xref: /linux/fs/nfsd/nfs4state.c (revision 776cfebb430c7b22c208b1b17add97f354d97cab)
1 /*
2 *  linux/fs/nfsd/nfs4state.c
3 *
4 *  Copyright (c) 2001 The Regents of the University of Michigan.
5 *  All rights reserved.
6 *
7 *  Kendrick Smith <kmsmith@umich.edu>
8 *  Andy Adamson <kandros@umich.edu>
9 *
10 *  Redistribution and use in source and binary forms, with or without
11 *  modification, are permitted provided that the following conditions
12 *  are met:
13 *
14 *  1. Redistributions of source code must retain the above copyright
15 *     notice, this list of conditions and the following disclaimer.
16 *  2. Redistributions in binary form must reproduce the above copyright
17 *     notice, this list of conditions and the following disclaimer in the
18 *     documentation and/or other materials provided with the distribution.
19 *  3. Neither the name of the University nor the names of its
20 *     contributors may be used to endorse or promote products derived
21 *     from this software without specific prior written permission.
22 *
23 *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
24 *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
25 *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
26 *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28 *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29 *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
30 *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
31 *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
32 *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
33 *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34 *
35 */
36 
37 #include <linux/param.h>
38 #include <linux/major.h>
39 #include <linux/slab.h>
40 
41 #include <linux/sunrpc/svc.h>
42 #include <linux/nfsd/nfsd.h>
43 #include <linux/nfsd/cache.h>
44 #include <linux/mount.h>
45 #include <linux/workqueue.h>
46 #include <linux/smp_lock.h>
47 #include <linux/kthread.h>
48 #include <linux/nfs4.h>
49 #include <linux/nfsd/state.h>
50 #include <linux/nfsd/xdr4.h>
51 
52 #define NFSDDBG_FACILITY                NFSDDBG_PROC
53 
54 /* Globals */
55 static time_t lease_time = 90;     /* default lease time */
56 static time_t old_lease_time = 90; /* past incarnation lease time */
57 static u32 nfs4_reclaim_init = 0;
58 time_t boot_time;
59 static time_t grace_end = 0;
60 static u32 current_clientid = 1;
61 static u32 current_ownerid = 1;
62 static u32 current_fileid = 1;
63 static u32 current_delegid = 1;
64 static u32 nfs4_init;
65 stateid_t zerostateid;             /* bits all 0 */
66 stateid_t onestateid;              /* bits all 1 */
67 
68 /* debug counters */
69 u32 list_add_perfile = 0;
70 u32 list_del_perfile = 0;
71 u32 add_perclient = 0;
72 u32 del_perclient = 0;
73 u32 alloc_file = 0;
74 u32 free_file = 0;
75 u32 vfsopen = 0;
76 u32 vfsclose = 0;
77 u32 alloc_delegation= 0;
78 u32 free_delegation= 0;
79 
80 /* forward declarations */
81 struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
82 static struct nfs4_delegation * find_delegation_stateid(struct inode *ino, stateid_t *stid);
83 static void release_stateid_lockowners(struct nfs4_stateid *open_stp);
84 
85 /* Locking:
86  *
87  * client_sema:
88  * 	protects clientid_hashtbl[], clientstr_hashtbl[],
89  * 	unconfstr_hashtbl[], uncofid_hashtbl[].
90  */
91 static DECLARE_MUTEX(client_sema);
92 
93 void
94 nfs4_lock_state(void)
95 {
96 	down(&client_sema);
97 }
98 
99 void
100 nfs4_unlock_state(void)
101 {
102 	up(&client_sema);
103 }
104 
105 static inline u32
106 opaque_hashval(const void *ptr, int nbytes)
107 {
108 	unsigned char *cptr = (unsigned char *) ptr;
109 
110 	u32 x = 0;
111 	while (nbytes--) {
112 		x *= 37;
113 		x += *cptr++;
114 	}
115 	return x;
116 }
117 
118 /* forward declarations */
119 static void release_stateowner(struct nfs4_stateowner *sop);
120 static void release_stateid(struct nfs4_stateid *stp, int flags);
121 static void release_file(struct nfs4_file *fp);
122 
123 /*
124  * Delegation state
125  */
126 
127 /* recall_lock protects the del_recall_lru */
128 spinlock_t recall_lock;
129 static struct list_head del_recall_lru;
130 
131 static struct nfs4_delegation *
132 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
133 {
134 	struct nfs4_delegation *dp;
135 	struct nfs4_file *fp = stp->st_file;
136 	struct nfs4_callback *cb = &stp->st_stateowner->so_client->cl_callback;
137 
138 	dprintk("NFSD alloc_init_deleg\n");
139 	if ((dp = kmalloc(sizeof(struct nfs4_delegation),
140 		GFP_KERNEL)) == NULL)
141 		return dp;
142 	INIT_LIST_HEAD(&dp->dl_del_perfile);
143 	INIT_LIST_HEAD(&dp->dl_del_perclnt);
144 	INIT_LIST_HEAD(&dp->dl_recall_lru);
145 	dp->dl_client = clp;
146 	dp->dl_file = fp;
147 	dp->dl_flock = NULL;
148 	get_file(stp->st_vfs_file);
149 	dp->dl_vfs_file = stp->st_vfs_file;
150 	dp->dl_type = type;
151 	dp->dl_recall.cbr_dp = NULL;
152 	dp->dl_recall.cbr_ident = cb->cb_ident;
153 	dp->dl_recall.cbr_trunc = 0;
154 	dp->dl_stateid.si_boot = boot_time;
155 	dp->dl_stateid.si_stateownerid = current_delegid++;
156 	dp->dl_stateid.si_fileid = 0;
157 	dp->dl_stateid.si_generation = 0;
158 	dp->dl_fhlen = current_fh->fh_handle.fh_size;
159 	memcpy(dp->dl_fhval, &current_fh->fh_handle.fh_base,
160 		        current_fh->fh_handle.fh_size);
161 	dp->dl_time = 0;
162 	atomic_set(&dp->dl_count, 1);
163 	list_add(&dp->dl_del_perfile, &fp->fi_del_perfile);
164 	list_add(&dp->dl_del_perclnt, &clp->cl_del_perclnt);
165 	alloc_delegation++;
166 	return dp;
167 }
168 
169 void
170 nfs4_put_delegation(struct nfs4_delegation *dp)
171 {
172 	if (atomic_dec_and_test(&dp->dl_count)) {
173 		dprintk("NFSD: freeing dp %p\n",dp);
174 		kfree(dp);
175 		free_delegation++;
176 	}
177 }
178 
179 /* Remove the associated file_lock first, then remove the delegation.
180  * lease_modify() is called to remove the FS_LEASE file_lock from
181  * the i_flock list, eventually calling nfsd's lock_manager
182  * fl_release_callback.
183  */
184 static void
185 nfs4_close_delegation(struct nfs4_delegation *dp)
186 {
187 	struct file *filp = dp->dl_vfs_file;
188 
189 	dprintk("NFSD: close_delegation dp %p\n",dp);
190 	dp->dl_vfs_file = NULL;
191 	/* The following nfsd_close may not actually close the file,
192 	 * but we want to remove the lease in any case. */
193 	if (dp->dl_flock)
194 		setlease(filp, F_UNLCK, &dp->dl_flock);
195 	nfsd_close(filp);
196 	vfsclose++;
197 }
198 
199 /* Called under the state lock. */
200 static void
201 unhash_delegation(struct nfs4_delegation *dp)
202 {
203 	list_del_init(&dp->dl_del_perfile);
204 	list_del_init(&dp->dl_del_perclnt);
205 	spin_lock(&recall_lock);
206 	list_del_init(&dp->dl_recall_lru);
207 	spin_unlock(&recall_lock);
208 	nfs4_close_delegation(dp);
209 	nfs4_put_delegation(dp);
210 }
211 
212 /*
213  * SETCLIENTID state
214  */
215 
216 /* Hash tables for nfs4_clientid state */
217 #define CLIENT_HASH_BITS                 4
218 #define CLIENT_HASH_SIZE                (1 << CLIENT_HASH_BITS)
219 #define CLIENT_HASH_MASK                (CLIENT_HASH_SIZE - 1)
220 
221 #define clientid_hashval(id) \
222 	((id) & CLIENT_HASH_MASK)
223 #define clientstr_hashval(name, namelen) \
224 	(opaque_hashval((name), (namelen)) & CLIENT_HASH_MASK)
225 /*
226  * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
227  * used in reboot/reset lease grace period processing
228  *
229  * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
230  * setclientid_confirmed info.
231  *
232  * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
233  * setclientid info.
234  *
235  * client_lru holds client queue ordered by nfs4_client.cl_time
236  * for lease renewal.
237  *
238  * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
239  * for last close replay.
240  */
241 static struct list_head	reclaim_str_hashtbl[CLIENT_HASH_SIZE];
242 static int reclaim_str_hashtbl_size = 0;
243 static struct list_head	conf_id_hashtbl[CLIENT_HASH_SIZE];
244 static struct list_head	conf_str_hashtbl[CLIENT_HASH_SIZE];
245 static struct list_head	unconf_str_hashtbl[CLIENT_HASH_SIZE];
246 static struct list_head	unconf_id_hashtbl[CLIENT_HASH_SIZE];
247 static struct list_head client_lru;
248 static struct list_head close_lru;
249 
250 static inline void
251 renew_client(struct nfs4_client *clp)
252 {
253 	/*
254 	* Move client to the end to the LRU list.
255 	*/
256 	dprintk("renewing client (clientid %08x/%08x)\n",
257 			clp->cl_clientid.cl_boot,
258 			clp->cl_clientid.cl_id);
259 	list_move_tail(&clp->cl_lru, &client_lru);
260 	clp->cl_time = get_seconds();
261 }
262 
263 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
264 static int
265 STALE_CLIENTID(clientid_t *clid)
266 {
267 	if (clid->cl_boot == boot_time)
268 		return 0;
269 	dprintk("NFSD stale clientid (%08x/%08x)\n",
270 			clid->cl_boot, clid->cl_id);
271 	return 1;
272 }
273 
274 /*
275  * XXX Should we use a slab cache ?
276  * This type of memory management is somewhat inefficient, but we use it
277  * anyway since SETCLIENTID is not a common operation.
278  */
279 static inline struct nfs4_client *
280 alloc_client(struct xdr_netobj name)
281 {
282 	struct nfs4_client *clp;
283 
284 	if ((clp = kmalloc(sizeof(struct nfs4_client), GFP_KERNEL))!= NULL) {
285 		memset(clp, 0, sizeof(*clp));
286 		if ((clp->cl_name.data = kmalloc(name.len, GFP_KERNEL)) != NULL) {
287 			memcpy(clp->cl_name.data, name.data, name.len);
288 			clp->cl_name.len = name.len;
289 		}
290 		else {
291 			kfree(clp);
292 			clp = NULL;
293 		}
294 	}
295 	return clp;
296 }
297 
298 static inline void
299 free_client(struct nfs4_client *clp)
300 {
301 	if (clp->cl_cred.cr_group_info)
302 		put_group_info(clp->cl_cred.cr_group_info);
303 	kfree(clp->cl_name.data);
304 	kfree(clp);
305 }
306 
307 void
308 put_nfs4_client(struct nfs4_client *clp)
309 {
310 	if (atomic_dec_and_test(&clp->cl_count))
311 		free_client(clp);
312 }
313 
314 static void
315 expire_client(struct nfs4_client *clp)
316 {
317 	struct nfs4_stateowner *sop;
318 	struct nfs4_delegation *dp;
319 	struct nfs4_callback *cb = &clp->cl_callback;
320 	struct rpc_clnt *clnt = clp->cl_callback.cb_client;
321 	struct list_head reaplist;
322 
323 	dprintk("NFSD: expire_client cl_count %d\n",
324 	                    atomic_read(&clp->cl_count));
325 
326 	/* shutdown rpc client, ending any outstanding recall rpcs */
327 	if (atomic_read(&cb->cb_set) == 1 && clnt) {
328 		rpc_shutdown_client(clnt);
329 		clnt = clp->cl_callback.cb_client = NULL;
330 	}
331 
332 	INIT_LIST_HEAD(&reaplist);
333 	spin_lock(&recall_lock);
334 	while (!list_empty(&clp->cl_del_perclnt)) {
335 		dp = list_entry(clp->cl_del_perclnt.next, struct nfs4_delegation, dl_del_perclnt);
336 		dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
337 				dp->dl_flock);
338 		list_del_init(&dp->dl_del_perclnt);
339 		list_move(&dp->dl_recall_lru, &reaplist);
340 	}
341 	spin_unlock(&recall_lock);
342 	while (!list_empty(&reaplist)) {
343 		dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
344 		list_del_init(&dp->dl_recall_lru);
345 		unhash_delegation(dp);
346 	}
347 	list_del(&clp->cl_idhash);
348 	list_del(&clp->cl_strhash);
349 	list_del(&clp->cl_lru);
350 	while (!list_empty(&clp->cl_perclient)) {
351 		sop = list_entry(clp->cl_perclient.next, struct nfs4_stateowner, so_perclient);
352 		release_stateowner(sop);
353 	}
354 	put_nfs4_client(clp);
355 }
356 
357 static struct nfs4_client *
358 create_client(struct xdr_netobj name) {
359 	struct nfs4_client *clp;
360 
361 	if (!(clp = alloc_client(name)))
362 		goto out;
363 	atomic_set(&clp->cl_count, 1);
364 	atomic_set(&clp->cl_callback.cb_set, 0);
365 	clp->cl_callback.cb_parsed = 0;
366 	INIT_LIST_HEAD(&clp->cl_idhash);
367 	INIT_LIST_HEAD(&clp->cl_strhash);
368 	INIT_LIST_HEAD(&clp->cl_perclient);
369 	INIT_LIST_HEAD(&clp->cl_del_perclnt);
370 	INIT_LIST_HEAD(&clp->cl_lru);
371 out:
372 	return clp;
373 }
374 
375 static void
376 copy_verf(struct nfs4_client *target, nfs4_verifier *source) {
377 	memcpy(target->cl_verifier.data, source->data, sizeof(target->cl_verifier.data));
378 }
379 
380 static void
381 copy_clid(struct nfs4_client *target, struct nfs4_client *source) {
382 	target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
383 	target->cl_clientid.cl_id = source->cl_clientid.cl_id;
384 }
385 
386 static void
387 copy_cred(struct svc_cred *target, struct svc_cred *source) {
388 
389 	target->cr_uid = source->cr_uid;
390 	target->cr_gid = source->cr_gid;
391 	target->cr_group_info = source->cr_group_info;
392 	get_group_info(target->cr_group_info);
393 }
394 
395 static int
396 cmp_name(struct xdr_netobj *n1, struct xdr_netobj *n2) {
397 	if (!n1 || !n2)
398 		return 0;
399 	return((n1->len == n2->len) && !memcmp(n1->data, n2->data, n2->len));
400 }
401 
402 static int
403 cmp_verf(nfs4_verifier *v1, nfs4_verifier *v2) {
404 	return(!memcmp(v1->data,v2->data,sizeof(v1->data)));
405 }
406 
407 static int
408 cmp_clid(clientid_t * cl1, clientid_t * cl2) {
409 	return((cl1->cl_boot == cl2->cl_boot) &&
410 	   	(cl1->cl_id == cl2->cl_id));
411 }
412 
413 /* XXX what about NGROUP */
414 static int
415 cmp_creds(struct svc_cred *cr1, struct svc_cred *cr2){
416 	return(cr1->cr_uid == cr2->cr_uid);
417 
418 }
419 
420 static void
421 gen_clid(struct nfs4_client *clp) {
422 	clp->cl_clientid.cl_boot = boot_time;
423 	clp->cl_clientid.cl_id = current_clientid++;
424 }
425 
426 static void
427 gen_confirm(struct nfs4_client *clp) {
428 	struct timespec 	tv;
429 	u32 *			p;
430 
431 	tv = CURRENT_TIME;
432 	p = (u32 *)clp->cl_confirm.data;
433 	*p++ = tv.tv_sec;
434 	*p++ = tv.tv_nsec;
435 }
436 
437 static int
438 check_name(struct xdr_netobj name) {
439 
440 	if (name.len == 0)
441 		return 0;
442 	if (name.len > NFS4_OPAQUE_LIMIT) {
443 		printk("NFSD: check_name: name too long(%d)!\n", name.len);
444 		return 0;
445 	}
446 	return 1;
447 }
448 
449 void
450 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
451 {
452 	unsigned int idhashval;
453 
454 	list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
455 	idhashval = clientid_hashval(clp->cl_clientid.cl_id);
456 	list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
457 	list_add_tail(&clp->cl_lru, &client_lru);
458 	clp->cl_time = get_seconds();
459 }
460 
461 void
462 move_to_confirmed(struct nfs4_client *clp)
463 {
464 	unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
465 	unsigned int strhashval;
466 
467 	dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
468 	list_del_init(&clp->cl_strhash);
469 	list_del_init(&clp->cl_idhash);
470 	list_add(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
471 	strhashval = clientstr_hashval(clp->cl_name.data,
472 			clp->cl_name.len);
473 	list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
474 	renew_client(clp);
475 }
476 
477 static struct nfs4_client *
478 find_confirmed_client(clientid_t *clid)
479 {
480 	struct nfs4_client *clp;
481 	unsigned int idhashval = clientid_hashval(clid->cl_id);
482 
483 	list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
484 		if (cmp_clid(&clp->cl_clientid, clid))
485 			return clp;
486 	}
487 	return NULL;
488 }
489 
490 static struct nfs4_client *
491 find_unconfirmed_client(clientid_t *clid)
492 {
493 	struct nfs4_client *clp;
494 	unsigned int idhashval = clientid_hashval(clid->cl_id);
495 
496 	list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
497 		if (cmp_clid(&clp->cl_clientid, clid))
498 			return clp;
499 	}
500 	return NULL;
501 }
502 
503 /* a helper function for parse_callback */
504 static int
505 parse_octet(unsigned int *lenp, char **addrp)
506 {
507 	unsigned int len = *lenp;
508 	char *p = *addrp;
509 	int n = -1;
510 	char c;
511 
512 	for (;;) {
513 		if (!len)
514 			break;
515 		len--;
516 		c = *p++;
517 		if (c == '.')
