 		SDT_PROBE(vfs, namecache, invalidate, done, ncp->nc_dvp,
387 		    0, 0, 0, 0);
388 
389 		ncp->nc_vp = NULL;
390 		ncp->nc_dvp = NULL;
391 		do {
392 			head = cache_gcqueue;
393 			ncp->nc_gcqueue = head;
394 		} while (atomic_cas_ptr(&cache_gcqueue, head, ncp) != head);
395 		atomic_inc_uint(&cache_gcpend);
396 	}
397 }
398 
399 /*
400  * Disassociate a namecache entry from any vnodes it is attached to,
401  * and remove from the global LRU list.
402  */
403 static void
404 cache_disassociate(struct namecache *ncp)
405 {
406 
407 	KASSERT(mutex_owned(namecache_lock));
408 	KASSERT(ncp->nc_dvp == NULL);
409 
410 	if (ncp->nc_lru.tqe_prev != NULL) {
411 		TAILQ_REMOVE(&nclruhead, ncp, nc_lru);
412 		ncp->nc_lru.tqe_prev = NULL;
413 	}
414 	if (ncp->nc_vhash.le_prev != NULL) {
415 		LIST_REMOVE(ncp, nc_vhash);
416 		ncp->nc_vhash.le_prev = NULL;
417 	}
418 	if (ncp->nc_vlist.le_prev != NULL) {
419 		LIST_REMOVE(ncp, nc_vlist);
420 		ncp->nc_vlist.le_prev = NULL;
421 	}
422 	if (ncp->nc_dvlist.le_prev != NULL) {
423 		LIST_REMOVE(ncp, nc_dvlist);
424 		ncp->nc_dvlist.le_prev = NULL;
425 	}
426 }
427 
428 /*
429  * Lock all CPUs to prevent any cache lookup activity.  Conceptually,
430  * this locks out all "readers".
431  */
432 static void
433 cache_lock_cpus(void)
434 {
435 	CPU_INFO_ITERATOR cii;
436 	struct cpu_info *ci;
437 	struct nchcpu *cpup;
438 
439 	/*
440 	 * Lock out all CPUs first, then harvest per-cpu stats.  This
441 	 * is probably not quite as cache-efficient as doing the lock
442 	 * and harvest at the same time, but allows cache_stat_sysctl()
443 	 * to make do with a per-cpu lock.
444 	 */
445 	for (CPU_INFO_FOREACH(cii, ci)) {
446 		cpup = ci->ci_data.cpu_nch;
447 		mutex_enter(&cpup->cpu_lock);
448 	}
449 	for (CPU_INFO_FOREACH(cii, ci)) {
450 		cpup = ci->ci_data.cpu_nch;
451 		UPDATE(cpup, ncs_goodhits);
452 		UPDATE(cpup, ncs_neghits);
453 		UPDATE(cpup, ncs_badhits);
454 		UPDATE(cpup, ncs_falsehits);
455 		UPDATE(cpup, ncs_miss);
456 		UPDATE(cpup, ncs_long);
457 		UPDATE(cpup, ncs_pass2);
458 		UPDATE(cpup, ncs_2passes);
459 		UPDATE(cpup, ncs_revhits);
460 		UPDATE(cpup, ncs_revmiss);
461 	}
462 }
463 
464 /*
465  * Release all CPU locks.
466  */
467 static void
468 cache_unlock_cpus(void)
469 {
470 	CPU_INFO_ITERATOR cii;
471 	struct cpu_info *ci;
472 	struct nchcpu *cpup;
473 
474 	for (CPU_INFO_FOREACH(cii, ci)) {
475 		cpup = ci->ci_data.cpu_nch;
476 		mutex_exit(&cpup->cpu_lock);
477 	}
478 }
479 
480 /*
481  * Find a single cache entry and return it locked.
482  * The caller needs to hold namecache_lock or a per-cpu lock to hold
483  * off cache_reclaim().
484  */
485 static struct namecache *
486 cache_lookup_entry(const struct vnode *dvp, const char *name, size_t namelen)
487 {
488 	struct nchashhead *ncpp;
489 	struct namecache *ncp;
490 	nchash_t hash;
491 
492 	KASSERT(dvp != NULL);
493 	hash = cache_hash(name, namelen);
494 	ncpp = &nchashtbl[NCHASH2(hash, dvp)];
495 
496 	LIST_FOREACH(ncp, ncpp, nc_hash) {
497 		membar_datadep_consumer();	/* for Alpha... */
498 		if (ncp->nc_dvp != dvp ||
499 		    ncp->nc_nlen != namelen ||
500 		    memcmp(ncp->nc_name, name, (u_int)ncp->nc_nlen))
501 		    	continue;
502 	    	mutex_enter(&ncp->nc_lock);
503 		if (__predict_true(ncp->nc_dvp == dvp)) {
504 			ncp->nc_hittime = hardclock_ticks;
505 			SDT_PROBE(vfs, namecache, lookup, hit, dvp,
506 			    name, namelen, 0, 0);
507 			return ncp;
508 		}
509 		/* Raced: entry has been nullified. */
510 		mutex_exit(&ncp->nc_lock);
511 	}
512 
513 	SDT_PROBE(vfs, namecache, lookup, miss, dvp,
514 	    name, namelen, 0, 0);
515 	return NULL;
516 }
517 
518 /*
519  * Look for a the name in the cache. We don't do this
520  * if the segment name is long, simply so the cache can avoid
521  * holding long names (which would either waste space, or
522  * add greatly to the complexity).
