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>
37 #include <linux/slab.h>
38 #include <linux/namei.h>
39 #include <linux/swap.h>
40 #include <linux/pagemap.h>
41 #include <linux/ratelimit.h>
42 #include <linux/sunrpc/svcauth_gss.h>
43 #include <linux/sunrpc/addr.h>
44 #include <linux/hash.h>
48 #include "current_stateid.h"
52 #define NFSDDBG_FACILITY NFSDDBG_PROC
54 #define all_ones {{~0,~0},~0}
55 static const stateid_t one_stateid = {
57 .si_opaque = all_ones,
59 static const stateid_t zero_stateid = {
62 static const stateid_t currentstateid = {
66 static u64 current_sessionid = 1;
68 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zero_stateid, sizeof(stateid_t)))
69 #define ONE_STATEID(stateid) (!memcmp((stateid), &one_stateid, sizeof(stateid_t)))
70 #define CURRENT_STATEID(stateid) (!memcmp((stateid), ¤tstateid, sizeof(stateid_t)))
72 /* forward declarations */
73 static bool check_for_locks(struct nfs4_file *fp, struct nfs4_lockowner *lowner);
74 static void nfs4_free_ol_stateid(struct nfs4_stid *stid);
78 /* Currently used for almost all code touching nfsv4 state: */
79 static DEFINE_MUTEX(client_mutex);
82 * Currently used for the del_recall_lru and file hash table. In an
83 * effort to decrease the scope of the client_mutex, this spinlock may
84 * eventually cover more:
86 static DEFINE_SPINLOCK(state_lock);
89 * A waitqueue for all in-progress 4.0 CLOSE operations that are waiting for
90 * the refcount on the open stateid to drop.
92 static DECLARE_WAIT_QUEUE_HEAD(close_wq);
94 static struct kmem_cache *openowner_slab;
95 static struct kmem_cache *lockowner_slab;
96 static struct kmem_cache *file_slab;
97 static struct kmem_cache *stateid_slab;
98 static struct kmem_cache *deleg_slab;
101 nfs4_lock_state(void)
103 mutex_lock(&client_mutex);
106 static void free_session(struct nfsd4_session *);
108 static bool is_session_dead(struct nfsd4_session *ses)
110 return ses->se_flags & NFS4_SESSION_DEAD;
113 static __be32 mark_session_dead_locked(struct nfsd4_session *ses, int ref_held_by_me)
115 if (atomic_read(&ses->se_ref) > ref_held_by_me)
116 return nfserr_jukebox;
117 ses->se_flags |= NFS4_SESSION_DEAD;
122 nfs4_unlock_state(void)
124 mutex_unlock(&client_mutex);
127 static bool is_client_expired(struct nfs4_client *clp)
129 return clp->cl_time == 0;
132 static __be32 mark_client_expired_locked(struct nfs4_client *clp)
134 if (atomic_read(&clp->cl_refcount))
135 return nfserr_jukebox;
140 static __be32 mark_client_expired(struct nfs4_client *clp)
142 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
145 spin_lock(&nn->client_lock);
146 ret = mark_client_expired_locked(clp);
147 spin_unlock(&nn->client_lock);
151 static __be32 get_client_locked(struct nfs4_client *clp)
153 if (is_client_expired(clp))
154 return nfserr_expired;
155 atomic_inc(&clp->cl_refcount);
159 /* must be called under the client_lock */
161 renew_client_locked(struct nfs4_client *clp)
163 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
165 if (is_client_expired(clp)) {
167 printk("%s: client (clientid %08x/%08x) already expired\n",
169 clp->cl_clientid.cl_boot,
170 clp->cl_clientid.cl_id);
174 dprintk("renewing client (clientid %08x/%08x)\n",
175 clp->cl_clientid.cl_boot,
176 clp->cl_clientid.cl_id);
177 list_move_tail(&clp->cl_lru, &nn->client_lru);
178 clp->cl_time = get_seconds();
182 renew_client(struct nfs4_client *clp)
184 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
186 spin_lock(&nn->client_lock);
187 renew_client_locked(clp);
188 spin_unlock(&nn->client_lock);
191 static void put_client_renew_locked(struct nfs4_client *clp)
193 if (!atomic_dec_and_test(&clp->cl_refcount))
195 if (!is_client_expired(clp))
196 renew_client_locked(clp);
199 static void put_client_renew(struct nfs4_client *clp)
201 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
203 if (!atomic_dec_and_lock(&clp->cl_refcount, &nn->client_lock))
205 if (!is_client_expired(clp))
206 renew_client_locked(clp);
207 spin_unlock(&nn->client_lock);
210 static __be32 nfsd4_get_session_locked(struct nfsd4_session *ses)
214 if (is_session_dead(ses))
215 return nfserr_badsession;
216 status = get_client_locked(ses->se_client);
219 atomic_inc(&ses->se_ref);
223 static void nfsd4_put_session_locked(struct nfsd4_session *ses)
225 struct nfs4_client *clp = ses->se_client;
227 if (atomic_dec_and_test(&ses->se_ref) && is_session_dead(ses))
229 put_client_renew_locked(clp);
232 static void nfsd4_put_session(struct nfsd4_session *ses)
234 struct nfs4_client *clp = ses->se_client;
235 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
237 spin_lock(&nn->client_lock);
238 nfsd4_put_session_locked(ses);
239 spin_unlock(&nn->client_lock);
243 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner)
245 return (sop->so_owner.len == owner->len) &&
246 0 == memcmp(sop->so_owner.data, owner->data, owner->len);
249 static struct nfs4_openowner *
250 find_openstateowner_str_locked(unsigned int hashval, struct nfsd4_open *open,
251 struct nfs4_client *clp)
253 struct nfs4_stateowner *so;
255 lockdep_assert_held(&clp->cl_lock);
257 list_for_each_entry(so, &clp->cl_ownerstr_hashtbl[hashval],
259 if (!so->so_is_open_owner)
261 if (same_owner_str(so, &open->op_owner)) {
262 atomic_inc(&so->so_count);
263 return openowner(so);
269 static struct nfs4_openowner *
270 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open,
271 struct nfs4_client *clp)
273 struct nfs4_openowner *oo;
275 spin_lock(&clp->cl_lock);
276 oo = find_openstateowner_str_locked(hashval, open, clp);
277 spin_unlock(&clp->cl_lock);
282 opaque_hashval(const void *ptr, int nbytes)
284 unsigned char *cptr = (unsigned char *) ptr;
294 static void nfsd4_free_file(struct nfs4_file *f)
296 kmem_cache_free(file_slab, f);
300 put_nfs4_file(struct nfs4_file *fi)
302 might_lock(&state_lock);
304 if (atomic_dec_and_lock(&fi->fi_ref, &state_lock)) {
305 hlist_del(&fi->fi_hash);
306 spin_unlock(&state_lock);
312 get_nfs4_file(struct nfs4_file *fi)
314 atomic_inc(&fi->fi_ref);
318 __nfs4_get_fd(struct nfs4_file *f, int oflag)
320 if (f->fi_fds[oflag])
321 return get_file(f->fi_fds[oflag]);
326 find_writeable_file_locked(struct nfs4_file *f)
330 lockdep_assert_held(&f->fi_lock);
332 ret = __nfs4_get_fd(f, O_WRONLY);
334 ret = __nfs4_get_fd(f, O_RDWR);
339 find_writeable_file(struct nfs4_file *f)
343 spin_lock(&f->fi_lock);
344 ret = find_writeable_file_locked(f);
345 spin_unlock(&f->fi_lock);
350 static struct file *find_readable_file_locked(struct nfs4_file *f)
354 lockdep_assert_held(&f->fi_lock);
356 ret = __nfs4_get_fd(f, O_RDONLY);
358 ret = __nfs4_get_fd(f, O_RDWR);
363 find_readable_file(struct nfs4_file *f)
367 spin_lock(&f->fi_lock);
368 ret = find_readable_file_locked(f);
369 spin_unlock(&f->fi_lock);
375 find_any_file(struct nfs4_file *f)
379 spin_lock(&f->fi_lock);
380 ret = __nfs4_get_fd(f, O_RDWR);
382 ret = __nfs4_get_fd(f, O_WRONLY);
384 ret = __nfs4_get_fd(f, O_RDONLY);
386 spin_unlock(&f->fi_lock);
390 static atomic_long_t num_delegations;
391 unsigned long max_delegations;
394 * Open owner state (share locks)
397 /* hash tables for lock and open owners */
398 #define OWNER_HASH_BITS 8
399 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
400 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
402 static unsigned int ownerstr_hashval(struct xdr_netobj *ownername)
406 ret = opaque_hashval(ownername->data, ownername->len);
407 return ret & OWNER_HASH_MASK;
410 /* hash table for nfs4_file */
411 #define FILE_HASH_BITS 8
412 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
414 static unsigned int nfsd_fh_hashval(struct knfsd_fh *fh)
416 return jhash2(fh->fh_base.fh_pad, XDR_QUADLEN(fh->fh_size), 0);
419 static unsigned int file_hashval(struct knfsd_fh *fh)
421 return nfsd_fh_hashval(fh) & (FILE_HASH_SIZE - 1);
424 static bool nfsd_fh_match(struct knfsd_fh *fh1, struct knfsd_fh *fh2)
426 return fh1->fh_size == fh2->fh_size &&
427 !memcmp(fh1->fh_base.fh_pad,
432 static struct hlist_head file_hashtbl[FILE_HASH_SIZE];
435 __nfs4_file_get_access(struct nfs4_file *fp, u32 access)
437 lockdep_assert_held(&fp->fi_lock);
439 if (access & NFS4_SHARE_ACCESS_WRITE)
440 atomic_inc(&fp->fi_access[O_WRONLY]);
441 if (access & NFS4_SHARE_ACCESS_READ)
442 atomic_inc(&fp->fi_access[O_RDONLY]);
446 nfs4_file_get_access(struct nfs4_file *fp, u32 access)
448 lockdep_assert_held(&fp->fi_lock);
450 /* Does this access mode make sense? */
451 if (access & ~NFS4_SHARE_ACCESS_BOTH)
454 /* Does it conflict with a deny mode already set? */
455 if ((access & fp->fi_share_deny) != 0)
456 return nfserr_share_denied;
458 __nfs4_file_get_access(fp, access);
462 static __be32 nfs4_file_check_deny(struct nfs4_file *fp, u32 deny)
464 /* Common case is that there is no deny mode. */
466 /* Does this deny mode make sense? */
467 if (deny & ~NFS4_SHARE_DENY_BOTH)
470 if ((deny & NFS4_SHARE_DENY_READ) &&
471 atomic_read(&fp->fi_access[O_RDONLY]))
472 return nfserr_share_denied;
474 if ((deny & NFS4_SHARE_DENY_WRITE) &&
475 atomic_read(&fp->fi_access[O_WRONLY]))
476 return nfserr_share_denied;
481 static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
483 might_lock(&fp->fi_lock);
485 if (atomic_dec_and_lock(&fp->fi_access[oflag], &fp->fi_lock)) {
486 struct file *f1 = NULL;
487 struct file *f2 = NULL;
489 swap(f1, fp->fi_fds[oflag]);
490 if (atomic_read(&fp->fi_access[1 - oflag]) == 0)
491 swap(f2, fp->fi_fds[O_RDWR]);
492 spin_unlock(&fp->fi_lock);
500 static void nfs4_file_put_access(struct nfs4_file *fp, u32 access)
502 WARN_ON_ONCE(access & ~NFS4_SHARE_ACCESS_BOTH);
504 if (access & NFS4_SHARE_ACCESS_WRITE)
505 __nfs4_file_put_access(fp, O_WRONLY);
506 if (access & NFS4_SHARE_ACCESS_READ)
507 __nfs4_file_put_access(fp, O_RDONLY);
510 static struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl,
511 struct kmem_cache *slab)
513 struct nfs4_stid *stid;
516 stid = kmem_cache_zalloc(slab, GFP_KERNEL);
520 idr_preload(GFP_KERNEL);
521 spin_lock(&cl->cl_lock);
522 new_id = idr_alloc_cyclic(&cl->cl_stateids, stid, 0, 0, GFP_NOWAIT);
523 spin_unlock(&cl->cl_lock);
527 stid->sc_client = cl;
528 stid->sc_stateid.si_opaque.so_id = new_id;
529 stid->sc_stateid.si_opaque.so_clid = cl->cl_clientid;
530 /* Will be incremented before return to client: */
531 atomic_set(&stid->sc_count, 1);
534 * It shouldn't be a problem to reuse an opaque stateid value.
535 * I don't think it is for 4.1. But with 4.0 I worry that, for
536 * example, a stray write retransmission could be accepted by
537 * the server when it should have been rejected. Therefore,
538 * adopt a trick from the sctp code to attempt to maximize the
539 * amount of time until an id is reused, by ensuring they always
540 * "increase" (mod INT_MAX):
544 kmem_cache_free(slab, stid);
548 static struct nfs4_ol_stateid * nfs4_alloc_open_stateid(struct nfs4_client *clp)
550 struct nfs4_stid *stid;
551 struct nfs4_ol_stateid *stp;
553 stid = nfs4_alloc_stid(clp, stateid_slab);
557 stp = openlockstateid(stid);
558 stp->st_stid.sc_free = nfs4_free_ol_stateid;
562 static void nfs4_free_deleg(struct nfs4_stid *stid)
564 kmem_cache_free(deleg_slab, stid);
565 atomic_long_dec(&num_delegations);
569 * When we recall a delegation, we should be careful not to hand it
570 * out again straight away.
571 * To ensure this we keep a pair of bloom filters ('new' and 'old')
572 * in which the filehandles of recalled delegations are "stored".
573 * If a filehandle appear in either filter, a delegation is blocked.
574 * When a delegation is recalled, the filehandle is stored in the "new"
576 * Every 30 seconds we swap the filters and clear the "new" one,
577 * unless both are empty of course.
579 * Each filter is 256 bits. We hash the filehandle to 32bit and use the
580 * low 3 bytes as hash-table indices.
582 * 'blocked_delegations_lock', which is always taken in block_delegations(),
583 * is used to manage concurrent access. Testing does not need the lock
584 * except when swapping the two filters.
586 static DEFINE_SPINLOCK(blocked_delegations_lock);
587 static struct bloom_pair {
588 int entries, old_entries;
590 int new; /* index into 'set' */
591 DECLARE_BITMAP(set[2], 256);
592 } blocked_delegations;
594 static int delegation_blocked(struct knfsd_fh *fh)
597 struct bloom_pair *bd = &blocked_delegations;
599 if (bd->entries == 0)
601 if (seconds_since_boot() - bd->swap_time > 30) {
602 spin_lock(&blocked_delegations_lock);
603 if (seconds_since_boot() - bd->swap_time > 30) {
604 bd->entries -= bd->old_entries;
605 bd->old_entries = bd->entries;
606 memset(bd->set[bd->new], 0,
609 bd->swap_time = seconds_since_boot();
611 spin_unlock(&blocked_delegations_lock);
613 hash = arch_fast_hash(&fh->fh_base, fh->fh_size, 0);
614 if (test_bit(hash&255, bd->set[0]) &&
615 test_bit((hash>>8)&255, bd->set[0]) &&
616 test_bit((hash>>16)&255, bd->set[0]))
619 if (test_bit(hash&255, bd->set[1]) &&
620 test_bit((hash>>8)&255, bd->set[1]) &&
621 test_bit((hash>>16)&255, bd->set[1]))
627 static void block_delegations(struct knfsd_fh *fh)
630 struct bloom_pair *bd = &blocked_delegations;
632 hash = arch_fast_hash(&fh->fh_base, fh->fh_size, 0);
634 spin_lock(&blocked_delegations_lock);
635 __set_bit(hash&255, bd->set[bd->new]);
636 __set_bit((hash>>8)&255, bd->set[bd->new]);
637 __set_bit((hash>>16)&255, bd->set[bd->new]);
638 if (bd->entries == 0)
639 bd->swap_time = seconds_since_boot();
641 spin_unlock(&blocked_delegations_lock);
644 static struct nfs4_delegation *
645 alloc_init_deleg(struct nfs4_client *clp, struct svc_fh *current_fh)
647 struct nfs4_delegation *dp;
650 dprintk("NFSD alloc_init_deleg\n");
651 n = atomic_long_inc_return(&num_delegations);
652 if (n < 0 || n > max_delegations)
654 if (delegation_blocked(¤t_fh->fh_handle))
656 dp = delegstateid(nfs4_alloc_stid(clp, deleg_slab));
660 dp->dl_stid.sc_free = nfs4_free_deleg;
662 * delegation seqid's are never incremented. The 4.1 special
663 * meaning of seqid 0 isn't meaningful, really, but let's avoid
664 * 0 anyway just for consistency and use 1:
666 dp->dl_stid.sc_stateid.si_generation = 1;
667 INIT_LIST_HEAD(&dp->dl_perfile);
668 INIT_LIST_HEAD(&dp->dl_perclnt);
669 INIT_LIST_HEAD(&dp->dl_recall_lru);
670 dp->dl_type = NFS4_OPEN_DELEGATE_READ;
671 INIT_WORK(&dp->dl_recall.cb_work, nfsd4_run_cb_recall);
674 atomic_long_dec(&num_delegations);
679 nfs4_put_stid(struct nfs4_stid *s)
681 struct nfs4_file *fp = s->sc_file;
682 struct nfs4_client *clp = s->sc_client;
684 might_lock(&clp->cl_lock);
686 if (!atomic_dec_and_lock(&s->sc_count, &clp->cl_lock)) {
687 wake_up_all(&close_wq);
690 idr_remove(&clp->cl_stateids, s->sc_stateid.si_opaque.so_id);
691 spin_unlock(&clp->cl_lock);
697 static void nfs4_put_deleg_lease(struct nfs4_file *fp)
699 lockdep_assert_held(&state_lock);
703 if (atomic_dec_and_test(&fp->fi_delegees)) {
704 vfs_setlease(fp->fi_deleg_file, F_UNLCK, &fp->fi_lease);
706 fput(fp->fi_deleg_file);
707 fp->fi_deleg_file = NULL;
711 static void unhash_stid(struct nfs4_stid *s)
717 hash_delegation_locked(struct nfs4_delegation *dp, struct nfs4_file *fp)
719 lockdep_assert_held(&state_lock);
720 lockdep_assert_held(&fp->fi_lock);
722 atomic_inc(&dp->dl_stid.sc_count);
723 dp->dl_stid.sc_type = NFS4_DELEG_STID;
724 list_add(&dp->dl_perfile, &fp->fi_delegations);
725 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
729 unhash_delegation_locked(struct nfs4_delegation *dp)
731 struct nfs4_file *fp = dp->dl_stid.sc_file;
733 lockdep_assert_held(&state_lock);
735 dp->dl_stid.sc_type = NFS4_CLOSED_DELEG_STID;
736 /* Ensure that deleg break won't try to requeue it */
738 spin_lock(&fp->fi_lock);
739 list_del_init(&dp->dl_perclnt);
740 list_del_init(&dp->dl_recall_lru);
741 list_del_init(&dp->dl_perfile);
742 spin_unlock(&fp->fi_lock);
744 nfs4_put_deleg_lease(fp);
747 static void destroy_delegation(struct nfs4_delegation *dp)
749 spin_lock(&state_lock);
750 unhash_delegation_locked(dp);
751 spin_unlock(&state_lock);
752 nfs4_put_stid(&dp->dl_stid);
755 static void revoke_delegation(struct nfs4_delegation *dp)
757 struct nfs4_client *clp = dp->dl_stid.sc_client;
759 WARN_ON(!list_empty(&dp->dl_recall_lru));
761 if (clp->cl_minorversion == 0)
762 nfs4_put_stid(&dp->dl_stid);
764 dp->dl_stid.sc_type = NFS4_REVOKED_DELEG_STID;
765 spin_lock(&clp->cl_lock);
766 list_add(&dp->dl_recall_lru, &clp->cl_revoked);
767 spin_unlock(&clp->cl_lock);
775 static unsigned int clientid_hashval(u32 id)
777 return id & CLIENT_HASH_MASK;
780 static unsigned int clientstr_hashval(const char *name)
782 return opaque_hashval(name, 8) & CLIENT_HASH_MASK;
786 * We store the NONE, READ, WRITE, and BOTH bits separately in the
787 * st_{access,deny}_bmap field of the stateid, in order to track not
788 * only what share bits are currently in force, but also what
789 * combinations of share bits previous opens have used. This allows us
790 * to enforce the recommendation of rfc 3530 14.2.19 that the server
791 * return an error if the client attempt to downgrade to a combination
792 * of share bits not explicable by closing some of its previous opens.
