2 * linux/fs/nfsd/nfs4state.c
4 * Copyright (c) 2001 The Regents of the University of Michigan.
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
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/file.h>
45 #include <linux/mount.h>
46 #include <linux/workqueue.h>
47 #include <linux/smp_lock.h>
48 #include <linux/kthread.h>
49 #include <linux/nfs4.h>
50 #include <linux/nfsd/state.h>
51 #include <linux/nfsd/xdr4.h>
52 #include <linux/namei.h>
53 #include <linux/swap.h>
54 #include <linux/mutex.h>
55 #include <linux/lockd/bind.h>
56 #include <linux/module.h>
57 #include <linux/sunrpc/svcauth_gss.h>
59 #define NFSDDBG_FACILITY NFSDDBG_PROC
62 static time_t lease_time = 90; /* default lease time */
63 static time_t user_lease_time = 90;
64 static time_t boot_time;
65 static u32 current_ownerid = 1;
66 static u32 current_fileid = 1;
67 static u32 current_delegid = 1;
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 char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery";
79 static void nfs4_set_recdir(char *recdir);
83 /* Currently used for almost all code touching nfsv4 state: */
84 static DEFINE_MUTEX(client_mutex);
87 * Currently used for the del_recall_lru and file hash table. In an
88 * effort to decrease the scope of the client_mutex, this spinlock may
89 * eventually cover more:
91 static DEFINE_SPINLOCK(recall_lock);
93 static struct kmem_cache *stateowner_slab = NULL;
94 static struct kmem_cache *file_slab = NULL;
95 static struct kmem_cache *stateid_slab = NULL;
96 static struct kmem_cache *deleg_slab = NULL;
101 mutex_lock(&client_mutex);
105 nfs4_unlock_state(void)
107 mutex_unlock(&client_mutex);
111 opaque_hashval(const void *ptr, int nbytes)
113 unsigned char *cptr = (unsigned char *) ptr;
123 static struct list_head del_recall_lru;
126 put_nfs4_file(struct nfs4_file *fi)
128 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
129 list_del(&fi->fi_hash);
130 spin_unlock(&recall_lock);
132 kmem_cache_free(file_slab, fi);
137 get_nfs4_file(struct nfs4_file *fi)
139 atomic_inc(&fi->fi_ref);
142 static int num_delegations;
143 unsigned int max_delegations;
146 * Open owner state (share locks)
149 /* hash tables for nfs4_stateowner */
150 #define OWNER_HASH_BITS 8
151 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
152 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
154 #define ownerid_hashval(id) \
155 ((id) & OWNER_HASH_MASK)
156 #define ownerstr_hashval(clientid, ownername) \
157 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
159 static struct list_head ownerid_hashtbl[OWNER_HASH_SIZE];
160 static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
162 /* hash table for nfs4_file */
163 #define FILE_HASH_BITS 8
164 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
165 #define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
166 /* hash table for (open)nfs4_stateid */
167 #define STATEID_HASH_BITS 10
168 #define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
169 #define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
171 #define file_hashval(x) \
172 hash_ptr(x, FILE_HASH_BITS)
173 #define stateid_hashval(owner_id, file_id) \
174 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
176 static struct list_head file_hashtbl[FILE_HASH_SIZE];
177 static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
179 static struct nfs4_delegation *
180 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
182 struct nfs4_delegation *dp;
183 struct nfs4_file *fp = stp->st_file;
184 struct nfs4_callback *cb = &stp->st_stateowner->so_client->cl_callback;
186 dprintk("NFSD alloc_init_deleg\n");
187 if (fp->fi_had_conflict)
189 if (num_delegations > max_delegations)
191 dp = kmem_cache_alloc(deleg_slab, GFP_KERNEL);
195 INIT_LIST_HEAD(&dp->dl_perfile);
196 INIT_LIST_HEAD(&dp->dl_perclnt);
197 INIT_LIST_HEAD(&dp->dl_recall_lru);
202 get_file(stp->st_vfs_file);
203 dp->dl_vfs_file = stp->st_vfs_file;
205 dp->dl_recall.cbr_dp = NULL;
206 dp->dl_recall.cbr_ident = cb->cb_ident;
207 dp->dl_recall.cbr_trunc = 0;
208 dp->dl_stateid.si_boot = boot_time;
209 dp->dl_stateid.si_stateownerid = current_delegid++;
210 dp->dl_stateid.si_fileid = 0;
211 dp->dl_stateid.si_generation = 0;
212 fh_copy_shallow(&dp->dl_fh, ¤t_fh->fh_handle);
214 atomic_set(&dp->dl_count, 1);
215 list_add(&dp->dl_perfile, &fp->fi_delegations);
216 list_add(&dp->dl_perclnt, &clp->cl_delegations);
221 nfs4_put_delegation(struct nfs4_delegation *dp)
223 if (atomic_dec_and_test(&dp->dl_count)) {
224 dprintk("NFSD: freeing dp %p\n",dp);
225 put_nfs4_file(dp->dl_file);
226 kmem_cache_free(deleg_slab, dp);
231 /* Remove the associated file_lock first, then remove the delegation.
232 * lease_modify() is called to remove the FS_LEASE file_lock from
233 * the i_flock list, eventually calling nfsd's lock_manager
234 * fl_release_callback.
237 nfs4_close_delegation(struct nfs4_delegation *dp)
239 struct file *filp = dp->dl_vfs_file;
241 dprintk("NFSD: close_delegation dp %p\n",dp);
242 dp->dl_vfs_file = NULL;
243 /* The following nfsd_close may not actually close the file,
244 * but we want to remove the lease in any case. */
246 vfs_setlease(filp, F_UNLCK, &dp->dl_flock);
250 /* Called under the state lock. */
252 unhash_delegation(struct nfs4_delegation *dp)
254 list_del_init(&dp->dl_perfile);
255 list_del_init(&dp->dl_perclnt);
256 spin_lock(&recall_lock);
257 list_del_init(&dp->dl_recall_lru);
258 spin_unlock(&recall_lock);
259 nfs4_close_delegation(dp);
260 nfs4_put_delegation(dp);
267 /* Hash tables for nfs4_clientid state */
268 #define CLIENT_HASH_BITS 4
269 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
270 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
272 #define clientid_hashval(id) \
273 ((id) & CLIENT_HASH_MASK)
274 #define clientstr_hashval(name) \
275 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
277 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
278 * used in reboot/reset lease grace period processing
280 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
281 * setclientid_confirmed info.
283 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
286 * client_lru holds client queue ordered by nfs4_client.cl_time
289 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
290 * for last close replay.
292 static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
293 static int reclaim_str_hashtbl_size = 0;
294 static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
295 static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
296 static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
297 static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
298 static struct list_head client_lru;
299 static struct list_head close_lru;
301 static void unhash_generic_stateid(struct nfs4_stateid *stp)
303 list_del(&stp->st_hash);
304 list_del(&stp->st_perfile);
305 list_del(&stp->st_perstateowner);
308 static void free_generic_stateid(struct nfs4_stateid *stp)
310 put_nfs4_file(stp->st_file);
311 kmem_cache_free(stateid_slab, stp);
314 static void release_lock_stateid(struct nfs4_stateid *stp)
316 unhash_generic_stateid(stp);
317 locks_remove_posix(stp->st_vfs_file, (fl_owner_t)stp->st_stateowner);
318 free_generic_stateid(stp);
321 static void unhash_lockowner(struct nfs4_stateowner *sop)
323 struct nfs4_stateid *stp;
325 list_del(&sop->so_idhash);
326 list_del(&sop->so_strhash);
327 list_del(&sop->so_perstateid);
328 while (!list_empty(&sop->so_stateids)) {
329 stp = list_first_entry(&sop->so_stateids,
330 struct nfs4_stateid, st_perstateowner);
331 release_lock_stateid(stp);
335 static void release_lockowner(struct nfs4_stateowner *sop)
337 unhash_lockowner(sop);
338 nfs4_put_stateowner(sop);
342 release_stateid_lockowners(struct nfs4_stateid *open_stp)
344 struct nfs4_stateowner *lock_sop;
346 while (!list_empty(&open_stp->st_lockowners)) {
347 lock_sop = list_entry(open_stp->st_lockowners.next,
348 struct nfs4_stateowner, so_perstateid);
349 /* list_del(&open_stp->st_lockowners); */
350 BUG_ON(lock_sop->so_is_open_owner);
351 release_lockowner(lock_sop);
355 static void release_open_stateid(struct nfs4_stateid *stp)
357 unhash_generic_stateid(stp);
358 release_stateid_lockowners(stp);
359 nfsd_close(stp->st_vfs_file);
360 free_generic_stateid(stp);
363 static void unhash_openowner(struct nfs4_stateowner *sop)
365 struct nfs4_stateid *stp;
367 list_del(&sop->so_idhash);
368 list_del(&sop->so_strhash);
369 list_del(&sop->so_perclient);
370 list_del(&sop->so_perstateid); /* XXX: necessary? */
371 while (!list_empty(&sop->so_stateids)) {
372 stp = list_first_entry(&sop->so_stateids,
373 struct nfs4_stateid, st_perstateowner);
374 release_open_stateid(stp);
378 static void release_openowner(struct nfs4_stateowner *sop)
380 unhash_openowner(sop);
381 list_del(&sop->so_close_lru);
382 nfs4_put_stateowner(sop);
386 release_session(struct nfsd4_session *ses)
388 list_del(&ses->se_hash);
389 list_del(&ses->se_perclnt);
390 nfsd4_put_session(ses);
394 free_session(struct kref *kref)
396 struct nfsd4_session *ses;
398 ses = container_of(kref, struct nfsd4_session, se_ref);
399 kfree(ses->se_slots);
404 renew_client(struct nfs4_client *clp)
407 * Move client to the end to the LRU list.
