2 * Copyright (c) 2001 The Regents of the University of Michigan.
5 * Kendrick Smith <kmsmith@umich.edu>
6 * Andy Adamson <kandros@umich.edu>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the University nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
22 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
23 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
24 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
28 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
29 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
30 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
31 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
35 #include <linux/file.h>
37 #include <linux/slab.h>
38 #include <linux/namei.h>
39 #include <linux/swap.h>
40 #include <linux/pagemap.h>
41 #include <linux/ratelimit.h>
42 #include <linux/sunrpc/svcauth_gss.h>
43 #include <linux/sunrpc/addr.h>
44 #include <linux/hash.h>
48 #include "current_stateid.h"
52 #define NFSDDBG_FACILITY NFSDDBG_PROC
54 #define all_ones {{~0,~0},~0}
55 static const stateid_t one_stateid = {
57 .si_opaque = all_ones,
59 static const stateid_t zero_stateid = {
62 static const stateid_t currentstateid = {
66 static u64 current_sessionid = 1;
68 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zero_stateid, sizeof(stateid_t)))
69 #define ONE_STATEID(stateid) (!memcmp((stateid), &one_stateid, sizeof(stateid_t)))
70 #define CURRENT_STATEID(stateid) (!memcmp((stateid), ¤tstateid, sizeof(stateid_t)))
72 /* forward declarations */
73 static bool check_for_locks(struct nfs4_file *fp, struct nfs4_lockowner *lowner);
74 static void nfs4_free_ol_stateid(struct nfs4_stid *stid);
78 /* Currently used for almost all code touching nfsv4 state: */
79 static DEFINE_MUTEX(client_mutex);
82 * Currently used for the del_recall_lru and file hash table. In an
83 * effort to decrease the scope of the client_mutex, this spinlock may
84 * eventually cover more:
86 static DEFINE_SPINLOCK(state_lock);
88 static struct kmem_cache *openowner_slab;
89 static struct kmem_cache *lockowner_slab;
90 static struct kmem_cache *file_slab;
91 static struct kmem_cache *stateid_slab;
92 static struct kmem_cache *deleg_slab;
97 mutex_lock(&client_mutex);
100 static void free_session(struct nfsd4_session *);
102 static bool is_session_dead(struct nfsd4_session *ses)
104 return ses->se_flags & NFS4_SESSION_DEAD;
107 static __be32 mark_session_dead_locked(struct nfsd4_session *ses, int ref_held_by_me)
109 if (atomic_read(&ses->se_ref) > ref_held_by_me)
110 return nfserr_jukebox;
111 ses->se_flags |= NFS4_SESSION_DEAD;
116 nfs4_unlock_state(void)
118 mutex_unlock(&client_mutex);
121 static bool is_client_expired(struct nfs4_client *clp)
123 return clp->cl_time == 0;
126 static __be32 mark_client_expired_locked(struct nfs4_client *clp)
128 if (atomic_read(&clp->cl_refcount))
129 return nfserr_jukebox;
134 static __be32 mark_client_expired(struct nfs4_client *clp)
136 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
139 spin_lock(&nn->client_lock);
140 ret = mark_client_expired_locked(clp);
141 spin_unlock(&nn->client_lock);
145 static __be32 get_client_locked(struct nfs4_client *clp)
147 if (is_client_expired(clp))
148 return nfserr_expired;
149 atomic_inc(&clp->cl_refcount);
153 /* must be called under the client_lock */
155 renew_client_locked(struct nfs4_client *clp)
157 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
159 if (is_client_expired(clp)) {
161 printk("%s: client (clientid %08x/%08x) already expired\n",
163 clp->cl_clientid.cl_boot,
164 clp->cl_clientid.cl_id);
168 dprintk("renewing client (clientid %08x/%08x)\n",
169 clp->cl_clientid.cl_boot,
170 clp->cl_clientid.cl_id);
171 list_move_tail(&clp->cl_lru, &nn->client_lru);
172 clp->cl_time = get_seconds();
176 renew_client(struct nfs4_client *clp)
178 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
180 spin_lock(&nn->client_lock);
181 renew_client_locked(clp);
182 spin_unlock(&nn->client_lock);
185 static void put_client_renew_locked(struct nfs4_client *clp)
187 if (!atomic_dec_and_test(&clp->cl_refcount))
189 if (!is_client_expired(clp))
190 renew_client_locked(clp);
193 static void put_client_renew(struct nfs4_client *clp)
195 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
197 if (!atomic_dec_and_lock(&clp->cl_refcount, &nn->client_lock))
199 if (!is_client_expired(clp))
200 renew_client_locked(clp);
201 spin_unlock(&nn->client_lock);
204 static __be32 nfsd4_get_session_locked(struct nfsd4_session *ses)
208 if (is_session_dead(ses))
209 return nfserr_badsession;
210 status = get_client_locked(ses->se_client);
213 atomic_inc(&ses->se_ref);
217 static void nfsd4_put_session_locked(struct nfsd4_session *ses)
219 struct nfs4_client *clp = ses->se_client;
221 if (atomic_dec_and_test(&ses->se_ref) && is_session_dead(ses))
223 put_client_renew_locked(clp);
226 static void nfsd4_put_session(struct nfsd4_session *ses)
228 struct nfs4_client *clp = ses->se_client;
229 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
231 spin_lock(&nn->client_lock);
232 nfsd4_put_session_locked(ses);
233 spin_unlock(&nn->client_lock);
238 opaque_hashval(const void *ptr, int nbytes)
240 unsigned char *cptr = (unsigned char *) ptr;
250 static void nfsd4_free_file(struct nfs4_file *f)
252 kmem_cache_free(file_slab, f);
256 put_nfs4_file(struct nfs4_file *fi)
258 might_lock(&state_lock);
260 if (atomic_dec_and_lock(&fi->fi_ref, &state_lock)) {
261 hlist_del(&fi->fi_hash);
262 spin_unlock(&state_lock);
268 get_nfs4_file(struct nfs4_file *fi)
270 atomic_inc(&fi->fi_ref);
274 __nfs4_get_fd(struct nfs4_file *f, int oflag)
276 if (f->fi_fds[oflag])
277 return get_file(f->fi_fds[oflag]);
282 find_writeable_file_locked(struct nfs4_file *f)
286 lockdep_assert_held(&f->fi_lock);
288 ret = __nfs4_get_fd(f, O_WRONLY);
290 ret = __nfs4_get_fd(f, O_RDWR);
295 find_writeable_file(struct nfs4_file *f)
299 spin_lock(&f->fi_lock);
300 ret = find_writeable_file_locked(f);
301 spin_unlock(&f->fi_lock);
306 static struct file *find_readable_file_locked(struct nfs4_file *f)
310 lockdep_assert_held(&f->fi_lock);
312 ret = __nfs4_get_fd(f, O_RDONLY);
314 ret = __nfs4_get_fd(f, O_RDWR);
319 find_readable_file(struct nfs4_file *f)
323 spin_lock(&f->fi_lock);
324 ret = find_readable_file_locked(f);
325 spin_unlock(&f->fi_lock);
331 find_any_file(struct nfs4_file *f)
335 spin_lock(&f->fi_lock);
336 ret = __nfs4_get_fd(f, O_RDWR);
338 ret = __nfs4_get_fd(f, O_WRONLY);
340 ret = __nfs4_get_fd(f, O_RDONLY);
342 spin_unlock(&f->fi_lock);
346 static atomic_long_t num_delegations;
347 unsigned long max_delegations;
350 * Open owner state (share locks)
353 /* hash tables for lock and open owners */
354 #define OWNER_HASH_BITS 8
355 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
356 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
358 static unsigned int ownerstr_hashval(u32 clientid, struct xdr_netobj *ownername)
362 ret = opaque_hashval(ownername->data, ownername->len);
364 return ret & OWNER_HASH_MASK;
367 /* hash table for nfs4_file */
368 #define FILE_HASH_BITS 8
369 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
371 static unsigned int nfsd_fh_hashval(struct knfsd_fh *fh)
373 return jhash2(fh->fh_base.fh_pad, XDR_QUADLEN(fh->fh_size), 0);
376 static unsigned int file_hashval(struct knfsd_fh *fh)
378 return nfsd_fh_hashval(fh) & (FILE_HASH_SIZE - 1);
381 static bool nfsd_fh_match(struct knfsd_fh *fh1, struct knfsd_fh *fh2)
383 return fh1->fh_size == fh2->fh_size &&
384 !memcmp(fh1->fh_base.fh_pad,
389 static struct hlist_head file_hashtbl[FILE_HASH_SIZE];
392 __nfs4_file_get_access(struct nfs4_file *fp, u32 access)
394 lockdep_assert_held(&fp->fi_lock);
396 if (access & NFS4_SHARE_ACCESS_WRITE)
397 atomic_inc(&fp->fi_access[O_WRONLY]);
398 if (access & NFS4_SHARE_ACCESS_READ)
399 atomic_inc(&fp->fi_access[O_RDONLY]);
403 nfs4_file_get_access(struct nfs4_file *fp, u32 access)
405 lockdep_assert_held(&fp->fi_lock);
407 /* Does this access mode make sense? */
408 if (access & ~NFS4_SHARE_ACCESS_BOTH)
411 /* Does it conflict with a deny mode already set? */
412 if ((access & fp->fi_share_deny) != 0)
413 return nfserr_share_denied;
415 __nfs4_file_get_access(fp, access);
419 static __be32 nfs4_file_check_deny(struct nfs4_file *fp, u32 deny)
421 /* Common case is that there is no deny mode. */
423 /* Does this deny mode make sense? */
424 if (deny & ~NFS4_SHARE_DENY_BOTH)
427 if ((deny & NFS4_SHARE_DENY_READ) &&
428 atomic_read(&fp->fi_access[O_RDONLY]))
429 return nfserr_share_denied;
431 if ((deny & NFS4_SHARE_DENY_WRITE) &&
432 atomic_read(&fp->fi_access[O_WRONLY]))
433 return nfserr_share_denied;
438 static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
440 might_lock(&fp->fi_lock);
442 if (atomic_dec_and_lock(&fp->fi_access[oflag], &fp->fi_lock)) {
443 struct file *f1 = NULL;
444 struct file *f2 = NULL;
446 swap(f1, fp->fi_fds[oflag]);
447 if (atomic_read(&fp->fi_access[1 - oflag]) == 0)
448 swap(f2, fp->fi_fds[O_RDWR]);
449 spin_unlock(&fp->fi_lock);
457 static void nfs4_file_put_access(struct nfs4_file *fp, u32 access)
459 WARN_ON_ONCE(access & ~NFS4_SHARE_ACCESS_BOTH);
461 if (access & NFS4_SHARE_ACCESS_WRITE)
462 __nfs4_file_put_access(fp, O_WRONLY);
463 if (access & NFS4_SHARE_ACCESS_READ)
464 __nfs4_file_put_access(fp, O_RDONLY);
467 static struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl,
468 struct kmem_cache *slab)
470 struct nfs4_stid *stid;
473 stid = kmem_cache_zalloc(slab, GFP_KERNEL);
477 idr_preload(GFP_KERNEL);
478 spin_lock(&cl->cl_lock);
479 new_id = idr_alloc_cyclic(&cl->cl_stateids, stid, 0, 0, GFP_NOWAIT);
480 spin_unlock(&cl->cl_lock);
484 stid->sc_client = cl;
485 stid->sc_stateid.si_opaque.so_id = new_id;
486 stid->sc_stateid.si_opaque.so_clid = cl->cl_clientid;
487 /* Will be incremented before return to client: */
488 atomic_set(&stid->sc_count, 1);
491 * It shouldn't be a problem to reuse an opaque stateid value.
492 * I don't think it is for 4.1. But with 4.0 I worry that, for
493 * example, a stray write retransmission could be accepted by
494 * the server when it should have been rejected. Therefore,
495 * adopt a trick from the sctp code to attempt to maximize the
496 * amount of time until an id is reused, by ensuring they always
497 * "increase" (mod INT_MAX):
501 kmem_cache_free(slab, stid);
505 static struct nfs4_ol_stateid * nfs4_alloc_open_stateid(struct nfs4_client *clp)
507 struct nfs4_stid *stid;
508 struct nfs4_ol_stateid *stp;
510 stid = nfs4_alloc_stid(clp, stateid_slab);
514 stp = openlockstateid(stid);
515 stp->st_stid.sc_free = nfs4_free_ol_stateid;
519 static void nfs4_free_deleg(struct nfs4_stid *stid)
521 kmem_cache_free(deleg_slab, stid);
522 atomic_long_dec(&num_delegations);
526 * When we recall a delegation, we should be careful not to hand it
527 * out again straight away.
528 * To ensure this we keep a pair of bloom filters ('new' and 'old')
529 * in which the filehandles of recalled delegations are "stored".
530 * If a filehandle appear in either filter, a delegation is blocked.
531 * When a delegation is recalled, the filehandle is stored in the "new"
533 * Every 30 seconds we swap the filters and clear the "new" one,
534 * unless both are empty of course.
536 * Each filter is 256 bits. We hash the filehandle to 32bit and use the
537 * low 3 bytes as hash-table indices.
539 * 'blocked_delegations_lock', which is always taken in block_delegations(),
540 * is used to manage concurrent access. Testing does not need the lock
541 * except when swapping the two filters.
543 static DEFINE_SPINLOCK(blocked_delegations_lock);
544 static struct bloom_pair {
545 int entries, old_entries;
547 int new; /* index into 'set' */
548 DECLARE_BITMAP(set[2], 256);
549 } blocked_delegations;
551 static int delegation_blocked(struct knfsd_fh *fh)
554 struct bloom_pair *bd = &blocked_delegations;
556 if (bd->entries == 0)
558 if (seconds_since_boot() - bd->swap_time > 30) {
559 spin_lock(&blocked_delegations_lock);
560 if (seconds_since_boot() - bd->swap_time > 30) {
561 bd->entries -= bd->old_entries;
562 bd->old_entries = bd->entries;
563 memset(bd->set[bd->new], 0,
566 bd->swap_time = seconds_since_boot();
568 spin_unlock(&blocked_delegations_lock);
570 hash = arch_fast_hash(&fh->fh_base, fh->fh_size, 0);
571 if (test_bit(hash&255, bd->set[0]) &&
572 test_bit((hash>>8)&255, bd->set[0]) &&
573 test_bit((hash>>16)&255, bd->set[0]))
576 if (test_bit(hash&255, bd->set[1]) &&
577 test_bit((hash>>8)&255, bd->set[1]) &&
578 test_bit((hash>>16)&255, bd->set[1]))
584 static void block_delegations(struct knfsd_fh *fh)
587 struct bloom_pair *bd = &blocked_delegations;
589 hash = arch_fast_hash(&fh->fh_base, fh->fh_size, 0);
591 spin_lock(&blocked_delegations_lock);
592 __set_bit(hash&255, bd->set[bd->new]);
593 __set_bit((hash>>8)&255, bd->set[bd->new]);
594 __set_bit((hash>>16)&255, bd->set[bd->new]);
595 if (bd->entries == 0)
596 bd->swap_time = seconds_since_boot();
598 spin_unlock(&blocked_delegations_lock);
601 static struct nfs4_delegation *
602 alloc_init_deleg(struct nfs4_client *clp, struct svc_fh *current_fh)
604 struct nfs4_delegation *dp;
607 dprintk("NFSD alloc_init_deleg\n");
608 n = atomic_long_inc_return(&num_delegations);
609 if (n < 0 || n > max_delegations)
611 if (delegation_blocked(¤t_fh->fh_handle))
613 dp = delegstateid(nfs4_alloc_stid(clp, deleg_slab));
617 dp->dl_stid.sc_free = nfs4_free_deleg;
619 * delegation seqid's are never incremented. The 4.1 special
620 * meaning of seqid 0 isn't meaningful, really, but let's avoid
621 * 0 anyway just for consistency and use 1:
623 dp->dl_stid.sc_stateid.si_generation = 1;
624 INIT_LIST_HEAD(&dp->dl_perfile);
625 INIT_LIST_HEAD(&dp->dl_perclnt);
626 INIT_LIST_HEAD(&dp->dl_recall_lru);
627 dp->dl_type = NFS4_OPEN_DELEGATE_READ;
628 INIT_WORK(&dp->dl_recall.cb_work, nfsd4_run_cb_recall);
631 atomic_long_dec(&num_delegations);
636 nfs4_put_stid(struct nfs4_stid *s)
638 struct nfs4_file *fp = s->sc_file;
639 struct nfs4_client *clp = s->sc_client;
641 might_lock(&clp->cl_lock);
643 if (!atomic_dec_and_lock(&s->sc_count, &clp->cl_lock))
645 idr_remove(&clp->cl_stateids, s->sc_stateid.si_opaque.so_id);
646 spin_unlock(&clp->cl_lock);
652 static void nfs4_put_deleg_lease(struct nfs4_file *fp)
654 lockdep_assert_held(&state_lock);
658 if (atomic_dec_and_test(&fp->fi_delegees)) {
659 vfs_setlease(fp->fi_deleg_file, F_UNLCK, &fp->fi_lease);
661 fput(fp->fi_deleg_file);
662 fp->fi_deleg_file = NULL;
666 static void unhash_stid(struct nfs4_stid *s)
672 hash_delegation_locked(struct nfs4_delegation *dp, struct nfs4_file *fp)
674 lockdep_assert_held(&state_lock);
675 lockdep_assert_held(&fp->fi_lock);
677 atomic_inc(&dp->dl_stid.sc_count);
678 dp->dl_stid.sc_type = NFS4_DELEG_STID;
679 list_add(&dp->dl_perfile, &fp->fi_delegations);
680 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
684 unhash_delegation_locked(struct nfs4_delegation *dp)
686 struct nfs4_file *fp = dp->dl_stid.sc_file;
688 lockdep_assert_held(&state_lock);
690 dp->dl_stid.sc_type = NFS4_CLOSED_DELEG_STID;
691 /* Ensure that deleg break won't try to requeue it */
693 spin_lock(&fp->fi_lock);
694 list_del_init(&dp->dl_perclnt);
695 list_del_init(&dp->dl_recall_lru);
696 list_del_init(&dp->dl_perfile);
697 spin_unlock(&fp->fi_lock);
699 nfs4_put_deleg_lease(fp);
702 static void destroy_delegation(struct nfs4_delegation *dp)
704 spin_lock(&state_lock);
705 unhash_delegation_locked(dp);
706 spin_unlock(&state_lock);
707 nfs4_put_stid(&dp->dl_stid);
710 static void revoke_delegation(struct nfs4_delegation *dp)
712 struct nfs4_client *clp = dp->dl_stid.sc_client;
714 WARN_ON(!list_empty(&dp->dl_recall_lru));
716 if (clp->cl_minorversion == 0)
717 nfs4_put_stid(&dp->dl_stid);
719 dp->dl_stid.sc_type = NFS4_REVOKED_DELEG_STID;
720 spin_lock(&clp->cl_lock);
721 list_add(&dp->dl_recall_lru, &clp->cl_revoked);
722 spin_unlock(&clp->cl_lock);
730 static unsigned int clientid_hashval(u32 id)
732 return id & CLIENT_HASH_MASK;
735 static unsigned int clientstr_hashval(const char *name)
737 return opaque_hashval(name, 8) & CLIENT_HASH_MASK;
741 * We store the NONE, READ, WRITE, and BOTH bits separately in the
742 * st_{access,deny}_bmap field of the stateid, in order to track not
743 * only what share bits are currently in force, but also what
744 * combinations of share bits previous opens have used. This allows us
745 * to enforce the recommendation of rfc 3530 14.2.19 that the server
746 * return an error if the client attempt to downgrade to a combination
747 * of share bits not explicable by closing some of its previous opens.
