2 * Copyright (c) 2001 The Regents of the University of Michigan.
5 * Kendrick Smith <kmsmith@umich.edu>
6 * Andy Adamson <kandros@umich.edu>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the University nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
22 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
23 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
24 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
28 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
29 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
30 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
31 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
35 #include <linux/file.h>
36 #include <linux/smp_lock.h>
37 #include <linux/slab.h>
38 #include <linux/namei.h>
39 #include <linux/swap.h>
40 #include <linux/sunrpc/svcauth_gss.h>
41 #include <linux/sunrpc/clnt.h>
45 #define NFSDDBG_FACILITY NFSDDBG_PROC
48 time_t nfsd4_lease = 90; /* default lease time */
49 time_t nfsd4_grace = 90;
50 static time_t boot_time;
51 static u32 current_ownerid = 1;
52 static u32 current_fileid = 1;
53 static u32 current_delegid = 1;
54 static stateid_t zerostateid; /* bits all 0 */
55 static stateid_t onestateid; /* bits all 1 */
56 static u64 current_sessionid = 1;
58 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
59 #define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
61 /* forward declarations */
62 static struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
63 static struct nfs4_delegation * find_delegation_stateid(struct inode *ino, stateid_t *stid);
64 static char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery";
65 static void nfs4_set_recdir(char *recdir);
69 /* Currently used for almost all code touching nfsv4 state: */
70 static DEFINE_MUTEX(client_mutex);
73 * Currently used for the del_recall_lru and file hash table. In an
74 * effort to decrease the scope of the client_mutex, this spinlock may
75 * eventually cover more:
77 static DEFINE_SPINLOCK(recall_lock);
79 static struct kmem_cache *stateowner_slab = NULL;
80 static struct kmem_cache *file_slab = NULL;
81 static struct kmem_cache *stateid_slab = NULL;
82 static struct kmem_cache *deleg_slab = NULL;
87 mutex_lock(&client_mutex);
91 nfs4_unlock_state(void)
93 mutex_unlock(&client_mutex);
97 opaque_hashval(const void *ptr, int nbytes)
99 unsigned char *cptr = (unsigned char *) ptr;
109 static struct list_head del_recall_lru;
112 put_nfs4_file(struct nfs4_file *fi)
114 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
115 list_del(&fi->fi_hash);
116 spin_unlock(&recall_lock);
118 kmem_cache_free(file_slab, fi);
123 get_nfs4_file(struct nfs4_file *fi)
125 atomic_inc(&fi->fi_ref);
128 static int num_delegations;
129 unsigned int max_delegations;
132 * Open owner state (share locks)
135 /* hash tables for nfs4_stateowner */
136 #define OWNER_HASH_BITS 8
137 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
138 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
140 #define ownerid_hashval(id) \
141 ((id) & OWNER_HASH_MASK)
142 #define ownerstr_hashval(clientid, ownername) \
143 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
145 static struct list_head ownerid_hashtbl[OWNER_HASH_SIZE];
146 static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
148 /* hash table for nfs4_file */
149 #define FILE_HASH_BITS 8
150 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
151 #define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
152 /* hash table for (open)nfs4_stateid */
153 #define STATEID_HASH_BITS 10
154 #define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
155 #define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
157 #define file_hashval(x) \
158 hash_ptr(x, FILE_HASH_BITS)
159 #define stateid_hashval(owner_id, file_id) \
160 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
162 static struct list_head file_hashtbl[FILE_HASH_SIZE];
163 static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
165 static void __nfs4_file_get_access(struct nfs4_file *fp, int oflag)
167 BUG_ON(!(fp->fi_fds[oflag] || fp->fi_fds[O_RDWR]));
168 atomic_inc(&fp->fi_access[oflag]);
171 static void nfs4_file_get_access(struct nfs4_file *fp, int oflag)
173 if (oflag == O_RDWR) {
174 __nfs4_file_get_access(fp, O_RDONLY);
175 __nfs4_file_get_access(fp, O_WRONLY);
177 __nfs4_file_get_access(fp, oflag);
180 static void nfs4_file_put_fd(struct nfs4_file *fp, int oflag)
182 if (fp->fi_fds[oflag]) {
183 fput(fp->fi_fds[oflag]);
184 fp->fi_fds[oflag] = NULL;
188 static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
190 if (atomic_dec_and_test(&fp->fi_access[oflag])) {
191 nfs4_file_put_fd(fp, O_RDWR);
192 nfs4_file_put_fd(fp, oflag);
196 static void nfs4_file_put_access(struct nfs4_file *fp, int oflag)
198 if (oflag == O_RDWR) {
199 __nfs4_file_put_access(fp, O_RDONLY);
200 __nfs4_file_put_access(fp, O_WRONLY);
202 __nfs4_file_put_access(fp, oflag);
205 static struct nfs4_delegation *
206 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
208 struct nfs4_delegation *dp;
209 struct nfs4_file *fp = stp->st_file;
211 dprintk("NFSD alloc_init_deleg\n");
213 * Major work on the lease subsystem (for example, to support
214 * calbacks on stat) will be required before we can support
215 * write delegations properly.
217 if (type != NFS4_OPEN_DELEGATE_READ)
219 if (fp->fi_had_conflict)
221 if (num_delegations > max_delegations)
223 dp = kmem_cache_alloc(deleg_slab, GFP_KERNEL);
227 INIT_LIST_HEAD(&dp->dl_perfile);
228 INIT_LIST_HEAD(&dp->dl_perclnt);
229 INIT_LIST_HEAD(&dp->dl_recall_lru);
233 nfs4_file_get_access(fp, O_RDONLY);
236 dp->dl_stateid.si_boot = boot_time;
237 dp->dl_stateid.si_stateownerid = current_delegid++;
238 dp->dl_stateid.si_fileid = 0;
239 dp->dl_stateid.si_generation = 0;
240 fh_copy_shallow(&dp->dl_fh, ¤t_fh->fh_handle);
242 atomic_set(&dp->dl_count, 1);
243 list_add(&dp->dl_perfile, &fp->fi_delegations);
244 list_add(&dp->dl_perclnt, &clp->cl_delegations);
245 INIT_WORK(&dp->dl_recall.cb_work, nfsd4_do_callback_rpc);
250 nfs4_put_delegation(struct nfs4_delegation *dp)
252 if (atomic_dec_and_test(&dp->dl_count)) {
253 dprintk("NFSD: freeing dp %p\n",dp);
254 put_nfs4_file(dp->dl_file);
255 kmem_cache_free(deleg_slab, dp);
260 /* Remove the associated file_lock first, then remove the delegation.
261 * lease_modify() is called to remove the FS_LEASE file_lock from
262 * the i_flock list, eventually calling nfsd's lock_manager
263 * fl_release_callback.
266 nfs4_close_delegation(struct nfs4_delegation *dp)
268 struct file *filp = find_readable_file(dp->dl_file);
270 dprintk("NFSD: close_delegation dp %p\n",dp);
272 vfs_setlease(filp, F_UNLCK, &dp->dl_flock);
273 nfs4_file_put_access(dp->dl_file, O_RDONLY);
276 /* Called under the state lock. */
278 unhash_delegation(struct nfs4_delegation *dp)
280 list_del_init(&dp->dl_perfile);
281 list_del_init(&dp->dl_perclnt);
282 spin_lock(&recall_lock);
283 list_del_init(&dp->dl_recall_lru);
284 spin_unlock(&recall_lock);
285 nfs4_close_delegation(dp);
286 nfs4_put_delegation(dp);
293 /* client_lock protects the client lru list and session hash table */
294 static DEFINE_SPINLOCK(client_lock);
296 /* Hash tables for nfs4_clientid state */
297 #define CLIENT_HASH_BITS 4
298 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
299 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
301 #define clientid_hashval(id) \
302 ((id) & CLIENT_HASH_MASK)
303 #define clientstr_hashval(name) \
304 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
306 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
307 * used in reboot/reset lease grace period processing
309 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
310 * setclientid_confirmed info.
312 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
315 * client_lru holds client queue ordered by nfs4_client.cl_time
318 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
319 * for last close replay.
321 static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
322 static int reclaim_str_hashtbl_size = 0;
323 static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
324 static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
325 static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
326 static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
327 static struct list_head client_lru;
328 static struct list_head close_lru;
330 static void unhash_generic_stateid(struct nfs4_stateid *stp)
332 list_del(&stp->st_hash);
333 list_del(&stp->st_perfile);
334 list_del(&stp->st_perstateowner);
337 static void free_generic_stateid(struct nfs4_stateid *stp)
339 put_nfs4_file(stp->st_file);
340 kmem_cache_free(stateid_slab, stp);
343 static void release_lock_stateid(struct nfs4_stateid *stp)
347 unhash_generic_stateid(stp);
348 file = find_any_file(stp->st_file);
350 locks_remove_posix(file, (fl_owner_t)stp->st_stateowner);
351 free_generic_stateid(stp);
354 static void unhash_lockowner(struct nfs4_stateowner *sop)
356 struct nfs4_stateid *stp;
358 list_del(&sop->so_idhash);
359 list_del(&sop->so_strhash);
360 list_del(&sop->so_perstateid);
361 while (!list_empty(&sop->so_stateids)) {
362 stp = list_first_entry(&sop->so_stateids,
363 struct nfs4_stateid, st_perstateowner);
364 release_lock_stateid(stp);
368 static void release_lockowner(struct nfs4_stateowner *sop)
370 unhash_lockowner(sop);
371 nfs4_put_stateowner(sop);
375 release_stateid_lockowners(struct nfs4_stateid *open_stp)
377 struct nfs4_stateowner *lock_sop;
379 while (!list_empty(&open_stp->st_lockowners)) {
380 lock_sop = list_entry(open_stp->st_lockowners.next,
381 struct nfs4_stateowner, so_perstateid);
382 /* list_del(&open_stp->st_lockowners); */
383 BUG_ON(lock_sop->so_is_open_owner);
384 release_lockowner(lock_sop);
389 * We store the NONE, READ, WRITE, and BOTH bits separately in the
390 * st_{access,deny}_bmap field of the stateid, in order to track not
391 * only what share bits are currently in force, but also what
392 * combinations of share bits previous opens have used. This allows us
393 * to enforce the recommendation of rfc 3530 14.2.19 that the server
394 * return an error if the client attempt to downgrade to a combination
395 * of share bits not explicable by closing some of its previous opens.
397 * XXX: This enforcement is actually incomplete, since we don't keep
398 * track of access/deny bit combinations; so, e.g., we allow:
400 * OPEN allow read, deny write
401 * OPEN allow both, deny none
402 * DOWNGRADE allow read, deny none
404 * which we should reject.
407 set_access(unsigned int *access, unsigned long bmap) {
411 for (i = 1; i < 4; i++) {
412 if (test_bit(i, &bmap))
418 set_deny(unsigned int *deny, unsigned long bmap) {
422 for (i = 0; i < 4; i++) {
423 if (test_bit(i, &bmap))
429 test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
430 unsigned int access, deny;
432 set_access(&access, stp->st_access_bmap);
433 set_deny(&deny, stp->st_deny_bmap);
434 if ((access & open->op_share_deny) || (deny & open->op_share_access))
439 static int nfs4_access_to_omode(u32 access)
441 switch (access & NFS4_SHARE_ACCESS_BOTH) {
442 case NFS4_SHARE_ACCESS_READ:
444 case NFS4_SHARE_ACCESS_WRITE:
446 case NFS4_SHARE_ACCESS_BOTH:
452 static int nfs4_access_bmap_to_omode(struct nfs4_stateid *stp)
456 set_access(&access, stp->st_access_bmap);
457 return nfs4_access_to_omode(access);
460 static void release_open_stateid(struct nfs4_stateid *stp)
462 int oflag = nfs4_access_bmap_to_omode(stp);
464 unhash_generic_stateid(stp);
465 release_stateid_lockowners(stp);
466 nfs4_file_put_access(stp->st_file, oflag);
467 free_generic_stateid(stp);
470 static void unhash_openowner(struct nfs4_stateowner *sop)
472 struct nfs4_stateid *stp;
474 list_del(&sop->so_idhash);
475 list_del(&sop->so_strhash);
476 list_del(&sop->so_perclient);
477 list_del(&sop->so_perstateid); /* XXX: necessary? */
478 while (!list_empty(&sop->so_stateids)) {
479 stp = list_first_entry(&sop->so_stateids,
480 struct nfs4_stateid, st_perstateowner);
481 release_open_stateid(stp);
485 static void release_openowner(struct nfs4_stateowner *sop)
487 unhash_openowner(sop);
488 list_del(&sop->so_close_lru);
489 nfs4_put_stateowner(sop);
492 #define SESSION_HASH_SIZE 512
493 static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
496 hash_sessionid(struct nfs4_sessionid *sessionid)
498 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
500 return sid->sequence % SESSION_HASH_SIZE;
504 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
506 u32 *ptr = (u32 *)(&sessionid->data[0]);
507 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
511 gen_sessionid(struct nfsd4_session *ses)
513 struct nfs4_client *clp = ses->se_client;
514 struct nfsd4_sessionid *sid;
516 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
517 sid->clientid = clp->cl_clientid;
518 sid->sequence = current_sessionid++;
523 * The protocol defines ca_maxresponssize_cached to include the size of
524 * the rpc header, but all we need to cache is the data starting after
525 * the end of the initial SEQUENCE operation--the rest we regenerate
526 * each time. Therefore we can advertise a ca_maxresponssize_cached
527 * value that is the number of bytes in our cache plus a few additional
528 * bytes. In order to stay on the safe side, and not promise more than
529 * we can cache, those additional bytes must be the minimum possible: 24
530 * bytes of rpc header (xid through accept state, with AUTH_NULL
531 * verifier), 12 for the compound header (with zero-length tag), and 44
532 * for the SEQUENCE op response:
534 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
537 free_session_slots(struct nfsd4_session *ses)
541 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
542 kfree(ses->se_slots[i]);
546 * We don't actually need to cache the rpc and session headers, so we
547 * can allocate a little less for each slot:
549 static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
551 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
554 static int nfsd4_sanitize_slot_size(u32 size)
556 size -= NFSD_MIN_HDR_SEQ_SZ; /* We don't cache the rpc header */
557 size = min_t(u32, size, NFSD_SLOT_CACHE_SIZE);
563 * XXX: If we run out of reserved DRC memory we could (up to a point)
564 * re-negotiate active sessions and reduce their slot usage to make
565 * rooom for new connections. For now we just fail the create session.
