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);
385 static DEFINE_SPINLOCK(sessionid_lock);
386 #define SESSION_HASH_SIZE 512
387 static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
390 hash_sessionid(struct nfs4_sessionid *sessionid)
392 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
394 return sid->sequence % SESSION_HASH_SIZE;
398 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
400 u32 *ptr = (u32 *)(&sessionid->data[0]);
401 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
404 /* caller must hold sessionid_lock */
405 static struct nfsd4_session *
406 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
408 struct nfsd4_session *elem;
411 dump_sessionid(__func__, sessionid);
412 idx = hash_sessionid(sessionid);
413 dprintk("%s: idx is %d\n", __func__, idx);
414 /* Search in the appropriate list */
415 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
416 dump_sessionid("list traversal", &elem->se_sessionid);
417 if (!memcmp(elem->se_sessionid.data, sessionid->data,
418 NFS4_MAX_SESSIONID_LEN)) {
423 dprintk("%s: session not found\n", __func__);
427 /* caller must hold sessionid_lock */
429 unhash_session(struct nfsd4_session *ses)
431 list_del(&ses->se_hash);
432 list_del(&ses->se_perclnt);
436 release_session(struct nfsd4_session *ses)
438 spin_lock(&sessionid_lock);
440 spin_unlock(&sessionid_lock);
441 nfsd4_put_session(ses);
445 free_session(struct kref *kref)
447 struct nfsd4_session *ses;
449 ses = container_of(kref, struct nfsd4_session, se_ref);
450 kfree(ses->se_slots);
455 renew_client(struct nfs4_client *clp)
458 * Move client to the end to the LRU list.
460 dprintk("renewing client (clientid %08x/%08x)\n",
461 clp->cl_clientid.cl_boot,
462 clp->cl_clientid.cl_id);
463 list_move_tail(&clp->cl_lru, &client_lru);
464 clp->cl_time = get_seconds();
467 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
469 STALE_CLIENTID(clientid_t *clid)
471 if (clid->cl_boot == boot_time)
473 dprintk("NFSD stale clientid (%08x/%08x)\n",
474 clid->cl_boot, clid->cl_id);
479 * XXX Should we use a slab cache ?
480 * This type of memory management is somewhat inefficient, but we use it
481 * anyway since SETCLIENTID is not a common operation.
483 static struct nfs4_client *alloc_client(struct xdr_netobj name)
485 struct nfs4_client *clp;
487 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
490 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
491 if (clp->cl_name.data == NULL) {
495 memcpy(clp->cl_name.data, name.data, name.len);
496 clp->cl_name.len = name.len;
501 shutdown_callback_client(struct nfs4_client *clp)
503 struct rpc_clnt *clnt = clp->cl_callback.cb_client;
507 * Callback threads take a reference on the client, so there
508 * should be no outstanding callbacks at this point.
510 clp->cl_callback.cb_client = NULL;
511 rpc_shutdown_client(clnt);
516 free_client(struct nfs4_client *clp)
518 shutdown_callback_client(clp);
519 if (clp->cl_cred.cr_group_info)
520 put_group_info(clp->cl_cred.cr_group_info);
521 kfree(clp->cl_principal);
522 kfree(clp->cl_name.data);
527 put_nfs4_client(struct nfs4_client *clp)
529 if (atomic_dec_and_test(&clp->cl_count))
534 expire_client(struct nfs4_client *clp)
536 struct nfs4_stateowner *sop;
537 struct nfs4_delegation *dp;
538 struct list_head reaplist;
540 dprintk("NFSD: expire_client cl_count %d\n",
541 atomic_read(&clp->cl_count));
543 INIT_LIST_HEAD(&reaplist);
544 spin_lock(&recall_lock);
545 while (!list_empty(&clp->cl_delegations)) {
546 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
547 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
549 list_del_init(&dp->dl_perclnt);
550 list_move(&dp->dl_recall_lru, &reaplist);
552 spin_unlock(&recall_lock);
553 while (!list_empty(&reaplist)) {
554 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
555 list_del_init(&dp->dl_recall_lru);
556 unhash_delegation(dp);
558 list_del(&clp->cl_idhash);
559 list_del(&clp->cl_strhash);
560 list_del(&clp->cl_lru);
561 while (!list_empty(&clp->cl_openowners)) {
562 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
563 release_openowner(sop);
565 while (!list_empty(&clp->cl_sessions)) {
566 struct nfsd4_session *ses;
567 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
569 release_session(ses);
571 put_nfs4_client(clp);
574 static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir)
576 struct nfs4_client *clp;
578 clp = alloc_client(name);
581 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
582 atomic_set(&clp->cl_count, 1);
583 atomic_set(&clp->cl_callback.cb_set, 0);
584 INIT_LIST_HEAD(&clp->cl_idhash);
585 INIT_LIST_HEAD(&clp->cl_strhash);
586 INIT_LIST_HEAD(&clp->cl_openowners);
587 INIT_LIST_HEAD(&clp->cl_delegations);
588 INIT_LIST_HEAD(&clp->cl_sessions);
589 INIT_LIST_HEAD(&clp->cl_lru);
593 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
595 memcpy(target->cl_verifier.data, source->data,
596 sizeof(target->cl_verifier.data));
599 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
601 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
602 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
605 static void copy_cred(struct svc_cred *target, struct svc_cred *source)
607 target->cr_uid = source->cr_uid;
608 target->cr_gid = source->cr_gid;
609 target->cr_group_info = source->cr_group_info;
610 get_group_info(target->cr_group_info);
613 static int same_name(const char *n1, const char *n2)
615 return 0 == memcmp(n1, n2, HEXDIR_LEN);
619 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
621 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
625 same_clid(clientid_t *cl1, clientid_t *cl2)
627 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
630 /* XXX what about NGROUP */
632 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
634 return cr1->cr_uid == cr2->cr_uid;
637 static void gen_clid(struct nfs4_client *clp)
639 static u32 current_clientid = 1;
641 clp->cl_clientid.cl_boot = boot_time;
642 clp->cl_clientid.cl_id = current_clientid++;
645 static void gen_confirm(struct nfs4_client *clp)
650 p = (u32 *)clp->cl_confirm.data;
651 *p++ = get_seconds();
655 static int check_name(struct xdr_netobj name)
659 if (name.len > NFS4_OPAQUE_LIMIT) {
660 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
667 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
669 unsigned int idhashval;
671 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
672 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
673 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
674 list_add_tail(&clp->cl_lru, &client_lru);
675 clp->cl_time = get_seconds();
679 move_to_confirmed(struct nfs4_client *clp)
681 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
682 unsigned int strhashval;
684 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
685 list_del_init(&clp->cl_strhash);
686 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
687 strhashval = clientstr_hashval(clp->cl_recdir);
688 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
692 static struct nfs4_client *
693 find_confirmed_client(clientid_t *clid)
695 struct nfs4_client *clp;
696 unsigned int idhashval = clientid_hashval(clid->cl_id);
698 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
699 if (same_clid(&clp->cl_clientid, clid))
705 static struct nfs4_client *
706 find_unconfirmed_client(clientid_t *clid)
708 struct nfs4_client *clp;
709 unsigned int idhashval = clientid_hashval(clid->cl_id);
711 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
712 if (same_clid(&clp->cl_clientid, clid))
718 static struct nfs4_client *
719 find_confirmed_client_by_str(const char *dname, unsigned int hashval)
721 struct nfs4_client *clp;
723 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
724 if (same_name(clp->cl_recdir, dname))
730 static struct nfs4_client *
731 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval)
733 struct nfs4_client *clp;
735 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
736 if (same_name(clp->cl_recdir, dname))
742 /* a helper function for parse_callback */
744 parse_octet(unsigned int *lenp, char **addrp)
746 unsigned int len = *lenp;
758 if ((c < '0') || (c > '9')) {
764 n = (n * 10) + (c - '0');
775 /* parse and set the setclientid ipv4 callback address */
777 parse_ipv4(unsigned int addr_len, char *addr_val, unsigned int *cbaddrp, unsigned short *cbportp)
782 u32 addrlen = addr_len;
783 char *addr = addr_val;
788 for(i = 4; i > 0 ; i--) {
789 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
792 cbaddr |= (temp << shift);
800 for(i = 2; i > 0 ; i--) {
801 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
804 cbport |= (temp << shift);
813 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se)
815 struct nfs4_callback *cb = &clp->cl_callback;
817 /* Currently, we only support tcp for the callback channel */
818 if ((se->se_callback_netid_len != 3) || memcmp((char *)se->se_callback_netid_val, "tcp", 3))
821 if ( !(parse_ipv4(se->se_callback_addr_len, se->se_callback_addr_val,
822 &cb->cb_addr, &cb->cb_port)))
824 cb->cb_prog = se->se_callback_prog;
825 cb->cb_ident = se->se_callback_ident;
828 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
829 "will not receive delegations\n",
830 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
836 * Set the exchange_id flags returned by the server.
