xref: /openbmc/linux/fs/nfsd/nfs4state.c (revision 8fa5723aa7e053d498336b48448b292fc2e0458b)
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/file.h>
45 #include <linux/mount.h>
46 #include <linux/workqueue.h>
47 #include <linux/smp_lock.h>
48 #include <linux/kthread.h>
49 #include <linux/nfs4.h>
50 #include <linux/nfsd/state.h>
51 #include <linux/nfsd/xdr4.h>
52 #include <linux/namei.h>
53 #include <linux/swap.h>
54 #include <linux/mutex.h>
55 #include <linux/lockd/bind.h>
56 #include <linux/module.h>
57 
58 #define NFSDDBG_FACILITY                NFSDDBG_PROC
59 
60 /* Globals */
61 static time_t lease_time = 90;     /* default lease time */
62 static time_t user_lease_time = 90;
63 static time_t boot_time;
64 static u32 current_ownerid = 1;
65 static u32 current_fileid = 1;
66 static u32 current_delegid = 1;
67 static u32 nfs4_init;
68 static stateid_t zerostateid;             /* bits all 0 */
69 static stateid_t onestateid;              /* bits all 1 */
70 
71 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
72 #define ONE_STATEID(stateid)  (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
73 
74 /* forward declarations */
75 static struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
76 static struct nfs4_delegation * find_delegation_stateid(struct inode *ino, stateid_t *stid);
77 static void release_stateid_lockowners(struct nfs4_stateid *open_stp);
78 static char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery";
79 static void nfs4_set_recdir(char *recdir);
80 
81 /* Locking:
82  *
83  * client_mutex:
84  * 	protects clientid_hashtbl[], clientstr_hashtbl[],
85  * 	unconfstr_hashtbl[], uncofid_hashtbl[].
86  */
87 static DEFINE_MUTEX(client_mutex);
88 
89 static struct kmem_cache *stateowner_slab = NULL;
90 static struct kmem_cache *file_slab = NULL;
91 static struct kmem_cache *stateid_slab = NULL;
92 static struct kmem_cache *deleg_slab = NULL;
93 
94 void
95 nfs4_lock_state(void)
96 {
97 	mutex_lock(&client_mutex);
98 }
99 
100 void
101 nfs4_unlock_state(void)
102 {
103 	mutex_unlock(&client_mutex);
104 }
105 
106 static inline u32
107 opaque_hashval(const void *ptr, int nbytes)
108 {
109 	unsigned char *cptr = (unsigned char *) ptr;
110 
111 	u32 x = 0;
112 	while (nbytes--) {
113 		x *= 37;
114 		x += *cptr++;
115 	}
116 	return x;
117 }
118 
119 /* forward declarations */
120 static void release_stateowner(struct nfs4_stateowner *sop);
121 static void release_stateid(struct nfs4_stateid *stp, int flags);
122 
123 /*
124  * Delegation state
125  */
126 
127 /* recall_lock protects the del_recall_lru */
128 static DEFINE_SPINLOCK(recall_lock);
129 static struct list_head del_recall_lru;
130 
131 static void
132 free_nfs4_file(struct kref *kref)
133 {
134 	struct nfs4_file *fp = container_of(kref, struct nfs4_file, fi_ref);
135 	list_del(&fp->fi_hash);
136 	iput(fp->fi_inode);
137 	kmem_cache_free(file_slab, fp);
138 }
139 
140 static inline void
141 put_nfs4_file(struct nfs4_file *fi)
142 {
143 	kref_put(&fi->fi_ref, free_nfs4_file);
144 }
145 
146 static inline void
147 get_nfs4_file(struct nfs4_file *fi)
148 {
149 	kref_get(&fi->fi_ref);
150 }
151 
152 static int num_delegations;
153 unsigned int max_delegations;
154 
155 /*
156  * Open owner state (share locks)
157  */
158 
159 /* hash tables for nfs4_stateowner */
160 #define OWNER_HASH_BITS              8
161 #define OWNER_HASH_SIZE             (1 << OWNER_HASH_BITS)
162 #define OWNER_HASH_MASK             (OWNER_HASH_SIZE - 1)
163 
164 #define ownerid_hashval(id) \
165         ((id) & OWNER_HASH_MASK)
166 #define ownerstr_hashval(clientid, ownername) \
167         (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
168 
169 static struct list_head	ownerid_hashtbl[OWNER_HASH_SIZE];
170 static struct list_head	ownerstr_hashtbl[OWNER_HASH_SIZE];
171 
172 /* hash table for nfs4_file */
173 #define FILE_HASH_BITS                   8
174 #define FILE_HASH_SIZE                  (1 << FILE_HASH_BITS)
175 #define FILE_HASH_MASK                  (FILE_HASH_SIZE - 1)
176 /* hash table for (open)nfs4_stateid */
177 #define STATEID_HASH_BITS              10
178 #define STATEID_HASH_SIZE              (1 << STATEID_HASH_BITS)
179 #define STATEID_HASH_MASK              (STATEID_HASH_SIZE - 1)
180 
181 #define file_hashval(x) \
182         hash_ptr(x, FILE_HASH_BITS)
183 #define stateid_hashval(owner_id, file_id)  \
184         (((owner_id) + (file_id)) & STATEID_HASH_MASK)
185 
186 static struct list_head file_hashtbl[FILE_HASH_SIZE];
187 static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
188 
189 static struct nfs4_delegation *
190 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
191 {
192 	struct nfs4_delegation *dp;
193 	struct nfs4_file *fp = stp->st_file;
194 	struct nfs4_callback *cb = &stp->st_stateowner->so_client->cl_callback;
195 
196 	dprintk("NFSD alloc_init_deleg\n");
197 	if (fp->fi_had_conflict)
198 		return NULL;
199 	if (num_delegations > max_delegations)
200 		return NULL;
201 	dp = kmem_cache_alloc(deleg_slab, GFP_KERNEL);
202 	if (dp == NULL)
203 		return dp;
204 	num_delegations++;
205 	INIT_LIST_HEAD(&dp->dl_perfile);
206 	INIT_LIST_HEAD(&dp->dl_perclnt);
207 	INIT_LIST_HEAD(&dp->dl_recall_lru);
208 	dp->dl_client = clp;
209 	get_nfs4_file(fp);
210 	dp->dl_file = fp;
211 	dp->dl_flock = NULL;
212 	get_file(stp->st_vfs_file);
213 	dp->dl_vfs_file = stp->st_vfs_file;
214 	dp->dl_type = type;
215 	dp->dl_recall.cbr_dp = NULL;
216 	dp->dl_recall.cbr_ident = cb->cb_ident;
217 	dp->dl_recall.cbr_trunc = 0;
218 	dp->dl_stateid.si_boot = boot_time;
219 	dp->dl_stateid.si_stateownerid = current_delegid++;
220 	dp->dl_stateid.si_fileid = 0;
221 	dp->dl_stateid.si_generation = 0;
222 	dp->dl_fhlen = current_fh->fh_handle.fh_size;
223 	memcpy(dp->dl_fhval, &current_fh->fh_handle.fh_base,
224 		        current_fh->fh_handle.fh_size);
225 	dp->dl_time = 0;
226 	atomic_set(&dp->dl_count, 1);
227 	list_add(&dp->dl_perfile, &fp->fi_delegations);
228 	list_add(&dp->dl_perclnt, &clp->cl_delegations);
229 	return dp;
230 }
231 
232 void
233 nfs4_put_delegation(struct nfs4_delegation *dp)
234 {
235 	if (atomic_dec_and_test(&dp->dl_count)) {
236 		dprintk("NFSD: freeing dp %p\n",dp);
237 		put_nfs4_file(dp->dl_file);
238 		kmem_cache_free(deleg_slab, dp);
239 		num_delegations--;
240 	}
241 }
242 
243 /* Remove the associated file_lock first, then remove the delegation.
244  * lease_modify() is called to remove the FS_LEASE file_lock from
245  * the i_flock list, eventually calling nfsd's lock_manager
246  * fl_release_callback.
247  */
248 static void
249 nfs4_close_delegation(struct nfs4_delegation *dp)
250 {
251 	struct file *filp = dp->dl_vfs_file;
252 
253 	dprintk("NFSD: close_delegation dp %p\n",dp);
254 	dp->dl_vfs_file = NULL;
255 	/* The following nfsd_close may not actually close the file,
256 	 * but we want to remove the lease in any case. */
257 	if (dp->dl_flock)
258 		vfs_setlease(filp, F_UNLCK, &dp->dl_flock);
259 	nfsd_close(filp);
260 }
261 
262 /* Called under the state lock. */
263 static void
264 unhash_delegation(struct nfs4_delegation *dp)
265 {
266 	list_del_init(&dp->dl_perfile);
267 	list_del_init(&dp->dl_perclnt);
268 	spin_lock(&recall_lock);
269 	list_del_init(&dp->dl_recall_lru);
270 	spin_unlock(&recall_lock);
271 	nfs4_close_delegation(dp);
272 	nfs4_put_delegation(dp);
273 }
274 
275 /*
276  * SETCLIENTID state
277  */
278 
279 /* Hash tables for nfs4_clientid state */
280 #define CLIENT_HASH_BITS                 4
281 #define CLIENT_HASH_SIZE                (1 << CLIENT_HASH_BITS)
282 #define CLIENT_HASH_MASK                (CLIENT_HASH_SIZE - 1)
283 
284 #define clientid_hashval(id) \
285 	((id) & CLIENT_HASH_MASK)
286 #define clientstr_hashval(name) \
287 	(opaque_hashval((name), 8) & CLIENT_HASH_MASK)
288 /*
289  * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
290  * used in reboot/reset lease grace period processing
291  *
292  * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
293  * setclientid_confirmed info.
294  *
295  * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
296  * setclientid info.
297  *
298  * client_lru holds client queue ordered by nfs4_client.cl_time
299  * for lease renewal.
300  *
301  * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
302  * for last close replay.
303  */
304 static struct list_head	reclaim_str_hashtbl[CLIENT_HASH_SIZE];
305 static int reclaim_str_hashtbl_size = 0;
306 static struct list_head	conf_id_hashtbl[CLIENT_HASH_SIZE];
307 static struct list_head	conf_str_hashtbl[CLIENT_HASH_SIZE];
308 static struct list_head	unconf_str_hashtbl[CLIENT_HASH_SIZE];
309 static struct list_head	unconf_id_hashtbl[CLIENT_HASH_SIZE];
310 static struct list_head client_lru;
311 static struct list_head close_lru;
312 
313 static inline void
314 renew_client(struct nfs4_client *clp)
315 {
316 	/*
317 	* Move client to the end to the LRU list.
318 	*/
319 	dprintk("renewing client (clientid %08x/%08x)\n",
320 			clp->cl_clientid.cl_boot,
321 			clp->cl_clientid.cl_id);
322 	list_move_tail(&clp->cl_lru, &client_lru);
323 	clp->cl_time = get_seconds();
324 }
325 
326 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
327 static int
328 STALE_CLIENTID(clientid_t *clid)
329 {
330 	if (clid->cl_boot == boot_time)
331 		return 0;
332 	dprintk("NFSD stale clientid (%08x/%08x)\n",
333 			clid->cl_boot, clid->cl_id);
334 	return 1;
335 }
336 
337 /*
338  * XXX Should we use a slab cache ?
339  * This type of memory management is somewhat inefficient, but we use it
340  * anyway since SETCLIENTID is not a common operation.
341  */
342 static struct nfs4_client *alloc_client(struct xdr_netobj name)
343 {
344 	struct nfs4_client *clp;
345 
346 	clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
347 	if (clp == NULL)
348 		return NULL;
349 	clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
350 	if (clp->cl_name.data == NULL) {
351 		kfree(clp);
352 		return NULL;
353 	}
354 	memcpy(clp->cl_name.data, name.data, name.len);
355 	clp->cl_name.len = name.len;
356 	return clp;
357 }
358 
359 static void
360 shutdown_callback_client(struct nfs4_client *clp)
361 {
362 	struct rpc_clnt *clnt = clp->cl_callback.cb_client;
363 
364 	if (clnt) {
365 		/*
366 		 * Callback threads take a reference on the client, so there
367 		 * should be no outstanding callbacks at this point.