567 static int nfsd4_get_drc_mem(int slotsize, u32 num)
571 num = min_t(u32, num, NFSD_MAX_SLOTS_PER_SESSION);
573 spin_lock(&nfsd_drc_lock);
574 avail = min_t(int, NFSD_MAX_MEM_PER_SESSION,
575 nfsd_drc_max_mem - nfsd_drc_mem_used);
576 num = min_t(int, num, avail / slotsize);
577 nfsd_drc_mem_used += num * slotsize;
578 spin_unlock(&nfsd_drc_lock);
583 static void nfsd4_put_drc_mem(int slotsize, int num)
585 spin_lock(&nfsd_drc_lock);
586 nfsd_drc_mem_used -= slotsize * num;
587 spin_unlock(&nfsd_drc_lock);
590 static struct nfsd4_session *alloc_session(int slotsize, int numslots)
592 struct nfsd4_session *new;
595 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
596 + sizeof(struct nfsd4_session) > PAGE_SIZE);
597 mem = numslots * sizeof(struct nfsd4_slot *);
599 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
602 /* allocate each struct nfsd4_slot and data cache in one piece */
603 for (i = 0; i < numslots; i++) {
604 mem = sizeof(struct nfsd4_slot) + slotsize;
605 new->se_slots[i] = kzalloc(mem, GFP_KERNEL);
606 if (!new->se_slots[i])
612 kfree(new->se_slots[i]);
617 static void init_forechannel_attrs(struct nfsd4_channel_attrs *new, struct nfsd4_channel_attrs *req, int numslots, int slotsize)
619 u32 maxrpc = nfsd_serv->sv_max_mesg;
621 new->maxreqs = numslots;
622 new->maxresp_cached = slotsize + NFSD_MIN_HDR_SEQ_SZ;
623 new->maxreq_sz = min_t(u32, req->maxreq_sz, maxrpc);
624 new->maxresp_sz = min_t(u32, req->maxresp_sz, maxrpc);
625 new->maxops = min_t(u32, req->maxops, NFSD_MAX_OPS_PER_COMPOUND);
628 static void free_conn(struct nfsd4_conn *c)
630 svc_xprt_put(c->cn_xprt);
634 static void nfsd4_conn_lost(struct svc_xpt_user *u)
636 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
637 struct nfs4_client *clp = c->cn_session->se_client;
639 spin_lock(&clp->cl_lock);
640 if (!list_empty(&c->cn_persession)) {
641 list_del(&c->cn_persession);
644 spin_unlock(&clp->cl_lock);
647 static __be32 nfsd4_new_conn(struct svc_rqst *rqstp, struct nfsd4_session *ses)
649 struct nfs4_client *clp = ses->se_client;
650 struct nfsd4_conn *conn;
652 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
654 return nfserr_jukebox;
655 conn->cn_flags = NFS4_CDFC4_FORE;
656 svc_xprt_get(rqstp->rq_xprt);
657 conn->cn_xprt = rqstp->rq_xprt;
658 conn->cn_session = ses;
660 spin_lock(&clp->cl_lock);
661 list_add(&conn->cn_persession, &ses->se_conns);
662 spin_unlock(&clp->cl_lock);
664 conn->cn_xpt_user.callback = nfsd4_conn_lost;
665 register_xpt_user(rqstp->rq_xprt, &conn->cn_xpt_user);
669 static void nfsd4_del_conns(struct nfsd4_session *s)
671 struct nfs4_client *clp = s->se_client;
672 struct nfsd4_conn *c;
674 spin_lock(&clp->cl_lock);
675 while (!list_empty(&s->se_conns)) {
676 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
677 list_del_init(&c->cn_persession);
678 spin_unlock(&clp->cl_lock);
680 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
683 spin_lock(&clp->cl_lock);
685 spin_unlock(&clp->cl_lock);
688 void free_session(struct kref *kref)
690 struct nfsd4_session *ses;
693 ses = container_of(kref, struct nfsd4_session, se_ref);
694 nfsd4_del_conns(ses);
695 spin_lock(&nfsd_drc_lock);
696 mem = ses->se_fchannel.maxreqs * slot_bytes(&ses->se_fchannel);
697 nfsd_drc_mem_used -= mem;
698 spin_unlock(&nfsd_drc_lock);
699 free_session_slots(ses);
704 static __be32 alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp, struct nfsd4_create_session *cses)
706 struct nfsd4_session *new;
707 struct nfsd4_channel_attrs *fchan = &cses->fore_channel;
708 int numslots, slotsize;
713 * Note decreasing slot size below client's request may
714 * make it difficult for client to function correctly, whereas
715 * decreasing the number of slots will (just?) affect
716 * performance. When short on memory we therefore prefer to
717 * decrease number of slots instead of their size.
719 slotsize = nfsd4_sanitize_slot_size(fchan->maxresp_cached);
720 numslots = nfsd4_get_drc_mem(slotsize, fchan->maxreqs);
722 new = alloc_session(slotsize, numslots);
724 nfsd4_put_drc_mem(slotsize, fchan->maxreqs);
725 return nfserr_jukebox;
727 init_forechannel_attrs(&new->se_fchannel, fchan, numslots, slotsize);
729 new->se_client = clp;
731 memcpy(clp->cl_sessionid.data, new->se_sessionid.data,
732 NFS4_MAX_SESSIONID_LEN);
734 INIT_LIST_HEAD(&new->se_conns);
736 new->se_flags = cses->flags;
737 kref_init(&new->se_ref);
738 idx = hash_sessionid(&new->se_sessionid);
739 spin_lock(&client_lock);
740 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
741 list_add(&new->se_perclnt, &clp->cl_sessions);
742 spin_unlock(&client_lock);
744 status = nfsd4_new_conn(rqstp, new);
746 free_session(&new->se_ref);
747 return nfserr_jukebox;
752 /* caller must hold client_lock */
753 static struct nfsd4_session *
754 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
756 struct nfsd4_session *elem;
759 dump_sessionid(__func__, sessionid);
760 idx = hash_sessionid(sessionid);
761 /* Search in the appropriate list */
762 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
763 if (!memcmp(elem->se_sessionid.data, sessionid->data,
764 NFS4_MAX_SESSIONID_LEN)) {
769 dprintk("%s: session not found\n", __func__);
773 /* caller must hold client_lock */
775 unhash_session(struct nfsd4_session *ses)
777 list_del(&ses->se_hash);
778 list_del(&ses->se_perclnt);
781 /* must be called under the client_lock */
783 renew_client_locked(struct nfs4_client *clp)
785 if (is_client_expired(clp)) {
786 dprintk("%s: client (clientid %08x/%08x) already expired\n",
788 clp->cl_clientid.cl_boot,
789 clp->cl_clientid.cl_id);
794 * Move client to the end to the LRU list.
796 dprintk("renewing client (clientid %08x/%08x)\n",
797 clp->cl_clientid.cl_boot,
798 clp->cl_clientid.cl_id);
799 list_move_tail(&clp->cl_lru, &client_lru);
800 clp->cl_time = get_seconds();
804 renew_client(struct nfs4_client *clp)
806 spin_lock(&client_lock);
807 renew_client_locked(clp);
808 spin_unlock(&client_lock);
811 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
813 STALE_CLIENTID(clientid_t *clid)
815 if (clid->cl_boot == boot_time)
817 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
818 clid->cl_boot, clid->cl_id, boot_time);
823 * XXX Should we use a slab cache ?
824 * This type of memory management is somewhat inefficient, but we use it
825 * anyway since SETCLIENTID is not a common operation.