749 * XXX: This enforcement is actually incomplete, since we don't keep
750 * track of access/deny bit combinations; so, e.g., we allow:
752 * OPEN allow read, deny write
753 * OPEN allow both, deny none
754 * DOWNGRADE allow read, deny none
756 * which we should reject.
759 bmap_to_share_mode(unsigned long bmap) {
761 unsigned int access = 0;
763 for (i = 1; i < 4; i++) {
764 if (test_bit(i, &bmap))
770 /* set share access for a given stateid */
772 set_access(u32 access, struct nfs4_ol_stateid *stp)
774 unsigned char mask = 1 << access;
776 WARN_ON_ONCE(access > NFS4_SHARE_ACCESS_BOTH);
777 stp->st_access_bmap |= mask;
780 /* clear share access for a given stateid */
782 clear_access(u32 access, struct nfs4_ol_stateid *stp)
784 unsigned char mask = 1 << access;
786 WARN_ON_ONCE(access > NFS4_SHARE_ACCESS_BOTH);
787 stp->st_access_bmap &= ~mask;
790 /* test whether a given stateid has access */
792 test_access(u32 access, struct nfs4_ol_stateid *stp)
794 unsigned char mask = 1 << access;
796 return (bool)(stp->st_access_bmap & mask);
799 /* set share deny for a given stateid */
801 set_deny(u32 deny, struct nfs4_ol_stateid *stp)
803 unsigned char mask = 1 << deny;
805 WARN_ON_ONCE(deny > NFS4_SHARE_DENY_BOTH);
806 stp->st_deny_bmap |= mask;
809 /* clear share deny for a given stateid */
811 clear_deny(u32 deny, struct nfs4_ol_stateid *stp)
813 unsigned char mask = 1 << deny;
815 WARN_ON_ONCE(deny > NFS4_SHARE_DENY_BOTH);
816 stp->st_deny_bmap &= ~mask;
819 /* test whether a given stateid is denying specific access */
821 test_deny(u32 deny, struct nfs4_ol_stateid *stp)
823 unsigned char mask = 1 << deny;
825 return (bool)(stp->st_deny_bmap & mask);
828 static int nfs4_access_to_omode(u32 access)
830 switch (access & NFS4_SHARE_ACCESS_BOTH) {
831 case NFS4_SHARE_ACCESS_READ:
833 case NFS4_SHARE_ACCESS_WRITE:
835 case NFS4_SHARE_ACCESS_BOTH:
843 * A stateid that had a deny mode associated with it is being released
844 * or downgraded. Recalculate the deny mode on the file.
847 recalculate_deny_mode(struct nfs4_file *fp)
849 struct nfs4_ol_stateid *stp;
851 spin_lock(&fp->fi_lock);
852 fp->fi_share_deny = 0;
853 list_for_each_entry(stp, &fp->fi_stateids, st_perfile)
854 fp->fi_share_deny |= bmap_to_share_mode(stp->st_deny_bmap);
855 spin_unlock(&fp->fi_lock);
859 reset_union_bmap_deny(u32 deny, struct nfs4_ol_stateid *stp)
864 for (i = 1; i < 4; i++) {
865 if ((i & deny) != i) {
871 /* Recalculate per-file deny mode if there was a change */
873 recalculate_deny_mode(stp->st_stid.sc_file);
876 /* release all access and file references for a given stateid */
878 release_all_access(struct nfs4_ol_stateid *stp)
881 struct nfs4_file *fp = stp->st_stid.sc_file;
883 if (fp && stp->st_deny_bmap != 0)
884 recalculate_deny_mode(fp);
886 for (i = 1; i < 4; i++) {
887 if (test_access(i, stp))
888 nfs4_file_put_access(stp->st_stid.sc_file, i);
889 clear_access(i, stp);
893 static void nfs4_put_stateowner(struct nfs4_stateowner *sop)
895 if (!atomic_dec_and_test(&sop->so_count))
897 sop->so_ops->so_unhash(sop);
898 kfree(sop->so_owner.data);
899 sop->so_ops->so_free(sop);
902 static void unhash_generic_stateid(struct nfs4_ol_stateid *stp)
904 struct nfs4_file *fp = stp->st_stid.sc_file;
906 lockdep_assert_held(&stp->st_stateowner->so_client->cl_lock);
908 spin_lock(&fp->fi_lock);
909 list_del(&stp->st_perfile);
910 spin_unlock(&fp->fi_lock);
911 list_del(&stp->st_perstateowner);
914 static void nfs4_free_ol_stateid(struct nfs4_stid *stid)
916 struct nfs4_ol_stateid *stp = openlockstateid(stid);
918 release_all_access(stp);
919 kmem_cache_free(stateid_slab, stid);
922 static void nfs4_free_lock_stateid(struct nfs4_stid *stid)
924 struct nfs4_ol_stateid *stp = openlockstateid(stid);
925 struct nfs4_lockowner *lo = lockowner(stp->st_stateowner);
928 file = find_any_file(stp->st_stid.sc_file);
930 filp_close(file, (fl_owner_t)lo);
931 if (stp->st_stateowner)
932 nfs4_put_stateowner(stp->st_stateowner);
933 nfs4_free_ol_stateid(stid);
936 static void __release_lock_stateid(struct nfs4_ol_stateid *stp)
938 struct nfs4_openowner *oo = openowner(stp->st_openstp->st_stateowner);
940 spin_lock(&oo->oo_owner.so_client->cl_lock);
941 list_del(&stp->st_locks);
942 unhash_generic_stateid(stp);
943 unhash_stid(&stp->st_stid);
944 spin_unlock(&oo->oo_owner.so_client->cl_lock);
945 nfs4_put_stid(&stp->st_stid);
948 static void unhash_lockowner(struct nfs4_lockowner *lo)
950 list_del_init(&lo->lo_owner.so_strhash);
953 static void release_lockowner_stateids(struct nfs4_lockowner *lo)
955 struct nfs4_ol_stateid *stp;
957 while (!list_empty(&lo->lo_owner.so_stateids)) {
958 stp = list_first_entry(&lo->lo_owner.so_stateids,
959 struct nfs4_ol_stateid, st_perstateowner);
960 __release_lock_stateid(stp);
964 static void release_lockowner(struct nfs4_lockowner *lo)
966 unhash_lockowner(lo);
967 release_lockowner_stateids(lo);
968 nfs4_put_stateowner(&lo->lo_owner);
971 static void release_lockowner_if_empty(struct nfs4_lockowner *lo)
973 if (list_empty(&lo->lo_owner.so_stateids))
974 release_lockowner(lo);
977 static void release_lock_stateid(struct nfs4_ol_stateid *stp)
979 struct nfs4_lockowner *lo;
981 lo = lockowner(stp->st_stateowner);
982 __release_lock_stateid(stp);
983 release_lockowner_if_empty(lo);
986 static void release_open_stateid_locks(struct nfs4_ol_stateid *open_stp)
987 __releases(&open_stp->st_stateowner->so_client->cl_lock)
988 __acquires(&open_stp->st_stateowner->so_client->cl_lock)
990 struct nfs4_ol_stateid *stp;
992 while (!list_empty(&open_stp->st_locks)) {
993 stp = list_entry(open_stp->st_locks.next,
994 struct nfs4_ol_stateid, st_locks);
995 spin_unlock(&open_stp->st_stateowner->so_client->cl_lock);
996 release_lock_stateid(stp);
997 spin_lock(&open_stp->st_stateowner->so_client->cl_lock);
1001 static void unhash_open_stateid(struct nfs4_ol_stateid *stp)
1003 spin_lock(&stp->st_stateowner->so_client->cl_lock);
1004 unhash_generic_stateid(stp);
1005 release_open_stateid_locks(stp);
1006 spin_unlock(&stp->st_stateowner->so_client->cl_lock);
1009 static void release_open_stateid(struct nfs4_ol_stateid *stp)
1011 unhash_open_stateid(stp);
1012 nfs4_put_stid(&stp->st_stid);
1015 static void unhash_openowner(struct nfs4_openowner *oo)
1017 list_del_init(&oo->oo_owner.so_strhash);
1018 list_del_init(&oo->oo_perclient);
1021 static void release_last_closed_stateid(struct nfs4_openowner *oo)
1023 struct nfs4_ol_stateid *s = oo->oo_last_closed_stid;
1026 nfs4_put_stid(&s->st_stid);
1027 oo->oo_last_closed_stid = NULL;
1031 static void release_openowner_stateids(struct nfs4_openowner *oo)
1033 struct nfs4_ol_stateid *stp;
1035 while (!list_empty(&oo->oo_owner.so_stateids)) {
1036 stp = list_first_entry(&oo->oo_owner.so_stateids,
1037 struct nfs4_ol_stateid, st_perstateowner);
1038 release_open_stateid(stp);
1042 static void release_openowner(struct nfs4_openowner *oo)
1044 unhash_openowner(oo);
1045 release_openowner_stateids(oo);
1046 list_del(&oo->oo_close_lru);
1047 release_last_closed_stateid(oo);
1048 nfs4_put_stateowner(&oo->oo_owner);
1052 hash_sessionid(struct nfs4_sessionid *sessionid)
1054 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
1056 return sid->sequence % SESSION_HASH_SIZE;
1061 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
1063 u32 *ptr = (u32 *)(&sessionid->data[0]);
1064 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
1068 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
1074 * Bump the seqid on cstate->replay_owner, and clear replay_owner if it
1075 * won't be used for replay.
1077 void nfsd4_bump_seqid(struct nfsd4_compound_state *cstate, __be32 nfserr)
1079 struct nfs4_stateowner *so = cstate->replay_owner;
1081 if (nfserr == nfserr_replay_me)
1084 if (!seqid_mutating_err(ntohl(nfserr))) {
1085 nfsd4_cstate_clear_replay(cstate);
1090 if (so->so_is_open_owner)
1091 release_last_closed_stateid(openowner(so));
1097 gen_sessionid(struct nfsd4_session *ses)
1099 struct nfs4_client *clp = ses->se_client;
1100 struct nfsd4_sessionid *sid;
1102 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
1103 sid->clientid = clp->cl_clientid;
1104 sid->sequence = current_sessionid++;
1109 * The protocol defines ca_maxresponssize_cached to include the size of
1110 * the rpc header, but all we need to cache is the data starting after
1111 * the end of the initial SEQUENCE operation--the rest we regenerate
1112 * each time. Therefore we can advertise a ca_maxresponssize_cached
1113 * value that is the number of bytes in our cache plus a few additional
1114 * bytes. In order to stay on the safe side, and not promise more than
1115 * we can cache, those additional bytes must be the minimum possible: 24
1116 * bytes of rpc header (xid through accept state, with AUTH_NULL
1117 * verifier), 12 for the compound header (with zero-length tag), and 44
1118 * for the SEQUENCE op response:
1120 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
1123 free_session_slots(struct nfsd4_session *ses)
1127 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
1128 kfree(ses->se_slots[i]);
1132 * We don't actually need to cache the rpc and session headers, so we
1133 * can allocate a little less for each slot:
1135 static inline u32 slot_bytes(struct nfsd4_channel_attrs *ca)
1139 if (ca->maxresp_cached < NFSD_MIN_HDR_SEQ_SZ)
1142 size = ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
1143 return size + sizeof(struct nfsd4_slot);
1147 * XXX: If we run out of reserved DRC memory we could (up to a point)
1148 * re-negotiate active sessions and reduce their slot usage to make
1149 * room for new connections. For now we just fail the create session.
1151 static u32 nfsd4_get_drc_mem(struct nfsd4_channel_attrs *ca)
1153 u32 slotsize = slot_bytes(ca);
1154 u32 num = ca->maxreqs;
1157 spin_lock(&nfsd_drc_lock);
1158 avail = min((unsigned long)NFSD_MAX_MEM_PER_SESSION,
1159 nfsd_drc_max_mem - nfsd_drc_mem_used);
1160 num = min_t(int, num, avail / slotsize);
1161 nfsd_drc_mem_used += num * slotsize;
1162 spin_unlock(&nfsd_drc_lock);
1167 static void nfsd4_put_drc_mem(struct nfsd4_channel_attrs *ca)
1169 int slotsize = slot_bytes(ca);
1171 spin_lock(&nfsd_drc_lock);
1172 nfsd_drc_mem_used -= slotsize * ca->maxreqs;
1173 spin_unlock(&nfsd_drc_lock);
1176 static struct nfsd4_session *alloc_session(struct nfsd4_channel_attrs *fattrs,
1177 struct nfsd4_channel_attrs *battrs)
1179 int numslots = fattrs->maxreqs;
1180 int slotsize = slot_bytes(fattrs);
1181 struct nfsd4_session *new;
1184 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
1185 + sizeof(struct nfsd4_session) > PAGE_SIZE);
1186 mem = numslots * sizeof(struct nfsd4_slot *);
1188 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
1191 /* allocate each struct nfsd4_slot and data cache in one piece */
1192 for (i = 0; i < numslots; i++) {
1193 new->se_slots[i] = kzalloc(slotsize, GFP_KERNEL);
1194 if (!new->se_slots[i])
1198 memcpy(&new->se_fchannel, fattrs, sizeof(struct nfsd4_channel_attrs));
1199 memcpy(&new->se_bchannel, battrs, sizeof(struct nfsd4_channel_attrs));
1204 kfree(new->se_slots[i]);
1209 static void free_conn(struct nfsd4_conn *c)
1211 svc_xprt_put(c->cn_xprt);
1215 static void nfsd4_conn_lost(struct svc_xpt_user *u)
1217 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
1218 struct nfs4_client *clp = c->cn_session->se_client;
1220 spin_lock(&clp->cl_lock);
1221 if (!list_empty(&c->cn_persession)) {
1222 list_del(&c->cn_persession);
1225 nfsd4_probe_callback(clp);
1226 spin_unlock(&clp->cl_lock);
1229 static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
1231 struct nfsd4_conn *conn;
1233 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
1236 svc_xprt_get(rqstp->rq_xprt);
1237 conn->cn_xprt = rqstp->rq_xprt;
1238 conn->cn_flags = flags;
1239 INIT_LIST_HEAD(&conn->cn_xpt_user.list);
1243 static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
1245 conn->cn_session = ses;
1246 list_add(&conn->cn_persession, &ses->se_conns);
1249 static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
1251 struct nfs4_client *clp = ses->se_client;
1253 spin_lock(&clp->cl_lock);
1254 __nfsd4_hash_conn(conn, ses);
1255 spin_unlock(&clp->cl_lock);
1258 static int nfsd4_register_conn(struct nfsd4_conn *conn)
1260 conn->cn_xpt_user.callback = nfsd4_conn_lost;
1261 return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
1264 static void nfsd4_init_conn(struct svc_rqst *rqstp, struct nfsd4_conn *conn, struct nfsd4_session *ses)
1268 nfsd4_hash_conn(conn, ses);
1269 ret = nfsd4_register_conn(conn);
1271 /* oops; xprt is already down: */
1272 nfsd4_conn_lost(&conn->cn_xpt_user);
1273 /* We may have gained or lost a callback channel: */
1274 nfsd4_probe_callback_sync(ses->se_client);
1277 static struct nfsd4_conn *alloc_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_create_session *cses)
1279 u32 dir = NFS4_CDFC4_FORE;
1281 if (cses->flags & SESSION4_BACK_CHAN)
1282 dir |= NFS4_CDFC4_BACK;
1283 return alloc_conn(rqstp, dir);
1286 /* must be called under client_lock */
1287 static void nfsd4_del_conns(struct nfsd4_session *s)
1289 struct nfs4_client *clp = s->se_client;
1290 struct nfsd4_conn *c;
1292 spin_lock(&clp->cl_lock);
1293 while (!list_empty(&s->se_conns)) {
1294 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
1295 list_del_init(&c->cn_persession);
1296 spin_unlock(&clp->cl_lock);
1298 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
1301 spin_lock(&clp->cl_lock);
1303 spin_unlock(&clp->cl_lock);
1306 static void __free_session(struct nfsd4_session *ses)
1308 free_session_slots(ses);
1312 static void free_session(struct nfsd4_session *ses)
1314 struct nfsd_net *nn = net_generic(ses->se_client->net, nfsd_net_id);
1316 lockdep_assert_held(&nn->client_lock);
1317 nfsd4_del_conns(ses);
1318 nfsd4_put_drc_mem(&ses->se_fchannel);
1319 __free_session(ses);
1322 static void init_session(struct svc_rqst *rqstp, struct nfsd4_session *new, struct nfs4_client *clp, struct nfsd4_create_session *cses)
1325 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1327 new->se_client = clp;
1330 INIT_LIST_HEAD(&new->se_conns);
1332 new->se_cb_seq_nr = 1;
1333 new->se_flags = cses->flags;
1334 new->se_cb_prog = cses->callback_prog;
1335 new->se_cb_sec = cses->cb_sec;
1336 atomic_set(&new->se_ref, 0);
1337 idx = hash_sessionid(&new->se_sessionid);
1338 spin_lock(&nn->client_lock);
1339 list_add(&new->se_hash, &nn->sessionid_hashtbl[idx]);
1340 spin_lock(&clp->cl_lock);
1341 list_add(&new->se_perclnt, &clp->cl_sessions);
1342 spin_unlock(&clp->cl_lock);
1343 spin_unlock(&nn->client_lock);
1345 if (cses->flags & SESSION4_BACK_CHAN) {
1346 struct sockaddr *sa = svc_addr(rqstp);
1348 * This is a little silly; with sessions there's no real
1349 * use for the callback address. Use the peer address
1350 * as a reasonable default for now, but consider fixing
1351 * the rpc client not to require an address in the
1354 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
1355 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1359 /* caller must hold client_lock */
1360 static struct nfsd4_session *
1361 __find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net)
1363 struct nfsd4_session *elem;
1365 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1367 dump_sessionid(__func__, sessionid);
1368 idx = hash_sessionid(sessionid);
1369 /* Search in the appropriate list */
1370 list_for_each_entry(elem, &nn->sessionid_hashtbl[idx], se_hash) {
1371 if (!memcmp(elem->se_sessionid.data, sessionid->data,
1372 NFS4_MAX_SESSIONID_LEN)) {
1377 dprintk("%s: session not found\n", __func__);
1381 static struct nfsd4_session *
1382 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net,
1385 struct nfsd4_session *session;
1386 __be32 status = nfserr_badsession;
1388 session = __find_in_sessionid_hashtbl(sessionid, net);
1391 status = nfsd4_get_session_locked(session);
1399 /* caller must hold client_lock */
1401 unhash_session(struct nfsd4_session *ses)
1403 list_del(&ses->se_hash);
1404 spin_lock(&ses->se_client->cl_lock);
1405 list_del(&ses->se_perclnt);
1406 spin_unlock(&ses->se_client->cl_lock);
1409 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1411 STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
1413 if (clid->cl_boot == nn->boot_time)
1415 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1416 clid->cl_boot, clid->cl_id, nn->boot_time);
1421 * XXX Should we use a slab cache ?
1422 * This type of memory management is somewhat inefficient, but we use it
1423 * anyway since SETCLIENTID is not a common operation.
