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 put_nfs4_client(clp);
517 static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir)
519 struct nfs4_client *clp;
521 clp = alloc_client(name);
524 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
525 atomic_set(&clp->cl_count, 1);
526 atomic_set(&clp->cl_callback.cb_set, 0);
527 INIT_LIST_HEAD(&clp->cl_idhash);
528 INIT_LIST_HEAD(&clp->cl_strhash);
529 INIT_LIST_HEAD(&clp->cl_openowners);
530 INIT_LIST_HEAD(&clp->cl_delegations);
531 INIT_LIST_HEAD(&clp->cl_lru);
535 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
537 memcpy(target->cl_verifier.data, source->data,
538 sizeof(target->cl_verifier.data));
541 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
543 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
544 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
547 static void copy_cred(struct svc_cred *target, struct svc_cred *source)
549 target->cr_uid = source->cr_uid;
550 target->cr_gid = source->cr_gid;
551 target->cr_group_info = source->cr_group_info;
552 get_group_info(target->cr_group_info);
555 static int same_name(const char *n1, const char *n2)
557 return 0 == memcmp(n1, n2, HEXDIR_LEN);
561 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
563 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
567 same_clid(clientid_t *cl1, clientid_t *cl2)
569 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
572 /* XXX what about NGROUP */
574 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
576 return cr1->cr_uid == cr2->cr_uid;
579 static void gen_clid(struct nfs4_client *clp)
581 static u32 current_clientid = 1;
583 clp->cl_clientid.cl_boot = boot_time;
584 clp->cl_clientid.cl_id = current_clientid++;
587 static void gen_confirm(struct nfs4_client *clp)
592 p = (u32 *)clp->cl_confirm.data;
593 *p++ = get_seconds();
597 static int check_name(struct xdr_netobj name)
601 if (name.len > NFS4_OPAQUE_LIMIT) {
602 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
609 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
611 unsigned int idhashval;
613 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
614 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
615 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
616 list_add_tail(&clp->cl_lru, &client_lru);
617 clp->cl_time = get_seconds();
621 move_to_confirmed(struct nfs4_client *clp)
623 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
624 unsigned int strhashval;
626 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
627 list_del_init(&clp->cl_strhash);
628 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
629 strhashval = clientstr_hashval(clp->cl_recdir);
630 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
634 static struct nfs4_client *
635 find_confirmed_client(clientid_t *clid)
637 struct nfs4_client *clp;
638 unsigned int idhashval = clientid_hashval(clid->cl_id);
640 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
641 if (same_clid(&clp->cl_clientid, clid))
647 static struct nfs4_client *
648 find_unconfirmed_client(clientid_t *clid)
650 struct nfs4_client *clp;
651 unsigned int idhashval = clientid_hashval(clid->cl_id);
653 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
654 if (same_clid(&clp->cl_clientid, clid))
660 static struct nfs4_client *
661 find_confirmed_client_by_str(const char *dname, unsigned int hashval)
663 struct nfs4_client *clp;
665 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
666 if (same_name(clp->cl_recdir, dname))
672 static struct nfs4_client *
673 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval)
675 struct nfs4_client *clp;
677 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
678 if (same_name(clp->cl_recdir, dname))
684 /* a helper function for parse_callback */
686 parse_octet(unsigned int *lenp, char **addrp)
688 unsigned int len = *lenp;
700 if ((c < '0') || (c > '9')) {
706 n = (n * 10) + (c - '0');
717 /* parse and set the setclientid ipv4 callback address */
719 parse_ipv4(unsigned int addr_len, char *addr_val, unsigned int *cbaddrp, unsigned short *cbportp)
724 u32 addrlen = addr_len;
725 char *addr = addr_val;
730 for(i = 4; i > 0 ; i--) {
731 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
734 cbaddr |= (temp << shift);
742 for(i = 2; i > 0 ; i--) {
743 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
746 cbport |= (temp << shift);
755 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se)
757 struct nfs4_callback *cb = &clp->cl_callback;
759 /* Currently, we only support tcp for the callback channel */
760 if ((se->se_callback_netid_len != 3) || memcmp((char *)se->se_callback_netid_val, "tcp", 3))
763 if ( !(parse_ipv4(se->se_callback_addr_len, se->se_callback_addr_val,
764 &cb->cb_addr, &cb->cb_port)))
766 cb->cb_prog = se->se_callback_prog;
767 cb->cb_ident = se->se_callback_ident;
770 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
771 "will not receive delegations\n",
772 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
778 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
779 struct nfsd4_setclientid *setclid)
781 struct sockaddr_in *sin = svc_addr_in(rqstp);
782 struct xdr_netobj clname = {
783 .len = setclid->se_namelen,
784 .data = setclid->se_name,
786 nfs4_verifier clverifier = setclid->se_verf;
787 unsigned int strhashval;
788 struct nfs4_client *conf, *unconf, *new;
791 char dname[HEXDIR_LEN];
793 if (!check_name(clname))
796 status = nfs4_make_rec_clidname(dname, &clname);
801 * XXX The Duplicate Request Cache (DRC) has been checked (??)
802 * We get here on a DRC miss.
805 strhashval = clientstr_hashval(dname);
808 conf = find_confirmed_client_by_str(dname, strhashval);
810 /* RFC 3530 14.2.33 CASE 0: */
811 status = nfserr_clid_inuse;
812 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
813 dprintk("NFSD: setclientid: string in use by client"
814 " at %pI4\n", &conf->cl_addr);
819 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
820 * has a description of SETCLIENTID request processing consisting
821 * of 5 bullet points, labeled as CASE0 - CASE4 below.
823 unconf = find_unconfirmed_client_by_str(dname, strhashval);
824 status = nfserr_resource;
827 * RFC 3530 14.2.33 CASE 4:
828 * placed first, because it is the normal case
831 expire_client(unconf);
832 new = create_client(clname, dname);
836 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
838 * RFC 3530 14.2.33 CASE 1:
839 * probable callback update
842 /* Note this is removing unconfirmed {*x***},
843 * which is stronger than RFC recommended {vxc**}.
844 * This has the advantage that there is at most
845 * one {*x***} in either list at any time.
847 expire_client(unconf);
849 new = create_client(clname, dname);
852 copy_clid(new, conf);
853 } else if (!unconf) {
855 * RFC 3530 14.2.33 CASE 2:
856 * probable client reboot; state will be removed if
859 new = create_client(clname, dname);
865 * RFC 3530 14.2.33 CASE 3:
866 * probable client reboot; state will be removed if
869 expire_client(unconf);
870 new = create_client(clname, dname);
875 copy_verf(new, &clverifier);
876 new->cl_addr = sin->sin_addr.s_addr;
877 new->cl_flavor = rqstp->rq_flavor;
878 princ = svc_gss_principal(rqstp);
880 new->cl_principal = kstrdup(princ, GFP_KERNEL);
881 if (new->cl_principal == NULL) {
886 copy_cred(&new->cl_cred, &rqstp->rq_cred);
888 gen_callback(new, setclid);
889 add_to_unconfirmed(new, strhashval);
890 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
891 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
892 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
901 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
902 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
903 * bullets, labeled as CASE1 - CASE4 below.
906 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
907 struct nfsd4_compound_state *cstate,
908 struct nfsd4_setclientid_confirm *setclientid_confirm)
910 struct sockaddr_in *sin = svc_addr_in(rqstp);
911 struct nfs4_client *conf, *unconf;
912 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
913 clientid_t * clid = &setclientid_confirm->sc_clientid;
916 if (STALE_CLIENTID(clid))
917 return nfserr_stale_clientid;
919 * XXX The Duplicate Request Cache (DRC) has been checked (??)
920 * We get here on a DRC miss.
925 conf = find_confirmed_client(clid);
926 unconf = find_unconfirmed_client(clid);
928 status = nfserr_clid_inuse;
929 if (conf && conf->cl_addr != sin->sin_addr.s_addr)
931 if (unconf && unconf->cl_addr != sin->sin_addr.s_addr)
935 * section 14.2.34 of RFC 3530 has a description of
936 * SETCLIENTID_CONFIRM request processing consisting
937 * of 4 bullet points, labeled as CASE1 - CASE4 below.
