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/jhash.h>
48 #include "current_stateid.h"
53 #define NFSDDBG_FACILITY NFSDDBG_PROC
55 #define all_ones {{~0,~0},~0}
56 static const stateid_t one_stateid = {
58 .si_opaque = all_ones,
60 static const stateid_t zero_stateid = {
63 static const stateid_t currentstateid = {
67 static u64 current_sessionid = 1;
69 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zero_stateid, sizeof(stateid_t)))
70 #define ONE_STATEID(stateid) (!memcmp((stateid), &one_stateid, sizeof(stateid_t)))
71 #define CURRENT_STATEID(stateid) (!memcmp((stateid), ¤tstateid, sizeof(stateid_t)))
73 /* forward declarations */
74 static bool check_for_locks(struct nfs4_file *fp, struct nfs4_lockowner *lowner);
75 static void nfs4_free_ol_stateid(struct nfs4_stid *stid);
80 * Currently used for the del_recall_lru and file hash table. In an
81 * effort to decrease the scope of the client_mutex, this spinlock may
82 * eventually cover more:
84 static DEFINE_SPINLOCK(state_lock);
87 * A waitqueue for all in-progress 4.0 CLOSE operations that are waiting for
88 * the refcount on the open stateid to drop.
90 static DECLARE_WAIT_QUEUE_HEAD(close_wq);
92 static struct kmem_cache *openowner_slab;
93 static struct kmem_cache *lockowner_slab;
94 static struct kmem_cache *file_slab;
95 static struct kmem_cache *stateid_slab;
96 static struct kmem_cache *deleg_slab;
97 static struct kmem_cache *odstate_slab;
99 static void free_session(struct nfsd4_session *);
101 static struct nfsd4_callback_ops nfsd4_cb_recall_ops;
103 static bool is_session_dead(struct nfsd4_session *ses)
105 return ses->se_flags & NFS4_SESSION_DEAD;
108 static __be32 mark_session_dead_locked(struct nfsd4_session *ses, int ref_held_by_me)
110 if (atomic_read(&ses->se_ref) > ref_held_by_me)
111 return nfserr_jukebox;
112 ses->se_flags |= NFS4_SESSION_DEAD;
116 static bool is_client_expired(struct nfs4_client *clp)
118 return clp->cl_time == 0;
121 static __be32 get_client_locked(struct nfs4_client *clp)
123 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
125 lockdep_assert_held(&nn->client_lock);
127 if (is_client_expired(clp))
128 return nfserr_expired;
129 atomic_inc(&clp->cl_refcount);
133 /* must be called under the client_lock */
135 renew_client_locked(struct nfs4_client *clp)
137 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
139 if (is_client_expired(clp)) {
141 printk("%s: client (clientid %08x/%08x) already expired\n",
143 clp->cl_clientid.cl_boot,
144 clp->cl_clientid.cl_id);
148 dprintk("renewing client (clientid %08x/%08x)\n",
149 clp->cl_clientid.cl_boot,
150 clp->cl_clientid.cl_id);
151 list_move_tail(&clp->cl_lru, &nn->client_lru);
152 clp->cl_time = get_seconds();
155 static void put_client_renew_locked(struct nfs4_client *clp)
157 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
159 lockdep_assert_held(&nn->client_lock);
161 if (!atomic_dec_and_test(&clp->cl_refcount))
163 if (!is_client_expired(clp))
164 renew_client_locked(clp);
167 static void put_client_renew(struct nfs4_client *clp)
169 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
171 if (!atomic_dec_and_lock(&clp->cl_refcount, &nn->client_lock))
173 if (!is_client_expired(clp))
174 renew_client_locked(clp);
175 spin_unlock(&nn->client_lock);
178 static __be32 nfsd4_get_session_locked(struct nfsd4_session *ses)
182 if (is_session_dead(ses))
183 return nfserr_badsession;
184 status = get_client_locked(ses->se_client);
187 atomic_inc(&ses->se_ref);
191 static void nfsd4_put_session_locked(struct nfsd4_session *ses)
193 struct nfs4_client *clp = ses->se_client;
194 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
196 lockdep_assert_held(&nn->client_lock);
198 if (atomic_dec_and_test(&ses->se_ref) && is_session_dead(ses))
200 put_client_renew_locked(clp);
203 static void nfsd4_put_session(struct nfsd4_session *ses)
205 struct nfs4_client *clp = ses->se_client;
206 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
208 spin_lock(&nn->client_lock);
209 nfsd4_put_session_locked(ses);
210 spin_unlock(&nn->client_lock);
213 static inline struct nfs4_stateowner *
214 nfs4_get_stateowner(struct nfs4_stateowner *sop)
216 atomic_inc(&sop->so_count);
221 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner)
223 return (sop->so_owner.len == owner->len) &&
224 0 == memcmp(sop->so_owner.data, owner->data, owner->len);
227 static struct nfs4_openowner *
228 find_openstateowner_str_locked(unsigned int hashval, struct nfsd4_open *open,
229 struct nfs4_client *clp)
231 struct nfs4_stateowner *so;
233 lockdep_assert_held(&clp->cl_lock);
235 list_for_each_entry(so, &clp->cl_ownerstr_hashtbl[hashval],
237 if (!so->so_is_open_owner)
239 if (same_owner_str(so, &open->op_owner))
240 return openowner(nfs4_get_stateowner(so));
245 static struct nfs4_openowner *
246 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open,
247 struct nfs4_client *clp)
249 struct nfs4_openowner *oo;
251 spin_lock(&clp->cl_lock);
252 oo = find_openstateowner_str_locked(hashval, open, clp);
253 spin_unlock(&clp->cl_lock);
258 opaque_hashval(const void *ptr, int nbytes)
260 unsigned char *cptr = (unsigned char *) ptr;
270 static void nfsd4_free_file_rcu(struct rcu_head *rcu)
272 struct nfs4_file *fp = container_of(rcu, struct nfs4_file, fi_rcu);
274 kmem_cache_free(file_slab, fp);
278 put_nfs4_file(struct nfs4_file *fi)
280 might_lock(&state_lock);
282 if (atomic_dec_and_lock(&fi->fi_ref, &state_lock)) {
283 hlist_del_rcu(&fi->fi_hash);
284 spin_unlock(&state_lock);
285 WARN_ON_ONCE(!list_empty(&fi->fi_clnt_odstate));
286 WARN_ON_ONCE(!list_empty(&fi->fi_delegations));
287 call_rcu(&fi->fi_rcu, nfsd4_free_file_rcu);
292 __nfs4_get_fd(struct nfs4_file *f, int oflag)
294 if (f->fi_fds[oflag])
295 return get_file(f->fi_fds[oflag]);
300 find_writeable_file_locked(struct nfs4_file *f)
304 lockdep_assert_held(&f->fi_lock);
306 ret = __nfs4_get_fd(f, O_WRONLY);
308 ret = __nfs4_get_fd(f, O_RDWR);
313 find_writeable_file(struct nfs4_file *f)
317 spin_lock(&f->fi_lock);
318 ret = find_writeable_file_locked(f);
319 spin_unlock(&f->fi_lock);
324 static struct file *find_readable_file_locked(struct nfs4_file *f)
328 lockdep_assert_held(&f->fi_lock);
330 ret = __nfs4_get_fd(f, O_RDONLY);
332 ret = __nfs4_get_fd(f, O_RDWR);
337 find_readable_file(struct nfs4_file *f)
341 spin_lock(&f->fi_lock);
342 ret = find_readable_file_locked(f);
343 spin_unlock(&f->fi_lock);
349 find_any_file(struct nfs4_file *f)
353 spin_lock(&f->fi_lock);
354 ret = __nfs4_get_fd(f, O_RDWR);
356 ret = __nfs4_get_fd(f, O_WRONLY);
358 ret = __nfs4_get_fd(f, O_RDONLY);
360 spin_unlock(&f->fi_lock);
364 static atomic_long_t num_delegations;
365 unsigned long max_delegations;
368 * Open owner state (share locks)
371 /* hash tables for lock and open owners */
372 #define OWNER_HASH_BITS 8
373 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
374 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
376 static unsigned int ownerstr_hashval(struct xdr_netobj *ownername)
380 ret = opaque_hashval(ownername->data, ownername->len);
381 return ret & OWNER_HASH_MASK;
384 /* hash table for nfs4_file */
385 #define FILE_HASH_BITS 8
386 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
388 static unsigned int nfsd_fh_hashval(struct knfsd_fh *fh)
390 return jhash2(fh->fh_base.fh_pad, XDR_QUADLEN(fh->fh_size), 0);
393 static unsigned int file_hashval(struct knfsd_fh *fh)
395 return nfsd_fh_hashval(fh) & (FILE_HASH_SIZE - 1);
398 static struct hlist_head file_hashtbl[FILE_HASH_SIZE];
401 __nfs4_file_get_access(struct nfs4_file *fp, u32 access)
403 lockdep_assert_held(&fp->fi_lock);
405 if (access & NFS4_SHARE_ACCESS_WRITE)
406 atomic_inc(&fp->fi_access[O_WRONLY]);
407 if (access & NFS4_SHARE_ACCESS_READ)
408 atomic_inc(&fp->fi_access[O_RDONLY]);
412 nfs4_file_get_access(struct nfs4_file *fp, u32 access)
414 lockdep_assert_held(&fp->fi_lock);
416 /* Does this access mode make sense? */
417 if (access & ~NFS4_SHARE_ACCESS_BOTH)
420 /* Does it conflict with a deny mode already set? */
421 if ((access & fp->fi_share_deny) != 0)
422 return nfserr_share_denied;
424 __nfs4_file_get_access(fp, access);
428 static __be32 nfs4_file_check_deny(struct nfs4_file *fp, u32 deny)
430 /* Common case is that there is no deny mode. */
432 /* Does this deny mode make sense? */
433 if (deny & ~NFS4_SHARE_DENY_BOTH)
436 if ((deny & NFS4_SHARE_DENY_READ) &&
437 atomic_read(&fp->fi_access[O_RDONLY]))
438 return nfserr_share_denied;
440 if ((deny & NFS4_SHARE_DENY_WRITE) &&
441 atomic_read(&fp->fi_access[O_WRONLY]))
442 return nfserr_share_denied;
447 static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
449 might_lock(&fp->fi_lock);
451 if (atomic_dec_and_lock(&fp->fi_access[oflag], &fp->fi_lock)) {
452 struct file *f1 = NULL;
453 struct file *f2 = NULL;
455 swap(f1, fp->fi_fds[oflag]);
456 if (atomic_read(&fp->fi_access[1 - oflag]) == 0)
457 swap(f2, fp->fi_fds[O_RDWR]);
458 spin_unlock(&fp->fi_lock);
466 static void nfs4_file_put_access(struct nfs4_file *fp, u32 access)
468 WARN_ON_ONCE(access & ~NFS4_SHARE_ACCESS_BOTH);
470 if (access & NFS4_SHARE_ACCESS_WRITE)
471 __nfs4_file_put_access(fp, O_WRONLY);
472 if (access & NFS4_SHARE_ACCESS_READ)
473 __nfs4_file_put_access(fp, O_RDONLY);
477 * Allocate a new open/delegation state counter. This is needed for
478 * pNFS for proper return on close semantics.
480 * Note that we only allocate it for pNFS-enabled exports, otherwise
481 * all pointers to struct nfs4_clnt_odstate are always NULL.
483 static struct nfs4_clnt_odstate *
484 alloc_clnt_odstate(struct nfs4_client *clp)
486 struct nfs4_clnt_odstate *co;
488 co = kmem_cache_zalloc(odstate_slab, GFP_KERNEL);
491 atomic_set(&co->co_odcount, 1);
497 hash_clnt_odstate_locked(struct nfs4_clnt_odstate *co)
499 struct nfs4_file *fp = co->co_file;
501 lockdep_assert_held(&fp->fi_lock);
502 list_add(&co->co_perfile, &fp->fi_clnt_odstate);
506 get_clnt_odstate(struct nfs4_clnt_odstate *co)
509 atomic_inc(&co->co_odcount);
513 put_clnt_odstate(struct nfs4_clnt_odstate *co)
515 struct nfs4_file *fp;
521 if (atomic_dec_and_lock(&co->co_odcount, &fp->fi_lock)) {
522 list_del(&co->co_perfile);
523 spin_unlock(&fp->fi_lock);
525 nfsd4_return_all_file_layouts(co->co_client, fp);
526 kmem_cache_free(odstate_slab, co);
530 static struct nfs4_clnt_odstate *
531 find_or_hash_clnt_odstate(struct nfs4_file *fp, struct nfs4_clnt_odstate *new)
533 struct nfs4_clnt_odstate *co;
534 struct nfs4_client *cl;
541 spin_lock(&fp->fi_lock);
542 list_for_each_entry(co, &fp->fi_clnt_odstate, co_perfile) {
543 if (co->co_client == cl) {
544 get_clnt_odstate(co);
550 hash_clnt_odstate_locked(new);
552 spin_unlock(&fp->fi_lock);
556 struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl,
557 struct kmem_cache *slab)
559 struct nfs4_stid *stid;
562 stid = kmem_cache_zalloc(slab, GFP_KERNEL);
566 idr_preload(GFP_KERNEL);
567 spin_lock(&cl->cl_lock);
568 new_id = idr_alloc_cyclic(&cl->cl_stateids, stid, 0, 0, GFP_NOWAIT);
569 spin_unlock(&cl->cl_lock);
573 stid->sc_client = cl;
574 stid->sc_stateid.si_opaque.so_id = new_id;
575 stid->sc_stateid.si_opaque.so_clid = cl->cl_clientid;
576 /* Will be incremented before return to client: */
577 atomic_set(&stid->sc_count, 1);
578 spin_lock_init(&stid->sc_lock);
581 * It shouldn't be a problem to reuse an opaque stateid value.
582 * I don't think it is for 4.1. But with 4.0 I worry that, for
583 * example, a stray write retransmission could be accepted by
584 * the server when it should have been rejected. Therefore,
585 * adopt a trick from the sctp code to attempt to maximize the
586 * amount of time until an id is reused, by ensuring they always
587 * "increase" (mod INT_MAX):
591 kmem_cache_free(slab, stid);
595 static struct nfs4_ol_stateid * nfs4_alloc_open_stateid(struct nfs4_client *clp)
597 struct nfs4_stid *stid;
598 struct nfs4_ol_stateid *stp;
600 stid = nfs4_alloc_stid(clp, stateid_slab);
604 stp = openlockstateid(stid);
605 stp->st_stid.sc_free = nfs4_free_ol_stateid;
609 static void nfs4_free_deleg(struct nfs4_stid *stid)
611 kmem_cache_free(deleg_slab, stid);
612 atomic_long_dec(&num_delegations);
616 * When we recall a delegation, we should be careful not to hand it
617 * out again straight away.
618 * To ensure this we keep a pair of bloom filters ('new' and 'old')
619 * in which the filehandles of recalled delegations are "stored".
620 * If a filehandle appear in either filter, a delegation is blocked.
621 * When a delegation is recalled, the filehandle is stored in the "new"
623 * Every 30 seconds we swap the filters and clear the "new" one,
624 * unless both are empty of course.
626 * Each filter is 256 bits. We hash the filehandle to 32bit and use the
627 * low 3 bytes as hash-table indices.
629 * 'blocked_delegations_lock', which is always taken in block_delegations(),
630 * is used to manage concurrent access. Testing does not need the lock
631 * except when swapping the two filters.
633 static DEFINE_SPINLOCK(blocked_delegations_lock);
634 static struct bloom_pair {
635 int entries, old_entries;
637 int new; /* index into 'set' */
638 DECLARE_BITMAP(set[2], 256);
639 } blocked_delegations;
641 static int delegation_blocked(struct knfsd_fh *fh)
644 struct bloom_pair *bd = &blocked_delegations;
646 if (bd->entries == 0)
648 if (seconds_since_boot() - bd->swap_time > 30) {
649 spin_lock(&blocked_delegations_lock);
650 if (seconds_since_boot() - bd->swap_time > 30) {
651 bd->entries -= bd->old_entries;
652 bd->old_entries = bd->entries;
653 memset(bd->set[bd->new], 0,
656 bd->swap_time = seconds_since_boot();
658 spin_unlock(&blocked_delegations_lock);
660 hash = jhash(&fh->fh_base, fh->fh_size, 0);
661 if (test_bit(hash&255, bd->set[0]) &&
662 test_bit((hash>>8)&255, bd->set[0]) &&
663 test_bit((hash>>16)&255, bd->set[0]))
666 if (test_bit(hash&255, bd->set[1]) &&
667 test_bit((hash>>8)&255, bd->set[1]) &&
668 test_bit((hash>>16)&255, bd->set[1]))
674 static void block_delegations(struct knfsd_fh *fh)
677 struct bloom_pair *bd = &blocked_delegations;
679 hash = jhash(&fh->fh_base, fh->fh_size, 0);
681 spin_lock(&blocked_delegations_lock);
682 __set_bit(hash&255, bd->set[bd->new]);
683 __set_bit((hash>>8)&255, bd->set[bd->new]);
684 __set_bit((hash>>16)&255, bd->set[bd->new]);
685 if (bd->entries == 0)
686 bd->swap_time = seconds_since_boot();
688 spin_unlock(&blocked_delegations_lock);
691 static struct nfs4_delegation *
692 alloc_init_deleg(struct nfs4_client *clp, struct svc_fh *current_fh,
693 struct nfs4_clnt_odstate *odstate)
695 struct nfs4_delegation *dp;
698 dprintk("NFSD alloc_init_deleg\n");
699 n = atomic_long_inc_return(&num_delegations);
700 if (n < 0 || n > max_delegations)
702 if (delegation_blocked(¤t_fh->fh_handle))
704 dp = delegstateid(nfs4_alloc_stid(clp, deleg_slab));
708 dp->dl_stid.sc_free = nfs4_free_deleg;
710 * delegation seqid's are never incremented. The 4.1 special
711 * meaning of seqid 0 isn't meaningful, really, but let's avoid
712 * 0 anyway just for consistency and use 1:
714 dp->dl_stid.sc_stateid.si_generation = 1;
715 INIT_LIST_HEAD(&dp->dl_perfile);
716 INIT_LIST_HEAD(&dp->dl_perclnt);
717 INIT_LIST_HEAD(&dp->dl_recall_lru);
718 dp->dl_clnt_odstate = odstate;
719 get_clnt_odstate(odstate);
720 dp->dl_type = NFS4_OPEN_DELEGATE_READ;
722 nfsd4_init_cb(&dp->dl_recall, dp->dl_stid.sc_client,
723 &nfsd4_cb_recall_ops, NFSPROC4_CLNT_CB_RECALL);
726 atomic_long_dec(&num_delegations);
731 nfs4_put_stid(struct nfs4_stid *s)
733 struct nfs4_file *fp = s->sc_file;
734 struct nfs4_client *clp = s->sc_client;
736 might_lock(&clp->cl_lock);
738 if (!atomic_dec_and_lock(&s->sc_count, &clp->cl_lock)) {
739 wake_up_all(&close_wq);
742 idr_remove(&clp->cl_stateids, s->sc_stateid.si_opaque.so_id);
743 spin_unlock(&clp->cl_lock);
750 nfs4_inc_and_copy_stateid(stateid_t *dst, struct nfs4_stid *stid)
752 stateid_t *src = &stid->sc_stateid;
754 spin_lock(&stid->sc_lock);
755 if (unlikely(++src->si_generation == 0))
756 src->si_generation = 1;
757 memcpy(dst, src, sizeof(*dst));
758 spin_unlock(&stid->sc_lock);
761 static void nfs4_put_deleg_lease(struct nfs4_file *fp)
763 struct file *filp = NULL;
765 spin_lock(&fp->fi_lock);
766 if (fp->fi_deleg_file && --fp->fi_delegees == 0)
767 swap(filp, fp->fi_deleg_file);
768 spin_unlock(&fp->fi_lock);
771 vfs_setlease(filp, F_UNLCK, NULL, (void **)&fp);
776 void nfs4_unhash_stid(struct nfs4_stid *s)
782 hash_delegation_locked(struct nfs4_delegation *dp, struct nfs4_file *fp)
784 lockdep_assert_held(&state_lock);
785 lockdep_assert_held(&fp->fi_lock);
787 atomic_inc(&dp->dl_stid.sc_count);
788 dp->dl_stid.sc_type = NFS4_DELEG_STID;
789 list_add(&dp->dl_perfile, &fp->fi_delegations);
790 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
794 unhash_delegation_locked(struct nfs4_delegation *dp)
796 struct nfs4_file *fp = dp->dl_stid.sc_file;
798 lockdep_assert_held(&state_lock);
800 if (list_empty(&dp->dl_perfile))
803 dp->dl_stid.sc_type = NFS4_CLOSED_DELEG_STID;
804 /* Ensure that deleg break won't try to requeue it */
806 spin_lock(&fp->fi_lock);
807 list_del_init(&dp->dl_perclnt);
808 list_del_init(&dp->dl_recall_lru);
809 list_del_init(&dp->dl_perfile);
810 spin_unlock(&fp->fi_lock);
814 static void destroy_delegation(struct nfs4_delegation *dp)
818 spin_lock(&state_lock);
819 unhashed = unhash_delegation_locked(dp);
820 spin_unlock(&state_lock);
822 put_clnt_odstate(dp->dl_clnt_odstate);
823 nfs4_put_deleg_lease(dp->dl_stid.sc_file);
824 nfs4_put_stid(&dp->dl_stid);
828 static void revoke_delegation(struct nfs4_delegation *dp)
830 struct nfs4_client *clp = dp->dl_stid.sc_client;
832 WARN_ON(!list_empty(&dp->dl_recall_lru));
834 put_clnt_odstate(dp->dl_clnt_odstate);
835 nfs4_put_deleg_lease(dp->dl_stid.sc_file);
837 if (clp->cl_minorversion == 0)
838 nfs4_put_stid(&dp->dl_stid);
840 dp->dl_stid.sc_type = NFS4_REVOKED_DELEG_STID;
841 spin_lock(&clp->cl_lock);
842 list_add(&dp->dl_recall_lru, &clp->cl_revoked);
843 spin_unlock(&clp->cl_lock);
851 static unsigned int clientid_hashval(u32 id)
853 return id & CLIENT_HASH_MASK;
856 static unsigned int clientstr_hashval(const char *name)
858 return opaque_hashval(name, 8) & CLIENT_HASH_MASK;
862 * We store the NONE, READ, WRITE, and BOTH bits separately in the
863 * st_{access,deny}_bmap field of the stateid, in order to track not
864 * only what share bits are currently in force, but also what
865 * combinations of share bits previous opens have used. This allows us
866 * to enforce the recommendation of rfc 3530 14.2.19 that the server
867 * return an error if the client attempt to downgrade to a combination
868 * of share bits not explicable by closing some of its previous opens.