523  *
524  * Lookup is called with DVP pointing to the directory to search,
525  * and CNP providing the name of the entry being sought: cn_nameptr
526  * is the name, cn_namelen is its length, and cn_flags is the flags
527  * word from the namei operation.
528  *
529  * DVP must be locked.
530  *
531  * There are three possible non-error return states:
532  *    1. Nothing was found in the cache. Nothing is known about
533  *       the requested name.
534  *    2. A negative entry was found in the cache, meaning that the
535  *       requested name definitely does not exist.
536  *    3. A positive entry was found in the cache, meaning that the
537  *       requested name does exist and that we are providing the
538  *       vnode.
539  * In these cases the results are:
540  *    1. 0 returned; VN is set to NULL.
541  *    2. 1 returned; VN is set to NULL.
542  *    3. 1 returned; VN is set to the vnode found.
543  *
544  * The additional result argument ISWHT is set to zero, unless a
545  * negative entry is found that was entered as a whiteout, in which
546  * case ISWHT is set to one.
547  *
548  * The ISWHT_RET argument pointer may be null. In this case an
549  * assertion is made that the whiteout flag is not set. File systems
550  * that do not support whiteouts can/should do this.
551  *
552  * Filesystems that do support whiteouts should add ISWHITEOUT to
553  * cnp->cn_flags if ISWHT comes back nonzero.
554  *
555  * When a vnode is returned, it is locked, as per the vnode lookup
556  * locking protocol.
557  *
558  * There is no way for this function to fail, in the sense of
559  * generating an error that requires aborting the namei operation.
560  *
561  * (Prior to October 2012, this function returned an integer status,
562  * and a vnode, and mucked with the flags word in CNP for whiteouts.
563  * The integer status was -1 for "nothing found", ENOENT for "a
564  * negative entry found", 0 for "a positive entry found", and possibly
565  * other errors, and the value of VN might or might not have been set
566  * depending on what error occurred.)
567  */
568 bool
569 cache_lookup(struct vnode *dvp, const char *name, size_t namelen,
570 	     uint32_t nameiop, uint32_t cnflags,
571 	     int *iswht_ret, struct vnode **vn_ret)
572 {
573 	struct namecache *ncp;
574 	struct vnode *vp;
575 	struct nchcpu *cpup;
576 	int error;
577 	bool hit;
578 
579 
580 	/* Establish default result values */
581 	if (iswht_ret != NULL) {
582 		*iswht_ret = 0;
583 	}
584 	*vn_ret = NULL;
585 
586 	if (__predict_false(!doingcache)) {
587 		return false;
588 	}
589 
590 	cpup = curcpu()->ci_data.cpu_nch;
591 	mutex_enter(&cpup->cpu_lock);
592 	if (__predict_false(namelen > NCHNAMLEN)) {
593 		SDT_PROBE(vfs, namecache, lookup, toolong, dvp,
594 		    name, namelen, 0, 0);
595 		COUNT(cpup, ncs_long);
596 		mutex_exit(&cpup->cpu_lock);
597 		/* found nothing */
598 		return false;
599 	}
600 
601 	ncp = cache_lookup_entry(dvp, name, namelen);
602 	if (__predict_false(ncp == NULL)) {
603 		COUNT(cpup, ncs_miss);
604 		mutex_exit(&cpup->cpu_lock);
605 		/* found nothing */
606 		return false;
607 	}
608 	if ((cnflags & MAKEENTRY) == 0) {
609 		COUNT(cpup, ncs_badhits);
610 		/*
611 		 * Last component and we are renaming or deleting,
612 		 * the cache entry is invalid, or otherwise don't
613 		 * want cache entry to exist.
614 		 */
615 		cache_invalidate(ncp);
616 		mutex_exit(&ncp->nc_lock);
617 		mutex_exit(&cpup->cpu_lock);
618 		/* found nothing */
619 		return false;
620 	}
621 	if (ncp->nc_vp == NULL) {
622 		if (iswht_ret != NULL) {
623 			/*
624 			 * Restore the ISWHITEOUT flag saved earlier.
625 			 */
626 			KASSERT((ncp->nc_flags & ~ISWHITEOUT) == 0);
627 			*iswht_ret = (ncp->nc_flags & ISWHITEOUT) != 0;
628 		} else {
629 			KASSERT(ncp->nc_flags == 0);
630 		}
631 
632 		if (__predict_true(nameiop != CREATE ||
633 		    (cnflags & ISLASTCN) == 0)) {
634 			COUNT(cpup, ncs_neghits);
635 			/* found neg entry; vn is already null from above */
636 			hit = true;
637 		} else {
638 			COUNT(cpup, ncs_badhits);
639 			/*
640 			 * Last component and we are preparing to create
641 			 * the named object, so flush the negative cache
642 			 * entry.