409 dprintk("renewing client (clientid %08x/%08x)\n",
410 clp->cl_clientid.cl_boot,
411 clp->cl_clientid.cl_id);
412 list_move_tail(&clp->cl_lru, &client_lru);
413 clp->cl_time = get_seconds();
416 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
418 STALE_CLIENTID(clientid_t *clid)
420 if (clid->cl_boot == boot_time)
422 dprintk("NFSD stale clientid (%08x/%08x)\n",
423 clid->cl_boot, clid->cl_id);
428 * XXX Should we use a slab cache ?
429 * This type of memory management is somewhat inefficient, but we use it
430 * anyway since SETCLIENTID is not a common operation.
432 static struct nfs4_client *alloc_client(struct xdr_netobj name)
434 struct nfs4_client *clp;
436 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
439 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
440 if (clp->cl_name.data == NULL) {
444 memcpy(clp->cl_name.data, name.data, name.len);
445 clp->cl_name.len = name.len;
450 shutdown_callback_client(struct nfs4_client *clp)
452 struct rpc_clnt *clnt = clp->cl_callback.cb_client;
456 * Callback threads take a reference on the client, so there
457 * should be no outstanding callbacks at this point.
459 clp->cl_callback.cb_client = NULL;
460 rpc_shutdown_client(clnt);
465 free_client(struct nfs4_client *clp)
467 shutdown_callback_client(clp);
468 if (clp->cl_cred.cr_group_info)
469 put_group_info(clp->cl_cred.cr_group_info);
470 kfree(clp->cl_principal);
471 kfree(clp->cl_name.data);
476 put_nfs4_client(struct nfs4_client *clp)
478 if (atomic_dec_and_test(&clp->cl_count))
483 expire_client(struct nfs4_client *clp)
485 struct nfs4_stateowner *sop;
486 struct nfs4_delegation *dp;
487 struct list_head reaplist;
489 dprintk("NFSD: expire_client cl_count %d\n",
490 atomic_read(&clp->cl_count));
492 INIT_LIST_HEAD(&reaplist);
493 spin_lock(&recall_lock);
494 while (!list_empty(&clp->cl_delegations)) {
495 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
496 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
498 list_del_init(&dp->dl_perclnt);
499 list_move(&dp->dl_recall_lru, &reaplist);
501 spin_unlock(&recall_lock);
502 while (!list_empty(&reaplist)) {
503 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
504 list_del_init(&dp->dl_recall_lru);
505 unhash_delegation(dp);
507 list_del(&clp->cl_idhash);
508 list_del(&clp->cl_strhash);
509 list_del(&clp->cl_lru);
510 while (!list_empty(&clp->cl_openowners)) {
511 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
512 release_openowner(sop);
514 while (!list_empty(&clp->cl_sessions)) {
515 struct nfsd4_session *ses;
516 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
518 release_session(ses);
520 put_nfs4_client(clp);
523 static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir)
525 struct nfs4_client *clp;
527 clp = alloc_client(name);
530 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
531 atomic_set(&clp->cl_count, 1);
532 atomic_set(&clp->cl_callback.cb_set, 0);
533 INIT_LIST_HEAD(&clp->cl_idhash);
534 INIT_LIST_HEAD(&clp->cl_strhash);
535 INIT_LIST_HEAD(&clp->cl_openowners);
536 INIT_LIST_HEAD(&clp->cl_delegations);
537 INIT_LIST_HEAD(&clp->cl_sessions);
538 INIT_LIST_HEAD(&clp->cl_lru);
542 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
544 memcpy(target->cl_verifier.data, source->data,
545 sizeof(target->cl_verifier.data));
548 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
550 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
551 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
554 static void copy_cred(struct svc_cred *target, struct svc_cred *source)
556 target->cr_uid = source->cr_uid;
557 target->cr_gid = source->cr_gid;
558 target->cr_group_info = source->cr_group_info;
559 get_group_info(target->cr_group_info);
562 static int same_name(const char *n1, const char *n2)
564 return 0 == memcmp(n1, n2, HEXDIR_LEN);
568 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
570 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
574 same_clid(clientid_t *cl1, clientid_t *cl2)
576 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
579 /* XXX what about NGROUP */
581 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
583 return cr1->cr_uid == cr2->cr_uid;
586 static void gen_clid(struct nfs4_client *clp)
588 static u32 current_clientid = 1;
590 clp->cl_clientid.cl_boot = boot_time;
591 clp->cl_clientid.cl_id = current_clientid++;
594 static void gen_confirm(struct nfs4_client *clp)
599 p = (u32 *)clp->cl_confirm.data;
600 *p++ = get_seconds();
604 static int check_name(struct xdr_netobj name)
608 if (name.len > NFS4_OPAQUE_LIMIT) {
609 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
616 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
618 unsigned int idhashval;
620 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
621 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
622 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
623 list_add_tail(&clp->cl_lru, &client_lru);
624 clp->cl_time = get_seconds();
628 move_to_confirmed(struct nfs4_client *clp)
630 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
631 unsigned int strhashval;
633 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
634 list_del_init(&clp->cl_strhash);
635 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
636 strhashval = clientstr_hashval(clp->cl_recdir);
637 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
641 static struct nfs4_client *
642 find_confirmed_client(clientid_t *clid)
644 struct nfs4_client *clp;
645 unsigned int idhashval = clientid_hashval(clid->cl_id);
647 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
648 if (same_clid(&clp->cl_clientid, clid))
654 static struct nfs4_client *
655 find_unconfirmed_client(clientid_t *clid)
657 struct nfs4_client *clp;
658 unsigned int idhashval = clientid_hashval(clid->cl_id);
660 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
661 if (same_clid(&clp->cl_clientid, clid))
667 static struct nfs4_client *
668 find_confirmed_client_by_str(const char *dname, unsigned int hashval)
670 struct nfs4_client *clp;
672 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
673 if (same_name(clp->cl_recdir, dname))
679 static struct nfs4_client *
680 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval)
682 struct nfs4_client *clp;
684 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
685 if (same_name(clp->cl_recdir, dname))
691 /* a helper function for parse_callback */
693 parse_octet(unsigned int *lenp, char **addrp)
695 unsigned int len = *lenp;
707 if ((c < '0') || (c > '9')) {
713 n = (n * 10) + (c - '0');
724 /* parse and set the setclientid ipv4 callback address */
726 parse_ipv4(unsigned int addr_len, char *addr_val, unsigned int *cbaddrp, unsigned short *cbportp)
731 u32 addrlen = addr_len;
732 char *addr = addr_val;
737 for(i = 4; i > 0 ; i--) {
738 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
741 cbaddr |= (temp << shift);
749 for(i = 2; i > 0 ; i--) {
750 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
753 cbport |= (temp << shift);
762 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se)
764 struct nfs4_callback *cb = &clp->cl_callback;
766 /* Currently, we only support tcp for the callback channel */
767 if ((se->se_callback_netid_len != 3) || memcmp((char *)se->se_callback_netid_val, "tcp", 3))
770 if ( !(parse_ipv4(se->se_callback_addr_len, se->se_callback_addr_val,
771 &cb->cb_addr, &cb->cb_port)))
773 cb->cb_prog = se->se_callback_prog;
774 cb->cb_ident = se->se_callback_ident;
777 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
778 "will not receive delegations\n",
779 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
785 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
786 struct nfsd4_setclientid *setclid)
788 struct sockaddr_in *sin = svc_addr_in(rqstp);
789 struct xdr_netobj clname = {
790 .len = setclid->se_namelen,
791 .data = setclid->se_name,
793 nfs4_verifier clverifier = setclid->se_verf;
794 unsigned int strhashval;
795 struct nfs4_client *conf, *unconf, *new;
798 char dname[HEXDIR_LEN];
800 if (!check_name(clname))
803 status = nfs4_make_rec_clidname(dname, &clname);
808 * XXX The Duplicate Request Cache (DRC) has been checked (??)
809 * We get here on a DRC miss.
812 strhashval = clientstr_hashval(dname);
815 conf = find_confirmed_client_by_str(dname, strhashval);
817 /* RFC 3530 14.2.33 CASE 0: */
818 status = nfserr_clid_inuse;
819 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
820 dprintk("NFSD: setclientid: string in use by client"
821 " at %pI4\n", &conf->cl_addr);
826 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
827 * has a description of SETCLIENTID request processing consisting
828 * of 5 bullet points, labeled as CASE0 - CASE4 below.
830 unconf = find_unconfirmed_client_by_str(dname, strhashval);
831 status = nfserr_resource;
834 * RFC 3530 14.2.33 CASE 4:
835 * placed first, because it is the normal case
838 expire_client(unconf);
839 new = create_client(clname, dname);
843 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
845 * RFC 3530 14.2.33 CASE 1:
846 * probable callback update
849 /* Note this is removing unconfirmed {*x***},
850 * which is stronger than RFC recommended {vxc**}.
851 * This has the advantage that there is at most
852 * one {*x***} in either list at any time.
854 expire_client(unconf);
856 new = create_client(clname, dname);
859 copy_clid(new, conf);
860 } else if (!unconf) {
862 * RFC 3530 14.2.33 CASE 2:
863 * probable client reboot; state will be removed if
866 new = create_client(clname, dname);
872 * RFC 3530 14.2.33 CASE 3:
873 * probable client reboot; state will be removed if
876 expire_client(unconf);
877 new = create_client(clname, dname);
882 copy_verf(new, &clverifier);
883 new->cl_addr = sin->sin_addr.s_addr;
884 new->cl_flavor = rqstp->rq_flavor;
885 princ = svc_gss_principal(rqstp);
887 new->cl_principal = kstrdup(princ, GFP_KERNEL);
888 if (new->cl_principal == NULL) {
893 copy_cred(&new->cl_cred, &rqstp->rq_cred);
895 gen_callback(new, setclid);
896 add_to_unconfirmed(new, strhashval);
897 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
898 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
899 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
908 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
909 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
910 * bullets, labeled as CASE1 - CASE4 below.