870 * XXX: This enforcement is actually incomplete, since we don't keep
871 * track of access/deny bit combinations; so, e.g., we allow:
873 * OPEN allow read, deny write
874 * OPEN allow both, deny none
875 * DOWNGRADE allow read, deny none
877 * which we should reject.
880 bmap_to_share_mode(unsigned long bmap) {
882 unsigned int access = 0;
884 for (i = 1; i < 4; i++) {
885 if (test_bit(i, &bmap))
891 /* set share access for a given stateid */
893 set_access(u32 access, struct nfs4_ol_stateid *stp)
895 unsigned char mask = 1 << access;
897 WARN_ON_ONCE(access > NFS4_SHARE_ACCESS_BOTH);
898 stp->st_access_bmap |= mask;
901 /* clear share access for a given stateid */
903 clear_access(u32 access, struct nfs4_ol_stateid *stp)
905 unsigned char mask = 1 << access;
907 WARN_ON_ONCE(access > NFS4_SHARE_ACCESS_BOTH);
908 stp->st_access_bmap &= ~mask;
911 /* test whether a given stateid has access */
913 test_access(u32 access, struct nfs4_ol_stateid *stp)
915 unsigned char mask = 1 << access;
917 return (bool)(stp->st_access_bmap & mask);
920 /* set share deny for a given stateid */
922 set_deny(u32 deny, struct nfs4_ol_stateid *stp)
924 unsigned char mask = 1 << deny;
926 WARN_ON_ONCE(deny > NFS4_SHARE_DENY_BOTH);
927 stp->st_deny_bmap |= mask;
930 /* clear share deny for a given stateid */
932 clear_deny(u32 deny, struct nfs4_ol_stateid *stp)
934 unsigned char mask = 1 << deny;
936 WARN_ON_ONCE(deny > NFS4_SHARE_DENY_BOTH);
937 stp->st_deny_bmap &= ~mask;
940 /* test whether a given stateid is denying specific access */
942 test_deny(u32 deny, struct nfs4_ol_stateid *stp)
944 unsigned char mask = 1 << deny;
946 return (bool)(stp->st_deny_bmap & mask);
949 static int nfs4_access_to_omode(u32 access)
951 switch (access & NFS4_SHARE_ACCESS_BOTH) {
952 case NFS4_SHARE_ACCESS_READ:
954 case NFS4_SHARE_ACCESS_WRITE:
956 case NFS4_SHARE_ACCESS_BOTH:
964 * A stateid that had a deny mode associated with it is being released
965 * or downgraded. Recalculate the deny mode on the file.
968 recalculate_deny_mode(struct nfs4_file *fp)
970 struct nfs4_ol_stateid *stp;
972 spin_lock(&fp->fi_lock);
973 fp->fi_share_deny = 0;
974 list_for_each_entry(stp, &fp->fi_stateids, st_perfile)
975 fp->fi_share_deny |= bmap_to_share_mode(stp->st_deny_bmap);
976 spin_unlock(&fp->fi_lock);
980 reset_union_bmap_deny(u32 deny, struct nfs4_ol_stateid *stp)
985 for (i = 1; i < 4; i++) {
986 if ((i & deny) != i) {
992 /* Recalculate per-file deny mode if there was a change */
994 recalculate_deny_mode(stp->st_stid.sc_file);
997 /* release all access and file references for a given stateid */
999 release_all_access(struct nfs4_ol_stateid *stp)
1002 struct nfs4_file *fp = stp->st_stid.sc_file;
1004 if (fp && stp->st_deny_bmap != 0)
1005 recalculate_deny_mode(fp);
1007 for (i = 1; i < 4; i++) {
1008 if (test_access(i, stp))
1009 nfs4_file_put_access(stp->st_stid.sc_file, i);
1010 clear_access(i, stp);
1014 static inline void nfs4_free_stateowner(struct nfs4_stateowner *sop)
1016 kfree(sop->so_owner.data);
1017 sop->so_ops->so_free(sop);
1020 static void nfs4_put_stateowner(struct nfs4_stateowner *sop)
1022 struct nfs4_client *clp = sop->so_client;
1024 might_lock(&clp->cl_lock);
1026 if (!atomic_dec_and_lock(&sop->so_count, &clp->cl_lock))
1028 sop->so_ops->so_unhash(sop);
1029 spin_unlock(&clp->cl_lock);
1030 nfs4_free_stateowner(sop);
1033 static bool unhash_ol_stateid(struct nfs4_ol_stateid *stp)
1035 struct nfs4_file *fp = stp->st_stid.sc_file;
1037 lockdep_assert_held(&stp->st_stateowner->so_client->cl_lock);
1039 if (list_empty(&stp->st_perfile))
1042 spin_lock(&fp->fi_lock);
1043 list_del_init(&stp->st_perfile);
1044 spin_unlock(&fp->fi_lock);
1045 list_del(&stp->st_perstateowner);
1049 static void nfs4_free_ol_stateid(struct nfs4_stid *stid)
1051 struct nfs4_ol_stateid *stp = openlockstateid(stid);
1053 put_clnt_odstate(stp->st_clnt_odstate);
1054 release_all_access(stp);
1055 if (stp->st_stateowner)
1056 nfs4_put_stateowner(stp->st_stateowner);
1057 kmem_cache_free(stateid_slab, stid);
1060 static void nfs4_free_lock_stateid(struct nfs4_stid *stid)
1062 struct nfs4_ol_stateid *stp = openlockstateid(stid);
1063 struct nfs4_lockowner *lo = lockowner(stp->st_stateowner);
1066 file = find_any_file(stp->st_stid.sc_file);
1068 filp_close(file, (fl_owner_t)lo);
1069 nfs4_free_ol_stateid(stid);
1073 * Put the persistent reference to an already unhashed generic stateid, while
1074 * holding the cl_lock. If it's the last reference, then put it onto the
1075 * reaplist for later destruction.
1077 static void put_ol_stateid_locked(struct nfs4_ol_stateid *stp,
1078 struct list_head *reaplist)
1080 struct nfs4_stid *s = &stp->st_stid;
1081 struct nfs4_client *clp = s->sc_client;
1083 lockdep_assert_held(&clp->cl_lock);
1085 WARN_ON_ONCE(!list_empty(&stp->st_locks));
1087 if (!atomic_dec_and_test(&s->sc_count)) {
1088 wake_up_all(&close_wq);
1092 idr_remove(&clp->cl_stateids, s->sc_stateid.si_opaque.so_id);
1093 list_add(&stp->st_locks, reaplist);
1096 static bool unhash_lock_stateid(struct nfs4_ol_stateid *stp)
1098 struct nfs4_openowner *oo = openowner(stp->st_openstp->st_stateowner);
1100 lockdep_assert_held(&oo->oo_owner.so_client->cl_lock);
1102 list_del_init(&stp->st_locks);
1103 nfs4_unhash_stid(&stp->st_stid);
1104 return unhash_ol_stateid(stp);
1107 static void release_lock_stateid(struct nfs4_ol_stateid *stp)
1109 struct nfs4_openowner *oo = openowner(stp->st_openstp->st_stateowner);
1112 spin_lock(&oo->oo_owner.so_client->cl_lock);
1113 unhashed = unhash_lock_stateid(stp);
1114 spin_unlock(&oo->oo_owner.so_client->cl_lock);
1116 nfs4_put_stid(&stp->st_stid);
1119 static void unhash_lockowner_locked(struct nfs4_lockowner *lo)
1121 struct nfs4_client *clp = lo->lo_owner.so_client;
1123 lockdep_assert_held(&clp->cl_lock);
1125 list_del_init(&lo->lo_owner.so_strhash);
1129 * Free a list of generic stateids that were collected earlier after being
1133 free_ol_stateid_reaplist(struct list_head *reaplist)
1135 struct nfs4_ol_stateid *stp;
1136 struct nfs4_file *fp;
1140 while (!list_empty(reaplist)) {
1141 stp = list_first_entry(reaplist, struct nfs4_ol_stateid,
1143 list_del(&stp->st_locks);
1144 fp = stp->st_stid.sc_file;
1145 stp->st_stid.sc_free(&stp->st_stid);
1151 static void release_lockowner(struct nfs4_lockowner *lo)
1153 struct nfs4_client *clp = lo->lo_owner.so_client;
1154 struct nfs4_ol_stateid *stp;
1155 struct list_head reaplist;
1157 INIT_LIST_HEAD(&reaplist);
1159 spin_lock(&clp->cl_lock);
1160 unhash_lockowner_locked(lo);
1161 while (!list_empty(&lo->lo_owner.so_stateids)) {
1162 stp = list_first_entry(&lo->lo_owner.so_stateids,
1163 struct nfs4_ol_stateid, st_perstateowner);
1164 WARN_ON(!unhash_lock_stateid(stp));
1165 put_ol_stateid_locked(stp, &reaplist);
1167 spin_unlock(&clp->cl_lock);
1168 free_ol_stateid_reaplist(&reaplist);
1169 nfs4_put_stateowner(&lo->lo_owner);
1172 static void release_open_stateid_locks(struct nfs4_ol_stateid *open_stp,
1173 struct list_head *reaplist)
1175 struct nfs4_ol_stateid *stp;
1177 lockdep_assert_held(&open_stp->st_stid.sc_client->cl_lock);
1179 while (!list_empty(&open_stp->st_locks)) {
1180 stp = list_entry(open_stp->st_locks.next,
1181 struct nfs4_ol_stateid, st_locks);
1182 WARN_ON(!unhash_lock_stateid(stp));
1183 put_ol_stateid_locked(stp, reaplist);
1187 static bool unhash_open_stateid(struct nfs4_ol_stateid *stp,
1188 struct list_head *reaplist)
1192 lockdep_assert_held(&stp->st_stid.sc_client->cl_lock);
1194 unhashed = unhash_ol_stateid(stp);
1195 release_open_stateid_locks(stp, reaplist);
1199 static void release_open_stateid(struct nfs4_ol_stateid *stp)
1201 LIST_HEAD(reaplist);
1203 spin_lock(&stp->st_stid.sc_client->cl_lock);
1204 if (unhash_open_stateid(stp, &reaplist))
1205 put_ol_stateid_locked(stp, &reaplist);
1206 spin_unlock(&stp->st_stid.sc_client->cl_lock);
1207 free_ol_stateid_reaplist(&reaplist);
1210 static void unhash_openowner_locked(struct nfs4_openowner *oo)
1212 struct nfs4_client *clp = oo->oo_owner.so_client;
1214 lockdep_assert_held(&clp->cl_lock);
1216 list_del_init(&oo->oo_owner.so_strhash);
1217 list_del_init(&oo->oo_perclient);
1220 static void release_last_closed_stateid(struct nfs4_openowner *oo)
1222 struct nfsd_net *nn = net_generic(oo->oo_owner.so_client->net,
1224 struct nfs4_ol_stateid *s;
1226 spin_lock(&nn->client_lock);
1227 s = oo->oo_last_closed_stid;
1229 list_del_init(&oo->oo_close_lru);
1230 oo->oo_last_closed_stid = NULL;
1232 spin_unlock(&nn->client_lock);
1234 nfs4_put_stid(&s->st_stid);
1237 static void release_openowner(struct nfs4_openowner *oo)
1239 struct nfs4_ol_stateid *stp;
1240 struct nfs4_client *clp = oo->oo_owner.so_client;
1241 struct list_head reaplist;
1243 INIT_LIST_HEAD(&reaplist);
1245 spin_lock(&clp->cl_lock);
1246 unhash_openowner_locked(oo);
1247 while (!list_empty(&oo->oo_owner.so_stateids)) {
1248 stp = list_first_entry(&oo->oo_owner.so_stateids,
1249 struct nfs4_ol_stateid, st_perstateowner);
1250 if (unhash_open_stateid(stp, &reaplist))
1251 put_ol_stateid_locked(stp, &reaplist);
1253 spin_unlock(&clp->cl_lock);
1254 free_ol_stateid_reaplist(&reaplist);
1255 release_last_closed_stateid(oo);
1256 nfs4_put_stateowner(&oo->oo_owner);
1260 hash_sessionid(struct nfs4_sessionid *sessionid)
1262 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
1264 return sid->sequence % SESSION_HASH_SIZE;
1267 #ifdef CONFIG_SUNRPC_DEBUG
1269 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
1271 u32 *ptr = (u32 *)(&sessionid->data[0]);
1272 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
1276 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
1282 * Bump the seqid on cstate->replay_owner, and clear replay_owner if it
1283 * won't be used for replay.
1285 void nfsd4_bump_seqid(struct nfsd4_compound_state *cstate, __be32 nfserr)
1287 struct nfs4_stateowner *so = cstate->replay_owner;
1289 if (nfserr == nfserr_replay_me)
1292 if (!seqid_mutating_err(ntohl(nfserr))) {
1293 nfsd4_cstate_clear_replay(cstate);
1298 if (so->so_is_open_owner)
1299 release_last_closed_stateid(openowner(so));
1305 gen_sessionid(struct nfsd4_session *ses)
1307 struct nfs4_client *clp = ses->se_client;
1308 struct nfsd4_sessionid *sid;
1310 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
1311 sid->clientid = clp->cl_clientid;
1312 sid->sequence = current_sessionid++;
1317 * The protocol defines ca_maxresponssize_cached to include the size of
1318 * the rpc header, but all we need to cache is the data starting after
1319 * the end of the initial SEQUENCE operation--the rest we regenerate
1320 * each time. Therefore we can advertise a ca_maxresponssize_cached
1321 * value that is the number of bytes in our cache plus a few additional
1322 * bytes. In order to stay on the safe side, and not promise more than
1323 * we can cache, those additional bytes must be the minimum possible: 24
1324 * bytes of rpc header (xid through accept state, with AUTH_NULL
1325 * verifier), 12 for the compound header (with zero-length tag), and 44
1326 * for the SEQUENCE op response:
1328 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
1331 free_session_slots(struct nfsd4_session *ses)
1335 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
1336 kfree(ses->se_slots[i]);
1340 * We don't actually need to cache the rpc and session headers, so we
1341 * can allocate a little less for each slot:
1343 static inline u32 slot_bytes(struct nfsd4_channel_attrs *ca)
1347 if (ca->maxresp_cached < NFSD_MIN_HDR_SEQ_SZ)
1350 size = ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
1351 return size + sizeof(struct nfsd4_slot);
1355 * XXX: If we run out of reserved DRC memory we could (up to a point)
1356 * re-negotiate active sessions and reduce their slot usage to make
1357 * room for new connections. For now we just fail the create session.
1359 static u32 nfsd4_get_drc_mem(struct nfsd4_channel_attrs *ca)
1361 u32 slotsize = slot_bytes(ca);
1362 u32 num = ca->maxreqs;
1365 spin_lock(&nfsd_drc_lock);
1366 avail = min((unsigned long)NFSD_MAX_MEM_PER_SESSION,
1367 nfsd_drc_max_mem - nfsd_drc_mem_used);
1368 num = min_t(int, num, avail / slotsize);
1369 nfsd_drc_mem_used += num * slotsize;
1370 spin_unlock(&nfsd_drc_lock);
1375 static void nfsd4_put_drc_mem(struct nfsd4_channel_attrs *ca)
1377 int slotsize = slot_bytes(ca);
1379 spin_lock(&nfsd_drc_lock);
1380 nfsd_drc_mem_used -= slotsize * ca->maxreqs;
1381 spin_unlock(&nfsd_drc_lock);
1384 static struct nfsd4_session *alloc_session(struct nfsd4_channel_attrs *fattrs,
1385 struct nfsd4_channel_attrs *battrs)
1387 int numslots = fattrs->maxreqs;
1388 int slotsize = slot_bytes(fattrs);
1389 struct nfsd4_session *new;
1392 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
1393 + sizeof(struct nfsd4_session) > PAGE_SIZE);
1394 mem = numslots * sizeof(struct nfsd4_slot *);
1396 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
1399 /* allocate each struct nfsd4_slot and data cache in one piece */
1400 for (i = 0; i < numslots; i++) {
1401 new->se_slots[i] = kzalloc(slotsize, GFP_KERNEL);
1402 if (!new->se_slots[i])
1406 memcpy(&new->se_fchannel, fattrs, sizeof(struct nfsd4_channel_attrs));
1407 memcpy(&new->se_bchannel, battrs, sizeof(struct nfsd4_channel_attrs));
1412 kfree(new->se_slots[i]);
1417 static void free_conn(struct nfsd4_conn *c)
1419 svc_xprt_put(c->cn_xprt);
1423 static void nfsd4_conn_lost(struct svc_xpt_user *u)
1425 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
1426 struct nfs4_client *clp = c->cn_session->se_client;
1428 spin_lock(&clp->cl_lock);
1429 if (!list_empty(&c->cn_persession)) {
1430 list_del(&c->cn_persession);
1433 nfsd4_probe_callback(clp);
1434 spin_unlock(&clp->cl_lock);
1437 static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
1439 struct nfsd4_conn *conn;
1441 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
1444 svc_xprt_get(rqstp->rq_xprt);
1445 conn->cn_xprt = rqstp->rq_xprt;
1446 conn->cn_flags = flags;
1447 INIT_LIST_HEAD(&conn->cn_xpt_user.list);
1451 static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
1453 conn->cn_session = ses;
1454 list_add(&conn->cn_persession, &ses->se_conns);
1457 static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
1459 struct nfs4_client *clp = ses->se_client;
1461 spin_lock(&clp->cl_lock);
1462 __nfsd4_hash_conn(conn, ses);
1463 spin_unlock(&clp->cl_lock);
1466 static int nfsd4_register_conn(struct nfsd4_conn *conn)
1468 conn->cn_xpt_user.callback = nfsd4_conn_lost;
1469 return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
1472 static void nfsd4_init_conn(struct svc_rqst *rqstp, struct nfsd4_conn *conn, struct nfsd4_session *ses)
1476 nfsd4_hash_conn(conn, ses);
1477 ret = nfsd4_register_conn(conn);
1479 /* oops; xprt is already down: */
1480 nfsd4_conn_lost(&conn->cn_xpt_user);
1481 /* We may have gained or lost a callback channel: */
1482 nfsd4_probe_callback_sync(ses->se_client);
1485 static struct nfsd4_conn *alloc_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_create_session *cses)
1487 u32 dir = NFS4_CDFC4_FORE;
1489 if (cses->flags & SESSION4_BACK_CHAN)
1490 dir |= NFS4_CDFC4_BACK;
1491 return alloc_conn(rqstp, dir);
1494 /* must be called under client_lock */
1495 static void nfsd4_del_conns(struct nfsd4_session *s)
1497 struct nfs4_client *clp = s->se_client;
1498 struct nfsd4_conn *c;
1500 spin_lock(&clp->cl_lock);
1501 while (!list_empty(&s->se_conns)) {
1502 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
1503 list_del_init(&c->cn_persession);
1504 spin_unlock(&clp->cl_lock);
1506 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
1509 spin_lock(&clp->cl_lock);
1511 spin_unlock(&clp->cl_lock);
1514 static void __free_session(struct nfsd4_session *ses)
1516 free_session_slots(ses);
1520 static void free_session(struct nfsd4_session *ses)
1522 nfsd4_del_conns(ses);
1523 nfsd4_put_drc_mem(&ses->se_fchannel);
1524 __free_session(ses);
1527 static void init_session(struct svc_rqst *rqstp, struct nfsd4_session *new, struct nfs4_client *clp, struct nfsd4_create_session *cses)
1530 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1532 new->se_client = clp;
1535 INIT_LIST_HEAD(&new->se_conns);
1537 new->se_cb_seq_nr = 1;
1538 new->se_flags = cses->flags;
1539 new->se_cb_prog = cses->callback_prog;
1540 new->se_cb_sec = cses->cb_sec;
1541 atomic_set(&new->se_ref, 0);
1542 idx = hash_sessionid(&new->se_sessionid);
1543 list_add(&new->se_hash, &nn->sessionid_hashtbl[idx]);
1544 spin_lock(&clp->cl_lock);
1545 list_add(&new->se_perclnt, &clp->cl_sessions);
1546 spin_unlock(&clp->cl_lock);
1549 struct sockaddr *sa = svc_addr(rqstp);
1551 * This is a little silly; with sessions there's no real
1552 * use for the callback address. Use the peer address
1553 * as a reasonable default for now, but consider fixing
1554 * the rpc client not to require an address in the
1557 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
1558 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1562 /* caller must hold client_lock */
1563 static struct nfsd4_session *
1564 __find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net)
1566 struct nfsd4_session *elem;
1568 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1570 lockdep_assert_held(&nn->client_lock);
1572 dump_sessionid(__func__, sessionid);
1573 idx = hash_sessionid(sessionid);
1574 /* Search in the appropriate list */
1575 list_for_each_entry(elem, &nn->sessionid_hashtbl[idx], se_hash) {
1576 if (!memcmp(elem->se_sessionid.data, sessionid->data,
1577 NFS4_MAX_SESSIONID_LEN)) {
1582 dprintk("%s: session not found\n", __func__);
1586 static struct nfsd4_session *
1587 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net,
1590 struct nfsd4_session *session;
1591 __be32 status = nfserr_badsession;
1593 session = __find_in_sessionid_hashtbl(sessionid, net);
1596 status = nfsd4_get_session_locked(session);
1604 /* caller must hold client_lock */
1606 unhash_session(struct nfsd4_session *ses)
1608 struct nfs4_client *clp = ses->se_client;
1609 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1611 lockdep_assert_held(&nn->client_lock);
1613 list_del(&ses->se_hash);
1614 spin_lock(&ses->se_client->cl_lock);
1615 list_del(&ses->se_perclnt);
1616 spin_unlock(&ses->se_client->cl_lock);
1619 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1621 STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
1624 * We're assuming the clid was not given out from a boot
1625 * precisely 2^32 (about 136 years) before this one. That seems
1626 * a safe assumption:
1628 if (clid->cl_boot == (u32)nn->boot_time)
1630 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1631 clid->cl_boot, clid->cl_id, nn->boot_time);
1636 * XXX Should we use a slab cache ?
1637 * This type of memory management is somewhat inefficient, but we use it
1638 * anyway since SETCLIENTID is not a common operation.