368 		 */
369 		clp->cl_callback.cb_client = NULL;
370 		rpc_shutdown_client(clnt);
371 	}
372 }
373 
374 static inline void
375 free_client(struct nfs4_client *clp)
376 {
377 	shutdown_callback_client(clp);
378 	if (clp->cl_cred.cr_group_info)
379 		put_group_info(clp->cl_cred.cr_group_info);
380 	kfree(clp->cl_name.data);
381 	kfree(clp);
382 }
383 
384 void
385 put_nfs4_client(struct nfs4_client *clp)
386 {
387 	if (atomic_dec_and_test(&clp->cl_count))
388 		free_client(clp);
389 }
390 
391 static void
392 expire_client(struct nfs4_client *clp)
393 {
394 	struct nfs4_stateowner *sop;
395 	struct nfs4_delegation *dp;
396 	struct list_head reaplist;
397 
398 	dprintk("NFSD: expire_client cl_count %d\n",
399 	                    atomic_read(&clp->cl_count));
400 
401 	INIT_LIST_HEAD(&reaplist);
402 	spin_lock(&recall_lock);
403 	while (!list_empty(&clp->cl_delegations)) {
404 		dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
405 		dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
406 				dp->dl_flock);
407 		list_del_init(&dp->dl_perclnt);
408 		list_move(&dp->dl_recall_lru, &reaplist);
409 	}
410 	spin_unlock(&recall_lock);
411 	while (!list_empty(&reaplist)) {
412 		dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
413 		list_del_init(&dp->dl_recall_lru);
414 		unhash_delegation(dp);
415 	}
416 	list_del(&clp->cl_idhash);
417 	list_del(&clp->cl_strhash);
418 	list_del(&clp->cl_lru);
419 	while (!list_empty(&clp->cl_openowners)) {
420 		sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
421 		release_stateowner(sop);
422 	}
423 	put_nfs4_client(clp);
424 }
425 
426 static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir)
427 {
428 	struct nfs4_client *clp;
429 
430 	clp = alloc_client(name);
431 	if (clp == NULL)
432 		return NULL;
433 	memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
434 	atomic_set(&clp->cl_count, 1);
435 	atomic_set(&clp->cl_callback.cb_set, 0);
436 	INIT_LIST_HEAD(&clp->cl_idhash);
437 	INIT_LIST_HEAD(&clp->cl_strhash);
438 	INIT_LIST_HEAD(&clp->cl_openowners);
439 	INIT_LIST_HEAD(&clp->cl_delegations);
440 	INIT_LIST_HEAD(&clp->cl_lru);
441 	return clp;
442 }
443 
444 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
445 {
446 	memcpy(target->cl_verifier.data, source->data,
447 			sizeof(target->cl_verifier.data));
448 }
449 
450 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
451 {
452 	target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
453 	target->cl_clientid.cl_id = source->cl_clientid.cl_id;
454 }
455 
456 static void copy_cred(struct svc_cred *target, struct svc_cred *source)
457 {
458 	target->cr_uid = source->cr_uid;
459 	target->cr_gid = source->cr_gid;
460 	target->cr_group_info = source->cr_group_info;
461 	get_group_info(target->cr_group_info);
462 }
463 
464 static int same_name(const char *n1, const char *n2)
465 {
466 	return 0 == memcmp(n1, n2, HEXDIR_LEN);
467 }
468 
469 static int
470 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
471 {
472 	return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
473 }
474 
475 static int
476 same_clid(clientid_t *cl1, clientid_t *cl2)
477 {
478 	return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
479 }
480 
481 /* XXX what about NGROUP */
482 static int
483 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
484 {
485 	return cr1->cr_uid == cr2->cr_uid;
486 }
487 
488 static void gen_clid(struct nfs4_client *clp)
489 {
490 	static u32 current_clientid = 1;
491 
492 	clp->cl_clientid.cl_boot = boot_time;
493 	clp->cl_clientid.cl_id = current_clientid++;
494 }
495 
496 static void gen_confirm(struct nfs4_client *clp)
497 {
498 	static u32 i;
499 	u32 *p;
500 
501 	p = (u32 *)clp->cl_confirm.data;
502 	*p++ = get_seconds();
503 	*p++ = i++;
504 }
505 
506 static int check_name(struct xdr_netobj name)
507 {
508 	if (name.len == 0)
509 		return 0;
510 	if (name.len > NFS4_OPAQUE_LIMIT) {
511 		dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
512 		return 0;
513 	}
514 	return 1;
515 }
516 
517 static void
518 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
519 {
520 	unsigned int idhashval;
521 
522 	list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
523 	idhashval = clientid_hashval(clp->cl_clientid.cl_id);
524 	list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
525 	list_add_tail(&clp->cl_lru, &client_lru);
526 	clp->cl_time = get_seconds();
527 }
528 
529 static void
530 move_to_confirmed(struct nfs4_client *clp)
531 {
532 	unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
533 	unsigned int strhashval;
534 
535 	dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
536 	list_del_init(&clp->cl_strhash);
537 	list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
538 	strhashval = clientstr_hashval(clp->cl_recdir);
539 	list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
540 	renew_client(clp);
541 }
542 
543 static struct nfs4_client *
544 find_confirmed_client(clientid_t *clid)
545 {
546 	struct nfs4_client *clp;
547 	unsigned int idhashval = clientid_hashval(clid->cl_id);
548 
549 	list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
550 		if (same_clid(&clp->cl_clientid, clid))
551 			return clp;
552 	}
553 	return NULL;
554 }
555 
556 static struct nfs4_client *
557 find_unconfirmed_client(clientid_t *clid)
558 {
559 	struct nfs4_client *clp;
560 	unsigned int idhashval = clientid_hashval(clid->cl_id);
561 
562 	list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
563 		if (same_clid(&clp->cl_clientid, clid))
564 			return clp;
565 	}
566 	return NULL;
567 }
568 
569 static struct nfs4_client *
570 find_confirmed_client_by_str(const char *dname, unsigned int hashval)
571 {
572 	struct nfs4_client *clp;
573 
574 	list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
575 		if (same_name(clp->cl_recdir, dname))
576 			return clp;
577 	}
578 	return NULL;
579 }
580 
581 static struct nfs4_client *
582 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval)
583 {
584 	struct nfs4_client *clp;
585 
586 	list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
587 		if (same_name(clp->cl_recdir, dname))
588 			return clp;
589 	}
590 	return NULL;
591 }
592 
593 /* a helper function for parse_callback */
594 static int
595 parse_octet(unsigned int *lenp, char **addrp)
596 {
597 	unsigned int len = *lenp;
598 	char *p = *addrp;
599 	int n = -1;
600 	char c;
601 
602 	for (;;) {
603 		if (!len)
604 			break;
605 		len--;
606 		c = *p++;
607 		if (c == '.')
608 			break;
609 		if ((c < '0') || (c > '9')) {
610 			n = -1;
611 			break;
612 		}
613 		if (n < 0)
614 			n = 0;
615 		n = (n * 10) + (c - '0');
616 		if (n > 255) {
617 			n = -1;
618 			break;
619 		}
620 	}
621 	*lenp = len;
622 	*addrp = p;
623 	return n;
624 }
625 
626 /* parse and set the setclientid ipv4 callback address */
627 static int
628 parse_ipv4(unsigned int addr_len, char *addr_val, unsigned int *cbaddrp, unsigned short *cbportp)
629 {
630 	int temp = 0;
631 	u32 cbaddr = 0;
632 	u16 cbport = 0;
633 	u32 addrlen = addr_len;
634 	char *addr = addr_val;
635 	int i, shift;
636 
637 	/* ipaddress */
638 	shift = 24;
639 	for(i = 4; i > 0  ; i--) {
640 		if ((temp = parse_octet(&addrlen, &addr)) < 0) {
641 			return 0;
642 		}
643 		cbaddr |= (temp << shift);
644 		if (shift > 0)
645 		shift -= 8;
646 	}
647 	*cbaddrp = cbaddr;
648 
649 	/* port */
650 	shift = 8;
651 	for(i = 2; i > 0  ; i--) {
652 		if ((temp = parse_octet(&addrlen, &addr)) < 0) {
653 			return 0;
654 		}
655 		cbport |= (temp << shift);
656 		if (shift > 0)
657 			shift -= 8;
658 	}
659 	*cbportp = cbport;
660 	return 1;
661 }
662 
663 static void
664 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se)
665 {
666 	struct nfs4_callback *cb = &clp->cl_callback;
667 
668 	/* Currently, we only support tcp for the callback channel */
669 	if ((se->se_callback_netid_len != 3) || memcmp((char *)se->se_callback_netid_val, "tcp", 3))
670 		goto out_err;
671 
672 	if ( !(parse_ipv4(se->se_callback_addr_len, se->se_callback_addr_val,
673 	                 &cb->cb_addr, &cb->cb_port)))
674 		goto out_err;
675 	cb->cb_prog = se->se_callback_prog;
676 	cb->cb_ident = se->se_callback_ident;
677 	return;
678 out_err:
679 	dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
680 		"will not receive delegations\n",
681 		clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
682 
683 	return;
684 }
685 
686 __be32
687 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
688 		  struct nfsd4_setclientid *setclid)
689 {
690 	struct sockaddr_in	*sin = svc_addr_in(rqstp);
691 	struct xdr_netobj 	clname = {
692 		.len = setclid->se_namelen,
693 		.data = setclid->se_name,
694 	};
695 	nfs4_verifier		clverifier = setclid->se_verf;
696 	unsigned int 		strhashval;
697 	struct nfs4_client	*conf, *unconf, *new;
698 	__be32 			status;
699 	char                    dname[HEXDIR_LEN];
700 
701 	if (!check_name(clname))
702 		return nfserr_inval;
703 
704 	status = nfs4_make_rec_clidname(dname, &clname);
705 	if (status)
706 		return status;
707 
708 	/*
709 	 * XXX The Duplicate Request Cache (DRC) has been checked (??)
710 	 * We get here on a DRC miss.
711 	 */
712 
713 	strhashval = clientstr_hashval(dname);
714 
715 	nfs4_lock_state();
716 	conf = find_confirmed_client_by_str(dname, strhashval);
717 	if (conf) {
718 		/* RFC 3530 14.2.33 CASE 0: */
719 		status = nfserr_clid_inuse;
720 		if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)
721 				|| conf->cl_addr != sin->sin_addr.s_addr) {
722 			dprintk("NFSD: setclientid: string in use by client"
723 				"at %u.%u.%u.%u\n", NIPQUAD(conf->cl_addr));
724 			goto out;
725 		}
726 	}
727 	/*
728 	 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
729 	 * has a description of SETCLIENTID request processing consisting
730 	 * of 5 bullet points, labeled as CASE0 - CASE4 below.
731 	 */
732 	unconf = find_unconfirmed_client_by_str(dname, strhashval);
733 	status = nfserr_resource;
734 	if (!conf) {
735 		/*
736 		 * RFC 3530 14.2.33 CASE 4:
737 		 * placed first, because it is the normal case
738 		 */
739 		if (unconf)
740 			expire_client(unconf);
741 		new = create_client(clname, dname);
742 		if (new == NULL)
743 			goto out;
744 		gen_clid(new);
745 	} else if (same_verf(&conf->cl_verifier, &clverifier)) {
746 		/*
747 		 * RFC 3530 14.2.33 CASE 1:
748 		 * probable callback update
749 		 */
750 		if (unconf) {
751 			/* Note this is removing unconfirmed {*x***},
752 			 * which is stronger than RFC recommended {vxc**}.
753 			 * This has the advantage that there is at most
754 			 * one {*x***} in either list at any time.
755 			 */
756 			expire_client(unconf);
757 		}
758 		new = create_client(clname, dname);
759 		if (new == NULL)
760 			goto out;
761 		copy_clid(new, conf);
762 	} else if (!unconf) {
763 		/*
764 		 * RFC 3530 14.2.33 CASE 2:
765 		 * probable client reboot; state will be removed if
766 		 * confirmed.
767 		 */
768 		new = create_client(clname, dname);
769 		if (new == NULL)
770 			goto out;
771 		gen_clid(new);
772 	} else {
773 		/*
774 		 * RFC 3530 14.2.33 CASE 3:
775 		 * probable client reboot; state will be removed if
776 		 * confirmed.
777 		 */
778 		expire_client(unconf);
779 		new = create_client(clname, dname);
780 		if (new == NULL)
781 			goto out;
782 		gen_clid(new);
783 	}
784 	copy_verf(new, &clverifier);
785 	new->cl_addr = sin->sin_addr.s_addr;
786 	copy_cred(&new->cl_cred, &rqstp->rq_cred);
787 	gen_confirm(new);
788 	gen_callback(new, setclid);
789 	add_to_unconfirmed(new, strhashval);
790 	setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
791 	setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
792 	memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
793 	status = nfs_ok;
794 out:
795 	nfs4_unlock_state();
796 	return status;
797 }
798 
799 
800 /*
801  * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
802  * a description of SETCLIENTID_CONFIRM request processing consisting of 4
803  * bullets, labeled as CASE1 - CASE4 below.
804  */
805 __be32
806 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
807 			 struct nfsd4_compound_state *cstate,
808 			 struct nfsd4_setclientid_confirm *setclientid_confirm)
809 {
810 	struct sockaddr_in *sin = svc_addr_in(rqstp);
811 	struct nfs4_client *conf, *unconf;
812 	nfs4_verifier confirm = setclientid_confirm->sc_confirm;
813 	clientid_t * clid = &setclientid_confirm->sc_clientid;
814 	__be32 status;
815 
816 	if (STALE_CLIENTID(clid))
817 		return nfserr_stale_clientid;
818 	/*
819 	 * XXX The Duplicate Request Cache (DRC) has been checked (??)
820 	 * We get here on a DRC miss.
821 	 */
822 
823 	nfs4_lock_state();
824 
825 	conf = find_confirmed_client(clid);
826 	unconf = find_unconfirmed_client(clid);
827 
828 	status = nfserr_clid_inuse;
829 	if (conf && conf->cl_addr != sin->sin_addr.s_addr)
830 		goto out;
831 	if (unconf && unconf->cl_addr != sin->sin_addr.s_addr)
832 		goto out;
833 
834 	/*
835 	 * section 14.2.34 of RFC 3530 has a description of
836 	 * SETCLIENTID_CONFIRM request processing consisting
837 	 * of 4 bullet points, labeled as CASE1 - CASE4 below.
838 	 */
839 	if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
840 		/*
841 		 * RFC 3530 14.2.34 CASE 1:
842 		 * callback update
843 		 */
844 		if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
845 			status = nfserr_clid_inuse;
846 		else {
847 			/* XXX: We just turn off callbacks until we can handle
848 			  * change request correctly. */
849 			atomic_set(&conf->cl_callback.cb_set, 0);
850 			gen_confirm(conf);
851 			nfsd4_remove_clid_dir(unconf);
852 			expire_client(unconf);
853 			status = nfs_ok;
854 
855 		}
856 	} else if (conf && !unconf) {
857 		/*
858 		 * RFC 3530 14.2.34 CASE 2:
859 		 * probable retransmitted request; play it safe and
860 		 * do nothing.
861 		 */
862 		if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
863 			status = nfserr_clid_inuse;
864 		else
865 			status = nfs_ok;
866 	} else if (!conf && unconf
867 			&& same_verf(&unconf->cl_confirm, &confirm)) {
868 		/*
869 		 * RFC 3530 14.2.34 CASE 3:
870 		 * Normal case; new or rebooted client:
871 		 */
872 		if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
873 			status = nfserr_clid_inuse;
874 		} else {
875 			unsigned int hash =
876 				clientstr_hashval(unconf->cl_recdir);
877 			conf = find_confirmed_client_by_str(unconf->cl_recdir,
878 									hash);
879 			if (conf) {
880 				nfsd4_remove_clid_dir(conf);
881 				expire_client(conf);
882 			}
883 			move_to_confirmed(unconf);
884 			conf = unconf;
885 			nfsd4_probe_callback(conf);
886 			status = nfs_ok;
887 		}
888 	} else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
889 	    && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
890 				    				&confirm)))) {
891 		/*
892 		 * RFC 3530 14.2.34 CASE 4:
893 		 * Client probably hasn't noticed that we rebooted yet.