827 static struct nfs4_client *alloc_client(struct xdr_netobj name)
829 struct nfs4_client *clp;
831 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
834 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
835 if (clp->cl_name.data == NULL) {
839 memcpy(clp->cl_name.data, name.data, name.len);
840 clp->cl_name.len = name.len;
845 free_client(struct nfs4_client *clp)
847 if (clp->cl_cred.cr_group_info)
848 put_group_info(clp->cl_cred.cr_group_info);
849 kfree(clp->cl_principal);
850 kfree(clp->cl_name.data);
855 release_session_client(struct nfsd4_session *session)
857 struct nfs4_client *clp = session->se_client;
859 if (!atomic_dec_and_lock(&clp->cl_refcount, &client_lock))
861 if (is_client_expired(clp)) {
863 session->se_client = NULL;
865 renew_client_locked(clp);
866 spin_unlock(&client_lock);
869 /* must be called under the client_lock */
871 unhash_client_locked(struct nfs4_client *clp)
873 mark_client_expired(clp);
874 list_del(&clp->cl_lru);
875 while (!list_empty(&clp->cl_sessions)) {
876 struct nfsd4_session *ses;
877 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
880 nfsd4_put_session(ses);
885 expire_client(struct nfs4_client *clp)
887 struct nfs4_stateowner *sop;
888 struct nfs4_delegation *dp;
889 struct list_head reaplist;
891 INIT_LIST_HEAD(&reaplist);
892 spin_lock(&recall_lock);
893 while (!list_empty(&clp->cl_delegations)) {
894 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
895 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
897 list_del_init(&dp->dl_perclnt);
898 list_move(&dp->dl_recall_lru, &reaplist);
900 spin_unlock(&recall_lock);
901 while (!list_empty(&reaplist)) {
902 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
903 list_del_init(&dp->dl_recall_lru);
904 unhash_delegation(dp);
906 while (!list_empty(&clp->cl_openowners)) {
907 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
908 release_openowner(sop);
910 nfsd4_shutdown_callback(clp);
911 if (clp->cl_cb_conn.cb_xprt)
912 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
913 list_del(&clp->cl_idhash);
914 list_del(&clp->cl_strhash);
915 spin_lock(&client_lock);
916 unhash_client_locked(clp);
917 if (atomic_read(&clp->cl_refcount) == 0)
919 spin_unlock(&client_lock);
922 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
924 memcpy(target->cl_verifier.data, source->data,
925 sizeof(target->cl_verifier.data));
928 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
930 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
931 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
934 static void copy_cred(struct svc_cred *target, struct svc_cred *source)
936 target->cr_uid = source->cr_uid;
937 target->cr_gid = source->cr_gid;
938 target->cr_group_info = source->cr_group_info;
939 get_group_info(target->cr_group_info);
942 static int same_name(const char *n1, const char *n2)
944 return 0 == memcmp(n1, n2, HEXDIR_LEN);
948 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
950 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
954 same_clid(clientid_t *cl1, clientid_t *cl2)
956 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
959 /* XXX what about NGROUP */
961 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
963 return cr1->cr_uid == cr2->cr_uid;
966 static void gen_clid(struct nfs4_client *clp)
968 static u32 current_clientid = 1;
970 clp->cl_clientid.cl_boot = boot_time;
971 clp->cl_clientid.cl_id = current_clientid++;
974 static void gen_confirm(struct nfs4_client *clp)
979 p = (u32 *)clp->cl_confirm.data;
980 *p++ = get_seconds();
984 static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
985 struct svc_rqst *rqstp, nfs4_verifier *verf)
987 struct nfs4_client *clp;
988 struct sockaddr *sa = svc_addr(rqstp);
991 clp = alloc_client(name);
995 princ = svc_gss_principal(rqstp);
997 clp->cl_principal = kstrdup(princ, GFP_KERNEL);
998 if (clp->cl_principal == NULL) {
1004 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1005 atomic_set(&clp->cl_refcount, 0);
1006 atomic_set(&clp->cl_cb_set, 0);
1007 INIT_LIST_HEAD(&clp->cl_idhash);
1008 INIT_LIST_HEAD(&clp->cl_strhash);
1009 INIT_LIST_HEAD(&clp->cl_openowners);
1010 INIT_LIST_HEAD(&clp->cl_delegations);
1011 INIT_LIST_HEAD(&clp->cl_sessions);
1012 INIT_LIST_HEAD(&clp->cl_lru);
1013 spin_lock_init(&clp->cl_lock);
1014 INIT_WORK(&clp->cl_cb_null.cb_work, nfsd4_do_callback_rpc);
1015 clp->cl_time = get_seconds();
1016 clear_bit(0, &clp->cl_cb_slot_busy);
1017 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1018 copy_verf(clp, verf);
1019 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1020 clp->cl_flavor = rqstp->rq_flavor;
1021 copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1027 static int check_name(struct xdr_netobj name)
1031 if (name.len > NFS4_OPAQUE_LIMIT) {
1032 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
1039 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
1041 unsigned int idhashval;
1043 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
1044 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1045 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
1050 move_to_confirmed(struct nfs4_client *clp)
1052 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1053 unsigned int strhashval;
1055 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1056 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
1057 strhashval = clientstr_hashval(clp->cl_recdir);
1058 list_move(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
1062 static struct nfs4_client *
1063 find_confirmed_client(clientid_t *clid)
1065 struct nfs4_client *clp;
1066 unsigned int idhashval = clientid_hashval(clid->cl_id);
1068 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
1069 if (same_clid(&clp->cl_clientid, clid))
1075 static struct nfs4_client *
1076 find_unconfirmed_client(clientid_t *clid)
1078 struct nfs4_client *clp;
1079 unsigned int idhashval = clientid_hashval(clid->cl_id);
1081 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
1082 if (same_clid(&clp->cl_clientid, clid))
1089 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
1090 * parameter. Matching is based on the fact the at least one of the
1091 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
1093 * FIXME: we need to unify the clientid namespaces for nfsv4.x
1094 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
1095 * and SET_CLIENTID{,_CONFIRM}
1098 match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
1100 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
1101 return use_exchange_id == has_exchange_flags;
1104 static struct nfs4_client *
1105 find_confirmed_client_by_str(const char *dname, unsigned int hashval,
1106 bool use_exchange_id)
1108 struct nfs4_client *clp;
1110 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
1111 if (same_name(clp->cl_recdir, dname) &&
1112 match_clientid_establishment(clp, use_exchange_id))
1118 static struct nfs4_client *
1119 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
1120 bool use_exchange_id)
1122 struct nfs4_client *clp;
1124 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
1125 if (same_name(clp->cl_recdir, dname) &&
1126 match_clientid_establishment(clp, use_exchange_id))
1133 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, u32 scopeid)
1135 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1136 unsigned short expected_family;
1138 /* Currently, we only support tcp and tcp6 for the callback channel */
1139 if (se->se_callback_netid_len == 3 &&
1140 !memcmp(se->se_callback_netid_val, "tcp", 3))
1141 expected_family = AF_INET;
1142 else if (se->se_callback_netid_len == 4 &&
1143 !memcmp(se->se_callback_netid_val, "tcp6", 4))
1144 expected_family = AF_INET6;
1148 conn->cb_addrlen = rpc_uaddr2sockaddr(se->se_callback_addr_val,
1149 se->se_callback_addr_len,
1150 (struct sockaddr *)&conn->cb_addr,
1151 sizeof(conn->cb_addr));
1153 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
1156 if (conn->cb_addr.ss_family == AF_INET6)
1157 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
1159 conn->cb_minorversion = 0;
1160 conn->cb_prog = se->se_callback_prog;
1161 conn->cb_ident = se->se_callback_ident;
1164 conn->cb_addr.ss_family = AF_UNSPEC;
1165 conn->cb_addrlen = 0;
1166 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1167 "will not receive delegations\n",
1168 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1174 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1177 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1179 struct nfsd4_slot *slot = resp->cstate.slot;
1182 dprintk("--> %s slot %p\n", __func__, slot);
1184 slot->sl_opcnt = resp->opcnt;
1185 slot->sl_status = resp->cstate.status;
1187 if (nfsd4_not_cached(resp)) {
1188 slot->sl_datalen = 0;
1191 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1192 base = (char *)resp->cstate.datap -
1193 (char *)resp->xbuf->head[0].iov_base;
1194 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1196 WARN("%s: sessions DRC could not cache compound\n", __func__);
1201 * Encode the replay sequence operation from the slot values.
1202 * If cachethis is FALSE encode the uncached rep error on the next
1203 * operation which sets resp->p and increments resp->opcnt for
1204 * nfs4svc_encode_compoundres.
1208 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1209 struct nfsd4_compoundres *resp)
1211 struct nfsd4_op *op;
1212 struct nfsd4_slot *slot = resp->cstate.slot;
1214 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__,
1215 resp->opcnt, resp->cstate.slot->sl_cachethis);
1217 /* Encode the replayed sequence operation */
1218 op = &args->ops[resp->opcnt - 1];
1219 nfsd4_encode_operation(resp, op);
1221 /* Return nfserr_retry_uncached_rep in next operation. */
1222 if (args->opcnt > 1 && slot->sl_cachethis == 0) {
1223 op = &args->ops[resp->opcnt++];
1224 op->status = nfserr_retry_uncached_rep;
1225 nfsd4_encode_operation(resp, op);
1231 * The sequence operation is not cached because we can use the slot and
1235 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1236 struct nfsd4_sequence *seq)
1238 struct nfsd4_slot *slot = resp->cstate.slot;
1241 dprintk("--> %s slot %p\n", __func__, slot);
1243 /* Either returns 0 or nfserr_retry_uncached */
1244 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1245 if (status == nfserr_retry_uncached_rep)
1248 /* The sequence operation has been encoded, cstate->datap set. */
1249 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1251 resp->opcnt = slot->sl_opcnt;
1252 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1253 status = slot->sl_status;
1259 * Set the exchange_id flags returned by the server.
1262 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1264 /* pNFS is not supported */
1265 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1267 /* Referrals are supported, Migration is not. */
1268 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1270 /* set the wire flags to return to client. */
1271 clid->flags = new->cl_exchange_flags;
1275 nfsd4_exchange_id(struct svc_rqst *rqstp,
1276 struct nfsd4_compound_state *cstate,
1277 struct nfsd4_exchange_id *exid)
1279 struct nfs4_client *unconf, *conf, *new;
1281 unsigned int strhashval;
1282 char dname[HEXDIR_LEN];
1283 char addr_str[INET6_ADDRSTRLEN];
1284 nfs4_verifier verf = exid->verifier;
1285 struct sockaddr *sa = svc_addr(rqstp);
1287 rpc_ntop(sa, addr_str, sizeof(addr_str));
1288 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1289 "ip_addr=%s flags %x, spa_how %d\n",
1290 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1291 addr_str, exid->flags, exid->spa_how);
1293 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1294 return nfserr_inval;
1296 /* Currently only support SP4_NONE */
1297 switch (exid->spa_how) {
1301 return nfserr_encr_alg_unsupp;
1303 BUG(); /* checked by xdr code */
1305 return nfserr_serverfault; /* no excuse :-/ */
1308 status = nfs4_make_rec_clidname(dname, &exid->clname);
1313 strhashval = clientstr_hashval(dname);
1318 conf = find_confirmed_client_by_str(dname, strhashval, true);
1320 if (!same_verf(&verf, &conf->cl_verifier)) {
1321 /* 18.35.4 case 8 */
1322 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1323 status = nfserr_not_same;
1326 /* Client reboot: destroy old state */
1327 expire_client(conf);
1330 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1331 /* 18.35.4 case 9 */
1332 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1333 status = nfserr_perm;
1336 expire_client(conf);
1340 * Set bit when the owner id and verifier map to an already
1341 * confirmed client id (18.35.3).
1343 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1346 * Falling into 18.35.4 case 2, possible router replay.
1347 * Leave confirmed record intact and return same result.
1349 copy_verf(conf, &verf);
1354 /* 18.35.4 case 7 */
1355 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1356 status = nfserr_noent;
1360 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
1363 * Possible retry or client restart. Per 18.35.4 case 4,
1364 * a new unconfirmed record should be generated regardless
1365 * of whether any properties have changed.
1367 expire_client(unconf);
1372 new = create_client(exid->clname, dname, rqstp, &verf);
1374 status = nfserr_jukebox;
1379 add_to_unconfirmed(new, strhashval);
1381 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1382 exid->clientid.cl_id = new->cl_clientid.cl_id;
1385 nfsd4_set_ex_flags(new, exid);
1387 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1388 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1392 nfs4_unlock_state();
1394 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1399 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1401 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1404 /* The slot is in use, and no response has been sent. */
1406 if (seqid == slot_seqid)
1407 return nfserr_jukebox;
1409 return nfserr_seq_misordered;
1412 if (likely(seqid == slot_seqid + 1))
1415 if (seqid == slot_seqid)
1416 return nfserr_replay_cache;
1418 if (seqid == 1 && (slot_seqid + 1) == 0)
1420 /* Misordered replay or misordered new request */
1421 return nfserr_seq_misordered;
1425 * Cache the create session result into the create session single DRC
1426 * slot cache by saving the xdr structure. sl_seqid has been set.
1427 * Do this for solo or embedded create session operations.
1430 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1431 struct nfsd4_clid_slot *slot, int nfserr)
1433 slot->sl_status = nfserr;
1434 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1438 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1439 struct nfsd4_clid_slot *slot)
1441 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1442 return slot->sl_status;
1446 nfsd4_create_session(struct svc_rqst *rqstp,
1447 struct nfsd4_compound_state *cstate,
1448 struct nfsd4_create_session *cr_ses)
1450 struct sockaddr *sa = svc_addr(rqstp);
1451 struct nfs4_client *conf, *unconf;
1452 struct nfsd4_clid_slot *cs_slot = NULL;
1456 unconf = find_unconfirmed_client(&cr_ses->clientid);
1457 conf = find_confirmed_client(&cr_ses->clientid);
1460 cs_slot = &conf->cl_cs_slot;
1461 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1462 if (status == nfserr_replay_cache) {
1463 dprintk("Got a create_session replay! seqid= %d\n",
1465 /* Return the cached reply status */
1466 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1468 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1469 status = nfserr_seq_misordered;
1470 dprintk("Sequence misordered!\n");
1471 dprintk("Expected seqid= %d but got seqid= %d\n",
1472 cs_slot->sl_seqid, cr_ses->seqid);
1475 cs_slot->sl_seqid++;
1476 } else if (unconf) {
1477 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1478 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1479 status = nfserr_clid_inuse;
1483 cs_slot = &unconf->cl_cs_slot;
1484 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1486 /* an unconfirmed replay returns misordered */
1487 status = nfserr_seq_misordered;
1491 cs_slot->sl_seqid++; /* from 0 to 1 */
1492 move_to_confirmed(unconf);
1494 if (cr_ses->flags & SESSION4_BACK_CHAN) {
1495 unconf->cl_cb_conn.cb_xprt = rqstp->rq_xprt;
1496 svc_xprt_get(rqstp->rq_xprt);
1498 (struct sockaddr *)&unconf->cl_cb_conn.cb_addr,
1500 unconf->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1501 unconf->cl_cb_conn.cb_minorversion =
1502 cstate->minorversion;
1503 unconf->cl_cb_conn.cb_prog = cr_ses->callback_prog;
1504 unconf->cl_cb_seq_nr = 1;
1505 nfsd4_probe_callback(unconf, &unconf->cl_cb_conn);
1509 status = nfserr_stale_clientid;
1514 * We do not support RDMA or persistent sessions
1516 cr_ses->flags &= ~SESSION4_PERSIST;
1517 cr_ses->flags &= ~SESSION4_RDMA;
1519 status = alloc_init_session(rqstp, conf, cr_ses);
1523 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1524 NFS4_MAX_SESSIONID_LEN);
1525 cr_ses->seqid = cs_slot->sl_seqid;
1528 /* cache solo and embedded create sessions under the state lock */
1529 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1531 nfs4_unlock_state();
1532 dprintk("%s returns %d\n", __func__, ntohl(status));
1536 static bool nfsd4_last_compound_op(struct svc_rqst *rqstp)
1538 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1539 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1541 return argp->opcnt == resp->opcnt;
1544 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1548 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1552 nfsd4_destroy_session(struct svc_rqst *r,
1553 struct nfsd4_compound_state *cstate,
1554 struct nfsd4_destroy_session *sessionid)
1556 struct nfsd4_session *ses;
1557 u32 status = nfserr_badsession;
1560 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1561 * - Should we return nfserr_back_chan_busy if waiting for
1562 * callbacks on to-be-destroyed session?
