xref: /linux/fs/nfsd/nfs4state.c (revision 14b42963f64b98ab61fa9723c03d71aa5ef4f862)
1 /*
2 *  linux/fs/nfsd/nfs4state.c
3 *
4 *  Copyright (c) 2001 The Regents of the University of Michigan.
5 *  All rights reserved.
6 *
7 *  Kendrick Smith <kmsmith@umich.edu>
8 *  Andy Adamson <kandros@umich.edu>
9 *
10 *  Redistribution and use in source and binary forms, with or without
11 *  modification, are permitted provided that the following conditions
12 *  are met:
13 *
14 *  1. Redistributions of source code must retain the above copyright
15 *     notice, this list of conditions and the following disclaimer.
16 *  2. Redistributions in binary form must reproduce the above copyright
17 *     notice, this list of conditions and the following disclaimer in the
18 *     documentation and/or other materials provided with the distribution.
19 *  3. Neither the name of the University nor the names of its
20 *     contributors may be used to endorse or promote products derived
21 *     from this software without specific prior written permission.
22 *
23 *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
24 *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
25 *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
26 *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28 *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29 *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
30 *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
31 *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
32 *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
33 *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34 *
35 */
36 
37 #include <linux/param.h>
38 #include <linux/major.h>
39 #include <linux/slab.h>
40 
41 #include <linux/sunrpc/svc.h>
42 #include <linux/nfsd/nfsd.h>
43 #include <linux/nfsd/cache.h>
44 #include <linux/mount.h>
45 #include <linux/workqueue.h>
46 #include <linux/smp_lock.h>
47 #include <linux/kthread.h>
48 #include <linux/nfs4.h>
49 #include <linux/nfsd/state.h>
50 #include <linux/nfsd/xdr4.h>
51 #include <linux/namei.h>
52 #include <linux/mutex.h>
53 
54 #define NFSDDBG_FACILITY                NFSDDBG_PROC
55 
56 /* Globals */
57 static time_t lease_time = 90;     /* default lease time */
58 static time_t user_lease_time = 90;
59 static time_t boot_time;
60 static int in_grace = 1;
61 static u32 current_clientid = 1;
62 static u32 current_ownerid = 1;
63 static u32 current_fileid = 1;
64 static u32 current_delegid = 1;
65 static u32 nfs4_init;
66 static stateid_t zerostateid;             /* bits all 0 */
67 static stateid_t onestateid;              /* bits all 1 */
68 
69 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
70 #define ONE_STATEID(stateid)  (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
71 
72 /* forward declarations */
73 static struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
74 static struct nfs4_delegation * find_delegation_stateid(struct inode *ino, stateid_t *stid);
75 static void release_stateid_lockowners(struct nfs4_stateid *open_stp);
76 static char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery";
77 static void nfs4_set_recdir(char *recdir);
78 
79 /* Locking:
80  *
81  * client_mutex:
82  * 	protects clientid_hashtbl[], clientstr_hashtbl[],
83  * 	unconfstr_hashtbl[], uncofid_hashtbl[].
84  */
85 static DEFINE_MUTEX(client_mutex);
86 
87 static kmem_cache_t *stateowner_slab = NULL;
88 static kmem_cache_t *file_slab = NULL;
89 static kmem_cache_t *stateid_slab = NULL;
90 static kmem_cache_t *deleg_slab = NULL;
91 
92 void
93 nfs4_lock_state(void)
94 {
95 	mutex_lock(&client_mutex);
96 }
97 
98 void
99 nfs4_unlock_state(void)
100 {
101 	mutex_unlock(&client_mutex);
102 }
103 
104 static inline u32
105 opaque_hashval(const void *ptr, int nbytes)
106 {
107 	unsigned char *cptr = (unsigned char *) ptr;
108 
109 	u32 x = 0;
110 	while (nbytes--) {
111 		x *= 37;
112 		x += *cptr++;
113 	}
114 	return x;
115 }
116 
117 /* forward declarations */
118 static void release_stateowner(struct nfs4_stateowner *sop);
119 static void release_stateid(struct nfs4_stateid *stp, int flags);
120 
121 /*
122  * Delegation state
123  */
124 
125 /* recall_lock protects the del_recall_lru */
126 static DEFINE_SPINLOCK(recall_lock);
127 static struct list_head del_recall_lru;
128 
129 static void
130 free_nfs4_file(struct kref *kref)
131 {
132 	struct nfs4_file *fp = container_of(kref, struct nfs4_file, fi_ref);
133 	list_del(&fp->fi_hash);
134 	iput(fp->fi_inode);
135 	kmem_cache_free(file_slab, fp);
136 }
137 
138 static inline void
139 put_nfs4_file(struct nfs4_file *fi)
140 {
141 	kref_put(&fi->fi_ref, free_nfs4_file);
142 }
143 
144 static inline void
145 get_nfs4_file(struct nfs4_file *fi)
146 {
147 	kref_get(&fi->fi_ref);
148 }
149 
150 static int num_delegations;
151 
152 /*
153  * Open owner state (share locks)
154  */
155 
156 /* hash tables for nfs4_stateowner */
157 #define OWNER_HASH_BITS              8
158 #define OWNER_HASH_SIZE             (1 << OWNER_HASH_BITS)
159 #define OWNER_HASH_MASK             (OWNER_HASH_SIZE - 1)
160 
161 #define ownerid_hashval(id) \
162         ((id) & OWNER_HASH_MASK)
163 #define ownerstr_hashval(clientid, ownername) \
164         (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
165 
166 static struct list_head	ownerid_hashtbl[OWNER_HASH_SIZE];
167 static struct list_head	ownerstr_hashtbl[OWNER_HASH_SIZE];
168 
169 /* hash table for nfs4_file */
170 #define FILE_HASH_BITS                   8
171 #define FILE_HASH_SIZE                  (1 << FILE_HASH_BITS)
172 #define FILE_HASH_MASK                  (FILE_HASH_SIZE - 1)
173 /* hash table for (open)nfs4_stateid */
174 #define STATEID_HASH_BITS              10
175 #define STATEID_HASH_SIZE              (1 << STATEID_HASH_BITS)
176 #define STATEID_HASH_MASK              (STATEID_HASH_SIZE - 1)
177 
178 #define file_hashval(x) \
179         hash_ptr(x, FILE_HASH_BITS)
180 #define stateid_hashval(owner_id, file_id)  \
181         (((owner_id) + (file_id)) & STATEID_HASH_MASK)
182 
183 static struct list_head file_hashtbl[FILE_HASH_SIZE];
184 static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
185 
186 static struct nfs4_delegation *
187 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
188 {
189 	struct nfs4_delegation *dp;
190 	struct nfs4_file *fp = stp->st_file;
191 	struct nfs4_callback *cb = &stp->st_stateowner->so_client->cl_callback;
192 
193 	dprintk("NFSD alloc_init_deleg\n");
194 	if (num_delegations > STATEID_HASH_SIZE * 4)
195 		return NULL;
196 	dp = kmem_cache_alloc(deleg_slab, GFP_KERNEL);
197 	if (dp == NULL)
198 		return dp;
199 	num_delegations++;
200 	INIT_LIST_HEAD(&dp->dl_perfile);
201 	INIT_LIST_HEAD(&dp->dl_perclnt);
202 	INIT_LIST_HEAD(&dp->dl_recall_lru);
203 	dp->dl_client = clp;
204 	get_nfs4_file(fp);
205 	dp->dl_file = fp;
206 	dp->dl_flock = NULL;
207 	get_file(stp->st_vfs_file);
208 	dp->dl_vfs_file = stp->st_vfs_file;
209 	dp->dl_type = type;
210 	dp->dl_recall.cbr_dp = NULL;
211 	dp->dl_recall.cbr_ident = cb->cb_ident;
212 	dp->dl_recall.cbr_trunc = 0;
213 	dp->dl_stateid.si_boot = boot_time;
214 	dp->dl_stateid.si_stateownerid = current_delegid++;
215 	dp->dl_stateid.si_fileid = 0;
216 	dp->dl_stateid.si_generation = 0;
217 	dp->dl_fhlen = current_fh->fh_handle.fh_size;
218 	memcpy(dp->dl_fhval, &current_fh->fh_handle.fh_base,
219 		        current_fh->fh_handle.fh_size);
220 	dp->dl_time = 0;
221 	atomic_set(&dp->dl_count, 1);
222 	list_add(&dp->dl_perfile, &fp->fi_delegations);
223 	list_add(&dp->dl_perclnt, &clp->cl_delegations);
224 	return dp;
225 }
226 
227 void
228 nfs4_put_delegation(struct nfs4_delegation *dp)
229 {
230 	if (atomic_dec_and_test(&dp->dl_count)) {
231 		dprintk("NFSD: freeing dp %p\n",dp);
232 		put_nfs4_file(dp->dl_file);
233 		kmem_cache_free(deleg_slab, dp);
234 		num_delegations--;
235 	}
236 }
237 
238 /* Remove the associated file_lock first, then remove the delegation.
239  * lease_modify() is called to remove the FS_LEASE file_lock from
240  * the i_flock list, eventually calling nfsd's lock_manager
241  * fl_release_callback.
242  */
243 static void
244 nfs4_close_delegation(struct nfs4_delegation *dp)
245 {
246 	struct file *filp = dp->dl_vfs_file;
247 
248 	dprintk("NFSD: close_delegation dp %p\n",dp);
249 	dp->dl_vfs_file = NULL;
250 	/* The following nfsd_close may not actually close the file,
251 	 * but we want to remove the lease in any case. */
252 	if (dp->dl_flock)
253 		setlease(filp, F_UNLCK, &dp->dl_flock);
254 	nfsd_close(filp);
255 }
256 
257 /* Called under the state lock. */
258 static void
259 unhash_delegation(struct nfs4_delegation *dp)
260 {
261 	list_del_init(&dp->dl_perfile);
262 	list_del_init(&dp->dl_perclnt);
263 	spin_lock(&recall_lock);
264 	list_del_init(&dp->dl_recall_lru);
265 	spin_unlock(&recall_lock);
266 	nfs4_close_delegation(dp);
267 	nfs4_put_delegation(dp);
268 }
269 
270 /*
271  * SETCLIENTID state
272  */
273 
274 /* Hash tables for nfs4_clientid state */
275 #define CLIENT_HASH_BITS                 4
276 #define CLIENT_HASH_SIZE                (1 << CLIENT_HASH_BITS)
277 #define CLIENT_HASH_MASK                (CLIENT_HASH_SIZE - 1)
278 
279 #define clientid_hashval(id) \
280 	((id) & CLIENT_HASH_MASK)
281 #define clientstr_hashval(name) \
282 	(opaque_hashval((name), 8) & CLIENT_HASH_MASK)
283 /*
284  * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
285  * used in reboot/reset lease grace period processing
286  *
287  * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
288  * setclientid_confirmed info.
289  *
290  * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
291  * setclientid info.
292  *
293  * client_lru holds client queue ordered by nfs4_client.cl_time
294  * for lease renewal.
295  *
296  * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
297  * for last close replay.
298  */
299 static struct list_head	reclaim_str_hashtbl[CLIENT_HASH_SIZE];
300 static int reclaim_str_hashtbl_size = 0;
301 static struct list_head	conf_id_hashtbl[CLIENT_HASH_SIZE];
302 static struct list_head	conf_str_hashtbl[CLIENT_HASH_SIZE];
303 static struct list_head	unconf_str_hashtbl[CLIENT_HASH_SIZE];
304 static struct list_head	unconf_id_hashtbl[CLIENT_HASH_SIZE];
305 static struct list_head client_lru;
306 static struct list_head close_lru;
307 
308 static inline void
309 renew_client(struct nfs4_client *clp)
310 {
311 	/*
312 	* Move client to the end to the LRU list.
313 	*/
314 	dprintk("renewing client (clientid %08x/%08x)\n",
315 			clp->cl_clientid.cl_boot,
316 			clp->cl_clientid.cl_id);
317 	list_move_tail(&clp->cl_lru, &client_lru);
318 	clp->cl_time = get_seconds();
319 }
320 
321 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
322 static int
323 STALE_CLIENTID(clientid_t *clid)
324 {
325 	if (clid->cl_boot == boot_time)
326 		return 0;
327 	dprintk("NFSD stale clientid (%08x/%08x)\n",
328 			clid->cl_boot, clid->cl_id);
329 	return 1;
330 }
331 
332 /*
333  * XXX Should we use a slab cache ?
334  * This type of memory management is somewhat inefficient, but we use it
335  * anyway since SETCLIENTID is not a common operation.