913 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
914 struct nfsd4_compound_state *cstate,
915 struct nfsd4_setclientid_confirm *setclientid_confirm)
917 struct sockaddr_in *sin = svc_addr_in(rqstp);
918 struct nfs4_client *conf, *unconf;
919 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
920 clientid_t * clid = &setclientid_confirm->sc_clientid;
923 if (STALE_CLIENTID(clid))
924 return nfserr_stale_clientid;
926 * XXX The Duplicate Request Cache (DRC) has been checked (??)
927 * We get here on a DRC miss.
932 conf = find_confirmed_client(clid);
933 unconf = find_unconfirmed_client(clid);
935 status = nfserr_clid_inuse;
936 if (conf && conf->cl_addr != sin->sin_addr.s_addr)
938 if (unconf && unconf->cl_addr != sin->sin_addr.s_addr)
942 * section 14.2.34 of RFC 3530 has a description of
943 * SETCLIENTID_CONFIRM request processing consisting
944 * of 4 bullet points, labeled as CASE1 - CASE4 below.
946 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
948 * RFC 3530 14.2.34 CASE 1:
951 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
952 status = nfserr_clid_inuse;
954 /* XXX: We just turn off callbacks until we can handle
955 * change request correctly. */
956 atomic_set(&conf->cl_callback.cb_set, 0);
958 nfsd4_remove_clid_dir(unconf);
959 expire_client(unconf);
963 } else if (conf && !unconf) {
965 * RFC 3530 14.2.34 CASE 2:
966 * probable retransmitted request; play it safe and
969 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
970 status = nfserr_clid_inuse;
973 } else if (!conf && unconf
974 && same_verf(&unconf->cl_confirm, &confirm)) {
976 * RFC 3530 14.2.34 CASE 3:
977 * Normal case; new or rebooted client:
979 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
980 status = nfserr_clid_inuse;
983 clientstr_hashval(unconf->cl_recdir);
984 conf = find_confirmed_client_by_str(unconf->cl_recdir,
987 nfsd4_remove_clid_dir(conf);
990 move_to_confirmed(unconf);
992 nfsd4_probe_callback(conf);
995 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
996 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
999 * RFC 3530 14.2.34 CASE 4:
1000 * Client probably hasn't noticed that we rebooted yet.
1002 status = nfserr_stale_clientid;
1004 /* check that we have hit one of the cases...*/
1005 status = nfserr_clid_inuse;
1008 nfs4_unlock_state();
1012 /* OPEN Share state helper functions */
1013 static inline struct nfs4_file *
1014 alloc_init_file(struct inode *ino)
1016 struct nfs4_file *fp;
1017 unsigned int hashval = file_hashval(ino);
1019 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1021 atomic_set(&fp->fi_ref, 1);
1022 INIT_LIST_HEAD(&fp->fi_hash);
1023 INIT_LIST_HEAD(&fp->fi_stateids);
1024 INIT_LIST_HEAD(&fp->fi_delegations);
1025 spin_lock(&recall_lock);
1026 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
1027 spin_unlock(&recall_lock);
1028 fp->fi_inode = igrab(ino);
1029 fp->fi_id = current_fileid++;
1030 fp->fi_had_conflict = false;
1037 nfsd4_free_slab(struct kmem_cache **slab)
1041 kmem_cache_destroy(*slab);
1046 nfsd4_free_slabs(void)
1048 nfsd4_free_slab(&stateowner_slab);
1049 nfsd4_free_slab(&file_slab);
1050 nfsd4_free_slab(&stateid_slab);
1051 nfsd4_free_slab(&deleg_slab);
1055 nfsd4_init_slabs(void)
1057 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
1058 sizeof(struct nfs4_stateowner), 0, 0, NULL);
1059 if (stateowner_slab == NULL)
1061 file_slab = kmem_cache_create("nfsd4_files",
1062 sizeof(struct nfs4_file), 0, 0, NULL);
1063 if (file_slab == NULL)
1065 stateid_slab = kmem_cache_create("nfsd4_stateids",
1066 sizeof(struct nfs4_stateid), 0, 0, NULL);
1067 if (stateid_slab == NULL)
1069 deleg_slab = kmem_cache_create("nfsd4_delegations",
1070 sizeof(struct nfs4_delegation), 0, 0, NULL);
1071 if (deleg_slab == NULL)
1076 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1081 nfs4_free_stateowner(struct kref *kref)
1083 struct nfs4_stateowner *sop =
1084 container_of(kref, struct nfs4_stateowner, so_ref);
1085 kfree(sop->so_owner.data);
1086 kmem_cache_free(stateowner_slab, sop);
1089 static inline struct nfs4_stateowner *
1090 alloc_stateowner(struct xdr_netobj *owner)
1092 struct nfs4_stateowner *sop;
1094 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1095 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1096 memcpy(sop->so_owner.data, owner->data, owner->len);
1097 sop->so_owner.len = owner->len;
1098 kref_init(&sop->so_ref);
1101 kmem_cache_free(stateowner_slab, sop);
1106 static struct nfs4_stateowner *
1107 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1108 struct nfs4_stateowner *sop;
1109 struct nfs4_replay *rp;
1110 unsigned int idhashval;
1112 if (!(sop = alloc_stateowner(&open->op_owner)))
1114 idhashval = ownerid_hashval(current_ownerid);
1115 INIT_LIST_HEAD(&sop->so_idhash);
1116 INIT_LIST_HEAD(&sop->so_strhash);
1117 INIT_LIST_HEAD(&sop->so_perclient);
1118 INIT_LIST_HEAD(&sop->so_stateids);
1119 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1120 INIT_LIST_HEAD(&sop->so_close_lru);
1122 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1123 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
1124 list_add(&sop->so_perclient, &clp->cl_openowners);
1125 sop->so_is_open_owner = 1;
1126 sop->so_id = current_ownerid++;
1127 sop->so_client = clp;
1128 sop->so_seqid = open->op_seqid;
1129 sop->so_confirmed = 0;
1130 rp = &sop->so_replay;
1131 rp->rp_status = nfserr_serverfault;
1133 rp->rp_buf = rp->rp_ibuf;
1138 init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1139 struct nfs4_stateowner *sop = open->op_stateowner;
1140 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1142 INIT_LIST_HEAD(&stp->st_hash);
1143 INIT_LIST_HEAD(&stp->st_perstateowner);
1144 INIT_LIST_HEAD(&stp->st_lockowners);
1145 INIT_LIST_HEAD(&stp->st_perfile);
1146 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
1147 list_add(&stp->st_perstateowner, &sop->so_stateids);
1148 list_add(&stp->st_perfile, &fp->fi_stateids);
1149 stp->st_stateowner = sop;
1152 stp->st_stateid.si_boot = boot_time;
1153 stp->st_stateid.si_stateownerid = sop->so_id;
1154 stp->st_stateid.si_fileid = fp->fi_id;
1155 stp->st_stateid.si_generation = 0;
1156 stp->st_access_bmap = 0;
1157 stp->st_deny_bmap = 0;
1158 __set_bit(open->op_share_access, &stp->st_access_bmap);
1159 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1160 stp->st_openstp = NULL;
1164 move_to_close_lru(struct nfs4_stateowner *sop)
1166 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1168 list_move_tail(&sop->so_close_lru, &close_lru);
1169 sop->so_time = get_seconds();
1173 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1176 return (sop->so_owner.len == owner->len) &&
1177 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1178 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1181 static struct nfs4_stateowner *
1182 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1184 struct nfs4_stateowner *so = NULL;
1186 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
1187 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1193 /* search file_hashtbl[] for file */
1194 static struct nfs4_file *
1195 find_file(struct inode *ino)
1197 unsigned int hashval = file_hashval(ino);
1198 struct nfs4_file *fp;
1200 spin_lock(&recall_lock);
1201 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
1202 if (fp->fi_inode == ino) {
1204 spin_unlock(&recall_lock);
1208 spin_unlock(&recall_lock);
1212 static inline int access_valid(u32 x)
1214 if (x < NFS4_SHARE_ACCESS_READ)
1216 if (x > NFS4_SHARE_ACCESS_BOTH)
1221 static inline int deny_valid(u32 x)
1223 /* Note: unlike access bits, deny bits may be zero. */
1224 return x <= NFS4_SHARE_DENY_BOTH;
1228 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1229 * st_{access,deny}_bmap field of the stateid, in order to track not
1230 * only what share bits are currently in force, but also what
1231 * combinations of share bits previous opens have used. This allows us
1232 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1233 * return an error if the client attempt to downgrade to a combination
1234 * of share bits not explicable by closing some of its previous opens.
1236 * XXX: This enforcement is actually incomplete, since we don't keep
1237 * track of access/deny bit combinations; so, e.g., we allow:
1239 * OPEN allow read, deny write
1240 * OPEN allow both, deny none
1241 * DOWNGRADE allow read, deny none
1243 * which we should reject.