839 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
841 /* pNFS is not supported */
842 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
844 /* Referrals are supported, Migration is not. */
845 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
847 /* set the wire flags to return to client. */
848 clid->flags = new->cl_exchange_flags;
852 nfsd4_exchange_id(struct svc_rqst *rqstp,
853 struct nfsd4_compound_state *cstate,
854 struct nfsd4_exchange_id *exid)
856 struct nfs4_client *unconf, *conf, *new;
858 unsigned int strhashval;
859 char dname[HEXDIR_LEN];
860 nfs4_verifier verf = exid->verifier;
861 u32 ip_addr = svc_addr_in(rqstp)->sin_addr.s_addr;
863 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
864 " ip_addr=%u flags %x, spa_how %d\n",
865 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
866 ip_addr, exid->flags, exid->spa_how);
868 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
871 /* Currently only support SP4_NONE */
872 switch (exid->spa_how) {
876 return nfserr_encr_alg_unsupp;
878 BUG(); /* checked by xdr code */
880 return nfserr_serverfault; /* no excuse :-/ */
883 status = nfs4_make_rec_clidname(dname, &exid->clname);
888 strhashval = clientstr_hashval(dname);
893 conf = find_confirmed_client_by_str(dname, strhashval);
895 if (!same_verf(&verf, &conf->cl_verifier)) {
897 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
898 status = nfserr_not_same;
901 /* Client reboot: destroy old state */
905 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
907 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
908 status = nfserr_perm;
914 if (ip_addr != conf->cl_addr &&
915 !(exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A)) {
916 /* Client collision. 18.35.4 case 3 */
917 status = nfserr_clid_inuse;
921 * Set bit when the owner id and verifier map to an already
922 * confirmed client id (18.35.3).
924 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
927 * Falling into 18.35.4 case 2, possible router replay.
928 * Leave confirmed record intact and return same result.
930 copy_verf(conf, &verf);
935 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
936 status = nfserr_noent;
941 unconf = find_unconfirmed_client_by_str(dname, strhashval);
944 * Possible retry or client restart. Per 18.35.4 case 4,
945 * a new unconfirmed record should be generated regardless
946 * of whether any properties have changed.
948 expire_client(unconf);
953 new = create_client(exid->clname, dname);
955 status = nfserr_resource;
959 copy_verf(new, &verf);
960 copy_cred(&new->cl_cred, &rqstp->rq_cred);
961 new->cl_addr = ip_addr;
964 add_to_unconfirmed(new, strhashval);
966 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
967 exid->clientid.cl_id = new->cl_clientid.cl_id;
969 new->cl_seqid = exid->seqid = 1;
970 nfsd4_set_ex_flags(new, exid);
972 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
973 new->cl_seqid, new->cl_exchange_flags);
979 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
984 nfsd4_create_session(struct svc_rqst *rqstp,
985 struct nfsd4_compound_state *cstate,
986 struct nfsd4_create_session *cr_ses)
988 return -1; /* stub */
992 nfsd4_destroy_session(struct svc_rqst *r,
993 struct nfsd4_compound_state *cstate,
994 struct nfsd4_destroy_session *sessionid)
996 return -1; /* stub */
1000 nfsd4_sequence(struct svc_rqst *r,
1001 struct nfsd4_compound_state *cstate,
1002 struct nfsd4_sequence *seq)
1004 return -1; /* stub */
1008 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1009 struct nfsd4_setclientid *setclid)
1011 struct sockaddr_in *sin = svc_addr_in(rqstp);
1012 struct xdr_netobj clname = {
1013 .len = setclid->se_namelen,
1014 .data = setclid->se_name,
1016 nfs4_verifier clverifier = setclid->se_verf;
1017 unsigned int strhashval;
1018 struct nfs4_client *conf, *unconf, *new;
1021 char dname[HEXDIR_LEN];
1023 if (!check_name(clname))
1024 return nfserr_inval;
1026 status = nfs4_make_rec_clidname(dname, &clname);
1031 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1032 * We get here on a DRC miss.
1035 strhashval = clientstr_hashval(dname);
1038 conf = find_confirmed_client_by_str(dname, strhashval);
1040 /* RFC 3530 14.2.33 CASE 0: */
1041 status = nfserr_clid_inuse;
1042 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1043 dprintk("NFSD: setclientid: string in use by client"
1044 " at %pI4\n", &conf->cl_addr);
1049 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1050 * has a description of SETCLIENTID request processing consisting
1051 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1053 unconf = find_unconfirmed_client_by_str(dname, strhashval);
1054 status = nfserr_resource;
1057 * RFC 3530 14.2.33 CASE 4:
1058 * placed first, because it is the normal case
1061 expire_client(unconf);
1062 new = create_client(clname, dname);
1066 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1068 * RFC 3530 14.2.33 CASE 1:
1069 * probable callback update
1072 /* Note this is removing unconfirmed {*x***},
1073 * which is stronger than RFC recommended {vxc**}.
1074 * This has the advantage that there is at most
1075 * one {*x***} in either list at any time.
1077 expire_client(unconf);
1079 new = create_client(clname, dname);
1082 copy_clid(new, conf);
1083 } else if (!unconf) {
1085 * RFC 3530 14.2.33 CASE 2:
1086 * probable client reboot; state will be removed if
1089 new = create_client(clname, dname);
1095 * RFC 3530 14.2.33 CASE 3:
1096 * probable client reboot; state will be removed if
1099 expire_client(unconf);
1100 new = create_client(clname, dname);
1105 copy_verf(new, &clverifier);
1106 new->cl_addr = sin->sin_addr.s_addr;
1107 new->cl_flavor = rqstp->rq_flavor;
1108 princ = svc_gss_principal(rqstp);
1110 new->cl_principal = kstrdup(princ, GFP_KERNEL);
1111 if (new->cl_principal == NULL) {
1116 copy_cred(&new->cl_cred, &rqstp->rq_cred);
1118 gen_callback(new, setclid);
1119 add_to_unconfirmed(new, strhashval);
1120 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1121 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1122 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1125 nfs4_unlock_state();
1131 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1132 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1133 * bullets, labeled as CASE1 - CASE4 below.
1136 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1137 struct nfsd4_compound_state *cstate,
1138 struct nfsd4_setclientid_confirm *setclientid_confirm)
1140 struct sockaddr_in *sin = svc_addr_in(rqstp);
1141 struct nfs4_client *conf, *unconf;
1142 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1143 clientid_t * clid = &setclientid_confirm->sc_clientid;
1146 if (STALE_CLIENTID(clid))
1147 return nfserr_stale_clientid;
1149 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1150 * We get here on a DRC miss.
1155 conf = find_confirmed_client(clid);
1156 unconf = find_unconfirmed_client(clid);
1158 status = nfserr_clid_inuse;
1159 if (conf && conf->cl_addr != sin->sin_addr.s_addr)
1161 if (unconf && unconf->cl_addr != sin->sin_addr.s_addr)
1165 * section 14.2.34 of RFC 3530 has a description of
1166 * SETCLIENTID_CONFIRM request processing consisting
1167 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1169 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
1171 * RFC 3530 14.2.34 CASE 1:
1174 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1175 status = nfserr_clid_inuse;
1177 /* XXX: We just turn off callbacks until we can handle
1178 * change request correctly. */
1179 atomic_set(&conf->cl_callback.cb_set, 0);
1181 nfsd4_remove_clid_dir(unconf);
1182 expire_client(unconf);
1186 } else if (conf && !unconf) {
1188 * RFC 3530 14.2.34 CASE 2:
1189 * probable retransmitted request; play it safe and
1192 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1193 status = nfserr_clid_inuse;
1196 } else if (!conf && unconf
1197 && same_verf(&unconf->cl_confirm, &confirm)) {
1199 * RFC 3530 14.2.34 CASE 3:
1200 * Normal case; new or rebooted client:
1202 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1203 status = nfserr_clid_inuse;
1206 clientstr_hashval(unconf->cl_recdir);
1207 conf = find_confirmed_client_by_str(unconf->cl_recdir,
1210 nfsd4_remove_clid_dir(conf);
1211 expire_client(conf);
1213 move_to_confirmed(unconf);
1215 nfsd4_probe_callback(conf);
1218 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1219 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
1222 * RFC 3530 14.2.34 CASE 4:
1223 * Client probably hasn't noticed that we rebooted yet.