336  */
337 static inline struct nfs4_client *
338 alloc_client(struct xdr_netobj name)
339 {
340 	struct nfs4_client *clp;
341 
342 	if ((clp = kmalloc(sizeof(struct nfs4_client), GFP_KERNEL))!= NULL) {
343 		memset(clp, 0, sizeof(*clp));
344 		if ((clp->cl_name.data = kmalloc(name.len, GFP_KERNEL)) != NULL) {
345 			memcpy(clp->cl_name.data, name.data, name.len);
346 			clp->cl_name.len = name.len;
347 		}
348 		else {
349 			kfree(clp);
350 			clp = NULL;
351 		}
352 	}
353 	return clp;
354 }
355 
356 static inline void
357 free_client(struct nfs4_client *clp)
358 {
359 	if (clp->cl_cred.cr_group_info)
360 		put_group_info(clp->cl_cred.cr_group_info);
361 	kfree(clp->cl_name.data);
362 	kfree(clp);
363 }
364 
365 void
366 put_nfs4_client(struct nfs4_client *clp)
367 {
368 	if (atomic_dec_and_test(&clp->cl_count))
369 		free_client(clp);
370 }
371 
372 static void
373 shutdown_callback_client(struct nfs4_client *clp)
374 {
375 	struct rpc_clnt *clnt = clp->cl_callback.cb_client;
376 
377 	/* shutdown rpc client, ending any outstanding recall rpcs */
378 	if (clnt) {
379 		clp->cl_callback.cb_client = NULL;
380 		rpc_shutdown_client(clnt);
381 		rpciod_down();
382 	}
383 }
384 
385 static void
386 expire_client(struct nfs4_client *clp)
387 {
388 	struct nfs4_stateowner *sop;
389 	struct nfs4_delegation *dp;
390 	struct list_head reaplist;
391 
392 	dprintk("NFSD: expire_client cl_count %d\n",
393 	                    atomic_read(&clp->cl_count));
394 
395 	shutdown_callback_client(clp);
396 
397 	INIT_LIST_HEAD(&reaplist);
398 	spin_lock(&recall_lock);
399 	while (!list_empty(&clp->cl_delegations)) {
400 		dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
401 		dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
402 				dp->dl_flock);
403 		list_del_init(&dp->dl_perclnt);
404 		list_move(&dp->dl_recall_lru, &reaplist);
405 	}
406 	spin_unlock(&recall_lock);
407 	while (!list_empty(&reaplist)) {
408 		dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
409 		list_del_init(&dp->dl_recall_lru);
410 		unhash_delegation(dp);
411 	}
412 	list_del(&clp->cl_idhash);
413 	list_del(&clp->cl_strhash);
414 	list_del(&clp->cl_lru);
415 	while (!list_empty(&clp->cl_openowners)) {
416 		sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
417 		release_stateowner(sop);
418 	}
419 	put_nfs4_client(clp);
420 }
421 
422 static struct nfs4_client *
423 create_client(struct xdr_netobj name, char *recdir) {
424 	struct nfs4_client *clp;
425 
426 	if (!(clp = alloc_client(name)))
427 		goto out;
428 	memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
429 	atomic_set(&clp->cl_count, 1);
430 	atomic_set(&clp->cl_callback.cb_set, 0);
431 	INIT_LIST_HEAD(&clp->cl_idhash);
432 	INIT_LIST_HEAD(&clp->cl_strhash);
433 	INIT_LIST_HEAD(&clp->cl_openowners);
434 	INIT_LIST_HEAD(&clp->cl_delegations);
435 	INIT_LIST_HEAD(&clp->cl_lru);
436 out:
437 	return clp;
438 }
439 
440 static void
441 copy_verf(struct nfs4_client *target, nfs4_verifier *source) {
442 	memcpy(target->cl_verifier.data, source->data, sizeof(target->cl_verifier.data));
443 }
444 
445 static void
446 copy_clid(struct nfs4_client *target, struct nfs4_client *source) {
447 	target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
448 	target->cl_clientid.cl_id = source->cl_clientid.cl_id;
449 }
450 
451 static void
452 copy_cred(struct svc_cred *target, struct svc_cred *source) {
453 
454 	target->cr_uid = source->cr_uid;
455 	target->cr_gid = source->cr_gid;
456 	target->cr_group_info = source->cr_group_info;
457 	get_group_info(target->cr_group_info);
458 }
459 
460 static inline int
461 same_name(const char *n1, const char *n2) {
462 	return 0 == memcmp(n1, n2, HEXDIR_LEN);
463 }
464 
465 static int
466 cmp_verf(nfs4_verifier *v1, nfs4_verifier *v2) {
467 	return(!memcmp(v1->data,v2->data,sizeof(v1->data)));
468 }
469 
470 static int
471 cmp_clid(clientid_t * cl1, clientid_t * cl2) {
472 	return((cl1->cl_boot == cl2->cl_boot) &&
473 	   	(cl1->cl_id == cl2->cl_id));
474 }
475 
476 /* XXX what about NGROUP */
477 static int
478 cmp_creds(struct svc_cred *cr1, struct svc_cred *cr2){
479 	return(cr1->cr_uid == cr2->cr_uid);
480 
481 }
482 
483 static void
484 gen_clid(struct nfs4_client *clp) {
485 	clp->cl_clientid.cl_boot = boot_time;
486 	clp->cl_clientid.cl_id = current_clientid++;
487 }
488 
489 static void
490 gen_confirm(struct nfs4_client *clp) {
491 	struct timespec 	tv;
492 	u32 *			p;
493 
494 	tv = CURRENT_TIME;
495 	p = (u32 *)clp->cl_confirm.data;
496 	*p++ = tv.tv_sec;
497 	*p++ = tv.tv_nsec;
498 }
499 
500 static int
501 check_name(struct xdr_netobj name) {
502 
503 	if (name.len == 0)
504 		return 0;
505 	if (name.len > NFS4_OPAQUE_LIMIT) {
506 		printk("NFSD: check_name: name too long(%d)!\n", name.len);
507 		return 0;
508 	}
509 	return 1;
510 }
511 
512 static void
513 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
514 {
515 	unsigned int idhashval;
516 
517 	list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
518 	idhashval = clientid_hashval(clp->cl_clientid.cl_id);
519 	list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
520 	list_add_tail(&clp->cl_lru, &client_lru);
521 	clp->cl_time = get_seconds();
522 }
523 
524 static void
525 move_to_confirmed(struct nfs4_client *clp)
526 {
527 	unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
528 	unsigned int strhashval;
529 
530 	dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
531 	list_del_init(&clp->cl_strhash);
532 	list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
533 	strhashval = clientstr_hashval(clp->cl_recdir);
534 	list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
535 	renew_client(clp);
536 }
537 
538 static struct nfs4_client *
539 find_confirmed_client(clientid_t *clid)
540 {
541 	struct nfs4_client *clp;
542 	unsigned int idhashval = clientid_hashval(clid->cl_id);
543 
544 	list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
545 		if (cmp_clid(&clp->cl_clientid, clid))
546 			return clp;
547 	}
548 	return NULL;
549 }
550 
551 static struct nfs4_client *
552 find_unconfirmed_client(clientid_t *clid)
553 {
554 	struct nfs4_client *clp;
555 	unsigned int idhashval = clientid_hashval(clid->cl_id);
556 
557 	list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
558 		if (cmp_clid(&clp->cl_clientid, clid))
559 			return clp;
560 	}
561 	return NULL;
562 }
563 
564 static struct nfs4_client *
565 find_confirmed_client_by_str(const char *dname, unsigned int hashval)
566 {
567 	struct nfs4_client *clp;
568 
569 	list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
570 		if (same_name(clp->cl_recdir, dname))
571 			return clp;
572 	}
573 	return NULL;
574 }
575 
576 static struct nfs4_client *
577 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval)
578 {
579 	struct nfs4_client *clp;
580 
581 	list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
582 		if (same_name(clp->cl_recdir, dname))
583 			return clp;
584 	}
585 	return NULL;
586 }
587 
588 /* a helper function for parse_callback */
589 static int
590 parse_octet(unsigned int *lenp, char **addrp)
591 {
592 	unsigned int len = *lenp;
593 	char *p = *addrp;
594 	int n = -1;
595 	char c;
596 
597 	for (;;) {
598 		if (!len)
599 			break;
600 		len--;
601 		c = *p++;
602 		if (c == '.')
603 			break;
604 		if ((c < '0') || (c > '9')) {
605 			n = -1;
606 			break;
607 		}
608 		if (n < 0)
609 			n = 0;
610 		n = (n * 10) + (c - '0');
611 		if (n > 255) {
612 			n = -1;
613 			break;
614 		}
615 	}
616 	*lenp = len;
617 	*addrp = p;
618 	return n;
619 }
620 
621 /* parse and set the setclientid ipv4 callback address */
622 static int
623 parse_ipv4(unsigned int addr_len, char *addr_val, unsigned int *cbaddrp, unsigned short *cbportp)
624 {
625 	int temp = 0;
626 	u32 cbaddr = 0;
627 	u16 cbport = 0;
628 	u32 addrlen = addr_len;
629 	char *addr = addr_val;
630 	int i, shift;
631 
632 	/* ipaddress */
633 	shift = 24;
634 	for(i = 4; i > 0  ; i--) {
635 		if ((temp = parse_octet(&addrlen, &addr)) < 0) {
636 			return 0;
637 		}
638 		cbaddr |= (temp << shift);
639 		if (shift > 0)
640 		shift -= 8;
641 	}
642 	*cbaddrp = cbaddr;
643 
644 	/* port */
645 	shift = 8;
646 	for(i = 2; i > 0  ; i--) {
647 		if ((temp = parse_octet(&addrlen, &addr)) < 0) {
648 			return 0;
649 		}
650 		cbport |= (temp << shift);
651 		if (shift > 0)
652 			shift -= 8;
653 	}
654 	*cbportp = cbport;
655 	return 1;
656 }
657 
658 static void
659 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se)
660 {
661 	struct nfs4_callback *cb = &clp->cl_callback;
662 
663 	/* Currently, we only support tcp for the callback channel */
664 	if ((se->se_callback_netid_len != 3) || memcmp((char *)se->se_callback_netid_val, "tcp", 3))
665 		goto out_err;
666 
667 	if ( !(parse_ipv4(se->se_callback_addr_len, se->se_callback_addr_val,
668 	                 &cb->cb_addr, &cb->cb_port)))
669 		goto out_err;
670 	cb->cb_prog = se->se_callback_prog;
671 	cb->cb_ident = se->se_callback_ident;
672 	return;
673 out_err:
674 	dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
675 		"will not receive delegations\n",
676 		clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
677 
678 	return;
679 }
680 
681 /*
682  * RFC 3010 has a complex implmentation description of processing a
683  * SETCLIENTID request consisting of 5 bullets, labeled as
684  * CASE0 - CASE4 below.
685  *
686  * NOTES:
687  * 	callback information will be processed in a future patch
688  *
689  *	an unconfirmed record is added when:
690  *      NORMAL (part of CASE 4): there is no confirmed nor unconfirmed record.
691  *	CASE 1: confirmed record found with matching name, principal,
692  *		verifier, and clientid.
693  *	CASE 2: confirmed record found with matching name, principal,
694  *		and there is no unconfirmed record with matching
695  *		name and principal
696  *
697  *      an unconfirmed record is replaced when:
698  *	CASE 3: confirmed record found with matching name, principal,
699  *		and an unconfirmed record is found with matching
700  *		name, principal, and with clientid and
701  *		confirm that does not match the confirmed record.
702  *	CASE 4: there is no confirmed record with matching name and
703  *		principal. there is an unconfirmed record with
704  *		matching name, principal.
705  *
706  *	an unconfirmed record is deleted when:
707  *	CASE 1: an unconfirmed record that matches input name, verifier,
708  *		and confirmed clientid.
709  *	CASE 4: any unconfirmed records with matching name and principal
710  *		that exist after an unconfirmed record has been replaced
711  *		as described above.
712  *
713  */
714 int
715 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_setclientid *setclid)
716 {
717 	u32 			ip_addr = rqstp->rq_addr.sin_addr.s_addr;
718 	struct xdr_netobj 	clname = {
719 		.len = setclid->se_namelen,
720 		.data = setclid->se_name,
721 	};
722 	nfs4_verifier		clverifier = setclid->se_verf;
723 	unsigned int 		strhashval;
724 	struct nfs4_client	*conf, *unconf, *new;
725 	int 			status;
726 	char                    dname[HEXDIR_LEN];
727 
728 	if (!check_name(clname))
729 		return nfserr_inval;
730 
731 	status = nfs4_make_rec_clidname(dname, &clname);
732 	if (status)
733 		return status;
734 
735 	/*
736 	 * XXX The Duplicate Request Cache (DRC) has been checked (??)
737 	 * We get here on a DRC miss.
738 	 */
739 
740 	strhashval = clientstr_hashval(dname);
741 
742 	nfs4_lock_state();
743 	conf = find_confirmed_client_by_str(dname, strhashval);
744 	if (conf) {
745 		/*
746 		 * CASE 0:
747 		 * clname match, confirmed, different principal
748 		 * or different ip_address
749 		 */
750 		status = nfserr_clid_inuse;
751 		if (!cmp_creds(&conf->cl_cred, &rqstp->rq_cred)
752 				|| conf->cl_addr != ip_addr) {
753 			printk("NFSD: setclientid: string in use by client"
754 			"(clientid %08x/%08x)\n",
755 			conf->cl_clientid.cl_boot, conf->cl_clientid.cl_id);
756 			goto out;
757 		}
758 	}
759 	unconf = find_unconfirmed_client_by_str(dname, strhashval);
760 	status = nfserr_resource;
761 	if (!conf) {
762 		/*
763 		 * CASE 4:
764 		 * placed first, because it is the normal case.
765 		 */
766 		if (unconf)
767 			expire_client(unconf);
768 		new = create_client(clname, dname);
769 		if (new == NULL)
770 			goto out;
771 		copy_verf(new, &clverifier);
772 		new->cl_addr = ip_addr;
773 		copy_cred(&new->cl_cred,&rqstp->rq_cred);
774 		gen_clid(new);
775 		gen_confirm(new);
776 		gen_callback(new, setclid);
777 		add_to_unconfirmed(new, strhashval);
778 	} else if (cmp_verf(&conf->cl_verifier, &clverifier)) {
779 		/*
780 		 * CASE 1:
781 		 * cl_name match, confirmed, principal match
782 		 * verifier match: probable callback update
783 		 *
784 		 * remove any unconfirmed nfs4_client with
785 		 * matching cl_name, cl_verifier, and cl_clientid
786 		 *
787 		 * create and insert an unconfirmed nfs4_client with same
788 		 * cl_name, cl_verifier, and cl_clientid as existing
789 		 * nfs4_client,  but with the new callback info and a
790 		 * new cl_confirm
791 		 */
792 		if (unconf) {
793 			/* Note this is removing unconfirmed {*x***},
794 			 * which is stronger than RFC recommended {vxc**}.
795 			 * This has the advantage that there is at most
796 			 * one {*x***} in either list at any time.
797 			 */
798 			expire_client(unconf);
799 		}
800 		new = create_client(clname, dname);
801 		if (new == NULL)
802 			goto out;
803 		copy_verf(new,&conf->cl_verifier);
804 		new->cl_addr = ip_addr;
805 		copy_cred(&new->cl_cred,&rqstp->rq_cred);
806 		copy_clid(new, conf);
807 		gen_confirm(new);
808 		gen_callback(new, setclid);
809 		add_to_unconfirmed(new,strhashval);
810 	} else if (!unconf) {
811 		/*
812 		 * CASE 2:
813 		 * clname match, confirmed, principal match
814 		 * verfier does not match
815 		 * no unconfirmed. create a new unconfirmed nfs4_client
816 		 * using input clverifier, clname, and callback info
817 		 * and generate a new cl_clientid and cl_confirm.
818 		 */
819 		new = create_client(clname, dname);
820 		if (new == NULL)
821 			goto out;
822 		copy_verf(new,&clverifier);
823 		new->cl_addr = ip_addr;
824 		copy_cred(&new->cl_cred,&rqstp->rq_cred);
825 		gen_clid(new);
826 		gen_confirm(new);
827 		gen_callback(new, setclid);
828 		add_to_unconfirmed(new, strhashval);
829 	} else if (!cmp_verf(&conf->cl_confirm, &unconf->cl_confirm)) {
830 		/*
831 		 * CASE3:
832 		 * confirmed found (name, principal match)
833 		 * confirmed verifier does not match input clverifier
834 		 *
835 		 * unconfirmed found (name match)
836 		 * confirmed->cl_confirm != unconfirmed->cl_confirm
837 		 *
838 		 * remove unconfirmed.