1563 * - Do we need to clear any callback info from previous session?
1566 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
1567 if (!nfsd4_last_compound_op(r))
1568 return nfserr_not_only_op;
1570 dump_sessionid(__func__, &sessionid->sessionid);
1571 spin_lock(&client_lock);
1572 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1574 spin_unlock(&client_lock);
1578 unhash_session(ses);
1579 spin_unlock(&client_lock);
1582 /* wait for callbacks */
1583 nfsd4_shutdown_callback(ses->se_client);
1584 nfs4_unlock_state();
1586 nfsd4_del_conns(ses);
1588 nfsd4_put_session(ses);
1591 dprintk("%s returns %d\n", __func__, ntohl(status));
1596 nfsd4_sequence(struct svc_rqst *rqstp,
1597 struct nfsd4_compound_state *cstate,
1598 struct nfsd4_sequence *seq)
1600 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1601 struct nfsd4_session *session;
1602 struct nfsd4_slot *slot;
1605 if (resp->opcnt != 1)
1606 return nfserr_sequence_pos;
1608 spin_lock(&client_lock);
1609 status = nfserr_badsession;
1610 session = find_in_sessionid_hashtbl(&seq->sessionid);
1614 status = nfserr_badslot;
1615 if (seq->slotid >= session->se_fchannel.maxreqs)
1618 slot = session->se_slots[seq->slotid];
1619 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1621 /* We do not negotiate the number of slots yet, so set the
1622 * maxslots to the session maxreqs which is used to encode
1623 * sr_highest_slotid and the sr_target_slot id to maxslots */
1624 seq->maxslots = session->se_fchannel.maxreqs;
1626 status = check_slot_seqid(seq->seqid, slot->sl_seqid, slot->sl_inuse);
1627 if (status == nfserr_replay_cache) {
1628 cstate->slot = slot;
1629 cstate->session = session;
1630 /* Return the cached reply status and set cstate->status
1631 * for nfsd4_proc_compound processing */
1632 status = nfsd4_replay_cache_entry(resp, seq);
1633 cstate->status = nfserr_replay_cache;
1639 /* Success! bump slot seqid */
1640 slot->sl_inuse = true;
1641 slot->sl_seqid = seq->seqid;
1642 slot->sl_cachethis = seq->cachethis;
1644 cstate->slot = slot;
1645 cstate->session = session;
1648 /* Hold a session reference until done processing the compound. */
1649 if (cstate->session) {
1650 nfsd4_get_session(cstate->session);
1651 atomic_inc(&session->se_client->cl_refcount);
1653 spin_unlock(&client_lock);
1654 dprintk("%s: return %d\n", __func__, ntohl(status));
1659 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
1661 if (rc->rca_one_fs) {
1662 if (!cstate->current_fh.fh_dentry)
1663 return nfserr_nofilehandle;
1665 * We don't take advantage of the rca_one_fs case.
1666 * That's OK, it's optional, we can safely ignore it.
1671 if (is_client_expired(cstate->session->se_client)) {
1672 nfs4_unlock_state();
1674 * The following error isn't really legal.
1675 * But we only get here if the client just explicitly
1676 * destroyed the client. Surely it no longer cares what
1677 * error it gets back on an operation for the dead
1680 return nfserr_stale_clientid;
1682 nfsd4_create_clid_dir(cstate->session->se_client);
1683 nfs4_unlock_state();
1688 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1689 struct nfsd4_setclientid *setclid)
1691 struct sockaddr *sa = svc_addr(rqstp);
1692 struct xdr_netobj clname = {
1693 .len = setclid->se_namelen,
1694 .data = setclid->se_name,
1696 nfs4_verifier clverifier = setclid->se_verf;
1697 unsigned int strhashval;
1698 struct nfs4_client *conf, *unconf, *new;
1700 char dname[HEXDIR_LEN];
1702 if (!check_name(clname))
1703 return nfserr_inval;
1705 status = nfs4_make_rec_clidname(dname, &clname);
1710 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1711 * We get here on a DRC miss.
1714 strhashval = clientstr_hashval(dname);
1717 conf = find_confirmed_client_by_str(dname, strhashval, false);
1719 /* RFC 3530 14.2.33 CASE 0: */
1720 status = nfserr_clid_inuse;
1721 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1722 char addr_str[INET6_ADDRSTRLEN];
1723 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
1725 dprintk("NFSD: setclientid: string in use by client "
1726 "at %s\n", addr_str);
1731 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1732 * has a description of SETCLIENTID request processing consisting
1733 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1735 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1736 status = nfserr_resource;
1739 * RFC 3530 14.2.33 CASE 4:
1740 * placed first, because it is the normal case
1743 expire_client(unconf);
1744 new = create_client(clname, dname, rqstp, &clverifier);
1748 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1750 * RFC 3530 14.2.33 CASE 1:
1751 * probable callback update
1754 /* Note this is removing unconfirmed {*x***},
1755 * which is stronger than RFC recommended {vxc**}.
1756 * This has the advantage that there is at most
1757 * one {*x***} in either list at any time.
1759 expire_client(unconf);
1761 new = create_client(clname, dname, rqstp, &clverifier);
1764 copy_clid(new, conf);
1765 } else if (!unconf) {
1767 * RFC 3530 14.2.33 CASE 2:
1768 * probable client reboot; state will be removed if
1771 new = create_client(clname, dname, rqstp, &clverifier);
1777 * RFC 3530 14.2.33 CASE 3:
1778 * probable client reboot; state will be removed if
1781 expire_client(unconf);
1782 new = create_client(clname, dname, rqstp, &clverifier);
1787 gen_callback(new, setclid, rpc_get_scope_id(sa));
1788 add_to_unconfirmed(new, strhashval);
1789 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1790 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1791 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1794 nfs4_unlock_state();
1800 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1801 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1802 * bullets, labeled as CASE1 - CASE4 below.
1805 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1806 struct nfsd4_compound_state *cstate,
1807 struct nfsd4_setclientid_confirm *setclientid_confirm)
1809 struct sockaddr *sa = svc_addr(rqstp);
1810 struct nfs4_client *conf, *unconf;
1811 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1812 clientid_t * clid = &setclientid_confirm->sc_clientid;
1815 if (STALE_CLIENTID(clid))
1816 return nfserr_stale_clientid;
1818 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1819 * We get here on a DRC miss.
1824 conf = find_confirmed_client(clid);
1825 unconf = find_unconfirmed_client(clid);
1827 status = nfserr_clid_inuse;
1828 if (conf && !rpc_cmp_addr((struct sockaddr *) &conf->cl_addr, sa))
1830 if (unconf && !rpc_cmp_addr((struct sockaddr *) &unconf->cl_addr, sa))
1834 * section 14.2.34 of RFC 3530 has a description of
1835 * SETCLIENTID_CONFIRM request processing consisting
1836 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1838 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
1840 * RFC 3530 14.2.34 CASE 1:
1843 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1844 status = nfserr_clid_inuse;
1846 atomic_set(&conf->cl_cb_set, 0);
1847 nfsd4_probe_callback(conf, &unconf->cl_cb_conn);
1848 expire_client(unconf);
1852 } else if (conf && !unconf) {
1854 * RFC 3530 14.2.34 CASE 2:
1855 * probable retransmitted request; play it safe and
1858 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1859 status = nfserr_clid_inuse;
1862 } else if (!conf && unconf
1863 && same_verf(&unconf->cl_confirm, &confirm)) {
1865 * RFC 3530 14.2.34 CASE 3:
1866 * Normal case; new or rebooted client:
1868 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1869 status = nfserr_clid_inuse;
1872 clientstr_hashval(unconf->cl_recdir);
1873 conf = find_confirmed_client_by_str(unconf->cl_recdir,
1876 nfsd4_remove_clid_dir(conf);
1877 expire_client(conf);
1879 move_to_confirmed(unconf);
1881 nfsd4_probe_callback(conf, &conf->cl_cb_conn);
1884 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1885 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
1888 * RFC 3530 14.2.34 CASE 4:
1889 * Client probably hasn't noticed that we rebooted yet.
1891 status = nfserr_stale_clientid;
1893 /* check that we have hit one of the cases...*/
1894 status = nfserr_clid_inuse;
1897 nfs4_unlock_state();
1901 /* OPEN Share state helper functions */
1902 static inline struct nfs4_file *
1903 alloc_init_file(struct inode *ino)
1905 struct nfs4_file *fp;
1906 unsigned int hashval = file_hashval(ino);
1908 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1910 atomic_set(&fp->fi_ref, 1);
1911 INIT_LIST_HEAD(&fp->fi_hash);
1912 INIT_LIST_HEAD(&fp->fi_stateids);
1913 INIT_LIST_HEAD(&fp->fi_delegations);
1914 fp->fi_inode = igrab(ino);
1915 fp->fi_id = current_fileid++;
1916 fp->fi_had_conflict = false;
1917 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
1918 memset(fp->fi_access, 0, sizeof(fp->fi_access));
1919 spin_lock(&recall_lock);
1920 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
1921 spin_unlock(&recall_lock);
1928 nfsd4_free_slab(struct kmem_cache **slab)
1932 kmem_cache_destroy(*slab);
1937 nfsd4_free_slabs(void)
1939 nfsd4_free_slab(&stateowner_slab);
1940 nfsd4_free_slab(&file_slab);
1941 nfsd4_free_slab(&stateid_slab);
1942 nfsd4_free_slab(&deleg_slab);
1946 nfsd4_init_slabs(void)
1948 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
1949 sizeof(struct nfs4_stateowner), 0, 0, NULL);
1950 if (stateowner_slab == NULL)
1952 file_slab = kmem_cache_create("nfsd4_files",
1953 sizeof(struct nfs4_file), 0, 0, NULL);
1954 if (file_slab == NULL)
1956 stateid_slab = kmem_cache_create("nfsd4_stateids",
1957 sizeof(struct nfs4_stateid), 0, 0, NULL);
1958 if (stateid_slab == NULL)
1960 deleg_slab = kmem_cache_create("nfsd4_delegations",
1961 sizeof(struct nfs4_delegation), 0, 0, NULL);
1962 if (deleg_slab == NULL)
1967 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1972 nfs4_free_stateowner(struct kref *kref)
1974 struct nfs4_stateowner *sop =
1975 container_of(kref, struct nfs4_stateowner, so_ref);
1976 kfree(sop->so_owner.data);
1977 kmem_cache_free(stateowner_slab, sop);
1980 static inline struct nfs4_stateowner *
1981 alloc_stateowner(struct xdr_netobj *owner)
1983 struct nfs4_stateowner *sop;
1985 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1986 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1987 memcpy(sop->so_owner.data, owner->data, owner->len);
1988 sop->so_owner.len = owner->len;
1989 kref_init(&sop->so_ref);
1992 kmem_cache_free(stateowner_slab, sop);
1997 static struct nfs4_stateowner *
1998 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1999 struct nfs4_stateowner *sop;
2000 struct nfs4_replay *rp;
2001 unsigned int idhashval;
2003 if (!(sop = alloc_stateowner(&open->op_owner)))
2005 idhashval = ownerid_hashval(current_ownerid);
2006 INIT_LIST_HEAD(&sop->so_idhash);
2007 INIT_LIST_HEAD(&sop->so_strhash);
2008 INIT_LIST_HEAD(&sop->so_perclient);
2009 INIT_LIST_HEAD(&sop->so_stateids);
2010 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
2011 INIT_LIST_HEAD(&sop->so_close_lru);
2013 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
2014 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
2015 list_add(&sop->so_perclient, &clp->cl_openowners);
2016 sop->so_is_open_owner = 1;
2017 sop->so_id = current_ownerid++;
2018 sop->so_client = clp;
2019 sop->so_seqid = open->op_seqid;
2020 sop->so_confirmed = 0;
2021 rp = &sop->so_replay;
2022 rp->rp_status = nfserr_serverfault;
2024 rp->rp_buf = rp->rp_ibuf;
2029 init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
2030 struct nfs4_stateowner *sop = open->op_stateowner;
2031 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
2033 INIT_LIST_HEAD(&stp->st_hash);
2034 INIT_LIST_HEAD(&stp->st_perstateowner);
2035 INIT_LIST_HEAD(&stp->st_lockowners);
2036 INIT_LIST_HEAD(&stp->st_perfile);
2037 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
2038 list_add(&stp->st_perstateowner, &sop->so_stateids);
2039 list_add(&stp->st_perfile, &fp->fi_stateids);
2040 stp->st_stateowner = sop;
2043 stp->st_stateid.si_boot = boot_time;
2044 stp->st_stateid.si_stateownerid = sop->so_id;
2045 stp->st_stateid.si_fileid = fp->fi_id;
2046 stp->st_stateid.si_generation = 0;
2047 stp->st_access_bmap = 0;
2048 stp->st_deny_bmap = 0;
2049 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
2050 &stp->st_access_bmap);
2051 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
2052 stp->st_openstp = NULL;
2056 move_to_close_lru(struct nfs4_stateowner *sop)
2058 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
2060 list_move_tail(&sop->so_close_lru, &close_lru);
2061 sop->so_time = get_seconds();
2065 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
2068 return (sop->so_owner.len == owner->len) &&
2069 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
2070 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
2073 static struct nfs4_stateowner *
2074 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
2076 struct nfs4_stateowner *so = NULL;
2078 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
2079 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
2085 /* search file_hashtbl[] for file */
2086 static struct nfs4_file *
2087 find_file(struct inode *ino)
2089 unsigned int hashval = file_hashval(ino);
2090 struct nfs4_file *fp;
2092 spin_lock(&recall_lock);
2093 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
2094 if (fp->fi_inode == ino) {
2096 spin_unlock(&recall_lock);
2100 spin_unlock(&recall_lock);
2104 static inline int access_valid(u32 x, u32 minorversion)
2106 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
2108 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
2110 x &= ~NFS4_SHARE_ACCESS_MASK;
2111 if (minorversion && x) {
2112 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
2114 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
2116 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
2123 static inline int deny_valid(u32 x)
2125 /* Note: unlike access bits, deny bits may be zero. */
2126 return x <= NFS4_SHARE_DENY_BOTH;
2130 * Called to check deny when READ with all zero stateid or
2131 * WRITE with all zero or all one stateid
2134 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2136 struct inode *ino = current_fh->fh_dentry->d_inode;
2137 struct nfs4_file *fp;
2138 struct nfs4_stateid *stp;
2141 dprintk("NFSD: nfs4_share_conflict\n");
2143 fp = find_file(ino);
2146 ret = nfserr_locked;
2147 /* Search for conflicting share reservations */
2148 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2149 if (test_bit(deny_type, &stp->st_deny_bmap) ||
2150 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
2160 nfs4_file_downgrade(struct nfs4_file *fp, unsigned int share_access)
2162 if (share_access & NFS4_SHARE_ACCESS_WRITE)
2163 nfs4_file_put_access(fp, O_WRONLY);
2164 if (share_access & NFS4_SHARE_ACCESS_READ)
2165 nfs4_file_put_access(fp, O_RDONLY);
2169 * Spawn a thread to perform a recall on the delegation represented
2170 * by the lease (file_lock)
2172 * Called from break_lease() with lock_kernel() held.