939 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
941 * RFC 3530 14.2.34 CASE 1:
944 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
945 status = nfserr_clid_inuse;
947 /* XXX: We just turn off callbacks until we can handle
948 * change request correctly. */
949 atomic_set(&conf->cl_callback.cb_set, 0);
951 nfsd4_remove_clid_dir(unconf);
952 expire_client(unconf);
956 } else if (conf && !unconf) {
958 * RFC 3530 14.2.34 CASE 2:
959 * probable retransmitted request; play it safe and
962 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
963 status = nfserr_clid_inuse;
966 } else if (!conf && unconf
967 && same_verf(&unconf->cl_confirm, &confirm)) {
969 * RFC 3530 14.2.34 CASE 3:
970 * Normal case; new or rebooted client:
972 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
973 status = nfserr_clid_inuse;
976 clientstr_hashval(unconf->cl_recdir);
977 conf = find_confirmed_client_by_str(unconf->cl_recdir,
980 nfsd4_remove_clid_dir(conf);
983 move_to_confirmed(unconf);
985 nfsd4_probe_callback(conf);
988 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
989 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
992 * RFC 3530 14.2.34 CASE 4:
993 * Client probably hasn't noticed that we rebooted yet.
995 status = nfserr_stale_clientid;
997 /* check that we have hit one of the cases...*/
998 status = nfserr_clid_inuse;
1001 nfs4_unlock_state();
1005 /* OPEN Share state helper functions */
1006 static inline struct nfs4_file *
1007 alloc_init_file(struct inode *ino)
1009 struct nfs4_file *fp;
1010 unsigned int hashval = file_hashval(ino);
1012 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1014 atomic_set(&fp->fi_ref, 1);
1015 INIT_LIST_HEAD(&fp->fi_hash);
1016 INIT_LIST_HEAD(&fp->fi_stateids);
1017 INIT_LIST_HEAD(&fp->fi_delegations);
1018 spin_lock(&recall_lock);
1019 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
1020 spin_unlock(&recall_lock);
1021 fp->fi_inode = igrab(ino);
1022 fp->fi_id = current_fileid++;
1023 fp->fi_had_conflict = false;
1030 nfsd4_free_slab(struct kmem_cache **slab)
1034 kmem_cache_destroy(*slab);
1039 nfsd4_free_slabs(void)
1041 nfsd4_free_slab(&stateowner_slab);
1042 nfsd4_free_slab(&file_slab);
1043 nfsd4_free_slab(&stateid_slab);
1044 nfsd4_free_slab(&deleg_slab);
1048 nfsd4_init_slabs(void)
1050 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
1051 sizeof(struct nfs4_stateowner), 0, 0, NULL);
1052 if (stateowner_slab == NULL)
1054 file_slab = kmem_cache_create("nfsd4_files",
1055 sizeof(struct nfs4_file), 0, 0, NULL);
1056 if (file_slab == NULL)
1058 stateid_slab = kmem_cache_create("nfsd4_stateids",
1059 sizeof(struct nfs4_stateid), 0, 0, NULL);
1060 if (stateid_slab == NULL)
1062 deleg_slab = kmem_cache_create("nfsd4_delegations",
1063 sizeof(struct nfs4_delegation), 0, 0, NULL);
1064 if (deleg_slab == NULL)
1069 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1074 nfs4_free_stateowner(struct kref *kref)
1076 struct nfs4_stateowner *sop =
1077 container_of(kref, struct nfs4_stateowner, so_ref);
1078 kfree(sop->so_owner.data);
1079 kmem_cache_free(stateowner_slab, sop);
1082 static inline struct nfs4_stateowner *
1083 alloc_stateowner(struct xdr_netobj *owner)
1085 struct nfs4_stateowner *sop;
1087 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1088 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1089 memcpy(sop->so_owner.data, owner->data, owner->len);
1090 sop->so_owner.len = owner->len;
1091 kref_init(&sop->so_ref);
1094 kmem_cache_free(stateowner_slab, sop);
1099 static struct nfs4_stateowner *
1100 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1101 struct nfs4_stateowner *sop;
1102 struct nfs4_replay *rp;
1103 unsigned int idhashval;
1105 if (!(sop = alloc_stateowner(&open->op_owner)))
1107 idhashval = ownerid_hashval(current_ownerid);
1108 INIT_LIST_HEAD(&sop->so_idhash);
1109 INIT_LIST_HEAD(&sop->so_strhash);
1110 INIT_LIST_HEAD(&sop->so_perclient);
1111 INIT_LIST_HEAD(&sop->so_stateids);
1112 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1113 INIT_LIST_HEAD(&sop->so_close_lru);
1115 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1116 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
1117 list_add(&sop->so_perclient, &clp->cl_openowners);
1118 sop->so_is_open_owner = 1;
1119 sop->so_id = current_ownerid++;
1120 sop->so_client = clp;
1121 sop->so_seqid = open->op_seqid;
1122 sop->so_confirmed = 0;
1123 rp = &sop->so_replay;
1124 rp->rp_status = nfserr_serverfault;
1126 rp->rp_buf = rp->rp_ibuf;
1131 init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1132 struct nfs4_stateowner *sop = open->op_stateowner;
1133 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1135 INIT_LIST_HEAD(&stp->st_hash);
1136 INIT_LIST_HEAD(&stp->st_perstateowner);
1137 INIT_LIST_HEAD(&stp->st_lockowners);
1138 INIT_LIST_HEAD(&stp->st_perfile);
1139 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
1140 list_add(&stp->st_perstateowner, &sop->so_stateids);
1141 list_add(&stp->st_perfile, &fp->fi_stateids);
1142 stp->st_stateowner = sop;
1145 stp->st_stateid.si_boot = boot_time;
1146 stp->st_stateid.si_stateownerid = sop->so_id;
1147 stp->st_stateid.si_fileid = fp->fi_id;
1148 stp->st_stateid.si_generation = 0;
1149 stp->st_access_bmap = 0;
1150 stp->st_deny_bmap = 0;
1151 __set_bit(open->op_share_access, &stp->st_access_bmap);
1152 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1153 stp->st_openstp = NULL;
1157 move_to_close_lru(struct nfs4_stateowner *sop)
1159 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1161 list_move_tail(&sop->so_close_lru, &close_lru);
1162 sop->so_time = get_seconds();
1166 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1169 return (sop->so_owner.len == owner->len) &&
1170 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1171 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1174 static struct nfs4_stateowner *
1175 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1177 struct nfs4_stateowner *so = NULL;
1179 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
1180 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1186 /* search file_hashtbl[] for file */
1187 static struct nfs4_file *
1188 find_file(struct inode *ino)
1190 unsigned int hashval = file_hashval(ino);
1191 struct nfs4_file *fp;
1193 spin_lock(&recall_lock);
1194 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
1195 if (fp->fi_inode == ino) {
1197 spin_unlock(&recall_lock);
1201 spin_unlock(&recall_lock);
1205 static inline int access_valid(u32 x)
1207 if (x < NFS4_SHARE_ACCESS_READ)
1209 if (x > NFS4_SHARE_ACCESS_BOTH)
1214 static inline int deny_valid(u32 x)
1216 /* Note: unlike access bits, deny bits may be zero. */
1217 return x <= NFS4_SHARE_DENY_BOTH;
1221 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1222 * st_{access,deny}_bmap field of the stateid, in order to track not
1223 * only what share bits are currently in force, but also what
1224 * combinations of share bits previous opens have used. This allows us
1225 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1226 * return an error if the client attempt to downgrade to a combination
1227 * of share bits not explicable by closing some of its previous opens.
1229 * XXX: This enforcement is actually incomplete, since we don't keep
1230 * track of access/deny bit combinations; so, e.g., we allow:
1232 * OPEN allow read, deny write
1233 * OPEN allow both, deny none
1234 * DOWNGRADE allow read, deny none
1236 * which we should reject.