1225 status = nfserr_stale_clientid;
1227 /* check that we have hit one of the cases...*/
1228 status = nfserr_clid_inuse;
1231 nfs4_unlock_state();
1235 /* OPEN Share state helper functions */
1236 static inline struct nfs4_file *
1237 alloc_init_file(struct inode *ino)
1239 struct nfs4_file *fp;
1240 unsigned int hashval = file_hashval(ino);
1242 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1244 atomic_set(&fp->fi_ref, 1);
1245 INIT_LIST_HEAD(&fp->fi_hash);
1246 INIT_LIST_HEAD(&fp->fi_stateids);
1247 INIT_LIST_HEAD(&fp->fi_delegations);
1248 spin_lock(&recall_lock);
1249 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
1250 spin_unlock(&recall_lock);
1251 fp->fi_inode = igrab(ino);
1252 fp->fi_id = current_fileid++;
1253 fp->fi_had_conflict = false;
1260 nfsd4_free_slab(struct kmem_cache **slab)
1264 kmem_cache_destroy(*slab);
1269 nfsd4_free_slabs(void)
1271 nfsd4_free_slab(&stateowner_slab);
1272 nfsd4_free_slab(&file_slab);
1273 nfsd4_free_slab(&stateid_slab);
1274 nfsd4_free_slab(&deleg_slab);
1278 nfsd4_init_slabs(void)
1280 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
1281 sizeof(struct nfs4_stateowner), 0, 0, NULL);
1282 if (stateowner_slab == NULL)
1284 file_slab = kmem_cache_create("nfsd4_files",
1285 sizeof(struct nfs4_file), 0, 0, NULL);
1286 if (file_slab == NULL)
1288 stateid_slab = kmem_cache_create("nfsd4_stateids",
1289 sizeof(struct nfs4_stateid), 0, 0, NULL);
1290 if (stateid_slab == NULL)
1292 deleg_slab = kmem_cache_create("nfsd4_delegations",
1293 sizeof(struct nfs4_delegation), 0, 0, NULL);
1294 if (deleg_slab == NULL)
1299 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1304 nfs4_free_stateowner(struct kref *kref)
1306 struct nfs4_stateowner *sop =
1307 container_of(kref, struct nfs4_stateowner, so_ref);
1308 kfree(sop->so_owner.data);
1309 kmem_cache_free(stateowner_slab, sop);
1312 static inline struct nfs4_stateowner *
1313 alloc_stateowner(struct xdr_netobj *owner)
1315 struct nfs4_stateowner *sop;
1317 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1318 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1319 memcpy(sop->so_owner.data, owner->data, owner->len);
1320 sop->so_owner.len = owner->len;
1321 kref_init(&sop->so_ref);
1324 kmem_cache_free(stateowner_slab, sop);
1329 static struct nfs4_stateowner *
1330 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1331 struct nfs4_stateowner *sop;
1332 struct nfs4_replay *rp;
1333 unsigned int idhashval;
1335 if (!(sop = alloc_stateowner(&open->op_owner)))
1337 idhashval = ownerid_hashval(current_ownerid);
1338 INIT_LIST_HEAD(&sop->so_idhash);
1339 INIT_LIST_HEAD(&sop->so_strhash);
1340 INIT_LIST_HEAD(&sop->so_perclient);
1341 INIT_LIST_HEAD(&sop->so_stateids);
1342 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1343 INIT_LIST_HEAD(&sop->so_close_lru);
1345 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1346 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
1347 list_add(&sop->so_perclient, &clp->cl_openowners);
1348 sop->so_is_open_owner = 1;
1349 sop->so_id = current_ownerid++;
1350 sop->so_client = clp;
1351 sop->so_seqid = open->op_seqid;
1352 sop->so_confirmed = 0;
1353 rp = &sop->so_replay;
1354 rp->rp_status = nfserr_serverfault;
1356 rp->rp_buf = rp->rp_ibuf;
1361 init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1362 struct nfs4_stateowner *sop = open->op_stateowner;
1363 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1365 INIT_LIST_HEAD(&stp->st_hash);
1366 INIT_LIST_HEAD(&stp->st_perstateowner);
1367 INIT_LIST_HEAD(&stp->st_lockowners);
1368 INIT_LIST_HEAD(&stp->st_perfile);
1369 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
1370 list_add(&stp->st_perstateowner, &sop->so_stateids);
1371 list_add(&stp->st_perfile, &fp->fi_stateids);
1372 stp->st_stateowner = sop;
1375 stp->st_stateid.si_boot = boot_time;
1376 stp->st_stateid.si_stateownerid = sop->so_id;
1377 stp->st_stateid.si_fileid = fp->fi_id;
1378 stp->st_stateid.si_generation = 0;
1379 stp->st_access_bmap = 0;
1380 stp->st_deny_bmap = 0;
1381 __set_bit(open->op_share_access, &stp->st_access_bmap);
1382 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1383 stp->st_openstp = NULL;
1387 move_to_close_lru(struct nfs4_stateowner *sop)
1389 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1391 list_move_tail(&sop->so_close_lru, &close_lru);
1392 sop->so_time = get_seconds();
1396 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1399 return (sop->so_owner.len == owner->len) &&
1400 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1401 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1404 static struct nfs4_stateowner *
1405 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1407 struct nfs4_stateowner *so = NULL;
1409 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
1410 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1416 /* search file_hashtbl[] for file */
1417 static struct nfs4_file *
1418 find_file(struct inode *ino)
1420 unsigned int hashval = file_hashval(ino);
1421 struct nfs4_file *fp;
1423 spin_lock(&recall_lock);
1424 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
1425 if (fp->fi_inode == ino) {
1427 spin_unlock(&recall_lock);
1431 spin_unlock(&recall_lock);
1435 static inline int access_valid(u32 x)
1437 if (x < NFS4_SHARE_ACCESS_READ)
1439 if (x > NFS4_SHARE_ACCESS_BOTH)
1444 static inline int deny_valid(u32 x)
1446 /* Note: unlike access bits, deny bits may be zero. */
1447 return x <= NFS4_SHARE_DENY_BOTH;
1451 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1452 * st_{access,deny}_bmap field of the stateid, in order to track not
1453 * only what share bits are currently in force, but also what
1454 * combinations of share bits previous opens have used. This allows us
1455 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1456 * return an error if the client attempt to downgrade to a combination
1457 * of share bits not explicable by closing some of its previous opens.
1459 * XXX: This enforcement is actually incomplete, since we don't keep
1460 * track of access/deny bit combinations; so, e.g., we allow:
1462 * OPEN allow read, deny write
1463 * OPEN allow both, deny none
1464 * DOWNGRADE allow read, deny none
1466 * which we should reject.
1469 set_access(unsigned int *access, unsigned long bmap) {
1473 for (i = 1; i < 4; i++) {
1474 if (test_bit(i, &bmap))
1480 set_deny(unsigned int *deny, unsigned long bmap) {
1484 for (i = 0; i < 4; i++) {
1485 if (test_bit(i, &bmap))
1491 test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1492 unsigned int access, deny;
1494 set_access(&access, stp->st_access_bmap);
1495 set_deny(&deny, stp->st_deny_bmap);
1496 if ((access & open->op_share_deny) || (deny & open->op_share_access))
1502 * Called to check deny when READ with all zero stateid or
1503 * WRITE with all zero or all one stateid
1506 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
1508 struct inode *ino = current_fh->fh_dentry->d_inode;
1509 struct nfs4_file *fp;
1510 struct nfs4_stateid *stp;
1513 dprintk("NFSD: nfs4_share_conflict\n");
1515 fp = find_file(ino);
1518 ret = nfserr_locked;
1519 /* Search for conflicting share reservations */
1520 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
1521 if (test_bit(deny_type, &stp->st_deny_bmap) ||
1522 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
1532 nfs4_file_downgrade(struct file *filp, unsigned int share_access)
1534 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
1535 drop_file_write_access(filp);
1536 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
1541 * Recall a delegation
1544 do_recall(void *__dp)
1546 struct nfs4_delegation *dp = __dp;
1548 dp->dl_file->fi_had_conflict = true;
1549 nfsd4_cb_recall(dp);
1554 * Spawn a thread to perform a recall on the delegation represented
1555 * by the lease (file_lock)
1557 * Called from break_lease() with lock_kernel() held.
1558 * Note: we assume break_lease will only call this *once* for any given
1562 void nfsd_break_deleg_cb(struct file_lock *fl)
1564 struct nfs4_delegation *dp= (struct nfs4_delegation *)fl->fl_owner;
1565 struct task_struct *t;
1567 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
1571 /* We're assuming the state code never drops its reference
1572 * without first removing the lease. Since we're in this lease
1573 * callback (and since the lease code is serialized by the kernel
1574 * lock) we know the server hasn't removed the lease yet, we know
1575 * it's safe to take a reference: */
1576 atomic_inc(&dp->dl_count);
1577 atomic_inc(&dp->dl_client->cl_count);
1579 spin_lock(&recall_lock);
1580 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
1581 spin_unlock(&recall_lock);
1583 /* only place dl_time is set. protected by lock_kernel*/
1584 dp->dl_time = get_seconds();
1587 * We don't want the locks code to timeout the lease for us;
1588 * we'll remove it ourself if the delegation isn't returned
1591 fl->fl_break_time = 0;
1593 t = kthread_run(do_recall, dp, "%s", "nfs4_cb_recall");
1595 struct nfs4_client *clp = dp->dl_client;
1597 printk(KERN_INFO "NFSD: Callback thread failed for "
1598 "for client (clientid %08x/%08x)\n",
1599 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1600 put_nfs4_client(dp->dl_client);
1601 nfs4_put_delegation(dp);
1606 * The file_lock is being reapd.
1608 * Called by locks_free_lock() with lock_kernel() held.
1611 void nfsd_release_deleg_cb(struct file_lock *fl)
1613 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1615 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
1617 if (!(fl->fl_flags & FL_LEASE) || !dp)
1619 dp->dl_flock = NULL;
1623 * Set the delegation file_lock back pointer.
1625 * Called from setlease() with lock_kernel() held.