839 		 *
840 		 * create an unconfirmed nfs4_client
841 		 * with same cl_name as existing confirmed nfs4_client,
842 		 * but with new callback info, new cl_clientid,
843 		 * new cl_verifier and a new cl_confirm
844 		 */
845 		expire_client(unconf);
846 		new = create_client(clname, dname);
847 		if (new == NULL)
848 			goto out;
849 		copy_verf(new,&clverifier);
850 		new->cl_addr = ip_addr;
851 		copy_cred(&new->cl_cred,&rqstp->rq_cred);
852 		gen_clid(new);
853 		gen_confirm(new);
854 		gen_callback(new, setclid);
855 		add_to_unconfirmed(new, strhashval);
856 	} else {
857 		/* No cases hit !!! */
858 		status = nfserr_inval;
859 		goto out;
860 
861 	}
862 	setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
863 	setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
864 	memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
865 	status = nfs_ok;
866 out:
867 	nfs4_unlock_state();
868 	return status;
869 }
870 
871 
872 /*
873  * RFC 3010 has a complex implmentation description of processing a
874  * SETCLIENTID_CONFIRM request consisting of 4 bullets describing
875  * processing on a DRC miss, labeled as CASE1 - CASE4 below.
876  *
877  * NOTE: callback information will be processed here in a future patch
878  */
879 int
880 nfsd4_setclientid_confirm(struct svc_rqst *rqstp, struct nfsd4_setclientid_confirm *setclientid_confirm)
881 {
882 	u32 ip_addr = rqstp->rq_addr.sin_addr.s_addr;
883 	struct nfs4_client *conf, *unconf;
884 	nfs4_verifier confirm = setclientid_confirm->sc_confirm;
885 	clientid_t * clid = &setclientid_confirm->sc_clientid;
886 	int status;
887 
888 	if (STALE_CLIENTID(clid))
889 		return nfserr_stale_clientid;
890 	/*
891 	 * XXX The Duplicate Request Cache (DRC) has been checked (??)
892 	 * We get here on a DRC miss.
893 	 */
894 
895 	nfs4_lock_state();
896 
897 	conf = find_confirmed_client(clid);
898 	unconf = find_unconfirmed_client(clid);
899 
900 	status = nfserr_clid_inuse;
901 	if (conf && conf->cl_addr != ip_addr)
902 		goto out;
903 	if (unconf && unconf->cl_addr != ip_addr)
904 		goto out;
905 
906 	if ((conf && unconf) &&
907 	    (cmp_verf(&unconf->cl_confirm, &confirm)) &&
908 	    (cmp_verf(&conf->cl_verifier, &unconf->cl_verifier)) &&
909 	    (same_name(conf->cl_recdir,unconf->cl_recdir))  &&
910 	    (!cmp_verf(&conf->cl_confirm, &unconf->cl_confirm))) {
911 		/* CASE 1:
912 		* unconf record that matches input clientid and input confirm.
913 		* conf record that matches input clientid.
914 		* conf and unconf records match names, verifiers
915 		*/
916 		if (!cmp_creds(&conf->cl_cred, &unconf->cl_cred))
917 			status = nfserr_clid_inuse;
918 		else {
919 			/* XXX: We just turn off callbacks until we can handle
920 			  * change request correctly. */
921 			atomic_set(&conf->cl_callback.cb_set, 0);
922 			gen_confirm(conf);
923 			nfsd4_remove_clid_dir(unconf);
924 			expire_client(unconf);
925 			status = nfs_ok;
926 
927 		}
928 	} else if ((conf && !unconf) ||
929 	    ((conf && unconf) &&
930 	     (!cmp_verf(&conf->cl_verifier, &unconf->cl_verifier) ||
931 	      !same_name(conf->cl_recdir, unconf->cl_recdir)))) {
932 		/* CASE 2:
933 		 * conf record that matches input clientid.
934 		 * if unconf record matches input clientid, then
935 		 * unconf->cl_name or unconf->cl_verifier don't match the
936 		 * conf record.
937 		 */
938 		if (!cmp_creds(&conf->cl_cred,&rqstp->rq_cred))
939 			status = nfserr_clid_inuse;
940 		else
941 			status = nfs_ok;
942 	} else if (!conf && unconf
943 			&& cmp_verf(&unconf->cl_confirm, &confirm)) {
944 		/* CASE 3:
945 		 * conf record not found.
946 		 * unconf record found.
947 		 * unconf->cl_confirm matches input confirm
948 		 */
949 		if (!cmp_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
950 			status = nfserr_clid_inuse;
951 		} else {
952 			unsigned int hash =
953 				clientstr_hashval(unconf->cl_recdir);
954 			conf = find_confirmed_client_by_str(unconf->cl_recdir,
955 									hash);
956 			if (conf) {
957 				nfsd4_remove_clid_dir(conf);
958 				expire_client(conf);
959 			}
960 			move_to_confirmed(unconf);
961 			conf = unconf;
962 			status = nfs_ok;
963 		}
964 	} else if ((!conf || (conf && !cmp_verf(&conf->cl_confirm, &confirm)))
965 	    && (!unconf || (unconf && !cmp_verf(&unconf->cl_confirm,
966 				    				&confirm)))) {
967 		/* CASE 4:
968 		 * conf record not found, or if conf, conf->cl_confirm does not
969 		 * match input confirm.
970 		 * unconf record not found, or if unconf, unconf->cl_confirm
971 		 * does not match input confirm.
972 		 */
973 		status = nfserr_stale_clientid;
974 	} else {
975 		/* check that we have hit one of the cases...*/
976 		status = nfserr_clid_inuse;
977 	}
978 out:
979 	if (!status)
980 		nfsd4_probe_callback(conf);
981 	nfs4_unlock_state();
982 	return status;
983 }
984 
985 /* OPEN Share state helper functions */
986 static inline struct nfs4_file *
987 alloc_init_file(struct inode *ino)
988 {
989 	struct nfs4_file *fp;
990 	unsigned int hashval = file_hashval(ino);
991 
992 	fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
993 	if (fp) {
994 		kref_init(&fp->fi_ref);
995 		INIT_LIST_HEAD(&fp->fi_hash);
996 		INIT_LIST_HEAD(&fp->fi_stateids);
997 		INIT_LIST_HEAD(&fp->fi_delegations);
998 		list_add(&fp->fi_hash, &file_hashtbl[hashval]);
999 		fp->fi_inode = igrab(ino);
1000 		fp->fi_id = current_fileid++;
1001 		return fp;
1002 	}
1003 	return NULL;
1004 }
1005 
1006 static void
1007 nfsd4_free_slab(kmem_cache_t **slab)
1008 {
1009 	int status;
1010 
1011 	if (*slab == NULL)
1012 		return;
1013 	status = kmem_cache_destroy(*slab);
1014 	*slab = NULL;
1015 	WARN_ON(status);
1016 }
1017 
1018 static void
1019 nfsd4_free_slabs(void)
1020 {
1021 	nfsd4_free_slab(&stateowner_slab);
1022 	nfsd4_free_slab(&file_slab);
1023 	nfsd4_free_slab(&stateid_slab);
1024 	nfsd4_free_slab(&deleg_slab);
1025 }
1026 
1027 static int
1028 nfsd4_init_slabs(void)
1029 {
1030 	stateowner_slab = kmem_cache_create("nfsd4_stateowners",
1031 			sizeof(struct nfs4_stateowner), 0, 0, NULL, NULL);
1032 	if (stateowner_slab == NULL)
1033 		goto out_nomem;
1034 	file_slab = kmem_cache_create("nfsd4_files",
1035 			sizeof(struct nfs4_file), 0, 0, NULL, NULL);
1036 	if (file_slab == NULL)
1037 		goto out_nomem;
1038 	stateid_slab = kmem_cache_create("nfsd4_stateids",
1039 			sizeof(struct nfs4_stateid), 0, 0, NULL, NULL);
1040 	if (stateid_slab == NULL)
1041 		goto out_nomem;
1042 	deleg_slab = kmem_cache_create("nfsd4_delegations",
1043 			sizeof(struct nfs4_delegation), 0, 0, NULL, NULL);
1044 	if (deleg_slab == NULL)
1045 		goto out_nomem;
1046 	return 0;
1047 out_nomem:
1048 	nfsd4_free_slabs();
1049 	dprintk("nfsd4: out of memory while initializing nfsv4\n");
1050 	return -ENOMEM;
1051 }
1052 
1053 void
1054 nfs4_free_stateowner(struct kref *kref)
1055 {
1056 	struct nfs4_stateowner *sop =
1057 		container_of(kref, struct nfs4_stateowner, so_ref);
1058 	kfree(sop->so_owner.data);
1059 	kmem_cache_free(stateowner_slab, sop);
1060 }
1061 
1062 static inline struct nfs4_stateowner *
1063 alloc_stateowner(struct xdr_netobj *owner)
1064 {
1065 	struct nfs4_stateowner *sop;
1066 
1067 	if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1068 		if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1069 			memcpy(sop->so_owner.data, owner->data, owner->len);
1070 			sop->so_owner.len = owner->len;
1071 			kref_init(&sop->so_ref);
1072 			return sop;
1073 		}
1074 		kmem_cache_free(stateowner_slab, sop);
1075 	}
1076 	return NULL;
1077 }
1078 
1079 static struct nfs4_stateowner *
1080 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1081 	struct nfs4_stateowner *sop;
1082 	struct nfs4_replay *rp;
1083 	unsigned int idhashval;
1084 
1085 	if (!(sop = alloc_stateowner(&open->op_owner)))
1086 		return NULL;
1087 	idhashval = ownerid_hashval(current_ownerid);
1088 	INIT_LIST_HEAD(&sop->so_idhash);
1089 	INIT_LIST_HEAD(&sop->so_strhash);
1090 	INIT_LIST_HEAD(&sop->so_perclient);
1091 	INIT_LIST_HEAD(&sop->so_stateids);
1092 	INIT_LIST_HEAD(&sop->so_perstateid);  /* not used */
1093 	INIT_LIST_HEAD(&sop->so_close_lru);
1094 	sop->so_time = 0;
1095 	list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1096 	list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
1097 	list_add(&sop->so_perclient, &clp->cl_openowners);
1098 	sop->so_is_open_owner = 1;
1099 	sop->so_id = current_ownerid++;
1100 	sop->so_client = clp;
1101 	sop->so_seqid = open->op_seqid;
1102 	sop->so_confirmed = 0;
1103 	rp = &sop->so_replay;
1104 	rp->rp_status = nfserr_serverfault;
1105 	rp->rp_buflen = 0;
1106 	rp->rp_buf = rp->rp_ibuf;
1107 	return sop;
1108 }
1109 
1110 static void
1111 release_stateid_lockowners(struct nfs4_stateid *open_stp)
1112 {
1113 	struct nfs4_stateowner *lock_sop;
1114 
1115 	while (!list_empty(&open_stp->st_lockowners)) {
1116 		lock_sop = list_entry(open_stp->st_lockowners.next,
1117 				struct nfs4_stateowner, so_perstateid);
1118 		/* list_del(&open_stp->st_lockowners);  */
1119 		BUG_ON(lock_sop->so_is_open_owner);
1120 		release_stateowner(lock_sop);
1121 	}
1122 }
1123 
1124 static void
1125 unhash_stateowner(struct nfs4_stateowner *sop)
1126 {
1127 	struct nfs4_stateid *stp;
1128 
1129 	list_del(&sop->so_idhash);
1130 	list_del(&sop->so_strhash);
1131 	if (sop->so_is_open_owner)
1132 		list_del(&sop->so_perclient);
1133 	list_del(&sop->so_perstateid);
1134 	while (!list_empty(&sop->so_stateids)) {
1135 		stp = list_entry(sop->so_stateids.next,
1136 			struct nfs4_stateid, st_perstateowner);
1137 		if (sop->so_is_open_owner)
1138 			release_stateid(stp, OPEN_STATE);
1139 		else
1140 			release_stateid(stp, LOCK_STATE);
1141 	}
1142 }
1143 
1144 static void
1145 release_stateowner(struct nfs4_stateowner *sop)
1146 {
1147 	unhash_stateowner(sop);
1148 	list_del(&sop->so_close_lru);
1149 	nfs4_put_stateowner(sop);
1150 }
1151 
1152 static inline void
1153 init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1154 	struct nfs4_stateowner *sop = open->op_stateowner;
1155 	unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1156 
1157 	INIT_LIST_HEAD(&stp->st_hash);
1158 	INIT_LIST_HEAD(&stp->st_perstateowner);
1159 	INIT_LIST_HEAD(&stp->st_lockowners);
1160 	INIT_LIST_HEAD(&stp->st_perfile);
1161 	list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
1162 	list_add(&stp->st_perstateowner, &sop->so_stateids);
1163 	list_add(&stp->st_perfile, &fp->fi_stateids);
1164 	stp->st_stateowner = sop;
1165 	get_nfs4_file(fp);
1166 	stp->st_file = fp;
1167 	stp->st_stateid.si_boot = boot_time;
1168 	stp->st_stateid.si_stateownerid = sop->so_id;
1169 	stp->st_stateid.si_fileid = fp->fi_id;
1170 	stp->st_stateid.si_generation = 0;
1171 	stp->st_access_bmap = 0;
1172 	stp->st_deny_bmap = 0;
1173 	__set_bit(open->op_share_access, &stp->st_access_bmap);
1174 	__set_bit(open->op_share_deny, &stp->st_deny_bmap);
1175 	stp->st_openstp = NULL;
1176 }
1177 
1178 static void
1179 release_stateid(struct nfs4_stateid *stp, int flags)
1180 {
1181 	struct file *filp = stp->st_vfs_file;
1182 
1183 	list_del(&stp->st_hash);
1184 	list_del(&stp->st_perfile);
1185 	list_del(&stp->st_perstateowner);
1186 	if (flags & OPEN_STATE) {
1187 		release_stateid_lockowners(stp);
1188 		stp->st_vfs_file = NULL;
1189 		nfsd_close(filp);
1190 	} else if (flags & LOCK_STATE)
1191 		locks_remove_posix(filp, (fl_owner_t) stp->st_stateowner);
1192 	put_nfs4_file(stp->st_file);
1193 	kmem_cache_free(stateid_slab, stp);
1194 }
1195 
1196 static void
1197 move_to_close_lru(struct nfs4_stateowner *sop)
1198 {
1199 	dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1200 
1201 	list_move_tail(&sop->so_close_lru, &close_lru);
1202 	sop->so_time = get_seconds();
1203 }
1204 
1205 static int
1206 cmp_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner, clientid_t *clid) {
1207 	return ((sop->so_owner.len == owner->len) &&
1208 	 !memcmp(sop->so_owner.data, owner->data, owner->len) &&
1209 	  (sop->so_client->cl_clientid.cl_id == clid->cl_id));
1210 }
1211 
1212 static struct nfs4_stateowner *
1213 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1214 {
1215 	struct nfs4_stateowner *so = NULL;
1216 
1217 	list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
1218 		if (cmp_owner_str(so, &open->op_owner, &open->op_clientid))
1219 			return so;
1220 	}
1221 	return NULL;
1222 }
1223 
1224 /* search file_hashtbl[] for file */
1225 static struct nfs4_file *
1226 find_file(struct inode *ino)
1227 {
1228 	unsigned int hashval = file_hashval(ino);
1229 	struct nfs4_file *fp;
1230 
1231 	list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
1232 		if (fp->fi_inode == ino) {
1233 			get_nfs4_file(fp);
1234 			return fp;
1235 		}
1236 	}
1237 	return NULL;
1238 }
1239 
1240 static int access_valid(u32 x)
1241 {
1242 	return (x > 0 && x < 4);
1243 }
1244 
1245 static int deny_valid(u32 x)
1246 {
1247 	return (x >= 0 && x < 5);
1248 }
1249 
1250 static void
1251 set_access(unsigned int *access, unsigned long bmap) {
1252 	int i;
1253 
1254 	*access = 0;
1255 	for (i = 1; i < 4; i++) {
1256 		if (test_bit(i, &bmap))
1257 			*access |= i;
1258 	}
1259 }
1260 
1261 static void
1262 set_deny(unsigned int *deny, unsigned long bmap) {
1263 	int i;
1264 
1265 	*deny = 0;
1266 	for (i = 0; i < 4; i++) {
1267 		if (test_bit(i, &bmap))
1268 			*deny |= i ;
1269 	}
1270 }
1271 
1272 static int
1273 test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1274 	unsigned int access, deny;
1275 
1276 	set_access(&access, stp->st_access_bmap);
1277 	set_deny(&deny, stp->st_deny_bmap);
1278 	if ((access & open->op_share_deny) || (deny & open->op_share_access))
1279 		return 0;
1280 	return 1;
1281 }
1282 
1283 /*
1284  * Called to check deny when READ with all zero stateid or
1285  * WRITE with all zero or all one stateid
1286  */
1287 static int
1288 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
1289 {
1290 	struct inode *ino = current_fh->fh_dentry->d_inode;
1291 	struct nfs4_file *fp;
1292 	struct nfs4_stateid *stp;
1293 	int ret;
1294 
1295 	dprintk("NFSD: nfs4_share_conflict\n");
1296 
1297 	fp = find_file(ino);
1298 	if (!fp)
1299 		return nfs_ok;
1300 	ret = nfserr_locked;
1301 	/* Search for conflicting share reservations */
1302 	list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
1303 		if (test_bit(deny_type, &stp->st_deny_bmap) ||
1304 		    test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
1305 			goto out;
1306 	}
1307 	ret = nfs_ok;
1308 out:
1309 	put_nfs4_file(fp);
1310 	return ret;
1311 }
1312 
1313 static inline void
1314 nfs4_file_downgrade(struct file *filp, unsigned int share_access)
1315 {
1316 	if (share_access & NFS4_SHARE_ACCESS_WRITE) {
1317 		put_write_access(filp->f_dentry->d_inode);
1318 		filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
1319 	}
1320 }
1321 
1322 /*
1323  * Recall a delegation
1324  */
1325 static int
1326 do_recall(void *__dp)
1327 {
1328 	struct nfs4_delegation *dp = __dp;
1329 
1330 	daemonize("nfsv4-recall");
1331 
1332 	nfsd4_cb_recall(dp);
1333 	return 0;
1334 }
1335 
1336 /*
1337  * Spawn a thread to perform a recall on the delegation represented
1338  * by the lease (file_lock)
1339  *
1340  * Called from break_lease() with lock_kernel() held.