1239 set_access(unsigned int *access, unsigned long bmap) {
1243 for (i = 1; i < 4; i++) {
1244 if (test_bit(i, &bmap))
1250 set_deny(unsigned int *deny, unsigned long bmap) {
1254 for (i = 0; i < 4; i++) {
1255 if (test_bit(i, &bmap))
1261 test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1262 unsigned int access, deny;
1264 set_access(&access, stp->st_access_bmap);
1265 set_deny(&deny, stp->st_deny_bmap);
1266 if ((access & open->op_share_deny) || (deny & open->op_share_access))
1272 * Called to check deny when READ with all zero stateid or
1273 * WRITE with all zero or all one stateid
1276 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
1278 struct inode *ino = current_fh->fh_dentry->d_inode;
1279 struct nfs4_file *fp;
1280 struct nfs4_stateid *stp;
1283 dprintk("NFSD: nfs4_share_conflict\n");
1285 fp = find_file(ino);
1288 ret = nfserr_locked;
1289 /* Search for conflicting share reservations */
1290 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
1291 if (test_bit(deny_type, &stp->st_deny_bmap) ||
1292 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
1302 nfs4_file_downgrade(struct file *filp, unsigned int share_access)
1304 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
1305 drop_file_write_access(filp);
1306 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
1311 * Recall a delegation
1314 do_recall(void *__dp)
1316 struct nfs4_delegation *dp = __dp;
1318 dp->dl_file->fi_had_conflict = true;
1319 nfsd4_cb_recall(dp);
1324 * Spawn a thread to perform a recall on the delegation represented
1325 * by the lease (file_lock)
1327 * Called from break_lease() with lock_kernel() held.
1328 * Note: we assume break_lease will only call this *once* for any given
1332 void nfsd_break_deleg_cb(struct file_lock *fl)
1334 struct nfs4_delegation *dp= (struct nfs4_delegation *)fl->fl_owner;
1335 struct task_struct *t;
1337 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
1341 /* We're assuming the state code never drops its reference
1342 * without first removing the lease. Since we're in this lease
1343 * callback (and since the lease code is serialized by the kernel
1344 * lock) we know the server hasn't removed the lease yet, we know
1345 * it's safe to take a reference: */
1346 atomic_inc(&dp->dl_count);
1347 atomic_inc(&dp->dl_client->cl_count);
1349 spin_lock(&recall_lock);
1350 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
1351 spin_unlock(&recall_lock);
1353 /* only place dl_time is set. protected by lock_kernel*/
1354 dp->dl_time = get_seconds();
1357 * We don't want the locks code to timeout the lease for us;
1358 * we'll remove it ourself if the delegation isn't returned
1361 fl->fl_break_time = 0;
1363 t = kthread_run(do_recall, dp, "%s", "nfs4_cb_recall");
1365 struct nfs4_client *clp = dp->dl_client;
1367 printk(KERN_INFO "NFSD: Callback thread failed for "
1368 "for client (clientid %08x/%08x)\n",
1369 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1370 put_nfs4_client(dp->dl_client);
1371 nfs4_put_delegation(dp);
1376 * The file_lock is being reapd.
1378 * Called by locks_free_lock() with lock_kernel() held.
1381 void nfsd_release_deleg_cb(struct file_lock *fl)
1383 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1385 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
1387 if (!(fl->fl_flags & FL_LEASE) || !dp)
1389 dp->dl_flock = NULL;
1393 * Set the delegation file_lock back pointer.
1395 * Called from setlease() with lock_kernel() held.
1398 void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
1400 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
1402 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
1409 * Called from setlease() with lock_kernel() held
1412 int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
1414 struct nfs4_delegation *onlistd =
1415 (struct nfs4_delegation *)onlist->fl_owner;
1416 struct nfs4_delegation *tryd =
1417 (struct nfs4_delegation *)try->fl_owner;
1419 if (onlist->fl_lmops != try->fl_lmops)
1422 return onlistd->dl_client == tryd->dl_client;
1427 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
1430 return lease_modify(onlist, arg);
1435 static struct lock_manager_operations nfsd_lease_mng_ops = {
1436 .fl_break = nfsd_break_deleg_cb,
1437 .fl_release_private = nfsd_release_deleg_cb,
1438 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
1439 .fl_mylease = nfsd_same_client_deleg_cb,
1440 .fl_change = nfsd_change_deleg_cb,
1445 nfsd4_process_open1(struct nfsd4_open *open)
1447 clientid_t *clientid = &open->op_clientid;
1448 struct nfs4_client *clp = NULL;
1449 unsigned int strhashval;
1450 struct nfs4_stateowner *sop = NULL;
1452 if (!check_name(open->op_owner))
1453 return nfserr_inval;
1455 if (STALE_CLIENTID(&open->op_clientid))
1456 return nfserr_stale_clientid;
1458 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
1459 sop = find_openstateowner_str(strhashval, open);
1460 open->op_stateowner = sop;
1462 /* Make sure the client's lease hasn't expired. */
1463 clp = find_confirmed_client(clientid);
1465 return nfserr_expired;
1468 if (!sop->so_confirmed) {
1469 /* Replace unconfirmed owners without checking for replay. */
1470 clp = sop->so_client;
1471 release_openowner(sop);
1472 open->op_stateowner = NULL;
1475 if (open->op_seqid == sop->so_seqid - 1) {
1476 if (sop->so_replay.rp_buflen)
1477 return nfserr_replay_me;
1478 /* The original OPEN failed so spectacularly
1479 * that we don't even have replay data saved!
1480 * Therefore, we have no choice but to continue
1481 * processing this OPEN; presumably, we'll
1482 * fail again for the same reason.
1484 dprintk("nfsd4_process_open1: replay with no replay cache\n");
1487 if (open->op_seqid != sop->so_seqid)
1488 return nfserr_bad_seqid;
1490 if (open->op_stateowner == NULL) {
1491 sop = alloc_init_open_stateowner(strhashval, clp, open);
1493 return nfserr_resource;
1494 open->op_stateowner = sop;
1496 list_del_init(&sop->so_close_lru);
1497 renew_client(sop->so_client);
1501 static inline __be32
1502 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
1504 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
1505 return nfserr_openmode;
1510 static struct nfs4_delegation *
1511 find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
1513 struct nfs4_delegation *dp;
1515 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
1516 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
1523 nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
1524 struct nfs4_delegation **dp)
1527 __be32 status = nfserr_bad_stateid;
1529 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
1532 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
1533 RD_STATE : WR_STATE;
1534 status = nfs4_check_delegmode(*dp, flags);
1538 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
1542 open->op_stateowner->so_confirmed = 1;
1547 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
1549 struct nfs4_stateid *local;
1550 __be32 status = nfserr_share_denied;
1551 struct nfs4_stateowner *sop = open->op_stateowner;
1553 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1554 /* ignore lock owners */
1555 if (local->st_stateowner->so_is_open_owner == 0)
1557 /* remember if we have seen this open owner */
1558 if (local->st_stateowner == sop)
1560 /* check for conflicting share reservations */
1561 if (!test_share(local, open))
1569 static inline struct nfs4_stateid *
1570 nfs4_alloc_stateid(void)
1572 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
1576 nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
1577 struct nfs4_delegation *dp,
1578 struct svc_fh *cur_fh, int flags)
1580 struct nfs4_stateid *stp;
1582 stp = nfs4_alloc_stateid();
1584 return nfserr_resource;
1587 get_file(dp->dl_vfs_file);
1588 stp->st_vfs_file = dp->dl_vfs_file;
1591 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
1594 if (status == nfserr_dropit)
1595 status = nfserr_jukebox;
1596 kmem_cache_free(stateid_slab, stp);
1604 static inline __be32
1605 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
1606 struct nfsd4_open *open)
1608 struct iattr iattr = {
1609 .ia_valid = ATTR_SIZE,
1612 if (!open->op_truncate)
1614 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
1615 return nfserr_inval;
1616 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
1620 nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
1622 struct file *filp = stp->st_vfs_file;
1623 struct inode *inode = filp->f_path.dentry->d_inode;
1624 unsigned int share_access, new_writer;
1627 set_access(&share_access, stp->st_access_bmap);
1628 new_writer = (~share_access) & open->op_share_access
1629 & NFS4_SHARE_ACCESS_WRITE;
1632 int err = get_write_access(inode);
1634 return nfserrno(err);
1635 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
1637 return nfserrno(err);
1638 file_take_write(filp);
1640 status = nfsd4_truncate(rqstp, cur_fh, open);
1643 put_write_access(inode);
1646 /* remember the open */
1647 filp->f_mode |= open->op_share_access;
1648 __set_bit(open->op_share_access, &stp->st_access_bmap);
1649 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1656 nfs4_set_claim_prev(struct nfsd4_open *open)
1658 open->op_stateowner->so_confirmed = 1;
1659 open->op_stateowner->so_client->cl_firststate = 1;
1663 * Attempt to hand out a delegation.