1246 set_access(unsigned int *access, unsigned long bmap) {
1250 for (i = 1; i < 4; i++) {
1251 if (test_bit(i, &bmap))
1257 set_deny(unsigned int *deny, unsigned long bmap) {
1261 for (i = 0; i < 4; i++) {
1262 if (test_bit(i, &bmap))
1268 test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1269 unsigned int access, deny;
1271 set_access(&access, stp->st_access_bmap);
1272 set_deny(&deny, stp->st_deny_bmap);
1273 if ((access & open->op_share_deny) || (deny & open->op_share_access))
1279 * Called to check deny when READ with all zero stateid or
1280 * WRITE with all zero or all one stateid
1283 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
1285 struct inode *ino = current_fh->fh_dentry->d_inode;
1286 struct nfs4_file *fp;
1287 struct nfs4_stateid *stp;
1290 dprintk("NFSD: nfs4_share_conflict\n");
1292 fp = find_file(ino);
1295 ret = nfserr_locked;
1296 /* Search for conflicting share reservations */
1297 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
1298 if (test_bit(deny_type, &stp->st_deny_bmap) ||
1299 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
1309 nfs4_file_downgrade(struct file *filp, unsigned int share_access)
1311 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
1312 drop_file_write_access(filp);
1313 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
1318 * Recall a delegation
1321 do_recall(void *__dp)
1323 struct nfs4_delegation *dp = __dp;
1325 dp->dl_file->fi_had_conflict = true;
1326 nfsd4_cb_recall(dp);
1331 * Spawn a thread to perform a recall on the delegation represented
1332 * by the lease (file_lock)
1334 * Called from break_lease() with lock_kernel() held.
1335 * Note: we assume break_lease will only call this *once* for any given
1339 void nfsd_break_deleg_cb(struct file_lock *fl)
1341 struct nfs4_delegation *dp= (struct nfs4_delegation *)fl->fl_owner;
1342 struct task_struct *t;
1344 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
1348 /* We're assuming the state code never drops its reference
1349 * without first removing the lease. Since we're in this lease
1350 * callback (and since the lease code is serialized by the kernel
1351 * lock) we know the server hasn't removed the lease yet, we know
1352 * it's safe to take a reference: */
1353 atomic_inc(&dp->dl_count);
1354 atomic_inc(&dp->dl_client->cl_count);
1356 spin_lock(&recall_lock);
1357 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
1358 spin_unlock(&recall_lock);
1360 /* only place dl_time is set. protected by lock_kernel*/
1361 dp->dl_time = get_seconds();
1364 * We don't want the locks code to timeout the lease for us;
1365 * we'll remove it ourself if the delegation isn't returned
1368 fl->fl_break_time = 0;
1370 t = kthread_run(do_recall, dp, "%s", "nfs4_cb_recall");
1372 struct nfs4_client *clp = dp->dl_client;
1374 printk(KERN_INFO "NFSD: Callback thread failed for "
1375 "for client (clientid %08x/%08x)\n",
1376 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1377 put_nfs4_client(dp->dl_client);
1378 nfs4_put_delegation(dp);
1383 * The file_lock is being reapd.
1385 * Called by locks_free_lock() with lock_kernel() held.
1388 void nfsd_release_deleg_cb(struct file_lock *fl)
1390 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1392 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
1394 if (!(fl->fl_flags & FL_LEASE) || !dp)
1396 dp->dl_flock = NULL;
1400 * Set the delegation file_lock back pointer.
1402 * Called from setlease() with lock_kernel() held.
1405 void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
1407 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
1409 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
1416 * Called from setlease() with lock_kernel() held
1419 int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
1421 struct nfs4_delegation *onlistd =
1422 (struct nfs4_delegation *)onlist->fl_owner;
1423 struct nfs4_delegation *tryd =
1424 (struct nfs4_delegation *)try->fl_owner;
1426 if (onlist->fl_lmops != try->fl_lmops)
1429 return onlistd->dl_client == tryd->dl_client;
1434 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
1437 return lease_modify(onlist, arg);
1442 static struct lock_manager_operations nfsd_lease_mng_ops = {
1443 .fl_break = nfsd_break_deleg_cb,
1444 .fl_release_private = nfsd_release_deleg_cb,
1445 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
1446 .fl_mylease = nfsd_same_client_deleg_cb,
1447 .fl_change = nfsd_change_deleg_cb,
1452 nfsd4_process_open1(struct nfsd4_open *open)
1454 clientid_t *clientid = &open->op_clientid;
1455 struct nfs4_client *clp = NULL;
1456 unsigned int strhashval;
1457 struct nfs4_stateowner *sop = NULL;
1459 if (!check_name(open->op_owner))
1460 return nfserr_inval;
1462 if (STALE_CLIENTID(&open->op_clientid))
1463 return nfserr_stale_clientid;
1465 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
1466 sop = find_openstateowner_str(strhashval, open);
1467 open->op_stateowner = sop;
1469 /* Make sure the client's lease hasn't expired. */
1470 clp = find_confirmed_client(clientid);
1472 return nfserr_expired;
1475 if (!sop->so_confirmed) {
1476 /* Replace unconfirmed owners without checking for replay. */
1477 clp = sop->so_client;
1478 release_openowner(sop);
1479 open->op_stateowner = NULL;
1482 if (open->op_seqid == sop->so_seqid - 1) {
1483 if (sop->so_replay.rp_buflen)
1484 return nfserr_replay_me;
1485 /* The original OPEN failed so spectacularly
1486 * that we don't even have replay data saved!
1487 * Therefore, we have no choice but to continue
1488 * processing this OPEN; presumably, we'll
1489 * fail again for the same reason.
1491 dprintk("nfsd4_process_open1: replay with no replay cache\n");
1494 if (open->op_seqid != sop->so_seqid)
1495 return nfserr_bad_seqid;
1497 if (open->op_stateowner == NULL) {
1498 sop = alloc_init_open_stateowner(strhashval, clp, open);
1500 return nfserr_resource;
1501 open->op_stateowner = sop;
1503 list_del_init(&sop->so_close_lru);
1504 renew_client(sop->so_client);
1508 static inline __be32
1509 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
1511 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
1512 return nfserr_openmode;
1517 static struct nfs4_delegation *
1518 find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
1520 struct nfs4_delegation *dp;
1522 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
1523 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
1530 nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
1531 struct nfs4_delegation **dp)
1534 __be32 status = nfserr_bad_stateid;
1536 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
1539 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
1540 RD_STATE : WR_STATE;
1541 status = nfs4_check_delegmode(*dp, flags);
1545 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
1549 open->op_stateowner->so_confirmed = 1;
1554 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
1556 struct nfs4_stateid *local;
1557 __be32 status = nfserr_share_denied;
1558 struct nfs4_stateowner *sop = open->op_stateowner;
1560 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1561 /* ignore lock owners */
1562 if (local->st_stateowner->so_is_open_owner == 0)
1564 /* remember if we have seen this open owner */
1565 if (local->st_stateowner == sop)
1567 /* check for conflicting share reservations */
1568 if (!test_share(local, open))
1576 static inline struct nfs4_stateid *
1577 nfs4_alloc_stateid(void)
1579 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
1583 nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
1584 struct nfs4_delegation *dp,
1585 struct svc_fh *cur_fh, int flags)
1587 struct nfs4_stateid *stp;
1589 stp = nfs4_alloc_stateid();
1591 return nfserr_resource;
1594 get_file(dp->dl_vfs_file);
1595 stp->st_vfs_file = dp->dl_vfs_file;
1598 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
1601 if (status == nfserr_dropit)
1602 status = nfserr_jukebox;
1603 kmem_cache_free(stateid_slab, stp);
1611 static inline __be32
1612 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
1613 struct nfsd4_open *open)
1615 struct iattr iattr = {
1616 .ia_valid = ATTR_SIZE,
1619 if (!open->op_truncate)
1621 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
1622 return nfserr_inval;
1623 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
1627 nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
1629 struct file *filp = stp->st_vfs_file;
1630 struct inode *inode = filp->f_path.dentry->d_inode;
1631 unsigned int share_access, new_writer;
1634 set_access(&share_access, stp->st_access_bmap);
1635 new_writer = (~share_access) & open->op_share_access
1636 & NFS4_SHARE_ACCESS_WRITE;
1639 int err = get_write_access(inode);
1641 return nfserrno(err);
1642 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
1644 return nfserrno(err);
1645 file_take_write(filp);
1647 status = nfsd4_truncate(rqstp, cur_fh, open);
1650 put_write_access(inode);
1653 /* remember the open */
1654 filp->f_mode |= open->op_share_access;
1655 __set_bit(open->op_share_access, &stp->st_access_bmap);
1656 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1663 nfs4_set_claim_prev(struct nfsd4_open *open)
1665 open->op_stateowner->so_confirmed = 1;
1666 open->op_stateowner->so_client->cl_firststate = 1;
1670 * Attempt to hand out a delegation.
1673 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
1675 struct nfs4_delegation *dp;
1676 struct nfs4_stateowner *sop = stp->st_stateowner;
1677 struct nfs4_callback *cb = &sop->so_client->cl_callback;
1678 struct file_lock fl, *flp = &fl;
1679 int status, flag = 0;
1681 flag = NFS4_OPEN_DELEGATE_NONE;
1682 open->op_recall = 0;
1683 switch (open->op_claim_type) {
1684 case NFS4_OPEN_CLAIM_PREVIOUS:
1685 if (!atomic_read(&cb->cb_set))
1686 open->op_recall = 1;
1687 flag = open->op_delegate_type;
1688 if (flag == NFS4_OPEN_DELEGATE_NONE)
1691 case NFS4_OPEN_CLAIM_NULL:
1692 /* Let's not give out any delegations till everyone's
1693 * had the chance to reclaim theirs.... */
1694 if (locks_in_grace())
1696 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
1698 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
1699 flag = NFS4_OPEN_DELEGATE_WRITE;
1701 flag = NFS4_OPEN_DELEGATE_READ;
1707 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
1709 flag = NFS4_OPEN_DELEGATE_NONE;
1712 locks_init_lock(&fl);
1713 fl.fl_lmops = &nfsd_lease_mng_ops;
1714 fl.fl_flags = FL_LEASE;
1715 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1716 fl.fl_end = OFFSET_MAX;
1717 fl.fl_owner = (fl_owner_t)dp;
1718 fl.fl_file = stp->st_vfs_file;
1719 fl.fl_pid = current->tgid;
1721 /* vfs_setlease checks to see if delegation should be handed out.
