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))
719 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
720 * parameter. Matching is based on the fact the at least one of the
721 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
723 * FIXME: we need to unify the clientid namespaces for nfsv4.x
724 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
725 * and SET_CLIENTID{,_CONFIRM}
728 match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
730 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
731 return use_exchange_id == has_exchange_flags;
734 static struct nfs4_client *
735 find_confirmed_client_by_str(const char *dname, unsigned int hashval,
736 bool use_exchange_id)
738 struct nfs4_client *clp;
740 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
741 if (same_name(clp->cl_recdir, dname) &&
742 match_clientid_establishment(clp, use_exchange_id))
748 static struct nfs4_client *
749 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
750 bool use_exchange_id)
752 struct nfs4_client *clp;
754 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
755 if (same_name(clp->cl_recdir, dname) &&
756 match_clientid_establishment(clp, use_exchange_id))
762 /* a helper function for parse_callback */
764 parse_octet(unsigned int *lenp, char **addrp)
766 unsigned int len = *lenp;
778 if ((c < '0') || (c > '9')) {
784 n = (n * 10) + (c - '0');
795 /* parse and set the setclientid ipv4 callback address */
797 parse_ipv4(unsigned int addr_len, char *addr_val, unsigned int *cbaddrp, unsigned short *cbportp)
802 u32 addrlen = addr_len;
803 char *addr = addr_val;
808 for(i = 4; i > 0 ; i--) {
809 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
812 cbaddr |= (temp << shift);
820 for(i = 2; i > 0 ; i--) {
821 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
824 cbport |= (temp << shift);
833 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se)
835 struct nfs4_callback *cb = &clp->cl_callback;
837 /* Currently, we only support tcp for the callback channel */
838 if ((se->se_callback_netid_len != 3) || memcmp((char *)se->se_callback_netid_val, "tcp", 3))
841 if ( !(parse_ipv4(se->se_callback_addr_len, se->se_callback_addr_val,
842 &cb->cb_addr, &cb->cb_port)))
844 cb->cb_prog = se->se_callback_prog;
845 cb->cb_ident = se->se_callback_ident;
848 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
849 "will not receive delegations\n",
850 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
856 nfsd4_set_statp(struct svc_rqst *rqstp, __be32 *statp)
858 struct nfsd4_compoundres *resp = rqstp->rq_resp;
860 resp->cstate.statp = statp;
864 * Dereference the result pages.
867 nfsd4_release_respages(struct page **respages, short resused)
871 dprintk("--> %s\n", __func__);
872 for (i = 0; i < resused; i++) {
875 put_page(respages[i]);
881 nfsd4_copy_pages(struct page **topages, struct page **frompages, short count)
885 for (i = 0; i < count; i++) {
886 topages[i] = frompages[i];
889 get_page(topages[i]);
894 * Cache the reply pages up to NFSD_PAGES_PER_SLOT + 1, clearing the previous
895 * pages. We add a page to NFSD_PAGES_PER_SLOT for the case where the total
896 * length of the XDR response is less than se_fmaxresp_cached
897 * (NFSD_PAGES_PER_SLOT * PAGE_SIZE) but the xdr_buf pages is used for a
898 * of the reply (e.g. readdir).
900 * Store the base and length of the rq_req.head[0] page
901 * of the NFSv4.1 data, just past the rpc header.
904 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
906 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
907 struct svc_rqst *rqstp = resp->rqstp;
908 struct nfsd4_compoundargs *args = rqstp->rq_argp;
909 struct nfsd4_op *op = &args->ops[resp->opcnt];
910 struct kvec *resv = &rqstp->rq_res.head[0];
912 dprintk("--> %s entry %p\n", __func__, entry);
914 /* Don't cache a failed OP_SEQUENCE. */
915 if (resp->opcnt == 1 && op->opnum == OP_SEQUENCE && resp->cstate.status)
917 nfsd4_release_respages(entry->ce_respages, entry->ce_resused);
918 entry->ce_resused = rqstp->rq_resused;
919 if (entry->ce_resused > NFSD_PAGES_PER_SLOT + 1)
920 entry->ce_resused = NFSD_PAGES_PER_SLOT + 1;
921 nfsd4_copy_pages(entry->ce_respages, rqstp->rq_respages,
923 entry->ce_status = resp->cstate.status;
924 entry->ce_datav.iov_base = resp->cstate.statp;
925 entry->ce_datav.iov_len = resv->iov_len - ((char *)resp->cstate.statp -
926 (char *)page_address(rqstp->rq_respages[0]));
927 entry->ce_opcnt = resp->opcnt;
928 /* Current request rpc header length*/
929 entry->ce_rpchdrlen = (char *)resp->cstate.statp -
930 (char *)page_address(rqstp->rq_respages[0]);
934 * We keep the rpc header, but take the nfs reply from the replycache.
937 nfsd41_copy_replay_data(struct nfsd4_compoundres *resp,
938 struct nfsd4_cache_entry *entry)
940 struct svc_rqst *rqstp = resp->rqstp;
941 struct kvec *resv = &resp->rqstp->rq_res.head[0];
944 /* Current request rpc header length*/
945 len = (char *)resp->cstate.statp -
946 (char *)page_address(rqstp->rq_respages[0]);
947 if (entry->ce_datav.iov_len + len > PAGE_SIZE) {
948 dprintk("%s v41 cached reply too large (%Zd).\n", __func__,
949 entry->ce_datav.iov_len);
952 /* copy the cached reply nfsd data past the current rpc header */
953 memcpy((char *)resv->iov_base + len, entry->ce_datav.iov_base,
954 entry->ce_datav.iov_len);
955 resv->iov_len = len + entry->ce_datav.iov_len;
960 * Keep the first page of the replay. Copy the NFSv4.1 data from the first
961 * cached page. Replace any futher replay pages from the cache.
964 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp)
966 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
969 dprintk("--> %s entry %p\n", __func__, entry);
972 if (!nfsd41_copy_replay_data(resp, entry)) {
974 * Not enough room to use the replay rpc header, send the
975 * cached header. Release all the allocated result pages.
977 svc_free_res_pages(resp->rqstp);
978 nfsd4_copy_pages(resp->rqstp->rq_respages, entry->ce_respages,
981 /* Release all but the first allocated result page */
983 resp->rqstp->rq_resused--;
984 svc_free_res_pages(resp->rqstp);
986 nfsd4_copy_pages(&resp->rqstp->rq_respages[1],
987 &entry->ce_respages[1],
988 entry->ce_resused - 1);
991 resp->rqstp->rq_resused = entry->ce_resused;
992 status = entry->ce_status;
998 * Set the exchange_id flags returned by the server.
1001 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1003 /* pNFS is not supported */
1004 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1006 /* Referrals are supported, Migration is not. */
1007 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1009 /* set the wire flags to return to client. */
1010 clid->flags = new->cl_exchange_flags;
1014 nfsd4_exchange_id(struct svc_rqst *rqstp,
1015 struct nfsd4_compound_state *cstate,
1016 struct nfsd4_exchange_id *exid)
1018 struct nfs4_client *unconf, *conf, *new;
1020 unsigned int strhashval;
1021 char dname[HEXDIR_LEN];
1022 nfs4_verifier verf = exid->verifier;
1023 u32 ip_addr = svc_addr_in(rqstp)->sin_addr.s_addr;
1025 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1026 " ip_addr=%u flags %x, spa_how %d\n",
1027 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1028 ip_addr, exid->flags, exid->spa_how);
1030 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1031 return nfserr_inval;
1033 /* Currently only support SP4_NONE */
1034 switch (exid->spa_how) {
1038 return nfserr_encr_alg_unsupp;
1040 BUG(); /* checked by xdr code */
1042 return nfserr_serverfault; /* no excuse :-/ */
1045 status = nfs4_make_rec_clidname(dname, &exid->clname);
1050 strhashval = clientstr_hashval(dname);
1055 conf = find_confirmed_client_by_str(dname, strhashval, true);
1057 if (!same_verf(&verf, &conf->cl_verifier)) {
1058 /* 18.35.4 case 8 */
1059 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1060 status = nfserr_not_same;
1063 /* Client reboot: destroy old state */
1064 expire_client(conf);
1067 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1068 /* 18.35.4 case 9 */
1069 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1070 status = nfserr_perm;
1073 expire_client(conf);
1076 if (ip_addr != conf->cl_addr &&
1077 !(exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A)) {
1078 /* Client collision. 18.35.4 case 3 */
1079 status = nfserr_clid_inuse;
1083 * Set bit when the owner id and verifier map to an already
1084 * confirmed client id (18.35.3).
1086 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1089 * Falling into 18.35.4 case 2, possible router replay.
1090 * Leave confirmed record intact and return same result.
1092 copy_verf(conf, &verf);
1096 /* 18.35.4 case 7 */
1097 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1098 status = nfserr_noent;
1103 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
1106 * Possible retry or client restart. Per 18.35.4 case 4,
1107 * a new unconfirmed record should be generated regardless
1108 * of whether any properties have changed.