794 * XXX: This enforcement is actually incomplete, since we don't keep
795 * track of access/deny bit combinations; so, e.g., we allow:
797 * OPEN allow read, deny write
798 * OPEN allow both, deny none
799 * DOWNGRADE allow read, deny none
801 * which we should reject.
804 bmap_to_share_mode(unsigned long bmap) {
806 unsigned int access = 0;
808 for (i = 1; i < 4; i++) {
809 if (test_bit(i, &bmap))
815 /* set share access for a given stateid */
817 set_access(u32 access, struct nfs4_ol_stateid *stp)
819 unsigned char mask = 1 << access;
821 WARN_ON_ONCE(access > NFS4_SHARE_ACCESS_BOTH);
822 stp->st_access_bmap |= mask;
825 /* clear share access for a given stateid */
827 clear_access(u32 access, struct nfs4_ol_stateid *stp)
829 unsigned char mask = 1 << access;
831 WARN_ON_ONCE(access > NFS4_SHARE_ACCESS_BOTH);
832 stp->st_access_bmap &= ~mask;
835 /* test whether a given stateid has access */
837 test_access(u32 access, struct nfs4_ol_stateid *stp)
839 unsigned char mask = 1 << access;
841 return (bool)(stp->st_access_bmap & mask);
844 /* set share deny for a given stateid */
846 set_deny(u32 deny, struct nfs4_ol_stateid *stp)
848 unsigned char mask = 1 << deny;
850 WARN_ON_ONCE(deny > NFS4_SHARE_DENY_BOTH);
851 stp->st_deny_bmap |= mask;
854 /* clear share deny for a given stateid */
856 clear_deny(u32 deny, struct nfs4_ol_stateid *stp)
858 unsigned char mask = 1 << deny;
860 WARN_ON_ONCE(deny > NFS4_SHARE_DENY_BOTH);
861 stp->st_deny_bmap &= ~mask;
864 /* test whether a given stateid is denying specific access */
866 test_deny(u32 deny, struct nfs4_ol_stateid *stp)
868 unsigned char mask = 1 << deny;
870 return (bool)(stp->st_deny_bmap & mask);
873 static int nfs4_access_to_omode(u32 access)
875 switch (access & NFS4_SHARE_ACCESS_BOTH) {
876 case NFS4_SHARE_ACCESS_READ:
878 case NFS4_SHARE_ACCESS_WRITE:
880 case NFS4_SHARE_ACCESS_BOTH:
888 * A stateid that had a deny mode associated with it is being released
889 * or downgraded. Recalculate the deny mode on the file.
892 recalculate_deny_mode(struct nfs4_file *fp)
894 struct nfs4_ol_stateid *stp;
896 spin_lock(&fp->fi_lock);
897 fp->fi_share_deny = 0;
898 list_for_each_entry(stp, &fp->fi_stateids, st_perfile)
899 fp->fi_share_deny |= bmap_to_share_mode(stp->st_deny_bmap);
900 spin_unlock(&fp->fi_lock);
904 reset_union_bmap_deny(u32 deny, struct nfs4_ol_stateid *stp)
909 for (i = 1; i < 4; i++) {
910 if ((i & deny) != i) {
916 /* Recalculate per-file deny mode if there was a change */
918 recalculate_deny_mode(stp->st_stid.sc_file);
921 /* release all access and file references for a given stateid */
923 release_all_access(struct nfs4_ol_stateid *stp)
926 struct nfs4_file *fp = stp->st_stid.sc_file;
928 if (fp && stp->st_deny_bmap != 0)
929 recalculate_deny_mode(fp);
931 for (i = 1; i < 4; i++) {
932 if (test_access(i, stp))
933 nfs4_file_put_access(stp->st_stid.sc_file, i);
934 clear_access(i, stp);
938 static void nfs4_put_stateowner(struct nfs4_stateowner *sop)
940 struct nfs4_client *clp = sop->so_client;
942 might_lock(&clp->cl_lock);
944 if (!atomic_dec_and_lock(&sop->so_count, &clp->cl_lock))
946 sop->so_ops->so_unhash(sop);
947 spin_unlock(&clp->cl_lock);
948 kfree(sop->so_owner.data);
949 sop->so_ops->so_free(sop);
952 static void unhash_ol_stateid(struct nfs4_ol_stateid *stp)
954 struct nfs4_file *fp = stp->st_stid.sc_file;
956 lockdep_assert_held(&stp->st_stateowner->so_client->cl_lock);
958 spin_lock(&fp->fi_lock);
959 list_del(&stp->st_perfile);
960 spin_unlock(&fp->fi_lock);
961 list_del(&stp->st_perstateowner);
964 static void nfs4_free_ol_stateid(struct nfs4_stid *stid)
966 struct nfs4_ol_stateid *stp = openlockstateid(stid);
968 release_all_access(stp);
969 if (stp->st_stateowner)
970 nfs4_put_stateowner(stp->st_stateowner);
971 kmem_cache_free(stateid_slab, stid);
974 static void nfs4_free_lock_stateid(struct nfs4_stid *stid)
976 struct nfs4_ol_stateid *stp = openlockstateid(stid);
977 struct nfs4_lockowner *lo = lockowner(stp->st_stateowner);
980 file = find_any_file(stp->st_stid.sc_file);
982 filp_close(file, (fl_owner_t)lo);
983 nfs4_free_ol_stateid(stid);
987 * Put the persistent reference to an already unhashed generic stateid, while
988 * holding the cl_lock. If it's the last reference, then put it onto the
989 * reaplist for later destruction.
991 static void put_ol_stateid_locked(struct nfs4_ol_stateid *stp,
992 struct list_head *reaplist)
994 struct nfs4_stid *s = &stp->st_stid;
995 struct nfs4_client *clp = s->sc_client;
997 lockdep_assert_held(&clp->cl_lock);
999 WARN_ON_ONCE(!list_empty(&stp->st_locks));
1001 if (!atomic_dec_and_test(&s->sc_count)) {
1002 wake_up_all(&close_wq);
1006 idr_remove(&clp->cl_stateids, s->sc_stateid.si_opaque.so_id);
1007 list_add(&stp->st_locks, reaplist);
1010 static void unhash_lock_stateid(struct nfs4_ol_stateid *stp)
1012 struct nfs4_openowner *oo = openowner(stp->st_openstp->st_stateowner);
1014 lockdep_assert_held(&oo->oo_owner.so_client->cl_lock);
1016 list_del_init(&stp->st_locks);
1017 unhash_ol_stateid(stp);
1018 unhash_stid(&stp->st_stid);
1021 static void release_lock_stateid(struct nfs4_ol_stateid *stp)
1023 struct nfs4_openowner *oo = openowner(stp->st_openstp->st_stateowner);
1025 spin_lock(&oo->oo_owner.so_client->cl_lock);
1026 unhash_lock_stateid(stp);
1027 spin_unlock(&oo->oo_owner.so_client->cl_lock);
1028 nfs4_put_stid(&stp->st_stid);
1031 static void unhash_lockowner_locked(struct nfs4_lockowner *lo)
1033 struct nfs4_client *clp = lo->lo_owner.so_client;
1035 lockdep_assert_held(&clp->cl_lock);
1037 list_del_init(&lo->lo_owner.so_strhash);
1041 * Free a list of generic stateids that were collected earlier after being
1045 free_ol_stateid_reaplist(struct list_head *reaplist)
1047 struct nfs4_ol_stateid *stp;
1051 while (!list_empty(reaplist)) {
1052 stp = list_first_entry(reaplist, struct nfs4_ol_stateid,
1054 list_del(&stp->st_locks);
1055 stp->st_stid.sc_free(&stp->st_stid);
1059 static void release_lockowner(struct nfs4_lockowner *lo)
1061 struct nfs4_client *clp = lo->lo_owner.so_client;
1062 struct nfs4_ol_stateid *stp;
1063 struct list_head reaplist;
1065 INIT_LIST_HEAD(&reaplist);
1067 spin_lock(&clp->cl_lock);
1068 unhash_lockowner_locked(lo);
1069 while (!list_empty(&lo->lo_owner.so_stateids)) {
1070 stp = list_first_entry(&lo->lo_owner.so_stateids,
1071 struct nfs4_ol_stateid, st_perstateowner);
1072 unhash_lock_stateid(stp);
1073 put_ol_stateid_locked(stp, &reaplist);
1075 spin_unlock(&clp->cl_lock);
1076 free_ol_stateid_reaplist(&reaplist);
1077 nfs4_put_stateowner(&lo->lo_owner);
1080 static void release_open_stateid_locks(struct nfs4_ol_stateid *open_stp,
1081 struct list_head *reaplist)
1083 struct nfs4_ol_stateid *stp;
1085 while (!list_empty(&open_stp->st_locks)) {
1086 stp = list_entry(open_stp->st_locks.next,
1087 struct nfs4_ol_stateid, st_locks);
1088 unhash_lock_stateid(stp);
1089 put_ol_stateid_locked(stp, reaplist);
1093 static void unhash_open_stateid(struct nfs4_ol_stateid *stp,
1094 struct list_head *reaplist)
1096 lockdep_assert_held(&stp->st_stid.sc_client->cl_lock);
1098 unhash_ol_stateid(stp);
1099 release_open_stateid_locks(stp, reaplist);
1102 static void release_open_stateid(struct nfs4_ol_stateid *stp)
1104 LIST_HEAD(reaplist);
1106 spin_lock(&stp->st_stid.sc_client->cl_lock);
1107 unhash_open_stateid(stp, &reaplist);
1108 put_ol_stateid_locked(stp, &reaplist);
1109 spin_unlock(&stp->st_stid.sc_client->cl_lock);
1110 free_ol_stateid_reaplist(&reaplist);
1113 static void unhash_openowner_locked(struct nfs4_openowner *oo)
1115 struct nfs4_client *clp = oo->oo_owner.so_client;
1117 lockdep_assert_held(&clp->cl_lock);
1119 list_del_init(&oo->oo_owner.so_strhash);
1120 list_del_init(&oo->oo_perclient);
1123 static void release_last_closed_stateid(struct nfs4_openowner *oo)
1125 struct nfs4_ol_stateid *s = oo->oo_last_closed_stid;
1128 list_del_init(&oo->oo_close_lru);
1129 oo->oo_last_closed_stid = NULL;
1130 nfs4_put_stid(&s->st_stid);
1134 static void release_openowner(struct nfs4_openowner *oo)
1136 struct nfs4_ol_stateid *stp;
1137 struct nfs4_client *clp = oo->oo_owner.so_client;
1138 struct list_head reaplist;
1140 INIT_LIST_HEAD(&reaplist);
1142 spin_lock(&clp->cl_lock);
1143 unhash_openowner_locked(oo);
1144 while (!list_empty(&oo->oo_owner.so_stateids)) {
1145 stp = list_first_entry(&oo->oo_owner.so_stateids,
1146 struct nfs4_ol_stateid, st_perstateowner);
1147 unhash_open_stateid(stp, &reaplist);
1148 put_ol_stateid_locked(stp, &reaplist);
1150 spin_unlock(&clp->cl_lock);
1151 free_ol_stateid_reaplist(&reaplist);
1152 release_last_closed_stateid(oo);
1153 nfs4_put_stateowner(&oo->oo_owner);
1157 hash_sessionid(struct nfs4_sessionid *sessionid)
1159 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
1161 return sid->sequence % SESSION_HASH_SIZE;
1166 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
1168 u32 *ptr = (u32 *)(&sessionid->data[0]);
1169 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
1173 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
1179 * Bump the seqid on cstate->replay_owner, and clear replay_owner if it
1180 * won't be used for replay.
1182 void nfsd4_bump_seqid(struct nfsd4_compound_state *cstate, __be32 nfserr)
1184 struct nfs4_stateowner *so = cstate->replay_owner;
1186 if (nfserr == nfserr_replay_me)
1189 if (!seqid_mutating_err(ntohl(nfserr))) {
1190 nfsd4_cstate_clear_replay(cstate);
1195 if (so->so_is_open_owner)
1196 release_last_closed_stateid(openowner(so));
1202 gen_sessionid(struct nfsd4_session *ses)
1204 struct nfs4_client *clp = ses->se_client;
1205 struct nfsd4_sessionid *sid;
1207 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
1208 sid->clientid = clp->cl_clientid;
1209 sid->sequence = current_sessionid++;
1214 * The protocol defines ca_maxresponssize_cached to include the size of
1215 * the rpc header, but all we need to cache is the data starting after
1216 * the end of the initial SEQUENCE operation--the rest we regenerate
1217 * each time. Therefore we can advertise a ca_maxresponssize_cached
1218 * value that is the number of bytes in our cache plus a few additional
1219 * bytes. In order to stay on the safe side, and not promise more than
1220 * we can cache, those additional bytes must be the minimum possible: 24
1221 * bytes of rpc header (xid through accept state, with AUTH_NULL
1222 * verifier), 12 for the compound header (with zero-length tag), and 44
1223 * for the SEQUENCE op response:
1225 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
1228 free_session_slots(struct nfsd4_session *ses)
1232 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
1233 kfree(ses->se_slots[i]);
1237 * We don't actually need to cache the rpc and session headers, so we
1238 * can allocate a little less for each slot:
1240 static inline u32 slot_bytes(struct nfsd4_channel_attrs *ca)
1244 if (ca->maxresp_cached < NFSD_MIN_HDR_SEQ_SZ)
1247 size = ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
1248 return size + sizeof(struct nfsd4_slot);
1252 * XXX: If we run out of reserved DRC memory we could (up to a point)
1253 * re-negotiate active sessions and reduce their slot usage to make
1254 * room for new connections. For now we just fail the create session.
1256 static u32 nfsd4_get_drc_mem(struct nfsd4_channel_attrs *ca)
1258 u32 slotsize = slot_bytes(ca);
1259 u32 num = ca->maxreqs;
1262 spin_lock(&nfsd_drc_lock);
1263 avail = min((unsigned long)NFSD_MAX_MEM_PER_SESSION,
1264 nfsd_drc_max_mem - nfsd_drc_mem_used);
1265 num = min_t(int, num, avail / slotsize);
1266 nfsd_drc_mem_used += num * slotsize;
1267 spin_unlock(&nfsd_drc_lock);
1272 static void nfsd4_put_drc_mem(struct nfsd4_channel_attrs *ca)
1274 int slotsize = slot_bytes(ca);
1276 spin_lock(&nfsd_drc_lock);
1277 nfsd_drc_mem_used -= slotsize * ca->maxreqs;
1278 spin_unlock(&nfsd_drc_lock);
1281 static struct nfsd4_session *alloc_session(struct nfsd4_channel_attrs *fattrs,
1282 struct nfsd4_channel_attrs *battrs)
1284 int numslots = fattrs->maxreqs;
1285 int slotsize = slot_bytes(fattrs);
1286 struct nfsd4_session *new;
1289 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
1290 + sizeof(struct nfsd4_session) > PAGE_SIZE);
1291 mem = numslots * sizeof(struct nfsd4_slot *);
1293 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
1296 /* allocate each struct nfsd4_slot and data cache in one piece */
1297 for (i = 0; i < numslots; i++) {
1298 new->se_slots[i] = kzalloc(slotsize, GFP_KERNEL);
1299 if (!new->se_slots[i])
1303 memcpy(&new->se_fchannel, fattrs, sizeof(struct nfsd4_channel_attrs));
1304 memcpy(&new->se_bchannel, battrs, sizeof(struct nfsd4_channel_attrs));
1309 kfree(new->se_slots[i]);
1314 static void free_conn(struct nfsd4_conn *c)
1316 svc_xprt_put(c->cn_xprt);
1320 static void nfsd4_conn_lost(struct svc_xpt_user *u)
1322 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
1323 struct nfs4_client *clp = c->cn_session->se_client;
1325 spin_lock(&clp->cl_lock);
1326 if (!list_empty(&c->cn_persession)) {
1327 list_del(&c->cn_persession);
1330 nfsd4_probe_callback(clp);
1331 spin_unlock(&clp->cl_lock);
1334 static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
1336 struct nfsd4_conn *conn;
1338 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
1341 svc_xprt_get(rqstp->rq_xprt);
1342 conn->cn_xprt = rqstp->rq_xprt;
1343 conn->cn_flags = flags;
1344 INIT_LIST_HEAD(&conn->cn_xpt_user.list);
1348 static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
1350 conn->cn_session = ses;
1351 list_add(&conn->cn_persession, &ses->se_conns);
1354 static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
1356 struct nfs4_client *clp = ses->se_client;
1358 spin_lock(&clp->cl_lock);
1359 __nfsd4_hash_conn(conn, ses);
1360 spin_unlock(&clp->cl_lock);
1363 static int nfsd4_register_conn(struct nfsd4_conn *conn)
1365 conn->cn_xpt_user.callback = nfsd4_conn_lost;
1366 return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
1369 static void nfsd4_init_conn(struct svc_rqst *rqstp, struct nfsd4_conn *conn, struct nfsd4_session *ses)
1373 nfsd4_hash_conn(conn, ses);
1374 ret = nfsd4_register_conn(conn);
1376 /* oops; xprt is already down: */
1377 nfsd4_conn_lost(&conn->cn_xpt_user);
1378 /* We may have gained or lost a callback channel: */
1379 nfsd4_probe_callback_sync(ses->se_client);
1382 static struct nfsd4_conn *alloc_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_create_session *cses)
1384 u32 dir = NFS4_CDFC4_FORE;
1386 if (cses->flags & SESSION4_BACK_CHAN)
1387 dir |= NFS4_CDFC4_BACK;
1388 return alloc_conn(rqstp, dir);
1391 /* must be called under client_lock */
1392 static void nfsd4_del_conns(struct nfsd4_session *s)
1394 struct nfs4_client *clp = s->se_client;
1395 struct nfsd4_conn *c;
1397 spin_lock(&clp->cl_lock);
1398 while (!list_empty(&s->se_conns)) {
1399 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
1400 list_del_init(&c->cn_persession);
1401 spin_unlock(&clp->cl_lock);
1403 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
1406 spin_lock(&clp->cl_lock);
1408 spin_unlock(&clp->cl_lock);
1411 static void __free_session(struct nfsd4_session *ses)
1413 free_session_slots(ses);
1417 static void free_session(struct nfsd4_session *ses)
1419 struct nfsd_net *nn = net_generic(ses->se_client->net, nfsd_net_id);
1421 lockdep_assert_held(&nn->client_lock);
1422 nfsd4_del_conns(ses);
1423 nfsd4_put_drc_mem(&ses->se_fchannel);
1424 __free_session(ses);
1427 static void init_session(struct svc_rqst *rqstp, struct nfsd4_session *new, struct nfs4_client *clp, struct nfsd4_create_session *cses)
1430 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1432 new->se_client = clp;
1435 INIT_LIST_HEAD(&new->se_conns);
1437 new->se_cb_seq_nr = 1;
1438 new->se_flags = cses->flags;
1439 new->se_cb_prog = cses->callback_prog;
1440 new->se_cb_sec = cses->cb_sec;
1441 atomic_set(&new->se_ref, 0);
1442 idx = hash_sessionid(&new->se_sessionid);
1443 spin_lock(&nn->client_lock);
1444 list_add(&new->se_hash, &nn->sessionid_hashtbl[idx]);
1445 spin_lock(&clp->cl_lock);
1446 list_add(&new->se_perclnt, &clp->cl_sessions);
1447 spin_unlock(&clp->cl_lock);
1448 spin_unlock(&nn->client_lock);
1450 if (cses->flags & SESSION4_BACK_CHAN) {
1451 struct sockaddr *sa = svc_addr(rqstp);
1453 * This is a little silly; with sessions there's no real
1454 * use for the callback address. Use the peer address
1455 * as a reasonable default for now, but consider fixing
1456 * the rpc client not to require an address in the
1459 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
1460 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1464 /* caller must hold client_lock */
1465 static struct nfsd4_session *
1466 __find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net)
1468 struct nfsd4_session *elem;
1470 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1472 dump_sessionid(__func__, sessionid);
1473 idx = hash_sessionid(sessionid);
1474 /* Search in the appropriate list */
1475 list_for_each_entry(elem, &nn->sessionid_hashtbl[idx], se_hash) {
1476 if (!memcmp(elem->se_sessionid.data, sessionid->data,
1477 NFS4_MAX_SESSIONID_LEN)) {
1482 dprintk("%s: session not found\n", __func__);
1486 static struct nfsd4_session *
1487 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net,
1490 struct nfsd4_session *session;
1491 __be32 status = nfserr_badsession;
1493 session = __find_in_sessionid_hashtbl(sessionid, net);
1496 status = nfsd4_get_session_locked(session);
1504 /* caller must hold client_lock */
1506 unhash_session(struct nfsd4_session *ses)
1508 list_del(&ses->se_hash);
1509 spin_lock(&ses->se_client->cl_lock);
1510 list_del(&ses->se_perclnt);
1511 spin_unlock(&ses->se_client->cl_lock);
1514 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1516 STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
1518 if (clid->cl_boot == nn->boot_time)
1520 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1521 clid->cl_boot, clid->cl_id, nn->boot_time);
1526 * XXX Should we use a slab cache ?