1640 static struct nfs4_client *alloc_client(struct xdr_netobj name)
1642 struct nfs4_client *clp;
1645 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
1648 clp->cl_name.data = kmemdup(name.data, name.len, GFP_KERNEL);
1649 if (clp->cl_name.data == NULL)
1651 clp->cl_ownerstr_hashtbl = kmalloc(sizeof(struct list_head) *
1652 OWNER_HASH_SIZE, GFP_KERNEL);
1653 if (!clp->cl_ownerstr_hashtbl)
1654 goto err_no_hashtbl;
1655 for (i = 0; i < OWNER_HASH_SIZE; i++)
1656 INIT_LIST_HEAD(&clp->cl_ownerstr_hashtbl[i]);
1657 clp->cl_name.len = name.len;
1658 INIT_LIST_HEAD(&clp->cl_sessions);
1659 idr_init(&clp->cl_stateids);
1660 atomic_set(&clp->cl_refcount, 0);
1661 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1662 INIT_LIST_HEAD(&clp->cl_idhash);
1663 INIT_LIST_HEAD(&clp->cl_openowners);
1664 INIT_LIST_HEAD(&clp->cl_delegations);
1665 INIT_LIST_HEAD(&clp->cl_lru);
1666 INIT_LIST_HEAD(&clp->cl_revoked);
1667 #ifdef CONFIG_NFSD_PNFS
1668 INIT_LIST_HEAD(&clp->cl_lo_states);
1670 spin_lock_init(&clp->cl_lock);
1671 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1674 kfree(clp->cl_name.data);
1681 free_client(struct nfs4_client *clp)
1683 while (!list_empty(&clp->cl_sessions)) {
1684 struct nfsd4_session *ses;
1685 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
1687 list_del(&ses->se_perclnt);
1688 WARN_ON_ONCE(atomic_read(&ses->se_ref));
1691 rpc_destroy_wait_queue(&clp->cl_cb_waitq);
1692 free_svc_cred(&clp->cl_cred);
1693 kfree(clp->cl_ownerstr_hashtbl);
1694 kfree(clp->cl_name.data);
1695 idr_destroy(&clp->cl_stateids);
1699 /* must be called under the client_lock */
1701 unhash_client_locked(struct nfs4_client *clp)
1703 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1704 struct nfsd4_session *ses;
1706 lockdep_assert_held(&nn->client_lock);
1708 /* Mark the client as expired! */
1710 /* Make it invisible */
1711 if (!list_empty(&clp->cl_idhash)) {
1712 list_del_init(&clp->cl_idhash);
1713 if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
1714 rb_erase(&clp->cl_namenode, &nn->conf_name_tree);
1716 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1718 list_del_init(&clp->cl_lru);
1719 spin_lock(&clp->cl_lock);
1720 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1721 list_del_init(&ses->se_hash);
1722 spin_unlock(&clp->cl_lock);
1726 unhash_client(struct nfs4_client *clp)
1728 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1730 spin_lock(&nn->client_lock);
1731 unhash_client_locked(clp);
1732 spin_unlock(&nn->client_lock);
1735 static __be32 mark_client_expired_locked(struct nfs4_client *clp)
1737 if (atomic_read(&clp->cl_refcount))
1738 return nfserr_jukebox;
1739 unhash_client_locked(clp);
1744 __destroy_client(struct nfs4_client *clp)
1746 struct nfs4_openowner *oo;
1747 struct nfs4_delegation *dp;
1748 struct list_head reaplist;
1750 INIT_LIST_HEAD(&reaplist);
1751 spin_lock(&state_lock);
1752 while (!list_empty(&clp->cl_delegations)) {
1753 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1754 WARN_ON(!unhash_delegation_locked(dp));
1755 list_add(&dp->dl_recall_lru, &reaplist);
1757 spin_unlock(&state_lock);
1758 while (!list_empty(&reaplist)) {
1759 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1760 list_del_init(&dp->dl_recall_lru);
1761 put_clnt_odstate(dp->dl_clnt_odstate);
1762 nfs4_put_deleg_lease(dp->dl_stid.sc_file);
1763 nfs4_put_stid(&dp->dl_stid);
1765 while (!list_empty(&clp->cl_revoked)) {
1766 dp = list_entry(clp->cl_revoked.next, struct nfs4_delegation, dl_recall_lru);
1767 list_del_init(&dp->dl_recall_lru);
1768 nfs4_put_stid(&dp->dl_stid);
1770 while (!list_empty(&clp->cl_openowners)) {
1771 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1772 nfs4_get_stateowner(&oo->oo_owner);
1773 release_openowner(oo);
1775 nfsd4_return_all_client_layouts(clp);
1776 nfsd4_shutdown_callback(clp);
1777 if (clp->cl_cb_conn.cb_xprt)
1778 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1783 destroy_client(struct nfs4_client *clp)
1786 __destroy_client(clp);
1789 static void expire_client(struct nfs4_client *clp)
1792 nfsd4_client_record_remove(clp);
1793 __destroy_client(clp);
1796 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1798 memcpy(target->cl_verifier.data, source->data,
1799 sizeof(target->cl_verifier.data));
1802 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1804 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1805 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1808 static int copy_cred(struct svc_cred *target, struct svc_cred *source)
1810 if (source->cr_principal) {
1811 target->cr_principal =
1812 kstrdup(source->cr_principal, GFP_KERNEL);
1813 if (target->cr_principal == NULL)
1816 target->cr_principal = NULL;
1817 target->cr_flavor = source->cr_flavor;
1818 target->cr_uid = source->cr_uid;
1819 target->cr_gid = source->cr_gid;
1820 target->cr_group_info = source->cr_group_info;
1821 get_group_info(target->cr_group_info);
1822 target->cr_gss_mech = source->cr_gss_mech;
1823 if (source->cr_gss_mech)
1824 gss_mech_get(source->cr_gss_mech);
1829 compare_blob(const struct xdr_netobj *o1, const struct xdr_netobj *o2)
1831 if (o1->len < o2->len)
1833 if (o1->len > o2->len)
1835 return memcmp(o1->data, o2->data, o1->len);
1838 static int same_name(const char *n1, const char *n2)
1840 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1844 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1846 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1850 same_clid(clientid_t *cl1, clientid_t *cl2)
1852 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1855 static bool groups_equal(struct group_info *g1, struct group_info *g2)
1859 if (g1->ngroups != g2->ngroups)
1861 for (i=0; i<g1->ngroups; i++)
1862 if (!gid_eq(GROUP_AT(g1, i), GROUP_AT(g2, i)))
1868 * RFC 3530 language requires clid_inuse be returned when the
1869 * "principal" associated with a requests differs from that previously
1870 * used. We use uid, gid's, and gss principal string as our best
1871 * approximation. We also don't want to allow non-gss use of a client
1872 * established using gss: in theory cr_principal should catch that
1873 * change, but in practice cr_principal can be null even in the gss case
1874 * since gssd doesn't always pass down a principal string.
1876 static bool is_gss_cred(struct svc_cred *cr)
1878 /* Is cr_flavor one of the gss "pseudoflavors"?: */
1879 return (cr->cr_flavor > RPC_AUTH_MAXFLAVOR);
1884 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1886 if ((is_gss_cred(cr1) != is_gss_cred(cr2))
1887 || (!uid_eq(cr1->cr_uid, cr2->cr_uid))
1888 || (!gid_eq(cr1->cr_gid, cr2->cr_gid))
1889 || !groups_equal(cr1->cr_group_info, cr2->cr_group_info))
1891 if (cr1->cr_principal == cr2->cr_principal)
1893 if (!cr1->cr_principal || !cr2->cr_principal)
1895 return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
1898 static bool svc_rqst_integrity_protected(struct svc_rqst *rqstp)
1900 struct svc_cred *cr = &rqstp->rq_cred;
1903 if (!cr->cr_gss_mech)
1905 service = gss_pseudoflavor_to_service(cr->cr_gss_mech, cr->cr_flavor);
1906 return service == RPC_GSS_SVC_INTEGRITY ||
1907 service == RPC_GSS_SVC_PRIVACY;
1910 static bool mach_creds_match(struct nfs4_client *cl, struct svc_rqst *rqstp)
1912 struct svc_cred *cr = &rqstp->rq_cred;
1914 if (!cl->cl_mach_cred)
1916 if (cl->cl_cred.cr_gss_mech != cr->cr_gss_mech)
1918 if (!svc_rqst_integrity_protected(rqstp))
1920 if (!cr->cr_principal)
1922 return 0 == strcmp(cl->cl_cred.cr_principal, cr->cr_principal);
1925 static void gen_confirm(struct nfs4_client *clp, struct nfsd_net *nn)
1930 * This is opaque to client, so no need to byte-swap. Use
1931 * __force to keep sparse happy
1933 verf[0] = (__force __be32)get_seconds();
1934 verf[1] = (__force __be32)nn->clverifier_counter++;
1935 memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
1938 static void gen_clid(struct nfs4_client *clp, struct nfsd_net *nn)
1940 clp->cl_clientid.cl_boot = nn->boot_time;
1941 clp->cl_clientid.cl_id = nn->clientid_counter++;
1942 gen_confirm(clp, nn);
1945 static struct nfs4_stid *
1946 find_stateid_locked(struct nfs4_client *cl, stateid_t *t)
1948 struct nfs4_stid *ret;
1950 ret = idr_find(&cl->cl_stateids, t->si_opaque.so_id);
1951 if (!ret || !ret->sc_type)
1956 static struct nfs4_stid *
1957 find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
1959 struct nfs4_stid *s;
1961 spin_lock(&cl->cl_lock);
1962 s = find_stateid_locked(cl, t);
1964 if (typemask & s->sc_type)
1965 atomic_inc(&s->sc_count);
1969 spin_unlock(&cl->cl_lock);
1973 static struct nfs4_client *create_client(struct xdr_netobj name,
1974 struct svc_rqst *rqstp, nfs4_verifier *verf)
1976 struct nfs4_client *clp;
1977 struct sockaddr *sa = svc_addr(rqstp);
1979 struct net *net = SVC_NET(rqstp);
1981 clp = alloc_client(name);
1985 ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1990 nfsd4_init_cb(&clp->cl_cb_null, clp, NULL, NFSPROC4_CLNT_CB_NULL);
1991 clp->cl_time = get_seconds();
1992 clear_bit(0, &clp->cl_cb_slot_busy);
1993 copy_verf(clp, verf);
1994 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1995 clp->cl_cb_session = NULL;
2001 add_clp_to_name_tree(struct nfs4_client *new_clp, struct rb_root *root)
2003 struct rb_node **new = &(root->rb_node), *parent = NULL;
2004 struct nfs4_client *clp;
2007 clp = rb_entry(*new, struct nfs4_client, cl_namenode);
2010 if (compare_blob(&clp->cl_name, &new_clp->cl_name) > 0)
2011 new = &((*new)->rb_left);
2013 new = &((*new)->rb_right);
2016 rb_link_node(&new_clp->cl_namenode, parent, new);
2017 rb_insert_color(&new_clp->cl_namenode, root);
2020 static struct nfs4_client *
2021 find_clp_in_name_tree(struct xdr_netobj *name, struct rb_root *root)
2024 struct rb_node *node = root->rb_node;
2025 struct nfs4_client *clp;
2028 clp = rb_entry(node, struct nfs4_client, cl_namenode);
2029 cmp = compare_blob(&clp->cl_name, name);
2031 node = node->rb_left;
2033 node = node->rb_right;
2041 add_to_unconfirmed(struct nfs4_client *clp)
2043 unsigned int idhashval;
2044 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2046 lockdep_assert_held(&nn->client_lock);
2048 clear_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
2049 add_clp_to_name_tree(clp, &nn->unconf_name_tree);
2050 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
2051 list_add(&clp->cl_idhash, &nn->unconf_id_hashtbl[idhashval]);
2052 renew_client_locked(clp);
2056 move_to_confirmed(struct nfs4_client *clp)
2058 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
2059 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2061 lockdep_assert_held(&nn->client_lock);
2063 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
2064 list_move(&clp->cl_idhash, &nn->conf_id_hashtbl[idhashval]);
2065 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
2066 add_clp_to_name_tree(clp, &nn->conf_name_tree);
2067 set_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
2068 renew_client_locked(clp);
2071 static struct nfs4_client *
2072 find_client_in_id_table(struct list_head *tbl, clientid_t *clid, bool sessions)
2074 struct nfs4_client *clp;
2075 unsigned int idhashval = clientid_hashval(clid->cl_id);
2077 list_for_each_entry(clp, &tbl[idhashval], cl_idhash) {
2078 if (same_clid(&clp->cl_clientid, clid)) {
2079 if ((bool)clp->cl_minorversion != sessions)
2081 renew_client_locked(clp);
2088 static struct nfs4_client *
2089 find_confirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
2091 struct list_head *tbl = nn->conf_id_hashtbl;
2093 lockdep_assert_held(&nn->client_lock);
2094 return find_client_in_id_table(tbl, clid, sessions);
2097 static struct nfs4_client *
2098 find_unconfirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
2100 struct list_head *tbl = nn->unconf_id_hashtbl;
2102 lockdep_assert_held(&nn->client_lock);
2103 return find_client_in_id_table(tbl, clid, sessions);
2106 static bool clp_used_exchangeid(struct nfs4_client *clp)
2108 return clp->cl_exchange_flags != 0;
2111 static struct nfs4_client *
2112 find_confirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
2114 lockdep_assert_held(&nn->client_lock);
2115 return find_clp_in_name_tree(name, &nn->conf_name_tree);
2118 static struct nfs4_client *
2119 find_unconfirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
2121 lockdep_assert_held(&nn->client_lock);
2122 return find_clp_in_name_tree(name, &nn->unconf_name_tree);
2126 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
2128 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
2129 struct sockaddr *sa = svc_addr(rqstp);
2130 u32 scopeid = rpc_get_scope_id(sa);
2131 unsigned short expected_family;
2133 /* Currently, we only support tcp and tcp6 for the callback channel */
2134 if (se->se_callback_netid_len == 3 &&
2135 !memcmp(se->se_callback_netid_val, "tcp", 3))
2136 expected_family = AF_INET;
2137 else if (se->se_callback_netid_len == 4 &&
2138 !memcmp(se->se_callback_netid_val, "tcp6", 4))
2139 expected_family = AF_INET6;
2143 conn->cb_addrlen = rpc_uaddr2sockaddr(clp->net, se->se_callback_addr_val,
2144 se->se_callback_addr_len,
2145 (struct sockaddr *)&conn->cb_addr,
2146 sizeof(conn->cb_addr));
2148 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
2151 if (conn->cb_addr.ss_family == AF_INET6)
2152 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
2154 conn->cb_prog = se->se_callback_prog;
2155 conn->cb_ident = se->se_callback_ident;
2156 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
2159 conn->cb_addr.ss_family = AF_UNSPEC;
2160 conn->cb_addrlen = 0;
2161 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
2162 "will not receive delegations\n",
2163 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
2169 * Cache a reply. nfsd4_check_resp_size() has bounded the cache size.
2172 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
2174 struct xdr_buf *buf = resp->xdr.buf;
2175 struct nfsd4_slot *slot = resp->cstate.slot;
2178 dprintk("--> %s slot %p\n", __func__, slot);
2180 slot->sl_opcnt = resp->opcnt;
2181 slot->sl_status = resp->cstate.status;
2183 slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
2184 if (nfsd4_not_cached(resp)) {
2185 slot->sl_datalen = 0;
2188 base = resp->cstate.data_offset;
2189 slot->sl_datalen = buf->len - base;
2190 if (read_bytes_from_xdr_buf(buf, base, slot->sl_data, slot->sl_datalen))
2191 WARN("%s: sessions DRC could not cache compound\n", __func__);
2196 * Encode the replay sequence operation from the slot values.
2197 * If cachethis is FALSE encode the uncached rep error on the next
2198 * operation which sets resp->p and increments resp->opcnt for
2199 * nfs4svc_encode_compoundres.
2203 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
2204 struct nfsd4_compoundres *resp)
2206 struct nfsd4_op *op;
2207 struct nfsd4_slot *slot = resp->cstate.slot;
2209 /* Encode the replayed sequence operation */
2210 op = &args->ops[resp->opcnt - 1];
2211 nfsd4_encode_operation(resp, op);
2213 /* Return nfserr_retry_uncached_rep in next operation. */
2214 if (args->opcnt > 1 && !(slot->sl_flags & NFSD4_SLOT_CACHETHIS)) {
2215 op = &args->ops[resp->opcnt++];
2216 op->status = nfserr_retry_uncached_rep;
2217 nfsd4_encode_operation(resp, op);
2223 * The sequence operation is not cached because we can use the slot and
2227 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
2228 struct nfsd4_sequence *seq)
2230 struct nfsd4_slot *slot = resp->cstate.slot;
2231 struct xdr_stream *xdr = &resp->xdr;
2235 dprintk("--> %s slot %p\n", __func__, slot);
2237 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
2241 p = xdr_reserve_space(xdr, slot->sl_datalen);
2244 return nfserr_serverfault;
2246 xdr_encode_opaque_fixed(p, slot->sl_data, slot->sl_datalen);
2247 xdr_commit_encode(xdr);
2249 resp->opcnt = slot->sl_opcnt;
2250 return slot->sl_status;
2254 * Set the exchange_id flags returned by the server.
2257 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
2259 #ifdef CONFIG_NFSD_PNFS
2260 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_PNFS_MDS;
2262 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
2265 /* Referrals are supported, Migration is not. */
2266 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
2268 /* set the wire flags to return to client. */
2269 clid->flags = new->cl_exchange_flags;
2272 static bool client_has_openowners(struct nfs4_client *clp)
2274 struct nfs4_openowner *oo;
2276 list_for_each_entry(oo, &clp->cl_openowners, oo_perclient) {
2277 if (!list_empty(&oo->oo_owner.so_stateids))
2283 static bool client_has_state(struct nfs4_client *clp)
2285 return client_has_openowners(clp)
2286 #ifdef CONFIG_NFSD_PNFS
2287 || !list_empty(&clp->cl_lo_states)
2289 || !list_empty(&clp->cl_delegations)
2290 || !list_empty(&clp->cl_sessions);
2294 nfsd4_exchange_id(struct svc_rqst *rqstp,
2295 struct nfsd4_compound_state *cstate,
2296 struct nfsd4_exchange_id *exid)
2298 struct nfs4_client *conf, *new;
2299 struct nfs4_client *unconf = NULL;
2301 char addr_str[INET6_ADDRSTRLEN];
2302 nfs4_verifier verf = exid->verifier;
2303 struct sockaddr *sa = svc_addr(rqstp);
2304 bool update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
2305 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2307 rpc_ntop(sa, addr_str, sizeof(addr_str));
2308 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
2309 "ip_addr=%s flags %x, spa_how %d\n",
2310 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
2311 addr_str, exid->flags, exid->spa_how);
2313 if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
2314 return nfserr_inval;
2316 switch (exid->spa_how) {
2318 if (!svc_rqst_integrity_protected(rqstp))
2319 return nfserr_inval;
2322 default: /* checked by xdr code */
2325 return nfserr_encr_alg_unsupp;
2328 new = create_client(exid->clname, rqstp, &verf);
2330 return nfserr_jukebox;
2332 /* Cases below refer to rfc 5661 section 18.35.4: */
2333 spin_lock(&nn->client_lock);
2334 conf = find_confirmed_client_by_name(&exid->clname, nn);
2336 bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
2337 bool verfs_match = same_verf(&verf, &conf->cl_verifier);
2340 if (!clp_used_exchangeid(conf)) { /* buggy client */
2341 status = nfserr_inval;
2344 if (!mach_creds_match(conf, rqstp)) {
2345 status = nfserr_wrong_cred;
2348 if (!creds_match) { /* case 9 */
2349 status = nfserr_perm;
2352 if (!verfs_match) { /* case 8 */
2353 status = nfserr_not_same;
2357 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
2360 if (!creds_match) { /* case 3 */
2361 if (client_has_state(conf)) {
2362 status = nfserr_clid_inuse;
2367 if (verfs_match) { /* case 2 */
2368 conf->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
2371 /* case 5, client reboot */
2376 if (update) { /* case 7 */
2377 status = nfserr_noent;
2381 unconf = find_unconfirmed_client_by_name(&exid->clname, nn);
2382 if (unconf) /* case 4, possible retry or client restart */
2383 unhash_client_locked(unconf);
2385 /* case 1 (normal case) */
2388 status = mark_client_expired_locked(conf);
2392 new->cl_minorversion = cstate->minorversion;
2393 new->cl_mach_cred = (exid->spa_how == SP4_MACH_CRED);
2396 add_to_unconfirmed(new);
2399 exid->clientid.cl_boot = conf->cl_clientid.cl_boot;
2400 exid->clientid.cl_id = conf->cl_clientid.cl_id;
2402 exid->seqid = conf->cl_cs_slot.sl_seqid + 1;
2403 nfsd4_set_ex_flags(conf, exid);
2405 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
2406 conf->cl_cs_slot.sl_seqid, conf->cl_exchange_flags);
2410 spin_unlock(&nn->client_lock);
2414 expire_client(unconf);
2419 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
2421 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
2424 /* The slot is in use, and no response has been sent. */
2426 if (seqid == slot_seqid)
2427 return nfserr_jukebox;
2429 return nfserr_seq_misordered;
2431 /* Note unsigned 32-bit arithmetic handles wraparound: */
2432 if (likely(seqid == slot_seqid + 1))
2434 if (seqid == slot_seqid)
2435 return nfserr_replay_cache;
2436 return nfserr_seq_misordered;
2440 * Cache the create session result into the create session single DRC
2441 * slot cache by saving the xdr structure. sl_seqid has been set.
2442 * Do this for solo or embedded create session operations.