643 			 */
644 			cache_invalidate(ncp);
645 			/* found nothing */
646 			hit = false;
647 		}
648 		mutex_exit(&ncp->nc_lock);
649 		mutex_exit(&cpup->cpu_lock);
650 		return hit;
651 	}
652 
653 	vp = ncp->nc_vp;
654 	mutex_enter(vp->v_interlock);
655 	mutex_exit(&ncp->nc_lock);
656 	mutex_exit(&cpup->cpu_lock);
657 
658 	/*
659 	 * Unlocked except for the vnode interlock.  Call vcache_tryvget().
660 	 */
661 	error = vcache_tryvget(vp);
662 	if (error) {
663 		KASSERT(error == EBUSY);
664 		/*
665 		 * This vnode is being cleaned out.
666 		 * XXX badhits?
667 		 */
668 		COUNT_UNL(cpup, ncs_falsehits);
669 		/* found nothing */
670 		return false;
671 	}
672 
673 	COUNT_UNL(cpup, ncs_goodhits);
674 	/* found it */
675 	*vn_ret = vp;
676 	return true;
677 }
678 
679 
680 /*
681  * Cut-'n-pasted version of the above without the nameiop argument.
682  */
683 bool
684 cache_lookup_raw(struct vnode *dvp, const char *name, size_t namelen,
685 		 uint32_t cnflags,
686 		 int *iswht_ret, struct vnode **vn_ret)
687 {
688 	struct namecache *ncp;
689 	struct vnode *vp;
690 	struct nchcpu *cpup;
691 	int error;
692 
693 	/* Establish default results. */
694 	if (iswht_ret != NULL) {
695 		*iswht_ret = 0;
696 	}
697 	*vn_ret = NULL;
698 
699 	if (__predict_false(!doingcache)) {
700 		/* found nothing */
701 		return false;
702 	}
703 
704 	cpup = curcpu()->ci_data.cpu_nch;
705 	mutex_enter(&cpup->cpu_lock);
706 	if (__predict_false(namelen > NCHNAMLEN)) {
707 		COUNT(cpup, ncs_long);
708 		mutex_exit(&cpup->cpu_lock);
709 		/* found nothing */
710 		return false;
711 	}
712 	ncp = cache_lookup_entry(dvp, name, namelen);
713 	if (__predict_false(ncp == NULL)) {
714 		COUNT(cpup, ncs_miss);
715 		mutex_exit(&cpup->cpu_lock);
716 		/* found nothing */
717 		return false;
718 	}
719 	vp = ncp->nc_vp;
720 	if (vp == NULL) {
721 		/*
722 		 * Restore the ISWHITEOUT flag saved earlier.
723 		 */
724 		if (iswht_ret != NULL) {
725 			KASSERT((ncp->nc_flags & ~ISWHITEOUT) == 0);
726 			/*cnp->cn_flags |= ncp->nc_flags;*/
727 			*iswht_ret = (ncp->nc_flags & ISWHITEOUT) != 0;
728 		}
729 		COUNT(cpup, ncs_neghits);
730 		mutex_exit(&ncp->nc_lock);
731 		mutex_exit(&cpup->cpu_lock);
732 		/* found negative entry; vn is already null from above */
733 		return true;
734 	}
735 	mutex_enter(vp->v_interlock);
736 	mutex_exit(&ncp->nc_lock);
737 	mutex_exit(&cpup->cpu_lock);
738 
739 	/*
740 	 * Unlocked except for the vnode interlock.  Call vcache_tryvget().
741 	 */
742 	error = vcache_tryvget(vp);
743 	if (error) {
744 		KASSERT(error == EBUSY);
745 		/*
746 		 * This vnode is being cleaned out.
747 		 * XXX badhits?
748 		 */
749 		COUNT_UNL(cpup, ncs_falsehits);
750 		/* found nothing */
751 		return false;
752 	}
753 
754 	COUNT_UNL(cpup, ncs_goodhits); /* XXX can be "badhits" */
755 	/* found it */
756 	*vn_ret = vp;
757 	return true;
758 }
759 
760 /*
761  * Scan cache looking for name of directory entry pointing at vp.
762  *
763  * If the lookup succeeds the vnode is referenced and stored in dvpp.
764  *
765  * If bufp is non-NULL, also place the name in the buffer which starts
766  * at bufp, immediately before *bpp, and move bpp backwards to point
767  * at the start of it.  (Yes, this is a little baroque, but it's done
768  * this way to cater to the whims of getcwd).
769  *
770  * Returns 0 on success, -1 on cache miss, positive errno on failure.
771  */
772 int
773 cache_revlookup(struct vnode *vp, struct vnode **dvpp, char **bpp, char *bufp)
774 {
775 	struct namecache *ncp;
776 	struct vnode *dvp;
777 	struct ncvhashhead *nvcpp;
778 	struct nchcpu *cpup;
779 	char *bp;
780 	int error, nlen;
781 
782 	if (!doingcache)
783 		goto out;
784 
785 	nvcpp = &ncvhashtbl[NCVHASH(vp)];
786 
787 	/*
788 	 * We increment counters in the local CPU's per-cpu stats.