1527 * This type of memory management is somewhat inefficient, but we use it
1528 * anyway since SETCLIENTID is not a common operation.
1530 static struct nfs4_client *alloc_client(struct xdr_netobj name)
1532 struct nfs4_client *clp;
1535 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
1538 clp->cl_name.data = kmemdup(name.data, name.len, GFP_KERNEL);
1539 if (clp->cl_name.data == NULL)
1541 clp->cl_ownerstr_hashtbl = kmalloc(sizeof(struct list_head) *
1542 OWNER_HASH_SIZE, GFP_KERNEL);
1543 if (!clp->cl_ownerstr_hashtbl)
1544 goto err_no_hashtbl;
1545 for (i = 0; i < OWNER_HASH_SIZE; i++)
1546 INIT_LIST_HEAD(&clp->cl_ownerstr_hashtbl[i]);
1547 clp->cl_name.len = name.len;
1548 INIT_LIST_HEAD(&clp->cl_sessions);
1549 idr_init(&clp->cl_stateids);
1550 atomic_set(&clp->cl_refcount, 0);
1551 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1552 INIT_LIST_HEAD(&clp->cl_idhash);
1553 INIT_LIST_HEAD(&clp->cl_openowners);
1554 INIT_LIST_HEAD(&clp->cl_delegations);
1555 INIT_LIST_HEAD(&clp->cl_lru);
1556 INIT_LIST_HEAD(&clp->cl_callbacks);
1557 INIT_LIST_HEAD(&clp->cl_revoked);
1558 spin_lock_init(&clp->cl_lock);
1559 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1562 kfree(clp->cl_name.data);
1569 free_client(struct nfs4_client *clp)
1571 struct nfsd_net __maybe_unused *nn = net_generic(clp->net, nfsd_net_id);
1573 lockdep_assert_held(&nn->client_lock);
1574 while (!list_empty(&clp->cl_sessions)) {
1575 struct nfsd4_session *ses;
1576 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
1578 list_del(&ses->se_perclnt);
1579 WARN_ON_ONCE(atomic_read(&ses->se_ref));
1582 rpc_destroy_wait_queue(&clp->cl_cb_waitq);
1583 free_svc_cred(&clp->cl_cred);
1584 kfree(clp->cl_ownerstr_hashtbl);
1585 kfree(clp->cl_name.data);
1586 idr_destroy(&clp->cl_stateids);
1590 /* must be called under the client_lock */
1592 unhash_client_locked(struct nfs4_client *clp)
1594 struct nfsd4_session *ses;
1596 list_del(&clp->cl_lru);
1597 spin_lock(&clp->cl_lock);
1598 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1599 list_del_init(&ses->se_hash);
1600 spin_unlock(&clp->cl_lock);
1604 destroy_client(struct nfs4_client *clp)
1606 struct nfs4_openowner *oo;
1607 struct nfs4_delegation *dp;
1608 struct list_head reaplist;
1609 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1611 INIT_LIST_HEAD(&reaplist);
1612 spin_lock(&state_lock);
1613 while (!list_empty(&clp->cl_delegations)) {
1614 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1615 unhash_delegation_locked(dp);
1616 list_add(&dp->dl_recall_lru, &reaplist);
1618 spin_unlock(&state_lock);
1619 while (!list_empty(&reaplist)) {
1620 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1621 list_del_init(&dp->dl_recall_lru);
1622 nfs4_put_stid(&dp->dl_stid);
1624 while (!list_empty(&clp->cl_revoked)) {
1625 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1626 list_del_init(&dp->dl_recall_lru);
1627 nfs4_put_stid(&dp->dl_stid);
1629 while (!list_empty(&clp->cl_openowners)) {
1630 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1631 atomic_inc(&oo->oo_owner.so_count);
1632 release_openowner(oo);
1634 nfsd4_shutdown_callback(clp);
1635 if (clp->cl_cb_conn.cb_xprt)
1636 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1637 list_del(&clp->cl_idhash);
1638 if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
1639 rb_erase(&clp->cl_namenode, &nn->conf_name_tree);
1641 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1642 spin_lock(&nn->client_lock);
1643 unhash_client_locked(clp);
1644 WARN_ON_ONCE(atomic_read(&clp->cl_refcount));
1646 spin_unlock(&nn->client_lock);
1649 static void expire_client(struct nfs4_client *clp)
1651 nfsd4_client_record_remove(clp);
1652 destroy_client(clp);
1655 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1657 memcpy(target->cl_verifier.data, source->data,
1658 sizeof(target->cl_verifier.data));
1661 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1663 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1664 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1667 static int copy_cred(struct svc_cred *target, struct svc_cred *source)
1669 if (source->cr_principal) {
1670 target->cr_principal =
1671 kstrdup(source->cr_principal, GFP_KERNEL);
1672 if (target->cr_principal == NULL)
1675 target->cr_principal = NULL;
1676 target->cr_flavor = source->cr_flavor;
1677 target->cr_uid = source->cr_uid;
1678 target->cr_gid = source->cr_gid;
1679 target->cr_group_info = source->cr_group_info;
1680 get_group_info(target->cr_group_info);
1681 target->cr_gss_mech = source->cr_gss_mech;
1682 if (source->cr_gss_mech)
1683 gss_mech_get(source->cr_gss_mech);
1688 compare_blob(const struct xdr_netobj *o1, const struct xdr_netobj *o2)
1692 res = o1->len - o2->len;
1695 return (long long)memcmp(o1->data, o2->data, o1->len);
1698 static int same_name(const char *n1, const char *n2)
1700 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1704 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1706 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1710 same_clid(clientid_t *cl1, clientid_t *cl2)
1712 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1715 static bool groups_equal(struct group_info *g1, struct group_info *g2)
1719 if (g1->ngroups != g2->ngroups)
1721 for (i=0; i<g1->ngroups; i++)
1722 if (!gid_eq(GROUP_AT(g1, i), GROUP_AT(g2, i)))
1728 * RFC 3530 language requires clid_inuse be returned when the
1729 * "principal" associated with a requests differs from that previously
1730 * used. We use uid, gid's, and gss principal string as our best
1731 * approximation. We also don't want to allow non-gss use of a client
1732 * established using gss: in theory cr_principal should catch that
1733 * change, but in practice cr_principal can be null even in the gss case
1734 * since gssd doesn't always pass down a principal string.
1736 static bool is_gss_cred(struct svc_cred *cr)
1738 /* Is cr_flavor one of the gss "pseudoflavors"?: */
1739 return (cr->cr_flavor > RPC_AUTH_MAXFLAVOR);
1744 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1746 if ((is_gss_cred(cr1) != is_gss_cred(cr2))
1747 || (!uid_eq(cr1->cr_uid, cr2->cr_uid))
1748 || (!gid_eq(cr1->cr_gid, cr2->cr_gid))
1749 || !groups_equal(cr1->cr_group_info, cr2->cr_group_info))
1751 if (cr1->cr_principal == cr2->cr_principal)
1753 if (!cr1->cr_principal || !cr2->cr_principal)
1755 return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
1758 static bool svc_rqst_integrity_protected(struct svc_rqst *rqstp)
1760 struct svc_cred *cr = &rqstp->rq_cred;
1763 if (!cr->cr_gss_mech)
1765 service = gss_pseudoflavor_to_service(cr->cr_gss_mech, cr->cr_flavor);
1766 return service == RPC_GSS_SVC_INTEGRITY ||
1767 service == RPC_GSS_SVC_PRIVACY;
1770 static bool mach_creds_match(struct nfs4_client *cl, struct svc_rqst *rqstp)
1772 struct svc_cred *cr = &rqstp->rq_cred;
1774 if (!cl->cl_mach_cred)
1776 if (cl->cl_cred.cr_gss_mech != cr->cr_gss_mech)
1778 if (!svc_rqst_integrity_protected(rqstp))
1780 if (!cr->cr_principal)
1782 return 0 == strcmp(cl->cl_cred.cr_principal, cr->cr_principal);
1785 static void gen_clid(struct nfs4_client *clp, struct nfsd_net *nn)
1787 static u32 current_clientid = 1;
1789 clp->cl_clientid.cl_boot = nn->boot_time;
1790 clp->cl_clientid.cl_id = current_clientid++;
1793 static void gen_confirm(struct nfs4_client *clp)
1799 * This is opaque to client, so no need to byte-swap. Use
1800 * __force to keep sparse happy
1802 verf[0] = (__force __be32)get_seconds();
1803 verf[1] = (__force __be32)i++;
1804 memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
1807 static struct nfs4_stid *
1808 find_stateid_locked(struct nfs4_client *cl, stateid_t *t)
1810 struct nfs4_stid *ret;
1812 ret = idr_find(&cl->cl_stateids, t->si_opaque.so_id);
1813 if (!ret || !ret->sc_type)
1818 static struct nfs4_stid *
1819 find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
1821 struct nfs4_stid *s;
1823 spin_lock(&cl->cl_lock);
1824 s = find_stateid_locked(cl, t);
1826 if (typemask & s->sc_type)
1827 atomic_inc(&s->sc_count);
1831 spin_unlock(&cl->cl_lock);
1835 static struct nfs4_client *create_client(struct xdr_netobj name,
1836 struct svc_rqst *rqstp, nfs4_verifier *verf)
1838 struct nfs4_client *clp;
1839 struct sockaddr *sa = svc_addr(rqstp);
1841 struct net *net = SVC_NET(rqstp);
1842 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1844 clp = alloc_client(name);
1848 ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1850 spin_lock(&nn->client_lock);
1852 spin_unlock(&nn->client_lock);
1855 INIT_WORK(&clp->cl_cb_null.cb_work, nfsd4_run_cb_null);
1856 clp->cl_time = get_seconds();
1857 clear_bit(0, &clp->cl_cb_slot_busy);
1858 copy_verf(clp, verf);
1859 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1861 clp->cl_cb_session = NULL;
1867 add_clp_to_name_tree(struct nfs4_client *new_clp, struct rb_root *root)
1869 struct rb_node **new = &(root->rb_node), *parent = NULL;
1870 struct nfs4_client *clp;
1873 clp = rb_entry(*new, struct nfs4_client, cl_namenode);
1876 if (compare_blob(&clp->cl_name, &new_clp->cl_name) > 0)
1877 new = &((*new)->rb_left);
1879 new = &((*new)->rb_right);
1882 rb_link_node(&new_clp->cl_namenode, parent, new);
1883 rb_insert_color(&new_clp->cl_namenode, root);
1886 static struct nfs4_client *
1887 find_clp_in_name_tree(struct xdr_netobj *name, struct rb_root *root)
1890 struct rb_node *node = root->rb_node;
1891 struct nfs4_client *clp;
1894 clp = rb_entry(node, struct nfs4_client, cl_namenode);
1895 cmp = compare_blob(&clp->cl_name, name);
1897 node = node->rb_left;
1899 node = node->rb_right;
1907 add_to_unconfirmed(struct nfs4_client *clp)
1909 unsigned int idhashval;
1910 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1912 clear_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1913 add_clp_to_name_tree(clp, &nn->unconf_name_tree);
1914 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1915 list_add(&clp->cl_idhash, &nn->unconf_id_hashtbl[idhashval]);
1920 move_to_confirmed(struct nfs4_client *clp)
1922 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1923 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1925 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1926 list_move(&clp->cl_idhash, &nn->conf_id_hashtbl[idhashval]);
1927 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1928 add_clp_to_name_tree(clp, &nn->conf_name_tree);
1929 set_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1933 static struct nfs4_client *
1934 find_client_in_id_table(struct list_head *tbl, clientid_t *clid, bool sessions)
1936 struct nfs4_client *clp;
1937 unsigned int idhashval = clientid_hashval(clid->cl_id);
1939 list_for_each_entry(clp, &tbl[idhashval], cl_idhash) {
1940 if (same_clid(&clp->cl_clientid, clid)) {
1941 if ((bool)clp->cl_minorversion != sessions)
1950 static struct nfs4_client *
1951 find_confirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1953 struct list_head *tbl = nn->conf_id_hashtbl;
1955 return find_client_in_id_table(tbl, clid, sessions);
1958 static struct nfs4_client *
1959 find_unconfirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1961 struct list_head *tbl = nn->unconf_id_hashtbl;
1963 return find_client_in_id_table(tbl, clid, sessions);
1966 static bool clp_used_exchangeid(struct nfs4_client *clp)
1968 return clp->cl_exchange_flags != 0;
1971 static struct nfs4_client *
1972 find_confirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1974 return find_clp_in_name_tree(name, &nn->conf_name_tree);
1977 static struct nfs4_client *
1978 find_unconfirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1980 return find_clp_in_name_tree(name, &nn->unconf_name_tree);
1984 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
1986 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1987 struct sockaddr *sa = svc_addr(rqstp);
1988 u32 scopeid = rpc_get_scope_id(sa);
1989 unsigned short expected_family;
1991 /* Currently, we only support tcp and tcp6 for the callback channel */
1992 if (se->se_callback_netid_len == 3 &&
1993 !memcmp(se->se_callback_netid_val, "tcp", 3))
1994 expected_family = AF_INET;
1995 else if (se->se_callback_netid_len == 4 &&
1996 !memcmp(se->se_callback_netid_val, "tcp6", 4))
1997 expected_family = AF_INET6;
2001 conn->cb_addrlen = rpc_uaddr2sockaddr(clp->net, se->se_callback_addr_val,
2002 se->se_callback_addr_len,
2003 (struct sockaddr *)&conn->cb_addr,
2004 sizeof(conn->cb_addr));
2006 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
2009 if (conn->cb_addr.ss_family == AF_INET6)
2010 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
2012 conn->cb_prog = se->se_callback_prog;
2013 conn->cb_ident = se->se_callback_ident;
2014 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
2017 conn->cb_addr.ss_family = AF_UNSPEC;
2018 conn->cb_addrlen = 0;
2019 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
2020 "will not receive delegations\n",
2021 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
2027 * Cache a reply. nfsd4_check_resp_size() has bounded the cache size.
2030 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
2032 struct xdr_buf *buf = resp->xdr.buf;
2033 struct nfsd4_slot *slot = resp->cstate.slot;
2036 dprintk("--> %s slot %p\n", __func__, slot);
2038 slot->sl_opcnt = resp->opcnt;
2039 slot->sl_status = resp->cstate.status;
2041 slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
2042 if (nfsd4_not_cached(resp)) {
2043 slot->sl_datalen = 0;
2046 base = resp->cstate.data_offset;
2047 slot->sl_datalen = buf->len - base;
2048 if (read_bytes_from_xdr_buf(buf, base, slot->sl_data, slot->sl_datalen))
2049 WARN("%s: sessions DRC could not cache compound\n", __func__);
2054 * Encode the replay sequence operation from the slot values.
2055 * If cachethis is FALSE encode the uncached rep error on the next
2056 * operation which sets resp->p and increments resp->opcnt for
2057 * nfs4svc_encode_compoundres.
2061 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
2062 struct nfsd4_compoundres *resp)
2064 struct nfsd4_op *op;
2065 struct nfsd4_slot *slot = resp->cstate.slot;
2067 /* Encode the replayed sequence operation */
2068 op = &args->ops[resp->opcnt - 1];
2069 nfsd4_encode_operation(resp, op);
2071 /* Return nfserr_retry_uncached_rep in next operation. */
2072 if (args->opcnt > 1 && !(slot->sl_flags & NFSD4_SLOT_CACHETHIS)) {
2073 op = &args->ops[resp->opcnt++];
2074 op->status = nfserr_retry_uncached_rep;
2075 nfsd4_encode_operation(resp, op);
2081 * The sequence operation is not cached because we can use the slot and
2085 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
2086 struct nfsd4_sequence *seq)
2088 struct nfsd4_slot *slot = resp->cstate.slot;
2089 struct xdr_stream *xdr = &resp->xdr;
2093 dprintk("--> %s slot %p\n", __func__, slot);
2095 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
2099 p = xdr_reserve_space(xdr, slot->sl_datalen);
2102 return nfserr_serverfault;
2104 xdr_encode_opaque_fixed(p, slot->sl_data, slot->sl_datalen);
2105 xdr_commit_encode(xdr);
2107 resp->opcnt = slot->sl_opcnt;
2108 return slot->sl_status;
2112 * Set the exchange_id flags returned by the server.
2115 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
2117 /* pNFS is not supported */
2118 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
2120 /* Referrals are supported, Migration is not. */
2121 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
2123 /* set the wire flags to return to client. */
2124 clid->flags = new->cl_exchange_flags;
2127 static bool client_has_state(struct nfs4_client *clp)
2130 * Note clp->cl_openowners check isn't quite right: there's no
2131 * need to count owners without stateid's.
2133 * Also note we should probably be using this in 4.0 case too.
2135 return !list_empty(&clp->cl_openowners)
2136 || !list_empty(&clp->cl_delegations)
2137 || !list_empty(&clp->cl_sessions);
2141 nfsd4_exchange_id(struct svc_rqst *rqstp,
2142 struct nfsd4_compound_state *cstate,
2143 struct nfsd4_exchange_id *exid)
2145 struct nfs4_client *unconf, *conf, *new;
2147 char addr_str[INET6_ADDRSTRLEN];
2148 nfs4_verifier verf = exid->verifier;
2149 struct sockaddr *sa = svc_addr(rqstp);
2150 bool update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
2151 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2153 rpc_ntop(sa, addr_str, sizeof(addr_str));
2154 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
2155 "ip_addr=%s flags %x, spa_how %d\n",
2156 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
2157 addr_str, exid->flags, exid->spa_how);
2159 if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
2160 return nfserr_inval;
2162 switch (exid->spa_how) {
2164 if (!svc_rqst_integrity_protected(rqstp))
2165 return nfserr_inval;
2168 default: /* checked by xdr code */
2171 return nfserr_encr_alg_unsupp;
2174 /* Cases below refer to rfc 5661 section 18.35.4: */
2176 conf = find_confirmed_client_by_name(&exid->clname, nn);
2178 bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
2179 bool verfs_match = same_verf(&verf, &conf->cl_verifier);
2182 if (!clp_used_exchangeid(conf)) { /* buggy client */
2183 status = nfserr_inval;
2186 if (!mach_creds_match(conf, rqstp)) {
2187 status = nfserr_wrong_cred;
2190 if (!creds_match) { /* case 9 */
2191 status = nfserr_perm;
2194 if (!verfs_match) { /* case 8 */
2195 status = nfserr_not_same;
2199 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
2203 if (!creds_match) { /* case 3 */
2204 if (client_has_state(conf)) {
2205 status = nfserr_clid_inuse;
2208 expire_client(conf);
2211 if (verfs_match) { /* case 2 */
2212 conf->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
2216 /* case 5, client reboot */
2220 if (update) { /* case 7 */
2221 status = nfserr_noent;
2225 unconf = find_unconfirmed_client_by_name(&exid->clname, nn);
2226 if (unconf) /* case 4, possible retry or client restart */
2227 expire_client(unconf);
2229 /* case 1 (normal case) */
2231 new = create_client(exid->clname, rqstp, &verf);
2233 status = nfserr_jukebox;
2236 new->cl_minorversion = cstate->minorversion;
2237 new->cl_mach_cred = (exid->spa_how == SP4_MACH_CRED);
2240 add_to_unconfirmed(new);
2242 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
2243 exid->clientid.cl_id = new->cl_clientid.cl_id;
2245 exid->seqid = new->cl_cs_slot.sl_seqid + 1;
2246 nfsd4_set_ex_flags(new, exid);
2248 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
2249 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
2253 nfs4_unlock_state();
2258 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
2260 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
2263 /* The slot is in use, and no response has been sent. */
2265 if (seqid == slot_seqid)
2266 return nfserr_jukebox;
2268 return nfserr_seq_misordered;
2270 /* Note unsigned 32-bit arithmetic handles wraparound: */
2271 if (likely(seqid == slot_seqid + 1))
2273 if (seqid == slot_seqid)
2274 return nfserr_replay_cache;
2275 return nfserr_seq_misordered;
2279 * Cache the create session result into the create session single DRC
2280 * slot cache by saving the xdr structure. sl_seqid has been set.