1666 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
1668 struct nfs4_delegation *dp;
1669 struct nfs4_stateowner *sop = stp->st_stateowner;
1670 struct nfs4_callback *cb = &sop->so_client->cl_callback;
1671 struct file_lock fl, *flp = &fl;
1672 int status, flag = 0;
1674 flag = NFS4_OPEN_DELEGATE_NONE;
1675 open->op_recall = 0;
1676 switch (open->op_claim_type) {
1677 case NFS4_OPEN_CLAIM_PREVIOUS:
1678 if (!atomic_read(&cb->cb_set))
1679 open->op_recall = 1;
1680 flag = open->op_delegate_type;
1681 if (flag == NFS4_OPEN_DELEGATE_NONE)
1684 case NFS4_OPEN_CLAIM_NULL:
1685 /* Let's not give out any delegations till everyone's
1686 * had the chance to reclaim theirs.... */
1687 if (locks_in_grace())
1689 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
1691 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
1692 flag = NFS4_OPEN_DELEGATE_WRITE;
1694 flag = NFS4_OPEN_DELEGATE_READ;
1700 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
1702 flag = NFS4_OPEN_DELEGATE_NONE;
1705 locks_init_lock(&fl);
1706 fl.fl_lmops = &nfsd_lease_mng_ops;
1707 fl.fl_flags = FL_LEASE;
1708 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1709 fl.fl_end = OFFSET_MAX;
1710 fl.fl_owner = (fl_owner_t)dp;
1711 fl.fl_file = stp->st_vfs_file;
1712 fl.fl_pid = current->tgid;
1714 /* vfs_setlease checks to see if delegation should be handed out.
1715 * the lock_manager callbacks fl_mylease and fl_change are used
1717 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1718 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
1719 unhash_delegation(dp);
1720 flag = NFS4_OPEN_DELEGATE_NONE;
1724 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
1726 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
1727 dp->dl_stateid.si_boot,
1728 dp->dl_stateid.si_stateownerid,
1729 dp->dl_stateid.si_fileid,
1730 dp->dl_stateid.si_generation);
1732 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
1733 && flag == NFS4_OPEN_DELEGATE_NONE
1734 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
1735 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1736 open->op_delegate_type = flag;
1740 * called with nfs4_lock_state() held.
1743 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
1745 struct nfs4_file *fp = NULL;
1746 struct inode *ino = current_fh->fh_dentry->d_inode;
1747 struct nfs4_stateid *stp = NULL;
1748 struct nfs4_delegation *dp = NULL;
1751 status = nfserr_inval;
1752 if (!access_valid(open->op_share_access)
1753 || !deny_valid(open->op_share_deny))
1756 * Lookup file; if found, lookup stateid and check open request,
1757 * and check for delegations in the process of being recalled.
1758 * If not found, create the nfs4_file struct
1760 fp = find_file(ino);
1762 if ((status = nfs4_check_open(fp, open, &stp)))
1764 status = nfs4_check_deleg(fp, open, &dp);
1768 status = nfserr_bad_stateid;
1769 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
1771 status = nfserr_resource;
1772 fp = alloc_init_file(ino);
1778 * OPEN the file, or upgrade an existing OPEN.
1779 * If truncate fails, the OPEN fails.
1782 /* Stateid was found, this is an OPEN upgrade */
1783 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
1786 update_stateid(&stp->st_stateid);
1788 /* Stateid was not found, this is a new OPEN */
1790 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
1791 flags |= NFSD_MAY_READ;
1792 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
1793 flags |= NFSD_MAY_WRITE;
1794 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
1797 init_stateid(stp, fp, open);
1798 status = nfsd4_truncate(rqstp, current_fh, open);
1800 release_open_stateid(stp);
1804 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
1807 * Attempt to hand out a delegation. No error return, because the
1808 * OPEN succeeds even if we fail.
1810 nfs4_open_delegation(current_fh, open, stp);
1814 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
1815 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
1816 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
1820 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
1821 nfs4_set_claim_prev(open);
1823 * To finish the open response, we just need to set the rflags.
1825 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
1826 if (!open->op_stateowner->so_confirmed)
1827 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
1833 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1836 struct nfs4_client *clp;
1840 dprintk("process_renew(%08x/%08x): starting\n",
1841 clid->cl_boot, clid->cl_id);
1842 status = nfserr_stale_clientid;
1843 if (STALE_CLIENTID(clid))
1845 clp = find_confirmed_client(clid);
1846 status = nfserr_expired;
1848 /* We assume the client took too long to RENEW. */
1849 dprintk("nfsd4_renew: clientid not found!\n");
1853 status = nfserr_cb_path_down;
1854 if (!list_empty(&clp->cl_delegations)
1855 && !atomic_read(&clp->cl_callback.cb_set))
1859 nfs4_unlock_state();
1863 struct lock_manager nfsd4_manager = {
1867 nfsd4_end_grace(void)
1869 dprintk("NFSD: end of grace period\n");
1870 nfsd4_recdir_purge_old();
1871 locks_end_grace(&nfsd4_manager);
1875 nfs4_laundromat(void)
1877 struct nfs4_client *clp;
1878 struct nfs4_stateowner *sop;
1879 struct nfs4_delegation *dp;
1880 struct list_head *pos, *next, reaplist;
1881 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
1882 time_t t, clientid_val = NFSD_LEASE_TIME;
1883 time_t u, test_val = NFSD_LEASE_TIME;
1887 dprintk("NFSD: laundromat service - starting\n");
1888 if (locks_in_grace())
1890 list_for_each_safe(pos, next, &client_lru) {
1891 clp = list_entry(pos, struct nfs4_client, cl_lru);
1892 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
1893 t = clp->cl_time - cutoff;
1894 if (clientid_val > t)
1898 dprintk("NFSD: purging unused client (clientid %08x)\n",
1899 clp->cl_clientid.cl_id);
1900 nfsd4_remove_clid_dir(clp);
1903 INIT_LIST_HEAD(&reaplist);
1904 spin_lock(&recall_lock);
1905 list_for_each_safe(pos, next, &del_recall_lru) {
1906 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
1907 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
1908 u = dp->dl_time - cutoff;
1913 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
1915 list_move(&dp->dl_recall_lru, &reaplist);
1917 spin_unlock(&recall_lock);
1918 list_for_each_safe(pos, next, &reaplist) {
1919 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
1920 list_del_init(&dp->dl_recall_lru);
1921 unhash_delegation(dp);
1923 test_val = NFSD_LEASE_TIME;
1924 list_for_each_safe(pos, next, &close_lru) {
1925 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
1926 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
1927 u = sop->so_time - cutoff;
1932 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
1934 release_openowner(sop);
1936 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
1937 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
1938 nfs4_unlock_state();
1939 return clientid_val;
1942 static struct workqueue_struct *laundry_wq;
1943 static void laundromat_main(struct work_struct *);
1944 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
1947 laundromat_main(struct work_struct *not_used)
1951 t = nfs4_laundromat();
1952 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
1953 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1956 static struct nfs4_stateowner *
1957 search_close_lru(u32 st_id, int flags)
1959 struct nfs4_stateowner *local = NULL;
1961 if (flags & CLOSE_STATE) {
1962 list_for_each_entry(local, &close_lru, so_close_lru) {
1963 if (local->so_id == st_id)
1971 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
1973 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1977 STALE_STATEID(stateid_t *stateid)
1979 if (stateid->si_boot == boot_time)
1981 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
1982 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
1983 stateid->si_generation);
1988 access_permit_read(unsigned long access_bmap)
1990 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
1991 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
1992 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
1996 access_permit_write(unsigned long access_bmap)
1998 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
1999 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2003 __be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
2005 __be32 status = nfserr_openmode;
2007 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2009 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2016 static inline __be32
2017 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2019 if (ONE_STATEID(stateid) && (flags & RD_STATE))
2021 else if (locks_in_grace()) {
2022 /* Answer in remaining cases depends on existance of
2023 * conflicting state; so we must wait out the grace period. */
2024 return nfserr_grace;
2025 } else if (flags & WR_STATE)
2026 return nfs4_share_conflict(current_fh,
2027 NFS4_SHARE_DENY_WRITE);
2028 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2029 return nfs4_share_conflict(current_fh,
2030 NFS4_SHARE_DENY_READ);
2034 * Allow READ/WRITE during grace period on recovered state only for files
2035 * that are not able to provide mandatory locking.
