2 * Copyright (c) 2001 The Regents of the University of Michigan.
5 * Kendrick Smith <kmsmith@umich.edu>
6 * Andy Adamson <kandros@umich.edu>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the University nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
22 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
23 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
24 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
28 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
29 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
30 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
31 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
35 #include <linux/file.h>
37 #include <linux/slab.h>
38 #include <linux/namei.h>
39 #include <linux/swap.h>
40 #include <linux/pagemap.h>
41 #include <linux/ratelimit.h>
42 #include <linux/sunrpc/svcauth_gss.h>
43 #include <linux/sunrpc/addr.h>
44 #include <linux/hash.h>
48 #include "current_stateid.h"
52 #define NFSDDBG_FACILITY NFSDDBG_PROC
54 #define all_ones {{~0,~0},~0}
55 static const stateid_t one_stateid = {
57 .si_opaque = all_ones,
59 static const stateid_t zero_stateid = {
62 static const stateid_t currentstateid = {
66 static u64 current_sessionid = 1;
68 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zero_stateid, sizeof(stateid_t)))
69 #define ONE_STATEID(stateid) (!memcmp((stateid), &one_stateid, sizeof(stateid_t)))
70 #define CURRENT_STATEID(stateid) (!memcmp((stateid), ¤tstateid, sizeof(stateid_t)))
72 /* forward declarations */
73 static bool check_for_locks(struct nfs4_file *fp, struct nfs4_lockowner *lowner);
74 static void nfs4_free_ol_stateid(struct nfs4_stid *stid);
79 * Currently used for the del_recall_lru and file hash table. In an
80 * effort to decrease the scope of the client_mutex, this spinlock may
81 * eventually cover more:
83 static DEFINE_SPINLOCK(state_lock);
86 * A waitqueue for all in-progress 4.0 CLOSE operations that are waiting for
87 * the refcount on the open stateid to drop.
89 static DECLARE_WAIT_QUEUE_HEAD(close_wq);
91 static struct kmem_cache *openowner_slab;
92 static struct kmem_cache *lockowner_slab;
93 static struct kmem_cache *file_slab;
94 static struct kmem_cache *stateid_slab;
95 static struct kmem_cache *deleg_slab;
97 static void free_session(struct nfsd4_session *);
99 static bool is_session_dead(struct nfsd4_session *ses)
101 return ses->se_flags & NFS4_SESSION_DEAD;
104 static __be32 mark_session_dead_locked(struct nfsd4_session *ses, int ref_held_by_me)
106 if (atomic_read(&ses->se_ref) > ref_held_by_me)
107 return nfserr_jukebox;
108 ses->se_flags |= NFS4_SESSION_DEAD;
112 static bool is_client_expired(struct nfs4_client *clp)
114 return clp->cl_time == 0;
117 static __be32 get_client_locked(struct nfs4_client *clp)
119 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
121 lockdep_assert_held(&nn->client_lock);
123 if (is_client_expired(clp))
124 return nfserr_expired;
125 atomic_inc(&clp->cl_refcount);
129 /* must be called under the client_lock */
131 renew_client_locked(struct nfs4_client *clp)
133 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
135 if (is_client_expired(clp)) {
137 printk("%s: client (clientid %08x/%08x) already expired\n",
139 clp->cl_clientid.cl_boot,
140 clp->cl_clientid.cl_id);
144 dprintk("renewing client (clientid %08x/%08x)\n",
145 clp->cl_clientid.cl_boot,
146 clp->cl_clientid.cl_id);
147 list_move_tail(&clp->cl_lru, &nn->client_lru);
148 clp->cl_time = get_seconds();
152 renew_client(struct nfs4_client *clp)
154 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
156 spin_lock(&nn->client_lock);
157 renew_client_locked(clp);
158 spin_unlock(&nn->client_lock);
161 static void put_client_renew_locked(struct nfs4_client *clp)
163 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
165 lockdep_assert_held(&nn->client_lock);
167 if (!atomic_dec_and_test(&clp->cl_refcount))
169 if (!is_client_expired(clp))
170 renew_client_locked(clp);
173 static void put_client_renew(struct nfs4_client *clp)
175 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
177 if (!atomic_dec_and_lock(&clp->cl_refcount, &nn->client_lock))
179 if (!is_client_expired(clp))
180 renew_client_locked(clp);
181 spin_unlock(&nn->client_lock);
184 static __be32 nfsd4_get_session_locked(struct nfsd4_session *ses)
188 if (is_session_dead(ses))
189 return nfserr_badsession;
190 status = get_client_locked(ses->se_client);
193 atomic_inc(&ses->se_ref);
197 static void nfsd4_put_session_locked(struct nfsd4_session *ses)
199 struct nfs4_client *clp = ses->se_client;
200 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
202 lockdep_assert_held(&nn->client_lock);
204 if (atomic_dec_and_test(&ses->se_ref) && is_session_dead(ses))
206 put_client_renew_locked(clp);
209 static void nfsd4_put_session(struct nfsd4_session *ses)
211 struct nfs4_client *clp = ses->se_client;
212 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
214 spin_lock(&nn->client_lock);
215 nfsd4_put_session_locked(ses);
216 spin_unlock(&nn->client_lock);
220 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner)
222 return (sop->so_owner.len == owner->len) &&
223 0 == memcmp(sop->so_owner.data, owner->data, owner->len);
226 static struct nfs4_openowner *
227 find_openstateowner_str_locked(unsigned int hashval, struct nfsd4_open *open,
228 struct nfs4_client *clp)
230 struct nfs4_stateowner *so;
232 lockdep_assert_held(&clp->cl_lock);
234 list_for_each_entry(so, &clp->cl_ownerstr_hashtbl[hashval],
236 if (!so->so_is_open_owner)
238 if (same_owner_str(so, &open->op_owner)) {
239 atomic_inc(&so->so_count);
240 return openowner(so);
246 static struct nfs4_openowner *
247 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open,
248 struct nfs4_client *clp)
250 struct nfs4_openowner *oo;
252 spin_lock(&clp->cl_lock);
253 oo = find_openstateowner_str_locked(hashval, open, clp);
254 spin_unlock(&clp->cl_lock);
259 opaque_hashval(const void *ptr, int nbytes)
261 unsigned char *cptr = (unsigned char *) ptr;
271 static void nfsd4_free_file(struct nfs4_file *f)
273 kmem_cache_free(file_slab, f);
277 put_nfs4_file(struct nfs4_file *fi)
279 might_lock(&state_lock);
281 if (atomic_dec_and_lock(&fi->fi_ref, &state_lock)) {
282 hlist_del(&fi->fi_hash);
283 spin_unlock(&state_lock);
289 get_nfs4_file(struct nfs4_file *fi)
291 atomic_inc(&fi->fi_ref);
295 __nfs4_get_fd(struct nfs4_file *f, int oflag)
297 if (f->fi_fds[oflag])
298 return get_file(f->fi_fds[oflag]);
303 find_writeable_file_locked(struct nfs4_file *f)
307 lockdep_assert_held(&f->fi_lock);
309 ret = __nfs4_get_fd(f, O_WRONLY);
311 ret = __nfs4_get_fd(f, O_RDWR);
316 find_writeable_file(struct nfs4_file *f)
320 spin_lock(&f->fi_lock);
321 ret = find_writeable_file_locked(f);
322 spin_unlock(&f->fi_lock);
327 static struct file *find_readable_file_locked(struct nfs4_file *f)
331 lockdep_assert_held(&f->fi_lock);
333 ret = __nfs4_get_fd(f, O_RDONLY);
335 ret = __nfs4_get_fd(f, O_RDWR);
340 find_readable_file(struct nfs4_file *f)
344 spin_lock(&f->fi_lock);
345 ret = find_readable_file_locked(f);
346 spin_unlock(&f->fi_lock);
352 find_any_file(struct nfs4_file *f)
356 spin_lock(&f->fi_lock);
357 ret = __nfs4_get_fd(f, O_RDWR);
359 ret = __nfs4_get_fd(f, O_WRONLY);
361 ret = __nfs4_get_fd(f, O_RDONLY);
363 spin_unlock(&f->fi_lock);
367 static atomic_long_t num_delegations;
368 unsigned long max_delegations;
371 * Open owner state (share locks)
374 /* hash tables for lock and open owners */
375 #define OWNER_HASH_BITS 8
376 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
377 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
379 static unsigned int ownerstr_hashval(struct xdr_netobj *ownername)
383 ret = opaque_hashval(ownername->data, ownername->len);
384 return ret & OWNER_HASH_MASK;
387 /* hash table for nfs4_file */
388 #define FILE_HASH_BITS 8
389 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
391 static unsigned int nfsd_fh_hashval(struct knfsd_fh *fh)
393 return jhash2(fh->fh_base.fh_pad, XDR_QUADLEN(fh->fh_size), 0);
396 static unsigned int file_hashval(struct knfsd_fh *fh)
398 return nfsd_fh_hashval(fh) & (FILE_HASH_SIZE - 1);
401 static bool nfsd_fh_match(struct knfsd_fh *fh1, struct knfsd_fh *fh2)
403 return fh1->fh_size == fh2->fh_size &&
404 !memcmp(fh1->fh_base.fh_pad,
409 static struct hlist_head file_hashtbl[FILE_HASH_SIZE];
412 __nfs4_file_get_access(struct nfs4_file *fp, u32 access)
414 lockdep_assert_held(&fp->fi_lock);
416 if (access & NFS4_SHARE_ACCESS_WRITE)
417 atomic_inc(&fp->fi_access[O_WRONLY]);
418 if (access & NFS4_SHARE_ACCESS_READ)
419 atomic_inc(&fp->fi_access[O_RDONLY]);
423 nfs4_file_get_access(struct nfs4_file *fp, u32 access)
425 lockdep_assert_held(&fp->fi_lock);
427 /* Does this access mode make sense? */
428 if (access & ~NFS4_SHARE_ACCESS_BOTH)
431 /* Does it conflict with a deny mode already set? */
432 if ((access & fp->fi_share_deny) != 0)
433 return nfserr_share_denied;
435 __nfs4_file_get_access(fp, access);
439 static __be32 nfs4_file_check_deny(struct nfs4_file *fp, u32 deny)
441 /* Common case is that there is no deny mode. */
443 /* Does this deny mode make sense? */
444 if (deny & ~NFS4_SHARE_DENY_BOTH)
447 if ((deny & NFS4_SHARE_DENY_READ) &&
448 atomic_read(&fp->fi_access[O_RDONLY]))
449 return nfserr_share_denied;
451 if ((deny & NFS4_SHARE_DENY_WRITE) &&
452 atomic_read(&fp->fi_access[O_WRONLY]))
453 return nfserr_share_denied;
458 static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
460 might_lock(&fp->fi_lock);
462 if (atomic_dec_and_lock(&fp->fi_access[oflag], &fp->fi_lock)) {
463 struct file *f1 = NULL;
464 struct file *f2 = NULL;
466 swap(f1, fp->fi_fds[oflag]);
467 if (atomic_read(&fp->fi_access[1 - oflag]) == 0)
468 swap(f2, fp->fi_fds[O_RDWR]);
469 spin_unlock(&fp->fi_lock);
477 static void nfs4_file_put_access(struct nfs4_file *fp, u32 access)
479 WARN_ON_ONCE(access & ~NFS4_SHARE_ACCESS_BOTH);
481 if (access & NFS4_SHARE_ACCESS_WRITE)
482 __nfs4_file_put_access(fp, O_WRONLY);
483 if (access & NFS4_SHARE_ACCESS_READ)
484 __nfs4_file_put_access(fp, O_RDONLY);
487 static struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl,
488 struct kmem_cache *slab)
490 struct nfs4_stid *stid;
493 stid = kmem_cache_zalloc(slab, GFP_KERNEL);
497 idr_preload(GFP_KERNEL);
498 spin_lock(&cl->cl_lock);
499 new_id = idr_alloc_cyclic(&cl->cl_stateids, stid, 0, 0, GFP_NOWAIT);
500 spin_unlock(&cl->cl_lock);
504 stid->sc_client = cl;
505 stid->sc_stateid.si_opaque.so_id = new_id;
506 stid->sc_stateid.si_opaque.so_clid = cl->cl_clientid;
507 /* Will be incremented before return to client: */
508 atomic_set(&stid->sc_count, 1);
511 * It shouldn't be a problem to reuse an opaque stateid value.
512 * I don't think it is for 4.1. But with 4.0 I worry that, for
513 * example, a stray write retransmission could be accepted by
514 * the server when it should have been rejected. Therefore,
515 * adopt a trick from the sctp code to attempt to maximize the
516 * amount of time until an id is reused, by ensuring they always
517 * "increase" (mod INT_MAX):
521 kmem_cache_free(slab, stid);
525 static struct nfs4_ol_stateid * nfs4_alloc_open_stateid(struct nfs4_client *clp)
527 struct nfs4_stid *stid;
528 struct nfs4_ol_stateid *stp;
530 stid = nfs4_alloc_stid(clp, stateid_slab);
534 stp = openlockstateid(stid);
535 stp->st_stid.sc_free = nfs4_free_ol_stateid;
539 static void nfs4_free_deleg(struct nfs4_stid *stid)
541 kmem_cache_free(deleg_slab, stid);
542 atomic_long_dec(&num_delegations);
546 * When we recall a delegation, we should be careful not to hand it
547 * out again straight away.
548 * To ensure this we keep a pair of bloom filters ('new' and 'old')
549 * in which the filehandles of recalled delegations are "stored".
550 * If a filehandle appear in either filter, a delegation is blocked.
551 * When a delegation is recalled, the filehandle is stored in the "new"
553 * Every 30 seconds we swap the filters and clear the "new" one,
554 * unless both are empty of course.
556 * Each filter is 256 bits. We hash the filehandle to 32bit and use the
557 * low 3 bytes as hash-table indices.
559 * 'blocked_delegations_lock', which is always taken in block_delegations(),
560 * is used to manage concurrent access. Testing does not need the lock
561 * except when swapping the two filters.
563 static DEFINE_SPINLOCK(blocked_delegations_lock);
564 static struct bloom_pair {
565 int entries, old_entries;
567 int new; /* index into 'set' */
568 DECLARE_BITMAP(set[2], 256);
569 } blocked_delegations;
571 static int delegation_blocked(struct knfsd_fh *fh)
574 struct bloom_pair *bd = &blocked_delegations;
576 if (bd->entries == 0)
578 if (seconds_since_boot() - bd->swap_time > 30) {
579 spin_lock(&blocked_delegations_lock);
580 if (seconds_since_boot() - bd->swap_time > 30) {
581 bd->entries -= bd->old_entries;
582 bd->old_entries = bd->entries;
583 memset(bd->set[bd->new], 0,
586 bd->swap_time = seconds_since_boot();
588 spin_unlock(&blocked_delegations_lock);
590 hash = arch_fast_hash(&fh->fh_base, fh->fh_size, 0);
591 if (test_bit(hash&255, bd->set[0]) &&
592 test_bit((hash>>8)&255, bd->set[0]) &&
593 test_bit((hash>>16)&255, bd->set[0]))
596 if (test_bit(hash&255, bd->set[1]) &&
597 test_bit((hash>>8)&255, bd->set[1]) &&
598 test_bit((hash>>16)&255, bd->set[1]))
604 static void block_delegations(struct knfsd_fh *fh)
607 struct bloom_pair *bd = &blocked_delegations;
609 hash = arch_fast_hash(&fh->fh_base, fh->fh_size, 0);
611 spin_lock(&blocked_delegations_lock);
612 __set_bit(hash&255, bd->set[bd->new]);
613 __set_bit((hash>>8)&255, bd->set[bd->new]);
614 __set_bit((hash>>16)&255, bd->set[bd->new]);
615 if (bd->entries == 0)
616 bd->swap_time = seconds_since_boot();
618 spin_unlock(&blocked_delegations_lock);
621 static struct nfs4_delegation *
622 alloc_init_deleg(struct nfs4_client *clp, struct svc_fh *current_fh)
624 struct nfs4_delegation *dp;
627 dprintk("NFSD alloc_init_deleg\n");
628 n = atomic_long_inc_return(&num_delegations);
629 if (n < 0 || n > max_delegations)
631 if (delegation_blocked(¤t_fh->fh_handle))
633 dp = delegstateid(nfs4_alloc_stid(clp, deleg_slab));
637 dp->dl_stid.sc_free = nfs4_free_deleg;
639 * delegation seqid's are never incremented. The 4.1 special
640 * meaning of seqid 0 isn't meaningful, really, but let's avoid
641 * 0 anyway just for consistency and use 1:
643 dp->dl_stid.sc_stateid.si_generation = 1;
644 INIT_LIST_HEAD(&dp->dl_perfile);
645 INIT_LIST_HEAD(&dp->dl_perclnt);
646 INIT_LIST_HEAD(&dp->dl_recall_lru);
647 dp->dl_type = NFS4_OPEN_DELEGATE_READ;
648 INIT_WORK(&dp->dl_recall.cb_work, nfsd4_run_cb_recall);
651 atomic_long_dec(&num_delegations);
656 nfs4_put_stid(struct nfs4_stid *s)
658 struct nfs4_file *fp = s->sc_file;
659 struct nfs4_client *clp = s->sc_client;
661 might_lock(&clp->cl_lock);
663 if (!atomic_dec_and_lock(&s->sc_count, &clp->cl_lock)) {
664 wake_up_all(&close_wq);
667 idr_remove(&clp->cl_stateids, s->sc_stateid.si_opaque.so_id);
668 spin_unlock(&clp->cl_lock);
674 static void nfs4_put_deleg_lease(struct nfs4_file *fp)
676 lockdep_assert_held(&state_lock);
680 if (atomic_dec_and_test(&fp->fi_delegees)) {
681 vfs_setlease(fp->fi_deleg_file, F_UNLCK, &fp->fi_lease);
683 fput(fp->fi_deleg_file);
684 fp->fi_deleg_file = NULL;
688 static void unhash_stid(struct nfs4_stid *s)
694 hash_delegation_locked(struct nfs4_delegation *dp, struct nfs4_file *fp)
696 lockdep_assert_held(&state_lock);
697 lockdep_assert_held(&fp->fi_lock);
699 atomic_inc(&dp->dl_stid.sc_count);
700 dp->dl_stid.sc_type = NFS4_DELEG_STID;
701 list_add(&dp->dl_perfile, &fp->fi_delegations);
702 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
706 unhash_delegation_locked(struct nfs4_delegation *dp)
708 struct nfs4_file *fp = dp->dl_stid.sc_file;
710 lockdep_assert_held(&state_lock);
712 dp->dl_stid.sc_type = NFS4_CLOSED_DELEG_STID;
713 /* Ensure that deleg break won't try to requeue it */
715 spin_lock(&fp->fi_lock);
716 list_del_init(&dp->dl_perclnt);
717 list_del_init(&dp->dl_recall_lru);
718 list_del_init(&dp->dl_perfile);
719 spin_unlock(&fp->fi_lock);
721 nfs4_put_deleg_lease(fp);
724 static void destroy_delegation(struct nfs4_delegation *dp)
726 spin_lock(&state_lock);
727 unhash_delegation_locked(dp);
728 spin_unlock(&state_lock);
729 nfs4_put_stid(&dp->dl_stid);
732 static void revoke_delegation(struct nfs4_delegation *dp)
734 struct nfs4_client *clp = dp->dl_stid.sc_client;
736 WARN_ON(!list_empty(&dp->dl_recall_lru));
738 if (clp->cl_minorversion == 0)
739 nfs4_put_stid(&dp->dl_stid);
741 dp->dl_stid.sc_type = NFS4_REVOKED_DELEG_STID;
742 spin_lock(&clp->cl_lock);
743 list_add(&dp->dl_recall_lru, &clp->cl_revoked);
744 spin_unlock(&clp->cl_lock);
752 static unsigned int clientid_hashval(u32 id)
754 return id & CLIENT_HASH_MASK;
757 static unsigned int clientstr_hashval(const char *name)
759 return opaque_hashval(name, 8) & CLIENT_HASH_MASK;
763 * We store the NONE, READ, WRITE, and BOTH bits separately in the
764 * st_{access,deny}_bmap field of the stateid, in order to track not
765 * only what share bits are currently in force, but also what
766 * combinations of share bits previous opens have used. This allows us
767 * to enforce the recommendation of rfc 3530 14.2.19 that the server
768 * return an error if the client attempt to downgrade to a combination
769 * of share bits not explicable by closing some of its previous opens.
771 * XXX: This enforcement is actually incomplete, since we don't keep
772 * track of access/deny bit combinations; so, e.g., we allow:
774 * OPEN allow read, deny write
775 * OPEN allow both, deny none
776 * DOWNGRADE allow read, deny none
778 * which we should reject.