2281 * Do this for solo or embedded create session operations.
2284 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
2285 struct nfsd4_clid_slot *slot, __be32 nfserr)
2287 slot->sl_status = nfserr;
2288 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
2292 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
2293 struct nfsd4_clid_slot *slot)
2295 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
2296 return slot->sl_status;
2299 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
2300 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
2301 1 + /* MIN tag is length with zero, only length */ \
2302 3 + /* version, opcount, opcode */ \
2303 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
2304 /* seqid, slotID, slotID, cache */ \
2305 4 ) * sizeof(__be32))
2307 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
2308 2 + /* verifier: AUTH_NULL, length 0 */\
2310 1 + /* MIN tag is length with zero, only length */ \
2311 3 + /* opcount, opcode, opstatus*/ \
2312 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
2313 /* seqid, slotID, slotID, slotID, status */ \
2314 5 ) * sizeof(__be32))
2316 static __be32 check_forechannel_attrs(struct nfsd4_channel_attrs *ca, struct nfsd_net *nn)
2318 u32 maxrpc = nn->nfsd_serv->sv_max_mesg;
2320 if (ca->maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ)
2321 return nfserr_toosmall;
2322 if (ca->maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ)
2323 return nfserr_toosmall;
2324 ca->headerpadsz = 0;
2325 ca->maxreq_sz = min_t(u32, ca->maxreq_sz, maxrpc);
2326 ca->maxresp_sz = min_t(u32, ca->maxresp_sz, maxrpc);
2327 ca->maxops = min_t(u32, ca->maxops, NFSD_MAX_OPS_PER_COMPOUND);
2328 ca->maxresp_cached = min_t(u32, ca->maxresp_cached,
2329 NFSD_SLOT_CACHE_SIZE + NFSD_MIN_HDR_SEQ_SZ);
2330 ca->maxreqs = min_t(u32, ca->maxreqs, NFSD_MAX_SLOTS_PER_SESSION);
2332 * Note decreasing slot size below client's request may make it
2333 * difficult for client to function correctly, whereas
2334 * decreasing the number of slots will (just?) affect
2335 * performance. When short on memory we therefore prefer to
2336 * decrease number of slots instead of their size. Clients that
2337 * request larger slots than they need will get poor results:
2339 ca->maxreqs = nfsd4_get_drc_mem(ca);
2341 return nfserr_jukebox;
2346 #define NFSD_CB_MAX_REQ_SZ ((NFS4_enc_cb_recall_sz + \
2347 RPC_MAX_HEADER_WITH_AUTH) * sizeof(__be32))
2348 #define NFSD_CB_MAX_RESP_SZ ((NFS4_dec_cb_recall_sz + \
2349 RPC_MAX_REPHEADER_WITH_AUTH) * sizeof(__be32))
2351 static __be32 check_backchannel_attrs(struct nfsd4_channel_attrs *ca)
2353 ca->headerpadsz = 0;
2356 * These RPC_MAX_HEADER macros are overkill, especially since we
2357 * don't even do gss on the backchannel yet. But this is still
2358 * less than 1k. Tighten up this estimate in the unlikely event
2359 * it turns out to be a problem for some client:
2361 if (ca->maxreq_sz < NFSD_CB_MAX_REQ_SZ)
2362 return nfserr_toosmall;
2363 if (ca->maxresp_sz < NFSD_CB_MAX_RESP_SZ)
2364 return nfserr_toosmall;
2365 ca->maxresp_cached = 0;
2367 return nfserr_toosmall;
2372 static __be32 nfsd4_check_cb_sec(struct nfsd4_cb_sec *cbs)
2374 switch (cbs->flavor) {
2380 * GSS case: the spec doesn't allow us to return this
2381 * error. But it also doesn't allow us not to support
2383 * I'd rather this fail hard than return some error the
2384 * client might think it can already handle:
2386 return nfserr_encr_alg_unsupp;
2391 nfsd4_create_session(struct svc_rqst *rqstp,
2392 struct nfsd4_compound_state *cstate,
2393 struct nfsd4_create_session *cr_ses)
2395 struct sockaddr *sa = svc_addr(rqstp);
2396 struct nfs4_client *conf, *unconf;
2397 struct nfsd4_session *new;
2398 struct nfsd4_conn *conn;
2399 struct nfsd4_clid_slot *cs_slot = NULL;
2401 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2403 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
2404 return nfserr_inval;
2405 status = nfsd4_check_cb_sec(&cr_ses->cb_sec);
2408 status = check_forechannel_attrs(&cr_ses->fore_channel, nn);
2411 status = check_backchannel_attrs(&cr_ses->back_channel);
2413 goto out_release_drc_mem;
2414 status = nfserr_jukebox;
2415 new = alloc_session(&cr_ses->fore_channel, &cr_ses->back_channel);
2417 goto out_release_drc_mem;
2418 conn = alloc_conn_from_crses(rqstp, cr_ses);
2420 goto out_free_session;
2423 unconf = find_unconfirmed_client(&cr_ses->clientid, true, nn);
2424 conf = find_confirmed_client(&cr_ses->clientid, true, nn);
2425 WARN_ON_ONCE(conf && unconf);
2428 status = nfserr_wrong_cred;
2429 if (!mach_creds_match(conf, rqstp))
2431 cs_slot = &conf->cl_cs_slot;
2432 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
2433 if (status == nfserr_replay_cache) {
2434 status = nfsd4_replay_create_session(cr_ses, cs_slot);
2436 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
2437 status = nfserr_seq_misordered;
2440 } else if (unconf) {
2441 struct nfs4_client *old;
2442 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
2443 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
2444 status = nfserr_clid_inuse;
2447 status = nfserr_wrong_cred;
2448 if (!mach_creds_match(unconf, rqstp))
2450 cs_slot = &unconf->cl_cs_slot;
2451 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
2453 /* an unconfirmed replay returns misordered */
2454 status = nfserr_seq_misordered;
2457 old = find_confirmed_client_by_name(&unconf->cl_name, nn);
2459 status = mark_client_expired(old);
2464 move_to_confirmed(unconf);
2467 status = nfserr_stale_clientid;
2472 * We do not support RDMA or persistent sessions
2474 cr_ses->flags &= ~SESSION4_PERSIST;
2475 cr_ses->flags &= ~SESSION4_RDMA;
2477 init_session(rqstp, new, conf, cr_ses);
2478 nfsd4_init_conn(rqstp, conn, new);
2480 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
2481 NFS4_MAX_SESSIONID_LEN);
2482 cs_slot->sl_seqid++;
2483 cr_ses->seqid = cs_slot->sl_seqid;
2485 /* cache solo and embedded create sessions under the state lock */
2486 nfsd4_cache_create_session(cr_ses, cs_slot, status);
2487 nfs4_unlock_state();
2490 nfs4_unlock_state();
2493 __free_session(new);
2494 out_release_drc_mem:
2495 nfsd4_put_drc_mem(&cr_ses->fore_channel);
2499 static __be32 nfsd4_map_bcts_dir(u32 *dir)
2502 case NFS4_CDFC4_FORE:
2503 case NFS4_CDFC4_BACK:
2505 case NFS4_CDFC4_FORE_OR_BOTH:
2506 case NFS4_CDFC4_BACK_OR_BOTH:
2507 *dir = NFS4_CDFC4_BOTH;
2510 return nfserr_inval;
2513 __be32 nfsd4_backchannel_ctl(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_backchannel_ctl *bc)
2515 struct nfsd4_session *session = cstate->session;
2516 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2519 status = nfsd4_check_cb_sec(&bc->bc_cb_sec);
2522 spin_lock(&nn->client_lock);
2523 session->se_cb_prog = bc->bc_cb_program;
2524 session->se_cb_sec = bc->bc_cb_sec;
2525 spin_unlock(&nn->client_lock);
2527 nfsd4_probe_callback(session->se_client);
2532 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
2533 struct nfsd4_compound_state *cstate,
2534 struct nfsd4_bind_conn_to_session *bcts)
2537 struct nfsd4_conn *conn;
2538 struct nfsd4_session *session;
2539 struct net *net = SVC_NET(rqstp);
2540 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2542 if (!nfsd4_last_compound_op(rqstp))
2543 return nfserr_not_only_op;
2545 spin_lock(&nn->client_lock);
2546 session = find_in_sessionid_hashtbl(&bcts->sessionid, net, &status);
2547 spin_unlock(&nn->client_lock);
2549 goto out_no_session;
2550 status = nfserr_wrong_cred;
2551 if (!mach_creds_match(session->se_client, rqstp))
2553 status = nfsd4_map_bcts_dir(&bcts->dir);
2556 conn = alloc_conn(rqstp, bcts->dir);
2557 status = nfserr_jukebox;
2560 nfsd4_init_conn(rqstp, conn, session);
2563 nfsd4_put_session(session);
2565 nfs4_unlock_state();
2569 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
2573 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
2577 nfsd4_destroy_session(struct svc_rqst *r,
2578 struct nfsd4_compound_state *cstate,
2579 struct nfsd4_destroy_session *sessionid)
2581 struct nfsd4_session *ses;
2583 int ref_held_by_me = 0;
2584 struct net *net = SVC_NET(r);
2585 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2588 status = nfserr_not_only_op;
2589 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
2590 if (!nfsd4_last_compound_op(r))
2594 dump_sessionid(__func__, &sessionid->sessionid);
2595 spin_lock(&nn->client_lock);
2596 ses = find_in_sessionid_hashtbl(&sessionid->sessionid, net, &status);
2598 goto out_client_lock;
2599 status = nfserr_wrong_cred;
2600 if (!mach_creds_match(ses->se_client, r))
2601 goto out_put_session;
2602 status = mark_session_dead_locked(ses, 1 + ref_held_by_me);
2604 goto out_put_session;
2605 unhash_session(ses);
2606 spin_unlock(&nn->client_lock);
2608 nfsd4_probe_callback_sync(ses->se_client);
2610 spin_lock(&nn->client_lock);
2613 nfsd4_put_session_locked(ses);
2615 spin_unlock(&nn->client_lock);
2617 nfs4_unlock_state();
2621 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
2623 struct nfsd4_conn *c;
2625 list_for_each_entry(c, &s->se_conns, cn_persession) {
2626 if (c->cn_xprt == xpt) {
2633 static __be32 nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
2635 struct nfs4_client *clp = ses->se_client;
2636 struct nfsd4_conn *c;
2637 __be32 status = nfs_ok;
2640 spin_lock(&clp->cl_lock);
2641 c = __nfsd4_find_conn(new->cn_xprt, ses);
2644 status = nfserr_conn_not_bound_to_session;
2645 if (clp->cl_mach_cred)
2647 __nfsd4_hash_conn(new, ses);
2648 spin_unlock(&clp->cl_lock);
2649 ret = nfsd4_register_conn(new);
2651 /* oops; xprt is already down: */
2652 nfsd4_conn_lost(&new->cn_xpt_user);
2655 spin_unlock(&clp->cl_lock);
2660 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
2662 struct nfsd4_compoundargs *args = rqstp->rq_argp;
2664 return args->opcnt > session->se_fchannel.maxops;
2667 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
2668 struct nfsd4_session *session)
2670 struct xdr_buf *xb = &rqstp->rq_arg;
2672 return xb->len > session->se_fchannel.maxreq_sz;
2676 nfsd4_sequence(struct svc_rqst *rqstp,
2677 struct nfsd4_compound_state *cstate,
2678 struct nfsd4_sequence *seq)
2680 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2681 struct xdr_stream *xdr = &resp->xdr;
2682 struct nfsd4_session *session;
2683 struct nfs4_client *clp;
2684 struct nfsd4_slot *slot;
2685 struct nfsd4_conn *conn;
2688 struct net *net = SVC_NET(rqstp);
2689 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2691 if (resp->opcnt != 1)
2692 return nfserr_sequence_pos;
2695 * Will be either used or freed by nfsd4_sequence_check_conn
2698 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
2700 return nfserr_jukebox;
2702 spin_lock(&nn->client_lock);
2703 session = find_in_sessionid_hashtbl(&seq->sessionid, net, &status);
2705 goto out_no_session;
2706 clp = session->se_client;
2708 status = nfserr_too_many_ops;
2709 if (nfsd4_session_too_many_ops(rqstp, session))
2710 goto out_put_session;
2712 status = nfserr_req_too_big;
2713 if (nfsd4_request_too_big(rqstp, session))
2714 goto out_put_session;
2716 status = nfserr_badslot;
2717 if (seq->slotid >= session->se_fchannel.maxreqs)
2718 goto out_put_session;
2720 slot = session->se_slots[seq->slotid];
2721 dprintk("%s: slotid %d\n", __func__, seq->slotid);
2723 /* We do not negotiate the number of slots yet, so set the
2724 * maxslots to the session maxreqs which is used to encode
2725 * sr_highest_slotid and the sr_target_slot id to maxslots */
2726 seq->maxslots = session->se_fchannel.maxreqs;
2728 status = check_slot_seqid(seq->seqid, slot->sl_seqid,
2729 slot->sl_flags & NFSD4_SLOT_INUSE);
2730 if (status == nfserr_replay_cache) {
2731 status = nfserr_seq_misordered;
2732 if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
2733 goto out_put_session;
2734 cstate->slot = slot;
2735 cstate->session = session;
2737 /* Return the cached reply status and set cstate->status
2738 * for nfsd4_proc_compound processing */
2739 status = nfsd4_replay_cache_entry(resp, seq);
2740 cstate->status = nfserr_replay_cache;
2744 goto out_put_session;
2746 status = nfsd4_sequence_check_conn(conn, session);
2749 goto out_put_session;
2751 buflen = (seq->cachethis) ?
2752 session->se_fchannel.maxresp_cached :
2753 session->se_fchannel.maxresp_sz;
2754 status = (seq->cachethis) ? nfserr_rep_too_big_to_cache :
2756 if (xdr_restrict_buflen(xdr, buflen - rqstp->rq_auth_slack))
2757 goto out_put_session;
2758 svc_reserve(rqstp, buflen);
2761 /* Success! bump slot seqid */
2762 slot->sl_seqid = seq->seqid;
2763 slot->sl_flags |= NFSD4_SLOT_INUSE;
2765 slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
2767 slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
2769 cstate->slot = slot;
2770 cstate->session = session;
2774 switch (clp->cl_cb_state) {
2776 seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
2778 case NFSD4_CB_FAULT:
2779 seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
2782 seq->status_flags = 0;
2784 if (!list_empty(&clp->cl_revoked))
2785 seq->status_flags |= SEQ4_STATUS_RECALLABLE_STATE_REVOKED;
2789 spin_unlock(&nn->client_lock);
2792 nfsd4_put_session_locked(session);
2793 goto out_no_session;
2797 nfsd4_sequence_done(struct nfsd4_compoundres *resp)
2799 struct nfsd4_compound_state *cs = &resp->cstate;
2801 if (nfsd4_has_session(cs)) {
2802 if (cs->status != nfserr_replay_cache) {
2803 nfsd4_store_cache_entry(resp);
2804 cs->slot->sl_flags &= ~NFSD4_SLOT_INUSE;
2806 /* Drop session reference that was taken in nfsd4_sequence() */
2807 nfsd4_put_session(cs->session);
2809 put_client_renew(cs->clp);
2813 nfsd4_destroy_clientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_destroy_clientid *dc)
2815 struct nfs4_client *conf, *unconf, *clp;
2817 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2820 unconf = find_unconfirmed_client(&dc->clientid, true, nn);
2821 conf = find_confirmed_client(&dc->clientid, true, nn);
2822 WARN_ON_ONCE(conf && unconf);
2827 if (client_has_state(conf)) {
2828 status = nfserr_clientid_busy;
2834 status = nfserr_stale_clientid;
2837 if (!mach_creds_match(clp, rqstp)) {
2838 status = nfserr_wrong_cred;
2843 nfs4_unlock_state();
2848 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
2852 if (rc->rca_one_fs) {
2853 if (!cstate->current_fh.fh_dentry)
2854 return nfserr_nofilehandle;
2856 * We don't take advantage of the rca_one_fs case.
2857 * That's OK, it's optional, we can safely ignore it.
2863 status = nfserr_complete_already;
2864 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
2865 &cstate->session->se_client->cl_flags))
2868 status = nfserr_stale_clientid;
2869 if (is_client_expired(cstate->session->se_client))
2871 * The following error isn't really legal.
