m68knommu: use IRQF_DISABLE in m68328serial.c
[wrt350n-kernel.git] / fs / nfsd / nfs4state.c
blob31673cd251c3c2936b99e566a34773f5a0dce0c1
1 /*
2 * linux/fs/nfsd/nfs4state.c
4 * Copyright (c) 2001 The Regents of the University of Michigan.
5 * All rights reserved.
7 * Kendrick Smith <kmsmith@umich.edu>
8 * Andy Adamson <kandros@umich.edu>
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
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.
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.
37 #include <linux/param.h>
38 #include <linux/major.h>
39 #include <linux/slab.h>
41 #include <linux/sunrpc/svc.h>
42 #include <linux/nfsd/nfsd.h>
43 #include <linux/nfsd/cache.h>
44 #include <linux/mount.h>
45 #include <linux/workqueue.h>
46 #include <linux/smp_lock.h>
47 #include <linux/kthread.h>
48 #include <linux/nfs4.h>
49 #include <linux/nfsd/state.h>
50 #include <linux/nfsd/xdr4.h>
51 #include <linux/namei.h>
52 #include <linux/swap.h>
53 #include <linux/mutex.h>
54 #include <linux/lockd/bind.h>
55 #include <linux/module.h>
57 #define NFSDDBG_FACILITY NFSDDBG_PROC
59 /* Globals */
60 static time_t lease_time = 90; /* default lease time */
61 static time_t user_lease_time = 90;
62 static time_t boot_time;
63 static int in_grace = 1;
64 static u32 current_clientid = 1;
65 static u32 current_ownerid = 1;
66 static u32 current_fileid = 1;
67 static u32 current_delegid = 1;
68 static u32 nfs4_init;
69 static stateid_t zerostateid; /* bits all 0 */
70 static stateid_t onestateid; /* bits all 1 */
72 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
73 #define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
75 /* forward declarations */
76 static struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
77 static struct nfs4_delegation * find_delegation_stateid(struct inode *ino, stateid_t *stid);
78 static void release_stateid_lockowners(struct nfs4_stateid *open_stp);
79 static char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery";
80 static void nfs4_set_recdir(char *recdir);
82 /* Locking:
84 * client_mutex:
85 * protects clientid_hashtbl[], clientstr_hashtbl[],
86 * unconfstr_hashtbl[], uncofid_hashtbl[].
88 static DEFINE_MUTEX(client_mutex);
90 static struct kmem_cache *stateowner_slab = NULL;
91 static struct kmem_cache *file_slab = NULL;
92 static struct kmem_cache *stateid_slab = NULL;
93 static struct kmem_cache *deleg_slab = NULL;
95 void
96 nfs4_lock_state(void)
98 mutex_lock(&client_mutex);
101 void
102 nfs4_unlock_state(void)
104 mutex_unlock(&client_mutex);
107 static inline u32
108 opaque_hashval(const void *ptr, int nbytes)
110 unsigned char *cptr = (unsigned char *) ptr;
112 u32 x = 0;
113 while (nbytes--) {
114 x *= 37;
115 x += *cptr++;
117 return x;
120 /* forward declarations */
121 static void release_stateowner(struct nfs4_stateowner *sop);
122 static void release_stateid(struct nfs4_stateid *stp, int flags);
125 * Delegation state
128 /* recall_lock protects the del_recall_lru */
129 static DEFINE_SPINLOCK(recall_lock);
130 static struct list_head del_recall_lru;
132 static void
133 free_nfs4_file(struct kref *kref)
135 struct nfs4_file *fp = container_of(kref, struct nfs4_file, fi_ref);
136 list_del(&fp->fi_hash);
137 iput(fp->fi_inode);
138 kmem_cache_free(file_slab, fp);
141 static inline void
142 put_nfs4_file(struct nfs4_file *fi)
144 kref_put(&fi->fi_ref, free_nfs4_file);
147 static inline void
148 get_nfs4_file(struct nfs4_file *fi)
150 kref_get(&fi->fi_ref);
153 static int num_delegations;
154 unsigned int max_delegations;
157 * Open owner state (share locks)
160 /* hash tables for nfs4_stateowner */
161 #define OWNER_HASH_BITS 8
162 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
163 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
165 #define ownerid_hashval(id) \
166 ((id) & OWNER_HASH_MASK)
167 #define ownerstr_hashval(clientid, ownername) \
168 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
170 static struct list_head ownerid_hashtbl[OWNER_HASH_SIZE];
171 static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
173 /* hash table for nfs4_file */
174 #define FILE_HASH_BITS 8
175 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
176 #define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
177 /* hash table for (open)nfs4_stateid */
178 #define STATEID_HASH_BITS 10
179 #define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
180 #define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
182 #define file_hashval(x) \
183 hash_ptr(x, FILE_HASH_BITS)
184 #define stateid_hashval(owner_id, file_id) \
185 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
187 static struct list_head file_hashtbl[FILE_HASH_SIZE];
188 static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
190 static struct nfs4_delegation *
191 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
193 struct nfs4_delegation *dp;
194 struct nfs4_file *fp = stp->st_file;
195 struct nfs4_callback *cb = &stp->st_stateowner->so_client->cl_callback;
197 dprintk("NFSD alloc_init_deleg\n");
198 if (fp->fi_had_conflict)
199 return NULL;
200 if (num_delegations > max_delegations)
201 return NULL;
202 dp = kmem_cache_alloc(deleg_slab, GFP_KERNEL);
203 if (dp == NULL)
204 return dp;
205 num_delegations++;
206 INIT_LIST_HEAD(&dp->dl_perfile);
207 INIT_LIST_HEAD(&dp->dl_perclnt);
208 INIT_LIST_HEAD(&dp->dl_recall_lru);
209 dp->dl_client = clp;
210 get_nfs4_file(fp);
211 dp->dl_file = fp;
212 dp->dl_flock = NULL;
213 get_file(stp->st_vfs_file);
214 dp->dl_vfs_file = stp->st_vfs_file;
215 dp->dl_type = type;
216 dp->dl_recall.cbr_dp = NULL;
217 dp->dl_recall.cbr_ident = cb->cb_ident;
218 dp->dl_recall.cbr_trunc = 0;
219 dp->dl_stateid.si_boot = boot_time;
220 dp->dl_stateid.si_stateownerid = current_delegid++;
221 dp->dl_stateid.si_fileid = 0;
222 dp->dl_stateid.si_generation = 0;
223 dp->dl_fhlen = current_fh->fh_handle.fh_size;
224 memcpy(dp->dl_fhval, &current_fh->fh_handle.fh_base,
225 current_fh->fh_handle.fh_size);
226 dp->dl_time = 0;
227 atomic_set(&dp->dl_count, 1);
228 list_add(&dp->dl_perfile, &fp->fi_delegations);
229 list_add(&dp->dl_perclnt, &clp->cl_delegations);
230 return dp;
233 void
234 nfs4_put_delegation(struct nfs4_delegation *dp)
236 if (atomic_dec_and_test(&dp->dl_count)) {
237 dprintk("NFSD: freeing dp %p\n",dp);
238 put_nfs4_file(dp->dl_file);
239 kmem_cache_free(deleg_slab, dp);
240 num_delegations--;
244 /* Remove the associated file_lock first, then remove the delegation.
245 * lease_modify() is called to remove the FS_LEASE file_lock from
246 * the i_flock list, eventually calling nfsd's lock_manager
247 * fl_release_callback.
249 static void
250 nfs4_close_delegation(struct nfs4_delegation *dp)
252 struct file *filp = dp->dl_vfs_file;
254 dprintk("NFSD: close_delegation dp %p\n",dp);
255 dp->dl_vfs_file = NULL;
256 /* The following nfsd_close may not actually close the file,
257 * but we want to remove the lease in any case. */
258 if (dp->dl_flock)
259 vfs_setlease(filp, F_UNLCK, &dp->dl_flock);
260 nfsd_close(filp);
263 /* Called under the state lock. */
264 static void
265 unhash_delegation(struct nfs4_delegation *dp)
267 list_del_init(&dp->dl_perfile);
268 list_del_init(&dp->dl_perclnt);
269 spin_lock(&recall_lock);
270 list_del_init(&dp->dl_recall_lru);
271 spin_unlock(&recall_lock);
272 nfs4_close_delegation(dp);
273 nfs4_put_delegation(dp);
277 * SETCLIENTID state
280 /* Hash tables for nfs4_clientid state */
281 #define CLIENT_HASH_BITS 4
282 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
283 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
285 #define clientid_hashval(id) \
286 ((id) & CLIENT_HASH_MASK)
287 #define clientstr_hashval(name) \
288 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
290 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
291 * used in reboot/reset lease grace period processing
293 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
294 * setclientid_confirmed info.
296 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
297 * setclientid info.
299 * client_lru holds client queue ordered by nfs4_client.cl_time
300 * for lease renewal.
302 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
303 * for last close replay.
305 static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
306 static int reclaim_str_hashtbl_size = 0;
307 static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
308 static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
309 static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
310 static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
311 static struct list_head client_lru;
312 static struct list_head close_lru;
314 static inline void
315 renew_client(struct nfs4_client *clp)
318 * Move client to the end to the LRU list.
320 dprintk("renewing client (clientid %08x/%08x)\n",
321 clp->cl_clientid.cl_boot,
322 clp->cl_clientid.cl_id);
323 list_move_tail(&clp->cl_lru, &client_lru);
324 clp->cl_time = get_seconds();
327 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
328 static int
329 STALE_CLIENTID(clientid_t *clid)
331 if (clid->cl_boot == boot_time)
332 return 0;
333 dprintk("NFSD stale clientid (%08x/%08x)\n",
334 clid->cl_boot, clid->cl_id);
335 return 1;
339 * XXX Should we use a slab cache ?
340 * This type of memory management is somewhat inefficient, but we use it
341 * anyway since SETCLIENTID is not a common operation.