518 			break;
519 		if ((c < '0') || (c > '9')) {
520 			n = -1;
521 			break;
522 		}
523 		if (n < 0)
524 			n = 0;
525 		n = (n * 10) + (c - '0');
526 		if (n > 255) {
527 			n = -1;
528 			break;
529 		}
530 	}
531 	*lenp = len;
532 	*addrp = p;
533 	return n;
534 }
535 
536 /* parse and set the setclientid ipv4 callback address */
537 int
538 parse_ipv4(unsigned int addr_len, char *addr_val, unsigned int *cbaddrp, unsigned short *cbportp)
539 {
540 	int temp = 0;
541 	u32 cbaddr = 0;
542 	u16 cbport = 0;
543 	u32 addrlen = addr_len;
544 	char *addr = addr_val;
545 	int i, shift;
546 
547 	/* ipaddress */
548 	shift = 24;
549 	for(i = 4; i > 0  ; i--) {
550 		if ((temp = parse_octet(&addrlen, &addr)) < 0) {
551 			return 0;
552 		}
553 		cbaddr |= (temp << shift);
554 		if (shift > 0)
555 		shift -= 8;
556 	}
557 	*cbaddrp = cbaddr;
558 
559 	/* port */
560 	shift = 8;
561 	for(i = 2; i > 0  ; i--) {
562 		if ((temp = parse_octet(&addrlen, &addr)) < 0) {
563 			return 0;
564 		}
565 		cbport |= (temp << shift);
566 		if (shift > 0)
567 			shift -= 8;
568 	}
569 	*cbportp = cbport;
570 	return 1;
571 }
572 
573 void
574 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se)
575 {
576 	struct nfs4_callback *cb = &clp->cl_callback;
577 
578 	/* Currently, we only support tcp for the callback channel */
579 	if ((se->se_callback_netid_len != 3) || memcmp((char *)se->se_callback_netid_val, "tcp", 3))
580 		goto out_err;
581 
582 	if ( !(parse_ipv4(se->se_callback_addr_len, se->se_callback_addr_val,
583 	                 &cb->cb_addr, &cb->cb_port)))
584 		goto out_err;
585 	cb->cb_prog = se->se_callback_prog;
586 	cb->cb_ident = se->se_callback_ident;
587 	cb->cb_parsed = 1;
588 	return;
589 out_err:
590 	printk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
591 		"will not receive delegations\n",
592 		clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
593 
594 	cb->cb_parsed = 0;
595 	return;
596 }
597 
598 /*
599  * RFC 3010 has a complex implmentation description of processing a
600  * SETCLIENTID request consisting of 5 bullets, labeled as
601  * CASE0 - CASE4 below.
602  *
603  * NOTES:
604  * 	callback information will be processed in a future patch
605  *
606  *	an unconfirmed record is added when:
607  *      NORMAL (part of CASE 4): there is no confirmed nor unconfirmed record.
608  *	CASE 1: confirmed record found with matching name, principal,
609  *		verifier, and clientid.
610  *	CASE 2: confirmed record found with matching name, principal,
611  *		and there is no unconfirmed record with matching
612  *		name and principal
613  *
614  *      an unconfirmed record is replaced when:
615  *	CASE 3: confirmed record found with matching name, principal,
616  *		and an unconfirmed record is found with matching
617  *		name, principal, and with clientid and
618  *		confirm that does not match the confirmed record.
619  *	CASE 4: there is no confirmed record with matching name and
620  *		principal. there is an unconfirmed record with
621  *		matching name, principal.
622  *
623  *	an unconfirmed record is deleted when:
624  *	CASE 1: an unconfirmed record that matches input name, verifier,
625  *		and confirmed clientid.
626  *	CASE 4: any unconfirmed records with matching name and principal
627  *		that exist after an unconfirmed record has been replaced
628  *		as described above.
629  *
630  */
631 int
632 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_setclientid *setclid)
633 {
634 	u32 			ip_addr = rqstp->rq_addr.sin_addr.s_addr;
635 	struct xdr_netobj 	clname = {
636 		.len = setclid->se_namelen,
637 		.data = setclid->se_name,
638 	};
639 	nfs4_verifier		clverifier = setclid->se_verf;
640 	unsigned int 		strhashval;
641 	struct nfs4_client *	conf, * unconf, * new, * clp;
642 	int 			status;
643 
644 	status = nfserr_inval;
645 	if (!check_name(clname))
646 		goto out;
647 
648 	/*
649 	 * XXX The Duplicate Request Cache (DRC) has been checked (??)
650 	 * We get here on a DRC miss.
651 	 */
652 
653 	strhashval = clientstr_hashval(clname.data, clname.len);
654 
655 	conf = NULL;
656 	nfs4_lock_state();
657 	list_for_each_entry(clp, &conf_str_hashtbl[strhashval], cl_strhash) {
658 		if (!cmp_name(&clp->cl_name, &clname))
659 			continue;
660 		/*
661 		 * CASE 0:
662 		 * clname match, confirmed, different principal
663 		 * or different ip_address
664 		 */
665 		status = nfserr_clid_inuse;
666 		if (!cmp_creds(&clp->cl_cred,&rqstp->rq_cred)) {
667 			printk("NFSD: setclientid: string in use by client"
668 			"(clientid %08x/%08x)\n",
669 			clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
670 			goto out;
671 		}
672 		if (clp->cl_addr != ip_addr) {
673 			printk("NFSD: setclientid: string in use by client"
674 			"(clientid %08x/%08x)\n",
675 			clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
676 			goto out;
677 		}
678 
679 		/*
680 	 	 * cl_name match from a previous SETCLIENTID operation
681 	 	 * XXX check for additional matches?
682 		 */
683 		conf = clp;
684 		break;
685 	}
686 	unconf = NULL;
687 	list_for_each_entry(clp, &unconf_str_hashtbl[strhashval], cl_strhash) {
688 		if (!cmp_name(&clp->cl_name, &clname))
689 			continue;
690 		/* cl_name match from a previous SETCLIENTID operation */
691 		unconf = clp;
692 		break;
693 	}
694 	status = nfserr_resource;
695 	if (!conf) {
696 		/*
697 		 * CASE 4:
698 		 * placed first, because it is the normal case.
699 		 */
700 		if (unconf)
701 			expire_client(unconf);
702 		if (!(new = create_client(clname)))
703 			goto out;
704 		copy_verf(new, &clverifier);
705 		new->cl_addr = ip_addr;
706 		copy_cred(&new->cl_cred,&rqstp->rq_cred);
707 		gen_clid(new);
708 		gen_confirm(new);
709 		gen_callback(new, setclid);
710 		add_to_unconfirmed(new, strhashval);
711 	} else if (cmp_verf(&conf->cl_verifier, &clverifier)) {
712 		/*
713 		 * CASE 1:
714 		 * cl_name match, confirmed, principal match
715 		 * verifier match: probable callback update
716 		 *
717 		 * remove any unconfirmed nfs4_client with
718 		 * matching cl_name, cl_verifier, and cl_clientid
719 		 *
720 		 * create and insert an unconfirmed nfs4_client with same
721 		 * cl_name, cl_verifier, and cl_clientid as existing
722 		 * nfs4_client,  but with the new callback info and a
723 		 * new cl_confirm
724 		 */
725 		if ((unconf) &&
726 		    cmp_verf(&unconf->cl_verifier, &conf->cl_verifier) &&
727 		     cmp_clid(&unconf->cl_clientid, &conf->cl_clientid)) {
728 				expire_client(unconf);
729 		}
730 		if (!(new = create_client(clname)))
731 			goto out;
732 		copy_verf(new,&conf->cl_verifier);
733 		new->cl_addr = ip_addr;
734 		copy_cred(&new->cl_cred,&rqstp->rq_cred);
735 		copy_clid(new, conf);
736 		gen_confirm(new);
737 		gen_callback(new, setclid);
738 		add_to_unconfirmed(new,strhashval);
739 	} else if (!unconf) {
740 		/*
741 		 * CASE 2:
742 		 * clname match, confirmed, principal match
743 		 * verfier does not match
744 		 * no unconfirmed. create a new unconfirmed nfs4_client
745 		 * using input clverifier, clname, and callback info
746 		 * and generate a new cl_clientid and cl_confirm.
747 		 */
748 		if (!(new = create_client(clname)))
749 			goto out;
750 		copy_verf(new,&clverifier);
751 		new->cl_addr = ip_addr;
752 		copy_cred(&new->cl_cred,&rqstp->rq_cred);
753 		gen_clid(new);
754 		gen_confirm(new);
755 		gen_callback(new, setclid);
756 		add_to_unconfirmed(new, strhashval);
757 	} else if (!cmp_verf(&conf->cl_confirm, &unconf->cl_confirm)) {
758 		/*
759 		 * CASE3:
760 		 * confirmed found (name, principal match)
761 		 * confirmed verifier does not match input clverifier
762 		 *
763 		 * unconfirmed found (name match)
764 		 * confirmed->cl_confirm != unconfirmed->cl_confirm
765 		 *
766 		 * remove unconfirmed.
767 		 *
768 		 * create an unconfirmed nfs4_client
769 		 * with same cl_name as existing confirmed nfs4_client,
770 		 * but with new callback info, new cl_clientid,
771 		 * new cl_verifier and a new cl_confirm
772 		 */
773 		expire_client(unconf);
774 		if (!(new = create_client(clname)))
775 			goto out;
776 		copy_verf(new,&clverifier);
777 		new->cl_addr = ip_addr;
778 		copy_cred(&new->cl_cred,&rqstp->rq_cred);
779 		gen_clid(new);
780 		gen_confirm(new);
781 		gen_callback(new, setclid);
782 		add_to_unconfirmed(new, strhashval);
783 	} else {
784 		/* No cases hit !!! */
785 		status = nfserr_inval;
786 		goto out;
787 
788 	}
789 	setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
790 	setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
791 	memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
792 	status = nfs_ok;
793 out:
794 	nfs4_unlock_state();
795 	return status;
796 }
797 
798 
799 /*
800  * RFC 3010 has a complex implmentation description of processing a
801  * SETCLIENTID_CONFIRM request consisting of 4 bullets describing
802  * processing on a DRC miss, labeled as CASE1 - CASE4 below.
803  *
804  * NOTE: callback information will be processed here in a future patch
805  */
806 int
807 nfsd4_setclientid_confirm(struct svc_rqst *rqstp, struct nfsd4_setclientid_confirm *setclientid_confirm)
808 {
809 	u32 ip_addr = rqstp->rq_addr.sin_addr.s_addr;
810 	struct nfs4_client *clp, *conf = NULL, *unconf = NULL;
811 	nfs4_verifier confirm = setclientid_confirm->sc_confirm;
812 	clientid_t * clid = &setclientid_confirm->sc_clientid;
813 	int status;
814 
815 	if (STALE_CLIENTID(clid))
816 		return nfserr_stale_clientid;
817 	/*
818 	 * XXX The Duplicate Request Cache (DRC) has been checked (??)
819 	 * We get here on a DRC miss.
820 	 */
821 
822 	nfs4_lock_state();
823 	clp = find_confirmed_client(clid);
824 	if (clp) {
825 		status = nfserr_inval;
826 		/*
827 		 * Found a record for this clientid. If the IP addresses
828 		 * don't match, return ERR_INVAL just as if the record had
829 		 * not been found.