2872 * But we only get here if the client just explicitly
2873 * destroyed the client. Surely it no longer cares what
2874 * error it gets back on an operation for the dead
2880 nfsd4_client_record_create(cstate->session->se_client);
2882 nfs4_unlock_state();
2887 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2888 struct nfsd4_setclientid *setclid)
2890 struct xdr_netobj clname = setclid->se_name;
2891 nfs4_verifier clverifier = setclid->se_verf;
2892 struct nfs4_client *conf, *unconf, *new;
2894 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2896 /* Cases below refer to rfc 3530 section 14.2.33: */
2898 conf = find_confirmed_client_by_name(&clname, nn);
2901 status = nfserr_clid_inuse;
2902 if (clp_used_exchangeid(conf))
2904 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
2905 char addr_str[INET6_ADDRSTRLEN];
2906 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
2908 dprintk("NFSD: setclientid: string in use by client "
2909 "at %s\n", addr_str);
2913 unconf = find_unconfirmed_client_by_name(&clname, nn);
2915 expire_client(unconf);
2916 status = nfserr_jukebox;
2917 new = create_client(clname, rqstp, &clverifier);
2920 if (conf && same_verf(&conf->cl_verifier, &clverifier))
2921 /* case 1: probable callback update */
2922 copy_clid(new, conf);
2923 else /* case 4 (new client) or cases 2, 3 (client reboot): */
2925 new->cl_minorversion = 0;
2926 gen_callback(new, setclid, rqstp);
2927 add_to_unconfirmed(new);
2928 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
2929 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
2930 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
2933 nfs4_unlock_state();
2939 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
2940 struct nfsd4_compound_state *cstate,
2941 struct nfsd4_setclientid_confirm *setclientid_confirm)
2943 struct nfs4_client *conf, *unconf;
2944 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
2945 clientid_t * clid = &setclientid_confirm->sc_clientid;
2947 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2949 if (STALE_CLIENTID(clid, nn))
2950 return nfserr_stale_clientid;
2953 conf = find_confirmed_client(clid, false, nn);
2954 unconf = find_unconfirmed_client(clid, false, nn);
2956 * We try hard to give out unique clientid's, so if we get an
2957 * attempt to confirm the same clientid with a different cred,
2958 * there's a bug somewhere. Let's charitably assume it's our
2961 status = nfserr_serverfault;
2962 if (unconf && !same_creds(&unconf->cl_cred, &rqstp->rq_cred))
2964 if (conf && !same_creds(&conf->cl_cred, &rqstp->rq_cred))
2966 /* cases below refer to rfc 3530 section 14.2.34: */
2967 if (!unconf || !same_verf(&confirm, &unconf->cl_confirm)) {
2968 if (conf && !unconf) /* case 2: probable retransmit */
2970 else /* case 4: client hasn't noticed we rebooted yet? */
2971 status = nfserr_stale_clientid;
2975 if (conf) { /* case 1: callback update */
2976 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
2977 nfsd4_probe_callback(conf);
2978 expire_client(unconf);
2979 } else { /* case 3: normal case; new or rebooted client */
2980 conf = find_confirmed_client_by_name(&unconf->cl_name, nn);
2982 status = mark_client_expired(conf);
2985 expire_client(conf);
2987 move_to_confirmed(unconf);
2988 nfsd4_probe_callback(unconf);
2991 nfs4_unlock_state();
2995 static struct nfs4_file *nfsd4_alloc_file(void)
2997 return kmem_cache_alloc(file_slab, GFP_KERNEL);
3000 /* OPEN Share state helper functions */
3001 static void nfsd4_init_file(struct nfs4_file *fp, struct knfsd_fh *fh)
3003 unsigned int hashval = file_hashval(fh);
3005 lockdep_assert_held(&state_lock);
3007 atomic_set(&fp->fi_ref, 1);
3008 spin_lock_init(&fp->fi_lock);
3009 INIT_LIST_HEAD(&fp->fi_stateids);
3010 INIT_LIST_HEAD(&fp->fi_delegations);
3011 fh_copy_shallow(&fp->fi_fhandle, fh);
3012 fp->fi_had_conflict = false;
3013 fp->fi_lease = NULL;
3014 fp->fi_share_deny = 0;
3015 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
3016 memset(fp->fi_access, 0, sizeof(fp->fi_access));
3017 hlist_add_head(&fp->fi_hash, &file_hashtbl[hashval]);
3021 nfsd4_free_slabs(void)
3023 kmem_cache_destroy(openowner_slab);
3024 kmem_cache_destroy(lockowner_slab);
3025 kmem_cache_destroy(file_slab);
3026 kmem_cache_destroy(stateid_slab);
3027 kmem_cache_destroy(deleg_slab);
3031 nfsd4_init_slabs(void)
3033 openowner_slab = kmem_cache_create("nfsd4_openowners",
3034 sizeof(struct nfs4_openowner), 0, 0, NULL);
3035 if (openowner_slab == NULL)
3037 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
3038 sizeof(struct nfs4_lockowner), 0, 0, NULL);
3039 if (lockowner_slab == NULL)
3040 goto out_free_openowner_slab;
3041 file_slab = kmem_cache_create("nfsd4_files",
3042 sizeof(struct nfs4_file), 0, 0, NULL);
3043 if (file_slab == NULL)
3044 goto out_free_lockowner_slab;
3045 stateid_slab = kmem_cache_create("nfsd4_stateids",
3046 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
3047 if (stateid_slab == NULL)
3048 goto out_free_file_slab;
3049 deleg_slab = kmem_cache_create("nfsd4_delegations",
3050 sizeof(struct nfs4_delegation), 0, 0, NULL);
3051 if (deleg_slab == NULL)
3052 goto out_free_stateid_slab;
3055 out_free_stateid_slab:
3056 kmem_cache_destroy(stateid_slab);
3058 kmem_cache_destroy(file_slab);
3059 out_free_lockowner_slab:
3060 kmem_cache_destroy(lockowner_slab);
3061 out_free_openowner_slab:
3062 kmem_cache_destroy(openowner_slab);
3064 dprintk("nfsd4: out of memory while initializing nfsv4\n");
3068 static void init_nfs4_replay(struct nfs4_replay *rp)
3070 rp->rp_status = nfserr_serverfault;
3072 rp->rp_buf = rp->rp_ibuf;
3073 mutex_init(&rp->rp_mutex);
3076 static void nfsd4_cstate_assign_replay(struct nfsd4_compound_state *cstate,
3077 struct nfs4_stateowner *so)
3079 if (!nfsd4_has_session(cstate)) {
3080 mutex_lock(&so->so_replay.rp_mutex);
3081 cstate->replay_owner = so;
3082 atomic_inc(&so->so_count);
3086 void nfsd4_cstate_clear_replay(struct nfsd4_compound_state *cstate)
3088 struct nfs4_stateowner *so = cstate->replay_owner;
3091 cstate->replay_owner = NULL;
3092 mutex_unlock(&so->so_replay.rp_mutex);
3093 nfs4_put_stateowner(so);
3097 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
3099 struct nfs4_stateowner *sop;
3101 sop = kmem_cache_alloc(slab, GFP_KERNEL);
3105 sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
3106 if (!sop->so_owner.data) {
3107 kmem_cache_free(slab, sop);
3110 sop->so_owner.len = owner->len;
3112 INIT_LIST_HEAD(&sop->so_stateids);
3113 sop->so_client = clp;
3114 init_nfs4_replay(&sop->so_replay);
3115 atomic_set(&sop->so_count, 1);
3119 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
3121 lockdep_assert_held(&clp->cl_lock);
3123 list_add(&oo->oo_owner.so_strhash,
3124 &clp->cl_ownerstr_hashtbl[strhashval]);
3125 list_add(&oo->oo_perclient, &clp->cl_openowners);
3128 static void nfs4_unhash_openowner(struct nfs4_stateowner *so)
3130 unhash_openowner_locked(openowner(so));
3133 static void nfs4_free_openowner(struct nfs4_stateowner *so)
3135 struct nfs4_openowner *oo = openowner(so);
3137 kmem_cache_free(openowner_slab, oo);
3140 static const struct nfs4_stateowner_operations openowner_ops = {
3141 .so_unhash = nfs4_unhash_openowner,
3142 .so_free = nfs4_free_openowner,
3145 static struct nfs4_openowner *
3146 alloc_init_open_stateowner(unsigned int strhashval, struct nfsd4_open *open,
3147 struct nfsd4_compound_state *cstate)
3149 struct nfs4_client *clp = cstate->clp;
3150 struct nfs4_openowner *oo, *ret;
3152 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
3155 oo->oo_owner.so_ops = &openowner_ops;
3156 oo->oo_owner.so_is_open_owner = 1;
3157 oo->oo_owner.so_seqid = open->op_seqid;
3159 if (nfsd4_has_session(cstate))
3160 oo->oo_flags |= NFS4_OO_CONFIRMED;
3162 oo->oo_last_closed_stid = NULL;
3163 INIT_LIST_HEAD(&oo->oo_close_lru);
3164 spin_lock(&clp->cl_lock);
3165 ret = find_openstateowner_str_locked(strhashval, open, clp);
3167 hash_openowner(oo, clp, strhashval);
3170 nfs4_free_openowner(&oo->oo_owner);
3171 spin_unlock(&clp->cl_lock);
3175 static void init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
3176 struct nfs4_openowner *oo = open->op_openowner;
3178 atomic_inc(&stp->st_stid.sc_count);
3179 stp->st_stid.sc_type = NFS4_OPEN_STID;
3180 INIT_LIST_HEAD(&stp->st_locks);
3181 stp->st_stateowner = &oo->oo_owner;
3182 atomic_inc(&stp->st_stateowner->so_count);
3184 stp->st_stid.sc_file = fp;
3185 stp->st_access_bmap = 0;
3186 stp->st_deny_bmap = 0;
3187 stp->st_openstp = NULL;
3188 spin_lock(&oo->oo_owner.so_client->cl_lock);
3189 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
3190 spin_lock(&fp->fi_lock);
3191 list_add(&stp->st_perfile, &fp->fi_stateids);
3192 spin_unlock(&fp->fi_lock);
3193 spin_unlock(&oo->oo_owner.so_client->cl_lock);
3197 * In the 4.0 case we need to keep the owners around a little while to handle
3198 * CLOSE replay. We still do need to release any file access that is held by
3199 * them before returning however.
3202 move_to_close_lru(struct nfs4_ol_stateid *s, struct net *net)
3204 struct nfs4_openowner *oo = openowner(s->st_stateowner);
3205 struct nfsd_net *nn = net_generic(s->st_stid.sc_client->net,
3208 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
3211 * We know that we hold one reference via nfsd4_close, and another
3212 * "persistent" reference for the client. If the refcount is higher
3213 * than 2, then there are still calls in progress that are using this
3214 * stateid. We can't put the sc_file reference until they are finished.
3215 * Wait for the refcount to drop to 2. Since it has been unhashed,
3216 * there should be no danger of the refcount going back up again at
3219 wait_event(close_wq, atomic_read(&s->st_stid.sc_count) == 2);
3221 release_all_access(s);
3222 if (s->st_stid.sc_file) {
3223 put_nfs4_file(s->st_stid.sc_file);
3224 s->st_stid.sc_file = NULL;
3226 release_last_closed_stateid(oo);
3227 oo->oo_last_closed_stid = s;
3228 list_move_tail(&oo->oo_close_lru, &nn->close_lru);
3229 oo->oo_time = get_seconds();
3232 /* search file_hashtbl[] for file */
3233 static struct nfs4_file *
3234 find_file_locked(struct knfsd_fh *fh)
3236 unsigned int hashval = file_hashval(fh);
3237 struct nfs4_file *fp;
3239 lockdep_assert_held(&state_lock);
3241 hlist_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
3242 if (nfsd_fh_match(&fp->fi_fhandle, fh)) {
3250 static struct nfs4_file *
3251 find_file(struct knfsd_fh *fh)
3253 struct nfs4_file *fp;
3255 spin_lock(&state_lock);
3256 fp = find_file_locked(fh);
3257 spin_unlock(&state_lock);
3261 static struct nfs4_file *
3262 find_or_add_file(struct nfs4_file *new, struct knfsd_fh *fh)
3264 struct nfs4_file *fp;
3266 spin_lock(&state_lock);
3267 fp = find_file_locked(fh);
3269 nfsd4_init_file(new, fh);
3272 spin_unlock(&state_lock);
3278 * Called to check deny when READ with all zero stateid or
3279 * WRITE with all zero or all one stateid
3282 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
3284 struct nfs4_file *fp;
3285 __be32 ret = nfs_ok;
3287 fp = find_file(¤t_fh->fh_handle);
3290 /* Check for conflicting share reservations */
3291 spin_lock(&fp->fi_lock);
3292 if (fp->fi_share_deny & deny_type)
3293 ret = nfserr_locked;
3294 spin_unlock(&fp->fi_lock);
3299 void nfsd4_prepare_cb_recall(struct nfs4_delegation *dp)
3301 struct nfsd_net *nn = net_generic(dp->dl_stid.sc_client->net,
3304 block_delegations(&dp->dl_stid.sc_file->fi_fhandle);
3307 * We can't do this in nfsd_break_deleg_cb because it is
3308 * already holding inode->i_lock.
3310 * If the dl_time != 0, then we know that it has already been
3311 * queued for a lease break. Don't queue it again.
3313 spin_lock(&state_lock);
3314 if (dp->dl_time == 0) {
3315 dp->dl_time = get_seconds();
3316 list_add_tail(&dp->dl_recall_lru, &nn->del_recall_lru);
3318 spin_unlock(&state_lock);
3321 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
3324 * We're assuming the state code never drops its reference
3325 * without first removing the lease. Since we're in this lease
3326 * callback (and since the lease code is serialized by the kernel
3327 * lock) we know the server hasn't removed the lease yet, we know
3328 * it's safe to take a reference.
3330 atomic_inc(&dp->dl_stid.sc_count);
3331 nfsd4_cb_recall(dp);
3334 /* Called from break_lease() with i_lock held. */
3335 static void nfsd_break_deleg_cb(struct file_lock *fl)
3337 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
3338 struct nfs4_delegation *dp;
3341 WARN(1, "(%p)->fl_owner NULL\n", fl);
3344 if (fp->fi_had_conflict) {
3345 WARN(1, "duplicate break on %p\n", fp);
3349 * We don't want the locks code to timeout the lease for us;
3350 * we'll remove it ourself if a delegation isn't returned
3353 fl->fl_break_time = 0;
3355 spin_lock(&fp->fi_lock);
3356 fp->fi_had_conflict = true;
3358 * If there are no delegations on the list, then we can't count on this
3359 * lease ever being cleaned up. Set the fl_break_time to jiffies so that
3360 * time_out_leases will do it ASAP. The fact that fi_had_conflict is now
3361 * true should keep any new delegations from being hashed.
3363 if (list_empty(&fp->fi_delegations))
3364 fl->fl_break_time = jiffies;
3366 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
3367 nfsd_break_one_deleg(dp);
3368 spin_unlock(&fp->fi_lock);
3372 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
3375 return lease_modify(onlist, arg);
3380 static const struct lock_manager_operations nfsd_lease_mng_ops = {
3381 .lm_break = nfsd_break_deleg_cb,
3382 .lm_change = nfsd_change_deleg_cb,
3385 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
3387 if (nfsd4_has_session(cstate))
3389 if (seqid == so->so_seqid - 1)
3390 return nfserr_replay_me;
3391 if (seqid == so->so_seqid)
3393 return nfserr_bad_seqid;
3396 static __be32 lookup_clientid(clientid_t *clid,
3397 struct nfsd4_compound_state *cstate,
3398 struct nfsd_net *nn)
3400 struct nfs4_client *found;
3403 found = cstate->clp;
3404 if (!same_clid(&found->cl_clientid, clid))
3405 return nfserr_stale_clientid;
3409 if (STALE_CLIENTID(clid, nn))
3410 return nfserr_stale_clientid;
3413 * For v4.1+ we get the client in the SEQUENCE op. If we don't have one
3414 * cached already then we know this is for is for v4.0 and "sessions"
3417 WARN_ON_ONCE(cstate->session);
3418 found = find_confirmed_client(clid, false, nn);
3420 return nfserr_expired;
3422 /* Cache the nfs4_client in cstate! */
3423 cstate->clp = found;
3424 atomic_inc(&found->cl_refcount);
3429 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
3430 struct nfsd4_open *open, struct nfsd_net *nn)
3432 clientid_t *clientid = &open->op_clientid;
3433 struct nfs4_client *clp = NULL;
3434 unsigned int strhashval;
3435 struct nfs4_openowner *oo = NULL;
3438 if (STALE_CLIENTID(&open->op_clientid, nn))
3439 return nfserr_stale_clientid;
3441 * In case we need it later, after we've already created the
3442 * file and don't want to risk a further failure:
3444 open->op_file = nfsd4_alloc_file();
3445 if (open->op_file == NULL)
3446 return nfserr_jukebox;
3448 status = lookup_clientid(clientid, cstate, nn);
3453 strhashval = ownerstr_hashval(&open->op_owner);
3454 oo = find_openstateowner_str(strhashval, open, clp);
3455 open->op_openowner = oo;
3459 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
3460 /* Replace unconfirmed owners without checking for replay. */
3461 release_openowner(oo);
3462 open->op_openowner = NULL;
3465 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
3470 oo = alloc_init_open_stateowner(strhashval, open, cstate);
3472 return nfserr_jukebox;
3473 open->op_openowner = oo;
3475 open->op_stp = nfs4_alloc_open_stateid(clp);
3477 return nfserr_jukebox;
3481 static inline __be32
3482 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
3484 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
3485 return nfserr_openmode;
3490 static int share_access_to_flags(u32 share_access)
3492 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
3495 static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
3497 struct nfs4_stid *ret;
3499 ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
3502 return delegstateid(ret);
3505 static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
3507 return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
3508 open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
3512 nfs4_check_deleg(struct nfs4_client *cl, struct nfsd4_open *open,
3513 struct nfs4_delegation **dp)
3516 __be32 status = nfserr_bad_stateid;
3517 struct nfs4_delegation *deleg;
3519 deleg = find_deleg_stateid(cl, &open->op_delegate_stateid);
3522 flags = share_access_to_flags(open->op_share_access);
3523 status = nfs4_check_delegmode(deleg, flags);
3525 nfs4_put_stid(&deleg->dl_stid);
3530 if (!nfsd4_is_deleg_cur(open))
3534 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3538 static struct nfs4_ol_stateid *
3539 nfsd4_find_existing_open(struct nfs4_file *fp, struct nfsd4_open *open)
3541 struct nfs4_ol_stateid *local, *ret = NULL;
3542 struct nfs4_openowner *oo = open->op_openowner;
3544 spin_lock(&fp->fi_lock);
3545 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
3546 /* ignore lock owners */
3547 if (local->st_stateowner->so_is_open_owner == 0)
3549 if (local->st_stateowner == &oo->oo_owner) {
3551 atomic_inc(&ret->st_stid.sc_count);
3555 spin_unlock(&fp->fi_lock);
3559 static inline int nfs4_access_to_access(u32 nfs4_access)
3563 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
3564 flags |= NFSD_MAY_READ;
3565 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
3566 flags |= NFSD_MAY_WRITE;
3570 static inline __be32
3571 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
3572 struct nfsd4_open *open)
3574 struct iattr iattr = {
3575 .ia_valid = ATTR_SIZE,
3578 if (!open->op_truncate)
3580 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
3581 return nfserr_inval;
3582 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
3585 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
3586 struct svc_fh *cur_fh, struct nfs4_ol_stateid *stp,
3587 struct nfsd4_open *open)
3589 struct file *filp = NULL;
3591 int oflag = nfs4_access_to_omode(open->op_share_access);
3592 int access = nfs4_access_to_access(open->op_share_access);
3593 unsigned char old_access_bmap, old_deny_bmap;
3595 spin_lock(&fp->fi_lock);
3598 * Are we trying to set a deny mode that would conflict with
3601 status = nfs4_file_check_deny(fp, open->op_share_deny);
3602 if (status != nfs_ok) {
3603 spin_unlock(&fp->fi_lock);
3607 /* set access to the file */
3608 status = nfs4_file_get_access(fp, open->op_share_access);
3609 if (status != nfs_ok) {
3610 spin_unlock(&fp->fi_lock);
3614 /* Set access bits in stateid */
3615 old_access_bmap = stp->st_access_bmap;
3616 set_access(open->op_share_access, stp);
3618 /* Set new deny mask */
3619 old_deny_bmap = stp->st_deny_bmap;
3620 set_deny(open->op_share_deny, stp);
3621 fp->fi_share_deny |= (open->op_share_deny & NFS4_SHARE_DENY_BOTH);
3623 if (!fp->fi_fds[oflag]) {
3624 spin_unlock(&fp->fi_lock);
3625 status = nfsd_open(rqstp, cur_fh, S_IFREG, access, &filp);
3627 goto out_put_access;
3628 spin_lock(&fp->fi_lock);
3629 if (!fp->fi_fds[oflag]) {
3630 fp->fi_fds[oflag] = filp;
3634 spin_unlock(&fp->fi_lock);
3638 status = nfsd4_truncate(rqstp, cur_fh, open);
3640 goto out_put_access;
3644 stp->st_access_bmap = old_access_bmap;
3645 nfs4_file_put_access(fp, open->op_share_access);
3646 reset_union_bmap_deny(bmap_to_share_mode(old_deny_bmap), stp);
3651 nfs4_upgrade_open(struct svc_rqst *rqstp, struct nfs4_file *fp, struct svc_fh *cur_fh, struct nfs4_ol_stateid *stp, struct nfsd4_open *open)
3654 unsigned char old_deny_bmap;
3656 if (!test_access(open->op_share_access, stp))
3657 return nfs4_get_vfs_file(rqstp, fp, cur_fh, stp, open);
3659 /* test and set deny mode */
3660 spin_lock(&fp->fi_lock);
3661 status = nfs4_file_check_deny(fp, open->op_share_deny);
3662 if (status == nfs_ok) {
3663 old_deny_bmap = stp->st_deny_bmap;
3664 set_deny(open->op_share_deny, stp);
3665 fp->fi_share_deny |=
3666 (open->op_share_deny & NFS4_SHARE_DENY_BOTH);
3668 spin_unlock(&fp->fi_lock);
3670 if (status != nfs_ok)
3673 status = nfsd4_truncate(rqstp, cur_fh, open);
3674 if (status != nfs_ok)
3675 reset_union_bmap_deny(old_deny_bmap, stp);
3680 nfs4_set_claim_prev(struct nfsd4_open *open, bool has_session)
3682 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3685 /* Should we give out recallable state?: */
3686 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
3688 if (clp->cl_cb_state == NFSD4_CB_UP)
3691 * In the sessions case, since we don't have to establish a
3692 * separate connection for callbacks, we assume it's OK
3693 * until we hear otherwise:
3695 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
3698 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_file *fp, int flag)
3700 struct file_lock *fl;
3702 fl = locks_alloc_lock();
3705 locks_init_lock(fl);
3706 fl->fl_lmops = &nfsd_lease_mng_ops;
3707 fl->fl_flags = FL_DELEG;
3708 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
3709 fl->fl_end = OFFSET_MAX;
3710 fl->fl_owner = (fl_owner_t)fp;
3711 fl->fl_pid = current->tgid;
3715 static int nfs4_setlease(struct nfs4_delegation *dp)
3717 struct nfs4_file *fp = dp->dl_stid.sc_file;
3718 struct file_lock *fl;
3722 fl = nfs4_alloc_init_lease(fp, NFS4_OPEN_DELEGATE_READ);
3725 filp = find_readable_file(fp);
3727 /* We should always have a readable file here */
3732 status = vfs_setlease(filp, fl->fl_type, &fl);
3734 locks_free_lock(fl);
3737 spin_lock(&state_lock);
3738 spin_lock(&fp->fi_lock);
3739 /* Did the lease get broken before we took the lock? */
3741 if (fp->fi_had_conflict)
3746 atomic_inc(&fp->fi_delegees);
3747 hash_delegation_locked(dp, fp);
3751 fp->fi_deleg_file = filp;
3752 atomic_set(&fp->fi_delegees, 1);
3753 hash_delegation_locked(dp, fp);
3754 spin_unlock(&fp->fi_lock);
3755 spin_unlock(&state_lock);
3758 spin_unlock(&fp->fi_lock);
3759 spin_unlock(&state_lock);
3765 static struct nfs4_delegation *
3766 nfs4_set_delegation(struct nfs4_client *clp, struct svc_fh *fh,
3767 struct nfs4_file *fp)
3770 struct nfs4_delegation *dp;
3772 if (fp->fi_had_conflict)
3773 return ERR_PTR(-EAGAIN);
3775 dp = alloc_init_deleg(clp, fh);
3777 return ERR_PTR(-ENOMEM);
3780 spin_lock(&state_lock);
3781 spin_lock(&fp->fi_lock);
3782 dp->dl_stid.sc_file = fp;
3783 if (!fp->fi_lease) {
3784 spin_unlock(&fp->fi_lock);
3785 spin_unlock(&state_lock);
3786 status = nfs4_setlease(dp);
3789 atomic_inc(&fp->fi_delegees);
3790 if (fp->fi_had_conflict) {
3794 hash_delegation_locked(dp, fp);
3797 spin_unlock(&fp->fi_lock);
3798 spin_unlock(&state_lock);
3801 nfs4_put_stid(&dp->dl_stid);
3802 return ERR_PTR(status);
3807 static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
3809 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3810 if (status == -EAGAIN)
3811 open->op_why_no_deleg = WND4_CONTENTION;
3813 open->op_why_no_deleg = WND4_RESOURCE;
3814 switch (open->op_deleg_want) {
3815 case NFS4_SHARE_WANT_READ_DELEG:
3816 case NFS4_SHARE_WANT_WRITE_DELEG:
3817 case NFS4_SHARE_WANT_ANY_DELEG:
3819 case NFS4_SHARE_WANT_CANCEL:
3820 open->op_why_no_deleg = WND4_CANCELLED;
3822 case NFS4_SHARE_WANT_NO_DELEG:
3829 * Attempt to hand out a delegation.