789 	 * We don't take the per-cpu lock, however, since this function
790 	 * is the only place these counters are incremented so no one
791 	 * will be racing with us to increment them.
792 	 */
793 	cpup = curcpu()->ci_data.cpu_nch;
794 	mutex_enter(namecache_lock);
795 	LIST_FOREACH(ncp, nvcpp, nc_vhash) {
796 		mutex_enter(&ncp->nc_lock);
797 		if (ncp->nc_vp == vp &&
798 		    (dvp = ncp->nc_dvp) != NULL &&
799 		    dvp != vp) { 		/* avoid pesky . entries.. */
800 
801 #ifdef DIAGNOSTIC
802 			if (ncp->nc_nlen == 1 &&
803 			    ncp->nc_name[0] == '.')
804 				panic("cache_revlookup: found entry for .");
805 
806 			if (ncp->nc_nlen == 2 &&
807 			    ncp->nc_name[0] == '.' &&
808 			    ncp->nc_name[1] == '.')
809 				panic("cache_revlookup: found entry for ..");
810 #endif
811 			COUNT(cpup, ncs_revhits);
812 			nlen = ncp->nc_nlen;
813 
814 			if (bufp) {
815 				bp = *bpp;
816 				bp -= nlen;
817 				if (bp <= bufp) {
818 					*dvpp = NULL;
819 					mutex_exit(&ncp->nc_lock);
820 					mutex_exit(namecache_lock);
821 					SDT_PROBE(vfs, namecache, revlookup,
822 					    fail, vp, ERANGE, 0, 0, 0);
823 					return (ERANGE);
824 				}
825 				memcpy(bp, ncp->nc_name, nlen);
826 				*bpp = bp;
827 			}
828 
829 			mutex_enter(dvp->v_interlock);
830 			mutex_exit(&ncp->nc_lock);
831 			mutex_exit(namecache_lock);
832 			error = vcache_tryvget(dvp);
833 			if (error) {
834 				KASSERT(error == EBUSY);
835 				if (bufp)
836 					(*bpp) += nlen;
837 				*dvpp = NULL;
838 				SDT_PROBE(vfs, namecache, revlookup, fail, vp,
839 				    error, 0, 0, 0);
840 				return -1;
841 			}
842 			*dvpp = dvp;
843 			SDT_PROBE(vfs, namecache, revlookup, success, vp, dvp,
844 			    0, 0, 0);
845 			return (0);
846 		}
847 		mutex_exit(&ncp->nc_lock);
848 	}
849 	COUNT(cpup, ncs_revmiss);
850 	mutex_exit(namecache_lock);
851  out:
852 	*dvpp = NULL;
853 	return (-1);
854 }
855 
856 /*
857  * Add an entry to the cache
858  */
859 void
860 cache_enter(struct vnode *dvp, struct vnode *vp,
861 	    const char *name, size_t namelen, uint32_t cnflags)
862 {
863 	struct namecache *ncp;
864 	struct namecache *oncp;
865 	struct nchashhead *ncpp;
866 	struct ncvhashhead *nvcpp;
867 	nchash_t hash;
868 
869 	/* First, check whether we can/should add a cache entry. */
870 	if ((cnflags & MAKEENTRY) == 0 ||
871 	    __predict_false(namelen > NCHNAMLEN || !doingcache)) {
872 		SDT_PROBE(vfs, namecache, enter, toolong, vp, name, namelen,
873 		    0, 0);
874 		return;
875 	}
876 
877 	SDT_PROBE(vfs, namecache, enter, done, vp, name, namelen, 0, 0);
878 	if (numcache > desiredvnodes) {
879 		mutex_enter(namecache_lock);
880 		cache_ev_forced.ev_count++;
881 		cache_reclaim();
882 		mutex_exit(namecache_lock);
883 	}
884 
885 	ncp = pool_cache_get(namecache_cache, PR_WAITOK);
886 	mutex_enter(namecache_lock);
887 	numcache++;
888 
889 	/*
890 	 * Concurrent lookups in the same directory may race for a
891 	 * cache entry.  if there's a duplicated entry, free it.
892 	 */
893 	oncp = cache_lookup_entry(dvp, name, namelen);
894 	if (oncp) {
895 		cache_invalidate(oncp);
896 		mutex_exit(&oncp->nc_lock);
897 	}
898 
899 	/* Grab the vnode we just found. */
900 	mutex_enter(&ncp->nc_lock);
901 	ncp->nc_vp = vp;
902 	ncp->nc_flags = 0;
903 	ncp->nc_hittime = 0;
904 	ncp->nc_gcqueue = NULL;
905 	if (vp == NULL) {
906 		/*
907 		 * For negative hits, save the ISWHITEOUT flag so we can
908 		 * restore it later when the cache entry is used again.