1341  * Note: we assume break_lease will only call this *once* for any given
1342  * lease.
1343  */
1344 static
1345 void nfsd_break_deleg_cb(struct file_lock *fl)
1346 {
1347 	struct nfs4_delegation *dp=  (struct nfs4_delegation *)fl->fl_owner;
1348 	struct task_struct *t;
1349 
1350 	dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
1351 	if (!dp)
1352 		return;
1353 
1354 	/* We're assuming the state code never drops its reference
1355 	 * without first removing the lease.  Since we're in this lease
1356 	 * callback (and since the lease code is serialized by the kernel
1357 	 * lock) we know the server hasn't removed the lease yet, we know
1358 	 * it's safe to take a reference: */
1359 	atomic_inc(&dp->dl_count);
1360 
1361 	spin_lock(&recall_lock);
1362 	list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
1363 	spin_unlock(&recall_lock);
1364 
1365 	/* only place dl_time is set. protected by lock_kernel*/
1366 	dp->dl_time = get_seconds();
1367 
1368 	/* XXX need to merge NFSD_LEASE_TIME with fs/locks.c:lease_break_time */
1369 	fl->fl_break_time = jiffies + NFSD_LEASE_TIME * HZ;
1370 
1371 	t = kthread_run(do_recall, dp, "%s", "nfs4_cb_recall");
1372 	if (IS_ERR(t)) {
1373 		struct nfs4_client *clp = dp->dl_client;
1374 
1375 		printk(KERN_INFO "NFSD: Callback thread failed for "
1376 			"for client (clientid %08x/%08x)\n",
1377 			clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1378 		nfs4_put_delegation(dp);
1379 	}
1380 }
1381 
1382 /*
1383  * The file_lock is being reapd.
1384  *
1385  * Called by locks_free_lock() with lock_kernel() held.
1386  */
1387 static
1388 void nfsd_release_deleg_cb(struct file_lock *fl)
1389 {
1390 	struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1391 
1392 	dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
1393 
1394 	if (!(fl->fl_flags & FL_LEASE) || !dp)
1395 		return;
1396 	dp->dl_flock = NULL;
1397 }
1398 
1399 /*
1400  * Set the delegation file_lock back pointer.
1401  *
1402  * Called from __setlease() with lock_kernel() held.
1403  */
1404 static
1405 void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
1406 {
1407 	struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
1408 
1409 	dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
1410 	if (!dp)
1411 		return;
1412 	dp->dl_flock = new;
1413 }
1414 
1415 /*
1416  * Called from __setlease() with lock_kernel() held
1417  */
1418 static
1419 int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
1420 {
1421 	struct nfs4_delegation *onlistd =
1422 		(struct nfs4_delegation *)onlist->fl_owner;
1423 	struct nfs4_delegation *tryd =
1424 		(struct nfs4_delegation *)try->fl_owner;
1425 
1426 	if (onlist->fl_lmops != try->fl_lmops)
1427 		return 0;
1428 
1429 	return onlistd->dl_client == tryd->dl_client;
1430 }
1431 
1432 
1433 static
1434 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
1435 {
1436 	if (arg & F_UNLCK)
1437 		return lease_modify(onlist, arg);
1438 	else
1439 		return -EAGAIN;
1440 }
1441 
1442 static struct lock_manager_operations nfsd_lease_mng_ops = {
1443 	.fl_break = nfsd_break_deleg_cb,
1444 	.fl_release_private = nfsd_release_deleg_cb,
1445 	.fl_copy_lock = nfsd_copy_lock_deleg_cb,
1446 	.fl_mylease = nfsd_same_client_deleg_cb,
1447 	.fl_change = nfsd_change_deleg_cb,
1448 };
1449 
1450 
1451 int
1452 nfsd4_process_open1(struct nfsd4_open *open)
1453 {
1454 	clientid_t *clientid = &open->op_clientid;
1455 	struct nfs4_client *clp = NULL;
1456 	unsigned int strhashval;
1457 	struct nfs4_stateowner *sop = NULL;
1458 
1459 	if (!check_name(open->op_owner))
1460 		return nfserr_inval;
1461 
1462 	if (STALE_CLIENTID(&open->op_clientid))
1463 		return nfserr_stale_clientid;
1464 
1465 	strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
1466 	sop = find_openstateowner_str(strhashval, open);
1467 	open->op_stateowner = sop;
1468 	if (!sop) {
1469 		/* Make sure the client's lease hasn't expired. */
1470 		clp = find_confirmed_client(clientid);
1471 		if (clp == NULL)
1472 			return nfserr_expired;
1473 		goto renew;
1474 	}
1475 	if (!sop->so_confirmed) {
1476 		/* Replace unconfirmed owners without checking for replay. */
1477 		clp = sop->so_client;
1478 		release_stateowner(sop);
1479 		open->op_stateowner = NULL;
1480 		goto renew;
1481 	}
1482 	if (open->op_seqid == sop->so_seqid - 1) {
1483 		if (sop->so_replay.rp_buflen)
1484 			return NFSERR_REPLAY_ME;
1485 		/* The original OPEN failed so spectacularly
1486 		 * that we don't even have replay data saved!
1487 		 * Therefore, we have no choice but to continue
1488 		 * processing this OPEN; presumably, we'll
1489 		 * fail again for the same reason.
1490 		 */
1491 		dprintk("nfsd4_process_open1: replay with no replay cache\n");
1492 		goto renew;
1493 	}
1494 	if (open->op_seqid != sop->so_seqid)
1495 		return nfserr_bad_seqid;
1496 renew:
1497 	if (open->op_stateowner == NULL) {
1498 		sop = alloc_init_open_stateowner(strhashval, clp, open);
1499 		if (sop == NULL)
1500 			return nfserr_resource;
1501 		open->op_stateowner = sop;
1502 	}
1503 	list_del_init(&sop->so_close_lru);
1504 	renew_client(sop->so_client);
1505 	return nfs_ok;
1506 }
1507 
1508 static inline int
1509 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
1510 {
1511 	if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
1512 		return nfserr_openmode;
1513 	else
1514 		return nfs_ok;
1515 }
1516 
1517 static struct nfs4_delegation *
1518 find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
1519 {
1520 	struct nfs4_delegation *dp;
1521 
1522 	list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
1523 		if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
1524 			return dp;
1525 	}
1526 	return NULL;
1527 }
1528 
1529 static int
1530 nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
1531 		struct nfs4_delegation **dp)
1532 {
1533 	int flags;
1534 	int status = nfserr_bad_stateid;
1535 
1536 	*dp = find_delegation_file(fp, &open->op_delegate_stateid);
1537 	if (*dp == NULL)
1538 		goto out;
1539 	flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
1540 						RD_STATE : WR_STATE;
1541 	status = nfs4_check_delegmode(*dp, flags);
1542 	if (status)
1543 		*dp = NULL;
1544 out:
1545 	if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
1546 		return nfs_ok;
1547 	if (status)
1548 		return status;
1549 	open->op_stateowner->so_confirmed = 1;
1550 	return nfs_ok;
1551 }
1552 
1553 static int
1554 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
1555 {
1556 	struct nfs4_stateid *local;
1557 	int status = nfserr_share_denied;
1558 	struct nfs4_stateowner *sop = open->op_stateowner;
1559 
1560 	list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1561 		/* ignore lock owners */
1562 		if (local->st_stateowner->so_is_open_owner == 0)
1563 			continue;
1564 		/* remember if we have seen this open owner */
1565 		if (local->st_stateowner == sop)
1566 			*stpp = local;
1567 		/* check for conflicting share reservations */
1568 		if (!test_share(local, open))
1569 			goto out;
1570 	}
1571 	status = 0;
1572 out:
1573 	return status;
1574 }
1575 
1576 static inline struct nfs4_stateid *
1577 nfs4_alloc_stateid(void)
1578 {
1579 	return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
1580 }
1581 
1582 static int
1583 nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
1584 		struct nfs4_delegation *dp,
1585 		struct svc_fh *cur_fh, int flags)
1586 {
1587 	struct nfs4_stateid *stp;
1588 
1589 	stp = nfs4_alloc_stateid();
1590 	if (stp == NULL)
1591 		return nfserr_resource;
1592 
1593 	if (dp) {
1594 		get_file(dp->dl_vfs_file);
1595 		stp->st_vfs_file = dp->dl_vfs_file;
1596 	} else {
1597 		int status;
1598 		status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
1599 				&stp->st_vfs_file);
1600 		if (status) {
1601 			if (status == nfserr_dropit)
1602 				status = nfserr_jukebox;
1603 			kmem_cache_free(stateid_slab, stp);
1604 			return status;
1605 		}
1606 	}
1607 	*stpp = stp;
1608 	return 0;
1609 }
1610 
1611 static inline int
1612 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
1613 		struct nfsd4_open *open)
1614 {
1615 	struct iattr iattr = {
1616 		.ia_valid = ATTR_SIZE,
1617 		.ia_size = 0,
1618 	};
1619 	if (!open->op_truncate)
1620 		return 0;
1621 	if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
1622 		return nfserr_inval;
1623 	return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
1624 }
1625 
1626 static int
1627 nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
1628 {
1629 	struct file *filp = stp->st_vfs_file;
1630 	struct inode *inode = filp->f_dentry->d_inode;
1631 	unsigned int share_access, new_writer;
1632 	int status;
1633 
1634 	set_access(&share_access, stp->st_access_bmap);
1635 	new_writer = (~share_access) & open->op_share_access
1636 			& NFS4_SHARE_ACCESS_WRITE;
1637 
1638 	if (new_writer) {
1639 		status = get_write_access(inode);
1640 		if (status)
1641 			return nfserrno(status);
1642 	}
1643 	status = nfsd4_truncate(rqstp, cur_fh, open);
1644 	if (status) {
1645 		if (new_writer)
1646 			put_write_access(inode);
1647 		return status;
1648 	}
1649 	/* remember the open */
1650 	filp->f_mode |= open->op_share_access;
1651 	set_bit(open->op_share_access, &stp->st_access_bmap);
1652 	set_bit(open->op_share_deny, &stp->st_deny_bmap);
1653 
1654 	return nfs_ok;
1655 }
1656 
1657 
1658 static void
1659 nfs4_set_claim_prev(struct nfsd4_open *open)
1660 {
1661 	open->op_stateowner->so_confirmed = 1;
1662 	open->op_stateowner->so_client->cl_firststate = 1;
1663 }
1664 
1665 /*
1666  * Attempt to hand out a delegation.