1722 * the lock_manager callbacks fl_mylease and fl_change are used
1724 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1725 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
1726 unhash_delegation(dp);
1727 flag = NFS4_OPEN_DELEGATE_NONE;
1731 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
1733 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
1734 dp->dl_stateid.si_boot,
1735 dp->dl_stateid.si_stateownerid,
1736 dp->dl_stateid.si_fileid,
1737 dp->dl_stateid.si_generation);
1739 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
1740 && flag == NFS4_OPEN_DELEGATE_NONE
1741 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
1742 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1743 open->op_delegate_type = flag;
1747 * called with nfs4_lock_state() held.
1750 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
1752 struct nfs4_file *fp = NULL;
1753 struct inode *ino = current_fh->fh_dentry->d_inode;
1754 struct nfs4_stateid *stp = NULL;
1755 struct nfs4_delegation *dp = NULL;
1758 status = nfserr_inval;
1759 if (!access_valid(open->op_share_access)
1760 || !deny_valid(open->op_share_deny))
1763 * Lookup file; if found, lookup stateid and check open request,
1764 * and check for delegations in the process of being recalled.
1765 * If not found, create the nfs4_file struct
1767 fp = find_file(ino);
1769 if ((status = nfs4_check_open(fp, open, &stp)))
1771 status = nfs4_check_deleg(fp, open, &dp);
1775 status = nfserr_bad_stateid;
1776 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
1778 status = nfserr_resource;
1779 fp = alloc_init_file(ino);
1785 * OPEN the file, or upgrade an existing OPEN.
1786 * If truncate fails, the OPEN fails.
1789 /* Stateid was found, this is an OPEN upgrade */
1790 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
1793 update_stateid(&stp->st_stateid);
1795 /* Stateid was not found, this is a new OPEN */
1797 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
1798 flags |= NFSD_MAY_READ;
1799 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
1800 flags |= NFSD_MAY_WRITE;
1801 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
1804 init_stateid(stp, fp, open);
1805 status = nfsd4_truncate(rqstp, current_fh, open);
1807 release_open_stateid(stp);
1811 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
1814 * Attempt to hand out a delegation. No error return, because the
1815 * OPEN succeeds even if we fail.
1817 nfs4_open_delegation(current_fh, open, stp);
1821 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
1822 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
1823 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
1827 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
1828 nfs4_set_claim_prev(open);
1830 * To finish the open response, we just need to set the rflags.
1832 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
1833 if (!open->op_stateowner->so_confirmed)
1834 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
1840 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1843 struct nfs4_client *clp;
1847 dprintk("process_renew(%08x/%08x): starting\n",
1848 clid->cl_boot, clid->cl_id);
1849 status = nfserr_stale_clientid;
1850 if (STALE_CLIENTID(clid))
1852 clp = find_confirmed_client(clid);
1853 status = nfserr_expired;
1855 /* We assume the client took too long to RENEW. */
1856 dprintk("nfsd4_renew: clientid not found!\n");
1860 status = nfserr_cb_path_down;
1861 if (!list_empty(&clp->cl_delegations)
1862 && !atomic_read(&clp->cl_callback.cb_set))
1866 nfs4_unlock_state();
1870 struct lock_manager nfsd4_manager = {
1874 nfsd4_end_grace(void)
1876 dprintk("NFSD: end of grace period\n");
1877 nfsd4_recdir_purge_old();
1878 locks_end_grace(&nfsd4_manager);
1882 nfs4_laundromat(void)
1884 struct nfs4_client *clp;
1885 struct nfs4_stateowner *sop;
1886 struct nfs4_delegation *dp;
1887 struct list_head *pos, *next, reaplist;
1888 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
1889 time_t t, clientid_val = NFSD_LEASE_TIME;
1890 time_t u, test_val = NFSD_LEASE_TIME;
1894 dprintk("NFSD: laundromat service - starting\n");
1895 if (locks_in_grace())
1897 list_for_each_safe(pos, next, &client_lru) {
1898 clp = list_entry(pos, struct nfs4_client, cl_lru);
1899 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
1900 t = clp->cl_time - cutoff;
1901 if (clientid_val > t)
1905 dprintk("NFSD: purging unused client (clientid %08x)\n",
1906 clp->cl_clientid.cl_id);
1907 nfsd4_remove_clid_dir(clp);
1910 INIT_LIST_HEAD(&reaplist);
1911 spin_lock(&recall_lock);
1912 list_for_each_safe(pos, next, &del_recall_lru) {
1913 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
1914 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
1915 u = dp->dl_time - cutoff;
1920 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
1922 list_move(&dp->dl_recall_lru, &reaplist);
1924 spin_unlock(&recall_lock);
1925 list_for_each_safe(pos, next, &reaplist) {
1926 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
1927 list_del_init(&dp->dl_recall_lru);
1928 unhash_delegation(dp);
1930 test_val = NFSD_LEASE_TIME;
1931 list_for_each_safe(pos, next, &close_lru) {
1932 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
1933 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
1934 u = sop->so_time - cutoff;
1939 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
1941 release_openowner(sop);
1943 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
1944 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
1945 nfs4_unlock_state();
1946 return clientid_val;
1949 static struct workqueue_struct *laundry_wq;
1950 static void laundromat_main(struct work_struct *);
1951 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
1954 laundromat_main(struct work_struct *not_used)
1958 t = nfs4_laundromat();
1959 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
1960 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1963 static struct nfs4_stateowner *
1964 search_close_lru(u32 st_id, int flags)
1966 struct nfs4_stateowner *local = NULL;
1968 if (flags & CLOSE_STATE) {
1969 list_for_each_entry(local, &close_lru, so_close_lru) {
1970 if (local->so_id == st_id)
1978 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
1980 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1984 STALE_STATEID(stateid_t *stateid)
1986 if (stateid->si_boot == boot_time)
1988 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
1989 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
1990 stateid->si_generation);
1995 access_permit_read(unsigned long access_bmap)
1997 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
1998 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
1999 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2003 access_permit_write(unsigned long access_bmap)
2005 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2006 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2010 __be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
2012 __be32 status = nfserr_openmode;
2014 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2016 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2023 static inline __be32
2024 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2026 if (ONE_STATEID(stateid) && (flags & RD_STATE))
2028 else if (locks_in_grace()) {
2029 /* Answer in remaining cases depends on existance of
2030 * conflicting state; so we must wait out the grace period. */
2031 return nfserr_grace;
2032 } else if (flags & WR_STATE)
2033 return nfs4_share_conflict(current_fh,
2034 NFS4_SHARE_DENY_WRITE);
2035 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2036 return nfs4_share_conflict(current_fh,
2037 NFS4_SHARE_DENY_READ);
2041 * Allow READ/WRITE during grace period on recovered state only for files
2042 * that are not able to provide mandatory locking.
2045 grace_disallows_io(struct inode *inode)
2047 return locks_in_grace() && mandatory_lock(inode);
2050 static int check_stateid_generation(stateid_t *in, stateid_t *ref)
2052 /* If the client sends us a stateid from the future, it's buggy: */
2053 if (in->si_generation > ref->si_generation)
2054 return nfserr_bad_stateid;
2056 * The following, however, can happen. For example, if the
2057 * client sends an open and some IO at the same time, the open
2058 * may bump si_generation while the IO is still in flight.
2059 * Thanks to hard links and renames, the client never knows what
2060 * file an open will affect. So it could avoid that situation
2061 * only by serializing all opens and IO from the same open
2062 * owner. To recover from the old_stateid error, the client
2063 * will just have to retry the IO:
2065 if (in->si_generation < ref->si_generation)
2066 return nfserr_old_stateid;
2070 static int is_delegation_stateid(stateid_t *stateid)
2072 return stateid->si_fileid == 0;
2076 * Checks for stateid operations
2079 nfs4_preprocess_stateid_op(struct svc_fh *current_fh, stateid_t *stateid, int flags, struct file **filpp)
2081 struct nfs4_stateid *stp = NULL;
2082 struct nfs4_delegation *dp = NULL;
2083 struct inode *ino = current_fh->fh_dentry->d_inode;
2089 if (grace_disallows_io(ino))
2090 return nfserr_grace;
2092 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2093 return check_special_stateids(current_fh, stateid, flags);
2095 status = nfserr_stale_stateid;
2096 if (STALE_STATEID(stateid))
2099 status = nfserr_bad_stateid;
2100 if (is_delegation_stateid(stateid)) {
2101 dp = find_delegation_stateid(ino, stateid);
2104 status = check_stateid_generation(stateid, &dp->dl_stateid);
2107 status = nfs4_check_delegmode(dp, flags);
2110 renew_client(dp->dl_client);
2112 *filpp = dp->dl_vfs_file;
2113 } else { /* open or lock stateid */
2114 stp = find_stateid(stateid, flags);
2117 if (nfs4_check_fh(current_fh, stp))
2119 if (!stp->st_stateowner->so_confirmed)
2121 status = check_stateid_generation(stateid, &stp->st_stateid);
2124 status = nfs4_check_openmode(stp, flags);
2127 renew_client(stp->st_stateowner->so_client);
2129 *filpp = stp->st_vfs_file;
2139 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2140 RD_STATE : WR_STATE;
2144 * Checks for sequence id mutating operations.