1628 void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
1630 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
1632 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
1639 * Called from setlease() with lock_kernel() held
1642 int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
1644 struct nfs4_delegation *onlistd =
1645 (struct nfs4_delegation *)onlist->fl_owner;
1646 struct nfs4_delegation *tryd =
1647 (struct nfs4_delegation *)try->fl_owner;
1649 if (onlist->fl_lmops != try->fl_lmops)
1652 return onlistd->dl_client == tryd->dl_client;
1657 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
1660 return lease_modify(onlist, arg);
1665 static struct lock_manager_operations nfsd_lease_mng_ops = {
1666 .fl_break = nfsd_break_deleg_cb,
1667 .fl_release_private = nfsd_release_deleg_cb,
1668 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
1669 .fl_mylease = nfsd_same_client_deleg_cb,
1670 .fl_change = nfsd_change_deleg_cb,
1675 nfsd4_process_open1(struct nfsd4_open *open)
1677 clientid_t *clientid = &open->op_clientid;
1678 struct nfs4_client *clp = NULL;
1679 unsigned int strhashval;
1680 struct nfs4_stateowner *sop = NULL;
1682 if (!check_name(open->op_owner))
1683 return nfserr_inval;
1685 if (STALE_CLIENTID(&open->op_clientid))
1686 return nfserr_stale_clientid;
1688 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
1689 sop = find_openstateowner_str(strhashval, open);
1690 open->op_stateowner = sop;
1692 /* Make sure the client's lease hasn't expired. */
1693 clp = find_confirmed_client(clientid);
1695 return nfserr_expired;
1698 if (!sop->so_confirmed) {
1699 /* Replace unconfirmed owners without checking for replay. */
1700 clp = sop->so_client;
1701 release_openowner(sop);
1702 open->op_stateowner = NULL;
1705 if (open->op_seqid == sop->so_seqid - 1) {
1706 if (sop->so_replay.rp_buflen)
1707 return nfserr_replay_me;
1708 /* The original OPEN failed so spectacularly
1709 * that we don't even have replay data saved!
1710 * Therefore, we have no choice but to continue
1711 * processing this OPEN; presumably, we'll
1712 * fail again for the same reason.
1714 dprintk("nfsd4_process_open1: replay with no replay cache\n");
1717 if (open->op_seqid != sop->so_seqid)
1718 return nfserr_bad_seqid;
1720 if (open->op_stateowner == NULL) {
1721 sop = alloc_init_open_stateowner(strhashval, clp, open);
1723 return nfserr_resource;
1724 open->op_stateowner = sop;
1726 list_del_init(&sop->so_close_lru);
1727 renew_client(sop->so_client);
1731 static inline __be32
1732 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
1734 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
1735 return nfserr_openmode;
1740 static struct nfs4_delegation *
1741 find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
1743 struct nfs4_delegation *dp;
1745 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
1746 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
1753 nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
1754 struct nfs4_delegation **dp)
1757 __be32 status = nfserr_bad_stateid;
1759 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
1762 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
1763 RD_STATE : WR_STATE;
1764 status = nfs4_check_delegmode(*dp, flags);
1768 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
1772 open->op_stateowner->so_confirmed = 1;
1777 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
1779 struct nfs4_stateid *local;
1780 __be32 status = nfserr_share_denied;
1781 struct nfs4_stateowner *sop = open->op_stateowner;
1783 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1784 /* ignore lock owners */
1785 if (local->st_stateowner->so_is_open_owner == 0)
1787 /* remember if we have seen this open owner */
1788 if (local->st_stateowner == sop)
1790 /* check for conflicting share reservations */
1791 if (!test_share(local, open))
1799 static inline struct nfs4_stateid *
1800 nfs4_alloc_stateid(void)
1802 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
1806 nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
1807 struct nfs4_delegation *dp,
1808 struct svc_fh *cur_fh, int flags)
1810 struct nfs4_stateid *stp;
1812 stp = nfs4_alloc_stateid();
1814 return nfserr_resource;
1817 get_file(dp->dl_vfs_file);
1818 stp->st_vfs_file = dp->dl_vfs_file;
1821 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
1824 if (status == nfserr_dropit)
1825 status = nfserr_jukebox;
1826 kmem_cache_free(stateid_slab, stp);
1834 static inline __be32
1835 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
1836 struct nfsd4_open *open)
1838 struct iattr iattr = {
1839 .ia_valid = ATTR_SIZE,
1842 if (!open->op_truncate)
1844 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
1845 return nfserr_inval;
1846 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
1850 nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
1852 struct file *filp = stp->st_vfs_file;
1853 struct inode *inode = filp->f_path.dentry->d_inode;
1854 unsigned int share_access, new_writer;
1857 set_access(&share_access, stp->st_access_bmap);
1858 new_writer = (~share_access) & open->op_share_access
1859 & NFS4_SHARE_ACCESS_WRITE;
1862 int err = get_write_access(inode);
1864 return nfserrno(err);
1865 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
1867 return nfserrno(err);
1868 file_take_write(filp);
1870 status = nfsd4_truncate(rqstp, cur_fh, open);
1873 put_write_access(inode);
1876 /* remember the open */
1877 filp->f_mode |= open->op_share_access;
1878 __set_bit(open->op_share_access, &stp->st_access_bmap);
1879 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1886 nfs4_set_claim_prev(struct nfsd4_open *open)
1888 open->op_stateowner->so_confirmed = 1;
1889 open->op_stateowner->so_client->cl_firststate = 1;
1893 * Attempt to hand out a delegation.
1896 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
1898 struct nfs4_delegation *dp;
1899 struct nfs4_stateowner *sop = stp->st_stateowner;
1900 struct nfs4_callback *cb = &sop->so_client->cl_callback;
1901 struct file_lock fl, *flp = &fl;
1902 int status, flag = 0;
1904 flag = NFS4_OPEN_DELEGATE_NONE;
1905 open->op_recall = 0;
1906 switch (open->op_claim_type) {
1907 case NFS4_OPEN_CLAIM_PREVIOUS:
1908 if (!atomic_read(&cb->cb_set))
1909 open->op_recall = 1;
1910 flag = open->op_delegate_type;
1911 if (flag == NFS4_OPEN_DELEGATE_NONE)
1914 case NFS4_OPEN_CLAIM_NULL:
1915 /* Let's not give out any delegations till everyone's
1916 * had the chance to reclaim theirs.... */
1917 if (locks_in_grace())
1919 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
1921 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
1922 flag = NFS4_OPEN_DELEGATE_WRITE;
1924 flag = NFS4_OPEN_DELEGATE_READ;
1930 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
1932 flag = NFS4_OPEN_DELEGATE_NONE;
1935 locks_init_lock(&fl);
1936 fl.fl_lmops = &nfsd_lease_mng_ops;
1937 fl.fl_flags = FL_LEASE;
1938 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1939 fl.fl_end = OFFSET_MAX;
1940 fl.fl_owner = (fl_owner_t)dp;
1941 fl.fl_file = stp->st_vfs_file;
1942 fl.fl_pid = current->tgid;
1944 /* vfs_setlease checks to see if delegation should be handed out.
1945 * the lock_manager callbacks fl_mylease and fl_change are used
1947 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1948 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
1949 unhash_delegation(dp);
1950 flag = NFS4_OPEN_DELEGATE_NONE;
1954 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
1956 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
1957 dp->dl_stateid.si_boot,
1958 dp->dl_stateid.si_stateownerid,
1959 dp->dl_stateid.si_fileid,
1960 dp->dl_stateid.si_generation);
1962 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
1963 && flag == NFS4_OPEN_DELEGATE_NONE
1964 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
1965 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1966 open->op_delegate_type = flag;
1970 * called with nfs4_lock_state() held.
1973 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
1975 struct nfs4_file *fp = NULL;
1976 struct inode *ino = current_fh->fh_dentry->d_inode;
1977 struct nfs4_stateid *stp = NULL;
1978 struct nfs4_delegation *dp = NULL;
1981 status = nfserr_inval;
1982 if (!access_valid(open->op_share_access)
1983 || !deny_valid(open->op_share_deny))
1986 * Lookup file; if found, lookup stateid and check open request,
1987 * and check for delegations in the process of being recalled.
1988 * If not found, create the nfs4_file struct
1990 fp = find_file(ino);
1992 if ((status = nfs4_check_open(fp, open, &stp)))
1994 status = nfs4_check_deleg(fp, open, &dp);
1998 status = nfserr_bad_stateid;
1999 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2001 status = nfserr_resource;
2002 fp = alloc_init_file(ino);
2008 * OPEN the file, or upgrade an existing OPEN.
2009 * If truncate fails, the OPEN fails.
2012 /* Stateid was found, this is an OPEN upgrade */
2013 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2016 update_stateid(&stp->st_stateid);
2018 /* Stateid was not found, this is a new OPEN */
2020 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
2021 flags |= NFSD_MAY_READ;
2022 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2023 flags |= NFSD_MAY_WRITE;
2024 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2027 init_stateid(stp, fp, open);
2028 status = nfsd4_truncate(rqstp, current_fh, open);
2030 release_open_stateid(stp);
2034 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2037 * Attempt to hand out a delegation. No error return, because the
2038 * OPEN succeeds even if we fail.