1667  */
1668 static void
1669 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
1670 {
1671 	struct nfs4_delegation *dp;
1672 	struct nfs4_stateowner *sop = stp->st_stateowner;
1673 	struct nfs4_callback *cb = &sop->so_client->cl_callback;
1674 	struct file_lock fl, *flp = &fl;
1675 	int status, flag = 0;
1676 
1677 	flag = NFS4_OPEN_DELEGATE_NONE;
1678 	open->op_recall = 0;
1679 	switch (open->op_claim_type) {
1680 		case NFS4_OPEN_CLAIM_PREVIOUS:
1681 			if (!atomic_read(&cb->cb_set))
1682 				open->op_recall = 1;
1683 			flag = open->op_delegate_type;
1684 			if (flag == NFS4_OPEN_DELEGATE_NONE)
1685 				goto out;
1686 			break;
1687 		case NFS4_OPEN_CLAIM_NULL:
1688 			/* Let's not give out any delegations till everyone's
1689 			 * had the chance to reclaim theirs.... */
1690 			if (nfs4_in_grace())
1691 				goto out;
1692 			if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
1693 				goto out;
1694 			if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
1695 				flag = NFS4_OPEN_DELEGATE_WRITE;
1696 			else
1697 				flag = NFS4_OPEN_DELEGATE_READ;
1698 			break;
1699 		default:
1700 			goto out;
1701 	}
1702 
1703 	dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
1704 	if (dp == NULL) {
1705 		flag = NFS4_OPEN_DELEGATE_NONE;
1706 		goto out;
1707 	}
1708 	locks_init_lock(&fl);
1709 	fl.fl_lmops = &nfsd_lease_mng_ops;
1710 	fl.fl_flags = FL_LEASE;
1711 	fl.fl_end = OFFSET_MAX;
1712 	fl.fl_owner =  (fl_owner_t)dp;
1713 	fl.fl_file = stp->st_vfs_file;
1714 	fl.fl_pid = current->tgid;
1715 
1716 	/* setlease checks to see if delegation should be handed out.
1717 	 * the lock_manager callbacks fl_mylease and fl_change are used
1718 	 */
1719 	if ((status = setlease(stp->st_vfs_file,
1720 		flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK, &flp))) {
1721 		dprintk("NFSD: setlease failed [%d], no delegation\n", status);
1722 		unhash_delegation(dp);
1723 		flag = NFS4_OPEN_DELEGATE_NONE;
1724 		goto out;
1725 	}
1726 
1727 	memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
1728 
1729 	dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
1730 	             dp->dl_stateid.si_boot,
1731 	             dp->dl_stateid.si_stateownerid,
1732 	             dp->dl_stateid.si_fileid,
1733 	             dp->dl_stateid.si_generation);
1734 out:
1735 	if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
1736 			&& flag == NFS4_OPEN_DELEGATE_NONE
1737 			&& open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
1738 		printk("NFSD: WARNING: refusing delegation reclaim\n");
1739 	open->op_delegate_type = flag;
1740 }
1741 
1742 /*
1743  * called with nfs4_lock_state() held.
1744  */
1745 int
1746 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
1747 {
1748 	struct nfs4_file *fp = NULL;
1749 	struct inode *ino = current_fh->fh_dentry->d_inode;
1750 	struct nfs4_stateid *stp = NULL;
1751 	struct nfs4_delegation *dp = NULL;
1752 	int status;
1753 
1754 	status = nfserr_inval;
1755 	if (!access_valid(open->op_share_access)
1756 			|| !deny_valid(open->op_share_deny))
1757 		goto out;
1758 	/*
1759 	 * Lookup file; if found, lookup stateid and check open request,
1760 	 * and check for delegations in the process of being recalled.
1761 	 * If not found, create the nfs4_file struct
1762 	 */
1763 	fp = find_file(ino);
1764 	if (fp) {
1765 		if ((status = nfs4_check_open(fp, open, &stp)))
1766 			goto out;
1767 		status = nfs4_check_deleg(fp, open, &dp);
1768 		if (status)
1769 			goto out;
1770 	} else {
1771 		status = nfserr_bad_stateid;
1772 		if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
1773 			goto out;
1774 		status = nfserr_resource;
1775 		fp = alloc_init_file(ino);
1776 		if (fp == NULL)
1777 			goto out;
1778 	}
1779 
1780 	/*
1781 	 * OPEN the file, or upgrade an existing OPEN.
1782 	 * If truncate fails, the OPEN fails.
1783 	 */
1784 	if (stp) {
1785 		/* Stateid was found, this is an OPEN upgrade */
1786 		status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
1787 		if (status)
1788 			goto out;
1789 		update_stateid(&stp->st_stateid);
1790 	} else {
1791 		/* Stateid was not found, this is a new OPEN */
1792 		int flags = 0;
1793 		if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
1794 			flags |= MAY_READ;
1795 		if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
1796 			flags |= MAY_WRITE;
1797 		status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
1798 		if (status)
1799 			goto out;
1800 		init_stateid(stp, fp, open);
1801 		status = nfsd4_truncate(rqstp, current_fh, open);
1802 		if (status) {
1803 			release_stateid(stp, OPEN_STATE);
1804 			goto out;
1805 		}
1806 	}
1807 	memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
1808 
1809 	/*
1810 	* Attempt to hand out a delegation. No error return, because the
1811 	* OPEN succeeds even if we fail.
1812 	*/
1813 	nfs4_open_delegation(current_fh, open, stp);
1814 
1815 	status = nfs_ok;
1816 
1817 	dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
1818 	            stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
1819 	            stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
1820 out:
1821 	if (fp)
1822 		put_nfs4_file(fp);
1823 	if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
1824 		nfs4_set_claim_prev(open);
1825 	/*
1826 	* To finish the open response, we just need to set the rflags.
1827 	*/
1828 	open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
1829 	if (!open->op_stateowner->so_confirmed)
1830 		open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
1831 
1832 	return status;
1833 }
1834 
1835 static struct workqueue_struct *laundry_wq;
1836 static struct work_struct laundromat_work;
1837 static void laundromat_main(void *);
1838 static DECLARE_WORK(laundromat_work, laundromat_main, NULL);
1839 
1840 int
1841 nfsd4_renew(clientid_t *clid)
1842 {
1843 	struct nfs4_client *clp;
1844 	int status;
1845 
1846 	nfs4_lock_state();
1847 	dprintk("process_renew(%08x/%08x): starting\n",
1848 			clid->cl_boot, clid->cl_id);
1849 	status = nfserr_stale_clientid;
1850 	if (STALE_CLIENTID(clid))
1851 		goto out;
1852 	clp = find_confirmed_client(clid);
1853 	status = nfserr_expired;
1854 	if (clp == NULL) {
1855 		/* We assume the client took too long to RENEW. */
1856 		dprintk("nfsd4_renew: clientid not found!\n");
1857 		goto out;
1858 	}
1859 	renew_client(clp);
1860 	status = nfserr_cb_path_down;
1861 	if (!list_empty(&clp->cl_delegations)
1862 			&& !atomic_read(&clp->cl_callback.cb_set))
1863 		goto out;
1864 	status = nfs_ok;
1865 out:
1866 	nfs4_unlock_state();
1867 	return status;
1868 }
1869 
1870 static void
1871 end_grace(void)
1872 {
1873 	dprintk("NFSD: end of grace period\n");
1874 	nfsd4_recdir_purge_old();
1875 	in_grace = 0;
1876 }
1877 
1878 static time_t
1879 nfs4_laundromat(void)
1880 {
1881 	struct nfs4_client *clp;
1882 	struct nfs4_stateowner *sop;
1883 	struct nfs4_delegation *dp;
1884 	struct list_head *pos, *next, reaplist;
1885 	time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
1886 	time_t t, clientid_val = NFSD_LEASE_TIME;
1887 	time_t u, test_val = NFSD_LEASE_TIME;
1888 
1889 	nfs4_lock_state();
1890 
1891 	dprintk("NFSD: laundromat service - starting\n");
1892 	if (in_grace)
1893 		end_grace();
1894 	list_for_each_safe(pos, next, &client_lru) {
1895 		clp = list_entry(pos, struct nfs4_client, cl_lru);
1896 		if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
1897 			t = clp->cl_time - cutoff;
1898 			if (clientid_val > t)
1899 				clientid_val = t;
1900 			break;
1901 		}
1902 		dprintk("NFSD: purging unused client (clientid %08x)\n",
1903 			clp->cl_clientid.cl_id);
1904 		nfsd4_remove_clid_dir(clp);
1905 		expire_client(clp);
1906 	}
1907 	INIT_LIST_HEAD(&reaplist);
1908 	spin_lock(&recall_lock);
1909 	list_for_each_safe(pos, next, &del_recall_lru) {
1910 		dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
1911 		if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
1912 			u = dp->dl_time - cutoff;
1913 			if (test_val > u)
1914 				test_val = u;
1915 			break;
1916 		}
1917 		dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
1918 			            dp, dp->dl_flock);
1919 		list_move(&dp->dl_recall_lru, &reaplist);
1920 	}
1921 	spin_unlock(&recall_lock);
1922 	list_for_each_safe(pos, next, &reaplist) {
1923 		dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
1924 		list_del_init(&dp->dl_recall_lru);
1925 		unhash_delegation(dp);
1926 	}
1927 	test_val = NFSD_LEASE_TIME;
1928 	list_for_each_safe(pos, next, &close_lru) {
1929 		sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
1930 		if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
1931 			u = sop->so_time - cutoff;
1932 			if (test_val > u)
1933 				test_val = u;
1934 			break;
1935 		}
1936 		dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
1937 			sop->so_id);
1938 		release_stateowner(sop);
1939 	}
1940 	if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
1941 		clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
1942 	nfs4_unlock_state();
1943 	return clientid_val;
1944 }
1945 
1946 void
1947 laundromat_main(void *not_used)
1948 {
1949 	time_t t;
1950 
1951 	t = nfs4_laundromat();
1952 	dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
1953 	queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1954 }
1955 
1956 static struct nfs4_stateowner *
1957 search_close_lru(u32 st_id, int flags)
1958 {
1959 	struct nfs4_stateowner *local = NULL;
1960 
1961 	if (flags & CLOSE_STATE) {
1962 		list_for_each_entry(local, &close_lru, so_close_lru) {
1963 			if (local->so_id == st_id)
1964 				return local;
1965 		}
1966 	}
1967 	return NULL;
1968 }
1969 
1970 static inline int
1971 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
1972 {
1973 	return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_dentry->d_inode;
1974 }
1975 
1976 static int
1977 STALE_STATEID(stateid_t *stateid)
1978 {
1979 	if (stateid->si_boot == boot_time)
1980 		return 0;
1981 	dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
1982 		stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
1983 		stateid->si_generation);
1984 	return 1;
1985 }
1986 
1987 static inline int
1988 access_permit_read(unsigned long access_bmap)
1989 {
1990 	return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
1991 		test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
1992 		test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
1993 }
1994 
1995 static inline int
1996 access_permit_write(unsigned long access_bmap)
1997 {
1998 	return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
1999 		test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2000 }
2001 
2002 static
2003 int nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
2004 {
2005         int status = nfserr_openmode;
2006 
2007 	if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2008                 goto out;
2009 	if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2010                 goto out;
2011 	status = nfs_ok;
2012 out:
2013 	return status;
2014 }
2015 
2016 static inline int
2017 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2018 {
2019 	/* Trying to call delegreturn with a special stateid? Yuch: */
2020 	if (!(flags & (RD_STATE | WR_STATE)))
2021 		return nfserr_bad_stateid;
2022 	else if (ONE_STATEID(stateid) && (flags & RD_STATE))
2023 		return nfs_ok;
2024 	else if (nfs4_in_grace()) {
2025 		/* Answer in remaining cases depends on existance of
2026 		 * conflicting state; so we must wait out the grace period. */
2027 		return nfserr_grace;
2028 	} else if (flags & WR_STATE)
2029 		return nfs4_share_conflict(current_fh,
2030 				NFS4_SHARE_DENY_WRITE);
2031 	else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2032 		return nfs4_share_conflict(current_fh,
2033 				NFS4_SHARE_DENY_READ);
2034 }
2035 
2036 /*
2037  * Allow READ/WRITE during grace period on recovered state only for files
2038  * that are not able to provide mandatory locking.
2039  */
2040 static inline int
2041 io_during_grace_disallowed(struct inode *inode, int flags)
2042 {
2043 	return nfs4_in_grace() && (flags & (RD_STATE | WR_STATE))
2044 		&& MANDATORY_LOCK(inode);
2045 }
2046 
2047 /*
2048 * Checks for stateid operations
2049 */
2050 int
2051 nfs4_preprocess_stateid_op(struct svc_fh *current_fh, stateid_t *stateid, int flags, struct file **filpp)
2052 {
2053 	struct nfs4_stateid *stp = NULL;
2054 	struct nfs4_delegation *dp = NULL;
2055 	stateid_t *stidp;
2056 	struct inode *ino = current_fh->fh_dentry->d_inode;
2057 	int status;
2058 
2059 	dprintk("NFSD: preprocess_stateid_op: stateid = (%08x/%08x/%08x/%08x)\n",
2060 		stateid->si_boot, stateid->si_stateownerid,
2061 		stateid->si_fileid, stateid->si_generation);
2062 	if (filpp)
2063 		*filpp = NULL;
2064 
2065 	if (io_during_grace_disallowed(ino, flags))
2066 		return nfserr_grace;
2067 
2068 	if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2069 		return check_special_stateids(current_fh, stateid, flags);
2070 
2071 	/* STALE STATEID */
2072 	status = nfserr_stale_stateid;
2073 	if (STALE_STATEID(stateid))
2074 		goto out;
2075 
2076 	/* BAD STATEID */
2077 	status = nfserr_bad_stateid;
2078 	if (!stateid->si_fileid) { /* delegation stateid */
2079 		if(!(dp = find_delegation_stateid(ino, stateid))) {
2080 			dprintk("NFSD: delegation stateid not found\n");
2081 			goto out;
2082 		}
2083 		stidp = &dp->dl_stateid;
2084 	} else { /* open or lock stateid */
2085 		if (!(stp = find_stateid(stateid, flags))) {
2086 			dprintk("NFSD: open or lock stateid not found\n");
2087 			goto out;
2088 		}
2089 		if ((flags & CHECK_FH) && nfs4_check_fh(current_fh, stp))
2090 			goto out;
2091 		if (!stp->st_stateowner->so_confirmed)
2092 			goto out;
2093 		stidp = &stp->st_stateid;
2094 	}
2095 	if (stateid->si_generation > stidp->si_generation)
2096 		goto out;
2097 
2098 	/* OLD STATEID */
2099 	status = nfserr_old_stateid;
2100 	if (stateid->si_generation < stidp->si_generation)
2101 		goto out;
2102 	if (stp) {
2103 		if ((status = nfs4_check_openmode(stp,flags)))
2104 			goto out;
2105 		renew_client(stp->st_stateowner->so_client);
2106 		if (filpp)
2107 			*filpp = stp->st_vfs_file;
2108 	} else if (dp) {
2109 		if ((status = nfs4_check_delegmode(dp, flags)))
2110 			goto out;
2111 		renew_client(dp->dl_client);
2112 		if (flags & DELEG_RET)
2113 			unhash_delegation(dp);
2114 		if (filpp)
2115 			*filpp = dp->dl_vfs_file;
2116 	}
2117 	status = nfs_ok;
2118 out:
2119 	return status;
2120 }
2121 
2122 static inline int
2123 setlkflg (int type)
2124 {
2125 	return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2126 		RD_STATE : WR_STATE;
2127 }
2128 
2129 /*
2130  * Checks for sequence id mutating operations.