2147 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)
2149 struct nfs4_stateid *stp;
2150 struct nfs4_stateowner *sop;
2153 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2154 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2155 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2156 stateid->si_generation);
2161 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2162 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
2163 return nfserr_bad_stateid;
2166 if (STALE_STATEID(stateid))
2167 return nfserr_stale_stateid;
2169 * We return BAD_STATEID if filehandle doesn't match stateid,
2170 * the confirmed flag is incorrecly set, or the generation
2171 * number is incorrect.
2173 stp = find_stateid(stateid, flags);
2176 * Also, we should make sure this isn't just the result of
2179 sop = search_close_lru(stateid->si_stateownerid, flags);
2181 return nfserr_bad_stateid;
2187 *sopp = sop = stp->st_stateowner;
2190 clientid_t *lockclid = &lock->v.new.clientid;
2191 struct nfs4_client *clp = sop->so_client;
2195 lkflg = setlkflg(lock->lk_type);
2197 if (lock->lk_is_new) {
2198 if (!sop->so_is_open_owner)
2199 return nfserr_bad_stateid;
2200 if (!same_clid(&clp->cl_clientid, lockclid))
2201 return nfserr_bad_stateid;
2202 /* stp is the open stateid */
2203 status = nfs4_check_openmode(stp, lkflg);
2207 /* stp is the lock stateid */
2208 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2214 if (nfs4_check_fh(current_fh, stp)) {
2215 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
2216 return nfserr_bad_stateid;
2220 * We now validate the seqid and stateid generation numbers.
2221 * For the moment, we ignore the possibility of
2222 * generation number wraparound.
2224 if (seqid != sop->so_seqid)
2227 if (sop->so_confirmed && flags & CONFIRM) {
2228 dprintk("NFSD: preprocess_seqid_op: expected"
2229 " unconfirmed stateowner!\n");
2230 return nfserr_bad_stateid;
2232 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2233 dprintk("NFSD: preprocess_seqid_op: stateowner not"
2234 " confirmed yet!\n");
2235 return nfserr_bad_stateid;
2237 status = check_stateid_generation(stateid, &stp->st_stateid);
2240 renew_client(sop->so_client);
2244 if (seqid == sop->so_seqid - 1) {
2245 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
2246 /* indicate replay to calling function */
2247 return nfserr_replay_me;
2249 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
2250 sop->so_seqid, seqid);
2252 return nfserr_bad_seqid;
2256 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2257 struct nfsd4_open_confirm *oc)
2260 struct nfs4_stateowner *sop;
2261 struct nfs4_stateid *stp;
2263 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
2264 (int)cstate->current_fh.fh_dentry->d_name.len,
2265 cstate->current_fh.fh_dentry->d_name.name);
2267 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
2273 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2274 oc->oc_seqid, &oc->oc_req_stateid,
2275 CONFIRM | OPEN_STATE,
2276 &oc->oc_stateowner, &stp, NULL)))
2279 sop = oc->oc_stateowner;
2280 sop->so_confirmed = 1;
2281 update_stateid(&stp->st_stateid);
2282 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
2283 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
2284 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
2285 stp->st_stateid.si_boot,
2286 stp->st_stateid.si_stateownerid,
2287 stp->st_stateid.si_fileid,
2288 stp->st_stateid.si_generation);
2290 nfsd4_create_clid_dir(sop->so_client);
2292 if (oc->oc_stateowner) {
2293 nfs4_get_stateowner(oc->oc_stateowner);
2294 cstate->replay_owner = oc->oc_stateowner;
2296 nfs4_unlock_state();
2302 * unset all bits in union bitmap (bmap) that
2303 * do not exist in share (from successful OPEN_DOWNGRADE)
2306 reset_union_bmap_access(unsigned long access, unsigned long *bmap)
2309 for (i = 1; i < 4; i++) {
2310 if ((i & access) != i)
2311 __clear_bit(i, bmap);
2316 reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
2319 for (i = 0; i < 4; i++) {
2320 if ((i & deny) != i)
2321 __clear_bit(i, bmap);
2326 nfsd4_open_downgrade(struct svc_rqst *rqstp,
2327 struct nfsd4_compound_state *cstate,
2328 struct nfsd4_open_downgrade *od)
2331 struct nfs4_stateid *stp;
2332 unsigned int share_access;
2334 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
2335 (int)cstate->current_fh.fh_dentry->d_name.len,
2336 cstate->current_fh.fh_dentry->d_name.name);
2338 if (!access_valid(od->od_share_access)
2339 || !deny_valid(od->od_share_deny))
2340 return nfserr_inval;
2343 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2347 &od->od_stateowner, &stp, NULL)))
2350 status = nfserr_inval;
2351 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
2352 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
2353 stp->st_access_bmap, od->od_share_access);
2356 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
2357 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
2358 stp->st_deny_bmap, od->od_share_deny);
2361 set_access(&share_access, stp->st_access_bmap);
2362 nfs4_file_downgrade(stp->st_vfs_file,
2363 share_access & ~od->od_share_access);
2365 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
2366 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
2368 update_stateid(&stp->st_stateid);
2369 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
2372 if (od->od_stateowner) {
2373 nfs4_get_stateowner(od->od_stateowner);
2374 cstate->replay_owner = od->od_stateowner;
2376 nfs4_unlock_state();
2381 * nfs4_unlock_state() called after encode
2384 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2385 struct nfsd4_close *close)
2388 struct nfs4_stateid *stp;
2390 dprintk("NFSD: nfsd4_close on file %.*s\n",
2391 (int)cstate->current_fh.fh_dentry->d_name.len,
2392 cstate->current_fh.fh_dentry->d_name.name);
2395 /* check close_lru for replay */
2396 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2399 OPEN_STATE | CLOSE_STATE,
2400 &close->cl_stateowner, &stp, NULL)))
2403 update_stateid(&stp->st_stateid);
2404 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
2406 /* release_stateid() calls nfsd_close() if needed */
2407 release_open_stateid(stp);
2409 /* place unused nfs4_stateowners on so_close_lru list to be
2410 * released by the laundromat service after the lease period
2411 * to enable us to handle CLOSE replay
2413 if (list_empty(&close->cl_stateowner->so_stateids))
2414 move_to_close_lru(close->cl_stateowner);
2416 if (close->cl_stateowner) {
2417 nfs4_get_stateowner(close->cl_stateowner);
2418 cstate->replay_owner = close->cl_stateowner;
2420 nfs4_unlock_state();
2425 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2426 struct nfsd4_delegreturn *dr)
2428 struct nfs4_delegation *dp;
2429 stateid_t *stateid = &dr->dr_stateid;
2430 struct inode *inode;
2433 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
2435 inode = cstate->current_fh.fh_dentry->d_inode;
2438 status = nfserr_bad_stateid;
2439 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2441 status = nfserr_stale_stateid;
2442 if (STALE_STATEID(stateid))
2444 status = nfserr_bad_stateid;
2445 if (!is_delegation_stateid(stateid))
2447 dp = find_delegation_stateid(inode, stateid);
2450 status = check_stateid_generation(stateid, &dp->dl_stateid);
2453 renew_client(dp->dl_client);
2455 unhash_delegation(dp);
2457 nfs4_unlock_state();
2464 * Lock owner state (byte-range locks)
2466 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
2467 #define LOCK_HASH_BITS 8
2468 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
2469 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
2472 end_offset(u64 start, u64 len)
2477 return end >= start ? end: NFS4_MAX_UINT64;
2480 /* last octet in a range */
2482 last_byte_offset(u64 start, u64 len)
2488 return end > start ? end - 1: NFS4_MAX_UINT64;
2491 #define lockownerid_hashval(id) \
2492 ((id) & LOCK_HASH_MASK)
2494 static inline unsigned int
2495 lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
2496 struct xdr_netobj *ownername)
2498 return (file_hashval(inode) + cl_id
2499 + opaque_hashval(ownername->data, ownername->len))
2503 static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
2504 static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
2505 static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
2507 static struct nfs4_stateid *
2508 find_stateid(stateid_t *stid, int flags)
2510 struct nfs4_stateid *local;
2511 u32 st_id = stid->si_stateownerid;
2512 u32 f_id = stid->si_fileid;
2513 unsigned int hashval;
2515 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
2516 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
2517 hashval = stateid_hashval(st_id, f_id);
2518 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
2519 if ((local->st_stateid.si_stateownerid == st_id) &&
2520 (local->st_stateid.si_fileid == f_id))
2525 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
2526 hashval = stateid_hashval(st_id, f_id);
2527 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
2528 if ((local->st_stateid.si_stateownerid == st_id) &&
2529 (local->st_stateid.si_fileid == f_id))
2536 static struct nfs4_delegation *
2537 find_delegation_stateid(struct inode *ino, stateid_t *stid)
2539 struct nfs4_file *fp;
2540 struct nfs4_delegation *dl;
2542 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
2543 stid->si_boot, stid->si_stateownerid,
2544 stid->si_fileid, stid->si_generation);
2546 fp = find_file(ino);
2549 dl = find_delegation_file(fp, stid);
2555 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
2556 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
2557 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
2558 * locking, this prevents us from being completely protocol-compliant. The
2559 * real solution to this problem is to start using unsigned file offsets in
2560 * the VFS, but this is a very deep change!