343 static inline struct nfs4_client *
344 alloc_client(struct xdr_netobj name)
346 struct nfs4_client *clp;
348 if ((clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL))!= NULL) {
349 if ((clp->cl_name.data = kmalloc(name.len, GFP_KERNEL)) != NULL) {
350 memcpy(clp->cl_name.data, name.data, name.len);
351 clp->cl_name.len = name.len;
353 else {
354 kfree(clp);
355 clp = NULL;
358 return clp;
361 static void
362 shutdown_callback_client(struct nfs4_client *clp)
364 struct rpc_clnt *clnt = clp->cl_callback.cb_client;
366 /* shutdown rpc client, ending any outstanding recall rpcs */
367 if (clnt) {
368 clp->cl_callback.cb_client = NULL;
369 rpc_shutdown_client(clnt);
373 static inline void
374 free_client(struct nfs4_client *clp)
376 shutdown_callback_client(clp);
377 if (clp->cl_cred.cr_group_info)
378 put_group_info(clp->cl_cred.cr_group_info);
379 kfree(clp->cl_name.data);
380 kfree(clp);
383 void
384 put_nfs4_client(struct nfs4_client *clp)
386 if (atomic_dec_and_test(&clp->cl_count))
387 free_client(clp);
390 static void
391 expire_client(struct nfs4_client *clp)
393 struct nfs4_stateowner *sop;
394 struct nfs4_delegation *dp;
395 struct list_head reaplist;
397 dprintk("NFSD: expire_client cl_count %d\n",
398 atomic_read(&clp->cl_count));
400 INIT_LIST_HEAD(&reaplist);
401 spin_lock(&recall_lock);
402 while (!list_empty(&clp->cl_delegations)) {
403 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
404 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
405 dp->dl_flock);
406 list_del_init(&dp->dl_perclnt);
407 list_move(&dp->dl_recall_lru, &reaplist);
409 spin_unlock(&recall_lock);
410 while (!list_empty(&reaplist)) {
411 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
412 list_del_init(&dp->dl_recall_lru);
413 unhash_delegation(dp);
415 list_del(&clp->cl_idhash);
416 list_del(&clp->cl_strhash);
417 list_del(&clp->cl_lru);
418 while (!list_empty(&clp->cl_openowners)) {
419 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
420 release_stateowner(sop);
422 put_nfs4_client(clp);
425 static struct nfs4_client *
426 create_client(struct xdr_netobj name, char *recdir) {
427 struct nfs4_client *clp;
429 if (!(clp = alloc_client(name)))
430 goto out;
431 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
432 atomic_set(&clp->cl_count, 1);
433 atomic_set(&clp->cl_callback.cb_set, 0);
434 INIT_LIST_HEAD(&clp->cl_idhash);
435 INIT_LIST_HEAD(&clp->cl_strhash);
436 INIT_LIST_HEAD(&clp->cl_openowners);
437 INIT_LIST_HEAD(&clp->cl_delegations);
438 INIT_LIST_HEAD(&clp->cl_lru);
439 out:
440 return clp;
443 static void
444 copy_verf(struct nfs4_client *target, nfs4_verifier *source) {
445 memcpy(target->cl_verifier.data, source->data, sizeof(target->cl_verifier.data));
448 static void
449 copy_clid(struct nfs4_client *target, struct nfs4_client *source) {
450 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
451 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
454 static void
455 copy_cred(struct svc_cred *target, struct svc_cred *source) {
457 target->cr_uid = source->cr_uid;
458 target->cr_gid = source->cr_gid;
459 target->cr_group_info = source->cr_group_info;
460 get_group_info(target->cr_group_info);
463 static inline int
464 same_name(const char *n1, const char *n2)
466 return 0 == memcmp(n1, n2, HEXDIR_LEN);
469 static int
470 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
472 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
475 static int
476 same_clid(clientid_t *cl1, clientid_t *cl2)
478 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
481 /* XXX what about NGROUP */
482 static int
483 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
485 return cr1->cr_uid == cr2->cr_uid;
488 static void
489 gen_clid(struct nfs4_client *clp) {
490 clp->cl_clientid.cl_boot = boot_time;
491 clp->cl_clientid.cl_id = current_clientid++;
494 static void
495 gen_confirm(struct nfs4_client *clp) {
496 struct timespec tv;
497 u32 * p;
499 tv = CURRENT_TIME;
500 p = (u32 *)clp->cl_confirm.data;
501 *p++ = tv.tv_sec;
502 *p++ = tv.tv_nsec;
505 static int
506 check_name(struct xdr_netobj name) {
508 if (name.len == 0)
509 return 0;
510 if (name.len > NFS4_OPAQUE_LIMIT) {
511 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
512 return 0;
514 return 1;
517 static void
518 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
520 unsigned int idhashval;
522 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
523 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
524 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
525 list_add_tail(&clp->cl_lru, &client_lru);
526 clp->cl_time = get_seconds();
529 static void
530 move_to_confirmed(struct nfs4_client *clp)
532 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
533 unsigned int strhashval;
535 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
536 list_del_init(&clp->cl_strhash);
537 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
538 strhashval = clientstr_hashval(clp->cl_recdir);
539 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
540 renew_client(clp);
543 static struct nfs4_client *
544 find_confirmed_client(clientid_t *clid)
546 struct nfs4_client *clp;
547 unsigned int idhashval = clientid_hashval(clid->cl_id);
549 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
550 if (same_clid(&clp->cl_clientid, clid))
551 return clp;
553 return NULL;
556 static struct nfs4_client *
557 find_unconfirmed_client(clientid_t *clid)
559 struct nfs4_client *clp;
560 unsigned int idhashval = clientid_hashval(clid->cl_id);
562 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
563 if (same_clid(&clp->cl_clientid, clid))
564 return clp;
566 return NULL;
569 static struct nfs4_client *
570 find_confirmed_client_by_str(const char *dname, unsigned int hashval)
572 struct nfs4_client *clp;
574 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
575 if (same_name(clp->cl_recdir, dname))
576 return clp;
578 return NULL;
581 static struct nfs4_client *
582 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval)
584 struct nfs4_client *clp;
586 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
587 if (same_name(clp->cl_recdir, dname))
588 return clp;
590 return NULL;
593 /* a helper function for parse_callback */
594 static int
595 parse_octet(unsigned int *lenp, char **addrp)
597 unsigned int len = *lenp;
598 char *p = *addrp;
599 int n = -1;
600 char c;
602 for (;;) {
603 if (!len)
604 break;
605 len--;
606 c = *p++;
607 if (c == '.')
608 break;
609 if ((c < '0') || (c > '9')) {
610 n = -1;
611 break;
613 if (n < 0)
614 n = 0;
615 n = (n * 10) + (c - '0');
616 if (n > 255) {
617 n = -1;
618 break;
621 *lenp = len;
622 *addrp = p;
623 return n;
626 /* parse and set the setclientid ipv4 callback address */
627 static int
628 parse_ipv4(unsigned int addr_len, char *addr_val, unsigned int *cbaddrp, unsigned short *cbportp)
630 int temp = 0;
631 u32 cbaddr = 0;
632 u16 cbport = 0;
633 u32 addrlen = addr_len;
634 char *addr = addr_val;
635 int i, shift;
637 /* ipaddress */
638 shift = 24;
639 for(i = 4; i > 0 ; i--) {
640 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
641 return 0;
643 cbaddr |= (temp << shift);
644 if (shift > 0)
645 shift -= 8;
647 *cbaddrp = cbaddr;
649 /* port */
650 shift = 8;
651 for(i = 2; i > 0 ; i--) {
652 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
653 return 0;
655 cbport |= (temp << shift);
656 if (shift > 0)
657 shift -= 8;
659 *cbportp = cbport;
660 return 1;
663 static void
664 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se)
666 struct nfs4_callback *cb = &clp->cl_callback;
668 /* Currently, we only support tcp for the callback channel */
669 if ((se->se_callback_netid_len != 3) || memcmp((char *)se->se_callback_netid_val, "tcp", 3))
670 goto out_err;
672 if ( !(parse_ipv4(se->se_callback_addr_len, se->se_callback_addr_val,
673 &cb->cb_addr, &cb->cb_port)))
674 goto out_err;
675 cb->cb_prog = se->se_callback_prog;
676 cb->cb_ident = se->se_callback_ident;
677 return;
678 out_err:
679 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
680 "will not receive delegations\n",
681 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
683 return;
687 * RFC 3010 has a complex implmentation description of processing a
688 * SETCLIENTID request consisting of 5 bullets, labeled as
689 * CASE0 - CASE4 below.
691 * NOTES:
692 * callback information will be processed in a future patch
694 * an unconfirmed record is added when:
695 * NORMAL (part of CASE 4): there is no confirmed nor unconfirmed record.
696 * CASE 1: confirmed record found with matching name, principal,
697 * verifier, and clientid.
698 * CASE 2: confirmed record found with matching name, principal,
699 * and there is no unconfirmed record with matching
700 * name and principal
702 * an unconfirmed record is replaced when:
703 * CASE 3: confirmed record found with matching name, principal,
704 * and an unconfirmed record is found with matching
705 * name, principal, and with clientid and
706 * confirm that does not match the confirmed record.
707 * CASE 4: there is no confirmed record with matching name and
708 * principal. there is an unconfirmed record with
709 * matching name, principal.
711 * an unconfirmed record is deleted when:
712 * CASE 1: an unconfirmed record that matches input name, verifier,
713 * and confirmed clientid.
714 * CASE 4: any unconfirmed records with matching name and principal
715 * that exist after an unconfirmed record has been replaced
716 * as described above.
719 __be32
720 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
721 struct nfsd4_setclientid *setclid)
723 struct sockaddr_in *sin = svc_addr_in(rqstp);
724 struct xdr_netobj clname = {
725 .len = setclid->se_namelen,
726 .data = setclid->se_name,
728 nfs4_verifier clverifier = setclid->se_verf;
729 unsigned int strhashval;
730 struct nfs4_client *conf, *unconf, *new;
731 __be32 status;
732 char dname[HEXDIR_LEN];
734 if (!check_name(clname))
735 return nfserr_inval;
737 status = nfs4_make_rec_clidname(dname, &clname);
738 if (status)
739 return status;
742 * XXX The Duplicate Request Cache (DRC) has been checked (??)
743 * We get here on a DRC miss.
746 strhashval = clientstr_hashval(dname);
748 nfs4_lock_state();
749 conf = find_confirmed_client_by_str(dname, strhashval);
750 if (conf) {
752 * CASE 0:
753 * clname match, confirmed, different principal
754 * or different ip_address
756 status = nfserr_clid_inuse;
757 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)
758 || conf->cl_addr != sin->sin_addr.s_addr) {
759 dprintk("NFSD: setclientid: string in use by client"
760 "at %u.%u.%u.%u\n", NIPQUAD(conf->cl_addr));
761 goto out;
764 unconf = find_unconfirmed_client_by_str(dname, strhashval);
765 status = nfserr_resource;
766 if (!conf) {
768 * CASE 4:
769 * placed first, because it is the normal case.
771 if (unconf)
772 expire_client(unconf);
773 new = create_client(clname, dname);
774 if (new == NULL)
775 goto out;
776 gen_clid(new);
777 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
779 * CASE 1:
780 * cl_name match, confirmed, principal match
781 * verifier match: probable callback update
783 * remove any unconfirmed nfs4_client with
784 * matching cl_name, cl_verifier, and cl_clientid
786 * create and insert an unconfirmed nfs4_client with same
787 * cl_name, cl_verifier, and cl_clientid as existing
788 * nfs4_client, but with the new callback info and a
789 * new cl_confirm
791 if (unconf) {
792 /* Note this is removing unconfirmed {*x***},
793 * which is stronger than RFC recommended {vxc**}.
794 * This has the advantage that there is at most
795 * one {*x***} in either list at any time.
797 expire_client(unconf);
799 new = create_client(clname, dname);
800 if (new == NULL)
801 goto out;
802 copy_clid(new, conf);
803 } else if (!unconf) {
805 * CASE 2:
806 * clname match, confirmed, principal match
807 * verfier does not match
808 * no unconfirmed. create a new unconfirmed nfs4_client
809 * using input clverifier, clname, and callback info
810 * and generate a new cl_clientid and cl_confirm.
812 new = create_client(clname, dname);
813 if (new == NULL)
814 goto out;
815 gen_clid(new);
816 } else if (!same_verf(&conf->cl_confirm, &unconf->cl_confirm)) {
818 * CASE3:
819 * confirmed found (name, principal match)
820 * confirmed verifier does not match input clverifier
822 * unconfirmed found (name match)
823 * confirmed->cl_confirm != unconfirmed->cl_confirm
825 * remove unconfirmed.
827 * create an unconfirmed nfs4_client
828 * with same cl_name as existing confirmed nfs4_client,
829 * but with new callback info, new cl_clientid,
830 * new cl_verifier and a new cl_confirm
832 expire_client(unconf);
833 new = create_client(clname, dname);
834 if (new == NULL)
835 goto out;
836 gen_clid(new);
837 } else {
838 /* No cases hit !!! */
839 status = nfserr_inval;
840 goto out;
843 copy_verf(new, &clverifier);
844 new->cl_addr = sin->sin_addr.s_addr;
845 copy_cred(&new->cl_cred, &rqstp->rq_cred);
846 gen_confirm(new);
847 gen_callback(new, setclid);
848 add_to_unconfirmed(new, strhashval);
849 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
850 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
851 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
852 status = nfs_ok;
853 out:
854 nfs4_unlock_state();
855 return status;
860 * RFC 3010 has a complex implmentation description of processing a
861 * SETCLIENTID_CONFIRM request consisting of 4 bullets describing
862 * processing on a DRC miss, labeled as CASE1 - CASE4 below.