909 		 */
910 		ncp->nc_flags = cnflags & ISWHITEOUT;
911 	}
912 
913 	/* Fill in cache info. */
914 	ncp->nc_dvp = dvp;
915 	LIST_INSERT_HEAD(&VNODE_TO_VIMPL(dvp)->vi_dnclist, ncp, nc_dvlist);
916 	if (vp)
917 		LIST_INSERT_HEAD(&VNODE_TO_VIMPL(vp)->vi_nclist, ncp, nc_vlist);
918 	else {
919 		ncp->nc_vlist.le_prev = NULL;
920 		ncp->nc_vlist.le_next = NULL;
921 	}
922 	KASSERT(namelen <= NCHNAMLEN);
923 	ncp->nc_nlen = namelen;
924 	memcpy(ncp->nc_name, name, (unsigned)ncp->nc_nlen);
925 	TAILQ_INSERT_TAIL(&nclruhead, ncp, nc_lru);
926 	hash = cache_hash(name, namelen);
927 	ncpp = &nchashtbl[NCHASH2(hash, dvp)];
928 
929 	/*
930 	 * Flush updates before making visible in table.  No need for a
931 	 * memory barrier on the other side: to see modifications the
932 	 * list must be followed, meaning a dependent pointer load.
933 	 * The below is LIST_INSERT_HEAD() inlined, with the memory
934 	 * barrier included in the correct place.
935 	 */
936 	if ((ncp->nc_hash.le_next = ncpp->lh_first) != NULL)
937 		ncpp->lh_first->nc_hash.le_prev = &ncp->nc_hash.le_next;
938 	ncp->nc_hash.le_prev = &ncpp->lh_first;
939 	membar_producer();
940 	ncpp->lh_first = ncp;
941 
942 	ncp->nc_vhash.le_prev = NULL;
943 	ncp->nc_vhash.le_next = NULL;
944 
945 	/*
946 	 * Create reverse-cache entries (used in getcwd) for directories.
947 	 * (and in linux procfs exe node)
948 	 */
949 	if (vp != NULL &&
950 	    vp != dvp &&
951 #ifndef NAMECACHE_ENTER_REVERSE
952 	    vp->v_type == VDIR &&
953 #endif
954 	    (ncp->nc_nlen > 2 ||
955 	    (ncp->nc_nlen > 1 && ncp->nc_name[1] != '.') ||
956 	    (/* ncp->nc_nlen > 0 && */ ncp->nc_name[0] != '.'))) {
957 		nvcpp = &ncvhashtbl[NCVHASH(vp)];
958 		LIST_INSERT_HEAD(nvcpp, ncp, nc_vhash);
959 	}
960 	mutex_exit(&ncp->nc_lock);
961 	mutex_exit(namecache_lock);
962 }
963 
964 /*
965  * Name cache initialization, from vfs_init() when we are booting
966  */
967 void
968 nchinit(void)
969 {
970 	int error;
971 
972 	TAILQ_INIT(&nclruhead);
973 	namecache_cache = pool_cache_init(sizeof(struct namecache),
974 	    coherency_unit, 0, 0, "ncache", NULL, IPL_NONE, cache_ctor,
975 	    cache_dtor, NULL);
976 	KASSERT(namecache_cache != NULL);
977 
978 	namecache_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
979 
980 	nchashtbl = hashinit(desiredvnodes, HASH_LIST, true, &nchash);
981 	ncvhashtbl =
982 #ifdef NAMECACHE_ENTER_REVERSE
983 	    hashinit(desiredvnodes, HASH_LIST, true, &ncvhash);
984 #else
985 	    hashinit(desiredvnodes/8, HASH_LIST, true, &ncvhash);
986 #endif
987 
988 	error = kthread_create(PRI_VM, KTHREAD_MPSAFE, NULL, cache_thread,
989 	    NULL, NULL, "cachegc");
990 	if (error != 0)
991 		panic("nchinit %d", error);
992 
993 	evcnt_attach_dynamic(&cache_ev_scan, EVCNT_TYPE_MISC, NULL,
994 	   "namecache", "entries scanned");
995 	evcnt_attach_dynamic(&cache_ev_gc, EVCNT_TYPE_MISC, NULL,
996 	   "namecache", "entries collected");
997 	evcnt_attach_dynamic(&cache_ev_over, EVCNT_TYPE_MISC, NULL,
998 	   "namecache", "over scan target");
999 	evcnt_attach_dynamic(&cache_ev_under, EVCNT_TYPE_MISC, NULL,
1000 	   "namecache", "under scan target");
1001 	evcnt_attach_dynamic(&cache_ev_forced, EVCNT_TYPE_MISC, NULL,
1002 	   "namecache", "forced reclaims");
1003 
1004 	sysctl_cache_stat_setup();
1005 }
1006 
1007 static int
1008 cache_ctor(void *arg, void *obj, int flag)
1009 {
1010 	struct namecache *ncp;
1011 
1012 	ncp = obj;
1013 	mutex_init(&ncp->nc_lock, MUTEX_DEFAULT, IPL_NONE);
1014 
1015 	return 0;
1016 }
1017 
1018 static void
1019 cache_dtor(void *arg, void *obj)
1020 {
1021 	struct namecache *ncp;
1022 
1023 	ncp = obj;
1024 	mutex_destroy(&ncp->nc_lock);
1025 }
1026 
1027 /*
1028  * Called once for each CPU in the system as attached.