830 		 */
831 		if (clp->cl_addr != ip_addr) {
832 			printk("NFSD: setclientid: string in use by client"
833 			"(clientid %08x/%08x)\n",
834 			clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
835 			goto out;
836 		}
837 		conf = clp;
838 	}
839 	clp = find_unconfirmed_client(clid);
840 	if (clp) {
841 		status = nfserr_inval;
842 		if (clp->cl_addr != ip_addr) {
843 			printk("NFSD: setclientid: string in use by client"
844 			"(clientid %08x/%08x)\n",
845 			clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
846 			goto out;
847 		}
848 		unconf = clp;
849 	}
850 	/* CASE 1:
851 	* unconf record that matches input clientid and input confirm.
852 	* conf record that matches input clientid.
853 	* conf  and unconf records match names, verifiers
854 	*/
855 	if ((conf && unconf) &&
856 	    (cmp_verf(&unconf->cl_confirm, &confirm)) &&
857 	    (cmp_verf(&conf->cl_verifier, &unconf->cl_verifier)) &&
858 	    (cmp_name(&conf->cl_name,&unconf->cl_name))  &&
859 	    (!cmp_verf(&conf->cl_confirm, &unconf->cl_confirm))) {
860 		if (!cmp_creds(&conf->cl_cred, &unconf->cl_cred))
861 			status = nfserr_clid_inuse;
862 		else {
863 			expire_client(conf);
864 			clp = unconf;
865 			move_to_confirmed(unconf);
866 			status = nfs_ok;
867 		}
868 		goto out;
869 	}
870 	/* CASE 2:
871 	 * conf record that matches input clientid.
872 	 * if unconf record that matches input clientid, then unconf->cl_name
873 	 * or unconf->cl_verifier don't match the conf record.
874 	 */
875 	if ((conf && !unconf) ||
876 	    ((conf && unconf) &&
877 	     (!cmp_verf(&conf->cl_verifier, &unconf->cl_verifier) ||
878 	      !cmp_name(&conf->cl_name, &unconf->cl_name)))) {
879 		if (!cmp_creds(&conf->cl_cred,&rqstp->rq_cred)) {
880 			status = nfserr_clid_inuse;
881 		} else {
882 			clp = conf;
883 			status = nfs_ok;
884 		}
885 		goto out;
886 	}
887 	/* CASE 3:
888 	 * conf record not found.
889 	 * unconf record found.
890 	 * unconf->cl_confirm matches input confirm
891 	 */
892 	if (!conf && unconf && cmp_verf(&unconf->cl_confirm, &confirm)) {
893 		if (!cmp_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
894 			status = nfserr_clid_inuse;
895 		} else {
896 			status = nfs_ok;
897 			clp = unconf;
898 			move_to_confirmed(unconf);
899 		}
900 		goto out;
901 	}
902 	/* CASE 4:
903 	 * conf record not found, or if conf, then conf->cl_confirm does not
904 	 * match input confirm.
905 	 * unconf record not found, or if unconf, then unconf->cl_confirm
906 	 * does not match input confirm.
907 	 */
908 	if ((!conf || (conf && !cmp_verf(&conf->cl_confirm, &confirm))) &&
909 	    (!unconf || (unconf && !cmp_verf(&unconf->cl_confirm, &confirm)))) {
910 		status = nfserr_stale_clientid;
911 		goto out;
912 	}
913 	/* check that we have hit one of the cases...*/
914 	status = nfserr_inval;
915 	goto out;
916 out:
917 	if (!status)
918 		nfsd4_probe_callback(clp);
919 	nfs4_unlock_state();
920 	return status;
921 }
922 
923 /*
924  * Open owner state (share locks)
925  */
926 
927 /* hash tables for nfs4_stateowner */
928 #define OWNER_HASH_BITS              8
929 #define OWNER_HASH_SIZE             (1 << OWNER_HASH_BITS)
930 #define OWNER_HASH_MASK             (OWNER_HASH_SIZE - 1)
931 
932 #define ownerid_hashval(id) \
933         ((id) & OWNER_HASH_MASK)
934 #define ownerstr_hashval(clientid, ownername) \
935         (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
936 
937 static struct list_head	ownerid_hashtbl[OWNER_HASH_SIZE];
938 static struct list_head	ownerstr_hashtbl[OWNER_HASH_SIZE];
939 
940 /* hash table for nfs4_file */
941 #define FILE_HASH_BITS                   8
942 #define FILE_HASH_SIZE                  (1 << FILE_HASH_BITS)
943 #define FILE_HASH_MASK                  (FILE_HASH_SIZE - 1)
944 /* hash table for (open)nfs4_stateid */
945 #define STATEID_HASH_BITS              10
946 #define STATEID_HASH_SIZE              (1 << STATEID_HASH_BITS)
947 #define STATEID_HASH_MASK              (STATEID_HASH_SIZE - 1)
948 
949 #define file_hashval(x) \
950         hash_ptr(x, FILE_HASH_BITS)
951 #define stateid_hashval(owner_id, file_id)  \
952         (((owner_id) + (file_id)) & STATEID_HASH_MASK)
953 
954 static struct list_head file_hashtbl[FILE_HASH_SIZE];
955 static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
956 
957 /* OPEN Share state helper functions */
958 static inline struct nfs4_file *
959 alloc_init_file(struct inode *ino)
960 {
961 	struct nfs4_file *fp;
962 	unsigned int hashval = file_hashval(ino);
963 
964 	if ((fp = kmalloc(sizeof(struct nfs4_file),GFP_KERNEL))) {
965 		INIT_LIST_HEAD(&fp->fi_hash);
966 		INIT_LIST_HEAD(&fp->fi_perfile);
967 		INIT_LIST_HEAD(&fp->fi_del_perfile);
968 		list_add(&fp->fi_hash, &file_hashtbl[hashval]);
969 		fp->fi_inode = igrab(ino);
970 		fp->fi_id = current_fileid++;
971 		alloc_file++;
972 		return fp;
973 	}
974 	return NULL;
975 }
976 
977 static void
978 release_all_files(void)
979 {
980 	int i;
981 	struct nfs4_file *fp;
982 
983 	for (i=0;i<FILE_HASH_SIZE;i++) {
984 		while (!list_empty(&file_hashtbl[i])) {
985 			fp = list_entry(file_hashtbl[i].next, struct nfs4_file, fi_hash);
986 			/* this should never be more than once... */
987 			if (!list_empty(&fp->fi_perfile) || !list_empty(&fp->fi_del_perfile)) {
988 				printk("ERROR: release_all_files: file %p is open, creating dangling state !!!\n",fp);
989 			}
990 			release_file(fp);
991 		}
992 	}
993 }
994 
995 kmem_cache_t *stateowner_slab = NULL;
996 
997 static int
998 nfsd4_init_slabs(void)
999 {
1000 	stateowner_slab = kmem_cache_create("nfsd4_stateowners",
1001 			sizeof(struct nfs4_stateowner), 0, 0, NULL, NULL);
1002 	if (stateowner_slab == NULL) {
1003 		dprintk("nfsd4: out of memory while initializing nfsv4\n");
1004 		return -ENOMEM;
1005 	}
1006 	return 0;
1007 }
1008 
1009 static void
1010 nfsd4_free_slabs(void)
1011 {
1012 	int status = 0;
1013 
1014 	if (stateowner_slab)
1015 		status = kmem_cache_destroy(stateowner_slab);
1016 	stateowner_slab = NULL;
1017 	BUG_ON(status);
1018 }
1019 
1020 void
1021 nfs4_free_stateowner(struct kref *kref)
1022 {
1023 	struct nfs4_stateowner *sop =
1024 		container_of(kref, struct nfs4_stateowner, so_ref);
1025 	kfree(sop->so_owner.data);
1026 	kmem_cache_free(stateowner_slab, sop);
1027 }
1028 
1029 static inline struct nfs4_stateowner *
1030 alloc_stateowner(struct xdr_netobj *owner)
1031 {
1032 	struct nfs4_stateowner *sop;
1033 
1034 	if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1035 		if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1036 			memcpy(sop->so_owner.data, owner->data, owner->len);
1037 			sop->so_owner.len = owner->len;
1038 			kref_init(&sop->so_ref);
1039 			return sop;
1040 		}
1041 		kmem_cache_free(stateowner_slab, sop);
1042 	}
1043 	return NULL;
1044 }
1045 
1046 static struct nfs4_stateowner *
1047 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1048 	struct nfs4_stateowner *sop;
1049 	struct nfs4_replay *rp;
1050 	unsigned int idhashval;
1051 
1052 	if (!(sop = alloc_stateowner(&open->op_owner)))
1053 		return NULL;
1054 	idhashval = ownerid_hashval(current_ownerid);
1055 	INIT_LIST_HEAD(&sop->so_idhash);
1056 	INIT_LIST_HEAD(&sop->so_strhash);
1057 	INIT_LIST_HEAD(&sop->so_perclient);
1058 	INIT_LIST_HEAD(&sop->so_perfilestate);
1059 	INIT_LIST_HEAD(&sop->so_perlockowner);  /* not used */
1060 	INIT_LIST_HEAD(&sop->so_close_lru);
1061 	sop->so_time = 0;
1062 	list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1063 	list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
1064 	list_add(&sop->so_perclient, &clp->cl_perclient);
1065 	add_perclient++;
1066 	sop->so_is_open_owner = 1;
1067 	sop->so_id = current_ownerid++;
1068 	sop->so_client = clp;
1069 	sop->so_seqid = open->op_seqid;
1070 	sop->so_confirmed = 0;
1071 	rp = &sop->so_replay;
1072 	rp->rp_status = NFSERR_SERVERFAULT;
1073 	rp->rp_buflen = 0;
1074 	rp->rp_buf = rp->rp_ibuf;
1075 	return sop;
1076 }
1077 
1078 static void
1079 release_stateid_lockowners(struct nfs4_stateid *open_stp)
1080 {
1081 	struct nfs4_stateowner *lock_sop;
1082 
1083 	while (!list_empty(&open_stp->st_perlockowner)) {
1084 		lock_sop = list_entry(open_stp->st_perlockowner.next,
1085 				struct nfs4_stateowner, so_perlockowner);
1086 		/* list_del(&open_stp->st_perlockowner);  */
1087 		BUG_ON(lock_sop->so_is_open_owner);
1088 		release_stateowner(lock_sop);
1089 	}
1090 }
1091 
1092 static void
1093 unhash_stateowner(struct nfs4_stateowner *sop)
1094 {
1095 	struct nfs4_stateid *stp;
1096 
1097 	list_del(&sop->so_idhash);
1098 	list_del(&sop->so_strhash);
1099 	if (sop->so_is_open_owner) {
1100 		list_del(&sop->so_perclient);
1101 		del_perclient++;
1102 	}
1103 	list_del(&sop->so_perlockowner);
1104 	while (!list_empty(&sop->so_perfilestate)) {
1105 		stp = list_entry(sop->so_perfilestate.next,
1106 			struct nfs4_stateid, st_perfilestate);
1107 		if (sop->so_is_open_owner)
1108 			release_stateid(stp, OPEN_STATE);
1109 		else
1110 			release_stateid(stp, LOCK_STATE);
1111 	}
1112 }
1113 
1114 static void
1115 release_stateowner(struct nfs4_stateowner *sop)
1116 {
1117 	unhash_stateowner(sop);
1118 	list_del(&sop->so_close_lru);
1119 	nfs4_put_stateowner(sop);
1120 }
1121 
1122 static inline void
1123 init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1124 	struct nfs4_stateowner *sop = open->op_stateowner;
1125 	unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1126 
1127 	INIT_LIST_HEAD(&stp->st_hash);
1128 	INIT_LIST_HEAD(&stp->st_perfilestate);
1129 	INIT_LIST_HEAD(&stp->st_perlockowner);
1130 	INIT_LIST_HEAD(&stp->st_perfile);
1131 	list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
1132 	list_add(&stp->st_perfilestate, &sop->so_perfilestate);
1133 	list_add_perfile++;
1134 	list_add(&stp->st_perfile, &fp->fi_perfile);
1135 	stp->st_stateowner = sop;
1136 	stp->st_file = fp;
1137 	stp->st_stateid.si_boot = boot_time;
1138 	stp->st_stateid.si_stateownerid = sop->so_id;
1139 	stp->st_stateid.si_fileid = fp->fi_id;
1140 	stp->st_stateid.si_generation = 0;
1141 	stp->st_access_bmap = 0;
1142 	stp->st_deny_bmap = 0;
1143 	__set_bit(open->op_share_access, &stp->st_access_bmap);
1144 	__set_bit(open->op_share_deny, &stp->st_deny_bmap);
1145 }
1146 
1147 static void
1148 release_stateid(struct nfs4_stateid *stp, int flags)
1149 {
1150 	struct file *filp = stp->st_vfs_file;
1151 
1152 	list_del(&stp->st_hash);
1153 	list_del_perfile++;
1154 	list_del(&stp->st_perfile);
1155 	list_del(&stp->st_perfilestate);
1156 	if (flags & OPEN_STATE) {
1157 		release_stateid_lockowners(stp);
1158 		stp->st_vfs_file = NULL;
1159 		nfsd_close(filp);
1160 		vfsclose++;
1161 	} else if (flags & LOCK_STATE)
1162 		locks_remove_posix(filp, (fl_owner_t) stp->st_stateowner);
1163 	kfree(stp);
1164 	stp = NULL;
1165 }
1166 
1167 static void
1168 release_file(struct nfs4_file *fp)
1169 {
1170 	free_file++;
1171 	list_del(&fp->fi_hash);
1172 	iput(fp->fi_inode);
1173 	kfree(fp);
1174 }
1175 
1176 void
1177 move_to_close_lru(struct nfs4_stateowner *sop)
1178 {
1179 	dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1180 
1181 	unhash_stateowner(sop);
1182 	list_add_tail(&sop->so_close_lru, &close_lru);
1183 	sop->so_time = get_seconds();
1184 }
1185 
1186 void
1187 release_state_owner(struct nfs4_stateid *stp, int flag)
1188 {
1189 	struct nfs4_stateowner *sop = stp->st_stateowner;
1190 	struct nfs4_file *fp = stp->st_file;
1191 
1192 	dprintk("NFSD: release_state_owner\n");
1193 	release_stateid(stp, flag);
1194 
1195 	/* place unused nfs4_stateowners on so_close_lru list to be
1196 	 * released by the laundromat service after the lease period
1197 	 * to enable us to handle CLOSE replay
1198 	 */
1199 	if (sop->so_confirmed && list_empty(&sop->so_perfilestate))
1200 		move_to_close_lru(sop);
1201 	/* unused nfs4_file's are releseed. XXX slab cache? */
1202 	if (list_empty(&fp->fi_perfile) && list_empty(&fp->fi_del_perfile)) {
1203 		release_file(fp);
1204 	}
1205 }
1206 
1207 static int
1208 cmp_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner, clientid_t *clid) {
1209 	return ((sop->so_owner.len == owner->len) &&
1210 	 !memcmp(sop->so_owner.data, owner->data, owner->len) &&
1211 	  (sop->so_client->cl_clientid.cl_id == clid->cl_id));
1212 }
1213 
1214 static struct nfs4_stateowner *
1215 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1216 {
1217 	struct nfs4_stateowner *so = NULL;
1218 
1219 	list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
1220 		if (cmp_owner_str(so, &open->op_owner, &open->op_clientid))
1221 			return so;
1222 	}
1223 	return NULL;
1224 }
1225 
1226 /* search file_hashtbl[] for file */
1227 static struct nfs4_file *
1228 find_file(struct inode *ino)
1229 {
1230 	unsigned int hashval = file_hashval(ino);
1231 	struct nfs4_file *fp;
1232 
1233 	list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
1234 		if (fp->fi_inode == ino)
1235 			return fp;
1236 	}
1237 	return NULL;
1238 }
1239 
1240 #define TEST_ACCESS(x) ((x > 0 || x < 4)?1:0)
1241 #define TEST_DENY(x) ((x >= 0 || x < 5)?1:0)
1242 
1243 void
1244 set_access(unsigned int *access, unsigned long bmap) {
1245 	int i;
1246 
1247 	*access = 0;
1248 	for (i = 1; i < 4; i++) {
1249 		if (test_bit(i, &bmap))
1250 			*access |= i;
1251 	}
1252 }
1253 
1254 void
1255 set_deny(unsigned int *deny, unsigned long bmap) {
1256 	int i;
1257 
1258 	*deny = 0;
1259 	for (i = 0; i < 4; i++) {
1260 		if (test_bit(i, &bmap))
1261 			*deny |= i ;
1262 	}
1263 }
1264 
1265 static int
1266 test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1267 	unsigned int access, deny;
1268 
1269 	set_access(&access, stp->st_access_bmap);
1270 	set_deny(&deny, stp->st_deny_bmap);
1271 	if ((access & open->op_share_deny) || (deny & open->op_share_access))
1272 		return 0;
1273 	return 1;
1274 }
1275 
1276 /*
1277  * Called to check deny when READ with all zero stateid or
1278  * WRITE with all zero or all one stateid
1279  */
1280 int
1281 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
1282 {
1283 	struct inode *ino = current_fh->fh_dentry->d_inode;
1284 	struct nfs4_file *fp;
1285 	struct nfs4_stateid *stp;
1286 
1287 	dprintk("NFSD: nfs4_share_conflict\n");
1288 
1289 	fp = find_file(ino);
1290 	if (fp) {
1291 	/* Search for conflicting share reservations */
1292 		list_for_each_entry(stp, &fp->fi_perfile, st_perfile) {
1293 			if (test_bit(deny_type, &stp->st_deny_bmap) ||
1294 			    test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
1295 				return nfserr_share_denied;
1296 		}
1297 	}
1298 	return nfs_ok;
1299 }
1300 
1301 static inline void
1302 nfs4_file_downgrade(struct file *filp, unsigned int share_access)
1303 {
1304 	if (share_access & NFS4_SHARE_ACCESS_WRITE) {
1305 		put_write_access(filp->f_dentry->d_inode);
1306 		filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
1307 	}
1308 }
1309 
1310 /*
1311  * Recall a delegation
1312  */
1313 static int
1314 do_recall(void *__dp)
1315 {
1316 	struct nfs4_delegation *dp = __dp;
1317 
1318 	daemonize("nfsv4-recall");
1319 
1320 	nfsd4_cb_recall(dp);
1321 	return 0;
1322 }
1323 
1324 /*
1325  * Spawn a thread to perform a recall on the delegation represented
1326  * by the lease (file_lock)
1327  *
1328  * Called from break_lease() with lock_kernel() held.