1110 expire_client(unconf);
1115 new = create_client(exid->clname, dname);
1117 status = nfserr_resource;
1121 copy_verf(new, &verf);
1122 copy_cred(&new->cl_cred, &rqstp->rq_cred);
1123 new->cl_addr = ip_addr;
1126 add_to_unconfirmed(new, strhashval);
1128 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1129 exid->clientid.cl_id = new->cl_clientid.cl_id;
1131 new->cl_seqid = exid->seqid = 1;
1132 nfsd4_set_ex_flags(new, exid);
1134 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1135 new->cl_seqid, new->cl_exchange_flags);
1139 nfs4_unlock_state();
1141 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1146 check_slot_seqid(u32 seqid, struct nfsd4_slot *slot)
1148 dprintk("%s enter. seqid %d slot->sl_seqid %d\n", __func__, seqid,
1151 /* The slot is in use, and no response has been sent. */
1152 if (slot->sl_inuse) {
1153 if (seqid == slot->sl_seqid)
1154 return nfserr_jukebox;
1156 return nfserr_seq_misordered;
1159 if (likely(seqid == slot->sl_seqid + 1))
1162 if (seqid == slot->sl_seqid)
1163 return nfserr_replay_cache;
1165 if (seqid == 1 && (slot->sl_seqid + 1) == 0)
1167 /* Misordered replay or misordered new request */
1168 return nfserr_seq_misordered;
1172 nfsd4_create_session(struct svc_rqst *rqstp,
1173 struct nfsd4_compound_state *cstate,
1174 struct nfsd4_create_session *cr_ses)
1176 return -1; /* stub */
1180 nfsd4_destroy_session(struct svc_rqst *r,
1181 struct nfsd4_compound_state *cstate,
1182 struct nfsd4_destroy_session *sessionid)
1184 return -1; /* stub */
1188 nfsd4_sequence(struct svc_rqst *rqstp,
1189 struct nfsd4_compound_state *cstate,
1190 struct nfsd4_sequence *seq)
1192 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1193 struct nfsd4_session *session;
1194 struct nfsd4_slot *slot;
1197 if (resp->opcnt != 1)
1198 return nfserr_sequence_pos;
1200 spin_lock(&sessionid_lock);
1201 status = nfserr_badsession;
1202 session = find_in_sessionid_hashtbl(&seq->sessionid);
1206 status = nfserr_badslot;
1207 if (seq->slotid >= session->se_fnumslots)
1210 slot = &session->se_slots[seq->slotid];
1211 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1213 status = check_slot_seqid(seq->seqid, slot);
1214 if (status == nfserr_replay_cache) {
1215 cstate->slot = slot;
1216 cstate->session = session;
1222 /* Success! bump slot seqid */
1223 slot->sl_inuse = true;
1224 slot->sl_seqid = seq->seqid;
1226 cstate->slot = slot;
1227 cstate->session = session;
1230 /* Renew the clientid on success and on replay.
1231 * Hold a session reference until done processing the compound:
1232 * nfsd4_put_session called only if the cstate slot is set.
1234 renew_client(session->se_client);
1235 nfsd4_get_session(session);
1237 spin_unlock(&sessionid_lock);
1238 dprintk("%s: return %d\n", __func__, ntohl(status));
1243 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1244 struct nfsd4_setclientid *setclid)
1246 struct sockaddr_in *sin = svc_addr_in(rqstp);
1247 struct xdr_netobj clname = {
1248 .len = setclid->se_namelen,
1249 .data = setclid->se_name,
1251 nfs4_verifier clverifier = setclid->se_verf;
1252 unsigned int strhashval;
1253 struct nfs4_client *conf, *unconf, *new;
1256 char dname[HEXDIR_LEN];
1258 if (!check_name(clname))
1259 return nfserr_inval;
1261 status = nfs4_make_rec_clidname(dname, &clname);
1266 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1267 * We get here on a DRC miss.
1270 strhashval = clientstr_hashval(dname);
1273 conf = find_confirmed_client_by_str(dname, strhashval, false);
1275 /* RFC 3530 14.2.33 CASE 0: */
1276 status = nfserr_clid_inuse;
1277 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1278 dprintk("NFSD: setclientid: string in use by client"
1279 " at %pI4\n", &conf->cl_addr);
1284 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1285 * has a description of SETCLIENTID request processing consisting
1286 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1288 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1289 status = nfserr_resource;
1292 * RFC 3530 14.2.33 CASE 4:
1293 * placed first, because it is the normal case
1296 expire_client(unconf);
1297 new = create_client(clname, dname);
1301 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1303 * RFC 3530 14.2.33 CASE 1:
1304 * probable callback update
1307 /* Note this is removing unconfirmed {*x***},
1308 * which is stronger than RFC recommended {vxc**}.
1309 * This has the advantage that there is at most
1310 * one {*x***} in either list at any time.
1312 expire_client(unconf);
1314 new = create_client(clname, dname);
1317 copy_clid(new, conf);
1318 } else if (!unconf) {
1320 * RFC 3530 14.2.33 CASE 2:
1321 * probable client reboot; state will be removed if
1324 new = create_client(clname, dname);
1330 * RFC 3530 14.2.33 CASE 3:
1331 * probable client reboot; state will be removed if
1334 expire_client(unconf);
1335 new = create_client(clname, dname);
1340 copy_verf(new, &clverifier);
1341 new->cl_addr = sin->sin_addr.s_addr;
1342 new->cl_flavor = rqstp->rq_flavor;
1343 princ = svc_gss_principal(rqstp);
1345 new->cl_principal = kstrdup(princ, GFP_KERNEL);
1346 if (new->cl_principal == NULL) {
1351 copy_cred(&new->cl_cred, &rqstp->rq_cred);
1353 gen_callback(new, setclid);
1354 add_to_unconfirmed(new, strhashval);
1355 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1356 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1357 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1360 nfs4_unlock_state();
1366 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1367 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1368 * bullets, labeled as CASE1 - CASE4 below.
1371 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1372 struct nfsd4_compound_state *cstate,
1373 struct nfsd4_setclientid_confirm *setclientid_confirm)
1375 struct sockaddr_in *sin = svc_addr_in(rqstp);
1376 struct nfs4_client *conf, *unconf;
1377 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1378 clientid_t * clid = &setclientid_confirm->sc_clientid;
1381 if (STALE_CLIENTID(clid))
1382 return nfserr_stale_clientid;
1384 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1385 * We get here on a DRC miss.
1390 conf = find_confirmed_client(clid);
1391 unconf = find_unconfirmed_client(clid);
1393 status = nfserr_clid_inuse;
1394 if (conf && conf->cl_addr != sin->sin_addr.s_addr)
1396 if (unconf && unconf->cl_addr != sin->sin_addr.s_addr)
1400 * section 14.2.34 of RFC 3530 has a description of
1401 * SETCLIENTID_CONFIRM request processing consisting
1402 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1404 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
1406 * RFC 3530 14.2.34 CASE 1:
1409 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1410 status = nfserr_clid_inuse;
1412 /* XXX: We just turn off callbacks until we can handle
1413 * change request correctly. */
1414 atomic_set(&conf->cl_callback.cb_set, 0);
1416 nfsd4_remove_clid_dir(unconf);
1417 expire_client(unconf);
1421 } else if (conf && !unconf) {
1423 * RFC 3530 14.2.34 CASE 2:
1424 * probable retransmitted request; play it safe and
1427 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1428 status = nfserr_clid_inuse;
1431 } else if (!conf && unconf
1432 && same_verf(&unconf->cl_confirm, &confirm)) {
1434 * RFC 3530 14.2.34 CASE 3:
1435 * Normal case; new or rebooted client:
1437 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1438 status = nfserr_clid_inuse;
1441 clientstr_hashval(unconf->cl_recdir);
1442 conf = find_confirmed_client_by_str(unconf->cl_recdir,
1445 nfsd4_remove_clid_dir(conf);
1446 expire_client(conf);
1448 move_to_confirmed(unconf);
1450 nfsd4_probe_callback(conf);
1453 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1454 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
1457 * RFC 3530 14.2.34 CASE 4:
1458 * Client probably hasn't noticed that we rebooted yet.