2173 * Note: we assume break_lease will only call this *once* for any given
2177 void nfsd_break_deleg_cb(struct file_lock *fl)
2179 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2181 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2185 /* We're assuming the state code never drops its reference
2186 * without first removing the lease. Since we're in this lease
2187 * callback (and since the lease code is serialized by the kernel
2188 * lock) we know the server hasn't removed the lease yet, we know
2189 * it's safe to take a reference: */
2190 atomic_inc(&dp->dl_count);
2192 spin_lock(&recall_lock);
2193 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2194 spin_unlock(&recall_lock);
2196 /* only place dl_time is set. protected by lock_kernel*/
2197 dp->dl_time = get_seconds();
2200 * We don't want the locks code to timeout the lease for us;
2201 * we'll remove it ourself if the delegation isn't returned
2204 fl->fl_break_time = 0;
2206 dp->dl_file->fi_had_conflict = true;
2207 nfsd4_cb_recall(dp);
2211 * The file_lock is being reapd.
2213 * Called by locks_free_lock() with lock_kernel() held.
2216 void nfsd_release_deleg_cb(struct file_lock *fl)
2218 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2220 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2222 if (!(fl->fl_flags & FL_LEASE) || !dp)
2224 dp->dl_flock = NULL;
2228 * Set the delegation file_lock back pointer.
2230 * Called from setlease() with lock_kernel() held.
2233 void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2235 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2237 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2244 * Called from setlease() with lock_kernel() held
2247 int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2249 struct nfs4_delegation *onlistd =
2250 (struct nfs4_delegation *)onlist->fl_owner;
2251 struct nfs4_delegation *tryd =
2252 (struct nfs4_delegation *)try->fl_owner;
2254 if (onlist->fl_lmops != try->fl_lmops)
2257 return onlistd->dl_client == tryd->dl_client;
2262 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2265 return lease_modify(onlist, arg);
2270 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2271 .fl_break = nfsd_break_deleg_cb,
2272 .fl_release_private = nfsd_release_deleg_cb,
2273 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2274 .fl_mylease = nfsd_same_client_deleg_cb,
2275 .fl_change = nfsd_change_deleg_cb,
2280 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2281 struct nfsd4_open *open)
2283 clientid_t *clientid = &open->op_clientid;
2284 struct nfs4_client *clp = NULL;
2285 unsigned int strhashval;
2286 struct nfs4_stateowner *sop = NULL;
2288 if (!check_name(open->op_owner))
2289 return nfserr_inval;
2291 if (STALE_CLIENTID(&open->op_clientid))
2292 return nfserr_stale_clientid;
2294 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2295 sop = find_openstateowner_str(strhashval, open);
2296 open->op_stateowner = sop;
2298 /* Make sure the client's lease hasn't expired. */
2299 clp = find_confirmed_client(clientid);
2301 return nfserr_expired;
2304 /* When sessions are used, skip open sequenceid processing */
2305 if (nfsd4_has_session(cstate))
2307 if (!sop->so_confirmed) {
2308 /* Replace unconfirmed owners without checking for replay. */
2309 clp = sop->so_client;
2310 release_openowner(sop);
2311 open->op_stateowner = NULL;
2314 if (open->op_seqid == sop->so_seqid - 1) {
2315 if (sop->so_replay.rp_buflen)
2316 return nfserr_replay_me;
2317 /* The original OPEN failed so spectacularly
2318 * that we don't even have replay data saved!
2319 * Therefore, we have no choice but to continue
2320 * processing this OPEN; presumably, we'll
2321 * fail again for the same reason.
2323 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2326 if (open->op_seqid != sop->so_seqid)
2327 return nfserr_bad_seqid;
2329 if (open->op_stateowner == NULL) {
2330 sop = alloc_init_open_stateowner(strhashval, clp, open);
2332 return nfserr_resource;
2333 open->op_stateowner = sop;
2335 list_del_init(&sop->so_close_lru);
2336 renew_client(sop->so_client);
2340 static inline __be32
2341 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2343 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2344 return nfserr_openmode;
2349 static struct nfs4_delegation *
2350 find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2352 struct nfs4_delegation *dp;
2354 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
2355 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2361 int share_access_to_flags(u32 share_access)
2363 share_access &= ~NFS4_SHARE_WANT_MASK;
2365 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
2369 nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2370 struct nfs4_delegation **dp)
2373 __be32 status = nfserr_bad_stateid;
2375 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2378 flags = share_access_to_flags(open->op_share_access);
2379 status = nfs4_check_delegmode(*dp, flags);
2383 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2387 open->op_stateowner->so_confirmed = 1;
2392 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2394 struct nfs4_stateid *local;
2395 __be32 status = nfserr_share_denied;
2396 struct nfs4_stateowner *sop = open->op_stateowner;
2398 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2399 /* ignore lock owners */
2400 if (local->st_stateowner->so_is_open_owner == 0)
2402 /* remember if we have seen this open owner */
2403 if (local->st_stateowner == sop)
2405 /* check for conflicting share reservations */
2406 if (!test_share(local, open))
2414 static inline struct nfs4_stateid *
2415 nfs4_alloc_stateid(void)
2417 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2420 static inline int nfs4_access_to_access(u32 nfs4_access)
2424 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
2425 flags |= NFSD_MAY_READ;
2426 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
2427 flags |= NFSD_MAY_WRITE;
2431 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file
2432 *fp, struct svc_fh *cur_fh, u32 nfs4_access)
2435 int oflag = nfs4_access_to_omode(nfs4_access);
2436 int access = nfs4_access_to_access(nfs4_access);
2438 if (!fp->fi_fds[oflag]) {
2439 status = nfsd_open(rqstp, cur_fh, S_IFREG, access,
2440 &fp->fi_fds[oflag]);
2441 if (status == nfserr_dropit)
2442 status = nfserr_jukebox;
2446 nfs4_file_get_access(fp, oflag);
2452 nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
2453 struct nfs4_file *fp, struct svc_fh *cur_fh,
2454 struct nfsd4_open *open)
2456 struct nfs4_stateid *stp;
2459 stp = nfs4_alloc_stateid();
2461 return nfserr_resource;
2463 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open->op_share_access);
2465 kmem_cache_free(stateid_slab, stp);
2472 static inline __be32
2473 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2474 struct nfsd4_open *open)
2476 struct iattr iattr = {
2477 .ia_valid = ATTR_SIZE,
2480 if (!open->op_truncate)
2482 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2483 return nfserr_inval;
2484 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2488 nfs4_upgrade_open(struct svc_rqst *rqstp, struct nfs4_file *fp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2490 u32 op_share_access = open->op_share_access & ~NFS4_SHARE_WANT_MASK;
2494 new_access = !test_bit(op_share_access, &stp->st_access_bmap);
2496 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, op_share_access);
2500 status = nfsd4_truncate(rqstp, cur_fh, open);
2503 int oflag = nfs4_access_to_omode(new_access);
2504 nfs4_file_put_access(fp, oflag);
2508 /* remember the open */
2509 __set_bit(op_share_access, &stp->st_access_bmap);
2510 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
2517 nfs4_set_claim_prev(struct nfsd4_open *open)
2519 open->op_stateowner->so_confirmed = 1;
2520 open->op_stateowner->so_client->cl_firststate = 1;
2524 * Attempt to hand out a delegation.
2527 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2529 struct nfs4_delegation *dp;
2530 struct nfs4_stateowner *sop = stp->st_stateowner;
2531 int cb_up = atomic_read(&sop->so_client->cl_cb_set);
2532 struct file_lock fl, *flp = &fl;
2533 int status, flag = 0;
2535 flag = NFS4_OPEN_DELEGATE_NONE;
2536 open->op_recall = 0;
2537 switch (open->op_claim_type) {
2538 case NFS4_OPEN_CLAIM_PREVIOUS:
2540 open->op_recall = 1;
2541 flag = open->op_delegate_type;
2542 if (flag == NFS4_OPEN_DELEGATE_NONE)
2545 case NFS4_OPEN_CLAIM_NULL:
2546 /* Let's not give out any delegations till everyone's
2547 * had the chance to reclaim theirs.... */
2548 if (locks_in_grace())
2550 if (!cb_up || !sop->so_confirmed)
2552 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2553 flag = NFS4_OPEN_DELEGATE_WRITE;
2555 flag = NFS4_OPEN_DELEGATE_READ;
2561 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2563 flag = NFS4_OPEN_DELEGATE_NONE;
2566 locks_init_lock(&fl);
2567 fl.fl_lmops = &nfsd_lease_mng_ops;
2568 fl.fl_flags = FL_LEASE;
2569 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2570 fl.fl_end = OFFSET_MAX;
2571 fl.fl_owner = (fl_owner_t)dp;
2572 fl.fl_file = find_readable_file(stp->st_file);
2573 BUG_ON(!fl.fl_file);
2574 fl.fl_pid = current->tgid;
2576 /* vfs_setlease checks to see if delegation should be handed out.
2577 * the lock_manager callbacks fl_mylease and fl_change are used
2579 if ((status = vfs_setlease(fl.fl_file, fl.fl_type, &flp))) {
2580 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
2581 unhash_delegation(dp);
2582 flag = NFS4_OPEN_DELEGATE_NONE;
2586 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2588 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2589 STATEID_VAL(&dp->dl_stateid));
2591 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2592 && flag == NFS4_OPEN_DELEGATE_NONE
2593 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2594 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2595 open->op_delegate_type = flag;
2599 * called with nfs4_lock_state() held.
2602 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2604 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2605 struct nfs4_file *fp = NULL;
2606 struct inode *ino = current_fh->fh_dentry->d_inode;
2607 struct nfs4_stateid *stp = NULL;
2608 struct nfs4_delegation *dp = NULL;
2611 status = nfserr_inval;
2612 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
2613 || !deny_valid(open->op_share_deny))
2616 * Lookup file; if found, lookup stateid and check open request,
2617 * and check for delegations in the process of being recalled.