1425 static struct nfs4_client *alloc_client(struct xdr_netobj name)
1427 struct nfs4_client *clp;
1429 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
1432 clp->cl_name.data = kmemdup(name.data, name.len, GFP_KERNEL);
1433 if (clp->cl_name.data == NULL) {
1437 clp->cl_name.len = name.len;
1438 INIT_LIST_HEAD(&clp->cl_sessions);
1439 idr_init(&clp->cl_stateids);
1440 atomic_set(&clp->cl_refcount, 0);
1441 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1442 INIT_LIST_HEAD(&clp->cl_idhash);
1443 INIT_LIST_HEAD(&clp->cl_openowners);
1444 INIT_LIST_HEAD(&clp->cl_delegations);
1445 INIT_LIST_HEAD(&clp->cl_lru);
1446 INIT_LIST_HEAD(&clp->cl_callbacks);
1447 INIT_LIST_HEAD(&clp->cl_revoked);
1448 spin_lock_init(&clp->cl_lock);
1449 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1454 free_client(struct nfs4_client *clp)
1456 struct nfsd_net __maybe_unused *nn = net_generic(clp->net, nfsd_net_id);
1458 lockdep_assert_held(&nn->client_lock);
1459 while (!list_empty(&clp->cl_sessions)) {
1460 struct nfsd4_session *ses;
1461 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
1463 list_del(&ses->se_perclnt);
1464 WARN_ON_ONCE(atomic_read(&ses->se_ref));
1467 rpc_destroy_wait_queue(&clp->cl_cb_waitq);
1468 free_svc_cred(&clp->cl_cred);
1469 kfree(clp->cl_name.data);
1470 idr_destroy(&clp->cl_stateids);
1474 /* must be called under the client_lock */
1476 unhash_client_locked(struct nfs4_client *clp)
1478 struct nfsd4_session *ses;
1480 list_del(&clp->cl_lru);
1481 spin_lock(&clp->cl_lock);
1482 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1483 list_del_init(&ses->se_hash);
1484 spin_unlock(&clp->cl_lock);
1488 destroy_client(struct nfs4_client *clp)
1490 struct nfs4_openowner *oo;
1491 struct nfs4_delegation *dp;
1492 struct list_head reaplist;
1493 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1495 INIT_LIST_HEAD(&reaplist);
1496 spin_lock(&state_lock);
1497 while (!list_empty(&clp->cl_delegations)) {
1498 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1499 unhash_delegation_locked(dp);
1500 list_add(&dp->dl_recall_lru, &reaplist);
1502 spin_unlock(&state_lock);
1503 while (!list_empty(&reaplist)) {
1504 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1505 list_del_init(&dp->dl_recall_lru);
1506 nfs4_put_stid(&dp->dl_stid);
1508 while (!list_empty(&clp->cl_revoked)) {
1509 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1510 list_del_init(&dp->dl_recall_lru);
1511 nfs4_put_stid(&dp->dl_stid);
1513 while (!list_empty(&clp->cl_openowners)) {
1514 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1515 release_openowner(oo);
1517 nfsd4_shutdown_callback(clp);
1518 if (clp->cl_cb_conn.cb_xprt)
1519 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1520 list_del(&clp->cl_idhash);
1521 if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
1522 rb_erase(&clp->cl_namenode, &nn->conf_name_tree);
1524 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1525 spin_lock(&nn->client_lock);
1526 unhash_client_locked(clp);
1527 WARN_ON_ONCE(atomic_read(&clp->cl_refcount));
1529 spin_unlock(&nn->client_lock);
1532 static void expire_client(struct nfs4_client *clp)
1534 nfsd4_client_record_remove(clp);
1535 destroy_client(clp);
1538 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1540 memcpy(target->cl_verifier.data, source->data,
1541 sizeof(target->cl_verifier.data));
1544 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1546 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1547 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1550 static int copy_cred(struct svc_cred *target, struct svc_cred *source)
1552 if (source->cr_principal) {
1553 target->cr_principal =
1554 kstrdup(source->cr_principal, GFP_KERNEL);
1555 if (target->cr_principal == NULL)
1558 target->cr_principal = NULL;
1559 target->cr_flavor = source->cr_flavor;
1560 target->cr_uid = source->cr_uid;
1561 target->cr_gid = source->cr_gid;
1562 target->cr_group_info = source->cr_group_info;
1563 get_group_info(target->cr_group_info);
1564 target->cr_gss_mech = source->cr_gss_mech;
1565 if (source->cr_gss_mech)
1566 gss_mech_get(source->cr_gss_mech);
1571 compare_blob(const struct xdr_netobj *o1, const struct xdr_netobj *o2)
1575 res = o1->len - o2->len;
1578 return (long long)memcmp(o1->data, o2->data, o1->len);
1581 static int same_name(const char *n1, const char *n2)
1583 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1587 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1589 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1593 same_clid(clientid_t *cl1, clientid_t *cl2)
1595 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1598 static bool groups_equal(struct group_info *g1, struct group_info *g2)
1602 if (g1->ngroups != g2->ngroups)
1604 for (i=0; i<g1->ngroups; i++)
1605 if (!gid_eq(GROUP_AT(g1, i), GROUP_AT(g2, i)))
1611 * RFC 3530 language requires clid_inuse be returned when the
1612 * "principal" associated with a requests differs from that previously
1613 * used. We use uid, gid's, and gss principal string as our best
1614 * approximation. We also don't want to allow non-gss use of a client
1615 * established using gss: in theory cr_principal should catch that
1616 * change, but in practice cr_principal can be null even in the gss case
1617 * since gssd doesn't always pass down a principal string.
1619 static bool is_gss_cred(struct svc_cred *cr)
1621 /* Is cr_flavor one of the gss "pseudoflavors"?: */
1622 return (cr->cr_flavor > RPC_AUTH_MAXFLAVOR);
1627 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1629 if ((is_gss_cred(cr1) != is_gss_cred(cr2))
1630 || (!uid_eq(cr1->cr_uid, cr2->cr_uid))
1631 || (!gid_eq(cr1->cr_gid, cr2->cr_gid))
1632 || !groups_equal(cr1->cr_group_info, cr2->cr_group_info))
1634 if (cr1->cr_principal == cr2->cr_principal)
1636 if (!cr1->cr_principal || !cr2->cr_principal)
1638 return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
1641 static bool svc_rqst_integrity_protected(struct svc_rqst *rqstp)
1643 struct svc_cred *cr = &rqstp->rq_cred;
1646 if (!cr->cr_gss_mech)
1648 service = gss_pseudoflavor_to_service(cr->cr_gss_mech, cr->cr_flavor);
1649 return service == RPC_GSS_SVC_INTEGRITY ||
1650 service == RPC_GSS_SVC_PRIVACY;
1653 static bool mach_creds_match(struct nfs4_client *cl, struct svc_rqst *rqstp)
1655 struct svc_cred *cr = &rqstp->rq_cred;
1657 if (!cl->cl_mach_cred)
1659 if (cl->cl_cred.cr_gss_mech != cr->cr_gss_mech)
1661 if (!svc_rqst_integrity_protected(rqstp))
1663 if (!cr->cr_principal)
1665 return 0 == strcmp(cl->cl_cred.cr_principal, cr->cr_principal);
1668 static void gen_clid(struct nfs4_client *clp, struct nfsd_net *nn)
1670 static u32 current_clientid = 1;
1672 clp->cl_clientid.cl_boot = nn->boot_time;
1673 clp->cl_clientid.cl_id = current_clientid++;
1676 static void gen_confirm(struct nfs4_client *clp)
1682 * This is opaque to client, so no need to byte-swap. Use
1683 * __force to keep sparse happy
1685 verf[0] = (__force __be32)get_seconds();
1686 verf[1] = (__force __be32)i++;
1687 memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
1690 static struct nfs4_stid *
1691 find_stateid_locked(struct nfs4_client *cl, stateid_t *t)
1693 struct nfs4_stid *ret;
1695 ret = idr_find(&cl->cl_stateids, t->si_opaque.so_id);
1696 if (!ret || !ret->sc_type)
1701 static struct nfs4_stid *
1702 find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
1704 struct nfs4_stid *s;
1706 spin_lock(&cl->cl_lock);
1707 s = find_stateid_locked(cl, t);
1709 if (typemask & s->sc_type)
1710 atomic_inc(&s->sc_count);
1714 spin_unlock(&cl->cl_lock);
1718 static struct nfs4_client *create_client(struct xdr_netobj name,
1719 struct svc_rqst *rqstp, nfs4_verifier *verf)
1721 struct nfs4_client *clp;
1722 struct sockaddr *sa = svc_addr(rqstp);
1724 struct net *net = SVC_NET(rqstp);
1725 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1727 clp = alloc_client(name);
1731 ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1733 spin_lock(&nn->client_lock);
1735 spin_unlock(&nn->client_lock);
1738 INIT_WORK(&clp->cl_cb_null.cb_work, nfsd4_run_cb_null);
1739 clp->cl_time = get_seconds();
1740 clear_bit(0, &clp->cl_cb_slot_busy);
1741 copy_verf(clp, verf);
1742 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1744 clp->cl_cb_session = NULL;
1750 add_clp_to_name_tree(struct nfs4_client *new_clp, struct rb_root *root)
1752 struct rb_node **new = &(root->rb_node), *parent = NULL;
1753 struct nfs4_client *clp;
1756 clp = rb_entry(*new, struct nfs4_client, cl_namenode);
1759 if (compare_blob(&clp->cl_name, &new_clp->cl_name) > 0)
1760 new = &((*new)->rb_left);
1762 new = &((*new)->rb_right);
1765 rb_link_node(&new_clp->cl_namenode, parent, new);
1766 rb_insert_color(&new_clp->cl_namenode, root);
1769 static struct nfs4_client *
1770 find_clp_in_name_tree(struct xdr_netobj *name, struct rb_root *root)
1773 struct rb_node *node = root->rb_node;
1774 struct nfs4_client *clp;
1777 clp = rb_entry(node, struct nfs4_client, cl_namenode);
1778 cmp = compare_blob(&clp->cl_name, name);
1780 node = node->rb_left;
1782 node = node->rb_right;
1790 add_to_unconfirmed(struct nfs4_client *clp)
1792 unsigned int idhashval;
1793 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1795 clear_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1796 add_clp_to_name_tree(clp, &nn->unconf_name_tree);
1797 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1798 list_add(&clp->cl_idhash, &nn->unconf_id_hashtbl[idhashval]);
1803 move_to_confirmed(struct nfs4_client *clp)
1805 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1806 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1808 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1809 list_move(&clp->cl_idhash, &nn->conf_id_hashtbl[idhashval]);
1810 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1811 add_clp_to_name_tree(clp, &nn->conf_name_tree);
1812 set_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1816 static struct nfs4_client *
1817 find_client_in_id_table(struct list_head *tbl, clientid_t *clid, bool sessions)
1819 struct nfs4_client *clp;
1820 unsigned int idhashval = clientid_hashval(clid->cl_id);
1822 list_for_each_entry(clp, &tbl[idhashval], cl_idhash) {
1823 if (same_clid(&clp->cl_clientid, clid)) {
1824 if ((bool)clp->cl_minorversion != sessions)
1833 static struct nfs4_client *
1834 find_confirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1836 struct list_head *tbl = nn->conf_id_hashtbl;
1838 return find_client_in_id_table(tbl, clid, sessions);
1841 static struct nfs4_client *
1842 find_unconfirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1844 struct list_head *tbl = nn->unconf_id_hashtbl;
1846 return find_client_in_id_table(tbl, clid, sessions);
1849 static bool clp_used_exchangeid(struct nfs4_client *clp)
1851 return clp->cl_exchange_flags != 0;
1854 static struct nfs4_client *
1855 find_confirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1857 return find_clp_in_name_tree(name, &nn->conf_name_tree);
1860 static struct nfs4_client *
1861 find_unconfirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1863 return find_clp_in_name_tree(name, &nn->unconf_name_tree);
1867 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
1869 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1870 struct sockaddr *sa = svc_addr(rqstp);
1871 u32 scopeid = rpc_get_scope_id(sa);
1872 unsigned short expected_family;
1874 /* Currently, we only support tcp and tcp6 for the callback channel */
1875 if (se->se_callback_netid_len == 3 &&
1876 !memcmp(se->se_callback_netid_val, "tcp", 3))
1877 expected_family = AF_INET;
1878 else if (se->se_callback_netid_len == 4 &&
1879 !memcmp(se->se_callback_netid_val, "tcp6", 4))
1880 expected_family = AF_INET6;
1884 conn->cb_addrlen = rpc_uaddr2sockaddr(clp->net, se->se_callback_addr_val,
1885 se->se_callback_addr_len,
1886 (struct sockaddr *)&conn->cb_addr,
1887 sizeof(conn->cb_addr));
1889 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
1892 if (conn->cb_addr.ss_family == AF_INET6)
1893 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
1895 conn->cb_prog = se->se_callback_prog;
1896 conn->cb_ident = se->se_callback_ident;
1897 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
1900 conn->cb_addr.ss_family = AF_UNSPEC;
1901 conn->cb_addrlen = 0;
1902 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1903 "will not receive delegations\n",
1904 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1910 * Cache a reply. nfsd4_check_resp_size() has bounded the cache size.
1913 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1915 struct xdr_buf *buf = resp->xdr.buf;
1916 struct nfsd4_slot *slot = resp->cstate.slot;
1919 dprintk("--> %s slot %p\n", __func__, slot);
1921 slot->sl_opcnt = resp->opcnt;
1922 slot->sl_status = resp->cstate.status;
1924 slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
1925 if (nfsd4_not_cached(resp)) {
1926 slot->sl_datalen = 0;
1929 base = resp->cstate.data_offset;
1930 slot->sl_datalen = buf->len - base;
1931 if (read_bytes_from_xdr_buf(buf, base, slot->sl_data, slot->sl_datalen))
1932 WARN("%s: sessions DRC could not cache compound\n", __func__);
1937 * Encode the replay sequence operation from the slot values.
1938 * If cachethis is FALSE encode the uncached rep error on the next
1939 * operation which sets resp->p and increments resp->opcnt for
1940 * nfs4svc_encode_compoundres.
1944 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1945 struct nfsd4_compoundres *resp)
1947 struct nfsd4_op *op;
1948 struct nfsd4_slot *slot = resp->cstate.slot;
1950 /* Encode the replayed sequence operation */
1951 op = &args->ops[resp->opcnt - 1];
1952 nfsd4_encode_operation(resp, op);
1954 /* Return nfserr_retry_uncached_rep in next operation. */
1955 if (args->opcnt > 1 && !(slot->sl_flags & NFSD4_SLOT_CACHETHIS)) {
1956 op = &args->ops[resp->opcnt++];
1957 op->status = nfserr_retry_uncached_rep;
1958 nfsd4_encode_operation(resp, op);
1964 * The sequence operation is not cached because we can use the slot and
1968 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1969 struct nfsd4_sequence *seq)
1971 struct nfsd4_slot *slot = resp->cstate.slot;
1972 struct xdr_stream *xdr = &resp->xdr;
1976 dprintk("--> %s slot %p\n", __func__, slot);
1978 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1982 p = xdr_reserve_space(xdr, slot->sl_datalen);
1985 return nfserr_serverfault;
1987 xdr_encode_opaque_fixed(p, slot->sl_data, slot->sl_datalen);
1988 xdr_commit_encode(xdr);
1990 resp->opcnt = slot->sl_opcnt;
1991 return slot->sl_status;
1995 * Set the exchange_id flags returned by the server.
1998 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
2000 /* pNFS is not supported */
2001 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
2003 /* Referrals are supported, Migration is not. */
2004 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
2006 /* set the wire flags to return to client. */
2007 clid->flags = new->cl_exchange_flags;
2010 static bool client_has_state(struct nfs4_client *clp)
2013 * Note clp->cl_openowners check isn't quite right: there's no
2014 * need to count owners without stateid's.
2016 * Also note we should probably be using this in 4.0 case too.
2018 return !list_empty(&clp->cl_openowners)
2019 || !list_empty(&clp->cl_delegations)
2020 || !list_empty(&clp->cl_sessions);
2024 nfsd4_exchange_id(struct svc_rqst *rqstp,
2025 struct nfsd4_compound_state *cstate,
2026 struct nfsd4_exchange_id *exid)
2028 struct nfs4_client *unconf, *conf, *new;
2030 char addr_str[INET6_ADDRSTRLEN];
2031 nfs4_verifier verf = exid->verifier;
2032 struct sockaddr *sa = svc_addr(rqstp);
2033 bool update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
2034 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2036 rpc_ntop(sa, addr_str, sizeof(addr_str));
2037 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
2038 "ip_addr=%s flags %x, spa_how %d\n",
2039 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
2040 addr_str, exid->flags, exid->spa_how);
2042 if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
2043 return nfserr_inval;
2045 switch (exid->spa_how) {
2047 if (!svc_rqst_integrity_protected(rqstp))
2048 return nfserr_inval;
2051 default: /* checked by xdr code */
2054 return nfserr_encr_alg_unsupp;
2057 /* Cases below refer to rfc 5661 section 18.35.4: */
2059 conf = find_confirmed_client_by_name(&exid->clname, nn);
2061 bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
2062 bool verfs_match = same_verf(&verf, &conf->cl_verifier);
2065 if (!clp_used_exchangeid(conf)) { /* buggy client */
2066 status = nfserr_inval;
2069 if (!mach_creds_match(conf, rqstp)) {
2070 status = nfserr_wrong_cred;
2073 if (!creds_match) { /* case 9 */
2074 status = nfserr_perm;
2077 if (!verfs_match) { /* case 8 */
2078 status = nfserr_not_same;
2082 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
2086 if (!creds_match) { /* case 3 */
2087 if (client_has_state(conf)) {
2088 status = nfserr_clid_inuse;
2091 expire_client(conf);
2094 if (verfs_match) { /* case 2 */
2095 conf->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
2099 /* case 5, client reboot */
2103 if (update) { /* case 7 */
2104 status = nfserr_noent;
2108 unconf = find_unconfirmed_client_by_name(&exid->clname, nn);
2109 if (unconf) /* case 4, possible retry or client restart */
2110 expire_client(unconf);
2112 /* case 1 (normal case) */
2114 new = create_client(exid->clname, rqstp, &verf);
2116 status = nfserr_jukebox;
2119 new->cl_minorversion = cstate->minorversion;
2120 new->cl_mach_cred = (exid->spa_how == SP4_MACH_CRED);
2123 add_to_unconfirmed(new);
2125 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
2126 exid->clientid.cl_id = new->cl_clientid.cl_id;
2128 exid->seqid = new->cl_cs_slot.sl_seqid + 1;
2129 nfsd4_set_ex_flags(new, exid);
2131 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
2132 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
2136 nfs4_unlock_state();
2141 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
2143 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
2146 /* The slot is in use, and no response has been sent. */
2148 if (seqid == slot_seqid)
2149 return nfserr_jukebox;
2151 return nfserr_seq_misordered;
2153 /* Note unsigned 32-bit arithmetic handles wraparound: */
2154 if (likely(seqid == slot_seqid + 1))
2156 if (seqid == slot_seqid)
2157 return nfserr_replay_cache;
2158 return nfserr_seq_misordered;
2162 * Cache the create session result into the create session single DRC
2163 * slot cache by saving the xdr structure. sl_seqid has been set.
2164 * Do this for solo or embedded create session operations.