1460 status = nfserr_stale_clientid;
1462 /* check that we have hit one of the cases...*/
1463 status = nfserr_clid_inuse;
1466 nfs4_unlock_state();
1470 /* OPEN Share state helper functions */
1471 static inline struct nfs4_file *
1472 alloc_init_file(struct inode *ino)
1474 struct nfs4_file *fp;
1475 unsigned int hashval = file_hashval(ino);
1477 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1479 atomic_set(&fp->fi_ref, 1);
1480 INIT_LIST_HEAD(&fp->fi_hash);
1481 INIT_LIST_HEAD(&fp->fi_stateids);
1482 INIT_LIST_HEAD(&fp->fi_delegations);
1483 spin_lock(&recall_lock);
1484 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
1485 spin_unlock(&recall_lock);
1486 fp->fi_inode = igrab(ino);
1487 fp->fi_id = current_fileid++;
1488 fp->fi_had_conflict = false;
1495 nfsd4_free_slab(struct kmem_cache **slab)
1499 kmem_cache_destroy(*slab);
1504 nfsd4_free_slabs(void)
1506 nfsd4_free_slab(&stateowner_slab);
1507 nfsd4_free_slab(&file_slab);
1508 nfsd4_free_slab(&stateid_slab);
1509 nfsd4_free_slab(&deleg_slab);
1513 nfsd4_init_slabs(void)
1515 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
1516 sizeof(struct nfs4_stateowner), 0, 0, NULL);
1517 if (stateowner_slab == NULL)
1519 file_slab = kmem_cache_create("nfsd4_files",
1520 sizeof(struct nfs4_file), 0, 0, NULL);
1521 if (file_slab == NULL)
1523 stateid_slab = kmem_cache_create("nfsd4_stateids",
1524 sizeof(struct nfs4_stateid), 0, 0, NULL);
1525 if (stateid_slab == NULL)
1527 deleg_slab = kmem_cache_create("nfsd4_delegations",
1528 sizeof(struct nfs4_delegation), 0, 0, NULL);
1529 if (deleg_slab == NULL)
1534 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1539 nfs4_free_stateowner(struct kref *kref)
1541 struct nfs4_stateowner *sop =
1542 container_of(kref, struct nfs4_stateowner, so_ref);
1543 kfree(sop->so_owner.data);
1544 kmem_cache_free(stateowner_slab, sop);
1547 static inline struct nfs4_stateowner *
1548 alloc_stateowner(struct xdr_netobj *owner)
1550 struct nfs4_stateowner *sop;
1552 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1553 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1554 memcpy(sop->so_owner.data, owner->data, owner->len);
1555 sop->so_owner.len = owner->len;
1556 kref_init(&sop->so_ref);
1559 kmem_cache_free(stateowner_slab, sop);
1564 static struct nfs4_stateowner *
1565 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1566 struct nfs4_stateowner *sop;
1567 struct nfs4_replay *rp;
1568 unsigned int idhashval;
1570 if (!(sop = alloc_stateowner(&open->op_owner)))
1572 idhashval = ownerid_hashval(current_ownerid);
1573 INIT_LIST_HEAD(&sop->so_idhash);
1574 INIT_LIST_HEAD(&sop->so_strhash);
1575 INIT_LIST_HEAD(&sop->so_perclient);
1576 INIT_LIST_HEAD(&sop->so_stateids);
1577 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1578 INIT_LIST_HEAD(&sop->so_close_lru);
1580 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1581 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
1582 list_add(&sop->so_perclient, &clp->cl_openowners);
1583 sop->so_is_open_owner = 1;
1584 sop->so_id = current_ownerid++;
1585 sop->so_client = clp;
1586 sop->so_seqid = open->op_seqid;
1587 sop->so_confirmed = 0;
1588 rp = &sop->so_replay;
1589 rp->rp_status = nfserr_serverfault;
1591 rp->rp_buf = rp->rp_ibuf;
1596 init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1597 struct nfs4_stateowner *sop = open->op_stateowner;
1598 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1600 INIT_LIST_HEAD(&stp->st_hash);
1601 INIT_LIST_HEAD(&stp->st_perstateowner);
1602 INIT_LIST_HEAD(&stp->st_lockowners);
1603 INIT_LIST_HEAD(&stp->st_perfile);
1604 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
1605 list_add(&stp->st_perstateowner, &sop->so_stateids);
1606 list_add(&stp->st_perfile, &fp->fi_stateids);
1607 stp->st_stateowner = sop;
1610 stp->st_stateid.si_boot = boot_time;
1611 stp->st_stateid.si_stateownerid = sop->so_id;
1612 stp->st_stateid.si_fileid = fp->fi_id;
1613 stp->st_stateid.si_generation = 0;
1614 stp->st_access_bmap = 0;
1615 stp->st_deny_bmap = 0;
1616 __set_bit(open->op_share_access, &stp->st_access_bmap);
1617 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1618 stp->st_openstp = NULL;
1622 move_to_close_lru(struct nfs4_stateowner *sop)
1624 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1626 list_move_tail(&sop->so_close_lru, &close_lru);
1627 sop->so_time = get_seconds();
1631 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1634 return (sop->so_owner.len == owner->len) &&
1635 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1636 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1639 static struct nfs4_stateowner *
1640 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1642 struct nfs4_stateowner *so = NULL;
1644 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
1645 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1651 /* search file_hashtbl[] for file */
1652 static struct nfs4_file *
1653 find_file(struct inode *ino)
1655 unsigned int hashval = file_hashval(ino);
1656 struct nfs4_file *fp;
1658 spin_lock(&recall_lock);
1659 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
1660 if (fp->fi_inode == ino) {
1662 spin_unlock(&recall_lock);
1666 spin_unlock(&recall_lock);
1670 static inline int access_valid(u32 x)
1672 if (x < NFS4_SHARE_ACCESS_READ)
1674 if (x > NFS4_SHARE_ACCESS_BOTH)
1679 static inline int deny_valid(u32 x)
1681 /* Note: unlike access bits, deny bits may be zero. */
1682 return x <= NFS4_SHARE_DENY_BOTH;
1686 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1687 * st_{access,deny}_bmap field of the stateid, in order to track not
1688 * only what share bits are currently in force, but also what
1689 * combinations of share bits previous opens have used. This allows us
1690 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1691 * return an error if the client attempt to downgrade to a combination
1692 * of share bits not explicable by closing some of its previous opens.
1694 * XXX: This enforcement is actually incomplete, since we don't keep
1695 * track of access/deny bit combinations; so, e.g., we allow:
1697 * OPEN allow read, deny write
1698 * OPEN allow both, deny none
1699 * DOWNGRADE allow read, deny none
1701 * which we should reject.
1704 set_access(unsigned int *access, unsigned long bmap) {
1708 for (i = 1; i < 4; i++) {
1709 if (test_bit(i, &bmap))
1715 set_deny(unsigned int *deny, unsigned long bmap) {
1719 for (i = 0; i < 4; i++) {
1720 if (test_bit(i, &bmap))
1726 test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1727 unsigned int access, deny;
1729 set_access(&access, stp->st_access_bmap);
1730 set_deny(&deny, stp->st_deny_bmap);
1731 if ((access & open->op_share_deny) || (deny & open->op_share_access))
1737 * Called to check deny when READ with all zero stateid or
1738 * WRITE with all zero or all one stateid
1741 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
1743 struct inode *ino = current_fh->fh_dentry->d_inode;
1744 struct nfs4_file *fp;
1745 struct nfs4_stateid *stp;
1748 dprintk("NFSD: nfs4_share_conflict\n");
1750 fp = find_file(ino);
1753 ret = nfserr_locked;
1754 /* Search for conflicting share reservations */
1755 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
1756 if (test_bit(deny_type, &stp->st_deny_bmap) ||
1757 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
1767 nfs4_file_downgrade(struct file *filp, unsigned int share_access)
1769 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
1770 drop_file_write_access(filp);
1771 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
1776 * Recall a delegation
1779 do_recall(void *__dp)
1781 struct nfs4_delegation *dp = __dp;
1783 dp->dl_file->fi_had_conflict = true;
1784 nfsd4_cb_recall(dp);
1789 * Spawn a thread to perform a recall on the delegation represented
1790 * by the lease (file_lock)
1792 * Called from break_lease() with lock_kernel() held.
1793 * Note: we assume break_lease will only call this *once* for any given
1797 void nfsd_break_deleg_cb(struct file_lock *fl)
1799 struct nfs4_delegation *dp= (struct nfs4_delegation *)fl->fl_owner;
1800 struct task_struct *t;
1802 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
1806 /* We're assuming the state code never drops its reference
1807 * without first removing the lease. Since we're in this lease
1808 * callback (and since the lease code is serialized by the kernel
1809 * lock) we know the server hasn't removed the lease yet, we know
1810 * it's safe to take a reference: */
1811 atomic_inc(&dp->dl_count);
1812 atomic_inc(&dp->dl_client->cl_count);
1814 spin_lock(&recall_lock);
1815 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
1816 spin_unlock(&recall_lock);
1818 /* only place dl_time is set. protected by lock_kernel*/
1819 dp->dl_time = get_seconds();
1822 * We don't want the locks code to timeout the lease for us;
1823 * we'll remove it ourself if the delegation isn't returned
1826 fl->fl_break_time = 0;
1828 t = kthread_run(do_recall, dp, "%s", "nfs4_cb_recall");
1830 struct nfs4_client *clp = dp->dl_client;
1832 printk(KERN_INFO "NFSD: Callback thread failed for "
1833 "for client (clientid %08x/%08x)\n",
1834 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1835 put_nfs4_client(dp->dl_client);
1836 nfs4_put_delegation(dp);
1841 * The file_lock is being reapd.
1843 * Called by locks_free_lock() with lock_kernel() held.
1846 void nfsd_release_deleg_cb(struct file_lock *fl)
1848 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1850 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
1852 if (!(fl->fl_flags & FL_LEASE) || !dp)
1854 dp->dl_flock = NULL;
1858 * Set the delegation file_lock back pointer.