864 * NOTE: callback information will be processed here in a future patch
866 __be32
867 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
868 struct nfsd4_compound_state *cstate,
869 struct nfsd4_setclientid_confirm *setclientid_confirm)
871 struct sockaddr_in *sin = svc_addr_in(rqstp);
872 struct nfs4_client *conf, *unconf;
873 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
874 clientid_t * clid = &setclientid_confirm->sc_clientid;
875 __be32 status;
877 if (STALE_CLIENTID(clid))
878 return nfserr_stale_clientid;
880 * XXX The Duplicate Request Cache (DRC) has been checked (??)
881 * We get here on a DRC miss.
884 nfs4_lock_state();
886 conf = find_confirmed_client(clid);
887 unconf = find_unconfirmed_client(clid);
889 status = nfserr_clid_inuse;
890 if (conf && conf->cl_addr != sin->sin_addr.s_addr)
891 goto out;
892 if (unconf && unconf->cl_addr != sin->sin_addr.s_addr)
893 goto out;
895 if ((conf && unconf) &&
896 (same_verf(&unconf->cl_confirm, &confirm)) &&
897 (same_verf(&conf->cl_verifier, &unconf->cl_verifier)) &&
898 (same_name(conf->cl_recdir,unconf->cl_recdir)) &&
899 (!same_verf(&conf->cl_confirm, &unconf->cl_confirm))) {
900 /* CASE 1:
901 * unconf record that matches input clientid and input confirm.
902 * conf record that matches input clientid.
903 * conf and unconf records match names, verifiers
905 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
906 status = nfserr_clid_inuse;
907 else {
908 /* XXX: We just turn off callbacks until we can handle
909 * change request correctly. */
910 atomic_set(&conf->cl_callback.cb_set, 0);
911 gen_confirm(conf);
912 nfsd4_remove_clid_dir(unconf);
913 expire_client(unconf);
914 status = nfs_ok;
917 } else if ((conf && !unconf) ||
918 ((conf && unconf) &&
919 (!same_verf(&conf->cl_verifier, &unconf->cl_verifier) ||
920 !same_name(conf->cl_recdir, unconf->cl_recdir)))) {
921 /* CASE 2:
922 * conf record that matches input clientid.
923 * if unconf record matches input clientid, then
924 * unconf->cl_name or unconf->cl_verifier don't match the
925 * conf record.
927 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
928 status = nfserr_clid_inuse;
929 else
930 status = nfs_ok;
931 } else if (!conf && unconf
932 && same_verf(&unconf->cl_confirm, &confirm)) {
933 /* CASE 3:
934 * conf record not found.
935 * unconf record found.
936 * unconf->cl_confirm matches input confirm
938 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
939 status = nfserr_clid_inuse;
940 } else {
941 unsigned int hash =
942 clientstr_hashval(unconf->cl_recdir);
943 conf = find_confirmed_client_by_str(unconf->cl_recdir,
944 hash);
945 if (conf) {
946 nfsd4_remove_clid_dir(conf);
947 expire_client(conf);
949 move_to_confirmed(unconf);
950 conf = unconf;
951 status = nfs_ok;
953 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
954 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
955 &confirm)))) {
956 /* CASE 4:
957 * conf record not found, or if conf, conf->cl_confirm does not
958 * match input confirm.
959 * unconf record not found, or if unconf, unconf->cl_confirm
960 * does not match input confirm.
962 status = nfserr_stale_clientid;
963 } else {
964 /* check that we have hit one of the cases...*/
965 status = nfserr_clid_inuse;
967 out:
968 if (!status)
969 nfsd4_probe_callback(conf);
970 nfs4_unlock_state();
971 return status;
974 /* OPEN Share state helper functions */
975 static inline struct nfs4_file *
976 alloc_init_file(struct inode *ino)
978 struct nfs4_file *fp;
979 unsigned int hashval = file_hashval(ino);
981 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
982 if (fp) {
983 kref_init(&fp->fi_ref);
984 INIT_LIST_HEAD(&fp->fi_hash);
985 INIT_LIST_HEAD(&fp->fi_stateids);
986 INIT_LIST_HEAD(&fp->fi_delegations);
987 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
988 fp->fi_inode = igrab(ino);
989 fp->fi_id = current_fileid++;
990 fp->fi_had_conflict = false;
991 return fp;
993 return NULL;
996 static void
997 nfsd4_free_slab(struct kmem_cache **slab)
999 if (*slab == NULL)
1000 return;
1001 kmem_cache_destroy(*slab);
1002 *slab = NULL;
1005 void
1006 nfsd4_free_slabs(void)
1008 nfsd4_free_slab(&stateowner_slab);
1009 nfsd4_free_slab(&file_slab);
1010 nfsd4_free_slab(&stateid_slab);
1011 nfsd4_free_slab(&deleg_slab);
1014 static int
1015 nfsd4_init_slabs(void)
1017 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
1018 sizeof(struct nfs4_stateowner), 0, 0, NULL);
1019 if (stateowner_slab == NULL)
1020 goto out_nomem;
1021 file_slab = kmem_cache_create("nfsd4_files",
1022 sizeof(struct nfs4_file), 0, 0, NULL);
1023 if (file_slab == NULL)
1024 goto out_nomem;
1025 stateid_slab = kmem_cache_create("nfsd4_stateids",
1026 sizeof(struct nfs4_stateid), 0, 0, NULL);
1027 if (stateid_slab == NULL)
1028 goto out_nomem;
1029 deleg_slab = kmem_cache_create("nfsd4_delegations",
1030 sizeof(struct nfs4_delegation), 0, 0, NULL);
1031 if (deleg_slab == NULL)
1032 goto out_nomem;
1033 return 0;
1034 out_nomem:
1035 nfsd4_free_slabs();
1036 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1037 return -ENOMEM;
1040 void
1041 nfs4_free_stateowner(struct kref *kref)
1043 struct nfs4_stateowner *sop =
1044 container_of(kref, struct nfs4_stateowner, so_ref);
1045 kfree(sop->so_owner.data);
1046 kmem_cache_free(stateowner_slab, sop);
1049 static inline struct nfs4_stateowner *
1050 alloc_stateowner(struct xdr_netobj *owner)
1052 struct nfs4_stateowner *sop;
1054 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1055 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1056 memcpy(sop->so_owner.data, owner->data, owner->len);
1057 sop->so_owner.len = owner->len;
1058 kref_init(&sop->so_ref);
1059 return sop;
1061 kmem_cache_free(stateowner_slab, sop);
1063 return NULL;
1066 static struct nfs4_stateowner *
1067 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1068 struct nfs4_stateowner *sop;
1069 struct nfs4_replay *rp;
1070 unsigned int idhashval;
1072 if (!(sop = alloc_stateowner(&open->op_owner)))
1073 return NULL;
1074 idhashval = ownerid_hashval(current_ownerid);
1075 INIT_LIST_HEAD(&sop->so_idhash);
1076 INIT_LIST_HEAD(&sop->so_strhash);
1077 INIT_LIST_HEAD(&sop->so_perclient);
1078 INIT_LIST_HEAD(&sop->so_stateids);
1079 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1080 INIT_LIST_HEAD(&sop->so_close_lru);
1081 sop->so_time = 0;
1082 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1083 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
1084 list_add(&sop->so_perclient, &clp->cl_openowners);
1085 sop->so_is_open_owner = 1;
1086 sop->so_id = current_ownerid++;
1087 sop->so_client = clp;
1088 sop->so_seqid = open->op_seqid;
1089 sop->so_confirmed = 0;
1090 rp = &sop->so_replay;
1091 rp->rp_status = nfserr_serverfault;
1092 rp->rp_buflen = 0;
1093 rp->rp_buf = rp->rp_ibuf;
1094 return sop;
1097 static void
1098 release_stateid_lockowners(struct nfs4_stateid *open_stp)
1100 struct nfs4_stateowner *lock_sop;
1102 while (!list_empty(&open_stp->st_lockowners)) {
1103 lock_sop = list_entry(open_stp->st_lockowners.next,
1104 struct nfs4_stateowner, so_perstateid);
1105 /* list_del(&open_stp->st_lockowners); */
1106 BUG_ON(lock_sop->so_is_open_owner);
1107 release_stateowner(lock_sop);
1111 static void
1112 unhash_stateowner(struct nfs4_stateowner *sop)
1114 struct nfs4_stateid *stp;
1116 list_del(&sop->so_idhash);
1117 list_del(&sop->so_strhash);
1118 if (sop->so_is_open_owner)
1119 list_del(&sop->so_perclient);
1120 list_del(&sop->so_perstateid);
1121 while (!list_empty(&sop->so_stateids)) {
1122 stp = list_entry(sop->so_stateids.next,
1123 struct nfs4_stateid, st_perstateowner);
1124 if (sop->so_is_open_owner)
1125 release_stateid(stp, OPEN_STATE);
1126 else
1127 release_stateid(stp, LOCK_STATE);
1131 static void
1132 release_stateowner(struct nfs4_stateowner *sop)
1134 unhash_stateowner(sop);
1135 list_del(&sop->so_close_lru);
1136 nfs4_put_stateowner(sop);
1139 static inline void
1140 init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1141 struct nfs4_stateowner *sop = open->op_stateowner;
1142 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1144 INIT_LIST_HEAD(&stp->st_hash);
1145 INIT_LIST_HEAD(&stp->st_perstateowner);
1146 INIT_LIST_HEAD(&stp->st_lockowners);
1147 INIT_LIST_HEAD(&stp->st_perfile);
1148 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
1149 list_add(&stp->st_perstateowner, &sop->so_stateids);
1150 list_add(&stp->st_perfile, &fp->fi_stateids);
1151 stp->st_stateowner = sop;
1152 get_nfs4_file(fp);
1153 stp->st_file = fp;
1154 stp->st_stateid.si_boot = boot_time;
1155 stp->st_stateid.si_stateownerid = sop->so_id;
1156 stp->st_stateid.si_fileid = fp->fi_id;
1157 stp->st_stateid.si_generation = 0;
1158 stp->st_access_bmap = 0;
1159 stp->st_deny_bmap = 0;
1160 __set_bit(open->op_share_access, &stp->st_access_bmap);
1161 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1162 stp->st_openstp = NULL;
1165 static void
1166 release_stateid(struct nfs4_stateid *stp, int flags)
1168 struct file *filp = stp->st_vfs_file;
1170 list_del(&stp->st_hash);
1171 list_del(&stp->st_perfile);
1172 list_del(&stp->st_perstateowner);
1173 if (flags & OPEN_STATE) {
1174 release_stateid_lockowners(stp);
1175 stp->st_vfs_file = NULL;
1176 nfsd_close(filp);
1177 } else if (flags & LOCK_STATE)
1178 locks_remove_posix(filp, (fl_owner_t) stp->st_stateowner);
1179 put_nfs4_file(stp->st_file);
1180 kmem_cache_free(stateid_slab, stp);
1183 static void
1184 move_to_close_lru(struct nfs4_stateowner *sop)
1186 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1188 list_move_tail(&sop->so_close_lru, &close_lru);
1189 sop->so_time = get_seconds();
1192 static int
1193 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1194 clientid_t *clid)
1196 return (sop->so_owner.len == owner->len) &&
1197 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1198 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1201 static struct nfs4_stateowner *
1202 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1204 struct nfs4_stateowner *so = NULL;
1206 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
1207 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1208 return so;
1210 return NULL;
1213 /* search file_hashtbl[] for file */
1214 static struct nfs4_file *
1215 find_file(struct inode *ino)
1217 unsigned int hashval = file_hashval(ino);
1218 struct nfs4_file *fp;
1220 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
1221 if (fp->fi_inode == ino) {
1222 get_nfs4_file(fp);
1223 return fp;
1226 return NULL;
1229 static int access_valid(u32 x)
1231 return (x > 0 && x < 4);
1234 static int deny_valid(u32 x)
1236 return (x >= 0 && x < 5);
1239 static void
1240 set_access(unsigned int *access, unsigned long bmap) {
1241 int i;
1243 *access = 0;
1244 for (i = 1; i < 4; i++) {
1245 if (test_bit(i, &bmap))
1246 *access |= i;
1250 static void
1251 set_deny(unsigned int *deny, unsigned long bmap) {
1252 int i;
1254 *deny = 0;
1255 for (i = 0; i < 4; i++) {
1256 if (test_bit(i, &bmap))
1257 *deny |= i ;
1261 static int
1262 test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1263 unsigned int access, deny;
1265 set_access(&access, stp->st_access_bmap);
1266 set_deny(&deny, stp->st_deny_bmap);
1267 if ((access & open->op_share_deny) || (deny & open->op_share_access))
1268 return 0;
1269 return 1;
1273 * Called to check deny when READ with all zero stateid or
1274 * WRITE with all zero or all one stateid
1276 static __be32
1277 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
1279 struct inode *ino = current_fh->fh_dentry->d_inode;
1280 struct nfs4_file *fp;
1281 struct nfs4_stateid *stp;
1282 __be32 ret;
1284 dprintk("NFSD: nfs4_share_conflict\n");
1286 fp = find_file(ino);
1287 if (!fp)
1288 return nfs_ok;
1289 ret = nfserr_locked;
1290 /* Search for conflicting share reservations */
1291 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
1292 if (test_bit(deny_type, &stp->st_deny_bmap) ||
1293 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
1294 goto out;
1296 ret = nfs_ok;
1297 out:
1298 put_nfs4_file(fp);
1299 return ret;
1302 static inline void
1303 nfs4_file_downgrade(struct file *filp, unsigned int share_access)
1305 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
1306 put_write_access(filp->f_path.dentry->d_inode);
1307 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
1312 * Recall a delegation
1314 static int
1315 do_recall(void *__dp)
1317 struct nfs4_delegation *dp = __dp;
1319 dp->dl_file->fi_had_conflict = true;
1320 nfsd4_cb_recall(dp);
1321 return 0;
1325 * Spawn a thread to perform a recall on the delegation represented
1326 * by the lease (file_lock)
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.