2167 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
2168 struct nfsd4_clid_slot *slot, __be32 nfserr)
2170 slot->sl_status = nfserr;
2171 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
2175 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
2176 struct nfsd4_clid_slot *slot)
2178 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
2179 return slot->sl_status;
2182 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
2183 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
2184 1 + /* MIN tag is length with zero, only length */ \
2185 3 + /* version, opcount, opcode */ \
2186 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
2187 /* seqid, slotID, slotID, cache */ \
2188 4 ) * sizeof(__be32))
2190 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
2191 2 + /* verifier: AUTH_NULL, length 0 */\
2193 1 + /* MIN tag is length with zero, only length */ \
2194 3 + /* opcount, opcode, opstatus*/ \
2195 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
2196 /* seqid, slotID, slotID, slotID, status */ \
2197 5 ) * sizeof(__be32))
2199 static __be32 check_forechannel_attrs(struct nfsd4_channel_attrs *ca, struct nfsd_net *nn)
2201 u32 maxrpc = nn->nfsd_serv->sv_max_mesg;
2203 if (ca->maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ)
2204 return nfserr_toosmall;
2205 if (ca->maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ)
2206 return nfserr_toosmall;
2207 ca->headerpadsz = 0;
2208 ca->maxreq_sz = min_t(u32, ca->maxreq_sz, maxrpc);
2209 ca->maxresp_sz = min_t(u32, ca->maxresp_sz, maxrpc);
2210 ca->maxops = min_t(u32, ca->maxops, NFSD_MAX_OPS_PER_COMPOUND);
2211 ca->maxresp_cached = min_t(u32, ca->maxresp_cached,
2212 NFSD_SLOT_CACHE_SIZE + NFSD_MIN_HDR_SEQ_SZ);
2213 ca->maxreqs = min_t(u32, ca->maxreqs, NFSD_MAX_SLOTS_PER_SESSION);
2215 * Note decreasing slot size below client's request may make it
2216 * difficult for client to function correctly, whereas
2217 * decreasing the number of slots will (just?) affect
2218 * performance. When short on memory we therefore prefer to
2219 * decrease number of slots instead of their size. Clients that
2220 * request larger slots than they need will get poor results:
2222 ca->maxreqs = nfsd4_get_drc_mem(ca);
2224 return nfserr_jukebox;
2229 #define NFSD_CB_MAX_REQ_SZ ((NFS4_enc_cb_recall_sz + \
2230 RPC_MAX_HEADER_WITH_AUTH) * sizeof(__be32))
2231 #define NFSD_CB_MAX_RESP_SZ ((NFS4_dec_cb_recall_sz + \
2232 RPC_MAX_REPHEADER_WITH_AUTH) * sizeof(__be32))
2234 static __be32 check_backchannel_attrs(struct nfsd4_channel_attrs *ca)
2236 ca->headerpadsz = 0;
2239 * These RPC_MAX_HEADER macros are overkill, especially since we
2240 * don't even do gss on the backchannel yet. But this is still
2241 * less than 1k. Tighten up this estimate in the unlikely event
2242 * it turns out to be a problem for some client:
2244 if (ca->maxreq_sz < NFSD_CB_MAX_REQ_SZ)
2245 return nfserr_toosmall;
2246 if (ca->maxresp_sz < NFSD_CB_MAX_RESP_SZ)
2247 return nfserr_toosmall;
2248 ca->maxresp_cached = 0;
2250 return nfserr_toosmall;
2255 static __be32 nfsd4_check_cb_sec(struct nfsd4_cb_sec *cbs)
2257 switch (cbs->flavor) {
2263 * GSS case: the spec doesn't allow us to return this
2264 * error. But it also doesn't allow us not to support
2266 * I'd rather this fail hard than return some error the
2267 * client might think it can already handle:
2269 return nfserr_encr_alg_unsupp;
2274 nfsd4_create_session(struct svc_rqst *rqstp,
2275 struct nfsd4_compound_state *cstate,
2276 struct nfsd4_create_session *cr_ses)
2278 struct sockaddr *sa = svc_addr(rqstp);
2279 struct nfs4_client *conf, *unconf;
2280 struct nfsd4_session *new;
2281 struct nfsd4_conn *conn;
2282 struct nfsd4_clid_slot *cs_slot = NULL;
2284 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2286 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
2287 return nfserr_inval;
2288 status = nfsd4_check_cb_sec(&cr_ses->cb_sec);
2291 status = check_forechannel_attrs(&cr_ses->fore_channel, nn);
2294 status = check_backchannel_attrs(&cr_ses->back_channel);
2296 goto out_release_drc_mem;
2297 status = nfserr_jukebox;
2298 new = alloc_session(&cr_ses->fore_channel, &cr_ses->back_channel);
2300 goto out_release_drc_mem;
2301 conn = alloc_conn_from_crses(rqstp, cr_ses);
2303 goto out_free_session;
2306 unconf = find_unconfirmed_client(&cr_ses->clientid, true, nn);
2307 conf = find_confirmed_client(&cr_ses->clientid, true, nn);
2308 WARN_ON_ONCE(conf && unconf);
2311 status = nfserr_wrong_cred;
2312 if (!mach_creds_match(conf, rqstp))
2314 cs_slot = &conf->cl_cs_slot;
2315 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
2316 if (status == nfserr_replay_cache) {
2317 status = nfsd4_replay_create_session(cr_ses, cs_slot);
2319 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
2320 status = nfserr_seq_misordered;
2323 } else if (unconf) {
2324 struct nfs4_client *old;
2325 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
2326 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
2327 status = nfserr_clid_inuse;
2330 status = nfserr_wrong_cred;
2331 if (!mach_creds_match(unconf, rqstp))
2333 cs_slot = &unconf->cl_cs_slot;
2334 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
2336 /* an unconfirmed replay returns misordered */
2337 status = nfserr_seq_misordered;
2340 old = find_confirmed_client_by_name(&unconf->cl_name, nn);
2342 status = mark_client_expired(old);
2347 move_to_confirmed(unconf);
2350 status = nfserr_stale_clientid;
2355 * We do not support RDMA or persistent sessions
2357 cr_ses->flags &= ~SESSION4_PERSIST;
2358 cr_ses->flags &= ~SESSION4_RDMA;
2360 init_session(rqstp, new, conf, cr_ses);
2361 nfsd4_init_conn(rqstp, conn, new);
2363 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
2364 NFS4_MAX_SESSIONID_LEN);
2365 cs_slot->sl_seqid++;
2366 cr_ses->seqid = cs_slot->sl_seqid;
2368 /* cache solo and embedded create sessions under the state lock */
2369 nfsd4_cache_create_session(cr_ses, cs_slot, status);
2370 nfs4_unlock_state();
2373 nfs4_unlock_state();
2376 __free_session(new);
2377 out_release_drc_mem:
2378 nfsd4_put_drc_mem(&cr_ses->fore_channel);
2382 static __be32 nfsd4_map_bcts_dir(u32 *dir)
2385 case NFS4_CDFC4_FORE:
2386 case NFS4_CDFC4_BACK:
2388 case NFS4_CDFC4_FORE_OR_BOTH:
2389 case NFS4_CDFC4_BACK_OR_BOTH:
2390 *dir = NFS4_CDFC4_BOTH;
2393 return nfserr_inval;
2396 __be32 nfsd4_backchannel_ctl(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_backchannel_ctl *bc)
2398 struct nfsd4_session *session = cstate->session;
2399 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2402 status = nfsd4_check_cb_sec(&bc->bc_cb_sec);
2405 spin_lock(&nn->client_lock);
2406 session->se_cb_prog = bc->bc_cb_program;
2407 session->se_cb_sec = bc->bc_cb_sec;
2408 spin_unlock(&nn->client_lock);
2410 nfsd4_probe_callback(session->se_client);
2415 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
2416 struct nfsd4_compound_state *cstate,
2417 struct nfsd4_bind_conn_to_session *bcts)
2420 struct nfsd4_conn *conn;
2421 struct nfsd4_session *session;
2422 struct net *net = SVC_NET(rqstp);
2423 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2425 if (!nfsd4_last_compound_op(rqstp))
2426 return nfserr_not_only_op;
2428 spin_lock(&nn->client_lock);
2429 session = find_in_sessionid_hashtbl(&bcts->sessionid, net, &status);
2430 spin_unlock(&nn->client_lock);
2432 goto out_no_session;
2433 status = nfserr_wrong_cred;
2434 if (!mach_creds_match(session->se_client, rqstp))
2436 status = nfsd4_map_bcts_dir(&bcts->dir);
2439 conn = alloc_conn(rqstp, bcts->dir);
2440 status = nfserr_jukebox;
2443 nfsd4_init_conn(rqstp, conn, session);
2446 nfsd4_put_session(session);
2448 nfs4_unlock_state();
2452 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
2456 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
2460 nfsd4_destroy_session(struct svc_rqst *r,
2461 struct nfsd4_compound_state *cstate,
2462 struct nfsd4_destroy_session *sessionid)
2464 struct nfsd4_session *ses;
2466 int ref_held_by_me = 0;
2467 struct net *net = SVC_NET(r);
2468 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2471 status = nfserr_not_only_op;
2472 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
2473 if (!nfsd4_last_compound_op(r))
2477 dump_sessionid(__func__, &sessionid->sessionid);
2478 spin_lock(&nn->client_lock);
2479 ses = find_in_sessionid_hashtbl(&sessionid->sessionid, net, &status);
2481 goto out_client_lock;
2482 status = nfserr_wrong_cred;
2483 if (!mach_creds_match(ses->se_client, r))
2484 goto out_put_session;
2485 status = mark_session_dead_locked(ses, 1 + ref_held_by_me);
2487 goto out_put_session;
2488 unhash_session(ses);
2489 spin_unlock(&nn->client_lock);
2491 nfsd4_probe_callback_sync(ses->se_client);
2493 spin_lock(&nn->client_lock);
2496 nfsd4_put_session_locked(ses);
2498 spin_unlock(&nn->client_lock);
2500 nfs4_unlock_state();
2504 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
2506 struct nfsd4_conn *c;
2508 list_for_each_entry(c, &s->se_conns, cn_persession) {
2509 if (c->cn_xprt == xpt) {
2516 static __be32 nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
2518 struct nfs4_client *clp = ses->se_client;
2519 struct nfsd4_conn *c;
2520 __be32 status = nfs_ok;
2523 spin_lock(&clp->cl_lock);
2524 c = __nfsd4_find_conn(new->cn_xprt, ses);
2527 status = nfserr_conn_not_bound_to_session;
2528 if (clp->cl_mach_cred)
2530 __nfsd4_hash_conn(new, ses);
2531 spin_unlock(&clp->cl_lock);
2532 ret = nfsd4_register_conn(new);
2534 /* oops; xprt is already down: */
2535 nfsd4_conn_lost(&new->cn_xpt_user);
2538 spin_unlock(&clp->cl_lock);
2543 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
2545 struct nfsd4_compoundargs *args = rqstp->rq_argp;
2547 return args->opcnt > session->se_fchannel.maxops;
2550 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
2551 struct nfsd4_session *session)
2553 struct xdr_buf *xb = &rqstp->rq_arg;
2555 return xb->len > session->se_fchannel.maxreq_sz;
2559 nfsd4_sequence(struct svc_rqst *rqstp,
2560 struct nfsd4_compound_state *cstate,
2561 struct nfsd4_sequence *seq)
2563 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2564 struct xdr_stream *xdr = &resp->xdr;
2565 struct nfsd4_session *session;
2566 struct nfs4_client *clp;
2567 struct nfsd4_slot *slot;
2568 struct nfsd4_conn *conn;
2571 struct net *net = SVC_NET(rqstp);
2572 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2574 if (resp->opcnt != 1)
2575 return nfserr_sequence_pos;
2578 * Will be either used or freed by nfsd4_sequence_check_conn
2581 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
2583 return nfserr_jukebox;
2585 spin_lock(&nn->client_lock);
2586 session = find_in_sessionid_hashtbl(&seq->sessionid, net, &status);
2588 goto out_no_session;
2589 clp = session->se_client;
2591 status = nfserr_too_many_ops;
2592 if (nfsd4_session_too_many_ops(rqstp, session))
2593 goto out_put_session;
2595 status = nfserr_req_too_big;
2596 if (nfsd4_request_too_big(rqstp, session))
2597 goto out_put_session;
2599 status = nfserr_badslot;
2600 if (seq->slotid >= session->se_fchannel.maxreqs)
2601 goto out_put_session;
2603 slot = session->se_slots[seq->slotid];
2604 dprintk("%s: slotid %d\n", __func__, seq->slotid);
2606 /* We do not negotiate the number of slots yet, so set the
2607 * maxslots to the session maxreqs which is used to encode
2608 * sr_highest_slotid and the sr_target_slot id to maxslots */
2609 seq->maxslots = session->se_fchannel.maxreqs;
2611 status = check_slot_seqid(seq->seqid, slot->sl_seqid,
2612 slot->sl_flags & NFSD4_SLOT_INUSE);
2613 if (status == nfserr_replay_cache) {
2614 status = nfserr_seq_misordered;
2615 if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
2616 goto out_put_session;
2617 cstate->slot = slot;
2618 cstate->session = session;
2620 /* Return the cached reply status and set cstate->status
2621 * for nfsd4_proc_compound processing */
2622 status = nfsd4_replay_cache_entry(resp, seq);
2623 cstate->status = nfserr_replay_cache;
2627 goto out_put_session;
2629 status = nfsd4_sequence_check_conn(conn, session);
2632 goto out_put_session;
2634 buflen = (seq->cachethis) ?
2635 session->se_fchannel.maxresp_cached :
2636 session->se_fchannel.maxresp_sz;
2637 status = (seq->cachethis) ? nfserr_rep_too_big_to_cache :
2639 if (xdr_restrict_buflen(xdr, buflen - rqstp->rq_auth_slack))
2640 goto out_put_session;
2641 svc_reserve(rqstp, buflen);
2644 /* Success! bump slot seqid */
2645 slot->sl_seqid = seq->seqid;
2646 slot->sl_flags |= NFSD4_SLOT_INUSE;
2648 slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
2650 slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
2652 cstate->slot = slot;
2653 cstate->session = session;
2657 switch (clp->cl_cb_state) {
2659 seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
2661 case NFSD4_CB_FAULT:
2662 seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
2665 seq->status_flags = 0;
2667 if (!list_empty(&clp->cl_revoked))
2668 seq->status_flags |= SEQ4_STATUS_RECALLABLE_STATE_REVOKED;
2672 spin_unlock(&nn->client_lock);
2675 nfsd4_put_session_locked(session);
2676 goto out_no_session;
2680 nfsd4_sequence_done(struct nfsd4_compoundres *resp)
2682 struct nfsd4_compound_state *cs = &resp->cstate;
2684 if (nfsd4_has_session(cs)) {
2685 if (cs->status != nfserr_replay_cache) {
2686 nfsd4_store_cache_entry(resp);
2687 cs->slot->sl_flags &= ~NFSD4_SLOT_INUSE;
2689 /* Drop session reference that was taken in nfsd4_sequence() */
2690 nfsd4_put_session(cs->session);
2692 put_client_renew(cs->clp);
2696 nfsd4_destroy_clientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_destroy_clientid *dc)
2698 struct nfs4_client *conf, *unconf, *clp;
2700 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2703 unconf = find_unconfirmed_client(&dc->clientid, true, nn);
2704 conf = find_confirmed_client(&dc->clientid, true, nn);
2705 WARN_ON_ONCE(conf && unconf);
2710 if (client_has_state(conf)) {
2711 status = nfserr_clientid_busy;
2717 status = nfserr_stale_clientid;
2720 if (!mach_creds_match(clp, rqstp)) {
2721 status = nfserr_wrong_cred;
2726 nfs4_unlock_state();
2731 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
2735 if (rc->rca_one_fs) {
2736 if (!cstate->current_fh.fh_dentry)
2737 return nfserr_nofilehandle;
2739 * We don't take advantage of the rca_one_fs case.
2740 * That's OK, it's optional, we can safely ignore it.
2746 status = nfserr_complete_already;
2747 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
2748 &cstate->session->se_client->cl_flags))
2751 status = nfserr_stale_clientid;
2752 if (is_client_expired(cstate->session->se_client))
2754 * The following error isn't really legal.