2445 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
2446 struct nfsd4_clid_slot *slot, __be32 nfserr)
2448 slot->sl_status = nfserr;
2449 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
2453 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
2454 struct nfsd4_clid_slot *slot)
2456 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
2457 return slot->sl_status;
2460 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
2461 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
2462 1 + /* MIN tag is length with zero, only length */ \
2463 3 + /* version, opcount, opcode */ \
2464 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
2465 /* seqid, slotID, slotID, cache */ \
2466 4 ) * sizeof(__be32))
2468 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
2469 2 + /* verifier: AUTH_NULL, length 0 */\
2471 1 + /* MIN tag is length with zero, only length */ \
2472 3 + /* opcount, opcode, opstatus*/ \
2473 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
2474 /* seqid, slotID, slotID, slotID, status */ \
2475 5 ) * sizeof(__be32))
2477 static __be32 check_forechannel_attrs(struct nfsd4_channel_attrs *ca, struct nfsd_net *nn)
2479 u32 maxrpc = nn->nfsd_serv->sv_max_mesg;
2481 if (ca->maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ)
2482 return nfserr_toosmall;
2483 if (ca->maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ)
2484 return nfserr_toosmall;
2485 ca->headerpadsz = 0;
2486 ca->maxreq_sz = min_t(u32, ca->maxreq_sz, maxrpc);
2487 ca->maxresp_sz = min_t(u32, ca->maxresp_sz, maxrpc);
2488 ca->maxops = min_t(u32, ca->maxops, NFSD_MAX_OPS_PER_COMPOUND);
2489 ca->maxresp_cached = min_t(u32, ca->maxresp_cached,
2490 NFSD_SLOT_CACHE_SIZE + NFSD_MIN_HDR_SEQ_SZ);
2491 ca->maxreqs = min_t(u32, ca->maxreqs, NFSD_MAX_SLOTS_PER_SESSION);
2493 * Note decreasing slot size below client's request may make it
2494 * difficult for client to function correctly, whereas
2495 * decreasing the number of slots will (just?) affect
2496 * performance. When short on memory we therefore prefer to
2497 * decrease number of slots instead of their size. Clients that
2498 * request larger slots than they need will get poor results:
2500 ca->maxreqs = nfsd4_get_drc_mem(ca);
2502 return nfserr_jukebox;
2507 #define NFSD_CB_MAX_REQ_SZ ((NFS4_enc_cb_recall_sz + \
2508 RPC_MAX_HEADER_WITH_AUTH) * sizeof(__be32))
2509 #define NFSD_CB_MAX_RESP_SZ ((NFS4_dec_cb_recall_sz + \
2510 RPC_MAX_REPHEADER_WITH_AUTH) * sizeof(__be32))
2512 static __be32 check_backchannel_attrs(struct nfsd4_channel_attrs *ca)
2514 ca->headerpadsz = 0;
2517 * These RPC_MAX_HEADER macros are overkill, especially since we
2518 * don't even do gss on the backchannel yet. But this is still
2519 * less than 1k. Tighten up this estimate in the unlikely event
2520 * it turns out to be a problem for some client:
2522 if (ca->maxreq_sz < NFSD_CB_MAX_REQ_SZ)
2523 return nfserr_toosmall;
2524 if (ca->maxresp_sz < NFSD_CB_MAX_RESP_SZ)
2525 return nfserr_toosmall;
2526 ca->maxresp_cached = 0;
2528 return nfserr_toosmall;
2533 static __be32 nfsd4_check_cb_sec(struct nfsd4_cb_sec *cbs)
2535 switch (cbs->flavor) {
2541 * GSS case: the spec doesn't allow us to return this
2542 * error. But it also doesn't allow us not to support
2544 * I'd rather this fail hard than return some error the
2545 * client might think it can already handle:
2547 return nfserr_encr_alg_unsupp;
2552 nfsd4_create_session(struct svc_rqst *rqstp,
2553 struct nfsd4_compound_state *cstate,
2554 struct nfsd4_create_session *cr_ses)
2556 struct sockaddr *sa = svc_addr(rqstp);
2557 struct nfs4_client *conf, *unconf;
2558 struct nfs4_client *old = NULL;
2559 struct nfsd4_session *new;
2560 struct nfsd4_conn *conn;
2561 struct nfsd4_clid_slot *cs_slot = NULL;
2563 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2565 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
2566 return nfserr_inval;
2567 status = nfsd4_check_cb_sec(&cr_ses->cb_sec);
2570 status = check_forechannel_attrs(&cr_ses->fore_channel, nn);
2573 status = check_backchannel_attrs(&cr_ses->back_channel);
2575 goto out_release_drc_mem;
2576 status = nfserr_jukebox;
2577 new = alloc_session(&cr_ses->fore_channel, &cr_ses->back_channel);
2579 goto out_release_drc_mem;
2580 conn = alloc_conn_from_crses(rqstp, cr_ses);
2582 goto out_free_session;
2584 spin_lock(&nn->client_lock);
2585 unconf = find_unconfirmed_client(&cr_ses->clientid, true, nn);
2586 conf = find_confirmed_client(&cr_ses->clientid, true, nn);
2587 WARN_ON_ONCE(conf && unconf);
2590 status = nfserr_wrong_cred;
2591 if (!mach_creds_match(conf, rqstp))
2593 cs_slot = &conf->cl_cs_slot;
2594 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
2596 if (status == nfserr_replay_cache)
2597 status = nfsd4_replay_create_session(cr_ses, cs_slot);
2600 } else if (unconf) {
2601 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
2602 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
2603 status = nfserr_clid_inuse;
2606 status = nfserr_wrong_cred;
2607 if (!mach_creds_match(unconf, rqstp))
2609 cs_slot = &unconf->cl_cs_slot;
2610 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
2612 /* an unconfirmed replay returns misordered */
2613 status = nfserr_seq_misordered;
2616 old = find_confirmed_client_by_name(&unconf->cl_name, nn);
2618 status = mark_client_expired_locked(old);
2624 move_to_confirmed(unconf);
2627 status = nfserr_stale_clientid;
2632 * We do not support RDMA or persistent sessions
2634 cr_ses->flags &= ~SESSION4_PERSIST;
2635 cr_ses->flags &= ~SESSION4_RDMA;
2637 init_session(rqstp, new, conf, cr_ses);
2638 nfsd4_get_session_locked(new);
2640 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
2641 NFS4_MAX_SESSIONID_LEN);
2642 cs_slot->sl_seqid++;
2643 cr_ses->seqid = cs_slot->sl_seqid;
2645 /* cache solo and embedded create sessions under the client_lock */
2646 nfsd4_cache_create_session(cr_ses, cs_slot, status);
2647 spin_unlock(&nn->client_lock);
2648 /* init connection and backchannel */
2649 nfsd4_init_conn(rqstp, conn, new);
2650 nfsd4_put_session(new);
2655 spin_unlock(&nn->client_lock);
2660 __free_session(new);
2661 out_release_drc_mem:
2662 nfsd4_put_drc_mem(&cr_ses->fore_channel);
2666 static __be32 nfsd4_map_bcts_dir(u32 *dir)
2669 case NFS4_CDFC4_FORE:
2670 case NFS4_CDFC4_BACK:
2672 case NFS4_CDFC4_FORE_OR_BOTH:
2673 case NFS4_CDFC4_BACK_OR_BOTH:
2674 *dir = NFS4_CDFC4_BOTH;
2677 return nfserr_inval;
2680 __be32 nfsd4_backchannel_ctl(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_backchannel_ctl *bc)
2682 struct nfsd4_session *session = cstate->session;
2683 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2686 status = nfsd4_check_cb_sec(&bc->bc_cb_sec);
2689 spin_lock(&nn->client_lock);
2690 session->se_cb_prog = bc->bc_cb_program;
2691 session->se_cb_sec = bc->bc_cb_sec;
2692 spin_unlock(&nn->client_lock);
2694 nfsd4_probe_callback(session->se_client);
2699 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
2700 struct nfsd4_compound_state *cstate,
2701 struct nfsd4_bind_conn_to_session *bcts)
2704 struct nfsd4_conn *conn;
2705 struct nfsd4_session *session;
2706 struct net *net = SVC_NET(rqstp);
2707 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2709 if (!nfsd4_last_compound_op(rqstp))
2710 return nfserr_not_only_op;
2711 spin_lock(&nn->client_lock);
2712 session = find_in_sessionid_hashtbl(&bcts->sessionid, net, &status);
2713 spin_unlock(&nn->client_lock);
2715 goto out_no_session;
2716 status = nfserr_wrong_cred;
2717 if (!mach_creds_match(session->se_client, rqstp))
2719 status = nfsd4_map_bcts_dir(&bcts->dir);
2722 conn = alloc_conn(rqstp, bcts->dir);
2723 status = nfserr_jukebox;
2726 nfsd4_init_conn(rqstp, conn, session);
2729 nfsd4_put_session(session);
2734 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
2738 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
2742 nfsd4_destroy_session(struct svc_rqst *r,
2743 struct nfsd4_compound_state *cstate,
2744 struct nfsd4_destroy_session *sessionid)
2746 struct nfsd4_session *ses;
2748 int ref_held_by_me = 0;
2749 struct net *net = SVC_NET(r);
2750 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2752 status = nfserr_not_only_op;
2753 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
2754 if (!nfsd4_last_compound_op(r))
2758 dump_sessionid(__func__, &sessionid->sessionid);
2759 spin_lock(&nn->client_lock);
2760 ses = find_in_sessionid_hashtbl(&sessionid->sessionid, net, &status);
2762 goto out_client_lock;
2763 status = nfserr_wrong_cred;
2764 if (!mach_creds_match(ses->se_client, r))
2765 goto out_put_session;
2766 status = mark_session_dead_locked(ses, 1 + ref_held_by_me);
2768 goto out_put_session;
2769 unhash_session(ses);
2770 spin_unlock(&nn->client_lock);
2772 nfsd4_probe_callback_sync(ses->se_client);
2774 spin_lock(&nn->client_lock);
2777 nfsd4_put_session_locked(ses);
2779 spin_unlock(&nn->client_lock);
2784 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
2786 struct nfsd4_conn *c;
2788 list_for_each_entry(c, &s->se_conns, cn_persession) {
2789 if (c->cn_xprt == xpt) {
2796 static __be32 nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
2798 struct nfs4_client *clp = ses->se_client;
2799 struct nfsd4_conn *c;
2800 __be32 status = nfs_ok;
2803 spin_lock(&clp->cl_lock);
2804 c = __nfsd4_find_conn(new->cn_xprt, ses);
2807 status = nfserr_conn_not_bound_to_session;
2808 if (clp->cl_mach_cred)
2810 __nfsd4_hash_conn(new, ses);
2811 spin_unlock(&clp->cl_lock);
2812 ret = nfsd4_register_conn(new);
2814 /* oops; xprt is already down: */
2815 nfsd4_conn_lost(&new->cn_xpt_user);
2818 spin_unlock(&clp->cl_lock);
2823 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
2825 struct nfsd4_compoundargs *args = rqstp->rq_argp;
2827 return args->opcnt > session->se_fchannel.maxops;
2830 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
2831 struct nfsd4_session *session)
2833 struct xdr_buf *xb = &rqstp->rq_arg;
2835 return xb->len > session->se_fchannel.maxreq_sz;
2839 nfsd4_sequence(struct svc_rqst *rqstp,
2840 struct nfsd4_compound_state *cstate,
2841 struct nfsd4_sequence *seq)
2843 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2844 struct xdr_stream *xdr = &resp->xdr;
2845 struct nfsd4_session *session;
2846 struct nfs4_client *clp;
2847 struct nfsd4_slot *slot;
2848 struct nfsd4_conn *conn;
2851 struct net *net = SVC_NET(rqstp);
2852 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2854 if (resp->opcnt != 1)
2855 return nfserr_sequence_pos;
2858 * Will be either used or freed by nfsd4_sequence_check_conn
2861 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
2863 return nfserr_jukebox;
2865 spin_lock(&nn->client_lock);
2866 session = find_in_sessionid_hashtbl(&seq->sessionid, net, &status);
2868 goto out_no_session;
2869 clp = session->se_client;
2871 status = nfserr_too_many_ops;
2872 if (nfsd4_session_too_many_ops(rqstp, session))
2873 goto out_put_session;
2875 status = nfserr_req_too_big;
2876 if (nfsd4_request_too_big(rqstp, session))
2877 goto out_put_session;
2879 status = nfserr_badslot;
2880 if (seq->slotid >= session->se_fchannel.maxreqs)
2881 goto out_put_session;
2883 slot = session->se_slots[seq->slotid];
2884 dprintk("%s: slotid %d\n", __func__, seq->slotid);
2886 /* We do not negotiate the number of slots yet, so set the
2887 * maxslots to the session maxreqs which is used to encode
2888 * sr_highest_slotid and the sr_target_slot id to maxslots */
2889 seq->maxslots = session->se_fchannel.maxreqs;
2891 status = check_slot_seqid(seq->seqid, slot->sl_seqid,
2892 slot->sl_flags & NFSD4_SLOT_INUSE);
2893 if (status == nfserr_replay_cache) {
2894 status = nfserr_seq_misordered;
2895 if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
2896 goto out_put_session;
2897 cstate->slot = slot;
2898 cstate->session = session;
2900 /* Return the cached reply status and set cstate->status
2901 * for nfsd4_proc_compound processing */
2902 status = nfsd4_replay_cache_entry(resp, seq);
2903 cstate->status = nfserr_replay_cache;
2907 goto out_put_session;
2909 status = nfsd4_sequence_check_conn(conn, session);
2912 goto out_put_session;
2914 buflen = (seq->cachethis) ?
2915 session->se_fchannel.maxresp_cached :
2916 session->se_fchannel.maxresp_sz;
2917 status = (seq->cachethis) ? nfserr_rep_too_big_to_cache :
2919 if (xdr_restrict_buflen(xdr, buflen - rqstp->rq_auth_slack))
2920 goto out_put_session;
2921 svc_reserve(rqstp, buflen);
2924 /* Success! bump slot seqid */
2925 slot->sl_seqid = seq->seqid;
2926 slot->sl_flags |= NFSD4_SLOT_INUSE;
2928 slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
2930 slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
2932 cstate->slot = slot;
2933 cstate->session = session;
2937 switch (clp->cl_cb_state) {
2939 seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
2941 case NFSD4_CB_FAULT:
2942 seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
2945 seq->status_flags = 0;
2947 if (!list_empty(&clp->cl_revoked))
2948 seq->status_flags |= SEQ4_STATUS_RECALLABLE_STATE_REVOKED;
2952 spin_unlock(&nn->client_lock);
2955 nfsd4_put_session_locked(session);
2956 goto out_no_session;
2960 nfsd4_sequence_done(struct nfsd4_compoundres *resp)
2962 struct nfsd4_compound_state *cs = &resp->cstate;
2964 if (nfsd4_has_session(cs)) {
2965 if (cs->status != nfserr_replay_cache) {
2966 nfsd4_store_cache_entry(resp);
2967 cs->slot->sl_flags &= ~NFSD4_SLOT_INUSE;
2969 /* Drop session reference that was taken in nfsd4_sequence() */
2970 nfsd4_put_session(cs->session);
2972 put_client_renew(cs->clp);
2976 nfsd4_destroy_clientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_destroy_clientid *dc)
2978 struct nfs4_client *conf, *unconf;
2979 struct nfs4_client *clp = NULL;
2981 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2983 spin_lock(&nn->client_lock);
2984 unconf = find_unconfirmed_client(&dc->clientid, true, nn);
2985 conf = find_confirmed_client(&dc->clientid, true, nn);
2986 WARN_ON_ONCE(conf && unconf);
2989 if (client_has_state(conf)) {
2990 status = nfserr_clientid_busy;
2993 status = mark_client_expired_locked(conf);
3000 status = nfserr_stale_clientid;
3003 if (!mach_creds_match(clp, rqstp)) {
3005 status = nfserr_wrong_cred;
3008 unhash_client_locked(clp);
3010 spin_unlock(&nn->client_lock);
3017 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
3021 if (rc->rca_one_fs) {
3022 if (!cstate->current_fh.fh_dentry)
3023 return nfserr_nofilehandle;
3025 * We don't take advantage of the rca_one_fs case.
3026 * That's OK, it's optional, we can safely ignore it.
3031 status = nfserr_complete_already;
3032 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
3033 &cstate->session->se_client->cl_flags))
3036 status = nfserr_stale_clientid;
3037 if (is_client_expired(cstate->session->se_client))
3039 * The following error isn't really legal.
3040 * But we only get here if the client just explicitly
3041 * destroyed the client. Surely it no longer cares what
3042 * error it gets back on an operation for the dead
3048 nfsd4_client_record_create(cstate->session->se_client);
3054 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3055 struct nfsd4_setclientid *setclid)
3057 struct xdr_netobj clname = setclid->se_name;
3058 nfs4_verifier clverifier = setclid->se_verf;
3059 struct nfs4_client *conf, *new;
3060 struct nfs4_client *unconf = NULL;
3062 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3064 new = create_client(clname, rqstp, &clverifier);
3066 return nfserr_jukebox;
3067 /* Cases below refer to rfc 3530 section 14.2.33: */
3068 spin_lock(&nn->client_lock);
3069 conf = find_confirmed_client_by_name(&clname, nn);
3070 if (conf && client_has_state(conf)) {
3072 status = nfserr_clid_inuse;
3073 if (clp_used_exchangeid(conf))
3075 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
3076 char addr_str[INET6_ADDRSTRLEN];
3077 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
3079 dprintk("NFSD: setclientid: string in use by client "
3080 "at %s\n", addr_str);
3084 unconf = find_unconfirmed_client_by_name(&clname, nn);
3086 unhash_client_locked(unconf);
3087 if (conf && same_verf(&conf->cl_verifier, &clverifier)) {
3088 /* case 1: probable callback update */
3089 copy_clid(new, conf);
3090 gen_confirm(new, nn);
3091 } else /* case 4 (new client) or cases 2, 3 (client reboot): */
3093 new->cl_minorversion = 0;
3094 gen_callback(new, setclid, rqstp);
3095 add_to_unconfirmed(new);
3096 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
3097 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
3098 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
3102 spin_unlock(&nn->client_lock);
3106 expire_client(unconf);
3112 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
3113 struct nfsd4_compound_state *cstate,
3114 struct nfsd4_setclientid_confirm *setclientid_confirm)
3116 struct nfs4_client *conf, *unconf;
3117 struct nfs4_client *old = NULL;
3118 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
3119 clientid_t * clid = &setclientid_confirm->sc_clientid;
3121 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3123 if (STALE_CLIENTID(clid, nn))
3124 return nfserr_stale_clientid;
3126 spin_lock(&nn->client_lock);
3127 conf = find_confirmed_client(clid, false, nn);
3128 unconf = find_unconfirmed_client(clid, false, nn);
3130 * We try hard to give out unique clientid's, so if we get an
3131 * attempt to confirm the same clientid with a different cred,
3132 * the client may be buggy; this should never happen.
3134 * Nevertheless, RFC 7530 recommends INUSE for this case:
3136 status = nfserr_clid_inuse;
3137 if (unconf && !same_creds(&unconf->cl_cred, &rqstp->rq_cred))
3139 if (conf && !same_creds(&conf->cl_cred, &rqstp->rq_cred))
3141 /* cases below refer to rfc 3530 section 14.2.34: */
3142 if (!unconf || !same_verf(&confirm, &unconf->cl_confirm)) {
3143 if (conf && !unconf) /* case 2: probable retransmit */
3145 else /* case 4: client hasn't noticed we rebooted yet? */
3146 status = nfserr_stale_clientid;
3150 if (conf) { /* case 1: callback update */
3152 unhash_client_locked(old);
3153 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
3154 } else { /* case 3: normal case; new or rebooted client */
3155 old = find_confirmed_client_by_name(&unconf->cl_name, nn);
3157 status = nfserr_clid_inuse;
3158 if (client_has_state(old)
3159 && !same_creds(&unconf->cl_cred,
3162 status = mark_client_expired_locked(old);
3168 move_to_confirmed(unconf);
3171 get_client_locked(conf);
3172 spin_unlock(&nn->client_lock);
3173 nfsd4_probe_callback(conf);
3174 spin_lock(&nn->client_lock);
3175 put_client_renew_locked(conf);
3177 spin_unlock(&nn->client_lock);
3183 static struct nfs4_file *nfsd4_alloc_file(void)
3185 return kmem_cache_alloc(file_slab, GFP_KERNEL);
3188 /* OPEN Share state helper functions */
3189 static void nfsd4_init_file(struct knfsd_fh *fh, unsigned int hashval,
3190 struct nfs4_file *fp)
3192 lockdep_assert_held(&state_lock);
3194 atomic_set(&fp->fi_ref, 1);
3195 spin_lock_init(&fp->fi_lock);
3196 INIT_LIST_HEAD(&fp->fi_stateids);
3197 INIT_LIST_HEAD(&fp->fi_delegations);
3198 INIT_LIST_HEAD(&fp->fi_clnt_odstate);
3199 fh_copy_shallow(&fp->fi_fhandle, fh);
3200 fp->fi_deleg_file = NULL;
3201 fp->fi_had_conflict = false;
3202 fp->fi_share_deny = 0;
3203 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
3204 memset(fp->fi_access, 0, sizeof(fp->fi_access));
3205 #ifdef CONFIG_NFSD_PNFS
3206 INIT_LIST_HEAD(&fp->fi_lo_states);
3207 atomic_set(&fp->fi_lo_recalls, 0);
3209 hlist_add_head_rcu(&fp->fi_hash, &file_hashtbl[hashval]);
3213 nfsd4_free_slabs(void)
3215 kmem_cache_destroy(odstate_slab);
3216 kmem_cache_destroy(openowner_slab);
3217 kmem_cache_destroy(lockowner_slab);
3218 kmem_cache_destroy(file_slab);
3219 kmem_cache_destroy(stateid_slab);
3220 kmem_cache_destroy(deleg_slab);
3224 nfsd4_init_slabs(void)
3226 openowner_slab = kmem_cache_create("nfsd4_openowners",
3227 sizeof(struct nfs4_openowner), 0, 0, NULL);
3228 if (openowner_slab == NULL)
3230 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
3231 sizeof(struct nfs4_lockowner), 0, 0, NULL);
3232 if (lockowner_slab == NULL)
3233 goto out_free_openowner_slab;
3234 file_slab = kmem_cache_create("nfsd4_files",
3235 sizeof(struct nfs4_file), 0, 0, NULL);
3236 if (file_slab == NULL)
3237 goto out_free_lockowner_slab;
3238 stateid_slab = kmem_cache_create("nfsd4_stateids",
3239 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
3240 if (stateid_slab == NULL)
3241 goto out_free_file_slab;
3242 deleg_slab = kmem_cache_create("nfsd4_delegations",
3243 sizeof(struct nfs4_delegation), 0, 0, NULL);
3244 if (deleg_slab == NULL)
3245 goto out_free_stateid_slab;
3246 odstate_slab = kmem_cache_create("nfsd4_odstate",
3247 sizeof(struct nfs4_clnt_odstate), 0, 0, NULL);
3248 if (odstate_slab == NULL)
3249 goto out_free_deleg_slab;
3252 out_free_deleg_slab:
3253 kmem_cache_destroy(deleg_slab);
3254 out_free_stateid_slab:
3255 kmem_cache_destroy(stateid_slab);
3257 kmem_cache_destroy(file_slab);
3258 out_free_lockowner_slab:
3259 kmem_cache_destroy(lockowner_slab);
3260 out_free_openowner_slab:
3261 kmem_cache_destroy(openowner_slab);
3263 dprintk("nfsd4: out of memory while initializing nfsv4\n");
3267 static void init_nfs4_replay(struct nfs4_replay *rp)
3269 rp->rp_status = nfserr_serverfault;
3271 rp->rp_buf = rp->rp_ibuf;
3272 mutex_init(&rp->rp_mutex);
3275 static void nfsd4_cstate_assign_replay(struct nfsd4_compound_state *cstate,
3276 struct nfs4_stateowner *so)
3278 if (!nfsd4_has_session(cstate)) {
3279 mutex_lock(&so->so_replay.rp_mutex);
3280 cstate->replay_owner = nfs4_get_stateowner(so);
3284 void nfsd4_cstate_clear_replay(struct nfsd4_compound_state *cstate)
3286 struct nfs4_stateowner *so = cstate->replay_owner;
3289 cstate->replay_owner = NULL;
3290 mutex_unlock(&so->so_replay.rp_mutex);
3291 nfs4_put_stateowner(so);
3295 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
3297 struct nfs4_stateowner *sop;
3299 sop = kmem_cache_alloc(slab, GFP_KERNEL);
3303 sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
3304 if (!sop->so_owner.data) {
3305 kmem_cache_free(slab, sop);
3308 sop->so_owner.len = owner->len;
3310 INIT_LIST_HEAD(&sop->so_stateids);
3311 sop->so_client = clp;
3312 init_nfs4_replay(&sop->so_replay);
3313 atomic_set(&sop->so_count, 1);
3317 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
3319 lockdep_assert_held(&clp->cl_lock);
3321 list_add(&oo->oo_owner.so_strhash,
3322 &clp->cl_ownerstr_hashtbl[strhashval]);
3323 list_add(&oo->oo_perclient, &clp->cl_openowners);
3326 static void nfs4_unhash_openowner(struct nfs4_stateowner *so)
3328 unhash_openowner_locked(openowner(so));
3331 static void nfs4_free_openowner(struct nfs4_stateowner *so)
3333 struct nfs4_openowner *oo = openowner(so);
3335 kmem_cache_free(openowner_slab, oo);
3338 static const struct nfs4_stateowner_operations openowner_ops = {
3339 .so_unhash = nfs4_unhash_openowner,
3340 .so_free = nfs4_free_openowner,
3343 static struct nfs4_openowner *
3344 alloc_init_open_stateowner(unsigned int strhashval, struct nfsd4_open *open,
3345 struct nfsd4_compound_state *cstate)
3347 struct nfs4_client *clp = cstate->clp;
3348 struct nfs4_openowner *oo, *ret;
3350 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
3353 oo->oo_owner.so_ops = &openowner_ops;
3354 oo->oo_owner.so_is_open_owner = 1;
3355 oo->oo_owner.so_seqid = open->op_seqid;
3357 if (nfsd4_has_session(cstate))
3358 oo->oo_flags |= NFS4_OO_CONFIRMED;
3360 oo->oo_last_closed_stid = NULL;
3361 INIT_LIST_HEAD(&oo->oo_close_lru);
3362 spin_lock(&clp->cl_lock);
3363 ret = find_openstateowner_str_locked(strhashval, open, clp);
3365 hash_openowner(oo, clp, strhashval);
3368 nfs4_free_stateowner(&oo->oo_owner);
3370 spin_unlock(&clp->cl_lock);
3374 static void init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
3375 struct nfs4_openowner *oo = open->op_openowner;
3377 atomic_inc(&stp->st_stid.sc_count);
3378 stp->st_stid.sc_type = NFS4_OPEN_STID;
3379 INIT_LIST_HEAD(&stp->st_locks);
3380 stp->st_stateowner = nfs4_get_stateowner(&oo->oo_owner);
3382 stp->st_stid.sc_file = fp;
3383 stp->st_access_bmap = 0;
3384 stp->st_deny_bmap = 0;
3385 stp->st_openstp = NULL;
3386 init_rwsem(&stp->st_rwsem);
3387 spin_lock(&oo->oo_owner.so_client->cl_lock);
3388 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
3389 spin_lock(&fp->fi_lock);
3390 list_add(&stp->st_perfile, &fp->fi_stateids);
3391 spin_unlock(&fp->fi_lock);
3392 spin_unlock(&oo->oo_owner.so_client->cl_lock);
3396 * In the 4.0 case we need to keep the owners around a little while to handle
3397 * CLOSE replay. We still do need to release any file access that is held by
3398 * them before returning however.