1860 * Called from setlease() with lock_kernel() held.
1863 void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
1865 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
1867 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
1874 * Called from setlease() with lock_kernel() held
1877 int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
1879 struct nfs4_delegation *onlistd =
1880 (struct nfs4_delegation *)onlist->fl_owner;
1881 struct nfs4_delegation *tryd =
1882 (struct nfs4_delegation *)try->fl_owner;
1884 if (onlist->fl_lmops != try->fl_lmops)
1887 return onlistd->dl_client == tryd->dl_client;
1892 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
1895 return lease_modify(onlist, arg);
1900 static struct lock_manager_operations nfsd_lease_mng_ops = {
1901 .fl_break = nfsd_break_deleg_cb,
1902 .fl_release_private = nfsd_release_deleg_cb,
1903 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
1904 .fl_mylease = nfsd_same_client_deleg_cb,
1905 .fl_change = nfsd_change_deleg_cb,
1910 nfsd4_process_open1(struct nfsd4_open *open)
1912 clientid_t *clientid = &open->op_clientid;
1913 struct nfs4_client *clp = NULL;
1914 unsigned int strhashval;
1915 struct nfs4_stateowner *sop = NULL;
1917 if (!check_name(open->op_owner))
1918 return nfserr_inval;
1920 if (STALE_CLIENTID(&open->op_clientid))
1921 return nfserr_stale_clientid;
1923 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
1924 sop = find_openstateowner_str(strhashval, open);
1925 open->op_stateowner = sop;
1927 /* Make sure the client's lease hasn't expired. */
1928 clp = find_confirmed_client(clientid);
1930 return nfserr_expired;
1933 if (!sop->so_confirmed) {
1934 /* Replace unconfirmed owners without checking for replay. */
1935 clp = sop->so_client;
1936 release_openowner(sop);
1937 open->op_stateowner = NULL;
1940 if (open->op_seqid == sop->so_seqid - 1) {
1941 if (sop->so_replay.rp_buflen)
1942 return nfserr_replay_me;
1943 /* The original OPEN failed so spectacularly
1944 * that we don't even have replay data saved!
1945 * Therefore, we have no choice but to continue
1946 * processing this OPEN; presumably, we'll
1947 * fail again for the same reason.
1949 dprintk("nfsd4_process_open1: replay with no replay cache\n");
1952 if (open->op_seqid != sop->so_seqid)
1953 return nfserr_bad_seqid;
1955 if (open->op_stateowner == NULL) {
1956 sop = alloc_init_open_stateowner(strhashval, clp, open);
1958 return nfserr_resource;
1959 open->op_stateowner = sop;
1961 list_del_init(&sop->so_close_lru);
1962 renew_client(sop->so_client);
1966 static inline __be32
1967 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
1969 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
1970 return nfserr_openmode;
1975 static struct nfs4_delegation *
1976 find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
1978 struct nfs4_delegation *dp;
1980 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
1981 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
1988 nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
1989 struct nfs4_delegation **dp)
1992 __be32 status = nfserr_bad_stateid;
1994 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
1997 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
1998 RD_STATE : WR_STATE;
1999 status = nfs4_check_delegmode(*dp, flags);
2003 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2007 open->op_stateowner->so_confirmed = 1;
2012 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2014 struct nfs4_stateid *local;
2015 __be32 status = nfserr_share_denied;
2016 struct nfs4_stateowner *sop = open->op_stateowner;
2018 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2019 /* ignore lock owners */
2020 if (local->st_stateowner->so_is_open_owner == 0)
2022 /* remember if we have seen this open owner */
2023 if (local->st_stateowner == sop)
2025 /* check for conflicting share reservations */
2026 if (!test_share(local, open))
2034 static inline struct nfs4_stateid *
2035 nfs4_alloc_stateid(void)
2037 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2041 nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
2042 struct nfs4_delegation *dp,
2043 struct svc_fh *cur_fh, int flags)
2045 struct nfs4_stateid *stp;
2047 stp = nfs4_alloc_stateid();
2049 return nfserr_resource;
2052 get_file(dp->dl_vfs_file);
2053 stp->st_vfs_file = dp->dl_vfs_file;
2056 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2059 if (status == nfserr_dropit)
2060 status = nfserr_jukebox;
2061 kmem_cache_free(stateid_slab, stp);
2069 static inline __be32
2070 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2071 struct nfsd4_open *open)
2073 struct iattr iattr = {
2074 .ia_valid = ATTR_SIZE,
2077 if (!open->op_truncate)
2079 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2080 return nfserr_inval;
2081 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2085 nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2087 struct file *filp = stp->st_vfs_file;
2088 struct inode *inode = filp->f_path.dentry->d_inode;
2089 unsigned int share_access, new_writer;
2092 set_access(&share_access, stp->st_access_bmap);
2093 new_writer = (~share_access) & open->op_share_access
2094 & NFS4_SHARE_ACCESS_WRITE;
2097 int err = get_write_access(inode);
2099 return nfserrno(err);
2100 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2102 return nfserrno(err);
2103 file_take_write(filp);
2105 status = nfsd4_truncate(rqstp, cur_fh, open);
2108 put_write_access(inode);
2111 /* remember the open */
2112 filp->f_mode |= open->op_share_access;
2113 __set_bit(open->op_share_access, &stp->st_access_bmap);
2114 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
2121 nfs4_set_claim_prev(struct nfsd4_open *open)
2123 open->op_stateowner->so_confirmed = 1;
2124 open->op_stateowner->so_client->cl_firststate = 1;
2128 * Attempt to hand out a delegation.
2131 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2133 struct nfs4_delegation *dp;
2134 struct nfs4_stateowner *sop = stp->st_stateowner;
2135 struct nfs4_callback *cb = &sop->so_client->cl_callback;
2136 struct file_lock fl, *flp = &fl;
2137 int status, flag = 0;
2139 flag = NFS4_OPEN_DELEGATE_NONE;
2140 open->op_recall = 0;
2141 switch (open->op_claim_type) {
2142 case NFS4_OPEN_CLAIM_PREVIOUS:
2143 if (!atomic_read(&cb->cb_set))
2144 open->op_recall = 1;
2145 flag = open->op_delegate_type;
2146 if (flag == NFS4_OPEN_DELEGATE_NONE)
2149 case NFS4_OPEN_CLAIM_NULL:
2150 /* Let's not give out any delegations till everyone's
2151 * had the chance to reclaim theirs.... */
2152 if (locks_in_grace())
2154 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
2156 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2157 flag = NFS4_OPEN_DELEGATE_WRITE;
2159 flag = NFS4_OPEN_DELEGATE_READ;
2165 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2167 flag = NFS4_OPEN_DELEGATE_NONE;
2170 locks_init_lock(&fl);
2171 fl.fl_lmops = &nfsd_lease_mng_ops;
2172 fl.fl_flags = FL_LEASE;
2173 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2174 fl.fl_end = OFFSET_MAX;
2175 fl.fl_owner = (fl_owner_t)dp;
2176 fl.fl_file = stp->st_vfs_file;
2177 fl.fl_pid = current->tgid;
2179 /* vfs_setlease checks to see if delegation should be handed out.
2180 * the lock_manager callbacks fl_mylease and fl_change are used
2182 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
2183 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
2184 unhash_delegation(dp);
2185 flag = NFS4_OPEN_DELEGATE_NONE;
2189 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2191 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
2192 dp->dl_stateid.si_boot,
2193 dp->dl_stateid.si_stateownerid,
2194 dp->dl_stateid.si_fileid,
2195 dp->dl_stateid.si_generation);
2197 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2198 && flag == NFS4_OPEN_DELEGATE_NONE
2199 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2200 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2201 open->op_delegate_type = flag;
2205 * called with nfs4_lock_state() held.
2208 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2210 struct nfs4_file *fp = NULL;
2211 struct inode *ino = current_fh->fh_dentry->d_inode;
2212 struct nfs4_stateid *stp = NULL;
2213 struct nfs4_delegation *dp = NULL;
2216 status = nfserr_inval;
2217 if (!access_valid(open->op_share_access)
2218 || !deny_valid(open->op_share_deny))
2221 * Lookup file; if found, lookup stateid and check open request,
2222 * and check for delegations in the process of being recalled.
2223 * If not found, create the nfs4_file struct
2225 fp = find_file(ino);
2227 if ((status = nfs4_check_open(fp, open, &stp)))
2229 status = nfs4_check_deleg(fp, open, &dp);
2233 status = nfserr_bad_stateid;
2234 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2236 status = nfserr_resource;
2237 fp = alloc_init_file(ino);
2243 * OPEN the file, or upgrade an existing OPEN.
2244 * If truncate fails, the OPEN fails.
2247 /* Stateid was found, this is an OPEN upgrade */
2248 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2251 update_stateid(&stp->st_stateid);
2253 /* Stateid was not found, this is a new OPEN */
2255 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
2256 flags |= NFSD_MAY_READ;
2257 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2258 flags |= NFSD_MAY_WRITE;
2259 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2262 init_stateid(stp, fp, open);
2263 status = nfsd4_truncate(rqstp, current_fh, open);
2265 release_open_stateid(stp);
2269 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2272 * Attempt to hand out a delegation. No error return, because the
2273 * OPEN succeeds even if we fail.