2618 * If not found, create the nfs4_file struct
2620 fp = find_file(ino);
2622 if ((status = nfs4_check_open(fp, open, &stp)))
2624 status = nfs4_check_deleg(fp, open, &dp);
2628 status = nfserr_bad_stateid;
2629 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2631 status = nfserr_resource;
2632 fp = alloc_init_file(ino);
2638 * OPEN the file, or upgrade an existing OPEN.
2639 * If truncate fails, the OPEN fails.
2642 /* Stateid was found, this is an OPEN upgrade */
2643 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
2646 update_stateid(&stp->st_stateid);
2648 status = nfs4_new_open(rqstp, &stp, fp, current_fh, open);
2651 init_stateid(stp, fp, open);
2652 status = nfsd4_truncate(rqstp, current_fh, open);
2654 release_open_stateid(stp);
2657 if (nfsd4_has_session(&resp->cstate))
2658 update_stateid(&stp->st_stateid);
2660 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2662 if (nfsd4_has_session(&resp->cstate))
2663 open->op_stateowner->so_confirmed = 1;
2666 * Attempt to hand out a delegation. No error return, because the
2667 * OPEN succeeds even if we fail.
2669 nfs4_open_delegation(current_fh, open, stp);
2673 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
2674 STATEID_VAL(&stp->st_stateid));
2678 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2679 nfs4_set_claim_prev(open);
2681 * To finish the open response, we just need to set the rflags.
2683 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
2684 if (!open->op_stateowner->so_confirmed &&
2685 !nfsd4_has_session(&resp->cstate))
2686 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2692 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2695 struct nfs4_client *clp;
2699 dprintk("process_renew(%08x/%08x): starting\n",
2700 clid->cl_boot, clid->cl_id);
2701 status = nfserr_stale_clientid;
2702 if (STALE_CLIENTID(clid))
2704 clp = find_confirmed_client(clid);
2705 status = nfserr_expired;
2707 /* We assume the client took too long to RENEW. */
2708 dprintk("nfsd4_renew: clientid not found!\n");
2712 status = nfserr_cb_path_down;
2713 if (!list_empty(&clp->cl_delegations)
2714 && !atomic_read(&clp->cl_cb_set))
2718 nfs4_unlock_state();
2722 struct lock_manager nfsd4_manager = {
2726 nfsd4_end_grace(void)
2728 dprintk("NFSD: end of grace period\n");
2729 nfsd4_recdir_purge_old();
2730 locks_end_grace(&nfsd4_manager);
2732 * Now that every NFSv4 client has had the chance to recover and
2733 * to see the (possibly new, possibly shorter) lease time, we
2734 * can safely set the next grace time to the current lease time:
2736 nfsd4_grace = nfsd4_lease;
2740 nfs4_laundromat(void)
2742 struct nfs4_client *clp;
2743 struct nfs4_stateowner *sop;
2744 struct nfs4_delegation *dp;
2745 struct list_head *pos, *next, reaplist;
2746 time_t cutoff = get_seconds() - nfsd4_lease;
2747 time_t t, clientid_val = nfsd4_lease;
2748 time_t u, test_val = nfsd4_lease;
2752 dprintk("NFSD: laundromat service - starting\n");
2753 if (locks_in_grace())
2755 INIT_LIST_HEAD(&reaplist);
2756 spin_lock(&client_lock);
2757 list_for_each_safe(pos, next, &client_lru) {
2758 clp = list_entry(pos, struct nfs4_client, cl_lru);
2759 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2760 t = clp->cl_time - cutoff;
2761 if (clientid_val > t)
2765 if (atomic_read(&clp->cl_refcount)) {
2766 dprintk("NFSD: client in use (clientid %08x)\n",
2767 clp->cl_clientid.cl_id);
2770 unhash_client_locked(clp);
2771 list_add(&clp->cl_lru, &reaplist);
2773 spin_unlock(&client_lock);
2774 list_for_each_safe(pos, next, &reaplist) {
2775 clp = list_entry(pos, struct nfs4_client, cl_lru);
2776 dprintk("NFSD: purging unused client (clientid %08x)\n",
2777 clp->cl_clientid.cl_id);
2778 nfsd4_remove_clid_dir(clp);
2781 spin_lock(&recall_lock);
2782 list_for_each_safe(pos, next, &del_recall_lru) {
2783 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2784 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2785 u = dp->dl_time - cutoff;
2790 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2792 list_move(&dp->dl_recall_lru, &reaplist);
2794 spin_unlock(&recall_lock);
2795 list_for_each_safe(pos, next, &reaplist) {
2796 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2797 list_del_init(&dp->dl_recall_lru);
2798 unhash_delegation(dp);
2800 test_val = nfsd4_lease;
2801 list_for_each_safe(pos, next, &close_lru) {
2802 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2803 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2804 u = sop->so_time - cutoff;
2809 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2811 release_openowner(sop);
2813 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2814 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2815 nfs4_unlock_state();
2816 return clientid_val;
2819 static struct workqueue_struct *laundry_wq;
2820 static void laundromat_main(struct work_struct *);
2821 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2824 laundromat_main(struct work_struct *not_used)
2828 t = nfs4_laundromat();
2829 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
2830 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
2833 static struct nfs4_stateowner *
2834 search_close_lru(u32 st_id, int flags)
2836 struct nfs4_stateowner *local = NULL;
2838 if (flags & CLOSE_STATE) {
2839 list_for_each_entry(local, &close_lru, so_close_lru) {
2840 if (local->so_id == st_id)
2848 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2850 return fhp->fh_dentry->d_inode != stp->st_file->fi_inode;
2854 STALE_STATEID(stateid_t *stateid)
2856 if (stateid->si_boot == boot_time)
2858 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
2859 STATEID_VAL(stateid));
2864 access_permit_read(unsigned long access_bmap)
2866 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2867 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2868 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2872 access_permit_write(unsigned long access_bmap)
2874 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2875 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2879 __be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
2881 __be32 status = nfserr_openmode;
2883 /* For lock stateid's, we test the parent open, not the lock: */
2884 if (stp->st_openstp)
2885 stp = stp->st_openstp;
2886 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2888 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2895 static inline __be32
2896 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2898 if (ONE_STATEID(stateid) && (flags & RD_STATE))
2900 else if (locks_in_grace()) {
2901 /* Answer in remaining cases depends on existance of
2902 * conflicting state; so we must wait out the grace period. */
2903 return nfserr_grace;
2904 } else if (flags & WR_STATE)
2905 return nfs4_share_conflict(current_fh,
2906 NFS4_SHARE_DENY_WRITE);
2907 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2908 return nfs4_share_conflict(current_fh,
2909 NFS4_SHARE_DENY_READ);
2913 * Allow READ/WRITE during grace period on recovered state only for files
2914 * that are not able to provide mandatory locking.
2917 grace_disallows_io(struct inode *inode)
2919 return locks_in_grace() && mandatory_lock(inode);
2922 static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
2925 * When sessions are used the stateid generation number is ignored
2928 if ((flags & HAS_SESSION) && in->si_generation == 0)
2931 /* If the client sends us a stateid from the future, it's buggy: */
2932 if (in->si_generation > ref->si_generation)
2933 return nfserr_bad_stateid;
2935 * The following, however, can happen. For example, if the
2936 * client sends an open and some IO at the same time, the open
2937 * may bump si_generation while the IO is still in flight.
2938 * Thanks to hard links and renames, the client never knows what
2939 * file an open will affect. So it could avoid that situation
2940 * only by serializing all opens and IO from the same open
2941 * owner. To recover from the old_stateid error, the client
2942 * will just have to retry the IO:
2944 if (in->si_generation < ref->si_generation)
2945 return nfserr_old_stateid;
2950 static int is_delegation_stateid(stateid_t *stateid)
2952 return stateid->si_fileid == 0;
2956 * Checks for stateid operations
2959 nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2960 stateid_t *stateid, int flags, struct file **filpp)
2962 struct nfs4_stateid *stp = NULL;
2963 struct nfs4_delegation *dp = NULL;
2964 struct svc_fh *current_fh = &cstate->current_fh;
2965 struct inode *ino = current_fh->fh_dentry->d_inode;
2971 if (grace_disallows_io(ino))
2972 return nfserr_grace;
2974 if (nfsd4_has_session(cstate))
2975 flags |= HAS_SESSION;
2977 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2978 return check_special_stateids(current_fh, stateid, flags);
2980 status = nfserr_stale_stateid;
2981 if (STALE_STATEID(stateid))
2984 status = nfserr_bad_stateid;
2985 if (is_delegation_stateid(stateid)) {
2986 dp = find_delegation_stateid(ino, stateid);
2989 status = check_stateid_generation(stateid, &dp->dl_stateid,
2993 status = nfs4_check_delegmode(dp, flags);
2996 renew_client(dp->dl_client);
2998 *filpp = find_readable_file(dp->dl_file);
3000 } else { /* open or lock stateid */
3001 stp = find_stateid(stateid, flags);
3004 if (nfs4_check_fh(current_fh, stp))
3006 if (!stp->st_stateowner->so_confirmed)
3008 status = check_stateid_generation(stateid, &stp->st_stateid,
3012 status = nfs4_check_openmode(stp, flags);
3015 renew_client(stp->st_stateowner->so_client);
3017 if (flags & RD_STATE)
3018 *filpp = find_readable_file(stp->st_file);
3020 *filpp = find_writeable_file(stp->st_file);
3031 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
3032 RD_STATE : WR_STATE;
3036 * Checks for sequence id mutating operations.
3039 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3040 stateid_t *stateid, int flags,
3041 struct nfs4_stateowner **sopp,
3042 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
3044 struct nfs4_stateid *stp;
3045 struct nfs4_stateowner *sop;
3046 struct svc_fh *current_fh = &cstate->current_fh;
3049 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
3050 seqid, STATEID_VAL(stateid));
3055 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
3056 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3057 return nfserr_bad_stateid;
3060 if (STALE_STATEID(stateid))
3061 return nfserr_stale_stateid;
3063 if (nfsd4_has_session(cstate))
3064 flags |= HAS_SESSION;
3067 * We return BAD_STATEID if filehandle doesn't match stateid,
3068 * the confirmed flag is incorrecly set, or the generation
3069 * number is incorrect.
3071 stp = find_stateid(stateid, flags);
3074 * Also, we should make sure this isn't just the result of
3077 sop = search_close_lru(stateid->si_stateownerid, flags);
3079 return nfserr_bad_stateid;
3085 *sopp = sop = stp->st_stateowner;
3088 clientid_t *lockclid = &lock->v.new.clientid;
3089 struct nfs4_client *clp = sop->so_client;
3093 lkflg = setlkflg(lock->lk_type);
3095 if (lock->lk_is_new) {
3096 if (!sop->so_is_open_owner)
3097 return nfserr_bad_stateid;
3098 if (!(flags & HAS_SESSION) &&
3099 !same_clid(&clp->cl_clientid, lockclid))
3100 return nfserr_bad_stateid;
3101 /* stp is the open stateid */
3102 status = nfs4_check_openmode(stp, lkflg);
3106 /* stp is the lock stateid */
3107 status = nfs4_check_openmode(stp->st_openstp, lkflg);
3113 if (nfs4_check_fh(current_fh, stp)) {
3114 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3115 return nfserr_bad_stateid;
3119 * We now validate the seqid and stateid generation numbers.
3120 * For the moment, we ignore the possibility of
3121 * generation number wraparound.
