1/*
2 * Copyright (c) 1989, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * This code is derived from software contributed to Berkeley by
6 * Rick Macklem at The University of Guelph.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
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 * 3. All advertising materials mentioning features or use of this software
17 * must display the following acknowledgement:
18 * This product includes software developed by the University of
19 * California, Berkeley and its contributors.
20 * 4. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * SUCH DAMAGE.
35 *
36 * @(#)nfs_node.c 8.6 (Berkeley) 5/22/95
37 * $FreeBSD: head/sys/nfsclient/nfs_node.c 83366 2001-09-12 08:38:13Z julian $
38 */
39
40
41#include <sys/param.h>
42#include <sys/systm.h>
43#include <sys/fnv_hash.h>
44#include <sys/lock.h>
45#include <sys/malloc.h>
46#include <sys/mount.h>
47#include <sys/namei.h>
48#include <sys/proc.h>
49#include <sys/socket.h>
50#include <sys/sysctl.h>
51#include <sys/vnode.h>
52
53#include <vm/vm_zone.h>
54
55#include <nfs/rpcv2.h>
56#include <nfs/nfsproto.h>
57#include
58#include
59#include
60
61static vm_zone_t nfsnode_zone;
62static LIST_HEAD(nfsnodehashhead, nfsnode) *nfsnodehashtbl;
63static u_long nfsnodehash;
64
65#define TRUE 1
66#define FALSE 0
67
68/*
69 * Grab an atomic snapshot of the nfsnode hash chain lengths
70 */
71SYSCTL_DECL(_debug_hashstat);
72static int
73sysctl_debug_hashstat_rawnfsnode(SYSCTL_HANDLER_ARGS)
74{
75 int error;
76 struct nfsnodehashhead *nnpp;
77 struct nfsnode *nnp;
78 int n_nfsnode;
79 int count;
80
81 n_nfsnode = nfsnodehash + 1; /* nfsnodehash = max index, not count */
82 if (!req->oldptr)
83 return SYSCTL_OUT(req, 0, n_nfsnode * sizeof(int));
84
85 /* Scan hash tables for applicable entries */
86 for (nnpp = nfsnodehashtbl; n_nfsnode > 0; n_nfsnode--, nnpp++) {
87 count = 0;
88 LIST_FOREACH(nnp, nnpp, n_hash) {
89 count++;
90 }
91 error = SYSCTL_OUT(req, (caddr_t)&count, sizeof(count));
92 if (error)
93 return (error);
94 }
95 return (0);
96}
97SYSCTL_PROC(_debug_hashstat, OID_AUTO, rawnfsnode, CTLTYPE_INT|CTLFLAG_RD,
98 0, 0, sysctl_debug_hashstat_rawnfsnode, "S,int", "nfsnode chain lengths");
99
100static int
101sysctl_debug_hashstat_nfsnode(SYSCTL_HANDLER_ARGS)
102{
103 int error;
104 struct nfsnodehashhead *nnpp;
105 struct nfsnode *nnp;
106 int n_nfsnode;
107 int count, maxlength, used, pct;
108
109 if (!req->oldptr)
110 return SYSCTL_OUT(req, 0, 4 * sizeof(int));
111
112 n_nfsnode = nfsnodehash + 1; /* nfsnodehash = max index, not count */
113 used = 0;
114 maxlength = 0;
115
116 /* Scan hash tables for applicable entries */
117 for (nnpp = nfsnodehashtbl; n_nfsnode > 0; n_nfsnode--, nnpp++) {
118 count = 0;
119 LIST_FOREACH(nnp, nnpp, n_hash) {
120 count++;
121 }
122 if (count)
123 used++;
124 if (maxlength < count)
125 maxlength = count;
126 }
127 n_nfsnode = nfsnodehash + 1;
128 pct = (used * 100 * 100) / n_nfsnode;
129 error = SYSCTL_OUT(req, (caddr_t)&n_nfsnode, sizeof(n_nfsnode));
130 if (error)
131 return (error);
132 error = SYSCTL_OUT(req, (caddr_t)&used, sizeof(used));
133 if (error)
134 return (error);
135 error = SYSCTL_OUT(req, (caddr_t)&maxlength, sizeof(maxlength));
136 if (error)
137 return (error);
138 error = SYSCTL_OUT(req, (caddr_t)&pct, sizeof(pct));
139 if (error)
140 return (error);
141 return (0);
142}
143SYSCTL_PROC(_debug_hashstat, OID_AUTO, nfsnode, CTLTYPE_INT|CTLFLAG_RD,
144 0, 0, sysctl_debug_hashstat_nfsnode, "I", "nfsnode chain lengths");
145
146/*
147 * Initialize hash links for nfsnodes
148 * and build nfsnode free list.