781 bmap_to_share_mode(unsigned long bmap) {
783 unsigned int access = 0;
785 for (i = 1; i < 4; i++) {
786 if (test_bit(i, &bmap))
792 /* set share access for a given stateid */
794 set_access(u32 access, struct nfs4_ol_stateid *stp)
796 unsigned char mask = 1 << access;
798 WARN_ON_ONCE(access > NFS4_SHARE_ACCESS_BOTH);
799 stp->st_access_bmap |= mask;
802 /* clear share access for a given stateid */
804 clear_access(u32 access, struct nfs4_ol_stateid *stp)
806 unsigned char mask = 1 << access;
808 WARN_ON_ONCE(access > NFS4_SHARE_ACCESS_BOTH);
809 stp->st_access_bmap &= ~mask;
812 /* test whether a given stateid has access */
814 test_access(u32 access, struct nfs4_ol_stateid *stp)
816 unsigned char mask = 1 << access;
818 return (bool)(stp->st_access_bmap & mask);
821 /* set share deny for a given stateid */
823 set_deny(u32 deny, struct nfs4_ol_stateid *stp)
825 unsigned char mask = 1 << deny;
827 WARN_ON_ONCE(deny > NFS4_SHARE_DENY_BOTH);
828 stp->st_deny_bmap |= mask;
831 /* clear share deny for a given stateid */
833 clear_deny(u32 deny, struct nfs4_ol_stateid *stp)
835 unsigned char mask = 1 << deny;
837 WARN_ON_ONCE(deny > NFS4_SHARE_DENY_BOTH);
838 stp->st_deny_bmap &= ~mask;
841 /* test whether a given stateid is denying specific access */
843 test_deny(u32 deny, struct nfs4_ol_stateid *stp)
845 unsigned char mask = 1 << deny;
847 return (bool)(stp->st_deny_bmap & mask);
850 static int nfs4_access_to_omode(u32 access)
852 switch (access & NFS4_SHARE_ACCESS_BOTH) {
853 case NFS4_SHARE_ACCESS_READ:
855 case NFS4_SHARE_ACCESS_WRITE:
857 case NFS4_SHARE_ACCESS_BOTH:
865 * A stateid that had a deny mode associated with it is being released
866 * or downgraded. Recalculate the deny mode on the file.
869 recalculate_deny_mode(struct nfs4_file *fp)
871 struct nfs4_ol_stateid *stp;
873 spin_lock(&fp->fi_lock);
874 fp->fi_share_deny = 0;
875 list_for_each_entry(stp, &fp->fi_stateids, st_perfile)
876 fp->fi_share_deny |= bmap_to_share_mode(stp->st_deny_bmap);
877 spin_unlock(&fp->fi_lock);
881 reset_union_bmap_deny(u32 deny, struct nfs4_ol_stateid *stp)
886 for (i = 1; i < 4; i++) {
887 if ((i & deny) != i) {
893 /* Recalculate per-file deny mode if there was a change */
895 recalculate_deny_mode(stp->st_stid.sc_file);
898 /* release all access and file references for a given stateid */
900 release_all_access(struct nfs4_ol_stateid *stp)
903 struct nfs4_file *fp = stp->st_stid.sc_file;
905 if (fp && stp->st_deny_bmap != 0)
906 recalculate_deny_mode(fp);
908 for (i = 1; i < 4; i++) {
909 if (test_access(i, stp))
910 nfs4_file_put_access(stp->st_stid.sc_file, i);
911 clear_access(i, stp);
915 static void nfs4_put_stateowner(struct nfs4_stateowner *sop)
917 struct nfs4_client *clp = sop->so_client;
919 might_lock(&clp->cl_lock);
921 if (!atomic_dec_and_lock(&sop->so_count, &clp->cl_lock))
923 sop->so_ops->so_unhash(sop);
924 spin_unlock(&clp->cl_lock);
925 kfree(sop->so_owner.data);
926 sop->so_ops->so_free(sop);
929 static void unhash_ol_stateid(struct nfs4_ol_stateid *stp)
931 struct nfs4_file *fp = stp->st_stid.sc_file;
933 lockdep_assert_held(&stp->st_stateowner->so_client->cl_lock);
935 spin_lock(&fp->fi_lock);
936 list_del(&stp->st_perfile);
937 spin_unlock(&fp->fi_lock);
938 list_del(&stp->st_perstateowner);
941 static void nfs4_free_ol_stateid(struct nfs4_stid *stid)
943 struct nfs4_ol_stateid *stp = openlockstateid(stid);
945 release_all_access(stp);
946 if (stp->st_stateowner)
947 nfs4_put_stateowner(stp->st_stateowner);
948 kmem_cache_free(stateid_slab, stid);
951 static void nfs4_free_lock_stateid(struct nfs4_stid *stid)
953 struct nfs4_ol_stateid *stp = openlockstateid(stid);
954 struct nfs4_lockowner *lo = lockowner(stp->st_stateowner);
957 file = find_any_file(stp->st_stid.sc_file);
959 filp_close(file, (fl_owner_t)lo);
960 nfs4_free_ol_stateid(stid);
964 * Put the persistent reference to an already unhashed generic stateid, while
965 * holding the cl_lock. If it's the last reference, then put it onto the
966 * reaplist for later destruction.
968 static void put_ol_stateid_locked(struct nfs4_ol_stateid *stp,
969 struct list_head *reaplist)
971 struct nfs4_stid *s = &stp->st_stid;
972 struct nfs4_client *clp = s->sc_client;
974 lockdep_assert_held(&clp->cl_lock);
976 WARN_ON_ONCE(!list_empty(&stp->st_locks));
978 if (!atomic_dec_and_test(&s->sc_count)) {
979 wake_up_all(&close_wq);
983 idr_remove(&clp->cl_stateids, s->sc_stateid.si_opaque.so_id);
984 list_add(&stp->st_locks, reaplist);
987 static void unhash_lock_stateid(struct nfs4_ol_stateid *stp)
989 struct nfs4_openowner *oo = openowner(stp->st_openstp->st_stateowner);
991 lockdep_assert_held(&oo->oo_owner.so_client->cl_lock);
993 list_del_init(&stp->st_locks);
994 unhash_ol_stateid(stp);
995 unhash_stid(&stp->st_stid);
998 static void release_lock_stateid(struct nfs4_ol_stateid *stp)
1000 struct nfs4_openowner *oo = openowner(stp->st_openstp->st_stateowner);
1002 spin_lock(&oo->oo_owner.so_client->cl_lock);
1003 unhash_lock_stateid(stp);
1004 spin_unlock(&oo->oo_owner.so_client->cl_lock);
1005 nfs4_put_stid(&stp->st_stid);
1008 static void unhash_lockowner_locked(struct nfs4_lockowner *lo)
1010 struct nfs4_client *clp = lo->lo_owner.so_client;
1012 lockdep_assert_held(&clp->cl_lock);
1014 list_del_init(&lo->lo_owner.so_strhash);
1018 * Free a list of generic stateids that were collected earlier after being
1022 free_ol_stateid_reaplist(struct list_head *reaplist)
1024 struct nfs4_ol_stateid *stp;
1025 struct nfs4_file *fp;
1029 while (!list_empty(reaplist)) {
1030 stp = list_first_entry(reaplist, struct nfs4_ol_stateid,
1032 list_del(&stp->st_locks);
1033 fp = stp->st_stid.sc_file;
1034 stp->st_stid.sc_free(&stp->st_stid);
1040 static void release_lockowner(struct nfs4_lockowner *lo)
1042 struct nfs4_client *clp = lo->lo_owner.so_client;
1043 struct nfs4_ol_stateid *stp;
1044 struct list_head reaplist;
1046 INIT_LIST_HEAD(&reaplist);
1048 spin_lock(&clp->cl_lock);
1049 unhash_lockowner_locked(lo);
1050 while (!list_empty(&lo->lo_owner.so_stateids)) {
1051 stp = list_first_entry(&lo->lo_owner.so_stateids,
1052 struct nfs4_ol_stateid, st_perstateowner);
1053 unhash_lock_stateid(stp);
1054 put_ol_stateid_locked(stp, &reaplist);
1056 spin_unlock(&clp->cl_lock);
1057 free_ol_stateid_reaplist(&reaplist);
1058 nfs4_put_stateowner(&lo->lo_owner);
1061 static void release_open_stateid_locks(struct nfs4_ol_stateid *open_stp,
1062 struct list_head *reaplist)
1064 struct nfs4_ol_stateid *stp;
1066 while (!list_empty(&open_stp->st_locks)) {
1067 stp = list_entry(open_stp->st_locks.next,
1068 struct nfs4_ol_stateid, st_locks);
1069 unhash_lock_stateid(stp);
1070 put_ol_stateid_locked(stp, reaplist);
1074 static void unhash_open_stateid(struct nfs4_ol_stateid *stp,
1075 struct list_head *reaplist)
1077 lockdep_assert_held(&stp->st_stid.sc_client->cl_lock);
1079 unhash_ol_stateid(stp);
1080 release_open_stateid_locks(stp, reaplist);
1083 static void release_open_stateid(struct nfs4_ol_stateid *stp)
1085 LIST_HEAD(reaplist);
1087 spin_lock(&stp->st_stid.sc_client->cl_lock);
1088 unhash_open_stateid(stp, &reaplist);
1089 put_ol_stateid_locked(stp, &reaplist);
1090 spin_unlock(&stp->st_stid.sc_client->cl_lock);
1091 free_ol_stateid_reaplist(&reaplist);
1094 static void unhash_openowner_locked(struct nfs4_openowner *oo)
1096 struct nfs4_client *clp = oo->oo_owner.so_client;
1098 lockdep_assert_held(&clp->cl_lock);
1100 list_del_init(&oo->oo_owner.so_strhash);
1101 list_del_init(&oo->oo_perclient);
1104 static void release_last_closed_stateid(struct nfs4_openowner *oo)
1106 struct nfsd_net *nn = net_generic(oo->oo_owner.so_client->net,
1108 struct nfs4_ol_stateid *s;
1110 spin_lock(&nn->client_lock);
1111 s = oo->oo_last_closed_stid;
1113 list_del_init(&oo->oo_close_lru);
1114 oo->oo_last_closed_stid = NULL;
1116 spin_unlock(&nn->client_lock);
1118 nfs4_put_stid(&s->st_stid);
1121 static void release_openowner(struct nfs4_openowner *oo)
1123 struct nfs4_ol_stateid *stp;
1124 struct nfs4_client *clp = oo->oo_owner.so_client;
1125 struct list_head reaplist;
1127 INIT_LIST_HEAD(&reaplist);
1129 spin_lock(&clp->cl_lock);
1130 unhash_openowner_locked(oo);
1131 while (!list_empty(&oo->oo_owner.so_stateids)) {
1132 stp = list_first_entry(&oo->oo_owner.so_stateids,
1133 struct nfs4_ol_stateid, st_perstateowner);
1134 unhash_open_stateid(stp, &reaplist);
1135 put_ol_stateid_locked(stp, &reaplist);
1137 spin_unlock(&clp->cl_lock);
1138 free_ol_stateid_reaplist(&reaplist);
1139 release_last_closed_stateid(oo);
1140 nfs4_put_stateowner(&oo->oo_owner);
1144 hash_sessionid(struct nfs4_sessionid *sessionid)
1146 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
1148 return sid->sequence % SESSION_HASH_SIZE;
1153 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
1155 u32 *ptr = (u32 *)(&sessionid->data[0]);
1156 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
1160 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
1166 * Bump the seqid on cstate->replay_owner, and clear replay_owner if it
1167 * won't be used for replay.
1169 void nfsd4_bump_seqid(struct nfsd4_compound_state *cstate, __be32 nfserr)
1171 struct nfs4_stateowner *so = cstate->replay_owner;
1173 if (nfserr == nfserr_replay_me)
1176 if (!seqid_mutating_err(ntohl(nfserr))) {
1177 nfsd4_cstate_clear_replay(cstate);
1182 if (so->so_is_open_owner)
1183 release_last_closed_stateid(openowner(so));
1189 gen_sessionid(struct nfsd4_session *ses)
1191 struct nfs4_client *clp = ses->se_client;
1192 struct nfsd4_sessionid *sid;
1194 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
1195 sid->clientid = clp->cl_clientid;
1196 sid->sequence = current_sessionid++;
1201 * The protocol defines ca_maxresponssize_cached to include the size of
1202 * the rpc header, but all we need to cache is the data starting after
1203 * the end of the initial SEQUENCE operation--the rest we regenerate
1204 * each time. Therefore we can advertise a ca_maxresponssize_cached
1205 * value that is the number of bytes in our cache plus a few additional
1206 * bytes. In order to stay on the safe side, and not promise more than
1207 * we can cache, those additional bytes must be the minimum possible: 24
1208 * bytes of rpc header (xid through accept state, with AUTH_NULL
1209 * verifier), 12 for the compound header (with zero-length tag), and 44
1210 * for the SEQUENCE op response:
1212 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
1215 free_session_slots(struct nfsd4_session *ses)
1219 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
1220 kfree(ses->se_slots[i]);
1224 * We don't actually need to cache the rpc and session headers, so we
1225 * can allocate a little less for each slot:
1227 static inline u32 slot_bytes(struct nfsd4_channel_attrs *ca)
1231 if (ca->maxresp_cached < NFSD_MIN_HDR_SEQ_SZ)
1234 size = ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
1235 return size + sizeof(struct nfsd4_slot);
1239 * XXX: If we run out of reserved DRC memory we could (up to a point)
1240 * re-negotiate active sessions and reduce their slot usage to make
1241 * room for new connections. For now we just fail the create session.
1243 static u32 nfsd4_get_drc_mem(struct nfsd4_channel_attrs *ca)
1245 u32 slotsize = slot_bytes(ca);
1246 u32 num = ca->maxreqs;
1249 spin_lock(&nfsd_drc_lock);
1250 avail = min((unsigned long)NFSD_MAX_MEM_PER_SESSION,
1251 nfsd_drc_max_mem - nfsd_drc_mem_used);
1252 num = min_t(int, num, avail / slotsize);
1253 nfsd_drc_mem_used += num * slotsize;
1254 spin_unlock(&nfsd_drc_lock);
1259 static void nfsd4_put_drc_mem(struct nfsd4_channel_attrs *ca)
1261 int slotsize = slot_bytes(ca);
1263 spin_lock(&nfsd_drc_lock);
1264 nfsd_drc_mem_used -= slotsize * ca->maxreqs;
1265 spin_unlock(&nfsd_drc_lock);
1268 static struct nfsd4_session *alloc_session(struct nfsd4_channel_attrs *fattrs,
1269 struct nfsd4_channel_attrs *battrs)
1271 int numslots = fattrs->maxreqs;
1272 int slotsize = slot_bytes(fattrs);
1273 struct nfsd4_session *new;
1276 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
1277 + sizeof(struct nfsd4_session) > PAGE_SIZE);
1278 mem = numslots * sizeof(struct nfsd4_slot *);
1280 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
1283 /* allocate each struct nfsd4_slot and data cache in one piece */
1284 for (i = 0; i < numslots; i++) {
1285 new->se_slots[i] = kzalloc(slotsize, GFP_KERNEL);
1286 if (!new->se_slots[i])
1290 memcpy(&new->se_fchannel, fattrs, sizeof(struct nfsd4_channel_attrs));
1291 memcpy(&new->se_bchannel, battrs, sizeof(struct nfsd4_channel_attrs));
1296 kfree(new->se_slots[i]);
1301 static void free_conn(struct nfsd4_conn *c)
1303 svc_xprt_put(c->cn_xprt);
1307 static void nfsd4_conn_lost(struct svc_xpt_user *u)
1309 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
1310 struct nfs4_client *clp = c->cn_session->se_client;
1312 spin_lock(&clp->cl_lock);
1313 if (!list_empty(&c->cn_persession)) {
1314 list_del(&c->cn_persession);
1317 nfsd4_probe_callback(clp);
1318 spin_unlock(&clp->cl_lock);
1321 static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
1323 struct nfsd4_conn *conn;
1325 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
1328 svc_xprt_get(rqstp->rq_xprt);
1329 conn->cn_xprt = rqstp->rq_xprt;
1330 conn->cn_flags = flags;
1331 INIT_LIST_HEAD(&conn->cn_xpt_user.list);
1335 static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
1337 conn->cn_session = ses;
1338 list_add(&conn->cn_persession, &ses->se_conns);
1341 static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
1343 struct nfs4_client *clp = ses->se_client;
1345 spin_lock(&clp->cl_lock);
1346 __nfsd4_hash_conn(conn, ses);
1347 spin_unlock(&clp->cl_lock);
1350 static int nfsd4_register_conn(struct nfsd4_conn *conn)
1352 conn->cn_xpt_user.callback = nfsd4_conn_lost;
1353 return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
1356 static void nfsd4_init_conn(struct svc_rqst *rqstp, struct nfsd4_conn *conn, struct nfsd4_session *ses)
1360 nfsd4_hash_conn(conn, ses);
1361 ret = nfsd4_register_conn(conn);
1363 /* oops; xprt is already down: */
1364 nfsd4_conn_lost(&conn->cn_xpt_user);
1365 /* We may have gained or lost a callback channel: */
1366 nfsd4_probe_callback_sync(ses->se_client);
1369 static struct nfsd4_conn *alloc_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_create_session *cses)
1371 u32 dir = NFS4_CDFC4_FORE;
1373 if (cses->flags & SESSION4_BACK_CHAN)
1374 dir |= NFS4_CDFC4_BACK;
1375 return alloc_conn(rqstp, dir);
1378 /* must be called under client_lock */
1379 static void nfsd4_del_conns(struct nfsd4_session *s)
1381 struct nfs4_client *clp = s->se_client;
1382 struct nfsd4_conn *c;
1384 spin_lock(&clp->cl_lock);
1385 while (!list_empty(&s->se_conns)) {
1386 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
1387 list_del_init(&c->cn_persession);
1388 spin_unlock(&clp->cl_lock);
1390 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
1393 spin_lock(&clp->cl_lock);
1395 spin_unlock(&clp->cl_lock);
1398 static void __free_session(struct nfsd4_session *ses)
1400 free_session_slots(ses);
1404 static void free_session(struct nfsd4_session *ses)
1406 nfsd4_del_conns(ses);
1407 nfsd4_put_drc_mem(&ses->se_fchannel);
1408 __free_session(ses);
1411 static void init_session(struct svc_rqst *rqstp, struct nfsd4_session *new, struct nfs4_client *clp, struct nfsd4_create_session *cses)
1414 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1416 new->se_client = clp;
1419 INIT_LIST_HEAD(&new->se_conns);
1421 new->se_cb_seq_nr = 1;
1422 new->se_flags = cses->flags;
1423 new->se_cb_prog = cses->callback_prog;
1424 new->se_cb_sec = cses->cb_sec;
1425 atomic_set(&new->se_ref, 0);
1426 idx = hash_sessionid(&new->se_sessionid);
1427 list_add(&new->se_hash, &nn->sessionid_hashtbl[idx]);
1428 spin_lock(&clp->cl_lock);
1429 list_add(&new->se_perclnt, &clp->cl_sessions);
1430 spin_unlock(&clp->cl_lock);
1432 if (cses->flags & SESSION4_BACK_CHAN) {
1433 struct sockaddr *sa = svc_addr(rqstp);
1435 * This is a little silly; with sessions there's no real
1436 * use for the callback address. Use the peer address
1437 * as a reasonable default for now, but consider fixing
1438 * the rpc client not to require an address in the
1441 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
1442 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1446 /* caller must hold client_lock */
1447 static struct nfsd4_session *
1448 __find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net)
1450 struct nfsd4_session *elem;
1452 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1454 lockdep_assert_held(&nn->client_lock);
1456 dump_sessionid(__func__, sessionid);
1457 idx = hash_sessionid(sessionid);
1458 /* Search in the appropriate list */
1459 list_for_each_entry(elem, &nn->sessionid_hashtbl[idx], se_hash) {
1460 if (!memcmp(elem->se_sessionid.data, sessionid->data,
1461 NFS4_MAX_SESSIONID_LEN)) {
1466 dprintk("%s: session not found\n", __func__);
1470 static struct nfsd4_session *
1471 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net,
1474 struct nfsd4_session *session;
1475 __be32 status = nfserr_badsession;
1477 session = __find_in_sessionid_hashtbl(sessionid, net);
1480 status = nfsd4_get_session_locked(session);
1488 /* caller must hold client_lock */
1490 unhash_session(struct nfsd4_session *ses)
1492 struct nfs4_client *clp = ses->se_client;
1493 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1495 lockdep_assert_held(&nn->client_lock);
1497 list_del(&ses->se_hash);
1498 spin_lock(&ses->se_client->cl_lock);
1499 list_del(&ses->se_perclnt);
1500 spin_unlock(&ses->se_client->cl_lock);
1503 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1505 STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
1507 if (clid->cl_boot == nn->boot_time)
1509 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1510 clid->cl_boot, clid->cl_id, nn->boot_time);
1515 * XXX Should we use a slab cache ?
1516 * This type of memory management is somewhat inefficient, but we use it
1517 * anyway since SETCLIENTID is not a common operation.