894 		 */
895 		status = nfserr_stale_clientid;
896 	} else {
897 		/* check that we have hit one of the cases...*/
898 		status = nfserr_clid_inuse;
899 	}
900 out:
901 	nfs4_unlock_state();
902 	return status;
903 }
904 
905 /* OPEN Share state helper functions */
906 static inline struct nfs4_file *
907 alloc_init_file(struct inode *ino)
908 {
909 	struct nfs4_file *fp;
910 	unsigned int hashval = file_hashval(ino);
911 
912 	fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
913 	if (fp) {
914 		kref_init(&fp->fi_ref);
915 		INIT_LIST_HEAD(&fp->fi_hash);
916 		INIT_LIST_HEAD(&fp->fi_stateids);
917 		INIT_LIST_HEAD(&fp->fi_delegations);
918 		list_add(&fp->fi_hash, &file_hashtbl[hashval]);
919 		fp->fi_inode = igrab(ino);
920 		fp->fi_id = current_fileid++;
921 		fp->fi_had_conflict = false;
922 		return fp;
923 	}
924 	return NULL;
925 }
926 
927 static void
928 nfsd4_free_slab(struct kmem_cache **slab)
929 {
930 	if (*slab == NULL)
931 		return;
932 	kmem_cache_destroy(*slab);
933 	*slab = NULL;
934 }
935 
936 void
937 nfsd4_free_slabs(void)
938 {
939 	nfsd4_free_slab(&stateowner_slab);
940 	nfsd4_free_slab(&file_slab);
941 	nfsd4_free_slab(&stateid_slab);
942 	nfsd4_free_slab(&deleg_slab);
943 }
944 
945 static int
946 nfsd4_init_slabs(void)
947 {
948 	stateowner_slab = kmem_cache_create("nfsd4_stateowners",
949 			sizeof(struct nfs4_stateowner), 0, 0, NULL);
950 	if (stateowner_slab == NULL)
951 		goto out_nomem;
952 	file_slab = kmem_cache_create("nfsd4_files",
953 			sizeof(struct nfs4_file), 0, 0, NULL);
954 	if (file_slab == NULL)
955 		goto out_nomem;
956 	stateid_slab = kmem_cache_create("nfsd4_stateids",
957 			sizeof(struct nfs4_stateid), 0, 0, NULL);
958 	if (stateid_slab == NULL)
959 		goto out_nomem;
960 	deleg_slab = kmem_cache_create("nfsd4_delegations",
961 			sizeof(struct nfs4_delegation), 0, 0, NULL);
962 	if (deleg_slab == NULL)
963 		goto out_nomem;
964 	return 0;
965 out_nomem:
966 	nfsd4_free_slabs();
967 	dprintk("nfsd4: out of memory while initializing nfsv4\n");
968 	return -ENOMEM;
969 }
970 
971 void
972 nfs4_free_stateowner(struct kref *kref)
973 {
974 	struct nfs4_stateowner *sop =
975 		container_of(kref, struct nfs4_stateowner, so_ref);
976 	kfree(sop->so_owner.data);
977 	kmem_cache_free(stateowner_slab, sop);
978 }
979 
980 static inline struct nfs4_stateowner *
981 alloc_stateowner(struct xdr_netobj *owner)
982 {
983 	struct nfs4_stateowner *sop;
984 
985 	if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
986 		if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
987 			memcpy(sop->so_owner.data, owner->data, owner->len);
988 			sop->so_owner.len = owner->len;
989 			kref_init(&sop->so_ref);
990 			return sop;
991 		}
992 		kmem_cache_free(stateowner_slab, sop);
993 	}
994 	return NULL;
995 }
996 
997 static struct nfs4_stateowner *
998 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
999 	struct nfs4_stateowner *sop;
1000 	struct nfs4_replay *rp;
1001 	unsigned int idhashval;
1002 
1003 	if (!(sop = alloc_stateowner(&open->op_owner)))
1004 		return NULL;
1005 	idhashval = ownerid_hashval(current_ownerid);
1006 	INIT_LIST_HEAD(&sop->so_idhash);
1007 	INIT_LIST_HEAD(&sop->so_strhash);
1008 	INIT_LIST_HEAD(&sop->so_perclient);
1009 	INIT_LIST_HEAD(&sop->so_stateids);
1010 	INIT_LIST_HEAD(&sop->so_perstateid);  /* not used */
1011 	INIT_LIST_HEAD(&sop->so_close_lru);
1012 	sop->so_time = 0;
1013 	list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1014 	list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
1015 	list_add(&sop->so_perclient, &clp->cl_openowners);
1016 	sop->so_is_open_owner = 1;
1017 	sop->so_id = current_ownerid++;
1018 	sop->so_client = clp;
1019 	sop->so_seqid = open->op_seqid;
1020 	sop->so_confirmed = 0;
1021 	rp = &sop->so_replay;
1022 	rp->rp_status = nfserr_serverfault;
1023 	rp->rp_buflen = 0;
1024 	rp->rp_buf = rp->rp_ibuf;
1025 	return sop;
1026 }
1027 
1028 static void
1029 release_stateid_lockowners(struct nfs4_stateid *open_stp)
1030 {
1031 	struct nfs4_stateowner *lock_sop;
1032 
1033 	while (!list_empty(&open_stp->st_lockowners)) {
1034 		lock_sop = list_entry(open_stp->st_lockowners.next,
1035 				struct nfs4_stateowner, so_perstateid);
1036 		/* list_del(&open_stp->st_lockowners);  */
1037 		BUG_ON(lock_sop->so_is_open_owner);
1038 		release_stateowner(lock_sop);
1039 	}
1040 }
1041 
1042 static void
1043 unhash_stateowner(struct nfs4_stateowner *sop)
1044 {
1045 	struct nfs4_stateid *stp;
1046 
1047 	list_del(&sop->so_idhash);
1048 	list_del(&sop->so_strhash);
1049 	if (sop->so_is_open_owner)
1050 		list_del(&sop->so_perclient);
1051 	list_del(&sop->so_perstateid);
1052 	while (!list_empty(&sop->so_stateids)) {
1053 		stp = list_entry(sop->so_stateids.next,
1054 			struct nfs4_stateid, st_perstateowner);
1055 		if (sop->so_is_open_owner)
1056 			release_stateid(stp, OPEN_STATE);
1057 		else
1058 			release_stateid(stp, LOCK_STATE);
1059 	}
1060 }
1061 
1062 static void
1063 release_stateowner(struct nfs4_stateowner *sop)
1064 {
1065 	unhash_stateowner(sop);
1066 	list_del(&sop->so_close_lru);
1067 	nfs4_put_stateowner(sop);
1068 }
1069 
1070 static inline void
1071 init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1072 	struct nfs4_stateowner *sop = open->op_stateowner;
1073 	unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1074 
1075 	INIT_LIST_HEAD(&stp->st_hash);
1076 	INIT_LIST_HEAD(&stp->st_perstateowner);
1077 	INIT_LIST_HEAD(&stp->st_lockowners);
1078 	INIT_LIST_HEAD(&stp->st_perfile);
1079 	list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
1080 	list_add(&stp->st_perstateowner, &sop->so_stateids);
1081 	list_add(&stp->st_perfile, &fp->fi_stateids);
1082 	stp->st_stateowner = sop;
1083 	get_nfs4_file(fp);
1084 	stp->st_file = fp;
1085 	stp->st_stateid.si_boot = boot_time;
1086 	stp->st_stateid.si_stateownerid = sop->so_id;
1087 	stp->st_stateid.si_fileid = fp->fi_id;
1088 	stp->st_stateid.si_generation = 0;
1089 	stp->st_access_bmap = 0;
1090 	stp->st_deny_bmap = 0;
1091 	__set_bit(open->op_share_access, &stp->st_access_bmap);
1092 	__set_bit(open->op_share_deny, &stp->st_deny_bmap);
1093 	stp->st_openstp = NULL;
1094 }
1095 
1096 static void
1097 release_stateid(struct nfs4_stateid *stp, int flags)
1098 {
1099 	struct file *filp = stp->st_vfs_file;
1100 
1101 	list_del(&stp->st_hash);
1102 	list_del(&stp->st_perfile);
1103 	list_del(&stp->st_perstateowner);
1104 	if (flags & OPEN_STATE) {
1105 		release_stateid_lockowners(stp);
1106 		stp->st_vfs_file = NULL;
1107 		nfsd_close(filp);
1108 	} else if (flags & LOCK_STATE)
1109 		locks_remove_posix(filp, (fl_owner_t) stp->st_stateowner);
1110 	put_nfs4_file(stp->st_file);
1111 	kmem_cache_free(stateid_slab, stp);
1112 }
1113 
1114 static void
1115 move_to_close_lru(struct nfs4_stateowner *sop)
1116 {
1117 	dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1118 
1119 	list_move_tail(&sop->so_close_lru, &close_lru);
1120 	sop->so_time = get_seconds();
1121 }
1122 
1123 static int
1124 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1125 							clientid_t *clid)
1126 {
1127 	return (sop->so_owner.len == owner->len) &&
1128 		0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1129 		(sop->so_client->cl_clientid.cl_id == clid->cl_id);
1130 }
1131 
1132 static struct nfs4_stateowner *
1133 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1134 {
1135 	struct nfs4_stateowner *so = NULL;
1136 
1137 	list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
1138 		if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1139 			return so;
1140 	}
1141 	return NULL;
1142 }
1143 
1144 /* search file_hashtbl[] for file */
1145 static struct nfs4_file *
1146 find_file(struct inode *ino)
1147 {
1148 	unsigned int hashval = file_hashval(ino);
1149 	struct nfs4_file *fp;
1150 
1151 	list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
1152 		if (fp->fi_inode == ino) {
1153 			get_nfs4_file(fp);
1154 			return fp;
1155 		}
1156 	}
1157 	return NULL;
1158 }
1159 
1160 static inline int access_valid(u32 x)
1161 {
1162 	if (x < NFS4_SHARE_ACCESS_READ)
1163 		return 0;
1164 	if (x > NFS4_SHARE_ACCESS_BOTH)
1165 		return 0;
1166 	return 1;
1167 }
1168 
1169 static inline int deny_valid(u32 x)
1170 {
1171 	/* Note: unlike access bits, deny bits may be zero. */
1172 	return x <= NFS4_SHARE_DENY_BOTH;
1173 }
1174 
1175 /*
1176  * We store the NONE, READ, WRITE, and BOTH bits separately in the
1177  * st_{access,deny}_bmap field of the stateid, in order to track not
1178  * only what share bits are currently in force, but also what
1179  * combinations of share bits previous opens have used.  This allows us
1180  * to enforce the recommendation of rfc 3530 14.2.19 that the server
1181  * return an error if the client attempt to downgrade to a combination
1182  * of share bits not explicable by closing some of its previous opens.
1183  *
1184  * XXX: This enforcement is actually incomplete, since we don't keep
1185  * track of access/deny bit combinations; so, e.g., we allow:
1186  *
1187  *	OPEN allow read, deny write
1188  *	OPEN allow both, deny none
1189  *	DOWNGRADE allow read, deny none
1190  *
1191  * which we should reject.
1192  */
1193 static void
1194 set_access(unsigned int *access, unsigned long bmap) {
1195 	int i;
1196 
1197 	*access = 0;
1198 	for (i = 1; i < 4; i++) {
1199 		if (test_bit(i, &bmap))
1200 			*access |= i;
1201 	}
1202 }
1203 
1204 static void
1205 set_deny(unsigned int *deny, unsigned long bmap) {
1206 	int i;
1207 
1208 	*deny = 0;
1209 	for (i = 0; i < 4; i++) {
1210 		if (test_bit(i, &bmap))
1211 			*deny |= i ;
1212 	}
1213 }
1214 
1215 static int
1216 test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1217 	unsigned int access, deny;
1218 
1219 	set_access(&access, stp->st_access_bmap);
1220 	set_deny(&deny, stp->st_deny_bmap);
1221 	if ((access & open->op_share_deny) || (deny & open->op_share_access))
1222 		return 0;
1223 	return 1;
1224 }
1225 
1226 /*
1227  * Called to check deny when READ with all zero stateid or
1228  * WRITE with all zero or all one stateid
1229  */
1230 static __be32
1231 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
1232 {
1233 	struct inode *ino = current_fh->fh_dentry->d_inode;
1234 	struct nfs4_file *fp;
1235 	struct nfs4_stateid *stp;
1236 	__be32 ret;
1237 
1238 	dprintk("NFSD: nfs4_share_conflict\n");
1239 
1240 	fp = find_file(ino);
1241 	if (!fp)
1242 		return nfs_ok;
1243 	ret = nfserr_locked;
1244 	/* Search for conflicting share reservations */
1245 	list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
1246 		if (test_bit(deny_type, &stp->st_deny_bmap) ||
1247 		    test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
1248 			goto out;
1249 	}
1250 	ret = nfs_ok;
1251 out:
1252 	put_nfs4_file(fp);
1253 	return ret;
1254 }
1255 
1256 static inline void
1257 nfs4_file_downgrade(struct file *filp, unsigned int share_access)
1258 {
1259 	if (share_access & NFS4_SHARE_ACCESS_WRITE) {
1260 		drop_file_write_access(filp);
1261 		filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
1262 	}
1263 }
1264 
1265 /*
1266  * Recall a delegation
1267  */
1268 static int
1269 do_recall(void *__dp)
1270 {
1271 	struct nfs4_delegation *dp = __dp;
1272 
1273 	dp->dl_file->fi_had_conflict = true;
1274 	nfsd4_cb_recall(dp);
1275 	return 0;
1276 }
1277 
1278 /*
1279  * Spawn a thread to perform a recall on the delegation represented
1280  * by the lease (file_lock)
1281  *
1282  * Called from break_lease() with lock_kernel() held.