1332 static
1333 void nfsd_break_deleg_cb(struct file_lock *fl)
1335 struct nfs4_delegation *dp= (struct nfs4_delegation *)fl->fl_owner;
1336 struct task_struct *t;
1338 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
1339 if (!dp)
1340 return;
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 atomic_inc(&dp->dl_client->cl_count);
1350 spin_lock(&recall_lock);
1351 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
1352 spin_unlock(&recall_lock);
1354 /* only place dl_time is set. protected by lock_kernel*/
1355 dp->dl_time = get_seconds();
1358 * We don't want the locks code to timeout the lease for us;
1359 * we'll remove it ourself if the delegation isn't returned
1360 * in time.
1362 fl->fl_break_time = 0;
1364 t = kthread_run(do_recall, dp, "%s", "nfs4_cb_recall");
1365 if (IS_ERR(t)) {
1366 struct nfs4_client *clp = dp->dl_client;
1368 printk(KERN_INFO "NFSD: Callback thread failed for "
1369 "for client (clientid %08x/%08x)\n",
1370 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1371 put_nfs4_client(dp->dl_client);
1372 nfs4_put_delegation(dp);
1377 * The file_lock is being reapd.
1379 * Called by locks_free_lock() with lock_kernel() held.
1381 static
1382 void nfsd_release_deleg_cb(struct file_lock *fl)
1384 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1386 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
1388 if (!(fl->fl_flags & FL_LEASE) || !dp)
1389 return;
1390 dp->dl_flock = NULL;
1394 * Set the delegation file_lock back pointer.
1396 * Called from setlease() with lock_kernel() held.
1398 static
1399 void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
1401 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
1403 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
1404 if (!dp)
1405 return;
1406 dp->dl_flock = new;
1410 * Called from setlease() with lock_kernel() held
1412 static
1413 int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
1415 struct nfs4_delegation *onlistd =
1416 (struct nfs4_delegation *)onlist->fl_owner;
1417 struct nfs4_delegation *tryd =
1418 (struct nfs4_delegation *)try->fl_owner;
1420 if (onlist->fl_lmops != try->fl_lmops)
1421 return 0;
1423 return onlistd->dl_client == tryd->dl_client;
1427 static
1428 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
1430 if (arg & F_UNLCK)
1431 return lease_modify(onlist, arg);
1432 else
1433 return -EAGAIN;
1436 static struct lock_manager_operations nfsd_lease_mng_ops = {
1437 .fl_break = nfsd_break_deleg_cb,
1438 .fl_release_private = nfsd_release_deleg_cb,
1439 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
1440 .fl_mylease = nfsd_same_client_deleg_cb,
1441 .fl_change = nfsd_change_deleg_cb,
1445 __be32
1446 nfsd4_process_open1(struct nfsd4_open *open)
1448 clientid_t *clientid = &open->op_clientid;
1449 struct nfs4_client *clp = NULL;
1450 unsigned int strhashval;
1451 struct nfs4_stateowner *sop = NULL;
1453 if (!check_name(open->op_owner))
1454 return nfserr_inval;
1456 if (STALE_CLIENTID(&open->op_clientid))
1457 return nfserr_stale_clientid;
1459 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
1460 sop = find_openstateowner_str(strhashval, open);
1461 open->op_stateowner = sop;
1462 if (!sop) {
1463 /* Make sure the client's lease hasn't expired. */
1464 clp = find_confirmed_client(clientid);
1465 if (clp == NULL)
1466 return nfserr_expired;
1467 goto renew;
1469 if (!sop->so_confirmed) {
1470 /* Replace unconfirmed owners without checking for replay. */
1471 clp = sop->so_client;
1472 release_stateowner(sop);
1473 open->op_stateowner = NULL;
1474 goto renew;
1476 if (open->op_seqid == sop->so_seqid - 1) {
1477 if (sop->so_replay.rp_buflen)
1478 return nfserr_replay_me;
1479 /* The original OPEN failed so spectacularly
1480 * that we don't even have replay data saved!
1481 * Therefore, we have no choice but to continue
1482 * processing this OPEN; presumably, we'll
1483 * fail again for the same reason.
1485 dprintk("nfsd4_process_open1: replay with no replay cache\n");
1486 goto renew;
1488 if (open->op_seqid != sop->so_seqid)
1489 return nfserr_bad_seqid;
1490 renew:
1491 if (open->op_stateowner == NULL) {
1492 sop = alloc_init_open_stateowner(strhashval, clp, open);
1493 if (sop == NULL)
1494 return nfserr_resource;
1495 open->op_stateowner = sop;
1497 list_del_init(&sop->so_close_lru);
1498 renew_client(sop->so_client);
1499 return nfs_ok;
1502 static inline __be32
1503 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
1505 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
1506 return nfserr_openmode;
1507 else
1508 return nfs_ok;
1511 static struct nfs4_delegation *
1512 find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
1514 struct nfs4_delegation *dp;
1516 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
1517 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
1518 return dp;
1520 return NULL;
1523 static __be32
1524 nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
1525 struct nfs4_delegation **dp)
1527 int flags;
1528 __be32 status = nfserr_bad_stateid;
1530 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
1531 if (*dp == NULL)
1532 goto out;
1533 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
1534 RD_STATE : WR_STATE;
1535 status = nfs4_check_delegmode(*dp, flags);
1536 if (status)
1537 *dp = NULL;
1538 out:
1539 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
1540 return nfs_ok;
1541 if (status)
1542 return status;
1543 open->op_stateowner->so_confirmed = 1;
1544 return nfs_ok;
1547 static __be32
1548 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
1550 struct nfs4_stateid *local;
1551 __be32 status = nfserr_share_denied;
1552 struct nfs4_stateowner *sop = open->op_stateowner;
1554 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1555 /* ignore lock owners */
1556 if (local->st_stateowner->so_is_open_owner == 0)
1557 continue;
1558 /* remember if we have seen this open owner */
1559 if (local->st_stateowner == sop)
1560 *stpp = local;
1561 /* check for conflicting share reservations */
1562 if (!test_share(local, open))
1563 goto out;
1565 status = 0;
1566 out:
1567 return status;
1570 static inline struct nfs4_stateid *
1571 nfs4_alloc_stateid(void)
1573 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
1576 static __be32
1577 nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
1578 struct nfs4_delegation *dp,
1579 struct svc_fh *cur_fh, int flags)
1581 struct nfs4_stateid *stp;
1583 stp = nfs4_alloc_stateid();
1584 if (stp == NULL)
1585 return nfserr_resource;
1587 if (dp) {
1588 get_file(dp->dl_vfs_file);
1589 stp->st_vfs_file = dp->dl_vfs_file;
1590 } else {
1591 __be32 status;
1592 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
1593 &stp->st_vfs_file);
1594 if (status) {
1595 if (status == nfserr_dropit)
1596 status = nfserr_jukebox;
1597 kmem_cache_free(stateid_slab, stp);
1598 return status;
1601 *stpp = stp;
1602 return 0;
1605 static inline __be32
1606 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
1607 struct nfsd4_open *open)
1609 struct iattr iattr = {
1610 .ia_valid = ATTR_SIZE,
1611 .ia_size = 0,
1613 if (!open->op_truncate)
1614 return 0;
1615 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
1616 return nfserr_inval;
1617 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
1620 static __be32
1621 nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
1623 struct file *filp = stp->st_vfs_file;
1624 struct inode *inode = filp->f_path.dentry->d_inode;
1625 unsigned int share_access, new_writer;
1626 __be32 status;
1628 set_access(&share_access, stp->st_access_bmap);
1629 new_writer = (~share_access) & open->op_share_access
1630 & NFS4_SHARE_ACCESS_WRITE;
1632 if (new_writer) {
1633 int err = get_write_access(inode);
1634 if (err)
1635 return nfserrno(err);
1637 status = nfsd4_truncate(rqstp, cur_fh, open);
1638 if (status) {
1639 if (new_writer)
1640 put_write_access(inode);
1641 return status;
1643 /* remember the open */
1644 filp->f_mode |= open->op_share_access;
1645 set_bit(open->op_share_access, &stp->st_access_bmap);
1646 set_bit(open->op_share_deny, &stp->st_deny_bmap);
1648 return nfs_ok;
1652 static void
1653 nfs4_set_claim_prev(struct nfsd4_open *open)
1655 open->op_stateowner->so_confirmed = 1;
1656 open->op_stateowner->so_client->cl_firststate = 1;
1660 * Attempt to hand out a delegation.
1662 static void
1663 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
1665 struct nfs4_delegation *dp;
1666 struct nfs4_stateowner *sop = stp->st_stateowner;
1667 struct nfs4_callback *cb = &sop->so_client->cl_callback;
1668 struct file_lock fl, *flp = &fl;
1669 int status, flag = 0;
1671 flag = NFS4_OPEN_DELEGATE_NONE;
1672 open->op_recall = 0;
1673 switch (open->op_claim_type) {
1674 case NFS4_OPEN_CLAIM_PREVIOUS:
1675 if (!atomic_read(&cb->cb_set))
1676 open->op_recall = 1;
1677 flag = open->op_delegate_type;
1678 if (flag == NFS4_OPEN_DELEGATE_NONE)
1679 goto out;
1680 break;
1681 case NFS4_OPEN_CLAIM_NULL:
1682 /* Let's not give out any delegations till everyone's
1683 * had the chance to reclaim theirs.... */
1684 if (nfs4_in_grace())
1685 goto out;
1686 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
1687 goto out;
1688 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
1689 flag = NFS4_OPEN_DELEGATE_WRITE;
1690 else
1691 flag = NFS4_OPEN_DELEGATE_READ;
1692 break;
1693 default:
1694 goto out;
1697 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
1698 if (dp == NULL) {
1699 flag = NFS4_OPEN_DELEGATE_NONE;
1700 goto out;
1702 locks_init_lock(&fl);
1703 fl.fl_lmops = &nfsd_lease_mng_ops;
1704 fl.fl_flags = FL_LEASE;
1705 fl.fl_end = OFFSET_MAX;
1706 fl.fl_owner = (fl_owner_t)dp;
1707 fl.fl_file = stp->st_vfs_file;
1708 fl.fl_pid = current->tgid;
1710 /* vfs_setlease checks to see if delegation should be handed out.