3831 * Note we don't support write delegations, and won't until the vfs has
3832 * proper support for them.
3835 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open,
3836 struct nfs4_ol_stateid *stp)
3838 struct nfs4_delegation *dp;
3839 struct nfs4_openowner *oo = openowner(stp->st_stateowner);
3840 struct nfs4_client *clp = stp->st_stid.sc_client;
3844 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
3845 open->op_recall = 0;
3846 switch (open->op_claim_type) {
3847 case NFS4_OPEN_CLAIM_PREVIOUS:
3849 open->op_recall = 1;
3850 if (open->op_delegate_type != NFS4_OPEN_DELEGATE_READ)
3853 case NFS4_OPEN_CLAIM_NULL:
3854 case NFS4_OPEN_CLAIM_FH:
3856 * Let's not give out any delegations till everyone's
3857 * had the chance to reclaim theirs....
3859 if (locks_in_grace(clp->net))
3861 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
3864 * Also, if the file was opened for write or
3865 * create, there's a good chance the client's
3866 * about to write to it, resulting in an
3867 * immediate recall (since we don't support
3868 * write delegations):
3870 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
3872 if (open->op_create == NFS4_OPEN_CREATE)
3878 dp = nfs4_set_delegation(clp, fh, stp->st_stid.sc_file);
3882 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
3884 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
3885 STATEID_VAL(&dp->dl_stid.sc_stateid));
3886 open->op_delegate_type = NFS4_OPEN_DELEGATE_READ;
3887 nfs4_put_stid(&dp->dl_stid);
3890 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE;
3891 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
3892 open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE) {
3893 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
3894 open->op_recall = 1;
3897 /* 4.1 client asking for a delegation? */
3898 if (open->op_deleg_want)
3899 nfsd4_open_deleg_none_ext(open, status);
3903 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
3904 struct nfs4_delegation *dp)
3906 if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
3907 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3908 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3909 open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
3910 } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
3911 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3912 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3913 open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
3915 /* Otherwise the client must be confused wanting a delegation
3916 * it already has, therefore we don't return
3917 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
3922 * called with nfs4_lock_state() held.
3925 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
3927 struct nfsd4_compoundres *resp = rqstp->rq_resp;
3928 struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
3929 struct nfs4_file *fp = NULL;
3930 struct nfs4_ol_stateid *stp = NULL;
3931 struct nfs4_delegation *dp = NULL;
3935 * Lookup file; if found, lookup stateid and check open request,
3936 * and check for delegations in the process of being recalled.
3937 * If not found, create the nfs4_file struct
3939 fp = find_or_add_file(open->op_file, ¤t_fh->fh_handle);
3940 if (fp != open->op_file) {
3941 status = nfs4_check_deleg(cl, open, &dp);
3944 stp = nfsd4_find_existing_open(fp, open);
3946 open->op_file = NULL;
3947 status = nfserr_bad_stateid;
3948 if (nfsd4_is_deleg_cur(open))
3950 status = nfserr_jukebox;
3954 * OPEN the file, or upgrade an existing OPEN.
3955 * If truncate fails, the OPEN fails.
3958 /* Stateid was found, this is an OPEN upgrade */
3959 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
3964 open->op_stp = NULL;
3965 init_open_stateid(stp, fp, open);
3966 status = nfs4_get_vfs_file(rqstp, fp, current_fh, stp, open);
3968 release_open_stateid(stp);
3972 update_stateid(&stp->st_stid.sc_stateid);
3973 memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3975 if (nfsd4_has_session(&resp->cstate)) {
3976 if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
3977 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3978 open->op_why_no_deleg = WND4_NOT_WANTED;
3984 * Attempt to hand out a delegation. No error return, because the
3985 * OPEN succeeds even if we fail.
3987 nfs4_open_delegation(current_fh, open, stp);
3991 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
3992 STATEID_VAL(&stp->st_stid.sc_stateid));
3994 /* 4.1 client trying to upgrade/downgrade delegation? */
3995 if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
3996 open->op_deleg_want)
3997 nfsd4_deleg_xgrade_none_ext(open, dp);
4001 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
4002 nfs4_set_claim_prev(open, nfsd4_has_session(&resp->cstate));
4004 * To finish the open response, we just need to set the rflags.
4006 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
4007 if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
4008 !nfsd4_has_session(&resp->cstate))
4009 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
4011 nfs4_put_stid(&dp->dl_stid);
4013 nfs4_put_stid(&stp->st_stid);
4018 void nfsd4_cleanup_open_state(struct nfsd4_compound_state *cstate,
4019 struct nfsd4_open *open, __be32 status)
4021 if (open->op_openowner) {
4022 struct nfs4_stateowner *so = &open->op_openowner->oo_owner;
4024 nfsd4_cstate_assign_replay(cstate, so);
4025 nfs4_put_stateowner(so);
4028 nfsd4_free_file(open->op_file);
4030 nfs4_put_stid(&open->op_stp->st_stid);
4034 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4037 struct nfs4_client *clp;
4039 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4042 dprintk("process_renew(%08x/%08x): starting\n",
4043 clid->cl_boot, clid->cl_id);
4044 status = lookup_clientid(clid, cstate, nn);
4048 status = nfserr_cb_path_down;
4049 if (!list_empty(&clp->cl_delegations)
4050 && clp->cl_cb_state != NFSD4_CB_UP)
4054 nfs4_unlock_state();
4059 nfsd4_end_grace(struct nfsd_net *nn)
4061 /* do nothing if grace period already ended */
4062 if (nn->grace_ended)
4065 dprintk("NFSD: end of grace period\n");
4066 nn->grace_ended = true;
4067 nfsd4_record_grace_done(nn, nn->boot_time);
4068 locks_end_grace(&nn->nfsd4_manager);
4070 * Now that every NFSv4 client has had the chance to recover and
4071 * to see the (possibly new, possibly shorter) lease time, we
4072 * can safely set the next grace time to the current lease time:
4074 nn->nfsd4_grace = nn->nfsd4_lease;
4078 nfs4_laundromat(struct nfsd_net *nn)
4080 struct nfs4_client *clp;
4081 struct nfs4_openowner *oo;
4082 struct nfs4_delegation *dp;
4083 struct list_head *pos, *next, reaplist;
4084 time_t cutoff = get_seconds() - nn->nfsd4_lease;
4085 time_t t, new_timeo = nn->nfsd4_lease;
4089 dprintk("NFSD: laundromat service - starting\n");
4090 nfsd4_end_grace(nn);
4091 INIT_LIST_HEAD(&reaplist);
4092 spin_lock(&nn->client_lock);
4093 list_for_each_safe(pos, next, &nn->client_lru) {
4094 clp = list_entry(pos, struct nfs4_client, cl_lru);
4095 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
4096 t = clp->cl_time - cutoff;
4097 new_timeo = min(new_timeo, t);
4100 if (mark_client_expired_locked(clp)) {
4101 dprintk("NFSD: client in use (clientid %08x)\n",
4102 clp->cl_clientid.cl_id);
4105 list_move(&clp->cl_lru, &reaplist);
4107 spin_unlock(&nn->client_lock);
4108 list_for_each_safe(pos, next, &reaplist) {
4109 clp = list_entry(pos, struct nfs4_client, cl_lru);
4110 dprintk("NFSD: purging unused client (clientid %08x)\n",
4111 clp->cl_clientid.cl_id);
4114 spin_lock(&state_lock);
4115 list_for_each_safe(pos, next, &nn->del_recall_lru) {
4116 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4117 if (net_generic(dp->dl_stid.sc_client->net, nfsd_net_id) != nn)
4119 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
4120 t = dp->dl_time - cutoff;
4121 new_timeo = min(new_timeo, t);
4124 unhash_delegation_locked(dp);
4125 list_add(&dp->dl_recall_lru, &reaplist);
4127 spin_unlock(&state_lock);
4128 while (!list_empty(&reaplist)) {
4129 dp = list_first_entry(&reaplist, struct nfs4_delegation,
4131 list_del_init(&dp->dl_recall_lru);
4132 revoke_delegation(dp);
4134 list_for_each_safe(pos, next, &nn->close_lru) {
4135 oo = container_of(pos, struct nfs4_openowner, oo_close_lru);
4136 if (time_after((unsigned long)oo->oo_time, (unsigned long)cutoff)) {
4137 t = oo->oo_time - cutoff;
4138 new_timeo = min(new_timeo, t);
4141 release_last_closed_stateid(oo);
4143 new_timeo = max_t(time_t, new_timeo, NFSD_LAUNDROMAT_MINTIMEOUT);
4144 nfs4_unlock_state();
4148 static struct workqueue_struct *laundry_wq;
4149 static void laundromat_main(struct work_struct *);
4152 laundromat_main(struct work_struct *laundry)
4155 struct delayed_work *dwork = container_of(laundry, struct delayed_work,
4157 struct nfsd_net *nn = container_of(dwork, struct nfsd_net,
4160 t = nfs4_laundromat(nn);
4161 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
4162 queue_delayed_work(laundry_wq, &nn->laundromat_work, t*HZ);
4165 static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *stp)
4167 if (!nfsd_fh_match(&fhp->fh_handle, &stp->st_stid.sc_file->fi_fhandle))
4168 return nfserr_bad_stateid;
4173 access_permit_read(struct nfs4_ol_stateid *stp)
4175 return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
4176 test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
4177 test_access(NFS4_SHARE_ACCESS_WRITE, stp);
4181 access_permit_write(struct nfs4_ol_stateid *stp)
4183 return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
4184 test_access(NFS4_SHARE_ACCESS_BOTH, stp);
4188 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
4190 __be32 status = nfserr_openmode;
4192 /* For lock stateid's, we test the parent open, not the lock: */
4193 if (stp->st_openstp)
4194 stp = stp->st_openstp;
4195 if ((flags & WR_STATE) && !access_permit_write(stp))
4197 if ((flags & RD_STATE) && !access_permit_read(stp))
4204 static inline __be32
4205 check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
4207 if (ONE_STATEID(stateid) && (flags & RD_STATE))
4209 else if (locks_in_grace(net)) {
4210 /* Answer in remaining cases depends on existence of
4211 * conflicting state; so we must wait out the grace period. */
4212 return nfserr_grace;
4213 } else if (flags & WR_STATE)
4214 return nfs4_share_conflict(current_fh,
4215 NFS4_SHARE_DENY_WRITE);
4216 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
4217 return nfs4_share_conflict(current_fh,
4218 NFS4_SHARE_DENY_READ);
4222 * Allow READ/WRITE during grace period on recovered state only for files
4223 * that are not able to provide mandatory locking.
4226 grace_disallows_io(struct net *net, struct inode *inode)
4228 return locks_in_grace(net) && mandatory_lock(inode);
4231 /* Returns true iff a is later than b: */
4232 static bool stateid_generation_after(stateid_t *a, stateid_t *b)
4234 return (s32)(a->si_generation - b->si_generation) > 0;
4237 static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
4240 * When sessions are used the stateid generation number is ignored
4243 if (has_session && in->si_generation == 0)
4246 if (in->si_generation == ref->si_generation)
4249 /* If the client sends us a stateid from the future, it's buggy: */
4250 if (stateid_generation_after(in, ref))
4251 return nfserr_bad_stateid;
4253 * However, we could see a stateid from the past, even from a
4254 * non-buggy client. For example, if the client sends a lock
4255 * while some IO is outstanding, the lock may bump si_generation
4256 * while the IO is still in flight. The client could avoid that
4257 * situation by waiting for responses on all the IO requests,
4258 * but better performance may result in retrying IO that
4259 * receives an old_stateid error if requests are rarely
4260 * reordered in flight:
4262 return nfserr_old_stateid;
4265 static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
4267 struct nfs4_stid *s;
4268 struct nfs4_ol_stateid *ols;
4269 __be32 status = nfserr_bad_stateid;
4271 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
4273 /* Client debugging aid. */
4274 if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid)) {
4275 char addr_str[INET6_ADDRSTRLEN];
4276 rpc_ntop((struct sockaddr *)&cl->cl_addr, addr_str,
4278 pr_warn_ratelimited("NFSD: client %s testing state ID "
4279 "with incorrect client ID\n", addr_str);
4282 spin_lock(&cl->cl_lock);
4283 s = find_stateid_locked(cl, stateid);
4286 status = check_stateid_generation(stateid, &s->sc_stateid, 1);
4289 switch (s->sc_type) {
4290 case NFS4_DELEG_STID:
4293 case NFS4_REVOKED_DELEG_STID:
4294 status = nfserr_deleg_revoked;
4296 case NFS4_OPEN_STID:
4297 case NFS4_LOCK_STID:
4298 ols = openlockstateid(s);
4299 if (ols->st_stateowner->so_is_open_owner
4300 && !(openowner(ols->st_stateowner)->oo_flags
4301 & NFS4_OO_CONFIRMED))
4302 status = nfserr_bad_stateid;
4307 printk("unknown stateid type %x\n", s->sc_type);
4309 case NFS4_CLOSED_STID:
4310 case NFS4_CLOSED_DELEG_STID:
4311 status = nfserr_bad_stateid;
4314 spin_unlock(&cl->cl_lock);
4319 nfsd4_lookup_stateid(struct nfsd4_compound_state *cstate,
4320 stateid_t *stateid, unsigned char typemask,
4321 struct nfs4_stid **s, struct nfsd_net *nn)
4325 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
4326 return nfserr_bad_stateid;
4327 status = lookup_clientid(&stateid->si_opaque.so_clid, cstate, nn);
4328 if (status == nfserr_stale_clientid) {
4329 if (cstate->session)
4330 return nfserr_bad_stateid;
4331 return nfserr_stale_stateid;
4335 *s = find_stateid_by_type(cstate->clp, stateid, typemask);
4337 return nfserr_bad_stateid;
4342 * Checks for stateid operations
4345 nfs4_preprocess_stateid_op(struct net *net, struct nfsd4_compound_state *cstate,
4346 stateid_t *stateid, int flags, struct file **filpp)
4348 struct nfs4_stid *s;
4349 struct nfs4_ol_stateid *stp = NULL;
4350 struct nfs4_delegation *dp = NULL;
4351 struct svc_fh *current_fh = &cstate->current_fh;
4352 struct inode *ino = current_fh->fh_dentry->d_inode;
4353 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4354 struct file *file = NULL;
4360 if (grace_disallows_io(net, ino))
4361 return nfserr_grace;
4363 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
4364 return check_special_stateids(net, current_fh, stateid, flags);
4368 status = nfsd4_lookup_stateid(cstate, stateid,
4369 NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID,
4373 status = check_stateid_generation(stateid, &s->sc_stateid, nfsd4_has_session(cstate));
4376 switch (s->sc_type) {
4377 case NFS4_DELEG_STID:
4378 dp = delegstateid(s);
4379 status = nfs4_check_delegmode(dp, flags);
4383 file = dp->dl_stid.sc_file->fi_deleg_file;
4386 status = nfserr_serverfault;
4392 case NFS4_OPEN_STID:
4393 case NFS4_LOCK_STID:
4394 stp = openlockstateid(s);
4395 status = nfs4_check_fh(current_fh, stp);
4398 if (stp->st_stateowner->so_is_open_owner
4399 && !(openowner(stp->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
4401 status = nfs4_check_openmode(stp, flags);
4405 struct nfs4_file *fp = stp->st_stid.sc_file;
4407 if (flags & RD_STATE)
4408 file = find_readable_file(fp);
4410 file = find_writeable_file(fp);
4414 status = nfserr_bad_stateid;
4423 nfs4_unlock_state();
4428 * Test if the stateid is valid
4431 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4432 struct nfsd4_test_stateid *test_stateid)
4434 struct nfsd4_test_stateid_id *stateid;
4435 struct nfs4_client *cl = cstate->session->se_client;
4438 list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
4439 stateid->ts_id_status =
4440 nfsd4_validate_stateid(cl, &stateid->ts_id_stateid);
4441 nfs4_unlock_state();
4447 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4448 struct nfsd4_free_stateid *free_stateid)
4450 stateid_t *stateid = &free_stateid->fr_stateid;
4451 struct nfs4_stid *s;
4452 struct nfs4_delegation *dp;
4453 struct nfs4_ol_stateid *stp;
4454 struct nfs4_client *cl = cstate->session->se_client;
4455 __be32 ret = nfserr_bad_stateid;
4458 spin_lock(&cl->cl_lock);
4459 s = find_stateid_locked(cl, stateid);
4462 switch (s->sc_type) {
4463 case NFS4_DELEG_STID:
4464 ret = nfserr_locks_held;
4466 case NFS4_OPEN_STID:
4467 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
4470 ret = nfserr_locks_held;
4472 case NFS4_LOCK_STID:
4473 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
4476 stp = openlockstateid(s);
4477 ret = nfserr_locks_held;
4478 if (check_for_locks(stp->st_stid.sc_file,
4479 lockowner(stp->st_stateowner)))
4481 unhash_lock_stateid(stp);
4482 spin_unlock(&cl->cl_lock);
4486 case NFS4_REVOKED_DELEG_STID:
4487 dp = delegstateid(s);
4488 list_del_init(&dp->dl_recall_lru);
4489 spin_unlock(&cl->cl_lock);
4493 /* Default falls through and returns nfserr_bad_stateid */
4496 spin_unlock(&cl->cl_lock);
4498 nfs4_unlock_state();
4505 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
4506 RD_STATE : WR_STATE;
4509 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
4511 struct svc_fh *current_fh = &cstate->current_fh;
4512 struct nfs4_stateowner *sop = stp->st_stateowner;
4515 status = nfsd4_check_seqid(cstate, sop, seqid);
4518 if (stp->st_stid.sc_type == NFS4_CLOSED_STID
4519 || stp->st_stid.sc_type == NFS4_REVOKED_DELEG_STID)
4521 * "Closed" stateid's exist *only* to return
4522 * nfserr_replay_me from the previous step, and
4523 * revoked delegations are kept only for free_stateid.