1519 static struct nfs4_client *alloc_client(struct xdr_netobj name)
1521 struct nfs4_client *clp;
1524 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
1527 clp->cl_name.data = kmemdup(name.data, name.len, GFP_KERNEL);
1528 if (clp->cl_name.data == NULL)
1530 clp->cl_ownerstr_hashtbl = kmalloc(sizeof(struct list_head) *
1531 OWNER_HASH_SIZE, GFP_KERNEL);
1532 if (!clp->cl_ownerstr_hashtbl)
1533 goto err_no_hashtbl;
1534 for (i = 0; i < OWNER_HASH_SIZE; i++)
1535 INIT_LIST_HEAD(&clp->cl_ownerstr_hashtbl[i]);
1536 clp->cl_name.len = name.len;
1537 INIT_LIST_HEAD(&clp->cl_sessions);
1538 idr_init(&clp->cl_stateids);
1539 atomic_set(&clp->cl_refcount, 0);
1540 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1541 INIT_LIST_HEAD(&clp->cl_idhash);
1542 INIT_LIST_HEAD(&clp->cl_openowners);
1543 INIT_LIST_HEAD(&clp->cl_delegations);
1544 INIT_LIST_HEAD(&clp->cl_lru);
1545 INIT_LIST_HEAD(&clp->cl_callbacks);
1546 INIT_LIST_HEAD(&clp->cl_revoked);
1547 spin_lock_init(&clp->cl_lock);
1548 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1551 kfree(clp->cl_name.data);
1558 free_client(struct nfs4_client *clp)
1560 while (!list_empty(&clp->cl_sessions)) {
1561 struct nfsd4_session *ses;
1562 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
1564 list_del(&ses->se_perclnt);
1565 WARN_ON_ONCE(atomic_read(&ses->se_ref));
1568 rpc_destroy_wait_queue(&clp->cl_cb_waitq);
1569 free_svc_cred(&clp->cl_cred);
1570 kfree(clp->cl_ownerstr_hashtbl);
1571 kfree(clp->cl_name.data);
1572 idr_destroy(&clp->cl_stateids);
1576 /* must be called under the client_lock */
1578 unhash_client_locked(struct nfs4_client *clp)
1580 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1581 struct nfsd4_session *ses;
1583 lockdep_assert_held(&nn->client_lock);
1585 /* Mark the client as expired! */
1587 /* Make it invisible */
1588 if (!list_empty(&clp->cl_idhash)) {
1589 list_del_init(&clp->cl_idhash);
1590 if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
1591 rb_erase(&clp->cl_namenode, &nn->conf_name_tree);
1593 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1595 list_del_init(&clp->cl_lru);
1596 spin_lock(&clp->cl_lock);
1597 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1598 list_del_init(&ses->se_hash);
1599 spin_unlock(&clp->cl_lock);
1603 unhash_client(struct nfs4_client *clp)
1605 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1607 spin_lock(&nn->client_lock);
1608 unhash_client_locked(clp);
1609 spin_unlock(&nn->client_lock);
1612 static __be32 mark_client_expired_locked(struct nfs4_client *clp)
1614 if (atomic_read(&clp->cl_refcount))
1615 return nfserr_jukebox;
1616 unhash_client_locked(clp);
1621 __destroy_client(struct nfs4_client *clp)
1623 struct nfs4_openowner *oo;
1624 struct nfs4_delegation *dp;
1625 struct list_head reaplist;
1627 INIT_LIST_HEAD(&reaplist);
1628 spin_lock(&state_lock);
1629 while (!list_empty(&clp->cl_delegations)) {
1630 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1631 unhash_delegation_locked(dp);
1632 list_add(&dp->dl_recall_lru, &reaplist);
1634 spin_unlock(&state_lock);
1635 while (!list_empty(&reaplist)) {
1636 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1637 list_del_init(&dp->dl_recall_lru);
1638 nfs4_put_stid(&dp->dl_stid);
1640 while (!list_empty(&clp->cl_revoked)) {
1641 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1642 list_del_init(&dp->dl_recall_lru);
1643 nfs4_put_stid(&dp->dl_stid);
1645 while (!list_empty(&clp->cl_openowners)) {
1646 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1647 atomic_inc(&oo->oo_owner.so_count);
1648 release_openowner(oo);
1650 nfsd4_shutdown_callback(clp);
1651 if (clp->cl_cb_conn.cb_xprt)
1652 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1657 destroy_client(struct nfs4_client *clp)
1660 __destroy_client(clp);
1663 static void expire_client(struct nfs4_client *clp)
1666 nfsd4_client_record_remove(clp);
1667 __destroy_client(clp);
1670 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1672 memcpy(target->cl_verifier.data, source->data,
1673 sizeof(target->cl_verifier.data));
1676 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1678 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1679 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1682 static int copy_cred(struct svc_cred *target, struct svc_cred *source)
1684 if (source->cr_principal) {
1685 target->cr_principal =
1686 kstrdup(source->cr_principal, GFP_KERNEL);
1687 if (target->cr_principal == NULL)
1690 target->cr_principal = NULL;
1691 target->cr_flavor = source->cr_flavor;
1692 target->cr_uid = source->cr_uid;
1693 target->cr_gid = source->cr_gid;
1694 target->cr_group_info = source->cr_group_info;
1695 get_group_info(target->cr_group_info);
1696 target->cr_gss_mech = source->cr_gss_mech;
1697 if (source->cr_gss_mech)
1698 gss_mech_get(source->cr_gss_mech);
1703 compare_blob(const struct xdr_netobj *o1, const struct xdr_netobj *o2)
1707 res = o1->len - o2->len;
1710 return (long long)memcmp(o1->data, o2->data, o1->len);
1713 static int same_name(const char *n1, const char *n2)
1715 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1719 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1721 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1725 same_clid(clientid_t *cl1, clientid_t *cl2)
1727 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1730 static bool groups_equal(struct group_info *g1, struct group_info *g2)
1734 if (g1->ngroups != g2->ngroups)
1736 for (i=0; i<g1->ngroups; i++)
1737 if (!gid_eq(GROUP_AT(g1, i), GROUP_AT(g2, i)))
1743 * RFC 3530 language requires clid_inuse be returned when the
1744 * "principal" associated with a requests differs from that previously
1745 * used. We use uid, gid's, and gss principal string as our best
1746 * approximation. We also don't want to allow non-gss use of a client
1747 * established using gss: in theory cr_principal should catch that
1748 * change, but in practice cr_principal can be null even in the gss case
1749 * since gssd doesn't always pass down a principal string.
1751 static bool is_gss_cred(struct svc_cred *cr)
1753 /* Is cr_flavor one of the gss "pseudoflavors"?: */
1754 return (cr->cr_flavor > RPC_AUTH_MAXFLAVOR);
1759 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1761 if ((is_gss_cred(cr1) != is_gss_cred(cr2))
1762 || (!uid_eq(cr1->cr_uid, cr2->cr_uid))
1763 || (!gid_eq(cr1->cr_gid, cr2->cr_gid))
1764 || !groups_equal(cr1->cr_group_info, cr2->cr_group_info))
1766 if (cr1->cr_principal == cr2->cr_principal)
1768 if (!cr1->cr_principal || !cr2->cr_principal)
1770 return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
1773 static bool svc_rqst_integrity_protected(struct svc_rqst *rqstp)
1775 struct svc_cred *cr = &rqstp->rq_cred;
1778 if (!cr->cr_gss_mech)
1780 service = gss_pseudoflavor_to_service(cr->cr_gss_mech, cr->cr_flavor);
1781 return service == RPC_GSS_SVC_INTEGRITY ||
1782 service == RPC_GSS_SVC_PRIVACY;
1785 static bool mach_creds_match(struct nfs4_client *cl, struct svc_rqst *rqstp)
1787 struct svc_cred *cr = &rqstp->rq_cred;
1789 if (!cl->cl_mach_cred)
1791 if (cl->cl_cred.cr_gss_mech != cr->cr_gss_mech)
1793 if (!svc_rqst_integrity_protected(rqstp))
1795 if (!cr->cr_principal)
1797 return 0 == strcmp(cl->cl_cred.cr_principal, cr->cr_principal);
1800 static void gen_confirm(struct nfs4_client *clp, struct nfsd_net *nn)
1805 * This is opaque to client, so no need to byte-swap. Use
1806 * __force to keep sparse happy
1808 verf[0] = (__force __be32)get_seconds();
1809 verf[1] = (__force __be32)nn->clientid_counter;
1810 memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
1813 static void gen_clid(struct nfs4_client *clp, struct nfsd_net *nn)
1815 clp->cl_clientid.cl_boot = nn->boot_time;
1816 clp->cl_clientid.cl_id = nn->clientid_counter++;
1817 gen_confirm(clp, nn);
1820 static struct nfs4_stid *
1821 find_stateid_locked(struct nfs4_client *cl, stateid_t *t)
1823 struct nfs4_stid *ret;
1825 ret = idr_find(&cl->cl_stateids, t->si_opaque.so_id);
1826 if (!ret || !ret->sc_type)
1831 static struct nfs4_stid *
1832 find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
1834 struct nfs4_stid *s;
1836 spin_lock(&cl->cl_lock);
1837 s = find_stateid_locked(cl, t);
1839 if (typemask & s->sc_type)
1840 atomic_inc(&s->sc_count);
1844 spin_unlock(&cl->cl_lock);
1848 static struct nfs4_client *create_client(struct xdr_netobj name,
1849 struct svc_rqst *rqstp, nfs4_verifier *verf)
1851 struct nfs4_client *clp;
1852 struct sockaddr *sa = svc_addr(rqstp);
1854 struct net *net = SVC_NET(rqstp);
1856 clp = alloc_client(name);
1860 ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1865 INIT_WORK(&clp->cl_cb_null.cb_work, nfsd4_run_cb_null);
1866 clp->cl_time = get_seconds();
1867 clear_bit(0, &clp->cl_cb_slot_busy);
1868 copy_verf(clp, verf);
1869 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1870 clp->cl_cb_session = NULL;
1876 add_clp_to_name_tree(struct nfs4_client *new_clp, struct rb_root *root)
1878 struct rb_node **new = &(root->rb_node), *parent = NULL;
1879 struct nfs4_client *clp;
1882 clp = rb_entry(*new, struct nfs4_client, cl_namenode);
1885 if (compare_blob(&clp->cl_name, &new_clp->cl_name) > 0)
1886 new = &((*new)->rb_left);
1888 new = &((*new)->rb_right);
1891 rb_link_node(&new_clp->cl_namenode, parent, new);
1892 rb_insert_color(&new_clp->cl_namenode, root);
1895 static struct nfs4_client *
1896 find_clp_in_name_tree(struct xdr_netobj *name, struct rb_root *root)
1899 struct rb_node *node = root->rb_node;
1900 struct nfs4_client *clp;
1903 clp = rb_entry(node, struct nfs4_client, cl_namenode);
1904 cmp = compare_blob(&clp->cl_name, name);
1906 node = node->rb_left;
1908 node = node->rb_right;
1916 add_to_unconfirmed(struct nfs4_client *clp)
1918 unsigned int idhashval;
1919 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1921 lockdep_assert_held(&nn->client_lock);
1923 clear_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1924 add_clp_to_name_tree(clp, &nn->unconf_name_tree);
1925 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1926 list_add(&clp->cl_idhash, &nn->unconf_id_hashtbl[idhashval]);
1927 renew_client_locked(clp);
1931 move_to_confirmed(struct nfs4_client *clp)
1933 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1934 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1936 lockdep_assert_held(&nn->client_lock);
1938 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1939 list_move(&clp->cl_idhash, &nn->conf_id_hashtbl[idhashval]);
1940 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1941 add_clp_to_name_tree(clp, &nn->conf_name_tree);
1942 set_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1943 renew_client_locked(clp);
1946 static struct nfs4_client *
1947 find_client_in_id_table(struct list_head *tbl, clientid_t *clid, bool sessions)
1949 struct nfs4_client *clp;
1950 unsigned int idhashval = clientid_hashval(clid->cl_id);
1952 list_for_each_entry(clp, &tbl[idhashval], cl_idhash) {
1953 if (same_clid(&clp->cl_clientid, clid)) {
1954 if ((bool)clp->cl_minorversion != sessions)
1956 renew_client_locked(clp);
1963 static struct nfs4_client *
1964 find_confirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1966 struct list_head *tbl = nn->conf_id_hashtbl;
1968 lockdep_assert_held(&nn->client_lock);
1969 return find_client_in_id_table(tbl, clid, sessions);
1972 static struct nfs4_client *
1973 find_unconfirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1975 struct list_head *tbl = nn->unconf_id_hashtbl;
1977 lockdep_assert_held(&nn->client_lock);
1978 return find_client_in_id_table(tbl, clid, sessions);
1981 static bool clp_used_exchangeid(struct nfs4_client *clp)
1983 return clp->cl_exchange_flags != 0;
1986 static struct nfs4_client *
1987 find_confirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1989 lockdep_assert_held(&nn->client_lock);
1990 return find_clp_in_name_tree(name, &nn->conf_name_tree);
1993 static struct nfs4_client *
1994 find_unconfirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1996 lockdep_assert_held(&nn->client_lock);
1997 return find_clp_in_name_tree(name, &nn->unconf_name_tree);
2001 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
2003 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
2004 struct sockaddr *sa = svc_addr(rqstp);
2005 u32 scopeid = rpc_get_scope_id(sa);
2006 unsigned short expected_family;
2008 /* Currently, we only support tcp and tcp6 for the callback channel */
2009 if (se->se_callback_netid_len == 3 &&
2010 !memcmp(se->se_callback_netid_val, "tcp", 3))
2011 expected_family = AF_INET;
2012 else if (se->se_callback_netid_len == 4 &&
2013 !memcmp(se->se_callback_netid_val, "tcp6", 4))
2014 expected_family = AF_INET6;
2018 conn->cb_addrlen = rpc_uaddr2sockaddr(clp->net, se->se_callback_addr_val,
2019 se->se_callback_addr_len,
2020 (struct sockaddr *)&conn->cb_addr,
2021 sizeof(conn->cb_addr));
2023 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
2026 if (conn->cb_addr.ss_family == AF_INET6)
2027 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
2029 conn->cb_prog = se->se_callback_prog;
2030 conn->cb_ident = se->se_callback_ident;
2031 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
2034 conn->cb_addr.ss_family = AF_UNSPEC;
2035 conn->cb_addrlen = 0;
2036 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
2037 "will not receive delegations\n",
2038 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
2044 * Cache a reply. nfsd4_check_resp_size() has bounded the cache size.
2047 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
2049 struct xdr_buf *buf = resp->xdr.buf;
2050 struct nfsd4_slot *slot = resp->cstate.slot;
2053 dprintk("--> %s slot %p\n", __func__, slot);
2055 slot->sl_opcnt = resp->opcnt;
2056 slot->sl_status = resp->cstate.status;
2058 slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
2059 if (nfsd4_not_cached(resp)) {
2060 slot->sl_datalen = 0;
2063 base = resp->cstate.data_offset;
2064 slot->sl_datalen = buf->len - base;
2065 if (read_bytes_from_xdr_buf(buf, base, slot->sl_data, slot->sl_datalen))
2066 WARN("%s: sessions DRC could not cache compound\n", __func__);
2071 * Encode the replay sequence operation from the slot values.
2072 * If cachethis is FALSE encode the uncached rep error on the next
2073 * operation which sets resp->p and increments resp->opcnt for
2074 * nfs4svc_encode_compoundres.
2078 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
2079 struct nfsd4_compoundres *resp)
2081 struct nfsd4_op *op;
2082 struct nfsd4_slot *slot = resp->cstate.slot;
2084 /* Encode the replayed sequence operation */
2085 op = &args->ops[resp->opcnt - 1];
2086 nfsd4_encode_operation(resp, op);
2088 /* Return nfserr_retry_uncached_rep in next operation. */
2089 if (args->opcnt > 1 && !(slot->sl_flags & NFSD4_SLOT_CACHETHIS)) {
2090 op = &args->ops[resp->opcnt++];
2091 op->status = nfserr_retry_uncached_rep;
2092 nfsd4_encode_operation(resp, op);
2098 * The sequence operation is not cached because we can use the slot and
2102 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
2103 struct nfsd4_sequence *seq)
2105 struct nfsd4_slot *slot = resp->cstate.slot;
2106 struct xdr_stream *xdr = &resp->xdr;
2110 dprintk("--> %s slot %p\n", __func__, slot);
2112 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
2116 p = xdr_reserve_space(xdr, slot->sl_datalen);
2119 return nfserr_serverfault;
2121 xdr_encode_opaque_fixed(p, slot->sl_data, slot->sl_datalen);
2122 xdr_commit_encode(xdr);
2124 resp->opcnt = slot->sl_opcnt;
2125 return slot->sl_status;
2129 * Set the exchange_id flags returned by the server.
2132 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
2134 /* pNFS is not supported */
2135 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
2137 /* Referrals are supported, Migration is not. */
2138 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
2140 /* set the wire flags to return to client. */
2141 clid->flags = new->cl_exchange_flags;
2144 static bool client_has_state(struct nfs4_client *clp)
2147 * Note clp->cl_openowners check isn't quite right: there's no
2148 * need to count owners without stateid's.
2150 * Also note we should probably be using this in 4.0 case too.
2152 return !list_empty(&clp->cl_openowners)
2153 || !list_empty(&clp->cl_delegations)
2154 || !list_empty(&clp->cl_sessions);
2158 nfsd4_exchange_id(struct svc_rqst *rqstp,
2159 struct nfsd4_compound_state *cstate,
2160 struct nfsd4_exchange_id *exid)
2162 struct nfs4_client *conf, *new;
2163 struct nfs4_client *unconf = NULL;
2165 char addr_str[INET6_ADDRSTRLEN];
2166 nfs4_verifier verf = exid->verifier;
2167 struct sockaddr *sa = svc_addr(rqstp);
2168 bool update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
2169 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2171 rpc_ntop(sa, addr_str, sizeof(addr_str));
2172 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
2173 "ip_addr=%s flags %x, spa_how %d\n",
2174 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
2175 addr_str, exid->flags, exid->spa_how);
2177 if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
2178 return nfserr_inval;
2180 switch (exid->spa_how) {
2182 if (!svc_rqst_integrity_protected(rqstp))
2183 return nfserr_inval;
2186 default: /* checked by xdr code */
2189 return nfserr_encr_alg_unsupp;
2192 new = create_client(exid->clname, rqstp, &verf);
2194 return nfserr_jukebox;
2196 /* Cases below refer to rfc 5661 section 18.35.4: */
2197 spin_lock(&nn->client_lock);
2198 conf = find_confirmed_client_by_name(&exid->clname, nn);
2200 bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
2201 bool verfs_match = same_verf(&verf, &conf->cl_verifier);
2204 if (!clp_used_exchangeid(conf)) { /* buggy client */
2205 status = nfserr_inval;
2208 if (!mach_creds_match(conf, rqstp)) {
2209 status = nfserr_wrong_cred;
2212 if (!creds_match) { /* case 9 */
2213 status = nfserr_perm;
2216 if (!verfs_match) { /* case 8 */
2217 status = nfserr_not_same;
2221 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
2224 if (!creds_match) { /* case 3 */
2225 if (client_has_state(conf)) {
2226 status = nfserr_clid_inuse;
2231 if (verfs_match) { /* case 2 */
2232 conf->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
2235 /* case 5, client reboot */
2240 if (update) { /* case 7 */
2241 status = nfserr_noent;
2245 unconf = find_unconfirmed_client_by_name(&exid->clname, nn);
2246 if (unconf) /* case 4, possible retry or client restart */
2247 unhash_client_locked(unconf);
2249 /* case 1 (normal case) */
2252 status = mark_client_expired_locked(conf);
2256 new->cl_minorversion = cstate->minorversion;
2257 new->cl_mach_cred = (exid->spa_how == SP4_MACH_CRED);
2260 add_to_unconfirmed(new);
2263 exid->clientid.cl_boot = conf->cl_clientid.cl_boot;
2264 exid->clientid.cl_id = conf->cl_clientid.cl_id;
2266 exid->seqid = conf->cl_cs_slot.sl_seqid + 1;
2267 nfsd4_set_ex_flags(conf, exid);
2269 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
2270 conf->cl_cs_slot.sl_seqid, conf->cl_exchange_flags);
2274 spin_unlock(&nn->client_lock);
2278 expire_client(unconf);
2283 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
2285 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
2288 /* The slot is in use, and no response has been sent. */
2290 if (seqid == slot_seqid)
2291 return nfserr_jukebox;
2293 return nfserr_seq_misordered;
2295 /* Note unsigned 32-bit arithmetic handles wraparound: */
2296 if (likely(seqid == slot_seqid + 1))
2298 if (seqid == slot_seqid)
2299 return nfserr_replay_cache;
2300 return nfserr_seq_misordered;
2304 * Cache the create session result into the create session single DRC
2305 * slot cache by saving the xdr structure. sl_seqid has been set.
2306 * Do this for solo or embedded create session operations.