1283  * Note: we assume break_lease will only call this *once* for any given
1284  * lease.
1285  */
1286 static
1287 void nfsd_break_deleg_cb(struct file_lock *fl)
1288 {
1289 	struct nfs4_delegation *dp=  (struct nfs4_delegation *)fl->fl_owner;
1290 	struct task_struct *t;
1291 
1292 	dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
1293 	if (!dp)
1294 		return;
1295 
1296 	/* We're assuming the state code never drops its reference
1297 	 * without first removing the lease.  Since we're in this lease
1298 	 * callback (and since the lease code is serialized by the kernel
1299 	 * lock) we know the server hasn't removed the lease yet, we know
1300 	 * it's safe to take a reference: */
1301 	atomic_inc(&dp->dl_count);
1302 	atomic_inc(&dp->dl_client->cl_count);
1303 
1304 	spin_lock(&recall_lock);
1305 	list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
1306 	spin_unlock(&recall_lock);
1307 
1308 	/* only place dl_time is set. protected by lock_kernel*/
1309 	dp->dl_time = get_seconds();
1310 
1311 	/*
1312 	 * We don't want the locks code to timeout the lease for us;
1313 	 * we'll remove it ourself if the delegation isn't returned
1314 	 * in time.
1315 	 */
1316 	fl->fl_break_time = 0;
1317 
1318 	t = kthread_run(do_recall, dp, "%s", "nfs4_cb_recall");
1319 	if (IS_ERR(t)) {
1320 		struct nfs4_client *clp = dp->dl_client;
1321 
1322 		printk(KERN_INFO "NFSD: Callback thread failed for "
1323 			"for client (clientid %08x/%08x)\n",
1324 			clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1325 		put_nfs4_client(dp->dl_client);
1326 		nfs4_put_delegation(dp);
1327 	}
1328 }
1329 
1330 /*
1331  * The file_lock is being reapd.
1332  *
1333  * Called by locks_free_lock() with lock_kernel() held.
1334  */
1335 static
1336 void nfsd_release_deleg_cb(struct file_lock *fl)
1337 {
1338 	struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1339 
1340 	dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
1341 
1342 	if (!(fl->fl_flags & FL_LEASE) || !dp)
1343 		return;
1344 	dp->dl_flock = NULL;
1345 }
1346 
1347 /*
1348  * Set the delegation file_lock back pointer.
1349  *
1350  * Called from setlease() with lock_kernel() held.
1351  */
1352 static
1353 void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
1354 {
1355 	struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
1356 
1357 	dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
1358 	if (!dp)
1359 		return;
1360 	dp->dl_flock = new;
1361 }
1362 
1363 /*
1364  * Called from setlease() with lock_kernel() held
1365  */
1366 static
1367 int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
1368 {
1369 	struct nfs4_delegation *onlistd =
1370 		(struct nfs4_delegation *)onlist->fl_owner;
1371 	struct nfs4_delegation *tryd =
1372 		(struct nfs4_delegation *)try->fl_owner;
1373 
1374 	if (onlist->fl_lmops != try->fl_lmops)
1375 		return 0;
1376 
1377 	return onlistd->dl_client == tryd->dl_client;
1378 }
1379 
1380 
1381 static
1382 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
1383 {
1384 	if (arg & F_UNLCK)
1385 		return lease_modify(onlist, arg);
1386 	else
1387 		return -EAGAIN;
1388 }
1389 
1390 static struct lock_manager_operations nfsd_lease_mng_ops = {
1391 	.fl_break = nfsd_break_deleg_cb,
1392 	.fl_release_private = nfsd_release_deleg_cb,
1393 	.fl_copy_lock = nfsd_copy_lock_deleg_cb,
1394 	.fl_mylease = nfsd_same_client_deleg_cb,
1395 	.fl_change = nfsd_change_deleg_cb,
1396 };
1397 
1398 
1399 __be32
1400 nfsd4_process_open1(struct nfsd4_open *open)
1401 {
1402 	clientid_t *clientid = &open->op_clientid;
1403 	struct nfs4_client *clp = NULL;
1404 	unsigned int strhashval;
1405 	struct nfs4_stateowner *sop = NULL;
1406 
1407 	if (!check_name(open->op_owner))
1408 		return nfserr_inval;
1409 
1410 	if (STALE_CLIENTID(&open->op_clientid))
1411 		return nfserr_stale_clientid;
1412 
1413 	strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
1414 	sop = find_openstateowner_str(strhashval, open);
1415 	open->op_stateowner = sop;
1416 	if (!sop) {
1417 		/* Make sure the client's lease hasn't expired. */
1418 		clp = find_confirmed_client(clientid);
1419 		if (clp == NULL)
1420 			return nfserr_expired;
1421 		goto renew;
1422 	}
1423 	if (!sop->so_confirmed) {
1424 		/* Replace unconfirmed owners without checking for replay. */
1425 		clp = sop->so_client;
1426 		release_stateowner(sop);
1427 		open->op_stateowner = NULL;
1428 		goto renew;
1429 	}
1430 	if (open->op_seqid == sop->so_seqid - 1) {
1431 		if (sop->so_replay.rp_buflen)
1432 			return nfserr_replay_me;
1433 		/* The original OPEN failed so spectacularly
1434 		 * that we don't even have replay data saved!
1435 		 * Therefore, we have no choice but to continue
1436 		 * processing this OPEN; presumably, we'll
1437 		 * fail again for the same reason.
1438 		 */
1439 		dprintk("nfsd4_process_open1: replay with no replay cache\n");
1440 		goto renew;
1441 	}
1442 	if (open->op_seqid != sop->so_seqid)
1443 		return nfserr_bad_seqid;
1444 renew:
1445 	if (open->op_stateowner == NULL) {
1446 		sop = alloc_init_open_stateowner(strhashval, clp, open);
1447 		if (sop == NULL)
1448 			return nfserr_resource;
1449 		open->op_stateowner = sop;
1450 	}
1451 	list_del_init(&sop->so_close_lru);
1452 	renew_client(sop->so_client);
1453 	return nfs_ok;
1454 }
1455 
1456 static inline __be32
1457 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
1458 {
1459 	if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
1460 		return nfserr_openmode;
1461 	else
1462 		return nfs_ok;
1463 }
1464 
1465 static struct nfs4_delegation *
1466 find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
1467 {
1468 	struct nfs4_delegation *dp;
1469 
1470 	list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
1471 		if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
1472 			return dp;
1473 	}
1474 	return NULL;
1475 }
1476 
1477 static __be32
1478 nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
1479 		struct nfs4_delegation **dp)
1480 {
1481 	int flags;
1482 	__be32 status = nfserr_bad_stateid;
1483 
1484 	*dp = find_delegation_file(fp, &open->op_delegate_stateid);
1485 	if (*dp == NULL)
1486 		goto out;
1487 	flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
1488 						RD_STATE : WR_STATE;
1489 	status = nfs4_check_delegmode(*dp, flags);
1490 	if (status)
1491 		*dp = NULL;
1492 out:
1493 	if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
1494 		return nfs_ok;
1495 	if (status)
1496 		return status;
1497 	open->op_stateowner->so_confirmed = 1;
1498 	return nfs_ok;
1499 }
1500 
1501 static __be32
1502 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
1503 {
1504 	struct nfs4_stateid *local;
1505 	__be32 status = nfserr_share_denied;
1506 	struct nfs4_stateowner *sop = open->op_stateowner;
1507 
1508 	list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1509 		/* ignore lock owners */
1510 		if (local->st_stateowner->so_is_open_owner == 0)
1511 			continue;
1512 		/* remember if we have seen this open owner */
1513 		if (local->st_stateowner == sop)
1514 			*stpp = local;
1515 		/* check for conflicting share reservations */
1516 		if (!test_share(local, open))
1517 			goto out;
1518 	}
1519 	status = 0;
1520 out:
1521 	return status;
1522 }
1523 
1524 static inline struct nfs4_stateid *
1525 nfs4_alloc_stateid(void)
1526 {
1527 	return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
1528 }
1529 
1530 static __be32
1531 nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
1532 		struct nfs4_delegation *dp,
1533 		struct svc_fh *cur_fh, int flags)
1534 {
1535 	struct nfs4_stateid *stp;
1536 
1537 	stp = nfs4_alloc_stateid();
1538 	if (stp == NULL)
1539 		return nfserr_resource;
1540 
1541 	if (dp) {
1542 		get_file(dp->dl_vfs_file);
1543 		stp->st_vfs_file = dp->dl_vfs_file;
1544 	} else {
1545 		__be32 status;
1546 		status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
1547 				&stp->st_vfs_file);
1548 		if (status) {
1549 			if (status == nfserr_dropit)
1550 				status = nfserr_jukebox;
1551 			kmem_cache_free(stateid_slab, stp);
1552 			return status;
1553 		}
1554 	}
1555 	*stpp = stp;
1556 	return 0;
1557 }
1558 
1559 static inline __be32
1560 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
1561 		struct nfsd4_open *open)
1562 {
1563 	struct iattr iattr = {
1564 		.ia_valid = ATTR_SIZE,
1565 		.ia_size = 0,
1566 	};
1567 	if (!open->op_truncate)
1568 		return 0;
1569 	if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
1570 		return nfserr_inval;
1571 	return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
1572 }
1573 
1574 static __be32
1575 nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
1576 {
1577 	struct file *filp = stp->st_vfs_file;
1578 	struct inode *inode = filp->f_path.dentry->d_inode;
1579 	unsigned int share_access, new_writer;
1580 	__be32 status;
1581 
1582 	set_access(&share_access, stp->st_access_bmap);
1583 	new_writer = (~share_access) & open->op_share_access
1584 			& NFS4_SHARE_ACCESS_WRITE;
1585 
1586 	if (new_writer) {
1587 		int err = get_write_access(inode);
1588 		if (err)
1589 			return nfserrno(err);
1590 		err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
1591 		if (err)
1592 			return nfserrno(err);
1593 		file_take_write(filp);
1594 	}
1595 	status = nfsd4_truncate(rqstp, cur_fh, open);
1596 	if (status) {
1597 		if (new_writer)
1598 			put_write_access(inode);
1599 		return status;
1600 	}
1601 	/* remember the open */
1602 	filp->f_mode |= open->op_share_access;
1603 	__set_bit(open->op_share_access, &stp->st_access_bmap);
1604 	__set_bit(open->op_share_deny, &stp->st_deny_bmap);
1605 
1606 	return nfs_ok;
1607 }
1608 
1609 
1610 static void
1611 nfs4_set_claim_prev(struct nfsd4_open *open)
1612 {
1613 	open->op_stateowner->so_confirmed = 1;
1614 	open->op_stateowner->so_client->cl_firststate = 1;
1615 }
1616 
1617 /*
1618  * Attempt to hand out a delegation.
1619  */
1620 static void
1621 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
1622 {
1623 	struct nfs4_delegation *dp;
1624 	struct nfs4_stateowner *sop = stp->st_stateowner;
1625 	struct nfs4_callback *cb = &sop->so_client->cl_callback;
1626 	struct file_lock fl, *flp = &fl;
1627 	int status, flag = 0;
1628 
1629 	flag = NFS4_OPEN_DELEGATE_NONE;
1630 	open->op_recall = 0;
1631 	switch (open->op_claim_type) {
1632 		case NFS4_OPEN_CLAIM_PREVIOUS:
1633 			if (!atomic_read(&cb->cb_set))
1634 				open->op_recall = 1;
1635 			flag = open->op_delegate_type;
1636 			if (flag == NFS4_OPEN_DELEGATE_NONE)
1637 				goto out;
1638 			break;
1639 		case NFS4_OPEN_CLAIM_NULL:
1640 			/* Let's not give out any delegations till everyone's
1641 			 * had the chance to reclaim theirs.... */
1642 			if (locks_in_grace())
1643 				goto out;
1644 			if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
1645 				goto out;
1646 			if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
1647 				flag = NFS4_OPEN_DELEGATE_WRITE;
1648 			else
1649 				flag = NFS4_OPEN_DELEGATE_READ;
1650 			break;
1651 		default:
1652 			goto out;
1653 	}
1654 
1655 	dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
1656 	if (dp == NULL) {
1657 		flag = NFS4_OPEN_DELEGATE_NONE;
1658 		goto out;
1659 	}
1660 	locks_init_lock(&fl);
1661 	fl.fl_lmops = &nfsd_lease_mng_ops;
1662 	fl.fl_flags = FL_LEASE;
1663 	fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1664 	fl.fl_end = OFFSET_MAX;
1665 	fl.fl_owner =  (fl_owner_t)dp;
1666 	fl.fl_file = stp->st_vfs_file;
1667 	fl.fl_pid = current->tgid;
1668 
1669 	/* vfs_setlease checks to see if delegation should be handed out.
1670 	 * the lock_manager callbacks fl_mylease and fl_change are used
1671 	 */
1672 	if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1673 		dprintk("NFSD: setlease failed [%d], no delegation\n", status);
1674 		unhash_delegation(dp);
1675 		flag = NFS4_OPEN_DELEGATE_NONE;
1676 		goto out;
1677 	}
1678 
1679 	memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
1680 
1681 	dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
1682 	             dp->dl_stateid.si_boot,
1683 	             dp->dl_stateid.si_stateownerid,
1684 	             dp->dl_stateid.si_fileid,
1685 	             dp->dl_stateid.si_generation);
1686 out:
1687 	if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
1688 			&& flag == NFS4_OPEN_DELEGATE_NONE
1689 			&& open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
1690 		dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1691 	open->op_delegate_type = flag;
1692 }
1693 
1694 /*
1695  * called with nfs4_lock_state() held.