3401 move_to_close_lru(struct nfs4_ol_stateid *s, struct net *net)
3403 struct nfs4_ol_stateid *last;
3404 struct nfs4_openowner *oo = openowner(s->st_stateowner);
3405 struct nfsd_net *nn = net_generic(s->st_stid.sc_client->net,
3408 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
3411 * We know that we hold one reference via nfsd4_close, and another
3412 * "persistent" reference for the client. If the refcount is higher
3413 * than 2, then there are still calls in progress that are using this
3414 * stateid. We can't put the sc_file reference until they are finished.
3415 * Wait for the refcount to drop to 2. Since it has been unhashed,
3416 * there should be no danger of the refcount going back up again at
3419 wait_event(close_wq, atomic_read(&s->st_stid.sc_count) == 2);
3421 release_all_access(s);
3422 if (s->st_stid.sc_file) {
3423 put_nfs4_file(s->st_stid.sc_file);
3424 s->st_stid.sc_file = NULL;
3427 spin_lock(&nn->client_lock);
3428 last = oo->oo_last_closed_stid;
3429 oo->oo_last_closed_stid = s;
3430 list_move_tail(&oo->oo_close_lru, &nn->close_lru);
3431 oo->oo_time = get_seconds();
3432 spin_unlock(&nn->client_lock);
3434 nfs4_put_stid(&last->st_stid);
3437 /* search file_hashtbl[] for file */
3438 static struct nfs4_file *
3439 find_file_locked(struct knfsd_fh *fh, unsigned int hashval)
3441 struct nfs4_file *fp;
3443 hlist_for_each_entry_rcu(fp, &file_hashtbl[hashval], fi_hash) {
3444 if (fh_match(&fp->fi_fhandle, fh)) {
3445 if (atomic_inc_not_zero(&fp->fi_ref))
3453 find_file(struct knfsd_fh *fh)
3455 struct nfs4_file *fp;
3456 unsigned int hashval = file_hashval(fh);
3459 fp = find_file_locked(fh, hashval);
3464 static struct nfs4_file *
3465 find_or_add_file(struct nfs4_file *new, struct knfsd_fh *fh)
3467 struct nfs4_file *fp;
3468 unsigned int hashval = file_hashval(fh);
3471 fp = find_file_locked(fh, hashval);
3476 spin_lock(&state_lock);
3477 fp = find_file_locked(fh, hashval);
3478 if (likely(fp == NULL)) {
3479 nfsd4_init_file(fh, hashval, new);
3482 spin_unlock(&state_lock);
3488 * Called to check deny when READ with all zero stateid or
3489 * WRITE with all zero or all one stateid
3492 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
3494 struct nfs4_file *fp;
3495 __be32 ret = nfs_ok;
3497 fp = find_file(¤t_fh->fh_handle);
3500 /* Check for conflicting share reservations */
3501 spin_lock(&fp->fi_lock);
3502 if (fp->fi_share_deny & deny_type)
3503 ret = nfserr_locked;
3504 spin_unlock(&fp->fi_lock);
3509 static void nfsd4_cb_recall_prepare(struct nfsd4_callback *cb)
3511 struct nfs4_delegation *dp = cb_to_delegation(cb);
3512 struct nfsd_net *nn = net_generic(dp->dl_stid.sc_client->net,
3515 block_delegations(&dp->dl_stid.sc_file->fi_fhandle);
3518 * We can't do this in nfsd_break_deleg_cb because it is
3519 * already holding inode->i_lock.
3521 * If the dl_time != 0, then we know that it has already been
3522 * queued for a lease break. Don't queue it again.
3524 spin_lock(&state_lock);
3525 if (dp->dl_time == 0) {
3526 dp->dl_time = get_seconds();
3527 list_add_tail(&dp->dl_recall_lru, &nn->del_recall_lru);
3529 spin_unlock(&state_lock);
3532 static int nfsd4_cb_recall_done(struct nfsd4_callback *cb,
3533 struct rpc_task *task)
3535 struct nfs4_delegation *dp = cb_to_delegation(cb);
3537 if (dp->dl_stid.sc_type == NFS4_CLOSED_DELEG_STID)
3540 switch (task->tk_status) {
3544 case -NFS4ERR_BAD_STATEID:
3546 * Race: client probably got cb_recall before open reply
3547 * granting delegation.
3549 if (dp->dl_retries--) {
3550 rpc_delay(task, 2 * HZ);
3559 static void nfsd4_cb_recall_release(struct nfsd4_callback *cb)
3561 struct nfs4_delegation *dp = cb_to_delegation(cb);
3563 nfs4_put_stid(&dp->dl_stid);
3566 static struct nfsd4_callback_ops nfsd4_cb_recall_ops = {
3567 .prepare = nfsd4_cb_recall_prepare,
3568 .done = nfsd4_cb_recall_done,
3569 .release = nfsd4_cb_recall_release,
3572 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
3575 * We're assuming the state code never drops its reference
3576 * without first removing the lease. Since we're in this lease
3577 * callback (and since the lease code is serialized by the kernel
3578 * lock) we know the server hasn't removed the lease yet, we know
3579 * it's safe to take a reference.
3581 atomic_inc(&dp->dl_stid.sc_count);
3582 nfsd4_run_cb(&dp->dl_recall);
3585 /* Called from break_lease() with i_lock held. */
3587 nfsd_break_deleg_cb(struct file_lock *fl)
3590 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
3591 struct nfs4_delegation *dp;
3594 WARN(1, "(%p)->fl_owner NULL\n", fl);
3597 if (fp->fi_had_conflict) {
3598 WARN(1, "duplicate break on %p\n", fp);
3602 * We don't want the locks code to timeout the lease for us;
3603 * we'll remove it ourself if a delegation isn't returned
3606 fl->fl_break_time = 0;
3608 spin_lock(&fp->fi_lock);
3609 fp->fi_had_conflict = true;
3611 * If there are no delegations on the list, then return true
3612 * so that the lease code will go ahead and delete it.
3614 if (list_empty(&fp->fi_delegations))
3617 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
3618 nfsd_break_one_deleg(dp);
3619 spin_unlock(&fp->fi_lock);
3624 nfsd_change_deleg_cb(struct file_lock *onlist, int arg,
3625 struct list_head *dispose)
3628 return lease_modify(onlist, arg, dispose);
3633 static const struct lock_manager_operations nfsd_lease_mng_ops = {
3634 .lm_break = nfsd_break_deleg_cb,
3635 .lm_change = nfsd_change_deleg_cb,
3638 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
3640 if (nfsd4_has_session(cstate))
3642 if (seqid == so->so_seqid - 1)
3643 return nfserr_replay_me;
3644 if (seqid == so->so_seqid)
3646 return nfserr_bad_seqid;
3649 static __be32 lookup_clientid(clientid_t *clid,
3650 struct nfsd4_compound_state *cstate,
3651 struct nfsd_net *nn)
3653 struct nfs4_client *found;
3656 found = cstate->clp;
3657 if (!same_clid(&found->cl_clientid, clid))
3658 return nfserr_stale_clientid;
3662 if (STALE_CLIENTID(clid, nn))
3663 return nfserr_stale_clientid;
3666 * For v4.1+ we get the client in the SEQUENCE op. If we don't have one
3667 * cached already then we know this is for is for v4.0 and "sessions"
3670 WARN_ON_ONCE(cstate->session);
3671 spin_lock(&nn->client_lock);
3672 found = find_confirmed_client(clid, false, nn);
3674 spin_unlock(&nn->client_lock);
3675 return nfserr_expired;
3677 atomic_inc(&found->cl_refcount);
3678 spin_unlock(&nn->client_lock);
3680 /* Cache the nfs4_client in cstate! */
3681 cstate->clp = found;
3686 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
3687 struct nfsd4_open *open, struct nfsd_net *nn)
3689 clientid_t *clientid = &open->op_clientid;
3690 struct nfs4_client *clp = NULL;
3691 unsigned int strhashval;
3692 struct nfs4_openowner *oo = NULL;
3695 if (STALE_CLIENTID(&open->op_clientid, nn))
3696 return nfserr_stale_clientid;
3698 * In case we need it later, after we've already created the
3699 * file and don't want to risk a further failure:
3701 open->op_file = nfsd4_alloc_file();
3702 if (open->op_file == NULL)
3703 return nfserr_jukebox;
3705 status = lookup_clientid(clientid, cstate, nn);
3710 strhashval = ownerstr_hashval(&open->op_owner);
3711 oo = find_openstateowner_str(strhashval, open, clp);
3712 open->op_openowner = oo;
3716 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
3717 /* Replace unconfirmed owners without checking for replay. */
3718 release_openowner(oo);
3719 open->op_openowner = NULL;
3722 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
3727 oo = alloc_init_open_stateowner(strhashval, open, cstate);
3729 return nfserr_jukebox;
3730 open->op_openowner = oo;
3732 open->op_stp = nfs4_alloc_open_stateid(clp);
3734 return nfserr_jukebox;
3736 if (nfsd4_has_session(cstate) &&
3737 (cstate->current_fh.fh_export->ex_flags & NFSEXP_PNFS)) {
3738 open->op_odstate = alloc_clnt_odstate(clp);
3739 if (!open->op_odstate)
3740 return nfserr_jukebox;
3746 static inline __be32
3747 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
3749 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
3750 return nfserr_openmode;
3755 static int share_access_to_flags(u32 share_access)
3757 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
3760 static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
3762 struct nfs4_stid *ret;
3764 ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
3767 return delegstateid(ret);
3770 static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
3772 return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
3773 open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
3777 nfs4_check_deleg(struct nfs4_client *cl, struct nfsd4_open *open,
3778 struct nfs4_delegation **dp)
3781 __be32 status = nfserr_bad_stateid;
3782 struct nfs4_delegation *deleg;
3784 deleg = find_deleg_stateid(cl, &open->op_delegate_stateid);
3787 flags = share_access_to_flags(open->op_share_access);
3788 status = nfs4_check_delegmode(deleg, flags);
3790 nfs4_put_stid(&deleg->dl_stid);
3795 if (!nfsd4_is_deleg_cur(open))
3799 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3803 static struct nfs4_ol_stateid *
3804 nfsd4_find_existing_open(struct nfs4_file *fp, struct nfsd4_open *open)
3806 struct nfs4_ol_stateid *local, *ret = NULL;
3807 struct nfs4_openowner *oo = open->op_openowner;
3809 spin_lock(&fp->fi_lock);
3810 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
3811 /* ignore lock owners */
3812 if (local->st_stateowner->so_is_open_owner == 0)
3814 if (local->st_stateowner == &oo->oo_owner) {
3816 atomic_inc(&ret->st_stid.sc_count);
3820 spin_unlock(&fp->fi_lock);
3824 static inline int nfs4_access_to_access(u32 nfs4_access)
3828 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
3829 flags |= NFSD_MAY_READ;
3830 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
3831 flags |= NFSD_MAY_WRITE;
3835 static inline __be32
3836 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
3837 struct nfsd4_open *open)
3839 struct iattr iattr = {
3840 .ia_valid = ATTR_SIZE,
3843 if (!open->op_truncate)
3845 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
3846 return nfserr_inval;
3847 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
3850 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
3851 struct svc_fh *cur_fh, struct nfs4_ol_stateid *stp,
3852 struct nfsd4_open *open)
3854 struct file *filp = NULL;
3856 int oflag = nfs4_access_to_omode(open->op_share_access);
3857 int access = nfs4_access_to_access(open->op_share_access);
3858 unsigned char old_access_bmap, old_deny_bmap;
3860 spin_lock(&fp->fi_lock);
3863 * Are we trying to set a deny mode that would conflict with
3866 status = nfs4_file_check_deny(fp, open->op_share_deny);
3867 if (status != nfs_ok) {
3868 spin_unlock(&fp->fi_lock);
3872 /* set access to the file */
3873 status = nfs4_file_get_access(fp, open->op_share_access);
3874 if (status != nfs_ok) {
3875 spin_unlock(&fp->fi_lock);
3879 /* Set access bits in stateid */
3880 old_access_bmap = stp->st_access_bmap;
3881 set_access(open->op_share_access, stp);
3883 /* Set new deny mask */
3884 old_deny_bmap = stp->st_deny_bmap;
3885 set_deny(open->op_share_deny, stp);
3886 fp->fi_share_deny |= (open->op_share_deny & NFS4_SHARE_DENY_BOTH);
3888 if (!fp->fi_fds[oflag]) {
3889 spin_unlock(&fp->fi_lock);
3890 status = nfsd_open(rqstp, cur_fh, S_IFREG, access, &filp);
3892 goto out_put_access;
3893 spin_lock(&fp->fi_lock);
3894 if (!fp->fi_fds[oflag]) {
3895 fp->fi_fds[oflag] = filp;
3899 spin_unlock(&fp->fi_lock);
3903 status = nfsd4_truncate(rqstp, cur_fh, open);
3905 goto out_put_access;
3909 stp->st_access_bmap = old_access_bmap;
3910 nfs4_file_put_access(fp, open->op_share_access);
3911 reset_union_bmap_deny(bmap_to_share_mode(old_deny_bmap), stp);
3916 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)
3919 unsigned char old_deny_bmap = stp->st_deny_bmap;
3921 if (!test_access(open->op_share_access, stp))
3922 return nfs4_get_vfs_file(rqstp, fp, cur_fh, stp, open);
3924 /* test and set deny mode */
3925 spin_lock(&fp->fi_lock);
3926 status = nfs4_file_check_deny(fp, open->op_share_deny);
3927 if (status == nfs_ok) {
3928 set_deny(open->op_share_deny, stp);
3929 fp->fi_share_deny |=
3930 (open->op_share_deny & NFS4_SHARE_DENY_BOTH);
3932 spin_unlock(&fp->fi_lock);
3934 if (status != nfs_ok)
3937 status = nfsd4_truncate(rqstp, cur_fh, open);
3938 if (status != nfs_ok)
3939 reset_union_bmap_deny(old_deny_bmap, stp);
3943 /* Should we give out recallable state?: */
3944 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
3946 if (clp->cl_cb_state == NFSD4_CB_UP)
3949 * In the sessions case, since we don't have to establish a
3950 * separate connection for callbacks, we assume it's OK
3951 * until we hear otherwise:
3953 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
3956 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_file *fp, int flag)
3958 struct file_lock *fl;
3960 fl = locks_alloc_lock();
3963 fl->fl_lmops = &nfsd_lease_mng_ops;
3964 fl->fl_flags = FL_DELEG;
3965 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
3966 fl->fl_end = OFFSET_MAX;
3967 fl->fl_owner = (fl_owner_t)fp;
3968 fl->fl_pid = current->tgid;
3972 static int nfs4_setlease(struct nfs4_delegation *dp)
3974 struct nfs4_file *fp = dp->dl_stid.sc_file;
3975 struct file_lock *fl;
3979 fl = nfs4_alloc_init_lease(fp, NFS4_OPEN_DELEGATE_READ);
3982 filp = find_readable_file(fp);
3984 /* We should always have a readable file here */
3986 locks_free_lock(fl);
3990 status = vfs_setlease(filp, fl->fl_type, &fl, NULL);
3992 locks_free_lock(fl);
3995 spin_lock(&state_lock);
3996 spin_lock(&fp->fi_lock);
3997 /* Did the lease get broken before we took the lock? */
3999 if (fp->fi_had_conflict)
4002 if (fp->fi_deleg_file) {
4005 hash_delegation_locked(dp, fp);
4008 fp->fi_deleg_file = filp;
4009 fp->fi_delegees = 1;
4010 hash_delegation_locked(dp, fp);
4011 spin_unlock(&fp->fi_lock);
4012 spin_unlock(&state_lock);
4015 spin_unlock(&fp->fi_lock);
4016 spin_unlock(&state_lock);
4022 static struct nfs4_delegation *
4023 nfs4_set_delegation(struct nfs4_client *clp, struct svc_fh *fh,
4024 struct nfs4_file *fp, struct nfs4_clnt_odstate *odstate)
4027 struct nfs4_delegation *dp;
4029 if (fp->fi_had_conflict)
4030 return ERR_PTR(-EAGAIN);
4032 dp = alloc_init_deleg(clp, fh, odstate);
4034 return ERR_PTR(-ENOMEM);
4037 spin_lock(&state_lock);
4038 spin_lock(&fp->fi_lock);
4039 dp->dl_stid.sc_file = fp;
4040 if (!fp->fi_deleg_file) {
4041 spin_unlock(&fp->fi_lock);
4042 spin_unlock(&state_lock);
4043 status = nfs4_setlease(dp);
4046 if (fp->fi_had_conflict) {
4051 hash_delegation_locked(dp, fp);
4054 spin_unlock(&fp->fi_lock);
4055 spin_unlock(&state_lock);
4058 put_clnt_odstate(dp->dl_clnt_odstate);
4059 nfs4_put_stid(&dp->dl_stid);
4060 return ERR_PTR(status);
4065 static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
4067 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
4068 if (status == -EAGAIN)
4069 open->op_why_no_deleg = WND4_CONTENTION;
4071 open->op_why_no_deleg = WND4_RESOURCE;
4072 switch (open->op_deleg_want) {
4073 case NFS4_SHARE_WANT_READ_DELEG:
4074 case NFS4_SHARE_WANT_WRITE_DELEG:
4075 case NFS4_SHARE_WANT_ANY_DELEG:
4077 case NFS4_SHARE_WANT_CANCEL:
4078 open->op_why_no_deleg = WND4_CANCELLED;
4080 case NFS4_SHARE_WANT_NO_DELEG:
4087 * Attempt to hand out a delegation.
4089 * Note we don't support write delegations, and won't until the vfs has
4090 * proper support for them.
4093 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open,
4094 struct nfs4_ol_stateid *stp)
4096 struct nfs4_delegation *dp;
4097 struct nfs4_openowner *oo = openowner(stp->st_stateowner);
4098 struct nfs4_client *clp = stp->st_stid.sc_client;
4102 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
4103 open->op_recall = 0;
4104 switch (open->op_claim_type) {
4105 case NFS4_OPEN_CLAIM_PREVIOUS:
4107 open->op_recall = 1;
4108 if (open->op_delegate_type != NFS4_OPEN_DELEGATE_READ)
4111 case NFS4_OPEN_CLAIM_NULL:
4112 case NFS4_OPEN_CLAIM_FH:
4114 * Let's not give out any delegations till everyone's
4115 * had the chance to reclaim theirs, *and* until
4116 * NLM locks have all been reclaimed:
4118 if (locks_in_grace(clp->net))
4120 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
4123 * Also, if the file was opened for write or
4124 * create, there's a good chance the client's
4125 * about to write to it, resulting in an
4126 * immediate recall (since we don't support
4127 * write delegations):
4129 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
4131 if (open->op_create == NFS4_OPEN_CREATE)
4137 dp = nfs4_set_delegation(clp, fh, stp->st_stid.sc_file, stp->st_clnt_odstate);
4141 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
4143 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
4144 STATEID_VAL(&dp->dl_stid.sc_stateid));
4145 open->op_delegate_type = NFS4_OPEN_DELEGATE_READ;
4146 nfs4_put_stid(&dp->dl_stid);
4149 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE;
4150 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
4151 open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE) {
4152 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
4153 open->op_recall = 1;
4156 /* 4.1 client asking for a delegation? */
4157 if (open->op_deleg_want)
4158 nfsd4_open_deleg_none_ext(open, status);
4162 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
4163 struct nfs4_delegation *dp)
4165 if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
4166 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
4167 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
4168 open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
4169 } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
4170 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
4171 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
4172 open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
4174 /* Otherwise the client must be confused wanting a delegation
4175 * it already has, therefore we don't return
4176 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
4181 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
4183 struct nfsd4_compoundres *resp = rqstp->rq_resp;
4184 struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
4185 struct nfs4_file *fp = NULL;
4186 struct nfs4_ol_stateid *stp = NULL;
4187 struct nfs4_delegation *dp = NULL;
4191 * Lookup file; if found, lookup stateid and check open request,
4192 * and check for delegations in the process of being recalled.