2755 * But we only get here if the client just explicitly
2756 * destroyed the client. Surely it no longer cares what
2757 * error it gets back on an operation for the dead
2763 nfsd4_client_record_create(cstate->session->se_client);
2765 nfs4_unlock_state();
2770 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2771 struct nfsd4_setclientid *setclid)
2773 struct xdr_netobj clname = setclid->se_name;
2774 nfs4_verifier clverifier = setclid->se_verf;
2775 struct nfs4_client *conf, *unconf, *new;
2777 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2779 /* Cases below refer to rfc 3530 section 14.2.33: */
2781 conf = find_confirmed_client_by_name(&clname, nn);
2784 status = nfserr_clid_inuse;
2785 if (clp_used_exchangeid(conf))
2787 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
2788 char addr_str[INET6_ADDRSTRLEN];
2789 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
2791 dprintk("NFSD: setclientid: string in use by client "
2792 "at %s\n", addr_str);
2796 unconf = find_unconfirmed_client_by_name(&clname, nn);
2798 expire_client(unconf);
2799 status = nfserr_jukebox;
2800 new = create_client(clname, rqstp, &clverifier);
2803 if (conf && same_verf(&conf->cl_verifier, &clverifier))
2804 /* case 1: probable callback update */
2805 copy_clid(new, conf);
2806 else /* case 4 (new client) or cases 2, 3 (client reboot): */
2808 new->cl_minorversion = 0;
2809 gen_callback(new, setclid, rqstp);
2810 add_to_unconfirmed(new);
2811 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
2812 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
2813 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
2816 nfs4_unlock_state();
2822 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
2823 struct nfsd4_compound_state *cstate,
2824 struct nfsd4_setclientid_confirm *setclientid_confirm)
2826 struct nfs4_client *conf, *unconf;
2827 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
2828 clientid_t * clid = &setclientid_confirm->sc_clientid;
2830 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2832 if (STALE_CLIENTID(clid, nn))
2833 return nfserr_stale_clientid;
2836 conf = find_confirmed_client(clid, false, nn);
2837 unconf = find_unconfirmed_client(clid, false, nn);
2839 * We try hard to give out unique clientid's, so if we get an
2840 * attempt to confirm the same clientid with a different cred,
2841 * there's a bug somewhere. Let's charitably assume it's our
2844 status = nfserr_serverfault;
2845 if (unconf && !same_creds(&unconf->cl_cred, &rqstp->rq_cred))
2847 if (conf && !same_creds(&conf->cl_cred, &rqstp->rq_cred))
2849 /* cases below refer to rfc 3530 section 14.2.34: */
2850 if (!unconf || !same_verf(&confirm, &unconf->cl_confirm)) {
2851 if (conf && !unconf) /* case 2: probable retransmit */
2853 else /* case 4: client hasn't noticed we rebooted yet? */
2854 status = nfserr_stale_clientid;
2858 if (conf) { /* case 1: callback update */
2859 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
2860 nfsd4_probe_callback(conf);
2861 expire_client(unconf);
2862 } else { /* case 3: normal case; new or rebooted client */
2863 conf = find_confirmed_client_by_name(&unconf->cl_name, nn);
2865 status = mark_client_expired(conf);
2868 expire_client(conf);
2870 move_to_confirmed(unconf);
2871 nfsd4_probe_callback(unconf);
2874 nfs4_unlock_state();
2878 static struct nfs4_file *nfsd4_alloc_file(void)
2880 return kmem_cache_alloc(file_slab, GFP_KERNEL);
2883 /* OPEN Share state helper functions */
2884 static void nfsd4_init_file(struct nfs4_file *fp, struct knfsd_fh *fh)
2886 unsigned int hashval = file_hashval(fh);
2888 lockdep_assert_held(&state_lock);
2890 atomic_set(&fp->fi_ref, 1);
2891 spin_lock_init(&fp->fi_lock);
2892 INIT_LIST_HEAD(&fp->fi_stateids);
2893 INIT_LIST_HEAD(&fp->fi_delegations);
2894 fh_copy_shallow(&fp->fi_fhandle, fh);
2895 fp->fi_had_conflict = false;
2896 fp->fi_lease = NULL;
2897 fp->fi_share_deny = 0;
2898 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
2899 memset(fp->fi_access, 0, sizeof(fp->fi_access));
2900 hlist_add_head(&fp->fi_hash, &file_hashtbl[hashval]);
2904 nfsd4_free_slabs(void)
2906 kmem_cache_destroy(openowner_slab);
2907 kmem_cache_destroy(lockowner_slab);
2908 kmem_cache_destroy(file_slab);
2909 kmem_cache_destroy(stateid_slab);
2910 kmem_cache_destroy(deleg_slab);
2914 nfsd4_init_slabs(void)
2916 openowner_slab = kmem_cache_create("nfsd4_openowners",
2917 sizeof(struct nfs4_openowner), 0, 0, NULL);
2918 if (openowner_slab == NULL)
2920 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
2921 sizeof(struct nfs4_lockowner), 0, 0, NULL);
2922 if (lockowner_slab == NULL)
2923 goto out_free_openowner_slab;
2924 file_slab = kmem_cache_create("nfsd4_files",
2925 sizeof(struct nfs4_file), 0, 0, NULL);
2926 if (file_slab == NULL)
2927 goto out_free_lockowner_slab;
2928 stateid_slab = kmem_cache_create("nfsd4_stateids",
2929 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
2930 if (stateid_slab == NULL)
2931 goto out_free_file_slab;
2932 deleg_slab = kmem_cache_create("nfsd4_delegations",
2933 sizeof(struct nfs4_delegation), 0, 0, NULL);
2934 if (deleg_slab == NULL)
2935 goto out_free_stateid_slab;
2938 out_free_stateid_slab:
2939 kmem_cache_destroy(stateid_slab);
2941 kmem_cache_destroy(file_slab);
2942 out_free_lockowner_slab:
2943 kmem_cache_destroy(lockowner_slab);
2944 out_free_openowner_slab:
2945 kmem_cache_destroy(openowner_slab);
2947 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2951 static void init_nfs4_replay(struct nfs4_replay *rp)
2953 rp->rp_status = nfserr_serverfault;
2955 rp->rp_buf = rp->rp_ibuf;
2956 mutex_init(&rp->rp_mutex);
2959 static void nfsd4_cstate_assign_replay(struct nfsd4_compound_state *cstate,
2960 struct nfs4_stateowner *so)
2962 if (!nfsd4_has_session(cstate)) {
2963 mutex_lock(&so->so_replay.rp_mutex);
2964 cstate->replay_owner = so;
2965 atomic_inc(&so->so_count);
2969 void nfsd4_cstate_clear_replay(struct nfsd4_compound_state *cstate)
2971 struct nfs4_stateowner *so = cstate->replay_owner;
2974 cstate->replay_owner = NULL;
2975 mutex_unlock(&so->so_replay.rp_mutex);
2976 nfs4_put_stateowner(so);
2980 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
2982 struct nfs4_stateowner *sop;
2984 sop = kmem_cache_alloc(slab, GFP_KERNEL);
2988 sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
2989 if (!sop->so_owner.data) {
2990 kmem_cache_free(slab, sop);
2993 sop->so_owner.len = owner->len;
2995 INIT_LIST_HEAD(&sop->so_stateids);
2996 sop->so_client = clp;
2997 init_nfs4_replay(&sop->so_replay);
2998 atomic_set(&sop->so_count, 1);
3002 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
3004 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
3006 list_add(&oo->oo_owner.so_strhash, &nn->ownerstr_hashtbl[strhashval]);
3007 list_add(&oo->oo_perclient, &clp->cl_openowners);
3010 static void nfs4_unhash_openowner(struct nfs4_stateowner *so)
3012 struct nfs4_openowner *oo = openowner(so);
3014 unhash_openowner(oo);
3017 static void nfs4_free_openowner(struct nfs4_stateowner *so)
3019 struct nfs4_openowner *oo = openowner(so);
3021 kmem_cache_free(openowner_slab, oo);
3024 static const struct nfs4_stateowner_operations openowner_ops = {
3025 .so_unhash = nfs4_unhash_openowner,
3026 .so_free = nfs4_free_openowner,
3029 static struct nfs4_openowner *
3030 alloc_init_open_stateowner(unsigned int strhashval, struct nfsd4_open *open,
3031 struct nfsd4_compound_state *cstate)
3033 struct nfs4_client *clp = cstate->clp;
3034 struct nfs4_openowner *oo;
3036 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
3039 oo->oo_owner.so_ops = &openowner_ops;
3040 oo->oo_owner.so_is_open_owner = 1;
3041 oo->oo_owner.so_seqid = open->op_seqid;
3042 oo->oo_flags = NFS4_OO_NEW;
3043 if (nfsd4_has_session(cstate))
3044 oo->oo_flags |= NFS4_OO_CONFIRMED;
3046 oo->oo_last_closed_stid = NULL;
3047 INIT_LIST_HEAD(&oo->oo_close_lru);
3048 hash_openowner(oo, clp, strhashval);
3052 static void init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
3053 struct nfs4_openowner *oo = open->op_openowner;
3055 atomic_inc(&stp->st_stid.sc_count);
3056 stp->st_stid.sc_type = NFS4_OPEN_STID;
3057 INIT_LIST_HEAD(&stp->st_locks);
3058 stp->st_stateowner = &oo->oo_owner;
3060 stp->st_stid.sc_file = fp;
3061 stp->st_access_bmap = 0;
3062 stp->st_deny_bmap = 0;
3063 stp->st_openstp = NULL;
3064 spin_lock(&oo->oo_owner.so_client->cl_lock);
3065 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
3066 spin_lock(&fp->fi_lock);
3067 list_add(&stp->st_perfile, &fp->fi_stateids);
3068 spin_unlock(&fp->fi_lock);
3069 spin_unlock(&oo->oo_owner.so_client->cl_lock);
3073 move_to_close_lru(struct nfs4_openowner *oo, struct net *net)
3075 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3077 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
3079 list_move_tail(&oo->oo_close_lru, &nn->close_lru);
3080 oo->oo_time = get_seconds();
3084 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
3087 return (sop->so_owner.len == owner->len) &&
3088 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
3089 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
3092 static struct nfs4_openowner *
3093 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open,
3094 bool sessions, struct nfsd_net *nn)
3096 struct nfs4_stateowner *so;
3097 struct nfs4_openowner *oo;
3098 struct nfs4_client *clp;
3100 list_for_each_entry(so, &nn->ownerstr_hashtbl[hashval], so_strhash) {
3101 if (!so->so_is_open_owner)
3103 if (same_owner_str(so, &open->op_owner, &open->op_clientid)) {
3105 clp = oo->oo_owner.so_client;
3106 if ((bool)clp->cl_minorversion != sessions)
3108 renew_client(oo->oo_owner.so_client);
3115 /* search file_hashtbl[] for file */
3116 static struct nfs4_file *
3117 find_file_locked(struct knfsd_fh *fh)
3119 unsigned int hashval = file_hashval(fh);
3120 struct nfs4_file *fp;
3122 lockdep_assert_held(&state_lock);
3124 hlist_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
3125 if (nfsd_fh_match(&fp->fi_fhandle, fh)) {
3133 static struct nfs4_file *
3134 find_file(struct knfsd_fh *fh)
3136 struct nfs4_file *fp;
3138 spin_lock(&state_lock);
3139 fp = find_file_locked(fh);
3140 spin_unlock(&state_lock);
3144 static struct nfs4_file *
3145 find_or_add_file(struct nfs4_file *new, struct knfsd_fh *fh)
3147 struct nfs4_file *fp;
3149 spin_lock(&state_lock);
3150 fp = find_file_locked(fh);
3152 nfsd4_init_file(new, fh);
3155 spin_unlock(&state_lock);
3161 * Called to check deny when READ with all zero stateid or
3162 * WRITE with all zero or all one stateid
3165 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
3167 struct nfs4_file *fp;
3168 __be32 ret = nfs_ok;
3170 fp = find_file(¤t_fh->fh_handle);
3173 /* Check for conflicting share reservations */
3174 spin_lock(&fp->fi_lock);
3175 if (fp->fi_share_deny & deny_type)
3176 ret = nfserr_locked;
3177 spin_unlock(&fp->fi_lock);
3182 void nfsd4_prepare_cb_recall(struct nfs4_delegation *dp)
3184 struct nfsd_net *nn = net_generic(dp->dl_stid.sc_client->net,
3187 block_delegations(&dp->dl_stid.sc_file->fi_fhandle);
3190 * We can't do this in nfsd_break_deleg_cb because it is
3191 * already holding inode->i_lock.
3193 * If the dl_time != 0, then we know that it has already been
3194 * queued for a lease break. Don't queue it again.
3196 spin_lock(&state_lock);
3197 if (dp->dl_time == 0) {
3198 dp->dl_time = get_seconds();
3199 list_add_tail(&dp->dl_recall_lru, &nn->del_recall_lru);
3201 spin_unlock(&state_lock);
3204 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
3207 * We're assuming the state code never drops its reference
3208 * without first removing the lease. Since we're in this lease
3209 * callback (and since the lease code is serialized by the kernel
3210 * lock) we know the server hasn't removed the lease yet, we know
3211 * it's safe to take a reference.
3213 atomic_inc(&dp->dl_stid.sc_count);
3214 nfsd4_cb_recall(dp);
3217 /* Called from break_lease() with i_lock held. */
3218 static void nfsd_break_deleg_cb(struct file_lock *fl)
3220 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
3221 struct nfs4_delegation *dp;
3224 WARN(1, "(%p)->fl_owner NULL\n", fl);
3227 if (fp->fi_had_conflict) {
3228 WARN(1, "duplicate break on %p\n", fp);
3232 * We don't want the locks code to timeout the lease for us;
3233 * we'll remove it ourself if a delegation isn't returned
3236 fl->fl_break_time = 0;
3238 spin_lock(&fp->fi_lock);
3239 fp->fi_had_conflict = true;
3241 * If there are no delegations on the list, then we can't count on this
3242 * lease ever being cleaned up. Set the fl_break_time to jiffies so that
3243 * time_out_leases will do it ASAP. The fact that fi_had_conflict is now
3244 * true should keep any new delegations from being hashed.
3246 if (list_empty(&fp->fi_delegations))
3247 fl->fl_break_time = jiffies;
3249 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
3250 nfsd_break_one_deleg(dp);
3251 spin_unlock(&fp->fi_lock);
3255 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
3258 return lease_modify(onlist, arg);
3263 static const struct lock_manager_operations nfsd_lease_mng_ops = {
3264 .lm_break = nfsd_break_deleg_cb,
3265 .lm_change = nfsd_change_deleg_cb,
3268 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
3270 if (nfsd4_has_session(cstate))
3272 if (seqid == so->so_seqid - 1)
3273 return nfserr_replay_me;
3274 if (seqid == so->so_seqid)
3276 return nfserr_bad_seqid;
3279 static __be32 lookup_clientid(clientid_t *clid,
3280 struct nfsd4_compound_state *cstate,
3281 struct nfsd_net *nn)
3283 struct nfs4_client *found;
3286 found = cstate->clp;
3287 if (!same_clid(&found->cl_clientid, clid))
3288 return nfserr_stale_clientid;
3292 if (STALE_CLIENTID(clid, nn))
3293 return nfserr_stale_clientid;
3296 * For v4.1+ we get the client in the SEQUENCE op. If we don't have one
3297 * cached already then we know this is for is for v4.0 and "sessions"
3300 WARN_ON_ONCE(cstate->session);
3301 found = find_confirmed_client(clid, false, nn);
3303 return nfserr_expired;
3305 /* Cache the nfs4_client in cstate! */
3306 cstate->clp = found;
3307 atomic_inc(&found->cl_refcount);
3312 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
3313 struct nfsd4_open *open, struct nfsd_net *nn)
3315 clientid_t *clientid = &open->op_clientid;
3316 struct nfs4_client *clp = NULL;
3317 unsigned int strhashval;
3318 struct nfs4_openowner *oo = NULL;
3321 if (STALE_CLIENTID(&open->op_clientid, nn))
3322 return nfserr_stale_clientid;
3324 * In case we need it later, after we've already created the
3325 * file and don't want to risk a further failure:
3327 open->op_file = nfsd4_alloc_file();
3328 if (open->op_file == NULL)
3329 return nfserr_jukebox;
3331 status = lookup_clientid(clientid, cstate, nn);
3336 strhashval = ownerstr_hashval(clientid->cl_id, &open->op_owner);
3337 oo = find_openstateowner_str(strhashval, open, cstate->minorversion, nn);
3338 open->op_openowner = oo;
3342 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
3343 /* Replace unconfirmed owners without checking for replay. */
3344 release_openowner(oo);
3345 open->op_openowner = NULL;
3348 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
3353 oo = alloc_init_open_stateowner(strhashval, open, cstate);
3355 return nfserr_jukebox;
3356 open->op_openowner = oo;
3358 open->op_stp = nfs4_alloc_open_stateid(clp);
3360 return nfserr_jukebox;
3364 static inline __be32
3365 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
3367 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
3368 return nfserr_openmode;
3373 static int share_access_to_flags(u32 share_access)
3375 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
3378 static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
3380 struct nfs4_stid *ret;
3382 ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
3385 return delegstateid(ret);
3388 static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
3390 return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
3391 open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
3395 nfs4_check_deleg(struct nfs4_client *cl, struct nfsd4_open *open,
3396 struct nfs4_delegation **dp)
3399 __be32 status = nfserr_bad_stateid;
3400 struct nfs4_delegation *deleg;
3402 deleg = find_deleg_stateid(cl, &open->op_delegate_stateid);
3405 flags = share_access_to_flags(open->op_share_access);
3406 status = nfs4_check_delegmode(deleg, flags);
3408 nfs4_put_stid(&deleg->dl_stid);
3413 if (!nfsd4_is_deleg_cur(open))
3417 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3421 static struct nfs4_ol_stateid *
3422 nfsd4_find_existing_open(struct nfs4_file *fp, struct nfsd4_open *open)
3424 struct nfs4_ol_stateid *local, *ret = NULL;
3425 struct nfs4_openowner *oo = open->op_openowner;
3427 spin_lock(&fp->fi_lock);
3428 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
3429 /* ignore lock owners */
3430 if (local->st_stateowner->so_is_open_owner == 0)
3432 if (local->st_stateowner == &oo->oo_owner) {
3434 atomic_inc(&ret->st_stid.sc_count);
3438 spin_unlock(&fp->fi_lock);
3442 static inline int nfs4_access_to_access(u32 nfs4_access)
3446 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
3447 flags |= NFSD_MAY_READ;
3448 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
3449 flags |= NFSD_MAY_WRITE;
3453 static inline __be32
3454 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
3455 struct nfsd4_open *open)
3457 struct iattr iattr = {
3458 .ia_valid = ATTR_SIZE,
3461 if (!open->op_truncate)
3463 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
3464 return nfserr_inval;
3465 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
3468 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
3469 struct svc_fh *cur_fh, struct nfs4_ol_stateid *stp,
3470 struct nfsd4_open *open)
3472 struct file *filp = NULL;
3474 int oflag = nfs4_access_to_omode(open->op_share_access);
3475 int access = nfs4_access_to_access(open->op_share_access);
3476 unsigned char old_access_bmap, old_deny_bmap;
3478 spin_lock(&fp->fi_lock);
3481 * Are we trying to set a deny mode that would conflict with
3484 status = nfs4_file_check_deny(fp, open->op_share_deny);
3485 if (status != nfs_ok) {
3486 spin_unlock(&fp->fi_lock);
3490 /* set access to the file */
3491 status = nfs4_file_get_access(fp, open->op_share_access);
3492 if (status != nfs_ok) {
3493 spin_unlock(&fp->fi_lock);
3497 /* Set access bits in stateid */
3498 old_access_bmap = stp->st_access_bmap;
3499 set_access(open->op_share_access, stp);
3501 /* Set new deny mask */
3502 old_deny_bmap = stp->st_deny_bmap;
3503 set_deny(open->op_share_deny, stp);
3504 fp->fi_share_deny |= (open->op_share_deny & NFS4_SHARE_DENY_BOTH);
3506 if (!fp->fi_fds[oflag]) {
3507 spin_unlock(&fp->fi_lock);
3508 status = nfsd_open(rqstp, cur_fh, S_IFREG, access, &filp);
3510 goto out_put_access;
3511 spin_lock(&fp->fi_lock);
3512 if (!fp->fi_fds[oflag]) {
3513 fp->fi_fds[oflag] = filp;
3517 spin_unlock(&fp->fi_lock);
3521 status = nfsd4_truncate(rqstp, cur_fh, open);
3523 goto out_put_access;
3527 stp->st_access_bmap = old_access_bmap;
3528 nfs4_file_put_access(fp, open->op_share_access);
3529 reset_union_bmap_deny(bmap_to_share_mode(old_deny_bmap), stp);
3534 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)
3537 unsigned char old_deny_bmap;
3539 if (!test_access(open->op_share_access, stp))
3540 return nfs4_get_vfs_file(rqstp, fp, cur_fh, stp, open);
3542 /* test and set deny mode */
3543 spin_lock(&fp->fi_lock);
3544 status = nfs4_file_check_deny(fp, open->op_share_deny);
3545 if (status == nfs_ok) {
3546 old_deny_bmap = stp->st_deny_bmap;
3547 set_deny(open->op_share_deny, stp);
3548 fp->fi_share_deny |=
3549 (open->op_share_deny & NFS4_SHARE_DENY_BOTH);
3551 spin_unlock(&fp->fi_lock);
3553 if (status != nfs_ok)
3556 status = nfsd4_truncate(rqstp, cur_fh, open);
3557 if (status != nfs_ok)
3558 reset_union_bmap_deny(old_deny_bmap, stp);
3563 nfs4_set_claim_prev(struct nfsd4_open *open, bool has_session)
3565 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3568 /* Should we give out recallable state?: */
3569 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
3571 if (clp->cl_cb_state == NFSD4_CB_UP)
3574 * In the sessions case, since we don't have to establish a
3575 * separate connection for callbacks, we assume it's OK
3576 * until we hear otherwise:
3578 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
3581 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_file *fp, int flag)
3583 struct file_lock *fl;
3585 fl = locks_alloc_lock();
3588 locks_init_lock(fl);
3589 fl->fl_lmops = &nfsd_lease_mng_ops;
3590 fl->fl_flags = FL_DELEG;
3591 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
3592 fl->fl_end = OFFSET_MAX;
3593 fl->fl_owner = (fl_owner_t)fp;
3594 fl->fl_pid = current->tgid;
3598 static int nfs4_setlease(struct nfs4_delegation *dp)
3600 struct nfs4_file *fp = dp->dl_stid.sc_file;
3601 struct file_lock *fl;
3605 fl = nfs4_alloc_init_lease(fp, NFS4_OPEN_DELEGATE_READ);
3608 filp = find_readable_file(fp);
3610 /* We should always have a readable file here */
3615 status = vfs_setlease(filp, fl->fl_type, &fl);
3617 locks_free_lock(fl);
3620 spin_lock(&state_lock);
3621 spin_lock(&fp->fi_lock);
3622 /* Did the lease get broken before we took the lock? */
3624 if (fp->fi_had_conflict)
3629 atomic_inc(&fp->fi_delegees);
3630 hash_delegation_locked(dp, fp);
3634 fp->fi_deleg_file = filp;
3635 atomic_set(&fp->fi_delegees, 1);
3636 hash_delegation_locked(dp, fp);
3637 spin_unlock(&fp->fi_lock);
3638 spin_unlock(&state_lock);
3641 spin_unlock(&fp->fi_lock);
3642 spin_unlock(&state_lock);
3648 static struct nfs4_delegation *
3649 nfs4_set_delegation(struct nfs4_client *clp, struct svc_fh *fh,
3650 struct nfs4_file *fp)
3653 struct nfs4_delegation *dp;
3655 if (fp->fi_had_conflict)
3656 return ERR_PTR(-EAGAIN);
3658 dp = alloc_init_deleg(clp, fh);
3660 return ERR_PTR(-ENOMEM);
3663 spin_lock(&state_lock);
3664 spin_lock(&fp->fi_lock);
3665 dp->dl_stid.sc_file = fp;
3666 if (!fp->fi_lease) {
3667 spin_unlock(&fp->fi_lock);
3668 spin_unlock(&state_lock);
3669 status = nfs4_setlease(dp);
3672 atomic_inc(&fp->fi_delegees);
3673 if (fp->fi_had_conflict) {
3677 hash_delegation_locked(dp, fp);
3680 spin_unlock(&fp->fi_lock);
3681 spin_unlock(&state_lock);
3684 nfs4_put_stid(&dp->dl_stid);
3685 return ERR_PTR(status);
3690 static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
3692 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3693 if (status == -EAGAIN)
3694 open->op_why_no_deleg = WND4_CONTENTION;
3696 open->op_why_no_deleg = WND4_RESOURCE;
3697 switch (open->op_deleg_want) {
3698 case NFS4_SHARE_WANT_READ_DELEG:
3699 case NFS4_SHARE_WANT_WRITE_DELEG:
3700 case NFS4_SHARE_WANT_ANY_DELEG:
3702 case NFS4_SHARE_WANT_CANCEL:
3703 open->op_why_no_deleg = WND4_CANCELLED;
3705 case NFS4_SHARE_WANT_NO_DELEG:
3712 * Attempt to hand out a delegation.