1696  */
1697 __be32
1698 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
1699 {
1700 	struct nfs4_file *fp = NULL;
1701 	struct inode *ino = current_fh->fh_dentry->d_inode;
1702 	struct nfs4_stateid *stp = NULL;
1703 	struct nfs4_delegation *dp = NULL;
1704 	__be32 status;
1705 
1706 	status = nfserr_inval;
1707 	if (!access_valid(open->op_share_access)
1708 			|| !deny_valid(open->op_share_deny))
1709 		goto out;
1710 	/*
1711 	 * Lookup file; if found, lookup stateid and check open request,
1712 	 * and check for delegations in the process of being recalled.
1713 	 * If not found, create the nfs4_file struct
1714 	 */
1715 	fp = find_file(ino);
1716 	if (fp) {
1717 		if ((status = nfs4_check_open(fp, open, &stp)))
1718 			goto out;
1719 		status = nfs4_check_deleg(fp, open, &dp);
1720 		if (status)
1721 			goto out;
1722 	} else {
1723 		status = nfserr_bad_stateid;
1724 		if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
1725 			goto out;
1726 		status = nfserr_resource;
1727 		fp = alloc_init_file(ino);
1728 		if (fp == NULL)
1729 			goto out;
1730 	}
1731 
1732 	/*
1733 	 * OPEN the file, or upgrade an existing OPEN.
1734 	 * If truncate fails, the OPEN fails.
1735 	 */
1736 	if (stp) {
1737 		/* Stateid was found, this is an OPEN upgrade */
1738 		status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
1739 		if (status)
1740 			goto out;
1741 		update_stateid(&stp->st_stateid);
1742 	} else {
1743 		/* Stateid was not found, this is a new OPEN */
1744 		int flags = 0;
1745 		if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
1746 			flags |= NFSD_MAY_READ;
1747 		if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
1748 			flags |= NFSD_MAY_WRITE;
1749 		status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
1750 		if (status)
1751 			goto out;
1752 		init_stateid(stp, fp, open);
1753 		status = nfsd4_truncate(rqstp, current_fh, open);
1754 		if (status) {
1755 			release_stateid(stp, OPEN_STATE);
1756 			goto out;
1757 		}
1758 	}
1759 	memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
1760 
1761 	/*
1762 	* Attempt to hand out a delegation. No error return, because the
1763 	* OPEN succeeds even if we fail.
1764 	*/
1765 	nfs4_open_delegation(current_fh, open, stp);
1766 
1767 	status = nfs_ok;
1768 
1769 	dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
1770 	            stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
1771 	            stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
1772 out:
1773 	if (fp)
1774 		put_nfs4_file(fp);
1775 	if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
1776 		nfs4_set_claim_prev(open);
1777 	/*
1778 	* To finish the open response, we just need to set the rflags.
1779 	*/
1780 	open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
1781 	if (!open->op_stateowner->so_confirmed)
1782 		open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
1783 
1784 	return status;
1785 }
1786 
1787 __be32
1788 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1789 	    clientid_t *clid)
1790 {
1791 	struct nfs4_client *clp;
1792 	__be32 status;
1793 
1794 	nfs4_lock_state();
1795 	dprintk("process_renew(%08x/%08x): starting\n",
1796 			clid->cl_boot, clid->cl_id);
1797 	status = nfserr_stale_clientid;
1798 	if (STALE_CLIENTID(clid))
1799 		goto out;
1800 	clp = find_confirmed_client(clid);
1801 	status = nfserr_expired;
1802 	if (clp == NULL) {
1803 		/* We assume the client took too long to RENEW. */
1804 		dprintk("nfsd4_renew: clientid not found!\n");
1805 		goto out;
1806 	}
1807 	renew_client(clp);
1808 	status = nfserr_cb_path_down;
1809 	if (!list_empty(&clp->cl_delegations)
1810 			&& !atomic_read(&clp->cl_callback.cb_set))
1811 		goto out;
1812 	status = nfs_ok;
1813 out:
1814 	nfs4_unlock_state();
1815 	return status;
1816 }
1817 
1818 struct lock_manager nfsd4_manager = {
1819 };
1820 
1821 static void
1822 nfsd4_end_grace(void)
1823 {
1824 	dprintk("NFSD: end of grace period\n");
1825 	nfsd4_recdir_purge_old();
1826 	locks_end_grace(&nfsd4_manager);
1827 }
1828 
1829 static time_t
1830 nfs4_laundromat(void)
1831 {
1832 	struct nfs4_client *clp;
1833 	struct nfs4_stateowner *sop;
1834 	struct nfs4_delegation *dp;
1835 	struct list_head *pos, *next, reaplist;
1836 	time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
1837 	time_t t, clientid_val = NFSD_LEASE_TIME;
1838 	time_t u, test_val = NFSD_LEASE_TIME;
1839 
1840 	nfs4_lock_state();
1841 
1842 	dprintk("NFSD: laundromat service - starting\n");
1843 	if (locks_in_grace())
1844 		nfsd4_end_grace();
1845 	list_for_each_safe(pos, next, &client_lru) {
1846 		clp = list_entry(pos, struct nfs4_client, cl_lru);
1847 		if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
1848 			t = clp->cl_time - cutoff;
1849 			if (clientid_val > t)
1850 				clientid_val = t;
1851 			break;
1852 		}
1853 		dprintk("NFSD: purging unused client (clientid %08x)\n",
1854 			clp->cl_clientid.cl_id);
1855 		nfsd4_remove_clid_dir(clp);
1856 		expire_client(clp);
1857 	}
1858 	INIT_LIST_HEAD(&reaplist);
1859 	spin_lock(&recall_lock);
1860 	list_for_each_safe(pos, next, &del_recall_lru) {
1861 		dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
1862 		if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
1863 			u = dp->dl_time - cutoff;
1864 			if (test_val > u)
1865 				test_val = u;
1866 			break;
1867 		}
1868 		dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
1869 			            dp, dp->dl_flock);
1870 		list_move(&dp->dl_recall_lru, &reaplist);
1871 	}
1872 	spin_unlock(&recall_lock);
1873 	list_for_each_safe(pos, next, &reaplist) {
1874 		dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
1875 		list_del_init(&dp->dl_recall_lru);
1876 		unhash_delegation(dp);
1877 	}
1878 	test_val = NFSD_LEASE_TIME;
1879 	list_for_each_safe(pos, next, &close_lru) {
1880 		sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
1881 		if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
1882 			u = sop->so_time - cutoff;
1883 			if (test_val > u)
1884 				test_val = u;
1885 			break;
1886 		}
1887 		dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
1888 			sop->so_id);
1889 		release_stateowner(sop);
1890 	}
1891 	if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
1892 		clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
1893 	nfs4_unlock_state();
1894 	return clientid_val;
1895 }
1896 
1897 static struct workqueue_struct *laundry_wq;
1898 static void laundromat_main(struct work_struct *);
1899 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
1900 
1901 static void
1902 laundromat_main(struct work_struct *not_used)
1903 {
1904 	time_t t;
1905 
1906 	t = nfs4_laundromat();
1907 	dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
1908 	queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1909 }
1910 
1911 static struct nfs4_stateowner *
1912 search_close_lru(u32 st_id, int flags)
1913 {
1914 	struct nfs4_stateowner *local = NULL;
1915 
1916 	if (flags & CLOSE_STATE) {
1917 		list_for_each_entry(local, &close_lru, so_close_lru) {
1918 			if (local->so_id == st_id)
1919 				return local;
1920 		}
1921 	}
1922 	return NULL;
1923 }
1924 
1925 static inline int
1926 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
1927 {
1928 	return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1929 }
1930 
1931 static int
1932 STALE_STATEID(stateid_t *stateid)
1933 {
1934 	if (stateid->si_boot == boot_time)
1935 		return 0;
1936 	dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
1937 		stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
1938 		stateid->si_generation);
1939 	return 1;
1940 }
1941 
1942 static inline int
1943 access_permit_read(unsigned long access_bmap)
1944 {
1945 	return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
1946 		test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
1947 		test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
1948 }
1949 
1950 static inline int
1951 access_permit_write(unsigned long access_bmap)
1952 {
1953 	return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
1954 		test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
1955 }
1956 
1957 static
1958 __be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1959 {
1960         __be32 status = nfserr_openmode;
1961 
1962 	if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
1963                 goto out;
1964 	if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
1965                 goto out;
1966 	status = nfs_ok;
1967 out:
1968 	return status;
1969 }
1970 
1971 static inline __be32
1972 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
1973 {
1974 	/* Trying to call delegreturn with a special stateid? Yuch: */
1975 	if (!(flags & (RD_STATE | WR_STATE)))
1976 		return nfserr_bad_stateid;
1977 	else if (ONE_STATEID(stateid) && (flags & RD_STATE))
1978 		return nfs_ok;
1979 	else if (locks_in_grace()) {
1980 		/* Answer in remaining cases depends on existance of
1981 		 * conflicting state; so we must wait out the grace period. */
1982 		return nfserr_grace;
1983 	} else if (flags & WR_STATE)
1984 		return nfs4_share_conflict(current_fh,
1985 				NFS4_SHARE_DENY_WRITE);
1986 	else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
1987 		return nfs4_share_conflict(current_fh,
1988 				NFS4_SHARE_DENY_READ);
1989 }
1990 
1991 /*
1992  * Allow READ/WRITE during grace period on recovered state only for files
1993  * that are not able to provide mandatory locking.
1994  */
1995 static inline int
1996 io_during_grace_disallowed(struct inode *inode, int flags)
1997 {
1998 	return locks_in_grace() && (flags & (RD_STATE | WR_STATE))
1999 		&& mandatory_lock(inode);
2000 }
2001 
2002 static int check_stateid_generation(stateid_t *in, stateid_t *ref)
2003 {
2004 	/* If the client sends us a stateid from the future, it's buggy: */
2005 	if (in->si_generation > ref->si_generation)
2006 		return nfserr_bad_stateid;
2007 	/*
2008 	 * The following, however, can happen.  For example, if the
2009 	 * client sends an open and some IO at the same time, the open
2010 	 * may bump si_generation while the IO is still in flight.
2011 	 * Thanks to hard links and renames, the client never knows what
2012 	 * file an open will affect.  So it could avoid that situation
2013 	 * only by serializing all opens and IO from the same open
2014 	 * owner.  To recover from the old_stateid error, the client
2015 	 * will just have to retry the IO:
2016 	 */
2017 	if (in->si_generation < ref->si_generation)
2018 		return nfserr_old_stateid;
2019 	return nfs_ok;
2020 }
2021 
2022 /*
2023 * Checks for stateid operations
2024 */
2025 __be32
2026 nfs4_preprocess_stateid_op(struct svc_fh *current_fh, stateid_t *stateid, int flags, struct file **filpp)
2027 {
2028 	struct nfs4_stateid *stp = NULL;
2029 	struct nfs4_delegation *dp = NULL;
2030 	stateid_t *stidp;
2031 	struct inode *ino = current_fh->fh_dentry->d_inode;
2032 	__be32 status;
2033 
2034 	dprintk("NFSD: preprocess_stateid_op: stateid = (%08x/%08x/%08x/%08x)\n",
2035 		stateid->si_boot, stateid->si_stateownerid,
2036 		stateid->si_fileid, stateid->si_generation);
2037 	if (filpp)
2038 		*filpp = NULL;
2039 
2040 	if (io_during_grace_disallowed(ino, flags))
2041 		return nfserr_grace;
2042 
2043 	if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2044 		return check_special_stateids(current_fh, stateid, flags);
2045 
2046 	/* STALE STATEID */
2047 	status = nfserr_stale_stateid;
2048 	if (STALE_STATEID(stateid))
2049 		goto out;
2050 
2051 	/* BAD STATEID */
2052 	status = nfserr_bad_stateid;
2053 	if (!stateid->si_fileid) { /* delegation stateid */
2054 		if(!(dp = find_delegation_stateid(ino, stateid))) {
2055 			dprintk("NFSD: delegation stateid not found\n");
2056 			goto out;
2057 		}
2058 		stidp = &dp->dl_stateid;
2059 	} else { /* open or lock stateid */
2060 		if (!(stp = find_stateid(stateid, flags))) {
2061 			dprintk("NFSD: open or lock stateid not found\n");
2062 			goto out;
2063 		}
2064 		if ((flags & CHECK_FH) && nfs4_check_fh(current_fh, stp))
2065 			goto out;
2066 		if (!stp->st_stateowner->so_confirmed)
2067 			goto out;
2068 		stidp = &stp->st_stateid;
2069 	}
2070 	status = check_stateid_generation(stateid, stidp);
2071 	if (status)
2072 		goto out;
2073 	if (stp) {
2074 		if ((status = nfs4_check_openmode(stp,flags)))
2075 			goto out;
2076 		renew_client(stp->st_stateowner->so_client);
2077 		if (filpp)
2078 			*filpp = stp->st_vfs_file;
2079 	} else {
2080 		if ((status = nfs4_check_delegmode(dp, flags)))
2081 			goto out;
2082 		renew_client(dp->dl_client);
2083 		if (flags & DELEG_RET)
2084 			unhash_delegation(dp);
2085 		if (filpp)
2086 			*filpp = dp->dl_vfs_file;
2087 	}
2088 	status = nfs_ok;
2089 out:
2090 	return status;
2091 }
2092 
2093 static inline int
2094 setlkflg (int type)
2095 {
2096 	return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2097 		RD_STATE : WR_STATE;
2098 }
2099 
2100 /*
2101  * Checks for sequence id mutating operations.
2102  */
2103 static __be32
2104 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)
2105 {
2106 	struct nfs4_stateid *stp;
2107 	struct nfs4_stateowner *sop;
2108 	__be32 status;
2109 
2110 	dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2111 			"stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2112 		stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2113 		stateid->si_generation);
2114 
2115 	*stpp = NULL;
2116 	*sopp = NULL;
2117 
2118 	if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2119 		dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
2120 		return nfserr_bad_stateid;
2121 	}
2122 
2123 	if (STALE_STATEID(stateid))
2124 		return nfserr_stale_stateid;
2125 	/*
2126 	* We return BAD_STATEID if filehandle doesn't match stateid,
2127 	* the confirmed flag is incorrecly set, or the generation
2128 	* number is incorrect.