2038 grace_disallows_io(struct inode *inode)
2040 return locks_in_grace() && mandatory_lock(inode);
2043 static int check_stateid_generation(stateid_t *in, stateid_t *ref)
2045 /* If the client sends us a stateid from the future, it's buggy: */
2046 if (in->si_generation > ref->si_generation)
2047 return nfserr_bad_stateid;
2049 * The following, however, can happen. For example, if the
2050 * client sends an open and some IO at the same time, the open
2051 * may bump si_generation while the IO is still in flight.
2052 * Thanks to hard links and renames, the client never knows what
2053 * file an open will affect. So it could avoid that situation
2054 * only by serializing all opens and IO from the same open
2055 * owner. To recover from the old_stateid error, the client
2056 * will just have to retry the IO:
2058 if (in->si_generation < ref->si_generation)
2059 return nfserr_old_stateid;
2063 static int is_delegation_stateid(stateid_t *stateid)
2065 return stateid->si_fileid == 0;
2069 * Checks for stateid operations
2072 nfs4_preprocess_stateid_op(struct svc_fh *current_fh, stateid_t *stateid, int flags, struct file **filpp)
2074 struct nfs4_stateid *stp = NULL;
2075 struct nfs4_delegation *dp = NULL;
2076 struct inode *ino = current_fh->fh_dentry->d_inode;
2082 if (grace_disallows_io(ino))
2083 return nfserr_grace;
2085 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2086 return check_special_stateids(current_fh, stateid, flags);
2088 status = nfserr_stale_stateid;
2089 if (STALE_STATEID(stateid))
2092 status = nfserr_bad_stateid;
2093 if (is_delegation_stateid(stateid)) {
2094 dp = find_delegation_stateid(ino, stateid);
2097 status = check_stateid_generation(stateid, &dp->dl_stateid);
2100 status = nfs4_check_delegmode(dp, flags);
2103 renew_client(dp->dl_client);
2105 *filpp = dp->dl_vfs_file;
2106 } else { /* open or lock stateid */
2107 stp = find_stateid(stateid, flags);
2110 if (nfs4_check_fh(current_fh, stp))
2112 if (!stp->st_stateowner->so_confirmed)
2114 status = check_stateid_generation(stateid, &stp->st_stateid);
2117 status = nfs4_check_openmode(stp, flags);
2120 renew_client(stp->st_stateowner->so_client);
2122 *filpp = stp->st_vfs_file;
2132 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2133 RD_STATE : WR_STATE;
2137 * Checks for sequence id mutating operations.
2140 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)
2142 struct nfs4_stateid *stp;
2143 struct nfs4_stateowner *sop;
2146 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2147 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2148 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2149 stateid->si_generation);
2154 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2155 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
2156 return nfserr_bad_stateid;
2159 if (STALE_STATEID(stateid))
2160 return nfserr_stale_stateid;
2162 * We return BAD_STATEID if filehandle doesn't match stateid,
2163 * the confirmed flag is incorrecly set, or the generation
2164 * number is incorrect.
2166 stp = find_stateid(stateid, flags);
2169 * Also, we should make sure this isn't just the result of
2172 sop = search_close_lru(stateid->si_stateownerid, flags);
2174 return nfserr_bad_stateid;
2180 *sopp = sop = stp->st_stateowner;
2183 clientid_t *lockclid = &lock->v.new.clientid;
2184 struct nfs4_client *clp = sop->so_client;
2188 lkflg = setlkflg(lock->lk_type);
2190 if (lock->lk_is_new) {
2191 if (!sop->so_is_open_owner)
2192 return nfserr_bad_stateid;
2193 if (!same_clid(&clp->cl_clientid, lockclid))
2194 return nfserr_bad_stateid;
2195 /* stp is the open stateid */
2196 status = nfs4_check_openmode(stp, lkflg);
2200 /* stp is the lock stateid */
2201 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2207 if (nfs4_check_fh(current_fh, stp)) {
2208 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
2209 return nfserr_bad_stateid;
2213 * We now validate the seqid and stateid generation numbers.
2214 * For the moment, we ignore the possibility of
2215 * generation number wraparound.
2217 if (seqid != sop->so_seqid)
2220 if (sop->so_confirmed && flags & CONFIRM) {
2221 dprintk("NFSD: preprocess_seqid_op: expected"
2222 " unconfirmed stateowner!\n");
2223 return nfserr_bad_stateid;
2225 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2226 dprintk("NFSD: preprocess_seqid_op: stateowner not"
2227 " confirmed yet!\n");
2228 return nfserr_bad_stateid;
2230 status = check_stateid_generation(stateid, &stp->st_stateid);
2233 renew_client(sop->so_client);
2237 if (seqid == sop->so_seqid - 1) {
2238 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
2239 /* indicate replay to calling function */
2240 return nfserr_replay_me;
2242 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
2243 sop->so_seqid, seqid);
2245 return nfserr_bad_seqid;
2249 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2250 struct nfsd4_open_confirm *oc)
2253 struct nfs4_stateowner *sop;
2254 struct nfs4_stateid *stp;
2256 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
2257 (int)cstate->current_fh.fh_dentry->d_name.len,
2258 cstate->current_fh.fh_dentry->d_name.name);
2260 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
2266 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2267 oc->oc_seqid, &oc->oc_req_stateid,
2268 CONFIRM | OPEN_STATE,
2269 &oc->oc_stateowner, &stp, NULL)))
2272 sop = oc->oc_stateowner;
2273 sop->so_confirmed = 1;
2274 update_stateid(&stp->st_stateid);
2275 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
2276 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
2277 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
2278 stp->st_stateid.si_boot,
2279 stp->st_stateid.si_stateownerid,
2280 stp->st_stateid.si_fileid,
2281 stp->st_stateid.si_generation);
2283 nfsd4_create_clid_dir(sop->so_client);
2285 if (oc->oc_stateowner) {
2286 nfs4_get_stateowner(oc->oc_stateowner);
2287 cstate->replay_owner = oc->oc_stateowner;
2289 nfs4_unlock_state();
2295 * unset all bits in union bitmap (bmap) that
2296 * do not exist in share (from successful OPEN_DOWNGRADE)
2299 reset_union_bmap_access(unsigned long access, unsigned long *bmap)
2302 for (i = 1; i < 4; i++) {
2303 if ((i & access) != i)
2304 __clear_bit(i, bmap);
2309 reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
2312 for (i = 0; i < 4; i++) {
2313 if ((i & deny) != i)
2314 __clear_bit(i, bmap);
2319 nfsd4_open_downgrade(struct svc_rqst *rqstp,
2320 struct nfsd4_compound_state *cstate,
2321 struct nfsd4_open_downgrade *od)
2324 struct nfs4_stateid *stp;
2325 unsigned int share_access;
2327 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
2328 (int)cstate->current_fh.fh_dentry->d_name.len,
2329 cstate->current_fh.fh_dentry->d_name.name);
2331 if (!access_valid(od->od_share_access)
2332 || !deny_valid(od->od_share_deny))
2333 return nfserr_inval;
2336 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2340 &od->od_stateowner, &stp, NULL)))
2343 status = nfserr_inval;
2344 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
2345 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
2346 stp->st_access_bmap, od->od_share_access);
2349 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
2350 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
2351 stp->st_deny_bmap, od->od_share_deny);
2354 set_access(&share_access, stp->st_access_bmap);
2355 nfs4_file_downgrade(stp->st_vfs_file,
2356 share_access & ~od->od_share_access);
2358 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
2359 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
2361 update_stateid(&stp->st_stateid);
2362 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
2365 if (od->od_stateowner) {
2366 nfs4_get_stateowner(od->od_stateowner);
2367 cstate->replay_owner = od->od_stateowner;
2369 nfs4_unlock_state();
2374 * nfs4_unlock_state() called after encode
2377 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2378 struct nfsd4_close *close)
2381 struct nfs4_stateid *stp;
2383 dprintk("NFSD: nfsd4_close on file %.*s\n",
2384 (int)cstate->current_fh.fh_dentry->d_name.len,
2385 cstate->current_fh.fh_dentry->d_name.name);
2388 /* check close_lru for replay */
2389 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2392 OPEN_STATE | CLOSE_STATE,
2393 &close->cl_stateowner, &stp, NULL)))
2396 update_stateid(&stp->st_stateid);
2397 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