1029  */
1030 void
1031 cache_cpu_init(struct cpu_info *ci)
1032 {
1033 	struct nchcpu *cpup;
1034 	size_t sz;
1035 
1036 	sz = roundup2(sizeof(*cpup), coherency_unit) + coherency_unit;
1037 	cpup = kmem_zalloc(sz, KM_SLEEP);
1038 	cpup = (void *)roundup2((uintptr_t)cpup, coherency_unit);
1039 	mutex_init(&cpup->cpu_lock, MUTEX_DEFAULT, IPL_NONE);
1040 	ci->ci_data.cpu_nch = cpup;
1041 }
1042 
1043 /*
1044  * Name cache reinitialization, for when the maximum number of vnodes increases.
1045  */
1046 void
1047 nchreinit(void)
1048 {
1049 	struct namecache *ncp;
1050 	struct nchashhead *oldhash1, *hash1;
1051 	struct ncvhashhead *oldhash2, *hash2;
1052 	u_long i, oldmask1, oldmask2, mask1, mask2;
1053 
1054 	hash1 = hashinit(desiredvnodes, HASH_LIST, true, &mask1);
1055 	hash2 =
1056 #ifdef NAMECACHE_ENTER_REVERSE
1057 	    hashinit(desiredvnodes, HASH_LIST, true, &mask2);
1058 #else
1059 	    hashinit(desiredvnodes/8, HASH_LIST, true, &mask2);
1060 #endif
1061 	mutex_enter(namecache_lock);
1062 	cache_lock_cpus();
1063 	oldhash1 = nchashtbl;
1064 	oldmask1 = nchash;
1065 	nchashtbl = hash1;
1066 	nchash = mask1;
1067 	oldhash2 = ncvhashtbl;
1068 	oldmask2 = ncvhash;
1069 	ncvhashtbl = hash2;
1070 	ncvhash = mask2;
1071 	for (i = 0; i <= oldmask1; i++) {
1072 		while ((ncp = LIST_FIRST(&oldhash1[i])) != NULL) {
1073 			LIST_REMOVE(ncp, nc_hash);
1074 			ncp->nc_hash.le_prev = NULL;
1075 		}
1076 	}
1077 	for (i = 0; i <= oldmask2; i++) {
1078 		while ((ncp = LIST_FIRST(&oldhash2[i])) != NULL) {
1079 			LIST_REMOVE(ncp, nc_vhash);
1080 			ncp->nc_vhash.le_prev = NULL;
1081 		}
1082 	}
1083 	cache_unlock_cpus();
1084 	mutex_exit(namecache_lock);
1085 	hashdone(oldhash1, HASH_LIST, oldmask1);
1086 	hashdone(oldhash2, HASH_LIST, oldmask2);
1087 }
1088 
1089 /*
1090  * Cache flush, a particular vnode; called when a vnode is renamed to
1091  * hide entries that would now be invalid
1092  */
1093 void
1094 cache_purge1(struct vnode *vp, const char *name, size_t namelen, int flags)
1095 {
1096 	struct namecache *ncp, *ncnext;
1097 
1098 	mutex_enter(namecache_lock);
1099 	if (flags & PURGE_PARENTS) {
1100 		SDT_PROBE(vfs, namecache, purge, parents, vp, 0, 0, 0, 0);
1101 
1102 		for (ncp = LIST_FIRST(&VNODE_TO_VIMPL(vp)->vi_nclist);
1103 		    ncp != NULL; ncp = ncnext) {
1104 			ncnext = LIST_NEXT(ncp, nc_vlist);
1105 			mutex_enter(&ncp->nc_lock);
1106 			cache_invalidate(ncp);
1107 			mutex_exit(&ncp->nc_lock);
1108 			cache_disassociate(ncp);
1109 		}
1110 	}
1111 	if (flags & PURGE_CHILDREN) {
1112 		SDT_PROBE(vfs, namecache, purge, children, vp, 0, 0, 0, 0);
1113 		for (ncp = LIST_FIRST(&VNODE_TO_VIMPL(vp)->vi_dnclist);
1114 		    ncp != NULL; ncp = ncnext) {
1115 			ncnext = LIST_NEXT(ncp, nc_dvlist);
1116 			mutex_enter(&ncp->nc_lock);
1117 			cache_invalidate(ncp);
1118 			mutex_exit(&ncp->nc_lock);
1119 			cache_disassociate(ncp);
1120 		}
1121 	}
1122 	if (name != NULL) {
1123 		SDT_PROBE(vfs, namecache, purge, name, name, namelen, 0, 0, 0);
1124 		ncp = cache_lookup_entry(vp, name, namelen);
1125 		if (ncp) {
1126 			cache_invalidate(ncp);
1127 			mutex_exit(&ncp->nc_lock);
1128 			cache_disassociate(ncp);
1129 		}
1130 	}
1131 	mutex_exit(namecache_lock);
1132 }
1133 
1134 /*
1135  * Cache flush, a whole filesystem; called when filesys is umounted to
1136  * remove entries that would now be invalid.