2129 	*/
2130 	stp = find_stateid(stateid, flags);
2131 	if (stp == NULL) {
2132 		/*
2133 		 * Also, we should make sure this isn't just the result of
2134 		 * a replayed close:
2135 		 */
2136 		sop = search_close_lru(stateid->si_stateownerid, flags);
2137 		if (sop == NULL)
2138 			return nfserr_bad_stateid;
2139 		*sopp = sop;
2140 		goto check_replay;
2141 	}
2142 
2143 	*stpp = stp;
2144 	*sopp = sop = stp->st_stateowner;
2145 
2146 	if (lock) {
2147 		clientid_t *lockclid = &lock->v.new.clientid;
2148 		struct nfs4_client *clp = sop->so_client;
2149 		int lkflg = 0;
2150 		__be32 status;
2151 
2152 		lkflg = setlkflg(lock->lk_type);
2153 
2154 		if (lock->lk_is_new) {
2155 			if (!sop->so_is_open_owner)
2156 				return nfserr_bad_stateid;
2157 			if (!same_clid(&clp->cl_clientid, lockclid))
2158 			       return nfserr_bad_stateid;
2159 			/* stp is the open stateid */
2160 			status = nfs4_check_openmode(stp, lkflg);
2161 			if (status)
2162 				return status;
2163 		} else {
2164 			/* stp is the lock stateid */
2165 			status = nfs4_check_openmode(stp->st_openstp, lkflg);
2166 			if (status)
2167 				return status;
2168                }
2169 	}
2170 
2171 	if (nfs4_check_fh(current_fh, stp)) {
2172 		dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
2173 		return nfserr_bad_stateid;
2174 	}
2175 
2176 	/*
2177 	*  We now validate the seqid and stateid generation numbers.
2178 	*  For the moment, we ignore the possibility of
2179 	*  generation number wraparound.
2180 	*/
2181 	if (seqid != sop->so_seqid)
2182 		goto check_replay;
2183 
2184 	if (sop->so_confirmed && flags & CONFIRM) {
2185 		dprintk("NFSD: preprocess_seqid_op: expected"
2186 				" unconfirmed stateowner!\n");
2187 		return nfserr_bad_stateid;
2188 	}
2189 	if (!sop->so_confirmed && !(flags & CONFIRM)) {
2190 		dprintk("NFSD: preprocess_seqid_op: stateowner not"
2191 				" confirmed yet!\n");
2192 		return nfserr_bad_stateid;
2193 	}
2194 	status = check_stateid_generation(stateid, &stp->st_stateid);
2195 	if (status)
2196 		return status;
2197 	renew_client(sop->so_client);
2198 	return nfs_ok;
2199 
2200 check_replay:
2201 	if (seqid == sop->so_seqid - 1) {
2202 		dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
2203 		/* indicate replay to calling function */
2204 		return nfserr_replay_me;
2205 	}
2206 	dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
2207 			sop->so_seqid, seqid);
2208 	*sopp = NULL;
2209 	return nfserr_bad_seqid;
2210 }
2211 
2212 __be32
2213 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2214 		   struct nfsd4_open_confirm *oc)
2215 {
2216 	__be32 status;
2217 	struct nfs4_stateowner *sop;
2218 	struct nfs4_stateid *stp;
2219 
2220 	dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
2221 			(int)cstate->current_fh.fh_dentry->d_name.len,
2222 			cstate->current_fh.fh_dentry->d_name.name);
2223 
2224 	status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
2225 	if (status)
2226 		return status;
2227 
2228 	nfs4_lock_state();
2229 
2230 	if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2231 					oc->oc_seqid, &oc->oc_req_stateid,
2232 					CONFIRM | OPEN_STATE,
2233 					&oc->oc_stateowner, &stp, NULL)))
2234 		goto out;
2235 
2236 	sop = oc->oc_stateowner;
2237 	sop->so_confirmed = 1;
2238 	update_stateid(&stp->st_stateid);
2239 	memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
2240 	dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
2241 		"stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
2242 		         stp->st_stateid.si_boot,
2243 		         stp->st_stateid.si_stateownerid,
2244 		         stp->st_stateid.si_fileid,
2245 		         stp->st_stateid.si_generation);
2246 
2247 	nfsd4_create_clid_dir(sop->so_client);
2248 out:
2249 	if (oc->oc_stateowner) {
2250 		nfs4_get_stateowner(oc->oc_stateowner);
2251 		cstate->replay_owner = oc->oc_stateowner;
2252 	}
2253 	nfs4_unlock_state();
2254 	return status;
2255 }
2256 
2257 
2258 /*
2259  * unset all bits in union bitmap (bmap) that
2260  * do not exist in share (from successful OPEN_DOWNGRADE)
2261  */
2262 static void
2263 reset_union_bmap_access(unsigned long access, unsigned long *bmap)
2264 {
2265 	int i;
2266 	for (i = 1; i < 4; i++) {
2267 		if ((i & access) != i)
2268 			__clear_bit(i, bmap);
2269 	}
2270 }
2271 
2272 static void
2273 reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
2274 {
2275 	int i;
2276 	for (i = 0; i < 4; i++) {
2277 		if ((i & deny) != i)
2278 			__clear_bit(i, bmap);
2279 	}
2280 }
2281 
2282 __be32
2283 nfsd4_open_downgrade(struct svc_rqst *rqstp,
2284 		     struct nfsd4_compound_state *cstate,
2285 		     struct nfsd4_open_downgrade *od)
2286 {
2287 	__be32 status;
2288 	struct nfs4_stateid *stp;
2289 	unsigned int share_access;
2290 
2291 	dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
2292 			(int)cstate->current_fh.fh_dentry->d_name.len,
2293 			cstate->current_fh.fh_dentry->d_name.name);
2294 
2295 	if (!access_valid(od->od_share_access)
2296 			|| !deny_valid(od->od_share_deny))
2297 		return nfserr_inval;
2298 
2299 	nfs4_lock_state();
2300 	if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2301 					od->od_seqid,
2302 					&od->od_stateid,
2303 					OPEN_STATE,
2304 					&od->od_stateowner, &stp, NULL)))
2305 		goto out;
2306 
2307 	status = nfserr_inval;
2308 	if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
2309 		dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
2310 			stp->st_access_bmap, od->od_share_access);
2311 		goto out;
2312 	}
2313 	if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
2314 		dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
2315 			stp->st_deny_bmap, od->od_share_deny);
2316 		goto out;
2317 	}
2318 	set_access(&share_access, stp->st_access_bmap);
2319 	nfs4_file_downgrade(stp->st_vfs_file,
2320 	                    share_access & ~od->od_share_access);
2321 
2322 	reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
2323 	reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
2324 
2325 	update_stateid(&stp->st_stateid);
2326 	memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
2327 	status = nfs_ok;
2328 out:
2329 	if (od->od_stateowner) {
2330 		nfs4_get_stateowner(od->od_stateowner);
2331 		cstate->replay_owner = od->od_stateowner;
2332 	}
2333 	nfs4_unlock_state();
2334 	return status;
2335 }
2336 
2337 /*
2338  * nfs4_unlock_state() called after encode
2339  */
2340 __be32
2341 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2342 	    struct nfsd4_close *close)
2343 {
2344 	__be32 status;
2345 	struct nfs4_stateid *stp;
2346 
2347 	dprintk("NFSD: nfsd4_close on file %.*s\n",
2348 			(int)cstate->current_fh.fh_dentry->d_name.len,
2349 			cstate->current_fh.fh_dentry->d_name.name);
2350 
2351 	nfs4_lock_state();
2352 	/* check close_lru for replay */
2353 	if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2354 					close->cl_seqid,
2355 					&close->cl_stateid,
2356 					OPEN_STATE | CLOSE_STATE,
2357 					&close->cl_stateowner, &stp, NULL)))
2358 		goto out;
2359 	status = nfs_ok;
2360 	update_stateid(&stp->st_stateid);
2361 	memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
2362 
2363 	/* release_stateid() calls nfsd_close() if needed */
2364 	release_stateid(stp, OPEN_STATE);
2365 
2366 	/* place unused nfs4_stateowners on so_close_lru list to be
2367 	 * released by the laundromat service after the lease period
2368 	 * to enable us to handle CLOSE replay
2369 	 */
2370 	if (list_empty(&close->cl_stateowner->so_stateids))
2371 		move_to_close_lru(close->cl_stateowner);
2372 out:
2373 	if (close->cl_stateowner) {
2374 		nfs4_get_stateowner(close->cl_stateowner);
2375 		cstate->replay_owner = close->cl_stateowner;
2376 	}
2377 	nfs4_unlock_state();
2378 	return status;
2379 }
2380 
2381 __be32
2382 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2383 		  struct nfsd4_delegreturn *dr)
2384 {
2385 	__be32 status;
2386 
2387 	if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
2388 		goto out;
2389 
2390 	nfs4_lock_state();
2391 	status = nfs4_preprocess_stateid_op(&cstate->current_fh,
2392 					    &dr->dr_stateid, DELEG_RET, NULL);
2393 	nfs4_unlock_state();
2394 out:
2395 	return status;
2396 }
2397 
2398 
2399 /*
2400  * Lock owner state (byte-range locks)
2401  */
2402 #define LOFF_OVERFLOW(start, len)      ((u64)(len) > ~(u64)(start))
2403 #define LOCK_HASH_BITS              8
2404 #define LOCK_HASH_SIZE             (1 << LOCK_HASH_BITS)
2405 #define LOCK_HASH_MASK             (LOCK_HASH_SIZE - 1)
2406 
2407 #define lockownerid_hashval(id) \
2408         ((id) & LOCK_HASH_MASK)
2409 
2410 static inline unsigned int
2411 lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
2412 		struct xdr_netobj *ownername)
2413 {
2414 	return (file_hashval(inode) + cl_id
2415 			+ opaque_hashval(ownername->data, ownername->len))
2416 		& LOCK_HASH_MASK;
2417 }
2418 
2419 static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
2420 static struct list_head	lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
2421 static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
2422 
2423 static struct nfs4_stateid *
2424 find_stateid(stateid_t *stid, int flags)
2425 {
2426 	struct nfs4_stateid *local = NULL;
2427 	u32 st_id = stid->si_stateownerid;
2428 	u32 f_id = stid->si_fileid;
2429 	unsigned int hashval;
2430 
2431 	dprintk("NFSD: find_stateid flags 0x%x\n",flags);
2432 	if ((flags & LOCK_STATE) || (flags & RD_STATE) || (flags & WR_STATE)) {
2433 		hashval = stateid_hashval(st_id, f_id);
2434 		list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
2435 			if ((local->st_stateid.si_stateownerid == st_id) &&
2436 			    (local->st_stateid.si_fileid == f_id))
2437 				return local;
2438 		}
2439 	}
2440 	if ((flags & OPEN_STATE) || (flags & RD_STATE) || (flags & WR_STATE)) {
2441 		hashval = stateid_hashval(st_id, f_id);
2442 		list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
2443 			if ((local->st_stateid.si_stateownerid == st_id) &&
2444 			    (local->st_stateid.si_fileid == f_id))
2445 				return local;
2446 		}
2447 	}
2448 	return NULL;
2449 }
2450 
2451 static struct nfs4_delegation *
2452 find_delegation_stateid(struct inode *ino, stateid_t *stid)
2453 {
2454 	struct nfs4_file *fp;
2455 	struct nfs4_delegation *dl;
2456 
2457 	dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
2458                     stid->si_boot, stid->si_stateownerid,
2459                     stid->si_fileid, stid->si_generation);
2460 
2461 	fp = find_file(ino);
2462 	if (!fp)
2463 		return NULL;
2464 	dl = find_delegation_file(fp, stid);
2465 	put_nfs4_file(fp);
2466 	return dl;
2467 }
2468 
2469 /*
2470  * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
2471  * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
2472  * byte, because of sign extension problems.  Since NFSv4 calls for 64-bit
2473  * locking, this prevents us from being completely protocol-compliant.  The
2474  * real solution to this problem is to start using unsigned file offsets in
2475  * the VFS, but this is a very deep change!
2476  */
2477 static inline void
2478 nfs4_transform_lock_offset(struct file_lock *lock)
2479 {
2480 	if (lock->fl_start < 0)
2481 		lock->fl_start = OFFSET_MAX;
2482 	if (lock->fl_end < 0)
2483 		lock->fl_end = OFFSET_MAX;
2484 }
2485 
2486 /* Hack!: For now, we're defining this just so we can use a pointer to it
2487  * as a unique cookie to identify our (NFSv4's) posix locks. */
2488 static struct lock_manager_operations nfsd_posix_mng_ops  = {
2489 };
2490 
2491 static inline void
2492 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
2493 {
2494 	struct nfs4_stateowner *sop;
2495 	unsigned int hval;
2496 
2497 	if (fl->fl_lmops == &nfsd_posix_mng_ops) {
2498 		sop = (struct nfs4_stateowner *) fl->fl_owner;
2499 		hval = lockownerid_hashval(sop->so_id);
2500 		kref_get(&sop->so_ref);
2501 		deny->ld_sop = sop;
2502 		deny->ld_clientid = sop->so_client->cl_clientid;
2503 	} else {
2504 		deny->ld_sop = NULL;
2505 		deny->ld_clientid.cl_boot = 0;
2506 		deny->ld_clientid.cl_id = 0;
2507 	}
2508 	deny->ld_start = fl->fl_start;
2509 	deny->ld_length = ~(u64)0;
2510 	if (fl->fl_end != ~(u64)0)
2511 		deny->ld_length = fl->fl_end - fl->fl_start + 1;
2512 	deny->ld_type = NFS4_READ_LT;
2513 	if (fl->fl_type != F_RDLCK)
2514 		deny->ld_type = NFS4_WRITE_LT;
2515 }
2516 
2517 static struct nfs4_stateowner *
2518 find_lockstateowner_str(struct inode *inode, clientid_t *clid,
2519 		struct xdr_netobj *owner)
2520 {
2521 	unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
2522 	struct nfs4_stateowner *op;
2523 
2524 	list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
2525 		if (same_owner_str(op, owner, clid))
2526 			return op;
2527 	}
2528 	return NULL;
2529 }
2530 
2531 /*
2532  * Alloc a lock owner structure.