1329  * Note: we assume break_lease will only call this *once* for any given
1330  * lease.
1331  */
1332 static
1333 void nfsd_break_deleg_cb(struct file_lock *fl)
1334 {
1335 	struct nfs4_delegation *dp=  (struct nfs4_delegation *)fl->fl_owner;
1336 	struct task_struct *t;
1337 
1338 	dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
1339 	if (!dp)
1340 		return;
1341 
1342 	/* We're assuming the state code never drops its reference
1343 	 * without first removing the lease.  Since we're in this lease
1344 	 * callback (and since the lease code is serialized by the kernel
1345 	 * lock) we know the server hasn't removed the lease yet, we know
1346 	 * it's safe to take a reference: */
1347 	atomic_inc(&dp->dl_count);
1348 
1349 	spin_lock(&recall_lock);
1350 	list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
1351 	spin_unlock(&recall_lock);
1352 
1353 	/* only place dl_time is set. protected by lock_kernel*/
1354 	dp->dl_time = get_seconds();
1355 
1356 	/* XXX need to merge NFSD_LEASE_TIME with fs/locks.c:lease_break_time */
1357 	fl->fl_break_time = jiffies + NFSD_LEASE_TIME * HZ;
1358 
1359 	t = kthread_run(do_recall, dp, "%s", "nfs4_cb_recall");
1360 	if (IS_ERR(t)) {
1361 		struct nfs4_client *clp = dp->dl_client;
1362 
1363 		printk(KERN_INFO "NFSD: Callback thread failed for "
1364 			"for client (clientid %08x/%08x)\n",
1365 			clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1366 		nfs4_put_delegation(dp);
1367 	}
1368 }
1369 
1370 /*
1371  * The file_lock is being reapd.
1372  *
1373  * Called by locks_free_lock() with lock_kernel() held.
1374  */
1375 static
1376 void nfsd_release_deleg_cb(struct file_lock *fl)
1377 {
1378 	struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1379 
1380 	dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
1381 
1382 	if (!(fl->fl_flags & FL_LEASE) || !dp)
1383 		return;
1384 	dp->dl_flock = NULL;
1385 }
1386 
1387 /*
1388  * Set the delegation file_lock back pointer.
1389  *
1390  * Called from __setlease() with lock_kernel() held.
1391  */
1392 static
1393 void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
1394 {
1395 	struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
1396 
1397 	dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
1398 	if (!dp)
1399 		return;
1400 	dp->dl_flock = new;
1401 }
1402 
1403 /*
1404  * Called from __setlease() with lock_kernel() held
1405  */
1406 static
1407 int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
1408 {
1409 	struct nfs4_delegation *onlistd =
1410 		(struct nfs4_delegation *)onlist->fl_owner;
1411 	struct nfs4_delegation *tryd =
1412 		(struct nfs4_delegation *)try->fl_owner;
1413 
1414 	if (onlist->fl_lmops != try->fl_lmops)
1415 		return 0;
1416 
1417 	return onlistd->dl_client == tryd->dl_client;
1418 }
1419 
1420 
1421 static
1422 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
1423 {
1424 	if (arg & F_UNLCK)
1425 		return lease_modify(onlist, arg);
1426 	else
1427 		return -EAGAIN;
1428 }
1429 
1430 struct lock_manager_operations nfsd_lease_mng_ops = {
1431 	.fl_break = nfsd_break_deleg_cb,
1432 	.fl_release_private = nfsd_release_deleg_cb,
1433 	.fl_copy_lock = nfsd_copy_lock_deleg_cb,
1434 	.fl_mylease = nfsd_same_client_deleg_cb,
1435 	.fl_change = nfsd_change_deleg_cb,
1436 };
1437 
1438 
1439 /*
1440  * nfsd4_process_open1()
1441  * 	lookup stateowner.
1442  * 		found:
1443  * 			check confirmed
1444  * 				confirmed:
1445  * 					check seqid
1446  * 				not confirmed:
1447  * 					delete owner
1448  * 					create new owner
1449  * 		notfound:
1450  * 			verify clientid
1451  * 			create new owner
1452  *
1453  * called with nfs4_lock_state() held.
1454  */
1455 int
1456 nfsd4_process_open1(struct nfsd4_open *open)
1457 {
1458 	int status;
1459 	clientid_t *clientid = &open->op_clientid;
1460 	struct nfs4_client *clp = NULL;
1461 	unsigned int strhashval;
1462 	struct nfs4_stateowner *sop = NULL;
1463 
1464 	status = nfserr_inval;
1465 	if (!check_name(open->op_owner))
1466 		goto out;
1467 
1468 	if (STALE_CLIENTID(&open->op_clientid))
1469 		return nfserr_stale_clientid;
1470 
1471 	strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
1472 	sop = find_openstateowner_str(strhashval, open);
1473 	if (sop) {
1474 		open->op_stateowner = sop;
1475 		/* check for replay */
1476 		if (open->op_seqid == sop->so_seqid){
1477 			if (sop->so_replay.rp_buflen)
1478 				return NFSERR_REPLAY_ME;
1479 			else {
1480 				/* The original OPEN failed so spectacularly
1481 				 * that we don't even have replay data saved!
1482 				 * Therefore, we have no choice but to continue
1483 				 * processing this OPEN; presumably, we'll
1484 				 * fail again for the same reason.
1485 				 */
1486 				dprintk("nfsd4_process_open1:"
1487 					" replay with no replay cache\n");
1488 				goto renew;
1489 			}
1490 		} else if (sop->so_confirmed) {
1491 			if (open->op_seqid == sop->so_seqid + 1)
1492 				goto renew;
1493 			status = nfserr_bad_seqid;
1494 			goto out;
1495 		} else {
1496 			/* If we get here, we received an OPEN for an
1497 			 * unconfirmed nfs4_stateowner. Since the seqid's are
1498 			 * different, purge the existing nfs4_stateowner, and
1499 			 * instantiate a new one.
1500 			 */
1501 			clp = sop->so_client;
1502 			release_stateowner(sop);
1503 		}
1504 	} else {
1505 		/* nfs4_stateowner not found.
1506 		 * Verify clientid and instantiate new nfs4_stateowner.
1507 		 * If verify fails this is presumably the result of the
1508 		 * client's lease expiring.
1509 		 */
1510 		status = nfserr_expired;
1511 		clp = find_confirmed_client(clientid);
1512 		if (clp == NULL)
1513 			goto out;
1514 	}
1515 	status = nfserr_resource;
1516 	sop = alloc_init_open_stateowner(strhashval, clp, open);
1517 	if (sop == NULL)
1518 		goto out;
1519 	open->op_stateowner = sop;
1520 renew:
1521 	status = nfs_ok;
1522 	renew_client(sop->so_client);
1523 out:
1524 	if (status && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
1525 		status = nfserr_reclaim_bad;
1526 	return status;
1527 }
1528 
1529 static int
1530 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
1531 {
1532 	struct nfs4_stateid *local;
1533 	int status = nfserr_share_denied;
1534 	struct nfs4_stateowner *sop = open->op_stateowner;
1535 
1536 	list_for_each_entry(local, &fp->fi_perfile, st_perfile) {
1537 		/* ignore lock owners */
1538 		if (local->st_stateowner->so_is_open_owner == 0)
1539 			continue;
1540 		/* remember if we have seen this open owner */
1541 		if (local->st_stateowner == sop)
1542 			*stpp = local;
1543 		/* check for conflicting share reservations */
1544 		if (!test_share(local, open))
1545 			goto out;
1546 	}
1547 	status = 0;
1548 out:
1549 	return status;
1550 }
1551 
1552 static int
1553 nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
1554 		struct svc_fh *cur_fh, int flags)
1555 {
1556 	struct nfs4_stateid *stp;
1557 	int status;
1558 
1559 	stp = kmalloc(sizeof(struct nfs4_stateid), GFP_KERNEL);
1560 	if (stp == NULL)
1561 		return nfserr_resource;
1562 
1563 	status = nfsd_open(rqstp, cur_fh, S_IFREG, flags, &stp->st_vfs_file);
1564 	if (status) {
1565 		if (status == nfserr_dropit)
1566 			status = nfserr_jukebox;
1567 		kfree(stp);
1568 		return status;
1569 	}
1570 	vfsopen++;
1571 	*stpp = stp;
1572 	return 0;
1573 }
1574 
1575 static inline int
1576 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
1577 		struct nfsd4_open *open)
1578 {
1579 	struct iattr iattr = {
1580 		.ia_valid = ATTR_SIZE,
1581 		.ia_size = 0,
1582 	};
1583 	if (!open->op_truncate)
1584 		return 0;
1585 	if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
1586 		return -EINVAL;
1587 	return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
1588 }
1589 
1590 static int
1591 nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
1592 {
1593 	struct file *filp = stp->st_vfs_file;
1594 	struct inode *inode = filp->f_dentry->d_inode;
1595 	unsigned int share_access;
1596 	int status;
1597 
1598 	set_access(&share_access, stp->st_access_bmap);
1599 	share_access = ~share_access;
1600 	share_access &= open->op_share_access;
1601 
1602 	if (!(share_access & NFS4_SHARE_ACCESS_WRITE))
1603 		return nfsd4_truncate(rqstp, cur_fh, open);
1604 
1605 	status = get_write_access(inode);
1606 	if (status)
1607 		return nfserrno(status);
1608 	status = nfsd4_truncate(rqstp, cur_fh, open);
1609 	if (status) {
1610 		put_write_access(inode);
1611 		return status;
1612 	}
1613 	/* remember the open */
1614 	filp->f_mode = (filp->f_mode | FMODE_WRITE) & ~FMODE_READ;
1615 	set_bit(open->op_share_access, &stp->st_access_bmap);
1616 	set_bit(open->op_share_deny, &stp->st_deny_bmap);
1617 
1618 	return nfs_ok;
1619 }
1620 
1621 
1622 /* decrement seqid on successful reclaim, it will be bumped in encode_open */
1623 static void
1624 nfs4_set_claim_prev(struct nfsd4_open *open, int *status)
1625 {
1626 	if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS) {
1627 		if (*status)
1628 			*status = nfserr_reclaim_bad;
1629 		else {
1630 			open->op_stateowner->so_confirmed = 1;
1631 			open->op_stateowner->so_seqid--;
1632 		}
1633 	}
1634 }
1635 
1636 /*
1637  * Attempt to hand out a delegation.
1638  */
1639 static void
1640 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
1641 {
1642 	struct nfs4_delegation *dp;
1643 	struct nfs4_stateowner *sop = stp->st_stateowner;
1644 	struct nfs4_callback *cb = &sop->so_client->cl_callback;
1645 	struct file_lock fl, *flp = &fl;
1646 	int status, flag = 0;
1647 
1648 	flag = NFS4_OPEN_DELEGATE_NONE;
1649 	if (open->op_claim_type != NFS4_OPEN_CLAIM_NULL
1650 	     || !atomic_read(&cb->cb_set) || !sop->so_confirmed)
1651 		goto out;
1652 
1653 	if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
1654 		flag = NFS4_OPEN_DELEGATE_WRITE;
1655 	else
1656 		flag = NFS4_OPEN_DELEGATE_READ;
1657 
1658 	dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
1659 	if (dp == NULL) {
1660 		flag = NFS4_OPEN_DELEGATE_NONE;
1661 		goto out;
1662 	}
1663 	locks_init_lock(&fl);
1664 	fl.fl_lmops = &nfsd_lease_mng_ops;
1665 	fl.fl_flags = FL_LEASE;
1666 	fl.fl_end = OFFSET_MAX;
1667 	fl.fl_owner =  (fl_owner_t)dp;
1668 	fl.fl_file = stp->st_vfs_file;
1669 	fl.fl_pid = current->tgid;
1670 
1671 	/* setlease checks to see if delegation should be handed out.