2399 /* release_stateid() calls nfsd_close() if needed */
2400 release_open_stateid(stp);
2402 /* place unused nfs4_stateowners on so_close_lru list to be
2403 * released by the laundromat service after the lease period
2404 * to enable us to handle CLOSE replay
2406 if (list_empty(&close->cl_stateowner->so_stateids))
2407 move_to_close_lru(close->cl_stateowner);
2409 if (close->cl_stateowner) {
2410 nfs4_get_stateowner(close->cl_stateowner);
2411 cstate->replay_owner = close->cl_stateowner;
2413 nfs4_unlock_state();
2418 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2419 struct nfsd4_delegreturn *dr)
2421 struct nfs4_delegation *dp;
2422 stateid_t *stateid = &dr->dr_stateid;
2423 struct inode *inode;
2426 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
2428 inode = cstate->current_fh.fh_dentry->d_inode;
2431 status = nfserr_bad_stateid;
2432 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2434 status = nfserr_stale_stateid;
2435 if (STALE_STATEID(stateid))
2437 status = nfserr_bad_stateid;
2438 if (!is_delegation_stateid(stateid))
2440 dp = find_delegation_stateid(inode, stateid);
2443 status = check_stateid_generation(stateid, &dp->dl_stateid);
2446 renew_client(dp->dl_client);
2448 unhash_delegation(dp);
2450 nfs4_unlock_state();
2457 * Lock owner state (byte-range locks)
2459 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
2460 #define LOCK_HASH_BITS 8
2461 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
2462 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
2465 end_offset(u64 start, u64 len)
2470 return end >= start ? end: NFS4_MAX_UINT64;
2473 /* last octet in a range */
2475 last_byte_offset(u64 start, u64 len)
2481 return end > start ? end - 1: NFS4_MAX_UINT64;
2484 #define lockownerid_hashval(id) \
2485 ((id) & LOCK_HASH_MASK)
2487 static inline unsigned int
2488 lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
2489 struct xdr_netobj *ownername)
2491 return (file_hashval(inode) + cl_id
2492 + opaque_hashval(ownername->data, ownername->len))
2496 static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
2497 static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
2498 static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
2500 static struct nfs4_stateid *
2501 find_stateid(stateid_t *stid, int flags)
2503 struct nfs4_stateid *local;
2504 u32 st_id = stid->si_stateownerid;
2505 u32 f_id = stid->si_fileid;
2506 unsigned int hashval;
2508 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
2509 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
2510 hashval = stateid_hashval(st_id, f_id);
2511 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
2512 if ((local->st_stateid.si_stateownerid == st_id) &&
2513 (local->st_stateid.si_fileid == f_id))
2518 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
2519 hashval = stateid_hashval(st_id, f_id);
2520 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
2521 if ((local->st_stateid.si_stateownerid == st_id) &&
2522 (local->st_stateid.si_fileid == f_id))
2529 static struct nfs4_delegation *
2530 find_delegation_stateid(struct inode *ino, stateid_t *stid)
2532 struct nfs4_file *fp;
2533 struct nfs4_delegation *dl;
2535 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
2536 stid->si_boot, stid->si_stateownerid,
2537 stid->si_fileid, stid->si_generation);
2539 fp = find_file(ino);
2542 dl = find_delegation_file(fp, stid);
2548 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
2549 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
2550 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
2551 * locking, this prevents us from being completely protocol-compliant. The
2552 * real solution to this problem is to start using unsigned file offsets in
2553 * the VFS, but this is a very deep change!
2556 nfs4_transform_lock_offset(struct file_lock *lock)
2558 if (lock->fl_start < 0)
2559 lock->fl_start = OFFSET_MAX;
2560 if (lock->fl_end < 0)
2561 lock->fl_end = OFFSET_MAX;
2564 /* Hack!: For now, we're defining this just so we can use a pointer to it
2565 * as a unique cookie to identify our (NFSv4's) posix locks. */
2566 static struct lock_manager_operations nfsd_posix_mng_ops = {
2570 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
2572 struct nfs4_stateowner *sop;
2575 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
2576 sop = (struct nfs4_stateowner *) fl->fl_owner;
2577 hval = lockownerid_hashval(sop->so_id);
2578 kref_get(&sop->so_ref);
2580 deny->ld_clientid = sop->so_client->cl_clientid;
2582 deny->ld_sop = NULL;
2583 deny->ld_clientid.cl_boot = 0;
2584 deny->ld_clientid.cl_id = 0;
2586 deny->ld_start = fl->fl_start;
2587 deny->ld_length = NFS4_MAX_UINT64;
2588 if (fl->fl_end != NFS4_MAX_UINT64)
2589 deny->ld_length = fl->fl_end - fl->fl_start + 1;
2590 deny->ld_type = NFS4_READ_LT;
2591 if (fl->fl_type != F_RDLCK)
2592 deny->ld_type = NFS4_WRITE_LT;
2595 static struct nfs4_stateowner *
2596 find_lockstateowner_str(struct inode *inode, clientid_t *clid,
2597 struct xdr_netobj *owner)
2599 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
2600 struct nfs4_stateowner *op;
2602 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
2603 if (same_owner_str(op, owner, clid))
2610 * Alloc a lock owner structure.
2611 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
2614 * strhashval = lock_ownerstr_hashval
2617 static struct nfs4_stateowner *
2618 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
2619 struct nfs4_stateowner *sop;
2620 struct nfs4_replay *rp;
2621 unsigned int idhashval;
2623 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
2625 idhashval = lockownerid_hashval(current_ownerid);
2626 INIT_LIST_HEAD(&sop->so_idhash);
2627 INIT_LIST_HEAD(&sop->so_strhash);
2628 INIT_LIST_HEAD(&sop->so_perclient);
2629 INIT_LIST_HEAD(&sop->so_stateids);
2630 INIT_LIST_HEAD(&sop->so_perstateid);
2631 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
2633 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
2634 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
2635 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
2636 sop->so_is_open_owner = 0;
2637 sop->so_id = current_ownerid++;
2638 sop->so_client = clp;
2639 /* It is the openowner seqid that will be incremented in encode in the
2640 * case of new lockowners; so increment the lock seqid manually: */
2641 sop->so_seqid = lock->lk_new_lock_seqid + 1;
2642 sop->so_confirmed = 1;
2643 rp = &sop->so_replay;
2644 rp->rp_status = nfserr_serverfault;
2646 rp->rp_buf = rp->rp_ibuf;
2650 static struct nfs4_stateid *
2651 alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
2653 struct nfs4_stateid *stp;
2654 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
2656 stp = nfs4_alloc_stateid();
2659 INIT_LIST_HEAD(&stp->st_hash);
2660 INIT_LIST_HEAD(&stp->st_perfile);
2661 INIT_LIST_HEAD(&stp->st_perstateowner);
2662 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
2663 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
2664 list_add(&stp->st_perfile, &fp->fi_stateids);
2665 list_add(&stp->st_perstateowner, &sop->so_stateids);
2666 stp->st_stateowner = sop;
2669 stp->st_stateid.si_boot = boot_time;
2670 stp->st_stateid.si_stateownerid = sop->so_id;
2671 stp->st_stateid.si_fileid = fp->fi_id;
2672 stp->st_stateid.si_generation = 0;
2673 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
2674 stp->st_access_bmap = open_stp->st_access_bmap;
2675 stp->st_deny_bmap = open_stp->st_deny_bmap;
2676 stp->st_openstp = open_stp;
2683 check_lock_length(u64 offset, u64 length)
2685 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
2686 LOFF_OVERFLOW(offset, length)));
2693 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2694 struct nfsd4_lock *lock)
2696 struct nfs4_stateowner *open_sop = NULL;
2697 struct nfs4_stateowner *lock_sop = NULL;
2698 struct nfs4_stateid *lock_stp;
2700 struct file_lock file_lock;
2701 struct file_lock conflock;
2703 unsigned int strhashval;
2707 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
2708 (long long) lock->lk_offset,
2709 (long long) lock->lk_length);
2711 if (check_lock_length(lock->lk_offset, lock->lk_length))
2712 return nfserr_inval;
2714 if ((status = fh_verify(rqstp, &cstate->current_fh,
2715 S_IFREG, NFSD_MAY_LOCK))) {
2716 dprintk("NFSD: nfsd4_lock: permission denied!\n");
2722 if (lock->lk_is_new) {
2724 * Client indicates that this is a new lockowner.