2275 nfs4_open_delegation(current_fh, open, stp);
2279 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
2280 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
2281 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
2285 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2286 nfs4_set_claim_prev(open);
2288 * To finish the open response, we just need to set the rflags.
2290 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
2291 if (!open->op_stateowner->so_confirmed)
2292 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2298 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2301 struct nfs4_client *clp;
2305 dprintk("process_renew(%08x/%08x): starting\n",
2306 clid->cl_boot, clid->cl_id);
2307 status = nfserr_stale_clientid;
2308 if (STALE_CLIENTID(clid))
2310 clp = find_confirmed_client(clid);
2311 status = nfserr_expired;
2313 /* We assume the client took too long to RENEW. */
2314 dprintk("nfsd4_renew: clientid not found!\n");
2318 status = nfserr_cb_path_down;
2319 if (!list_empty(&clp->cl_delegations)
2320 && !atomic_read(&clp->cl_callback.cb_set))
2324 nfs4_unlock_state();
2328 struct lock_manager nfsd4_manager = {
2332 nfsd4_end_grace(void)
2334 dprintk("NFSD: end of grace period\n");
2335 nfsd4_recdir_purge_old();
2336 locks_end_grace(&nfsd4_manager);
2340 nfs4_laundromat(void)
2342 struct nfs4_client *clp;
2343 struct nfs4_stateowner *sop;
2344 struct nfs4_delegation *dp;
2345 struct list_head *pos, *next, reaplist;
2346 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
2347 time_t t, clientid_val = NFSD_LEASE_TIME;
2348 time_t u, test_val = NFSD_LEASE_TIME;
2352 dprintk("NFSD: laundromat service - starting\n");
2353 if (locks_in_grace())
2355 list_for_each_safe(pos, next, &client_lru) {
2356 clp = list_entry(pos, struct nfs4_client, cl_lru);
2357 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2358 t = clp->cl_time - cutoff;
2359 if (clientid_val > t)
2363 dprintk("NFSD: purging unused client (clientid %08x)\n",
2364 clp->cl_clientid.cl_id);
2365 nfsd4_remove_clid_dir(clp);
2368 INIT_LIST_HEAD(&reaplist);
2369 spin_lock(&recall_lock);
2370 list_for_each_safe(pos, next, &del_recall_lru) {
2371 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2372 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2373 u = dp->dl_time - cutoff;
2378 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2380 list_move(&dp->dl_recall_lru, &reaplist);
2382 spin_unlock(&recall_lock);
2383 list_for_each_safe(pos, next, &reaplist) {
2384 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2385 list_del_init(&dp->dl_recall_lru);
2386 unhash_delegation(dp);
2388 test_val = NFSD_LEASE_TIME;
2389 list_for_each_safe(pos, next, &close_lru) {
2390 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2391 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2392 u = sop->so_time - cutoff;
2397 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2399 release_openowner(sop);
2401 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2402 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2403 nfs4_unlock_state();
2404 return clientid_val;
2407 static struct workqueue_struct *laundry_wq;
2408 static void laundromat_main(struct work_struct *);
2409 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2412 laundromat_main(struct work_struct *not_used)
2416 t = nfs4_laundromat();
2417 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
2418 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
2421 static struct nfs4_stateowner *
2422 search_close_lru(u32 st_id, int flags)
2424 struct nfs4_stateowner *local = NULL;
2426 if (flags & CLOSE_STATE) {
2427 list_for_each_entry(local, &close_lru, so_close_lru) {
2428 if (local->so_id == st_id)
2436 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2438 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
2442 STALE_STATEID(stateid_t *stateid)
2444 if (stateid->si_boot == boot_time)
2446 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
2447 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2448 stateid->si_generation);
2453 access_permit_read(unsigned long access_bmap)
2455 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2456 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2457 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2461 access_permit_write(unsigned long access_bmap)
2463 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2464 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2468 __be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
2470 __be32 status = nfserr_openmode;
2472 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2474 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2481 static inline __be32
2482 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2484 if (ONE_STATEID(stateid) && (flags & RD_STATE))
2486 else if (locks_in_grace()) {
2487 /* Answer in remaining cases depends on existance of
2488 * conflicting state; so we must wait out the grace period. */
2489 return nfserr_grace;
2490 } else if (flags & WR_STATE)
2491 return nfs4_share_conflict(current_fh,
2492 NFS4_SHARE_DENY_WRITE);
2493 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2494 return nfs4_share_conflict(current_fh,
2495 NFS4_SHARE_DENY_READ);
2499 * Allow READ/WRITE during grace period on recovered state only for files
2500 * that are not able to provide mandatory locking.
2503 grace_disallows_io(struct inode *inode)
2505 return locks_in_grace() && mandatory_lock(inode);
2508 static int check_stateid_generation(stateid_t *in, stateid_t *ref)
2510 /* If the client sends us a stateid from the future, it's buggy: */
2511 if (in->si_generation > ref->si_generation)
2512 return nfserr_bad_stateid;
2514 * The following, however, can happen. For example, if the
2515 * client sends an open and some IO at the same time, the open
2516 * may bump si_generation while the IO is still in flight.
2517 * Thanks to hard links and renames, the client never knows what
2518 * file an open will affect. So it could avoid that situation
2519 * only by serializing all opens and IO from the same open
2520 * owner. To recover from the old_stateid error, the client
2521 * will just have to retry the IO:
2523 if (in->si_generation < ref->si_generation)
2524 return nfserr_old_stateid;
2528 static int is_delegation_stateid(stateid_t *stateid)
2530 return stateid->si_fileid == 0;
2534 * Checks for stateid operations
2537 nfs4_preprocess_stateid_op(struct svc_fh *current_fh, stateid_t *stateid, int flags, struct file **filpp)
2539 struct nfs4_stateid *stp = NULL;
2540 struct nfs4_delegation *dp = NULL;
2541 struct inode *ino = current_fh->fh_dentry->d_inode;
2547 if (grace_disallows_io(ino))
2548 return nfserr_grace;
2550 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2551 return check_special_stateids(current_fh, stateid, flags);
2553 status = nfserr_stale_stateid;
2554 if (STALE_STATEID(stateid))
2557 status = nfserr_bad_stateid;
2558 if (is_delegation_stateid(stateid)) {
2559 dp = find_delegation_stateid(ino, stateid);
2562 status = check_stateid_generation(stateid, &dp->dl_stateid);
2565 status = nfs4_check_delegmode(dp, flags);
2568 renew_client(dp->dl_client);
2570 *filpp = dp->dl_vfs_file;
2571 } else { /* open or lock stateid */
2572 stp = find_stateid(stateid, flags);
2575 if (nfs4_check_fh(current_fh, stp))
2577 if (!stp->st_stateowner->so_confirmed)
2579 status = check_stateid_generation(stateid, &stp->st_stateid);
2582 status = nfs4_check_openmode(stp, flags);
2585 renew_client(stp->st_stateowner->so_client);
2587 *filpp = stp->st_vfs_file;
2597 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2598 RD_STATE : WR_STATE;
2602 * Checks for sequence id mutating operations.
2605 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)
2607 struct nfs4_stateid *stp;
2608 struct nfs4_stateowner *sop;
2611 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2612 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2613 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2614 stateid->si_generation);
2619 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2620 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
2621 return nfserr_bad_stateid;
2624 if (STALE_STATEID(stateid))
2625 return nfserr_stale_stateid;
2627 * We return BAD_STATEID if filehandle doesn't match stateid,
2628 * the confirmed flag is incorrecly set, or the generation
2629 * number is incorrect.
2631 stp = find_stateid(stateid, flags);
2634 * Also, we should make sure this isn't just the result of
2637 sop = search_close_lru(stateid->si_stateownerid, flags);
2639 return nfserr_bad_stateid;
2645 *sopp = sop = stp->st_stateowner;
2648 clientid_t *lockclid = &lock->v.new.clientid;
2649 struct nfs4_client *clp = sop->so_client;
2653 lkflg = setlkflg(lock->lk_type);
2655 if (lock->lk_is_new) {
2656 if (!sop->so_is_open_owner)
2657 return nfserr_bad_stateid;
2658 if (!same_clid(&clp->cl_clientid, lockclid))
2659 return nfserr_bad_stateid;
2660 /* stp is the open stateid */
2661 status = nfs4_check_openmode(stp, lkflg);
2665 /* stp is the lock stateid */
2666 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2672 if (nfs4_check_fh(current_fh, stp)) {
2673 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
2674 return nfserr_bad_stateid;
2678 * We now validate the seqid and stateid generation numbers.
2679 * For the moment, we ignore the possibility of
2680 * generation number wraparound.