2309 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
2310 struct nfsd4_clid_slot *slot, __be32 nfserr)
2312 slot->sl_status = nfserr;
2313 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
2317 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
2318 struct nfsd4_clid_slot *slot)
2320 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
2321 return slot->sl_status;
2324 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
2325 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
2326 1 + /* MIN tag is length with zero, only length */ \
2327 3 + /* version, opcount, opcode */ \
2328 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
2329 /* seqid, slotID, slotID, cache */ \
2330 4 ) * sizeof(__be32))
2332 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
2333 2 + /* verifier: AUTH_NULL, length 0 */\
2335 1 + /* MIN tag is length with zero, only length */ \
2336 3 + /* opcount, opcode, opstatus*/ \
2337 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
2338 /* seqid, slotID, slotID, slotID, status */ \
2339 5 ) * sizeof(__be32))
2341 static __be32 check_forechannel_attrs(struct nfsd4_channel_attrs *ca, struct nfsd_net *nn)
2343 u32 maxrpc = nn->nfsd_serv->sv_max_mesg;
2345 if (ca->maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ)
2346 return nfserr_toosmall;
2347 if (ca->maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ)
2348 return nfserr_toosmall;
2349 ca->headerpadsz = 0;
2350 ca->maxreq_sz = min_t(u32, ca->maxreq_sz, maxrpc);
2351 ca->maxresp_sz = min_t(u32, ca->maxresp_sz, maxrpc);
2352 ca->maxops = min_t(u32, ca->maxops, NFSD_MAX_OPS_PER_COMPOUND);
2353 ca->maxresp_cached = min_t(u32, ca->maxresp_cached,
2354 NFSD_SLOT_CACHE_SIZE + NFSD_MIN_HDR_SEQ_SZ);
2355 ca->maxreqs = min_t(u32, ca->maxreqs, NFSD_MAX_SLOTS_PER_SESSION);
2357 * Note decreasing slot size below client's request may make it
2358 * difficult for client to function correctly, whereas
2359 * decreasing the number of slots will (just?) affect
2360 * performance. When short on memory we therefore prefer to
2361 * decrease number of slots instead of their size. Clients that
2362 * request larger slots than they need will get poor results:
2364 ca->maxreqs = nfsd4_get_drc_mem(ca);
2366 return nfserr_jukebox;
2371 #define NFSD_CB_MAX_REQ_SZ ((NFS4_enc_cb_recall_sz + \
2372 RPC_MAX_HEADER_WITH_AUTH) * sizeof(__be32))
2373 #define NFSD_CB_MAX_RESP_SZ ((NFS4_dec_cb_recall_sz + \
2374 RPC_MAX_REPHEADER_WITH_AUTH) * sizeof(__be32))
2376 static __be32 check_backchannel_attrs(struct nfsd4_channel_attrs *ca)
2378 ca->headerpadsz = 0;
2381 * These RPC_MAX_HEADER macros are overkill, especially since we
2382 * don't even do gss on the backchannel yet. But this is still
2383 * less than 1k. Tighten up this estimate in the unlikely event
2384 * it turns out to be a problem for some client:
2386 if (ca->maxreq_sz < NFSD_CB_MAX_REQ_SZ)
2387 return nfserr_toosmall;
2388 if (ca->maxresp_sz < NFSD_CB_MAX_RESP_SZ)
2389 return nfserr_toosmall;
2390 ca->maxresp_cached = 0;
2392 return nfserr_toosmall;
2397 static __be32 nfsd4_check_cb_sec(struct nfsd4_cb_sec *cbs)
2399 switch (cbs->flavor) {
2405 * GSS case: the spec doesn't allow us to return this
2406 * error. But it also doesn't allow us not to support
2408 * I'd rather this fail hard than return some error the
2409 * client might think it can already handle:
2411 return nfserr_encr_alg_unsupp;
2416 nfsd4_create_session(struct svc_rqst *rqstp,
2417 struct nfsd4_compound_state *cstate,
2418 struct nfsd4_create_session *cr_ses)
2420 struct sockaddr *sa = svc_addr(rqstp);
2421 struct nfs4_client *conf, *unconf;
2422 struct nfs4_client *old = NULL;
2423 struct nfsd4_session *new;
2424 struct nfsd4_conn *conn;
2425 struct nfsd4_clid_slot *cs_slot = NULL;
2427 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2429 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
2430 return nfserr_inval;
2431 status = nfsd4_check_cb_sec(&cr_ses->cb_sec);
2434 status = check_forechannel_attrs(&cr_ses->fore_channel, nn);
2437 status = check_backchannel_attrs(&cr_ses->back_channel);
2439 goto out_release_drc_mem;
2440 status = nfserr_jukebox;
2441 new = alloc_session(&cr_ses->fore_channel, &cr_ses->back_channel);
2443 goto out_release_drc_mem;
2444 conn = alloc_conn_from_crses(rqstp, cr_ses);
2446 goto out_free_session;
2448 spin_lock(&nn->client_lock);
2449 unconf = find_unconfirmed_client(&cr_ses->clientid, true, nn);
2450 conf = find_confirmed_client(&cr_ses->clientid, true, nn);
2451 WARN_ON_ONCE(conf && unconf);
2454 status = nfserr_wrong_cred;
2455 if (!mach_creds_match(conf, rqstp))
2457 cs_slot = &conf->cl_cs_slot;
2458 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
2459 if (status == nfserr_replay_cache) {
2460 status = nfsd4_replay_create_session(cr_ses, cs_slot);
2462 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
2463 status = nfserr_seq_misordered;
2466 } else if (unconf) {
2467 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
2468 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
2469 status = nfserr_clid_inuse;
2472 status = nfserr_wrong_cred;
2473 if (!mach_creds_match(unconf, rqstp))
2475 cs_slot = &unconf->cl_cs_slot;
2476 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
2478 /* an unconfirmed replay returns misordered */
2479 status = nfserr_seq_misordered;
2482 old = find_confirmed_client_by_name(&unconf->cl_name, nn);
2484 status = mark_client_expired_locked(old);
2490 move_to_confirmed(unconf);
2493 status = nfserr_stale_clientid;
2498 * We do not support RDMA or persistent sessions
2500 cr_ses->flags &= ~SESSION4_PERSIST;
2501 cr_ses->flags &= ~SESSION4_RDMA;
2503 init_session(rqstp, new, conf, cr_ses);
2504 nfsd4_get_session_locked(new);
2506 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
2507 NFS4_MAX_SESSIONID_LEN);
2508 cs_slot->sl_seqid++;
2509 cr_ses->seqid = cs_slot->sl_seqid;
2511 /* cache solo and embedded create sessions under the client_lock */
2512 nfsd4_cache_create_session(cr_ses, cs_slot, status);
2513 spin_unlock(&nn->client_lock);
2514 /* init connection and backchannel */
2515 nfsd4_init_conn(rqstp, conn, new);
2516 nfsd4_put_session(new);
2521 spin_unlock(&nn->client_lock);
2526 __free_session(new);
2527 out_release_drc_mem:
2528 nfsd4_put_drc_mem(&cr_ses->fore_channel);
2532 static __be32 nfsd4_map_bcts_dir(u32 *dir)
2535 case NFS4_CDFC4_FORE:
2536 case NFS4_CDFC4_BACK:
2538 case NFS4_CDFC4_FORE_OR_BOTH:
2539 case NFS4_CDFC4_BACK_OR_BOTH:
2540 *dir = NFS4_CDFC4_BOTH;
2543 return nfserr_inval;
2546 __be32 nfsd4_backchannel_ctl(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_backchannel_ctl *bc)
2548 struct nfsd4_session *session = cstate->session;
2549 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2552 status = nfsd4_check_cb_sec(&bc->bc_cb_sec);
2555 spin_lock(&nn->client_lock);
2556 session->se_cb_prog = bc->bc_cb_program;
2557 session->se_cb_sec = bc->bc_cb_sec;
2558 spin_unlock(&nn->client_lock);
2560 nfsd4_probe_callback(session->se_client);
2565 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
2566 struct nfsd4_compound_state *cstate,
2567 struct nfsd4_bind_conn_to_session *bcts)
2570 struct nfsd4_conn *conn;
2571 struct nfsd4_session *session;
2572 struct net *net = SVC_NET(rqstp);
2573 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2575 if (!nfsd4_last_compound_op(rqstp))
2576 return nfserr_not_only_op;
2577 spin_lock(&nn->client_lock);
2578 session = find_in_sessionid_hashtbl(&bcts->sessionid, net, &status);
2579 spin_unlock(&nn->client_lock);
2581 goto out_no_session;
2582 status = nfserr_wrong_cred;
2583 if (!mach_creds_match(session->se_client, rqstp))
2585 status = nfsd4_map_bcts_dir(&bcts->dir);
2588 conn = alloc_conn(rqstp, bcts->dir);
2589 status = nfserr_jukebox;
2592 nfsd4_init_conn(rqstp, conn, session);
2595 nfsd4_put_session(session);
2600 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
2604 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
2608 nfsd4_destroy_session(struct svc_rqst *r,
2609 struct nfsd4_compound_state *cstate,
2610 struct nfsd4_destroy_session *sessionid)
2612 struct nfsd4_session *ses;
2614 int ref_held_by_me = 0;
2615 struct net *net = SVC_NET(r);
2616 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2618 status = nfserr_not_only_op;
2619 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
2620 if (!nfsd4_last_compound_op(r))
2624 dump_sessionid(__func__, &sessionid->sessionid);
2625 spin_lock(&nn->client_lock);
2626 ses = find_in_sessionid_hashtbl(&sessionid->sessionid, net, &status);
2628 goto out_client_lock;
2629 status = nfserr_wrong_cred;
2630 if (!mach_creds_match(ses->se_client, r))
2631 goto out_put_session;
2632 status = mark_session_dead_locked(ses, 1 + ref_held_by_me);
2634 goto out_put_session;
2635 unhash_session(ses);
2636 spin_unlock(&nn->client_lock);
2638 nfsd4_probe_callback_sync(ses->se_client);
2640 spin_lock(&nn->client_lock);
2643 nfsd4_put_session_locked(ses);
2645 spin_unlock(&nn->client_lock);
2650 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
2652 struct nfsd4_conn *c;
2654 list_for_each_entry(c, &s->se_conns, cn_persession) {
2655 if (c->cn_xprt == xpt) {
2662 static __be32 nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
2664 struct nfs4_client *clp = ses->se_client;
2665 struct nfsd4_conn *c;
2666 __be32 status = nfs_ok;
2669 spin_lock(&clp->cl_lock);
2670 c = __nfsd4_find_conn(new->cn_xprt, ses);
2673 status = nfserr_conn_not_bound_to_session;
2674 if (clp->cl_mach_cred)
2676 __nfsd4_hash_conn(new, ses);
2677 spin_unlock(&clp->cl_lock);
2678 ret = nfsd4_register_conn(new);
2680 /* oops; xprt is already down: */
2681 nfsd4_conn_lost(&new->cn_xpt_user);
2684 spin_unlock(&clp->cl_lock);
2689 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
2691 struct nfsd4_compoundargs *args = rqstp->rq_argp;
2693 return args->opcnt > session->se_fchannel.maxops;
2696 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
2697 struct nfsd4_session *session)
2699 struct xdr_buf *xb = &rqstp->rq_arg;
2701 return xb->len > session->se_fchannel.maxreq_sz;
2705 nfsd4_sequence(struct svc_rqst *rqstp,
2706 struct nfsd4_compound_state *cstate,
2707 struct nfsd4_sequence *seq)
2709 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2710 struct xdr_stream *xdr = &resp->xdr;
2711 struct nfsd4_session *session;
2712 struct nfs4_client *clp;
2713 struct nfsd4_slot *slot;
2714 struct nfsd4_conn *conn;
2717 struct net *net = SVC_NET(rqstp);
2718 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2720 if (resp->opcnt != 1)
2721 return nfserr_sequence_pos;
2724 * Will be either used or freed by nfsd4_sequence_check_conn
2727 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
2729 return nfserr_jukebox;
2731 spin_lock(&nn->client_lock);
2732 session = find_in_sessionid_hashtbl(&seq->sessionid, net, &status);
2734 goto out_no_session;
2735 clp = session->se_client;
2737 status = nfserr_too_many_ops;
2738 if (nfsd4_session_too_many_ops(rqstp, session))
2739 goto out_put_session;
2741 status = nfserr_req_too_big;
2742 if (nfsd4_request_too_big(rqstp, session))
2743 goto out_put_session;
2745 status = nfserr_badslot;
2746 if (seq->slotid >= session->se_fchannel.maxreqs)
2747 goto out_put_session;
2749 slot = session->se_slots[seq->slotid];
2750 dprintk("%s: slotid %d\n", __func__, seq->slotid);
2752 /* We do not negotiate the number of slots yet, so set the
2753 * maxslots to the session maxreqs which is used to encode
2754 * sr_highest_slotid and the sr_target_slot id to maxslots */
2755 seq->maxslots = session->se_fchannel.maxreqs;
2757 status = check_slot_seqid(seq->seqid, slot->sl_seqid,
2758 slot->sl_flags & NFSD4_SLOT_INUSE);
2759 if (status == nfserr_replay_cache) {
2760 status = nfserr_seq_misordered;
2761 if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
2762 goto out_put_session;
2763 cstate->slot = slot;
2764 cstate->session = session;
2766 /* Return the cached reply status and set cstate->status
2767 * for nfsd4_proc_compound processing */
2768 status = nfsd4_replay_cache_entry(resp, seq);
2769 cstate->status = nfserr_replay_cache;
2773 goto out_put_session;
2775 status = nfsd4_sequence_check_conn(conn, session);
2778 goto out_put_session;
2780 buflen = (seq->cachethis) ?
2781 session->se_fchannel.maxresp_cached :
2782 session->se_fchannel.maxresp_sz;
2783 status = (seq->cachethis) ? nfserr_rep_too_big_to_cache :
2785 if (xdr_restrict_buflen(xdr, buflen - rqstp->rq_auth_slack))
2786 goto out_put_session;
2787 svc_reserve(rqstp, buflen);
2790 /* Success! bump slot seqid */
2791 slot->sl_seqid = seq->seqid;
2792 slot->sl_flags |= NFSD4_SLOT_INUSE;
2794 slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
2796 slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
2798 cstate->slot = slot;
2799 cstate->session = session;
2803 switch (clp->cl_cb_state) {
2805 seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
2807 case NFSD4_CB_FAULT:
2808 seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
2811 seq->status_flags = 0;
2813 if (!list_empty(&clp->cl_revoked))
2814 seq->status_flags |= SEQ4_STATUS_RECALLABLE_STATE_REVOKED;
2818 spin_unlock(&nn->client_lock);
2821 nfsd4_put_session_locked(session);
2822 goto out_no_session;
2826 nfsd4_sequence_done(struct nfsd4_compoundres *resp)
2828 struct nfsd4_compound_state *cs = &resp->cstate;
2830 if (nfsd4_has_session(cs)) {
2831 if (cs->status != nfserr_replay_cache) {
2832 nfsd4_store_cache_entry(resp);
2833 cs->slot->sl_flags &= ~NFSD4_SLOT_INUSE;
2835 /* Drop session reference that was taken in nfsd4_sequence() */
2836 nfsd4_put_session(cs->session);
2838 put_client_renew(cs->clp);
2842 nfsd4_destroy_clientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_destroy_clientid *dc)
2844 struct nfs4_client *conf, *unconf;
2845 struct nfs4_client *clp = NULL;
2847 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2849 spin_lock(&nn->client_lock);
2850 unconf = find_unconfirmed_client(&dc->clientid, true, nn);
2851 conf = find_confirmed_client(&dc->clientid, true, nn);
2852 WARN_ON_ONCE(conf && unconf);
2855 if (client_has_state(conf)) {
2856 status = nfserr_clientid_busy;
2859 status = mark_client_expired_locked(conf);
2866 status = nfserr_stale_clientid;
2869 if (!mach_creds_match(clp, rqstp)) {
2871 status = nfserr_wrong_cred;
2874 unhash_client_locked(clp);
2876 spin_unlock(&nn->client_lock);
2883 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
2887 if (rc->rca_one_fs) {
2888 if (!cstate->current_fh.fh_dentry)
2889 return nfserr_nofilehandle;
2891 * We don't take advantage of the rca_one_fs case.
2892 * That's OK, it's optional, we can safely ignore it.
2897 status = nfserr_complete_already;
2898 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
2899 &cstate->session->se_client->cl_flags))
2902 status = nfserr_stale_clientid;
2903 if (is_client_expired(cstate->session->se_client))
2905 * The following error isn't really legal.
2906 * But we only get here if the client just explicitly
2907 * destroyed the client. Surely it no longer cares what
2908 * error it gets back on an operation for the dead
2914 nfsd4_client_record_create(cstate->session->se_client);
2920 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2921 struct nfsd4_setclientid *setclid)
2923 struct xdr_netobj clname = setclid->se_name;
2924 nfs4_verifier clverifier = setclid->se_verf;
2925 struct nfs4_client *conf, *new;
2926 struct nfs4_client *unconf = NULL;
2928 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2930 new = create_client(clname, rqstp, &clverifier);
2932 return nfserr_jukebox;
2933 /* Cases below refer to rfc 3530 section 14.2.33: */
2934 spin_lock(&nn->client_lock);
2935 conf = find_confirmed_client_by_name(&clname, nn);
2938 status = nfserr_clid_inuse;
2939 if (clp_used_exchangeid(conf))
2941 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
2942 char addr_str[INET6_ADDRSTRLEN];
2943 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
2945 dprintk("NFSD: setclientid: string in use by client "
2946 "at %s\n", addr_str);
2950 unconf = find_unconfirmed_client_by_name(&clname, nn);
2952 unhash_client_locked(unconf);
2953 if (conf && same_verf(&conf->cl_verifier, &clverifier))
2954 /* case 1: probable callback update */
2955 copy_clid(new, conf);
2956 else /* case 4 (new client) or cases 2, 3 (client reboot): */
2958 new->cl_minorversion = 0;
2959 gen_callback(new, setclid, rqstp);
2960 add_to_unconfirmed(new);
2961 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
2962 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
2963 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
2967 spin_unlock(&nn->client_lock);
2971 expire_client(unconf);
2977 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
2978 struct nfsd4_compound_state *cstate,
2979 struct nfsd4_setclientid_confirm *setclientid_confirm)
2981 struct nfs4_client *conf, *unconf;
2982 struct nfs4_client *old = NULL;
2983 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
2984 clientid_t * clid = &setclientid_confirm->sc_clientid;
2986 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2988 if (STALE_CLIENTID(clid, nn))
2989 return nfserr_stale_clientid;
2991 spin_lock(&nn->client_lock);
2992 conf = find_confirmed_client(clid, false, nn);
2993 unconf = find_unconfirmed_client(clid, false, nn);
2995 * We try hard to give out unique clientid's, so if we get an
2996 * attempt to confirm the same clientid with a different cred,
2997 * there's a bug somewhere. Let's charitably assume it's our
3000 status = nfserr_serverfault;
3001 if (unconf && !same_creds(&unconf->cl_cred, &rqstp->rq_cred))
3003 if (conf && !same_creds(&conf->cl_cred, &rqstp->rq_cred))
3005 /* cases below refer to rfc 3530 section 14.2.34: */
3006 if (!unconf || !same_verf(&confirm, &unconf->cl_confirm)) {
3007 if (conf && !unconf) /* case 2: probable retransmit */
3009 else /* case 4: client hasn't noticed we rebooted yet? */
3010 status = nfserr_stale_clientid;
3014 if (conf) { /* case 1: callback update */
3016 unhash_client_locked(old);
3017 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
3018 } else { /* case 3: normal case; new or rebooted client */
3019 old = find_confirmed_client_by_name(&unconf->cl_name, nn);
3021 status = mark_client_expired_locked(old);
3027 move_to_confirmed(unconf);
3030 get_client_locked(conf);
3031 spin_unlock(&nn->client_lock);
3032 nfsd4_probe_callback(conf);
3033 spin_lock(&nn->client_lock);
3034 put_client_renew_locked(conf);
3036 spin_unlock(&nn->client_lock);
3042 static struct nfs4_file *nfsd4_alloc_file(void)
3044 return kmem_cache_alloc(file_slab, GFP_KERNEL);
3047 /* OPEN Share state helper functions */
3048 static void nfsd4_init_file(struct nfs4_file *fp, struct knfsd_fh *fh)
3050 unsigned int hashval = file_hashval(fh);
3052 lockdep_assert_held(&state_lock);
3054 atomic_set(&fp->fi_ref, 1);
3055 spin_lock_init(&fp->fi_lock);
3056 INIT_LIST_HEAD(&fp->fi_stateids);
3057 INIT_LIST_HEAD(&fp->fi_delegations);
3058 fh_copy_shallow(&fp->fi_fhandle, fh);
3059 fp->fi_had_conflict = false;
3060 fp->fi_lease = NULL;
3061 fp->fi_share_deny = 0;
3062 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
3063 memset(fp->fi_access, 0, sizeof(fp->fi_access));
3064 hlist_add_head(&fp->fi_hash, &file_hashtbl[hashval]);
3068 nfsd4_free_slabs(void)
3070 kmem_cache_destroy(openowner_slab);
3071 kmem_cache_destroy(lockowner_slab);
3072 kmem_cache_destroy(file_slab);
3073 kmem_cache_destroy(stateid_slab);
3074 kmem_cache_destroy(deleg_slab);
3078 nfsd4_init_slabs(void)
3080 openowner_slab = kmem_cache_create("nfsd4_openowners",
3081 sizeof(struct nfs4_openowner), 0, 0, NULL);
3082 if (openowner_slab == NULL)
3084 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
3085 sizeof(struct nfs4_lockowner), 0, 0, NULL);
3086 if (lockowner_slab == NULL)
3087 goto out_free_openowner_slab;
3088 file_slab = kmem_cache_create("nfsd4_files",
3089 sizeof(struct nfs4_file), 0, 0, NULL);
3090 if (file_slab == NULL)
3091 goto out_free_lockowner_slab;
3092 stateid_slab = kmem_cache_create("nfsd4_stateids",
3093 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
3094 if (stateid_slab == NULL)
3095 goto out_free_file_slab;
3096 deleg_slab = kmem_cache_create("nfsd4_delegations",
3097 sizeof(struct nfs4_delegation), 0, 0, NULL);
3098 if (deleg_slab == NULL)
3099 goto out_free_stateid_slab;
3102 out_free_stateid_slab:
3103 kmem_cache_destroy(stateid_slab);
3105 kmem_cache_destroy(file_slab);
3106 out_free_lockowner_slab:
3107 kmem_cache_destroy(lockowner_slab);
3108 out_free_openowner_slab:
3109 kmem_cache_destroy(openowner_slab);
3111 dprintk("nfsd4: out of memory while initializing nfsv4\n");
3115 static void init_nfs4_replay(struct nfs4_replay *rp)
3117 rp->rp_status = nfserr_serverfault;
3119 rp->rp_buf = rp->rp_ibuf;
3120 mutex_init(&rp->rp_mutex);
3123 static void nfsd4_cstate_assign_replay(struct nfsd4_compound_state *cstate,
3124 struct nfs4_stateowner *so)
3126 if (!nfsd4_has_session(cstate)) {
3127 mutex_lock(&so->so_replay.rp_mutex);
3128 cstate->replay_owner = so;
3129 atomic_inc(&so->so_count);
3133 void nfsd4_cstate_clear_replay(struct nfsd4_compound_state *cstate)
3135 struct nfs4_stateowner *so = cstate->replay_owner;
3138 cstate->replay_owner = NULL;
3139 mutex_unlock(&so->so_replay.rp_mutex);
3140 nfs4_put_stateowner(so);
3144 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
3146 struct nfs4_stateowner *sop;
3148 sop = kmem_cache_alloc(slab, GFP_KERNEL);
3152 sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
3153 if (!sop->so_owner.data) {
3154 kmem_cache_free(slab, sop);
3157 sop->so_owner.len = owner->len;
3159 INIT_LIST_HEAD(&sop->so_stateids);
3160 sop->so_client = clp;
3161 init_nfs4_replay(&sop->so_replay);
3162 atomic_set(&sop->so_count, 1);
3166 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
3168 lockdep_assert_held(&clp->cl_lock);
3170 list_add(&oo->oo_owner.so_strhash,
3171 &clp->cl_ownerstr_hashtbl[strhashval]);
3172 list_add(&oo->oo_perclient, &clp->cl_openowners);
3175 static void nfs4_unhash_openowner(struct nfs4_stateowner *so)
3177 unhash_openowner_locked(openowner(so));
3180 static void nfs4_free_openowner(struct nfs4_stateowner *so)
3182 struct nfs4_openowner *oo = openowner(so);
3184 kmem_cache_free(openowner_slab, oo);
3187 static const struct nfs4_stateowner_operations openowner_ops = {
3188 .so_unhash = nfs4_unhash_openowner,
3189 .so_free = nfs4_free_openowner,
3192 static struct nfs4_openowner *
3193 alloc_init_open_stateowner(unsigned int strhashval, struct nfsd4_open *open,
3194 struct nfsd4_compound_state *cstate)
3196 struct nfs4_client *clp = cstate->clp;
3197 struct nfs4_openowner *oo, *ret;
3199 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
3202 oo->oo_owner.so_ops = &openowner_ops;
3203 oo->oo_owner.so_is_open_owner = 1;
3204 oo->oo_owner.so_seqid = open->op_seqid;
3206 if (nfsd4_has_session(cstate))
3207 oo->oo_flags |= NFS4_OO_CONFIRMED;
3209 oo->oo_last_closed_stid = NULL;
3210 INIT_LIST_HEAD(&oo->oo_close_lru);
3211 spin_lock(&clp->cl_lock);
3212 ret = find_openstateowner_str_locked(strhashval, open, clp);
3214 hash_openowner(oo, clp, strhashval);
3217 nfs4_free_openowner(&oo->oo_owner);
3218 spin_unlock(&clp->cl_lock);
3222 static void init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
3223 struct nfs4_openowner *oo = open->op_openowner;
3225 atomic_inc(&stp->st_stid.sc_count);
3226 stp->st_stid.sc_type = NFS4_OPEN_STID;
3227 INIT_LIST_HEAD(&stp->st_locks);
3228 stp->st_stateowner = &oo->oo_owner;
3229 atomic_inc(&stp->st_stateowner->so_count);
3231 stp->st_stid.sc_file = fp;
3232 stp->st_access_bmap = 0;
3233 stp->st_deny_bmap = 0;
3234 stp->st_openstp = NULL;
3235 spin_lock(&oo->oo_owner.so_client->cl_lock);
3236 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
3237 spin_lock(&fp->fi_lock);
3238 list_add(&stp->st_perfile, &fp->fi_stateids);
3239 spin_unlock(&fp->fi_lock);
3240 spin_unlock(&oo->oo_owner.so_client->cl_lock);
3244 * In the 4.0 case we need to keep the owners around a little while to handle
3245 * CLOSE replay. We still do need to release any file access that is held by
3246 * them before returning however.