2725 * Use open owner and open stateid to create lock owner and
2728 struct nfs4_stateid *open_stp = NULL;
2729 struct nfs4_file *fp;
2731 status = nfserr_stale_clientid;
2732 if (STALE_CLIENTID(&lock->lk_new_clientid))
2735 /* validate and update open stateid and open seqid */
2736 status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2737 lock->lk_new_open_seqid,
2738 &lock->lk_new_open_stateid,
2740 &lock->lk_replay_owner, &open_stp,
2744 open_sop = lock->lk_replay_owner;
2745 /* create lockowner and lock stateid */
2746 fp = open_stp->st_file;
2747 strhashval = lock_ownerstr_hashval(fp->fi_inode,
2748 open_sop->so_client->cl_clientid.cl_id,
2749 &lock->v.new.owner);
2750 /* XXX: Do we need to check for duplicate stateowners on
2751 * the same file, or should they just be allowed (and
2752 * create new stateids)? */
2753 status = nfserr_resource;
2754 lock_sop = alloc_init_lock_stateowner(strhashval,
2755 open_sop->so_client, open_stp, lock);
2756 if (lock_sop == NULL)
2758 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
2759 if (lock_stp == NULL)
2762 /* lock (lock owner + lock stateid) already exists */
2763 status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2764 lock->lk_old_lock_seqid,
2765 &lock->lk_old_lock_stateid,
2767 &lock->lk_replay_owner, &lock_stp, lock);
2770 lock_sop = lock->lk_replay_owner;
2772 /* lock->lk_replay_owner and lock_stp have been created or found */
2773 filp = lock_stp->st_vfs_file;
2775 status = nfserr_grace;
2776 if (locks_in_grace() && !lock->lk_reclaim)
2778 status = nfserr_no_grace;
2779 if (!locks_in_grace() && lock->lk_reclaim)
2782 locks_init_lock(&file_lock);
2783 switch (lock->lk_type) {
2786 file_lock.fl_type = F_RDLCK;
2790 case NFS4_WRITEW_LT:
2791 file_lock.fl_type = F_WRLCK;
2795 status = nfserr_inval;
2798 file_lock.fl_owner = (fl_owner_t)lock_sop;
2799 file_lock.fl_pid = current->tgid;
2800 file_lock.fl_file = filp;
2801 file_lock.fl_flags = FL_POSIX;
2802 file_lock.fl_lmops = &nfsd_posix_mng_ops;
2804 file_lock.fl_start = lock->lk_offset;
2805 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
2806 nfs4_transform_lock_offset(&file_lock);
2809 * Try to lock the file in the VFS.
2810 * Note: locks.c uses the BKL to protect the inode's lock list.
2813 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
2815 case 0: /* success! */
2816 update_stateid(&lock_stp->st_stateid);
2817 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
2821 case (EAGAIN): /* conflock holds conflicting lock */
2822 status = nfserr_denied;
2823 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
2824 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
2827 status = nfserr_deadlock;
2830 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
2831 status = nfserr_resource;
2835 if (status && lock->lk_is_new && lock_sop)
2836 release_lockowner(lock_sop);
2837 if (lock->lk_replay_owner) {
2838 nfs4_get_stateowner(lock->lk_replay_owner);
2839 cstate->replay_owner = lock->lk_replay_owner;
2841 nfs4_unlock_state();
2846 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
2847 * so we do a temporary open here just to get an open file to pass to
2848 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
2851 static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
2856 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
2859 err = vfs_test_lock(file, lock);
2868 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2869 struct nfsd4_lockt *lockt)
2871 struct inode *inode;
2872 struct file_lock file_lock;
2876 if (locks_in_grace())
2877 return nfserr_grace;
2879 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
2880 return nfserr_inval;
2882 lockt->lt_stateowner = NULL;
2885 status = nfserr_stale_clientid;
2886 if (STALE_CLIENTID(&lockt->lt_clientid))
2889 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
2890 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
2891 if (status == nfserr_symlink)
2892 status = nfserr_inval;
2896 inode = cstate->current_fh.fh_dentry->d_inode;
2897 locks_init_lock(&file_lock);
2898 switch (lockt->lt_type) {
2901 file_lock.fl_type = F_RDLCK;
2904 case NFS4_WRITEW_LT:
2905 file_lock.fl_type = F_WRLCK;
2908 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
2909 status = nfserr_inval;
2913 lockt->lt_stateowner = find_lockstateowner_str(inode,
2914 &lockt->lt_clientid, &lockt->lt_owner);
2915 if (lockt->lt_stateowner)
2916 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
2917 file_lock.fl_pid = current->tgid;
2918 file_lock.fl_flags = FL_POSIX;
2920 file_lock.fl_start = lockt->lt_offset;
2921 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
2923 nfs4_transform_lock_offset(&file_lock);
2926 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
2928 status = nfserrno(error);
2931 if (file_lock.fl_type != F_UNLCK) {
2932 status = nfserr_denied;
2933 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
2936 nfs4_unlock_state();
2941 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2942 struct nfsd4_locku *locku)
2944 struct nfs4_stateid *stp;
2945 struct file *filp = NULL;
2946 struct file_lock file_lock;
2950 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
2951 (long long) locku->lu_offset,
2952 (long long) locku->lu_length);
2954 if (check_lock_length(locku->lu_offset, locku->lu_length))
2955 return nfserr_inval;
2959 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2963 &locku->lu_stateowner, &stp, NULL)))
2966 filp = stp->st_vfs_file;
2968 locks_init_lock(&file_lock);
2969 file_lock.fl_type = F_UNLCK;
2970 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
2971 file_lock.fl_pid = current->tgid;
2972 file_lock.fl_file = filp;
2973 file_lock.fl_flags = FL_POSIX;
2974 file_lock.fl_lmops = &nfsd_posix_mng_ops;
2975 file_lock.fl_start = locku->lu_offset;
2977 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
2978 nfs4_transform_lock_offset(&file_lock);
2981 * Try to unlock the file in the VFS.
2983 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
2985 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
2989 * OK, unlock succeeded; the only thing left to do is update the stateid.
2991 update_stateid(&stp->st_stateid);
2992 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
2995 if (locku->lu_stateowner) {
2996 nfs4_get_stateowner(locku->lu_stateowner);
2997 cstate->replay_owner = locku->lu_stateowner;
2999 nfs4_unlock_state();
3003 status = nfserrno(err);
3009 * 1: locks held by lockowner
3010 * 0: no locks held by lockowner
3013 check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3015 struct file_lock **flpp;
3016 struct inode *inode = filp->f_path.dentry->d_inode;
3020 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
3021 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
3032 nfsd4_release_lockowner(struct svc_rqst *rqstp,
3033 struct nfsd4_compound_state *cstate,
3034 struct nfsd4_release_lockowner *rlockowner)
3036 clientid_t *clid = &rlockowner->rl_clientid;
3037 struct nfs4_stateowner *sop;
3038 struct nfs4_stateid *stp;
3039 struct xdr_netobj *owner = &rlockowner->rl_owner;
3040 struct list_head matches;
3044 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3045 clid->cl_boot, clid->cl_id);
3047 /* XXX check for lease expiration */
3049 status = nfserr_stale_clientid;
3050 if (STALE_CLIENTID(clid))
3055 status = nfserr_locks_held;
3056 /* XXX: we're doing a linear search through all the lockowners.