1711 * the lock_manager callbacks fl_mylease and fl_change are used
1713 if ((status = vfs_setlease(stp->st_vfs_file,
1714 flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK, &flp))) {
1715 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
1716 unhash_delegation(dp);
1717 flag = NFS4_OPEN_DELEGATE_NONE;
1718 goto out;
1721 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
1723 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
1724 dp->dl_stateid.si_boot,
1725 dp->dl_stateid.si_stateownerid,
1726 dp->dl_stateid.si_fileid,
1727 dp->dl_stateid.si_generation);
1728 out:
1729 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
1730 && flag == NFS4_OPEN_DELEGATE_NONE
1731 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
1732 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1733 open->op_delegate_type = flag;
1737 * called with nfs4_lock_state() held.
1739 __be32
1740 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
1742 struct nfs4_file *fp = NULL;
1743 struct inode *ino = current_fh->fh_dentry->d_inode;
1744 struct nfs4_stateid *stp = NULL;
1745 struct nfs4_delegation *dp = NULL;
1746 __be32 status;
1748 status = nfserr_inval;
1749 if (!access_valid(open->op_share_access)
1750 || !deny_valid(open->op_share_deny))
1751 goto out;
1753 * Lookup file; if found, lookup stateid and check open request,
1754 * and check for delegations in the process of being recalled.
1755 * If not found, create the nfs4_file struct
1757 fp = find_file(ino);
1758 if (fp) {
1759 if ((status = nfs4_check_open(fp, open, &stp)))
1760 goto out;
1761 status = nfs4_check_deleg(fp, open, &dp);
1762 if (status)
1763 goto out;
1764 } else {
1765 status = nfserr_bad_stateid;
1766 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
1767 goto out;
1768 status = nfserr_resource;
1769 fp = alloc_init_file(ino);
1770 if (fp == NULL)
1771 goto out;
1775 * OPEN the file, or upgrade an existing OPEN.
1776 * If truncate fails, the OPEN fails.
1778 if (stp) {
1779 /* Stateid was found, this is an OPEN upgrade */
1780 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
1781 if (status)
1782 goto out;
1783 update_stateid(&stp->st_stateid);
1784 } else {
1785 /* Stateid was not found, this is a new OPEN */
1786 int flags = 0;
1787 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
1788 flags |= MAY_READ;
1789 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
1790 flags |= MAY_WRITE;
1791 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
1792 if (status)
1793 goto out;
1794 init_stateid(stp, fp, open);
1795 status = nfsd4_truncate(rqstp, current_fh, open);
1796 if (status) {
1797 release_stateid(stp, OPEN_STATE);
1798 goto out;
1801 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
1804 * Attempt to hand out a delegation. No error return, because the
1805 * OPEN succeeds even if we fail.
1807 nfs4_open_delegation(current_fh, open, stp);
1809 status = nfs_ok;
1811 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
1812 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
1813 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
1814 out:
1815 if (fp)
1816 put_nfs4_file(fp);
1817 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
1818 nfs4_set_claim_prev(open);
1820 * To finish the open response, we just need to set the rflags.
1822 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
1823 if (!open->op_stateowner->so_confirmed)
1824 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
1826 return status;
1829 static struct workqueue_struct *laundry_wq;
1830 static void laundromat_main(struct work_struct *);
1831 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
1833 __be32
1834 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1835 clientid_t *clid)
1837 struct nfs4_client *clp;
1838 __be32 status;
1840 nfs4_lock_state();
1841 dprintk("process_renew(%08x/%08x): starting\n",
1842 clid->cl_boot, clid->cl_id);
1843 status = nfserr_stale_clientid;
1844 if (STALE_CLIENTID(clid))
1845 goto out;
1846 clp = find_confirmed_client(clid);
1847 status = nfserr_expired;
1848 if (clp == NULL) {
1849 /* We assume the client took too long to RENEW. */
1850 dprintk("nfsd4_renew: clientid not found!\n");
1851 goto out;
1853 renew_client(clp);
1854 status = nfserr_cb_path_down;
1855 if (!list_empty(&clp->cl_delegations)
1856 && !atomic_read(&clp->cl_callback.cb_set))
1857 goto out;
1858 status = nfs_ok;
1859 out:
1860 nfs4_unlock_state();
1861 return status;
1864 static void
1865 end_grace(void)
1867 dprintk("NFSD: end of grace period\n");
1868 nfsd4_recdir_purge_old();
1869 in_grace = 0;
1872 static time_t
1873 nfs4_laundromat(void)
1875 struct nfs4_client *clp;
1876 struct nfs4_stateowner *sop;
1877 struct nfs4_delegation *dp;
1878 struct list_head *pos, *next, reaplist;
1879 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
1880 time_t t, clientid_val = NFSD_LEASE_TIME;
1881 time_t u, test_val = NFSD_LEASE_TIME;
1883 nfs4_lock_state();
1885 dprintk("NFSD: laundromat service - starting\n");
1886 if (in_grace)
1887 end_grace();
1888 list_for_each_safe(pos, next, &client_lru) {
1889 clp = list_entry(pos, struct nfs4_client, cl_lru);
1890 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
1891 t = clp->cl_time - cutoff;
1892 if (clientid_val > t)
1893 clientid_val = t;
1894 break;
1896 dprintk("NFSD: purging unused client (clientid %08x)\n",
1897 clp->cl_clientid.cl_id);
1898 nfsd4_remove_clid_dir(clp);
1899 expire_client(clp);
1901 INIT_LIST_HEAD(&reaplist);
1902 spin_lock(&recall_lock);
1903 list_for_each_safe(pos, next, &del_recall_lru) {
1904 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
1905 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
1906 u = dp->dl_time - cutoff;
1907 if (test_val > u)
1908 test_val = u;
1909 break;
1911 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
1912 dp, dp->dl_flock);
1913 list_move(&dp->dl_recall_lru, &reaplist);
1915 spin_unlock(&recall_lock);
1916 list_for_each_safe(pos, next, &reaplist) {
1917 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
1918 list_del_init(&dp->dl_recall_lru);
1919 unhash_delegation(dp);
1921 test_val = NFSD_LEASE_TIME;
1922 list_for_each_safe(pos, next, &close_lru) {
1923 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
1924 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
1925 u = sop->so_time - cutoff;
1926 if (test_val > u)
1927 test_val = u;
1928 break;
1930 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
1931 sop->so_id);
1932 release_stateowner(sop);
1934 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
1935 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
1936 nfs4_unlock_state();
1937 return clientid_val;
1940 void
1941 laundromat_main(struct work_struct *not_used)
1943 time_t t;
1945 t = nfs4_laundromat();
1946 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
1947 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1950 static struct nfs4_stateowner *
1951 search_close_lru(u32 st_id, int flags)
1953 struct nfs4_stateowner *local = NULL;
1955 if (flags & CLOSE_STATE) {
1956 list_for_each_entry(local, &close_lru, so_close_lru) {
1957 if (local->so_id == st_id)
1958 return local;
1961 return NULL;
1964 static inline int
1965 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
1967 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1970 static int
1971 STALE_STATEID(stateid_t *stateid)
1973 if (stateid->si_boot == boot_time)
1974 return 0;
1975 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
1976 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
1977 stateid->si_generation);
1978 return 1;
1981 static inline int
1982 access_permit_read(unsigned long access_bmap)
1984 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
1985 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
1986 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
1989 static inline int
1990 access_permit_write(unsigned long access_bmap)
1992 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
1993 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
1996 static
1997 __be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1999 __be32 status = nfserr_openmode;
2001 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2002 goto out;
2003 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2004 goto out;
2005 status = nfs_ok;
2006 out:
2007 return status;
2010 static inline __be32
2011 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2013 /* Trying to call delegreturn with a special stateid? Yuch: */
2014 if (!(flags & (RD_STATE | WR_STATE)))
2015 return nfserr_bad_stateid;
2016 else if (ONE_STATEID(stateid) && (flags & RD_STATE))
2017 return nfs_ok;
2018 else if (nfs4_in_grace()) {
2019 /* Answer in remaining cases depends on existance of
2020 * conflicting state; so we must wait out the grace period. */
2021 return nfserr_grace;
2022 } else if (flags & WR_STATE)
2023 return nfs4_share_conflict(current_fh,
2024 NFS4_SHARE_DENY_WRITE);
2025 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2026 return nfs4_share_conflict(current_fh,
2027 NFS4_SHARE_DENY_READ);
2031 * Allow READ/WRITE during grace period on recovered state only for files
2032 * that are not able to provide mandatory locking.
2034 static inline int
2035 io_during_grace_disallowed(struct inode *inode, int flags)
2037 return nfs4_in_grace() && (flags & (RD_STATE | WR_STATE))
2038 && mandatory_lock(inode);
2042 * Checks for stateid operations
2044 __be32
2045 nfs4_preprocess_stateid_op(struct svc_fh *current_fh, stateid_t *stateid, int flags, struct file **filpp)
2047 struct nfs4_stateid *stp = NULL;
2048 struct nfs4_delegation *dp = NULL;
2049 stateid_t *stidp;
2050 struct inode *ino = current_fh->fh_dentry->d_inode;
2051 __be32 status;
2053 dprintk("NFSD: preprocess_stateid_op: stateid = (%08x/%08x/%08x/%08x)\n",
2054 stateid->si_boot, stateid->si_stateownerid,
2055 stateid->si_fileid, stateid->si_generation);
2056 if (filpp)
2057 *filpp = NULL;
2059 if (io_during_grace_disallowed(ino, flags))
2060 return nfserr_grace;
2062 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2063 return check_special_stateids(current_fh, stateid, flags);
2065 /* STALE STATEID */
2066 status = nfserr_stale_stateid;
2067 if (STALE_STATEID(stateid))
2068 goto out;
2070 /* BAD STATEID */
2071 status = nfserr_bad_stateid;
2072 if (!stateid->si_fileid) { /* delegation stateid */
2073 if(!(dp = find_delegation_stateid(ino, stateid))) {
2074 dprintk("NFSD: delegation stateid not found\n");
2075 goto out;
2077 stidp = &dp->dl_stateid;
2078 } else { /* open or lock stateid */
2079 if (!(stp = find_stateid(stateid, flags))) {
2080 dprintk("NFSD: open or lock stateid not found\n");
2081 goto out;
2083 if ((flags & CHECK_FH) && nfs4_check_fh(current_fh, stp))
2084 goto out;
2085 if (!stp->st_stateowner->so_confirmed)
2086 goto out;
2087 stidp = &stp->st_stateid;
2089 if (stateid->si_generation > stidp->si_generation)
2090 goto out;
2092 /* OLD STATEID */
2093 status = nfserr_old_stateid;
2094 if (stateid->si_generation < stidp->si_generation)
2095 goto out;
2096 if (stp) {
2097 if ((status = nfs4_check_openmode(stp,flags)))
2098 goto out;
2099 renew_client(stp->st_stateowner->so_client);
2100 if (filpp)
2101 *filpp = stp->st_vfs_file;
2102 } else if (dp) {
2103 if ((status = nfs4_check_delegmode(dp, flags)))
2104 goto out;
2105 renew_client(dp->dl_client);
2106 if (flags & DELEG_RET)
2107 unhash_delegation(dp);
2108 if (filpp)
2109 *filpp = dp->dl_vfs_file;
2111 status = nfs_ok;
2112 out:
2113 return status;
2116 static inline int
2117 setlkflg (int type)
2119 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2120 RD_STATE : WR_STATE;
2124 * Checks for sequence id mutating operations.