2682 if (seqid != sop->so_seqid)
2685 if (sop->so_confirmed && flags & CONFIRM) {
2686 dprintk("NFSD: preprocess_seqid_op: expected"
2687 " unconfirmed stateowner!\n");
2688 return nfserr_bad_stateid;
2690 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2691 dprintk("NFSD: preprocess_seqid_op: stateowner not"
2692 " confirmed yet!\n");
2693 return nfserr_bad_stateid;
2695 status = check_stateid_generation(stateid, &stp->st_stateid);
2698 renew_client(sop->so_client);
2702 if (seqid == sop->so_seqid - 1) {
2703 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
2704 /* indicate replay to calling function */
2705 return nfserr_replay_me;
2707 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
2708 sop->so_seqid, seqid);
2710 return nfserr_bad_seqid;
2714 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2715 struct nfsd4_open_confirm *oc)
2718 struct nfs4_stateowner *sop;
2719 struct nfs4_stateid *stp;
2721 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
2722 (int)cstate->current_fh.fh_dentry->d_name.len,
2723 cstate->current_fh.fh_dentry->d_name.name);
2725 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
2731 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2732 oc->oc_seqid, &oc->oc_req_stateid,
2733 CONFIRM | OPEN_STATE,
2734 &oc->oc_stateowner, &stp, NULL)))
2737 sop = oc->oc_stateowner;
2738 sop->so_confirmed = 1;
2739 update_stateid(&stp->st_stateid);
2740 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
2741 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
2742 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
2743 stp->st_stateid.si_boot,
2744 stp->st_stateid.si_stateownerid,
2745 stp->st_stateid.si_fileid,
2746 stp->st_stateid.si_generation);
2748 nfsd4_create_clid_dir(sop->so_client);
2750 if (oc->oc_stateowner) {
2751 nfs4_get_stateowner(oc->oc_stateowner);
2752 cstate->replay_owner = oc->oc_stateowner;
2754 nfs4_unlock_state();
2760 * unset all bits in union bitmap (bmap) that
2761 * do not exist in share (from successful OPEN_DOWNGRADE)
2764 reset_union_bmap_access(unsigned long access, unsigned long *bmap)
2767 for (i = 1; i < 4; i++) {
2768 if ((i & access) != i)
2769 __clear_bit(i, bmap);
2774 reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
2777 for (i = 0; i < 4; i++) {
2778 if ((i & deny) != i)
2779 __clear_bit(i, bmap);
2784 nfsd4_open_downgrade(struct svc_rqst *rqstp,
2785 struct nfsd4_compound_state *cstate,
2786 struct nfsd4_open_downgrade *od)
2789 struct nfs4_stateid *stp;
2790 unsigned int share_access;
2792 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
2793 (int)cstate->current_fh.fh_dentry->d_name.len,
2794 cstate->current_fh.fh_dentry->d_name.name);
2796 if (!access_valid(od->od_share_access)
2797 || !deny_valid(od->od_share_deny))
2798 return nfserr_inval;
2801 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2805 &od->od_stateowner, &stp, NULL)))
2808 status = nfserr_inval;
2809 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
2810 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
2811 stp->st_access_bmap, od->od_share_access);
2814 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
2815 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
2816 stp->st_deny_bmap, od->od_share_deny);
2819 set_access(&share_access, stp->st_access_bmap);
2820 nfs4_file_downgrade(stp->st_vfs_file,
2821 share_access & ~od->od_share_access);
2823 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
2824 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
2826 update_stateid(&stp->st_stateid);
2827 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
2830 if (od->od_stateowner) {
2831 nfs4_get_stateowner(od->od_stateowner);
2832 cstate->replay_owner = od->od_stateowner;
2834 nfs4_unlock_state();
2839 * nfs4_unlock_state() called after encode
2842 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2843 struct nfsd4_close *close)
2846 struct nfs4_stateid *stp;
2848 dprintk("NFSD: nfsd4_close on file %.*s\n",
2849 (int)cstate->current_fh.fh_dentry->d_name.len,
2850 cstate->current_fh.fh_dentry->d_name.name);
2853 /* check close_lru for replay */
2854 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2857 OPEN_STATE | CLOSE_STATE,
2858 &close->cl_stateowner, &stp, NULL)))
2861 update_stateid(&stp->st_stateid);
2862 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
2864 /* release_stateid() calls nfsd_close() if needed */
2865 release_open_stateid(stp);
2867 /* place unused nfs4_stateowners on so_close_lru list to be
2868 * released by the laundromat service after the lease period
2869 * to enable us to handle CLOSE replay
2871 if (list_empty(&close->cl_stateowner->so_stateids))
2872 move_to_close_lru(close->cl_stateowner);
2874 if (close->cl_stateowner) {
2875 nfs4_get_stateowner(close->cl_stateowner);
2876 cstate->replay_owner = close->cl_stateowner;
2878 nfs4_unlock_state();
2883 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2884 struct nfsd4_delegreturn *dr)
2886 struct nfs4_delegation *dp;
2887 stateid_t *stateid = &dr->dr_stateid;
2888 struct inode *inode;
2891 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
2893 inode = cstate->current_fh.fh_dentry->d_inode;
2896 status = nfserr_bad_stateid;
2897 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2899 status = nfserr_stale_stateid;
2900 if (STALE_STATEID(stateid))
2902 status = nfserr_bad_stateid;
2903 if (!is_delegation_stateid(stateid))
2905 dp = find_delegation_stateid(inode, stateid);
2908 status = check_stateid_generation(stateid, &dp->dl_stateid);
2911 renew_client(dp->dl_client);
2913 unhash_delegation(dp);
2915 nfs4_unlock_state();
2922 * Lock owner state (byte-range locks)
2924 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
2925 #define LOCK_HASH_BITS 8
2926 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
2927 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
2930 end_offset(u64 start, u64 len)
2935 return end >= start ? end: NFS4_MAX_UINT64;
2938 /* last octet in a range */
2940 last_byte_offset(u64 start, u64 len)
2946 return end > start ? end - 1: NFS4_MAX_UINT64;
2949 #define lockownerid_hashval(id) \
2950 ((id) & LOCK_HASH_MASK)
2952 static inline unsigned int
2953 lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
2954 struct xdr_netobj *ownername)
2956 return (file_hashval(inode) + cl_id
2957 + opaque_hashval(ownername->data, ownername->len))
2961 static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
2962 static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
2963 static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
2965 static struct nfs4_stateid *
2966 find_stateid(stateid_t *stid, int flags)
2968 struct nfs4_stateid *local;
2969 u32 st_id = stid->si_stateownerid;
2970 u32 f_id = stid->si_fileid;
2971 unsigned int hashval;
2973 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
2974 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
2975 hashval = stateid_hashval(st_id, f_id);
2976 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
2977 if ((local->st_stateid.si_stateownerid == st_id) &&
2978 (local->st_stateid.si_fileid == f_id))
2983 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
2984 hashval = stateid_hashval(st_id, f_id);
2985 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
2986 if ((local->st_stateid.si_stateownerid == st_id) &&
2987 (local->st_stateid.si_fileid == f_id))
2994 static struct nfs4_delegation *
2995 find_delegation_stateid(struct inode *ino, stateid_t *stid)
2997 struct nfs4_file *fp;
2998 struct nfs4_delegation *dl;
3000 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
3001 stid->si_boot, stid->si_stateownerid,
3002 stid->si_fileid, stid->si_generation);
3004 fp = find_file(ino);
3007 dl = find_delegation_file(fp, stid);
3013 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3014 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3015 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3016 * locking, this prevents us from being completely protocol-compliant. The
3017 * real solution to this problem is to start using unsigned file offsets in
3018 * the VFS, but this is a very deep change!
3021 nfs4_transform_lock_offset(struct file_lock *lock)
3023 if (lock->fl_start < 0)
3024 lock->fl_start = OFFSET_MAX;
3025 if (lock->fl_end < 0)
3026 lock->fl_end = OFFSET_MAX;
3029 /* Hack!: For now, we're defining this just so we can use a pointer to it
3030 * as a unique cookie to identify our (NFSv4's) posix locks. */
3031 static struct lock_manager_operations nfsd_posix_mng_ops = {
3035 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3037 struct nfs4_stateowner *sop;
3040 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3041 sop = (struct nfs4_stateowner *) fl->fl_owner;
3042 hval = lockownerid_hashval(sop->so_id);
3043 kref_get(&sop->so_ref);
3045 deny->ld_clientid = sop->so_client->cl_clientid;
3047 deny->ld_sop = NULL;
3048 deny->ld_clientid.cl_boot = 0;
3049 deny->ld_clientid.cl_id = 0;
3051 deny->ld_start = fl->fl_start;
3052 deny->ld_length = NFS4_MAX_UINT64;
3053 if (fl->fl_end != NFS4_MAX_UINT64)
3054 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3055 deny->ld_type = NFS4_READ_LT;
3056 if (fl->fl_type != F_RDLCK)
3057 deny->ld_type = NFS4_WRITE_LT;
3060 static struct nfs4_stateowner *
3061 find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3062 struct xdr_netobj *owner)
3064 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3065 struct nfs4_stateowner *op;
3067 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
3068 if (same_owner_str(op, owner, clid))
3075 * Alloc a lock owner structure.