2533  * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
2534  * occured.
2535  *
2536  * strhashval = lock_ownerstr_hashval
2537  */
2538 
2539 static struct nfs4_stateowner *
2540 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
2541 	struct nfs4_stateowner *sop;
2542 	struct nfs4_replay *rp;
2543 	unsigned int idhashval;
2544 
2545 	if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
2546 		return NULL;
2547 	idhashval = lockownerid_hashval(current_ownerid);
2548 	INIT_LIST_HEAD(&sop->so_idhash);
2549 	INIT_LIST_HEAD(&sop->so_strhash);
2550 	INIT_LIST_HEAD(&sop->so_perclient);
2551 	INIT_LIST_HEAD(&sop->so_stateids);
2552 	INIT_LIST_HEAD(&sop->so_perstateid);
2553 	INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
2554 	sop->so_time = 0;
2555 	list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
2556 	list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
2557 	list_add(&sop->so_perstateid, &open_stp->st_lockowners);
2558 	sop->so_is_open_owner = 0;
2559 	sop->so_id = current_ownerid++;
2560 	sop->so_client = clp;
2561 	/* It is the openowner seqid that will be incremented in encode in the
2562 	 * case of new lockowners; so increment the lock seqid manually: */
2563 	sop->so_seqid = lock->lk_new_lock_seqid + 1;
2564 	sop->so_confirmed = 1;
2565 	rp = &sop->so_replay;
2566 	rp->rp_status = nfserr_serverfault;
2567 	rp->rp_buflen = 0;
2568 	rp->rp_buf = rp->rp_ibuf;
2569 	return sop;
2570 }
2571 
2572 static struct nfs4_stateid *
2573 alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
2574 {
2575 	struct nfs4_stateid *stp;
2576 	unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
2577 
2578 	stp = nfs4_alloc_stateid();
2579 	if (stp == NULL)
2580 		goto out;
2581 	INIT_LIST_HEAD(&stp->st_hash);
2582 	INIT_LIST_HEAD(&stp->st_perfile);
2583 	INIT_LIST_HEAD(&stp->st_perstateowner);
2584 	INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
2585 	list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
2586 	list_add(&stp->st_perfile, &fp->fi_stateids);
2587 	list_add(&stp->st_perstateowner, &sop->so_stateids);
2588 	stp->st_stateowner = sop;
2589 	get_nfs4_file(fp);
2590 	stp->st_file = fp;
2591 	stp->st_stateid.si_boot = boot_time;
2592 	stp->st_stateid.si_stateownerid = sop->so_id;
2593 	stp->st_stateid.si_fileid = fp->fi_id;
2594 	stp->st_stateid.si_generation = 0;
2595 	stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
2596 	stp->st_access_bmap = open_stp->st_access_bmap;
2597 	stp->st_deny_bmap = open_stp->st_deny_bmap;
2598 	stp->st_openstp = open_stp;
2599 
2600 out:
2601 	return stp;
2602 }
2603 
2604 static int
2605 check_lock_length(u64 offset, u64 length)
2606 {
2607 	return ((length == 0)  || ((length != ~(u64)0) &&
2608 	     LOFF_OVERFLOW(offset, length)));
2609 }
2610 
2611 /*
2612  *  LOCK operation
2613  */
2614 __be32
2615 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2616 	   struct nfsd4_lock *lock)
2617 {
2618 	struct nfs4_stateowner *open_sop = NULL;
2619 	struct nfs4_stateowner *lock_sop = NULL;
2620 	struct nfs4_stateid *lock_stp;
2621 	struct file *filp;
2622 	struct file_lock file_lock;
2623 	struct file_lock conflock;
2624 	__be32 status = 0;
2625 	unsigned int strhashval;
2626 	unsigned int cmd;
2627 	int err;
2628 
2629 	dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
2630 		(long long) lock->lk_offset,
2631 		(long long) lock->lk_length);
2632 
2633 	if (check_lock_length(lock->lk_offset, lock->lk_length))
2634 		 return nfserr_inval;
2635 
2636 	if ((status = fh_verify(rqstp, &cstate->current_fh,
2637 				S_IFREG, NFSD_MAY_LOCK))) {
2638 		dprintk("NFSD: nfsd4_lock: permission denied!\n");
2639 		return status;
2640 	}
2641 
2642 	nfs4_lock_state();
2643 
2644 	if (lock->lk_is_new) {
2645 		/*
2646 		 * Client indicates that this is a new lockowner.
2647 		 * Use open owner and open stateid to create lock owner and
2648 		 * lock stateid.
2649 		 */
2650 		struct nfs4_stateid *open_stp = NULL;
2651 		struct nfs4_file *fp;
2652 
2653 		status = nfserr_stale_clientid;
2654 		if (STALE_CLIENTID(&lock->lk_new_clientid))
2655 			goto out;
2656 
2657 		/* validate and update open stateid and open seqid */
2658 		status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2659 				        lock->lk_new_open_seqid,
2660 		                        &lock->lk_new_open_stateid,
2661 					OPEN_STATE,
2662 		                        &lock->lk_replay_owner, &open_stp,
2663 					lock);
2664 		if (status)
2665 			goto out;
2666 		open_sop = lock->lk_replay_owner;
2667 		/* create lockowner and lock stateid */
2668 		fp = open_stp->st_file;
2669 		strhashval = lock_ownerstr_hashval(fp->fi_inode,
2670 				open_sop->so_client->cl_clientid.cl_id,
2671 				&lock->v.new.owner);
2672 		/* XXX: Do we need to check for duplicate stateowners on
2673 		 * the same file, or should they just be allowed (and
2674 		 * create new stateids)? */
2675 		status = nfserr_resource;
2676 		lock_sop = alloc_init_lock_stateowner(strhashval,
2677 				open_sop->so_client, open_stp, lock);
2678 		if (lock_sop == NULL)
2679 			goto out;
2680 		lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
2681 		if (lock_stp == NULL)
2682 			goto out;
2683 	} else {
2684 		/* lock (lock owner + lock stateid) already exists */
2685 		status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2686 				       lock->lk_old_lock_seqid,
2687 				       &lock->lk_old_lock_stateid,
2688 				       LOCK_STATE,
2689 				       &lock->lk_replay_owner, &lock_stp, lock);
2690 		if (status)
2691 			goto out;
2692 		lock_sop = lock->lk_replay_owner;
2693 	}
2694 	/* lock->lk_replay_owner and lock_stp have been created or found */
2695 	filp = lock_stp->st_vfs_file;
2696 
2697 	status = nfserr_grace;
2698 	if (locks_in_grace() && !lock->lk_reclaim)
2699 		goto out;
2700 	status = nfserr_no_grace;
2701 	if (!locks_in_grace() && lock->lk_reclaim)
2702 		goto out;
2703 
2704 	locks_init_lock(&file_lock);
2705 	switch (lock->lk_type) {
2706 		case NFS4_READ_LT:
2707 		case NFS4_READW_LT:
2708 			file_lock.fl_type = F_RDLCK;
2709 			cmd = F_SETLK;
2710 		break;
2711 		case NFS4_WRITE_LT:
2712 		case NFS4_WRITEW_LT:
2713 			file_lock.fl_type = F_WRLCK;
2714 			cmd = F_SETLK;
2715 		break;
2716 		default:
2717 			status = nfserr_inval;
2718 		goto out;
2719 	}
2720 	file_lock.fl_owner = (fl_owner_t)lock_sop;
2721 	file_lock.fl_pid = current->tgid;
2722 	file_lock.fl_file = filp;
2723 	file_lock.fl_flags = FL_POSIX;
2724 	file_lock.fl_lmops = &nfsd_posix_mng_ops;
2725 
2726 	file_lock.fl_start = lock->lk_offset;
2727 	if ((lock->lk_length == ~(u64)0) ||
2728 			LOFF_OVERFLOW(lock->lk_offset, lock->lk_length))
2729 		file_lock.fl_end = ~(u64)0;
2730 	else
2731 		file_lock.fl_end = lock->lk_offset + lock->lk_length - 1;
2732 	nfs4_transform_lock_offset(&file_lock);
2733 
2734 	/*
2735 	* Try to lock the file in the VFS.
2736 	* Note: locks.c uses the BKL to protect the inode's lock list.
2737 	*/
2738 
2739 	err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
2740 	switch (-err) {
2741 	case 0: /* success! */
2742 		update_stateid(&lock_stp->st_stateid);
2743 		memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
2744 				sizeof(stateid_t));
2745 		status = 0;
2746 		break;
2747 	case (EAGAIN):		/* conflock holds conflicting lock */
2748 		status = nfserr_denied;
2749 		dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
2750 		nfs4_set_lock_denied(&conflock, &lock->lk_denied);
2751 		break;
2752 	case (EDEADLK):
2753 		status = nfserr_deadlock;
2754 		break;
2755 	default:
2756 		dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
2757 		status = nfserr_resource;
2758 		break;
2759 	}
2760 out:
2761 	if (status && lock->lk_is_new && lock_sop)
2762 		release_stateowner(lock_sop);
2763 	if (lock->lk_replay_owner) {
2764 		nfs4_get_stateowner(lock->lk_replay_owner);
2765 		cstate->replay_owner = lock->lk_replay_owner;
2766 	}
2767 	nfs4_unlock_state();
2768 	return status;
2769 }
2770 
2771 /*
2772  * LOCKT operation
2773  */
2774 __be32
2775 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2776 	    struct nfsd4_lockt *lockt)
2777 {
2778 	struct inode *inode;
2779 	struct file file;
2780 	struct file_lock file_lock;
2781 	int error;
2782 	__be32 status;
2783 
2784 	if (locks_in_grace())
2785 		return nfserr_grace;
2786 
2787 	if (check_lock_length(lockt->lt_offset, lockt->lt_length))
2788 		 return nfserr_inval;
2789 
2790 	lockt->lt_stateowner = NULL;
2791 	nfs4_lock_state();
2792 
2793 	status = nfserr_stale_clientid;
2794 	if (STALE_CLIENTID(&lockt->lt_clientid))
2795 		goto out;
2796 
2797 	if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
2798 		dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
2799 		if (status == nfserr_symlink)
2800 			status = nfserr_inval;
2801 		goto out;
2802 	}
2803 
2804 	inode = cstate->current_fh.fh_dentry->d_inode;
2805 	locks_init_lock(&file_lock);
2806 	switch (lockt->lt_type) {
2807 		case NFS4_READ_LT:
2808 		case NFS4_READW_LT:
2809 			file_lock.fl_type = F_RDLCK;
2810 		break;
2811 		case NFS4_WRITE_LT:
2812 		case NFS4_WRITEW_LT:
2813 			file_lock.fl_type = F_WRLCK;
2814 		break;
2815 		default:
2816 			dprintk("NFSD: nfs4_lockt: bad lock type!\n");
2817 			status = nfserr_inval;
2818 		goto out;
2819 	}
2820 
2821 	lockt->lt_stateowner = find_lockstateowner_str(inode,
2822 			&lockt->lt_clientid, &lockt->lt_owner);
2823 	if (lockt->lt_stateowner)
2824 		file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
2825 	file_lock.fl_pid = current->tgid;
2826 	file_lock.fl_flags = FL_POSIX;
2827 	file_lock.fl_lmops = &nfsd_posix_mng_ops;
2828 
2829 	file_lock.fl_start = lockt->lt_offset;
2830 	if ((lockt->lt_length == ~(u64)0) || LOFF_OVERFLOW(lockt->lt_offset, lockt->lt_length))
2831 		file_lock.fl_end = ~(u64)0;
2832 	else
2833 		file_lock.fl_end = lockt->lt_offset + lockt->lt_length - 1;
2834 
2835 	nfs4_transform_lock_offset(&file_lock);
2836 
2837 	/* vfs_test_lock uses the struct file _only_ to resolve the inode.
2838 	 * since LOCKT doesn't require an OPEN, and therefore a struct
2839 	 * file may not exist, pass vfs_test_lock a struct file with
2840 	 * only the dentry:inode set.
2841 	 */
2842 	memset(&file, 0, sizeof (struct file));
2843 	file.f_path.dentry = cstate->current_fh.fh_dentry;
2844 
2845 	status = nfs_ok;
2846 	error = vfs_test_lock(&file, &file_lock);
2847 	if (error) {
2848 		status = nfserrno(error);
2849 		goto out;
2850 	}
2851 	if (file_lock.fl_type != F_UNLCK) {
2852 		status = nfserr_denied;
2853 		nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
2854 	}
2855 out:
2856 	nfs4_unlock_state();
2857 	return status;
2858 }
2859 
2860 __be32
2861 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2862 	    struct nfsd4_locku *locku)
2863 {
2864 	struct nfs4_stateid *stp;
2865 	struct file *filp = NULL;
2866 	struct file_lock file_lock;
2867 	__be32 status;
2868 	int err;
2869 
2870 	dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
2871 		(long long) locku->lu_offset,
2872 		(long long) locku->lu_length);
2873 
2874 	if (check_lock_length(locku->lu_offset, locku->lu_length))
2875 		 return nfserr_inval;
2876 
2877 	nfs4_lock_state();
2878 
2879 	if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2880 					locku->lu_seqid,
2881 					&locku->lu_stateid,
2882 					LOCK_STATE,
2883 					&locku->lu_stateowner, &stp, NULL)))
2884 		goto out;
2885 
2886 	filp = stp->st_vfs_file;
2887 	BUG_ON(!filp);
2888 	locks_init_lock(&file_lock);
2889 	file_lock.fl_type = F_UNLCK;
2890 	file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
2891 	file_lock.fl_pid = current->tgid;
2892 	file_lock.fl_file = filp;
2893 	file_lock.fl_flags = FL_POSIX;
2894 	file_lock.fl_lmops = &nfsd_posix_mng_ops;
2895 	file_lock.fl_start = locku->lu_offset;
2896 
2897 	if ((locku->lu_length == ~(u64)0) || LOFF_OVERFLOW(locku->lu_offset, locku->lu_length))
2898 		file_lock.fl_end = ~(u64)0;
2899 	else
2900 		file_lock.fl_end = locku->lu_offset + locku->lu_length - 1;
2901 	nfs4_transform_lock_offset(&file_lock);
2902 
2903 	/*
2904 	*  Try to unlock the file in the VFS.