149 */
150void
151nfs_nhinit()
152{
153 nfsnode_zone = zinit("NFSNODE", sizeof(struct nfsnode), 0, 0, 1);
154 nfsnodehashtbl = hashinit(desiredvnodes, M_NFSHASH, &nfsnodehash);
155}
156
157/*
158 * Look up a vnode/nfsnode by file handle.
159 * Callers must check for mount points!!
160 * In all cases, a pointer to a
161 * nfsnode structure is returned.
162 */
163static int nfs_node_hash_lock;
164
165int
166nfs_nget(mntp, fhp, fhsize, npp)
167 struct mount *mntp;
168 register nfsfh_t *fhp;
169 int fhsize;
170 struct nfsnode **npp;
171{
172 struct thread *td = curthread; /* XXX */
173 struct nfsnode *np, *np2;
174 struct nfsnodehashhead *nhpp;
175 register struct vnode *vp;
176 struct vnode *nvp;
177 int error;
178 int rsflags;
179 struct nfsmount *nmp;
180
181 /*
182 * Calculate nfs mount point and figure out whether the rslock should
183 * be interruptable or not.
184 */
185 nmp = VFSTONFS(mntp);
186 if (nmp->nm_flag & NFSMNT_INT)
187 rsflags = PCATCH;
188 else
189 rsflags = 0;
190
191retry:
192 nhpp = NFSNOHASH(fnv_32_buf(fhp->fh_bytes, fhsize, FNV1_32_INIT));
193loop:
194 for (np = nhpp->lh_first; np != 0; np = np->n_hash.le_next) {
195 if (mntp != NFSTOV(np)->v_mount || np->n_fhsize != fhsize ||
196 bcmp((caddr_t)fhp, (caddr_t)np->n_fhp, fhsize))
197 continue;
198 vp = NFSTOV(np);
199 if (vget(vp, LK_EXCLUSIVE, td))
200 goto loop;
201 *npp = np;
202 return(0);
203 }
204 /*
205 * Obtain a lock to prevent a race condition if the getnewvnode()
206 * or MALLOC() below happens to block.
207 */
208 if (nfs_node_hash_lock) {
209 while (nfs_node_hash_lock) {
210 nfs_node_hash_lock = -1;
211 tsleep(&nfs_node_hash_lock, PVM, "nfsngt", 0);
212 }
213 goto loop;
214 }
215 nfs_node_hash_lock = 1;
216
217 /*
218 * Allocate before getnewvnode since doing so afterward
219 * might cause a bogus v_data pointer to get dereferenced
220 * elsewhere if zalloc should block.