3123 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
3126 if (sop->so_confirmed && flags & CONFIRM) {
3127 dprintk("NFSD: preprocess_seqid_op: expected"
3128 " unconfirmed stateowner!\n");
3129 return nfserr_bad_stateid;
3131 if (!sop->so_confirmed && !(flags & CONFIRM)) {
3132 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3133 " confirmed yet!\n");
3134 return nfserr_bad_stateid;
3136 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
3139 renew_client(sop->so_client);
3143 if (seqid == sop->so_seqid - 1) {
3144 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
3145 /* indicate replay to calling function */
3146 return nfserr_replay_me;
3148 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3149 sop->so_seqid, seqid);
3151 return nfserr_bad_seqid;
3155 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3156 struct nfsd4_open_confirm *oc)
3159 struct nfs4_stateowner *sop;
3160 struct nfs4_stateid *stp;
3162 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3163 (int)cstate->current_fh.fh_dentry->d_name.len,
3164 cstate->current_fh.fh_dentry->d_name.name);
3166 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3172 if ((status = nfs4_preprocess_seqid_op(cstate,
3173 oc->oc_seqid, &oc->oc_req_stateid,
3174 CONFIRM | OPEN_STATE,
3175 &oc->oc_stateowner, &stp, NULL)))
3178 sop = oc->oc_stateowner;
3179 sop->so_confirmed = 1;
3180 update_stateid(&stp->st_stateid);
3181 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
3182 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3183 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stateid));
3185 nfsd4_create_clid_dir(sop->so_client);
3187 if (oc->oc_stateowner) {
3188 nfs4_get_stateowner(oc->oc_stateowner);
3189 cstate->replay_owner = oc->oc_stateowner;
3191 nfs4_unlock_state();
3197 * unset all bits in union bitmap (bmap) that
3198 * do not exist in share (from successful OPEN_DOWNGRADE)
3201 reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3204 for (i = 1; i < 4; i++) {
3205 if ((i & access) != i)
3206 __clear_bit(i, bmap);
3211 reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3214 for (i = 0; i < 4; i++) {
3215 if ((i & deny) != i)
3216 __clear_bit(i, bmap);
3221 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3222 struct nfsd4_compound_state *cstate,
3223 struct nfsd4_open_downgrade *od)
3226 struct nfs4_stateid *stp;
3227 unsigned int share_access;
3229 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3230 (int)cstate->current_fh.fh_dentry->d_name.len,
3231 cstate->current_fh.fh_dentry->d_name.name);
3233 if (!access_valid(od->od_share_access, cstate->minorversion)
3234 || !deny_valid(od->od_share_deny))
3235 return nfserr_inval;
3238 if ((status = nfs4_preprocess_seqid_op(cstate,
3242 &od->od_stateowner, &stp, NULL)))
3245 status = nfserr_inval;
3246 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3247 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3248 stp->st_access_bmap, od->od_share_access);
3251 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3252 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3253 stp->st_deny_bmap, od->od_share_deny);
3256 set_access(&share_access, stp->st_access_bmap);
3257 nfs4_file_downgrade(stp->st_file, share_access & ~od->od_share_access);
3259 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3260 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3262 update_stateid(&stp->st_stateid);
3263 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3266 if (od->od_stateowner) {
3267 nfs4_get_stateowner(od->od_stateowner);
3268 cstate->replay_owner = od->od_stateowner;
3270 nfs4_unlock_state();
3275 * nfs4_unlock_state() called after encode
3278 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3279 struct nfsd4_close *close)
3282 struct nfs4_stateid *stp;
3284 dprintk("NFSD: nfsd4_close on file %.*s\n",
3285 (int)cstate->current_fh.fh_dentry->d_name.len,
3286 cstate->current_fh.fh_dentry->d_name.name);
3289 /* check close_lru for replay */
3290 if ((status = nfs4_preprocess_seqid_op(cstate,
3293 OPEN_STATE | CLOSE_STATE,
3294 &close->cl_stateowner, &stp, NULL)))
3297 update_stateid(&stp->st_stateid);
3298 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3300 /* release_stateid() calls nfsd_close() if needed */
3301 release_open_stateid(stp);
3303 /* place unused nfs4_stateowners on so_close_lru list to be
3304 * released by the laundromat service after the lease period
3305 * to enable us to handle CLOSE replay
3307 if (list_empty(&close->cl_stateowner->so_stateids))
3308 move_to_close_lru(close->cl_stateowner);
3310 if (close->cl_stateowner) {
3311 nfs4_get_stateowner(close->cl_stateowner);
3312 cstate->replay_owner = close->cl_stateowner;
3314 nfs4_unlock_state();
3319 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3320 struct nfsd4_delegreturn *dr)
3322 struct nfs4_delegation *dp;
3323 stateid_t *stateid = &dr->dr_stateid;
3324 struct inode *inode;
3328 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
3330 inode = cstate->current_fh.fh_dentry->d_inode;
3332 if (nfsd4_has_session(cstate))
3333 flags |= HAS_SESSION;
3335 status = nfserr_bad_stateid;
3336 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3338 status = nfserr_stale_stateid;
3339 if (STALE_STATEID(stateid))
3341 status = nfserr_bad_stateid;
3342 if (!is_delegation_stateid(stateid))
3344 dp = find_delegation_stateid(inode, stateid);
3347 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
3350 renew_client(dp->dl_client);
3352 unhash_delegation(dp);
3354 nfs4_unlock_state();
3361 * Lock owner state (byte-range locks)
3363 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3364 #define LOCK_HASH_BITS 8
3365 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3366 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3369 end_offset(u64 start, u64 len)
3374 return end >= start ? end: NFS4_MAX_UINT64;
3377 /* last octet in a range */
3379 last_byte_offset(u64 start, u64 len)
3385 return end > start ? end - 1: NFS4_MAX_UINT64;
3388 #define lockownerid_hashval(id) \
3389 ((id) & LOCK_HASH_MASK)
3391 static inline unsigned int
3392 lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3393 struct xdr_netobj *ownername)
3395 return (file_hashval(inode) + cl_id
3396 + opaque_hashval(ownername->data, ownername->len))
3400 static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3401 static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3402 static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3404 static struct nfs4_stateid *
3405 find_stateid(stateid_t *stid, int flags)
3407 struct nfs4_stateid *local;
3408 u32 st_id = stid->si_stateownerid;
3409 u32 f_id = stid->si_fileid;
3410 unsigned int hashval;
3412 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
3413 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
3414 hashval = stateid_hashval(st_id, f_id);
3415 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3416 if ((local->st_stateid.si_stateownerid == st_id) &&
3417 (local->st_stateid.si_fileid == f_id))
3422 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
3423 hashval = stateid_hashval(st_id, f_id);
3424 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3425 if ((local->st_stateid.si_stateownerid == st_id) &&
3426 (local->st_stateid.si_fileid == f_id))
3433 static struct nfs4_delegation *
3434 find_delegation_stateid(struct inode *ino, stateid_t *stid)
3436 struct nfs4_file *fp;
3437 struct nfs4_delegation *dl;
3439 dprintk("NFSD: %s: stateid=" STATEID_FMT "\n", __func__,
3442 fp = find_file(ino);
3445 dl = find_delegation_file(fp, stid);
3451 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3452 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3453 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3454 * locking, this prevents us from being completely protocol-compliant. The
3455 * real solution to this problem is to start using unsigned file offsets in
3456 * the VFS, but this is a very deep change!
3459 nfs4_transform_lock_offset(struct file_lock *lock)
3461 if (lock->fl_start < 0)
3462 lock->fl_start = OFFSET_MAX;
3463 if (lock->fl_end < 0)
3464 lock->fl_end = OFFSET_MAX;
3467 /* Hack!: For now, we're defining this just so we can use a pointer to it
3468 * as a unique cookie to identify our (NFSv4's) posix locks. */
3469 static const struct lock_manager_operations nfsd_posix_mng_ops = {
3473 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3475 struct nfs4_stateowner *sop;
3477 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3478 sop = (struct nfs4_stateowner *) fl->fl_owner;
3479 kref_get(&sop->so_ref);
3481 deny->ld_clientid = sop->so_client->cl_clientid;
3483 deny->ld_sop = NULL;
3484 deny->ld_clientid.cl_boot = 0;
3485 deny->ld_clientid.cl_id = 0;
3487 deny->ld_start = fl->fl_start;
3488 deny->ld_length = NFS4_MAX_UINT64;
3489 if (fl->fl_end != NFS4_MAX_UINT64)
3490 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3491 deny->ld_type = NFS4_READ_LT;
3492 if (fl->fl_type != F_RDLCK)
3493 deny->ld_type = NFS4_WRITE_LT;
3496 static struct nfs4_stateowner *
3497 find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3498 struct xdr_netobj *owner)
3500 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3501 struct nfs4_stateowner *op;
3503 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
3504 if (same_owner_str(op, owner, clid))
3511 * Alloc a lock owner structure.
3512 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3515 * strhashval = lock_ownerstr_hashval
3518 static struct nfs4_stateowner *
3519 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3520 struct nfs4_stateowner *sop;
3521 struct nfs4_replay *rp;
3522 unsigned int idhashval;
3524 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3526 idhashval = lockownerid_hashval(current_ownerid);
3527 INIT_LIST_HEAD(&sop->so_idhash);
3528 INIT_LIST_HEAD(&sop->so_strhash);
3529 INIT_LIST_HEAD(&sop->so_perclient);
3530 INIT_LIST_HEAD(&sop->so_stateids);
3531 INIT_LIST_HEAD(&sop->so_perstateid);
3532 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3534 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3535 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
3536 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
3537 sop->so_is_open_owner = 0;
3538 sop->so_id = current_ownerid++;
3539 sop->so_client = clp;
3540 /* It is the openowner seqid that will be incremented in encode in the
3541 * case of new lockowners; so increment the lock seqid manually: */
3542 sop->so_seqid = lock->lk_new_lock_seqid + 1;
3543 sop->so_confirmed = 1;
3544 rp = &sop->so_replay;
3545 rp->rp_status = nfserr_serverfault;
3547 rp->rp_buf = rp->rp_ibuf;
3551 static struct nfs4_stateid *
3552 alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3554 struct nfs4_stateid *stp;
3555 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3557 stp = nfs4_alloc_stateid();
3560 INIT_LIST_HEAD(&stp->st_hash);
3561 INIT_LIST_HEAD(&stp->st_perfile);
3562 INIT_LIST_HEAD(&stp->st_perstateowner);
3563 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
3564 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
3565 list_add(&stp->st_perfile, &fp->fi_stateids);
3566 list_add(&stp->st_perstateowner, &sop->so_stateids);
3567 stp->st_stateowner = sop;
3570 stp->st_stateid.si_boot = boot_time;
3571 stp->st_stateid.si_stateownerid = sop->so_id;
3572 stp->st_stateid.si_fileid = fp->fi_id;
3573 stp->st_stateid.si_generation = 0;
3574 stp->st_deny_bmap = open_stp->st_deny_bmap;
3575 stp->st_openstp = open_stp;
3582 check_lock_length(u64 offset, u64 length)
3584 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
3585 LOFF_OVERFLOW(offset, length)));
3592 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3593 struct nfsd4_lock *lock)
3595 struct nfs4_stateowner *open_sop = NULL;
3596 struct nfs4_stateowner *lock_sop = NULL;
3597 struct nfs4_stateid *lock_stp;
3598 struct nfs4_file *fp;
3599 struct file *filp = NULL;
3600 struct file_lock file_lock;
3601 struct file_lock conflock;
3603 unsigned int strhashval;
3607 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3608 (long long) lock->lk_offset,
3609 (long long) lock->lk_length);
3611 if (check_lock_length(lock->lk_offset, lock->lk_length))
3612 return nfserr_inval;
3614 if ((status = fh_verify(rqstp, &cstate->current_fh,
3615 S_IFREG, NFSD_MAY_LOCK))) {
3616 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3622 if (lock->lk_is_new) {
3624 * Client indicates that this is a new lockowner.
3625 * Use open owner and open stateid to create lock owner and
3628 struct nfs4_stateid *open_stp = NULL;
3630 status = nfserr_stale_clientid;
3631 if (!nfsd4_has_session(cstate) &&
3632 STALE_CLIENTID(&lock->lk_new_clientid))
3635 /* validate and update open stateid and open seqid */
3636 status = nfs4_preprocess_seqid_op(cstate,
3637 lock->lk_new_open_seqid,
3638 &lock->lk_new_open_stateid,
3640 &lock->lk_replay_owner, &open_stp,
3644 open_sop = lock->lk_replay_owner;
3645 /* create lockowner and lock stateid */
3646 fp = open_stp->st_file;
3647 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3648 open_sop->so_client->cl_clientid.cl_id,
3649 &lock->v.new.owner);
3650 /* XXX: Do we need to check for duplicate stateowners on
3651 * the same file, or should they just be allowed (and
3652 * create new stateids)? */
3653 status = nfserr_resource;
3654 lock_sop = alloc_init_lock_stateowner(strhashval,
3655 open_sop->so_client, open_stp, lock);
3656 if (lock_sop == NULL)
3658 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
3659 if (lock_stp == NULL)
3662 /* lock (lock owner + lock stateid) already exists */
3663 status = nfs4_preprocess_seqid_op(cstate,
3664 lock->lk_old_lock_seqid,
3665 &lock->lk_old_lock_stateid,
3667 &lock->lk_replay_owner, &lock_stp, lock);
3670 lock_sop = lock->lk_replay_owner;
3671 fp = lock_stp->st_file;
3673 /* lock->lk_replay_owner and lock_stp have been created or found */
3675 status = nfserr_grace;
3676 if (locks_in_grace() && !lock->lk_reclaim)
3678 status = nfserr_no_grace;
3679 if (!locks_in_grace() && lock->lk_reclaim)
3682 locks_init_lock(&file_lock);
3683 switch (lock->lk_type) {
3686 if (find_readable_file(lock_stp->st_file)) {
3687 nfs4_get_vfs_file(rqstp, fp, &cstate->current_fh, NFS4_SHARE_ACCESS_READ);
3688 filp = find_readable_file(lock_stp->st_file);
3690 file_lock.fl_type = F_RDLCK;
3694 case NFS4_WRITEW_LT:
3695 if (find_writeable_file(lock_stp->st_file)) {
3696 nfs4_get_vfs_file(rqstp, fp, &cstate->current_fh, NFS4_SHARE_ACCESS_WRITE);
3697 filp = find_writeable_file(lock_stp->st_file);
3699 file_lock.fl_type = F_WRLCK;
3703 status = nfserr_inval;
3707 status = nfserr_openmode;
3710 file_lock.fl_owner = (fl_owner_t)lock_sop;
3711 file_lock.fl_pid = current->tgid;
3712 file_lock.fl_file = filp;
3713 file_lock.fl_flags = FL_POSIX;
3714 file_lock.fl_lmops = &nfsd_posix_mng_ops;
3716 file_lock.fl_start = lock->lk_offset;
3717 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
3718 nfs4_transform_lock_offset(&file_lock);
3721 * Try to lock the file in the VFS.