2126 static __be32
2127 nfs4_preprocess_seqid_op(struct svc_fh *current_fh, u32 seqid, stateid_t *stateid, int flags, struct nfs4_stateowner **sopp, struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
2129 struct nfs4_stateid *stp;
2130 struct nfs4_stateowner *sop;
2132 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2133 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2134 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2135 stateid->si_generation);
2137 *stpp = NULL;
2138 *sopp = NULL;
2140 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2141 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
2142 return nfserr_bad_stateid;
2145 if (STALE_STATEID(stateid))
2146 return nfserr_stale_stateid;
2148 * We return BAD_STATEID if filehandle doesn't match stateid,
2149 * the confirmed flag is incorrecly set, or the generation
2150 * number is incorrect.
2152 stp = find_stateid(stateid, flags);
2153 if (stp == NULL) {
2155 * Also, we should make sure this isn't just the result of
2156 * a replayed close:
2158 sop = search_close_lru(stateid->si_stateownerid, flags);
2159 if (sop == NULL)
2160 return nfserr_bad_stateid;
2161 *sopp = sop;
2162 goto check_replay;
2165 if (lock) {
2166 struct nfs4_stateowner *sop = stp->st_stateowner;
2167 clientid_t *lockclid = &lock->v.new.clientid;
2168 struct nfs4_client *clp = sop->so_client;
2169 int lkflg = 0;
2170 __be32 status;
2172 lkflg = setlkflg(lock->lk_type);
2174 if (lock->lk_is_new) {
2175 if (!sop->so_is_open_owner)
2176 return nfserr_bad_stateid;
2177 if (!same_clid(&clp->cl_clientid, lockclid))
2178 return nfserr_bad_stateid;
2179 /* stp is the open stateid */
2180 status = nfs4_check_openmode(stp, lkflg);
2181 if (status)
2182 return status;
2183 } else {
2184 /* stp is the lock stateid */
2185 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2186 if (status)
2187 return status;
2191 if ((flags & CHECK_FH) && nfs4_check_fh(current_fh, stp)) {
2192 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
2193 return nfserr_bad_stateid;
2196 *stpp = stp;
2197 *sopp = sop = stp->st_stateowner;
2200 * We now validate the seqid and stateid generation numbers.
2201 * For the moment, we ignore the possibility of
2202 * generation number wraparound.
2204 if (seqid != sop->so_seqid)
2205 goto check_replay;
2207 if (sop->so_confirmed && flags & CONFIRM) {
2208 dprintk("NFSD: preprocess_seqid_op: expected"
2209 " unconfirmed stateowner!\n");
2210 return nfserr_bad_stateid;
2212 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2213 dprintk("NFSD: preprocess_seqid_op: stateowner not"
2214 " confirmed yet!\n");
2215 return nfserr_bad_stateid;
2217 if (stateid->si_generation > stp->st_stateid.si_generation) {
2218 dprintk("NFSD: preprocess_seqid_op: future stateid?!\n");
2219 return nfserr_bad_stateid;
2222 if (stateid->si_generation < stp->st_stateid.si_generation) {
2223 dprintk("NFSD: preprocess_seqid_op: old stateid!\n");
2224 return nfserr_old_stateid;
2226 renew_client(sop->so_client);
2227 return nfs_ok;
2229 check_replay:
2230 if (seqid == sop->so_seqid - 1) {
2231 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
2232 /* indicate replay to calling function */
2233 return nfserr_replay_me;
2235 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
2236 sop->so_seqid, seqid);
2237 *sopp = NULL;
2238 return nfserr_bad_seqid;
2241 __be32
2242 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2243 struct nfsd4_open_confirm *oc)
2245 __be32 status;
2246 struct nfs4_stateowner *sop;
2247 struct nfs4_stateid *stp;
2249 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
2250 (int)cstate->current_fh.fh_dentry->d_name.len,
2251 cstate->current_fh.fh_dentry->d_name.name);
2253 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
2254 if (status)
2255 return status;
2257 nfs4_lock_state();
2259 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2260 oc->oc_seqid, &oc->oc_req_stateid,
2261 CHECK_FH | CONFIRM | OPEN_STATE,
2262 &oc->oc_stateowner, &stp, NULL)))
2263 goto out;
2265 sop = oc->oc_stateowner;
2266 sop->so_confirmed = 1;
2267 update_stateid(&stp->st_stateid);
2268 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
2269 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
2270 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
2271 stp->st_stateid.si_boot,
2272 stp->st_stateid.si_stateownerid,
2273 stp->st_stateid.si_fileid,
2274 stp->st_stateid.si_generation);
2276 nfsd4_create_clid_dir(sop->so_client);
2277 out:
2278 if (oc->oc_stateowner) {
2279 nfs4_get_stateowner(oc->oc_stateowner);
2280 cstate->replay_owner = oc->oc_stateowner;
2282 nfs4_unlock_state();
2283 return status;
2288 * unset all bits in union bitmap (bmap) that
2289 * do not exist in share (from successful OPEN_DOWNGRADE)
2291 static void
2292 reset_union_bmap_access(unsigned long access, unsigned long *bmap)
2294 int i;
2295 for (i = 1; i < 4; i++) {
2296 if ((i & access) != i)
2297 __clear_bit(i, bmap);
2301 static void
2302 reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
2304 int i;
2305 for (i = 0; i < 4; i++) {
2306 if ((i & deny) != i)
2307 __clear_bit(i, bmap);
2311 __be32
2312 nfsd4_open_downgrade(struct svc_rqst *rqstp,
2313 struct nfsd4_compound_state *cstate,
2314 struct nfsd4_open_downgrade *od)
2316 __be32 status;
2317 struct nfs4_stateid *stp;
2318 unsigned int share_access;
2320 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
2321 (int)cstate->current_fh.fh_dentry->d_name.len,
2322 cstate->current_fh.fh_dentry->d_name.name);
2324 if (!access_valid(od->od_share_access)
2325 || !deny_valid(od->od_share_deny))
2326 return nfserr_inval;
2328 nfs4_lock_state();
2329 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2330 od->od_seqid,
2331 &od->od_stateid,
2332 CHECK_FH | OPEN_STATE,
2333 &od->od_stateowner, &stp, NULL)))
2334 goto out;
2336 status = nfserr_inval;
2337 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
2338 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
2339 stp->st_access_bmap, od->od_share_access);
2340 goto out;
2342 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
2343 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
2344 stp->st_deny_bmap, od->od_share_deny);
2345 goto out;
2347 set_access(&share_access, stp->st_access_bmap);
2348 nfs4_file_downgrade(stp->st_vfs_file,
2349 share_access & ~od->od_share_access);
2351 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
2352 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
2354 update_stateid(&stp->st_stateid);
2355 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
2356 status = nfs_ok;
2357 out:
2358 if (od->od_stateowner) {
2359 nfs4_get_stateowner(od->od_stateowner);
2360 cstate->replay_owner = od->od_stateowner;
2362 nfs4_unlock_state();
2363 return status;
2367 * nfs4_unlock_state() called after encode
2369 __be32
2370 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2371 struct nfsd4_close *close)
2373 __be32 status;
2374 struct nfs4_stateid *stp;
2376 dprintk("NFSD: nfsd4_close on file %.*s\n",
2377 (int)cstate->current_fh.fh_dentry->d_name.len,
2378 cstate->current_fh.fh_dentry->d_name.name);
2380 nfs4_lock_state();
2381 /* check close_lru for replay */
2382 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2383 close->cl_seqid,
2384 &close->cl_stateid,
2385 CHECK_FH | OPEN_STATE | CLOSE_STATE,
2386 &close->cl_stateowner, &stp, NULL)))
2387 goto out;
2388 status = nfs_ok;
2389 update_stateid(&stp->st_stateid);
2390 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
2392 /* release_stateid() calls nfsd_close() if needed */
2393 release_stateid(stp, OPEN_STATE);
2395 /* place unused nfs4_stateowners on so_close_lru list to be
2396 * released by the laundromat service after the lease period
2397 * to enable us to handle CLOSE replay
2399 if (list_empty(&close->cl_stateowner->so_stateids))
2400 move_to_close_lru(close->cl_stateowner);
2401 out:
2402 if (close->cl_stateowner) {
2403 nfs4_get_stateowner(close->cl_stateowner);
2404 cstate->replay_owner = close->cl_stateowner;
2406 nfs4_unlock_state();
2407 return status;
2410 __be32
2411 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2412 struct nfsd4_delegreturn *dr)
2414 __be32 status;
2416 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
2417 goto out;
2419 nfs4_lock_state();
2420 status = nfs4_preprocess_stateid_op(&cstate->current_fh,
2421 &dr->dr_stateid, DELEG_RET, NULL);
2422 nfs4_unlock_state();
2423 out:
2424 return status;
2429 * Lock owner state (byte-range locks)
2431 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
2432 #define LOCK_HASH_BITS 8
2433 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
2434 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
2436 #define lockownerid_hashval(id) \
2437 ((id) & LOCK_HASH_MASK)
2439 static inline unsigned int
2440 lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
2441 struct xdr_netobj *ownername)
2443 return (file_hashval(inode) + cl_id
2444 + opaque_hashval(ownername->data, ownername->len))
2445 & LOCK_HASH_MASK;
2448 static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
2449 static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
2450 static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
2452 static struct nfs4_stateid *
2453 find_stateid(stateid_t *stid, int flags)
2455 struct nfs4_stateid *local = NULL;
2456 u32 st_id = stid->si_stateownerid;
2457 u32 f_id = stid->si_fileid;
2458 unsigned int hashval;
2460 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
2461 if ((flags & LOCK_STATE) || (flags & RD_STATE) || (flags & WR_STATE)) {
2462 hashval = stateid_hashval(st_id, f_id);
2463 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
2464 if ((local->st_stateid.si_stateownerid == st_id) &&
2465 (local->st_stateid.si_fileid == f_id))
2466 return local;
2469 if ((flags & OPEN_STATE) || (flags & RD_STATE) || (flags & WR_STATE)) {
2470 hashval = stateid_hashval(st_id, f_id);
2471 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
2472 if ((local->st_stateid.si_stateownerid == st_id) &&
2473 (local->st_stateid.si_fileid == f_id))
2474 return local;
2477 return NULL;
2480 static struct nfs4_delegation *
2481 find_delegation_stateid(struct inode *ino, stateid_t *stid)
2483 struct nfs4_file *fp;
2484 struct nfs4_delegation *dl;
2486 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
2487 stid->si_boot, stid->si_stateownerid,
2488 stid->si_fileid, stid->si_generation);
2490 fp = find_file(ino);
2491 if (!fp)
2492 return NULL;
2493 dl = find_delegation_file(fp, stid);
2494 put_nfs4_file(fp);
2495 return dl;
2499 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
2500 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
2501 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
2502 * locking, this prevents us from being completely protocol-compliant. The
2503 * real solution to this problem is to start using unsigned file offsets in
2504 * the VFS, but this is a very deep change!
2506 static inline void
2507 nfs4_transform_lock_offset(struct file_lock *lock)
2509 if (lock->fl_start < 0)
2510 lock->fl_start = OFFSET_MAX;
2511 if (lock->fl_end < 0)
2512 lock->fl_end = OFFSET_MAX;
2515 /* Hack!: For now, we're defining this just so we can use a pointer to it
2516 * as a unique cookie to identify our (NFSv4's) posix locks. */
2517 static struct lock_manager_operations nfsd_posix_mng_ops = {
2520 static inline void
2521 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
2523 struct nfs4_stateowner *sop;
2524 unsigned int hval;
2526 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
2527 sop = (struct nfs4_stateowner *) fl->fl_owner;
2528 hval = lockownerid_hashval(sop->so_id);
2529 kref_get(&sop->so_ref);
2530 deny->ld_sop = sop;
2531 deny->ld_clientid = sop->so_client->cl_clientid;
2532 } else {
2533 deny->ld_sop = NULL;
2534 deny->ld_clientid.cl_boot = 0;
2535 deny->ld_clientid.cl_id = 0;
2537 deny->ld_start = fl->fl_start;
2538 deny->ld_length = ~(u64)0;
2539 if (fl->fl_end != ~(u64)0)
2540 deny->ld_length = fl->fl_end - fl->fl_start + 1;
2541 deny->ld_type = NFS4_READ_LT;
2542 if (fl->fl_type != F_RDLCK)
2543 deny->ld_type = NFS4_WRITE_LT;
2546 static struct nfs4_stateowner *
2547 find_lockstateowner_str(struct inode *inode, clientid_t *clid,
2548 struct xdr_netobj *owner)
2550 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
2551 struct nfs4_stateowner *op;
2553 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
2554 if (same_owner_str(op, owner, clid))
2555 return op;
2557 return NULL;
2561 * Alloc a lock owner structure.