221 */
222 np = zalloc(nfsnode_zone);
223
224 error = getnewvnode(VT_NFS, mntp, nfsv2_vnodeop_p, &nvp);
225 if (error) {
226 if (nfs_node_hash_lock < 0)
227 wakeup(&nfs_node_hash_lock);
228 nfs_node_hash_lock = 0;
229 *npp = 0;
230 zfree(nfsnode_zone, np);
231 return (error);
232 }
233 vp = nvp;
234 bzero((caddr_t)np, sizeof *np);
235 vp->v_data = np;
236 np->n_vnode = vp;
237 /*
238 * Insert the nfsnode in the hash queue for its new file handle
239 */
240 for (np2 = nhpp->lh_first; np2 != 0; np2 = np2->n_hash.le_next) {
241 if (mntp != NFSTOV(np2)->v_mount || np2->n_fhsize != fhsize ||
242 bcmp((caddr_t)fhp, (caddr_t)np2->n_fhp, fhsize))
243 continue;
244 vrele(vp);
245 if (nfs_node_hash_lock < 0)
246 wakeup(&nfs_node_hash_lock);
247 nfs_node_hash_lock = 0;
248 zfree(nfsnode_zone, np);
249 goto retry;
250 }
251 LIST_INSERT_HEAD(nhpp, np, n_hash);
252 if (fhsize > NFS_SMALLFH) {
253 MALLOC(np->n_fhp, nfsfh_t *, fhsize, M_NFSBIGFH, M_WAITOK);
254 } else
255 np->n_fhp = &np->n_fh;
256 bcopy((caddr_t)fhp, (caddr_t)np->n_fhp, fhsize);
257 np->n_fhsize = fhsize;
258 lockinit(&np->n_rslock, PVFS | rsflags, "nfrslk", 0, LK_NOPAUSE);
259 lockinit(&vp->v_lock, PVFS, "nfsnlk", 0, LK_NOPAUSE);
260 *npp = np;
261
262 if (nfs_node_hash_lock < 0)
263 wakeup(&nfs_node_hash_lock);
264 nfs_node_hash_lock = 0;
265
266 /*
267 * Lock the new nfsnode.
268 */
269 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
270
271 return (0);
272}
273
274int
275nfs_inactive(ap)
276 struct vop_inactive_args /* {
277 struct vnode *a_vp;
278 struct thread *a_td;
279 } */ *ap;
280{
281 register struct nfsnode *np;
282 register struct sillyrename *sp;
283 struct thread *td = curthread; /* XXX */
284
285 np = VTONFS(ap->a_vp);
286 if (prtactive && ap->a_vp->v_usecount != 0)
287 vprint("nfs_inactive: pushing active", ap->a_vp);
288 if (ap->a_vp->v_type != VDIR) {
289 sp = np->n_sillyrename;
290 np->n_sillyrename = (struct sillyrename *)0;
291 } else
292 sp = (struct sillyrename *)0;
293 if (sp) {
294 /*
295 * We need a reference to keep the vnode from being
296 * recycled by getnewvnode while we do the I/O
297 * associated with discarding the buffers unless we
298 * are being forcibly unmounted in which case we already
299 * have our own reference.
300 */
301 if (ap->a_vp->v_usecount > 0)
302 (void) nfs_vinvalbuf(ap->a_vp, 0, sp->s_cred, td, 1);
303 else if (vget(ap->a_vp, 0, td))
304 panic("nfs_inactive: lost vnode");
305 else {
306 (void) nfs_vinvalbuf(ap->a_vp, 0, sp->s_cred, td, 1);
307 vrele(ap->a_vp);
308 }
309 /*
310 * Remove the silly file that was rename'd earlier
311 */
312 nfs_removeit(sp);
313 crfree(sp->s_cred);
314 vrele(sp->s_dvp);
315 FREE((caddr_t)sp, M_NFSREQ);
316 }
317 np->n_flag &= (NMODIFIED | NFLUSHINPROG | NFLUSHWANT | NQNFSEVICTED |
318 NQNFSNONCACHE | NQNFSWRITE);
319 VOP_UNLOCK(ap->a_vp, 0, ap->a_td);
320 return (0);
321}
322
323/*
324 * Reclaim an nfsnode so that it can be used for other purposes.
325 */
326int
327nfs_reclaim(ap)
328 struct vop_reclaim_args /* {
329 struct vnode *a_vp;
330 } */ *ap;
331{
332 register struct vnode *vp = ap->a_vp;
333 register struct nfsnode *np = VTONFS(vp);
334 register struct nfsmount *nmp = VFSTONFS(vp->v_mount);
335 register struct nfsdmap *dp, *dp2;
336
337 if (prtactive && vp->v_usecount != 0)
338 vprint("nfs_reclaim: pushing active", vp);
339
340 if (np->n_hash.le_prev != NULL)
341 LIST_REMOVE(np, n_hash);
342
343 /*
344 * For nqnfs, take it off the timer queue as required.