2040 nfs4_open_delegation(current_fh, open, stp);
2044 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
2045 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
2046 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
2050 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2051 nfs4_set_claim_prev(open);
2053 * To finish the open response, we just need to set the rflags.
2055 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
2056 if (!open->op_stateowner->so_confirmed)
2057 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2063 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2066 struct nfs4_client *clp;
2070 dprintk("process_renew(%08x/%08x): starting\n",
2071 clid->cl_boot, clid->cl_id);
2072 status = nfserr_stale_clientid;
2073 if (STALE_CLIENTID(clid))
2075 clp = find_confirmed_client(clid);
2076 status = nfserr_expired;
2078 /* We assume the client took too long to RENEW. */
2079 dprintk("nfsd4_renew: clientid not found!\n");
2083 status = nfserr_cb_path_down;
2084 if (!list_empty(&clp->cl_delegations)
2085 && !atomic_read(&clp->cl_callback.cb_set))
2089 nfs4_unlock_state();
2093 struct lock_manager nfsd4_manager = {
2097 nfsd4_end_grace(void)
2099 dprintk("NFSD: end of grace period\n");
2100 nfsd4_recdir_purge_old();
2101 locks_end_grace(&nfsd4_manager);
2105 nfs4_laundromat(void)
2107 struct nfs4_client *clp;
2108 struct nfs4_stateowner *sop;
2109 struct nfs4_delegation *dp;
2110 struct list_head *pos, *next, reaplist;
2111 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
2112 time_t t, clientid_val = NFSD_LEASE_TIME;
2113 time_t u, test_val = NFSD_LEASE_TIME;
2117 dprintk("NFSD: laundromat service - starting\n");
2118 if (locks_in_grace())
2120 list_for_each_safe(pos, next, &client_lru) {
2121 clp = list_entry(pos, struct nfs4_client, cl_lru);
2122 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2123 t = clp->cl_time - cutoff;
2124 if (clientid_val > t)
2128 dprintk("NFSD: purging unused client (clientid %08x)\n",
2129 clp->cl_clientid.cl_id);
2130 nfsd4_remove_clid_dir(clp);
2133 INIT_LIST_HEAD(&reaplist);
2134 spin_lock(&recall_lock);
2135 list_for_each_safe(pos, next, &del_recall_lru) {
2136 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2137 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2138 u = dp->dl_time - cutoff;
2143 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2145 list_move(&dp->dl_recall_lru, &reaplist);
2147 spin_unlock(&recall_lock);
2148 list_for_each_safe(pos, next, &reaplist) {
2149 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2150 list_del_init(&dp->dl_recall_lru);
2151 unhash_delegation(dp);
2153 test_val = NFSD_LEASE_TIME;
2154 list_for_each_safe(pos, next, &close_lru) {
2155 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2156 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2157 u = sop->so_time - cutoff;
2162 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2164 release_openowner(sop);
2166 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2167 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2168 nfs4_unlock_state();
2169 return clientid_val;
2172 static struct workqueue_struct *laundry_wq;
2173 static void laundromat_main(struct work_struct *);
2174 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2177 laundromat_main(struct work_struct *not_used)
2181 t = nfs4_laundromat();
2182 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
2183 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
2186 static struct nfs4_stateowner *
2187 search_close_lru(u32 st_id, int flags)
2189 struct nfs4_stateowner *local = NULL;
2191 if (flags & CLOSE_STATE) {
2192 list_for_each_entry(local, &close_lru, so_close_lru) {
2193 if (local->so_id == st_id)
2201 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2203 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
2207 STALE_STATEID(stateid_t *stateid)
2209 if (stateid->si_boot == boot_time)
2211 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
2212 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2213 stateid->si_generation);
2218 access_permit_read(unsigned long access_bmap)
2220 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2221 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2222 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2226 access_permit_write(unsigned long access_bmap)
2228 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2229 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2233 __be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
2235 __be32 status = nfserr_openmode;
2237 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2239 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2246 static inline __be32
2247 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2249 if (ONE_STATEID(stateid) && (flags & RD_STATE))
2251 else if (locks_in_grace()) {
2252 /* Answer in remaining cases depends on existance of
2253 * conflicting state; so we must wait out the grace period. */
2254 return nfserr_grace;
2255 } else if (flags & WR_STATE)
2256 return nfs4_share_conflict(current_fh,
2257 NFS4_SHARE_DENY_WRITE);
2258 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2259 return nfs4_share_conflict(current_fh,
2260 NFS4_SHARE_DENY_READ);
2264 * Allow READ/WRITE during grace period on recovered state only for files
2265 * that are not able to provide mandatory locking.
2268 grace_disallows_io(struct inode *inode)
2270 return locks_in_grace() && mandatory_lock(inode);
2273 static int check_stateid_generation(stateid_t *in, stateid_t *ref)
2275 /* If the client sends us a stateid from the future, it's buggy: */
2276 if (in->si_generation > ref->si_generation)
2277 return nfserr_bad_stateid;
2279 * The following, however, can happen. For example, if the
2280 * client sends an open and some IO at the same time, the open
2281 * may bump si_generation while the IO is still in flight.
2282 * Thanks to hard links and renames, the client never knows what
2283 * file an open will affect. So it could avoid that situation
2284 * only by serializing all opens and IO from the same open
2285 * owner. To recover from the old_stateid error, the client
2286 * will just have to retry the IO:
2288 if (in->si_generation < ref->si_generation)
2289 return nfserr_old_stateid;
2293 static int is_delegation_stateid(stateid_t *stateid)
2295 return stateid->si_fileid == 0;
2299 * Checks for stateid operations
2302 nfs4_preprocess_stateid_op(struct svc_fh *current_fh, stateid_t *stateid, int flags, struct file **filpp)
2304 struct nfs4_stateid *stp = NULL;
2305 struct nfs4_delegation *dp = NULL;
2306 struct inode *ino = current_fh->fh_dentry->d_inode;
2312 if (grace_disallows_io(ino))
2313 return nfserr_grace;
2315 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2316 return check_special_stateids(current_fh, stateid, flags);
2318 status = nfserr_stale_stateid;
2319 if (STALE_STATEID(stateid))
2322 status = nfserr_bad_stateid;
2323 if (is_delegation_stateid(stateid)) {
2324 dp = find_delegation_stateid(ino, stateid);
2327 status = check_stateid_generation(stateid, &dp->dl_stateid);
2330 status = nfs4_check_delegmode(dp, flags);
2333 renew_client(dp->dl_client);
2335 *filpp = dp->dl_vfs_file;
2336 } else { /* open or lock stateid */
2337 stp = find_stateid(stateid, flags);
2340 if (nfs4_check_fh(current_fh, stp))
2342 if (!stp->st_stateowner->so_confirmed)
2344 status = check_stateid_generation(stateid, &stp->st_stateid);
2347 status = nfs4_check_openmode(stp, flags);
2350 renew_client(stp->st_stateowner->so_client);
2352 *filpp = stp->st_vfs_file;
2362 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2363 RD_STATE : WR_STATE;
2367 * Checks for sequence id mutating operations.
2370 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)
2372 struct nfs4_stateid *stp;
2373 struct nfs4_stateowner *sop;
2376 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2377 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2378 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2379 stateid->si_generation);
2384 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2385 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
2386 return nfserr_bad_stateid;
2389 if (STALE_STATEID(stateid))
2390 return nfserr_stale_stateid;
2392 * We return BAD_STATEID if filehandle doesn't match stateid,
2393 * the confirmed flag is incorrecly set, or the generation
2394 * number is incorrect.
2396 stp = find_stateid(stateid, flags);
2399 * Also, we should make sure this isn't just the result of
2402 sop = search_close_lru(stateid->si_stateownerid, flags);
2404 return nfserr_bad_stateid;
2410 *sopp = sop = stp->st_stateowner;
2413 clientid_t *lockclid = &lock->v.new.clientid;
2414 struct nfs4_client *clp = sop->so_client;
2418 lkflg = setlkflg(lock->lk_type);
2420 if (lock->lk_is_new) {
2421 if (!sop->so_is_open_owner)
2422 return nfserr_bad_stateid;
2423 if (!same_clid(&clp->cl_clientid, lockclid))
2424 return nfserr_bad_stateid;
2425 /* stp is the open stateid */
2426 status = nfs4_check_openmode(stp, lkflg);
2430 /* stp is the lock stateid */
2431 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2437 if (nfs4_check_fh(current_fh, stp)) {
2438 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
2439 return nfserr_bad_stateid;
2443 * We now validate the seqid and stateid generation numbers.
2444 * For the moment, we ignore the possibility of
2445 * generation number wraparound.