1672 	 * the lock_manager callbacks fl_mylease and fl_change are used
1673 	 */
1674 	if ((status = setlease(stp->st_vfs_file,
1675 		flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK, &flp))) {
1676 		dprintk("NFSD: setlease failed [%d], no delegation\n", status);
1677 		unhash_delegation(dp);
1678 		flag = NFS4_OPEN_DELEGATE_NONE;
1679 		goto out;
1680 	}
1681 
1682 	memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
1683 
1684 	dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
1685 	             dp->dl_stateid.si_boot,
1686 	             dp->dl_stateid.si_stateownerid,
1687 	             dp->dl_stateid.si_fileid,
1688 	             dp->dl_stateid.si_generation);
1689 out:
1690 	open->op_delegate_type = flag;
1691 }
1692 
1693 /*
1694  * called with nfs4_lock_state() held.
1695  */
1696 int
1697 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
1698 {
1699 	struct nfs4_file *fp = NULL;
1700 	struct inode *ino = current_fh->fh_dentry->d_inode;
1701 	struct nfs4_stateid *stp = NULL;
1702 	int status;
1703 
1704 	status = nfserr_inval;
1705 	if (!TEST_ACCESS(open->op_share_access) || !TEST_DENY(open->op_share_deny))
1706 		goto out;
1707 	/*
1708 	 * Lookup file; if found, lookup stateid and check open request,
1709 	 * and check for delegations in the process of being recalled.
1710 	 * If not found, create the nfs4_file struct
1711 	 */
1712 	fp = find_file(ino);
1713 	if (fp) {
1714 		if ((status = nfs4_check_open(fp, open, &stp)))
1715 			goto out;
1716 	} else {
1717 		status = nfserr_resource;
1718 		fp = alloc_init_file(ino);
1719 		if (fp == NULL)
1720 			goto out;
1721 	}
1722 
1723 	/*
1724 	 * OPEN the file, or upgrade an existing OPEN.
1725 	 * If truncate fails, the OPEN fails.
1726 	 */
1727 	if (stp) {
1728 		/* Stateid was found, this is an OPEN upgrade */
1729 		status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
1730 		if (status)
1731 			goto out;
1732 	} else {
1733 		/* Stateid was not found, this is a new OPEN */
1734 		int flags = 0;
1735 		if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
1736 			flags = MAY_WRITE;
1737 		else
1738 			flags = MAY_READ;
1739 		if ((status = nfs4_new_open(rqstp, &stp, current_fh, flags)))
1740 			goto out;
1741 		init_stateid(stp, fp, open);
1742 		status = nfsd4_truncate(rqstp, current_fh, open);
1743 		if (status) {
1744 			release_stateid(stp, OPEN_STATE);
1745 			goto out;
1746 		}
1747 	}
1748 	memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
1749 
1750 	/*
1751 	* Attempt to hand out a delegation. No error return, because the
1752 	* OPEN succeeds even if we fail.
1753 	*/
1754 	nfs4_open_delegation(current_fh, open, stp);
1755 
1756 	status = nfs_ok;
1757 
1758 	dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
1759 	            stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
1760 	            stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
1761 out:
1762 	/* take the opportunity to clean up unused state */
1763 	if (fp && list_empty(&fp->fi_perfile) && list_empty(&fp->fi_del_perfile))
1764 		release_file(fp);
1765 
1766 	/* CLAIM_PREVIOUS has different error returns */
1767 	nfs4_set_claim_prev(open, &status);
1768 	/*
1769 	* To finish the open response, we just need to set the rflags.
1770 	*/
1771 	open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
1772 	if (!open->op_stateowner->so_confirmed)
1773 		open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
1774 
1775 	return status;
1776 }
1777 
1778 static struct work_struct laundromat_work;
1779 static void laundromat_main(void *);
1780 static DECLARE_WORK(laundromat_work, laundromat_main, NULL);
1781 
1782 int
1783 nfsd4_renew(clientid_t *clid)
1784 {
1785 	struct nfs4_client *clp;
1786 	int status;
1787 
1788 	nfs4_lock_state();
1789 	dprintk("process_renew(%08x/%08x): starting\n",
1790 			clid->cl_boot, clid->cl_id);
1791 	status = nfserr_stale_clientid;
1792 	if (STALE_CLIENTID(clid))
1793 		goto out;
1794 	clp = find_confirmed_client(clid);
1795 	status = nfserr_expired;
1796 	if (clp == NULL) {
1797 		/* We assume the client took too long to RENEW. */
1798 		dprintk("nfsd4_renew: clientid not found!\n");
1799 		goto out;
1800 	}
1801 	renew_client(clp);
1802 	status = nfserr_cb_path_down;
1803 	if (!list_empty(&clp->cl_del_perclnt)
1804 			&& !atomic_read(&clp->cl_callback.cb_set))
1805 		goto out;
1806 	status = nfs_ok;
1807 out:
1808 	nfs4_unlock_state();
1809 	return status;
1810 }
1811 
1812 time_t
1813 nfs4_laundromat(void)
1814 {
1815 	struct nfs4_client *clp;
1816 	struct nfs4_stateowner *sop;
1817 	struct nfs4_delegation *dp;
1818 	struct list_head *pos, *next, reaplist;
1819 	time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
1820 	time_t t, clientid_val = NFSD_LEASE_TIME;
1821 	time_t u, test_val = NFSD_LEASE_TIME;
1822 
1823 	nfs4_lock_state();
1824 
1825 	dprintk("NFSD: laundromat service - starting\n");
1826 	list_for_each_safe(pos, next, &client_lru) {
1827 		clp = list_entry(pos, struct nfs4_client, cl_lru);
1828 		if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
1829 			t = clp->cl_time - cutoff;
1830 			if (clientid_val > t)
1831 				clientid_val = t;
1832 			break;
1833 		}
1834 		dprintk("NFSD: purging unused client (clientid %08x)\n",
1835 			clp->cl_clientid.cl_id);
1836 		expire_client(clp);
1837 	}
1838 	INIT_LIST_HEAD(&reaplist);
1839 	spin_lock(&recall_lock);
1840 	list_for_each_safe(pos, next, &del_recall_lru) {
1841 		dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
1842 		if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
1843 			u = dp->dl_time - cutoff;
1844 			if (test_val > u)
1845 				test_val = u;
1846 			break;
1847 		}
1848 		dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
1849 			            dp, dp->dl_flock);
1850 		list_move(&dp->dl_recall_lru, &reaplist);
1851 	}
1852 	spin_unlock(&recall_lock);
1853 	list_for_each_safe(pos, next, &reaplist) {
1854 		dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
1855 		list_del_init(&dp->dl_recall_lru);
1856 		unhash_delegation(dp);
1857 	}
1858 	test_val = NFSD_LEASE_TIME;
1859 	list_for_each_safe(pos, next, &close_lru) {
1860 		sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
1861 		if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
1862 			u = sop->so_time - cutoff;
1863 			if (test_val > u)
1864 				test_val = u;
1865 			break;
1866 		}
1867 		dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
1868 			sop->so_id);
1869 		list_del(&sop->so_close_lru);
1870 		nfs4_put_stateowner(sop);
1871 	}
1872 	if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
1873 		clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
1874 	nfs4_unlock_state();
1875 	return clientid_val;
1876 }
1877 
1878 void
1879 laundromat_main(void *not_used)
1880 {
1881 	time_t t;
1882 
1883 	t = nfs4_laundromat();
1884 	dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
1885 	schedule_delayed_work(&laundromat_work, t*HZ);
1886 }
1887 
1888 /* search ownerid_hashtbl[] and close_lru for stateid owner
1889  * (stateid->si_stateownerid)
1890  */
1891 struct nfs4_stateowner *
1892 find_openstateowner_id(u32 st_id, int flags) {
1893 	struct nfs4_stateowner *local = NULL;
1894 
1895 	dprintk("NFSD: find_openstateowner_id %d\n", st_id);
1896 	if (flags & CLOSE_STATE) {
1897 		list_for_each_entry(local, &close_lru, so_close_lru) {
1898 			if (local->so_id == st_id)
1899 				return local;
1900 		}
1901 	}
1902 	return NULL;
1903 }
1904 
1905 static inline int
1906 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
1907 {
1908 	return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_dentry->d_inode;
1909 }
1910 
1911 static int
1912 STALE_STATEID(stateid_t *stateid)
1913 {
1914 	if (stateid->si_boot == boot_time)
1915 		return 0;
1916 	printk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
1917 		stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
1918 		stateid->si_generation);
1919 	return 1;
1920 }
1921 
1922 static inline int
1923 access_permit_read(unsigned long access_bmap)
1924 {
1925 	return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
1926 		test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
1927 		test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
1928 }
1929 
1930 static inline int
1931 access_permit_write(unsigned long access_bmap)
1932 {
1933 	return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
1934 		test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
1935 }
1936 
1937 static
1938 int nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1939 {
1940         int status = nfserr_openmode;
1941 
1942 	if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
1943                 goto out;
1944 	if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
1945                 goto out;
1946 	status = nfs_ok;
1947 out:
1948 	return status;
1949 }
1950 
1951 static inline int
1952 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
1953 {
1954 	if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
1955 		return nfserr_openmode;
1956 	else
1957 		return nfs_ok;
1958 }
1959 
1960 static inline int
1961 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
1962 {
1963 	/* Trying to call delegreturn with a special stateid? Yuch: */
1964 	if (!(flags & (RD_STATE | WR_STATE)))
1965 		return nfserr_bad_stateid;
1966 	else if (ONE_STATEID(stateid) && (flags & RD_STATE))
1967 		return nfs_ok;
1968 	else if (nfs4_in_grace()) {
1969 		/* Answer in remaining cases depends on existance of
1970 		 * conflicting state; so we must wait out the grace period. */
1971 		return nfserr_grace;
1972 	} else if (flags & WR_STATE)
1973 		return nfs4_share_conflict(current_fh,
1974 				NFS4_SHARE_DENY_WRITE);
1975 	else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
1976 		return nfs4_share_conflict(current_fh,
1977 				NFS4_SHARE_DENY_READ);
1978 }
1979 
1980 /*
1981  * Allow READ/WRITE during grace period on recovered state only for files
1982  * that are not able to provide mandatory locking.
1983  */
1984 static inline int
1985 io_during_grace_disallowed(struct inode *inode, int flags)
1986 {
1987 	return nfs4_in_grace() && (flags & (RD_STATE | WR_STATE))
1988 		&& MANDATORY_LOCK(inode);
1989 }
1990 
1991 /*
1992 * Checks for stateid operations
1993 */
1994 int
1995 nfs4_preprocess_stateid_op(struct svc_fh *current_fh, stateid_t *stateid, int flags, struct file **filpp)
1996 {
1997 	struct nfs4_stateid *stp = NULL;
1998 	struct nfs4_delegation *dp = NULL;
1999 	stateid_t *stidp;
2000 	struct inode *ino = current_fh->fh_dentry->d_inode;
2001 	int status;
2002 
2003 	dprintk("NFSD: preprocess_stateid_op: stateid = (%08x/%08x/%08x/%08x)\n",
2004 		stateid->si_boot, stateid->si_stateownerid,
2005 		stateid->si_fileid, stateid->si_generation);
2006 	if (filpp)
2007 		*filpp = NULL;
2008 
2009 	if (io_during_grace_disallowed(ino, flags))
2010 		return nfserr_grace;
2011 
2012 	if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2013 		return check_special_stateids(current_fh, stateid, flags);
2014 
2015 	/* STALE STATEID */
2016 	status = nfserr_stale_stateid;
2017 	if (STALE_STATEID(stateid))
2018 		goto out;
2019 
2020 	/* BAD STATEID */
2021 	status = nfserr_bad_stateid;
2022 	if (!stateid->si_fileid) { /* delegation stateid */
2023 		if(!(dp = find_delegation_stateid(ino, stateid))) {
2024 			dprintk("NFSD: delegation stateid not found\n");
2025 			if (nfs4_in_grace())
2026 				status = nfserr_grace;
2027 			goto out;
2028 		}
2029 		stidp = &dp->dl_stateid;
2030 	} else { /* open or lock stateid */
2031 		if (!(stp = find_stateid(stateid, flags))) {
2032 			dprintk("NFSD: open or lock stateid not found\n");
2033 			if (nfs4_in_grace())
2034 				status = nfserr_grace;
2035 			goto out;
2036 		}
2037 		if ((flags & CHECK_FH) && nfs4_check_fh(current_fh, stp))
2038 			goto out;
2039 		if (!stp->st_stateowner->so_confirmed)
2040 			goto out;
2041 		stidp = &stp->st_stateid;
2042 	}
2043 	if (stateid->si_generation > stidp->si_generation)
2044 		goto out;
2045 
2046 	/* OLD STATEID */
2047 	status = nfserr_old_stateid;
2048 	if (stateid->si_generation < stidp->si_generation)
2049 		goto out;
2050 	if (stp) {
2051 		if ((status = nfs4_check_openmode(stp,flags)))
2052 			goto out;
2053 		renew_client(stp->st_stateowner->so_client);
2054 		if (filpp)
2055 			*filpp = stp->st_vfs_file;
2056 	} else if (dp) {
2057 		if ((status = nfs4_check_delegmode(dp, flags)))
2058 			goto out;
2059 		renew_client(dp->dl_client);
2060 		if (flags & DELEG_RET)
2061 			unhash_delegation(dp);
2062 		if (filpp)
2063 			*filpp = dp->dl_vfs_file;
2064 	}
2065 	status = nfs_ok;
2066 out:
2067 	return status;
2068 }
2069 
2070 
2071 /*
2072  * Checks for sequence id mutating operations.
2073  */
2074 int
2075 nfs4_preprocess_seqid_op(struct svc_fh *current_fh, u32 seqid, stateid_t *stateid, int flags, struct nfs4_stateowner **sopp, struct nfs4_stateid **stpp, clientid_t *lockclid)
2076 {
2077 	int status;
2078 	struct nfs4_stateid *stp;
2079 	struct nfs4_stateowner *sop;
2080 
2081 	dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2082 			"stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2083 		stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2084 		stateid->si_generation);
2085 
2086 	*stpp = NULL;
2087 	*sopp = NULL;
2088 
2089 	status = nfserr_bad_stateid;
2090 	if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2091 		printk("NFSD: preprocess_seqid_op: magic stateid!\n");
2092 		goto out;
2093 	}
2094 
2095 	status = nfserr_stale_stateid;
2096 	if (STALE_STATEID(stateid))
2097 		goto out;
2098 	/*
2099 	* We return BAD_STATEID if filehandle doesn't match stateid,
2100 	* the confirmed flag is incorrecly set, or the generation
2101 	* number is incorrect.