2563 nfs4_transform_lock_offset(struct file_lock *lock)
2565 if (lock->fl_start < 0)
2566 lock->fl_start = OFFSET_MAX;
2567 if (lock->fl_end < 0)
2568 lock->fl_end = OFFSET_MAX;
2571 /* Hack!: For now, we're defining this just so we can use a pointer to it
2572 * as a unique cookie to identify our (NFSv4's) posix locks. */
2573 static struct lock_manager_operations nfsd_posix_mng_ops = {
2577 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
2579 struct nfs4_stateowner *sop;
2582 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
2583 sop = (struct nfs4_stateowner *) fl->fl_owner;
2584 hval = lockownerid_hashval(sop->so_id);
2585 kref_get(&sop->so_ref);
2587 deny->ld_clientid = sop->so_client->cl_clientid;
2589 deny->ld_sop = NULL;
2590 deny->ld_clientid.cl_boot = 0;
2591 deny->ld_clientid.cl_id = 0;
2593 deny->ld_start = fl->fl_start;
2594 deny->ld_length = NFS4_MAX_UINT64;
2595 if (fl->fl_end != NFS4_MAX_UINT64)
2596 deny->ld_length = fl->fl_end - fl->fl_start + 1;
2597 deny->ld_type = NFS4_READ_LT;
2598 if (fl->fl_type != F_RDLCK)
2599 deny->ld_type = NFS4_WRITE_LT;
2602 static struct nfs4_stateowner *
2603 find_lockstateowner_str(struct inode *inode, clientid_t *clid,
2604 struct xdr_netobj *owner)
2606 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
2607 struct nfs4_stateowner *op;
2609 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
2610 if (same_owner_str(op, owner, clid))
2617 * Alloc a lock owner structure.
2618 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
2621 * strhashval = lock_ownerstr_hashval
2624 static struct nfs4_stateowner *
2625 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
2626 struct nfs4_stateowner *sop;
2627 struct nfs4_replay *rp;
2628 unsigned int idhashval;
2630 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
2632 idhashval = lockownerid_hashval(current_ownerid);
2633 INIT_LIST_HEAD(&sop->so_idhash);
2634 INIT_LIST_HEAD(&sop->so_strhash);
2635 INIT_LIST_HEAD(&sop->so_perclient);
2636 INIT_LIST_HEAD(&sop->so_stateids);
2637 INIT_LIST_HEAD(&sop->so_perstateid);
2638 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
2640 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
2641 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
2642 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
2643 sop->so_is_open_owner = 0;
2644 sop->so_id = current_ownerid++;
2645 sop->so_client = clp;
2646 /* It is the openowner seqid that will be incremented in encode in the
2647 * case of new lockowners; so increment the lock seqid manually: */
2648 sop->so_seqid = lock->lk_new_lock_seqid + 1;
2649 sop->so_confirmed = 1;
2650 rp = &sop->so_replay;
2651 rp->rp_status = nfserr_serverfault;
2653 rp->rp_buf = rp->rp_ibuf;
2657 static struct nfs4_stateid *
2658 alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
2660 struct nfs4_stateid *stp;
2661 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
2663 stp = nfs4_alloc_stateid();
2666 INIT_LIST_HEAD(&stp->st_hash);
2667 INIT_LIST_HEAD(&stp->st_perfile);
2668 INIT_LIST_HEAD(&stp->st_perstateowner);
2669 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
2670 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
2671 list_add(&stp->st_perfile, &fp->fi_stateids);
2672 list_add(&stp->st_perstateowner, &sop->so_stateids);
2673 stp->st_stateowner = sop;
2676 stp->st_stateid.si_boot = boot_time;
2677 stp->st_stateid.si_stateownerid = sop->so_id;
2678 stp->st_stateid.si_fileid = fp->fi_id;
2679 stp->st_stateid.si_generation = 0;
2680 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
2681 stp->st_access_bmap = open_stp->st_access_bmap;
2682 stp->st_deny_bmap = open_stp->st_deny_bmap;
2683 stp->st_openstp = open_stp;
2690 check_lock_length(u64 offset, u64 length)
2692 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
2693 LOFF_OVERFLOW(offset, length)));
2700 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2701 struct nfsd4_lock *lock)
2703 struct nfs4_stateowner *open_sop = NULL;
2704 struct nfs4_stateowner *lock_sop = NULL;
2705 struct nfs4_stateid *lock_stp;
2707 struct file_lock file_lock;
2708 struct file_lock conflock;
2710 unsigned int strhashval;
2714 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
2715 (long long) lock->lk_offset,
2716 (long long) lock->lk_length);
2718 if (check_lock_length(lock->lk_offset, lock->lk_length))
2719 return nfserr_inval;
2721 if ((status = fh_verify(rqstp, &cstate->current_fh,
2722 S_IFREG, NFSD_MAY_LOCK))) {
2723 dprintk("NFSD: nfsd4_lock: permission denied!\n");
2729 if (lock->lk_is_new) {
2731 * Client indicates that this is a new lockowner.
2732 * Use open owner and open stateid to create lock owner and
2735 struct nfs4_stateid *open_stp = NULL;
2736 struct nfs4_file *fp;
2738 status = nfserr_stale_clientid;
2739 if (STALE_CLIENTID(&lock->lk_new_clientid))
2742 /* validate and update open stateid and open seqid */
2743 status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2744 lock->lk_new_open_seqid,
2745 &lock->lk_new_open_stateid,
2747 &lock->lk_replay_owner, &open_stp,
2751 open_sop = lock->lk_replay_owner;
2752 /* create lockowner and lock stateid */
2753 fp = open_stp->st_file;
2754 strhashval = lock_ownerstr_hashval(fp->fi_inode,
2755 open_sop->so_client->cl_clientid.cl_id,
2756 &lock->v.new.owner);
2757 /* XXX: Do we need to check for duplicate stateowners on
2758 * the same file, or should they just be allowed (and
2759 * create new stateids)? */
2760 status = nfserr_resource;
2761 lock_sop = alloc_init_lock_stateowner(strhashval,
2762 open_sop->so_client, open_stp, lock);
2763 if (lock_sop == NULL)
2765 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
2766 if (lock_stp == NULL)
2769 /* lock (lock owner + lock stateid) already exists */
2770 status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2771 lock->lk_old_lock_seqid,
2772 &lock->lk_old_lock_stateid,
2774 &lock->lk_replay_owner, &lock_stp, lock);
2777 lock_sop = lock->lk_replay_owner;
2779 /* lock->lk_replay_owner and lock_stp have been created or found */
2780 filp = lock_stp->st_vfs_file;
2782 status = nfserr_grace;
2783 if (locks_in_grace() && !lock->lk_reclaim)
2785 status = nfserr_no_grace;
2786 if (!locks_in_grace() && lock->lk_reclaim)
2789 locks_init_lock(&file_lock);
2790 switch (lock->lk_type) {
2793 file_lock.fl_type = F_RDLCK;
2797 case NFS4_WRITEW_LT:
2798 file_lock.fl_type = F_WRLCK;
2802 status = nfserr_inval;
2805 file_lock.fl_owner = (fl_owner_t)lock_sop;
2806 file_lock.fl_pid = current->tgid;
2807 file_lock.fl_file = filp;
2808 file_lock.fl_flags = FL_POSIX;
2809 file_lock.fl_lmops = &nfsd_posix_mng_ops;
2811 file_lock.fl_start = lock->lk_offset;
2812 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
2813 nfs4_transform_lock_offset(&file_lock);
2816 * Try to lock the file in the VFS.
2817 * Note: locks.c uses the BKL to protect the inode's lock list.
2820 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
2822 case 0: /* success! */
2823 update_stateid(&lock_stp->st_stateid);
2824 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
2828 case (EAGAIN): /* conflock holds conflicting lock */
2829 status = nfserr_denied;
2830 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
2831 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
2834 status = nfserr_deadlock;
2837 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
2838 status = nfserr_resource;
2842 if (status && lock->lk_is_new && lock_sop)
2843 release_lockowner(lock_sop);
2844 if (lock->lk_replay_owner) {
2845 nfs4_get_stateowner(lock->lk_replay_owner);
2846 cstate->replay_owner = lock->lk_replay_owner;
2848 nfs4_unlock_state();
2853 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
2854 * so we do a temporary open here just to get an open file to pass to
2855 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
2858 static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
2863 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
2866 err = vfs_test_lock(file, lock);
2875 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2876 struct nfsd4_lockt *lockt)
2878 struct inode *inode;
2879 struct file_lock file_lock;
2883 if (locks_in_grace())
2884 return nfserr_grace;
2886 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
2887 return nfserr_inval;
2889 lockt->lt_stateowner = NULL;
2892 status = nfserr_stale_clientid;
2893 if (STALE_CLIENTID(&lockt->lt_clientid))
2896 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
2897 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
2898 if (status == nfserr_symlink)
2899 status = nfserr_inval;
2903 inode = cstate->current_fh.fh_dentry->d_inode;
2904 locks_init_lock(&file_lock);
2905 switch (lockt->lt_type) {
2908 file_lock.fl_type = F_RDLCK;
2911 case NFS4_WRITEW_LT:
2912 file_lock.fl_type = F_WRLCK;
2915 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
2916 status = nfserr_inval;
2920 lockt->lt_stateowner = find_lockstateowner_str(inode,
2921 &lockt->lt_clientid, &lockt->lt_owner);
2922 if (lockt->lt_stateowner)
2923 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
2924 file_lock.fl_pid = current->tgid;
2925 file_lock.fl_flags = FL_POSIX;
2927 file_lock.fl_start = lockt->lt_offset;
2928 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
2930 nfs4_transform_lock_offset(&file_lock);
2933 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
2935 status = nfserrno(error);
2938 if (file_lock.fl_type != F_UNLCK) {
2939 status = nfserr_denied;
2940 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
2943 nfs4_unlock_state();
2948 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2949 struct nfsd4_locku *locku)
2951 struct nfs4_stateid *stp;
2952 struct file *filp = NULL;
2953 struct file_lock file_lock;
2957 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
2958 (long long) locku->lu_offset,
2959 (long long) locku->lu_length);
2961 if (check_lock_length(locku->lu_offset, locku->lu_length))
2962 return nfserr_inval;
2966 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2970 &locku->lu_stateowner, &stp, NULL)))
2973 filp = stp->st_vfs_file;
2975 locks_init_lock(&file_lock);
2976 file_lock.fl_type = F_UNLCK;
2977 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
2978 file_lock.fl_pid = current->tgid;
2979 file_lock.fl_file = filp;
2980 file_lock.fl_flags = FL_POSIX;
2981 file_lock.fl_lmops = &nfsd_posix_mng_ops;
2982 file_lock.fl_start = locku->lu_offset;
2984 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
2985 nfs4_transform_lock_offset(&file_lock);
2988 * Try to unlock the file in the VFS.