2447 if (seqid != sop->so_seqid)
2450 if (sop->so_confirmed && flags & CONFIRM) {
2451 dprintk("NFSD: preprocess_seqid_op: expected"
2452 " unconfirmed stateowner!\n");
2453 return nfserr_bad_stateid;
2455 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2456 dprintk("NFSD: preprocess_seqid_op: stateowner not"
2457 " confirmed yet!\n");
2458 return nfserr_bad_stateid;
2460 status = check_stateid_generation(stateid, &stp->st_stateid);
2463 renew_client(sop->so_client);
2467 if (seqid == sop->so_seqid - 1) {
2468 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
2469 /* indicate replay to calling function */
2470 return nfserr_replay_me;
2472 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
2473 sop->so_seqid, seqid);
2475 return nfserr_bad_seqid;
2479 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2480 struct nfsd4_open_confirm *oc)
2483 struct nfs4_stateowner *sop;
2484 struct nfs4_stateid *stp;
2486 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
2487 (int)cstate->current_fh.fh_dentry->d_name.len,
2488 cstate->current_fh.fh_dentry->d_name.name);
2490 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
2496 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2497 oc->oc_seqid, &oc->oc_req_stateid,
2498 CONFIRM | OPEN_STATE,
2499 &oc->oc_stateowner, &stp, NULL)))
2502 sop = oc->oc_stateowner;
2503 sop->so_confirmed = 1;
2504 update_stateid(&stp->st_stateid);
2505 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
2506 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
2507 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
2508 stp->st_stateid.si_boot,
2509 stp->st_stateid.si_stateownerid,
2510 stp->st_stateid.si_fileid,
2511 stp->st_stateid.si_generation);
2513 nfsd4_create_clid_dir(sop->so_client);
2515 if (oc->oc_stateowner) {
2516 nfs4_get_stateowner(oc->oc_stateowner);
2517 cstate->replay_owner = oc->oc_stateowner;
2519 nfs4_unlock_state();
2525 * unset all bits in union bitmap (bmap) that
2526 * do not exist in share (from successful OPEN_DOWNGRADE)
2529 reset_union_bmap_access(unsigned long access, unsigned long *bmap)
2532 for (i = 1; i < 4; i++) {
2533 if ((i & access) != i)
2534 __clear_bit(i, bmap);
2539 reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
2542 for (i = 0; i < 4; i++) {
2543 if ((i & deny) != i)
2544 __clear_bit(i, bmap);
2549 nfsd4_open_downgrade(struct svc_rqst *rqstp,
2550 struct nfsd4_compound_state *cstate,
2551 struct nfsd4_open_downgrade *od)
2554 struct nfs4_stateid *stp;
2555 unsigned int share_access;
2557 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
2558 (int)cstate->current_fh.fh_dentry->d_name.len,
2559 cstate->current_fh.fh_dentry->d_name.name);
2561 if (!access_valid(od->od_share_access)
2562 || !deny_valid(od->od_share_deny))
2563 return nfserr_inval;
2566 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2570 &od->od_stateowner, &stp, NULL)))
2573 status = nfserr_inval;
2574 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
2575 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
2576 stp->st_access_bmap, od->od_share_access);
2579 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
2580 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
2581 stp->st_deny_bmap, od->od_share_deny);
2584 set_access(&share_access, stp->st_access_bmap);
2585 nfs4_file_downgrade(stp->st_vfs_file,
2586 share_access & ~od->od_share_access);
2588 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
2589 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
2591 update_stateid(&stp->st_stateid);
2592 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
2595 if (od->od_stateowner) {
2596 nfs4_get_stateowner(od->od_stateowner);
2597 cstate->replay_owner = od->od_stateowner;
2599 nfs4_unlock_state();
2604 * nfs4_unlock_state() called after encode
2607 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2608 struct nfsd4_close *close)
2611 struct nfs4_stateid *stp;
2613 dprintk("NFSD: nfsd4_close on file %.*s\n",
2614 (int)cstate->current_fh.fh_dentry->d_name.len,
2615 cstate->current_fh.fh_dentry->d_name.name);
2618 /* check close_lru for replay */
2619 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2622 OPEN_STATE | CLOSE_STATE,
2623 &close->cl_stateowner, &stp, NULL)))
2626 update_stateid(&stp->st_stateid);
2627 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
2629 /* release_stateid() calls nfsd_close() if needed */
2630 release_open_stateid(stp);
2632 /* place unused nfs4_stateowners on so_close_lru list to be
2633 * released by the laundromat service after the lease period
2634 * to enable us to handle CLOSE replay
2636 if (list_empty(&close->cl_stateowner->so_stateids))
2637 move_to_close_lru(close->cl_stateowner);
2639 if (close->cl_stateowner) {
2640 nfs4_get_stateowner(close->cl_stateowner);
2641 cstate->replay_owner = close->cl_stateowner;
2643 nfs4_unlock_state();
2648 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2649 struct nfsd4_delegreturn *dr)
2651 struct nfs4_delegation *dp;
2652 stateid_t *stateid = &dr->dr_stateid;
2653 struct inode *inode;
2656 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
2658 inode = cstate->current_fh.fh_dentry->d_inode;
2661 status = nfserr_bad_stateid;
2662 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2664 status = nfserr_stale_stateid;
2665 if (STALE_STATEID(stateid))
2667 status = nfserr_bad_stateid;
2668 if (!is_delegation_stateid(stateid))
2670 dp = find_delegation_stateid(inode, stateid);
2673 status = check_stateid_generation(stateid, &dp->dl_stateid);
2676 renew_client(dp->dl_client);
2678 unhash_delegation(dp);
2680 nfs4_unlock_state();
2687 * Lock owner state (byte-range locks)
2689 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
2690 #define LOCK_HASH_BITS 8
2691 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
2692 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
2695 end_offset(u64 start, u64 len)
2700 return end >= start ? end: NFS4_MAX_UINT64;
2703 /* last octet in a range */
2705 last_byte_offset(u64 start, u64 len)
2711 return end > start ? end - 1: NFS4_MAX_UINT64;
2714 #define lockownerid_hashval(id) \
2715 ((id) & LOCK_HASH_MASK)
2717 static inline unsigned int
2718 lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
2719 struct xdr_netobj *ownername)
2721 return (file_hashval(inode) + cl_id
2722 + opaque_hashval(ownername->data, ownername->len))
2726 static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
2727 static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
2728 static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
2730 static struct nfs4_stateid *
2731 find_stateid(stateid_t *stid, int flags)
2733 struct nfs4_stateid *local;
2734 u32 st_id = stid->si_stateownerid;
2735 u32 f_id = stid->si_fileid;
2736 unsigned int hashval;
2738 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
2739 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
2740 hashval = stateid_hashval(st_id, f_id);
2741 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
2742 if ((local->st_stateid.si_stateownerid == st_id) &&
2743 (local->st_stateid.si_fileid == f_id))
2748 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
2749 hashval = stateid_hashval(st_id, f_id);
2750 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
2751 if ((local->st_stateid.si_stateownerid == st_id) &&
2752 (local->st_stateid.si_fileid == f_id))
2759 static struct nfs4_delegation *
2760 find_delegation_stateid(struct inode *ino, stateid_t *stid)
2762 struct nfs4_file *fp;
2763 struct nfs4_delegation *dl;
2765 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
2766 stid->si_boot, stid->si_stateownerid,
2767 stid->si_fileid, stid->si_generation);
2769 fp = find_file(ino);
2772 dl = find_delegation_file(fp, stid);
2778 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
2779 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
2780 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
2781 * locking, this prevents us from being completely protocol-compliant. The
2782 * real solution to this problem is to start using unsigned file offsets in
2783 * the VFS, but this is a very deep change!
2786 nfs4_transform_lock_offset(struct file_lock *lock)
2788 if (lock->fl_start < 0)
2789 lock->fl_start = OFFSET_MAX;
2790 if (lock->fl_end < 0)
2791 lock->fl_end = OFFSET_MAX;
2794 /* Hack!: For now, we're defining this just so we can use a pointer to it
2795 * as a unique cookie to identify our (NFSv4's) posix locks. */
2796 static struct lock_manager_operations nfsd_posix_mng_ops = {
2800 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
2802 struct nfs4_stateowner *sop;
2805 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
2806 sop = (struct nfs4_stateowner *) fl->fl_owner;
2807 hval = lockownerid_hashval(sop->so_id);
2808 kref_get(&sop->so_ref);
2810 deny->ld_clientid = sop->so_client->cl_clientid;
2812 deny->ld_sop = NULL;
2813 deny->ld_clientid.cl_boot = 0;
2814 deny->ld_clientid.cl_id = 0;
2816 deny->ld_start = fl->fl_start;
2817 deny->ld_length = NFS4_MAX_UINT64;
2818 if (fl->fl_end != NFS4_MAX_UINT64)
2819 deny->ld_length = fl->fl_end - fl->fl_start + 1;
2820 deny->ld_type = NFS4_READ_LT;
2821 if (fl->fl_type != F_RDLCK)
2822 deny->ld_type = NFS4_WRITE_LT;
2825 static struct nfs4_stateowner *
2826 find_lockstateowner_str(struct inode *inode, clientid_t *clid,
2827 struct xdr_netobj *owner)
2829 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
2830 struct nfs4_stateowner *op;
2832 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
2833 if (same_owner_str(op, owner, clid))
2840 * Alloc a lock owner structure.