3714 * Note we don't support write delegations, and won't until the vfs has
3715 * proper support for them.
3718 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open,
3719 struct nfs4_ol_stateid *stp)
3721 struct nfs4_delegation *dp;
3722 struct nfs4_openowner *oo = openowner(stp->st_stateowner);
3723 struct nfs4_client *clp = stp->st_stid.sc_client;
3727 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
3728 open->op_recall = 0;
3729 switch (open->op_claim_type) {
3730 case NFS4_OPEN_CLAIM_PREVIOUS:
3732 open->op_recall = 1;
3733 if (open->op_delegate_type != NFS4_OPEN_DELEGATE_READ)
3736 case NFS4_OPEN_CLAIM_NULL:
3737 case NFS4_OPEN_CLAIM_FH:
3739 * Let's not give out any delegations till everyone's
3740 * had the chance to reclaim theirs....
3742 if (locks_in_grace(clp->net))
3744 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
3747 * Also, if the file was opened for write or
3748 * create, there's a good chance the client's
3749 * about to write to it, resulting in an
3750 * immediate recall (since we don't support
3751 * write delegations):
3753 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
3755 if (open->op_create == NFS4_OPEN_CREATE)
3761 dp = nfs4_set_delegation(clp, fh, stp->st_stid.sc_file);
3765 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
3767 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
3768 STATEID_VAL(&dp->dl_stid.sc_stateid));
3769 open->op_delegate_type = NFS4_OPEN_DELEGATE_READ;
3770 nfs4_put_stid(&dp->dl_stid);
3773 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE;
3774 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
3775 open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE) {
3776 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
3777 open->op_recall = 1;
3780 /* 4.1 client asking for a delegation? */
3781 if (open->op_deleg_want)
3782 nfsd4_open_deleg_none_ext(open, status);
3786 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
3787 struct nfs4_delegation *dp)
3789 if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
3790 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3791 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3792 open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
3793 } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
3794 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3795 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3796 open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
3798 /* Otherwise the client must be confused wanting a delegation
3799 * it already has, therefore we don't return
3800 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
3805 * called with nfs4_lock_state() held.
3808 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
3810 struct nfsd4_compoundres *resp = rqstp->rq_resp;
3811 struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
3812 struct nfs4_file *fp = NULL;
3813 struct nfs4_ol_stateid *stp = NULL;
3814 struct nfs4_delegation *dp = NULL;
3818 * Lookup file; if found, lookup stateid and check open request,
3819 * and check for delegations in the process of being recalled.
3820 * If not found, create the nfs4_file struct
3822 fp = find_or_add_file(open->op_file, ¤t_fh->fh_handle);
3823 if (fp != open->op_file) {
3824 status = nfs4_check_deleg(cl, open, &dp);
3827 stp = nfsd4_find_existing_open(fp, open);
3829 open->op_file = NULL;
3830 status = nfserr_bad_stateid;
3831 if (nfsd4_is_deleg_cur(open))
3833 status = nfserr_jukebox;
3837 * OPEN the file, or upgrade an existing OPEN.
3838 * If truncate fails, the OPEN fails.
3841 /* Stateid was found, this is an OPEN upgrade */
3842 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
3847 open->op_stp = NULL;
3848 init_open_stateid(stp, fp, open);
3849 status = nfs4_get_vfs_file(rqstp, fp, current_fh, stp, open);
3851 release_open_stateid(stp);
3855 update_stateid(&stp->st_stid.sc_stateid);
3856 memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3858 if (nfsd4_has_session(&resp->cstate)) {
3859 if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
3860 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3861 open->op_why_no_deleg = WND4_NOT_WANTED;
3867 * Attempt to hand out a delegation. No error return, because the
3868 * OPEN succeeds even if we fail.
3870 nfs4_open_delegation(current_fh, open, stp);
3874 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
3875 STATEID_VAL(&stp->st_stid.sc_stateid));
3877 /* 4.1 client trying to upgrade/downgrade delegation? */
3878 if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
3879 open->op_deleg_want)
3880 nfsd4_deleg_xgrade_none_ext(open, dp);
3884 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
3885 nfs4_set_claim_prev(open, nfsd4_has_session(&resp->cstate));
3887 * To finish the open response, we just need to set the rflags.
3889 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
3890 if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
3891 !nfsd4_has_session(&resp->cstate))
3892 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
3894 nfs4_put_stid(&dp->dl_stid);
3896 nfs4_put_stid(&stp->st_stid);
3901 void nfsd4_cleanup_open_state(struct nfsd4_compound_state *cstate,
3902 struct nfsd4_open *open, __be32 status)
3904 if (open->op_openowner) {
3905 struct nfs4_openowner *oo = open->op_openowner;
3907 if (!list_empty(&oo->oo_owner.so_stateids))
3908 list_del_init(&oo->oo_close_lru);
3909 if (oo->oo_flags & NFS4_OO_NEW) {
3911 release_openowner(oo);
3912 open->op_openowner = NULL;
3914 oo->oo_flags &= ~NFS4_OO_NEW;
3916 if (open->op_openowner)
3917 nfsd4_cstate_assign_replay(cstate, &oo->oo_owner);
3920 nfsd4_free_file(open->op_file);
3922 nfs4_put_stid(&open->op_stp->st_stid);
3926 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3929 struct nfs4_client *clp;
3931 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3934 dprintk("process_renew(%08x/%08x): starting\n",
3935 clid->cl_boot, clid->cl_id);
3936 status = lookup_clientid(clid, cstate, nn);
3940 status = nfserr_cb_path_down;
3941 if (!list_empty(&clp->cl_delegations)
3942 && clp->cl_cb_state != NFSD4_CB_UP)
3946 nfs4_unlock_state();
3951 nfsd4_end_grace(struct nfsd_net *nn)
3953 /* do nothing if grace period already ended */
3954 if (nn->grace_ended)
3957 dprintk("NFSD: end of grace period\n");
3958 nn->grace_ended = true;
3959 nfsd4_record_grace_done(nn, nn->boot_time);
3960 locks_end_grace(&nn->nfsd4_manager);
3962 * Now that every NFSv4 client has had the chance to recover and
3963 * to see the (possibly new, possibly shorter) lease time, we
3964 * can safely set the next grace time to the current lease time:
3966 nn->nfsd4_grace = nn->nfsd4_lease;
3970 nfs4_laundromat(struct nfsd_net *nn)
3972 struct nfs4_client *clp;
3973 struct nfs4_openowner *oo;
3974 struct nfs4_delegation *dp;
3975 struct list_head *pos, *next, reaplist;
3976 time_t cutoff = get_seconds() - nn->nfsd4_lease;
3977 time_t t, new_timeo = nn->nfsd4_lease;
3981 dprintk("NFSD: laundromat service - starting\n");
3982 nfsd4_end_grace(nn);
3983 INIT_LIST_HEAD(&reaplist);
3984 spin_lock(&nn->client_lock);
3985 list_for_each_safe(pos, next, &nn->client_lru) {
3986 clp = list_entry(pos, struct nfs4_client, cl_lru);
3987 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
3988 t = clp->cl_time - cutoff;
3989 new_timeo = min(new_timeo, t);
3992 if (mark_client_expired_locked(clp)) {
3993 dprintk("NFSD: client in use (clientid %08x)\n",
3994 clp->cl_clientid.cl_id);
3997 list_move(&clp->cl_lru, &reaplist);
3999 spin_unlock(&nn->client_lock);
4000 list_for_each_safe(pos, next, &reaplist) {
4001 clp = list_entry(pos, struct nfs4_client, cl_lru);
4002 dprintk("NFSD: purging unused client (clientid %08x)\n",
4003 clp->cl_clientid.cl_id);
4006 spin_lock(&state_lock);
4007 list_for_each_safe(pos, next, &nn->del_recall_lru) {
4008 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4009 if (net_generic(dp->dl_stid.sc_client->net, nfsd_net_id) != nn)
4011 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
4012 t = dp->dl_time - cutoff;
4013 new_timeo = min(new_timeo, t);
4016 unhash_delegation_locked(dp);
4017 list_add(&dp->dl_recall_lru, &reaplist);
4019 spin_unlock(&state_lock);
4020 while (!list_empty(&reaplist)) {
4021 dp = list_first_entry(&reaplist, struct nfs4_delegation,
4023 list_del_init(&dp->dl_recall_lru);
4024 revoke_delegation(dp);
4026 list_for_each_safe(pos, next, &nn->close_lru) {
4027 oo = container_of(pos, struct nfs4_openowner, oo_close_lru);
4028 if (time_after((unsigned long)oo->oo_time, (unsigned long)cutoff)) {
4029 t = oo->oo_time - cutoff;
4030 new_timeo = min(new_timeo, t);
4033 release_openowner(oo);
4035 new_timeo = max_t(time_t, new_timeo, NFSD_LAUNDROMAT_MINTIMEOUT);
4036 nfs4_unlock_state();
4040 static struct workqueue_struct *laundry_wq;
4041 static void laundromat_main(struct work_struct *);
4044 laundromat_main(struct work_struct *laundry)
4047 struct delayed_work *dwork = container_of(laundry, struct delayed_work,
4049 struct nfsd_net *nn = container_of(dwork, struct nfsd_net,
4052 t = nfs4_laundromat(nn);
4053 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
4054 queue_delayed_work(laundry_wq, &nn->laundromat_work, t*HZ);
4057 static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *stp)
4059 if (!nfsd_fh_match(&fhp->fh_handle, &stp->st_stid.sc_file->fi_fhandle))
4060 return nfserr_bad_stateid;
4065 access_permit_read(struct nfs4_ol_stateid *stp)
4067 return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
4068 test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
4069 test_access(NFS4_SHARE_ACCESS_WRITE, stp);
4073 access_permit_write(struct nfs4_ol_stateid *stp)
4075 return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
4076 test_access(NFS4_SHARE_ACCESS_BOTH, stp);
4080 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
4082 __be32 status = nfserr_openmode;
4084 /* For lock stateid's, we test the parent open, not the lock: */
4085 if (stp->st_openstp)
4086 stp = stp->st_openstp;
4087 if ((flags & WR_STATE) && !access_permit_write(stp))
4089 if ((flags & RD_STATE) && !access_permit_read(stp))
4096 static inline __be32
4097 check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
4099 if (ONE_STATEID(stateid) && (flags & RD_STATE))
4101 else if (locks_in_grace(net)) {
4102 /* Answer in remaining cases depends on existence of
4103 * conflicting state; so we must wait out the grace period. */
4104 return nfserr_grace;
4105 } else if (flags & WR_STATE)
4106 return nfs4_share_conflict(current_fh,
4107 NFS4_SHARE_DENY_WRITE);
4108 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
4109 return nfs4_share_conflict(current_fh,
4110 NFS4_SHARE_DENY_READ);
4114 * Allow READ/WRITE during grace period on recovered state only for files
4115 * that are not able to provide mandatory locking.
4118 grace_disallows_io(struct net *net, struct inode *inode)
4120 return locks_in_grace(net) && mandatory_lock(inode);
4123 /* Returns true iff a is later than b: */
4124 static bool stateid_generation_after(stateid_t *a, stateid_t *b)
4126 return (s32)(a->si_generation - b->si_generation) > 0;
4129 static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
4132 * When sessions are used the stateid generation number is ignored
4135 if (has_session && in->si_generation == 0)
4138 if (in->si_generation == ref->si_generation)
4141 /* If the client sends us a stateid from the future, it's buggy: */
4142 if (stateid_generation_after(in, ref))
4143 return nfserr_bad_stateid;
4145 * However, we could see a stateid from the past, even from a
4146 * non-buggy client. For example, if the client sends a lock
4147 * while some IO is outstanding, the lock may bump si_generation
4148 * while the IO is still in flight. The client could avoid that
4149 * situation by waiting for responses on all the IO requests,
4150 * but better performance may result in retrying IO that
4151 * receives an old_stateid error if requests are rarely
4152 * reordered in flight:
4154 return nfserr_old_stateid;
4157 static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
4159 struct nfs4_stid *s;
4160 struct nfs4_ol_stateid *ols;
4161 __be32 status = nfserr_bad_stateid;
4163 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
4165 /* Client debugging aid. */
4166 if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid)) {
4167 char addr_str[INET6_ADDRSTRLEN];
4168 rpc_ntop((struct sockaddr *)&cl->cl_addr, addr_str,
4170 pr_warn_ratelimited("NFSD: client %s testing state ID "
4171 "with incorrect client ID\n", addr_str);
4174 spin_lock(&cl->cl_lock);
4175 s = find_stateid_locked(cl, stateid);
4178 status = check_stateid_generation(stateid, &s->sc_stateid, 1);
4181 switch (s->sc_type) {
4182 case NFS4_DELEG_STID:
4185 case NFS4_REVOKED_DELEG_STID:
4186 status = nfserr_deleg_revoked;
4188 case NFS4_OPEN_STID:
4189 case NFS4_LOCK_STID:
4190 ols = openlockstateid(s);
4191 if (ols->st_stateowner->so_is_open_owner
4192 && !(openowner(ols->st_stateowner)->oo_flags
4193 & NFS4_OO_CONFIRMED))
4194 status = nfserr_bad_stateid;
4199 printk("unknown stateid type %x\n", s->sc_type);
4201 case NFS4_CLOSED_STID:
4202 case NFS4_CLOSED_DELEG_STID:
4203 status = nfserr_bad_stateid;
4206 spin_unlock(&cl->cl_lock);
4211 nfsd4_lookup_stateid(struct nfsd4_compound_state *cstate,
4212 stateid_t *stateid, unsigned char typemask,
4213 struct nfs4_stid **s, struct nfsd_net *nn)
4217 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
4218 return nfserr_bad_stateid;
4219 status = lookup_clientid(&stateid->si_opaque.so_clid, cstate, nn);
4220 if (status == nfserr_stale_clientid) {
4221 if (cstate->session)
4222 return nfserr_bad_stateid;
4223 return nfserr_stale_stateid;
4227 *s = find_stateid_by_type(cstate->clp, stateid, typemask);
4229 return nfserr_bad_stateid;
4234 * Checks for stateid operations
4237 nfs4_preprocess_stateid_op(struct net *net, struct nfsd4_compound_state *cstate,
4238 stateid_t *stateid, int flags, struct file **filpp)
4240 struct nfs4_stid *s;
4241 struct nfs4_ol_stateid *stp = NULL;
4242 struct nfs4_delegation *dp = NULL;
4243 struct svc_fh *current_fh = &cstate->current_fh;
4244 struct inode *ino = current_fh->fh_dentry->d_inode;
4245 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4246 struct file *file = NULL;
4252 if (grace_disallows_io(net, ino))
4253 return nfserr_grace;
4255 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
4256 return check_special_stateids(net, current_fh, stateid, flags);
4260 status = nfsd4_lookup_stateid(cstate, stateid,
4261 NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID,
4265 status = check_stateid_generation(stateid, &s->sc_stateid, nfsd4_has_session(cstate));
4268 switch (s->sc_type) {
4269 case NFS4_DELEG_STID:
4270 dp = delegstateid(s);
4271 status = nfs4_check_delegmode(dp, flags);
4275 file = dp->dl_stid.sc_file->fi_deleg_file;
4278 status = nfserr_serverfault;
4284 case NFS4_OPEN_STID:
4285 case NFS4_LOCK_STID:
4286 stp = openlockstateid(s);
4287 status = nfs4_check_fh(current_fh, stp);
4290 if (stp->st_stateowner->so_is_open_owner
4291 && !(openowner(stp->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
4293 status = nfs4_check_openmode(stp, flags);
4297 struct nfs4_file *fp = stp->st_stid.sc_file;
4299 if (flags & RD_STATE)
4300 file = find_readable_file(fp);
4302 file = find_writeable_file(fp);
4306 status = nfserr_bad_stateid;
4315 nfs4_unlock_state();
4320 nfsd4_free_lock_stateid(struct nfs4_ol_stateid *stp)
4322 struct nfs4_lockowner *lo = lockowner(stp->st_stateowner);
4324 if (check_for_locks(stp->st_stid.sc_file, lo))
4325 return nfserr_locks_held;
4326 release_lockowner_if_empty(lo);
4331 * Test if the stateid is valid
4334 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4335 struct nfsd4_test_stateid *test_stateid)
4337 struct nfsd4_test_stateid_id *stateid;
4338 struct nfs4_client *cl = cstate->session->se_client;
4341 list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
4342 stateid->ts_id_status =
4343 nfsd4_validate_stateid(cl, &stateid->ts_id_stateid);
4344 nfs4_unlock_state();
4350 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4351 struct nfsd4_free_stateid *free_stateid)
4353 stateid_t *stateid = &free_stateid->fr_stateid;
4354 struct nfs4_stid *s;
4355 struct nfs4_delegation *dp;
4356 struct nfs4_client *cl = cstate->session->se_client;
4357 __be32 ret = nfserr_bad_stateid;
4360 spin_lock(&cl->cl_lock);
4361 s = find_stateid_locked(cl, stateid);
4364 switch (s->sc_type) {
4365 case NFS4_DELEG_STID:
4366 ret = nfserr_locks_held;
4368 case NFS4_OPEN_STID:
4369 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
4372 ret = nfserr_locks_held;
4374 case NFS4_LOCK_STID:
4375 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
4378 spin_unlock(&cl->cl_lock);
4379 ret = nfsd4_free_lock_stateid(openlockstateid(s));
4381 case NFS4_REVOKED_DELEG_STID:
4382 dp = delegstateid(s);
4383 list_del_init(&dp->dl_recall_lru);
4384 spin_unlock(&cl->cl_lock);
4388 /* Default falls through and returns nfserr_bad_stateid */
4391 spin_unlock(&cl->cl_lock);
4393 nfs4_unlock_state();
4400 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
4401 RD_STATE : WR_STATE;
4404 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
4406 struct svc_fh *current_fh = &cstate->current_fh;
4407 struct nfs4_stateowner *sop = stp->st_stateowner;
4410 status = nfsd4_check_seqid(cstate, sop, seqid);
4413 if (stp->st_stid.sc_type == NFS4_CLOSED_STID
4414 || stp->st_stid.sc_type == NFS4_REVOKED_DELEG_STID)
4416 * "Closed" stateid's exist *only* to return
4417 * nfserr_replay_me from the previous step, and
4418 * revoked delegations are kept only for free_stateid.
4420 return nfserr_bad_stateid;
4421 status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
4424 return nfs4_check_fh(current_fh, stp);
4428 * Checks for sequence id mutating operations.