2990 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
2992 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
2996 * OK, unlock succeeded; the only thing left to do is update the stateid.
2998 update_stateid(&stp->st_stateid);
2999 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3002 if (locku->lu_stateowner) {
3003 nfs4_get_stateowner(locku->lu_stateowner);
3004 cstate->replay_owner = locku->lu_stateowner;
3006 nfs4_unlock_state();
3010 status = nfserrno(err);
3016 * 1: locks held by lockowner
3017 * 0: no locks held by lockowner
3020 check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3022 struct file_lock **flpp;
3023 struct inode *inode = filp->f_path.dentry->d_inode;
3027 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
3028 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
3039 nfsd4_release_lockowner(struct svc_rqst *rqstp,
3040 struct nfsd4_compound_state *cstate,
3041 struct nfsd4_release_lockowner *rlockowner)
3043 clientid_t *clid = &rlockowner->rl_clientid;
3044 struct nfs4_stateowner *sop;
3045 struct nfs4_stateid *stp;
3046 struct xdr_netobj *owner = &rlockowner->rl_owner;
3047 struct list_head matches;
3051 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3052 clid->cl_boot, clid->cl_id);
3054 /* XXX check for lease expiration */
3056 status = nfserr_stale_clientid;
3057 if (STALE_CLIENTID(clid))
3062 status = nfserr_locks_held;
3063 /* XXX: we're doing a linear search through all the lockowners.
3064 * Yipes! For now we'll just hope clients aren't really using
3065 * release_lockowner much, but eventually we have to fix these
3066 * data structures. */
3067 INIT_LIST_HEAD(&matches);
3068 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3069 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
3070 if (!same_owner_str(sop, owner, clid))
3072 list_for_each_entry(stp, &sop->so_stateids,
3074 if (check_for_locks(stp->st_vfs_file, sop))
3076 /* Note: so_perclient unused for lockowners,
3077 * so it's OK to fool with here. */
3078 list_add(&sop->so_perclient, &matches);
3082 /* Clients probably won't expect us to return with some (but not all)
3083 * of the lockowner state released; so don't release any until all
3084 * have been checked. */
3086 while (!list_empty(&matches)) {
3087 sop = list_entry(matches.next, struct nfs4_stateowner,
3089 /* unhash_stateowner deletes so_perclient only
3090 * for openowners. */
3091 list_del(&sop->so_perclient);
3092 release_lockowner(sop);
3095 nfs4_unlock_state();
3099 static inline struct nfs4_client_reclaim *
3102 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
3106 nfs4_has_reclaimed_state(const char *name)
3108 unsigned int strhashval = clientstr_hashval(name);
3109 struct nfs4_client *clp;
3111 clp = find_confirmed_client_by_str(name, strhashval);
3116 * failure => all reset bets are off, nfserr_no_grace...
3119 nfs4_client_to_reclaim(const char *name)
3121 unsigned int strhashval;
3122 struct nfs4_client_reclaim *crp = NULL;
3124 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3125 crp = alloc_reclaim();
3128 strhashval = clientstr_hashval(name);
3129 INIT_LIST_HEAD(&crp->cr_strhash);
3130 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
3131 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
3132 reclaim_str_hashtbl_size++;
3137 nfs4_release_reclaim(void)
3139 struct nfs4_client_reclaim *crp = NULL;
3142 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3143 while (!list_empty(&reclaim_str_hashtbl[i])) {
3144 crp = list_entry(reclaim_str_hashtbl[i].next,
3145 struct nfs4_client_reclaim, cr_strhash);
3146 list_del(&crp->cr_strhash);
3148 reclaim_str_hashtbl_size--;
3151 BUG_ON(reclaim_str_hashtbl_size);
3155 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
3156 static struct nfs4_client_reclaim *
3157 nfs4_find_reclaim_client(clientid_t *clid)
3159 unsigned int strhashval;
3160 struct nfs4_client *clp;
3161 struct nfs4_client_reclaim *crp = NULL;
3164 /* find clientid in conf_id_hashtbl */
3165 clp = find_confirmed_client(clid);
3169 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3170 clp->cl_name.len, clp->cl_name.data,
3173 /* find clp->cl_name in reclaim_str_hashtbl */
3174 strhashval = clientstr_hashval(clp->cl_recdir);
3175 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
3176 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
3184 * Called from OPEN. Look for clientid in reclaim list.
3187 nfs4_check_open_reclaim(clientid_t *clid)
3189 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
3192 /* initialization to perform at module load time: */
3195 nfs4_state_init(void)
3199 status = nfsd4_init_slabs();
3202 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3203 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3204 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3205 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3206 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
3208 for (i = 0; i < FILE_HASH_SIZE; i++) {
3209 INIT_LIST_HEAD(&file_hashtbl[i]);
3211 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3212 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3213 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3215 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3216 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3217 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3219 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3220 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3221 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3223 memset(&onestateid, ~0, sizeof(stateid_t));
3224 INIT_LIST_HEAD(&close_lru);
3225 INIT_LIST_HEAD(&client_lru);
3226 INIT_LIST_HEAD(&del_recall_lru);
3227 for (i = 0; i < CLIENT_HASH_SIZE; i++)
3228 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
3229 reclaim_str_hashtbl_size = 0;
3234 nfsd4_load_reboot_recovery_data(void)
3239 nfsd4_init_recdir(user_recovery_dirname);
3240 status = nfsd4_recdir_load();
3241 nfs4_unlock_state();
3243 printk("NFSD: Failure reading reboot recovery data\n");
3247 get_nfs4_grace_period(void)
3249 return max(user_lease_time, lease_time) * HZ;
3253 * Since the lifetime of a delegation isn't limited to that of an open, a
3254 * client may quite reasonably hang on to a delegation as long as it has
3255 * the inode cached. This becomes an obvious problem the first time a
3256 * client's inode cache approaches the size of the server's total memory.
3258 * For now we avoid this problem by imposing a hard limit on the number
3259 * of delegations, which varies according to the server's memory size.
3262 set_max_delegations(void)
3265 * Allow at most 4 delegations per megabyte of RAM. Quick
3266 * estimates suggest that in the worst case (where every delegation
3267 * is for a different inode), a delegation could take about 1.5K,
3268 * giving a worst case usage of about 6% of memory.
3270 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
3273 /* initialization to perform when the nfsd service is started: */
3276 __nfs4_state_start(void)
3278 unsigned long grace_time;
3280 boot_time = get_seconds();
3281 grace_time = get_nfs4_grace_period();
3282 lease_time = user_lease_time;
3283 locks_start_grace(&nfsd4_manager);
3284 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
3286 laundry_wq = create_singlethread_workqueue("nfsd4");
3287 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
3288 set_max_delegations();
3292 nfs4_state_start(void)
3296 nfsd4_load_reboot_recovery_data();
3297 __nfs4_state_start();
3303 nfs4_lease_time(void)
3309 __nfs4_state_shutdown(void)
3312 struct nfs4_client *clp = NULL;
3313 struct nfs4_delegation *dp = NULL;
3314 struct list_head *pos, *next, reaplist;
3316 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3317 while (!list_empty(&conf_id_hashtbl[i])) {
3318 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
3321 while (!list_empty(&unconf_str_hashtbl[i])) {
3322 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
3326 INIT_LIST_HEAD(&reaplist);
3327 spin_lock(&recall_lock);
3328 list_for_each_safe(pos, next, &del_recall_lru) {
3329 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3330 list_move(&dp->dl_recall_lru, &reaplist);
3332 spin_unlock(&recall_lock);
3333 list_for_each_safe(pos, next, &reaplist) {
3334 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3335 list_del_init(&dp->dl_recall_lru);
3336 unhash_delegation(dp);
3339 nfsd4_shutdown_recdir();
3344 nfs4_state_shutdown(void)
3346 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
3347 destroy_workqueue(laundry_wq);
3348 locks_end_grace(&nfsd4_manager);
3350 nfs4_release_reclaim();
3351 __nfs4_state_shutdown();
3352 nfs4_unlock_state();
3356 * user_recovery_dirname is protected by the nfsd_mutex since it's only
3357 * accessed when nfsd is starting.
3360 nfs4_set_recdir(char *recdir)
3362 strcpy(user_recovery_dirname, recdir);
3366 * Change the NFSv4 recovery directory to recdir.
3369 nfs4_reset_recoverydir(char *recdir)
3374 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
3378 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
3379 nfs4_set_recdir(recdir);
3387 nfs4_recoverydir(void)
3389 return user_recovery_dirname;
3393 * Called when leasetime is changed.
3395 * The only way the protocol gives us to handle on-the-fly lease changes is to
3396 * simulate a reboot. Instead of doing that, we just wait till the next time
3397 * we start to register any changes in lease time. If the administrator
3398 * really wants to change the lease time *now*, they can go ahead and bring
3399 * nfsd down and then back up again after changing the lease time.
3401 * user_lease_time is protected by nfsd_mutex since it's only really accessed
3402 * when nfsd is starting
3405 nfs4_reset_lease(time_t leasetime)
3407 user_lease_time = leasetime;