345 */
346 if ((nmp->nm_flag & NFSMNT_NQNFS) && TAILQ_NEXT(np, n_timer) != 0) {
347 TAILQ_REMOVE(&nmp->nm_timerhead, np, n_timer);
348 }
349
350 /*
351 * Free up any directory cookie structures and
352 * large file handle structures that might be associated with
353 * this nfs node.
354 */
355 if (vp->v_type == VDIR) {
356 dp = np->n_cookies.lh_first;
357 while (dp) {
358 dp2 = dp;
359 dp = dp->ndm_list.le_next;
360 FREE((caddr_t)dp2, M_NFSDIROFF);
361 }
362 }
363 if (np->n_fhsize > NFS_SMALLFH) {
364 FREE((caddr_t)np->n_fhp, M_NFSBIGFH);
365 }
366
367 lockdestroy(&np->n_rslock);
368
369 cache_purge(vp);
370 zfree(nfsnode_zone, vp->v_data);
371 vp->v_data = (void *)0;
372 return (0);
373}
374
375#if 0
376/*
377 * Lock an nfsnode
378 */
379int
380nfs_lock(ap)
381 struct vop_lock_args /* {
382 struct vnode *a_vp;
383 } */ *ap;
384{
385 register struct vnode *vp = ap->a_vp;
386
387 /*
388 * Ugh, another place where interruptible mounts will get hung.
389 * If you make this sleep interruptible, then you have to fix all
390 * the VOP_LOCK() calls to expect interruptibility.
391 */
392 while (vp->v_flag & VXLOCK) {
393 vp->v_flag |= VXWANT;
394 (void) tsleep((caddr_t)vp, PINOD, "nfslck", 0);
395 }
396 if (vp->v_tag == VT_NON)
397 return (ENOENT);
398
399#if 0
400 /*
401 * Only lock regular files. If a server crashed while we were
402 * holding a directory lock, we could easily end up sleeping
403 * until the server rebooted while holding a lock on the root.
404 * Locks are only needed for protecting critical sections in
405 * VMIO at the moment.
406 * New vnodes will have type VNON but they should be locked
407 * since they may become VREG. This is checked in loadattrcache
408 * and unwanted locks are released there.
409 */
410 if (vp->v_type == VREG || vp->v_type == VNON) {
411 while (np->n_flag & NLOCKED) {
412 np->n_flag |= NWANTED;
413 (void) tsleep((caddr_t) np, PINOD, "nfslck2", 0);
414 /*
415 * If the vnode has transmuted into a VDIR while we
416 * were asleep, then skip the lock.
417 */
418 if (vp->v_type != VREG && vp->v_type != VNON)
419 return (0);
420 }
421 np->n_flag |= NLOCKED;
422 }
423#endif
424
425 return (0);
426}
427
428/*
429 * Unlock an nfsnode
430 */
431int
432nfs_unlock(ap)
433 struct vop_unlock_args /* {
434 struct vnode *a_vp;
435 } */ *ap;
436{
437#if 0
438 struct vnode* vp = ap->a_vp;
439 struct nfsnode* np = VTONFS(vp);
440
441 if (vp->v_type == VREG || vp->v_type == VNON) {
442 if (!(np->n_flag & NLOCKED))
443 panic("nfs_unlock: nfsnode not locked");
444 np->n_flag &= ~NLOCKED;
445 if (np->n_flag & NWANTED) {
446 np->n_flag &= ~NWANTED;
447 wakeup((caddr_t) np);
448 }
449 }
450#endif
451
452 return (0);
453}
454
455/*
456 * Check for a locked nfsnode
457 */
458int
459nfs_islocked(ap)
460 struct vop_islocked_args /* {
461 struct vnode *a_vp;
462 struct thread *a_td;
463 } */ *ap;
464{
465 return VTONFS(ap->a_vp)->n_flag & NLOCKED ? 1 : 0;
466}
467#endif
468