3249 move_to_close_lru(struct nfs4_ol_stateid *s, struct net *net)
3251 struct nfs4_ol_stateid *last;
3252 struct nfs4_openowner *oo = openowner(s->st_stateowner);
3253 struct nfsd_net *nn = net_generic(s->st_stid.sc_client->net,
3256 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
3259 * We know that we hold one reference via nfsd4_close, and another
3260 * "persistent" reference for the client. If the refcount is higher
3261 * than 2, then there are still calls in progress that are using this
3262 * stateid. We can't put the sc_file reference until they are finished.
3263 * Wait for the refcount to drop to 2. Since it has been unhashed,
3264 * there should be no danger of the refcount going back up again at
3267 wait_event(close_wq, atomic_read(&s->st_stid.sc_count) == 2);
3269 release_all_access(s);
3270 if (s->st_stid.sc_file) {
3271 put_nfs4_file(s->st_stid.sc_file);
3272 s->st_stid.sc_file = NULL;
3275 spin_lock(&nn->client_lock);
3276 last = oo->oo_last_closed_stid;
3277 oo->oo_last_closed_stid = s;
3278 list_move_tail(&oo->oo_close_lru, &nn->close_lru);
3279 oo->oo_time = get_seconds();
3280 spin_unlock(&nn->client_lock);
3282 nfs4_put_stid(&last->st_stid);
3285 /* search file_hashtbl[] for file */
3286 static struct nfs4_file *
3287 find_file_locked(struct knfsd_fh *fh)
3289 unsigned int hashval = file_hashval(fh);
3290 struct nfs4_file *fp;
3292 lockdep_assert_held(&state_lock);
3294 hlist_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
3295 if (nfsd_fh_match(&fp->fi_fhandle, fh)) {
3303 static struct nfs4_file *
3304 find_file(struct knfsd_fh *fh)
3306 struct nfs4_file *fp;
3308 spin_lock(&state_lock);
3309 fp = find_file_locked(fh);
3310 spin_unlock(&state_lock);
3314 static struct nfs4_file *
3315 find_or_add_file(struct nfs4_file *new, struct knfsd_fh *fh)
3317 struct nfs4_file *fp;
3319 spin_lock(&state_lock);
3320 fp = find_file_locked(fh);
3322 nfsd4_init_file(new, fh);
3325 spin_unlock(&state_lock);
3331 * Called to check deny when READ with all zero stateid or
3332 * WRITE with all zero or all one stateid
3335 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
3337 struct nfs4_file *fp;
3338 __be32 ret = nfs_ok;
3340 fp = find_file(¤t_fh->fh_handle);
3343 /* Check for conflicting share reservations */
3344 spin_lock(&fp->fi_lock);
3345 if (fp->fi_share_deny & deny_type)
3346 ret = nfserr_locked;
3347 spin_unlock(&fp->fi_lock);
3352 void nfsd4_prepare_cb_recall(struct nfs4_delegation *dp)
3354 struct nfsd_net *nn = net_generic(dp->dl_stid.sc_client->net,
3357 block_delegations(&dp->dl_stid.sc_file->fi_fhandle);
3360 * We can't do this in nfsd_break_deleg_cb because it is
3361 * already holding inode->i_lock.
3363 * If the dl_time != 0, then we know that it has already been
3364 * queued for a lease break. Don't queue it again.
3366 spin_lock(&state_lock);
3367 if (dp->dl_time == 0) {
3368 dp->dl_time = get_seconds();
3369 list_add_tail(&dp->dl_recall_lru, &nn->del_recall_lru);
3371 spin_unlock(&state_lock);
3374 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
3377 * We're assuming the state code never drops its reference
3378 * without first removing the lease. Since we're in this lease
3379 * callback (and since the lease code is serialized by the kernel
3380 * lock) we know the server hasn't removed the lease yet, we know
3381 * it's safe to take a reference.
3383 atomic_inc(&dp->dl_stid.sc_count);
3384 nfsd4_cb_recall(dp);
3387 /* Called from break_lease() with i_lock held. */
3388 static void nfsd_break_deleg_cb(struct file_lock *fl)
3390 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
3391 struct nfs4_delegation *dp;
3394 WARN(1, "(%p)->fl_owner NULL\n", fl);
3397 if (fp->fi_had_conflict) {
3398 WARN(1, "duplicate break on %p\n", fp);
3402 * We don't want the locks code to timeout the lease for us;
3403 * we'll remove it ourself if a delegation isn't returned
3406 fl->fl_break_time = 0;
3408 spin_lock(&fp->fi_lock);
3409 fp->fi_had_conflict = true;
3411 * If there are no delegations on the list, then we can't count on this
3412 * lease ever being cleaned up. Set the fl_break_time to jiffies so that
3413 * time_out_leases will do it ASAP. The fact that fi_had_conflict is now
3414 * true should keep any new delegations from being hashed.
3416 if (list_empty(&fp->fi_delegations))
3417 fl->fl_break_time = jiffies;
3419 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
3420 nfsd_break_one_deleg(dp);
3421 spin_unlock(&fp->fi_lock);
3425 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
3428 return lease_modify(onlist, arg);
3433 static const struct lock_manager_operations nfsd_lease_mng_ops = {
3434 .lm_break = nfsd_break_deleg_cb,
3435 .lm_change = nfsd_change_deleg_cb,
3438 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
3440 if (nfsd4_has_session(cstate))
3442 if (seqid == so->so_seqid - 1)
3443 return nfserr_replay_me;
3444 if (seqid == so->so_seqid)
3446 return nfserr_bad_seqid;
3449 static __be32 lookup_clientid(clientid_t *clid,
3450 struct nfsd4_compound_state *cstate,
3451 struct nfsd_net *nn)
3453 struct nfs4_client *found;
3456 found = cstate->clp;
3457 if (!same_clid(&found->cl_clientid, clid))
3458 return nfserr_stale_clientid;
3462 if (STALE_CLIENTID(clid, nn))
3463 return nfserr_stale_clientid;
3466 * For v4.1+ we get the client in the SEQUENCE op. If we don't have one
3467 * cached already then we know this is for is for v4.0 and "sessions"
3470 WARN_ON_ONCE(cstate->session);
3471 spin_lock(&nn->client_lock);
3472 found = find_confirmed_client(clid, false, nn);
3474 spin_unlock(&nn->client_lock);
3475 return nfserr_expired;
3477 atomic_inc(&found->cl_refcount);
3478 spin_unlock(&nn->client_lock);
3480 /* Cache the nfs4_client in cstate! */
3481 cstate->clp = found;
3486 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
3487 struct nfsd4_open *open, struct nfsd_net *nn)
3489 clientid_t *clientid = &open->op_clientid;
3490 struct nfs4_client *clp = NULL;
3491 unsigned int strhashval;
3492 struct nfs4_openowner *oo = NULL;
3495 if (STALE_CLIENTID(&open->op_clientid, nn))
3496 return nfserr_stale_clientid;
3498 * In case we need it later, after we've already created the
3499 * file and don't want to risk a further failure:
3501 open->op_file = nfsd4_alloc_file();
3502 if (open->op_file == NULL)
3503 return nfserr_jukebox;
3505 status = lookup_clientid(clientid, cstate, nn);
3510 strhashval = ownerstr_hashval(&open->op_owner);
3511 oo = find_openstateowner_str(strhashval, open, clp);
3512 open->op_openowner = oo;
3516 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
3517 /* Replace unconfirmed owners without checking for replay. */
3518 release_openowner(oo);
3519 open->op_openowner = NULL;
3522 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
3527 oo = alloc_init_open_stateowner(strhashval, open, cstate);
3529 return nfserr_jukebox;
3530 open->op_openowner = oo;
3532 open->op_stp = nfs4_alloc_open_stateid(clp);
3534 return nfserr_jukebox;
3538 static inline __be32
3539 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
3541 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
3542 return nfserr_openmode;
3547 static int share_access_to_flags(u32 share_access)
3549 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
3552 static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
3554 struct nfs4_stid *ret;
3556 ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
3559 return delegstateid(ret);
3562 static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
3564 return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
3565 open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
3569 nfs4_check_deleg(struct nfs4_client *cl, struct nfsd4_open *open,
3570 struct nfs4_delegation **dp)
3573 __be32 status = nfserr_bad_stateid;
3574 struct nfs4_delegation *deleg;
3576 deleg = find_deleg_stateid(cl, &open->op_delegate_stateid);
3579 flags = share_access_to_flags(open->op_share_access);
3580 status = nfs4_check_delegmode(deleg, flags);
3582 nfs4_put_stid(&deleg->dl_stid);
3587 if (!nfsd4_is_deleg_cur(open))
3591 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3595 static struct nfs4_ol_stateid *
3596 nfsd4_find_existing_open(struct nfs4_file *fp, struct nfsd4_open *open)
3598 struct nfs4_ol_stateid *local, *ret = NULL;
3599 struct nfs4_openowner *oo = open->op_openowner;
3601 spin_lock(&fp->fi_lock);
3602 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
3603 /* ignore lock owners */
3604 if (local->st_stateowner->so_is_open_owner == 0)
3606 if (local->st_stateowner == &oo->oo_owner) {
3608 atomic_inc(&ret->st_stid.sc_count);
3612 spin_unlock(&fp->fi_lock);
3616 static inline int nfs4_access_to_access(u32 nfs4_access)
3620 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
3621 flags |= NFSD_MAY_READ;
3622 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
3623 flags |= NFSD_MAY_WRITE;
3627 static inline __be32
3628 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
3629 struct nfsd4_open *open)
3631 struct iattr iattr = {
3632 .ia_valid = ATTR_SIZE,
3635 if (!open->op_truncate)
3637 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
3638 return nfserr_inval;
3639 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
3642 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
3643 struct svc_fh *cur_fh, struct nfs4_ol_stateid *stp,
3644 struct nfsd4_open *open)
3646 struct file *filp = NULL;
3648 int oflag = nfs4_access_to_omode(open->op_share_access);
3649 int access = nfs4_access_to_access(open->op_share_access);
3650 unsigned char old_access_bmap, old_deny_bmap;
3652 spin_lock(&fp->fi_lock);
3655 * Are we trying to set a deny mode that would conflict with
3658 status = nfs4_file_check_deny(fp, open->op_share_deny);
3659 if (status != nfs_ok) {
3660 spin_unlock(&fp->fi_lock);
3664 /* set access to the file */
3665 status = nfs4_file_get_access(fp, open->op_share_access);
3666 if (status != nfs_ok) {
3667 spin_unlock(&fp->fi_lock);
3671 /* Set access bits in stateid */
3672 old_access_bmap = stp->st_access_bmap;
3673 set_access(open->op_share_access, stp);
3675 /* Set new deny mask */
3676 old_deny_bmap = stp->st_deny_bmap;
3677 set_deny(open->op_share_deny, stp);
3678 fp->fi_share_deny |= (open->op_share_deny & NFS4_SHARE_DENY_BOTH);
3680 if (!fp->fi_fds[oflag]) {
3681 spin_unlock(&fp->fi_lock);
3682 status = nfsd_open(rqstp, cur_fh, S_IFREG, access, &filp);
3684 goto out_put_access;
3685 spin_lock(&fp->fi_lock);
3686 if (!fp->fi_fds[oflag]) {
3687 fp->fi_fds[oflag] = filp;
3691 spin_unlock(&fp->fi_lock);
3695 status = nfsd4_truncate(rqstp, cur_fh, open);
3697 goto out_put_access;
3701 stp->st_access_bmap = old_access_bmap;
3702 nfs4_file_put_access(fp, open->op_share_access);
3703 reset_union_bmap_deny(bmap_to_share_mode(old_deny_bmap), stp);
3708 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)
3711 unsigned char old_deny_bmap;
3713 if (!test_access(open->op_share_access, stp))
3714 return nfs4_get_vfs_file(rqstp, fp, cur_fh, stp, open);
3716 /* test and set deny mode */
3717 spin_lock(&fp->fi_lock);
3718 status = nfs4_file_check_deny(fp, open->op_share_deny);
3719 if (status == nfs_ok) {
3720 old_deny_bmap = stp->st_deny_bmap;
3721 set_deny(open->op_share_deny, stp);
3722 fp->fi_share_deny |=
3723 (open->op_share_deny & NFS4_SHARE_DENY_BOTH);
3725 spin_unlock(&fp->fi_lock);
3727 if (status != nfs_ok)
3730 status = nfsd4_truncate(rqstp, cur_fh, open);
3731 if (status != nfs_ok)
3732 reset_union_bmap_deny(old_deny_bmap, stp);
3737 nfs4_set_claim_prev(struct nfsd4_open *open, bool has_session)
3739 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3742 /* Should we give out recallable state?: */
3743 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
3745 if (clp->cl_cb_state == NFSD4_CB_UP)
3748 * In the sessions case, since we don't have to establish a
3749 * separate connection for callbacks, we assume it's OK
3750 * until we hear otherwise:
3752 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
3755 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_file *fp, int flag)
3757 struct file_lock *fl;
3759 fl = locks_alloc_lock();
3762 locks_init_lock(fl);
3763 fl->fl_lmops = &nfsd_lease_mng_ops;
3764 fl->fl_flags = FL_DELEG;
3765 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
3766 fl->fl_end = OFFSET_MAX;
3767 fl->fl_owner = (fl_owner_t)fp;
3768 fl->fl_pid = current->tgid;
3772 static int nfs4_setlease(struct nfs4_delegation *dp)
3774 struct nfs4_file *fp = dp->dl_stid.sc_file;
3775 struct file_lock *fl;
3779 fl = nfs4_alloc_init_lease(fp, NFS4_OPEN_DELEGATE_READ);
3782 filp = find_readable_file(fp);
3784 /* We should always have a readable file here */
3789 status = vfs_setlease(filp, fl->fl_type, &fl);
3791 locks_free_lock(fl);
3794 spin_lock(&state_lock);
3795 spin_lock(&fp->fi_lock);
3796 /* Did the lease get broken before we took the lock? */
3798 if (fp->fi_had_conflict)
3803 atomic_inc(&fp->fi_delegees);
3804 hash_delegation_locked(dp, fp);
3808 fp->fi_deleg_file = filp;
3809 atomic_set(&fp->fi_delegees, 1);
3810 hash_delegation_locked(dp, fp);
3811 spin_unlock(&fp->fi_lock);
3812 spin_unlock(&state_lock);
3815 spin_unlock(&fp->fi_lock);
3816 spin_unlock(&state_lock);
3822 static struct nfs4_delegation *
3823 nfs4_set_delegation(struct nfs4_client *clp, struct svc_fh *fh,
3824 struct nfs4_file *fp)
3827 struct nfs4_delegation *dp;
3829 if (fp->fi_had_conflict)
3830 return ERR_PTR(-EAGAIN);
3832 dp = alloc_init_deleg(clp, fh);
3834 return ERR_PTR(-ENOMEM);
3837 spin_lock(&state_lock);
3838 spin_lock(&fp->fi_lock);
3839 dp->dl_stid.sc_file = fp;
3840 if (!fp->fi_lease) {
3841 spin_unlock(&fp->fi_lock);
3842 spin_unlock(&state_lock);
3843 status = nfs4_setlease(dp);
3846 atomic_inc(&fp->fi_delegees);
3847 if (fp->fi_had_conflict) {
3851 hash_delegation_locked(dp, fp);
3854 spin_unlock(&fp->fi_lock);
3855 spin_unlock(&state_lock);
3858 nfs4_put_stid(&dp->dl_stid);
3859 return ERR_PTR(status);
3864 static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
3866 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3867 if (status == -EAGAIN)
3868 open->op_why_no_deleg = WND4_CONTENTION;
3870 open->op_why_no_deleg = WND4_RESOURCE;
3871 switch (open->op_deleg_want) {
3872 case NFS4_SHARE_WANT_READ_DELEG:
3873 case NFS4_SHARE_WANT_WRITE_DELEG:
3874 case NFS4_SHARE_WANT_ANY_DELEG:
3876 case NFS4_SHARE_WANT_CANCEL:
3877 open->op_why_no_deleg = WND4_CANCELLED;
3879 case NFS4_SHARE_WANT_NO_DELEG:
3886 * Attempt to hand out a delegation.
3888 * Note we don't support write delegations, and won't until the vfs has
3889 * proper support for them.
3892 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open,
3893 struct nfs4_ol_stateid *stp)
3895 struct nfs4_delegation *dp;
3896 struct nfs4_openowner *oo = openowner(stp->st_stateowner);
3897 struct nfs4_client *clp = stp->st_stid.sc_client;
3901 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
3902 open->op_recall = 0;
3903 switch (open->op_claim_type) {
3904 case NFS4_OPEN_CLAIM_PREVIOUS:
3906 open->op_recall = 1;
3907 if (open->op_delegate_type != NFS4_OPEN_DELEGATE_READ)
3910 case NFS4_OPEN_CLAIM_NULL:
3911 case NFS4_OPEN_CLAIM_FH:
3913 * Let's not give out any delegations till everyone's
3914 * had the chance to reclaim theirs....
3916 if (locks_in_grace(clp->net))
3918 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
3921 * Also, if the file was opened for write or
3922 * create, there's a good chance the client's
3923 * about to write to it, resulting in an
3924 * immediate recall (since we don't support
3925 * write delegations):
3927 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
3929 if (open->op_create == NFS4_OPEN_CREATE)
3935 dp = nfs4_set_delegation(clp, fh, stp->st_stid.sc_file);
3939 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
3941 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
3942 STATEID_VAL(&dp->dl_stid.sc_stateid));
3943 open->op_delegate_type = NFS4_OPEN_DELEGATE_READ;
3944 nfs4_put_stid(&dp->dl_stid);
3947 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE;
3948 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
3949 open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE) {
3950 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
3951 open->op_recall = 1;
3954 /* 4.1 client asking for a delegation? */
3955 if (open->op_deleg_want)
3956 nfsd4_open_deleg_none_ext(open, status);
3960 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
3961 struct nfs4_delegation *dp)
3963 if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
3964 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3965 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3966 open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
3967 } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
3968 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3969 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3970 open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
3972 /* Otherwise the client must be confused wanting a delegation
3973 * it already has, therefore we don't return
3974 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
3979 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
3981 struct nfsd4_compoundres *resp = rqstp->rq_resp;
3982 struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
3983 struct nfs4_file *fp = NULL;
3984 struct nfs4_ol_stateid *stp = NULL;
3985 struct nfs4_delegation *dp = NULL;
3989 * Lookup file; if found, lookup stateid and check open request,
3990 * and check for delegations in the process of being recalled.