2562 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
2563 * occured.
2565 * strhashval = lock_ownerstr_hashval
2568 static struct nfs4_stateowner *
2569 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
2570 struct nfs4_stateowner *sop;
2571 struct nfs4_replay *rp;
2572 unsigned int idhashval;
2574 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
2575 return NULL;
2576 idhashval = lockownerid_hashval(current_ownerid);
2577 INIT_LIST_HEAD(&sop->so_idhash);
2578 INIT_LIST_HEAD(&sop->so_strhash);
2579 INIT_LIST_HEAD(&sop->so_perclient);
2580 INIT_LIST_HEAD(&sop->so_stateids);
2581 INIT_LIST_HEAD(&sop->so_perstateid);
2582 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
2583 sop->so_time = 0;
2584 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
2585 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
2586 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
2587 sop->so_is_open_owner = 0;
2588 sop->so_id = current_ownerid++;
2589 sop->so_client = clp;
2590 /* It is the openowner seqid that will be incremented in encode in the
2591 * case of new lockowners; so increment the lock seqid manually: */
2592 sop->so_seqid = lock->lk_new_lock_seqid + 1;
2593 sop->so_confirmed = 1;
2594 rp = &sop->so_replay;
2595 rp->rp_status = nfserr_serverfault;
2596 rp->rp_buflen = 0;
2597 rp->rp_buf = rp->rp_ibuf;
2598 return sop;
2601 static struct nfs4_stateid *
2602 alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
2604 struct nfs4_stateid *stp;
2605 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
2607 stp = nfs4_alloc_stateid();
2608 if (stp == NULL)
2609 goto out;
2610 INIT_LIST_HEAD(&stp->st_hash);
2611 INIT_LIST_HEAD(&stp->st_perfile);
2612 INIT_LIST_HEAD(&stp->st_perstateowner);
2613 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
2614 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
2615 list_add(&stp->st_perfile, &fp->fi_stateids);
2616 list_add(&stp->st_perstateowner, &sop->so_stateids);
2617 stp->st_stateowner = sop;
2618 get_nfs4_file(fp);
2619 stp->st_file = fp;
2620 stp->st_stateid.si_boot = boot_time;
2621 stp->st_stateid.si_stateownerid = sop->so_id;
2622 stp->st_stateid.si_fileid = fp->fi_id;
2623 stp->st_stateid.si_generation = 0;
2624 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
2625 stp->st_access_bmap = open_stp->st_access_bmap;
2626 stp->st_deny_bmap = open_stp->st_deny_bmap;
2627 stp->st_openstp = open_stp;
2629 out:
2630 return stp;
2633 static int
2634 check_lock_length(u64 offset, u64 length)
2636 return ((length == 0) || ((length != ~(u64)0) &&
2637 LOFF_OVERFLOW(offset, length)));
2641 * LOCK operation
2643 __be32
2644 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2645 struct nfsd4_lock *lock)
2647 struct nfs4_stateowner *open_sop = NULL;
2648 struct nfs4_stateowner *lock_sop = NULL;
2649 struct nfs4_stateid *lock_stp;
2650 struct file *filp;
2651 struct file_lock file_lock;
2652 struct file_lock conflock;
2653 __be32 status = 0;
2654 unsigned int strhashval;
2655 unsigned int cmd;
2656 int err;
2658 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
2659 (long long) lock->lk_offset,
2660 (long long) lock->lk_length);
2662 if (check_lock_length(lock->lk_offset, lock->lk_length))
2663 return nfserr_inval;
2665 if ((status = fh_verify(rqstp, &cstate->current_fh,
2666 S_IFREG, MAY_LOCK))) {
2667 dprintk("NFSD: nfsd4_lock: permission denied!\n");
2668 return status;
2671 nfs4_lock_state();
2673 if (lock->lk_is_new) {
2675 * Client indicates that this is a new lockowner.
2676 * Use open owner and open stateid to create lock owner and
2677 * lock stateid.
2679 struct nfs4_stateid *open_stp = NULL;
2680 struct nfs4_file *fp;
2682 status = nfserr_stale_clientid;
2683 if (STALE_CLIENTID(&lock->lk_new_clientid))
2684 goto out;
2686 /* validate and update open stateid and open seqid */
2687 status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2688 lock->lk_new_open_seqid,
2689 &lock->lk_new_open_stateid,
2690 CHECK_FH | OPEN_STATE,
2691 &lock->lk_replay_owner, &open_stp,
2692 lock);
2693 if (status)
2694 goto out;
2695 open_sop = lock->lk_replay_owner;
2696 /* create lockowner and lock stateid */
2697 fp = open_stp->st_file;
2698 strhashval = lock_ownerstr_hashval(fp->fi_inode,
2699 open_sop->so_client->cl_clientid.cl_id,
2700 &lock->v.new.owner);
2701 /* XXX: Do we need to check for duplicate stateowners on
2702 * the same file, or should they just be allowed (and
2703 * create new stateids)? */
2704 status = nfserr_resource;
2705 lock_sop = alloc_init_lock_stateowner(strhashval,
2706 open_sop->so_client, open_stp, lock);
2707 if (lock_sop == NULL)
2708 goto out;
2709 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
2710 if (lock_stp == NULL)
2711 goto out;
2712 } else {
2713 /* lock (lock owner + lock stateid) already exists */
2714 status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2715 lock->lk_old_lock_seqid,
2716 &lock->lk_old_lock_stateid,
2717 CHECK_FH | LOCK_STATE,
2718 &lock->lk_replay_owner, &lock_stp, lock);
2719 if (status)
2720 goto out;
2721 lock_sop = lock->lk_replay_owner;
2723 /* lock->lk_replay_owner and lock_stp have been created or found */
2724 filp = lock_stp->st_vfs_file;
2726 status = nfserr_grace;
2727 if (nfs4_in_grace() && !lock->lk_reclaim)
2728 goto out;
2729 status = nfserr_no_grace;
2730 if (!nfs4_in_grace() && lock->lk_reclaim)
2731 goto out;
2733 locks_init_lock(&file_lock);
2734 switch (lock->lk_type) {
2735 case NFS4_READ_LT:
2736 case NFS4_READW_LT:
2737 file_lock.fl_type = F_RDLCK;
2738 cmd = F_SETLK;
2739 break;
2740 case NFS4_WRITE_LT:
2741 case NFS4_WRITEW_LT:
2742 file_lock.fl_type = F_WRLCK;
2743 cmd = F_SETLK;
2744 break;
2745 default:
2746 status = nfserr_inval;
2747 goto out;
2749 file_lock.fl_owner = (fl_owner_t)lock_sop;
2750 file_lock.fl_pid = current->tgid;
2751 file_lock.fl_file = filp;
2752 file_lock.fl_flags = FL_POSIX;
2753 file_lock.fl_lmops = &nfsd_posix_mng_ops;
2755 file_lock.fl_start = lock->lk_offset;
2756 if ((lock->lk_length == ~(u64)0) ||
2757 LOFF_OVERFLOW(lock->lk_offset, lock->lk_length))
2758 file_lock.fl_end = ~(u64)0;
2759 else
2760 file_lock.fl_end = lock->lk_offset + lock->lk_length - 1;
2761 nfs4_transform_lock_offset(&file_lock);
2764 * Try to lock the file in the VFS.
2765 * Note: locks.c uses the BKL to protect the inode's lock list.
2768 /* XXX?: Just to divert the locks_release_private at the start of
2769 * locks_copy_lock: */
2770 locks_init_lock(&conflock);
2771 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
2772 switch (-err) {
2773 case 0: /* success! */
2774 update_stateid(&lock_stp->st_stateid);
2775 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
2776 sizeof(stateid_t));
2777 status = 0;
2778 break;
2779 case (EAGAIN): /* conflock holds conflicting lock */
2780 status = nfserr_denied;
2781 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
2782 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
2783 break;
2784 case (EDEADLK):
2785 status = nfserr_deadlock;
2786 break;
2787 default:
2788 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
2789 status = nfserr_resource;
2790 break;
2792 out:
2793 if (status && lock->lk_is_new && lock_sop)
2794 release_stateowner(lock_sop);
2795 if (lock->lk_replay_owner) {
2796 nfs4_get_stateowner(lock->lk_replay_owner);
2797 cstate->replay_owner = lock->lk_replay_owner;
2799 nfs4_unlock_state();
2800 return status;
2804 * LOCKT operation
2806 __be32
2807 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2808 struct nfsd4_lockt *lockt)
2810 struct inode *inode;
2811 struct file file;
2812 struct file_lock file_lock;
2813 int error;
2814 __be32 status;
2816 if (nfs4_in_grace())
2817 return nfserr_grace;
2819 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
2820 return nfserr_inval;
2822 lockt->lt_stateowner = NULL;
2823 nfs4_lock_state();
2825 status = nfserr_stale_clientid;
2826 if (STALE_CLIENTID(&lockt->lt_clientid))
2827 goto out;
2829 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
2830 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
2831 if (status == nfserr_symlink)
2832 status = nfserr_inval;
2833 goto out;
2836 inode = cstate->current_fh.fh_dentry->d_inode;
2837 locks_init_lock(&file_lock);
2838 switch (lockt->lt_type) {
2839 case NFS4_READ_LT:
2840 case NFS4_READW_LT:
2841 file_lock.fl_type = F_RDLCK;
2842 break;
2843 case NFS4_WRITE_LT:
2844 case NFS4_WRITEW_LT:
2845 file_lock.fl_type = F_WRLCK;
2846 break;
2847 default:
2848 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
2849 status = nfserr_inval;
2850 goto out;
2853 lockt->lt_stateowner = find_lockstateowner_str(inode,
2854 &lockt->lt_clientid, &lockt->lt_owner);
2855 if (lockt->lt_stateowner)
2856 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
2857 file_lock.fl_pid = current->tgid;
2858 file_lock.fl_flags = FL_POSIX;
2859 file_lock.fl_lmops = &nfsd_posix_mng_ops;
2861 file_lock.fl_start = lockt->lt_offset;
2862 if ((lockt->lt_length == ~(u64)0) || LOFF_OVERFLOW(lockt->lt_offset, lockt->lt_length))
2863 file_lock.fl_end = ~(u64)0;
2864 else
2865 file_lock.fl_end = lockt->lt_offset + lockt->lt_length - 1;
2867 nfs4_transform_lock_offset(&file_lock);
2869 /* vfs_test_lock uses the struct file _only_ to resolve the inode.
2870 * since LOCKT doesn't require an OPEN, and therefore a struct
2871 * file may not exist, pass vfs_test_lock a struct file with
2872 * only the dentry:inode set.