4193 * If not found, create the nfs4_file struct
4195 fp = find_or_add_file(open->op_file, ¤t_fh->fh_handle);
4196 if (fp != open->op_file) {
4197 status = nfs4_check_deleg(cl, open, &dp);
4200 stp = nfsd4_find_existing_open(fp, open);
4202 open->op_file = NULL;
4203 status = nfserr_bad_stateid;
4204 if (nfsd4_is_deleg_cur(open))
4209 * OPEN the file, or upgrade an existing OPEN.
4210 * If truncate fails, the OPEN fails.
4213 /* Stateid was found, this is an OPEN upgrade */
4214 down_read(&stp->st_rwsem);
4215 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
4217 up_read(&stp->st_rwsem);
4222 open->op_stp = NULL;
4223 init_open_stateid(stp, fp, open);
4224 down_read(&stp->st_rwsem);
4225 status = nfs4_get_vfs_file(rqstp, fp, current_fh, stp, open);
4227 up_read(&stp->st_rwsem);
4228 release_open_stateid(stp);
4232 stp->st_clnt_odstate = find_or_hash_clnt_odstate(fp,
4234 if (stp->st_clnt_odstate == open->op_odstate)
4235 open->op_odstate = NULL;
4237 nfs4_inc_and_copy_stateid(&open->op_stateid, &stp->st_stid);
4238 up_read(&stp->st_rwsem);
4240 if (nfsd4_has_session(&resp->cstate)) {
4241 if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
4242 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
4243 open->op_why_no_deleg = WND4_NOT_WANTED;
4249 * Attempt to hand out a delegation. No error return, because the
4250 * OPEN succeeds even if we fail.
4252 nfs4_open_delegation(current_fh, open, stp);
4256 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
4257 STATEID_VAL(&stp->st_stid.sc_stateid));
4259 /* 4.1 client trying to upgrade/downgrade delegation? */
4260 if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
4261 open->op_deleg_want)
4262 nfsd4_deleg_xgrade_none_ext(open, dp);
4266 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
4267 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
4269 * To finish the open response, we just need to set the rflags.
4271 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
4272 if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
4273 !nfsd4_has_session(&resp->cstate))
4274 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
4276 nfs4_put_stid(&dp->dl_stid);
4278 nfs4_put_stid(&stp->st_stid);
4283 void nfsd4_cleanup_open_state(struct nfsd4_compound_state *cstate,
4284 struct nfsd4_open *open)
4286 if (open->op_openowner) {
4287 struct nfs4_stateowner *so = &open->op_openowner->oo_owner;
4289 nfsd4_cstate_assign_replay(cstate, so);
4290 nfs4_put_stateowner(so);
4293 kmem_cache_free(file_slab, open->op_file);
4295 nfs4_put_stid(&open->op_stp->st_stid);
4296 if (open->op_odstate)
4297 kmem_cache_free(odstate_slab, open->op_odstate);
4301 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4304 struct nfs4_client *clp;
4306 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4308 dprintk("process_renew(%08x/%08x): starting\n",
4309 clid->cl_boot, clid->cl_id);
4310 status = lookup_clientid(clid, cstate, nn);
4314 status = nfserr_cb_path_down;
4315 if (!list_empty(&clp->cl_delegations)
4316 && clp->cl_cb_state != NFSD4_CB_UP)
4324 nfsd4_end_grace(struct nfsd_net *nn)
4326 /* do nothing if grace period already ended */
4327 if (nn->grace_ended)
4330 dprintk("NFSD: end of grace period\n");
4331 nn->grace_ended = true;
4333 * If the server goes down again right now, an NFSv4
4334 * client will still be allowed to reclaim after it comes back up,
4335 * even if it hasn't yet had a chance to reclaim state this time.
4338 nfsd4_record_grace_done(nn);
4340 * At this point, NFSv4 clients can still reclaim. But if the
4341 * server crashes, any that have not yet reclaimed will be out
4342 * of luck on the next boot.
4344 * (NFSv4.1+ clients are considered to have reclaimed once they
4345 * call RECLAIM_COMPLETE. NFSv4.0 clients are considered to
4346 * have reclaimed after their first OPEN.)
4348 locks_end_grace(&nn->nfsd4_manager);
4350 * At this point, and once lockd and/or any other containers
4351 * exit their grace period, further reclaims will fail and
4352 * regular locking can resume.
4357 nfs4_laundromat(struct nfsd_net *nn)
4359 struct nfs4_client *clp;
4360 struct nfs4_openowner *oo;
4361 struct nfs4_delegation *dp;
4362 struct nfs4_ol_stateid *stp;
4363 struct list_head *pos, *next, reaplist;
4364 time_t cutoff = get_seconds() - nn->nfsd4_lease;
4365 time_t t, new_timeo = nn->nfsd4_lease;
4367 dprintk("NFSD: laundromat service - starting\n");
4368 nfsd4_end_grace(nn);
4369 INIT_LIST_HEAD(&reaplist);
4370 spin_lock(&nn->client_lock);
4371 list_for_each_safe(pos, next, &nn->client_lru) {
4372 clp = list_entry(pos, struct nfs4_client, cl_lru);
4373 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
4374 t = clp->cl_time - cutoff;
4375 new_timeo = min(new_timeo, t);
4378 if (mark_client_expired_locked(clp)) {
4379 dprintk("NFSD: client in use (clientid %08x)\n",
4380 clp->cl_clientid.cl_id);
4383 list_add(&clp->cl_lru, &reaplist);
4385 spin_unlock(&nn->client_lock);
4386 list_for_each_safe(pos, next, &reaplist) {
4387 clp = list_entry(pos, struct nfs4_client, cl_lru);
4388 dprintk("NFSD: purging unused client (clientid %08x)\n",
4389 clp->cl_clientid.cl_id);
4390 list_del_init(&clp->cl_lru);
4393 spin_lock(&state_lock);
4394 list_for_each_safe(pos, next, &nn->del_recall_lru) {
4395 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4396 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
4397 t = dp->dl_time - cutoff;
4398 new_timeo = min(new_timeo, t);
4401 WARN_ON(!unhash_delegation_locked(dp));
4402 list_add(&dp->dl_recall_lru, &reaplist);
4404 spin_unlock(&state_lock);
4405 while (!list_empty(&reaplist)) {
4406 dp = list_first_entry(&reaplist, struct nfs4_delegation,
4408 list_del_init(&dp->dl_recall_lru);
4409 revoke_delegation(dp);
4412 spin_lock(&nn->client_lock);
4413 while (!list_empty(&nn->close_lru)) {
4414 oo = list_first_entry(&nn->close_lru, struct nfs4_openowner,
4416 if (time_after((unsigned long)oo->oo_time,
4417 (unsigned long)cutoff)) {
4418 t = oo->oo_time - cutoff;
4419 new_timeo = min(new_timeo, t);
4422 list_del_init(&oo->oo_close_lru);
4423 stp = oo->oo_last_closed_stid;
4424 oo->oo_last_closed_stid = NULL;
4425 spin_unlock(&nn->client_lock);
4426 nfs4_put_stid(&stp->st_stid);
4427 spin_lock(&nn->client_lock);
4429 spin_unlock(&nn->client_lock);
4431 new_timeo = max_t(time_t, new_timeo, NFSD_LAUNDROMAT_MINTIMEOUT);
4435 static struct workqueue_struct *laundry_wq;
4436 static void laundromat_main(struct work_struct *);
4439 laundromat_main(struct work_struct *laundry)
4442 struct delayed_work *dwork = container_of(laundry, struct delayed_work,
4444 struct nfsd_net *nn = container_of(dwork, struct nfsd_net,
4447 t = nfs4_laundromat(nn);
4448 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
4449 queue_delayed_work(laundry_wq, &nn->laundromat_work, t*HZ);
4452 static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stid *stp)
4454 if (!fh_match(&fhp->fh_handle, &stp->sc_file->fi_fhandle))
4455 return nfserr_bad_stateid;
4460 access_permit_read(struct nfs4_ol_stateid *stp)
4462 return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
4463 test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
4464 test_access(NFS4_SHARE_ACCESS_WRITE, stp);
4468 access_permit_write(struct nfs4_ol_stateid *stp)
4470 return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
4471 test_access(NFS4_SHARE_ACCESS_BOTH, stp);
4475 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
4477 __be32 status = nfserr_openmode;
4479 /* For lock stateid's, we test the parent open, not the lock: */
4480 if (stp->st_openstp)
4481 stp = stp->st_openstp;
4482 if ((flags & WR_STATE) && !access_permit_write(stp))
4484 if ((flags & RD_STATE) && !access_permit_read(stp))
4491 static inline __be32
4492 check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
4494 if (ONE_STATEID(stateid) && (flags & RD_STATE))
4496 else if (opens_in_grace(net)) {
4497 /* Answer in remaining cases depends on existence of
4498 * conflicting state; so we must wait out the grace period. */
4499 return nfserr_grace;
4500 } else if (flags & WR_STATE)
4501 return nfs4_share_conflict(current_fh,
4502 NFS4_SHARE_DENY_WRITE);
4503 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
4504 return nfs4_share_conflict(current_fh,
4505 NFS4_SHARE_DENY_READ);
4509 * Allow READ/WRITE during grace period on recovered state only for files
4510 * that are not able to provide mandatory locking.
4513 grace_disallows_io(struct net *net, struct inode *inode)
4515 return opens_in_grace(net) && mandatory_lock(inode);
4518 /* Returns true iff a is later than b: */
4519 static bool stateid_generation_after(stateid_t *a, stateid_t *b)
4521 return (s32)(a->si_generation - b->si_generation) > 0;
4524 static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
4527 * When sessions are used the stateid generation number is ignored
4530 if (has_session && in->si_generation == 0)
4533 if (in->si_generation == ref->si_generation)
4536 /* If the client sends us a stateid from the future, it's buggy: */
4537 if (stateid_generation_after(in, ref))
4538 return nfserr_bad_stateid;
4540 * However, we could see a stateid from the past, even from a
4541 * non-buggy client. For example, if the client sends a lock
4542 * while some IO is outstanding, the lock may bump si_generation
4543 * while the IO is still in flight. The client could avoid that
4544 * situation by waiting for responses on all the IO requests,
4545 * but better performance may result in retrying IO that
4546 * receives an old_stateid error if requests are rarely
4547 * reordered in flight:
4549 return nfserr_old_stateid;
4552 static __be32 nfsd4_check_openowner_confirmed(struct nfs4_ol_stateid *ols)
4554 if (ols->st_stateowner->so_is_open_owner &&
4555 !(openowner(ols->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
4556 return nfserr_bad_stateid;
4560 static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
4562 struct nfs4_stid *s;
4563 __be32 status = nfserr_bad_stateid;
4565 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
4567 /* Client debugging aid. */
4568 if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid)) {
4569 char addr_str[INET6_ADDRSTRLEN];
4570 rpc_ntop((struct sockaddr *)&cl->cl_addr, addr_str,
4572 pr_warn_ratelimited("NFSD: client %s testing state ID "
4573 "with incorrect client ID\n", addr_str);
4576 spin_lock(&cl->cl_lock);
4577 s = find_stateid_locked(cl, stateid);
4580 status = check_stateid_generation(stateid, &s->sc_stateid, 1);
4583 switch (s->sc_type) {
4584 case NFS4_DELEG_STID:
4587 case NFS4_REVOKED_DELEG_STID:
4588 status = nfserr_deleg_revoked;
4590 case NFS4_OPEN_STID:
4591 case NFS4_LOCK_STID:
4592 status = nfsd4_check_openowner_confirmed(openlockstateid(s));
4595 printk("unknown stateid type %x\n", s->sc_type);
4597 case NFS4_CLOSED_STID:
4598 case NFS4_CLOSED_DELEG_STID:
4599 status = nfserr_bad_stateid;
4602 spin_unlock(&cl->cl_lock);
4607 nfsd4_lookup_stateid(struct nfsd4_compound_state *cstate,
4608 stateid_t *stateid, unsigned char typemask,
4609 struct nfs4_stid **s, struct nfsd_net *nn)
4613 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
4614 return nfserr_bad_stateid;
4615 status = lookup_clientid(&stateid->si_opaque.so_clid, cstate, nn);
4616 if (status == nfserr_stale_clientid) {
4617 if (cstate->session)
4618 return nfserr_bad_stateid;
4619 return nfserr_stale_stateid;
4623 *s = find_stateid_by_type(cstate->clp, stateid, typemask);
4625 return nfserr_bad_stateid;
4629 static struct file *
4630 nfs4_find_file(struct nfs4_stid *s, int flags)
4635 switch (s->sc_type) {
4636 case NFS4_DELEG_STID:
4637 if (WARN_ON_ONCE(!s->sc_file->fi_deleg_file))
4639 return get_file(s->sc_file->fi_deleg_file);
4640 case NFS4_OPEN_STID:
4641 case NFS4_LOCK_STID:
4642 if (flags & RD_STATE)
4643 return find_readable_file(s->sc_file);
4645 return find_writeable_file(s->sc_file);
4653 nfs4_check_olstateid(struct svc_fh *fhp, struct nfs4_ol_stateid *ols, int flags)
4657 status = nfsd4_check_openowner_confirmed(ols);
4660 return nfs4_check_openmode(ols, flags);
4664 nfs4_check_file(struct svc_rqst *rqstp, struct svc_fh *fhp, struct nfs4_stid *s,
4665 struct file **filpp, bool *tmp_file, int flags)
4667 int acc = (flags & RD_STATE) ? NFSD_MAY_READ : NFSD_MAY_WRITE;
4671 file = nfs4_find_file(s, flags);
4673 status = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry,
4674 acc | NFSD_MAY_OWNER_OVERRIDE);
4682 status = nfsd_open(rqstp, fhp, S_IFREG, acc, filpp);
4694 * Checks for stateid operations
4697 nfs4_preprocess_stateid_op(struct svc_rqst *rqstp,
4698 struct nfsd4_compound_state *cstate, stateid_t *stateid,
4699 int flags, struct file **filpp, bool *tmp_file)
4701 struct svc_fh *fhp = &cstate->current_fh;
4702 struct inode *ino = d_inode(fhp->fh_dentry);
4703 struct net *net = SVC_NET(rqstp);
4704 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4705 struct nfs4_stid *s = NULL;
4713 if (grace_disallows_io(net, ino))
4714 return nfserr_grace;
4716 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
4717 status = check_special_stateids(net, fhp, stateid, flags);
4721 status = nfsd4_lookup_stateid(cstate, stateid,
4722 NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID,
4726 status = check_stateid_generation(stateid, &s->sc_stateid,
4727 nfsd4_has_session(cstate));
4731 switch (s->sc_type) {
4732 case NFS4_DELEG_STID:
4733 status = nfs4_check_delegmode(delegstateid(s), flags);
4735 case NFS4_OPEN_STID:
4736 case NFS4_LOCK_STID:
4737 status = nfs4_check_olstateid(fhp, openlockstateid(s), flags);
4740 status = nfserr_bad_stateid;
4745 status = nfs4_check_fh(fhp, s);
4748 if (!status && filpp)
4749 status = nfs4_check_file(rqstp, fhp, s, filpp, tmp_file, flags);
4757 * Test if the stateid is valid
4760 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4761 struct nfsd4_test_stateid *test_stateid)
4763 struct nfsd4_test_stateid_id *stateid;
4764 struct nfs4_client *cl = cstate->session->se_client;
4766 list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
4767 stateid->ts_id_status =
4768 nfsd4_validate_stateid(cl, &stateid->ts_id_stateid);
4774 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4775 struct nfsd4_free_stateid *free_stateid)
4777 stateid_t *stateid = &free_stateid->fr_stateid;
4778 struct nfs4_stid *s;
4779 struct nfs4_delegation *dp;
4780 struct nfs4_ol_stateid *stp;
4781 struct nfs4_client *cl = cstate->session->se_client;
4782 __be32 ret = nfserr_bad_stateid;
4784 spin_lock(&cl->cl_lock);
4785 s = find_stateid_locked(cl, stateid);
4788 switch (s->sc_type) {
4789 case NFS4_DELEG_STID:
4790 ret = nfserr_locks_held;
4792 case NFS4_OPEN_STID:
4793 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
4796 ret = nfserr_locks_held;
4798 case NFS4_LOCK_STID:
4799 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
4802 stp = openlockstateid(s);
4803 ret = nfserr_locks_held;
4804 if (check_for_locks(stp->st_stid.sc_file,
4805 lockowner(stp->st_stateowner)))
4807 WARN_ON(!unhash_lock_stateid(stp));
4808 spin_unlock(&cl->cl_lock);
4812 case NFS4_REVOKED_DELEG_STID:
4813 dp = delegstateid(s);
4814 list_del_init(&dp->dl_recall_lru);
4815 spin_unlock(&cl->cl_lock);
4819 /* Default falls through and returns nfserr_bad_stateid */
4822 spin_unlock(&cl->cl_lock);
4830 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
4831 RD_STATE : WR_STATE;
4834 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
4836 struct svc_fh *current_fh = &cstate->current_fh;
4837 struct nfs4_stateowner *sop = stp->st_stateowner;
4840 status = nfsd4_check_seqid(cstate, sop, seqid);
4843 if (stp->st_stid.sc_type == NFS4_CLOSED_STID
4844 || stp->st_stid.sc_type == NFS4_REVOKED_DELEG_STID)
4846 * "Closed" stateid's exist *only* to return
4847 * nfserr_replay_me from the previous step, and
4848 * revoked delegations are kept only for free_stateid.
4850 return nfserr_bad_stateid;
4851 down_write(&stp->st_rwsem);
4852 status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
4853 if (status == nfs_ok)
4854 status = nfs4_check_fh(current_fh, &stp->st_stid);
4855 if (status != nfs_ok)
4856 up_write(&stp->st_rwsem);
4861 * Checks for sequence id mutating operations.
4864 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
4865 stateid_t *stateid, char typemask,
4866 struct nfs4_ol_stateid **stpp,
4867 struct nfsd_net *nn)
4870 struct nfs4_stid *s;
4871 struct nfs4_ol_stateid *stp = NULL;
4873 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
4874 seqid, STATEID_VAL(stateid));
4877 status = nfsd4_lookup_stateid(cstate, stateid, typemask, &s, nn);
4880 stp = openlockstateid(s);
4881 nfsd4_cstate_assign_replay(cstate, stp->st_stateowner);
4883 status = nfs4_seqid_op_checks(cstate, stateid, seqid, stp);
4887 nfs4_put_stid(&stp->st_stid);
4891 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
4892 stateid_t *stateid, struct nfs4_ol_stateid **stpp, struct nfsd_net *nn)
4895 struct nfs4_openowner *oo;
4896 struct nfs4_ol_stateid *stp;
4898 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
4899 NFS4_OPEN_STID, &stp, nn);
4902 oo = openowner(stp->st_stateowner);
4903 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
4904 up_write(&stp->st_rwsem);
4905 nfs4_put_stid(&stp->st_stid);
4906 return nfserr_bad_stateid;
4913 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4914 struct nfsd4_open_confirm *oc)
4917 struct nfs4_openowner *oo;
4918 struct nfs4_ol_stateid *stp;
4919 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4921 dprintk("NFSD: nfsd4_open_confirm on file %pd\n",
4922 cstate->current_fh.fh_dentry);
4924 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
4928 status = nfs4_preprocess_seqid_op(cstate,
4929 oc->oc_seqid, &oc->oc_req_stateid,
4930 NFS4_OPEN_STID, &stp, nn);
4933 oo = openowner(stp->st_stateowner);
4934 status = nfserr_bad_stateid;
4935 if (oo->oo_flags & NFS4_OO_CONFIRMED) {
4936 up_write(&stp->st_rwsem);
4939 oo->oo_flags |= NFS4_OO_CONFIRMED;
4940 nfs4_inc_and_copy_stateid(&oc->oc_resp_stateid, &stp->st_stid);
4941 up_write(&stp->st_rwsem);
4942 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
4943 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
4945 nfsd4_client_record_create(oo->oo_owner.so_client);
4948 nfs4_put_stid(&stp->st_stid);
4950 nfsd4_bump_seqid(cstate, status);
4954 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
4956 if (!test_access(access, stp))
4958 nfs4_file_put_access(stp->st_stid.sc_file, access);
4959 clear_access(access, stp);
4962 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
4964 switch (to_access) {
4965 case NFS4_SHARE_ACCESS_READ:
4966 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
4967 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
4969 case NFS4_SHARE_ACCESS_WRITE:
4970 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
4971 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
4973 case NFS4_SHARE_ACCESS_BOTH:
4981 nfsd4_open_downgrade(struct svc_rqst *rqstp,
4982 struct nfsd4_compound_state *cstate,
4983 struct nfsd4_open_downgrade *od)
4986 struct nfs4_ol_stateid *stp;
4987 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4989 dprintk("NFSD: nfsd4_open_downgrade on file %pd\n",
4990 cstate->current_fh.fh_dentry);
4992 /* We don't yet support WANT bits: */
4993 if (od->od_deleg_want)
4994 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
4997 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
4998 &od->od_stateid, &stp, nn);
5001 status = nfserr_inval;
5002 if (!test_access(od->od_share_access, stp)) {
5003 dprintk("NFSD: access not a subset of current bitmap: 0x%hhx, input access=%08x\n",
5004 stp->st_access_bmap, od->od_share_access);
5007 if (!test_deny(od->od_share_deny, stp)) {
5008 dprintk("NFSD: deny not a subset of current bitmap: 0x%hhx, input deny=%08x\n",
5009 stp->st_deny_bmap, od->od_share_deny);
5012 nfs4_stateid_downgrade(stp, od->od_share_access);
5013 reset_union_bmap_deny(od->od_share_deny, stp);
5014 nfs4_inc_and_copy_stateid(&od->od_stateid, &stp->st_stid);
5017 up_write(&stp->st_rwsem);
5018 nfs4_put_stid(&stp->st_stid);
5020 nfsd4_bump_seqid(cstate, status);
5024 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
5026 struct nfs4_client *clp = s->st_stid.sc_client;
5028 LIST_HEAD(reaplist);
5030 s->st_stid.sc_type = NFS4_CLOSED_STID;
5031 spin_lock(&clp->cl_lock);
5032 unhashed = unhash_open_stateid(s, &reaplist);
5034 if (clp->cl_minorversion) {
5036 put_ol_stateid_locked(s, &reaplist);
5037 spin_unlock(&clp->cl_lock);
5038 free_ol_stateid_reaplist(&reaplist);
5040 spin_unlock(&clp->cl_lock);
5041 free_ol_stateid_reaplist(&reaplist);
5043 move_to_close_lru(s, clp->net);
5048 * nfs4_unlock_state() called after encode
5051 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5052 struct nfsd4_close *close)
5055 struct nfs4_ol_stateid *stp;
5056 struct net *net = SVC_NET(rqstp);
5057 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5059 dprintk("NFSD: nfsd4_close on file %pd\n",
5060 cstate->current_fh.fh_dentry);
5062 status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
5064 NFS4_OPEN_STID|NFS4_CLOSED_STID,
5066 nfsd4_bump_seqid(cstate, status);
5069 nfs4_inc_and_copy_stateid(&close->cl_stateid, &stp->st_stid);
5070 up_write(&stp->st_rwsem);
5072 nfsd4_close_open_stateid(stp);
5074 /* put reference from nfs4_preprocess_seqid_op */
5075 nfs4_put_stid(&stp->st_stid);
5081 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5082 struct nfsd4_delegreturn *dr)
5084 struct nfs4_delegation *dp;
5085 stateid_t *stateid = &dr->dr_stateid;
5086 struct nfs4_stid *s;
5088 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
5090 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
5093 status = nfsd4_lookup_stateid(cstate, stateid, NFS4_DELEG_STID, &s, nn);
5096 dp = delegstateid(s);
5097 status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
5101 destroy_delegation(dp);
5103 nfs4_put_stid(&dp->dl_stid);
5109 end_offset(u64 start, u64 len)
5114 return end >= start ? end: NFS4_MAX_UINT64;
5117 /* last octet in a range */
5119 last_byte_offset(u64 start, u64 len)
5125 return end > start ? end - 1: NFS4_MAX_UINT64;
5129 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
5130 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
5131 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
5132 * locking, this prevents us from being completely protocol-compliant. The
5133 * real solution to this problem is to start using unsigned file offsets in
5134 * the VFS, but this is a very deep change!