2102 	* If there is no entry in the openfile table for this id,
2103 	* we can't always return BAD_STATEID;
2104 	* this might be a retransmitted CLOSE which has arrived after
2105 	* the openfile has been released.
2106 	*/
2107 	if (!(stp = find_stateid(stateid, flags)))
2108 		goto no_nfs4_stateid;
2109 
2110 	status = nfserr_bad_stateid;
2111 
2112 	/* for new lock stateowners:
2113 	 * check that the lock->v.new.open_stateid
2114 	 * refers to an open stateowner
2115 	 *
2116 	 * check that the lockclid (nfs4_lock->v.new.clientid) is the same
2117 	 * as the open_stateid->st_stateowner->so_client->clientid
2118 	 */
2119 	if (lockclid) {
2120 		struct nfs4_stateowner *sop = stp->st_stateowner;
2121 		struct nfs4_client *clp = sop->so_client;
2122 
2123 		if (!sop->so_is_open_owner)
2124 			goto out;
2125 		if (!cmp_clid(&clp->cl_clientid, lockclid))
2126 			goto out;
2127 	}
2128 
2129 	if ((flags & CHECK_FH) && nfs4_check_fh(current_fh, stp)) {
2130 		printk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
2131 		goto out;
2132 	}
2133 
2134 	*stpp = stp;
2135 	*sopp = sop = stp->st_stateowner;
2136 
2137 	/*
2138 	*  We now validate the seqid and stateid generation numbers.
2139 	*  For the moment, we ignore the possibility of
2140 	*  generation number wraparound.
2141 	*/
2142 	if (seqid != sop->so_seqid + 1)
2143 		goto check_replay;
2144 
2145 	if (sop->so_confirmed) {
2146 		if (flags & CONFIRM) {
2147 			printk("NFSD: preprocess_seqid_op: expected unconfirmed stateowner!\n");
2148 			goto out;
2149 		}
2150 	}
2151 	else {
2152 		if (!(flags & CONFIRM)) {
2153 			printk("NFSD: preprocess_seqid_op: stateowner not confirmed yet!\n");
2154 			goto out;
2155 		}
2156 	}
2157 	if (stateid->si_generation > stp->st_stateid.si_generation) {
2158 		printk("NFSD: preprocess_seqid_op: future stateid?!\n");
2159 		goto out;
2160 	}
2161 
2162 	status = nfserr_old_stateid;
2163 	if (stateid->si_generation < stp->st_stateid.si_generation) {
2164 		printk("NFSD: preprocess_seqid_op: old stateid!\n");
2165 		goto out;
2166 	}
2167 	/* XXX renew the client lease here */
2168 	status = nfs_ok;
2169 
2170 out:
2171 	return status;
2172 
2173 no_nfs4_stateid:
2174 
2175 	/*
2176 	* We determine whether this is a bad stateid or a replay,
2177 	* starting by trying to look up the stateowner.
2178 	* If stateowner is not found - stateid is bad.
2179 	*/
2180 	if (!(sop = find_openstateowner_id(stateid->si_stateownerid, flags))) {
2181 		printk("NFSD: preprocess_seqid_op: no stateowner or nfs4_stateid!\n");
2182 		status = nfserr_bad_stateid;
2183 		goto out;
2184 	}
2185 	*sopp = sop;
2186 
2187 check_replay:
2188 	if (seqid == sop->so_seqid) {
2189 		printk("NFSD: preprocess_seqid_op: retransmission?\n");
2190 		/* indicate replay to calling function */
2191 		status = NFSERR_REPLAY_ME;
2192 	} else  {
2193 		printk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d\n", sop->so_seqid +1, seqid);
2194 
2195 		*sopp = NULL;
2196 		status = nfserr_bad_seqid;
2197 	}
2198 	goto out;
2199 }
2200 
2201 int
2202 nfsd4_open_confirm(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open_confirm *oc)
2203 {
2204 	int status;
2205 	struct nfs4_stateowner *sop;
2206 	struct nfs4_stateid *stp;
2207 
2208 	dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
2209 			(int)current_fh->fh_dentry->d_name.len,
2210 			current_fh->fh_dentry->d_name.name);
2211 
2212 	if ((status = fh_verify(rqstp, current_fh, S_IFREG, 0)))
2213 		goto out;
2214 
2215 	nfs4_lock_state();
2216 
2217 	if ((status = nfs4_preprocess_seqid_op(current_fh, oc->oc_seqid,
2218 					&oc->oc_req_stateid,
2219 					CHECK_FH | CONFIRM | OPEN_STATE,
2220 					&oc->oc_stateowner, &stp, NULL)))
2221 		goto out;
2222 
2223 	sop = oc->oc_stateowner;
2224 	sop->so_confirmed = 1;
2225 	update_stateid(&stp->st_stateid);
2226 	memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
2227 	dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
2228 		"stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
2229 		         stp->st_stateid.si_boot,
2230 		         stp->st_stateid.si_stateownerid,
2231 		         stp->st_stateid.si_fileid,
2232 		         stp->st_stateid.si_generation);
2233 out:
2234 	if (oc->oc_stateowner)
2235 		nfs4_get_stateowner(oc->oc_stateowner);
2236 	nfs4_unlock_state();
2237 	return status;
2238 }
2239 
2240 
2241 /*
2242  * unset all bits in union bitmap (bmap) that
2243  * do not exist in share (from successful OPEN_DOWNGRADE)
2244  */
2245 static void
2246 reset_union_bmap_access(unsigned long access, unsigned long *bmap)
2247 {
2248 	int i;
2249 	for (i = 1; i < 4; i++) {
2250 		if ((i & access) != i)
2251 			__clear_bit(i, bmap);
2252 	}
2253 }
2254 
2255 static void
2256 reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
2257 {
2258 	int i;
2259 	for (i = 0; i < 4; i++) {
2260 		if ((i & deny) != i)
2261 			__clear_bit(i, bmap);
2262 	}
2263 }
2264 
2265 int
2266 nfsd4_open_downgrade(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open_downgrade *od)
2267 {
2268 	int status;
2269 	struct nfs4_stateid *stp;
2270 	unsigned int share_access;
2271 
2272 	dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
2273 			(int)current_fh->fh_dentry->d_name.len,
2274 			current_fh->fh_dentry->d_name.name);
2275 
2276 	if (!TEST_ACCESS(od->od_share_access) || !TEST_DENY(od->od_share_deny))
2277 		return nfserr_inval;
2278 
2279 	nfs4_lock_state();
2280 	if ((status = nfs4_preprocess_seqid_op(current_fh, od->od_seqid,
2281 					&od->od_stateid,
2282 					CHECK_FH | OPEN_STATE,
2283 					&od->od_stateowner, &stp, NULL)))
2284 		goto out;
2285 
2286 	status = nfserr_inval;
2287 	if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
2288 		dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
2289 			stp->st_access_bmap, od->od_share_access);
2290 		goto out;
2291 	}
2292 	if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
2293 		dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
2294 			stp->st_deny_bmap, od->od_share_deny);
2295 		goto out;
2296 	}
2297 	set_access(&share_access, stp->st_access_bmap);
2298 	nfs4_file_downgrade(stp->st_vfs_file,
2299 	                    share_access & ~od->od_share_access);
2300 
2301 	reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
2302 	reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
2303 
2304 	update_stateid(&stp->st_stateid);
2305 	memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
2306 	status = nfs_ok;
2307 out:
2308 	if (od->od_stateowner)
2309 		nfs4_get_stateowner(od->od_stateowner);
2310 	nfs4_unlock_state();
2311 	return status;
2312 }
2313 
2314 /*
2315  * nfs4_unlock_state() called after encode
2316  */
2317 int
2318 nfsd4_close(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_close *close)
2319 {
2320 	int status;
2321 	struct nfs4_stateid *stp;
2322 
2323 	dprintk("NFSD: nfsd4_close on file %.*s\n",
2324 			(int)current_fh->fh_dentry->d_name.len,
2325 			current_fh->fh_dentry->d_name.name);
2326 
2327 	nfs4_lock_state();
2328 	/* check close_lru for replay */
2329 	if ((status = nfs4_preprocess_seqid_op(current_fh, close->cl_seqid,
2330 					&close->cl_stateid,
2331 					CHECK_FH | OPEN_STATE | CLOSE_STATE,
2332 					&close->cl_stateowner, &stp, NULL)))
2333 		goto out;
2334 	/*
2335 	*  Return success, but first update the stateid.
2336 	*/
2337 	status = nfs_ok;
2338 	update_stateid(&stp->st_stateid);
2339 	memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
2340 
2341 	/* release_state_owner() calls nfsd_close() if needed */
2342 	release_state_owner(stp, OPEN_STATE);
2343 out:
2344 	if (close->cl_stateowner)
2345 		nfs4_get_stateowner(close->cl_stateowner);
2346 	nfs4_unlock_state();
2347 	return status;
2348 }
2349 
2350 int
2351 nfsd4_delegreturn(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_delegreturn *dr)
2352 {
2353 	int status;
2354 
2355 	if ((status = fh_verify(rqstp, current_fh, S_IFREG, 0)))
2356 		goto out;
2357 
2358 	nfs4_lock_state();
2359 	status = nfs4_preprocess_stateid_op(current_fh, &dr->dr_stateid, DELEG_RET, NULL);
2360 	nfs4_unlock_state();
2361 out:
2362 	return status;
2363 }
2364 
2365 
2366 /*
2367  * Lock owner state (byte-range locks)
2368  */
2369 #define LOFF_OVERFLOW(start, len)      ((u64)(len) > ~(u64)(start))
2370 #define LOCK_HASH_BITS              8
2371 #define LOCK_HASH_SIZE             (1 << LOCK_HASH_BITS)
2372 #define LOCK_HASH_MASK             (LOCK_HASH_SIZE - 1)
2373 
2374 #define lockownerid_hashval(id) \
2375         ((id) & LOCK_HASH_MASK)
2376 
2377 static inline unsigned int
2378 lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
2379 		struct xdr_netobj *ownername)
2380 {
2381 	return (file_hashval(inode) + cl_id
2382 			+ opaque_hashval(ownername->data, ownername->len))
2383 		& LOCK_HASH_MASK;
2384 }
2385 
2386 static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
2387 static struct list_head	lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
2388 static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
2389 
2390 struct nfs4_stateid *
2391 find_stateid(stateid_t *stid, int flags)
2392 {
2393 	struct nfs4_stateid *local = NULL;
2394 	u32 st_id = stid->si_stateownerid;
2395 	u32 f_id = stid->si_fileid;
2396 	unsigned int hashval;
2397 
2398 	dprintk("NFSD: find_stateid flags 0x%x\n",flags);
2399 	if ((flags & LOCK_STATE) || (flags & RD_STATE) || (flags & WR_STATE)) {
2400 		hashval = stateid_hashval(st_id, f_id);
2401 		list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
2402 			if ((local->st_stateid.si_stateownerid == st_id) &&
2403 			    (local->st_stateid.si_fileid == f_id))
2404 				return local;
2405 		}
2406 	}
2407 	if ((flags & OPEN_STATE) || (flags & RD_STATE) || (flags & WR_STATE)) {
2408 		hashval = stateid_hashval(st_id, f_id);
2409 		list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
2410 			if ((local->st_stateid.si_stateownerid == st_id) &&
2411 			    (local->st_stateid.si_fileid == f_id))
2412 				return local;
2413 		}
2414 	} else
2415 		printk("NFSD: find_stateid: ERROR: no state flag\n");
2416 	return NULL;
2417 }
2418 
2419 static struct nfs4_delegation *
2420 find_delegation_stateid(struct inode *ino, stateid_t *stid)
2421 {
2422 	struct nfs4_delegation *dp = NULL;
2423 	struct nfs4_file *fp = NULL;
2424 	u32 st_id;
2425 
2426 	dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
2427                     stid->si_boot, stid->si_stateownerid,
2428                     stid->si_fileid, stid->si_generation);
2429 
2430 	st_id = stid->si_stateownerid;
2431 	fp = find_file(ino);
2432 	if (fp) {
2433 		list_for_each_entry(dp, &fp->fi_del_perfile, dl_del_perfile) {
2434 			if(dp->dl_stateid.si_stateownerid == st_id) {
2435 				dprintk("NFSD: find_delegation dp %p\n",dp);
2436 				return dp;
2437 			}
2438 		}
2439 	}
2440 	return NULL;
2441 }
2442 
2443 /*
2444  * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
2445  * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
2446  * byte, because of sign extension problems.  Since NFSv4 calls for 64-bit
2447  * locking, this prevents us from being completely protocol-compliant.  The
2448  * real solution to this problem is to start using unsigned file offsets in
2449  * the VFS, but this is a very deep change!
2450  */
2451 static inline void
2452 nfs4_transform_lock_offset(struct file_lock *lock)
2453 {
2454 	if (lock->fl_start < 0)
2455 		lock->fl_start = OFFSET_MAX;
2456 	if (lock->fl_end < 0)
2457 		lock->fl_end = OFFSET_MAX;
2458 }
2459 
2460 int
2461 nfs4_verify_lock_stateowner(struct nfs4_stateowner *sop, unsigned int hashval)
2462 {
2463 	struct nfs4_stateowner *local = NULL;
2464 	int status = 0;
2465 
2466 	if (hashval >= LOCK_HASH_SIZE)
2467 		goto out;
2468 	list_for_each_entry(local, &lock_ownerid_hashtbl[hashval], so_idhash) {
2469 		if (local == sop) {
2470 			status = 1;
2471 			goto out;
2472 		}
2473 	}
2474 out:
2475 	return status;
2476 }
2477 
2478 
2479 static inline void
2480 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
2481 {
2482 	struct nfs4_stateowner *sop = (struct nfs4_stateowner *) fl->fl_owner;
2483 	unsigned int hval = lockownerid_hashval(sop->so_id);
2484 
2485 	deny->ld_sop = NULL;
2486 	if (nfs4_verify_lock_stateowner(sop, hval)) {
2487 		kref_get(&sop->so_ref);
2488 		deny->ld_sop = sop;
2489 		deny->ld_clientid = sop->so_client->cl_clientid;
2490 	}
2491 	deny->ld_start = fl->fl_start;
2492 	deny->ld_length = ~(u64)0;
2493 	if (fl->fl_end != ~(u64)0)
2494 		deny->ld_length = fl->fl_end - fl->fl_start + 1;
2495 	deny->ld_type = NFS4_READ_LT;
2496 	if (fl->fl_type != F_RDLCK)
2497 		deny->ld_type = NFS4_WRITE_LT;
2498 }
2499 
2500 static struct nfs4_stateowner *
2501 find_lockstateowner(struct xdr_netobj *owner, clientid_t *clid)
2502 {
2503 	struct nfs4_stateowner *local = NULL;
2504 	int i;
2505 
2506 	for (i = 0; i < LOCK_HASH_SIZE; i++) {
2507 		list_for_each_entry(local, &lock_ownerid_hashtbl[i], so_idhash) {
2508 			if (!cmp_owner_str(local, owner, clid))
2509 				continue;
2510 			return local;
2511 		}
2512 	}
2513 	return NULL;
2514 }
2515 
2516 static struct nfs4_stateowner *
2517 find_lockstateowner_str(struct inode *inode, clientid_t *clid,
2518 		struct xdr_netobj *owner)
2519 {
2520 	unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
2521 	struct nfs4_stateowner *op;
2522 
2523 	list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
2524 		if (cmp_owner_str(op, owner, clid))
2525 			return op;
2526 	}
2527 	return NULL;
2528 }
2529 
2530 /*
2531  * Alloc a lock owner structure.