3076 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3079 * strhashval = lock_ownerstr_hashval
3082 static struct nfs4_stateowner *
3083 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3084 struct nfs4_stateowner *sop;
3085 struct nfs4_replay *rp;
3086 unsigned int idhashval;
3088 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3090 idhashval = lockownerid_hashval(current_ownerid);
3091 INIT_LIST_HEAD(&sop->so_idhash);
3092 INIT_LIST_HEAD(&sop->so_strhash);
3093 INIT_LIST_HEAD(&sop->so_perclient);
3094 INIT_LIST_HEAD(&sop->so_stateids);
3095 INIT_LIST_HEAD(&sop->so_perstateid);
3096 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3098 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3099 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
3100 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
3101 sop->so_is_open_owner = 0;
3102 sop->so_id = current_ownerid++;
3103 sop->so_client = clp;
3104 /* It is the openowner seqid that will be incremented in encode in the
3105 * case of new lockowners; so increment the lock seqid manually: */
3106 sop->so_seqid = lock->lk_new_lock_seqid + 1;
3107 sop->so_confirmed = 1;
3108 rp = &sop->so_replay;
3109 rp->rp_status = nfserr_serverfault;
3111 rp->rp_buf = rp->rp_ibuf;
3115 static struct nfs4_stateid *
3116 alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3118 struct nfs4_stateid *stp;
3119 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3121 stp = nfs4_alloc_stateid();
3124 INIT_LIST_HEAD(&stp->st_hash);
3125 INIT_LIST_HEAD(&stp->st_perfile);
3126 INIT_LIST_HEAD(&stp->st_perstateowner);
3127 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
3128 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
3129 list_add(&stp->st_perfile, &fp->fi_stateids);
3130 list_add(&stp->st_perstateowner, &sop->so_stateids);
3131 stp->st_stateowner = sop;
3134 stp->st_stateid.si_boot = boot_time;
3135 stp->st_stateid.si_stateownerid = sop->so_id;
3136 stp->st_stateid.si_fileid = fp->fi_id;
3137 stp->st_stateid.si_generation = 0;
3138 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3139 stp->st_access_bmap = open_stp->st_access_bmap;
3140 stp->st_deny_bmap = open_stp->st_deny_bmap;
3141 stp->st_openstp = open_stp;
3148 check_lock_length(u64 offset, u64 length)
3150 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
3151 LOFF_OVERFLOW(offset, length)));
3158 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3159 struct nfsd4_lock *lock)
3161 struct nfs4_stateowner *open_sop = NULL;
3162 struct nfs4_stateowner *lock_sop = NULL;
3163 struct nfs4_stateid *lock_stp;
3165 struct file_lock file_lock;
3166 struct file_lock conflock;
3168 unsigned int strhashval;
3172 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3173 (long long) lock->lk_offset,
3174 (long long) lock->lk_length);
3176 if (check_lock_length(lock->lk_offset, lock->lk_length))
3177 return nfserr_inval;
3179 if ((status = fh_verify(rqstp, &cstate->current_fh,
3180 S_IFREG, NFSD_MAY_LOCK))) {
3181 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3187 if (lock->lk_is_new) {
3189 * Client indicates that this is a new lockowner.
3190 * Use open owner and open stateid to create lock owner and
3193 struct nfs4_stateid *open_stp = NULL;
3194 struct nfs4_file *fp;
3196 status = nfserr_stale_clientid;
3197 if (STALE_CLIENTID(&lock->lk_new_clientid))
3200 /* validate and update open stateid and open seqid */
3201 status = nfs4_preprocess_seqid_op(&cstate->current_fh,
3202 lock->lk_new_open_seqid,
3203 &lock->lk_new_open_stateid,
3205 &lock->lk_replay_owner, &open_stp,
3209 open_sop = lock->lk_replay_owner;
3210 /* create lockowner and lock stateid */
3211 fp = open_stp->st_file;
3212 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3213 open_sop->so_client->cl_clientid.cl_id,
3214 &lock->v.new.owner);
3215 /* XXX: Do we need to check for duplicate stateowners on
3216 * the same file, or should they just be allowed (and
3217 * create new stateids)? */
3218 status = nfserr_resource;
3219 lock_sop = alloc_init_lock_stateowner(strhashval,
3220 open_sop->so_client, open_stp, lock);
3221 if (lock_sop == NULL)
3223 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
3224 if (lock_stp == NULL)
3227 /* lock (lock owner + lock stateid) already exists */
3228 status = nfs4_preprocess_seqid_op(&cstate->current_fh,
3229 lock->lk_old_lock_seqid,
3230 &lock->lk_old_lock_stateid,
3232 &lock->lk_replay_owner, &lock_stp, lock);
3235 lock_sop = lock->lk_replay_owner;
3237 /* lock->lk_replay_owner and lock_stp have been created or found */
3238 filp = lock_stp->st_vfs_file;
3240 status = nfserr_grace;
3241 if (locks_in_grace() && !lock->lk_reclaim)
3243 status = nfserr_no_grace;
3244 if (!locks_in_grace() && lock->lk_reclaim)
3247 locks_init_lock(&file_lock);
3248 switch (lock->lk_type) {
3251 file_lock.fl_type = F_RDLCK;
3255 case NFS4_WRITEW_LT:
3256 file_lock.fl_type = F_WRLCK;
3260 status = nfserr_inval;
3263 file_lock.fl_owner = (fl_owner_t)lock_sop;
3264 file_lock.fl_pid = current->tgid;
3265 file_lock.fl_file = filp;
3266 file_lock.fl_flags = FL_POSIX;
3267 file_lock.fl_lmops = &nfsd_posix_mng_ops;
3269 file_lock.fl_start = lock->lk_offset;
3270 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
3271 nfs4_transform_lock_offset(&file_lock);
3274 * Try to lock the file in the VFS.
3275 * Note: locks.c uses the BKL to protect the inode's lock list.
3278 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
3280 case 0: /* success! */
3281 update_stateid(&lock_stp->st_stateid);
3282 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3286 case (EAGAIN): /* conflock holds conflicting lock */
3287 status = nfserr_denied;
3288 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3289 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3292 status = nfserr_deadlock;
3295 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
3296 status = nfserr_resource;
3300 if (status && lock->lk_is_new && lock_sop)
3301 release_lockowner(lock_sop);
3302 if (lock->lk_replay_owner) {
3303 nfs4_get_stateowner(lock->lk_replay_owner);
3304 cstate->replay_owner = lock->lk_replay_owner;
3306 nfs4_unlock_state();
3311 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3312 * so we do a temporary open here just to get an open file to pass to
3313 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3316 static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3321 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3324 err = vfs_test_lock(file, lock);
3333 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3334 struct nfsd4_lockt *lockt)
3336 struct inode *inode;
3337 struct file_lock file_lock;
3341 if (locks_in_grace())
3342 return nfserr_grace;
3344 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3345 return nfserr_inval;
3347 lockt->lt_stateowner = NULL;
3350 status = nfserr_stale_clientid;
3351 if (STALE_CLIENTID(&lockt->lt_clientid))
3354 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
3355 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
3356 if (status == nfserr_symlink)
3357 status = nfserr_inval;
3361 inode = cstate->current_fh.fh_dentry->d_inode;
3362 locks_init_lock(&file_lock);
3363 switch (lockt->lt_type) {
3366 file_lock.fl_type = F_RDLCK;
3369 case NFS4_WRITEW_LT:
3370 file_lock.fl_type = F_WRLCK;
3373 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
3374 status = nfserr_inval;
3378 lockt->lt_stateowner = find_lockstateowner_str(inode,
3379 &lockt->lt_clientid, &lockt->lt_owner);
3380 if (lockt->lt_stateowner)
3381 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3382 file_lock.fl_pid = current->tgid;
3383 file_lock.fl_flags = FL_POSIX;
3385 file_lock.fl_start = lockt->lt_offset;
3386 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
3388 nfs4_transform_lock_offset(&file_lock);
3391 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
3393 status = nfserrno(error);
3396 if (file_lock.fl_type != F_UNLCK) {
3397 status = nfserr_denied;
3398 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
3401 nfs4_unlock_state();
3406 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3407 struct nfsd4_locku *locku)
3409 struct nfs4_stateid *stp;
3410 struct file *filp = NULL;
3411 struct file_lock file_lock;
3415 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3416 (long long) locku->lu_offset,
3417 (long long) locku->lu_length);
3419 if (check_lock_length(locku->lu_offset, locku->lu_length))
3420 return nfserr_inval;
3424 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
3428 &locku->lu_stateowner, &stp, NULL)))
3431 filp = stp->st_vfs_file;
3433 locks_init_lock(&file_lock);
3434 file_lock.fl_type = F_UNLCK;
3435 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3436 file_lock.fl_pid = current->tgid;
3437 file_lock.fl_file = filp;
3438 file_lock.fl_flags = FL_POSIX;
3439 file_lock.fl_lmops = &nfsd_posix_mng_ops;
3440 file_lock.fl_start = locku->lu_offset;
3442 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
3443 nfs4_transform_lock_offset(&file_lock);
3446 * Try to unlock the file in the VFS.