1137  */
1138 void
1139 cache_purgevfs(struct mount *mp)
1140 {
1141 	struct namecache *ncp, *nxtcp;
1142 
1143 	SDT_PROBE(vfs, namecache, purge, vfs, mp, 0, 0, 0, 0);
1144 	mutex_enter(namecache_lock);
1145 	for (ncp = TAILQ_FIRST(&nclruhead); ncp != NULL; ncp = nxtcp) {
1146 		nxtcp = TAILQ_NEXT(ncp, nc_lru);
1147 		mutex_enter(&ncp->nc_lock);
1148 		if (ncp->nc_dvp != NULL && ncp->nc_dvp->v_mount == mp) {
1149 			/* Free the resources we had. */
1150 			cache_invalidate(ncp);
1151 			cache_disassociate(ncp);
1152 		}
1153 		mutex_exit(&ncp->nc_lock);
1154 	}
1155 	cache_reclaim();
1156 	mutex_exit(namecache_lock);
1157 }
1158 
1159 /*
1160  * Scan global list invalidating entries until we meet a preset target.
1161  * Prefer to invalidate entries that have not scored a hit within
1162  * cache_hottime seconds.  We sort the LRU list only for this routine's
1163  * benefit.
1164  */
1165 static void
1166 cache_prune(int incache, int target)
1167 {
1168 	struct namecache *ncp, *nxtcp, *sentinel;
1169 	int items, recent, tryharder;
1170 
1171 	KASSERT(mutex_owned(namecache_lock));
1172 
1173 	SDT_PROBE(vfs, namecache, prune, done, incache, target, 0, 0, 0);
1174 	items = 0;
1175 	tryharder = 0;
1176 	recent = hardclock_ticks - hz * cache_hottime;
1177 	sentinel = NULL;
1178 	for (ncp = TAILQ_FIRST(&nclruhead); ncp != NULL; ncp = nxtcp) {
1179 		if (incache <= target)
1180 			break;
1181 		items++;
1182 		nxtcp = TAILQ_NEXT(ncp, nc_lru);
1183 		if (ncp == sentinel) {
1184 			/*
1185 			 * If we looped back on ourself, then ignore
1186 			 * recent entries and purge whatever we find.
1187 			 */
1188 			tryharder = 1;
1189 		}
1190 		if (ncp->nc_dvp == NULL)
1191 			continue;
1192 		if (!tryharder && (ncp->nc_hittime - recent) > 0) {
1193 			if (sentinel == NULL)
1194 				sentinel = ncp;
1195 			TAILQ_REMOVE(&nclruhead, ncp, nc_lru);
1196 			TAILQ_INSERT_TAIL(&nclruhead, ncp, nc_lru);
1197 			continue;
1198 		}
1199 		mutex_enter(&ncp->nc_lock);
1200 		if (ncp->nc_dvp != NULL) {
1201 			cache_invalidate(ncp);
1202 			cache_disassociate(ncp);
1203 			incache--;
1204 		}
1205 		mutex_exit(&ncp->nc_lock);
1206 	}
1207 	cache_ev_scan.ev_count += items;
1208 }
1209 
1210 /*
1211  * Collect dead cache entries from all CPUs and garbage collect.
1212  */
1213 static void
1214 cache_reclaim(void)
1215 {
1216 	struct namecache *ncp, *next;
1217 	int items;
1218 
1219 	KASSERT(mutex_owned(namecache_lock));
1220 
1221 	/*
1222 	 * If the number of extant entries not awaiting garbage collection
1223 	 * exceeds the high water mark, then reclaim stale entries until we
1224 	 * reach our low water mark.
1225 	 */
1226 	items = numcache - cache_gcpend;
1227 	if (items > (uint64_t)desiredvnodes * cache_hiwat / 100) {
1228 		cache_prune(items, (int)((uint64_t)desiredvnodes *
1229 		    cache_lowat / 100));
1230 		cache_ev_over.ev_count++;
1231 	} else
1232 		cache_ev_under.ev_count++;
1233 
1234 	/*
1235 	 * Stop forward lookup activity on all CPUs and garbage collect dead
1236 	 * entries.
1237 	 */
1238 	cache_lock_cpus();
1239 	ncp = cache_gcqueue;
1240 	cache_gcqueue = NULL;
1241 	items = cache_gcpend;
1242 	cache_gcpend = 0;
1243 	while (ncp != NULL) {
1244 		next = ncp->nc_gcqueue;
1245 		cache_disassociate(ncp);
1246 		KASSERT(ncp->nc_dvp == NULL);
1247 		if (ncp->nc_hash.le_prev != NULL) {
1248 			LIST_REMOVE(ncp, nc_hash);
1249 			ncp->nc_hash.le_prev = NULL;
1250 		}
1251 		pool_cache_put(namecache_cache, ncp);
1252 		ncp = next;
1253 	}
1254 	cache_unlock_cpus();
1255 	numcache -= items;
1256 	cache_ev_gc.ev_count += items;
1257 }
1258 
1259 /*
1260  * Cache maintainence thread, awakening once per second to:
1261  *
1262  * => keep number of entries below the high water mark
1263  * => sort pseudo-LRU list
1264  * => garbage collect dead entries
1265  */
1266 static void
1267 cache_thread(void *arg)
1268 {
1269 
1270 	mutex_enter(namecache_lock);
1271 	for (;;) {
1272 		cache_reclaim();
1273 		kpause("cachegc", false, hz, namecache_lock);
1274 	}
1275 }
1276 
1277 #ifdef DDB
1278 void
1279 namecache_print(struct vnode *vp, void (*pr)(const char *, ...))