3057 * Yipes! For now we'll just hope clients aren't really using
3058 * release_lockowner much, but eventually we have to fix these
3059 * data structures. */
3060 INIT_LIST_HEAD(&matches);
3061 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3062 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
3063 if (!same_owner_str(sop, owner, clid))
3065 list_for_each_entry(stp, &sop->so_stateids,
3067 if (check_for_locks(stp->st_vfs_file, sop))
3069 /* Note: so_perclient unused for lockowners,
3070 * so it's OK to fool with here. */
3071 list_add(&sop->so_perclient, &matches);
3075 /* Clients probably won't expect us to return with some (but not all)
3076 * of the lockowner state released; so don't release any until all
3077 * have been checked. */
3079 while (!list_empty(&matches)) {
3080 sop = list_entry(matches.next, struct nfs4_stateowner,
3082 /* unhash_stateowner deletes so_perclient only
3083 * for openowners. */
3084 list_del(&sop->so_perclient);
3085 release_lockowner(sop);
3088 nfs4_unlock_state();
3092 static inline struct nfs4_client_reclaim *
3095 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
3099 nfs4_has_reclaimed_state(const char *name)
3101 unsigned int strhashval = clientstr_hashval(name);
3102 struct nfs4_client *clp;
3104 clp = find_confirmed_client_by_str(name, strhashval);
3109 * failure => all reset bets are off, nfserr_no_grace...
3112 nfs4_client_to_reclaim(const char *name)
3114 unsigned int strhashval;
3115 struct nfs4_client_reclaim *crp = NULL;
3117 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3118 crp = alloc_reclaim();
3121 strhashval = clientstr_hashval(name);
3122 INIT_LIST_HEAD(&crp->cr_strhash);
3123 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
3124 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
3125 reclaim_str_hashtbl_size++;
3130 nfs4_release_reclaim(void)
3132 struct nfs4_client_reclaim *crp = NULL;
3135 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3136 while (!list_empty(&reclaim_str_hashtbl[i])) {
3137 crp = list_entry(reclaim_str_hashtbl[i].next,
3138 struct nfs4_client_reclaim, cr_strhash);
3139 list_del(&crp->cr_strhash);
3141 reclaim_str_hashtbl_size--;
3144 BUG_ON(reclaim_str_hashtbl_size);
3148 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
3149 static struct nfs4_client_reclaim *
3150 nfs4_find_reclaim_client(clientid_t *clid)
3152 unsigned int strhashval;
3153 struct nfs4_client *clp;
3154 struct nfs4_client_reclaim *crp = NULL;
3157 /* find clientid in conf_id_hashtbl */
3158 clp = find_confirmed_client(clid);
3162 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3163 clp->cl_name.len, clp->cl_name.data,
3166 /* find clp->cl_name in reclaim_str_hashtbl */
3167 strhashval = clientstr_hashval(clp->cl_recdir);
3168 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
3169 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
3177 * Called from OPEN. Look for clientid in reclaim list.
3180 nfs4_check_open_reclaim(clientid_t *clid)
3182 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
3185 /* initialization to perform at module load time: */
3188 nfs4_state_init(void)
3192 status = nfsd4_init_slabs();
3195 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3196 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3197 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3198 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3199 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
3201 for (i = 0; i < FILE_HASH_SIZE; i++) {
3202 INIT_LIST_HEAD(&file_hashtbl[i]);
3204 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3205 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3206 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3208 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3209 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3210 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3212 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3213 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3214 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3216 memset(&onestateid, ~0, sizeof(stateid_t));
3217 INIT_LIST_HEAD(&close_lru);
3218 INIT_LIST_HEAD(&client_lru);
3219 INIT_LIST_HEAD(&del_recall_lru);
3220 for (i = 0; i < CLIENT_HASH_SIZE; i++)
3221 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
3222 reclaim_str_hashtbl_size = 0;
3227 nfsd4_load_reboot_recovery_data(void)
3232 nfsd4_init_recdir(user_recovery_dirname);
3233 status = nfsd4_recdir_load();
3234 nfs4_unlock_state();
3236 printk("NFSD: Failure reading reboot recovery data\n");
3240 get_nfs4_grace_period(void)
3242 return max(user_lease_time, lease_time) * HZ;
3246 * Since the lifetime of a delegation isn't limited to that of an open, a
3247 * client may quite reasonably hang on to a delegation as long as it has
3248 * the inode cached. This becomes an obvious problem the first time a
3249 * client's inode cache approaches the size of the server's total memory.
3251 * For now we avoid this problem by imposing a hard limit on the number
3252 * of delegations, which varies according to the server's memory size.
3255 set_max_delegations(void)
3258 * Allow at most 4 delegations per megabyte of RAM. Quick
3259 * estimates suggest that in the worst case (where every delegation
3260 * is for a different inode), a delegation could take about 1.5K,
3261 * giving a worst case usage of about 6% of memory.
3263 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
3266 /* initialization to perform when the nfsd service is started: */
3269 __nfs4_state_start(void)
3271 unsigned long grace_time;
3273 boot_time = get_seconds();
3274 grace_time = get_nfs4_grace_period();
3275 lease_time = user_lease_time;
3276 locks_start_grace(&nfsd4_manager);
3277 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
3279 laundry_wq = create_singlethread_workqueue("nfsd4");
3280 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
3281 set_max_delegations();
3285 nfs4_state_start(void)
3289 nfsd4_load_reboot_recovery_data();
3290 __nfs4_state_start();
3296 nfs4_lease_time(void)
3302 __nfs4_state_shutdown(void)
3305 struct nfs4_client *clp = NULL;
3306 struct nfs4_delegation *dp = NULL;
3307 struct list_head *pos, *next, reaplist;
3309 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3310 while (!list_empty(&conf_id_hashtbl[i])) {
3311 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
3314 while (!list_empty(&unconf_str_hashtbl[i])) {
3315 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
3319 INIT_LIST_HEAD(&reaplist);
3320 spin_lock(&recall_lock);
3321 list_for_each_safe(pos, next, &del_recall_lru) {
3322 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3323 list_move(&dp->dl_recall_lru, &reaplist);
3325 spin_unlock(&recall_lock);
3326 list_for_each_safe(pos, next, &reaplist) {
3327 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3328 list_del_init(&dp->dl_recall_lru);
3329 unhash_delegation(dp);
3332 nfsd4_shutdown_recdir();
3337 nfs4_state_shutdown(void)
3339 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
3340 destroy_workqueue(laundry_wq);
3341 locks_end_grace(&nfsd4_manager);
3343 nfs4_release_reclaim();
3344 __nfs4_state_shutdown();
3345 nfs4_unlock_state();
3349 * user_recovery_dirname is protected by the nfsd_mutex since it's only
3350 * accessed when nfsd is starting.
3353 nfs4_set_recdir(char *recdir)
3355 strcpy(user_recovery_dirname, recdir);
3359 * Change the NFSv4 recovery directory to recdir.
3362 nfs4_reset_recoverydir(char *recdir)
3367 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
3371 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
3372 nfs4_set_recdir(recdir);
3380 nfs4_recoverydir(void)
3382 return user_recovery_dirname;
3386 * Called when leasetime is changed.
3388 * The only way the protocol gives us to handle on-the-fly lease changes is to
3389 * simulate a reboot. Instead of doing that, we just wait till the next time
3390 * we start to register any changes in lease time. If the administrator
3391 * really wants to change the lease time *now*, they can go ahead and bring
3392 * nfsd down and then back up again after changing the lease time.
3394 * user_lease_time is protected by nfsd_mutex since it's only really accessed
3395 * when nfsd is starting
3398 nfs4_reset_lease(time_t leasetime)
3400 user_lease_time = leasetime;