2905 	*/
2906 	err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
2907 	if (err) {
2908 		dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
2909 		goto out_nfserr;
2910 	}
2911 	/*
2912 	* OK, unlock succeeded; the only thing left to do is update the stateid.
2913 	*/
2914 	update_stateid(&stp->st_stateid);
2915 	memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
2916 
2917 out:
2918 	if (locku->lu_stateowner) {
2919 		nfs4_get_stateowner(locku->lu_stateowner);
2920 		cstate->replay_owner = locku->lu_stateowner;
2921 	}
2922 	nfs4_unlock_state();
2923 	return status;
2924 
2925 out_nfserr:
2926 	status = nfserrno(err);
2927 	goto out;
2928 }
2929 
2930 /*
2931  * returns
2932  * 	1: locks held by lockowner
2933  * 	0: no locks held by lockowner
2934  */
2935 static int
2936 check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
2937 {
2938 	struct file_lock **flpp;
2939 	struct inode *inode = filp->f_path.dentry->d_inode;
2940 	int status = 0;
2941 
2942 	lock_kernel();
2943 	for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
2944 		if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
2945 			status = 1;
2946 			goto out;
2947 		}
2948 	}
2949 out:
2950 	unlock_kernel();
2951 	return status;
2952 }
2953 
2954 __be32
2955 nfsd4_release_lockowner(struct svc_rqst *rqstp,
2956 			struct nfsd4_compound_state *cstate,
2957 			struct nfsd4_release_lockowner *rlockowner)
2958 {
2959 	clientid_t *clid = &rlockowner->rl_clientid;
2960 	struct nfs4_stateowner *sop;
2961 	struct nfs4_stateid *stp;
2962 	struct xdr_netobj *owner = &rlockowner->rl_owner;
2963 	struct list_head matches;
2964 	int i;
2965 	__be32 status;
2966 
2967 	dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
2968 		clid->cl_boot, clid->cl_id);
2969 
2970 	/* XXX check for lease expiration */
2971 
2972 	status = nfserr_stale_clientid;
2973 	if (STALE_CLIENTID(clid))
2974 		return status;
2975 
2976 	nfs4_lock_state();
2977 
2978 	status = nfserr_locks_held;
2979 	/* XXX: we're doing a linear search through all the lockowners.
2980 	 * Yipes!  For now we'll just hope clients aren't really using
2981 	 * release_lockowner much, but eventually we have to fix these
2982 	 * data structures. */
2983 	INIT_LIST_HEAD(&matches);
2984 	for (i = 0; i < LOCK_HASH_SIZE; i++) {
2985 		list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
2986 			if (!same_owner_str(sop, owner, clid))
2987 				continue;
2988 			list_for_each_entry(stp, &sop->so_stateids,
2989 					st_perstateowner) {
2990 				if (check_for_locks(stp->st_vfs_file, sop))
2991 					goto out;
2992 				/* Note: so_perclient unused for lockowners,
2993 				 * so it's OK to fool with here. */
2994 				list_add(&sop->so_perclient, &matches);
2995 			}
2996 		}
2997 	}
2998 	/* Clients probably won't expect us to return with some (but not all)
2999 	 * of the lockowner state released; so don't release any until all
3000 	 * have been checked. */
3001 	status = nfs_ok;
3002 	while (!list_empty(&matches)) {
3003 		sop = list_entry(matches.next, struct nfs4_stateowner,
3004 								so_perclient);
3005 		/* unhash_stateowner deletes so_perclient only
3006 		 * for openowners. */
3007 		list_del(&sop->so_perclient);
3008 		release_stateowner(sop);
3009 	}
3010 out:
3011 	nfs4_unlock_state();
3012 	return status;
3013 }
3014 
3015 static inline struct nfs4_client_reclaim *
3016 alloc_reclaim(void)
3017 {
3018 	return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
3019 }
3020 
3021 int
3022 nfs4_has_reclaimed_state(const char *name)
3023 {
3024 	unsigned int strhashval = clientstr_hashval(name);
3025 	struct nfs4_client *clp;
3026 
3027 	clp = find_confirmed_client_by_str(name, strhashval);
3028 	return clp ? 1 : 0;
3029 }
3030 
3031 /*
3032  * failure => all reset bets are off, nfserr_no_grace...
3033  */
3034 int
3035 nfs4_client_to_reclaim(const char *name)
3036 {
3037 	unsigned int strhashval;
3038 	struct nfs4_client_reclaim *crp = NULL;
3039 
3040 	dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3041 	crp = alloc_reclaim();
3042 	if (!crp)
3043 		return 0;
3044 	strhashval = clientstr_hashval(name);
3045 	INIT_LIST_HEAD(&crp->cr_strhash);
3046 	list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
3047 	memcpy(crp->cr_recdir, name, HEXDIR_LEN);
3048 	reclaim_str_hashtbl_size++;
3049 	return 1;
3050 }
3051 
3052 static void
3053 nfs4_release_reclaim(void)
3054 {
3055 	struct nfs4_client_reclaim *crp = NULL;
3056 	int i;
3057 
3058 	for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3059 		while (!list_empty(&reclaim_str_hashtbl[i])) {
3060 			crp = list_entry(reclaim_str_hashtbl[i].next,
3061 			                struct nfs4_client_reclaim, cr_strhash);
3062 			list_del(&crp->cr_strhash);
3063 			kfree(crp);
3064 			reclaim_str_hashtbl_size--;
3065 		}
3066 	}
3067 	BUG_ON(reclaim_str_hashtbl_size);
3068 }
3069 
3070 /*
3071  * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
3072 static struct nfs4_client_reclaim *
3073 nfs4_find_reclaim_client(clientid_t *clid)
3074 {
3075 	unsigned int strhashval;
3076 	struct nfs4_client *clp;
3077 	struct nfs4_client_reclaim *crp = NULL;
3078 
3079 
3080 	/* find clientid in conf_id_hashtbl */
3081 	clp = find_confirmed_client(clid);
3082 	if (clp == NULL)
3083 		return NULL;
3084 
3085 	dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3086 		            clp->cl_name.len, clp->cl_name.data,
3087 			    clp->cl_recdir);
3088 
3089 	/* find clp->cl_name in reclaim_str_hashtbl */
3090 	strhashval = clientstr_hashval(clp->cl_recdir);
3091 	list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
3092 		if (same_name(crp->cr_recdir, clp->cl_recdir)) {
3093 			return crp;
3094 		}
3095 	}
3096 	return NULL;
3097 }
3098 
3099 /*
3100 * Called from OPEN. Look for clientid in reclaim list.
3101 */
3102 __be32
3103 nfs4_check_open_reclaim(clientid_t *clid)
3104 {
3105 	return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
3106 }
3107 
3108 /* initialization to perform at module load time: */
3109 
3110 int
3111 nfs4_state_init(void)
3112 {
3113 	int i, status;
3114 
3115 	status = nfsd4_init_slabs();
3116 	if (status)
3117 		return status;
3118 	for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3119 		INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3120 		INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3121 		INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3122 		INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
3123 	}
3124 	for (i = 0; i < FILE_HASH_SIZE; i++) {
3125 		INIT_LIST_HEAD(&file_hashtbl[i]);
3126 	}
3127 	for (i = 0; i < OWNER_HASH_SIZE; i++) {
3128 		INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3129 		INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3130 	}
3131 	for (i = 0; i < STATEID_HASH_SIZE; i++) {
3132 		INIT_LIST_HEAD(&stateid_hashtbl[i]);
3133 		INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3134 	}
3135 	for (i = 0; i < LOCK_HASH_SIZE; i++) {
3136 		INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3137 		INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3138 	}
3139 	memset(&onestateid, ~0, sizeof(stateid_t));
3140 	INIT_LIST_HEAD(&close_lru);
3141 	INIT_LIST_HEAD(&client_lru);
3142 	INIT_LIST_HEAD(&del_recall_lru);
3143 	for (i = 0; i < CLIENT_HASH_SIZE; i++)
3144 		INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
3145 	reclaim_str_hashtbl_size = 0;
3146 	return 0;
3147 }
3148 
3149 static void
3150 nfsd4_load_reboot_recovery_data(void)
3151 {
3152 	int status;
3153 
3154 	nfs4_lock_state();
3155 	nfsd4_init_recdir(user_recovery_dirname);
3156 	status = nfsd4_recdir_load();
3157 	nfs4_unlock_state();
3158 	if (status)
3159 		printk("NFSD: Failure reading reboot recovery data\n");
3160 }
3161 
3162 unsigned long
3163 get_nfs4_grace_period(void)
3164 {
3165 	return max(user_lease_time, lease_time) * HZ;
3166 }
3167 
3168 /*
3169  * Since the lifetime of a delegation isn't limited to that of an open, a
3170  * client may quite reasonably hang on to a delegation as long as it has
3171  * the inode cached.  This becomes an obvious problem the first time a
3172  * client's inode cache approaches the size of the server's total memory.
3173  *
3174  * For now we avoid this problem by imposing a hard limit on the number
3175  * of delegations, which varies according to the server's memory size.
3176  */
3177 static void
3178 set_max_delegations(void)
3179 {
3180 	/*
3181 	 * Allow at most 4 delegations per megabyte of RAM.  Quick
3182 	 * estimates suggest that in the worst case (where every delegation
3183 	 * is for a different inode), a delegation could take about 1.5K,
3184 	 * giving a worst case usage of about 6% of memory.
3185 	 */
3186 	max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
3187 }
3188 
3189 /* initialization to perform when the nfsd service is started: */
3190 
3191 static void
3192 __nfs4_state_start(void)
3193 {
3194 	unsigned long grace_time;
3195 
3196 	boot_time = get_seconds();
3197 	grace_time = get_nfs4_grace_period();
3198 	lease_time = user_lease_time;
3199 	locks_start_grace(&nfsd4_manager);
3200 	printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
3201 	       grace_time/HZ);
3202 	laundry_wq = create_singlethread_workqueue("nfsd4");
3203 	queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
3204 	set_max_delegations();
3205 }
3206 
3207 void
3208 nfs4_state_start(void)
3209 {
3210 	if (nfs4_init)
3211 		return;
3212 	nfsd4_load_reboot_recovery_data();
3213 	__nfs4_state_start();
3214 	nfs4_init = 1;
3215 	return;
3216 }
3217 
3218 time_t
3219 nfs4_lease_time(void)
3220 {
3221 	return lease_time;
3222 }
3223 
3224 static void
3225 __nfs4_state_shutdown(void)
3226 {
3227 	int i;
3228 	struct nfs4_client *clp = NULL;
3229 	struct nfs4_delegation *dp = NULL;
3230 	struct list_head *pos, *next, reaplist;
3231 
3232 	for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3233 		while (!list_empty(&conf_id_hashtbl[i])) {
3234 			clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
3235 			expire_client(clp);
3236 		}
3237 		while (!list_empty(&unconf_str_hashtbl[i])) {
3238 			clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
3239 			expire_client(clp);
3240 		}
3241 	}
3242 	INIT_LIST_HEAD(&reaplist);
3243 	spin_lock(&recall_lock);
3244 	list_for_each_safe(pos, next, &del_recall_lru) {
3245 		dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3246 		list_move(&dp->dl_recall_lru, &reaplist);
3247 	}
3248 	spin_unlock(&recall_lock);
3249 	list_for_each_safe(pos, next, &reaplist) {
3250 		dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3251 		list_del_init(&dp->dl_recall_lru);
3252 		unhash_delegation(dp);
3253 	}
3254 
3255 	nfsd4_shutdown_recdir();
3256 	nfs4_init = 0;
3257 }
3258 
3259 void
3260 nfs4_state_shutdown(void)
3261 {
3262 	cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
3263 	destroy_workqueue(laundry_wq);
3264 	nfs4_lock_state();
3265 	nfs4_release_reclaim();
3266 	__nfs4_state_shutdown();
3267 	nfs4_unlock_state();
3268 }
3269 
3270 /*
3271  * user_recovery_dirname is protected by the nfsd_mutex since it's only
3272  * accessed when nfsd is starting.
3273  */
3274 static void
3275 nfs4_set_recdir(char *recdir)
3276 {
3277 	strcpy(user_recovery_dirname, recdir);
3278 }
3279 
3280 /*
3281  * Change the NFSv4 recovery directory to recdir.
3282  */
3283 int
3284 nfs4_reset_recoverydir(char *recdir)
3285 {
3286 	int status;
3287 	struct path path;
3288 
3289 	status = kern_path(recdir, LOOKUP_FOLLOW, &path);
3290 	if (status)
3291 		return status;
3292 	status = -ENOTDIR;
3293 	if (S_ISDIR(path.dentry->d_inode->i_mode)) {
3294 		nfs4_set_recdir(recdir);
3295 		status = 0;
3296 	}
3297 	path_put(&path);
3298 	return status;
3299 }
3300 
3301 char *
3302 nfs4_recoverydir(void)
3303 {
3304 	return user_recovery_dirname;
3305 }
3306 
3307 /*
3308  * Called when leasetime is changed.
3309  *
3310  * The only way the protocol gives us to handle on-the-fly lease changes is to
3311  * simulate a reboot.  Instead of doing that, we just wait till the next time
3312  * we start to register any changes in lease time.  If the administrator
3313  * really wants to change the lease time *now*, they can go ahead and bring
3314  * nfsd down and then back up again after changing the lease time.
3315  *
3316  * user_lease_time is protected by nfsd_mutex since it's only really accessed
3317  * when nfsd is starting
3318  */
3319 void
3320 nfs4_reset_lease(time_t leasetime)
3321 {
3322 	user_lease_time = leasetime;
3323 }
3324