2131  */
2132 static int
2133 nfs4_preprocess_seqid_op(struct svc_fh *current_fh, u32 seqid, stateid_t *stateid, int flags, struct nfs4_stateowner **sopp, struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
2134 {
2135 	struct nfs4_stateid *stp;
2136 	struct nfs4_stateowner *sop;
2137 
2138 	dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2139 			"stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2140 		stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2141 		stateid->si_generation);
2142 
2143 	*stpp = NULL;
2144 	*sopp = NULL;
2145 
2146 	if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2147 		printk("NFSD: preprocess_seqid_op: magic stateid!\n");
2148 		return nfserr_bad_stateid;
2149 	}
2150 
2151 	if (STALE_STATEID(stateid))
2152 		return nfserr_stale_stateid;
2153 	/*
2154 	* We return BAD_STATEID if filehandle doesn't match stateid,
2155 	* the confirmed flag is incorrecly set, or the generation
2156 	* number is incorrect.
2157 	*/
2158 	stp = find_stateid(stateid, flags);
2159 	if (stp == NULL) {
2160 		/*
2161 		 * Also, we should make sure this isn't just the result of
2162 		 * a replayed close:
2163 		 */
2164 		sop = search_close_lru(stateid->si_stateownerid, flags);
2165 		if (sop == NULL)
2166 			return nfserr_bad_stateid;
2167 		*sopp = sop;
2168 		goto check_replay;
2169 	}
2170 
2171 	if (lock) {
2172 		struct nfs4_stateowner *sop = stp->st_stateowner;
2173 		clientid_t *lockclid = &lock->v.new.clientid;
2174 		struct nfs4_client *clp = sop->so_client;
2175 		int lkflg = 0;
2176 		int status;
2177 
2178 		lkflg = setlkflg(lock->lk_type);
2179 
2180 		if (lock->lk_is_new) {
2181                        if (!sop->so_is_open_owner)
2182 			       return nfserr_bad_stateid;
2183                        if (!cmp_clid(&clp->cl_clientid, lockclid))
2184 			       return nfserr_bad_stateid;
2185                        /* stp is the open stateid */
2186                        status = nfs4_check_openmode(stp, lkflg);
2187                        if (status)
2188 			       return status;
2189                } else {
2190                        /* stp is the lock stateid */
2191                        status = nfs4_check_openmode(stp->st_openstp, lkflg);
2192                        if (status)
2193 			       return status;
2194                }
2195 
2196 	}
2197 
2198 	if ((flags & CHECK_FH) && nfs4_check_fh(current_fh, stp)) {
2199 		printk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
2200 		return nfserr_bad_stateid;
2201 	}
2202 
2203 	*stpp = stp;
2204 	*sopp = sop = stp->st_stateowner;
2205 
2206 	/*
2207 	*  We now validate the seqid and stateid generation numbers.
2208 	*  For the moment, we ignore the possibility of
2209 	*  generation number wraparound.
2210 	*/
2211 	if (seqid != sop->so_seqid)
2212 		goto check_replay;
2213 
2214 	if (sop->so_confirmed && flags & CONFIRM) {
2215 		printk("NFSD: preprocess_seqid_op: expected"
2216 				" unconfirmed stateowner!\n");
2217 		return nfserr_bad_stateid;
2218 	}
2219 	if (!sop->so_confirmed && !(flags & CONFIRM)) {
2220 		printk("NFSD: preprocess_seqid_op: stateowner not"
2221 				" confirmed yet!\n");
2222 		return nfserr_bad_stateid;
2223 	}
2224 	if (stateid->si_generation > stp->st_stateid.si_generation) {
2225 		printk("NFSD: preprocess_seqid_op: future stateid?!\n");
2226 		return nfserr_bad_stateid;
2227 	}
2228 
2229 	if (stateid->si_generation < stp->st_stateid.si_generation) {
2230 		printk("NFSD: preprocess_seqid_op: old stateid!\n");
2231 		return nfserr_old_stateid;
2232 	}
2233 	renew_client(sop->so_client);
2234 	return nfs_ok;
2235 
2236 check_replay:
2237 	if (seqid == sop->so_seqid - 1) {
2238 		dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
2239 		/* indicate replay to calling function */
2240 		return NFSERR_REPLAY_ME;
2241 	}
2242 	printk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
2243 			sop->so_seqid, seqid);
2244 	*sopp = NULL;
2245 	return nfserr_bad_seqid;
2246 }
2247 
2248 int
2249 nfsd4_open_confirm(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open_confirm *oc, struct nfs4_stateowner **replay_owner)
2250 {
2251 	int status;
2252 	struct nfs4_stateowner *sop;
2253 	struct nfs4_stateid *stp;
2254 
2255 	dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
2256 			(int)current_fh->fh_dentry->d_name.len,
2257 			current_fh->fh_dentry->d_name.name);
2258 
2259 	status = fh_verify(rqstp, current_fh, S_IFREG, 0);
2260 	if (status)
2261 		return status;
2262 
2263 	nfs4_lock_state();
2264 
2265 	if ((status = nfs4_preprocess_seqid_op(current_fh, oc->oc_seqid,
2266 					&oc->oc_req_stateid,
2267 					CHECK_FH | CONFIRM | OPEN_STATE,
2268 					&oc->oc_stateowner, &stp, NULL)))
2269 		goto out;
2270 
2271 	sop = oc->oc_stateowner;
2272 	sop->so_confirmed = 1;
2273 	update_stateid(&stp->st_stateid);
2274 	memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
2275 	dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
2276 		"stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
2277 		         stp->st_stateid.si_boot,
2278 		         stp->st_stateid.si_stateownerid,
2279 		         stp->st_stateid.si_fileid,
2280 		         stp->st_stateid.si_generation);
2281 
2282 	nfsd4_create_clid_dir(sop->so_client);
2283 out:
2284 	if (oc->oc_stateowner) {
2285 		nfs4_get_stateowner(oc->oc_stateowner);
2286 		*replay_owner = oc->oc_stateowner;
2287 	}
2288 	nfs4_unlock_state();
2289 	return status;
2290 }
2291 
2292 
2293 /*
2294  * unset all bits in union bitmap (bmap) that
2295  * do not exist in share (from successful OPEN_DOWNGRADE)
2296  */
2297 static void
2298 reset_union_bmap_access(unsigned long access, unsigned long *bmap)
2299 {
2300 	int i;
2301 	for (i = 1; i < 4; i++) {
2302 		if ((i & access) != i)
2303 			__clear_bit(i, bmap);
2304 	}
2305 }
2306 
2307 static void
2308 reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
2309 {
2310 	int i;
2311 	for (i = 0; i < 4; i++) {
2312 		if ((i & deny) != i)
2313 			__clear_bit(i, bmap);
2314 	}
2315 }
2316 
2317 int
2318 nfsd4_open_downgrade(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open_downgrade *od, struct nfs4_stateowner **replay_owner)
2319 {
2320 	int status;
2321 	struct nfs4_stateid *stp;
2322 	unsigned int share_access;
2323 
2324 	dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
2325 			(int)current_fh->fh_dentry->d_name.len,
2326 			current_fh->fh_dentry->d_name.name);
2327 
2328 	if (!access_valid(od->od_share_access)
2329 			|| !deny_valid(od->od_share_deny))
2330 		return nfserr_inval;
2331 
2332 	nfs4_lock_state();
2333 	if ((status = nfs4_preprocess_seqid_op(current_fh, od->od_seqid,
2334 					&od->od_stateid,
2335 					CHECK_FH | OPEN_STATE,
2336 					&od->od_stateowner, &stp, NULL)))
2337 		goto out;
2338 
2339 	status = nfserr_inval;
2340 	if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
2341 		dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
2342 			stp->st_access_bmap, od->od_share_access);
2343 		goto out;
2344 	}
2345 	if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
2346 		dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
2347 			stp->st_deny_bmap, od->od_share_deny);
2348 		goto out;
2349 	}
2350 	set_access(&share_access, stp->st_access_bmap);
2351 	nfs4_file_downgrade(stp->st_vfs_file,
2352 	                    share_access & ~od->od_share_access);
2353 
2354 	reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
2355 	reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
2356 
2357 	update_stateid(&stp->st_stateid);
2358 	memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
2359 	status = nfs_ok;
2360 out:
2361 	if (od->od_stateowner) {
2362 		nfs4_get_stateowner(od->od_stateowner);
2363 		*replay_owner = od->od_stateowner;
2364 	}
2365 	nfs4_unlock_state();
2366 	return status;
2367 }
2368 
2369 /*
2370  * nfs4_unlock_state() called after encode
2371  */
2372 int
2373 nfsd4_close(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_close *close, struct nfs4_stateowner **replay_owner)
2374 {
2375 	int status;
2376 	struct nfs4_stateid *stp;
2377 
2378 	dprintk("NFSD: nfsd4_close on file %.*s\n",
2379 			(int)current_fh->fh_dentry->d_name.len,
2380 			current_fh->fh_dentry->d_name.name);
2381 
2382 	nfs4_lock_state();
2383 	/* check close_lru for replay */
2384 	if ((status = nfs4_preprocess_seqid_op(current_fh, close->cl_seqid,
2385 					&close->cl_stateid,
2386 					CHECK_FH | OPEN_STATE | CLOSE_STATE,
2387 					&close->cl_stateowner, &stp, NULL)))
2388 		goto out;
2389 	status = nfs_ok;
2390 	update_stateid(&stp->st_stateid);
2391 	memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
2392 
2393 	/* release_stateid() calls nfsd_close() if needed */
2394 	release_stateid(stp, OPEN_STATE);
2395 
2396 	/* place unused nfs4_stateowners on so_close_lru list to be
2397 	 * released by the laundromat service after the lease period
2398 	 * to enable us to handle CLOSE replay
2399 	 */
2400 	if (list_empty(&close->cl_stateowner->so_stateids))
2401 		move_to_close_lru(close->cl_stateowner);
2402 out:
2403 	if (close->cl_stateowner) {
2404 		nfs4_get_stateowner(close->cl_stateowner);
2405 		*replay_owner = close->cl_stateowner;
2406 	}
2407 	nfs4_unlock_state();
2408 	return status;
2409 }
2410 
2411 int
2412 nfsd4_delegreturn(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_delegreturn *dr)
2413 {
2414 	int status;
2415 
2416 	if ((status = fh_verify(rqstp, current_fh, S_IFREG, 0)))
2417 		goto out;
2418 
2419 	nfs4_lock_state();
2420 	status = nfs4_preprocess_stateid_op(current_fh, &dr->dr_stateid, DELEG_RET, NULL);
2421 	nfs4_unlock_state();
2422 out:
2423 	return status;
2424 }
2425 
2426 
2427 /*
2428  * Lock owner state (byte-range locks)
2429  */
2430 #define LOFF_OVERFLOW(start, len)      ((u64)(len) > ~(u64)(start))
2431 #define LOCK_HASH_BITS              8
2432 #define LOCK_HASH_SIZE             (1 << LOCK_HASH_BITS)
2433 #define LOCK_HASH_MASK             (LOCK_HASH_SIZE - 1)
2434 
2435 #define lockownerid_hashval(id) \
2436         ((id) & LOCK_HASH_MASK)
2437 
2438 static inline unsigned int
2439 lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
2440 		struct xdr_netobj *ownername)
2441 {
2442 	return (file_hashval(inode) + cl_id
2443 			+ opaque_hashval(ownername->data, ownername->len))
2444 		& LOCK_HASH_MASK;
2445 }
2446 
2447 static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
2448 static struct list_head	lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
2449 static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
2450 
2451 static struct nfs4_stateid *
2452 find_stateid(stateid_t *stid, int flags)
2453 {
2454 	struct nfs4_stateid *local = NULL;
2455 	u32 st_id = stid->si_stateownerid;
2456 	u32 f_id = stid->si_fileid;
2457 	unsigned int hashval;
2458 
2459 	dprintk("NFSD: find_stateid flags 0x%x\n",flags);
2460 	if ((flags & LOCK_STATE) || (flags & RD_STATE) || (flags & WR_STATE)) {
2461 		hashval = stateid_hashval(st_id, f_id);
2462 		list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
2463 			if ((local->st_stateid.si_stateownerid == st_id) &&
2464 			    (local->st_stateid.si_fileid == f_id))
2465 				return local;
2466 		}
2467 	}
2468 	if ((flags & OPEN_STATE) || (flags & RD_STATE) || (flags & WR_STATE)) {
2469 		hashval = stateid_hashval(st_id, f_id);
2470 		list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
2471 			if ((local->st_stateid.si_stateownerid == st_id) &&
2472 			    (local->st_stateid.si_fileid == f_id))
2473 				return local;
2474 		}
2475 	}
2476 	return NULL;
2477 }
2478 
2479 static struct nfs4_delegation *
2480 find_delegation_stateid(struct inode *ino, stateid_t *stid)
2481 {
2482 	struct nfs4_file *fp;
2483 	struct nfs4_delegation *dl;
2484 
2485 	dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
2486                     stid->si_boot, stid->si_stateownerid,
2487                     stid->si_fileid, stid->si_generation);
2488 
2489 	fp = find_file(ino);
2490 	if (!fp)
2491 		return NULL;
2492 	dl = find_delegation_file(fp, stid);
2493 	put_nfs4_file(fp);
2494 	return dl;
2495 }
2496 
2497 /*
2498  * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
2499  * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
2500  * byte, because of sign extension problems.  Since NFSv4 calls for 64-bit
2501  * locking, this prevents us from being completely protocol-compliant.  The
2502  * real solution to this problem is to start using unsigned file offsets in
2503  * the VFS, but this is a very deep change!