3991 * If not found, create the nfs4_file struct
3993 fp = find_or_add_file(open->op_file, ¤t_fh->fh_handle);
3994 if (fp != open->op_file) {
3995 status = nfs4_check_deleg(cl, open, &dp);
3998 stp = nfsd4_find_existing_open(fp, open);
4000 open->op_file = NULL;
4001 status = nfserr_bad_stateid;
4002 if (nfsd4_is_deleg_cur(open))
4004 status = nfserr_jukebox;
4008 * OPEN the file, or upgrade an existing OPEN.
4009 * If truncate fails, the OPEN fails.
4012 /* Stateid was found, this is an OPEN upgrade */
4013 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
4018 open->op_stp = NULL;
4019 init_open_stateid(stp, fp, open);
4020 status = nfs4_get_vfs_file(rqstp, fp, current_fh, stp, open);
4022 release_open_stateid(stp);
4026 update_stateid(&stp->st_stid.sc_stateid);
4027 memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4029 if (nfsd4_has_session(&resp->cstate)) {
4030 if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
4031 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
4032 open->op_why_no_deleg = WND4_NOT_WANTED;
4038 * Attempt to hand out a delegation. No error return, because the
4039 * OPEN succeeds even if we fail.
4041 nfs4_open_delegation(current_fh, open, stp);
4045 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
4046 STATEID_VAL(&stp->st_stid.sc_stateid));
4048 /* 4.1 client trying to upgrade/downgrade delegation? */
4049 if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
4050 open->op_deleg_want)
4051 nfsd4_deleg_xgrade_none_ext(open, dp);
4055 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
4056 nfs4_set_claim_prev(open, nfsd4_has_session(&resp->cstate));
4058 * To finish the open response, we just need to set the rflags.
4060 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
4061 if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
4062 !nfsd4_has_session(&resp->cstate))
4063 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
4065 nfs4_put_stid(&dp->dl_stid);
4067 nfs4_put_stid(&stp->st_stid);
4072 void nfsd4_cleanup_open_state(struct nfsd4_compound_state *cstate,
4073 struct nfsd4_open *open, __be32 status)
4075 if (open->op_openowner) {
4076 struct nfs4_stateowner *so = &open->op_openowner->oo_owner;
4078 nfsd4_cstate_assign_replay(cstate, so);
4079 nfs4_put_stateowner(so);
4082 nfsd4_free_file(open->op_file);
4084 nfs4_put_stid(&open->op_stp->st_stid);
4088 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4091 struct nfs4_client *clp;
4093 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4095 dprintk("process_renew(%08x/%08x): starting\n",
4096 clid->cl_boot, clid->cl_id);
4097 status = lookup_clientid(clid, cstate, nn);
4101 status = nfserr_cb_path_down;
4102 if (!list_empty(&clp->cl_delegations)
4103 && clp->cl_cb_state != NFSD4_CB_UP)
4111 nfsd4_end_grace(struct nfsd_net *nn)
4113 /* do nothing if grace period already ended */
4114 if (nn->grace_ended)
4117 dprintk("NFSD: end of grace period\n");
4118 nn->grace_ended = true;
4119 nfsd4_record_grace_done(nn, nn->boot_time);
4120 locks_end_grace(&nn->nfsd4_manager);
4122 * Now that every NFSv4 client has had the chance to recover and
4123 * to see the (possibly new, possibly shorter) lease time, we
4124 * can safely set the next grace time to the current lease time:
4126 nn->nfsd4_grace = nn->nfsd4_lease;
4130 nfs4_laundromat(struct nfsd_net *nn)
4132 struct nfs4_client *clp;
4133 struct nfs4_openowner *oo;
4134 struct nfs4_delegation *dp;
4135 struct nfs4_ol_stateid *stp;
4136 struct list_head *pos, *next, reaplist;
4137 time_t cutoff = get_seconds() - nn->nfsd4_lease;
4138 time_t t, new_timeo = nn->nfsd4_lease;
4140 dprintk("NFSD: laundromat service - starting\n");
4141 nfsd4_end_grace(nn);
4142 INIT_LIST_HEAD(&reaplist);
4143 spin_lock(&nn->client_lock);
4144 list_for_each_safe(pos, next, &nn->client_lru) {
4145 clp = list_entry(pos, struct nfs4_client, cl_lru);
4146 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
4147 t = clp->cl_time - cutoff;
4148 new_timeo = min(new_timeo, t);
4151 if (mark_client_expired_locked(clp)) {
4152 dprintk("NFSD: client in use (clientid %08x)\n",
4153 clp->cl_clientid.cl_id);
4156 list_add(&clp->cl_lru, &reaplist);
4158 spin_unlock(&nn->client_lock);
4159 list_for_each_safe(pos, next, &reaplist) {
4160 clp = list_entry(pos, struct nfs4_client, cl_lru);
4161 dprintk("NFSD: purging unused client (clientid %08x)\n",
4162 clp->cl_clientid.cl_id);
4163 list_del_init(&clp->cl_lru);
4166 spin_lock(&state_lock);
4167 list_for_each_safe(pos, next, &nn->del_recall_lru) {
4168 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4169 if (net_generic(dp->dl_stid.sc_client->net, nfsd_net_id) != nn)
4171 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
4172 t = dp->dl_time - cutoff;
4173 new_timeo = min(new_timeo, t);
4176 unhash_delegation_locked(dp);
4177 list_add(&dp->dl_recall_lru, &reaplist);
4179 spin_unlock(&state_lock);
4180 while (!list_empty(&reaplist)) {
4181 dp = list_first_entry(&reaplist, struct nfs4_delegation,
4183 list_del_init(&dp->dl_recall_lru);
4184 revoke_delegation(dp);
4187 spin_lock(&nn->client_lock);
4188 while (!list_empty(&nn->close_lru)) {
4189 oo = list_first_entry(&nn->close_lru, struct nfs4_openowner,
4191 if (time_after((unsigned long)oo->oo_time,
4192 (unsigned long)cutoff)) {
4193 t = oo->oo_time - cutoff;
4194 new_timeo = min(new_timeo, t);
4197 list_del_init(&oo->oo_close_lru);
4198 stp = oo->oo_last_closed_stid;
4199 oo->oo_last_closed_stid = NULL;
4200 spin_unlock(&nn->client_lock);
4201 nfs4_put_stid(&stp->st_stid);
4202 spin_lock(&nn->client_lock);
4204 spin_unlock(&nn->client_lock);
4206 new_timeo = max_t(time_t, new_timeo, NFSD_LAUNDROMAT_MINTIMEOUT);
4210 static struct workqueue_struct *laundry_wq;
4211 static void laundromat_main(struct work_struct *);
4214 laundromat_main(struct work_struct *laundry)
4217 struct delayed_work *dwork = container_of(laundry, struct delayed_work,
4219 struct nfsd_net *nn = container_of(dwork, struct nfsd_net,
4222 t = nfs4_laundromat(nn);
4223 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
4224 queue_delayed_work(laundry_wq, &nn->laundromat_work, t*HZ);
4227 static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *stp)
4229 if (!nfsd_fh_match(&fhp->fh_handle, &stp->st_stid.sc_file->fi_fhandle))
4230 return nfserr_bad_stateid;
4235 access_permit_read(struct nfs4_ol_stateid *stp)
4237 return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
4238 test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
4239 test_access(NFS4_SHARE_ACCESS_WRITE, stp);
4243 access_permit_write(struct nfs4_ol_stateid *stp)
4245 return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
4246 test_access(NFS4_SHARE_ACCESS_BOTH, stp);
4250 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
4252 __be32 status = nfserr_openmode;
4254 /* For lock stateid's, we test the parent open, not the lock: */
4255 if (stp->st_openstp)
4256 stp = stp->st_openstp;
4257 if ((flags & WR_STATE) && !access_permit_write(stp))
4259 if ((flags & RD_STATE) && !access_permit_read(stp))
4266 static inline __be32
4267 check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
4269 if (ONE_STATEID(stateid) && (flags & RD_STATE))
4271 else if (locks_in_grace(net)) {
4272 /* Answer in remaining cases depends on existence of
4273 * conflicting state; so we must wait out the grace period. */
4274 return nfserr_grace;
4275 } else if (flags & WR_STATE)
4276 return nfs4_share_conflict(current_fh,
4277 NFS4_SHARE_DENY_WRITE);
4278 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
4279 return nfs4_share_conflict(current_fh,
4280 NFS4_SHARE_DENY_READ);
4284 * Allow READ/WRITE during grace period on recovered state only for files
4285 * that are not able to provide mandatory locking.
4288 grace_disallows_io(struct net *net, struct inode *inode)
4290 return locks_in_grace(net) && mandatory_lock(inode);
4293 /* Returns true iff a is later than b: */
4294 static bool stateid_generation_after(stateid_t *a, stateid_t *b)
4296 return (s32)(a->si_generation - b->si_generation) > 0;
4299 static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
4302 * When sessions are used the stateid generation number is ignored
4305 if (has_session && in->si_generation == 0)
4308 if (in->si_generation == ref->si_generation)
4311 /* If the client sends us a stateid from the future, it's buggy: */
4312 if (stateid_generation_after(in, ref))
4313 return nfserr_bad_stateid;
4315 * However, we could see a stateid from the past, even from a
4316 * non-buggy client. For example, if the client sends a lock
4317 * while some IO is outstanding, the lock may bump si_generation
4318 * while the IO is still in flight. The client could avoid that
4319 * situation by waiting for responses on all the IO requests,
4320 * but better performance may result in retrying IO that
4321 * receives an old_stateid error if requests are rarely
4322 * reordered in flight:
4324 return nfserr_old_stateid;
4327 static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
4329 struct nfs4_stid *s;
4330 struct nfs4_ol_stateid *ols;
4331 __be32 status = nfserr_bad_stateid;
4333 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
4335 /* Client debugging aid. */
4336 if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid)) {
4337 char addr_str[INET6_ADDRSTRLEN];
4338 rpc_ntop((struct sockaddr *)&cl->cl_addr, addr_str,
4340 pr_warn_ratelimited("NFSD: client %s testing state ID "
4341 "with incorrect client ID\n", addr_str);
4344 spin_lock(&cl->cl_lock);
4345 s = find_stateid_locked(cl, stateid);
4348 status = check_stateid_generation(stateid, &s->sc_stateid, 1);
4351 switch (s->sc_type) {
4352 case NFS4_DELEG_STID:
4355 case NFS4_REVOKED_DELEG_STID:
4356 status = nfserr_deleg_revoked;
4358 case NFS4_OPEN_STID:
4359 case NFS4_LOCK_STID:
4360 ols = openlockstateid(s);
4361 if (ols->st_stateowner->so_is_open_owner
4362 && !(openowner(ols->st_stateowner)->oo_flags
4363 & NFS4_OO_CONFIRMED))
4364 status = nfserr_bad_stateid;
4369 printk("unknown stateid type %x\n", s->sc_type);
4371 case NFS4_CLOSED_STID:
4372 case NFS4_CLOSED_DELEG_STID:
4373 status = nfserr_bad_stateid;
4376 spin_unlock(&cl->cl_lock);
4381 nfsd4_lookup_stateid(struct nfsd4_compound_state *cstate,
4382 stateid_t *stateid, unsigned char typemask,
4383 struct nfs4_stid **s, struct nfsd_net *nn)
4387 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
4388 return nfserr_bad_stateid;
4389 status = lookup_clientid(&stateid->si_opaque.so_clid, cstate, nn);
4390 if (status == nfserr_stale_clientid) {
4391 if (cstate->session)
4392 return nfserr_bad_stateid;
4393 return nfserr_stale_stateid;
4397 *s = find_stateid_by_type(cstate->clp, stateid, typemask);
4399 return nfserr_bad_stateid;
4404 * Checks for stateid operations
4407 nfs4_preprocess_stateid_op(struct net *net, struct nfsd4_compound_state *cstate,
4408 stateid_t *stateid, int flags, struct file **filpp)
4410 struct nfs4_stid *s;
4411 struct nfs4_ol_stateid *stp = NULL;
4412 struct nfs4_delegation *dp = NULL;
4413 struct svc_fh *current_fh = &cstate->current_fh;
4414 struct inode *ino = current_fh->fh_dentry->d_inode;
4415 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4416 struct file *file = NULL;
4422 if (grace_disallows_io(net, ino))
4423 return nfserr_grace;
4425 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
4426 return check_special_stateids(net, current_fh, stateid, flags);
4428 status = nfsd4_lookup_stateid(cstate, stateid,
4429 NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID,
4433 status = check_stateid_generation(stateid, &s->sc_stateid, nfsd4_has_session(cstate));
4436 switch (s->sc_type) {
4437 case NFS4_DELEG_STID:
4438 dp = delegstateid(s);
4439 status = nfs4_check_delegmode(dp, flags);
4443 file = dp->dl_stid.sc_file->fi_deleg_file;
4446 status = nfserr_serverfault;
4452 case NFS4_OPEN_STID:
4453 case NFS4_LOCK_STID:
4454 stp = openlockstateid(s);
4455 status = nfs4_check_fh(current_fh, stp);
4458 if (stp->st_stateowner->so_is_open_owner
4459 && !(openowner(stp->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
4461 status = nfs4_check_openmode(stp, flags);
4465 struct nfs4_file *fp = stp->st_stid.sc_file;
4467 if (flags & RD_STATE)
4468 file = find_readable_file(fp);
4470 file = find_writeable_file(fp);
4474 status = nfserr_bad_stateid;
4486 * Test if the stateid is valid
4489 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4490 struct nfsd4_test_stateid *test_stateid)
4492 struct nfsd4_test_stateid_id *stateid;
4493 struct nfs4_client *cl = cstate->session->se_client;
4495 list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
4496 stateid->ts_id_status =
4497 nfsd4_validate_stateid(cl, &stateid->ts_id_stateid);
4503 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4504 struct nfsd4_free_stateid *free_stateid)
4506 stateid_t *stateid = &free_stateid->fr_stateid;
4507 struct nfs4_stid *s;
4508 struct nfs4_delegation *dp;
4509 struct nfs4_ol_stateid *stp;
4510 struct nfs4_client *cl = cstate->session->se_client;
4511 __be32 ret = nfserr_bad_stateid;
4513 spin_lock(&cl->cl_lock);
4514 s = find_stateid_locked(cl, stateid);
4517 switch (s->sc_type) {
4518 case NFS4_DELEG_STID:
4519 ret = nfserr_locks_held;
4521 case NFS4_OPEN_STID:
4522 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
4525 ret = nfserr_locks_held;
4527 case NFS4_LOCK_STID:
4528 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
4531 stp = openlockstateid(s);
4532 ret = nfserr_locks_held;
4533 if (check_for_locks(stp->st_stid.sc_file,
4534 lockowner(stp->st_stateowner)))
4536 unhash_lock_stateid(stp);
4537 spin_unlock(&cl->cl_lock);
4541 case NFS4_REVOKED_DELEG_STID:
4542 dp = delegstateid(s);
4543 list_del_init(&dp->dl_recall_lru);
4544 spin_unlock(&cl->cl_lock);
4548 /* Default falls through and returns nfserr_bad_stateid */
4551 spin_unlock(&cl->cl_lock);
4559 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
4560 RD_STATE : WR_STATE;
4563 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
4565 struct svc_fh *current_fh = &cstate->current_fh;
4566 struct nfs4_stateowner *sop = stp->st_stateowner;
4569 status = nfsd4_check_seqid(cstate, sop, seqid);
4572 if (stp->st_stid.sc_type == NFS4_CLOSED_STID
4573 || stp->st_stid.sc_type == NFS4_REVOKED_DELEG_STID)
4575 * "Closed" stateid's exist *only* to return
4576 * nfserr_replay_me from the previous step, and
4577 * revoked delegations are kept only for free_stateid.
4579 return nfserr_bad_stateid;
4580 status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
4583 return nfs4_check_fh(current_fh, stp);
4587 * Checks for sequence id mutating operations.
4590 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
4591 stateid_t *stateid, char typemask,
4592 struct nfs4_ol_stateid **stpp,
4593 struct nfsd_net *nn)
4596 struct nfs4_stid *s;
4597 struct nfs4_ol_stateid *stp = NULL;
4599 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
4600 seqid, STATEID_VAL(stateid));
4603 status = nfsd4_lookup_stateid(cstate, stateid, typemask, &s, nn);
4606 stp = openlockstateid(s);
4607 nfsd4_cstate_assign_replay(cstate, stp->st_stateowner);
4609 status = nfs4_seqid_op_checks(cstate, stateid, seqid, stp);
4613 nfs4_put_stid(&stp->st_stid);
4617 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
4618 stateid_t *stateid, struct nfs4_ol_stateid **stpp, struct nfsd_net *nn)
4621 struct nfs4_openowner *oo;
4622 struct nfs4_ol_stateid *stp;
4624 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
4625 NFS4_OPEN_STID, &stp, nn);
4628 oo = openowner(stp->st_stateowner);
4629 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
4630 nfs4_put_stid(&stp->st_stid);
4631 return nfserr_bad_stateid;
4638 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4639 struct nfsd4_open_confirm *oc)
4642 struct nfs4_openowner *oo;
4643 struct nfs4_ol_stateid *stp;
4644 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4646 dprintk("NFSD: nfsd4_open_confirm on file %pd\n",
4647 cstate->current_fh.fh_dentry);
4649 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
4653 status = nfs4_preprocess_seqid_op(cstate,
4654 oc->oc_seqid, &oc->oc_req_stateid,
4655 NFS4_OPEN_STID, &stp, nn);
4658 oo = openowner(stp->st_stateowner);
4659 status = nfserr_bad_stateid;
4660 if (oo->oo_flags & NFS4_OO_CONFIRMED)
4662 oo->oo_flags |= NFS4_OO_CONFIRMED;
4663 update_stateid(&stp->st_stid.sc_stateid);
4664 memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4665 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
4666 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
4668 nfsd4_client_record_create(oo->oo_owner.so_client);
4671 nfs4_put_stid(&stp->st_stid);
4673 nfsd4_bump_seqid(cstate, status);
4677 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
4679 if (!test_access(access, stp))
4681 nfs4_file_put_access(stp->st_stid.sc_file, access);
4682 clear_access(access, stp);
4685 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
4687 switch (to_access) {
4688 case NFS4_SHARE_ACCESS_READ:
4689 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
4690 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
4692 case NFS4_SHARE_ACCESS_WRITE:
4693 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
4694 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
4696 case NFS4_SHARE_ACCESS_BOTH:
4704 nfsd4_open_downgrade(struct svc_rqst *rqstp,
4705 struct nfsd4_compound_state *cstate,
4706 struct nfsd4_open_downgrade *od)
4709 struct nfs4_ol_stateid *stp;
4710 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4712 dprintk("NFSD: nfsd4_open_downgrade on file %pd\n",
4713 cstate->current_fh.fh_dentry);
4715 /* We don't yet support WANT bits: */
4716 if (od->od_deleg_want)
4717 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
4720 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
4721 &od->od_stateid, &stp, nn);
4724 status = nfserr_inval;
4725 if (!test_access(od->od_share_access, stp)) {
4726 dprintk("NFSD: access not a subset of current bitmap: 0x%hhx, input access=%08x\n",
4727 stp->st_access_bmap, od->od_share_access);
4730 if (!test_deny(od->od_share_deny, stp)) {
4731 dprintk("NFSD: deny not a subset of current bitmap: 0x%hhx, input deny=%08x\n",
4732 stp->st_deny_bmap, od->od_share_deny);
4735 nfs4_stateid_downgrade(stp, od->od_share_access);
4737 reset_union_bmap_deny(od->od_share_deny, stp);
4739 update_stateid(&stp->st_stid.sc_stateid);
4740 memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4743 nfs4_put_stid(&stp->st_stid);
4745 nfsd4_bump_seqid(cstate, status);
4749 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
4751 struct nfs4_client *clp = s->st_stid.sc_client;
4752 LIST_HEAD(reaplist);
4754 s->st_stid.sc_type = NFS4_CLOSED_STID;
4755 spin_lock(&clp->cl_lock);
4756 unhash_open_stateid(s, &reaplist);
4758 if (clp->cl_minorversion) {
4759 put_ol_stateid_locked(s, &reaplist);
4760 spin_unlock(&clp->cl_lock);
4761 free_ol_stateid_reaplist(&reaplist);
4763 spin_unlock(&clp->cl_lock);
4764 free_ol_stateid_reaplist(&reaplist);
4765 move_to_close_lru(s, clp->net);
4770 * nfs4_unlock_state() called after encode
4773 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4774 struct nfsd4_close *close)
4777 struct nfs4_ol_stateid *stp;
4778 struct net *net = SVC_NET(rqstp);
4779 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4781 dprintk("NFSD: nfsd4_close on file %pd\n",
4782 cstate->current_fh.fh_dentry);
4784 status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
4786 NFS4_OPEN_STID|NFS4_CLOSED_STID,
4788 nfsd4_bump_seqid(cstate, status);
4791 update_stateid(&stp->st_stid.sc_stateid);
4792 memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4794 nfsd4_close_open_stateid(stp);
4796 /* put reference from nfs4_preprocess_seqid_op */
4797 nfs4_put_stid(&stp->st_stid);
4803 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4804 struct nfsd4_delegreturn *dr)
4806 struct nfs4_delegation *dp;
4807 stateid_t *stateid = &dr->dr_stateid;
4808 struct nfs4_stid *s;
4810 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4812 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4815 status = nfsd4_lookup_stateid(cstate, stateid, NFS4_DELEG_STID, &s, nn);
4818 dp = delegstateid(s);
4819 status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
4823 destroy_delegation(dp);
4825 nfs4_put_stid(&dp->dl_stid);
4831 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
4834 end_offset(u64 start, u64 len)
4839 return end >= start ? end: NFS4_MAX_UINT64;
4842 /* last octet in a range */
4844 last_byte_offset(u64 start, u64 len)
4850 return end > start ? end - 1: NFS4_MAX_UINT64;
4854 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
4855 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
4856 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
4857 * locking, this prevents us from being completely protocol-compliant. The
4858 * real solution to this problem is to start using unsigned file offsets in
4859 * the VFS, but this is a very deep change!