5137 nfs4_transform_lock_offset(struct file_lock *lock)
5139 if (lock->fl_start < 0)
5140 lock->fl_start = OFFSET_MAX;
5141 if (lock->fl_end < 0)
5142 lock->fl_end = OFFSET_MAX;
5146 nfsd4_fl_get_owner(fl_owner_t owner)
5148 struct nfs4_lockowner *lo = (struct nfs4_lockowner *)owner;
5150 nfs4_get_stateowner(&lo->lo_owner);
5155 nfsd4_fl_put_owner(fl_owner_t owner)
5157 struct nfs4_lockowner *lo = (struct nfs4_lockowner *)owner;
5160 nfs4_put_stateowner(&lo->lo_owner);
5163 static const struct lock_manager_operations nfsd_posix_mng_ops = {
5164 .lm_get_owner = nfsd4_fl_get_owner,
5165 .lm_put_owner = nfsd4_fl_put_owner,
5169 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
5171 struct nfs4_lockowner *lo;
5173 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
5174 lo = (struct nfs4_lockowner *) fl->fl_owner;
5175 deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
5176 lo->lo_owner.so_owner.len, GFP_KERNEL);
5177 if (!deny->ld_owner.data)
5178 /* We just don't care that much */
5180 deny->ld_owner.len = lo->lo_owner.so_owner.len;
5181 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
5184 deny->ld_owner.len = 0;
5185 deny->ld_owner.data = NULL;
5186 deny->ld_clientid.cl_boot = 0;
5187 deny->ld_clientid.cl_id = 0;
5189 deny->ld_start = fl->fl_start;
5190 deny->ld_length = NFS4_MAX_UINT64;
5191 if (fl->fl_end != NFS4_MAX_UINT64)
5192 deny->ld_length = fl->fl_end - fl->fl_start + 1;
5193 deny->ld_type = NFS4_READ_LT;
5194 if (fl->fl_type != F_RDLCK)
5195 deny->ld_type = NFS4_WRITE_LT;
5198 static struct nfs4_lockowner *
5199 find_lockowner_str_locked(struct nfs4_client *clp, struct xdr_netobj *owner)
5201 unsigned int strhashval = ownerstr_hashval(owner);
5202 struct nfs4_stateowner *so;
5204 lockdep_assert_held(&clp->cl_lock);
5206 list_for_each_entry(so, &clp->cl_ownerstr_hashtbl[strhashval],
5208 if (so->so_is_open_owner)
5210 if (same_owner_str(so, owner))
5211 return lockowner(nfs4_get_stateowner(so));
5216 static struct nfs4_lockowner *
5217 find_lockowner_str(struct nfs4_client *clp, struct xdr_netobj *owner)
5219 struct nfs4_lockowner *lo;
5221 spin_lock(&clp->cl_lock);
5222 lo = find_lockowner_str_locked(clp, owner);
5223 spin_unlock(&clp->cl_lock);
5227 static void nfs4_unhash_lockowner(struct nfs4_stateowner *sop)
5229 unhash_lockowner_locked(lockowner(sop));
5232 static void nfs4_free_lockowner(struct nfs4_stateowner *sop)
5234 struct nfs4_lockowner *lo = lockowner(sop);
5236 kmem_cache_free(lockowner_slab, lo);
5239 static const struct nfs4_stateowner_operations lockowner_ops = {
5240 .so_unhash = nfs4_unhash_lockowner,
5241 .so_free = nfs4_free_lockowner,
5245 * Alloc a lock owner structure.
5246 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
5249 * strhashval = ownerstr_hashval
5251 static struct nfs4_lockowner *
5252 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp,
5253 struct nfs4_ol_stateid *open_stp,
5254 struct nfsd4_lock *lock)
5256 struct nfs4_lockowner *lo, *ret;
5258 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
5261 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
5262 lo->lo_owner.so_is_open_owner = 0;
5263 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid;
5264 lo->lo_owner.so_ops = &lockowner_ops;
5265 spin_lock(&clp->cl_lock);
5266 ret = find_lockowner_str_locked(clp, &lock->lk_new_owner);
5268 list_add(&lo->lo_owner.so_strhash,
5269 &clp->cl_ownerstr_hashtbl[strhashval]);
5272 nfs4_free_stateowner(&lo->lo_owner);
5274 spin_unlock(&clp->cl_lock);
5279 init_lock_stateid(struct nfs4_ol_stateid *stp, struct nfs4_lockowner *lo,
5280 struct nfs4_file *fp, struct inode *inode,
5281 struct nfs4_ol_stateid *open_stp)
5283 struct nfs4_client *clp = lo->lo_owner.so_client;
5285 lockdep_assert_held(&clp->cl_lock);
5287 atomic_inc(&stp->st_stid.sc_count);
5288 stp->st_stid.sc_type = NFS4_LOCK_STID;
5289 stp->st_stateowner = nfs4_get_stateowner(&lo->lo_owner);
5291 stp->st_stid.sc_file = fp;
5292 stp->st_stid.sc_free = nfs4_free_lock_stateid;
5293 stp->st_access_bmap = 0;
5294 stp->st_deny_bmap = open_stp->st_deny_bmap;
5295 stp->st_openstp = open_stp;
5296 init_rwsem(&stp->st_rwsem);
5297 list_add(&stp->st_locks, &open_stp->st_locks);
5298 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
5299 spin_lock(&fp->fi_lock);
5300 list_add(&stp->st_perfile, &fp->fi_stateids);
5301 spin_unlock(&fp->fi_lock);
5304 static struct nfs4_ol_stateid *
5305 find_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp)
5307 struct nfs4_ol_stateid *lst;
5308 struct nfs4_client *clp = lo->lo_owner.so_client;
5310 lockdep_assert_held(&clp->cl_lock);
5312 list_for_each_entry(lst, &lo->lo_owner.so_stateids, st_perstateowner) {
5313 if (lst->st_stid.sc_file == fp) {
5314 atomic_inc(&lst->st_stid.sc_count);
5321 static struct nfs4_ol_stateid *
5322 find_or_create_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fi,
5323 struct inode *inode, struct nfs4_ol_stateid *ost,
5326 struct nfs4_stid *ns = NULL;
5327 struct nfs4_ol_stateid *lst;
5328 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
5329 struct nfs4_client *clp = oo->oo_owner.so_client;
5331 spin_lock(&clp->cl_lock);
5332 lst = find_lock_stateid(lo, fi);
5334 spin_unlock(&clp->cl_lock);
5335 ns = nfs4_alloc_stid(clp, stateid_slab);
5339 spin_lock(&clp->cl_lock);
5340 lst = find_lock_stateid(lo, fi);
5342 lst = openlockstateid(ns);
5343 init_lock_stateid(lst, lo, fi, inode, ost);
5348 spin_unlock(&clp->cl_lock);
5355 check_lock_length(u64 offset, u64 length)
5357 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
5358 (length > ~offset)));
5361 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
5363 struct nfs4_file *fp = lock_stp->st_stid.sc_file;
5365 lockdep_assert_held(&fp->fi_lock);
5367 if (test_access(access, lock_stp))
5369 __nfs4_file_get_access(fp, access);
5370 set_access(access, lock_stp);
5374 lookup_or_create_lock_state(struct nfsd4_compound_state *cstate,
5375 struct nfs4_ol_stateid *ost,
5376 struct nfsd4_lock *lock,
5377 struct nfs4_ol_stateid **lst, bool *new)
5380 struct nfs4_file *fi = ost->st_stid.sc_file;
5381 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
5382 struct nfs4_client *cl = oo->oo_owner.so_client;
5383 struct inode *inode = d_inode(cstate->current_fh.fh_dentry);
5384 struct nfs4_lockowner *lo;
5385 unsigned int strhashval;
5387 lo = find_lockowner_str(cl, &lock->lk_new_owner);
5389 strhashval = ownerstr_hashval(&lock->lk_new_owner);
5390 lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
5392 return nfserr_jukebox;
5394 /* with an existing lockowner, seqids must be the same */
5395 status = nfserr_bad_seqid;
5396 if (!cstate->minorversion &&
5397 lock->lk_new_lock_seqid != lo->lo_owner.so_seqid)
5401 *lst = find_or_create_lock_stateid(lo, fi, inode, ost, new);
5403 status = nfserr_jukebox;
5408 nfs4_put_stateowner(&lo->lo_owner);
5416 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5417 struct nfsd4_lock *lock)
5419 struct nfs4_openowner *open_sop = NULL;
5420 struct nfs4_lockowner *lock_sop = NULL;
5421 struct nfs4_ol_stateid *lock_stp = NULL;
5422 struct nfs4_ol_stateid *open_stp = NULL;
5423 struct nfs4_file *fp;
5424 struct file *filp = NULL;
5425 struct file_lock *file_lock = NULL;
5426 struct file_lock *conflock = NULL;
5431 struct net *net = SVC_NET(rqstp);
5432 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5434 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
5435 (long long) lock->lk_offset,
5436 (long long) lock->lk_length);
5438 if (check_lock_length(lock->lk_offset, lock->lk_length))
5439 return nfserr_inval;
5441 if ((status = fh_verify(rqstp, &cstate->current_fh,
5442 S_IFREG, NFSD_MAY_LOCK))) {
5443 dprintk("NFSD: nfsd4_lock: permission denied!\n");
5447 if (lock->lk_is_new) {
5448 if (nfsd4_has_session(cstate))
5449 /* See rfc 5661 18.10.3: given clientid is ignored: */
5450 memcpy(&lock->lk_new_clientid,
5451 &cstate->session->se_client->cl_clientid,
5452 sizeof(clientid_t));
5454 status = nfserr_stale_clientid;
5455 if (STALE_CLIENTID(&lock->lk_new_clientid, nn))
5458 /* validate and update open stateid and open seqid */
5459 status = nfs4_preprocess_confirmed_seqid_op(cstate,
5460 lock->lk_new_open_seqid,
5461 &lock->lk_new_open_stateid,
5465 up_write(&open_stp->st_rwsem);
5466 open_sop = openowner(open_stp->st_stateowner);
5467 status = nfserr_bad_stateid;
5468 if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
5469 &lock->lk_new_clientid))
5471 status = lookup_or_create_lock_state(cstate, open_stp, lock,
5473 if (status == nfs_ok)
5474 down_write(&lock_stp->st_rwsem);
5476 status = nfs4_preprocess_seqid_op(cstate,
5477 lock->lk_old_lock_seqid,
5478 &lock->lk_old_lock_stateid,
5479 NFS4_LOCK_STID, &lock_stp, nn);
5483 lock_sop = lockowner(lock_stp->st_stateowner);
5485 lkflg = setlkflg(lock->lk_type);
5486 status = nfs4_check_openmode(lock_stp, lkflg);
5490 status = nfserr_grace;
5491 if (locks_in_grace(net) && !lock->lk_reclaim)
5493 status = nfserr_no_grace;
5494 if (!locks_in_grace(net) && lock->lk_reclaim)
5497 file_lock = locks_alloc_lock();
5499 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
5500 status = nfserr_jukebox;
5504 fp = lock_stp->st_stid.sc_file;
5505 switch (lock->lk_type) {
5508 spin_lock(&fp->fi_lock);
5509 filp = find_readable_file_locked(fp);
5511 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
5512 spin_unlock(&fp->fi_lock);
5513 file_lock->fl_type = F_RDLCK;
5516 case NFS4_WRITEW_LT:
5517 spin_lock(&fp->fi_lock);
5518 filp = find_writeable_file_locked(fp);
5520 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
5521 spin_unlock(&fp->fi_lock);
5522 file_lock->fl_type = F_WRLCK;
5525 status = nfserr_inval;
5529 status = nfserr_openmode;
5533 file_lock->fl_owner = (fl_owner_t)lockowner(nfs4_get_stateowner(&lock_sop->lo_owner));
5534 file_lock->fl_pid = current->tgid;
5535 file_lock->fl_file = filp;
5536 file_lock->fl_flags = FL_POSIX;
5537 file_lock->fl_lmops = &nfsd_posix_mng_ops;
5538 file_lock->fl_start = lock->lk_offset;
5539 file_lock->fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
5540 nfs4_transform_lock_offset(file_lock);
5542 conflock = locks_alloc_lock();
5544 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
5545 status = nfserr_jukebox;
5549 err = vfs_lock_file(filp, F_SETLK, file_lock, conflock);
5551 case 0: /* success! */
5552 nfs4_inc_and_copy_stateid(&lock->lk_resp_stateid, &lock_stp->st_stid);
5555 case (EAGAIN): /* conflock holds conflicting lock */
5556 status = nfserr_denied;
5557 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
5558 nfs4_set_lock_denied(conflock, &lock->lk_denied);
5561 status = nfserr_deadlock;
5564 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
5565 status = nfserrno(err);
5572 /* Bump seqid manually if the 4.0 replay owner is openowner */
5573 if (cstate->replay_owner &&
5574 cstate->replay_owner != &lock_sop->lo_owner &&
5575 seqid_mutating_err(ntohl(status)))
5576 lock_sop->lo_owner.so_seqid++;
5578 up_write(&lock_stp->st_rwsem);
5581 * If this is a new, never-before-used stateid, and we are
5582 * returning an error, then just go ahead and release it.
5585 release_lock_stateid(lock_stp);
5587 nfs4_put_stid(&lock_stp->st_stid);
5590 nfs4_put_stid(&open_stp->st_stid);
5591 nfsd4_bump_seqid(cstate, status);
5593 locks_free_lock(file_lock);
5595 locks_free_lock(conflock);
5600 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
5601 * so we do a temporary open here just to get an open file to pass to
5602 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
5605 static __be32 nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
5608 __be32 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
5610 err = nfserrno(vfs_test_lock(file, lock));
5620 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5621 struct nfsd4_lockt *lockt)
5623 struct file_lock *file_lock = NULL;
5624 struct nfs4_lockowner *lo = NULL;
5626 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
5628 if (locks_in_grace(SVC_NET(rqstp)))
5629 return nfserr_grace;
5631 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
5632 return nfserr_inval;
5634 if (!nfsd4_has_session(cstate)) {
5635 status = lookup_clientid(&lockt->lt_clientid, cstate, nn);
5640 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
5643 file_lock = locks_alloc_lock();
5645 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
5646 status = nfserr_jukebox;
5650 switch (lockt->lt_type) {
5653 file_lock->fl_type = F_RDLCK;
5656 case NFS4_WRITEW_LT:
5657 file_lock->fl_type = F_WRLCK;
5660 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
5661 status = nfserr_inval;
5665 lo = find_lockowner_str(cstate->clp, &lockt->lt_owner);
5667 file_lock->fl_owner = (fl_owner_t)lo;
5668 file_lock->fl_pid = current->tgid;
5669 file_lock->fl_flags = FL_POSIX;
5671 file_lock->fl_start = lockt->lt_offset;
5672 file_lock->fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
5674 nfs4_transform_lock_offset(file_lock);
5676 status = nfsd_test_lock(rqstp, &cstate->current_fh, file_lock);
5680 if (file_lock->fl_type != F_UNLCK) {
5681 status = nfserr_denied;
5682 nfs4_set_lock_denied(file_lock, &lockt->lt_denied);
5686 nfs4_put_stateowner(&lo->lo_owner);
5688 locks_free_lock(file_lock);
5693 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5694 struct nfsd4_locku *locku)
5696 struct nfs4_ol_stateid *stp;
5697 struct file *filp = NULL;
5698 struct file_lock *file_lock = NULL;
5701 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
5703 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
5704 (long long) locku->lu_offset,
5705 (long long) locku->lu_length);
5707 if (check_lock_length(locku->lu_offset, locku->lu_length))
5708 return nfserr_inval;
5710 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
5711 &locku->lu_stateid, NFS4_LOCK_STID,
5715 filp = find_any_file(stp->st_stid.sc_file);
5717 status = nfserr_lock_range;
5720 file_lock = locks_alloc_lock();
5722 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
5723 status = nfserr_jukebox;
5727 file_lock->fl_type = F_UNLCK;
5728 file_lock->fl_owner = (fl_owner_t)lockowner(nfs4_get_stateowner(stp->st_stateowner));
5729 file_lock->fl_pid = current->tgid;
5730 file_lock->fl_file = filp;
5731 file_lock->fl_flags = FL_POSIX;
5732 file_lock->fl_lmops = &nfsd_posix_mng_ops;
5733 file_lock->fl_start = locku->lu_offset;
5735 file_lock->fl_end = last_byte_offset(locku->lu_offset,
5737 nfs4_transform_lock_offset(file_lock);
5739 err = vfs_lock_file(filp, F_SETLK, file_lock, NULL);
5741 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
5744 nfs4_inc_and_copy_stateid(&locku->lu_stateid, &stp->st_stid);
5748 up_write(&stp->st_rwsem);
5749 nfs4_put_stid(&stp->st_stid);
5751 nfsd4_bump_seqid(cstate, status);
5753 locks_free_lock(file_lock);
5757 status = nfserrno(err);
5763 * true: locks held by lockowner
5764 * false: no locks held by lockowner
5767 check_for_locks(struct nfs4_file *fp, struct nfs4_lockowner *lowner)
5769 struct file_lock *fl;
5771 struct file *filp = find_any_file(fp);
5772 struct inode *inode;
5773 struct file_lock_context *flctx;
5776 /* Any valid lock stateid should have some sort of access */
5781 inode = file_inode(filp);
5782 flctx = inode->i_flctx;
5784 if (flctx && !list_empty_careful(&flctx->flc_posix)) {
5785 spin_lock(&flctx->flc_lock);
5786 list_for_each_entry(fl, &flctx->flc_posix, fl_list) {
5787 if (fl->fl_owner == (fl_owner_t)lowner) {
5792 spin_unlock(&flctx->flc_lock);
5799 nfsd4_release_lockowner(struct svc_rqst *rqstp,
5800 struct nfsd4_compound_state *cstate,
5801 struct nfsd4_release_lockowner *rlockowner)
5803 clientid_t *clid = &rlockowner->rl_clientid;
5804 struct nfs4_stateowner *sop;
5805 struct nfs4_lockowner *lo = NULL;
5806 struct nfs4_ol_stateid *stp;
5807 struct xdr_netobj *owner = &rlockowner->rl_owner;
5808 unsigned int hashval = ownerstr_hashval(owner);
5810 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
5811 struct nfs4_client *clp;
5813 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
5814 clid->cl_boot, clid->cl_id);
5816 status = lookup_clientid(clid, cstate, nn);
5821 /* Find the matching lock stateowner */
5822 spin_lock(&clp->cl_lock);
5823 list_for_each_entry(sop, &clp->cl_ownerstr_hashtbl[hashval],
5826 if (sop->so_is_open_owner || !same_owner_str(sop, owner))
5829 /* see if there are still any locks associated with it */
5830 lo = lockowner(sop);
5831 list_for_each_entry(stp, &sop->so_stateids, st_perstateowner) {
5832 if (check_for_locks(stp->st_stid.sc_file, lo)) {
5833 status = nfserr_locks_held;
5834 spin_unlock(&clp->cl_lock);
5839 nfs4_get_stateowner(sop);
5842 spin_unlock(&clp->cl_lock);
5844 release_lockowner(lo);
5848 static inline struct nfs4_client_reclaim *
5851 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
5855 nfs4_has_reclaimed_state(const char *name, struct nfsd_net *nn)
5857 struct nfs4_client_reclaim *crp;
5859 crp = nfsd4_find_reclaim_client(name, nn);
5860 return (crp && crp->cr_clp);
5864 * failure => all reset bets are off, nfserr_no_grace...