3722 * Note: locks.c uses the BKL to protect the inode's lock list.
3725 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
3727 case 0: /* success! */
3728 update_stateid(&lock_stp->st_stateid);
3729 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3733 case (EAGAIN): /* conflock holds conflicting lock */
3734 status = nfserr_denied;
3735 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3736 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3739 status = nfserr_deadlock;
3742 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
3743 status = nfserr_resource;
3747 if (status && lock->lk_is_new && lock_sop)
3748 release_lockowner(lock_sop);
3749 if (lock->lk_replay_owner) {
3750 nfs4_get_stateowner(lock->lk_replay_owner);
3751 cstate->replay_owner = lock->lk_replay_owner;
3753 nfs4_unlock_state();
3758 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3759 * so we do a temporary open here just to get an open file to pass to
3760 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3763 static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3768 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3771 err = vfs_test_lock(file, lock);
3780 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3781 struct nfsd4_lockt *lockt)
3783 struct inode *inode;
3784 struct file_lock file_lock;
3788 if (locks_in_grace())
3789 return nfserr_grace;
3791 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3792 return nfserr_inval;
3794 lockt->lt_stateowner = NULL;
3797 status = nfserr_stale_clientid;
3798 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
3801 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
3802 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
3803 if (status == nfserr_symlink)
3804 status = nfserr_inval;
3808 inode = cstate->current_fh.fh_dentry->d_inode;
3809 locks_init_lock(&file_lock);
3810 switch (lockt->lt_type) {
3813 file_lock.fl_type = F_RDLCK;
3816 case NFS4_WRITEW_LT:
3817 file_lock.fl_type = F_WRLCK;
3820 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
3821 status = nfserr_inval;
3825 lockt->lt_stateowner = find_lockstateowner_str(inode,
3826 &lockt->lt_clientid, &lockt->lt_owner);
3827 if (lockt->lt_stateowner)
3828 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3829 file_lock.fl_pid = current->tgid;
3830 file_lock.fl_flags = FL_POSIX;
3832 file_lock.fl_start = lockt->lt_offset;
3833 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
3835 nfs4_transform_lock_offset(&file_lock);
3838 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
3840 status = nfserrno(error);
3843 if (file_lock.fl_type != F_UNLCK) {
3844 status = nfserr_denied;
3845 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
3848 nfs4_unlock_state();
3853 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3854 struct nfsd4_locku *locku)
3856 struct nfs4_stateid *stp;
3857 struct file *filp = NULL;
3858 struct file_lock file_lock;
3862 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3863 (long long) locku->lu_offset,
3864 (long long) locku->lu_length);
3866 if (check_lock_length(locku->lu_offset, locku->lu_length))
3867 return nfserr_inval;
3871 if ((status = nfs4_preprocess_seqid_op(cstate,
3875 &locku->lu_stateowner, &stp, NULL)))
3878 filp = find_any_file(stp->st_file);
3880 status = nfserr_lock_range;
3884 locks_init_lock(&file_lock);
3885 file_lock.fl_type = F_UNLCK;
3886 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3887 file_lock.fl_pid = current->tgid;
3888 file_lock.fl_file = filp;
3889 file_lock.fl_flags = FL_POSIX;
3890 file_lock.fl_lmops = &nfsd_posix_mng_ops;
3891 file_lock.fl_start = locku->lu_offset;
3893 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
3894 nfs4_transform_lock_offset(&file_lock);
3897 * Try to unlock the file in the VFS.
3899 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
3901 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
3905 * OK, unlock succeeded; the only thing left to do is update the stateid.
3907 update_stateid(&stp->st_stateid);
3908 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3911 if (locku->lu_stateowner) {
3912 nfs4_get_stateowner(locku->lu_stateowner);
3913 cstate->replay_owner = locku->lu_stateowner;
3915 nfs4_unlock_state();
3919 status = nfserrno(err);
3925 * 1: locks held by lockowner
3926 * 0: no locks held by lockowner
3929 check_for_locks(struct nfs4_file *filp, struct nfs4_stateowner *lowner)
3931 struct file_lock **flpp;
3932 struct inode *inode = filp->fi_inode;
3936 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
3937 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
3948 nfsd4_release_lockowner(struct svc_rqst *rqstp,
3949 struct nfsd4_compound_state *cstate,
3950 struct nfsd4_release_lockowner *rlockowner)
3952 clientid_t *clid = &rlockowner->rl_clientid;
3953 struct nfs4_stateowner *sop;
3954 struct nfs4_stateid *stp;
3955 struct xdr_netobj *owner = &rlockowner->rl_owner;
3956 struct list_head matches;
3960 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3961 clid->cl_boot, clid->cl_id);
3963 /* XXX check for lease expiration */
3965 status = nfserr_stale_clientid;
3966 if (STALE_CLIENTID(clid))
3971 status = nfserr_locks_held;
3972 /* XXX: we're doing a linear search through all the lockowners.
3973 * Yipes! For now we'll just hope clients aren't really using
3974 * release_lockowner much, but eventually we have to fix these
3975 * data structures. */
3976 INIT_LIST_HEAD(&matches);
3977 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3978 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
3979 if (!same_owner_str(sop, owner, clid))
3981 list_for_each_entry(stp, &sop->so_stateids,
3983 if (check_for_locks(stp->st_file, sop))
3985 /* Note: so_perclient unused for lockowners,
3986 * so it's OK to fool with here. */
3987 list_add(&sop->so_perclient, &matches);
3991 /* Clients probably won't expect us to return with some (but not all)
3992 * of the lockowner state released; so don't release any until all
3993 * have been checked. */
3995 while (!list_empty(&matches)) {
3996 sop = list_entry(matches.next, struct nfs4_stateowner,
3998 /* unhash_stateowner deletes so_perclient only
3999 * for openowners. */
4000 list_del(&sop->so_perclient);
4001 release_lockowner(sop);
4004 nfs4_unlock_state();
4008 static inline struct nfs4_client_reclaim *
4011 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
4015 nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
4017 unsigned int strhashval = clientstr_hashval(name);
4018 struct nfs4_client *clp;
4020 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
4025 * failure => all reset bets are off, nfserr_no_grace...
4028 nfs4_client_to_reclaim(const char *name)
4030 unsigned int strhashval;
4031 struct nfs4_client_reclaim *crp = NULL;
4033 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
4034 crp = alloc_reclaim();
4037 strhashval = clientstr_hashval(name);
4038 INIT_LIST_HEAD(&crp->cr_strhash);
4039 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
4040 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
4041 reclaim_str_hashtbl_size++;
4046 nfs4_release_reclaim(void)
4048 struct nfs4_client_reclaim *crp = NULL;
4051 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4052 while (!list_empty(&reclaim_str_hashtbl[i])) {
4053 crp = list_entry(reclaim_str_hashtbl[i].next,
4054 struct nfs4_client_reclaim, cr_strhash);
4055 list_del(&crp->cr_strhash);
4057 reclaim_str_hashtbl_size--;
4060 BUG_ON(reclaim_str_hashtbl_size);
4064 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4065 static struct nfs4_client_reclaim *
4066 nfs4_find_reclaim_client(clientid_t *clid)
4068 unsigned int strhashval;
4069 struct nfs4_client *clp;
4070 struct nfs4_client_reclaim *crp = NULL;
4073 /* find clientid in conf_id_hashtbl */
4074 clp = find_confirmed_client(clid);
4078 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
4079 clp->cl_name.len, clp->cl_name.data,
4082 /* find clp->cl_name in reclaim_str_hashtbl */
4083 strhashval = clientstr_hashval(clp->cl_recdir);
4084 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
4085 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
4093 * Called from OPEN. Look for clientid in reclaim list.
4096 nfs4_check_open_reclaim(clientid_t *clid)
4098 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
4101 /* initialization to perform at module load time: */
4104 nfs4_state_init(void)
4108 status = nfsd4_init_slabs();
4111 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4112 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
4113 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
4114 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
4115 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
4116 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
4118 for (i = 0; i < SESSION_HASH_SIZE; i++)
4119 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
4120 for (i = 0; i < FILE_HASH_SIZE; i++) {
4121 INIT_LIST_HEAD(&file_hashtbl[i]);
4123 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4124 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
4125 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
4127 for (i = 0; i < STATEID_HASH_SIZE; i++) {
4128 INIT_LIST_HEAD(&stateid_hashtbl[i]);
4129 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
4131 for (i = 0; i < LOCK_HASH_SIZE; i++) {
4132 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
4133 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
4135 memset(&onestateid, ~0, sizeof(stateid_t));
4136 INIT_LIST_HEAD(&close_lru);
4137 INIT_LIST_HEAD(&client_lru);
4138 INIT_LIST_HEAD(&del_recall_lru);
4139 reclaim_str_hashtbl_size = 0;
4144 nfsd4_load_reboot_recovery_data(void)
4149 nfsd4_init_recdir(user_recovery_dirname);
4150 status = nfsd4_recdir_load();
4151 nfs4_unlock_state();
4153 printk("NFSD: Failure reading reboot recovery data\n");
4157 * Since the lifetime of a delegation isn't limited to that of an open, a
4158 * client may quite reasonably hang on to a delegation as long as it has
4159 * the inode cached. This becomes an obvious problem the first time a
4160 * client's inode cache approaches the size of the server's total memory.
4162 * For now we avoid this problem by imposing a hard limit on the number
4163 * of delegations, which varies according to the server's memory size.
4166 set_max_delegations(void)
4169 * Allow at most 4 delegations per megabyte of RAM. Quick
4170 * estimates suggest that in the worst case (where every delegation
4171 * is for a different inode), a delegation could take about 1.5K,
4172 * giving a worst case usage of about 6% of memory.
4174 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4177 /* initialization to perform when the nfsd service is started: */
4180 __nfs4_state_start(void)
4184 boot_time = get_seconds();
4185 locks_start_grace(&nfsd4_manager);
4186 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4188 ret = set_callback_cred();
4191 laundry_wq = create_singlethread_workqueue("nfsd4");
4192 if (laundry_wq == NULL)
4194 ret = nfsd4_create_callback_queue();
4196 goto out_free_laundry;
4197 queue_delayed_work(laundry_wq, &laundromat_work, nfsd4_grace * HZ);
4198 set_max_delegations();
4201 destroy_workqueue(laundry_wq);
4206 nfs4_state_start(void)
4208 nfsd4_load_reboot_recovery_data();
4209 return __nfs4_state_start();
4213 __nfs4_state_shutdown(void)
4216 struct nfs4_client *clp = NULL;
4217 struct nfs4_delegation *dp = NULL;
4218 struct list_head *pos, *next, reaplist;
4220 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4221 while (!list_empty(&conf_id_hashtbl[i])) {
4222 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4225 while (!list_empty(&unconf_str_hashtbl[i])) {
4226 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4230 INIT_LIST_HEAD(&reaplist);
4231 spin_lock(&recall_lock);
4232 list_for_each_safe(pos, next, &del_recall_lru) {
4233 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4234 list_move(&dp->dl_recall_lru, &reaplist);
4236 spin_unlock(&recall_lock);
4237 list_for_each_safe(pos, next, &reaplist) {
4238 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4239 list_del_init(&dp->dl_recall_lru);
4240 unhash_delegation(dp);
4243 nfsd4_shutdown_recdir();
4247 nfs4_state_shutdown(void)
4249 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4250 destroy_workqueue(laundry_wq);
4251 locks_end_grace(&nfsd4_manager);
4253 nfs4_release_reclaim();
4254 __nfs4_state_shutdown();
4255 nfs4_unlock_state();
4256 nfsd4_destroy_callback_queue();
4260 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4261 * accessed when nfsd is starting.
4264 nfs4_set_recdir(char *recdir)
4266 strcpy(user_recovery_dirname, recdir);
4270 * Change the NFSv4 recovery directory to recdir.
4273 nfs4_reset_recoverydir(char *recdir)
4278 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
4282 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
4283 nfs4_set_recdir(recdir);
4291 nfs4_recoverydir(void)
4293 return user_recovery_dirname;