2841 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
2844 * strhashval = lock_ownerstr_hashval
2847 static struct nfs4_stateowner *
2848 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
2849 struct nfs4_stateowner *sop;
2850 struct nfs4_replay *rp;
2851 unsigned int idhashval;
2853 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
2855 idhashval = lockownerid_hashval(current_ownerid);
2856 INIT_LIST_HEAD(&sop->so_idhash);
2857 INIT_LIST_HEAD(&sop->so_strhash);
2858 INIT_LIST_HEAD(&sop->so_perclient);
2859 INIT_LIST_HEAD(&sop->so_stateids);
2860 INIT_LIST_HEAD(&sop->so_perstateid);
2861 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
2863 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
2864 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
2865 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
2866 sop->so_is_open_owner = 0;
2867 sop->so_id = current_ownerid++;
2868 sop->so_client = clp;
2869 /* It is the openowner seqid that will be incremented in encode in the
2870 * case of new lockowners; so increment the lock seqid manually: */
2871 sop->so_seqid = lock->lk_new_lock_seqid + 1;
2872 sop->so_confirmed = 1;
2873 rp = &sop->so_replay;
2874 rp->rp_status = nfserr_serverfault;
2876 rp->rp_buf = rp->rp_ibuf;
2880 static struct nfs4_stateid *
2881 alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
2883 struct nfs4_stateid *stp;
2884 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
2886 stp = nfs4_alloc_stateid();
2889 INIT_LIST_HEAD(&stp->st_hash);
2890 INIT_LIST_HEAD(&stp->st_perfile);
2891 INIT_LIST_HEAD(&stp->st_perstateowner);
2892 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
2893 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
2894 list_add(&stp->st_perfile, &fp->fi_stateids);
2895 list_add(&stp->st_perstateowner, &sop->so_stateids);
2896 stp->st_stateowner = sop;
2899 stp->st_stateid.si_boot = boot_time;
2900 stp->st_stateid.si_stateownerid = sop->so_id;
2901 stp->st_stateid.si_fileid = fp->fi_id;
2902 stp->st_stateid.si_generation = 0;
2903 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
2904 stp->st_access_bmap = open_stp->st_access_bmap;
2905 stp->st_deny_bmap = open_stp->st_deny_bmap;
2906 stp->st_openstp = open_stp;
2913 check_lock_length(u64 offset, u64 length)
2915 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
2916 LOFF_OVERFLOW(offset, length)));
2923 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2924 struct nfsd4_lock *lock)
2926 struct nfs4_stateowner *open_sop = NULL;
2927 struct nfs4_stateowner *lock_sop = NULL;
2928 struct nfs4_stateid *lock_stp;
2930 struct file_lock file_lock;
2931 struct file_lock conflock;
2933 unsigned int strhashval;
2937 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
2938 (long long) lock->lk_offset,
2939 (long long) lock->lk_length);
2941 if (check_lock_length(lock->lk_offset, lock->lk_length))
2942 return nfserr_inval;
2944 if ((status = fh_verify(rqstp, &cstate->current_fh,
2945 S_IFREG, NFSD_MAY_LOCK))) {
2946 dprintk("NFSD: nfsd4_lock: permission denied!\n");
2952 if (lock->lk_is_new) {
2954 * Client indicates that this is a new lockowner.
2955 * Use open owner and open stateid to create lock owner and
2958 struct nfs4_stateid *open_stp = NULL;
2959 struct nfs4_file *fp;
2961 status = nfserr_stale_clientid;
2962 if (STALE_CLIENTID(&lock->lk_new_clientid))
2965 /* validate and update open stateid and open seqid */
2966 status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2967 lock->lk_new_open_seqid,
2968 &lock->lk_new_open_stateid,
2970 &lock->lk_replay_owner, &open_stp,
2974 open_sop = lock->lk_replay_owner;
2975 /* create lockowner and lock stateid */
2976 fp = open_stp->st_file;
2977 strhashval = lock_ownerstr_hashval(fp->fi_inode,
2978 open_sop->so_client->cl_clientid.cl_id,
2979 &lock->v.new.owner);
2980 /* XXX: Do we need to check for duplicate stateowners on
2981 * the same file, or should they just be allowed (and
2982 * create new stateids)? */
2983 status = nfserr_resource;
2984 lock_sop = alloc_init_lock_stateowner(strhashval,
2985 open_sop->so_client, open_stp, lock);
2986 if (lock_sop == NULL)
2988 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
2989 if (lock_stp == NULL)
2992 /* lock (lock owner + lock stateid) already exists */
2993 status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2994 lock->lk_old_lock_seqid,
2995 &lock->lk_old_lock_stateid,
2997 &lock->lk_replay_owner, &lock_stp, lock);
3000 lock_sop = lock->lk_replay_owner;
3002 /* lock->lk_replay_owner and lock_stp have been created or found */
3003 filp = lock_stp->st_vfs_file;
3005 status = nfserr_grace;
3006 if (locks_in_grace() && !lock->lk_reclaim)
3008 status = nfserr_no_grace;
3009 if (!locks_in_grace() && lock->lk_reclaim)
3012 locks_init_lock(&file_lock);
3013 switch (lock->lk_type) {
3016 file_lock.fl_type = F_RDLCK;
3020 case NFS4_WRITEW_LT:
3021 file_lock.fl_type = F_WRLCK;
3025 status = nfserr_inval;
3028 file_lock.fl_owner = (fl_owner_t)lock_sop;
3029 file_lock.fl_pid = current->tgid;
3030 file_lock.fl_file = filp;
3031 file_lock.fl_flags = FL_POSIX;
3032 file_lock.fl_lmops = &nfsd_posix_mng_ops;
3034 file_lock.fl_start = lock->lk_offset;
3035 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
3036 nfs4_transform_lock_offset(&file_lock);
3039 * Try to lock the file in the VFS.
3040 * Note: locks.c uses the BKL to protect the inode's lock list.
3043 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
3045 case 0: /* success! */
3046 update_stateid(&lock_stp->st_stateid);
3047 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3051 case (EAGAIN): /* conflock holds conflicting lock */
3052 status = nfserr_denied;
3053 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3054 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3057 status = nfserr_deadlock;
3060 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
3061 status = nfserr_resource;
3065 if (status && lock->lk_is_new && lock_sop)
3066 release_lockowner(lock_sop);
3067 if (lock->lk_replay_owner) {
3068 nfs4_get_stateowner(lock->lk_replay_owner);
3069 cstate->replay_owner = lock->lk_replay_owner;
3071 nfs4_unlock_state();
3076 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3077 * so we do a temporary open here just to get an open file to pass to
3078 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3081 static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3086 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3089 err = vfs_test_lock(file, lock);
3098 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3099 struct nfsd4_lockt *lockt)
3101 struct inode *inode;
3102 struct file_lock file_lock;
3106 if (locks_in_grace())
3107 return nfserr_grace;
3109 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3110 return nfserr_inval;
3112 lockt->lt_stateowner = NULL;
3115 status = nfserr_stale_clientid;
3116 if (STALE_CLIENTID(&lockt->lt_clientid))
3119 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
3120 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
3121 if (status == nfserr_symlink)
3122 status = nfserr_inval;
3126 inode = cstate->current_fh.fh_dentry->d_inode;
3127 locks_init_lock(&file_lock);
3128 switch (lockt->lt_type) {
3131 file_lock.fl_type = F_RDLCK;
3134 case NFS4_WRITEW_LT:
3135 file_lock.fl_type = F_WRLCK;
3138 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
3139 status = nfserr_inval;
3143 lockt->lt_stateowner = find_lockstateowner_str(inode,
3144 &lockt->lt_clientid, &lockt->lt_owner);
3145 if (lockt->lt_stateowner)
3146 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3147 file_lock.fl_pid = current->tgid;
3148 file_lock.fl_flags = FL_POSIX;
3150 file_lock.fl_start = lockt->lt_offset;
3151 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
3153 nfs4_transform_lock_offset(&file_lock);
3156 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
3158 status = nfserrno(error);
3161 if (file_lock.fl_type != F_UNLCK) {
3162 status = nfserr_denied;
3163 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
3166 nfs4_unlock_state();
3171 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3172 struct nfsd4_locku *locku)
3174 struct nfs4_stateid *stp;
3175 struct file *filp = NULL;
3176 struct file_lock file_lock;
3180 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3181 (long long) locku->lu_offset,
3182 (long long) locku->lu_length);
3184 if (check_lock_length(locku->lu_offset, locku->lu_length))
3185 return nfserr_inval;
3189 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
3193 &locku->lu_stateowner, &stp, NULL)))
3196 filp = stp->st_vfs_file;
3198 locks_init_lock(&file_lock);
3199 file_lock.fl_type = F_UNLCK;
3200 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3201 file_lock.fl_pid = current->tgid;
3202 file_lock.fl_file = filp;
3203 file_lock.fl_flags = FL_POSIX;
3204 file_lock.fl_lmops = &nfsd_posix_mng_ops;
3205 file_lock.fl_start = locku->lu_offset;
3207 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
3208 nfs4_transform_lock_offset(&file_lock);
3211 * Try to unlock the file in the VFS.