4525 return nfserr_bad_stateid;
4526 status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
4529 return nfs4_check_fh(current_fh, stp);
4533 * Checks for sequence id mutating operations.
4536 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
4537 stateid_t *stateid, char typemask,
4538 struct nfs4_ol_stateid **stpp,
4539 struct nfsd_net *nn)
4542 struct nfs4_stid *s;
4543 struct nfs4_ol_stateid *stp = NULL;
4545 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
4546 seqid, STATEID_VAL(stateid));
4549 status = nfsd4_lookup_stateid(cstate, stateid, typemask, &s, nn);
4552 stp = openlockstateid(s);
4553 nfsd4_cstate_assign_replay(cstate, stp->st_stateowner);
4555 status = nfs4_seqid_op_checks(cstate, stateid, seqid, stp);
4559 nfs4_put_stid(&stp->st_stid);
4563 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
4564 stateid_t *stateid, struct nfs4_ol_stateid **stpp, struct nfsd_net *nn)
4567 struct nfs4_openowner *oo;
4568 struct nfs4_ol_stateid *stp;
4570 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
4571 NFS4_OPEN_STID, &stp, nn);
4574 oo = openowner(stp->st_stateowner);
4575 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
4576 nfs4_put_stid(&stp->st_stid);
4577 return nfserr_bad_stateid;
4584 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4585 struct nfsd4_open_confirm *oc)
4588 struct nfs4_openowner *oo;
4589 struct nfs4_ol_stateid *stp;
4590 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4592 dprintk("NFSD: nfsd4_open_confirm on file %pd\n",
4593 cstate->current_fh.fh_dentry);
4595 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
4601 status = nfs4_preprocess_seqid_op(cstate,
4602 oc->oc_seqid, &oc->oc_req_stateid,
4603 NFS4_OPEN_STID, &stp, nn);
4606 oo = openowner(stp->st_stateowner);
4607 status = nfserr_bad_stateid;
4608 if (oo->oo_flags & NFS4_OO_CONFIRMED)
4610 oo->oo_flags |= NFS4_OO_CONFIRMED;
4611 update_stateid(&stp->st_stid.sc_stateid);
4612 memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4613 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
4614 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
4616 nfsd4_client_record_create(oo->oo_owner.so_client);
4619 nfs4_put_stid(&stp->st_stid);
4621 nfsd4_bump_seqid(cstate, status);
4622 nfs4_unlock_state();
4626 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
4628 if (!test_access(access, stp))
4630 nfs4_file_put_access(stp->st_stid.sc_file, access);
4631 clear_access(access, stp);
4634 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
4636 switch (to_access) {
4637 case NFS4_SHARE_ACCESS_READ:
4638 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
4639 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
4641 case NFS4_SHARE_ACCESS_WRITE:
4642 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
4643 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
4645 case NFS4_SHARE_ACCESS_BOTH:
4653 nfsd4_open_downgrade(struct svc_rqst *rqstp,
4654 struct nfsd4_compound_state *cstate,
4655 struct nfsd4_open_downgrade *od)
4658 struct nfs4_ol_stateid *stp;
4659 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4661 dprintk("NFSD: nfsd4_open_downgrade on file %pd\n",
4662 cstate->current_fh.fh_dentry);
4664 /* We don't yet support WANT bits: */
4665 if (od->od_deleg_want)
4666 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
4670 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
4671 &od->od_stateid, &stp, nn);
4674 status = nfserr_inval;
4675 if (!test_access(od->od_share_access, stp)) {
4676 dprintk("NFSD: access not a subset of current bitmap: 0x%hhx, input access=%08x\n",
4677 stp->st_access_bmap, od->od_share_access);
4680 if (!test_deny(od->od_share_deny, stp)) {
4681 dprintk("NFSD: deny not a subset of current bitmap: 0x%hhx, input deny=%08x\n",
4682 stp->st_deny_bmap, od->od_share_deny);
4685 nfs4_stateid_downgrade(stp, od->od_share_access);
4687 reset_union_bmap_deny(od->od_share_deny, stp);
4689 update_stateid(&stp->st_stid.sc_stateid);
4690 memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4693 nfs4_put_stid(&stp->st_stid);
4695 nfsd4_bump_seqid(cstate, status);
4696 nfs4_unlock_state();
4700 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
4702 struct nfs4_client *clp = s->st_stid.sc_client;
4703 LIST_HEAD(reaplist);
4705 s->st_stid.sc_type = NFS4_CLOSED_STID;
4706 spin_lock(&clp->cl_lock);
4707 unhash_open_stateid(s, &reaplist);
4709 if (clp->cl_minorversion) {
4710 put_ol_stateid_locked(s, &reaplist);
4711 spin_unlock(&clp->cl_lock);
4712 free_ol_stateid_reaplist(&reaplist);
4714 spin_unlock(&clp->cl_lock);
4715 free_ol_stateid_reaplist(&reaplist);
4716 move_to_close_lru(s, clp->net);
4721 * nfs4_unlock_state() called after encode
4724 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4725 struct nfsd4_close *close)
4728 struct nfs4_ol_stateid *stp;
4729 struct net *net = SVC_NET(rqstp);
4730 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4732 dprintk("NFSD: nfsd4_close on file %pd\n",
4733 cstate->current_fh.fh_dentry);
4736 status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
4738 NFS4_OPEN_STID|NFS4_CLOSED_STID,
4740 nfsd4_bump_seqid(cstate, status);
4743 update_stateid(&stp->st_stid.sc_stateid);
4744 memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4746 nfsd4_close_open_stateid(stp);
4748 /* put reference from nfs4_preprocess_seqid_op */
4749 nfs4_put_stid(&stp->st_stid);
4751 nfs4_unlock_state();
4756 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4757 struct nfsd4_delegreturn *dr)
4759 struct nfs4_delegation *dp;
4760 stateid_t *stateid = &dr->dr_stateid;
4761 struct nfs4_stid *s;
4763 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4765 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4769 status = nfsd4_lookup_stateid(cstate, stateid, NFS4_DELEG_STID, &s, nn);
4772 dp = delegstateid(s);
4773 status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
4777 destroy_delegation(dp);
4779 nfs4_put_stid(&dp->dl_stid);
4781 nfs4_unlock_state();
4787 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
4790 end_offset(u64 start, u64 len)
4795 return end >= start ? end: NFS4_MAX_UINT64;
4798 /* last octet in a range */
4800 last_byte_offset(u64 start, u64 len)
4806 return end > start ? end - 1: NFS4_MAX_UINT64;
4810 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
4811 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
4812 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
4813 * locking, this prevents us from being completely protocol-compliant. The
4814 * real solution to this problem is to start using unsigned file offsets in
4815 * the VFS, but this is a very deep change!
4818 nfs4_transform_lock_offset(struct file_lock *lock)
4820 if (lock->fl_start < 0)
4821 lock->fl_start = OFFSET_MAX;
4822 if (lock->fl_end < 0)
4823 lock->fl_end = OFFSET_MAX;
4826 /* Hack!: For now, we're defining this just so we can use a pointer to it
4827 * as a unique cookie to identify our (NFSv4's) posix locks. */
4828 static const struct lock_manager_operations nfsd_posix_mng_ops = {
4832 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
4834 struct nfs4_lockowner *lo;
4836 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
4837 lo = (struct nfs4_lockowner *) fl->fl_owner;
4838 deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
4839 lo->lo_owner.so_owner.len, GFP_KERNEL);
4840 if (!deny->ld_owner.data)
4841 /* We just don't care that much */
4843 deny->ld_owner.len = lo->lo_owner.so_owner.len;
4844 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
4847 deny->ld_owner.len = 0;
4848 deny->ld_owner.data = NULL;
4849 deny->ld_clientid.cl_boot = 0;
4850 deny->ld_clientid.cl_id = 0;
4852 deny->ld_start = fl->fl_start;
4853 deny->ld_length = NFS4_MAX_UINT64;
4854 if (fl->fl_end != NFS4_MAX_UINT64)
4855 deny->ld_length = fl->fl_end - fl->fl_start + 1;
4856 deny->ld_type = NFS4_READ_LT;
4857 if (fl->fl_type != F_RDLCK)
4858 deny->ld_type = NFS4_WRITE_LT;
4861 static struct nfs4_lockowner *
4862 find_lockowner_str_locked(clientid_t *clid, struct xdr_netobj *owner,
4863 struct nfs4_client *clp)
4865 unsigned int strhashval = ownerstr_hashval(owner);
4866 struct nfs4_stateowner *so;
4868 list_for_each_entry(so, &clp->cl_ownerstr_hashtbl[strhashval],
4870 if (so->so_is_open_owner)
4872 if (!same_owner_str(so, owner))
4874 atomic_inc(&so->so_count);
4875 return lockowner(so);
4880 static struct nfs4_lockowner *
4881 find_lockowner_str(clientid_t *clid, struct xdr_netobj *owner,
4882 struct nfs4_client *clp)
4884 struct nfs4_lockowner *lo;
4886 spin_lock(&clp->cl_lock);
4887 lo = find_lockowner_str_locked(clid, owner, clp);
4888 spin_unlock(&clp->cl_lock);
4892 static void nfs4_unhash_lockowner(struct nfs4_stateowner *sop)
4894 unhash_lockowner_locked(lockowner(sop));
4897 static void nfs4_free_lockowner(struct nfs4_stateowner *sop)
4899 struct nfs4_lockowner *lo = lockowner(sop);
4901 kmem_cache_free(lockowner_slab, lo);
4904 static const struct nfs4_stateowner_operations lockowner_ops = {
4905 .so_unhash = nfs4_unhash_lockowner,
4906 .so_free = nfs4_free_lockowner,
4910 * Alloc a lock owner structure.
4911 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
4914 * strhashval = ownerstr_hashval
4916 static struct nfs4_lockowner *
4917 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp,
4918 struct nfs4_ol_stateid *open_stp,
4919 struct nfsd4_lock *lock)
4921 struct nfs4_lockowner *lo, *ret;
4923 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
4926 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
4927 lo->lo_owner.so_is_open_owner = 0;
4928 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid;
4929 lo->lo_owner.so_ops = &lockowner_ops;
4930 spin_lock(&clp->cl_lock);
4931 ret = find_lockowner_str_locked(&clp->cl_clientid,
4932 &lock->lk_new_owner, clp);
4934 list_add(&lo->lo_owner.so_strhash,
4935 &clp->cl_ownerstr_hashtbl[strhashval]);
4938 nfs4_free_lockowner(&lo->lo_owner);
4939 spin_unlock(&clp->cl_lock);
4944 init_lock_stateid(struct nfs4_ol_stateid *stp, struct nfs4_lockowner *lo,
4945 struct nfs4_file *fp, struct inode *inode,
4946 struct nfs4_ol_stateid *open_stp)
4948 struct nfs4_client *clp = lo->lo_owner.so_client;
4950 lockdep_assert_held(&clp->cl_lock);
4952 atomic_inc(&stp->st_stid.sc_count);
4953 stp->st_stid.sc_type = NFS4_LOCK_STID;
4954 stp->st_stateowner = &lo->lo_owner;
4955 atomic_inc(&lo->lo_owner.so_count);
4957 stp->st_stid.sc_file = fp;
4958 stp->st_stid.sc_free = nfs4_free_lock_stateid;
4959 stp->st_access_bmap = 0;
4960 stp->st_deny_bmap = open_stp->st_deny_bmap;
4961 stp->st_openstp = open_stp;
4962 list_add(&stp->st_locks, &open_stp->st_locks);
4963 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
4964 spin_lock(&fp->fi_lock);
4965 list_add(&stp->st_perfile, &fp->fi_stateids);
4966 spin_unlock(&fp->fi_lock);
4969 static struct nfs4_ol_stateid *
4970 find_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp)
4972 struct nfs4_ol_stateid *lst;
4973 struct nfs4_client *clp = lo->lo_owner.so_client;
4975 lockdep_assert_held(&clp->cl_lock);
4977 list_for_each_entry(lst, &lo->lo_owner.so_stateids, st_perstateowner) {
4978 if (lst->st_stid.sc_file == fp) {
4979 atomic_inc(&lst->st_stid.sc_count);
4986 static struct nfs4_ol_stateid *
4987 find_or_create_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fi,
4988 struct inode *inode, struct nfs4_ol_stateid *ost,
4991 struct nfs4_stid *ns = NULL;
4992 struct nfs4_ol_stateid *lst;
4993 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
4994 struct nfs4_client *clp = oo->oo_owner.so_client;
4996 spin_lock(&clp->cl_lock);
4997 lst = find_lock_stateid(lo, fi);
4999 spin_unlock(&clp->cl_lock);
5000 ns = nfs4_alloc_stid(clp, stateid_slab);
5004 spin_lock(&clp->cl_lock);
5005 lst = find_lock_stateid(lo, fi);
5007 lst = openlockstateid(ns);
5008 init_lock_stateid(lst, lo, fi, inode, ost);
5013 spin_unlock(&clp->cl_lock);
5020 check_lock_length(u64 offset, u64 length)
5022 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
5023 LOFF_OVERFLOW(offset, length)));
5026 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
5028 struct nfs4_file *fp = lock_stp->st_stid.sc_file;
5030 lockdep_assert_held(&fp->fi_lock);
5032 if (test_access(access, lock_stp))
5034 __nfs4_file_get_access(fp, access);
5035 set_access(access, lock_stp);
5039 lookup_or_create_lock_state(struct nfsd4_compound_state *cstate,
5040 struct nfs4_ol_stateid *ost,
5041 struct nfsd4_lock *lock,
5042 struct nfs4_ol_stateid **lst, bool *new)
5045 struct nfs4_file *fi = ost->st_stid.sc_file;
5046 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
5047 struct nfs4_client *cl = oo->oo_owner.so_client;
5048 struct inode *inode = cstate->current_fh.fh_dentry->d_inode;
5049 struct nfs4_lockowner *lo;
5050 unsigned int strhashval;
5052 lo = find_lockowner_str(&cl->cl_clientid, &lock->v.new.owner, cl);
5054 strhashval = ownerstr_hashval(&lock->v.new.owner);
5055 lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
5057 return nfserr_jukebox;
5059 /* with an existing lockowner, seqids must be the same */
5060 status = nfserr_bad_seqid;
5061 if (!cstate->minorversion &&
5062 lock->lk_new_lock_seqid != lo->lo_owner.so_seqid)
5066 *lst = find_or_create_lock_stateid(lo, fi, inode, ost, new);
5068 status = nfserr_jukebox;
5073 nfs4_put_stateowner(&lo->lo_owner);
5081 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5082 struct nfsd4_lock *lock)
5084 struct nfs4_openowner *open_sop = NULL;
5085 struct nfs4_lockowner *lock_sop = NULL;
5086 struct nfs4_ol_stateid *lock_stp = NULL;
5087 struct nfs4_ol_stateid *open_stp = NULL;
5088 struct nfs4_file *fp;
5089 struct file *filp = NULL;
5090 struct file_lock *file_lock = NULL;
5091 struct file_lock *conflock = NULL;
5096 struct net *net = SVC_NET(rqstp);
5097 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5099 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
5100 (long long) lock->lk_offset,
5101 (long long) lock->lk_length);
5103 if (check_lock_length(lock->lk_offset, lock->lk_length))
5104 return nfserr_inval;
5106 if ((status = fh_verify(rqstp, &cstate->current_fh,
5107 S_IFREG, NFSD_MAY_LOCK))) {
5108 dprintk("NFSD: nfsd4_lock: permission denied!\n");
5114 if (lock->lk_is_new) {
5115 if (nfsd4_has_session(cstate))
5116 /* See rfc 5661 18.10.3: given clientid is ignored: */
5117 memcpy(&lock->v.new.clientid,
5118 &cstate->session->se_client->cl_clientid,
5119 sizeof(clientid_t));
5121 status = nfserr_stale_clientid;
5122 if (STALE_CLIENTID(&lock->lk_new_clientid, nn))
5125 /* validate and update open stateid and open seqid */
5126 status = nfs4_preprocess_confirmed_seqid_op(cstate,
5127 lock->lk_new_open_seqid,
5128 &lock->lk_new_open_stateid,
5132 open_sop = openowner(open_stp->st_stateowner);
5133 status = nfserr_bad_stateid;
5134 if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
5135 &lock->v.new.clientid))
5137 status = lookup_or_create_lock_state(cstate, open_stp, lock,
5140 status = nfs4_preprocess_seqid_op(cstate,
5141 lock->lk_old_lock_seqid,
5142 &lock->lk_old_lock_stateid,
5143 NFS4_LOCK_STID, &lock_stp, nn);
5147 lock_sop = lockowner(lock_stp->st_stateowner);
5149 lkflg = setlkflg(lock->lk_type);
5150 status = nfs4_check_openmode(lock_stp, lkflg);
5154 status = nfserr_grace;
5155 if (locks_in_grace(net) && !lock->lk_reclaim)
5157 status = nfserr_no_grace;
5158 if (!locks_in_grace(net) && lock->lk_reclaim)
5161 file_lock = locks_alloc_lock();
5163 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
5164 status = nfserr_jukebox;
5168 fp = lock_stp->st_stid.sc_file;
5169 locks_init_lock(file_lock);
5170 switch (lock->lk_type) {
5173 spin_lock(&fp->fi_lock);
5174 filp = find_readable_file_locked(fp);
5176 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
5177 spin_unlock(&fp->fi_lock);
5178 file_lock->fl_type = F_RDLCK;
5181 case NFS4_WRITEW_LT:
5182 spin_lock(&fp->fi_lock);
5183 filp = find_writeable_file_locked(fp);
5185 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
5186 spin_unlock(&fp->fi_lock);
5187 file_lock->fl_type = F_WRLCK;
5190 status = nfserr_inval;
5194 status = nfserr_openmode;
5197 file_lock->fl_owner = (fl_owner_t)lock_sop;
5198 file_lock->fl_pid = current->tgid;
5199 file_lock->fl_file = filp;
5200 file_lock->fl_flags = FL_POSIX;
5201 file_lock->fl_lmops = &nfsd_posix_mng_ops;
5202 file_lock->fl_start = lock->lk_offset;
5203 file_lock->fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
5204 nfs4_transform_lock_offset(file_lock);
5206 conflock = locks_alloc_lock();
5208 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
5209 status = nfserr_jukebox;
5213 err = vfs_lock_file(filp, F_SETLK, file_lock, conflock);
5215 case 0: /* success! */
5216 update_stateid(&lock_stp->st_stid.sc_stateid);
5217 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stid.sc_stateid,
5221 case (EAGAIN): /* conflock holds conflicting lock */
5222 status = nfserr_denied;
5223 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
5224 nfs4_set_lock_denied(conflock, &lock->lk_denied);
5227 status = nfserr_deadlock;
5230 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
5231 status = nfserrno(err);
5238 /* Bump seqid manually if the 4.0 replay owner is openowner */
5239 if (cstate->replay_owner &&
5240 cstate->replay_owner != &lock_sop->lo_owner &&
5241 seqid_mutating_err(ntohl(status)))
5242 lock_sop->lo_owner.so_seqid++;
5245 * If this is a new, never-before-used stateid, and we are
5246 * returning an error, then just go ahead and release it.