2874 memset(&file, 0, sizeof (struct file));
2875 file.f_path.dentry = cstate->current_fh.fh_dentry;
2877 status = nfs_ok;
2878 error = vfs_test_lock(&file, &file_lock);
2879 if (error) {
2880 status = nfserrno(error);
2881 goto out;
2883 if (file_lock.fl_type != F_UNLCK) {
2884 status = nfserr_denied;
2885 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
2887 out:
2888 nfs4_unlock_state();
2889 return status;
2892 __be32
2893 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2894 struct nfsd4_locku *locku)
2896 struct nfs4_stateid *stp;
2897 struct file *filp = NULL;
2898 struct file_lock file_lock;
2899 __be32 status;
2900 int err;
2902 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
2903 (long long) locku->lu_offset,
2904 (long long) locku->lu_length);
2906 if (check_lock_length(locku->lu_offset, locku->lu_length))
2907 return nfserr_inval;
2909 nfs4_lock_state();
2911 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2912 locku->lu_seqid,
2913 &locku->lu_stateid,
2914 CHECK_FH | LOCK_STATE,
2915 &locku->lu_stateowner, &stp, NULL)))
2916 goto out;
2918 filp = stp->st_vfs_file;
2919 BUG_ON(!filp);
2920 locks_init_lock(&file_lock);
2921 file_lock.fl_type = F_UNLCK;
2922 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
2923 file_lock.fl_pid = current->tgid;
2924 file_lock.fl_file = filp;
2925 file_lock.fl_flags = FL_POSIX;
2926 file_lock.fl_lmops = &nfsd_posix_mng_ops;
2927 file_lock.fl_start = locku->lu_offset;
2929 if ((locku->lu_length == ~(u64)0) || LOFF_OVERFLOW(locku->lu_offset, locku->lu_length))
2930 file_lock.fl_end = ~(u64)0;
2931 else
2932 file_lock.fl_end = locku->lu_offset + locku->lu_length - 1;
2933 nfs4_transform_lock_offset(&file_lock);
2936 * Try to unlock the file in the VFS.
2938 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
2939 if (err) {
2940 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
2941 goto out_nfserr;
2944 * OK, unlock succeeded; the only thing left to do is update the stateid.
2946 update_stateid(&stp->st_stateid);
2947 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
2949 out:
2950 if (locku->lu_stateowner) {
2951 nfs4_get_stateowner(locku->lu_stateowner);
2952 cstate->replay_owner = locku->lu_stateowner;
2954 nfs4_unlock_state();
2955 return status;
2957 out_nfserr:
2958 status = nfserrno(err);
2959 goto out;
2963 * returns
2964 * 1: locks held by lockowner
2965 * 0: no locks held by lockowner
2967 static int
2968 check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
2970 struct file_lock **flpp;
2971 struct inode *inode = filp->f_path.dentry->d_inode;
2972 int status = 0;
2974 lock_kernel();
2975 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
2976 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
2977 status = 1;
2978 goto out;
2981 out:
2982 unlock_kernel();
2983 return status;
2986 __be32
2987 nfsd4_release_lockowner(struct svc_rqst *rqstp,
2988 struct nfsd4_compound_state *cstate,
2989 struct nfsd4_release_lockowner *rlockowner)
2991 clientid_t *clid = &rlockowner->rl_clientid;
2992 struct nfs4_stateowner *sop;
2993 struct nfs4_stateid *stp;
2994 struct xdr_netobj *owner = &rlockowner->rl_owner;
2995 struct list_head matches;
2996 int i;
2997 __be32 status;
2999 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3000 clid->cl_boot, clid->cl_id);
3002 /* XXX check for lease expiration */
3004 status = nfserr_stale_clientid;
3005 if (STALE_CLIENTID(clid))
3006 return status;
3008 nfs4_lock_state();
3010 status = nfserr_locks_held;
3011 /* XXX: we're doing a linear search through all the lockowners.
3012 * Yipes! For now we'll just hope clients aren't really using
3013 * release_lockowner much, but eventually we have to fix these
3014 * data structures. */
3015 INIT_LIST_HEAD(&matches);
3016 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3017 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
3018 if (!same_owner_str(sop, owner, clid))
3019 continue;
3020 list_for_each_entry(stp, &sop->so_stateids,
3021 st_perstateowner) {
3022 if (check_for_locks(stp->st_vfs_file, sop))
3023 goto out;
3024 /* Note: so_perclient unused for lockowners,
3025 * so it's OK to fool with here. */
3026 list_add(&sop->so_perclient, &matches);
3030 /* Clients probably won't expect us to return with some (but not all)
3031 * of the lockowner state released; so don't release any until all
3032 * have been checked. */
3033 status = nfs_ok;
3034 while (!list_empty(&matches)) {
3035 sop = list_entry(matches.next, struct nfs4_stateowner,
3036 so_perclient);
3037 /* unhash_stateowner deletes so_perclient only
3038 * for openowners. */
3039 list_del(&sop->so_perclient);
3040 release_stateowner(sop);
3042 out:
3043 nfs4_unlock_state();
3044 return status;
3047 static inline struct nfs4_client_reclaim *
3048 alloc_reclaim(void)
3050 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
3054 nfs4_has_reclaimed_state(const char *name)
3056 unsigned int strhashval = clientstr_hashval(name);
3057 struct nfs4_client *clp;
3059 clp = find_confirmed_client_by_str(name, strhashval);
3060 return clp ? 1 : 0;
3064 * failure => all reset bets are off, nfserr_no_grace...
3067 nfs4_client_to_reclaim(const char *name)
3069 unsigned int strhashval;
3070 struct nfs4_client_reclaim *crp = NULL;
3072 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3073 crp = alloc_reclaim();
3074 if (!crp)
3075 return 0;
3076 strhashval = clientstr_hashval(name);
3077 INIT_LIST_HEAD(&crp->cr_strhash);
3078 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
3079 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
3080 reclaim_str_hashtbl_size++;
3081 return 1;
3084 static void
3085 nfs4_release_reclaim(void)
3087 struct nfs4_client_reclaim *crp = NULL;
3088 int i;
3090 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3091 while (!list_empty(&reclaim_str_hashtbl[i])) {
3092 crp = list_entry(reclaim_str_hashtbl[i].next,
3093 struct nfs4_client_reclaim, cr_strhash);
3094 list_del(&crp->cr_strhash);
3095 kfree(crp);
3096 reclaim_str_hashtbl_size--;
3099 BUG_ON(reclaim_str_hashtbl_size);
3103 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
3104 static struct nfs4_client_reclaim *
3105 nfs4_find_reclaim_client(clientid_t *clid)
3107 unsigned int strhashval;
3108 struct nfs4_client *clp;
3109 struct nfs4_client_reclaim *crp = NULL;
3112 /* find clientid in conf_id_hashtbl */
3113 clp = find_confirmed_client(clid);
3114 if (clp == NULL)
3115 return NULL;
3117 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3118 clp->cl_name.len, clp->cl_name.data,
3119 clp->cl_recdir);
3121 /* find clp->cl_name in reclaim_str_hashtbl */
3122 strhashval = clientstr_hashval(clp->cl_recdir);
3123 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
3124 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
3125 return crp;
3128 return NULL;
3132 * Called from OPEN. Look for clientid in reclaim list.
3134 __be32
3135 nfs4_check_open_reclaim(clientid_t *clid)
3137 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
3140 /* initialization to perform at module load time: */
3143 nfs4_state_init(void)
3145 int i, status;
3147 status = nfsd4_init_slabs();
3148 if (status)
3149 return status;
3150 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3151 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3152 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3153 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3154 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
3156 for (i = 0; i < FILE_HASH_SIZE; i++) {
3157 INIT_LIST_HEAD(&file_hashtbl[i]);
3159 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3160 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3161 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3163 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3164 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3165 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3167 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3168 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3169 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3171 memset(&onestateid, ~0, sizeof(stateid_t));
3172 INIT_LIST_HEAD(&close_lru);
3173 INIT_LIST_HEAD(&client_lru);
3174 INIT_LIST_HEAD(&del_recall_lru);
3175 for (i = 0; i < CLIENT_HASH_SIZE; i++)
3176 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
3177 reclaim_str_hashtbl_size = 0;
3178 return 0;
3181 static void
3182 nfsd4_load_reboot_recovery_data(void)
3184 int status;
3186 nfs4_lock_state();
3187 nfsd4_init_recdir(user_recovery_dirname);
3188 status = nfsd4_recdir_load();
3189 nfs4_unlock_state();
3190 if (status)
3191 printk("NFSD: Failure reading reboot recovery data\n");
3194 unsigned long
3195 get_nfs4_grace_period(void)
3197 return max(user_lease_time, lease_time) * HZ;
3201 * Since the lifetime of a delegation isn't limited to that of an open, a
3202 * client may quite reasonably hang on to a delegation as long as it has
3203 * the inode cached. This becomes an obvious problem the first time a
3204 * client's inode cache approaches the size of the server's total memory.
3206 * For now we avoid this problem by imposing a hard limit on the number
3207 * of delegations, which varies according to the server's memory size.
3209 static void
3210 set_max_delegations(void)
3213 * Allow at most 4 delegations per megabyte of RAM. Quick
3214 * estimates suggest that in the worst case (where every delegation
3215 * is for a different inode), a delegation could take about 1.5K,
3216 * giving a worst case usage of about 6% of memory.
3218 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
3221 /* initialization to perform when the nfsd service is started: */
3223 static void
3224 __nfs4_state_start(void)
3226 unsigned long grace_time;
3228 boot_time = get_seconds();
3229 grace_time = get_nfs_grace_period();
3230 lease_time = user_lease_time;
3231 in_grace = 1;
3232 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
3233 grace_time/HZ);
3234 laundry_wq = create_singlethread_workqueue("nfsd4");
3235 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
3236 set_max_delegations();
3239 void
3240 nfs4_state_start(void)
3242 if (nfs4_init)
3243 return;
3244 nfsd4_load_reboot_recovery_data();
3245 __nfs4_state_start();
3246 nfs4_init = 1;
3247 return;
3251 nfs4_in_grace(void)
3253 return in_grace;
3256 time_t
3257 nfs4_lease_time(void)
3259 return lease_time;
3262 static void
3263 __nfs4_state_shutdown(void)
3265 int i;
3266 struct nfs4_client *clp = NULL;
3267 struct nfs4_delegation *dp = NULL;
3268 struct list_head *pos, *next, reaplist;
3270 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3271 while (!list_empty(&conf_id_hashtbl[i])) {
3272 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
3273 expire_client(clp);
3275 while (!list_empty(&unconf_str_hashtbl[i])) {
3276 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
3277 expire_client(clp);
3280 INIT_LIST_HEAD(&reaplist);
3281 spin_lock(&recall_lock);
3282 list_for_each_safe(pos, next, &del_recall_lru) {
3283 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3284 list_move(&dp->dl_recall_lru, &reaplist);
3286 spin_unlock(&recall_lock);
3287 list_for_each_safe(pos, next, &reaplist) {
3288 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3289 list_del_init(&dp->dl_recall_lru);
3290 unhash_delegation(dp);
3293 nfsd4_shutdown_recdir();
3294 nfs4_init = 0;
3297 void
3298 nfs4_state_shutdown(void)
3300 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
3301 destroy_workqueue(laundry_wq);
3302 nfs4_lock_state();
3303 nfs4_release_reclaim();
3304 __nfs4_state_shutdown();
3305 nfs4_unlock_state();
3308 static void
3309 nfs4_set_recdir(char *recdir)
3311 nfs4_lock_state();
3312 strcpy(user_recovery_dirname, recdir);
3313 nfs4_unlock_state();
3317 * Change the NFSv4 recovery directory to recdir.
3320 nfs4_reset_recoverydir(char *recdir)
3322 int status;
3323 struct nameidata nd;
3325 status = path_lookup(recdir, LOOKUP_FOLLOW, &nd);
3326 if (status)
3327 return status;
3328 status = -ENOTDIR;
3329 if (S_ISDIR(nd.dentry->d_inode->i_mode)) {
3330 nfs4_set_recdir(recdir);
3331 status = 0;
3333 path_release(&nd);
3334 return status;
3338 * Called when leasetime is changed.
3340 * The only way the protocol gives us to handle on-the-fly lease changes is to
3341 * simulate a reboot. Instead of doing that, we just wait till the next time
3342 * we start to register any changes in lease time. If the administrator
3343 * really wants to change the lease time *now*, they can go ahead and bring
3344 * nfsd down and then back up again after changing the lease time.
3346 void
3347 nfs4_reset_lease(time_t leasetime)
3349 lock_kernel();
3350 user_lease_time = leasetime;
3351 unlock_kernel();