4431 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
4432 stateid_t *stateid, char typemask,
4433 struct nfs4_ol_stateid **stpp,
4434 struct nfsd_net *nn)
4437 struct nfs4_stid *s;
4438 struct nfs4_ol_stateid *stp = NULL;
4440 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
4441 seqid, STATEID_VAL(stateid));
4444 status = nfsd4_lookup_stateid(cstate, stateid, typemask, &s, nn);
4447 stp = openlockstateid(s);
4448 nfsd4_cstate_assign_replay(cstate, stp->st_stateowner);
4450 status = nfs4_seqid_op_checks(cstate, stateid, seqid, stp);
4454 nfs4_put_stid(&stp->st_stid);
4458 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
4459 stateid_t *stateid, struct nfs4_ol_stateid **stpp, struct nfsd_net *nn)
4462 struct nfs4_openowner *oo;
4463 struct nfs4_ol_stateid *stp;
4465 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
4466 NFS4_OPEN_STID, &stp, nn);
4469 oo = openowner(stp->st_stateowner);
4470 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
4471 nfs4_put_stid(&stp->st_stid);
4472 return nfserr_bad_stateid;
4479 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4480 struct nfsd4_open_confirm *oc)
4483 struct nfs4_openowner *oo;
4484 struct nfs4_ol_stateid *stp;
4485 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4487 dprintk("NFSD: nfsd4_open_confirm on file %pd\n",
4488 cstate->current_fh.fh_dentry);
4490 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
4496 status = nfs4_preprocess_seqid_op(cstate,
4497 oc->oc_seqid, &oc->oc_req_stateid,
4498 NFS4_OPEN_STID, &stp, nn);
4501 oo = openowner(stp->st_stateowner);
4502 status = nfserr_bad_stateid;
4503 if (oo->oo_flags & NFS4_OO_CONFIRMED)
4505 oo->oo_flags |= NFS4_OO_CONFIRMED;
4506 update_stateid(&stp->st_stid.sc_stateid);
4507 memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4508 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
4509 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
4511 nfsd4_client_record_create(oo->oo_owner.so_client);
4514 nfs4_put_stid(&stp->st_stid);
4516 nfsd4_bump_seqid(cstate, status);
4517 nfs4_unlock_state();
4521 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
4523 if (!test_access(access, stp))
4525 nfs4_file_put_access(stp->st_stid.sc_file, access);
4526 clear_access(access, stp);
4529 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
4531 switch (to_access) {
4532 case NFS4_SHARE_ACCESS_READ:
4533 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
4534 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
4536 case NFS4_SHARE_ACCESS_WRITE:
4537 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
4538 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
4540 case NFS4_SHARE_ACCESS_BOTH:
4548 nfsd4_open_downgrade(struct svc_rqst *rqstp,
4549 struct nfsd4_compound_state *cstate,
4550 struct nfsd4_open_downgrade *od)
4553 struct nfs4_ol_stateid *stp;
4554 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4556 dprintk("NFSD: nfsd4_open_downgrade on file %pd\n",
4557 cstate->current_fh.fh_dentry);
4559 /* We don't yet support WANT bits: */
4560 if (od->od_deleg_want)
4561 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
4565 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
4566 &od->od_stateid, &stp, nn);
4569 status = nfserr_inval;
4570 if (!test_access(od->od_share_access, stp)) {
4571 dprintk("NFSD: access not a subset of current bitmap: 0x%hhx, input access=%08x\n",
4572 stp->st_access_bmap, od->od_share_access);
4575 if (!test_deny(od->od_share_deny, stp)) {
4576 dprintk("NFSD: deny not a subset of current bitmap: 0x%hhx, input deny=%08x\n",
4577 stp->st_deny_bmap, od->od_share_deny);
4580 nfs4_stateid_downgrade(stp, od->od_share_access);
4582 reset_union_bmap_deny(od->od_share_deny, stp);
4584 update_stateid(&stp->st_stid.sc_stateid);
4585 memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4588 nfs4_put_stid(&stp->st_stid);
4590 nfsd4_bump_seqid(cstate, status);
4591 nfs4_unlock_state();
4595 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
4597 struct nfs4_client *clp = s->st_stid.sc_client;
4598 struct nfs4_openowner *oo = openowner(s->st_stateowner);
4600 s->st_stid.sc_type = NFS4_CLOSED_STID;
4601 unhash_open_stateid(s);
4603 if (clp->cl_minorversion) {
4604 if (list_empty(&oo->oo_owner.so_stateids))
4605 release_openowner(oo);
4606 nfs4_put_stid(&s->st_stid);
4609 * In the 4.0 case we need to keep the owners around a
4610 * little while to handle CLOSE replay. We still do need
4611 * to release any file access that is held by them
4612 * before returning however.
4614 release_all_access(s);
4615 if (s->st_stid.sc_file) {
4616 put_nfs4_file(s->st_stid.sc_file);
4617 s->st_stid.sc_file = NULL;
4619 oo->oo_last_closed_stid = s;
4620 if (list_empty(&oo->oo_owner.so_stateids))
4621 move_to_close_lru(oo, clp->net);
4626 * nfs4_unlock_state() called after encode
4629 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4630 struct nfsd4_close *close)
4633 struct nfs4_ol_stateid *stp;
4634 struct net *net = SVC_NET(rqstp);
4635 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4637 dprintk("NFSD: nfsd4_close on file %pd\n",
4638 cstate->current_fh.fh_dentry);
4641 status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
4643 NFS4_OPEN_STID|NFS4_CLOSED_STID,
4645 nfsd4_bump_seqid(cstate, status);
4648 update_stateid(&stp->st_stid.sc_stateid);
4649 memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4651 nfsd4_close_open_stateid(stp);
4653 /* put reference from nfs4_preprocess_seqid_op */
4654 nfs4_put_stid(&stp->st_stid);
4656 nfs4_unlock_state();
4661 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4662 struct nfsd4_delegreturn *dr)
4664 struct nfs4_delegation *dp;
4665 stateid_t *stateid = &dr->dr_stateid;
4666 struct nfs4_stid *s;
4668 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4670 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4674 status = nfsd4_lookup_stateid(cstate, stateid, NFS4_DELEG_STID, &s, nn);
4677 dp = delegstateid(s);
4678 status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
4682 destroy_delegation(dp);
4684 nfs4_put_stid(&dp->dl_stid);
4686 nfs4_unlock_state();
4692 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
4695 end_offset(u64 start, u64 len)
4700 return end >= start ? end: NFS4_MAX_UINT64;
4703 /* last octet in a range */
4705 last_byte_offset(u64 start, u64 len)
4711 return end > start ? end - 1: NFS4_MAX_UINT64;
4715 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
4716 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
4717 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
4718 * locking, this prevents us from being completely protocol-compliant. The
4719 * real solution to this problem is to start using unsigned file offsets in
4720 * the VFS, but this is a very deep change!
4723 nfs4_transform_lock_offset(struct file_lock *lock)
4725 if (lock->fl_start < 0)
4726 lock->fl_start = OFFSET_MAX;
4727 if (lock->fl_end < 0)
4728 lock->fl_end = OFFSET_MAX;
4731 /* Hack!: For now, we're defining this just so we can use a pointer to it
4732 * as a unique cookie to identify our (NFSv4's) posix locks. */
4733 static const struct lock_manager_operations nfsd_posix_mng_ops = {
4737 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
4739 struct nfs4_lockowner *lo;
4741 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
4742 lo = (struct nfs4_lockowner *) fl->fl_owner;
4743 deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
4744 lo->lo_owner.so_owner.len, GFP_KERNEL);
4745 if (!deny->ld_owner.data)
4746 /* We just don't care that much */
4748 deny->ld_owner.len = lo->lo_owner.so_owner.len;
4749 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
4752 deny->ld_owner.len = 0;
4753 deny->ld_owner.data = NULL;
4754 deny->ld_clientid.cl_boot = 0;
4755 deny->ld_clientid.cl_id = 0;
4757 deny->ld_start = fl->fl_start;
4758 deny->ld_length = NFS4_MAX_UINT64;
4759 if (fl->fl_end != NFS4_MAX_UINT64)
4760 deny->ld_length = fl->fl_end - fl->fl_start + 1;
4761 deny->ld_type = NFS4_READ_LT;
4762 if (fl->fl_type != F_RDLCK)
4763 deny->ld_type = NFS4_WRITE_LT;
4766 static struct nfs4_lockowner *
4767 find_lockowner_str(clientid_t *clid, struct xdr_netobj *owner,
4768 struct nfsd_net *nn)
4770 unsigned int strhashval = ownerstr_hashval(clid->cl_id, owner);
4771 struct nfs4_stateowner *so;
4773 list_for_each_entry(so, &nn->ownerstr_hashtbl[strhashval], so_strhash) {
4774 if (so->so_is_open_owner)
4776 if (!same_owner_str(so, owner, clid))
4778 atomic_inc(&so->so_count);
4779 return lockowner(so);
4784 static void nfs4_unhash_lockowner(struct nfs4_stateowner *sop)
4786 unhash_lockowner(lockowner(sop));
4789 static void nfs4_free_lockowner(struct nfs4_stateowner *sop)
4791 struct nfs4_lockowner *lo = lockowner(sop);
4793 kmem_cache_free(lockowner_slab, lo);
4796 static const struct nfs4_stateowner_operations lockowner_ops = {
4797 .so_unhash = nfs4_unhash_lockowner,
4798 .so_free = nfs4_free_lockowner,
4802 * Alloc a lock owner structure.
4803 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
4806 * strhashval = ownerstr_hashval
4808 static struct nfs4_lockowner *
4809 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp, struct nfsd4_lock *lock) {
4810 struct nfs4_lockowner *lo;
4811 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
4813 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
4816 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
4817 lo->lo_owner.so_is_open_owner = 0;
4818 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid;
4819 lo->lo_owner.so_ops = &lockowner_ops;
4820 list_add(&lo->lo_owner.so_strhash, &nn->ownerstr_hashtbl[strhashval]);
4825 init_lock_stateid(struct nfs4_ol_stateid *stp, struct nfs4_lockowner *lo,
4826 struct nfs4_file *fp, struct inode *inode,
4827 struct nfs4_ol_stateid *open_stp)
4829 struct nfs4_client *clp = lo->lo_owner.so_client;
4831 lockdep_assert_held(&clp->cl_lock);
4833 atomic_inc(&stp->st_stid.sc_count);
4834 stp->st_stid.sc_type = NFS4_LOCK_STID;
4835 stp->st_stateowner = &lo->lo_owner;
4836 atomic_inc(&lo->lo_owner.so_count);
4838 stp->st_stid.sc_file = fp;
4839 stp->st_stid.sc_free = nfs4_free_lock_stateid;
4840 stp->st_access_bmap = 0;
4841 stp->st_deny_bmap = open_stp->st_deny_bmap;
4842 stp->st_openstp = open_stp;
4843 list_add(&stp->st_locks, &open_stp->st_locks);
4844 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
4845 spin_lock(&fp->fi_lock);
4846 list_add(&stp->st_perfile, &fp->fi_stateids);
4847 spin_unlock(&fp->fi_lock);
4850 static struct nfs4_ol_stateid *
4851 find_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp)
4853 struct nfs4_ol_stateid *lst;
4854 struct nfs4_client *clp = lo->lo_owner.so_client;
4856 lockdep_assert_held(&clp->cl_lock);
4858 list_for_each_entry(lst, &lo->lo_owner.so_stateids, st_perstateowner) {
4859 if (lst->st_stid.sc_file == fp) {
4860 atomic_inc(&lst->st_stid.sc_count);
4867 static struct nfs4_ol_stateid *
4868 find_or_create_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fi,
4869 struct inode *inode, struct nfs4_ol_stateid *ost,
4872 struct nfs4_stid *ns = NULL;
4873 struct nfs4_ol_stateid *lst;
4874 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
4875 struct nfs4_client *clp = oo->oo_owner.so_client;
4877 spin_lock(&clp->cl_lock);
4878 lst = find_lock_stateid(lo, fi);
4880 spin_unlock(&clp->cl_lock);
4881 ns = nfs4_alloc_stid(clp, stateid_slab);
4885 spin_lock(&clp->cl_lock);
4886 lst = find_lock_stateid(lo, fi);
4888 lst = openlockstateid(ns);
4889 init_lock_stateid(lst, lo, fi, inode, ost);
4894 spin_unlock(&clp->cl_lock);
4901 check_lock_length(u64 offset, u64 length)
4903 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
4904 LOFF_OVERFLOW(offset, length)));
4907 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
4909 struct nfs4_file *fp = lock_stp->st_stid.sc_file;
4911 lockdep_assert_held(&fp->fi_lock);
4913 if (test_access(access, lock_stp))
4915 __nfs4_file_get_access(fp, access);
4916 set_access(access, lock_stp);
4920 lookup_or_create_lock_state(struct nfsd4_compound_state *cstate,
4921 struct nfs4_ol_stateid *ost,
4922 struct nfsd4_lock *lock,
4923 struct nfs4_ol_stateid **lst, bool *new)
4926 struct nfs4_file *fi = ost->st_stid.sc_file;
4927 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
4928 struct nfs4_client *cl = oo->oo_owner.so_client;
4929 struct inode *inode = cstate->current_fh.fh_dentry->d_inode;
4930 struct nfs4_lockowner *lo;
4931 unsigned int strhashval;
4932 struct nfsd_net *nn = net_generic(cl->net, nfsd_net_id);
4934 lo = find_lockowner_str(&cl->cl_clientid, &lock->v.new.owner, nn);
4936 strhashval = ownerstr_hashval(cl->cl_clientid.cl_id,
4937 &lock->v.new.owner);
4938 lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
4940 return nfserr_jukebox;
4941 /* FIXME: extra reference for new lockowners for the client */
4942 atomic_inc(&lo->lo_owner.so_count);
4944 /* with an existing lockowner, seqids must be the same */
4945 status = nfserr_bad_seqid;
4946 if (!cstate->minorversion &&
4947 lock->lk_new_lock_seqid != lo->lo_owner.so_seqid)
4951 *lst = find_or_create_lock_stateid(lo, fi, inode, ost, new);
4953 release_lockowner_if_empty(lo);
4954 status = nfserr_jukebox;
4959 nfs4_put_stateowner(&lo->lo_owner);
4967 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4968 struct nfsd4_lock *lock)
4970 struct nfs4_openowner *open_sop = NULL;
4971 struct nfs4_lockowner *lock_sop = NULL;
4972 struct nfs4_ol_stateid *lock_stp = NULL;
4973 struct nfs4_ol_stateid *open_stp = NULL;
4974 struct nfs4_file *fp;
4975 struct file *filp = NULL;
4976 struct file_lock *file_lock = NULL;
4977 struct file_lock *conflock = NULL;
4982 struct net *net = SVC_NET(rqstp);
4983 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4985 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
4986 (long long) lock->lk_offset,
4987 (long long) lock->lk_length);
4989 if (check_lock_length(lock->lk_offset, lock->lk_length))
4990 return nfserr_inval;
4992 if ((status = fh_verify(rqstp, &cstate->current_fh,
4993 S_IFREG, NFSD_MAY_LOCK))) {
4994 dprintk("NFSD: nfsd4_lock: permission denied!\n");
5000 if (lock->lk_is_new) {
5001 if (nfsd4_has_session(cstate))
5002 /* See rfc 5661 18.10.3: given clientid is ignored: */
5003 memcpy(&lock->v.new.clientid,
5004 &cstate->session->se_client->cl_clientid,
5005 sizeof(clientid_t));
5007 status = nfserr_stale_clientid;
5008 if (STALE_CLIENTID(&lock->lk_new_clientid, nn))
5011 /* validate and update open stateid and open seqid */
5012 status = nfs4_preprocess_confirmed_seqid_op(cstate,
5013 lock->lk_new_open_seqid,
5014 &lock->lk_new_open_stateid,
5018 open_sop = openowner(open_stp->st_stateowner);
5019 status = nfserr_bad_stateid;
5020 if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
5021 &lock->v.new.clientid))
5023 status = lookup_or_create_lock_state(cstate, open_stp, lock,
5026 status = nfs4_preprocess_seqid_op(cstate,
5027 lock->lk_old_lock_seqid,
5028 &lock->lk_old_lock_stateid,
5029 NFS4_LOCK_STID, &lock_stp, nn);
5033 lock_sop = lockowner(lock_stp->st_stateowner);
5035 lkflg = setlkflg(lock->lk_type);
5036 status = nfs4_check_openmode(lock_stp, lkflg);
5040 status = nfserr_grace;
5041 if (locks_in_grace(net) && !lock->lk_reclaim)
5043 status = nfserr_no_grace;
5044 if (!locks_in_grace(net) && lock->lk_reclaim)
5047 file_lock = locks_alloc_lock();
5049 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
5050 status = nfserr_jukebox;
5054 fp = lock_stp->st_stid.sc_file;
5055 locks_init_lock(file_lock);
5056 switch (lock->lk_type) {
5059 spin_lock(&fp->fi_lock);
5060 filp = find_readable_file_locked(fp);
5062 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
5063 spin_unlock(&fp->fi_lock);
5064 file_lock->fl_type = F_RDLCK;
5067 case NFS4_WRITEW_LT:
5068 spin_lock(&fp->fi_lock);
5069 filp = find_writeable_file_locked(fp);
5071 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
5072 spin_unlock(&fp->fi_lock);
5073 file_lock->fl_type = F_WRLCK;
5076 status = nfserr_inval;
5080 status = nfserr_openmode;
5083 file_lock->fl_owner = (fl_owner_t)lock_sop;
5084 file_lock->fl_pid = current->tgid;
5085 file_lock->fl_file = filp;
5086 file_lock->fl_flags = FL_POSIX;
5087 file_lock->fl_lmops = &nfsd_posix_mng_ops;
5088 file_lock->fl_start = lock->lk_offset;
5089 file_lock->fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
5090 nfs4_transform_lock_offset(file_lock);
5092 conflock = locks_alloc_lock();
5094 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
5095 status = nfserr_jukebox;
5099 err = vfs_lock_file(filp, F_SETLK, file_lock, conflock);
5101 case 0: /* success! */
5102 update_stateid(&lock_stp->st_stid.sc_stateid);
5103 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stid.sc_stateid,
5107 case (EAGAIN): /* conflock holds conflicting lock */
5108 status = nfserr_denied;
5109 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
5110 nfs4_set_lock_denied(conflock, &lock->lk_denied);
5113 status = nfserr_deadlock;
5116 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
5117 status = nfserrno(err);
5124 /* Bump seqid manually if the 4.0 replay owner is openowner */
5125 if (cstate->replay_owner &&
5126 cstate->replay_owner != &lock_sop->lo_owner &&
5127 seqid_mutating_err(ntohl(status)))
5128 lock_sop->lo_owner.so_seqid++;
5131 * If this is a new, never-before-used stateid, and we are
5132 * returning an error, then just go ahead and release it.