5249 release_lock_stateid(lock_stp);
5251 nfs4_put_stid(&lock_stp->st_stid);
5254 nfs4_put_stid(&open_stp->st_stid);
5255 nfsd4_bump_seqid(cstate, status);
5256 nfs4_unlock_state();
5258 locks_free_lock(file_lock);
5260 locks_free_lock(conflock);
5265 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
5266 * so we do a temporary open here just to get an open file to pass to
5267 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
5270 static __be32 nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
5273 __be32 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
5275 err = nfserrno(vfs_test_lock(file, lock));
5285 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5286 struct nfsd4_lockt *lockt)
5288 struct file_lock *file_lock = NULL;
5289 struct nfs4_lockowner *lo = NULL;
5291 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
5293 if (locks_in_grace(SVC_NET(rqstp)))
5294 return nfserr_grace;
5296 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
5297 return nfserr_inval;
5301 if (!nfsd4_has_session(cstate)) {
5302 status = lookup_clientid(&lockt->lt_clientid, cstate, nn);
5307 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
5310 file_lock = locks_alloc_lock();
5312 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
5313 status = nfserr_jukebox;
5316 locks_init_lock(file_lock);
5317 switch (lockt->lt_type) {
5320 file_lock->fl_type = F_RDLCK;
5323 case NFS4_WRITEW_LT:
5324 file_lock->fl_type = F_WRLCK;
5327 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
5328 status = nfserr_inval;
5332 lo = find_lockowner_str(&lockt->lt_clientid, &lockt->lt_owner,
5335 file_lock->fl_owner = (fl_owner_t)lo;
5336 file_lock->fl_pid = current->tgid;
5337 file_lock->fl_flags = FL_POSIX;
5339 file_lock->fl_start = lockt->lt_offset;
5340 file_lock->fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
5342 nfs4_transform_lock_offset(file_lock);
5344 status = nfsd_test_lock(rqstp, &cstate->current_fh, file_lock);
5348 if (file_lock->fl_type != F_UNLCK) {
5349 status = nfserr_denied;
5350 nfs4_set_lock_denied(file_lock, &lockt->lt_denied);
5354 nfs4_put_stateowner(&lo->lo_owner);
5355 nfs4_unlock_state();
5357 locks_free_lock(file_lock);
5362 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5363 struct nfsd4_locku *locku)
5365 struct nfs4_ol_stateid *stp;
5366 struct file *filp = NULL;
5367 struct file_lock *file_lock = NULL;
5370 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
5372 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
5373 (long long) locku->lu_offset,
5374 (long long) locku->lu_length);
5376 if (check_lock_length(locku->lu_offset, locku->lu_length))
5377 return nfserr_inval;
5381 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
5382 &locku->lu_stateid, NFS4_LOCK_STID,
5386 filp = find_any_file(stp->st_stid.sc_file);
5388 status = nfserr_lock_range;
5391 file_lock = locks_alloc_lock();
5393 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
5394 status = nfserr_jukebox;
5397 locks_init_lock(file_lock);
5398 file_lock->fl_type = F_UNLCK;
5399 file_lock->fl_owner = (fl_owner_t)lockowner(stp->st_stateowner);
5400 file_lock->fl_pid = current->tgid;
5401 file_lock->fl_file = filp;
5402 file_lock->fl_flags = FL_POSIX;
5403 file_lock->fl_lmops = &nfsd_posix_mng_ops;
5404 file_lock->fl_start = locku->lu_offset;
5406 file_lock->fl_end = last_byte_offset(locku->lu_offset,
5408 nfs4_transform_lock_offset(file_lock);
5410 err = vfs_lock_file(filp, F_SETLK, file_lock, NULL);
5412 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
5415 update_stateid(&stp->st_stid.sc_stateid);
5416 memcpy(&locku->lu_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
5420 nfs4_put_stid(&stp->st_stid);
5422 nfsd4_bump_seqid(cstate, status);
5423 nfs4_unlock_state();
5425 locks_free_lock(file_lock);
5429 status = nfserrno(err);
5435 * true: locks held by lockowner
5436 * false: no locks held by lockowner
5439 check_for_locks(struct nfs4_file *fp, struct nfs4_lockowner *lowner)
5441 struct file_lock **flpp;
5443 struct file *filp = find_any_file(fp);
5444 struct inode *inode;
5447 /* Any valid lock stateid should have some sort of access */
5452 inode = file_inode(filp);
5454 spin_lock(&inode->i_lock);
5455 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
5456 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
5461 spin_unlock(&inode->i_lock);
5467 nfsd4_release_lockowner(struct svc_rqst *rqstp,
5468 struct nfsd4_compound_state *cstate,
5469 struct nfsd4_release_lockowner *rlockowner)
5471 clientid_t *clid = &rlockowner->rl_clientid;
5472 struct nfs4_stateowner *sop;
5473 struct nfs4_lockowner *lo = NULL;
5474 struct nfs4_ol_stateid *stp;
5475 struct xdr_netobj *owner = &rlockowner->rl_owner;
5476 unsigned int hashval = ownerstr_hashval(owner);
5478 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
5479 struct nfs4_client *clp;
5481 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
5482 clid->cl_boot, clid->cl_id);
5486 status = lookup_clientid(clid, cstate, nn);
5491 /* Find the matching lock stateowner */
5492 spin_lock(&clp->cl_lock);
5493 list_for_each_entry(sop, &clp->cl_ownerstr_hashtbl[hashval],
5496 if (sop->so_is_open_owner || !same_owner_str(sop, owner))
5499 /* see if there are still any locks associated with it */
5500 lo = lockowner(sop);
5501 list_for_each_entry(stp, &sop->so_stateids, st_perstateowner) {
5502 if (check_for_locks(stp->st_stid.sc_file, lo)) {
5503 status = nfserr_locks_held;
5504 spin_unlock(&clp->cl_lock);
5509 atomic_inc(&sop->so_count);
5512 spin_unlock(&clp->cl_lock);
5514 release_lockowner(lo);
5516 nfs4_unlock_state();
5520 static inline struct nfs4_client_reclaim *
5523 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
5527 nfs4_has_reclaimed_state(const char *name, struct nfsd_net *nn)
5529 struct nfs4_client_reclaim *crp;
5531 crp = nfsd4_find_reclaim_client(name, nn);
5532 return (crp && crp->cr_clp);
5536 * failure => all reset bets are off, nfserr_no_grace...
5538 struct nfs4_client_reclaim *
5539 nfs4_client_to_reclaim(const char *name, struct nfsd_net *nn)
5541 unsigned int strhashval;
5542 struct nfs4_client_reclaim *crp;
5544 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
5545 crp = alloc_reclaim();
5547 strhashval = clientstr_hashval(name);
5548 INIT_LIST_HEAD(&crp->cr_strhash);
5549 list_add(&crp->cr_strhash, &nn->reclaim_str_hashtbl[strhashval]);
5550 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
5552 nn->reclaim_str_hashtbl_size++;
5558 nfs4_remove_reclaim_record(struct nfs4_client_reclaim *crp, struct nfsd_net *nn)
5560 list_del(&crp->cr_strhash);
5562 nn->reclaim_str_hashtbl_size--;
5566 nfs4_release_reclaim(struct nfsd_net *nn)
5568 struct nfs4_client_reclaim *crp = NULL;
5571 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
5572 while (!list_empty(&nn->reclaim_str_hashtbl[i])) {
5573 crp = list_entry(nn->reclaim_str_hashtbl[i].next,
5574 struct nfs4_client_reclaim, cr_strhash);
5575 nfs4_remove_reclaim_record(crp, nn);
5578 WARN_ON_ONCE(nn->reclaim_str_hashtbl_size);
5582 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
5583 struct nfs4_client_reclaim *
5584 nfsd4_find_reclaim_client(const char *recdir, struct nfsd_net *nn)
5586 unsigned int strhashval;
5587 struct nfs4_client_reclaim *crp = NULL;
5589 dprintk("NFSD: nfs4_find_reclaim_client for recdir %s\n", recdir);
5591 strhashval = clientstr_hashval(recdir);
5592 list_for_each_entry(crp, &nn->reclaim_str_hashtbl[strhashval], cr_strhash) {
5593 if (same_name(crp->cr_recdir, recdir)) {
5601 * Called from OPEN. Look for clientid in reclaim list.
5604 nfs4_check_open_reclaim(clientid_t *clid,
5605 struct nfsd4_compound_state *cstate,
5606 struct nfsd_net *nn)
5610 /* find clientid in conf_id_hashtbl */
5611 status = lookup_clientid(clid, cstate, nn);
5613 return nfserr_reclaim_bad;
5615 if (nfsd4_client_record_check(cstate->clp))
5616 return nfserr_reclaim_bad;
5621 #ifdef CONFIG_NFSD_FAULT_INJECTION
5623 u64 nfsd_forget_client(struct nfs4_client *clp, u64 max)
5625 if (mark_client_expired(clp))
5631 u64 nfsd_print_client(struct nfs4_client *clp, u64 num)
5633 char buf[INET6_ADDRSTRLEN];
5634 rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, sizeof(buf));
5635 printk(KERN_INFO "NFS Client: %s\n", buf);
5639 static void nfsd_print_count(struct nfs4_client *clp, unsigned int count,
5642 char buf[INET6_ADDRSTRLEN];
5643 rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, sizeof(buf));
5644 printk(KERN_INFO "NFS Client: %s has %u %s\n", buf, count, type);
5647 static u64 nfsd_foreach_client_lock(struct nfs4_client *clp, u64 max,
5648 void (*func)(struct nfs4_ol_stateid *))
5650 struct nfs4_openowner *oop;
5651 struct nfs4_ol_stateid *stp, *st_next;
5652 struct nfs4_ol_stateid *lst, *lst_next;
5655 list_for_each_entry(oop, &clp->cl_openowners, oo_perclient) {
5656 list_for_each_entry_safe(stp, st_next,
5657 &oop->oo_owner.so_stateids, st_perstateowner) {
5658 list_for_each_entry_safe(lst, lst_next,
5659 &stp->st_locks, st_locks) {
5671 u64 nfsd_forget_client_locks(struct nfs4_client *clp, u64 max)
5673 return nfsd_foreach_client_lock(clp, max, release_lock_stateid);
5676 u64 nfsd_print_client_locks(struct nfs4_client *clp, u64 max)
5678 u64 count = nfsd_foreach_client_lock(clp, max, NULL);
5679 nfsd_print_count(clp, count, "locked files");
5683 static u64 nfsd_foreach_client_open(struct nfs4_client *clp, u64 max, void (*func)(struct nfs4_openowner *))
5685 struct nfs4_openowner *oop, *next;
5688 list_for_each_entry_safe(oop, next, &clp->cl_openowners, oo_perclient) {
5698 u64 nfsd_forget_client_openowners(struct nfs4_client *clp, u64 max)
5700 return nfsd_foreach_client_open(clp, max, release_openowner);
5703 u64 nfsd_print_client_openowners(struct nfs4_client *clp, u64 max)
5705 u64 count = nfsd_foreach_client_open(clp, max, NULL);
5706 nfsd_print_count(clp, count, "open files");
5710 static u64 nfsd_find_all_delegations(struct nfs4_client *clp, u64 max,
5711 struct list_head *victims)
5713 struct nfs4_delegation *dp, *next;
5716 lockdep_assert_held(&state_lock);
5717 list_for_each_entry_safe(dp, next, &clp->cl_delegations, dl_perclnt) {
5720 * It's not safe to mess with delegations that have a
5721 * non-zero dl_time. They might have already been broken
5722 * and could be processed by the laundromat outside of
5723 * the state_lock. Just leave them be.
5725 if (dp->dl_time != 0)
5728 unhash_delegation_locked(dp);
5729 list_add(&dp->dl_recall_lru, victims);
5737 u64 nfsd_forget_client_delegations(struct nfs4_client *clp, u64 max)
5739 struct nfs4_delegation *dp, *next;
5743 spin_lock(&state_lock);
5744 count = nfsd_find_all_delegations(clp, max, &victims);
5745 spin_unlock(&state_lock);
5747 list_for_each_entry_safe(dp, next, &victims, dl_recall_lru) {
5748 list_del_init(&dp->dl_recall_lru);
5749 revoke_delegation(dp);
5755 u64 nfsd_recall_client_delegations(struct nfs4_client *clp, u64 max)
5757 struct nfs4_delegation *dp;
5761 spin_lock(&state_lock);
5762 count = nfsd_find_all_delegations(clp, max, &victims);
5763 while (!list_empty(&victims)) {
5764 dp = list_first_entry(&victims, struct nfs4_delegation,
5766 list_del_init(&dp->dl_recall_lru);
5768 nfsd_break_one_deleg(dp);
5770 spin_unlock(&state_lock);
5775 u64 nfsd_print_client_delegations(struct nfs4_client *clp, u64 max)
5779 spin_lock(&state_lock);
5780 count = nfsd_find_all_delegations(clp, max, NULL);
5781 spin_unlock(&state_lock);
5783 nfsd_print_count(clp, count, "delegations");
5787 u64 nfsd_for_n_state(u64 max, u64 (*func)(struct nfs4_client *, u64))
5789 struct nfs4_client *clp, *next;
5791 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns, nfsd_net_id);
5793 if (!nfsd_netns_ready(nn))
5796 list_for_each_entry_safe(clp, next, &nn->client_lru, cl_lru) {
5797 count += func(clp, max - count);
5798 if ((max != 0) && (count >= max))
5805 struct nfs4_client *nfsd_find_client(struct sockaddr_storage *addr, size_t addr_size)
5807 struct nfs4_client *clp;
5808 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns, nfsd_net_id);
5810 if (!nfsd_netns_ready(nn))
5813 list_for_each_entry(clp, &nn->client_lru, cl_lru) {
5814 if (memcmp(&clp->cl_addr, addr, addr_size) == 0)
5820 #endif /* CONFIG_NFSD_FAULT_INJECTION */
5823 * Since the lifetime of a delegation isn't limited to that of an open, a
5824 * client may quite reasonably hang on to a delegation as long as it has
5825 * the inode cached. This becomes an obvious problem the first time a
5826 * client's inode cache approaches the size of the server's total memory.
5828 * For now we avoid this problem by imposing a hard limit on the number
5829 * of delegations, which varies according to the server's memory size.
5832 set_max_delegations(void)
5835 * Allow at most 4 delegations per megabyte of RAM. Quick
5836 * estimates suggest that in the worst case (where every delegation
5837 * is for a different inode), a delegation could take about 1.5K,
5838 * giving a worst case usage of about 6% of memory.
5840 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
5843 static int nfs4_state_create_net(struct net *net)
5845 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5848 nn->conf_id_hashtbl = kmalloc(sizeof(struct list_head) *
5849 CLIENT_HASH_SIZE, GFP_KERNEL);
5850 if (!nn->conf_id_hashtbl)
5852 nn->unconf_id_hashtbl = kmalloc(sizeof(struct list_head) *
5853 CLIENT_HASH_SIZE, GFP_KERNEL);
5854 if (!nn->unconf_id_hashtbl)
5856 nn->sessionid_hashtbl = kmalloc(sizeof(struct list_head) *
5857 SESSION_HASH_SIZE, GFP_KERNEL);
5858 if (!nn->sessionid_hashtbl)
5861 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
5862 INIT_LIST_HEAD(&nn->conf_id_hashtbl[i]);
5863 INIT_LIST_HEAD(&nn->unconf_id_hashtbl[i]);
5865 for (i = 0; i < SESSION_HASH_SIZE; i++)
5866 INIT_LIST_HEAD(&nn->sessionid_hashtbl[i]);
5867 nn->conf_name_tree = RB_ROOT;
5868 nn->unconf_name_tree = RB_ROOT;
5869 INIT_LIST_HEAD(&nn->client_lru);
5870 INIT_LIST_HEAD(&nn->close_lru);
5871 INIT_LIST_HEAD(&nn->del_recall_lru);
5872 spin_lock_init(&nn->client_lock);
5874 INIT_DELAYED_WORK(&nn->laundromat_work, laundromat_main);
5880 kfree(nn->unconf_id_hashtbl);
5882 kfree(nn->conf_id_hashtbl);
5888 nfs4_state_destroy_net(struct net *net)
5891 struct nfs4_client *clp = NULL;
5892 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5894 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
5895 while (!list_empty(&nn->conf_id_hashtbl[i])) {
5896 clp = list_entry(nn->conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
5897 destroy_client(clp);
5901 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
5902 while (!list_empty(&nn->unconf_id_hashtbl[i])) {
5903 clp = list_entry(nn->unconf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
5904 destroy_client(clp);
5908 kfree(nn->sessionid_hashtbl);
5909 kfree(nn->unconf_id_hashtbl);
5910 kfree(nn->conf_id_hashtbl);
5915 nfs4_state_start_net(struct net *net)
5917 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5920 ret = nfs4_state_create_net(net);
5923 nfsd4_client_tracking_init(net);
5924 nn->boot_time = get_seconds();
5925 locks_start_grace(net, &nn->nfsd4_manager);
5926 nn->grace_ended = false;
5927 printk(KERN_INFO "NFSD: starting %ld-second grace period (net %p)\n",
5928 nn->nfsd4_grace, net);
5929 queue_delayed_work(laundry_wq, &nn->laundromat_work, nn->nfsd4_grace * HZ);
5933 /* initialization to perform when the nfsd service is started: */
5936 nfs4_state_start(void)
5940 ret = set_callback_cred();
5943 laundry_wq = create_singlethread_workqueue("nfsd4");
5944 if (laundry_wq == NULL) {
5948 ret = nfsd4_create_callback_queue();
5950 goto out_free_laundry;
5952 set_max_delegations();
5957 destroy_workqueue(laundry_wq);
5963 nfs4_state_shutdown_net(struct net *net)
5965 struct nfs4_delegation *dp = NULL;
5966 struct list_head *pos, *next, reaplist;
5967 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5969 cancel_delayed_work_sync(&nn->laundromat_work);
5970 locks_end_grace(&nn->nfsd4_manager);
5973 INIT_LIST_HEAD(&reaplist);
5974 spin_lock(&state_lock);
5975 list_for_each_safe(pos, next, &nn->del_recall_lru) {
5976 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
5977 unhash_delegation_locked(dp);
5978 list_add(&dp->dl_recall_lru, &reaplist);
5980 spin_unlock(&state_lock);
5981 list_for_each_safe(pos, next, &reaplist) {
5982 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
5983 list_del_init(&dp->dl_recall_lru);
5984 nfs4_put_stid(&dp->dl_stid);
5987 nfsd4_client_tracking_exit(net);
5988 nfs4_state_destroy_net(net);
5989 nfs4_unlock_state();
5993 nfs4_state_shutdown(void)
5995 destroy_workqueue(laundry_wq);
5996 nfsd4_destroy_callback_queue();
6000 get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
6002 if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG) && CURRENT_STATEID(stateid))
6003 memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
6007 put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
6009 if (cstate->minorversion) {
6010 memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
6011 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
6016 clear_current_stateid(struct nfsd4_compound_state *cstate)
6018 CLEAR_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
6022 * functions to set current state id
6025 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
6027 put_stateid(cstate, &odp->od_stateid);
6031 nfsd4_set_openstateid(struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
6033 put_stateid(cstate, &open->op_stateid);
6037 nfsd4_set_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
6039 put_stateid(cstate, &close->cl_stateid);
6043 nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate, struct nfsd4_lock *lock)
6045 put_stateid(cstate, &lock->lk_resp_stateid);
6049 * functions to consume current state id
6053 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
6055 get_stateid(cstate, &odp->od_stateid);
6059 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate, struct nfsd4_delegreturn *drp)
6061 get_stateid(cstate, &drp->dr_stateid);
6065 nfsd4_get_freestateid(struct nfsd4_compound_state *cstate, struct nfsd4_free_stateid *fsp)
6067 get_stateid(cstate, &fsp->fr_stateid);
6071 nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate, struct nfsd4_setattr *setattr)
6073 get_stateid(cstate, &setattr->sa_stateid);
6077 nfsd4_get_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
6079 get_stateid(cstate, &close->cl_stateid);
6083 nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate, struct nfsd4_locku *locku)
6085 get_stateid(cstate, &locku->lu_stateid);
6089 nfsd4_get_readstateid(struct nfsd4_compound_state *cstate, struct nfsd4_read *read)
6091 get_stateid(cstate, &read->rd_stateid);
6095 nfsd4_get_writestateid(struct nfsd4_compound_state *cstate, struct nfsd4_write *write)
6097 get_stateid(cstate, &write->wr_stateid);