4862 nfs4_transform_lock_offset(struct file_lock *lock)
4864 if (lock->fl_start < 0)
4865 lock->fl_start = OFFSET_MAX;
4866 if (lock->fl_end < 0)
4867 lock->fl_end = OFFSET_MAX;
4870 /* Hack!: For now, we're defining this just so we can use a pointer to it
4871 * as a unique cookie to identify our (NFSv4's) posix locks. */
4872 static const struct lock_manager_operations nfsd_posix_mng_ops = {
4876 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
4878 struct nfs4_lockowner *lo;
4880 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
4881 lo = (struct nfs4_lockowner *) fl->fl_owner;
4882 deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
4883 lo->lo_owner.so_owner.len, GFP_KERNEL);
4884 if (!deny->ld_owner.data)
4885 /* We just don't care that much */
4887 deny->ld_owner.len = lo->lo_owner.so_owner.len;
4888 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
4891 deny->ld_owner.len = 0;
4892 deny->ld_owner.data = NULL;
4893 deny->ld_clientid.cl_boot = 0;
4894 deny->ld_clientid.cl_id = 0;
4896 deny->ld_start = fl->fl_start;
4897 deny->ld_length = NFS4_MAX_UINT64;
4898 if (fl->fl_end != NFS4_MAX_UINT64)
4899 deny->ld_length = fl->fl_end - fl->fl_start + 1;
4900 deny->ld_type = NFS4_READ_LT;
4901 if (fl->fl_type != F_RDLCK)
4902 deny->ld_type = NFS4_WRITE_LT;
4905 static struct nfs4_lockowner *
4906 find_lockowner_str_locked(clientid_t *clid, struct xdr_netobj *owner,
4907 struct nfs4_client *clp)
4909 unsigned int strhashval = ownerstr_hashval(owner);
4910 struct nfs4_stateowner *so;
4912 lockdep_assert_held(&clp->cl_lock);
4914 list_for_each_entry(so, &clp->cl_ownerstr_hashtbl[strhashval],
4916 if (so->so_is_open_owner)
4918 if (!same_owner_str(so, owner))
4920 atomic_inc(&so->so_count);
4921 return lockowner(so);
4926 static struct nfs4_lockowner *
4927 find_lockowner_str(clientid_t *clid, struct xdr_netobj *owner,
4928 struct nfs4_client *clp)
4930 struct nfs4_lockowner *lo;
4932 spin_lock(&clp->cl_lock);
4933 lo = find_lockowner_str_locked(clid, owner, clp);
4934 spin_unlock(&clp->cl_lock);
4938 static void nfs4_unhash_lockowner(struct nfs4_stateowner *sop)
4940 unhash_lockowner_locked(lockowner(sop));
4943 static void nfs4_free_lockowner(struct nfs4_stateowner *sop)
4945 struct nfs4_lockowner *lo = lockowner(sop);
4947 kmem_cache_free(lockowner_slab, lo);
4950 static const struct nfs4_stateowner_operations lockowner_ops = {
4951 .so_unhash = nfs4_unhash_lockowner,
4952 .so_free = nfs4_free_lockowner,
4956 * Alloc a lock owner structure.
4957 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
4960 * strhashval = ownerstr_hashval
4962 static struct nfs4_lockowner *
4963 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp,
4964 struct nfs4_ol_stateid *open_stp,
4965 struct nfsd4_lock *lock)
4967 struct nfs4_lockowner *lo, *ret;
4969 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
4972 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
4973 lo->lo_owner.so_is_open_owner = 0;
4974 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid;
4975 lo->lo_owner.so_ops = &lockowner_ops;
4976 spin_lock(&clp->cl_lock);
4977 ret = find_lockowner_str_locked(&clp->cl_clientid,
4978 &lock->lk_new_owner, clp);
4980 list_add(&lo->lo_owner.so_strhash,
4981 &clp->cl_ownerstr_hashtbl[strhashval]);
4984 nfs4_free_lockowner(&lo->lo_owner);
4985 spin_unlock(&clp->cl_lock);
4990 init_lock_stateid(struct nfs4_ol_stateid *stp, struct nfs4_lockowner *lo,
4991 struct nfs4_file *fp, struct inode *inode,
4992 struct nfs4_ol_stateid *open_stp)
4994 struct nfs4_client *clp = lo->lo_owner.so_client;
4996 lockdep_assert_held(&clp->cl_lock);
4998 atomic_inc(&stp->st_stid.sc_count);
4999 stp->st_stid.sc_type = NFS4_LOCK_STID;
5000 stp->st_stateowner = &lo->lo_owner;
5001 atomic_inc(&lo->lo_owner.so_count);
5003 stp->st_stid.sc_file = fp;
5004 stp->st_stid.sc_free = nfs4_free_lock_stateid;
5005 stp->st_access_bmap = 0;
5006 stp->st_deny_bmap = open_stp->st_deny_bmap;
5007 stp->st_openstp = open_stp;
5008 list_add(&stp->st_locks, &open_stp->st_locks);
5009 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
5010 spin_lock(&fp->fi_lock);
5011 list_add(&stp->st_perfile, &fp->fi_stateids);
5012 spin_unlock(&fp->fi_lock);
5015 static struct nfs4_ol_stateid *
5016 find_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp)
5018 struct nfs4_ol_stateid *lst;
5019 struct nfs4_client *clp = lo->lo_owner.so_client;
5021 lockdep_assert_held(&clp->cl_lock);
5023 list_for_each_entry(lst, &lo->lo_owner.so_stateids, st_perstateowner) {
5024 if (lst->st_stid.sc_file == fp) {
5025 atomic_inc(&lst->st_stid.sc_count);
5032 static struct nfs4_ol_stateid *
5033 find_or_create_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fi,
5034 struct inode *inode, struct nfs4_ol_stateid *ost,
5037 struct nfs4_stid *ns = NULL;
5038 struct nfs4_ol_stateid *lst;
5039 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
5040 struct nfs4_client *clp = oo->oo_owner.so_client;
5042 spin_lock(&clp->cl_lock);
5043 lst = find_lock_stateid(lo, fi);
5045 spin_unlock(&clp->cl_lock);
5046 ns = nfs4_alloc_stid(clp, stateid_slab);
5050 spin_lock(&clp->cl_lock);
5051 lst = find_lock_stateid(lo, fi);
5053 lst = openlockstateid(ns);
5054 init_lock_stateid(lst, lo, fi, inode, ost);
5059 spin_unlock(&clp->cl_lock);
5066 check_lock_length(u64 offset, u64 length)
5068 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
5069 LOFF_OVERFLOW(offset, length)));
5072 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
5074 struct nfs4_file *fp = lock_stp->st_stid.sc_file;
5076 lockdep_assert_held(&fp->fi_lock);
5078 if (test_access(access, lock_stp))
5080 __nfs4_file_get_access(fp, access);
5081 set_access(access, lock_stp);
5085 lookup_or_create_lock_state(struct nfsd4_compound_state *cstate,
5086 struct nfs4_ol_stateid *ost,
5087 struct nfsd4_lock *lock,
5088 struct nfs4_ol_stateid **lst, bool *new)
5091 struct nfs4_file *fi = ost->st_stid.sc_file;
5092 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
5093 struct nfs4_client *cl = oo->oo_owner.so_client;
5094 struct inode *inode = cstate->current_fh.fh_dentry->d_inode;
5095 struct nfs4_lockowner *lo;
5096 unsigned int strhashval;
5098 lo = find_lockowner_str(&cl->cl_clientid, &lock->v.new.owner, cl);
5100 strhashval = ownerstr_hashval(&lock->v.new.owner);
5101 lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
5103 return nfserr_jukebox;
5105 /* with an existing lockowner, seqids must be the same */
5106 status = nfserr_bad_seqid;
5107 if (!cstate->minorversion &&
5108 lock->lk_new_lock_seqid != lo->lo_owner.so_seqid)
5112 *lst = find_or_create_lock_stateid(lo, fi, inode, ost, new);
5114 status = nfserr_jukebox;
5119 nfs4_put_stateowner(&lo->lo_owner);
5127 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5128 struct nfsd4_lock *lock)
5130 struct nfs4_openowner *open_sop = NULL;
5131 struct nfs4_lockowner *lock_sop = NULL;
5132 struct nfs4_ol_stateid *lock_stp = NULL;
5133 struct nfs4_ol_stateid *open_stp = NULL;
5134 struct nfs4_file *fp;
5135 struct file *filp = NULL;
5136 struct file_lock *file_lock = NULL;
5137 struct file_lock *conflock = NULL;
5142 struct net *net = SVC_NET(rqstp);
5143 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5145 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
5146 (long long) lock->lk_offset,
5147 (long long) lock->lk_length);
5149 if (check_lock_length(lock->lk_offset, lock->lk_length))
5150 return nfserr_inval;
5152 if ((status = fh_verify(rqstp, &cstate->current_fh,
5153 S_IFREG, NFSD_MAY_LOCK))) {
5154 dprintk("NFSD: nfsd4_lock: permission denied!\n");
5158 if (lock->lk_is_new) {
5159 if (nfsd4_has_session(cstate))
5160 /* See rfc 5661 18.10.3: given clientid is ignored: */
5161 memcpy(&lock->v.new.clientid,
5162 &cstate->session->se_client->cl_clientid,
5163 sizeof(clientid_t));
5165 status = nfserr_stale_clientid;
5166 if (STALE_CLIENTID(&lock->lk_new_clientid, nn))
5169 /* validate and update open stateid and open seqid */
5170 status = nfs4_preprocess_confirmed_seqid_op(cstate,
5171 lock->lk_new_open_seqid,
5172 &lock->lk_new_open_stateid,
5176 open_sop = openowner(open_stp->st_stateowner);
5177 status = nfserr_bad_stateid;
5178 if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
5179 &lock->v.new.clientid))
5181 status = lookup_or_create_lock_state(cstate, open_stp, lock,
5184 status = nfs4_preprocess_seqid_op(cstate,
5185 lock->lk_old_lock_seqid,
5186 &lock->lk_old_lock_stateid,
5187 NFS4_LOCK_STID, &lock_stp, nn);
5191 lock_sop = lockowner(lock_stp->st_stateowner);
5193 lkflg = setlkflg(lock->lk_type);
5194 status = nfs4_check_openmode(lock_stp, lkflg);
5198 status = nfserr_grace;
5199 if (locks_in_grace(net) && !lock->lk_reclaim)
5201 status = nfserr_no_grace;
5202 if (!locks_in_grace(net) && lock->lk_reclaim)
5205 file_lock = locks_alloc_lock();
5207 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
5208 status = nfserr_jukebox;
5212 fp = lock_stp->st_stid.sc_file;
5213 locks_init_lock(file_lock);
5214 switch (lock->lk_type) {
5217 spin_lock(&fp->fi_lock);
5218 filp = find_readable_file_locked(fp);
5220 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
5221 spin_unlock(&fp->fi_lock);
5222 file_lock->fl_type = F_RDLCK;
5225 case NFS4_WRITEW_LT:
5226 spin_lock(&fp->fi_lock);
5227 filp = find_writeable_file_locked(fp);
5229 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
5230 spin_unlock(&fp->fi_lock);
5231 file_lock->fl_type = F_WRLCK;
5234 status = nfserr_inval;
5238 status = nfserr_openmode;
5241 file_lock->fl_owner = (fl_owner_t)lock_sop;
5242 file_lock->fl_pid = current->tgid;
5243 file_lock->fl_file = filp;
5244 file_lock->fl_flags = FL_POSIX;
5245 file_lock->fl_lmops = &nfsd_posix_mng_ops;
5246 file_lock->fl_start = lock->lk_offset;
5247 file_lock->fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
5248 nfs4_transform_lock_offset(file_lock);
5250 conflock = locks_alloc_lock();
5252 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
5253 status = nfserr_jukebox;
5257 err = vfs_lock_file(filp, F_SETLK, file_lock, conflock);
5259 case 0: /* success! */
5260 update_stateid(&lock_stp->st_stid.sc_stateid);
5261 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stid.sc_stateid,
5265 case (EAGAIN): /* conflock holds conflicting lock */
5266 status = nfserr_denied;
5267 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
5268 nfs4_set_lock_denied(conflock, &lock->lk_denied);
5271 status = nfserr_deadlock;
5274 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
5275 status = nfserrno(err);
5282 /* Bump seqid manually if the 4.0 replay owner is openowner */
5283 if (cstate->replay_owner &&
5284 cstate->replay_owner != &lock_sop->lo_owner &&
5285 seqid_mutating_err(ntohl(status)))
5286 lock_sop->lo_owner.so_seqid++;
5289 * If this is a new, never-before-used stateid, and we are
5290 * returning an error, then just go ahead and release it.
5293 release_lock_stateid(lock_stp);
5295 nfs4_put_stid(&lock_stp->st_stid);
5298 nfs4_put_stid(&open_stp->st_stid);
5299 nfsd4_bump_seqid(cstate, status);
5301 locks_free_lock(file_lock);
5303 locks_free_lock(conflock);
5308 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
5309 * so we do a temporary open here just to get an open file to pass to
5310 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
5313 static __be32 nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
5316 __be32 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
5318 err = nfserrno(vfs_test_lock(file, lock));
5328 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5329 struct nfsd4_lockt *lockt)
5331 struct file_lock *file_lock = NULL;
5332 struct nfs4_lockowner *lo = NULL;
5334 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
5336 if (locks_in_grace(SVC_NET(rqstp)))
5337 return nfserr_grace;
5339 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
5340 return nfserr_inval;
5342 if (!nfsd4_has_session(cstate)) {
5343 status = lookup_clientid(&lockt->lt_clientid, cstate, nn);
5348 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
5351 file_lock = locks_alloc_lock();
5353 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
5354 status = nfserr_jukebox;
5357 locks_init_lock(file_lock);
5358 switch (lockt->lt_type) {
5361 file_lock->fl_type = F_RDLCK;
5364 case NFS4_WRITEW_LT:
5365 file_lock->fl_type = F_WRLCK;
5368 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
5369 status = nfserr_inval;
5373 lo = find_lockowner_str(&lockt->lt_clientid, &lockt->lt_owner,
5376 file_lock->fl_owner = (fl_owner_t)lo;
5377 file_lock->fl_pid = current->tgid;
5378 file_lock->fl_flags = FL_POSIX;
5380 file_lock->fl_start = lockt->lt_offset;
5381 file_lock->fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
5383 nfs4_transform_lock_offset(file_lock);
5385 status = nfsd_test_lock(rqstp, &cstate->current_fh, file_lock);
5389 if (file_lock->fl_type != F_UNLCK) {
5390 status = nfserr_denied;
5391 nfs4_set_lock_denied(file_lock, &lockt->lt_denied);
5395 nfs4_put_stateowner(&lo->lo_owner);
5397 locks_free_lock(file_lock);
5402 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5403 struct nfsd4_locku *locku)
5405 struct nfs4_ol_stateid *stp;
5406 struct file *filp = NULL;
5407 struct file_lock *file_lock = NULL;
5410 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
5412 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
5413 (long long) locku->lu_offset,
5414 (long long) locku->lu_length);
5416 if (check_lock_length(locku->lu_offset, locku->lu_length))
5417 return nfserr_inval;
5419 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
5420 &locku->lu_stateid, NFS4_LOCK_STID,
5424 filp = find_any_file(stp->st_stid.sc_file);
5426 status = nfserr_lock_range;
5429 file_lock = locks_alloc_lock();
5431 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
5432 status = nfserr_jukebox;
5435 locks_init_lock(file_lock);
5436 file_lock->fl_type = F_UNLCK;
5437 file_lock->fl_owner = (fl_owner_t)lockowner(stp->st_stateowner);
5438 file_lock->fl_pid = current->tgid;
5439 file_lock->fl_file = filp;
5440 file_lock->fl_flags = FL_POSIX;
5441 file_lock->fl_lmops = &nfsd_posix_mng_ops;
5442 file_lock->fl_start = locku->lu_offset;
5444 file_lock->fl_end = last_byte_offset(locku->lu_offset,
5446 nfs4_transform_lock_offset(file_lock);
5448 err = vfs_lock_file(filp, F_SETLK, file_lock, NULL);
5450 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
5453 update_stateid(&stp->st_stid.sc_stateid);
5454 memcpy(&locku->lu_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
5458 nfs4_put_stid(&stp->st_stid);
5460 nfsd4_bump_seqid(cstate, status);
5462 locks_free_lock(file_lock);
5466 status = nfserrno(err);
5472 * true: locks held by lockowner
5473 * false: no locks held by lockowner
5476 check_for_locks(struct nfs4_file *fp, struct nfs4_lockowner *lowner)
5478 struct file_lock **flpp;
5480 struct file *filp = find_any_file(fp);
5481 struct inode *inode;
5484 /* Any valid lock stateid should have some sort of access */
5489 inode = file_inode(filp);
5491 spin_lock(&inode->i_lock);
5492 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
5493 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
5498 spin_unlock(&inode->i_lock);
5504 nfsd4_release_lockowner(struct svc_rqst *rqstp,
5505 struct nfsd4_compound_state *cstate,
5506 struct nfsd4_release_lockowner *rlockowner)
5508 clientid_t *clid = &rlockowner->rl_clientid;
5509 struct nfs4_stateowner *sop;
5510 struct nfs4_lockowner *lo = NULL;
5511 struct nfs4_ol_stateid *stp;
5512 struct xdr_netobj *owner = &rlockowner->rl_owner;
5513 unsigned int hashval = ownerstr_hashval(owner);
5515 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
5516 struct nfs4_client *clp;
5518 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
5519 clid->cl_boot, clid->cl_id);
5521 status = lookup_clientid(clid, cstate, nn);
5526 /* Find the matching lock stateowner */
5527 spin_lock(&clp->cl_lock);
5528 list_for_each_entry(sop, &clp->cl_ownerstr_hashtbl[hashval],
5531 if (sop->so_is_open_owner || !same_owner_str(sop, owner))
5534 /* see if there are still any locks associated with it */
5535 lo = lockowner(sop);
5536 list_for_each_entry(stp, &sop->so_stateids, st_perstateowner) {
5537 if (check_for_locks(stp->st_stid.sc_file, lo)) {
5538 status = nfserr_locks_held;
5539 spin_unlock(&clp->cl_lock);
5544 atomic_inc(&sop->so_count);
5547 spin_unlock(&clp->cl_lock);
5549 release_lockowner(lo);
5553 static inline struct nfs4_client_reclaim *
5556 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
5560 nfs4_has_reclaimed_state(const char *name, struct nfsd_net *nn)
5562 struct nfs4_client_reclaim *crp;
5564 crp = nfsd4_find_reclaim_client(name, nn);
5565 return (crp && crp->cr_clp);
5569 * failure => all reset bets are off, nfserr_no_grace...
5571 struct nfs4_client_reclaim *
5572 nfs4_client_to_reclaim(const char *name, struct nfsd_net *nn)
5574 unsigned int strhashval;
5575 struct nfs4_client_reclaim *crp;
5577 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
5578 crp = alloc_reclaim();
5580 strhashval = clientstr_hashval(name);
5581 INIT_LIST_HEAD(&crp->cr_strhash);
5582 list_add(&crp->cr_strhash, &nn->reclaim_str_hashtbl[strhashval]);
5583 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
5585 nn->reclaim_str_hashtbl_size++;
5591 nfs4_remove_reclaim_record(struct nfs4_client_reclaim *crp, struct nfsd_net *nn)
5593 list_del(&crp->cr_strhash);
5595 nn->reclaim_str_hashtbl_size--;
5599 nfs4_release_reclaim(struct nfsd_net *nn)
5601 struct nfs4_client_reclaim *crp = NULL;
5604 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
5605 while (!list_empty(&nn->reclaim_str_hashtbl[i])) {
5606 crp = list_entry(nn->reclaim_str_hashtbl[i].next,
5607 struct nfs4_client_reclaim, cr_strhash);
5608 nfs4_remove_reclaim_record(crp, nn);
5611 WARN_ON_ONCE(nn->reclaim_str_hashtbl_size);
5615 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
5616 struct nfs4_client_reclaim *
5617 nfsd4_find_reclaim_client(const char *recdir, struct nfsd_net *nn)
5619 unsigned int strhashval;
5620 struct nfs4_client_reclaim *crp = NULL;
5622 dprintk("NFSD: nfs4_find_reclaim_client for recdir %s\n", recdir);
5624 strhashval = clientstr_hashval(recdir);
5625 list_for_each_entry(crp, &nn->reclaim_str_hashtbl[strhashval], cr_strhash) {
5626 if (same_name(crp->cr_recdir, recdir)) {
5634 * Called from OPEN. Look for clientid in reclaim list.