2504  */
2505 static inline void
2506 nfs4_transform_lock_offset(struct file_lock *lock)
2507 {
2508 	if (lock->fl_start < 0)
2509 		lock->fl_start = OFFSET_MAX;
2510 	if (lock->fl_end < 0)
2511 		lock->fl_end = OFFSET_MAX;
2512 }
2513 
2514 /* Hack!: For now, we're defining this just so we can use a pointer to it
2515  * as a unique cookie to identify our (NFSv4's) posix locks. */
2516 static struct lock_manager_operations nfsd_posix_mng_ops  = {
2517 };
2518 
2519 static inline void
2520 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
2521 {
2522 	struct nfs4_stateowner *sop;
2523 	unsigned int hval;
2524 
2525 	if (fl->fl_lmops == &nfsd_posix_mng_ops) {
2526 		sop = (struct nfs4_stateowner *) fl->fl_owner;
2527 		hval = lockownerid_hashval(sop->so_id);
2528 		kref_get(&sop->so_ref);
2529 		deny->ld_sop = sop;
2530 		deny->ld_clientid = sop->so_client->cl_clientid;
2531 	} else {
2532 		deny->ld_sop = NULL;
2533 		deny->ld_clientid.cl_boot = 0;
2534 		deny->ld_clientid.cl_id = 0;
2535 	}
2536 	deny->ld_start = fl->fl_start;
2537 	deny->ld_length = ~(u64)0;
2538 	if (fl->fl_end != ~(u64)0)
2539 		deny->ld_length = fl->fl_end - fl->fl_start + 1;
2540 	deny->ld_type = NFS4_READ_LT;
2541 	if (fl->fl_type != F_RDLCK)
2542 		deny->ld_type = NFS4_WRITE_LT;
2543 }
2544 
2545 static struct nfs4_stateowner *
2546 find_lockstateowner_str(struct inode *inode, clientid_t *clid,
2547 		struct xdr_netobj *owner)
2548 {
2549 	unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
2550 	struct nfs4_stateowner *op;
2551 
2552 	list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
2553 		if (cmp_owner_str(op, owner, clid))
2554 			return op;
2555 	}
2556 	return NULL;
2557 }
2558 
2559 /*
2560  * Alloc a lock owner structure.
2561  * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
2562  * occured.
2563  *
2564  * strhashval = lock_ownerstr_hashval
2565  */
2566 
2567 static struct nfs4_stateowner *
2568 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
2569 	struct nfs4_stateowner *sop;
2570 	struct nfs4_replay *rp;
2571 	unsigned int idhashval;
2572 
2573 	if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
2574 		return NULL;
2575 	idhashval = lockownerid_hashval(current_ownerid);
2576 	INIT_LIST_HEAD(&sop->so_idhash);
2577 	INIT_LIST_HEAD(&sop->so_strhash);
2578 	INIT_LIST_HEAD(&sop->so_perclient);
2579 	INIT_LIST_HEAD(&sop->so_stateids);
2580 	INIT_LIST_HEAD(&sop->so_perstateid);
2581 	INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
2582 	sop->so_time = 0;
2583 	list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
2584 	list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
2585 	list_add(&sop->so_perstateid, &open_stp->st_lockowners);
2586 	sop->so_is_open_owner = 0;
2587 	sop->so_id = current_ownerid++;
2588 	sop->so_client = clp;
2589 	/* It is the openowner seqid that will be incremented in encode in the
2590 	 * case of new lockowners; so increment the lock seqid manually: */
2591 	sop->so_seqid = lock->lk_new_lock_seqid + 1;
2592 	sop->so_confirmed = 1;
2593 	rp = &sop->so_replay;
2594 	rp->rp_status = nfserr_serverfault;
2595 	rp->rp_buflen = 0;
2596 	rp->rp_buf = rp->rp_ibuf;
2597 	return sop;
2598 }
2599 
2600 static struct nfs4_stateid *
2601 alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
2602 {
2603 	struct nfs4_stateid *stp;
2604 	unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
2605 
2606 	stp = nfs4_alloc_stateid();
2607 	if (stp == NULL)
2608 		goto out;
2609 	INIT_LIST_HEAD(&stp->st_hash);
2610 	INIT_LIST_HEAD(&stp->st_perfile);
2611 	INIT_LIST_HEAD(&stp->st_perstateowner);
2612 	INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
2613 	list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
2614 	list_add(&stp->st_perfile, &fp->fi_stateids);
2615 	list_add(&stp->st_perstateowner, &sop->so_stateids);
2616 	stp->st_stateowner = sop;
2617 	get_nfs4_file(fp);
2618 	stp->st_file = fp;
2619 	stp->st_stateid.si_boot = boot_time;
2620 	stp->st_stateid.si_stateownerid = sop->so_id;
2621 	stp->st_stateid.si_fileid = fp->fi_id;
2622 	stp->st_stateid.si_generation = 0;
2623 	stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
2624 	stp->st_access_bmap = open_stp->st_access_bmap;
2625 	stp->st_deny_bmap = open_stp->st_deny_bmap;
2626 	stp->st_openstp = open_stp;
2627 
2628 out:
2629 	return stp;
2630 }
2631 
2632 static int
2633 check_lock_length(u64 offset, u64 length)
2634 {
2635 	return ((length == 0)  || ((length != ~(u64)0) &&
2636 	     LOFF_OVERFLOW(offset, length)));
2637 }
2638 
2639 /*
2640  *  LOCK operation
2641  */
2642 int
2643 nfsd4_lock(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_lock *lock, struct nfs4_stateowner **replay_owner)
2644 {
2645 	struct nfs4_stateowner *open_sop = NULL;
2646 	struct nfs4_stateowner *lock_sop = NULL;
2647 	struct nfs4_stateid *lock_stp;
2648 	struct file *filp;
2649 	struct file_lock file_lock;
2650 	struct file_lock conflock;
2651 	int status = 0;
2652 	unsigned int strhashval;
2653 
2654 	dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
2655 		(long long) lock->lk_offset,
2656 		(long long) lock->lk_length);
2657 
2658 	if (check_lock_length(lock->lk_offset, lock->lk_length))
2659 		 return nfserr_inval;
2660 
2661 	if ((status = fh_verify(rqstp, current_fh, S_IFREG, MAY_LOCK))) {
2662 		dprintk("NFSD: nfsd4_lock: permission denied!\n");
2663 		return status;
2664 	}
2665 
2666 	nfs4_lock_state();
2667 
2668 	if (lock->lk_is_new) {
2669 		/*
2670 		 * Client indicates that this is a new lockowner.
2671 		 * Use open owner and open stateid to create lock owner and
2672 		 * lock stateid.
2673 		 */
2674 		struct nfs4_stateid *open_stp = NULL;
2675 		struct nfs4_file *fp;
2676 
2677 		status = nfserr_stale_clientid;
2678 		if (STALE_CLIENTID(&lock->lk_new_clientid))
2679 			goto out;
2680 
2681 		/* validate and update open stateid and open seqid */
2682 		status = nfs4_preprocess_seqid_op(current_fh,
2683 				        lock->lk_new_open_seqid,
2684 		                        &lock->lk_new_open_stateid,
2685 		                        CHECK_FH | OPEN_STATE,
2686 		                        &lock->lk_replay_owner, &open_stp,
2687 					lock);
2688 		if (status)
2689 			goto out;
2690 		open_sop = lock->lk_replay_owner;
2691 		/* create lockowner and lock stateid */
2692 		fp = open_stp->st_file;
2693 		strhashval = lock_ownerstr_hashval(fp->fi_inode,
2694 				open_sop->so_client->cl_clientid.cl_id,
2695 				&lock->v.new.owner);
2696 		/* XXX: Do we need to check for duplicate stateowners on
2697 		 * the same file, or should they just be allowed (and
2698 		 * create new stateids)? */
2699 		status = nfserr_resource;
2700 		lock_sop = alloc_init_lock_stateowner(strhashval,
2701 				open_sop->so_client, open_stp, lock);
2702 		if (lock_sop == NULL)
2703 			goto out;
2704 		lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
2705 		if (lock_stp == NULL)
2706 			goto out;
2707 	} else {
2708 		/* lock (lock owner + lock stateid) already exists */
2709 		status = nfs4_preprocess_seqid_op(current_fh,
2710 				       lock->lk_old_lock_seqid,
2711 				       &lock->lk_old_lock_stateid,
2712 				       CHECK_FH | LOCK_STATE,
2713 				       &lock->lk_replay_owner, &lock_stp, lock);
2714 		if (status)
2715 			goto out;
2716 		lock_sop = lock->lk_replay_owner;
2717 	}
2718 	/* lock->lk_replay_owner and lock_stp have been created or found */
2719 	filp = lock_stp->st_vfs_file;
2720 
2721 	status = nfserr_grace;
2722 	if (nfs4_in_grace() && !lock->lk_reclaim)
2723 		goto out;
2724 	status = nfserr_no_grace;
2725 	if (!nfs4_in_grace() && lock->lk_reclaim)
2726 		goto out;
2727 
2728 	locks_init_lock(&file_lock);
2729 	switch (lock->lk_type) {
2730 		case NFS4_READ_LT:
2731 		case NFS4_READW_LT:
2732 			file_lock.fl_type = F_RDLCK;
2733 		break;
2734 		case NFS4_WRITE_LT:
2735 		case NFS4_WRITEW_LT:
2736 			file_lock.fl_type = F_WRLCK;
2737 		break;
2738 		default:
2739 			status = nfserr_inval;
2740 		goto out;
2741 	}
2742 	file_lock.fl_owner = (fl_owner_t)lock_sop;
2743 	file_lock.fl_pid = current->tgid;
2744 	file_lock.fl_file = filp;
2745 	file_lock.fl_flags = FL_POSIX;
2746 	file_lock.fl_lmops = &nfsd_posix_mng_ops;
2747 
2748 	file_lock.fl_start = lock->lk_offset;
2749 	if ((lock->lk_length == ~(u64)0) ||
2750 			LOFF_OVERFLOW(lock->lk_offset, lock->lk_length))
2751 		file_lock.fl_end = ~(u64)0;
2752 	else
2753 		file_lock.fl_end = lock->lk_offset + lock->lk_length - 1;
2754 	nfs4_transform_lock_offset(&file_lock);
2755 
2756 	/*
2757 	* Try to lock the file in the VFS.
2758 	* Note: locks.c uses the BKL to protect the inode's lock list.
2759 	*/
2760 
2761 	/* XXX?: Just to divert the locks_release_private at the start of
2762 	 * locks_copy_lock: */
2763 	conflock.fl_ops = NULL;
2764 	conflock.fl_lmops = NULL;
2765 	status = posix_lock_file_conf(filp, &file_lock, &conflock);
2766 	dprintk("NFSD: nfsd4_lock: posix_lock_file_conf status %d\n",status);
2767 	switch (-status) {
2768 	case 0: /* success! */
2769 		update_stateid(&lock_stp->st_stateid);
2770 		memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
2771 				sizeof(stateid_t));
2772 		break;
2773 	case (EAGAIN):		/* conflock holds conflicting lock */
2774 		status = nfserr_denied;
2775 		dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
2776 		nfs4_set_lock_denied(&conflock, &lock->lk_denied);
2777 		break;
2778 	case (EDEADLK):
2779 		status = nfserr_deadlock;
2780 		break;
2781 	default:
2782 		dprintk("NFSD: nfsd4_lock: posix_lock_file_conf() failed! status %d\n",status);
2783 		status = nfserr_resource;
2784 		break;
2785 	}
2786 out:
2787 	if (status && lock->lk_is_new && lock_sop)
2788 		release_stateowner(lock_sop);
2789 	if (lock->lk_replay_owner) {
2790 		nfs4_get_stateowner(lock->lk_replay_owner);
2791 		*replay_owner = lock->lk_replay_owner;
2792 	}
2793 	nfs4_unlock_state();
2794 	return status;
2795 }
2796 
2797 /*
2798  * LOCKT operation
2799  */
2800 int
2801 nfsd4_lockt(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_lockt *lockt)
2802 {
2803 	struct inode *inode;
2804 	struct file file;
2805 	struct file_lock file_lock;
2806 	struct file_lock conflock;
2807 	int status;
2808 
2809 	if (nfs4_in_grace())
2810 		return nfserr_grace;
2811 
2812 	if (check_lock_length(lockt->lt_offset, lockt->lt_length))
2813 		 return nfserr_inval;
2814 
2815 	lockt->lt_stateowner = NULL;
2816 	nfs4_lock_state();
2817 
2818 	status = nfserr_stale_clientid;
2819 	if (STALE_CLIENTID(&lockt->lt_clientid))
2820 		goto out;
2821 
2822 	if ((status = fh_verify(rqstp, current_fh, S_IFREG, 0))) {
2823 		dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
2824 		if (status == nfserr_symlink)
2825 			status = nfserr_inval;
2826 		goto out;
2827 	}
2828 
2829 	inode = current_fh->fh_dentry->d_inode;
2830 	locks_init_lock(&file_lock);
2831 	switch (lockt->lt_type) {
2832 		case NFS4_READ_LT:
2833 		case NFS4_READW_LT:
2834 			file_lock.fl_type = F_RDLCK;
2835 		break;
2836 		case NFS4_WRITE_LT:
2837 		case NFS4_WRITEW_LT:
2838 			file_lock.fl_type = F_WRLCK;
2839 		break;
2840 		default:
2841 			printk("NFSD: nfs4_lockt: bad lock type!\n");
2842 			status = nfserr_inval;
2843 		goto out;
2844 	}
2845 
2846 	lockt->lt_stateowner = find_lockstateowner_str(inode,
2847 			&lockt->lt_clientid, &lockt->lt_owner);
2848 	if (lockt->lt_stateowner)
2849 		file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
2850 	file_lock.fl_pid = current->tgid;
2851 	file_lock.fl_flags = FL_POSIX;
2852 	file_lock.fl_lmops = &nfsd_posix_mng_ops;
2853 
2854 	file_lock.fl_start = lockt->lt_offset;
2855 	if ((lockt->lt_length == ~(u64)0) || LOFF_OVERFLOW(lockt->lt_offset, lockt->lt_length))
2856 		file_lock.fl_end = ~(u64)0;
2857 	else
2858 		file_lock.fl_end = lockt->lt_offset + lockt->lt_length - 1;
2859 
2860 	nfs4_transform_lock_offset(&file_lock);
2861 
2862 	/* posix_test_lock uses the struct file _only_ to resolve the inode.
2863 	 * since LOCKT doesn't require an OPEN, and therefore a struct
2864 	 * file may not exist, pass posix_test_lock a struct file with
2865 	 * only the dentry:inode set.