5135 release_lock_stateid(lock_stp);
5137 nfs4_put_stid(&lock_stp->st_stid);
5140 nfs4_put_stid(&open_stp->st_stid);
5141 nfsd4_bump_seqid(cstate, status);
5142 nfs4_unlock_state();
5144 locks_free_lock(file_lock);
5146 locks_free_lock(conflock);
5151 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
5152 * so we do a temporary open here just to get an open file to pass to
5153 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
5156 static __be32 nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
5159 __be32 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
5161 err = nfserrno(vfs_test_lock(file, lock));
5171 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5172 struct nfsd4_lockt *lockt)
5174 struct file_lock *file_lock = NULL;
5175 struct nfs4_lockowner *lo = NULL;
5177 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
5179 if (locks_in_grace(SVC_NET(rqstp)))
5180 return nfserr_grace;
5182 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
5183 return nfserr_inval;
5187 if (!nfsd4_has_session(cstate)) {
5188 status = lookup_clientid(&lockt->lt_clientid, cstate, nn);
5193 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
5196 file_lock = locks_alloc_lock();
5198 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
5199 status = nfserr_jukebox;
5202 locks_init_lock(file_lock);
5203 switch (lockt->lt_type) {
5206 file_lock->fl_type = F_RDLCK;
5209 case NFS4_WRITEW_LT:
5210 file_lock->fl_type = F_WRLCK;
5213 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
5214 status = nfserr_inval;
5218 lo = find_lockowner_str(&lockt->lt_clientid, &lockt->lt_owner, nn);
5220 file_lock->fl_owner = (fl_owner_t)lo;
5221 file_lock->fl_pid = current->tgid;
5222 file_lock->fl_flags = FL_POSIX;
5224 file_lock->fl_start = lockt->lt_offset;
5225 file_lock->fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
5227 nfs4_transform_lock_offset(file_lock);
5229 status = nfsd_test_lock(rqstp, &cstate->current_fh, file_lock);
5233 if (file_lock->fl_type != F_UNLCK) {
5234 status = nfserr_denied;
5235 nfs4_set_lock_denied(file_lock, &lockt->lt_denied);
5239 nfs4_put_stateowner(&lo->lo_owner);
5240 nfs4_unlock_state();
5242 locks_free_lock(file_lock);
5247 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5248 struct nfsd4_locku *locku)
5250 struct nfs4_ol_stateid *stp;
5251 struct file *filp = NULL;
5252 struct file_lock *file_lock = NULL;
5255 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
5257 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
5258 (long long) locku->lu_offset,
5259 (long long) locku->lu_length);
5261 if (check_lock_length(locku->lu_offset, locku->lu_length))
5262 return nfserr_inval;
5266 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
5267 &locku->lu_stateid, NFS4_LOCK_STID,
5271 filp = find_any_file(stp->st_stid.sc_file);
5273 status = nfserr_lock_range;
5276 file_lock = locks_alloc_lock();
5278 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
5279 status = nfserr_jukebox;
5282 locks_init_lock(file_lock);
5283 file_lock->fl_type = F_UNLCK;
5284 file_lock->fl_owner = (fl_owner_t)lockowner(stp->st_stateowner);
5285 file_lock->fl_pid = current->tgid;
5286 file_lock->fl_file = filp;
5287 file_lock->fl_flags = FL_POSIX;
5288 file_lock->fl_lmops = &nfsd_posix_mng_ops;
5289 file_lock->fl_start = locku->lu_offset;
5291 file_lock->fl_end = last_byte_offset(locku->lu_offset,
5293 nfs4_transform_lock_offset(file_lock);
5295 err = vfs_lock_file(filp, F_SETLK, file_lock, NULL);
5297 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
5300 update_stateid(&stp->st_stid.sc_stateid);
5301 memcpy(&locku->lu_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
5305 nfs4_put_stid(&stp->st_stid);
5307 nfsd4_bump_seqid(cstate, status);
5308 nfs4_unlock_state();
5310 locks_free_lock(file_lock);
5314 status = nfserrno(err);
5320 * true: locks held by lockowner
5321 * false: no locks held by lockowner
5324 check_for_locks(struct nfs4_file *fp, struct nfs4_lockowner *lowner)
5326 struct file_lock **flpp;
5328 struct file *filp = find_any_file(fp);
5329 struct inode *inode;
5332 /* Any valid lock stateid should have some sort of access */
5337 inode = file_inode(filp);
5339 spin_lock(&inode->i_lock);
5340 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
5341 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
5346 spin_unlock(&inode->i_lock);
5352 nfsd4_release_lockowner(struct svc_rqst *rqstp,
5353 struct nfsd4_compound_state *cstate,
5354 struct nfsd4_release_lockowner *rlockowner)
5356 clientid_t *clid = &rlockowner->rl_clientid;
5357 struct nfs4_stateowner *sop = NULL, *tmp;
5358 struct nfs4_lockowner *lo;
5359 struct nfs4_ol_stateid *stp;
5360 struct xdr_netobj *owner = &rlockowner->rl_owner;
5361 unsigned int hashval = ownerstr_hashval(clid->cl_id, owner);
5363 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
5365 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
5366 clid->cl_boot, clid->cl_id);
5370 status = lookup_clientid(clid, cstate, nn);
5374 status = nfserr_locks_held;
5376 /* Find the matching lock stateowner */
5377 list_for_each_entry(tmp, &nn->ownerstr_hashtbl[hashval], so_strhash) {
5378 if (tmp->so_is_open_owner)
5380 if (same_owner_str(tmp, owner, clid)) {
5386 /* No matching owner found, maybe a replay? Just declare victory... */
5392 lo = lockowner(sop);
5393 /* see if there are still any locks associated with it */
5394 list_for_each_entry(stp, &sop->so_stateids, st_perstateowner) {
5395 if (check_for_locks(stp->st_stid.sc_file, lo))
5400 release_lockowner(lo);
5402 nfs4_unlock_state();
5406 static inline struct nfs4_client_reclaim *
5409 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
5413 nfs4_has_reclaimed_state(const char *name, struct nfsd_net *nn)
5415 struct nfs4_client_reclaim *crp;
5417 crp = nfsd4_find_reclaim_client(name, nn);
5418 return (crp && crp->cr_clp);
5422 * failure => all reset bets are off, nfserr_no_grace...
5424 struct nfs4_client_reclaim *
5425 nfs4_client_to_reclaim(const char *name, struct nfsd_net *nn)
5427 unsigned int strhashval;
5428 struct nfs4_client_reclaim *crp;
5430 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
5431 crp = alloc_reclaim();
5433 strhashval = clientstr_hashval(name);
5434 INIT_LIST_HEAD(&crp->cr_strhash);
5435 list_add(&crp->cr_strhash, &nn->reclaim_str_hashtbl[strhashval]);
5436 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
5438 nn->reclaim_str_hashtbl_size++;
5444 nfs4_remove_reclaim_record(struct nfs4_client_reclaim *crp, struct nfsd_net *nn)
5446 list_del(&crp->cr_strhash);
5448 nn->reclaim_str_hashtbl_size--;
5452 nfs4_release_reclaim(struct nfsd_net *nn)
5454 struct nfs4_client_reclaim *crp = NULL;
5457 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
5458 while (!list_empty(&nn->reclaim_str_hashtbl[i])) {
5459 crp = list_entry(nn->reclaim_str_hashtbl[i].next,
5460 struct nfs4_client_reclaim, cr_strhash);
5461 nfs4_remove_reclaim_record(crp, nn);
5464 WARN_ON_ONCE(nn->reclaim_str_hashtbl_size);
5468 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
5469 struct nfs4_client_reclaim *
5470 nfsd4_find_reclaim_client(const char *recdir, struct nfsd_net *nn)
5472 unsigned int strhashval;
5473 struct nfs4_client_reclaim *crp = NULL;
5475 dprintk("NFSD: nfs4_find_reclaim_client for recdir %s\n", recdir);
5477 strhashval = clientstr_hashval(recdir);
5478 list_for_each_entry(crp, &nn->reclaim_str_hashtbl[strhashval], cr_strhash) {
5479 if (same_name(crp->cr_recdir, recdir)) {
5487 * Called from OPEN. Look for clientid in reclaim list.
5490 nfs4_check_open_reclaim(clientid_t *clid,
5491 struct nfsd4_compound_state *cstate,
5492 struct nfsd_net *nn)
5496 /* find clientid in conf_id_hashtbl */
5497 status = lookup_clientid(clid, cstate, nn);
5499 return nfserr_reclaim_bad;
5501 if (nfsd4_client_record_check(cstate->clp))
5502 return nfserr_reclaim_bad;
5507 #ifdef CONFIG_NFSD_FAULT_INJECTION
5509 u64 nfsd_forget_client(struct nfs4_client *clp, u64 max)
5511 if (mark_client_expired(clp))
5517 u64 nfsd_print_client(struct nfs4_client *clp, u64 num)
5519 char buf[INET6_ADDRSTRLEN];
5520 rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, sizeof(buf));
5521 printk(KERN_INFO "NFS Client: %s\n", buf);
5525 static void nfsd_print_count(struct nfs4_client *clp, unsigned int count,
5528 char buf[INET6_ADDRSTRLEN];
5529 rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, sizeof(buf));
5530 printk(KERN_INFO "NFS Client: %s has %u %s\n", buf, count, type);
5533 static u64 nfsd_foreach_client_lock(struct nfs4_client *clp, u64 max,
5534 void (*func)(struct nfs4_ol_stateid *))
5536 struct nfs4_openowner *oop;
5537 struct nfs4_ol_stateid *stp, *st_next;
5538 struct nfs4_ol_stateid *lst, *lst_next;
5541 list_for_each_entry(oop, &clp->cl_openowners, oo_perclient) {
5542 list_for_each_entry_safe(stp, st_next,
5543 &oop->oo_owner.so_stateids, st_perstateowner) {
5544 list_for_each_entry_safe(lst, lst_next,
5545 &stp->st_locks, st_locks) {
5557 u64 nfsd_forget_client_locks(struct nfs4_client *clp, u64 max)
5559 return nfsd_foreach_client_lock(clp, max, release_lock_stateid);
5562 u64 nfsd_print_client_locks(struct nfs4_client *clp, u64 max)
5564 u64 count = nfsd_foreach_client_lock(clp, max, NULL);
5565 nfsd_print_count(clp, count, "locked files");
5569 static u64 nfsd_foreach_client_open(struct nfs4_client *clp, u64 max, void (*func)(struct nfs4_openowner *))
5571 struct nfs4_openowner *oop, *next;
5574 list_for_each_entry_safe(oop, next, &clp->cl_openowners, oo_perclient) {
5584 u64 nfsd_forget_client_openowners(struct nfs4_client *clp, u64 max)
5586 return nfsd_foreach_client_open(clp, max, release_openowner);
5589 u64 nfsd_print_client_openowners(struct nfs4_client *clp, u64 max)
5591 u64 count = nfsd_foreach_client_open(clp, max, NULL);
5592 nfsd_print_count(clp, count, "open files");
5596 static u64 nfsd_find_all_delegations(struct nfs4_client *clp, u64 max,
5597 struct list_head *victims)
5599 struct nfs4_delegation *dp, *next;
5602 lockdep_assert_held(&state_lock);
5603 list_for_each_entry_safe(dp, next, &clp->cl_delegations, dl_perclnt) {
5606 * It's not safe to mess with delegations that have a
5607 * non-zero dl_time. They might have already been broken
5608 * and could be processed by the laundromat outside of
5609 * the state_lock. Just leave them be.
5611 if (dp->dl_time != 0)
5614 unhash_delegation_locked(dp);
5615 list_add(&dp->dl_recall_lru, victims);
5623 u64 nfsd_forget_client_delegations(struct nfs4_client *clp, u64 max)
5625 struct nfs4_delegation *dp, *next;
5629 spin_lock(&state_lock);
5630 count = nfsd_find_all_delegations(clp, max, &victims);
5631 spin_unlock(&state_lock);
5633 list_for_each_entry_safe(dp, next, &victims, dl_recall_lru) {
5634 list_del_init(&dp->dl_recall_lru);
5635 revoke_delegation(dp);
5641 u64 nfsd_recall_client_delegations(struct nfs4_client *clp, u64 max)
5643 struct nfs4_delegation *dp;
5647 spin_lock(&state_lock);
5648 count = nfsd_find_all_delegations(clp, max, &victims);
5649 while (!list_empty(&victims)) {
5650 dp = list_first_entry(&victims, struct nfs4_delegation,
5652 list_del_init(&dp->dl_recall_lru);
5654 nfsd_break_one_deleg(dp);
5656 spin_unlock(&state_lock);
5661 u64 nfsd_print_client_delegations(struct nfs4_client *clp, u64 max)
5665 spin_lock(&state_lock);
5666 count = nfsd_find_all_delegations(clp, max, NULL);
5667 spin_unlock(&state_lock);
5669 nfsd_print_count(clp, count, "delegations");
5673 u64 nfsd_for_n_state(u64 max, u64 (*func)(struct nfs4_client *, u64))
5675 struct nfs4_client *clp, *next;
5677 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns, nfsd_net_id);
5679 if (!nfsd_netns_ready(nn))
5682 list_for_each_entry_safe(clp, next, &nn->client_lru, cl_lru) {
5683 count += func(clp, max - count);
5684 if ((max != 0) && (count >= max))
5691 struct nfs4_client *nfsd_find_client(struct sockaddr_storage *addr, size_t addr_size)
5693 struct nfs4_client *clp;
5694 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns, nfsd_net_id);
5696 if (!nfsd_netns_ready(nn))
5699 list_for_each_entry(clp, &nn->client_lru, cl_lru) {
5700 if (memcmp(&clp->cl_addr, addr, addr_size) == 0)
5706 #endif /* CONFIG_NFSD_FAULT_INJECTION */
5709 * Since the lifetime of a delegation isn't limited to that of an open, a
5710 * client may quite reasonably hang on to a delegation as long as it has
5711 * the inode cached. This becomes an obvious problem the first time a
5712 * client's inode cache approaches the size of the server's total memory.
5714 * For now we avoid this problem by imposing a hard limit on the number
5715 * of delegations, which varies according to the server's memory size.
5718 set_max_delegations(void)
5721 * Allow at most 4 delegations per megabyte of RAM. Quick
5722 * estimates suggest that in the worst case (where every delegation
5723 * is for a different inode), a delegation could take about 1.5K,
5724 * giving a worst case usage of about 6% of memory.
5726 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
5729 static int nfs4_state_create_net(struct net *net)
5731 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5734 nn->conf_id_hashtbl = kmalloc(sizeof(struct list_head) *
5735 CLIENT_HASH_SIZE, GFP_KERNEL);
5736 if (!nn->conf_id_hashtbl)
5738 nn->unconf_id_hashtbl = kmalloc(sizeof(struct list_head) *
5739 CLIENT_HASH_SIZE, GFP_KERNEL);
5740 if (!nn->unconf_id_hashtbl)
5742 nn->ownerstr_hashtbl = kmalloc(sizeof(struct list_head) *
5743 OWNER_HASH_SIZE, GFP_KERNEL);
5744 if (!nn->ownerstr_hashtbl)
5746 nn->sessionid_hashtbl = kmalloc(sizeof(struct list_head) *
5747 SESSION_HASH_SIZE, GFP_KERNEL);
5748 if (!nn->sessionid_hashtbl)
5751 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
5752 INIT_LIST_HEAD(&nn->conf_id_hashtbl[i]);
5753 INIT_LIST_HEAD(&nn->unconf_id_hashtbl[i]);
5755 for (i = 0; i < OWNER_HASH_SIZE; i++)
5756 INIT_LIST_HEAD(&nn->ownerstr_hashtbl[i]);
5757 for (i = 0; i < SESSION_HASH_SIZE; i++)
5758 INIT_LIST_HEAD(&nn->sessionid_hashtbl[i]);
5759 nn->conf_name_tree = RB_ROOT;
5760 nn->unconf_name_tree = RB_ROOT;
5761 INIT_LIST_HEAD(&nn->client_lru);
5762 INIT_LIST_HEAD(&nn->close_lru);
5763 INIT_LIST_HEAD(&nn->del_recall_lru);
5764 spin_lock_init(&nn->client_lock);
5766 INIT_DELAYED_WORK(&nn->laundromat_work, laundromat_main);
5772 kfree(nn->ownerstr_hashtbl);
5774 kfree(nn->unconf_id_hashtbl);
5776 kfree(nn->conf_id_hashtbl);
5782 nfs4_state_destroy_net(struct net *net)
5785 struct nfs4_client *clp = NULL;
5786 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5788 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
5789 while (!list_empty(&nn->conf_id_hashtbl[i])) {
5790 clp = list_entry(nn->conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
5791 destroy_client(clp);
5795 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
5796 while (!list_empty(&nn->unconf_id_hashtbl[i])) {
5797 clp = list_entry(nn->unconf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
5798 destroy_client(clp);
5802 kfree(nn->sessionid_hashtbl);
5803 kfree(nn->ownerstr_hashtbl);
5804 kfree(nn->unconf_id_hashtbl);
5805 kfree(nn->conf_id_hashtbl);
5810 nfs4_state_start_net(struct net *net)
5812 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5815 ret = nfs4_state_create_net(net);
5818 nfsd4_client_tracking_init(net);
5819 nn->boot_time = get_seconds();
5820 locks_start_grace(net, &nn->nfsd4_manager);
5821 nn->grace_ended = false;
5822 printk(KERN_INFO "NFSD: starting %ld-second grace period (net %p)\n",
5823 nn->nfsd4_grace, net);
5824 queue_delayed_work(laundry_wq, &nn->laundromat_work, nn->nfsd4_grace * HZ);
5828 /* initialization to perform when the nfsd service is started: */
5831 nfs4_state_start(void)
5835 ret = set_callback_cred();
5838 laundry_wq = create_singlethread_workqueue("nfsd4");
5839 if (laundry_wq == NULL) {
5843 ret = nfsd4_create_callback_queue();
5845 goto out_free_laundry;
5847 set_max_delegations();
5852 destroy_workqueue(laundry_wq);
5858 nfs4_state_shutdown_net(struct net *net)
5860 struct nfs4_delegation *dp = NULL;
5861 struct list_head *pos, *next, reaplist;
5862 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5864 cancel_delayed_work_sync(&nn->laundromat_work);
5865 locks_end_grace(&nn->nfsd4_manager);
5868 INIT_LIST_HEAD(&reaplist);
5869 spin_lock(&state_lock);
5870 list_for_each_safe(pos, next, &nn->del_recall_lru) {
5871 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
5872 unhash_delegation_locked(dp);
5873 list_add(&dp->dl_recall_lru, &reaplist);
5875 spin_unlock(&state_lock);
5876 list_for_each_safe(pos, next, &reaplist) {
5877 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
5878 list_del_init(&dp->dl_recall_lru);
5879 nfs4_put_stid(&dp->dl_stid);
5882 nfsd4_client_tracking_exit(net);
5883 nfs4_state_destroy_net(net);
5884 nfs4_unlock_state();
5888 nfs4_state_shutdown(void)
5890 destroy_workqueue(laundry_wq);
5891 nfsd4_destroy_callback_queue();
5895 get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
5897 if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG) && CURRENT_STATEID(stateid))
5898 memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
5902 put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
5904 if (cstate->minorversion) {
5905 memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
5906 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
5911 clear_current_stateid(struct nfsd4_compound_state *cstate)
5913 CLEAR_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
5917 * functions to set current state id
5920 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
5922 put_stateid(cstate, &odp->od_stateid);
5926 nfsd4_set_openstateid(struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
5928 put_stateid(cstate, &open->op_stateid);
5932 nfsd4_set_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
5934 put_stateid(cstate, &close->cl_stateid);
5938 nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate, struct nfsd4_lock *lock)
5940 put_stateid(cstate, &lock->lk_resp_stateid);
5944 * functions to consume current state id
5948 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
5950 get_stateid(cstate, &odp->od_stateid);
5954 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate, struct nfsd4_delegreturn *drp)
5956 get_stateid(cstate, &drp->dr_stateid);
5960 nfsd4_get_freestateid(struct nfsd4_compound_state *cstate, struct nfsd4_free_stateid *fsp)
5962 get_stateid(cstate, &fsp->fr_stateid);
5966 nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate, struct nfsd4_setattr *setattr)
5968 get_stateid(cstate, &setattr->sa_stateid);
5972 nfsd4_get_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
5974 get_stateid(cstate, &close->cl_stateid);
5978 nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate, struct nfsd4_locku *locku)
5980 get_stateid(cstate, &locku->lu_stateid);
5984 nfsd4_get_readstateid(struct nfsd4_compound_state *cstate, struct nfsd4_read *read)
5986 get_stateid(cstate, &read->rd_stateid);
5990 nfsd4_get_writestateid(struct nfsd4_compound_state *cstate, struct nfsd4_write *write)
5992 get_stateid(cstate, &write->wr_stateid);