5637 nfs4_check_open_reclaim(clientid_t *clid,
5638 struct nfsd4_compound_state *cstate,
5639 struct nfsd_net *nn)
5643 /* find clientid in conf_id_hashtbl */
5644 status = lookup_clientid(clid, cstate, nn);
5646 return nfserr_reclaim_bad;
5648 if (nfsd4_client_record_check(cstate->clp))
5649 return nfserr_reclaim_bad;
5654 #ifdef CONFIG_NFSD_FAULT_INJECTION
5656 put_client(struct nfs4_client *clp)
5658 atomic_dec(&clp->cl_refcount);
5661 static struct nfs4_client *
5662 nfsd_find_client(struct sockaddr_storage *addr, size_t addr_size)
5664 struct nfs4_client *clp;
5665 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
5668 if (!nfsd_netns_ready(nn))
5671 list_for_each_entry(clp, &nn->client_lru, cl_lru) {
5672 if (memcmp(&clp->cl_addr, addr, addr_size) == 0)
5679 nfsd_inject_print_clients(void)
5681 struct nfs4_client *clp;
5683 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
5685 char buf[INET6_ADDRSTRLEN];
5687 if (!nfsd_netns_ready(nn))
5690 spin_lock(&nn->client_lock);
5691 list_for_each_entry(clp, &nn->client_lru, cl_lru) {
5692 rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, sizeof(buf));
5693 pr_info("NFS Client: %s\n", buf);
5696 spin_unlock(&nn->client_lock);
5702 nfsd_inject_forget_client(struct sockaddr_storage *addr, size_t addr_size)
5705 struct nfs4_client *clp;
5706 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
5709 if (!nfsd_netns_ready(nn))
5712 spin_lock(&nn->client_lock);
5713 clp = nfsd_find_client(addr, addr_size);
5715 if (mark_client_expired_locked(clp) == nfs_ok)
5720 spin_unlock(&nn->client_lock);
5729 nfsd_inject_forget_clients(u64 max)
5732 struct nfs4_client *clp, *next;
5733 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
5735 LIST_HEAD(reaplist);
5737 if (!nfsd_netns_ready(nn))
5740 spin_lock(&nn->client_lock);
5741 list_for_each_entry_safe(clp, next, &nn->client_lru, cl_lru) {
5742 if (mark_client_expired_locked(clp) == nfs_ok) {
5743 list_add(&clp->cl_lru, &reaplist);
5744 if (max != 0 && ++count >= max)
5748 spin_unlock(&nn->client_lock);
5750 list_for_each_entry_safe(clp, next, &reaplist, cl_lru)
5756 static void nfsd_print_count(struct nfs4_client *clp, unsigned int count,
5759 char buf[INET6_ADDRSTRLEN];
5760 rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, sizeof(buf));
5761 printk(KERN_INFO "NFS Client: %s has %u %s\n", buf, count, type);
5765 nfsd_inject_add_lock_to_list(struct nfs4_ol_stateid *lst,
5766 struct list_head *collect)
5768 struct nfs4_client *clp = lst->st_stid.sc_client;
5769 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
5775 lockdep_assert_held(&nn->client_lock);
5776 atomic_inc(&clp->cl_refcount);
5777 list_add(&lst->st_locks, collect);
5780 static u64 nfsd_foreach_client_lock(struct nfs4_client *clp, u64 max,
5781 struct list_head *collect,
5782 void (*func)(struct nfs4_ol_stateid *))
5784 struct nfs4_openowner *oop;
5785 struct nfs4_ol_stateid *stp, *st_next;
5786 struct nfs4_ol_stateid *lst, *lst_next;
5789 spin_lock(&clp->cl_lock);
5790 list_for_each_entry(oop, &clp->cl_openowners, oo_perclient) {
5791 list_for_each_entry_safe(stp, st_next,
5792 &oop->oo_owner.so_stateids, st_perstateowner) {
5793 list_for_each_entry_safe(lst, lst_next,
5794 &stp->st_locks, st_locks) {
5797 nfsd_inject_add_lock_to_list(lst,
5802 * Despite the fact that these functions deal
5803 * with 64-bit integers for "count", we must
5804 * ensure that it doesn't blow up the
5805 * clp->cl_refcount. Throw a warning if we
5806 * start to approach INT_MAX here.
5808 WARN_ON_ONCE(count == (INT_MAX / 2));
5815 spin_unlock(&clp->cl_lock);
5821 nfsd_collect_client_locks(struct nfs4_client *clp, struct list_head *collect,
5824 return nfsd_foreach_client_lock(clp, max, collect, unhash_lock_stateid);
5828 nfsd_print_client_locks(struct nfs4_client *clp)
5830 u64 count = nfsd_foreach_client_lock(clp, 0, NULL, NULL);
5831 nfsd_print_count(clp, count, "locked files");
5836 nfsd_inject_print_locks(void)
5838 struct nfs4_client *clp;
5840 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
5843 if (!nfsd_netns_ready(nn))
5846 spin_lock(&nn->client_lock);
5847 list_for_each_entry(clp, &nn->client_lru, cl_lru)
5848 count += nfsd_print_client_locks(clp);
5849 spin_unlock(&nn->client_lock);
5855 nfsd_reap_locks(struct list_head *reaplist)
5857 struct nfs4_client *clp;
5858 struct nfs4_ol_stateid *stp, *next;
5860 list_for_each_entry_safe(stp, next, reaplist, st_locks) {
5861 list_del_init(&stp->st_locks);
5862 clp = stp->st_stid.sc_client;
5863 nfs4_put_stid(&stp->st_stid);
5869 nfsd_inject_forget_client_locks(struct sockaddr_storage *addr, size_t addr_size)
5871 unsigned int count = 0;
5872 struct nfs4_client *clp;
5873 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
5875 LIST_HEAD(reaplist);
5877 if (!nfsd_netns_ready(nn))
5880 spin_lock(&nn->client_lock);
5881 clp = nfsd_find_client(addr, addr_size);
5883 count = nfsd_collect_client_locks(clp, &reaplist, 0);
5884 spin_unlock(&nn->client_lock);
5885 nfsd_reap_locks(&reaplist);
5890 nfsd_inject_forget_locks(u64 max)
5893 struct nfs4_client *clp;
5894 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
5896 LIST_HEAD(reaplist);
5898 if (!nfsd_netns_ready(nn))
5901 spin_lock(&nn->client_lock);
5902 list_for_each_entry(clp, &nn->client_lru, cl_lru) {
5903 count += nfsd_collect_client_locks(clp, &reaplist, max - count);
5904 if (max != 0 && count >= max)
5907 spin_unlock(&nn->client_lock);
5908 nfsd_reap_locks(&reaplist);
5913 nfsd_foreach_client_openowner(struct nfs4_client *clp, u64 max,
5914 struct list_head *collect,
5915 void (*func)(struct nfs4_openowner *))
5917 struct nfs4_openowner *oop, *next;
5918 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
5922 lockdep_assert_held(&nn->client_lock);
5924 spin_lock(&clp->cl_lock);
5925 list_for_each_entry_safe(oop, next, &clp->cl_openowners, oo_perclient) {
5929 atomic_inc(&clp->cl_refcount);
5930 list_add(&oop->oo_perclient, collect);
5935 * Despite the fact that these functions deal with
5936 * 64-bit integers for "count", we must ensure that
5937 * it doesn't blow up the clp->cl_refcount. Throw a
5938 * warning if we start to approach INT_MAX here.
5940 WARN_ON_ONCE(count == (INT_MAX / 2));
5944 spin_unlock(&clp->cl_lock);
5950 nfsd_print_client_openowners(struct nfs4_client *clp)
5952 u64 count = nfsd_foreach_client_openowner(clp, 0, NULL, NULL);
5954 nfsd_print_count(clp, count, "openowners");
5959 nfsd_collect_client_openowners(struct nfs4_client *clp,
5960 struct list_head *collect, u64 max)
5962 return nfsd_foreach_client_openowner(clp, max, collect,
5963 unhash_openowner_locked);
5967 nfsd_inject_print_openowners(void)
5969 struct nfs4_client *clp;
5971 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
5974 if (!nfsd_netns_ready(nn))
5977 spin_lock(&nn->client_lock);
5978 list_for_each_entry(clp, &nn->client_lru, cl_lru)
5979 count += nfsd_print_client_openowners(clp);
5980 spin_unlock(&nn->client_lock);
5986 nfsd_reap_openowners(struct list_head *reaplist)
5988 struct nfs4_client *clp;
5989 struct nfs4_openowner *oop, *next;
5991 list_for_each_entry_safe(oop, next, reaplist, oo_perclient) {
5992 list_del_init(&oop->oo_perclient);
5993 clp = oop->oo_owner.so_client;
5994 release_openowner(oop);
6000 nfsd_inject_forget_client_openowners(struct sockaddr_storage *addr,
6003 unsigned int count = 0;
6004 struct nfs4_client *clp;
6005 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6007 LIST_HEAD(reaplist);
6009 if (!nfsd_netns_ready(nn))
6012 spin_lock(&nn->client_lock);
6013 clp = nfsd_find_client(addr, addr_size);
6015 count = nfsd_collect_client_openowners(clp, &reaplist, 0);
6016 spin_unlock(&nn->client_lock);
6017 nfsd_reap_openowners(&reaplist);
6022 nfsd_inject_forget_openowners(u64 max)
6025 struct nfs4_client *clp;
6026 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6028 LIST_HEAD(reaplist);
6030 if (!nfsd_netns_ready(nn))
6033 spin_lock(&nn->client_lock);
6034 list_for_each_entry(clp, &nn->client_lru, cl_lru) {
6035 count += nfsd_collect_client_openowners(clp, &reaplist,
6037 if (max != 0 && count >= max)
6040 spin_unlock(&nn->client_lock);
6041 nfsd_reap_openowners(&reaplist);
6045 static u64 nfsd_find_all_delegations(struct nfs4_client *clp, u64 max,
6046 struct list_head *victims)
6048 struct nfs4_delegation *dp, *next;
6049 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6053 lockdep_assert_held(&nn->client_lock);
6055 spin_lock(&state_lock);
6056 list_for_each_entry_safe(dp, next, &clp->cl_delegations, dl_perclnt) {
6059 * It's not safe to mess with delegations that have a
6060 * non-zero dl_time. They might have already been broken
6061 * and could be processed by the laundromat outside of
6062 * the state_lock. Just leave them be.
6064 if (dp->dl_time != 0)
6067 atomic_inc(&clp->cl_refcount);
6068 unhash_delegation_locked(dp);
6069 list_add(&dp->dl_recall_lru, victims);
6073 * Despite the fact that these functions deal with
6074 * 64-bit integers for "count", we must ensure that
6075 * it doesn't blow up the clp->cl_refcount. Throw a
6076 * warning if we start to approach INT_MAX here.
6078 WARN_ON_ONCE(count == (INT_MAX / 2));
6082 spin_unlock(&state_lock);
6087 nfsd_print_client_delegations(struct nfs4_client *clp)
6089 u64 count = nfsd_find_all_delegations(clp, 0, NULL);
6091 nfsd_print_count(clp, count, "delegations");
6096 nfsd_inject_print_delegations(void)
6098 struct nfs4_client *clp;
6100 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6103 if (!nfsd_netns_ready(nn))
6106 spin_lock(&nn->client_lock);
6107 list_for_each_entry(clp, &nn->client_lru, cl_lru)
6108 count += nfsd_print_client_delegations(clp);
6109 spin_unlock(&nn->client_lock);
6115 nfsd_forget_delegations(struct list_head *reaplist)
6117 struct nfs4_client *clp;
6118 struct nfs4_delegation *dp, *next;
6120 list_for_each_entry_safe(dp, next, reaplist, dl_recall_lru) {
6121 list_del_init(&dp->dl_recall_lru);
6122 clp = dp->dl_stid.sc_client;
6123 revoke_delegation(dp);
6129 nfsd_inject_forget_client_delegations(struct sockaddr_storage *addr,
6133 struct nfs4_client *clp;
6134 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6136 LIST_HEAD(reaplist);
6138 if (!nfsd_netns_ready(nn))
6141 spin_lock(&nn->client_lock);
6142 clp = nfsd_find_client(addr, addr_size);
6144 count = nfsd_find_all_delegations(clp, 0, &reaplist);
6145 spin_unlock(&nn->client_lock);
6147 nfsd_forget_delegations(&reaplist);
6152 nfsd_inject_forget_delegations(u64 max)
6155 struct nfs4_client *clp;
6156 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6158 LIST_HEAD(reaplist);
6160 if (!nfsd_netns_ready(nn))
6163 spin_lock(&nn->client_lock);
6164 list_for_each_entry(clp, &nn->client_lru, cl_lru) {
6165 count += nfsd_find_all_delegations(clp, max - count, &reaplist);
6166 if (max != 0 && count >= max)
6169 spin_unlock(&nn->client_lock);
6170 nfsd_forget_delegations(&reaplist);
6175 nfsd_recall_delegations(struct list_head *reaplist)
6177 struct nfs4_client *clp;
6178 struct nfs4_delegation *dp, *next;
6180 list_for_each_entry_safe(dp, next, reaplist, dl_recall_lru) {
6181 list_del_init(&dp->dl_recall_lru);
6182 clp = dp->dl_stid.sc_client;
6184 * We skipped all entries that had a zero dl_time before,
6185 * so we can now reset the dl_time back to 0. If a delegation
6186 * break comes in now, then it won't make any difference since
6187 * we're recalling it either way.
6189 spin_lock(&state_lock);
6191 spin_unlock(&state_lock);
6192 nfsd_break_one_deleg(dp);
6198 nfsd_inject_recall_client_delegations(struct sockaddr_storage *addr,
6202 struct nfs4_client *clp;
6203 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6205 LIST_HEAD(reaplist);
6207 if (!nfsd_netns_ready(nn))
6210 spin_lock(&nn->client_lock);
6211 clp = nfsd_find_client(addr, addr_size);
6213 count = nfsd_find_all_delegations(clp, 0, &reaplist);
6214 spin_unlock(&nn->client_lock);
6216 nfsd_recall_delegations(&reaplist);
6221 nfsd_inject_recall_delegations(u64 max)
6224 struct nfs4_client *clp, *next;
6225 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6227 LIST_HEAD(reaplist);
6229 if (!nfsd_netns_ready(nn))
6232 spin_lock(&nn->client_lock);
6233 list_for_each_entry_safe(clp, next, &nn->client_lru, cl_lru) {
6234 count += nfsd_find_all_delegations(clp, max - count, &reaplist);
6235 if (max != 0 && ++count >= max)
6238 spin_unlock(&nn->client_lock);
6239 nfsd_recall_delegations(&reaplist);
6242 #endif /* CONFIG_NFSD_FAULT_INJECTION */
6245 * Since the lifetime of a delegation isn't limited to that of an open, a
6246 * client may quite reasonably hang on to a delegation as long as it has
6247 * the inode cached. This becomes an obvious problem the first time a
6248 * client's inode cache approaches the size of the server's total memory.
6250 * For now we avoid this problem by imposing a hard limit on the number
6251 * of delegations, which varies according to the server's memory size.
6254 set_max_delegations(void)
6257 * Allow at most 4 delegations per megabyte of RAM. Quick
6258 * estimates suggest that in the worst case (where every delegation
6259 * is for a different inode), a delegation could take about 1.5K,
6260 * giving a worst case usage of about 6% of memory.
6262 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
6265 static int nfs4_state_create_net(struct net *net)
6267 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
6270 nn->conf_id_hashtbl = kmalloc(sizeof(struct list_head) *
6271 CLIENT_HASH_SIZE, GFP_KERNEL);
6272 if (!nn->conf_id_hashtbl)
6274 nn->unconf_id_hashtbl = kmalloc(sizeof(struct list_head) *
6275 CLIENT_HASH_SIZE, GFP_KERNEL);
6276 if (!nn->unconf_id_hashtbl)
6278 nn->sessionid_hashtbl = kmalloc(sizeof(struct list_head) *
6279 SESSION_HASH_SIZE, GFP_KERNEL);
6280 if (!nn->sessionid_hashtbl)
6283 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
6284 INIT_LIST_HEAD(&nn->conf_id_hashtbl[i]);
6285 INIT_LIST_HEAD(&nn->unconf_id_hashtbl[i]);
6287 for (i = 0; i < SESSION_HASH_SIZE; i++)
6288 INIT_LIST_HEAD(&nn->sessionid_hashtbl[i]);
6289 nn->conf_name_tree = RB_ROOT;
6290 nn->unconf_name_tree = RB_ROOT;
6291 INIT_LIST_HEAD(&nn->client_lru);
6292 INIT_LIST_HEAD(&nn->close_lru);
6293 INIT_LIST_HEAD(&nn->del_recall_lru);
6294 spin_lock_init(&nn->client_lock);
6296 INIT_DELAYED_WORK(&nn->laundromat_work, laundromat_main);
6302 kfree(nn->unconf_id_hashtbl);
6304 kfree(nn->conf_id_hashtbl);
6310 nfs4_state_destroy_net(struct net *net)
6313 struct nfs4_client *clp = NULL;
6314 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
6316 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
6317 while (!list_empty(&nn->conf_id_hashtbl[i])) {
6318 clp = list_entry(nn->conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
6319 destroy_client(clp);
6323 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
6324 while (!list_empty(&nn->unconf_id_hashtbl[i])) {
6325 clp = list_entry(nn->unconf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
6326 destroy_client(clp);
6330 kfree(nn->sessionid_hashtbl);
6331 kfree(nn->unconf_id_hashtbl);
6332 kfree(nn->conf_id_hashtbl);
6337 nfs4_state_start_net(struct net *net)
6339 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
6342 ret = nfs4_state_create_net(net);
6345 nfsd4_client_tracking_init(net);
6346 nn->boot_time = get_seconds();
6347 locks_start_grace(net, &nn->nfsd4_manager);
6348 nn->grace_ended = false;
6349 printk(KERN_INFO "NFSD: starting %ld-second grace period (net %p)\n",
6350 nn->nfsd4_grace, net);
6351 queue_delayed_work(laundry_wq, &nn->laundromat_work, nn->nfsd4_grace * HZ);
6355 /* initialization to perform when the nfsd service is started: */
6358 nfs4_state_start(void)
6362 ret = set_callback_cred();
6365 laundry_wq = create_singlethread_workqueue("nfsd4");
6366 if (laundry_wq == NULL) {
6370 ret = nfsd4_create_callback_queue();
6372 goto out_free_laundry;
6374 set_max_delegations();
6379 destroy_workqueue(laundry_wq);
6385 nfs4_state_shutdown_net(struct net *net)
6387 struct nfs4_delegation *dp = NULL;
6388 struct list_head *pos, *next, reaplist;
6389 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
6391 cancel_delayed_work_sync(&nn->laundromat_work);
6392 locks_end_grace(&nn->nfsd4_manager);
6394 INIT_LIST_HEAD(&reaplist);
6395 spin_lock(&state_lock);
6396 list_for_each_safe(pos, next, &nn->del_recall_lru) {
6397 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
6398 unhash_delegation_locked(dp);
6399 list_add(&dp->dl_recall_lru, &reaplist);
6401 spin_unlock(&state_lock);
6402 list_for_each_safe(pos, next, &reaplist) {
6403 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
6404 list_del_init(&dp->dl_recall_lru);
6405 nfs4_put_stid(&dp->dl_stid);
6408 nfsd4_client_tracking_exit(net);
6409 nfs4_state_destroy_net(net);
6413 nfs4_state_shutdown(void)
6415 destroy_workqueue(laundry_wq);
6416 nfsd4_destroy_callback_queue();
6420 get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
6422 if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG) && CURRENT_STATEID(stateid))
6423 memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
6427 put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
6429 if (cstate->minorversion) {
6430 memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
6431 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
6436 clear_current_stateid(struct nfsd4_compound_state *cstate)
6438 CLEAR_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
6442 * functions to set current state id
6445 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
6447 put_stateid(cstate, &odp->od_stateid);
6451 nfsd4_set_openstateid(struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
6453 put_stateid(cstate, &open->op_stateid);
6457 nfsd4_set_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
6459 put_stateid(cstate, &close->cl_stateid);
6463 nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate, struct nfsd4_lock *lock)
6465 put_stateid(cstate, &lock->lk_resp_stateid);
6469 * functions to consume current state id
6473 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
6475 get_stateid(cstate, &odp->od_stateid);
6479 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate, struct nfsd4_delegreturn *drp)
6481 get_stateid(cstate, &drp->dr_stateid);
6485 nfsd4_get_freestateid(struct nfsd4_compound_state *cstate, struct nfsd4_free_stateid *fsp)
6487 get_stateid(cstate, &fsp->fr_stateid);
6491 nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate, struct nfsd4_setattr *setattr)
6493 get_stateid(cstate, &setattr->sa_stateid);
6497 nfsd4_get_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
6499 get_stateid(cstate, &close->cl_stateid);
6503 nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate, struct nfsd4_locku *locku)
6505 get_stateid(cstate, &locku->lu_stateid);
6509 nfsd4_get_readstateid(struct nfsd4_compound_state *cstate, struct nfsd4_read *read)
6511 get_stateid(cstate, &read->rd_stateid);
6515 nfsd4_get_writestateid(struct nfsd4_compound_state *cstate, struct nfsd4_write *write)
6517 get_stateid(cstate, &write->wr_stateid);