2866 	 */
2867 	memset(&file, 0, sizeof (struct file));
2868 	file.f_dentry = current_fh->fh_dentry;
2869 
2870 	status = nfs_ok;
2871 	if (posix_test_lock(&file, &file_lock, &conflock)) {
2872 		status = nfserr_denied;
2873 		nfs4_set_lock_denied(&conflock, &lockt->lt_denied);
2874 	}
2875 out:
2876 	nfs4_unlock_state();
2877 	return status;
2878 }
2879 
2880 int
2881 nfsd4_locku(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_locku *locku, struct nfs4_stateowner **replay_owner)
2882 {
2883 	struct nfs4_stateid *stp;
2884 	struct file *filp = NULL;
2885 	struct file_lock file_lock;
2886 	int status;
2887 
2888 	dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
2889 		(long long) locku->lu_offset,
2890 		(long long) locku->lu_length);
2891 
2892 	if (check_lock_length(locku->lu_offset, locku->lu_length))
2893 		 return nfserr_inval;
2894 
2895 	nfs4_lock_state();
2896 
2897 	if ((status = nfs4_preprocess_seqid_op(current_fh,
2898 					locku->lu_seqid,
2899 					&locku->lu_stateid,
2900 					CHECK_FH | LOCK_STATE,
2901 					&locku->lu_stateowner, &stp, NULL)))
2902 		goto out;
2903 
2904 	filp = stp->st_vfs_file;
2905 	BUG_ON(!filp);
2906 	locks_init_lock(&file_lock);
2907 	file_lock.fl_type = F_UNLCK;
2908 	file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
2909 	file_lock.fl_pid = current->tgid;
2910 	file_lock.fl_file = filp;
2911 	file_lock.fl_flags = FL_POSIX;
2912 	file_lock.fl_lmops = &nfsd_posix_mng_ops;
2913 	file_lock.fl_start = locku->lu_offset;
2914 
2915 	if ((locku->lu_length == ~(u64)0) || LOFF_OVERFLOW(locku->lu_offset, locku->lu_length))
2916 		file_lock.fl_end = ~(u64)0;
2917 	else
2918 		file_lock.fl_end = locku->lu_offset + locku->lu_length - 1;
2919 	nfs4_transform_lock_offset(&file_lock);
2920 
2921 	/*
2922 	*  Try to unlock the file in the VFS.
2923 	*/
2924 	status = posix_lock_file(filp, &file_lock);
2925 	if (status) {
2926 		dprintk("NFSD: nfs4_locku: posix_lock_file failed!\n");
2927 		goto out_nfserr;
2928 	}
2929 	/*
2930 	* OK, unlock succeeded; the only thing left to do is update the stateid.
2931 	*/
2932 	update_stateid(&stp->st_stateid);
2933 	memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
2934 
2935 out:
2936 	if (locku->lu_stateowner) {
2937 		nfs4_get_stateowner(locku->lu_stateowner);
2938 		*replay_owner = locku->lu_stateowner;
2939 	}
2940 	nfs4_unlock_state();
2941 	return status;
2942 
2943 out_nfserr:
2944 	status = nfserrno(status);
2945 	goto out;
2946 }
2947 
2948 /*
2949  * returns
2950  * 	1: locks held by lockowner
2951  * 	0: no locks held by lockowner
2952  */
2953 static int
2954 check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
2955 {
2956 	struct file_lock **flpp;
2957 	struct inode *inode = filp->f_dentry->d_inode;
2958 	int status = 0;
2959 
2960 	lock_kernel();
2961 	for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
2962 		if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
2963 			status = 1;
2964 			goto out;
2965 		}
2966 	}
2967 out:
2968 	unlock_kernel();
2969 	return status;
2970 }
2971 
2972 int
2973 nfsd4_release_lockowner(struct svc_rqst *rqstp, struct nfsd4_release_lockowner *rlockowner)
2974 {
2975 	clientid_t *clid = &rlockowner->rl_clientid;
2976 	struct nfs4_stateowner *sop;
2977 	struct nfs4_stateid *stp;
2978 	struct xdr_netobj *owner = &rlockowner->rl_owner;
2979 	struct list_head matches;
2980 	int i;
2981 	int status;
2982 
2983 	dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
2984 		clid->cl_boot, clid->cl_id);
2985 
2986 	/* XXX check for lease expiration */
2987 
2988 	status = nfserr_stale_clientid;
2989 	if (STALE_CLIENTID(clid))
2990 		return status;
2991 
2992 	nfs4_lock_state();
2993 
2994 	status = nfserr_locks_held;
2995 	/* XXX: we're doing a linear search through all the lockowners.
2996 	 * Yipes!  For now we'll just hope clients aren't really using
2997 	 * release_lockowner much, but eventually we have to fix these
2998 	 * data structures. */
2999 	INIT_LIST_HEAD(&matches);
3000 	for (i = 0; i < LOCK_HASH_SIZE; i++) {
3001 		list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
3002 			if (!cmp_owner_str(sop, owner, clid))
3003 				continue;
3004 			list_for_each_entry(stp, &sop->so_stateids,
3005 					st_perstateowner) {
3006 				if (check_for_locks(stp->st_vfs_file, sop))
3007 					goto out;
3008 				/* Note: so_perclient unused for lockowners,
3009 				 * so it's OK to fool with here. */
3010 				list_add(&sop->so_perclient, &matches);
3011 			}
3012 		}
3013 	}
3014 	/* Clients probably won't expect us to return with some (but not all)
3015 	 * of the lockowner state released; so don't release any until all
3016 	 * have been checked. */
3017 	status = nfs_ok;
3018 	while (!list_empty(&matches)) {
3019 		sop = list_entry(matches.next, struct nfs4_stateowner,
3020 								so_perclient);
3021 		/* unhash_stateowner deletes so_perclient only
3022 		 * for openowners. */
3023 		list_del(&sop->so_perclient);
3024 		release_stateowner(sop);
3025 	}
3026 out:
3027 	nfs4_unlock_state();
3028 	return status;
3029 }
3030 
3031 static inline struct nfs4_client_reclaim *
3032 alloc_reclaim(void)
3033 {
3034 	return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
3035 }
3036 
3037 int
3038 nfs4_has_reclaimed_state(const char *name)
3039 {
3040 	unsigned int strhashval = clientstr_hashval(name);
3041 	struct nfs4_client *clp;
3042 
3043 	clp = find_confirmed_client_by_str(name, strhashval);
3044 	return clp ? 1 : 0;
3045 }
3046 
3047 /*
3048  * failure => all reset bets are off, nfserr_no_grace...
3049  */
3050 int
3051 nfs4_client_to_reclaim(const char *name)
3052 {
3053 	unsigned int strhashval;
3054 	struct nfs4_client_reclaim *crp = NULL;
3055 
3056 	dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3057 	crp = alloc_reclaim();
3058 	if (!crp)
3059 		return 0;
3060 	strhashval = clientstr_hashval(name);
3061 	INIT_LIST_HEAD(&crp->cr_strhash);
3062 	list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
3063 	memcpy(crp->cr_recdir, name, HEXDIR_LEN);
3064 	reclaim_str_hashtbl_size++;
3065 	return 1;
3066 }
3067 
3068 static void
3069 nfs4_release_reclaim(void)
3070 {
3071 	struct nfs4_client_reclaim *crp = NULL;
3072 	int i;
3073 
3074 	for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3075 		while (!list_empty(&reclaim_str_hashtbl[i])) {
3076 			crp = list_entry(reclaim_str_hashtbl[i].next,
3077 			                struct nfs4_client_reclaim, cr_strhash);
3078 			list_del(&crp->cr_strhash);
3079 			kfree(crp);
3080 			reclaim_str_hashtbl_size--;
3081 		}
3082 	}
3083 	BUG_ON(reclaim_str_hashtbl_size);
3084 }
3085 
3086 /*
3087  * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
3088 static struct nfs4_client_reclaim *
3089 nfs4_find_reclaim_client(clientid_t *clid)
3090 {
3091 	unsigned int strhashval;
3092 	struct nfs4_client *clp;
3093 	struct nfs4_client_reclaim *crp = NULL;
3094 
3095 
3096 	/* find clientid in conf_id_hashtbl */
3097 	clp = find_confirmed_client(clid);
3098 	if (clp == NULL)
3099 		return NULL;
3100 
3101 	dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3102 		            clp->cl_name.len, clp->cl_name.data,
3103 			    clp->cl_recdir);
3104 
3105 	/* find clp->cl_name in reclaim_str_hashtbl */
3106 	strhashval = clientstr_hashval(clp->cl_recdir);
3107 	list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
3108 		if (same_name(crp->cr_recdir, clp->cl_recdir)) {
3109 			return crp;
3110 		}
3111 	}
3112 	return NULL;
3113 }
3114 
3115 /*
3116 * Called from OPEN. Look for clientid in reclaim list.
3117 */
3118 int
3119 nfs4_check_open_reclaim(clientid_t *clid)
3120 {
3121 	return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
3122 }
3123 
3124 /* initialization to perform at module load time: */
3125 
3126 void
3127 nfs4_state_init(void)
3128 {
3129 	int i;
3130 
3131 	for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3132 		INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3133 		INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3134 		INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3135 		INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
3136 	}
3137 	for (i = 0; i < FILE_HASH_SIZE; i++) {
3138 		INIT_LIST_HEAD(&file_hashtbl[i]);
3139 	}
3140 	for (i = 0; i < OWNER_HASH_SIZE; i++) {
3141 		INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3142 		INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3143 	}
3144 	for (i = 0; i < STATEID_HASH_SIZE; i++) {
3145 		INIT_LIST_HEAD(&stateid_hashtbl[i]);
3146 		INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3147 	}
3148 	for (i = 0; i < LOCK_HASH_SIZE; i++) {
3149 		INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3150 		INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3151 	}
3152 	memset(&onestateid, ~0, sizeof(stateid_t));
3153 	INIT_LIST_HEAD(&close_lru);
3154 	INIT_LIST_HEAD(&client_lru);
3155 	INIT_LIST_HEAD(&del_recall_lru);
3156 	for (i = 0; i < CLIENT_HASH_SIZE; i++)
3157 		INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
3158 	reclaim_str_hashtbl_size = 0;
3159 }
3160 
3161 static void
3162 nfsd4_load_reboot_recovery_data(void)
3163 {
3164 	int status;
3165 
3166 	nfs4_lock_state();
3167 	nfsd4_init_recdir(user_recovery_dirname);
3168 	status = nfsd4_recdir_load();
3169 	nfs4_unlock_state();
3170 	if (status)
3171 		printk("NFSD: Failure reading reboot recovery data\n");
3172 }
3173 
3174 /* initialization to perform when the nfsd service is started: */
3175 
3176 static void
3177 __nfs4_state_start(void)
3178 {
3179 	time_t grace_time;
3180 
3181 	boot_time = get_seconds();
3182 	grace_time = max(user_lease_time, lease_time);
3183 	lease_time = user_lease_time;
3184 	in_grace = 1;
3185 	printk("NFSD: starting %ld-second grace period\n", grace_time);
3186 	laundry_wq = create_singlethread_workqueue("nfsd4");
3187 	queue_delayed_work(laundry_wq, &laundromat_work, grace_time*HZ);
3188 }
3189 
3190 int
3191 nfs4_state_start(void)
3192 {
3193 	int status;
3194 
3195 	if (nfs4_init)
3196 		return 0;
3197 	status = nfsd4_init_slabs();
3198 	if (status)
3199 		return status;
3200 	nfsd4_load_reboot_recovery_data();
3201 	__nfs4_state_start();
3202 	nfs4_init = 1;
3203 	return 0;
3204 }
3205 
3206 int
3207 nfs4_in_grace(void)
3208 {
3209 	return in_grace;
3210 }
3211 
3212 time_t
3213 nfs4_lease_time(void)
3214 {
3215 	return lease_time;
3216 }
3217 
3218 static void
3219 __nfs4_state_shutdown(void)
3220 {
3221 	int i;
3222 	struct nfs4_client *clp = NULL;
3223 	struct nfs4_delegation *dp = NULL;
3224 	struct list_head *pos, *next, reaplist;
3225 
3226 	for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3227 		while (!list_empty(&conf_id_hashtbl[i])) {
3228 			clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
3229 			expire_client(clp);
3230 		}
3231 		while (!list_empty(&unconf_str_hashtbl[i])) {
3232 			clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
3233 			expire_client(clp);
3234 		}
3235 	}
3236 	INIT_LIST_HEAD(&reaplist);
3237 	spin_lock(&recall_lock);
3238 	list_for_each_safe(pos, next, &del_recall_lru) {
3239 		dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3240 		list_move(&dp->dl_recall_lru, &reaplist);
3241 	}
3242 	spin_unlock(&recall_lock);
3243 	list_for_each_safe(pos, next, &reaplist) {
3244 		dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3245 		list_del_init(&dp->dl_recall_lru);
3246 		unhash_delegation(dp);
3247 	}
3248 
3249 	cancel_delayed_work(&laundromat_work);
3250 	nfsd4_shutdown_recdir();
3251 	nfs4_init = 0;
3252 }
3253 
3254 void
3255 nfs4_state_shutdown(void)
3256 {
3257 	cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
3258 	destroy_workqueue(laundry_wq);
3259 	nfs4_lock_state();
3260 	nfs4_release_reclaim();
3261 	__nfs4_state_shutdown();
3262 	nfsd4_free_slabs();
3263 	nfs4_unlock_state();
3264 }
3265 
3266 static void
3267 nfs4_set_recdir(char *recdir)
3268 {
3269 	nfs4_lock_state();
3270 	strcpy(user_recovery_dirname, recdir);
3271 	nfs4_unlock_state();
3272 }
3273 
3274 /*
3275  * Change the NFSv4 recovery directory to recdir.
3276  */
3277 int
3278 nfs4_reset_recoverydir(char *recdir)
3279 {
3280 	int status;
3281 	struct nameidata nd;
3282 
3283 	status = path_lookup(recdir, LOOKUP_FOLLOW, &nd);
3284 	if (status)
3285 		return status;
3286 	status = -ENOTDIR;
3287 	if (S_ISDIR(nd.dentry->d_inode->i_mode)) {
3288 		nfs4_set_recdir(recdir);
3289 		status = 0;
3290 	}
3291 	path_release(&nd);
3292 	return status;
3293 }
3294 
3295 /*
3296  * Called when leasetime is changed.
3297  *
3298  * The only way the protocol gives us to handle on-the-fly lease changes is to
3299  * simulate a reboot.  Instead of doing that, we just wait till the next time
3300  * we start to register any changes in lease time.  If the administrator
3301  * really wants to change the lease time *now*, they can go ahead and bring
3302  * nfsd down and then back up again after changing the lease time.
3303  */
3304 void
3305 nfs4_reset_lease(time_t leasetime)
3306 {
3307 	lock_kernel();
3308 	user_lease_time = leasetime;
3309 	unlock_kernel();
3310 }
3311