5866 struct nfs4_client_reclaim *
5867 nfs4_client_to_reclaim(const char *name, struct nfsd_net *nn)
5869 unsigned int strhashval;
5870 struct nfs4_client_reclaim *crp;
5872 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
5873 crp = alloc_reclaim();
5875 strhashval = clientstr_hashval(name);
5876 INIT_LIST_HEAD(&crp->cr_strhash);
5877 list_add(&crp->cr_strhash, &nn->reclaim_str_hashtbl[strhashval]);
5878 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
5880 nn->reclaim_str_hashtbl_size++;
5886 nfs4_remove_reclaim_record(struct nfs4_client_reclaim *crp, struct nfsd_net *nn)
5888 list_del(&crp->cr_strhash);
5890 nn->reclaim_str_hashtbl_size--;
5894 nfs4_release_reclaim(struct nfsd_net *nn)
5896 struct nfs4_client_reclaim *crp = NULL;
5899 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
5900 while (!list_empty(&nn->reclaim_str_hashtbl[i])) {
5901 crp = list_entry(nn->reclaim_str_hashtbl[i].next,
5902 struct nfs4_client_reclaim, cr_strhash);
5903 nfs4_remove_reclaim_record(crp, nn);
5906 WARN_ON_ONCE(nn->reclaim_str_hashtbl_size);
5910 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
5911 struct nfs4_client_reclaim *
5912 nfsd4_find_reclaim_client(const char *recdir, struct nfsd_net *nn)
5914 unsigned int strhashval;
5915 struct nfs4_client_reclaim *crp = NULL;
5917 dprintk("NFSD: nfs4_find_reclaim_client for recdir %s\n", recdir);
5919 strhashval = clientstr_hashval(recdir);
5920 list_for_each_entry(crp, &nn->reclaim_str_hashtbl[strhashval], cr_strhash) {
5921 if (same_name(crp->cr_recdir, recdir)) {
5929 * Called from OPEN. Look for clientid in reclaim list.
5932 nfs4_check_open_reclaim(clientid_t *clid,
5933 struct nfsd4_compound_state *cstate,
5934 struct nfsd_net *nn)
5938 /* find clientid in conf_id_hashtbl */
5939 status = lookup_clientid(clid, cstate, nn);
5941 return nfserr_reclaim_bad;
5943 if (test_bit(NFSD4_CLIENT_RECLAIM_COMPLETE, &cstate->clp->cl_flags))
5944 return nfserr_no_grace;
5946 if (nfsd4_client_record_check(cstate->clp))
5947 return nfserr_reclaim_bad;
5952 #ifdef CONFIG_NFSD_FAULT_INJECTION
5954 put_client(struct nfs4_client *clp)
5956 atomic_dec(&clp->cl_refcount);
5959 static struct nfs4_client *
5960 nfsd_find_client(struct sockaddr_storage *addr, size_t addr_size)
5962 struct nfs4_client *clp;
5963 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
5966 if (!nfsd_netns_ready(nn))
5969 list_for_each_entry(clp, &nn->client_lru, cl_lru) {
5970 if (memcmp(&clp->cl_addr, addr, addr_size) == 0)
5977 nfsd_inject_print_clients(void)
5979 struct nfs4_client *clp;
5981 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
5983 char buf[INET6_ADDRSTRLEN];
5985 if (!nfsd_netns_ready(nn))
5988 spin_lock(&nn->client_lock);
5989 list_for_each_entry(clp, &nn->client_lru, cl_lru) {
5990 rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, sizeof(buf));
5991 pr_info("NFS Client: %s\n", buf);
5994 spin_unlock(&nn->client_lock);
6000 nfsd_inject_forget_client(struct sockaddr_storage *addr, size_t addr_size)
6003 struct nfs4_client *clp;
6004 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6007 if (!nfsd_netns_ready(nn))
6010 spin_lock(&nn->client_lock);
6011 clp = nfsd_find_client(addr, addr_size);
6013 if (mark_client_expired_locked(clp) == nfs_ok)
6018 spin_unlock(&nn->client_lock);
6027 nfsd_inject_forget_clients(u64 max)
6030 struct nfs4_client *clp, *next;
6031 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6033 LIST_HEAD(reaplist);
6035 if (!nfsd_netns_ready(nn))
6038 spin_lock(&nn->client_lock);
6039 list_for_each_entry_safe(clp, next, &nn->client_lru, cl_lru) {
6040 if (mark_client_expired_locked(clp) == nfs_ok) {
6041 list_add(&clp->cl_lru, &reaplist);
6042 if (max != 0 && ++count >= max)
6046 spin_unlock(&nn->client_lock);
6048 list_for_each_entry_safe(clp, next, &reaplist, cl_lru)
6054 static void nfsd_print_count(struct nfs4_client *clp, unsigned int count,
6057 char buf[INET6_ADDRSTRLEN];
6058 rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, sizeof(buf));
6059 printk(KERN_INFO "NFS Client: %s has %u %s\n", buf, count, type);
6063 nfsd_inject_add_lock_to_list(struct nfs4_ol_stateid *lst,
6064 struct list_head *collect)
6066 struct nfs4_client *clp = lst->st_stid.sc_client;
6067 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6073 lockdep_assert_held(&nn->client_lock);
6074 atomic_inc(&clp->cl_refcount);
6075 list_add(&lst->st_locks, collect);
6078 static u64 nfsd_foreach_client_lock(struct nfs4_client *clp, u64 max,
6079 struct list_head *collect,
6080 bool (*func)(struct nfs4_ol_stateid *))
6082 struct nfs4_openowner *oop;
6083 struct nfs4_ol_stateid *stp, *st_next;
6084 struct nfs4_ol_stateid *lst, *lst_next;
6087 spin_lock(&clp->cl_lock);
6088 list_for_each_entry(oop, &clp->cl_openowners, oo_perclient) {
6089 list_for_each_entry_safe(stp, st_next,
6090 &oop->oo_owner.so_stateids, st_perstateowner) {
6091 list_for_each_entry_safe(lst, lst_next,
6092 &stp->st_locks, st_locks) {
6095 nfsd_inject_add_lock_to_list(lst,
6100 * Despite the fact that these functions deal
6101 * with 64-bit integers for "count", we must
6102 * ensure that it doesn't blow up the
6103 * clp->cl_refcount. Throw a warning if we
6104 * start to approach INT_MAX here.
6106 WARN_ON_ONCE(count == (INT_MAX / 2));
6113 spin_unlock(&clp->cl_lock);
6119 nfsd_collect_client_locks(struct nfs4_client *clp, struct list_head *collect,
6122 return nfsd_foreach_client_lock(clp, max, collect, unhash_lock_stateid);
6126 nfsd_print_client_locks(struct nfs4_client *clp)
6128 u64 count = nfsd_foreach_client_lock(clp, 0, NULL, NULL);
6129 nfsd_print_count(clp, count, "locked files");
6134 nfsd_inject_print_locks(void)
6136 struct nfs4_client *clp;
6138 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6141 if (!nfsd_netns_ready(nn))
6144 spin_lock(&nn->client_lock);
6145 list_for_each_entry(clp, &nn->client_lru, cl_lru)
6146 count += nfsd_print_client_locks(clp);
6147 spin_unlock(&nn->client_lock);
6153 nfsd_reap_locks(struct list_head *reaplist)
6155 struct nfs4_client *clp;
6156 struct nfs4_ol_stateid *stp, *next;
6158 list_for_each_entry_safe(stp, next, reaplist, st_locks) {
6159 list_del_init(&stp->st_locks);
6160 clp = stp->st_stid.sc_client;
6161 nfs4_put_stid(&stp->st_stid);
6167 nfsd_inject_forget_client_locks(struct sockaddr_storage *addr, size_t addr_size)
6169 unsigned int count = 0;
6170 struct nfs4_client *clp;
6171 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6173 LIST_HEAD(reaplist);
6175 if (!nfsd_netns_ready(nn))
6178 spin_lock(&nn->client_lock);
6179 clp = nfsd_find_client(addr, addr_size);
6181 count = nfsd_collect_client_locks(clp, &reaplist, 0);
6182 spin_unlock(&nn->client_lock);
6183 nfsd_reap_locks(&reaplist);
6188 nfsd_inject_forget_locks(u64 max)
6191 struct nfs4_client *clp;
6192 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6194 LIST_HEAD(reaplist);
6196 if (!nfsd_netns_ready(nn))
6199 spin_lock(&nn->client_lock);
6200 list_for_each_entry(clp, &nn->client_lru, cl_lru) {
6201 count += nfsd_collect_client_locks(clp, &reaplist, max - count);
6202 if (max != 0 && count >= max)
6205 spin_unlock(&nn->client_lock);
6206 nfsd_reap_locks(&reaplist);
6211 nfsd_foreach_client_openowner(struct nfs4_client *clp, u64 max,
6212 struct list_head *collect,
6213 void (*func)(struct nfs4_openowner *))
6215 struct nfs4_openowner *oop, *next;
6216 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6220 lockdep_assert_held(&nn->client_lock);
6222 spin_lock(&clp->cl_lock);
6223 list_for_each_entry_safe(oop, next, &clp->cl_openowners, oo_perclient) {
6227 atomic_inc(&clp->cl_refcount);
6228 list_add(&oop->oo_perclient, collect);
6233 * Despite the fact that these functions deal with
6234 * 64-bit integers for "count", we must ensure that
6235 * it doesn't blow up the clp->cl_refcount. Throw a
6236 * warning if we start to approach INT_MAX here.
6238 WARN_ON_ONCE(count == (INT_MAX / 2));
6242 spin_unlock(&clp->cl_lock);
6248 nfsd_print_client_openowners(struct nfs4_client *clp)
6250 u64 count = nfsd_foreach_client_openowner(clp, 0, NULL, NULL);
6252 nfsd_print_count(clp, count, "openowners");
6257 nfsd_collect_client_openowners(struct nfs4_client *clp,
6258 struct list_head *collect, u64 max)
6260 return nfsd_foreach_client_openowner(clp, max, collect,
6261 unhash_openowner_locked);
6265 nfsd_inject_print_openowners(void)
6267 struct nfs4_client *clp;
6269 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6272 if (!nfsd_netns_ready(nn))
6275 spin_lock(&nn->client_lock);
6276 list_for_each_entry(clp, &nn->client_lru, cl_lru)
6277 count += nfsd_print_client_openowners(clp);
6278 spin_unlock(&nn->client_lock);
6284 nfsd_reap_openowners(struct list_head *reaplist)
6286 struct nfs4_client *clp;
6287 struct nfs4_openowner *oop, *next;
6289 list_for_each_entry_safe(oop, next, reaplist, oo_perclient) {
6290 list_del_init(&oop->oo_perclient);
6291 clp = oop->oo_owner.so_client;
6292 release_openowner(oop);
6298 nfsd_inject_forget_client_openowners(struct sockaddr_storage *addr,
6301 unsigned int count = 0;
6302 struct nfs4_client *clp;
6303 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6305 LIST_HEAD(reaplist);
6307 if (!nfsd_netns_ready(nn))
6310 spin_lock(&nn->client_lock);
6311 clp = nfsd_find_client(addr, addr_size);
6313 count = nfsd_collect_client_openowners(clp, &reaplist, 0);
6314 spin_unlock(&nn->client_lock);
6315 nfsd_reap_openowners(&reaplist);
6320 nfsd_inject_forget_openowners(u64 max)
6323 struct nfs4_client *clp;
6324 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6326 LIST_HEAD(reaplist);
6328 if (!nfsd_netns_ready(nn))
6331 spin_lock(&nn->client_lock);
6332 list_for_each_entry(clp, &nn->client_lru, cl_lru) {
6333 count += nfsd_collect_client_openowners(clp, &reaplist,
6335 if (max != 0 && count >= max)
6338 spin_unlock(&nn->client_lock);
6339 nfsd_reap_openowners(&reaplist);
6343 static u64 nfsd_find_all_delegations(struct nfs4_client *clp, u64 max,
6344 struct list_head *victims)
6346 struct nfs4_delegation *dp, *next;
6347 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6351 lockdep_assert_held(&nn->client_lock);
6353 spin_lock(&state_lock);
6354 list_for_each_entry_safe(dp, next, &clp->cl_delegations, dl_perclnt) {
6357 * It's not safe to mess with delegations that have a
6358 * non-zero dl_time. They might have already been broken
6359 * and could be processed by the laundromat outside of
6360 * the state_lock. Just leave them be.
6362 if (dp->dl_time != 0)
6365 atomic_inc(&clp->cl_refcount);
6366 WARN_ON(!unhash_delegation_locked(dp));
6367 list_add(&dp->dl_recall_lru, victims);
6371 * Despite the fact that these functions deal with
6372 * 64-bit integers for "count", we must ensure that
6373 * it doesn't blow up the clp->cl_refcount. Throw a
6374 * warning if we start to approach INT_MAX here.
6376 WARN_ON_ONCE(count == (INT_MAX / 2));
6380 spin_unlock(&state_lock);
6385 nfsd_print_client_delegations(struct nfs4_client *clp)
6387 u64 count = nfsd_find_all_delegations(clp, 0, NULL);
6389 nfsd_print_count(clp, count, "delegations");
6394 nfsd_inject_print_delegations(void)
6396 struct nfs4_client *clp;
6398 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6401 if (!nfsd_netns_ready(nn))
6404 spin_lock(&nn->client_lock);
6405 list_for_each_entry(clp, &nn->client_lru, cl_lru)
6406 count += nfsd_print_client_delegations(clp);
6407 spin_unlock(&nn->client_lock);
6413 nfsd_forget_delegations(struct list_head *reaplist)
6415 struct nfs4_client *clp;
6416 struct nfs4_delegation *dp, *next;
6418 list_for_each_entry_safe(dp, next, reaplist, dl_recall_lru) {
6419 list_del_init(&dp->dl_recall_lru);
6420 clp = dp->dl_stid.sc_client;
6421 revoke_delegation(dp);
6427 nfsd_inject_forget_client_delegations(struct sockaddr_storage *addr,
6431 struct nfs4_client *clp;
6432 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6434 LIST_HEAD(reaplist);
6436 if (!nfsd_netns_ready(nn))
6439 spin_lock(&nn->client_lock);
6440 clp = nfsd_find_client(addr, addr_size);
6442 count = nfsd_find_all_delegations(clp, 0, &reaplist);
6443 spin_unlock(&nn->client_lock);
6445 nfsd_forget_delegations(&reaplist);
6450 nfsd_inject_forget_delegations(u64 max)
6453 struct nfs4_client *clp;
6454 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6456 LIST_HEAD(reaplist);
6458 if (!nfsd_netns_ready(nn))
6461 spin_lock(&nn->client_lock);
6462 list_for_each_entry(clp, &nn->client_lru, cl_lru) {
6463 count += nfsd_find_all_delegations(clp, max - count, &reaplist);
6464 if (max != 0 && count >= max)
6467 spin_unlock(&nn->client_lock);
6468 nfsd_forget_delegations(&reaplist);
6473 nfsd_recall_delegations(struct list_head *reaplist)
6475 struct nfs4_client *clp;
6476 struct nfs4_delegation *dp, *next;
6478 list_for_each_entry_safe(dp, next, reaplist, dl_recall_lru) {
6479 list_del_init(&dp->dl_recall_lru);
6480 clp = dp->dl_stid.sc_client;
6482 * We skipped all entries that had a zero dl_time before,
6483 * so we can now reset the dl_time back to 0. If a delegation
6484 * break comes in now, then it won't make any difference since
6485 * we're recalling it either way.
6487 spin_lock(&state_lock);
6489 spin_unlock(&state_lock);
6490 nfsd_break_one_deleg(dp);
6496 nfsd_inject_recall_client_delegations(struct sockaddr_storage *addr,
6500 struct nfs4_client *clp;
6501 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6503 LIST_HEAD(reaplist);
6505 if (!nfsd_netns_ready(nn))
6508 spin_lock(&nn->client_lock);
6509 clp = nfsd_find_client(addr, addr_size);
6511 count = nfsd_find_all_delegations(clp, 0, &reaplist);
6512 spin_unlock(&nn->client_lock);
6514 nfsd_recall_delegations(&reaplist);
6519 nfsd_inject_recall_delegations(u64 max)
6522 struct nfs4_client *clp, *next;
6523 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6525 LIST_HEAD(reaplist);
6527 if (!nfsd_netns_ready(nn))
6530 spin_lock(&nn->client_lock);
6531 list_for_each_entry_safe(clp, next, &nn->client_lru, cl_lru) {
6532 count += nfsd_find_all_delegations(clp, max - count, &reaplist);
6533 if (max != 0 && ++count >= max)
6536 spin_unlock(&nn->client_lock);
6537 nfsd_recall_delegations(&reaplist);
6540 #endif /* CONFIG_NFSD_FAULT_INJECTION */
6543 * Since the lifetime of a delegation isn't limited to that of an open, a
6544 * client may quite reasonably hang on to a delegation as long as it has
6545 * the inode cached. This becomes an obvious problem the first time a
6546 * client's inode cache approaches the size of the server's total memory.
6548 * For now we avoid this problem by imposing a hard limit on the number
6549 * of delegations, which varies according to the server's memory size.
6552 set_max_delegations(void)
6555 * Allow at most 4 delegations per megabyte of RAM. Quick
6556 * estimates suggest that in the worst case (where every delegation
6557 * is for a different inode), a delegation could take about 1.5K,
6558 * giving a worst case usage of about 6% of memory.
6560 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
6563 static int nfs4_state_create_net(struct net *net)
6565 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
6568 nn->conf_id_hashtbl = kmalloc(sizeof(struct list_head) *
6569 CLIENT_HASH_SIZE, GFP_KERNEL);
6570 if (!nn->conf_id_hashtbl)
6572 nn->unconf_id_hashtbl = kmalloc(sizeof(struct list_head) *
6573 CLIENT_HASH_SIZE, GFP_KERNEL);
6574 if (!nn->unconf_id_hashtbl)
6576 nn->sessionid_hashtbl = kmalloc(sizeof(struct list_head) *
6577 SESSION_HASH_SIZE, GFP_KERNEL);
6578 if (!nn->sessionid_hashtbl)
6581 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
6582 INIT_LIST_HEAD(&nn->conf_id_hashtbl[i]);
6583 INIT_LIST_HEAD(&nn->unconf_id_hashtbl[i]);
6585 for (i = 0; i < SESSION_HASH_SIZE; i++)
6586 INIT_LIST_HEAD(&nn->sessionid_hashtbl[i]);
6587 nn->conf_name_tree = RB_ROOT;
6588 nn->unconf_name_tree = RB_ROOT;
6589 INIT_LIST_HEAD(&nn->client_lru);
6590 INIT_LIST_HEAD(&nn->close_lru);
6591 INIT_LIST_HEAD(&nn->del_recall_lru);
6592 spin_lock_init(&nn->client_lock);
6594 INIT_DELAYED_WORK(&nn->laundromat_work, laundromat_main);
6600 kfree(nn->unconf_id_hashtbl);
6602 kfree(nn->conf_id_hashtbl);
6608 nfs4_state_destroy_net(struct net *net)
6611 struct nfs4_client *clp = NULL;
6612 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
6614 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
6615 while (!list_empty(&nn->conf_id_hashtbl[i])) {
6616 clp = list_entry(nn->conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
6617 destroy_client(clp);
6621 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
6622 while (!list_empty(&nn->unconf_id_hashtbl[i])) {
6623 clp = list_entry(nn->unconf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
6624 destroy_client(clp);
6628 kfree(nn->sessionid_hashtbl);
6629 kfree(nn->unconf_id_hashtbl);
6630 kfree(nn->conf_id_hashtbl);
6635 nfs4_state_start_net(struct net *net)
6637 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
6640 ret = nfs4_state_create_net(net);
6643 nn->boot_time = get_seconds();
6644 nn->grace_ended = false;
6645 nn->nfsd4_manager.block_opens = true;
6646 locks_start_grace(net, &nn->nfsd4_manager);
6647 nfsd4_client_tracking_init(net);
6648 printk(KERN_INFO "NFSD: starting %ld-second grace period (net %p)\n",
6649 nn->nfsd4_grace, net);
6650 queue_delayed_work(laundry_wq, &nn->laundromat_work, nn->nfsd4_grace * HZ);
6654 /* initialization to perform when the nfsd service is started: */
6657 nfs4_state_start(void)
6661 ret = set_callback_cred();
6664 laundry_wq = alloc_workqueue("%s", WQ_UNBOUND, 0, "nfsd4");
6665 if (laundry_wq == NULL) {
6669 ret = nfsd4_create_callback_queue();
6671 goto out_free_laundry;
6673 set_max_delegations();
6678 destroy_workqueue(laundry_wq);
6684 nfs4_state_shutdown_net(struct net *net)
6686 struct nfs4_delegation *dp = NULL;
6687 struct list_head *pos, *next, reaplist;
6688 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
6690 cancel_delayed_work_sync(&nn->laundromat_work);
6691 locks_end_grace(&nn->nfsd4_manager);
6693 INIT_LIST_HEAD(&reaplist);
6694 spin_lock(&state_lock);
6695 list_for_each_safe(pos, next, &nn->del_recall_lru) {
6696 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
6697 WARN_ON(!unhash_delegation_locked(dp));
6698 list_add(&dp->dl_recall_lru, &reaplist);
6700 spin_unlock(&state_lock);
6701 list_for_each_safe(pos, next, &reaplist) {
6702 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
6703 list_del_init(&dp->dl_recall_lru);
6704 put_clnt_odstate(dp->dl_clnt_odstate);
6705 nfs4_put_deleg_lease(dp->dl_stid.sc_file);
6706 nfs4_put_stid(&dp->dl_stid);
6709 nfsd4_client_tracking_exit(net);
6710 nfs4_state_destroy_net(net);
6714 nfs4_state_shutdown(void)
6716 destroy_workqueue(laundry_wq);
6717 nfsd4_destroy_callback_queue();
6721 get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
6723 if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG) && CURRENT_STATEID(stateid))
6724 memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
6728 put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
6730 if (cstate->minorversion) {
6731 memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
6732 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
6737 clear_current_stateid(struct nfsd4_compound_state *cstate)
6739 CLEAR_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
6743 * functions to set current state id
6746 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
6748 put_stateid(cstate, &odp->od_stateid);
6752 nfsd4_set_openstateid(struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
6754 put_stateid(cstate, &open->op_stateid);
6758 nfsd4_set_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
6760 put_stateid(cstate, &close->cl_stateid);
6764 nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate, struct nfsd4_lock *lock)
6766 put_stateid(cstate, &lock->lk_resp_stateid);
6770 * functions to consume current state id
6774 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
6776 get_stateid(cstate, &odp->od_stateid);
6780 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate, struct nfsd4_delegreturn *drp)
6782 get_stateid(cstate, &drp->dr_stateid);
6786 nfsd4_get_freestateid(struct nfsd4_compound_state *cstate, struct nfsd4_free_stateid *fsp)
6788 get_stateid(cstate, &fsp->fr_stateid);
6792 nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate, struct nfsd4_setattr *setattr)
6794 get_stateid(cstate, &setattr->sa_stateid);
6798 nfsd4_get_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
6800 get_stateid(cstate, &close->cl_stateid);
6804 nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate, struct nfsd4_locku *locku)
6806 get_stateid(cstate, &locku->lu_stateid);
6810 nfsd4_get_readstateid(struct nfsd4_compound_state *cstate, struct nfsd4_read *read)
6812 get_stateid(cstate, &read->rd_stateid);
6816 nfsd4_get_writestateid(struct nfsd4_compound_state *cstate, struct nfsd4_write *write)
6818 get_stateid(cstate, &write->wr_stateid);