1280 {
1281 	struct vnode *dvp = NULL;
1282 	struct namecache *ncp;
1283 
1284 	TAILQ_FOREACH(ncp, &nclruhead, nc_lru) {
1285 		if (ncp->nc_vp == vp && ncp->nc_dvp != NULL) {
1286 			(*pr)("name %.*s\n", ncp->nc_nlen, ncp->nc_name);
1287 			dvp = ncp->nc_dvp;
1288 		}
1289 	}
1290 	if (dvp == NULL) {
1291 		(*pr)("name not found\n");
1292 		return;
1293 	}
1294 	vp = dvp;
1295 	TAILQ_FOREACH(ncp, &nclruhead, nc_lru) {
1296 		if (ncp->nc_vp == vp) {
1297 			(*pr)("parent %.*s\n", ncp->nc_nlen, ncp->nc_name);
1298 		}
1299 	}
1300 }
1301 #endif
1302 
1303 void
1304 namecache_count_pass2(void)
1305 {
1306 	struct nchcpu *cpup = curcpu()->ci_data.cpu_nch;
1307 
1308 	COUNT_UNL(cpup, ncs_pass2);
1309 }
1310 
1311 void
1312 namecache_count_2passes(void)
1313 {
1314 	struct nchcpu *cpup = curcpu()->ci_data.cpu_nch;
1315 
1316 	COUNT_UNL(cpup, ncs_2passes);
1317 }
1318 
1319 /*
1320  * Fetch the current values of the stats.  We return the most
1321  * recent values harvested into nchstats by cache_reclaim(), which
1322  * will be less than a second old.
1323  */
1324 static int
1325 cache_stat_sysctl(SYSCTLFN_ARGS)
1326 {
1327 	struct nchstats stats;
1328 	struct nchcpu *my_cpup;
1329 #ifdef CACHE_STATS_CURRENT
1330 	CPU_INFO_ITERATOR cii;
1331 	struct cpu_info *ci;
1332 #endif	/* CACHE_STATS_CURRENT */
1333 
1334 	if (oldp == NULL) {
1335 		*oldlenp = sizeof(stats);
1336 		return 0;
1337 	}
1338 
1339 	if (*oldlenp < sizeof(stats)) {
1340 		*oldlenp = 0;
1341 		return 0;
1342 	}
1343 
1344 	/*
1345 	 * Take this CPU's per-cpu lock to hold off cache_reclaim()
1346 	 * from doing a stats update while doing minimal damage to
1347 	 * concurrent operations.
1348 	 */
1349 	sysctl_unlock();
1350 	my_cpup = curcpu()->ci_data.cpu_nch;
1351 	mutex_enter(&my_cpup->cpu_lock);
1352 	stats = nchstats;
1353 #ifdef CACHE_STATS_CURRENT
1354 	for (CPU_INFO_FOREACH(cii, ci)) {
1355 		struct nchcpu *cpup = ci->ci_data.cpu_nch;
1356 
1357 		ADD(stats, cpup, ncs_goodhits);
1358 		ADD(stats, cpup, ncs_neghits);
1359 		ADD(stats, cpup, ncs_badhits);
1360 		ADD(stats, cpup, ncs_falsehits);
1361 		ADD(stats, cpup, ncs_miss);
1362 		ADD(stats, cpup, ncs_long);
1363 		ADD(stats, cpup, ncs_pass2);
1364 		ADD(stats, cpup, ncs_2passes);
1365 		ADD(stats, cpup, ncs_revhits);
1366 		ADD(stats, cpup, ncs_revmiss);
1367 	}
1368 #endif	/* CACHE_STATS_CURRENT */
1369 	mutex_exit(&my_cpup->cpu_lock);
1370 	sysctl_relock();
1371 
1372 	*oldlenp = sizeof(stats);
1373 	return sysctl_copyout(l, &stats, oldp, sizeof(stats));
1374 }
1375 
1376 static void
1377 sysctl_cache_stat_setup(void)
1378 {
1379 
1380 	KASSERT(sysctllog == NULL);
1381 	sysctl_createv(&sysctllog, 0, NULL, NULL,
1382 		       CTLFLAG_PERMANENT,
1383 		       CTLTYPE_STRUCT, "namecache_stats",
1384 		       SYSCTL_DESCR("namecache statistics"),
1385 		       cache_stat_sysctl, 0, NULL, 0,
1386 		       CTL_VFS, CTL_CREATE, CTL_EOL);
1387 }
1388