2532  * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
2533  * occured.
2534  *
2535  * strhashval = lock_ownerstr_hashval
2536  * so_seqid = lock->lk_new_lock_seqid - 1: it gets bumped in encode
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_perfilestate);
2552 	INIT_LIST_HEAD(&sop->so_perlockowner);
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_perlockowner, &open_stp->st_perlockowner);
2558 	sop->so_is_open_owner = 0;
2559 	sop->so_id = current_ownerid++;
2560 	sop->so_client = clp;
2561 	sop->so_seqid = lock->lk_new_lock_seqid - 1;
2562 	sop->so_confirmed = 1;
2563 	rp = &sop->so_replay;
2564 	rp->rp_status = NFSERR_SERVERFAULT;
2565 	rp->rp_buflen = 0;
2566 	rp->rp_buf = rp->rp_ibuf;
2567 	return sop;
2568 }
2569 
2570 struct nfs4_stateid *
2571 alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
2572 {
2573 	struct nfs4_stateid *stp;
2574 	unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
2575 
2576 	if ((stp = kmalloc(sizeof(struct nfs4_stateid),
2577 					GFP_KERNEL)) == NULL)
2578 		goto out;
2579 	INIT_LIST_HEAD(&stp->st_hash);
2580 	INIT_LIST_HEAD(&stp->st_perfile);
2581 	INIT_LIST_HEAD(&stp->st_perfilestate);
2582 	INIT_LIST_HEAD(&stp->st_perlockowner); /* not used */
2583 	list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
2584 	list_add(&stp->st_perfile, &fp->fi_perfile);
2585 	list_add_perfile++;
2586 	list_add(&stp->st_perfilestate, &sop->so_perfilestate);
2587 	stp->st_stateowner = sop;
2588 	stp->st_file = fp;
2589 	stp->st_stateid.si_boot = boot_time;
2590 	stp->st_stateid.si_stateownerid = sop->so_id;
2591 	stp->st_stateid.si_fileid = fp->fi_id;
2592 	stp->st_stateid.si_generation = 0;
2593 	stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
2594 	stp->st_access_bmap = open_stp->st_access_bmap;
2595 	stp->st_deny_bmap = open_stp->st_deny_bmap;
2596 
2597 out:
2598 	return stp;
2599 }
2600 
2601 int
2602 check_lock_length(u64 offset, u64 length)
2603 {
2604 	return ((length == 0)  || ((length != ~(u64)0) &&
2605 	     LOFF_OVERFLOW(offset, length)));
2606 }
2607 
2608 /*
2609  *  LOCK operation
2610  */
2611 int
2612 nfsd4_lock(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_lock *lock)
2613 {
2614 	struct nfs4_stateowner *lock_sop = NULL, *open_sop = NULL;
2615 	struct nfs4_stateid *lock_stp;
2616 	struct file *filp;
2617 	struct file_lock file_lock;
2618 	struct file_lock *conflock;
2619 	int status = 0;
2620 	unsigned int strhashval;
2621 
2622 	dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
2623 		(long long) lock->lk_offset,
2624 		(long long) lock->lk_length);
2625 
2626 	if (nfs4_in_grace() && !lock->lk_reclaim)
2627 		return nfserr_grace;
2628 	if (!nfs4_in_grace() && lock->lk_reclaim)
2629 		return nfserr_no_grace;
2630 
2631 	if (check_lock_length(lock->lk_offset, lock->lk_length))
2632 		 return nfserr_inval;
2633 
2634 	nfs4_lock_state();
2635 
2636 	if (lock->lk_is_new) {
2637 	/*
2638 	 * Client indicates that this is a new lockowner.
2639 	 * Use open owner and open stateid to create lock owner and lock
2640 	 * stateid.
2641 	 */
2642 		struct nfs4_stateid *open_stp = NULL;
2643 		struct nfs4_file *fp;
2644 
2645 		status = nfserr_stale_clientid;
2646 		if (STALE_CLIENTID(&lock->lk_new_clientid)) {
2647 			printk("NFSD: nfsd4_lock: clientid is stale!\n");
2648 			goto out;
2649 		}
2650 
2651 		/* is the new lock seqid presented by the client zero? */
2652 		status = nfserr_bad_seqid;
2653 		if (lock->v.new.lock_seqid != 0)
2654 			goto out;
2655 
2656 		/* validate and update open stateid and open seqid */
2657 		status = nfs4_preprocess_seqid_op(current_fh,
2658 				        lock->lk_new_open_seqid,
2659 		                        &lock->lk_new_open_stateid,
2660 		                        CHECK_FH | OPEN_STATE,
2661 		                        &open_sop, &open_stp,
2662 					&lock->v.new.clientid);
2663 		if (status) {
2664 			if (lock->lk_reclaim)
2665 				status = nfserr_reclaim_bad;
2666 			goto out;
2667 		}
2668 		/* create lockowner and lock stateid */
2669 		fp = open_stp->st_file;
2670 		strhashval = lock_ownerstr_hashval(fp->fi_inode,
2671 				open_sop->so_client->cl_clientid.cl_id,
2672 				&lock->v.new.owner);
2673 		/*
2674 		 * If we already have this lock owner, the client is in
2675 		 * error (or our bookeeping is wrong!)
2676 		 * for asking for a 'new lock'.
2677 		 */
2678 		status = nfserr_bad_stateid;
2679 		lock_sop = find_lockstateowner(&lock->v.new.owner,
2680 						&lock->v.new.clientid);
2681 		if (lock_sop)
2682 			goto out;
2683 		status = nfserr_resource;
2684 		if (!(lock->lk_stateowner = alloc_init_lock_stateowner(strhashval, open_sop->so_client, open_stp, lock)))
2685 			goto out;
2686 		if ((lock_stp = alloc_init_lock_stateid(lock->lk_stateowner,
2687 						fp, open_stp)) == NULL) {
2688 			release_stateowner(lock->lk_stateowner);
2689 			lock->lk_stateowner = NULL;
2690 			goto out;
2691 		}
2692 		/* bump the open seqid used to create the lock */
2693 		open_sop->so_seqid++;
2694 	} else {
2695 		/* lock (lock owner + lock stateid) already exists */
2696 		status = nfs4_preprocess_seqid_op(current_fh,
2697 				       lock->lk_old_lock_seqid,
2698 				       &lock->lk_old_lock_stateid,
2699 				       CHECK_FH | LOCK_STATE,
2700 				       &lock->lk_stateowner, &lock_stp, NULL);
2701 		if (status)
2702 			goto out;
2703 	}
2704 	/* lock->lk_stateowner and lock_stp have been created or found */
2705 	filp = lock_stp->st_vfs_file;
2706 
2707 	if ((status = fh_verify(rqstp, current_fh, S_IFREG, MAY_LOCK))) {
2708 		printk("NFSD: nfsd4_lock: permission denied!\n");
2709 		goto out;
2710 	}
2711 
2712 	locks_init_lock(&file_lock);
2713 	switch (lock->lk_type) {
2714 		case NFS4_READ_LT:
2715 		case NFS4_READW_LT:
2716 			file_lock.fl_type = F_RDLCK;
2717 		break;
2718 		case NFS4_WRITE_LT:
2719 		case NFS4_WRITEW_LT:
2720 			file_lock.fl_type = F_WRLCK;
2721 		break;
2722 		default:
2723 			status = nfserr_inval;
2724 		goto out;
2725 	}
2726 	file_lock.fl_owner = (fl_owner_t) lock->lk_stateowner;
2727 	file_lock.fl_pid = current->tgid;
2728 	file_lock.fl_file = filp;
2729 	file_lock.fl_flags = FL_POSIX;
2730 
2731 	file_lock.fl_start = lock->lk_offset;
2732 	if ((lock->lk_length == ~(u64)0) ||
2733 			LOFF_OVERFLOW(lock->lk_offset, lock->lk_length))
2734 		file_lock.fl_end = ~(u64)0;
2735 	else
2736 		file_lock.fl_end = lock->lk_offset + lock->lk_length - 1;
2737 	nfs4_transform_lock_offset(&file_lock);
2738 
2739 	/*
2740 	* Try to lock the file in the VFS.
2741 	* Note: locks.c uses the BKL to protect the inode's lock list.
2742 	*/
2743 
2744 	status = posix_lock_file(filp, &file_lock);
2745 	if (file_lock.fl_ops && file_lock.fl_ops->fl_release_private)
2746 		file_lock.fl_ops->fl_release_private(&file_lock);
2747 	dprintk("NFSD: nfsd4_lock: posix_lock_file status %d\n",status);
2748 	switch (-status) {
2749 	case 0: /* success! */
2750 		update_stateid(&lock_stp->st_stateid);
2751 		memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
2752 				sizeof(stateid_t));
2753 		goto out;
2754 	case (EAGAIN):
2755 		goto conflicting_lock;
2756 	case (EDEADLK):
2757 		status = nfserr_deadlock;
2758 	default:
2759 		dprintk("NFSD: nfsd4_lock: posix_lock_file() failed! status %d\n",status);
2760 		goto out_destroy_new_stateid;
2761 	}
2762 
2763 conflicting_lock:
2764 	dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
2765 	status = nfserr_denied;
2766 	/* XXX There is a race here. Future patch needed to provide
2767 	 * an atomic posix_lock_and_test_file
2768 	 */
2769 	if (!(conflock = posix_test_lock(filp, &file_lock))) {
2770 		status = nfserr_serverfault;
2771 		goto out;
2772 	}
2773 	nfs4_set_lock_denied(conflock, &lock->lk_denied);
2774 
2775 out_destroy_new_stateid:
2776 	if (lock->lk_is_new) {
2777 		dprintk("NFSD: nfsd4_lock: destroy new stateid!\n");
2778 	/*
2779 	* An error encountered after instantiation of the new
2780 	* stateid has forced us to destroy it.
2781 	*/
2782 		if (!seqid_mutating_err(status))
2783 			open_sop->so_seqid--;
2784 
2785 		release_state_owner(lock_stp, LOCK_STATE);
2786 	}
2787 out:
2788 	if (lock->lk_stateowner)
2789 		nfs4_get_stateowner(lock->lk_stateowner);
2790 	nfs4_unlock_state();
2791 	return status;
2792 }
2793 
2794 /*
2795  * LOCKT operation
2796  */
2797 int
2798 nfsd4_lockt(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_lockt *lockt)
2799 {
2800 	struct inode *inode;
2801 	struct file file;
2802 	struct file_lock file_lock;
2803 	struct file_lock *conflicting_lock;
2804 	int status;
2805 
2806 	if (nfs4_in_grace())
2807 		return nfserr_grace;
2808 
2809 	if (check_lock_length(lockt->lt_offset, lockt->lt_length))
2810 		 return nfserr_inval;
2811 
2812 	lockt->lt_stateowner = NULL;
2813 	nfs4_lock_state();
2814 
2815 	status = nfserr_stale_clientid;
2816 	if (STALE_CLIENTID(&lockt->lt_clientid)) {
2817 		printk("NFSD: nfsd4_lockt: clientid is stale!\n");
2818 		goto out;
2819 	}
2820 
2821 	if ((status = fh_verify(rqstp, current_fh, S_IFREG, 0))) {
2822 		printk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
2823 		if (status == nfserr_symlink)
2824 			status = nfserr_inval;
2825 		goto out;
2826 	}
2827 
2828 	inode = current_fh->fh_dentry->d_inode;
2829 	locks_init_lock(&file_lock);
2830 	switch (lockt->lt_type) {
2831 		case NFS4_READ_LT:
2832 		case NFS4_READW_LT:
2833 			file_lock.fl_type = F_RDLCK;
2834 		break;
2835 		case NFS4_WRITE_LT:
2836 		case NFS4_WRITEW_LT:
2837 			file_lock.fl_type = F_WRLCK;
2838 		break;
2839 		default:
2840 			printk("NFSD: nfs4_lockt: bad lock type!\n");
2841 			status = nfserr_inval;
2842 		goto out;
2843 	}
2844 
2845 	lockt->lt_stateowner = find_lockstateowner_str(inode,
2846 			&lockt->lt_clientid, &lockt->lt_owner);
2847 	if (lockt->lt_stateowner)
2848 		file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
2849 	file_lock.fl_pid = current->tgid;
2850 	file_lock.fl_flags = FL_POSIX;
2851 
2852 	file_lock.fl_start = lockt->lt_offset;
2853 	if ((lockt->lt_length == ~(u64)0) || LOFF_OVERFLOW(lockt->lt_offset, lockt->lt_length))
2854 		file_lock.fl_end = ~(u64)0;
2855 	else
2856 		file_lock.fl_end = lockt->lt_offset + lockt->lt_length - 1;
2857 
2858 	nfs4_transform_lock_offset(&file_lock);
2859 
2860 	/* posix_test_lock uses the struct file _only_ to resolve the inode.
2861 	 * since LOCKT doesn't require an OPEN, and therefore a struct
2862 	 * file may not exist, pass posix_test_lock a struct file with
2863 	 * only the dentry:inode set.