3213 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
3215 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
3219 * OK, unlock succeeded; the only thing left to do is update the stateid.
3221 update_stateid(&stp->st_stateid);
3222 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3225 if (locku->lu_stateowner) {
3226 nfs4_get_stateowner(locku->lu_stateowner);
3227 cstate->replay_owner = locku->lu_stateowner;
3229 nfs4_unlock_state();
3233 status = nfserrno(err);
3239 * 1: locks held by lockowner
3240 * 0: no locks held by lockowner
3243 check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3245 struct file_lock **flpp;
3246 struct inode *inode = filp->f_path.dentry->d_inode;
3250 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
3251 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
3262 nfsd4_release_lockowner(struct svc_rqst *rqstp,
3263 struct nfsd4_compound_state *cstate,
3264 struct nfsd4_release_lockowner *rlockowner)
3266 clientid_t *clid = &rlockowner->rl_clientid;
3267 struct nfs4_stateowner *sop;
3268 struct nfs4_stateid *stp;
3269 struct xdr_netobj *owner = &rlockowner->rl_owner;
3270 struct list_head matches;
3274 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3275 clid->cl_boot, clid->cl_id);
3277 /* XXX check for lease expiration */
3279 status = nfserr_stale_clientid;
3280 if (STALE_CLIENTID(clid))
3285 status = nfserr_locks_held;
3286 /* XXX: we're doing a linear search through all the lockowners.
3287 * Yipes! For now we'll just hope clients aren't really using
3288 * release_lockowner much, but eventually we have to fix these
3289 * data structures. */
3290 INIT_LIST_HEAD(&matches);
3291 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3292 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
3293 if (!same_owner_str(sop, owner, clid))
3295 list_for_each_entry(stp, &sop->so_stateids,
3297 if (check_for_locks(stp->st_vfs_file, sop))
3299 /* Note: so_perclient unused for lockowners,
3300 * so it's OK to fool with here. */
3301 list_add(&sop->so_perclient, &matches);
3305 /* Clients probably won't expect us to return with some (but not all)
3306 * of the lockowner state released; so don't release any until all
3307 * have been checked. */
3309 while (!list_empty(&matches)) {
3310 sop = list_entry(matches.next, struct nfs4_stateowner,
3312 /* unhash_stateowner deletes so_perclient only
3313 * for openowners. */
3314 list_del(&sop->so_perclient);
3315 release_lockowner(sop);
3318 nfs4_unlock_state();
3322 static inline struct nfs4_client_reclaim *
3325 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
3329 nfs4_has_reclaimed_state(const char *name)
3331 unsigned int strhashval = clientstr_hashval(name);
3332 struct nfs4_client *clp;
3334 clp = find_confirmed_client_by_str(name, strhashval);
3339 * failure => all reset bets are off, nfserr_no_grace...
3342 nfs4_client_to_reclaim(const char *name)
3344 unsigned int strhashval;
3345 struct nfs4_client_reclaim *crp = NULL;
3347 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3348 crp = alloc_reclaim();
3351 strhashval = clientstr_hashval(name);
3352 INIT_LIST_HEAD(&crp->cr_strhash);
3353 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
3354 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
3355 reclaim_str_hashtbl_size++;
3360 nfs4_release_reclaim(void)
3362 struct nfs4_client_reclaim *crp = NULL;
3365 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3366 while (!list_empty(&reclaim_str_hashtbl[i])) {
3367 crp = list_entry(reclaim_str_hashtbl[i].next,
3368 struct nfs4_client_reclaim, cr_strhash);
3369 list_del(&crp->cr_strhash);
3371 reclaim_str_hashtbl_size--;
3374 BUG_ON(reclaim_str_hashtbl_size);
3378 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
3379 static struct nfs4_client_reclaim *
3380 nfs4_find_reclaim_client(clientid_t *clid)
3382 unsigned int strhashval;
3383 struct nfs4_client *clp;
3384 struct nfs4_client_reclaim *crp = NULL;
3387 /* find clientid in conf_id_hashtbl */
3388 clp = find_confirmed_client(clid);
3392 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3393 clp->cl_name.len, clp->cl_name.data,
3396 /* find clp->cl_name in reclaim_str_hashtbl */
3397 strhashval = clientstr_hashval(clp->cl_recdir);
3398 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
3399 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
3407 * Called from OPEN. Look for clientid in reclaim list.
3410 nfs4_check_open_reclaim(clientid_t *clid)
3412 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
3415 /* initialization to perform at module load time: */
3418 nfs4_state_init(void)
3422 status = nfsd4_init_slabs();
3425 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3426 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3427 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3428 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3429 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
3431 for (i = 0; i < SESSION_HASH_SIZE; i++)
3432 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
3433 for (i = 0; i < FILE_HASH_SIZE; i++) {
3434 INIT_LIST_HEAD(&file_hashtbl[i]);
3436 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3437 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3438 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3440 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3441 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3442 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3444 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3445 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3446 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3448 memset(&onestateid, ~0, sizeof(stateid_t));
3449 INIT_LIST_HEAD(&close_lru);
3450 INIT_LIST_HEAD(&client_lru);
3451 INIT_LIST_HEAD(&del_recall_lru);
3452 for (i = 0; i < CLIENT_HASH_SIZE; i++)
3453 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
3454 reclaim_str_hashtbl_size = 0;
3459 nfsd4_load_reboot_recovery_data(void)
3464 nfsd4_init_recdir(user_recovery_dirname);
3465 status = nfsd4_recdir_load();
3466 nfs4_unlock_state();
3468 printk("NFSD: Failure reading reboot recovery data\n");
3472 get_nfs4_grace_period(void)
3474 return max(user_lease_time, lease_time) * HZ;
3478 * Since the lifetime of a delegation isn't limited to that of an open, a
3479 * client may quite reasonably hang on to a delegation as long as it has
3480 * the inode cached. This becomes an obvious problem the first time a
3481 * client's inode cache approaches the size of the server's total memory.
3483 * For now we avoid this problem by imposing a hard limit on the number
3484 * of delegations, which varies according to the server's memory size.
3487 set_max_delegations(void)
3490 * Allow at most 4 delegations per megabyte of RAM. Quick
3491 * estimates suggest that in the worst case (where every delegation
3492 * is for a different inode), a delegation could take about 1.5K,
3493 * giving a worst case usage of about 6% of memory.
3495 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
3498 /* initialization to perform when the nfsd service is started: */
3501 __nfs4_state_start(void)
3503 unsigned long grace_time;
3505 boot_time = get_seconds();
3506 grace_time = get_nfs4_grace_period();
3507 lease_time = user_lease_time;
3508 locks_start_grace(&nfsd4_manager);
3509 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
3511 laundry_wq = create_singlethread_workqueue("nfsd4");
3512 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
3513 set_max_delegations();
3517 nfs4_state_start(void)
3521 nfsd4_load_reboot_recovery_data();
3522 __nfs4_state_start();
3528 nfs4_lease_time(void)
3534 __nfs4_state_shutdown(void)
3537 struct nfs4_client *clp = NULL;
3538 struct nfs4_delegation *dp = NULL;
3539 struct list_head *pos, *next, reaplist;
3541 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3542 while (!list_empty(&conf_id_hashtbl[i])) {
3543 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
3546 while (!list_empty(&unconf_str_hashtbl[i])) {
3547 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
3551 INIT_LIST_HEAD(&reaplist);
3552 spin_lock(&recall_lock);
3553 list_for_each_safe(pos, next, &del_recall_lru) {
3554 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3555 list_move(&dp->dl_recall_lru, &reaplist);
3557 spin_unlock(&recall_lock);
3558 list_for_each_safe(pos, next, &reaplist) {
3559 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3560 list_del_init(&dp->dl_recall_lru);
3561 unhash_delegation(dp);
3564 nfsd4_shutdown_recdir();
3569 nfs4_state_shutdown(void)
3571 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
3572 destroy_workqueue(laundry_wq);
3573 locks_end_grace(&nfsd4_manager);
3575 nfs4_release_reclaim();
3576 __nfs4_state_shutdown();
3577 nfs4_unlock_state();
3581 * user_recovery_dirname is protected by the nfsd_mutex since it's only
3582 * accessed when nfsd is starting.
3585 nfs4_set_recdir(char *recdir)
3587 strcpy(user_recovery_dirname, recdir);
3591 * Change the NFSv4 recovery directory to recdir.
3594 nfs4_reset_recoverydir(char *recdir)
3599 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
3603 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
3604 nfs4_set_recdir(recdir);
3612 nfs4_recoverydir(void)
3614 return user_recovery_dirname;
3618 * Called when leasetime is changed.
3620 * The only way the protocol gives us to handle on-the-fly lease changes is to
3621 * simulate a reboot. Instead of doing that, we just wait till the next time
3622 * we start to register any changes in lease time. If the administrator
3623 * really wants to change the lease time *now*, they can go ahead and bring
3624 * nfsd down and then back up again after changing the lease time.
3626 * user_lease_time is protected by nfsd_mutex since it's only really accessed
3627 * when nfsd is starting
3630 nfs4_reset_lease(time_t leasetime)
3632 user_lease_time = leasetime;