3448 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
3450 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
3454 * OK, unlock succeeded; the only thing left to do is update the stateid.
3456 update_stateid(&stp->st_stateid);
3457 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3460 if (locku->lu_stateowner) {
3461 nfs4_get_stateowner(locku->lu_stateowner);
3462 cstate->replay_owner = locku->lu_stateowner;
3464 nfs4_unlock_state();
3468 status = nfserrno(err);
3474 * 1: locks held by lockowner
3475 * 0: no locks held by lockowner
3478 check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3480 struct file_lock **flpp;
3481 struct inode *inode = filp->f_path.dentry->d_inode;
3485 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
3486 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
3497 nfsd4_release_lockowner(struct svc_rqst *rqstp,
3498 struct nfsd4_compound_state *cstate,
3499 struct nfsd4_release_lockowner *rlockowner)
3501 clientid_t *clid = &rlockowner->rl_clientid;
3502 struct nfs4_stateowner *sop;
3503 struct nfs4_stateid *stp;
3504 struct xdr_netobj *owner = &rlockowner->rl_owner;
3505 struct list_head matches;
3509 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3510 clid->cl_boot, clid->cl_id);
3512 /* XXX check for lease expiration */
3514 status = nfserr_stale_clientid;
3515 if (STALE_CLIENTID(clid))
3520 status = nfserr_locks_held;
3521 /* XXX: we're doing a linear search through all the lockowners.
3522 * Yipes! For now we'll just hope clients aren't really using
3523 * release_lockowner much, but eventually we have to fix these
3524 * data structures. */
3525 INIT_LIST_HEAD(&matches);
3526 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3527 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
3528 if (!same_owner_str(sop, owner, clid))
3530 list_for_each_entry(stp, &sop->so_stateids,
3532 if (check_for_locks(stp->st_vfs_file, sop))
3534 /* Note: so_perclient unused for lockowners,
3535 * so it's OK to fool with here. */
3536 list_add(&sop->so_perclient, &matches);
3540 /* Clients probably won't expect us to return with some (but not all)
3541 * of the lockowner state released; so don't release any until all
3542 * have been checked. */
3544 while (!list_empty(&matches)) {
3545 sop = list_entry(matches.next, struct nfs4_stateowner,
3547 /* unhash_stateowner deletes so_perclient only
3548 * for openowners. */
3549 list_del(&sop->so_perclient);
3550 release_lockowner(sop);
3553 nfs4_unlock_state();
3557 static inline struct nfs4_client_reclaim *
3560 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
3564 nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
3566 unsigned int strhashval = clientstr_hashval(name);
3567 struct nfs4_client *clp;
3569 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
3574 * failure => all reset bets are off, nfserr_no_grace...
3577 nfs4_client_to_reclaim(const char *name)
3579 unsigned int strhashval;
3580 struct nfs4_client_reclaim *crp = NULL;
3582 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3583 crp = alloc_reclaim();
3586 strhashval = clientstr_hashval(name);
3587 INIT_LIST_HEAD(&crp->cr_strhash);
3588 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
3589 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
3590 reclaim_str_hashtbl_size++;
3595 nfs4_release_reclaim(void)
3597 struct nfs4_client_reclaim *crp = NULL;
3600 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3601 while (!list_empty(&reclaim_str_hashtbl[i])) {
3602 crp = list_entry(reclaim_str_hashtbl[i].next,
3603 struct nfs4_client_reclaim, cr_strhash);
3604 list_del(&crp->cr_strhash);
3606 reclaim_str_hashtbl_size--;
3609 BUG_ON(reclaim_str_hashtbl_size);
3613 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
3614 static struct nfs4_client_reclaim *
3615 nfs4_find_reclaim_client(clientid_t *clid)
3617 unsigned int strhashval;
3618 struct nfs4_client *clp;
3619 struct nfs4_client_reclaim *crp = NULL;
3622 /* find clientid in conf_id_hashtbl */
3623 clp = find_confirmed_client(clid);
3627 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3628 clp->cl_name.len, clp->cl_name.data,
3631 /* find clp->cl_name in reclaim_str_hashtbl */
3632 strhashval = clientstr_hashval(clp->cl_recdir);
3633 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
3634 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
3642 * Called from OPEN. Look for clientid in reclaim list.
3645 nfs4_check_open_reclaim(clientid_t *clid)
3647 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
3650 /* initialization to perform at module load time: */
3653 nfs4_state_init(void)
3657 status = nfsd4_init_slabs();
3660 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3661 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3662 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3663 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3664 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
3666 for (i = 0; i < SESSION_HASH_SIZE; i++)
3667 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
3668 for (i = 0; i < FILE_HASH_SIZE; i++) {
3669 INIT_LIST_HEAD(&file_hashtbl[i]);
3671 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3672 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3673 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3675 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3676 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3677 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3679 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3680 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3681 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3683 memset(&onestateid, ~0, sizeof(stateid_t));
3684 INIT_LIST_HEAD(&close_lru);
3685 INIT_LIST_HEAD(&client_lru);
3686 INIT_LIST_HEAD(&del_recall_lru);
3687 for (i = 0; i < CLIENT_HASH_SIZE; i++)
3688 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
3689 reclaim_str_hashtbl_size = 0;
3694 nfsd4_load_reboot_recovery_data(void)
3699 nfsd4_init_recdir(user_recovery_dirname);
3700 status = nfsd4_recdir_load();
3701 nfs4_unlock_state();
3703 printk("NFSD: Failure reading reboot recovery data\n");
3707 get_nfs4_grace_period(void)
3709 return max(user_lease_time, lease_time) * HZ;
3713 * Since the lifetime of a delegation isn't limited to that of an open, a
3714 * client may quite reasonably hang on to a delegation as long as it has
3715 * the inode cached. This becomes an obvious problem the first time a
3716 * client's inode cache approaches the size of the server's total memory.
3718 * For now we avoid this problem by imposing a hard limit on the number
3719 * of delegations, which varies according to the server's memory size.
3722 set_max_delegations(void)
3725 * Allow at most 4 delegations per megabyte of RAM. Quick
3726 * estimates suggest that in the worst case (where every delegation
3727 * is for a different inode), a delegation could take about 1.5K,
3728 * giving a worst case usage of about 6% of memory.
3730 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
3733 /* initialization to perform when the nfsd service is started: */
3736 __nfs4_state_start(void)
3738 unsigned long grace_time;
3740 boot_time = get_seconds();
3741 grace_time = get_nfs4_grace_period();
3742 lease_time = user_lease_time;
3743 locks_start_grace(&nfsd4_manager);
3744 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
3746 laundry_wq = create_singlethread_workqueue("nfsd4");
3747 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
3748 set_max_delegations();
3752 nfs4_state_start(void)
3756 nfsd4_load_reboot_recovery_data();
3757 __nfs4_state_start();
3763 nfs4_lease_time(void)
3769 __nfs4_state_shutdown(void)
3772 struct nfs4_client *clp = NULL;
3773 struct nfs4_delegation *dp = NULL;
3774 struct list_head *pos, *next, reaplist;
3776 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3777 while (!list_empty(&conf_id_hashtbl[i])) {
3778 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
3781 while (!list_empty(&unconf_str_hashtbl[i])) {
3782 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
3786 INIT_LIST_HEAD(&reaplist);
3787 spin_lock(&recall_lock);
3788 list_for_each_safe(pos, next, &del_recall_lru) {
3789 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3790 list_move(&dp->dl_recall_lru, &reaplist);
3792 spin_unlock(&recall_lock);
3793 list_for_each_safe(pos, next, &reaplist) {
3794 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3795 list_del_init(&dp->dl_recall_lru);
3796 unhash_delegation(dp);
3799 nfsd4_shutdown_recdir();
3804 nfs4_state_shutdown(void)
3806 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
3807 destroy_workqueue(laundry_wq);
3808 locks_end_grace(&nfsd4_manager);
3810 nfs4_release_reclaim();
3811 __nfs4_state_shutdown();
3812 nfs4_unlock_state();
3816 * user_recovery_dirname is protected by the nfsd_mutex since it's only
3817 * accessed when nfsd is starting.
3820 nfs4_set_recdir(char *recdir)
3822 strcpy(user_recovery_dirname, recdir);
3826 * Change the NFSv4 recovery directory to recdir.
3829 nfs4_reset_recoverydir(char *recdir)
3834 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
3838 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
3839 nfs4_set_recdir(recdir);
3847 nfs4_recoverydir(void)
3849 return user_recovery_dirname;
3853 * Called when leasetime is changed.
3855 * The only way the protocol gives us to handle on-the-fly lease changes is to
3856 * simulate a reboot. Instead of doing that, we just wait till the next time
3857 * we start to register any changes in lease time. If the administrator
3858 * really wants to change the lease time *now*, they can go ahead and bring
3859 * nfsd down and then back up again after changing the lease time.
3861 * user_lease_time is protected by nfsd_mutex since it's only really accessed
3862 * when nfsd is starting
3865 nfs4_reset_lease(time_t leasetime)
3867 user_lease_time = leasetime;