2864 	 */
2865 	memset(&file, 0, sizeof (struct file));
2866 	file.f_dentry = current_fh->fh_dentry;
2867 
2868 	status = nfs_ok;
2869 	conflicting_lock = posix_test_lock(&file, &file_lock);
2870 	if (conflicting_lock) {
2871 		status = nfserr_denied;
2872 		nfs4_set_lock_denied(conflicting_lock, &lockt->lt_denied);
2873 	}
2874 out:
2875 	nfs4_unlock_state();
2876 	return status;
2877 }
2878 
2879 int
2880 nfsd4_locku(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_locku *locku)
2881 {
2882 	struct nfs4_stateid *stp;
2883 	struct file *filp = NULL;
2884 	struct file_lock file_lock;
2885 	int status;
2886 
2887 	dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
2888 		(long long) locku->lu_offset,
2889 		(long long) locku->lu_length);
2890 
2891 	if (check_lock_length(locku->lu_offset, locku->lu_length))
2892 		 return nfserr_inval;
2893 
2894 	nfs4_lock_state();
2895 
2896 	if ((status = nfs4_preprocess_seqid_op(current_fh,
2897 					locku->lu_seqid,
2898 					&locku->lu_stateid,
2899 					CHECK_FH | LOCK_STATE,
2900 					&locku->lu_stateowner, &stp, NULL)))
2901 		goto out;
2902 
2903 	filp = stp->st_vfs_file;
2904 	BUG_ON(!filp);
2905 	locks_init_lock(&file_lock);
2906 	file_lock.fl_type = F_UNLCK;
2907 	file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
2908 	file_lock.fl_pid = current->tgid;
2909 	file_lock.fl_file = filp;
2910 	file_lock.fl_flags = FL_POSIX;
2911 	file_lock.fl_start = locku->lu_offset;
2912 
2913 	if ((locku->lu_length == ~(u64)0) || LOFF_OVERFLOW(locku->lu_offset, locku->lu_length))
2914 		file_lock.fl_end = ~(u64)0;
2915 	else
2916 		file_lock.fl_end = locku->lu_offset + locku->lu_length - 1;
2917 	nfs4_transform_lock_offset(&file_lock);
2918 
2919 	/*
2920 	*  Try to unlock the file in the VFS.
2921 	*/
2922 	status = posix_lock_file(filp, &file_lock);
2923 	if (file_lock.fl_ops && file_lock.fl_ops->fl_release_private)
2924 		file_lock.fl_ops->fl_release_private(&file_lock);
2925 	if (status) {
2926 		printk("NFSD: nfs4_locku: posix_lock_file failed!\n");
2927 		goto out_nfserr;
2928 	}
2929 	/*
2930 	* OK, unlock succeeded; the only thing left to do is update the stateid.
2931 	*/
2932 	update_stateid(&stp->st_stateid);
2933 	memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
2934 
2935 out:
2936 	if (locku->lu_stateowner)
2937 		nfs4_get_stateowner(locku->lu_stateowner);
2938 	nfs4_unlock_state();
2939 	return status;
2940 
2941 out_nfserr:
2942 	status = nfserrno(status);
2943 	goto out;
2944 }
2945 
2946 /*
2947  * returns
2948  * 	1: locks held by lockowner
2949  * 	0: no locks held by lockowner
2950  */
2951 static int
2952 check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
2953 {
2954 	struct file_lock **flpp;
2955 	struct inode *inode = filp->f_dentry->d_inode;
2956 	int status = 0;
2957 
2958 	lock_kernel();
2959 	for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
2960 		if ((*flpp)->fl_owner == (fl_owner_t)lowner)
2961 			status = 1;
2962 			goto out;
2963 	}
2964 out:
2965 	unlock_kernel();
2966 	return status;
2967 }
2968 
2969 int
2970 nfsd4_release_lockowner(struct svc_rqst *rqstp, struct nfsd4_release_lockowner *rlockowner)
2971 {
2972 	clientid_t *clid = &rlockowner->rl_clientid;
2973 	struct nfs4_stateowner *local = NULL;
2974 	struct xdr_netobj *owner = &rlockowner->rl_owner;
2975 	int status;
2976 
2977 	dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
2978 		clid->cl_boot, clid->cl_id);
2979 
2980 	/* XXX check for lease expiration */
2981 
2982 	status = nfserr_stale_clientid;
2983 	if (STALE_CLIENTID(clid)) {
2984 		printk("NFSD: nfsd4_release_lockowner: clientid is stale!\n");
2985 		return status;
2986 	}
2987 
2988 	nfs4_lock_state();
2989 
2990 	status = nfs_ok;
2991 	local = find_lockstateowner(owner, clid);
2992 	if (local) {
2993 		struct nfs4_stateid *stp;
2994 
2995 		/* check for any locks held by any stateid
2996 		 * associated with the (lock) stateowner */
2997 		status = nfserr_locks_held;
2998 		list_for_each_entry(stp, &local->so_perfilestate,
2999 				st_perfilestate) {
3000 			if (check_for_locks(stp->st_vfs_file, local))
3001 				goto out;
3002 		}
3003 		/* no locks held by (lock) stateowner */
3004 		status = nfs_ok;
3005 		release_stateowner(local);
3006 	}
3007 out:
3008 	nfs4_unlock_state();
3009 	return status;
3010 }
3011 
3012 static inline struct nfs4_client_reclaim *
3013 alloc_reclaim(int namelen)
3014 {
3015 	struct nfs4_client_reclaim *crp = NULL;
3016 
3017 	crp = kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
3018 	if (!crp)
3019 		return NULL;
3020 	crp->cr_name.data = kmalloc(namelen, GFP_KERNEL);
3021 	if (!crp->cr_name.data) {
3022 		kfree(crp);
3023 		return NULL;
3024 	}
3025 	return crp;
3026 }
3027 
3028 /*
3029  * failure => all reset bets are off, nfserr_no_grace...
3030  */
3031 static int
3032 nfs4_client_to_reclaim(char *name, int namlen)
3033 {
3034 	unsigned int strhashval;
3035 	struct nfs4_client_reclaim *crp = NULL;
3036 
3037 	dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", namlen, name);
3038 	crp = alloc_reclaim(namlen);
3039 	if (!crp)
3040 		return 0;
3041 	strhashval = clientstr_hashval(name, namlen);
3042 	INIT_LIST_HEAD(&crp->cr_strhash);
3043 	list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
3044 	memcpy(crp->cr_name.data, name, namlen);
3045 	crp->cr_name.len = namlen;
3046 	reclaim_str_hashtbl_size++;
3047 	return 1;
3048 }
3049 
3050 static void
3051 nfs4_release_reclaim(void)
3052 {
3053 	struct nfs4_client_reclaim *crp = NULL;
3054 	int i;
3055 
3056 	BUG_ON(!nfs4_reclaim_init);
3057 	for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3058 		while (!list_empty(&reclaim_str_hashtbl[i])) {
3059 			crp = list_entry(reclaim_str_hashtbl[i].next,
3060 			                struct nfs4_client_reclaim, cr_strhash);
3061 			list_del(&crp->cr_strhash);
3062 			kfree(crp->cr_name.data);
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 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\n",
3086 		            clp->cl_name.len, clp->cl_name.data);
3087 
3088 	/* find clp->cl_name in reclaim_str_hashtbl */
3089 	strhashval = clientstr_hashval(clp->cl_name.data, clp->cl_name.len);
3090 	list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
3091 		if (cmp_name(&crp->cr_name, &clp->cl_name)) {
3092 			return crp;
3093 		}
3094 	}
3095 	return NULL;
3096 }
3097 
3098 /*
3099 * Called from OPEN. Look for clientid in reclaim list.
3100 */
3101 int
3102 nfs4_check_open_reclaim(clientid_t *clid)
3103 {
3104 	struct nfs4_client_reclaim *crp;
3105 
3106 	if ((crp = nfs4_find_reclaim_client(clid)) == NULL)
3107 		return nfserr_reclaim_bad;
3108 	return nfs_ok;
3109 }
3110 
3111 
3112 /*
3113  * Start and stop routines
3114  */
3115 
3116 static void
3117 __nfs4_state_init(void)
3118 {
3119 	int i;
3120 	time_t grace_time;
3121 
3122 	if (!nfs4_reclaim_init) {
3123 		for (i = 0; i < CLIENT_HASH_SIZE; i++)
3124 			INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
3125 		reclaim_str_hashtbl_size = 0;
3126 		nfs4_reclaim_init = 1;
3127 	}
3128 	for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3129 		INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3130 		INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3131 		INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3132 		INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
3133 	}
3134 	for (i = 0; i < FILE_HASH_SIZE; i++) {
3135 		INIT_LIST_HEAD(&file_hashtbl[i]);
3136 	}
3137 	for (i = 0; i < OWNER_HASH_SIZE; i++) {
3138 		INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3139 		INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3140 	}
3141 	for (i = 0; i < STATEID_HASH_SIZE; i++) {
3142 		INIT_LIST_HEAD(&stateid_hashtbl[i]);
3143 		INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3144 	}
3145 	for (i = 0; i < LOCK_HASH_SIZE; i++) {
3146 		INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3147 		INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3148 	}
3149 	memset(&zerostateid, 0, sizeof(stateid_t));
3150 	memset(&onestateid, ~0, sizeof(stateid_t));
3151 
3152 	INIT_LIST_HEAD(&close_lru);
3153 	INIT_LIST_HEAD(&client_lru);
3154 	INIT_LIST_HEAD(&del_recall_lru);
3155 	spin_lock_init(&recall_lock);
3156 	boot_time = get_seconds();
3157 	grace_time = max(old_lease_time, lease_time);
3158 	if (reclaim_str_hashtbl_size == 0)
3159 		grace_time = 0;
3160 	if (grace_time)
3161 		printk("NFSD: starting %ld-second grace period\n", grace_time);
3162 	grace_end = boot_time + grace_time;
3163 	INIT_WORK(&laundromat_work,laundromat_main, NULL);
3164 	schedule_delayed_work(&laundromat_work, NFSD_LEASE_TIME*HZ);
3165 }
3166 
3167 int
3168 nfs4_state_init(void)
3169 {
3170 	int status;
3171 
3172 	if (nfs4_init)
3173 		return 0;
3174 	status = nfsd4_init_slabs();
3175 	if (status)
3176 		return status;
3177 	__nfs4_state_init();
3178 	nfs4_init = 1;
3179 	return 0;
3180 }
3181 
3182 int
3183 nfs4_in_grace(void)
3184 {
3185 	return get_seconds() < grace_end;
3186 }
3187 
3188 void
3189 set_no_grace(void)
3190 {
3191 	printk("NFSD: ERROR in reboot recovery.  State reclaims will fail.\n");
3192 	grace_end = get_seconds();
3193 }
3194 
3195 time_t
3196 nfs4_lease_time(void)
3197 {
3198 	return lease_time;
3199 }
3200 
3201 static void
3202 __nfs4_state_shutdown(void)
3203 {
3204 	int i;
3205 	struct nfs4_client *clp = NULL;
3206 	struct nfs4_delegation *dp = NULL;
3207 	struct nfs4_stateowner *sop = NULL;
3208 	struct list_head *pos, *next, reaplist;
3209 
3210 	list_for_each_safe(pos, next, &close_lru) {
3211 		sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
3212 		list_del(&sop->so_close_lru);
3213 		nfs4_put_stateowner(sop);
3214 	}
3215 
3216 	for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3217 		while (!list_empty(&conf_id_hashtbl[i])) {
3218 			clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
3219 			expire_client(clp);
3220 		}
3221 		while (!list_empty(&unconf_str_hashtbl[i])) {
3222 			clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
3223 			expire_client(clp);
3224 		}
3225 	}
3226 	INIT_LIST_HEAD(&reaplist);
3227 	spin_lock(&recall_lock);
3228 	list_for_each_safe(pos, next, &del_recall_lru) {
3229 		dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3230 		list_move(&dp->dl_recall_lru, &reaplist);
3231 	}
3232 	spin_unlock(&recall_lock);
3233 	list_for_each_safe(pos, next, &reaplist) {
3234 		dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3235 		list_del_init(&dp->dl_recall_lru);
3236 		unhash_delegation(dp);
3237 	}
3238 
3239 	release_all_files();
3240 	cancel_delayed_work(&laundromat_work);
3241 	flush_scheduled_work();
3242 	nfs4_init = 0;
3243 	dprintk("NFSD: list_add_perfile %d list_del_perfile %d\n",
3244 			list_add_perfile, list_del_perfile);
3245 	dprintk("NFSD: add_perclient %d del_perclient %d\n",
3246 			add_perclient, del_perclient);
3247 	dprintk("NFSD: alloc_file %d free_file %d\n",
3248 			alloc_file, free_file);
3249 	dprintk("NFSD: vfsopen %d vfsclose %d\n",
3250 			vfsopen, vfsclose);
3251 	dprintk("NFSD: alloc_delegation %d free_delegation %d\n",
3252 			alloc_delegation, free_delegation);
3253 
3254 }
3255 
3256 void
3257 nfs4_state_shutdown(void)
3258 {
3259 	nfs4_lock_state();
3260 	nfs4_release_reclaim();
3261 	__nfs4_state_shutdown();
3262 	nfsd4_free_slabs();
3263 	nfs4_unlock_state();
3264 }
3265 
3266 /*
3267  * Called when leasetime is changed.
3268  *
3269  * if nfsd is not started, simply set the global lease.
3270  *
3271  * if nfsd(s) are running, lease change requires nfsv4 state to be reset.
3272  * e.g: boot_time is reset, existing nfs4_client structs are
3273  * used to fill reclaim_str_hashtbl, then all state (except for the
3274  * reclaim_str_hashtbl) is re-initialized.
3275  *
3276  * if the old lease time is greater than the new lease time, the grace
3277  * period needs to be set to the old lease time to allow clients to reclaim
3278  * their state. XXX - we may want to set the grace period == lease time
3279  * after an initial grace period == old lease time
3280  *
3281  * if an error occurs in this process, the new lease is set, but the server
3282  * will not honor OPEN or LOCK reclaims, and will return nfserr_no_grace
3283  * which means OPEN/LOCK/READ/WRITE will fail during grace period.
3284  *
3285  * clients will attempt to reset all state with SETCLIENTID/CONFIRM, and
3286  * OPEN and LOCK reclaims.
3287  */
3288 void
3289 nfs4_reset_lease(time_t leasetime)
3290 {
3291 	struct nfs4_client *clp;
3292 	int i;
3293 
3294 	printk("NFSD: New leasetime %ld\n",leasetime);
3295 	if (!nfs4_init)
3296 		return;
3297 	nfs4_lock_state();
3298 	old_lease_time = lease_time;
3299 	lease_time = leasetime;
3300 
3301 	nfs4_release_reclaim();
3302 
3303 	/* populate reclaim_str_hashtbl with current confirmed nfs4_clientid */
3304 	for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3305 		list_for_each_entry(clp, &conf_id_hashtbl[i], cl_idhash) {
3306 			if (!nfs4_client_to_reclaim(clp->cl_name.data,
3307 						clp->cl_name.len)) {
3308 				nfs4_release_reclaim();
3309 				goto init_state;
3310 			}
3311 		}
3312 	}
3313 init_state:
3314 	__nfs4_state_shutdown();
3315 	__nfs4_state_init();
3316 	nfs4_unlock_state();
3317 }
3318 
3319